ContentslistsavailableatScienceDirect
Small Ruminant Research
j o u r n a l ho me p ag e :w w w . e l s e v i e r . c o m / l o c a t e / s m a l l r u m r e s
Estimates of genetic parameters and genetic trends for pre-weaning growth traits in Sardi sheep
I. Boujenane
a,∗, I.T. Diallo
baDepartmentofAnimalProductionandBiotechnology,InstitutAgronomiqueetVétérinaireHassanII,P.O.Box6202Rabat-Instituts,10101Rabat,Morocco
bPROGEBE,Guinea
a r t i c l e i n f o
Articlehistory:
Received6May2016 Receivedinrevisedform 28November2016 Accepted1December2016 Availableonline5December2016
Keywords:
Sheep Sardibreed Growthtrait Maternaleffects Heritability Genetictrend
a b s t r a c t
Theaimofthepresentstudywastoestimate(co)variancecomponentsforweightsatbirth(BW),30days (W30d)and90days(W90d)ofage,aswellasaveragedailygainfrombirthto30days(ADG1)andfrom 30to90days(ADG2)ofSardisheep.Recordsof3709lambsthatdescendedfrom1170damsand85 sireswerecollectedoveraperiodof21yearsintheKraKrasheepfarm.(Co)variancecomponentsand thecorrespondinggeneticparameterswereestimatedunderRestrictedMaximumLikelihood(REML) methodbyfittingofsixdifferentanimalmodelswithignoringorincludingmaternaleffectsincluding maternalgeneticormaternalpermanentenvironmentaleffects.Thebestmodelwaschosenbasedon Akaike’sInformationCriterion(AIC).Bivariateanalysiswasperformedusingthemostappropriatemodels obtainedinunivariateanalysis.Ageofdam,typeofbirth,sexoflambandyearofbirthshowedsignificant effectsonstudiedtraits,sotheywereconsideredasfixedeffectsinanalyzingmodels.Directheritability estimatesforBW,W30d,W90d,ADG1andADG2were0.07±0.02,0.09±0.03,0.05±0.02,0.01±0.01 and0.05±0.02,respectively.Maternalheritabilityestimateswere0.13±0.03and0.21±0.04forBWand W30d,respectively.TheseresultsindicatethatmaternaleffectsonBWandW30dofSardilambsneedto beconsideredinanyselectionprogrammeundertakeninthisbreed.Maternalpermanentenvironmental effectsaccountedfor11%,8%and8%ofphenotypicvarianceforW90d,ADG1andADG2,respectively.
Thedirectgeneticcorrelationsbetweenstudiedtraitswerepositiveinallcasesandrangedfrom0.23 forBW-ADG1to0.98forW90d-ADG2.Thecorrespondingphenotypiccorrelationsvariedfrom−0.01for BW-ADG1to0.92forW30d-ADG1.ThematernaladditivegeneticcorrelationbetweenBWandW30d was0.66.Also,theenvironmentalcorrelationsamongstudiedtraitsvariedfrom−0.04forBW-ADG1to 0.93forW30d-ADG1.Annualgeneticgainsfordirectandmaternaladditivegeneticeffectsofstudied traitswereclosetozero.Thustorealizeageneticprogress,selectionshouldbebasedonadditivegenetic valuesinsteadofphenotypiccharacteristics.
©2016ElsevierB.V.Allrightsreserved.
1. Introduction
TheSardiisthemostimportantsheepbreedinMorocco.Itis anexcellentmeatbreedanditishighlydesiredforthecelebra- tionofthereligiousAidAl-Adha.Thismayexplaintheincreaseof itsnumber,reachingabout2.15millionhead,andtheextension ofitsgeographicalarea,goingfromthecentretothenorthofthe country.Theheightatwithersofadultramsandewesaveraged 80–90cmand55–65cm,respectively.Thishighsize,however,is notaccompaniedbyaheavyweight,sinceadultweightsoframs andewesaveragedonly65kgand45kg,respectively(Boujenane,
∗Correspondingauthor.
E-mailaddresses:[email protected],[email protected] (I.Boujenane),[email protected](I.T.Diallo).
2005),indicatingthatthereisenoughpotentialforincreasingof bodyweightandmeatproduction.Therefore,thereisacrucialneed fortheimprovementofgrowthpre-weaningtraitsofSardisheep.
However,inordertodesignbreedingstrategiesforSardisheep, geneticandnon-geneticparametersforgrowthtraitsshouldbe estimated.Boujenaneetal.(2001)havealreadyestimatedgenetic parametersforgrowthtraitsofSardisheepusingasiremodel.How- ever,numerousstudies(Mohammadietal.,2010;Savar-Soflaetal., 2011;Boujenaneetal.,2015)havedemonstratedthattheinclusion of maternal effects in animal models for growth traits, partic- ularly duringthepre-weaning period,hasanimportant impact ontheestimatesofdirectheritability.However,ignoringmater- naleffectsmightresultinoverestimationofgeneticparameters and mightincrease thebiasinpredictedresponsestoselection (Ghafouri-KesbiandEskandarinasab,2008).Thus,toachieveopti-
http://dx.doi.org/10.1016/j.smallrumres.2016.12.002 0921-4488/©2016ElsevierB.V.Allrightsreserved.
mumprogressinaselectionprogramme,bothdirectandmaternal componentsshouldbetakenintoaccount,especiallyinthepres- enceofanantagonisticrelationshipbetweenthem(Ekizetal.,2004;
Rashidietal.,2008).
Therefore,themainobjectiveofthecurrentstudywastoesti- matethegeneticparameters fordirect andmaternal effectson growthtraitsinSardi sheepbyfittingdifferentmodels,includ- ingdirectadditivegenetic,maternaladditivegeneticandmaternal permanentenvironmentaleffects.
