Animal model evaluation of dairy traits in US sheep breeds, their crosses and three synthetic populations

Animal model evaluation of dairy traits in US sheep breeds, their crosses and three synthetic populations

Small Ruminant Research 34 (1999) 1±9 Animal model evaluation of dairy traits in US sheep breeds, their crosses and three synthetic populations H. Sa...

89KB Sizes 3 Downloads 58 Views

Small Ruminant Research 34 (1999) 1±9

Animal model evaluation of dairy traits in US sheep breeds, their crosses and three synthetic populations H. Sakula,1, W.J. Boylana,2, J.N.B. Shresthab,*,3 a

Department of Animal Science, University of Minnesota, St. Paul, MN 55108, USA Centre for Food and Animal Research, Agriculture and Agri-Food Canada, Ottawa, Ont., K1A 0C6 Canada

b

Accepted 15 April 1999

Abstract Multi-trait animal model procedures allowed simultaneous estimation of year of record, age of ewe, litter size and genetic group, and the prediction of direct additive genetic merit, and non-genetic permanent maternal environmental effects of the dams with multiple lactations. Mixed model methodology was used to obtain estimated breeding values and estimated producing ability for total milk yield and percentages of fat, protein and lactose in the Dorset (D), Finnsheep (F), Lincoln (L) and Rambouillet (Ra) breeds and some of their crosses, the Romanov, Suffolk and Targhee breeds and three synthetic populations derived from crossbred foundations: Synthetic I (FXL), Synthetic II (DXRa) and Synthetic III (FXL)X(DXRa). Ranking of breeds was based on the mean values of mixed model solutions for direct additive genetic merit as deviations from their respective groups and non-genetic permanent maternal environmental effects of the dams. Among the purebreds, Suffolk ewes produced the highest volume of milk while the Finnish Landrace and Romanov ewes produced the lowest volume (6.8 vs ÿ14.5 and ÿ11.5, respectively). Productivity of synthetic populations was comparable to that of their purebred parents but lower than the crossbred parents. Crossbred ewes exceeded their purebred parents in milk yield (0.32 vs ÿ6.15), possibly due to heterosis. In contrast, the percentages of fat (ÿ0.07 vs ÿ0.02), protein (ÿ0.04 vs 0.02) and lactose content (0.01 vs ÿ0.07) in the crossbred ewes were similar to those in the purebred ewes. Increased milk production in sheep could be bene®cial in raising more lambs from multiple births common in the sheep breeds with a genetic background of the Finnsheep and Romanov inheritance. The genetic variation between and within domestic breeds reported here indicates an opportunity to develop higher milk producing strains for a commercial dairy sheep industry in North America. However, importation of European dairy sheep breeds is an alternative approach which might achieve that goal more rapidly and which is now being evaluated. Published by Elsevier Science Ireland Ltd. Keywords: Dairy sheep; Animal model; Milk yield; Composition

*Corresponding author. Tel: +1-819-565-9171; fax: +1-819-564-5507; e-mail: [email protected] Present address: Parke-Davis Laboratory for Molecular Genetics, 1501 Harbor Bay Parkway, Alameda, California 94502, USA. 2 Retired. 3 Present address: Dairy and Swine Research and Development Centre, Agriculture and Agri-Food Canada, PO Box 90, 2000 Route 108 est, Lennoxville, PQ J1M 1Z3, Canada. 1

0921-4488/99/$ ± see front matter Published by Elsevier Science Ireland Ltd. PII: S 0 9 2 1 - 4 4 8 8 ( 9 9 ) 0 0 0 4 8 - 6

