DDT and pyrethroids share the same mode of action on the insect nervous system, targeting the neuronal voltage-gated sodium ion channels. Molecular characterizations have revealed that various mutations in the S1-S6 transmembrane segments of domain II of the sodium ion channel gene give rise to resistance to these insecticides in a number of insect species 123. In the major Afrotropical malaria vector, Anopheles gambiae sensu stricto (hereafter named simply An. gambiae), two point mutations at amino acid position 1014 of the voltage-gated sodium channel gene have been described, resulting in either a leucine-phenylalanine (L1014F) 2, or a leucine-serine (L1014S) 4 substitution. Both mutations have been shown to be linked with DDT and pyrethroid resistance phenotypes in field An. gambiae populations 2456789. A relatively small number of studies have been conducted to determine the epidemiological impact of kdr target site insensitivity on the efficacy of pyrethroid and DDT based vector control 910111213. Although a consensus has not been reached yet, some groups suggest that it may be associated with decreased efficacy of ITNs and pyrethroid based IRS 1011. In the latter case, programme staff were able to utilize carbamate insecticides, which have a different mode of action, and thereby overcome the resistance problem. Therefore, studies for determining the epidemiological impact and the extent of distribution of kdr mutations in An. gambiae populations are considered relevant for the design of insecticide-based control programmes and for informing the debate about the re-introduction of DDT into the vector control.

A number of studies, with limited geographical sampling, have detailed the distribution of kdr mutations in An. gambiae: most have either screened for the L1014F allele in west African countries 26141516171819202122, or the L1014S mutation in east Africa 4. A few studies have screened for the presence of both resistance alleles in Nigeria 5, Cameroon 23, Equatorial Guinea 24, Gabon 25, Uganda 26, Kenya and Tanzania 27, Malawi and Mozambique 28.

The picture that emerges from these studies is that the kdr mutations are not homogeneously distributed in the two molecular forms of An. gambiae (termed M and S), which are considered as incipient species 29. In fact in early studies, the absence of the L1014F allele in the M-form was considered one of the major pieces of evidence for a severe restriction of gene flow between the two forms 30. To date, the L1014F mutation has been reported in several S-form populations from western and western-central Africa 6183031 and recently from Uganda 26. The L1014S mutation, originally recorded in Kenya 427, has recently been found in populations from Uganda 26, Gabon 25, Cameroon 23 and Equatorial Guinea 24. In most of the few M-form populations surveyed, kdr alleles were seldom found, even in areas where the mutations occur at high frequencies in sympatric S-form populations 5162022. Where the L1014F allele was found in M-form individuals it was attributed to introgression between the two molecular forms 17213132, with the exception of two cases, Bioko Island (Equatorial Guinea) and Douala (Cameroon), where the L1014F mutation may have arisen independently in the M-form 192333. The L1014S allele has not been observed in the M-form although the number of individuals analysed was small 52324.

This manuscript details the first effort to analyse the distribution of the L1014F and L1014S mutations in An. gambiae populations from Sub-Saharan Africa west of the Rift Valley. It is envisaged that the data will serve as baseline for both researchers and, more importantly, for control staff who manage the deployment of insecticides, the efficacy of which could be compromised by the presence of kdr alleles.

Eight-hundred and two An. gambiae were kdr genotyped using the HOLA approach and 193 specimens from Gabon and São Tomé and Príncipe Islands were analysed by AS-PCR. The kdr genotype was successfully determined for more that 98% of the specimens.

Table 2 shows the kdr allele frequencies observed in the S-form samples (N = 628). The L1014F allele was present in all samples analysed, with the exception of samples from Kikudo (N = 46) and Saurimo (N = 15) in Angola. No difference in the L1014F frequency was observed between samples collected in Kedougou (Senegal) in 2001 and 2005 [chi-square = 0.57, degrees-of-freedom (df) = 1, P = 0.32]. The L1014S allele was found in samples from Cameroon, Gabon and north-western Angola (Cabinda). Observed genotypic frequencies were in agreement with the Hardy-Weinberg expectations in all samples with N>20.

All M-form samples genotyped (N = 354) were characterized by the exclusive presence of the wild-type 1014L allele, with the exception of the samples from Benin (Dassa, N = 5), Nigeria (Kobape, Olugbo, N = 41) and Cameroon (Kribi, N = 48), showing a frequency of 40.0% [95% Confidence Interval (CI) = 16.8–68.7], 19.5% (CI = 12.4–29.4) and 6.3% (CI = 2.9–13.0) of the L1014F allele, respectively.

