UNIVERSITAgrave CATTOLICA DEL SACRO CUORE Sede di Piacenza

Scuola di Dottorato per il Sistema Agro-alimentare

Doctoral School on the Agro-Food System

cycle XXVII

SSD AGR17

High-throughput sequencing technologies and genomics insights into the bovine species

Candidate Marco Milanesi Matr n 4011148

Academic Year 20132014

Scuola di Dottorato per il Sistema Agro-alimentare

Doctoral School on the Agro-Food System

cycle XXVII

SSD AGR17

High-throughput sequencing technologies and genomics insights into the bovine species

Coordinator Chmo Prof Antonio Albanese

_______________________________________

Candidate Marco Milanesi Matriculation n 4011148

Tutor Prof Paolo AjmoneMarsan PhD Lorenzo Bomba PhD Stefano Capomaccio PhD Ezequiel Louis Nicolazzi

Academic Year 20132014

Alla$mia$famiglia$

A$mio$nonno$Mario$

INDEXamp

Chapteramp1ndashIntroduction 3

Chapteramp2Relativeextendedhaplotypehomozygosity(rEHH)signalsacrossbreedsrevealdairyandbeefspecificsignaturesofselection

11

Chapteramp3SearchingnewsignalsforproductivetraitsthroughgenebasedassociationanalysisinthreeItaliancattlebreeds

49

Chapteramp4ampMUGBASaspeciesfreegenebasedassociationsuite

81

Chapteramp5Imputationaccuracyisrobusttocattlereferencegenomeupdates

89

Chapteramp6QuestfordeleteriousrecessivevariantsinItalianHolsteinbullscombiningexomesequencingandhighdensitySNPanalysis

99

Chapteramp7ndashConclusion 167

amp

Acknowledgmentsampamp 173

CHAPTER1

Introduction

Background

The bovine species counts almost 15 billion of heads classified in some 800

breeds worldwide (httpfaostat3faoorg) The species is adapted to a wide

varietyofdifferentenvironmentandhusbandrysystemsandisoftenrearedfor

multiple purposes (Taberlet et al 2008) draft power milk meat leather

productionandotheruses(Elsiketal2009)

European cattle (Bos$ taurus) derive from a single domestication event of the

auroch (Bos$primigenius)occurred in theFertileCrescent around10000years

ago(Taberletetal2011)FollowingdomesticationcattlefollowedtheNeolithic

expansionofagricultureandcolonisedEurope(seeAjmoneMMarsanetal2010

for a review) Many thousand years of natural and artificial selection have

shaped cattle phenotype and genotype to meet environmental conditions and

human needs and together with demographic events and genetic drift have

started thedifferentiation of the bovine species in diverse and locally adapted

populations This differentiation process accelerated in the last two centuries

whenthebreedconceptwasinventedsothatnowthisspeciescountshundreds

of different breeds with high diversity in coat colour body size production

aptitudeadaptationtoclimate tolerancetodiseaseetc(Taberletetal2008

Elsiketal2009TheBovineHapMapConsortium2009)

Thisdiversityisnowbeingprogressivelylostwiththeextinctionofmanylocal

populationsmainlyforeconomicreasonsSelectioneffortshavethereforebeen

focussedonafewoutperformingindustrialbreedsthatarepresentlyselectedfor

specialiseddairyorbeefproductionandsometimesdoublepurpose

In these breeds breeding schemes are well developed and have been

traditionally based on the genetic evaluation of sires and dams using data

collectedfromperformanceandprogenytestingandunbiasedlinearprediction

modelsThesearefoundedontheFisherinfinitesimalmodellingofquantitative

trait variation and need no genomic information to estimate breeding values

(EBVs) The genetic progress recorded using this approach has been relevant

particularly for traits having medium to high heritability In Italian Holstein

genetic trends for themain production traits (milk protein and fat yield) are

positiveHoweveraccurateestimateofbreedingvalueshasahighcostinterms

4

of investment and time so that 5 to 6 years are needed to have a dairy bullprogenytested(wwwanafiit)Theadventofgenomicselection(Meuwissenetal2001)markedamilestoneindairycattlebreedingTheideawasdevelopedbeforethemoleculartoolswereinplaceforimplementitinpracticeandhasbecomearealityonlyrecentlyInthisapproach progeny tested bulls are genotyped at many thousand loci and thevalue of their EBVs used to estimate the value of each allele at each locusgenotyped These values are then used to predict the genetic value of younganimals on the basis of their genotype As a result animals from the newgeneration receive an EBV with the same accuracy of estimates obtained byprogenytestingbutmuchearlierThislowerthegenerationintervalandgreatlyincreasestherateofgeneticprogress indairypopulationsThe initial ideawasso good that nowadays genomic selection is widely applied in industrialcountriestothemajordairycattlebreedsWiththedecreaseofgenotypingcostit is likely that its use will be soon expanded to animals selected for otherpurposes (eg beef cattle) and other species (eg pigs) However genomicselection is an extension of traditional selection It is extremely useful inbreedingforcomplextraitsbutgivesnobiologicalinformationonthegenesthatcontrolthemIfthetraditionalselectionviewsasiregenomeasabigblackboxhaving a value but no clue why genomic selection views the same genomedisassembled inmany small boxes but still black and stillwithno clueon thereasonoftheirvalueIt is reasonable to think that a deeper understanding of the biologicalmechanisms controlling traits may further increase the accuracy of genomicevaluationsIndeedthenewtechnologiesoffernewopportunitiesinthissensebyloweringcostsandincreasingprecisionofdataproductionThis thesis is exploring the use of these technologies for retrieving biologicalinformationfromgenomicdataItusesmethodsnowconsideredldquotraditionalrdquoingenomics (GWAS and analysis of selection sweeps) and less traditional ones(gene centered GWAS and exome analysis) All data used here have beenproducedwithin twonationalprojectson farmanimalgenomics fundedby theItalianMinistryofAgriculture

5

M SELMOL ldquoRicerca e innovazione nelle attivitagrave di miglioramento genetico

animalemediante tecniche di geneticamolecolare per la competitivitagrave del

sistemazootecniconazionaleIrdquo(ResearchandInnovationinanimalbreeding

sheepgoatbuffalohorseanddonkeyI)

M INNOVAGENldquoRicercaeinnovazionenelleattivitagravedimiglioramentogenetico

animalemediante tecniche di geneticamolecolare per la competitivitagrave del

sistema zootecnico nazionale IIrdquo (Research and Innovation in animal

breedingsheepgoatbuffalohorseanddonkeyII)

In these projects the research group I work with M Istituto di Zootecnica at

Universitagrave Cattolica del Sacro Cuore di Piacenza M coordinated all the research

activitiesconcerningtheapplicationofgenomicstechniquesindairycattle

Contentsofthethesis

The thesis is structured into a short introduction (this first chapter) 5 core

chapters(Chapter2to6)andashortConclusion(Chapter7)Withtheexception

of Chapter 6 still in preparation the other core chapters containmanuscripts

that at time of writing are under review for their publication in international

peerMreviewedjournalsormanuscript

InChapter2selectionsignatureindairyandbeefcattleareinvestigatedwiththe

Illumina BovineSNP50 Genotyping BeadChip Five Italian cattle breeds (Italian

Holstein Italian Brown Italian Simmental Marchigiana and Piedmontese) are

analyzedwiththerEHHmethod(Sabetietal2002)tofindsignaturesofrecent

selectionCandidategenesinthevicinityofsignaturessharedbyeitherdairyor

beefbreeds are identified and theirpossible role indairy andbeefproduction

discussed

In Chapter 3 a GWAS analysis is run to investigate production traits in three

Italian dairy cattle breeds (Italian Holstein Italian Brown Italian Simmental)

Data are analysed with a well established method (Amin et al 2007)

6

implemented in theGenABELRpackage (Aulchenkoet al 2007) andwith the

newgeneMbasedmethodAssociationisinvestigatedbetween50KSNPsandmilk

production(milkyield)andqualitytraits(percentageofmilkproteinandfat)

InChapter4thegeneMbasedmethodusedinthepreviouschapter3inspiredby

theVEGASapproach(Liuetal2010)isdescribedThesoftwarewasdeveloped

andreMcodedinourlaboratoryinordertopermittheanalysisofanyspeciesin

additiontohumanMultiMtestingcorrectionandplotrepresentationcompletethe

toolsmadeavailable

InChapter5wetesttheimpactoftheuseofdifferentreferencesequencesinthe

imputation step Italian Simmental data are used to test variation in genomeM

wide inputation accuracy using different cattle reference sequences (BTAU42

UMD31 and BTAU46) To increase the analyses reliability four different

inputationsoftwaresaretested

InChapter6Holsteindataareanalysedcombiningexomesequencesand800K

SNPs data to seek deleterious recessive variantsWhile in natural populations

the destiny of these variants is to disappear in livestock breeds they can be

maintained and disseminated if carried by a highly productive sire The

intersection of deleterious mutations identified by sequencing and lack of

homozygousregionsidentifiedbytheanalysisofthepopulationwiththeHDSNP

chipidentifiedgenescandidatetocarryrecessivedeleteriousvariants

Objective

Themainobjectiveofthethesiswastoinvestigatethebovinegenomewiththe

aid new highMthroughput technology and to explore new strategies for linking

biology (genes and their function) to traits This linkage once validated may

increase the accuracy of genomic prediction and find application in selection

schemesegineradicatingthemostdeleteriousvariantsandorintheplanning

ofmatings

7

ReferencesAjmoneMMarsanPGarciaJFLenstraJA2010OntheoriginofcattleHow

aurochsbecamecattleandcolonizedtheworldEvolAnthropol19148ndash157doi101002evan20267

AminNvanDuijnCMAulchenkoYS2007AGenomicBackgroundBasedMethodforAssociationAnalysisinRelatedIndividualsPLoSONE2e1274doi101371journalpone0001274

AulchenkoYSKoningDMJdeHaleyC2007GenomewideRapidAssociationUsingMixedModelandRegressionAFastandSimpleMethodForGenomewidePedigreeMBasedQuantitativeTraitLociAssociationAnalysisGenetics177577ndash585doi101534genetics107075614

ElsikCGTellamRLWorleyKC2009TheGenomeSequenceofTaurineCattleAWindowtoRuminantBiologyandEvolutionScience324522ndash528doi101126science1169588

LiuJZMcraeAFNyholtDRMedlandSEWrayNRBrownKMHaywardNKMontgomeryGWVisscherPMMartinNGMacgregorS2010AVersatileGeneMBasedTestforGenomeMwideAssociationStudiesTheAmericanJournalofHumanGenetics87139ndash145doi101016jajhg201006009

MeuwissenTHHayesBJGoddardME2001PredictionoftotalgeneticvalueusinggenomeMwidedensemarkermapsGenetics1571819ndash1829

SabetiPCReichDEHigginsJMLevineHZPRichterDJSchaffnerSFGabrielSBPlatkoJVPattersonNJMcDonaldGJAckermanHCCampbellSJAltshulerDCooperRKwiatkowskiDWardRLanderES2002DetectingrecentpositiveselectioninthehumangenomefromhaplotypestructureNature419832ndash837doi101038nature01140

TaberletPCoissacEPansuJPompanonF2011ConservationgeneticsofcattlesheepandgoatsCRBiol334247ndash254doi101016jcrvi201012007

8

TaberletPValentiniARezaeiHRNaderiSPompanonFNegriniR

AjmoneMMarsanP2008Arecattlesheepandgoatsendangered

speciesMolEcol17275ndash284doi101111j1365M294X200703475x

TheBovineHapMapConsortium2009GenomeMWideSurveyofSNPVariation

UncoverstheGeneticStructureofCattleBreedsScience324528ndash532

doi101126science1167936

9

CHAPTER2

Relative(extendedamphaplotype(homozygosity)(rEHH)ampsignalsampacrossampbreedsamprevealampdairyampand$beef$specificampsignaturesampofampselection

LBomba1ELNicolazzi2MMilanesi1etal1IstitutodiZootecnicaUCSCviaEmiliaParmense8429122PiacenzaItaly2FondazioneParcoTecnologicoPadanoViaEinsteinLocCascinaCodazza26900LodiItalyTheseauthorscontributedequallytothiswork

____________________SubmittedtoGeneticSelectionEvolution

Abstract

Background

Nowadays many methods have been implemented to scan for signatures of

selection at genomescale level within and across breeds Extended haplotype

homozygosity is a proficient approach to detect genome regions under recent

selective pressure within breeds The objective of this study is to identify

common regions under positive recent selection shared by most represented

dairyandbeefItaliancattlebreeds

Results

A total of 3220 animals from Italian Holstein (2179) Italian Brown (775)

Simmental (493) Marchigiana (485) and Piedmontese (379) breeds were

genotyped with the Illumina BovineSNP50 BeadChip v1 in the frame of the

ItalianlivestockgenomicprojectsldquoSelMolrdquoandldquoProzoordquoAfterstandardcleaning

proceduresgenotypeswerephasedandcorehaplotypesidentifiedThedecayof

Linkage Disequilibrium (LD) for each core haplotype was assessedmeasuring

the extended haplotype homozygosity (EHH) To correct for the lack of good

estimatesof local recombinationrates therelativeEHH(rEHH)wascalculated

for each corehaplotypeGenomic regionsunder recentpositive selectionwere

identifiedasthosehighlyfrequentcorehaplotypeswithsignificantrEHHvalues

Significant independentcorehaplotypeswerealignedacrossbreeds to identify

signalsofrecentselectionsharedbydairyandsharedbybeefbreedsOverall82

and 87 common regions under selectionwere detected among dairy and beef

cattlebreedsrespectivelyBioinformaticsanalysisidentified244and232genes

mapping in these common genomic regions Pathway and network analysis of

significant genes revealed molecular functions biologically related to signal of

selectionsdetectedinmilkormeatproductiontypes

Conclusions

Resultssuggest that themultibreedapproach isamethodto identifyselection

signatures in cattle breeds under selection for the same production goal

12

allowing to better pinpoint the genomic regions of interest in dairy and beef

production

Background

The advent of genomic technologies and the consequent availability of single

nucleotidepolymorphism(SNP)datahaveenabledthestudyofselectionevents

on the cattle genome (Hayes et al 2008 Qanbari et al 2010) Such selection

signals arise from environmental or anthropogenic pressures and help to

understand the causes that led to breed formation These studies are usually

addressedinaldquotopdownrdquoapproach(ShahzadandLoor2012)fromgenotypeto

phenotypeinwhichgenomicdataisstatisticallyevaluatedtodetectdirectional

selectionThephenotypeneededtorunselectionsignaturesisintendedinavery

generalsenseabreedaproductionaptitudeorevenageographicareaoforigin

Thisapproachholdsthepotentialto investigatetraitsasadaptationtoextreme

climatesanddifferentfeedingandhusbandrysystemsresiliencetodiseasesetc

thatareveryexpensivedifficultandsometimesimpossibletostudywithclassic

GWAS (GenomeWide Association Study) approaches Therefore it is expected

thatgenomic informationretrievedinthesestudiesonlypartiallyoverlapsthat

identified by GWAS on specific traits Searching for these ldquosignatures of

selectionrdquo is of interest for understanding molecular mechanisms underlying

important biological processes and providing new insights to functional

biologicalinformation(egdiseaseandorimportanteconomictraits(Nielsenet

al 2007)) To date many methods have been implemented to scan these

selectionsignaturesatthegenomiclevelMethodsdifferwithrespecttoseveral

properties (Lenstra et al 2012) there are methods that perform analyses

withinbreedoracrossbreedsthatcomparealleleorhaplotypefrequenciesand

size that detect recent or ancient selection events either still segregating or

fixatedintopopulations(Lenstraetal2012OrsquoBrienetal2014Utsunomiyaet

al2013)Differentmethodshavedifferentsensitivityandrobustnesseg they

maybeinfluencedatadifferentextentbymarkerascertainmentbiasanduneven

distributionofrecombinationhotspotsalongthegenome

13

Extended haplotype homozygosity (EHH) is a method that has been used in

many animal species including cattle (Pan et al 2013 Qanbari et al 2010

Zhangetal2006)TheEHHmethodwasfirstdevelopedbySabetietal(2002)

for human population analysis The main objective of this methodology is to

identifylongrangehaplotypesinheritedwithoutrecombinationeventsMorein

detailunderaneutralevolutionmodelchangesinallelefrequenciesaredriven

onlybygeneticdriftInthisscenarioanewvariantwouldrequirealongtimeto

reach high frequency in the population and the Linkage Disequilibrium (LD)

surroundingitwoulddecayduetorecombinationevents(Kimura1984)Inthe

caseofpositiveselectionaquickriseinfrequencyofabeneficialmutationina

relatively short time will preserve the original haplotype structure as the

number of recombination events would be limited Therefore a signature of

positiveselectionisdefinedasaregionwithanalleleinlongrangeLDlocatedin

anuncommonlyhighfrequencyhaplotype

One of themost interesting features of thismethod is that it is able to detect

genomic regions that are under recent selection These recent selection

signatures can be therefore used to identify the genomic regions involved in

breedformationInadditionEHHmethoddoesnotrequireanydefinitionofan

ancestralallele(asneededintheintegratedhaplotypescoreiHS(Voightetal

2006) and is suited to be applied to SNP data because it is less sensitive to

ascertainmentbiasthanothermethods(Nielsenetal2007)HowevertheEHH

methodgeneratesa largenumberof falsepositiveandnegative resultsdue to

heterogeneous recombination rates along the genome (Qanbari et al 2010)

Moreover another drawback not specific of EHH but shared by all selection

signaturemethod is the lack of robust inferences able to discern true signals

fromthosegeneratedbychance(Kemperetal2014)

To partially account for these limitations Sabeti et al (2002) developed the

relative extended haplotype homozygosity (rEHH) including an empirical

methodologytoobtainsignificantsignalsTherEHHofaputativecorehaplotype

compares its original EHH valuewith that of other haplotypes at that specific

locus using all other haplotypes as control for local variation in the

recombination rate Therefore it identifies genomic regions carrying variants

14

under selection that are still segregating in the population analysed Although

thesemethodswere conceived for human populations theywere successfully

appliedto livestockspeciessuchaspig(Maetal2012)andcattle(Qanbariet

al2010)

After domestication about 10000 years ago in the fertile crescent taurine

cattle have colonized Europe and Africa and have been selected for different

human needs (Taberlet et al 2011) Particularly in the last centuries the

anthropic pressure led to the formation of hundreds of specialized breeds

adapted to different environmental conditions and linked to local tradition

constitutingagenepooldeservingattentionforconservation(AjmoneMarsanet

al2010)Someofthesebreedshavebeenselectedspecificallyfordairybeefor

bothproductiontypesfollowingstrongartificialselectionforthesetraits(Hayes

etal2009)Thepresentstudyaimsat identifyingsignalsof recentdirectional

selectionusingrEHHmethodindairyandbeefproductiontypesusingdatafrom

five Italian dairy beef and dual purpose cattle breeds We focused on the

significant core haplotypes scanned by rEHH that are shared among breeds of

thesameproductiontypeThenweusedabioinformaticsapproachto identify

positionalcandidategeneslyingwithinthegenomicregionsunderselectionand

investigatetheirbiologicalrole

Methods

SampleAnimalsandGenotyping

A total of 4311 bulls of five Italian dairy beef anddual purpose breedswere

genotyped with the Illumina BovineSNP50 BeadChip v1 (Illumina San Diego

CA)joiningthegenotypingeffortsoftwoItalianprojects(namelyldquoSelMolrdquoand

ldquoProzoordquo)Thedatasetincluded101replicatesand773siresonspairsusedfor

downstreamqualitycheckofdataproducedIndetailgenotypesof2954dairy

(2179 ItalianHolsteinand775 ItalianBrown)864beef (485Marchigianaand

379 Piedmontese) and 493 dual purpose (Italian Simmental) bulls were

availableDataqualitycontrol(QC)wasperformedintwostagesafirststageon

15

animals independently for each breed applying the same methods and

thresholdsandasecondstageappliedonmarkersacrossall theindividualsof

thedatasetThefirststageexcludedindividualswithunexpectedlyhigh(gt02)

Mendelian errors for fatherson couples andwith low call rates (lt95)The

secondstageexcludedi)SNPswithle25missingvaluesinthewholedataset

orcompletelymissing inonebreed ii)SNPswithminorallele frequencyle5

and iii) SNPs in sexual chromosomes and with unassigned chromosome or

physicalposition

EstimationofrEHH

HaplotypeswereobtainedusingfastPHASEsoftwareconsideringdefaultoptions

(ScheetandStephens2006)andwererunbreedwiseandchromosomewisein

eachbreedPedigreeinformationforallbullswereprovidedbytheirrespective

breederassociations andwereused to filteroutdirect relatives (in fatherson

pairsthesonwasmaintainedinthedatasetandthefatherdiscarded)andover

representedfamilies(amaximumof5randomlychosenindividualsperhalfsib

familywasallowed)Thefinaldatasetcontainingtheseldquolessrelatedrdquoanimalswill

behenceforthcalled ldquononredundantrdquodatasetThisnonredundantdatasetwas

used also to calculate within breed pairwise wholegenome Linkage

Disequilibrium(LD)Ther2statisticforallpairsofmarkerswasobtainedusing

PLINK software v107 (Purcell et al 2007)Thedecayof LDwas assessedby

averagingr2valuesupto1Mbdistance

To test if the population structure influenced the detection of rEHH we

replicated all analyses on the whole Italian Holstein dataset without any

population correction (ie ldquoredundantrdquo dataset) focusing on genes or gene

clustersthatarewellknowntobeunderrecentselectionincattle(ieldquocandidate

regionsrdquo) In particularwe focused on the casein polled gene cluster and coat

colourgenes(MC1RandKIT(Kemperetal2014Qanbarietal2010))

The EHH and rEHH calculations were performed using Sweep v11 software

(Sabetietal2002)Somedefaultprogramsettingshadtobemodifiedtoadapt

theseanalyses to thebovinegenomeas thesoftwarewasoriginallydeveloped

forhumangeneticanalysesSpecificallylocalrecombinationratesbetweenSNPs

16

wereapproximatedto1cMperMbEHHandrEHHcalculationswereperformed

breedwise and chromosomewise using an automatic core selection with

defaultoptions(ieconsideringlongestnonoverlappingcoresandlimitingcores

toatleast3andnomorethan20SNPs)asinQanbarietal(2010)AlthoughEHH

and rEHH values were obtained for all core haplotypes only those with

frequency ge 25 were retained for further analyses The (empirical) rEHH

significancethresholdwasobtainedbydividingtherEHHvaluesinto20binsof

5 frequency range each logtransforming withinbin values to achieve

normality Core haplotypes with pvalues lt 005 were considered significant

DetectionofcommonselectionsignatureswasobtainedcomparingrEHHvalues

acrossgenomicregionsacrossbreeds

Breedgroupingaccordingtoproductiontype

Toexamine thesignals common toeachof the twoproduction types (iedairy

andbeef) corehaplotypes thatsharedoneormoreSNP inat least twobreeds

from the same production type were selected The dual purpose Italian

Simmentalwas included inbothdairy (ItalianHolsteinand ItalianBrown)and

beef groups (Piedmontese and Marchigiana) since potentially it possesses

haplotypes selected for both production types All downstream analyses were

performedseparatelyforthedairyandthebeefproductiontypes

Detectionandannotationofcandidategenes

Genome annotation was performed using Biomart

(httpwwwbiomartorgindexhtml) a comprehensive source of gene

annotationformanyspeciesprovidedbyEnsembl(wwwensemblorg)Thelist

ofsharedhaplotypesregions(inbp)fordairyandbeefwasusedasinputfilein

theBiomartwebinterfaceThesetofgenesretrievedbyBiomartwasthenused

as input for theCanonicalpathwayandregulatorynetworkanalyses Ingenuity

PathwayAnalysis toolversion80(IPA IngenuityregSystems IncRedwoodCity

CA httpwwwingenuitycom) and a substantial examination of published

literaturewasusedtoexaminethefunctionalrelationshipsamongtheresulting

genes The IPA operates with a proprietary knowledge database providing

complementarypathwayanalysisforseveralspeciesincludingcattleIntheIPA

17

analyses the significance of each biological function identified was estimated

using Fisherrsquos exact test A Benjamini and Hochberg correction for multiple

testingwas calculated for function and canonical pathways For eachNetwork

IPAcomputedascore(pscore=log10(pvalue))fittingthesetstudiedgenesand

a list of biological functionspresent in the IPAknowledgedatabase The score

considered thenumberofgenes in thenetworkand thesizeof thenetwork to

estimatehowrelevantthisnetworkwasaccordingtothelistofgenesprovided

Thenthescorewasusedtorankthenetworks

Results

DatasetQC

Therepeatabilityobtainedfromthe101replicatespresentinthewholedataset

washigherthan998AfterthetwoQCstages105individualsand9730SNPs

were discarded Moreover after phasing 1292 individuals were removed to

reducethe largenumberofsibfamiliespresent intheredundantdatasetAfter

quality control procedures the final dataset had 44271 SNPs and 1132 514

393 410 and 364 individuals from Italian Holstein Italian Brown Italian

SimmentalMarchigianaandPiedmontesebullsrespectively(Table1)

