Introduction

Arboviruses (Arthropod-borne virus) infections, such as dengue, Zika, chikungunya and yellow fever are causing severe impacts on global health, especially in tropical low-income countries1. In Brazil, dengue is considered as hyperendemic, with regional and seasonal circulation of the four serotypes (I–IV), but other newly emerged viruses such as Zika (ZIKV) and chikungunya (CHIKV) are also circulating2. In 2022, Brazil reached the highest number of deaths caused by dengue fever (1016 deaths) ever recorded in the country3, and the recent increase in yellow fever cases in regions bordering urban centers4 highlights the imminence of re-urbanization of this disease5. As there are no effective vaccines available for most of these arboviral diseases (excluding yellow fever), controlling the vector—the mosquito Aedes aegypti (Diptera, Culicidae: Linnaeus, 1762)—is the method of choice for reducing the risk of arbovirus transmission.

Chemical control is still the cornerstone of any vector borne-diseases control programme worldwide6,7. Historically, insecticides have played an essential role in the decline of various diseases, including malaria and dengue8. However, the massive and repeated use of chemicals for decades has favored the spread of insecticide-resistance in Aedes mosquitoes9. Currently, insecticide resistance (IR) is seen as one of the most important threats for the control of diseases caused by viruses transmitted by Aedes, as it may reduce the efficacy of chemical-based vector control interventions10,11. In Brazil, nationwide insecticide resistance monitoring (MRI) has reported strong resistance of Ae. aegypti to all pesticides used by the Ministry of Health, such as the larvicide temephos (organophosphate), the insect growth regulator pyriproxyfen, and the adulticides deltamethrin (pyrethroid) and malathion (organophosphate)12,13.

Among the genetic alterations involved in IR, the most common are single nucleotide changes in the voltage-gated sodium channel (NaV), which cause resistance to the knockdown effect of pyrethroids, and are therefore called the kdr mutations (knockdown resistance)14. In Brazil, at least three kdr mutations have been reported in Ae. aegypti: a substitution of Valine to Isoleucine at position 1016 (V1016I), a Phenylalanine to Cysteine at position 1534 (F1534C), and a Valine to Leucine at position 410 (V410L)15,16. These mutations were classified as NaVR1 and NaVR2, where NaVR1 is a haplotype containing the F1534C kdr mutation only while NaVR2 exhibits F1534C + V410L + V1016I mutations17. These alleles give pyrethroid resistance under a recessive trait, and the NaVR2 causes the highest level of resistance to deltamethrin18. Other SNPs, such as V1016G, S989P and T1520I, are common in Asian populations of Ae. aegypti19,20 but they are absent in Brazil and neighboring countries.

Monitoring the presence and spread of kdr mutations through molecular methods is relevant because pyrethroids are widely used for vector control and household domestic purposes21,22,16. Additionally, the analysis of the genetic structure and gene flow among vector populations can contribute to a better understanding of the evolutionary forces driving the dispersion of resistance alleles23.

In Brazil, differences in the genetic structure and spatial distribution of pyrethroid (kdr) resistance have been reported in Aedes aegypti with at least three well defined clustered regions within the country16. The Amapá State (AP), located in the Northern part of the country, bordering the French Guiana, is an important gateway between Brazil and the Caribbean to arboviruses and vector populations with distinct characteristics, including insecticide resistance24. By the way, this was one of the entrance doors for the CHIKV virus in 2015 in Brazil25. In addition, the geographical aspects of the region raise relevant questions for population genetic studies since locations infested with Ae. aegypti are relatively isolated from each other’s due to the vast Amazonian vegetation. In AP, Ae. aegypti from Oiapoque (the city bordering the French Guiana) is among the Brazilian mosquito populations exhibiting the highest levels of resistance to deltamethrin, temephos and malathion, while the Ae. aegypti population from the AP State capital, Macapá city, is far less resistant12,26,27.These cities are 577 km distant, connected by a road (BR-156) that crosses the dense Amazon Forest. The frequency of kdr alleles is distinct between them, with a predominance of kdr NaVR1 and kdr NaVR2 in Macapá and Oiapoque, respectively24. The presence and frequency of these kdr alleles in other cities of the AP State is however unknown.

