selectionjusewang

OPENORIGINALARTICLENeuregulinsignalingpathwayinsmokingbehaviorRGupta1,BQaiser1,LHe2,3,TSHiekkalinna1,4,ABZheutlin5,STherman4,MOllikainen1,6,SRipatti1,6,MPerola4,VSalomaa4,LMilani7,TDCannon5,PAFMadden8,TKorhonen1,4,6,9,JKaprio1,6andALoukola1Understandingmolecularprocessesthatlinkcomorbidtraitssuchasaddictionsandmentaldisorderscanprovidenoveltherapeutictargets.
Neuregulinsignalingpathway(NSP)haspreviouslybeenimplicatedinschizophrenia,aneurodevelopmentaldisorderwithhighcomorbiditytosmoking.
UsingaFinnishtwinfamilysample,wehavepreviouslydetectedassociationbetweennicotinedependenceandERBB4(aneuregulinreceptor),andlinkageforsmokinginitiationattheERBB4locuson2q33.
Further,Neuregulin3hasrecentlybeenshowntoassociatewithnicotinewithdrawalinabehavioralmousemodel.
Inthisstudy,wescrutinizedassociationandlinkagebetween15036common,lowfrequencyandraregeneticvariantsin10NSPgenesandphenotypesencompassingsmokingandalcoholuse.
UsingtheFinnishtwinfamilysample(N=1998from740families),wedetected66variants(representing23LDblocks)signicantlyassociated(falsediscoveryratePo0.
05)withsmokinginitiation,nicotinedependenceandnicotinewithdrawal.
Wecomprehensivelyannotatedtheassociatedvariantsusingexpression(eQTL)andmethylationquantitativetraitloci(meQTL)analysesinaFinnishpopulationsample.
Amongthe66variants,weidentied25eQTLs(inNRG1andERBB4),22meQTLs(inNRG3,ERBB4andPSENEN),amissensevariantinNRG1(rs113317778)andasplicingdisruptionvariantinERBB4(rs13385826).
MajorityoftheQTLsinbloodwerereplicatedinsilicousingpubliclyavailabledatabases,withadditionalQTLsobservedinbrain.
Inconclusion,ourresultssupporttheinvolvementofNSPinsmokingbehaviorbutnotinalcoholuseandabuse,anddisclosefunctionalpotentialfor56ofthe66associatedsingle-nucleotidepolymorphism.
TranslationalPsychiatry(2017)7,e1212;doi:10.
1038/tp.
2017.
183;publishedonline22August2017INTRODUCTIONSmokingisamajorriskfactorforavarietyofsomaticdiseasesandstronglyassociateswithseveralneuropsychiatricdisorders,1–3themostprominentcomorbiditybeingalcoholuseanddependence.
4Oneofthekeyfactorsdrivingpersistentsmokingisnicotinedependence(ND),manifestedbydevelopmentoftolerance,symptomsofcravinganduncontrollableuseduetothehighaddictivepotentialofnicotine.
5Abstainingfromsmokingresultsinnicotinewithdrawal(NW)symptomsincluding,forexample,irritability,depressedmoodandrestlessness,6largelycontributingtothehighrelapseratesamongsmokerstryingtoquit.
7Despitetheavailabilityofseveralsmokingcessationpharmacotherapiesincludingnicotinereplacementtherapyandmedicationssuchasvarenicline,bupropionandcytisine,6-monthabstinenceratesareatbestonlytwo-tothreefoldcomparedwithpharmacologicallyunassistedquitattempts.
8Neuregulinsignalingpathway(NSP)isinvolvedinmodulatingneuronalmigrationanddifferentiation.
Thekeyfunctionalcomponentsofthispathwayareneuregulins(NRG1,NRG3)andtheirreceptor(ERBB4),aswellasbetasecretase(BACE1)andthegamma-secretasecomplex(comprisesPSEN1,PSEN2,APH1A,APH1B,PSENENandNCSTN)(reviewedinref.
9)).
DeviationfromanoptimallevelofNRG/ERBBsignalinginthebrainisshowntoimpairbrainfunctions.
10Interestingly,severalNSPgeneshavepreviouslybeenimplicatedinschizophrenia(SCZ),9–11aneuro-developmentaldisorderwithhighcomorbiditytosmoking.
EvidenceisemergingfortheinvolvementoftheNSPinsmokingbehavior.
Recently,weperformedagenome-wideassociationstudy(GWAS)amongsmokersfromtheFinnishTwinCohort(N=1104)anddetectedassociationbetweenNDdenedbyDiagnosticandStatisticalManualofMentalDisorders,4thedition(DSM-IV)12andERBB4.
13Usingapartlyoverlappingsubsetoffamilies(N=505twinsandrstdegreerelatives),wehavepreviouslyidentiedlinkagebetweensmokinginitiation(SI)andmicrosatellitemarkersat2q33,overlappingtheERBB4locus.
14Further,Nrg3hasrecentlybeenshowntoassociatewithNWinabehavioralmousemodel.
15Therefore,comprehensivescrutinyoftheNSPmayaidinidentifyingacommonlinkforsmokingbehaviorandcomorbiddisorders.
16GWAShasenableddiscoveryofgeneticvariantsassociatedwithdiseasetraits.
However,majorityoftheidentiedvariantsresideinnon-codingregions.
Itisessentialtoidentifythefunctionality/regulatorypotentialofthedetectedvariants,forexample,viamechanismssuchasepigeneticregulation.
Functionalannota-tionofvariantshasboostedprogressfromgeneticstudiestobiologicalunderstanding,aidingindevelopmentoftherapeuticmeasures.
17–19Inthisstudy,wescrutinizedassociationandlinkagebetweencommon,lowfrequencyandraregeneticvariantsinthe10genesencodingthekeyfunctionalcomponentsoftheNSPinaFinnishtwinfamilysample(N=1998from740families).
Tocomprehen-sivelytestwhetherNSPvariantsassociatewithnicotineuse,weincludedninesmoking-relatedphenotypesinouranalyses.
Inoursecondaryanalyses,weincludedvealcohol-relatedphenotypestoevaluate(i)whetherNSPvariantsassociatewithnicotineuse1InstituteforMolecularMedicineFinland,UniversityofHelsinki,Helsinki,Finland;2BroadInstituteofMITandHarvard,Cambridge,MA,USA;3ComputerScienceandArticialIntelligenceLaboratory,MIT,Cambridge,MA,USA;4NationalInstituteforHealthandWelfare,Helsinki,Finland;5DepartmentofPsychology,YaleUniversity,NewHaven,CT,USA;6DepartmentofPublicHealth,UniversityofHelsinki,Helsinki,Finland;7EstonianGenomeCenter,UniversityofTartu,Tartu,Estonia;8DepartmentofPsychiatry,WashingtonUniversitySchoolofMedicine,SaintLouis,MO,USAand9InstituteofPublicHealthandClinicalNutrition,UniversityofEasternFinland,Kuopio,Finland.
Correspondence:RGupta,InstituteforMolecularMedicineFinland,UniversityofHelsinki,Tukholmankatu8,Biomedicum2B,HelsinkiFIN-00014,Finland.
E-mail:richa.
gupta@helsinki.
Received29December2016;revised26June2017;accepted6July2017Citation:TranslPsychiatry(2017)7,e1212;doi:10.
1038/tp.
2017.
183www.
nature.
com/tpspecically,orrelatetogeneralliabilitytoaddictions,and(ii)whethersignalsobservedforsmoking-relatedphenotypesareconfoundedbycomorbidalcoholuse.
Wedetectedstatisticallysignicantassociationwithsingle-nucleotidepolymorphisms(SNPs)andSI,DSM-IVNDandDSM-IVNW,infullagreementwithpreviousstudiesofNSPinsmokingbehavior.
13–15Further,weperformedcomprehensiveannotationofthedetectedassocia-tions,andrevealedpotentialfunctionsformajorityoftheassociatingSNPs.
MATERIALSANDMETHODSDiscoverysampleTheNAG-FINdiscoverysamplehasbeenpreviouslydescribedindetail.
13,14,20Briey,thestudysamplewasascertainedfromtheFinnishTwinCohortconsistingof35834adulttwinsbornin1938–1957.
Onthebasisofearlierdata,thetwinpairsconcordantforever-smokingwereidentiedandrecruitedalongwiththeirfamilymembers(mainlysiblings)fortheNicotineAddictionGenetics(NAG)Finlandstudy(N=2265).
Twinpairsconcordantforheavysmokingwereprimarilytargetedtoincreasethegeneticload.
Datacollectiontookplacein2001–2005.
Theassociationandlinkagestudysampleconsistedof1998individualsfrom740families,including980dizygotic(DZ)twinindividuals(bothco-twinsincluded),139monozygotic(MZ)twinindividuals(oneco-twinpertwinpairwasincluded),46individualswithunconrmedzygosity(duetoalackofDNAsamplefromtheco-twin)and833otherfamilymembers(mostlysiblings).
