The implementation of long-lasting insecticidal bed nets has differential effects on the genetic structure of the African malaria vectors in the Anopheles gambiae complex in Dielmo, Senegal

Background Mosquitoes belonging to the Anopheles gambiae complex are the main vectors of malaria in sub-Saharan Africa. Among these, An. gambiae, Anopheles coluzzii and Anopheles arabiensis are the most efficient vectors and are largely distributed in sympatric locations. However, these species present ecological and behavioural differences that impact their vectorial capacity and complicate vector-control efforts, mainly based on long-lasting insecticidal bed nets (LLINs) and indoor residual spraying (IRS). In this study, the genetic structure of these three species in a Senegalese village (Dielmo) was investigated using microsatellite data in samples collected in 2006 before implementation of LLINs, in 2008, when they were introduced, and in 2010, 2 years after the use of LLINs. Results In this study 611 individuals were included, namely 136 An. coluzzii, 101 An. gambiae, 6 An. coluzzii/An. gambiae hybrids and 368 An. arabiensis. According to the species, the effect of the implementation of LLINs in Dielmo is differentiated. Populations of the sister species An. coluzzii and An. gambiae regularly experienced bottleneck events, but without significant inbreeding. The Fst values suggested in 2006 a breakdown of assortative mating resulting in hybrids, but the introduction of LLINs was followed by a decrease in the number of hybrids. This suggests a decrease in mating success of hybrids, ecological maladaptation, or a lesser probability of mating between species due to a decrease in An. coluzzii population size. By contrast, the introduction of LLINs has favoured the sibling species An. arabiensis. In this study, some spatial and temporal structuration between An. arabiensis populations were detected, especially in 2008, and the higher genetic diversity observed could result from a diversifying selection. Conclusions This work demonstrates the complexity of the malaria context and shows the need to study the genetic structure of Anopheles populations to evaluate the effectiveness of vector-control tools and successful management of malaria vector control. Electronic supplementary material The online version of this article (doi:10.1186/s12936-017-1992-8) contains supplementary material, which is available to authorized users.


Background
Over the last decade, impressive progress has been made in controlling malaria vectors in sub-Saharan Africa mainly using long-lasting insecticidal bed nets (LLINs) and indoor residual spraying (IRS) [1]. Both tools have remarkably contributed to the decrease of malaria prevalence since 2000 [2] in sub-Saharan Africa, where the burden of malaria is heaviest [3]. Three species of the Anopheles gambiae complex, namely Anopheles gambiae, Anopheles coluzzii and Anopheles arabiensis are the primary vectors of malaria in this region [4]. Indeed, these species coexist in sympatry throughout their range and are closely associated with human habitats [5]. The two former species, An. gambiae and An. coluzzii, were previously identified as the two An. gambiae molecular forms, S and M, respectively [6]. Evidence of hybridization or introgression between these species raised questions about their taxonomic status [7][8][9][10][11][12][13][14]. The recognized sibling species An. arabiensis was never divided into chromosomal forms, although it is widely distributed in the Afrotropical region [15]. Behavioural differences that impact their vectorial capacity as malaria vectors have been recorded between these sibling species. Indeed, An. gambiae and An. coluzzii are highly anthropophilic and endophilic [5].
Anopheles arabiensis displays similar patterns, but presents a competitive advantage of being more zoophilic and exophilic [16]. These phenotypic differences highlight genetic heterogeneities within the An. gambiae complex that could affect the species habitat preference and their adaptive responses against malaria control tools. Several studies have largely reported selective pressure on malaria vectors after the widespread use of vector-control tools using insecticide compounds. The most obvious example is the selection and the emergence of insecticide resistance genes in the main vectors in sub-Saharan Africa [17][18][19][20][21][22]. Furthermore, when the LLINs are used, a selective advantage has been reported in Anopheles species that exhibit a tendency to feed outdoors, earlier in the evening, during diurnal hours and on non-human hosts [23][24][25]. Indeed, a previous study carried out in Nigeria and based on cytogenetic analyses has reported the presence of chromosomal inversions correlated with the indoor or outdoor biting preferences in the main vectors of the An. gambiae complex [26]. Furthermore, in Burkina Faso, in 2011, an outdoor-resting subgroup of the An. gambiae complex, named "Goundry", genetically isolated from both the M (An. coluzzii) and S (An. gambiae) forms was found exclusively in peridomestic human habitats [27]. The authors suggest that this outdoor subgroup would be favoured by selective pressure of insecticide-treated bed nets and indoor residual spraying used in the area.
Some studies showed the persistence of An. arabiensis following the implementation of insecticide-treated bed nets and its potential involvement in residual malaria transmission [23,28]. These events could jeopardize the effectiveness of vector-control tools using insecticide compounds due to a non-uniform exposure of vectors to insecticide and worsen malaria transmission. Thus, in the context of scaling up malaria vector-control tools using insecticide-treated bed nets and indoor residual spraying, it is essential to know whether selective pressure affects the genetic structure of the main vectors of the An. gambiae complex relative to their habitat feeding place, a paramount factor in understanding the malaria epidemiology.
Concerning An. gambiae and An. coluzzii, previous investigations on the genetic differentiation of both species have reported divergent findings, even though the geographical scale, the pattern of genomic tools used and the genomic island analysed were not homologous [29][30][31][32][33][34][35][36]. Genetic differentiation between populations of the sibling species, An. arabiensis was reported according to geographical distance [37,38]. But, in Tanzania, high levels of differentiation were reported within An. arabiensis sympatric samples [39]. Senegalese populations of the genetic structure of the An. gambiae species complex were investigated only in two An. arabiensis populations collected in Dielmo and Barkedji villages (250 km apart) [40] between which low but significant levels of genetic differentiation were found.
As in other African malaria-endemic regions, Senegal has adopted universal coverage of insecticide-treated bed nets in its strategic national program to accelerate efforts toward malaria pre-elimination. Thus, several studies have reported on the malaria transmission dynamics and the monitoring of entomological parameters on Anopheles vectors [28,41,42]. However, no information was available on the impact of insecticide-treated bed nets on the genetic structure of the main vectors of the An. gambiae complex.
This study focuses on the impact of long-lasting insecticidal treated bed nets on the genetic structure of An. coluzzii and An. gambiae species and their sibling species An. arabiensis through a spatiotemporal survey covering 3 years of entomological surveys. The occurrence of putative gene flows or hybridization between An. coluzzii and An. gambiae is also considered. The implication of the control strategy is discussed.