2. Materialandmethods 2.1. Animalsandmanagement
ThestudywascarriedoutattheKraKrasheepselectionfarm locatedinthecentreofMorocco.Thefarmhasasemi-aridclimate, withanaverageannualrainof200mm.Growthdatawerecollected overaperiodof21years(1983–2004,except1995).Eachyearabout 200eweswerejoinedto6ramsforaperiodof75days,frommid- JunetotheendofAugust.Anocturnesinglesirematingsystemwas practisedwitharatioofabout40ewesperram.Atmating,aram offamilywasnotallowedtomateewesofitsownfamily.Lambing periodoccurredfromNovembertoFebruary.Ewelambswerebred at18monthsofage.Throughoutthestudy,replacementanimals wereselectedbasedonweaningweightandphysicalconformation.
Animalsweregrazed onnatural pasturesduringthe day.In general,grazingpasturegrassescoveredabout70%ofbreedingani- malsfeedrequirements.Theremainingof30%wascoveredbya supplementationprovidedduringthemating(600g/day),thelast monthof pregnancyandthebeginning oflactation (750g/day).
Ramsreceivedabout1kgperdayofconcentratemixture.Allani- malswereprovidedwithminerallicksandwateringwasallowed twiceaday;beforeandaftergrazing.
Sheepwerevaccinatedagainstenterotoxaemiaandsheep-pox andtreatedagainstinternalandexternalparasites.Also,sickani- malswereusuallytreated.
2.2. Datarecordingandstudiedtraits
Atbirth,lambs wereear-taggedand informationconcerning theirdam,birthdate,sexandtypeofbirthwereregistered.Lambs ranwiththedamsupto3monthswhentheyareweaned.Lambs wereweighedatbirthandevery3weeksuntilweaning.Investi- gatedtraitswereweightsatbirth(BW),30days(W30d)and90days (W90d).Thelasttwoweightswerecomputedbyapplyingthelinear interpolationtoweightsobtainedatweightingsrealizedjustbefore andaftertheconsideredage(30or90days).Averagedailygains frombirthto30days(ADG1)andfrom30to90days(ADG2)were calculatedassuminglineargrowthratebetweentheappropriate weights.
2.3. Statisticalanalyses
Growth data were collected over a period of 21 years (1983–2004,except1995) on3709Sardilambsbornfrom1170 damsand85siresregisteredintheflockbookofSardibreed.Ini- tially,the datasetincluded4323 records.However,lambs with implausible birth dates, unknown sex, unknown type of birth or born from unknown parents were discarded. The structure ofdata set,whichis used inthecurrent study,is presented in Table1. Toidentify fixedeffects tobe includedin themodels, least-squaresanalyseswereconductedusingtheGLMprocedure (SAS,2002).Thiswasperformedonamodelincludingfixedeffects of age of dam (≤30months,>30–≤42months,>42–≤54months and>54months),typeofbirth(singleandtwin),sexoflamb(male
andfemale)andperiodofbirth(from1982/83to2003/04).Inter- actionsbetween fixedeffects werenot tested, and hencewere assumedtobenegligible.Whenaneffectwasdeterminedtobe significant(P<0.05),differencesamongleast-squaresmeanswere examinedbytheTukeymethodformultiplecomparisons.Allthe effectsweresignificantforalltraits,and thereforetheseeffects wereincludedinthemodel.
Variancecomponentswereestimatedforeachtraitseparately withderivative-freerestrictedmaximumlikelihood(REML)pro- ceduresusingtheMTDFREMLprogramme(Boldmanetal.,1995).
Six different models were fitted for each trait, by ignoring or including maternal additive genetic effect, covariance between direct-maternaladditivegeneticeffectandmaternalpermanent environmentaleffect.Thesemodelsare:
y= Xb+Zaa+e (1)
y= Xb+Zaa+Wpepe+e (2)
y= Xb+Zaa+Zmm+e Cov(a,m) =0 (3) y= Xb+Zaa+Zmm+e Cov(a,m) =Aam (4) y= Xb+Zaa+Zmm+Wpepe+e Cov(a,m) =0 (5) y= Xb+Zaa+Zmm+Wpepe+e Cov(a,m) =Aam (6) whereyisavectorofobservations,bisavectoroffixedeffectswith incidencematrixX,a∼N(0,Aa2)andm∼N(0,Am2)arevectorsof directandmaternaladditivegeneticeffectswithincidencematri- cesZaandZm,respectively,pe∼N(0,Idpe2)isavectorofrandom maternalpermanentenvironmentaleffectswithincidencematrix Wpe,ande∼N(0,Ine2)isavectorofrandomresidualeffects.Also,a2 isthedirectadditivegeneticvariance,m2 isthematernaladditive geneticvariance,amisthecovariancebetweendirectandmaternal additivegeneticeffects, pe2 isthematernalpermanentenviron- mentalvariance,e2istheresidualvariance,Aistheadditivegenetic relationshipmatrix,andId andIn areidentitymatricesoforder equaltothenumberofdamsandthenumberofrecords,respec- tively.Convergencewasconsideredreachedwhenthevarianceof functionvalues(−2logL)inthesimplex wasless than10−8.To ensurethataglobalmaximumwasreached,severalotherroundsof iterationswereusedusingresultsfromthepreviousroundasstart- ingvalues.Whenestimatesdidnotchangeattheseconddecimal, convergencewasconfirmed.
Inunivariateanalysis,theAkaike’sinformationcriterion(AIC) wasusedtodeterminethemostappropriatemodelforestimating (co)variancecomponentsforeachtraitasfollows(Akaike,1973):
AICi= −2logLi+2pi
wherelogLiisthemaximizedloglikelihoodofmodeliatconver- genceandpiisthenumberofindependentlyestimatedparameters ofmodeli.ThemodelwiththesmallestAICwasconsideredasthe mostappropriatemodel.