2

H. Sakul et al. / Small Ruminant Research 34 (1999) 1±9

1. Introduction Raising sheep for milk production is a major undertaking in many countries of Europe, Asia and the Middle East. In 1990, more than 18,000 tonnes of cheese made from sheep milk was imported into the United States of America at a value exceeding 70 million dollars (USDA, 1990). In North America, there was no interest in milking sheep until 1984, when the ®rst dairy sheep research program was initiated at the University of Minnesota and opportunities for the development of a dairy sheep enterprise were identi®ed (Boylan, 1987). Currently, a commercial dairy sheep industry which can supply the current demand does not exist in North America. Dairy sheep research in Europe and the Middle East has resulted in the development of East Friesian, Lacaune, Lacha and Awassi, ``dairy'' type sheep breeds (Flamant and Barillet, 1982; Aleandri et al., 1986; Barillet et al., 1986). However, use of these breeds to develop a dairy sheep industry in North America is currently not feasable due to quarantine and ®nancial constraints on importation of sheep from continental Europe. Evaluation and selection within and among domestic breeds or the development of synthetic populations from imported and domestic dairy sheep breeds are alternatives. Animal model procedures (Quaas and Pollak, 1980) allow simultaneous evaluation of sires, dams and progeny for direct additive genetic merit. Animals without records can also be evaluated using their relationships to animals with records. This procedure is used widely for genetic evaluation in dairy cattle and has been used in dairy goats (Wiggans, 1989), but was not commonly used in dairy sheep when this study was completed. In France, single trait animal model procedures have been used for dairy sheep to evaluate their ef®ciency in estimating breeding values (Barillet et al., 1990). The results indicated that it was more ef®cient to use animal model procedures than the Modi®ed Contemporary Comparison method. The ®rst in a series of papers on evaluation of US sheep breeds dealt with milk production and milk composition in seven sheep breeds and three synthetic populations (Sakul and Boylan, 1993). In the present study reporting additional data from the same project, a multi-trait animal model for total milk yield and percentages of fat, protein and lactose was used to

predict direct additive genetic merit in ewes, and nongenetic permanent maternal environmental effects in dams with multiple lactations in seven sheep breeds, some of their crosses, and the three synthetic populations. 2. Materials and methods 2.1. Experimental animals and protocol The study included data collected at the University of Minnesota over ®ve years (1985 to 1989) from seven pure breeds, i.e., Dorset (D), Finnsheep (F), Lincoln (L), Rambouillet (Ra), Romanov, Suffolk and Targhee, three synthetic populations developed from crossbred foundations, i.e., Synthetic I (FXL)2 i.e. inter se matings of F and L crossbreds, Synthetic II (DXRa)2 and Synthetic III [(FXL)X(DXRa)]2, and 12 crossbred groups obtained from reciprocal matings of D, F, L and Ra breeds. Ewes of the synthetic populations had been established from inter se matings of the initial crosses in the foundation ¯ock. Most of these ewes providing data for the present study were the result of four generations of such matings. There had been no intentional selection for milk yield and composition. A random sample of approximately 30 ewes from each breed and some of their crosses, and three synthetic populations were chosen for the study (Table 2). These ewes had been weaned when their lambs were approximately 30 days of age and machine-milked twice daily for about four months from late April or early May to ®rst week of September. Individual milk yield was recorded weekly in the morning and afternoon. Recording days were treated as mid-week days to ensure least bias in obtaining weekly milk yield. During the ®nal week of lactation, ewes were milked only once a day to facilitate drying off. Ewes that produced less than 10 ml of milk for two consecutive weeks were removed from the test. Total milk production was obtained from average daily milk yield and number of days on milk trial. Milk composition was determined from individual samples taken every two weeks with a Multispec M Infrared Tester. To meet nutritional requirement of the ewe, at each milking ewes were fed alfalfa-brome hay and 350 g of concentrate mixture consisting of shelled corn and

H. Sakul et al. / Small Ruminant Research 34 (1999) 1±9

soybean meal in pelleted form. During the ®ve years of the study, all purebred, crossbred and synthetic populations were fed the same diets, raised in adjoining pens sharing the same environment and were handled in the same manner. Traits analyzed were total milk yield (liter) during a lactation and percentages of fat, protein and lactose in milk. Care and handling of the sheep used in this study conformed to the guidelines established by the American Association for Accreditation of Laboratory Animal Care, which has guidelines on animal experimentation similar to those of the Canadian Council on Animal Care. 2.2. Statistical analyses Simultaneous genetic evaluation of 279 ewes, their 75 sires and 28 dams for total milk yield and percentages of fat, protein and lactose was based on a mixed model analysis of 447 production records (200 purebreds, 166 crossbreds and 81 synthetic populations) over all available lactations. The ®xed effects included were year of record (1985 through 1989), age of ewe (2, 3, 4, and 5 years or older) and total number of lambs born per ewe lambing (1, 2, 3, and 4 or more lambs). Random effects were direct additive genetic merit as deviations from their respective genetic groups, non-genetic permanent maternal environmental effects for those ewes that had more than one milking record, and residual error. Phantom parents were assigned to those animals with pedigree information missing on one or both parents (Westell and Van Vleck, 1987; Shrestha et al., 1996). Genetic groups, corresponding to the four selection paths of genetic gain i.e. paternal and maternal grand-sires and grand-dams, were de®ned by generation number, year of birth, breed composition and the sex of the missing parent. Parents of ewes from the synthetic populations or crossbreds were either crossbred or purebred and were assigned to appropriate genetic groups. Assignment of parents with records provided a direct solution. Parents with no measurement on themselves and only one daughter were assigned to genetic groups according to their year of birth and breed composition. Previous selection, if any, or origin of foundation sires and dams was accounted for when assigning such base animals to ®xed genetic groups. The number of phantom parent groups created was 31 for male parents, and 39 for

