Estimation of (co) variance components and genetic parameters for growth traits in Arman sheep

Introduction

Small ruminants play an important role in the livelihood of a sizeable portion of the human population in the tropics, where the small ruminants are mainly kept by local pastoralists under low-input production systems. Therefore, coordinated attempts, in terms of managerial practices and genetic improvement to promote production efficiency, are of crucial importance (Kosgey and Okeyo, 2007). Mutton production in Iran, as the main source of red meat, does not meet the increasing demand. Iranian native sheep are mainly kept under low-efficient production systems, relying on limited low-quality rangelands. Such low efficiency necessitates designing appropriate breeding programs for improved performance of lambs as an appropriate alternative for enhancing meat production. Sheep has relatively high biodiversity in the Middle East. For example, sheep population in Iran comprises about 26 breed populations. Such genetic diversity provides favorable opportunities for enhancing production efficiency by means of crossbreeding strategies that exploit breed diversities, heterosis, and breed complementary (Freking and Leymaster, 2004). There has been a revival of interest in the creation of composite populations from crossbreeding of prolific exotic sheep breeds with local breeds, with the aim of exploiting production efficiency to the benefit of the livestock industries (Shrestha, 2005).

Arman sheep was synthesized by crossbreeding of four sheep breeds including Baluchi, Ghezel (two Iranian native breeds), Chios and Suffolk in Abbasabad breeding station, located in Khorasan Razavi province, north-eastern of Iran. The aiming of improvement was mainly increasing litter size, mutton production and tolerance to harsh and unfavorable environmental conditions, prevalent in the area. The project was started from the 1975. As breed fixation was obtained by selection and inbreeding.

Growth rate and body weight of lambs at different ages have deterministic effects on the profitability of sheep production enterprises. Therefore, these traits may be taken into account as efficient selection criteria in any sheep breeding system. Riggio et al. (2008) estimated genetic parameters on body weight of Scottish Blackface sheep and suggested that taking live body weight on older lambs into account as selection criterion would increase selection accuracy. Such appropriate selective procedure requires accurate estimates of (co)variance components and genetic parameters. Genetic parameters for growth traits of different sheep breeds have been reported (Safari et al., 2005; Miraei-Ashtiani et al., 2007; Rashidi et al., 2008; Mohammadi et al., 2010; Gowane et al., 2010). The results of these studies indicated that the inclusion of maternal effects in the models considered for genetic evaluation of growth traits, especially for pre-weaning traits, is of crucial importance. They showed that the exclusion of maternal effects leads to upward biased estimates for (co)variance components. Thus, accurate estimation of (co)variance components is a perquisite for designing any breeding program and genetic evaluation system.

Because of paucity of such estimates for growth traits of Arman sheep, as a new composite sheep breed in Iran, the objective of the current research was to estimate the (co)variance components and corresponding genetic parameters for growth traits in Arman sheep.

Materials and methods

Flock management

Animals were kept under conventional managerial practices similar to local flocks. Breeding season extends from late August to late October. Therefore, lambing occurs late in January to late March. Breeding rams and ewes were selected mainly based on phenotypic appraisal such as visual body conformation at yearling age and nearest three generations of pedigree information on birth type of their lambs. Maiden ewes were exposed to fertile rams at approximately 18 months of age under a fully supervised mating strategy. The estrous ewes were detected by teaser rams, at a ratio of 20-25 ewes per ram. The ewes were kept for a maximum of 7 parities (until 8 years of age) and the rams for a maximum of 2 mating seasons. To avoid inbreeding, rams were allocated rotationally to each group of ewes. Lambs were ear-tagged and weighed at lambing or within 24 h of birth. The ewes and their lambs were housed in separate pens for a few days. The lambs were creep-fed and grazed on the range. They kept together until weaning time; approximately 3 months of age. All lambs were weaned on the same day, but not necessarily at the same age. During spring and summer, the flock was kept in pastures and in the autumn it was grazed on wheat and barley stubbles. During winter, the lambs were kept indoors and hand- fed. Supplementary feeding was offered to all animals during winter and to ewes late in pregnancy including a ration composed of wheat and barley straw, alfalfa hay, sugar beet pulp and concentrate.