Study area and mosquito sampling
The study was conducted in Dielmo village (13°45′N, 16°25′W), where a longitudinal survey has been carried out since 1990 to identify all malaria episodes and to investigate the relationship between malaria parasites, human hosts and vectors [43]. The village is located 280 km Southeast of Dakar and about 15 km North of the Gambian border. In Dielmo, the rainy season occurs mostly from June to October. In July 2008, long-lasting insecticide-bed nets (LLINs) were distributed to all villagers. This was followed by a decrease to malaria incidence between August 2008 and August 2010 [44]. However, two episodic resurgences of malaria attacks were noted in adults and children aged up to 10 years old in 2010 and 2013 [44,45].
In the village, since the establishment of the project, mosquitoes have been captured monthly using human landing catches (HLC) from indoor and outdoor habitations between 7 p.m. and 7 a.m. for three consecutive nights. All anopheline species collected were identified in the field using the morphological identification keys of Gilles and Coetzee [46] and the specimens were individually stored in Eppendorf tubes containing silica gel and then taken for further molecular analysis. In this study, mosquitoes were sampled during the rainy season in September and October 2006 (2 years before LLINs), in July 2008 (date of implementation of LLINs), September 2008 and September and October 2010 (2 years after LLINs were first used).

DNA extraction and species identification
Genomic DNA from each mosquito was extracted using the kit NucleoSpin Tissue XS made by Machery-Nagel. DNA extracts were amplified by polymerase chain reaction (PCR) to discriminate species within the An. gambiae complex (Table 1) using the protocol of Wilkins et al. [47], with intentional mismatch primers recognising single nucleotide polymorphisms (SNPs) at the 3-prime end (IMP-PCR).
In order to have a sufficient sample size for microsatellite analysis, i.e. no less than 15 individuals, some samples were pooled as follows ( Table 2): An. coluzzii and An. gambiae samples were analysed according to the location of capture, indoor and outdoor in each year, and if necessary indoor and outdoor samples of each species were pooled per year. Anopheles arabiensis samples were analysed according to the location of capture in each month (July; September; October) except in 2006. The identified hybrids between An. coluzzii and An. gambiae (six individuals) were not included in the general microsatellite analysis in order to avoid bias in gene flow estimation, but were subjected to the assignment test.
Several studies have reported different molecular approaches to identify the An. coluzzii and An. gambiae species. However, Santolomazza et al. [48] reported that these methods are not entirely interchangeable and some differences have been reported among results. For this purpose, to avoid biases that could affect the interpretation of our genetic analyses, a subset of specimens, were randomly sampled and identified according to the polymorphism of nearly 200 bp-long the Short Interspersed Elements (SINE200) in division 6 of the X-chromosome as described by Santolamazza et al. [49].

