A Multistage Testing Strategy for Detection of Quantitative Trait Loci Affe Disease Resistance in At

Copyright2004bytheGeneticsSocietyofAmericaDOI:10.1534/genetics.103.013227AMultistageTestingStrategyforDetectionofQuantitativeTraitLociAffectingDiseaseResistanceinAtlanticSalmonThomasMoen,*,1KjerstiT.Fjalestad,*HegeMunck*andLuisGomez-Raya†*AKVAFORSK(InstituteofAquacultureResearch),N-1432Aas,Norwayand†`AreadeProduccio´Animal,CentreUdL-IRTA,25198Lleida,SpainManuscriptreceivedJanuary10,2003AcceptedforpublicationFebruary1,2004ABSTRACTAmultistagetestingstrategytodetectQTLforresistancetoinfectioussalmonanemia(ISA)inAtlanticsalmonisproposed.First,genotypingofamplifiedfragmentlengthpolymorphisms(AFLP)andatransmis-siondisequilibriumtest(TDT)werecarriedoutusingdeadoffspringfromadiseaseresistancechallengetest.Second,AFLPgenotypingamongsurvivorsfollowedbyaMendeliansegregationtestwasperformed.Third,within-familysurvivalanalysesusingalloffspringweredevelopedandappliedtosignificantTDTmarkerswithMendelianinheritance.Maximum-likelihoodmethodologywasdevelopedforTDTwithdominantmarkerstoexploitlinkagedisequilibriumwithinfamilies.Thestrategywastestedwithtwofull-sibfamiliesofAtlanticsalmonsiredbythesamemaleandconsistingof79offspringintotal.Alldeadoffspringfromthetwofamiliesweretypedfor64primercombinations,resultingin340scoredmarkers.Therewere26significantresultsoutof401TDTsusingdeadoffspring.Inthesecondstage,only17markerfamiliesshowedMendeliansegregationandweretestedinsurvivalanalysis.ApermutationtestwasperformedforallsurvivalanalysestocomputeexperimentwiseP-values.Twomarkers,aaccac356andagccta150,weresignificantatPϽ0.05whenaccountingformultipletestinginthesurvivalanalyses.Theproposedstrategymightbemorepowerfulthancurrentmappingstrategiesbecauseitreducesthenumberofteststobeperformedinthelasttestingstage.HEspreadofdiseasesisamajorprobleminaqua-cultureproduction,anddeterminingthegeneticarchitectureofdiseaseresistancetraitsisofgreatinter-esttogeneticistsworkingonaquaculturespecies.Map-pingofquantitativetraitloci(QTL)fordiseaseresis-tanceisoneapproachthatcanprovideinformationonthelocationandeffectsofthegenesinfluencingthetrait.QTLcanbeidentifiedthroughagenomescaninwhichthesegregationofalargenumberofmarkers,distributedoverthewholegenome,istestedforassocia-tionswiththerecordedphenotypes.Ideally,themarkersshouldbecodominantandhavehighheterozygosity.MicrosatellitemarkersfulfillboththesecriteriaandareregardedbymanyasthemarkerofchoiceforQTLresearch(e.g.,Jacksonetal.1998;Ozakietal.2001).Thedevelopmentofmicrosatellitesis,however,are-source-demandingprocess,andformanyspecies,high-densitymicrosatellitemapsarenotavailable.Amongtheaquaculturespecies,microsatellitemapshavebeenpublishedonrainbowtrout(Youngetal.1998;Saka-motoetal.2000;Nicholsetal.2003),catfish(Wald-bieseretal.2001),tilapia(Kocheretal.1998;Agrestietal.2000),andJapaneseflounder(Coimbraetal.2003),butnotonAtlanticsalmon.Correspondingauthor:AKVAFORSK,Box5010,N-1432Aas,Norway.E-mail:thomas.moen@akvaforsk.nlh.noGenetics167:851–858(June2004)1TWhenageneticmapisnotavailable,analternativestrategyforagenomescanistousemarkersthatcanbedevelopedwithoutpriorknowledgeoftheDNAse-quenceinthemarkerregion.Amongthesemarkers,amplifiedfragmentlengthpolymorphisms(AFLP)markersarethemostwidelyused.AFLPmarkers(Vosetal.1995)areknowntobehighlyreproducible(Jonesetal.1997;Bagleyetal.2001),andseveralmarkerscanbegenotypedatatime.AFLPs,however,aredominantmarkers,meaningthatoneofthehomozygousgeno-typescannotbedistinguishedfromtheheterozygousgenotypewithouttheuseofspecialequipment/software(Castiglionietal.1999;PiephoandKoch2000).ThedominantnatureofAFLPmarkersmakestheiruseinmappingexperimentsmoredifficult.QTLmappingfordiseaseresistanceinfishtypicallyreliesontheuseofchallengetestsinwhichoffspringgroupsareexposedtothedisease.Thesurvivaltimesarerecorded,oreachanimalisclassifiedsimplyassus-ceptibleorresistant.Thedatacanbeanalyzedbydiffer-entmethods.Weproposeinthisarticleatestingstrategybasedontwostatisticalmethodsusedinresearchonhumans.Thetransmissiondisequilibriumtest(TDT;Spielmanetal.1993)comparesthenumberoftimesthatamarkeralleleistransmittedornottransmittedfromaheterozygousparenttoanaffectedoffspring,andthereforeonlytheaffectedoffspringareconsid-852T.Moenetal.ered.TheTDThasbeenappliedonlytocodominantmarkerssuchasmicrosatellites,becausethetestrequiresthatallelescanbetracedfromparentstooffspring.Survivalanalysis(Altman1991)hasbeenwidelyusedinhumanmedicinefortestingofmedicaltreatments.SurvivalanalysisisexpectedtobemorepowerfulthantheTDT,sincethevariableusedisthesurvivaltimeofeachanimalinsteadofthecategoricalcharacteristics“susceptible/resistant.”However,itismorelaborde-manding,sinceitrequiresthegenotypingofbothsus-ceptibleandresistantanimals.AcommonprobleminQTLmappingexperimentsattemptingalargecoverageofthegenomeisreducedpowerofdetectionduetomultipletesting;i.e.,whenmanymarkersaretested,sometestsareexpectedtogivesignificantresultsjustbychance.WeproposeinthisarticleastrategyfordetectionofQTLthattakesintoaccountboththemultipletestingproblemandthepossiblelackofageneticmap.Thetestingstagesofthisschemeare(1)affectedfishfromadisease-resistancechallengetestaregenotypedforAFLPmarkersandthedataareanalyzedwiththeTDT;(2)resistantfisharegenotypedforsignificantmarkersfromstage1,andatestforMendeliansegregationisperformed;and(3)allTDT-significantmarkersthatfollowMendeliansegre-gationareanalyzedusingsurvivalanalysis.Thestrategymakesitpossibletoperformagenomescaninarela-tivelyfastandcost-effectiveway.Furthermore,itreducesthemultipletestingproblembecauseonlyasubsetofmarkersistakentothefinalteststage.Objectivesneededtomakethestrategyfeasiblewere(a)todevelopamaximum-likelihoodmethodtoallowtransmissiondisequilibriumtestingusingdominantmarkersand(b)todevelopwithin-familysurvivalanalysisforQTLdetec-tionwithdominantmarkers.ThestrategyisageneralapproachtodetectQTLinspecieswithnoavailablemaps.However,wehavedevelopedandappliedthestrategytodetectQTLforresistanceagainstinfectioussalmonanemia(ISA)inAtlanticsalmon.ISAcausessevereproblemsintheAt-lanticsalmonfarmingindustryinNorway,Canada,Scot-land,andtheUnitedStates(ThorudandDjupviketal.1988;Mullinsetal.1998;Rodgeretal.1998;Bouch-ardetal.2001).Thecausalagentbelongstotheortho-myxoviridaefamilyofRNAviruses(Falketal.1997),andtypicalpathologicalchangesincludesevereanemia,leucopenia,ascites,andhemorrhagiclivernecrosis(ThorudandDjupviketal.1988;Evensenetal.1991).Theaccumulatedmortalityofanattackrangesfrom15to100%(Thorud1991).Atlanticsalmonistheonlyspeciesknowntodevelopthedisease,althoughthevirusmaysurviveandreplicateinothersalmonidsaswell(Nylundetal.1997).MATERIALSANDMETHODScameChallengefromthetest1995forclassresistanceofthebreedingtoISA:ThepopulationgeneticmaterialofAqua-GenoffamiliespaternalA/S,Hemne,Norway.Thebreedingpopulationconsistson(Refstiehalf-sib1990).families,eachcomprisedoftwofull-sibpartaroutineChallengetestsforISAarecarriedoutchallengeofthebasisatVESOVikanAkvaVet,Namsos,Norway,astaltransponderconditionstest,procedureandthefishforbroodstockselection.Priortotheindividuallywerekepttaggedunderstandardenvironmen-feeding.Aftertags.8daysTheoftestacclimatizationwasperformedwithat8passivemonthsintegratedthetestafterfacilities,firstthefishwereintraperitoneallyinjectedwith