Evidence for convergent nucleotide evolution and high allelic turnover rates at the complementary sex determiner (csd) gene of western and Asian honey bees

Our understanding of the impact of recombination, mutation, genetic drift and selection on the evolution of a single gene is still limited. Here we investigate the impact of all of these evolutionary forces at the complementary sex determiner (csd) gene which evolves under a balancing mode of selection. Females are heterozygous at the csd gene and males are hemizygous; diploid males are lethal and occur when csd is homozygous. Rare alleles thus have a selective advantage, are seldom lost by the effect of genetic drift and are maintained over extended periods of time when compared to neutral polymorphisms. Here, we report on the analysis of 17, 19 and 15 csd alleles of Apis cerana, Apis dorsata and Apis mellifera honey bees respectively. We observed great heterogeneity of synonymous (pi S) and nonsynonymous (pi N) polymorphisms across the gene, with a consistent peak in exon 6 and 7. We propose that exons 6 and 7 encode the potential specifying domain (csd-PSD) which has accumulated elevated nucleotide polymorphisms over time by balancing selection. We observed no direct evidence that balancing selection favors the accumulation of nonsynonymous changes at csd-PSD (pi N/pi S ratios are all < 1, ranging from 0.6 to 0.95). We observed an excess of shared nonsynonymous changes, which suggests that strong evolutionary constraints are operating at csd-PSD resulting in the independent accumulation of the same nonsynonymous changes in different alleles across species (convergent evolution). Analysis of a csd-PSD genealogy revealed relatively short average coalescence times (~6 million years), low average synonymous nucleotide diversity (pi S < 0.09) and a lack of trans-specific alleles which substantially contrasts with previously analyzed loci under strong balancing selection. We excluded the possibility of a burst of diversification after population bottlenecking and intragenic recombination as explanatory factors, leaving high turn-over rates as the explanation for this observation. By comparing observed allele richness and average coalescence times with a simplified model of csd-coalescence, we found that small long term population sizes (i.e. Ne