Tc

SIa

b

a

ARRAA

KSBB

1

f

MH

0d

Animal Reproduction Science 124 (2011) 28–38

Contents lists available at ScienceDirect

Animal Reproduction Science

journal homepage: www.elsevier.com/locate/anireprosci

ime trends, environmental factors and genetic basis of semen traitsollected in Holstein bulls under commercial conditions

ofiene Karouia, Clara Díaza, Magdalena Serranoa, Roger Cuea,1,doia Celorriob, María J. Carabanoa,∗

Animal Breeding Department, National Institute for Agriculture and Food Research and Technology, Ctra. de La Coruna Km 7.5, 28040 Madrid, SpainABEREKIN, S.A. Parque Tecnológico, Edificio 600, 48160 Derio, Spain

r t i c l e i n f o

rticle history:eceived 30 July 2010eceived in revised form 25 January 2011ccepted 7 February 2011vailable online 13 February 2011

eywords:emen qualityreeding soundness evaluationayesian analysis

a b s t r a c t

The fact that results of artificial insemination (AI) are declining in highly selected dairycattle populations has added a renewed interest to the evaluation of male fertility. Datafrom 42,348 ejaculates collected from 1990 to 2007 on 502 Holstein bulls were analysed ina Bayesian framework to provide estimates of the evolution of semen traits routinely col-lected in AI centres throughout the last decades of intense selection for production traits andestimate genetic parameters. The traits under consideration were volume (VOL), concen-tration (CONC), number of spermatozoa per ejaculate (NESPZ), mass motility score (MM),individual motility (IM), and post-thawing motility (PTM). The environmental factors stud-ied were year-season and week of collection, which account for changes in environmentaland technical conditions along time, age at collection, ejaculate order, time from previ-ous collection (TPC) and time between collection and freezing (TCF) (only for PTM). Bull’sinbreeding coefficient (Fi), bull’s permanent environmental and additive genetic effectswere also considered. The use of reduced models was evaluated using the Bayes factor.For all the systematic effects tested, strong or very strong evidence in favour of includ-ing the effect in the model was obtained, except for Fi for motility traits and TCF forPTM. No systematic time trends for environment or bull effects were observed, except forPTM, which showed an increasing environmental trend, associated with improvements in

freezing–thawing protocols. Heritability estimates were moderate (0.16–0.22), except forIM, which presented a low value (0.07). Genetic correlations among motilities and betweenmotilities and CONC were large and positive [0.38–0.87], VOL showed a negative correlationwith CONC (−0.13) but with ample HPD95%. The magnitude of heritabilities would allow

n if reqt of bu

an efficient selectioviability componen

. Introduction

Bull fertility has been a concern for the dairy industryor decades. AI centres rely on the ability of their bulls to

∗ Corresponding author. Tel.: +34 913476742; fax: +34 913478743.E-mail address: mjc@inia.es (M.J. Carabano).

1 Present address: Department of Animal Science, McGill University,acdonald Campus, 21111 Lakeshore Road Ste. Anne de Bellevue, QC,9X 3V9, Canada.

378-4320/$ – see front matter © 2011 Elsevier B.V. All rights reserved.oi:10.1016/j.anireprosci.2011.02.008

uired and grants the use of these traits as indicators of the spermlls breeding soundness.

© 2011 Elsevier B.V. All rights reserved.

produce a sufficient amount of semen with a good fertil-izing potential. Culling of high merit bulls due to impairedfertility may result in important economic losses for thewhole dairy industry. More recently, the fact that resultsof AI are declining in highly selected dairy cattle popula-tions (see, e.g., Lucy, 2001) has added a renewed interest

to the evaluation of male fertility. Consequences of theintense selection for production traits on female fertil-ity have been the subject of numerous studies. However,to the authors’ knowledge, no results about trends fortraits affecting the fertilizing capacity of the sperm or,

roductio

S. Karoui et al. / Animal Rep

more importantly, for male fertility in the last decades areavailable.

Bull breeding soundness evaluation is usually carriedout based on semen traits routinely collected in the AI cen-tres. These traits have been associated with bull fertility,but are affected by environmental factors that can bias eval-uation of the bull’s merit. Optimising the statistical modelthat allows a better appraisal of the bull’s own merit infield conditions is important. Numerous studies have deter-mined the effect of several environmental factors suchas age of bull, ejaculate order, season, collection team orinterval between collections on semen traits (Everett et al.,1978; Fuente et al., 1984; Taylor et al., 1985; Diarra et al.,1997; Mathevon et al., 1998; Brito et al., 2002a,b; Fuerst-Waltl et al., 2006; Hallap et al., 2006; Gredler et al., 2007;Druet et al., 2009; Koivisto et al., 2009; Fiaz et al., 2010;Mandal et al., 2010). However, many of those studies havebeen carried out in a relatively low number of animalsand under experimental conditions. Moreover, additionaleffects might need to be considered. In this sense, thehighly oscillating nature of commercial semen productionrequires identifying short term factors affecting produc-tion and quality around the collection day. In addition, theincreasing rate of inbreeding in Holstein populations war-rants its consideration for the statistical modelling.

The effectiveness of using semen traits to indirectlyimprove bull breeding soundness and the emphasis tobe placed on each of them depends on the heritability,repeatability and correlations among them and betweeneach seminal trait and the selection criteria. Estimates ofheritability of semen traits in cattle are widely variable,ranging from low to moderate, depending on the trait,population studied, age of the bull and whether individ-ual ejaculate measures or average production per bull areanalysed. A significant relationship between bull fertilitymeasured through results of AI and semen traits (morespecifically sperm motility) has been found in several stud-ies (see, e.g., Christensen et al., 1999, 2005; Januskauskaset al., 2000).

This study aimed to evaluate the statistical modellingincluding a short term contemporary group and inbreed-ing level, investigate time trends and to study geneticparameters for semen traits of Holstein bulls collected incommercial conditions in an AI centre from 1990 to 2007.Bayesian methods, which provide a flexible frameworkto make inferences about the parameters describing thedata generation process and functions of them and a com-prehensive set of tools to describe the uncertainty in theestimation of the quantities of interest and to comparealternative models for the analysis of noisy measures andunbalanced designs that characterize field data, were usedin this study.

2. Materials and methods

2.1. Data

Data on five semen traits, volume of ejaculate (VOL),concentration (CON), mass motility score (MM), individ-ual motility (IM) and post thawing progressive individualmotility (PTM), which were routinely collected in a Spanish

n Science 124 (2011) 28–38 29

AI centre (Aberekin, S.A.) from 1990 through 2007, wereprovided by this AI centre. Number of spermatozoa perejaculate (NESPZ) was then obtained from the product ofVOL and CONC. Only records from Holstein bulls were anal-ysed.

In this AI centre, semen is routinely collected using anartificial vagina in a nearly regular weekly basis. Ejaculatevolume (milliliters) is measured directly from the collect-ing container. During the semen analysis, collected samplesare kept in a 32 ◦C bath. Concentration is determined usingan IMV ACCEULL spectrophotometer (IMV InternationalCorporation, Maple Grove, MN, USA). Percent progressivemotility, which has been named as individual motility, isestimated by examining unstained diluted semen usinga 20× magnification and light microscopical observationequipped with a warm stage. Mass motility or motilityscore is subjectively assessed by a trained technician forundiluted unstained semen under microscope (10×) usinga scale from 0 to 5 (best motility).