Studied traits

The data set used in the present study was collected during 11years period, from 1999 to 2010, at Abbasabad Sheep Breeding Station, Khorasan Razavi province, located in north-eastern Iran. Investigated traits were body weights of lambs at birth (BW), body weight at 3 months of age as weaning weight (WW), body weight at 6 months of age (6MW), body weight at 9 months of age (9MW), yearling age (YW), average daily gain from birth to weaning (ADG) and the Kleiber ratio (KR) from birth to weaning; defined as ADG/WW^0.75. The structure of the data set used in the present study is set out in Table 1.

Statistical analysis

Fixed effects

Significance testing of fixed effects to be included in the operational model for each trait, was carried out using the general linear model (GLM) procedure (SAS Institute, 2002), and the least squares means of the traits were obtained. Considered fixed effects in the model were gender of lamb (male and female), birth year in 11 classes (1999–2010), dam age at lambing in 6 classes (2-7 years old), birth type in 3 classes (single, twin and triplet>) and age of lamb at 3, 6, 9 and 12 months (in days) as a linear covariate for WW, 6MW, 9MW and YW, respectively. The interactions between fixed effects were not significant and therefore excluded.

Estimation of (co)variance components and genetic parameters

The restricted maximum likelihood (REML) procedure, under a derivative free algorithm, was used to estimate the (co)variance components and corresponding genetic parameters applying WOMBAT program (Meyer, 2006). Tested models (in matrix notation) were as follows:

In which, y is a vector of records for studied traits; b, a, m, c, l and e are vectors of fixed, direct genetic, maternal genetic, maternal permanent environmental, maternal temporary environmental (common litter) and the residual effects, respectively. X, Z₁,Z₂,Z₃ and Z₄ are corresponding design matrices associating the fixed, direct genetic, maternal genetic, maternal permanent environmental and maternal temporary environmental effects to vector of y. Temporary environmental effects, also called common environmental effects, are a portion of maternal environmental effects that is common among full-sibs in a certain year and differ throughout years.

It was assumed that direct additive genetic, maternal additive genetic, maternal permanent environmental, maternal temporary environmental and residual effects were normally distributed with a mean 0 and variance of Aσ²_a, Aσ²_m, I_dσ²_c, I_lσ²_l and I_nσ²_e, respectively. Additionally, σ²_a, σ²_m, σ²_c, σ²_land σ²_e are direct additive genetic variance, maternal additive genetic variance, maternal permanent environmental variance (half sibs across years), maternal temporary environmental variance (full sibs within a year) and residual variance, respectively. A is the additive numerator relationship matrix, I_d, I_l and I_n are identity matrices that have order equal to the number of dams, litters and number of records, respectively, and σ_am refers to the covariance between direct genetic and maternal genetic effects.

An Akaike's information criterion (AIC) test was applied to determine the most appropriate model for estimating of (co)variance components for each trait as follows (Akaike, 1974):

AIC_i = -2 log L_i + 2 p_i

In which, log L_i is the maximized log likelihood of the respective modeli at convergence and p_i is the number of parameters obtained from each model; the model with the lowest AIC was considered the most appropriate model. Total heritability () for the studied traits was calculated as the following”

= (σ²_a +0.5σ²_m +1.5σ_am) / σ²_p

In which σ²_a, σ²_m,  σ_am  and σ²_p are direct additive genetic variance, maternal additive genetic variance phenotypic variance, covariance between direct and maternal additive genetic variance and phenotypic variance, respectively.

 Maternal across year repeatability for ewe performance (t_m) was estimated as follow (Gowane et al., 2010):

t_m =1/4+++ (m r_am h)

In which, , , , r_am , m and are denote direct heritability, maternal heritability, ratio of maternal permanent environmental variance to phenotypic variance, correlation between direct and maternal additive genetic effects, square root of direct heritability and square root of maternal heritability, respectively. Genetic and phenotypic correlations were estimated using bivariate analyses applying the best model, determined in univariate analyses. When the value of -2 log likelihood variance in the AIREML function was below 10⁻⁸; convergence was assumed to have been achieved.

