Evidence for late Pleistocene origin of Astyanax mexicanus cavefish


previous item | next item

Citation
Evidence for late Pleistocene origin of Astyanax mexicanus cavefish

Material Information

Title:
Evidence for late Pleistocene origin of Astyanax mexicanus cavefish
Series Title:
BMC Evolutionary Biology
Creator:
Fumey, Julien
Hinaux, Hélène
Noirot, Céline
Thermes, Claude
Rétaux, Sylvie
Casane, Didier
Publication Date:
Language:
English

Subjects

Subjects / Keywords:
Cavefish ( local )
Adaptation ( local )
High-Throughput Sequencing ( local )
Microsatellites ( local )
Snps ( local )
Molecular Dating ( local )
Genre:
serial ( sobekcm )

Notes

Abstract:
Cavefish populations belonging to the Mexican tetra species Astyanax mexicanus are outstanding models to study the tempo and mode of adaptation to a radical environmental change. They are currently assigned to two main groups, the so-called “old” and “new” lineages, which would have populated several caves independently and at different times. However, we do not have yet accurate estimations of the time frames of evolution of these populations.We reanalyzed the geographic distribution of mitochondrial and nuclear DNA polymorphisms and we found that these data do not support the existence of two cavefish lineages. Using IMa2, a program that allows dating population divergence in addition to demographic parameters, we found that microsatellite polymorphism strongly supports a very recent origin of cave populations (< 20,000 years). We identified a large number of single-nucleotide polymorphisms (SNPs) in transcript sequences of pools of embryos (Pool-seq) belonging to Pachón cave population and a surface population from Texas. Based on summary statistics that can be computed with this SNP data set together with simulations of evolution of SNP polymorphisms in two recently isolated populations, we looked for sets of demographic parameters that allow the computation of summary statistics with simulated populations that are similar to the ones with the sampled populations. In most simulations for which we could find a good fit between the summary statistics of observed and simulated data, the best fit occurred when the divergence between simulated populations was less than 30,000 years.Although it is often assumed that some cave populations have a very ancient origin, a recent origin of these populations is strongly supported by our analyses of independent sets of nuclear DNA polymorphism. Moreover, the observation of two divergent haplogroups of mitochondrial and nuclear genes with different geographic distributions support a recent admixture of two divergent surface populations, before the isolation of cave populati
Original Version:
BMC Evolutionary Biology, Vol. 18, no. 43 (2018-04-18).

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
This item is licensed with the Creative Commons Attribution License. This license lets others distribute, remix, tweak, and build upon this work, even commercially, as long as they credit the author for the original creation.
Resource Identifier:
K26-05078 ( USFLDC: LOCAL DOI )
k26.5078 ( USFLDC: LOCAL Handle )

USFLDC Membership

Aggregations:
University of South Florida
Karst Information Portal

Postcard Information

Format:
serial

Downloads

This item is only available as the following downloads:


Full Text

PAGE 1

RESEARCHARTICLEOpenAccess EvidenceforlatePleistoceneoriginofAstyanaxmexicanuscavefishJulienFumey1,2,HélèneHinaux3,CélineNoirot4,ClaudeThermes2,SylvieRétaux3andDidierCasane1,5*AbstractBackground:CavefishpopulationsbelongingtotheMexicantetraspeciesAstyanaxmexicanusareoutstandingmodelstostudythetempoandmodeofadaptationtoaradicalenvironmentalchange.Theyarecurrentlyassignedtotwomaingroups,theso-called“old”and“new”lineages,whichwouldhavepopulatedseveralcavesindependentlyandatdifferenttimes.However,wedonothaveyetaccurateestimationsofthetimeframesofevolutionofthesepopulations.Results:WereanalyzedthegeographicdistributionofmitochondrialandnuclearDNApolymorphismsandwefoundthatthesedatadonotsupporttheexistenceoftwocavefishlineages.UsingIMa2,aprogramthatallowsdatingpopulationdivergenceinadditiontodemographicparameters,wefoundthatmicrosatellitepolymorphismstronglysupportsaveryrecentoriginofcavepopulations(<20,000years).Weidentifiedalargenumberofsingle-nucleotidepolymorphisms(SNPs)intranscriptsequencesofpoolsofembryos(Pool-seq)belongingtoPachóncavepopulationandasurfacepopulationfromTexas.BasedonsummarystatisticsthatcanbecomputedwiththisSNPdatasettogetherwithsimulationsofevolutionofSNPpolymorphismsintworecentlyisolatedpopulations,welookedforsetsofdemographicparametersthatallowthecomputationofsummarystatisticswithsimulatedpopulationsthataresimilartotheoneswiththesampledpopulations.Inmostsimulationsforwhichwecouldfindagoodfitbetweenthesummarystatisticsofobservedandsimulateddata,thebestfitoccurredwhenthedivergencebetweensimulatedpopulationswaslessthan30,000years.Conclusions:Althoughitisoftenassumedthatsomecavepopulationshaveaveryancientorigin,arecentoriginofthesepopulationsisstronglysupportedbyouranalysesofindependentsetsofnuclearDNApolymorphism.Moreover,theobservationoftwodivergenthaplogroupsofmitochondrialandnucleargeneswithdifferentgeographicdistributionssupportarecentadmixtureoftwodivergentsurfacepopulations,beforetheisolationofcavepopulations.Ifcavepopulationsareindeedonlyseveralthousandyearsold,manyphenotypicchangesobservedincavefishwouldthushavemainlyinvolvedthefixationofgeneticvariantspresentinsurfacefishpopulationsandwithinaveryshortperiodoftime.Keywords:Cavefish,Adaptation,High-throughputsequencing,Microsatellites,SNPs,MoleculardatingBackgroundTwowell-differentiatedmorphotypes,surfacefishandcavefish,arefoundinthespeciesAstyanaxmexicanus.Twenty-ninecavefishpopulationshavebeendiscoveredsofarinlimestonecavesintheSierradeElAbraregionofnortheasternMexico[1,2](Fig.1).Cavefishdifferfromtheirsurfacecounterpartsinnumerousmorphological,physiologicalandbehavioraltraits,themoststrikingbeingthatmostcavefishlackfunctionaleyesandaredepigmen-ted[3].Mostcavesinhabitedbycavefishshareanumberofabioticandbioticcharacteristicssuchasconstantdark-nessandabsenceofpredators,andmostcavefishshowevolutionofanumberofcharacters[4],eitherbecausetheyaredispensable-regressivetraits-suchaslossofeyesandpigmentation[5],orbecausetheyareinvolvedintheadaptation-constructivetraits-tothisenvironmentwhichisinhospitableformostfishes.Forexample,cave-fishhavealowermetabolicrate[6–8],producelargereggs[9],havemoreandlargersuperficialneuromastsinvolvedinvibrationattractionbehavior[10–12],sleepverylittle[13,14],haveshiftedfromfightingtoforagingbehavior *Correspondence:Didier.Casane@egce.cnrs-gif.fr1Évolution,Génomes,Comportement,Écologie,CNRS,IRD,UnivParis-Sud.UniversitéParis-Saclay,F-91198Gif-sur-Yvette,France5UniversitéParisDiderot,SorbonneParisCité,Paris,FranceFulllistofauthorinformationisavailableattheendofthearticle ©TheAuthor(s).2018OpenAccessThisarticleisdistributedunderthetermsoftheCreativeCommonsAttribution4.0InternationalLicense(http://creativecommons.org/licenses/by/4.0/),whichpermitsunrestricteduse,distribution,andreproductioninanymedium,providedyougiveappropriatecredittotheoriginalauthor(s)andthesource,providealinktotheCreativeCommonslicense,andindicateifchangesweremade.TheCreativeCommonsPublicDomainDedicationwaiver(http://creativecommons.org/publicdomain/zero/1.0/)appliestothedatamadeavailableinthisarticle,unlessotherwisestated.Fumeyetal.BMCEvolutionaryBiology (2018) 18:43 https://doi.org/10.1186/s12862-018-1156-7

PAGE 2

[15],havelargernumbersoftastebuds[16,17],haveenhancedchemosensorycapabilities[18]andhaveen-hancedpreycaptureskillatboththelarvalandadultstages[11,19,20].Verysignificantadvanceshavebeenmadeinidentify-ingproximalmechanisms[21],thatismutationsthathavechangedphysiological,developmental,andbehaviortraitsofcavefishandnewmoleculartoolsavailabletodaywillallowustoidentifysuchmutationsataneverin-creasingpace[22–26].Howeveritismuchmoretrickytodisentangledistalmechanisms[21],i.e.evolutionarymechanisms.Werethesemutationsalreadypresentatlowfrequencyinsurfacefishstandingvariationordidtheyappearaftersettlements?Arepleiotropiceffectsandepistaticinteractionsimportantintheseevolution-aryprocesses?Whatistheimpactofrecombination,geneticdrift,selectionandmigrationincavefishevolu-tion?Thesequestionshavefueleddiscussionsontherelativeimportanceofthesedifferentevolutionarymech-anisms[12,17,27–31].Inordertoanalyzeseveraloftheseissuessuchastherelativeweightofselection,migrationandgeneticdrift,itwouldbeveryusefultohaveaccurateestimationsofsomedemographicandpopulationgeneticparameterstodescribethedynamicofcavefishevolution.Geneflowfromthesurfacepopulationshasbeenestimatedtobefromverylow,ifany,toveryhigh,dependingonthecavepopulationexamined.Somestudieshavealsofoundsignificantandhighergeneflowfromcavetosurfacepopulationsthanintheoppositedirection[32–37].Moreover,assomecavesareveryclosetoeachother,fishmigrationswithincaveclustersarelikely.Amongmodelparametersparticularlyimportanttodescribetheevolutionofacavefishpopulationare:1)thetimeatwhichsettlementoccurredand2)howlongittookforsurfacefishtoadapttothecaveenvironment.Asshownbelow,noreliableageswereavailablebutAs-tyanaxmexicanuscavepopulationshaveneverthelessbeenassignedtotwogroups,theso-called“old”and“new”lineages,whichwouldhavepopulatedseveralcavesindependentlyandatdifferenttimes[37–39],reviewedin[2].However,theageofcavefishsettlementhasbeenesti-matedfortwopopulationsonly,thoseinhabitingthePachónandLosSabinoscaves,whichbothbelongtothe“old”lineage.Onthebasisofallozymepolymorphism[32]andapopulationgeneticmethodspecificallydesignedtoestimatethetimeafterdivergencebetweenincompletelyisolatedpopulationsofunequalsizes(suchascaveandsurfacepopulations),thesepopulationswereestimatedtobe710,000and525,000yearsold,respectively,suggestingthattheycouldbeancient[40].However,thesmallnum-beroflocistudiedatthattime(17allozymelociscored),theabsenceofpolymorphisminPachónandverylowpolymorphisminLosSabinosdidnotallowaccurate Fig.1Mapsshowingcaveandsurfacesamplingsites.aSitesinMexicoandTexas.bSitesintheSierradeElAbraregioninMexico.Surfacefish:S1toS4.Cavepopulations:O1=Pachón,O2=Yerbaniz,O3=Japonés,O4=Arroyo,O5=Tinaja;O6=Curva,O7=Toro,O8=Chica,N1=Molino,N2=CaballoMoro,N3=Subterráneo(See[33]foramoredetaileddescriptionofthesesamplingsites),Sa=Sabinos,Pi=Piedras.Outercircleinred:“G”mtDNA.MapadaptedfromFig.5in[1]Fumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page2of19

