Predictive ability and mediation effects of the rumen microbiome on feed efficiency and methane traits in Hereford beef cattle

The ruminant genome exerts moderate control over rumen microbial composition, which is a major determinant of feed efficiency and methane emissions. However, the integration of new omics data, whether for phenotypic prediction, selective breeding, or both, is still under discussion. This study aimed to (1) estimate the heritability and microbiability for residual feed intake (RFI), dry matter intake (DMI), BW, and methane yield (MY); (2) assess the predictive ability of models including host genome and microbiome information; and (3) evaluate the mediation effects of the rumen microbiome on feed efficiency and performance traits. The data set consisted of 537 Hereford bulls and steers with RFI, DMI, and BW records, as well as a subset of them (n = 123) with MY records. All animals were genotyped using 100 k single-nucleotide polymorphism panels, and rumen microbial abundances were determined through enzyme-restriction reduced representation sequencing (ER-RRS) analysis. Heritability estimates ranged from 0.18 for RFI to 0.36 for BW, while microbiability values were moderate (0.16-0.32), indicating that both host genetics and the microbiome significantly contribute to trait variation. We found that the use of genome and rumen microbiome information improved predictive ability for BW (r = 0.48-0.52), as assessed by Pearson's correlation between observed and predicted values, but not for RFI (r = 0.15-0.14), DMI (r = 0.72-0.73), or MY (r = 0.68-0.69). We also identified several amplicon sequence variants (ASVs) with moderate genetic control and potential mediation effects on RFI, DMI, and BW. However, given the large number of tests performed, these findings should be interpreted with caution due to the increased risk of false positives. Interestingly, our findings show better results for the use of rumen microbiome for selective breeding than for phenotypic prediction in beef cattle. Additionally, we highlighted that using genomics and rumen microbiome data within structural equation models provides new biological insights into animal performance. However, the assumption of no environmental covariance between the host and the microbiome is strict but necessary. Further exploration includes the use of instrumental auxiliary variables that allow for the inclusion of the environmental covariance between the traits of interest.