Background

Malaria, caused by Plasmodium parasites and transmitted by Anopheles mosquitoes, is an important public health problem in the Americas, where ~ 505,600 cases were diagnosed in 2023 alone1. Approximately 1 in 5 of those malaria cases occurred in Brazil1, where the number of cases rose by 3% between 2015 and 2020 (137,982 vs. 142,112)2. The transmission of Plasmodium species in Brazil predominantly occurs in the Amazon region2,3. Plasmodium vivax parasites cause the majority (~ 83%) of malaria cases, followed by P. falciparum (~ 17%), and then other Plasmodium species (0.1%). Over the last 7 years, the proportion of cases caused by P. falciparum, the species that causes most severe disease, has increased by 7%2.

The primary malaria vector in Brazil, Anopheles darlingi (also called Nyssorhynchus darlingi), is highly susceptible to Plasmodium infection and can maintain malaria transmission even when parasites are at low densities4,5,6,7,8,9. This mosquito species is highly adaptable to recently anthropized environments and exhibits both exophagic and endophagic behavior. An. darlingi primarily feeds on humans and can occupy ecological niches left empty by other Anopheles species (spp.)8,10. Since 2017, deforestation has increased in the Brazilian Amazon, with previous studies showing these newly deforested areas or “frontiers” have higher An. darlingi abundance and increased malaria transmission11. The adaptability of An. darlingi mosquitoes is believed to be leading to year-round malaria transmission. Previously, forest populations of An. darlingi would peak during, and towards the end of the rainy season due to the increased availability of larval habitats in flooded areas near rivers. However, environmental changes produced by humans have created permanently available larval habitats, thereby supporting perennial malaria transmission12,13. These challenges represent a clear risk for Brazil to accomplish its National Elimination plan and achieve the World Health Organization (WHO) goal of reducing malaria cases by 90% within the next 7 years14.

Typical malaria control practice in Brazil uses insecticides, especially indoor residual spraying (IRS) or long lasting insecticide treated nets (LLINs)15,16. Pyrethroids such as cyperpermethrin and deltamethrin, have been the insecticide class applied in recent years using both methodologies, but organophosphates (malathion) and carbamates are used on Aedes aegypti as part of arbovirus control. Since the reintroduction of dengue in Brazil and the occurrence of outbreaks of other arboviruses (e.g., Zika and chikungunya), Ae. aegypti has been part of a national insecticide resistance screening programme to optimise elimination strategies, but no such programme exists for Anopheles17,18. Resistance to pyrethroids has been reported across Anopheles species globally, and for An. darlingi in countries surrounding Brazil, including Bolivia, Peru, Colombia, and French Guiana19,20,21. Resistance to carbamates has also been reported in Peru and Bolivia, and organochloride resistance has been observed in Peru and Colombia22,23. No resistance has been reported for organophosphates in Brazil, and no data is available for pyrroles or neonicotinoids. The lack of reporting on insecticide resistance in An. darlingi is a cause for concern, due to the high levels of associated Plasmodium spp. transmission by this vector.

The main insecticide resistance mechanisms observed by mosquito species are target site, metabolic and cuticular, and behavioural avoidance24. Target site resistance is mediated by mutations in insecticide target genes, such as the acetylcholinesterase-1 (ace-1), γ-aminobutyric acid (GABA) receptor (rdl), and voltage-gated sodium channel (vgsc). Particularly well-studied are the knockdown resistance (kdr) mutations associated with Dichlorodiphenyltrichloroethane (DDT) and pyrethroid-based insecticides, including the L1014 mutation in vgsc25,26. Point mutations in the glutathione-s-transferase epsilon (GSTe2) gene, which encodes an insecticide metabolising enzyme, are also associated with these insecticides, and the L119F and I114T mutations lead to resistance to permethrin in An. funestus27,28. Mutations and duplications in the ace-1 gene result in organophosphate and carbamate resistance, and amino acid alterations in rdl have been associated with resistance to organochlorines, particularly dieldrin29,30,31. Molecular surveillance studies of insecticide resistance mutations in An. darlingi are scarce, with only a few investigations exploring target regions in vgsc and ace-1 genes32,33. None of the single nucleotide polymorphisms (SNPs) previously found to be associated with insecticide resistance in other Anopheles species have been observed in An. darlingi, including in pyrethroid resistant populations33. It is possible that genetic variants in other regions within these genes or in additional candidate genes, may contribute to the observed outcomes.

