Selection on Aedes aegypti alters Wolbachia-mediated dengue virus blocking and fitness

Abstract

The dengue, Zika and chikungunya viruses are transmitted by the mosquito Aedes aegypti and pose a substantial threat to global public health. Current vaccines and mosquito control strategies have limited efficacy, so novel interventions are needed^1,2. Wolbachia are bacteria that inhabit insect cells and have been found to reduce viral infection-a phenotype that is referred to as viral 'blocking'³. Although not naturally found in A. aegypti⁴, Wolbachia were stably introduced into this mosquito in 2011^4,5 and were shown to reduce the transmission potential of dengue, Zika and chikungunya^6,7. Subsequent field trials showed Wolbachia's ability to spread through A. aegypti populations and reduce the local incidence of dengue fever⁸. Despite these successes, the evolutionary stability of viral blocking is unknown. Here, we utilized artificial selection to reveal genetic variation in the mosquito that affects Wolbachia-mediated dengue blocking. We found that mosquitoes exhibiting weaker blocking also have reduced fitness, suggesting the potential for natural selection to maintain blocking. We also identified A. aegypti genes that affect blocking strength, shedding light on a possible mechanism for the trait. These results will inform the use of Wolbachia as biocontrol agents against mosquito-borne viruses and direct further research into measuring and improving their efficacy.

Fig. 1 |. Illustration of the selection treatments.

We selected for low and high values of Wolbachia-mediated dengue blocking alongside a control treatment where mosquitoes were selected at random. Each treatment consisted of three independent replicate populations randomly generated from the same ancestral population of Wolbachia-infected A. aegypti (wMel.F strain) that represented the genetic diversity in Queensland, Australia. Circles represent the population of females whose eggs were selected based on their infection load to seed the subsequent generation of breeding. Arrows indicate the direction of selection. Blue represents selection for high blocking, red for low blocking; and yellow represents random selection.

Fig. 2 |. Evolution of Wolbachia-mediated blocking of the dengue virus in A. aegypti.

All mosquitoes were injected with the same dose of virus. a, log₁₀[copies of dengue virus per mosquito] in the high-, low- and random-blocking treatments after four rounds of selection (mixed-effects model: treatment: χ² = 9.8; d.f. = 5; P = 0.0073; post-hoc Tukey’s test: low versus high: P = 0.0017; low versus random: P = 0.0005; random versus high: not significant (NS; P = 0.94); Wolbachia density: χ² = 1.36; d.f. = 6; P = 0.24; treatment × Wolbachia density: χ² = 0.25; d.f. = 8; P = 0.88). b, log₁₀[copies of dengue virus per mosquito] in the evolved populations either treated with the antibiotic tetracycline to remove Wolbachia (Wolbachia⁻) or not (Wolbachia⁺). These mosquitoes were all tested four generations after the selection experiment (Wolbachia⁻ mixed-effects model: treatment: χ² = 1.01; d.f. = 5; P = 0.603; FDR-corrected for multiple comparisons P = 0.603; Wolbachia⁺ mixed-effects model: treatment: χ² = 13; d.f. = 5; P = 0.0015; FDRcorrected for multiple comparisons P = 0.003; post-hoc Tukey’s test: high versus random: P = 0.95; random versus low: P < 1 × 10⁻⁵; high versus low: P = 2.14 × 10⁻⁵). Sample sizes and further statistical information are listed in Supplementary Table 1. Replicate population IDs are noted under each error bar. Error bars represent 2 s.e. *P < 0.01; **P < 0.001; ***P < 0.0001.

Fig. 3 |. Relationship between the fitness of Wolbachia-infected mosquitoes and their ability to block dengue virus across the evolved mosquito populations.

The intrinsic rate of natural increase (r) for each A. aegypti population was measured in the absence of dengue infection. This was calculated from an age-structured Leslie matrix model that combined different fitness measures that were collected empirically. Life-history data were collected in triplicate for each replicate population, and populations were analysed in batches (1–3) for logistical purposes. Replicate populations were assigned to batches randomly (mixed-effects regression: log10[copies of dengue virus per mosquito]: χ² = 9.04; d.f. = 5; P = 0.0026). Experimental batch and treatment were included as random factors to control for non-independence. Data are presented by batch, which was the largest random source of variation. The colour of each point represents the selection treatment. Replicate population IDs are noted next to each point. Black lines fitted across treatments within each batch show the relationship between fitness and dengue virus load. Sample sizes and further statistical information are listed in Supplementary Table 1.

Fig. 4 |. Genetic variation in A. aegypti associated with Wolbachia-mediated dengue virus blocking.

a, Manhattan plot showing SNPs in the A. aegypti genome that were differentiated between the high- and low-blocking populations. The SNPs were analysed with a generalized linear model, and genomewide significance was estimated using empirical significance thresholds based on exhaustive permutation of our experimental data. The red line is the 5% FDR threshold based on all possible permutations (39.3). b, Expression of cadherin relative to the housekeeping gene RPS17 in the body (mixed-effects model: treatment: d.f. = 5; χ² = 19.07; P = 7.228 × 10⁻⁵; post-hoc Tukey’s test: low versus high: P < 0.001; low versus random: P < 0.001; random versus high: P = 0.042). c, Expression of α-mannosidase 2a relative to RPS17 in the body (mixed-effects model: treatment: d.f. = 5; χ² = 2.49; P = 0.29). d, Frequencies of thymine at position 238,777,216 on A. aegypti chromosome 1 plotted against the average dengue load for each population. This is the most significant SNP from the high versus low comparison within the cadherin gene. Replicate population IDs are noted under each error bar. Sample sizes and further statistical information are listed in Supplementary Table 1. Error bars represent 2 s.e. *P < 0.01; **P < 0.001; ***P < 0.0001.