Genome-wide association screens for Achilles tendon and ACL tears and tendinopathy

16 Pages·2015·1.12 MB·English

RESEARCHARTICLE
Genome-wide association screens for Achilles
tendon and ACL tears and tendinopathy
StuartK.Kim1*,ThomasR.Roos1,2,AndrewK.Roos1,2,JohnP.Kleimeyer3,Marwa
A.Ahmed3,GabrielleT.Goodlin4,MichaelFredericson3,JohnP.A.Ioannidis5,6,7,Andrew
L.Avins8,JasonL.Dragoo3*
1 DepartmentDevelopmentalBiology,StanfordUniversityMedicalCenter,StanfordCA,UnitedStatesof
America,2 DepartmentHealthResearchandPolicy,DivisionofEpidemiology,StanfordUniversityMedical
Center,StanfordCA,UnitedStatesofAmerica,3 DepartmentOrthopaedicSurgery,StanfordUniversity
MedicalCenter,StanfordCA,UnitedStatesofAmerica,4 CollegeofMedicine,CaliforniaNorthstate
a1111111111 University,ElkGroveCA,UnitedStatesofAmerica,5 DepartmentofMedicine,StanfordPrevention
a1111111111 ResearchCenter,StanfordUniversitySchoolofMedicine,StanfordCA,UnitedStatesofAmerica,
6 DepartmentofHealthResearchandPolicy,DivisionofEpidemiology,StanfordUniversitySchoolof
a1111111111
Medicine,StanfordCA,UnitedStatesofAmerica,7 DepartmentofStatistics,StanfordUniversitySchoolof
a1111111111
HumanitiesandSciences,StanfordCA,UnitedStatesofAmerica,8 KaiserPermanenteNorthernCalifornia,
a1111111111 DivisionofResearch,Oakland,California,UnitedStatesofAmerica
*[email protected](SKK);[email protected](JLD)
OPENACCESS Abstract
Citation:KimSK,RoosTR,RoosAK,Kleimeyer
Achillestendinopathyorruptureandanteriorcruciateligament(ACL)rupturearesubstantial
JP,AhmedMA,GoodlinGT,etal.(2017)Genome-
wideassociationscreensforAchillestendonand injuriesaffectingathletes,associatedwithdelayedrecoveryorinabilitytoreturntocompeti-
ACLtearsandtendinopathy.PLoSONE12(3): tion.Toidentifygeneticmarkersthatmightbeusedtopredictriskfortheseinjuries,weper-
e0170422.https://doi.org/10.1371/journal.
formedgenome-wideassociationscreensfortheseinjuriesusingdatafromtheGenetic
pone.0170422
EpidemiologyResearchonAdultHealthandAging(GERA)cohortconsistingof102,979
Editor:JamesH-CWang,UniversityofPittsburgh,
individuals.Wedidnotfindanysinglenucleotidepolymorphisms(SNPs)associatedwith
UNITEDSTATES
eitheroftheseinjurieswithap-valuethatwasgenome-widesignificant(p<5x10-8).We
Received:September2,2016
found,however,fourandthreepolymorphismswithp-valuesthatwereborderlinesignificant
Accepted:January4,2017 (p<10−6)forAchillestendoninjuryandACLrupture,respectively.WethentestedSNPs
Published:March30,2017 previouslyreportedtobeassociatedwitheitherAchillestendoninjuryorACLrupture.
Noneshowedanassociationinourcohortwithafalsediscoveryrateoflessthan5%.We
Copyright:©2017Kimetal.Thisisanopen
accessarticledistributedunderthetermsofthe obtained,however,moderatetoweakevidenceforreplicationinonecase;specifically,
CreativeCommonsAttributionLicense,which rs4919510inMIR608hadap-valueof5.1x10-3forassociationwithAchillestendoninjury,
permitsunrestricteduse,distribution,and
correspondingtoa7%chanceoffalsereplication.Finally,wetested2855SNPsin90candi-
reproductioninanymedium,providedtheoriginal
authorandsourcearecredited. dategenesformusculoskeletalinjury,butdidnotfindanythatshowedasignificantassocia-
tionbelowafalsediscoveryrateof5%.Weprovidedatacontainingsummarystatisticsfor
DataAvailabilityStatement:Allrelevantdataare
withinthemanuscript,supportinginformationfiles, theentiregenome,whichwillbeusefulforfuturegeneticstudiesontheseinjuries.
andhostedatthefollowingURL:http://cmgm.
stanford.edu/~kimlab/ACL/Achilles_ACL.html.Data
willalsobeavailableatNIHGRASP:https://grasp.
nhlbi.nih.gov/FullResults.aspx.
Funding:ThisworkwassupportedbyNational
Introduction
InstituteofAging(5RO1AG025941)toSKK.
