Anopheles bionomics, insecticide resistance and malaria transmission in southwest Burkina Faso: A pre-intervention study

Background Twenty-seven villages were selected in southwest Burkina Faso to implement new vector control strategies in addition to long lasting insecticidal nets (LLINs) through a Randomized Controlled Trial (RCT). We conducted entomological surveys in the villages during the dry cold season (January 2017), dry hot season (March 2017) and rainy season (June 2017) to describe malaria vectors bionomics, insecticide resistance and transmission prior to this trial. Methods We carried out hourly catches (from 17:00 to 09:00) inside and outside 4 houses in each village using the Human Landing Catch technique. Mosquitoes were identified using morphological taxonomic keys. Specimens belonging to the Anopheles gambiae complex and Anopheles funestus group were identified using molecular techniques as well as detection of Plasmodium falciparum infection and insecticide resistance target-site mutations. Results Eight Anopheles species were detected in the area. Anopheles funestus s.s was the main vector during the dry cold season. It was replaced by Anopheles coluzzii during the dry hot season whereas An. coluzzii and An. gambiae s.s. were the dominant species during the rainy season. Species composition of the Anopheles population varied significantly among seasons. All insecticide resistance mechanisms (kdr-w, kdr-e and ace-1 target site mutations) investigated were found in each members of the An. gambiae complex but at different frequencies. We observed early and late biting phenotypes in the main malaria vector species. Entomological inoculation rates were 2.61, 2.67 and 11.25 infected bites per human per month during dry cold season, dry hot season and rainy season, respectively. Conclusion The entomological indicators of malaria transmission were high despite the universal coverage with LLINs. We detected early and late biting phenotypes in the main malaria vector species as well as physiological insecticide resistance mechanisms. These data will be used to evaluate the impact of complementary tools to LLINs in an upcoming RCT.


Introduction
The World Health Organization (WHO) has reported 228 million malaria cases and 405 000 deaths worldwide in 2018 [1]. Significant progress has been made in the era of malaria control and elimination between 2000 and 2017 in all regions of the world [2]. The number of malaria cases worldwide has decreased by 62% between 2000 and 2015 but seems to rebound since 2016 [2][3][4]. Sub-Saharan Africa accounted for about 93% of cases and deaths in 2017 [1]. In 2017, the health ministry of Burkina Faso recorded 11.9 million cases and 4144 deaths attributed to malaria [5].
Malaria control is mainly based on symptomatic and preventive treatments (with artemisinin-based combination therapies: ACTs) and vector control. Vector control aims at reducing malaria transmission by targeting Anopheles mosquitoes that transmit Plasmodium parasites. Core vector control measures rely mostly on the mass distribution of long-lasting insecticidal nets (LLINs) and high coverage of indoor residual spraying (IRS) to reduce the risk of malaria infection by targeting indoor biting mosquitoes [4]. It was estimated that LLINs have contributed to 68% of the decline in malaria cases observed between 2000 and 2015 in Africa [6] despite moderate use rates (LLIN use rate reached 55% in 2015 in sub-Saharan Africa [7]. However, the emergence of physiological [8][9][10] and behavioral [11,12] insecticide resistance mechanisms in Anopheles mosquitoes, as observed in most parts of Africa, could both compromise the effectiveness of LLINs and explain the recent rebound in malaria cases. Consequently, there is an urgent need for increasing or complementing the protection provided by the LLINs. Complementary strategies exists but before being included into strategic plans by national malaria control programs (NMCPs), supported by international donors and implemented in endemic countries, they need to be evaluated through rigorous and independent process [13]. In this context, Institut de Recherche en Sciences de la Santé (IRSS), Institut de Recherche pour le Développement (IRD) and Institut Pierre Richet (IPR) have been funded to conduct the REACT (Gestion de la REsistAnce aux inseCTicides au Burkina Faso et en Côte d'Ivoire) project to evaluate in Burkina Faso and Côte d'Ivoire four complementary strategies to LLINs trough a randomized controlled trial (RCT). These strategies included: i) larviciding with Bacillus thuringiensis israelensis (Bti) to target immature stages of Anopheles species, ii) Indoor residual spraying (IRS) with pirimiphos-methy to target endophilic malaria vectors, iii) information, education, community (IEC) to improve LLINs use iv) ivermectin administration to animals, a One health approach to tackle zoophagic behavior of malaria vectors and to improve animal health. The current study conducted entomological surveys to describe malaria vector bionomics, insecticide resistance and transmission prior to the implementation of the REACT RCT.

