Population-based resequencing of experimentally evolved populations reveals the genetic basis of body size variation in Drosophila melanogaster

Abstract

Body size is a classic quantitative trait with evolutionarily significant variation within many species. Locating the alleles responsible for this variation would help understand the maintenance of variation in body size in particular, as well as quantitative traits in general. However, successful genome-wide association of genotype and phenotype may require very large sample sizes if alleles have low population frequencies or modest effects. As a complementary approach, we propose that population-based resequencing of experimentally evolved populations allows for considerable power to map functional variation. Here, we use this technique to investigate the genetic basis of natural variation in body size in Drosophila melanogaster. Significant differentiation of hundreds of loci in replicate selection populations supports the hypothesis that the genetic basis of body size variation is very polygenic in D. melanogaster. Significantly differentiated variants are limited to single genes at some loci, allowing precise hypotheses to be formed regarding causal polymorphisms, while other significant regions are large and contain many genes. By using significantly associated polymorphisms as a priori candidates in follow-up studies, these data are expected to provide considerable power to determine the genetic basis of natural variation in body size.

Figure 1. Response to selection.

Above: Distribution of thorax length in each population after >100 generations of selection. Below: Distribution of sieve depth in each population. Median (line), 75% quartile (box), and range (whiskers) are shown. C1, C2 = controls, L1, L2 = large lines, S1, S2 = small lines.

Figure 2. Frequency histogram of differentiation between populations, on a log scale.

Pair-wise comparisons are shown between the two control populations, and between the two independent comparisons of a large- and small-selected population. Red = control population 1 versus control population 2; green = large population 1 versus small population 1; blue = large population 2 versus small population 2.

Figure 3. Frequency histogram of differentiation between treatments, on a log scale.

The distribution of the diffStat is shown, when the two large- and two small-selected populations are considered together; this is compared to the expected distribution obtained via simulation. Blue = observed, red = simulated drift. Note the log scale.

Figure 4. Differentiation on chromosome arm 3R.

The diffStat is shown for each variant that had higher or lower allele frequencies in the large-selected lines compared to the small-selected lines. Above: Color coding indicates significance: black = nonsignificant variants, blue = significant variants at the permissive FDR threshold (FDR<10%); gold = significant variants at the restrictive FDR threshold (FDR<5%); red = peak variants. Below: Color coding indicates estimated starting allele frequency: black = all variants, gold = variants with an average control frequency <0.05; red circles indicate peak variants, as in A.

Figure 5. Fine-scale mapping of candidate causal variants.

The diffStat is shown for each variant at each locus; black = nonsignificant variants, blue = FDR<10%; purple = FDR<5%; red = peak variants. Above: Eip63E; Below: dre4. For the candidate gene at each locus, the exons are shown as linked grey boxes; only one transcript is shown for simplicity. The arrow in B indicates a serine-tryptophan replacement discussed in the text.