Theaveragenumberofindividualsperfamilywas3(range1–10),andthedatacontained185familieswithatleast4familymembersincludedinthestudy.
SampledescriptionispresentedinTable1.
ThestudywasapprovedbytheethicscommitteeofthehospitaldistrictofHelsinkiandUusimaa,Finlandin2001and2016,andbytheIRBofWashingtonUniversity,StLouis,MO,USA.
Theparticipantshaveprovidedwritteninformedconsent.
NAG-FINgenotypedata.
GenotypingfortheNAG-FINsamplewasperformedwiththeHuman670-QuadCustomIlluminaBeadChip(Illumina,SanDiego,CA,USA)(N=1097)attheWellcomeTrustSangerInstitute,UK,andwiththeIlluminaHumanCoreExomeBeadChip(N=901)attheWellcomeTrustSangerInstitute,andattheBroadInstituteofMITandHarvard,USA.
Qualitycontrols(QC)forthegenotypedatahavebeenpreviouslydescribed21andarealsopresentedinSupplementaryTable1.
Pre-phasingofthedatawasdonewithSHAPEIT222andimputationwithIMPUTE223usingthe1000GenomesPhaseIintegratedhaplotypesreferencepanel.
24Foranalysesofthe10NSPgenes(NRG1,NRG3,ERBB4,BACE1,PSEN1,PSEN2,APH1A,APH1B,PSENENandNCSTN),weextractedSNPswithinthegeneregions(accordingtothelongestisoformreportedattheUCSCGenomebrowser)with50kbankingregions.
Geneboundaries(accordingtoGRCh37/hg19)arelistedinSupplementaryTable1,alongwiththenumberofSNPsincludedforeachgene.
Onlyvariantswithminorallelefrequency(MAF)o0.
01locatedincodingregions,splicesites,promotersoruntranslatedregions(UTRs)wereincludedintherarevariantanalysis.
Altogether,15036SNPswereanalyzedinourdiscoveryphase.
NAG-FINphenotypedata.
ToevaluatetheroleofgeneticvariantswithintheNSPgenesinaddictions,weinitiallytestedninephenotypesencom-passingsmokingbehaviorinourdiscoverysample(SupplementaryTable2(I)).
TotestwhethervariantsinNSPgenesassociatewithnicotineusespecically,orrelatetogeneralliabilitytoaddictions,wealsotestedvealcoholusephenotypes(SupplementaryTable2(II)).
Datacollectionforthephenotypeshasbeenpreviouslydescribedindetail.
6,20AnnotationsampleForfunctionalannotationoftheassociationsdetectedinthediscoveryphase,weusedtheDILGOM2007(Dietary,LifestyleandGeneticdeterminantsofObesityandMetabolicsyndrome)studysample,whichoriginatesfromthepopulation-basednationalFINRISK2007study.
TheDILGOMstudysamplehasbeenpreviouslydescribedindetail,25,26includingatotalof631unrelatedFinnishindividualsaged25–74yearsfromtheHelsinkiarea.
Forannotationanalyses,weused512individualswithgenome-widegenotype,geneexpressionandDNAmethylationdataavailable,allmeasuredfromperipheralbloodleukocytes(46%males,meanage52years(s.
d.
13.
7)).
DILGOMsamplewasusedtoperformpopulation-specicexpression(eQTL)andmethylationquantitativetraitloci(meQTL)analysisinbloodtissue.
Wefurtheranalyzeddifferentialtranscriptomeandmethylomeamongsmokersandneversmokersinthisdataset.
Onthebasisofself-reportedsmokingstatustheindividualsinthesamplecanbecategorizedintocurrentdailysmokers(N=84),currentoccasionalsmokers(N=34),recentquitters(1–6monthsofabstinence)(N=13),formersmokers(46monthsofabstinence)(N=133),neversmokers(N=245)andthreewithmissingsmokingstatus(asdescribedinref.
27).
TheDILGOMparticipantshaveprovidedwritteninformedconsent.
TheprotocolwasdesignedandperformedaccordingtotheprinciplesoftheHelsinkiDeclarationandwasapprovedbythecoordinatingethicscommitteeofthehospitaldistrictofHelsinkiandUusimaa,Finland.
DILGOMgenotypedata.
GenotypingfortheDILGOMsamplewasperformedwiththeIllumina610-QuadSNParray.
ImputationwasdonewithIMPUTE2usingthe1000GenomesPhaseIintegratedhaplotypesreferencepanel.
23,24StringentQCthresholdswereappliedtolteroutlowqualitySNPsandsamplesaspreviouslydescribed.
25QCsandimputationforallFinnishGWASdataweredonecentrallyattheInstituteforMolecularMedicineFinland(FIMM),UniversityofHelsinki,Finland.
DILGOMgene-expressiondata.
RNAprotocolsanddataprocessingfortheIlluminaHT-12expressionarrayhavebeenpreviouslydescribed.
25Briey,peripheralbloodwasusedasasourceofRNA;allarrayswerequantilenormalizedatthestrip-levelandtechnicalreplicateswerecombinedviabead-countweightedaverage.
Probeswereremovediftheymappedtoanon-autosomalchromosome,erythrocyteglobincomponentsormorethanonegenomicposition.
Atotalof19probesmappedtothe10NSPgenes,ofwhich17passedQCthresholds;noneofthePSENENprobespassedQC.
Oneprobepergenewasselectedbasedonhighestinterquartilerange(IQR)representinghighestvariancefordifferentialexpressionanalysis.
DILGOMDNAmethylationdata.
DNAextractedfromperipheralbloodwasbisulteconvertedusingEZ-96DNAMethylation-GoldKit(ZymoResearch,Irvine,CA,USA)accordingtothemanufacturer'sinstructions.
IlluminaInniumHumanMethylation450BeadChipwasusedtomeasuretheDNAmethylationlevelsusingtheInniumprotocolformethylationworkow.
28DNAmethylationdatawaspre-processedandnormalizedusingthebioconductorRpackage'min',29withtheSubset-quantileWithinArrayNormalization(SWAN)method.
ProbeswithdetectionP-value40.
01inanysamplewerediscarded.
Probesreportedascross-reactive(mappingtomultiplegenomiclocationsandknownrepeatregions)andprobeswithSNPswerealsoremoved,aspreviouslysuggested.
29Atotalof254CpGprobesmappedtothetenNSPgenes,ofwhich226passedQCthresholds.
Onaverage,therewere25probespergene(range13–69).
InsilicodatabasesforreplicatingQTLsidentiedinbloodWeutilizedblood-derivedgeneexpressionandgenotypicdataavailableintheGenotype-TissueExpression(GTEx)projectdatabase(gtexportal.
org)30toreplicatedetectedeQTLsintheDILGOMsample.
ForreplicatingTable1.
DiscoverysamplecharacteristicsSamplecharacteristicsDescriptivestatisticsTotalsamplesize(%males)1998(52%)Numberoffamilies740Meanage(range,s.
d.
)56(30–92,10.
1)Fulllingsmokinginitiationa(%)1660(83%)Meanageatsmokinginitiation(range,s.
d.
)18.
5(7–56,4.
96)MeanCPD(range,s.
d.
)b18.
8(1.
5–45,10.
2)MeanDSM-IVNDsymptoms(range,s.
d.
)b2.
9(0–7,1.
7)FulllingDSM-IVNDdiagnosis(%)b844(42%)MeanDSM-IVNWsymptoms(range,s.
d.
)c2.
3(0–8,2.
1)FulllingDSM-IVNWdiagnosisc522(26%)Abbreviations:ND,nicotinedependence;NW,nicotinewithdrawal.
aSmoked100cigarettesduringlifetime.
bAmongthosewhohaveinitiatedsmoking.
cAmongthosewhohaveinitiatedsmoking,andhaveattemptedquitting.
NSPinsmokingbehaviorRGuptaetal2TranslationalPsychiatry(2017),1–8detectedmeQTLsintheDILGOMsample,wequeriedthemQTLdbdatabase(mqtldb.
org)31containingblood-derivedmethylationandgenotypedataonmother–childpairs.
InsilicodatabasesforidentifyingQTLsinthebrainGiventheneuropsychiatricnatureofourphenotypes,wetestedforeQTLsindifferentbraintissuesusing13braintissuesavailableinGTExandin10brainregionsavailableintheBraineQTLAlmanac(BRAINEAC;braineac.
org)database.
32FormeQTLs,wequeriedtheepigenome-widesignicantmeQTLsreportedbyHannonetal.
33infetalbrainsamples(epigenetics.
essex.
ac.
uk/mQTL).
SchizophreniatwinsampleIntheexpressionanalysesofthe10NSPgenes,wealsoutilizedaSCZtwinsample,previouslydescribedindetail.
34Thedataconsistedof18SCZcasesand55controls(includingunaffectedco-twinsandadditionalunaffectedtwinpairs),withgenome-widegene-expressiondatageneratedwithIlluminaHumanWG6v3.
