Fine-scale spatial and temporal dynamics of kdr haplotypes in Aedes aegypti from Mexico
As resistance to insecticides increases in disease vectors, it has become exceedingly important to monitor populations for susceptibility. Most studies of field populations of Aedes aegypti have largely characterized resistance patterns at the spatial scale of the city or country, which may not be completely informative given that insecticide application occurs at the scale of the house or city block. Phenotypic resistance to pyrethroids dominates in Ae. aegypti, and it has been partially explained by mutations in the voltage-gated sodium channel gene. Here, we assess community-level patterns of four knockdown resistance (kdr) haplotypes (C1534/I1016, F1534/I1016, C1534/V1016 and F1534/V1016) in Ae. aegypti in 24 randomly chosen city blocks from a city in Yucatán State, Mexico, during both the dry and wet season and over two years.
Three of the four haplotypes, C1534/I1016, C1534/V1016 and F1534/V1016 were heterogeneous between city blocks at all four sampling time points, and the double mutant haplotype, C1534/I1016, showed a significant increase following the wet season. The F1534/I1016 haplotype was rarely detected, similar to other studies. However, when haplotype frequencies were aggregated to a coarser spatial scale, the differences in space and time were obscured.
Our results provide empirical evidence that the selection of kdr alleles is occurring at fine spatial scales, indicating that future studies should include this scale to better understand evolutionary processes of resistance in natural populations.
KeywordsAedes aegypti kdr Pyrethroid Insecticide resistance
The recent introduction of Zika and chikungunya into the Americas, along with the global persistence of dengue, has made Aedes aeygpti one of the most important disease vectors worldwide . Aedes aegypti are highly anthropophilic mosquitoes that live in close association with humans, primarily in urban areas. They live and breed in and around houses, with immature stages developing in a wide variety of peridomestic water-holding containers, making them very difficult to control . Aedes aegypti control programs employ a combination of methods including ultra-low volume spraying (ULV), indoor space spraying (ISS), indoor residual spraying (IRS), source reduction, and larviciding [3, 4, 5]. Unfortunately, most vector control measures, which are the only means to prevent most diseases spread by Ae. aegypti, have largely failed in controlling disease transmission in regions where dengue is endemic [6, 7].
Reliance on chemical control strategies has led to the development of resistance to many of the insecticides used to control Ae. aegypti throughout the world [3, 8]. Historically, the use of pyrethroid insecticides for vector control has been widespread due to their low cost and low mammalian toxicity  and consequently pyrethroid resistance has become a major source of treatment failure . In Ae. aegypti, resistance to pyrethroid insecticides most commonly arises due to increased metabolic activity or structural alterations at the target site of the insecticide . Point mutations in the para-orthologous sodium channel gene are associated with resistance to both pyrethroid and organochlorine insecticides in Ae. aegypti . Collectively termed “knockdown resistance”, or kdr, these mutations reduce the ability of insecticides to bind to the sodium channel . Several kdr mutations have been identified in Ae. aegypti  and two have been previously associated with pyrethroid resistance in Ae. aegypti from Mexico, specifically the valine to isoleucine mutation on codon 1016 of the voltage-gated sodium channel gene and a phenylalanine to cysteine mutation on codon 1534 [11, 12, 13]. Both the I1016 and C1534 kdr mutations are well established in the literature as significantly associated with pyrethroid resistant phenotypes [3, 14, 15, 16], and recently it has been suggested that there is co-evolution of the two markers , yet their combined impact on phenotype is not clear.
The frequency of kdr mutations has been increasing rapidly in time and space throughout the world. For example, in Mexico, the frequency of I1016 increased from < 0.1% in 1996–2000 to 88.3% in 2007–2009 . Brazil experienced a similar increase in both the C1534 and I1016 alleles: in 2002 neither allele was present, in 2006 both were detected at low frequencies, and by 2012 both alleles were widely prevalent . Not only are kdr alleles increasing in frequency in locations where they had previously been detected, but they are also being detected in new locations at relatively high frequencies. For example, the first report of kdr in Indian Ae. aegypti in 2015 detected a frequency of C1534 of up to 0.79 in one location in the country . Similarly, the G1016 and P989 kdr alleles were recently widely detected above 80% frequency in Myanmar . Given these increases, it has become important to understand the patterns of resistance associated with kdr and their persistence in natural populations.