Total heritability estimates were calculated according to Willham(1972)as:
h2t = a2+0.5m2+1.5am
p2
Similarderivative-freeREMLprocedureswereusedforbivariate analysestoestimatecovariancesbetweeneachpairoftraits.The modelsappliedinbivariateanalyseswerethosefittedforeachof theunderlyingtraitsintheunivariateanalyses.Firstly,weusethe simplexvarianceoflessthan10−3toreachconvergence.Thenthe (co)varianceestimatesobtainedfromthepreviousrunwereused asstartingvaluestoreachconvergencewiththesimplexvariance oflessthan10−5.Finally,(co)variancecomponentsconsideredin thepresentstudywerethoseobtainedatthelastiteration.
Table1
ThecharacteristicsofthedatastructureforgrowthtraitsofSardisheep.
Item Traita
BW(kg) W30d(kg) W90d(kg) ADG1(g/day) ADG2(g/day)
Numberofrecords 3709 3537 3378 3537 3372
Numberofanimals 4323 4195 4059 4195 4053
Numberofsires 85 85 85 85 85
Numberofdams 1170 1157 1147 1157 1146
Averagenumberofprogenypersire 35.5 35.0 35.0 35.0 34.9
Averagenumberofprogenyperdam 3.14 3.03 2.92 3.03 2.92
Arithmeticmean 3.71 10.3 22.2 218 198
Standarddeviation 0.71 2.05 5.13 59.7 60.4
Coefficientofvariation,% 14.3 13.4 12.5 19.6 17.1
aBW=birthweight;W30d=weightat30days;W90d=weightat90days;ADG1=averagedailygainfrombirthto30days;ADG2=averagedailygainfrom30to90days.
Using thebest model for each trait, breedingvalues ofani- malswerepredictedwithBestLinearUnbiasedPrediction(BLUP) methodology.Theannualgeneticgainwasestimatedbycalculat- ingtheregressioncoefficientofadditivegeneticvalues(directand maternal)ofeachanimalonitsbirthyearusingtheREGprocedure (SAS,2002).
3. Resultsanddiscussion 3.1. Non-geneticeffects
Least-squaresmeansandrespectivestandarderrorsforstud- iedgrowthtraitsarepresentedinTable2.Allnon-geneticfactors hadsignificanteffectsonthestudiedtraits(P<0.001).Lambsborn fromyoungdamshadlowergrowthperformancesthanthoseborn fromadultewes.Theincreaseinlambgrowthwiththeincrease indam’sageisexplainedbytheimprovementofmotheringabil- ityandmilkproductionofmatureewes.Also,BW,W30d,W90d, ADG1andADG2 oflambsbornas singleswere1.03kg,1.92kg, 3.3kg,28g/dand24g/d,respectivelyhigherthanthoseoflambs born as twins. The increase in litter size frequentlyleads to a decreaseingrowthperformancesbecauseofcompetitionbetween twinsformilkproduction.Maleswereheavierandgrewfasterthan females.Differencesbetweenthetwosexeswere0.17kg,0.55kg, 1.7kg,12g/dand17g/dforBW,W30d,W90d,ADG1andADG2, respectively.GrowthtraitsofSardilambsvariedamongbirthyears.
Thoseborn in 1996/97and 2003/2004 had the highest perfor- mances,whereasthosebornin1982/83and1992/93hadthelowest performances.Thesedifferencesbetweenbirthperiodsoccurred probablybecauseofchangesthataffecttheprevailingclimaticcon- ditions,feedingandgeneralmanagementpractices.Non-genetic effectsongrowthtraitsareinagreementwiththosefoundinthe literature(Eskandarinasabetal.,2010;Mohammadietal.,2013;
Boujenaneetal.,2015).
3.2. Modelscomparison
Estimatesof(co)variancecomponentsandgeneticparameters forgrowthtraitsandAICvaluesforeachofthesixmodelsaresum- marizedinTables3and4.Model1,whichignoredmaternaleffects, resultedinhigherestimates fordirectadditivegeneticvariance (a2)anddirectheritability(h2d)thandidtheothermodelsforall traits,exceptforW30dandADG1.Introducingeitheramaternal permanentenvironmentaleffect(Model2)oramaternalgenetic effect(Model3)increasedtheloglikelihoodvaluesanddecreased estimatesfora2 andh2d.Theadditionofdirect-maternalgenetic covarianceincreaseda2andh2d(Model4)forW30dandADG1,but notforBW,W90dandADG2.Whenmaternalgeneticandmaternal permanentenvironmentaleffects,withorwithoutdirect-maternal geneticcovariance,wereincludedinthemodel(Models5–6),a2 andh2dweresimilartothoseinModels2–4forBW,W90dandADG2.
OnthebasisoftheAICtestresultsandthenumberofparameters usedinthepresentstudy,Model4,whichincludeddirectadditive geneticandmaternaladditivegeneticeffects,anddirect-maternal geneticcovariance, wasthemostappropriatemodeltodescribe BW and W30d, whereas Model 2, which included direct addi- tivegeneticand maternalpermanentenvironmentaleffect,was themostappropriatemodeltodescribeW90d,ADG1andADG2.
Thusforthelattertraits,thepermanentenvironmentaleffectof thedamwasdeterminedtobemoreimportantthanthematernal geneticeffect.Ghafouri-KesbiandBaneh(2012)reportedthatfor IranianMakooeisheep,Model2withthedirectadditivegenetic andmaternalpermanentenvironmentaleffectsandModel5with thedirectadditivegenetic,maternaladditivegeneticandmater- nalpermanentenvironmentaleffectswerethemostappropriate modelsforbirthweightandweaningweight,respectively,whereas Jafaroghlietal.(2010)showedthatmodels2–6werenotsignifi- cantlydifferentintermsofloglikelihoodwhenestimatingvariance componentsandgeneticparametersforgrowthtraitsinMoghani sheep.