Microsatellite genotyping
Fourteen microsatellite loci (Table 3) selected from the genetic map of An. gambiae published by Zheng et al. [50] were used for genotyping studies. These markers were located on chromosome 3 to avoid potential bias resulting from selective pressures associated with chromosomal inversions on chromosome 2 [26] and genomic island of divergence regions on the X chromosome [29]. Microsatellite markers were amplified for each specimen by multiplexed PCR using fluorescently labeled (PET, NED, FAM, VIC) forward primer in a final volume of 10 µl containing 5 µl of 2× PCR Master Mix (Applied Biosystems ® ), 1 µl of primer mix, 3 µl of RNase free water and 1 µl of genomic DNA. Four primer mixes were constituted according to the annealing temperature (Ta) of markers in a final volume of 400 µl, consisting of 12.5 pmol of each primer. Amplification reactions were performed with an initial denaturation step of 10 min at 94 °C followed by 25-30 cycles of 30 s at 94 °C, 30 s at Ta °C, 30 s at 72 °C, and then an extension step of 5 min at 72 °C. PCR products were mixed with Genescan-500 Liz size standard and deionized formamide (Applied Biosystem). Amplified fragments were separated by capillary electrophoresis in an automatic sequencer (ABI 3130xl Genetic Analyser) and the allele's size scored using Gen-eMapper software (Applied Biosystem).

Data analysis
Microchecker 2.2.3 software developed by Van Oosterhout et al. [51] was initially used to examine possible genotyping errors due to null alleles. Genetic diversity was assessed per locus, per population, by estimating the number of observed alleles (Na), and by the mean number of alleles per locus (A) using Arlequin.3.5.1 [52]. The same software was used to test deviation from Hardy-Weinberg equilibrium (HWE) for each locus in each population and to estimate the observed heterozygosity (Ho) and expected heterozygosity (He). Statistical significance of HWE was assessed by the exact probability test available in Genepop 3.2 software [53] and the same software was used to estimate the inbreeding coefficient (Fis) and the linkage disequilibrium between each pair of loci in each population. A possibly significant heterozygosity excess (the signature of a bottleneck) was tested with the Bottleneck program [54] using two mutational models: an infinite allele model (IAM) and a stepwise mutation model (SMM). The probabilities for the sign tests (ps) for heterozygosity excess and for pw Wilcoxon test (two-tailed for heterozygote excess or deficiency) were calculated. Genetic differentiation between populations was assessed using pairwise estimates Fst values according to Weir and Cockerham, computed using Arlequin software [52]. Additionally, an assignment test implemented in the Geneclass 2 program [54] and developed by Paetkau et al. [55] was used for populations of each species to estimate the likelihood of an individual's multilocus genotype being assigned to the population in which it has the highest likelihood of belonging, compared to the likelihood of being assigned to other populations. The test was thus used including all the individuals of both An. coluzzii and An. gambiae. This partition was used as a reference to assign hybrids to a most likely An. coluzzii or/and An. gambiae population using the Geneclass 2 program [54].
For the six An. gambiae/An. coluzzii hybrids only one was identified as a hybrid with the second method. Therefore, the species identified according to IMP-PCR which could discriminate more hybrids were analyzed with microsatellite markers. Hybrids individuals were then discarded for the general population genetic studies but not for the assignment test.

