| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
* Animal Science Unit, Gembloux Agricultural University, B-5030 Gembloux, Belgium
Livestock and Pasture Office, 1002 Tunis Belvedere, Tunisia
Ecole Supérieure d’Agriculture de Mateur, 7030 Mateur, Tunisia
CONVIS Herdbuch, Service Elevage et Génétique, L-9004 Ettelbruck, Luxembourg
# National Fund for Scientific Research, B-1000 Brussels, Belgium
1 Corresponding author: gengler.n{at}fsagx.ac.be
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Key Words: milk yield • genetic parameter • genotype x environment interaction • Holstein
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
It has been found that G x E occurs when the performance of different genotypes is not identically affected by different environments. The definition of environment should not include only physical and climatic conditions, but also production and health management, economic constraints, and prevailing agricultural policies (Stanton et al., 1991). Studies on G x E vary from controlled experiments with a few hundred animals (Veerkamp et al., 1995; Beerda et al., 2007) to modeling large field data sets (Weigel et al., 2001; Zwald et al., 2003). Only scaling effects caused by heterogeneous genetic variances were reported by almost all within-country analyses of milk yield in Holsteins (Calus et al., 2002; Raffrenato et al., 2003; Fahey et al., 2007). Furthermore, high genetic correlations were found between countries from the same ecological zone (Weigel et al., 2001) with no evidence of G x E. On the other hand, studies on G x E between countries with different climatic conditions and production systems (Stanton et al., 1991; Costa et al., 2000; Ojango and Pollott, 2002) were rare. In these studies, genetic correlations between countries suggest the existence of G x E for milk yield in Holsteins. Moreover, Stanton et al. (1991) and Costa et al. (2000) found that the response to selection was smaller in low-input systems than in high-input ones.
Milk yield has been the main breeding objective in Tunisia, whereas a composite index that includes durability, health, and reproduction traits in addition to milk yield is the selection criterion currently used in Luxembourg (Miglior et al., 2005). The Luxembourg Holstein population originally included one-third non-Holstein Red and White dairy cows. Breeders in Luxembourg imported heifers mainly from Germany. Tunisia started importing purebred pregnant Friesian heifers from the Netherlands in 1970. Holstein semen and heifers were then imported from Canada, the United States, and some European countries (Hammami et al., 2007). As in Tunisia, breeders in Luxembourg are currently using semen from mostly North American and European Holstein sires. Therefore, average additive genetic relationships and genetic similarity between the Luxembourg and Tunisian cow populations have increased with time as a result of continuously using common sires proven in foreign high-input environments (Hammami et al., 2007). The Tunisian environment can be described as a low-to medium-input system, whereas the Luxembourg one can be described as a high-input system. Quantifying interactions between imported germplasm and Luxembourg and Tunisian environments is important for both populations to evaluate and adjust their breeding strategies if necessary. The objective of this study was to assess G x E for first-lactation milk yield in Holsteins using Luxembourg and Tunisian field data.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
.
|
Age at calving and the calving season were defined to account for specificities of each population. Four seasons (September to November, December to February, March to May, and June to August) for Tunisia and 3 seasons (January to March, April to August, September to December) for Luxembourg were identified. Additionally, age at calving was classified into 5 classes (<28 mo, 28 to 30, 31 to 33, 34 to 36, and >36 mo) for the Luxembourg population and 6 classes (<26 mo, 26 to 27, 28 to 29, 30 to 31, 32 to 33, and >33 mo) for the Tunisian population, respectively. Different age at calving classes were used to account for dissimilar ranges of calving age in the 2 countries (Table 1
).
Analysis
Genetic parameters and evaluations were obtained from single and joint analyses of Luxembourg and Tunisian data by a random regression TD animal model. In matrix notation, the within-country model was:
y = Xb + Q (Za + Zp + Wh) + e,
where y = a vector of TD milk yields; b = a vector of the fixed effects: herd x test-date, age x season of calving x classes of 25 DIM, and classes of 5 DIM; a = a vector of random regression coefficients for animal genetic (AG) effect; p = a vector of random regression coefficients for permanent environmental (PE) effect; h = a vector of random regression coefficients for herd-year of calving common environmental effect (HY); e = a vector of residual effects; Q = a matrix of Legendre polynomials; and X, Z, and W = incidence matrices relating observations to various effects. Residuals were assumed to be constant within DIM intervals. Legendre polynomials were of order 4 with: qt0 = 1,
The same model from above was used for the joint analysis but was extended to a bivariate model; records from Luxembourg and Tunisia being considered 2 distinct, correlated traits. Definitions of fixed and random effects remained the same but were trait-specific and nested within country.
