| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
ANIMAL GENETICS |
* Dipartimento S.En.Fi.Mi.Zo.–Sezione Produzioni Animali, Università degli Studi di Palermo, Viale delle Scienze–Parco d’Orleans, 90128 Palermo, Italy; and
Roslin Institute and Royal (Dick) School of Veterinary Studies, Roslin Biocentre, Roslin, Midlothian EH25 9PS, United Kingdom
 |
Abstract |
|---|
 Top Abstract INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: binary trait • genetic parameter • lamb survival • live weight
|
INTRODUCTION |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
). In a study of economic weights for hill sheep in the United Kingdom, Conington et al. (2004) found that lamb survival was one of the main influences on overall productivity. Further, Amer et al. (1999) demonstrated that for New Zealand sheep, the economic value of ewe prolificacy was greater for flocks with high rates of lamb survival.
For this reason, survival has been the objective of many studies, especially in countries where extensive husbandry systems predominate. In particular, sheep production is very extensive in the United Kingdom and New Zealand. In these countries, the death rate before weaning has been estimated at 5 to 40% (Eales et al., 1983) and 3 to 25% (Hight and Jury, 1970; Dalton et al., 1980), respectively. Therefore, the necessity of improving the lamb survival trait, and consequently the inclusion of this trait as a selection criterion in sheep breeding programs, appears clear.
The major causes of deceased lambs are starvation, mismothering, and exposure (Haughey, 1993). In addition, lamb mortality may vary because of location, birth type, year, and season of birth (e.g., Wilson and Murayi, 1988), and between and within breeds (Smith, 1977; Dalton et al., 1980). Some studies have also determined the relative importance of sex of lamb, age of dam, maternal behavior, genotype of parents, and birth weight on lamb survival in New Zealand flocks (e.g., Dalton et al., 1980; Hinch et al., 1983). Although many studies have been carried out on lamb survival, apart from Sawalha et al. (2007), there are few comparable published estimates of genetic and nongenetic parameters for this trait, and little has been published on the genetic correlations between lamb survival and growth. The objective of this study was to estimate the heritability of lamb survival and BW at different ages, and the genetic relationship between lamb survival and early growth in Scottish Blackface sheep.
|
MATERIALS AND METHODS |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Data Set
The data set comprised 4,459 records collected from 1988 to 2003 in a Scottish Blackface flock (Bishop, 1993) at Roslin Institute’s Blythbank Farm in the Border Region of Scotland. Animal identification, state of the animal (alive or dead), sex, line (i.e., pure Blackface or sire of different or unknown breed), date of birth and of death, dam, age of dam in years, sire, and litter size at birth were reported in the data. In addition, a group variable was created; from 1997 onward, the flock was split into 2 grazing groups: those born in the first 2 wk of the lambing season (classified as group 1) and those born subsequently (group 2). Before 1997, all animals grazed together.
Six binary traits were created, showing whether lambs were alive (1) or dead (0) at certain ages. First, survival was defined either by perinatal mortality (i.e., mortalities recorded on the day of birth, including still-births), or postnatal mortality (i.e., mortalities from d 1 to weaning at 12 wk). Second, cumulative survival to wk 1, 4, 8, and 12 was calculated. This trait is of considerable practical importance, because net economic loss increases with the age of the animal at death. Mortality after weaning was negligible and was not considered in this study. Records of live BW (kg) were available for most lambs, with measurements taken at birth, 4, 8, 12, 16, 20, and 24 wk. Lambs were not weighed at death, with the exception of stillborn lambs. Therefore, for a lamb that died at 6 wk, only BW at birth and at 4 wk were available for the analysis. The pedigree file comprised 1,416 dams and 178 sires. For the majority of lambs, both parents were known (95.9%). In a few cases, both parents were unknown (0.6%), and in other cases, the sire was unknown (3.5%). Records of lambs with both parents unknown were not considered in the analysis.
Statistical Analyses
For the statistical analyses, Genstat for Windows v. 7.0 software (GenStat 7 Committee, 2003) was used. Frequencies for the survival data were calculated to illustrate cumulative mortality across time (Table 1). Detailed survival data were not available for the 315 lambs born in 2003. Three kinds of analyses were undertaken on the data set, as described in the following sections.
|
A linear model was used to identify nongenetic effects on variation of lamb BW. The model was
where yijklmn was an observation on BW at birth, 4, 8, 12, 16, 20, or 24 wk; µ was the intercept of the model; ADi was the fixed effect of the age of the dam (i = 1, ..., 4, corresponding to 2, 3, 4, and 5 yr of age); Sexj was the fixed effect of sex (j = 1, 2); YBk was the fixed effect of the year of birth (k = 1, ..., 16); LSl was the fixed effect of the litter size (l = 1, 2, 3); (YB x LS x G)klm was the fixed effect of the year of birth x litter size x group interaction (m = 1, 2); DB was a covariate effect of day of birth and β1 was its regression coefficient; β2k was the regression coefficient for day of birth specific to year k (YB x DB)k; and eijklmn was the residual random effect.
