Download K26-05222-ncomms6307 ( .pdf )

PAGE 1

ARTICLEReceived31Jul2014 | Accepted17Sep2014 | Published20Oct2014Thecave“shgenomerevealscandidategenes foreyelossSuzanneE.McGaugh1, w,JoshuaB.Gross2,BronwenAken3,4,MarylineBlin5,RichardBorowsky6, DomitilleChalopin7,He ´ le ` neHinaux5,WilliamR.Jeffery8,AlexKeene9,LiMa8,PatrickMinx1,DanielMurphy3,4, KellyE.OQuin10,SylvieRe ´ taux5,NicolasRohner11,SteveM.J.Searle3,BethanyA.Stahl2,CliffTabin11, Jean-NicolasVolff7,MasatoYoshizawa9&WesleyC.Warren1Naturalpopulationssubjectedtostrongenvironmentalselectionpressuresofferawindow intothegeneticunderpinningsofevolutionarychange.Cave“shpopulations, Astyanax mexicanus (Teleostei:Characiphysi),exhibitrepeated,independentevolutionforavarietyof traitsincludingeyedegeneration,pigmentloss,increasedsizeandnumberoftastebudsand mechanosensoryorgans,andshiftsinmanybehaviouraltraits.Surfaceandcaveformsare interfertilemakingthissystemamenabletogeneticinterrogation;however,lackofareference genomehashamperedeffortstoidentifygenesresponsibleforchangesincaveformsof A.mexicanus. Herewepresentthe“rst denovo genomeassemblyfor Astyanaxmexicanus cave“sh,contrastrepeatelementstootherteleostgenomes,identifycandidategenes underlyingquantitativetraitloci(QTL),andassaythesecandidategenesforpotential functionalandexpressiondifferences.Weexpectthecave“shgenometoadvanceunderstandingoftheevolutionaryprocess,aswellas,analogoushumandiseaseincludingretinal dysfunction. DOI:10.1038/ncomms6307 OPEN 1TheGenomeInstitute,WashingtonUniversity,CampusBox8501,StLouis,Missouri63108,USA.2DepartmentofBiologicalSciences,Universityof Cincinnati,711BRieveschlHall,312CollegeDrive,Cincinnati,Ohio45221,USA.3WellcomeTrustSangerInstitute,WellcomeTrustGenomeCampus,Hinxton, CambridgeCB101SA,UK.4EuropeanMolecularBiologyLaboratory,EuropeanBioinformaticsInstitute,WellcomeTrustGenomeCampus,Hinxton, CambridgeCB101SD,UK.5DECAgroup,NeurobiologyandDevelopmentLaboratory,CNRS-InstitutdeNeurobiologieAlfredFessard,91198Gif-sur-Yvette, France.6DepartmentofBiology,NewYorkUniversity,NewYork,NewYork10003-6688,USA.7InstitutdeGe ´ nomiqueFonctionnelledeLyon,EcoleNormale Supe ´ rieuredeLyon,CNRS,UMR5242,UCBL,46alle ´ edItalie,LyonF-69364,France.8DepartmentofBiology,UniversityofMaryland,CollegePark, Maryland20742,USA.9DepartmentofBiology,UniversityofNevada,Reno,Nevada89557,USA.10DepartmentofBiology,CentreCollege,600West WalnutSt,Danville,Kentucky40422,USA.11HarvardMedicalSchoolDepartmentofGenetics,77AvenueLouisPasteur;NRB360,Boston,Massachusetts 02115,USA. w Presentaddress:Ecology,Evolution,andBehavior,UniversityofMinnesota,100EcologyBuilding,1987UpperBufordCir,FalconHeights, Minnesota55108,USA.CorrespondenceandrequestsformaterialsshouldbeaddressedtoS.E.M.(email:smcgaugh@umn.edu). NATURECOMMUNICATIONS |5:5307|DOI:10.1038/ncomms6307|www.nature.com/naturecommunications1& 2014 MacmillanPublishersLimited.Allrightsreserved.

PAGE 2

Someofthemostfundamentalquestionsinevolutionarybiologyinvolvehoworganismscanadapttonewenviron-ments.Naturalpopulationsunderstrongselectionmaybeespeciallyusefulindecipheringthegeneticvariantsunderpinningtheseevolutionaryresponses.Yet,fewsystemspossessdramaticphenotypicchangesthatcanbede“nitivelyattributedtoselectionpressuresofanewenvironment.Evenfewerspeciescanbeusedtounderstandhowevolutionproceedswhenrepeatedinseparatepopulations.Caveanimalsofferoneofthemostexcitingsystemsinwhichtostudythesequestions1.Speci“cally,surfaceformsoftheMexicantetra,Astyanaxmexicanus,colonizedmultiplecavesinnortheasternMexicoandevolvedextremecave-associatedtraitsatleastfourindependenttimesoverthepast2…3Myr(refs2,3).Cave“shpopulationsexhibitrepeatedmorphologicalevolutionforavarietyoftraitsincludingeyedegeneration2,4,pigmentloss5,6,increasedsizeandnumberofspecializedmechanosensoryorganscalledneuromasts7andincreasednumbersoftastebuds4.Cave“shhavealsoevolvedbehaviouraldifferencesrelativetotheirsurface-dwellingcounterpartsincludingincreasedattractiontovibrations7,increasedolfactorycapabilities8,alteredfeedingangles9andlossofschoolingandaggression10,11.Further,cave“shlosebodyweightlessquicklythansurfacemorphs8andshowdramaticsleepreductionscomparedtosurface“sh12.Thepolarityofthesetraitchangesisknown(derivedincave“sh).Therefore,thesenaturalreplicatesofferauniqueopportunitytostudythegeneticbasesofparallelandconvergentevolutionarychanges1.Further,A.mexicanusisamenabletomoleculargeneticmanipulationinthelab13,14,andpriorQTL(quantitativetraitlocus)analysesofsurfaceandcave“shcrossesidenti“edgenomicregionsregulatingnumerousbehaviouralandmorphologicaltraits4,7…10,15.Inthiswork,wepresentthe“rstdenovogenomeassemblyforA.mexicanuscave“shtoallowformorepreciseidenti“cationofcandidategenesunderlyingQTLthanwaspreviouslypossiblewithsynteniccomparisonstozebra“sh15,16.Wedemonstratetheeffectivenessofthisapproachbyidentifyingcandidategenesforeyedevelopmentandothercave-derivedtraits.WefurtheranalyzeRNAseqdatatosurveythesecandidategenesforpotentialcodingandexpressionleveldifferencesbetweenthesurfaceandcavepopulations.Formanytraits,weexpectthatthecave“shgenomewillprovideatoolfordiscoveryoftheroleofindividualgenesandpathways.ResultsSequencingandannotation.Thesequencedcave“shindividualwasthe“rst-generationoffspringoftwowild-caughtparents,whichoriginatedfromPacho´ncave,Tamaulipas,Mexico.Whilethereareatleast29cavesthatcontainAstyanaxcave“sh,thePacho´ncave“sharethemoststudiedandexhibitthemostextremetroglomorphicphenotypesrelativetotheothercaves2.Thisgenomedraftwasassembledtoasizeof964Mb,whichissimilarinsizetoacongenericinBrazil17.Thedraftgenomecontains10,735scaffolds(N50contig¼14.7kb;N50scaffold¼1.775Mb),andthelongestscaffoldsizewas9.823Mb(SupplementaryData1).UsingtheEnsemblannotationpipeline18andRNAseqtranscriptevidence(eightuniquetissues;SupplementaryData2),wepredictedatotalof23,042protein-codinggenes,similartoothersequencedteleost“shes.Zebra“shistheclosestsequencedrelativetocave“sh(divergedapproximately250Myr)19,andweannotated16,480one-to-oneorthologswithzebra“sh.Toestimategenerepresentationinthedraftgenome,weusedassembledcave“shtranscriptsandevolutionarilyconservedgenemodels.AlignmentoftheAstyanaxbestopenreadingframes(SupplementaryData2)tothegenomescaffoldsfoundthatacrosstissue-speci“ctranscriptomes,amedianof81%oftranscriptsalignoveratleast75%oftheirlengthwithatleast90%identity.Further,CEGMAanalyses20indicatedthat95%ofthe248ultra-conservedcoreeukaryoticgenesarepresentinthegenomeassembly,and69%ofthe248ultra-conservedcoreeukaryoticgeneswereconsideredcompletegenes.Collectively,thissuggeststhattheassemblyhascapturedmuchoftheprotein-codingsequenceinthecave“shgenome.Transposableelementannotation.One-thirdofthecave“shgenomeiscomposedoftransposableelements(TEs)(SupplementaryData3and4).Thisrepetitivecontentiscom-parabletomostpublished“shgenomes(SupplementaryData3),withtheexceptionsofzebra“sh(52.2%TEs)21andFugu(2.8%)(ref.22).Inthecave“sh,DNAtransposonsaremoreabundantanddiversi“edthanretrotransposons,asthereareatleast12differentsuperfamiliesofDNAtransposonsrepresenting12.7%ofthegenome.Incontrast,retrotransposonscompriseonly2.3%ofthegenome(SupplementaryData4).Itappearsthatarecentwaveoftranspositionoccurredinthecave“shgenome(Fig.1)andwascomposedmostlyofTc-MarinerandhATsuperfamilies,whichcurrentlycompriseapproximately9.5%ofthecave“shgenome.Similarly,zebra“shexperiencedarecentlargeexpansionofrepeatfamilies,includingTc-MarinerandhATsuperfamilies,whereasanothercommonmodel,stickleback,hasnot(RepeatMaskerGenomicDatasets).WeestimatedthepotentialageofthedifferentofcopiesforeachTE-relatedsuperfamiliesbycalculatingKimuradistancesassum-ingthatmostofthemutationsinTEcopiesareneutral.Theratesoftransversions(q)andtransitions(p)weretransformedinKimuradistancesusing[K¼½ln(12pq) 14ln(12q)].Thecave“shgenomediffersincomparisonwithzebra“shinthatitappearstolackveryyoungelements(asindicatedbytheKimuradistancefromtheconsensus,Fig.1,RepeatMaskerGenomicDatasets).Giventhecaveatsofpossibleassemblyartefacts,lackofveryyoungelementsindicatesthatitisunlikelythatmanycopiesofTc-MarinerandhATsuperfamiliesarestill 1.00.80.60.40.20.010Kimura distance (from 0 to 50)SINE/tRNA–LysSINE/otherLINE/Rex–BabarLINE/L2RC/HelitronDNA/TcMarinerDNA/hATDNA/KolobokDNA/HarbingerDNA/CMCDNA/othersPercentage of TEs in the Cavefish genome234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950 Figure1|TEsuperfamilieshistoryinthecave“shgenome.Onlysuperfamiliesthatshowcontenthigherthan0.1%inSupplementaryData4wereused.Kimuradistancesarerangedfromvalue0representingrecentTEcopiesto50fortheoldTEinsertions. ARTICLENATURECOMMUNICATIONS|DOI:10.1038/ncomms63072NATURECOMMUNICATIONS|5:5307|DOI:10.1038/ncomms6307|www.nature.com/naturecommunications&2014MacmillanPublishersLimited.Allrightsreserved.