18

Table1JNumberofanimalsgenotypedbeforeandafterQCTotalnumberofanimalgenotypedandrelativenumberofanimaldiscardedafterQCanalysis

Detectionofselectionsignatures

Sweepv11 softwaredetectedrEHHsignalswitha frequencyhigher than25overlapping the test candidate regions in both datasets corrected or not forpopulation structure In the redundant dataset we found only one significantrEHHsignal(caseincluster)whereasinthenonredundantdatasetweidentified5significantrEHHsignalsoverlappingallthetestcandidateregionsconsidered(caseinclusterpolledclusterMC1R)(Table2)

BreedTotal

Genotyped

ED15

misAN

ED1

REPL

ED1

MENDCleaned

HOL 2179 40 31 5 2093

BRW 775 6 16 4 749

SIM 493 6 6 2 479

MAR 485 37 38 410

PIE 379 5 10 364

19

Table2JComparisonofcandidateregionrEHHinnonJredundantandredundantdatasetrEHH overlapping candidate regions in both corrected (nonredundant) and not corrected

(redundant)datasetsforpopulation

Redundant NonRedundant

Candidate

geneBTA1

Closest

SNP(bp)CH2range CH2Frequency

rEHHJ

log(pval)CH2Frequency

rEHHJ

log(pval)

POLLED

Cluster1 1981154 18974181981154 H1079 093037 H1078 167174

MC1R 18 13657912 1331772014007505 H1069 120096 H1069 077167

KIT_BOVIN 6 72821175 7250492172821175 H1054 030030 H1054 033048

Casein

Cluster6 88427760 8835009888452835 H1047 094139 H1046 148133

Casein

Cluster6 88427760 8835009888452835 H2032 002003 H2033 002003

In total5526567847724388and4049corehaplotypeswitha frequency

higher than 25 were detected on Italian Holstein Italian Brown Italian

Simmental Marchigiana and Piedmontese breeds respectively A total of 838

866 740 692 and 613 core haplotypes were found significant outliers (see

Materials and Methods) in the aforementioned breeds Table 3 shows the

distribution of the total and significant core haplotypes per chromosome and

breed

20

Tableamp3amp(ampCoreamphaplotypesampdistributionampDistributionoftotalandsignificantcorehaplotypesperchromosomeandbreed

ampamp ItalianampampHolsteinamp

ItalianampampBrownamp

ItalianampSimmentalamp Marchigianaamp Piedmonteseamp

BTA1amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

1amp 389 66 389 68 335 49 310 51 288 46

2amp 312 47 314 52 267 44 257 42 238 33

3amp 292 48 313 62 264 46 236 34 227 34

4amp 268 46 279 50 245 41 237 39 229 39

5amp 223 32 231 29 200 30 183 27 171 22

6amp 291 51 307 40 265 45 252 38 233 42

7amp 249 37 253 46 214 39 195 33 204 27

8amp 268 39 270 38 235 35 223 32 209 31

9amp 227 37 228 41 210 37 165 22 187 30

10amp 219 36 234 34 192 25 196 26 168 25

11amp 248 39 246 34 209 30 205 33 187 25

12amp 162 24 176 18 148 21 148 25 135 22

13amp 205 32 200 19 155 18 178 35 135 19

14amp 175 17 189 25 168 29 143 22 129 19

15amp 180 31 185 29 148 23 139 13 130 19

16amp 176 28 192 29 160 25 149 19 127 15

17amp 170 27 183 29 168 31 135 22 123 16

18amp 133 18 153 19 119 14 108 10 82 12

19amp 137 22 130 23 118 16 99 18 92 15

20amp 184 35 172 33 145 23 124 23 124 17

21amp 145 23 147 30 113 18 110 21 94 13

22amp 140 24 145 19 115 21 112 18 104 13

23amp 106 13 100 16 67 10 64 9 57 8

24amp 129 11 140 20 124 17 97 23 101 16

25amp 91 12 98 11 68 10 77 12 70 6

26amp 125 7 114 14 93 12 96 15 90 16

27amp 91 12 91 15 75 11 84 12 60 10

28amp 88 9 96 10 70 10 66 9 55 11

29amp 103 15 103 13 82 10 72 9 67 12

TOTamp 5526 838 5678 866 4772 740 4388 692 4049 613

1BostaurusAutosomes2Detectedcorehaplotypeswithafrequencyinthebreedhigherthan253Significantcorehaplotypesatplt005

Comparisonamptoamppreviouslyampreportedampdataamp

AnumberofstudieshavesearchedselectionsweepsinHolstein(Qanbarietal

2010) Brown (Qanbari et al 2011) and Simmental (Fan et al 2014) breeds

Since different methods are expected to identify signatures having different

characteristicsthecomparisonwithpreviousstudiesislimitedtothosestudies

using thesamemethodand thesamebreed(s)analysed inourstudyOnlyone

study analysed (German) HolsteinRFriesian cattle with rEHH (Qanbari et al

2010)Qanbarietal(2010)reportedcandidategenesandgeneclustersunder

Scuola di Dottorato per il Sistema Agro-alimentare

Doctoral School on the Agro-Food System

cycle XXVII

SSD AGR17

High-throughput sequencing technologies and genomics insights into the bovine species

Coordinator Chmo Prof Antonio Albanese

_______________________________________

Candidate Marco Milanesi Matriculation n 4011148

Tutor Prof Paolo AjmoneMarsan PhD Lorenzo Bomba PhD Stefano Capomaccio PhD Ezequiel Louis Nicolazzi

Academic Year 20132014

Alla$mia$famiglia$

A$mio$nonno$Mario$

INDEXamp

Chapteramp1ndashIntroduction 3

Chapteramp2Relativeextendedhaplotypehomozygosity(rEHH)signalsacrossbreedsrevealdairyandbeefspecificsignaturesofselection

11

Chapteramp3SearchingnewsignalsforproductivetraitsthroughgenebasedassociationanalysisinthreeItaliancattlebreeds

49

Chapteramp4ampMUGBASaspeciesfreegenebasedassociationsuite

81

Chapteramp5Imputationaccuracyisrobusttocattlereferencegenomeupdates

89

Chapteramp6QuestfordeleteriousrecessivevariantsinItalianHolsteinbullscombiningexomesequencingandhighdensitySNPanalysis

99

Chapteramp7ndashConclusion 167

amp

Acknowledgmentsampamp 173

CHAPTER1

Introduction

Background

The bovine species counts almost 15 billion of heads classified in some 800

breeds worldwide (httpfaostat3faoorg) The species is adapted to a wide

varietyofdifferentenvironmentandhusbandrysystemsandisoftenrearedfor

multiple purposes (Taberlet et al 2008) draft power milk meat leather

productionandotheruses(Elsiketal2009)

European cattle (Bos$ taurus) derive from a single domestication event of the

auroch (Bos$primigenius)occurred in theFertileCrescent around10000years

ago(Taberletetal2011)FollowingdomesticationcattlefollowedtheNeolithic

expansionofagricultureandcolonisedEurope(seeAjmoneMMarsanetal2010

for a review) Many thousand years of natural and artificial selection have

shaped cattle phenotype and genotype to meet environmental conditions and

human needs and together with demographic events and genetic drift have

started thedifferentiation of the bovine species in diverse and locally adapted

populations This differentiation process accelerated in the last two centuries

whenthebreedconceptwasinventedsothatnowthisspeciescountshundreds

of different breeds with high diversity in coat colour body size production

aptitudeadaptationtoclimate tolerancetodiseaseetc(Taberletetal2008

Elsiketal2009TheBovineHapMapConsortium2009)

Thisdiversityisnowbeingprogressivelylostwiththeextinctionofmanylocal

populationsmainlyforeconomicreasonsSelectioneffortshavethereforebeen

focussedonafewoutperformingindustrialbreedsthatarepresentlyselectedfor

specialiseddairyorbeefproductionandsometimesdoublepurpose

In these breeds breeding schemes are well developed and have been

traditionally based on the genetic evaluation of sires and dams using data

collectedfromperformanceandprogenytestingandunbiasedlinearprediction

modelsThesearefoundedontheFisherinfinitesimalmodellingofquantitative

trait variation and need no genomic information to estimate breeding values

(EBVs) The genetic progress recorded using this approach has been relevant

particularly for traits having medium to high heritability In Italian Holstein

genetic trends for themain production traits (milk protein and fat yield) are

positiveHoweveraccurateestimateofbreedingvalueshasahighcostinterms

4

of investment and time so that 5 to 6 years are needed to have a dairy bullprogenytested(wwwanafiit)Theadventofgenomicselection(Meuwissenetal2001)markedamilestoneindairycattlebreedingTheideawasdevelopedbeforethemoleculartoolswereinplaceforimplementitinpracticeandhasbecomearealityonlyrecentlyInthisapproach progeny tested bulls are genotyped at many thousand loci and thevalue of their EBVs used to estimate the value of each allele at each locusgenotyped These values are then used to predict the genetic value of younganimals on the basis of their genotype As a result animals from the newgeneration receive an EBV with the same accuracy of estimates obtained byprogenytestingbutmuchearlierThislowerthegenerationintervalandgreatlyincreasestherateofgeneticprogress indairypopulationsThe initial ideawasso good that nowadays genomic selection is widely applied in industrialcountriestothemajordairycattlebreedsWiththedecreaseofgenotypingcostit is likely that its use will be soon expanded to animals selected for otherpurposes (eg beef cattle) and other species (eg pigs) However genomicselection is an extension of traditional selection It is extremely useful inbreedingforcomplextraitsbutgivesnobiologicalinformationonthegenesthatcontrolthemIfthetraditionalselectionviewsasiregenomeasabigblackboxhaving a value but no clue why genomic selection views the same genomedisassembled inmany small boxes but still black and stillwithno clueon thereasonoftheirvalueIt is reasonable to think that a deeper understanding of the biologicalmechanisms controlling traits may further increase the accuracy of genomicevaluationsIndeedthenewtechnologiesoffernewopportunitiesinthissensebyloweringcostsandincreasingprecisionofdataproductionThis thesis is exploring the use of these technologies for retrieving biologicalinformationfromgenomicdataItusesmethodsnowconsideredldquotraditionalrdquoingenomics (GWAS and analysis of selection sweeps) and less traditional ones(gene centered GWAS and exome analysis) All data used here have beenproducedwithin twonationalprojectson farmanimalgenomics fundedby theItalianMinistryofAgriculture

5

M SELMOL ldquoRicerca e innovazione nelle attivitagrave di miglioramento genetico

animalemediante tecniche di geneticamolecolare per la competitivitagrave del

sistemazootecniconazionaleIrdquo(ResearchandInnovationinanimalbreeding

sheepgoatbuffalohorseanddonkeyI)

M INNOVAGENldquoRicercaeinnovazionenelleattivitagravedimiglioramentogenetico

animalemediante tecniche di geneticamolecolare per la competitivitagrave del

sistema zootecnico nazionale IIrdquo (Research and Innovation in animal

breedingsheepgoatbuffalohorseanddonkeyII)

In these projects the research group I work with M Istituto di Zootecnica at

Universitagrave Cattolica del Sacro Cuore di Piacenza M coordinated all the research

activitiesconcerningtheapplicationofgenomicstechniquesindairycattle

Contentsofthethesis

The thesis is structured into a short introduction (this first chapter) 5 core

chapters(Chapter2to6)andashortConclusion(Chapter7)Withtheexception

of Chapter 6 still in preparation the other core chapters containmanuscripts

that at time of writing are under review for their publication in international

peerMreviewedjournalsormanuscript

InChapter2selectionsignatureindairyandbeefcattleareinvestigatedwiththe

Illumina BovineSNP50 Genotyping BeadChip Five Italian cattle breeds (Italian

Holstein Italian Brown Italian Simmental Marchigiana and Piedmontese) are

analyzedwiththerEHHmethod(Sabetietal2002)tofindsignaturesofrecent

selectionCandidategenesinthevicinityofsignaturessharedbyeitherdairyor

beefbreeds are identified and theirpossible role indairy andbeefproduction

discussed

In Chapter 3 a GWAS analysis is run to investigate production traits in three

Italian dairy cattle breeds (Italian Holstein Italian Brown Italian Simmental)

Data are analysed with a well established method (Amin et al 2007)

6

implemented in theGenABELRpackage (Aulchenkoet al 2007) andwith the

newgeneMbasedmethodAssociationisinvestigatedbetween50KSNPsandmilk

production(milkyield)andqualitytraits(percentageofmilkproteinandfat)

InChapter4thegeneMbasedmethodusedinthepreviouschapter3inspiredby

theVEGASapproach(Liuetal2010)isdescribedThesoftwarewasdeveloped

andreMcodedinourlaboratoryinordertopermittheanalysisofanyspeciesin

additiontohumanMultiMtestingcorrectionandplotrepresentationcompletethe

toolsmadeavailable

InChapter5wetesttheimpactoftheuseofdifferentreferencesequencesinthe

imputation step Italian Simmental data are used to test variation in genomeM

wide inputation accuracy using different cattle reference sequences (BTAU42

UMD31 and BTAU46) To increase the analyses reliability four different

inputationsoftwaresaretested

InChapter6Holsteindataareanalysedcombiningexomesequencesand800K

SNPs data to seek deleterious recessive variantsWhile in natural populations

the destiny of these variants is to disappear in livestock breeds they can be

maintained and disseminated if carried by a highly productive sire The

intersection of deleterious mutations identified by sequencing and lack of

homozygousregionsidentifiedbytheanalysisofthepopulationwiththeHDSNP

chipidentifiedgenescandidatetocarryrecessivedeleteriousvariants

Objective

Themainobjectiveofthethesiswastoinvestigatethebovinegenomewiththe

aid new highMthroughput technology and to explore new strategies for linking

biology (genes and their function) to traits This linkage once validated may

increase the accuracy of genomic prediction and find application in selection

schemesegineradicatingthemostdeleteriousvariantsandorintheplanning

ofmatings

7

ReferencesAjmoneMMarsanPGarciaJFLenstraJA2010OntheoriginofcattleHow

aurochsbecamecattleandcolonizedtheworldEvolAnthropol19148ndash157doi101002evan20267

AminNvanDuijnCMAulchenkoYS2007AGenomicBackgroundBasedMethodforAssociationAnalysisinRelatedIndividualsPLoSONE2e1274doi101371journalpone0001274

AulchenkoYSKoningDMJdeHaleyC2007GenomewideRapidAssociationUsingMixedModelandRegressionAFastandSimpleMethodForGenomewidePedigreeMBasedQuantitativeTraitLociAssociationAnalysisGenetics177577ndash585doi101534genetics107075614

ElsikCGTellamRLWorleyKC2009TheGenomeSequenceofTaurineCattleAWindowtoRuminantBiologyandEvolutionScience324522ndash528doi101126science1169588

LiuJZMcraeAFNyholtDRMedlandSEWrayNRBrownKMHaywardNKMontgomeryGWVisscherPMMartinNGMacgregorS2010AVersatileGeneMBasedTestforGenomeMwideAssociationStudiesTheAmericanJournalofHumanGenetics87139ndash145doi101016jajhg201006009

MeuwissenTHHayesBJGoddardME2001PredictionoftotalgeneticvalueusinggenomeMwidedensemarkermapsGenetics1571819ndash1829

SabetiPCReichDEHigginsJMLevineHZPRichterDJSchaffnerSFGabrielSBPlatkoJVPattersonNJMcDonaldGJAckermanHCCampbellSJAltshulerDCooperRKwiatkowskiDWardRLanderES2002DetectingrecentpositiveselectioninthehumangenomefromhaplotypestructureNature419832ndash837doi101038nature01140

TaberletPCoissacEPansuJPompanonF2011ConservationgeneticsofcattlesheepandgoatsCRBiol334247ndash254doi101016jcrvi201012007

8

TaberletPValentiniARezaeiHRNaderiSPompanonFNegriniR

AjmoneMMarsanP2008Arecattlesheepandgoatsendangered

speciesMolEcol17275ndash284doi101111j1365M294X200703475x

TheBovineHapMapConsortium2009GenomeMWideSurveyofSNPVariation

UncoverstheGeneticStructureofCattleBreedsScience324528ndash532

doi101126science1167936

9

CHAPTER2

Relative(extendedamphaplotype(homozygosity)(rEHH)ampsignalsampacrossampbreedsamprevealampdairyampand$beef$specificampsignaturesampofampselection

LBomba1ELNicolazzi2MMilanesi1etal1IstitutodiZootecnicaUCSCviaEmiliaParmense8429122PiacenzaItaly2FondazioneParcoTecnologicoPadanoViaEinsteinLocCascinaCodazza26900LodiItalyTheseauthorscontributedequallytothiswork

____________________SubmittedtoGeneticSelectionEvolution

Abstract

Background

Nowadays many methods have been implemented to scan for signatures of

selection at genomescale level within and across breeds Extended haplotype

homozygosity is a proficient approach to detect genome regions under recent

selective pressure within breeds The objective of this study is to identify

common regions under positive recent selection shared by most represented

dairyandbeefItaliancattlebreeds

Results

A total of 3220 animals from Italian Holstein (2179) Italian Brown (775)

Simmental (493) Marchigiana (485) and Piedmontese (379) breeds were

genotyped with the Illumina BovineSNP50 BeadChip v1 in the frame of the

ItalianlivestockgenomicprojectsldquoSelMolrdquoandldquoProzoordquoAfterstandardcleaning

proceduresgenotypeswerephasedandcorehaplotypesidentifiedThedecayof

Linkage Disequilibrium (LD) for each core haplotype was assessedmeasuring

the extended haplotype homozygosity (EHH) To correct for the lack of good

estimatesof local recombinationrates therelativeEHH(rEHH)wascalculated

for each corehaplotypeGenomic regionsunder recentpositive selectionwere

identifiedasthosehighlyfrequentcorehaplotypeswithsignificantrEHHvalues

Significant independentcorehaplotypeswerealignedacrossbreeds to identify

signalsofrecentselectionsharedbydairyandsharedbybeefbreedsOverall82

and 87 common regions under selectionwere detected among dairy and beef

cattlebreedsrespectivelyBioinformaticsanalysisidentified244and232genes

mapping in these common genomic regions Pathway and network analysis of

significant genes revealed molecular functions biologically related to signal of

selectionsdetectedinmilkormeatproductiontypes

Conclusions

Resultssuggest that themultibreedapproach isamethodto identifyselection

signatures in cattle breeds under selection for the same production goal

12

allowing to better pinpoint the genomic regions of interest in dairy and beef

production

Background

The advent of genomic technologies and the consequent availability of single

nucleotidepolymorphism(SNP)datahaveenabledthestudyofselectionevents

on the cattle genome (Hayes et al 2008 Qanbari et al 2010) Such selection

signals arise from environmental or anthropogenic pressures and help to

understand the causes that led to breed formation These studies are usually

addressedinaldquotopdownrdquoapproach(ShahzadandLoor2012)fromgenotypeto

phenotypeinwhichgenomicdataisstatisticallyevaluatedtodetectdirectional

selectionThephenotypeneededtorunselectionsignaturesisintendedinavery

generalsenseabreedaproductionaptitudeorevenageographicareaoforigin

Thisapproachholdsthepotentialto investigatetraitsasadaptationtoextreme

climatesanddifferentfeedingandhusbandrysystemsresiliencetodiseasesetc

thatareveryexpensivedifficultandsometimesimpossibletostudywithclassic

GWAS (GenomeWide Association Study) approaches Therefore it is expected

thatgenomic informationretrievedinthesestudiesonlypartiallyoverlapsthat

identified by GWAS on specific traits Searching for these ldquosignatures of

selectionrdquo is of interest for understanding molecular mechanisms underlying

important biological processes and providing new insights to functional

biologicalinformation(egdiseaseandorimportanteconomictraits(Nielsenet

al 2007)) To date many methods have been implemented to scan these

selectionsignaturesatthegenomiclevelMethodsdifferwithrespecttoseveral

properties (Lenstra et al 2012) there are methods that perform analyses

withinbreedoracrossbreedsthatcomparealleleorhaplotypefrequenciesand

size that detect recent or ancient selection events either still segregating or

fixatedintopopulations(Lenstraetal2012OrsquoBrienetal2014Utsunomiyaet

al2013)Differentmethodshavedifferentsensitivityandrobustnesseg they

maybeinfluencedatadifferentextentbymarkerascertainmentbiasanduneven

distributionofrecombinationhotspotsalongthegenome

13

Extended haplotype homozygosity (EHH) is a method that has been used in

many animal species including cattle (Pan et al 2013 Qanbari et al 2010

Zhangetal2006)TheEHHmethodwasfirstdevelopedbySabetietal(2002)

for human population analysis The main objective of this methodology is to

identifylongrangehaplotypesinheritedwithoutrecombinationeventsMorein

detailunderaneutralevolutionmodelchangesinallelefrequenciesaredriven

onlybygeneticdriftInthisscenarioanewvariantwouldrequirealongtimeto

reach high frequency in the population and the Linkage Disequilibrium (LD)

surroundingitwoulddecayduetorecombinationevents(Kimura1984)Inthe

caseofpositiveselectionaquickriseinfrequencyofabeneficialmutationina

relatively short time will preserve the original haplotype structure as the

number of recombination events would be limited Therefore a signature of

positiveselectionisdefinedasaregionwithanalleleinlongrangeLDlocatedin

anuncommonlyhighfrequencyhaplotype

One of themost interesting features of thismethod is that it is able to detect

genomic regions that are under recent selection These recent selection

signatures can be therefore used to identify the genomic regions involved in

breedformationInadditionEHHmethoddoesnotrequireanydefinitionofan

ancestralallele(asneededintheintegratedhaplotypescoreiHS(Voightetal

2006) and is suited to be applied to SNP data because it is less sensitive to

ascertainmentbiasthanothermethods(Nielsenetal2007)HowevertheEHH

methodgeneratesa largenumberof falsepositiveandnegative resultsdue to

heterogeneous recombination rates along the genome (Qanbari et al 2010)

Moreover another drawback not specific of EHH but shared by all selection

signaturemethod is the lack of robust inferences able to discern true signals

fromthosegeneratedbychance(Kemperetal2014)

To partially account for these limitations Sabeti et al (2002) developed the

relative extended haplotype homozygosity (rEHH) including an empirical

methodologytoobtainsignificantsignalsTherEHHofaputativecorehaplotype

compares its original EHH valuewith that of other haplotypes at that specific

locus using all other haplotypes as control for local variation in the

recombination rate Therefore it identifies genomic regions carrying variants

14

under selection that are still segregating in the population analysed Although

thesemethodswere conceived for human populations theywere successfully

appliedto livestockspeciessuchaspig(Maetal2012)andcattle(Qanbariet

al2010)

After domestication about 10000 years ago in the fertile crescent taurine

cattle have colonized Europe and Africa and have been selected for different

human needs (Taberlet et al 2011) Particularly in the last centuries the

anthropic pressure led to the formation of hundreds of specialized breeds

adapted to different environmental conditions and linked to local tradition

constitutingagenepooldeservingattentionforconservation(AjmoneMarsanet

al2010)Someofthesebreedshavebeenselectedspecificallyfordairybeefor

bothproductiontypesfollowingstrongartificialselectionforthesetraits(Hayes

etal2009)Thepresentstudyaimsat identifyingsignalsof recentdirectional

selectionusingrEHHmethodindairyandbeefproductiontypesusingdatafrom

five Italian dairy beef and dual purpose cattle breeds We focused on the

significant core haplotypes scanned by rEHH that are shared among breeds of

thesameproductiontypeThenweusedabioinformaticsapproachto identify

positionalcandidategeneslyingwithinthegenomicregionsunderselectionand

investigatetheirbiologicalrole

Methods

SampleAnimalsandGenotyping

A total of 4311 bulls of five Italian dairy beef anddual purpose breedswere

genotyped with the Illumina BovineSNP50 BeadChip v1 (Illumina San Diego

CA)joiningthegenotypingeffortsoftwoItalianprojects(namelyldquoSelMolrdquoand

ldquoProzoordquo)Thedatasetincluded101replicatesand773siresonspairsusedfor

downstreamqualitycheckofdataproducedIndetailgenotypesof2954dairy

(2179 ItalianHolsteinand775 ItalianBrown)864beef (485Marchigianaand

379 Piedmontese) and 493 dual purpose (Italian Simmental) bulls were

availableDataqualitycontrol(QC)wasperformedintwostagesafirststageon

15

animals independently for each breed applying the same methods and

thresholdsandasecondstageappliedonmarkersacrossall theindividualsof

thedatasetThefirststageexcludedindividualswithunexpectedlyhigh(gt02)