PAGE 3

estimations.Thestandarderror(SE)wasverylarge,460,000and330,000years,respectively.Assuminganor-maldistribution[41],the95%confidenceintervalis±1.96xSE.Itimpliesthatthesepopulationscouldbeeitherveryrecent(acoupleofthousandyearsandevenless)orveryancient(about1.5millionyears).Basedonthisanalysis,theonlysafeconclusionisthatthesecavepopulationsarenotmillionsofyearsold.Thelargeuncertaintyassociatedtotheseestimationsisprobablythereasonwhytheyarerarelycitedbyinvestigatorsworkingonthesecavefish.Thehypothesisofaveryancientoriginofthe“old”cavefishlineage,i.e.millionsofyearsago,reliesonlyondiscussionsofmitochondrialDNA(mtDNA)phylogeniesofsurfacefishandcavefish,showingtwohighlydivergentmitochondrialhaplogroups[37,39,42].However,ancientcoalescenceofmtDNAhaplogroupsdoesnotnecessarilyimplyanancientisolationofsomecavepopulations,i.e.thetimeofseparationofthepopulationsisnotnecessarilyequal,notevenclose,tothetimeofcoalescenceofthemtDNAsequences.Analternativehypothesisthatwouldleadtothesameobservationisarecentadmixtureoftwodivergentsurfacepopulationsfollowedbyoneorseveralfishsettlementsincaves.MorerecentlyaphylogeneticanalysiswasperformedusingalargeSNPdatasetinordertoestimatethenumberofin-dependentcavesettlementsattheoriginoffivecavefishpop-ulationsfromthreedistinctregions[43].Thisanalysisdidnotsupportthetwolineageshypothesisbutindicatedatleastfourindependentoriginsforthesecavefishpopulations.Nevertheless,assuminganancientoriginofcavepopulationsandthusthatsurfaceandcavepopula-tionsareatmutation/migration/driftequilibrium,esti-mationofdifferentiation[32–34,38]andmigrationratesamongpopulations[33,35]wereperformedusingmicrosatellitepolymorphism.Insummary,andexceptanattemptusinganallozymesdatasetunfortunatelytoosmalltogiveaccurateestima-tions,nodatinghaseverbeenperformedusingnuclearmarkerstodirectlytesttheassumptionthatsomecave-fishpopulationsaremillionsofyearsold.Severalobservationsledustodoubtaboutitsaccur-acy.Inparticular,lookingatsequencesofPachóncave-fish,wedidnotfindmanyobviousloss-of-functionmutations,suchasframeshiftsandstopcodons,ineye-specificcrystallingenes[26]andopsingenes[44–46](unpublishedresults),anunexpectedobservationifthispopulationwasestablishedforseveralhundredthousandyears,andwhichbecomesveryunlikelyifitwasestab-lishedmorethanonemillionyearsago[47].Indeed,otherfishthatcouldbeconfinedintocavesformillionsofyearshavefixedloss-of-functionmutationsinseveralopsinsandcrystallinsgenes[48–50].Here,weanalyzedpublishedsequencedatasetsandwefoundthatdifferentnuclearlocihavedifferentphylogeniesthatarenotcongruentwiththemtDNAphylogeny.More-over,usingapublishedmicrosatellitedatasetandanap-proachthatallowdatingtheisolationofcloselyrelatedpopulationswhenthereisgeneflow,weobtainedgoodevidenceofarecentoriginofallcavefishpopulationsana-lyzed,e.g.likelylessthan20,000yearsago,notwithstand-ingtheir“old”or“young”classification.Intheseanalyses,estimationsofeffectivepopulationsizesandmigrationratesweremorecoherentwithexpectationsthaninprevi-ousanalyses:effectivepopulationsizesforcavefishwereatleastoneorderofmagnitudesmallerthanthoseforsur-facefish;geneflowswerefromthesurfacetocavesandnottheotherwayaround.Inordertocorroboratethesenovelestimationswithanindependentdatasetandusingaverydifferentmethod,weidentifiedandanalyzedalargenumberofsingle-nucleotidepolymorphisms(SNPs)intranscriptsequencesoftwopoolsofembryos(Pool-seq)belongingtothePachóncavepopulationandasurfacepopulationfromTexas.Thecomparisonofthesedatawithsimula-tionssuggeststhatthePachóncavepopulationhasprob-ablybeenundergroundlessthan30,000years.BothdatingmethodsgavecongruentestimationsoftheageofthePachóncavepopulation,andtheypointedtoaveryrecentorigin.ThenewtimeframeweproposefortheevolutionofA.mexicanuscavefishsuggeststhatthemanypheno-typicchangesobservedinthesecavefishmayhavemainlyinvolvedthefixationofgeneticvariantspresentinsurfacefishpopulations,andwithinaveryshortperiodoftime.ResultsNuclearhaplotypephylogeniesTherearecurrentlyfewnucleargenesforwhichse-quencesfromdifferentcaveandsurfacefishhavebeenpublished.AfragmentofthecodingsequenceofRag1isavailableforalargesampleofsurfaceAstyanaxspp.butonlyonecavefish[42].Aphylogenybasedonthisgenealonehasnotbeenpublished.Weperformedthisphylo-geneticanalysis(Additionalfile1:FigureS1).Thisphyl-ogenyissoverypoorlyresolvedthatitcannotsupportthemtDNAphylogeny(Additionalfile1:FigureS2)noranyotherphylogeny.Thenweanalyzedaseriesofgenes(Mc1r,Oca2,Mc4r,Mc3r,Lepb,Lepr,Pomcb)forwhichatleastfoursequenceswereavailable,threefromcave-fishandonefromasurfacefish(Additionalfile2).ForMc4r,Mc3r,Lepb,Lepr,Pomcb,sequenceswereob-tainedforthesamelocalities(Surface,Pachón,TinajaandMolino)[51](Fig.1).Thematrixofinformativesitesineachgeneisshown(Fig.2a).Whenmorethanonein-formativesitewasfoundwithinagene,allinformativesitessupportedthesametreetopologybutdifferentgenessupporteddifferenttreetopologies.OveralltheFumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page3of19

PAGE 4

threepossibleunrootedtreesweresimilarlysupported(Fig.2b,Additionalfile2).Partialorcompletecodingsequencesoffiveothergenes(Per1,Per2,Tef1,Cry1aandCpdphotolyase)fromthreelocalities(PachónandChicacavesandasurfacelocalityclosetoMicos)werealsoavailable[52].ThesesequenceswerealignedwiththesequenceoftheTexassurfacefishandthemostparsimoniousunrootedtreewasreconstructedforeachgene.Despitethefactthattwodivergenthaplotypeswerefoundforsomegenes,thetwosurfacefishsequenceswerealwaysveryclose(Additionalfile2).Insum,theincongruenceofthephy-logeniesofthesegenessuggeststhattheyhaveinde-pendentevolutionaryhistories.DatingwithmicrosatellitesWenextre-analyzedapreviouslypublisheddatasetofmicrosatellitepolymorphisms[33]usingIMa2[53].Thisprogram,whichimplementsamethodbasedonsimula-tionsofcoalescenceofsamplesofalleles,allowstheesti-mationofthemarginalposteriorprobabilitydensityofpopulationsizes,migrationratesanddivergencetimes.Weperformedaseriesofpairwiseanalysesinvolvingacavepopulationandasurfacepopulation.Wealsoana-lyzedthedivergenceofthreecavepopulations.Theesti-matedmarginalposteriorprobabilitydensitiesofthemodelparametersobtainedwithPachón(O1)andasur-facepopulation(S3),using22lociandarandomsampleof60allelesperlocusperpopulationareshown(Fig.3).SimilardistributionswereobtainedusingthefirsthalfandthesecondhalfoftheMCMCchain,suggestingthatthesamplingprocesshasbeenlongenoughtogetstableposteriorprobabilitydistributions.Moreover,thesepos-teriordistributionshaveasinglesharppeakandtheprobabilitiesarelowfortheextremesvaluesofthepriordistributions,suggestingthattheselectionofthemax-imumvalueofeachparameterwassuitable.InIMa2,demographicparametersarescaledtothemutationrate.Estimationsoftheseparametersthusdependsonprioronthemutationrate.Assumingthatthemutationrateofthemicrosatellitesis5×104[33,54,55],estimationsofpopulationsizes,migrationratesanddivergencetimecanbeobtained(Table1).Theeffectivepopulationsize(Ne)was150[150–1150]and10,750[6650–16,350]forPachóncavepopulationandsurfacefishpopulationrespectively(themaximumlikelihoodvalueisgivenwiththe95%highestposteriordensityintervalbetweenbrackets).Theancestraleffectivepopulationsizewas44,850[32,350–73,250].Assumingagenerationtimeequalsto3.5years,whichisbasedonarecentestimationofthedistributionoftheagesoffishcapturedinthewild[56],thedivergencetimewas5110years[1302–18,214].Themigrationrateswereverylow,butabout100timeshigherfromthesurfacethanfromthecave.Withthesamesamplesize(i.e.numberoflociandnumberofalleles/locus/population),thesameana-lyseswereperformedwithChicacave(O8)andthesamesurfacepopulation(S3)(Additionalfile1:FigureS3)andwithPachón(O1)andChica(O8)caves(Additionalfile1:FigureS4).Suchalargenumberofallelesforalargenumberoflociwasnotavailableforotherusefulana-lyses(Additionalfile1:TableS1),wethususedsmallersamples.Withasampleof14lociand40allelesperlocusperpopulation,thesameanalysiswasperformedwithMolinocave(N1)andasurfacepopulation(S1)(Additionalfile1:FigureS5).Withasampleof18lociand40allelesperlocusperpopulation,thesameana-lysiswasperformedwithCaballoMorocave(N2)andasurfacepopulation(S2)(Additionalfile1:FigureS6).Withasampleof21lociand40allelesperlocusperpopulation,thesameanalysiswasperformedwithSubterráneocave(N3)andasurfacepopulation(S4)(Additionalfile1:FigureS7).Withasampleof14lociand20allelesperlocusperpopulation,thesameanalysiswasperformedwithCurvacave(O6)andasurfacepopulation(S3)(Additionalfile1:FigureS8),andwithCurvacave(O6)andPachóncave(O1)(Additionalfile1:FigureS9).Theresultsoftheseanalysesarehighlyconsistent(Table2).Caveeffectivepopulationsize(Ne)wasalwayslow,afewhundreds.ExtantsurfacepopulationNewasconsist-entlyfoundtobeabout10,000whateverthesamplinglocality.TheancestralsurfacepopulationNewascon-sistentlyfoundlargerthantheextantsurfacepopula-tionNe,i.e.between2and4timeslarger.Themigrationrateswerealwayslow.Themaximumlikeli-hoodofthedivergencetimewassimilarinallcases,intherangeof1000to10,000years,theposteriordistri-butionslargelyoverlapping.Theseresultssuggestthatallthecavepopulationsana-lyzedareveryrecent,regardlessoftheirclassificationasancient,recent,isolatedormixed,andregardlessoftheirmtDNAhaplotype. ab Fig.2Incongruenceoffourtaxonnucleargenephylogenies.aInformativesitesinMc4r,Mc3r,LeprandPomcb.bUnrootedtreeswithfourtaxa.InformativesitesequallysupportthethreepossibletreesFumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page4of19

PAGE 5

SNPsandsubstitutionratesinsurfaceandcavepopulationsNext,wesetouttocorroboratethesepreviousfindingswithanindependentmethod,throughtheanalysesofSNPsfoundinsurfacefish(SF)andPachóncavefish(CF)transcriptomes,coupledtosimulations.Weusedtranscriptomesequencedatasetsfrompooledembryos(Additionalfile1:FigureS10).Wedefinedeightclassesofpolymorphicsitesaccordingtothepresenceofanan-cestraland/oraderivedalleleinSFandCFpopulations,usingtheBuenosAirestetra(Hyphessobryconanisitsi)asanoutgroup(Fig.4).WeestimatedthefrequenciesoftheseeightSNPclassesatsynonymous,non-codingandnon-synonymoussites(Table3).ThefrequenciesofSNPsintheeightclasseswererobustaccordingtodifferentparameterthresholdsusedtoincludeSNPsintheanalysis(Methods,Additionalfile1:FigureS11,TableS2aandTableS2b).Theratio(SF/CF)ofsynonymous,non-codingandnon-synonymouspolymorphismwas3.08,2.71and2.34,re-spectively,andtheratio(CF/SF)ofderivedfixedal-leleswas2.34,1.45and1.52,respectively.TheseresultsindicatedthatthelevelofpolymorphismwashigherintheSFpopulation,butthenumberoffixedderivedalleleswashigherintheCFpopulation. Table1EstimateddemographicparameterswithIMa2HiPtHPD95%q0(4NeuPachón)0.3000.300–2.300q1(4NeuSurface)21.50013.300–32.700q2(4NeuAncestral)89.70064.700–146.500t0(tu)0.7300.186–2.602m0>1(m0->1/u)1.7020.948–4.232m1>0(m1->0/u)0.0130.000–0.308NePachón150150–1150NeSurface10,7506650–16,350Neancestral44,85032,350–73,250t51101302–18,214mS->P8.5×1044.7–21.2×104mP->S6.5×1060.0–1.5×104Weassumedamutationrate(u)of0.0005andameangenerationtime(g)of3.5years(2yearsforsurfacefishand5yearsforcavefish).HiPt:thehighestvalueinthehistogramoftheposteriormarginaldensity.HPD95%:thelowerandupperboundoftheestimated95%highestposteriordensity(HPD)interval.Effectivepopulationsize(Ne)Pachón=q0/4u;Nesurface=q1/4u;Neancestral=q2/4u;t(divergencetimeinyears)=t0xg/u;mS->P(migrationratefromsurfacetoPachón)=m0>1xu;mP->S(migrationratefromPachóntosurface)=m1>0xu.m0>1andm1>0aremigrationratebackwardintime,whereasmS->PandmP->Saremigrationrateforwardintime Fig.3PosteriormarginaldensityplotsofthedemographicparametersestimatedfromtheisolationwithmigrationmodelofPachóncavefishandS3_surfacefish.q0=4N0(whereN0istheeffectivesizeofpopulation0[i.e.Pachóncavefish]andisthemutationrate),q1=4N1(whereN1istheeffectivesizeofpopulation1[i.e.S3_Surfacefish]),q2=4N2(whereN2istheeffectivesizeofpopulation2[i.e.ancestralpopulation]),t0=t/g(wheretisthenumberofgenerationsincedivergenceandgisthegenerationtime),m0>1(-scaledmigrationratefrompopulation0topopulation1backwardintime=-scaledmigrationratefrompopulation1topopulation0forwardintime),m1>0(-scaledmigrationratefrompopulation1topopulation0backwardintime=-scaledmigrationratefrompopulation0topopulation1forwardintime).Commandline:./IMa2–iinfileO8S3_n60.u-ooutput_O1S3_n60_longrun.out-b200000-l100000-d100–m10-q200-t4-c1-p23567-r25-u1-hfg-hn50-ha0.999-hb0.3Fumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page5of19