Themagnitudeofmaternalpermanentenvironmentalvariance asaproportionofphenotypicvariance(pe2)obtainedbydiffer- entmodels(2,5and6)waschangedfrom3to9%forBW,0–11%
forW30d,6–11%forW90d,0–8%forADG1and0–8%forADG2, respectively.Moreover,theinclusionofmaternalpermanentenvi- ronmentaleffectsinmodel2,5and6reducedmaternaladditive geneticvarianceforalltraitsanddirectadditivegeneticvariance for BW, W30d and ADG1.As inbreeding mayreduce theaddi- tivegeneticvarianceandtheheritability(Kristensenetal.,2005), thereductionobservedinthecurrentstudycouldnotbedueto inbreedingsinceitsmeanwaszerointhestudiedflock.Thus if significant,pe2 shouldbeconsideredinordertoobtainaccurate estimatesofdirectandmaternalheritabilities.
3.3. Geneticparameters
Basedonthemostappropriatemodels,directheritabilityesti- matesforBW,W30d,W90d,ADG1andADG2were0.07,0.09,0.05, 0.01and0.05,respectively.Theseestimatesforbodyweightswere intherangeofthosereportedinseveralstudies(Boujenaneand Kansari,2002;Jafaroghlietal.,2010;Rashidietal.,2008;Boujenane etal.,2015).However,theywerelower thanthoseobtainedby Miraei-Ashtianietal.(2007)inSangsarisheep(0.33),Kushwaha etal.(2009)inChoklasheep(0.20)andGamasaeeetal.(2010)in Mehrabansheep(0.30).DirectheritabilityestimatesforADG1and ADG2inthecurrentstudywereclosetothosereportedbyMiraei- Ashtianietal.(2007)foraveragedailygainfrombirthtoweaning inSangsarisheep,butlowerthanthosefoundbyJafaroghlietal.
(2010),Abegazetal.(2005)andRashidietal.(2008)foraverage dailygainfrombirthtoweaninginMoghani,HorroandKermani sheepbreeds,respectively.Thelowdirectheritabilityestimatesfor growthtraitsobtainedinthecurrentresearchmaybeattributed
Table2
Numberofobservations(N),least-squaresmeans(LSM)±standarderrors(SE)forgrowthtraitsinSardilambsa.
Factorsofvariation BW(kg) W30d(kg) W90d(kg) ADG1(g/day) ADG2(g/day)
N LSM±SE N LSM±SE N LSM±SE N LSM±SE N LSM±,SE
Ageofdam(years) *** *** *** * ***
Age≤2.5 1145 3.23±0.02c 1051 9.38±0.08b 1028 20.9±0.22c 1051 206±2,62b 1028 191±2,64c 2.5<Age≤3.5 936 3.40±0.02b 908 9.67±0.09a 845 21.3±0.22b 908 210±2,73a 844 193±2,73bc 3.5<Age≤4.5 928 3.46±0.02a 905 9.79±0.09a 839 21.70.22a 905 212±2,77a 838 198±2,73a Age>4.5 700 3.45±0.02ab 673 9.65±0.09a 666 21.5±0.23ab 673 208±2,89ab 662 196±2,81ab
Typeofbirth *** *** *** *** ***
Single 3135 3.90±0.01a 2994 10.6±0.07a 2875 23.0±0.20a 2994 223±2,32a 2869 207±2,44a
Twin 574 2.87±0.02b 543 8.68±0.09b 503 19.7±0.23b 543 195±2,89b 503 183±2,84b
Sexoflamb *** *** *** *** ***
Female 1905 3.30±0.02b 1825 9.35±0.08b 1756 20.6±0.21b 1825 203±2,52b 1754 186±2,59b
Male 1804 3.47±0.02a 1712 9.90±0.08a 1622 22.2±0.21a 1712 215±2,51a 1618 203±2,57a
Periodofbirth *** *** *** *** ***
aLeast-squaresmeanswithinacolumnthatdonothaveacommonsuperscript(a–c)aresignificantlydifferent(P<0.05).Fortraitsabbreviations,seefootnoteofTable1.
* P<0.05.
***P<0.001.
Table3
Estimatesof(co)variancecomponentsandgeneticparameterestimatesforgrowthtraitsinSardisheepfromunivariateanalyses.a
Trait Model a2 m2 am 2pe 2e 2p h2d±SE h2m±SE ram pe2 h2t AIC
BW 1 0.03 0.26 0.28 0.09±0.02 0.09 −864
2 0.02 0.02 0.24 0.28 0.06±0.02 0.09 0.06 −898
3 0.02 0.02 0.24 0.28 0.06±0.02 0.09±0.02 0.11 −898
4 0.02 0.04 −0.02 0.24 0.28 0.07±0.02 0.13±0.03 −0.60 0.04 −899
5 0.02 0.01 0.02 0.24 0.28 0.06±0.02 0.03±0.02 0.06 0.09 −896
6 0.02 0.03 −0.01 0.01 0.24 0.28 0.07±0.02 0.10±0.03 −0.67 0.03 0.02 −897
W30d 1 0.12 1.80 1.92 0.06±0.02 0.06 5905
2 0.05 0.20 1.66 1.92 0.03±0.02 0.11 0.03 5865
3 0.05 0.20 1.66 1.92 0.03±0.02 0.11±0.02 0.03 5865
4 0.17 0.41 −0.27 1.60 1.91 0.09±0.03 0.21±0.04 −1.00 0.02 5832
5 0.05 0.13 0.08 1.66 1.92 0.03±0.02 0.07±0.01 0.04 0.06 5867
6 0.14 0.19 0.16 0.00 1.58 2.07 0.07±0.03 0.09±0.01 1.00 0.00 0.23 5903
W90d 1 0.92 6.93 7.85 0.12±0.03 0.12 10321
2 0.36 0.89 6.52 7.77 0.05±0.02 0.11 0.05 10272
3 0.36 0.89 6.52 7.76 0.05±0.02 0.11±0.02 0.10 10272
4 0.34 0.80 0.10 6.53 7.78 0.04±0.02 0.10±0.03 0.20 0.11 10274
5 0.36 0.11 0.77 6.52 7.77 0.05±0.02 0.01±0.02 0.10 0.05 10274
6 0.34 0.33 0.11 0.47 6.53 7.78 0.04±0.02 0.04±0.03 0.32 0.06 0.09 10276
2a=directadditivegeneticvariance;2m=maternaladditivegeneticvariance;am=direct-maternalgeneticcovariance;pe2=maternalpermanentenvironmentalvariance;
2e=residualvariance;p2=phenotypicvariance;h2d= a22 p
=heritabilityofdirectgeneticeffects;h2m= m22 p
=heritabilityofmaternalgeneticeffects;ram= aamm=correlation betweendirectandmaternalgeneticeffects;pe2=2c2
p
=maternalpermanentenvironmentalvarianceasaproportionofphenotypicvariance;h2t= 2a+0.5m22+1.5am p
.,Fortraits abbreviations,seefootnoteofTable1.