In this study, we conducted mosquito collection in cities located along the BR-156 road of the Amapá State to assess the genetic structure and the spatial distribution of kdr mutations in Ae. aegypti along a South-North transect. The study was carried out to better understand local adaptation of dengue vector populations in the Amapá State and guide national authority in the selection of the most judicious insecticides to use for vector control.

Results

Kdr genotyping

We genotyped 272 Ae. aegypti mosquitoes for the three kdr SNPs V410L, V1016I and F1534C. In total, we found eight genotypes (Fig. 1): the six genotypes expected by the combination of wild-type NaVS (VVF), kdr NaVR1 (VVC) and kdr NaVR2 (LIC) alleles, and two additional. The genotype R2X1 (LL + VI + CC), composed of kdr NaVR2 and the herein called kdr NaVX1 (LVC) alleles, was found in OIA and CAL. We could not accurately determine the allelic composition of the genotype Y (LL + VI + FC), because it may be composed of LIC/LVF or LVC/LIF. This genotype was exclusively found in OIA. We named the possible alleles LFV and LIF as kdr NaVX2 and kdr NaVX3, respectively (Supplementary Tables S1, S2 and Fig. 1).

Figure 1
figure 1

Kdr genotypes in Aedes aegypti from the Amapá State, Brazil. (a) Scheme of the voltage-gated sodium channel (NaV), indicating the four domains (D1–D4), each with six transmembrane segments (S1–S6); (b) The genotypes observed in each of the three SNP sites in IS6, IIS6 and IIIS6 NaV segments. The wild-type and the kdr genotypes are in blue and red, respectively; (c) The total kdr genotypic frequencies, considering all samples (272 mosquitos genotyped for V410L, V1016I and F1534C) from the Amapá State; (d) Possible alleles that constitute the genotypes found here. The "Y" genotype is obtained by combining R2/X2 or X1/X3 alleles.

Full size image

In general, the lowest genotype frequency found in the Ae. aegypti populations was the wild-type homozygous SS (VV + VV + FF), only observed in MAC (2.3%). The most frequent genotype was R1R1 (VV + VV + CC), ranging from 2.2% in OIA to 65.1% in MAC. All populations showed high frequencies of other “resistant genotypes” (i.e. R1R1, R1R2, and R2R2), whereas their sum ranged from 63% in OIA to 100% in PGR. MAC was the only population without the homozygous R2R2 genotype (R1R1 + R1R2 = 76.7%). The genotype R2X1 was observed in CAL (8.7%) and OIA (2.2%). The genotype Y (LL + VI + FC) was found only in OIA, although at a relatively low frequency (28.3%) (Supplementary Table S1 and Fig. 1). The populations showed opposite trends regarding the kdr allelic frequencies according to their location along the South-North transect. For example, MAC presented the higher kdr NaVR1 (77.9%) and the lower kdr NaVR2 (9.3%) frequencies, while OIA showed the lower kdr NaVR1 (15.2%) and the higher kdr NaVR2 (52.2%) frequencies (Supplementary Table S2). The genotypic frequencies of each population are presented in Supplementary Figure S1.

Kdr haplotype sequences

We amplified and sequenced the corresponding IS6, IIS6 and IIIS6 NaV fragments of some homozygous R1R1 (4 MAC and 1 OIA) and R2R2 (4 OIA) mosquitoes. The IS6, IIS6 and IIIS6 sequences of R1R1 samples were similar to previously published sequences (GenBank accession numbers: LC557528, MN602762, MN602780, respectively). The R2R2 sequences of samples from OIA, IS6, IIS6, and IIIS6 fragments were also similar to known sequences (GenBank accession numbers: KY747530, MN602754, MN602780, respectively).

Microsatellite analyses

We genotyped 12 microsatellite loci in a total of 288 Ae. aegypti divided into six populations from the Amapá State. We observed a total of 60 alleles, varying from two (AC4 and B2) to 17 (Ag2) (Supplementary Table S3). Some markers were not under HWE, even after Bonferroni corrections, in some populations: AG2 (CAL, OAI), A9 (TTZ, CAL), A1 (FGO), AC1 (FGO, TTZ), AG5 (FGO, CAL) and AG1 (FGO) (Supplementary Table S4). In this case, the populations showed a Fis > 0 (Table 1 and Supplementary Table S4), indicating a heterozygous deficit. Although the LD test showed 33 tests significant out of the 396 evaluated combinations (8.3%), none loci pairs were consistently correlated in all populations after Bonferroni correction (Supplementary Table S5).

