Abstract: (Co)variance components and genetic parameters were estimated using Derivative-free Restricted Maximum Likelihood (DFREML) approach for milk yield and fat yield of Iranian Holsteins of Isfahan province. Data was consisted of first lactation records on milk and fat yields of 12,047 Holsteins from 45 herds, which were calved from 1995 to the end of 2001. Records were pre-adjusted to mature equivalent yields (ME-2X-305d). Six different animal models were fitted, which were differentiated by including or excluding maternal additive genetic effects, maternal permanent environmental effects and direct-maternal genetic covariance. Animal Models included Herd-Year-Season and direct additive genetic effects as fixed and random effects, respectively. Direct heritability estimates (h2) ranged from 0.157 to 0.229 and 0.203 to 0.243 for milk yield and fat yield, respectively. The estimates were substantially higher when maternal effects were ignored from the model. Mean estimates of the maternal genetic and maternal permanent environment variances and direct-maternal genetic covariance as fractions to the phenotypic variances were 0.07, 0.02 and 0.025 for milk yield and 0.01, 0.05 and 0.016 for fat yield, respectively, while, positive direct-maternal relationships (Covam) were obtained. The results of this study showed that maternal additive genetic and maternal permanent environmental effects had not any important influence on the yield of dairy cows. However, they can improve genetic evaluations. Also, there is no need for the inclusion of direct-maternal genetic covariance in the animal models for dairy cows.
INTRODUCTION
The performance of individual animals is affected by genetic and environmental factors. Quantitative traits are generally assumed to be under the control of many genes, each with small additive effects (Rorato et al., 1999; Simm, 1998). Estimates of genetic parameters are important in the design of animal breeding schemes aimed to maximizing genetic gain. Heritability of a trait is a fraction of the genetic variation (additive genetic variation in the Narrow sense) to the phenotypic variation, which indicates, on average, how much of the superiority of selected animals as the next generation parents is passed to the progeny generation. Heritability is also used to evaluate the direct genetic response or correlated response to selection (Edriss and Khosravinia, 2000; Simm, 1998).
In farm mammals, traits are influenced by genes of the individual animal, the environment provided by the dam and other environmental effects (Albuquerque and Meyer, 2001). According to Willham (1963), in mammals, environmental variation surrounding the offspring is partially due to the genetics of the dam. Quantitative traits can be influenced by two genetic components: Animal genotype (direct genetic effects) and dam genotype (maternal genetic effects). Maternal effects have been defined as any influence of dam on its offspring, except the effects of directly transmitted genes that affect performance of the progeny (Legates, 1972). Edriss and Khosravinia (2000) declared that some maternal effects that have influence on pre birth growth and development of the calf, include: uterine capacity, age of dam, dam's health, litter size and dam's nutritional state. Schutz et al. (1992) suggested that milk yield, intrauterine environment and mothering ability in mammals are common components of maternal effects, which may be both genetically and environmentally determined. Estimates of heritability obtained from daughter-dam regression were usually greater than those obtained from correlation among paternal half-sibs (Seykora and McDaniel, 1983; Van Vleck and Bradford, 1966). Van Vleck and Bradford (1966) hypothesized that the larger heritability estimates from daughter-dam regression may be due to maternal genetic effects.
Some evidence suggests that maternal genetic effects are not important for yield traits in dairy cattle (Albuquerque et al., 1998; Schutz et al., 1992; Van Vleck and Bradford, 1966; Van Vleck and Hart, 1966). Albuquerque et al. (1998) concluded that additive maternal genetic effects and covariance between maternal and direct genetic effects do not seem to make important contributions to the phenotypic variances of milk yield, fat yield and fat percentage of New York Holsteins. Estimates of heritability obtained by animal model were greater than those obtained by sire model, which is probably due to selection effects on sires (Rorato et al., 1999; Dong et al., 1988) and taking into account of all pedigree relationships between animals. Many researches have been based on maternal effects on sheep and beef cattle. However, the number of these kinds of studies on dairy cattle is not enough. Also, there is a lack of investigation to estimate the importance of maternal effects on the animal model fitness.