Expectations and covariance structure for random effects were defined as follows:
E(y) = Xb, E(a) = 0, E(p) = 0, E(h) = 0, E(e) = 0,
and
V(a) = G, V(p) = P, V(h) = H, V(e) = E,
where
Genetic parameters from both single and 2-country models were estimated with a Bayesian approach via a Gibbs sampling algorithm (Misztal et al., 2002). Single chains of 120,000 samples (with 20,000 discarded) were generated for separate analyses on the data of each country. For the joint analysis, a chain of 160,000 samples (with 30,000 as burn-in period) was generated. Convergence of Gibbs chains was monitored by inspecting plots of selected realizations.
Persistency of lactation can be defined as the ability of a cow to maintain milk production after peak yield. This parameter was defined in various ways in the literature (Gengler, 1996), with no consensus reached yet on the suitable definition. Lately, this parameter has been calculated as a by-product of the random regression model (Jamrozik et al., 2002; Druet et al., 2005). In this study, persistency was defined as the breeding value on DIM 280 minus the breeding value for milk yield on DIM 80. In the definition by Jamrozik et al. (1998), DIM 80 was chosen to replace DIM 60 because average peak yield occurs in DIM 73 and DIM 65 in Luxembourg and Tunisian Holsteins, respectively (Table 1). Heritability [
(c)] for parameter i (daily, 305-d period, and persistency) in country c (Luxembourg or Tunisia) was calculated as:
The constant k multiplying the residual variance (
where G(l,t) = the genetic covariance matrix of curve parameters for milk yield in Luxembourg and Tunisia and G(l,l) and G(t,t) = the genetic covariance matrices of the same curve parameters in Luxembourg and Tunisia, respectively. The genetic correlation for persistency was obtained by replacing q305-d with qpers in the formula above.
Estimated breeding values of animals for milk yield at a given DIM (t) were computed as qtâs, where âs = (â0s, â1s, â2s, â3s, â4s) is the vector of solutions for the additive genetic random regression coefficients of animal s. The EBV for 305-d milk yield (EBV305-d) were obtained by summing for each animal the EBV from 1 to 305 DIM. Those of persistency of lactation (EBVpers) were calculated as EBV280 – EBV80. Product-moment and rank correlations between EBV of sires with at least 4 daughters with records in each of the 2 countries were calculated using PROC CORR (SAS Institute, 2002).
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
. Milk production differed between the 2 countries. A first-lactating Holstein cow in Luxembourg produced nearly 4 kg more milk per day than its counterpart in Tunisia. This difference summed up to 1,744 kg of milk over a whole lactation for all completed lactations in 2004 (ICAR, 2007). Peak yield was also greater for Luxembourg than for Tunisian first-lactation cows. Furthermore, primiparous cows in Tunisia tended to reach their peak of production earlier than those under Luxembourg management conditions (Table 1). Zwald et al. (2001) found that the number of days to peak yield was larger in countries with a high milk production level than in countries with a low milk production level. The average age at first calving was around 30 mo in both populations. Although the frequencies of calving were nearly uniform over the year in Tunisia, they were more variable in Luxembourg, with a peak (35%) during September to November and a decline (13%) in the March to May period. Herd sizes were larger in Tunisia than in Luxembourg, and the average number of recorded cows per herd in Tunisia was 3x greater than that in Luxembourg in 2004 (ICAR, 2007). Luxembourg had a small average number of TD records per herd-test date class (9.4), but this level is in the range found in bordering countries. Herds in Belgium, the Netherlands, and Germany, have 5.4, 9.6, and 10.3 records per TD, on average, respectively (Zwald et al., 2001). This same parameter was 36.1 records in Tunisia, comparable to that observed on Israeli (34.0) first-lactation Holstein cows (Zwald et al., 2001). The period of high temperatures lasts from May to September in Tunisia, with peaks during mainly the summer season. High temperature and humidity may compromise milk production of Holstein cows for a long period of time in Tunisia. In Luxembourg, only the month of August may constitute a period of discomfort for Holstein cows because of heat. Negative effects of heat stress on milk production were reported by Ravagnolo et al. (2000). Moreover, the average rainfall was greater in Luxembourg than in Tunisia. Limited water resources may constrain the quantity and quality of forage production. This may consequently hinder high-yielding breeds from expressing their potentials. Reproduction parameters could also be affected, which would help explain longer calving intervals in Tunisia in 2004 (ICAR, 2007).