Table 1 Genetic diversity of six Aedes aegypti populations from Amapá State, Brazil, based on the analysis of 12 microsatellites.
Full size table

Genetic diversity and differentiation

Table 1 shows the genetic diversity in the Ae. aegypti populations from the Amapá State. The average allelic diversity varied from 3.58 (MAC) to 3.91 (TTG), the average number of effective alleles (Ne) from 2.33 (FGO) to 2.62 (OIA), the number of private alleles per population (Np) from 0 (CAL and PGR) to 0.25 (MAC), and the number of allelic richness (R) from 3.56 (CAL) to 3.88 (TTZ). The genetic diversity of a mosquito population usually is positively related to the expected heterozygosity (He) and observed heterozygosity (Ho) values. Here the He varied from 0.510 to 0.567 and the Ho varied from 0.528 to 0.576 (Table 1).

Regarding the genetic differentiation as measured by the Fst, the values of the pairs ranged from 0.004 (PGR-TTZ) to 0.084 (OIA-PGR) (Table 2 and Supplementary Table S6). Except for the pairs OIA-FGO (0.818) and OIA-PGR (0.884), the number of migrants (Nm) values were greater than one, indicating a certain degree of gene flow between the Ae. aegypti populations along the BR-156 road. The pairs associated with OIA showed the highest Fst and, conversely, the lowest Nm values, which can be justified by the fact that Oiapoque is the city located in the extreme north of the BR-156 road. Interestingly, OIA-MAC showed the highest Nm value among all pairs with OIA, although Macapá and Oiapoque are the most distant cities. The AMOVA showed Fis and Fit values of 0.92 and 1.57 (both p < 0.001), respectively (Supplementary Table S7), indicating a moderate differentiation within the populations. At least 55% of the genetic distance among the populations may be attributed to isolation by distance (IBD), according to the Mantel test (0.549, p > 0.031).

Table 2 Genetic differentiation (Fst) and number of migrant (Nm) indexes of pairwise comparison of Aedes aegypti from Amapá State, Brazil.
Full size table

Genetic structure

The Bayesian analysis conducted on the mosquito populations from Amapá suggests that Ae. aegypti is most likely divided into two genetic clusters (K = 2) (Supplementary Figure S2). We designed the structure plot considering K = 2 and K = 3 (Fig. 2a). In both scenarios, TTZ, FGO, PGR and MAC are the most homogeneous populations. With K = 2, OIA and CAL seem more related; with K = 3, OIA stands out as a more isolated population. The pie charts with the cluster frequencies in each population over the map (Fig. 2b) facilitate observing the populations’ genetic structure along the BR-156 road. For example, with K = 2, we observed that the most homogeneous populations were OIA and FGO, with 95.7% of the genetic diversity assigned back to cluster 2 in OIA while 95.5% to cluster 1 in FGO. In addition, the DAPC analysis plot evidenced three groups, where each OIA and CAL formed isolated groups, and TTZ, FGO, PGR and MAC were all mixed in a third group (Fig. 2c). These genetic structure analyses complemented the genetic differentiation indexes, suggesting that OIA and CAL are more structured than the other populations. This trend can be explained by the likely higher gene flow among the mosquitoes from TTZ, FGO and PGR and the capital MAC compared to OIA, which is a distant and isolated city located in the border area with French Guiana.

Figure 2
figure 2

Genetic structure of Aedes aegypti populations from Amapá State, Brazil, based on 12 microsatellite markers. (a) Structure plot: Bayesian clustering analysis where each bar represents the inferred ancestry of each individual with two (K = 2) or three (K = 3) genetic clusters; (b) Pie charts of the global inferred ancestry value of each population with K = 2 and K = 3; (c) Discriminating analyses of principal components (DAPC) plot, with the dots and colors representing individuals and groups, respectively. Eigenvalues indicate the number of principal components that best explain the differences between individuals (4 components were indicated for the 6 populations).

Full size image

Kdr and genetic structure

Considering the genetic structure based on the SSR analyses into two clusters (K = 2), the populations from clusters 1 and 2 presented the kdr alleles NaVR1 and NaVR2, whilst the additional kdr alleles were present only in the populations from cluster 2 (Fig. 3).

