Generation of knock-down resistance alleles of the voltage-gated sodium ion channel in Drosophila
Pyrethroids and DDT bind to the voltage-gated sodium channels (VGSCs) triggering abnormal channel activity but are unable to bind or disrupt function of Knockdown-resistant (Kdr) mutant channels. The most widespread kdr alleles in Anophelines include L1014F, whereas I1011M/V are prevalent in Aedes aegypti5. While attempts have been made to evaluate the effect of these mutations in field isolates, their direct impact on insect fitness and physiology has been challenging to assess due to genetic heterogeneity of wild insect populations6,7,8,9. As the protein sequences of VGSCs are highly conserved across insect orders10, with nearly 100% identity at the insecticide-binding site (Supplementary Fig. 1), we analyzed the effects of prevalent kdr field variants on insecticide susceptibility and fitness in the genetic model Drosophila melanogaster. As a first step toward this goal, we generated fly lines carrying mutations equivalent to L1014F, I1011M and I1011V alleles in the Drosophila vgsc gene ortholog paralytic (para) using CRISPR-Cas9-mediated gene editing (Fig. 1A).
Since the L1014F, I1011M and I1011V kdr mutations have not been reported in IR Drosophila field isolates, we first tested whether they conferred resistance to DDT and pyrethroids (permethrin and deltamethrin). The three mutants exhibited differential insecticide resistance profiles: paraL1014F flies were highly resistant to DDT (Fig. 1B), moderately resistant to permethrin (Supplementary Fig. 2A), but were susceptible to deltamethrin (Supplementary Fig. 2B). In contrast, paraI1011M flies displayed moderate resistance to DDT and permethrin (Fig. 1B, Supplementary Fig. 2A), while paraI1011V mutants remained susceptible to all three insecticides even at low concentrations (Fig. 1B, Supplementary Fig. 2A, Supplementary Fig. 2B). Recent molecular modeling studies of the Na+ ion channel domain of the VGSC and the conformational consequences of prevalent kdr mutations (L1014F, L1014H and L1014S) indicate that substitution of different amino acids even at the same position can have different impacts on DDT binding11. Similar alternative conformational effects may explain the observed differential resistance profiles for the paraL1014F, paraI1011M, and paraI1011V alleles we observed.
Different kdr alleles are associated with distinct fitness costs
We reasoned that the absence of paraL1014F and paraI1011M mutations in wild Drosophila populations could be due to associated fitness costs. We tested this hypothesis by measuring the effect of these mutations on various fitness parameters. In the absence of insecticide exposure, both paraL1014F and paraI1011M significantly reduced lifespan in males (Fig. 1C), but had no detectable impact on embryonic viability (Fig. 1D). Because strain-specific sensitivities to DDT can alter oxidative stress responses in flies12, we tested whether paraL1014F or paraI1011M mutations impacted resistance to oxidative stress induced by the herbicide paraquat13,14. In these experiments, paraL1014F mutants exhibited significantly elevated resistance to paraquat compared to WT flies, while paraI1011M mutants remained susceptible (Fig. 1E), revealing again distinct functional outcomes of these vgsc mutations. Previous observations also linked mutations in para with temperature sensitivity, wherein mutant flies display paralysis at higher temperatures due to altered neuronal activity15. Furthermore, a subset of these temperature-sensitive mutants also exhibited resistance to insecticides16. Given the ubiquity of the IR L1014F allele in other insects, we hypothesized that temperature sensitivity may contribute to the absence of this particular kdr mutation in wild Drosophila populations, despite the robust resistance it confers to DDT (Fig. 1B) and permethrin (Supplementary Fig. 2A) exposure. Indeed, paraL1014F mutants displayed paralysis at higher temperature, while paraI1011M male mutants were surprisingly more resistant than wild-type (Fig. 1F).
In humans, mutations in VGSC are linked to altered sleep and seizure disorders17. Fly models for these disease-causing mutations also manifest significant sleep defects, altered neuronal activity, and neurodegeneration18,19,20. As for the phenotypes discussed above, differential effects on sleep profiles were observed for the various IR alleles we generated. While paraL1014F mutants followed normal daytime sleep patterns (cumulative and temporal profiles), their night-time sleep was significantly elevated. In contrast, paraI1011M mutants displayed delayed day-time sleep and elevated night-time sleep (Fig. 1G, H, Supplementary Fig. 3). These distinct physiological phenotypes of the two IR resistant mutations may arise from their differential impact on channel activity21 and consequent neuronal functions. Further studies will be required to ascertain the mechanistic basis for these alternative behavioral profiles.
