Genome-wide association studies reveal novel loci associated with pyrethroid and organophosphate resistance in Anopheles gambiae s.l.

Resistance to insecticides in Anopheles mosquitoes threatens the effectiveness of the most widespread tools currently used to control malaria. The genetic underpinnings of resistance are still only partially understood, with much of the variance in resistance phenotype left unexplained. We performed a multi-country large scale genome-wide association study of resistance to two insecticides widely used in malaria control: deltamethrin and pirimiphos-methyl. Using a bioassay methodology designed to maximise the phenotypic difference between resistant and susceptible samples, we sequenced 969 phenotyped female An. gambiae and An. coluzzii from ten locations across four countries in West Africa (Benin, Côte d’Ivoire, Ghana and Togo), identifying single nucleotide polymorphisms (SNPs) and copy number variants (CNVs) segregating in the populations. The patterns of resistance association were highly multiallelic and variable between populations, with different genomic regions contributing to resistance, as well as different mutations within a given region. While the strongest and most consistent association with deltamethrin resistance came from the region around Cyp6aa1, this resistance was based on a combination of several independent CNVs in An. coluzzii, and on a non-CNV bearing haplotype in An. gambiae. Further signals involved a range of cytochrome P450, mitochondrial, and immunity genes. Similarly, for pirimiphos-methyl, while the strongest signal came from the region of Ace1, more widespread signals included cytochrome P450s, glutathione S-transferases, and a subunit of the nAChR target site of neonicotinoid insecticides. The regions around Cyp9k1 and the Tep family of immune genes were associated with resistance to both insecticide classes, suggesting possible cross-resistance mechanisms. These locally-varying, multigenic and multiallelic patterns highlight the challenges involved in genomic monitoring and surveillance of resistance, and form the basis for improvement of methods used to detect and predict resistance. Based on simulations of resistance variants, we recommend that yet larger scale studies, exceeding 500 phenotyped samples per population, are required to better identify associated genomic regions.