3

female parents. Group coef®cients were included in constructing the inverse of the numerator relationship matrix as described by Henderson (1976) to account for genetic selection in the parents. The rules for incorporation of the numerator relationship matrix and for coef®cients due to groups was according to Westell et al. (1988). The animal model provided a direct solution for ewes that made records with joint prediction of genetic merit of sires and dams without records based on relationship to ewes with records through the inverse of the numerator relationship matrix. The animal and non-genetic permanent maternal environmental effects were assumed to have constant variance and to be uncorrelated with residual effects. The residual effects, which varied according to the number of traits measured, were assumed to have constant variances and to be uncorrelated with each other. The estimates of additive genetic, non-genetic permanent maternal environmental, residual and phenotypic variance-covariance components of milk production per lactation and milk composition (Table 1) were derived from heritabilities, repeatabilities, genetic and phenotypic correlations (Sakul and Boylan, 1993). The covariance among animal effects was included in the mixed model equations through the use of the numerator relationship matrix. The number of equations exceeded one thousand, therefore, Gauss± Seidel iteration with no constraints to the system was applied to obtain solutions (Shrestha et al., 1996). Mean values of estimated breeding value (EBV) and estimated producing ability (EPA) for the pure breeds, their crosses and the synthetic populations were computed from mixed model solutions. Standard errors for each breed or breed cross were computed from the variance among mixed model solutions of ewes within breed or breed cross. The multiple range test devised by Duncan (1955) and extended by Kramer (1957) was used to test for signi®cant breed differences among three or more means. The heterosis effect in the crossbred ewes was estimated by subtracting the average EBV of the purebreds from the corresponding average of their reciprocal crosses. 3. Results Overall means and standard deviations were 7327.5 l for total milk yield per lactation,

4

H. Sakul et al. / Small Ruminant Research 34 (1999) 1±9

Table 1 Estimates of additive genetic (G), non-genetic permanent maternal environmental (D), residual (R) and phenotypic (P) variance-covariance components for milk production per lactation and milk compositiona Milk production and composition Total milk yield (l)

G D R Pd

Fatb (%)

G D R P

Protein (%)

G D R P

Lactosec (%)

G D R P

Milk yield (l)

Fatb (%)

Protein (%)

Lactosec (%)

106.3768 82.5974 170.1448 359.1190

0.6718 ÿ0.4974 ÿ2.7686 ÿ1.1051

0.0600 ÿ0.0045 ÿ0.7492 ÿ0.7933

0.9505 0.1212 ÿ0.9776 1.2940

0.0795 0.0651 0.2168 0.3614

0.0321 ÿ0.0079 ÿ0.0500 0.0765

ÿ0.0085 ÿ0.0100 0.0002 ÿ0.0284

0.0247 0.0084 0.0722 0.1053

ÿ0.0049 0.0013 0.0034 ÿ0.0112 0.0151 0.0042 0.0192 0.0385

a

Estimates derived from heritabilities, repeatabilities, genetic and phenotypic correlations (Sakul and Boylan, 1993). Repeatability of 0.4 was assumed. c Repeatability of 0.5 was assumed. d PˆG‡D‡R. b

6.00.7% for fat content, 5.80.4% for protein content and 4.90.3% for lactose content. More than 800 rounds of iteration were required for the mixed model equations to converge. Following convergence, mean values of EBVs and EPAs and their standard errors for total milk yield and percentages of fat, protein and lactose were computed for the pure breeds, their crosses and the synthetic populations (Tables 2 and 3). 3.1. Purebred sheep Ewes of the Suffolk breed had signi®cantly greater mean EBVs for total milk yield than ewes of the Romanov, Finnish Landrace and Lincoln breeds. Rambouillet ewes did not differ (P>0.05) from ewes of any other breed. Suffolk, Targhee and Dorset ewes produced signi®cantly more milk than Finnish Landrace ewes. The mean EBVs for percentage of fat in ewes of the Finnish Landrace breed were signi®cantly lower than in all other breeds. Percentage of protein for Suffolk, Romanov, Dorset, Lincoln and

