CRISPR/Cas9 modified An. gambiae carrying kdr mutation L1014F functionally validate its contribution in insecticide resistance and interaction with metabolic enzymes

Insecticide resistance in Anopheles mosquitoes is a major obstacle in maintaining the momentum in reducing the malaria burden; mitigating strategies require improved understanding of the underlying mechanisms. Mutations in the target site of insecticides (the voltage gated sodium channel for the most widely used pyrethroid class) and over-expression of detoxification enzymes are commonly reported, but their relative contribution to phenotypic resistance remain poorly understood. Here we present a genome editing pipeline to introduce single nucleotide polymorphisms in An. gambiae which we have used to study the effect of the classical kdr mutation L1014F (L995F based on An. gambiae numbering), one of the most widely distributed resistance alleles. Introduction of 1014F in an otherwise fully susceptible genetic background increased levels of resistance to all tested pyrethroids and DDT ranging from 9.9-fold for permethrin to >24-fold for DDT. The introduction of the 1014F allele was sufficient to reduce mortality of mosquitoes after exposure to deltamethrin treated bednets, even as the only resistance mechanism present. When 1014F was combined with over-expression of glutathione transferase Gste2, resistance to permethrin increased further demonstrating the critical combined effect between target site resistance and detoxification enzymes in vivo. We also show that mosquitoes carrying the 1014F allele in homozygosity showed fitness disadvantages including increased mortality at the larval stage and a reduction in fecundity and adult longevity, which can have consequences for the strength of selection that will apply to this allele in the field. Author Summary Escalation of pyrethroid resistance in Anopheles mosquitoes threatens to reduce the effectiveness of our most important tools in malaria control. Studying the mechanisms underlying insecticide resistance is critical to design mitigation strategies. Here, using genome modified mosquitoes, we functionally characterize the most prevalent mutation in resistant mosquitoes, showing that it confers substantial levels of resistance to all tested pyrethroids and undermines the performance of pyrethroid-treated nets. Furthermore, we show that combining this mutation with elevated levels of a detoxification enzyme further increases resistance. The pipeline we have developed provides a robust approach to quantifying the contribution of different combinations of resistance mechanisms to the overall phenotype, providing the missing link between resistance monitoring and predictions of resistance impact.


13
Insecticide resistance in Anopheles mosquitoes is a major obstacle in maintaining the momentum in 14 reducing the malaria burden; mitigating strategies require improved understanding of the underlying 15 mechanisms. Mutations in the target site of insecticides (the voltage gated sodium channel for the most 16 widely used pyrethroid class) and over-expression of detoxification enzymes are commonly reported, but 17 their relative contribution to phenotypic resistance remain poorly understood. Here we present a genome 18 editing pipeline to introduce single nucleotide polymorphisms in An. gambiae which we have used to 19 study the effect of the classical kdr mutation L1014F (L995F based on An. gambiae numbering), one of the 20 most widely distributed resistance alleles. Introduction of 1014F in an otherwise fully susceptible genetic 21 background increased levels of resistance to all tested pyrethroids and DDT ranging from 9.9-fold for 22 permethrin to >24-fold for DDT. The introduction of the 1014F allele was sufficient to reduce mortality of 23 mosquitoes after exposure to deltamethrin treated bednets, even as the only resistance mechanism 24 present. When 1014F was combined with over-expression of glutathione transferase Gste2, resistance to 25 permethrin increased further demonstrating the critical combined effect between target site resistance 26 and detoxification enzymes in vivo. We also show that mosquitoes carrying the 1014F allele in 27 homozygosity showed fitness disadvantages including increased mortality at the larval stage and a 28 reduction in fecundity and adult longevity, which can have consequences for the strength of selection that 29 will apply to this allele in the field.

30
Author Summary

31
Escalation of pyrethroid resistance in Anopheles mosquitoes threatens to reduce the effectiveness of our 32 most important tools in malaria control. Studying the mechanisms underlying insecticide resistance is 33 critical to design mitigation strategies. Here, using genome modified mosquitoes, we functionally 34 characterize the most prevalent mutation in resistant mosquitoes, showing that it confers substantial 35 levels of resistance to all tested pyrethroids and undermines the performance of pyrethroid-treated nets. 36 Furthermore, we show that combining this mutation with elevated levels of a detoxification enzyme 37 further increases resistance. The pipeline we have developed provides a robust approach to quantifying Introduction

