Genome-wide single nucleotide polymorphism (SNP) identification and characterization in a non-model organism, the African buffalo (Syncerus caffer), using next generation sequencing

(1)

Genome-wide single nucleotide polymorphism (SNP) identiﬁcation and characterization in a non-model organism, the African buffalo (Syncerus caffer), using next generation sequencing

Nathalie Smitz

^a,b,∗

, Pim Van Hooft

^c

, Rasmus Heller

^d

, Daniel Cornélis

^e,f

, Philippe Chardonnet

^g

, Robert Kraus

^h,i

, Ben Greyling

^j

, Richard Crooijmans

^k

, Martien Groenen

^k

, Johan Michaux

^a,e

aConservationGenetics,UniversityofLiège,BoulevardduRectorat26,4000Liège,Belgium

bJointExperimentalMolecularUnit,RoyalMuseumforCentralAfrica,Leuvensesteenweg11-17,3080Tervuren,Belgium

cResourceEcologyGroup,WageningenUniversity,P.O.Box47,6700AAWageningen,TheNetherlands

dBioinformatics,DepartmentofBiology,UniversityofCopenhagen,OleMaaløesVej5,2200Copenhagen,Denmark

eCentredeCoopérationInternationaleenRechercheAgronomiquepourleDéveloppement(CIRAD),UPRAGIRS,CampusInternationaldeBaillarguet, F-34398Montpellier,France

fCentredeCoopérationInternationaleenRechercheAgronomiquepourleDéveloppement(CIRAD)-RP-PCP,UniversityofZimbabwe,Harare,Zimbabwe

gInternationalFoundationfortheConservationofWildlife(IGF),RuedeTéhéran15,75008Paris,France

hDepartmentofBiology,UniversityofKonstanz,78457Konstanz,Germany

iDepartmentofMigrationandImmuno-Ecology,MaxPlanckInstituteforOrnithology,AmObstberg1,78315Radolfzell,Germany

jAgriculturalResearchCouncil,OldOlifantsfonteinRoad,IreneCenturion0062,SouthAfrica

kAnimalBreedingandGenomicsCentre,WageningenUniversity,Droevendaalsesteeg1,Wageningen,6708PB,TheNetherlands

Keywords:

Populationgenomics Conservation Diseaseecology Molecularmarkers

a b s t r a c t

ThisstudyaimedtodevelopasetofSNPmarkerswithhighresolutionandaccuracywithintheAfrican buffalo.Suchasetcanbeused,amongothers,todepictsubtlepopulationgeneticstructureforabetter understandingofbuffalopopulationdynamics.Intotal,18.5millionDNAsequencesof76bpweregen- eratedbynextgenerationsequencingonanIlluminaGenomeAnalyzerIIfromareducedrepresentation libraryusingDNAfromapanelof13Africanbuffalorepresentativeofthefoursubspecies.Weidentified 2534SNPswithhighconfidencewithinthepanelbyaligningtheshortsequencestothecattlegenome (Bostaurus).Theaveragesequencingdepthofthecompletealignedsetofreadswasestimatedat5x,and at13xwhenonlyconsideringthefinalsetofputativeSNPsthatpassedthefilteringcriterion.Oursetof SNPswasvalidatedbyPCRamplificationandSangersequencingof15SNPs.Ofthese15SNPs,14ampli- fiedsuccessfullyand13wereshowntobepolymorphic(successrate:87%).Thefidelityoftheidentified setofSNPsandpotentialfutureapplicationsarefinallydiscussed.

Introduction

TheAfricanbuffalo(Synceruscaffer)hassufferedofmajorpop- ulationlossesduringthelastcentury,impactingallbutunevenly subspecies.Habitatloss,climaticchanges,poachinganddiseases are the main challenges currently threatening the species sur- vival,contributingtolocalbuffalopopulationsdecimation.Direct competitionforspaceandresourcesgraduallyappearedwiththe

∗Correspondingauthor.Presentaddress:MRAC,Leuvensesteenweg11-17,3080, Tervuren,Belgium.

E-mailaddress:nathalie.smitz@africamuseum.be(N.Smitz).

expansionoflivestockfarmingandagriculture.Currentlyaround 75%oftheglobalAfricanbuffalopopulationislocatedinprotected areas(East,1999).Theresultingdisruptionofnaturalwildlifepop- ulationadmixtureislikelyresponsibleforgeneticerosion(Young and Clarke, 2000; Frankham et al., 2002). Isolated populations arelikelytohavelowerreproductiveﬁtnessandlosetheiradap- tivegeneticvariation,whilepresentingahigherriskofextinction (Frankhametal.,2002).Conservationgeneticshelpinidentifying andpromotingappropriatemanagementmethodstoreducethe risksof speciesextinction throughthe studyofthespatial dis- tributionof mutationsbetweenand amongpopulations. Recent technologicaladvanceshaverevolutionizedthegenerationofthese geneticresources,allowingDNA-libraryconstruction,large-scale

Konstanzer Online-Publikations-System (KOPS) URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-0-379374

Erschienen in: Mammalian Biology - Zeitschrift für Säugetierkunde ; 81 (2016), 6. - S. 595-603 https://dx.doi.org/10.1016/j.mambio.2016.07.047

(2)

596

sequencingandidentiﬁcationofsinglenucleotidepolymorphism (SNP)geneticmarkers(Seebetal.,2011).SNPswereshowntocon- stitutehighlyinformativemarkers(Morinetal.,2009)andlead toabetterinferenceofpopulationstructurethanmicrosatellites (Liuetal.,2005;Santureetal.,2010).Attentionhasbeguntoshift towardSNPsaspreferredgeneticmarkersduetotheirincreased powerofresolution and accuracy forstudying ﬁnescalepopu- lationstructure (Schlötterer, 2004).Thisis based ontheirhigh abundancethroughout thegenome,simple mutationcharacter- istics,lowmutationrates,usabilityonnon-invasivesamplesand historicalDNA,andstandardizationpossibilitiesbetweenlabora- tories(Krausetal.,2014;Morinetal.,2007a,b,2004;Luikartetal., 2003).SNPshavebecomeanestablishedmarkerinmolecularecol- ogy,evolutionarygenetics,andanimalbreeding(Daveyetal.,2011;

Krausetal.,2014,2012;Morinetal.,2004;Santureetal.,2010).