Study site and design
The study was conducted in the Diébougou health district located in southwest Burkina Faso. The natural vegetation is mainly wooded savannah dotted with clear forest gallery. The climate is tropical with two seasons: one dry season from October to April and one rainy season from May to September. The average annual rainfall is about 1200 mm. The dry season is "cold" from December to February (with average minimal and maximal temperatures of 18˚C and 35˚C, respectively) and "hot" from March to April (with average minimal and maximal temperatures of 25˚C and 40˚C, respectively). Agriculture (cotton growing and cereals) is the main economic activity in the area, followed by artisanal gold mining and production of coal and wood [14,15]. The study took place in 27 villages that were selected considering accessibility during the rainy season, a size of 200-500 inhabitants per village and a distance between two villages higher than two kilometers. According to a census carried out by our team in August 2016 one month after a universal distribution by the NMCP, the proportion of households with at least one LLIN and with at least one LLIN for every two inhabitants were 96.15% and 71.17%, respectively with 8680 inhabitants in the 27 selected villages. The proportion of households with access to an LLIN in their household was 87.60%. We calculated these indicators according to WHO formulas [16].

Mosquito collection and determination
We carried out three rounds of mosquito collection in January 2017 (dry cold season), March 2017 (dry hot season) and June 2017 (rainy season) using Human Landing Collections (HLC). The procedure for conducting HLC was for a person to sit on a stool, and mosquitoes to alight on his exposed legs where they were then collected using a hemolysis tube [17,18]. Mosquitoes were collected from 17:00 to 09:00 both indoors and outdoors at 4 sites per village (sites of collection remained the same during the three surveys). We performed collections in three villages simultaneously for one night leading to 9 nights of collections needed to survey all 27 villages during each survey. Due to a long duration of collection (16h), two teams of 8 collectors worked in each village. The first team worked from 17:00 to 01:00 when they were replaced by the second team from 01:00 to 09:00. Collectors were rotated among the collection points every hour. Indoor collection points were rooms that meet the following criteria: being usually inhabited; quiet without excessive movement of peoples; open to the outside through a door or a window. Outdoor collection was conducted in areas usually occupied by people but are sheltered from wind, traffic, fires and are not large meeting areas. The distance between collection sites was at least 50 m. The distance between indoors and outdoors collection points in one site was at least 10 m to minimize competition between mosquito collectors. Mosquitoes were collected in individual tubes plugged with cotton and stored in hourly bags. Independent staff supervised rotation of the mosquito collection and regularly checked for the quality of the mosquito collection. The following criteria were checked: respected collection location, collector at his post, collector awake, collector in a correct position, collector adequately dressed, correct hourly bags used. If one of the criteria was not respected, required arrangements were immediately made by the supervisor.

Morphological identification and dissection
Mosquitoes were morphologically identified where possible in the field to genus and species levels using morphological keys [19,20]. A subsample of 100 non blood-fed Anopheles spp. individuals was randomly selected per survey and per village and dissected to identify their parity state (parous or nulliparous) [21]. Parous female are those that have laid eggs at least once. All females belonging to the Anopheles genus were stored in individual tubes with silicagel and preserved at -20˚C for further analyses.

Molecular analysis
DNA extracted from head-thorax of Anopheles spp. individuals was used to detect Plasmodium falciparum infection using quantitative polymerase chain reaction (PCR) assay [22]. Individuals belonging to the Anopheles gambiae complex and the Anopheles funestus group were identified to species by PCR [23][24][25].