42
The widespread use of insecticides in indoor residual spraying and insecticide-treated bednets (ITNs) has 43 been a critical driver in the reduction of malaria cases in the last decades (1). These tools have been so 44 effective because they reduce the density and the lifespan of mosquitoes and thus their ability to transmit 45 the Plasmodium parasite. However, gains in malaria control are now stalling (2) and this has been 46 attributed, at least partially, to increasing levels of insecticide resistance in Anopheles vectors (3). As 47 malaria vector control relies on a limited range of chemicals and new insecticides need years to be 48 developed and approved, it is critical to preserve the effectiveness of available compounds. To do that we 49 need to understand the mechanisms by which insects have evolved resistance and design mitigation 50 strategies.

51
The molecular basis of insecticide resistance is complex with multiple mechanisms co-existing. The two 52 most widely reported adaptations include mutations at the target site that reduce the insecticide's binding 53 affinity and increased production of detoxification enzymes, like P450s, esterases and GSTs (glutathione 54 transferases) that inactivate the insecticide molecules and enhance their excretion (4). More recently 55 additional mechanisms have been described, including cuticular modifications that reduce the 56 insecticide's penetration rate (5) and insecticide sequestration from chemosensory proteins (6). Although 57 in many cases the association of these mechanisms with insecticide resistance is clear, we are still lacking 58 critical knowledge about the effect size of each mechanism and importantly the combined effect of 59 different mechanisms. It has been hypothesized that the synergistic effect of different mechanisms is what 60 leads to operational control failure, (7) but there is little data to support this prediction. This knowledge 61 gap reduces our ability to interpret the results of the various molecular diagnostics that have been 62 developed to screen for the presence of insecticide resistance mechanisms in field populations, simply 63 because the predictive value of the markers used is unclear (8).

64
Insecticide resistance mechanisms often carry fitness costs. Therefore, removal of the insecticide, as a 65 selection pressure, is predicted to reduce the frequency of resistance alleles in the population. This is the 66 basis of insecticide resistance management strategies that alternate the use of chemistries with different 67 mode of action. However, not all resistance alleles pose high fitness costs, in which case they can persist 68 in populations, and in some cases additional mutations with a compensating role can be selected. Thus, it 69 is important to evaluate the cost of each documented resistance mechanism and make evidence-based 70 decisions on insecticide alternations. A key bottleneck is the ability to compare the effect of specific 71 mutations on defined genetic backgrounds. To date, several studies have documented fitness costs for 72 target-site resistance mutations, but these studies are rarely performed on genetically related strains, 73 which complicates the interpretation of results and often the establishment of a direct link (9).

74
In this study we examine the contribution to insecticide resistance and the associated fitness costs for 75 mutation L1014F on the An. gambiae voltage gated sodium channel (VGSC). Mutation L1014F (or L995F 76 using the Anopheles gambiae codon numbering), also known as classical kdr (knock-down resistance), was 77 among the first mechanisms associated with resistance to the organochlorine DDT and to pyrethroids 78 (10). Pyrethroids are a particularly important insecticide class as they are used in all ITNs, even the newer nets that combine two chemistries, due to their low mammalian toxicity and fast mode of action. 80 Therefore, it is commonplace to genotype for this mutation as an adjunct to resistance prevalence 81 bioassays. Several studies have tried to measure the effect size of L1014F in pyrethroid resistance, either 82 by associating its presence with survival to insecticide exposure (11-13), by introgressing the mutation in 83 an insecticide susceptible strain (14,15) or by genetically modifying Drosophila melanogaster (16). The 84 outcomes of these studies vary, with some reporting a low effect size, while others a moderate to high 85 (17). However, in all cases the results could be influenced by confounding factors arising either from 86 differences in the genetic background of the mosquito strains compared, or due to the use of a model 87 organism that might not reflect the exact response in mosquitoes.

88
Here, we have used CRISPR/Cas9 to introduce the L1014F mutation in an An. gambiae insecticide 89 susceptible strain providing the opportunity to investigate the direct effect of this mutation on several 90 traits with the minimum possible confounding effects. This is the first time to our knowledge a mosquito 91 strain has been genome edited to functionally validate a single mutation and we provide an experimental 92 pipeline for studies wishing to do the same. We also report the generation of a transgenic An. gambiae 93 line in which we have combined the L1014F mutation with over-expression of the detoxification enzyme 94 Gste2, which permits functional validation of the combined effect of target site and detoxification 95 enzymes in vivo.