0ExpressionBeadChip,aspreviouslydescribed.
34OneprobepergenewasselectedinaccordancewiththeprobeselectioninDILGOMsampleforexpressionanalyses.
Ageandsexwasavailableforallparticipants,whereassmokingstatus(denedassmokerversusnon-smoker)wasavailablefor13SCZcasesand41controls.
InsilicodatabaseforSCZtranscriptomeanalysisWequeriedthedatabaseforSchizophreniaGeneticResearch(SZDB;szdb.
org)35toidentifydifferentiallyexpressedNSPgenesbetweenSCZcasesandcontrols.
DiscoveryanalysesAltogether,9924common(MAF40.
05),4106lowfrequency(0.
01MAF0.
05)and1006rare(MAFo0.
01)variantsinthetenNSPgenesweretestedforassociationinitiallywithninesmokingphenotypesandassecondaryanalyseswithvealcohol-relatedphenotypes.
Wealsoexaminedcommonvariantsforlinkageaswellasjointlinkageandlinkagedisequilibrium(LD).
Toaccountformultipletesting,weusedtheBenjaminiandHochbergmethod36andconsideredfalsediscoveryrate(FDR)adjustedP-valuesbelow0.
05asstatisticallysignicant.
Tovisualizethegeneregionsshowingsignicantassociationsinourdiscoveryanalyses,weusedLocustrack37andHaploviewfordeningLDblocksbasedonthe'solidspineofLD'option.
38Geneticassociationanalysisofcommonandlowfrequencyvariants.
Uni-variatelinearmixedmodel(LMM)implementedinGEMMA(genome-wideefcientmixed-modelassociation)39wasusedtotestassociationofSNPswiththequantitativetraits.
Genotypeuncertaintywasaccountedforusingallelicdosagedata.
Ageandsexwereusedascovariates(xedeffectspartofthemodel).
Populationstraticationandgeneticcorrelationwithinthesamplewasmodeledwithadditionalrandomeffectsusingastandardizedrelatednessmatrixcalculatedfromgenome-widegenotypedata.
P-valuesfromWaldtestwereusedtoassesstheassociationbetweeneachSNPandthephenotype.
LinkageandjointlinkageandLDanalysisofcommonvariants.
Two-pointlinkageanalysisaswellasjointlinkageandLDanalysiswasperformedforbinarytraitsusingthePSEUDOMARKERsoftware,40assumingarecessivemodeofinheritance(identicaltonon-parametricaffectedsib-pairanalysis(ASP),whichisaspecialcaseofparametriclinkageanalysis).
41,42PSEUDOMARKERhasbeenevidencedasthemostpowerfulfamily-basedassociationanalysismethodforbinarytraitsimplementingtheElston–Stewartalgorithmforfull-likelihoodanalysis.
43OnlycommonvariantswithMAF40.
05wereincludedintheseanalyses,toavoidanalysisofmonomorphicanduninformativemarkersinthesample.
Inourlinkageanalyses,weleveragedtheextendedtwinfamilydata,asourdatacontained185familieswithatleastfourfamilymembersincludedinthestudy.
Rarevariantassociationanalysis.
Therarevariantassociation(RVA)analysiswasdividedintosingle-variantandgene-basedtests.
Anexpectedkinshipmatrixwascalculatedusingthepedigreeinformationandweincorporatedageandsexascovariates.
Forsingle-variantRVAanalysisofquantitativetraits,weusedthe'lmekin'functionfromthe'coxme'Rpackage.
44ToanalyzebinarytraitsRpackage'pedigreemm'wasemployed.
45Earlierstudieshaveshownthatsingle-varianttestssufferfromlossofstatisticalpower.
46Consequently,gene-basedtestsusingSNP-set(Sequence)KernelAssociationTest(SKAT)47forquantitativetraitsandHierarchicalBayesianMultipleRegressionmodel(HBMR)48forbothquantitativeandbinarytraitswereperformedusingRpackages.
Inthegene-basedtests,onlyonevariantwasselectedwhenevermultiplevariantswereinfullLD,andeachvariantincludedhadtobeatleastasingleton(thatis,havinganimputedgenotype40.
5foratleastoneindividual).
HBMRoutputsresultsasBayesfactors(BFs),andwedeclaredanominallysignicantndingwhenBFexceededathresholdof2.
45(correspondingtoaP-valueo0.
05),aspreviouslysuggested.
49,50AnnotationsAllSNPsshowingstatisticallysignicantassociationorlinkage(FDRPo0.
05)inthediscoveryanalyseswereannotatedtoinferpotentialfunctionalconsequences.
Forthis,weusedquantitativetraitlocianalysisinapopulation-basedsample(blood-deriveddata),andconsideredFDRadjustedP-valuesbelow0.
05asstatisticallysignicant.
ToreplicateQTLsdetectedinblood,weusedpubliclyavailabledatabases.
Further,asourphenotypesareneuropsychiatricdisorders,wewantedtoexploretheeffectofassociatingSNPsusingpubliclyavailablebraintissuedata.
ExpressionquantitativetraitlocianalysisintheDILGOMsample.
Expressionquantitativetraitloci(eQTL)analysiswasperformedintheDILGOMsampletotesttheeffectofgenotypesofthehighlightedSNPsonexpressionlevelsofNSPgenes.
Rpackage'MatrixEQTL'wasusedwithalinearmodelsetting.
51Age,sex,bodymassindex(BMI)andsmokingstatuswereusedascovariatestotestcis-actingeQTLs.
Normalizedexpressionvalueswerelog2transformedandtestedagainstSNPgenotypescodedas0,1or2copiesoftheeffectallele.
MethylationquantitativetraitlocianalysisintheDILGOMsample.
Methyla-tionquantitativetraitloci(meQTL)analysiswasperformedintheDILGOMsampletotesttheeffectofgenotypesofthehighlightedSNPsonmethylationlevelsofNSPgenes.
Wetestedcis-actingmeQTLsusingtheRpackage'MatrixEQTL'withlinearmodelsetting.
Age,sex,BMI,smokingstatusandwhitebloodcellcountsestimatedusinghousemanalgorithm52wereaddedascovariates,aspreviouslysuggested.
31,53Normalizedmethylationβ-valuesweretestedagainstSNPgenotypescodedas0,1or2copiesoftheeffectallele.
Insilicoexpressionquantitativetraitlocianalysis.
ToreplicateoureQTLsdetectedintheDILGOMsample(blood),weutilizedGTExwhole-blooddata.
ToexploretheeffectsofourhighlightedSNPsinthebrain,wetestedtheSNPsascis-eQTLsin13braintissuesavailableinGTEx,andin10brainregionsavailableinBRAINEAC.
AllGTExanalyses(bloodandbraintissues)wereperformedwiththe'TestYourOwn'eQTLsoption.
Insilicomethylationquantitativetraitlocianalysis.
ToreplicateourmeQTLsdetectedintheDILGOMsample(blood),weusedthemQTLdbassessingcis-meQTLs.
Totesttheeffectofgenotypesinthebrain,wequeriedthefetalbrainmeQTLsreportedbyHannonandcolleagues.
Otherfunctionalannotations.
EnsemblVariantEffectPredictor(VEP)54wasusedtoprocurepredictedfunctionalannotationsforthehighlightedSNPs.
HaploRegv4wasusedtoannotatetheSNPsforregulatoryregions(promoterandenhancerhistonemarks,DNaseIhypersensitivity,proteinbindingandregulatorymotifs).
55Inaddition,SPANR56scoresforsplicingdisruptionpotentialwereextractedforthehighlightedSNPs.
DifferentialexpressionandmethylationanalysesAswedetectedassociationbetweensmokingbehaviorandseveralNSPgenes,wewantedtoexplorewhetherNSPgeneexpressionandmethylationlevelsareaffectedbysmokingstatus.
Weperformeddifferentialexpressionandmethylationanalysisbetweennever(N=245)andcurrentdailysmokers(N=84)intheDILGOMsample.
Wealsocomparedneversmokers(N=245)againstapooledsampleofcurrentdailyandoccasionalsmokers(N=118).
Weemployedlinearmodelsusingthe'lm'functioninRwhileadjustingforage,sexandBMI.
Whencomparingmethylationlevels,wefurtheradjustedforestimatedwhitebloodcellcounts.
52WeconsideredFDRadjustedPo0.
05asevidenceforsignicantdifference.
OwingtothehighcomorbiditybetweenSCZandsmoking,wewantedtoexploretheeffectofsmokingonNSPgeneexpressionintheSCZtwinNSPinsmokingbehaviorRGuptaetal3TranslationalPsychiatry(2017),1–8sample.
WerstexaminedexpressiondifferencesbetweenSCZcases(N=18)andcontrols(N=55),whileaccountingforrelatednessandusingageandsexascovariates.
Wethenaddedsmokingstatusasacovariate.