Parameters measured
We calculated the human biting rate (HBR; the number of vectors' bites per human per night (b.h -1 .n -1 ), the sporozoite infection rate (SIR: the proportion of Anopheles infected by P. falciparum), the entomological inoculation rate (EIR; the number of infected bites per human per night (ib.h -1 .n -1 ), the endophagy rate (ER; the proportion of Anopheles females collected indoors) and the parous rate (PR; the proportion of parous females over the total dissected) for each Anopheles species and for overall Anopheles spp. From PRs, we deduced daily survival rates (p) using the Davidson's method [29] (p ¼ ffi ffi ffi ffi ffi ffi PR g p with the duration of the gonotrophic cycle g assumed to be 2 days). Knowing survival rates, we were able to estimate the proportion P d of the vector population that reach the epidemiologically dangerous age according to the formulae given by MacDonald (Expression 5 p.16 in [30]: P d = p n ) assuming the sporogony duration n of malaria parasites in mosquitoes takes 11 days after which the parasites can be transmitted back to humans.

Statistical analyses
We assessed HBR, ER, SIR, EIR and PR using generalized mixed effect models (GLMM) with collection points and villages as nested random intercept. All models were fitted using the 'glmer' function of the package 'lme4' [31] using R software [32]. For all models, we performed backward stepwise deletion of the fixed terms followed by Likelihood ratio tests. Term removals that significantly reduced explanatory power (p < 0.05) were retained in the minimal adequate model. We used the post-hoc Tukey method to do multiple comparison among modalities of the fixed terms and calculated estimated marginal means (EMM) using the 'emmeans' function of the 'emmeans' package [33]. We considered main effects and interactions of the fixed terms in all the models.
We performed HBR and EIR analyses using negative-binomial models fitted on nightly counts of all Anopheles and of Anopheles individuals found to be infected by P. falciparum, respectively. Fixed terms were the season (dry cold, dry hot or rainy) and the collection position (indoor or outdoor). The results of final models are presented (S1 File).
We performed SIR, ER and PR analyses using binomial models fitted on individual status of the Anopheles (infected vs. uninfected, collected indoors vs. outdoors or parous vs. nulliparous, respectively). Fixed terms for these binomial models were the season and the individual species. Rate ratio (RR) for Negative Binomial models and Odds ratio (OR) for Binomial models were computed with 95% confidence intervals.
The relationship between Anopheles species composition and seasons was tested using a Fisher's exact test.
We assessed nightly activity of each major Anopheles species by comparing their Median Catching Time (MCT), which represents the time at which 50% of the individuals were collected [11], using a Kruskal-Wallis test and a Dunn's post-hoc test for multiple comparisons (function 'dunnTest' of the 'VIA' package) in R [34].
We compared the genoptypes frequencies of kdr and ace-1 mutations among seasons and species using the G-test [35] implemented in Genepop 4.7 and run in R [36]. Genotypic frequencies at the kdr and ace-1 mutations were tested for conformity to Hardy-Weinberg equilibrium using the "Exact HW test" [37]. In case of disequilibrium, we tested heterozygote excess and deficiency using the score test [38]. We tested the relationship between infectiousness and the genotypes for the mutations kdr-w, kdr-e and ace1 using Pearson's Chi-square tests (or a Fisher's exact test when one or more count were < 5).

Ethics approval and consent to participate
The protocol of this study was reviewed and approved by the Institutional Ethics Committee of the Institut de Recherche en Sciences de la Santé (IEC-IRSS) and registered as N˚A06/2016/ CEIRES. Mosquito collectors and supervisors gave their written informed consent. They received a vaccine against yellow fever as a prophylactic measure. Collectors were treated free of charge for malaria according to WHO recommendations.

