|Title||Detection of quantitative trait loci in broilers|
|Author(s)||Kaam, J.T. van|
|Source||Agricultural University. Promotor(en): E.W. Brascamp; J.A.M. van Arendonk; R.L. Quaas. - S.l. : S.n. - ISBN 9789058082633 - 175|
Animal Breeding and Genomics
|Publication type||Dissertation, internally prepared|
|Keyword(s)||vleeskuikens - genen - kwantitatieve kenmerken - lichaamsgewicht - groei - voeropname - genetische kartering - genetische merkers - regressieanalyse - broilers - genes - quantitative traits - body weight - growth - feed intake - genetic mapping - genetic markers - regression analysis|
This dissertation deals with the development and application of methods for the detection of genes with a substantial influence on quantitative traits, so called quantitative trait loci (QTLs) in broilers. For the purpose of detection of QTLs, an experiment was initiated. A three generation full sib-half sib experimental population consisting of 10 full sib families originating from a cross between two broiler dam lines was set up. Genotypes of up to 437 microsatellite markers on 28 linkage groups were determined on all 20 generation one and 451 generation two full sib animals. Generation three half sib animals were divided in hatches and phenotypic observations on several traits were collected in different experiments. Data from a feed efficiency and a carcass experiment were used in the QTL analyses. In both experiments approximately 2,000 phenotypic observations were collected per trait.
The data were analysed using a two step procedure: first average adjusted progeny trait values were calculated, and secondly QTL analysis was performed using the average adjusted progeny trait values as the dependent variable. Large differences in mean and variance of male and female body weight were found. Prior adjustment of these differences is necessary to ensure that each observation has a similar effect within the QTL analysis. Therefore, a bivariate analysis was used to estimate variances, fixed and genetic effects. These estimated effects were used to calculate average adjusted progeny trait values for all generation two animals by averaging progeny observations, which were standardised after adjusting for fixed and maternal genetic effects and for the additive genetic contribution of the other parent. A full sib regression interval mapping approach was applied, because it enables a quick initial scan of the entire genome and simultaneously includes the segregation of alleles from both generation one parents. The QTL analyses were across family and average adjusted progeny trait values were weighted to account for the number of third generation observations included. In total, 24 autosomal linkage groups were analysed in this chapter. The most likely QTL position was found between markers MCW0058 and LEI0071 on chromosome 1.
This approach was applied on all traits in a feed efficiency experiment. These traits were body weight at 23 and 48 days, growth between 23 and 48 days, feed intake between 23 and 48 days, the same feed intake adjusted for body weight, and feed efficiency. In total 27 autosomal linkage groups were analysed and four QTLs for body weight, growth and feed intake traits were found. The most significant QTL was located between markers UMA1.107 and MCW0058 on chromosome 1 and had a 4% genomewise significance for feed intake between 23 and 48 days. Furthermore, this QTL exceeded suggestive linkage for growth between 23 and 48 days and body weight at 48 days. The other QTLs showed suggestive linkage. The second QTL, affecting feed intake between 23 and 48 days, was located between markers ADL0289 and ADL0262 on linkage group WAU26. On chromosome 4, between markers MCW0085 and LEI0122, a third QTL was found, which had an effect on both feed intake traits. Finally, a fourth QTL, which affected feed intake adjusted for body weight, was located between markers MCW0082 and MCW0341 on chromosome 2.
In a similar way, the analyses of all traits in a carcass experiment were performed. These traits were body weight at 48 days, carcass weight, carcass percentage, breast meat colour unadjusted and adjusted for body weight, original leg scores, transformed leg scores and transformed leg scores adjusted for body weight. Two suggestive QTLs for carcass percentage and meat colour were detected. The QTL affecting carcass percentage was located between markers ADL0183 and LEI0079 on chromosome 1. The QTL for meat colour was located on chromosome 2 and gave a peak between markers MCW0185 and MCW0234 and between markers MCW0264 and ADL0164.
The sex chromosomes were omitted from the previous genome scans. Later the Z chromosome was analysed for growth and carcass traits. Additionally, feathering was analysed. For the Z chromosome, only the segregation of male chromosomes provides information on the presence of genes and therefore a half sib interval mapping approach was used. No QTLs were found which affected growth or carcass traits. For feathering, however, a huge QTL effect was found. The feathering gene was located between markers ADL0022 and MCW0331.
For a more detailed analysis, an existing Bayesian method is extended to enable the analysis of the experimental broiler data accounting for the heterogeneity of variance between sexes. Heterogeneity is accounted for by including separate scale parameters for the polygenic and QTL allelic effects per sex and by separate error variances per sex. A Bayesian analysis is undertaken on chromosomal regions where QTLs were found with the initial regression analyses. Advantages of the Bayesian method in comparison with the regression analysis are that normally distributed random polygenic and QTL effects are modelled and dispersion parameters are estimated for all random terms in the model. Furthermore, individual observations are used instead of offspring averages and mate correction is no longer necessary, because all genetic relations are taken into account through relationship matrices. By simultaneous sampling of all model parameters, uncertainties are taken into account. The use of a reduced animal model enables the analysis of complex populations. Markov Chain Monte Carlo algorithms were applied to obtain solutions. The Bayesian method was successful in finding QTLs in all regions previously detected.
The Bayesian method is extended even further to enable a bivariate analysis of body weight data obtained in both experiments. Combining data from both experiments is expected to improve the QTL detection power and estimation accuracy. For each sex-trait combination separate error variances and separate scale parameters for the polygenic and QTL allelic effects were included. Furthermore, a polygenic correlation was included. Broiler body weight data measured at 48 days was used to illustrate the method. The QTL on chromosome 1 found previously in the feed efficiency experiment but not in the carcass experiment, was now detected in both experiments demonstrating that the QTL detection power indeed increased. The most likely QTL location, however, was in a different marker bracket for both experiments.
Finally, the number of QTLs and the power of the design are discussed. Differences between the regression and the Bayesian method are mentioned and potential extensions on both methods are discussed. With the regression method, a two QTL analysis was applied to increase the power and bootstrapping was used to provide confidence intervals of the QTL position. For the Bayesian method, the most important extensions to be implemented are the sampling of the QTL position, the inclusion of correlated residuals, which would enable bivariate analysis of traits measured on the same individuals, and the ability to handle imprinting.