The ejaculates containing less than 300 × 106 spz/ml,4 MM score and 75% of IM are discarded and notincluded in the pooled collections for freezing. The cut-off points are a ‘rule of thumb’ and not a strict guideline.Individual ejaculates are diluted with egg-yolk-tris exten-der containing glycerol and antibiotics (Bioxcell®, IMVTechnologies). Final concentration of diluted samples is120 × 106 spz/ml.

After gradually cooling to 5 ◦C (at a speed of0.25–0.30 ◦C/min), the semen is packaged in 0.25 cc straws(30 million spz per straw) and frozen using a programmablebio-freezer (model 5300 3T, IMV Technologies). For post-thaw semen evaluation, two straws per ejaculate arethawed in a water bath of 36 ◦C and evaluated individuallyfor the percentage of progressively motile spermatozoa,using a computerized Sperm Class Analyzer® (Microptic,S.L.) from year 2004.

Some changes in the routine collection and laboratoryequipment have occurred along the period of collectionof data. Up to year 1997, freezing was done using atraditional vapour method. From that date, a computer-ized programmable freezer has been used, as previouslyreported. The spectrophotometer used to measure CONwas changed in years 1997 and 2006. The cooling systemprevious to freezing was improved in year 2003. The exten-der used to dilute the semen samples was changed fromTriladyl® (minitüb) to the present one in year 2001. For allthe equipment and protocols an improvement in repeata-bility and accuracy of measurements has been observedfrom year 2000, when the AI centre obtained the ISO certi-fication.

For analyses, records received from the AI centre wereedited according to the following criteria. All the recordscorresponding to ejaculates below the threshold of 1 mlper ejaculate or 300 million spz/ml or above the thresh-old of 20 ml and 4000 million spz/ml were discarded. Alsothe registries of bulls with age of collection less than 12

months and all the data of bulls with less than 5 observa-tions were eliminated. Discarded records represented 4% ofthe initial data. Finally, 42,348 ejaculates corresponding to502 bulls were analysed. A summary of the data statisticsis presented in Table 1.

30 S. Karoui et al. / Animal Reproduction Science 124 (2011) 28–38

Table 1Summary statistics for number of records per bull and for the measures of traits used in the study.

VOLa CONCa NESPZa MMa IMa PTMa

Overall mean 5.49 1267.02 6.90 4.20 85.18 48.83[Minimum–maximum] [1–20] [300–3815] [0.3–46.3] [1–5] [0.99] [3–95]Standard deviation 2.61 521.78 4.25 0.76 10.57 12.57Total number of records 42,348 41,212 41,212 42,026 42,244 23,793Number of bulls 501 498 498 501 501 497Number of records/bull

Mean 84.53 82.75 82.75 83.88 84.31 47.87

ermatoM g motili

2

w3hocfitbE

4fw

2

ea

y

wl(otccawm1ed1dwTwEa

Standard deviation 84.83 84.55[Minimum–maximum] [6–761] [4–739]

a VOL = volume of ejaculate (ml); CONC = concentration (millions of spM = mass motility (1–5); IM = individual motility (%); PTM = post thawin

.2. Pedigree information

For the 502 bulls with records in the data file, ancestorsere traced back as many generations as possible (up to

7 generations maximum). On average, bulls with recordsad five complete generations with 4 and 11 bulls withnly one and two complete generations, respectively. Theomplete pedigree file used to estimate inbreeding coef-cients (Fi) included 8645 animals. Fi for each animal inhe pedigree was calculated using the algorithm describedy Meuwissen and Luo (1992) implemented in programNDOG (Gutiérrez and Goyache, 2005).

The 502 bulls with records were sired by 173 sires and20 dams, and were grouped in 26, 96 and 80 families ofull, paternal and maternal half sibs, respectively. Ten bullsith records had also progeny with data.

.3. Statistical analyses

Statistical analyses were carried out under linear mod-ls and Bayesian methods. The full model (FM) considereds a starting point for alternative reduced models was:

ijklmrost = YSi + ACj + TPCk + ENl

+ Fim + WCr + ao + pes + eijklmrost (1)

here yijklmrost is the semen trait recorded in one ejacu-ate (VOL, CONC, NESPZ, MM, IM) or from both ejaculatesPTM) in the collection day; YSi, i = 1 . . . 72, was the effectf year-season of collection. A total of 18 years of collec-ion and four seasons, defined as 3 months periods, wereonsidered. ACj, j = 1 . . . 98, was the effect of age of bull atollection. Age classes were defined by month up to classges above 103 months for which two or more monthsere grouped to have enough records for an accurate esti-ation. Minimum and maximum ages at collection were

2 and 138 months, respectively. TPCk, k = 1 . . . 9, was theffect of the time from previous collection. Class 1 wasefined by the interval [1–3] days, classes 2–5 representedday increment, class 6 was defined by the interval [8–13]ays; class 7 was [14–21], class 8 was [22–60] and class 9

as assigned to records with a TPC greater than 60 days.

he mode was 7 days and larger intervals were associatedith young bulls waiting for the results of progeny test.

Nl, l = 1, 2, was the ejaculate order. This effect did notffect PTM, given that both ejaculates in one collection

84.55 84.71 84.86 56.47[4–739] [6–754] [6–759] [3–687]

zoa/ml); NESPZ = number of spermatozoa per ejaculate (×103 millions);ty (%).

date are pooled together before freezing. Fim, m = 1 . . . 7,was the effect of the inbreeding coefficient. Fi classes weredefined by increments of 2.5% of the level of inbreeding. Thefirst class was defined as Fi < 1% and the last as Fi ≥ 12.5%.TCFn, n = 1 . . . 5 was the effect of time between collectionand freezing. This factor only affects PTM and classes weredefined by 1 h increments starting in 5 h. The last classwas TCF > 9 h. WCr, r = 1 . . . 934 was the short term effectof week of collection. This effect accounts for managementor other environmental factors affecting collections withina 1 week period. ao, o = 1 . . . 8645 was the additive geneticeffect. pes, s = 1 . . . 502 was the permanent environmen-tal and non additive genetic effects. This effect representsenvironmental factors and non additive effects affecting allmeasures of the same bull. eijklmrost is the residual effect.

In Bayesian inference, all unobserved quantities areassumed to possess distributional properties and thereforeare treated as random variables. However, one can assigndifferent prior knowledge to different effects. In our case,vague, flat priors, which assign equal probability to all lev-els of each factor and are indicative of a vague knowledgeof the a priori distribution were chosen for YS, AC, TPC,EN and Fi. Gaussian priors with associated null means andunknown variances, which were also estimated in the anal-yses, were assigned to WC, additive genetic and permanentenvironmental effects. Gaussian priors are informative pri-ors that shrink the estimates of the unknown parameterstowards the prior mean when the amount of informationin the data is scarce. A normal likelihood was assumed forthe observed data. Inference about unknown parameterswas obtained from their marginal posterior distributionsvia Gibbs sampling algorithm, which produces an empir-ical estimate of the posterior distribution of interest byiteratively sampling from full conditional distributions.