In the two sampling sites where M and S forms were sympatric, L1014F frequencies in Nigeria were similar (16.3% in S and 19.5% in M-form) whilst higher frequencies were observed in the S-form in Cameroon (66.7% in S and 6.3% in M-form; Fisher's Exact Probability Test, P << 0.001).

The S-form specimens genotyped by both the HOLA and the PCR assay for the L1014F allele 30 were compared (N = 267). The results of the PCR genotyping were consistent with those of HOLA in 98% (125/127), 79% (27/34) and 89% (43/48) of 1014L/1014L, 1014L/L1014F and L1014F/L1014F genotypes, respectively. When the L1014S allele was present (sample from Cabinda, Angola), the PCR assay interpreted as 1014L/1014L all the 1014L/L1014S (31/31) and L1014S/L1014S (12/12) specimens, while L1014S/L1014F genotypes were interpreted either as L1014F/L1014F (13/15) or as 1014L/L1014F (2/15). Sequencing of the fragment of the domain II of the voltage-gated sodium channel gene containing the kdr codon from 18 of the specimens providing inconsistent results confirmed the interpretation obtained by HOLA, with a single exception (i.e. a 1014L/1014L HOLA-specimen which was 1014L/L1014F by both PCR and sequencing).

This study shows that kdr mutations are widespread in An. gambiae S-form in west and west-central regions of sub-Saharan Africa. The L1014F allele was constantly present from the western limits of the S-form geographic range in central Senegal to the sub-equatorial areas of north-western Angola. The L1014S allele was not found in S-form samples collected in western sampling sites (i.e. west of 10°W), while it was found together with the L1014F allele in western-central sites, in agreement with previous preliminary reports from Gabon 25, Cameroon 23, Equatorial Guinea 24, and Uganda 26. Since no departures from Hardy-Weinberg expectations have been observed in these samples, the significant L1014F/L1014S heterozygote excess reported in Gabon by Pinto et al 25 could likely be a transient and/or localized phenomenon, rather than the result of a possible selective advantage of carriers of both mutations in the heterozygous state. Further surveys will be required to ascertain the epidemiological impact of the co-occurrence of the two kdr mutations, at the level of both the individual and population, in the west-central African region.

Merging the above results with all the data available to date on the simultaneous occurrence of the two kdr mutations in S-form 232425262728, three main geographic areas can be distinguished based on kdr allele distribution and frequencies (Figure 1a): i) a western area, from Senegal to Nigeria, in which only the L1014F allele is present with frequencies >50% in most sites; ii) a west-central area from Cameroon in the north to Cabinda (Angola) in the south, extending eastward to Uganda, in which both kdr mutations are present at variable frequencies in most sites; iii) an area to the east of the Rift-Valley, where previous reports show L1014S frequencies below 10% in Kenya and where to the south both resistance mutations are likely absent.

The data obtained do not allow to determine whether the geographical differences in kdr allele frequencies among S-form populations is a result of non-uniform selective pressures due to variation in insecticide use, as recently hypothesized by Pinto et al 28. However, it is interesting to note that the data from Angola suggest an impact of human activities on kdr distribution. All samples from Angola were characterized by the wild-type 1014L/1014L genotype, with the only exception of an S-form sample collected in Cabinda, less than 100 km apart from the sampling site of Kikudo 37. Although it is not possible to completely rule out the presence of kdr alleles in the relatively small Kikudo sample (N = 46), as there is a 95% probability that an allele with a frequency of 3% or below would not be detected in the analysed sample (equivalent to mis-scoring one heterozygote genotype), the difference in the kdr allele frequencies between these sites is significant (chi-square = 67.9, df = 2, P < 0.001). This likely reflects the fact that the Kikudo area was involved in the civil war that ravaged Angola for the last 30 years, whilst Cabinda is located in an Angolan enclave within the Democratic Republic of Congo, which has been only marginally involved in the war, thus allowing for a larger scale use of pyrethroids against agricultural pests and/or malaria vectors 38.

