Surveys of DNA sequence variation have shown that the level of genetic variability in a genomic region is often strongly positively correlated with its rate of crossing over (CO) [1][2][3]. This pattern is caused by selection acting on linked sites, which reduces genetic variability and can also cause the frequency distribution of segregating variants to contain more rare variants than expected without selection (skew). These effects of selection may involve the spread of beneficial mutations (selective sweeps, SSWs), the elimination of deleterious mutations (background selection, BGS) or both together, and are expected to be stronger with lower rates of crossing over [1][2][3]. However, in a recent study of human populations, the skew was reduced in the lowest CO regions compared with regions with somewhat higher CO rates [4]. A similar pattern is seen in the population genomic studies of Drosophila simulans described here. We propose an explanation for this paradoxical observation, and validate it using computer simulations. This explanation is based on the finding that partially recessive, linked deleterious mutations can increase rather than reduce neutral variability when the product of the effective population size (N e ) and the selection coefficient against homozygous carriers of mutations (s) is ≤ 1, i.e. there is associative overdominance (AOD) rather than BGS [5]. We show that AOD can operate in a genomic region with a low rate of CO, opening up a new perspective on how selection affects patterns of variability at linked sites.
Results and Discussion
Diversity Statistics in Relation to CO Rates in Drosophila simulansThe top two panels of Figure 1 show the relations between the rate of crossing over and mean pairwise diversity at four-fold degenerate nucleotide sites (π 4 ), for two populations of Drosophila simulans. Consistent with previous studies of many species [1-3, 6], there is a significant positive relation between CO rate and nucleotide site diversity for both X chromosome (X) and autosomes (A). As reported previously [7], X has much lower diversity than A for all bins of CO rates. Diversity is higher for the Madagascan than the Kenyan population, consistent with the latter having been founded as a relatively small population by flies descended from the putatively ancestral Madagascan population [7].The bottom panels of Figure 1 show the relations between CO rate and a measure of the skew of the site frequency spectrum (SFS) towards rare variants, Δθwhere π 4 and θw4 are the estimates of diversity based on the mean pairwise difference between alleles [8] and the number of segregating sites [9], respectively. Other measures of skew behave similarly (Supplemental Figure S1). As was also found in a Rwandan population of D. melanogaster [10], there is much greater skew on X than A, although the absolute level of skew for both X and A is much higher than in D. melanogaster, suggesting a more intense recent population expansion in D. simulans. The skew is also larger in the Madagascan than the Kenya...