PAGE 3

activeinthecave“shgenome(con“rmedbyanalysesoftranscriptomes,SupplementaryData5…9).Identi“cationofcandidategenesunderQTL.Perhapsthemostdistincttraitincave“shisthereduced,nearlyabsenteye(Fig.2),whichisindependentlyderivedinmultiple,independentcaveinvasions2,3,23.Incave“sh,earlyeyedevelopmentislargelysimilartotheeyedevelopmentinsurface“shinthatlensvesiclesandopticcupsform,albeit,theyaresmallerincave“shevenatveryearlystages(14h.p.f.,hourspostfertilization24).Thelensapoptosisbeginsafter25h.p.f.(refs24,25),andtheretinaundergoessigni“cantapoptosisatabout35h.p.f.(ref.24).Thisapoptosiscontinuesfordaystoweeks,andleadstoanarrestofeyedevelopment25,26.WeexaminethegenomeforgenesunderQTLforeyesizefromPacho´ncave“shsurface“shcrossesfromvariousstudies4,6…8,15,16.Acrossstudies,wecountatotalof15non-overlappingQTLforeye-relatedphenotypesdiscoveredinthePacho´npopulation4,7…10,15,16(SupplementaryFig.1).ScaffoldsoftendidnotspantheentirecriticalregioncomprisingaQTL;thus,eachQTLcriticalregionmaybedistributedacrossseveralscaffolds.AllgenesonascaffoldcontainingamarkerlinkedtotheQTLwereincluded.Intotal,2,408genesoutofthe23,042genesannotatedinthisdraftofthegenomewereassociatedwiththesegenomicregions.Itislikelythatasigni“cantportionofthesegenesthatarephysicallylinkedtothecausalvariantarenotresponsibleforthephenotype.Tonarrowthelistofcandidategenes,weexaminedthegeneexpressioninsurfaceandcavepopulationswithadevelopmentaltimecoursetakenat10h.p.f.,24h.p.f.,1.5dayspostfertilization(d.p.f.),and3d.p.f.RNAfromeachtimeperiodwasextractedfrom50whole,pooledindividualsandIlluminareadsweregeneratedforcave“shandsurface“shpoolsseparately.Animportantcaveattointerpretingthegeneexpressiondataisthatevenearlyindevelopment,cave“sheyesaresmallerthansurface“sheyes,andlowernumbersoftranscriptsmayre”ectsmallereyesandnotnecessarilydownregulation.Thetranscriptsequenceswerealsousedforobtainingcodingvariantdifferencesbetweensurface“shandcave“sh.Duetotheenormityofde“ninggenetoQTLassociationsformanytroglomorphictraits,weprimarilyfocusedontheeyephenotype.Hereweusedexpressiondataandintegratedpathwayanalysis27topredictlikelyphenotypesandthegenespotentiallyunderlyingthosephenotypes.Utilizingpriorknowledgeofpredictedoutcomebetweentranscriptionalregulatorsandtheirtargetgenes27,weimplicate30genesundertheQTLtoresultincongenitaleyeanomalies.Thedirectionofgeneexpressionchangebetweensurface“shandcave“shsupportsanincreasedlikelihoodofeyeanomaliesincave“shrelativetosurface“shat10h.p.f.,24h.p.f.,and1.5d.p.f.(forexample,12/27,12/19,11/30geneshaveexpressiondirectionconsistentwithincreasedcongenitalanomalyoftheeye,respectively;biased-correctedzscoreZ2.266,Po0.0001inallcases,seeref.27fordetailsofzscorecalculation).Atthelastsampledtimepoint(3d.p.f.),theexpressiondataareconsistentwithincreasedcave“sheyeanomalies,butinterestingly,thezscoresbecomesmallerwiththeprogressionofdevelopment(SupplementaryData10…16)andarenotsigni“cantat3d.p.f.WeperformedanenrichmenttestwithdatacombinedacrosstimepointsandfoundthattheQTLwereenrichedforgenesinvolvedincongenitalanomalyoftheeye,(30/1,560relativeto159/12,040inthetotalexpressiondataset;w2-testwithYatescorrectionPvalueo0.034,w2¼4.48,oddsratio¼1.57,95%con“denceintervalofoddsratio¼1.05…2.35).Additionalgenesinvolvedineyedevelopment,functionanddiseasewereenrichedintheQTLset,thoughnotsigni“cantlyso(129/1,560relativeto921/12,040intotaldataset;w2-testwithYatescorrectionPvalue¼0.35,w2¼0.88,oddsratio¼1.10,95%con“denceintervalofoddsratio¼0.91…1.34).Therefore,wecontendthattheeye-relatedQTLarequalitativelyenrichedforeye-relatedgenesrelativetotherestofthegenome,buttheeye-relatedQTLarequantitativelymorelikelytocontaingenesassociatedwithcongenitaleyedefects.Speci“ccandidategenesundereye-relatedQTL.SeveralgenesfoundundertheQTLareclassiccandidatesforeyedevelopment,andwehighlightseveral,whichmaybeparticularlypromising.WenarroweddownthelistofcandidategenesundertheQTLbyfocusingonthosewithexpressiondifferencesbetweencave“shandsurface“sh.StatisticalcomparisonsofgeneexpressionlevelswereperformedusingthemeasureoflogfoldchangeperformedinCuffdiff2.1.0(ref.28)(seeCuffdiff2.1.0documentationforadditionaldetailsoftest).Unlessotherwisenoted,allPvaluesgivenbelowfordifferentialexpressionbetweencaveandsurface“shweregeneratedbythistest.Linkagegroup(LG)namesareinconsistentacrossstudies;thus,theLGsgivenbelowcorrespondtothenamingschemeintheoriginalstudyinwhichtheQTLwasfoundandthosestudiesarecitedaftertheLGname.Oneofthesecandidategenesidenti“edbythismethodiscryaa,anantiapoptoticchaperoneproteinwhoseabsenceofgeneexpressionwashypothesizedtoplayaroleincave“sheyedegeneration26.CryaafallsunderaQTLforeyesizeonLG27(scaffoldcontainingmarkerAm229b)fromProtasetal.4Next,pitx3isessentialforlensdevelopmentinzebra“sh29,30andknockdownexperimentsresultinzebra“shwithdegeneratelensandretinasandmisshapenlowerjaws29.Cave“shexhibitsigni“cantlylowerexpressionofpitx3at24h.p.f.and3d.p.f.(Po0.002atbothtimepoints,qualitativelyloweratalltimes),butthereareonlytwosynonymousdifferencesbetweensurfaceandcave“shpitx3.Pitx3islocatedundertheQTLforlenslengthonLG14(ref.4)andforeyesizeonLG4(ref.7).Similarly,rx3islocatedunderaQTLforeyesizeonLG4(refs4,8)andunderliesalossofeyesinzebra“sh(chokh)andmedaka(eyeless)mutants31,32.Rx3exhibitssigni“cantlylessexpressionincave“shthaninsurface“shat10h.p.f.and3d.p.f.(Po0.0003atbothtimepoints,qualitativelyloweratalltimes)andnocodingvariants.Likewise,undertheQTLforeyesizeonLG4(refs4,8)arethegenesolfm2aandolfml2a.Zebra“shknockdownsofolfm2resultinabnormalitiesintheolfactorypits,eyesandoptictectumaswellasreducedandless-de“nedPax6expressionintheeye33,andolfm2aexhibitedsigni“cantlylowerexpressionincave“shat3d.p.f.(Po0.001).Wedidnotdetectcodingdifferencesinolfm2a,anddatawereunavailableforolfml2a.Lastly,BCoRisfoundonLG19(refs4,8).BCoRislinkedwithocularcolobomas Figure2|Photographsofsurfaceandcave“sh.(a,b)Surface“sh(line152)(c,d)Pacho´ncave“sh(line45).Scalebarfora,cis1cm.Scalebarforb,dis0.25cm.PhotosbyB.A.S. NATURECOMMUNICATIONS|DOI:10.1038/ncomms6307ARTICLENATURECOMMUNICATIONS|5:5307|DOI:10.1038/ncomms6307|www.nature.com/naturecommunications3&2014MacmillanPublishersLimited.Allrightsreserved.

