Last update: January 19, 2011 01:47:28 PM E-mail Print

 

GENETIC PARAMETERS FOR FITNESS TRAITS IN SOUTH AFRICAN MERINO SHEEP


 W.J. Olivier* & M.A. Snyman

Grootfontein Agricultural Development Institute,

 Private Bag X529, Middelburg E.C., 5900, South Africa

 

J.B. van Wyk & G.J. Erasmus

Department of Animal Science, University of the Free State,

 P.O. Box 339, Bloemfontein, 9300, South Africa

 


Abstract

The main objectives of this study were to estimate heritabilities and the range of sire breeding values for reproductive traits and survival rate in two Merino flocks using a threshold model. A GFCAT set of programmes was used to analyse reproductive data collected on the Grootfontein Merino stud (from 1968 to 1996) and the Carnarvon Merino flock (from 1964 to 1983). Fitness traits analysed were fertility (whether a ewe lambed or not; 0 or 1), fecundity (number of lambs born to a ewe that lambed; 1, 2 or 3), litter size at birth (number of lambs born per ewe mated; 0, 1, 2 or 3), litter size at weaning (number of lambs weaned per ewe mated; 0, 1, 2 or 3) and survival rate (whether a lamb born alive, was dead or alive at weaning; 0 or 1). The estimated heritabilities on the underlying scale for the respective traits were 0.072, 0.173, 0.131, 0.092 and 0.000 for the Grootfontein Merino stud and 0.203, 0.311, 0.201, 0.183 and 0.000 for the Carnarvon Merino flock. The heritability estimates and the range in sire breeding values indicate that it would be possible to improve reproduction, but not survival rate, genetically through selection.

 

Key words: Merino sheep, threshold model, sire evaluation, reproductive traits, survival rate

 

Introduction

Selection objectives and criteria in sheep farming enterprises differed from breed to breed, country to country, and even from farm to farm. Reproduction and survival rate, however, are the only traits that are universally important, as they are the most important factors determining the efficiency of lamb production (Large, 1970;  Snyman et al., 1997a). It was suggested by Dickerson (1978) that there is more potential for increasing both biological and economic efficiency of lamb production through genetic improvement in reproduction than in either growth rate or body composition. Due to various reasons, of which the difficulty in accurately measuring reproductive performance at an early age and the lack of suitable computer software, selection for reproduction was neglected in the past. This is especially true for Merino sheep, as they are pre-eminently wool producers. The availability of modern statistical software and computer hardware has made it possible to obtain more accurate estimates of genetic parameters and breeding values for reproductive traits.

Reproductive traits are, by definition, threshold traits. Threshold traits are not continuous in their expression, but exhibit categorical phenotypes. The understanding of the inheritance of such traits lies in the idea that the trait has an underlying continuity with a threshold which imposes a discontinuity on the visible expression of the trait (Falconer & Mackay, 1996). The relationship between polygenes and the expression of discontinuous characters comes about through the establishment of thresholds. That is, those polygenically determined genotypes that have values below the threshold show no expression of the character (Strickberger, 1990).

In the past linear model methodology, such as Henderson=s method III, were the most frequently used for the analysis of discontinuous as well as continuous data. The problem when analysing discontinuous data with linear procedures, is that the method of analyses is based on continuous phenotypic distribution and does not take the discontinuity of threshold traits into consideration. Gianola (1982) stated that the main theoretical reason for not using BLUP (Best Linear Unbiased Prediction) with categorical data is that breeding values and residuals are not independent of each other. Threshold procedures should therefore be more suitable when analysing reproduction data, as threshold model sire breeding values are expressed in units of residual standard deviation of the underlying scale.

Fitness traits analysed in this study included fertility, fecundity, litter size at birth, litter size at weaning and survival rate. The objectives of this study were to estimate heritabilities and to determine the range of sire breeding values for reproductive traits and survival rate in South African Merino sheep using a threshold model.

 

Material and methods

Data

Data collected on the Grootfontein Merino stud (from 1968 to 1996) and the Carnarvon Merino flock (from 1964 to 1983) were used for this study. Detailed descriptions of the management and selection procedures followed in these flocks were reported earlier for the Grootfontein Merino stud by Olivier (1989) and for the Carnarvon Merino flock by Erasmus et al. (1990).