Two sets of analyses were carried out. In the first one,uni-variate analyses were carried out to determine thebest way of modelling semen traits individually. The Bayesfactor (BF) was estimated to choose between the FM andalternative reduced models excluding one factor at a timeand a model that considered a flat prior for WC instead ofthe Gaussian informative prior used in FM. The BF is thequotient between the posterior probabilities of two alter-

native models and can be interpreted as a summary of theevidence provided by the data in favour of the referencemodel (FM in our case) or against the alternative model(reduced models and the model with flat prior for WC).Kass and Raftery (1995) scale on the natural logarithm of

S. Karoui et al. / Animal Reproduction Science 124 (2011) 28–38 31

Table 2Logarithm of the Bayes factor (LN(BF)) and evidence (Evid.) against the model removing the corresponding factor compared with the full model (FM) forthe analysed traits.a

Factorremovedd

VOLb CONCb NESPZb MMb MIb PTMb

Ln(BF)a Evid.c Ln(BF) a Evid.c Ln(BF)a Evid.c Ln(BF)a Evid.c Ln(BF)a Evid.c Ln(BF)a Evid.c

YS 160.97 VS 87.5 VS 117.56 VS 68.43 VS 64.76 VS 43.64 VSWC 138.38 VS 662.68 VS 313.44 VS 734.35 VS 444.67 VS 298.95 VSYS + WC 651.85 VS 2903.53 VS 1761.13 VS 1947.95 VS 941.33 VS 661.74 VSAC 762.83 VS 626.82 VS 1083.55 VS 699.76 VS 237.94 VS 268.64 VSTPC 139.25 VS 47.03 VS 94.85 VS 51.84 VS 42.93 VS 7.03 VSEN 572.76 VS 4904.04 VS 4111.84 VS 207.37 VS 13.38 VS – –TCF – – – – – – – – – – −0.68 NFi 4.47 S 12.43 VS 5,53 S 0.48 B −3.25 N 0.98 BYS (WCf) −13.43 N −13.1 N −13.19 N −14.38 N −20.91 N 14.31 VS

a Positive values of the Ln(BF) indicate that the FM is preferred over the alternative model and vice versa.b VOL: volume of ejaculate; CONC: concentration; NESPZ: number of spermatozoa per ejaculate; MM: mass motility; IM: individual motility; PTM:

y strongction; TPear-seas

post-thawing motility.c N: negative; B: not worth more than a bare mention S: strong VS: verd YS: year-season of collection; WC: week of collection; AC: age at colle

collection and freezing; Fi: inbreeding level; YS(WCf): model excluding y

the BF was used to assess the strength of the evidence infavour of the FM. For these uni-variate analyses, softwarepreviously developed by the authors was used. More detailson the computing strategy can be found in López-Romeroet al. (2003).

In a second analysis, a multi trait approach usingthe model selected from the uni-variate analyses andconsidering all six semen traits jointly was carried out.Freely available software TM (available upon request tothe authors: Andres.legarra@toulouse.inra.fr) was used tocarry out the multi-trait analyses.

In the uni-variate analyses, single chains with 500,000

samples, discarding the first 100,000 as burn-in were usedfor inference. For the multiple trait analysis, more roundsof the Gibbs sampling scheme were required to reach con-vergence, particularly for the genetic covariances. Finally,

VOLUME (ml)

0

0.5

1

1.5

2

2.5

3

3.5

4

03/0

1/1

990

03/0

1/1

991

03/0

1/1

992

03/0

1/1

993

03/0

1/1

994

03/0

1/1

995

03/0

1/1

996

03/0

1/1

997

03/0

1/1

998

03/0

1/1

999

03/0

1/2

000

03/0

1/2

001

03/0

1/2

002

03/0

1/2

003

03/0

1/2

004

03/0

1/2

005

03/0

1/2

006

03/0

1/2

007

Date of collection

Es

tim

ate

d v

alu

e

1

1

1

1

2

2

Es

tim

ate

d v

alu

e

MASS MOTILITY (1-5)

2.5

3

3.5

4

4.5

5

03/0

1/1

990

03/0

1/1

991

03/0

1/1

992

03/0

1/1

993

03/0

1/1

994

03/0

1/1

995

03/0

1/1

996

03/0

1/1

997

03/0

1/1

998

03/0

1/1

999

03/0

1/2

000

03/0

1/2

001

03/0

1/2

002

03/0

1/2

003

03/0

1/2

004

03/0

1/2

005

03/0

1/2

006

03/0

1/2

007

Date of collection

Esti

mate

d v

alu

e

Esti

mate

d v

alu

e

Fig. 1. Estimated effect for date of collect obtained from posterior means f

.C: time to previous collection; EN: ejaculate number; TCF: time betweenon of collection and including WC with a flat prior.

5,000,000 total samples with 3,000,0000 as burn-in wereobtained.

Trends for the traits under consideration by bull’s year ofbirth were obtained from posterior means for bull effects(adding genetic and permanent environmental solutionsfrom the uni-variate analyses) to investigate the evolutionof bulls’ breeding soundness over the last decades. Giventhe fact that obtaining unbiased estimates of genetic trendswould require information on the selection criteria, whichwas not available, the sum of both additive genetic and per-manent environmental effects was used to estimate trendsalong time. The sum of these components, what we have

called bull effect, is equivalent to the average phenotypicvalue of a bull for each trait adjusted by the systematiceffects included in the model and it is expected to be lessbiased by selection than the bull additive genetic compo-

CONCENTRATION (106 sperm/ml)

500

750

000

250

500

750

000

250

03/0

1/1

990

03/0

1/1

991

03/0

1/1

992

03/0

1/1

993

03/0

1/1

994

03/0

1/1

995

03/0

1/1

996

03/0

1/1

997

03/0

1/1

998

03/0

1/1

999

03/0

1/2

000

03/0

1/2

001

03/0

1/2

002

03/0

1/2

003

03/0

1/2

004

03/0

1/2

005

03/0

1/2

006

03/0

1/2

007

Date of collection

POST THAWING MOTILITY (%)

45

50

55

60

65

70

75

03/0

1/1

990

03/0

1/1

991

03/0

1/1

992

03/0

1/1

993

03/0

1/1

994

03/0

1/1

995

03/0

1/1

996

03/0

1/1

997

03/0

1/1

998

03/0

1/1

999

03/0

1/2

000

03/0

1/2

001

03/0

1/2

002

03/0

1/2

003

03/0

1/2

004

03/0

1/2

005

03/0

1/2

006

03/0

1/2

007

Date of collection

or year season and year week effects on the analysed semen traits.

32 S. Karoui et al. / Animal Reproduction Science 124 (2011) 28–38

Table 3Posterior means (percentage over phenotypic means in parenthesis) and 95% high probability density region (in brackets in lower row) of contrasts forselected levels of age at collection (AC), time to previous collection (TPC), ejaculate number (EN) and inbreeding level (Fi).