97
Generation of a genome modified An. gambiae line bearing the L1014F VGSC mutation in an insecticide 98 susceptible background.

99
Embryos of the insecticide susceptible Kisumu strain were injected with a CRISPR/Donor plasmid mix 100 (Materials and Methods) and 32 larvae (G 0 ) were obtained with RFP (the Red Fluorescent Protein-marker 101 present on the CRISPR plasmid) expression predominantly in their anal papillae, that indicated these 102 fluorescent individuals had actively transcribed plasmids delivered. A subset of each female's G 1 progeny 103 was screened for the presence of genetically modified alleles using a LNA diagnostic assay (18). A positive 104 transformation event was found in progeny of one G 0 female at a frequency of 18% (Table S1) and verified 105 through sequencing. Sequencing of the vgsc region in the transformed line also revealed, based on the 106 presence in the donor construct of silent SNPs (Supplementary Figure 1), that DNA resection during 107 CRISPR reached from one side at least 199bp and from the other side less than 158bp. Thus, even if a 108 gRNA target is not available at or close (within 20 bp) to the mutation site, which is the preferred option, 109 it is still possible to recover transformants with gRNAs targeting a region further away. A summary of the 110 strategy used to generate the Kisumu-F/F line (which is para 1014F/F homozygous) is illustrated in Figure 1.

128
The extent of the resistance phenotype conferred by mutation L1014F was initially tested using WHO 129 discriminating dose assays. The concentration of insecticides in these assays is fixed at twice the lethal 130 concentration that kills 99% of the susceptible mosquitoes after 1h of exposure and mortality of less than 131 90% is the threshold to define resistance (19). Based on these criteria L1014F confers WHO defined 132 resistance to the pyrethroids, permethrin and α -cypermethrin only in homozygosity (recessive character) 133 ( Figure 2A, B and C), while it confers resistance to DDT in heterozygosity (dominant character) ( Figure 2D). 134 A reduction in mortality for Kisumu-F/F individuals against deltamethrin, was observed although not 135 statistically significant ( Figure 2D). The percentage of mosquitoes being knocked down immediately after 136 the 1h exposure period was also recorded for each insecticide (SI Appendix, Figure S2). In all cases except 137 deltamethrin, there is significant inhibition of knock down after 1hr exposure in F/F mosquitoes, but in 138 F/L genotypes a knockdown resistance phenotype was only observed against DDT. There was no 139 significant difference between 1hr knock down and 24 hr mortality for any insecticide or genotype (SI 140 Appendix, Figure S2). As expected, no resistance was observed against the carbamate insecticide 141 bendiocarb (Supporting dataset 1) that targets acetylcholinesterase and not the VGSC. To further quantify the level of resistance conferred by L1014F to the insecticides we performed time 151 response assays and estimated the time required to obtain 50% mortality (LT 50 ) for each of the three 152 genotypes (Table 1). Kisumu-F/F exhibited the highest predicted Resistance Ratio (RR 50 ) against DDT (> 153 24.5-fold) although accurate determination of the LT 50 for DDT was not possible in mutant homozygotes, 154 as mortality of Kisumu-F/F was less than 23% after 9 hours of exposure to DDT, which was the latest time 155 point measured. The RR 50 for α-cypermethrin was 19.7-fold, for deltamethrin 14.6-fold and for permethrin 156 9.9-fold. Lower RR 50 were observed for the heterozygote (Kisumu-L/F) mosquitoes in all cases (Table 1),  157 thus L1014F under the exposure conditions used has a semidominant effect on resistance.

178
Pre-exposure to PBO increases mortality of the Kisumu-F/F strain to deltamethrin. pre-exposure to PBO resulted in a significant increase in mortality (mean mortality of 84.7±0.86%) ( Figure  191 4).

