An evaluation of maternal in fl uences on growth traits : the Zandi sheep breed of Iran as an example

An animal model was used to estimate direct and maternal genetic parameters for birth weight (BW), weaning weight (WW), 9-month weight (W9) and yearling weight (YW) for a population of Zandi sheep. The magnitude of estimates of these parameters was used to evaluate the importance of direct and maternal infl uences on body weight at various stages of Zandi lambs’ life. Results confi rmed the potential maternal infl uences to change pattern of body weight’s heritability with age. Maternal infl uences were signifi cant for early growth traits. The importance of maternal infl uences diminished with age, although for W9 and YW those effects did not disappear completely. In contrast to variation due to maternal effects, the variance of direct effects increased from birth to a year of age. Estimates of maternal breeding values averaged by year of birth decreased throughout the experiment for both BW and WW. It was concluded that for more precise designing selection programs and, in consequence, increasing effi ciency of selection, inclusion of maternal effects as well as direct-maternal genetic covariance into the model, especially for early growth traits, is necessary.


INTRODUCTION
Most livestock are mammals with long periods of maternal dependence, so that an offspring's phenotype is determined not only by its own genotype and the random environmental conditions it experiences during development, but also by the environment provided by its parents.The infl uence of the mother on her offspring has been apparent since the beginning of animal husbandry.Many erroneous ideas have been handed down over the years, particularly with regard to maternal impressions.For example, one of the earliest recorded statements is the belief that effects due to events which the mother was subjected to during mating or pregnancy are transmitted to the foetus as maternal infl uences and affect pre-and post-natal phenotypes of the offspring (Legates, 1972).Now, maternal effects can be described as any contribution of a mother that affects the phenotypic expression of her offspring.The mother's milk production, health status, size, intrauterine environments, and mothering ability are all common components which are controlled by her genotype and associated environmental factors (Maniatis and Pollott, 2003).Therefore, maternal effects are divided in to genetic and environmental components.However, for the offspring, maternal effects are considered to be environmental.So, there are indirect genetic (heritable maternal effects) and environmental effects (Szwaczkowski et al., 2006).Before the 1990s, due to computational limitations and other problems in estimating maternal components (Willham, 1980), for most studies of growth, maternal effects were ignored from the analyses and only direct effects considered (see e.g., Atkins, 1986).With increasing computing power and availability of REML algorithms for analyses to fi t an animal model has been simplifi ed estimation of (co)variance components due to maternal effects (Meyer, 1997).Hence, during the last decade, most estimates of (co)variance components have been obtained for various sheep breeds using an animal model incorporating maternal effects (Snyman et al., 1995;Van Vleck et al., 2003;Ekiz, 2004;Szwaczkowski et al., 2006;Miraei-Ashtiani et al., 2007).These studies have shown the biases that are introduced by using the simple direct additive genetic animal model and the importance of modeling maternal inheritance in order to obtain accurate estimates of (co)variance components.However, in most studies, estimates of genetic parameters due to maternal effects have generally been restricted to traits describing growth till weaning.
The aims of the present paper were to investigate the importance of direct and maternal effects on body weight at different ages of Zandi lambs and to develop a suitable model for the data set used.

Data and management
Data and pedigree information on Zandi sheep collected at the Breeding Station of Zandi sheep over the period from 1987 to 2005 were used in this research.The station is situated in Khojir national park between Tehran and Abali at 35 °45´E and 51 °40´N, 1547 m above see level.Only pedigree information was recorded prior to 1990.In the experiment, maiden ewes were exposed to rams at about 18 months of age and kept in the fl ock until death or the apparent infertility.Rams were used for 3 or 4 breeding years and kept separated from ewes, except in the mating season.In the mating season (commenced in August), initially artifi cial insemination (AI) was performed, but animals that did not conceive by AI were allocated to natural service.In the latter case, each group of the ewes (detected in oestrus) was allotted to one fertile ram in a separate mating pen.At birth, lambs were weighed, tagged, sexed, and identifi ed to their parents.Birth date was also recorded.During the suckling period, lambs were kept indoors and allowed to nurse their mothers twice a day.The suckling stage lasted for 90 days on average.Animals were kept on natural pasture during the spring, summer and autumn seasons and indoors during the winter.Selection was based on weaning weight and, to some extent, body conformation.
The traits analysed were birth weight (BW), weaning weight (WW), body weight at 9 months of age (W9) and yearling weight (YW).Age at weighing was used as a covariate to account the differences in age at each time of weighing.Characteristics of the data used in the analyses are given in Table 1.