Results

Fixed effects and model comparisons

 The descriptive statistics are summarized in Table 1. Approximately 9.25 % of the lambs lost from birth to weaning age. Least squares means of the traits are shown in Table 2. Singleton, twin and triplet lambs constituted 29.6%, 59.0% and 11.4% of total lambs, respectively. As indicated in Table 2, birth year, gender and birth type of lambs had significant effects on all studied traits (p0.05). Age of lambs (in days) at 3, 6, 9 and 12 months of age as a linear covariate, significantly influenced the WW, 6MW, 9MW and YW, respectively.

The AIC values under different tested models are presented in Table 3. The most appropriate model for BW and WW included direct additive genetic, maternal additive genetic, maternal permanent environmental and common litter effects, without considering covariance between direct additive and maternal additive genetic effects (Model 8). The most appropriate model for ADG was Model 6; including direct additive, maternal additive and maternal permanent environmental effects. The Model including direct additive genetic effects and maternal permanent environmental effects (Model 2) was determined as the best model for KR and 6MW. Maternal effects did not influence the 9MW and YW; resulting in selection of the simplest model, which included direct additive genetic effects as the sole random effects, for 9MW and YW.

Estimates of genetic parameters

Estimation of genetic parameters, based on the best model under univariate analyses, is shown in Table 4. Total heritability estimates ranged from 0.04 for KR to 0.19 for WW and ADG; while those related to repeatability of ewe performance varied from 0.02 for 9MW to 0.26 for BW. Correlation estimates are set out in Table 5. Direct additive genetic correlations were positive and ranged from 0.08±0.10 for KR-YW to 0.83±0.18 for WW-ADG. Maternal additive genetic correlations were relatively medium estimates; ranging from 0.21±0.08 for BW-ADG to 0.38±0.21 for BW-WW. Maternal permanent environmental correlation estimates were low (0.09±0.12 for BW-6MW) to high (0.72±0.15 for BW-ADG). A high estimate of 0.85±0.26 was obtained for maternal temporary environmental correlation between BW and WW. Phenotypic correlations were ranged from 0.19±0.11 (KR-9MW) to 0.96±0.14 (WW-ADG) and environmental ones from 0.08±0.09 (KR-YW) to 0.78±0.14 (WW-ADG).

Discussion

Superiority of male lambs to female lambs in terms of studied traits can be partly ascribed to differences in endocrine system of male and female lambs that tends to become more pronounced as lambs approach maturity (Matika et al., 2003; Yilmaz et al., 2007). The significant effects of dam age on the studied traits can be explained to some extent by differences in maternal effects and maternal behavior of ewes at different ages (Abbasi et al., 2011). Differences in managing practices, feed availability, climatic conditions and breeding systems through years, are possible reasons for significant effects of year on the considered traits (Matika et al., 2003). Competition for milk consumption among the twins and triplets leads to significant effect of birth type of lambs on the pre-weaning studied traits (Yilmaz et al., 2007). The significant effect of birth type on 6MW, 9MW and YW may be due to the existence of twin and triplet lambs at post-weaning period (at 6, 9 and 12 months of age).  

As indicated in Table 2, BW and WW of twins were higher than those of triplets but there were no significant differences among twin and triplet lambs in terms of 6MW, 9MW and YW. Significant influences of fixed effects on body weight of different sheep breed have been well documented (Yazdi et al., 1997; Abegaz et al., 2005; Rashidi et al., 2008; Jafaroghli et al., 2010). 

Duguma et al. (2002) pointed out that if maternal effects constitute a sizable part of genetic variation ignoring those results in upward biased estimates. In the present study, most of the dams had their own records, approximately 74 % at birth to 51% at yearling age (Table 1). Maniatis and Pollot (2003) remembered that the accuracy of partitioning maternal effects into genetic and environmental components may be affected by the number of records per dam and the proportion of dams with own records.