Mendelian errors for fatherson couples andwith low call rates (lt95)The

secondstageexcludedi)SNPswithle25missingvaluesinthewholedataset

orcompletelymissing inonebreed ii)SNPswithminorallele frequencyle5

and iii) SNPs in sexual chromosomes and with unassigned chromosome or

physicalposition

EstimationofrEHH

HaplotypeswereobtainedusingfastPHASEsoftwareconsideringdefaultoptions

(ScheetandStephens2006)andwererunbreedwiseandchromosomewisein

eachbreedPedigreeinformationforallbullswereprovidedbytheirrespective

breederassociations andwereused to filteroutdirect relatives (in fatherson

pairsthesonwasmaintainedinthedatasetandthefatherdiscarded)andover

representedfamilies(amaximumof5randomlychosenindividualsperhalfsib

familywasallowed)Thefinaldatasetcontainingtheseldquolessrelatedrdquoanimalswill

behenceforthcalled ldquononredundantrdquodatasetThisnonredundantdatasetwas

used also to calculate within breed pairwise wholegenome Linkage

Disequilibrium(LD)Ther2statisticforallpairsofmarkerswasobtainedusing

PLINK software v107 (Purcell et al 2007)Thedecayof LDwas assessedby

averagingr2valuesupto1Mbdistance

To test if the population structure influenced the detection of rEHH we

replicated all analyses on the whole Italian Holstein dataset without any

population correction (ie ldquoredundantrdquo dataset) focusing on genes or gene

clustersthatarewellknowntobeunderrecentselectionincattle(ieldquocandidate

regionsrdquo) In particularwe focused on the casein polled gene cluster and coat

colourgenes(MC1RandKIT(Kemperetal2014Qanbarietal2010))

The EHH and rEHH calculations were performed using Sweep v11 software

(Sabetietal2002)Somedefaultprogramsettingshadtobemodifiedtoadapt

theseanalyses to thebovinegenomeas thesoftwarewasoriginallydeveloped

forhumangeneticanalysesSpecificallylocalrecombinationratesbetweenSNPs

16

wereapproximatedto1cMperMbEHHandrEHHcalculationswereperformed

breedwise and chromosomewise using an automatic core selection with

defaultoptions(ieconsideringlongestnonoverlappingcoresandlimitingcores

toatleast3andnomorethan20SNPs)asinQanbarietal(2010)AlthoughEHH

and rEHH values were obtained for all core haplotypes only those with

frequency ge 25 were retained for further analyses The (empirical) rEHH

significancethresholdwasobtainedbydividingtherEHHvaluesinto20binsof

5 frequency range each logtransforming withinbin values to achieve

normality Core haplotypes with pvalues lt 005 were considered significant

DetectionofcommonselectionsignatureswasobtainedcomparingrEHHvalues

acrossgenomicregionsacrossbreeds

Breedgroupingaccordingtoproductiontype

Toexamine thesignals common toeachof the twoproduction types (iedairy

andbeef) corehaplotypes thatsharedoneormoreSNP inat least twobreeds

from the same production type were selected The dual purpose Italian

Simmentalwas included inbothdairy (ItalianHolsteinand ItalianBrown)and

beef groups (Piedmontese and Marchigiana) since potentially it possesses

haplotypes selected for both production types All downstream analyses were

performedseparatelyforthedairyandthebeefproductiontypes

Detectionandannotationofcandidategenes

Genome annotation was performed using Biomart

(httpwwwbiomartorgindexhtml) a comprehensive source of gene

annotationformanyspeciesprovidedbyEnsembl(wwwensemblorg)Thelist

ofsharedhaplotypesregions(inbp)fordairyandbeefwasusedasinputfilein

theBiomartwebinterfaceThesetofgenesretrievedbyBiomartwasthenused

as input for theCanonicalpathwayandregulatorynetworkanalyses Ingenuity

PathwayAnalysis toolversion80(IPA IngenuityregSystems IncRedwoodCity

CA httpwwwingenuitycom) and a substantial examination of published

literaturewasusedtoexaminethefunctionalrelationshipsamongtheresulting

genes The IPA operates with a proprietary knowledge database providing

complementarypathwayanalysisforseveralspeciesincludingcattleIntheIPA

17

analyses the significance of each biological function identified was estimated

using Fisherrsquos exact test A Benjamini and Hochberg correction for multiple

testingwas calculated for function and canonical pathways For eachNetwork

IPAcomputedascore(pscore=log10(pvalue))fittingthesetstudiedgenesand

a list of biological functionspresent in the IPAknowledgedatabase The score

considered thenumberofgenes in thenetworkand thesizeof thenetwork to

estimatehowrelevantthisnetworkwasaccordingtothelistofgenesprovided

Thenthescorewasusedtorankthenetworks

Results

DatasetQC

Therepeatabilityobtainedfromthe101replicatespresentinthewholedataset

washigherthan998AfterthetwoQCstages105individualsand9730SNPs

were discarded Moreover after phasing 1292 individuals were removed to

reducethe largenumberofsibfamiliespresent intheredundantdatasetAfter

quality control procedures the final dataset had 44271 SNPs and 1132 514

393 410 and 364 individuals from Italian Holstein Italian Brown Italian

SimmentalMarchigianaandPiedmontesebullsrespectively(Table1)

18

Table1JNumberofanimalsgenotypedbeforeandafterQCTotalnumberofanimalgenotypedandrelativenumberofanimaldiscardedafterQCanalysis

Detectionofselectionsignatures

Sweepv11 softwaredetectedrEHHsignalswitha frequencyhigher than25overlapping the test candidate regions in both datasets corrected or not forpopulation structure In the redundant dataset we found only one significantrEHHsignal(caseincluster)whereasinthenonredundantdatasetweidentified5significantrEHHsignalsoverlappingallthetestcandidateregionsconsidered(caseinclusterpolledclusterMC1R)(Table2)

BreedTotal

Genotyped

ED15

misAN

ED1

REPL

ED1

MENDCleaned

HOL 2179 40 31 5 2093

BRW 775 6 16 4 749

SIM 493 6 6 2 479

MAR 485 37 38 410

PIE 379 5 10 364

19

Table2JComparisonofcandidateregionrEHHinnonJredundantandredundantdatasetrEHH overlapping candidate regions in both corrected (nonredundant) and not corrected

(redundant)datasetsforpopulation

Redundant NonRedundant

Candidate

geneBTA1

Closest

SNP(bp)CH2range CH2Frequency

rEHHJ

log(pval)CH2Frequency

rEHHJ

log(pval)

POLLED

Cluster1 1981154 18974181981154 H1079 093037 H1078 167174

MC1R 18 13657912 1331772014007505 H1069 120096 H1069 077167

KIT_BOVIN 6 72821175 7250492172821175 H1054 030030 H1054 033048

Casein

Cluster6 88427760 8835009888452835 H1047 094139 H1046 148133

Casein

Cluster6 88427760 8835009888452835 H2032 002003 H2033 002003

In total5526567847724388and4049corehaplotypeswitha frequency

higher than 25 were detected on Italian Holstein Italian Brown Italian

Simmental Marchigiana and Piedmontese breeds respectively A total of 838

866 740 692 and 613 core haplotypes were found significant outliers (see

Materials and Methods) in the aforementioned breeds Table 3 shows the

distribution of the total and significant core haplotypes per chromosome and

breed

20

Tableamp3amp(ampCoreamphaplotypesampdistributionampDistributionoftotalandsignificantcorehaplotypesperchromosomeandbreed

ampamp ItalianampampHolsteinamp

ItalianampampBrownamp

ItalianampSimmentalamp Marchigianaamp Piedmonteseamp

BTA1amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

1amp 389 66 389 68 335 49 310 51 288 46

2amp 312 47 314 52 267 44 257 42 238 33

3amp 292 48 313 62 264 46 236 34 227 34

4amp 268 46 279 50 245 41 237 39 229 39

5amp 223 32 231 29 200 30 183 27 171 22

6amp 291 51 307 40 265 45 252 38 233 42

7amp 249 37 253 46 214 39 195 33 204 27

8amp 268 39 270 38 235 35 223 32 209 31

9amp 227 37 228 41 210 37 165 22 187 30

10amp 219 36 234 34 192 25 196 26 168 25

11amp 248 39 246 34 209 30 205 33 187 25

12amp 162 24 176 18 148 21 148 25 135 22

13amp 205 32 200 19 155 18 178 35 135 19

14amp 175 17 189 25 168 29 143 22 129 19

15amp 180 31 185 29 148 23 139 13 130 19

16amp 176 28 192 29 160 25 149 19 127 15

17amp 170 27 183 29 168 31 135 22 123 16

18amp 133 18 153 19 119 14 108 10 82 12

19amp 137 22 130 23 118 16 99 18 92 15

20amp 184 35 172 33 145 23 124 23 124 17

21amp 145 23 147 30 113 18 110 21 94 13

22amp 140 24 145 19 115 21 112 18 104 13

23amp 106 13 100 16 67 10 64 9 57 8

24amp 129 11 140 20 124 17 97 23 101 16

25amp 91 12 98 11 68 10 77 12 70 6

26amp 125 7 114 14 93 12 96 15 90 16

27amp 91 12 91 15 75 11 84 12 60 10

28amp 88 9 96 10 70 10 66 9 55 11

29amp 103 15 103 13 82 10 72 9 67 12

TOTamp 5526 838 5678 866 4772 740 4388 692 4049 613

1BostaurusAutosomes2Detectedcorehaplotypeswithafrequencyinthebreedhigherthan253Significantcorehaplotypesatplt005

Comparisonamptoamppreviouslyampreportedampdataamp

AnumberofstudieshavesearchedselectionsweepsinHolstein(Qanbarietal

2010) Brown (Qanbari et al 2011) and Simmental (Fan et al 2014) breeds

Since different methods are expected to identify signatures having different

characteristicsthecomparisonwithpreviousstudiesislimitedtothosestudies

using thesamemethodand thesamebreed(s)analysed inourstudyOnlyone

study analysed (German) HolsteinRFriesian cattle with rEHH (Qanbari et al

2010)Qanbarietal(2010)reportedcandidategenesandgeneclustersunder

Alla$mia$famiglia$

A$mio$nonno$Mario$

INDEXamp

Chapteramp1ndashIntroduction 3

Chapteramp2Relativeextendedhaplotypehomozygosity(rEHH)signalsacrossbreedsrevealdairyandbeefspecificsignaturesofselection

11

Chapteramp3SearchingnewsignalsforproductivetraitsthroughgenebasedassociationanalysisinthreeItaliancattlebreeds

49

Chapteramp4ampMUGBASaspeciesfreegenebasedassociationsuite

81

Chapteramp5Imputationaccuracyisrobusttocattlereferencegenomeupdates

89

Chapteramp6QuestfordeleteriousrecessivevariantsinItalianHolsteinbullscombiningexomesequencingandhighdensitySNPanalysis

99

Chapteramp7ndashConclusion 167

amp

Acknowledgmentsampamp 173

CHAPTER1

Introduction

Background

The bovine species counts almost 15 billion of heads classified in some 800

breeds worldwide (httpfaostat3faoorg) The species is adapted to a wide

varietyofdifferentenvironmentandhusbandrysystemsandisoftenrearedfor

multiple purposes (Taberlet et al 2008) draft power milk meat leather

productionandotheruses(Elsiketal2009)

European cattle (Bos$ taurus) derive from a single domestication event of the

auroch (Bos$primigenius)occurred in theFertileCrescent around10000years

ago(Taberletetal2011)FollowingdomesticationcattlefollowedtheNeolithic

expansionofagricultureandcolonisedEurope(seeAjmoneMMarsanetal2010

for a review) Many thousand years of natural and artificial selection have

shaped cattle phenotype and genotype to meet environmental conditions and

human needs and together with demographic events and genetic drift have

started thedifferentiation of the bovine species in diverse and locally adapted

populations This differentiation process accelerated in the last two centuries

whenthebreedconceptwasinventedsothatnowthisspeciescountshundreds

of different breeds with high diversity in coat colour body size production

aptitudeadaptationtoclimate tolerancetodiseaseetc(Taberletetal2008

Elsiketal2009TheBovineHapMapConsortium2009)

Thisdiversityisnowbeingprogressivelylostwiththeextinctionofmanylocal

populationsmainlyforeconomicreasonsSelectioneffortshavethereforebeen

focussedonafewoutperformingindustrialbreedsthatarepresentlyselectedfor

specialiseddairyorbeefproductionandsometimesdoublepurpose

In these breeds breeding schemes are well developed and have been

traditionally based on the genetic evaluation of sires and dams using data

collectedfromperformanceandprogenytestingandunbiasedlinearprediction

modelsThesearefoundedontheFisherinfinitesimalmodellingofquantitative

trait variation and need no genomic information to estimate breeding values

(EBVs) The genetic progress recorded using this approach has been relevant

particularly for traits having medium to high heritability In Italian Holstein

genetic trends for themain production traits (milk protein and fat yield) are

positiveHoweveraccurateestimateofbreedingvalueshasahighcostinterms

4

of investment and time so that 5 to 6 years are needed to have a dairy bullprogenytested(wwwanafiit)Theadventofgenomicselection(Meuwissenetal2001)markedamilestoneindairycattlebreedingTheideawasdevelopedbeforethemoleculartoolswereinplaceforimplementitinpracticeandhasbecomearealityonlyrecentlyInthisapproach progeny tested bulls are genotyped at many thousand loci and thevalue of their EBVs used to estimate the value of each allele at each locusgenotyped These values are then used to predict the genetic value of younganimals on the basis of their genotype As a result animals from the newgeneration receive an EBV with the same accuracy of estimates obtained byprogenytestingbutmuchearlierThislowerthegenerationintervalandgreatlyincreasestherateofgeneticprogress indairypopulationsThe initial ideawasso good that nowadays genomic selection is widely applied in industrialcountriestothemajordairycattlebreedsWiththedecreaseofgenotypingcostit is likely that its use will be soon expanded to animals selected for otherpurposes (eg beef cattle) and other species (eg pigs) However genomicselection is an extension of traditional selection It is extremely useful inbreedingforcomplextraitsbutgivesnobiologicalinformationonthegenesthatcontrolthemIfthetraditionalselectionviewsasiregenomeasabigblackboxhaving a value but no clue why genomic selection views the same genomedisassembled inmany small boxes but still black and stillwithno clueon thereasonoftheirvalueIt is reasonable to think that a deeper understanding of the biologicalmechanisms controlling traits may further increase the accuracy of genomicevaluationsIndeedthenewtechnologiesoffernewopportunitiesinthissensebyloweringcostsandincreasingprecisionofdataproductionThis thesis is exploring the use of these technologies for retrieving biologicalinformationfromgenomicdataItusesmethodsnowconsideredldquotraditionalrdquoingenomics (GWAS and analysis of selection sweeps) and less traditional ones(gene centered GWAS and exome analysis) All data used here have beenproducedwithin twonationalprojectson farmanimalgenomics fundedby theItalianMinistryofAgriculture

5

M SELMOL ldquoRicerca e innovazione nelle attivitagrave di miglioramento genetico

animalemediante tecniche di geneticamolecolare per la competitivitagrave del

sistemazootecniconazionaleIrdquo(ResearchandInnovationinanimalbreeding

sheepgoatbuffalohorseanddonkeyI)

M INNOVAGENldquoRicercaeinnovazionenelleattivitagravedimiglioramentogenetico

animalemediante tecniche di geneticamolecolare per la competitivitagrave del

sistema zootecnico nazionale IIrdquo (Research and Innovation in animal

breedingsheepgoatbuffalohorseanddonkeyII)

In these projects the research group I work with M Istituto di Zootecnica at

Universitagrave Cattolica del Sacro Cuore di Piacenza M coordinated all the research

activitiesconcerningtheapplicationofgenomicstechniquesindairycattle

Contentsofthethesis

The thesis is structured into a short introduction (this first chapter) 5 core

chapters(Chapter2to6)andashortConclusion(Chapter7)Withtheexception

of Chapter 6 still in preparation the other core chapters containmanuscripts

that at time of writing are under review for their publication in international

peerMreviewedjournalsormanuscript

InChapter2selectionsignatureindairyandbeefcattleareinvestigatedwiththe

Illumina BovineSNP50 Genotyping BeadChip Five Italian cattle breeds (Italian

Holstein Italian Brown Italian Simmental Marchigiana and Piedmontese) are

analyzedwiththerEHHmethod(Sabetietal2002)tofindsignaturesofrecent

selectionCandidategenesinthevicinityofsignaturessharedbyeitherdairyor

beefbreeds are identified and theirpossible role indairy andbeefproduction

discussed

In Chapter 3 a GWAS analysis is run to investigate production traits in three

Italian dairy cattle breeds (Italian Holstein Italian Brown Italian Simmental)

Data are analysed with a well established method (Amin et al 2007)

6

implemented in theGenABELRpackage (Aulchenkoet al 2007) andwith the

newgeneMbasedmethodAssociationisinvestigatedbetween50KSNPsandmilk

production(milkyield)andqualitytraits(percentageofmilkproteinandfat)

InChapter4thegeneMbasedmethodusedinthepreviouschapter3inspiredby

theVEGASapproach(Liuetal2010)isdescribedThesoftwarewasdeveloped

andreMcodedinourlaboratoryinordertopermittheanalysisofanyspeciesin

additiontohumanMultiMtestingcorrectionandplotrepresentationcompletethe

toolsmadeavailable

InChapter5wetesttheimpactoftheuseofdifferentreferencesequencesinthe

imputation step Italian Simmental data are used to test variation in genomeM

wide inputation accuracy using different cattle reference sequences (BTAU42

UMD31 and BTAU46) To increase the analyses reliability four different

inputationsoftwaresaretested

InChapter6Holsteindataareanalysedcombiningexomesequencesand800K

SNPs data to seek deleterious recessive variantsWhile in natural populations

the destiny of these variants is to disappear in livestock breeds they can be

maintained and disseminated if carried by a highly productive sire The

intersection of deleterious mutations identified by sequencing and lack of

homozygousregionsidentifiedbytheanalysisofthepopulationwiththeHDSNP

chipidentifiedgenescandidatetocarryrecessivedeleteriousvariants

Objective

Themainobjectiveofthethesiswastoinvestigatethebovinegenomewiththe

aid new highMthroughput technology and to explore new strategies for linking

biology (genes and their function) to traits This linkage once validated may

increase the accuracy of genomic prediction and find application in selection

schemesegineradicatingthemostdeleteriousvariantsandorintheplanning

ofmatings

7

ReferencesAjmoneMMarsanPGarciaJFLenstraJA2010OntheoriginofcattleHow

aurochsbecamecattleandcolonizedtheworldEvolAnthropol19148ndash157doi101002evan20267

AminNvanDuijnCMAulchenkoYS2007AGenomicBackgroundBasedMethodforAssociationAnalysisinRelatedIndividualsPLoSONE2e1274doi101371journalpone0001274

AulchenkoYSKoningDMJdeHaleyC2007GenomewideRapidAssociationUsingMixedModelandRegressionAFastandSimpleMethodForGenomewidePedigreeMBasedQuantitativeTraitLociAssociationAnalysisGenetics177577ndash585doi101534genetics107075614

ElsikCGTellamRLWorleyKC2009TheGenomeSequenceofTaurineCattleAWindowtoRuminantBiologyandEvolutionScience324522ndash528doi101126science1169588

LiuJZMcraeAFNyholtDRMedlandSEWrayNRBrownKMHaywardNKMontgomeryGWVisscherPMMartinNGMacgregorS2010AVersatileGeneMBasedTestforGenomeMwideAssociationStudiesTheAmericanJournalofHumanGenetics87139ndash145doi101016jajhg201006009

MeuwissenTHHayesBJGoddardME2001PredictionoftotalgeneticvalueusinggenomeMwidedensemarkermapsGenetics1571819ndash1829

SabetiPCReichDEHigginsJMLevineHZPRichterDJSchaffnerSFGabrielSBPlatkoJVPattersonNJMcDonaldGJAckermanHCCampbellSJAltshulerDCooperRKwiatkowskiDWardRLanderES2002DetectingrecentpositiveselectioninthehumangenomefromhaplotypestructureNature419832ndash837doi101038nature01140

TaberletPCoissacEPansuJPompanonF2011ConservationgeneticsofcattlesheepandgoatsCRBiol334247ndash254doi101016jcrvi201012007

8

TaberletPValentiniARezaeiHRNaderiSPompanonFNegriniR

AjmoneMMarsanP2008Arecattlesheepandgoatsendangered

speciesMolEcol17275ndash284doi101111j1365M294X200703475x

TheBovineHapMapConsortium2009GenomeMWideSurveyofSNPVariation

UncoverstheGeneticStructureofCattleBreedsScience324528ndash532

doi101126science1167936

9

CHAPTER2

Relative(extendedamphaplotype(homozygosity)(rEHH)ampsignalsampacrossampbreedsamprevealampdairyampand$beef$specificampsignaturesampofampselection

LBomba1ELNicolazzi2MMilanesi1etal1IstitutodiZootecnicaUCSCviaEmiliaParmense8429122PiacenzaItaly2FondazioneParcoTecnologicoPadanoViaEinsteinLocCascinaCodazza26900LodiItalyTheseauthorscontributedequallytothiswork

____________________SubmittedtoGeneticSelectionEvolution

Abstract

Background

Nowadays many methods have been implemented to scan for signatures of

selection at genomescale level within and across breeds Extended haplotype

homozygosity is a proficient approach to detect genome regions under recent

selective pressure within breeds The objective of this study is to identify

common regions under positive recent selection shared by most represented

dairyandbeefItaliancattlebreeds

Results

A total of 3220 animals from Italian Holstein (2179) Italian Brown (775)

Simmental (493) Marchigiana (485) and Piedmontese (379) breeds were

genotyped with the Illumina BovineSNP50 BeadChip v1 in the frame of the

ItalianlivestockgenomicprojectsldquoSelMolrdquoandldquoProzoordquoAfterstandardcleaning

proceduresgenotypeswerephasedandcorehaplotypesidentifiedThedecayof

Linkage Disequilibrium (LD) for each core haplotype was assessedmeasuring

the extended haplotype homozygosity (EHH) To correct for the lack of good

estimatesof local recombinationrates therelativeEHH(rEHH)wascalculated

for each corehaplotypeGenomic regionsunder recentpositive selectionwere

identifiedasthosehighlyfrequentcorehaplotypeswithsignificantrEHHvalues

Significant independentcorehaplotypeswerealignedacrossbreeds to identify

signalsofrecentselectionsharedbydairyandsharedbybeefbreedsOverall82

and 87 common regions under selectionwere detected among dairy and beef

cattlebreedsrespectivelyBioinformaticsanalysisidentified244and232genes

mapping in these common genomic regions Pathway and network analysis of

significant genes revealed molecular functions biologically related to signal of

selectionsdetectedinmilkormeatproductiontypes

Conclusions

Resultssuggest that themultibreedapproach isamethodto identifyselection

signatures in cattle breeds under selection for the same production goal

12

allowing to better pinpoint the genomic regions of interest in dairy and beef

production

Background

The advent of genomic technologies and the consequent availability of single

nucleotidepolymorphism(SNP)datahaveenabledthestudyofselectionevents

on the cattle genome (Hayes et al 2008 Qanbari et al 2010) Such selection

signals arise from environmental or anthropogenic pressures and help to

understand the causes that led to breed formation These studies are usually

addressedinaldquotopdownrdquoapproach(ShahzadandLoor2012)fromgenotypeto

phenotypeinwhichgenomicdataisstatisticallyevaluatedtodetectdirectional

selectionThephenotypeneededtorunselectionsignaturesisintendedinavery

generalsenseabreedaproductionaptitudeorevenageographicareaoforigin

Thisapproachholdsthepotentialto investigatetraitsasadaptationtoextreme

climatesanddifferentfeedingandhusbandrysystemsresiliencetodiseasesetc

thatareveryexpensivedifficultandsometimesimpossibletostudywithclassic

GWAS (GenomeWide Association Study) approaches Therefore it is expected

thatgenomic informationretrievedinthesestudiesonlypartiallyoverlapsthat

identified by GWAS on specific traits Searching for these ldquosignatures of

selectionrdquo is of interest for understanding molecular mechanisms underlying

important biological processes and providing new insights to functional

biologicalinformation(egdiseaseandorimportanteconomictraits(Nielsenet

al 2007)) To date many methods have been implemented to scan these

selectionsignaturesatthegenomiclevelMethodsdifferwithrespecttoseveral

properties (Lenstra et al 2012) there are methods that perform analyses

withinbreedoracrossbreedsthatcomparealleleorhaplotypefrequenciesand

size that detect recent or ancient selection events either still segregating or

fixatedintopopulations(Lenstraetal2012OrsquoBrienetal2014Utsunomiyaet

al2013)Differentmethodshavedifferentsensitivityandrobustnesseg they

maybeinfluencedatadifferentextentbymarkerascertainmentbiasanduneven

distributionofrecombinationhotspotsalongthegenome

13

Extended haplotype homozygosity (EHH) is a method that has been used in

many animal species including cattle (Pan et al 2013 Qanbari et al 2010

Zhangetal2006)TheEHHmethodwasfirstdevelopedbySabetietal(2002)

for human population analysis The main objective of this methodology is to

identifylongrangehaplotypesinheritedwithoutrecombinationeventsMorein

detailunderaneutralevolutionmodelchangesinallelefrequenciesaredriven

onlybygeneticdriftInthisscenarioanewvariantwouldrequirealongtimeto

reach high frequency in the population and the Linkage Disequilibrium (LD)

surroundingitwoulddecayduetorecombinationevents(Kimura1984)Inthe

caseofpositiveselectionaquickriseinfrequencyofabeneficialmutationina

relatively short time will preserve the original haplotype structure as the

number of recombination events would be limited Therefore a signature of

positiveselectionisdefinedasaregionwithanalleleinlongrangeLDlocatedin

anuncommonlyhighfrequencyhaplotype

One of themost interesting features of thismethod is that it is able to detect

genomic regions that are under recent selection These recent selection

signatures can be therefore used to identify the genomic regions involved in

breedformationInadditionEHHmethoddoesnotrequireanydefinitionofan

ancestralallele(asneededintheintegratedhaplotypescoreiHS(Voightetal

2006) and is suited to be applied to SNP data because it is less sensitive to

ascertainmentbiasthanothermethods(Nielsenetal2007)HowevertheEHH

methodgeneratesa largenumberof falsepositiveandnegative resultsdue to

heterogeneous recombination rates along the genome (Qanbari et al 2010)

