Strong mutational bias toward deletions in the Drosophila melanogaster genome is compensated by selection

Abstract

Insertions and deletions (collectively indels) obviously have a major impact on genome evolution. However, before large-scale data on indel polymorphism became available, it was difficult to estimate the strength of selection acting on indel mutations. Here, we analyze indel polymorphism and divergence in different compartments of the Drosophila melanogaster genome: exons, introns of different lengths, and intergenic regions. Data on low-frequency polymorphisms indicate that 0.036-0.039 short (1-30 nt) insertion mutations and 0.085-0.092 short deletion mutations, with mean lengths 3.23 and 4.78, respectively, occur per single-nucleotide substitution. The excess of short deletion over short insertion mutations implies that indel mutations of these lengths should lead to a loss of approximately 0.30 nt per single-nucleotide replacement. However, polymorphism and divergence data show that this deletion bias is almost completely compensated by selection: Negative selection is stronger against deletions, whereas insertions are more likely to be favored by positive selection. Among the inframe low-frequency polymorphic mutations in exons, long introns, and intergenic regions, selection prevents a larger fraction of deletions (80-87%, depending on the type of the compartment) than of insertions (70-82%) or single-nucleotide substitutions (49-73%), from reaching high frequencies. The corresponding fractions were the lowest in short introns: 66%, 47%, and 15%, respectively, consistent with the weakest selective constraint in them. The McDonald-Kreitman test shows that 32-46% of the deletions and 60-73% of the insertions that were fixed in the recent evolution of D. melanogaster are adaptive, whereas this fraction is only 0-29% for single-nucleotide substitutions.

Fig. 1.—

Prevalence of insertions and deletions of different lengths per single-nucleotide substitution observed at very low derived allele frequencies (DAF < 3%) in introns of lengths 70–300 nt (A) and in intergenic regions (B). Error bars are 95% confidence intervals based on 1,000 bootstrap trials.

Fig. 2.—

Numbers of polymorphic and fixed indels and single-nucleotide substitutions in genome compartments of different kinds. Top row (A–D), polymorphisms with DAF < 15%; middle row (E–H), polymorphisms with DAF > 15%; and bottom row (I–L), fixed mutations. A, E, I: exons (inframe indels and missense substitutions); B, F, J: intergenic regions; C, G, K: long (>300 nt) introns; and D, H, L: short (70–300 nt) introns. Light gray, insertions; dark gray, deletions; yellow, amino acid substitutions; and blue, single-nucleotide substitutions in noncoding regions. Broken lines show the expected values if polymorphism and the rate of divergence in the compartment were the same as in the short introns. In each panel, the left vertical axis shows the number of indels, and the right vertical axis shows the number of single-nucleotide substitutions.

Fig. 3.—

ξ for indels and single-nucleotide substitutions. Low values of ξ correspond to a high fraction of deleterious mutations and vice versa. (A) Mean ξ in different genome compartments: exons (inframe indels or missense substitutions), intergenic regions, long introns (>300 nt), and short introns (70–300 nt). (B) ξ for indel mutations in short and very short introns, depending on the length of the intron after the indel. Light gray, insertions; dark gray, deletions; yellow, missense substitutions; and blue, single-nucleotide substitutions in noncoding regions. Error bars are 95% confidence intervals based on 1,000 bootstrap trials.

Fig. 4.—

α for indels and single-nucleotide substitutions in the different genome compartments: exons (inframe indels or missense substitutions), intergenic regions, long introns (>300 nt), and short introns (70–300 nt). Light gray, insertions; dark gray, deletions; yellow, missense substitutions; and blue, single-nucleotide substitutions within the noncoding regions. Error bars are 95% confidence intervals based on 1,000 bootstrap trials.

Fig. 5.—

Negative selection on synonymous sites. Panels A–C show the numbers of single-nucleotide substitutions per nucleotide site. Green bars correspond to mutations within 8–30 nt in introns ≤65 nt, and blue bars correspond to mutations within the 4-fold-degenerate sites. (A) polymorphisms with DAF<15%; (B) polymorphisms with DAF>15%; and (C) fixed mutations. In panel D, each bar is the ratio of the per-site numbers of single-nucleotide substitutions observed within the 4-fold-degenerate sites and in positions 8–30 of short introns. Error bars are 95% confidence intervals based on 1,000 bootstrap trials.