The main objective of the present study was to estimate the importance of direct additive genetic, maternal additive genetic and maternal permanent environmental effects for milk yield and fat yield traits. Also a comparison was performed between the utilized animal models and the effectiveness of including maternal effects in the animal models was studied for dairy cows.
MATERIALS AND METHODS
The data was collected by the Animal Breeding Center of Iran in Isfahan province, including pedigree and performance information of Holstein cows calved from 1995 to the end of 2001 in 45 herds. After limiting age at first calving to be from 21 to 39 months of age, a dataset consisted of pedigree information and first lactation records on milk and fat yields of 12,047 cows were remained. Records were corrected based on mature equivalent yields (ME-2X-305d). The structure of the data is presented in Table 1.
Variance components were estimated using derivative-free restricted maximum likelihood approach applying the DFREML 3.0β package (Meyer, 1998). Genetic analyses were performed for the traits considered in this study, using six different single-trait animal models.
| Table 1: | Summery of data structure |
Herd-Year-Season (HYS) and direct additive genetic effects were considered as fixed and random effects, respectively. But the animal models were differentiated by either including or excluding, maternal genetic and maternal permanent environmental effects and direct-maternal genetic covariance as random effects. The animal models considered in this study were:
| Model 1: | y = Xβ+Z1a+ε | |
| Model 2: | y = Xβ+Z1a+Z2c+ε | |
| Model 3: | y = Xβ+Z1a+Z2m+ε | cov (a,m) = 0 |
| Model 4: | y = Xβ+Z1a+Z2m+ε | cov (a,m) = Aσam |
| Model 5: | y = Xβ+Z1a+Z2m+Z3c+ε | cov (a,m) = 0 |
| Model 6: | y = Xβ+Z1a+Z2m+Z3c+ε | cov (a,m) = Aσam |
Where: y is the vector of observations, β is the vector of fixed contemporary groups, a is the vector of random direct additive genetic effects, m is the vector of random maternal additive genetic effects, c is the vector of random maternal permanent environmental effects and ε is the vector of random residual effects. X and Z1 are the incidence matrices relating records to fixed and direct additive genetic effects, respectively. Z2 and Z3 are the incidence matrices relating records to related maternal effects.
Each model had its own variance components, for instance the variance-covariance components of Model 6 were as follows:
Where: A is the numerator relationship matrix between animals and I is the identity matrix with the order of number of records. σ2a, σ2m, σ2c, σ2ε and σam are V(a), V(m), V(c), V(ε) and cov(a,m), respectively.
Direct and maternal heritabilities were obtained as σ2a/σ2p and σ2m/σ2p, respectively; where, σ2p is the sum of all variance components (phenotypic variance) estimated by the model of analysis.
RESULTS AND DISCUSSION
The means±standard deviations were estimated (7385.7±1542.1) and (205.47±42.52) for milk yield and fat yield, respectively. According to the coefficients of variation (20.88 and 21.18 for milk and fat, respectively), milk and fat yield variations did not differ considerably in the studied population.
| Table 2: | Estimation of variance components for milk yield and fat yield, using six animal models |
| M = Model; σ2a = direct additive genetic variance; σ2m = maternal additive genetic variance; σam = direct-maternal genetic covariance; σ2c = maternal permanent environmental variance; σ2ε = random residual variance; σ2p = phenotypic variance; h2 = direct heritability; m2 = maternal heritability; c2 = ratio of maternal permanent environmental variance to phenotypic variance; am = direct-maternal genetic covariance as a proportion of phenotypic variance; Log L = Log Likelihood | |
Variance and covariance components, as well as direct and maternal heritabilities and Log L values estimated by six different animal models for the considered traits are presented in Table 2. Animal models with higher Log L values are more accurate, so it could be concluded that by considering more parameters in the models, there would be more improvement for the animal model's accuracy.