The Grootfontein Merino stud (GMS) is kept at the Grootfontein Agricultural Development Institute near Middelburg (31°28'S,25°1'E) in the North-eastern Karoo region. The animals are run under favourable nutrional conditions, which include irrigated pastures and supplementary feeding.

The Carnarvon Merino flock (CMF) was kept on natural pastures at the Departmental Experimental Station near Carnarvon (30°59'S,20°9'E) in the North-western Karoo region. The vegetation consists of mixed grass and karoo shrub and is described as arid karoo. The average rainfall is 209mm and it occurs mainly during the autumn months. The official grazing capacity is 5,5 hectare per small stock unit.

The reproductive traits analysed were fertility (whether a ewe lambed or not; 0 or 1), fecundity (number of lambs born to a ewe that lambed; 1,2 or 3), litter size at birth (number of lambs born to a ewe mated; 0,1,2 or 3), litter size at weaning (number of lambs weaned to a ewe mated; 0,1,2 or 3) and survival rate (whether a lamb born alive, was dead or alive at weaning; 0 or 1). Only sires with more than ten records for each trait were retained for analyses. A description of the data sets is given in Table 1.

 

Table 1. Description of the data sets

 

 

No. of records

No. of sires

No. of categories

No. of thresholds

GMS

CMF

GMS

CMF

Fertility

8590

8777

193

420

2

1

Fecundity

7432

6237

185

420

3

2

LSB

8590

8777

193

420

4

3

LSW

8590

8777

193

420

4

3

Survival rate

10210

9174

217

493

2

1

 

The average litter size at birth over the experimental period was 0.74 for the Carnarvon Merino flock and 1.34 for the Grootfontein Merino stud. The corresponding litter size at weaning was 0.64 and 1.11, while fecundity for the two flocks was 1.14 and 1.56 respectively.



Statistical analyses

Sources of non-genetic variation

The categorical data modelling (CATMOD) procedure of SAS (1989) was used to determine the importance of non-genetic sources of variation. The effects of  year (19 for CMF) / year-season (36 for GMS), age of dam in years (7 for GMS and 6 for CMF) and birth status (3 in both flocks; fitted for survival rate only) were included in the analyses.

The following model was fitted for each trait :

   Yijk = ai + bj + sk +eijk

where

   Yijk       = an observation of a trait of the sire of the ith age of dam group, the jth birth status and the kth year-season / year,

   ai         = effect of ith age of dam,

   bj         = effect of the jth birth status,

   sk         = effect of the kth year-season/ year, and

   eijk       = random residual.

 

Threshold model analyses

Heritability estimates and estimated sire breeding values (EBV=s) were obtained by means of a GFCAT programme (Konstantinov, 1995). GFCAT is a set of programmes for the analyses of Amixed@ threshold models with support for REML-type variance component estimation based on the methods of Gianola & Foulley (1983). Under these models, the respective traits occur as a result of an underlying unobserved phenotype exceeding a given threshold (Konstantinov et al., 1994). The unobserved continuous phenotypes are assumed to be normally distributed. For each trait a vector, F, of means corresponding to subpopulations determined by combinations of levels of Afixed@ b and Arandom@ s factors, is modelled as:

   F = Xb + Zs

where 

   F is a vector of underlying means,

   b is a vector associated with the effects of year (CMF) /  year-season (GMS), age of dam  (not for survival rate) and birth status (only for survival rate),

   s is a vector of sire effects and

   X and Z are design matrices.

The sire effects are assumed to be normally distributed, with E(s) = 0 and Var(s) = Aδ2s, where A is the numerator relationship matrix.

The solutions for thresholds and vectors b and s were computed as described by Konstantinov et al (1994). The different traits for each of the flocks were analysed separately. The heritability (h2) was estimated by setting the error variance to unity:


  h2 =    
    4 x δ2s
  _______
   (1 + δ2s)

The expression of the solutions for the thresholds, all the effects and the estimated sire breeding values are in units of residual standard deviation of the underlying scale due to the fact that the underlying scale is unknown

 

Results and discussion

Sources of non-genetic variation

The probability values (P) of significance for non-genetic sources of variation for the different traits in each flock are given in Table 2. In GMS the year-season effect and the year effect in CMF had a significant effect on all the traits. Age of dam had a significant effect on all the reproductive traits in both flocks, but no influence on survival rate. Birth status had a significant effect on survival rate in both flocks.