PAGE 4

inhumanandzebra“sh34anditsbindingpartner,BCL6,hasbeen showntocontrolopticcupmorphogenesisthroughregulationof p53 inzebra“sh35.Cave“shexhibitsigni“cantlylowerexpression of BCoR at10h.p.f.( P o 0.013),andfournonsynonymouscoding differencesexistbetweensurfaceandcave“sh,thoughallappear tobeinevolutionarylabilesites.Importantly,thesegenes representonlyasubsetoftheinterestingcandidatesunderQTL. CandidategenesinQTLwithpotentiallypleiotropiceffects . ForseveralQTL,multipletroglomorphicphenotypesco-localize witheyesize,andthisco-localizationhasbeensuggestedasan evidencethatselectionforsomecave-adaptedtraitsresultedin pleiotropicdegenerationofeyes7.OneoftheseQTLisinvolvedin vibrationattractionbehaviour,eyesizeandsuper“cialneuromast numberattheorbitonLG2(ref.7).ThissameQTLforeyesize hasbeenidenti“edinmultiplestudies(LG7(ref.4),LG8(refs4,8) andLG1(ref.16))andaQTLforthethicknessoftheinner nuclearlayeroftheretinaonLG2(ref.15).TheseLGsfrom variousstudiesallcorrespondtothesamegenomicregion,and herewecountthisregionasasingleQTL.Wemainlyconcentrate ongenesthatareexpressedinbothcaveandsurface“shand appeartohavenotbeenpseudogenizedincave“sh,astheseare genesmostlikelytohavepleiotropiceffectsandtobegood candidatesfordrivingmultiplephenotypesthatco-localizetothe sameQTL. Oneofthemoreinterestingcandidategenesinthisregionis shisa2 ,whichinhibitsWntand“broblastgrowthfactorsignalling byretainingtheirrespectivereceptors,Frizzledand“broblast growthfactorreceptor,intheendoplasmicreticulum36.Cave“sh expressionof shisa2 isqualitativelyhigherthansurface“shatall timepoints(signi“cantlysoat10h.p.f.,24h.p.f.and1.5d.p.f., P o 0.005),but shisa2 containsonlyasinglesynonymouschange betweencaveandsurface“sh.Aduplicatecopyof shisa2 isalso underaneyeQTLfoundonLG6(refs4,8),andthisparalog exhibitsnocodingdifferencesandelevated,butmostlynonsigni“cant,expressionincave“sh(24h.p.f., P o 0.0003,not signi“cantat10h.p.f.and1.5d.p.f.)andlowerexpressionin cave“shat3d.p.f.( P o 0.0002). Because shisa2 interactswithmajordriversofdevelopment,we furtherassessedquantitativeandspatialdifferencesofexpression forthetwo shisa2 genes(LG2andLG6)byquantitativePCR (qPCR)and insitu hybridizationon Astyanax embryos.Forboth genes,qPCRexperimentsdidnotdetectsigni“cantdifferencesat 36h.p.f.betweenthetwomorphs,suggestinganexpression differenceoflessthantwofold(Fig.3a;Mann…Whitney U -test, P 4 0.05).Atthisstage, shisa2 -LG2wasexpressedthroughoutthe epidermisaswellasintheolfactoryepitheliumandthelensin surface“sh,butlensexpressionwasnotablymissingincave“sh (Fig.3b). Shisa2 -LG6hadamorecomplexexpressionpattern, reminiscentofwhatwasdescribedin Xenopus37andzebra“sh38, andincludedexpressioninthebranchialarches,cranialganglia, epidermis,olfactoryepithelium(like shisa2 -LG2),retinaandlens (Fig.3c).Noobviousdifferencewasobservedbetweensurface“sh andcave“shembryosconcerning shisa2 -LG6expressionpattern. Insum,anatomicalanalysisdetectedalackofs hisa2 (LG2) expressioninthecave“shlens,whichsuggestschangesinthe regulatoryregionofthisgenemaycontributetothelossofeyesin cave“sh. In Xenopus embryos,s hisa2 morpholinoknockdownormRNA injectionelicittheexpressionchangesfor otx2, akeyhomeobox geneforheadandeyedevelopment36.We,therefore,also compared otx2 expressionpatternsandlevelsin Astyanax cave“shandsurface“shembryos,thoughwecannotlocalize thisgeneontoaspeci“cLG.While otx2 patternissimilarinthe twomorphsduringheadandbraindevelopment(Fig.4b),lens expressionismuchweakerincave“shat48h.p.f.(Fig.4c),aswell aswhenassessedbywhole-organismsemi-quantitativereverse transcriptase-PCR(Fig.4a,SupplementaryFig.2).Wehave, therefore,identi“edapotentialdevelopmentalregulatorycascade thatmayleadtothecave“sheyelossandthatwouldinvolve shisa2 and otx2 inthedevelopinglens. Inadditionto shisa2 ,weidenti“edcandidategenesunderthis potentiallypleiotropicQTL.Severalgenesmeetingourcriteria underthisparticularQTLinclude prox1 and AIFM1 .Two additionalgenesfoundintheQTLanalysisofOQuin etal.15, crxa and Tbx2a ,arealsopresentunderthisQTLinouranalysis. Prox1 regulatesmanyprocessesindevelopmentincludinglens “breelongationanddifferentiationandtheexitofretinal progenitorcellsfromthecellcyclereviewedinref.39.The knockdownof prox1 resultsinthedisruptionofthelens-speci“c g -crystallinexpressionandsubsequentlensapoptosis40.Cave“sh expressionof prox1 exhibitsasimilarspatialpatterntosurface “shinthedevelopinglens,andforthisreason prox1 was previouslyconsideredunlikelytoplayaroleinthecave-speci“c eyedegeneration41. Prox1 isexpressedinsensoryhaircellsofthe neuromastandtastereceptorcellsoftastebuds,bothofwhichare morenumerousincave“shrelativetosurface,but prox1 expressioninthesestructuresdoesnotoccuruntil96h.p.f. (ref.41).Wedetectednosequencedifferencesbetweencaveand surface“shfor prox1 .However,inourwhole-organismRNAseq data,signi“cantlylowerexpressionincave“shwasobservedat 24h.p.f.,1.5d.p.f.and3d.p.f.( P o 0.022inallcases),while marginallynon-signi“canthigherexpressionincave“shwas observedattheearliestsampledtimepoint(10h.p.f., P ¼ 0.083). Signi“cantlylowerexpressionof prox1 duringthese developmentaltimepointsisconsistentwithincreasedlens apoptosisincave“sh.Therefore,are-examinationofthe contributionof prox1 ,inlightofitslocationunderthisQTL forsuborbitalneuromastcellnumber,VABandeyesize7andits quantitativeexpressiondifferences,maybewarranted. AIFM1 isimplicatedinsigni“cantandprogressiveoptic atrophyinmutantHarlequinmice,andthismutantphenotype canberescuedbyinjectionofanexpressionvectorcontaining AIFM1 (ref.42).Cave“shexhibitsigni“cantlylowerexpression of AIFM1 at24h.p.f.( P o 0.003)thansurface“sh,andthis geneexhibitsanintronicspliceregionvariantand“ve nonsynonymousvariants,twoofwhichappearderivedin cave“sh.Thesevariantswereallpredictedtobetoleratedbya computationalmethodthatattemptstodetermineifanamino acidsubstitutionisdetrimentaltoproteinfunction(SIFT43). Interestingly,theparalogofthisgene, AIFM2 ,isalsolocated undertheQTLforeyesizefoundonLG14(ref.4)andLG4 (ref.7). AIFM2 hassigni“cantlyreducedexpressionincave“sh relativetosurface“shatmosttimepoints(10h.p.f.,24h.p.f., 3d.p.f.; P o 0.022inallcases)withqualitativelylowerexpression at1.5d.p.f.( P ¼ 0.095).Further,thetwospliceregionvariantsare “xedbetweenthesurfaceandcave“sh,oneofwhichalsoresults inanonsynonymouschangethatisputativelyderivedincave“sh, thoughthischangeispredictedbySIFTtobetolerated. Crxa inducesretinalstemcellstodifferentiateintofunctional photoreceptors44.Whenknockeddowninzebra“sh, crxa promptsthedownregulationofgenesinthephototransduction cascade45,andisimplicatedineyereductionexperiencedby anothertroglomorphic“sh, Sinocyclocheilusanophtalmus46.This geneexhibitssigni“cantlyreducedexpressionincave“shat1.5 and3d.p.f.( P o 0.001;attheothertimepoints,expressioncould notbetested). Crxa containednosequencedifferencesbetween caveandsurface“sh. Tbx2a exhibitslocalizedexpressioninzebra“shmainlyinthe oticplacode,opticvesicle,oticvesicleandretina(alsoinventral mesodermandpectoral“nbud)38.T bx2 resultsinsmalleroptic ARTICLENATURECOMMUNICATIONS|DOI:10.1038/ncomms63074NATURECOMMUNICATIONS |5:5307|DOI:10.1038/ncomms6307|www.nature.com/naturecommunications& 2014 MacmillanPublishersLimited.Allrightsreserved.

PAGE 5

cupswhenmutatedinmice47,andtwocopiesexistinzebra“sh.Tbx2aisinvolvedincraniofacialandpharyngealarchdevelopment48,anditsparalogTbx2bisrequiredforproperretinalneuronalformationinzebra“sh49.Tbx2aexhibitslowerexpressionincave“shatalltimepoints(Po0.07at1.5d.p.f.,Po0.001atallothertimepoints)andthreenonsynonymousdifferencesbetweencaveandsurface.Onlyonenonsynonymousdifferenceisputativelyderivedincave“sh(D401E),andsuchanaminoacidreplacementispredictedbySIFTtobetolerated.Underthesecondco-localizingQTLforthetraitsvibrationattractionbehaviour,super“cialneuromastnumberatorbitandeyesizelocatedonLG17inYoshizawaetal.7,welackedscaffoldcoverageforseveralmarkersinthecenteroftheQTL(208e,205dand221a;SupplementaryData17).Thereareseveralinterestinggenesinthisregion(SupplementaryData11),butfewareascompellingasgenesfoundontheco-localizingQTLonLG2ofYoshizawaetal.7Weexpectfuturedraftsofthegenometouncoveradditionalcandidategenesinthisregion.Candidategenesforadditionalcavephenotypes.Lastly,wesoughttobrie”yinvestigateotherdistinctivetraitsforcave“sh,includingreducedpigmentation5,6.First,wefoundthatoneofthemostfamouspigmentationgenes,mc1r,knowntobemutatedinPacho´ncave“sh5(thepopulationfromwhichtheQTLweremapped),islocatedunderthecriticalregionoftheQTLfor 2abchijlmkdfge10Shisa2LG2(QTL)Shisa2LG6(paralog)SF, 36 h.p.f.EpidEpidEpidEpidEpidOlf epitOlf epitOlf epitOlf epitEpidOlf epitLensLensLensLensLensRetinaRetinaLensRetinaLensRetinaCF, 36 h.p.f.qPCR on Shisa2 genes at 36 h.p.f.Relative mRNA levelsShisa2 (LG2)Shisa2 (LG6)SFCF Figure3|Expressionpatternsofshisa2.(a)QuantitativePCRforshisa2geneson36h.p.f.wholelarvaeofsurface“sh(blue)andcave“sh(red).Nosigni“cantdifferencewasfoundbetweencave“shandsurface“shexpression(Mann…WhitneyU-test,P40.05).Theerrorbaristhes.e.ofthemean,andthesamplesizeisthreeineachcase(eachtriplicateisapoolof10…1536h.p.f.larvae).Photographsofinsituhybridizationfortheindicatedshisa2mRNAat36h.p.f.onsurface“sh(b…d,h…j)andcave“sh(e…g,k…m)embryos,focusingonheadandeyeexpression.Thebottompictures(c,d,f,g,i,j,l,m)arecenteredontheeyeregion,withfocuseitheronthelens/retinaorontheoverlyingskin.Inallpanels,anteriorisleftanddorsalisup.Inb,e,h,k,thephotographsweretakenfromlightlylabelledembryos(theepidermisisbarelylabelled)andinc,d,f,g,i,j,l,m,thephotographsweretakenfrommorestronglylabelledembryos(epidermisexpressionisvisible).Epid,epidermis;olfepit,olfactoryepithelium(nose).Thescalebarsare25mmforpanelsb,e,h,kand10mmfortheotherpanels. NATURECOMMUNICATIONS|DOI:10.1038/ncomms6307ARTICLENATURECOMMUNICATIONS|5:5307|DOI:10.1038/ncomms6307|www.nature.com/naturecommunications5&2014MacmillanPublishersLimited.Allrightsreserved.

