Identification and fine mapping of quantitative trait loci for backfat on bovine chromosomes 2, 5, 6, 19, 21, and 23 in a commercial line of Bos taurusLi, C.;Basarab, J.;Snelling, W. M.;Benkel, B.;Kneeland, J.;Murdoch, B.;Hansen, C.;Moore, S. S.
doi: 10.2527/2004.824967xpmid: 15080315
Abstract Backfat thickness is one of the major quantitative traits that affects carcass quality in beef cattle. In this study, we identified and fine-mapped QTL for backfat EBV on bovine chromosomes 2, 5, 6, 19, 21, and 23 using an identical-by-descent haplotype-sharing analysis in a commercial line of Bos taurus. Eleven haplotypes were found to have significant associations with backfat EBV at the comparison-wise P-value threshold, and one at the chromosome-wise P-value threshold on bovine chromosomes 5, 6, 19, 21, and 23. On average, the 12 significant haplotypes had an effect of 0.62 SD on backfat EBV, ranging from 0.38 SD to 1.33 SD. The 12 significant haplotypes spanned nine chromosomal regions, one on chromosome 5 (65.4 to 70.0 cM), three on 6 (8.2 to 11.8 cM, 63.6 to 68.1 cM, and 81.5 to 83.0 cM), three on 19 (4.8 to 15.9 cM, 39.4 to 46.5 cM, and 65.7 to 99.5 cM), one on 21 (46.1 to 53.1 cM), and one on 23 (45.1 to 50.9 cM). Among the nine chromosomal regions, six were new QTL regions and three showed remarkable agreement with QTL regions that were previously reported. Eight of the nine QTL regions were localized to less than or close to 10 cM in genetic distance. The results provide a useful reference for further positional candidate gene research and marker-assisted selection for backfat. Introduction The amount and distribution of fat has an important impact on carcass and meat quality in beef cattle (Powell and Huffman, 1973; Wheeler et al., 1994; Lozeman et al., 2001). Beef carcasses can be stratified by using subcutaneous fat thickness alone or in combination with marbling (Jeremiah, 1996). Breeding for optimal fat is therefore one of the major goals toward better profitability in the beef industry. Mapping of QTL and identification of causative genes that affect fat metabolism will enhance the progress toward this goal. Quantitative trait loci mapping for quantitative traits related to fat has been reported in a number of studies (Stone et al., 1999; Casas et al., 2001; MacNeil and Grosz, 2002). Fine mapping and verification of those QTL for fat are therefore necessary for further candidate gene research. Individuals within a semiclosed population, such as a commercial line of cattle, are expected to be derived from one or a limited number of founders. Some common haplotypes originating from the common ancestors should therefore carry on and segregate among the individuals of the breeding line, particularly when selection is applied. These common haplotypes may harbor QTL of interest and make it possible to locate QTL segregating in the line. This identical-by-descent (IBD) haplotype-sharing QTL mapping strategy avoids the generation of a well-designed mapping population. Obtaining such well-designed mapping populations is costly and may be impractical in commercial herds. In our previous studies, we have successfully identified and fine-mapped QTL for growth on bovine chromosome (BTA) 5 (Li et al., 2002a,b), as well as for backfat EBV on BTA14 (Moore et al., 2003) in commercial lines of Bos taurus using the IBD haplotype-sharing analysis method. We report here the identification and fine mapping of QTL for backfat EBV on BTA2, 5, 6, 19, 21, and 23 in a commercial population of Bos taurus. Materials and Methods Animals and Phenotypic Data Animals were from the M1 line of Beefbooster Inc. (Calgary, Canada) and were born in 1998. The M1 line was developed from an Angus base and has been under selection for over 30 yr. The selection criteria for the lines are based on indices described by MacNeil and Newman (1994). A 10-mL blood sample was collected by venipuncture from each male calf and the potential sires, and the DNA from each blood sample was extracted and kept for later parentage identification. Sire identification was carried out by the Saskatchewan Research Council, Canada, using DNA microsatellite markers. The backfat EBV of each male calf was calculated based on a BLUP procedure by Beefbooster Inc. Genotyping One hundred seventy-six male calves and their 12 respective sires (nine to 30 calves of each sire) of the M1 line were genotyped using 76 microsatellite markers, 13 from BTA2, 16 from BTA5, 16 from BTA6, 14 from BTA19, 8 from BTA21, and 9 from BTA23. The microsatellite markers genotyped on each chromosome (Figure 1) were chosen at an approximately even genetic distance and spanned on average 93.4% of the chromosomes, with a range of 83.6 (chromosome 6) to 99.7% (chromosome 23). Marker locations and primers designed for genotyping were based on the microsatellite marker information available on the USMARC Website (http://www.marc.usda.gov/genome/genome.html). Figure 1. View largeDownload slide Haplotypes with lowest P-values between two adjacent loci along bovine chromosomes 2, 5, 6, 19, 21, and 23 for backfat EBV in the M1 commercial line of Bos taurus from Beefboster Inc., Calgary, Canada. Haplotypes were defined by two alleles of a pair of loci. For example, haplotype BMS490-4 and ETH10-2 represents a segment of chromosome having allele 4 of BMS490 and allele 2 of ETH10. The genetic map distance was indicated in centimorgan. The dashed line represented the comparison-wise P-value threshold level, whereas the solid line represented the chromosome-wise P-value threshold level. Figure 1. View largeDownload slide Haplotypes with lowest P-values between two adjacent loci along bovine chromosomes 2, 5, 6, 19, 21, and 23 for backfat EBV in the M1 commercial line of Bos taurus from Beefboster Inc., Calgary, Canada. Haplotypes were defined by two alleles of a pair of loci. For example, haplotype BMS490-4 and ETH10-2 represents a segment of chromosome having allele 4 of BMS490 and allele 2 of ETH10. The genetic map distance was indicated in centimorgan. The dashed line represented the comparison-wise P-value threshold level, whereas the solid line represented the chromosome-wise P-value threshold level. Haplotype Identification and Fine Mapping of QTL for Backfat Haplotype identification and fine mapping of QTL for backfat EBV on chromosomes 2, 5, 6, 19, 21, and 23 were carried out using the IBD haplotype sharing analysis as described by Li et al. (2002a,b). Genotypes of the genotyped microsatellite markers of each of the 176 male calves were checked against the calf's sire to verify the sire inheritance. Alleles of each locus contributed by the sire, as well as by the dam, were identified for each calf by examining the genotype of their sires. The haplotypes (allele linkage phases) of each male calf were then established along chromosomes 2, 5, 6, 19, 21, and 23. The GLM procedure of SAS (SAS Inst., Inc., Cary, NC) was used to test the association between each of the most commonly observed haplotypes and the backfat EBV. The linear model was: Yij = μ + Hi + Eij, where Yij = the backfat EBV of animal j for haplotype i, μ = overall experimental mean, Hi = fixed effect corresponding to the haplotype effect under test (1 when the individual has the haplotype or 0 when the individual is without the haplotype), and Eij = residual error. Because the number of animals carrying two copies of a haplotype were small in the data set, animals carrying two copies of a haplotype were grouped with animals carrying one copy of a haplotype as haplotype class“1.” Animals with uncertain haplotypes were considered to be missing values and were deleted from the analysis. Type III sums of squares were used in all F-tests. The haplotype effect in standard deviations was estimated by dividing the difference of backfat EBV least squares means between haplotype classes“1” and“0” by the standard deviation of the trait. The comparison-wise and chromosome-wise thresholds of the P-value were generated empirically from the permutation method outlined by Churchill and Doerge (1994) and as described by Li et al. (2002b). Type I error rates of 0.05 and 0.10 were used for calculating comparison-wise and chromosome-wise P-value thresholds, respectively. Results On average, 6.5 alleles were detected for each locus on the chromosomes, with a range of 2 to 14 alleles per locus. Associations between a haplotype and backfat EBV were analyzed only for the haplotypes of two adjacent loci with frequencies of 8% or higher. Haplotypes that extended more than two loci were not included in the final haplotype association analysis because of their low frequencies in the dataset compared with those of haplotypes of two adjacent loci. Table 1 lists the haplotypes that had significant associations with backfat EBV above the comparison-wise and chromosome-wise thresholds, and Figure 1 depicts the haplotypes of adjacent loci with the lowest P-values along BTA2, 5, 6, 19, 21 and 23. Table 1. Significant association between haplotypes and backfat EBV on bovine chromosomes 5, 6, 19, 21, and 23 in the M1 line of Bos taurusa Haplotypeb Chromosome P-valuec Haplotype effectd BMS490-4, ETH10-2 5 0.022* +0.62 SD (0.2007) BMS490-2, ETH10-3 5 0.007* −0.72 SD (0.2331) INRA133-208, ILSTS090-149 6 0.019* +0.38 SD (0.1230) BMS470-68, BMS360-8 6 0.043* +0.43 SD (0.1392) OAREL03-4, ILSTS087-2 6 0.024* +0.42 SD (0.1360) BM6000-6, BMS745-2 19 0.025* −0.67 SD (0.2169) RM222-0, BP20-3 19 0.003** −1.33 SD (0.4305) CSSM65-2, BMS1069-3 19 0.027* −0.38 SD (0.1230) BMS1069-3, RM388-2 19 0.031* −0.57 SD (0.1845) RM388-4, BMS601-3 19 0.019* +0.47 SD (0.1521) BMS2815-99, ILSTS092-174 21 0.022* −0.73 SD (0.2363) RM185-103, BM1818-263 23 0.020* +0.69 SD (0.2234) Haplotypeb Chromosome P-valuec Haplotype effectd BMS490-4, ETH10-2 5 0.022* +0.62 SD (0.2007) BMS490-2, ETH10-3 5 0.007* −0.72 SD (0.2331) INRA133-208, ILSTS090-149 6 0.019* +0.38 SD (0.1230) BMS470-68, BMS360-8 6 0.043* +0.43 SD (0.1392) OAREL03-4, ILSTS087-2 6 0.024* +0.42 SD (0.1360) BM6000-6, BMS745-2 19 0.025* −0.67 SD (0.2169) RM222-0, BP20-3 19 0.003** −1.33 SD (0.4305) CSSM65-2, BMS1069-3 19 0.027* −0.38 SD (0.1230) BMS1069-3, RM388-2 19 0.031* −0.57 SD (0.1845) RM388-4, BMS601-3 19 0.019* +0.47 SD (0.1521) BMS2815-99, ILSTS092-174 21 0.022* −0.73 SD (0.2363) RM185-103, BM1818-263 23 0.020* +0.69 SD (0.2234) a Beefbooster Inc., Calgary, Canada. b The haplotypes were named by two alleles of a pair of loci. For example, haplotype BMS490-4 and ETH10-2 represented a segment of chromosome having allele 4 of BMS490 and allele 2 of ETH10. c Indicated the effects were significant at the comparison-wise (*) and chromosome-wise (**) P-value thresholds. d SD = standard deviation, + and − represented the positive and negative effects, respectively. The actual haplotype effects on the backfat EBV in mm are shown in parentheses. View Large Table 1. Significant association between haplotypes and backfat EBV on bovine chromosomes 5, 6, 19, 21, and 23 in the M1 line of Bos taurusa Haplotypeb Chromosome P-valuec Haplotype effectd BMS490-4, ETH10-2 5 0.022* +0.62 SD (0.2007) BMS490-2, ETH10-3 5 0.007* −0.72 SD (0.2331) INRA133-208, ILSTS090-149 6 0.019* +0.38 SD (0.1230) BMS470-68, BMS360-8 6 0.043* +0.43 SD (0.1392) OAREL03-4, ILSTS087-2 6 0.024* +0.42 SD (0.1360) BM6000-6, BMS745-2 19 0.025* −0.67 SD (0.2169) RM222-0, BP20-3 19 0.003** −1.33 SD (0.4305) CSSM65-2, BMS1069-3 19 0.027* −0.38 SD (0.1230) BMS1069-3, RM388-2 19 0.031* −0.57 SD (0.1845) RM388-4, BMS601-3 19 0.019* +0.47 SD (0.1521) BMS2815-99, ILSTS092-174 21 0.022* −0.73 SD (0.2363) RM185-103, BM1818-263 23 0.020* +0.69 SD (0.2234) Haplotypeb Chromosome P-valuec Haplotype effectd BMS490-4, ETH10-2 5 0.022* +0.62 SD (0.2007) BMS490-2, ETH10-3 5 0.007* −0.72 SD (0.2331) INRA133-208, ILSTS090-149 6 0.019* +0.38 SD (0.1230) BMS470-68, BMS360-8 6 0.043* +0.43 SD (0.1392) OAREL03-4, ILSTS087-2 6 0.024* +0.42 SD (0.1360) BM6000-6, BMS745-2 19 0.025* −0.67 SD (0.2169) RM222-0, BP20-3 19 0.003** −1.33 SD (0.4305) CSSM65-2, BMS1069-3 19 0.027* −0.38 SD (0.1230) BMS1069-3, RM388-2 19 0.031* −0.57 SD (0.1845) RM388-4, BMS601-3 19 0.019* +0.47 SD (0.1521) BMS2815-99, ILSTS092-174 21 0.022* −0.73 SD (0.2363) RM185-103, BM1818-263 23 0.020* +0.69 SD (0.2234) a Beefbooster Inc., Calgary, Canada. b The haplotypes were named by two alleles of a pair of loci. For example, haplotype BMS490-4 and ETH10-2 represented a segment of chromosome having allele 4 of BMS490 and allele 2 of ETH10. c Indicated the effects were significant at the comparison-wise (*) and chromosome-wise (**) P-value thresholds. d SD = standard deviation, + and − represented the positive and negative effects, respectively. The actual haplotype effects on the backfat EBV in mm are shown in parentheses. View Large In total, 11 haplotypes were found to have significant associations with backfat EBV at the comparison-wise P-value threshold and one was found at the chromosome-wise P-value threshold on BTA5, 6, 19, 21, and 23 (Table 1). On average, the 12 significant haplotypes had an effect of 0.62 SD on the backfat, ranging from 0.38 to 1.33 SD (Table 1). None of the haplotypes on bovine chromosome 2 showed significant effects on the backfat EBV at the comparison-wise threshold. On BTA5, haplotype BMS490-4, ETH10-2 had a significant positive effect on the backfat EBV at the comparison-wise threshold. The haplotype represented the chromosomal region of 65.4 to 70.0 cM. Animals with the haplotype had a backfat EBV 0.62 SD higher than animals without the haplotype. In the same chromosomal region, an alternative haplotype, BMS490-2, ETH10-3, had a negative effect on the backfat EBV at the comparison-wise threshold, decreasing the backfat EBV by 0.72 SD. Three haplotypes on BTA6 were found to have significant effects on the backfat EBV at the comparison-wise thresholds. The three haplotypes, INRA133-208, ILSTS090-149; BMS470-68, BMS360-8; and OAREL03-4, ILSTS087-2 were located in three chromosomal regions of 8.2 to 11.8 cM, 63.6 to 68.1 cM, and 81.5 to 83.0 cM, respectively (Figure 1). All three haplotypes showed significant positive effects on backfat, increasing the backfat EBV by 0.38, 0.43, and 0.42 SD, respectively. Similarly, three chromosomal regions on BTA19 were identified as having significant effects on the backfat EBV. The three chromosomal regions were represented by five haplotypes (Figure 1). In the chromosomal region of 4.8 to 15.9 cM, haplotype BM6000-6, BMS745-2 had a significant negative effect on the backfat EBV at the comparison-wise threshold, decreasing it by 0.67 SD. In the chromosomal region of 39.4 to 46.5 cM, haplotype RM222-0, BP20-3 also had a negative effect on the backfat EBV, and the significance level reached the chromosome-wise threshold. Animals with the haplotype had 1.33 SD lower backfat EBV than the animals without the haplotype. In the chromosomal region of 65.7 to 99.5 cM, three haplotypes were found to have significant effects on the backfat EBV at the comparison-wise threshold. Haplotypes CSSM65-2, BMS1069-3 and BMS1069-3, RM388-2 had significant negative effects on the backfat EBV, decreasing it by 0.38 and 0.57 SD, respectively. Haplotype RM388-4, BMS601-3, however, showed a significant positive effect on the backfat EBV, increasing it by 0.47 SD. On BTA21, haplotype BMS2815-99, ILSTS092-174 in the chromosomal region of 46.1 to 53.1 cM was the only haplotype that showed a significant effect on the backfat EBV at the comparison-wise level. The haplotype had a negative effect on the backfat EBV, decreasing it by 0.73 SD. Haplotype RM185-103, BM1818-263 was the only haplotype on BTA23 that showed a significant association with backfat at the comparison-wise threshold. The haplotype spanned the chromosomal region of 45.1 to 50.9 cM and had a significant positive effect on the backfat EBV. Animals with the haplotype had a backfat EBV 0.69 SD higher than the animals without the haplotype. Discussion The successful application of marker-assisted selection in commercial animal populations will depend on a number of factors. Among these are the ability to identify the genes or closely linked markers to the genes underlying the QTL, the ability to test whether allelic variations at these loci are segregating in the population, and an understanding of how these genes interact with the environment or with other genes affecting economic traits. All this must be done in an efficient and cost-effective manner in order for the technology to be adopted by the livestock industries. Identity-by-descent QTL mapping using haplotype sharing has been successfully demonstrated in humans (de Vries et al., 1996; Fallin et al., 2001) and cattle (Riquet et al., 1999). The method takes advantage of linkage disequilibrium in populations with limited outbreeding, in which common chromosome segments are shared by individuals in populations that originated from a few common founders. Thus, chromosome segments that house the QTL can be identified through direct haplotype comparison. The feasibility of using haplotype-mapping methods depends on the extent of the linkage disequilibrium. Farnir et al. (2000) reported that linkage disequilibrium in a Holstein-Friesian dairy cattle population extended over several tens of centimorgans. In this study, we observed a level of linkage disequilibrium similar to that seen in dairy cattle, and some haplotypes between two adjacent markers had much higher frequencies than others in the M1 line (data not shown). Such a phenomenon may be attributed to the introduction of a limited number of founders and artificial selection over generations, a common breeding practice in beef cattle as well as in dairy cattle. In a commercial breeding line, selection may play an even more important role in maintaining linkage disequilibrium. Selection that is in favor of desired traits increases the percentage of IBD haplotypes housing the corresponding genes, and thus makes IBD mapping based on haplotype sharing analysis even more feasible. In our previous studies, we successfully mapped QTL for birth weight, preweaning ADG, and ADG on feed in both the M1 and M3 commercial lines of Beefbooster Inc. using the IBD haplotype-sharing analysis, and narrowed down some of the QTL regions to less than 10 cM (Li et al., 2002a,b). The IBD haplotype-sharing analysis detected the same, but better defined, QTL regions in comparison to the interval-mapping method (Li et al., 2002a). In addition to the actual phenotypic data, we have also used the birth weight EBV data for QTL fine mapping on BTA5 and found that the QTL regions for birth weight identified using the primary phenotypic data were in very good agreement with those detected using EBV data (Li et al., 2002a,b). We also mapped a QTL region for backfat on bovine chromosome 14 using backfat EBV data and found that the QTL region was consistent with other studies (Moore et al., 2003). In this study, we identified a total of nine chromosomal regions, one on BTA5 (65.4 to 70.0 cM), three on 6 (8.2 to 11.8 cM, 63.6 to 68.1 cM, and 81.5 to 83.0 cM), three on 19 (4.8 to 15.9 cM, 39.4 to 46.5 cM, and 65.7 to 99.5 cM), one on 21 (46.1 to 53.1 cM), and one on 23 (45.1 to 50.9 cM) that had significant associations with backfat EBV. Among the nine QTL regions, three QTL regions showed remarkable consistency with those identified by other studies and six were new QTL regions for backfat. Casas et al. (2000) reported a QTL for fat depth in the chromosomal region of 40 to 80 cM on bovine chromosome 5. We confirmed the QTL region in the chromosomal region of 65.4 to 70.0 cM and narrowed it down to about 5 cM. On bovine chromosome 6, Wiener et al. (2000) identified one QTL for milk fat yield in the region of 73 to 91 cM in a Holstein-Friesian family, similar to the QTL region of 81.5 to 83.0 cM that we identified for backfat in this study. Whether the two QTL regions represent the same QTL or separate QTL for milk fat yield and backfat, however, remains to be determined. On BTA19, Taylor et al. (1998) reported that QTL for subcutaneous fat and ether-extractable fat were located in the chromosome region of approximately 60 to 80 cM that harbored the growth hormone 1 gene. In this study, we identified a similar chromosomal region of 65.7 to 99.5 cM that showed a significant association with backfat. Such consistency strongly indicates the effectiveness of identification and fine mapping QTL in commercial lines of livestock using the IBD haplotype sharing method. The M1 line has been developed as a maternal component of a commercial crossbreeding scheme. Selection is based on an index described by MacNeil and Newman (1994), along with independent culling levels specifying minimum and maximum birth weight, minimum preweaning ADG, and minimum ADG on feed in M1 line individuals. The selection index was constructed based on 18 different traits and selection was in favor of greater fat depth in the M1 line (MacNeil and Newman, 1994). Among the 12 haplotypes with higher frequencies and also showing significant associations with the backfat, six haplotypes had positive effects and six had negative effects on backfat, which suggests that selection in favor of backfat may not be strong in the M1 line. This emphasizes the care that must be taken in implementing marker-assisted selection when only one or a few markers are considered. Selection on a marker may also have negative effects on other traits due to pleiotropic effects of the gene or due to other genes closely linked to the marker affecting the other traits. Implications Quantitative trait loci for backfat in beef cattle have been identified and fine mapped on bovine chromosomes 5, 6, 19, 21, and 23 in a commercial line of Bos taurus. In total, nine regions on these five chromosomes were identified to have significant associations with backfat estimated breeding value. The results should provide a valuable reference for further positional candidate gene research and marker-assisted selection. Literature Cited Casas, E., S. D. Shackelford, J. W. Keele, R. T. Stone, S. M. Kappes, and M. Koohmaraie 2000. Quantitative trait loci affecting growth and carcass composition of cattle segregating alternative forms of myostatin. J. Anim. Sci. 78: 560– 569. Google Scholar CrossRef Search ADS PubMed Casas, E., R. T. Stone, J. W. Keele, S. D. Shackelford, S. M. Kappes, and M. Koohmaraie 2001. A comprehensive search for quantitative trait loci affecting growth and carcass composition of cattle segregating alternative forms of the myostatin gene. J. Anim. Sci. 79: 854– 860. Google Scholar CrossRef Search ADS PubMed Churchill, G. A., and R. W. Doerge 1994. Empirical threshold values for quantitative trait mapping. Genetics 138: 963– 971. Google Scholar PubMed de Vries, H. G., M. A. van der Meulen, R. Rozen, D. J. J. Hally, H. Scheffer, L. P. ten Kate, C. H. C. M. Buys, and G. J. te Meerman 1996. Haplotype identity between individuals who share a CFTR mutation allele ‘identical by descent’: Demonstration of the usefulness of the haplotype-sharing concept for gene mapping in real populations. Hum. Genet. 98: 304– 309. Google Scholar CrossRef Search ADS PubMed Fallin, D., A. Cohen, L. Essioux, I. Chumakov, M. Blumenfeld, D. Cohen, and N. J. Schork 2001. Genetic analysis of case/control data using estimated haplotype frequencies: Application to APOE locus variation and Alzheimer's disease. Genome Res. 11: 143– 151. Google Scholar CrossRef Search ADS PubMed Farnir, F., W. Coppieters, J. Arranz, P. Berzi, N. Cambisano, B. Grisart, L. Karim, F. Marcq, L. Moreau, M. Mni, C. Nezer, P. Simon, P. Vanmanshoven, D. Wagenaar, and M. Georges 2000. Extensive genome-wide linkage disequlibrium in cattle. Genome Res. 10: 220– 227. Google Scholar CrossRef Search ADS PubMed Jeremiah, L. E. 1996. The influence of subcutaneous fat thickness and marbling on beef. Food Res. Int. 29: 513– 520. Google Scholar CrossRef Search ADS Li, C., J. Basarab, W. M. Snelling, B. Benkel, B. Murdoch, J. Kneeland, C. Hansen, and S. S. Moore 2002a. Identical by descent haplotype sharing analysis: Application in fine mapping of QTLs for birth weight in commerical lines of Bos taurus. Pages 481–484 in Proc. 7th World Congr. Genet. Appl. Livest. Prod. Montpellier, France. Li, C., J. Basarab, W. M. Snelling, B. Benkel, B. Murdoch, and S. S. Moore 2002b. The identification of common haplotypes on bovine chromosome 5 within commercial lines of Bos taurus and their associations with growth traits. J. Anim. Sci. 80: 1187– 1194. Google Scholar CrossRef Search ADS Lozeman, F. J., C. K. Middleton, J. Deng, E. C. Kazala, C. Verhaege, P. S. Mir, A. Laroche, D. R. C. Bailey, and R. J. Weselake 2001. Characterization of microsomal diacylglycerol acytransferase activity from bovine adipose and muscle tissue. Comp. Biochem. Physiol. Part B 130: 105– 115. Google Scholar CrossRef Search ADS MacNeil, M. D., and M. D. Grosz 2002. Genome-wide scans for QTL affecting carcass traits in Hereford × composite double backcross populations. J. Anim. Sci. 80: 2316– 2324. Google Scholar PubMed MacNeil, M. D., and S. Newman 1994. Selection indices for Canadian beef production using specialized sire and dam lines. Can. J. Anim. Sci. 74: 419– 424. Google Scholar CrossRef Search ADS Moore, S. S., C. Li, J. Basarab, W. M. Snelling, J. Kneeland, B. Murdoch, C. Hansen, and B. Benkel 2003. Fine mapping of quantitative trait loci and assessment of positional candidate genes for backfat on bovine chromosome 14 in a commercial line of Bos taurus. J. Anim. Sci. 81: 1919– 1925. Google Scholar CrossRef Search ADS PubMed Powell W. E., and D. L. Huffman 1973. Predicting chemical composition of beef carcass from easily obtainable carcass variables. J. Anim. Sci. 36: 1069– 1076. Google Scholar CrossRef Search ADS Riquet, J., W. Coppieters, N. Cambisano, J.-J. Arranz, P. Berzi, S. K. Davis, B. Grisart, F. Farnir, L. Karim, M. Mni, P. Simon, J. F. Tayor, P. Vanmanshoven, D. Wagenaar, J. E. Womack, and M. Georges 1999. Fine-mapping of quantitative trait loci by identity by descent in outbred populations: Application to milk production in dairy cattle. Proc. Natl. Acad. Sci. USA 96: 9252– 9257. Google Scholar CrossRef Search ADS Stone, R., T. J. W. Keele, S. D. Shackelford, S. M. Kappes, and M. Koohmaraie 1999. A primary screen of the bovine genome for quantitative trait loci affecting carcass and growth traits. J. Anim. Sci. 77: 1379– 1384. Google Scholar CrossRef Search ADS PubMed Taylor, J. F., L. L. Coutinho, K. L. Herring, D. S. Gallagher Jr., R. A. Brenneman, N. Burney, J. O. Sanders, J. W. Turner, S. B. Smith, R. K. Miller, J. W. Savell, and S. K. Davis 1998. Candidate gene analysis of GH1 for effects on growth and carcass composition of cattle. Anim. Genet. 29: 194– 201. Google Scholar CrossRef Search ADS PubMed Wheeler T. L., L. V. Cundiff, and R. M. Koch 1994. Effect of marbling degree on beef palatability in Bos taurus and Bos indicus cattle. J. Anim. Sci. 72: 3145– 3151. Google Scholar CrossRef Search ADS PubMed Wiener, P., I. Maclean, J. L. Williams, and J. A. Woolliams 2000. Testing for the presence of previously identified QTL for milk production traits in new populations. Anim. Genet. 31: 385– 395. Google Scholar CrossRef Search ADS PubMed Footnotes 1 This work was supported through grant No. 2000M624 awarded to S. S. Moore through the Alberta Agricultural Research Institute and grant No. ACC-99AB343 awarded to B. Benkel and S. S. Moore through the Canada-Alberta Beef Industry Development Fund. Copyright 2004 Journal of Animal Science
Identification and fine mapping of quantitative trait loci for backfat on bovine chromosomes 2, 5, 6, 19, 21, and 23 in a commercial line of Bos taurusLi, C.;Basarab, J.;Snelling, W. M.;Benkel, B.;Kneeland, J.;Murdoch, B.;Hansen, C.;Moore, S. S.
doi: 10.1093/ansci/82.4.967pmid: N/A
Abstract Backfat thickness is one of the major quantitative traits that affects carcass quality in beef cattle. In this study, we identified and fine-mapped QTL for backfat EBV on bovine chromosomes 2, 5, 6, 19, 21, and 23 using an identical-by-descent haplotype-sharing analysis in a commercial line of Bos taurus. Eleven haplotypes were found to have significant associations with backfat EBV at the comparison-wise P-value threshold, and one at the chromosome-wise P-value threshold on bovine chromosomes 5, 6, 19, 21, and 23. On average, the 12 significant haplotypes had an effect of 0.62 SD on backfat EBV, ranging from 0.38 SD to 1.33 SD. The 12 significant haplotypes spanned nine chromosomal regions, one on chromosome 5 (65.4 to 70.0 cM), three on 6 (8.2 to 11.8 cM, 63.6 to 68.1 cM, and 81.5 to 83.0 cM), three on 19 (4.8 to 15.9 cM, 39.4 to 46.5 cM, and 65.7 to 99.5 cM), one on 21 (46.1 to 53.1 cM), and one on 23 (45.1 to 50.9 cM). Among the nine chromosomal regions, six were new QTL regions and three showed remarkable agreement with QTL regions that were previously reported. Eight of the nine QTL regions were localized to less than or close to 10 cM in genetic distance. The results provide a useful reference for further positional candidate gene research and marker-assisted selection for backfat. Introduction The amount and distribution of fat has an important impact on carcass and meat quality in beef cattle (Powell and Huffman, 1973; Wheeler et al., 1994; Lozeman et al., 2001). Beef carcasses can be stratified by using subcutaneous fat thickness alone or in combination with marbling (Jeremiah, 1996). Breeding for optimal fat is therefore one of the major goals toward better profitability in the beef industry. Mapping of QTL and identification of causative genes that affect fat metabolism will enhance the progress toward this goal. Quantitative trait loci mapping for quantitative traits related to fat has been reported in a number of studies (Stone et al., 1999; Casas et al., 2001; MacNeil and Grosz, 2002). Fine mapping and verification of those QTL for fat are therefore necessary for further candidate gene research. Individuals within a semiclosed population, such as a commercial line of cattle, are expected to be derived from one or a limited number of founders. Some common haplotypes originating from the common ancestors should therefore carry on and segregate among the individuals of the breeding line, particularly when selection is applied. These common haplotypes may harbor QTL of interest and make it possible to locate QTL segregating in the line. This identical-by-descent (IBD) haplotype-sharing QTL mapping strategy avoids the generation of a well-designed mapping population. Obtaining such well-designed mapping populations is costly and may be impractical in commercial herds. In our previous studies, we have successfully identified and fine-mapped QTL for growth on bovine chromosome (BTA) 5 (Li et al., 2002a,b), as well as for backfat EBV on BTA14 (Moore et al., 2003) in commercial lines of Bos taurus using the IBD haplotype-sharing analysis method. We report here the identification and fine mapping of QTL for backfat EBV on BTA2, 5, 6, 19, 21, and 23 in a commercial population of Bos taurus. Materials and Methods Animals and Phenotypic Data Animals were from the M1 line of Beefbooster Inc. (Calgary, Canada) and were born in 1998. The M1 line was developed from an Angus base and has been under selection for over 30 yr. The selection criteria for the lines are based on indices described by MacNeil and Newman (1994). A 10-mL blood sample was collected by venipuncture from each male calf and the potential sires, and the DNA from each blood sample was extracted and kept for later parentage identification. Sire identification was carried out by the Saskatchewan Research Council, Canada, using DNA microsatellite markers. The backfat EBV of each male calf was calculated based on a BLUP procedure by Beefbooster Inc. Genotyping One hundred seventy-six male calves and their 12 respective sires (nine to 30 calves of each sire) of the M1 line were genotyped using 76 microsatellite markers, 13 from BTA2, 16 from BTA5, 16 from BTA6, 14 from BTA19, 8 from BTA21, and 9 from BTA23. The microsatellite markers genotyped on each chromosome (Figure 1) were chosen at an approximately even genetic distance and spanned on average 93.4% of the chromosomes, with a range of 83.6 (chromosome 6) to 99.7% (chromosome 23). Marker locations and primers designed for genotyping were based on the microsatellite marker information available on the USMARC Website (http://www.marc.usda.gov/genome/genome.html). Figure 1. View largeDownload slide Haplotypes with lowest P-values between two adjacent loci along bovine chromosomes 2, 5, 6, 19, 21, and 23 for backfat EBV in the M1 commercial line of Bos taurus from Beefboster Inc., Calgary, Canada. Haplotypes were defined by two alleles of a pair of loci. For example, haplotype BMS490-4 and ETH10-2 represents a segment of chromosome having allele 4 of BMS490 and allele 2 of ETH10. The genetic map distance was indicated in centimorgan. The dashed line represented the comparison-wise P-value threshold level, whereas the solid line represented the chromosome-wise P-value threshold level. Figure 1. View largeDownload slide Haplotypes with lowest P-values between two adjacent loci along bovine chromosomes 2, 5, 6, 19, 21, and 23 for backfat EBV in the M1 commercial line of Bos taurus from Beefboster Inc., Calgary, Canada. Haplotypes were defined by two alleles of a pair of loci. For example, haplotype BMS490-4 and ETH10-2 represents a segment of chromosome having allele 4 of BMS490 and allele 2 of ETH10. The genetic map distance was indicated in centimorgan. The dashed line represented the comparison-wise P-value threshold level, whereas the solid line represented the chromosome-wise P-value threshold level. Haplotype Identification and Fine Mapping of QTL for Backfat Haplotype identification and fine mapping of QTL for backfat EBV on chromosomes 2, 5, 6, 19, 21, and 23 were carried out using the IBD haplotype sharing analysis as described by Li et al. (2002a,b). Genotypes of the genotyped microsatellite markers of each of the 176 male calves were checked against the calf's sire to verify the sire inheritance. Alleles of each locus contributed by the sire, as well as by the dam, were identified for each calf by examining the genotype of their sires. The haplotypes (allele linkage phases) of each male calf were then established along chromosomes 2, 5, 6, 19, 21, and 23. The GLM procedure of SAS (SAS Inst., Inc., Cary, NC) was used to test the association between each of the most commonly observed haplotypes and the backfat EBV. The linear model was: Yij = μ + Hi + Eij, where Yij = the backfat EBV of animal j for haplotype i, μ = overall experimental mean, Hi = fixed effect corresponding to the haplotype effect under test (1 when the individual has the haplotype or 0 when the individual is without the haplotype), and Eij = residual error. Because the number of animals carrying two copies of a haplotype were small in the data set, animals carrying two copies of a haplotype were grouped with animals carrying one copy of a haplotype as haplotype class“1.” Animals with uncertain haplotypes were considered to be missing values and were deleted from the analysis. Type III sums of squares were used in all F-tests. The haplotype effect in standard deviations was estimated by dividing the difference of backfat EBV least squares means between haplotype classes“1” and“0” by the standard deviation of the trait. The comparison-wise and chromosome-wise thresholds of the P-value were generated empirically from the permutation method outlined by Churchill and Doerge (1994) and as described by Li et al. (2002b). Type I error rates of 0.05 and 0.10 were used for calculating comparison-wise and chromosome-wise P-value thresholds, respectively. Results On average, 6.5 alleles were detected for each locus on the chromosomes, with a range of 2 to 14 alleles per locus. Associations between a haplotype and backfat EBV were analyzed only for the haplotypes of two adjacent loci with frequencies of 8% or higher. Haplotypes that extended more than two loci were not included in the final haplotype association analysis because of their low frequencies in the dataset compared with those of haplotypes of two adjacent loci. Table 1 lists the haplotypes that had significant associations with backfat EBV above the comparison-wise and chromosome-wise thresholds, and Figure 1 depicts the haplotypes of adjacent loci with the lowest P-values along BTA2, 5, 6, 19, 21 and 23. Table 1. Significant association between haplotypes and backfat EBV on bovine chromosomes 5, 6, 19, 21, and 23 in the M1 line of Bos taurusa Haplotypeb Chromosome P-valuec Haplotype effectd BMS490-4, ETH10-2 5 0.022* +0.62 SD (0.2007) BMS490-2, ETH10-3 5 0.007* −0.72 SD (0.2331) INRA133-208, ILSTS090-149 6 0.019* +0.38 SD (0.1230) BMS470-68, BMS360-8 6 0.043* +0.43 SD (0.1392) OAREL03-4, ILSTS087-2 6 0.024* +0.42 SD (0.1360) BM6000-6, BMS745-2 19 0.025* −0.67 SD (0.2169) RM222-0, BP20-3 19 0.003** −1.33 SD (0.4305) CSSM65-2, BMS1069-3 19 0.027* −0.38 SD (0.1230) BMS1069-3, RM388-2 19 0.031* −0.57 SD (0.1845) RM388-4, BMS601-3 19 0.019* +0.47 SD (0.1521) BMS2815-99, ILSTS092-174 21 0.022* −0.73 SD (0.2363) RM185-103, BM1818-263 23 0.020* +0.69 SD (0.2234) Haplotypeb Chromosome P-valuec Haplotype effectd BMS490-4, ETH10-2 5 0.022* +0.62 SD (0.2007) BMS490-2, ETH10-3 5 0.007* −0.72 SD (0.2331) INRA133-208, ILSTS090-149 6 0.019* +0.38 SD (0.1230) BMS470-68, BMS360-8 6 0.043* +0.43 SD (0.1392) OAREL03-4, ILSTS087-2 6 0.024* +0.42 SD (0.1360) BM6000-6, BMS745-2 19 0.025* −0.67 SD (0.2169) RM222-0, BP20-3 19 0.003** −1.33 SD (0.4305) CSSM65-2, BMS1069-3 19 0.027* −0.38 SD (0.1230) BMS1069-3, RM388-2 19 0.031* −0.57 SD (0.1845) RM388-4, BMS601-3 19 0.019* +0.47 SD (0.1521) BMS2815-99, ILSTS092-174 21 0.022* −0.73 SD (0.2363) RM185-103, BM1818-263 23 0.020* +0.69 SD (0.2234) a Beefbooster Inc., Calgary, Canada. b The haplotypes were named by two alleles of a pair of loci. For example, haplotype BMS490-4 and ETH10-2 represented a segment of chromosome having allele 4 of BMS490 and allele 2 of ETH10. c Indicated the effects were significant at the comparison-wise (*) and chromosome-wise (**) P-value thresholds. d SD = standard deviation, + and − represented the positive and negative effects, respectively. The actual haplotype effects on the backfat EBV in mm are shown in parentheses. View Large Table 1. Significant association between haplotypes and backfat EBV on bovine chromosomes 5, 6, 19, 21, and 23 in the M1 line of Bos taurusa Haplotypeb Chromosome P-valuec Haplotype effectd BMS490-4, ETH10-2 5 0.022* +0.62 SD (0.2007) BMS490-2, ETH10-3 5 0.007* −0.72 SD (0.2331) INRA133-208, ILSTS090-149 6 0.019* +0.38 SD (0.1230) BMS470-68, BMS360-8 6 0.043* +0.43 SD (0.1392) OAREL03-4, ILSTS087-2 6 0.024* +0.42 SD (0.1360) BM6000-6, BMS745-2 19 0.025* −0.67 SD (0.2169) RM222-0, BP20-3 19 0.003** −1.33 SD (0.4305) CSSM65-2, BMS1069-3 19 0.027* −0.38 SD (0.1230) BMS1069-3, RM388-2 19 0.031* −0.57 SD (0.1845) RM388-4, BMS601-3 19 0.019* +0.47 SD (0.1521) BMS2815-99, ILSTS092-174 21 0.022* −0.73 SD (0.2363) RM185-103, BM1818-263 23 0.020* +0.69 SD (0.2234) Haplotypeb Chromosome P-valuec Haplotype effectd BMS490-4, ETH10-2 5 0.022* +0.62 SD (0.2007) BMS490-2, ETH10-3 5 0.007* −0.72 SD (0.2331) INRA133-208, ILSTS090-149 6 0.019* +0.38 SD (0.1230) BMS470-68, BMS360-8 6 0.043* +0.43 SD (0.1392) OAREL03-4, ILSTS087-2 6 0.024* +0.42 SD (0.1360) BM6000-6, BMS745-2 19 0.025* −0.67 SD (0.2169) RM222-0, BP20-3 19 0.003** −1.33 SD (0.4305) CSSM65-2, BMS1069-3 19 0.027* −0.38 SD (0.1230) BMS1069-3, RM388-2 19 0.031* −0.57 SD (0.1845) RM388-4, BMS601-3 19 0.019* +0.47 SD (0.1521) BMS2815-99, ILSTS092-174 21 0.022* −0.73 SD (0.2363) RM185-103, BM1818-263 23 0.020* +0.69 SD (0.2234) a Beefbooster Inc., Calgary, Canada. b The haplotypes were named by two alleles of a pair of loci. For example, haplotype BMS490-4 and ETH10-2 represented a segment of chromosome having allele 4 of BMS490 and allele 2 of ETH10. c Indicated the effects were significant at the comparison-wise (*) and chromosome-wise (**) P-value thresholds. d SD = standard deviation, + and − represented the positive and negative effects, respectively. The actual haplotype effects on the backfat EBV in mm are shown in parentheses. View Large In total, 11 haplotypes were found to have significant associations with backfat EBV at the comparison-wise P-value threshold and one was found at the chromosome-wise P-value threshold on BTA5, 6, 19, 21, and 23 (Table 1). On average, the 12 significant haplotypes had an effect of 0.62 SD on the backfat, ranging from 0.38 to 1.33 SD (Table 1). None of the haplotypes on bovine chromosome 2 showed significant effects on the backfat EBV at the comparison-wise threshold. On BTA5, haplotype BMS490-4, ETH10-2 had a significant positive effect on the backfat EBV at the comparison-wise threshold. The haplotype represented the chromosomal region of 65.4 to 70.0 cM. Animals with the haplotype had a backfat EBV 0.62 SD higher than animals without the haplotype. In the same chromosomal region, an alternative haplotype, BMS490-2, ETH10-3, had a negative effect on the backfat EBV at the comparison-wise threshold, decreasing the backfat EBV by 0.72 SD. Three haplotypes on BTA6 were found to have significant effects on the backfat EBV at the comparison-wise thresholds. The three haplotypes, INRA133-208, ILSTS090-149; BMS470-68, BMS360-8; and OAREL03-4, ILSTS087-2 were located in three chromosomal regions of 8.2 to 11.8 cM, 63.6 to 68.1 cM, and 81.5 to 83.0 cM, respectively (Figure 1). All three haplotypes showed significant positive effects on backfat, increasing the backfat EBV by 0.38, 0.43, and 0.42 SD, respectively. Similarly, three chromosomal regions on BTA19 were identified as having significant effects on the backfat EBV. The three chromosomal regions were represented by five haplotypes (Figure 1). In the chromosomal region of 4.8 to 15.9 cM, haplotype BM6000-6, BMS745-2 had a significant negative effect on the backfat EBV at the comparison-wise threshold, decreasing it by 0.67 SD. In the chromosomal region of 39.4 to 46.5 cM, haplotype RM222-0, BP20-3 also had a negative effect on the backfat EBV, and the significance level reached the chromosome-wise threshold. Animals with the haplotype had 1.33 SD lower backfat EBV than the animals without the haplotype. In the chromosomal region of 65.7 to 99.5 cM, three haplotypes were found to have significant effects on the backfat EBV at the comparison-wise threshold. Haplotypes CSSM65-2, BMS1069-3 and BMS1069-3, RM388-2 had significant negative effects on the backfat EBV, decreasing it by 0.38 and 0.57 SD, respectively. Haplotype RM388-4, BMS601-3, however, showed a significant positive effect on the backfat EBV, increasing it by 0.47 SD. On BTA21, haplotype BMS2815-99, ILSTS092-174 in the chromosomal region of 46.1 to 53.1 cM was the only haplotype that showed a significant effect on the backfat EBV at the comparison-wise level. The haplotype had a negative effect on the backfat EBV, decreasing it by 0.73 SD. Haplotype RM185-103, BM1818-263 was the only haplotype on BTA23 that showed a significant association with backfat at the comparison-wise threshold. The haplotype spanned the chromosomal region of 45.1 to 50.9 cM and had a significant positive effect on the backfat EBV. Animals with the haplotype had a backfat EBV 0.69 SD higher than the animals without the haplotype. Discussion The successful application of marker-assisted selection in commercial animal populations will depend on a number of