 

Table 2. Probability values (P) of significance for non-genetic sources of variation for the different traits for each flock

Effect

 

Fertility

Fecundity

LSB

LSW

Survival rate

df

P

df

P

df

P

df

P

df

P

Age of dam

(GMS)

6

**

6

**

12

**

12

**

5

ns

Age of dam

(CMF)

5

**

5

**

10

**

10

**

5

ns

Year-season

(GMS)

35

**

35

**

66

**

70

**

33

**

Year

(CMF)

18

**

17

**

34

**

34

**

20

**

Birth status

(GMS)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2

**

Birth status

(CMF)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2

**

df = Degrees of freedom

LSB = Litter size at birth

LSW = Litter size at weaning

** = P < 0.001; ns = P > 0.05

 

Those effects which had a significant influence on the respective traits were included in the subsequent threshold analysis.

 

Threshold solutions

Solutions for thresholds, age of dam and birth status for each flock are given in Table 2 and 3 for GMS and CMF respectively.

Table 3. Thresholds and solutions for age of dam and birth status for GMS

 

Fertility

Fecundity

Litter size at birth

Litter size at weaning

Survival rate

Thresholds

 

 

 

 

 

1

0.0000

0.0000

0.0000

0.0000

0.0000

2

 

2.1871

1.3460

1.3194

 

3

 

 

3.3808

3.3109

 

 

 

 

 

 

 

Age of dam

 

 

 

 

 

1

-0.0187

-1.1606

0.0956

-0.3385

 

2

1.1229

-0.2335

1.0519

0.5720

 

3

1.2180

0.2403

1.3572

.08784

 

4

1.3045

0.4681

1.5431

0.9520

 

5

1.2343

0.5274

1.5510

0.9451

 

6

1.0105

0.4770

1.3973

0.7922

 

7

0.7880

0.4845

1.2499

0.6540

 

 

 

 

 

 

 

Birth status

 

 

 

 

 

1

 

 

 

 

1.8735

2

 

 

 

 

1.0933

3

 

 

 

 

0.7718

 

Table 4. Thresholds and solutions for age of dam and birth status for CMF

 

Fertility

Fecundity

Litter size at birth

Litter size at weaning

Survival rate

Thresholds

 

 

 

 

 

  1

0.0000

0.0000

0.0000

0.0000

0.0000

  2

 

2.0875

2.0842

1.9422

 

  3

 

 

3.9910

4.0342

 

 

 

 

 

 


Age of dam

 

 

 

 

 

  2

0.0026

-2.7390

-0.1911

-0.3546

 

  3

0.4767

-2.1466

0.2747

0.0795

 

  4

0.6493

-1.8521

0.4960

0.3081

 

  5

0.7551

-1.6821

0.6318

0.4318

 

  6

0.7469

-1.4906

0.7111

0.5090

 

  7

0.5766

-1.4684

0.5917

0.3567

 

 

 

 

 

 

 

Birth status

 

 

 

 

 

  1

 

 

 

 

1.5382

  2

 

 

 

 

1.1588

  3

 

 

 

 

1.1635

It is evident from Table 3 that there was a distinct pattern for the solutions of age of dam in GMS. There was an increase in fertility, fecundity and litter size at birth and weaning with an increase in age of dam until four or five years of age. In CMF a similar pattern was followed for all traits until five or six years of age. As expected, the youngest ewes had the lowest fertility, fecundity and litter size in both flocks. The decrease in survival rate with an increase in birth status was also expected.

 

Heritabilities

Estimates of sire variances and heritabilities for the different traits in each flock are given in Table 5.

Table 5. Sire variance and heritabilities for fitness traits in the two flocks

 

GMS

CMF

Trait

Sire Variance

Heritability

Sire Variance

Heritability

Fertility

0.01836

0.072

0.05334

0.203

Fecundity

0.04527

0.173

0.08416

0.311

LSB

0.03374

0.131

0.05276

0.201

LSW

0.02373

0.092

0.04780

0.183

Survival rate

0.00000

0.000

0.00000

0.000

The highest heritability in both flocks was obtained for fecundity, which was 10% higher than that estimated for fertility in the respective flocks. This supported claims that selection for reproduction should be based on fecundity and not fertility. Furthermore,  heritability estimates for litter size at birth were also lower than those estimated for fecundity. This could be expected, as records from ewes that failed to lamb were also included for the estimation of heritability for  litter size at birth. Heritabilities obtained for  litter size at weaning were lower than those estimated for fecundity, but similar to the estimates obtained for fertility in both flocks. The heritability for survival rate was zero in both the GMS and CMF.