Anopheles densities and composition
During the three surveys, we collected a total of 2591 mosquitoes belonging to four genus: During the dry cold season, An. funestus s.s was the most abundant species representing 78.64% (n = 221/281), followed by An. gambiae s.s (6.04%, n = 17/281), An. coluzzii (4.98%, n = 14/281) and An. pharoensis (4.98%, n = 14/281) ( Fig 1A). The other species were found at very low percentages (~1%) (Fig 1A). The relative abundance and species composition of the Anopheles population varied from one village to another. We collected Anopheles mosquitoes in 19 villages out of the 27 surveyed with the highest densities registered in Diagnon (n = 149, 4 species) and Kpédia (n = 39, 3 species) ( Fig 1A). We observed the highest Anopheles species diversity in Niaba (n = 15, 6 species) where six species were identified. At the opposite in nine villages, only one species of Anopheles was found (either An. funestus s.s, An. coluzzii or An. pharoensis).
During the dry hot season, An. coluzzii (75.86%, n = 220/290) almost totally replaced An. funestus s.s (8.27%, n = 24/290) while proportions of other species do not vary substantially. During this season, Anopheles mosquitoes were collected in only 15 villages (Fig 1B). Highest densities and species diversities were observed in Diagnon (n = 179, four species) and Kpédia (n = 65, four species) ( Fig 1B). We identified three Anopheles species in both Nouvielgane and Nipodja villages and two Anopheles species in Yelbelela and Niaba. In all other villages, only one Anopheles species was present at low density.

Plasmodium infection and transmission intensity
We analyzed a total of 1808 head-thoraxes of Anopheles for the research of P. falciparum infection. The overall sporozoite infection rate (SIR) was 0.07 [0.02-0.14] (19/272), 0.06 [0.01-0.11]    Table). These values of parous rate were used to calculated the percentage of Anopheles which could be found beyond the epidemiologically dangerous age and could therefore transmit malaria parasites. The percentages of Anopheles beyond the epidemiologically dangerous age were 22.10%, 27.16% and 10.78% in the dry cold, dry hot and rainy seasons, respectively. The average parous rate of Anopheles mosquitoes during both cold and hot dry seasons was significantly higher than during the rainy season (OR = 1. 8 Table). The parous rate did not differ significantly among the species (Likelihood ratio test χ 2 = 2.51, df = 3, p = 0.47).

Frequencies of L1014F kdr, L1014S kdr and G119S ace-1 mutations in An. gambiae s.l.
Numbers of individuals of each genotype of the three mutations and their frequencies in An. arabiensis, An. gambiae s.s. and An. coluzzii are presented for each season (Table 4). We were not able to find a significant relationship between infectiousness and the genotypes of the kdrw mutation (χ 2 = 1.44, df = 2, p = 0.48), kdr-e mutation (Fisher's exact test, p = 0.16) and ace-1 mutation (Fisher's exact test, p = 0.17) in the An. gambiae s.l. population.
In An. arabiensis, the kdr-w mutation did not vary significantly among seasons (exact G test p-values > 0.15) nor among villages (exact G test p-values > 0.08). The population did not differ significantly from the Hardy-Weinberg equilibrium (HWE) whatever the season (exact HW test p > 0.31).
The kdr-w frequency in An. coluzzii during the dry cold season was 0.61, it increased (not significantly, exact G test p = 0.41) to 0.69 during dry hot season and then decreased significantly to 0.55 during the rainy season (exact G test p<0.001). During the dry hot season, when the frequency of the kdr-w mutation was the highest, the population was not at the HWE In An. gambiae s.s, the kdr-w mutation frequency was 0.97 during dry cold season and significantly decrease to 0.59 in dry hot season (exact G test p < 0.001). During the rainy season when mosquito densities were very high, the kdr-w frequency rose up to 0.88. That was significantly higher than during dry hot season (exact G test p < 0.001) but not different than during dry cold season (exact G test p = 0.12). This population was not at the HWE for the kdr-w mutation (exact HW test p < 0.001) due to heterozygote deficiency (exact HW test p < 0.001). Heterozygote deficiency was observed in most of the villages during each season (S2 File).
In An. arabiensis, the kdr-e mutation was detected only during the rainy season at a frequency of 0.30. The frequency of the kdr-e mutation did not vary significantly among villages (exact G test p > 0.10). A significant deviation to HWE was observed (exact HW test p = 0.002) due to heterozygote excess in rainy season (exact HW test p = 0.001). However, we did not observe this heterozygote excess at the village scale (S3 File).
In An. coluzzii, the kdr-e mutation was not detected during the dry-cold season and only one heterozygous individual was collected during the dry hot season corresponding to a frequency of 0.002. The frequency increased significantly to 0.12 during the rainy season (exact G test p < 0.001). A significant deviation to HWE was observed for the kdr-e mutation (exact HW test p < 0.001) due to heterozygote excess in most of the villages during each season (S4 File).
In An. gambiae s.s, the kdr-e mutation was detected only during the rainy season at a frequency of 0.03. The kdr-e mutation did not vary significantly among villages (exact G test p > 0.11). The population did not differ significantly from the HWE (exact HW test p > 0.05) in rainy season.
In An. arabiensis, the ace-1 mutation was detected only during the dry hot season at a frequency of 0.12. The frequency of the ace-1 mutation did not vary significantly among villages (exact G test p > 0.48). The population did not differ significantly from the HWE (exact HW test p > 0.05) in dry hot season.
The frequency of the ace-1 allele in An. coluzzii was 0.07, 0.02 and 0.02 in dry cold, dry hot and rainy seasons, respectively. There were no significant difference in the ace-1 allele Table 4. Allele frequency of kdr L1014F, kdr L1014S and ace-1 G119S mutations in Anopheles gambiae s.l. populations.