factors. Among these are the ability to identify the genes or closely linked markers to the genes underlying the QTL, the ability to test whether allelic variations at these loci are segregating in the population, and an understanding of how these genes interact with the environment or with other genes affecting economic traits. All this must be done in an efficient and cost-effective manner in order for the technology to be adopted by the livestock industries. Identity-by-descent QTL mapping using haplotype sharing has been successfully demonstrated in humans (de Vries et al., 1996; Fallin et al., 2001) and cattle (Riquet et al., 1999). The method takes advantage of linkage disequilibrium in populations with limited outbreeding, in which common chromosome segments are shared by individuals in populations that originated from a few common founders. Thus, chromosome segments that house the QTL can be identified through direct haplotype comparison. The feasibility of using haplotype-mapping methods depends on the extent of the linkage disequilibrium. Farnir et al. (2000) reported that linkage disequilibrium in a Holstein-Friesian dairy cattle population extended over several tens of centimorgans. In this study, we observed a level of linkage disequilibrium similar to that seen in dairy cattle, and some haplotypes between two adjacent markers had much higher frequencies than others in the M1 line (data not shown). Such a phenomenon may be attributed to the introduction of a limited number of founders and artificial selection over generations, a common breeding practice in beef cattle as well as in dairy cattle. In a commercial breeding line, selection may play an even more important role in maintaining linkage disequilibrium. Selection that is in favor of desired traits increases the percentage of IBD haplotypes housing the corresponding genes, and thus makes IBD mapping based on haplotype sharing analysis even more feasible. In our previous studies, we successfully mapped QTL for birth weight, preweaning ADG, and ADG on feed in both the M1 and M3 commercial lines of Beefbooster Inc. using the IBD haplotype-sharing analysis, and narrowed down some of the QTL regions to less than 10 cM (Li et al., 2002a,b). The IBD haplotype-sharing analysis detected the same, but better defined, QTL regions in comparison to the interval-mapping method (Li et al., 2002a). In addition to the actual phenotypic data, we have also used the birth weight EBV data for QTL fine mapping on BTA5 and found that the QTL regions for birth weight identified using the primary phenotypic data were in very good agreement with those detected using EBV data (Li et al., 2002a,b). We also mapped a QTL region for backfat on bovine chromosome 14 using backfat EBV data and found that the QTL region was consistent with other studies (Moore et al., 2003). In this study, we identified a total of nine chromosomal regions, one on BTA5 (65.4 to 70.0 cM), three on 6 (8.2 to 11.8 cM, 63.6 to 68.1 cM, and 81.5 to 83.0 cM), three on 19 (4.8 to 15.9 cM, 39.4 to 46.5 cM, and 65.7 to 99.5 cM), one on 21 (46.1 to 53.1 cM), and one on 23 (45.1 to 50.9 cM) that had significant associations with backfat EBV. Among the nine QTL regions, three QTL regions showed remarkable consistency with those identified by other studies and six were new QTL regions for backfat. Casas et al. (2000) reported a QTL for fat depth in the chromosomal region of 40 to 80 cM on bovine chromosome 5. We confirmed the QTL region in the chromosomal region of 65.4 to 70.0 cM and narrowed it down to about 5 cM. On bovine chromosome 6, Wiener et al. (2000) identified one QTL for milk fat yield in the region of 73 to 91 cM in a Holstein-Friesian family, similar to the QTL region of 81.5 to 83.0 cM that we identified for backfat in this study. Whether the two QTL regions represent the same QTL or separate QTL for milk fat yield and backfat, however, remains to be determined. On BTA19, Taylor et al. (1998) reported that QTL for subcutaneous fat and ether-extractable fat were located in the chromosome region of approximately 60 to 80 cM that harbored the growth hormone 1 gene. In this study, we identified a similar chromosomal region of 65.7 to 99.5 cM that showed a significant association with backfat. Such consistency strongly indicates the effectiveness of identification and fine mapping QTL in commercial lines of livestock using the IBD haplotype sharing method. The M1 line has been developed as a maternal component of a commercial crossbreeding scheme. Selection is based on an index described by MacNeil and Newman (1994), along with independent culling levels specifying minimum and maximum birth weight, minimum preweaning ADG, and minimum ADG on feed in M1 line individuals. The selection index was constructed based on 18 different traits and selection was in favor of greater fat depth in the M1 line (MacNeil and Newman, 1994). Among the 12 haplotypes with higher frequencies and also showing significant associations with the backfat, six haplotypes had positive effects and six had negative effects on backfat, which suggests that selection in favor of backfat may not be strong in the M1 line. This emphasizes the care that must be taken in implementing marker-assisted selection when only one or a few markers are considered. Selection on a marker may also have negative effects on other traits due to pleiotropic effects of the gene or due to other genes closely linked to the marker affecting the other traits. Implications Quantitative trait loci for backfat in beef cattle have been identified and fine mapped on bovine chromosomes 5, 6, 19, 21, and 23 in a commercial line of Bos taurus. In total, nine regions on these five chromosomes were identified to have significant associations with backfat estimated breeding value. The results should provide a valuable reference for further positional candidate gene research and marker-assisted selection. Literature Cited Casas, E., S. D. Shackelford, J. W. Keele, R. T. Stone, S. M. Kappes, and M. Koohmaraie 2000. Quantitative trait loci affecting growth and carcass composition of cattle segregating alternative forms of myostatin. J. Anim. Sci. 78: 560– 569. Google Scholar CrossRef Search ADS PubMed Casas, E., R. T. Stone, J. W. Keele, S. D. Shackelford, S. M. Kappes, and M. Koohmaraie 2001. A comprehensive search for quantitative trait loci affecting growth and carcass composition of cattle segregating alternative forms of the myostatin gene. J. Anim. Sci. 79: 854– 860. Google Scholar CrossRef Search ADS PubMed Churchill, G. A., and R. W. Doerge 1994. Empirical threshold values for quantitative trait mapping. Genetics 138: 963– 971. Google Scholar PubMed de Vries, H. G., M. A. van der Meulen, R. Rozen, D. J. J. Hally, H. Scheffer, L. P. ten Kate, C. H. C. M. Buys, and G. J. te Meerman 1996. Haplotype identity between individuals who share a CFTR mutation allele ‘identical by descent’: Demonstration of the usefulness of the haplotype-sharing concept for gene mapping in real populations. Hum. Genet. 98: 304– 309. Google Scholar CrossRef Search ADS PubMed Fallin, D., A. Cohen, L. Essioux, I. Chumakov, M. Blumenfeld, D. Cohen, and N. J. Schork 2001. Genetic analysis of case/control data using estimated haplotype frequencies: Application to APOE locus variation and Alzheimer's disease. Genome Res. 11: 143– 151. Google Scholar CrossRef Search ADS PubMed Farnir, F., W. Coppieters, J. Arranz, P. Berzi, N. Cambisano, B. Grisart, L. Karim, F. Marcq, L. Moreau, M. Mni, C. Nezer, P. Simon, P. Vanmanshoven, D. Wagenaar, and M. Georges 2000. Extensive genome-wide linkage disequlibrium in cattle. Genome Res. 10: 220– 227. Google Scholar CrossRef Search ADS PubMed Jeremiah, L. E. 1996. The influence of subcutaneous fat thickness and marbling on beef. Food Res. Int. 29: 513– 520. Google Scholar CrossRef Search ADS Li, C., J. Basarab, W. M. Snelling, B. Benkel, B. Murdoch, J. Kneeland, C. Hansen, and S. S. Moore 2002a. Identical by descent haplotype sharing analysis: Application in fine mapping of QTLs for birth weight in commerical lines of Bos taurus. Pages 481–484 in Proc. 7th World Congr. Genet. Appl. Livest. Prod. Montpellier, France. Li, C., J. Basarab, W. M. Snelling, B. Benkel, B. Murdoch, and S. S. Moore 2002b. The identification of common haplotypes on bovine chromosome 5 within commercial lines of Bos taurus and their associations with growth traits. J. Anim. Sci. 80: 1187– 1194. Google Scholar CrossRef Search ADS Lozeman, F. J., C. K. Middleton, J. Deng, E. C. Kazala, C. Verhaege, P. S. Mir, A. Laroche, D. R. C. Bailey, and R. J. Weselake 2001. Characterization of microsomal diacylglycerol acytransferase activity from bovine adipose and muscle tissue. Comp. Biochem. Physiol. Part B 130: 105– 115. Google Scholar CrossRef Search ADS MacNeil, M. D., and M. D. Grosz 2002. Genome-wide scans for QTL affecting carcass traits in Hereford × composite double backcross populations. J. Anim. Sci. 80: 2316– 2324. Google Scholar CrossRef Search ADS PubMed MacNeil, M. D., and S. Newman 1994. Selection indices for Canadian beef production using specialized sire and dam lines. Can. J. Anim. Sci. 74: 419– 424. Google Scholar CrossRef Search ADS Moore, S. S., C. Li, J. Basarab, W. M. Snelling, J. Kneeland, B. Murdoch, C. Hansen, and B. Benkel 2003. Fine mapping of quantitative trait loci and assessment of positional candidate genes for backfat on bovine chromosome 14 in a commercial line of Bos taurus. J. Anim. Sci. 81: 1919– 1925. Google Scholar CrossRef Search ADS PubMed Powell W. E., and D. L. Huffman 1973. Predicting chemical composition of beef carcass from easily obtainable carcass variables. J. Anim. Sci. 36: 1069– 1076. Google Scholar CrossRef Search ADS Riquet, J., W. Coppieters, N. Cambisano, J.-J. Arranz, P. Berzi, S. K. Davis, B. Grisart, F. Farnir, L. Karim, M. Mni, P. Simon, J. F. Tayor, P. Vanmanshoven, D. Wagenaar, J. E. Womack, and M. Georges 1999. Fine-mapping of quantitative trait loci by identity by descent in outbred populations: Application to milk production in dairy cattle. Proc. Natl. Acad. Sci. USA 96: 9252– 9257. Google Scholar CrossRef Search ADS Stone, R., T. J. W. Keele, S. D. Shackelford, S. M. Kappes, and M. Koohmaraie 1999. A primary screen of the bovine genome for quantitative trait loci affecting carcass and growth traits. J. Anim. Sci. 77: 1379– 1384. Google Scholar CrossRef Search ADS PubMed Taylor, J. F., L. L. Coutinho, K. L. Herring, D. S. Gallagher Jr., R. A. Brenneman, N. Burney, J. O. Sanders, J. W. Turner, S. B. Smith, R. K. Miller, J. W. Savell, and S. K. Davis 1998. Candidate gene analysis of GH1 for effects on growth and carcass composition of cattle. Anim. Genet. 29: 194– 201. Google Scholar CrossRef Search ADS PubMed Wheeler T. L., L. V. Cundiff, and R. M. Koch 1994. Effect of marbling degree on beef palatability in Bos taurus and Bos indicus cattle. J. Anim. Sci. 72: 3145– 3151. Google Scholar CrossRef Search ADS PubMed Wiener, P., I. Maclean, J. L. Williams, and J. A. Woolliams 2000. Testing for the presence of previously identified QTL for milk production traits in new populations. Anim. Genet. 31: 385– 395. Google Scholar CrossRef Search ADS PubMed Footnotes 1 This work was supported through grant No. 2000M624 awarded to S. S. Moore through the Alberta Agricultural Research Institute and grant No. ACC-99AB343 awarded to B. Benkel and S. S. Moore through the Canada-Alberta Beef Industry Development Fund. Copyright 2004 Journal of Animal Science
A resource allocation model describing consequences of artificial selection under metabolic stressvan der Waaij, E. H.
doi: 10.2527/2004.824973xpmid: 15080316
Abstract Long-term selection on production results in increased environmental sensitivity. This often is expressed through decreased fertility and increased health problems. The phenomenon has been described in all common farm animal species. One theory is that potential resource intake is insufficient to express production potential. Additional resources are drawn away from fitness-related traits, such as fertility and health, to further increase observed production. In addition, resources for maintaining fitness depend on the demands by the environment. In a harsh environment, more resources are required for fitness-related traits than in an optimal environment. Literature results show that selection in an optimal environment will increase sensitivity to less optimal environments. The objectives of this paper were to increase understanding of the underlying mechanism behind the development of environmental sensitivity and to gain insight into correlated response(s) when selection is on observed production. A resource allocation model was defined where observed production depended on production potential, resource intake potential, and the allocation of resources to production or fitness, including maintenance, health, and reproduction. Penalties for reproductive performance and probability of survival were included when the proportion of resources assigned to fitness dropped below a certain, environment-related, threshold. Mass selection was practiced on observed production during 40 generations using stochastic simulation. Depending on the heritabilities of the underlying components and on the environment, selection on observed production resulted in a decrease in reproductive rate and in the development of environmental sensitivity when resource intake becomes limiting. Correlations of observed production with underlying components changed across generations, following a nonlinear pattern. The proposed model is simple, but increases the understanding of underlying mechanisms and consequences of selection for production when resources are limiting. Introduction In animal production, negative correlations are observed between production and fitness-related traits, such as fertility and health (Rauw et al., 1998). In lactating animals, poor BCS during the period of negative energy balance results in decreased fertility (e.g., Pryce et al., 2000). In poultry, long-term selection for increased growth resulted in decreased fertility (Marks, 1996; Nestor et al., 1996). It seems that energy allocated to production cannot be applied to other body functions, resulting in increased health and fertility problems (Collard et al., 2000). Apart from the balance between production and health and fertility, the selection environment influences animal performance. For example, the effect of the negative energy balance has been partly compensated for by improving the environment. However, despite these actions, negatively correlated responses to increased production are becoming stronger: environmental sensitivity increases and is especially expressed in decreased fertility. Animals tend to adapt to the environment they are selected in, which may result in the development of a genotype × environment (G × E) interaction. Results presented in the literature suggest that the size of difference between environments especially determines whether a G × E interaction will develop (e.g., Emanuelson et al., 1999; Cameron et al., 2000, 2003). It is clear that selection for production may lead to problems in health and fertility, and that under some circumstances, G × E may develop, but there is still little understanding of the mechanism behind these matters. This paper presents a model to describe the long-term consequences of artificial selection on observed production, based on the allocation of recourses. The aim of this article is to increase understanding of the mechanism behind the development of environmental sensitivity and to gain insight into the correlated responses to selection for observed production. Materials and Methods Model A model was developed after De Jong and Van Noordwijk. (1992) describing the allocation of resources (to a production trait and fitness of an animal (Van der Waaij et al., 2002). Note that all traits are expressed in energy units. In this study,“resources” was defined as“potential feed intake.” It was assumed that sufficient resources were available at all times.“Production” is defined as the potential production level (Pp) of an animal that can be realized under optimal metabolic circumstances (i.e., when resource intake enables full expression of production and creates no limitations to fitness).“Fitness” was defined as a combination of maintenance, health, and reproduction. A proportion of resources was assigned to fitness, rather than an absolute amount, for it was assumed that increased production would lead to proportionally increased resource demand for fitness. Resource (R) allocation is controlled by a factor“c,” so that resources for fitness (Rf) and for production (Rp) are as follows: \begin{eqnarray*}&&Rp\ =\ \mathit{c}\ {\times}\ R\\&&Rf\ =\ (1\ {-}\ \mathit{c})\ {\times}\ R\end{eqnarray*} so that R = Rf + Rp. It was assumed that c, Pp, and R were uncorrelated. The observed production (Po) was defined as the production level that is expressed. When Pp ≤ Rp, or when the demand for resources for expression of Pp is smaller than the amount of resources available for production, Po will be equal to Pp, and resource intake is assumed to be expressed accordingly, so not to its full potential. In reality, the extra resource intake potential most likely would be used for fitness, if required, or for building up body reserves, although this is not taken into account here. When Pp > Rp, or the demand is larger than the genetic potential for resource intake, Po will be equal to Rp, and the animal is in metabolic stress. It was assumed that, in contrast to natural selection, production gets first, but not unlimited, priority. When selection is on Po and resource intake becomes limited, selection pressure, consequently, is shifted toward resource intake and allocation of proportionally more resources for production and away from fitness. An insufficient proportion of resources allocated to fitness may result in decreased health, fertility, and energy available for maintenance, with consequences for reproduction rate and probability of survival. It was assumed that in a good environment, a smaller proportion of resources is required for expression of full fitness potential (fp) than in a poor environment (e.g., temperature difference, disease pressure). Fitness potential was defined as the probability of survival and reproduction. It was assumed that there are two thresholds for (1 − c), with defined values depending on the environment. The lower (L) threshold represents the value for (1 − c) below which the animal will not survive (fp = 0, and consequently Po = 0). The upper (U) threshold represents the value for (1 − c) above which animals survive and show full reproduction potential (fp = 1). In between both thresholds, the probability of survival and reproduction was reduced linearly going from U to L: \[fp\ =\ [(1\ {-}\ \mathit{c})\ {-}\ L]/(U\ {-}\ L)\] For each animal in a defined environment (male and female), fp was determined based on its value for (1 − c). It was assumed that reduced reproduction rate only occurred in females, but reduced survival probability occurred in both sexes. Consequences for survival and reproduction were determined for each animal individually. For survival, if the animal had a value for (1 − c) that was between both thresholds, a random number was drawn from a uniform distribution (0, 1) for each animal that had a value for (1 − c) that was between both thresholds. The animal survived when the random number was smaller than the animal's value for fp. Animals with values for (1 − c) above U survived, and those below L died. For reproduction, each dam had a potential maximal number of offspring. For dams with value for (1 − c) above U, all offspring were born. For dams with a value for (1 − c) that was between thresholds, a random number was drawn from a uniform distribution (0, 1) for each potential offspring separately. Each individual offspring was born when the accompanying random number was smaller than fp. As discrete generations were assumed, decreased reproduction rate can be regarded as either decreased litter size (e.g., due to poor quality oocytes) but with only one reproduction cycle per dam, or multiple reproduction cycles of a single potential offspring each, but with decreased success rate. Females with a value for (1 − c) below L had already died and thus were never selected. In this study, the same thresholds were applied to define survival probability and reproduction probability. In reality, these thresholds may be different, depending on the priority given to reproduction. To summarize, Pp, R, and c are causal, partly heritable parameters; U and L are set by the environment; and fitness (represented by survival and female reproductive performance) and Po rates are resulting phenotypic parameters. This model allows for the development over time of negative relationships between fitness and traits under artificial selection that are counteracted by natural selection. The size of the conflict between fitness and production will depend on the contribution of R and c to the genetic variability in Po; large variance due to c will increase conflicts, whereas large variance due to R will avoid conflicts. Population and Parameters An initial population was simulated with 240 males and 240 females. In total, 40 discrete generations were simulated (200 replicates), and in each generation, the aim was to select 24 males and 120 females on their own performance for observed production (mass selection). No adjustment was made to the maximal number of offspring per female or to the maximal number of females selected, even if the desired number of selected animals could not be met, resulting in a fluctuating population size. Each dam had a maximum of four offspring. The average number of offspring per dam was calculated across all dams selected. Each offspring born had a probability of 0.5 to be male, otherwise it was female. Mean and phenotypic SD were 0.2 and 0.05 for Pp, 1.0 and 0.1 for R, and 0.5 and 0.05 for c, respectively. The heritabilities for Pp and R were set to 0.3, and for c, heritability was 0.1, 0.3, or 0.5. Changing the genetic variance varied heritabilities; the phenotypic variance was assumed to remain constant. The initial means and variances were chosen such that a proportion of animals was not under metabolic stress during the first generations of selection. Two types of environments were assumed, good or poor, and were indicated by the values of the thresholds. The L threshold was either 0.4, representing a good environment, or 0.5, representing a poor environment. The distance between the L and U thresholds varied by 1 or 3 phenotypic SD (depending on the type of environmental stress), resulting in U = 0.45 (1 SD) or 0.55 (3 SD), representing a good environment, and U = 0.55 (1 SD) or 0.65 (3 SD), representing a poor environment. Animals were fed ad libitum with feed of constant quality in all cases, and it was assumed that energy intake was not negatively affected due to loss of appetite. In other words, each animal was able to express full resource intake potential. Results and Discussion Reproduction Rate Population size and average number of progeny per selected dam were approximately constant across generations in the good environment and very similar for both heritabilities for c. The average number of offspring of selected dams was 3.9 across generations and population size remained larger than 450 animals at all times for \(h^{2}_{c}\ =\ 0.1\) . For \(h^{2}_{c}\ =\ 0.5\) , the average number of offspring decreased from 3.9 during the first 24 generations to 3.7 in generation 40 and population size decreased from around 470 to around 450 in generation 40. In the poor environment for \(h^{2}_{c}\ =\ 0.1\) , the initial average number of offspring per selected dam was lower (3.2) and population size was low (177). Consequently, the number of selection candidates was much smaller than intended (only 71 in the first generation), and the number of dams available only became equal to the target of 120 in generation 17. For \(h^{2}_{c}\ =\ 0.5\) , initial average number of offspring was 3.3, increasing to 3.9 in generation 11 and decreasing again to 3.7 in generation 40. The number of dams available was insufficient during the first three generations and population size increased from 199 in the initial generation to 450 in generation four, after which there was little variation. Each dam got four chances to produce offspring, often resulting in some dams not having any offspring, whereas many had four. Therefore, a strong natural selection (i.e., correlated response) for fertility rate occurred, in both the poor and good environments. Overall, there seems to be a constant, suboptimal reproduction rate of 3.7 that develops in both environments, given the current selection strategy. This can be explained by the long-term genetic contributions theory (Wray and Thompson, 1990). Animals with a high genetic potential for reproduction rate that are selected for breeding (based on observed production) have a higher potential of passing on their genes to the next generation. Also, those offspring are more likely to reproduce than are animals whose dams had a lower reproduction rate. Finally, a larger number of offspring will result in a larger chance that one of them will be selected to produce the next generation. These factors together are probably the underlying reason for the relatively high average number of offspring in later generations. In practical breeding, the best producing dams that are less reproductive often are offered more chances to reproduce, resulting in a decreased, or even negative, natural selection pressure on reproduction. Selection Response Figure 1 shows the results of 40 generations of selection on observed production, where the heritability for c is either 0.1 or 0.5 (results for \(h^{2}_{c}\ =\ 0.3\) were between these results and are therefore not shown) and the heritability for R is 0.3. Two environments are compared and the distance between thresholds is one SD. The starting point was an unselected population. Observed production in the figure is determined as the average production for all animals in the generation, including those with production equal to zero. Therefore, in the poor environment, Po initially is much lower than Pp due to insufficient values for (1 − c). Selection for increased observed production thus results in selection pressure on (1 − c). Selection pressure on (1 − c) remains until this is no longer a limiting factor. This will occur when (1 − c) has values around the upper threshold. Due to chance, there will always be selected parents with values for (1 − c) below the U threshold that still manage to have high Po and be reproductive. Also, parents with values for (1 − c) above the U threshold may have offspring with values below. Consequently, the population mean for (1 − c) in the final generations in Figure 1 remains below the U threshold. Figure 1. View largeDownload slide Consequences for fitness and production when selection is on observed production. Means are shown for resource intake (triangle), production potential (square), observed production (open diamond), resource allocation factor (×), and survival probability (circle), following selection in a good (a and c) or a poor (b and d) environment, and for a heritability for resource allocation factor (c) of 0.1 (a and b) or 0.5 (c and d). Heritabilities for production potential and resource intake potential were 0.3. Figure 1. View largeDownload slide Consequences for fitness and production when selection is on observed production. Means are shown for resource intake (triangle), production potential (square), observed production (open diamond), resource allocation factor (×), and survival probability (circle), following selection in a good (a and c) or a poor (b and d) environment, and for a heritability for resource allocation factor (c) of 0.1 (a and b) or 0.5 (c and d). Heritabilities for production potential and resource intake potential were 0.3. After a number of generations, resource intake potential becomes limiting in both environments and selection for observed production results in increased resource intake potential. The number of generations until resource intake potential becomes limiting depends on the heritability for c and on the environment, although environment is the most important factor. A low heritability for c may be positive in the good environment because the fitness potential remains higher and genetic gain in production is mainly caused by a correlated response in resource intake rather than in c. However, in the poor environment, the low heritability for c also results in a slow increase in fitness potential. Phenotypic Correlations Figure 2 shows the phenotypic correlations between Po and Pp, resource intake potential, and the resource allocation factor. For the correlation with survival probability, values for Po were defined as if all animals would have been able to produce, regardless of survival. All results come from the same 200 replicates as in Figure 1. Because Pp, resource intake potential, and resource allocation factor were assumed to be genetically uncorrelated and the environmental variance was assumed to be constant, the phenotypic correlations between these traits were zero by definition (and in the simulation results). In the good environment, the correlation between observed and potential production is initially high, but decreases across generations to approximately 0.10. When resource intake becomes limiting (i.e., under selection pressure), the correlation between resource intake potential and Po increases to approximately 0.10. Simultaneously, the correlation with c increases. In the poor environment, Po is the factor most strongly correlated to resource allocation (0.6 to 0.75 for \(h^{2}_{c}\ =\ 0.1\) , and 0.42 to 0.75 for \(h^{2}_{c}\ =\ 0.5\) ). Selection immediately increases pressure on resources for fitness, therefore metabolic stress, and thus the development of environmental sensitivity, occurs at an earlier generation in the good environment. Figure 2. View largeDownload slide Consequences for correlations with observed production. Correlations are shown for observed production with production potential (closed diamond), resource intake (open diamond), resource allocation factor (closed circle), and survival probability (closed triangle), following selection in a good (a and c) or a poor (b and d) environment, and for a heritability for resource allocation factor (c) of 0.1 (a and b) or 0.5 (c and d). Heritabilities for production potential and resource intake potential were 0.3. Figure 2. View largeDownload slide Consequences for correlations with observed production. Correlations are shown for observed production with production potential (closed diamond), resource intake (open diamond), resource allocation factor (closed circle), and survival probability (closed triangle), following selection in a good (a and c) or a poor (b and d) environment, and for a heritability for resource allocation factor (c) of 0.1 (a and b) or 0.5 (c and d). Heritabilities for production potential and resource intake potential were 0.3. Beilharz et al. (1993) and Knap and Bishop (2000) argue that when resources become limiting, a negative correlation between production traits and fitness-related traits will result. Results in this study indicate that environmental sensitivity, indicated by the negative correlation between Po and survival probability, develops as soon as there is metabolic stress. This finding is in agreement with Kolmodin et al. (2002), who found a similar trend in environmental sensitivity in Scandinavian dairy cattle. Apparently, the animals with the highest Po (trait under selection) tend to be the animals with poorer values for c, and thus are those with increased environmental sensitivity. Switched Environments Long-term selection for production in one environment will lead to increased production in that particular environment. However, it does not necessarily mean that those improved animals would perform as well in a second environment with different production circumstances. The main reason is a difference in natural selection pressure on traits underlying observed production in both environments (Van der Waaij et al., 2000). Thorpe and Luiting (2000) argue that selection in a poor environment will result in more robust animals. Rauw et al. (1998) have summarized situations in practical pig, poultry, and dairy cattle breeding where this may be the case. Kolmodin et al. (2003) have shown that selection for high phenotypic values does result in increased environmental sensitivity in dairy cattle, especially when the environment is improved during the generations of selection. To validate whether the model presented here would also predict this, animals were selected in one environment for 10 or 40 generations, transferred to the opposite environment, and subsequently selected to produce generation 11 or 41. Results in Table 1 indicate that the difference in the quality of the environments and the number of generations of selection in those environments are important factors determining the presence and size of environmental sensitivity and, consequently, the development of a G × E interaction. The heritability for c, representing the potential selection response in c, has a larger influence in the short term (10 generations) than in the long term (40 generations). Resource intake potential and resource allocation factor do not change going from one environment to the other. However, Po, survival probability, and reproduction probability do, mainly because of a change in the importance of (1 − c). Table 1. Consequences of changing environment following 10 or 40 generations of mass selection for average observed production in either a good or a poor environment when the heritability for c is either 0.1 or 0.5, the heritability for resource intake and production potential is 0.3, and the distance between thresholds is 1 SD Responsesa \(h^{2}_{c}\) Changeb Genc Po R c surv rpo,pp rpo,R rpo,c rpo,surv noff 10 generations 0.1 P→G 10 0.13 1.00 0.53 0.52 0.21 0.00 0.74 0.011 3.5 11 0.26 1.00 0.53 0.98 0.79 0.01 0.21 0.009 3.5 G→P 10 0.31 1.01 0.51 0.94 0.51 0.02 0.38 −0.006 3.9 11 0.13 1.01 0.51 0.38 0.09 0.00 0.73 −0.024 3.9 0.5 P→G 10 0.28 1.00 0.59 0.90 0.40 0.04 0.46 −0.03 3.9 11 0.32 1.01 0.59 1.00 0.92 0.11 −0.12 −0.007 3.9 G→P 10 0.32 1.00 0.53 0.98 0.68 0.03 0.22 −0.003 4.0 11 0.18 1.01 0.53 0.52 0.13 0.02 0.74 −0.043 4.0 40 generations 0.1 P→G 40 0.39 1.24 0.57 0.78 0.07 0.09 0.58 −0.37 3.7 41 0.51 1.26 0.57 0.99 0.42 0.35 −0.44 −0.06 3.7 G→P 40 0.56 1.33 0.49 0.87 0.09 0.10 0.47 −0.27 3.7 41 0.13 1.35 0.49 0.22 0.00 0.04 0.66 −0.58 3.7 0.5 P→G 40 0.43 1.36 0.56 0.78 0.06 0.08 0.59 −0.38 3.8 41 0.56 1.37 0.56 1.00 0.41 0.36 −0.49 −0.05 3.8 G→P 40 0.53 1.32 0.47 0.81 0.07 0.10 0.58 −0.28 3.7 41 0.08 1.33 0.47 0.13 0.00 0.03 0.59 −0.50 3.7 Responsesa \(h^{2}_{c}\) Changeb Genc Po R c surv rpo,pp rpo,R rpo,c rpo,surv noff 10 generations 0.1 P→G 10 0.13 1.00 0.53 0.52 0.21 0.00 0.74 0.011 3.5 11 0.26 1.00 0.53 0.98 0.79 0.01 0.21 0.009 3.5 G→P 10 0.31 1.01 0.51 0.94 0.51 0.02 0.38 −0.006 3.9 11 0.13 1.01 0.51 0.38 0.09 0.00 0.73 −0.024 3.9 0.5 P→G 10 0.28 1.00 0.59 0.90 0.40 0.04 0.46 −0.03 3.9 11 0.32 1.01 0.59 1.00 0.92 0.11 −0.12 −0.007 3.9 G→P 10 0.32 1.00 0.53 0.98 0.68 0.03 0.22 −0.003 4.0 11 0.18 1.01 0.53 0.52 0.13 0.02 0.74 −0.043 4.0 40 generations 0.1 P→G 40 0.39 1.24 0.57 0.78 0.07 0.09 0.58 −0.37 3.7 41 0.51 1.26 0.57 0.99 0.42 0.35 −0.44 −0.06 3.7 G→P 40 0.56 1.33 0.49 0.87 0.09 0.10 0.47 −0.27 3.7 41 0.13 1.35 0.49 0.22 0.00 0.04 0.66 −0.58 3.7 0.5 P→G 40 0.43 1.36 0.56 0.78 0.06 0.08 0.59 −0.38 3.8 41 0.56 1.37 0.56 1.00 0.41 0.36 −0.49 −0.05 3.8 G→P 40 0.53 1.32 0.47 0.81 0.07 0.10 0.58 −0.28 3.7 41 0.08 1.33 0.47 0.13 0.00 0.03 0.59 −0.50 3.7 a Po = observed production; R = resource intake potential; c = resource allocation factor; fp = survival probability; rpo,pp = correlation between Po and Pp; rpo,R = correlation between Po and R; rpo,c = correlation between Po and c; rpo,surv = correlation between Po and survival probability; and noff = average number of offspring per selected dam. b Change = transfer from one environment to the other followed by one round of selection, G = good environment, and P = poor environment. c Gen = generation. View Large Table 1. Consequences of changing environment following 10 or 40 generations of mass selection for average observed production in either a good or a poor environment when the heritability for c is either 0.1 or 0.5, the heritability for resource intake and production potential is 0.3, and the distance between thresholds is 1 SD Responsesa \(h^{2}_{c}\) Changeb Genc Po R c surv rpo,pp rpo,R rpo,c rpo,surv noff 10 generations 0.1 P→G 10 0.13 1.00 0.53 0.52 0.21 0.00 0.74 0.011 3.5 11 0.26 1.00 0.53 0.98 0.79 0.01 0.21 0.009 3.5 G→P 10 0.31 1.01 0.51 0.94 0.51 0.02 0.38 −0.006 3.9 11 0.13 1.01 0.51 0.38 0.09 0.00 0.73 −0.024 3.9 0.5 P→G 10 0.28 1.00 0.59 0.90 0.40 0.04 0.46 −0.03 3.9 11 0.32 1.01 0.59 1.00 0.92 0.11 −0.12 −0.007 3.9 G→P 10 0.32 1.00 0.53 0.98 0.68 0.03 0.22 −0.003 4.0 11 0.18 1.01 0.53 0.52 0.13 0.02 0.74 −0.043 4.0 40 generations 0.1 P→G 40 0.39 1.24 0.57 0.78 0.07 0.09 0.58 −0.37 3.7 41 0.51 1.26 0.57 0.99 0.42 0.35 −0.44 −0.06 3.7 G→P 40 0.56 1.33 0.49 0.87 0.09 0.10 0.47 −0.27 3.7 41 0.13 1.35 0.49 0.22 0.00 0.04 0.66 −0.58 3.7 0.5 P→G 40 0.43 1.36 0.56 0.78 0.06 0.08 0.59 −0.38 3.8 41 0.56 1.37 0.56 1.00 0.41 0.36 −0.49 −0.05 3.8 G→P 40 0.53 1.32 0.47 0.81 0.07 0.10 0.58 −0.28 3.7 41 0.08 1.33 0.47 0.13 0.00 0.03 0.59 −0.50 3.7 Responsesa \(h^{2}_{c}\) Changeb Genc Po R c surv rpo,pp rpo,R rpo,c rpo,surv noff 10 generations 0.1 P→G 10 0.13 1.00 0.53 0.52 0.21 0.00 0.74 0.011 3.5 11 0.26 1.00 0.53 0.98 0.79 0.01 0.21 0.009 3.5 G→P 10 0.31 1.01 0.51 0.94 0.51 0.02 0.38 −0.006 3.9 11 0.13 1.01 0.51 0.38 0.09 0.00 0.73 −0.024 3.9 0.5 P→G 10 0.28 1.00 0.59 0.90 0.40 0.04 0.46 −0.03 3.9 11 0.32 1.01 0.59 1.00 0.92 0.11 −0.12 −0.007 3.9 G→P 10 0.32 1.00 0.53 0.98 0.68 0.03 0.22 −0.003 4.0 11 0.18 1.01 0.53 0.52 0.13 0.02 0.74 −0.043 4.0 40 generations 0.1 P→G 40 0.39 1.24 0.57 0.78 0.07 0.09 0.58 −0.37 3.7 41 0.51 1.26 0.57 0.99 0.42 0.35 −0.44 −0.06 3.7 G→P 40 0.56 1.33 0.49 0.87 0.09 0.10 0.47 −0.27 3.7 41 0.13 1.35 0.49 0.22 0.00 0.04 0.66 −0.58 3.7 0.5 P→G 40 0.43 1.36 0.56 0.78 0.06 0.08 0.59 −0.38 3.8 41 0.56 1.37 0.56 1.00 0.41 0.36 −0.49 −0.05 3.8 G→P 40 0.53 1.32 0.47 0.81 0.07 0.10 0.58 −0.28 3.7 41 0.08 1.33 0.47 0.13 0.00 0.03 0.59 −0.50 3.7 a Po = observed production; R = resource intake potential; c = resource allocation factor; fp = survival probability; rpo,pp = correlation between Po and Pp; rpo,R = correlation between Po and R; rpo,c = correlation between Po and c; rpo,surv = correlation between Po and survival probability; and noff = average number of offspring per selected dam. b Change = transfer from one environment to the other followed by one round of selection, G = good environment, and P = poor environment. c Gen = generation. View Large The correlation between Po and Pp changed considerably going from one environment to the other, the change being largest for \(h^{2}_{c}\ =\ 0.1\) , going from the poor to the good environment after 10 generations of selection (0.21 to 0.79). In all cases, the correlation was lowest in the poor environment and never equal to 1 in the good environment, indicating that observed production is dependent not only on Pp, especially under harsher conditions. The correlation with resource intake generally was low, and only moderately high when transferred from a poor to a good environment after 40 generations of selection. The correlation with the resource allocation factor changes considerably going from one environment to the other, the most extreme change occurring when going from a poor to a good environment after 40 generations of selection and \(h^{2}_{c}\ =\ 0.5\) ; this results in a correlation change from 0.59 to −0.49. Going from a poor to a good environment, the resource intake factor, determining how much energy is assigned to fitness-related traits, can be decreased without negative consequences for reproduction probability, and thus for Po. Transferring from a good to a poor environment causes a decrease in the correlation between Po and survival probability, especially after 40 generations of selection. Kolmodin et al. (2002) argue that environmental sensitivity develops following phenotypic selection in the presence of G × E, but the present results show that environmental sensitivity actually is the reason for the presence of G × E. Given the results in Figure 2 and Table 1, it can thus be concluded that the size of G × E depends on the number of generations of selection in separate environments. In relation to this, it is interesting to see that the (size of) change in phenotypic correlations between Po and c, resource intake, Pp, and survival probability has an important influence due to the environment, rather than to the underlying genetic parameters of c, resource intake potential, and Pp. Long-term selection in the same environment, as, for example, in the case of indigenous breeds, causes some changes in the relationships as described above. Selection in the good environment usually results in a higher Po than when selection is in the poor environment and animals are transferred to a good environment after a number of generations. However, when \(h^{2}_{c}\ =\ 0.5\) and after 40 generations of selection in the poor environment, Po was higher after transfer to the good environment than after 40 generations of selection in the good environment. It should be noted that resource intake also is higher, with economic consequences. Also, in the present study, it was assumed that resources were never limited and were of equal quality in both environments. This may be valid when differences between environments are due to, for example, climate conditions in the housing system or to a difference in infection pressure. However, when the quality and/or quantity of resources in the poor environment is lower than that in the good environment, a maximum is set to resource intake. Most likely, there will be a point in time where Po no longer increases as a consequence of limiting resources, and reproduction rate may move to a lower equilibrium than in case of ad libitum feeding. Varying Type of Environmental Stress Differences in type of environmental stress were mimicked by varying the distance between thresholds (move the U threshold). The larger the distance, the more gradual the influence of a decrease in (1 − c) will be. For \(h^{2}_{c}\ =\ 0.1\) , a distance between thresholds of 2 or 3 SD, compared with the 1 SD as discussed above, had drastic consequences. Especially in the case of 3 SD, the reproductive rate in the poor environment decreased too much and the average number of offspring per selected dam did not exceed two in any generation with a survival rate of approximately 17% across generations. The survival rate decreased a bit in the good environment as well when the distance between thresholds was increased from 1 to 3 SD, but the average number of offspring was 3.3. After 40 generations, the average Po was 0.43 and the survival and reproduction probability 0.80. The average value for (1 − c) was 0.53, and for feed intake potential, it was 1.24, instead of 0.49 