Achillestendinopathyorruptureandanteriorcruciateligament(ACL)rupturesarefrequent
Competinginterests:Theauthorshavedeclared
thatnocompetinginterestsexist. sourcesofpainanddysfunctioninrecreationalandeliteathletes.Recentstudieshavetesteda
PLOSONE|https://doi.org/10.1371/journal.pone.0170422 March30,2017 1/16
Genome-wideassociationscreensforAchillestendonandACLtearsandtendinopathy
fewsingle-nucleotidepolymorphisms(SNPs)inasmallnumberofcandidategenesforassoci-
ationwithAchillestendoninjuryorACLruptureinathletes.ForAchillestendoninjury,
studiesfound19DNAvariationsresidingin12genesthatwereassociatedwithAchillestendi-
nopathyatp<0.05usingcohortscontainingbetween52and184athletes[1–12].Inonecase
(rs12722inCOL5A1),theassociationwithAchillestendinopathywasfoundtoreplicateinan
independentcohort(p=.024)[4].ForACLrupture,studieshavepreviouslyfoundnineDNA
variationsineightgeneswithp<.05wherethenumberofcasesrangedbetween86and358
[7,13,14–22].rs180012inCOL1A1wasreplicatedinseveralstudies[14–18].Noneofthe
remainingSNPshavebeenreplicatedforeitherAchillestendoninjuryorACLrupture.
Thepurposeofthisstudywastoperformascreenoftheentiregenomeforpolymorphisms
associatedwitheitherAchillestendoninjuryorACLruptureinalargedataset.Wealsoana-
lyzedspecificSNPspreviouslyreportedtobeassociatedwitheitherinjury.
Materialsandmethods
Genome-wideassociationscreenswereperformedforeitherAchillestendoninjury(defined
astendinopathy,ruptureorrepair)orACLruptureusingdatafromtheGeneticEpidemiology
ResearchonAdultHealthandAging(GERA)cohort.Thegenerationofthedataandpipeline
fordataanalysishavebeenpreviouslydescribedinJorgensonetal.,2015andRoosetal.,2016
[23][24].
GERAcohort
TheGERAcohortiscomprisedof110,266adultmenandwomenmembersofKaiserPerma-
nenteNorthernCalifornia(KPNC)MedicalCarePlan.ItisacomponentoftheKPNC
ResearchProgramonGenes,EnvironmentandHealth(RPGEH).Acompletedescriptionof
thecohortandstudydesigncanbefoundindbGaP(StudyAccession:phs000674.v1.p1).The
averageageoftheparticipantsatthetimeofsamplecollectionwas62.9yearsold(standard
deviation=13.8years).LengthofmembershipinKPNCaveraged23.5years.Themedical
recordcontainstheentirepatienthistory,includinginjuriesthatoccurredpriortoenrollment
inKPNC,ifreportedbythepatientandrecordedbythephysician.
Ouranalysiscohort(n=102,979)includes59,671females,43,242males,and66individuals
ofambiguoussex.SexwasdeterminedbyheterozygosityontheXchromosome[23].For66
individuals,heterozygosityoftheXchromosomewasindeterminate,sonosexcallwasmade.
Moreover,ouranalysiscohortisethnicallydiverse,including83,264European-White(EUR);
8,560Latino(LAT);7,518EastAsian(EAS);3,161African-American(AFR);and476South
Asian(SAS)individualsbasedonancestryprinciplecomponents.
Genotyping,qualitycontrol,imputation,geneticancestry
Participantsweregenotypedatover650,000SNPsonfourancestrygroup-specificAffymetrix
Axiomgenome-widearraysoptimizedforindividualsofEuropean(EUR),African-American
(AFR),EastAsian(EAS),andLatino(LAT)ancestrygroup[25,26].Genotypequalitycontrol
proceduresfortheGERAcohortwereperformedasdescribedpreviously,exceptthatonly
SNPswithaminorallelefrequencyof>1%wereanalyzed[27].ThefinalnumberofSNPsthat
weredirectlygenotypedwas670,572forEUR;802,186forLAT;708,373forEAS;and878,176
forAFRarrays.
Genotypeswerepre-phasedandthenimputedtoacosmopolitanreferencepanelconsisting
ofallindividualsfromthe1000GenomesProject[24].ThefinalnumberofSNPsthatwere
imputedwas9,207,988forEUR(9,878,560total);10,380,912forLAT(11,183,098total);
8,355,578forEAS(9,063,951total);and16,659,640forAFRarrays(17,537,816total).
PLOSONE|https://doi.org/10.1371/journal.pone.0170422 March30,2017 2/16
Genome-wideassociationscreensforAchillestendonandACLtearsandtendinopathy
Determinationofgeneticancestrywasperformedbyprincipalcomponentanalysis(PCA),
aspreviouslydescribed[24].TheseancestryprincipalcomponentswereusedintheGWASto
adjustforgeneticancestry.