Rambouillet ewes was signi®cantly greater than for Targhee and Finnish Landrace ewes. In contrast, percentage of lactose did not vary signi®cantly among breeds. The mean EPAs of total milk yield for Romanov and Finnish Landrace ewes were signi®cantly lower than Suffolk and Targhee ewes. Finnish Landrace ewes were signi®cantly lower in mean EPA for percentages of fat and protein compared to the remaining breeds which were similar. There was no signi®cant difference among breeds for percentage of lactose. 3.2. Crossbred sheep Though not signi®cant, the mean EBVs among crossbred ewes for total milk yield (ÿ7.4 to 11.4) and percentage of lactose (ÿ0.07 to 0.13) were variable. Ewes of DXL and DXRa crosses had higher mean EBVs for fat and protein percent than FXL ewes. The mean EBV of crossbred ewes exceeded that of the purebreds for total milk yield (0.32 vs ÿ6.15). In contrast, there was no signi®cant difference between

H. Sakul et al. / Small Ruminant Research 34 (1999) 1±9

5

Table 2 Mean estimated breeding values (SE) for total milk yield per lactation and percentages of fat, protein and lactose in purebred ewes and their crosses, and ewes from synthetic populations Breed

n

Total milk yield (l)

Pure breeds Suffolk Targhee Romanov

31 30 16

6.8a1.8 1.0ab2.1 ÿ11.5bc1.7

0.24a0.04 0.01a0.06 0.32a0.08

0.04ab0.02 ÿ0.03b0.03 0.18ab0.04

ÿ0.00a0.02 0.00a0.02 ÿ0.09a0.03

Pure breeds Dorset Finnish Landrace Lincoln Rambouillet Mean

29 31 32 31

ÿ1.2ab2.5 ÿ14.5c2.0 ÿ6.8bc1.7 ÿ2.1abc2.4 ÿ6.15

0.25a0.07 ÿ0.40b0.04 ÿ0.01a0.05 0.10a0.05 ÿ0.02

0.20a0.04 ÿ0.25b0.02 ÿ0.00ab0.03 0.11ab0.03 0.02

ÿ0.07a0.03 ÿ0.11a0.02 ÿ0.08a0.01 ÿ0.00a0.02 ÿ0.07

Crossbreds DXF DXL DXRa FXL FXRa LXRa Mean

32 27 24 30 27 26

ÿ7.4a4.2 7.0a3.9 1.0a4.9 ÿ3.7a3.3 11.4a4.1 ÿ6.4a2.6 0.32

ÿ0.19ab0.04 0.16a0.09 0.22a0.11 ÿ0.39b0.07 ÿ0.10ab0.11 ÿ0.14ab0.06 ÿ0.07

ÿ0.07ab0.03 0.07a0.04 0.18a0.06 ÿ0.25b0.06 ÿ0.10ab0.06 ÿ0.04ab0.04 ÿ0.04

ÿ0.07a0.05 0.05a0.04 ÿ0.01a0.05 0.01a0.05 0.13a0.04 ÿ0.03a0.04 0.01

28 26 27

ÿ17.5a2.2 0.0b2.2 ÿ3.3b2.5

ÿ0.38a0.07 ÿ0.01a0.09 ÿ0.13a0.06

ÿ0.16a0.04 0.09b0.04 ÿ0.03ab0.03

ÿ0.13a0.03 ÿ0.01a0.02 ÿ0.03a0.03

Synthetic populations Synthetic I Synthetic II Synthetic III

Fat (%)

Protein (%)

Lactose (%)

Breed abbreviations: DˆDorset, FˆFinnish Landrace, LˆLincoln, RaˆRambouillet, RoˆRomanov, SˆSuffolk, TˆTarghee, Synthetic Iˆ(FXL)2, Synthetic IIˆ(DXRa)2 and Synthetic IIIˆ((FXL)X(DXRa))2. The two-breed cross is an average of the reciprocal crosses. Means within a column and within class not followed by the same letter differ (P<0.05).

