Plasmodium falciparum, one of the deadliest pathogens circulating in Niger and many other sub-Saharan countries, is contributing significantly to the high estimated child mortality rate (United Nations Children's Fund 2005). The major vectors responsible for P. falciparum transmission in Niger are Anopheles gambiae sensu lato (s.l.) and Anopheles funestus, the former being far more widespread over the country whereas the last is found in more limited areas 1. Of the sibling species and forms constituting the Anopheles gambiae complex, Anopheles arabiensis and both molecular M and S forms of An. gambiae sensu stricto (s.s.) have been found to date in Niger 23.

At the end of 2005, a nationwide long-lasting insecticide-treated net (LLIN) distribution targeting the most vulnerable members of the population was organized in conjunction with an integrated poliomyelitis vaccination campaign 4. This LLIN coverage aimed to reduce morbidity and mortality caused by malaria clinical cases. Such a reduction can be observed when the bed net usage rate in the population is high enough 56789. Although bed nets were already quite commonly used by the population, principally for protection against biting nuisance mosquitoes, this campaign strongly increased insecticide-treated bed net usage in households all over the country 10.

The effectiveness of bed nets impregnated with pyrethroid insecticides derives from several mechanisms that decrease the probability of a host-seeking female mosquito to succeed in taking a blood-meal. Besides the physical barrier constituted by the net, a deterrent effect limits unfed mosquitoes entering houses where a treated net is present, and in case of tarsal contact with the treated net, the insecticide compound could repel, hurt, or kill the mosquito. These properties have led to substantial reductions in indoor mosquito densities, and biting, feeding and survival rates in field trials 11121314 and experimental huts evaluations 15161718. Some of these studies have also demonstrated a reduction in P. falciparum transmission 111314.

However, a number of studies throughout Africa have reported the existence of anopheline populations resistant to various insecticides 151920212223 involving two major types of resistance mechanisms : insecticide target-site insensitivity due to single nucleotide polymorphisms (SNP), and increased enzymatic metabolization of insecticidal compounds 24. Two SNP at amino-acid position 1014 in the voltage-gated sodium channel gene have been described in An. gambiae 2526, leading to amino-acid substitutions involved in reducing the affinity of DDT and pyrethroid insecticides for their target site in the insect nervous system. These SNP, named knock-down resistance (kdr) mutations in relation to their phenotypic expression, were found in many field populations of An. gambiae s.s. and An. arabiensis covering West, Central and East Africa 272829303132333435363738394041424344454647484950. In West and West-Central Africa, it was shown that the kdr-w (or L1014F) allele was far more frequent and widespread in the S molecular form of An. gambiae 232728293233343536373850, whereas only few M form populations from a limited area near the Gulf of Guinea presented kdr-w alleles at low frequencies 23313236373850, except in few urban and peri-urban coastal areas where it reached high frequencies 2035404142434445. However, to date there have been almost no studies on kdr mutations in sahelian An. gambiae populations except one in Burkina Faso that included two sites in the sahelian zone of the country 32. No kdr M forms were found whereas one village exibited kdr-w S forms.

The monitoring of insecticide resistance in malaria vectors is of prime importance especially where control programmes are planned or already running, in order to assess potential selection effects of insecticidal compounds on vector populations, and to take appropriate measures such as switching to other classes of compounds. For this goal, the presence and frequency of the kdr mutations constitute a valuable and useful resistance marker for two main reasons. First, it provides an early warning of resistance development as the mutation arises well before any effect on phenotype can be detected in a population 51. Indeed, the expression of the 24 h-survival diagnostic phenotype 52 appear to be recessive 4553. Therefore, a population presenting a low kdr frequency mainly in heterozygous state is likely to show a high mortality rate during bioassays. As further evidences supporting the advantage of kdr genotyping over bioassays for emergent resistance detection, Chandre et al. 53 found a significant mortality reduction only when heterozygous females proportion reached 60%, and a significant increase of Knockdown time (KdT) only with 40% heterozygous females. Secondly, the kdr mutations seem to be well correlated with resistance phenotype 25384553 in both An. gambiae molecular forms, even if metabolic resistance mechanisms could also be involved in increased tolerance to pyrethroids 27.