Table 2 gives origins of common sires and maternal grandsires for both populations. Luxembourg and Tunisia had 231 common sires, with between 4 and 903 daughters. In addition, 114 maternal grandsires were in common (with at least 4 granddaughters in each country), of which 67 were also common sires. Common sires and maternal grandsires came from the United States, Canada, Germany, the Netherlands, Italy, and France. Sires from the United States and Germany were used the most among all foreign bulls. Nevertheless, some variability was observed in sire usage between the 2 countries (Hammami et al., 2007). The number of daughters from the United States common sires was greater in Tunisia, and the number of daughters of sires from Germany and the Netherlands was greater in Luxembourg.
|
. Estimates obtained for Tunisia in this study were very similar to those for first-lactation milk yield obtained by Hammami et al. (2008) using a 3-trait-3-lactation random regression TD model. The AG variance estimates for the first 3 coefficients from the Luxembourg data were 46 to 69% larger than corresponding estimates from the Tunisian data. On the other hand, Tunisian PE variances for the same first 3 coefficients were 52 to 65% larger than Luxembourg ones. Although the joint analysis provided greater AG and PE variance estimates for the same random coefficients in both countries, proportional differences in these estimates between populations were maintained. The largest variances of PE effect in Tunisia could be due to the poor management practices and feeding fluctuations during the year, which introduce additional variation that is permanently associated with each cow compared with its counterpart in Luxembourg. On the other hand, the decreased AG variances in the Tunisian environment can be possibly caused by difficulties encountered by daughters of superior sires to express their genetic potential under harsh conditions. We can speculate that the within-herd correlations of genotype x management (essentially feeding system) in Luxembourg are large because of the available quantity and quality of forages. The most common buffer feeds [i.e., maize (Zea mays) silage and brewers grains] used in Luxembourg farms are available only for very few Tunisian farms in limited quantity and with low nutritional values.
|
. The first principal component of the AG effect explained 86 and 89% of total variations in milk yield in Tunisia and Luxembourg, respectively. The cumulative proportion of AG variance explained by the first 3 principal components was over 98% of the total variance in each of the 2 countries. However, 4 components were needed to account for that much (98%) variation in the PE effect. This was in agreement with Pool et al. (2000), who reported that a third-order Legendre polynomial for AG and a fourth polynomial for PE variances should be used to accurately model variance curves across lactation. The first 2 components accounted for 97 and 95% of total AG variances in Luxembourg and Tunisia, respectively. These 2 components have been used to represent total yield and persistency in some studies (Jamrozik et al., 2002; Druet et al., 2005). Jamrozik et al. (2002) found only small differences with respect to the first 2 principal components among the Australian, Canadian, Italian, and New Zealand Holstein populations when using fourth-order Legendre polynomials. Based on a random regression TD model with the same order of Legendre polynomials, Druet et al. (2005) also found that the first 2 components explained more than 95% of variation. They used only 2 eigenvectors, associated with the first 2 eigenvalues, as covariates to estimate covariance components of daily milk traits in French Holsteins. In this study (Table 4), a very large proportion of the AG variances (99% for Luxembourg, 98% for Tunisia) was explained by the first 3 eigenvalues, and the relatively small eigenvalues for higher orders show that the rank reduction of the actual model by using only the third order for the AG part may be sensible. On the other hand, a relatively smaller proportion of the PE variance (95% for both countries)
|