crossbred and purebred ewes for percentage fat (ÿ0.07 vs ÿ0.02), protein (ÿ0.04 vs 0.02) and lactose (0.01 vs ÿ0.07). Heterosis was greatest in FXRa ewes followed by DXL ewes. This has resulted in higher total milk yield and percentage of lactose compared to all other crossbred ewes. Ewes of DXRa crosses produced greater percentages of fat and protein while DXF crosses ranked lowest in total milk yield and lactose, and FXL ranked lowest in percentages of fat and protein. Though not signi®cant among crossbred ewes, mean EPAs for total milk yield (ÿ8.2 to 18.9) and percentages of fat (ÿ0.44 to 0.21) and lactose (ÿ0.06 to 0.14) were variable. In contrast, percentage of protein in DXRa ewes differed signi®cantly from FXL ewes. The mean EPA of crossbred ewes exceeded that of the purebreds for total milk yield (2.83 vs

ÿ6.68), and percentages of fat (ÿ0.107 vs ÿ0.013), protein (ÿ0.03 vs 0.01) and lactose (0.018 vs ÿ0.065). The rankings of crossbred ewes for mean EBV were consistent with those for mean EPA. 3.3. Synthetic sheep The mean EBV for total milk yield was signi®cantly greater for Synthetic II and III ewes than Synthetic I ewes. Synthetic II ewes had a higher percentage of protein than Synthetic I ewes. In contrast, there was no signi®cant difference among Synthetic sheep for percentage of fat and lactose. The mean EPAs for total milk yield and percentages of fat and lactose for ewes of the three synthetic sheep were similar, however, percentage of protein for Synthetic II ewes was higher than for Synthetic I ewes.

6

H. Sakul et al. / Small Ruminant Research 34 (1999) 1±9

Table 3 Mean estimated producing abilities (SE) for total milk yield per lactation and percentages of fat, protein and lactose in purebred ewes and their crosses, and ewes from synthetic populations Breed

n

Total milk yield (l)

Pure breeds Suffolk Targhee Romanov

31 30 16

8.0a2.6 1.7a3.0 ÿ15.4b3.0

0.28a0.05 0.02a0.07 0.37a0.13

0.04a0.02 ÿ0.03a0.03 0.18a0.05

ÿ0.01a0.02 0.0a0.02 ÿ0.09a0.04

Pure breeds Dorset Finnish Landrace Lincoln Rambouillet Mean

29 31 32 31

ÿ1.6ab3.4 ÿ15.8b2.9 ÿ7.4ab2.7 ÿ1.9ab3.7 ÿ6.68

0.26a0.08 ÿ0.39b0.05 ÿ0.02a0.07 0.10a0.06 ÿ0.01

0.21a0.05 ÿ0.26b0.02 ÿ0.02a0.03 0.11a0.03 0.01

ÿ0.07a0.03 ÿ0.11a0.02 ÿ0.08a0.01 0.0a0.03 ÿ0.07

Crossbreds DXF DXL DXRa FXL FXRa LXRa Mean

32 27 24 30 27 26

ÿ5.4a5.5 11.6a5.6 1.5a5.8 ÿ1.4a5.3 18.9a7.5 ÿ8.2a3.9 2.83

ÿ0.28a0.07 0.17a0.12 0.21a0.13 ÿ0.44a0.09 ÿ0.14a0.14 ÿ0.16a0.08 ÿ0.11

ÿ0.05ab0.04 0.07ab0.05 0.21a0.07 ÿ0.27b0.07 ÿ0.11ab0.06 ÿ0.04ab0.05 ÿ0.03

ÿ0.06a0.05 0.05a0.04 0.0a0.06 0.01a0.05 0.14a0.04 ÿ0.03a0.04 0.02

28 26 27

ÿ17.3a3.1 0.9a3.9 ÿ1.4a3.9

ÿ0.40a0.10 ÿ0.06a0.12 ÿ0.16a0.08

ÿ0.17a0.04 0.10b0.05 ÿ0.03ab0.04

ÿ0.12a0.03 0.0a0.03 ÿ0.03a0.03

Synthetic populations Synthetic I Synthetic II Synthetic III

Fat (%)

Protein (%)

Lactose (%)

Breed abbreviations: DˆDorset, FˆFinnish Landrace, LˆLincoln, RaˆRambouillet, RoˆRomanov, SˆSuffolk, TˆTarghee, Synthetic Iˆ(FXL)2, Synthetic IIˆ(DXRa)2 and Synthetic IIIˆ((FXL)X(DXRa))2. The two-breed cross is an average of the reciprocal crosses. Means within a column and within class not followed by the same letter differ (P<0.05).