193
We investigated the effect of mutation L1014F on several life history traits. In terms of growth, a 194 significantly lower percentage of Kisumu-F/F larvae (mean of 75%) reached the pupae stage compared to 195 wild type Kisumu-L/L (mean of 91%; P<0.0001) ( Figure 5A). In addition, the fecundity of Kisumu-F/F 196 females was significantly reduced with 35.7% of females not producing eggs, after one blood meal, 197 compared to 10.7% for the Kisumu-L/L females (P=0.003) ( Figure 5B). In contrast the fertility (number of 198 eggs laid) was not significantly different between the two genotypes (mean number of eggs laid by 199 Kisumu-F/F was 72 (±24 SD) compared to 83 (± 37 SD) for Kisumu-L/L) ( Figure 5C

237
In vivo functional validation is critical in establishing the importance of candidate mechanisms, alone and 238 in combination, in phenotypic insecticide resistance. Here we have used CRISPR/Cas9 to introduce the 239 VGSC mutation L1014F, which is widespread in insecticide resistant An. gambiae populations in Africa, in 240 an insecticide susceptible genetic background. Ordinarily genome modification strategies use a dominant 241 fluorescent marker to identify rare transformants. However, studying the role of single nucleotide 242 polymorphisms precludes the introduction of additional sequences, as these could confound the 243 interpretation of results. In this study we show that the CRISPR/Cas9 strategy we followed achieves 244 transformation rates high enough to recover transformants without the aid of a dominant marker gene 245 and thus can be used in An. gambiae to study the role of single nucleotide polymorphisms. Its applicability 246 to other mosquito species will depend on the rates of homology directed repair, which can be lower than 247 we achieved in An. gambiae (22). The use of CRISPR/Cas9 to study the effect of target site mutations in 248 insecticide resistance has been predominantly performed using the model organism Drosophila 249 melanogaster (16,(23)(24)(25) and in one agricultural pest (26), thus our study reports, to the best of our 250 knowledge, the first time this approach being used in a major insect vector of human disease pathogens. 251 Several target-site resistance mutations have been described in these insects whose effect size needs to 252 be clarified, but which could be assessed following the approach described here; these include mutations 253 in acetylcholinesterase, the target of organophosphate and carbamate insecticides, the GABA-gated 254 chloride channel that is associated with dieldrin resistance (27) and the chitin synthase that is associated 255 with resistance to diflubenzuron and other benzoylurea insecticides inhibiting this enzyme (25,28).

256
Introduction of L1014F in the Kisumu susceptible background increased insecticide resistance against all 257 pyrethroid insecticides tested and DDT, with resistance levels ranging from 9.9-fold for permethrin to 258 >24.5-fold for DDT in quantitative assays. The very high levels of resistance to DDT, observed even for 259 heterozygous individuals, suggests that L1014F was originally selected by the widespread use of DDT and 260 retained in the populations after introduction of pyrethroids, by conferring lower, but still substantial 261 levels of resistance. The resistance observed for Kisumu-F/F against permethrin and deltamethrin are very 262 similar to the levels of insensitivity mutation L1014F provides to Ae. aegypti VGSCs expressed in the in 263 vitro Xenopus oocyte system (8-fold for permethrin and 14-fold for deltamethrin) (29). In that system 264 insensitivity is calculated based on the percentage of channels whose activity is being modified by 265 pyrethroids. In the case of deltamethrin although the quantitative bioassay showed a 14-fold increase in 266 resistance, similar to the 13-fold reported for transgenic Drosophila carrying the equivalent mutation (16) quantitative bioassay data also demonstrate that characterizing resistance as a recessive or dominant trait 278 depends on the type of assay and conditions used. For example, the 1014F allele is dominant for DDT 279 resistance using the standard WHO bioassay, but recessive based on the quantitative analysis. 280 We also showed that the 1014F allele when in homozygosity is sufficient to induce reduced mortality after 281 mosquitoes are exposed to commercial, deltamethrin treated ITNs. We recorded a 67.5% mortality in the 282 standard WHO cone bioassay for Kisumu-F/F. This is higher than the 26% mortality that was previously 283 reported (14) for the kdr-Kisumu line, which was generated by introgressing the L1014F mutation from a 284 resistant field strain into Kisumu (15). Although introgression substantially dilutes the resistance genetic 285 background, it is impossible, even with multiple rounds of crossing, to achieve the resolution that allows 286 one to look at the effects of a mutation in isolation. This is particularly relevant for centromeric loci located 287 in low recombination regions, such as the vgsc (30). In addition, whole genome sequencing data from the 288 Anopheles gambiae 1000 Genomes Project, have shown a high genetic variation in the vgsc gene itself. 289 Twenty non-synonymous substitutions were identified, thirteen of which were found to occur almost 290 exclusively on haplotypes carrying the L1014F resistance allele and may enhance or compensate for the 291 L1014F resistance phenotype (31). Thus, the higher levels of resistance obtained for the introgressed line 292 could be related to the presence of additional unscreened mutations in the vgsc allele originating from 293 the parental resistant population or the presence of other genes that are carried over during 294 introgression.