Figure 3
figure 3

Kdr allelic frequencies and genetic clustering of Aedes aegypti populations from Amapá State, Brazil. Kdr allelic frequencies are indicated in pie charts for each respective locality. The localities are shaded according to their genetic clustering (K = 2, see Fig. 2). OIA: Oiapoque, CAL: Calçoene, TTZ: Tartarugalzinho, FGO: Ferreira Gomes, PGR: Porto Grande, MAC: Macapá. S: Wild-type NaVS (VVF), R1: kdr NaVR1 (VVC), R2: kdr NaVR2 (LIC), X1: kdr NaVX1 (LVC), Others: alleles possibly composing the Y genotype: R2X2 (LIC/LVF) or X1X3 (LVC/LIF), see Fig. 1b.

Full size image

Discussion

In this study, we investigated the frequency of kdr genotypes and the genetic structure of Ae. aegypti populations along a South-North transect of the Amapá State, starting from the capital city Macapá to Oiapoque, located on the border between Brazil and French Guiana. The Amapá State is particularly isolated from the other States and covered by > 90% by the Amazon Forest. Our study showed that the dengue vector Ae. aegypti is divided into two well-defined genetic clusters in this State, and the kdr allelic composition roughly followed this clustering. We also found evidence of new kdr genotype arrangements in at least two of the six mosquito populations analyzed. Better understanding of the genetic structure of Ae. aegypti populations is relevant not only to better understand the possible dispersion of resistance genes in this region but also to guide decision-making for vector control and insecticide resistance management.

Historically in Brazil, the organophosphate malathion replaced pyrethroid insecticides around 2006–200912. However, malathion was never utilized in the Amapá State (operational constraints), and until recently, only pyrethroids were employed. The first cases of chikungunya in Brazil were recorded in 2014 in the cities of Oiapoque (Amapá State) and Feira de Santana (Bahia State), from two distinct events25. As a result, the use of pyrethroid insecticide (deltamethrin) was intensified in Oiapoque, which likely contributed to the increase in the selection pressure on resistance genes. Indeed, Ae. aegypti populations collected from Amapá State in 2014/2015 showed high resistance to deltamethrin with estimated RR50 (resistance ratio) of 46.4 and 143.9 in Macapá and Oiapoque, respectively26. Recently, Cielo has been used in ULV applications since 2019, and Spinosad was introduced in Amapá State in 2021 (personal communication—Health Secretariat of Amapá State). In French Guiana, deltamethrin is the only adulticide authorized in vector control campaigns since 201128.

Here, we observed that all Ae. aegypti populations from Amapá presented high frequencies of kdr alleles, considering the three important SNPs in the voltage-gated sodium channel gene (NaV): V410L, V1016I and F1534C. Interestingly, we found two additional kdr genotypes composition or arrangement, LL + VI + CC in Calçoene and Oiapoque, and LL + VI + FC in Oiapoque. These distinct genotypes were probably present in previous surveys; however, they were not detected since the V410L had not been investigated in mosquitoes from Amapá16. The LL + VI + CC (R2X1) was previously observed in populations from Peru (Palomino, 2022—unpublished data), Colombia29, Mexico30, and USA31. The LL + VI + FC (Y) was also previously observed in populations from the countries mentioned above, and in distinct Brazilian regions, such as Natal, Salvador, Vitória and Brasília16. Regardless of the location, their frequencies were generally low: varying from 1.1% in the USA to 5.8% in Colombia (R2X1), and from 0.2% in Peru to 2.5% in the USA (Y). In Amapá the frequencies were 28.3% (Y) and 2.2% (R2X1) in Oiapoque, and 8.7% (R2X1) in Calçoene. Complementary studies conducted by our research team suggest that these genotypes may result from gene duplications, hence maintaining distinct copies of the NaV gene in the same chromosome32,18,16. Their relationship with pyrethroid resistance requires further investigation.

We also showed that the kdr NaVR2 allelic frequency remained very high in Oiapoque (67% in 2014/201526, 68.5% in 201816 and 52.2% in 2020) and it increased significantly in Calçoene (from 7.8% in 201816 to 35.9% in 2020). While the kdr NaVR2 was absent in Macapá in 2014/201526, it has been found at low frequency (5.6%) in 201816, and at 9.3% in 2020, as shown in this study. The homozygous R2R2 is still absent from Macapá but it has been reported in Calçoene for the first time. In the localities of Ferreira Gomes, Porto Grande and Tartarugalzinho kdr NaVR2 varied from 42.4 to 48.9% and the R2R2 genotype ranged from 21.7 to 31.1%. It is worth noting that all localities along the BR-156 road had higher kdr NaVR2 frequency than Macapá, which is the State capital and therefore highly connected to these localities. Considering the genetic cost of the kdr NaVR2 under insecticide-free environment18, we assume that the insecticide pressure by pyrethroids might be lower in Macapá compared to other cities, hence prevailing the introduction and establishment of the kdr NaVR2 in Ae. aegypti in the State capital.