PAGE 6

DatingwithSNPsInordertomakeanestimationoftheageofthePachóncavepopulationwiththeSNPsthatisindependentoftheestimationmadewithmicrosatellitepolymorphism,wecomparedtheobservedsummarystatisticsofsyn-onymouspolymorphismwiththesummarystatisticsofneutralpolymorphisminsimulatedpopulationsusingthemodelofevolutionimplementedinIMa2,i.e.anan-cestralpopulationdividedintotwopopulationsatsomepointinthepast.Eachpopulationhaditsowneffectivepopulationsize(Fig.4).Migrationsbetweenpopulationswereallowed.WhereasIMa2isbasedonsimulationofcoalescenceofasampleofalleles(backwardintime),wedidsimulationsforwardintimeofgeneticdriftofallelefrequenciesinwholepopulations,whichwasmoreadaptedtoanalyzeourSNPdataset.Bothapproacheshavetheirownadvantagesandlimits[57].Simulationsofgeneticdriftintwopopulationsrecentlyandpartiallyisolatedallowedtoestimatesummarystatisticsthatcouldbeeasilycomparedwiththeobservedsummarystatistics.Inaddition,theimplementationofourownsimulationprogramallowedtotakeintoaccountachangeofgener-ationtimeincavefishandtocomputemoreeasilytheevo-lutionofsummarystatisticsthroughtime.Wetestedwiderangesofparametersvalues(populationsizes,migrationrates,isolationtime)thatweredefinedtakingintoaccounttheresultsobtainedwithIMa2,previ-ouspopulationsgeneticsstudiesandinformationgatheredthroughseveraltripstothePachóncave.Afewsetofparametersallowedagoodfitbetweenthesummarystatisticsoftheobservedandsimulatedpoly-morphism(Additionalfile1:TableS3a,TableS3b,TableS3candTableS3d).Asanexample,wecanconsideracase,i.e.aparameterset,whichgaveagoodfit(inAdditionalfile1:TableS3a,framedinyellow).Inthissimulationtheancestralpopulationsizewassetto10,000andwasatmutation/driftequilibrium;aftertheseparationofthesurfaceandcavepopulations,thecavepopulationsizewassetto1250andtheTexassurfacepopulationsizewassetto10,000;theprobabilityofmigra-tionperyearfromsurfacetocavewas0.001andthenum-berofmigrantswas1%ofthecavepopulationsize(i.e.12fish);thegenerationtimeofthecavefishwassetto5years Table2EstimateddemographicparameterswithIMa2Populationsp0–p1O1–S3p0–p1O6–S3p0–p1O8–S3p0–p1O1–O6p0–p1O1–O8p0–p1N1–S1p0–p1N2–S2p0–p1N3–S4Nep0HiPt15050250250250150850450HPD95%[150–1150][0–950][150–1250][150–1250][150–1150][0–950][350–1850][250–1650]Nep1HiPt10,7506950585035055010,55038506950HPD95%[6650–16,350][2150–83,150][1950–60,350][150–1350][150–1550][6050–15,950][2150–6950][1550–14,350]NeancestralHiPt44,85031,25042,45024,05026,65047,05037,95043,350HPD95%[32,350–73,250][19,850–49,650][28,750–62,650][11,950–50,750][14,150–44,150][28,950–81,550][25,250–65,350][20,350–69,150]tHiPt511036827982540127410,1504074770HPD95%[1302–18,214][350–13,874][30–8302][860–5980][266–2422][2002–24,990][1834–10,626][406–17,234]m1->0HiPt8.5×10412.8×10411.2×1048.6×1043.4×1044.9×1043.0×10429.0×104HPD95%[4.4–21.2×104][4.7–89.0×104][6.4–50.0×104][0.5–24.5×104][0.0–8.7×104][1.0–15.2×104][0.0–8.3×104][9.0–57.2×104]m0->1HiPt6.5×1041.0×10–47.93×1042.5×1064.3×1045.0×1061.5×10610.2×104HPD95%[0.0–1.5×104][0.0–11.2×104][1.3–10.7×104][0.0–8.5×104][0.0–10.1×104][0.0–1.3×104][0.0–1.8×104][3.8–17.2×104]SeeTable1fordefinitionofparametersandFig.1forlocalizationofpopulations ab Fig.4AnalysisofpolymorphisminAstyanaxmexicanusTexassurfacevsPachóncavepopulation,usingHyphessobryconanisitsiasoutgroup.aEvolutionarymodel.bTheeightSNPclassescorrespondtothepolymorphismpatternsthatcanbefoundwithinandbetweentwopopulations.Class1:Differentfixedallelesineachpopulation,derivedalleleincavefish;Class2:Differentfixedallelesineachpopulation,derivedalleleinsurfacefish;Class3:Polymorphismincavefish,ancestralfixedalleleinsurfacefish;Class4:Polymorphismincavefish,derivedfixedalleleinsurfacefish;Class5:Polymorphisminsurfacefish,ancestralfixedalleleincavefish;Class6:Polymorphisminsurfacefish,derivedfixedalleleincavefish;Class7:Sharedpolymorphism;Class8:Divergentpolymorphism.x,yandzcanbeoneofthefournucleotidesA,T,G,CFumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page6of19

PAGE 7

andthegenerationtimeofthesurfacefishwassetto2years.Every100years(i.e.50SFgenerations,or20CFgenerations),10fishweresampledineachpopulationtosimulatethesamplingprocesswhenthelabpopulationswereestablished.Eachlabpopulationwasthensetwithaconstanteffectivepopulationsizeof10over10genera-tions.ThenwecomparedthefrequencyofeachSNPclassinthesimulatedlabpopulationswiththeobservedfre-quency.Itwasthuspossibletocheckthefitofthesum-marystatisticsthroughtimeusingagoodnessoffitscore.Inthissimulation,thebestfit(thelowestvalueofthescore)occurredwhentheageofthecavepopulationwas21,500years(Fig.5a).AllSNPclassfrequenciesinthesimulatedpopulationsfitwell(goodnessoffitscore=0.68)withtheobservedfrequencies(Fig.5b).Then,theolderwasthedivergenceofthepopulationsandtheworsewasthefit(Fig.5a).Inthissimulation,aswellasinallothersimulations,themutationratepergeneration(u),thatistheprob-abilityofappearanceofanewalleleatanewlocusinonehaploidgenomeatagivengeneration,wassetto2.102.ThenumberofnewSNPsthatappearedpergenerationinapopulationofsizeNwas2Nu,eachwithafrequencyof1/2N.Thismeansthatinthesurfacepopulationtherewas2×10,000×2×102=400newSNPsateachgeneration,andthatthese400newSNPsappearedwithaninitialfrequencyof1/(2×10,000)=5×105.Inparallel,50newSNPsappearedwithinitialfrequencyof4×104inthecavepopulationateachgeneration.Alllociwereindependent.ItisnoteworthythatthefitoftheactualandsimulatedpolymorphismdidnotdependonthemutationratebecausewecomparedtherelativefrequenciesofSNPclassesratherthantheirabsolutenumbers.Inotherwords,ifthemutationratewashigher,thenumberofSNPsineachclasswashigher,buttherelativefrequencyofeachclassremainedthesame.Thusthescoreofgoodnessoffitdidnotdependonthemutationrate.Themutationrateweusedwasatrade-offbetweentheaccuracyoftheSNPclassfrequencyestimationsinthesimulatedpopulationsandthetimetorunasimulation(thehigherthemutationrate,thehigherthenumberofpolymorphicsitesforwhichallelefrequencyevolutionwassimulated).Finally,theestimationoftheageofthecavepopulationdependsonthegenerationtimeineachpopulationasthisage,inyears,istheproductofthenumberofgenerationmultipliedbythegenerationtime.AstheanalyseswithIMa2suggestedthattheancestralpopulationsizewaslargerthantheextantsurfacepopu-lationsize,wetestedancestralpopulationintherangeof10,000to100,000.Surfacepopulationsizewassetintherangeof5000to20,000.Cavefishpopulationsizewassetintherangeof75to10,000.Wealsotookintoaccountmigrationfromthesurfacetothecave:theprobabilityofmigrationvariedbetween0.1and0.0001peryearandthepercentageofsurfacefishthatmigratedintothecavevariedbetween1%and0.01%ofthecave-fishpopulationsize.WeconsideredthatthemigrationrateandthenumberofmigrantsateachmigrationfromthecavetosurfacewasnegligibleasitissuggestedbytheanalysiswithIMa2andourobservationsinthefield.Insummary,goodfitscouldbefoundbetweenob-servedandsimulatedsummarystatisticswhenthe Table3ClassificationofpolymorphismsinAstyanaxmexicanusTexassurfacevsPachóncavepopulationsSynonymousNon-codingNon-synonymousClassn%n%n%AncestralfixedSF,derivedfixedCF(1)54012.815712.730114.3AncestralfixedCF,derivedfixedSF(2)2806.71119.021010.0PolymorphismCF,ancestralfixedSF(3)47611.314611.830214.5PolymorphismCF,derivedfixedSF(4)1192.8574.6914.4PolymorphismSF,ancestralfixedCF(5)208649.660148.592343.6PolymorphismSF,derivedfixedCF(6)3939.3877.01597.8Sharedpolymorphism(7)3097.4806.51235.9Divergent(8)10.000.000.0Total4204100.01239100.02110100.0PolymorphismSF(5+6+7)27887681205PolymorphismCF(3+4+7)904283516RatioSF/CF3.082.712.34DerivedandfixedSF(2+4)399168302DerivedandfixedCF(1+6)933244460RatioCF/SF2.341.451.52Thresholds:100;MAF>5%;ScoreBlast<10–5;interval>50bp(seematerialsandmethodsforthresholddefinitions).CF:Cavefish;SF:Surfacefish;numbersinbracketsareclassidentifiersdescribedinFig.4Fumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page7of19

PAGE 8

effectivepopulationsizeofthecavepopulationwasmuchsmallerthanthesurfaceeffectivepopulationsizeandwhenthedivergencewasrecent,inmostcasesaround20,000years(Additionalfile1:TableS3a,TableS3b,TableS3candTableS3d).DiscussionEvidenceagainsttheexistenceofoldandnewlineagesofA.mexicanus.WediscussbelowthatthepresenceoftwodivergentmtDNAhaplotypesisnotperseastrongsupportfortheexistenceoftwofishlineages.Moreover,theincongru-enceofthemtDNAphylogenywithphylogeniesob-tainedwithseveralindependentnuclearlocidefinitivelyinvalidatesthishypothesis.AwidelyacceptedscenariointhecommunityworkingonA.mexicanuscavefishisthatsomecavepopulationsareancient,i.e.hundredsofthousandsorevenmillionsofyearsold,andrelatedtoextinctsurfacefish,whereasothercavepopulationsaremorerecentandrelatedtoextantsurfacefish.Wedemonstratebelowthatthishy-pothesisreliesonlyontheexistenceoftwodivergentmtDNAhaplogroupsthataresupposedtoreflecttheex-istenceoftwodivergentfishlineages.ThehypothesisthatcavefishoriginatedfromtwoseparatesurfacefishstockswasfirstformulatedonthebasisofaNADHde-hydrogenase2(ND2)phylogenyofcaveandsurfacefish[39].OntheonehandallsurfacefishfromtheSierradeElAbrabelongedtoahaplogroupnamed“lineageA”,aswellastwosurfacefishfromTexasandasurfacefishfromtheCoahuilastate,innortheasternMéxico.Pachón Fig.5Goodnessoffittothedata.Themodelparametersare:SFpopulationsize=10,000;CFpopulationsize=625;%migrantsfromsurfacetocave=0.1;migrationratefromsurfacetocave=0.001/year;SFgenerationtime=2years;CFgenerationtime=5years;labpopulationparameters:10fish,10generations.aScoreofgoodnessoffitaccordingtotheageofthecavepopulation,thebestfitiswhenthecavefishpopulationis21,500yearsold.bEvolutionoftheSNPclassfrequenciesduringthesimulation.HorizontaldottedlinesaretheobservedSNPclassfrequencies.Observedandsimulatedfrequenciesattheageofthebestfitareshowninthetoprightcorner.cEvolutionofthenumberofpolymorphicsitesinSFandCFduringthesimulation.dEvolutionofthenumberofderivedallelesthatwerefixedinSFandCFduringthesimulation.eEvolutionoftheSF/CFpolymorphismratioandtheCF/SFderivedalleleratiothatreachedfixationduringthesimulation.Horizontaldottedlinesaretheobservedratios.TheverticaldottedlineistheageofthecavefishpopulationforwhichthebestfitwasobservedFumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page8of19