295
The P450 inhibitor PBO is commonly used to estimate the contribution of P450 metabolism to resistance 296 phenotypes. Here we showed that pre-exposure to PBO increased mortality of the Kis-L1014F strain after 297 deltamethrin exposure. This enhancement in deltamethrin efficacy, even in the absence of metabolic 298 resistance, could be either mediated through the inhibition of the endogenous levels of P450s, that might 299 still contribute to resistance when in combination with the L1014F mutation, or by enhancing the 300 penetration, and thus bioavailability, of the insecticide through the cuticle (20). Either way, this result 301 suggests that demonstration of elevated P450 levels is not a necessary pre-requisite for use of PBO 302 containing products to mitigate against resistance, as has previously been recommended.

303
Combining mutation L1014F with over-expression of the detoxification enzyme Gste2 resulted in 304 increased levels of resistance to permethrin. Transgenic mosquitoes over-expressing Gste2 (allele carrying 305 the I114T variant associated with DDT resistance (32)) , as the only resistance mechanism have previously 306 been shown to be resistant to DDT and the organophosphate insecticide fenitrothion, but not to 307 permethrin or deltamethrin (21). However metabolic assays with recombinant An. funestus Gste2 have 308 shown its ability to directly metabolise DDT and permethrin in vitro (33). Thus, the An. gambiae Gste2 is 309 either a poor permethrin metaboliser that needs the presence of target-site resistance to show a 310 phenotypic effect or it mainly metabolises secondary metabolites of permethrin that can also have toxic 311 effects. We had intended to test the combined effect of Gste2 and L1014F in DDT resistance, but the 312 insensitivity of the Kisumu-F/F line even after long exposure times and high doses of DDT in topical 313 applications, precluded detection of any further increase in resistance with the addition of Gste2. 314 However, the permethrin data support the hypothesis (16) that the different mechanisms can interact in 315 An. gambiae, and that their combined effect ultimately shapes the resistance phenotype in mosquitoes.
Although the mechanisms of this interaction have not been elucidated yet, there are several hypotheses 317 as described in (16) . The kdr mutation, by reducing the binding affinity of the insecticide, could provide 318 more time for detoxification enzymes to act and avoid saturation. Alternatively, the detoxification 319 enzymes could reduce the number of parental insecticide molecules reaching the nervous system or 320 generate less-toxic metabolites with an even further reduced binding affinity to the mutated VGSC.

321
In addition to the impact on insecticide resistance we showed that L1014F has pleiotropic effects on 322 fitness related traits. It has long been hypothesized that L1014F carries fitness costs (34), but providing a 323 direct association had been complicated by the comparison of populations with different genetic 324 backgrounds (35). Although we cannot rule out the possibility of CRIPSR induced off-targets or haplotype-325 specific effects in the Kisumu-F/F line, the comparison we perform involves the least possible confounding 326 genetic effects. We show that significantly fewer individuals from the Kisumu-F/F strain are able to 327 develop from early instar larvae to pupae. This could also explain, at least partially, the lower than 328 expected number of homozygous individuals observed in the process of establishing the Kisumu-F/F line. 329 In addition, a significantly lower number of females from the Kisumu-F/F line oviposited. Although we did 330 not measure insemination rates, this may be related to the reported association of VGSC mutations with 331 reduced male mating success (36), and warrants further investigation in future. A reduced lifespan was 332 also observed for Kisumu-F/F females, which could be even more pronounced under field conditions. 333 However, it should be noted that fitness costs related to the L1014F mutation could be ameliorated by 334 compensatory mechanisms in wild populations. For example, several additional non-synonymous 335 mutations are found on the vgsc gene in linkage with L1014F, which is consistent with them either 336 conferring additional resistance or compensating for the deleterious effects of L1014F (31, 37, 38), and 337 can be examined by further mutagenesis of the Kisumu-F/F line generated herein.