Moreover another drawback not specific of EHH but shared by all selection

signaturemethod is the lack of robust inferences able to discern true signals

fromthosegeneratedbychance(Kemperetal2014)

To partially account for these limitations Sabeti et al (2002) developed the

relative extended haplotype homozygosity (rEHH) including an empirical

methodologytoobtainsignificantsignalsTherEHHofaputativecorehaplotype

compares its original EHH valuewith that of other haplotypes at that specific

locus using all other haplotypes as control for local variation in the

recombination rate Therefore it identifies genomic regions carrying variants

14

under selection that are still segregating in the population analysed Although

thesemethodswere conceived for human populations theywere successfully

appliedto livestockspeciessuchaspig(Maetal2012)andcattle(Qanbariet

al2010)

After domestication about 10000 years ago in the fertile crescent taurine

cattle have colonized Europe and Africa and have been selected for different

human needs (Taberlet et al 2011) Particularly in the last centuries the

anthropic pressure led to the formation of hundreds of specialized breeds

adapted to different environmental conditions and linked to local tradition

constitutingagenepooldeservingattentionforconservation(AjmoneMarsanet

al2010)Someofthesebreedshavebeenselectedspecificallyfordairybeefor

bothproductiontypesfollowingstrongartificialselectionforthesetraits(Hayes

etal2009)Thepresentstudyaimsat identifyingsignalsof recentdirectional

selectionusingrEHHmethodindairyandbeefproductiontypesusingdatafrom

five Italian dairy beef and dual purpose cattle breeds We focused on the

significant core haplotypes scanned by rEHH that are shared among breeds of

thesameproductiontypeThenweusedabioinformaticsapproachto identify

positionalcandidategeneslyingwithinthegenomicregionsunderselectionand

investigatetheirbiologicalrole

Methods

SampleAnimalsandGenotyping

A total of 4311 bulls of five Italian dairy beef anddual purpose breedswere

genotyped with the Illumina BovineSNP50 BeadChip v1 (Illumina San Diego

CA)joiningthegenotypingeffortsoftwoItalianprojects(namelyldquoSelMolrdquoand

ldquoProzoordquo)Thedatasetincluded101replicatesand773siresonspairsusedfor

downstreamqualitycheckofdataproducedIndetailgenotypesof2954dairy

(2179 ItalianHolsteinand775 ItalianBrown)864beef (485Marchigianaand

379 Piedmontese) and 493 dual purpose (Italian Simmental) bulls were

availableDataqualitycontrol(QC)wasperformedintwostagesafirststageon

15

animals independently for each breed applying the same methods and

thresholdsandasecondstageappliedonmarkersacrossall theindividualsof

thedatasetThefirststageexcludedindividualswithunexpectedlyhigh(gt02)

Mendelian errors for fatherson couples andwith low call rates (lt95)The

secondstageexcludedi)SNPswithle25missingvaluesinthewholedataset

orcompletelymissing inonebreed ii)SNPswithminorallele frequencyle5

and iii) SNPs in sexual chromosomes and with unassigned chromosome or

physicalposition

EstimationofrEHH

HaplotypeswereobtainedusingfastPHASEsoftwareconsideringdefaultoptions

(ScheetandStephens2006)andwererunbreedwiseandchromosomewisein

eachbreedPedigreeinformationforallbullswereprovidedbytheirrespective

breederassociations andwereused to filteroutdirect relatives (in fatherson

pairsthesonwasmaintainedinthedatasetandthefatherdiscarded)andover

representedfamilies(amaximumof5randomlychosenindividualsperhalfsib

familywasallowed)Thefinaldatasetcontainingtheseldquolessrelatedrdquoanimalswill

behenceforthcalled ldquononredundantrdquodatasetThisnonredundantdatasetwas

used also to calculate within breed pairwise wholegenome Linkage

Disequilibrium(LD)Ther2statisticforallpairsofmarkerswasobtainedusing

PLINK software v107 (Purcell et al 2007)Thedecayof LDwas assessedby

averagingr2valuesupto1Mbdistance

To test if the population structure influenced the detection of rEHH we

replicated all analyses on the whole Italian Holstein dataset without any

population correction (ie ldquoredundantrdquo dataset) focusing on genes or gene

clustersthatarewellknowntobeunderrecentselectionincattle(ieldquocandidate

regionsrdquo) In particularwe focused on the casein polled gene cluster and coat

colourgenes(MC1RandKIT(Kemperetal2014Qanbarietal2010))

The EHH and rEHH calculations were performed using Sweep v11 software

(Sabetietal2002)Somedefaultprogramsettingshadtobemodifiedtoadapt

theseanalyses to thebovinegenomeas thesoftwarewasoriginallydeveloped

forhumangeneticanalysesSpecificallylocalrecombinationratesbetweenSNPs

16

wereapproximatedto1cMperMbEHHandrEHHcalculationswereperformed

breedwise and chromosomewise using an automatic core selection with

defaultoptions(ieconsideringlongestnonoverlappingcoresandlimitingcores

toatleast3andnomorethan20SNPs)asinQanbarietal(2010)AlthoughEHH

and rEHH values were obtained for all core haplotypes only those with

frequency ge 25 were retained for further analyses The (empirical) rEHH

significancethresholdwasobtainedbydividingtherEHHvaluesinto20binsof

5 frequency range each logtransforming withinbin values to achieve

normality Core haplotypes with pvalues lt 005 were considered significant

DetectionofcommonselectionsignatureswasobtainedcomparingrEHHvalues

acrossgenomicregionsacrossbreeds

Breedgroupingaccordingtoproductiontype

Toexamine thesignals common toeachof the twoproduction types (iedairy

andbeef) corehaplotypes thatsharedoneormoreSNP inat least twobreeds

from the same production type were selected The dual purpose Italian

Simmentalwas included inbothdairy (ItalianHolsteinand ItalianBrown)and

beef groups (Piedmontese and Marchigiana) since potentially it possesses

haplotypes selected for both production types All downstream analyses were

performedseparatelyforthedairyandthebeefproductiontypes

Detectionandannotationofcandidategenes

Genome annotation was performed using Biomart

(httpwwwbiomartorgindexhtml) a comprehensive source of gene

annotationformanyspeciesprovidedbyEnsembl(wwwensemblorg)Thelist

ofsharedhaplotypesregions(inbp)fordairyandbeefwasusedasinputfilein

theBiomartwebinterfaceThesetofgenesretrievedbyBiomartwasthenused

as input for theCanonicalpathwayandregulatorynetworkanalyses Ingenuity

PathwayAnalysis toolversion80(IPA IngenuityregSystems IncRedwoodCity

CA httpwwwingenuitycom) and a substantial examination of published

literaturewasusedtoexaminethefunctionalrelationshipsamongtheresulting

genes The IPA operates with a proprietary knowledge database providing

complementarypathwayanalysisforseveralspeciesincludingcattleIntheIPA

17

analyses the significance of each biological function identified was estimated

using Fisherrsquos exact test A Benjamini and Hochberg correction for multiple

testingwas calculated for function and canonical pathways For eachNetwork

IPAcomputedascore(pscore=log10(pvalue))fittingthesetstudiedgenesand

a list of biological functionspresent in the IPAknowledgedatabase The score

considered thenumberofgenes in thenetworkand thesizeof thenetwork to

estimatehowrelevantthisnetworkwasaccordingtothelistofgenesprovided

Thenthescorewasusedtorankthenetworks

Results

DatasetQC

Therepeatabilityobtainedfromthe101replicatespresentinthewholedataset

washigherthan998AfterthetwoQCstages105individualsand9730SNPs

were discarded Moreover after phasing 1292 individuals were removed to

reducethe largenumberofsibfamiliespresent intheredundantdatasetAfter

quality control procedures the final dataset had 44271 SNPs and 1132 514

393 410 and 364 individuals from Italian Holstein Italian Brown Italian

SimmentalMarchigianaandPiedmontesebullsrespectively(Table1)

18

Table1JNumberofanimalsgenotypedbeforeandafterQCTotalnumberofanimalgenotypedandrelativenumberofanimaldiscardedafterQCanalysis

Detectionofselectionsignatures

Sweepv11 softwaredetectedrEHHsignalswitha frequencyhigher than25overlapping the test candidate regions in both datasets corrected or not forpopulation structure In the redundant dataset we found only one significantrEHHsignal(caseincluster)whereasinthenonredundantdatasetweidentified5significantrEHHsignalsoverlappingallthetestcandidateregionsconsidered(caseinclusterpolledclusterMC1R)(Table2)

BreedTotal

Genotyped

ED15

misAN

ED1

REPL

ED1

MENDCleaned

HOL 2179 40 31 5 2093

BRW 775 6 16 4 749

SIM 493 6 6 2 479

MAR 485 37 38 410

PIE 379 5 10 364

19

Table2JComparisonofcandidateregionrEHHinnonJredundantandredundantdatasetrEHH overlapping candidate regions in both corrected (nonredundant) and not corrected

(redundant)datasetsforpopulation

Redundant NonRedundant

Candidate

geneBTA1

Closest

SNP(bp)CH2range CH2Frequency

rEHHJ

log(pval)CH2Frequency

rEHHJ

log(pval)

POLLED

Cluster1 1981154 18974181981154 H1079 093037 H1078 167174

MC1R 18 13657912 1331772014007505 H1069 120096 H1069 077167

KIT_BOVIN 6 72821175 7250492172821175 H1054 030030 H1054 033048

Casein

Cluster6 88427760 8835009888452835 H1047 094139 H1046 148133

Casein

Cluster6 88427760 8835009888452835 H2032 002003 H2033 002003

In total5526567847724388and4049corehaplotypeswitha frequency

higher than 25 were detected on Italian Holstein Italian Brown Italian

Simmental Marchigiana and Piedmontese breeds respectively A total of 838

866 740 692 and 613 core haplotypes were found significant outliers (see

Materials and Methods) in the aforementioned breeds Table 3 shows the

distribution of the total and significant core haplotypes per chromosome and

breed

20

Tableamp3amp(ampCoreamphaplotypesampdistributionampDistributionoftotalandsignificantcorehaplotypesperchromosomeandbreed

ampamp ItalianampampHolsteinamp

ItalianampampBrownamp

ItalianampSimmentalamp Marchigianaamp Piedmonteseamp

BTA1amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

1amp 389 66 389 68 335 49 310 51 288 46

2amp 312 47 314 52 267 44 257 42 238 33

3amp 292 48 313 62 264 46 236 34 227 34

4amp 268 46 279 50 245 41 237 39 229 39

5amp 223 32 231 29 200 30 183 27 171 22

6amp 291 51 307 40 265 45 252 38 233 42

7amp 249 37 253 46 214 39 195 33 204 27

8amp 268 39 270 38 235 35 223 32 209 31

9amp 227 37 228 41 210 37 165 22 187 30

10amp 219 36 234 34 192 25 196 26 168 25

11amp 248 39 246 34 209 30 205 33 187 25

12amp 162 24 176 18 148 21 148 25 135 22

13amp 205 32 200 19 155 18 178 35 135 19

14amp 175 17 189 25 168 29 143 22 129 19

15amp 180 31 185 29 148 23 139 13 130 19

16amp 176 28 192 29 160 25 149 19 127 15

17amp 170 27 183 29 168 31 135 22 123 16

18amp 133 18 153 19 119 14 108 10 82 12

19amp 137 22 130 23 118 16 99 18 92 15

20amp 184 35 172 33 145 23 124 23 124 17

21amp 145 23 147 30 113 18 110 21 94 13

22amp 140 24 145 19 115 21 112 18 104 13

23amp 106 13 100 16 67 10 64 9 57 8

24amp 129 11 140 20 124 17 97 23 101 16

25amp 91 12 98 11 68 10 77 12 70 6

26amp 125 7 114 14 93 12 96 15 90 16

27amp 91 12 91 15 75 11 84 12 60 10

28amp 88 9 96 10 70 10 66 9 55 11

29amp 103 15 103 13 82 10 72 9 67 12

TOTamp 5526 838 5678 866 4772 740 4388 692 4049 613

1BostaurusAutosomes2Detectedcorehaplotypeswithafrequencyinthebreedhigherthan253Significantcorehaplotypesatplt005

Comparisonamptoamppreviouslyampreportedampdataamp

AnumberofstudieshavesearchedselectionsweepsinHolstein(Qanbarietal

2010) Brown (Qanbari et al 2011) and Simmental (Fan et al 2014) breeds

Since different methods are expected to identify signatures having different

characteristicsthecomparisonwithpreviousstudiesislimitedtothosestudies

using thesamemethodand thesamebreed(s)analysed inourstudyOnlyone

study analysed (German) HolsteinRFriesian cattle with rEHH (Qanbari et al

2010)Qanbarietal(2010)reportedcandidategenesandgeneclustersunder

INDEXamp

Chapteramp1ndashIntroduction 3

Chapteramp2Relativeextendedhaplotypehomozygosity(rEHH)signalsacrossbreedsrevealdairyandbeefspecificsignaturesofselection

11

Chapteramp3SearchingnewsignalsforproductivetraitsthroughgenebasedassociationanalysisinthreeItaliancattlebreeds

49

Chapteramp4ampMUGBASaspeciesfreegenebasedassociationsuite

81

Chapteramp5Imputationaccuracyisrobusttocattlereferencegenomeupdates

89

Chapteramp6QuestfordeleteriousrecessivevariantsinItalianHolsteinbullscombiningexomesequencingandhighdensitySNPanalysis

99

Chapteramp7ndashConclusion 167

amp

Acknowledgmentsampamp 173

CHAPTER1

Introduction

Background

The bovine species counts almost 15 billion of heads classified in some 800

breeds worldwide (httpfaostat3faoorg) The species is adapted to a wide

varietyofdifferentenvironmentandhusbandrysystemsandisoftenrearedfor

multiple purposes (Taberlet et al 2008) draft power milk meat leather

productionandotheruses(Elsiketal2009)

European cattle (Bos$ taurus) derive from a single domestication event of the

auroch (Bos$primigenius)occurred in theFertileCrescent around10000years

ago(Taberletetal2011)FollowingdomesticationcattlefollowedtheNeolithic

expansionofagricultureandcolonisedEurope(seeAjmoneMMarsanetal2010

for a review) Many thousand years of natural and artificial selection have

shaped cattle phenotype and genotype to meet environmental conditions and

human needs and together with demographic events and genetic drift have

started thedifferentiation of the bovine species in diverse and locally adapted

populations This differentiation process accelerated in the last two centuries

whenthebreedconceptwasinventedsothatnowthisspeciescountshundreds

of different breeds with high diversity in coat colour body size production

aptitudeadaptationtoclimate tolerancetodiseaseetc(Taberletetal2008

Elsiketal2009TheBovineHapMapConsortium2009)

Thisdiversityisnowbeingprogressivelylostwiththeextinctionofmanylocal

populationsmainlyforeconomicreasonsSelectioneffortshavethereforebeen

focussedonafewoutperformingindustrialbreedsthatarepresentlyselectedfor

specialiseddairyorbeefproductionandsometimesdoublepurpose

In these breeds breeding schemes are well developed and have been

traditionally based on the genetic evaluation of sires and dams using data

collectedfromperformanceandprogenytestingandunbiasedlinearprediction

modelsThesearefoundedontheFisherinfinitesimalmodellingofquantitative

trait variation and need no genomic information to estimate breeding values

(EBVs) The genetic progress recorded using this approach has been relevant

particularly for traits having medium to high heritability In Italian Holstein

genetic trends for themain production traits (milk protein and fat yield) are

positiveHoweveraccurateestimateofbreedingvalueshasahighcostinterms

4

of investment and time so that 5 to 6 years are needed to have a dairy bullprogenytested(wwwanafiit)Theadventofgenomicselection(Meuwissenetal2001)markedamilestoneindairycattlebreedingTheideawasdevelopedbeforethemoleculartoolswereinplaceforimplementitinpracticeandhasbecomearealityonlyrecentlyInthisapproach progeny tested bulls are genotyped at many thousand loci and thevalue of their EBVs used to estimate the value of each allele at each locusgenotyped These values are then used to predict the genetic value of younganimals on the basis of their genotype As a result animals from the newgeneration receive an EBV with the same accuracy of estimates obtained byprogenytestingbutmuchearlierThislowerthegenerationintervalandgreatlyincreasestherateofgeneticprogress indairypopulationsThe initial ideawasso good that nowadays genomic selection is widely applied in industrialcountriestothemajordairycattlebreedsWiththedecreaseofgenotypingcostit is likely that its use will be soon expanded to animals selected for otherpurposes (eg beef cattle) and other species (eg pigs) However genomicselection is an extension of traditional selection It is extremely useful inbreedingforcomplextraitsbutgivesnobiologicalinformationonthegenesthatcontrolthemIfthetraditionalselectionviewsasiregenomeasabigblackboxhaving a value but no clue why genomic selection views the same genomedisassembled inmany small boxes but still black and stillwithno clueon thereasonoftheirvalueIt is reasonable to think that a deeper understanding of the biologicalmechanisms controlling traits may further increase the accuracy of genomicevaluationsIndeedthenewtechnologiesoffernewopportunitiesinthissensebyloweringcostsandincreasingprecisionofdataproductionThis thesis is exploring the use of these technologies for retrieving biologicalinformationfromgenomicdataItusesmethodsnowconsideredldquotraditionalrdquoingenomics (GWAS and analysis of selection sweeps) and less traditional ones(gene centered GWAS and exome analysis) All data used here have beenproducedwithin twonationalprojectson farmanimalgenomics fundedby theItalianMinistryofAgriculture

5

M SELMOL ldquoRicerca e innovazione nelle attivitagrave di miglioramento genetico

animalemediante tecniche di geneticamolecolare per la competitivitagrave del

sistemazootecniconazionaleIrdquo(ResearchandInnovationinanimalbreeding

sheepgoatbuffalohorseanddonkeyI)

M INNOVAGENldquoRicercaeinnovazionenelleattivitagravedimiglioramentogenetico

animalemediante tecniche di geneticamolecolare per la competitivitagrave del

sistema zootecnico nazionale IIrdquo (Research and Innovation in animal

breedingsheepgoatbuffalohorseanddonkeyII)

In these projects the research group I work with M Istituto di Zootecnica at

Universitagrave Cattolica del Sacro Cuore di Piacenza M coordinated all the research

activitiesconcerningtheapplicationofgenomicstechniquesindairycattle

Contentsofthethesis

The thesis is structured into a short introduction (this first chapter) 5 core

chapters(Chapter2to6)andashortConclusion(Chapter7)Withtheexception

of Chapter 6 still in preparation the other core chapters containmanuscripts

that at time of writing are under review for their publication in international

peerMreviewedjournalsormanuscript

InChapter2selectionsignatureindairyandbeefcattleareinvestigatedwiththe

Illumina BovineSNP50 Genotyping BeadChip Five Italian cattle breeds (Italian

Holstein Italian Brown Italian Simmental Marchigiana and Piedmontese) are

analyzedwiththerEHHmethod(Sabetietal2002)tofindsignaturesofrecent

selectionCandidategenesinthevicinityofsignaturessharedbyeitherdairyor

beefbreeds are identified and theirpossible role indairy andbeefproduction

discussed

In Chapter 3 a GWAS analysis is run to investigate production traits in three

Italian dairy cattle breeds (Italian Holstein Italian Brown Italian Simmental)

Data are analysed with a well established method (Amin et al 2007)

6

implemented in theGenABELRpackage (Aulchenkoet al 2007) andwith the

newgeneMbasedmethodAssociationisinvestigatedbetween50KSNPsandmilk

production(milkyield)andqualitytraits(percentageofmilkproteinandfat)

InChapter4thegeneMbasedmethodusedinthepreviouschapter3inspiredby

theVEGASapproach(Liuetal2010)isdescribedThesoftwarewasdeveloped

andreMcodedinourlaboratoryinordertopermittheanalysisofanyspeciesin

additiontohumanMultiMtestingcorrectionandplotrepresentationcompletethe

toolsmadeavailable

InChapter5wetesttheimpactoftheuseofdifferentreferencesequencesinthe

imputation step Italian Simmental data are used to test variation in genomeM

wide inputation accuracy using different cattle reference sequences (BTAU42

UMD31 and BTAU46) To increase the analyses reliability four different

inputationsoftwaresaretested

InChapter6Holsteindataareanalysedcombiningexomesequencesand800K

SNPs data to seek deleterious recessive variantsWhile in natural populations

the destiny of these variants is to disappear in livestock breeds they can be

maintained and disseminated if carried by a highly productive sire The

intersection of deleterious mutations identified by sequencing and lack of

homozygousregionsidentifiedbytheanalysisofthepopulationwiththeHDSNP

chipidentifiedgenescandidatetocarryrecessivedeleteriousvariants

Objective

Themainobjectiveofthethesiswastoinvestigatethebovinegenomewiththe

aid new highMthroughput technology and to explore new strategies for linking

biology (genes and their function) to traits This linkage once validated may

increase the accuracy of genomic prediction and find application in selection

schemesegineradicatingthemostdeleteriousvariantsandorintheplanning

ofmatings

7

ReferencesAjmoneMMarsanPGarciaJFLenstraJA2010OntheoriginofcattleHow

aurochsbecamecattleandcolonizedtheworldEvolAnthropol19148ndash157doi101002evan20267

AminNvanDuijnCMAulchenkoYS2007AGenomicBackgroundBasedMethodforAssociationAnalysisinRelatedIndividualsPLoSONE2e1274doi101371journalpone0001274

AulchenkoYSKoningDMJdeHaleyC2007GenomewideRapidAssociationUsingMixedModelandRegressionAFastandSimpleMethodForGenomewidePedigreeMBasedQuantitativeTraitLociAssociationAnalysisGenetics177577ndash585doi101534genetics107075614

ElsikCGTellamRLWorleyKC2009TheGenomeSequenceofTaurineCattleAWindowtoRuminantBiologyandEvolutionScience324522ndash528doi101126science1169588

LiuJZMcraeAFNyholtDRMedlandSEWrayNRBrownKMHaywardNKMontgomeryGWVisscherPMMartinNGMacgregorS2010AVersatileGeneMBasedTestforGenomeMwideAssociationStudiesTheAmericanJournalofHumanGenetics87139ndash145doi101016jajhg201006009

MeuwissenTHHayesBJGoddardME2001PredictionoftotalgeneticvalueusinggenomeMwidedensemarkermapsGenetics1571819ndash1829

SabetiPCReichDEHigginsJMLevineHZPRichterDJSchaffnerSFGabrielSBPlatkoJVPattersonNJMcDonaldGJAckermanHCCampbellSJAltshulerDCooperRKwiatkowskiDWardRLanderES2002DetectingrecentpositiveselectioninthehumangenomefromhaplotypestructureNature419832ndash837doi101038nature01140

TaberletPCoissacEPansuJPompanonF2011ConservationgeneticsofcattlesheepandgoatsCRBiol334247ndash254doi101016jcrvi201012007

8

TaberletPValentiniARezaeiHRNaderiSPompanonFNegriniR

AjmoneMMarsanP2008Arecattlesheepandgoatsendangered

speciesMolEcol17275ndash284doi101111j1365M294X200703475x

TheBovineHapMapConsortium2009GenomeMWideSurveyofSNPVariation

UncoverstheGeneticStructureofCattleBreedsScience324528ndash532

doi101126science1167936

9

CHAPTER2

Relative(extendedamphaplotype(homozygosity)(rEHH)ampsignalsampacrossampbreedsamprevealampdairyampand$beef$specificampsignaturesampofampselection

LBomba1ELNicolazzi2MMilanesi1etal1IstitutodiZootecnicaUCSCviaEmiliaParmense8429122PiacenzaItaly2FondazioneParcoTecnologicoPadanoViaEinsteinLocCascinaCodazza26900LodiItalyTheseauthorscontributedequallytothiswork

____________________SubmittedtoGeneticSelectionEvolution

Abstract

Background

Nowadays many methods have been implemented to scan for signatures of

selection at genomescale level within and across breeds Extended haplotype

homozygosity is a proficient approach to detect genome regions under recent

selective pressure within breeds The objective of this study is to identify

common regions under positive recent selection shared by most represented

dairyandbeefItaliancattlebreeds

Results

A total of 3220 animals from Italian Holstein (2179) Italian Brown (775)

