Evidence of epistatic effects on weaning weight in crossbred beef cattle.

Abstract

Weaning weights (WW; N=1962) of 20 genotypes comprising purebred, F1, backcrosses, crisscrosses, F2, F3 and F4 generations of inter se combinations of Brahman (B) and Hereford (H) were used to estimate (direct and maternal) additive, joint additive, heterotic and epistatic effects and their interactions with environment. Results are from additional data, plus data previously presented by Arthur et al. (1999) where experimental details can be found. For brevity, only estimates for the 15 genotypes produced in Grafton (subtropical NE of NSW) are presented here. At this location, there were 5 dam genotypes: H, B, firstcross BxH (F1), backcross BxF1 or HxF1 mated to sires H, B or F1. Covariates for joint additive (genes for adaptation and genes for high metabolism) effects on WW are calcu-lated by proportion B x proportion H and conceptua-lized as complementarity or profit (Kinghorn, 1993). Dickerson (1974) defined coefficient of recombination loss as the average fraction of independently segregating pairs of loci in gametes from both parents which are expected to be non-parental combinations; numerical values for these coefficients and for hetero-zygosity in stabilized synthetic populations are equal. Given all the intermediate genotypes in the present work, individual and maternal coefficients of epistasis (E and E) were calculated as the average hetero-zygosity in parents of an individual and of the dam. The best possible model, with respect to the genetic components, is the genotypic model (GM), when the effects of each of the 15 genotypes are estimated and the best possible (100%) fit is reached. A more parsimonious representation in terms of genetic components is needed. The benchmark model (BM) contains 4 genetic para-meters: individual additive, maternal additive, indi-vidual heterosis and maternal heterosis. This is a com-monly used model in analysis of crossbred data and it explained 75.7% of the variation amongst genotypes. The full model (FM) required 8 parameters to be estimated and the 4 extra ones to the BM are: individual joint-additive, maternal joint-additive, E and E. FM could explain 90.1 % of the variation between genotypes. About 60% of the lack-of-fit or total distance (24.3%) between BM and GM can be reduced by fitting FM. Part of the remaining distance between FM and GM (9.9%) may be due to sampling errors in the estimation of each of the 15 points and part to other genetic effects. The estimation of all genetic components, showing severe multicollinearity on this data structure, was possible since ridge regression (λ=10) was used. Individual additive effects were small for WW: 0.8 kg advantage for pure B calves in comparison with pure H (if all other genetic components were the same) but maternal additive was much larger: 6.9kg. Individual and maternal joint additive effects were positive indicating that intermediate genetic compositions are complementary. The individual jointadditive benefit of a .5B .5H was 1.3 kg with respect to either purebred. If the dam was .5B .5H then the benefit was 3.3 kg with respect to either pure dam. Individual heterosis was estimated as 19.4 kg (9.8%) for a F1 calf and maternal heterosis as 26.0 kg (13.1%) for a F1 dam. Maximum epistatic loss occurs in F2 individuals and it was estimated as -9.5 kg (4.8%). This reduction in WW was independent of the one observed due to reducing heterozygosity from 1.0 to 0.5 (from F1 to F2). F2 calves still perform well because their F1 dam expresses maximum maternal heterosis. But F2 dams show full epistatic loss, estimated as -12.6 kg (6.4%). It is easy to realize the importance of considering epistatic effects for genetic evaluation if, for example, progeny of a given sire out of F1 or out of F2 dams are compared. Even if reduction in heterosis is considered in the BM, margin for error is large. Koch et al. (1985), working with advanced genera-tions of crosses among Angus and Hereford cattle, found that heterosis retained in F3 vs. F1 indicated that net epistatic effects were relatively negligible for weaning weight. Differences between their results and epistasis estimates found in this study may arise from different interpretations of Dickerson's recombination loss. Their coefficients for individual two-locus additive x additive epistatic effects in parental breeds were all equal in F1, F2 and F3; and the same for maternal epistatic effects in F2 and F3. The design of the present set of experiments allowed for the verification that more heterozygosity in one generation is concomitant with higher frequency of recombinant configurations in the gametes; therefore, more epistatic effects in the next generation. Currently used BM captures basic features of the surface from the GM while FM provides a better fit. Similar benefits and gains in accuracy can be expected if E and E are added to the models in: (1) predicting results of advanced generations of designed crossbred populations; and (2) acrossor multibreed genetic evaluation analysis.