Boldvaluesindicateestimatesfromthemostappropriatemodel.
aVariancecomponentsofweightsareinkg2.
tothelowqualityofpasturesonwhichtheflockwasmaintained, resultinginahighenvironmentalvariance.Furthermore,according totheobtaineddirectheritabilityestimatesinthepresentstudy, growthtraitsinSardisheepareclassifiedaslowheritabletraits, indicatingthatperformanceofanimalsareless usefulinidenti- fyingtheindividualswiththehighgeneticmerit,andtherefore, lowgeneticprogresswouldbeexpectedthroughphenotypicselec- tionprogrammes.Thus,todetectthebestanimalsandtorealizea geneticprogress,selectionshouldbebasedonestimatesofbreed- ingvaluesandnotonrawperformanceoflambs.
AsAICvaluesclearlyindicate,maternaladditivegeneticeffects are important for BW and W30d since when using the most appropriatemodel(Model4),h2mestimateswere0.13and 0.21, respectively,buttheseeffectswerenotessentialforW90d,ADG1 andADG2.Itisnotablethatdatastructurehasagreatimpacton theaccuracyofmaternaleffectsestimation.ManiatisandPollott (2003)reportedthat foraccurately separatingmaternalgenetic componentsandmaternalpermanentenvironmentaleffectsfrom
combinedanddirecteffects, alargedataset,severalwell-linked generationsofrecordsandmanyrelationshipsbetweenrelatives relatedtothemotherareneeded.Thesmallimpactofmaternal effectsonW90dandADG2inthepresentstudymaybeduetothe factthatsucklinglambsrelymainlyontheirmother’smilkfrom birthto30days,whereasafterthisage,theimportanceofmilkyield ofdamsdecreasesmorerapidlyandlambsdependmoreonthem- selves.Sotheimpactofmaternalgeneticpotentialonbodyweight gainoflambs,asreportedbyGizawetal.(2007),wasdecreasedand thelamb’sphenotypeforgrowthwasmorerelatedtolamb’saddi- tivegeneticeffects.Theh2mestimateforBWinthecurrentstudy wasinagreementwiththosereportedbyothers(Ekizetal.,2004;
Mandaletal.,2006b;Kushwahaetal.,2009),butlowerthanesti- matesofRashidietal.(2008),Gamasaeeetal.(2010)andJafaroghli etal.(2010)indifferentsheepbreeds.Moreover,thelowmaternal geneticinfluenceongrwthtraitsmaybeduetothelackofadequate pasturethatcouldbemaskingtheexpressionofthematernalability oftheewe(Ghafouri-Kesbietal.,2008).
Table4
Estimatesof(co)variancecomponentsandgeneticparameterestimatesforgrowthtraitsinSardisheepfromunivariateanalyses.a
Trait Model 2a 2m am pe2 e2 p2 h2d±SE h2m±SE ram pe2 h2t AIC
ADG1 1 56 1772 1829 0.03±0.02 0.03 30006
2 14 145 1668 1827 0.01±0.01 0.08 0.01 29983
3 14 145 1668 1827 0.01±0.01 0.08±0.02 0.05 29983
4 158 143 150 1539 1990 0.08±0.04 0.07±0.03 1.00 0.23 30030
5 14 6 139 1667 1827 0.01±0.01 0.00±0.00 0.08 0.01 29985
6 119 128 124 0 1570 1941 0.06±0.03 0.07±0.03 1.00 0.00 0.19 30017
ADG2 1 114 1045 1159 0.10±0.03 0.10 27029
2 58 90 1003 1150 0.05±0.02 0.08 0.05 27005
3 58 90 1002 1151 0.05±0.02 0.08±0.02 0.09 27005
4 49 52 50 1007 1158 0.04±0.02 0.04±0.03 1.00 0.13 27005
5 58 0 90 1003 1151 0.05±0.02 0.00±0.00 0.08 0.05 27005
6 49 52 50 0 1008 1160 0.04±0.02 0.05±0.01 1.00 0.00 0.13 27007
a2=directadditivegeneticvariance;m2=maternaladditivegeneticvariance;am=direct-maternalgeneticcovariance;pe2=maternalpermanentenvironmentalvariance;
e2=residualvariance;2p=phenotypicvariance;h2d= 2a2
p
=heritabilityofdirectgeneticeffects;h2m= 2m2
p
=heritabilityofmaternalgeneticeffects;ram= aamm=correlation betweendirectandmaternalgeneticeffects;pe2=c22
p
=maternalpermanentenvironmentalvarianceasaproportionofphenotypicvariance,h2t= 2a+0.52m2+1.5am p
.Fortraits abbreviations,seefootnoteofTable1.
Boldvaluesindicateestimatesfromthemostappropriatemodel.
aVariancecomponentsofADGs’areing2.
