Population genomics reveals that an anthropophilic population of Aedes aegypti mosquitoes in West Africa recently gave rise to American and Asian populations of this major disease vector

Abstract

Background: The mosquito Aedes aegypti is the main vector of dengue, Zika, chikungunya and yellow fever viruses. This major disease vector is thought to have arisen when the African subspecies Ae. aegypti formosus evolved from being zoophilic and living in forest habitats into a form that specialises on humans and resides near human population centres. The resulting domestic subspecies, Ae. aegypti aegypti, is found throughout the tropics and largely blood-feeds on humans.

Results: To understand this transition, we have sequenced the exomes of mosquitoes collected from five populations from around the world. We found that Ae. aegypti specimens from an urban population in Senegal in West Africa were more closely related to populations in Mexico and Sri Lanka than they were to a nearby forest population. We estimate that the populations in Senegal and Mexico split just a few hundred years ago, and we found no evidence of Ae. aegypti aegypti mosquitoes migrating back to Africa from elsewhere in the tropics. The out-of-Africa migration was accompanied by a dramatic reduction in effective population size, resulting in a loss of genetic diversity and rare genetic variants.

Conclusions: We conclude that a domestic population of Ae. aegypti in Senegal and domestic populations on other continents are more closely related to each other than to other African populations. This suggests that an ancestral population of Ae. aegypti evolved to become a human specialist in Africa, giving rise to the subspecies Ae. aegypti aegypti. The descendants of this population are still found in West Africa today, and the rest of the world was colonised when mosquitoes from this population migrated out of Africa. This is the first report of an African population of Ae. aegypti aegypti mosquitoes that is closely related to Asian and American populations. As the two subspecies differ in their ability to vector disease, their existence side by side in West Africa may have important implications for disease transmission.

Keywords: Aedes aegypti; Anthropophilic; Arboviral diseases; Dengue virus; Mosquito evolution; Vector-borne diseases; Zika virus.

Fig. 1

Nucleotide diversity (a) and site frequency spectrum (b) of five populations of Ae. aegypti. a Nucleotide diversity (π) was estimated for non-coding sites >500 bp from exons. b The site frequency spectrum was estimated for 10 individuals from each population using third codon positions and non-coding sites >100 bp from exons. Ae. bromeliae was used to polarize sites. The grey bars are the expected frequencies assuming variant sites are neutral and the effective population size is constant. In both panels, error bars are 95% confidence intervals from block-bootstrapping

Fig. 2

Genetic structure in Ae. aegypti populations. a Principal component analysis of Ae. aegypti exome sequences from five populations. The PCA was calculated from a covariance matrix calculated from all variants in the genome while accounting for genotype uncertainty. The percentage of the variance explained by each component is shown on the axis. b Ancestry proportions for Ae. aegypti individuals from five populations. Ancestry is conditional on the number of genetic clusters (K = 2–5) and is inferred from all sites in our dataset

Fig. 3

Differences in allele frequencies between populations. a Two-dimensional site frequency spectra. Colours represent the number of sites at a given frequency within each population (0-20) with frequency increasing from left to right and bottom to top in each spectrum. Allele frequencies were estimated using 10 randomly sampled individuals from each population. b Pairwise F _ST

Fig. 4

Historical relationships between Ae. aegypti populations. a Neighbour-joining tree of Ae. aegypti exome sequences from five populations. The tree is rooted with the sequence of Ae. bromeliae. Branches leading to samples from different populations are colour-coded. The scale is genetic distance (D _xy). b Relationships between populations. The branch lengths are proportional to the amount of genetic drift that has occurred. The scale bar shows ten times the average standard error of the entries in the sample covariance matrix. The numbers on branches are percent bootstrap support calculated by resampling blocks of 100 SNPs. The population tree was reconstructed using allele frequency data using the TreeMix program [37]. Both panels use all sites in our dataset

Fig. 5

Demographic modelling for African and non-African populations does not support admixture-back-to-Africa model. a Statistical support for four demographic models. Log likelihood indicates likelihood of data given each model, with higher values corresponding to better fit. Lower Akaike information criterion (AIC) values indicate better support for model (AIC = 2d – 2(Log Likelihood), where d is the number of model parameters estimated). b Admixture analysis of data simulated under best-fit demographic model generates evidence for mixed ancestry in Senegal Urban similar to Fig. 2, despite including no admixture in model. Five thousand 500-bp exons were simulated using fastsimcoal2 and analysed using admixture [67]. c Schematic representing the maximum likelihood estimated model. Parameters are effective population sizes, and times when populations split or changed in size. d Confidence intervals (CIs) for model parameters