Effect AC (months) TPC (day) EN Fi (%)

Contrast 12–24 24–36 36–>120 [1,3]–>60 1–2 <1–[10.0, 12.5]

VOLa −1.47 (−26.87)[−1.73, −1.23]

−0.88 (−16.12)[−1.13, −0.62]

−1.59 (−29.04)[−2.09, −1.12]

−0.82 (−15.02)[−0.98, −0.68]

0.58 (10.48)[0.54, 0.61]

0.59 (10.50)[−0.28, 1.48]

CONCa −440.59 (−34.77)[−496.71, −386.30]

−90.02 (−7.11)[−147.97, −33.90]

209.89 (16.56)[105.01, 313.82]

−26.23 (−2.07)[−61.11, 5.99]

383.71 (30.28)[376.51, 391.27]

68.91 (5.44)[−135.28, 269.34]

NESPZa −3.46 (−50.21)[−3.87, −2.97]

−1.71 (−24.83)[−2.16, −1.23]

0.46 (6.62)[−0.39, 1.27]

−1.08 (−15.69)[−1.36, −0.82]

2.84 (41.27)[2.79, 2.90]

1.30 (18.81)[0.01, 2.68]

MMa −0.73 (0.04)[−0.82, −0.65] −17.59

−0.16 (−3.78)[−0.24, −0.07]

0.57 (13.7)[0.41, 0.74]

0.15 (3.61)[0.10, 0.20]

0.11 (2.62)[0.10, 0.12]

0.23 (5.54)[−0.08, 0.54]

MIa −7.19 (−8.45)[−8.50, −5.87]

−1.03 (−1.21)[−2.30, 0.38]

4.94 (5.80)[2.28, 7.46]

1.97 (2.31)[1.16, 2.78]

0.51 (0.59)[0.33, 0.67]

4.50 (5.27)[−1.30, 9.80]

a 09).80]

permatM hawing

nce

3

3

dTefl

tvittFmtsc

ttWiu

aemisaetofc

PTM −6.7 (−13.77)[−8.71, −4.98]

−0.98 (−2.01)[−2.94, 1.03]

14.20 (29.[10.76, 17

a VOL = volume of ejaculate (ml); CONC = concentration (millions of sM = mass motility score (1–5); IM = individual motility (%); PTM = post t

ent. In addition, raw phenotypic trends were obtained andompared with the previously estimated trends for bullffects.

. Results

.1. Model selection

The logarithm of the BF and the strength of the evi-ence against alternative models are shown in Table 2.he model excluding YS and using a flat prior for the WCffect was named YS (WCf), with the letter f standing forat prior.

For all traits, all factors considered, except Fi for motilityraits (MM, IM and PTM) and TCF for PTM, showed strong orery strong evidence against the simplified models, indicat-ng that the omitted factor/s had a significant contributiono the posterior probability of the FM. The effects showinghe largest contribution to the posterior probability of theM were AC for VOL, EN for CONC and NESPZ and WC for theotility traits (MM, IM and PTM). On the other hand, from

he factors showing strong evidence favouring their inclu-ion in the model, TPC was the factor showing the smallestontribution.

Except for PTM (which has around half of the observa-ions than the other traits because it results from mixinghe two ejaculates in a collection day), using a flat prior for

C and excluding the YS effect (model YS (WCf)) resultedn a slightly larger posterior probability of the model thansing a Gaussian prior as in FM.

Fi had a slightly negative or bare contribution to the FMnd TCF showed a slightly negative contribution for mod-lling PTM. These results indicate that inbreeding was not aajor factor to be considered when analysing semen traits

n this population. It is important to note that the figureshown for Ln(BF) are approximations obtained in an iter-tive process and not the exact values (see López-Romero

t al., 2003 for details). Therefore, around zero values forhe approximated differences in the posterior probabilityf the models could have a different sign than the exact dif-erence. In these cases, a strong conclusion on model choiceannot be drawn.

2.53 (5.19)[1.45, 3.79]

– 5.55 (11.37)[−1.01, 11.72]

ozoa/ml); NESPZ = number of spermatozoa per ejaculate (103millions);motility (%).

3.2. Management and environmental effect

Estimates of management and environmental effectsalong collection years are shown in Fig. 1. The values ofthe management effects have been obtained by adding thesolutions (posterior means in the Bayesian context) forthe effects of YS and WC corresponding to each date ofcollection. The YS effect represents the average mediumterm trend in semen traits and the WC effect is a shortterm deviation from that average trend due to effects thataffect these traits around the collection day. No system-atic trend was observed for any of the traits, except forPTM, which showed an increasing trend in the studiedperiod. The largest studentised range (difference betweenlargest and smallest estimated value for the week of col-lection effect over the standard deviation of the estimatedvalues) was obtained for CONC (range = 1095.8 million ofsperm/ml, st range = 9.3), followed by PTM (range = 31.4%,st range = 4.0). Average of estimates of YS plus WC solu-tions corresponding to 3 month seasons did not showevidence of any seasonal trend for the traits analysed.

3.3. Other environmental effects

Table 3 shows posterior means and 95% high probabil-ity density regions (95HPD) of differences between levelsof interest for environmental factors other than the effectsused to calculate environmental trends, and for Fi. Differ-ences between levels as percentage of the phenotypic mean(in Table 1) of the trait are also shown. AC was the effectthat showed the largest impact on all semen traits anal-ysed, followed by EN, concordant with the results from theLn(BF) in Table 2.

Fig. 2 shows the posterior means for AC classes. VOLincreased with age for the whole age range analysed. How-ever, a larger increase was observed during the first 3 yearsof age. VOL became more stable at around 5 years of age.

CONC and motility traits showed a different pattern witha steep increment up to the second year of age, a plateaufrom 3 to 5 years of age and a slow decline afterwards. Asharp decline was observed for the last age class, composedof animals with more than 10 years of age. The first phase

S. Karoui et al. / Animal Reproduction Science 124 (2011) 28–38 33

VOLUME (ml)

4

6

8

10

12

97857361493725131

Age class

Esti

mate

d v

alu

e

CONCENTRATION (106 sperm/ml)

-500

-400

-300

-200

-100

0

100

200

97857361493725131

Age class

Esti

mate

d v

alu

e

MASS MOTILITY (1-5)

1.8

2.1

2.4

2.7

3

97

Esti

mate

d v

alu

e

POST THAWING MOTILITY (%)

5

10

15

20

Esti

mate

d v

alu

eect age

number of observations (835 observations from 8 bulls)and yielded unexpected results for those traits. Therefore,not enough information was present in this data set toestimate the effect of high levels of inbreeding on sementraits.

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

Fi

1.5

857361493725131

Age class

Fig. 2. Posterior means for levels of the eff

corresponds to the sexual development of the individualsand the second to stability of the production due to sex-ual maturity and the acquisition of the reflex to producesemen. Overall, traits measuring semen output, VOL, CONCand NESPZ, were more affected than motility traits by AC,according with the differences in the traits at different agesexpressed as percentage of the phenotypic mean shown inTable 3.

The effect of EN generally affected all semen traits neg-atively, i.e., second ejaculates were worse in quantity andquality of sperm, but not with the same importance for all ofthem. CONC and NESPZ were the characters most affectedby this effect. The difference between estimated values forfirst and second ejaculate was 383.7 millions of spermato-zoa/ml. The difference between first and second ejaculatefor VOL was estimated at 0.58 ml. Motilities were nearlynot affected by EN.

The effect of TPC on semen traits was much less notori-ous than the other environmental effects. VOL and NESPZwere the traits more affected by TPC. VOL showed a trendto increase as TPC increased, but the estimated differencebetween extreme levels (1–3 days and more than 60 days)was only 0.8 ml. For CONC, an optimum for this effect wasobserved for an interval between 3 and 7 days, althoughthe difference between the maximum and minimum valuefor this effect was very small. For the motility traits, nearlyflat trends were observed apart from a slight decrease inmotility when the TPC was beyond 60 days, which wouldcorrespond to bulls that have produced the required stockof semen and are waiting for the results of the progenytest. Except for VOL and NESPZ, TPC was responsible fordifferences near or below 5% of the phenotypic mean ofthe trait.

3.4. Inbreeding

Average level of inbreeding in this population was5.51%, ranging from 0.22% to 16.9%, for the bulls that pro-

0

97857361493725131

Age class

at collection on the analysed semen traits.

duce data. The 70th percentile of the distribution had alimit threshold at 6.25%. Fig. 3 shows the average Fi forsuccessive years of birth of bulls with records, showingan increase over time, but, at different rates. A sharperincrease is observed between years 1984 and 1992. From1992 onwards, inbreeding increase slowed down. Overall,average rate of increase in Fi was 0.27% per year.

