Fitness costs in the presence and absence of insecticide use explains abundance of two common Aedes aegypti kdr resistance alleles found in the Americas

Aedes aegypti is the vector of viruses such as chikungunya, dengue, yellow fever and Zika that have a critical impact on human health. Control of adult mosquitoes is widely done using pyrethroids, but resistance has reduced the effectiveness of this class of insecticides. Resistance to pyrethroids in mosquitoes is commonly due to mutations in the voltage-gated sodium channel (Vgsc) gene (these mutations are known as knockdown resistance, kdr). In the Americas and the Caribbean, the most common kdr alleles are 410L+1016I+1534C and 1534C. In this study, we conducted a population cage experiment to evaluate changes in the allele and genotype frequencies of the 410L+1016I+1534C allele by crossing two congenic strains; one carrying the 410L+1016I+1534C and another with the 1534C allele. Changes in allele frequencies were measured over 10 generations in the absence of insecticide exposure. We also applied one cycle of selection with deltamethrin at F9 to evaluate the changes in allele and genotype frequencies. Our findings indicate that fitness costs were higher with the 410L+1016I+1534C allele, relative to the 1534C allele, in the absence of deltamethrin exposure, but that the 410L+1016I+1534C allele provides a stronger advantage when exposed to deltamethrin relative to the 1534C allele. Changes in genotype frequencies were not in Hardy-Weinberg equilibrium and could not be explained by drift. Our results suggest the diametrically opposed fitness costs in the presence and absence of insecticides is a reason for the variations in frequencies between the 410L+1016I+1534C and 1534C alleles in field populations.


Introduction
Aedes aegypti is the vector of viruses that have devastating impacts on public health and insecticide control of adult mosquitoes is most commonly carried out using pyrethroids.A. aegypti is the vector of viruses such as chikungunya, dengue, yellow fever and Zika [1][2][3][4][5].To reduce the burden caused by these diseases, control of adult mosquitoes is mainly done using a class of insecticides known as pyrethroids, with deltamethrin being a common pyrethroid applied for control of A. aegypti in the field.Pyrethroid toxicity occurs due to the binding and disruption of the voltage-gated sodium channel (VGSC).
Relative to insecticide susceptible alleles, resistance alleles commonly have a fitness cost in the absence of insecticide exposure [19].However, field populations often include multiple resistance alleles.Thus, the fitness of the different resistance alleles to each other (in both the presence and absence of insecticide) is important to understanding the evolution of resistance.For example, in Musca domestica the relative fitness in the presence of pyrethroids of the different kdr alleles is super-kdr (M918T+L1014F) > kdr (L1014F) > kdr-his (L1014H), but in the absence of insecticide the relative fitness is susceptible > kdr-his > kdr > super-kdr [20], suggesting that alleles with multiple mutations might have a greater fitness cost than those with a single mutation.In A. aegypti, the relative fitness costs of different kdr alleles (to other kdr alleles) is much less studied.This is problematic because many resistance alleles coexist in field populations, and some potentially impose more detrimental disadvantages relative to other resistance alleles in the absence of insecticide exposure.Therefore, it is valuable to understand not only the relative fitness of kdr alleles relative to the susceptible alleles, but also to understand the relative fitness of the different kdr alleles to each other.Such information is needed to better understand the patterns of alleles found in field populations.
In this study, we investigated the relative fitness of two A. aegypti kdr alleles (410L+1016I+1534C and 1534C) relative to each other, in the absence and presence of insecticide exposure.We wanted to assess whether a resistance allele with multiple mutations imposed greater fitness costs relative to the allele with only one mutation (in accordance with what was previously found in M. domestica).Overall, the 410L+1016I+1534C allele imposes both an astonishing fitness cost in the absence of insecticide, and a strong advantage relative to 1534C in the presence of deltamethrin exposure.

Mosquito strains
Two congenic strains of A. aegypti were used in this study.Both strains share the same pyrethroid-susceptible background from the Rockefeller strain (ROCK), but carry different kdr alleles.The LMRKDR:ROCK (LKR) is a pyrethroid-resistant strain which is congenic to ROCK, but contains the kdr allele 410L+1016I+1534C [18].The 1534C:ROCK is a pyrethroidresistant strain which is congenic to ROCK, but contains the F1534C allele [17].In both congenic strains, the kdr alleles are the only mechanism of resistance to pyrethroids [17,18].

