Hitchhiking effects of recurrent beneficial amino acid substitutions in the Drosophila melanogaster genome

Abstract

Several recent studies have estimated that a large fraction of amino acid divergence between species of Drosophila was fixed by positive selection, using statistical approaches based on the McDonald-Kreitman test. However, little is known about associated selection coefficients of beneficial amino acid mutations. Recurrent selective sweeps associated with adaptive substitutions should leave a characteristic signature in genome variability data that contains information about the frequency and strength of selection. Here, I document a significant negative correlation between the level and the frequency of synonymous site polymorphism and the rate of protein evolution in highly recombining regions of the X chromosome of D. melanogaster. This pattern is predicted by recurrent adaptive protein evolution and suggests that adaptation is an important determinant of patterns of neutral variation genome-wide. Using a maximum likelihood approach, I estimate the product of the rate and strength of selection under a recurrent genetic hitchhiking model, lambda2N(e)s approximately 3 x 10(-8). Using an approach based on the McDonald-Kreitman test, I estimate that approximately 50% of divergent amino acids were driven to fixation by positive selection, implying that beneficial amino acid substitutions are of weak effect on average, on the order of 10(-5) (i.e., 2N(e)s approximately 40). Two implications of these results are that most adaptive substitutions will be difficult to detect in genome scans of selection and that population size (and genetic drift) may be an important determinant of the evolutionary dynamics of protein adaptation.

Figure 1.

Reduced levels of synonymous site diversity associated with in high rates of protein evolution in D. melanogaster. (A) Raw estimates of synonymous site diversity (π_S) vs. nonsynonymous divergence (K_a). (B) Partial correlation residuals of π_S versus K_a, correcting for K_S. Negative residuals have been set to zero. (Dashed lines) Weighted average diversity across loci.

Figure 2.

Reduced levels of codon usage bias associated with high rates of protein evolution. Plotted is the frequency of preferred codons (Fop) versus partial regression residuals of K_a (corrected for K_s). (Dashed line) Average Fop across loci.

Figure 3.

Joint likelihood surface for θ and αs under the recurrent hitchhiking model. (White circle) Maximum likelihood estimate. Each contour represents two units of log likelihood.

Figure 4.

The fit of the recurrent hitchhiking model to the data using maximum likelihood estimates of θ and αs, and locus-specific parameters. (A) The fit to average levels of synonymous site diversity (π_s). (B) The fit to average Tajima’s D. Observed and simulated data are binned into three categories based on levels of amino acid divergence (K_a). Plotted are the mean of 1000 simulated replicates (white circles, gray lines) and the mean for the observed data (black circles, solid lines) with standard errors. (Horizontal dotted lines) Expectation under the standard neutral model with recombination.

Figure 5.

The fraction of amino acid mutations fixed by positive selection (α) versus their selective advantage s for αs = 6 × 10⁻⁶ (solid line). (Arrow) Maximum likelihood estimate of α using the approach of Bierne and Eyre-Walker (2004) (see Supplemental materials 6), (shaded area) 95% CI for the estimate of α, (vertical dotted line) minimum possible selection coefficent (α = 1; s = 6 × 10⁻⁶).