Statistical analysis
The fi xed part of the models included birth year (15 years), season of birth (winter and spring), sex (male and female), type of birth (single, twin and triplet) and age of dam at lambing (2-10 years old).Least square analyses using the GLM procedure of SAS ( 2004) showed that all these fi xed effects were signifi cant (P<0.05) for all traits.These effects then were included in the linear mixed models.The same fi xed effects were used for models with different random effects.A series of six single-trait linear animal models were considered that differed in the (co)variance components fi tted to assess the importance of maternal effects.The models expanded from a simple animal model (Model 1) to a comprehensive maternal effects model (Model 6).The random models used are summarized in Table 2. Depending on the model fi tted, direct additive  ), maternal additive genetic variance (σ 2 m ), covariance between direct and maternal genetic effects (σ a,m ), residual variance (σ 2 e ), phenotypic variance (σ 2 p ), direct heritability (h 2 ), maternal heritability (m 2 ), the ratio of maternal permanent environmental variance to phenotypic variance (c 2 ), and correlation between direct and maternal genetic effects (r am ) were estimated, accordingly.
Estimates of (co)variance components, and log likelihoods values, were obtained using a derivative-free REML algorithm as implemented in the DFREML program (Meyer, 2001).In fi tting the models, the Simplex procedure was employed to locate the maximum of the log likelihood (log L).Convergence was considered to have been reached when the variance of function values (-2 log L) in the Simplex was less than 10 -8 .Standard errors were obtained for the estimated parameters from a part of the DFREML program.
The information criterion (AIC) of Akaike (1977) was computed to rank the models according to their power to fi t the data.The criterion for model i is defi ned as AIC i = -2 log L i + 2 (number of independently estimated parameters i ), which log L i denoting maximum of the like lihood of model i.The model yielding the smallest AIC fi ts the data best.
Estimates of maternal breeding values were obtained with Best Linear Unbiased Prediction (BLUP) method based on a two-trait animal model for the traits which signifi cantly infl uenced by maternal effects (BW and WW).Means for estimated maternal breeding values of the animals by year of birth were calculated to evaluate the maternal genetic trends.

RESULTS
Estimates of genetic parameters for each pair of traits (related to early and later stages of lambs' life) obtained via the six models with corresponding AIC value for each analysis are summarized in Tables 3 and 4, respectively.For all traits, estimates of direct heritability were infl uenced by the statistical model.With the simple animal model (Model 1), h 2 estimates were overestimate; but when maternal effects were included in the models, estimates of h 2 decreased, especially for BW and WW.For BW and WW models 5 and 4, respectively, had lowest AIC.Latter result shows that maternal effects on body weight early in life are signifi cant.The inclusion of maternal genetic effects in the model (Model 3), led to a greater decrease in direct heritability than when maternal permanent environmental effects were in the model (Model 2).Hence, the maternal genetic infl uence of the dam was determined to be the most important maternal effect.At birth, maternal heritability (m2 ) was greater than direct heritability (0.12 vs 0.11, Model 5), diminished to 0.02 for weaning weight (Model 4).Likewise, the importance of maternal permanent environmental effects (c 2 ) decreased with age at weighing.Fitting the direct-maternal genetic covariance in the models (Models 4 and 6, respectively) resulted to negative estimates of directmaternal genetic correlation (r am ) for BW and positive estimates for WW.For W9 and YW, although maternal effects were present, but as clearly demonstrated by AIC values, those effects did not signifi cantly increase the likelihood over that for a simple animal model.On the other hand, sampling errors of direct heritability increased when maternal effects were estimated for W9 and YW.These effects, therefore, should not be included if they have no signifi cant infl uence on the interested trait.