Development of an allelic drive system to replace the kdr 1014 F allele
We recently developed an efficient CRISPR-based allelic-drive strategy in Drosophila22 and here, we adapt this system to revert an IR mutation back to the insecticide-susceptible wild-type allele. Because the widespread paraL1014F allele conferred the greatest resistance to DDT and permethrin, we focused on this mutation for allelic correction. In these experiments, we designed a guide RNA (gRNA-F, Fig. 2B) carried on a gene-drive cassette (y<CC|pF|>) that selectively targets the paraL1014F allele for cleavage, leading to its repair from the uncleavable wild-type paraL1014 (=paraWT) allele. We placed gRNA–F under control of the ubiquitously active U6 promoter in the y<CC|pF|> transgenic drive element, which also expresses an additional gRNA (gRNA-y) that targets the insertion site of the y<CC|pF|> element for cleavage within the yellow (y) locus to promote self-copying and super-Mendelian inheritance of that element (Fig. 2A). We followed inheritance of the y<CC|pF|> element by the DsRed fluorescent protein marker, expressed in eyes under the control of the 3XP3 promoter.
We first assessed performance of the y<CC|pF|> allelic-drive element ±Cas9 in two-generation test crosses (Fig. 2C). We chose a mini-white insertion tightly linked (~0.5 cM) with the uncleavable paraWT donor allele to distinguish it from the cut-sensitive paraL1014F receiver allele. F2 progeny inheriting the white− (w−) “receiver” chromosome (hereafter referred as F2 w− receiver progeny) were either unmodified (paraL1014F) or converted (paraWT) (Fig. 2A, C).
We assayed super-Mendelian transmission of both the y<CC|pF|> gene cassette and allelic-drive of the paraWT allele. As per the cross-scheme depicted in Fig. 2C, a Cas9 expressing transgene (vasa-Cas9) was provided from an unlinked third chromosomal source. Introduction of Cas9 led to a substantial increase of y<CC|pF|> (DsRed+) transmission in F2 progeny, compared to F2 progeny from control (-Cas9) crosses, (46.8% to 85.1% in females and 39.9% to 84.5% in males, Fig. 2D), confirming that the drive element copies with similar efficiency to other drives inserted at this same site22,23,24,25. The fraction of F2 w− receiver progeny, however, was not altered significantly in presence of the Cas9 source (~50% for both control and drive) (Supplementary Fig. 4), indicating that DNA cleavage directed by gRNA-F did not result in frequent production of lethal mutations. Next, we tested F2 w− adult receiver progeny for insecticide resistance. Such receiver progeny derived from y<CC|pF|>; Cas9 bearing mothers displayed a significantly lower percent survival in presence of DDT (39.9% in males, 62% in females) compared to F2 w− receiver flies from control crosses (75.5% in males, Fig. 2E, 85.18% in females, Supplementary Fig. 5B). These results demonstrate biased inheritance of the uncleavable paraWT allele mediated by cleavage and conversion of the IR paraL1014F allele to paraWT allele. Cas9-dependent conversion from F- > L at the 1014 site was confirmed by genotyping individual F2 males. Among F2 w− receiver males from y<CC|pF|> control mothers, the proportion of L1014 wild-type allele was very low (2.7%), consistent with the low recombination rate expected from the close association (~60 kb) between the para locus and the w+mc insertion. In contrast, F2 w− receiver males derived from y<CC|pF|>; Cas9 mothers exhibited a >10-fold increase (29.0%) in the proportion of the uncleavable, wild-type L1014 allele (Fig. 2F, Supplementary Fig. 5D). A similar trend was observed in F2 w− receiver females, where we observed a modest but significant decrease in the percentage of the 1014F allele (Supplementary Fig. 5A, C). However, because females have two X chromosomes, it was difficult to assess the precise conversion frequencies, which are more accurately determined in males. Cumulatively, these results demonstrate a ~30% frequency of allelic conversion from 1014F to L1014, resulting from Cas9-dependent allelic-drive induced by the y<CC|pF|> drive element.
The kdr 1014 F allelic-drive reverses insecticide resistance in population cages
Given the efficient super-Mendelian inheritance of both the y<CC|pF|> drive element and paraWT allele in two-generation test crosses, we next examined their performance over multiple generations in population cages. We seeded three replicate cages with equal numbers of males and females. Half of the females (25% of total flies) were heterozygous for the y<CC|pF|> allelic-drive, Cas9, and L1014 paraWT allele (comprising 16.7% of all para alleles) and the other half of the females (25% of total flies) as well as all males (50% of total flies) carried the IR 1014 F allele. One half of the F1 progeny were monitored for DsRed and DDT resistance phenotypes as well as for genotypes (L1014 versus 1014F sequences) at each generation, while the other half of that population was used to seed the next generation (without sorting for gender).