Allele competition experiments
To conduct the allele competition experiments (also known as population cage experiments), the two congenic strains were crossed: Cross A (LKR females x 1534C:ROCK males) and reciprocal cross B (1534C:ROCK females x LKR males)(Fig 1).For each reciprocal cross, 400 unmated females and 200 males were released into a cage and were allowed to mate en masse for seven days.The offspring resulting from crosses A and B were split into three cages (A1, A2, A3, B1, B2 and B3) as shown in Fig 1 .Mosquitoes were reared at 25.0-28.5˚C(average = 25.6), 26.8-54.5% (average = 47.9%)relative humidity, and a 14:10-h (light/dark) photoperiod.All replicates were run at the same time.Mosquitoes were reared as previously described [21].Approximately 800 pupae from the larval containers were placed into cages for each following generation.Additional emerging males and females were stored separately in 2 ml Eppendorf tubes at -80˚C until they were genotyped.
The genotyping of the mosquitoes was done by allele specific polymerase chain reaction (ASPCR) using primers that were allele specific at codon 1016 of the Vgsc (one allele has 410L +1016I+1534C and allele 1534C has V410+V1016+1534C).The genomic DNA from mosquito heads was extracted as previously described [22] and ASPCR was done using the following thermocycler conditions: For the 1534C allele, 94˚C for 3 min, 36 cycles of 94˚C for 30 sec, 62˚C for 30 sec and 72˚C for 1 min, and 72˚C (7 min).For the 410L+1016I+1534C allele, 94˚C for 3 min, 34 x (94˚C for 30 sec, 65˚C for 30 sec, 72˚C for 1 min) and 72˚C (7 min).We used 1 μL of extracted DNA as templates for ASPCR.The primers used for each PCR reaction are shown in S2 Table .Each PCR reaction was evaluated on a 1% agarose gel and was scored as homozygous V1016 (PCR band only with V1016 primers), homozygous for 1016I (ASPCR band only with 1016I primers) or heterozygous (PCR band with both 1016V and 1016I primers).For each biological replicate, 45 males and 45 females were genotyped.Every PCR plate included control DNA from 2 samples of LKR, 1534C:ROCK and F 1 (control DNA was previously sent for sequencing to validate the expected codon V1016I for each corresponding strain or F 1 ).In addition to the controls, the codon V1016I was verified for 16 samples from A2F5 and A1F10, eight samples each).The sequencing results for the controls and samples agreed with the results from the gels.Each sample was genotyped using two technical replicates by running PCR reactions and gels twice.
The genotyping results were scored independently by three judges (JJS, JGS and CRF).Each judge assigned a genotype for each mosquito in both technical replicates.If the technical replicates agreed, a final genotype was assigned.A consensus genotype was determined if there was agreement between two out of three judges.If three judges disagreed on the final genotype, then those samples were considered for a third technical replicate.Third replicates were done if there were more than 10 samples per biological replicate for which there was no consensus genotype; otherwise, these samples were dropped from analysis.

Selection of F 9 using deltamethrin
We selected 3-7-day old males and females with deltamethrin.The doses used for males and females (0.039 ng/mosquito and 0.078 ng/mosquito, respectively) were chosen to give approximately 80% kill.To conduct the selections, mosquitoes were knocked down with CO 2 and placed in a paper cup on a bucket filled with ice to keep them anesthetized.

Allele competition data analysis
Differences between allele and genotype frequencies across generations were tested using linear mixed models and checked for significance from F-values generated from ANOVA in R as previously described [21].
Deviation from Hardy-Weinberg equilibrium (HWE) within generations was assessed using a chi-square test (χ2) to compare the observed genotype counts relative to the expected genotype counts calculated using the allele frequencies of the same generation [21].The level of significance for statistical analyses was p < 0.05.Simulations were used to evaluate the likelihood of the observed allele frequency changes due to genetic drift.Simulations were done with R software [23] and scripts applied as in a previous study [24].The model assumed a diploid and panmictic population with a fixed size of 800 individuals and the initial 410L+1016I +1534C allele frequency for each generation was defined as the observed frequency in the previous interval.Simulations were repeated 50,000 times per generation and p-values were estimated as the number of simulations in which ending allele frequency was equal or more extreme than the initial value of the interval divided by the total number of simulations.The null hypothesis assumed all fluctuations in allele frequencies could be explained by genetic drift.The p-values obtained for the HWE and genetic drift were adjusted for multiple comparisons using Holm's test [25].