Posterior means of heritability estimates of 305-d yield and persistency as defined in this study for first lactation obtained from univariate and bivariate (Luxembourg and Tunisia used as 2 separate traits) analyses are given in Table 5. Heritabilities obtained from the joint analysis were larger than those obtained from separate analyses. Tunisia had consistently smaller heritabilities than Luxembourg for milk yield and persistency. For instance, heritability for 305-d yield in Tunisia was 53% lower than that in Luxembourg. Ojango and Pollott (2002) reported that heritability estimates for 305-d milk yield in Kenya were 42% lower than that in the United Kingdom when doing a joint analysis. Carabaño et al. (1989) also found that Spanish first-lactation 305-d milk yield heritability estimated using a sire lactation model (0.16 from within Spain and 0.12 from between Spain and the US analyses) was smaller than that found in the US population (0.33 from within the United States and 0.26 from between Spain and the US analyses). These authors concluded that low heritability estimates obtained for the Spanish Holsteins were caused by poor management conditions. Differences in heritabilities between Luxembourg and Tunisia may be caused by differences in production levels resulting from differences in climatic conditions and management. Jamrozik et al. (2002) also reported variable estimates of lactation curve parameters among countries with different production systems. These authors linked differences in estimates not only to absolute production levels but also to relationships among lactation curve parameters found for any given population. Persistency as defined by Jamrozik et al. (2002) clearly depended on production levels of the considered population. Heritability estimates of persistency for both populations were lower than the 0.30 reported by Jamrozik et al. (1998) for the Canadian Holsteins using the linear slope between yields at DIM 60 and DIM 280 as a measure of persistency. Low heritability of persistency obtained in Tunisia could reflect difficulties encountered by cows to maintain high milk production after the peak, thereby suppressing expression of genetic variance.
|
. The genetic correlation between Luxembourg and Tunisian 305-d milk yields (0.60) was low compared with estimates in some studies (Weigel et al., 2001; Jamrozik et al., 2002). Jamrozik et al. (2002) reported genetic correlations for total (mostly 305-d) milk yield. Results ranged from 0.66 to 0.83 among Australian, New Zealand, Canadian, and Italian first-lactation Holsteins using a random regression TD animal model. Likewise, Weigel et al. (2001) reported values greater than 0.80 between total yield correlation estimates obtained by a multitrait analysis of 16 million first-lactation records from 17 Interbull country members. Estimates reported by Weigel et al. (2001) were >0.90 between countries with predominantly grazing systems (Ireland, Australia, and New Zealand). Those correlations were greater than 0.91 among countries with high milk production levels (United States, Canada, Belgium, the Netherlands, and Italy). Correlations between total yields in the remaining Interbull countries ranged between 0.80 and 0.90 (Weigel et al., 2001). The genetic correlation between 305-d milk yields in Luxembourg and Tunisia is still in the range of estimates obtained between pairs of countries known with harsh climatic conditions or diverse production systems (e.g., the genetic correlation between the United Kingdom and Kenya was only 0.49; Ojango and Pollott, 2002). Cienfuegos-Rivas et al. (1999) reported a genetic correlation of 0.63 between the United States and Mexico. A slightly larger genetic correlation (0.66) between Italy and New Zealand was reported by Jamrozik et al. (2002). The genetic correlation between persistency, as defined in this study, in Luxembourg and Tunisia was 0.36, which is lower than 0.55 obtained between the Canadian and Italian first-lactation cows in a study of 4 populations (Jamrozik et al., 2002), in which persistency was defined as a byproduct of the random regression model. These authors found the lowest estimate (0.16) between the Italian and New Zealand Holsteins. According to Robertson (1959), low genetic correlations suggest evidence of G x E. In this study, low genetic correlations (<0.80) between parameters (305-d milk yield and persistency) of Tunisian and Luxembourg first lactation indicate that superior animals on a trait in one environment are not necessarily as superior in the other and vice versa.