0.2-mldosesoftheinfectiousmaterialdilutedin0.85%PBS.Thechallengewasdoneinfreshwaterwithanapproximatewatertempera-tureof12Њ.Mortalitywasmonitoreduntil50%oftheoffspringoverallweredeadandthentheexperimentwasterminated.Thecauseofdeathwasverifiedbybacteriologicaltestsof10%ofthefishusingblood-agarplates.Plateswithandwithout2%NaClwereused,sinceotherpotentialpathogenscouldhavearangeofsaltpreferences.DeadfishwerecollectedeachdayandstoredatϪ20ЊforlaterDNAanalyses.Afterterminationofthetest,thesurvivorswerekilledandstoredatϪ20Њ.AllprocedureswereaccordingtostandardoperationproceduresatVESOVikanAkvaVet,whichisoperatedincom-pliancewiththeOECDprinciplesofGoodLaboratoryPracticewithregardtoEuropeanUnionCouncilDirective88/320/EEC.Fromthelargersetoffamiliestested,tissuewascollectedfromparentsandalloffspringfrom30randomlyselectedpaternalhalf-sibfamiliestobeusedinthisstudy.Fromamongthese30families,1half-sibfamily,composedof2full-sibfamiliesof40and39offspring,waschosenonthebasisof(1)thenumberofinformativeAFLPmarkersobservedinthethreeparentsand(2)thedistributionofsusceptible/resistantindividualswithineachfull-sibfamily(1:1ratioconsideredoptimal).DNAextraction:DNAwasextractedusingtheDNeasykit(96-wellformat)fromQIAGEN(Chatsworth,CA).TheDNAconcentrationwasestimatedfromtheabsorbanceat260nm(A260),measuredonaGeneQuantIIspectrophotometer(AmershamPharmacia).ThequalityoftheDNAwasinspectedusingagarosegelelectrophoresisandtheAVosetal.1995)260/A280ratio.AFLPgenotyping:TheAFLP(genotypingwasdoneusingtheAFLPAnalysisSystemIandtheAFLPstarterprimerkitfromLifeTechnologiesandaccordingtothe␥-33kitmanual.TheEcoRIprimerswereendlabeledwithP.DNAfragmentswereseparatedon6%polyacrylamidegels(19:1acrylamide:bis;7.5murea,1ϫTBEbuffer).Theelectrophoresiswasrunat60Wconstantpower.Onlymarkersthatwerereproducibleandhadafragmentsizeof50–600bpwerescored.TheAFLPswerescoredasdominantmarkers.Thegenotypesofparentswiththeband-presentphenotypeforagivenmarkerwereinferredfromtheAFLPphenotypesofoffspring.TheAFLPmarkerswerenamedbythethreeselectivenucleotidesoftheEcoRIprimer,followedbythethreeselectivenucleotidesoftheMseIprimer,withthefragmentlength(inbasepairs)attheend.Multistagetestingstrategy:Thestrategywasbasedondis-cardingmarkersinthefollowingsequentialtests:transmissiondisequilibriumtest,Mendeliansegregationtest,andQTLde-tectionusingsurvivalanalysiswithinfamilies(Figure1).Transmissiondisequilibriumtest:AnapplicationoftheTDTwasdevelopedtodealwithdominantmarkersandthefamilystructurecurrentlyusedinchallengetestsbytheNorwegiansalmonindustry(onesirematedtotwodams).TheformoftheTDTdependedontheparentalϭ(TϪgenotypes.NT)2ThestandardTDTisaMcNemartest:␹2/(TϩNT),whereTisthenumberoftimesthatthealleleinquestionwastrans-mittedfromaheterozygousparenttoanaffectedoffspring,andNTisthenumberoftimesthattheallelewasnottransmit-ted(fromaheterozygousparentcarryingthealleletoanaffectedoffspring).TheMcNemartestwasperformedforAQTLMappingStrategy853Figure1.—ExperimentdesignforthechallengetestforISAresistanceinAtlanticsalmon.sireandbothdamswereheterozygous.InTable1,vs,vd1,vd2arethetransmissionparametersfromthesire,dam1,anddam2,respectively.Atransmissionparameteristheprobabilityoftransmissionofoneoftheparentalallelestoitsoffspring.Underthenullhypothesisvsϭ0.5,vd1ϭ0.5,andvd2ϭ0.5;K1andK2areconstants.Thelikelihoodequationcorrespond-ingtothecaseinwhichthesireandbothdamswereheterozy-gouscanbefoundinthefirstrowofTable2.Atestforthetransmissiondisequilibriumbasedonthelikelihoodratiostatisticswasperformedbylikelihoodratiotest(LRT)فϪ2lnL(0.5,0.5,0.5)/L(vs,vd1,vd2)with2d.f.Similarequationsandtestswerederivedfortheothersituations(Table2).Mendeliansegregationtest:AtestfortheMendeliansegrega-tionwascarriedoutforeachmarkerthatturnedoutsignificantintheTDT.Theexpectedsegregationratiosweretestedusingchi-squaregoodnessoffit.Thesignificancelevelsusedforthetestwere0.10,0.05,and0.01.Markerssignificantatanyofthoselevelswerenotusedinthefollowingtestingstage.Theuseofthresholdat0.10wasdonetoassurethatmarkersinthesurvivalanalysesweresegregatinginaMendelianfashion.QTLdetectionusingsurvivalanalysiswithinfamilies:SurvivalanalysisutilizesmoreinformationthanTDTdoes,sincethevariableusedisthenumberofdayseachfishsurvived.Also,itconsidersbothsusceptibleandresistantfish.Allfishsharedthesameenvironmentandasurvivalanalysiswithinfamilieswasthereforeappropriate.Thetwogroupstobecomparedwere(1)offspringwiththeband-presentmarkerphenotypewithinafull-sibfamilyand(2)offspringwiththeband-absentmarkerphenotypewithinthesamefull-sibfamily.Hypothesistestingwascarriedoutusingalog-ranktest,LRANKϭ(OaϪEa)2/Eaϩ(OpϪEp)2/Ep,whereEaϭ͚tiϭ1Eai,Epϭ͚tiϭ1Epi,whereEaiϭdi(rai/ri);Epiϭdi(rpi/ri);diisthetotalnumberofcases(fishthatdied)duringdayi;riisthetotalnumberoffishatrisk(stillalive)atthebeginningofdayi;raiisthenumberoffishatriskintheband-absentgroupatthebeginningofdayi;rpiisthenumberoffishatriskintheband-presentgroupatthebeginningofdayi;OaandOparethetotalnumberofcasesintheband-absentandband-presentgroups,respectively(EaandEparethecorre-spondingexpectedcounts).LRANKisdistributedasa␹2distri-butionwith1d.f.Kaplan-Meiersurvivalcurveswerecon-structedforbothgroupsbycomputingthecumulativeproportionofoffspringwithinthegroupstillaliveatanytime;Sa(ti)ϭSa(tiϪ1)(raiϪdai)/raifortheband-absentgroup,andSp(ti)ϭSp(tiϪ1)(rpiϪdpi)/apifortheband-presentgroup.Thedifferencebetweenthesurvivalcurvesforthetwogroupswasestimatedbythehazardratio:markershavingoneofthefollowingparentalmarkergenotypeconfigurations:(1)sireheterozygous(Aa)andbothdamshomozygousforthenullallele(aa;thetestwasthendoneonthehalf-sibfamily);(2)sireAa,oneofthedamsaaandtheotherdamhomozygousforthebandallele(AA;thetestwasthendoneonthefull-sibfamilywithmixedband-present/band-absentoffspring);(3)sireaaandoneortwodamsAa(thetestwasthendoneonthefull-sibfamily/familieswithmixedband-present/band-absentoffspring).McNemar’stestapproximatelyfollowsachi-squaredistributionwith1d.f.Fortheothermarkers,oneorbothfull-sibgroupshadtwoheterozygousparents.Maximum-likelihoodmethodsweredevelopedtodetectdistortedsegregationofallelesamongtheaffectedoffspring.Table1liststheexpectedgenotypefrequenciesusedtoconstructthelikelihoodequationwhenTABLE1Theexpectedgenotypefrequencieswithinfull-sibfamilywhensireandbothdamswereheterozygousAllelefromsireAAaaAAaaAllelefromdam1AAAA————Allelefromdam2————AaAaExpectedgenotypefrequencyvsvd1vs(1Ϫvd1)(1Ϫvs)vd1(1Ϫvs)(1Ϫvd1)vsvd2vs(1Ϫvd2)(1Ϫvs)vd2(1Ϫvs)(1Ϫvd2)GenotypeAAAaAAAaAAAaAAAaPhenotypeBandBandBandBandBandBandBandBandpresentpresentpresentabsentpresentpresentpresentabsentvs,vd1,andvd2arethetransmissionparametersfromthesireanddam1and2,respectively.854T.Moenetal.TABLE2LikelihoodandMcNemarequationsfortransmissiondisequilibriumtestingforthepossiblecombinationsofparentalgenotypesforonesirematedtotwodamsParentalgenotypesDegreesoffreedom211111SireAaAaAaAaAaaaDam1AaAaAaaaaaAaDam2AaaaAAaaAA—Likelihoodor␹equationsL(vs,vd1,vd2)ϭK1((1Ϫvs)(1Ϫvd1))na1(1Ϫ(1Ϫvs)(1Ϫvd1))np1ϫK2((1Ϫvs)(1Ϫvd2))na2(1Ϫ(1Ϫvs)(1Ϫvd2))np1L(vs,vd1)ϭK1((1Ϫvs)(1Ϫvd1))na1(1Ϫ(1Ϫvs)(1Ϫvd1))np1p2ϫK2(1Ϫvs)na2vnsL(vs,vd1)ϭK1((1Ϫvs)(1Ϫvd1))na1(1Ϫ(1Ϫvs)(1Ϫvd1))np1␹2ϭ(np1ϩnp2Ϫna1Ϫna2)2/(np1ϩnp2ϩna1ϩna2)␹2ϭ(np1Ϫna1)2/(np1ϩna1)␹2ϭ(np1Ϫna1)2/(np1ϩna1)2K1andK2areconstants;vs,vd1,andvd2arethetransmissionparametersfromthesire,dam1,anddam2,respectively;na1andna2arethenumbersofaffectedoffspringhavingtheband-absentmarkerphenotypeinfamilies1and2,respectively;np1andnp2arethenumbersofaffectedoffspringhavingtheband-presentmarkerphenotypeinfamilies1and2,respectively.hϭOp/EpOa/Ea.Anapproximate95%confidenceintervalforhwasesti-matedbythebacktransformationofthestandarderrorofloge(h)withvalueSE(loge(h))ϭΊ11ϩ.EpEaPermutationtestusingsurvivalanalysis:FollowingChurchillandDoerge(1994),apermutationtestwasperformedtoaccountformultipletesting.Theusualprocedureistoshufflethetraitvaluesrandomlyandthenreassigneachtraitvaluetoanewindividualwhileretainingthegenotypeatalltypedmarkersforeachindividual.Inthepresentpermutationtest,theindividualswereshuffledandthenumberofsurvivingdayswasreassignedinthesameorderastherealdata.Thiswasdoneforcomputationaleasesinceshufflingthetraitvalueswouldrequireorderingaccordingtothenumberofdaysalive.Theshufflingwasdone100,000timeswithineachfull-sibfamilyandforallmarkerstestedinthesurvivalanalysis.Alog-ranktestwascomputedforeachmarkerandpermutedsample.Eachpermutationsetcorrespondingtoallmarkerswasor-deredandthehighestvaluetaken.TheexperimentwisecriticalvaluewascomputedasthenumberoftimesthattheobservedLRTvalue,whenanalyzingthetruedata,washigherthanthehighestvalueofeachpermutationsetdividedby100,000.Inaddition,onepermutationtestwithonemillionreplicatesbutusingonesinglemarkerwasperformedtocompareempir-icalandtabulatedthresholdsinthesurvivalanalysis.RESULTSParentsandaffectedoffspringofapaternalhalf-sibfamilycomprisingoftwofull-sibfamiliesweregeno-typedwithAFLPmarkers.Atotalof64AFLPprimercombinationswereused,resultingin340scoredmark-ers(apolymorphismrateof5.3%).Atotalof401trans-missiondisequilibriumtestswereperformedusingsus-ceptiblefishandthe340polymorphicmarkers.Thisfirsttestingstageresultedin6markerssignificantatPϽ0.01and20significantatPϽ0.05(Table3).Thisresultindicatesthattherateofsignificantmarkersishigherthanexpectedbychance;thereisan“excess”ofsignificantmarkersthatcouldbeattributedtotheassociationofalleleswithdiseaseresistance.Thesecondtestingstageconsistedofgenotypingsur-vivorsforallsignificantmarkersatstage1andcarryingoutaMendeliansegregationtest.Thistestwasper-formedtoreducethenumberoftestsinthethirdstageandtoidentifymarkerswithnon-Mendelianinheri-tance.ItisshowninTable3thatsix,five,andfouroftheTDTsignificantmarkersdepartedfromtheexpectedMendeliansegregationatPϽ0.01,PϽ0.05,andPϽ0.10,respectively.Thesemarkerswereeliminatedfromthethirdstageofanalyses.ThethirdstagewastoperformsurvivalanalysesformarkersthatweresignificantintheTDTandthatalsofollowedMendelianinheritance.Atotalof17markersweretestedusingsurvivalanalysis,fromwhich2and7weresignificantusingthetabulatedthresholdsatPϽ0.1andPϽ0.05,respectively(Table4).Apermutationtestforthesurvivalanalysiswithonemillionreplicateswas,first,performedforonesinglemarkertocomparetabulatedandempiricalthresholds.Table5showsthattabulatedthresholdsareconserva-tive,inparticular,atlowsignificancelevels.Next,apermutationtestwasperformedsimulatingandtestingall17markersinthesurvivalanalysistoaccountformultipletesting.Theresultsofthepermuta-tiontestareshowninTable4.Onlythe2markersaaccac74andagccta290remainedsignificantatPϽ0.05afterthepermutationtest.Thehazardratiowasف0.25forbothmarkers;i.e.,areductionofthemortalityrateofف25%wasobservedamongfishinheritingtheband-presentalleleforeithermarker.TheKaplan-MeiersurvivalcurvesformarkersaacAQTLMappingStrategyTABLE3LRTafterTDTamongsusceptibleoffspringandMendeliansegregationtestNo.ofresistantoffspringp6177,586,413,16816,1114,109,1117,13991115,313,146611,1012812781317a10213,11714,117,193,54,412,33,311444,