Several authors have studied the effect of insecticide treated nets (ITNs) with pyrethroids on An. gambiae populations and the possible selection of kdr alleles either in experimental huts trials 18455354 or laboratory experiments 4553, helping in the understanding of advantages conferred by the mutations on survival and blood-feeding in areas of ITN use. However, few data are available on long-term and/or large-scale ITN coverage effect on resistance and kdr mutations in natural settings. In East Africa, no selection effect of long-term ITNs use on phenotypic resistance was noticed 5556, whereas Stump et al. 57 reported a significant increase of kdr-east mutation frequency in An. gambiae S form populations from Kenya after four years of ITNs community use. In West-Central Africa, a rapid increase of kdr-w mutation was observed in M forms from the island of Bioko following a large-scale insecticide residual spraying (IRS) programme 4158.

The distribution of the kdr-w mutation was studied in An. gambiae s.l. populations from the sahelian area of Niger during three consecutive wet seasons, allowing to monitor temporal variations following the nationwide LLIN coverage implemented after the first collection event. This study reports the first documented case of kdr mutation in An. gambiae populations of Niger and provides crucial information about potential effects of wide-scale LLIN coverage on kdr mutation selection.

This study describes the first case of kdr mutation-harbouring An. gambiae populations from Niger, to contribute to the description of the spatial distribution of kdr-w mutation across West Africa, especially in the northern limits of An. gambiae distribution area where such data are very rare. Fully susceptible genotypes were found in every An. arabiensis sample. This result is consistent with other studies from neighbouring countries where An. arabiensis exhibited very few if any kdr alleles 212932. Despite the fact that our S form sample was too limited to describe spatial and temporal dynamics of kdr-w mutation, its high prevalence is nevertheless interesting. The cause of mutation maintenance without significant insecticide pressure is unknown, as the mutation frequency was actually already 39% (n = 13) before the LLIN distribution.

The collected specimens allowed a more detailed study of M form populations throughout the country. It suggests that kdr-w mutation was already present at a low level in M form populations before the nationwide LLIN distribution, in various localities distributed all over the sahelian zone, at longitudes ranging approximately from 1°E to 10°E and latitudes below 16°N. This finding greatly extends the area of known kdr-carrying M form populations, that was to date limited to more humid areas in Côte d'Ivoire 43, Ghana 36, Burkina Faso 3132, Benin 1950, Nigeria 2250, Cameroon 3750, Equatorial Guinea 4146 and Angola 47. The mutation frequencies encountered are consistent with studies from neighbouring countries where kdr-w frequency was usually low in M molecular form : 6% in Malanville, Northern Benin, near the Nigerien border 19, 0.7% in Koubri 36 and 2% in VK7 32, southern Burkina Faso.