PAGE 6

numberofmelanocytesinfourregionsofthebody(LG9(refs4,8)).Second,cave“shhaveanincreasednumberoftastebudsandincreasednumberofmaxillaryteeth4,8.AQTLfornumberoftastebudscontainstheserotoninreceptorhtr2a(LG5(refs4,8)),andtastecelldevelopmentandsignaltransductioninvolvesserotoninsignalling50,51.Third,aQTLforthenumberofmaxillaryteethincave“sh(LG13(refs4,8))containsdact2,whichsigni“cantlyinhibitsdlx2duringtoothformationinmouse52.Whenknockedoutinmice,dlx2producesadecreaseinthenumberofmolars53,supportingthenotionthatthisgenemayhaveaconservedroleintheregulationoftoothformation.Analysesofputativegenelosses.Weinvestigatedgenesthatwereputativelylostinthecave“shlineagesincethedivergenceofcave“shandzebra“sh,byexamininggenesthatwerepresentinzebra“shandeightadditionalactinopterygianteleostsavailableinEnsembl(SupplementaryData18).Thesegeneswerenotenrichedfor305geneontologyaccessionsrelatedtoeyedevelopmentorfunction,andsimilarresultswereobtainedforZFINanatomicalexpressiondataandZFIN-predictedphenotype.Transcriptomedatafromtheeighttissuesusedforgeneannotationandthedevelopmentalsurface“shandcave“shtimeserieswereassembledusingTrinity54foratotalof10separatetranscriptomes.OpenreadingframeswerepredictedfromtheseassembliesusingTransdecoderintheTrinitypackage.WeconstructedaBLASTdatabasefromthecodingregionsofzebra“shfromEnsemblGenes74andqueriedthisdatabaseusingeachofthetranscriptsinthelongest_orfs.cds“leswithBLASTn.Weusedastronge-valuecutoff(cutoffo1E-100),andresultswererobustforallvaluesweexaminedfrom1E-20to1E-100.Inthisway,weidenti“edwhethertheputativelymissinggeneinthecave“shgenome(butpresentinthezebra“shgenome)waspotentiallypresentinthesurfaceorcave-derivedRNAseqdata.Forseveralgenesthatwerepotentialcandidatesforloss,wecouldnot“ndarepresentativetranscriptforcave“shbutcould“ndatranscriptcopyamongthesurface“shtranscriptomedata(SupplementaryData19).Weattemptedtocon“rmthelackofatranscriptincave“shusingreversetranscription.However,forallcasesthatwetriedtocon“rmaputativelymissingcave“shtranscript,acave“shtranscriptwasdetected.Althoughnotaddingevidenceforcave“sh-speci“closs,forseverallargegenefamilies,oneorseveralmemberswerenotannotatedinthegenomesequenceandwerenotdetectedinsurfaceorcave“shtranscriptomedata.Whiletheseresultsareverypreliminary,potentialcandidatesforgenelossincludemembersofgenefamiliesinvolvedinvisionsuchasretinoldehydrogenases,crystallins,sineoculishomeoboxes,opsins/rhodopsins(includingmelanopsinwhosetruncatingmutationisimplicatedinthelossofalight-entrainableclockinSomaliancave“sh55),development,regulationofsleepandcircadianclocks(including“broblastgrowthfactors,gamma-aminobutyricacidAreceptors,anddopaminereceptors).Likewise,cave“shexhibitexcessivelocomotoractivitycomparedwithsurface“sh56,andseveralgenesthatinducehyperlocomotionwhenknockedoutorblockedinmiceorzebra“shdonotappeartobepresentinthecurrentcave“shgenomeannotationortranscriptomedata(SupplementaryData19).Interestingly,thenakedmolerat,aspeciesthatalsolivesindarknessandhasreducedeyes,hasalsoexperiencedlossesinsimilargenefamilies57.Assemblyandannotationerrorsoflargegenefamiliesarecommonindraftgenomes;thus,amoreextensiveandde“nitiveexplorationofthesecomplexgenefamiliesawaitsfuturestudies.Weprovidealistfromtheinitial,preliminaryanalysis(SupplementaryData19)tofacilitatefuturestudies.DiscussionInthiswork,wepresentadraftgenomeoftheMexicancave“sh,Astyanaxmexicanusandidentifycandidategenesforsomeofthespeciesmosticonicphenotypes.Pasteffortshavefocusedonmappingtraitstogenomicregions4,7…10,15.Byleveragingthesepaststudies,wedemonstratetheutilityofthegenomeforcandidategenediscovery,andhighlightseveralpotentialregulatorsofeyedevelopmentthatwerepreviouslynotimplicatedincave“sheyedegeneration.WealsoanalysedRNAseqdatatoidentifycodingvariantsbetweencave“shandsurface“shandnarrowthelistofcandidategenesthatpotentiallyimpactdegenerationoftheeye.Identi“cationofcandidategenesfrompastQTLworkisespeciallyexcitinginA.mexicanusbecausecave“shareamenabletoahostofmoleculargenetictechniquesthatcanbeusedtovalidatealleliceffects(forexample,injectionofmessengerRNAintodevelopingembryo13,meganuclease-andtransposase-basedtransgenesis14)andadditionalexperimentaltechniquescanbeaccomplishedusingthecloserelativeandlaboratorymodelzebra“sh5(forexample,geneeditingtechnologiessuchasTALENs).Thus,cave“shrepresentapowerfulsystemforexaminingthegeneticbasesofevolutionarychange,andweexpectprogressincandidate aSemi quantitative PCR on otx2 gene at 40 h.p.f.otx2 mRNA18S rRNASFCFSF10 h.p.f.12.5 h.p.f.40 h.p.f.48 h.p.f.40 h.p.f.48 h.p.f.10 h.p.f.12.5 h.p.f.CFSFLensLensCFbc Figure4|Expressionpatternsofotx2.(a)Semi-quantitativereversetranscriptase-PCRfortheoxt2geneson40h.p.f.wholeembryos.Cave“sh(CF)otx2transcriptsareslightlylessabundantthanthoseofsurface“sh(SF)comparedwithan18SrRNAstandard.(b)Photographsofinsituhybridizationsforotx2mRNAat10,12.5,40and48h.p.f.onsurface“sh(SF)andCFembryos,focusingonheadandeyeexpression.Inallpanels,anteriorisontheleft.Inlowerpanels,dorsalisup.Scalebarsare100mmforpanelslabelled40and48h.p.f.in(b).Scalebarsare250mmforpanelslabelled10and12.5h.p.f.inb.(c)Sectionsofinsitu-hybridizedSFandCFlarvaeat48h.p.f.showstrongotx2downregulationinthecave“shlens.Scalebarsare100mmforpanelsinc. ARTICLENATURECOMMUNICATIONS|DOI:10.1038/ncomms63076NATURECOMMUNICATIONS|5:5307|DOI:10.1038/ncomms6307|www.nature.com/naturecommunications&2014MacmillanPublishersLimited.Allrightsreserved.