Genetic structure of Anopheles populations
Fourteen microsatellites previously defined on An. gambiae were tested (Table 3). Polymorphism parameters for An. gambiae, An. coluzzii and An. arabiensis samples (all pooled populations) are given in Table 3. All microsatellites were polymorphic except H127, which was monomorphic in the three species, and H242 in An. arabiensis. Moreover, the locus H88 failed to amplify properly, particularly in An. coluzzii and An. gambiae individuals, even after optimization of PCR conditions. Thus, these loci were removed from further analyses in the concerned species. Moreover, Microchecker analysis performed on each Anopheles population showed the presence of null alleles at the H812 and H758 loci within An. gambiae and An. coluzzii populations, and H817, H128, and H555 in An. arabiensis populations (Table 3). Therefore, these loci were also discarded. A total of 11 polymorphic loci were analysed in An. gambiae and An. coluzzii populations and 10 in An. arabiensis populations. For the sister species, the total number of observed alleles (Na) for the polymorphic locus ranged from 7 (H242 and H817) to 25 (H128) for An. coluzzii and 6 (H817) to 24 (H128) for An. gambiae, with a mean of 10.48 and 10.21 alleles per locus respectively ( Table 3) Table 4).
The Hardy-Weinberg equilibrium tested in each population for each locus showed 17/44 and 17/33 significant departures from Hardy-Weinberg proportions in equilibrium within population of An. coluzzii and An. gambiae, respectively (Additional file 1: Table S1).
Concerning the An. arabiensis species, a higher polymorphism than for the two sister species is shown with a total number of alleles ranging from 10 (H119) to 30 (H93), with a mean value of 20.9 and a mean number of alleles per locus of 11.14 ( Table 3). The multilocus study performed for each population showed the minimum mean number of alleles of all loci in AROUTSEP06 and AROUTSEP08 populations (8.50) and the maximum in ARINDJUL08 and AROUTJUL08 (14.000 and 14.400) ( Table 4). The mean observed heterozygosis (Ho) in An. arabiensis samples ranged from 0.560 (AROUTSEP06) to 0.738 (ARINDJUL08). All the samples presented a deficit in heterozygotes from the Hardy-Weinberg equilibrium. Positive inbreeding coefficients (Fis) ranged from 0.036 to 0.27 but were not significant according to the confidence interval (Table 4). The Hardy-Weinberg equilibrium tested for each locus showed 40/90 significant departures from Hardy-Weinberg proportions in equilibrium (Additional file 1: Table S2).

Genetic differentiation between populations
Estimates of pairwise Fst values were calculated between populations of each sister species (An. coluzzii, An. gambiae), and between populations of the two species (Table 5). Except for two An. gambiae populations (AGIND10 and AGOUT10), no significant differentiation within An. coluzzii or An. gambiae populations (P > 0.05) was observed, indicating for each species no genetic differentiation between populations sampled indoors and outdoors, and/or in different years, including before (2006)

Table 4 Genetic variability for each population of An. coluzzii (AC), An. gambiae (AG), and An. arabiensis with: sample size (N), mean number of alleles for all loci (A), observed heterozygosity (Ho), expected heterozygosity (He), P value in Hardy-Weinberg equilibrium and inbreeding coefficient (Fis) and confidence interval (CI), probability for H excess or deficiency for the sign tests (p s ) with, in brackets, the ratio of the number of loci with heterozygote excess to the number with heterozygote deficiency and for the one-tailed Wilcoxon test for H excess (p w ) for the infinite allele model (IAM) and the stepwise mutation model (SMM)
Those in italic indicate only the significant P value for H excess

Populations
N A ± se Ho ± se He ± se P value ± se Fis CI 95%

Assignment test
The results of the assignment test using Geneclass 2 revealed that for An. coluzzii populations, less than a quarter of individuals (about 8-13.33%) are classified in their original population, except one population (ACOUT06). For An. gambiae populations, around half of the individuals were assigned to their original population except AG06 (21.74%) ( Table 7). However, when all samples of An. coluzzii are pooled, 75.63% of individuals are correctly assigned to An. coluzzii, and a similar percentage is observed for An. gambiae (73.17%). The assignment test in hybrids showed that four individuals were highly assigned to An. coluzzii (percentage assignment >95%) while two were highly assigned to An. gambiae (Table 9). For An. arabiensis populations, 7.4-22% of individuals are assigned to their original population (Table 8).