Based on strong supporting results, several authors 19212933343648 hypothesized that past and current agricultural use of pyrethroids and DDT for crop protection led to the selection of resistant individuals by challenging larval stages with residual insecticide products accumulating in water bodies around cultivated areas. This hypothesis was recently supported 44 by showing indirectly the presence of pesticide residues in soil and water from vegetable gardens and farms in Benin that limited the emergence rate of challenged larvae. We could relate our pre-LLIN low global kdr frequency to the presumed limited environmental selection pressure on An. gambiae populations, as the rural areas studied are usually surrounded by seasonal millet and/or sorghum subsistence cultivation for local consumption. These farming practices employ negligible amounts of pesticides. Unreliable precipitation and limited commercial demand tend to keep the use of inputs such as chemical fertilizer, pesticides and hired labour to a minimum 64. However, several localized mosquito populations might experience more pyrethroid exposure, especially in irrigation and gardening zones where agricultural production demands and allows financially insecticide use. Urban domestic use of pyrethroids for personal protection was also suggested to favor the emergence of resistant individuals within mosquito populations 2939. Indeed, the highest kdr frequencies ever detected in M forms were found around important cities in Côte d'Ivoire 43 and Benin 193544, corroborating the hypothesis of high insecticide pressure within urban environnements. In addition, it seems that M form populations breeding inside the city of Niamey also present high kdr-w frequencies, far greater than any analysed rural site. Some of the sampled larval populations were breeding near a small stream that crosses the city and is surrounded by year-round vegetable cultivation areas. We therefore cannot attribute potential selection effects to gardening and/or domestic insecticide use, but strongly suspect far higher insecticide exposure due to crop protection treatments. This hypothesis is sustained by the higher kdr frequency found near the cultivation areas compared to other larval habitats in Niamey. We conducted a basic interview in the cultivation areas, all 24 people reported repeated insecticide use along the year, bought in local market places. These multiple potential sources of pyrethroid pressure are common in many sub-Saharan urban and peri-urban areas 19354344, and are primarily due to the presence of cultivated zones in the outskirts of cities. These local situations prevent a clear identification of factors responsible for the high resistance levels detected.

Compared to the pre-LLIN 2005 collections, a significant four-fold increase of global kdr-w mutation frequency was observed in the 2006 wet season, around 7–8 months after the nationwide LLIN distribution, again followed by a similar four-fold increase between 2006 and 2007 wet seasons. Although suspected, the selective pressure exerted by the nationwide LLIN coverage causing the kdr increase within mosquito populations cannot be demonstrated (mainly due to absence of control zones). Even though it is at a quite low level, this fast and linear increase could potentially lead to high kdr frequencies within a few years. A similar trend was reported in An. gambiae S form populations from West Kenya 57 after four years of ITNs trials in a 200 km² area whereas no variation was observed in control zones without ITNs. Sharp et al. 58 also reported an increasing kdr-w frequency in M forms from Bioko Island (Equatorial Guinea) in response to an IRS programme. On the same island, Reimer et al. 41 found a high allelic frequency in urban and peri-urban areas one year after the beginning of residual pyrethroids spraying, that they compare to the absence of kdr alleles reported before the IRS programme 65. However, it should be noted that Berzosa et al. 65 analysed only ten larvae at each collection site, resulting in a high probability of missing kdr alleles present at low to moderate frequency. Indeed, a null kdr frequency within 10 individuals gives a 95% Confidence Interval upper bound of 16.8%, that means a 95% probability to find no kdr allele in the sample although present at a frequency below 16.8% in the population. A similar rapid kdr frequency increase was reported in Abidjan 252843 from no kdr alleles around 1998 (n = 30) to 39% around 1999 (n = 27) and 70% in 2004, with the last sample collected in an outdoor deltamethrin-spraying area (n = 103). Unfortunately the two first studies pre-dated the description of An. gambiae molecular forms so we cannot exclude an effect of variation in molecular form proportions on kdr frenquency. These studies and our results suggest a selection effect of large-scale insecticide-based control programmes on kdr mosquitoes. However, a recent study in Mali 34 showed increasing kdr frequencies in S forms from three villages between 1993 and 2004 that were attributed to agricultural and domestic pyrethroid use in absence of any wide-scale control programme. Also, under the hypothesis of a recent introgression of the kdr-w allele from S to M form in southern forest areas, we cannot rule out the possibility that the global increase we described is simply due to current spatial expansion of the mutation in West Africa.

Marked seasonality and a short rainy season condition the annual expansion of mosquito populations and the start of a new malaria transmission season. Each new malaria transmission season's increased mosquito abundance stimulates the Sahel inhabitants to use their impregnated bed nets. Indeed, the nationwide LLINs distribution campaign took place during the beginning of the dry season when usually very low mosquito densities are found throughout the sahelian zone. Our personal observations of low LLIN usage before the wet season (R. Labbo, unpublished) were confirmed by the recently published coverage and usage study 10 and are consistent with this low biting nuisance. This fact could have important implications in terms of potential selective pressure exerted by the LLIN, that would consequently be highly seasonal and co-occurrent with the fast increase of mosquito population size soon after the beginning of the rainy season. The long-term effect of such explosive patterns of mosquito populations dynamics and LLIN usage on insecticide resistance is unknown, but could potentially differ from more humid zones where populations are more stable along the year. Also, this delay between bed nets distribution and maximal usage rate (around six months after) could also have delayed the start of the resistance increase, and could explain the low kdr frequency in 2006 only a few months after the nets began to be used largely.