146,310119,631298943LRT13.76**11.84**10.00**8.89**8.80**8.76*8.05**7.86*7.08*6.84*6.49*6.37*6.37*6.24*6.20*6.16*5.76*5.76*4.57*4.55*4.26*4.00*3.86*3.86*3.86*3.85*MST4.56*23.68**13.47**5.76*15.21**3.79†,8.21**3.79***1.72,8.53**4.38*4.9*,0.00.214.33*8.00**2.372.37,7.81**0.11,1.921.320.950.890.030.643.27***2.78***2.630.000.44855ParentalgenotypesMarkeracactg299acacat393aggcat376aaccta206actcta476aacctg366agcctt142aggcat346aggctt110agccta219acactg369aagcag276aggcaa241aggctt391acgcag88acacaa404aagctt275acactg289aagcta477agccta290acacta397acgcac188actcaa133acccaa181aaccac74acgcag89SireaaAaAaaaAaAaaaAaAaAaAaaaaaAaAaAaaaaaAaaaaaaaaaaaaaAaDamAaaaaa,aaAaaa,aaaa,AaAaAa,aaAa,Aaaa,AaAa,AaAaAaAaAa,aaaa,AaAaAaaa,aaAaAaAaAaAaAaAaNo.ofsusceptibleoffspringp19175,5166,512,211716,1710,134,1615,114152016,75,1916166,6614126669a2213,17313,176,143,59,915,64,1115411,1314,35512,1316541515158p,band-presentphenotype;a,band-absentphenotype.Parentalgenotypes:Aaisheterozygousandaaishomozygousfortheband-absentallele.MST,Mendaliansegregationtest.*PϽ0.05;**PϽ0.01;***PϽ0.10.cac74andagccta290areshowninFigure2.Onlyattheverybeginningdidthelinescorrespondingtotheband-presentandband-absentgroupscrosseachotherformarkeragccta290,indicatingpossibledeparturefromtheproportionalhazardassumption.Thisassumptionmeansthatthereisaconstantrelativeriskalongtimeforeachtestedgroup.However,thecrossingofthetwolinestakesplaceattheverybeginningandcouldbeattributedtochance.DISCUSSIONThepurposeofthisstudywastoproposeastrategytodetectQTLfordiseaseresistanceinaquaculturespe-ciesforwhichgeneticmapsarenotavailable.Thestrat-egyismultistage:(1)TDTindeceasedoffspringisap-pliedusingAFLPs;(2)onlymarkersthatweresignificantintheaffectedoffspringaretypedforthesurvivorsandatestforMendeliansegregationisperformed;and(3)asurvivalanalysisisperformedformarkersthatweresignificantfortheTDTbutsegregatedaccordingtoMendelianinheritance.Theadvantagesofthisstrategyare,first,thatlargepartsofthegenomecanbecoveredusingfastandeasilyapplicabletyping.Second,thenum-beroftestsinthelaststageisreduced,allowingamorepowerfuldesign.Theneedtoaccountformultipletestingisageneralproblemingenome-widetestingforQTL.Theproposedstrategyalleviatestheproblembymakingalargereduc-tioninthenumberoftestsperformedinthesurvivalanalysis.However,themultistagemethodwasnotopti-mizedinthesensethatthresholdsattheTDTstagecouldbemadelargerorsmallertomaximizetheproba-bilityofnotmissingaQTLwhilelimitingthenumberofteststobeperformedinthesurvivalanalysis.Ifthethresholdismadesmallerthanthatinthisstudy(e.g.,top20%rankingmarkersattheTDTstage),thenthenumberofteststobeperformedislargerinthesurvivalanalysisandtheoverallpowermightbereduced.Moreresearchisneededtoestablishwhichthresholdsshouldbeusedtoachievemorepowerfuldesigns.TheapproachusedinthisstudyassumesthattheTDT856T.Moenetal.TABLE4Survivalanalysisresultsfortheresistanceallele,hazardratio,confidenceintervalofhazardratioat5%,LRANK,andexperimentwiseP-valuesResistancealleleBandBandBandBandBandBandBandBandBandBandBandBandBandBandBandBandBandpresentpresentabsentpresentabsentpresentpresentpresentpresentabsentpresentabsentpresentabsentpresentpresentabsentHazardratioDam1acacaa440acgcag89acacta397aagcta477acgcag88acactg369aggcat346agccta290aaccac74aagctt275aagcta477acactg289acactg369aggctt391agccta219acccaa181acacaa4400.300.330.340.570.360.771.00Dam20.250.270.320.340.340.400.270.620.650.760.11–0.560.11–0.640.14–0.760.14–0.850.14–0.790.16–1.020.08–0.870.21–1.800.27–1.570.25–2.319.625**8.724**5.441*5.17*4.950*4.818*1.8921.0260.8100.1930.0310.0460.2330.2650.2930.3120.9320.9970.9991.0000.12–0.740.11–1.020.14–0.830.22–1.440.09–1.370