Heritability estimates obtained by model (1) were 0.229 and 0.242 for milk yield and fat yield, respectively. Weigel et al. (1999) reported heritability estimates of 0.23 and 0.17 for milk and fat yields, respectively. Heritability estimates were 0.32 and 0.205 in Spanish dairy cattle for milk and fat yields, respectively (Perez et al., 1999). The heritabilities decreased by the inclusion of more parameters in the models, except for the inclusion of maternal permanent environmental effects for fat yield.
Contrasting heritability estimates obtained by model 1 with models 2 and 3 shows that for both traits, considering maternal genetic effects in the model resulted in a further reduction in direct additive genetic variance and so direct heritability, rather than considering maternal permanent environmental effects. Thus, it seems that variance of direct additive genetic effects is more broken into σ2m rather than σ2c. Also, a comparison between Log L values of models 2 and 3, demonstrated that the inclusion of maternal permanent environmental effects resulted in a more suitable model rather than maternal genetic effects.
Comparing model 1 vs. 3 and model 2 vs. 5 indicated that inclusion of maternal genetic effects in the animal model resulted in 0.054 and 0.02 reductions in direct heritabilities of milk and fat yields, respectively. Albuquerque et al. (1998) reported increases by 0.014 and 0.021 in direct additive heritabilities by the removal of maternal genetic effects and covariance for maternal-direct effects from the animal model for milk and fat yields, respectively. Comparisons between models (1, 2) models (3, 5) and models (4, 6) showed that the inclusion of maternal permanent environmental effects in the model on average results in 0.01 reduction in milk heritability; while, there would be 0.01 increase in fat heritability. On average, the inclusion of direct-maternal genetic covariance in the model (from model 3 to 4 and from model 5 to 6) resulted in 0.01 reduction in direct heritabilities of the traits. Also, differences between Log L values of models 3 vs. 4 and between models 5 vs. 6, showed that direct-maternal genetic covariance did not have any important effect on the model fitness and animal models which only differ in the existence of direct-maternal genetic covariance do not seem to have any significant differences. Due to the low impact of direct-maternal genetic covariance on heritabilities and the animal model fitness, as well as leading to more computational time and process, inclusion of this covariance in the animal models (models 4 and 6) is not recommended for milk production traits in dairy cattle. Therefore, it could be concluded that Model 5 is the best model for genetic evaluations of the studied traits.
In this study, maternal heritabilities (m2) were found to be around 0.07 and 0.01 for milk yield and fat yield, respectively. So, it can be concluded that maternal genetic effects are more important for milk than fat. Albuquerque et al. (1998) analyzed a dataset that consisted of lactation records for New York Holstein cows and obtained maternal heritabilities of 0.008 and 0.006 for milk and fat yields, respectively. Schutz et al. (1992) using an animal model found ratios of maternal genetic effect variances to phenotypic variances 2.58% and 6.5% for milk and fat yields, respectively. Maternal heritabilities decreased by the inclusion of direct-maternal genetic covariances in the model (from model 3 to 4 and model 5 to 6) which may be due to absorption a part of maternal additive genetic variance to create direct-maternal genetic covariance. This hypothesis is supported by the reduction of σ2m from model 3 to 4 and model 5 to 6. These low maternal heritabilities showed that a slight proportion of phenotypic variation in the progeny was attributed to the genetic variation in dams; and the maternal genetic variation did not have any important influence on phenotypic variation in the progenies, especially for fat yield. Another consideration is that each cow can only calve once a year, but because of widespread use of Artificial Insemination (AI), each bull can have many more calves per year and bulls contribute more to the gene pool rather than cows. In addition, the selection intensity achieved on bulls is greater than for cows, therefore the phenotypic variation in progeny is more influenced by genetic variation in bulls rather than cows and cows have a smaller share in both genetic and phenotypic variations. In animals like sheep and to some extent beef cattle in which the usage of artificial insemination is fewer than dairy cattle, the role of maternal genetic effects is more evident; for instance, Duangjinda et al. (2001) estimated 0.12, 0.07 and 0.15 as maternal heritabilities for Hereford, Gelbvieh and Charolais weaning weights.
