Figure 2. Plasmodium falciparum infection is associated with the 2La inversion genotype in the Anopheles gambiae complex across Africa.
(A) Oocyst infection prevalence. Sampled mosquitoes were dissected to count midgut oocysts and were molecularly genotyped for the 2La inversion. Differences in oocyst infection prevalence between all three inversion genotypes are indicated by vertical bars, and statistical difference between inversion homozygotes, 2La/2La and 2L+a/2+a, at each geographic sampling site are indicated by p-values. Pairwise statistical comparisons between all inversion genotypes are presented in Figure 2—figure supplement 1. Oocyst infections were measured in wild-caught adult mosquitoes after uncontrolled exposure to bloodmeals and parasites in nature (Kenya and Guinea-Conakry), or in adults grown from wild larvae, then membrane-fed on blood from natural gametocyte carriers (Burkina Faso). In Kenya, mosquitoes were held for 7 days before dissection, whereas in Guinea-Conakry, mosquitoes were dissected upon capture. Association was measured in test sets including all infected mosquitoes and an equal number of random non-infected mosquitoes from the same collection, to normalize average oocyst infection prevalence to ≥50% for each site (see Materials and methods). This analysis provides the statistical power to detect both increase and decrease of in prevalence, and normalizes statistical power across all sites. The Burkina Faso collection was comprised of multiple taxa, which are presented and analyzed separately in Figure 2—figure supplement 2. (B) Oocyst infection intensity. The numbers of midgut oocysts 7–8 days after infection are shown for all individuals with ≥1 midgut oocyst. Statistical analysis by non-parametric Wilcoxon-Mann-Whitney tests indicates that within the Burkina Faso samples, the 2La/2La mosquitoes had significantly lower parasite loads as compared to 2La/2L+a or 2L+a/2L+a. A similar trend is evident though non-significant in the Kenya samples (all p values shown in Figure 2—figure supplement 1). Boxplots delineate the first and third quartiles, median is indicated by the dashed line within the box and error bars are 1.5 time the interquartile range. (C) 2La genotype frequency as a function of infection outcome. At each sampling site, 2La/2La homozygotes are more prevalent in the uninfected group, whereas 2L+a/2L+a homozygotes are more prevalent in the infected group. 2La inversion frequencies are significantly different between the uninfected and infected samples at each of the three sampling sites.