Whole genome sequencing (WGS) has been applied to many vectors to better understand their genomic landscapes and identify candidate genes to unravel mechanisms of insecticide resistance. However, Anopheles genomes are large (~ 300 Mbp), and WGS is an expensive method that requires high quantities of DNA to gain suitable genomic insights, meaning it is inappropriate as a high-throughput surveillance method. The application of next-generation sequencing to targeted PCR amplicons, in tandem with dual-index barcoding, has been successfully used in other Anopheles spp., Aedes spp., and Plasmodium spp., as a high-throughput and low-cost screening method for insecticide or drug resistance mutations in target loci34,35,36,37. Targeting several candidate genes in many samples permits the tracking of emerging resistance and spread of known mutations in the population. This approach also allows for an analysis of genotype-phenotype associations to identify novel mutations linked to insecticide phenotypic assays.

Here we have designed an amplicon-sequencing (“amp-seq”) assay, consisting of a panel of 11 amplicons (each ~ 500 bp) covering multiple regions across 4 genes (vgsc, ace-1, rdl, and GSTe2) commonly associated with insecticide resistance, and a further 2 genes (Its2 and cox1) used for species identification and phylogenetic analysis. Population genetics studies of An. darlingi have focused on cox1 and its2, therefore we include both markers to compare the genetic structure of the An. darlingi populations examined here with existing available data from other countries. This approach allows for a broader understanding of the species’ genetic diversity across regions.

This multitarget assay was used to screen 200 An. darlingi mosquitoes collected in Brazil, revealing several new mutations. Our assay represents a cost-effective method to confirm mosquito species and conduct insecticide resistance surveillance, with the potential to inform control strategies for an understudied vector that is responsible for high levels of malaria transmission in South America.

Methods

Sample collection, species identification and DNA extraction

An. darlingi field mosquitoes were mostly collected from localities in Rondônia state (n = 171; Candeias do Jamari n = 116 (GPS: 8°39’41.0” S 63°01’58.8” W, 8°41’00.5” S 63°11’08.8” W), Porto Velho (GPS: 8 47’08” S, 63 55’04” W) n = 55) but we also include four samples from the Amazonas state (n = 4; Manaus n = 1 (GPS: − 2°58’52.7"S 60°03’03.4"W), Manacapuru (GPS: 3°16’34.3"S 60°37’36.9"W) n = 3)38,39. The collections were performed during studies of vector density in malaria endemic areas of Rondônia 2018–201938,39. Mosquitoes from the An. darlingi colony of Porto Velho/Rondônia (colony generations : F2-F4, F9-F11, F21, F33-F35, F39, F40-F42) were also included in the study (n = 25)40. In total, 200 mosquitoes were screened. The field samples were selected based on whether the localities had used insecticides (Porto Velho/ Manaus) or had little/no recent insecticide usage (Candeias do Jamari, Manacapuru). During the sample collection period in Porto Velho, alpha-cypermethrin was applied for both IRS and LLINs, while cypermethrin was used for thermal fogging. In Manaus, etofenprox was employed for IRS, and alpha-cypermethrin was applied to LLINs. For thermal fogging, both cypermethrin and lambda-cyhalothrin were used. The specimens were initially identified by stereoscopic microscopy, using the established dichotomous keys41. Genomic DNA was extracted from whole mosquitoes using the Qiagen DNeasy ® Tissue and Blood kit (Qiagen, Hilden, Germany), according to the manufacturer’s instructions. Briefly, each mosquito was grounded with 30 µL of lysis buffer and the DNA extract was resuspended in 50 µL elution buffer.