PAGE 9

andChicacavefishalsobelongedtothishaplogroupA.Ontheotherhand,Curva,TinajaandSabinoscavefish,livingincavesthataregeographicallyclosetoeachother,belongedtoanotherandwelldifferentiatedhap-logroupnamed“lineageB”.Theauthorsconcludedthatthecavefishbelongtoanoldstockoffish,“thelineageB”thatwaspresentatthesurfacealongtimeago,butnowextinctandreplacedbysurfacefishwithhaplotypesbelongingtohaplogroupA.NoteworthythishypothesisimpliesthatthemtDNAhaplotypeA1foundinPachónandChicacavefish(ahaplotypefoundinmostsurfacelocalities)istheresultofrecentmtDNAintrogressionsintothesecaves.Theauthorsofthispublicationpro-posedanotherexplanation:whereasitislikelythatintro-gressioncanoccurinChicacavewheresurfacefishandhybridshavebeenfound,theysuggestedthatPachóncavefish,thatseemmuchmoreisolated,haveevolvedin-dependentlyandmorerecentlythanhaplogroupBcave-fish,andtheyareundergoingtroglomorphicevolutionmorerapidlythanothercavefishpopulations[39].Thesehypotheseswereamongthemostparsimoniousthatcouldbeformulatedatthattimewiththisdataset.Itisimportanttonotethat,andevenifweaccepttheadhochypothesisthatasurfacepopulationhasbeenreplacedbyanothersurfacepopulationwithmtDNAsbelongingtoadivergenthaplogroup,itdoesnotnecessarilyimplythattheageofthecavefishrelatedtothefirstpopulationisthetimeofcoalescenceofthetwodivergentmtDNAhaplogroups.Indeed,twosurfacepopulationscouldhaveevolvedindependentlyforaverylongtimeintwosepa-ratedMexicanregions,allowingtheevolutionoftwodi-vergenthaplogroups,butonepopulation(theextantpopulation)couldhavereplacedthefirstone(extinctpopulation)onlyveryrecently.Moreover,somecavepopulationscouldhaveevolvedfromthefirstextinctsurfacepopulationrecentlytoo,butofcoursebeforeitsextinction.Veryrecentlytoo,othercavepopulationscouldhaveevolvedfromtheextantsurfacepopulation.Insum-mary,underthehypothesisofreplacementofasurfacepopulationbyanotherone,thecoalescencetimeofmtDNAcangive,atbest,theolderagepossibleforcave-fishdescendantsoftheextinctsurfacefish.Nevertheless,evenifallcavesettlementsareveryrecentitdoesnotpre-cludefindingfishcarryingdivergentmtDNA.Anotherstudyusingapartialsequenceofthecyto-chromebgeneconfirmedtheexistenceoftwodivergentmtDNAhaplogroups[36].Thisresultwasexpectedasmitochondrialgenesarecompletelylinkedandauniquephylogenyisexpectedformitochondrialgenomes.Moreover,inthisstudyathirdhaplogroupwasidentifiedinYucatan.UsingamorecomprehensivesampleandthesamemtDNAmarker[37],uptosevendivergenthaplogroupswerefoundinMexico(AtoG,haplogroupGforcytbcorrespondingtohaplogroupBwithND2)withahighlystructuredgeographicdistributionsuggest-ingpastfragmentationand/orastrongisolationbydis-tance.Inthisstudy,haplogroupGwasstillcavespecific(Piedras,Sabinos,TinajaandCurvathatarecavesclosetoeachother)andhaplogroupAwasstillNorthernGulfcoastandcavespecific(Fig.1).Howeveramorerecentanalysis[42],expandingfurtherthesampleofpopula-tions,allowedtheidentificationofsurfacefishwithhap-lotypesveryclosetohaplogroupG(namedCladeIIlineageIe)andhaplogroupA(namedCladeIIa)insym-patryinasamewaterbodies,i.e.MezquitalandAgana-val,innorthwesternMexico.ThisfindinginvalidatesthehypothesisthathaplogroupGevolvedintheElAbrare-gionalongtimeago,wentextinct,andwasreplacedbyhaplogroupA.Moreover,haplotypes“Glike”werealsofoundinsurfacefishlocalities(RasconandTamasopo)closetoSierradeElAbra[42].HaplogroupsAandGarehighlydivergent,supportingamodelinwhichtheyaccumulatedmutationsintwopopulationsisolatedforalongperiodoftime.Thepres-enceofbothhaplogroupsinnorthwesternandnorth-easternMexicosuggeststhatthesepopulationsmixedrecently,atthetimeofasecondarycontact.Datingre-sultsdiscussedbelowsuggestthatduringthelastglaci-ation,twoallopatricpopulationsfromnorthMexico,onecarryinghaplotypesbelongingtohaplogroupAandtheothercarryinghaplotypebelongingtohaplogroupG,mighthavemovedsouthandmixedthere.Afterthisglaciationtheymighthavemovednorthagain,nowshar-inghaplotypesbelongingtohaplogroupAandG(thishaplotypemixtureisactuallyobservedinthenorthwest-ernregion,i.e.MezquitalandAganavalwaterbodies).Inthenortheasternregion,haplotypesbelongingtothehaplogroupGhaveuptonowbeenfoundonlyinseveralcavesinarestrictedgeographicareaandhaplotypes“Glike”insurfacelocalitiesalsoinarestrictedarea.Note-worthy,suchrecentsecondarycontactofdivergenthap-logroupswerealsoobservedatseveralotherplacesinsouthMexico[34,42]suggestingthatseveralpopula-tionsofAstyanaxmexicanuswereisolatedforalongtimeindifferentregionsinMexicoandCentralAmericaandhavebeenrecentlyundergoingsecondarycontact.Insummarywethinkthat,consideringthemtDNApolymorphismalone,thereisnoreasontobelievethatthecoalescencetimeofthemtDNAhaplogroupsshouldcorrespondtotheageofthemostancientcavefishpopula-tions.Onthecontrary,takingintoaccountthemostre-centpublications,itsuggestsarecentadmixtureoftwodivergentpopulations.Thisadmixtureshouldberecentenoughtoallowthemaintenanceofbothhaplogroupsatdifferentgeographicscale(innorthMexicoasawholeandinnorthwesternandnortheasternMexicoindepend-ently),asgeneticdriftshouldhaveeliminated,atasmallgeographicscale,oneofthemafteralongperiodoftime.Fumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page9of19

PAGE 10

Moreover,theexistenceoftwofishlineagesimpliesthatthemtDNAphylogenyshouldbecongruentwithunlinkednuDNAphylogenies,whereasarecentadmix-tureoftwosurfacepopulationsbeforefishsettlementsincavesshouldleadtorandomfixationofallelesatdif-ferentunlinkedlociandthusincongruentphylogeniesbetweenmtDNAandnuDNAlociaswellasbetweendifferentunlinkednuDNAloci.Noteworthy,aphylogeneticanalysiswasperformedusingalargeSNPdatasetinordertoestimatethenum-berofindependentcavesettlementsattheoriginoffivecavefishpopulationsfromthreedistinctregions[43].Thisanalysisdidnotsupportthetwolineageshypothesisbutindicatedatleastfourindependentoriginsforthesecavefishpopulations.However,thisstudycouldnotevaluateifcongruentphylogeniesareobtainedwithsam-plesofsequencesfromdifferentnuclearloci.Moreover,datingtheageoftheinternalnodesofthephylogenetictreewasnotpossible.WecomparedmtDNAandnuDNAphylogeniesusingpublishedsequencesofseveralnucleargenes.First,wereconstructedamaximumlikelihoodphylogenywithRag1.Theresolutionissolowthatitprecludesanyphylogeneticinference,andevenspeciesdefinedusingmtDNAarenotsupported.Thecongruenceofthephy-logenieswithmtDNAandmtDNA+Rag1[42]isthere-sultoftheverylowquantityofphylogeneticsignalinRag1comparedtomtDNAanditdoesnotsupportthecongruenceofmtDNAandRag1phylogenies.Thenweexaminedphylogeniesobtainedwithothernu-cleargenes(Mc3r,Mc4r,Lepb,Lepr,Pomcb).Thesephy-logeniesarebasedonfoursequences,butthereareneverthelesshighlyinformative.Foreachgenewefoundauniquephylogenywithouthomoplasysuggestingnore-combinationwithineachlocus.Moreover,theincongru-enceofthephylogeniesobtainedwiththeseunlinkedlocisupportstheirindependentevolutionaryhistories.Theseresultsdonotsupporttwowelldefinedfishlineageswhereasadmixtureofgenephylogeniesisexpectedwhensampledlocalitiesarepoorlyisolatedand/orhavebeenseparatedforashortperiodoftime.Partialorcompletecodingsequencesoffiveothergenes(Per1,Per2,Tef1,Cry1aandCpdphotolyase)fromthreelocalities(PachónandChicacavesandasurfacelocalityclosetoMicos)werealsoavailable[52].ThesesequenceswerealignedwiththesequenceoftheTexassurfacefishandthemostparsimonioustreerecon-structedforeachgene.Surfacefishsequenceswereal-waysverycloseconfirmingthemtDNAevidencethatthesurfacepopulationsampledinTexasisgeneticallyveryclosetoSierradeElAbrasurfacefish.Thesephy-logeniesarealsointerestingbecausetheyhighlightan-otherfact.Whereasforsomegenes,allthehaplotypesarealmostidentical(veryfewmutationsinMc1r,Mc4r,Lepb,Pomcb,Per2,Tef1,Cpdphotolyase),wecaniden-tifytwo,andonlytwo,divergenthaplotypesforMc3r,Lepr,Per1andCry1a.Moreover,thedistributionofdi-vergenthaplotypesisnotthesamefordifferentloci(adivergenthaplotypeofMc3risfoundinTinajacaveonly,adivergenthaplotypeofLeprinMolinocaveonly,diver-genthaplotypesofbothPer1andCry1aatthesurfaceonly).TakingintoaccounttheexistenceoftwodivergentmtDNA(“G”haplotypeinTinajaand“A”haplotypeinPachónandsurfacefish),thesephylogeniessuggestthattwodivergentfishlineageswithwelldifferentiatedge-nomesmixedanddivergentallelesateachlocussegre-gatedrandomlyatthesurfaceandincaves.Whentherearenodivergentallelesatalocus(Mc1r,Mc4r,Lepb,Pomcb,Per2,Tef1,Cpdphotolyase),onecansupposethatallelesfromoneancestralpopulationwentextinct.Onthisbasiswecametotheconclusionthatcavefishcouldbemuchmorerecentthanusuallythought.Inordertomakeaquantitativeanalysisofthishypothesisweappliedtwodifferentapproachestoestimatetheageofsomecavepopulationsusingmultipleunlinkednuclearloci.DatingisolationtimesofcavepopulationsDatingtheageofarecentlyisolatedpopulationthatcanexchangemigrantswiththe“source”populationisadif-ficulttask[58].Ifdivergenceislowandthereissharedpolymorphismbetweentwopopulations,itcanbetheresultofregularmigrationbetweenthesepopulationsthatdivergedalongtimeago,ortheconsequenceofarecentdivergenceofcompletelyisolatedpopulations,orsomethinginbetween.Onecanthusestimatehowlongagothepopulationsdiverged(assumingnogeneflow)usingphylogeneticmethods,oronecanestimatethegeneflow(assumingthatthepopulationsareatmutation/mi-gration/driftequilibrium,i.e.theyhavebeenseparatedforaverylongtime,migrationsoccurredregularlyandthusthephylogeneticsignalhasbeenerased)usingpopulationgeneticmethods.SuchmethodshavebeenappliedtostudytheevolutionofA.mexicanuscavefish[32–39,42,43,59],butneitheroneisofmuchuseandoftenmisleadingifthegoalistode-velopafullpicturethatincludesestimatesofrecentseparationtimeandgeneflow[58].Insuchcaseitisnecessarytoconsidernon-equilibriummodelsandmethodsallowingthejointestimationofdemographicparameters(populationssizesandmigrationrates)anddi-vergencetime[58].Accordingly,weusedIMa2,awidelyusedprogramfor“isolationwithmigration”(IM)modelanalyses[53]toestimatedivergencetimebetweensurfaceandcavepopulationswithadatasetofmulti-locusmicro-satellitepolymorphism.IMa2isbasedonbackwardsimu-lationsofcoalescenceofsamplesofalleles.ForPachóncavepopulation,weestimatedthedivergencetimeusinganalternativeapproachbasedonforwardsimulationsofFumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page10of19

PAGE 11

evolutionofSNPs.Analysesofmicrosatellitepolymorph-ismwithIMa2supportedarecentoriginofallcavepopu-lations.AnalysesofSNPsconfirmedarecentoriginofPachóncavefish.DatingwithmicrosatellitesThemicrosatellitedatasetwaskindlyprovidedbyM.BradicandR.Borowsky[33].Weperformedaseriesofpairwiseanalysesimplyingacavepopulationandasur-facepopulationortwocavepopulationsusingIMa2inordertoestimatethemarginalposteriorprobabilitydensityofmodelparameters(i.e.populationsizes,mi-grationratesanddivergencetime).WhereasthecurrentversionofIMa2canhandlemorethantwopopulations,thephylogeneticrelationshipsbetweenpopulationsmustbeknown.InthecaseofAstyanaxmexicanus,asshownabove,thereisnoobviousphylogeneticrelationshipsbe-tweensurfaceandcavepopulations.Inaddition,asthenumberofparametersincreasesveryfastwiththenum-berofpopulationsanalyzed,itisaverydifficulttasktoanalyzemorethantwopopulations.However,itispos-sibletostudyalargecomplexdivergenceproblemthatinvolvesmultiplecloselyrelatedpopulationsbyanalyz-ingpairsofpopulations[53,60].FirstwefocusedonthedivergencetimeofPachóncavepopulation(O1,accordingtoBradicetal.nomen-clature[33])andasampleofsurfacefishfromcloselo-calities(S3)(Fig.1)becauselargesamplesofalleleshadbeengenotypedformanylocianditmayallowmoreac-curateestimationsofmodelparametersthanforcavepopulationsforwhichalimitednumberofallelesweregenotyped.Assumingamutationrate,estimationsoftheeffectivepopulationsizes,migrationratesanddivergencetimecanbeobtained.Inmostpopulationgeneticstud-ies,includingA.mexicanus[33],themutationrateofmicrosatellitelociisassumedtobeabout5×104.Thisisaquitehighmutationrate,butitisverylikelyasthelociretainedforpopulationgeneticanalysesarethemostvariable,thusthosewiththehighestmutationrate.Theestimationoftheeffectivepopulationsize(Ne)was150[150–1150]and10,750[6650–16,350]forPachóncavepopulationandsurfacefishpopulation,respectively.TheseestimationsmakesenseasitisobviousafterseveraltripsinSierradeElAbrathatthecensuspopulationsize(Nc)ofsurfacefishismuchhigherthanthecensuspopulationsizeofthecavefish.NchasbeenestimatedforPachóncavefish(8502;95%confidencelimits[1279–18,283])[1]butithasneverbeenestimatedforsurfacefish.OntheonehandweexpectthatNeiscorrelatedwithNc,butforfishthatcanpotentiallylayorfertilizehundredsofeggsornoeggsatallduringtheirlifesuchasA.mexicanusfish,thevarianceofthenumberofdescendantsisprobablyhigh(intherangeof101to102)andthusNemightbeonetotwoordersofmagnitudesmallerthanNc[61,62].IfNcisintheorderofmagnitudeof104,itisexpectedthatNeisintheorderofmagnitudeof102to103asfoundinthepresentstudywithIMa2.Previousstudies,usingthesamedatasetandtheprogramMigrate[63],foundsimilarNeforseveralcaves,includingPachón[33,35].TheresultsobtainedwithIMa2suggestedthatthemi-grationratefromthecavetosurfaceisnegligiblewhereasthemigrationratefromthesurfacetothecaveislow.Thisisalsoexpected.Iffishcouldeasilyexitthecave,noevolutionofcavefishwouldhaveoccurred.Moreover,itisdifficulttoimaginethatblindcavefishwhofoundtheirwaytothesurfacewillhaveagoodfit-nessthere.Concerningthemigrationrateofsurfacefishintothecave,itislikelythatthefitnessofsurfacefishinacaveislowcomparedtowell-adaptedcavefish(LuisEspinasa,personalcommunication).So,evenifthemi-grationrateofadultsurfacefishintocaveisnotnegli-gible,the“effectivemigrationrate”,thatistherateofmigrationofsurfacefishthatactuallyreproduceinthecaveisprobablyextremelylow.Accordingly,surfacefishhasneverbeenobservedinPachóncave,butfishthatwerelikelyhybridswerereportedduringacoupleofyearsinthe80’s[64].Thisobservationmadebyonlyonegroupofinvestigatorshadneverbeenmadebyany-bodyelse,beforeorafter.TheIMa2estimationofdivergencetimeof5110years[1302–18,214]suggeststhatthePachónpopulationismuchyoungerthanusuallythought.Indeed,andwithoutanycomplexcomputations,asimpleglanceatthemicro-satellitedataset(Additionalfile3)actuallysupportsare-centoriginofthiscavefish:1)thereisnodivergenceofthedistributionofallelesizesfoundinthesurfacefishandPachóncavefish,2)theallelesfoundinthecavearealsopresentatthesurface.Thedifferencebetweenthesepopu-lationsisthat,ateachlocus,therearemanydifferentallelesizesatthesurfacebutamuchsmallernumberofallelesizesinthecave.Thisdifferentiationwithoutdivergencecanbeeasilyexplainedbyamuchhighergeneticdriftinthesmallcavepopulationthaninthelargesurfacepopula-tion.Ofnote,evenifweconsiderthatthemutationrateis10timeslowerthantheratetakentomaketheseestima-tions,theoriginofthePachóncavefishwouldstillbemuchmorerecentthanusuallythought.Similarresultswereobtainedwithallthecavepopula-tionsstudied.TheyallappearedhavinglowNeandare-centorigin,lessthan10,000yearsago.Takingintoaccountthelargevarianceofthedivergencetimeestima-tions,wecanconcludethattheyaremostlikelyalllessthan20,000yearsold.Inaddition,takingalsointoaccounttheuncertaintyonthemeanmutationrateofthemicrosa-tellitesthatcouldbeabout5timeslowerthanassumed,thelimittotheestimationoftheageofthecavepopula-tionscouldbepushedtoabout100,000years.Fumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page11of19