Despitetheirattractiveness,somedifficultieshavebeenexperi- encedindevelopingSNPinnon-modelorganismsduetothelimited ornogenomicresourcesavailable,leadingtocomplexlaboratory screeningofsegmentsofthegenomefrommultipleindividualsto yieldonlyasmallnumberofindependentSNPs.Next-generation sequencing (NGS) allows to overcome this issue by providing large-scalegenomevariation studiesbasedondeepsequencing ofrelativelylargegenomefractions(>1%)oreven thecomplete genome (Seeb et al., 2011). However, not so long ago, within non-modelorganisms,thepredominanttechniquehasbeenthe targetedgeneapproach,usingregularSangersequencing(Sanger etal.,1977),sinceitdoesnotrequirespecies-specificpre-existing DNAdataandisapplicabletomanytaxa.AfewhundredSNPswere identifiedusingthisapproachfornumerousspecies(e.g.,158SNPs, Sceloporusundulates;112SNPs,Salmosalar;768SNPs,Pusahisp- idahispida;168SNPs,Thryothoruspleurostictus)(Andreassenetal., 2010;Crameretal.,2008;Olsenetal.,2011;Rosenblumetal.,2006).

OnlyafewSNPsperspecies(<100)havebeendevelopedusingthe targetedgeneapproachforanimalsofconservationconcernsuch asthemarmoset(Saguinusoedipus),thedhole(Cuonalpinus)and theelephant(Loxodontaafricana)(Aitkenetal.,2004).Thetargeted geneapproach,althoughstillwidelyused,islaborious,timecon- suming,costlyandyieldsonlyafairlylimitedamountofcandidate SNPsincontrasttoNGS.

TheReducedRepresentationLibrary(RRL)approachisa NGS methodthatinvolvesa digestionstepofmultiplegenomicDNA sampleswithrestrictionenzyme(s),a selectionof theresulting restrictionfragmentsandasequencingstep.RRLapproacheshave beenusedtogeneratetensofthousandstomillionsofcandidate SNPswithagenome-widecoverageforexampleincattle(Tassell etal.,2008),turkey(Kerstensetal.,2009)andgreattit(VanBers etal.,2010).Alternatively,SNPresourcesfromonespeciescouldbe usedinacloselyrelatedspecies.AnIlluminaBovineSNP50Bead- Chiphasbeendevelopedforcattle(Bostaurus),acloserelativeto theAfricanbuffalo(Matukumallietal.,2009).ThisBeadChipscores 54,001informativeSNPsthatareuniformlydistributedalongthe entirecattlegenome.Ithasahighcrossamplificationsuccessrate acrosscattlebreeds (Matukumalliet al.,2009).However,when used onother bovid species, the number of polymorphic sites decreasessubstantially.OnlyafewpercentofallSNPsonthechip werestillpolymorphic(Milleretal.,2010)whentestedonother speciessuchasthewaterbuffalo(Bubalusbubalis−1159SNPs), theYak(Bosgrunniens−949), theNorth AmericanBison (Bison bison−1604),andtheBanteng(Bosjavanicus−1429)(Michelizzi etal.,2011).SimilarresultswereobtainedwhentestingtheOvi- neSNP50BeadChip,developedfordomesticsheep,ontworelated ovidspecies(Milleretal.,2010).Cross-speciesamplificationofSNP assaysusuallydoesnotworkwellcomparedtocross-speciesampli- ficationofmicrosatellites(Krausetal.,2012).Evenifgenotypingis successful,manypolymorphismsin onespeciesarefixedinthe other.Moreover,cross-speciesSNPsmayharborextremebiases

inallelefrequencies,sincetheymaypredominantlybefoundin regionsofthegenomeundernaturalselectionfavoringpolymor- phism(e.g.,balancingselection).

Sincecross-speciesgenotypingofSNPsoftenseemsproblem- aticorbiased,thisstudyaimstocharacterizeagenome-widesetof SNPsspecificallyfortheAfricanbuffalooveritswholedistribution area(sub-SaharanAfrica).ApreviousstudyconductedbyLeRoex etal.(2012)alreadyaimedatidentifyingSNPsintheAfricanbuffalo, howevertheirsamplingwaslimitedtotheCapebuffalosubspecies (Synceruscaffercaffer)andtotheHluhluwe-iMfoloziNationalPark (NP).ThebuffalopopulationwithinthisNationalParkisknown tobeaffectedbystrongnon-equilibriumconditionslinkedtoa founderevent(Smitzetal.,2014;DuToit,1954;Kappmeieretal., 1998).Inthepresentstudy,NextGenerationSequencingofreduced representationlibrariesforSNPdiscoverywasused.Thegenome ofanotherBovidspecies,Bostaurus,whichdivergedfromAfrican buffaloapproximately12millionyearsago,wasusedasareference formappingthereads(HassaninandRopiquet,2004;Pitraetal., 2002;RobinsonandRopiquet,2011;TimeTreesoftware-Hedges etal.,2006;KumarandHedges,2011).Thepresentstudyallowed theidentificationof2534SNPswithhighconfidencebyaligning shortsequencesoftheAfricanbuffalo(Synceruscaffer)tothecattle genome(Bostaurus).

Materialandmethods

Samplecollectionandlibrarypreparation

A geographically large and diverse panel of African buffalo wassampled:6fromEastand SouthernAfrica[SouthAfrica(2), Uganda(1), Kenya(1), Ethiopia(1), Namibia(1)] belonging tothe Synceruscaffercaffersubspecies,and7fromWestandCentralAfrica [CentralAfricanRepublic(1), Niger(3),Chad(2),BurkinaFaso(1)]

belongingtotheS.c.nanus,S.c.brachycerosandS.c.aequinoctialis subspeciesrespectively(Fig.1).These subspeciesweregrouped togetherbecausephylogeneticstudiesshowedthattheyformone cladewithonlyminortomoderateF_STdifferentiationbetweensub- species,rangingbetween0.02 and0.12 (Smitzetal.,2013;Van Hooftetal.,2002).Sampleextraction, selectionandRRLlibrary preparationproceduresareavailableasSupplementaryinforma- tion(Supplementaryﬁle1).

Sequenceﬁltering

Priorto thesequence alignment steps,differentﬁlters were appliedtotherawIllumina sequencedataaccording toseveral criteria.First,sequenceswereexpectedtostartwitha CTdinu- cleotidebecauseoftheAluIrestrictionsite(betweenATandCT).