and 1.33, respectively. For \(h^{2}_{c}\ =\ 0.5\) , the situation was slightly different. In the poor environment, the average number of offspring again was below two at first, but increased across generations, and from generation 19 onward, the population size started to rapidly increase. By then, (1 − c) had increased from 0.51 to 0.64. In generation 40, average Po was 0.32, (1 − c) was 0.62, resource intake capacity was 1.24, and survival and reproduction probability was 73%. In the good environment, the average number of offspring again was below two at first, but quickly increased to above two in generation 3. In generation 40, average Po reached 0.44, (1 − c) was 0.52, resource intake potential was 1.30, and survival and reproduction probability was 77%. General Discussion In this article, a model is proposed to describe the consequences of selection on Po for allocation of resources to production and fitness-related traits. By assuming simple additive relationships between the underlying components, it is shown that nonlinear relationships develop among resulting components. Depending on the heritabilities of the underlying components and on the environment, it is shown that selection on Po eventually results in a decrease in survival and reproductive rate and in the development of environmental sensitivity. Even though the proposed model is very simple, it helps increase the understanding of underlying mechanisms and consequences of selection for production when resources are limiting. The consequences of long-term selection for a production trait have been described in the literature. In poultry (Japanese quail and turkey), long-term selection for increased premature BW resulted in increased adult BW, feed intake, and feed efficiency (Marks, 1996), egg weight, and eating bouts (Nestor et al., 1996), but also in decreased hatchability and egg production (Marks, 1996), semen production, ability to walk, and resistance to infection (Nestor et al., 1996). Selection for higher milk yields resulted in increased feed intake and decreased energy balance in cattle, sheep, and pigs (reviewed by Veerkamp, 2002). In dairy cattle, where selection has been predominantly on milk production for the past decades, fertility has become a limiting factor (Pryce et al., 2000; Royal et al., 2002). The genetic correlations between various fertility-related traits and BCS, milk, fat and protein in Holstein cattle are moderately to strongly negative (e.g., Pryce et al., 2001, 2002; Royal et al., 2002), and increase unfavorably under heat stress (López-Gatius, 2003). Health is also influenced by a shift in the allocation of resources, as can be illustrated by ascites in broilers. The correlation between BW and ascites-related traits under cold conditions are moderate to high (Pakdel et al., 2002). These correlations are different for animals under cold vs. normal climate conditions (De Greef et al., 2001). Also, the resistance to infection with Pasteurella multocida and Newcastle virus decreased following generations of selection for increased premature BW in turkeys (Nestor et al., 1996). On the other hand, the metabolic balance may also shift toward increased fitness as a result of selection, as can be illustrated by chickens that have been under long-term selection for increased immune response. These animals are substantially smaller than control animals (Parmentier et al., 1996). Animals tend to adapt to their environment. For example, selection in dairy cattle in more temperate regions will decrease heat tolerance in the warmer regions (Ravagnolo and Misztal, 2002), fast-growing broilers show less heat tolerance than slower growing ones (Yalçin et al., 2001), and the indigenous Red Maasai sheep is much more productive than the Dorper in the humid coastal environment of east Africa, whereas the hybrid Dorper is only a little more productive in the semi-arid environment (Baker, et al., 2002). Results like these show that G × E interactions develop, especially when the difference between environments is substantial. The direction of selection response obtained from the simulation seems in agreement with selection results as presented in the literature; however, we were not able to find any results in the literature on the changing correlations between Po and underlying traits across generations. Validation of the model, therefore, is difficult at present, though could be done by setting up a selection experiment, for example, in mice, which are less expensive. Several papers have shown that selection for increased preweaning growth (i.e., milk production of the mother) will have increased BW and feed intake as a correlated response (e.g., Bünger et al., 1998). However, in these studies, offspring were suckled by their own mother, resulting in entanglement of genetic effects of milk production, growth, and litter size The pups would need to be cross fostered in order to eliminate this effect. Selection should be on preweaning growth of the cross-fostered pups, as a representation of milk production. Feed intake should be measured, as well as fertility (time to pregnancy), maternal BW (as an additional indication of negative energy balance), and cumulative pup weight at a number of times (milk production). Genetic parameters could then be estimated and the simulation results could be validated. In the study, it was assumed that c, Pp, and R were genetically uncorrelated. This assumption may or may not be realistic, depending on the situation. The reason for assuming zero correlation is that we do not know what the real correlations would be. Also, the correlations may be different across situations or traits considered (e.g., production traits such as growth, milk production, wool production). By assuming zero correlation between the underlying traits (i.e., c, Pp, and R), results give more insight into the size of the correlations that develop over time between those traits and the resulting traits (i.e., Po, reproduction probability). In a very poor environment, reproduction rate was not always sufficient to cover replacement rates and, as a consequence, population size decreased. Because of natural selection, in later generations, population size often recovered. As a consequence, the selected proportion was not always comparable across environments or the various genetic parameters, resulting in differing selection pressure overall, regardless of the underlying components. A decreasing population size often is not acceptable in practical animal breeding, and management would be adjusted to limit loss in population size. The purpose of this study was to gain understanding of underlying mechanisms and, therefore, results are presented when management is not adjusted, regardless of the results. Selection response in feed intake in the present article occurs as a correlated response to selection for increased observed production. The fact that it is a correlated response with a time lag may explain why the response often is not sufficient to fully express the trait under selection. In cattle breeding, an increased feed intake is not considered a negative consequence, and often is stimulated. In pig and poultry breeding, however, an increased feed intake is not always considered to be advantageous. Therefore, if one fails to allow for increased feed intake, but does select for increased production, energy for fitness will be diminished. The consequences of this selection practice would be increased selection pressure on resource allocation toward the trait with highest selection weight (i.e., production), resulting in an increase in environmental sensitivity. Putting no limits on feed intake does not automatically solve the problem. Because of the time lag, not limiting feed intake is not enough, as presented in this research. By also putting sufficient selection pressure directly on feed intake, selection pressure on c may be relieved. However, it may be difficult to balance selection pressure because resource allocation of individual animals cannot be measured and too large a feed intake would result in animals that grow fat. Selection for an optimal allocation of resources, given a certain environment, would be desirable, even though it is not clear yet how that should be accomplished. Production is a trait that is determined by many underlying components. Selection for production puts the highest selection pressure on the most limiting component. Improvement in that component may reveal a second component as most limiting, which may have, for example, a different heritability. The consequence of this shift in importance of underlying components is that correlations between Po and these components change as well. In this study, we have only discriminated between Pp, resource allocation factor, and feed intake, but in reality there are many more underlying components—all of them, together with the environment, determining the final observed production. One way of determining the size of the effect of underlying components on the trait of interest (i.e., observed production in this case) in practice would be to estimate the correlations between Po and underlying components across generations, provided these underlying components can be measured. A change in the correlations indicates a change in the importance of one of the underlying components. Implications Even though the present model simplifies reality, it gives insight into the consequences of selection for production in a good or poor environment. Selection always puts the greatest selection pressure on the most limiting underlying trait, resulting in a changing correlation between underlying traits across generations. Long-term phenotypic selection on observed production increases environmental sensitivity, which causes problems, especially when the animals are subsequently transferred to a poorer environment. The direction of correlated selection response agrees with literature results on reaction-norm models, but it remains important to validate the model with historical data, both to quantify the elements of the model (e.g., the heritability of the resource allocation factor) and to assess the predictive ability of the model. Literature Cited Baker, R. L., J. M. Mugambi, J. O. Audho, A. B. Carles, and W. Thorpe 2002. Comparison of Red Maasai and Dorper sheep for resistance to gastro-intestinal nematode parasites, productivity and efficiency in a humid and semi-arid environment in Kenya. Proc. 7th World Congr. Genet. Appl. Livest. Prod., Montpellier, France, Communication No. 13–10. Beilharz, R. G., B. G. Luxford, and J. L. Wilkinson 1993. Quantitative genetics and evolution: Is our understanding of genetics sufficient to explain evolution? J. Anim. Br. Gen. 110: 161– 170. Google Scholar CrossRef Search ADS Bünger L., U. Renne, G. Dietl, and S. Kuhla 1998. Long-term selection for protein amount over 70 generations in mice. Genet. Res. 72: 93– 109. Google Scholar CrossRef Search ADS PubMed Butler, W. R. 2003. Energy balance relationships with follicular development, ovulation and fertility in postpartum dairy cows. Livest. Prod. Sci. 83: 211– 218. Google Scholar CrossRef Search ADS Cameron, N. D., M. Enser, G. R. Nute, F. M. Whittington, J. C. Penman, A. C. Fisken, A. M. Perry, and J. D. Wood 2000. Genotype with nutrition interaction on fatty acid composition of intramuscular fat and the relationship with flavor of pig meat. Meat Sci. 55: 187– 195. Google Scholar CrossRef Search ADS PubMed Cameron, N. D., G. B. Garth, J. C. Penman, and A. Fiskin 2003. Sensitivity to dietary lysine: energy content in pigs divergently selected for components of efficient lean growth rate. Anim. Sci. 76: 175– 190. Google Scholar CrossRef Search ADS Collard, B. L., P. J. Boettcher, J. C. M. Dekkers, D. Petitclerc, and L. R. Schaeffer 2000. Relationships between energy balance and health traits of dairy cattle in early lactation. J. Dairy Sci. 83: 2683– 2690. Google Scholar CrossRef Search ADS PubMed De Greef, K. H., L. L. G. Janss, A. L. J. Vereijken, R. Pit, and C. L. M. Gerritsen 2001. Disease induced variability of genetic correlations: ascites in broilers as a case study. J. Anim. Sci. 79: 1723– 1733. Google Scholar CrossRef Search ADS PubMed De Jong, G., and A. J. Van Noordwijk 1992. Acquisition and allocation of resources: Genetic (co)variances, selection, and life histories. Am. Nat. 139: 749– 770. Google Scholar CrossRef Search ADS Emanuelson, U., G. Banos, and J. Philipsson 1999. Interbull Centre Report. Pages 1–6 in Interbull Bulletin 22. Proc. of the 1999 Interbull Mtg., Zürich, Switzerland. Knap, P. W., and S. C. Bishop 2000. Relations between genetic change and infectious disease in domestic livestock. Pages 65–80 in Occas. Pub. Br. Soc. Anim. Sci. No. 27. BSAS, Edinburgh, Scotland. Kolmodin, R., E. Strandberg, H. Jorjani, and B. Danell 2003. Selection in presence of genotype by environment interaction: Response in environmental sensitivity. Anim. Sci. 76: 375– 386. Google Scholar CrossRef Search ADS Kolmodin, R., E. Strandberg, E. Madsen, J. Jensen, and H. Jorjani 2002. Genotype by environment interaction in Nordic dairy cattle studied using reaction norms. Acta Agric. Scand. 52: 11– 24. López-Gatius, F. 2003. Is fertility declining in dairy cattle? A retrospective study in north-eastern Spain. Theriogenology 60: 89– 99. Google Scholar CrossRef Search ADS PubMed Luiting, P., and E. M. Urff 1991. Optimization of a model to estimate residual feed consumption in the laying hen. Livest. Prod. Sci., 27: 321– 338. Google Scholar CrossRef Search ADS Marks, H. L. 1996. Long-term selection for body weight in Japanese quail under different environments. Poult. Sci. 75: 1198– 1203. Google Scholar CrossRef Search ADS PubMed Nestor, K. E., D. O. Noble, N. J. Zhu, and Y. Moritsu 1996. Direct and correlated responses to long-term selection for increased body weight and egg production in turkeys. Poult. Sci. 75: 1180– 1191. Google Scholar CrossRef Search ADS PubMed Pakdel, A., J. A. M. Van Arendonk, A. L. J. Vereijken, and H. Bovenhuis 2002. Genetic and phenotypic correlations for ascites related traits in broilers. Communication No. 13–16. Proc. 7th World Congr. Genet. Appl. Livest. Prod., Montpellier, France. Parmentier, H. K., M. G. B. Nieuwland, E. Rijke, G. De Vries Reilingh, and J. W. Schrama 1996. Divergent antibody responses to vaccines and divergent body weight of chicken lines selected for high and low humoral responsiveness to sheep red blood cells. Avian Dis. 40: 634– 644. Google Scholar CrossRef Search ADS PubMed Pryce, J. E., M. P. Coffey, and S. Brotherstone 2000. The genetic relationship between calving interval, body condition score and linear type and management traits in registered Holsteins. J. Dairy Sci. 83: 2664– 2671. Google Scholar CrossRef Search ADS PubMed Pryce, J. E., M. P. Coffey, S. H. Brotherstone, and J. A. Wooliams 2002. Genetic relationships between calving interval and body condition score conditional on milk yield. J. Dairy Sci. 85: 1590– 1595. Google Scholar CrossRef Search ADS PubMed Pryce, J. E., M. P. Coffey, and G. Simm 2001. The relationship between body condition score and reproductive performance. J. Dairy Sci. 84: 1508– 1515. Google Scholar CrossRef Search ADS PubMed Rauw, W. M., E. Kanis, E. N. Noordhuizen-Stassen, and F. J. Grommers 1998. Undesirable side effects of selection for high production efficiency in farm animals: A review. Livest. Prod. Sci. 56: 15– 33. Google Scholar CrossRef Search ADS Ravagnolo, O., and I. Misztal 2002. Genetic component of heat stress in dairy cattle, parameter estimation. J. Dairy Sci., 83: 2126– 2130. Google Scholar CrossRef Search ADS Royal, M. D., J. E. Pryce, J. A. Woolliams, and A. P. F. Flint 2002. The genetic relationship between commencement of luteal activity and calving interval, body condition score, production, and linear type traits in Holstein-Friesian dairy cattle. J. Dairy Sci. 85: 3071– 3080. Google Scholar CrossRef Search ADS PubMed Thorpe, B. H., and E. Luiting 2000. Breeding for resistance to production diseases in poultry. Pages 357–377 in Breeding for Disease Resistance in Farm Animals. 2nd ed. R. F. E. Axford, S. C. Bishop, F. W. Nicholas, and J. B. Owen ed. CAB Int., Oxon, U.K. Van der Waaij, E. H., P. Bijma, S. C. Bishop, and J. A. M. Van Arendonk 2000. Modeling selection for production traits under selection under constant infection pressure. J. Anim. Sci. 78: 2809– 2820. Google Scholar CrossRef Search ADS PubMed Van der Waaij, E. H., L. L. G. Janss, and P. Bijma 2002. A resource allocation model describing consequences of artificial selection under metabolic stress. Communication No. 23–11. Proc. 7th World Congr. Genet. Appl. Livest. Prod., Montpellier, France. Veerkamp, R. F. 2002. Feed intake and energy balance in lactating animals. Proc. 7th World Congr. Genet. Appl. Livest. Prod., Montpellier, France, Communication No. 10–01. Wray, N. R., and R. Thompson 1990. Prediction of rates of inbreeding in selected populations. Genet. Res. Camb. 55: 41– 54. Google Scholar CrossRef Search ADS Yalçin, S., S. Özkan, T. Türknut, and P. B. Siegel 2001. Responses to heat stress in commercial and local broiler stocks. Br. Poult. Sci. 42: 149– 152. Google Scholar CrossRef Search ADS PubMed Copyright 2004 Journal of Animal Science
A resource allocation model describing consequences of artificial selection under metabolic stressvan der Waaij, E. H.
doi: 10.1093/ansci/82.4.973pmid: N/A
Abstract Long-term selection on production results in increased environmental sensitivity. This often is expressed through decreased fertility and increased health problems. The phenomenon has been described in all common farm animal species. One theory is that potential resource intake is insufficient to express production potential. Additional resources are drawn away from fitness-related traits, such as fertility and health, to further increase observed production. In addition, resources for maintaining fitness depend on the demands by the environment. In a harsh environment, more resources are required for fitness-related traits than in an optimal environment. Literature results show that selection in an optimal environment will increase sensitivity to less optimal environments. The objectives of this paper were to increase understanding of the underlying mechanism behind the development of environmental sensitivity and to gain insight into correlated response(s) when selection is on observed production. A resource allocation model was defined where observed production depended on production potential, resource intake potential, and the allocation of resources to production or fitness, including maintenance, health, and reproduction. Penalties for reproductive performance and probability of survival were included when the proportion of resources assigned to fitness dropped below a certain, environment-related, threshold. Mass selection was practiced on observed production during 40 generations using stochastic simulation. Depending on the heritabilities of the underlying components and on the environment, selection on observed production resulted in a decrease in reproductive rate and in the development of environmental sensitivity when resource intake becomes limiting. Correlations of observed production with underlying components changed across generations, following a nonlinear pattern. The proposed model is simple, but increases the understanding of underlying mechanisms and consequences of selection for production when resources are limiting. Introduction In animal production, negative correlations are observed between production and fitness-related traits, such as fertility and health (Rauw et al., 1998). In lactating animals, poor BCS during the period of negative energy balance results in decreased fertility (e.g., Pryce et al., 2000). In poultry, long-term selection for increased growth resulted in decreased fertility (Marks, 1996; Nestor et al., 1996). It seems that energy allocated to production cannot be applied to other body functions, resulting in increased health and fertility problems (Collard et al., 2000). Apart from the balance between production and health and fertility, the selection environment influences animal performance. For example, the effect of the negative energy balance has been partly compensated for by improving the environment. However, despite these actions, negatively correlated responses to increased production are becoming stronger: environmental sensitivity increases and is especially expressed in decreased fertility. Animals tend to adapt to the environment they are selected in, which may result in the development of a genotype × environment (G × E) interaction. Results presented in the literature suggest that the size of difference between environments especially determines whether a G × E interaction will develop (e.g., Emanuelson et al., 1999; Cameron et al., 2000, 2003). It is clear that selection for production may lead to problems in health and fertility, and that under some circumstances, G × E may develop, but there is still little understanding of the mechanism behind these matters. This paper presents a model to describe the long-term consequences of artificial selection on observed production, based on the allocation of recourses. The aim of this article is to increase understanding of the mechanism behind the development of environmental sensitivity and to gain insight into the correlated responses to selection for observed production. Materials and Methods Model A model was developed after De Jong and Van Noordwijk. (1992) describing the allocation of resources (to a production trait and fitness of an animal (Van der Waaij et al., 2002). Note that all traits are expressed in energy units. In this study,“resources” was defined as“potential feed intake.” It was assumed that sufficient resources were available at all times.“Production” is defined as the potential production level (Pp) of an animal that can be realized under optimal metabolic circumstances (i.e., when resource intake enables full expression of production and creates no limitations to fitness).“Fitness” was defined as a combination of maintenance, health, and reproduction. A proportion of resources was assigned to fitness, rather than an absolute amount, for it was assumed that increased production would lead to proportionally increased resource demand for fitness. Resource (R) allocation is controlled by a factor“c,” so that resources for fitness (Rf) and for production (Rp) are as follows: \begin{eqnarray*}&&Rp\ =\ \mathit{c}\ {\times}\ R\\&&Rf\ =\ (1\ {-}\ \mathit{c})\ {\times}\ R\end{eqnarray*} so that R = Rf + Rp. It was assumed that c, Pp, and R were uncorrelated. The observed production (Po) was defined as the production level that is expressed. When Pp ≤ Rp, or when the demand for resources for expression of Pp is smaller than the amount of resources available for production, Po will be equal to Pp, and resource intake is assumed to be expressed accordingly, so not to its full potential. In reality, the extra resource intake potential most likely would be used for fitness, if required, or for building up body reserves, although this is not taken into account here. When Pp > Rp, or the demand is larger than the genetic potential for resource intake, Po will be equal to Rp, and the animal is in metabolic stress. It was assumed that, in contrast to natural selection, production gets first, but not unlimited, priority. When selection is on Po and resource intake becomes limited, selection pressure, consequently, is shifted toward resource intake and allocation of proportionally more resources for production and away from fitness. An insufficient proportion of resources allocated to fitness may result in decreased health, fertility, and energy available for maintenance, with consequences for reproduction rate and probability of survival. It was assumed that in a good environment, a smaller proportion of resources is required for expression of full fitness potential (fp) than in a poor environment (e.g., temperature difference, disease pressure). Fitness potential was defined as the probability of survival and reproduction. It was assumed that there are two thresholds for (1 − c), with defined values depending on the environment. The lower (L) threshold represents the value for (1 − c) below which the animal will not survive (fp = 0, and consequently Po = 0). The upper (U) threshold represents the value for (1 − c) above which animals survive and show full reproduction potential (fp = 1). In between both thresholds, the probability of survival and reproduction was reduced linearly going from U to L: \[fp\ =\ [(1\ {-}\ \mathit{c})\ {-}\ L]/(U\ {-}\ L)\] For each animal in a defined environment (male and female), fp was determined based on its value for (1 − c). It was assumed that reduced reproduction rate only occurred in females, but reduced survival probability occurred in both sexes. Consequences for survival and reproduction were determined for each animal individually. For survival, if the animal had a value for (1 − c) that was between both thresholds, a random number was drawn from a uniform distribution (0, 1) for each animal that had a value for (1 − c) that was between both thresholds. The animal survived when the random number was smaller than the animal's value for fp. Animals with values for (1 − c) above U survived, and those below L died. For reproduction, each dam had a potential maximal number of offspring. For dams with value for (1 − c) above U, all offspring were born. For dams with a value for (1 − c) that was between thresholds, a random number was drawn from a uniform distribution (0, 1) for each potential offspring separately. Each individual offspring was born when the accompanying random number was smaller than fp. As discrete generations were assumed, decreased reproduction rate can be regarded as either decreased litter size (e.g., due to poor quality oocytes) but with only one reproduction cycle per dam, or multiple reproduction cycles of a single potential offspring each, but with decreased success rate. Females with a value for (1 − c) below L had already died and thus were never selected. In this study, the same thresholds were applied to define survival probability and reproduction probability. In reality, these thresholds may be different, depending on the priority given to reproduction. To summarize, Pp, R, and c are causal, partly heritable parameters; U and L are set by the environment; and fitness (represented by survival and female reproductive performance) and Po rates are resulting phenotypic parameters. This model allows for the development over time of negative relationships between fitness and traits under artificial selection that are counteracted by natural selection. The size of the conflict between fitness and production will depend on the contribution of R and c to the genetic variability in Po; large variance due to c will increase conflicts, whereas large variance due to R will avoid conflicts. Population and Parameters An initial population was simulated with 240 males and 240 females. In total, 40 discrete generations were simulated (200 replicates), and in each generation, the aim was to select 24 males and 120 females on their own performance for observed production (mass selection). No adjustment was made to the maximal number of offspring per female or to the maximal number of females selected, even if the desired number of selected animals could not be met, resulting in a fluctuating population size. Each dam had a maximum of four offspring. The average number of offspring per dam was calculated across all dams selected. Each offspring born had a probability of 0.5 to be male, otherwise it was female. Mean and phenotypic SD were 0.2 and 0.05 for Pp, 1.0 and 0.1 for R, and 0.5 and 0.05 for c, respectively. The heritabilities for Pp and R were set to 0.3, and for c, heritability was 0.1, 0.3, or 0.5. Changing the genetic variance varied heritabilities; the phenotypic variance was assumed to remain constant. The initial means and variances were chosen such that a proportion of animals was not under metabolic stress during the first generations of selection. Two types of environments were assumed, good or poor, and were indicated by the values of the thresholds. The L threshold was either 0.4, representing a good environment, or 0.5, representing a poor environment. The distance between the L and U thresholds varied by 1 or 3 phenotypic SD (depending on the type of environmental stress), resulting in U = 0.45 (1 SD) or 0.55 (3 SD), representing a good environment, and U = 0.55 (1 SD) or 0.65 (3 SD), representing a poor environment. Animals were fed ad libitum with feed of constant quality in all cases, and it was assumed that energy intake was not negatively affected due to loss of appetite. In other words, each animal was able to express full resource intake potential. Results and Discussion Reproduction Rate Population size and average number of progeny per selected dam were approximately constant across generations in the good environment and very similar for both heritabilities for c. The average number of offspring of selected dams was 3.9 across generations and population size remained larger than 450 animals at all times for \(h^{2}_{c}\ =\ 0.1\) . For \(h^{2}_{c}\ =\ 0.5\) , the average number of offspring decreased from 3.9 during the first 24 generations to 3.7 in generation 40 and population size decreased from around 470 to around 450 in generation 40. In the poor environment for \(h^{2}_{c}\ =\ 0.1\) , the initial average number of offspring per selected dam was lower (3.2) and population size was low (177). Consequently, the number of selection candidates was much smaller than intended (only 71 in the first generation), and the number of dams available only became equal to the target of 120 in generation 17. For \(h^{2}_{c}\ =\ 0.5\) , initial average number of offspring was 3.3, increasing to 3.9 in generation 11 and decreasing again to 3.7 in generation 40. The number of dams available was insufficient during the first three generations and population size increased from 199 in the initial generation to 450 in generation four, after which there was little variation. Each dam got four chances to produce offspring, often resulting in some dams not having any offspring, whereas many had four. Therefore, a strong natural selection (i.e., correlated response) for fertility rate occurred, in both the poor and good environments. Overall, there seems to be a constant, suboptimal reproduction rate of 3.7 that develops in both environments, given the current selection strategy. This can be explained by the long-term genetic contributions theory (Wray and Thompson, 1990). Animals with a high genetic potential for reproduction rate that are selected for breeding (based on observed production) have a higher potential of passing on their genes to the next generation. Also, those offspring are more likely to reproduce than are animals whose dams had a lower reproduction rate. Finally, a larger number of offspring will result in a larger chance that one of them will be selected to produce the next generation. These factors together are probably the underlying reason for the relatively high average number of offspring in later generations. In practical breeding, the best producing dams that are less reproductive often are offered more chances to reproduce, resulting in a decreased, or even negative, natural selection pressure on reproduction. Selection Response Figure 1 shows the results of 40 generations of selection on observed production, where the heritability for c is either 0.1 or 0.5 (results for \(h^{2}_{c}\ =\ 0.3\) were between these results and are therefore not shown) and the heritability for R is 0.3. Two environments are compared and the distance between thresholds is one SD. The starting point was an unselected population. Observed production in the figure is determined as the average production for all animals in the generation, including those with production equal to zero. Therefore, in the poor environment, Po initially is much lower than Pp due to insufficient values for (1 − c). Selection for increased observed production thus results in selection pressure on (1 − c). Selection pressure on (1 − c) remains until this is no longer a limiting factor. This will occur when (1 − c) has values around the upper threshold. Due to chance, there will always be selected parents with values for (1 − c) below the U threshold that still manage to have high Po and be reproductive. Also, parents with values for (1 − c) above the U threshold may have offspring with values below. Consequently, the population mean for (1 − c) in the final generations in Figure 1 remains below the U threshold. Figure 1. View largeDownload slide Consequences for fitness and production when selection is on observed production. Means are shown for resource intake (triangle), production potential (square), observed production (open diamond), resource allocation factor (×), and survival probability (circle), following selection in a good (a and c) or a poor (b and d) environment, and for a heritability for resource allocation factor (c) of 0.1 (a and b) or 0.5 (c and d). Heritabilities for production potential and resource intake potential were 0.3. Figure 1. View largeDownload slide Consequences for fitness and production when selection is on observed production. Means are shown for resource intake (triangle), production potential (square), observed production (open diamond), resource allocation factor (×), and survival probability (circle), following selection in a good (a and c) or a poor (b and d) environment, and for a heritability for resource allocation factor (c) of 0.1 (a and b) or 0.5 (c and d). Heritabilities for production potential and resource intake potential were 0.3. After a number of generations, resource intake potential becomes limiting in both environments and selection for observed production results in increased resource intake potential. The number of generations until resource intake potential becomes limiting depends on the heritability for c and on the environment, although environment is the most important factor. A low heritability for c may be positive in the good environment because the fitness potential remains higher and genetic gain in production is mainly caused by a correlated response in resource intake rather than in c. However, in the poor environment, the low heritability for c also results in a slow increase in fitness potential. Phenotypic Correlations Figure 2 shows the phenotypic correlations between Po and Pp, resource intake potential, and the resource allocation factor. For the correlation with survival probability, values for Po were defined as if all animals would have been able to produce, regardless of survival. All results come from the same 200 replicates as in Figure 1. Because Pp, resource intake potential, and resource allocation factor were assumed to be genetically uncorrelated and the environmental variance was assumed to be constant, the phenotypic correlations between these traits were zero by definition (and in the simulation results). In the good environment, the correlation between observed and potential production is initially high, but decreases across generations to approximately 0.10. When resource intake becomes limiting (i.e., under selection pressure), the correlation between resource intake potential and Po increases to approximately 0.10. Simultaneously, the correlation with c increases. In the poor environment, Po is the factor most strongly correlated to resource allocation (0.6 to 0.75 for \(h^{2}_{c}\ =\ 0.1\) , and 0.42 to 0.75 for \(h^{2}_{c}\ =\ 0.5\) ). Selection immediately increases pressure on resources for fitness, therefore metabolic stress, and thus the development of environmental sensitivity, occurs at an earlier generation in the good environment. Figure 2. View largeDownload slide Consequences for correlations with observed production. Correlations are shown for observed production with production potential (closed diamond), resource intake (open diamond), resource allocation factor (closed circle), and survival probability (closed triangle), following selection in a good (a and c) or a poor (b and d) environment, and for a heritability for resource allocation factor (c) of 0.1 (a and b) or 0.5 (c and d). Heritabilities for production potential and resource intake potential were 0.3. Figure 2. View largeDownload slide Consequences for correlations with observed production. Correlations are shown for observed production with production potential (closed diamond), resource intake (open diamond), resource allocation factor (closed circle), and survival probability (closed triangle), following selection in a good (a and c) or a poor (b and d) environment, and for a heritability for resource allocation factor (c) of 0.1 (a and b) or 0.5 (c and d). Heritabilities for production potential and resource intake potential were 0.3. Beilharz et al. (1993) and Knap and Bishop (2000) argue that when resources become limiting, a negative correlation between production traits and fitness-related traits will result. Results in this study indicate that environmental sensitivity, indicated by the negative correlation between Po and survival probability, develops as soon as there is metabolic stress. This finding is in agreement with Kolmodin et al. (2002), who found a similar trend in environmental sensitivity in Scandinavian dairy cattle. Apparently, the animals with the highest Po (trait under selection) tend to be the animals with poorer values for c, and thus are those with increased environmental sensitivity. Switched Environments Long-term selection for production in one environment will lead to increased production in that particular environment. However, it does not necessarily mean that those improved animals would perform as well in a second environment with different production circumstances. The main reason is a difference in natural selection pressure on traits underlying observed production in both environments (Van der Waaij et al., 2000). Thorpe and Luiting (2000) argue that selection in a poor environment will result in more robust animals. Rauw et al. (1998) have summarized situations in practical pig, poultry, and dairy cattle breeding where this may be the case. Kolmodin et al. (2003) have shown that selection for high phenotypic values does result in increased environmental sensitivity in dairy cattle, especially when the environment is improved during the generations of selection. To validate whether the model presented here would also predict this, animals were selected in one environment for 10 or 40 generations, transferred to the opposite environment, and subsequently selected to produce generation 11 or 41. Results in Table 1 indicate that the difference in the quality of the environments and the number of generations of selection in those environments are important factors determining the presence and size of environmental sensitivity and, consequently, the development of a G × E interaction. The heritability for c, representing the potential selection response in c, has a larger influence in the short term (10 generations) than in the long term (40 generations). Resource intake potential and resource allocation factor do not change going from one environment to the other. However, Po, survival probability, and reproduction probability do, mainly because of a change in the importance of (1 − c). Table 1. Consequences of changing environment following 10 or 40 generations of mass selection for average observed production in either a good or a poor environment when the heritability for c is either 0.1 or 0.5, the heritability for resource intake and production potential is 0.3, and the distance between thresholds is 1 SD Responsesa \(h^{2}_{c}\) Changeb Genc Po R c surv rpo,pp rpo,R rpo,c rpo,surv noff 10 generations 0.1 P→G 10 0.13 1.00 0.53 0.52 0.21 0.00 0.74 0.011 3.5 11 0.26 1.00 0.53 0.98 0.79 0.01 0.21 0.009 3.5 G→P 10 0.31 1.01 0.51 0.94 0.51 0.02 0.38 −0.006 3.9 11 0.13 1.01 0.51 0.38 0.09 0.00 0.73 −0.024 3.9 0.5 P→G 10 0.28 1.00 0.59 0.90 0.40 0.04 0.46 −0.03 3.9 11 0.32 1.01 0.59 1.00 0.92 0.11 −0.12 −0.007 3.9 G→P 10 0.32 1.00 0.53 0.98 0.68 0.03 0.22 −0.003 4.0 11 0.18 1.01 0.53 0.52 0.13 0.02 0.74 −0.043 4.0 40 generations 0.1 P→G 40 0.39 1.24 0.57 0.78 0.07 0.09 0.58 −0.37 3.7 41 0.51 1.26 0.57 0.99 0.42 0.35 −0.44 −0.06 3.7 G→P 40 0.56 1.33 0.49 0.87 0.09 0.10 0.47 −0.27 3.7 41 0.13 1.35 0.49 0.22 0.00 0.04 0.66 −0.58 3.7 0.5 P→G 40 0.43 1.36 0.56 0.78 0.06 0.08 0.59 −0.38 3.8 41 0.56 1.37 0.56 1.00 0.41 0.36 −0.49 −0.05 3.8 G→P 40 0.53 1.32 0.47 0.81 0.07 0.10 0.58 −0.28 3.7 41 0.08 1.33 0.47 0.13 0.00 0.03 0.59 −0.50 3.7 Responsesa \(h^{2}_{c}\) Changeb Genc Po R c surv rpo,pp rpo,R rpo,c rpo,surv noff 10 generations 0.1 P→G 10 0.13 1.00 0.53 0.52 0.21 0.00 0.74 0.011 3.5 11 0.26 1.00 0.53 0.98 0.79 0.01 0.21 0.009 3.5 G→P 10 0.31 1.01 0.51 0.94 0.51 0.02 0.38 −0.006 3.9 11 0.13 1.01 0.51 0.38 0.09 0.00 0.73 −0.024 3.9 0.5 P→G 10 0.28 1.00 0.59 0.90 0.40 0.04 0.46 −0.03 3.9 11 0.32 1.01 0.59 1.00 0.92 0.11 −0.12 −0.007 3.9 G→P 10 0.32 1.00 0.53 0.98 0.68 0.03 0.22 −0.003 4.0 11 0.18 1.01 0.53 0.52 0.13 0.02 0.74 −0.043 4.0 40 generations 0.1 P→G 40 0.39 1.24 0.57 0.78 0.07 0.09 0.58 −0.37 3.7 41 0.51 1.26 0.57 0.99 0.42 0.35 −0.44 −0.06 3.7 G→P 40 0.56 1.33 0.49 0.87 0.09 0.10 0.47 −0.27 3.7 41 0.13 1.35 0.49 0.22 0.00 0.04 0.66 −0.58 3.7 0.5 P→G 40 0.43 1.36 0.56 0.78 0.06 0.08 0.59 −0.38 3.8 41 0.56 1.37 0.56 1.00 0.41 0.36 −0.49 −0.05 3.8 G→P 40 0.53 1.32 0.47 0.81 0.07 0.10 0.58 −0.28 3.7 41 0.08 1.33 0.47 0.13 0.00 0.03 0.59 −0.50 3.7 a Po = observed production; R = resource intake potential; c = resource allocation factor; fp = survival probability; rpo,pp = correlation between Po and Pp; rpo,R = correlation between Po and R; rpo,c = correlation between Po and c; rpo,surv = correlation between Po and survival probability; and noff = average number of offspring per selected dam. b Change = transfer from one environment to the other followed by one round of selection, G = good environment, and P = poor environment. c Gen = generation. View Large Table 1. Consequences of changing environment following 10 or 40 generations of mass selection for average observed production in either a good or a poor environment when the heritability for c is either 0.1 or 0.5, the heritability for resource intake and production potential is 0.3, and the distance between thresholds is 1 SD Responsesa \(h^{2}_{c}\) Changeb Genc Po R c surv rpo,pp rpo,R rpo,c rpo,surv noff 10 generations 0.1 P→G 10 0.13 1.00 0.53 0.52 0.21 0.00 0.74 0.011 3.5 11 0.26 1.00 0.53 0.98 0.79 0.01 0.21 0.009 3.5 G→P 10 0.31 1.01 0.51 0.94 0.51 0.02 0.38 −0.006 3.9 11 0.13 1.01 0.51 0.38 0.09 0.00 0.73 −0.024 3.9 0.5 P→G 10 0.28 1.00 0.59 0.90 0.40 0.04 0.46 −0.03 3.9 11 0.32 1.01 0.59 1.00 0.92 0.11 −0.12 −0.007 3.9 G→P 10 0.32 1.00 0.53 0.98 0.68 0.03 0.22 −0.003 4.0 11 0.18 1.01 0.53 0.52 0.13 0.02 0.74 −0.043 4.0 40 generations 0.1 P→G 40 0.39 1.24 0.57 0.78 0.07 0.09 0.58 −0.37 3.7 41 0.51 1.26 0.57 0.99 0.42 0.35 −0.44 −0.06 3.7 G→P 40 0.56 1.33 0.49 0.87 0.09 0.10 0.47 −0.27 3.7 41 0.13 1.35 0.49 0.22 0.00 0.04 0.66 −0.58 3.7 0.5 P→G 40 0.43 1.36 0.56 0.78 0.06 0.08 0.59 −0.38 3.8 41 0.56 1.37 0.56 1.00 0.41 0.36 −0.49 −0.05 3.8 G→P 40 0.53 1.32 0.47 0.81 0.07 0.10 0.58 −0.28 3.7 41 0.08 1.33 0.47 0.13 