PAGE 12

DatingwithSNPsEventhoughtherecentevolutionofA.mexicanuscave-fishiswellsupportedbyanalysesofmultiplemicrosatel-litelociwithIMa2,thisestimationwassoatoddswiththecurrentopinionofantiquityofmostcavepopula-tionsthatweconsiderednecessarytobringadditionalevidenceusingacompletelydifferentapproachandato-tallydifferentsetofdata.Acongruentestimationofthemodelparametersofinterest,inparticularthediver-gencetime,wouldgreatlystrengthenourconclusion.WefocusedondatingPachóncavepopulationforwhichweidentifiedalargesampleofSNPsinRNA-seqofpooledembryosoffishmaintainedinthelab.Weper-formedthedatingofthedivergencetimeofthispopula-tionwithaTexassurfacepopulationoffishforwhichSNPswerealsoidentifiedusingthesameapproach.Asdiscussedabove,alldata(mtDNAandnuDNA)showedunambiguouslythatthesesurfacefisharecloselyrelatedtothesurfacefishsampledintheSierradeElAbraregion.Despitethisevidenceandtheabsenceofahighstructurationofthegeneticdiversityofthesur-facefishintheSierradeElAbraregion,ifthereisageneticallyclosersurfacepopulationlivingnearthePachóncave,thestraightforwardconclusionwouldbethattheresultweobtainedisanoverestimateoftheageofthePachóncavepopulation.Ofnote,theTexassurfacefishsharedabout7%oftheirSNPswithPachóncavefish.Assharedpolymorph-ismisexpectedtodecreasequicklyifatleastonepopu-lationhasaloweffectivepopulationsizeandthetwopopulationsarecompletelyisolated,itsuggeststhatthedivergenceisrecentorthemigrationrateishigh.Inordertomakeestimationsofthemodelparameters(i.e.populationsizes,migrationratesanddivergencetime)thatcouldexplainthedistributionofSNPswithinandbetweenpopulations,weransimulationsoftheevo-lutionoftheSNPsforwardintimeintwopopulations,allowingmigrationfromthesurfacetothecave.Run-ningthesesimulationswithdifferentsetsofparametersallowedfindingsimulationsforwhich,afteragivennumberofyearsofdivergence,thedistributionoftheSNPswithinandamongsimulatedpopulationswasverysimilartothedistributionobservedwithinandamongtherealpopulations.Thetimeofdivergencewastakenasanestimationoftheageofthecavepopulation.Theuseofsimulationsinordertoestimateunknownvaluesofmodelparametersiscommoninpopulationgenetics,forexampleinApproximateBayesianComputationmethods[65].Therationaleofthisanalysisisthatwhenapopulationisdividedintotwopopulations,ancestralpolymorphismissharedbythedaughterpopulationsforaperiodoftime,evenifthepopulationsaretotallyisolated.Asdivergenceproceeds,eachdescendantpopulationexperiencefixationandlossofallelesatlocithatwerepolymorphic,andthisrandomsortingofallelesispartofthewaypopulationsdi-verge.Thedivergenceofthepopulationsalsoincreasesthroughtimebecausenewallelesappearatnewanddifferentpolymorphicsitesinbothpopulations.How-ever,migrationallowssharingoftheancestralandnewpolymorphismandcanrestrictthedivergenceofthepopulations.Asalreadydiscusseditisthuschal-lengingtoestimateifsharedpolymorphismisduetoarecentsplit,genefloworboth[66].However,someobservationsonSNPsdistributionwithinandamongthecaveandsurfacepopulationssuggestedarecentdivergence.Wefound2.34timesmorederivedallelesthatreachedfixation(substitutions)atsynonymouspolymorphicsitesinthecavefishpopulationthaninthesurfacefishpopu-lation.Anexcessofnon-synonymousandnon-codingsubstitutionswerealsoobserved.Wesearchedforanex-planationfortheseobservationsthatcouldappearatfirstglanceunexpected,inparticularforsynonymoussubstitutionswhichareformostofthemprobablyneu-tralornearlyneutral.Itiswellknownthatiftwopopu-lationshavedivergedforalongtimeandifthemutationrateisthesameinbothpopulations,theneutralsubsti-tutionratesshouldbeequalandindependentofthepopulationsizes[67].Nevertheless,asimpleexplanation,totallycompliantwiththisfundamentalresultoftheoret-icalpopulationgenetics,reliesonthefactthatwhenanancestralpopulationisdividedintoalarge(surface)andasmall(cave)population,theprobabilityoffixationofaneutralalleleisthesameinbothpopulations(itistheallelefrequency)buttheprocessoffixationisfasterinthesmallthaninthelargepopulation.Indeedthemeantimetofixationis t1ðpÞ¼4Nð 1ppÞlnð1pÞ(whereNisthepopulationsize,pistheallelefrequency)[68].Thestraightforwardconsequenceisatransientaccelerationofthesubstitutionpaceinthesmallpopulationthatisnotanymoreobservedafteralongperiodoftime[67].Inotherwords,itisexpectedthatduringashortperiodoftimeaftertheirseparationasmallpopulationwouldhadfixedmorederivedallelespresentintheancestralpopulationthanthelargepopulation.Wethoughtthatthisinformation,togetherwithotherinformationaboutthedistributionofpolymorphismwithinandbetweenpopulations,couldbeusedfordivergencedating.Wedefinedsummarystatisticsdescribingthepoly-morphismandthedivergenceoftwopopulationsthatcouldbeaccuratelyestimatedusingpooledRNA-seq[69].TheevolutionarymodelisidenticaltothemodelanalyzedwithIMa2andreliesonthesameparameters(threeeffectivepopulationsizes,twomigrationrates,andadivergencetime).Ofnote,wesetthemigrationFumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page12of19

PAGE 13

ratefromcavetosurfacetozeroastheanalyseswithIMa2andseveraltripstothiscaveconvincedusthattheimpactofmigrationofPachóncavefishonsurfacefishDNApolymorphismisnegligible.Inoursimulations,wesetthegenerationtimetotwoandfiveyearsforthesurfaceandcavepopulations,re-spectively.ThissurfacefishgenerationtimeistwicetheestimationsobtainedforotherAstyanaxspecies[70]andthecavefishgenerationtimeisthevalueestimatedbyP.Sadoglu,unpublishedbutreportedasapersonalcommunication[40].Thisestimationisbasedonthehy-pothesisthatcavefishmayliveandremainfertileforalongtime,about15years.Itisunlikelythatthesegener-ationtimesareunderestimatesandtheycouldactuallybeoverestimatesofthetruegenerationtimes,inparticu-larforPachóncavefish[56].AstheestimationoftheageofthePachóncavefishpopulationdirectlydependsonthesegenerationtimes,thedivergencetimeswediscussbelowaremorelikelyoverestimatesthanunderesti-mates.Wetookintoaccountmigrationratefromsur-facetocave.Thismigrationratedependedontwoparameters:theprobabilityofmigration/yearandthepercentageofcavefishthatweresurfacemigrantswhenamigrationoccurred.Thiswayitispossibletosimulatedifferentpossibilitiessuchasfewmigrantsthatenterthecaveveryoftenormanymigrantsinveryrareoccasions,andallintermediatecasesbetweenthesetwoextremes.Wealsotookintoaccountthatgeneticdriftoccurredduringseveralgenerationsinthelab.Inordertoestimatetheeffectofgeneticdriftinthelabandtheaccuracyofal-lelesfrequencyestimationsusingpooledRNA-seq,wecomparedthefrequenciesoftwoalleles(Pro106Leu)ofMAOgeneestimatedusingasampleofwildcaughtPachónfish,asampleofadultfishmaintainedinthelabwithfrequenciesestimatedusingpooledRNA-seqoflabembryos.AlthoughithasbeenpublishedthatPachónpopulationisnotpolymorphicatthislocus[71],wefoundpolymorphismwithalargersampleofwildcaughtfishandsimilarfrequencieswithbothadultandembryolabfish(datanotshown)suggestingthatgeneticdriftdidnotremovefromthelabpopulationapolymorphismknowntobepresentinthePachónpopulation.OfnotewhereastheestimationofthedivergencetimedependsonthegenerationtimeandthemutationratewithIMa2,ourestimationdependsonlyonthegenerationtime.Thisisduetothefactthatthesummarystatisticsweused,i.e.SNPsclassfrequencies,tocomparesimulatedandrealpopulations,donotdependonthemutationrate.WesetamutationrateallowinganumberofSNPswithinandbetweensimulatedpopulationslargeenoughtohaveaccurateestimationsofthesummarystatistics.Withoutmigration,sharedpolymorphismswerequicklylostandthebestfitofthemodeltothedatawasobtainedwhenthecavefishpopulationsize(1250)wassmallerthanthesurfacefishpopulation(10,000)andtheageofthecavefishpopulationwas20,600years.Whenmigrationwasincluded,goodfitwiththedataalsoimpliedlargedifferencesinpopulationsizes,alowmigrationrateandlownumbersofmigrants.Theverybestfitwasob-servedforasurfacefishpopulationsizeof20,000,acavepopulationsizeof1250,andacavepopulationageof54,200years.Nevertheless,mostverygoodfitswereobtainedwithadivergencetimeintherangeof20,000to30,000years.Asexpectednofitwasfoundwhenthesurfaceandcavepopulationshadsimilarsizes.Thebestfitwasob-servedwhentheancestralpopulationsizewassimilartothesurfacepopulationsize.Itisatoddswiththeestima-tionofamuchlargerancestralpopulationsizeestimatedwithIMa2.Wedonothaveaclearexplanationforthisdiscrepancy,butitcouldbetheconsequenceofthead-mixtureoftwodivergentpopulationsattheoriginoftheancestralpopulation.Suchadmixturecouldhavein-creasedthenumberofallelesateachmicrosatelliteloci,andIMa2thusinferredalargerancestralpopulationsize.ForSNPs,admixtureofdivergentpopulationsre-sultsinmorepolymorphicsites,butthenumberofal-lelesatagivenlocusisnotincreasedastheprobabilitythattwoparallelmutationsoccurredindependentlyatthesamelocusintwopopulationsisextremelylow.OtherevidenceforarecentoriginofthePachóncavefishInarecentanalysisoftheexpressionof14crystallingenesinthePachóncavefish,4genesarenotexpressedorexpressedataverylowlevel,butnostopcodonorframeshiftcouldbeidentified[26].Thisresultisinac-cordancewitharecentoriginofthispopulation,assev-eralloss-of-functionmutationsshouldhavereachedfixationafterseveralhundredthousandyearsofevolu-tionofgenesthatwouldnolongerbeunderselection,astheyarenotnecessaryinthedark[47].Indeed,otherfishspeciesthatarelikelyconfinedintocavesformil-lionsofyearshavefixedloss-of-functionmutationsinseveralopsinsandcrystallinsgenes[48–50].Wearecur-rentlyworkingondatingcavefishpopulationsusingthefrequencyoflossoffunctionmutationsingenesthataredispensableinthedark.Second,arecentstudyhasshownthattheheatshockprotein90(HSP90)couldrestricttheexpressionofeye-sizepolymorphisminsurfacepopulations[72].Indeed,HSP90inhibitionallowedtheobservationofalargereye-sizevariation.Moreover,ithasbeenpossibletose-lectfishwithareduced-eyephenotypewhichcanbeob-servedeveninthepresenceofHSP90activity.Thisresultsuggeststhatstandinggeneticvariationinextantsurfacepopulationscouldhaveplayedaroleintheevo-lutionofeyelossincavefish.Thisisalsocompatiblewitharecentoriginofthecavepopulation.Fumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page13of19