Allsequencesnotbearingthispatternwerediscardedaspotential contamination.Secondly, averagequalityscoreswerecalculated foreachreadbytakingthemeanofallindividualscoresateachof the76positions.Readspresentinglowoverallphredqualityscores wereremoved(EwingandGreen,1998).Moreover,endofreads displayingtwosuccessivereadpositionswithaveragephredqual- ityscoreslowerthan20weretrimmedfromtheﬁrstreadposition withaphred<20.

SequencemappingandSNPdiscovery

Quality ﬁltered and trimmed sequence reads were aligned to the bovine reference genome (Bos taurus; UCSC Genome Bioinformatics; http://genome.ucsc.edu/http://genome.ucsc.edu/

(21/01/2015)) since an African buffalo genome sequence is notavailable. TheMosaikAssemblersoftware (Mosaik 1.0.1388- Stromberg, 2010) wasused withdefault settings, specifying a medianfragmentlengthof50bp(i.e.,innermatedistance)with

(3)

Fig.1.SamplinglocalitiesofAfricanbuffaloincludedinthisstudy.West-centralclade(S.c.nanus,S.cbrachyceros,S.c.aequinoctialis)andsouth-easternlineage(S.c.caffer) weresampeled;1,PamaReserve;2,WNP;3,AoukNPandZakoumaNP;4,N’gotto-Kindo;5,OmoNP;6,NakuruNP;7,QueenElizabethNP;8,WaterbergNP;9,Hluhluwe UmfoloziNP.

a searchradius of50bptosearchfor missing mate andfor an alignmentthatconformstothepaired-endorientation.TheALL alignmentmodewasusedwithahashsizeof15(allhashingstrat- egy),a maximumpercentage ofmismatchesallowedof15, and a minimum clustersize of 35. Consensus candidateSNPs were extractedfromthedatasetusingtheSAMToolssoftwarewiththe pileupfunction(SAMTools0.1.7–SequenceAlignment/Map-(Li etal.,2009a,b)).CandidateSNPs werethenﬁlteredforhavinga phredqualityscoreabove20(qualityofbasecalling>99%),forhav- ingamappingqualityscoreabove30,andaminoralleleoccurrence atthepolymorphicpositionofatleastthreetimes.Positionsthat weremonomorphicwithintheSNPdiscoverypoolbutshoweda ﬁxeddifferencewiththereferencegenome(Bostaurus)weredis- carded.Finally, SNPs witha fourtimeshigher read depththan theaveragereaddepthoftheRRL(averagetotalnumberofreads alignedwiththereferencegenometoauniqueposition)werealso discarded(Kerstensetal.,2009),asthesearelikelytobefalseSNPs thataretheresultofalignmentofparologoussequences.

For thevalidation oftheﬁlteredSNPset,primers hadtobe designedinthedirectﬂankingregionaroundtheSNPs.Therefore,a buffaloconsensussequencewasgeneratedfromtheRRLsequences

and ﬂanking sequences around SNP positions were extracted.

Wherepossible,flankingregionsweregeneratedbasedonthespe- cificAfricanbuffaloconsensussequence.IfaSNPwassituatedclose tothe beginningor end of thereads,flankingregions for each SNPweregeneratedusingpartoftheBostaurusgenomeandcon- catenatedwithAfricanbuffaloconsensusinformation(‘chimeric flankingsequences’(Jonkeretal.,2012)).

Validation

Our set of SNPs was validated by randomly selecting 15 SNPs scattered on the whole genome and by amplify- ing them by standard PCR. Only SNPs that reached at least an Illumina design score of 0.6 in Illumina’s Assay Design Tool available at http://www.illumina.com/support/array/array software(21/01/2015)wereselected.Primersweredesignedusing Primer3(RozenandSkaletsky,2000;http://simgene.com/Primer3 (21/01/2015)),enteringourchimericﬂankingsequences(Table1).

Asthecattlegenomewasusedtoalignourmergedsequences,we wantedtospeciﬁcallytestthesequenceconservationbetweenBos taurusandSynceruscaffer.Ifsequenceconservationishigh,ampliﬁ-

(4)

598

Table1

PrimersetsusedforSNPvalidationstep(n/a:noampliﬁcation).

Chrnr.bpposition Primerset SNP Observedpolymorphism

Chr1.75025101 TTTGGATCAGGAGGAACCAGCCCCTTTGGTGGAACATTTA A/G Yes

Chr5.14429154 AAAGGATTTCTGTTGGTGGAGAGATTTGCCTTCTCAAACTGGA A/G Yes

Chr6.119157451 TGAAATCTAACTGCCTGGGACTCAGGTGTGCTGGTTTACAGG C/T Yes

Chr9.106373371 AGTCTGCCTAAAAAGCCCATTCCTCCCACGCACAGACTC A/C Yes

Chr10.4010071 TCACCTGAATCCCACCCTTACTCGAGAAGGGCTTTGTGAC A/C Yes

Chr11.72584688 AACACCCCACCTTAATGCAGGTCAGGAGAGGGCTGTCAAG C/T Yes

Chr11.70625560 GCCATAAGGGTGGTGTCATCCCATGGACATCCTTTTCCTG C/T No

Chr12.20644129 TCCATGCCCATCTGAGATTTCCTGGCCTGACTCTGAGGTA A/C Yes

Chr14.80545216 GAGATCCCACTCGGCTGTTAAACCGTGAGCGAAGTGAGAG A/T Yes

Chr15.23524504 GATGGACTTGGTGGCAATTTGCCTCAGGACCATTTTCAGA A/G Yes

Chr15.81001941 GCTTGTTCAGATGGCACAGAGCCAGTACTCCCCCTAGCTC A/C n/a

Chr16.62257511 GCGTTCCTTCAACAACCAAGGCCATCTTGATTTCCTTCCA C/T Yes

Chr17.4151052 TCCCAGAGCAGACAGTCTCACGGTGATCATCTGCTAATGC A/C Yes

Chr17.74836302 CCCTCCACTAGCTTCTCAGCAGTGGAGCTGAGGTCTTGGA C/T Yes

Chr19.7345443 CATAATCCCAGCCAGTCTCCGAGAGCACCCCTGAGTTGAA C/T Yes

cationofSNPswillbesuccessfulwhenprimersaredesignedwithin theadjacentregionofthebovinegenomesurroundingthereadcon- tainingtheSNPcandidate.Therefore,threesituations(fiveSNPs persituation)weretested:(1)bothforwardandreverseprimers designedonthebovinegenomewherenoRRLreadsaligned,(2) oneforwardprimer designedonthereferencegenomeandone reverseprimerdesignedontheRRLreads(Synceruscaffer),and(3) bothprimersdesignedsolelyfromtheRRLreads.Thus,validation requiredSNPsfirstlytoamplifysuccessfully andsecondlytobe polymorphic.Eachofthese15SNPswassequencedinall13indi- vidualsusedintheoriginaldiscoverypool.ThePCRreactiontook placeinatotalvolumeof12␮l,consistingof3␮lDNA(10ng/␮l), 0.5␮lofprimers(0.03␮g/␮l),5.2␮lMastermix(ThermoScientific) and0.3␮lAmpliTaq^® DNAPolymerase. Cyclingconditionscon- sistedof36cyclesfor30sat95^◦C,30sat50^◦Cand30sat72^◦C.