Table5
Phenotypic,directgenetic,maternalgenetic,maternalpermanentenvironmentalandenvironmentalcorrelationestimatesamongstudiedtraitsunderbivariateanimal models.a
Traits1 rp ra rm re
BW–W30d 0.37 0.72±0.18 0.66±0.14 0.32±0.02
BW–W90d 0.27 0.68±0.18 – 0.22±0.02
BW–ADG1 −0.01 0.23±0.34 – −0.04±0.02
BW–ADG2 0.12 0.54±0.23 – 0.07±0.02
W30d–W90d 0.72 0.55±0.23 – 0.73±0.01
W30d–ADG1 0.92 0.86±0.08 – 0.93±0.01
W30d–ADG2 0.30 0.28±0.31 – 0.29±0.02
W90d–ADG1 0.66 0.79±0.14 – 0.67±0.01
W90d–ADG2 0.88 0.98±0.02 – 0.88±0.01
ADG1–ADG2 0.28 0.62±0.23 – 0.26±0.02
1Fortraitsabbreviations,seefootnoteofTable1.
arp:phenotypiccorrelation;ra:directgeneticcorrelation;rm:maternalgeneticcorrelation;re:environmentalcorrelation.
Table6
AnnualgeneticprogressforgrowthperformanceofSardilambsfrom1983to2004a. Traitb Directadditivegeneticeffects Maternaladditivegeneticeffects BW(g) −1.12±0.18*** 1.08±0.19***
W30d(g) 0.17±0.40NS −0.26±0.62NS W90d(g) −2.65±0.88** –
ADG1(g) −0.01±0.004** – ADG2(g) −0.03±0.01** –
aNS:P>0.05.
bFortraitsabbreviations,seefootnoteofTable1.
**P<0.01.
***P<0.001.
Atbirthand30days,h2mwasgreaterthanh2d.Estimateswere 0.13vs.0.07forBWand0.21vs.0.09forW30d.Similartendency wasobservedbyBahreiniBehzadietal.(2007)andGhafouri-Kesbi andEskandarinasab(2008)indicatingthattheinclusionofmater- nalgeneticeffectsinthemodelledtoagreaterdecreaseindirect heritability.Therefore,ifmaternaleffectsarepresent,theyshould befittedinthemodel,otherwise,estimatesofdirectheritability willbeoverestimatedandthetruegeneticpotentialoflambsis maskedbymaternalperformancemakingtheselectionofsuperior animalsdifficult(Ghafouri-KesbiandBaneh,2012).
Fromappropriatemodels,fractionsofvarianceduetomaternal permanentenvironmentaleffectsonphenotypicvariancewere11%
forW90dand8%forADG1andADG2.Ghafouri-Kesbietal.(2008) reportedthatestimatesofc2were4%forWWandJafaroghlietal.
(2010)foundthecorrespondingestimateofabout0.08foraverage dailygainfrombirthtoweaning.
The best model (Model 4) showed that fitting the direct- maternal genetic covariance resulted to estimates of direct- maternalgeneticcorrelation(ram)of−0.60and−1.00forBWand W30d,respectively.ManiatisandPollot(2003)reportedthatthe highernegativeestimatesmaybeattributedtothedatastructure.
Nevertheless, moderate to high negative estimates of correla- tionsbetweendirectandmaternalgeneticeffectsforpre-weaning growthtraitswerecommonlyreportedintheliterature(Abegaz et al.,2002; Boujenaneand Kansari,2002; Rashidiet al.,2008;
Kariukietal.,2010).Likewise,verylargenegativeestimateswere alsoreportedinanumberofotherstudies(Ekizetal.,2004;Mandal etal.,2006b).Cundiff(1972)postulatedthatfromanevolutionary pointofview,thenegativecovariancebetweendirectandmaternal geneticeffectspreventsspeciesfrombecomingincreasinglylarger.
Asaconsequenceofthenegativecorrelationbetweendirectand maternalgeneticeffect,differentrankingsofindividualsmaybe obtainedwhenthematernalcontributionisignoredintheevalua- tionprocedure(Szwaczkowskietal.,2006).
Estimatesoftotalheritability(h2t)forBW,W30d,W90d,ADG1 andADG2were0.04,0.02,0.05,0.01and0.05,respectively.These estimatesareverylow,suggestingthatmassselectionwouldnot beveryeffectiveinimprovingthesetraits.Withregard topub- lishedestimates,thetotalheritabilityforgrowthtraitsobserved inthisstudywerecomparablewiththeresultsofseveralstudies (BoujenaneandKansari,2002inTimahditesheep;Ekizetal.,2004 inTurkishMerinosheep;Mandaletal.,2006ainMuzaffarnagari
-200 -150 -100 -50 0 50 100 150
1983 1985
1987 1989
1991 1993 1996
1998 2000 2002
2004
Year of birth
Average direct genetic value (g)
BW W30d W90d ADG1 ADG2
Fig.1.DirectgenetictrendsforBW,W30d,W90d,ADG1andADG2intheSardibreed.
-150 -100 -50 0 50 100
1983 1985
1987 1989
1991 1993
1996 1998 2000
2002 2004
Year of birth Average maternal genetic value (g)
BW W30d
Fig.2.MaternalgenetictrendsforBWandW30dintheSardibreed.
sheep),butlowerthantheresultsofseveralstudies(Neseretal., 2001;Safari etal.,2005;Matikaetal.,2003)indifferentbreeds ofsheep.Willham(1972)advocatedthatwherematernalgenetic effectsarepresent,thepotentialresponsetoselectionmightbet- terbeexpressedbyh2t.However,ifthereisnegativecovariance betweendirectandmaternalgeneticeffects,whichisthecaseinthe presentstudyforBWandW30d,phenotypicresponsetoselection maybediminished(Wolfetal.,1998).