Results
There were no significant differences in the allele frequencies over generations between reciprocal crosses A and B (Fig 1 ) for the unselected (ANOVA, p-value = 0. 951) or deltamethrinselected generations (ANOVA, p-value = 0.174).There were no significant differences in the allele frequencies over generations between males and females in the absence (ANOVA, pvalue = 0.888) nor presence of deltamethrin selection (ANOVA, p-value = 0.774.For this reason, allele frequencies from crosses A and B (and from males and females) were pooled for further analysis.
There was a remarkable fitness disadvantage to the 410L+1016I+1534C allele, relative to the 1534C allele, in the absence of insecticide.The fitness cost was most dramatic from the F 1 to F 3 and F 5 where the 410L+1016I+1534C allele decreased from an average frequency of 50.0% to 21.3% and 11.5%, respectively (S1 Table ).The frequency of the 410L+1016I+1534C allele remained relatively unchanged after the F 5 (Fig 2 ).The frequency of the 410L +1016I+1534C allele decreased in a trend that is best fitted by a power function model (F 1,34 = 276.3,r 2 = 0.89, ANOVA = 2.20x10 -16 ).Changes in genotype frequencies over time are shown in Fig 3 .The genotype frequencies were not in HWE (within generation) at the F 3 , were largely in HWE in the F 5 and F 7 , and were variable between replicates for the F 7 and F 9 (S1 Table ).Our simulations indicated that the changes in allele frequencies from the F 1 could not be accounted for by drift (all p-values = 0).However, as the changes in allele frequency between generations became more subtle (F 7 -F 10 ), drift was found to possibly explain 8 out of 18 results (comparisons made to previous generation) (S1 Table ).

and S1 Table
The 410L+1016I+1534C allele exhibited an extraordinarily strong fitness advantage after exposure to deltamethrin with the allele frequency rising from 16% to 62.1% (Fig 2).As expected for a population under selection, the genotype frequencies do not follow the assumptions of HWE (p-values were < 0.0001) (S1 Table ).These results are consistent with the observation that the 410L+1016I+1534C allele gave higher levels of deltamethrin resistance than the 1534C allele [18].
The frequencies of the 410L+1016I+1534C and 1534C alleles are strongly influenced by the fitness disadvantage of the 410L+1016I+1534C allele relative to the 1534C allele in the absence of insecticides.In contrast in the presence of deltamethrin, the 410L+1016I+1534C allele has an enormous fitness advantage.During periods of use of pyrethroids such as cyhalothrin (to which the 410L+1016I+1534C and 1534C alleles confer similar levels of resistance [17,18]), both alleles would be favored, relative to susceptible alleles.These results bear some similarity to kdr alleles in house fly where the super-kdr allele (having two mutations) confers higher levels of resistance, but also has a greater fitness cost in the absence of insecticide, compared to kdr-type alleles with a single mutation [20].Given the importance of the VGSC to neurological function, it would be interesting to explore what neurophysiological changes might underline the fitness costs of kdr alleles.
The fitness cost of the 410L+1016I+1534C allele (relative to the 1534C allele) was remarkably higher than has been found for comparisons of kdr versus susceptible alleles.For example, using a similar population cage experiment, the 989P+1016G [26] and the F1534C [21] alleles were shown to have a fitness cost relative to a susceptible allele, but the change in allele frequencies across generations was far less pronounced (frequency of the kdr allele decreased in a linear manner and was >0.3 by the F 7 ) than we observed in this study.Another study used a strain of A. aegypti congenic to the pyrethroid-susceptible ROCK strain, but carrying an unknown kdr allele (although it was known to carry the V1016I mutation) and measured fitness costs.The frequency of the allele containing the 1016I mutation decreased through 15 generations in the absence of pyrethroid exposure (cross A: r 2 = 0.527, p = 0.0006 and cross B: r 2 = 0.569, p = 0.0003, respectively) [27]  Our findings show a strong fitness cost associated with the 410L+1016I+1534C allele in the absence of insecticide exposure, but this allele provides a fitness advantage when exposed to deltamethrin, relative to the 1534C allele.Given the dramatic fitness cost of the 410L+1016I +1534C allele relative to the 1534C allele in the absence of insecticide, it is possible that the functionality of the VGSC is more drastically impaired for the former allele compared to the latter and investigating the neurophysiological differences between these alleles would be insightful.

Fig 2 .
Fig 2. Frequency changes of the 410L+1016I+1534C allele over 10 generations in the absence and presence of deltamethrin selection.The solid black line represents the changes in allele frequencies in the absence of deltamethrin exposure are represented by a power function model (F 1,34 = 276.3,r 2 = 0.89).The dashed orange line represents the changes in allele frequencies from F 9 to F 10 in the presence of deltamethrin selection (linear model F 1,10 = 186.1,r 2 = 0.95).Different letters represent significant differences based on Tukey's test.https://doi.org/10.1371/journal.pntd.0011741.g002

Fig 3 .
Fig 3. Changes in genotype frequencies over generations in the absence and presence of deltamethrin selection.Different letters represent significant differences based on Tukey's test.https://doi.org/10.1371/journal.pntd.0011741.g003 at a rate similar to what was seen for the 1534C and 989P+1016G alleles.It would be helpful to know what the allele was in those experiments, as well as what role (if any) the enhanced detoxification enzymes played in the changes in allele frequency.