Aninitialdenaturationstepprecededtheprocess(5minat95^◦C), andaﬁnalextensionstepfollowedtheprocess(10minat72^◦C).

SequencingwasperformedonanABI3730XLcapillarysequencer.

TheresultingsequenceswerealignedusingCLUSTALX(Thompson etal.,1997)asimplementedinBIOEDITv.7.09(Hall,1999)andSNPs werevalidatedvisually.

Results

SequencingoftheRRLandreadﬁltering

TheAluIrestrictionenzymewaschosenfortheconstructionof theAfricanbuffaloRRLsince itmaximizedthequantityoffrag- mentssituatedinthetargetedsizerangeof100–200bp,evaluated performinganinsilicodigestionoftheBostaurusgenome.Cor- respondencesbetweeninsilicoandinvitroobservedrestriction enzymecleavagepatternswerepreviouslydemonstratedwithin othermammalspecies(Abdurashitovetal.,2006,2007).Intotal, 18.5millionpaired-endsequencesof76bplengthweregenerated bytheIlluminaGenomeAnalyzerIIontwolanes.Thegenomecov- erage(Mosaik1.0.1388,Stromberg,2010)wasestimatedatabout 5%ofcodingandnon-codingregions(Fig.2).Theaveragephred qualityscoreperreadpositiondroppedbelow20afterposition55 forabout5millionreads.TomaintainsufﬁcientqualityforSNP detection,thosereadsweretrimmedafterposition55.Theaverage sequencingdepthofthewholealignedsetofreadswasestimated at5xandat13xafterﬁlteringsteps.

Readalignmenttothereferencegenome

Thecattlegenomeconsistsof29autosomesandthesexchro- mosomes,withatotalestimatedgenomesizeof2.87Gb(Liuetal., 2009).Intotal,about60%ofthereadswerenotretainedbecause

theydidnotpasstheﬁlteringcriteriaofthealignment:theywere tooshort,werenotunique(i.e.,aligntomorethanonelocation) orcontainedtoomanynucleotidedifferences.Eventually,6.9mil- lionreadsremainedfortheSNPdiscoveryandcouldbesuccessfully alignedtotheBostaurusreferencegenome,correspondingto836.5 millionbp.Fromthesereads,22%wereorphans(i.e.,onlyoneof thepairedreadalignedtothereference,whiletheotherdidnot), while14%hadonepairedreadthatwasnon-unique.Thephysical distributionoftheidentiﬁedSNPsacrossthebuffalogenomewas estimatedusingthecattlegenomeasthereference.

SNPdetection

Atotalof318,091putativeSNPsweredetected.Fixeddiffer- encesbetweenthediscoverypanelofAfricanbuffaloandthecattle referencegenomewerediscarded(i.e.,22,472).ThefewSNPs(1.7%) withmorethantwo alleleswerealsodiscardedastri-ortetra- allelicSNPsareuncommonandaremorelikelytobetheresultof sequencingerrorsthanrealpolymorphism (Brookes,1999).Fur- thermore,mostgenotypingassaydesignsdonotallowformore thantwoalleles.AfterfilteringtheseSNPsforminorallelecount (minoralleleoccurringatleastthreetimes),forminimumphred qualityscoreof20andforminimummappingqualityscoreof30, 2534SNPsinwhichweplacehighconfidenceremaineddistributed acrosstheentiregenome(Table2).Thesequencingdepthhadan averageof13readsafterfilteringsteps.Atotal numberof1837 SNPshadanIlluminadesignscore≥0.6.

SNPqualityassessment

Theratiooftransitions (TS;i.e.,C/TorA/G)versustransver- sions(TV;i.e.,A/T,G/C,A/CorG/T)wasestimatedasameasurefor thequalityoftheSNPdiscovery.TheTS:TVratioobservedwithin ourdatasetwas2.38:1(1784transitionsversus750transversions), withanearlyequal numberofA/G andT/Cmutations(889A/G and895T/C),andthefourTVchangesoccurringatsimilarfrequen- cies.Thisistheexpectedempiricalratio,whileratiossubstantially lowerthan2canbeindicativeofrandomgenotypingerror(Kraus etal.,2012).TheTS:TVratioremainedsimilarwhenplottedper readposition(Fig.3),whichisagoodindicationthattherewasno readpositionbias,suchasfalseSNPsduetolowsequencingquality towardstheendsofreads(Krausetal.,2012).Moreover,SNPpre- dictionsweretestedbydeterminingwhetherparticularregionsof sequencereadspresentedmoreSNPcandidatesthanothers.Previ- ousstudieshaveshownthattailsofreadspresentexcessivelymore sequencingerrors,leadingtofalseSNPsidentiﬁcation(Dohmetal., 2008;VanBersetal.,2010).WithintheSNPsetpassingallﬁltering

(5)

Fig.2.Referencesequencecoverageofthe30bovinechromosomes(Bostaurus).

(6)

600

Table2

NumberofhighconﬁdenceSNPsoneachofthe30bovinechromosomes(Bostaurus).

Chrnr SNPnr

1 89

2 98

3 135

4 92

5 105

6 51

7 103

8 70

9 54

10 73

11 114

12 58

13 97

14 69

15 80

16 68

17 72

18 90

19 112

20 50

21 102

22 106

23 89

24 89

25 119

26 84

27 73

28 61

29 92

X 39

Fig.3. AverageTS:TVratioateachreadposition.

criteria,nooverrepresentationofSNPcandidatesintheendsofthe readswasobserved.