Despite the increasing inbreeding level in this popu-lation, inbreeding depression on semen traits could notbe considered important. Results in Table 2 indicate thata model excluding Fi might fit this data nearly as wellas the full model. In terms of changes in expected pro-duction, Fi seems to be responsible for changes rangingfrom 5.4% for CONC to 18.8% for NESPZ and 11.4% for PTMof the phenotypic mean, respectively (see Table 3). Theseexpected differences in values of the traits were obtainedfor the class of nearly non inbred animals and of animalswith inbreeding level between 10% and 12.5%. The classof animals with inbreeding level over 12.5% had a small

0.0

200620042002200019981996199419921990198819861984

Year of birth

Fig. 3. Average inbreeding coefficient per year of birth of the bulls withdata.

34 S. Karoui et al. / Animal Reproduction Science 124 (2011) 28–38

VOLUME (ml)

4

5

6

7

8

2006200520042003200220012000199919981997199619951994199319921991199019891988198719861985

Year of birth

CONCENTRATION (106 sperm/ml)

800

1000

1200

1400

1600

2006200520042003200220012000199919981997199619951994199319921991199019891988198719861985

Year of birth

MASS MOTILITY (1-5)

3.5

4

4.5

5

2006200520042003200220012000199919981997199619951994199319921991199019891988198719861985

Year of birth

POST THAWING MOTILITY (%)

40

45

50

55

2006200520042003200220012000199919981997199619951994199319921991199019891988198719861985

Year of birth

F en traitss sociated

3

aoattpasbrb

ebMftafg(irteo

3

rtE

ig. 4. Estimated trends by year of birth of the bulls for the analysed semolutions for the sum of genetic and permanent environmental effects as

.5. Bull effects

Posterior means for permanent environmental anddditive genetic effects associated with the bull were alsobtained from the statistical analyses: the sum of geneticnd permanent effect represents the ability of each bullo produce semen, both from the bull’s genotype for theseraits, and from the non genetic factors that affect semenroduction (quantity and quality) of a particular bull alongll its productive life. It could therefore be used as a mea-ure of bull breeding soundness, given that it quantifies theull’s expected sperm production free of the general envi-onmental factors affecting semen production and that cane recorded in field conditions.

Trends for bull effects obtained from raw data and fromstimated additive genetic and permanent environmentalull effects are shown in Fig. 4. Both trends were similar forM, IM but not for VOL, CONC and PTM. Trends obtained

rom estimated bull effects did not show any steady pat-ern for all traits while phenotypic raw estimates showeddeclining trend for VOL and a slightly increasing trend

or CONC and PTM. As previously mentioned, estimation ofenetic trends when information on the selection criteriamilk production in this case) is not present may be biasedf the selection criteria and the trait under analysis are cor-elated. Nevertheless, the genetic correlation between milkraits and semen characteristics, although unknown, is notxpected to be very large given the fact that these two typesf traits are, in principle, not physiologically related.

.6. Genetic parameter estimates

Table 4 shows the posterior features for heritability,epeatability, genetic and phenotypic correlations underhe full model for each trait in a multi-trait analysis.stimated heritabilities ranged from low (0.09) for IM to

obtained from raw averages (solid line) and from averages of estimatedwith each bull.

moderate (0.22) for VOL. Estimated repeatabilities, whichrepresents the correlation between measures of the sameanimal, and also the proportion of bull effects variation overthe total variance (obtained by adding genetic, permanentenvironmental and residual variance) ranged from 0.27 forNESPZ to 0.45 for PTM.

Estimated genetic correlations show that correlationsbetween VOL and CONC were negative and small withmotilities, while CONC and NESPZ somehow showed largerestimated correlations with motilities than VOL. The esti-mated correlation between NESPZ with quality traitsranged from 0.32 for PTM to 0.57 for CONC and it was 0.60with VOL. The largest correlations, 0.85, were estimatedbetween the motility parameters.

Estimates of correlations required a much largernumber of rounds of the Gibbs sampling process toachieve convergence. Moreover, some posterior correla-tions showed large 95HPD and a non null probability ofbeing zero. VOL had the largest 95HPD values and correla-tions among motilities the lowest.

4. Discussion

Statistical modelling is an important feature to obtainunbiased and precise predictions of bull fertility merit.According to the estimated Bayes factors and to thesolutions obtained in this study, the short and mediumterm management/environment effect (WC and YS) hadthe largest impact on semen production compared withthe other factors analysed. This effect accounts for man-agement practices (nutrition, bull handling, changes in

laboratory personnel over time, differences in the temper-ature of the artificial vagina used to collect the semen, thespecific procedures and equipment used in ejaculate freez-ing, changes in extenders and cryoprotectants, etc.) andenvironmental conditions (temperature, humidity, sani-

S. Karoui et al. / Animal Reproductio

Tab

le4

Post

erio

rm

ean

san

d95

%h

igh

pro

babi

lity

den

sity

regi

on(i

nbr

acke

ts)

ofh

erit

abil

itie

s/re

pet

abil

itie

s(d

iago

nal

),ge

net

ic(a

bove

dia

gon

al)

and

ph

enot

ypic

(bel

owd

iago

nal

)co

rrel

atio

ns.

Trai

taV

OL

CO

NC

NES

PZM

MIM

PTM

VO

L0.

22/0

.31

−0.1

30.

660.

130.

070.

13[0

.13,

0.32

]/[0

.28,

0.34

][−

0.47

,0.1

8][0

.47,

0.82

][−

0.22

,0.4

5][−

0.41

,0.5

3][−

0.24

,0.5

0]C

ON

C−0

.06

0.19

/0.3

00.

600.

730.

540.

38[−

0.09

,−0.

03]

[0.0

7,0.

28]/

[0.3

0,0.

38]

[0.3

9,0.

79]

[0.4

6,0.

92]

[0.1

7,0.

89]

[0.0

2,0.

78]

NES

PZ0.

630.

630.

18/0

.27

0.57

0.38

0.32

[0.6

1,0.

65]

[0.6

2,0.

66]

[0.1

0,0.

23]/

[0.2

4,0.

27]

[0.3

2,0.

80]

[0.0

0,0.

78]

[0.0

0,0.

68]

MM

0.05

0.52

0.39

0.16

/0.3

70.

850.

79[0

.01,

0.07

][0

.49,

0.54

][0

.36,

0.41

][0

.05,

0.27

]/[0

.33,

0.40

][0

.74,

0.97

][0

.64,

0.93

]IM

0.04

0.33

0.25

0.69

0.09

/0.3

50.

87[0

.01,

0.07

][0

.29,

0.37

][0

.22,

0.28

][0

.67,

0.70

][0

.03,

0.18

]/[0

.31,

0.39

][0

.74,

0.97

]PT

M0.

020.

190.

150.

430.

410.

22/0

.45

[0.0

0,0.

07]

[0.1

4,0.

24]

[0.1

1,0.

19]

[0.3

9,0.

47]

[0.3

7,0.

45]

[0.1

1,0.

33]/

[0.4

1,0.