Estimated heritabilities for the reproductive traits in the CMF were higher than the corresponding heritabilities obtained in the GMS. This may be due to the more natural conditions under which the Carnarvon flock was kept. In the GMS, artificial insemination is done, while natural mating was practised in the CMF.

Very little threshold heritability estimates for reproduction traits are available in the literature. Snyman et al (1997b) estimated heritabilities with a threshold model of 0.20, 0.42, 0.27, 0.19 and 0.02 for fertility, fecundity, litter size at birth, litter size at weaning and survival rate respectively. Jorgensen (1994) reported heritability estimates for litter size at birth and litter size at weaning ranging from 0.14 to 0.19 and 0.09 to 0.27 respectively, while Konstantinov et al. (1994) obtained an estimate of 0.24 for litter size at birth. The heritabilities for fertility, fecundity and survival rate estimated in this study are lower than those of the reported heritabilities, while those for litter size at birth and litter size at weaning are within the range of reported heritabilities.

Reported heritabilities, estimated from linear model paternal half-sib analyses, range from 0.01 (Atkins, 1986) to 0.22 (Martin et al, 1981) for fertility, from -0.12 (Basuthakur et al, 1973) to 0.35 (Abdulkhaliq et al, 1989) for litter size at birth and from 0.02 (Martin et al, 1981) to 0.26 (Abdulkhaliq et al, 1989) for litter size at weaning. The heritability estimates in this study for fertility and litter size at birth and litter size at weaning are with-in the range of these published heritabilities.

Snyman et al. (1997a) proposed that ewe selection should concentrate on reproductive performance. In flocks with an average lambing percentage of less than 100%, such as the CMF, selection for improved reproduction should be based on fertility, i.e. culling all ewes that fail to lamb. Due to a lack of sufficient numbers in flocks with a very poor reproductive performance, it would in many instances be necessary to keep some of the ewes that failed to lamb.

Maximum exploitation of the available genetic variation in a threshold trait such as fertility is, however, not very easy. Selection intensity is usually lower in traits with only one threshold, compared to those with two or three thresholds or continuous traits (Bourdon, 1997). This is especially true when a larger or smaller proportion need to be selected than the number of animals available of the favourable phenotype. Selection response would be greater if animals closer to the threshold on the underlying liability scale could be identified and selected. This is, however, not possible as there is, in the case of fertility, only two distinct phenotypes. This problem could to a large extent be eliminated by selection based on threshold estimated sire breeding values for the trait in question. Using estimated sire breeding values would not only increase selection intensity, but also accuracy of selection.

In ewe flocks with a high reproductive rate, such as the GMS, selection for improved reproduction performance should be based on fecundity or litter size at birth (combination of fertility and fecundity). These traits have  a higher heritability than fertility, and an increased selection intensity with these traits are possible, as they have more than one threshold. Alternatively, selection could be based on a genetically correlated continuous trait, such as total weight of lamb weaned (Snyman et al., 1997a), for which animal model EBV=s for ewes can also be estimated.

 

Sire breeding values

Sire breeding values (expressed in units of residual standard deviation of the underlying scale) for the best and worst sire for the different traits are given in Table 6. It is interesting to note that in GMS the best sire for each trait had the most daughters with reproduction records, while in CMF the number of daughters with reproduction records for the best and worst sires was almost equal.

Table 6. Range of estimated sire breeding valuesa for each trait.

 

GMS

CMF

Trait

Best sire

Worst sire

Best sire

Worst sire

 

(n)b

(n)

(n)

(n)

Fertility

0.1819

-0.2102

0.3273

-0.4700

 

(99)

(67)

(38)

(18)

Fecundity

0.4814

-0.3073

0.7936

-0.3430

 

(92)

(43)

(31)

(25)

LSB

0.4349

-0.3206

0.5834

-0.4266

 

(99)

(67)

(36)

(28)

LSW

0.2606

-0.2606

0.6007

-0.4049

 

(117)

(42)

(36)

(28)

a  Expressed in units of residual standard deviation of the underlying scale

b  Number of daughters with reproduction records

The higher heritability estimates obtained in the CMF are reflected in the wider range of sire breeding values compared to those of GMS. The range of sire breeding values indicate that there  is genetic variation between sires with regard to the reproductive traits which could be exploited during selection.