In addition, the decrease of the wild-type NaVS allele in all populations is of great concern. In Macapá, the frequency of the S allele was 15.7% in 2014/201526, 17.8% in 201816 and 12.8% in 2020. We also showed that the wild-type allele NaVS is now absent in Porto Grande and less than 13% in other populations tested. It is worth noting that the NaVS allele was not detected in any Ae. aegypti population from the Amazonian region in the previous nationwide surveillance (2018)26.

Regarding the kdr alleles nucleotide composition, we showed that the sequences of R1R1 and R2R2 (homozygous) samples from Macapá and Oiapoque were similar to the ones previously described17,33. Evidence shows that at least two haplotypes emerged independently with the 1534C kdr mutation and one haplotype with the kdr mutation 1016I17. The 1534C is present in the herein called kdr alleles NaVR1 (VVC) and NaVR2 (LIC) and the 1016I in the kdr NaVR2 in Ae. aegypti Brazilian populations16.

Overall, we suspect that the high resistance to pyrethroids in Oiapoque is associated with a higher frequency of kdr NaVR2 and the occurrence of rare kdr genotypes, as well as to the higher expression of detoxifying genes. Indeed, Ae. aegypti populations collected in the transborder city of Saint Georges de Oiapoque in French Guiana showed marked amplification of several CYP6s and CYP9Js playing a role in pyrethroid resistance34. We then cannot discard the possibility that these genes may be present in the Oiapoque population. Indeed Oiapoque (AP-Brazil) and Saint-Georges (French Guiana) are separated by a river which probably does not avoid the intense gene flow of Ae. aegypti between both sides, facilitated by the transportation of people and goods, which directly impacts the genetic structure of the mosquito24. Further work is needed to assess the detoxification pathway and genomic changes underlying the resistance mechanism of Ae. aegypti in the transborder area between Brazil and France that exhibits extremely high levels of resistance to all public health insecticides.

Finally, we analyzed the genetic structure and gene flow between Ae. aegypti from Amapá using 12 microsatellites markers. We showed that the Fst values among the six populations of Ae. aegypti were higher in the pairs with Oiapoque, which was isolated in a single genetic group in the Bayesian analysis (k = 2) and clearly isolated in the DAPC. Calçoene, the closest city to Oiapoque, was also represented in an isolated group in the DAPC. The isolation by distance (IBD), confirmed by the Mantel test, and the dense Amazon Forest remain the most likely explanation for the genetic structure and kdr frequency differences among the populations alongside the BR-156 Road. Interestingly, Oiapoque presented the lowest Fst and the highest Nm with Macapá, that are yet distant from 580 km. This can be explained by the fact that Macapá is the State capital, and the flow of people and goods between those cities is very intense, hence facilitating the passive transportation of Ae. aegypti.

Our findings align with other studies that support the hypothesis attributing the long-distance dispersal of Ae. aegypti to human activities, particularly vehicular traffic on highways35,36, as well as boat37,38 and airplane39 transportation. A prior study identified the spread of resistance and kdr mutations within Ae. aegypti populations along a major road in central Brazil. The movement of people and goods between the state capital and other cities was claimed as the factor facilitating the dispersal35. A mitochondrial DNA based study of Ae. aegypti populations from Argentina suggested that this mosquito passively commuted the long distance of approximately 60 km in a year36. In the Peruvian Amazon, the Ae. aegypti gene flow was higher between localities with heavy boat traffic compared to those with less intense movement37, similarly to our findings, however with the transportation in the road encased within a dense forest.