PLOS ONE
frequency among seasons (exact G test p > 0.11) nor among villages (exact G test p > 0.13).
The ace-1 allele frequency in An. coluzzii did not differ from the HWE (exact HW test p > 0.05).
In An. gambiae s.s, the frequencies of the ace-1 allele were 0.22, 0.11 and 0.19 in dry cold, dry hot and rainy seasons, respectively. The frequency did not vary significantly among seasons (exact G test p > 0.23) nor among villages (exact G test p > 0.07). The population did not differ significantly from the HWE (exact HW test p > 0.12) whatever the season.
Three hundred and fifty four (354) individuals An. gambiae s.l carried multiple insecticide resistance mechanisms. Two hundred and seven (207) carried at least one mutated allele for both kdr-w and kdr-e mutations (belonging the An. gambiae s.s, An. coluzzii and An. arabiensis species). One hundred and forty-two (142) carried at least one mutated allele for both kdr-w and ace-1 mutations (belonging the An. gambiae s.s, An. coluzzii and An. arabiensis species) and 16 carried at least one mutated allele for both kdr-e and ace-1 mutations (belonging the An. gambiae s.s and An. coluzzii species). Eleven (11) individuals carried the three mutations (belonging the An. gambiae s.s and An. coluzzii species).

Discussion
This study showed that the malaria vector species and abundance in the Diébougou area varied significantly according to the season. Anopheles funestus s.s was the predominant vector during the dry cold season (January 2017) but most individuals were collected in one village (Diagnon) that is close to swamps on the edge of the dam of Bapla. Anopheles funestus s.s densities were ten times lower two month later during the dry hot season (March 2017) and it almost disappeared in June 2017 during the rainy season. Anopheles funestus s.s is known to breed in large permanent or semi-permanent pools preferentially with emergent vegetation on its margins [39,40]. In Burkina Faso, two chromosomal forms of An. funestus have been described named Folonzo and Kiribina [41,42]. Folonzo form of An. funestus is mainly associated with the presence of water reservoirs containing natural vegetation, such as swamps. The swamps near Diagnon are upstream from the Bapla dam and become dry at the end of the dry season. Provided that An. funestus in our study area is of the Folonzo form, this may explain why An. funestus s.s almost disappeared during the dry hot season and until swamps become green and flooded again.
Anopheles coluzzii was shown at very low densities during the dry cold season but became the predominant malaria vector species from the dry hot season (March 2017) to the rainy season (June 2017). During the dry hot season, most of individuals were collected in Diagnon, the same village where An. funestus s.s densities were simultaneously falling. In Burkina Faso, An. coluzzii is known to breed in permanent or semi-permanent breeding sites [43,44]. Its presence in Diagnon during the dry season is certainly linked to the dam. However, it is not clear why An. coluzzii was present in very low densities during January 2017 and became numerous two months later. We hypothesize that the reduction of the breeding sites favorable to An. funestus s.s may have increased An. coluzzii competitiveness against An. funestus s.s around the Bapla dam. This result is in accordance with the high variability of An. coluzzii densities during the dry season as described by Dao et al. [45] in the neighboring Mali.
Densities of An. gambiae s.s. and An. arabiensis were low during both dry (cold and hot) seasons and increased substantially during the rainy season (in the largest extent for An. gambiae s.s.). This is consistent with the preference of both these species to breed in temporary rain-dependent pools and puddles [40,43,46].
