Fig. 1

Experimental design and overall strategy. Fertile and subfertile bulls are shown in red and blue, respectively. A distribution of the corrected non-return rate at 56 days post-insemination (NRR 56) for all the Montbéliarde bulls born between 2011 and 2014 (519 bulls in total, in grey) and for fertile (n = 57) and subfertile bulls (n = 43) included in the main cohort. B The main cohort comprised 100 bulls with contrasting fertility, based on the differences in corrected NRR 56 (Wilcoxon test, p < 0.05). C for each bull, straws representing several ejaculates prepared for AI were thawed and pooled. Semen functional parameters corrected for batch preparation effects (see “Methods”) were assessed on these pools and no significant difference in mitochondrial status, motility or viability could be detected between fertile and subfertile bulls (Wilcoxon test, p ≥ 0.05). D DNA methylation was analyzed on these samples by RRBS, enabling the selection of a subset of CpGs at which DNA methylation could be measured with sufficient precision (CpGs10). These CpGs10 were then subjected to differential analysis and fertility-related DMCs were identified. E Using the methylation values at DMCs, a predictive model for fertility was constructed and validated. F Distribution of the corrected NRR 56 for all the Montbéliarde bulls born between 2009 and 2012 (485 bulls in total, in grey) and for fertile (n = 16) and subfertile bulls (n = 4) included in the independent cohort. G The independent cohort included 20 bulls with contrasting fertility based on the differences in corrected NRR 56 (Wilcoxon test, p < 0.05), and was only used to evaluate the potential for field application of the predictive model built on the main cohort. H Semen functional analysis and RRBS were performed on one ejaculate per bull. The model previously built on the main cohort was applied to the methylation values obtained in the independent cohort at CpGs10 identified as DMCs using the main cohort. Model quality indicators assessing the consistency between the actual and predicted fertility were then calculated for both the main and independent cohorts