Primer design

Amplicon primers were designed using Primer3 software, against sequences downloaded from VectorBase42. The primers aimed to amplify an approximate 500 bp region containing SNPs previously described as associated with insecticide resistance in Anopheles or Aedes mosquito’s species. This resulted in a panel of 9 primers targeting 4 genes: (i) vgsc (4 amplicons, targeting four domains); (ii) rdl (2 amplicons, targeting 3 SNPs); (iii) ace-1 (2 amplicons, targeting two SNPs, including G280S32); and (iv) GSTe2 (1 amplicon, targeting two SNPs). When possible, these primers were designed to bind to exonic regions. Two other amplicons were designed to target genes commonly used for species identification and phylogenetic investigation: the ribosomally encoded gene internal transcribed spacer 2 (its2); and cytochrome c oxidase I (cox1), a locus found in the mitochondria. This resulted in a final panel of 11 amplicons, covering 6 genes (Table 1). Each primer sequence was concatenated with a unique 5’ barcode (8 bp) designed in house34. Samples were assigned a unique forward and reverse barcode combination used for the generation of amplicons to enable sample pooling before sequencing. To identify amplicons suitable for multiplexing the ThermoFisher Scientific Multiple Primer Analyser was used with sensitivity for dimer detection set to one.

Table 1 Primers and targets of An. darlingi amp-seq panel.
Full size table

Amplicon generation

Multiplex Polymerase chain reactions (PCR) were performed using NEB Q5 hot start polymerase (New England BioLabs, UK) with a total volume of 25µl per reactions. Sample volume of 1µL (~ 2ng/µL) was used, with an average final primer concentration of 0.5µm in each PCR. The amplification was conducted as follows: hot-start polymerase activation for 3 minutes at 95 °C, followed by 30 cycles of 95 °C for 10 seconds, 58 °C for 30 seconds and 72 °C for 45 seconds, followed by a final elongation step of 72 °C for 5 minutes. Post-multiplex PCR reaction, amplicons were visualised on a 1% agarose gel to confirm amplification, alongside band size and intensity. Amplicons for each sample contained a unique 5’ barcode (8 bp) designed in-house34, making it possible to pool all amplicons across samples. Pools were purified using Roche Kapa beads following manufacturer’s instructions. A bead to sample ratio of 0.7:1 was used to remove excess primers and PCR reagents. The Qubit 2.0 fluorimeter HS DNA kit was used to quantify the pool concentration. Illumina barcodes and adaptors were further added to the sample pools as part of the Illumina-based Amplicon-EZ service (Genewiz, UK). Pools contained a maximum of 200 amplicons to maximise coverage. Each indexed pool was sequenced using a 2 × 250 bp (paired-end) configuration on an Illumina MiSeq. A minimum of 50,000 paired-end reads were attained per pool, which equates to at least 450 reads per amplicon in a pool of 110 amplicons, at a cost of < US$0.5 per amplicon. For variant confirmation with Sanger sequencing, amplicons were generated using the same primers but were amplified in simplex reactions, before being sequenced by Genewiz, UK.

Amplicon analysis

Raw fastq files were de-multiplexed using the unique barcode combination assigned to each pool and each sample, using an in-house python script (https://github.com/LSHTMPathogenSeqLab/amplicon-seq/blob/main/scripts/amplicon_script.py). The resulting sample fastq files were then analyfirst trimmed using Trimmomatic software, then mapped to the reference sequence (idAnoDarlMG_H_01, from NCBI: https://www.ncbi.nlm.nih.gov/datasets/genome/GCF_943734745.1/) using the bwa-mem package, and reads are then filtered using Samclip software43,44,45. GATK HaplotypeCaller (v4.1.4.1, default parameters) and Freebayes (v1.3.5, --haplotype-length − 1) software were used to call variants46,47. The SNPs and small insertions/deletions (INDELs) identified were then filtered using bcftools48. To pass quality control checks, a minimum depth of 30 reads, phred score of > 30 per base, and a minimum allele depth of 10 was required. Variants had to be present in > 1 sample, and across > 1 of the sample pools sequenced. The SnpEff tool was applied to annotate variants using a database built from the idAnoDarlMG_H_01 reference genome49. Variants were then genotyped based on the proportion of alternative allele to total depth coverage, called as homozygous reference (< 20% alternate allele reads), heterozygous (20–80% alternate allele), or homozygous alternate (> 80% alternate allele reads)34.