PAGE 7

geneidenti“cationwillbemoreef“cientwiththeadditionofthe genome. Whileacandidategeneapproachhasstrongpotentialinthe cave“shsystem,ourexpressionanalyseshighlightthediscoveries enabledbyapathwayapproach.Inoneexample,ouranalysis predictedthereduced-eyephenotypeincave“shrelativeto surface“shbyutilizingthedirectionofexpressionwithineye developmentpathways.Notably,thisresultwasoneofthemost signi“cantphenotypespredictedasadownstreamphenotypic effectfromourdevelopmentalgeneexpressiontimecourseof wholeembryos(SupplementaryData13…16).Nonetheless,many databases,includingtheonesusedinthisstudy(Ingenuity PathwayAnalysis(IPA)),containonlyorthologuesfromhuman, mouseandrat,andapproximately89%ofthegenes(2013/2408) inourQTLdatasetcontainedmatchesintheIPAdatabase.This underscorestheneedforfutureiterationsofthesedatabasesto includenonmammalianmodelspecies(forexample,zebra“sh, Drosophila )toincreasehomologymatchesand,therefore,enable pathwayanalysisforalargeswathoforganisms. Weanticipatethatthisgenomicresourcewillbecoupledwith oneofthelargeststrengthsofthesystem,therepeatedevolution ofsimilarcave-associatedtraitsinindependentlyderivedcave populations2.Crossesbetween“shfromdifferentcaves complementandrestorecertaincave-derivedphenotypes(for example,rudimentaryeyes)23;thus,atleastsomeofthegenetic changesaccountingforcave-associatedtraitsareuniquetoeach cavelineage9.Inaddition,surfacepopulationsprovideapoolof standinggeneticvariationforthecaves58,andthecave“shsystem offersaninterestingsystemforstudyingadaptationinthefaceof gene”ow2.Toinvestigatethesequestions,ongoingworkthatis beyondthescopeofthispaperincludesapopulationgenomic effortfromseveralcaveandsurfacelocalities.Toenhancethese efforts,thecave“shgenomewillneedtobeanchoredtoa physical,chromosome-scalemap.Worktoproduceahigherqualitydraftusinglong-readtechnologyandagenotyping-bysequencinglinkagemapiscurrentlyunderway.Weexpectthat thisupcoming,reviseddraftwillfurtheraidinvestigationsofthe impactofselection,drift,migrationandgeneticarchitecturein creatingthesesreplicatedphenotypes. Inconclusion,the Astyanax genomepresentedherewillallow fordissectionofthegeneticbasesofconstructiveanddegenerativetraitsthatmakethecave“shdistinctive,willfacilitate futurestudiesinvestigatingthepathsofrepeatedevolutionand mayadvanceunderstandingofhumanmaladies(forexample, sleepdisorders,congenitaleyedefects)forwhichthecave“shcan serveasapowerfulnaturalmodelsystem. MethodsSourcematerial.SourceDNAwasobtainedfromtheJefferyLab.DNAwas collectedfromheart,liver,spleenandgillofasingle7…year-oldadultfemale cave“sh(Pacho ´ n)usingtheGenomic-TipTissueMidikits(Qiagen,Valencia,CA). RNAfromeighttissueswasextractedwithRNALipidMidikitsandRNeasykits (Qiagen).AnimalusecompliedwithethicalstandardsandwasapprovedbyThe UniversityofMarylandInstitutionalAnimalCareandUseCommitteeprotocol numberR-12…53toW.R.Jeffery. Genomesequencingandassembly.Usingagenomesizeestimateof1.19Gb, totalrawsequencecoverageofIlluminareadswas B 95 (shortinsertpaired-end reads,3,8and40kbmate-pairedlibraries,SupplementaryData1).Thecombined sequencereadswereassembledusingALLPATHSsoftware59andtheassembled coveragewas70 .Thisdraftassemblywasreferredtoas Astyanaxmexicanus 1.0.2.Thisassemblyhasbeengap“lledwithaversionofImage60,modi“edfor largegenomesandcleanedofcontaminatingcontigsbyperforminga MegaBLAST61ofthecontigsagainstadapter,bacterialandvertebratedatabases. IdentifyinglocationsofQTLingenomeassembly.Manymarkersoverlapped betweenpreviouslypublishedQTLanalyses,renderingitpossibletocompare coarseQTLlocationsacrossreplicatedmaps4,7,8,15,16,eventhoughLGnameswere notconsistentbetweenstudies.Markers”ankingtheQTLwerelocalizedto scaffoldsviabestBLASThit.Brie”y,acombinationof689RAD-tagsequences, microsatellitemarkersandcDNAswithlinkagemappositions15werealignedto thescaffoldsusingBLASTn. Astyanaxmexicanus hasahaploidchromosome numberof25(ref.62),andtherewereatotalon24LGsrepresentedbythese markers(SupplementaryFig.1). Allmarkerswererequiredtohaveane-valueof1E-20exceptforasubsetof microsatelliteandcDNAmarkerswhereonlytheforwardprimer,reverseprimer andsometimestherepeatmotifsequencewereavailable.Forthesemicrosatellites, wordsizewasreducedto7,andhitswererequiredtohaveane-valueoflessthan 1E-1andidentityofgreaterthan99%.TwocDNAstakenfrom Daniorerio were alsoallowedweakeridentity(85%orgreater)andane-valuecutoffof1E-1. OnlythetopBLASThitforeachmarkerwasrecorded.Scaffoldsthatmappedto differentLGswereexcluded( n ¼ 9).Inthecasewherethreeormoremarkers mappedthescaffoldtooneLGandonlyasinglemarkermappedthescaffoldtoa differentLG( n ¼ 3),onlytheincongruentmarkerwasexcluded.Intotal,340 scaffoldswerelocalizedtoLGsrepresenting574Mbofsequence(Supplementary Data17).Scaffoldswithonlyasinglemarker(219scaffolds)wereorderedalongthe chromosomeinlinewiththegeneticmapasdescribedinref.15andthe orientationwasassignedrandomly.Orientationwasassignedforscaffoldswithtwo ormoremarkers(121scaffolds, B 339Mb)physicallyandgeneticallymapped (SupplementaryData17;SupplementaryFig.1).Suchamapisverysimilartowhat isgivenintheSupplementaryMaterialsofref.9. Repetitivelandscape.Repeats,includingTEs,wereidenti“edandannotated usingRepeatModelersoftwarewithdefaultparameters.Theannotationfollowsthe universalclassi“cation63.Theautomaticlibrarywasscreenedto“lteranddiscard sequencessharinghighsimilaritieswithUniprotprotein.Paralleltotheautomatic annotation,potentiallyabsentfamiliesthatwerenotfoundusingRepeatModeler weremanuallysearchedbyBLASTusingknownTEproteins.Theunplaced scaffoldsweremaskedusingRepeatMasker3.3.0(http://www.repeatmasker.org/) withthecave“sh-speci“crepeatedlibraryusing…liband…alignfunctions.Results wereparsedtodeterminecopynumberandcoverageofTEsuperfamiliesusingthe RepeatMaskerout“les. ToinvestigatethelevelofTEtranscription,weanalysedthreedifferent assembledtranscriptomes:muscle,brainandwhole-eyesurface“sh. TranscriptomesweremaskedusingRepeatMasker3.3.0usingthespeci“ccave“sh repeatlibrary,aswasdoneforthegenomicanalyses.Theproportionofthevarious classes(forexample,DNA,LINE,SINE,LTRandunknownelements)were comparedwiththeirrespectiveproportionsinthegenome(camembertgraphs). Weassayedforover-andunder-representationofTEsuperfamiliesbycomparing therespectiveproportionofeachfamilyandsuperfamilyinthegenomeandinthe transcriptomes.Thefollowingequationwasused:(percentageoftheTEfamilyin thegenome[ortranscriptome]*100)/Totalrepeatcontentinthegenome[or transcriptome]). Genepredictionandannotation.Iterativestepsthatrelyonsimilarityevidence frompriorteleostgenemodelsand abinitio genepredictionalgorithmswere followedtobuildgenemodelsaccordingtoestablishedmethodsatEnsembl18. Protein-codingmodelswereextendedintoUTRregionsandcompletedexon modelswerevalidatedwithRNAseqdata(RNAseqanalysessectionbelow)from diversetissuetypes.Additionalmethodsfollowedhereforgeneratinggenebuilds byEnsemblarelocatedat:http://useast.ensembl.org/info/genome/genebuild/ 2013_10_cave“sh_genebuild.pdf.Althoughnotusedinthesestudies,asecondgene setproducedwiththeNCBIgeneannotationpipelineisavailableat:ftp:// ftp.ncbi.nlm.nih.gov/genomes/Astyanax_mexicanus/. RNAseqanalyses.RNAseqdatawasobtainedfromtwodifferentsequencing efforts.The“rstconsistedoftissue-speci“c100-bppaired-endIlluminareads whichwereusedforgenomeannotationbyEnsembl(SupplementaryData2). Theseincludedsamplesfrombrain,heart,kidney,liver,muscle,nasalcavityand skinfromadultPacho ´ ncave“shandeyesfromadultsurface“shfromTexas.For alltissues,multiple(onetosix)individualswerepooled,exceptforeyeswhereone surfaceindividualwasused. ThesecondRNAseqsequencingeffortconsistedofadevelopmentaltimecourse ofembryostakenfrom10h.p.f.,24h.p.f.,1.5d.p.f.and3d.p.f.Fiftyindividuals werepooledforeachtimepoint.ThreeseparateTruSeq2Illuminalibrarieswere madeforeachtimepointfromthesamepoolofRNA,providingtechnical replicates. Time-courseRNAseqdatawerecleanedbytrimmingthe“rstbasewith Fastx_trimmer(http://hannonlab.cshl.edu/fastx_toolkit/),trimmingwith Trimmomaticv0.30(ref.64)usingtheadapterlibraryforTruSeq2,allowinga qualityscoreof30acrossa4-bpslidingwindowandremovingallreads o 30 nucleotidesinlengthafterprocessing.Readswerealignedtothereferencegenome usingTopHat2(ref.65)withdefaultparametersexceptthatthemaximumintron lengthwassetto10,000.Cuf”inks2.1.0(ref.28)wasusedtocalculatedifferencesin expressionbetweencaveandsurfaceRNAseqdata.Cuf”inkswasusedwiththe parameters:--frag-bias-correct--multi-read-correct--upper-quartile-norm-compatible-hits-norm(withthegtf“leforthegenome).Cuffdiffwasusedwiththe NATURECOMMUNICATIONS|DOI:10.1038/ncomms6307ARTICLENATURECOMMUNICATIONS |5:5307|DOI:10.1038/ncomms6307|www.nature.com/naturecommunications7& 2014 MacmillanPublishersLimited.Allrightsreserved.