Phenotypedefinition
AchillestendoninjurieswereidentifiedintheGERAcohortbasedonclinicaldiagnosesand
surgicalprocedurescapturedintheKPNCElectronicHealthRecordsystem.Caseswere
definedasindividualswithatleastoneInternationalClassificationofDisease,NinthRevision
(ICD9)orCommonProcedureTerminology,FourthEdition(CPT4)code:ICD726_71,
ICD727_67,and/orCPT27650,describingAchillesbursitisortendinitis,nontraumaticrup-
tureofAchillestendon,andprimaryrepairofAchillestendonrupture,respectively(Table1).
TheElectronicMedicalRecordincludesinjuriesovertheentirelifetimeofthepatients;i.e.
thosethatoccurredpriortoenrollmentinKPNCaswellasthosethatoccurredafterthegeno-
typinganalysiswasperformed,anduntilthisstudywasinitiatedJuly22,2015.
ForACLrupture,KPNCpatientswithanypotentialACLinjurywereidentifiedbysearch
oftheelectronicmedicalrecordfortheICD9codes717.83and844.2andtheCPT4codes
29888and27407atanytime.Theimagingstudiesandsurgicalreportsfromthesepatients
werereviewedandthosewhohadstrongevidenceforafullorpartialACLruptureonMRI
and/orunderwentACLreconstructionwereconsideredtohavehadanACLrupture.The
chartsoftheremainingpatientswerethenmanuallyreviewedbyoneofthestudyphysicians
(AA)withtheassistanceofthestudyorthopedist(JD).Thosepatientswhohadconfirmatory
evidenceofACLrupture(e.g.,unambiguoushistoryinaprogressnote)werealsoclassifiedas
havinghadanACLrupture.PatientswithambiguoushistoriesorACLinjurywithoutrupture
werenotconsideredtohavehadafullorpartialACLrupture.
Genome-wideassociationandmeta-analysis
WetestedforSNPassociationswithalogisticregressionmodelusingalleleinanadditive
geneticmodelforeachofthefiveancestrygroupsseparately,asinRoosetal.,2016[24].The
modelwasadjustedforsex,ageatenrollmentintotheRPGEHcohort,ancestrygroupusing
principalcomponents,andgenotypingchipversionforallpopulations.
ForAchillestendinopathyorrupture,theprincipalcomponentsusedwere:AFR(6PCs),
EAS(6PCs),EUR(10PCs),LAT(6PCs)andSAS(6PCs).ForACLrupture,therewereonly
7casesintheSASandAFRpopulations,andsothesepopulationswerenotanalyzedfurther.
FortheremainingpopulationsforACLrupture,theprincipalcomponentsusedwere:EAS(2
PCs),EUR(4PCs)andLAT(3PCs).10,551,193SNPswereprocessedinthemeta-analysisfor
Achillestendinopathyorruptureand8,303,052SNPswereprocessedforACLrupturemeta-
analysis.
Table1. InternationalStatisticalClassificationofDiseasesandRelatedHealthProblemsandCurrent
ProceduralTerminologycodesforAchillesTendinopathy.
Description Code N(%)a
Achillesbursitisortendinitis ICD726_71 4,949(96)
Non-traumaticruptureofAchillestendon ICD727_67 105(2)
Repair,primary,openorpercutaneous,rupturedAchillestendon CPT27650 215(4)
Totalnumberindividuals 5,148(100)
aNumberofindividualswiththatcodeintheirelectronicmedicalrecords,andpercentofthetotalnumberof
individualsinparentheses.Someindividualshadmorethanonecode.
https://doi.org/10.1371/journal.pone.0170422.t001
PLOSONE|https://doi.org/10.1371/journal.pone.0170422 March30,2017 3/16
Genome-wideassociationscreensforAchillestendonandACLtearsandtendinopathy
Toassessforinflationduetopopulationstratification,thegenomiccontrolparameter(λ)
wascalculated[28].ForAchillestendinopathyorrupture,genomicinflationwassmallfor
GWAanalysesintheEUR,LAT,EAS,AFRpopulations(λbetween.99and1.02),butsubstan-
tialintheSASpopulation(λ~0.37).ForACLrupture,λwasbetween.93and1.0fortheEAS,
EURandLATpopulations.Subsequently,p-valueswereadjustedforgenomiccontrolineach
population.Thelinkagedisequilibriumscoremethodwasusedtoevaluatethegenomicinfla-
tionfactorusingthepythonscriptldscfromBulik-Sullivanetal.2015[29].LDHubwasused
tocomputethegeneticcorrelationbetweentheGWASforthetwoinjuriesfromthispaper
andthosefrom198publishedstudies(http://ldsc.broadinstitute.org/)[30].Theinputdatacon-
tainedtheunadjustedp-valuesfromtheEuropeanpopulationsfromeithertheAchillestendon
injuryortheACLinjuryGWAS.SNPsthatoverlappedthoseinHAPMAP3wereretainedfor
analysis.ForboththeAchillesandACLinjuryphenotypes,therewasalowinterceptoftheLD
scoreregression,indicatinglackofbiasfrompopulationstructure(S1Table).Resultsfrom
eachpopulationwerecombinedbyinverse-variance,fixed-effectsmeta-analysisandbyin-
verse-variance,random-effectsmeta-analysis.Resultsarereportedusingfixed-effectsmeta-
analysisp-values.SNPswerediscardedfromthemeta-analysisifdataweremissingfromthe
EURpopulation,whichcomprisesover80%ofthecohort.