Phylogenetic analysis

For the Its2 and cox1 amplicons, SNP calls with > 50-fold read depth were converted to fasta files using an in-house pipeline (https://github.com/LSHTMPathogenSeqLab/fastq2matrix). Only sequences with SNP calls that reached this depth were included in phylogenetic analysis. For each gene, sequence data was aligned using the MAFFT tool. Sequences from the NCBI from other countries were included in the resulting alignments50. For cox-1, 67 sequences were added from Brazil, Honduras, Belize, Colombia, Panama, Ecuador, and Peru. For its2, an additional 26 sequences were aligned from Brazil, Colombia, Belize, and Bolivia. For tree generation, sequences from both genes were concatenated, and the resulting alignments were viewed and trimmed using Aliview51. Phylogenetic trees were constructed using RAxML software52. The trees were built using a maximum-likelihood method, with the GTRGAMMA option. This approach assumes a GTR model of nucleotide substitution, and a gamma model of rate heterogeneity. A bootstrap value of 1000 was used for tree construction, and the resulting tree was visualised using iTOL software53.

Haplotype networks and maps

To construct the haplotype network, fasta sequences for cox-1 and its2 were aligned for all samples, and then the packages Pegas and ggplot2 R54,55 were applied. Pegas was used to calculate nucleotide diversity, haplotype diversity, Tajima’s D statistic, fixation indel (Fst), heterozygosity and linkage disequilibrium55.

Results

Detection of novel SNPs in genes associated with insecticide resistance

A total of 200 An. darlingi samples were sequenced, with the resulting average amplicon coverage ranging from 171- to 5621-fold (Table 2). From the alignments, 246 SNPs and 20 INDELs passed all quality control measures, the majority of which were either synonymous (37.8%) or intronic variants (45.9%). Ten SNPs and one INDEL were annotated as non-synonymous and resulted in an amino acid change (Table 3 & Supplementary Table S1). Missense SNPs were only found in ace-1 and GSTe2 genes and all were present in at least two samples and in two or more populations (Table 3). These SNPs have not been previously reported. In the ace-1 gene, 111 SNPs were found across the two amplicons, including five missense SNPs (V243I, N294H, S673N, S674N, and S674T), with S674N occurring at the highest frequency (19.6% heterozygous and 2.7% homozygous samples). Three of these five mutations (V243I, N294H and S674T) occurred only in field populations, while S673N and S674N appeared in field populations and all colony samples (Table 4). A further five non-synonymous SNPs were found in the GSTe2 gene. Three of the missense SNPs (I114V, T166I, and T179I) occurred at frequencies below 23%, and two others (D128E and T205A) appear to be at or approaching fixation as no samples were genotyped as homozygous reference, and over 90% were homozygous alternate for both amino acid substitutions. Two of these amino acid alterations in the GSTe2 gene were found only in field populations (I114V and T166I), and the remaining three (D128E, T191I, and T205A) were also observed in colony samples.

Table 2 Average amplicon coverage, and number of genetic variants identified.
Full size table
Table 3 Locations and allelic frequencies of detected non-synonymous variants.
Full size table
Table 4 Genotype frequencies of non-synonymous SNPs in the four field collection sites.
Full size table

For vgsc, the analysis revealed a INDEL caused by a 3 bp deletion, resulting in an isoleucine deletion at position 422 in the first domain. This mutation has not previously been reported and occurred at a frequency of 25% across field and colony samples. Sanger sequencing confirmed these mutations, ruling out sequencing artifacts. No other non-synonymous variations were observed across the four amplicons investigated in the vgsc gene. No missense polymorphisms were detected in the gaba gene (rdl).