Based on the use of microsatellite markers, some studies concluded that the reinfestation of Ae. aegypti populations in Brazil after the eradication programme of the 1970s may have come from at least two distinct genetically differentiated groups, i.e. one from Venezuela to northern Brazil, and one from the Caribbean populations to the southeast of Brazil40,41. From this perspective, the Amapá State should have been invaded by Ae. aegypti from the two genetically distinct groups: in the north (Oaipoque and Calçoene) by the ‘Caribbean group’ through French Guiana, and in the south (Macapá, Ferreira Gomes, Porto Grande and Tartarugalzinho) by the ‘Venezuelan group’ through other Amazonian Brazilian locality, probably Pará State.

In this study, we used two types of molecular genotyping, kdr and microsatellite markers, to address distinct questions. The kdr SNPs were genotyped to obtain their genotypic frequencies in the populations. As kdr is under strong selection pressure, it is not an accurate marker to infer the population genetic structure, which was achieved by neutral microsatellite genotyping. By combining these two analyses, we observed that kdr alleles are spreading, likely by the Ae. aegypti ability to disperse itself and its eggs passively, however, without disrupting the overall genetic structure of the isolated populations.

Microsatellites and kdr genotyping analyses evidenced genetic differences among Ae. aegypti populations relatively isolated by a dense forest, connected by an only principal road. We found that rare kdr genotypes are increasing in frequency in some regions of Amapá State and this relationship with pyrethroid resistance deserves further investigation. More additional work is also needed to better understand the environmental and landscape determinants involved in the evolution and spatial distribution of insecticide resistance including kdr mutations in the dengue vector Aedes aegypti in Amapá.

Methods

Ae. aegypti collections

The field team of Laboratório de Entomologia Médica do Instituto de Pesquisas Científicas e Tecnológicas do Estado do Amapá (IEPA) installed 50 eggtraps in each of the six Amapá cities: Oiapoque (OIA) (03°49′53'' N, 51°50′07'' W), Calçoene (CAL) (02°29′53'' N, 50°56′59'' W), Tartarugalzinho (TTZ) (01°30′21” N, 50°54′41'' W), Ferreira Gomes (FGO) (00º51′14 “ N, 51º11′39″ W), Porto Grande (PGR) (00°42′16'' N, 51°24′35'' W) and Macapá (MAC) (00°02′04″ N, 51°03′60″ W) (Fig. 4), following the methodology developed by the Ministry of Health13. Collections were made in August (OIA, CAL, TTZ, FGO, PGR) and October/2020 (MAC). The eggtraps in CAL, FGO, PGR and TTZ were randomly distributed in the whole city. In MAC and OIA, the eggtraps were preferentially placed on the sidelines of BR-156 road. The eggs were stimulated to hatch in IEPA laboratory, and the resulting adults (F0) were packed in silica gel and shipped to Laboratório de Biologia, Controle e Vigilância de Insetos Vetores (LBCVIV) at Instituto Oswaldo Cruz (IOC/Fiocruz) for further analysis.

Figure 4
figure 4

Map showing the location of Amapá State, Brazil, and French Guiana, France. The delimitations of Amapá cities are also shown including the cities where Aedes aegypti were collected (in colour). The BR-156 road is indicated by a black line.

Full size image

Kdr genotyping

We extracted DNA from male mosquitoes from the F0 generation (from egg collection), individually titrated with TNES buffer, as described elsewhere42. The DNA of each sample was quantified in a Nanodrop One C spectrophotometer (Thermo Scientific) and aliquoted at 20 ng/µL in ultra-pure water. About 45 individuals from each population were genotyped for the kdr SNPs V410L, V1016I and F1543C through real-time TaqMan qPCR approach, essentially as described elsewhere16 (See Supplementary Table S8 for primers and probes sequences). The genotyping callings of each SNP were resolved by the online software Genotype Analysis Module v3.9 (Thermo Fisher), setting the endpoint to CT 40. Allelic and genotypic frequency charts were performed on GraphPad Prism v9.2.0 (www.graphpad.com). We considered the three SNPs (410 + 1016 + 1534) to determine the genotype of each sample16,20.