SNPvalidation

OnlyoneprimerpairdesignedwithintheRRLreadsfailedto yieldanamplificationproduct.Withinthe14remainingamplifi- cationproducts,onedidnotcontaintheexpectedpolymorphism, whichmeansthat87%oftheexpectedSNPswereconfirmedby Sangersequencingoftheindividualsinthediscoverypanel.This highpercentageofsuccessfulPCRamplificationissimilartothat observedingeese(93%of384SNPstested)usingthesamechimeric technique (Jonker et al., 2012). The ten primer set entirely or partlydesignedwithinthecattlegenomegavePCRproductswith expectedSNPsobserved.Thiscorroboratesthehighgenomecon- servationbetweencattleandbuffalo.

Discussion

ModelspeciesreferencegenomeandSNPvalidation

Thepresentstudyenabledtheidentificationof2534SNPswith highconfidenceinanon-modelorganism.1837SNPshadanIllu- mina design score ≥0.6, reflecting a high likelihood that assay designwillbesuccessfulonamodernhighthroughputSNPgeno- typingplatform.About30%ofthesequencereadscouldbealigned tothebovinegenome,aclosely-relatedspecies.ThestudyofJonker etal.(2012)usedthesametechniquetoidentifySNPsintheBar- nacleGoose(Brantaleucopsis)byaligning1.77millionreadstothe Mallard(Anasplatyrhynchos)genome(divergencetime30million years)(Huangetal.,2013).Inthatstudy,16.1%ofthereadssuc- cessfullyaligned,subsequentlyallowingtheidentificationof2188 highconfidenceSNPs.IntheAfricanbuffalo,LeRoexetal.(2012) mapped19–23%oftheirshortreads(50bpreads)tothedomestic cowgenome(Bostaurus).Ourstudyconfirmsthatusingagenome ofacloselyrelatedspeciesasareferencestandardprovidesasuffi- cientnumberofhighconfidenceSNPsandoffersagoodalternative tocharacterizeSNPs innon-modelspecieswithoutcarryingout tediousstepsofdeepassemblyofredundantcontigs(Kerstensetal., 2009;VanBersetal.,2010).

The chimeric ﬂanking sequences obtained from the cattle genomewerealsousedtogenerateprimersforvalidationsteps.

HighPCRamplificationsuccesswithchimericprimersindicated sufficientconservationbetweenthegenomesofthetwospecies tousethemforgenotypingassaydesign.ThehighPCRamplifica- tionsuccesscouldlikelybeattributedtothecorrespondenceofthe alignedfilteredreadstohighly conservedsequences.VanHooft etal.(1999)previouslydemonstratedhighgenomeconservation whenusingmicrosatellitesprimersdevelopedforcattleonAfrican buffalo,with83%successfulamplification.

Ascertainmentbias

Ascertainmentbiascanresultfromtheselectionoflocifrom anunrepresentativesample ofindividuals. Tolimit thiskindof bias,arelativelylargepoolofsamplescoveringthewholedistribu- tionareaofthetargetedspecieswasselected,comprisingallfour currentlytaxonomicallyrecognizedsubspeciesofAfricanbuffalos.

However,ascertainmentbiascanalsobeintroducedbylimitedread depth.Bystipulatingaminorallelecountofthreeinourprotocol, sequencingdepthshouldatleastcomprisesixreadsforaSNPto beretained.Ouraveragesequencingdepthofthewholeputative SNPdatabasewasestimatedat5x,whichincreasedto13xwhen estimatedontheSNPsetthatpassedtheﬁlteringcriterion.This sequencingdepthremainslowcomparedtootherstudies(e.g.,25x (VanBersetal.,2010),58x(Kerstensetal.,2009)).Thestudiesmost similartoours,toourknowledge,arethatofJonkeretal.(2012) (Brantaleucopsis),whichyieldedanaveragesequencedepthof9.9x, andthatofLeRoexetal.(2012)(S.c.caffer),whichyieldanaverage sequencedepthof2.7x.Ourlowsequencedepthmaybeexplained byanover-representationofsize-fractionatedfragmentsranging between100and200bpslicedfromthepolyacrylamidegel.Conse- quently,manytruerarevariantsmaycertainlyhavebeenrejected.

Moreover,thelowdepthofcoveragealsoimpliesthatonlySNPs presentinmultiplesamplesofourDNApoolhadachancetobe identiﬁed.OverrepresentationofcommonSNPsoverrareSNPsis thusexpectedtointroduce biasintoourSNPset.Thisneedsto betakenintoaccountwheninterpretinggenotypicdatainfuture projects.However,byourgeographicallybroadsamplingdesignwe avoidthesubstantialgeographicascertainmentbiasthatislikely presentinLeRoexetal.(2012),becausetheirSNPdiscoverypanel waslimitedingeographicextent.

(7)

Fig.4. CumulativenumberofSNPoccurrenceateachreadposition.

Sequencingerror

Differentestimatorswereusedtoevaluatetheriskoffalseposi- tivesintheSNPdiscoveryanalysis.AhighTS:TVratioisconsidered a goodmeasure of SNPvalidity.A TS:TVratio of1:2 wouldbe detectedifmutationswererandomandisthereforeanindication forsequencingerrors. HigherratesofC/Tmutationsdue tothe deaminationofmethylcytosinesinCpGdinucleotidesareresponsi- bleforahigherTS:TVratioinrealdata(Cooperetal.,2010;Scarano etal.,1967;Vignaletal.,2002).Usually,theratioof2.1:1isobserved inmammals(DePristoetal.,2011).Aratiosigniﬁcantlylowerthan thislastonecanthereforebeanindicatorofpoorqualitysequenc- ingdata.OurTS:TVratioof2.4:1issimilartotheresultsobtained forexampleinthestudyofKrausetal.(2011)(2.3:1)andJonker etal.(2012)(2.7:1).ItthusindicatesthatmostdetectedSNPcalls werenotrandom,whichreﬂectsthatourSNPslikelyrepresenttrue nucleotidepolymorphism.

MisidentiﬁcationofSNPsduetosequencingerrorsisavoided byexcludingvariationwithalowphredscore.Tailsofreadsoften displayincreasinglymoresequencingerrorsusingIllumina’stech- nology.Eventhoughwetrimmedourreadsduringqualitychecks, adecreaseinthenumberofpredictedSNPsinthetailofthereads wasobservedandmaybeexplainedbyanassociateddecreaseof theassociatedphredscore(Fig.4).ThisdecreaseinpredictedSNPs perpositioninIlluminasequencingreadswasalsoreportedinear- lierstudies(Kerstensetal.,2009;Ramosetal.,2009;VanBersetal., 2010),onwhichthecurrentSNPdetectionpipelineisbased.

