Intragenic Hill-Robertson interference influences selection intensity on synonymous mutations in Drosophila

Natural selection influences synonymous mutations and synonymous codon usage in many eukaryotes to improve the efficiency of translation in highly expressed genes. Recent studies of gene composition in eukaryotes have shown that codon usage also varies independently of expression levels, both among genes and at the intragenic level. Here, we investigate rates of evolution (Ks) and intensity of selection (gamma(s)) on synonymous mutations in two groups of genes that differ greatly in the length of their exons, but with equivalent levels of gene expression and rates of crossing-over in Drosophila melanogaster. We estimate gamma(s) using patterns of divergence and polymorphism in 50 Drosophila genes (100 kb of coding sequence) to take into account possible variation in mutation trends across the genome, among genes or among codons. We show that genes with long exons exhibit higher Ks and reduced gamma(s) compared to genes with short exons. We also show that Ks and gamma(s) vary significantly across long exons, with higher Ks and reduced gamma(s) in the central region compared to flanking regions of the same exons, hence indicating that the difference between genes with short and long exons can be mostly attributed to the central region of these long exons. Although amino acid composition can also play a significant role when estimating Ks and gamma(s), our analyses show that the differences in Ks and gamma(s) between genes with short and long exons and across long exons cannot be explained by differences in protein composition. All these results are consistent with the Interference Selection (IS) model that proposes that the Hill-Robertson (HR) effect caused by many weakly selected mutations has detectable evolutionary consequences at the intragenic level in genomes with recombination. Under the IS model, exon size and exon-intron structure influence the effectiveness of selection, with long exons showing reduced effectiveness of selection when compared to small exons and the central region of long exons showing reduced intensity of selection compared to flanking coding regions. Finally, our results further stress the need to consider selection on synonymous mutations and its variation--among and across genes and exons--in studies of protein evolution.