Na V sequencing

We amplified and sequenced fragments corresponding to the IS6, IIS6, and IIIS6 NaV segments of elected samples to determine the respective kdr haplotypes circulating in Ae. aegypti populations from Amapá. We selected nine samples genotyped as homozygous to kdr NaVR1 or kdr NaVR2 alleles (see Results) from MAC and OIA. For the PCR amplification, we used the kit Phusion High-Fidelity PCR (New England, Biolabs), containing Phusion Buffer 1X, DMSO 3%, the respective primers pair30,43,44 (Supplementary Table S9), 0.5 µm each and ultra-pure water q.s. 25 µL. The thermal-cycle conditions were 98 °C/30″ in the first step, followed by 35 cycles at 98 °C/10″ for denaturation, 60 °C (IS6), or 57 °C (IIS6), or 61 °C (IIS6)/15″ for primers annealing, and 72 °C/30″ for the Polymerase amplification, followed by a final extension step at 72 °C/7’. The amplicons were purified with the Qiaquick PCR Purification kit (Qiagen), according to the conditions provided by the manufacturer, and subjected to the sequencing reaction with the Bigdye Terminator V3.1 kit (Invitrogen, Thermo Fisher) using 1 µL of the purified amplicon and 1 µM of one of the respective primers. The sequencing reaction products of both strands of each sample were sent to the Fiocruz DNA Sequencing Facility (an ABI 3730 equipment). The sequences were analyzed using Geneious v9.1.845.

Microsatellites genotyping

The same DNA samples used for kdr genotyping were also genotyped for 12 well-characterized and polymorphic microsatellite loci that are widely used in studies of Ae. aegypti populations from the American continent: AC1, AC2, AC4, AC5, AG1, AG2, AG5, CT2, A1, A9, B2 and B3 (Supplementary Table S10)46,47. We performed the reactions with the Type-It PCR kit (Qiagen), according to a protocol standardized by Brown et al.46, using multiplex primers, originally proposed by Schuelke48, using an M13 tail at the primers 5' end and marked with FAM or HEX fluorescence (Supplementary Table S10). For each reaction, we used the Type-It Multiplex PCR Master Mix (Qiagen) 1x, each forward primer at 0.025 μM, each reverse primer at 0.25 μM, each probe at 0.5 μM, 1 µL of DNA (20 ng) and ultra-pure water q.s. 10 µl. The thermocycling conditions were 94 °C/10’ followed by 35 cycles of 94 °C/30″, 54 °C/30″ and 72 °C/30″, followed by the final step of 72 °C/5’. The product of each PCR was diluted 1:10, and 1 µL of the product was used for genotyping. Each amplicon received 0.5 μM the dye size standard GeneScan 500 LIZ (Applied Biosystems) and was submitted to the Genotyping/ Fragment Analysis Facility at Fiocruz (equipment 3130xl Genetic Analyser, Applied Biosystems).

Data analysis

We used the software Geneious 9.1.845 with the plug-in developed for ABI fragment analyses (Geneious Microsatellite Plugin) to obtain the genotype callings of all microsatellite loci, which were exported in a .csv file. We tested the loci for Hardy–Weinberg equilibrium (HWE) and linkage disequilibrium in the Genepop v4.7.5 software, with significance levels adjusted by Bonferroni correction49,50. The presence of null alleles in each locus was verified with Microchecker Software v2.2.351. The genetic diversity parameters: average number of different alleles (Na), number of effective alleles (Ne), private allelic richness (Np), expected heterozygosity (He), observed heterozygosity (Ho), endogamy coefficients (Fis) and the number of migrants (Nm) for each population were estimated using the software Genetic Analysis in Excel (Genalex) v.6.50352. This same software was used for analysis of molecular variance (AMOVA). The allelic richness (R) was calculated with the HP-Rare v1.1 software53. We estimated the magnitude of genetic differentiation with paired Fst values using Arlequin v3.5.254 and Freena55, respectively, both with 10,000 permutations. To test the assumption of isolation by distance (IBD), we run the Mantel test, with Arlequin v3.5.2 with 10,000 permutations54, correlating genetic and geographic data (geographical distance per km × Fst).

We also used the Structure v2.3.4 software56 to assess the population's genetic structuring based on the number of genetic clusters (K). The best K value was obtained based on ten independent runs with 500,000 Monte Carlo chains (MCMC) iterations, excluding 20% of the initial chains (burn-in). The output data was analyzed in the Structure Harvester 2.3 software57 to determine the best number of genetic clusters (K) based on an analysis of the ΔK chart58. Finally, we submitted the output files to Clumpp v1.1.259 and Distruct V1.160 software to plot the genetic structure of the evaluated populations. A multivariate statistical analysis (discriminant analysis of principal components—DAPC) was performed with the Adegenet package61,62 in the R platform63.