Simmental (493) Marchigiana (485) and Piedmontese (379) breeds were

genotyped with the Illumina BovineSNP50 BeadChip v1 in the frame of the

ItalianlivestockgenomicprojectsldquoSelMolrdquoandldquoProzoordquoAfterstandardcleaning

proceduresgenotypeswerephasedandcorehaplotypesidentifiedThedecayof

Linkage Disequilibrium (LD) for each core haplotype was assessedmeasuring

the extended haplotype homozygosity (EHH) To correct for the lack of good

estimatesof local recombinationrates therelativeEHH(rEHH)wascalculated

for each corehaplotypeGenomic regionsunder recentpositive selectionwere

identifiedasthosehighlyfrequentcorehaplotypeswithsignificantrEHHvalues

Significant independentcorehaplotypeswerealignedacrossbreeds to identify

signalsofrecentselectionsharedbydairyandsharedbybeefbreedsOverall82

and 87 common regions under selectionwere detected among dairy and beef

cattlebreedsrespectivelyBioinformaticsanalysisidentified244and232genes

mapping in these common genomic regions Pathway and network analysis of

significant genes revealed molecular functions biologically related to signal of

selectionsdetectedinmilkormeatproductiontypes

Conclusions

Resultssuggest that themultibreedapproach isamethodto identifyselection

signatures in cattle breeds under selection for the same production goal

12

allowing to better pinpoint the genomic regions of interest in dairy and beef

production

Background

The advent of genomic technologies and the consequent availability of single

nucleotidepolymorphism(SNP)datahaveenabledthestudyofselectionevents

on the cattle genome (Hayes et al 2008 Qanbari et al 2010) Such selection

signals arise from environmental or anthropogenic pressures and help to

understand the causes that led to breed formation These studies are usually

addressedinaldquotopdownrdquoapproach(ShahzadandLoor2012)fromgenotypeto

phenotypeinwhichgenomicdataisstatisticallyevaluatedtodetectdirectional

selectionThephenotypeneededtorunselectionsignaturesisintendedinavery

generalsenseabreedaproductionaptitudeorevenageographicareaoforigin

Thisapproachholdsthepotentialto investigatetraitsasadaptationtoextreme

climatesanddifferentfeedingandhusbandrysystemsresiliencetodiseasesetc

thatareveryexpensivedifficultandsometimesimpossibletostudywithclassic

GWAS (GenomeWide Association Study) approaches Therefore it is expected

thatgenomic informationretrievedinthesestudiesonlypartiallyoverlapsthat

identified by GWAS on specific traits Searching for these ldquosignatures of

selectionrdquo is of interest for understanding molecular mechanisms underlying

important biological processes and providing new insights to functional

biologicalinformation(egdiseaseandorimportanteconomictraits(Nielsenet

al 2007)) To date many methods have been implemented to scan these

selectionsignaturesatthegenomiclevelMethodsdifferwithrespecttoseveral

properties (Lenstra et al 2012) there are methods that perform analyses

withinbreedoracrossbreedsthatcomparealleleorhaplotypefrequenciesand

size that detect recent or ancient selection events either still segregating or

fixatedintopopulations(Lenstraetal2012OrsquoBrienetal2014Utsunomiyaet

al2013)Differentmethodshavedifferentsensitivityandrobustnesseg they

maybeinfluencedatadifferentextentbymarkerascertainmentbiasanduneven

distributionofrecombinationhotspotsalongthegenome

13

Extended haplotype homozygosity (EHH) is a method that has been used in

many animal species including cattle (Pan et al 2013 Qanbari et al 2010

Zhangetal2006)TheEHHmethodwasfirstdevelopedbySabetietal(2002)

for human population analysis The main objective of this methodology is to

identifylongrangehaplotypesinheritedwithoutrecombinationeventsMorein

detailunderaneutralevolutionmodelchangesinallelefrequenciesaredriven

onlybygeneticdriftInthisscenarioanewvariantwouldrequirealongtimeto

reach high frequency in the population and the Linkage Disequilibrium (LD)

surroundingitwoulddecayduetorecombinationevents(Kimura1984)Inthe

caseofpositiveselectionaquickriseinfrequencyofabeneficialmutationina

relatively short time will preserve the original haplotype structure as the

number of recombination events would be limited Therefore a signature of

positiveselectionisdefinedasaregionwithanalleleinlongrangeLDlocatedin

anuncommonlyhighfrequencyhaplotype

One of themost interesting features of thismethod is that it is able to detect

genomic regions that are under recent selection These recent selection

signatures can be therefore used to identify the genomic regions involved in

breedformationInadditionEHHmethoddoesnotrequireanydefinitionofan

ancestralallele(asneededintheintegratedhaplotypescoreiHS(Voightetal

2006) and is suited to be applied to SNP data because it is less sensitive to

ascertainmentbiasthanothermethods(Nielsenetal2007)HowevertheEHH

methodgeneratesa largenumberof falsepositiveandnegative resultsdue to

heterogeneous recombination rates along the genome (Qanbari et al 2010)

Moreover another drawback not specific of EHH but shared by all selection

signaturemethod is the lack of robust inferences able to discern true signals

fromthosegeneratedbychance(Kemperetal2014)

To partially account for these limitations Sabeti et al (2002) developed the

relative extended haplotype homozygosity (rEHH) including an empirical

methodologytoobtainsignificantsignalsTherEHHofaputativecorehaplotype

compares its original EHH valuewith that of other haplotypes at that specific

locus using all other haplotypes as control for local variation in the

recombination rate Therefore it identifies genomic regions carrying variants

14

under selection that are still segregating in the population analysed Although

thesemethodswere conceived for human populations theywere successfully

appliedto livestockspeciessuchaspig(Maetal2012)andcattle(Qanbariet

al2010)

After domestication about 10000 years ago in the fertile crescent taurine

cattle have colonized Europe and Africa and have been selected for different

human needs (Taberlet et al 2011) Particularly in the last centuries the

anthropic pressure led to the formation of hundreds of specialized breeds

adapted to different environmental conditions and linked to local tradition

constitutingagenepooldeservingattentionforconservation(AjmoneMarsanet

al2010)Someofthesebreedshavebeenselectedspecificallyfordairybeefor

bothproductiontypesfollowingstrongartificialselectionforthesetraits(Hayes

etal2009)Thepresentstudyaimsat identifyingsignalsof recentdirectional

selectionusingrEHHmethodindairyandbeefproductiontypesusingdatafrom

five Italian dairy beef and dual purpose cattle breeds We focused on the

significant core haplotypes scanned by rEHH that are shared among breeds of

thesameproductiontypeThenweusedabioinformaticsapproachto identify

positionalcandidategeneslyingwithinthegenomicregionsunderselectionand

investigatetheirbiologicalrole

Methods

SampleAnimalsandGenotyping

A total of 4311 bulls of five Italian dairy beef anddual purpose breedswere

genotyped with the Illumina BovineSNP50 BeadChip v1 (Illumina San Diego

CA)joiningthegenotypingeffortsoftwoItalianprojects(namelyldquoSelMolrdquoand

ldquoProzoordquo)Thedatasetincluded101replicatesand773siresonspairsusedfor

downstreamqualitycheckofdataproducedIndetailgenotypesof2954dairy

(2179 ItalianHolsteinand775 ItalianBrown)864beef (485Marchigianaand

379 Piedmontese) and 493 dual purpose (Italian Simmental) bulls were

availableDataqualitycontrol(QC)wasperformedintwostagesafirststageon

15

animals independently for each breed applying the same methods and

thresholdsandasecondstageappliedonmarkersacrossall theindividualsof

thedatasetThefirststageexcludedindividualswithunexpectedlyhigh(gt02)

Mendelian errors for fatherson couples andwith low call rates (lt95)The

secondstageexcludedi)SNPswithle25missingvaluesinthewholedataset

orcompletelymissing inonebreed ii)SNPswithminorallele frequencyle5

and iii) SNPs in sexual chromosomes and with unassigned chromosome or

physicalposition

EstimationofrEHH

HaplotypeswereobtainedusingfastPHASEsoftwareconsideringdefaultoptions

(ScheetandStephens2006)andwererunbreedwiseandchromosomewisein

eachbreedPedigreeinformationforallbullswereprovidedbytheirrespective

breederassociations andwereused to filteroutdirect relatives (in fatherson

pairsthesonwasmaintainedinthedatasetandthefatherdiscarded)andover

representedfamilies(amaximumof5randomlychosenindividualsperhalfsib

familywasallowed)Thefinaldatasetcontainingtheseldquolessrelatedrdquoanimalswill

behenceforthcalled ldquononredundantrdquodatasetThisnonredundantdatasetwas

used also to calculate within breed pairwise wholegenome Linkage

Disequilibrium(LD)Ther2statisticforallpairsofmarkerswasobtainedusing

PLINK software v107 (Purcell et al 2007)Thedecayof LDwas assessedby

averagingr2valuesupto1Mbdistance

To test if the population structure influenced the detection of rEHH we

replicated all analyses on the whole Italian Holstein dataset without any

population correction (ie ldquoredundantrdquo dataset) focusing on genes or gene

clustersthatarewellknowntobeunderrecentselectionincattle(ieldquocandidate

regionsrdquo) In particularwe focused on the casein polled gene cluster and coat

colourgenes(MC1RandKIT(Kemperetal2014Qanbarietal2010))

The EHH and rEHH calculations were performed using Sweep v11 software

(Sabetietal2002)Somedefaultprogramsettingshadtobemodifiedtoadapt

theseanalyses to thebovinegenomeas thesoftwarewasoriginallydeveloped

forhumangeneticanalysesSpecificallylocalrecombinationratesbetweenSNPs

16

wereapproximatedto1cMperMbEHHandrEHHcalculationswereperformed

breedwise and chromosomewise using an automatic core selection with

defaultoptions(ieconsideringlongestnonoverlappingcoresandlimitingcores

toatleast3andnomorethan20SNPs)asinQanbarietal(2010)AlthoughEHH

and rEHH values were obtained for all core haplotypes only those with

frequency ge 25 were retained for further analyses The (empirical) rEHH

significancethresholdwasobtainedbydividingtherEHHvaluesinto20binsof

5 frequency range each logtransforming withinbin values to achieve

normality Core haplotypes with pvalues lt 005 were considered significant

DetectionofcommonselectionsignatureswasobtainedcomparingrEHHvalues

acrossgenomicregionsacrossbreeds

Breedgroupingaccordingtoproductiontype

Toexamine thesignals common toeachof the twoproduction types (iedairy

andbeef) corehaplotypes thatsharedoneormoreSNP inat least twobreeds

from the same production type were selected The dual purpose Italian

Simmentalwas included inbothdairy (ItalianHolsteinand ItalianBrown)and

beef groups (Piedmontese and Marchigiana) since potentially it possesses

haplotypes selected for both production types All downstream analyses were

performedseparatelyforthedairyandthebeefproductiontypes

Detectionandannotationofcandidategenes

Genome annotation was performed using Biomart

(httpwwwbiomartorgindexhtml) a comprehensive source of gene

annotationformanyspeciesprovidedbyEnsembl(wwwensemblorg)Thelist

ofsharedhaplotypesregions(inbp)fordairyandbeefwasusedasinputfilein

theBiomartwebinterfaceThesetofgenesretrievedbyBiomartwasthenused

as input for theCanonicalpathwayandregulatorynetworkanalyses Ingenuity

PathwayAnalysis toolversion80(IPA IngenuityregSystems IncRedwoodCity

CA httpwwwingenuitycom) and a substantial examination of published

literaturewasusedtoexaminethefunctionalrelationshipsamongtheresulting

genes The IPA operates with a proprietary knowledge database providing

complementarypathwayanalysisforseveralspeciesincludingcattleIntheIPA

17

analyses the significance of each biological function identified was estimated

using Fisherrsquos exact test A Benjamini and Hochberg correction for multiple

testingwas calculated for function and canonical pathways For eachNetwork

IPAcomputedascore(pscore=log10(pvalue))fittingthesetstudiedgenesand

a list of biological functionspresent in the IPAknowledgedatabase The score

considered thenumberofgenes in thenetworkand thesizeof thenetwork to

estimatehowrelevantthisnetworkwasaccordingtothelistofgenesprovided

Thenthescorewasusedtorankthenetworks

Results

DatasetQC

Therepeatabilityobtainedfromthe101replicatespresentinthewholedataset

washigherthan998AfterthetwoQCstages105individualsand9730SNPs

were discarded Moreover after phasing 1292 individuals were removed to

reducethe largenumberofsibfamiliespresent intheredundantdatasetAfter

quality control procedures the final dataset had 44271 SNPs and 1132 514

393 410 and 364 individuals from Italian Holstein Italian Brown Italian

SimmentalMarchigianaandPiedmontesebullsrespectively(Table1)

18

Table1JNumberofanimalsgenotypedbeforeandafterQCTotalnumberofanimalgenotypedandrelativenumberofanimaldiscardedafterQCanalysis

Detectionofselectionsignatures

Sweepv11 softwaredetectedrEHHsignalswitha frequencyhigher than25overlapping the test candidate regions in both datasets corrected or not forpopulation structure In the redundant dataset we found only one significantrEHHsignal(caseincluster)whereasinthenonredundantdatasetweidentified5significantrEHHsignalsoverlappingallthetestcandidateregionsconsidered(caseinclusterpolledclusterMC1R)(Table2)

BreedTotal

Genotyped

ED15

misAN

ED1

REPL

ED1

MENDCleaned

HOL 2179 40 31 5 2093

BRW 775 6 16 4 749

SIM 493 6 6 2 479

MAR 485 37 38 410

PIE 379 5 10 364

19

Table2JComparisonofcandidateregionrEHHinnonJredundantandredundantdatasetrEHH overlapping candidate regions in both corrected (nonredundant) and not corrected

(redundant)datasetsforpopulation

Redundant NonRedundant

Candidate

geneBTA1

Closest

SNP(bp)CH2range CH2Frequency

rEHHJ

log(pval)CH2Frequency

rEHHJ

log(pval)

POLLED

Cluster1 1981154 18974181981154 H1079 093037 H1078 167174

MC1R 18 13657912 1331772014007505 H1069 120096 H1069 077167

KIT_BOVIN 6 72821175 7250492172821175 H1054 030030 H1054 033048

Casein

Cluster6 88427760 8835009888452835 H1047 094139 H1046 148133

Casein

Cluster6 88427760 8835009888452835 H2032 002003 H2033 002003

In total5526567847724388and4049corehaplotypeswitha frequency

higher than 25 were detected on Italian Holstein Italian Brown Italian

Simmental Marchigiana and Piedmontese breeds respectively A total of 838

866 740 692 and 613 core haplotypes were found significant outliers (see

Materials and Methods) in the aforementioned breeds Table 3 shows the

distribution of the total and significant core haplotypes per chromosome and

breed

20

Tableamp3amp(ampCoreamphaplotypesampdistributionampDistributionoftotalandsignificantcorehaplotypesperchromosomeandbreed

ampamp ItalianampampHolsteinamp

ItalianampampBrownamp

ItalianampSimmentalamp Marchigianaamp Piedmonteseamp

BTA1amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

1amp 389 66 389 68 335 49 310 51 288 46

2amp 312 47 314 52 267 44 257 42 238 33

3amp 292 48 313 62 264 46 236 34 227 34

4amp 268 46 279 50 245 41 237 39 229 39

5amp 223 32 231 29 200 30 183 27 171 22

6amp 291 51 307 40 265 45 252 38 233 42

7amp 249 37 253 46 214 39 195 33 204 27

8amp 268 39 270 38 235 35 223 32 209 31

9amp 227 37 228 41 210 37 165 22 187 30

10amp 219 36 234 34 192 25 196 26 168 25

11amp 248 39 246 34 209 30 205 33 187 25

12amp 162 24 176 18 148 21 148 25 135 22

13amp 205 32 200 19 155 18 178 35 135 19

14amp 175 17 189 25 168 29 143 22 129 19

15amp 180 31 185 29 148 23 139 13 130 19

16amp 176 28 192 29 160 25 149 19 127 15

17amp 170 27 183 29 168 31 135 22 123 16

18amp 133 18 153 19 119 14 108 10 82 12

19amp 137 22 130 23 118 16 99 18 92 15

20amp 184 35 172 33 145 23 124 23 124 17

21amp 145 23 147 30 113 18 110 21 94 13

22amp 140 24 145 19 115 21 112 18 104 13

23amp 106 13 100 16 67 10 64 9 57 8

24amp 129 11 140 20 124 17 97 23 101 16

25amp 91 12 98 11 68 10 77 12 70 6

26amp 125 7 114 14 93 12 96 15 90 16

27amp 91 12 91 15 75 11 84 12 60 10

28amp 88 9 96 10 70 10 66 9 55 11

29amp 103 15 103 13 82 10 72 9 67 12

TOTamp 5526 838 5678 866 4772 740 4388 692 4049 613

1BostaurusAutosomes2Detectedcorehaplotypeswithafrequencyinthebreedhigherthan253Significantcorehaplotypesatplt005

Comparisonamptoamppreviouslyampreportedampdataamp

AnumberofstudieshavesearchedselectionsweepsinHolstein(Qanbarietal

2010) Brown (Qanbari et al 2011) and Simmental (Fan et al 2014) breeds

Since different methods are expected to identify signatures having different

characteristicsthecomparisonwithpreviousstudiesislimitedtothosestudies

using thesamemethodand thesamebreed(s)analysed inourstudyOnlyone

study analysed (German) HolsteinRFriesian cattle with rEHH (Qanbari et al

2010)Qanbarietal(2010)reportedcandidategenesandgeneclustersunder

CHAPTER1

Introduction

Background

The bovine species counts almost 15 billion of heads classified in some 800

breeds worldwide (httpfaostat3faoorg) The species is adapted to a wide

varietyofdifferentenvironmentandhusbandrysystemsandisoftenrearedfor

multiple purposes (Taberlet et al 2008) draft power milk meat leather

productionandotheruses(Elsiketal2009)

European cattle (Bos$ taurus) derive from a single domestication event of the

auroch (Bos$primigenius)occurred in theFertileCrescent around10000years

ago(Taberletetal2011)FollowingdomesticationcattlefollowedtheNeolithic

expansionofagricultureandcolonisedEurope(seeAjmoneMMarsanetal2010

for a review) Many thousand years of natural and artificial selection have

shaped cattle phenotype and genotype to meet environmental conditions and

human needs and together with demographic events and genetic drift have

started thedifferentiation of the bovine species in diverse and locally adapted

populations This differentiation process accelerated in the last two centuries

whenthebreedconceptwasinventedsothatnowthisspeciescountshundreds

of different breeds with high diversity in coat colour body size production

aptitudeadaptationtoclimate tolerancetodiseaseetc(Taberletetal2008

Elsiketal2009TheBovineHapMapConsortium2009)

Thisdiversityisnowbeingprogressivelylostwiththeextinctionofmanylocal

populationsmainlyforeconomicreasonsSelectioneffortshavethereforebeen

focussedonafewoutperformingindustrialbreedsthatarepresentlyselectedfor

specialiseddairyorbeefproductionandsometimesdoublepurpose

In these breeds breeding schemes are well developed and have been

traditionally based on the genetic evaluation of sires and dams using data

collectedfromperformanceandprogenytestingandunbiasedlinearprediction

modelsThesearefoundedontheFisherinfinitesimalmodellingofquantitative

trait variation and need no genomic information to estimate breeding values

(EBVs) The genetic progress recorded using this approach has been relevant

particularly for traits having medium to high heritability In Italian Holstein

genetic trends for themain production traits (milk protein and fat yield) are

positiveHoweveraccurateestimateofbreedingvalueshasahighcostinterms

4

of investment and time so that 5 to 6 years are needed to have a dairy bullprogenytested(wwwanafiit)Theadventofgenomicselection(Meuwissenetal2001)markedamilestoneindairycattlebreedingTheideawasdevelopedbeforethemoleculartoolswereinplaceforimplementitinpracticeandhasbecomearealityonlyrecentlyInthisapproach progeny tested bulls are genotyped at many thousand loci and thevalue of their EBVs used to estimate the value of each allele at each locusgenotyped These values are then used to predict the genetic value of younganimals on the basis of their genotype As a result animals from the newgeneration receive an EBV with the same accuracy of estimates obtained byprogenytestingbutmuchearlierThislowerthegenerationintervalandgreatlyincreasestherateofgeneticprogress indairypopulationsThe initial ideawasso good that nowadays genomic selection is widely applied in industrialcountriestothemajordairycattlebreedsWiththedecreaseofgenotypingcostit is likely that its use will be soon expanded to animals selected for otherpurposes (eg beef cattle) and other species (eg pigs) However genomicselection is an extension of traditional selection It is extremely useful inbreedingforcomplextraitsbutgivesnobiologicalinformationonthegenesthatcontrolthemIfthetraditionalselectionviewsasiregenomeasabigblackboxhaving a value but no clue why genomic selection views the same genomedisassembled inmany small boxes but still black and stillwithno clueon thereasonoftheirvalueIt is reasonable to think that a deeper understanding of the biologicalmechanisms controlling traits may further increase the accuracy of genomicevaluationsIndeedthenewtechnologiesoffernewopportunitiesinthissensebyloweringcostsandincreasingprecisionofdataproductionThis thesis is exploring the use of these technologies for retrieving biologicalinformationfromgenomicdataItusesmethodsnowconsideredldquotraditionalrdquoingenomics (GWAS and analysis of selection sweeps) and less traditional ones(gene centered GWAS and exome analysis) All data used here have beenproducedwithin twonationalprojectson farmanimalgenomics fundedby theItalianMinistryofAgriculture

5

M SELMOL ldquoRicerca e innovazione nelle attivitagrave di miglioramento genetico

animalemediante tecniche di geneticamolecolare per la competitivitagrave del

sistemazootecniconazionaleIrdquo(ResearchandInnovationinanimalbreeding

sheepgoatbuffalohorseanddonkeyI)

M INNOVAGENldquoRicercaeinnovazionenelleattivitagravedimiglioramentogenetico

animalemediante tecniche di geneticamolecolare per la competitivitagrave del

sistema zootecnico nazionale IIrdquo (Research and Innovation in animal

breedingsheepgoatbuffalohorseanddonkeyII)

In these projects the research group I work with M Istituto di Zootecnica at

Universitagrave Cattolica del Sacro Cuore di Piacenza M coordinated all the research

activitiesconcerningtheapplicationofgenomicstechniquesindairycattle

Contentsofthethesis

The thesis is structured into a short introduction (this first chapter) 5 core

chapters(Chapter2to6)andashortConclusion(Chapter7)Withtheexception

of Chapter 6 still in preparation the other core chapters containmanuscripts

that at time of writing are under review for their publication in international

peerMreviewedjournalsormanuscript

InChapter2selectionsignatureindairyandbeefcattleareinvestigatedwiththe

Illumina BovineSNP50 Genotyping BeadChip Five Italian cattle breeds (Italian

Holstein Italian Brown Italian Simmental Marchigiana and Piedmontese) are

analyzedwiththerEHHmethod(Sabetietal2002)tofindsignaturesofrecent

selectionCandidategenesinthevicinityofsignaturessharedbyeitherdairyor

beefbreeds are identified and theirpossible role indairy andbeefproduction

discussed

In Chapter 3 a GWAS analysis is run to investigate production traits in three

Italian dairy cattle breeds (Italian Holstein Italian Brown Italian Simmental)

Data are analysed with a well established method (Amin et al 2007)

6

implemented in theGenABELRpackage (Aulchenkoet al 2007) andwith the

newgeneMbasedmethodAssociationisinvestigatedbetween50KSNPsandmilk

production(milkyield)andqualitytraits(percentageofmilkproteinandfat)

InChapter4thegeneMbasedmethodusedinthepreviouschapter3inspiredby

theVEGASapproach(Liuetal2010)isdescribedThesoftwarewasdeveloped

andreMcodedinourlaboratoryinordertopermittheanalysisofanyspeciesin

additiontohumanMultiMtestingcorrectionandplotrepresentationcompletethe

toolsmadeavailable

InChapter5wetesttheimpactoftheuseofdifferentreferencesequencesinthe

imputation step Italian Simmental data are used to test variation in genomeM

wide inputation accuracy using different cattle reference sequences (BTAU42

UMD31 and BTAU46) To increase the analyses reliability four different

inputationsoftwaresaretested

InChapter6Holsteindataareanalysedcombiningexomesequencesand800K

SNPs data to seek deleterious recessive variantsWhile in natural populations

the destiny of these variants is to disappear in livestock breeds they can be

maintained and disseminated if carried by a highly productive sire The

intersection of deleterious mutations identified by sequencing and lack of

homozygousregionsidentifiedbytheanalysisofthepopulationwiththeHDSNP

chipidentifiedgenescandidatetocarryrecessivedeleteriousvariants

Objective

Themainobjectiveofthethesiswastoinvestigatethebovinegenomewiththe

aid new highMthroughput technology and to explore new strategies for linking

biology (genes and their function) to traits This linkage once validated may

increase the accuracy of genomic prediction and find application in selection

schemesegineradicatingthemostdeleteriousvariantsandorintheplanning

ofmatings

7

ReferencesAjmoneMMarsanPGarciaJFLenstraJA2010OntheoriginofcattleHow

aurochsbecamecattleandcolonizedtheworldEvolAnthropol19148ndash157doi101002evan20267