G x E
Genetic variances and heritability estimates revealed unequal genetic expression of Holstein genes under Luxembourg and Tunisian management circumstances. Several studies (Veerkamp and Goddard, 1998; Gengler et al., 2005) reported that the genetic variance for milk yield is greater in high-input systems compared with that in low-input systems. Others (Raffrenato et al., 2003; Ben Gara et al., 2006) found low genetic variance and heritability estimates for production traits under low-input production environments. Low genetic variances for milk yield throughout lactation observed in Tunisian Holsteins translate into smaller differences in breeding values among animals. Performances of cows in Tunisia are limited because of harsh climatic conditions (discomfort) and limited feed resources compared with less constraining factors in Luxembourg, where cows are managed under conditions similar to those where their sires were selected. Currently, 70% of total bulls used in the Luxembourg cow population were selected in bordering countries with production conditions similar to those found in Luxembourg. Selection responses to the use of US Holstein sires for milk production in Latin America were estimated to be 53 to 78% of the response observed in the United States (Stanton et al., 1991). The relative response to sire selection expected in Kenya based on United Kingdom breeding values was only about 44% (Ojango and Pollott, 2002). Therefore, environmental factors appear to constrain expression of genetic merit in tropical or subtropical regions relying on imported proven Holstein semen.
Differences in additive genetic variances obtained for Luxembourg and Tunisian populations imply that a scaling effect exists for EBV of sires across these environments. Low genetic and rank correlations translate a reranking of sires across these populations. Tunisian EBV of sires for 305-d yields were regressed on corresponding Luxembourg values. Figure 1 shows a plot of these milk yield EBV of the 231 common sires estimated in Tunisia against their EBV in Luxembourg. The slope (b) in the regression equation could be viewed as the expected response measured on daughters in Tunisia from sire selection on EBV in Luxembourg. This regression coefficient was only 0.18 in the current study. This result is less than that (b = 0.32) reported by Ojango and Pollott (2002) from regressing EBV of sires in Kenya on the EBV of the same sires in the United Kingdom. The expected response (b = 0.57) measured on daughters of US sires in the low-input Mexican environment was less than responses of daughters of the same US sires in all US environments (Stanton et al., 1991). These authors concluded that the US Holstein genes had restrained expression under the Mexican management circumstances.
|
) between EBV305-d from within-country analyses were lower than corresponding genetic correlations (0.60). Cienfuegos-Rivas et al. (1999) also found low rank correlation coefficient (0.59) between herds in low milk production level in Mexico and all herds in the United States. They concluded that this result was evidence for a significant G x E interaction and that sires were ranked differently in the Mexican environment compared with their ranking in the United States. Figure 2 shows the EBV305-d of 15 common sires with more than 30 daughters in Luxembourg. It appears that considerable reranking of bulls with respect to milk yield was observed. Moreover, product moment and rank correlation based only on these 15 sires were around 0.42. Restricting computation to only this group of sires did not cause any significant change in correlation estimates between EBV found in Luxembourg and Tunisia. Mulder et al. (2004) reported that the reranking effect is most likely observed with top sires. As on milk yield EBV305-d, there were also differences in the ranking of sires on EBVpers between the 2 environments (Table 6). Product moment (0.27) and rank correlations (0.26) between EBVpers from within-country analyses were lower than those between EBV305-d (Table 5). Moreover, the rank correlation between Tunisian sire EBV305-d and Luxembourg sire EBVpers was 0.21, and that between Luxembourg sires EBV305-d and Tunisian EBVpers was 0.29. These low correlations indicate that superior animals for 305-d milk yield are not necessarily as superior for persistency of lactation. These results may also imply that animals with high 305-d yield in a favorable environment may have lower yield but greater persistency in a less favorable environment.
|
|
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
0.60) showed that paying premium prices for elite animals from temperate environments or semen from those animals is not necessarily a good strategy. The issue is to transfer correctly the genetic progress from temperate high-input levels into different environments. Moreover, importing countries should use more effort to meet the elementary needs of high-yielding breeds under their conditions.
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Received for publication March 4, 2008. Accepted for publication May 28, 2008.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  H. Hammami, B. Rekik, C. Bastin, H. Soyeurt, J. Bormann, J. Stoll, and N. Gengler Environmental sensitivity for milk yield in Luxembourg and Tunisian Holsteins by herd management level J Dairy Sci, September 1, 2009; 92(9): 4604 - 4612. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