Previous studies of kdr in field populations of Ae. aegypti have largely extrapolated regional patterns of kdr frequencies from sampling relatively few locations in a country [17, 18, 20], or they offer a snapshot through cross-sectional studies that report kdr frequencies taken at a single time point [19, 21, 22]. These sampling schemes, which pool specimens from multiple breeding sites, would be appropriate if Ae. aegypti populations are panmictic at coarse spatial scales (e.g. province or country), as previous research has suggested , or if sampling occurred within the limited flight range of Ae. aegypti (approximately 50–150 m ). However, the heterogeneity of insecticide selection pressures may render this assumption invalid. One hypothesis is that the use of ULV applications of insecticides to control Ae. aegypti would lead to homogeneous distributions of resistance within larger geographic areas (e.g. neighborhoods). Conversely, the use of ISS or IRS for the control of Ae. aegypti may lead to a more heterogeneous pattern, driven by the extent and frequency of highly focused insecticide applications . Given the reactive nature of dengue control, and the potential for strong variability in insecticide selection pressure in space and time, there is a need to understand the spatial scale of insecticide resistance dynamics.
A recent study by Deming et al.  found significant heterogeneity in kdr frequencies between interconnected towns in Yucatán State, Mexico, with some towns displaying heterogeneity between city blocks. Building on this research, a comprehensive sampling of Ae. aegypti was conducted in one of the study towns over a two year period, aiming to answer two questions: (i) are kdr frequencies heterogeneous at the spatial scale of city blocks, or are they more accurately assessed at the neighborhood level, and (ii) to what extent can kdr frequencies vary over a relatively short timeframe, namely between the wet and dry season within a year? Understanding the spatial and temporal dynamics of kdr frequencies is not only important for monitoring insecticide susceptibility, but also for designing interventions to mitigate and manage resistance while controlling populations effectively.
Mexico’s national vector control program utilizes ULV, ISS and larviciding to control Ae. aegypti. In the year 2000, the program began using permethrin-based products, but began transitioning to non-permethrin products in 2009 [11, 26]. At the time of the study in Hunucmá, the insecticides malathion, chlorpyrifos, bendiocarb and deltamethrin were used for adult Ae. aegypti control (Secretary of Health, Yucatán). Previous research indicated that in Hunucmá in 2013, the Ae. aegypti population was resistant to deltamethrin and the frequency of the C1534 kdr allele was 0.568 and I1016 was 0.425 .
Insecticide use data
The Yucatán’s Secretary of Health provided information from their Entomological Surveillance and Vector Control Database  about the Ae. aegypti control efforts in Hunucmá during the duration of the study. These data included the types of insecticide, mode of application (ULV, ISS, or larvicide), date and location of application at the city block level. Insecticide-based interventions were typically conducted reactively in response to reported dengue or chikungunya cases.
We used a Prokopack aspirator  to collect adult mosquitoes from each house for 10 minutes or until we sampled from every available room. All collected mosquitoes were transported in plastic cups to the laboratory at the Universidad Autonoma de Yucatán in Merida, where they were placed at -20 °C for euthanization. We identified each mosquito to the species level and sex, and Ae. aegypti were stored in RNAlater® Stabilization Solution (Thermo Fisher Scientific, Waltham, MA, USA) for future kdr genotyping.
