Restoration of pyrethroid susceptibility in a highly resistant Aedes aegypti population

Insecticide resistance has evolved in disease vectors worldwide, creating the urgent need to either develop new control methods or restore insecticide susceptibility to regain use of existing tools. Here we show that phenotypic susceptibility can be restored in a highly resistant field-derived strain of Aedes aegypti in only 10 generations through rearing them in the absence of insecticide.


Introduction
Insecticide resistance is an increasing challenge for disease control [1]. Of particular concern is resistance to pyrethroid insecticides, given their widespread use and low mammalian toxicity [1]. Populations of Aedes aegypti, the main dengue, chikungunya and Zika vector, are exhibiting increasingly high levels of pyrethroid resistance, commonly measured by the knock-down resistance (kdr) mutations [2][3][4]. These point mutations in the para-orthologous sodium channel gene disrupt insecticide binding to the voltage-gated sodium channels [5]. Two kdr mutations strongly associated with pyrethroid resistance in Ae. aegypti are the phenylalanine to cysteine mutation in 1534 (F1534C) and the valine to isoleucine mutation in 1016 (V1016I) [6,7].
Despite strong insecticide selection pressure, polymorphism is maintained at the kdr locus in Ae. aegypti field populations at fine spatial scales [8][9][10][11]. Such polymorphism may be indicative of a fitness cost to the mutations in the absence of insecticide. While fitness costs in resistant strains of Ae. aegypti have been demonstrated [12 -14], there is limited empirical evidence demonstrating restoration of pyrethroid susceptibility without insecticide pressure. Instead, studies tend to focus on selection towards resistance, describing increases in both kdr allele frequencies with pyrethroid exposure [13,14] along with reductions in larval viability [12]. Here, we use semi-natural experiments to assess whether susceptibility can be restored in a resistant strain of Ae. aegypti.

Methods (a) Experimental design
Lasting Insecticidal Nets (MGK) containing Olyset TM Technology with 2% permethrin only. We replicated each treatment five times; all replicates were conducted simultaneously inside an uninhabited room in a typical house in Merida, Mexico to simulate semi-natural conditions.
For each replicate, we placed a 2 l white bucket containing 1 l of domestic water and 500 -800 eggs into an experimental tent. We fed larvae bovine liver powder (MP Biomedicals) ad libitum to minimize effects of larval competition. A 10% sucrose solution was provided to adults daily. Emerged females were blood-fed once a week for two weeks by a human volunteer who placed his or her arm directly inside the cage. Females were allowed to lay eggs into black oviposition cups and each week eggs were removed from the tent to dry. After two weeks, all remaining eggs and adults were removed from the tent. Eggs were left to dry for one day, and then 500 -800 were selected at random and placed into a clean 2 l bucket to start the next generation. Surviving adult mosquitoes were euthanized in a 2208C freezer. Fifteen female and 15 male surviving adults from each replicate were selected at random for kdr genotyping. This process was repeated for 10 generations (1 year). After generation 10, eggs from tents with insecticide were placed into tents without insecticide, and vice versa, to reverse selection pressure.

(b) Strain characterization
Source Ae. aegypti were F1 from eggs collected in Merida, Mexico. The frequency of I1016 and C1534 alleles in the population was 0.595 and 0.937 respectively, and the population exhibited 13.7% knock-down to permethrin according to CDC bioassay protocols [15].

(c) Resistance assays
At generations F0 and F10, we conducted CDC bottle bioassays (15) on 100 females per experimental replicate (four bottle replicates of 25 mosquitoes) to test for phenotypic resistance to technical grade permethrin at 15 mg ml 21 . A susceptible reference strain (Cienega de Flores) and an acetone-coated bottle were used as controls. The proportion of mosquitoes knocked-down at 30 min was recorded, and a Welch two-sample t-test (used for unequal variance between groups) assessed the difference in knock-down proportions between treatments. Allele-specific realtime PCR determined the kdr genotypes for generations 1, 3, 7 and 10 following protocols outlined in Saavedra-Rodriguez et al. [7] for F1534C and Yanola et al. [16] for V1016I.

