Detection and characterisation of quantitative trait loci affecting muscle and growth phenotypes in sheep

Abstract

This thesis addresses the dissection and characterisation of quantitative trait loci (QTL) affecting production traits in sheep. Firstly, the association between specific genetic polymorphisms and complex variation in weight, muscle and fat depositions was investigated. Research concentrated on assessing the presence, correspondence and significance of two single nucleotide polymorphisms (SNPs) in the GDF8 region of ovine chromosome 2, reportedly affecting muscle production. Commercial populations of British Texel, Suffolk and Charollais sheep were studied. The SNPs were absent in Suffolk and almost fixed in Texel breeds. In the Charollais population, the SNPs segregated at intermediate frequencies and a significant association was found between these polymorphisms and muscle depth. The previously proposed causative allele at one of the loci resulted in increased muscle depth and, at allele frequency of 0.5, this locus would explain one third of the additive genetic variance for the trait. Partial recessive allelic expression is proposed by genotypic value predictions and is consistent with the previously postulated molecular mechanism by which it gives rise to muscle changes. Secondly, the thesis focused on detection of QTL associated with growth. Live weight is a composite of growth rates over time, with inter-age genetic correlations for live weight decreasing as time between weight measurements increases. To explore whether observed genetic correlation patterns translate into distinct loci acting on weight at different growth stages, a novel method was developed and the applicability of a second proposed method was explored. Both methods allowed simultaneous analysis of multiple live weights per animal, while accounting differently for the correlation among measurements ordered in time. In the first approach, a growth curve technique was developed and employed to map growth QTL for curve parameters and predicted growth descriptors. A study of actual live weights identified significant QTL at different ages on distinct chromosomes, with QTL significance and variance changing over time. Further application of this technique on a simulated dataset validated its effectiveness in detecting age-dependent QTL. An extension of the procedure resulted in a novel technique for genomic evaluation of longitudinal traits. In the second method examined, random regression (RR) models were applied for dissection of growth QTL. Systematic model selection and inclusion of relevant random effects resulted in apparently significant QTL, but the method was computationally demanding, model choice proved challenging and the results were questioned. To further explore the method, RR models were applied to various simulated growth phenotypes composed of time-dependent QTL trajectories, polygenic and environmental effects. Statistically optimal RR models succeeded in identifying significant QTL and predicting the simulated time-dependence for most scenarios. However, the issue of model choice was again prominent, as suboptimal models resulted in unreliable QTL variance trajectories and pronounced confounding between different time-dependent effects. Thus, the growth curve approach appeared to be the more flexible and robust process for analysing longitudinal data to map agedependent QTL.