During the dry season collections in January of each year, there were low numbers of adult mosquitoes collected inside homes. To augment sampling efforts and supplement Prokopack collections, we placed oviposition traps (ovitraps) outside four houses per block (one on each side of the block) for up to 4 weeks to collect Ae. aegypti eggs. The eggs were then aggregated by block and reared in the insectary at the Universidad Autonoma de Yucatán. Thirty-five to fifty adults per block emerging from the eggs collected from the ovitraps were selected at random for subsequent kdr genotyping. Adults from ovitraps were only used to complement block collections with less than 30 adults collected by Prokopacks. Our strategy of using four traps per block over a period of 4 weeks aimed at minimizing bias typical of ovitrap collections (i.e. sampling related individuals at a rate higher than what we would collect indoors with Prokopacks). One city block had enough mosquitoes to compare adult and ovitrap data and yielded no significant differences in the frequency of kdr mutations in both samples (for Prokopack, the frequency of C1534 was 18/30 and with the ovitrap it was 20/26; χ2 = 1.13, df = 1, P = 0.2867; for I1016 it was 14/30 with the Prokopack and 10/26 with the ovitrap; χ2 = 0.121, df = 1, P = 0.7278).
Both male and female mosquitoes collected from blocks were used for genotyping since they both contribute to the genetic pool of the next generation. Individual mosquitoes were placed in individual wells of a 96-well plate in a 50 μl solution containing 5 μl of Taq 10× buffer (containing 500 mM KCl, 100 mM tris HCl, 15 mM MgCl2 and 1% Triton X-100) and 45 μl of sterile ddH2O and heated in an Eppendorf Mastercycler© Pro thermocycler at 95 °C for 15 min. Allele-specific real time PCR was conducted using a BioRad© CFX96 real-time PCR machine (Hercules, CA, USA) based on the protocols and primer sequences described by Saavedra-Rodriguez et al.  for position 1016 and Yanola et al.  for position 1534. Reaction conditions were as follows: for locus 1016, each 20 μL reaction consisted of 8 μl of PerfeCTa® SYBR® Green FastMix (Quanta Biosciences, Gaithersburg, MD, USA), 8.86 μl of ddH2O, 0.34 μl of 10 μM Val1016 forward primer, 0.4 μl of the 10 μM Iso1016 forward primer, 0.4 μl of the 10 μM Iso1016 reverse primer and 2 μl of template DNA; for locus 1534, each 20 μl reaction consisted of 9 μl of PerfeCTa® SYBR® Green FastMix (Quanta Biosciences), 7.15 μl of ddH2O, 0.65 μl of 10 μM Cys1534 forward primer, 0.60 μl of the 10 μM Phe1534 forward primer, 0.60 μl of the 10 μM Cys1534 reverse primer and 2 μl of template DNA. Melting curve analyses determined the genotypes of the mosquito at each codon [14, 15]. Each PCR plate included three control Ae. aegypti mosquitoes of known genotypes at the two kdr loci: a homozygous wild-type individual, a heterozygous individual and a homozygous mutant individual (separate controls were used for each mutation).
The F1534C and V1016I mutations are located close together on the chromosome, so we tested for linkage disequilibrium between the two markers by using the equations outlined in Gillespie  to calculate the coefficient D, r2, and the Chi-square statistic (χ2) with one degree of freedom. The mutations were in linkage disequilibrium for all time points (First time point: D = 0.137, r2 = 0.31, χ2 = 541.9, P < 0.0001), so we estimated haplotype frequencies for each block and sector at each time point. Full methods for haplotype calculations are found in Additional file 1.
Haplotype frequencies, except for F1534/I1016 which was rarely detected, were mapped using QGIS 2.18 (QGIS Development Team, 2016) at both the block and sector level for each sampling time point. Sampling time points in June-July were characterized as the wet season and time points in January were defined as the dry season.
We used a multiple proportions test with a Chi-square distribution to test if at least one city block or sector had significantly different haplotype frequencies than the others for each time point. The same test was used to see which blocks or sectors had different haplotype frequencies at least one point in time. The presence of spatial autocorrelation between block-level allele frequencies was assessed with the Moran’s I statistic using an inverse distance weighted scheme in the R package ape . Moran’s I is similar to Pearson’s r correlation, yet it tests correlation of values in space. To test for an overall effect of time on block-level allele frequencies, we used a linear mixed model with time point as the predictor and block nested within sector as the random intercept using the R package nlme . Additionally, the effect of season on block-level allele frequencies was analyzed by combining the data from the two dry and two wet season collections. For this analysis, we used a linear mixed model with block as a random intercept and season as fixed effect. Lastly, we used a linear mixed model with block as a random intercept to assess the relationship between the mean number of adult mosquitoes collected in the houses per block and the allele frequencies per block. Only wet season data were used for this analysis due to the limited number of mosquitoes collected inside houses during the dry season.