DISCUSSION
Numerous studies have investigated the importance of maternal effects for various mammalian species.For livestock species such as sheep, where there is -direct heritability; m 2 -maternal heritability; c 2 -maternal permanent environmental variance as proportion of phenotypic variance; r am -direct-maternal genetic correlation; h t variation in litter size and suckling period is substantial, maternal effects would be more important than for cattle or swine (Bradford, 1972).As AIC values clearly indicate, maternal effects are important for BW and WW.For W9 and YW maternal effects did not disappear completely, although they were not signifi cant.This may be due to carry-over effects from weaning weight.Such carry-over of maternal effects beyond weaning are biologically sensible; particularly when lamb growth to weaning is restricted by low dam milk production either due to high litter size or seasonal constraints on feed (Bradford, 1972;Snyman et al., 1995).
In the absence of maternal effects, estimates of heritability were 0.33, 0.16, 0.22 and 0.26 for BW, WW, W9 and YW, respectively.This shows that in case of body weight the genetic control diminishes with age.Dobek et al. (2004) reported similar fi ndings.The general paucity of literature on the subject of estimates of heritability for models including or not including maternal effects in sheep over a similar range of ages makes comparisons diffi cult.Bosso et al. (2007) studied the growth traits in Djallonke sheep and with analyses not including maternal effects reported heritability estimates of 0.50, 0.43 and 0.30 for body weight at birth, 120 and 360 days of age, respectively.In addition, while a study by Dobek et al. (2004), also ignoring maternal effects, resulted in heritability estimates of 0.31 and 0.28 for birth weight and body weight at four weeks of age for a population of pure and crossbreeding lambs, Szwaczkowski et al. (2006) who worked on the population studied by Dobek et al. (2004), of course with a larger data set, for analyses that contained maternal effects, reported direct heritability estimates of 0.21 and 0.22 for body weight at similar ages.Moreover, Bahreini-Behzadi et al. (2007) studied the importance of direct and maternal effects on body weight at different ages of Kermani sheep and with a model not including maternal effects reported heritability estimates of 0.62, 0.59, 0.66, 0.41 and 0.27 for birth weigh, weaning weight, 6-month weight, 9-month weight and yearling weight, respectively; while when maternal effects were considered, those estimates reduced to 0.10, 0.22, 0.09, 0.13 and 0.14, respectively.These reports provide evidence that disregarding maternal effects when they are actually present will cause to an upward bias in estimates of direct heritability because estimated additive genetic variance will capture at least part of the maternal variation.
When maternal effects are included in the analyses and so that maternal components of variance can be separated from the direct component, a new parameter which has been called ''direct heritability'' can be extracted.In that case direct heritability increased with age.All else being equal, maternal effects will change the general pattern of body weight's heritability with age when they allowed being presence in the analyses.
Maternal genetic effects arise from allelic differences between individual mothers at loci infl uencing offspring phenotype, and are themselves a heritable component of phenotypic variance (Wolf et al., 1998).Maternal heritability was highest at birth and comparable with direct heritability, but decreased with age and for W9 and YW was small in comparison with direct heritability.With respect to the magnitude of maternal heritability, Miraei-Ashtiani et al. (2007) reported a surprisingly greater estimate of m 2 for birth weight (0.65) compared with the estimates for weaning weight (0.08) and body weight at 6 months of age (0.11).In the current study, the low estimate of maternal heritability for WW may be due to pre-weaning feed conditions.Lambs receiving limited supplies of ewe's milk may be encouraged to start eating creep feed earlier.In such a situation, the importance of milk yield of dams would decrease more rapidly when lambs were offered a high quality pasture (in the spring) or grass hay and concentrate (in the winter).This presumably would lead to a smaller estimate of maternal heritability at weaning.