These four species (An. funestus s.s, An. coluzzii, An. gambiae s.s and An. arabiensis) were responsible for all the P. falciparum transmission recorded in this study. Sporozoite infection rate (SIR) did not vary among seasons but EIR increase from an average of about 3 infected bites per human per month in dry (cold and hot) seasons to more than 10 bites per human per month during the rainy season. The values of EIR recorded in our study are consistent with previous works in the same region of Burkina Faso [47][48][49]. Such levels of transmission (EIR > 100 infected bites per person per year) are relatively high when put in the African context [50]. As SIR was stable over the study, the increase in EIR in rainy season was mathematically due to the increase in vectors densities. According to our results, malaria transmission (as measured by EIR) persists year round in the study area despite high spatial disparities. Indeed, during the dry cold and dry hot seasons, the spatial distribution of An. funestus s.s. and An. coluzzii, that were the main malaria vectors, was restrained to a small number of villages that concentrated malaria transmission. During the rainy season, the spatial distribution of malaria vector (with predominance of An. coluzzii and An. gambiae s.s.) was more homogeneous.
In addition to these well-known primary malaria vectors, we recorded the presence of four other Anopheles species in low densities (An. nili, An. pharoensis, An. rufipes, An. squamosus). While we were not able to detect Plasmodium parasites in the few individuals belonging to these species, they are all potential vectors of Plasmodium [40,[51][52][53][54][55]. Because they are considered secondary vectors [56,57] (due to their main trophic behavior including exophagy and/ or zoophagy) or found in low densities, these species are often neglected in most of the recent studies carried out in Burkina Faso [58]. However, these vectors could possibly maintain residual levels of transmission [57,59]. Indeed, drastic reduction of the major vector densities may free up their ecological niches for secondary vectors [60][61][62][63]. Therefore, secondary vectors deserve particular attention [64]. Their bionomics and competence for P. falciparum transmission remain to be investigated.
Regarding the behavior of the main malaria vector species, we observed that An. coluzzii, An. gambiae s.s and An. funestus s.s were slightly but mainly endophagic (50 to 65% depending on the season) that is in accordance with previous results recorded in another location in southwest Burkina Faso [47]. This indicates that indoor vector control measures (such as LLINs and IRS) are expected to target a significant part of the vector population but probably insufficient to stop the transmission. The peak of aggressiveness of the main Anopheles species occurred during the second part of the night. This observation is consistent with the usual patterns of hourly biting aggressiveness of these species [65]. This period corresponds to the moments of deep sleep of the human populations who are potentially protected by LLINs [46,66]. However, we observed early-and late-biting phenotypes in the main malaria vector species (An. coluzzii, An. funestus s.s. and An. gambiae s.s). These phenotypes might mediate residual transmission and we might expect to see them selected by the massive use of insecticide-based vector control tools such as LLINs [67,68]. It is therefore crucial to monitor malaria vector behavior in this area when implementing malaria vector control strategies in order to track the possible emergence of behavioral resistances.
This study showed high parous rates of malaria vectors (> 60%) in the study area, regardless of the season. The parous rate was significantly lower in the rainy season than in the cold and hot dry seasons, indicating that young females were more prevalent in the rainy season. The lower parous rate of Anopheles vectors during the rainy season may be due to the presence of more breeding habitats which yield more nulliparous mosquitoes in vector populations [69]. It is also possible that a higher usage of LLINs during the rainy season as a result of a higher nuisance and lower temperatures might result in smaller proportions of parous vectors [70]. The three target-site mutations kdr-w, kdr-e and ace-1 that were tested for were detected in the three species of the An. gambiae complex but at varying frequencies among species and seasons. In June 2017 during the rainy season (when the populations were most abundant), the kdr-w mutation frequency was very high in An. gambiae s.s. (almost fixed, f = 0.88). Since the discovery of this mutation [26] in