0.00 0.03 0.59 −0.50 3.7 Responsesa \(h^{2}_{c}\) Changeb Genc Po R c surv rpo,pp rpo,R rpo,c rpo,surv noff 10 generations 0.1 P→G 10 0.13 1.00 0.53 0.52 0.21 0.00 0.74 0.011 3.5 11 0.26 1.00 0.53 0.98 0.79 0.01 0.21 0.009 3.5 G→P 10 0.31 1.01 0.51 0.94 0.51 0.02 0.38 −0.006 3.9 11 0.13 1.01 0.51 0.38 0.09 0.00 0.73 −0.024 3.9 0.5 P→G 10 0.28 1.00 0.59 0.90 0.40 0.04 0.46 −0.03 3.9 11 0.32 1.01 0.59 1.00 0.92 0.11 −0.12 −0.007 3.9 G→P 10 0.32 1.00 0.53 0.98 0.68 0.03 0.22 −0.003 4.0 11 0.18 1.01 0.53 0.52 0.13 0.02 0.74 −0.043 4.0 40 generations 0.1 P→G 40 0.39 1.24 0.57 0.78 0.07 0.09 0.58 −0.37 3.7 41 0.51 1.26 0.57 0.99 0.42 0.35 −0.44 −0.06 3.7 G→P 40 0.56 1.33 0.49 0.87 0.09 0.10 0.47 −0.27 3.7 41 0.13 1.35 0.49 0.22 0.00 0.04 0.66 −0.58 3.7 0.5 P→G 40 0.43 1.36 0.56 0.78 0.06 0.08 0.59 −0.38 3.8 41 0.56 1.37 0.56 1.00 0.41 0.36 −0.49 −0.05 3.8 G→P 40 0.53 1.32 0.47 0.81 0.07 0.10 0.58 −0.28 3.7 41 0.08 1.33 0.47 0.13 0.00 0.03 0.59 −0.50 3.7 a Po = observed production; R = resource intake potential; c = resource allocation factor; fp = survival probability; rpo,pp = correlation between Po and Pp; rpo,R = correlation between Po and R; rpo,c = correlation between Po and c; rpo,surv = correlation between Po and survival probability; and noff = average number of offspring per selected dam. b Change = transfer from one environment to the other followed by one round of selection, G = good environment, and P = poor environment. c Gen = generation. View Large The correlation between Po and Pp changed considerably going from one environment to the other, the change being largest for \(h^{2}_{c}\ =\ 0.1\) , going from the poor to the good environment after 10 generations of selection (0.21 to 0.79). In all cases, the correlation was lowest in the poor environment and never equal to 1 in the good environment, indicating that observed production is dependent not only on Pp, especially under harsher conditions. The correlation with resource intake generally was low, and only moderately high when transferred from a poor to a good environment after 40 generations of selection. The correlation with the resource allocation factor changes considerably going from one environment to the other, the most extreme change occurring when going from a poor to a good environment after 40 generations of selection and \(h^{2}_{c}\ =\ 0.5\) ; this results in a correlation change from 0.59 to −0.49. Going from a poor to a good environment, the resource intake factor, determining how much energy is assigned to fitness-related traits, can be decreased without negative consequences for reproduction probability, and thus for Po. Transferring from a good to a poor environment causes a decrease in the correlation between Po and survival probability, especially after 40 generations of selection. Kolmodin et al. (2002) argue that environmental sensitivity develops following phenotypic selection in the presence of G × E, but the present results show that environmental sensitivity actually is the reason for the presence of G × E. Given the results in Figure 2 and Table 1, it can thus be concluded that the size of G × E depends on the number of generations of selection in separate environments. In relation to this, it is interesting to see that the (size of) change in phenotypic correlations between Po and c, resource intake, Pp, and survival probability has an important influence due to the environment, rather than to the underlying genetic parameters of c, resource intake potential, and Pp. Long-term selection in the same environment, as, for example, in the case of indigenous breeds, causes some changes in the relationships as described above. Selection in the good environment usually results in a higher Po than when selection is in the poor environment and animals are transferred to a good environment after a number of generations. However, when \(h^{2}_{c}\ =\ 0.5\) and after 40 generations of selection in the poor environment, Po was higher after transfer to the good environment than after 40 generations of selection in the good environment. It should be noted that resource intake also is higher, with economic consequences. Also, in the present study, it was assumed that resources were never limited and were of equal quality in both environments. This may be valid when differences between environments are due to, for example, climate conditions in the housing system or to a difference in infection pressure. However, when the quality and/or quantity of resources in the poor environment is lower than that in the good environment, a maximum is set to resource intake. Most likely, there will be a point in time where Po no longer increases as a consequence of limiting resources, and reproduction rate may move to a lower equilibrium than in case of ad libitum feeding. Varying Type of Environmental Stress Differences in type of environmental stress were mimicked by varying the distance between thresholds (move the U threshold). The larger the distance, the more gradual the influence of a decrease in (1 − c) will be. For \(h^{2}_{c}\ =\ 0.1\) , a distance between thresholds of 2 or 3 SD, compared with the 1 SD as discussed above, had drastic consequences. Especially in the case of 3 SD, the reproductive rate in the poor environment decreased too much and the average number of offspring per selected dam did not exceed two in any generation with a survival rate of approximately 17% across generations. The survival rate decreased a bit in the good environment as well when the distance between thresholds was increased from 1 to 3 SD, but the average number of offspring was 3.3. After 40 generations, the average Po was 0.43 and the survival and reproduction probability 0.80. The average value for (1 − c) was 0.53, and for feed intake potential, it was 1.24, instead of 0.49 and 1.33, respectively. For \(h^{2}_{c}\ =\ 0.5\) , the situation was slightly different. In the poor environment, the average number of offspring again was below two at first, but increased across generations, and from generation 19 onward, the population size started to rapidly increase. By then, (1 − c) had increased from 0.51 to 0.64. In generation 40, average Po was 0.32, (1 − c) was 0.62, resource intake capacity was 1.24, and survival and reproduction probability was 73%. In the good environment, the average number of offspring again was below two at first, but quickly increased to above two in generation 3. In generation 40, average Po reached 0.44, (1 − c) was 0.52, resource intake potential was 1.30, and survival and reproduction probability was 77%. General Discussion In this article, a model is proposed to describe the consequences of selection on Po for allocation of resources to production and fitness-related traits. By assuming simple additive relationships between the underlying components, it is shown that nonlinear relationships develop among resulting components. Depending on the heritabilities of the underlying components and on the environment, it is shown that selection on Po eventually results in a decrease in survival and reproductive rate and in the development of environmental sensitivity. Even though the proposed model is very simple, it helps increase the understanding of underlying mechanisms and consequences of selection for production when resources are limiting. The consequences of long-term selection for a production trait have been described in the literature. In poultry (Japanese quail and turkey), long-term selection for increased premature BW resulted in increased adult BW, feed intake, and feed efficiency (Marks, 1996), egg weight, and eating bouts (Nestor et al., 1996), but also in decreased hatchability and egg production (Marks, 1996), semen production, ability to walk, and resistance to infection (Nestor et al., 1996). Selection for higher milk yields resulted in increased feed intake and decreased energy balance in cattle, sheep, and pigs (reviewed by Veerkamp, 2002). In dairy cattle, where selection has been predominantly on milk production for the past decades, fertility has become a limiting factor (Pryce et al., 2000; Royal et al., 2002). The genetic correlations between various fertility-related traits and BCS, milk, fat and protein in Holstein cattle are moderately to strongly negative (e.g., Pryce et al., 2001, 2002; Royal et al., 2002), and increase unfavorably under heat stress (López-Gatius, 2003). Health is also influenced by a shift in the allocation of resources, as can be illustrated by ascites in broilers. The correlation between BW and ascites-related traits under cold conditions are moderate to high (Pakdel et al., 2002). These correlations are different for animals under cold vs. normal climate conditions (De Greef et al., 2001). Also, the resistance to infection with Pasteurella multocida and Newcastle virus decreased following generations of selection for increased premature BW in turkeys (Nestor et al., 1996). On the other hand, the metabolic balance may also shift toward increased fitness as a result of selection, as can be illustrated by chickens that have been under long-term selection for increased immune response. These animals are substantially smaller than control animals (Parmentier et al., 1996). Animals tend to adapt to their environment. For example, selection in dairy cattle in more temperate regions will decrease heat tolerance in the warmer regions (Ravagnolo and Misztal, 2002), fast-growing broilers show less heat tolerance than slower growing ones (Yalçin et al., 2001), and the indigenous Red Maasai sheep is much more productive than the Dorper in the humid coastal environment of east Africa, whereas the hybrid Dorper is only a little more productive in the semi-arid environment (Baker, et al., 2002). Results like these show that G × E interactions develop, especially when the difference between environments is substantial. The direction of selection response obtained from the simulation seems in agreement with selection results as presented in the literature; however, we were not able to find any results in the literature on the changing correlations between Po and underlying traits across generations. Validation of the model, therefore, is difficult at present, though could be done by setting up a selection experiment, for example, in mice, which are less expensive. Several papers have shown that selection for increased preweaning growth (i.e., milk production of the mother) will have increased BW and feed intake as a correlated response (e.g., Bünger et al., 1998). However, in these studies, offspring were suckled by their own mother, resulting in entanglement of genetic effects of milk production, growth, and litter size The pups would need to be cross fostered in order to eliminate this effect. Selection should be on preweaning growth of the cross-fostered pups, as a representation of milk production. Feed intake should be measured, as well as fertility (time to pregnancy), maternal BW (as an additional indication of negative energy balance), and cumulative pup weight at a number of times (milk production). Genetic parameters could then be estimated and the simulation results could be validated. In the study, it was assumed that c, Pp, and R were genetically uncorrelated. This assumption may or may not be realistic, depending on the situation. The reason for assuming zero correlation is that we do not know what the real correlations would be. Also, the correlations may be different across situations or traits considered (e.g., production traits such as growth, milk production, wool production). By assuming zero correlation between the underlying traits (i.e., c, Pp, and R), results give more insight into the size of the correlations that develop over time between those traits and the resulting traits (i.e., Po, reproduction probability). In a very poor environment, reproduction rate was not always sufficient to cover replacement rates and, as a consequence, population size decreased. Because of natural selection, in later generations, population size often recovered. As a consequence, the selected proportion was not always comparable across environments or the various genetic parameters, resulting in differing selection pressure overall, regardless of the underlying components. A decreasing population size often is not acceptable in practical animal breeding, and management would be adjusted to limit loss in population size. The purpose of this study was to gain understanding of underlying mechanisms and, therefore, results are presented when management is not adjusted, regardless of the results. Selection response in feed intake in the present article occurs as a correlated response to selection for increased observed production. The fact that it is a correlated response with a time lag may explain why the response often is not sufficient to fully express the trait under selection. In cattle breeding, an increased feed intake is not considered a negative consequence, and often is stimulated. In pig and poultry breeding, however, an increased feed intake is not always considered to be advantageous. Therefore, if one fails to allow for increased feed intake, but does select for increased production, energy for fitness will be diminished. The consequences of this selection practice would be increased selection pressure on resource allocation toward the trait with highest selection weight (i.e., production), resulting in an increase in environmental sensitivity. Putting no limits on feed intake does not automatically solve the problem. Because of the time lag, not limiting feed intake is not enough, as presented in this research. By also putting sufficient selection pressure directly on feed intake, selection pressure on c may be relieved. However, it may be difficult to balance selection pressure because resource allocation of individual animals cannot be measured and too large a feed intake would result in animals that grow fat. Selection for an optimal allocation of resources, given a certain environment, would be desirable, even though it is not clear yet how that should be accomplished. Production is a trait that is determined by many underlying components. Selection for production puts the highest selection pressure on the most limiting component. Improvement in that component may reveal a second component as most limiting, which may have, for example, a different heritability. The consequence of this shift in importance of underlying components is that correlations between Po and these components change as well. In this study, we have only discriminated between Pp, resource allocation factor, and feed intake, but in reality there are many more underlying components—all of them, together with the environment, determining the final observed production. One way of determining the size of the effect of underlying components on the trait of interest (i.e., observed production in this case) in practice would be to estimate the correlations between Po and underlying components across generations, provided these underlying components can be measured. A change in the correlations indicates a change in the importance of one of the underlying components. Implications Even though the present model simplifies reality, it gives insight into the consequences of selection for production in a good or poor environment. Selection always puts the greatest selection pressure on the most limiting underlying trait, resulting in a changing correlation between underlying traits across generations. Long-term phenotypic selection on observed production increases environmental sensitivity, which causes problems, especially when the animals are subsequently transferred to a poorer environment. The direction of correlated selection response agrees with literature results on reaction-norm models, but it remains important to validate the model with historical data, both to quantify the elements of the model (e.g., the heritability of the resource allocation factor) and to assess the predictive ability of the model. Literature Cited Baker, R. L., J. M. Mugambi, J. O. Audho, A. B. Carles, and W. Thorpe 2002. Comparison of Red Maasai and Dorper sheep for resistance to gastro-intestinal nematode parasites, productivity and efficiency in a humid and semi-arid environment in Kenya. Proc. 7th World Congr. Genet. Appl. Livest. Prod., Montpellier, France, Communication No. 13–10. Beilharz, R. G., B. G. Luxford, and J. L. Wilkinson 1993. Quantitative genetics and evolution: Is our understanding of genetics sufficient to explain evolution? J. Anim. Br. Gen. 110: 161– 170. Google Scholar CrossRef Search ADS Bünger L., U. Renne, G. Dietl, and S. Kuhla 1998. Long-term selection for protein amount over 70 generations in mice. Genet. Res. 72: 93– 109. Google Scholar CrossRef Search ADS PubMed Butler, W. R. 2003. Energy balance relationships with follicular development, ovulation and fertility in postpartum dairy cows. Livest. Prod. Sci. 83: 211– 218. Google Scholar CrossRef Search ADS Cameron, N. D., M. Enser, G. R. Nute, F. M. Whittington, J. C. Penman, A. C. Fisken, A. M. Perry, and J. D. Wood 2000. Genotype with nutrition interaction on fatty acid composition of intramuscular fat and the relationship with flavor of pig meat. Meat Sci. 55: 187– 195. Google Scholar CrossRef Search ADS PubMed Cameron, N. D., G. B. Garth, J. C. Penman, and A. Fiskin 2003. Sensitivity to dietary lysine: energy content in pigs divergently selected for components of efficient lean growth rate. Anim. Sci. 76: 175– 190. Google Scholar CrossRef Search ADS Collard, B. L., P. J. Boettcher, J. C. M. Dekkers, D. Petitclerc, and L. R. Schaeffer 2000. Relationships between energy balance and health traits of dairy cattle in early lactation. J. Dairy Sci. 83: 2683– 2690. Google Scholar CrossRef Search ADS PubMed De Greef, K. H., L. L. G. Janss, A. L. J. Vereijken, R. Pit, and C. L. M. Gerritsen 2001. Disease induced variability of genetic correlations: ascites in broilers as a case study. J. Anim. Sci. 79: 1723– 1733. Google Scholar CrossRef Search ADS PubMed De Jong, G., and A. J. Van Noordwijk 1992. Acquisition and allocation of resources: Genetic (co)variances, selection, and life histories. Am. Nat. 139: 749– 770. Google Scholar CrossRef Search ADS Emanuelson, U., G. Banos, and J. Philipsson 1999. Interbull Centre Report. Pages 1–6 in Interbull Bulletin 22. Proc. of the 1999 Interbull Mtg., Zürich, Switzerland. Knap, P. W., and S. C. Bishop 2000. Relations between genetic change and infectious disease in domestic livestock. Pages 65–80 in Occas. Pub. Br. Soc. Anim. Sci. No. 27. BSAS, Edinburgh, Scotland. Kolmodin, R., E. Strandberg, H. Jorjani, and B. Danell 2003. Selection in presence of genotype by environment interaction: Response in environmental sensitivity. Anim. Sci. 76: 375– 386. Google Scholar CrossRef Search ADS Kolmodin, R., E. Strandberg, E. Madsen, J. Jensen, and H. Jorjani 2002. Genotype by environment interaction in Nordic dairy cattle studied using reaction norms. Acta Agric. Scand. 52: 11– 24. López-Gatius, F. 2003. Is fertility declining in dairy cattle? A retrospective study in north-eastern Spain. Theriogenology 60: 89– 99. Google Scholar CrossRef Search ADS PubMed Luiting, P., and E. M. Urff 1991. Optimization of a model to estimate residual feed consumption in the laying hen. Livest. Prod. Sci., 27: 321– 338. Google Scholar CrossRef Search ADS Marks, H. L. 1996. Long-term selection for body weight in Japanese quail under different environments. Poult. Sci. 75: 1198– 1203. Google Scholar CrossRef Search ADS PubMed Nestor, K. E., D. O. Noble, N. J. Zhu, and Y. Moritsu 1996. Direct and correlated responses to long-term selection for increased body weight and egg production in turkeys. Poult. Sci. 75: 1180– 1191. Google Scholar CrossRef Search ADS PubMed Pakdel, A., J. A. M. Van Arendonk, A. L. J. Vereijken, and H. Bovenhuis 2002. Genetic and phenotypic correlations for ascites related traits in broilers. Communication No. 13–16. Proc. 7th World Congr. Genet. Appl. Livest. Prod., Montpellier, France. Parmentier, H. K., M. G. B. Nieuwland, E. Rijke, G. De Vries Reilingh, and J. W. Schrama 1996. Divergent antibody responses to vaccines and divergent body weight of chicken lines selected for high and low humoral responsiveness to sheep red blood cells. Avian Dis. 40: 634– 644. Google Scholar CrossRef Search ADS PubMed Pryce, J. E., M. P. Coffey, and S. Brotherstone 2000. The genetic relationship between calving interval, body condition score and linear type and management traits in registered Holsteins. J. Dairy Sci. 83: 2664– 2671. Google Scholar CrossRef Search ADS PubMed Pryce, J. E., M. P. Coffey, S. H. Brotherstone, and J. A. Wooliams 2002. Genetic relationships between calving interval and body condition score conditional on milk yield. J. Dairy Sci. 85: 1590– 1595. Google Scholar CrossRef Search ADS PubMed Pryce, J. E., M. P. Coffey, and G. Simm 2001. The relationship between body condition score and reproductive performance. J. Dairy Sci. 84: 1508– 1515. Google Scholar CrossRef Search ADS PubMed Rauw, W. M., E. Kanis, E. N. Noordhuizen-Stassen, and F. J. Grommers 1998. Undesirable side effects of selection for high production efficiency in farm animals: A review. Livest. Prod. Sci. 56: 15– 33. Google Scholar CrossRef Search ADS Ravagnolo, O., and I. Misztal 2002. Genetic component of heat stress in dairy cattle, parameter estimation. J. Dairy Sci., 83: 2126– 2130. Google Scholar CrossRef Search ADS Royal, M. D., J. E. Pryce, J. A. Woolliams, and A. P. F. Flint 2002. The genetic relationship between commencement of luteal activity and calving interval, body condition score, production, and linear type traits in Holstein-Friesian dairy cattle. J. Dairy Sci. 85: 3071– 3080. Google Scholar CrossRef Search ADS PubMed Thorpe, B. H., and E. Luiting 2000. Breeding for resistance to production diseases in poultry. Pages 357–377 in Breeding for Disease Resistance in Farm Animals. 2nd ed. R. F. E. Axford, S. C. Bishop, F. W. Nicholas, and J. B. Owen ed. CAB Int., Oxon, U.K. Van der Waaij, E. H., P. Bijma, S. C. Bishop, and J. A. M. Van Arendonk 2000. Modeling selection for production traits under selection under constant infection pressure. J. Anim. Sci. 78: 2809– 2820. Google Scholar CrossRef Search ADS PubMed Van der Waaij, E. H., L. L. G. Janss, and P. Bijma 2002. A resource allocation model describing consequences of artificial selection under metabolic stress. Communication No. 23–11. Proc. 7th World Congr. Genet. Appl. Livest. Prod., Montpellier, France. Veerkamp, R. F. 2002. Feed intake and energy balance in lactating animals. Proc. 7th World Congr. Genet. Appl. Livest. Prod., Montpellier, France, Communication No. 10–01. Wray, N. R., and R. Thompson 1990. Prediction of rates of inbreeding in selected populations. Genet. Res. Camb. 55: 41– 54. Google Scholar CrossRef Search ADS Yalçin, S., S. Özkan, T. Türknut, and P. B. Siegel 2001. Responses to heat stress in commercial and local broiler stocks. Br. Poult. Sci. 42: 149– 152. Google Scholar CrossRef Search ADS PubMed Copyright 2004 Journal of Animal Science
Effect of inbreeding on the incidence of retained placenta in Friesian horsesSevinga, M.;Vrijenhoek, T.;Hesselink, J. W.;Barkema, H. W.;Groen, A. F.
doi: 10.1093/ansci/82.4.982pmid: N/A
Abstract This study was motivated by the hypothesis that the incidence of retained placenta (RP) in Friesian horses is associated with inbreeding. The objectives were to 1) calculate the inbreeding rate in the total registered Friesian horse population; 2) study the association of the inbreeding coefficient of the foal and the mare with the incidence of RP; and 3) study the heritability of RP in Friesian mares after normal foalings. Data from the total registered Friesian horse population from 1879 to 2000 (52,392 individuals) were collected from the registration files of the Friesian Horse Studbook. In 1999 and 2000, 495 parturitions in 436 mares were studied. From 1979 to 2000, the inbreeding rate of the total population was 1.9% per generation. The regression coefficients for the regression of the incidence of RP on inbreeding coefficients of the foal and the mare were 0.12 ± 0.052 and −0.016 ± 0.019, respectively. Mean heritability estimates of RP as a foal trait and as a mare trait were 0.046 ± 0.088 and 0.105 ± 0.123, respectively. It was concluded that, in order to avoid a further increase in the incidence of RP in Friesian mares, a decrease in the inbreeding rate by increasing the effective breeding population is required. Furthermore, the findings indicate that the high incidence of RP in Friesian horses is at least partly a result of inbreeding. Introduction In Friesian mares, there is a high incidence (54%) of retained placenta (RP) after normal foalings (Sevinga et al., 2003). This contrasts with previously reported incidences of equine RP of 2 to 10% (Vandeplassche et al., 1971; Provencher et al., 1988). Retained placenta is commonly defined as a failure to expel all fetal membranes within 3 h of delivery (Blanchard and Varner, 1993). In a previous study, no association was found between RP in Friesian mares and potential risk factors (Sevinga et al., 2003). In cattle, little is known of underlying mechanisms of causal factors for RP (Laven and Peters, 1996). Even though estimates of heritability of RP in cattle were low (Distl et al., 1991; Schnitzenlehner et al., 1998), genetic effects are assumed to contribute considerably to the occurrence of RP. Since the origin of the breed, the number of Friesian horses has declined during several critical periods (Osinga, 2000). The increased agricultural mechanization in the early 1900s caused a dramatic decrease in interest in Friesian horses. In 1917, only three stallions were registered. A second bottleneck occurred in the late 1960s, when approximately 1,000 horses and 500 matings were registered. The present-day number of approximately 30,000 registered horses originates from this small population. This history suggests the occurrence of inbreeding and genetic drift (Nicholas, 1996). These phenomena may cause a decrease of the level of heterozygosity in a population, which may lead to inbreeding depression. We hypothesize that the incidence of RP in Friesian horses is associated with inbreeding. Therefore, the objectives of the present study were to 1) calculate the rate of inbreeding in the total registered Friesian horse population; 2) study the association of the inbreeding coefficient of the foal and the mare with the incidence of RP; and 3) estimate the heritability of RP. Materials and Methods Total Friesian Horse Population Data from the total registered Friesian horse population from 1879 to 2000 were collected from the registration files of the Friesian Horse Studbook. Individual inbreeding coefficients and a sequential ID file were generated using Pedigree Viewer (http://www.personal.une.edu.au/~bkinghor/pedigree.htm). Average inbreeding coefficients for each year (Ft), and rate of inbreeding ([Ft − Ft − 1]/[1 − Ft]) were calculated (Falconer, 1989). The base population was defined as consisting of parents of unknown descent with an inbreeding coefficient of 0.0. The effective population size (Ne) was calculated as Ne = ½δF (Falconer, 1989), where δF = inbreeding trend per generation. All four selection paths were included when estimating the generation interval used for calculating the inbreeding trend per generation. Study Population for Inbreeding and Retained Placenta Before the foaling season of 1999 and 2000, owners (n = 198) of Friesian brood mares (Equus caballus) in 12 veterinary practices in the Dutch provinces of Fryslân and Noord-Brabant (n = 11 and n = 1, respectively) were willing and able to participate in this study. Participating owners agreed to register all parturitions of their Friesian mares and to not treat RP within 3 h after delivery of the foal. After at least 3 h following delivery, the mares with RP were, if necessary, treated systemically with oxytocin or oxytocin dissolved in a calcium-magnesium-borogluconate solution (Sevinga et al., 2002a), and the placenta was removed manually when this treatment was not successful. In case of dystocia, mares were excluded from the study. To classify the mares with and without RP, the time of delivery of the foal and the time of expulsion and/or removal of the placenta were registered. Retained placenta was defined as a failure to expel all or part of the fetal membranes for at least 3 h after delivery. The registered data were checked for completeness and adequacy, such as certainty about the time of expulsion of the fetal membranes and the interval between the delivery of the foal and the initiation of treatment. Statistical Analyses Regression analysis was performed using the ASREML software package (http://www.vsn-intl.com/ASReml/). Regression coefficients of the incidence of RP on the inbreeding coefficient of the foal (foal regression), as well as that of the mare (mare regression), were estimated separately. Based on preliminary analyses, the following regression models were used: \[Foal\ regression:\ s\ +\ se\ +\ ysp\ +\ bF_{f}\ +\ e\] \[Mare\ regression:\ sm\ +\ se\ +\ ysp\ +\ bF_{m}\ +\ e\] where s represents the sire effect, sm is the effect of the maternal grandsire, se is the sex of the foal, ysp is the interaction effect of year, season, and veterinary practice, b is the regression coefficient of the occurrence of RP on the inbreeding coefficient, Ff is the inbreeding coefficient of the foal, Fm is the inbreeding coefficient of the mare, and e is the environmental effect. Logistic regression was performed in order to account for the binary character of RP. The following mixed sire model was derived from both regression models: \[y\ =\ Xb\ +\ Z_{1}s\ +\ Z_{2}ysp\ +\ e\] where the vector y consists of all observations with (yi = 1) or without (yi = 0) the occurrence of RP; vector b represents one fixed effect, sex of the foal, with two classes (male and female); and vector s contains random sire effects, with 52 classes (sires) in the case of foal regression and 69 classes (maternal grandsires) in the case of mare regression. The variance of s is defined as \(A{\sigma}^{2}_{s}\) , where A is the matrix with additive genetic relationships between sires. The matrix A is not assumed to be identical to I, the identity matrix, because pedigree information from two generations was included. Therefore not all sires are unrelated. The vector ysp contained random effects of year × season year × practice (72 combinations). The variance of ysp is defined as \(I{\sigma}^{2}_{ysp}\) . Vector e refers to the residual effects, with variance \(I{\sigma}^{2}_{e}\) . The incidence matrices X, Z1, Z2 relate observations to corresponding sex, sire and ysp effect levels. Calculated variances were used to estimate heritability, using the following formula: \[h^{2}\ =\ 4\ {\sigma}^{2}_{s}/{\sigma}^{2}_{p};\] where h2 is the heritability of RP, \({\sigma}^{2}_{s}\) is the estimated sire variance, and \({\sigma}^{2}_{p}\) is the total phenotypic variance. The total phenotypic variance is made up of the sire variance \(({\sigma}^{2}_{p})\) , the year, year × season year × practice variance \(({\sigma}^{2}_{ysp})\) , and the error variance \({\sigma}^{2}_{e})\) . Results Total Friesian Horse Population The total Friesian horse population comprised 52,392 individuals, with 21,991 stallions, 547 (2.5%) of which have been used as sires. For every individual, the entire pedigree was registered. Mean total number of offspring per sire was 92. In 1979 and 2000, mean number of offspring per sire was 15 and 54, being 3.4 and 1.4%, respectively, of the total annual offspring. Three stallions produced more than 1,000 total offspring. The base population comprised 1,938 individuals. The number of parturitions per year is shown in Figure 1, and was 4,178 in 1999 and 3,722 in 2000. Figure 1. View largeDownload slide Number of parturitions per year. Figure 1. View largeDownload slide Number of parturitions per year. Mean inbreeding coefficients of the foals born in 1999 and 2000 were 0.156 ± 0.019 and 0.157 ± 0.018, respectively. As shown in Figure 2, before 1940, the inbreeding coefficient increased only moderately. However, from 1940 to 1979, the inbreeding coefficient showed a substantial, though not constant, increase from 0.04 to approximately 0.12. The inbreeding rate from 1976 to 2000 is shown in Figure 3, and was on average 0.002/yr. With a mean generation interval of 9.6 yr, the inbreeding rate was 0.019 per generation. The Ne was 27 individuals. Figure 2. View largeDownload slide Inbreeding in the Friesian horse population from 1863 to 2000. Figure 2. View largeDownload slide Inbreeding in the Friesian horse population from 1863 to 2000. Figure 3. View largeDownload slide Inbreeding rate ([Ft − Ft − 1]/[1 − Ft]) from 1976 to 2000. Figure 3. View largeDownload slide Inbreeding rate ([Ft − Ft − 1]/[1 − Ft]) from 1976 to 2000. Study Population for Inbreeding and Retained Placenta After correcting for inadequate data, 495 observations in 436 mares were included in the study. Numbers of sires and dams' sires were 52 and 69, respectively. Mean number of offspring per sire was 9.5, with 10 sires accounting for 50% of the foals. Two stallions were sire and dam's sire as well. Incidence of RP between sires and dam's sires ranged from 29.4 to 72.0% and from 42.1% to 66.7%, respectively. The generation interval was 9 yr (mean age of sires 9.2, and of dams 8.7 yr). Of 495 normal parturitions in 436 mares, 267 cases of RP (53.9% with a 95% confidence interval of 49.5 to 58.4%) were observed. Mean inbreeding coefficients of the foals and mares were 0.158 ± 0.018, and 0.145 ± 0.023, respectively. The slopes of the regressions of the incidence of RP on inbreeding of the foal and of the mare were equal to 0.12 ± 0.052 and −0.016 ± 0.019, respectively. Indications of the slopes of the regressions of the incidence of RP on the inbreeding coefficient of the foal and of the mare are shown in Figure 4a and 4b, respectively. These figures were created by grouping the data according to inbreeding coefficient (as indicated in the figures). For each group, the incidence of RP was calculated and plotted against the inbreeding coefficient. Mean heritability estimates of RP as a foal trait and as a mare trait were 0.046 ± 0.088 and 0.105 ± 0.123, respectively. Figure 4. View largeDownload slide Indication of the regression of the incidence of retained placenta on inbreeding percentage of (A) the foal and (B) the mare. Figure 4. View largeDownload slide Indication of the regression of the incidence of retained placenta on inbreeding percentage of (A) the foal and (B) the mare. Discussion In our study, mean inbreeding coefficients of all foals of the total population born in 1999 and 2000, were 0.156 and 0.157, respectively, which is even higher than the level of inbreeding resulting from half-sib mating (0.125). Inbreeding decreases the level of heterozygosity, leading to a reduction in the potential for improving that population by selection (Nicholas, 1996). This consequence may undermine the aims of horse breeders. Rather than the level (Ft) and increment of inbreeding (Ft − Ft − 1), the inbreeding rate ([Ft − Ft − 1]/[1 − Ft]) is the essential population parameter (Falconer, 1989). Although detailed knowledge of relevant parameters for putting a constraint on the inbreeding rate is lacking, an acceptable level may be between 0.5 and 1% per generation (Bijma, 2000). From 1979 to 2000, the inbreeding rate of the total population was 1.9% per generation. The effective population size was 27 individuals, predominantly sires. However, the mean number of sires used in this period was 49, which, with equal use of the sires, would have resulted in an inbreeding rate per generation of 1%. Firstly, this shows that only a portion of the stallions contributes substantially to the population, which is also reflected by the finding that, in the studied population, only 10 of 52 sires used accounted for 50% of the offspring. Moreover, with this lower inbreeding rate, risk of losing genetic variance in the population is lower and the expected effects of inbreeding depression may be smaller as well. Therefore, increasing the effective population size would be advantageous. From 1979 to 2000, the number of offspring per stallion increased from 15 to 54, although the relative proportion per sire decreased from 3.4 to 1.4%. The latter is probably in part a result of the Friesian Horse Studbook providing inbreeding coefficients for new offspring since 1979, and advising against using sire–dam combinations that produce offspring with an inbreeding coefficient higher than 5% (calculated over five generations). However, considering the present inbreeding rate, and the relationship between the incidence of RP and the inbreeding coefficient of the foal, this policy has not been sufficiently successful. Therefore, if this breeding policy remains unchanged, the incidence of RP is likely to increase in the future, provided that there is no selection against RP. However, a change in breeding policy might be counteracted by the relatively small practical consequences of RP in Friesian mares (Sevinga et al., 2002a,b), which, at the same time, make selection against RP unlikely. Likewise, the estimated heritability of RP is relatively low, which would likely result in slow progress due to selection. The findings of our study show a positive linear relationship between a high incidence of RP and the inbreeding coefficient of the foal in the studied population of Friesian horses. In contrast, the regression coefficient of the incidence of RP on the inbreeding coefficient of the mare was negative, with relatively large standard error. The incidence of RP is therefore assumed to be uninfluenced by the inbreeding coefficient of the mare, and RP could thus be regarded as a foal trait. It is suggested that, in cattle, the maturation and separation process of the placenta in normal at term deliveries could be guided by a maternal × fetal immunological interaction, possibly dependent on (in)compatibility of major histocompatibility complex (MHC) class I products between cow and calf (Joosten et al., 1991). The compatibility of MHC antigens, and thus degree of relationship between cow and calf, might be associated with RP. The probability that a foal inherits a paternal haplotype that is identical to that of the mare is increased in a foal with a relatively high inbreeding coefficient. This might determine the degree to which the mare mounts an immune response against the paternal haplotypes/alleles that are inherited by the foal, resulting in inhibited maturation and a delayed separation and expulsion of the placenta. To study the potential relationship between MHC class I genes and RP, comparative molecular studies on (markers for) MHC class I genes of sires, dams, and foals with regard to RP are indicated. Furthermore, altered placental maturation might lead to, or be a consequence of, histo-pathological changes of placental tissues. Therefore, this subject also needs further investigation. Heritability estimates of RP in the present study were not very precise, which was caused by the relatively small population studied. Regarding the previously reported low incidences of RP in other breeds of horses (2 to 10%) (Vandeplassche et al., 1971; Provencher et al., 1988), a more precise estimate of heritability might be difficult to obtain. In conclusion, the findings of the present study show a strong inbreeding rate in the total Friesian horse population over the last decades. The present findings indicate that the high incidence of RP in Friesian horses is, at least partly, a result of inbreeding. Reduction of the inbreeding rate by increasing the effective breeding population is required in order to avoid a further increase in the incidence of RP in Friesian mares. Implications The findings of our study support our hypothesis that the incidence of retained placenta in Friesian horses is associated with inbreeding. Due to this association, combined with the determined strong inbreeding trend over the last decades, a further increase of the high incidence of retained placenta is to be expected, unless appropriate measures are taken. Reduction of the inbreeding trend is required, which can be achieved by increasing the effective breeding population. A substantial increase of the number of sires used would be the most effective measure. Emphasizing the usefulness of sire/dam combinations that produce offspring with relatively low inbreeding coefficients should be part of the breeding policy. Because the estimated heritability is low, selection against retained placenta would likely result in slow progress. Literature Cited Bijma, P. 2000. Long-term genetic contributions: Prediction of rates of inbreeding and genetic gain in selected populations. Ph.D. Diss., Wageningen Univ., Wageningen. Universal Press, Veenendaal, The Netherlands. Blanchard, T. L., and D. D. Varner 1993. Therapy for retained placenta in the mare. Vet. Med. 88: 55– 59. Distl, O., M. Ron, and G. Francos 1991. Genetic analysis of reproductive disorders in Israeli Holstein dairy cows. Theriogenology 35: 827– 836. Google Scholar CrossRef Search ADS PubMed Falconer, D. S. 1989. Introduction to Quantitative Genetics. 3rd ed. Longman Group, Harlow, Essex, U.K. Joosten, I., M. F. Sanders, and E. J. Hensen 1991. Involvement of major histocompatibilty complex class I compatibility between dam and calf in the aetiology of bovine retained placenta. Anim. Genet. 22: 455– 463. Google Scholar CrossRef Search ADS PubMed Laven, R. A., and A. R. Peters 1996. Bovine retained placenta: aetiology, pathogenesis and econonmic loss. Vet. Rec. 139: 465– 471. Google Scholar CrossRef Search ADS PubMed Nicholas, F. W. 1996. Introduction to Veterinary Genetics. Oxford Univ. Press, Oxford, U.K. Osinga, A. 2000. Het fokken van het Friese paard. Stichting It Fryske Hoars, Leeuwarden, The Netherlands. Provencher, R., W. R. Threlfall, P. W. Murdick, and W. K. Wearly 1988. Retained fetal membranes in the mare: A retrospective study. Can. Vet. J. 29: 903– 910. Google Scholar PubMed Schnitzenlehner, S., A. Essl, and J. Sölkner 1998. Retained placenta: estimation of nongenetic effects, heritability and correlations to important traits in cattle. J. Anim. Breed Genet. 115: 467– 478. Google Scholar CrossRef Search ADS Sevinga, M., H. W. Barkema, and J. W. Hesselink 2002a. Serum calcium and magnesium concentrations and the use of a calcium-magnesium-borogluconate solution in the treatment of Friesian mares with retained placenta. Theriogenology 57: 941– 947. Google Scholar CrossRef Search ADS Sevinga, M., H. W. Barkema, H. Stryhn, and J. W. Hesselink 2003. Retained placenta in Friesian mares: incidence, and potential risk factors with special emphasis on gestational length. Theriogenology (in press). Sevinga, M., J. W. Hesselink, and H. W. Barkema 2002b. Reproductive performance of Friesian mares after retained placenta and manual removal of the placenta. Theriogenology 57: 923– 930. Google Scholar CrossRef Search ADS Vandeplassche, M., J. Spincemaille, and R. Bouters 1971. Aethiology, pathogenesis and treatment of retained placenta in the mare. Equine Vet. J. 3: 144– 147. Google Scholar CrossRef Search ADS PubMed Copyright 2004 Journal of Animal Science
Effect of inbreeding on the incidence of retained placenta in Friesian horsesSevinga, M.;Vrijenhoek, T.;Hesselink, J. W.;Barkema, H. W.;Groen, A. F.
