Recent and recurrent selective sweeps of the antiviral RNAi gene Argonaute-2 in three species of Drosophila

Abstract

Antagonistic host-parasite interactions can drive rapid adaptive evolution in genes of the immune system, and such arms races may be an important force shaping polymorphism in the genome. The RNA interference pathway gene Argonaute-2 (AGO2) is a key component of antiviral defense in Drosophila, and we have previously shown that genes in this pathway experience unusually high rates of adaptive substitution. Here we study patterns of genetic variation in a 100-kbp region around AGO2 in three different species of Drosophila. Our data suggest that recent independent selective sweeps in AGO2 have reduced genetic variation across a region of more than 50 kbp in Drosophila melanogaster, D. simulans, and D. yakuba, and we estimate that selection has fixed adaptive substitutions in this gene every 30-100 thousand years. The strongest signal of recent selection is evident in D. simulans, where we estimate that the most recent selective sweep involved an allele with a selective advantage of the order of 0.5-1% and occurred roughly 13-60 Kya. To evaluate the potential consequences of the recent substitutions on the structure and function of AGO2, we used fold-recognition and homology-based modeling to derive a structural model for the Drosophila protein, and this suggests that recent substitutions in D. simulans are overrepresented at the protein surface. In summary, our results show that selection by parasites can consistently target the same genes in multiple species, resulting in areas of the genome that have markedly reduced genetic diversity.

FIG. 1.

Genomic positions and gene trees for loci surrounding AGO2. In each tree, the upper clade is Drosophila melanogaster and the lower clade is D. simulans. The size and position of amplified regions are shown by white boxes, and gray boxes show the corresponding genes (note that the amplified fragment from Yellow-k partially overlaps locus CG7945). The total length of the surveyed region was approximately 123 kbp, and the total length of amplified sequence per individual was approximately 13.5 kbp. Gene trees were constructed using neighbor joining (MEGA v. 3.1, Kumar et al. 2004), based on coding sites only and were rooted using D. yakuba. All trees are drawn to the same scale. The shallow within-species genealogy associated with recent selective sweeps in AGO2 is clear in both species.

FIG. 2.

Genetic diversity around AGO2. Genetic diversity at all sites (upper row) and synonymous sites (middle row) is considerably reduced around AGO2 (positioned at zero on the x axis) in all three species. This is also reflected in the diversity/divergence ratio at synonymous sites (lower row: synonymous site diversity within species, θ_s, divided by divergence between species, K_S). Loci that are significantly different from AGO2 in individual pairwise HKA tests (Hudson et al. 1987) are marked on the diversity/divergence graphs with asterisks: *P < 0.05, **P < 0.01, ***P < 0.001.

Fig. 3.

Composite likelihood profile. The CLR between a standard neutral model and selective sweep model, considering each site in turn (Li and Stephan 2005). The region surrounding AGO2 is shown for Drosophila melanogaster (upper panel) and D. simulans (lower panel). Gray regions are those for which sequence data are available, and the thin horizontal line shows the most stringent 5% significance threshold for this statistic derived a range of plausible population-growth scenarios (see Materials and Methods). The maximum likelihood estimate of the focal site for the sweep is given by a vertical dashed line; note that under this model there is no significant evidence of a recent sweep in D. melanogaster but that D. simulans shows strong evidence of a sweep in AGO2.

FIG. 4.

Recent amino acid substitutions in D. simulans AGO2. The surface structure of Drosophila AGO2 derived from published archean and Drosophila Argonaute structures by fold-recognition and homology modeling (see Materials and Methods). Moving down the figure, the four panels are successive 90° rotations about the vertical axis. The PAZ domain is indicated in green, the PIWI domain is indicated in blue, and the amino acid substitutions that occurred in D. simulans since the split from D. sechellia (ca. 250 kya) are shown in red. The two remaining substitutions at L106 and S404 are buried within the structure (see also supplementary fig. S2B, Supplementary Material online).