Inthissample,wecouldonlyclassifythesubjectsintosmokersandnon-smokers,astherewasnoinformationonformersmoking.
Owingtothesmallsamplesize,nocorrectionformultipletestingwasapplied.
RESULTSDiscoveryanalysesWetested15036SNPsmappingtothe10NSPgenes(with50kbankingregions)forlinkage,jointlinkageandLD,andassociationidentifyingaltogether66SNPsshowingstatisticallysignicantsignal(FDRPo0.
05)withSI,NDandNWphenotypes(Supplemen-taryTable3).
OnthebasisofLDstructures,the66SNPsrepresent23LDblocks(SupplementaryFigure1).
AsummaryoftheresultsfromthediscoveryanalysesispresentedinTable2.
WedetectedevidenceoflinkageforSI,NDdiagnosisandNWdiagnosiswithmultipleSNPsinERRB4,withthestrongestsignalemerginginrs1836721forSI(logoftheodds(LOD)=3.
32)andND(LOD=1.
94),andinrs147786642forNW(LOD=1.
12).
JointlinkageandLDtestsperformedwithPSEUDOMARKERdetectedasignalforSIinsevenofthetenNSPgenes(APH1A,PSEN2,ERBB4,NRG1,NRG3,BACE1andPSENEN;FDRPo0.
05)(SupplementaryFigure2),andsignalforNDdiagnosisinERBB4(FDRP=0.
002forrs13385826)(SupplementaryFigure3).
AssociationtestsperformedwithGEMMAidentiedasignalforNWsymptomcountinERBB4(FDRP=0.
008,β=0.
517forrs13001305,andforthreehighlycorrelatedSNPs:rs73989053,rs13006797,rs17328083)(Figure1),suggestingthateachcopyoftheeffectallelelowersnumberofNWsymptomsbyhalfacountonascaleof0–8counts.
SimilarresultswereobtainedforNRG3SNPrs11192578(FDRP=0.
008,β=0.
773)(SupplementaryTable3),suggestingthateachcopyoftheeffectallelelowersnumberofNWsymptomsby0.
8counts.
Nosignicantsignalwasdetectedinsingle-variantRVAanalysesaftercorrectingformultipletesting(datanotshown).
Inthegene-basedRVAanalysis,neithermethods(SKATorHBMR)detectedsignicantlyassociatedvariantsforNDorNWsymptomcounts.
Incontrast,HBMRidentiedassociationbetweenSIandbothNRG1(BF=3.
073,Po0.
0360)andPSEN1(BF=9.
693,Po0.
0078),andbetweenNWdiagnosisandERBB4(BF=3.
799,Po0.
0267).
AnnotationTotestwhetherthe66SNPshighlightedinthediscoveryphaseassociatedwithgeneexpressionorDNAmethylationlevelsoftheNSPgenes,weperformedciseQTLandmeQTLanalysesintheDILGOMsample(blood-deriveddata).
Outofthe66SNPs,twoSNPsinERBB4and23SNPsinNRG1wereidentiedaseQTLs(FDRPo0.
05)(SupplementaryTable4).
Further,12SNPsinERBB4(eightofthoseformingmeQTLswithmultipleCpGsites),oneSNPinPSENEN,andoneSNPinNRG3(FDRPo0.
05)wereidentiedasaltogether22meQTLs(SupplementaryTable5).
ToreplicateoureQTLndings,wetestedthe66highlightedSNPsforcis-eQTLsinwhole-blooddataavailableinGTEx.
Weobserved17eQTLsinNRG1,overlappingthe23NRG1eQTLsidentiedinDILGOMsample,andadditionaleQTLsinBACE1andPSENEN(FDRPo0.
05)(SupplementaryTable6).
Wethentestedall66SNPsforcis-eQTLsinRNAexpressiondatafrom13brainregionsavailableintheGTExand10brainregionsavailableintheBRAINEACdatabase.
Altogether,vebraineQTLsweredetected:oneERBB4SNP(rs13385826)inspinalcord,oneNRG3SNP(rs12774918)inamygdala,oneBACE1SNP(rs1261780)incerebellarhemisphereandtwoERBB4SNPs(rs192584214,rs112465988)incerebellarcortex(FDRPo0.
05)(SupplementaryTable2.
Summaryofstatisticallysignicant(FDRPo0.
05)resultsfromdiscoveryphaseGenesymbolGenenameChrNo.
ofSNPshighlightedPhenotypesshowingassociationSNPswithlowestP-valuersnumberP-valueAPH1AAph-1homologA12SmokingInitiationrs1834238661.
8E09PSEN2Presenilin211SmokingInitiationrs109160532.
1E04ERBB4Erb-B2ReceptorTyrosineKinase4227SmokingInitiationrs134130994.
0E27NicotineDepdendencers133858261.
7E07NicotineWithdrawalrs130067972.
1E06NRG1Neuregulin1824SmokingInitiationrs43292359.
3E13NRG3Neuregulin3109SmokingInitiationrs115281792.
9E07NicotineWithdrawalrs111925781.
2E06BACE1Beta-secretase1112SmokingInitiationrs1911092954.
0E06PSENENPresenilinenhancer191SmokingInitiationrs8074831.
7E04Abbreviations:FDR,falsediscoveryrate;ND,nicotinedependence;NSP,neuregulinsignalingpathway;NW,nicotinewithdrawal;SI,smokinginitiation;SNP,singlenucleotidepolymorphism.
Altogether,66SNPswerehighlightedinsevenofthetenNSPgeneswithSI,DSM-IVNDdiagnosisandDSM-IVNWsymptomcountphenotypes.
Foreachgene,SNPwiththelowestP-valuesarepresented.
Completeresultsfor66SNPsarepresentedinSupplementaryTable3.
Figure1.
Regionalplotforrs11192578inERBB4geneshowingassociationwithnicotinewithdrawal(NW)symptomcount.
NSPinsmokingbehaviorRGuptaetal4TranslationalPsychiatry(2017),1–8Table7).
OverlapbetweeneQTLsdetectedinbrainandblood(GTEx)wasobservedonlyforrs1261780inBACE1.
ToreplicateourmeQTLsndings,wetestedthe66highlightedSNPsforcis-meQTLsinmQTLdb.
TwelvemeQTLsidentiedinDILGOMsampleoverlappedwithresultsfrommQTLdbwithadditionalmeQTLsobservedforrs807483inPSENEN(SupplementaryTable8).
Inaddition,amongtheepigenome-widesignicantmeQTLsreportedinfetalbrainbyHannonandcolleagues,wefoundfouroftheERBB4SNPs(rs73989053,rs17328083,rs13001305andrs75489550)(SupplementaryTable9),alloverlappingwithmeQTLsintheDILGOMsample,andallshowingassociationwithNWinourdiscoverysample.
Amongthehighlighted66SNPs,EnsemblVariantEffectPredictor(VEP)indicatedamissensevariantinNRG1(rs113317778),promoterregionvariantsinNRG1(rs75673683)andERBB4(rs73989053),andtwoAPH1ASNPs(rs183423866,rs187135585)thatoverlapwithtranscriptionfactorbindingsites(SupplementaryTable10).
AccordingtoHaploReg,asignicantportionofERBB4andNRG3SNPsoverlapwithpromoterand/orenhancersinbraintissue(SupplementaryTable11).
Mostofthehighlighted66SNPswereeithertoofarfromsplicejunctionsorhadnooverlappingcodingtranscriptsforsplicingdisruptionanalysiswithSPANR.
OnlyERBB4variantrs13385826hadadeltapsiscoreof0.
09indicatingsplicingdisruptionpotential(absolutedeltapsi40.
05).
Annotationresultsforthe66SNPsaresummarizedinFigure2andinSupplementaryTable3.
DifferentialexpressionandmethylationanalysesWhenexaminingtheexpressionofthetengenes,weobservedhigherexpressionlevelsofNRG1andPSEN1amongcurrentdailyversusneversmokers(FDRPo0.
05)(SupplementaryTable12).
Whenoccasionalsmokerswerepooledtogetherwithcurrentdailysmokers,thedifferenceswerenolongersignicant.
Expressionlevelsinoccasionalsmokersresembledthoseseeninneversmokers(SupplementaryFigure5).
Nosignicantdifferenceinmethylationlevelsbetweencurrentdailyversusneversmokerswereobserved(SupplementaryTable13),includingoccasionalsmokersamongcurrentdailysmokersdidnotaffectthisresult.
IntheSCZtwinsample,NRG1,NRG3andAPH1Bshowedatrend(P-valueo0.
1)fordifferentialexpressionbetweenSCZcasesandcontrolsinanalysesadjustedforageandsex(SupplementaryTable14).
Whensmokingstatuswasaddedtothemodel,thetrenddisappeared.
DISCUSSIONSmokingbehaviorandNDarecomplextraitswithseveralgenesandpathwayshavingcrucialroles.