PAGE 14

Non-equilibriummodelsandcavefishpopulationgeneticsArecentoriginoftheso-called“old”Pachónpopulationcansolveaconundrumputforwardbypreviousandthepresentanalyses.First,attheSNPandmicrosatellitelevel,thediversityisnotthatlowinPachóncavewhencomparedwithsurfacepopulations,i.e.aboutonethird.Ifthepopulationsareatmigration/driftequilibrium,itmeansthattheeffectivepopulationsizeofPachóncave-fishisaboutonethirdofthesurfacepopulations,andthisisatoddswiththelikelylargedifferenceincensuspopulationsizes[1,33].Ofcourse,wecanproposeadhochypothesestoexplainthisdiscrepancy.Cavefishmayhaveamuchlowerreproductivesuccessvariancethansurfacefish,orsurfacefishcouldhavelargerpopu-lationsizefluctuationsthroughtimethancavefish.Insuchcases,theeffectivepopulationsizescouldbemuchclosertooneanotherthancensuspopulationsizesbe-causeitiswellestablishedthatlargevarianceinrepro-ductivesuccessandlargepopulationsizefluctuationshugelyreducetheeffectivepopulationsize[62].AnalternativeexplanationisthatthegeneticdiversityinthePachóncaveisactuallyhigherthanexpectedatmutation/drift/migrationequilibrium.Ourresultssug-gestthattheeffectivepopulationsizeofthesurfacefishisatleastoneorderofmagnitudelargerthantheeffectivepopulationsizeofcavefish,aratiothatismoreinaccordancewiththeunknownbutcertainlyverydifferentlongtermcensuspopulationsizes.Thepresentstudyisastrikingillustrationofhowmislead-ingcanbeanalysesofevolutionaryprocessesthatdonottakeaccountthatbiologicalsystemsarenotnecessar-ilyatequilibrium.Thetwoanalyseswedescribedaboverelyonapproachesthatdonotsupposemutation/migra-tion/driftequilibrium.Theyallowedtheestimationofdemographicparametersthataremoreinlinewithexpec-tationsbasedonfieldobservationsthanpreviousestima-tions.Therearemuchmoresurfacefishthancavefishandtheimpactofmigrationofcavefishonsurfacefishdiver-sityislikelyextremelylow,andmostlikelynull.Thenewtimeframeweproposefortheevolutionofthecavefishpopulationswouldnotallowenoughtimeforthefixationofmanydenovomutationsandmostde-rivedallelesthatreachedfixationincaveswereprobablyalreadypresentintheancestralpopulation.Thiswasalsosuggestedbyarecentpopulationgenomicstudy[73].Thismayimplythatthecavephenotypeevolvedmainlybychangesinthefrequenciesofallelesthatwererareintheancestralsurfacepopulation.Inpar-ticular,someofthesealleleswouldhavebeenloss-of-functionordeleteriousmutationsthatcouldnotreachhighfrequencyinsurfacepopulationsbuttheycouldreachhighfrequencyorfixationquicklyinasmallcavepopulationwheretheyareneutralorevenadvantageous.ConclusionAlthoughweestimatedthatallcavefishpopulationsareprobablyrecent,lessthan20,000yearsold,thenumberofindependentandapproximativelysimultaneousadap-tationstocaveandevolutionofcavephenotypeisstillanopenquestion.Theevolutioninashortperiodoftimeofthephenotypeofindividualsbelongingtoapopu-lationadaptingtoanewenvironment,isactuallynotthatunexpectedandhasalreadybeenobservedinotherfishspeciessuchasthestickleback[74],dwarfwhitefishes[75]andAfricancichlids[76,77].Recently,thefirstEuropeancavefish,withawelldifferentiatedcavephenotype,hasbeendescribed.ThephylogeneticanalysisofmtDNAhap-lotypes,theanalysisofgeneticdifferentiationusingmicro-satellitelociandtherecentglacialhistoryoftheregionsuggeststhatthesefishpopulationishighlyisolatedbutforlessthan20,000years[78].Mexicancavefishcouldthusbeanotherandstrikingil-lustrationthatmanyphenotypicchangescanaccumulateinparallelandinashortperiodoftimethankstostandinggeneticvariation[79].Therelativerolesofselectionanddriftinallelicfrequencychangesisnotyetunderstood,butiftherecentoriginofcavefishpopulationsiscon-firmed,theywouldbeanexcellentmodeltoanalyzethisissueusingpopulationgenomicstoolssuchasthequanti-ficationofselectivesweeparoundcandidatelocimostlikelyinvolvedintheadaptationtoacaveenvironment.MethodsDatingwithIMa2Withamultilocusmicrosatellitepolymorphismdataset[33]andtheprogramIMa2[53]divergencetimesofpairsofpopulationswereestimated.Theprogramisbasedonanisolation-with-migration(IM)modelandusesMetropolis-coupledMarkovchain(MCMC)tech-niquestoestimatetheposteriordensitiesofthetimeofdivergence,populationsizesandgeneflow[58].Themodelassumesrandompopulationsamples,astepwisemutationmodel,neutralmutation,freelyrecombininglociandconstantpopulationsizesandgeneexchangerates.Althoughmodellingconstantpopulationsizesandgeneexchangeratesmightnotbeidealinthecaseofcavefishpopulations,modellinggeneralpatternsrequiressimplificationsandthisispresentlytheonlyoptioninprogramssuchasIMa2andMigratewhichisthepro-grampreviouslyusedtoestimategeneflow[33,63].Itwasalsonotpossibletotakeintoaccountchangesingenerationtimeaftertheseparationofthepopulations.Itisthusmoretheorderofmagnitudeoftheparameterestimationsthatcanbediscussedratherthanmaximumlikelihoodvaluesperse.Astheupperlimitofthepriordistributionofeachmodelparametermustbeset,weranshortMCMCchainstotestseveralsetsofparame-tersthatallowedtheidentificationofsuitablevaluesofFumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page14of19

PAGE 15

upperlimitsofmutation-scaledeffectivepopulationsizes,geneflowsanddivergencetime.Itallowedalsotoestimatethelengthoftheburn-inperiod.Inordertousealargeandevennumberofalleles/locus/populationswesampledatrandomthesamenumberofallelesateachlocusineachpopulation.Moreover,werestrainedthenumberofallelesto60asittookmorethanonemonthtogettheresultswith22lociand60alleles/locus/popula-tion.Suchalargesampleofallelesformanylociwasnotavailableformanycomparisons.Inthatcases,weusedsmallersamplesoflociandallele/locus.Nevertheless,wedidnotretainanalyseswithlessthan20alleles/locusastheresultswerenotrobustaccordingtoaresamplingprocess.Theparametersetsusedinthedifferentanalysesareindicatedasacommandlineinthelegendofthefig-uresthatsummarizetheoutputofIMa2.Thenumberoflociandthenumberofalleles/locus/populationarealsoindicated(Additionalfile1:FigureS3toFigureS9).SampledpopulationsforSNPsanalysesForfifteenyearswehavemaintainedlaboratorystocksofAstyanaxmexicanuscavefishandsurfacefish,foundedwithfishcollectedrespectivelyinthePachóncave(SierradeElAbra,Mexico)andattheSanSolomonSpring(Texas,USA)(Fig.1),andobtainedfromW.RJefferyin2004.SeveralstudiesshowedthatthedifferentiationofthesurfacepopulationsintheElAbraregionandTexasisverylow[33,37,39],suggestingthatthecomparisonofanyofthemwithacavefishpopulationshouldgiveaboutthesameresult.Indeed,wefoundalsoaverylowmtDNAdivergencebetweentheTexaspopulationandthePachóncavepopulation(Additionalfile1:FigureS2).Inaddition,wefoundthattheyshareabout7%oftheirSNPs(Table3).Bothobservationssuggestthattheyhavenotbeenisolatedforaverylongtime.In2012,thirtyHyphessobryconanisitsifishwerepur-chasedfromChalet-HéraultAquariophilie(Nuaille,France).RNAsamplesandRNA-seqInordertoidentifypolymorphismsatthepopulationlevelbasedonaPool-seqapproach[69],foreachpopulation,50to200embryos/larvaefromseveralindependentspawningeventsandatdifferentdevelopmentalstages(6hpost-fertilizationtotwoweekspost-fertilization)werepooledandtotalRNAisolated(Additionalfile1:FigureS10).EachRNAsamplewassequencedonanIlluminaHiSeq2000platform(2×100bppaired-end).Thepooledembryosam-pleshadbeenpreviouslysequencedusingtheSangerand454methods[80](Additionalfile1:FigureS10).TranscriptomeassemblyandannotationTheAstyanaxmexicanustranscriptomewasassembledwithNewblerver.2.8(Roche454)sequenceanalysissoftwareusing454sequences(2.106reads)ofboththePachóncaveandsurfacefishpooledembryos.Weobtained33,400contigs(meancontiglength=824bp).WealsotriedtogenerateatranscriptomeassemblyusingtheIlluminasequences,butwhereasthisresultedinmorecontigs(49,728)thanthe454sequences,manyofthemwereconcatenationsofdifferenttranscriptsandinsomecasesthesametranscriptwasfoundinmorethanonecontig.WethereforemappedtheIlluminasequencesontothe454contigstoidentifyandannotateSNPs.Putativecodingsequencesineachcontigwereidentifiedusingthezebrafish(Zv9)proteomeavailableatEnsEMBL73asareference[81].Acontigwasconsideredproteincodingifthee-valueforthebesthitwas<105.Wefound13,240proteincodingcontigs(contigmeanlength=530bp).Weidentifiedcontigscontainingdomainsthatmatcheddifferentzebrafishproteinsandwhichweremostlikelychimericcontigs.Thesecontigswereremoved(369,i.e.3%oftheproteincodingcontigs).Intotal,weanalyzed12,871putativeproteincodingcontigs.SNPidentificationandannotationIlluminasequenceswerealignedtocontigswithBWA[82]usingthedefaultparametersforpaired-endreads.Hyphes-sobryconanisitsisequenceswerealignedtoAstyanaxcon-tigsusingalowermaximumeditdistance(n=0.001).SNPscallingwasperformedusingGATKUnifiedGen-otyperv2.4.9[83].BecausewefilteredSNPsafterdetec-tionusingdifferentparameterthresholdsdescribedbelow,weusedtheallowPotentiallyMisencodedQualsand–rfBadCigaroptions.Wedetected299,101SNPsincluding141,490SNPsinannotatedcontigs.Whenacompletecodingsequencewasidentified,i.e.fromthestartcodontothestopcodonandcorrespond-ingtoacompletezebrafishprotein,wecouldidentifythenon-codingflankingsequences(containing18,743SNPs),otherwiseonlythesequencematchingthecodingsequenceofthezebrafishwasannotatedascodingandtheflankingsequenceswerenotannotated.The55,950SNPsinthecodingsequenceswereannotatedassyn-onymousornon-synonymous,accordingtowhichaminoacidwascodedforbythealternativecodonsresultingfromtheSNP.Theancestralalleleandthederivedallelewereinferredaccordingtotheallelefoundintheout-groupHyphessobryconanisitsi(Fig.4).SNPsforwhichtheancestralalleleandderivedallelecouldnotbeidentified,eitherbecauseinHyphessobryconanisitsinosequencecouldbeidentifiedortherewasanotherallelepresentortheallelewaspolymorphic,werediscarded.SNPclassificationTheSNPsidentifiedinAstyanaxmexicanusSFandCFwereclassifiedintoeightclasses(Fig.4).ThenumberofSNPsinthedifferentclassesdependedonthethresholdsusedtoconsideraSNPasreliableandpolymorphicinFumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page15of19