doi: 10.2527/2004.824982xpmid: 15080317
Abstract This study was motivated by the hypothesis that the incidence of retained placenta (RP) in Friesian horses is associated with inbreeding. The objectives were to 1) calculate the inbreeding rate in the total registered Friesian horse population; 2) study the association of the inbreeding coefficient of the foal and the mare with the incidence of RP; and 3) study the heritability of RP in Friesian mares after normal foalings. Data from the total registered Friesian horse population from 1879 to 2000 (52,392 individuals) were collected from the registration files of the Friesian Horse Studbook. In 1999 and 2000, 495 parturitions in 436 mares were studied. From 1979 to 2000, the inbreeding rate of the total population was 1.9% per generation. The regression coefficients for the regression of the incidence of RP on inbreeding coefficients of the foal and the mare were 0.12 ± 0.052 and −0.016 ± 0.019, respectively. Mean heritability estimates of RP as a foal trait and as a mare trait were 0.046 ± 0.088 and 0.105 ± 0.123, respectively. It was concluded that, in order to avoid a further increase in the incidence of RP in Friesian mares, a decrease in the inbreeding rate by increasing the effective breeding population is required. Furthermore, the findings indicate that the high incidence of RP in Friesian horses is at least partly a result of inbreeding. Introduction In Friesian mares, there is a high incidence (54%) of retained placenta (RP) after normal foalings (Sevinga et al., 2003). This contrasts with previously reported incidences of equine RP of 2 to 10% (Vandeplassche et al., 1971; Provencher et al., 1988). Retained placenta is commonly defined as a failure to expel all fetal membranes within 3 h of delivery (Blanchard and Varner, 1993). In a previous study, no association was found between RP in Friesian mares and potential risk factors (Sevinga et al., 2003). In cattle, little is known of underlying mechanisms of causal factors for RP (Laven and Peters, 1996). Even though estimates of heritability of RP in cattle were low (Distl et al., 1991; Schnitzenlehner et al., 1998), genetic effects are assumed to contribute considerably to the occurrence of RP. Since the origin of the breed, the number of Friesian horses has declined during several critical periods (Osinga, 2000). The increased agricultural mechanization in the early 1900s caused a dramatic decrease in interest in Friesian horses. In 1917, only three stallions were registered. A second bottleneck occurred in the late 1960s, when approximately 1,000 horses and 500 matings were registered. The present-day number of approximately 30,000 registered horses originates from this small population. This history suggests the occurrence of inbreeding and genetic drift (Nicholas, 1996). These phenomena may cause a decrease of the level of heterozygosity in a population, which may lead to inbreeding depression. We hypothesize that the incidence of RP in Friesian horses is associated with inbreeding. Therefore, the objectives of the present study were to 1) calculate the rate of inbreeding in the total registered Friesian horse population; 2) study the association of the inbreeding coefficient of the foal and the mare with the incidence of RP; and 3) estimate the heritability of RP. Materials and Methods Total Friesian Horse Population Data from the total registered Friesian horse population from 1879 to 2000 were collected from the registration files of the Friesian Horse Studbook. Individual inbreeding coefficients and a sequential ID file were generated using Pedigree Viewer (http://www.personal.une.edu.au/~bkinghor/pedigree.htm). Average inbreeding coefficients for each year (Ft), and rate of inbreeding ([Ft − Ft − 1]/[1 − Ft]) were calculated (Falconer, 1989). The base population was defined as consisting of parents of unknown descent with an inbreeding coefficient of 0.0. The effective population size (Ne) was calculated as Ne = ½δF (Falconer, 1989), where δF = inbreeding trend per generation. All four selection paths were included when estimating the generation interval used for calculating the inbreeding trend per generation. Study Population for Inbreeding and Retained Placenta Before the foaling season of 1999 and 2000, owners (n = 198) of Friesian brood mares (Equus caballus) in 12 veterinary practices in the Dutch provinces of Fryslân and Noord-Brabant (n = 11 and n = 1, respectively) were willing and able to participate in this study. Participating owners agreed to register all parturitions of their Friesian mares and to not treat RP within 3 h after delivery of the foal. After at least 3 h following delivery, the mares with RP were, if necessary, treated systemically with oxytocin or oxytocin dissolved in a calcium-magnesium-borogluconate solution (Sevinga et al., 2002a), and the placenta was removed manually when this treatment was not successful. In case of dystocia, mares were excluded from the study. To classify the mares with and without RP, the time of delivery of the foal and the time of expulsion and/or removal of the placenta were registered. Retained placenta was defined as a failure to expel all or part of the fetal membranes for at least 3 h after delivery. The registered data were checked for completeness and adequacy, such as certainty about the time of expulsion of the fetal membranes and the interval between the delivery of the foal and the initiation of treatment. Statistical Analyses Regression analysis was performed using the ASREML software package (http://www.vsn-intl.com/ASReml/). Regression coefficients of the incidence of RP on the inbreeding coefficient of the foal (foal regression), as well as that of the mare (mare regression), were estimated separately. Based on preliminary analyses, the following regression models were used: \[Foal\ regression:\ s\ +\ se\ +\ ysp\ +\ bF_{f}\ +\ e\] \[Mare\ regression:\ sm\ +\ se\ +\ ysp\ +\ bF_{m}\ +\ e\] where s represents the sire effect, sm is the effect of the maternal grandsire, se is the sex of the foal, ysp is the interaction effect of year, season, and veterinary practice, b is the regression coefficient of the occurrence of RP on the inbreeding coefficient, Ff is the inbreeding coefficient of the foal, Fm is the inbreeding coefficient of the mare, and e is the environmental effect. Logistic regression was performed in order to account for the binary character of RP. The following mixed sire model was derived from both regression models: \[y\ =\ Xb\ +\ Z_{1}s\ +\ Z_{2}ysp\ +\ e\] where the vector y consists of all observations with (yi = 1) or without (yi = 0) the occurrence of RP; vector b represents one fixed effect, sex of the foal, with two classes (male and female); and vector s contains random sire effects, with 52 classes (sires) in the case of foal regression and 69 classes (maternal grandsires) in the case of mare regression. The variance of s is defined as \(A{\sigma}^{2}_{s}\) , where A is the matrix with additive genetic relationships between sires. The matrix A is not assumed to be identical to I, the identity matrix, because pedigree information from two generations was included. Therefore not all sires are unrelated. The vector ysp contained random effects of year × season year × practice (72 combinations). The variance of ysp is defined as \(I{\sigma}^{2}_{ysp}\) . Vector e refers to the residual effects, with variance \(I{\sigma}^{2}_{e}\) . The incidence matrices X, Z1, Z2 relate observations to corresponding sex, sire and ysp effect levels. Calculated variances were used to estimate heritability, using the following formula: \[h^{2}\ =\ 4\ {\sigma}^{2}_{s}/{\sigma}^{2}_{p};\] where h2 is the heritability of RP, \({\sigma}^{2}_{s}\) is the estimated sire variance, and \({\sigma}^{2}_{p}\) is the total phenotypic variance. The total phenotypic variance is made up of the sire variance \(({\sigma}^{2}_{p})\) , the year, year × season year × practice variance \(({\sigma}^{2}_{ysp})\) , and the error variance \({\sigma}^{2}_{e})\) . Results Total Friesian Horse Population The total Friesian horse population comprised 52,392 individuals, with 21,991 stallions, 547 (2.5%) of which have been used as sires. For every individual, the entire pedigree was registered. Mean total number of offspring per sire was 92. In 1979 and 2000, mean number of offspring per sire was 15 and 54, being 3.4 and 1.4%, respectively, of the total annual offspring. Three stallions produced more than 1,000 total offspring. The base population comprised 1,938 individuals. The number of parturitions per year is shown in Figure 1, and was 4,178 in 1999 and 3,722 in 2000. Figure 1. View largeDownload slide Number of parturitions per year. Figure 1. View largeDownload slide Number of parturitions per year. Mean inbreeding coefficients of the foals born in 1999 and 2000 were 0.156 ± 0.019 and 0.157 ± 0.018, respectively. As shown in Figure 2, before 1940, the inbreeding coefficient increased only moderately. However, from 1940 to 1979, the inbreeding coefficient showed a substantial, though not constant, increase from 0.04 to approximately 0.12. The inbreeding rate from 1976 to 2000 is shown in Figure 3, and was on average 0.002/yr. With a mean generation interval of 9.6 yr, the inbreeding rate was 0.019 per generation. The Ne was 27 individuals. Figure 2. View largeDownload slide Inbreeding in the Friesian horse population from 1863 to 2000. Figure 2. View largeDownload slide Inbreeding in the Friesian horse population from 1863 to 2000. Figure 3. View largeDownload slide Inbreeding rate ([Ft − Ft − 1]/[1 − Ft]) from 1976 to 2000. Figure 3. View largeDownload slide Inbreeding rate ([Ft − Ft − 1]/[1 − Ft]) from 1976 to 2000. Study Population for Inbreeding and Retained Placenta After correcting for inadequate data, 495 observations in 436 mares were included in the study. Numbers of sires and dams' sires were 52 and 69, respectively. Mean number of offspring per sire was 9.5, with 10 sires accounting for 50% of the foals. Two stallions were sire and dam's sire as well. Incidence of RP between sires and dam's sires ranged from 29.4 to 72.0% and from 42.1% to 66.7%, respectively. The generation interval was 9 yr (mean age of sires 9.2, and of dams 8.7 yr). Of 495 normal parturitions in 436 mares, 267 cases of RP (53.9% with a 95% confidence interval of 49.5 to 58.4%) were observed. Mean inbreeding coefficients of the foals and mares were 0.158 ± 0.018, and 0.145 ± 0.023, respectively. The slopes of the regressions of the incidence of RP on inbreeding of the foal and of the mare were equal to 0.12 ± 0.052 and −0.016 ± 0.019, respectively. Indications of the slopes of the regressions of the incidence of RP on the inbreeding coefficient of the foal and of the mare are shown in Figure 4a and 4b, respectively. These figures were created by grouping the data according to inbreeding coefficient (as indicated in the figures). For each group, the incidence of RP was calculated and plotted against the inbreeding coefficient. Mean heritability estimates of RP as a foal trait and as a mare trait were 0.046 ± 0.088 and 0.105 ± 0.123, respectively. Figure 4. View largeDownload slide Indication of the regression of the incidence of retained placenta on inbreeding percentage of (A) the foal and (B) the mare. Figure 4. View largeDownload slide Indication of the regression of the incidence of retained placenta on inbreeding percentage of (A) the foal and (B) the mare. Discussion In our study, mean inbreeding coefficients of all foals of the total population born in 1999 and 2000, were 0.156 and 0.157, respectively, which is even higher than the level of inbreeding resulting from half-sib mating (0.125). Inbreeding decreases the level of heterozygosity, leading to a reduction in the potential for improving that population by selection (Nicholas, 1996). This consequence may undermine the aims of horse breeders. Rather than the level (Ft) and increment of inbreeding (Ft − Ft − 1), the inbreeding rate ([Ft − Ft − 1]/[1 − Ft]) is the essential population parameter (Falconer, 1989). Although detailed knowledge of relevant parameters for putting a constraint on the inbreeding rate is lacking, an acceptable level may be between 0.5 and 1% per generation (Bijma, 2000). From 1979 to 2000, the inbreeding rate of the total population was 1.9% per generation. The effective population size was 27 individuals, predominantly sires. However, the mean number of sires used in this period was 49, which, with equal use of the sires, would have resulted in an inbreeding rate per generation of 1%. Firstly, this shows that only a portion of the stallions contributes substantially to the population, which is also reflected by the finding that, in the studied population, only 10 of 52 sires used accounted for 50% of the offspring. Moreover, with this lower inbreeding rate, risk of losing genetic variance in the population is lower and the expected effects of inbreeding depression may be smaller as well. Therefore, increasing the effective population size would be advantageous. From 1979 to 2000, the number of offspring per stallion increased from 15 to 54, although the relative proportion per sire decreased from 3.4 to 1.4%. The latter is probably in part a result of the Friesian Horse Studbook providing inbreeding coefficients for new offspring since 1979, and advising against using sire–dam combinations that produce offspring with an inbreeding coefficient higher than 5% (calculated over five generations). However, considering the present inbreeding rate, and the relationship between the incidence of RP and the inbreeding coefficient of the foal, this policy has not been sufficiently successful. Therefore, if this breeding policy remains unchanged, the incidence of RP is likely to increase in the future, provided that there is no selection against RP. However, a change in breeding policy might be counteracted by the relatively small practical consequences of RP in Friesian mares (Sevinga et al., 2002a,b), which, at the same time, make selection against RP unlikely. Likewise, the estimated heritability of RP is relatively low, which would likely result in slow progress due to selection. The findings of our study show a positive linear relationship between a high incidence of RP and the inbreeding coefficient of the foal in the studied population of Friesian horses. In contrast, the regression coefficient of the incidence of RP on the inbreeding coefficient of the mare was negative, with relatively large standard error. The incidence of RP is therefore assumed to be uninfluenced by the inbreeding coefficient of the mare, and RP could thus be regarded as a foal trait. It is suggested that, in cattle, the maturation and separation process of the placenta in normal at term deliveries could be guided by a maternal × fetal immunological interaction, possibly dependent on (in)compatibility of major histocompatibility complex (MHC) class I products between cow and calf (Joosten et al., 1991). The compatibility of MHC antigens, and thus degree of relationship between cow and calf, might be associated with RP. The probability that a foal inherits a paternal haplotype that is identical to that of the mare is increased in a foal with a relatively high inbreeding coefficient. This might determine the degree to which the mare mounts an immune response against the paternal haplotypes/alleles that are inherited by the foal, resulting in inhibited maturation and a delayed separation and expulsion of the placenta. To study the potential relationship between MHC class I genes and RP, comparative molecular studies on (markers for) MHC class I genes of sires, dams, and foals with regard to RP are indicated. Furthermore, altered placental maturation might lead to, or be a consequence of, histo-pathological changes of placental tissues. Therefore, this subject also needs further investigation. Heritability estimates of RP in the present study were not very precise, which was caused by the relatively small population studied. Regarding the previously reported low incidences of RP in other breeds of horses (2 to 10%) (Vandeplassche et al., 1971; Provencher et al., 1988), a more precise estimate of heritability might be difficult to obtain. In conclusion, the findings of the present study show a strong inbreeding rate in the total Friesian horse population over the last decades. The present findings indicate that the high incidence of RP in Friesian horses is, at least partly, a result of inbreeding. Reduction of the inbreeding rate by increasing the effective breeding population is required in order to avoid a further increase in the incidence of RP in Friesian mares. Implications The findings of our study support our hypothesis that the incidence of retained placenta in Friesian horses is associated with inbreeding. Due to this association, combined with the determined strong inbreeding trend over the last decades, a further increase of the high incidence of retained placenta is to be expected, unless appropriate measures are taken. Reduction of the inbreeding trend is required, which can be achieved by increasing the effective breeding population. A substantial increase of the number of sires used would be the most effective measure. Emphasizing the usefulness of sire/dam combinations that produce offspring with relatively low inbreeding coefficients should be part of the breeding policy. Because the estimated heritability is low, selection against retained placenta would likely result in slow progress. Literature Cited Bijma, P. 2000. Long-term genetic contributions: Prediction of rates of inbreeding and genetic gain in selected populations. Ph.D. Diss., Wageningen Univ., Wageningen. Universal Press, Veenendaal, The Netherlands. Blanchard, T. L., and D. D. Varner 1993. Therapy for retained placenta in the mare. Vet. Med. 88: 55– 59. Distl, O., M. Ron, and G. Francos 1991. Genetic analysis of reproductive disorders in Israeli Holstein dairy cows. Theriogenology 35: 827– 836. Google Scholar CrossRef Search ADS PubMed Falconer, D. S. 1989. Introduction to Quantitative Genetics. 3rd ed. Longman Group, Harlow, Essex, U.K. Joosten, I., M. F. Sanders, and E. J. Hensen 1991. Involvement of major histocompatibilty complex class I compatibility between dam and calf in the aetiology of bovine retained placenta. Anim. Genet. 22: 455– 463. Google Scholar CrossRef Search ADS PubMed Laven, R. A., and A. R. Peters 1996. Bovine retained placenta: aetiology, pathogenesis and econonmic loss. Vet. Rec. 139: 465– 471. Google Scholar CrossRef Search ADS PubMed Nicholas, F. W. 1996. Introduction to Veterinary Genetics. Oxford Univ. Press, Oxford, U.K. Osinga, A. 2000. Het fokken van het Friese paard. Stichting It Fryske Hoars, Leeuwarden, The Netherlands. Provencher, R., W. R. Threlfall, P. W. Murdick, and W. K. Wearly 1988. Retained fetal membranes in the mare: A retrospective study. Can. Vet. J. 29: 903– 910. Google Scholar PubMed Schnitzenlehner, S., A. Essl, and J. Sölkner 1998. Retained placenta: estimation of nongenetic effects, heritability and correlations to important traits in cattle. J. Anim. Breed Genet. 115: 467– 478. Google Scholar CrossRef Search ADS Sevinga, M., H. W. Barkema, and J. W. Hesselink 2002a. Serum calcium and magnesium concentrations and the use of a calcium-magnesium-borogluconate solution in the treatment of Friesian mares with retained placenta. Theriogenology 57: 941– 947. Google Scholar CrossRef Search ADS Sevinga, M., H. W. Barkema, H. Stryhn, and J. W. Hesselink 2003. Retained placenta in Friesian mares: incidence, and potential risk factors with special emphasis on gestational length. Theriogenology (in press). Sevinga, M., J. W. Hesselink, and H. W. Barkema 2002b. Reproductive performance of Friesian mares after retained placenta and manual removal of the placenta. Theriogenology 57: 923– 930. Google Scholar CrossRef Search ADS Vandeplassche, M., J. Spincemaille, and R. Bouters 1971. Aethiology, pathogenesis and treatment of retained placenta in the mare. Equine Vet. J. 3: 144– 147. Google Scholar CrossRef Search ADS PubMed Copyright 2004 Journal of Animal Science
Threshold-linear analysis of measures of fertility in artificial insemination data and days to calving in beef cattleDonoghue, K. A.;Rekaya, R.;Bertrand, J. K.;Misztal, I.
doi: 10.2527/2004.824987xpmid: 15080318
Abstract Mating and calving records for 47,533 first-calf heifers in Australian Angus herds were used to examine the relationship between days to calving (DC) and two measures of fertility in AI data: 1) calving to first insemination (CFI) and 2) calving success (CS). Calving to first insemination and calving success were defined as binary traits. A threshold-linear Bayesian model was employed for both analyses: 1) DC and CFI and 2) DC and CS. Posterior means (SD) of additive covariance and corresponding genetic correlation between the DC and CFI were −0.62 d (0.19 d) and −0.66 (0.12), respectively. The corresponding point estimates between the DC and CS were −0.70 d (0.14 d) and −0.73 (0.06), respectively. These genetic correlations indicate a strong, negative relationship between DC and both measures of fertility in AI data. Selecting for animals with shorter DC intervals genetically will lead to correlated increases in both CS and CFI. Posterior means (SD) for additive and residual variance and heritability for DC for the DC-CFI analysis were 23.5 d2 (4.1 d2), 363.2 d2 (4.8 d2), and 0.06 (0.01), respectively. The corresponding parameter estimates for the DC-CS analysis were very similar. Posterior means (SD) for additive, herd-year and service sire variance and heritability for CFI were 0.04 (0.01), 0.06 (0.06), 0.14 (0.16), and 0.03 (0.01), respectively. Posterior means (SD) for additive, herd-year, and service sire variance and heritability for CS were 0.04 (0.01), 0.07 (0.07), 0.14 (0.16), and 0.03 (0.01), respectively. The similarity of the parameter estimates for CFI and CS suggest that either trait could be used as a measure of fertility in AI data. However, the definition of CFI allows the identification of animals that not only record a calving event, but calve to their first insemination, and the value of this trait would be even greater in a more complete dataset than that used in this study. The magnitude of the correlations between DC and CS-CFI suggest that it may be possible to use a multitrait approach in the evaluation of AI and natural service data, and to report one genetic value that could be used for selection purposes. Introduction Little attention has been given to genetic evaluations of fertility that incorporate both natural service (NS) and AI information in one analysis. Previous work investigating reproductive performance under NS matings (e.g., Buddenberg et al., 1990; Johnston and Bunter, 1996) has focused on the traits of calving date and days to calving (DC), whereas studies using information from AI mating data have primarily evaluated binary traits, such as heifer pregnancy and calving success (CS). Donoghue et al. (2004b) investigated calving to first insemination (CFI) as a potential trait from which information from both natural and artificial matings could be combined and compared. These authors reported a genetic correlation of 0.82 between CFI under the two types of mating data and concluded that cows with a higher probability of CFI when mated using AI also had a high probability of CFI when mated via NS. One alternative method for combining both sources of mating information may be to fit a threshold-linear bivariate analysis to the data and report one genetic value for fertility. Continuous measures of fertility, such as DC, have several advantages over binary traits such as CS and CFI. They not only allow identification of animals more likely to conceive, but also allow for the identification of animals that will conceive early in the breeding season. It may be desirable to report only the genetic value for DC when AI and NS mating are combined, given that reasonable genetic relationships exist with the binary trait. The objective of this study was to investigate the genetic relationships between DC and measures of fertility in AI data (CS and CFI). Two bivariate analyses were conducted: DC with CFI and DC with CS. Genetic correlations between DC and CS or CFI were used to examine the magnitude of the genetic relationships. In addition, correlations between breeding values of CS or CFI and DC were compared for sires with progeny records. Materials and Methods Data The data comprised mating and calving records for first-calf females from Angus herds located in predominantly temperate regions of Australia. The Australian Angus database has field records and is a total female inventory recording system, for which mating details on every female in the herd are available. Animals had either an AI or a NS record, but not both. The majority of animals with AI records had a single mating record, whereas a small number (n = 2,277) had two mating records: AI mating followed by NS mating. The traits of CS and CFI were defined only for AI records, whereas DC was defined only for NS records. Only animals that had their first mating record between 270 and 625 d of age were included in the analysis. After editing, records from 47,533 females born between 1987 and 2000 were available for analysis. Edits performed for all traits included removal of 1) animals with incomplete records and 2) mating records resulting in multiple births. In addition, single-record contemporary groups were removed from the AI data, whereas contemporary groups with less than four records and contemporary groups consisting of only noncalvers were removed from the NS data. The final AI and NS datasets comprised 16,358 and 31,175 records, respectively. The total number of animals, including ancestors, was 78,912. A total of 2,239 and 3,945 sires was represented in the AI and NS datasets, respectively, with 1,547 sires having progeny in both datasets. The measures of fertility for AI matings were recorded for the same animals, and only differed in definition: CFI vs. CS. The structures of the datasets are shown in Table 1, with summary statistics for CFI and CS combined. Table 1. Descriptive statistics of the datasets for calving success and calving to first insemination (CS-CFI) and days to calving (DC) Variable CS-CFI DC Number of records 16,358 31,175 Number of sires 2,239 3,945 Number of contemporary groups 1,136 1,730 Number of service sires 687 2,539 Variable CS-CFI DC Number of records 16,358 31,175 Number of sires 2,239 3,945 Number of contemporary groups 1,136 1,730 Number of service sires 687 2,539 View Large Table 1. Descriptive statistics of the datasets for calving success and calving to first insemination (CS-CFI) and days to calving (DC) Variable CS-CFI DC Number of records 16,358 31,175 Number of sires 2,239 3,945 Number of contemporary groups 1,136 1,730 Number of service sires 687 2,539 Variable CS-CFI DC Number of records 16,358 31,175 Number of sires 2,239 3,945 Number of contemporary groups 1,136 1,730 Number of service sires 687 2,539 View Large Calving to first insemination was recorded only for AI data and was defined as a binary trait. Gestation length records within two SD of the mean (by sex of calf) were coded as 1, whereas records more than two SD above the mean, including animals that failed to calve, were coded as 0. Gestation length was computed as the difference between the insemination date and subsequent calving date, and was averaged by sex of calf. Mating records whose gestation length was more than two SD lower than the mean (by sex of calf) were considered outliers and removed from the dataset (Kadarmideen and Coffey, 2001). Nongenetic effects were linear and quadratic covariates for age at mating, month of mating, and random effects of service sire and contemporary group. Contemporary group was defined to include animals from the same herd mated in the same year. The mean incidence of calving to first insemination was 79%. Calving success was recorded only for AI data, and was defined as a binary trait. Females that calved, whether to an AI mating or a subsequent NS mating, were coded as 1, whereas animals that failed to calve were coded as 0. Nongenetic effects were linear and quadratic covariates for age at mating, month of mating, and random effects of service sire and contemporary group. Service sire was designated as the first sire to which the female was mated (i.e., the AI mating sire). Contemporary group was defined to include animals from the same herd mated in the same year. The mean incidence of calving success was 92%. The continuous trait of DC was recorded only for NS data and was defined as the number of days between the time a bull was turned out in the pasture and the subsequent parturition of the female. Nongenetic effects were linear and quadratic covariates for age at mating, sex of calf, and contemporary group. Contemporary group was defined to include animals from the same herd that were mated to the same service sire in the same month and year. The sex of calf effect was randomly assigned to either male or female for all animals with censored records. Using the approach of Donoghue et al. (2004a), trait values for censored records were simulated from their respective predictive distributions (truncated normal distributions). For all animals in the same contemporary group, the truncation point was the largest observed DC record. The predicted DC for a censored record was between the truncation point and positive infinity. Thus, an animal with a censored record could not receive a simulated record that was smaller than an uncensored record within its contemporary group. The number of days added to this truncation point for each of the censored records was determined by drawing samples at random from the truncated distribution, and depended on the fixed effects in the model, as well as its relationships with other animals. Model Two analyses were undertaken: 1) DC with CFI and 2) DC with CS. Both models were bivariate, with CFI or CS being binary (threshold) and DC being Gaussian. It was postulated that liabilities of CS or CFI and DC were jointly Gaussian. Foulley et al. (1983) developed this model. The following mixed linear animal model was used for both analyses of traits: \[\mathbf{y}\ =\ \mathbf{X{\beta}\ +\ Z_{s}}\mathbf{s}\ +\ \mathbf{Z_{u}}\mathbf{u}\ +\ \mathbf{e}\] where y was a vector of DC observations or unobserved liabilities of CS or CFI; \(\mathbf{{\beta}\ =}\ \mathbf{({\beta}}^{{^\prime}}_{1}\mathbf{,{\beta}}^{{^\prime}}_{2}\mathbf{){^\prime}}\) was the vector of systematic effects; s was the vector of service sire effects (for CFI and CS only); u was the vector of additive genetic values; e was the vector of residual terms; X, Zs, and Zu were known incidence matrices with the appropriate dimensions. For CFI and CS, β1 included herd × year effects, month of mating effects, and linear and quadratic covariates for age at mating. For DC, β2 included contemporary group effects, sex of calf effects, and linear and quadratic covariates for age at mating. The random effect of service sire was not included for the trait of DC. This effect was included as part of the contemporary group definition for DC, which is the current practice in the genetic evaluation program in Australia (Schneeberger et al., 1991). Conditionally, on the model parameters, it was assumed that the sampling distribution of observations was as follows: \[p(\mathbf{y}|\mathbf{{\beta},s},\mathbf{u},\mathbf{R_{0}})\ ~\ N(\mathbf{X{\beta}\ +\ Z_{s}}\mathbf{s}\ +\ \mathbf{Z_{u}}\mathbf{u},\ \mathbf{I}\ {\otimes}\ \mathbf{R_{0}})\] where R0 is a 2 × 2 variance–covariance matrix with the following structure: \[\mathbf{R_{0}}\ =\ \left[\begin{array}{ll}{\sigma}^{2}_{e1}&{\sigma}_{e12}\\{\sigma}_{e21}&{\sigma}^{2}_{e2}\end{array}\right]\] Given the nonidentifiability problem of threshold models, at least two restrictions were needed (Cox and Snell, 1989; Sorensen et al., 1995). In this study for the traits of CFI and CS, the threshold and residual variance \(({\sigma}^{2}_{e1})\) were arbitrarily set to zero and one, respectively. Furthermore, all animals had either an AI or a NS record; none had both traits measured. Consequently, the residual covariance (σe12 = σe21) cannot be inferred and was set to zero. The prior distribution for the residual variance for DC was: \[p({\sigma}^{2}_{e2})\ ~\ U[0,10,000]\] For the traits of CS and CFI, the vector of systematic effects was partitioned into subvectors \(\mathbf{{\beta}_{1}}\ =\ (\mathbf{{\beta}^{{^\prime}}_{H}},\ \mathbf{{\beta}^{{^\prime}}_{R}}){^\prime}\) , where βH was a vector of herd × year effects and βR was a vector containing month of mating effects and linear and quadratic covariates for age at mating. The average number of records per herd × year class was small, and many herd × year classes were either all zeros or all ones for any of the two binary traits. This can lead to poor inferences, and is generally referred to as the extreme category problem (ECP) (Misztal et al., 1989). Hence, herd × year effects were assigned a normal prior with unknown mean and variance. This is based on the results of Rekaya et al. (2000), who found that this prior distribution alleviated ECP in such data. The prior distribution for the vector βH was: \[p({\beta}_{H}|{\eta},{\sigma}^{2}_{h})\ ~\ N({\eta},\ \mathbf{I}{\sigma}^{2}_{h})\] where η and \({\sigma}^{2}_{h}\) are the mean and variance of herd × year effects, respectively. Both η and \({\sigma}^{2}_{h}\) were assumed unknown and hence, priors were specified as follows: \begin{eqnarray*}&&p({\eta})\ ~\ U[{-}10,\ 10]\\&&p(<F>{\sigma}^{2}_{h}</F>)\ ~\ U[0,1]\end{eqnarray*} where U [.] is the uniform distribution. The prior distribution for vector βR was: \[p({\beta}_{R}|{\sigma}^{2}_{R})\ ~\ N(\mathbf{0},\mathbf{I}{\sigma}^{2}_{R})\] with \({\sigma}^{2}_{R}\ =\ 10^{5}\) . As \({\sigma}^{2}_{R}\) is large relative to the residual variance, this prior distribution conveys vague prior knowledge about each of the elements of βR. For DC, the following prior distribution was assumed for the systematic effects: \[p({\beta}\mathbf{_{2}})\ ~\ N(0,10^{6})\] A normal distribution was used as the prior for the effect of service sire for the traits of CFI and CS: \[p(\mathbf{s}|{\sigma}^{2}_{s})\ ~\ N(0,\mathbf{I}{\sigma}^{2}_{s})\] The following uniform-bounded prior was assigned to the service sire variance for these traits: \[p({\sigma}^{2}_{s})\ ~\ U[0,1]\] A multivariate normal distribution was used as the prior for the animal effects: \[p(\mathbf{u}\ |\ \mathbf{A},\ \mathbf{G}_{0})\ ~\ N(\mathbf{0},\ \mathbf{A}\ {\otimes}\ \mathbf{G}_{0})\] where \(\mathbf{G}_{0}\ =\ \left[\begin{array}{ll}{\sigma}^{2}_{u1}&{\sigma}_{u12}\\{\sigma}_{u21}&{\sigma}^{2}_{u2}\end{array}\right]\) was the additive (co)variance matrix, and A was the additive relationship matrix between animals. A conjugate proper prior was assumed. The scaling factor for the prior of G0, shown below, was taken from the literature (Johnston et al., 2001). The degree of belief a priori was set to 5 conveying little weight to the prior information. \[\mathbf{G_{0}}\ =\ \left[\begin{array}{ll}25&{-}0.8\\&0.05\end{array}\right]\] The joint prior density had the form: \[\begin{array}{l}p(\mathbf{{\beta}_{1}},\ \mathbf{{\beta}_{R}},\ \mathbf{{\eta}},\ {\sigma}^{2}_{h},\ \mathbf{s},\ \mathbf{G_{0}},\ {\sigma}^{2}_{e2})\ =\\p(\mathbf{{\beta}_{1}})\ p(\mathbf{{\beta}_{R}})\ p(\mathbf{{\beta}_{H}}|\mathbf{{\eta}},{\sigma}^{2}_{h})\ p(\mathbf{{\eta}})\ p({\sigma}^{2}_{h})\\p(\mathbf{s}|{\sigma}^{2}_{s})\ p({\sigma}^{2}_{s})\ p(\mathbf{u}|\mathbf{G_{0}})\ p(\mathbf{G_{0}})\ p({\sigma}^{2}_{e2})\end{array}\] The joint posterior density is proportional to the product of the density of the conditional distribution of the observation × the joint prior density. Draws from the conditional posterior distribution of all the parameters were obtained using a Gibbs sampler with data augmentation (Sorensen et al., 1995). The joint posterior was augmented with the univariate normal liabilities for CS or CFI. After augmentation, all the fully conditional posterior distributions of model parameters can be derived as described by Albert and Chib (1993) and Sorensen et al. (1995). These distributions are normal for the systematic parameters (service sire and animal effects), truncated normal for the liabilities, scaled-inverted χ2 for the residual variance for DC, and scaled-inverted Wishart distributions for the dispersion parameters. Liabilities were sampled from their truncated normal distribution using inverse cumulative distribution function technique (Devroye, 1986). The posterior distribution was augmented with the unobserved calving dates corresponding to the censored DC observations as described in Donoghue et al. (2004a). (Co)variance components were estimated for both analyses, and genetic correlations of CS or CFI with DC were compared. Breeding values were predicted for all animals, and correlations between breeding values of CS or CFI and DC were compared for sires with progeny records. Results and Discussion For all analyses, convergence was assessed using methodology presented by Raftery and Lewis (1992). The required length of the burn-in period was always less then 5,000 iterations for all parameters. Thus, 200,000 iterations of the sampler were run with a conservative 50,000 iterations discarded as burn-in; the remaining 150,000 iterations were retained with thinning for post-Gibbs analysis. The mean and SD of DC for uncensored females was 302 ± 19 d, whereas the corresponding statistic for censored females was 354 ± 25 d. The number of days added to the largest observed DC record within a contemporary group ranged from 5 to 26 d. The majority of censored animals (63%) received a record ranging from 10 to 16 d greater than the largest observed DC record within their contemporary group, whereas 21, 14, and 2% of censored females received records with 5 to 9, 17 to 21, and 21 to 26 d added, respectively. Summaries of the posterior distributions of (co)variance components, heritabilities, and genetic correlation from the CFI-DC bivariate analysis are presented in Table 2. The posterior mean (SD) of the additive covariance between CFI and DC was −0.62 d (0.19 d), and the corresponding genetic correlation was −0.66 (0.12). These results suggest a high, negative correlation between probability of CFI and DC. A large, negative correlation indicates that, genetically, cows with a higher probability of calving to their first insemination will also record a shorter DC interval. Thus, selection for increased probability of CFI would result in correlated decreases in DC interval. This value is similar to that reported by Johnston et al. (2001) for the genetic correlation between DC and calving success (−0.66). Morris et al. (2000) also found a similar relationship between higher pregnancy rates and early calving dates in Angus heifers, but did not report the magnitude of this association. Table 2. Means, standard deviations, and bounds of high posterior density (HPD) intervals (95%) of the posterior distribution of (co)variance components, heritabilities, and correlations for bivariate analysis of calving to first insemination (CFI) and days to calving (DC) Trait Parametera Mean SD HPD (95%) CFIb \({\sigma}^{2}_{h}\) 0.06 0.06 0.01 to 0.19 \({\sigma}^{2}_{s}\) 0.14 0.16 0.01 to 0.49 \({\sigma}^{2}_{u}\) 0.04 0.01 0.02 to 0.07 h2 0.03 0.01 0.01 to 0.05 DCc \({\sigma}^{2}_{e}\) , d2 363.2 4.8 353.6 to 372.4 \({\sigma}^{2}_{u}\) , d2 23.5 4.1 16.5 to 32.6 h2 0.06 0.01 0.04 to 0.08 CFI-DC σu12, d −0.62 0.19 −1.04 to −0.31 rg −0.66 0.12 −0.89 to −0.44 Trait Parametera Mean SD HPD (95%) CFIb \({\sigma}^{2}_{h}\) 0.06 0.06 0.01 to 0.19 \({\sigma}^{2}_{s}\) 0.14 0.16 0.01 to 0.49 \({\sigma}^{2}_{u}\) 0.04 0.01 0.02 to 0.07 h2 0.03 0.01 0.01 to 0.05 DCc \({\sigma}^{2}_{e}\) , d2 363.2 4.8 353.6 to 372.4 \({\sigma}^{2}_{u}\) , d2 23.5 4.1 16.5 to 32.6 h2 0.06 0.01 0.04 to 0.08 CFI-DC σu12, d −0.62 0.19 −1.04 to −0.31 rg −0.66 0.12 −0.89 to −0.44 a \({\sigma}^{2}_{h}\) = herd-year variance; \({\sigma}^{2}_{s}\) = service sire variance; \({\sigma}^{2}_{u}\) = additive variance; \({\sigma}^{2}_{e}\) = residual variance; σu12 = additive covariance between the two traits; rg = genetic correlation. b \(h^{2}\ =\ {\sigma}^{2}_{u}/({\sigma}^{2}_{h}\ +\ {\sigma}^{2}_{s}\ +\ {\sigma}^{2}_{u}\ +\ 1)\) . c \(h^{2}\ =\ {\sigma}^{2}_{u}/({\sigma}^{2}_{u}\ +\ {\sigma}^{2}_{e})\) . View Large Table 2. Means, standard deviations, and bounds of high posterior density (HPD) intervals (95%) of the posterior distribution of (co)variance components, heritabilities, and correlations for bivariate analysis of calving to first insemination (CFI) and days to calving (DC) Trait Parametera Mean SD HPD (95%) CFIb \({\sigma}^{2}_{h}\) 0.06 0.06 0.01 to 0.19 \({\sigma}^{2}_{s}\) 0.14 0.16 0.01 to 0.49 \({\sigma}^{2}_{u}\) 0.04 0.01 0.02 to 0.07 h2 0.03 0.01 0.01 to 0.05 DCc \({\sigma}^{2}_{e}\) , d2 363.2 4.8 353.6 to 372.4 \({\sigma}^{2}_{u}\) , d2 23.5 4.1 16.5 to 32.6 h2 0.06 0.01 0.04 to 0.08 CFI-DC σu12, d −0.62 0.19 −1.04 to −0.31 rg −0.66 0.12 −0.89 to −0.44 Trait Parametera Mean SD HPD (95%) CFIb \({\sigma}^{2}_{h}\) 0.06 0.06 0.01 to 0.19 \({\sigma}^{2}_{s}\) 0.14 0.16 0.01 to 0.49 \({\sigma}^{2}_{u}\) 0.04 0.01 0.02 to 0.07 h2 0.03 0.01 0.01 to 0.05 DCc \({\sigma}^{2}_{e}\) , d2 363.2 4.8 353.6 to 372.4 \({\sigma}^{2}_{u}\) , d2 23.5 4.1 16.5 to 32.6 h2 0.06 0.01 0.04 to 0.08 CFI-DC σu12, d −0.62 0.19 −1.04 to −0.31 rg −0.66 0.12 −0.89 to −0.44 a \({\sigma}^{2}_{h}\) = herd-year variance; \({\sigma}^{2}_{s}\) = service sire variance; \({\sigma}^{2}_{u}\) = additive variance; \({\sigma}^{2}_{e}\) = residual variance; σu12 = additive covariance between the two traits; rg = genetic correlation. b \(h^{2}\ =\ {\sigma}^{2}_{u}/({\sigma}^{2}_{h}\ +\ {\sigma}^{2}_{s}\ +\ {\sigma}^{2}_{u}\ +\ 1)\) . c \(h^{2}\ =\ {\sigma}^{2}_{u}/({\sigma}^{2}_{u}\ +\ {\sigma}^{2}_{e})\) . View Large The posterior means (SD) for additive variance and heritability for DC were similar to previous estimates reported for this population (Donoghue et al., 2004a); furthermore, both estimates were within the high posterior density (HPD) (95%) intervals presented in the earlier study. The point estimates of heritability were smaller than estimates reported in the literature for DC or the equivalent trait of calving date for analyses when censored records are included (Johnston and Bunter, 1996; Morris et al., 2000; Johnston et al., 2001); however, our estimates were within the range of standard error associated with these literature estimates. The point estimate of residual variance for DC was similar to estimates previously reported for this population (Donoghue et al., 2004a). In the current study, the effect of service sire was included as part of the contemporary group definition to align with current industry practices for evaluation of DC. An additional analysis was undertaken with the random effect of service sire included in the model. The results from this analysis (not shown) were very similar to the results obtained in the current study. Fitting service sire as an additional random effect allowed for the formation of larger contemporary groups and had little effect on the estimation of additive variance or on heritability of the trait. Thus, including service sire as a random effect may be a reasonable option to avoid small contemporary groups. The posterior means for additive variance and heritability for CFI were similar to previous estimates reported for this population (Donoghue et al., 2004b); furthermore, both estimates were within the HPD (95%) intervals presented in the earlier study. The point estimate of h2 in the current study was lower than estimates reported in literature for heifer fertility; however, it is within the 90% CI reported by Evans et al. (1999) for the trait of heifer pregnancy. The low estimates of h2 observed in this study could result from more appropriate analytical procedures for data analysis, Bayesian approach vs. Method R for a small dataset, or perhaps are a reflection of the slight differences in trait definitions between CFI and heifer pregnancy. The posterior mean of herd-year variance for CFI in this study was much smaller than reported in an earlier study for the same population: 0.06 vs. 0.84 (Donoghue et al., 2004b). The variable nature of this parameter is most likely caused by a high incidence of ECP in the AI data; 50% of herd-year classes contained observations that fell into the same category for CFI. Possible reasons for this high incidence of ECP in the AI data were discussed in the earlier study, and included incomplete data recording and implementation of different management levels under AI matings. The highly fluctuating nature of this parameter suggests that caution should be used in the interpretation of the herd-year variance in the presence of ECP. Despite the large difference in this parameter observed between this study and the earlier study for CFI, additive variance and heritability point estimates were very similar in both studies, indicating that estimation of these parameters appears stable. The posterior mean of service sire variance for CFI was larger in magnitude than both herd-year and additive variances. The point estimate in the current study was higher than the estimate reported in the earlier study for this population and was outside the HPD (95%) interval for that study. However, examination of the SD and HPD (95%) interval associated with this point estimate indicates a lack of statistical evidence proving it was significantly different from zero. This result was most likely a reflection of the small number of service sires (n = 687) represented in the AI data. Summaries of the posterior distributions of (co)variance components, heritabilities, and genetic correlation from the CS-DC bivariate analysis are presented in Table 3. The posterior mean (SD) of the additive covariance between CS and DC was −0.70 d (0.14 d), and the corresponding genetic correlation was −0.73 (0.06). These results suggest a high, negative correlation between probability of CS and DC. A large, negative correlation indicates that, genetically, females with a higher probability of calving success will also record a shorter DC interval. Thus, selection for increased probability of CS would result in correlated decreases in DC interval. This value is similar in magnitude to that reported by Johnston et al. (2001) for the genetic correlation between the same traits (−0.66), and is slightly larger than the genetic correlation between CFI and DC (−0.66) in the current study. However, the SD and HPD (95%) intervals associated with the point estimates of genetic correlations between CS or CFI and DC in the current study indicate that they were not significantly different from each other. Table 3. Means, standard deviations, and bounds of high posterior density (HPD) intervals (95%) of the posterior distribution of (co)variance components, heritabilities, and correlations for bivariate analysis of calving success (CS) and days to calving (DC) Trait Parametera Mean SD HPD (95%) CSb \({\sigma}^{2}_{h}\) 0.07 0.07 0.06 to 0.19 \({\sigma}^{2}_{s}\) 0.14 0.16 0.004 to 0.44 \({\sigma}^{2}_{u}\) 0.04 0.01 0.01 to 0.06 h2 0.03 0.01 0.01 to 0.05 DCc \({\sigma}^{2}_{e}\) , d2 363.2 4.8 353.8 to 372.6 \({\sigma}^{2}_{u}\) , d2 23.5 4.1 15.5 to 31.5 h2 0.06 0.01 0.04 to 0.08 CS-DC σu12, d −0.70 0.14 −1.06 to −0.29 rg −0.73 0.06 −0.91 to −0.45 Trait Parametera Mean SD HPD (95%) CSb \({\sigma}^{2}_{h}\) 0.07 0.07 0.06 to 0.19 \({\sigma}^{2}_{s}\) 0.14 0.16 0.004 to 0.44 \({\sigma}^{2}_{u}\) 0.04 0.01 0.01 to 0.06 h2 0.03 0.01 0.01 to 0.05 DCc \({\sigma}^{2}_{e}\) , d2 363.2 4.8 353.8 to 372.6 \({\sigma}^{2}_{u}\) , d2 23.5 4.1 15.5 to 31.5 h2 0.06 0.01 0.04 to 0.08 CS-DC σu12, d −0.70 0.14 −1.06 to −0.29 rg −0.73 0.06 −0.91 to −0.45 a \({\sigma}^{2}_{h}\) = herd-year variance; \({\sigma}^{2}_{s}\) = service sire variance; \({\sigma}^{2}_{u}\) = additive variance; \({\sigma}^{2}_{e}\) = residual variance; σu12 = additive covariance between the two traits; rg = genetic correlation. b \(h^{2}\ =\ {\sigma}^{2}_{u}/({\sigma}^{2}_{h}\ +\ {\sigma}^{2}_{s}\ +\ {\sigma}^{2}_{u}\ +\ 1)\) . c \(h^{2}\ =\ {\sigma}^{2}_{u}/({\sigma}^{2}_{u}\ +\ {\sigma}^{2}_{e})\) . View Large Table 3. Means, standard deviations, and bounds of high posterior density (HPD) intervals (95%) of the posterior distribution of (co)variance components, heritabilities, and correlations for bivariate analysis of calving success (CS) and days to calving (DC) Trait Parametera Mean SD HPD (95%) CSb \({\sigma}^{2}_{h}\) 0.07 0.07 0.06 to 0.19 \({\sigma}^{2}_{s}\) 0.14 0.16 0.004 to 0.44 \({\sigma}^{2}_{u}\) 0.04 0.01 0.01 to 0.06 h2 0.03 0.01 0.01 to 0.05 DCc \({\sigma}^{2}_{e}\) , d2 363.2 4.8 353.8 to 372.6 \({\sigma}^{2}_{u}\) , d2 23.5 4.1 15.5 to 31.5 h2 0.06 0.01 0.04 to 0.08 CS-DC σu12, d −0.70 0.14 −1.06 to −0.29 rg −0.73 0.06 −0.91 to −0.45 Trait Parametera Mean SD HPD (95%) CSb \({\sigma}^{2}_{h}\) 0.07 0.07 0.06 to 0.19 \({\sigma}^{2}_{s}\) 0.14 0.16 0.004 to 0.44 \({\sigma}^{2}_{u}\) 0.04 0.01 0.01 to 0.06 h2 0.03 0.01 0.01 to 0.05 DCc \({\sigma}^{2}_{e}\) , d2 363.2 4.8 353.8 to 372.6 \({\sigma}^{2}_{u}\) , d2 23.5 4.1 15.5 to 31.5 h2 0.06 0.01 0.04 to 0.08 CS-DC σu12, d −0.70 0.14 −1.06 to −0.29 rg −0.73 0.06 −0.91 to −0.45 a \({\sigma}^{2}_{h}\) = herd-year variance; \({\sigma}^{2}_{s}\) = service sire variance; \({\sigma}^{2}_{u}\) = additive variance; \({\sigma}^{2}_{e}\) = residual variance; σu12 = additive covariance between the two traits; rg = genetic correlation. b \(h^{2}\ =\ {\sigma}^{2}_{u}/({\sigma}^{2}_{h}\ +\ {\sigma}^{2}_{s}\ +\ {\sigma}^{2}_{u}\ +\ 1)\) . c \(h^{2}\ =\ {\sigma}^{2}_{u}/({\sigma}^{2}_{u}\ +\ {\sigma}^{2}_{e})\) . View Large The posterior means for additive and residual variances for DC were very similar to estimates for DC under the DC-CFI bivariate. Furthermore, all estimates were within the high posterior density (HPD) (95%) intervals for this analysis. The posterior means for additive variance and heritability for CS in this study were low in magnitude. The point estimate of h2 for CS is very similar to the estimate for the same trait reported by Johnston et al. (2001) when the categorical nature of the trait was taken into account. Higher estimates of h2 for CS have been reported in the literature when the trait was analyzed without adjustment for the categorical nature (Johnston and Bunter, 1996). The point estimates for all parameters for CS were very close to the corresponding estimates for CFI, with all estimates within the HPD (95%) intervals for CFI. These results may reflect the similar definitions of the two traits and the fact that only a small number of animals changed from a trait value of 0 under CFI to a trait value of 1 under CS (n = 2,219). The level of ECP was high under both trait definitions; 50 and 74% of herd-year classes had observations that fell into the same category for CFI and CS, respectively. The similarity of results observed in this study between CFI and CS may not hold for a dataset when complete recording is available (i.e., both successful and unsuccessful AI matings are reported, as well as information regarding NS matings that may follow). The incidence of CFI would be expected to be lower under complete recording, with a lower incidence of ECP. There were 183 sires with more than 10 daughters with CS or CFI records, as well as more than 10 daughters with DC records. The mean (SD) DC breeding value (BV) for these sires was 0.30 d (2.57 d) with a range from −5.60 to 8.67 d. These BV were predicted using parameters from the DC-CFI analysis and are not reported for DC BV under the DC-CS analysis because of the similarity of the parameters. The mean (SD) CS BV was −0.02% (0.08), ranging from −0.28 to 0.17%. The mean (SD) CFI BV was −0.01% (0.08), ranging from −0.25 to 0.16%. The correlations between DC and CS BV and DC and CFI BV for these sires were −0.993 and −0.997, respectively. These results indicate that sires whose daughters have either an increased probability of calving success or an increased probability of calving to first insemination also produce daughters with shorter DC records. The regression coefficient of CFI on DC was −0.03% successful calving to first insemination per day; for every 1 d decrease in DC BV, there is a 0.03% increase in CFI BV, and a similar result was observed for the regression coefficient between CS and DC. Implications The potential of calving to first insemination as a measure of fertility in artificial insemination data has been confirmed. Parameter estimates for this trait were close to those for calving success. The former trait allows the identification of animals that not only record a calving event, but also calve to their first insemination, and the value would be even greater in a dataset more complete than that used in this study. The genetic correlations reported between days to calving and both measures of fertility in artificial insemination data indicate a strong, negative relationship. Selecting for animals with shorter days to calving intervals genetically will lead to correlated increases in both calving success and calving to first insemination. The magnitude of these correlations suggest that it may be possible to use a multitrait approach to the evaluation of artificial insemination and natural service data, but report one genetic value that could be used for selection purposes. Literature Cited Albert, J., and S. Chib 1993. Bayesian analysis of binary and polychotomous response data. J. Am. Statist. Assoc. 88: 669– 679. Google Scholar CrossRef Search ADS Buddenberg, B. J., C. J. Brown, and A. H. Brown 1990. Heritability estimates of calving date in Hereford cattle maintained on range under natural mating. J. Anim. Sci. 68: 70– 74. Google Scholar CrossRef Search ADS PubMed Cox, D. R., and E. J. Snell 1989. Analysis of binary data. CRC Press, London, U.K. Devroye, L. 1986. Non-Uniform Random Variate Generation. Springer-Verlag, New York, NY. Google Scholar CrossRef Search ADS Donoghue, K. A., R. Rekaya, and J. K. Bertrand 2004a. Comparison of methods for handling censored records in beef fertility data: Field data. J. Anim. Sci. 82: 357– 361. Google Scholar CrossRef Search ADS Donoghue, K. A., R. Rekaya, J. K. Bertrand, and I. Misztal 2004b. Genetic evaluation of calving to first insemination using natural and artificial insemination mating data. J. Anim. Sci. 82: 391– 396. Evans, J. L., B. L. Golden, R. M. Bourdon, and K. L. Long 1999. Additive genetic relationships between heifer pregnancy and scrotal circumference in Hereford cattle. J. Anim. Sci. 77: 2621– 2628. Google Scholar CrossRef Search ADS PubMed Foulley, J. L., D. Gianola, and R. Thompson 1983. Prediction of genetic merit from data on binary and quantitative variates with an application to calving difficulty, birth weight, and pelvic opening. Genet. Sel. Evol. 15: 401– 424. Google Scholar CrossRef Search ADS PubMed Johnston, D. J., and K. L. Bunter 1996. Days to calving in Angus cattle: Genetic and environmental effects, and covariances with other traits. Livest. Prod. Sci. 45: 13– 22. Google Scholar CrossRef Search ADS Johnston, D. J., J. M. Henshall, and B. Tier 2001. Estimate of the genetic correlation between calving success and days to calving in Angus females. Pages 353–355 in Proc. 14th Conf. Assoc. Advmt. Anim. Breed. Genet., Queenstown, New Zealand. Kadarmideen, H. N., and M. P. Coffey 2001. Quality and validation of insemination data for national genetic evaluations for dairy cow fertility in the United Kingdom. Interbull Bulletin 27: 133– 138. Misztal, I., D. Gianola, and J. L. Foulley 1989. Computing aspects of nonlinear methods of sire evaluation for categorical data. J. Dairy Sci. 72: 1557– 1568. Google Scholar CrossRef Search ADS Morris, C. A., J. A. Wilson, G. L. Bennett, N. G. Cullen, S. M. Hickey, and J. C. Hunter 2000. Genetic parameters for growth, puberty, and beef cow reproductive traits in a puberty selection experiment. N. Z. J. Agric. Res. 43: 83– 91. Google Scholar CrossRef Search ADS Raftery, A. E., and S. Lewis 1992. How many iterations in the Gibbs sampler? Pages 763–773 in Bayesian Statistics 4. J. M. Bernando, J. O. Berger, A. P. Dawid, and A. F. M. Smith ed. Oxford Univ. Press, New York, NY. Rekaya, R., K. A. Weigel, D. Gianola, B. Heringstad, and G. Klemetsdal 2000. Methods for attenuating bias of variance component estimates in threshold models when herds are small. J. Dairy Sci. 83(Suppl. 1): 56– 57. (Abstr.) Google Scholar CrossRef Search ADS Schneeberger, M., B. Tier, and K. Hammond 1991. Introducing the third generation of BREEDPLAN and GROUP BREEDPLAN. Pages 194–199 in Proc. 9th Conf. Aust. Assoc. Anim. Breeding Genetics, Melbourne, Australia. Sorensen, D. A., S. Andersen, D. Gianola, and I. Korsgaard 1995. Bayesian inference in threshold using Gibbs sampling. Genet. Sel. Evol. 27: 229– 249. Google Scholar CrossRef Search ADS Footnotes 1 Appreciation is extended to the Angus Society of Australia for providing the data; Meat and Livestock Australia for the research scholarship provided to the first author, and to D. J. Johnston and C. Teseling for their contributions. Copyright 2004 Journal of Animal Science
Threshold-linear analysis of measures of fertility in artificial insemination data and days to calving in beef cattleDonoghue, K. A.;Rekaya, R.;Bertrand, J. K.;Misztal, I.