Thefixed-effectsandrandom-effectsp-valuesforalloftheSNPstestedinthisstudyare
availablefromNIHGRASP:https://grasp.nhlbi.nih.gov/FullResults.aspx.Thesedatacouldbe
usefulforfuturestudiesonthegeneticriskforAchillestendoninjuryorACLrupture.
ReplicationanalysisofpreviouslyreportedSNPs
Wesearchedourmeta-analysisresultsforpreviously-publishedgeneticassociationswith
eitherAchillesinjuryorACLrupture[1,2,4–12,13,14–22].Wesearchedthepublicdatabases
PubMed/MEDLINE(http://www.ncbi.nlm.nih.gov/pubmed)andScopus(http://www.scopus.
com/)forpreviouslypublishedstudiesonthegeneticsofAchillestendoninjuryorACLrup-
ture.Inclusioncriteriaforfilteringresultswere:researchpaperswithoriginaldata,Englishlan-
guage,full-textarticleavailable(orabstractwithenoughdata),geneticassociationstudieswith
humansubjects,caseswithAchillestendoninjuryorACLrupturephenotypes,andgeneor
SNPassociationsreported.Articleswereincludedregardlessofwhetherornottheassociations
weresignificantandregardlessofthesizeofthestudypopulation.Furthermore,thereferences
fromeachpublicationwereusedtofindotherarticlesnotcapturedintheinitialsearch.The
finalsearcheswereconductedonJune1,2016.ABenjamini-Hochbergfalsediscoveryrate
threshold(q=0.05)wasusedtoaccountformultipletesting[31].
AnalysisofnovelcandidateSNPs
BasedonourknowledgeofbiologicalfunctionsoftheAchillestendonandACL,wegenerated
alistof90candidategenespotentiallyassociatedwitheitherAchillesorACLinjury.These
genescodeforstructuralcomponentsorthedevelopmentofligamentsand/ortendons.Aset
ofallknownSNPsfoundwithin2kbofthestartandendofthesegeneswasgeneratedusing
SCAN(http://www.scandb.org/)[31].RareSNPs(MAF<0.005)wereremovedbecausethey
wouldnothaveenoughstatisticalpowertoshowanassociationingeneassociationstudies.
SNPsnotfoundinourGWASresultswereremoved,yielding14,734SNPs.Toselectasingle
tagSNPineachlinkagedisequilibriumblock,SNPpruningwithanLDthreshold(r2>0.5)was
performedinPLINKv1.90(b3.34),resultingin2,855candidateSNPs.Forthecandidategenes
thathadbeentestedearlier,previousstudiestestedonlyoneorafewSNPsfromeachgene
whereasourlistcontainsmanytagSNPsspanningtheentiregene.Becausewetestedmany
candidateSNPs,weusedtheBenjamini-Hochbergmethodtosetthefalsediscoveryrateto
PLOSONE|https://doi.org/10.1371/journal.pone.0170422 March30,2017 4/16
Genome-wideassociationscreensforAchillestendonandACLtearsandtendinopathy
q=.05;i.e.thetopSNPinthelistwouldrequireap-valueof1.7x10-5inordertobedeemed
significant[32].
Gene-basedtestingwasperformedusingtheVErsatileGene-bASedtestVEGAS2[33].
2855SNPsfrom90candidategenesandtheirfixed-effectsp-valuesfromeithertheACLrup-
tureortheAchillestendoninjurymeta-analyseswereused.DNAvariantsthatwerenotSNPs
werefilteredfromtheinputfiles.Gene-basedanalysiswasperformedusingtheonline
VEGAS2softwarefromhttps://vegas2.qimrberghofer.edu.au/[33].
Ethicalconsiderations
ThisstudyanalyzedstoreddatafromRPGEH.Thehealthandgenotypedataforthesubjects
werede-identified.AllstudyprocedureswereapprovedbytheInstitutionalReviewBoardof
theKaiserFoundationResearchInstitute.