Inadditiontothewell-establishedroleofgeneticvariationinnicotinicreceptorgenes57–59andnicotinemetabolizingenzymes,60–62theNSPisemergingasacontributorinsmokingbehavior.
ByusingaFinnishtwinfamilysample,wehavepreviouslyidentiedlinkageofSIattheERBB4locus,14andassociationbetweenNDandERBB4.
13Turneretal.
15haverecentlyshowntheinvolvementofNrg3andErbb4intheanxietyeffectsofNWinabehavioralmousemodel,andassociationbetweenNRG3SNPswithsmokingcessationsuccessinaclinicaltrial.
Inthecurrentstudy,weincluded14phenotypesassessingsmokingbehaviorandalcoholuseinaFinnishtwinfamilysample(N=1998individualsfrom740families).
Wescrutinizedaltogether15036common,lowfrequencyandrarevariantsin10NSPgenes,andidentied66SNPs(representing23LDblocks)signicantlyassociatedwithSI,NDandNW.
Wethencomprehensivelyannotatedthefunctionalpotentialofthehighlightedvariantsinanindependentpopulation-basedsampleofsmokersandnon-smokers.
OnthebasisofLDguresgeneratedforthesevengenes,weestimatethatthe66highlightedSNPsrepresent23LDblocks.
Forallbutthreegenes(APH1A,PSEN2andPSENEN),SNPsinmultipleLDblockswerehighlighted.
WedetectedlinkageforSIwithmultipleSNPsinERBB4,strengtheningourpreviouslinkagendings.
14Oursamplesizeincreasedbetweentheoriginallinkagestudyandthecurrentstudy(N=505versusN=1998),andtheevidenceforlinkageincreasedcorrespondingly(maxparametricLOD=2.
56versusmaxLOD=3.
32),despitethefactthatbiallelicmarkers(thatis,SNPs)usedinthecurrentstudyprovidelesslinkageinformationindividuallycomparedtomultiallelicmarkers(thatis,microsatellitemarkers)63usedintheoriginallinkagestudy.
14ERBB4variantsalsoshowedlinkageforNDdiagnosis(LOD=1.
94)andNWdiagnosis(LOD=1.
12),jointlinkageandLDforSIandNDdiagnosis,andassociationforNWsymptomcount.
Thesendingsfurthersupportourpreviousassociationresults13andprovidesubstantiveevidenceforfuturestudiestocharacter-izetheroleofERBB4inaddictions.
InadditiontohighlightingERBB4,ourresultsprovideevidencefortheinvolvementofotherNSPgenesinsmokingbehavior.
A2NESPPSENEN668324381sr585531781sr35061901srrs1836721rs13385826rs184443533rs6729616rs187745320rs13413099rs71350752rs192584214rs112465988rs185444317rs10167304rs12694263rs1581385rs187445356rs4993337rs141486719rs78391944rs71422776rs12990946rs35225704rs73989053rs13006797rs17328083rs13001305rs75489550rs36001505rs147786642rs113317778rs4327825rs4329235rs4436087rs10954844rs77085193rs17643260rs7000656rs17716671rs11780575rs17644185rs17716920rs10808325rs75974189rs11779465rs58182668rs11787140rs4733344rs12545304rs11777461rs75673683rs4733346rs7840005rs77752517rs79886978rs12774918rs11528179rs191742951rs11192578rs34434628rs60197472rs59013794rs7908968rs1261780rs191109295rs807483eQTLinblood(DILGOM)meQTLinblood(DILGOM)Splicingdisruptive(SPANR)Promoteroverlap(HaploReg)eQTLinblood(GTEx)meQTLinblood(mQTLdb)MissenseSNP(VEP)Enhanceroverlap(HaploReg)eQTLinbrain(GTEx,BRAINEAC)meQTLinfetalbrain(Hannonetal2016)RegulatoryregionSNP(VEP)A1HPAERBB4NRG1NRG3BACE1Figure2.
Summaryguredepictingthefunctionalannotationofthe66single-nucleotidepolymorphisms(SNPs)highlightedinthediscoveryphase.
Thegurepresentsthesummaryoffunctionalannotationresultsforthe66SNPshighlightedinthediscoveryanalyseswhileshowingtheoverlapbetweenmultiplefunctionalannotationsforeachSNP.
InadditiontotheeQTLsandmeQTLsidentiedinbloodandbrain,weidentiedonesplicingdisruptionvariant(rs13385826)withSPANR,onemissensevariant(rs113317778),andfourvariantsinregulatoryregions(promoter(andanking)regionandtranscriptionfactorbindingsites)withsomeSNPsoverlappingpromoterandenhancerinbloodandbrain.
NSPinsmokingbehaviorRGuptaetal5TranslationalPsychiatry(2017),1–8SNPinNRG3wasassociatedwithNWsymptomcount,inlinewithndingsfromanimalmodelsandassociationanalysesreportedbyTurneretal.
15TheeffectsizesdetectedforERBB4andNRG3SNPsshowingassociationwithNWsymptomcountareprominent,correspondingtoadecreaseof0.
5–0.
8countsonascaleof0–8counts,pereachcopyoftheeffectallele.
Further,jointlinkageandLDanalysisdetectedasignalforSIinsevenoftheNSPgenes(majorityofthehighlightedSNPs),likelyduetoourdiscoverysamplebeingheavilyenrichedforsmoking(83%havinginitiatedsmoking)providingamplestatisticalpowerinourfamilysample.
Inourrarevariantgene-basedanalysesusingHMBR,weobservedsignicantassociationofERBB4,NRG1andPSEN1withSIandNWdiagnosis.
Overall,weprovidestrongevidenceimplyingtheNSPinsmokingbehaviorwhilehighlightingnovelassociationsinvegenes(NRG1,BACE1,APH1A,PSEN2andPSENEN)forsmokingbehavior.
Itisnoteworthythatamongthe14testedphenotypes(nineassessingsmokingbehaviorandvealcoholuse),thesignalconsistentlyemergesfor(i)mokinginitiation,supportingourpreviouslinkagendingforsmokinginitiationon2q33,over-lappingtheERBB4locus,14(ii)DSM-IVND,supportingourpreviouslyreportedassociationbetweenERBB4andDSM-IVND13and(iii)NW,supportingpreviousndingsbyTurneretal.
15NosignicantsignalwasdetectedforNDassessedbyFTNDorsmokingquantity.
Further,nosignalwasdetectedforalcoholusephenotypes,suggestingthattheroleoftheNSPisspecictonicotine,ratherthanbeingbroadlyinvolvedinaddictionsingeneral.
Lackofsignalforalcoholusephenotypesalsosuggeststhatsignalsobservedforsmoking-relatedpheno-typesarenotconfoundedbycomorbidalcoholuse.
FuturestudiesareneededtoassesstheinvolvementoftheNSPinuseanddependenceofothersubstancesbesidesnicotineandalcohol.
Inourdiscoverydataset,cannabisorotherillicitdruguseisveryrare.
64Altogether,38ofthe66highlightedSNPswereidentiedeitheraseQTLsormeQTLsintheDILGOMsample,suggestingthattheymayaffectexpressionandmethylationlevelsoftheNSPgenes.
Also,veofthe66SNPsweredetectedaseQTLsindifferentbraintissues(cerebellarcortex,cerebellarhemisphere,amygdalaandspinalcord)andfourasmeQTLsinfetalbrain,indicatingthatatleastsomeofourhighlightedSNPsmayaffectgeneexpressionormethylationinthebrain.
Threebrainregions(amygdala,cerebellarcortexandcerebellarhemisphere)showingeQTLsinourstudyhavepreviouslybeenimplicatedinaddictionstudies.
65–68Althoughspinalcord,anothertissueshowingeQTLsinourstudy,hasnotbeenstudiedpreviouslywithrespecttoaddictions,ourndingsfromGTExdatasuggestspinalcordshouldbeconsideredinaddictionstudies.
30Smokinghasaprominenteffectonmethylation,69andaltersgeneexpression.
70Unfortunately,mostpubliclyavailableeQTLandmeQTLdatabasesdonotcurrentlyprovideinformationonsmokingstatus.
Nevertheless,wecouldreplicatemosteQTLsandmeQTLsdetectedintheDILGOMsampleinblood.
Forexample,outofthe23NRG1eQTLsdetectedintheDILGOMsample(blood),17replicatedinGTEx(blood)butnonewereobservedinbraintissue(GTEx,BRAINEAC),however,amongthe20ERBB4meQTLsdetectedintheDILGOMsample(blood),17replicatedinmQTLdbdata(blood)and4ofwerealsoobservedinfetalbrain.
Suchinconsistentoverlapmaystemfromdifferencesingeneexpres-sionandmethylationlevelsacrosstissues.
AdditionalfactorscontributingtowardsincompleteoverlapofQTLresultsbetweenblood-derivedresultsinDILGOMversuspublicdatabases(GTExandmQTLdb)mayincludepopulation-specicgeneticback-groundandlackofdetailedphenotypicinformation(likesmokingstatus)availableinthedatabases.