Total milk yield, and percentages of fat, protein and lactose for Synthetic I ewes as a deviation from their parental purebred ewes were ÿ6.85, ÿ0.18, ÿ0.04, and ÿ0.04, respectively. Comparable values were 1.65, ÿ0.19, ÿ0.07 and 0.03, respectively, for Synthetic II ewes, and 2.85, ÿ0.12, ÿ0.05 and 0.04, respectively, for Synthetic III ewes. Correspondingly, total milk yield, fat, protein and lactose of Synthetic I ewes as a deviation from the average of crossbred ewes constituting the parental breeds were ÿ13.8, 0.01, 0.09, and ÿ0.14, respectively. Comparable values were ÿ1.0, ÿ0.23, ÿ0.09 and 0.0, respectively, for Synthetic II ewes, and ÿ2.0, ÿ0.05%, 0.01 and ÿ0.03, respectively, for Synthetic III ewes. Total milk yield, fat, protein and lactose of Synthetic III ewes as a deviation from the average of Synthetic I and II ewes were 5.45, 0.07, 0.01 and 0.04, respectively.

4. Discussion The animal model permitted the simultaneous evaluation of multiple traits in all animals, not only on the basis of their own records but also, through the inclusion of the inverse of the relationship matrix, on the performance of all relatives. This procedure was able to predict genetic effects when certain breeds or breed crosses were not well represented in all classes of the ®xed effects. The additional advantage of being able to evaluate animals for genetic merit that do not have records of their own, even those that are not yet born, on the basis of the records of their relatives gives animal breeders a powerful tool indeed. An animal model was used to obtain estimated breeding values for all ewes and their parents (Henderson and Quaas, 1976). Multi-trait animal model metho-

H. Sakul et al. / Small Ruminant Research 34 (1999) 1±9

dology has been used by Shrestha et al. (1996) to measure genetic trends in body weight of lambs. Shafto et al. (1996) have shown that multi-trait animal model solutions would be preferable to the least squares model when estimating parameters and breeding values from an unbalanced data set with unequal subclass frequencies. Furthermore, there is bene®t from using the animal model when data have been collected from individuals subjected to selection. Theoretically, the animal model is expected to yield best linear unbiased estimates of ®xed effects and best linear unbiased prediction of random effects. However, the quality of the analysis performed by the animal model is dependent on the accuracy of the genetic parameters. Furthermore, calculation of actual standard error terms for the animal model solutions is not possible due to the size of the matrix to be inverted. Therefore, approximate standard errors based on the variation within animal model solutions for each breed were utilized in this study. The superior milking ability of ewes of the Suffolk breed followed by the Targhee breed detected in this study is in agreement with the ®ndings of Slen et al. (1963). The average body weights of the Suffolk, Targhee and Rambouillet breeds that were 78, 76 and 71 kg, respectively (Sakul and Boylan, 1992) were confounded with the breed. In contrast, ewes from two fecund type sheep breeds, Finnish Landrace and Romanov with average body weights of 57 and 59 kg, respectively, were relatively low in total milk yield (Tables 2 and 3), an observation also reported by Flamant and Bonaiti (1979) and Ricordeau et al. (1988) for these breeds. Composition of milk tended to vary between the two most proli®c sheep breeds. Romanov ewes showed the highest percentage of fat while Finnish Landrace ewes had the lowest fat and protein content. Ewes of the Lincoln breed weighing 70 kg were poor in milk yield and composition while Dorset ewes weighing 62 kg were good milkers, producing milk with high fat and protein content. Ewes from breeds established in North America such as Dorset, Rambouillet and Suffolk were intermediate in fat and protein content while Lincoln and Targhee ewes tended to produce less fat and protein. However, percentage of lactose was consistent among the breeds. These ®ndings for US sheep breeds are in agreement with the earlier study on pure breeds by Sakul and Boylan (1992). Shrestha and Hansen (1998)