Discussion
The three major vectors of An. gambiae complex, An. coluzzii, An. gambiae and An. arabiensis, have an overlapping distribution in Dielmo as described in many parts of West Africa [5]. The aim of the present study was to elucidate whether implementation of LLINs in Dielmo  has an impact on the genetic structure of these populations. Using microsatellite markers, the genetic structure of mosquito populations was analysed before (2006), during (2008) and after (2010) the implementation of LLINs in Dielmo. These microsatellites mapped throughout the chromosome 3 have been chosen to avoid confounding patterns of genetic structure associated to linkedmarkers on polymorphic inversions on chromosome 2 and putative genes of reproductive isolation on the X chromosome [26,29]. The results of An. coluzzii and An. gambiae populations should be interpreted with caution due to the low population size of samples compared with those of An. arabiensis, and the resulting pooling of some samples. However, the observed heterozygosity is quite similar whatever the year and the species with significant deficits in heterozygotes in all populations. Various factors could explain the observed heterozygote deficit compared to the expected heterozygosity under HWE. A Wahlund effect due to sample bias cannot be discarded for some pooled samples, but genetic drift due to repetitive reduction in population size resulting from bottlenecks is likely, since all the An. coluzzii and An. gambiae populations exhibit heterozygote excess under the IAM model using the Wilcoxon test with Bottleneck software. This is congruent with entomological data from a previous study showing that after the implementation of LLINs in Dielmo, the relative abundance of Anopheles populations fluctuated substantially in our study area, with a dramatic decrease of An. coluzzii and An. gambiae, while An. arabiensis increased and remained the prevalent species [28]. However, no significant inbreeding is noticed (F IS values), so at each generation, gene flow would be sufficient to prevent inbreeding. Between pairs of populations within each species no significant genetic structuration (F ST values) was observed except the two An. gambiae populations sampled indoors and outdoors in 2010 after the implementation of LLINs in Dielmo. Between the sister species, the lack of genetic differentiation between some populations reveals the occurrence of a gene flow. Because the results of the assignation test using neutral markers fails to shown a clear pattern of hybridization (for hybrids, the assignation is not equal for the two species), results are in favour of introgressed individuals, with backcrossing of hybrids with parental species. The asymmetrical situation is observed with a higher proportion of hybrids assigned to An. coluzzii, in favour of the introgression of the genome of An. coluzzii by An. gambiae, except for the only one hybrid in 2010. From a methodological point of view, it should be noted that the IMP-PCR and SINE200 methods are quite equivalent for the species identification but the IMP-PCR method seems more acute to detect hybrids. Previous studies in West and Central African countries where An. coluzzii and An. gambiae are frequently  sympatric reported a strong deficit of hybrids in nature that was below 1%, resulting in reproductive isolation between both species [56,57]. However, hybrids rates ranged from 7% in Gambia [57] and over 20% in Guinea Bissau [7,8,36,58]. In the study, the figure of 2.5% of hybrids is in agreement with a previous study performed in Senegal [42]. In Guinea Bissau high genetic divergence was found between An. coluzzii and An. gambiae (Fst value = 0.348), but 60% of field-collected hybrids were backcrossed [8]. Indeed, Lee et al. [11] reported that the assortative mating between An. coluzzii and An. gambiae periodically breaks down, explaining high levels of hybridization in various sites in West Africa lying outside coastal areas (5.2-96.9%). Additionally in the laboratory, Diabaté et al. [59] found that F1 hybrids are fully fertile and viable. In Benin, the occurrence of the Kdr mutation in An. coluzzii populations was attributed to an introgression process involving An. gambiae, which already presented the kdr mutation [60]. In Mali Norris et al. [14] showed before the implementation of ITNs and IRS, a breakdown of assortative mating resulting in high levels of hybrids in 2006 and the introgression of the Kdr mutation in An. coluzzii populations. This led to an increase in the frequency of selectively advantaged backcrossed individuals after the widespread introduction of ITNs. In our study, between An. coluzzii and An. gambiae populations, significant gene flow was found in 2006 and 2008, but for 2010, all the Fst values were significant. Thus, as for Mali populations, there was a breakdown of assortative mating resulting in hybrids in the same year 2006, but the introduction of LLINs had a negative impact on hybridization frequency, which is corroborated by the decrease in the number of hybrids observed in the data after 2008. This study suggests decreased mating success of hybrids or ecological maladaptation, but also a lesser probability of mating between species due to a decrease in An. coluzzii population size. The lack of An. coluzzii could explain the occurrence in 2010 of a unique hybrid assigned to An. gambiae, probably resulting from hybrid backcrossing with the most populous species, An. gambiae. Then, the genetic divergence could be maintained between the two species despite gene flow, which is consistent the speciation-with-gene flow model as proposed by Turner et al. [29].
Regarding the sibling species An. arabiensis, results suggested some spatial and temporal structuration between populations, especially in 2008, of the seven significant F ST pairwise comparisons, six involved in 2008 collected populations. If An. arabiensis, the most widespread member of the An. gambiae complex was described generally to be genetically less structured compared to the sibling species An. coluzzii and An. gambiae populations, some studies demonstrated genetic distinct subpopulations [37,39]. The level of genetic differentiation in An. arabiensis populations varies significantly according to geographical distance, as demonstrated in Senegalese and Indian Ocean island populations [40] or in Sudanese to Mozambican populations [61], but also on a local scale in a village in Tanzania due to ecological diversification [39]. Polymorphic inversions were found at differential frequencies in West Africa than in East Africa Anopheles populations [62,63] and between outdoor and indoor populations [26]. Inversions were involved in local adaptation, with the selection of co-adapted genes affecting behaviour activities [64,65]. Marsden et al. [66] suggested that chromosomal inversions contribute to population structure in An. arabiensis. This species, although generally considered to be a less efficient vector, was a major malaria vector in Dielmo due to its high human biting rates [67]. This species presents behaviour plasticity according its feeding activities, resting places and also to its great tendency to survive during the dry season [68]. After the implementation of LLINs in 2008 to Dielmo, a reconfiguration of vector abundance was noted, a large decrease in population size was observed in An. coluzzii and An. gambiae, but an expansion of An. arabiensis [28]. The lack of bottleneck effects nor significant inbreeding in populations are in adequacy with large population size. A higher genetic diversity was observed in July 2008 compared with 2006 for both indoor and outdoor populations. This could result from the migration of individuals coming from genetically divergent populations. Indeed, an entomological study conducted in villages around the Dielmo area showed that most of the Anopheles populations in these areas belong to An. arabiensis [69]. The higher genetic diversity coincides with the date of LLIN implementation. A diversifying selection could be envisaged, favouring individuals with genotypes diverging from the mean genotype of the population. Various scenarios could be proposed, such as the occurrence of various resistance genes conferring individual advantage and then their selection, while nonresistant individuals were eliminated. The selection of individuals with divergent behaviour in biting activities could be envisaged, since it was demonstrated in Dielmo for another vector, Anopheles funestus, where a shift to diurnal feeding, essentially after the introduction of LLINs, was observed. This highlights the need to explore the evolution of the genetic basis of behavioural traits in Anopheles populations in the context of vector control of malaria transmission.

