Investigations on major gene by polygene and gene by environment interaction in German Holstein dairy cattle

摘要

Putative interaction effects between DGAT1 K232A mutation and the polygenic terms (all genes except DGAT1) were investigated in chapter one. This was done for five milk production traits (milk yield, protein yield, fat yield, protein percentage and fat percentage) in the German Holstein dairy cattle population. Therefore, mixed models are used. The test for interaction relied on the comparison of polygenic variance components depending on the sire?s genotypes at DGAT1 K232A. Found substitution effects were highly significant for all traits. Significant interactions between DGAT1 K232A and the polygenic term were found for milk fat and protein percentage. These interactions could be used in breeding schemes. Depending on the DGAT1 K232A genotypes of the sample, in which the sire will be used, three polygenic breeding values of a sire can be calculated. Because the genotypes of the samples are often unknown and usually heterogeneous, this is not a practical approach. Rank correlations between the three polygenic EBVs were always above 0.95, which suggested very little re-ranking. GxE were studied in chapter two. For this, reaction norm random regression sire models were used in the German Holstein dairy cattle population. Around 2300 sires with a minimum of 50 daughters per sire and at minimum seven first-lactation test day observations per daughter were analyzed. As traits, corrected test day records for milk yield, protein yield, fat yield and somatic cell score (SCS) were used. As environmental descriptors, we used herd test day solutions (htds) for milk traits, milk energy yield or SCS. Second-order orthogonal polynomial regressions were applied to the sire effects. Results showed significant slope variances of the reaction norms, which caused a non-constant additive genetic variance across the environmental ranges considered, which pointed to the presence of minor GxE effects. When the environment improved, the additive genetic variance increased, meaning higher (lower) htds for milk traits (SCS). This was also influenced by pure scaling effects, because the non-genetic variance increased in an improved environment and the heritability was less influenced by the environment. For the environmental ranges considered in this study, GxE effects caused very little re-ranking of the sires. To obtain unbiased genetic parameters, it was important to model heterogeneous residual variances. A large genome-wide association analysis was conducted in chapter three to identify SNPs that affect general production (GP) and environmental sensitivity (ES) of milk traits. Around 13 million daughter records were used to calculate sire estimates for GP and ES with help of linear reaction norm models. Daughters were offspring from 2297 sires. The sires were genotyped with a 54k SNP chip. As environmental descriptor, the average milk energy yield performance of the herds at the time where the daughter observations were recorded was used. The sire estimates were used as observations in genome-wide association analyses using 1797 sires. With help of an independent validation set (500 sires of the same population), significant SNPs were confirmed. To separate GxE scaling and other GxE effects, the observations were log-transformed. GxE effects could be found with help of reaction norm models and numerous significant SNPs could be validated for GP and ES, whereas many SNPs affecting GP also affected ES. ES of milk traits is a typical quantitative trait, which is controlled by many genes with small effects and few genes with larger effect. Effects of some SNPs for ES were not removable by log-transformation of observations, indicating that these are not solely scaling effects. Positions of founded clusters were often in well-known candidate regions affecting milk traits. No SNPs, which show effects for GP and ES in opposite directions could be found. Environmental descriptor in GxE analyses is often modelled by average herd milk production levels. Another possibility could be the level of hygiene and udder health. In chapter four, the same models were used as in chapter three. A genome-wide association analysis was done using htds for SCS as an environmental descriptor. With help of this, several SNP clusters affecting intercept and slope could be detected and confirmed. Many SNPs or clusters affecting intercept and slope could be identified, but in total, the number of SNPs affecting intercept was larger. The same SNPs could be detected and validated with and without considering GxE in reaction norm models. Some SNPs affecting only slope were found. For slope, nearly the same SNPs could be found with SCS as an environmental descriptor as presented in chapter three, although both environmental descriptors were only slightly correlated.