Results
Studypopulationandgenotypeinformation
WeobtainedaccesstoinjuryinformationandgenotypedatafromtheResearchProgramon
Genes,EnvironmentandHealthcohort(MaterialsandMethods).Thisprogramincludes
genotypeandmedicaldatafrom102,979patientsintheKaiser-PermanentesysteminNorth-
ernCalifornia.ForAchillestendoninjury,weinterrogatedtheelectronicmedicalrecordsof
theseindividualsforthosethathadincurredtendinopathyorarupture(Table1)(Materials
andMethods).ForACLrupture,casesweredefinedaspatientsthatshowedACLruptureby
MRIorthathadundergoneaprocedureforACLreconstruction(Methods).Intotal,there
were5,148casesofAchillestendoninjury(consistingof290withrupturesand4,858withbur-
sitisortendinitis)from102,979individuals.ForACLrupture,therewereonly7caseseach
fromtheAFRandSASpopulations,sotheseancestrygroupswereexcludedfromfurtheranal-
ysis.Thisleft598casesofACLrupturefrom99,342individualsintheEAS,EURandLAT
ancestrygroups.DescriptivefactorsfortheseinjuriesareshowninTable2.ForACLrupture,
theaverageageofthecasesis10yearslessthanthatofthecontrols(p<10−100).Onepossible
explanationisthattherecouldbeanascertainmentbiasagainstACLrupturesinolderpatients
thatincurredtheinjurywhentheywereyoungbeforeMRIwascommonlyused.
Genome-widestudiesforassociationwitheitherachillestendonInjuryor
ACLrupture
Fig1showsQQplotswiththeobservedp-valuesfromtheGERAcohortcomparedtothep-
valuesthatwouldbeexpectedbychance.Foreitherinjury,thep-valuesfromtheassociation
studyroughlyoverlapthosethatwouldbeexpectedbychance.
WeusedLinkageDisequilibriumScoreRegressiontoanalyzetheGWASdatafor:1)herita-
bility,2)fractionofthegeneticassociationsignalderivedfrompolygenicassociations,and3)
overlapwithGWASdataforothertraits[29,30].First,geneticheritabilityreferstoamountof
thetraitthatcanbeexplainedbythesumofalloftheSNPsintheGWAS.ForbothAchilles
tendoninjuryandACLrupture,theheritabilitywaslow(0.5–1.1%)(S1Table).Second,LD
scoreregressionisabletodiscernwhethertheobservedgeneticassociationsareduetoassocia-
tionswithsmalleffectsfrommanylociversusassociationsduetopopulationstructure.Ifthe
associationsareduetopolygenicassociations,thentheinterceptfromLDscoreregressionwill
belessthanthegenomicinflationfactor.ForbothAchillestendoninjuryandACLrupture,
LDscoreregressionsuggeststhatpopulationstructuredidnothaveastronginfluenceinthe
observedassociationresults.Third,theamountofgeneticcorrelationbetweentheresultsfrom
PLOSONE|https://doi.org/10.1371/journal.pone.0170422 March30,2017 5/16
Genome-wideassociationscreensforAchillestendonandACLtearsandtendinopathy
Table2. DescriptivefactorsforAchillestendoninjuryandACLrupture.
Achillestendoninjury
Cases(%) Controls(%) Total
Overall 5,148(5.0%) 97,831(95.0%) 102,979
AncestryGroup
European 4,258(5.1%) 79,006(94.9%) 83,264
Latin-American 413(4.8%) 8,147(95.2%) 8,560
EastAsian 268(3.6%) 7,250(96.4%) 7,518
African 192(6.1%) 2,969(93.9%) 3,161
South-EastAsian 17(3.6%) 459(96.4%) 476
Sex
Female 2,934(4.9%) 56,737(95.1%) 59,671
Male 2,211(5.1%) 41,031(94.9%) 43,242
Unknown 3(4.5%) 63(95.5%) 66
Agea 62.3(62.2–62.4)b 62.7(62.6–62.8) 62.6(62.5–62.7)
ACLrupture
Cases(%) Controls(%) Total
Overall 598(.61%) 98,744(99.39%) 99,342
AncestryGroup
European 495(.59%) 82,769(99.61%) 83,264
Latin-American 54(.63%) 8,506(99.37%) 8,560
EastAsian 49(.65%) 7,469(99.35%) 7,518
Sex
Female 349(.61%) 57,257(99.39%) 57,606
Male 249(.59%) 41,421(99.41%) 41,670
Unknown 0(0%) 66(100%) 66
Agea 52.0(50.8–53.1)c 62.7(62.6–62.8) 62.6(62.5–62.7)
aMeanage(years)atenrollment(95%CI).
bThereisasmalldifferenceinagesbetweencasesandcontrols(p=0.01).
cThedifferenceinagesbetweenthecasesandcontrolsishighlysignificant(p<10−100).
https://doi.org/10.1371/journal.pone.0170422.t002
thetwostudiesanalyzedherewiththosefrom198previouslypublishedstudieswasdeter-
minedusingLDHub(MaterialsandMethods)[30].NeitheroftheGWASresultsfromAchilles
tendoninjuryorACLruptureshowedasignificantgeneticcorrelationwithanyofthe198
datasets(S2Table).