Summary statistics for all Ae. aegypti collected and analyzed from Hunucmá, Mexico
No. of females collected
No. of males collected
Total no. collected
No. of C1534 genotyped
No. of I1016 genotyped
kdr haplotype frequencies in space
kdr haplotype frequencies over time
The block-level haplotype frequencies also significantly changed over time. Using the first sampling point (wet season 2014) as the reference with a null hypothesis of no change over sampling points, there was a significant increase in the C1534/I1016 haplotype frequency of 0.113 ± 0.04 during the following dry season in 2015 (hierarchical linear mixed-effects model, time effect t = 2.67, P = 0.009). Accordingly, the wild-type haplotype, F1534/V1016, showed a significant decrease in block-level frequencies of 0.09 ± 0.04 between the wet season of 2014 and the following dry season in 2015 (hierarchical linear mixed-effects model, time effect t = -2.11, P = 0.038). Aggregating by sector obscured the differences in the wild-type haplotype frequencies over time, showing no significant differences between any of the time points. However, there was still a significant increase of 0.127 ± 0.04 in the frequency of C1534/I1016 between the wet season of 2014 and the dry season of 2015 at the sector level (hierarchical linear mixed-effects model, time effect t = 3.10, P = 0.0127).
Significant heterogeneity in the frequency of kdr haplotypes was detected between city blocks of a dengue endemic town in the Yucatán, suggesting that selection for these haplotypes is occurring at a fine spatial scale. Municipal insecticide application in the study area is highly variable in space and time, creating a patchy mosaic of selection pressures. Theory predicts that in such an environment, if the rate of dispersal or migration between areas is greater than the strength of selection, there will be no local adaptation and the allele with the best average fitness across habitats will increase towards fixation, homogenizing populations . However, if the strength of selection is higher, then local adaptation can occur, with immigration limiting the probability of fixation of an advantageous allele in a given habitat . Population genetics studies have previously shown that Ae. aegypti populations in the Yucatán (including Hunucmá) are panmictic [23, 35], suggesting that migration is high and therefore kdr frequencies should be similar within towns. Yet, evidence from town-level data  and from the present study suggest that local adaptation to pyrethroid insecticides is occurring.
Our data also show temporal heterogeneity in kdr haplotype frequencies throughout the year, suggesting rapid selection for these markers. Overall, town blocks showed higher kdr haplotypes and allele frequencies during the dry season, which occurs directly following the wet season that coincides with disease transmission. During the wet season, there was an increase in insecticide application to control mosquitoes, subjecting the population to higher selection pressure than during the dry season. The use of household aerosol insecticides, highly prevalent in the Yucatán , could have acted as another important selective pressure during the transmission season, though this study unfortunately did not quantify use. However, a previous study from Merida, Mexico found that approximately 87% of households use commercially-available pyrethroid products to control mosquitoes in their homes . Although we were unable to statistically assess the association between insecticide application and kdr allele frequencies, it is likely that increased insecticide application during the wet season contributed to the increase in kdr haplotypes and allele frequencies that were detected during the dry season.
The relationship between insecticide use and kdr frequencies has been well-documented in Anopheles spp.; many researchers have detected an increase in kdr following deployment of insecticide-treated bednets (ITN) or long-lasting insecticide-treated nets (LLIN), both containing pyrethroids, for malaria prevention [38, 39, 40]. For example, Czeher et al.  sampled An. gambiae in Niger over the course of three years, pre- and post-LLIN intervention, and found that the frequency of the L1014F kdr allele was heterogeneous over 14 study sites and increased over time. Similarly, Stump et al.  found an increase in the L1014S kdr allele in An. gambiae in western Kenya following a 15 year period of ITN use, though only in the intervention village and its nearest neighbor (~5 km apart); villages at least 20 km away did not exhibit an increase in kdr frequencies. These studies illustrate that insecticide exposures can drive kdr frequencies in Anopheles spp., and provides justification for further research into these associations in Ae. aegypti.