Thehighvalidationsuccessinourstudycanalsoillustratethe qualityof thepredictedSNPs. Fromthesetof15, onlyone did notamplifywhoseprimersweredesignedbasedonthegenerated reads(Table1).Therefore,a closelyrelatedspeciesgenomecan beusedbothformappingandSNPdiscovery,butalsotodesign theprimersforthegenotypingstep.AnotherSNPappearedtobe monomorphic,yieldingaconversionrateof87%ofpolymorphic SNPs.ThisvalidationsuccessrateishigherthanthatofLeRoex etal.(2012)workingontheCapebuffalo(S.c.caffer).There,within thesetof173SNPs usedfor thevalidation,143amplifiedsuc- cessfullyandonly75werepolymorphic.ThefalsepositivesinSNP discoveryinthestudyofLeRoexetal.(2012)seemstobelinked tothelowcoverage(mean2.7x),andtothefactthataSNPwas inferredifthenucleotidevariantwassupportedbyaminimumof tworeads(vs.6inourstudy).Usingsuchalowcut-offmaysig- nificantlyincreasetheriskofidentifyingfalsepositives.Applying aminorallelecountofatleastthreeminimizesfalsepositivesin theSNPdiscoveryanalysis.Thisapproach,however,alsodramati- callyreducestheidentificationoftruenucleotidepolymorphisms thatcould,inprinciple,bedetectedevenifitwouldalsoincrease chancesofidentifyingasequencingerrorasanSNP.

UtilityofSNPsinAfricanbuffalo

SNPmarkerscanprovidemajorinsightsintoanimaldispersal patterns.Thisisespeciallyrelevantinthelightofrecentconserva- tioninitiativesaimingtorestoregeneticdiversityofwildlifestocks by re-establishing demographic connectivity between wildlife populations of different NP (e.g., Great Limpopo Transfrontier Conservation Area). Dispersal beyond traditional conservancy boundaries,andalsonationalborders,mayposeariskasfaras thespreadofpathogensareconcerned(Crossetal.,2004,2005).

Amongwildlifespecies,buffalosareknowntobeoneofthemain wildlifereservoirsfordiseases(Rodwelletal.,2001).Sincebuf- faloarecloselyrelatedtocattle,andmaytransmitdiseasedirectly orindirectly, buffaloalsorepresentan importantthreatfor the Africanlivestockindustry,fromaconservation,sanitaryandeco- nomicpointofview(Garine-Wichatitskyetal.,2010;Jollesetal., 2005;Micheletal.,2006).Forﬁnescaleinference,alargernumber ofSNPsmayberequired,astheinformationcontentofoneSNPis lessthanthatofonemicrosatellite(i.e.,bi-allelicvs.multi-allelic markers).Previousstudiesrevealedthatfourtotwelvetimesmore SNPsareneededforpopulationstructureinferencetomatchthe statisticalpowerofonemicrosatellite(Liuetal.,2005).Forhighly dispersiveorganismsithasbeenshownthatthedetectionoflow levelsofdifferentiationispossiblewithaminimumof 80SNPs (Morinetal.,2009;Rymanetal.,2006).Ourlargesetofhundreds ofSNPsshouldthusallowtoscalethegeneticmarkersystemto theneedsoffuturestudiesoftheinteractionbetweenlandscape featuresandmicroevolutionaryprocesses(Maneletal.,2003).

ThisSNPdatabasemayalsobeofbenefitinthecontextofselec- tivebreeding.Indeed,selectivebreedingofspecificphenotypesof theCapebuffalosubspecies(S.c.caffer-South-EasternAfrica)has becomeanintricatebusinesswithinprivategamefarming.Females arebeingselectedforhorn length,milkproduction andregular calvingintervals,whilemalesarebeingselectedforhornsize,body massandshape,whicharedesirabletotrophyhunters.However, suchapproaches mayleadto distortionof evolutionarynatural processesandmayreducethespeciesgeneticvariabilitythereby weakeningthespeciesresilienceinthewild.Thesepracticesare notbelievedtobenefittheconservationofglobalbiodiversity,and mayevenbecomeproblematicifgeneticdilutionoccursthrough escapesofselectedindividualsintoneighboringwildpopulations ofbuffalo.Futuredevelopmentofguidelinesincollaborationwith thegame-farmingbreedersshouldallowfindingcompromisesfor thelong-termconservationofthewildlifespecies.

Conclusions

WithinahighlymobilespeciessuchastheAfricanbuffalo,the SNPsetdeveloped inthis studyshouldprovidehighlyvaluable andreliabletoolsforgaininginsightintothemigratorypatternof thisspecies,knowntobeadiseasereservoir.Ourapproachyielded higherqualitySNPs(asjudgedbyassayconversionrate)andless geographicallybiasedSNPsthanapreviousstudy(LeRoexetal., 2012).Furthermore,theconstructionofchimericﬂankingsequence wasshown toincreasethenumberofusableSNPsbyproviding sufﬁcientlylargeregionsforthegenotypingassay.

Conﬂictofinterest

Theauthorsdeclaretohavenoconﬂictsofinterest.Theydevel- opedallaspectsofthisstudy.Thesponsorsoftheissuehadnorole inthestudydesign,datacollectionandanalysis,decisiontopublish, orpreparationofthemanuscript.

(8)

602

Authorcontributions

Thepresentresearchstudywasdesignedincollaborationwith PimVanHooft,RasmusHeller,JohanMichaux,RichardCrooijmans, MartienGroenenandBenGreyling.Partialsampleswereprovided byDanielCornélisandPhilippeChardonnet.Statisticalanalysisand interpretationwasperformedbyNathalieSmitz,withassistance ofRobertKraus,RichardCrooijmansandMartienGroenen.Allco- authorsparticipatedtothepaperwriting.

Acknowledgements

WeliketothankthesupportoftheResearchPlatform“Produc- tionandConservationinPartnership”(RP-PCP).Wewouldalsolike tothankF.Jori,B.VanVuuren,K.L.Kanapeckasandallcollectorsfor providingusthesamplesusedforthediscoveryoftheSNPset.Tech- nicalassistanceinthelaboratorywasprovidedbyBertDibbits,and RudyJonkerhelpedwithconstructingchimericﬂankingsequences.

Proofreading assistancewasprovided byVirginie Winant.This projectwassupportedbythenetwork “BibliothèqueduVivant”

fundedbytheCNRS,the“MuséeNationald’HistoireNaturelle”,the INRAandthe“CentreNationaldeSéquenc¸age”.Thisstudyissup- portedbyresearchgrantsfromtheFRS-FNRSofBelgiumprovided toJ.R.MichauxandN.M.R.Smitz.