AminNvanDuijnCMAulchenkoYS2007AGenomicBackgroundBasedMethodforAssociationAnalysisinRelatedIndividualsPLoSONE2e1274doi101371journalpone0001274

AulchenkoYSKoningDMJdeHaleyC2007GenomewideRapidAssociationUsingMixedModelandRegressionAFastandSimpleMethodForGenomewidePedigreeMBasedQuantitativeTraitLociAssociationAnalysisGenetics177577ndash585doi101534genetics107075614

ElsikCGTellamRLWorleyKC2009TheGenomeSequenceofTaurineCattleAWindowtoRuminantBiologyandEvolutionScience324522ndash528doi101126science1169588

LiuJZMcraeAFNyholtDRMedlandSEWrayNRBrownKMHaywardNKMontgomeryGWVisscherPMMartinNGMacgregorS2010AVersatileGeneMBasedTestforGenomeMwideAssociationStudiesTheAmericanJournalofHumanGenetics87139ndash145doi101016jajhg201006009

MeuwissenTHHayesBJGoddardME2001PredictionoftotalgeneticvalueusinggenomeMwidedensemarkermapsGenetics1571819ndash1829

SabetiPCReichDEHigginsJMLevineHZPRichterDJSchaffnerSFGabrielSBPlatkoJVPattersonNJMcDonaldGJAckermanHCCampbellSJAltshulerDCooperRKwiatkowskiDWardRLanderES2002DetectingrecentpositiveselectioninthehumangenomefromhaplotypestructureNature419832ndash837doi101038nature01140

TaberletPCoissacEPansuJPompanonF2011ConservationgeneticsofcattlesheepandgoatsCRBiol334247ndash254doi101016jcrvi201012007

8

TaberletPValentiniARezaeiHRNaderiSPompanonFNegriniR

AjmoneMMarsanP2008Arecattlesheepandgoatsendangered

speciesMolEcol17275ndash284doi101111j1365M294X200703475x

TheBovineHapMapConsortium2009GenomeMWideSurveyofSNPVariation

UncoverstheGeneticStructureofCattleBreedsScience324528ndash532

doi101126science1167936

9

CHAPTER2

Relative(extendedamphaplotype(homozygosity)(rEHH)ampsignalsampacrossampbreedsamprevealampdairyampand$beef$specificampsignaturesampofampselection

LBomba1ELNicolazzi2MMilanesi1etal1IstitutodiZootecnicaUCSCviaEmiliaParmense8429122PiacenzaItaly2FondazioneParcoTecnologicoPadanoViaEinsteinLocCascinaCodazza26900LodiItalyTheseauthorscontributedequallytothiswork

____________________SubmittedtoGeneticSelectionEvolution

Abstract

Background

Nowadays many methods have been implemented to scan for signatures of

selection at genomescale level within and across breeds Extended haplotype

homozygosity is a proficient approach to detect genome regions under recent

selective pressure within breeds The objective of this study is to identify

common regions under positive recent selection shared by most represented

dairyandbeefItaliancattlebreeds

Results

A total of 3220 animals from Italian Holstein (2179) Italian Brown (775)

Simmental (493) Marchigiana (485) and Piedmontese (379) breeds were

genotyped with the Illumina BovineSNP50 BeadChip v1 in the frame of the

ItalianlivestockgenomicprojectsldquoSelMolrdquoandldquoProzoordquoAfterstandardcleaning

proceduresgenotypeswerephasedandcorehaplotypesidentifiedThedecayof

Linkage Disequilibrium (LD) for each core haplotype was assessedmeasuring

the extended haplotype homozygosity (EHH) To correct for the lack of good

estimatesof local recombinationrates therelativeEHH(rEHH)wascalculated

for each corehaplotypeGenomic regionsunder recentpositive selectionwere

identifiedasthosehighlyfrequentcorehaplotypeswithsignificantrEHHvalues

Significant independentcorehaplotypeswerealignedacrossbreeds to identify

signalsofrecentselectionsharedbydairyandsharedbybeefbreedsOverall82

and 87 common regions under selectionwere detected among dairy and beef

cattlebreedsrespectivelyBioinformaticsanalysisidentified244and232genes

mapping in these common genomic regions Pathway and network analysis of

significant genes revealed molecular functions biologically related to signal of

selectionsdetectedinmilkormeatproductiontypes

Conclusions

Resultssuggest that themultibreedapproach isamethodto identifyselection

signatures in cattle breeds under selection for the same production goal

12

allowing to better pinpoint the genomic regions of interest in dairy and beef

production

Background

The advent of genomic technologies and the consequent availability of single

nucleotidepolymorphism(SNP)datahaveenabledthestudyofselectionevents

on the cattle genome (Hayes et al 2008 Qanbari et al 2010) Such selection

signals arise from environmental or anthropogenic pressures and help to

understand the causes that led to breed formation These studies are usually

addressedinaldquotopdownrdquoapproach(ShahzadandLoor2012)fromgenotypeto

phenotypeinwhichgenomicdataisstatisticallyevaluatedtodetectdirectional

selectionThephenotypeneededtorunselectionsignaturesisintendedinavery

generalsenseabreedaproductionaptitudeorevenageographicareaoforigin

Thisapproachholdsthepotentialto investigatetraitsasadaptationtoextreme

climatesanddifferentfeedingandhusbandrysystemsresiliencetodiseasesetc

thatareveryexpensivedifficultandsometimesimpossibletostudywithclassic

GWAS (GenomeWide Association Study) approaches Therefore it is expected

thatgenomic informationretrievedinthesestudiesonlypartiallyoverlapsthat

identified by GWAS on specific traits Searching for these ldquosignatures of

selectionrdquo is of interest for understanding molecular mechanisms underlying

important biological processes and providing new insights to functional

biologicalinformation(egdiseaseandorimportanteconomictraits(Nielsenet

al 2007)) To date many methods have been implemented to scan these

selectionsignaturesatthegenomiclevelMethodsdifferwithrespecttoseveral

properties (Lenstra et al 2012) there are methods that perform analyses

withinbreedoracrossbreedsthatcomparealleleorhaplotypefrequenciesand

size that detect recent or ancient selection events either still segregating or

fixatedintopopulations(Lenstraetal2012OrsquoBrienetal2014Utsunomiyaet

al2013)Differentmethodshavedifferentsensitivityandrobustnesseg they

maybeinfluencedatadifferentextentbymarkerascertainmentbiasanduneven

distributionofrecombinationhotspotsalongthegenome

13

Extended haplotype homozygosity (EHH) is a method that has been used in

many animal species including cattle (Pan et al 2013 Qanbari et al 2010

Zhangetal2006)TheEHHmethodwasfirstdevelopedbySabetietal(2002)

for human population analysis The main objective of this methodology is to

identifylongrangehaplotypesinheritedwithoutrecombinationeventsMorein

detailunderaneutralevolutionmodelchangesinallelefrequenciesaredriven

onlybygeneticdriftInthisscenarioanewvariantwouldrequirealongtimeto

reach high frequency in the population and the Linkage Disequilibrium (LD)

surroundingitwoulddecayduetorecombinationevents(Kimura1984)Inthe

caseofpositiveselectionaquickriseinfrequencyofabeneficialmutationina

relatively short time will preserve the original haplotype structure as the

number of recombination events would be limited Therefore a signature of

positiveselectionisdefinedasaregionwithanalleleinlongrangeLDlocatedin

anuncommonlyhighfrequencyhaplotype

One of themost interesting features of thismethod is that it is able to detect

genomic regions that are under recent selection These recent selection

signatures can be therefore used to identify the genomic regions involved in

breedformationInadditionEHHmethoddoesnotrequireanydefinitionofan

ancestralallele(asneededintheintegratedhaplotypescoreiHS(Voightetal

2006) and is suited to be applied to SNP data because it is less sensitive to

ascertainmentbiasthanothermethods(Nielsenetal2007)HowevertheEHH

methodgeneratesa largenumberof falsepositiveandnegative resultsdue to

heterogeneous recombination rates along the genome (Qanbari et al 2010)

Moreover another drawback not specific of EHH but shared by all selection

signaturemethod is the lack of robust inferences able to discern true signals

fromthosegeneratedbychance(Kemperetal2014)

To partially account for these limitations Sabeti et al (2002) developed the

relative extended haplotype homozygosity (rEHH) including an empirical

methodologytoobtainsignificantsignalsTherEHHofaputativecorehaplotype

compares its original EHH valuewith that of other haplotypes at that specific

locus using all other haplotypes as control for local variation in the

recombination rate Therefore it identifies genomic regions carrying variants

14

under selection that are still segregating in the population analysed Although

thesemethodswere conceived for human populations theywere successfully

appliedto livestockspeciessuchaspig(Maetal2012)andcattle(Qanbariet

al2010)

After domestication about 10000 years ago in the fertile crescent taurine

cattle have colonized Europe and Africa and have been selected for different

human needs (Taberlet et al 2011) Particularly in the last centuries the

anthropic pressure led to the formation of hundreds of specialized breeds

adapted to different environmental conditions and linked to local tradition

constitutingagenepooldeservingattentionforconservation(AjmoneMarsanet

al2010)Someofthesebreedshavebeenselectedspecificallyfordairybeefor

bothproductiontypesfollowingstrongartificialselectionforthesetraits(Hayes

etal2009)Thepresentstudyaimsat identifyingsignalsof recentdirectional

selectionusingrEHHmethodindairyandbeefproductiontypesusingdatafrom

five Italian dairy beef and dual purpose cattle breeds We focused on the

significant core haplotypes scanned by rEHH that are shared among breeds of

thesameproductiontypeThenweusedabioinformaticsapproachto identify

positionalcandidategeneslyingwithinthegenomicregionsunderselectionand

investigatetheirbiologicalrole

Methods

SampleAnimalsandGenotyping

A total of 4311 bulls of five Italian dairy beef anddual purpose breedswere

genotyped with the Illumina BovineSNP50 BeadChip v1 (Illumina San Diego

CA)joiningthegenotypingeffortsoftwoItalianprojects(namelyldquoSelMolrdquoand

ldquoProzoordquo)Thedatasetincluded101replicatesand773siresonspairsusedfor

downstreamqualitycheckofdataproducedIndetailgenotypesof2954dairy

(2179 ItalianHolsteinand775 ItalianBrown)864beef (485Marchigianaand

379 Piedmontese) and 493 dual purpose (Italian Simmental) bulls were

availableDataqualitycontrol(QC)wasperformedintwostagesafirststageon

15

animals independently for each breed applying the same methods and

thresholdsandasecondstageappliedonmarkersacrossall theindividualsof

thedatasetThefirststageexcludedindividualswithunexpectedlyhigh(gt02)

Mendelian errors for fatherson couples andwith low call rates (lt95)The

secondstageexcludedi)SNPswithle25missingvaluesinthewholedataset

orcompletelymissing inonebreed ii)SNPswithminorallele frequencyle5

and iii) SNPs in sexual chromosomes and with unassigned chromosome or

physicalposition

EstimationofrEHH

HaplotypeswereobtainedusingfastPHASEsoftwareconsideringdefaultoptions

(ScheetandStephens2006)andwererunbreedwiseandchromosomewisein

eachbreedPedigreeinformationforallbullswereprovidedbytheirrespective

breederassociations andwereused to filteroutdirect relatives (in fatherson

pairsthesonwasmaintainedinthedatasetandthefatherdiscarded)andover

representedfamilies(amaximumof5randomlychosenindividualsperhalfsib

familywasallowed)Thefinaldatasetcontainingtheseldquolessrelatedrdquoanimalswill

behenceforthcalled ldquononredundantrdquodatasetThisnonredundantdatasetwas

used also to calculate within breed pairwise wholegenome Linkage

Disequilibrium(LD)Ther2statisticforallpairsofmarkerswasobtainedusing

PLINK software v107 (Purcell et al 2007)Thedecayof LDwas assessedby

averagingr2valuesupto1Mbdistance

To test if the population structure influenced the detection of rEHH we

replicated all analyses on the whole Italian Holstein dataset without any

population correction (ie ldquoredundantrdquo dataset) focusing on genes or gene

clustersthatarewellknowntobeunderrecentselectionincattle(ieldquocandidate

regionsrdquo) In particularwe focused on the casein polled gene cluster and coat

colourgenes(MC1RandKIT(Kemperetal2014Qanbarietal2010))

The EHH and rEHH calculations were performed using Sweep v11 software

(Sabetietal2002)Somedefaultprogramsettingshadtobemodifiedtoadapt

theseanalyses to thebovinegenomeas thesoftwarewasoriginallydeveloped

forhumangeneticanalysesSpecificallylocalrecombinationratesbetweenSNPs

16

wereapproximatedto1cMperMbEHHandrEHHcalculationswereperformed

breedwise and chromosomewise using an automatic core selection with

defaultoptions(ieconsideringlongestnonoverlappingcoresandlimitingcores

toatleast3andnomorethan20SNPs)asinQanbarietal(2010)AlthoughEHH

and rEHH values were obtained for all core haplotypes only those with

frequency ge 25 were retained for further analyses The (empirical) rEHH

significancethresholdwasobtainedbydividingtherEHHvaluesinto20binsof

5 frequency range each logtransforming withinbin values to achieve

normality Core haplotypes with pvalues lt 005 were considered significant

DetectionofcommonselectionsignatureswasobtainedcomparingrEHHvalues

acrossgenomicregionsacrossbreeds

Breedgroupingaccordingtoproductiontype

Toexamine thesignals common toeachof the twoproduction types (iedairy

andbeef) corehaplotypes thatsharedoneormoreSNP inat least twobreeds

from the same production type were selected The dual purpose Italian

Simmentalwas included inbothdairy (ItalianHolsteinand ItalianBrown)and

beef groups (Piedmontese and Marchigiana) since potentially it possesses

haplotypes selected for both production types All downstream analyses were

performedseparatelyforthedairyandthebeefproductiontypes

Detectionandannotationofcandidategenes

Genome annotation was performed using Biomart

(httpwwwbiomartorgindexhtml) a comprehensive source of gene

annotationformanyspeciesprovidedbyEnsembl(wwwensemblorg)Thelist

ofsharedhaplotypesregions(inbp)fordairyandbeefwasusedasinputfilein

theBiomartwebinterfaceThesetofgenesretrievedbyBiomartwasthenused

as input for theCanonicalpathwayandregulatorynetworkanalyses Ingenuity

PathwayAnalysis toolversion80(IPA IngenuityregSystems IncRedwoodCity

CA httpwwwingenuitycom) and a substantial examination of published

literaturewasusedtoexaminethefunctionalrelationshipsamongtheresulting

genes The IPA operates with a proprietary knowledge database providing

complementarypathwayanalysisforseveralspeciesincludingcattleIntheIPA

17

analyses the significance of each biological function identified was estimated

using Fisherrsquos exact test A Benjamini and Hochberg correction for multiple

testingwas calculated for function and canonical pathways For eachNetwork

IPAcomputedascore(pscore=log10(pvalue))fittingthesetstudiedgenesand

a list of biological functionspresent in the IPAknowledgedatabase The score

considered thenumberofgenes in thenetworkand thesizeof thenetwork to

estimatehowrelevantthisnetworkwasaccordingtothelistofgenesprovided

Thenthescorewasusedtorankthenetworks

Results

DatasetQC

Therepeatabilityobtainedfromthe101replicatespresentinthewholedataset

washigherthan998AfterthetwoQCstages105individualsand9730SNPs

were discarded Moreover after phasing 1292 individuals were removed to

reducethe largenumberofsibfamiliespresent intheredundantdatasetAfter

quality control procedures the final dataset had 44271 SNPs and 1132 514

393 410 and 364 individuals from Italian Holstein Italian Brown Italian

SimmentalMarchigianaandPiedmontesebullsrespectively(Table1)

18

Table1JNumberofanimalsgenotypedbeforeandafterQCTotalnumberofanimalgenotypedandrelativenumberofanimaldiscardedafterQCanalysis

Detectionofselectionsignatures

Sweepv11 softwaredetectedrEHHsignalswitha frequencyhigher than25overlapping the test candidate regions in both datasets corrected or not forpopulation structure In the redundant dataset we found only one significantrEHHsignal(caseincluster)whereasinthenonredundantdatasetweidentified5significantrEHHsignalsoverlappingallthetestcandidateregionsconsidered(caseinclusterpolledclusterMC1R)(Table2)

BreedTotal

Genotyped

ED15

misAN

ED1

REPL

ED1

MENDCleaned

HOL 2179 40 31 5 2093

BRW 775 6 16 4 749

SIM 493 6 6 2 479

MAR 485 37 38 410

PIE 379 5 10 364

19

Table2JComparisonofcandidateregionrEHHinnonJredundantandredundantdatasetrEHH overlapping candidate regions in both corrected (nonredundant) and not corrected

(redundant)datasetsforpopulation

Redundant NonRedundant

Candidate

geneBTA1

Closest

SNP(bp)CH2range CH2Frequency

rEHHJ

log(pval)CH2Frequency

rEHHJ

log(pval)

POLLED

Cluster1 1981154 18974181981154 H1079 093037 H1078 167174

MC1R 18 13657912 1331772014007505 H1069 120096 H1069 077167

KIT_BOVIN 6 72821175 7250492172821175 H1054 030030 H1054 033048

Casein

Cluster6 88427760 8835009888452835 H1047 094139 H1046 148133

Casein

Cluster6 88427760 8835009888452835 H2032 002003 H2033 002003

In total5526567847724388and4049corehaplotypeswitha frequency

higher than 25 were detected on Italian Holstein Italian Brown Italian

Simmental Marchigiana and Piedmontese breeds respectively A total of 838

866 740 692 and 613 core haplotypes were found significant outliers (see

Materials and Methods) in the aforementioned breeds Table 3 shows the

distribution of the total and significant core haplotypes per chromosome and

breed

20

Tableamp3amp(ampCoreamphaplotypesampdistributionampDistributionoftotalandsignificantcorehaplotypesperchromosomeandbreed

ampamp ItalianampampHolsteinamp

ItalianampampBrownamp

ItalianampSimmentalamp Marchigianaamp Piedmonteseamp

BTA1amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

1amp 389 66 389 68 335 49 310 51 288 46

2amp 312 47 314 52 267 44 257 42 238 33

3amp 292 48 313 62 264 46 236 34 227 34

4amp 268 46 279 50 245 41 237 39 229 39

5amp 223 32 231 29 200 30 183 27 171 22

6amp 291 51 307 40 265 45 252 38 233 42

7amp 249 37 253 46 214 39 195 33 204 27

8amp 268 39 270 38 235 35 223 32 209 31

9amp 227 37 228 41 210 37 165 22 187 30

10amp 219 36 234 34 192 25 196 26 168 25

11amp 248 39 246 34 209 30 205 33 187 25

12amp 162 24 176 18 148 21 148 25 135 22

13amp 205 32 200 19 155 18 178 35 135 19

14amp 175 17 189 25 168 29 143 22 129 19

15amp 180 31 185 29 148 23 139 13 130 19

16amp 176 28 192 29 160 25 149 19 127 15

17amp 170 27 183 29 168 31 135 22 123 16

18amp 133 18 153 19 119 14 108 10 82 12

19amp 137 22 130 23 118 16 99 18 92 15

20amp 184 35 172 33 145 23 124 23 124 17

21amp 145 23 147 30 113 18 110 21 94 13

22amp 140 24 145 19 115 21 112 18 104 13

23amp 106 13 100 16 67 10 64 9 57 8

24amp 129 11 140 20 124 17 97 23 101 16

25amp 91 12 98 11 68 10 77 12 70 6

26amp 125 7 114 14 93 12 96 15 90 16

27amp 91 12 91 15 75 11 84 12 60 10

28amp 88 9 96 10 70 10 66 9 55 11

29amp 103 15 103 13 82 10 72 9 67 12

TOTamp 5526 838 5678 866 4772 740 4388 692 4049 613

1BostaurusAutosomes2Detectedcorehaplotypeswithafrequencyinthebreedhigherthan253Significantcorehaplotypesatplt005

Comparisonamptoamppreviouslyampreportedampdataamp

AnumberofstudieshavesearchedselectionsweepsinHolstein(Qanbarietal

2010) Brown (Qanbari et al 2011) and Simmental (Fan et al 2014) breeds

Since different methods are expected to identify signatures having different

characteristicsthecomparisonwithpreviousstudiesislimitedtothosestudies

using thesamemethodand thesamebreed(s)analysed inourstudyOnlyone

study analysed (German) HolsteinRFriesian cattle with rEHH (Qanbari et al

2010)Qanbarietal(2010)reportedcandidategenesandgeneclustersunder

Background

The bovine species counts almost 15 billion of heads classified in some 800

breeds worldwide (httpfaostat3faoorg) The species is adapted to a wide

varietyofdifferentenvironmentandhusbandrysystemsandisoftenrearedfor

multiple purposes (Taberlet et al 2008) draft power milk meat leather

productionandotheruses(Elsiketal2009)

European cattle (Bos$ taurus) derive from a single domestication event of the

auroch (Bos$primigenius)occurred in theFertileCrescent around10000years

ago(Taberletetal2011)FollowingdomesticationcattlefollowedtheNeolithic

expansionofagricultureandcolonisedEurope(seeAjmoneMMarsanetal2010

for a review) Many thousand years of natural and artificial selection have

shaped cattle phenotype and genotype to meet environmental conditions and

human needs and together with demographic events and genetic drift have

started thedifferentiation of the bovine species in diverse and locally adapted

populations This differentiation process accelerated in the last two centuries

whenthebreedconceptwasinventedsothatnowthisspeciescountshundreds

of different breeds with high diversity in coat colour body size production

aptitudeadaptationtoclimate tolerancetodiseaseetc(Taberletetal2008

Elsiketal2009TheBovineHapMapConsortium2009)

Thisdiversityisnowbeingprogressivelylostwiththeextinctionofmanylocal

populationsmainlyforeconomicreasonsSelectioneffortshavethereforebeen

focussedonafewoutperformingindustrialbreedsthatarepresentlyselectedfor

specialiseddairyorbeefproductionandsometimesdoublepurpose

In these breeds breeding schemes are well developed and have been

traditionally based on the genetic evaluation of sires and dams using data

collectedfromperformanceandprogenytestingandunbiasedlinearprediction

modelsThesearefoundedontheFisherinfinitesimalmodellingofquantitative

trait variation and need no genomic information to estimate breeding values

(EBVs) The genetic progress recorded using this approach has been relevant

particularly for traits having medium to high heritability In Italian Holstein

genetic trends for themain production traits (milk protein and fat yield) are

positiveHoweveraccurateestimateofbreedingvalueshasahighcostinterms

4

of investment and time so that 5 to 6 years are needed to have a dairy bullprogenytested(wwwanafiit)Theadventofgenomicselection(Meuwissenetal2001)markedamilestoneindairycattlebreedingTheideawasdevelopedbeforethemoleculartoolswereinplaceforimplementitinpracticeandhasbecomearealityonlyrecentlyInthisapproach progeny tested bulls are genotyped at many thousand loci and thevalue of their EBVs used to estimate the value of each allele at each locusgenotyped These values are then used to predict the genetic value of younganimals on the basis of their genotype As a result animals from the newgeneration receive an EBV with the same accuracy of estimates obtained byprogenytestingbutmuchearlierThislowerthegenerationintervalandgreatlyincreasestherateofgeneticprogress indairypopulationsThe initial ideawasso good that nowadays genomic selection is widely applied in industrialcountriestothemajordairycattlebreedsWiththedecreaseofgenotypingcostit is likely that its use will be soon expanded to animals selected for otherpurposes (eg beef cattle) and other species (eg pigs) However genomicselection is an extension of traditional selection It is extremely useful inbreedingforcomplextraitsbutgivesnobiologicalinformationonthegenesthatcontrolthemIfthetraditionalselectionviewsasiregenomeasabigblackboxhaving a value but no clue why genomic selection views the same genomedisassembled inmany small boxes but still black and stillwithno clueon thereasonoftheirvalueIt is reasonable to think that a deeper understanding of the biologicalmechanisms controlling traits may further increase the accuracy of genomicevaluationsIndeedthenewtechnologiesoffernewopportunitiesinthissensebyloweringcostsandincreasingprecisionofdataproductionThis thesis is exploring the use of these technologies for retrieving biologicalinformationfromgenomicdataItusesmethodsnowconsideredldquotraditionalrdquoingenomics (GWAS and analysis of selection sweeps) and less traditional ones(gene centered GWAS and exome analysis) All data used here have beenproducedwithin twonationalprojectson farmanimalgenomics fundedby theItalianMinistryofAgriculture

5

M SELMOL ldquoRicerca e innovazione nelle attivitagrave di miglioramento genetico

animalemediante tecniche di geneticamolecolare per la competitivitagrave del

sistemazootecniconazionaleIrdquo(ResearchandInnovationinanimalbreeding

sheepgoatbuffalohorseanddonkeyI)