PAGE 8

parameters:--frag-bias-correct--multi-read-correct--FDR0.1--dispersionmethodper-condition. Time-courseRNAseqworkwasperformedunderaprotocolapprovedbythe InstitutionalAnimalCareandUseCommittee(IACUC)oftheUniversityof Cincinnati;ProtocolNumber10-01-21-01toJ.B.G.andcompliedwithethical regulationfortreatmentofanimals.Samplesforgenomeannotationweretakenin theJefferylabunderanimalcareprotocolsreferencedabove. Transcriptvariantanalyses.Atmost,15Pacho ´ ncave“shcontributedtothe offspringusedintheRNAseqexperiment(thesameninebreedingindividualsfrom Pacho ´ nLine163wereusedtogenerateembryosfor10h.p.f.,24h.p.f.,3d.p.f.time points;sixbreedingindividualsfromPacho ´ nLine138wereusedfor1.5d.p.f.)and atotalofthreeindividualscontributedtothesurfaceembryosfromwhichRNAseq datawerecollected.These“sharelaboratorystocksandlikelysomewhatinbred; thus,wehadlittlepowertoassignallelefrequenciesbetweencaveandsurface pooledsamples.Forallanalyses,weconcentrateondifferencesthatwerecompletely“xedinourdatasetbetweenthesurfaceRNAseqgenotypesandthecave referencegenome þ caveRNAseqgenotypes. Forvariantcalling,weconcatenatedallexpressiondataforsurfaceand separatelyconcatenatedallexpressiondataforcaveandmappedthesesurfaceand cavereadstothereferencegenomeusingTopHat2.Thesealignmentswerepassed toSamtoolsv0.1.19(ref.66)tocreatempileup“lesforcaveandsurfacewhichwere usedbyVarScanv2.3.6(ref.67)tocallsomaticmutationswithsurfacepileups designatedasnormalandcaveastumour.Variantswerethenusedasinputfor EnsemblsstandaloneVariantEffectPredictorv73(ref.68)topredicttheclass (forexample,synonymous,nonsynonymous)ofeachvariant.Todetermineifthe substitutionsidenti“edbyVariantEffectPredictorv73werelikelyderivedin cave“sh,peptidesequencefromorthologuesofzebra“sh,coelcanth,spottedgar, sticklebackandplaty“shwereobtainedfromBiomartandalignedwithClustalW69. Inmostcases,itwasstraightforwardtoclassifywhetherthesurfaceorcaveamino acidwaslikelyderived,andinallothercasesthesiteappearedevolutionarilylabile. Thisanalysisshouldbeinterpretedwiththecaveatthatitdoesnotaccountforthe possibilitythatthevariantclassi“edasderivedincave“shisactuallypresentinthe standinggeneticvariationofthesurface“shwhichwasnotsampledinourdata. SIFTSequencewasusedtopredictthefunctionalimpactofnonsynonymous substitutions43. Identi“cationofcandidategenes.Allpathwayanalyseswereperformedwith theIPAsuiteoftoolsavailableathttp://www.ingenuity.com/products/ipa.The entireanalysis-readypoolcontainedonly65%ofgenesinourQTLdataset (1,560/2,408)(SupplementaryData12)assomeofthegenesunderourQTLandin theIPAdatabasedidnothavesuf“cientexpressiondataforanalysis.Forall enrichmenttests,weusedonlytheanalysis-readygeneset“lteredbyIPA,which doesnotincludemultiplegeneswiththesameEntrezgenenameorgeneslacking expressiondatainourdataset. InadditiontotheIPAanalyses,whichdidnotannotateallofthegenesunder theQTL,weconductedindependentliteraturesearchesongenesandprioritized thosethatwere(1)differentiallyexpressedinatleastoneofthedevelopmentaltime points;(2)containedatleastone“xednonsenseormissensedifferencebetween caveandsurface“sh;or(3)exhibitedexpressioninaneye-relatedstructureduring developmentofthezebra“sh(ZFINanatomicaldatabase)orhadageneontology annotationordescriptionrelatedtoeye,retina,lensoropticfunction. QuantitativePCRfor shisa2.Forthe shisa2insitu hybridizationandqPCR experiments,laboratorystocksof A.mexicanus surface“shoriginatedfromSan SolomonSpring,BalmorheaStatePark,Texas.Cave“shfromPacho ´ ncavewere obtainedin2004…2006fromtheJefferylaboratoryattheUniversityofMaryland, CollegePark,MD,andweresincethenbredintheGIFanimalfacility. TotalRNAwasextractedfrom36h.p.f.cave“shorsurface“shembryoswith TRIzolreagent(Invitrogen)followedbypuri“cationandDNasetreatmentwiththe MachereyNagelNucleoSpinRNAIIkit.RNAamountsweredeterminedbythe Nanodrop2000cspectrophotometer(ThermoScienti“c).TotalRNA(1 m g)was reversetranscribedina20m l“nalreactionvolumeusingtheHighCapacitycDNA ReverseTranscriptionKit(LifeTechnologies,GrandIsland,NY,USA)withRNase inhibitorandrandomprimersfollowingthemanufacturersinstructions. QuantitativePCRwasperformedonaQuantStudio12KFlexReal-TimePCR SystemwithaSYBRgreendetectionprotocol.cDNA(3ng)wasmixedwithFast SYBRGreenMasterMixand500nMofeachprimerina“nalvolumeof10 m l.The reactionmixturewassubmittedto40cyclesofPCR(95 C,20s;(95 C,1s;60 C, 20s) 40)followedbyafusioncycletoanalyzethemeltingcurveofthePCR products.Negativecontrolswithoutthereversetranscriptasewereintroducedto verifytheabsenceofgenomicDNAcontaminants.Primersweredesignedbyusing thePrimer-BLASTtoolfromNCBIandthePrimerExpress3.0software(Life Technologies).Primerswerede“nedeitherinoneexonandoneexon…exon junctionorintwoexonsspannedbyalargeintron.Speci“cityandtheabsenceof multi-locusmatchingattheprimersitewereveri“edbyBLASTanalysis.The ampli“cationef“cienciesofprimersweregeneratedusingtheslopesofstandard curvesobtainedbyafour-folddilutionseries.Ampli“cationspeci“cityforeach real-timePCRreactionwascon“rmedbyanalysisofthedissociationcurves. Determined Ctvalueswerethenexploitedforfurtheranalysis,withthe Gapdh and Actb1 genesasreferences.Eachsamplemeasurementwasmadeatleastin duplicate.PrimersequencesforLG6shisa2 were0974-AM-LG6-F150-CGCAGTG CCCATCTACGTG-30and0975-AM-LG6-R150-TGTTTGGGTCGCAGAC AGC-30.ForLG2shisa2 ,theprimersequenceswere0982-AM-LG2-F350-GGGCA CCACAGTTTTTCCAA-30and0983-AM-LG2-R350-CTGTCCGTGTGCCTG ACTGA-30.For Gapdh and Actb1 ,primerswere0970-AMgapdh-F150-GTTGGC ATCAACGGATTTGG-30and0971-AMgapdh-R150-CCAGGTCAATGAAGG GGTCA-30and0972-AMactb1-F250-GCCATCATGCGTCTTGACCT-30and 0973-AMactb1-R250-ATCTCACGCTCAGCGGTTGT-30,respectively. For shisa2 work,animalsweretreatedaccordingtotheFrenchandEuropean regulationsforhandlingofanimalsinresearch.Authorizationforuseofanimals forthisworkwasprovidedbyParisCentre-SudEthicCommittee(authorization number2012-0052)toS.R.(number91…116). QuantitativePCRfor otx2.TotalRNAwasisolatedfrom40h.p.f.surface“shand Pacho ´ ncave“shlarvaeusingTRIzol(LifeTechnologies).cDNAwassynthesized usingtheSuperScriptTMIIIFirst-StrandSynthesisSuperMixKitandoligo(dT)20primers(LifeTechnologies).Forsemi-quantitativereversetranscriptase-PCR,part ofthe otx2 codingregionwasampli“edfromcDNAwithprimers50-ATGATGT CGTATCTCAAGCAACC-30(forward)and50-TAATCCAAGCAGTCGGCGTT GAAG-30(reverse)usingPCRMaster(RocheAppliedScience,Indianapolis,IN, USA),whichyieldedan otx2 PCRproductof857bp.ThePCRcyclingconditions were:onecycleofinitialdenaturationat94 Cfor5min,followedby35cyclesof denaturation(94 Cfor30s),annealing(58 Cfor30s)andelongation(72 Cfor 45s)anda“nalelongationstepat72 Cfor7min.Ampli“cationofthecontrol18S rRNAwascarriedoutusing1 m lofthesynthesizedcDNAwithprimersina50m l reactionvolumeusingPCRMaster(Roche).The18SrRNAprimerswere50-GAG TATGGTTGCAAAGCTGAAA-30(forward)and50-CCGGACATCTAAGGG CATCA-30(reverse),whichyieldedaPCRproductof343bp.ThePCRcycling conditionswere:onecycleofinitialdenaturationat94 Cfor5min,followedwith 25cyclesofdenaturation(94 Cfor30s),annealing(at62 Cfor30s)andelongation(at72 Cfor30s),followedbya“nalelongationstepat72 Cfor7min. Whole-mount insitu hybridizationfor shisa2.cDNAswereampli“edbyPCR frompCMV-Sport6plasmidspickedfromourcDNAlibraryusingSP6andT7 primersanddigoxygenin-riboprobesweresynthesizedfromPCRtemplates.A protocolforautomatedwhole-mount insitu hybridization(Intavis)wasperformed. Brie”y,embryoswereprogressivelyrehydrated,permeabilizedbyproteinaseK (Sigma)treatmentbeforebeingincubatedovernightat68 Cinhybridization buffercontainingtheappropriate shisa2 probe.Afterstringentwashes,the hybridizedprobesweredetectedbyimmunohistochemistryusinganalkaline phosphatase-conjugatedantibodyagainstdigoxygenin(Roche)andaNBT/BCIP chromogenicsubstrate(Roche). Whole-mount insitu hybridizationfor otx2.Forprobepreparation,the otx2 codingregionfragmentwasampli“edfromsurface“shcDNAwithPCRMaster (Roche)accordingtotheHotstartPCRprotocolusingthe otx2 primersdescribed above.ThePCRcyclingconditionswere:onecycleofinitialdenaturation(94 C) for2min,followedby32cyclesofdenaturation(94 Cfor30s),annealing(58 C for30s)andelongation(72 Cfor45s)anda“nalelongationstep(72 Cfor 7min).The“rstPCRproductwasusedasthetemplateforasecondcycleof PCRampli“cationusingsameconditions.Theresulting857bpPCRproductwas clonedintotheTOPOvectorintheTPOTACloningKitDualPromotor(Life Technologies)andcon“rmedbysequencing. Insitu hybridizationwasperformedaccordingtoref.70withsome modi“cations.TheplasmidDNAwaslinearizedwithrestrictionenzymes Bam H1 and Xho I(LifeTechnologies)at37 Cfor1handpuri“edwiththeQIAquickPCR Puri“cationKit(Qiagen).Senseandantisensedigoxigenin(DIG)-labelledRNAs weretranscribedwithSP6RNAandT7RNAPolymerases(Roche).The invitro transcriptionreactionswereconductedaccordingtotheDIGRNALabelingMix (Roche)protocol.Thereactionswereterminatedwith0.2MEDTA(pH8.0),and RNAwasprecipitatedwith4MLiClandwashedinprechilled70%ethanol.The RNAprobewasdenaturedfor3minat95 C,quicklycooledonicefor5minand thenaddedtotheHYB þ (seebelow)toobtainaconcentratedstock(10 m gml 1). Theembryoswere“xedwith4%paraformaldehydeinPBSovernightat4 C, dehydratedinanincreasingmethanolseriesandstoredat 20 C.Rehydrated embryosweretreatedwithproteinaseK(10 m gml 1inPBST(PBSplus0.1% Tween20))for5…10minatroomtemperature,washedtwicewithPBST,post“xed for20minwith4%paraformaldehydeinPBSTandwashed5timeswithPBST (5mineach).TheembryoswerepretreatedwithHYB (50%formamide,5 SSC,0.1%Tween20)for5minat60 Cwithoutshaking.TheHYB wasreplaced withHYB þ (HYB ,1mgml 1yeastRNA,50 m gml 1heparin)andthe embryoswereprehybridizedat60 Cfor4hwithgentleshaking.The prehybridizationmixwasremovedandreplacedwith1ng m l 1of otx2 senseor antisenseprobeinHYB þ .Hybridizationwascarriedoutat60 Covernight withgentleshaking.Theembryoswerethenwashedtwiceat60 Cwith50% formamide/2 SSCT(salinesodiumcitrateplus0.1%Tween20)for30mineach, oncewith2 SSCTfor15minat60 C,twicewith0.2 SSCT(20mineach)at ARTICLENATURECOMMUNICATIONS|DOI:10.1038/ncomms63078NATURECOMMUNICATIONS |5:5307|DOI:10.1038/ncomms6307|www.nature.com/naturecommunications& 2014 MacmillanPublishersLimited.Allrightsreserved.