AppendixA. Supplementarydata

Supplementarydataassociatedwiththisarticlecanbefound,in theonlineversion,athttp://dx.doi.org/10.1016/j.mambio.2016.07.

047.

References

Abdurashitov,M.A.,Tomilov,V.N.,Chernukhin,V.A.,Gonchar,D.A.,Degtyarev,S.

Kh.,2006.MammalianchromosomalDNAdigestionwithrestriction endonucleasesinsilico.OvchinnikovBull.Biotechnol.Phys.Chem.Biol.2, 29–38.

Abdurashitov,M.A.,Tomilov,V.N.,Chernukhin,V.A.,Gonchar,D.A.,Degtyarev,S.

Kh.,2007.ComparativeanalysisofhumanchromosomalDNAdigestionwith restrictionendonucleasesinvitroandinsilico.Med.Genet.6,29–36.

Aitken,N.,Smith,S.,Schwarz,C.,Morin,P.A.,2004.Singlenucleotide

polymorphism(SNP)discoveryinmammals:atargeted-geneapproach.Mol.

Ecol.13,1423–1431,http://dx.doi.org/10.1111/j.1365-294X.2004.02159.x.

Andreassen,R.,Lunner,S.,Høyheim,B.,2010.TargetedSNPdiscoveryinAtlantic salmon(Salmosalar)genesusinga3UTR-primedSNPdetectionapproach.

BMCGenomics11,706,http://dx.doi.org/10.1186/1471-2164-11-706.

Brookes,A.J.,1999.TheessenceofSNPs.Gene234,177–186.

Cooper,D.N.,Mort,M.,Stenson,P.D.,Ball,E.V.,Chuzhanova,N.A.,2010.

Methylation-mediateddeaminationof5-methylcytosineappearstogiverise tomutationscausinghumaninheriteddiseaseinCpNpGtrinucleotides,aswell asinCpGdinucleotides.Hum.Genomics4,406–410.

Cramer,E.R.A.,Stenzler,L.,Talaba,A.L.,Makarewich,C.A.,Vehrencamp,S.L., Lovette,I.J.,2008.IsolationandcharacterizationofSNPvariationat90 anonymouslociinthebandedwren(Thryothoruspleurostictus).Conserv.

Genet.9,1657–1660,http://dx.doi.org/10.1007/s10592-008-9511-7.

Cross,P.,Lloyd-Smith,J.,Bowers,J.,Hay,C.,Hofmeyr,M.,Getz,W.,2004.

Integratingassociationdataanddiseasedynamics:anillustrationusing AfricanbuffaloinKrugerNationalPark.Ann.Zool.Fennici41,879–892.

Cross,P.C.,Lloyd-Smith,J.O.,Getz,W.M.,2005.Disentanglingassociationpatterns inﬁssion-fusionsocietiesusingAfricanbuffaloasanexample.Anim.Behav.69, 499–506,http://dx.doi.org/10.1016/j.anbehav.2004.08.006.

Davey,J.W.,Hohenlohe,P.A.,Etter,P.D.,Boone,J.Q.,Catchen,J.M.,Blaxter,M.L., 2011.Genome-widegeneticmarkerdiscoveryandgenotypingusing next-generationsequencing.Nat.Rev.Genet.12,499–510,http://dx.doi.org/

10.1038/nrg3012.

DePristo,M.A.,Banks,E.,Poplin,R.,Garimella,K.,Maguire,V.,Hartl,J.R., Philippakis,C.,Angel,A.A.,del,G.,Rivas,M.A.,Hanna,M.,McKenna,A.,Fennell, T.J.,Kernytsky,A.M.,Sivachenko,A.Y.,Cibulskis,K.,Gabriel,S.B.,Altshuler,D., Daly,M.J.,2011.Aframeworkforvariationdiscoveryandgenotypingusing next-generationDNAsequencingdata.Nat.Genet.43,491–498,http://dx.doi.

org/10.1038/ng.806.

Dohm,J.C.,Lottaz,C.,Borodina,T.,Himmelbauer,H.,2008.Substantialbiasesin ultra-shortreaddatasetsfromhigh-throughputDNAsequencing.Nucleic AcidsRes.36,e105,http://dx.doi.org/10.1093/nar/gkn425.

DuToit,R.,1954.TrypanosomiasisinZululandandthecontroloftsetseﬂiesby chemicalmeans.OnderstepoortJ.Vet.Res.26,317–387.

East,R.,1999.AfricanAntelopeDatabase1999.Gland:IUCN,Switzerlandand Cambridge.

Ewing,B.,Green,P.,1998.Base-callingofautomatedsequencertracesusingphred.

II.Errorprobabilities.GenomeRes.8,186–194.

Frankham,R.,Ballou,J.D.,Briscoe,D.A.,2002.IntroductiontoConservation Genetics.CambridgeUniversityPress,Cambridge.

Garine-Wichatitsky,M.,deCaron,A.,Gomo,C.,Foggin,C.,Dutlow,K.,Pfukenyi,D., Lane,E.,Bel,S.,LeHofmeyr,M.,Hlokwe,T.,Michel,A.,2010.Bovine tuberculosisinbuffaloes,SouthernAfrica.Emerg.Infect.Dis.16,884–885, http://dx.doi.org/10.1890/02-5266.

Hall,T.A.,1999.BioEdit:auser-friendlybiologicalsequencealignmenteditorand analysisprogramforWindows95/98/NT.NucleicAcidsSymp.Ser.41,95–98.

Hassanin,A.,Ropiquet,A.,2004.MolecularphylogenyofthetribeBovini(Bovidae, Bovinae)andthetaxonomicstatusoftheKouprey,BossauveliUrbain1937.

Mol.Phylogenet.Evol.33,896–907,http://dx.doi.org/10.1016/j.ympev.2004.

08.009.

Hedges,S.B.,Dudley,J.,Kumar,S.,2006.TimeTree:apublicknowledge-baseof divergencetimesamongorganisms.Bioinformatics22,2971–2972,http://dx.

doi.org/10.1093/bioinformatics/btl505.

Huang,Y.,Li,Y.,Burt,D.W.,Chen,H.,Zhang,Y.,etal.,2013.Theduckgenomeand transcriptomeprovideinsightintoanavianinﬂuenzavirusreservoirspecies.