Theresultsobtainedfrombivariateanalysesarepresentedin Table5.Estimatesofdirectgeneticcorrelationsbetweengrowth traitswerepositiveandmediumtohigh,varyingfrom0.23between BW and ADG1 to0.98 betweenW90d and ADG2. The positive directgeneticcorrelationsamongthestudiedtraitssuggestthat geneticfactorswhichinfluencethesetraitswereinsimilardirec- tionandthatselectionforanyofthesetraitswillbringoutapositive responsetoselectionforothers.Moreover,thedecreaseofdirect geneticcorrelationestimatesofBWwithW30dandW90dwith ageisinagreementwiththepublishedvaluesinotherliteratures (Mohammadietal.,2010;Abbasietal.,2012;Mohammadietal., 2013).Similartoourestimates,highandpositivegeneticcorrela- tionsbetweengrowthtraitshavebeenreportedbyseveralauthors
invarioussheepbreeds(Miraei-Ashtianietal.,2007;Rashidietal., 2008;Jafaroghlietal.,2010).Consequently,itislogicaltosuggest thatthetraitstobeincludedinthesheeprecordingsystemcould mainlybethose measuredearlyinlifeofthelambs, e.g.before weaning.
ExceptthematernalgeneticcorrelationbetweenBWandW30d thatwasestimatedto0.66,thosebetweentheothertraitswerenot estimatedbecausetheirbestmodelsdidnotincludethematernal geneticeffects.Thiscorrelationindicatesthatmaternaleffectsare partlyoriginatingfromtheprenatalperiodandcouldhavesome favourableeffectsonpost-natalgrowthtraitsofSardilambs.This estimateisinagreementwiththatofBahreiniBehzadietal.(2007) (0.68inKermanilambs)buthigherthanthatofMohammadietal.
(2010)(0.30inShallambs).
Phenotypiccorrelationsvariedfrom0.27and0.72amongbody weightsandvariedfrom−0.01and0.92betweenweightsandADG.
Thehighestphenotypiccorrelation(0.92)wasrecordedbetween W30dandADG1andthelowest(−0.01)wasfoundbetweenBW and ADG1.Theseestimates werecomparable tothosereported intheliterature(BahreiniBehzadi etal.,2007;Gamasaeeetal., 2010).Further,exceptcorrelationsbetweenW30dandADG1and
betweenW30dandADG2,theotherestimateswerelowerthan theircorrespondinggeneticcorrelationestimates.
Environmentalcorrelations amonggrowthtraitsvariedfrom
−0.04betweenBWandADG1to0.93betweenW30dandADG1.The estimatesofenvironmentalcorrelationsamonginvestigatedtraits werepositive(excepttheonebetweenBWandADG1)andgener- allylowerthanthoseofgeneticcorrelations.BahreiniBehzadietal.
(2007)andAbbasietal.(2012)reportedthattheenvironmental correlationsamonggrowthtraitswereallpositive.
3.4. Estimationofgeneticgain
Theannualgenetictrendfluctuatedthroughoutthestudyperiod (Figs. 1and2).Estimatesofdirectandmaternalannualgenetic trend obtained under univariate analyses for growth traits are showninTable6.Theannualdirectgenetictrendswerenegative andclosetozeroforallgrowthtraits.Estimateofdirectgenetic trendforBW (−1.12g/year)wasgenerallysimilartothelowest valuesreportedbyZishirietal.(2010)inSouthAfricanDormer breed(−2.1g/year)andbyDietal.(2014)inChinesesuperfine Merinosheep(−1g/year).However,itwaslowerthanestimates reportedinvariousothersheepbreeds(0g/yearinIranianBaluchi breed,GholizadehandGhafouri-Kesbi, 2015; 2g/yearin Iranian Kermani,Mokhtariand Rashidi,2010; 5g/yearin SouthAfrican DohneMerino,KlerkandHeydenrych,1990;21g/yearinIranian Afsharisheep,Ghafouri-Kesbietal.,2009).
TheestimateofdirectgenetictrendforW30dwas0.17g/year (P>0.05).Kariukietal.(2010)reportedanestimateof107g/year inDorpersheep.ForW90d,ADG1andADG2,directgenetictrends werenegativeandsignificant(P<0.01)and estimatedto−2.65,
−0.01 and −0.03g/year, repectively. While there are very few reportsdealingwithgenetictrendestimatesforADGinliterature, estimatesofdirect genetictrendforweaning weightinvarious sheep breeds are numerous. Bosso et al. (2007) and Ghafouri- Kesbietal.(2009)reportedthatthegenetictrendforpre-weaning growthrate(0–4months)increasedby0.01and0.001kgperyear, respectively.Thegenetictrendestimatesforweaningweightwere 7g/yearinBaluchibreed(GholizadehandGhafouri-Kesbi,2015), 40.6g/yearinDormerbreed(Zishirietal.,2010),96g/yearinDor- persheep(Kariukietal.,2010),128g/yearinKurdisheep(Rashidi andAkheshi,2007)and167ginAfsharisheep(Ghafouri-Kesbietal., 2009).
Further,maternalgenetictrendwaspositiveandhighlysignif- icant(P<0.001)andwas1.08g/yearforBWandnegativeandnot significant(P>0.05)andwas−0.26g/yearforW30d.Ahighervalue formaternalgeneticresponse(3g/year)wasreportedbyMokhtari andRashidi(2010)inIranianKermanibreed.Also,maternalgenetic trendestimateforBWwashigherthanthedirectone.Thismaybe duetomaternaleffectsinSardisheepthathavemorepronounced effectonBWthanthedirectadditivegeneticeffects.
Thelowandirrelevantrealizedannualgenetictrendsforgrowth traitsinthepresentstudycouldbeattributedprincipallytothe lowheritabilityestimatesandtothepoorselectionpractisedthat largelyreliesonphenotypiccharacteristicsinsteadofonanimals’
additivegeneticvalues.Inthisfarm,siresareselectedeveryyear atabout18month’sagebasedonconformation,weightandcon- formitytoexternalappearanceofbreed,butnotontheirgenetic merit,andyetselectionofanimalswithhigherphenotypicvalues doesn’tguaranteetheselectionofanimalswithhigherbreeding values.Thelowannualgeneticgainsobtainedinthecurrentstudy mightbealsoduetotheharshrearingenvironmentofSardisheep thatisnotverysuitablefortheexteriorisationoflambs’genetic potential.Infact,Sardisheepareraisedinasemi-aridzonechar- acterizedbyahightemperatureinsummerandpoorqualityand quantitypastures.