7

recently reviewed the newly created synthetic sheep breeds (Canadian, Outaouais and Rideau Arcotts) developed in Canada under an accelerated lambing program and concluded that these new breeds demonstrated potential for improvement of the sheep industry. Thus, there appears to be potential for improving milk yield and composition of sheep breeds based on genetic resources that exist today in North America. The crossbred ewes surpassed their purebred parents in EBV and EPA for total milk yield. Though not signi®cant, FXRa ewes exceeded all other crossbred ewes in total milk yield and percentage of lactose while DXF ewes were lowest. There was a tendency for FXRa and DXL ewes to have a higher mean EBV of total milk yield, surpassing the mean value of their respective purebreds (11.4 vs ÿ8.3 l for FXRa, 7.0 vs ÿ4.0 l for DXL). Fat and protein content was signi®cantly lower in the FXL ewes. DXRa and DXL ewes tended to have higher fat and protein content while FXRa, LXRa and DXF ewes were intermediate (Tables 2 and 3). These average values were consistent for EBV and EPA. The higher values for the crossbred ewes compared to the purebreds may be due to heterosis resulting from multiplicative epistasis or dominance. These ®ndings are in agreement with a number of studies reporting higher production of milk, fat and protein from crossbred than purebred ewes (Mavrogenis and Louca, 1980; Niznikowski et al., 1988). However, the apparent superiority of crossbreds over purebred ewes was not observed in the present study for percentages of fat, protein and lactose. The lack of adequate number of ewes in the speci®c reciprocal crosses increased the sampling error associated with those estimates; therefore, reciprocal effects were not estimated. Furthermore, it is evident that when EBVs are used for selection, crossbreds and purebreds should be selected separately to avoid any confounding effect due to heterotic in¯uence. Overall, synthetic populations ranked intermediate for EBVs and EPAs for total milk yield and percentage of lactose. Lower mean EBVs of total milk yield and protein content in Synthetic I ewes may be associated with the lower genetic merit of Finnish Landrace and Lincoln ewes which were the parental breeds. In contrast, performance of Dorset and Rambouillet ewes which were the parental breeds of the Synthetic II ewes were similar to their parental breeds. In general,

8

H. Sakul et al. / Small Ruminant Research 34 (1999) 1±9

Synthetic II and III ewes with average body weights of 71 and 74 kg, respectively, ranked higher in EBVs and EPAs for all traits than Synthetic I ewes that weighed 70 kg. The crossbred ewes with the same parental breed composition were more productive than the synthetic populations. There was a tendency for Synthetic III ewes to be more productive for milk yield and composition than Synthetic I ewes but were similar to Synthetic II ewes. New breeds developed from a crossbred foundation, therefore, have the potential to be more productive when based on a broad genetic base of more productive breeds. 5. Conclusion It is feasible to use an animal model for evaluation of multiple traits in dairy sheep based on multiple lactations. Obtaining averages for estimated breeding value and estimated producing abilities can be helpful in ranking and comparing breeds and their crosses. The information presented is of major importance to breeders of dairy sheep and represents the best and probably the only available information in North America on milk yield and composition in sheep. Further studies are required to evaluate breeds and their crosses based on a larger volume of data to verify the present ®ndings. Improvements in milk production can be readily achieved by developing new breeds when exotic milk sheep breeds from continental Europe are crossed with sheep genetic resources in North America. An alternative approach would be importation of dairy sheep breeds from continental Europe in large numbers. Increasing milk production in ewes could be bene®cial in raising more lambs from multiple births and possibly in developing a sheep dairy industry that could replace a portion of the imported cheese. Acknowledgements The authors wishes to thank the International Development Research Centre (IDRC), Ottawa, Canada for awarding a Postdoctoral Fellowship to the senior author. Paper no. 19,616 Scienti®c Journal Series of the University of Minnesota Agricultural Experiment Station, and Contribution no. of the Dairy

and Swine Research and Development Centre, Agriculture and Agri-Food Canada, Lennoxville, QueÂbec, Canada, in cooperation with AHRD, ARS, USDA as a contribution to the Regional Project NC-111, Increased Ef®ciency of Lamb Production.