PAGE 9

60 CandtwicewithMABT(150mMmaleicacid,100mMNaCl,pH7.5,0.1% Tween20)for5mineachatroomtemperature.Theembryoswereincubatedwith blockingsolution(MABT,2%blockingreagent)overnightat4 Cwithrockingand thenwithAnti-DIG-APFabfragments(1:5,000;Roche)inblockingsolution overnightat4 Cwithgentlerocking.TheembryoswerewashedoncewithMABT containing10%sheepserumatroomtemperaturefor25minandeightmoretimes (45…60mineach)withMABTatroomtemperaturewithgentlyshaking.Then,the embryoswerewashedwithPBSTandincubatedinBMPurpleAPSubstrate (Roche)atroomtemperatureinthedark.Afterthesignaldeveloped,thereaction wasterminatedbyrinsingtheembryosseveraltimesinPBST.Embryoswere processedthroughanincreasingglycerolseriesinPBS(30…50…80%)andimagedby microscopy. Insitu -hybridizedembryosweredehydratedthroughanethanolseries(from 30,50,70,85,95%,andthree100%steps)for20mineachatroomtemperature. Thedehydratedembryoswereincubatedinethanol:Histo-Clear(1:1)withrotation for20min,intwochangesofHisto-Clearfor30mineach),inparaf“n:Histo-Clear (1:1)at62 Cfor1hand“nally100%paraf“nat62 Cfor2h.Theblocks containingembeddedembryoswerecutinto15m msections,andthesectionswere dewaxedandviewedbymicroscopy.References1.Elmer,K.R.&Meyer,A.Adaptationintheageofecologicalgenomics:insights fromparallelismandconvergence. TrendsEcol.Evol. 26, 298…306(2011). 2.Bradic,M.,Beerli,P.,Leo ´ n,F.G.-d.,Esquivel-Bobadilla,S.&Borowsky,R.Gene ”owandpopulationstructureintheMexicanblindcave“shcomplex( Astyanax mexicanus ). BMCEvol.Biol. 12, 9…9(2012). 3.Coghill,L.M.,DarrinHulsey,C.,Chaves-Campos,J.,Garcš ´ adeLeon,F.J.& Johnson,S.G.Nextgenerationphylogeographyofcaveandsurface Astyanax mexicanus . Mol.Phylogenet.Evol. 79C, 368…374(2014). 4.Protas,M.,Conrad,M.,Gross,J.B.,Tabin,C.&Borowsky,R.Regressive evolutionintheMexicancavetetra, Astyanaxmexicanus . Curr.Biol. 17, 452…454(2007). 5.Gross,J.B.,Borowsky,R.&Tabin,C.J.Anovelrolefor Mc1r intheparallel evolutionofdepigmentationinindependentpopulationsofthecave“sh Astyanaxmexicanus . PLoSGenet. 5, e1000326…e1000326(2009). 6.Protas,M.E. etal. Geneticanalysisofcave“shrevealsmolecularconvergencein theevolutionofalbinism. Nat.Genet. 38, 107…111(2006). 7.Yoshizawa,M.,Yamamoto,Y.,OQuin,K.E.&Jeffery,W.R.Evolutionofan adaptivebehavioranditssensoryreceptorspromoteseyeregressioninblind cave“sh. BMCBiol. 10, 108(2012). 8.Protas,M. etal. Multi-traitevolutioninacave“sh, Astyanaxmexicanus . Evol. Dev. 10, 196…209(2008). 9.Kowalko,J.E. etal. Convergenceinfeedingpostureoccursthroughdifferent geneticlociinindependentlyevolvedcavepopulationsof Astyanaxmexicanus . Proc.NatlAcad.Sci.USA 110, 16933…16938(2013). 10.Kowalko,J.E. etal. Lossofschoolingbehaviorincave“shthroughsightdependentandsight-independentmechanisms. Curr.Biol. 23, 1874…1883 (2013). 11.Elipot,Y.,Hinaux,H.,Callebert,J.&Re ´ taux,S.Evolutionaryshiftfrom“ghting toforaginginblindcave“shthroughchangesintheserotoninnetwork. Curr. Biol. 23, 1…10(2013). 12.Duboue ´ ,E.R.,Keene,A.C.&Borowsky,R.L.Evolutionaryconvergenceon sleeplossincave“shpopulations. Curr.Biol. 21, 671…676(2011). 13.Yamamoto,Y.,Stock,D.W.&Jeffery,W.R.Hedgehogsignallingcontrolseye degenerationinblindcave“sh. Nature 431, 844…847(2004). 14.Elipot,Y.,Legendre,L.,Pe ` re,S.,Sohm,F.&Re´ taux,S. Astyanax transgenesis andhusbandry:howcave“shentersthelaboratory. Zebra“sh 11, 291…299 (2014). 15.OQuin,K.E.,Yoshizawa,M.,Doshi,P.&Jeffery,W.R.Quantitativegenetic analysisofretinaldegenerationintheblindcave“sh Astyanaxmexicanus . PLoS ONE 8, e57281(2013). 16.Gross,J.B. etal. Syntenyandcandidategenepredictionusingananchored linkagemapof Astyanaxmexicanus . Proc.NatlAcad.Sci.USA 105, 20106…20111(2008). 17.Carvalho,M.L.,Oliveira,C.,Navarrete,M.C.,Froehlich,O.&Foresti,F. NuclearDNAcontentdeterminationinCharaciformes“sh(Teleostei, Ostariophysi)fromtheNeotropicalregion. Genet.Mol.Biol. 25, 49…55(2002). 18.Flicek,P. etal. Ensembl2012. NucleicAcidsRes. 40, D84…D90(2012). 19.Nakatani,M.,Miya,M.,Mabuchi,K.,Saitoh,K.&Nishida,M.Evolutionary historyofOtophysi(Teleostei),amajorcladeofthemodernfreshwater“shes: PangaeanoriginandMesozoicradiation. BMCEvol.Biol. 11, 177(2011). 20.Parra,G.,Bradnam,K.&Korf,I.CEGMA:apipelinetoaccuratelyannotate coregenesineukaryoticgenomes. Bioinformatics 23, 1061…1067(2007). 21.Howe,K. etal. Thezebra“shreferencegenomesequenceanditsrelationshipto thehumangenome. Nature 496, 498…503(2013). 22.Aparicio,S. etal. Whole-genomeshotgunassemblyandanalysisofthegenome of Fugurubripes . Science 297, 1301…1310(2002). 23.Borowsky,R.Restoringsightinblindcave“sh. Curr.Biol. 19, R23…R24(2008). 24.Alunni,A. etal. Developmentalmechanismsforretinaldegenerationinthe blindcave“sh Astyanaxmexicanus . J.Comp.Neurol. 505, 221…233(2007). 25.Jeffery,W.R.&Martasian,D.P.Evolutionofeyeregressioninthecave“sh Astyanax :apoptosisandthe Pax-6 gene. Am.Zool. 38, 685…696(1998). 26.Strickler,A.G.,Byerly,M.S.&Jeffery,W.R.Lensgeneexpressionanalysis revealsdownregulationoftheanti-apoptoticchaperone a A-crystallinduring cave“sheyedegeneration. Dev.GenesEvol. 217, 771…782(2007). 27.Kra ¨ mer,A.,Green,J.,Pollard,J.&Tugendreich,S.Causalanalysisapproaches inIngenuityPathwayAnalysis. Bioinformatics 30, 523…530(2014). 28.Trapnell,C.etal. Differentialanalysisofgeneregulationattranscriptresolution withRNA-seq. Nat.Biotechnol. 31, 46…53(2013). 29.Shi,X. etal. Zebra“sh pitx3 isnecessaryfornormallensandretinal development. Mech.Dev. 122, 513…527(2005). 30.Zilinski,C.A.,Shah,R.,Lane,M.E.&Jamrich,M.Modulationofzebra“sh pitx3 expressionintheprimordiaofthepituitary,lens,olfactoryepitheliumand cranialgangliaby hedgehog and nodal signaling. Genesis 41, 33…40(2005). 31.Loosli,F. etal. Lossofeyesinzebra“shcausedbymutationof chokh/rx3 . EMBORep. 4, 894…899(2003). 32.Loosli,F. etal. Medakaeyelessisthekeyfactorlinkingretinaldetermination andeyegrowth. Development 128, 4035…4044(2001). 33.Lee,J.-A.,Anholt,R.R.&Cole,G.J.Olfactomedin-2mediatesdevelopmentof theanteriorcentralnervoussystemandheadstructuresinzebra“sh. Mech.Dev. 125, 167…181(2008). 34.Ng,D. etal. OculofaciocardiodentalandLenzmicrophthalmiasyndromes resultfromdistinctclassesofmutationsin BCOR . Nat.Genet. 36, 411…416 (2004). 35.Lee,J.,Lee,B.-K.&Gross,J.M.Bcl6afunctionisrequiredduringopticcup formationtopreventp53-dependentapoptosisandcolobomata. Hum.Mol. Genet. 22, 3568…3582(2013). 36.Yamamoto,A.,Nagano,T.,Takehara,S.,Hibi,M.&Aizawa,S.Shisapromotes headformationthroughtheinhibitionofreceptorproteinmaturationforthe caudalizingfactors,WntandFGF. Cell 120, 223…235(2005). 37.Silva,A.,Filipe,M.,Vitorino,M.,Steinbeisser,H.&Belo,J.Developmental expressionof Shisa-2 in Xenopuslaevis . Int.J.Dev.Biol. 50, 575…579(2006). 38.Thisse,B.&Thisse,C. FastReleaseClones:AHighThroughputExpression Analysis (ZFINDirectDataSubmission.,2004). 39.Ogino,H.,Ochi,H.,Reza,H.M.&Yasuda,K.Transcriptionfactorsinvolvedin lensdevelopmentfromthepreplacodalectoderm. Dev.Biol. 363, 333…347 (2012). 40.Wigle,J.T.,Chowdhury,K.,Gruss,P.&Oliver,G. Prox1 functioniscrucialfor mouselens-“breelongation. Nat.Genet. 21, 318…322(1999). 41.Jeffery,W.R.,Strickler,A.G.,Guiney,S.,Heyser,D.G.&Tomarev,S.I. Prox1 ineyedegenerationandsensoryorgancompensationduringdevelopmentand evolutionofthecave“sh Astyanax . Dev.GenesEvol.210, 223…230(2000). 42.Bouaita,A. etal. Downregulationofapoptosis-inducingfactorinHarlequin miceinducesprogressiveandsevereopticatrophywhichisdurablyprevented byAAV2AIF1 genetherapy. Brain 135, 35…52(2012). 43.Kumar,P.,Henikoff,S.&Ng,P.C.PredictingtheeffectsofcodingnonsynonymousvariantsonproteinfunctionusingtheSIFTalgorithm. Nat. Protoc. 4, 1073…1081(2009). 44.Jomary,C.&Jones,S.E.Inductionoffunctionalphotoreceptorphenotypeby exogenous Crx expressioninmouseretinalstemcells. Invest.Ophthalmol.Vis. Sci. 49, 429…437(2008). 45.Shen,Y.-c.&Raymond,P.A.Zebra“sh cone-rod ( crx )homeoboxgene promotesretinogenesis. Dev.Biol. 269, 237…251(2004). 46.Meng,F. etal. Evolutionoftheeyetranscriptomeunderconstantdarknessin Sinocyclocheilus cave“sh. Mol.Biol.Evol. 30, 1527…1543(2013). 47.Behesti,H.,Papaioannou,V.&Sowden,J.Lossof Tbx2 delaysopticvesicle invaginationleadingtosmallopticcups. Dev.Biol. 333, 360…372(2009). 48.Thu,H.N.T.,Tien,S.F.H.,Loh,S.L.,Yan,J.S.B.&Korzh,V. tbx2a Is requiredforspeci“cationofendodermalpouchesduringdevelopmentofthe pharyngealarches. PLoSONE 8, e77171(2013). 49.Gross,J.M.&Dowling,J.E. Tbx2b isessentialforneuronaldifferentiation alongthedorsal/ventralaxisofthezebra“shretina. Proc.NatlAcad.Sci.USA 102, 4371…4376(2005). 50.Roper,S.D.Tastebudsasperipheralchemosensoryprocessors. Semin.Cell Dev.Biol. 24, 71…79(2013). 51.Ortš ´ z-Alvarado,R. etal. Expressionoftryptophanhydroxylaseindeveloping mousetastepapillae. FEBSLett. 580, 5371…5376(2006). 52.Li,X.,Florez,S.,Wang,J.,Cao,H.&Amendt,B.Dact2repressesPITX2 transcriptionalactivationandcellproliferationthroughWnt/beta-catenin signalingduringodontogenesis. PLoSONE 8, e54868(2013). 53.Qiu,M. etal. RoleoftheDlxhomeoboxgenesinproximodistalpatterningof thebranchialarches:mutationsofDlx-1,Dlx-2,andDlx-1and-2alter morphogenesisofproximalskeletalandsofttissuestructuresderivedfromthe “rstandsecondarches. Dev.Biol. 185, 165…184(1997). 54.Grabherr,M.G. etal. Full-lengthtranscriptomeassemblyfromRNA-Seqdata withoutareferencegenome. Nat.Biotechnol. 29, 644…652(2011). NATURECOMMUNICATIONS|DOI:10.1038/ncomms6307ARTICLENATURECOMMUNICATIONS |5:5307|DOI:10.1038/ncomms6307|www.nature.com/naturecommunications9& 2014 MacmillanPublishersLimited.Allrightsreserved.