Weplottedthep-valueforeverySNPonManhattanplots(Fig2).WedidnotfindanySNP
thatmatchedthegenome-widestatisticalthresholdofp<5x10-8.Table3liststheindependent
SNPsthatreachp<10−6.TherewerefourSNPsforAchillestendoninjuryandthreeSNPsfor
ACLrupture.
Re-testingcandidateSNPsforassociationwithachillestendonInjuryor
ACLrupture
PreviousstudieshavereportedcandidateSNPsthathaveshownanassociationwitheither
AchillestendoninjuryorACLruptureusingp<0.05asacutoff.Specifically,19SNPsin12
candidategeneshavebeenreportedtoshowanassociationwithAchillestendinopathy(using
p<0.05asacutoff)[1,2,4–12].Likewise,nineSNPsineightcandidategeneshavebeenreported
toshowanassociationwithACLrupture[7,13,14–22].
PLOSONE|https://doi.org/10.1371/journal.pone.0170422 March30,2017 6/16
Genome-wideassociationscreensforAchillestendonandACLtearsandtendinopathy
Fig1. QQplotsforAchillestendoninjury(A)orACLrupture(B).Theplotsshowtheobservedp-values
plottedonthey-axis(blackdots)againstthep-valuesexpectedbychanceplottedonthex-axis(redline).The
blackdotsaligncloselywiththeredlineindicatingthatthereislittleornostatisticalsignalineithertheAchilles
tendoninjuryorACLrupturecohorts.
https://doi.org/10.1371/journal.pone.0170422.g001
ForAchillestendinopathyorrupture,14oftheseSNPswerecontainedinourdataset
whereasforACLrupture,6SNPswerepresent.Weattemptedtoreplicatethepreviousresults
forthesecandidateSNPsusingourdataset.Becausewearetestingonlyasmallnumberof
SNPs,thethresholdforstatisticalsignificancecanbemuchlowerthanthegenome-wide
thresholdthatweusedforthegenome-widestudyabove(p<5x10-8).Itisstillimportant,
however,toadjustthep-valuethresholdtocompensateformultipletesting.Weusedthe
PLOSONE|https://doi.org/10.1371/journal.pone.0170422 March30,2017 7/16
Genome-wideassociationscreensforAchillestendonandACLtearsandtendinopathy
Fig2. ManhattanplotsforAchillestendoninjury(A)orACLrupture(B).Shownarethep-valuesforassociationwithAchillestendoninjuryorACL
rupture.Chromosomenumberisindicatedacrossthebottom.Y-axesshowthe-log p-valueforassociationwitheitherAchillestendoninjuryorACL
10
rupture.Theredlineindicatesthethresholdforgenome-widesignificance(p=5x10-8)andthebluelineindicatesp=5x10-5.
https://doi.org/10.1371/journal.pone.0170422.g002
Benjamini-Hochbergmethodtosetthefalsediscoveryratetoq=0.05.Inaddition,theas-
sociationshouldshowthesamedirectionofeffect(i.e.sameriskallele)asintheprevious
publication.
Table3. Tophitsfromthegenome-widescreenforAchillestendoninjuryandACLrupture(p<1x10-6).
Achillestendoninjury
SNP EAa MAFb P(FE)c OR(FE)(95%CI)d P(RE) OR(RE)(95%CI)
rs1937810 C 0.17 1.5x10-7 1.16(1.11–1.21) 1.5x10-7 1.16(1.11–1.21)
rs57104447 C 0.04 3.0x10-7 1.28(1.19–1.37) .01 1.25(1.16–1.34)
rs57224706 G 0.03 6.0x10-7 1.36(1.24–1.48) 6.0x10-7 1.36(1.24–1.48)
rs60713544 C 0.02 7.2x10-7 1.43(1.29–1.57) 7.2x10-7 1.43(1.29–1.57)
ACLrupture
SNP EAa MAFb P(FE)c OR(FE)(95%CI)d P(RE) OR(RE)(95%CI)
rs4067493 G 0.04 2.4x10-7 1.94(1.69–2.19) 3.1x10-5 2.06(1.81–2.31)
rs113435565 C 0.03 4.0x10-7 1.91(1.66–2.16) 4.0x10-7 1.91(1.66–2.16)
rs11960097 G 0.08 5.9x10-7 1.54(1.37–1.71) 5.9x10-7 1.54(1.37–1.71)
aEffectallele.