Estimated value for direct heritability of BW was in agreement with the estimates of Rashidi et al. (2008) in Kermani sheep and Jafaroghli et al. (2010) in Moghani sheep. The low direct heritability estimate for BW denotes the fact that direct genetic effects constitute a negligible portion of the phenotypic variance for BW of Arman lambs; suggesting that slow genetic progress would be expected through direct selection. Such low direct heritability is possibly due to the inclusion of maternal effects in the selected model. Contrary to the present findings, Safari et al. (2005) reported weighted mean of direct heritability, obtained from the literature, for BW of meat, dual-purpose and wool type sheep at 0.15, 0.19 and 0.21, respectively.

Estimated direct heritability values for WW and ADG were relatively similar in magnitude. Direct heritability estimate of WW accords well with literature (Miraei-Ashtiani et al., 2007; Gowane et al., 2010; Mohammadi et al., 2010; Mohammadi et al., 2011). The obtained value for direct heritability of ADG was generally accord with estimates of Rashidi et al. (2008) in Kermani sheep (0.15) and Mohammadi et al. (2011) in Sanjabi sheep (0.14). Safari et al. (2005) reported weighted mean estimate of 0.18 for direct heritability of WW in both dual-purpose and meat-type breeds of sheep which are in general agreement with the estimated value in the present study. A low direct heritability value was estimated for KR (0.04); which generally accords well with those obtained by Rashidi et al. (2008) in Kermani sheep and by Matika et al. (2003) in Sabi sheep. The Kleiber ratio has been proposed as an efficient selection criterion for feed efficiency under low-input range conditions and provides good indication of how economically an animal grows (Scholtz and Roux, 1988).

At post-weaning period, direct heritability estimates decreased from 0.15 at 6 months of age to 0.08 at 9 months of age and increased to a value of 0.16 at yearling age. Direct heritability estimate value for 6MW was generally concordant with estimates of Vatankhah and Talebi (2008) in Lori-Bakhtiari sheep (0.19) and Abegaz et al. (2005) in Horro sheep (0.18). Higher (Mirae-Ashtiani et al., 2007; Mokhtari et al., 2008; Gowane et al., 2010) and lower (Mohammadi et al., 2010; Eskandarinassab et al., 2010) estimates were also reported. A low estimate of 0.08 obtained for direct heritability of 9MW; generally agreed with the value of 0.03 in Kermani sheep (Rashidi et al., 2008). Obtained direct heritability estimate of YW (0.16) generally agreed with estimate of Mokhtari et al. (2008) in Kermani sheep (0.15) and Miraei-Ashtiani et al. (2007) in Sangsari sheep (0.10). Higher estimates also were reported by others (Snyman et al., 1995; Yazdi et al., 1997; Abegaz et al., 2005).

Accurate genetic evaluation of growth traits requires adopting models that contain direct, maternal genetic and maternal environmental effects. Where multiple births are relatively common, partitioning maternal environmental effects into across year effect (maternal permanent environmental) and litter effect (within year common environmental effect specified to the litter) is of paramount importance in terms breeding (Safari et al., 2005). Safari et al. (2005) suggested that for traits affected by maternal effects, interpretation of genetic parameters under animal model is mainly dependent on both data structure and the analytical model used.