Logistic Regression of Survival Traits.
Because survival traits were coded as binary traits, a generalized linear model analysis was performed, assuming a Bernoulli distribution of the data and using a logit link function. Several models were fitted and successively compared by using deviance values. In model 1, the same fixed effects used in the linear model were fitted. The other models were similar to model 1, except that in each model, all terms including 1 of 4 different fixed effects were eliminated (Sex, LS, AD, and YB from models 2, 3, 4, and 5, respectively) to determine the significance of the fixed effect using a 
2 test.
Estimation of Genetic Parameters.
Genetic parameters for lamb BW were estimated by using the ASREML package (Gilmour et al., 2000), fitting an animal model with all known pedigree relationships, including the fixed effects that were significant (P < 0.05) in the preliminary linear model analysis. Additional models were obtained by including a litter effect, a maternal genetic effect, or both litter and maternal genetic effects. Log-likelihood ratio tests were conducted to determine the most suitable model for each BW trait in univariate analyses.
Variance components and heritabilities for survival traits were first estimated by using a linear sire model accounting for the same fixed effects included in the linear models used for BW. A threshold sire model was then fitted, assuming a probit link function.
Phenotypic and genetic correlations for BW were estimated by using bivariate analyses. The same fixed effects as described previously were fitted in the model, using the animal as a random effect. Phenotypic and genetic correlations between survival traits and BW were estimated by using a Threshold Model (TM) program (available on request from the author at andres.legarra{at}toulouse.inra.fr), using a Bayesian analysis, and performing numerical integration through the Gibbs sampler. Models including the maternal effect, the litter effect, or both were also fitted to estimate the phenotypic and genetic correlations both for BW and between BW and survival; however, these analyses did not converge. Moreover, the TM program does not handle covariates. Therefore, in this case the model was simplified and the covariates of day of birth and day of birth x year of birth interaction were excluded. Flat priors were used for both fixed effects and variance components. A chain of 100,000 iterations was used, with a burn-in of 30,000 rounds, saving a sample every 100 iterations. The mean of the estimated marginal posterior density was used as a point estimate of the genetic parameters of interest.
|
RESULTS |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Table 2 shows a summary of basic statistics and coefficients of determination obtained for lamb BW. The coefficients of determination ranged from 35.6 to 64.9%. The highest coefficient of determination was estimated for BW at 20 wk of age.
|
The deviance values obtained with the binary analyses enabled us to determine the most significant model by using the 2 test (data not shown); the deviance can be interpreted as a measure of discrepancy or "distance" between the model prediction and the data. Model 1 was more significant than model 2 and model 5, but it was less significant than model 3 and model 4 for survival at birth. These results demonstrated that the most important factors for perinatal survival (i.e., survival at birth) were sex and year of birth.
Table 3 shows the relative risks of death for litter size, sex, and age of dam for all survival traits. The results showed that single-born and twin-born lambs were more likely to survive than triplet-born lambs at each age; male lambs were less likely to survive than females; and lambs born from 2-yr-old ewes were less likely to survive than lambs born from older ewes, although, when using a dam age of 5 yr as the baseline, significant contrasts were seen only at 1 wk of age.
|
Direct and maternal heritability estimates for lamb BW traits and ratios of litter (common environment) variance to total variance are reported in Table 4. The model including direct and maternal genetic effects, and the litter effect gave the best fit for lamb BW at birth, 12 wk, and 16 wk. The model including direct and maternal genetic effects was the most suitable model for lamb BW at 4 and 8 wk, whereas the model including direct genetic and litter effects ensured the best fit for lamb BW at 20 and 24 wk. The heritability estimates for the direct effect ranged from 0.08 to 0.26, being substantially higher after 20 wk, with standard errors ranging from 0.04 to 0.07. The heritability estimates for the maternal effect ranged from 0.06 to 0.21, with standard errors between 0.04 and 0.05, although these estimates were not significant from 20 wk of age onward. The estimates for the common environmental effect were between 0.04 and 0.16, with standard errors between 0.03 and 0.04, but this effect was not significant for the BW at 4 and 8 wk.
