Assessing connectivity despite high diversity in island populations of a malaria mosquito

Abstract

Documenting isolation is notoriously difficult for species with vast polymorphic populations. High proportions of shared variation impede estimation of connectivity, even despite leveraging information from many genetic markers. We overcome these impediments by combining classical analysis of neutral variation with assays of the structure of selected variation, demonstrated using populations of the principal African malaria vector Anopheles gambiae. Accurate estimation of mosquito migration is crucial for efforts to combat malaria. Modeling and cage experiments suggest that mosquito gene drive systems will enable malaria eradication, but establishing safety and efficacy requires identification of isolated populations in which to conduct field testing. We assess Lake Victoria islands as candidate sites, finding one island 30 km offshore is as differentiated from mainland samples as populations from across the continent. Collectively, our results suggest sufficient contemporary isolation of these islands to warrant consideration as field-testing locations and illustrate shared adaptive variation as a useful proxy for connectivity in highly polymorphic species.

Keywords: Anopheles gambiae; gene drive technology; gene flow; malaria; migration.

Figure 1

Map of Lake Victoria Basin (LVB) study area. Map of study area showing sampling localities on Ssese Islands (blue) and mainland localities (red) in LVB. The Ag1000G reference population, Nagongera, Tororo District, is not shown, but lies 111 km NE of Kiyindi, 57 km from the shore of Lake Victoria. Map data copyright 2018 Google

Figure 2

Population structure in the Lake Victoria Basin (LVB). Analyses are based on chromosome 3 to avoid segregating inversions on other chromosomes, unless otherwise noted. (a) ADMIXTURE‐inferred ancestry of individuals in LVB. Results based on analysis of LVB and Ag1000G merged dataset. Analysis is restricted to Anopheles gambiae s. s. Clustering shown for k = 6 clusters, which minimizes cross‐validation error, and k = 9 clusters, the lowest k for which island individuals have the majority of their ancestry assigned to an island‐specific cluster. (b) Results of the clustering analysis with k = 9 clusters for LVB individuals, split by sampling locality. (c) Plot of first two components of PCA of Lake Victoria Basin individuals showing locality of origin. Mainland individuals are colored red, while island individuals are blue, and point shape indicates sampling locality. Based on these results and that of ADMIXTURE analysis, the island sample of Bugala was split into mainland‐ and island‐like subpopulations (“Bugala (M)" and “Bugala (I)," respectively) for subsequent analyses (Figure S4). (d) Heatmap of F _ST between sites (lower triangle) and associated z‐score computed via block jackknife (upper triangle). “Bugala (M)" and “Bugala (I)" are the mainland‐ and island‐like subpopulations of Bugala. (e) Proportion of genomewide pairwise IBD sharing between individuals, based on the full genome. Each small square represents a comparison between two individuals, and darker colors indicate a higher proportion of the two genomes is in IBD, shaded on a logarithmic scale. Individuals are grouped by locality

Figure 3

Diversity metrics in the Lake Victoria Basin samples. Shown are a (a) boxplot of nucleotide diversity (π; in 10 kb windows), (b) boxplot of Tajima's D (in 10 kb windows), (c) boxplot of inbreeding statistic (F), (d) boxplot of length of runs of homozygosity (F _ROH), (e) histogram of minor allele frequency (MAF), and (f) decay in linkage disequilibrium (r²), all grouped by sampling locality. For all boxplots, outlier points are not shown

Figure 4

Population history of the Lake Victoria Basin samples. (a) Long‐term evolutionary population histories inferred via stairway plots for island and mainland samples. (b) Contemporary or short‐term effective population size (N _e) history inferred using sharing of regions that are identical by descent (IBD). Wamala, a mainland locality showing island‐like fluctuations in population size, is indicated with a dashed line. Plot truncated to exclude implausibly high estimates that are likely an artifact of sample size