A Large Panel of Drosophila simulans Reveals an Abundance of Common Variants

Abstract

The rapidly expanding availability of large NGS data sets provides an opportunity to investigate population genetics at an unprecedented scale. Drosophila simulans is the sister species of the model organism Drosophila melanogaster, and is often presumed to share similar demographic history. However, previous population genetic and ecological work suggests very different signatures of selection and demography. Here, we sequence a new panel of 170 inbred genotypes of a North American population of D. simulans, a valuable complement to the DGRP and other D. melanogaster panels. We find some unexpected signatures of demography, in the form of excess intermediate frequency polymorphisms. Simulations suggest that this is possibly due to a recent population contraction and selection. We examine the outliers in the D. simulans genome determined by a haplotype test to attempt to parse the contribution of demography and selection to the patterns observed in this population. Untangling the relative contribution of demography and selection to genomic patterns of variation is challenging, however, it is clear that although D. melanogaster was thought to share demographic history with D. simulans different forces are at work in shaping genomic variation in this population of D. simulans.

Keywords: Drosophila simulans; population genetics; selection.

Fig. 1.—

Summary statistics. (a) For each chromosome, Tajima’s D (pink) and$\pi$ (blue) are plotted in 10-kb windows. The genomic coordinates are split into 10-kb windows along the x-axis. Each data point represents a single window. Note that the axis for π (blue) is on the left-hand side while Tajima’s D (pink) is on the right. For clarity, chromosome 4 is on a different y-axis for π, due to its small size. Location of centromeres and telomeres for each chromosome are indicated at the lower edges of each graph. All lines were smoothed using local regression in R. (b) The median and distribution (boxes are first and third quartile) of Tajima’s D for each annotation category. Notches at the median are calculated as described in boxplot.stats in R and correspond roughly to the 95% confidence interval for the medians.

Fig. 2.—

Each simulation was performed for a region 10⁵ bp and Tajima’s D was calculated in windows of 10 kb. Recombination was $\rho =$5×10⁻⁷ and mutation $\mu =$3×10⁻⁹. Here, Tajima’s D in demographic simulations is shown compared with Drosophila simulans. (a) Admixture simulations, including admixture with a bottleneck prior to splitting, and a bottleneck along one lineage prior to admixing (bottleneck with admixture), and admixture with bottlenecks in both populations after isolation (admixture with bottleneck). (b) Constant population size and population contractions of different degrees and durations. A recent bottleneck to 0.1% for 60 years resembles the data as much as slightly older bottlenecks, and is used in concert with the simulations of selection in the following figure. (c) Tajima’s D in different simulations of selection compared with D. simulans. Not all of the simulations listed in supplementary table S2, Supplementary Material online, are included here, only those that most resemble the D. simulans data or illustrate the difference between selection alone and selection with demography. All population contractions were to 0.1% of the population for 60 years, terminating 40 years prior to sampling. For all simulations s = 0.1, ƒ refers to the starting frequency of the selected mutation, and t is its time of introduction with units of 4N_e.

Fig. 3.—

Linkage disequilibrium. (a) Decay of linkage disequilibrium on the X and autosomes calculated by distance. LD is binned initially in windows of 20 bp until 300 bp and 150 bp thereafter. Shown is the log(LD) of the mean of each of these bins. Regions of low recombination have been filtered out. (b) The relationship between LD and outliers of Tajima’s D with regression lines plotted for negative and positive values of Tajima’s D separately. This is to highlight the possibility for different relationships between LD and Tajima’s D depending upon the selective regime—for example, that lower (higher) values of Tajima’s D may be associated with higher (lower) values of LD. The R² value of for each of these lines is shown in the upper right hand corner of each graph.

Fig. 4.—

Haplotype Diversity. Shown are the results of the H12 scan for each chromosome other than the fourth, due to its overall reduced recombination rate. On the right-hand side are the scans in windows of 400 bases, with window centers separated by 50 SNPs. Gray regions indicate regions of the chromosome arms with reduced recombination that were not included in the final analysis. Red points indicate the top peaks for each chromosome. The dotted red lines indicate two positive controls from Drosophila melanogaster, Cyp6g1, and Ace. On the left-hand side, the haplotype frequency spectra for the top ten intervals from each chromosome are shown. The height of the top shaded region in light blue indicates the proportion of that haplotype in the population. Colors after that indicate the second, third, and so on most frequent haplotypes while regions in gray indicate singletons. Most sweeps, with the exception of Cyp6g1, have second haplotypes sorting at appreciable frequencies.

Fig. 5.—

Demographic simulations and H12. (a) The spread of H12 values in each of the demographic scenarios that were simulated. Demography does not substantially elevate Tajima’s D though population contraction does have more haplotype structure than the other demographic scenarios. (b) An example of the haplotype structure under each of the demographic scenarios, with the highest H12 value for each scenario shown. Again, the most frequent haplotype is shown in light blue and the second most frequent in dark blue. Gray regions indicate singletons.

Fig. 6.—

Simulations of selection and H12. (a) The spread of H12 values for each of the simulations of selection. While demography and selection elevated Tajima’s D the most, they do not elevate H12 values the most. This may be due to the timing of selection, which had to be modeled differently for combinations of selection and demography. (b) Patterns of haplotype diversity in the top values of H12 from each simulation and Drosophila simulans. Note that we include the highest values of H12 for D. simulans from the autosomes, as these simulations pertain principally to autosomal variation. These selection scenarios are more similar to that observed in D. simulans than demography, however, the high frequency of the second haplotype in D. simulans was not recapitulated in any simulation.

Fig. 7.—

H12. (a) Tajima’s D versus H12 and H2/H1. The average Tajima’s D value was calculated for intervals identified as the top 50 H12 scores for each chromosome. Intervals are defined as the smallest and largest edges of each peak, thus adjacent intervals with high H12 scores are considered a single peak. The slight negative relationship between H12 and Tajima’s D is significant for all chromosomes (P < 0.0001), and is expected given the decreasing ability of H12 to detect sweeps as they become softer (supplementary fig. S5, Supplementary Material online). A high value of H2/H1 and H12 indicates a soft selective sweep, so a positive relationship between Tajima’s D and H2/H1 indicates that soft selective sweeps are responsible for the high Tajima’s D values. The relationship is significant at P < 0.0003–0.002 for all chromosomes other than 3L (supplementary fig. S6, Supplementary Material online).