Where maternal genetic effects are present, the potential response to selection might better be expressed by ''total heritability'' (h t 2 ) defi ned as the ratio of the sum (σ 2 a + 0.5σ 2 m + 1.5σ a,m ) to phenotypic variance (Willham, 1972).Since maternal genetic effects are heritable, the h t 2 of weight may be somewhat higher than the estimates of direct heritability reported, unless the direct-maternal genetic covariance is large and negative.Total heritability has the intuitive interpretation of describing total potential response to selection, replacing direct heritability when maternal genetic effects are present (Wilson et al., 2005).Presence of maternal genetic variance would increase the total heritability and hence the potential response to selection.Therefore, as a result, phenotypic response to selection may be accelerated by maternal genetic effects, but can also be dampened, if there is negative covariance between direct and maternal genetic effects (Wolf et al., 1998).At a mechanistic level such a negative covariance might occur through pleiotropy, for example if a gene has a positive effect on an offspring trait but a negative effect on maternal performance for that trait (Wilson et al., 2005).
Estimates of maternal breeding values, as an indication of maternal performance, averaged by year of birth for BW and WW are shown in Figure 1.For both BW and WW, means of estimates of maternal breeding value by year of birth decreased over time.Young and Legates (1965) and Legates (1972) reported that if the correlation between direct and maternal genetic effects is negative, maternal performance may decline under continuous selection for growth to the extent that pre-weaning viability and growth may be retarded.Such retardation, if extreme, would not likely be offset by compensatory growth during a later period despite the presence of genes for rapid growth.From a genetic analysis on growth traits of beef cattle, Vesely and Robison (1971) showed that selection effectiveness for body weight at 205 days was reduced by approximately 20% due to the negative r am .For WW, because maternal genetic variance is low, the strong positive correlation (+0.99) between direct and maternal genetic effects must be an overestimate, makes interpretation diffi cult.
When the genetic correlation between growth and maternal performance is strongly negative, other breeding methods might be more desirable for overcoming the genetic antagonism between the two traits.With a large negative covariance, selection for maternal performance in one line and for growth in a second with eventual crossing has been suggested in such situations, although its applicability depends upon the nature of the action of the genes infl uencing the traits and the reproductive pattern of the species (Young and Legates, 1965).On the other hand, a negative correlation between direct and maternal genetic effects indicates that individuals will be ranked differently when the maternal contribution is omitted in the selection procedure (Szwaczkowski et al., 2006).Therefore, knowledge of the relationship between growth and postnatal maternal performance is most important for an accurate interpretation of selection response and in devising breeding programs for traits affected by maternal infl uences.
Maternal permanent environmental component was more important for BW than other weights.The permanent environmental effect of the dam on birth weight is mainly determined by uterine capacity, feeding level at late gestation and the maternal behaviour of the dam.Estimation of c 2 is necessary to obtain accurate estimates of m 2 .Snyman et al. (1995) and Ekiz (2004) indicated that the exclusion of the maternal permanent environmental effects could cause maternal heritability to be overestimated.

CONCLUSIONS
Maternal effects play an important role in the early growth of Zandi lambs.Those effects diminish after weaning, when the infl uence of genes of the lamb increases, but due to carry-over effects, some maternal variation might persist in the post-weaning growth period.Due to the fact that signifi cant variation in body weight, particularly before weaning, arise from maternal effects and possibly a negative correlation between direct and maternal genetic effects, genetic evaluation should be based on a model including maternal effects.It would seem that an index using estimated breeding values of all traits weighed by their net economic values is needed.For example, increased lambing diffi culty associated with greater BW might lead to negative economic values for both direct and maternal genetic values.Birth weight by itself has no direct economic value.Its value comes indirectly from lambing diffi culty and/or survival.Finally, the difference in estimates of genetic parameters determined in different models shows that model choice is an important aspect for obtaining accurate estimates of breeding values that are going to be used when deciding on a breeding program.

Figure 1 .
Figure 1.Maternal genetic trends for birth weight (BW) and weaning weight (WW)

Table 1 .
Characteristics of the data set

Table 2 .
(Co)variance components fi tted to models used in the analyses Model

Table 3 .
Genetic parameters a ± SE for birth weight (BW) and weaning weight (WW) b AIC -Akaike's information criterion