doi: 10.1093/ansci/82.4.987pmid: N/A
Abstract Mating and calving records for 47,533 first-calf heifers in Australian Angus herds were used to examine the relationship between days to calving (DC) and two measures of fertility in AI data: 1) calving to first insemination (CFI) and 2) calving success (CS). Calving to first insemination and calving success were defined as binary traits. A threshold-linear Bayesian model was employed for both analyses: 1) DC and CFI and 2) DC and CS. Posterior means (SD) of additive covariance and corresponding genetic correlation between the DC and CFI were −0.62 d (0.19 d) and −0.66 (0.12), respectively. The corresponding point estimates between the DC and CS were −0.70 d (0.14 d) and −0.73 (0.06), respectively. These genetic correlations indicate a strong, negative relationship between DC and both measures of fertility in AI data. Selecting for animals with shorter DC intervals genetically will lead to correlated increases in both CS and CFI. Posterior means (SD) for additive and residual variance and heritability for DC for the DC-CFI analysis were 23.5 d2 (4.1 d2), 363.2 d2 (4.8 d2), and 0.06 (0.01), respectively. The corresponding parameter estimates for the DC-CS analysis were very similar. Posterior means (SD) for additive, herd-year and service sire variance and heritability for CFI were 0.04 (0.01), 0.06 (0.06), 0.14 (0.16), and 0.03 (0.01), respectively. Posterior means (SD) for additive, herd-year, and service sire variance and heritability for CS were 0.04 (0.01), 0.07 (0.07), 0.14 (0.16), and 0.03 (0.01), respectively. The similarity of the parameter estimates for CFI and CS suggest that either trait could be used as a measure of fertility in AI data. However, the definition of CFI allows the identification of animals that not only record a calving event, but calve to their first insemination, and the value of this trait would be even greater in a more complete dataset than that used in this study. The magnitude of the correlations between DC and CS-CFI suggest that it may be possible to use a multitrait approach in the evaluation of AI and natural service data, and to report one genetic value that could be used for selection purposes. Introduction Little attention has been given to genetic evaluations of fertility that incorporate both natural service (NS) and AI information in one analysis. Previous work investigating reproductive performance under NS matings (e.g., Buddenberg et al., 1990; Johnston and Bunter, 1996) has focused on the traits of calving date and days to calving (DC), whereas studies using information from AI mating data have primarily evaluated binary traits, such as heifer pregnancy and calving success (CS). Donoghue et al. (2004b) investigated calving to first insemination (CFI) as a potential trait from which information from both natural and artificial matings could be combined and compared. These authors reported a genetic correlation of 0.82 between CFI under the two types of mating data and concluded that cows with a higher probability of CFI when mated using AI also had a high probability of CFI when mated via NS. One alternative method for combining both sources of mating information may be to fit a threshold-linear bivariate analysis to the data and report one genetic value for fertility. Continuous measures of fertility, such as DC, have several advantages over binary traits such as CS and CFI. They not only allow identification of animals more likely to conceive, but also allow for the identification of animals that will conceive early in the breeding season. It may be desirable to report only the genetic value for DC when AI and NS mating are combined, given that reasonable genetic relationships exist with the binary trait. The objective of this study was to investigate the genetic relationships between DC and measures of fertility in AI data (CS and CFI). Two bivariate analyses were conducted: DC with CFI and DC with CS. Genetic correlations between DC and CS or CFI were used to examine the magnitude of the genetic relationships. In addition, correlations between breeding values of CS or CFI and DC were compared for sires with progeny records. Materials and Methods Data The data comprised mating and calving records for first-calf females from Angus herds located in predominantly temperate regions of Australia. The Australian Angus database has field records and is a total female inventory recording system, for which mating details on every female in the herd are available. Animals had either an AI or a NS record, but not both. The majority of animals with AI records had a single mating record, whereas a small number (n = 2,277) had two mating records: AI mating followed by NS mating. The traits of CS and CFI were defined only for AI records, whereas DC was defined only for NS records. Only animals that had their first mating record between 270 and 625 d of age were included in the analysis. After editing, records from 47,533 females born between 1987 and 2000 were available for analysis. Edits performed for all traits included removal of 1) animals with incomplete records and 2) mating records resulting in multiple births. In addition, single-record contemporary groups were removed from the AI data, whereas contemporary groups with less than four records and contemporary groups consisting of only noncalvers were removed from the NS data. The final AI and NS datasets comprised 16,358 and 31,175 records, respectively. The total number of animals, including ancestors, was 78,912. A total of 2,239 and 3,945 sires was represented in the AI and NS datasets, respectively, with 1,547 sires having progeny in both datasets. The measures of fertility for AI matings were recorded for the same animals, and only differed in definition: CFI vs. CS. The structures of the datasets are shown in Table 1, with summary statistics for CFI and CS combined. Table 1. Descriptive statistics of the datasets for calving success and calving to first insemination (CS-CFI) and days to calving (DC) Variable CS-CFI DC Number of records 16,358 31,175 Number of sires 2,239 3,945 Number of contemporary groups 1,136 1,730 Number of service sires 687 2,539 Variable CS-CFI DC Number of records 16,358 31,175 Number of sires 2,239 3,945 Number of contemporary groups 1,136 1,730 Number of service sires 687 2,539 View Large Table 1. Descriptive statistics of the datasets for calving success and calving to first insemination (CS-CFI) and days to calving (DC) Variable CS-CFI DC Number of records 16,358 31,175 Number of sires 2,239 3,945 Number of contemporary groups 1,136 1,730 Number of service sires 687 2,539 Variable CS-CFI DC Number of records 16,358 31,175 Number of sires 2,239 3,945 Number of contemporary groups 1,136 1,730 Number of service sires 687 2,539 View Large Calving to first insemination was recorded only for AI data and was defined as a binary trait. Gestation length records within two SD of the mean (by sex of calf) were coded as 1, whereas records more than two SD above the mean, including animals that failed to calve, were coded as 0. Gestation length was computed as the difference between the insemination date and subsequent calving date, and was averaged by sex of calf. Mating records whose gestation length was more than two SD lower than the mean (by sex of calf) were considered outliers and removed from the dataset (Kadarmideen and Coffey, 2001). Nongenetic effects were linear and quadratic covariates for age at mating, month of mating, and random effects of service sire and contemporary group. Contemporary group was defined to include animals from the same herd mated in the same year. The mean incidence of calving to first insemination was 79%. Calving success was recorded only for AI data, and was defined as a binary trait. Females that calved, whether to an AI mating or a subsequent NS mating, were coded as 1, whereas animals that failed to calve were coded as 0. Nongenetic effects were linear and quadratic covariates for age at mating, month of mating, and random effects of service sire and contemporary group. Service sire was designated as the first sire to which the female was mated (i.e., the AI mating sire). Contemporary group was defined to include animals from the same herd mated in the same year. The mean incidence of calving success was 92%. The continuous trait of DC was recorded only for NS data and was defined as the number of days between the time a bull was turned out in the pasture and the subsequent parturition of the female. Nongenetic effects were linear and quadratic covariates for age at mating, sex of calf, and contemporary group. Contemporary group was defined to include animals from the same herd that were mated to the same service sire in the same month and year. The sex of calf effect was randomly assigned to either male or female for all animals with censored records. Using the approach of Donoghue et al. (2004a), trait values for censored records were simulated from their respective predictive distributions (truncated normal distributions). For all animals in the same contemporary group, the truncation point was the largest observed DC record. The predicted DC for a censored record was between the truncation point and positive infinity. Thus, an animal with a censored record could not receive a simulated record that was smaller than an uncensored record within its contemporary group. The number of days added to this truncation point for each of the censored records was determined by drawing samples at random from the truncated distribution, and depended on the fixed effects in the model, as well as its relationships with other animals. Model Two analyses were undertaken: 1) DC with CFI and 2) DC with CS. Both models were bivariate, with CFI or CS being binary (threshold) and DC being Gaussian. It was postulated that liabilities of CS or CFI and DC were jointly Gaussian. Foulley et al. (1983) developed this model. The following mixed linear animal model was used for both analyses of traits: \[\mathbf{y}\ =\ \mathbf{X{\beta}\ +\ Z_{s}}\mathbf{s}\ +\ \mathbf{Z_{u}}\mathbf{u}\ +\ \mathbf{e}\] where y was a vector of DC observations or unobserved liabilities of CS or CFI; \(\mathbf{{\beta}\ =}\ \mathbf{({\beta}}^{{^\prime}}_{1}\mathbf{,{\beta}}^{{^\prime}}_{2}\mathbf{){^\prime}}\) was the vector of systematic effects; s was the vector of service sire effects (for CFI and CS only); u was the vector of additive genetic values; e was the vector of residual terms; X, Zs, and Zu were known incidence matrices with the appropriate dimensions. For CFI and CS, β1 included herd × year effects, month of mating effects, and linear and quadratic covariates for age at mating. For DC, β2 included contemporary group effects, sex of calf effects, and linear and quadratic covariates for age at mating. The random effect of service sire was not included for the trait of DC. This effect was included as part of the contemporary group definition for DC, which is the current practice in the genetic evaluation program in Australia (Schneeberger et al., 1991). Conditionally, on the model parameters, it was assumed that the sampling distribution of observations was as follows: \[p(\mathbf{y}|\mathbf{{\beta},s},\mathbf{u},\mathbf{R_{0}})\ ~\ N(\mathbf{X{\beta}\ +\ Z_{s}}\mathbf{s}\ +\ \mathbf{Z_{u}}\mathbf{u},\ \mathbf{I}\ {\otimes}\ \mathbf{R_{0}})\] where R0 is a 2 × 2 variance–covariance matrix with the following structure: \[\mathbf{R_{0}}\ =\ \left[\begin{array}{ll}{\sigma}^{2}_{e1}&{\sigma}_{e12}\\{\sigma}_{e21}&{\sigma}^{2}_{e2}\end{array}\right]\] Given the nonidentifiability problem of threshold models, at least two restrictions were needed (Cox and Snell, 1989; Sorensen et al., 1995). In this study for the traits of CFI and CS, the threshold and residual variance \(({\sigma}^{2}_{e1})\) were arbitrarily set to zero and one, respectively. Furthermore, all animals had either an AI or a NS record; none had both traits measured. Consequently, the residual covariance (σe12 = σe21) cannot be inferred and was set to zero. The prior distribution for the residual variance for DC was: \[p({\sigma}^{2}_{e2})\ ~\ U[0,10,000]\] For the traits of CS and CFI, the vector of systematic effects was partitioned into subvectors \(\mathbf{{\beta}_{1}}\ =\ (\mathbf{{\beta}^{{^\prime}}_{H}},\ \mathbf{{\beta}^{{^\prime}}_{R}}){^\prime}\) , where βH was a vector of herd × year effects and βR was a vector containing month of mating effects and linear and quadratic covariates for age at mating. The average number of records per herd × year class was small, and many herd × year classes were either all zeros or all ones for any of the two binary traits. This can lead to poor inferences, and is generally referred to as the extreme category problem (ECP) (Misztal et al., 1989). Hence, herd × year effects were assigned a normal prior with unknown mean and variance. This is based on the results of Rekaya et al. (2000), who found that this prior distribution alleviated ECP in such data. The prior distribution for the vector βH was: \[p({\beta}_{H}|{\eta},{\sigma}^{2}_{h})\ ~\ N({\eta},\ \mathbf{I}{\sigma}^{2}_{h})\] where η and \({\sigma}^{2}_{h}\) are the mean and variance of herd × year effects, respectively. Both η and \({\sigma}^{2}_{h}\) were assumed unknown and hence, priors were specified as follows: \begin{eqnarray*}&&p({\eta})\ ~\ U[{-}10,\ 10]\\&&p(<F>{\sigma}^{2}_{h}</F>)\ ~\ U[0,1]\end{eqnarray*} where U [.] is the uniform distribution. The prior distribution for vector βR was: \[p({\beta}_{R}|{\sigma}^{2}_{R})\ ~\ N(\mathbf{0},\mathbf{I}{\sigma}^{2}_{R})\] with \({\sigma}^{2}_{R}\ =\ 10^{5}\) . As \({\sigma}^{2}_{R}\) is large relative to the residual variance, this prior distribution conveys vague prior knowledge about each of the elements of βR. For DC, the following prior distribution was assumed for the systematic effects: \[p({\beta}\mathbf{_{2}})\ ~\ N(0,10^{6})\] A normal distribution was used as the prior for the effect of service sire for the traits of CFI and CS: \[p(\mathbf{s}|{\sigma}^{2}_{s})\ ~\ N(0,\mathbf{I}{\sigma}^{2}_{s})\] The following uniform-bounded prior was assigned to the service sire variance for these traits: \[p({\sigma}^{2}_{s})\ ~\ U[0,1]\] A multivariate normal distribution was used as the prior for the animal effects: \[p(\mathbf{u}\ |\ \mathbf{A},\ \mathbf{G}_{0})\ ~\ N(\mathbf{0},\ \mathbf{A}\ {\otimes}\ \mathbf{G}_{0})\] where \(\mathbf{G}_{0}\ =\ \left[\begin{array}{ll}{\sigma}^{2}_{u1}&{\sigma}_{u12}\\{\sigma}_{u21}&{\sigma}^{2}_{u2}\end{array}\right]\) was the additive (co)variance matrix, and A was the additive relationship matrix between animals. A conjugate proper prior was assumed. The scaling factor for the prior of G0, shown below, was taken from the literature (Johnston et al., 2001). The degree of belief a priori was set to 5 conveying little weight to the prior information. \[\mathbf{G_{0}}\ =\ \left[\begin{array}{ll}25&{-}0.8\\&0.05\end{array}\right]\] The joint prior density had the form: \[\begin{array}{l}p(\mathbf{{\beta}_{1}},\ \mathbf{{\beta}_{R}},\ \mathbf{{\eta}},\ {\sigma}^{2}_{h},\ \mathbf{s},\ \mathbf{G_{0}},\ {\sigma}^{2}_{e2})\ =\\p(\mathbf{{\beta}_{1}})\ p(\mathbf{{\beta}_{R}})\ p(\mathbf{{\beta}_{H}}|\mathbf{{\eta}},{\sigma}^{2}_{h})\ p(\mathbf{{\eta}})\ p({\sigma}^{2}_{h})\\p(\mathbf{s}|{\sigma}^{2}_{s})\ p({\sigma}^{2}_{s})\ p(\mathbf{u}|\mathbf{G_{0}})\ p(\mathbf{G_{0}})\ p({\sigma}^{2}_{e2})\end{array}\] The joint posterior density is proportional to the product of the density of the conditional distribution of the observation × the joint prior density. Draws from the conditional posterior distribution of all the parameters were obtained using a Gibbs sampler with data augmentation (Sorensen et al., 1995). The joint posterior was augmented with the univariate normal liabilities for CS or CFI. After augmentation, all the fully conditional posterior distributions of model parameters can be derived as described by Albert and Chib (1993) and Sorensen et al. (1995). These distributions are normal for the systematic parameters (service sire and animal effects), truncated normal for the liabilities, scaled-inverted χ2 for the residual variance for DC, and scaled-inverted Wishart distributions for the dispersion parameters. Liabilities were sampled from their truncated normal distribution using inverse cumulative distribution function technique (Devroye, 1986). The posterior distribution was augmented with the unobserved calving dates corresponding to the censored DC observations as described in Donoghue et al. (2004a). (Co)variance components were estimated for both analyses, and genetic correlations of CS or CFI with DC were compared. Breeding values were predicted for all animals, and correlations between breeding values of CS or CFI and DC were compared for sires with progeny records. Results and Discussion For all analyses, convergence was assessed using methodology presented by Raftery and Lewis (1992). The required length of the burn-in period was always less then 5,000 iterations for all parameters. Thus, 200,000 iterations of the sampler were run with a conservative 50,000 iterations discarded as burn-in; the remaining 150,000 iterations were retained with thinning for post-Gibbs analysis. The mean and SD of DC for uncensored females was 302 ± 19 d, whereas the corresponding statistic for censored females was 354 ± 25 d. The number of days added to the largest observed DC record within a contemporary group ranged from 5 to 26 d. The majority of censored animals (63%) received a record ranging from 10 to 16 d greater than the largest observed DC record within their contemporary group, whereas 21, 14, and 2% of censored females received records with 5 to 9, 17 to 21, and 21 to 26 d added, respectively. Summaries of the posterior distributions of (co)variance components, heritabilities, and genetic correlation from the CFI-DC bivariate analysis are presented in Table 2. The posterior mean (SD) of the additive covariance between CFI and DC was −0.62 d (0.19 d), and the corresponding genetic correlation was −0.66 (0.12). These results suggest a high, negative correlation between probability of CFI and DC. A large, negative correlation indicates that, genetically, cows with a higher probability of calving to their first insemination will also record a shorter DC interval. Thus, selection for increased probability of CFI would result in correlated decreases in DC interval. This value is similar to that reported by Johnston et al. (2001) for the genetic correlation between DC and calving success (−0.66). Morris et al. (2000) also found a similar relationship between higher pregnancy rates and early calving dates in Angus heifers, but did not report the magnitude of this association. Table 2. Means, standard deviations, and bounds of high posterior density (HPD) intervals (95%) of the posterior distribution of (co)variance components, heritabilities, and correlations for bivariate analysis of calving to first insemination (CFI) and days to calving (DC) Trait Parametera Mean SD HPD (95%) CFIb \({\sigma}^{2}_{h}\) 0.06 0.06 0.01 to 0.19 \({\sigma}^{2}_{s}\) 0.14 0.16 0.01 to 0.49 \({\sigma}^{2}_{u}\) 0.04 0.01 0.02 to 0.07 h2 0.03 0.01 0.01 to 0.05 DCc \({\sigma}^{2}_{e}\) , d2 363.2 4.8 353.6 to 372.4 \({\sigma}^{2}_{u}\) , d2 23.5 4.1 16.5 to 32.6 h2 0.06 0.01 0.04 to 0.08 CFI-DC σu12, d −0.62 0.19 −1.04 to −0.31 rg −0.66 0.12 −0.89 to −0.44 Trait Parametera Mean SD HPD (95%) CFIb \({\sigma}^{2}_{h}\) 0.06 0.06 0.01 to 0.19 \({\sigma}^{2}_{s}\) 0.14 0.16 0.01 to 0.49 \({\sigma}^{2}_{u}\) 0.04 0.01 0.02 to 0.07 h2 0.03 0.01 0.01 to 0.05 DCc \({\sigma}^{2}_{e}\) , d2 363.2 4.8 353.6 to 372.4 \({\sigma}^{2}_{u}\) , d2 23.5 4.1 16.5 to 32.6 h2 0.06 0.01 0.04 to 0.08 CFI-DC σu12, d −0.62 0.19 −1.04 to −0.31 rg −0.66 0.12 −0.89 to −0.44 a \({\sigma}^{2}_{h}\) = herd-year variance; \({\sigma}^{2}_{s}\) = service sire variance; \({\sigma}^{2}_{u}\) = additive variance; \({\sigma}^{2}_{e}\) = residual variance; σu12 = additive covariance between the two traits; rg = genetic correlation. b \(h^{2}\ =\ {\sigma}^{2}_{u}/({\sigma}^{2}_{h}\ +\ {\sigma}^{2}_{s}\ +\ {\sigma}^{2}_{u}\ +\ 1)\) . c \(h^{2}\ =\ {\sigma}^{2}_{u}/({\sigma}^{2}_{u}\ +\ {\sigma}^{2}_{e})\) . View Large Table 2. Means, standard deviations, and bounds of high posterior density (HPD) intervals (95%) of the posterior distribution of (co)variance components, heritabilities, and correlations for bivariate analysis of calving to first insemination (CFI) and days to calving (DC) Trait Parametera Mean SD HPD (95%) CFIb \({\sigma}^{2}_{h}\) 0.06 0.06 0.01 to 0.19 \({\sigma}^{2}_{s}\) 0.14 0.16 0.01 to 0.49 \({\sigma}^{2}_{u}\) 0.04 0.01 0.02 to 0.07 h2 0.03 0.01 0.01 to 0.05 DCc \({\sigma}^{2}_{e}\) , d2 363.2 4.8 353.6 to 372.4 \({\sigma}^{2}_{u}\) , d2 23.5 4.1 16.5 to 32.6 h2 0.06 0.01 0.04 to 0.08 CFI-DC σu12, d −0.62 0.19 −1.04 to −0.31 rg −0.66 0.12 −0.89 to −0.44 Trait Parametera Mean SD HPD (95%) CFIb \({\sigma}^{2}_{h}\) 0.06 0.06 0.01 to 0.19 \({\sigma}^{2}_{s}\) 0.14 0.16 0.01 to 0.49 \({\sigma}^{2}_{u}\) 0.04 0.01 0.02 to 0.07 h2 0.03 0.01 0.01 to 0.05 DCc \({\sigma}^{2}_{e}\) , d2 363.2 4.8 353.6 to 372.4 \({\sigma}^{2}_{u}\) , d2 23.5 4.1 16.5 to 32.6 h2 0.06 0.01 0.04 to 0.08 CFI-DC σu12, d −0.62 0.19 −1.04 to −0.31 rg −0.66 0.12 −0.89 to −0.44 a \({\sigma}^{2}_{h}\) = herd-year variance; \({\sigma}^{2}_{s}\) = service sire variance; \({\sigma}^{2}_{u}\) = additive variance; \({\sigma}^{2}_{e}\) = residual variance; σu12 = additive covariance between the two traits; rg = genetic correlation. b \(h^{2}\ =\ {\sigma}^{2}_{u}/({\sigma}^{2}_{h}\ +\ {\sigma}^{2}_{s}\ +\ {\sigma}^{2}_{u}\ +\ 1)\) . c \(h^{2}\ =\ {\sigma}^{2}_{u}/({\sigma}^{2}_{u}\ +\ {\sigma}^{2}_{e})\) . View Large The posterior means (SD) for additive variance and heritability for DC were similar to previous estimates reported for this population (Donoghue et al., 2004a); furthermore, both estimates were within the high posterior density (HPD) (95%) intervals presented in the earlier study. The point estimates of heritability were smaller than estimates reported in the literature for DC or the equivalent trait of calving date for analyses when censored records are included (Johnston and Bunter, 1996; Morris et al., 2000; Johnston et al., 2001); however, our estimates were within the range of standard error associated with these literature estimates. The point estimate of residual variance for DC was similar to estimates previously reported for this population (Donoghue et al., 2004a). In the current study, the effect of service sire was included as part of the contemporary group definition to align with current industry practices for evaluation of DC. An additional analysis was undertaken with the random effect of service sire included in the model. The results from this analysis (not shown) were very similar to the results obtained in the current study. Fitting service sire as an additional random effect allowed for the formation of larger contemporary groups and had little effect on the estimation of additive variance or on heritability of the trait. Thus, including service sire as a random effect may be a reasonable option to avoid small contemporary groups. The posterior means for additive variance and heritability for CFI were similar to previous estimates reported for this population (Donoghue et al., 2004b); furthermore, both estimates were within the HPD (95%) intervals presented in the earlier study. The point estimate of h2 in the current study was lower than estimates reported in literature for heifer fertility; however, it is within the 90% CI reported by Evans et al. (1999) for the trait of heifer pregnancy. The low estimates of h2 observed in this study could result from more appropriate analytical procedures for data analysis, Bayesian approach vs. Method R for a small dataset, or perhaps are a reflection of the slight differences in trait definitions between CFI and heifer pregnancy. The posterior mean of herd-year variance for CFI in this study was much smaller than reported in an earlier study for the same population: 0.06 vs. 0.84 (Donoghue et al., 2004b). The variable nature of this parameter is most likely caused by a high incidence of ECP in the AI data; 50% of herd-year classes contained observations that fell into the same category for CFI. Possible reasons for this high incidence of ECP in the AI data were discussed in the earlier study, and included incomplete data recording and implementation of different management levels under AI matings. The highly fluctuating nature of this parameter suggests that caution should be used in the interpretation of the herd-year variance in the presence of ECP. Despite the large difference in this parameter observed between this study and the earlier study for CFI, additive variance and heritability point estimates were very similar in both studies, indicating that estimation of these parameters appears stable. The posterior mean of service sire variance for CFI was larger in magnitude than both herd-year and additive variances. The point estimate in the current study was higher than the estimate reported in the earlier study for this population and was outside the HPD (95%) interval for that study. However, examination of the SD and HPD (95%) interval associated with this point estimate indicates a lack of statistical evidence proving it was significantly different from zero. This result was most likely a reflection of the small number of service sires (n = 687) represented in the AI data. Summaries of the posterior distributions of (co)variance components, heritabilities, and genetic correlation from the CS-DC bivariate analysis are presented in Table 3. The posterior mean (SD) of the additive covariance between CS and DC was −0.70 d (0.14 d), and the corresponding genetic correlation was −0.73 (0.06). These results suggest a high, negative correlation between probability of CS and DC. A large, negative correlation indicates that, genetically, females with a higher probability of calving success will also record a shorter DC interval. Thus, selection for increased probability of CS would result in correlated decreases in DC interval. This value is similar in magnitude to that reported by Johnston et al. (2001) for the genetic correlation between the same traits (−0.66), and is slightly larger than the genetic correlation between CFI and DC (−0.66) in the current study. However, the SD and HPD (95%) intervals associated with the point estimates of genetic correlations between CS or CFI and DC in the current study indicate that they were not significantly different from each other. Table 3. Means, standard deviations, and bounds of high posterior density (HPD) intervals (95%) of the posterior distribution of (co)variance components, heritabilities, and correlations for bivariate analysis of calving success (CS) and days to calving (DC) Trait Parametera Mean SD HPD (95%) CSb \({\sigma}^{2}_{h}\) 0.07 0.07 0.06 to 0.19 \({\sigma}^{2}_{s}\) 0.14 0.16 0.004 to 0.44 \({\sigma}^{2}_{u}\) 0.04 0.01 0.01 to 0.06 h2 0.03 0.01 0.01 to 0.05 DCc \({\sigma}^{2}_{e}\) , d2 363.2 4.8 353.8 to 372.6 \({\sigma}^{2}_{u}\) , d2 23.5 4.1 15.5 to 31.5 h2 0.06 0.01 0.04 to 0.08 CS-DC σu12, d −0.70 0.14 −1.06 to −0.29 rg −0.73 0.06 −0.91 to −0.45 Trait Parametera Mean SD HPD (95%) CSb \({\sigma}^{2}_{h}\) 0.07 0.07 0.06 to 0.19 \({\sigma}^{2}_{s}\) 0.14 0.16 0.004 to 0.44 \({\sigma}^{2}_{u}\) 0.04 0.01 0.01 to 0.06 h2 0.03 0.01 0.01 to 0.05 DCc \({\sigma}^{2}_{e}\) , d2 363.2 4.8 353.8 to 372.6 \({\sigma}^{2}_{u}\) , d2 23.5 4.1 15.5 to 31.5 h2 0.06 0.01 0.04 to 0.08 CS-DC σu12, d −0.70 0.14 −1.06 to −0.29 rg −0.73 0.06 −0.91 to −0.45 a \({\sigma}^{2}_{h}\) = herd-year variance; \({\sigma}^{2}_{s}\) = service sire variance; \({\sigma}^{2}_{u}\) = additive variance; \({\sigma}^{2}_{e}\) = residual variance; σu12 = additive covariance between the two traits; rg = genetic correlation. b \(h^{2}\ =\ {\sigma}^{2}_{u}/({\sigma}^{2}_{h}\ +\ {\sigma}^{2}_{s}\ +\ {\sigma}^{2}_{u}\ +\ 1)\) . c \(h^{2}\ =\ {\sigma}^{2}_{u}/({\sigma}^{2}_{u}\ +\ {\sigma}^{2}_{e})\) . View Large Table 3. Means, standard deviations, and bounds of high posterior density (HPD) intervals (95%) of the posterior distribution of (co)variance components, heritabilities, and correlations for bivariate analysis of calving success (CS) and days to calving (DC) Trait Parametera Mean SD HPD (95%) CSb \({\sigma}^{2}_{h}\) 0.07 0.07 0.06 to 0.19 \({\sigma}^{2}_{s}\) 0.14 0.16 0.004 to 0.44 \({\sigma}^{2}_{u}\) 0.04 0.01 0.01 to 0.06 h2 0.03 0.01 0.01 to 0.05 DCc \({\sigma}^{2}_{e}\) , d2 363.2 4.8 353.8 to 372.6 \({\sigma}^{2}_{u}\) , d2 23.5 4.1 15.5 to 31.5 h2 0.06 0.01 0.04 to 0.08 CS-DC σu12, d −0.70 0.14 −1.06 to −0.29 rg −0.73 0.06 −0.91 to −0.45 Trait Parametera Mean SD HPD (95%) CSb \({\sigma}^{2}_{h}\) 0.07 0.07 0.06 to 0.19 \({\sigma}^{2}_{s}\) 0.14 0.16 0.004 to 0.44 \({\sigma}^{2}_{u}\) 0.04 0.01 0.01 to 0.06 h2 0.03 0.01 0.01 to 0.05 DCc \({\sigma}^{2}_{e}\) , d2 363.2 4.8 353.8 to 372.6 \({\sigma}^{2}_{u}\) , d2 23.5 4.1 15.5 to 31.5 h2 0.06 0.01 0.04 to 0.08 CS-DC σu12, d −0.70 0.14 −1.06 to −0.29 rg −0.73 0.06 −0.91 to −0.45 a \({\sigma}^{2}_{h}\) = herd-year variance; \({\sigma}^{2}_{s}\) = service sire variance; \({\sigma}^{2}_{u}\) = additive variance; \({\sigma}^{2}_{e}\) = residual variance; σu12 = additive covariance between the two traits; rg = genetic correlation. b \(h^{2}\ =\ {\sigma}^{2}_{u}/({\sigma}^{2}_{h}\ +\ {\sigma}^{2}_{s}\ +\ {\sigma}^{2}_{u}\ +\ 1)\) . c \(h^{2}\ =\ {\sigma}^{2}_{u}/({\sigma}^{2}_{u}\ +\ {\sigma}^{2}_{e})\) . View Large The posterior means for additive and residual variances for DC were very similar to estimates for DC under the DC-CFI bivariate. Furthermore, all estimates were within the high posterior density (HPD) (95%) intervals for this analysis. The posterior means for additive variance and heritability for CS in this study were low in magnitude. The point estimate of h2 for CS is very similar to the estimate for the same trait reported by Johnston et al. (2001) when the categorical nature of the trait was taken into account. Higher estimates of h2 for CS have been reported in the literature when the trait was analyzed without adjustment for the categorical nature (Johnston and Bunter, 1996). The point estimates for all parameters for CS were very close to the corresponding estimates for CFI, with all estimates within the HPD (95%) intervals for CFI. These results may reflect the similar definitions of the two traits and the fact that only a small number of animals changed from a trait value of 0 under CFI to a trait value of 1 under CS (n = 2,219). The level of ECP was high under both trait definitions; 50 and 74% of herd-year classes had observations that fell into the same category for CFI and CS, respectively. The similarity of results observed in this study between CFI and CS may not hold for a dataset when complete recording is available (i.e., both successful and unsuccessful AI matings are reported, as well as information regarding NS matings that may follow). The incidence of CFI would be expected to be lower under complete recording, with a lower incidence of ECP. There were 183 sires with more than 10 daughters with CS or CFI records, as well as more than 10 daughters with DC records. The mean (SD) DC breeding value (BV) for these sires was 0.30 d (2.57 d) with a range from −5.60 to 8.67 d. These BV were predicted using parameters from the DC-CFI analysis and are not reported for DC BV under the DC-CS analysis because of the similarity of the parameters. The mean (SD) CS BV was −0.02% (0.08), ranging from −0.28 to 0.17%. The mean (SD) CFI BV was −0.01% (0.08), ranging from −0.25 to 0.16%. The correlations between DC and CS BV and DC and CFI BV for these sires were −0.993 and −0.997, respectively. These results indicate that sires whose daughters have either an increased probability of calving success or an increased probability of calving to first insemination also produce daughters with shorter DC records. The regression coefficient of CFI on DC was −0.03% successful calving to first insemination per day; for every 1 d decrease in DC BV, there is a 0.03% increase in CFI BV, and a similar result was observed for the regression coefficient between CS and DC. Implications The potential of calving to first insemination as a measure of fertility in artificial insemination data has been confirmed. Parameter estimates for this trait were close to those for calving success. The former trait allows the identification of animals that not only record a calving event, but also calve to their first insemination, and the value would be even greater in a dataset more complete than that used in this study. The genetic correlations reported between days to calving and both measures of fertility in artificial insemination data indicate a strong, negative relationship. Selecting for animals with shorter days to calving intervals genetically will lead to correlated increases in both calving success and calving to first insemination. The magnitude of these correlations suggest that it may be possible to use a multitrait approach to the evaluation of artificial insemination and natural service data, but report one genetic value that could be used for selection purposes. Literature Cited Albert, J., and S. Chib 1993. Bayesian analysis of binary and polychotomous response data. J. Am. Statist. Assoc. 88: 669– 679. Google Scholar CrossRef Search ADS Buddenberg, B. J., C. J. Brown, and A. H. Brown 1990. Heritability estimates of calving date in Hereford cattle maintained on range under natural mating. J. Anim. Sci. 68: 70– 74. Google Scholar CrossRef Search ADS PubMed Cox, D. R., and E. J. Snell 1989. Analysis of binary data. CRC Press, London, U.K. Devroye, L. 1986. Non-Uniform Random Variate Generation. Springer-Verlag, New York, NY. Google Scholar CrossRef Search ADS Donoghue, K. A., R. Rekaya, and J. K. Bertrand 2004a. Comparison of methods for handling censored records in beef fertility data: Field data. J. Anim. Sci. 82: 357– 361. Google Scholar CrossRef Search ADS Donoghue, K. A., R. Rekaya, J. K. Bertrand, and I. Misztal 2004b. Genetic evaluation of calving to first insemination using natural and artificial insemination mating data. J. Anim. Sci. 82: 391– 396. Evans, J. L., B. L. Golden, R. M. Bourdon, and K. L. Long 1999. Additive genetic relationships between heifer pregnancy and scrotal circumference in Hereford cattle. J. Anim. Sci. 77: 2621– 2628. Google Scholar CrossRef Search ADS PubMed Foulley, J. L., D. Gianola, and R. Thompson 1983. Prediction of genetic merit from data on binary and quantitative variates with an application to calving difficulty, birth weight, and pelvic opening. Genet. Sel. Evol. 15: 401– 424. Google Scholar CrossRef Search ADS PubMed Johnston, D. J., and K. L. Bunter 1996. Days to calving in Angus cattle: Genetic and environmental effects, and covariances with other traits. Livest. Prod. Sci. 45: 13– 22. Google Scholar CrossRef Search ADS Johnston, D. J., J. M. Henshall, and B. Tier 2001. Estimate of the genetic correlation between calving success and days to calving in Angus females. Pages 353–355 in Proc. 14th Conf. Assoc. Advmt. Anim. Breed. Genet., Queenstown, New Zealand. Kadarmideen, H. N., and M. P. Coffey 2001. Quality and validation of insemination data for national genetic evaluations for dairy cow fertility in the United Kingdom. Interbull Bulletin 27: 133– 138. Misztal, I., D. Gianola, and J. L. Foulley 1989. Computing aspects of nonlinear methods of sire evaluation for categorical data. J. Dairy Sci. 72: 1557– 1568. Google Scholar CrossRef Search ADS Morris, C. A., J. A. Wilson, G. L. Bennett, N. G. Cullen, S. M. Hickey, and J. C. Hunter 2000. Genetic parameters for growth, puberty, and beef cow reproductive traits in a puberty selection experiment. N. Z. J. Agric. Res. 43: 83– 91. Google Scholar CrossRef Search ADS Raftery, A. E., and S. Lewis 1992. How many iterations in the Gibbs sampler? Pages 763–773 in Bayesian Statistics 4. J. M. Bernando, J. O. Berger, A. P. Dawid, and A. F. M. Smith ed. Oxford Univ. Press, New York, NY. Rekaya, R., K. A. Weigel, D. Gianola, B. Heringstad, and G. Klemetsdal 2000. Methods for attenuating bias of variance component estimates in threshold models when herds are small. J. Dairy Sci. 83(Suppl. 1): 56– 57. (Abstr.) Google Scholar CrossRef Search ADS Schneeberger, M., B. Tier, and K. Hammond 1991. Introducing the third generation of BREEDPLAN and GROUP BREEDPLAN. Pages 194–199 in Proc. 9th Conf. Aust. Assoc. Anim. Breeding Genetics, Melbourne, Australia. Sorensen, D. A., S. Andersen, D. Gianola, and I. Korsgaard 1995. Bayesian inference in threshold using Gibbs sampling. Genet. Sel. Evol. 27: 229– 249. Google Scholar CrossRef Search ADS Footnotes 1 Appreciation is extended to the Angus Society of Australia for providing the data; Meat and Livestock Australia for the research scholarship provided to the first author, and to D. J. Johnston and C. Teseling for their contributions. Copyright 2004 Journal of Animal Science
Genetic correlation between serum insulin-like growth factor-1 concentration and performance and meat quality traits in Duroc pigsSuzuki, K.;Nakagawa, M.;Katoh, K.;Kadowaki, H.;Shibata, T.;Uchida, H.;Obara, Y.;Nishida, A.