Genetic diversity of An. darlingi populations in Brazil

Sequences for mitochondrial cox-1 gene and ribosomal its2 were generated for genetic diversity analysis. In the its2 gene, six SNPs were identified. Originally a total of 91 SNPs were identified in the cox-1 amplicon, but upon further inspection, 71 of these SNPs were present in one sample (AnDar600), which was subsequently identified as an An. peryassui isolate (Blast score: 99.2% identity) and excluded from further analysis. Twenty SNPs were identified in the cox-1 gene, 19 of which appeared in the Candeias do Jamari population, 13 in colony samples, 10 in the Porto Velho populations, and six in the State of Amazonas samples. A fixation index analysis for each SNP revealed no significant population differentiation across these genes (Fst < 0.032).

Phylogenetic analysis with the its2 gene reflected the small number of SNPs shown in these populations, and very little differentiation was observed between Brazilian isolates and those from other countries (Fig. 1). The tree separated into two main clades, the first of which contains three publicly available samples from Brazil, Belize, and Bolivia. The second clade contained several subclades that included all sequences generated in this study (n = 198) along with the remaining publicly available Brazilian and Colombian samples (n = 26). No differentiation between the different Brazilian populations was observed, which was supported by low nucleotide diversity (π = 0.00536) (Table 5).

Fig. 1
figure 1

Maximum-likelihood tree constructed using its2 gene sequences generated in this study (total = 198, Br_Rondonia = 171, Br_Amazonas = 4, Br_Colony = 23), alongside other publicly available An. darlingi its2 sequences (n = 26, Brazil, Colombia, Belize, and Bolivia). The tree was built using the maximum-likelihood method assuming GTR model of nucleotide substitution, with the gamma model of heterogeneity rate.

Full size image
Table 5 Nucleotide and haplotype diversities of ITS2 per population.
Full size table

For the cox-1 gene, a total of 128 sequences from this study alongside 67 publicly available samples were analised. (Fig. 2). Using this gene, it was possible to see clusters separating most Colombian, Honduran and Brazilian samples. Although, sequences from other countries can be seen interspersed particularly within the Brazilian cox-1 sequences. There was both low genetic and nucleotide diversity (0.00697) in cox-1 (Table 6), similar to the its2 results.

Fig. 2
figure 2

Maximum-likelihood tree constructed using cox-1 gene sequences generated in this study (total = 129, Br_Rondonia = 106, Br_Amazonas = 3, Br_Colony = 20), alongside other publicly available An. darlingi cox-1 sequences (n = 67, Brazil, Honduras, Belize, Colombia, Panama, Ecuador, and Peru). This tree also has a group of Anopheles spp. including An. albimanus, An. arabiensis, An. coluzzi, An. dirus, An. funestus, and An. gambiae. The tree was built using the maximum-likelihood method assuming GTR model of nucleotide substitution, with the gamma model of heterogeneity rate.

Full size image
Table 6 Nucleotide and haplotype diversities of cox-1 per population.
Full size table

A phylogenetic based on concatenated its2 and cox-1 sequences revealed little differentiation between samples across the geographical regions (Supplementary figure S1). Both loci demonstrated high haplotype diversity (Tables 5 and 6), with 44 haplotypes identified for the cox-1 gene, and 31 for its2. Most of the haplotypes, 77.3% for cox-1 (n = 34), and 54.8% for Its2 (n = 17), were present in only one sample (singletons). The high number of singleton haplotypes reflects the high proportion of SNPs occurring at low frequency in the populations. Haplotype 34, representing the cox-1 gene, was the most frequent, and was present in samples from the colony, and states of Amazonas and Rondônia (Supplementary Figure S2). A higher number of its2 haplotypes (n = 14, 45.2%) were present in more than one sample, compared to cox1 (n = 10, 22.7%). Haplotypes 1 and 9 were the most frequent, present in both colony and state of Rondônia populations (Supplementary Figure S3). The samples from the state of Amazonas shared haplotypes with both colony and state of Rondônia samples. The networks revealed shared haplotypes for both genes across the three populations, and included several other samples available from other Brazilian states.