Further,smallsamplesizesfortissuesthataredifculttoaccess,suchasbrain,reducethepowertoidentifysignals.
Inaddition,datamadeavailablefrompublishedstudiesoftenonlycontain(epi)genome-widesignicantresultstherebylimitingthepossibilityofreplicatingresultsoftargetedstudieslikeours.
Nonetheless,epigenome-widesignicantmeQTLsreportedforthehighlighted66SNPsinfetalbrainoverlapwithmeQTLsobservedinDILGOMsample(blooddata),indicatingveryrobustassociationsbetweenSNPsandmethylationlevelsofthesegenes.
ThefourERBB4promoterregionmeQTLsdetectedinbothbloodandfetalbrainshowedassociationwithNW.
ThesemeQTLswerenotdetectedinadultbraintissuesinthesamestudybyHannonetal.
;33assmokingwasnotaccountedforinthisstudy,gene-by-environmentinteractionsinducedbysmokingexposurecannotberuledout.
IffuturestudiesrevealthatmethylationattheseCpGsitesreacttosmokingexposure,thislocuscouldbeapotentialtargetforepigenetictherapyforNW.
Owingtothelimitedaccessandavailabilityofbraintissue,theuseofbloodasasubstitutefortranscriptlevelanalyseshasbeenevaluated,andamodestoverlapof~19%hasbeenobserved,emphasizingthevalueoftissuesthatarespecictothepathophysiologyofthetraitofinterest.
71,72However,bloodcanbeusedassurrogateforbraintissues,especiallyforgenesthatareco-expressedbetweenthetwotissues.
73AccordingtoGTEx,sevenoutofthetenNSPgenesareco-expressedinbrainandblood.
However,theexpressionlevelsdiffersignicantlyforalmostallthegenesbetweenbrainandblood.
AlternativesplicingofERBB4hasbeenreportedtoresultinfunctionallydistinctisoformsthatcanalterdownstreamsignaling.
74Interestingly,theND-associatedSNPrs13385826inERBB4ispredictedtobeasplicingdisruptionvariantaccordingtoSPANR,plausiblyinducingalternativesplicingofexon21encodingforpartofthekinasedomainofERBB4,crucialfordownstreamsignaling.
75Functionalvalidationofthesplicingproductofrs13385826isneededtoconrmthespecicERBB4isoformanditsconsequencestothepropertiesoftheprotein.
DownstreamtargetsoftheNSParebeingevaluatedasdrugtargetsinSCZ.
76,77DespiteonlyoneSNPbeinginthecodingregion,thereappearstoberegulatorypotentialin56ofthehighlighted66SNPsbymeansofeQTLs,meQTLsandotherregulatoryfeatures(overlappingenhancerand/orpromoterortranscriptionfactorbindingsites)(Figure2).
However,furthervalidationstudiesarerequiredtoconrmthepredictedfunctionoftheseSNPs.
Inourexpressionanalyses,weobserveddifferentialexpressionofNRG1andPSEN1betweencurrentdailyandneversmokersintheDILGOMsample,suggestingthatsmokinginuencestheexpressionofatleastsomeoftheNSPgenes.
Interestingly,mergingcurrentdailyandoccasionalsmokersdilutedthesignal.
Further,ourdatashowthatexpressionlevelsofoccasionalsmokersresemblemoreneversmokerscomparedwiththedailysmokers.
Occasionalsmokersbydenitionsmokeirregularlyandhencealternatebetweenexposureandnon-exposuretotobaccoanditschemicals.
Thisunderliestheimportanceofcarefuldenitionofsmokingstatusingene-expressionstudies.
DifferentialexpressioninNRG1,NRG3,PSEN1,PSEN2,PSENENandAPH1A(FDRPo0.
05)hasbeenreportedbetweenSCZcasesandcontrolsinSZDB;35strikingly,smokingstatushasnotbeenaccountedforinanyofthestudiesincludedinthisdatabase.
Inthecurrentstudy,wedetectedatrendfordifferentialexpressionofNRG1,NRG3andAPH1BinoursmallSCZtwinsample(N=73).
However,afteradjustingforsmokingstatusthetrenddisap-peared.
Giventhehighco-occurrenceandsharedgeneticcomponentofSCZandsmoking,78andourndings,theresultsfromSZDBmaypartlyreectsmokingstatusratherthanthediseasestatus,advocatingtheimportanceofaccountingforconfoundingeffectsofsmoking.
WedetectednosignicantdifferencesinNSPgenemethylationlevelsbetweencurrentdailyandneversmokersintheDILGOMsample.
ThissuggeststhatthemicroarrayusedinthisstudymaynothavecoveredrelevantCpGsites,orthatotherregulatorymechan-ismsbesidesmethylationaccountforthedetecteddifferencesinexpressionlevels.
PoolingoccasionalsmokerstogetherwithcurrentNSPinsmokingbehaviorRGuptaetal6TranslationalPsychiatry(2017),1–8dailysmokershadnoeffect,suggestingthatmethylationofNSPgenesdoesnotchangerapidlyinresponsetoabstinence.
Inouranalyses,weutilizedaFinnishtwinfamilysample,heavilyenrichedforsmoking.
Astheparticipantswereonaverage56yearsoldatthetimeofenrollment,wewerenotabletorecruitparentsbutrathertookadvantageofthelargenumberofaffectedsib-ships.
WeapplieddiligentqualitycontrolstoexcludelowqualitySNPs.
Althoughthewidelyusedthresholdofimputationinfoscoreof40.
4wasapplied,theaverageinfoscoreofthe66highlightedSNPswas0.
89(median0.
93),andonlytwoSNPshadaninfoscoreo0.
7.
NRG3SNPrs11192578showedmildLDwithneighboringSNPsplausiblyexplainedbysuggestiveevidenceofpositiveselection(iHSscore1.
7obtainedfromHaplotter(haplotter.
uchicago.
edu)79andlowminorallelefrequency(MAF=0.
04)inoursample.
ERBB4SNPrs13385826(showingassociationwithNDdiagnosis)appearstobeanorphansignalwithnosupportfromsurroundingSNPsdespitehavingseveralSNPsinhighLD.
Inconclusion,ourtwinfamilysampleprovidedfurtherevidencefortheinvolvementoftheNSPinsmokingbehaviorbutnotinalcoholuseandabusephenotypes.
Ourdifferentialexpressionanalyseshighlightedtheimportanceofcarefullydeningthesmokingstatusingene-expressionstudies.
OurresultsunderlinedtheinvolvementofERBB4inSI,NDandNW.
Further,wehighlightedNRG3forSIandNW,andadditionalveNSPgenesforSI.
Byusingbothin-housedataandpubliclyavailabledatabaseswedepictedpotentialfunctionforanexceptionallyhighproportion(56/66)oftheassociatedSNPs,suggestingthepotentialofthesevariantsasfuturedrugtargets.
CONFLICTOFINTERESTDrKaprioandDrKorhonenhaveprovidedconsultationtoPzeronnicotinedependenceanditstreatment.
DrCannonisaconsultanttotheLosAngelesCountyDepartmentofMentalHealthandtoBoehringerIngelheimPharmaceuticals.
Theremainingauthorsdeclarenoconictofinterest.
ACKNOWLEDGMENTSWewarmlythanktheparticipatingtwinpairsandtheirfamilymembers,aswellastheparticipantsoftheDILGOMstudyfortheircontribution.
DatacollectionoftheFinnishtwincohortsampleshasbeensupportedbytheAcademyofFinlandCenterofExcellenceinComplexDiseaseGenetics(Grants213506and129680),theAcademyofFinland(Grants265240and263278toJK),NIHGrantDA12854(toPAFM),SigridJuseliusFoundation(toJK),GlobalResearchAwardforNicotineDependence,Pzer(toJK),theWellcomeTrustSangerInstitute,UKandtheBroadInstituteofMITandHarvard,USA.
TheDILGOMstudyhasbeensupportedbytheAcademyofFinlandResearch(Grants255935toMP)forDNAmethylationdataandSigridJuséliusFoundationandYrjJahnssonFoundationforgene-expressiondata.
DNAmethyla-tiondatafortheDILGOMsamplewasgeneratedattheEstonianGenomeCenter,UniversityofTartu,underthesupervisionofProfAndresMetspalu.
REFERENCES1CentersforDiseaseControlandPrevention(US),NationalCenterforChronicDiseasePreventionandHealthPromotion(US),OfceonSmokingandHealth(US).
HowTobaccoSmokeCausesDisease:TheBiologyandBehavioralBasisforSmoking-AttributableDisease:AReportoftheSurgeonGeneral.
CentersforDiseaseControlandPrevention(US):Atlanta,GA,2010.
2DomeP,LazaryJ,KalaposMP,RihmerZ.
Smoking,nicotineandneuropsychiatricdisorders.