bMinorAlleleFrequencyinthecontrolpopulation.
cP-valuefromfixedeffects(FE)orrandomeffects(RE)meta-analysis.
dAllelicoddsratio(95%confidenceinterval)fortheeffectallele.FE,fixedeffect;RE,randomeffect.
https://doi.org/10.1371/journal.pone.0170422.t003
PLOSONE|https://doi.org/10.1371/journal.pone.0170422 March30,2017 8/16
Genome-wideassociationscreensforAchillestendonandACLtearsandtendinopathy
ForAchillestendinopathyorrupture,noneofthe14SNPswerefoundtobesignificant
usingtheBenjamini-Hochbergthreshold(Table4).However,rs1045485andrs4789932
showedsmallp-valuesofp<0.01,butinbothcasesthedirectionoftheeffectinourstudywas
oppositetothedirectionreportedpreviously(Table4).ForACLrupture,noneoftheSNPs
showedanassociationinourdatasetusingtheBenamini-Hochbergcutoff(Table4).
Inourgeneassociationstudy,weusedsex,ageandancestrygroupascovariateswhereas
previousstudiesdidnotusethesecovariates.Tocontrolforthisdifference,were-testedthe
candidateSNPswithoutusingcovariatesinordertoalignouranalysiswiththosethathad
beendonepreviously(S3Table).ForAchillestendinopathyorrupture,oneSNPinMIR608
(rs4919510)showedalowp-value(p=5.1x10-3)andhadthesamedirectionofeffect(Gispro-
tective)asseenpreviously[5].Thep-valueforrs4919510(p=5.1x10-3)isclosetothep-value
thresholdsetusingtheBenjamini-Hochbergmethod(p=3.5x10-3);i.e.ap-valueof5.1x10-3
hasabouta7%chancetobeafalsereplicationusing14tests.Additionally,fourotherSNPs
(rs1045485,rs591058,rs679620andrs4789932)showedasmallp-valuebuttheirdirectionof
effectwasoppositetotheeffectseenpreviously,sotheseSNPsdonotprovidesupportforthe
previousfindings(S3Table).
ForACLrupture,whenwerepeatedtheanalysiswithoutusingcovariates,wefoundthat
rs1800255inCOL3A1hadasmallp-valueof0.03withthesamedirectionasfoundpreviously
Table4. Re-testingcandidategenesforassociationwithAchillestendoninjuryorACLrupture.
Achillestendoninjury
SNP Gene EAa P-valueb OR(95%CI) Refc
rs1045485d CASP8 C 2.6E-03 1.10(1.04–1.16) [2]
rs4747096 ADAMTS14 G 0.55 0.98(0.93–1.04) [10]
rs2761884 BMP4 T 0.22 1.03(0.98–1.07) [1]
rs1134170 COL5A1 A 0.45 0.97(0.88–1.05) [5]
rs12722 COL5A1 T 0.62 0.97(0.88–1.06) [4]
rs3196378 COL5A1 A 0.75 1.01(0.93–1.10) [4]
rs1559186 COL5A3 C 0.75 1.01(0.97–1.05) [6]
rs331079 FBN2 C 0.32 1.03(0.93–1.10) [7]
rs4919510 MIR608 G 0.35 0.98(0.92–1.03) [5]
rs591058 MMP3 T 0.34 1.06(1.02–1.10) [9]
rs679620 MMP3 T 0.35 1.06(1.02–1.10) [9]
rs4789932d TIMP2 G 8.7E-03 1.08(1.02–1.13) [10]
rs1330363 TNC C 0.26 1.02(.98–1.07) [12]
rs2104772 TNC A 0.29 0.98(.94–1.02) [12]
ACLrupture
SNP Gene EAa P-valueb OR(95%CI) Refc
rs1516797 ACAN T 0.99 1.00(0.77–1.43) [13]
rs516115 DCN A 0.29 1.07(0.94–1.18) [13]
rs970547 COL12A1 T 0.11 1.11(0.98–1.25) [19]
rs2276109 MMP12 T 0.44 1.07(0.90-.1.24) [20]
rs1800255 COL3A1 A 0.17 1.09(1.00–1.31) [21]
rs331079 FBN2 G 0.45 1.07(.91–1.25) [7]
aEffectAllele.
bP-valuefromthisstudy.
cReferenceshowingoriginalassociationofcandidateSNPwithAchillestendinopathy.
dForrs1045485andrs4789932,thedirectionoftheeffectwasoppositetothepublishedresults;i.e.theminorCalleleofrs1045485andtheGalleleof
rs4789932areassociatedwithincreasedriskinthisworkbutdecreasedriskinthepriorwork[2][10].
https://doi.org/10.1371/journal.pone.0170422.t004
PLOSONE|https://doi.org/10.1371/journal.pone.0170422 March30,2017 9/16
Genome-wideassociationscreensforAchillestendonandACLtearsandtendinopathy
(S3Table)[21].Withseventests,thisp-valuedoesnotmeettheBenjamini-Hochbergthreshold
(p=7x10-3)andhasabouta21%chanceofbeingafalsediscovery.