49]

aV

OL

=vo

lum

eof

ejac

ula

te;

CO

NC

=co

nce

ntr

atio

n;

NES

PZ=

nu

mbe

rof

sper

mat

ozoa

per

ejac

ula

te;

MM

=m

ass

mot

ilit

y;IM

=in

div

idu

alm

otil

ity;

PTM

=p

ost

thaw

ing

mot

ilit

y.

n Science 124 (2011) 28–38 35

tary status of the stud, environmental contaminants, etc.)around the date of collection. This effect has been littlestudied previously but it is expected to play an importantrole in obtaining unbiased estimates of semen character-istics in field conditions. Tusell et al. (2010) have recentlyfound evidence of the importance of this kind of environ-mental effect in buck fertility. The statistical treatment ofthe effect depends on the amount of information availablefor each class. In our case, the use of a flat prior on WC ina model with no YS effect provided a nearly as good fit ofthe data (except for PTM) as a model with YS and WC witha Gaussian prior, the full model. This might be due to thefact that more information is available in the data to esti-mate the WC effect for all traits compared with the numberof records available for PTM. When the number of data isscarce for estimating WC classes, a model including bothfactors, a YS effect and a WC effect with a Gaussian prior,might be advantageous because YS should normally havea relatively large number of observations and counterbal-ance the lack of accuracy of the estimates of WC classes.Moreover, in medium or large AI centres, a day of collec-tion effect instead of WC might be considered to improvethe environmental adjustment on bull fertility. Connected-ness issues, i.e., to ensure that enough number of bulls aremeasured in each day of collection with genetic connec-tions to bulls in other contemporary groups, would need tobe considered in order to avoid confounding effects.

The trend estimated from YS and WC effects reflects acompendium of short and long term environmental effectsacting on semen production along time. In our study,the traits showing more changes across time were CONCand PTM and the least affected was VOL. The recordedinformation for CONC and PTM clearly depends on themeasurement equipment and on technical protocols forfreezing and thawing semen and on the technician expe-rience. On the other hand, VOL should be more affectedby bull handling during collection and MM and MI bythe laboratory technician. CONC showed a steady increasefrom year 1982 through year 1996. Then, a sharp decreaseoccurred, followed by more stable measurements. Thesharp change in trend coincided with the beginning of theuse of a new spectrophotometer in the laboratory. For PTM,the improvement in the measures in the last period wasassociated with an improvement in the freezing protocol.Because the AI centre studied is of small-medium size, onlyone bull handler and two consecutive technicians werepresent for the whole studied period, which might explainthe smaller impact of this effect on VOL, MM and MI.

Long term environmental trends were not observed inthis study. The presence of environmental chemicals thatdisrupt the endocrine system of animals has been hypoth-esized as a cause of decline in human male fertility andstudied by Wahl and Reif (2009) in bulls as an animal modelfor reproductive abnormalities in humans. However, in thisstudy, the only trait showing a persistent trend was PTM,but of positive sign, and therefore could not be related to

the presence of environmental contaminants, which wouldhave induced a negative trend. These findings are in agree-ment with early studies aiming at estimating time trends inlarge series of years (Setchell, 1997; Stolla and Trombach,1999) and with the recent findings of Wahl and Reif (2009).

3 oduction

Fsctp

lmdibpAstfsfaeTdttwmedpoaaplctatTumtstActcpabNslpi

simo

6 S. Karoui et al. / Animal Repr

rom results in those studies and in the present work, iteems that despite the potential exposure of bulls in AIentres to environmental pollutants, there is no consis-ent trend towards declining quality and quantity of spermroduction.

Another type of trend was obtained in this study, byooking at the raw averages and average estimated bull

erit by year of birth to investigate changes in semen pro-uction due to quality of bulls for these traits. Changes

n bull production would reflect how selection criteria forulls entering progeny test along time have affected semenroduction. In the centre providing the data, as in mostI centres around the world, selection of young Holsteinires has been based mostly on production traits. The twoypes of estimated trends (from raw phenotypic data androm estimated bull effects) showed different patterns, ashown in Fig. 4. These discrepancies can be explained by theact that bull’s merit, obtained from its estimated geneticnd permanent environment values, is free of the generalnvironmental factors affecting semen production (AC, EN,PC, Fi and TCF in our study). Thus, for VOL and CONC, theifference between the two curves might be explained byhe correction for the AC effect. Phenotypic measures athe beginning of the time series (years of birth 1985–1988)ere obtained after year 1990, when bulls are adult ani-als with a larger expected semen volume. The opposite is

xpected for the last years of birth, for which young bullata is more frequent. For CONC, bulls born in the firsteriod are expected to have collection records when theptimum CONC level has been passed (beyond 3 years ofge), while bulls born in the last period had collections atges near optimum for semen quality. For PTM, the rawhenotypic trend might be reflecting the improvement in

aboratory techniques implemented in the last period ofollected records. The use of AC and short and mediumerm environmental effects (YS and WC) is expected todjust the raw records for these factors and would explainhe lack of trend observed from the estimated bull effects.herefore, failing to account for environmental factors, i.e.,sing raw data, may produce biased results for the esti-ated trends for bull soundness with time. The lack of

rend observed from the bull effects would mean that theelection practiced mainly for production does not seemo have affected the semen production of the bulls in thisI centre. This might be due to an eventually low geneticorrelation between production and semen traits and/oro the fact that low fertility bulls are screened and dis-arded according to their semen quality and this might bereventing deterioration in semen characteristics. Ducrocqnd Humblot (1997) reported that the elimination of youngulls based on semen traits was about 10–15% per year inormande bulls. The bulls producing the data used in this

tud have been sampled from the world Holstein popu-ation using international genetic evaluations mainly forroduction traits and might therefore reflect the situation

n other Holstein bull studs.

A consequence of the intense selection in the Hol-

tein population is the increase of the inbreeding leveln this population. In the bull population studied the

ean inbreeding level (5.51%) was in the upper limitf average inbreeding estimates obtained for other Hol-

Science 124 (2011) 28–38

stein populations, where Fi ranged from 2.64% to 4.7%(Miglior and Burnside, 1995; Van Raden, 2002; Kearneyet al., 2004; Sørensen et al., 2005; Sewalem et al., 2006).Moreover, the average rate of increase in Fi was 0.27%per year, a larger rate than the 0.17% reported by Kearneyet al. (2004), and than the 1% per generation reported bySørensen et al. (2005) for British and Danish Holstein pop-ulations, respectively. However, despite the slightly largerrate of increase in inbreeding observed in this population,inbreeding depression on semen traits was almost negligi-ble. Diarra et al. (1997), studying Holstein bulls, and VanEldik et al. (2006) for Shetland pony stallions, did not findsignificant effects of Fi on the volume of ejaculate but othertraits connected to semen quality (motility and abnormalsperm morphology) showed a detrimental effect for highvalues of Fi. The small effect of Fi found in this study mightbe due to the fact that although average level of inbreedingwas relatively high, only 8 animals showed Fi values over12.5%. The maximum level of inbreeding observed in ourstudy (16.9%) was not as high as in other studies probablydue to the inbreeding monitoring practiced in the analysedAI centre. González-Recio et al. (2007), in their study on thepopulation of cows and bulls associated with the AI centreinvolved in this study, found average levels of Fi of 3%, buta maximum level of 39%.