Discussion

The application of our amplicon sequencing panel to Brazilian field and colony An. darlingi samples has demonstrated its potential utility for species identification, and the discovery of SNPs in genes associated with insecticide resistance. Whilst no previously reported SNPs associated with insecticide resistance in other Anopheles species were found in this study, ten other non-synonymous SNPs were detected. Of the five SNPs found in the GSTe2 gene, all except one (D128E) are either in amino acid positions that are highly variable across Anopheles spp. or the mutation results in a change to an amino acid that is present in the reference of another Anopheles spp. The I114V substitution is in the same location as a previously reported amino acid alteration in An. gambiae56. In An. gambiae, the mutation results in a I114T substitution, which is hypothesised to cause resistance through the introduction of hydroxyl (-OH) group on the substituted threonine. The hydroxyl group decreases product affinity in the hydrophobic DDT binding site, thereby increasing metabolic turnover of the insecticide. In this instance the valine substitution does not introduce this same hydroxyl group, and valine is present as a reference amino acid in An. atroparvus. The D128E mutation occurs at a highly conserved site across Anopheles spp., with aspartic acid (D) present as the reference for all. The alteration to glutamic acid results in a similar amino acid structure with the addition of an extra carbon. This mutation appears to be near fixation as 92% of samples were genotyped as homozygous alternate, and the remaining 8% as heterozygous.

Pyrethroids are the predominant insecticide class being applied for malaria-focussed vector control in Brazil16. This implies that if target site mutations were to arise, they would predominantly occur in the vgsc gene. However, in Brazilian dengue control programmes, Aedes spp. are targeted with both pyrethroids and organophosphates, and this usage could impact Anopheles spp. While malaria and Aedes-transmitted arboviruses tend to occur in different ecological zones in Brazil—malaria is more prevalent in rural areas while arboviruses are more common in urban environments—there can still be some overlap in the intervention areas. This overlap is especially evident in peri-urban regions and towns where both vector species might be present and where rapid urbanization and migration could increase the risk of both diseases. This could explain the three non-synonymous SNPs found in the ace-1 gene, a target for organophosphates. The previous reported target-site mutations G119S and N485I (positions in Torpedo califonica, G305S and S642I) alter susceptibility to the organophosphate and carbamate classes of insecticide29. An. darlingi resistant to carbamates have been reported in Bolivia, close to the Brazilian border22. Of the five amino acid alterations detected in the ace-1 gene (V243I, N294H, S673N, and S674N/T), in the Anopheles darlingi from our study, none have been previously reported. The S673N and S674N/T ace-1 mutations occur approximately 49 residues away from one of the three catalytic sites (H440; H625 in An. darlingi) and were present in field and colony samples for which the resistance status is not known. The N294H alteration occurs 9 amino acids upstream of the G119S (G305S) mutation, and results in a change from an amino acid with a polar uncharged side chain (asparagine) to one with a charged side chain with an aromatic imidazole ring. For all five of these amino acid substitutions, the reference amino acid is conserved across other Anopheles species (results from sequence alignment of An. darlingi, An. albamanus, An. gambiae, An. funestus, An. stephensi, and An. minimus). An amino acid conserved across species suggests is could be important for protein function. Further studies, which include bioassays, are needed to confirm if these mutations result in organophosphate or carbamate resistance. It is not possible at present to exclude potential candidate SNPs on the basis of their presence in the colony samples, as their insecticide resistance profiles are still under evaluation.