PAGE 16

eachpopulation.Therationaleforthesetofthresholdsselectedisgivenbelow.Thepopulationsbeingcloselyrelated(theybelongtothesamespecies)andthemutationrateforaSNPoriginbeingverylow(~108),wewouldexpectthattheeighthclass(divergentpolymorphism)ofSNPswouldbeaveryrareoutcomebecauseitistheresultoftwoindependentmutationsatthesamesite,eitherintheancestralpopulationorintheCFandSFpopulations.WefoundonlyoneSNPsinthisclass(Table3).ItsuggeststhatIlluminasequencingdidnotgenerateanumberofsequencingerrorsthatwouldsignificantlyinflatethenumberofSNPsidentified.ParameterthresholdsforSNPselectionWeexaminedtheeffectofthethresholdsappliedtopa-rametersusedtodiscardSNPsbeforetheirclassificationandpopulationgenomicsanalyses.Firstwelookedattheeffectofsequencingdepth.Whereasthemeansequencingdepthwas820,thestand-arddeviationwasverylarge(9730).WhentheminimalnumberofreadsperpopulationataSNPsitewassetto100orhigher,therelativefrequenciesoftheeightSNPclasseswereverystable,indicatingthat100wasagoodcompromisebetweenthestabilityofthedistributionoftheSNPsintodifferentclassesandthenumberofSNPsdiscarded(Additionalfile1:FigureS11).Wethenconsideredtheeffectofthee-valueoftheblastbetweentheAstyanaxcontigandthezebrafishse-quenceusedforannotation,inordertodiscardpoorlyconservedsequencesthatweremisidentifiedasproteincoding.ItappearedthattheSNPclassificationwasstablewhicheverthethresholdwasused,i.e.e-value<105(Additionalfile1:FigureS11).WealsoexaminedtheeffectoftheintervalbetweenSNPs,becausewewouldexpectclustersofspuriousSNPsinpoorlysequencedregions.WetestedtheeffectofselectingSNPsinregionswithoutanyotherSNPs.Asexpected,therewasanexcessofsharedpolymorphisms(class7)withasmallwindowsize.Whenthethresholdwassetto>50bponeachsideoftheSNP,thedistribu-tionwasstable(Additionalfile1:FigureS11).Finally,weconsideredthatthelowestvalueofminorallelefrequency(MAF)inthelabpopulationsshouldbesetaround5%becausetheeffectivepopulationsizeinthelabislow.Alltheabovethresholds,apartfromthatforMAF,aretrade-offsbetweenqualityandquantityofthedata.ThelowestMAFvaluepossibleinthepooledembryosamplesdependsontheunknownnumberofparentsoftheembryos,andtheMAFthresholdof>5%couldthere-forebeconsideredarbitrary.Nevertheless,usingMAFthresholdsof1%,5%and10%weobtainedsimilarSNPclassfrequencies(Table3andAdditionalfile1:TableS2aandTableS2b).Theresultswerethusalsorobustaccordingtothisparameter,andtheuseofdifferentsetsofparametersledtosimilardistributionofSNPclassesthatledtothesameconclusion.Therefore,allanalysesinthispaperwereperformedusingthefollowingthresholds:MAF>5%;depth>100;e-value<105;SNPisolation>50bp.SimulationsoftheevolutionofneutralpolymorphismsintwopopulationsInordertoestimatetheageofthePachóncavepopula-tion,wecomparedthedistributionofSNPsintosevenclasses(thedivergentpolymorphismclasswasemptyandthusexcluded)definedabovewiththedistributionobtainedinsimulationsoftheevolutionaryprocess(Fig.3).MoredetailedontherationaleofthemethodisgiveninAdditionalfile4.Theevolutionarymodelanditsparame-tersarethoseofIMa2,i.e.effectivepopulationsizes,migra-tionratesanddivergencetime.Therangeofvaluestestedweredefinedaccordingobservationinthefield,estimationsfoundintheliteratureandtheresultsoftheanalysiswithIMa2.Thefullmodelisasfollowing:anancestralpopula-tionwithagivensize(10,000,20,000,50,000or100,000)andatmutation/driftequilibrium(whichdependsonthemutationrateandthepopulationsize)wassplitintotwopopulationsthatcouldhavedifferentsizes.Therecouldbemigrationsfromthesurfacetothecave.Thesmallestprobabilityofmigrationwassetto0.00001andthehighestto0.1peryear.Thesmallestpercentageofcavefishthataremigrantswhenamigrationoccurswassetto0andthehighestwassetto1%.Wealsotookintoaccountthatgen-eticdriftcouldhaveoccurredinthelaboratorystocks(ef-fectivepopulationsizeinthelabwassetto10andthenumberofgenerationwassetto10).Allmutationswereneutral(frequencychangesateachlocusweredrivenbygeneticdriftonly)andeachlocuswereevolvingindepend-ently.Foragivensetofparameters,eachtengenerationsaftertheisolationoftwopopulations,weestimatedthefre-quencyofSNPsineachcategoryandweestimatedascoreofgoodnessoffitwiththefrequenciesobtainedwiththerealSNPdataset.Weranthesimulationandthetestofgoodnessoffitwithdifferentsetsofparametersinordertoidentifythesetsofparameters,includingtheageofthePachóncavepopulation,thatresultedinSNPclassfre-quenciesthatfittedwellwithobservedfrequencies.TheprogramwaswritteninCandisavailableonGithubwithitsdocumentation(http://julienfumey.github.io/popsim).AdditionalfilesAdditionalfile1:Supplementaryfiguresandtables.(PDF1092kb)Additionalfile2:Fourtaxonunrootednucleargenephylogenies.(PDF101kb)Additionalfile3:Allelefrequencydistributionsat25microsatelliteloci.(PDF2321kb)Additionalfile4:Rationaleofdatingwithpool-seqSNPs.(PDF420kb)Fumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page16of19

PAGE 17

Abbreviationbp:basepair;BWA:Burrow-WheelerAlgorithm;CF:Cavefish;Cpdphotolyase:Cyclobutanepyrimidinedimerphotolyase;Cry1a:cryptochrome1A;GATK:GenomeAnalysisToolKit;HSP90:HeatShockProtein90;i.e.:idest.;IM:Isolationwithmigration;Lepb:signalpeptidaseI;Lepr:LeptinReceptor;MAF:minorallelefrequency;Mc1r:MelanoCortin1Receptor;Mc3r:MelanoCortin3Receptor;Mc4r:MelanoCortin4Receptor;MCMC:MetropolisCoupledMarkovChain;mtDNA:mitochondrialdesoxyribonucleicacid;Nc:Censuspopulationsize;Ne:Effectivepopulationsize;nuDNA:nucleardesoxyribonucleicacid;Oca2:OculocutaneousalbinismII;Per1:periodcircadianproteinhomolog1;Per2:periodcircadianproteinhomolog2;Pomcb:proopiomelanocortinb;Rag1:Recombinationactivatinggene1;RNA:Ribonucleicacid;SE:StandardError;SF:Surfacefish;SNP:SingleNucleotidePolymorphism;Spp.:speciespluralis;Tef1:translationelongationfactorEF-1alpha;USA:UnitedStatesofAmericaAcknowledgmentsThisworkhasbenefitedfromthefacilitiesandexpertiseofthehighthroughputsequencingplatformofI2BC.ConsentforpublicationsNotapplicableFundingThisworkwassupportedbyacollaborativegrantfromANR(AgenceNationaledelaRecherche)BLINDTEST(toSRandDC)andfromIDEEV(toDC).Thefundingbodieshadnoroleinthedesignofthestudyandcollection,analysis,andinterpretationofdataandinwritingthemanuscript.AvailabilityofdataandmaterialsMicrosatellitedataareavailablefrom[33].SNPsandtheirannotationsarestoredinaMySQLdatabaseandareavailableonlineathttp://ngspipelines.toulouse.inra.fr:9022.PerlandRscriptsforthedataanalysesandgraphicsareavailableuponrequest.Illuminaand454transcriptomicsequencesareavailableonArrayExpress:https://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-5142/.ScriptusedcanbefoundonGitHub:http://github.com/julienfumeyAuthors’contributionsDCandSRdesignedthestudy.JFwrotetheprogramofsimulation.DCandJFanalyzedthedata.DC,CT,SR,HHcollectedthedata.CNandJFgeneratedthedatabases.DCdraftedthemanuscript.Allauthorscontributedtothewritingofthemanuscript.Allauthorsreadandapprovedthefinalmanuscript.EthicsapprovalandconsenttoparticipateSR’sauthorizationforuseofanimalsinresearchisnumber91–116,andincludesa“Certificatdecapacitépourl’élevagedefaunesauvage”.ExperimentswereperformedaccordingtoParisCentre-SudEthicCommitteeauthorizationnumbers2012–0053and2012–0056.Fin-clipsfromwild-caughtanimalswerecollectedundertheauspicesofMexicanpermit02241/13deliveredtoSRbySecretariadeMedioAmbienteyRecursosNaturales.CompetinginterestsTheauthorsdeclarethattheyhavenocompetinginterest.Publisher’sNoteSpringerNatureremainsneutralwithregardtojurisdictionalclaimsinpublishedmapsandinstitutionalaffiliations.Authordetails1Évolution,Génomes,Comportement,Écologie,CNRS,IRD,UnivParis-Sud.UniversitéParis-Saclay,F-91198Gif-sur-Yvette,France.2InstituteforIntegrativeBiologyoftheCell(I2BC),CEA,CNRS,UniversitéParis-Sud,UMR9198,FRC3115,AvenuedelaTerrasse,Bâtiment24,Gif-sur-Yvette,F-91198Paris,France.3DECAgroup,Paris-SaclayInstituteofNeuroscience,UMR9197,CNRS,GifsurYvette,France.4PlateformeBioinformatiqueToulouse,Midi-Pyrénées,UBIA,INRA,AuzevilleCastanet-Tolosan,France.5UniversitéParisDiderot,SorbonneParisCité,Paris,France.Received:5December2017Accepted:19March2018 References1.MitchellRW,RussellWH,ElliottWR.Mexicaneyelesscharacinfishes,genusAstyanax:environment,distributionandevolution.SpecPublMusTexasTechnUniversity.1977;12:1–89.2.GrossJB.ThecomplexoriginofAstyanaxcavefish.BMCEvolBiol.2012;12:105.3.JefferyWR.RegressiveevolutioninAstyanaxcavefish.AnnuRevGenet.2009;43:25–47.4.JefferyWR.Emergingmodelsystemsinevo-devo:cavefishandmicroevolutionofdevelopment.EvolDev.2008;10(3):265–72.5.WilkensH,StreckerU.ConvergentevolutionofthecavefishAstyanax(Characidae,Teleostei):geneticevidencefromreducedeye-sizeandpigmentation.BiolJLinnSoc.2003;80(4):545–54.6.HüppopK.Oxygen-consumptionofAstyanaxfasciatus(Characidae,Pisces)-acomparisonofepigeanandhypogeanpopulations.EnvironBiolFish.1986;17(4):299–308.7.MoranD,SoftleyR,WarrantEJ.EyelessMexicancavefishsaveenergybyeliminatingthecircadianrhythminmetabolism.PLoSOne.2014;9(9):e107877.8.SalinK,VoituronY,MourinJ,HervantF.Cavecolonizationwithoutfastingcapacities:anexamplewiththefishAstyanaxfasciatusmexicanus.CompBiochemPhysiolAMolIntegrPhysiol.2010;156(4):451–7.9.HüppopK,WilkensH.BiggereggsinsubterraneanAstyanaxfasciatus(Characidae,Pisces)-theirsignificanceandgenetics.ZeitschriftFurZoologischeSystematikUndEvolutionsforschung.1991;29(4):280–8.10.TeykeT.MorphologicaldifferencesinneuromastsoftheblindcavefishAstyanaxhubbsiandthesightedriverfishAstyanaxmexicanus.BrainBehaviorandEvolution.1990;35(1):23–30.11.YoshizawaM,GorickiS,SoaresD,JefferyWR.Evolutionofabehavioralshiftmediatedbysuperficialneuromastshelpscavefishfindfoodindarkness.CurrBiol.2010;20(18):1631–6.12.YoshizawaM,YamamotoY,O'QuinKE,JefferyWR.Evolutionofanadaptivebehavioranditssensoryreceptorspromoteseyeregressioninblindcavefish.BMCBiol.2012;10:108.13.DubouéER,BorowskyRL,KeeneAC.Beta-adrenergicsignalingregulatesevolutionarilyderivedsleeplossintheMexicancavefish.BrainBehavEvol.2012;80(4):233–43.14.DubouéER,KeeneAC,BorowskyRL.Evolutionaryconvergenceonsleeplossincavefishpopulations.CurrBiol.2011;21(8):671–6.15.ElipotY,HinauxH,CallebertJ,RetauxS.Evolutionaryshiftfromfightingtoforaginginblindcavefishthroughchangesintheserotoninnetwork.CurrBiol.2013;23(1):1–10.16.VaratharasanN,CrollRP,Franz-OdendaalT.TastebuddevelopmentandpatterninginsightedandblindmorphsofAstyanaxmexicanus.DevDyn.2009;238(12):3056–64.17.YamamotoY,ByerlyMS,JackmanWR,JefferyWR.Pleiotropicfunctionsofembryonicsonichedgehogexpressionlinkjawandtastebudamplificationwitheyelossduringcavefishevolution.DevBiol.2009;330(1):200–11.18.BibliowiczJ,AlieA,EspinasaL,YoshizawaM,BlinM,HinauxH,LegendreL,PereS,RetauxS.DifferencesinchemosensoryresponsebetweeneyedandeyelessAstyanaxmexicanusoftheRioSubterraneocave.EvoDevo.2013;4(1):25.19.EspinasaL,BibliowiczJ,JefferyW,RetauxS.EnhancedpreycaptureskillsinAstyanaxcavefishlarvaeareindependentfromeyeloss.EvoDevo.2014;5(1):35.20.HüppopK.Foodfindingabilityincavefish(Astyanaxfasciatus).IntJSpeleol.1987;18:59–66.21.MayrE.Causeandeffectinbiology.Science.1961;134(3489):1501–6.22.CasaneD,RétauxS.EvolutionarygeneticsofthecavefishAstyanaxmexicanus.In:NicholasSF,editor.AdvancesinGenetics,vol.95.Cambridge:AcademicPress;2016.p.117–59.23.MaL,JefferyWR,EssnerJJ,KowalkoJE.GenomeeditingusingTALENsinblindMexicancavefish,Astyanaxmexicanus.PLoSOne.2015;10(3):e0119370.24.McGaughSE,GrossJB,AkenB,BlinM,BorowskyR,ChalopinD,HinauxH,JefferyWR,KeeneA,MaL,etal.Thecavefishgenomerevealscandidategenesforeyeloss.NatCommun.2014;5:5307.25.O'QuinKE,YoshizawaM,DoshiP,JefferyWR.QuantitativegeneticanalysisofretinaldegenerationintheblindcavefishAstyanaxmexicanus.PLoSOne.2013;8(2):e57281.26.HinauxH,BlinM,FumeyJ,LegendreL,HeuzeA,CasaneD,RetauxS.LensdefectsinAstyanaxmexicanuscavefish:evolutionofCrystallinsandaroleforalphaA-Crystallin.DevelopmentalNeurobiology.2015;75(5):505–21.Fumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page17of19