|
shows the heritabilities, estimated considering survival as both a binary and a continuous trait, and the expected values calculated with the formula of Dempster and Lerner (1950). The heritability estimates obtained with the probit model ranged from 0.01 to 0.33, and it should be noted that the heritabilities for cumulative survival declined from birth onward. This is explicable from a consideration of perinatal and postnatal survival. Whereas perinatal survival was moderately heritable, postnatal survival had a heritability close to zero. As expected, heritabilities from the normal analysis were somewhat lower, and it was recognized that the assumption of the trait as being continuous with normally distributed residuals was violated. However, expected values derived from these heritabilities were close to those from the binary analysis, confirming that the impact of departures from normality was predictable.
|
. The genetic correlations ranged from 0.43 to 0.94, indicating a moderate to high positive correlation between these traits. The standard errors for genetic correlation estimates were always less than or equal to 0.06, except for genetic correlations estimated between birth weight and other BW traits, where standard errors were higher, reaching 0.10 between birth weight and BW at 24 wk. Genetic and phenotypic correlations were high when the traits were measured close together in time but decreased as the interval between measurements increased.
|
shows the genetic and phenotypic correlations between survival and BW traits; however, phenotypic correlations are presented only for uncensored data sets (i.e., pairs of traits in which the BW age is less than the survival age). The phenotypic correlations were always positive and ranged from 0.16 to 0.54, whereas the estimates for the genetic correlations were similar, ranging from 0.04 to 0.45. Therefore, genetically improving live BW should improve survival, and vice versa.
|
|
DISCUSSION |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
The results from the generalized linear model analyses showed that sex, age of dam, and litter size had an influence on survival. Male lambs had lower survival rates than females, in agreement with the results obtained by Hight and Jury (1970) and Dalton et al. (1980). This differs from the results reported by Atkins (1980), who found that males and females have similar survival rates. Two-year-old ewes had lower lamb survival than older ewes, probably because these ewes are inexperienced mothers and their lambs receive less attention. This result was consistent with those reported by Hight and Jury (1970) and Atkins (1980). The effect of litter size on lamb survival is much debated. It has been shown that mortality usually increases with litter size (i.e., Hinch et al., 1983), which is in agreement with the results obtained in this study. There is evidence in sheep for an antagonistic genetic correlation between litter size and lamb survival (Cundiff et al., 1982). However, Bradford (1972) suggested that litter size does not affect lamb survival directly but through its effects on birth weight. Litter size can also influence the distribution of the ewe’s grooming behavior; O’Connor et al. (1992) demonstrated that twin-born lambs received less overall grooming attention than singleton lambs. Year of birth was also significant in the current study, by accounting for variation in climatic conditions that affect lamb survival either through exposure effects during inclement weather or through the effects on the nutritional status of the grazing ewe and lamb (Donnelly, 1984).
The heritability estimates for BW traits fall within the range reported in the literature for corresponding periods. Snyman et al. (1995) and Al-Shorepy and Notter (1998) reported direct heritability estimates (ha2) for birth weight of 0.22 and 0.23, which are higher than the 0.15 obtained in this study. Moreover, their estimates for the maternal genetic effect ( hm2) were 0.09 and 0.07, respectively, and lower than the 0.18 found in this study.
In general, our ha2 estimates for BW at 4 and 8 wk were in the ranges (from 0.06 to 0.11, and from 0.13 and 0.21, respectively) reported in the literature for BW at 30 and 60 d of age, whereas our hm2 estimates for BW at 4 and 8 wk were 0.21 and 0.19, respectively, and were higher than the estimates reported in the literature (Notter, 1998; Janssens et al., 2000; Matika et al., 2003).
Many authors have demonstrated that, apart from birth weight, direct additive genetic effects for BW increase in importance with increasing age (e.g., Snyman et al., 1995; Näsholm and Danell, 1996). This agrees with the estimates obtained in this study, if we look at the estimates for older ages (i.e., BW at 20 and 24 wk). Therefore, these results suggest that live BW measurements made on older lambs would increase the accuracy of selection.