(d) Analysis
These mutations are close on the chromosome, approximately 44.5 kb apart [17], so we calculated linkage disequilibrium at each generation and for each replicate (see electronic supplementary material for equations) [18]. Using the maximum-likelihood estimation of linkage disequilibrium, D, we estimated haplotype frequencies for each generation and each replicate. To calculate fitness of each haplotype, we aggregated all replicates from each generation to increase sample size and calculated overall haplotype frequency. In the insecticide treatment, the fitness of each haplotype was calculated as the average haplotype frequency with insecticide divided by the average haplotype frequency without insecticide. We estimated the fitness of each haplotype without insecticide by dividing the average haplotype frequency at generation F10 by the average haplotype frequency at generation F1 without insecticide. Relative fitness was calculated by normalizing each haplotype to the haplotype with the highest fitness in the treatment, which assumes that the haplotype with the highest frequency carries the highest fitness.

Results (a) Phenotypic resistance
The mean proportion of knocked-down mosquitoes in the initial population (F0) was 0.14 + 0.13, remaining unchanged over the course of 10 generations with insecticide exposure, ending with a knock-down proportion of 0.12 + 0.17 ( figure 1). However, the treatment without insecticide was 5.8 times more susceptible, displaying a knock-down proportion of 0.70 + 0.14 (Welch t-test; t ¼ 25.72, d.f. ¼ 7.6, p-value 0.001). Additionally, 100% mortality was observed in F11 adults originating from no-insecticide treatments when placed into insecticide treatments; adults from F11 originating from insecticide treatments survived when placed into non-insecticide tents.

Discussion
We found a significant shift towards phenotypic susceptibility in populations without insecticide pressure. While populations did not meet the criterion for susceptibility defined by the World Health Organization, which is less than 90% knock-down [19], the significant loss of resistance in only 10 generations suggests that a longer duration   [6,7,16,20,21], yet kdr is not the only genetic mechanism underlying resistance. Saavedra-Rodriguez et al. [22] found that while I1016 was the largest contributor to variance in kdr phenotype (approx. 58.6%), loci involved with metabolic detoxification of insecticide were also responsible. Additionally, an overexpression of mixed-function oxidases has been associated with deltamethrin-selected Ae. aegypti [14] and an upregulation of cytochrome P 450 genes in permethrin-selected strains [23]. The lack of a strong association between kdr alleles and phenotypic resistance in our data suggests that metabolic resistance may be driving phenotype in our populations.
The relative fitness estimates of the haplotypes with and without insecticide give insight into the individual allele fitness, providing evidence of a fitness cost to C1534. Without insecticide, the wild-type haplotype, F1534/V1016, had the highest fitness. The C1534/V1016 haplotype, however, only had a fitness of one-half that of the wild-type haplotype, indicating a cost to C1534. Additionally, the absence of the F1534/I1016 haplotype indicates that either (1) I1016 is costly, or (2) the mutations are sequential in nature, with I1016 occurring only after the presence of C1534. Vera-Maloof et al. [24] also found a near absence of the F1534/ I1016 haplotype in a linkage disequilibrium analysis of kdr mutations in field-caught Mexican Ae. aegypti, and similarly concluded that the mutations are likely evolved sequentially.
Overall, our results give compelling evidence of a fitness cost to the C1534 mutation and show that susceptibility can be restored in a highly permethrin-resistant Ae. aegypti population in the absence of insecticide. Only two other studies have evaluated the loss of pyrethroid resistance in Ae. aegypti. Both found a change after 15 generations: one quantified a decrease in I1016 frequency from 0.75 to 0.20 [13], yet did not evaluate phenotype, and the other found that the phenotype of a formerly permethrin-resistant population approached that of the susceptible strain [25]. It is important to note that kdr mutations are recessive to wild-type alleles, conferring  resistance in the homozygous state [1]. Even when resistance is lost, kdr mutations may still be present in heterozygotes and could be selected for given a subsequent increase in insecticide pressure. Future studies should evaluate the loss of resistance in field populations, as they are subject to ecological forces such as density-dependent competition [26] and environmental variation [27] that may modify the strength of selection for resistance mutations. Even though this study was conducted under laboratory conditions, the results support a vector control strategy that rotates chemicals in time and/or space, providing areas where resistant populations can revert to susceptibility while still using effective chemicals to suppress overall population abundance [28].
Ethics. Experimental protocols were approved by Emory University's ethics committee under protocol ID IRB00082848 and by the Comisió n Estatal de Bioetica de Yucatán. Informed consent was provided by human volunteers who were involved in mosquito blood-feeding.