NeurosciBiobehavRev2010;34:295–342.
3LoukolaA,HallforsJ,KorhonenT,KaprioJ.
Geneticsandsmoking.
CurrAddictRep2014;1:75–82.
4AdamsS.
Psychopharmacologyoftobaccoandalcoholcomorbidity:areviewofcurrentevidence.
CurrAddictRep2017;4:25–34.
5BenowitzNL.
Pharmacologyofnicotine:addiction,smoking-induceddisease,andtherapeutics.
AnnuRevPharmacolToxicol2009;49:57–71.
6PergadiaML,AgrawalA,LoukolaA,MontgomeryGW,BromsU,SacconeSFetal.
GeneticlinkagendingsforDSM-IVnicotinewithdrawalintwopopulations.
AmJMedGenetBNeuropsychiatrGenet2009;150B:950–959.
7WestRJ,HajekP,BelcherM.
Severityofwithdrawalsymptomsasapredictorofoutcomeofanattempttoquitsmoking.
PsychologicalMed1989;19:981–985.
8CahillK,Lindson-HawleyN,ThomasKH,FanshaweTR,LancasterT.
Nicotinereceptorpartialagonistsforsmokingcessation.
CochraneDatabaseSystRev2016;5:CD006103.
9HatzimanolisA,McGrathJA,WangR,LiT,WongPC,NestadtGetal.
Multiplevariantsaggregateintheneuregulinsignalingpathwayinasubsetofschizo-phreniapatients.
TranslPsychiatry2013;3:e264.
10MeiL,NaveKA.
Neuregulin-ERBBsignalinginthenervoussystemandneuro-psychiatricdiseases.
Neuron2014;83:27–49.
11BuonannoA.
Theneuregulinsignalingpathwayandschizophrenia:fromgenestosynapsesandneuralcircuits.
BrainResBull2010;83:122–131.
12AmericanPsychiatricAssociation.
DiagnosticandStatisticalManualofMentalDisorders:DSM-IV,4thedn.
AmericanPsychiatricAssociation:Washington,DC,1994.
13LoukolaA,WedenojaJ,Keskitalo-VuokkoK,BromsU,KorhonenT,RipattiSetal.
Genome-wideassociationstudyondetailedprolesofsmokingbehaviorandnicotinedependenceinatwinsample.
MolPsychiatry2014;19:615–624.
14LoukolaA,BromsU,MaunuH,WidenE,HeikkilaK,SiivolaMetal.
Linkageofnicotinedependenceandsmokingbehavioron10q,7qand11pintwinswithhomogeneousgeneticbackground.
PharmacogenomicsJ2008;8:209–219.
15TurnerJR,RayR,LeeB,EverettL,XiangJ,JepsonCetal.
Evidencefrommouseandmanforaroleofneuregulin3innicotinedependence.
MolPsychiatry2014;19:801–810.
16FisherML,LoukolaA,KaprioJ,TurnerJR.
Roleoftheneuregulinsignalingpath-wayinnicotinedependenceandco-morbiddisorders.
IntRevNeurobiol2015;124:113–131.
17EdwardsSL,BeesleyJ,FrenchJD,DunningAM.
BeyondGWASs:illuminatingthedarkroadfromassociationtofunction.
AmJHumGenet2013;93:779–797.
18HuangQ.
Geneticstudyofcomplexdiseasesinthepost-GWASera.
JGenetGenomics2015;42:87–98.
19ChenCY,ChangIS,HsiungCA,WassermanWW.
OntheidenticationofpotentialregulatoryvariantswithingenomewideassociationcandidateSNPsets.
BMCMedGenomics2014;7:34.
20BromsU,WedenojaJ,LargeauMR,KorhonenT,PitkaniemiJ,Keskitalo-VuokkoKetal.
AnalysisofdetailedphenotypeprolesrevealsCHRNA5-CHRNA3-CHRNB4geneclusterassociationwithseveralnicotinedependencetraits.
NicotineTobRes2012;14:720–733.
21HeL,PitkaniemiJ,HeikkilaK,ChouYL,MaddenPA,KorhonenTetal.
Genome-widetime-to-eventanalysisonsmokingprogressionstagesinafamily-basedstudy.
BrainBehav2016;6:e00462.
22DelaneauO,HowieB,CoxAJ,ZaguryJF,MarchiniJ.
Haplotypeestimationusingsequencingreads.
AmJHumGenet2013;93:687–696.
23HowieBN,DonnellyP,MarchiniJ.
Aexibleandaccurategenotypeimputationmethodforthenextgenerationofgenome-wideassociationstudies.
PLoSGenet2009;5:e1000529.
24GenomesProjectConsortium,AbecasisGR,AutonA,BrooksLD,DePristoMA,DurbinRMetal.
Anintegratedmapofgeneticvariationfrom1,092humangenomes.
Nature2012;491:56–65.
25InouyeM,SilanderK,HamalainenE,SalomaaV,HaraldK,JousilahtiPetal.
Animmuneresponsenetworkassociatedwithbloodlipidlevels.
PLoSGenet2010;6:e1001113.
26InouyeM,KettunenJ,SoininenP,SilanderK,RipattiS,KumpulaLSetal.
Meta-bonomic,transcriptomic,andgenomicvariationofapopulationcohort.
MolSystBiol2010;6:441.
27LuostarinenM,TuovinenEL,SaarniSE,KinnunenT,HukkinenM,HaukkalaAetal.
WeightconcernsamongFinnishever-smokers:apopulation-basedstudy.
NicotineTobRes2013;15:1696–1704.
28BibikovaM,BarnesB,TsanC,HoV,KlotzleB,LeJMetal.
HighdensityDNAmethylationarraywithsingleCpGsiteresolution.
Genomics2011;98:288–295.
29AryeeMJ,JaffeAE,Corrada-BravoH,Ladd-AcostaC,FeinbergAP,HansenKDetal.
Min:aexibleandcomprehensiveBioconductorpackagefortheanalysisofInniumDNAmethylationmicroarrays.
Bioinformatics2014;30:1363–1369.
30GTExConsortium.
Humangenomics.
TheGenotype-TissueExpression(GTEx)pilotanalysis:multitissuegeneregulationinhumans.
Science2015;348:648–660.
31GauntTR,ShihabHA,HemaniG,MinJL,WoodwardG,LyttletonOetal.
Sys-tematicidenticationofgeneticinuencesonmethylationacrossthehumanlifecourse.
GenomeBiol2016;17:61.
32RamasamyA,TrabzuniD,GuelS,VargheseV,SmithC,WalkerRetal.
Geneticvariabilityintheregulationofgeneexpressionintenregionsofthehumanbrain.
NatNeurosci2014;17:1418–1428.
33HannonE,SpiersH,VianaJ,PidsleyR,BurrageJ,MurphyTMetal.
MethylationQTLsinthedevelopingbrainandtheirenrichmentinschizophreniariskloci.
NatNeurosci2016;19:48–54.
NSPinsmokingbehaviorRGuptaetal7TranslationalPsychiatry(2017),1–834ZheutlinAB,ViehmanRW,FortgangR,BorgJ,SmithDJ,SuvisaariJetal.
Cognitiveendophenotypesinformgenome-wideexpressionprolinginschizophrenia.
Neuropsychology2016;30:40–52.
35WuY,YaoYG,LuoXJ.
SZDB:adatabaseforschizophreniageneticresearch.
SchizophrBull2016;43:459–471.
36BenjaminiY,HochbergY.
Controllingthefalsediscoveryrate:apracticalandpowerfulapproachtomultipletesting.
JRStatSocB1995;57:289–300.
37Cuellar-PartidaG,RenteriaME,MacGregorS.
LocusTrack:IntegratedvisualizationofGWASresultsandgenomicannotation.
SourceCodeBiolMed2015;10:1.
38BarrettJC.
Haploview:visualizationandanalysisofSNPgenotypedata.
ColdSpringHarbProtoc2009;2009:pdb.
ip71.
39ZhouX,StephensM.
Genome-wideefcientmixed-modelanalysisforassociationstudies.
NatGenet2012;44:821–824.
40GertzEM,HiekkalinnaT,DigabelSL,AudetC,TerwilligerJD,SchafferAA.
PSEU-DOMARKER2.
0:efcientcomputationoflikelihoodsusingNOMAD.
BMCBioin-formatics2014;15:47.
41HiekkalinnaT,GoringHH,TerwilligerJD.
Onthevalidityofthelikelihoodratiotestandconsistencyofresultingparameterestimatesinjointlinkageandlinkagedisequilibriumanalysisunderimproperlyspeciedparametricmodels.
AnnHumGenet2012;76:63–73.
42GoringHH,TerwilligerJD.
LinkageanalysisinthepresenceoferrorsIV:jointpseudomarkeranalysisoflinkageand/orlinkagedisequilibriumonamixtureofpedigreesandsingletonswhenthemodeofinheritancecannotbeaccuratelyspecied.