An. gambiae s.s, the kdr-w frequencies have always been high in this species in Southwest Burkina Faso [8,10,71] indicating that the mutation has a low cost for this species or that strong selection pressures occur. Selection pressure likely to occur might be due to vector control (universal distribution of LLINs is implemented in Burkina Faso since 2010 [72]) and massive use of pesticides in agriculture [73]. Indeed, the intensive use of insecticides in cotton cultivation might be a major factor driving the selection of pyrethroid-resistant specimens in Southwest Burkina Faso [10]. Most of the populations of An. gambiae s.s. from this study showed heterozygote deficiency for the kdr-w mutation that could indicate a Wahlund effect resulting from the presence of several populations within each given village. This might be explained by the colonization of temporary breeding sites by different vector populations during the rainy season as evidence the annual reappearance of An. gambiae s.s. in the Sahel after the dry season [45]. To note that no disequilibrium was observed for the kdr-e and ace-1 mutations in this species indicating that these populations, if they exists, resemble the local population in terms of genotypic composition for these loci.
During the rainy season, the kdr-w mutation frequency was lower in An. coluzzii (f = 0.55) and An. arabiensis (f = 0.21) than in An. gambiae s.s. But in counterpart, both An. coluzzii (f = 0.19) and An. arabiensis (f = 0.29) showed higher prevalence of the kdr-e mutations than An. gambiae s.s. (f = 0.03). In An. coluzzii and An. arabiensis, a heterozygote excess was observed for the kdr-e mutation. This disequilibrium may result from a heterozygote advantage as it has been observed in mating in An. coluzzii specimen from Southwest Burkina Faso carrying the kdr-w mutation [74]. The causes of this disequilibrium remain to be explored.
Both kdr mutations are well "implanted" in the vector population of our study area. These mutations provide physiological resistance to the pyrethroids insecticides, the only family of compounds authorized to impregnate LLINs. Although the impact of pyrethroids resistance on the efficacy of malaria control remains disputable [75], the development and the implementation of resistance management strategies need to be encouraged.
Regarding the ace-1 mutation that confers resistance to carbamates and organophosphates insecticides, it was present in the three species at a moderate frequency in An. gambiae s.s. but at very low frequencies in An. coluzzii and An. arabiensis. In Burkina Faso, the three species may express high frequencies of the ace-1 mutation depending on the location [76]. In our study area, An. gambiae s.s. seems to suffer more selection pressure possibly due to a wide use of carbamates and organophosphates particularly for cotton growing.
Recent studies in Tanzania and Senegal showed a relationship between the genotype for the kdr mutation of wild Anopheles gambiae s.l. and Plasmodium falciparum infection [77]. The mosquitoes carrying the mutation were more infected than those carrying the wild alleles. In contrast, we were not able to show such a relationship whatever the mutation studied (kdr-w, kdr-e or ace-1). This may be explained by the complex interactions occurring between insecticide resistance and Plasmodium infection as summarized in [78]. Indeed, while an insecticide resistance mutation may be associated with increased vector competence [79], the infection by Plasmodium falciparum tends to decrease survival of vectors carrying this mutation [80].

Conclusions
Malaria transmission in the Diébougou area was mainly due to An. funestus s.s, An. coluzzii, An. gambiae s.s and An. arabiensis with high spatio-temporal heterogeneities. The entomological indicators of malaria transmission were high despite the presence of LLINs. We observed early and late biting phenotypes in the main malaria vector species. These phenotypes might mediate residual transmission. Three mutations kdr-w, kdr-e and ace-1 were present in the three species of the An. gambiae complex with high frequency variability between species. These data will be used to evaluate the impact of complementary tools to LLINs in an upcoming RCT.