The lower kdr haplotype frequencies detected in this study during the early wet season, where the preceding months had low insecticide application, may be indicative of a fitness cost associated with the kdr alleles [41, 42, 43]. It is encouraging that frequencies of kdr haplotypes can be reduced in the field within relatively few mosquito generations, which may facilitate the eventual restoration of insecticide susceptibility if kdr is the primary mechanism responsible for insecticide resistance in this population. While phenotypic resistance was not measured in the present study, a previous study found a significant association between the kdr alleles and phenotypic resistance to deltamethrin at the same study site . Since kdr is not the only mechanism that confers pyrethroid resistance, the role of metabolic resistance mechanisms in this population needs to be studied in greater detail.
The fine-scale spatial and temporal heterogeneity in the frequency of alleles associated with insecticide resistance has important implications for monitoring susceptibility in field populations. When kdr haplotype frequencies were analyzed at the sector level, the differences in time and space were obscured, displaying a more homogeneous pattern. This suggests that important insecticide resistance-related trends may be missed if single sites are assumed to be broadly representative of larger geographies. The marked spatial randomness found in the kdr haplotype frequencies at the block level requires a sampling design that selects multiple, random blocks to provide a truly representative picture of kdr in an area. Operationally, this would require modifications to existing insecticide resistance surveillance strategies. Additionally, seasonal changes in kdr frequencies suggest that cross-sectional studies, or even longitudinal studies that sample at the same time point every year, may not capture evolutionary dynamics. Moreover, how phenotypic resistance is affected by the observed changes in allele frequency needs to be described. Ultimately, understanding the processes underlying the patterns that are observed in the field can inform the improvement of mosquito control strategies and mitigate the negative impacts of the rapid evolution of resistance.
We would like to thank the entire staff at Unidad Colaborativa para Bioensayos Entomologicos at Universidad Autonoma de Yucatán for their assistance in sample collection and processing. The findings and conclusions in this paper are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.
This project was funded by grants from National Science Foundation (NSF/DEB/DDIG #1601520), Mexico’s CONACYT (Project #255141), and the ARCS Foundation.
Availability of data and materials
Data are publicly available at the Dryad Digital Repository
MKG, AL, PMS and GMVP designed the study. MKG, JR and AMB collected data and managed field teams. MKG and JR conducted laboratory assays and analyzed data. MKG and GMVP drafted the manuscript. All authors read and approved the final manuscript.
Ethics approval and consent to participate
Emory University Institutional Review Board (IRB00082848) approved all study protocols. Written informed consent was obtained from the house owner, and houses that did not consent to the study were not included.
Consent for publication
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 15.Yanola J, Somboon P, Walton C, Nachaiwieng W, Somwang P, Prapanthadara L-A. High-throughput assays for detection of the F1534C mutation in the voltage-gated sodium channel gene in permethrin-resistant Aedes aegypti and the distribution of this mutation throughout Thailand. Trop Med Int Health. 2011;16:501–9.CrossRefGoogle Scholar
- 17.Linss JGB, Brito LP, Garcia GA, Araki AS, Bruno RV, Lima JBP, et al. Distribution and dissemination of the Val1016Ile and Phe1534Cys kdr mutations in Aedes aegypti Brazilian natural populations. Parasit Vectors. 2014;7:25.Google Scholar
- 25.World Health Organization. Dengue: Guidelines for diagnosis, treatment, prevention and control. Geneva: World Health Organization; 2009.Google Scholar
- 31.Gillespie JH. Population Genetics. Second edition. Baltimore, MD: John Hopkins University Press; 2004.Google Scholar
- 33.Pinheiro J, Bates D, DebRoy S, Sarkar D, R Core Team. nlme: Linear and Nonlinear Mixed Effects Models. R package version 3. p. 1–137. https://CRAN.R-project.org/package=nlme
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.