Maternal additive genetic effects disappeared after weaning. The estimated value for maternal heritability of BW was in concordance with estimates of Rashidi et al. (2008) in Kermani sheep (0.24) and Eskandarinassab et al. (2010) in Afshari sheep (0.22). Lower estimates also reported by Mohmmadi et al. (2010) in Sanjabi sheep (0.14) and by Mohammadi et al. (2011) in Zandi sheep (0.13). As expected, maternal effects constitute an integral part of variation for BW, probably reflecting differences in the uterine conditions, mainly with respect to the quality and capacity of the uterine space for growth of the fetus (Gowane et al., 2010). Maternal heritability estimate for WW was lower than direct one (0.13 vs. 0.15) and generally agreed with estimate of Zamani and Mohammadi (2008) in Iranian Mehraban sheep (0.08). Higher (Mohammadi et al., 2010) and lower estimate (Ozcan et al., 2005; Miraei-Ashtiani et al., 2007) were also reported. Low obtained value for maternal heritability of ADG was in general agreement with Ozcan et al. (2005) in Turkish Merino sheep (0.04) and Ghafouri-Kesbi et al. (2011) in Zandi sheep (0.03). Maternal permanent environmental effects influenced all pre-weaning traits and 6MW. These effects may be due to uterine environmental and multiple birth effects on milk production of ewes, level of nutrition at final stages of gestation and maternal behavior (Maria et al., 1993; Snyman et al., 1995). Estimated values for c² of BW, ADG and WW were generally in agreement with estimates of Abbasi et al. (2012) in Iranian Baluchi sheep. Mokhtari et al. (2008) reported a value of 0.09 for c² of 6MW in Kermani sheep which is in accordance with obtained value in the present study. Consistent with our estimates, Abegaz et al. (2005) reported value of 0.13 for l² estimate of weaning weight in Horro sheep. Abbasi et al. (2012) reported a value of 0.19 for l² of BW in Bauchi sheep which is higher than values recorded in the present study.

Total heritability estimates are model sensitive (Gowane et al., 2010). Abegaz et al. (2005) pointed out that when maternal effects are important in the expression of a trait total heritability is of crucial importance in terms breeding and is useful in estimation of selection response based on phenotypic values. The obtained estimates of  and t_m for BW and for WW were in general agreement with estimated values reported by Gowane et al. (2010) in Malpura sheep. Estimates of t_m for post-weaning body weights were generally higher than estimated values by Gowane et al. (2010) in Malpura sheep.

Birth weight had positive and low to medium direct genetic correlations with other studied traits, ranging from 0.09 (BW-KR) to 0.49 (BW-WW). Gowane et al. (2010) reported positive and medium direct genetic correlation estimates among BW and other body weights, including weaning weight, 6 months weight, 9 months weight and yearling weight, in Malpura sheep which were generally in agreement with estimates in the present study.

Direct genetic correlation estimate of BW with other body weight traits decreased with age. A low direct genetic correlation estimate was obtained for BW-KR which was in agreement with estimate of Mohammadi et al., (2010) in Sanjabi sheep (0.02). Medium estimate of 0.25 for direct genetic correlation of BW and ADG was in agreement with estimate of Mohammadi et al. (2010) in Sanjabi sheep (0.19) and Mohammadi et al. (2011) in Zandi sheep (0.21). Higher estimate were reported by Rashidi et al. (2008) in Kermani sheep.

Direct genetic correlation estimate between WW and ADG was high (0.83) and of similar magnitude to estimates of Duguma et al. (2002) in Tygerhoek Merino sheep (0.99) and Rashidi et al. (2008) in Kermani sheep (0.86). Duguma et al. (2002) pointed out that WW and ADG are genetically the same trait; thus selection can be carried out based on either one. Lower estimate of 0.59 for direct genetic correlation among WW and ADG also were reported by Maria et al. (1993). Medium and positive direct genetic correlations were found between WW and other traits (Table 5). Estimated values for direct genetic correlations of WW with 6MW, 9MW and YW generally agreed with estimates of Gowane et al. (2010) in Malpura sheep.

Weaning weight had a medium and positive direct genetic correlation with KR (0.55), the corresponding estimate was generally in agreement with estimate of Mohammadi et al. (2010) in Sanjabi sheep (0.73). A relatively large direct genetic correlation estimate was obtained for ADG-KR (0.75) which was in agreement with those obtained by Rashidi et al. (2008) in Kermani sheep (0.76) and Mohammadi et al. (2011) in Zandi sheep (0.84). Direct genetic correlations of post-weaning body weight traits with ADG were higher than those of post-weaning ones with KR. Similar pattern were recorded by Mohammadi et al. (2010) in Sanjabi sheep.

