Higher male than female recombination rate in cattle is controlled by genetic variants effective in both sexes

We herein study genetic recombination in three dairy cattle populations from France, New-Zealand and the Netherlands. We apply a new phasing algorithm extracting familial information suited for large half-sib families to reconstruct haplotypes and detect cross-overs (CO). The software is robust to genotyping and map errors. We identify more than 2,000,000 CO events in sperm cells transmitted by 3008 sires to 94,603 offspring, and more than 500,000 CO events in oocytes transmitted by 11,497 cows to 25,390 offspring. When measured in identical family structures, the average number of CO in males (24.0) was found to be larger than in females (21.8). In males, recombination rates were higher closer to telomeres whereas in females, recombination rates dropped at both centromeres and telomeres (probably as a result of lower informativity). The heritability of the global recombination rate (GRR) was close to 0.20 in males and to 0.08 in females. Genetic correlation ranged from 0.38 to 0.69 depending on the population, indicating that shared variants are influencing GRR in both genders. Haplotype-based genome-wide association studies revealed four genome-wide significant QTL, including two previously identified ones (involving REC8 and RNF212). For all QTLs, there was a positive correlation between haplotype effects across sexes, ranging from 0.35 to 0.68. We selected two reference panels of respectively 122 and 215 bulls sequenced at cover > 15x to impute variants in the New-Zealand and French populations. All variants identified by next-generating sequencing in 5 Mb windows encompassing the QTL peaks were imputed with Beagle in order to perform a sequence-based association study. For three QTLs, we identified missense mutations in genes known to be involved in meiosis among the most significantly associated variants. These variants were perfectly associated with the haplotypes underlying the QTL effects. The variant identified in RNF212 had already been reported, whereas missense mutations in MLH3 (N408S) and HFM1 (S1189L) are new findings. Surprisingly, variants previously identified in REC8 did not capture the QTL effect whereas variants in RNF212B, PPP1R3E, BCL2L2, HOMEZ and PABPN1 had much stronger association with the phenotype. The three missense mutations were significant in both genders with two of them accounting for approximately 10% of the genetic variance in males (the allelic substitution effect being approximately equal to one additional CO per genome). Our results are very different from reports of recombination in other species. For instance, in human, recombination rate is higher in females, distinct variants affect recombination rate in males and females and the genetic correlation is close to 0 whereas in cattle, we observed a higher recombination rate in males controlled by shared variants effective in both sexes.