Our major finding regarding the genetic control of survival was that perinatal survival was moderately heritable, whereas postnatal survival had very low heritability. As a direct consequence, the heritability of cumulative survival declined with age as it became increasingly influenced by postnatal mortalities. Several results on heritability estimates for lamb survival have been reported in the published literature, and these analyses have taken mainly the direct and maternal genetic effects into account (i.e., Barwick et al., 1990; Burfening, 1993; Lopez-Villalobos and Garrick, 1999). It has been demonstrated that lamb survival is a complex trait influenced by the capacity of the individual lamb to survive and by the rearing ability of the dam (Cundiff et al., 1982; Piper et al., 1982). Furthermore, a sire model has been fitted by Konstantinov et al. (1994), Olivier et al. (1998), and Matika et al. (2003). Sawalha et al. (2007) provided a comprehensive genetic analysis of lamb survival in Scottish Blackface sheep, although their definition of survival, broken into discrete periods, differs from our cumulative survival traits. In addition to results for survival analyzed as a binary trait, Sawalha et al. (2007) presented results for an analysis of survival time in Scottish Blackface lambs. Using both approaches, Sawalha et al. (2007) found survival to be a trait with somewhat low heritability, although when transformed to the underlying scale, the heritabilities ranged from 0.06 to 0.30. In contrast to our results, Sawalha et al. (2007) found survival at d 0 to be a trait with low heritability, although the reason for this discrepancy between the two studies is not clear.
The declining heritability estimates for survival traits from birth onward in our study show clearly that at birth the genetic component has a more important influence than later on; this confirms the large influence of environmental factors that accumulate with time. Therefore, in principle, substantial improvement in lamb survival may be more readily achieved by improving lambing and rearing management practices, although such improvements are often elusive in practice. In general, the estimates found in the literature were lower than those obtained in this study (i.e., Burfening, 1993; Konstantinov et al., 1994; Lopez-Villalobos and Garrick, 1999; Matika et al., 2003; Sawalha et al., 2007).
One of the reasons for the low heritability usually observed for lamb mortality is that it is a composite trait, combining many factors that may lead to death. Some authors (e.g., Sawalha et al., 2007) address this by subdividing survival into different age periods; however, this does not overcome the fact that mortality at any point may still be multifactorial. The choice of time period is also somewhat arbitrary. In addition to peri-and postnatal survival, we chose to analyze cumulative mortality (or survival), because this is the trait of economic importance to the farmer. Additionally, survival or mortality may be analyzed in several ways (e.g., by using binary analyses, as in this paper, or by survival analyses in which length of time until death is the trait of interest). We chose the former method because whether a lamb is alive or dead at a certain age is of greater practical relevance than the length of time until death. In fact, from an economic perspective a lamb that lives longer before dying carries a greater financial penalty than a lamb that dies at birth.
Genetic (and phenotypic) correlation estimates obtained between BW traits and between survival and BW traits were always positive. Some authors have highlighted the fact that there is a positive phenotypic relationship between survival and birth weight at all dam ages (e.g., Smith, 1977; Mukasa-Mugerwa et al., 2000). The only published estimates for the genetic correlation between live BW and survival are given by Sawalha et al. (2007), in which positive genetic correlations are seen between survival at d 0 and live BW. For survival at older ages, Sawalha et al. (2007) did not observe any genetic correlations that were significantly different from zero; however, differences in the definition of survival make comparisons with our study difficult.
The positive correlation between survival and live BW found in our study is favorable, because it indicates that selection for increased live BW should increase survival, and selection for improved survival would also have beneficial effects on live BW. Furthermore, the results indicate that survival is best assessed on the day of birth, before environmental effects have an untoward effect on survival, whereas live BW is better assessed at older ages (e.g., 20 or 24 wk of age).
Implications
The results show that lamb survival is a heritable trait in extensively reared lambs, and it could easily be included in a BLUP analysis in which information on relatives will be exploited. Additionally, because the results suggest that lamb survival is most heritable at birth or within 24 h of birth, the simplest approach is simply to record whether a lamb is dead or alive at, or immediately after, birth. This will enable good genetic progress and it will minimize the effort required to record lamb survival. Furthermore, use of this information in a multitrait BLUP analysis will benefit selection for increased live BW, because early lamb survival is favorably genetically correlated with subsequent lamb growth.
|
Footnotes |
|---|
2 Corresponding author: vriggio{at}unipa.it
Received for publication March 2, 2007. Accepted for publication March 31, 2008.
|
LITERATURE CITED |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. W. P. Cloete, I. Misztal, and J. J. Olivier Genetic parameters and trends for lamb survival and birth weight in a Merino flock divergently selected for multiple rearing ability J Anim Sci, July 1, 2009; 87(7): 2196 - 2208. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Maxa, A. R. Sharifi, J. Pedersen, M. Gauly, H. Simianer, and E. Norberg Genetic parameters and factors influencing survival to twenty-four hours after birth in Danish meat sheep breeds J Anim Sci, June 1, 2009; 87(6): 1888 - 1895. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