Estimated direct genetic correlations among post-weaning body weights (6MW, 9MW and YW) were positive and medium in magnitude and generally lower than other published estimates (Mokhtari et al., 2008; Mohammadi et al., 2010). In general, the existence of positive direct genetic correlations among the studied traits suggests that genetic factors which influence these traits were in similar direction. Positive estimates for maternal genetic correlation among pre-weaning traits (except for KR) indicated that maternal additive genetic effects, which favor the growth of fetus, could have some favorable effects on post-natal growth traits of Arman lambs. Body weight and growth rate from birth to weaning are influenced by similar genes in terms of maternal genetic effects. Obtained estimates of maternal genetic correlation were in agreement with estimates of Abbasi et al. (2012) in Baluchi sheep. Positive values were recorded for maternal permanent environmental correlations among pre-weaning growth traits and 6MW, suggesting that good management conditions and favorable maternal behavior would have beneficial effects on body weight of lambs from birth to 6 months of age (Gowane et al., 2010). Maternal temporary environmental correlations were high in magnitude and found only between BW and WW. Phenotypic and/or environmental correlation estimates among the studied traits generally agreed with those of Abegaz et al., (2005) in Horro sheep and Mohammadi et al., (2010) in Sanjabi sheep. Positive genetic (direct and/or maternal), phenotypic and environmental correlations among body weight traits indicated that there was no genetic, phenotypic and environmental antagonist relationship among considered traits. Therefore, selection for any of these body weights will bring about positive response to selection in terms of genetic and phenotypic values.

Conclusions

Different models for estimating of (co)variance components and genetic parameters have been compared. Low direct genetic variations were found for all studied traits. Thus, a relatively low genetic gain would be expected through mass selection. As Arman sheep is a crossbred one it seems that non-additive genetic effects have sizeable impacts on the expression of body weight and growth rate traits. The obtained results revealed that the maternal environmental effects should be portioned into permanent and temporary components until six months of age.

Acknowledgement

Assistance from the staff of Abbasabad Sheep Breeding Station for providing the data set used in the present study is deeply acknowledged. 

Table 1. Summary of descriptive statistics for the traits studied in Arman sheep

 ^a BW: Birth weight, WW: Weaning weight (3 months weight), ADG: Average daily gain from birth to weaning, KR: Kleiber ratio from birth to weaning, 6MW: Six-month weight, 9MW: nine-month weight, YW: Yearling weight

^b S.D.: Standard deviation, C.V. : Coefficient of variation

  Table 2. Least squares means ±S.E. for the pre-weaning and post-weaning growth traits in Arman lambs

Means with similar letters in each subclass within a column do not differ; *^, ** significant effect at P§ Regression coefficient of body weight on lamb age (in days). ^¥ For trait abbreviations see footnote to Table 1.

Table 3. AIC values for pre-weaning and post-weaning growth traits in Arman lambs under different univariate animal models

¥ The best model determined for each trait is shown in bold face

^¥¥ For trait abbreviations see footnote to Table 1

Table 4. Genetic parameter estimates for pre-weaning and post-weaning growth traits under the most appropriate univariate animal model

 : phenotypic variance;  : direct heritability; : maternal heritability; : ratio of maternal permanent environmental effects to phenotypic variance; : ratio of common litter effects to phenotypic variance; S. E.: standard error; =Total heritability = (σ²_a +0.5σ²_m +1.5σ_am ) / σ²_p ; t_m = ( 1/4+++m r_am h)

 ^¥ For trait abbreviations see footnote to Table 1

Table 5. Correlation estimates among studied traits under bivariate animal models

r_p: phenotypic correlation; r_a: direct genetic correlation; r_m: maternal genetic correlation r_c: maternal permanent environmental correlations; r _l: common litter effect correlation r_e: environmental correlation.

^¥ For trait abbreviations see footnote to Table 1