Concerning our finding of lower kdr frequency within resting compared to host-seeking mosquitoes, the most likely hypothesis is directly linked to the collection method. The resting collections used pyrethroids (0.25% tetramethrin/0.05% cyphenothrin/0.04% prallethrin) and could therefore have preferentially killed susceptible kds/kds females whereas at least a portion of kdr females could have escaped or stayed on walls and ceilings without being knocked down by the spraying. If this hypothesis was true, one should consider the collection method employed when comparing published studies or planning field collections as it could bias the kdr frequency values. In addition, It seems unlikely that the observed variation between collection methods is due to different sub-populations within the domestic environnement. Resting females may have taken a blood-meal one or two days before and constitute the same population that blood-seeking females collected by indoor landing catches.

Once kdr mutations are found in a population, one important issue would be to determine to what extent it could decrease the benefits of control programmes. The relationship between kdr genotypes and resistance phenotypes was partly unraveled by laboratory studies 2528384553 and experimental huts trials 1845, but we are far from fully understanding the role of kdr mutations on mosquito survival in the field. Therefore, we can only speculate on the kdr frequency required to measure a significant effect on insecticide resistance. Some studies give however interesting indications about the personal and/or community protection provided by insecticide materials or treatments in areas of kdr populations. Sharp et al. 58 reported a reduced infection rate after an IRS programme with pyrethroids on An. gambiae M form populations exhibiting around 40% kdr frequency, however the abundances were only decreased after a shift towards carbamate spraying. The recent study of N'guessan et al. 15, although conducted in semi-field conditions with experimental huts and artificially holed bed nets, gives interesting clues because the same experiments were run in two contrasting environments of kdr frequency in M form populations. The personal protection provided by the treated holed nets was good when kdr frequency was 6%, but was much decreased when kdr frequency was around 80%. In addition, in a peri-urban area of Abidjan, Côte d'Ivoire, with M form populations with a kdr frequency of 70% 43, a spatial spraying programme of deltamethrin and fenitrothion in conjunction with deltamethrin-treated bed net usage in a french military camp did not significantly reduce the mosquito biting rates.

Therefore, as the present results indicate still low kdr frequencies mainly in the heterozygous state, it is unlikely that the current spread of the mutation has to date any significant effect on resistance phenotype and resulting efficacy of LLIN. However, it is feared that this continuous spread of kdr-w mutation could rapidly impede current efforts to reduce malaria transmission by implementing large-scale pyrethroid-treated nets coverage.

These results are of prime importance in our effort to document multiple effects of operational control programmes on mosquito vectors, and to conceive sustainable control strategies for the future. As an increasing number of African countries plan to develop and scale up malaria control strategies including large vector control implementation, continued monitoring for insecticide resistance will be of utmost importance in a context of increasing extensive pyrethroid exposure. The documentation of the factors contributing to resistance selection within those populations is also highly important. Nevertheless, the long-term sustainability of such programmes will not be achieved without the availability of alternative insecticidal compounds that could replace pyrethroids for bed net impregnation where they would appear inefficient due to kdr mutations and/or other resistance mechanisms.

The authors declare that they have no competing interests.

CC carried out the molecular processing, the analysis and interpretation of data, and contributed to the drafting of the manuscript; RL participated to the conception of the study and the field samples collection and identification; IA participated to the collection and identification of the samples and the molecular processing; JBD conceived and coordinated the study, and contributed to the interpretation of data and the drafting of the manuscript. All authors read and approved the final manuscript.