PAGE 10

55.Cavallari,N. etal. Ablindcircadianclockincave“shrevealsthatopsins mediateperipheralclockphotoreception. PLoSBiol. 9, e1001142(2011). 56.Sharma,S.,Coombs,S.,Patton,P.&BurtdePerera,T.Thefunctionof wall-followingbehaviorsintheMexicanblindcave“shandasightedrelative, theMexicantetra( Astyanax ). J.Comp.Physiol.A 195, 225…240(2009). 57.Kim,E.B. etal. Genomesequencingrevealsinsightsintophysiologyand longevityofthenakedmolerat. Nature 479, 223…227(2011). 58.Gross,J.&Wilkens,H.Albinisminphylogeneticallyandgeographically distinctpopulationsofAstyanaxcave“sharisesthroughthesameloss-offunctionOca2allele. Heredity 111, 122…130(2013). 59.Gnerre,S. etal. High-qualitydraftassembliesofmammaliangenomesfrom massivelyparallelsequencedata. Proc.NatlAcad.Sci.USA 108, 1513…1518 (2011). 60.Tsai,I.J.,Otto,T.D.&Berriman,M.Method:Improvingdraftassembliesby iterativemappingandassemblyofshortreadstoeliminategaps. GenomeBiol. 11, R41(2010). 61.Zhang,Z.,Schwartz,S.,Wagner,L.&Miller,W.Agreedyalgorithmfor aligningDNAsequences. J.Comput.Biol. 7, 203…214(2000). 62.Kavalco,K.F.&DeAlmeida-Toledo,L.F.Molecularcytogeneticsofblind mexicantetraandcommentsonthekaryotypiccharacteristicsofgenus Astyanax (Teleostei,Characidae). Zebra“sh 4, 103…111(2007). 63.Wicker,T. etal. Auni“edclassi“cationsystemforeukaryotictransposable elements. Nat.Rev.Genet. 8, 973…982(2007). 64.Lohse,M. etal. RobiNA:auser-friendly,integratedsoftwaresolutionfor RNA-Seq-basedtranscriptomics. NucleicAcidsRes. 40 (WebServerissue): W622…W627(2012). 65.Kim,D. etal. TopHat2:accuratealignmentoftranscriptomesinthepresenceof insertions,deletionsandgenefusions. GenomeBiol. 14, R36(2013). 66.Li,H. etal. TheSequencealignment/map(SAM)formatandSAMtools. Bioinformatics 25, 2078…2079(2009). 67.Koboldt,D. etal. VarScan2:Somaticmutationandcopynumberalteration discoveryincancerbyexomesequencing. GenomeRes. 22, 568…576(2012). 68.McLaren,W. etal. Derivingtheconsequencesofgenomicvariantswiththe EnsemblAPIandSNPEffectPredictor. BMCBioinformatics 26, 2069…2070 (2010). 69.Thompson,J.,Higgins,D.&Gibson,T.CLUSTALW:improvingthesensitivity ofprogressivemultiplesequencealignmentthroughsequenceweighting, position-speci“cgappenaltiesandweightmatrixchoice. NucleicAcidsRes. 22, 4673…4680(1994). 70.Strickler,A.G.,Yamamoto,Y.&Jeffery,W.R.EarlyandlatechangesinPax6 expressionaccompanyeyedegenerationduringcave“shdevelopment. Dev. GenesEvol. 211, 138…144(2001).AcknowledgementsThisworkwassupportedbyNIHgrantR24RR032658-01toW.C.W.andTheGenome InstituteatWashingtonUniversitySchoolofMedicine.Collectionswereconductedwith MexicanPermitNumber040396-213-03grantedtoW.R.J.Thisworkwasalsosupported bytheWellcomeTrust(grantnumbersWT095908andWT098051)andtheEuropean MolecularBiologyLaboratory. Shisa2 qPCRworkbene“tedfromthefacilitiesand expertiseoftheQPCRplatformofIMAGIF(CentredeRecherchedeGif-www. imagif.cnrs.fr).ThisworkwassupportedinpartbytheNationalInstitutesofHealth (NIDCR)grantDE022403toJ.B.G.WethankJ.Tabinfortechnicalassistance.Weare gratefulforresourcesfromtheUniversityofMinnesotaSupercomputingInstitute.AuthorcontributionsS.E.M.andW.C.W.aretheprincipalinvestigatorswhoconceivedtheproject,analysed thedataandwrotethemanuscript.W.C.W.andP.M.sequencedandassembledthe genome.J.B.G.andB.A.S.providedRNAseqdata,identi“edcandidategenesandaidedin writingthemanuscript.C.T.andN.R.validatedgenelosscandidates,identi“edcandidate genesandaidedinwritingthemanuscript.D.C.andJ.-N.V.carriedouttransposable elementanalyses.W.R.J.,K.E.O.andM.Y.providedtissueforDNAandRNAsequencing andaidedinwritingthemanuscript.W.R.J.andL.M.carriedoutthe otx2 validation work.A.K.investigatedcandidategenesandA.K.andR.B.aidedinwritingthe manuscript.S.R.,H.H.andM.B.carriedoutthe shisa2 validationworkandaidedin writingthemanuscript.S.M.J.S.ledtheEnsemblgeneannotationandB.A.andD.M. performedthegenomeannotation.AdditionalinformationAccessioncodes: AllgenomicdataareassociatedwithbioprojectPRJNA89115andhave beendepositedinGenBank/EMBL/DDBJNucleotidedatabaseundertheaccessioncode APWO00000000.RNAseqdatahavebeendepositedintheGenBank/EMBL/DDBJ sequencereadarchiveundertheaccessioncodesPRJNA177689(tissue-speci“ctranscriptomes)andPRJNA258661(developmentaltimecourse).Geneannotationscanbe foundathttp://www.ensembl.org/Astyanax_mexicanus/Info/Indexandhavebeen depositedintheGenBank/EMBL/DDBJAssemblydatabaseundertheaccessioncode PRJNA237016. SupplementaryInformation accompaniesthispaperathttp://www.nature.com/ naturecommunications Competing“nancialinterests: Theauthorsdeclarenocompeting“nancialinterests. Reprintsandpermission informationisavailableonlineathttp://npg.nature.com/ reprintsandpermissions/ Howtocitethisarticle :McGaughS.E., etal. Thecave“shgenomerevealscandidate genesforeyeloss. Nat.Commun. 5:5307doi:10.1038/ncomms6307(2014). ThisworkislicensedunderaCreativeCommonsAttributionNonCommercial-ShareAlike4.0InternationalLicense.Theimagesor otherthirdpartymaterialinthisarticleareincludedinthearticlesCreativeCommons license,unlessindicatedotherwiseinthecreditline;ifthematerialisnotincludedunder theCreativeCommonslicense,userswillneedtoobtainpermissionfromthelicense holdertoreproducethematerial.Toviewacopyofthislicense,visithttp:// creativecommons.org/licenses/by-nc-sa/4.0/ ARTICLENATURECOMMUNICATIONS|DOI:10.1038/ncomms630710NATURECOMMUNICATIONS |5:5307|DOI:10.1038/ncomms6307|www.nature.com/naturecommunications& 2014 MacmillanPublishersLimited.Allrightsreserved.