AmJHumGenet2000;66:1310–1327.
43HiekkalinnaT,GoringHH,LambertB,WeissKM,NorrgrannP,SchafferAAetal.
Onthestatisticalpropertiesoffamily-basedassociationtestsindatasetscontainingbothpedigreesandunrelatedcase-controlsamples.
EurJHumGenet2012;20:217–223.
44TherneauTM.
coxme:MixedEffectsCoxModels.
RpackageVersion2.
2-3,2015.
Availablefrom:http://CRAN.
R-project.
org/package=coxme.
45VazquezAI,BatesDM,RosaGJ,GianolaD,WeigelKA.
Technicalnote:anRpackageforttinggeneralizedlinearmixedmodelsinanimalbreeding.
JAnimSci2010;88:497–504.
46LiB,LealSM.
Methodsfordetectingassociationswithrarevariantsforcommondiseases:applicationtoanalysisofsequencedata.
AmJHumGenet2008;83:311–321.
47WuMC,LeeS,CaiT,LiY,BoehnkeM,LinX.
Rare-variantassociationtestingforsequencingdatawiththesequencekernelassociationtest.
AmJHumGenet2011;89:82–93.
48HeL,PitkaniemiJ,SarinAP,SalomaaV,SillanpaaMJ,RipattiS.
HierarchicalBayesianmodelforrarevariantassociationanalysisintegratinggenotypeuncertaintyinhumansequencedata.
GenetEpidemiol2015;39:89–100.
49ThomasSellkeMJB,BergerJO.
Calibrationofρvaluesfortestingprecisenullhypotheses.
AmStat2001;55:62–71.
50StephensM,BaldingDJ.
Bayesianstatisticalmethodsforgeneticassociationstudies.
NatRevGenet2009;10:681–690.
51ShabalinAA.
MatrixeQTL:ultrafasteQTLanalysisvialargematrixoperations.
Bioinformatics2012;28:1353–1358.
52HousemanEA,AccomandoWP,KoestlerDC,ChristensenBC,MarsitCJ,NelsonHHetal.
DNAmethylationarraysassurrogatemeasuresofcellmixturedistribution.
BMCBioinformatics2012;13:86.
53ZeilingerS,KuhnelB,KloppN,BaurechtH,KleinschmidtA,GiegerCetal.
Tobaccosmokingleadstoextensivegenome-widechangesinDNAmethylation.
PLoSONE2013;8:e63812.
54McLarenW,GilL,HuntSE,RiatHS,RitchieGRS,ThormannAetal.
Theensemblvarianteffectpredictor.
GenomeBiol2016;17:122.
55WardLD,KellisM.
HaploRegv4:systematicminingofputativecausalvariants,celltypes,regulatorsandtargetgenesforhumancomplextraitsanddisease.
NucleicAcidsRes2016;44:D877–D881.
56XiongHY,AlipanahiB,LeeLJ,BretschneiderH,MericoD,YuenRKetal.
RNAsplicing.
Thehumansplicingcoderevealsnewinsightsintothegeneticdeter-minantsofdisease.
Science2015;347:1254806.
57TobaccoandGeneticsConsortium.
Genome-widemeta-analysesidentifymultiplelociassociatedwithsmokingbehavior.
NatGenet2010;42:441–447.
58LiuJZ,TozziF,WaterworthDM,PillaiSG,MugliaP,MiddletonLetal.
Meta-analysisandimputationrenestheassociationof15q25withsmokingquantity.
NatGenet2010;42:436–440.
59ThorgeirssonTE,GudbjartssonDF,SurakkaI,VinkJM,AminN,GellerFetal.
SequencevariantsatCHRNB3-CHRNA6andCYP2A6affectsmokingbehavior.
NatGenet2010;42:448–453.
60GoldAB,LermanC.
Pharmacogeneticsofsmokingcessation:roleofnicotinetargetandmetabolismgenes.
HumGenet2012;131:857–876.
61RayR,TyndaleRF,LermanC.
Nicotinedependencepharmacogenetics:roleofgeneticvariationinnicotine-metabolizingenzymes.
JNeurogenet2009;23:252–261.
62HoMK,TyndaleRF.
Overviewofthepharmacogenomicsofcigarettesmoking.
PharmacogenomicsJ2007;7:81–98.
63XiongM,JinL.
Comparisonofthepowerandaccuracyofbiallelicandmicro-satellitemarkersinpopulation-basedgene-mappingmethods.
AmJHumGenet1999;64:629–640.
64AgrawalA,PergadiaML,SacconeSF,LynskeyMT,WangJC,MartinNGetal.
Anautosomallinkagescanforcannabisusedisordersinthenicotineaddictiongeneticsproject.
ArchGenPsychiatry2008;65:713–721.
65VolkowND,FowlerJS,WangGJ.
Theaddictedhumanbrain:insightsfromima-gingstudies.
JClinInvest2003;111:1444–1451.
66MihovY,HurlemannR.
Alteredamygdalafunctioninnicotineaddiction:insightsfromhumanneuroimagingstudies.
Neuropsychologia2012;50:1719–1729.
67MiquelM,Vazquez-SanromanD,Carbo-GasM,Gil-MiravetI,Sanchis-SeguraC,CarulliDetal.
HavewebeenignoringtheelephantintheroomSevenargu-mentsforconsideringthecerebellumaspartofaddictioncircuitry.
NeurosciBiobehavRev2016;60:1–11.
68KuhnS,RomanowskiA,SchillingC,MobascherA,WarbrickT,WintererGetal.
Braingreymatterdecitsinsmokers:focusonthecerebellum.
BrainStructFunct2012;217:517–522.
69JoehanesR,JustAC,MarioniRE,PillingLC,ReynoldsLM,MandaviyaPRetal.
Epigeneticsignaturesofcigarettesmoking.
CircCardiovascGenet2016;9:436–447.
70HuanT,JoehanesR,SchurmannC,SchrammK,PillingLC,PetersMJetal.
Awhole-bloodtranscriptomemeta-analysisidentiesgeneexpressionsignaturesofcigarettesmoking.
HumMolGenet2016;25:4611–4623.
71McKenzieM,HendersAK,CaracellaA,WrayNR,PowellJE.
Overlapofexpressionquantitativetraitloci(eQTL)inhumanbrainandblood.
BMCMedGenomics2014;7:31.
72HernandezDG,NallsMA,MooreM,ChongS,DillmanA,TrabzuniDetal.
Inte-grationofGWASSNPsandtissuespecicexpressionprolingrevealdiscreteeQTLsforhumantraitsinbloodandbrain.
NeurobiolDis2012;47:20–28.
73DaviesMN,LawnS,WhatleyS,FernandesC,WilliamsRW,SchalkwykLC.
Towhatextentisbloodareasonablesurrogateforbrainingeneexpressionstudies:estimationfrommousehippocampusandspleen.
FrontNeurosci2009;3:54.
74VeikkolainenV,VaparantaK,HalkilahtiK,IljinK,SundvallM,EleniusK.
FunctionofERBB4isdeterminedbyalternativesplicing.
CellCycle2011;10:2647–2657.
75BoseR,ZhangX.
TheErbBkinasedomain:structuralperspectivesintokinaseactivationandinhibition.
ExpCellRes2009;315:649–658.
76PapaleoF,YangF,PatersonC,PalumboS,CarrGV,WangYetal.
Behavioral,neurophysiological,andsynapticimpairmentinatransgenicNeuregulin1(NRG1-IV)MurineSchizophreniaModel.
JNeurosci2016;36:4859–4875.
77RicoB.
Findingadruggabletargetforschizophrenia.
ProcNatlAcadSciUSA2012;109:11902–11903.
78HartzSM,HortonAC,HancockDB,BakerTB,CaporasoNE,ChenLSetal.
Geneticcorrelationbetweensmokingbehaviorsandschizophrenia.
SchizophrRes2017;pii:S0920-9964(17)30111-1.
doi:10.
1016/j.
schres.
2017.
02.
022[e-pubaheadofprint].
79VoightBF,KudaravalliS,WenX,PritchardJK.
Amapofrecentpositiveselectioninthehumangenome.
PLoSBiol2006;4:e72.
ThisworkislicensedunderaCreativeCommonsAttribution4.
0InternationalLicense.
Theimagesorotherthirdpartymaterialinthisarticleareincludedinthearticle'sCreativeCommonslicense,unlessindicatedotherwiseinthecreditline;ifthematerialisnotincludedundertheCreativeCommonslicense,userswillneedtoobtainpermissionfromthelicenseholdertoreproducethematerial.
Toviewacopyofthislicense,visithttp://creativecommons.
org/licenses/by/4.
0/TheAuthor(s)2017SupplementaryInformationaccompaniesthepaperontheTranslationalPsychiatrywebsite(http://www.
nature.
com/tp)NSPinsmokingbehaviorRGuptaetal8TranslationalPsychiatry(2017),1–8