Cage experiments were conducted in y− w− genetic backgrounds to preclude known mating and fitness advantages associated with wild-type y+ and w+ alleles26,27. Drive cages included the y<CC|pF|> element on the X and vasaCas9-GFP on the 3rd chromosome, while control cages lacked the Cas9 source (Fig. 3A). This scheme permitted assessment of the relative fitness costs associated with different alleles (for example the y<CC|pF|> element versus a y− point mutation, or the para1014L versus para1014F alleles) in the presence or absence of Cas9, as well as evaluating Cas9-dependent drive performance. The fitness costs calculated from the control (–Cas9) cages were used for predictive modeling of population dynamics for the y<CC|pF|> drive (Fig. 3B, Supplementary Fig. 6C) and 1014F allele (Fig. 3D, Supplementary Fig. 6D) and were found to be consistent with two of the three observed cage replicates. The initial divergence of one cage from the others, and hence the averaging model, could stem from variation in certain fitness costs or reflect statistical variation due to sampling, however, this discrepancy disappeared by generation 10. Thus, the y<CC|pF|> element achieved ~80% introduction (final frequency) in all +Cas9 drive cages while increasing only modestly in control (−Cas9) cages (Fig. 3C). In such experiments, generation of uncleavable mutations by non-homologous end joining (NHEJ) events at the yellow locus may limit the spread of the y<CC|pF|> element as previously reported22,23,24. The frequency of the unlinked Cas9 (GFP+) source remained stable over 10 generations suggesting that the vasaCas9 transgene did not impose a detectable fitness cost, even when associated with the drive element (Supplementary Fig. 6A), a phenomenon that has been observed in other contexts28,29. The total population size and sex ratio also remained stable for both the drive and control cages throughout the experiments, suggesting that no significant fitness-induced distortions were operative (Supplementary Fig. 6B).
Regarding the primary objective of these experiments, we also evaluated the DDT resistance status of drive (+Cas9) and control (−Cas9) cage populations over time. Based on prior studies in which equilibrium final frequencies between split-drive cassettes and NHEJ alleles were attained by generations 8/928, we tested generations 9–12 for DDT resistance. We observed a dramatic reversal of DDT resistance from the initial introduced percentage of 83.3% to 15.2 % in males of the drive containing cages (+Cas9) and a more modest, but significant, reduction in females (32%). Control (−Cas9) cages displayed less dramatic reductions in IR (43.3% for males; 55.9% for females), which presumably reflects negative selection based on the fitness costs associated with the 1014F allele (Fig. 3C, Supplementary Fig. 6E, Supplementary Fig. 6F, Supplementary Table 5, Supplementary Table 6). This IR reversal was further confirmed by sequence analysis of males and females from drive and control populations at the 1014 site at three sampled generations (generations 1, 5, and 9). At generation 5, we observed modest parallel decrements in 1014F allele in both +Cas9 drive and −Cas9 controls. However, by generation 9, the frequency of the 1014F allele was greatly reduced in both male and female drive populations. Among males, 1014F representation had dropped precipitously to 13.4% in the drive cages, while decreasing only to 49.3% in controls (Fig. 3F). Similarly, among females, the 1014F frequencies fell to 17.5% in the drive populations compared to 48.3% in control populations (Fig. 3E, Supplementary Fig. 7). Although IR allelic frequencies were not significantly different between drive males and females (i.e., 13.4% versus 17.5%, respectively), we observed a notable difference in their percent survival when exposed to 50 ppm DDT (15.2% versus 32.54%, respectively – Fig. 3C). This greater residual survival observed in females most likely reflects the presence of paraWT/paraL1014F heterozygotes, which display intermediate IR levels (Supplementary Fig. 8). Additionally, physiological differences between males and females may contribute to their differential resistance to DDT30.
In these experiments, recovery of NHEJ events at the para 1014 site was exceedingly rare (Fig. S5A, D). One possible explanation is that most NHEJ mutations that may have been created at the conserved gRNA target site are likely to be non-functional, resulting in lethal mosaicism in females22, and lethality in males. However, as mentioned above, we did not find a significantly biased reduction in the proportion of F2 w− receiver progeny (Supplementary Fig. 4), suggesting that such mutant lethal alleles are not likely to be created at an appreciable rate. Our results from cage trial experiments show that 90% introgression of the L1014 wild-type allele is achieved at generation 9, which is delayed compared to the rapid spread of the drive element (~90% introgression at generation 3). This kinetic difference suggests that cleavage induced by gRNA-F may be less efficient than for gRNA-y, as inferred by estimates from model fitting (Supplementary Tables 5 and 6). However, because gRNA-F generates few NHEJ alleles, its modest, but clean, cleavage profile supports a delayed but eventually superior drive outcome. Another factor likely to contribute to the extensive reversal in IR obtained in these drive experiments is the synergistic action of two independent selective processes24 wherein the allelic drive promoting Super-Mendelian inheritance of the L1014 allele (e.g., mean copying efficiency to receiver chromosomes = 0.29) acts in combination with the greater relative fitness of the wild-type L1014 versus 1014F allele (e.g., median fitness cost of 1014F allele in males = 0.28) (Fig. 1C–F, Supplementary Tables 5 and 6), a feature also likely to be relevant in field contexts5,7,9,31,32,33,34. We note that this synergy between the independently acting gene-drive and positive selection for the 1014 allele in the driving configuration (+Cas9) also may be reflected by the reduced variance in allelic frequencies between cages observed in generation 9 (Fig. 3F, most obvious in males) compared to the action of positive selection alone (−Cas9), a phenomenon we have observed in other studies24. Presumably, the likelihood of 1014 L allele inheritance is increased as a result of these two cooperative processes.