4. Conclusion
Thepresentstudycontributestothecomparison ofdifferent models for estimation of (co)variancecomponents and genetic parameters in Sardi sheep.Although only BW and W30d were affectedbymaternaladditivegeneticeffects,theresultsconfirmed thenecessityofincludingmaternaleffectsinmodelsappliedfor geneticevaluationofearlygrowthtraitsofSardisheep.Thelowher- itabilityestimatesobtainedforinvestigatedtraitsinthisresearch suggestthatmassselectionbasedonthesetraitsmayresultinslow geneticprogressforgrowthtraits.Thedifferenceinestimatesof geneticparametersdeterminedindifferentmodelsshowedthat modelchoiceisanimportantaspectforobtainingaccuratepredic- tionsofbreedingvaluesthataregoingtobeusedwhendeciding onabreedingprogramme.Finally,duetotheexistenceofgenetic variationforgrowthtraitsandgenerallypositiveandmediumto highgeneticcorrelationsamongtheinvestigatedtraits,itcanbe concludedthatimprovementofgrowthtraitsofSardisheepseems feasibleinselectionprogrammes.Nevertheless,itisrecommended toimprovethemanagementofflockinordertoreducetheenviron- mentalvarianceandtoincreaseheritabilityestimates,andalsoto selectreplacementanimalsbasedontheirgeneticmeritbutnoton theirrawperformancesbecausewithoutaselectiononthegenetic merit,nogeneticimprovementwillbeexpected.
Acknowledgment
Theco-authorwouldliketothankthePROGEBE-GUINEAProject forfinancinghisMasterstudies.
References
Abbasi,M.A.,Abdollahi-Arpanahi,R.,Maghsoudi,A.,VaezTorshizi,A., Nejati-Javaremi,A.,2012.Evaluationofmodelsforestimationofgenetic parametersandmaternaleffectsforearlygrowthtraitsofIranianBaluchi sheep.SmallRumin.Res.104,62–69.
Abegaz,S.,Negussie,E.,Duguma,G.,Rege,J.E.O.,2002.Geneticparameter estimatesforgrowthtraitsinHorrosheep.J.Anim.Breed.Genet.119,35–45.
Abegaz,S.,vanWyk,J.B.,Olivier,J.J.,2005.Modelcomparisonsandgeneticand environmentalparameterestimatesofgrowthandtheKleiberratioinHorro sheep.S.Afr.J.Anim.Sci.35(1),30–40.
Akaike,H.,1973.Informationtheoryandanextensionofthemaximumlikelihood principle.In:Petrov,B.N.,Csaki,F.(Eds.),Proc.2ndInt.Symp.Information Theory.AkademiaiKiado,Budapest,Hungary,pp.267–281.
BahreiniBehzadi,M.R.,Shahroudi,F.E.,VanVleck,L.D.,2007.Estimatesofgenetic parametersforgrowthtraitsinKermanisheep.J.Anim.Breed.Genet.124, 296–301.
Boldman,K.G.,Kriese,L.A.,VanVleck,L.D.,VanTassell,C.P.,Kachman,S.D.,1995.A ManualforUseofMTDFREML.ASetofProgramstoObtainEstimatesof VariancesandCovariances.USDA/ARS,Washington,DC,USA(Draft).
Bosso,N.A.,Cissé,M.F.,vanderWaaij,E.H.,Falla,A.,vanArendonk,J.A.M.,2007.
GeneticandphenotypicparametersofbodyweightinWestAfricanDwarfgoat andDjallonkésheep.SmallRumin.Res.67,271–278.
Boujenane,I.,Kansari,J.,2002.Estimatesof(co)varianceduetodirectand maternaleffectsforbodyweightsinTimahditesheep.Anim.Sci.74,409–414.
Boujenane,I.,M’zian,S.,Sadik,M.,2001.Estimationdesparamètresgénétiqueset phénotypiquesdelacroissancedesovinsderaceSardi.ActesInst.Agron.Vet.
(Maroc.)21(3),177–183.
Boujenane,I.,Chikhi,A.,Ibnelbachyr,M.,Mouh,F.Z.,2015.Estimationofgenetic parametersandmaternaleffectsforbodyweightatdifferentagesinD’man sheep.SmallRumin.Res.130,27–35.
Boujenane,I.,2005.SmallruminantbreedsofMorocco.In:Iniguez,L.(Ed.), CharacterizationofSmallRuminantBreedsinWestAsiaandNorthAfrica,vol.
2.NorthAfrica.InternationalCenterforAgriculturalResearchintheDryAreas (ICARDA),Aleppo,Syria,pp.5–54.
Cundiff,L.V.,1972.Theroleofmaternaleffectsinanimalbreeding:VIII.
Comparativeaspectsofmaternaleffects.J.Anim.Sci.35,1335–1337.
Di,J.,Ainiwaer,L.,Xu,X.,Zhang,Y.,Yu,L.,Li,W.,2014.Genetictrendsforgrowth andwooltraitsofChinesesuperfineMerinosheepusingamulti-traitanimal model.SmallRumin.Res.117,47–51.
Ekiz,B.,Ozcan,M.,Yilmaz,A.,2004.Estimatesofgeneticparametersfordirectand maternaleffectswithsixdifferentmodelsonbirthandweaningweightsof TurkishMerinolambs.Turk.J.Vet.Anim.Sci.28,383–389.
Eskandarinasab,M.,Ghafouri-Kesbi,F.,Abbasi,M.A.,2010.Differentmodelsfor evaluationofgrowthtraitsandKleiberratioinanexperimentalflockofIranian fat-tailedAfsharisheep.J.Anim.Breed.Genet.127,26–33.