doi: 10.2527/2004.824994xpmid: 15080319
Abstract This study was intended to examine whether serum IGF-I concentration is appropriate for use as a physiological predictor for genetic improvement of meat production and meat quality traits in pigs. Heritabilities and genetic correlations were estimated for these traits. The Duroc breed used in this study was selected for seven generations for average daily BW gain (DG) from 30 to 105 kg of BW, loin-eye muscle area (EM), backfat thickness (BF), and intramuscular fat (IMF) content. Serum IGF-I concentration of boars and gilts at the fourth generation of selection and that of boars, gilts, and barrows from the fifth to seventh generations of selection were measured at 8 wk (IGFI-8W) for 832 animals and again at the time they reached 105 kg of BW (IGFI-105KG) for 834 animals. A multivariate REML procedure was used to estimate genetic parameters with a model incorporating generation of selection, sex, common environmental effect of litter, and individual additive genetic effects. Heritability estimates for IGFI-8W and IGFI-105KG were 0.23 ± 0.02 and 0.26 ± 0.03, respectively. The estimates of common environmental effect for IGFI-8W and IGFI-105KG were 0.20 ± 0.02 and 0.03 ± 0.01, respectively. Positive genetic correlations were estimated between IGFI-8W and DG (0.26 ± 0.08), EM (0.22 ± 0.10), and IMF (0.32 ± 0.10). Moreover, the positive genetic correlation between IGFI-105KG and EM was 0.42 ± 0.08. These results indicate that serum IGF-I concentration at an early stage of growth was effective for prediction of IMF, but it was not a reliable physiological predictor of genetic merit of meat production traits. Introduction Bioactive substances in blood, such as IGF-I, would be useful as selection indexes of production traits because it is easy to collect serum samples from live animals. Furthermore, concentrations measured in young animals would be especially useful for selection if they could predict future performance of animals. Insulin-like growth factor-I is secreted mainly from liver and is stimulated by GH. The main action of IGF-I is mediating GH function. It facilitates cartilage ossification and growth promotion. Use of IGF-I as a physiological criterion for genetic animal improvement is possible because IGF-I concentration increases steadily during animal growth, in contrast to the large circadian variation of GH (Scanes et al., 1987). Low-realized heritabilities of 0.15 and 0.10 for IGF-I concentration were estimated in direct selection for seven (Blair et al., 1989) and five generations (Baker et al., 1991), respectively, in mice. Those studies also reported a positive genetic correlation between BW and serum IGF-I concentration in mice. Bunter et al. (2002) reported that heritabilities for serum IGF-I concentration were in the range of 0.20 to 0.58, and that IGF-I concentration is genetically correlated with backfat depth and feed conversion ratio in pigs. Conversely, negative genetic correlations between IGF-I concentration and growth (Davis and Simmen, 1997) and backfat thickness (BF; Davis and Simmen, 2000) in beef cattle have been estimated. Notwithstanding, little is known regarding genetic correlations between IGF-I concentration and meat production traits. Therefore, this study was intended to estimate heritability of IGF-I and genetic correlations between serum IGF-I concentration and meat production traits. It also investigated whether serum IGF-I concentration is effective as a physiological criterion of selection for higher growth rate using Duroc pigs selected for meat production traits. Materials and Methods Animals and Performance Testing Procedures Duroc pigs used in this experiment were of a line selected for seven generations at the Miyagi Prefecture Animal Industry Experiment Station from 1995 to 2001 (Suzuki et al., 2002). Selection criteria traits were daily gain from 30 to 105 kg of BW (DG), loin-eye muscle area (EM), BF at 105 kg of BW measured by ultrasound technology, and intramuscular fat content (IMF) measured on slaughtered sib pigs. Average population size of each generation was 14 boars and 42 gilts. Gilts farrowed only once, and boars were retained for one 4- to 6-wk breeding period; therefore, a new generation was obtained each year. Pigs were weaned at 4 wk. At 8 wk, one to two male piglets (total 50 piglets) and two to four female piglets (total 100 piglets) from each litter were selected as candidates for boars and gilts based on BW at 8 wk (BW8W). At the same time, approximately 80 piglets in total, comprising mainly boars and sometimes gilts from each litter, were selected for full-sib testing in each generation. This first stage of selection was conducted within litter. Each pig's blood was collected at the first stage of selection and BF was measured at the half body point and 2 cm away from midline by an A-mode ultrasound machine (EPOCH 2002, Panametrics Japan Co., Ltd., Tokyo, Japan) at 8 wk (BF8W). Boars for full-sib testing were subsequently castrated. The performance tests began when BW reached 30 kg and ended at 105 kg. Therefore, DG was from 30 to 105 kg of BW. Backfat thickness and EM were measured on 105-kg animals on the left side at half body length using an ultrasound (B-mode) color-scanning scope (SR-100; Kaijo Corp., Tokyo, Japan). Computer software determined EM. Feed intake was measured with boars reared individually, and feed conversion ratio (FCR) was calculated from 30 to 105 kg of BW. Blood samples also were collected from candidate boars and gilts and from two full sibs, mainly castrated full brothers at 105 kg of BW without fasting. Pigs were provided ad libitum access to a commercial diet (15% CP, 78% total digestible nutrition, 0.76 % lysine content, DM basis) during the testing periods from 30 to 105 kg of live weight. Pigs had free access to water. Boars were reared individually in performance-testing pens. Gilts and barrows were reared in growing pens and group fed in a concrete-floored building with eight pigs per pen, which provided floor space of 1.2 m2 per pig. Selection Method The objectives of this selection were to produce an excellent Duroc line for use as terminal sires in meat production with high meat quality traits, and to supply these Duroc boars to pig farmers as commercial terminal sires. Therefore, selection was conducted without a control line at the Miyagi Prefecture Animal Industry Experiment Station in Japan. The first and second generations of selection were performed by an index selection method based on relative desired gains (Yamada et al., 1975). Traits selected for were DG, EM, BF, and IMF. Genetic and phenotypic parameters used to derive the selection criteria were obtained from the performance test data of the first and second generation, respectively. The means of DG, EM, BF, and IMF at the first generation were 865 g, 36.1 cm2, 2.34 cm, and 4.3%, respectively. Relative desired gain was established as 135 g, 3.9 cm2, −0.54 cm, and 0.7% for DG, EM, BF and IMF, respectively. To avoid rapid disappearance of the base generation's genes from the population, selection was made within sires for boars and within litters for gilts at the first generation. Breeding values of DG, EM, BF, and IMF were estimated by multiple-trait, animal-model BLUP, from the third generation onward. The breeding values were calculated using the PEST3.1 program (Groeneveld and Kovac, 1990) after estimating genetic parameters by the VCE4 program (Newmaier and Groeneveld, 1998), with models of generation and sex as fixed effects and random effects of individual additive genetic effect and error. Relative economic weights of selection traits were calculated from the relative desired gain. The relative desired gains of DG, EM, BF, and IMF were established from the performance test data of the first generation as described before. The aggregate breeding values were calculated by multiplying the relative economic weights to the EBV of each trait, and the selection was executed. Approximately 15 boars and 50 gilts were selected at each generation. In each generation, inbreeding coefficients for individual pigs were computed. Based on inbreeding information, all matings were planned to minimize the rate of increase in inbreeding. Carcass Dissection and Meat Quality Measurement Pigs for full-sib testing (barrows and gilts) were slaughtered by manual low-voltage (200 V) electrical stunning 24 h after feed removal with free access to water. Carcasses were placed in a conventional chiller at 4°C for 24 h. Subsequently, for measuring meat quality in the LM, a 7- to 10-cm-long piece of the loin (two thoracic vertebrae sections above the last rib) was taken from the left half carcass of each pig. External loin adipose tissue was removed. Two pieces of meat, 2 × 2 × 5 cm each, were weighed and vacuum-packaged in polyethylene bags. They were then heated with a water bath of 70°C for 30 min. Then, after cooling at room temperature, two cooked pieces per animal were cut to 1 × 1 × 5 cm. We measured the tenderness (TS, kgf/cm2) with a Tensipresser (TTP-50BXII; Taketomo Electric Co., Tokyo, Japan) developed by Nakai et al. (1992). This machine was developed to accurately evaluate meat tenderness using an up and down motion to imitate chewing action. To determine IMF, two minced loin meat samples of approximately 20 g each were analyzed using the Soxhlet method. Measurement of Serum IGF-I Concentration At the first stage of selection at 8 wk of age and at the end of testing when all pigs reached 105 kg of BW, blood samples were collected from the jugular vein of pigs from the first to the seventh generation of selection. However, IGF-I was measured only for boars and gilts at the fourth generation and IGF-I of boars, gilts, and barrows was measured from the fifth to seventh generation. After centrifugal separation for 15 min at 4°C, the serum was preserved at −20°C until used for IGF-I assay. Serum samples were extracted by acid–ethanol extraction to dissociate IGF-I from binding proteins before assaying (Daughaday et al., 1980). Then IGF-I concentrations were assayed in duplicate by RIA following the double-antibody method (Kuhara et al., 1992). Intra- and interassay coefficients were 3.3 and 7.8%, respectively. Minimum detectable concentration was 20 pg/mL. Serum IGF-I concentration measured at 8 wk and that measured at 105 kg of BW were defined as IGFI-8W and IGFI-105KG, respectively. Table 1 shows the total number of IGF-I measurements. Table 1. Number of animals, means, standard deviation, heritability, common environmental effect (c2), and phenotypic standard deviation (σp) in Duroc pigs Traits No. Mean SD h2 ± SE c2 ± SE σp Daily gain, g/d 1,642 873.6 109.3 0.49 ± 0.03 0.04 ± 0.01 80.7 Loin muscle area, cm2 1,639 37.00 4.05 0.43 ± 0.03 0.02 ± 0.01 3.69 Backfat thickness, cm 1,642 2.37 0.43 0.73 ± 0.02 0.02 ± 0.01 0.39 Intramuscular fat, % 543 4.25 1.46 0.39 ± 0.03 0.08 ± 0.02 1.40 Tenderness, kgf/cm2 544 72.51 12.71 0.39 ± 0.04 0.08 ± 0.02 12.7 Body weight at 8 wk, kg 1,642 22.69 3.76 0.24 ± 0.02 0.36 ± 0.03 3.18 Backfat thickness at 8 wk, cm 1,639 1.02 0.20 0.41 ± 0.02 0.17 ± 0.02 1.66 Feed conversion ratio 379 2.65 0.17 0.35 ± 0.04 0.15 ± 0.04 0.17 IGFI-8W, ng/mLa 832 102.8 62.0 0.23 ± 0.03 0.20 ± 0.02 40.0 IGFI-105KG, ng/mLb 834 128.5 68.9 0.26 ± 0.03 0.03 ± 0.01 44.5 Traits No. Mean SD h2 ± SE c2 ± SE σp Daily gain, g/d 1,642 873.6 109.3 0.49 ± 0.03 0.04 ± 0.01 80.7 Loin muscle area, cm2 1,639 37.00 4.05 0.43 ± 0.03 0.02 ± 0.01 3.69 Backfat thickness, cm 1,642 2.37 0.43 0.73 ± 0.02 0.02 ± 0.01 0.39 Intramuscular fat, % 543 4.25 1.46 0.39 ± 0.03 0.08 ± 0.02 1.40 Tenderness, kgf/cm2 544 72.51 12.71 0.39 ± 0.04 0.08 ± 0.02 12.7 Body weight at 8 wk, kg 1,642 22.69 3.76 0.24 ± 0.02 0.36 ± 0.03 3.18 Backfat thickness at 8 wk, cm 1,639 1.02 0.20 0.41 ± 0.02 0.17 ± 0.02 1.66 Feed conversion ratio 379 2.65 0.17 0.35 ± 0.04 0.15 ± 0.04 0.17 IGFI-8W, ng/mLa 832 102.8 62.0 0.23 ± 0.03 0.20 ± 0.02 40.0 IGFI-105KG, ng/mLb 834 128.5 68.9 0.26 ± 0.03 0.03 ± 0.01 44.5 a IGF-I concentration at 8 wk. b IGF-I concentration at 105 kg BW. View Large Table 1. Number of animals, means, standard deviation, heritability, common environmental effect (c2), and phenotypic standard deviation (σp) in Duroc pigs Traits No. Mean SD h2 ± SE c2 ± SE σp Daily gain, g/d 1,642 873.6 109.3 0.49 ± 0.03 0.04 ± 0.01 80.7 Loin muscle area, cm2 1,639 37.00 4.05 0.43 ± 0.03 0.02 ± 0.01 3.69 Backfat thickness, cm 1,642 2.37 0.43 0.73 ± 0.02 0.02 ± 0.01 0.39 Intramuscular fat, % 543 4.25 1.46 0.39 ± 0.03 0.08 ± 0.02 1.40 Tenderness, kgf/cm2 544 72.51 12.71 0.39 ± 0.04 0.08 ± 0.02 12.7 Body weight at 8 wk, kg 1,642 22.69 3.76 0.24 ± 0.02 0.36 ± 0.03 3.18 Backfat thickness at 8 wk, cm 1,639 1.02 0.20 0.41 ± 0.02 0.17 ± 0.02 1.66 Feed conversion ratio 379 2.65 0.17 0.35 ± 0.04 0.15 ± 0.04 0.17 IGFI-8W, ng/mLa 832 102.8 62.0 0.23 ± 0.03 0.20 ± 0.02 40.0 IGFI-105KG, ng/mLb 834 128.5 68.9 0.26 ± 0.03 0.03 ± 0.01 44.5 Traits No. Mean SD h2 ± SE c2 ± SE σp Daily gain, g/d 1,642 873.6 109.3 0.49 ± 0.03 0.04 ± 0.01 80.7 Loin muscle area, cm2 1,639 37.00 4.05 0.43 ± 0.03 0.02 ± 0.01 3.69 Backfat thickness, cm 1,642 2.37 0.43 0.73 ± 0.02 0.02 ± 0.01 0.39 Intramuscular fat, % 543 4.25 1.46 0.39 ± 0.03 0.08 ± 0.02 1.40 Tenderness, kgf/cm2 544 72.51 12.71 0.39 ± 0.04 0.08 ± 0.02 12.7 Body weight at 8 wk, kg 1,642 22.69 3.76 0.24 ± 0.02 0.36 ± 0.03 3.18 Backfat thickness at 8 wk, cm 1,639 1.02 0.20 0.41 ± 0.02 0.17 ± 0.02 1.66 Feed conversion ratio 379 2.65 0.17 0.35 ± 0.04 0.15 ± 0.04 0.17 IGFI-8W, ng/mLa 832 102.8 62.0 0.23 ± 0.03 0.20 ± 0.02 40.0 IGFI-105KG, ng/mLb 834 128.5 68.9 0.26 ± 0.03 0.03 ± 0.01 44.5 a IGF-I concentration at 8 wk. b IGF-I concentration at 105 kg BW. View Large Statistical Analysis Performance data for DG, EM, BF, BW8W, BF8W, and FCR and meat quality data for IMF and TS, and IGFI-8W and IGFI-105KG were used for the final estimation of genetic parameters. Table 1 shows the number of pigs for each measurement. The following multiple trait animal model was used for analysis to estimate genetic parameters: \[\mathit{Y_{ijklm}}\ =\ {\mu}_{\mathit{i}}\ +\ \mathit{G_{ij}}\ +\ \mathit{S_{ik}}\ +\ \mathit{c_{il}}\ +\ \mathit{a_{im}}\ +\ \mathit{e_{ijklm}}\] where Yijklm = observation for traits i; μi = common constant for trait i; and Gij = fixed effect of selection generation j for trait i (this selection generation effect included the genetic effect of selection and the environmental effect at each generation); Sik = fixed effect of sex k for trait i; cil = random effect of common environment l of littermates for trait i; aim = random additive genetic effect of animal m for trait i; and eijklm = random residual effect for trait i. Seven generations of pedigree information of 1,642 animals with data and of 152 ancestors born before the fourth generation (total 1,794 animals) were included in this analysis. The VCE4.25 program (Neumaier and Groeneveld, 1998) was used to estimate (co)variance components and their respective standard error. Standard errors of heritability estimates and genetic correlations were also estimated with the VCE4.25 program. The GLM procedure of SAS (SAS Inst., Inc., Cary, NC) was used to obtain sex × generation least squares means of IGFI-8W and IGFI-105KG accounting for fixed effect of sex, generation, and sex × generation interaction and to test significance. Results Table 1 shows heritability estimates for meat production traits and serum IGF-I concentration. At the fourth generation of selection, IGF-I concentration was measured with boar and gilt candidates; from the fifth to seventh generations of selection, it was measured with these candidates and their two full sibs. Heritability estimates of IGFI-8W, IGFI-105KG, and BW8W were similar (0.23, 0.26, and 0.24, respectively). Table 1 also shows estimates of the proportion of the common environmental variance (c2). Its estimates for IGFI-8W and BW8W were higher than that for IGFI-105KG. Table 2 presents genetic and phenotypic correlations of IGF-I concentration (IGFI-8W and IGFI-105KG) with selection traits (DG, EM, BF, and IMF), and correlated traits (TS, BW8W, BF8W, and FCR). The genetic correlation between IGFI-8W and BW8W was moderate and those between IGFI-8W and DG, EM, IMF, BF8W, and FCR were low. In addition, genetic correlations of IGFI-8W with BF and TS were low. Phenotypic correlations between them were 0.1 or less, except for correlations between IGFI-8W and BW8W, and BF8W. Although IGFI-105KG showed moderately high genetic correlations with EM and TS, it had low genetic correlations with BF, BW8W, and BF8W. In addition, the estimated genetic correlation between IGFI-8W and IGFI-105 KG was high, but the estimated phenotypic correlation was low. Table 2. Genetic correlation (rG) and phenotypic correlation (rp) in Duroc pigs IGFI-8Wa IGFI-105KGb Traits rG ± SE rp rG ± SE rp Daily gain 0.26 ± 0.08 0.08 0.22 ± 0.07 0.05 Loin muscle area 0.22 ± 0.10 0.02 0.42 ± 0.08 0.19 Backfat thickness 0.13 ± 0.08 0.10 −0.02 ± 0.07 −0.11 Intramuscular fat 0.32 ± 0.10 0.04 0.26 ± 0.09 −0.17 Tenderness −0.05 ± 0.12 0.02 0.36 ± 0.10 0.30 Body weight at 8 wk 0.45 ± 0.08 0.40 0.09 ± 0.09 −0.08 Backfat thickness at 8wk 0.33 ± 0.06 0.28 0.00 ± 0.06 −0.09 Feed conversion ratio 0.20 ± 0.08 0.13 −0.17 ± 0.07 −0.21 IGF-I concentration at 8 wk — — 0.73 ± 0.08 0.15 IGFI-8Wa IGFI-105KGb Traits rG ± SE rp rG ± SE rp Daily gain 0.26 ± 0.08 0.08 0.22 ± 0.07 0.05 Loin muscle area 0.22 ± 0.10 0.02 0.42 ± 0.08 0.19 Backfat thickness 0.13 ± 0.08 0.10 −0.02 ± 0.07 −0.11 Intramuscular fat 0.32 ± 0.10 0.04 0.26 ± 0.09 −0.17 Tenderness −0.05 ± 0.12 0.02 0.36 ± 0.10 0.30 Body weight at 8 wk 0.45 ± 0.08 0.40 0.09 ± 0.09 −0.08 Backfat thickness at 8wk 0.33 ± 0.06 0.28 0.00 ± 0.06 −0.09 Feed conversion ratio 0.20 ± 0.08 0.13 −0.17 ± 0.07 −0.21 IGF-I concentration at 8 wk — — 0.73 ± 0.08 0.15 a IGF-I concentration at 8 wk of age. b IGF-I concentration at 105 kg of BW. View Large Table 2. Genetic correlation (rG) and phenotypic correlation (rp) in Duroc pigs IGFI-8Wa IGFI-105KGb Traits rG ± SE rp rG ± SE rp Daily gain 0.26 ± 0.08 0.08 0.22 ± 0.07 0.05 Loin muscle area 0.22 ± 0.10 0.02 0.42 ± 0.08 0.19 Backfat thickness 0.13 ± 0.08 0.10 −0.02 ± 0.07 −0.11 Intramuscular fat 0.32 ± 0.10 0.04 0.26 ± 0.09 −0.17 Tenderness −0.05 ± 0.12 0.02 0.36 ± 0.10 0.30 Body weight at 8 wk 0.45 ± 0.08 0.40 0.09 ± 0.09 −0.08 Backfat thickness at 8wk 0.33 ± 0.06 0.28 0.00 ± 0.06 −0.09 Feed conversion ratio 0.20 ± 0.08 0.13 −0.17 ± 0.07 −0.21 IGF-I concentration at 8 wk — — 0.73 ± 0.08 0.15 IGFI-8Wa IGFI-105KGb Traits rG ± SE rp rG ± SE rp Daily gain 0.26 ± 0.08 0.08 0.22 ± 0.07 0.05 Loin muscle area 0.22 ± 0.10 0.02 0.42 ± 0.08 0.19 Backfat thickness 0.13 ± 0.08 0.10 −0.02 ± 0.07 −0.11 Intramuscular fat 0.32 ± 0.10 0.04 0.26 ± 0.09 −0.17 Tenderness −0.05 ± 0.12 0.02 0.36 ± 0.10 0.30 Body weight at 8 wk 0.45 ± 0.08 0.40 0.09 ± 0.09 −0.08 Backfat thickness at 8wk 0.33 ± 0.06 0.28 0.00 ± 0.06 −0.09 Feed conversion ratio 0.20 ± 0.08 0.13 −0.17 ± 0.07 −0.21 IGF-I concentration at 8 wk — — 0.73 ± 0.08 0.15 a IGF-I concentration at 8 wk of age. b IGF-I concentration at 105 kg of BW. View Large Table 3 lists the least squares means of IGFI-8W and IGFI-105KG for the fourth to the seventh generations of selection and for the three sexes. Results of ANOVA in the IGFI-8W indicated that the effect of generation and sex was significant. The IGFI-8W concentration in boars and gilts increased (P < 0.01) from the fourth to the sixth generation, but the concentration at the seventh generation decreased significantly compared with the sixth generation. The average BW8W of the seventh generation decreased significantly from the previous generation (BW for these four generations were 24.1, 27.2, 26.1, and 23.7 kg in boars and 22.4, 25.7, 24.8, and 22.2 kg in gilts, respectively). Decreased IGFI-8W seems to be related to this decrease in growth. The mean IGFI-8W of boars was significantly higher than those of gilts for all generations except the fourth generation. Analysis of variance indicated that the effects of generation, sex, and generation × sex interaction were significant for the IGFI-105KG. This generation × sex interaction resulted from a high concentration (P < 0.01) of IGFI-105KG in seventh-generation boars. There were significant differences in mean IGFI-105KG among boars, gilts, and barrows. The serum IGFI-105KG increased (P < 0.01) compared with IGFI-8W. In addition, serum IGF-I concentration of gilts increased (P < 0.01) even though the rate of increase was less than that for boars. However, IGF-I concentration of barrows did not change with their growth at the fifth and seventh generations; also, it decreased (P < 0.01) at the sixth generation of selection. Table 3. Least squares means ± standard errors for serum IGF-I concentration among sex, growth stage, and generation of selection in Duroc pigs IGFI-8W, ng/mLa IGFI-105KG, ng/mLb Generation Boars Gilts Boarsc Boars Gilts Barrows 4 45.4 ± 6.2 44.9 ± 4.1 67.3 ± 6.4f 49.8 ± 4.7g 5 104.7 ± 5.7 75.3 ± 3.7e 90.1 ± 5.5d 159.3 ± 6.0f 101.1 ± 3.8g 86.2 ± 5.7h 6 158.9 ± 5.9d 129.3 ± 3.6e 133.6 ± 5.6e 184.4 ± 5.9f 132.5 ± 3.7g 98.2 ± 5.7h 7 126.6 ± 4.4d 107.5 ± 3.5e 117.3 ± 5.6de 256.4 ± 4.5f 146.4 ± 3.5g 116.2 ± 5.7h IGFI-8W, ng/mLa IGFI-105KG, ng/mLb Generation Boars Gilts Boarsc Boars Gilts Barrows 4 45.4 ± 6.2 44.9 ± 4.1 67.3 ± 6.4f 49.8 ± 4.7g 5 104.7 ± 5.7 75.3 ± 3.7e 90.1 ± 5.5d 159.3 ± 6.0f 101.1 ± 3.8g 86.2 ± 5.7h 6 158.9 ± 5.9d 129.3 ± 3.6e 133.6 ± 5.6e 184.4 ± 5.9f 132.5 ± 3.7g 98.2 ± 5.7h 7 126.6 ± 4.4d 107.5 ± 3.5e 117.3 ± 5.6de 256.4 ± 4.5f 146.4 ± 3.5g 116.2 ± 5.7h a IGF-I concentration at 8 wk of age. b IGF-I concentration at 105 kg BW. c These boars are selected as full sib test and they were castrated after blood collection. d,e Means with different superscripts within the same row in IGFI-8W differ by sex (P < 0.05). f,g,h Means with different superscripts within the same row in IGFI-105KG differ by sex (P < 0.05). View Large Table 3. Least squares means ± standard errors for serum IGF-I concentration among sex, growth stage, and generation of selection in Duroc pigs IGFI-8W, ng/mLa IGFI-105KG, ng/mLb Generation Boars Gilts Boarsc Boars Gilts Barrows 4 45.4 ± 6.2 44.9 ± 4.1 67.3 ± 6.4f 49.8 ± 4.7g 5 104.7 ± 5.7 75.3 ± 3.7e 90.1 ± 5.5d 159.3 ± 6.0f 101.1 ± 3.8g 86.2 ± 5.7h 6 158.9 ± 5.9d 129.3 ± 3.6e 133.6 ± 5.6e 184.4 ± 5.9f 132.5 ± 3.7g 98.2 ± 5.7h 7 126.6 ± 4.4d 107.5 ± 3.5e 117.3 ± 5.6de 256.4 ± 4.5f 146.4 ± 3.5g 116.2 ± 5.7h IGFI-8W, ng/mLa IGFI-105KG, ng/mLb Generation Boars Gilts Boarsc Boars Gilts Barrows 4 45.4 ± 6.2 44.9 ± 4.1 67.3 ± 6.4f 49.8 ± 4.7g 5 104.7 ± 5.7 75.3 ± 3.7e 90.1 ± 5.5d 159.3 ± 6.0f 101.1 ± 3.8g 86.2 ± 5.7h 6 158.9 ± 5.9d 129.3 ± 3.6e 133.6 ± 5.6e 184.4 ± 5.9f 132.5 ± 3.7g 98.2 ± 5.7h 7 126.6 ± 4.4d 107.5 ± 3.5e 117.3 ± 5.6de 256.4 ± 4.5f 146.4 ± 3.5g 116.2 ± 5.7h a IGF-I concentration at 8 wk of age. b IGF-I concentration at 105 kg BW. c These boars are selected as full sib test and they were castrated after blood collection. d,e Means with different superscripts within the same row in IGFI-8W differ by sex (P < 0.05). f,g,h Means with different superscripts within the same row in IGFI-105KG differ by sex (P < 0.05). View Large Discussion Present results suggest that the heritability for serum IGF-I concentration was low and the common environmental effect on the growth performance was large in early stage of growth in pig. Blair et al. (1989) and Baker et al. (1991) reported heritability estimates of 0.15 for serum IGF-I concentration at 6 wk of age, and of 0.10 for the serum IGF-I concentration at 12 wk in mice. Bunter et al. (2002) reviewed heritability estimates ranging from 0.20 to 0.58 and pooled estimates of 0.28 and 0.32, including and excluding estimates from restricted feeding data. Hermesch et al. (2001) estimated the heritability (0.24) and common environmental effect (0.13) for juvenile IGF-I concentration at 4 wk. In addition, Cameron et al. (2003) reported a high and significant common environment variance component (0.46) for serum IGF-I concentration at 6 wk. Cameron et al. (2003) pointed out that a genetic analysis of serum IGF-I concentration data, measured at an early age, should include a common environment effect in the model to avoid overestimating heritability. If the common environmental effect had not been included in the present analysis, the estimated heritability of IGFI-8W would have been higher (0.37). Heritability is inferred to be low judging from the present result and other reports. This research was mainly intended to examine whether the IGF-I concentration measured at an early stage of growth is effective as a physiological predictor of meat production traits at a later stage of growth. Therefore, we estimated genetic and phenotypic correlations between serum IGF-I concentration and meat production and meat quality traits. Results suggest that serum IGFI-8W was a useful index of growth and fat accumulation up to 8 wk because IGFI-8W showed a moderate, positive genetic and phenotypic correlation with BW8W and BF8W. However, genetic correlations of IGFI-8W with DG, FCR, EM, BF, and TENDERNESS were lower than those with BW8 and BF8W. The present estimates of genetic correlation with these traits are different in the magnitude and direction from previously reported estimates. For growth rate, Baker et al. (1991) reported a realized genetic correlation of 0.58 and a phenotypic correlation of 0.38 between IGF-I concentration and BW at 12 wk in mice. On the other hand, Hermesch et al. (2001) estimated a genetic correlation of 0.12 ± 0.12 between IGF-I measured at 4 wk and the lifetime ADG. Bunter et al. (2002) also reported a low genetic correlation of 0.15 between IGF-I concentration measured at around 4 to 5 wk and lifetime ADG in their review. It seems that serum IGF-I concentration at an early stage of growth is not an effective physiological predictor of genetic merit for production traits during performance test. A high genetic correlation of 0.50 between backfat thickness and serum IGF-I concentration was reported (Bunter et al., 2002). However, the genetic correlation between IGFI-8W and BF at 105 kg of BW was low in the present study. The result of administering serum IGF-I to Meishan pigs (Klindt et al., 1998) was a significant increase in backfat thickness. On the other hand, Buonomo and Klindt (1993) reported that selection for increased backfat thickness of the pig increased serum IGF-2 concentration. In addition, Owens et al. (1999) reported that IGF-I controls growth of lean meat and that IGF-2 controls adipose tissue growth. In most countries, the main objectives of pig improvement have been increasing lean growth rate and decreasing backfat thickness. In contrast, moderate thickness in the dressed carcass is important in Japan. Therefore, the BF of the Duroc breed used in the present study was considered to be thick (average 2.37 cm). Apparently, the genetic difference in the fat accumulation of the pig sample used may influence the magnitude of genetic correlation. Along with backfat thickness, high genetic correlations between IGF-I concentration and FCR have been reported in pig (Bunter et al., 2002) and beef cattle (Johnston et al., 2002). The present study estimated lower genetic and phenotypic correlations. A moderate genetic correlation of 0.32 between IGFI-8W and IMF was estimated. No reports have addressed the relation between serum IGF-I concentration and meat quality traits. If IMF could be estimated from serum IGF-I concentration, such information would be important and relatively easily obtained. Duroc pigs used in present study can accumulate more intramuscular fat than other breeds or lines. Therefore, it is necessary to confirm the relationship between IGF-I and IMF with other breeds showing low intramuscular fat accumulation. Compared with the genetic correlations of EM and TS with IGFI-8W measured at 8 wk, those with IGFI-105KG measured at an older age increased. These results suggest that serum IGF-I concentration is related to an increase in EM thickness and the amount of the lean meat. Serum IGF-I concentration of boars increased considerably from 8 wk to 105 kg of BW compared with gilts and barrows in every generation. That result suggests that the expression of IGF-I concentration is limited by other hormonal factors. Genetic correlations of BW8W, BF8W, BF at 105 kg, and FCR with IGFI-105KG were lower than with IGFI-8W. Moreover, the genetic correlations of DG and IMF with IGFI-105KG did not changed compared to those with IGFI-8W. These changes suggest that the physiological function of IGF-I depends on the animal's age. Cameron et al. (2003) also reported that the serum IGF-I concentration measured at 6 wk was positively related with ultrasonic backfat depth and the FCR, whereas serum IGF-I concentration measured at the end of test (90 kg of BW) was negatively correlated with BF and the FCR. Implications A positive genetic correlation of serum insulin-like growth factor I concentration at 8 wk of age in pigs with intramuscular fat was estimated. Furthermore, a moderate genetic correlation between serum insulin-like growth factor I concentration at 105 kg of body weight and loin muscle area also was estimated. These results suggest that serum insulin-like growth factor I concentration has some relation to growth traits and meat production traits. It may be useful as an indirect selection criterion in the early growth stage to evaluate future ability of intramuscular fat accumulation in pig breeding. Literature Cited Baker, R. L., A. J. Peterson, J. J. Bass, N. C. Amyes, B. H. Breier, and P. D. Gluckman 1991. Replicated selection for insulin-like growth factor-1 and body weight in mice. Theor. Appl. Genet. 81: 685– 692. Google Scholar CrossRef Search ADS PubMed Blair, H. T., S. N. McCutcheon, D. D. S. Mackenzie, P. D. Gluckman, J. E. Ormsby, and B. H. Brier 1989. Responses to divergent selection for plasma concentrations of insulin-like growth factor-1 in mice. Genet. Res. (Camb.) 53: 187– 191. Google Scholar CrossRef Search ADS Bunter, K., S. Hermesch, B. G. Luxford, K. Lahti, and E. Sutcliffe 2002. IGF-I concentration measured in juvenile pigs provides information for breeding programs: A mini review. Communication No. 03–09 in Proc. 7th World Cong. Genet. Appl. Livest. Prod., Montpellier, France. Buonomo, F. C., and J. Klindt 1993. Ontogeny of growth hormone (GH), insulin-like growth factors (IGF-I and IGF-II) and IGF binding protein-2 (IGFBP-2) in genetically lean and obese swine. Domest. Anim. Endocrinol. 10: 257– 265. Google Scholar CrossRef Search ADS PubMed Cameron, N. D., E. McCullough, K. Troup, and C. Penman 2003. Serum insulin-like growth factor-1 concentration in pigs divergently selelcted for daily food intake or lean growth rate. J. Anim. Breed. Genet. 120: 1– 10. Google Scholar CrossRef Search ADS Daughaday, W. H., I. K. Mariz, and S. L. Blethen 1980. Inhibition of access of bound somatomedin to membrane receptor and immunobinding sites: A comparison of radioreceptor and radioimmunoassay of somatomedin in native and acid-ethanol-extracted serum. J. Clin. Endocrinol. Metab. 51: 781– 788. Google Scholar CrossRef Search ADS PubMed Davis, M. E., and R. C. M. Simmen 1997. Genetic parameter estimates for serum insulin-like growth factor 1 concentration and performance traits in Angus beef cattle. J. Anim. Sci. 75: 317– 324. Google Scholar CrossRef Search ADS PubMed Davis, M. E., and R. C. M. Simmen 2000. Genetic parameter estimates for serum insulin-like growth factor 1 concentration and carcass traits in Angus beef cattle. J. Anim. Sci. 78: 2305– 2313. Google Scholar CrossRef Search ADS PubMed Groeneveld, E., and M. Kovac 1990. A generalized computing procedure for setting up and solving mixed linear models. J. Dairy Sci. 73: 513– 531. Google Scholar CrossRef Search ADS Hermesch, S., K. L. Bunter, and B. G. Luxford 2001. Estimates of genetic correlations between IGF-I recorded at 4 wk of age and individual piglet weights at birth and 14 days, along with lifetime growth rate and backfat. Pages 227–230 in Proc. Assoc. Advan. Anim. Breed. Genet., Queenstown, New Zealand. Johnston, D. J. R. M. Herd, M. J. Kadel, H-U. Graser, P. F. Arthur, and J. A. Archer 2002. Evidence of IGF-I as a genetic predictor of feed efficiency traits in beef cattle. Communication No. 10–16 in Proc. 7th World Cong. Genet. Appl. Livest. Prod., Montpellier, France. Klindt, J., J. T. Yen, F. C. Buonomo, A. J. Roberts, and T. Wise 1998. Growth, body composition, and endocrine responses to chronic administration of insulin-like growth factor and(or) porcine growth hormone in pigs. J. Anim. Sci. 76: 2368– 2381. Google Scholar CrossRef Search ADS PubMed Kuhara, T., K. Katoh, S. Oda, A. Ohneda, and Y. Sasaki 1992. Responses of metabolic hormones to prolonged intraduodenal infusion of amino acids in sheep. Anim. Sci. Technol. 63: 1123– 1133. Nakai, H., R. Tanabe, T. Ikeda, and M. Nishizawa 1992. Development of a technique for measuring tenderness in meat using a“Tensipresser.” Pages 947–950 in Proc. 38th Int. Cong. Meat Sci. Technol., Clermont-Ferrand, France. Neumaier, A., and E. Groeneveld 1998. Restricted maximum likelihood estimation of covariance in sparse linear models. Genet. Sel. Evol. 30: 3– 26. Google Scholar CrossRef Search ADS Owens, P. C., K. L. Gatford, P. E. Walton, W. Morley, and R. G. Campbell 1999. The relationship between endogenous insulin-like growth factors and growth in pigs. J. Anim. Sci. 77: 2098– 2103. Google Scholar CrossRef Search ADS PubMed Scanes, C. G., D. Lazarus, S. Bowen, F. C. Buonomo, and R. L. Gilbreath 1987. Postnatal changes in circulating concentrations of growth hormone, somatomedin C and thyroid hormones in pigs. Domest. Anim. Endocrinol. 4: 253– 257. Google Scholar CrossRef Search ADS PubMed Suzuki, K., H. Kadowaki, T. Shibata, H. Uchida, and Y. Sato 2002. Selection for daily gain, loin-eye area, backfat thickness and intramuscular fat in 7 generations of Duroc pigs. Communication No. 11–15 in Proc. 7th World Cong. Genet. Appl. Livest. Prod., Montpellier, France. Yamada, Y., K. Yokouchi, and A. Nishida 1975. Selection index when genetic gains of individual traits are of primary concern. Japan. J. Genetics. 50: 33– 41. Google Scholar CrossRef Search ADS Footnotes 1 We gratefully acknowledge A. F. Parlow for the IGF-I antibody provided by NIDDK, USA, and N. D. Cameron for comments on the manuscript. Copyright 2004 Journal of Animal Science
Genetic correlation between serum insulin-like growth factor-1 concentration and performance and meat quality traits in Duroc pigsSuzuki, K.;Nakagawa, M.;Katoh, K.;Kadowaki, H.;Shibata, T.;Uchida, H.;Obara, Y.;Nishida, A.