The mean of direct-maternal genetic covariance as a proportion of total (am) was estimated 0.025 and 0.016 for milk and fat yields, respectively, which were both positive. According to Albuquerque et al. (1998) some confounding between direct and maternal genetic effects should be expected because the dam that contributes the maternal genetic effects also transmits half of her genetic value for direct effects to her daughter. On average, the ratio of maternal permanent environmental variance to the phenotypic variance (c2) was 0.02 and 0.05 for milk and fat yields, respectively. So, it can be concluded that maternal permanent environmental effects are more important for fat than milk. Although these ratios are very low for both traits, this situation was not really unexpected, because in most dairy industry systems in the world, including Iran, dairy calves are separated from their dams at birth, so the influence of maternal permanent environment does not seem to have any important effect on the progeny and would be effective only during pre-birth period. Such effects would include: dam's health, dam's nutritional state, age of dam, intrauterine environment and capacity, transmitting immunity from dam to its offspring and the post-birth effect of colostrums consumption by the calf from birth to 24 h. It seems that these effects can affect more on livability of dairy calves rather than their subsequent lactation performances. Albuquerque et al. (1998) pointed out that for dairy calves maternal effects act only through intrauterine environment. Dairy calves are deprived from post-birth environmental effects provided by their dams, like maternal care and maternal ability. It is evident that for other mammals such as sheep or beef cattle that the offspring remain with their dams to the weaning age, these effects are more important. Duguma et al. (2002) reported that maternal effects may be expected to be more important in sheep than cattle because of the greater relative variation in litter size in sheep and the competition between lambs for their mother's milk supply. They obtained ratios of maternal permanent environment variances to phenotypic variances ranged from 0.1 to 0.27 and 0.06 to 0.1 for birth weight and weaning weight, respectively in Merino sheep. For weaning weight in Hereford, Gelbvieh and Charolais cattle breeds the ratios were 0.13, 0.15 and 0.15, respectively (Duangjinda et al., 2001).
Overall, the results of this study showed that maternal effects are not very important for milk production traits in Holsteins, which are in agreement with the previous results (Albuquerque et al., 1998; Schutz et al., 1992; Van Vleck and Bradford, 1966; Van Vleck and Hart, 1966). However, they can improve the accuracy of animal models.
CONCLUSION
One of the most important practices in animal breeding strategies is to find more accurate and practical models for estimating genetic and environmental parameters. This study provides useful information on operational models and genetic parameters based on the actual data needed for predicting breeding values. Animal models are more suitable than sire models; because of considering all pedigree relationships rather than only through sire lines. However, animal models are differentiated by considering or ignoring some genetic and environmental effects. A notable feature of the results of this study was that, model 5 was the best fitted model for genetic analyses for the studied traits. However, time and computational limitations should be considered. The results indicate that maternal additive genetic and maternal permanent environmental effects and direct-maternal genetic covariance do not seem to have any considerable effect on milk and fat yields; probably because dairy calves do not benefit from the maternal effects. The inclusion of direct-maternal genetic covariance in the model had a slight impact on heritability estimates and model fitness; therefore, introducing this effect in the animal model is not recommended. Although, in this study the effects of maternal additive genetic and maternal permanent environmental parameters were not found to be important, the inclusion of these effects could improve the accuracy of genetic evaluations through improving the model fitness. So, the inclusion of these effects is recommended to improve genetic evaluations for milk production traits in dairy cattle in situations where there is no time and computational limitations (e.g., no very large data and multi-trait or repeated records analyses).
ACKNOWLEDGMENTS
Financial support of Isfahan University of Technology (Project AGA812) is acknowledged.