The main disadvantage of selection based on estimated sire breeding values is that these EBV=s for reproductive traits are only available after the first parity of a sire=s first daughters. In normal sheep enterprises, this would be when the sire is already four years of age. At that stage, the rams have already been used for two years in the flock. The most important application of sire EBV=s in practice would be in the identification of merit rams for use as national AI-sires.

 

Conclusion

The heritability estimates and range of sire breeding values obtained in this study indicate that it would be possible to improve reproduction rate, but not survival rate, genetically through selection in these flocks. Threshold model sire breeding values for reproductive traits could be used most effectively in the identification of rams for use as national AI-sires. When the estimation of threshold animal model breeding values become possible, it would lead to a further increase in selection response if ewes could also be selected on EBV=s for the respective traits.  A further advantage of using these animal model EBV=s, is that they can be estimated much earlier than sire EBV=s, and sire selection could be done on the performance of his dam.

 

References

ABDULKHALIQ, A.M., HARVEY, W.R. & PARKER, C.F., 1989. Genetic parameters for ewe productivity traits in the Columbia, Suffolk and Targhee breeds. J. Anim. Sci. 67 : 3250

ATKINS, K.D., 1986. A genetic analysis of the components of lifetime productivity in Scottish Blackface sheep. Anim. Prod. 1986. 43 : 405

BASUTHAKUR, A.K., BURFENING, P.J., VAN HORN, J.L. & BLACKWELL, R.L., 1973. A study of some aspects of lifetime production in Targhee and Columbia sheep. J. Anim. Sci. 36 : 813

BOURDON, R.M., 1997. Understanding animal breeding. Prentice-Hall.

DICKERSON, G.E. , 1978. Animal size and efficiency: Basic concepts. Anim. Prod. 27 : 367

ERASMUS, G.J., DE LANGE, A.O., DELPORT, G.J. & OLIVIER, J.J., 1990. Genetic and phenotypic parameter estimates of production traits in Merino sheep in an arid environment. S.  Afr.  J.  Anim.  Sci. 20(1) : 31

FALCONER, D.S. & MACKAY, T.F.C., 1996. Introduction to quantitative genetics. Longman, New York

GIANOLA, D., 1982. Theory and analyses of threshold characters. J. Anim. Sci. 54 : 1078

GIANOLA, D. & FOULLY, J.L., 1983. Sire evaluation for ordered data with a threshold model. Genet.Sel. Evol. 15 : 201

JORGENSEN, J.N., 1994. Estimation of genetic parameters for litter size in sheep using a non-linear method. Acta Agric. Scand. Seet. A. Animal Sci. 44 :  8

KONSTANTINOV, K.V., ERASMUS, G.J. & VAN WYK, J.B., 1994. Evaluation of Dormer sires for litter size and lamb mortality using a threshold model. S. Afr. J. Anim. Sci. 24(4) : 119

KONSTANTINOV, K.V., 1995. GFCAT: User=s guide. University of the Orange Free State, Bloemfontein

LARGE, R.V., 1970. The biological efficiency of meat production in sheep. Anim. Prod. 12 : 393

MARTIN, T.G., NICHOLSON, D., SMITH, G. & SALES, D.I., 1981. Phenotypic and genetic parameters for reproductive performance in a synthetic line of sheep. J. Agric. Sci. 96 : 107

OLIVIER, J.J., 1989. Genetic and environmental trends in the Grootfontein Merino stud. PhD-thesis, University of the Orange Free State, Bloemfontein

SAS Institute Inc., SAS/STAT Users Guide, Version 6, Fourth Edition, Volume 1, Cary, NC:SAS Institute Inc., 1989 : 405

SNYMAN, M.A., ERASMUS, G.J., VAN WYK, J.B., & OLIVIER, J.J., 1997a. Genetic parameter estimates for total weight of lamb weaned in Afrino and Merino sheep. Livest. Prod. Sci. 48(2) : 111

SNYMAN, M.A., ERASMUS, G.J. & VAN WYK, J.B., 1997b. An investigation into the possible genetic improvement of reproduction and survival rate in Afrino sheep using a Threshold model.  S. Afr. J. Anim. Sci.  (Submitted)

STRICKBERGER, M.W., 1990. Genetics, Third Edition. Macmillian Publishing Company. New York.