This is the first regional survey of M-form specimens with respect to both resistance mutations. The results show the absence of the L1014S allele in all samples analysed and the presence of the L1014F allele in samples from a relatively restricted geographic region in the central part of Gulf of Guinea (i.e. Benin, Nigeria and Cameroon) (Figure 1b). Interestingly, the first report of the presence of L1014F in M-form was from Benin in 1998 39, where it was shown to have introgressed from sympatric S-form populations 21 and where it has recently reached frequencies around 80% 15. Inter alia, it can be hypothesized that the mutation is spreading from this region eastwards (i.e. Nigeria and later in Cameroon, where the L1014F frequency is 19% and 6%, respectively) and westwards in Burkina Faso 7. Moreover, the hypothesis of a further "possible subdivision within the M molecular form" 40 may further complicate the picture.

Moreover, the different distribution of the two kdr mutations between sympatric M and S populations could also reflect different ecological/behavioural traits between M and S-forms, that might promote different exposure to insecticide selective pressures. For example, the M-form may be more adapted to urbanized, man-influenced ecological settings, whereas the S-form tends to prevail in rural settings 2941, where a use of insecticides for agricultural purposes is expected to be greater.

From the technical point of view, the results highlight the high sensitivity and specificity of both the HOLA and AS-PCR approaches, which efficiently and simultaneously detected both kdr alleles, as already suggested on a smaller scale by Lynd et al 35 and Pinto et al 25. Moreover, the results reinforce the notion that genotyping kdr alleles only with a single primer set (i.e. either the Martinez-Torres et al 2, or the Ranson et al 4 methods), is an incomplete and possibly misleading analysis. For a full description of kdr resistance in a given region, it is essential to screen for both kdr alleles.

Although the results obtained refer to relatively few An. gambiae samples from a very large geographical range, they represent a first effort to analyse the overall distribution of the L1014F and L1014S mutations in sub-Saharan Africa, particularly west of the Rift Valley, where both molecular forms of this species co-occur in most of their range of distribution.

The results on S-form populations, combined with previously published data (Figure 1a), show a non-uniform distribution of kdr mutations throughout the continent: the L1014F is the only allele present west of 10°W latitude, while the two mutations co-exist in a wide geographic area in western-central Africa (ranging approximately from 6°N to 6°S).

The L1014F allele is present also in the M-form in an area comprised approximately between 5°W and 15°E, showing higher frequencies (>75%) in Benin and in the island of Bioko 1533, and lower frequencies (<20%) in Burkina Faso, Nigeria and Cameroon. The absence of the L1014S allele in the M-form samples analysed will serve as baseline data for further studies on the possible occurrence and dispersal of this mutation in this form (Figure 1b).

The overall picture strongly suggests that the origin and spread of kdr alleles in An. gambiae is an ongoing process. There is evidence that both kdr alleles have originated in the S-form through multiple mutation events 28, whereas the L1014F allele may have arisen in the M-form either through introgression with the S-form or through independent mutation events 192142. Insect control activities, such as the large scale use of pyretroid insecticides for agricultural purposes, and possibly for domestic protection, may contribute largely to this process, as suggested by Pinto et al 28, who hypothesized a possible link between the recent expansion of cotton production in West Africa and an increased selective pressure due to pyrethroid insecticide use. This highlights the need to carefully monitor the evolution and spread of kdr target site insensitivity with a sensitive and precise method for the detection of both alleles and through more specifically designed spatial and temporal transects. Such surveys should be accompanied with the molecular analysis of non-target site resistance mechanisms for An. gambiae. This would facilitate an analysis of the relative contribution of kdr mutations into the resistance phenotype. Given the large geographic area where the two mutations co-occur in the S-form, it is also extremely important to undertake insecticide resistance surveys to correlate kdr genotypes to DDT and pyrethroid resistance phenotypes. This would allow to ascertain the phenotypic effect of both kdr mutations, particularly when co-occurring in the same individual, and to evaluate their possible effect on the efficacy of ITNs and pyrethroid based IRS, thus allowing control programme managers to make informed choices about the chemicals they use for vector control.

FS carried out the molecular processing, participated in the analysis and interpretation of data, and contributed to the drafting of the manuscript; MC and EB helped with the molecular processing; JE and ID contributed to sample collections and molecular processing; AC, MJD and VP participated in the analysis and interpretation of data, and contributed to the drafting of the manuscript; FS and JP collected part of the samples and participated to the molecular processing and to the analysis and interpretation of data, and contributed to the drafting of the manuscript; AdT conceived and coordinated the study. All authors read and approved the final manuscript.