339
Understanding the mechanisms of insecticide resistance and their phenotypic contribution can provide 340 predictive value for molecular diagnostic markers to be used as resistance prevalence monitoring tools 341 and assist in control program decision making. This work reports the functional validation of mutation 342 L1014F, which is widely used as a marker of pyrethroid resistance in malaria vectors, but without a clear 343 understanding of how precisely it affects the performance of insecticide-based vector control tools. We 344 show that L1014F even in isolation, provides substantial levels of insecticide resistance that impact the 345 performance of ITNs, but critically resistance levels are further increased when combined with the over-346 expression of a metabolic enzyme, the Gste2. This highlights the importance of interpreting molecular 347 diagnostics carefully, considering the synergistic or additive effects of the different mechanisms. Further 348 work is needed to investigate the complex interactions of the different insecticide resistance mechanisms 349 and the different mutations on the vgsc gene; the ability, demonstrated here, to introduce precise and 350 defined mutations in any combination, on any given haplotype offers a powerful tool to finally dissect 351 these interactions. This in turn will greatly increase the predictive value of multi-locus diagnostic panels 352 to detect and track the emergence of insecticide resistance, with the potential to be transformative for 353 insecticide resistance management programs. (https://chopchop.rc.fas.harvard.edu) websites. As described in Figure S1, the guide RNA that proved 374 successful in generating mutants was gRNA 3.

375
Generation of Donor for HDR: A 1,600 bb region of the Voltage Gated Sodium Channel (vgsc, AGAP004707-376 RA) gene having the 1014 codon in the middle (homology arms extending 800 bp either direction) was 377 synthesized de novo by Genescript in a puc19 vector, including the A→T transversion that generates the 378 L1014F mutation (codon alteration TTA→TTT) and other synonymous SNPs in the gRNA target sites to 379 avoid cleavage of the vector by Cas9 (SI Appendix, Figure S1).

380
Mosquito embryo injections and identification of G 1 transformants.

381
Freshly laid eggs of the Kisumu strain were microinjected with a mix of the Cas9 and donor plasmids 382 (300ng/ul each). In order to enrich for those individuals injected with appreciable quantities of the Cas9 383 and donor plasmid mix, surviving G 0 larvae expressing RFP (present on the Cas9 plasmid as a 3xP3:RFP) 384 transiently in the anal papillae were reared separately and backcrossed with the Kisumu strain in sex-385 specific cages. Females of both crosses were put in individually egg-laying tubes and a subset of their 386 progeny (G 1 ) was used to screen for positive transformants. DNA was extracted from G 1 larvae in pools of 387 two by grinding them in 30ul of STE buffer ( exposure and mortality after a 24hour recovery time. At least 3 replicates of 20 females (2-5 days old) 419 each were performed for each time point. A control tube with no insecticide was included in each test for 420 each strain. LT 50 , Resistance Ratio and associated statistical parameters were calculated from log-logistic 421 2 parameter dose response models using the drc package from (41) and the drm(), compParm(), ED() and 422 EDcomp() functions in R statistical software (version 3.4.3).

423
Cone bioassays 424 Cone bioassays were performed as described in (42) using PermaNet 2.0 nets (obtained directly by 425 Vestergaard). Νine replicates of five female mosquitoes, 2-5 days old were tested using at least three 426 randomly selected pieces of the nets (control and treated). Mosquitoes were exposed for 3min and then 427 transferred to recovery cups. Mortality was recorded 24h later. Kisumu-F/F homozygous strains. Mosquitoes were fed five days later with our standard 'blood' supply 441 (50:50 mix of human research red cells and plasma, NHS blood and transplant service) using a Hemotek 442 membrane feeding system. Those that did not feed were removed from the experiment. Two days post 443 blood feeding females were placed in individual laying cups containing 20 ml of water. Females were 444 aspirated out of the cups two days later and oviposited eggs were counted under a stereoscope. The 445 number of females that laid (at least one egg) and did not lay were recorded and analyzed using a Fisher's 446 exact test (GraphPad Prism 9.0.0). Larvae that hatched were counted four days later to include any with 447 a delayed hatching. Statistical comparison of egg and larvae counts was done using a Mann-Whitney U 448 test (GraphPad Prism 9.0.0).

449
Adult longevity 450 Thirty females from each strain, were transferred in pools of five in six replicate cups and fed ad libitum 451 with 10% (w/v) sugar solution. Mortality was recorded every day and dead mosquitoes removed. 452 Longevity was assessed using GraphPad Prism 9.0.0 using a Mantel-Cox test.

454
We would like to thank Beth Poulton (LSTM) for assistance with the R code used to analyze the time 455 response assays and her help in rearing the Gste2 transgenic colony. We would also like to thank Fraser 456 Colman (