The major histocompatibility complex (MHC), a genomic region containing genes crucial to immunity 1, is known to be involved in mate choice in numerous vertebrate species. More precisely, it has been shown that mate choice is influenced by MHC in ways that favor the mating of MHC-dissimilar individuals 2. This pattern is believed to improve the immune resistance of the offspring conceived through such matings. In humans, the question of MHC-dependent mate choice remains highly controversial (see 3 for a review). Nowadays, through the availability of high-density genomic datasets providing genotypes for couples, such as the HapMap datasets 4, 5, MHC-dependent mating patterns in humans can be re-considered.
In a recent commentary 6, Derti and Roth wrote that the “HapMap genotypes do not confidently support a role for the MHC locus in human mate selection”. This statement follows a previous article of Derti et al. 7 entitled “Absence of evidence for MHC-dependent mate selection within HapMap populations”. This article and commentary contrast with previous publications by Chaix et al. 8 and by us 9 reporting lines of evidence compatible with MHC-disassortative mating among HapMap European American spouses.
More specifically, Derti and Roth raise the following main criticisms:
- i)Outliers may drive the observed significance
- ii)This significance is lost when minor modifications are made to the analysis procedure
- iii)In the HapMap 2 dataset, non-mate pairs show a dissimilarity signal as strong as mates.
Here, we reaffirm our viewpoint that such criticisms are unjustified for the following reasons:
- i)Regarding outliers, the distribution of MHC relatedness among mates shows no such outliers, and Derti et al. agree on this. Consequently, we see no reason for excluding the most MHC dissimilar couple from the distribution, as done by Derti et al. Similarly, Derti et al. propose excluding opposite-sex non-mate pairs showing extreme MHC similarity. Even if the distribution of MHC relatedness coefficients among opposite-sex non-mates appears to be skewed toward MHC similarity, it does not show outliers. Moreover, close relatives were already excluded from the analysis. Consequently, we see no reason for excluding the most MHC-similar opposite-sex non-mates. The sample size of the HapMap datasets are considerably smaller than sample sizes usually used in behavioral/evolutionary ecology studies, and removing even a single couple from the analysis is expected to decrease significance. Moreover, assuming the significance was actually driven by outliers, as claimed by Derti et al., this could mean that MHC influences some cases of mate choice, not all, which would still be an interesting biological effect.
- ii)Similarly, some of the other modifications made by Derti et al. to our analysis procedure are unjustified: to quantify the MHC dissimilarity among spouses, Derti et al. computed the median over all couples rather than the mean. Given the sample size, we feel it is not surprising that the significance is decreased when using less powerful statistics. On the other hand, Derti et al. increased the number of permutations, which seems a relevant proposition to us, and significance is conserved when doing so (p = 0.022, reported by Derti et al.), even after multiple testing correction.
- iii)Derti and Roth report that in the HapMap 2 dataset the MHC locus is almost as extremely dissimilar in non-mates as it is in mates. To support this statement, they report that the fraction of genomic segments with relatedness lower than the MHC locus is 0.6 and 3.9%, respectively, for mates and non-mates, and that “similar results were obtained when considering the recombination rate (not shown)”. We cannot agree with this latter statement since when considering the recombination rate, the fractions are actually 1.6 and 6.3%. Taking the recombination rate into account is essential since the statistical properties of the relatedness coefficient depend on it 10, 11. Consequently, we reaffirm here that in the HapMap2 and HapMap3 European American datasets, the MHC relatedness among mates, but not among opposite-sex non-mates, is extreme in comparison to the genome.
Given the points above, contrary to Derti and Roth, we believe that the HapMap European American genotypes are compatible with the hypothesis of MHC-disassortative mating. Mate choice is a complex process involving biological and cultural factors. Larger datasets collected throughout a large number of populations may help unravel the impact of biology and culture on human mate choice.