doi: 10.1093/ansci/82.4.994pmid: N/A
Abstract This study was intended to examine whether serum IGF-I concentration is appropriate for use as a physiological predictor for genetic improvement of meat production and meat quality traits in pigs. Heritabilities and genetic correlations were estimated for these traits. The Duroc breed used in this study was selected for seven generations for average daily BW gain (DG) from 30 to 105 kg of BW, loin-eye muscle area (EM), backfat thickness (BF), and intramuscular fat (IMF) content. Serum IGF-I concentration of boars and gilts at the fourth generation of selection and that of boars, gilts, and barrows from the fifth to seventh generations of selection were measured at 8 wk (IGFI-8W) for 832 animals and again at the time they reached 105 kg of BW (IGFI-105KG) for 834 animals. A multivariate REML procedure was used to estimate genetic parameters with a model incorporating generation of selection, sex, common environmental effect of litter, and individual additive genetic effects. Heritability estimates for IGFI-8W and IGFI-105KG were 0.23 ± 0.02 and 0.26 ± 0.03, respectively. The estimates of common environmental effect for IGFI-8W and IGFI-105KG were 0.20 ± 0.02 and 0.03 ± 0.01, respectively. Positive genetic correlations were estimated between IGFI-8W and DG (0.26 ± 0.08), EM (0.22 ± 0.10), and IMF (0.32 ± 0.10). Moreover, the positive genetic correlation between IGFI-105KG and EM was 0.42 ± 0.08. These results indicate that serum IGF-I concentration at an early stage of growth was effective for prediction of IMF, but it was not a reliable physiological predictor of genetic merit of meat production traits. Introduction Bioactive substances in blood, such as IGF-I, would be useful as selection indexes of production traits because it is easy to collect serum samples from live animals. Furthermore, concentrations measured in young animals would be especially useful for selection if they could predict future performance of animals. Insulin-like growth factor-I is secreted mainly from liver and is stimulated by GH. The main action of IGF-I is mediating GH function. It facilitates cartilage ossification and growth promotion. Use of IGF-I as a physiological criterion for genetic animal improvement is possible because IGF-I concentration increases steadily during animal growth, in contrast to the large circadian variation of GH (Scanes et al., 1987). Low-realized heritabilities of 0.15 and 0.10 for IGF-I concentration were estimated in direct selection for seven (Blair et al., 1989) and five generations (Baker et al., 1991), respectively, in mice. Those studies also reported a positive genetic correlation between BW and serum IGF-I concentration in mice. Bunter et al. (2002) reported that heritabilities for serum IGF-I concentration were in the range of 0.20 to 0.58, and that IGF-I concentration is genetically correlated with backfat depth and feed conversion ratio in pigs. Conversely, negative genetic correlations between IGF-I concentration and growth (Davis and Simmen, 1997) and backfat thickness (BF; Davis and Simmen, 2000) in beef cattle have been estimated. Notwithstanding, little is known regarding genetic correlations between IGF-I concentration and meat production traits. Therefore, this study was intended to estimate heritability of IGF-I and genetic correlations between serum IGF-I concentration and meat production traits. It also investigated whether serum IGF-I concentration is effective as a physiological criterion of selection for higher growth rate using Duroc pigs selected for meat production traits. Materials and Methods Animals and Performance Testing Procedures Duroc pigs used in this experiment were of a line selected for seven generations at the Miyagi Prefecture Animal Industry Experiment Station from 1995 to 2001 (Suzuki et al., 2002). Selection criteria traits were daily gain from 30 to 105 kg of BW (DG), loin-eye muscle area (EM), BF at 105 kg of BW measured by ultrasound technology, and intramuscular fat content (IMF) measured on slaughtered sib pigs. Average population size of each generation was 14 boars and 42 gilts. Gilts farrowed only once, and boars were retained for one 4- to 6-wk breeding period; therefore, a new generation was obtained each year. Pigs were weaned at 4 wk. At 8 wk, one to two male piglets (total 50 piglets) and two to four female piglets (total 100 piglets) from each litter were selected as candidates for boars and gilts based on BW at 8 wk (BW8W). At the same time, approximately 80 piglets in total, comprising mainly boars and sometimes gilts from each litter, were selected for full-sib testing in each generation. This first stage of selection was conducted within litter. Each pig's blood was collected at the first stage of selection and BF was measured at the half body point and 2 cm away from midline by an A-mode ultrasound machine (EPOCH 2002, Panametrics Japan Co., Ltd., Tokyo, Japan) at 8 wk (BF8W). Boars for full-sib testing were subsequently castrated. The performance tests began when BW reached 30 kg and ended at 105 kg. Therefore, DG was from 30 to 105 kg of BW. Backfat thickness and EM were measured on 105-kg animals on the left side at half body length using an ultrasound (B-mode) color-scanning scope (SR-100; Kaijo Corp., Tokyo, Japan). Computer software determined EM. Feed intake was measured with boars reared individually, and feed conversion ratio (FCR) was calculated from 30 to 105 kg of BW. Blood samples also were collected from candidate boars and gilts and from two full sibs, mainly castrated full brothers at 105 kg of BW without fasting. Pigs were provided ad libitum access to a commercial diet (15% CP, 78% total digestible nutrition, 0.76 % lysine content, DM basis) during the testing periods from 30 to 105 kg of live weight. Pigs had free access to water. Boars were reared individually in performance-testing pens. Gilts and barrows were reared in growing pens and group fed in a concrete-floored building with eight pigs per pen, which provided floor space of 1.2 m2 per pig. Selection Method The objectives of this selection were to produce an excellent Duroc line for use as terminal sires in meat production with high meat quality traits, and to supply these Duroc boars to pig farmers as commercial terminal sires. Therefore, selection was conducted without a control line at the Miyagi Prefecture Animal Industry Experiment Station in Japan. The first and second generations of selection were performed by an index selection method based on relative desired gains (Yamada et al., 1975). Traits selected for were DG, EM, BF, and IMF. Genetic and phenotypic parameters used to derive the selection criteria were obtained from the performance test data of the first and second generation, respectively. The means of DG, EM, BF, and IMF at the first generation were 865 g, 36.1 cm2, 2.34 cm, and 4.3%, respectively. Relative desired gain was established as 135 g, 3.9 cm2, −0.54 cm, and 0.7% for DG, EM, BF and IMF, respectively. To avoid rapid disappearance of the base generation's genes from the population, selection was made within sires for boars and within litters for gilts at the first generation. Breeding values of DG, EM, BF, and IMF were estimated by multiple-trait, animal-model BLUP, from the third generation onward. The breeding values were calculated using the PEST3.1 program (Groeneveld and Kovac, 1990) after estimating genetic parameters by the VCE4 program (Newmaier and Groeneveld, 1998), with models of generation and sex as fixed effects and random effects of individual additive genetic effect and error. Relative economic weights of selection traits were calculated from the relative desired gain. The relative desired gains of DG, EM, BF, and IMF were established from the performance test data of the first generation as described before. The aggregate breeding values were calculated by multiplying the relative economic weights to the EBV of each trait, and the selection was executed. Approximately 15 boars and 50 gilts were selected at each generation. In each generation, inbreeding coefficients for individual pigs were computed. Based on inbreeding information, all matings were planned to minimize the rate of increase in inbreeding. Carcass Dissection and Meat Quality Measurement Pigs for full-sib testing (barrows and gilts) were slaughtered by manual low-voltage (200 V) electrical stunning 24 h after feed removal with free access to water. Carcasses were placed in a conventional chiller at 4°C for 24 h. Subsequently, for measuring meat quality in the LM, a 7- to 10-cm-long piece of the loin (two thoracic vertebrae sections above the last rib) was taken from the left half carcass of each pig. External loin adipose tissue was removed. Two pieces of meat, 2 × 2 × 5 cm each, were weighed and vacuum-packaged in polyethylene bags. They were then heated with a water bath of 70°C for 30 min. Then, after cooling at room temperature, two cooked pieces per animal were cut to 1 × 1 × 5 cm. We measured the tenderness (TS, kgf/cm2) with a Tensipresser (TTP-50BXII; Taketomo Electric Co., Tokyo, Japan) developed by Nakai et al. (1992). This machine was developed to accurately evaluate meat tenderness using an up and down motion to imitate chewing action. To determine IMF, two minced loin meat samples of approximately 20 g each were analyzed using the Soxhlet method. Measurement of Serum IGF-I Concentration At the first stage of selection at 8 wk of age and at the end of testing when all pigs reached 105 kg of BW, blood samples were collected from the jugular vein of pigs from the first to the seventh generation of selection. However, IGF-I was measured only for boars and gilts at the fourth generation and IGF-I of boars, gilts, and barrows was measured from the fifth to seventh generation. After centrifugal separation for 15 min at 4°C, the serum was preserved at −20°C until used for IGF-I assay. Serum samples were extracted by acid–ethanol extraction to dissociate IGF-I from binding proteins before assaying (Daughaday et al., 1980). Then IGF-I concentrations were assayed in duplicate by RIA following the double-antibody method (Kuhara et al., 1992). Intra- and interassay coefficients were 3.3 and 7.8%, respectively. Minimum detectable concentration was 20 pg/mL. Serum IGF-I concentration measured at 8 wk and that measured at 105 kg of BW were defined as IGFI-8W and IGFI-105KG, respectively. Table 1 shows the total number of IGF-I measurements. Table 1. Number of animals, means, standard deviation, heritability, common environmental effect (c2), and phenotypic standard deviation (σp) in Duroc pigs Traits No. Mean SD h2 ± SE c2 ± SE σp Daily gain, g/d 1,642 873.6 109.3 0.49 ± 0.03 0.04 ± 0.01 80.7 Loin muscle area, cm2 1,639 37.00 4.05 0.43 ± 0.03 0.02 ± 0.01 3.69 Backfat thickness, cm 1,642 2.37 0.43 0.73 ± 0.02 0.02 ± 0.01 0.39 Intramuscular fat, % 543 4.25 1.46 0.39 ± 0.03 0.08 ± 0.02 1.40 Tenderness, kgf/cm2 544 72.51 12.71 0.39 ± 0.04 0.08 ± 0.02 12.7 Body weight at 8 wk, kg 1,642 22.69 3.76 0.24 ± 0.02 0.36 ± 0.03 3.18 Backfat thickness at 8 wk, cm 1,639 1.02 0.20 0.41 ± 0.02 0.17 ± 0.02 1.66 Feed conversion ratio 379 2.65 0.17 0.35 ± 0.04 0.15 ± 0.04 0.17 IGFI-8W, ng/mLa 832 102.8 62.0 0.23 ± 0.03 0.20 ± 0.02 40.0 IGFI-105KG, ng/mLb 834 128.5 68.9 0.26 ± 0.03 0.03 ± 0.01 44.5 Traits No. Mean SD h2 ± SE c2 ± SE σp Daily gain, g/d 1,642 873.6 109.3 0.49 ± 0.03 0.04 ± 0.01 80.7 Loin muscle area, cm2 1,639 37.00 4.05 0.43 ± 0.03 0.02 ± 0.01 3.69 Backfat thickness, cm 1,642 2.37 0.43 0.73 ± 0.02 0.02 ± 0.01 0.39 Intramuscular fat, % 543 4.25 1.46 0.39 ± 0.03 0.08 ± 0.02 1.40 Tenderness, kgf/cm2 544 72.51 12.71 0.39 ± 0.04 0.08 ± 0.02 12.7 Body weight at 8 wk, kg 1,642 22.69 3.76 0.24 ± 0.02 0.36 ± 0.03 3.18 Backfat thickness at 8 wk, cm 1,639 1.02 0.20 0.41 ± 0.02 0.17 ± 0.02 1.66 Feed conversion ratio 379 2.65 0.17 0.35 ± 0.04 0.15 ± 0.04 0.17 IGFI-8W, ng/mLa 832 102.8 62.0 0.23 ± 0.03 0.20 ± 0.02 40.0 IGFI-105KG, ng/mLb 834 128.5 68.9 0.26 ± 0.03 0.03 ± 0.01 44.5 a IGF-I concentration at 8 wk. b IGF-I concentration at 105 kg BW. View Large Table 1. Number of animals, means, standard deviation, heritability, common environmental effect (c2), and phenotypic standard deviation (σp) in Duroc pigs Traits No. Mean SD h2 ± SE c2 ± SE σp Daily gain, g/d 1,642 873.6 109.3 0.49 ± 0.03 0.04 ± 0.01 80.7 Loin muscle area, cm2 1,639 37.00 4.05 0.43 ± 0.03 0.02 ± 0.01 3.69 Backfat thickness, cm 1,642 2.37 0.43 0.73 ± 0.02 0.02 ± 0.01 0.39 Intramuscular fat, % 543 4.25 1.46 0.39 ± 0.03 0.08 ± 0.02 1.40 Tenderness, kgf/cm2 544 72.51 12.71 0.39 ± 0.04 0.08 ± 0.02 12.7 Body weight at 8 wk, kg 1,642 22.69 3.76 0.24 ± 0.02 0.36 ± 0.03 3.18 Backfat thickness at 8 wk, cm 1,639 1.02 0.20 0.41 ± 0.02 0.17 ± 0.02 1.66 Feed conversion ratio 379 2.65 0.17 0.35 ± 0.04 0.15 ± 0.04 0.17 IGFI-8W, ng/mLa 832 102.8 62.0 0.23 ± 0.03 0.20 ± 0.02 40.0 IGFI-105KG, ng/mLb 834 128.5 68.9 0.26 ± 0.03 0.03 ± 0.01 44.5 Traits No. Mean SD h2 ± SE c2 ± SE σp Daily gain, g/d 1,642 873.6 109.3 0.49 ± 0.03 0.04 ± 0.01 80.7 Loin muscle area, cm2 1,639 37.00 4.05 0.43 ± 0.03 0.02 ± 0.01 3.69 Backfat thickness, cm 1,642 2.37 0.43 0.73 ± 0.02 0.02 ± 0.01 0.39 Intramuscular fat, % 543 4.25 1.46 0.39 ± 0.03 0.08 ± 0.02 1.40 Tenderness, kgf/cm2 544 72.51 12.71 0.39 ± 0.04 0.08 ± 0.02 12.7 Body weight at 8 wk, kg 1,642 22.69 3.76 0.24 ± 0.02 0.36 ± 0.03 3.18 Backfat thickness at 8 wk, cm 1,639 1.02 0.20 0.41 ± 0.02 0.17 ± 0.02 1.66 Feed conversion ratio 379 2.65 0.17 0.35 ± 0.04 0.15 ± 0.04 0.17 IGFI-8W, ng/mLa 832 102.8 62.0 0.23 ± 0.03 0.20 ± 0.02 40.0 IGFI-105KG, ng/mLb 834 128.5 68.9 0.26 ± 0.03 0.03 ± 0.01 44.5 a IGF-I concentration at 8 wk. b IGF-I concentration at 105 kg BW. View Large Statistical Analysis Performance data for DG, EM, BF, BW8W, BF8W, and FCR and meat quality data for IMF and TS, and IGFI-8W and IGFI-105KG were used for the final estimation of genetic parameters. Table 1 shows the number of pigs for each measurement. The following multiple trait animal model was used for analysis to estimate genetic parameters: \[\mathit{Y_{ijklm}}\ =\ {\mu}_{\mathit{i}}\ +\ \mathit{G_{ij}}\ +\ \mathit{S_{ik}}\ +\ \mathit{c_{il}}\ +\ \mathit{a_{im}}\ +\ \mathit{e_{ijklm}}\] where Yijklm = observation for traits i; μi = common constant for trait i; and Gij = fixed effect of selection generation j for trait i (this selection generation effect included the genetic effect of selection and the environmental effect at each generation); Sik = fixed effect of sex k for trait i; cil = random effect of common environment l of littermates for trait i; aim = random additive genetic effect of animal m for trait i; and eijklm = random residual effect for trait i. Seven generations of pedigree information of 1,642 animals with data and of 152 ancestors born before the fourth generation (total 1,794 animals) were included in this analysis. The VCE4.25 program (Neumaier and Groeneveld, 1998) was used to estimate (co)variance components and their respective standard error. Standard errors of heritability estimates and genetic correlations were also estimated with the VCE4.25 program. The GLM procedure of SAS (SAS Inst., Inc., Cary, NC) was used to obtain sex × generation least squares means of IGFI-8W and IGFI-105KG accounting for fixed effect of sex, generation, and sex × generation interaction and to test significance. Results Table 1 shows heritability estimates for meat production traits and serum IGF-I concentration. At the fourth generation of selection, IGF-I concentration was measured with boar and gilt candidates; from the fifth to seventh generations of selection, it was measured with these candidates and their two full sibs. Heritability estimates of IGFI-8W, IGFI-105KG, and BW8W were similar (0.23, 0.26, and 0.24, respectively). Table 1 also shows estimates of the proportion of the common environmental variance (c2). Its estimates for IGFI-8W and BW8W were higher than that for IGFI-105KG. Table 2 presents genetic and phenotypic correlations of IGF-I concentration (IGFI-8W and IGFI-105KG) with selection traits (DG, EM, BF, and IMF), and correlated traits (TS, BW8W, BF8W, and FCR). The genetic correlation between IGFI-8W and BW8W was moderate and those between IGFI-8W and DG, EM, IMF, BF8W, and FCR were low. In addition, genetic correlations of IGFI-8W with BF and TS were low. Phenotypic correlations between them were 0.1 or less, except for correlations between IGFI-8W and BW8W, and BF8W. Although IGFI-105KG showed moderately high genetic correlations with EM and TS, it had low genetic correlations with BF, BW8W, and BF8W. In addition, the estimated genetic correlation between IGFI-8W and IGFI-105 KG was high, but the estimated phenotypic correlation was low. Table 2. Genetic correlation (rG) and phenotypic correlation (rp) in Duroc pigs IGFI-8Wa IGFI-105KGb Traits rG ± SE rp rG ± SE rp Daily gain 0.26 ± 0.08 0.08 0.22 ± 0.07 0.05 Loin muscle area 0.22 ± 0.10 0.02 0.42 ± 0.08 0.19 Backfat thickness 0.13 ± 0.08 0.10 −0.02 ± 0.07 −0.11 Intramuscular fat 0.32 ± 0.10 0.04 0.26 ± 0.09 −0.17 Tenderness −0.05 ± 0.12 0.02 0.36 ± 0.10 0.30 Body weight at 8 wk 0.45 ± 0.08 0.40 0.09 ± 0.09 −0.08 Backfat thickness at 8wk 0.33 ± 0.06 0.28 0.00 ± 0.06 −0.09 Feed conversion ratio 0.20 ± 0.08 0.13 −0.17 ± 0.07 −0.21 IGF-I concentration at 8 wk — — 0.73 ± 0.08 0.15 IGFI-8Wa IGFI-105KGb Traits rG ± SE rp rG ± SE rp Daily gain 0.26 ± 0.08 0.08 0.22 ± 0.07 0.05 Loin muscle area 0.22 ± 0.10 0.02 0.42 ± 0.08 0.19 Backfat thickness 0.13 ± 0.08 0.10 −0.02 ± 0.07 −0.11 Intramuscular fat 0.32 ± 0.10 0.04 0.26 ± 0.09 −0.17 Tenderness −0.05 ± 0.12 0.02 0.36 ± 0.10 0.30 Body weight at 8 wk 0.45 ± 0.08 0.40 0.09 ± 0.09 −0.08 Backfat thickness at 8wk 0.33 ± 0.06 0.28 0.00 ± 0.06 −0.09 Feed conversion ratio 0.20 ± 0.08 0.13 −0.17 ± 0.07 −0.21 IGF-I concentration at 8 wk — — 0.73 ± 0.08 0.15 a IGF-I concentration at 8 wk of age. b IGF-I concentration at 105 kg of BW. View Large Table 2. Genetic correlation (rG) and phenotypic correlation (rp) in Duroc pigs IGFI-8Wa IGFI-105KGb Traits rG ± SE rp rG ± SE rp Daily gain 0.26 ± 0.08 0.08 0.22 ± 0.07 0.05 Loin muscle area 0.22 ± 0.10 0.02 0.42 ± 0.08 0.19 Backfat thickness 0.13 ± 0.08 0.10 −0.02 ± 0.07 −0.11 Intramuscular fat 0.32 ± 0.10 0.04 0.26 ± 0.09 −0.17 Tenderness −0.05 ± 0.12 0.02 0.36 ± 0.10 0.30 Body weight at 8 wk 0.45 ± 0.08 0.40 0.09 ± 0.09 −0.08 Backfat thickness at 8wk 0.33 ± 0.06 0.28 0.00 ± 0.06 −0.09 Feed conversion ratio 0.20 ± 0.08 0.13 −0.17 ± 0.07 −0.21 IGF-I concentration at 8 wk — — 0.73 ± 0.08 0.15 IGFI-8Wa IGFI-105KGb Traits rG ± SE rp rG ± SE rp Daily gain 0.26 ± 0.08 0.08 0.22 ± 0.07 0.05 Loin muscle area 0.22 ± 0.10 0.02 0.42 ± 0.08 0.19 Backfat thickness 0.13 ± 0.08 0.10 −0.02 ± 0.07 −0.11 Intramuscular fat 0.32 ± 0.10 0.04 0.26 ± 0.09 −0.17 Tenderness −0.05 ± 0.12 0.02 0.36 ± 0.10 0.30 Body weight at 8 wk 0.45 ± 0.08 0.40 0.09 ± 0.09 −0.08 Backfat thickness at 8wk 0.33 ± 0.06 0.28 0.00 ± 0.06 −0.09 Feed conversion ratio 0.20 ± 0.08 0.13 −0.17 ± 0.07 −0.21 IGF-I concentration at 8 wk — — 0.73 ± 0.08 0.15 a IGF-I concentration at 8 wk of age. b IGF-I concentration at 105 kg of BW. View Large Table 3 lists the least squares means of IGFI-8W and IGFI-105KG for the fourth to the seventh generations of selection and for the three sexes. Results of ANOVA in the IGFI-8W indicated that the effect of generation and sex was significant. The IGFI-8W concentration in boars and gilts increased (P < 0.01) from the fourth to the sixth generation, but the concentration at the seventh generation decreased significantly compared with the sixth generation. The average BW8W of the seventh generation decreased significantly from the previous generation (BW for these four generations were 24.1, 27.2, 26.1, and 23.7 kg in boars and 22.4, 25.7, 24.8, and 22.2 kg in gilts, respectively). Decreased IGFI-8W seems to be related to this decrease in growth. The mean IGFI-8W of boars was significantly higher than those of gilts for all generations except the fourth generation. Analysis of variance indicated that the effects of generation, sex, and generation × sex interaction were significant for the IGFI-105KG. This generation × sex interaction resulted from a high concentration (P < 0.01) of IGFI-105KG in seventh-generation boars. There were significant differences in mean IGFI-105KG among boars, gilts, and barrows. The serum IGFI-105KG increased (P < 0.01) compared with IGFI-8W. In addition, serum IGF-I concentration of gilts increased (P < 0.01) even though the rate of increase was less than that for boars. However, IGF-I concentration of barrows did not change with their growth at the fifth and seventh generations; also, it decreased (P < 0.01) at the sixth generation of selection. Table 3. Least squares means ± standard errors for serum IGF-I concentration among sex, growth stage, and generation of selection in Duroc pigs IGFI-8W, ng/mLa IGFI-105KG, ng/mLb Generation Boars Gilts Boarsc Boars Gilts Barrows 4 45.4 ± 6.2 44.9 ± 4.1 67.3 ± 6.4f 49.8 ± 4.7g 5 104.7 ± 5.7 75.3 ± 3.7e 90.1 ± 5.5d 159.3 ± 6.0f 101.1 ± 3.8g 86.2 ± 5.7h 6 158.9 ± 5.9d 129.3 ± 3.6e 133.6 ± 5.6e 184.4 ± 5.9f 132.5 ± 3.7g 98.2 ± 5.7h 7 126.6 ± 4.4d 107.5 ± 3.5e 117.3 ± 5.6de 256.4 ± 4.5f 146.4 ± 3.5g 116.2 ± 5.7h IGFI-8W, ng/mLa IGFI-105KG, ng/mLb Generation Boars Gilts Boarsc Boars Gilts Barrows 4 45.4 ± 6.2 44.9 ± 4.1 67.3 ± 6.4f 49.8 ± 4.7g 5 104.7 ± 5.7 75.3 ± 3.7e 90.1 ± 5.5d 159.3 ± 6.0f 101.1 ± 3.8g 86.2 ± 5.7h 6 158.9 ± 5.9d 129.3 ± 3.6e 133.6 ± 5.6e 184.4 ± 5.9f 132.5 ± 3.7g 98.2 ± 5.7h 7 126.6 ± 4.4d 107.5 ± 3.5e 117.3 ± 5.6de 256.4 ± 4.5f 146.4 ± 3.5g 116.2 ± 5.7h a IGF-I concentration at 8 wk of age. b IGF-I concentration at 105 kg BW. c These boars are selected as full sib test and they were castrated after blood collection. d,e Means with different superscripts within the same row in IGFI-8W differ by sex (P < 0.05). f,g,h Means with different superscripts within the same row in IGFI-105KG differ by sex (P < 0.05). View Large Table 3. Least squares means ± standard errors for serum IGF-I concentration among sex, growth stage, and generation of selection in Duroc pigs IGFI-8W, ng/mLa IGFI-105KG, ng/mLb Generation Boars Gilts Boarsc Boars Gilts Barrows 4 45.4 ± 6.2 44.9 ± 4.1 67.3 ± 6.4f 49.8 ± 4.7g 5 104.7 ± 5.7 75.3 ± 3.7e 90.1 ± 5.5d 159.3 ± 6.0f 101.1 ± 3.8g 86.2 ± 5.7h 6 158.9 ± 5.9d 129.3 ± 3.6e 133.6 ± 5.6e 184.4 ± 5.9f 132.5 ± 3.7g 98.2 ± 5.7h 7 126.6 ± 4.4d 107.5 ± 3.5e 117.3 ± 5.6de 256.4 ± 4.5f 146.4 ± 3.5g 116.2 ± 5.7h IGFI-8W, ng/mLa IGFI-105KG, ng/mLb Generation Boars Gilts Boarsc Boars Gilts Barrows 4 45.4 ± 6.2 44.9 ± 4.1 67.3 ± 6.4f 49.8 ± 4.7g 5 104.7 ± 5.7 75.3 ± 3.7e 90.1 ± 5.5d 159.3 ± 6.0f 101.1 ± 3.8g 86.2 ± 5.7h 6 158.9 ± 5.9d 129.3 ± 3.6e 133.6 ± 5.6e 184.4 ± 5.9f 132.5 ± 3.7g 98.2 ± 5.7h 7 126.6 ± 4.4d 107.5 ± 3.5e 117.3 ± 5.6de 256.4 ± 4.5f 146.4 ± 3.5g 116.2 ± 5.7h a IGF-I concentration at 8 wk of age. b IGF-I concentration at 105 kg BW. c These boars are selected as full sib test and they were castrated after blood collection. d,e Means with different superscripts within the same row in IGFI-8W differ by sex (P < 0.05). f,g,h Means with different superscripts within the same row in IGFI-105KG differ by sex (P < 0.05). View Large Discussion Present results suggest that the heritability for serum IGF-I concentration was low and the common environmental effect on the growth performance was large in early stage of growth in pig. Blair et al. (1989) and Baker et al. (1991) reported heritability estimates of 0.15 for serum IGF-I concentration at 6 wk of age, and of 0.10 for the serum IGF-I concentration at 12 wk in mice. Bunter et al. (2002) reviewed heritability estimates ranging from 0.20 to 0.58 and pooled estimates of 0.28 and 0.32, including and excluding estimates from restricted feeding data. Hermesch et al. (2001) estimated the heritability (0.24) and common environmental effect (0.13) for juvenile IGF-I concentration at 4 wk. In addition, Cameron et al. (2003) reported a high and significant common environment variance component (0.46) for serum IGF-I concentration at 6 wk. Cameron et al. (2003) pointed out that a genetic analysis of serum IGF-I concentration data, measured at an early age, should include a common environment effect in the model to avoid overestimating heritability. If the common environmental effect had not been included in the present analysis, the estimated heritability of IGFI-8W would have been higher (0.37). Heritability is inferred to be low judging from the present result and other reports. This research was mainly intended to examine whether the IGF-I concentration measured at an early stage of growth is effective as a physiological predictor of meat production traits at a later stage of growth. Therefore, we estimated genetic and phenotypic correlations between serum IGF-I concentration and meat production and meat quality traits. Results suggest that serum IGFI-8W was a useful index of growth and fat accumulation up to 8 wk because IGFI-8W showed a moderate, positive genetic and phenotypic correlation with BW8W and BF8W. However, genetic correlations of IGFI-8W with DG, FCR, EM, BF, and TENDERNESS were lower than those with BW8 and BF8W. The present estimates of genetic correlation with these traits are different in the magnitude and direction from previously reported estimates. For growth rate, Baker et al. (1991) reported a realized genetic correlation of 0.58 and a phenotypic correlation of 0.38 between IGF-I concentration and BW at 12 wk in mice. On the other hand, Hermesch et al. (2001) estimated a genetic correlation of 0.12 ± 0.12 between IGF-I measured at 4 wk and the lifetime ADG. Bunter et al. (2002) also reported a low genetic correlation of 0.15 between IGF-I concentration measured at around 4 to 5 wk and lifetime ADG in their review. It seems that serum IGF-I concentration at an early stage of growth is not an effective physiological predictor of genetic merit for production traits during performance test. A high genetic correlation of 0.50 between backfat thickness and serum IGF-I concentration was reported (Bunter et al., 2002). However, the genetic correlation between IGFI-8W and BF at 105 kg of BW was low in the present study. The result of administering serum IGF-I to Meishan pigs (Klindt et al., 1998) was a significant increase in backfat thickness. On the other hand, Buonomo and Klindt (1993) reported that selection for increased backfat thickness of the pig increased serum IGF-2 concentration. In addition, Owens et al. (1999) reported that IGF-I controls growth of lean meat and that IGF-2 controls adipose tissue growth. In most countries, the main objectives of pig improvement have been increasing lean growth rate and decreasing backfat thickness. In contrast, moderate thickness in the dressed carcass is important in Japan. Therefore, the BF of the Duroc breed used in the present study was considered to be thick (average 2.37 cm). Apparently, the genetic difference in the fat accumulation of the pig sample used may influence the magnitude of genetic correlation. Along with backfat thickness, high genetic correlations between IGF-I concentration and FCR have been reported in pig (Bunter et al., 2002) and beef cattle (Johnston et al., 2002). The present study estimated lower genetic and phenotypic correlations. A moderate genetic correlation of 0.32 between IGFI-8W and IMF was estimated. No reports have addressed the relation between serum IGF-I concentration and meat quality traits. If IMF could be estimated from serum IGF-I concentration, such information would be important and relatively easily obtained. Duroc pigs used in present study can accumulate more intramuscular fat than other breeds or lines. Therefore, it is necessary to confirm the relationship between IGF-I and IMF with other breeds showing low intramuscular fat accumulation. Compared with the genetic correlations of EM and TS with IGFI-8W measured at 8 wk, those with IGFI-105KG measured at an older age increased. These results suggest that serum IGF-I concentration is related to an increase in EM thickness and the amount of the lean meat. Serum IGF-I concentration of boars increased considerably from 8 wk to 105 kg of BW compared with gilts and barrows in every generation. That result suggests that the expression of IGF-I concentration is limited by other hormonal factors. Genetic correlations of BW8W, BF8W, BF at 105 kg, and FCR with IGFI-105KG were lower than with IGFI-8W. Moreover, the genetic correlations of DG and IMF with IGFI-105KG did not changed compared to those with IGFI-8W. These changes suggest that the physiological function of IGF-I depends on the animal's age. Cameron et al. (2003) also reported that the serum IGF-I concentration measured at 6 wk was positively related with ultrasonic backfat depth and the FCR, whereas serum IGF-I concentration measured at the end of test (90 kg of BW) was negatively correlated with BF and the FCR. Implications A positive genetic correlation of serum insulin-like growth factor I concentration at 8 wk of age in pigs with intramuscular fat was estimated. Furthermore, a moderate genetic correlation between serum insulin-like growth factor I concentration at 105 kg of body weight and loin muscle area also was estimated. These results suggest that serum insulin-like growth factor I concentration has some relation to growth traits and meat production traits. It may be useful as an indirect selection criterion in the early growth stage to evaluate future ability of intramuscular fat accumulation in pig breeding. Literature Cited Baker, R. L., A. J. Peterson, J. J. Bass, N. C. Amyes, B. H. Breier, and P. D. Gluckman 1991. Replicated selection for insulin-like growth factor-1 and body weight in mice. Theor. Appl. Genet. 81: 685– 692. Google Scholar CrossRef Search ADS PubMed Blair, H. T., S. N. McCutcheon, D. D. S. Mackenzie, P. D. Gluckman, J. E. Ormsby, and B. H. Brier 1989. Responses to divergent selection for plasma concentrations of insulin-like growth factor-1 in mice. Genet. Res. (Camb.) 53: 187– 191. Google Scholar CrossRef Search ADS Bunter, K., S. Hermesch, B. G. Luxford, K. Lahti, and E. Sutcliffe 2002. IGF-I concentration measured in juvenile pigs provides information for breeding programs: A mini review. Communication No. 03–09 in Proc. 7th World Cong. Genet. Appl. Livest. Prod., Montpellier, France. Buonomo, F. C., and J. Klindt 1993. Ontogeny of growth hormone (GH), insulin-like growth factors (IGF-I and IGF-II) and IGF binding protein-2 (IGFBP-2) in genetically lean and obese swine. Domest. Anim. Endocrinol. 10: 257– 265. Google Scholar CrossRef Search ADS PubMed Cameron, N. D., E. McCullough, K. Troup, and C. Penman 2003. Serum insulin-like growth factor-1 concentration in pigs divergently selelcted for daily food intake or lean growth rate. J. Anim. Breed. Genet. 120: 1– 10. Google Scholar CrossRef Search ADS Daughaday, W. H., I. K. Mariz, and S. L. Blethen 1980. Inhibition of access of bound somatomedin to membrane receptor and immunobinding sites: A comparison of radioreceptor and radioimmunoassay of somatomedin in native and acid-ethanol-extracted serum. J. Clin. Endocrinol. Metab. 51: 781– 788. Google Scholar CrossRef Search ADS PubMed Davis, M. E., and R. C. M. Simmen 1997. Genetic parameter estimates for serum insulin-like growth factor 1 concentration and performance traits in Angus beef cattle. J. Anim. Sci. 75: 317– 324. Google Scholar CrossRef Search ADS PubMed Davis, M. E., and R. C. M. Simmen 2000. Genetic parameter estimates for serum insulin-like growth factor 1 concentration and carcass traits in Angus beef cattle. J. Anim. Sci. 78: 2305– 2313. Google Scholar CrossRef Search ADS PubMed Groeneveld, E., and M. Kovac 1990. A generalized computing procedure for setting up and solving mixed linear models. J. Dairy Sci. 73: 513– 531. Google Scholar CrossRef Search ADS Hermesch, S., K. L. Bunter, and B. G. Luxford 2001. Estimates of genetic correlations between IGF-I recorded at 4 wk of age and individual piglet weights at birth and 14 days, along with lifetime growth rate and backfat. Pages 227–230 in Proc. Assoc. Advan. Anim. Breed. Genet., Queenstown, New Zealand. Johnston, D. J. R. M. Herd, M. J. Kadel, H-U. Graser, P. F. Arthur, and J. A. Archer 2002. Evidence of IGF-I as a genetic predictor of feed efficiency traits in beef cattle. Communication No. 10–16 in Proc. 7th World Cong. Genet. Appl. Livest. Prod., Montpellier, France. Klindt, J., J. T. Yen, F. C. Buonomo, A. J. Roberts, and T. Wise 1998. Growth, body composition, and endocrine responses to chronic administration of insulin-like growth factor and(or) porcine growth hormone in pigs. J. Anim. Sci. 76: 2368– 2381. Google Scholar CrossRef Search ADS PubMed Kuhara, T., K. Katoh, S. Oda, A. Ohneda, and Y. Sasaki 1992. Responses of metabolic hormones to prolonged intraduodenal infusion of amino acids in sheep. Anim. Sci. Technol. 63: 1123– 1133. Nakai, H., R. Tanabe, T. Ikeda, and M. Nishizawa 1992. Development of a technique for measuring tenderness in meat using a“Tensipresser.” Pages 947–950 in Proc. 38th Int. Cong. Meat Sci. Technol., Clermont-Ferrand, France. Neumaier, A., and E. Groeneveld 1998. Restricted maximum likelihood estimation of covariance in sparse linear models. Genet. Sel. Evol. 30: 3– 26. Google Scholar CrossRef Search ADS Owens, P. C., K. L. Gatford, P. E. Walton, W. Morley, and R. G. Campbell 1999. The relationship between endogenous insulin-like growth factors and growth in pigs. J. Anim. Sci. 77: 2098– 2103. Google Scholar CrossRef Search ADS PubMed Scanes, C. G., D. Lazarus, S. Bowen, F. C. Buonomo, and R. L. Gilbreath 1987. Postnatal changes in circulating concentrations of growth hormone, somatomedin C and thyroid hormones in pigs. Domest. Anim. Endocrinol. 4: 253– 257. Google Scholar CrossRef Search ADS PubMed Suzuki, K., H. Kadowaki, T. Shibata, H. Uchida, and Y. Sato 2002. Selection for daily gain, loin-eye area, backfat thickness and intramuscular fat in 7 generations of Duroc pigs. Communication No. 11–15 in Proc. 7th World Cong. Genet. Appl. Livest. Prod., Montpellier, France. Yamada, Y., K. Yokouchi, and A. Nishida 1975. Selection index when genetic gains of individual traits are of primary concern. Japan. J. Genetics. 50: 33– 41. Google Scholar CrossRef Search ADS Footnotes 1 We gratefully acknowledge A. F. Parlow for the IGF-I antibody provided by NIDDK, USA, and N. D. Cameron for comments on the manuscript. Copyright 2004 Journal of Animal Science