Once records have been adjusted for systematic envi-ronmental effects, the efficiency of selection of semencharacteristics depends on the heritability of the traits.The literature estimates of genetic parameters for semenproduction traits vary in a wide range of values becauseof differences in the methods of estimation, populationsand ways to measure these traits. In general, larger esti-mates are found for volume of ejaculate, ranging fromvalues under 0.10 in Simmental (Miroslav et al., 2000) andHereford (Kealey et al., 2006) bulls to 0.65 in the studyof Ducrocq and Humblot (1995) on Holstein bulls, whereaverage values of at least 10 measures were used as depen-dent variables (with a subsequent increase in heritabilityestimates). Motility score or mass motility, subjectivelymeasured in a discrete scale, showed the lowest heritabil-ity values, under 0.05 in Taylor et al. (1985) for Holsteinand in Gredler et al. (2007) for Simmental bulls, althoughlarge values (0.50 for motility score and 0.43 for progres-sive motility) have also been reported (Druet et al., 2009).In our study, the lowest heritability value was found for IM,which was measured in a subjective way by the AI centretechnician. Repeatabilities were within the limit of valuesfound in the literature and indicate that semen produc-tion is only moderately repeatable. Mathevon et al. (1998)found lower repeatability estimates for young animals thanfor adult bulls which was explained by these authors by thefact that semen production is expected to be more stable inadult bulls than in young growing animals. In our case, near70% of the observations came from bulls under 30 monthsof age, which might partially explain the moderate repeata-bility values. Low repeatability values have also been found

by Gredler et al., 2007 in dual purpose Simmental bullsaveraging less than 3 years of age, but not so by Druetet al. (2009) who found large heritability estimates and,therefore, large repeatabilities, in young Holstein bulls.As mentioned by Rodríguez-Martínez (2003), the sperm

roductio

S. Karoui et al. / Animal Rep

population in a particular ejaculate is heterogeneous bynature because it represents the sum of cells that developin various spermatogenic waves. Therefore, the expectedrepeatability of semen characteristics in measures of thesame bull is not expected to be one even if we could fix theexternal environmental and technical factors that deter-mine the final sperm production observed. In this sense,identification of those external factors and a good statisti-cal modelling can help to improve repeatability of the traits,and therefore, reduce inaccurate bull’s breeding soundnessevaluations.

For the semen production traits, estimates of geneticcorrelations published previously show an antagonisticcorrelation between semen volume and concentration.In consequence, these two traits are expected to showopposite relationships with the remaining traits. In ourstudy, the estimated genetic correlation between VOL andCONC was slightly negative (with a 73% probability of tak-ing values smaller than 0) and close to zero for VOL vs.positive for CONC and the motility traits. Estimates of cor-relations between VOL, CONC or NESPZ and the motilitytraits showed larger 95HPD than estimates among motil-ity traits. Accuracy of estimates tended to be inverselyrelated to the magnitude of the correlation. Estimates ofcorrelations between traits that are expected to be highlycorrelated, such as the motility traits, are obtained withmore accuracy but traits with an intermediate or low cor-relation showed a wide range of variation in the sampledvalues and large autocorrelation between successive sam-ples. In addition, observation of the Gibbs sampling chainsindicated a certain degree of complementariness betweenestimates of additive genetic and permanent environmen-tal effects, which occurs when identifiability problemsexist. Identification or separation of those two compo-nents associated with the bull is only possible througha good pedigree information. In our case, records from arelatively large number of families were used, 96 fami-lies of paternal half sibs and 80 families of maternal halfsibs, but the family size was low (between 2 and 19 withan average of 5.2 sons). Other authors (Mathevon et al.,1998; Gredler et al., 2007) have mentioned the possibleconfounding between genetic and environmental compo-nents associated with a bull for semen traits because ofthe low amount of family information present. Gredleret al. (2007) found a close to zero and positive geneticcorrelation between VOL and CONC and positive valuesfor the correlation between volume and motility score butnegative phenotypic correlations. In that case informationon 301 bulls sired by 114 animals was used, with verysmall family size. Correlations among motility traits werelarge and positive, as expected, and more accurately esti-mated.

Although semen characteristics measured in commer-cial conditions such as the ones considered in this studyhave shown only moderate correlations with fertilityin some studies (Rodríguez-Martínez, 2003), their use

presents some advantages such as the easy and low costof measurement, and moderate heritability and repeatabil-ity. On the other hand, the use of results of AI, which couldbe considered as a true measure of bull fertility, requiresthe collection of a very large number of inseminations per

n Science 124 (2011) 28–38 37

bull to have a good accuracy and is obtained later than thesemen characteristics. AI results depend on many factorsthat are difficult to adjust for and, typically, the male con-tribution represents a very small proportion of the totalvariability. An improved evaluation of bull fertility could beobtained by the joint evaluation of results of AI collected ona large scale in many breeding schemes and semen traitscollected in the AI centres, to produce a combined index ofbull fertility soundness.

5. Conclusions

This study allowed confirmation of the large influ-ence of three environmental factors, the period of thecollection, the age at collection and the ejaculate order,on semen production traits. VOL and CONC were moreaffected by age and ejaculate order, respectively. On theother hand, motility traits, indicators of sperm viability,were mostly affected by the environmental conditionsaround period of collection. Inbreeding depression onsemen traits could not be considered relevant despite theincrease of this parameter over time. Moreover, a sys-tematic trend for the individual bull effect, composed ofits genetic and permanent environmental components,was not observed for semen traits. These results indicatethat neither the increased rates of inbreeding, nor thecorrelated responses that might result from the intenseselection practiced on other traits have had a negativeimpact on the analysed semen traits in this population.Long term environmental trends, which combine man-agement practices and environmental conditions, werenot observed for these traits, except for post-thawingmotility, which largely depends on the technical prac-tices in the laboratory. In fact, most of the changesobserved in the estimated environmental trends for alltraits were associated with changes in laboratory proto-cols. The larger values for heritabilities and repeatabilitiesobtained in this and other studies compared with estimatesfor those parameters for bull fertility indices obtained fromAI results confirm the use of regularly collected semencharacteristics as efficient, low cost and early predictorsof sperm viability, one of the bull breeding soundnesscomponents.

Finally, the use of simple linear models such as theone proposed in this study can help AI centres to opti-mise semen production by monitoring environmentalfactors and bull soundness in order to guarantee a goodseminal industry. Inclusion of other traits indicators ofsperm ability to fecundate the oocyte and produce viableembryos, combined with the routinely collected traitsmight be important in order to maintain bull fertil-ity, preserving the selection priorities in the dairy cattlesector.

Acknowledgements

The financial support of the project INIA-FEDER:RTA2007-0071 is greatly acknowledged. S. Karoui wasfunded by an INIA scholarship.

3 oduction

R

B

B

C

C

D

D

D

D

E

F

F

F

G

G

G

H

J

K

1159–1170.

