Reconciling Differences in Pool-GWAS Between Populations: A Case Study of Female Abdominal Pigmentation in Drosophila melanogaster

Abstract

The degree of concordance between populations in the genetic architecture of a given trait is an important issue in medical and evolutionary genetics. Here, we address this problem, using a replicated pooled genome-wide association study approach (Pool-GWAS) to compare the genetic basis of variation in abdominal pigmentation in female European and South African Drosophila melanogaster. We find that, in both the European and the South African flies, variants near the tan and bric-à-brac 1 (bab1) genes are most strongly associated with pigmentation. However, the relative contribution of these loci differs: in the European populations, tan outranks bab1, while the converse is true for the South African flies. Using simulations, we show that this result can be explained parsimoniously, without invoking different causal variants between the populations, by a combination of frequency differences between the two populations and dominance for the causal alleles at the bab1 locus. Our results demonstrate the power of cost-effective, replicated Pool-GWAS to shed light on differences in the genetic architecture of a given trait between populations.

Keywords: Drosophila melanogaster; GWAS; dominance; genome-wide association study; pigmentation; pooled sequencing.

Figure 1

Manhattan plots for the effect of SNPs on female abdominal pigmentation for segment A7. The horizontal axis shows the genomic location of each tested SNP, with chromosomal location indicated alternately by black and gray and with vertical lines indicating the locations of the tan (red), pdm3 (yellow), bab1 (green), and ebony (blue) genes. The vertical axis shows the –log₁₀ P-value from the CMH test, with a horizontal line indicating the 0.05 FDR threshold. (A–D) Results for the experiments with European (A) and South African (B) flies, with detailed views of regions of the tan (C) and bab (D) loci that contain SNPs showing strong associations. In the detailed views, only SNPs with P-values <0.1 are shown. Gray rectangles mark regions containing coding sequence of the genes, and white rectangles indicate known cis-regulatory elements. The arrowheads indicate the direction of transcription. Male-specific enhancer (MSE), dimorphic enhancer (DME), and anterior element (AE) indicate previously described cis-regulatory regions (Jeong et al. 2008; Williams et al. 2008).

Figure 2

Consistency between European and South African samples. SNPs were ranked by CMH test P-values, with different cutoffs shown on the x-axis. SNPs were ranked according to the results in the European (red line) or South African populations (blue line). One thousand random samples of unassociated SNPs—matched in number, allele frequency, and chromosomal distribution with the analyzed SNPs—were used as a comparison (dashed lines). (A) The fraction of SNPs with consistent effects, i.e., where light and dark alleles inferred between the two samples were the same. (B) Correlation coefficients (Spearman’s ρ) for the logOR values of the two populations. The logOR is a measure of the effect of the SNP, with the direction of the effect indicated by its sign.

Figure 3

Rejection sampling results. Heatmap shows the percentage of accepted simulations using an ABC rejection method with 2,000,000 simulations uniformly sampling parameter space and an acceptance rate of 1%, using these criteria: logOR(tan)/logOR(bab1), Europe = 2.51, South Africa = 0.68; logOR(tan_EU)/logOR(tan_SA) = 2.28; and logOR(bab1_EU)/logOR(bab1_SA) = 0.62. Parameters were sampled uniformly from [−0.125, 1.125] for dominance and [0.25, 4.5] for the additive effect of bab1 relative to tan. For presentation in a grid, the parameter values were binned into intervals of 0.25 for dominance and 1 for relative additive effect. The axis labels indicate the center of each binning interval. The numbers in the cells give the percentage of accepted simulations in each binned parameter interval. Each cell represents a particular range of combinations of dominance of the light alleles at bab1 (top axis) and tan (bottom axis) and the relative additive effect of bab1 to tan loci (left axis).