.23–0.540.29–3.436.037*5.743*4.821*1.3071.0710.2230.0000.1680.1970.3120.9870.9961.0001.000ConfidenceintervalExperimentwiseP-valueMarkerLRANK*PϽ0.05;**PϽ0.01.performedonsusceptiblefishandthesurvivalanalysisusedonthewholedatasetareindependenttests.Thisassumptionisnotlikelytobefullycorrect,sincetherewillbesomedependencebetweenthevariablesused(af-fected/resistantandnumberofdayssurvived).Theas-sumptionis,however,supportedbythefactthatthetwovariablesaredifferentvariablesandalsobecausetheanimalstestedintheTDTareonlyasubsetoftheanimalstestedinthesurvivalanalysis.ThesignificantmarkersfromTDTweretestedfordeparturesfromMendeliansegregationafterincorpo-ratingresistantanimals.MarkersnotfollowingMende-lianinheritancewereremovedfromthenexttestingTABLE5␹2tabulatedandempiricalthresholdsatsignificancelevelsof0.005,0.01,0.05,and0.10aftercarryingapermutationtestwithonemarkerandonemillionreplicatesSignificancelevel0.0050.0100.0500.100ThresholdEmpirical7.106.013.502.46Tabulated7.886.643.842.71stage(survivalanalysis).Amongthemarkersthatwerediscardedbecauseofnon-Mendelianinheritance,sev-eralmightbesegregatinganomalouslyjustbychance.Anotherpossiblereasonforthenon-Mendeliansegrega-tionisthatAtlanticsalmonisapartlytetraploidorgan-ism(Wrightetal.1983;AllendorfandThorgaard1984;AllendorfandDanzmann1997).AsaresultoftheapplicationoftheQTLtestingstrat-egy,twoputativeQTLaffectingresistancetoISAwerefoundwithinoneofthefull-sibfamilies.AfirstcrudeanalysisofthesegregationofthetwoQTLshowedlink-agebetweenthem(resultsnotshown).AdensegeneticmaparoundthisareawouldhelpforfinemappingoftheputativeQTL.Thelackofanavailablegeneticmapmakesitdifficulttolocatemanymarkersaroundthetwosignificantmarkers.OnepossibilitytospeedupthisprocessistheconstructionofalinkagemapstartingbyestablishinglinkagebetweeneachofthetwosignificantAFLPandtherestoftheAFLPpolymorphisms.Inaddi-tion,thegenotypingofavailablemicrosatellitesinourmaterialmightbeveryusefultoconstructageneticmaparoundthisarea.ThegeneralconclusionofthisstudyisthatQTLaf-fectingdiseaseresistancecanbedetectedinspeciesinwhichgeneticmapsarenotavailable.ThisrepresentsashortcutcomparedtoQTLmappingstrategiescurrentlyusedinterrestrialfarmanimals.Finally,themultistageQTLtestingstrategyispotentiallymorepowerfulthanAQTLMappingStrategy857andFigureagccta290.2.—Kaplan-Meiersurvivalcurvesformarkersaaccac74conventionalgenome-widescanswherethresholdsmustbesetveryhightoaccountforthetestingofalargenumberofgeneticmarkers.WearegratefultothepersonnelofVESOVikanAkvaVetandAquaGenforthebiologicalmaterialandforthechallengetesting.ThisworkhasbeensupportedbytheNorwegianResearchCouncilwithprojectno.130162/130titled“StrategicQTLResearchPlanforDiseaseResistanceinAtlanticSalmonandCattle.”LITERATURECITEDAgresti,J.J.,S.Seki,A.Cnaani,S.Poompuang,E.M.Hallermanetal.,2000Breedingnewstrainsoftilapia:developmentofanartificialcenteroforiginandlinkagemapbasedonAFLPandmicrosatelliteloci.Aquaculture185:43–56.Allendorf,F.W.,andR.G.Danzmann,1997SecondarytetrasomicsegregationofMDH-Bandpreferentialpairingofhomeologuesinrainbowtrout.Genetics145:1083–1092.Allendorf,F.W.,andG.H.Thorgaard,1984Polyploidyandtheevolutionofsalmonidfishes,pp.1–53inTheEvolutionaryGeneticsofFishes,editedbyJ.B.Turner.PlenumPress,NewYork.Altman,D.G.,1991PracticalStatisticsforMedicalResearch.Chap-man&Hall,London.Bagley,M.J.,S.L.AndersonandB.May,2001Choiceofmethod-ologyforassessinggeneticimpactsofenvironmentalstressors:polymorphismandreproducibilityofRAPDandAFLPfinger-prints.Ecotoxicology10:239–244.Bouchard,D.,K.Brockway,C.Giray,W.KeleherandP.L.Mer-rill,2001Firstreportofinfectioussalmonanemia(ISA)intheUnitedStates.Bull.Eur.