Characterizing recurrent positive selection at fast-evolving genes in Drosophila miranda and Drosophila pseudoobscura

Abstract

Characterizing the distribution of selection coefficients in natural populations remains a central challenge in evolutionary biology. We resequenced a subset of 19 fast-evolving protein-coding genes in the sister species Drosophila miranda and D. pseudoobscura and their flanking regions to characterize the spatial footprint left by recurrent and recent selection. Consistent with previous findings, fast-evolving genes and their flanking regions show reduced levels of neutral diversity compared with randomly chosen genes, as expected under recurrent selection models. Applying a variety of statistical tests designed for the detection of selection at different evolutionary timescales, we attempt to characterize parameters of adaptive evolution. In D. miranda, fast-evolving genes generally show evidence of increased rates of adaptive evolution relative to random genes, whereas this pattern is somewhat less pronounced in D. pseudoobscura. Our results suggest that fast-evolving genes are not characterized by significantly different selection coefficients but rather a shift in the distribution of the rate of fixation.

FIG. 1.—

Plots of the expected reduction in variation due to recurrent hitchhiking based upon the theoretical prediction of Wiehe and Stephan (1993): where θ = 4N_eμ (where N_e is the effective population size and μ is the mutation rate per site per generation), κ = 0.075 (a constant), λ = the rate of adaptive substitutions per site per generation, and γ = 2N_es (where s is the selection coefficient). By fixing μ, r, and λ, the effects of varying N_e on levels of polymorphisms are plotted for different values of s.

FIG. 2.—

Plot of synonymous diversity π_s versus amino acid divergence K_a. (a) Drosophila miranda: The pooled loci of both this study as well as the randomly selected genes of Bachtrog et al. (2009) are shown. (b) D. pseudoobscura: The pooled loci of homologous genes across the X chromosome. The solid line indicates the significant correlation between these measures of synonymous polymorphism and nonsynonymous divergence.

FIG. 3.—

A visual schematic of both the empirical and the simulation approaches taken here. On the x axis, the black box represents the coding region. In gray, a pair of regions are sequenced 5 kb up- and downstream. In white, another pair of regions are sequenced 10 kb up- and downstream. Table 2 compares the following three results: 1) black region only, 2) black + gray regions, and 3) black + gray + white regions. The solid line gives the expected levels of variation at these distances for a selective sweep occurring at position 0 (i.e., in the coding region)—where the length of the sequence = 1 kb, the beneficial mutation fixed at time (τ) = 0.0001 4N generations, θ = 0.01, ρ = 0.1, and α = 2Ns = 1000. Ten thousand replicates were simulated under each scenario. Overlaid are the observed relative reductions in variation among the 19 fast-evolving genes at each of the five regions for Drosophila miranda (dashed) and D. pseudoobscura (dotted).

FIG. 4.—

Approximate Bayesian estimation of both the strength and the rate of recurrent positive selection among 19 fast-evolving genes, compared with randomly selected genes, in Drosophila miranda and D. pseudoobcura. Estimation is based on 10⁶ draws from the prior. Given are the marginal distributions, with fast-evolving genes in black and random genes in gray. Although there is no significant difference in the estimated strength of selection, there is over an order of magnitude increase in the estimated rate of selection among fast-evolving genes in D. miranda relative to random genes. Consistent with observed levels of reduction, frequency spectrum, and LD-based tests of selection, this hitchhiking pattern—though qualitatively similar—appears less pronounced in D. pseudoobscura among the fast-evolving data set.