M INNOVAGENldquoRicercaeinnovazionenelleattivitagravedimiglioramentogenetico

animalemediante tecniche di geneticamolecolare per la competitivitagrave del

sistema zootecnico nazionale IIrdquo (Research and Innovation in animal

breedingsheepgoatbuffalohorseanddonkeyII)

In these projects the research group I work with M Istituto di Zootecnica at

Universitagrave Cattolica del Sacro Cuore di Piacenza M coordinated all the research

activitiesconcerningtheapplicationofgenomicstechniquesindairycattle

Contentsofthethesis

The thesis is structured into a short introduction (this first chapter) 5 core

chapters(Chapter2to6)andashortConclusion(Chapter7)Withtheexception

of Chapter 6 still in preparation the other core chapters containmanuscripts

that at time of writing are under review for their publication in international

peerMreviewedjournalsormanuscript

InChapter2selectionsignatureindairyandbeefcattleareinvestigatedwiththe

Illumina BovineSNP50 Genotyping BeadChip Five Italian cattle breeds (Italian

Holstein Italian Brown Italian Simmental Marchigiana and Piedmontese) are

analyzedwiththerEHHmethod(Sabetietal2002)tofindsignaturesofrecent

selectionCandidategenesinthevicinityofsignaturessharedbyeitherdairyor

beefbreeds are identified and theirpossible role indairy andbeefproduction

discussed

In Chapter 3 a GWAS analysis is run to investigate production traits in three

Italian dairy cattle breeds (Italian Holstein Italian Brown Italian Simmental)

Data are analysed with a well established method (Amin et al 2007)

6

implemented in theGenABELRpackage (Aulchenkoet al 2007) andwith the

newgeneMbasedmethodAssociationisinvestigatedbetween50KSNPsandmilk

production(milkyield)andqualitytraits(percentageofmilkproteinandfat)

InChapter4thegeneMbasedmethodusedinthepreviouschapter3inspiredby

theVEGASapproach(Liuetal2010)isdescribedThesoftwarewasdeveloped

andreMcodedinourlaboratoryinordertopermittheanalysisofanyspeciesin

additiontohumanMultiMtestingcorrectionandplotrepresentationcompletethe

toolsmadeavailable

InChapter5wetesttheimpactoftheuseofdifferentreferencesequencesinthe

imputation step Italian Simmental data are used to test variation in genomeM

wide inputation accuracy using different cattle reference sequences (BTAU42

UMD31 and BTAU46) To increase the analyses reliability four different

inputationsoftwaresaretested

InChapter6Holsteindataareanalysedcombiningexomesequencesand800K

SNPs data to seek deleterious recessive variantsWhile in natural populations

the destiny of these variants is to disappear in livestock breeds they can be

maintained and disseminated if carried by a highly productive sire The

intersection of deleterious mutations identified by sequencing and lack of

homozygousregionsidentifiedbytheanalysisofthepopulationwiththeHDSNP

chipidentifiedgenescandidatetocarryrecessivedeleteriousvariants

Objective

Themainobjectiveofthethesiswastoinvestigatethebovinegenomewiththe

aid new highMthroughput technology and to explore new strategies for linking

biology (genes and their function) to traits This linkage once validated may

increase the accuracy of genomic prediction and find application in selection

schemesegineradicatingthemostdeleteriousvariantsandorintheplanning

ofmatings

7

ReferencesAjmoneMMarsanPGarciaJFLenstraJA2010OntheoriginofcattleHow

aurochsbecamecattleandcolonizedtheworldEvolAnthropol19148ndash157doi101002evan20267

AminNvanDuijnCMAulchenkoYS2007AGenomicBackgroundBasedMethodforAssociationAnalysisinRelatedIndividualsPLoSONE2e1274doi101371journalpone0001274

AulchenkoYSKoningDMJdeHaleyC2007GenomewideRapidAssociationUsingMixedModelandRegressionAFastandSimpleMethodForGenomewidePedigreeMBasedQuantitativeTraitLociAssociationAnalysisGenetics177577ndash585doi101534genetics107075614

ElsikCGTellamRLWorleyKC2009TheGenomeSequenceofTaurineCattleAWindowtoRuminantBiologyandEvolutionScience324522ndash528doi101126science1169588

LiuJZMcraeAFNyholtDRMedlandSEWrayNRBrownKMHaywardNKMontgomeryGWVisscherPMMartinNGMacgregorS2010AVersatileGeneMBasedTestforGenomeMwideAssociationStudiesTheAmericanJournalofHumanGenetics87139ndash145doi101016jajhg201006009

MeuwissenTHHayesBJGoddardME2001PredictionoftotalgeneticvalueusinggenomeMwidedensemarkermapsGenetics1571819ndash1829

SabetiPCReichDEHigginsJMLevineHZPRichterDJSchaffnerSFGabrielSBPlatkoJVPattersonNJMcDonaldGJAckermanHCCampbellSJAltshulerDCooperRKwiatkowskiDWardRLanderES2002DetectingrecentpositiveselectioninthehumangenomefromhaplotypestructureNature419832ndash837doi101038nature01140

TaberletPCoissacEPansuJPompanonF2011ConservationgeneticsofcattlesheepandgoatsCRBiol334247ndash254doi101016jcrvi201012007

8

TaberletPValentiniARezaeiHRNaderiSPompanonFNegriniR

AjmoneMMarsanP2008Arecattlesheepandgoatsendangered

speciesMolEcol17275ndash284doi101111j1365M294X200703475x

TheBovineHapMapConsortium2009GenomeMWideSurveyofSNPVariation

UncoverstheGeneticStructureofCattleBreedsScience324528ndash532

doi101126science1167936

9

CHAPTER2

Relative(extendedamphaplotype(homozygosity)(rEHH)ampsignalsampacrossampbreedsamprevealampdairyampand$beef$specificampsignaturesampofampselection

LBomba1ELNicolazzi2MMilanesi1etal1IstitutodiZootecnicaUCSCviaEmiliaParmense8429122PiacenzaItaly2FondazioneParcoTecnologicoPadanoViaEinsteinLocCascinaCodazza26900LodiItalyTheseauthorscontributedequallytothiswork

____________________SubmittedtoGeneticSelectionEvolution

Abstract

Background

Nowadays many methods have been implemented to scan for signatures of

selection at genomescale level within and across breeds Extended haplotype

homozygosity is a proficient approach to detect genome regions under recent

selective pressure within breeds The objective of this study is to identify

common regions under positive recent selection shared by most represented

dairyandbeefItaliancattlebreeds

Results

A total of 3220 animals from Italian Holstein (2179) Italian Brown (775)

Simmental (493) Marchigiana (485) and Piedmontese (379) breeds were

genotyped with the Illumina BovineSNP50 BeadChip v1 in the frame of the

ItalianlivestockgenomicprojectsldquoSelMolrdquoandldquoProzoordquoAfterstandardcleaning

proceduresgenotypeswerephasedandcorehaplotypesidentifiedThedecayof

Linkage Disequilibrium (LD) for each core haplotype was assessedmeasuring

the extended haplotype homozygosity (EHH) To correct for the lack of good

estimatesof local recombinationrates therelativeEHH(rEHH)wascalculated

for each corehaplotypeGenomic regionsunder recentpositive selectionwere

identifiedasthosehighlyfrequentcorehaplotypeswithsignificantrEHHvalues

Significant independentcorehaplotypeswerealignedacrossbreeds to identify

signalsofrecentselectionsharedbydairyandsharedbybeefbreedsOverall82

and 87 common regions under selectionwere detected among dairy and beef

cattlebreedsrespectivelyBioinformaticsanalysisidentified244and232genes

mapping in these common genomic regions Pathway and network analysis of

significant genes revealed molecular functions biologically related to signal of

selectionsdetectedinmilkormeatproductiontypes

Conclusions

Resultssuggest that themultibreedapproach isamethodto identifyselection

signatures in cattle breeds under selection for the same production goal

12

allowing to better pinpoint the genomic regions of interest in dairy and beef

production

Background

The advent of genomic technologies and the consequent availability of single

nucleotidepolymorphism(SNP)datahaveenabledthestudyofselectionevents

on the cattle genome (Hayes et al 2008 Qanbari et al 2010) Such selection

signals arise from environmental or anthropogenic pressures and help to

understand the causes that led to breed formation These studies are usually

addressedinaldquotopdownrdquoapproach(ShahzadandLoor2012)fromgenotypeto

phenotypeinwhichgenomicdataisstatisticallyevaluatedtodetectdirectional

selectionThephenotypeneededtorunselectionsignaturesisintendedinavery

generalsenseabreedaproductionaptitudeorevenageographicareaoforigin

Thisapproachholdsthepotentialto investigatetraitsasadaptationtoextreme

climatesanddifferentfeedingandhusbandrysystemsresiliencetodiseasesetc

thatareveryexpensivedifficultandsometimesimpossibletostudywithclassic

GWAS (GenomeWide Association Study) approaches Therefore it is expected

thatgenomic informationretrievedinthesestudiesonlypartiallyoverlapsthat

identified by GWAS on specific traits Searching for these ldquosignatures of

selectionrdquo is of interest for understanding molecular mechanisms underlying

important biological processes and providing new insights to functional

biologicalinformation(egdiseaseandorimportanteconomictraits(Nielsenet

al 2007)) To date many methods have been implemented to scan these

selectionsignaturesatthegenomiclevelMethodsdifferwithrespecttoseveral

properties (Lenstra et al 2012) there are methods that perform analyses

withinbreedoracrossbreedsthatcomparealleleorhaplotypefrequenciesand

size that detect recent or ancient selection events either still segregating or

fixatedintopopulations(Lenstraetal2012OrsquoBrienetal2014Utsunomiyaet

al2013)Differentmethodshavedifferentsensitivityandrobustnesseg they

maybeinfluencedatadifferentextentbymarkerascertainmentbiasanduneven

distributionofrecombinationhotspotsalongthegenome

13

Extended haplotype homozygosity (EHH) is a method that has been used in

many animal species including cattle (Pan et al 2013 Qanbari et al 2010

Zhangetal2006)TheEHHmethodwasfirstdevelopedbySabetietal(2002)

for human population analysis The main objective of this methodology is to

identifylongrangehaplotypesinheritedwithoutrecombinationeventsMorein

detailunderaneutralevolutionmodelchangesinallelefrequenciesaredriven

onlybygeneticdriftInthisscenarioanewvariantwouldrequirealongtimeto

reach high frequency in the population and the Linkage Disequilibrium (LD)

surroundingitwoulddecayduetorecombinationevents(Kimura1984)Inthe

caseofpositiveselectionaquickriseinfrequencyofabeneficialmutationina

relatively short time will preserve the original haplotype structure as the

number of recombination events would be limited Therefore a signature of

positiveselectionisdefinedasaregionwithanalleleinlongrangeLDlocatedin

anuncommonlyhighfrequencyhaplotype

One of themost interesting features of thismethod is that it is able to detect

genomic regions that are under recent selection These recent selection

signatures can be therefore used to identify the genomic regions involved in

breedformationInadditionEHHmethoddoesnotrequireanydefinitionofan

ancestralallele(asneededintheintegratedhaplotypescoreiHS(Voightetal

2006) and is suited to be applied to SNP data because it is less sensitive to

ascertainmentbiasthanothermethods(Nielsenetal2007)HowevertheEHH

methodgeneratesa largenumberof falsepositiveandnegative resultsdue to

heterogeneous recombination rates along the genome (Qanbari et al 2010)

Moreover another drawback not specific of EHH but shared by all selection

signaturemethod is the lack of robust inferences able to discern true signals

fromthosegeneratedbychance(Kemperetal2014)

To partially account for these limitations Sabeti et al (2002) developed the

relative extended haplotype homozygosity (rEHH) including an empirical

methodologytoobtainsignificantsignalsTherEHHofaputativecorehaplotype

compares its original EHH valuewith that of other haplotypes at that specific

locus using all other haplotypes as control for local variation in the

recombination rate Therefore it identifies genomic regions carrying variants

14

under selection that are still segregating in the population analysed Although

thesemethodswere conceived for human populations theywere successfully

appliedto livestockspeciessuchaspig(Maetal2012)andcattle(Qanbariet

al2010)

After domestication about 10000 years ago in the fertile crescent taurine

cattle have colonized Europe and Africa and have been selected for different

human needs (Taberlet et al 2011) Particularly in the last centuries the

anthropic pressure led to the formation of hundreds of specialized breeds

adapted to different environmental conditions and linked to local tradition

constitutingagenepooldeservingattentionforconservation(AjmoneMarsanet

al2010)Someofthesebreedshavebeenselectedspecificallyfordairybeefor

bothproductiontypesfollowingstrongartificialselectionforthesetraits(Hayes

etal2009)Thepresentstudyaimsat identifyingsignalsof recentdirectional

selectionusingrEHHmethodindairyandbeefproductiontypesusingdatafrom

five Italian dairy beef and dual purpose cattle breeds We focused on the

significant core haplotypes scanned by rEHH that are shared among breeds of

thesameproductiontypeThenweusedabioinformaticsapproachto identify

positionalcandidategeneslyingwithinthegenomicregionsunderselectionand

investigatetheirbiologicalrole

Methods

SampleAnimalsandGenotyping

A total of 4311 bulls of five Italian dairy beef anddual purpose breedswere

genotyped with the Illumina BovineSNP50 BeadChip v1 (Illumina San Diego

CA)joiningthegenotypingeffortsoftwoItalianprojects(namelyldquoSelMolrdquoand

ldquoProzoordquo)Thedatasetincluded101replicatesand773siresonspairsusedfor

downstreamqualitycheckofdataproducedIndetailgenotypesof2954dairy

(2179 ItalianHolsteinand775 ItalianBrown)864beef (485Marchigianaand

379 Piedmontese) and 493 dual purpose (Italian Simmental) bulls were

availableDataqualitycontrol(QC)wasperformedintwostagesafirststageon

15

animals independently for each breed applying the same methods and

thresholdsandasecondstageappliedonmarkersacrossall theindividualsof

thedatasetThefirststageexcludedindividualswithunexpectedlyhigh(gt02)

Mendelian errors for fatherson couples andwith low call rates (lt95)The

secondstageexcludedi)SNPswithle25missingvaluesinthewholedataset

orcompletelymissing inonebreed ii)SNPswithminorallele frequencyle5

and iii) SNPs in sexual chromosomes and with unassigned chromosome or

physicalposition

EstimationofrEHH

HaplotypeswereobtainedusingfastPHASEsoftwareconsideringdefaultoptions

(ScheetandStephens2006)andwererunbreedwiseandchromosomewisein

eachbreedPedigreeinformationforallbullswereprovidedbytheirrespective

breederassociations andwereused to filteroutdirect relatives (in fatherson

pairsthesonwasmaintainedinthedatasetandthefatherdiscarded)andover

representedfamilies(amaximumof5randomlychosenindividualsperhalfsib

familywasallowed)Thefinaldatasetcontainingtheseldquolessrelatedrdquoanimalswill

behenceforthcalled ldquononredundantrdquodatasetThisnonredundantdatasetwas

used also to calculate within breed pairwise wholegenome Linkage

Disequilibrium(LD)Ther2statisticforallpairsofmarkerswasobtainedusing

PLINK software v107 (Purcell et al 2007)Thedecayof LDwas assessedby

averagingr2valuesupto1Mbdistance

To test if the population structure influenced the detection of rEHH we

replicated all analyses on the whole Italian Holstein dataset without any

population correction (ie ldquoredundantrdquo dataset) focusing on genes or gene

clustersthatarewellknowntobeunderrecentselectionincattle(ieldquocandidate

regionsrdquo) In particularwe focused on the casein polled gene cluster and coat

colourgenes(MC1RandKIT(Kemperetal2014Qanbarietal2010))

The EHH and rEHH calculations were performed using Sweep v11 software

(Sabetietal2002)Somedefaultprogramsettingshadtobemodifiedtoadapt

theseanalyses to thebovinegenomeas thesoftwarewasoriginallydeveloped

forhumangeneticanalysesSpecificallylocalrecombinationratesbetweenSNPs

16

wereapproximatedto1cMperMbEHHandrEHHcalculationswereperformed

breedwise and chromosomewise using an automatic core selection with

defaultoptions(ieconsideringlongestnonoverlappingcoresandlimitingcores

toatleast3andnomorethan20SNPs)asinQanbarietal(2010)AlthoughEHH

and rEHH values were obtained for all core haplotypes only those with

frequency ge 25 were retained for further analyses The (empirical) rEHH

significancethresholdwasobtainedbydividingtherEHHvaluesinto20binsof

5 frequency range each logtransforming withinbin values to achieve

normality Core haplotypes with pvalues lt 005 were considered significant

DetectionofcommonselectionsignatureswasobtainedcomparingrEHHvalues

acrossgenomicregionsacrossbreeds

Breedgroupingaccordingtoproductiontype

Toexamine thesignals common toeachof the twoproduction types (iedairy

andbeef) corehaplotypes thatsharedoneormoreSNP inat least twobreeds

from the same production type were selected The dual purpose Italian

Simmentalwas included inbothdairy (ItalianHolsteinand ItalianBrown)and

beef groups (Piedmontese and Marchigiana) since potentially it possesses

haplotypes selected for both production types All downstream analyses were

performedseparatelyforthedairyandthebeefproductiontypes

Detectionandannotationofcandidategenes

Genome annotation was performed using Biomart

(httpwwwbiomartorgindexhtml) a comprehensive source of gene

annotationformanyspeciesprovidedbyEnsembl(wwwensemblorg)Thelist

ofsharedhaplotypesregions(inbp)fordairyandbeefwasusedasinputfilein

theBiomartwebinterfaceThesetofgenesretrievedbyBiomartwasthenused

as input for theCanonicalpathwayandregulatorynetworkanalyses Ingenuity

PathwayAnalysis toolversion80(IPA IngenuityregSystems IncRedwoodCity

CA httpwwwingenuitycom) and a substantial examination of published

literaturewasusedtoexaminethefunctionalrelationshipsamongtheresulting

genes The IPA operates with a proprietary knowledge database providing

complementarypathwayanalysisforseveralspeciesincludingcattleIntheIPA

17

analyses the significance of each biological function identified was estimated

using Fisherrsquos exact test A Benjamini and Hochberg correction for multiple

testingwas calculated for function and canonical pathways For eachNetwork

IPAcomputedascore(pscore=log10(pvalue))fittingthesetstudiedgenesand

a list of biological functionspresent in the IPAknowledgedatabase The score

considered thenumberofgenes in thenetworkand thesizeof thenetwork to

estimatehowrelevantthisnetworkwasaccordingtothelistofgenesprovided

Thenthescorewasusedtorankthenetworks

Results

DatasetQC

Therepeatabilityobtainedfromthe101replicatespresentinthewholedataset

washigherthan998AfterthetwoQCstages105individualsand9730SNPs

were discarded Moreover after phasing 1292 individuals were removed to

reducethe largenumberofsibfamiliespresent intheredundantdatasetAfter

quality control procedures the final dataset had 44271 SNPs and 1132 514

393 410 and 364 individuals from Italian Holstein Italian Brown Italian

SimmentalMarchigianaandPiedmontesebullsrespectively(Table1)

18

Table1JNumberofanimalsgenotypedbeforeandafterQCTotalnumberofanimalgenotypedandrelativenumberofanimaldiscardedafterQCanalysis

Detectionofselectionsignatures

Sweepv11 softwaredetectedrEHHsignalswitha frequencyhigher than25overlapping the test candidate regions in both datasets corrected or not forpopulation structure In the redundant dataset we found only one significantrEHHsignal(caseincluster)whereasinthenonredundantdatasetweidentified5significantrEHHsignalsoverlappingallthetestcandidateregionsconsidered(caseinclusterpolledclusterMC1R)(Table2)

BreedTotal

Genotyped

ED15

misAN

ED1

REPL

ED1

MENDCleaned

HOL 2179 40 31 5 2093

BRW 775 6 16 4 749

SIM 493 6 6 2 479

MAR 485 37 38 410

PIE 379 5 10 364

19

Table2JComparisonofcandidateregionrEHHinnonJredundantandredundantdatasetrEHH overlapping candidate regions in both corrected (nonredundant) and not corrected

(redundant)datasetsforpopulation

Redundant NonRedundant

Candidate

geneBTA1

Closest

SNP(bp)CH2range CH2Frequency

rEHHJ

log(pval)CH2Frequency

rEHHJ

log(pval)

POLLED

Cluster1 1981154 18974181981154 H1079 093037 H1078 167174

MC1R 18 13657912 1331772014007505 H1069 120096 H1069 077167

KIT_BOVIN 6 72821175 7250492172821175 H1054 030030 H1054 033048

Casein

Cluster6 88427760 8835009888452835 H1047 094139 H1046 148133

Casein

Cluster6 88427760 8835009888452835 H2032 002003 H2033 002003

In total5526567847724388and4049corehaplotypeswitha frequency

higher than 25 were detected on Italian Holstein Italian Brown Italian

Simmental Marchigiana and Piedmontese breeds respectively A total of 838

866 740 692 and 613 core haplotypes were found significant outliers (see

Materials and Methods) in the aforementioned breeds Table 3 shows the

distribution of the total and significant core haplotypes per chromosome and

breed

20

Tableamp3amp(ampCoreamphaplotypesampdistributionampDistributionoftotalandsignificantcorehaplotypesperchromosomeandbreed

ampamp ItalianampampHolsteinamp

ItalianampampBrownamp

ItalianampSimmentalamp Marchigianaamp Piedmonteseamp

BTA1amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

1amp 389 66 389 68 335 49 310 51 288 46

2amp 312 47 314 52 267 44 257 42 238 33

3amp 292 48 313 62 264 46 236 34 227 34

4amp 268 46 279 50 245 41 237 39 229 39

5amp 223 32 231 29 200 30 183 27 171 22

6amp 291 51 307 40 265 45 252 38 233 42

7amp 249 37 253 46 214 39 195 33 204 27

8amp 268 39 270 38 235 35 223 32 209 31

9amp 227 37 228 41 210 37 165 22 187 30

10amp 219 36 234 34 192 25 196 26 168 25

11amp 248 39 246 34 209 30 205 33 187 25

12amp 162 24 176 18 148 21 148 25 135 22

13amp 205 32 200 19 155 18 178 35 135 19

14amp 175 17 189 25 168 29 143 22 129 19

15amp 180 31 185 29 148 23 139 13 130 19

16amp 176 28 192 29 160 25 149 19 127 15

17amp 170 27 183 29 168 31 135 22 123 16

18amp 133 18 153 19 119 14 108 10 82 12

19amp 137 22 130 23 118 16 99 18 92 15

20amp 184 35 172 33 145 23 124 23 124 17

21amp 145 23 147 30 113 18 110 21 94 13

22amp 140 24 145 19 115 21 112 18 104 13

23amp 106 13 100 16 67 10 64 9 57 8

24amp 129 11 140 20 124 17 97 23 101 16

25amp 91 12 98 11 68 10 77 12 70 6

26amp 125 7 114 14 93 12 96 15 90 16

27amp 91 12 91 15 75 11 84 12 60 10

28amp 88 9 96 10 70 10 66 9 55 11

29amp 103 15 103 13 82 10 72 9 67 12

TOTamp 5526 838 5678 866 4772 740 4388 692 4049 613

1BostaurusAutosomes2Detectedcorehaplotypeswithafrequencyinthebreedhigherthan253Significantcorehaplotypesatplt005

Comparisonamptoamppreviouslyampreportedampdataamp

AnumberofstudieshavesearchedselectionsweepsinHolstein(Qanbarietal

2010) Brown (Qanbari et al 2011) and Simmental (Fan et al 2014) breeds

Since different methods are expected to identify signatures having different

characteristicsthecomparisonwithpreviousstudiesislimitedtothosestudies

using thesamemethodand thesamebreed(s)analysed inourstudyOnlyone

study analysed (German) HolsteinRFriesian cattle with rEHH (Qanbari et al

2010)Qanbarietal(2010)reportedcandidategenesandgeneclustersunder

of investment and time so that 5 to 6 years are needed to have a dairy bullprogenytested(wwwanafiit)Theadventofgenomicselection(Meuwissenetal2001)markedamilestoneindairycattlebreedingTheideawasdevelopedbeforethemoleculartoolswereinplaceforimplementitinpracticeandhasbecomearealityonlyrecentlyInthisapproach progeny tested bulls are genotyped at many thousand loci and thevalue of their EBVs used to estimate the value of each allele at each locusgenotyped These values are then used to predict the genetic value of younganimals on the basis of their genotype As a result animals from the newgeneration receive an EBV with the same accuracy of estimates obtained byprogenytestingbutmuchearlierThislowerthegenerationintervalandgreatlyincreasestherateofgeneticprogress indairypopulationsThe initial ideawasso good that nowadays genomic selection is widely applied in industrialcountriestothemajordairycattlebreedsWiththedecreaseofgenotypingcostit is likely that its use will be soon expanded to animals selected for otherpurposes (eg beef cattle) and other species (eg pigs) However genomicselection is an extension of traditional selection It is extremely useful inbreedingforcomplextraitsbutgivesnobiologicalinformationonthegenesthatcontrolthemIfthetraditionalselectionviewsasiregenomeasabigblackboxhaving a value but no clue why genomic selection views the same genomedisassembled inmany small boxes but still black and stillwithno clueon thereasonoftheirvalueIt is reasonable to think that a deeper understanding of the biologicalmechanisms controlling traits may further increase the accuracy of genomicevaluationsIndeedthenewtechnologiesoffernewopportunitiesinthissensebyloweringcostsandincreasingprecisionofdataproductionThis thesis is exploring the use of these technologies for retrieving biologicalinformationfromgenomicdataItusesmethodsnowconsideredldquotraditionalrdquoingenomics (GWAS and analysis of selection sweeps) and less traditional ones(gene centered GWAS and exome analysis) All data used here have beenproducedwithin twonationalprojectson farmanimalgenomics fundedby theItalianMinistryofAgriculture

