Assessing genotype by feed interactions for milk production traits in dairy cattle

Dairy farming systems are evolving toward more sustainable production practices, and these changes may enhance genotype-by-environment interaction (G × E). Analyses of G × E would benefit from detailed information on feeding practices, however, such data are rarely incorporated, largely because they are not routinely available. This study investigated G × E for milk production traits using detailed information on feeding practices using reaction norm models, and assessed differences between organic and conventional farming systems using character state models. Environmental descriptors were available for 973 Dutch dairy farms over the period 2016-2021. These descriptors included the proportion of dietary concentrate (% of diet DM), nitrogen use efficiency (NUE, %), phosphorus use efficiency (PUE, %), and dietary net energy for lactation (NE_L, MJ/kg DM). For comparison, we also included the more commonly used descriptor herd average milk yield per cow (HMY, kg/cow/year). The data set included 101,979 first lactation milk production records, of which 6,094 originated from organic farms and 95,885 from conventional farms. Heritability (±SE) of milk yield was 0.36 ± 0.04 in organic and 0.44 ± 0.01 in conventional farming systems. Heritability of milk yield increased from 0.17 ± 0.02 when feeding 13.0% to 0.54 ± 0.01 when feeding 43.0% dietary concentrate (DM basis). Similarly, heritability of milk, fat, and protein yield increased with increasing NUE, PUE, NE_L and HMY. In our study, the residual variance was assumed constant across environmental descriptors; therefore, the increase in heritability was due to an increase in genetic variance. Heritability of milk fat and protein content and somatic cell score showed limited response to changes in environmental descriptors. Genetic correlations between the same trait across environmental descriptors were generally near unity, except for yield traits across HMY levels. These results indicate G × E arising from scaling effects, but provide little evidence of genotype re-ranking across environments. Ignoring scaling effects leads to biased EBVs for environments deviating from the average. Multivariate analyses that include traits beyond milk production are needed to quantify the potential implications for selective breeding programs.