8 S. Karoui et al. / Animal Repr

eferences

rito, L.F.C., Silva, A.E.D.F., Rodrigues, L.H., Vieira, F.V., Deragon, L.A.G.,Kastelic, J.P., 2002a. Effect of age and genetic group on characteris-tics of the scrotum, testes and testicular vascular cones, and on spermproduction and semen quality in AI bulls in Brazil. Theriogenology 58,1175–1186.

rito, L.F.C., Silva, A.E.D.F., Rodrigues, L.H., Vieira, F.V., Deragon, L.A.G.,Kastelic, J.P., 2002b. Effects of environmental factors, age and geno-type on sperm production and semen quality in Bos indicus and Bostaurus AI bulls in Brazil. Anim. Reprod. Sci. 70, 181–190.

hristensen, P., Brockhoff, P.B., Lehn-Jensen, H., 1999. The relationshipbetween semen quality and the nonreturn rate of bulls. Reprod.Domest. Anim. 23, 492–496.

hristensen, P., Boelling, D., Pedersen, K.M., Korsgaard, I.R., Jensen, J., 2005.Relationship between sperm viability as determined by flow cytome-try and nonreturn rate of dairy bulls. J. Androl. 26, 98–106.

iarra, S., Paré, J.P., Roy, G., 1997. Facteurs genetiques et environnemen-taux affectant la qualité de la semence de jeunes taureaux Holstein.Can. J. Anim. Sci. 77, 77–85.

ruet, T., Fritz, S., Sellem, E., Basso, B., Gerard, O., Salas-Cortes, L., Humblot,P., Druart, X., Eggen, A., 2009. Estimation of genetic parameters andgenome scan for 15 semen characteristics traits of Holstein bulls. J.Anim. Breed. Genet. 126, 269–277.

ucrocq, V., Humblot, P., 1995. Genetic characteristics and evolution ofsemen production of young Normande bulls. Livest. Prod. Sci. 41, 1–10.

ucrocq, V., Humblot, P., 1997. Is selection on semen characteris-tics needed? In: International Workshop on Genetic Improvementof Functional Traits in Cattle: Fertility and Reproduction. Groub,Germany, November 23–25.

verett, R.W., Bean, B., y Foote, R.H., 1978. Sources of variation of semenoutput. J. Dairy Sci. 61, 90–95.

iaz, M., Usmani, R.H., Abdullah, M., Ahmad, T., 2010. Evaluation of semenquality of Holstein Friesian and Jersey bulls maintained under sub-tropical environment. Pakistan Vet. J. 30, 75–78.

uente, L.F., García, L., Vallejo, M., 1984. Reproductive characters in Gali-cian Blonds. I. Characters of semen used for artificial insemination.Anal. Fac. Vet. León 30, 119–125.

uerst-Waltl, B., Schwarzenbacher, H., Perner, C., Sölkner, J., 2006. Effectsof age environmental factors on semen production and semen qualityof Austrian Simmental bulls. Anim. Reprod. Sci. 95, 27–37.

onzález-Recio, O., López de Maturana, E., Gutiérrez, J.P., 2007. Inbreedingdepression on female fertility and calving ease in Spanish dairy cattle.J. Dairy Sci. 90, 5744–5752.

redler, B., Fuerst, C., Fuerst-Walt, B., Schwarzenbacher, H., Sölkner, J.,2007. Genetic parameters for semen production traits in Austriandual-purpose Simmental bulls. Reprod. Domest. Anim. 42, 326–328.

utiérrez, J.P., Goyache, H., 2005. A note on ENDOG: a computer pro-gram for analyzing pedigree information. J. Anim. Breed. Genet. 122,172–176.

allap, T., Jaakma, Ü., Rodriguez-Martinez, H., 2006. Changes in semenquality in Estonian Holstein AI bulls at 3, 5 and 7 years of age. Reprod.Domest. Anim. 41, 214–218.

anuskauskas, A., Johannisson, A., Soderquist, L., Rodriguez-Martinez, H.,2000. Assessment of sperm characteristics postthaw and response tocalcium ionophore in relation to fertility in Swedish dairy AI bulls.Theriogenology 53, 859–875.

ass, E.R., Raftery, A., 1995. Bayes factors. J. Amer. Stat. Assoc. 430,773–795.

Science 124 (2011) 28–38

Kealey, C.G., MacNeil, M.D., Tess, M.W., Geary, T.W., Bellows, R.A.,2006. Genetic parameter estimates for scrotal circumference andsemen characteristics of line 1 Hereford bulls. J. Anim. Sci. 8,283–290.

Kearney, J.F., Wall, R., Villanueva, B., Coffey, M.P., 2004. Inbreeding trendsand application of optimized selection in the UK Holstein population.J. Dairy Sci. 87, 3503–3509.

Koivisto, M.B., Costa, M.T.A., Perri, S.H.V., Vicente, W.R.R., 2009. The effectof season on semen characteristics and freezability in Bos indicus andBos taurus bulls in the South Eastern region of Brazil. Reprod. Domest.Anim. 44, 587–592.

López-Romero, P., Rekaya, R., Carabano, M.J., 2003. Assessment ofhomogeneity vs. heterogeneity of residual variance in randomregression test-day models in a Bayesian analysis. J. Dairy Sci. 86,3371–3385.

Lucy, M.C., 2001. Reproductive loss in high-producing dairy cattle: wherewill it end? J. Dairy Sci. 84, 1277–1293.

Mandal, D.K., Kumar, M., Tyagi, S., 2010. Effect of age on spermiogram ofHolstein Friesian Sahiwal crossbred bulls. Animal 4, 595–603.

Mathevon, M., Buhr, M.M., Dekkers, J.C., 1998. Environmental, manage-ment, and genetic factors affecting semen production in Holstein bulls.J. Dairy Sci. 81 (12), 3321–3330.

Meuwissen, T.I., Luo, Z., 1992. Computing inbreeding coefficients in largepopulations. Genet. Sel. Evol. 24, 305–313.

Miglior, F., Burnside, E.B., 1995. Inbreeding of Canadian Holstein cattle. J.Dairy Sci. 78, 1163–1167.

Miroslav, K., Marijan, P., Nikola, S., Bozo, M., 2000. Genetic evaluation ofsemen and growth traits of young Simmental bulls in performancetest. Agric. Conspect. Sci. 64, 15–20.

Rodríguez-Martínez, H., 2003. Laboratory semen assessment andprediction of fertility: still utopia? Reprod. Domest. Anim. 38,312–318.

Setchell, B.P., 1997. Sperm counts in semen of farm animals. Int. J. Androl.20, 209–214.

Sewalem, A., Kistemaker, G.J., Miglior, F., Van Doormaal, B.J., 2006. Anal-ysis of inbreeding and its relationship with functional longevity inCanadian dairy cattle. J. Dairy Sci. 89, 2210–2216.

Sørensen, A.C., Sørensen, M.K., Berg, P., 2005. Inbreeding in Danish dairycattle breeds. J. Dairy Sci. 88, 1865–1872.

Stolla, R., Trombach, H., 1999. Changes in testicular function? Semenquality in AI bulls over a 25 years period. Reprod. Domest. Anim. 34,343–350.

Taylor, J.F., Bean, B., Marshall, E., y Sullivan, J.J., 1985. Genetic andenvironmental components of semen production traits of artificialinsemination Holstein bulls. J. Dairy Sci. 68, 2703–2722.

Tusell, L., Garcia-Tomas, M., Baselga, M., Rekaya, R., Rafel, O., Ramon, J.,Lopez-Bejar, M., Piles, M., 2010. Interaction of genotype × artificialinsemination conditions for male effect on fertility and prolificacy.J. Anim. Sci. 88, 3475–3485.

Van Eldik, P., Van der Waaij, E.H., Ducro, B., Kooper, A.W., Scout,T.A.E., Colenbrander, B., 2006. Possible negative effects of inbreed-ing on semen quality in Shetland pony stallions. Theriogenology 65,

Van Raden, P., 2002. AIPL Changes to Evaluation and Editing Sys-tem – February 2000. Inbreeding sample. http://aipl.arsusda.gov/reference/changes/chng0002.html (accessed 15.09.09).

Wahl, R.L., Reif, J.S., 2009. Temporal trends in bull semen quality: a com-parative model for human health? Environ. Res. 109, 273–280.