5

M SELMOL ldquoRicerca e innovazione nelle attivitagrave di miglioramento genetico

animalemediante tecniche di geneticamolecolare per la competitivitagrave del

sistemazootecniconazionaleIrdquo(ResearchandInnovationinanimalbreeding

sheepgoatbuffalohorseanddonkeyI)

M INNOVAGENldquoRicercaeinnovazionenelleattivitagravedimiglioramentogenetico

animalemediante tecniche di geneticamolecolare per la competitivitagrave del

sistema zootecnico nazionale IIrdquo (Research and Innovation in animal

breedingsheepgoatbuffalohorseanddonkeyII)

In these projects the research group I work with M Istituto di Zootecnica at

Universitagrave Cattolica del Sacro Cuore di Piacenza M coordinated all the research

activitiesconcerningtheapplicationofgenomicstechniquesindairycattle

Contentsofthethesis

The thesis is structured into a short introduction (this first chapter) 5 core

chapters(Chapter2to6)andashortConclusion(Chapter7)Withtheexception

of Chapter 6 still in preparation the other core chapters containmanuscripts

that at time of writing are under review for their publication in international

peerMreviewedjournalsormanuscript

InChapter2selectionsignatureindairyandbeefcattleareinvestigatedwiththe

Illumina BovineSNP50 Genotyping BeadChip Five Italian cattle breeds (Italian

Holstein Italian Brown Italian Simmental Marchigiana and Piedmontese) are

analyzedwiththerEHHmethod(Sabetietal2002)tofindsignaturesofrecent

selectionCandidategenesinthevicinityofsignaturessharedbyeitherdairyor

beefbreeds are identified and theirpossible role indairy andbeefproduction

discussed

In Chapter 3 a GWAS analysis is run to investigate production traits in three

Italian dairy cattle breeds (Italian Holstein Italian Brown Italian Simmental)

Data are analysed with a well established method (Amin et al 2007)

6

implemented in theGenABELRpackage (Aulchenkoet al 2007) andwith the

newgeneMbasedmethodAssociationisinvestigatedbetween50KSNPsandmilk

production(milkyield)andqualitytraits(percentageofmilkproteinandfat)

InChapter4thegeneMbasedmethodusedinthepreviouschapter3inspiredby

theVEGASapproach(Liuetal2010)isdescribedThesoftwarewasdeveloped

andreMcodedinourlaboratoryinordertopermittheanalysisofanyspeciesin

additiontohumanMultiMtestingcorrectionandplotrepresentationcompletethe

toolsmadeavailable

InChapter5wetesttheimpactoftheuseofdifferentreferencesequencesinthe

imputation step Italian Simmental data are used to test variation in genomeM

wide inputation accuracy using different cattle reference sequences (BTAU42

UMD31 and BTAU46) To increase the analyses reliability four different

inputationsoftwaresaretested

InChapter6Holsteindataareanalysedcombiningexomesequencesand800K

SNPs data to seek deleterious recessive variantsWhile in natural populations

the destiny of these variants is to disappear in livestock breeds they can be

maintained and disseminated if carried by a highly productive sire The

intersection of deleterious mutations identified by sequencing and lack of

homozygousregionsidentifiedbytheanalysisofthepopulationwiththeHDSNP

chipidentifiedgenescandidatetocarryrecessivedeleteriousvariants

Objective

Themainobjectiveofthethesiswastoinvestigatethebovinegenomewiththe

aid new highMthroughput technology and to explore new strategies for linking

biology (genes and their function) to traits This linkage once validated may

increase the accuracy of genomic prediction and find application in selection

schemesegineradicatingthemostdeleteriousvariantsandorintheplanning

ofmatings

7

ReferencesAjmoneMMarsanPGarciaJFLenstraJA2010OntheoriginofcattleHow

aurochsbecamecattleandcolonizedtheworldEvolAnthropol19148ndash157doi101002evan20267

AminNvanDuijnCMAulchenkoYS2007AGenomicBackgroundBasedMethodforAssociationAnalysisinRelatedIndividualsPLoSONE2e1274doi101371journalpone0001274

AulchenkoYSKoningDMJdeHaleyC2007GenomewideRapidAssociationUsingMixedModelandRegressionAFastandSimpleMethodForGenomewidePedigreeMBasedQuantitativeTraitLociAssociationAnalysisGenetics177577ndash585doi101534genetics107075614

ElsikCGTellamRLWorleyKC2009TheGenomeSequenceofTaurineCattleAWindowtoRuminantBiologyandEvolutionScience324522ndash528doi101126science1169588

LiuJZMcraeAFNyholtDRMedlandSEWrayNRBrownKMHaywardNKMontgomeryGWVisscherPMMartinNGMacgregorS2010AVersatileGeneMBasedTestforGenomeMwideAssociationStudiesTheAmericanJournalofHumanGenetics87139ndash145doi101016jajhg201006009

MeuwissenTHHayesBJGoddardME2001PredictionoftotalgeneticvalueusinggenomeMwidedensemarkermapsGenetics1571819ndash1829

SabetiPCReichDEHigginsJMLevineHZPRichterDJSchaffnerSFGabrielSBPlatkoJVPattersonNJMcDonaldGJAckermanHCCampbellSJAltshulerDCooperRKwiatkowskiDWardRLanderES2002DetectingrecentpositiveselectioninthehumangenomefromhaplotypestructureNature419832ndash837doi101038nature01140

TaberletPCoissacEPansuJPompanonF2011ConservationgeneticsofcattlesheepandgoatsCRBiol334247ndash254doi101016jcrvi201012007

8

TaberletPValentiniARezaeiHRNaderiSPompanonFNegriniR

AjmoneMMarsanP2008Arecattlesheepandgoatsendangered

speciesMolEcol17275ndash284doi101111j1365M294X200703475x

TheBovineHapMapConsortium2009GenomeMWideSurveyofSNPVariation

UncoverstheGeneticStructureofCattleBreedsScience324528ndash532

doi101126science1167936

9

CHAPTER2

Relative(extendedamphaplotype(homozygosity)(rEHH)ampsignalsampacrossampbreedsamprevealampdairyampand$beef$specificampsignaturesampofampselection

LBomba1ELNicolazzi2MMilanesi1etal1IstitutodiZootecnicaUCSCviaEmiliaParmense8429122PiacenzaItaly2FondazioneParcoTecnologicoPadanoViaEinsteinLocCascinaCodazza26900LodiItalyTheseauthorscontributedequallytothiswork

____________________SubmittedtoGeneticSelectionEvolution

Abstract

Background

Nowadays many methods have been implemented to scan for signatures of

selection at genomescale level within and across breeds Extended haplotype

homozygosity is a proficient approach to detect genome regions under recent

selective pressure within breeds The objective of this study is to identify

common regions under positive recent selection shared by most represented

dairyandbeefItaliancattlebreeds

Results

A total of 3220 animals from Italian Holstein (2179) Italian Brown (775)

Simmental (493) Marchigiana (485) and Piedmontese (379) breeds were

genotyped with the Illumina BovineSNP50 BeadChip v1 in the frame of the

ItalianlivestockgenomicprojectsldquoSelMolrdquoandldquoProzoordquoAfterstandardcleaning

proceduresgenotypeswerephasedandcorehaplotypesidentifiedThedecayof

Linkage Disequilibrium (LD) for each core haplotype was assessedmeasuring

the extended haplotype homozygosity (EHH) To correct for the lack of good

estimatesof local recombinationrates therelativeEHH(rEHH)wascalculated

for each corehaplotypeGenomic regionsunder recentpositive selectionwere

identifiedasthosehighlyfrequentcorehaplotypeswithsignificantrEHHvalues

Significant independentcorehaplotypeswerealignedacrossbreeds to identify

signalsofrecentselectionsharedbydairyandsharedbybeefbreedsOverall82

and 87 common regions under selectionwere detected among dairy and beef

cattlebreedsrespectivelyBioinformaticsanalysisidentified244and232genes

mapping in these common genomic regions Pathway and network analysis of

significant genes revealed molecular functions biologically related to signal of

selectionsdetectedinmilkormeatproductiontypes

Conclusions

Resultssuggest that themultibreedapproach isamethodto identifyselection

signatures in cattle breeds under selection for the same production goal

12

allowing to better pinpoint the genomic regions of interest in dairy and beef

production

Background

The advent of genomic technologies and the consequent availability of single

nucleotidepolymorphism(SNP)datahaveenabledthestudyofselectionevents

on the cattle genome (Hayes et al 2008 Qanbari et al 2010) Such selection

signals arise from environmental or anthropogenic pressures and help to

understand the causes that led to breed formation These studies are usually

addressedinaldquotopdownrdquoapproach(ShahzadandLoor2012)fromgenotypeto

phenotypeinwhichgenomicdataisstatisticallyevaluatedtodetectdirectional

selectionThephenotypeneededtorunselectionsignaturesisintendedinavery

generalsenseabreedaproductionaptitudeorevenageographicareaoforigin

Thisapproachholdsthepotentialto investigatetraitsasadaptationtoextreme

climatesanddifferentfeedingandhusbandrysystemsresiliencetodiseasesetc

thatareveryexpensivedifficultandsometimesimpossibletostudywithclassic

GWAS (GenomeWide Association Study) approaches Therefore it is expected

thatgenomic informationretrievedinthesestudiesonlypartiallyoverlapsthat

identified by GWAS on specific traits Searching for these ldquosignatures of

selectionrdquo is of interest for understanding molecular mechanisms underlying

important biological processes and providing new insights to functional

biologicalinformation(egdiseaseandorimportanteconomictraits(Nielsenet

al 2007)) To date many methods have been implemented to scan these

selectionsignaturesatthegenomiclevelMethodsdifferwithrespecttoseveral

properties (Lenstra et al 2012) there are methods that perform analyses

withinbreedoracrossbreedsthatcomparealleleorhaplotypefrequenciesand

size that detect recent or ancient selection events either still segregating or

fixatedintopopulations(Lenstraetal2012OrsquoBrienetal2014Utsunomiyaet

al2013)Differentmethodshavedifferentsensitivityandrobustnesseg they

maybeinfluencedatadifferentextentbymarkerascertainmentbiasanduneven

distributionofrecombinationhotspotsalongthegenome

13

Extended haplotype homozygosity (EHH) is a method that has been used in

many animal species including cattle (Pan et al 2013 Qanbari et al 2010

Zhangetal2006)TheEHHmethodwasfirstdevelopedbySabetietal(2002)

for human population analysis The main objective of this methodology is to

identifylongrangehaplotypesinheritedwithoutrecombinationeventsMorein

detailunderaneutralevolutionmodelchangesinallelefrequenciesaredriven

onlybygeneticdriftInthisscenarioanewvariantwouldrequirealongtimeto

reach high frequency in the population and the Linkage Disequilibrium (LD)

surroundingitwoulddecayduetorecombinationevents(Kimura1984)Inthe

caseofpositiveselectionaquickriseinfrequencyofabeneficialmutationina

relatively short time will preserve the original haplotype structure as the

number of recombination events would be limited Therefore a signature of

positiveselectionisdefinedasaregionwithanalleleinlongrangeLDlocatedin

anuncommonlyhighfrequencyhaplotype

One of themost interesting features of thismethod is that it is able to detect

genomic regions that are under recent selection These recent selection

signatures can be therefore used to identify the genomic regions involved in

breedformationInadditionEHHmethoddoesnotrequireanydefinitionofan

ancestralallele(asneededintheintegratedhaplotypescoreiHS(Voightetal

2006) and is suited to be applied to SNP data because it is less sensitive to

ascertainmentbiasthanothermethods(Nielsenetal2007)HowevertheEHH

methodgeneratesa largenumberof falsepositiveandnegative resultsdue to

heterogeneous recombination rates along the genome (Qanbari et al 2010)

Moreover another drawback not specific of EHH but shared by all selection

signaturemethod is the lack of robust inferences able to discern true signals

fromthosegeneratedbychance(Kemperetal2014)

To partially account for these limitations Sabeti et al (2002) developed the

relative extended haplotype homozygosity (rEHH) including an empirical

methodologytoobtainsignificantsignalsTherEHHofaputativecorehaplotype

compares its original EHH valuewith that of other haplotypes at that specific

locus using all other haplotypes as control for local variation in the

recombination rate Therefore it identifies genomic regions carrying variants

14

under selection that are still segregating in the population analysed Although

thesemethodswere conceived for human populations theywere successfully

appliedto livestockspeciessuchaspig(Maetal2012)andcattle(Qanbariet

al2010)

After domestication about 10000 years ago in the fertile crescent taurine

cattle have colonized Europe and Africa and have been selected for different

human needs (Taberlet et al 2011) Particularly in the last centuries the

anthropic pressure led to the formation of hundreds of specialized breeds

adapted to different environmental conditions and linked to local tradition

constitutingagenepooldeservingattentionforconservation(AjmoneMarsanet

al2010)Someofthesebreedshavebeenselectedspecificallyfordairybeefor

bothproductiontypesfollowingstrongartificialselectionforthesetraits(Hayes

etal2009)Thepresentstudyaimsat identifyingsignalsof recentdirectional

selectionusingrEHHmethodindairyandbeefproductiontypesusingdatafrom

five Italian dairy beef and dual purpose cattle breeds We focused on the

significant core haplotypes scanned by rEHH that are shared among breeds of

thesameproductiontypeThenweusedabioinformaticsapproachto identify

positionalcandidategeneslyingwithinthegenomicregionsunderselectionand

investigatetheirbiologicalrole

Methods

SampleAnimalsandGenotyping

A total of 4311 bulls of five Italian dairy beef anddual purpose breedswere

genotyped with the Illumina BovineSNP50 BeadChip v1 (Illumina San Diego

CA)joiningthegenotypingeffortsoftwoItalianprojects(namelyldquoSelMolrdquoand

ldquoProzoordquo)Thedatasetincluded101replicatesand773siresonspairsusedfor

downstreamqualitycheckofdataproducedIndetailgenotypesof2954dairy

(2179 ItalianHolsteinand775 ItalianBrown)864beef (485Marchigianaand

379 Piedmontese) and 493 dual purpose (Italian Simmental) bulls were

availableDataqualitycontrol(QC)wasperformedintwostagesafirststageon

15

animals independently for each breed applying the same methods and

thresholdsandasecondstageappliedonmarkersacrossall theindividualsof

thedatasetThefirststageexcludedindividualswithunexpectedlyhigh(gt02)

Mendelian errors for fatherson couples andwith low call rates (lt95)The

secondstageexcludedi)SNPswithle25missingvaluesinthewholedataset

orcompletelymissing inonebreed ii)SNPswithminorallele frequencyle5

and iii) SNPs in sexual chromosomes and with unassigned chromosome or

physicalposition

EstimationofrEHH

HaplotypeswereobtainedusingfastPHASEsoftwareconsideringdefaultoptions

(ScheetandStephens2006)andwererunbreedwiseandchromosomewisein

eachbreedPedigreeinformationforallbullswereprovidedbytheirrespective

breederassociations andwereused to filteroutdirect relatives (in fatherson

pairsthesonwasmaintainedinthedatasetandthefatherdiscarded)andover

representedfamilies(amaximumof5randomlychosenindividualsperhalfsib

familywasallowed)Thefinaldatasetcontainingtheseldquolessrelatedrdquoanimalswill

behenceforthcalled ldquononredundantrdquodatasetThisnonredundantdatasetwas

used also to calculate within breed pairwise wholegenome Linkage

Disequilibrium(LD)Ther2statisticforallpairsofmarkerswasobtainedusing

PLINK software v107 (Purcell et al 2007)Thedecayof LDwas assessedby

averagingr2valuesupto1Mbdistance

To test if the population structure influenced the detection of rEHH we

replicated all analyses on the whole Italian Holstein dataset without any

population correction (ie ldquoredundantrdquo dataset) focusing on genes or gene

clustersthatarewellknowntobeunderrecentselectionincattle(ieldquocandidate

regionsrdquo) In particularwe focused on the casein polled gene cluster and coat

colourgenes(MC1RandKIT(Kemperetal2014Qanbarietal2010))

The EHH and rEHH calculations were performed using Sweep v11 software

(Sabetietal2002)Somedefaultprogramsettingshadtobemodifiedtoadapt

theseanalyses to thebovinegenomeas thesoftwarewasoriginallydeveloped

forhumangeneticanalysesSpecificallylocalrecombinationratesbetweenSNPs

16

wereapproximatedto1cMperMbEHHandrEHHcalculationswereperformed

breedwise and chromosomewise using an automatic core selection with

defaultoptions(ieconsideringlongestnonoverlappingcoresandlimitingcores

toatleast3andnomorethan20SNPs)asinQanbarietal(2010)AlthoughEHH

and rEHH values were obtained for all core haplotypes only those with

frequency ge 25 were retained for further analyses The (empirical) rEHH

significancethresholdwasobtainedbydividingtherEHHvaluesinto20binsof

5 frequency range each logtransforming withinbin values to achieve

normality Core haplotypes with pvalues lt 005 were considered significant

DetectionofcommonselectionsignatureswasobtainedcomparingrEHHvalues

acrossgenomicregionsacrossbreeds

Breedgroupingaccordingtoproductiontype

Toexamine thesignals common toeachof the twoproduction types (iedairy

andbeef) corehaplotypes thatsharedoneormoreSNP inat least twobreeds

from the same production type were selected The dual purpose Italian

Simmentalwas included inbothdairy (ItalianHolsteinand ItalianBrown)and

beef groups (Piedmontese and Marchigiana) since potentially it possesses

haplotypes selected for both production types All downstream analyses were

performedseparatelyforthedairyandthebeefproductiontypes

Detectionandannotationofcandidategenes

Genome annotation was performed using Biomart

(httpwwwbiomartorgindexhtml) a comprehensive source of gene

annotationformanyspeciesprovidedbyEnsembl(wwwensemblorg)Thelist

ofsharedhaplotypesregions(inbp)fordairyandbeefwasusedasinputfilein

theBiomartwebinterfaceThesetofgenesretrievedbyBiomartwasthenused

as input for theCanonicalpathwayandregulatorynetworkanalyses Ingenuity

PathwayAnalysis toolversion80(IPA IngenuityregSystems IncRedwoodCity

CA httpwwwingenuitycom) and a substantial examination of published

literaturewasusedtoexaminethefunctionalrelationshipsamongtheresulting

genes The IPA operates with a proprietary knowledge database providing

complementarypathwayanalysisforseveralspeciesincludingcattleIntheIPA

17

analyses the significance of each biological function identified was estimated

using Fisherrsquos exact test A Benjamini and Hochberg correction for multiple

testingwas calculated for function and canonical pathways For eachNetwork

IPAcomputedascore(pscore=log10(pvalue))fittingthesetstudiedgenesand

a list of biological functionspresent in the IPAknowledgedatabase The score

considered thenumberofgenes in thenetworkand thesizeof thenetwork to

estimatehowrelevantthisnetworkwasaccordingtothelistofgenesprovided

Thenthescorewasusedtorankthenetworks

Results

DatasetQC

Therepeatabilityobtainedfromthe101replicatespresentinthewholedataset

washigherthan998AfterthetwoQCstages105individualsand9730SNPs

were discarded Moreover after phasing 1292 individuals were removed to

reducethe largenumberofsibfamiliespresent intheredundantdatasetAfter

quality control procedures the final dataset had 44271 SNPs and 1132 514

393 410 and 364 individuals from Italian Holstein Italian Brown Italian

SimmentalMarchigianaandPiedmontesebullsrespectively(Table1)

18

Table1JNumberofanimalsgenotypedbeforeandafterQCTotalnumberofanimalgenotypedandrelativenumberofanimaldiscardedafterQCanalysis

Detectionofselectionsignatures

Sweepv11 softwaredetectedrEHHsignalswitha frequencyhigher than25overlapping the test candidate regions in both datasets corrected or not forpopulation structure In the redundant dataset we found only one significantrEHHsignal(caseincluster)whereasinthenonredundantdatasetweidentified5significantrEHHsignalsoverlappingallthetestcandidateregionsconsidered(caseinclusterpolledclusterMC1R)(Table2)

BreedTotal

Genotyped

ED15

misAN

ED1

REPL

ED1

MENDCleaned

HOL 2179 40 31 5 2093

BRW 775 6 16 4 749

SIM 493 6 6 2 479

MAR 485 37 38 410

PIE 379 5 10 364

19

Table2JComparisonofcandidateregionrEHHinnonJredundantandredundantdatasetrEHH overlapping candidate regions in both corrected (nonredundant) and not corrected

(redundant)datasetsforpopulation

Redundant NonRedundant

Candidate

geneBTA1

Closest

SNP(bp)CH2range CH2Frequency

rEHHJ

log(pval)CH2Frequency

rEHHJ

log(pval)

POLLED

Cluster1 1981154 18974181981154 H1079 093037 H1078 167174

MC1R 18 13657912 1331772014007505 H1069 120096 H1069 077167

KIT_BOVIN 6 72821175 7250492172821175 H1054 030030 H1054 033048

Casein

Cluster6 88427760 8835009888452835 H1047 094139 H1046 148133

Casein

Cluster6 88427760 8835009888452835 H2032 002003 H2033 002003

In total5526567847724388and4049corehaplotypeswitha frequency

higher than 25 were detected on Italian Holstein Italian Brown Italian

Simmental Marchigiana and Piedmontese breeds respectively A total of 838

866 740 692 and 613 core haplotypes were found significant outliers (see

Materials and Methods) in the aforementioned breeds Table 3 shows the

distribution of the total and significant core haplotypes per chromosome and

breed

20

Tableamp3amp(ampCoreamphaplotypesampdistributionampDistributionoftotalandsignificantcorehaplotypesperchromosomeandbreed

ampamp ItalianampampHolsteinamp

ItalianampampBrownamp

ItalianampSimmentalamp Marchigianaamp Piedmonteseamp

BTA1amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

Hampgtamp252amp

SignampH3amp

1amp 389 66 389 68 335 49 310 51 288 46

2amp 312 47 314 52 267 44 257 42 238 33

3amp 292 48 313 62 264 46 236 34 227 34

4amp 268 46 279 50 245 41 237 39 229 39

5amp 223 32 231 29 200 30 183 27 171 22

6amp 291 51 307 40 265 45 252 38 233 42

7amp 249 37 253 46 214 39 195 33 204 27

8amp 268 39 270 38 235 35 223 32 209 31

9amp 227 37 228 41 210 37 165 22 187 30

10amp 219 36 234 34 192 25 196 26 168 25

11amp 248 39 246 34 209 30 205 33 187 25

12amp 162 24 176 18 148 21 148 25 135 22

13amp 205 32 200 19 155 18 178 35 135 19

14amp 175 17 189 25 168 29 143 22 129 19

15amp 180 31 185 29 148 23 139 13 130 19

16amp 176 28 192 29 160 25 149 19 127 15

17amp 170 27 183 29 168 31 135 22 123 16

18amp 133 18 153 19 119 14 108 10 82 12

19amp 137 22 130 23 118 16 99 18 92 15

20amp 184 35 172 33 145 23 124 23 124 17

21amp 145 23 147 30 113 18 110 21 94 13

22amp 140 24 145 19 115 21 112 18 104 13

23amp 106 13 100 16 67 10 64 9 57 8

24amp 129 11 140 20 124 17 97 23 101 16

25amp 91 12 98 11 68 10 77 12 70 6

26amp 125 7 114 14 93 12 96 15 90 16

27amp 91 12 91 15 75 11 84 12 60 10

28amp 88 9 96 10 70 10 66 9 55 11

29amp 103 15 103 13 82 10 72 9 67 12

TOTamp 5526 838 5678 866 4772 740 4388 692 4049 613

1BostaurusAutosomes2Detectedcorehaplotypeswithafrequencyinthebreedhigherthan253Significantcorehaplotypesatplt005

Comparisonamptoamppreviouslyampreportedampdataamp

AnumberofstudieshavesearchedselectionsweepsinHolstein(Qanbarietal

2010) Brown (Qanbari et al 2011) and Simmental (Fan et al 2014) breeds

Since different methods are expected to identify signatures having different

characteristicsthecomparisonwithpreviousstudiesislimitedtothosestudies

using thesamemethodand thesamebreed(s)analysed inourstudyOnlyone

study analysed (German) HolsteinRFriesian cattle with rEHH (Qanbari et al

2010)Qanbarietal(2010)reportedcandidategenesandgeneclustersunder