Nat.Genet.45,776–784,http://dx.doi.org/10.1038/ng.2657.

Jolles,A.E.,Cooper,D.V.,Levin,S.A.,2005.Hiddeneffectsofchronictuberculosisin Africanbuffalo.Ecology86,2358–2364,http://dx.doi.org/10.1890/05-0038.

Jonker,R.M.,Zhang,Q.,VanHooft,P.,Loonen,M.J.J.E.,VanderJeugd,H.P., Crooijmans,R.P.M.A.,Groenen,M.A.M.,Prins,H.H.T.,Kraus,R.H.S.,2012.The developmentofagenomewideSNPsetfortheBarnaclegooseBranta leucopsis.PLoSOne7,e38412,http://dx.doi.org/10.1371/journal.pone.

0038412.

Kappmeier,K.,Nevill,E.M.,Bagnall,R.J.,1998.Reviewoftsetseﬂiesand trypanosomosisinSouthAfrica.OnderstepoortJ.Vet.Res.65,195–203.

Kerstens,H.H.D.,Crooijmans,R.P.M.,Veenendaal,A.,Dibbits,B.W.,Chin-A-Woeng, T.F.C.,denDunnen,J.T.,Groenen,M.A.M.,2009.Largescalesinglenucleotide polymorphismdiscoveryinunsequencedgenomesusingsecondgeneration highthroughputsequencingtechnology:appliedtoturkey.BMCGenomics10, 479,http://dx.doi.org/10.1186/1471-2164-10-479.

Kraus,R.H.S.,Kerstens,H.H.D.,VanHooft,P.,Crooijmans,R.P.M.A.,VanDerPoel,J.J., Elmberg,J.,Vignal,A.,Huang,Y.,Li,N.,Prins,H.H.T.,Groenen,M.A.M.,2011.

GenomewideSNPdiscovery,analysisandevaluationinmallard(Anas platyrhynchos).BMCGenomics12,150,http://dx.doi.org/10.1186/1471-2164- 12-150.

Kraus,R.H.S.,Kerstens,H.H.D.,vanHooft,P.,Megens,H.-J.,Elmberg,J.,Tsvey,A., Sartakov,D.,Soloviev,S.A.,Crooijmans,R.P.M.A.,Groenen,M.A.M.,Ydenberg, R.C.,Prins,H.H.T.,2012.Widespreadhorizontalgenomicexchangedoesnot erodespeciesbarriersamongsympatricducks.BMCEvol.Biol.12,45,http://

dx.doi.org/10.1186/1471-2148-12-45.

Kraus,R.H.S.,Vonholdt,B.,Cocchiararo,B.,Harms,V.,Bayerl,H.,Uhn,R.K.,Orster, D.W.F.,Roos,C.,2014.Asingle-nucleotidepolymorphism-basedapproachfor rapidandcost-effectivegeneticwolfmonitoringinEuropebasedon noninvasivelycollectedsamples.Mol.Ecol.Resour.,http://dx.doi.org/10.1111/

1755-0998.12307.

Kumar,S.,Hedges,S.B.,2011.TimeTree2:speciesdivergencetimesontheiPhone.

Bioinformatics27,2023–2024,http://dx.doi.org/10.1093/bioinformatics/

btr315.

LeRoex,N.,Noyes,H.,Brass,A.,Bradley,D.G.,Kemp,S.J.,Kay,S.,vanHelden,P.D., Hoal,E.G.,2012.NovelSNPdiscoveryinAfricanbuffalosynceruscaffer,using high-throughputsequencing.PLoSOne7,e48792,http://dx.doi.org/10.1371/

journal.pone.0048792.

Li,H.,Handsaker,B.,Wysoker,A.,Fennell,T.,Ruan,J.,Homer,N.,Marth,G., Abecasis,G.,Durbin,R.,2009a.Thesequencealignment/map(SAM)formatand SAMtools.Bioinformatics25,2078–2079.

Li,S.,Wan,H.,Ji,H.,Zhou,K.,Yang,G.,2009b.SNPdiscoverybasedonCATSand genotypingintheﬁnlessporpoise(Neophocaenaphocaenoides).Conserv.

Genet.10,2013–2019.

Liu,N.,Chen,L.,Wang,S.,Oh,C.,Zhao,H.,2005.Comparisonofsingle-nucleotide polymorphismsandmicrosatellitesininferenceofpopulationstructure.BMC Genet.6(Suppl.1),S26,http://dx.doi.org/10.1186/1471-2156-6-s1-s26.

Liu,Y.,Qin,X.,Song,X.-Z.H.,Jiang,H.,Shen,Y.,Durbin,K.J.,Lien,S.,Kent,M.P., Sodeland,M.,Ren,Y.,Zhang,L.,Sodergren,E.,Havlak,P.,Worley,K.C., Weinstock,G.M.,Gibbs,R.A.,2009.Bostaurusgenomeassembly.BMC Genomics10,180,http://dx.doi.org/10.1186/1471-2164-10-180.

Luikart,G.,England,P.R.,Tallmon,D.,Jordan,S.,Taberlet,P.,2003.Thepowerand promiseofpopulationgenomics:fromgenotypingtogenometyping.Nat.Rev.

Genet.4,981–994,http://dx.doi.org/10.1038/nrg1226.

Manel,S.,Schwartz,M.K.,Luikart,G.,Taberlet,P.,2003.Landscapegenetics:

combininglandscapeecologyandpopulationgenetics.TrendsEcol.Evol.18, 189–197,http://dx.doi.org/10.1016/S0169-5347(03)00008-9.

Matukumalli,L.K.,Lawley,C.T.,Schnabel,R.D.,Taylor,J.F.,Allan,M.F.,Heaton,M.P., O’Connell,J.,Moore,S.S.,Smith,T.P.L.,Sonstegard,T.S.,VanTassell,C.P.,2009.

DevelopmentandcharacterizationofahighdensitySNPgenotypingassayfor cattle.PLoSOne4,e5350,http://dx.doi.org/10.1371/journal.pone.0005350.

Michel,A.L.,Bengis,R.G.,Keet,D.F.,Hofmeyr,M.,DEKlerk,L.M.,Cross,P.C.,Jolles, A.E.,Cooper,D.,Whyte,I.J.,Buss,P.,Godfroid,J.,2006.Wildlifetuberculosisin SouthAfricanconservationareas:implicationsandchallenges.Vet.Microbiol.

112,91–100,http://dx.doi.org/10.1016/j.vetmic.2005.11.035.