Conclusions
Several studies in Africa carried out after the implementation of interventions strategies have focused on the survival, the status of insecticide resistance and behavioural biting changes of Anopheles and on vector composition. However, little information is available about the genetic structure of the main malaria vectors of the An. gambiae complex involved in Senegal, which is particularly worrying in the context of high coverage of insecticide-treated bed nets in the malaria control programme. In this study, the potential effects of the implementation of LLINs on the genetic structure of An. gambiae, An. coluzzii and An. arabiensis were demonstrated. According to the species, the effect after the implementation of LLINs in Dielmo is differentiated. The sister species An. coluzzii and An. gambiae populations regularly experienced bottleneck, but without significant inbreeding. Since a breakdown of assortative mating resulted in hybrids, the introduction of ITNs had a negative impact on hybridization frequency. As pointed by Marsden et al. [8] the occurrence of a various level of hybrids across West Africa probably results from a "geographic mosaic of reproductive isolation" but this phenomenon could have considerable implications for transgenic control strategies.
Regarding the sibling species An. arabiensis, the study suggested some spatial and temporal structuration between populations, especially in 2008, coinciding with the date of LLINs implementation, which could result from diversifying selection favouring the expansion of this species. Taking into account that vector control is the cornerstone for reducing the burden of malaria disease and that current tools are mainly based on insecticide-treated bed nets, multidisciplinary studies combining epidemiology, ecology, and population genetics are needed to define successful management of malaria vector control.

Authors' contributions
The study was designed by SS, PMS, CS, MH. SS performed the experiments. SS, MH analysed the data. SS and MH drafted the manuscript. SD, CS, ND, PMS Additional file Additional file 1: Table S1. Genetic variability for each locus within An. coluzzii and An. gambiae populations, observed heterozygosity (Ho), expected heterozygosity (He), P value in Hardy-Weinberg equilibrium and inbreeding coefficients (Fis). In bold: locus in Hardy-Weinberg disequilibrium. Table S2. Genetic variability for each locus within An. arabiensis populations, observed heterozygosity (Ho), expected heterozygosity (He), P value in Hardy-Weinberg equilibrium and inbreeding coefficients (Fis). In bold: locus in Hardy-Weinberg disequilibrium.
contributed to data interpretation and manuscript draft. All authors read and approved the final manuscript.