References Aleandri, R., Nardone, A., Russo, V., 1986. Milk yield for the cheesemaking process: Quantitative trait loci and selection strategies. Proc. of the third World Cong. on Genet. Appl. Livest. Prod. 12, 64±69. Barillet, F., Elsen, J.M., Roussely, M., 1986. Optimization of a selection scheme for milk composition and yield in milking ewes: Example of the Lacaune breed. Proc. of the third World Cong. on Genet. Appl. to Livest. Prod. 9, 658±664. Barillet, F., Boichard, D., Barbat, A., Astruc, J.M., Bonaiti, B., 1990. Use of animal model for genetic evaluation of dairy sheep. Proc. 41st Ann. Meet. of the EAAP, Toulouse, France, pp. 1±7. Boylan, W.J., 1987. An alternative Canadian enterprise±dairy sheep. Sheep Canada 12, 8. Duncan, D.B., 1955. Mutiple range and multiple F tests. Biometrics 11, 1±42. Flamant, J.C., Barillet, F., 1982. Adaptation of the principles of selection for milk production to milking ewes: A review. Livest. Prod. Sci. 9, 549±559. Flamant, J.C., Bonaiti, B., 1979. Evaluation des aptitudes laitieÁres des brebis de race pure ou croiseÂes Romanov [Evaluation of the milking abilities of purebred and crossbred Romanov ewes]. Annales de GeÂneÂtique et de SeÂleÂction Animale 11, 223±240. Henderson, C.R., 1976. A simple method for computing the inverse of a numerator relationship matrix used in prediction of breeding values. Biometrics 32, 69±83. Henderson, C.R., Quaas, R.L., 1976. Multiple trait evaluation using relatives' records. J. Anim. Sci. 43, 1188±1197. Kramer, C.Y., 1957. Extension of multiple range tests to group correlated adjusted means. Biometrics 13, 31±38. Mavrogenis, A.P., Louca, A., 1980. Effects of different husbandry systems on milk production of purebred and crossbred sheep. Anim. Prod. 31, 171±176. Niznikowski, R., Tyszka, Z.J., Janikowski, W., 1988. Meat, wool and milk utilization together with comparison of F1 coming from Finnsheep rams with F1 coming from other prolific breeds. J. Agric. Sci. (Finland) 60, 534±537. Quaas, R.L., Pollak, E.J., 1980. Mixed model methodology for farm and ranch beef cattle testing programs. J. Anim. Sci. 51, 1277±1287. Ricordeau, G., Tchamitchian, L., Poivey, J.P., 1988. Performance, development and use of Romanov in France. J. Agric. Sci. (Finland) 60, 558±565. Sakul, H., Boylan, W.J., 1992. Evaluation of US sheep breeds for milk production and milk composition. Small Rumin. Res. 7, 195±201.

H. Sakul et al. / Small Ruminant Research 34 (1999) 1±9 Sakul, H., Boylan, W.J., 1993. Genetic and phenotypic parameters of milk traits in sheep. Proceedings of the Seventh World Conference on Animal Production, vol. 2, Edmonton, Canada, pp. 5±6. Shafto, A.M., Crow, G.H., Shrestha, J.N.B., Parker, R.J., McVetty, P.B., Palmer, W.M., 1996. Genetic evaluation of lamb performance in Outaouais Arcott and Suffolk sheep, their crosses and Canadian Arcott- or Hampshire sired three breed crosses. Can. J. Anim. Sci. 76, 15±22. Shrestha, J.N.B., Hansen, C., 1998. Developing synthetic breeds of sheep: A review of the Canadian experience. Proc. of the sixth World Cong. on Genet. Appl. Livest. Prod 24, 121±124. Shrestha, J.N.B., Peters, H.F., Heaney, D.P., Van Vleck, L.D., 1996. Genetic trends over 20 years of selection in the three synthetic Arcotts, Suffolk and Finnish Landrace sheep breeds. 1. Early growth traits. Can. J. Anim. Sci. 76, 23±34.

9

Slen, S.B., Clark, R.D., Hironaka, R., 1963. A comparison of milk production and its relation to lamb growth in five breeds of sheep. Can. J. Anim. Sci. 43, 16±21. United States Department of Agriculture, 1990. US trade and prospects: Dairy, livestock and poultry products. Foreign Agricultural Service, Circular Series FDLP 12±90. Westell, R.A., Van Vleck, L.D., 1987. Simultaneous genetic evaluation of sires and cows for a large population of dairy cattle. J. Dairy Sci. 70, 1006±1017. Westell, R.A., Quaas, R.L., Van Vleck, L.D., 1988. Genetic groups in an animal model. J. Dairy Sci. 71, 1310±1318. Wiggans, G.R., 1989. Animal model evaluation of dairy goats for milk, fat, and protein yields with crossbred animals included. J. Dairy Sci. 72, 2411±2416.