Assoc.FishPathol.21:86–88.Castiglioni,P.,P.Ajmonemarsan,R.VanwijkandM.Motto,1999AFLPmarkersinamolecularlinkagemapofmaize:codominantscoringandlinkagegroupdistribution.Theor.Appl.Genet.99:425–431.Churchill,G.A.,andR.W.Doerge,1994Empiricalthresholdvaluesforquantitativetraitmapping.Genetics138:963–971.Coimbra,M.R.M.,K.Kobayashi,S.Koretsugu,O.Hasegawa,E.O’Haraetal.,2003AgeneticlinkagemapoftheJapaneseflounder,Paralichthysolivaceus.Aquaculture220:203–218.Evensen,Ø.,K.E.ThorudandY.A.Olsen,1991AmorphologicalstudyofthegrossandlightmicroscopiclesionsofinfectiousanaemiainAtlanticsalmon(SalmosalarL.).Res.Vet.Sci.51:215–222.Falk,K.,E.Namork,E.Rimstad,S.MjaalandandB.H.Dannevig,1997Characterizationofinfectioussalmonanaemiavirus,anorthomyxo-likevirusisolatedfromAtlanticsalmon(SalmosalarL).J.Virol.71:9016–9023.Jackson,T.R.,M.M.Ferguson,R.G.Danzmann,A.G.Fishback,P.E.Ihssenetal.,1998IdentificationoftwoQTLinfluencinguppertemperaturetoleranceinthreerainbowtrout(Oncorhyn-chusmykiss)half-sibfamilies.Heredity80:143–151.Jones,C.J.,K.J.Edwards,S.Castaglione,M.O.Winfield,F.Salaetal.,1997ReproducibilitytestingofRAPD,AFLPandSSRmarkersinplantsbyanetworkofEuropeanlaboratories.Mol.Breed.3:381–390.Kocher,T.D.,W.D.Lee,H.Sobolewska,D.PenmanandB.McAn-drew,1998Ageneticlinkagemapofacichlidfish,thetilapia.Genetics148:1225–1232.Mullins,J.E.,D.GromanandD.Wadowska,1998InfectioussalmonanemiainsaltwaterAtlanticsalmon(SalmosalarL.)inNewBrunswick,Canada.Bull.Eur.Assoc.FishPathol.18:110–114.Nichols,K.M.,W.P.Young,R.G.Danzmann,B.D.Robison,C.Rexroadetal.,2003Aconsolidatedlinkagemapforrainbowtrout(Oncorhynchusmykiss).Anim.Genet.34:102–115.Nylund,A.,A.M.Kvenseth,B.KrossoyandK.Hodneland,1997Replicationoftheinfectioussalmonanaemiavirus(ISAV)inrainbowtroutOncorhynchusmykiss(Walbaum).J.FishDis.20:275–279.Ozaki,A.,T.Sakamoto,S.Khoo,K.Nakamura,M.R.M.Coimbraetal.,2001Quantitativetraitloci(QTLs)associatedwithresis-tance/susceptibilitytoinfectiouspancreaticnecrosisvirus(IPNV)inrainbowtrout(Oncorhynchusmykiss).Mol.Genet.Genomics265:23–31.Piepho,H.P.,andG.Koch,2000Codominantanalysisofbandingdatafromadominantmarkersystembynormalmixtures.Genet-ics155:1459–1468.Refstie,T.,1990Applicationofbreedingschemes.Aquaculture85:163–169.Rodger,H.D.,T.Turnbull,F.Muir,S.MillarandR.H.Richards,1998Infectioussalmonanemia(ISA)intheUnitedKingdom.Bull.Eur.Assoc.FishPathol.18:115–116.Sakamoto,T.,R.G.Danzmann,K.Gharbi,P.Howard,A.Ozakietal.,2000Amicrosatellitelinkagemapofrainbowtrout(Oncor-hynchusmykiss)characterizedbylargesex-specificdifferencesinrecombinationrates.Genetics155:1331–1345.Spielman,R.S.,R.E.McGinnisandW.J.Ewens,1993Transmissiontestforlinkagedisequilibrium:theinsulingeneregionandinsu-lin-dependentdiabetesmellitus(IDDM).Am.J.Hum.Genet.52:506–516.Thorud,K.,1991Infectioussalmonanemiavirus.Ph.D.Thesis,NorwegianCollegeofVeterinaryMedicine,Oslo.Thorud,K.,andH.O.Djupvik,1988InfectioussalmonanemiainAtlanticsalmon(SalmosalarL.).Bull.Eur.Assoc.FishPathol.8:109–111.Vos,P.,R.Hogers,M.Bleeker,M.Reijans,T.VanDeLeeetal.,858T.Moenetal.somepairingintetraploidderivativesalmonidgenomes.IsozymesCurr.Top.Biol.Med.Res.10:239–260.Young,W.P.,P.A.Wheeler,V.H.Coryell,P.KeimandG.H.Thorgaard,1998Adetailedlinkagemapofrainbowtroutproducedusingdoubledhaploids.Genetics148:839–850.Communicatingeditor:G.A.Churchill1995AFLP:anewtechniqueforDNAfingerprinting.NucleicAcidsRes.23:4407–4414.Waldbieser,G.C.,B.G.Bosworth,D.J.NonnemanandW.R.Wolters,2001Amicrosatellite-basedgeneticlinkagemapforchannelcatfish,Ictaluruspunctatus.Genetics159:727–734.Wright,J.E.,K.Johnson,A.HollisterandB.May,1983Meioticmodelstoexplainclassicallinkage,pseudolinkage,andchromo-