The four study locations had varying insecticide usage, with two sites, Porto Velho (State of Rondônia), and Manaus (State of Amazonia), having a history of intensive use of pyrethroids and carbamates, particularly the application of cypermethrin related counpounds for residual spraying effors, with lambda-cyhalothrin used since 2012. In contrast, Candeias do Jamari (State of Rondônia) and Manacapuru (State of Amazonas), had little to no insecticide spraying. All missense SNPs identified in this study were found in locations with intensive insecticide spraying, and locations with little to no insecticide usage. Additional studies that combine phenotypic and molecular surveillance data are needed to understand the impact of the SNPs detected here on insecticide resistance in An. darlingi.

The lack of SNPs found in the rdl gene may reflect the reduced selective pressure exerted by dieldrin.

Mutations in gaba result in dieldrin resistance, an organochloride that has been prohibited from use for at least the past decade due to its adverse effects on human health. However, rdl SNPs have been documented in other Anopheles species, which also have no longer exposure to dieldrin. These mutations highlight the potential for persistence in populations even after the cessation of insecticide use, driven by environmental residues or cross-resistance with other insecticides. Continued monitoring of rdl SNPs across Anopheles species is essential to understand the dynamics of resistance and its potential implications for vector control strategies. In relation to the vgsc gene, four amplicons were investigated, but no kdr mutations were detected in the populations surveyed here. This observation was also reported in pyrethroid resistant populations of An. darlingi from Colombia33. The lack of kdr mutations suggests that SNPs in vgsc may play a reduced role in An. darlingi insecticide resistance compared to other Anopheles spp., such as An. gambiae. An INDEL resulting in isoleucine deletion at position 422 was detected in 25% of sample, across both field and colony sources. Further studies are needed to understand the involvement of this deletion in insecticide resistance.

The lack of known resistance associated SNPs in An. darlingi could be due to the gene flow between geographically close populations with varying levels of insecticide use. The mixing of possible resistant populations with insecticide-sensitive populations may minimise the frequency of resistant alleles, a dilution effect hypotesised by Vezenegho et al.19. Another hypothesis is that other genetic variants may play a role in resistance, or that different mechanisms, such as differential gene expression, could be involved in this species resistance status.

In relation to the genetic diversity across populations, there was little differentiation between An. darlingi from Brazil and other regions using its2 gene data. However, the phylogenetic tree constructed using cox-1 gene data revealed that samples generally grouped by country. It has been hypothesised that physical barriers like the Atlantic forest mountain range or the Amazon River, prevent the mixing of these populations and so they appear genetically distinct57,58,59. Data from the cox-1 gene revealed an outlying sample, which aligned with another Anopheleles species (An. peryassui).

Within the Brazilian population, low nucleotide diversity was seen in tandem with high haplotype diversity for both cox-1 and its2 genes, indicating many low frequency variants. This observation is consistent with previous studies conducted on cox-1 across Central and South American, and within Brazilian and Colombian An. darlingi populations13,57,60. The geographical proximity of the collection sites of these mosquitoes may also contribute to the low genetic diversity observed here. The inclusion of study sites that are more geographically distant or from other countries may give greater resolution to the population dynamics of this species.

Overall, our amp-seq panel provides a tool to investigate the genetic diversity of this understudied An. darlingi vector. It is a high-throughput, low-cost assay for species identification and the detection of novel SNPs in insecticide resistance associated genes. Further investigation is required to identify whether these SNPs contribute to insecticide resistance in An. darlingi. A clear limitation of this method is that it currently only includes target regions of known loci associated with insecticide resistance. However the panel is easily adaptable, enabling the inclusion of additional targets in these or other genes, including further metabolic markers. The panel can be used in tandem with phenotypic assays to identify SNPs that result in functional changes. Large-scale surveillance methods are urgently needed to inform malaria vector control methods in Brazil, particularly to assist initiatives to reduce malaria transmission. Our panel represents the first steps towards a high-throughput, multitarget molecular surveillance method for tracking known and identify potential new markers of resistance.