Secondary contact and local adaptation contribute to genome-wide patterns of clinal variation in Drosophila melanogaster

Abstract

Populations arrayed along broad latitudinal gradients often show patterns of clinal variation in phenotype and genotype. Such population differentiation can be generated and maintained by both historical demographic events and local adaptation. These evolutionary forces are not mutually exclusive and can in some cases produce nearly identical patterns of genetic differentiation among populations. Here, we investigate the evolutionary forces that generated and maintain clinal variation genome-wide among populations of Drosophila melanogaster sampled in North America and Australia. We contrast patterns of clinal variation in these continents with patterns of differentiation among ancestral European and African populations. Using established and novel methods we derive here, we show that recently derived North America and Australia populations were likely founded by both European and African lineages and that this hybridization event likely contributed to genome-wide patterns of parallel clinal variation between continents. The pervasive effects of admixture mean that differentiation at only several hundred loci can be attributed to the operation of spatially varying selection using an FST outlier approach. Our results provide novel insight into the well-studied system of clinal differentiation in D. melanogaster and provide a context for future studies seeking to identify loci contributing to local adaptation in a wide variety of organisms, including other invasive species as well as temperate endemics.

Keywords: Drosophila melanogaster; adaptation; latitudinal clines; parallelism; secondary contact.

Fig. 1

Map of collection locales (squares) of D. melanogaster populations used in this study.

Fig. 2

Estimated population tree of sampled locales. Loci were sampled across all chromosomes, focusing on SNPs residing within (+), outside (−) large, cosmopolitan inversions or both (±). See Fig. S1 (Supporting information) for population trees based on each chromosome.

Fig. 3

Proportion of African and European ancestry in North American and Australian populations. Thin grey lines represent the ancestry estimate for any particular African or European population averaged over all European and African populations. Black line represents the average ancestry proportions from all African and European populations. The blue circle represents the spring Pennsylvania sample, and the red circle represents the fall Pennsylvanian sample. Correlations (r) between the average ancestry estimates (thick black line) and latitude are reported (P-value < 0.01:**; P-value < 0.05:*; P-value < 0.1:⁺; P-value > 0.1:^n.s). See Table S1 (Supporting information) for ancestry estimates for each pairwise combination of African and European population, and for SNPs on each chromosome, and with locations inside and/or outside inversions.

Fig. 4

Observed standardized Z score of (A) f₃ and (B) D statistics for each North American and Australian population when considering combinations of African and European populations and a putative donor population. For simplicity, only the Austrian population is shown as a putative European donor population. See Fig. S4 and Table S2 (Supporting information) for every pairwise combination of European and African populations. Dashed line represents the Bonferroni significance threshold at alpha <0.05. Values below the threshold in (A) are significantly different from 0. Values above or below the threshold in (B) are significantly different from 0. African populations are colour-coded based on location: West Africa (red); Central Africa (green); Southern Africa (blue).

Fig. 5

Number of significantly differentiated SNPs in North America and Australia at various FDR (q-value) thresholds.

Fig. 6

Patterns of codifferentiation and parallelism between North American, Australian and the Old world populations. (A) log₂ odds ratio that SNPs fall above the F_ST quantile cut-off (x-axis) in both sets of populations (NA: North America; AUS: Australia; OW: Old World). (B) Proportion of SNPs that vary in a parallel way given that they fall above the F_ST quantile cut-off in both sets of populations. Confidence bands represent 95% confidence intervals.