Proofs for genotype by environment interactions considering pedigree and genomic data from organic and conventional cow reference populations

The aim of the present study was to prove genotype by environment interactions (G × E) for production, longevity, and health traits considering conventional and organic German Holstein dairy cattle subpopulations. The full data set included 141,778 Holstein cows from 57 conventional herds and 7,915 cows from 9 organic herds. The analyzed traits were first-lactation milk yield and fat percentage (FP), the length of productive life (LPL) and the health traits mastitis, ovarian cycle disorders, and digital dermatitis in first lactation. A subset of phenotyped cows was genotyped and used for the implementation of separate cow reference populations. After SNP quality controls, the cow reference sets considered 40,830 SNP from 19,700 conventional cows and the same 40,830 SNP from 1,282 organic cows. The proof of possible G × E was made via multiple-trait model applications, considering same traits from the conventional and organic population as different traits. In this regard, pedigree (A), genomic (G) and combined relationship (H) matrices were constructed. For the production traits, heritabilities were very similar in both organic and conventional populations (i.e., close to 0.70 for FP and close to 0.40 for milk yield). For low heritability health traits and LPL, stronger heritability fluctuations were observed, especially for digital dermatitis with 0.05 ± 0.01 (organic, A matrix) to 0.33 ± 0.04 (conventional, G matrix). Quite large genetic correlations between same traits from the 2 environments were estimated for production traits, especially for high heritability FP. For LPL, the genetic correlation was 0.67 (A matrix) and 0.66 (H matrix). The genetic correlation between LPL organic with LPL conventional was 0.94 when considering the G matrix, but only 213 genotyped cows were included. For health traits, genetic correlations were throughout lower than 0.80, indicating possible G × E. Genetic correlations from the different matrices A, G, and H for health and production traits followed the same pattern, but the estimates from G for health traits were associated with quite large standard errors. In genome-wide association studies, significantly associated SNP for production traits overlapped in the conventional and organic population. In contrast, for low heritability LPL and health traits, significantly associated SNP and annotated potential candidate genes differed in both populations. In this regard, significantly associated SNP for mastitis from conven