PAGE 18

27.JefferyWR.Pleiotropyandeyedegenerationincavefish.Heredity.2010;105(5):495–6.28.WilkensH.Genes,modulesandtheevolutionofcavefish.Heredity.2010;105(5):413–22.29.BorowskyR.EyeregressioninblindAstyanaxcavefishmayfacilitatetheevolutionofanadaptivebehavioranditssensoryreceptors.BMCBiol.2013;11(1):81.30.GrossJB,PowersAK,DavisEM,KaplanSA.ApleiotropicinteractionbetweenvisionlossandhypermelanisminAstyanaxmexicanuscavexsurfacehybrids.BMCEvolBiol.2016;16(1):1–16.31.RetauxS,CasaneD.Evolutionofeyedevelopmentinthedarknessofcaves:adaptation,drift,orboth?EvoDevo.2013;4(1):26.32.AviseJC,SelanderRK.Evolutionarygeneticsofcave-dwellingfishesofgenusAstyanax.Evolution.1972;26(1):1–19.33.BradicM,BeerliP,Garcia-deLeonFJ,Esquivel-BobadillaS,BorowskyRL.GeneflowandpopulationstructureintheMexicanblindcavefishcomplex(Astyanaxmexicanus).BMCEvolBiol.2012;12:9.34.HausdorfB,WilkensH,StreckerU.PopulationgeneticpatternsrevealedbymicrosatellitedatachallengethemitochondrialDNAbasedtaxonomyofAstyanaxinMexico(Characidae,Teleostei).MolPhylogenetEvol.2011;60(1):89–97.35.PanaramK,BorowskyR.GeneflowandgeneticvariabilityincaveandsurfacepopulationsoftheMexicantetra,Astyanaxmexicanus(Telcostei:Characidae).Copeia.2005;(2):409–16.36.StreckerU,BernatchezL,WilkensH.GeneticdivergencebetweencaveandsurfacepopulationsofAstyanaxinMexico(Characidae,Teleostei).MolEcol.2003;12(3):699–710.37.StreckerU,FaundezVH,WilkensH.PhylogeographyofsurfaceandcaveAstyanax(Teleostei)fromcentralandNorthAmericabasedoncytochromebsequencedata.MolPhylogenetEvol.2004;33(2):469–81.38.StreckerU,HausdorfB,WilkensH.ParallelspeciationinAstyanaxcavefish(Teleostei)innorthernMexico.MolPhylogenetEvol.2012;62(1):62–70.39.DowlingTE,MartasianDP,JefferyWR.Evidenceformultiplegeneticformswithsimilareyelessphenotypesintheblindcavefish,Astyanaxmexicanus.MolBiolEvol.2002;19(4):446–55.40.ChakrabortyR,NeiM.Dynamicsofgenedifferentiationbetweenincompletelyisolatedpopulationsofunequalsizes.TheorPopulBiol.1974;5(3):460–9.41.NeiM,RoychoudhuryAK.Samplingvariancesofheterozygosityandgeneticdistance.Genetics.1974;76(2):379–90.42.Ornelas-GarcíaCP,Domínguez-DomínguezO,DoadrioI.EvolutionaryhistoryofthefishgenusAstyanaxBaird&Girard(1854)(Actinopterygii,Characidae)inMesoamericarevealsmultiplemorphologicalhomoplasies.BMCEvolBiol.2008;8(1):1–17.43.CoghillLM,DarrinHulseyC,Chaves-CamposJ,GarcíadeLeonFJ,JohnsonSG.NextgenerationphylogeographyofcaveandsurfaceAstyanaxmexicanus.MolPhylogenetEvol.2014;79(0):368–74.44.YokoyamaR,YokoyamaS.Convergentevolutionofthered-andgreen-likevisualpigmentgenesinfish,Astyanaxfasciatus,andhuman.ProcNatlAcadSciUSA.1990;87(23):9315–8.45.YokoyamaR,YokoyamaS.MolecularcharacterizationofabluevisualpigmentgeneinthefishAstyanaxfasciatus.FEBSLett.1993;334(1):27–31.46.YokoyamaS,MeanyA,WilkensH,YokoyamaR.Initialmutationalstepstowardlossofopsingenefunctionincavefish.MolBiolEvol.1995;12(4):527–32.47.LiW-H,NeiM.Persistenceofcommonallelesintworelatedpopulationsorspecies.Genetics.1977;86(4):901–14.48.CavallariN,FrigatoE,ValloneD,FroehlichN,FernandoLopez-OlmedaJ,FoaA,BertiR,JavierSanchez-VazquezF,BertolucciC,FoulkesNS.Ablindcircadianclockincavefishrevealsthatopsinsmediateperipheralclockphotoreception.PLoSBiol.2011;9(9):e1001142.49.YangJ,ChenX,BaiJ,FangD,QiuY,JiangW,YuanH,BianC,LuJ,HeS,etal.TheSinocyclocheiluscavefishgenomeprovidesinsightsintocaveadaptation.BMCBiol.2016;14(1):1–13.50.NiemillerML,FitzpatrickBM,ShahP,SchmitzL,NearTJ.Evidenceforrepeatedlossofselectiveconstraintinrhodopsinofamblyopsidcavefishes(teleostei:amblyopsidae).Evolution.2013;67(3):732–48.51.AspirasAC,RohnerN,MartineauB,BorowskyRL,TabinCJ.Melanocortin4receptormutationscontributetotheadaptationofcavefishtonutrient-poorconditions.ProcNatlAcadSci.2015;112(31):9668–73.52.BealeA,GuibalC,TamaiTK,KlotzL,CowenS,PeyricE,ReynosoVH,YamamotoY,WhitmoreD.CircadianrhythmsinMexicanblindcavefishAstyanaxmexicanusinthelabandinthefield.NatCommun.2013;4:2769.53.HeyJ.Isolationwithmigrationmodelsformorethantwopopulations.MolBiolEvol.2010;27(4):905–20.54.YueGH,DavidL,OrbanL.Mutationrateandpatternofmicrosatellitesincommoncarp(CyprinuscarpioL.).Genetica.2007;129(3):329–31.55.O’reillyPT,HerbingerC,WrightJM.AnalysisofparentagedeterminationinAtlanticsalmon(Salmosalar)usingmicrosatellites.AnimGenet.1998;29(5):363–70.56.SimonV,ElleboodeR,MahéK,LegendreL,Ornelas-GarciaP,EspinasaL,RétauxS.ComparinggrowthinsurfaceandcavemorphsofthespeciesAstyanaxmexicanus:insightsfromscales.EvoDevo.2017;8(1):23.57.HobanS,BertorelleG,GaggiottiOE.Computersimulations:toolsforpopulationandevolutionarygenetics.NatRevGenet.2012;13(2):110–22.58.HeyJ,NielsenR.Multilocusmethodsforestimatingpopulationsizes,migrationratesanddivergencetime,withapplicationstothedivergenceofDrosophilapseudoobscuraandD-persimilis.Genetics.2004;167(2):747–60.59.EspinasaL,BorowskyRB.OriginsandrelationshipofcavepopulationsoftheblindMexicantetra,Astyanaxfasciatus,inthesierradeElAbra.EnvironBiolFish.2001;62(1–3):233–7.60.RunemarkA,HeyJ,HanssonB,SvenssonEI.Vicariancedivergenceandgeneflowamongisletpopulationsofanendemiclizard.MolEcol.2012;21(1):117–29.61.HedgecockD.Doesvarianceinreproductivesuccesslimiteffectivepopulationsizesofmarineorganisms?In:BeaumontAR,editor.Geneticsandevolutionofaquaticorganisms.London:Chapman&Hall;1994.p.122–34.62.CrowJF,KimuraM.Anintroductiontopopulationgeneticstheory.NewYork:Harper&Row;1970.63.BeerliP,FelsensteinJ.Maximumlikelihoodestimationofamigrationmatrixandeffectivepopulationsizesinnsubpopulationsbyusingacoalescentapproach.ProcNatlAcadSciUSA.2001;98(8):4563–8.64.LangeckerTG,WilkensH,JungeP.IntrogressivehybridizationinthePachoncavepopulationofAstyanaxfasciatus(Teleostei:Characidae).IchthyologicalExplorationofFreshwaters.1991;2:209–12.65.BeaumontMA.ApproximateBayesiancomputationinevolutionandecology.AnnuRevEcolEvolSyst.2010;41(1):379–406.66.PinhoC,HeyJ.Divergencewithgeneflow:modelsanddata.In:FutuymaDJ,ShaferHB,SimberloffD,editors.AnnualReviewofEcology,Evolution,andSystematics,Vol41,vol.41;2010.p.215–30.67.KimuraM.Evolutionaryrateatmolecularlevel.Nature.1968;217(5129):624–6.68.KimuraM,OhtaT.Averagenumberofgenerationsuntilfixationofamutantgeneinafinitepopulation.Genetics.1969;61(3):763–71.69.SchlottererC,ToblerR,KoflerR,NolteV.Sequencingpoolsofindividuals-mininggenome-widepolymorphismdatawithoutbigfunding.NatRevGenet.2014;15(11):749–63.70.WinemillerKO.Patternsofvariationinlife-historyamongsouth-Americanfishesinseasonalenvironments.Oecologia.1989;81(2):225–41.71.ElipotY,HinauxH,CallebertJ,LaunayJ-M,BlinM,RétauxS.AmutationintheenzymemonoamineoxidaseexplainspartoftheAstyanaxcavefishbehaviouralsyndrome.NatCommun.2014;5:3647.72.RohnerN,JaroszDF,KowalkoJE,YoshizawaM,JefferyWR,BorowskyRL,LindquistS,TabinCJ.Crypticvariationinmorphologicalevolution:HSP90asacapacitorforlossofeyesincavefish.Science.2013;342(6164):1372–5.73.BradicM,TeotónioH,BorowskyRL.ThepopulationgenomicsofrepeatedevolutionintheblindcavefishAstyanaxmexicanus.MolBiolEvol.2013;30(11):2383–400.74.HohenlohePA,BasshamS,EtterPD,StifflerN,JohnsonEA,CreskoWA.PopulationgenomicsofparalleladaptationinThreespinesticklebackusingsequencedRADtags.PLoSGenet.2010;6(2):e1000862.75.RenautS,NolteAW,RogersSM,DeromeN,BernatchezL.SNPsignaturesofselectiononstandinggeneticvariationandtheirassociationwithadaptivephenotypesalonggradientsofecologicalspeciationinlakewhitefishspeciespairs(Coregonusspp.).MolEcol.2011;20(3):545–59.76.JohnsonTC,ScholzCA,TalbotMR,KeltsK,RickettsRD,NgobiG,BeuningK,SsemmandaI,McGillJW.LatepleistocenedesiccationofLakeVictoriaandrapidevolutionofcichlidfishes.Science.1996;273(5278):1091–3.77.BrawandD,WagnerCE,LiYI,MalinskyM,KellerI,FanSH,SimakovO,NgAY,LimZW,BezaultE,etal.ThegenomicsubstrateforadaptiveradiationinAfricancichlidfish.Nature.2014;513(7518):375–81.78.Behrmann-GodelJ,NolteAW,KreiselmaierJ,BerkaR,FreyhofJ.ThefirstEuropeancavefish.CurrentBiology.27(7):R257–8.79.PaabyAB,RockmanMV.Crypticgeneticvariation:evolution'shiddensubstrate.NatRevGenet.2014;15(4):247–58.Fumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page18of19

PAGE 19

80.HinauxH,PoulainJ,DaSilvaC,NoirotC,JefferyWR,CasaneD,RetauxS.DenovosequencingofAstyanaxmexicanussurfacefishandPachoncavefishtranscriptomesrevealsenrichmentofmutationsincavefishputativeeyegenes.PLoSOne.2013;8(1):e53553.81.FlicekP,AhmedI,AmodeMR,BarrellD,BealK,BrentS,Carvalho-SilvaD,ClaphamP,CoatesG,FairleyS,etal.Ensembl2013.NucleicAcidsRes.2013;41(D1):D48–55.82.LiH,DurbinR.Fastandaccurateshortreadalignmentwithburrows-wheelertransform.Bioinformatics.2009;25(14):1754–60.83.McKennaA,HannaM,BanksE,SivachenkoA,CibulskisK,KernytskyA,GarimellaK,AltshulerD,GabrielS,DalyM,etal.ThegenomeanalysisToolkit:aMapReduceframeworkforanalyzingnext-generationDNAsequencingdata.GenomeRes.2010;20(9):1297–303. • We accept pre-submission inquiries  Our selector tool helps you to Þnd the most relevant journal We provide round the clock customer support  Convenient online submission Thorough peer review Inclusion in PubMed and all major indexing services  Maximum visibility for your researchSubmit your manuscript atwww.biomedcentral.com/submitSubmit your next manuscript to BioMed Central and we will help you at every step: Fumeyetal.BMCEvolutionaryBiology (2018) 18:43 Page19of19


printinsert_linkshareget_appmore_horiz

Download Options

close


  • info Info

    There are only PDFs associated with this resource.

  • link PDF(s)



Cite this item close

APA

Cras ut cursus ante, a fringilla nunc. Mauris lorem nunc, cursus sit amet enim ac, vehicula vestibulum mi. Mauris viverra nisl vel enim faucibus porta. Praesent sit amet ornare diam, non finibus nulla.

MLA

Cras efficitur magna et sapien varius, luctus ullamcorper dolor convallis. Orci varius natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Fusce sit amet justo ut erat laoreet congue sed a ante.

CHICAGO

Phasellus ornare in augue eu imperdiet. Donec malesuada sapien ante, at vehicula orci tempor molestie. Proin vitae urna elit. Pellentesque vitae nisi et diam euismod malesuada aliquet non erat.

WIKIPEDIA

Nunc fringilla dolor ut dictum placerat. Proin ac neque rutrum, consectetur ligula id, laoreet ligula. Nulla lorem massa, consectetur vitae consequat in, lobortis at dolor. Nunc sed leo odio.