TestingnewcandidateSNPsforassociationwithachillestendonInjury
orACLrupture
Todemonstratehowthesegenome-widedatacanbeapplied,weusedthedatatoevaluatea
newsetofcandidategenesforassociationwithAchillestendinopathyorruptureandwith
ACLrupture.Wecreatedalistof90genesencodingproteinsinvolvedintheformationofliga-
mentsortendons,includingthesetofcandidategenesfrompreviouspublications(S4Table).
Weassembledalistofallindependent,commonSNPsinthese90genes.Wetestedeachofthe
candidateSNPsforassociationwiththeseinjuriesbylookinguptheirp-valuesinthemeta-
analysisresults(S5Table).NoneofthecandidateSNPshadasignificantp-valueforeither
injury(usingtheBenjamini-Hochbergfalsediscoveryrateofq=0.05).ThetopSNPforAchil-
lestendoninjury(rs4660148)hadp=5x10-5,whichcorrespondstoafalsediscoveryrateof
about14%.ThetopSNPforACLrupture(rs8090)hadp=6x10-4,whichcorrespondstoa
falsediscoveryrategreaterthan50%(S5Table).
BesidesevaluatingeachSNPindividually,SNPswereanalyzedaspartofgenesusinggene-
basedanalysis[33].Ingene-basedanalysis,SNPsareassignedtoindividualgenesinclose
proximity,andtheneachgeneistestedforassociationwithaphenotype.Gene-basedtesting
confersanadvantageincaseswherethereisallelicheterogeneity–whentherearemultiple,
independentvariantsinfluencingatraitatthesamelocus.Withgene-basedtesting,thesignals
forgeneticassociationfromindependentvariantsarecombined,possiblyyieldingap-value
belowagivensignificancethresholdeventhoughtheindividualSNPsinthegenemaynot
themselveshavesignificantp-values.Weusedgene-basedanalysistotestthe90candidate
genesforassociationwitheitherAchillestendinopathyorruptureorwithACLrupture(Meth-
ods).ForAchillestendinopathyorrupture,noneofthegenesshowedasignificantassociation
(S6Table).
ForACLrupture,GLT25D1(alsoreferredtoasCOLGALT1)initiallyshowedaborderline
significantassociationusinggene-basedanalysis.Specifically,GLT25D1hadap-valueof5.7x
10-4among87genestested,whichisjustbelowthethresholdforBonferronisignificance(p=
5.8x10-4)(S6Table).GLT25D1encodescollagenbeta(1-O)galactosyltransferase1,whichtrans-
fersgalactosetohydroxylysineresiduesofcollagen[34].FourSNPs(rs2082001,rs2375637,
rs8110571andrs55960725)withinGLT25D1showedanassociationwithACLrupturewith
p<0.05(S1Fig,S7Table).OnepossibilityisthateachoftheseSNPsmightinfluenceACLrup-
tureindependently(allelicheterogeneity;fourindependentalleles).Anotherpossibilityisthat
onlyoneSNPisresponsiblefortheassociationandtheotherSNPsmayshowanassociation
duetoweaklinkage(r2wasbetween0.03and0.43betweenthefourSNPs).Todistinguish
betweenthesetwopossibilities,weperformedaconditionalanalysisinwhichthegenotype
oftheSNPwiththestrongestp-value(rs55960725)wasaddedasacovariateinthelogistic
regressionforACLrupture.Intheconditionalanalysis,noneoftheotherSNPs(rs2082001,
rs2375637orrs8110571)hadp-valuesthatfellbelow0.05andtheireffectsizesshrank,indicat-
ingthattheirassociationwithACLrupturewaspartiallydependentonrs55960725(S7Table).
Discussion
AchillestendonandACLinjuryarecommoninrecreationalandeliteathletes,andevenin
non-athletes[35–40].Priorstudieshaveidentified12genesassociatedwithAchillestendino-
pathyand8genesassociatedwithACLrupture[1,2,4–12,13,14–22].Thesepriorstudies,how-
ever,evaluatedasmallnumberofcandidategenesamongsmallcohortsofathletes.Withthe
PLOSONE|https://doi.org/10.1371/journal.pone.0170422 March30,2017 10/16