Aging partially restores the efficacy of malaria vector control in insecticide-resistant populations of Anopheles gambiae s.l. from Burkina Faso

Study site and mosquito collections

Blood-fed female Anopheles gambiae were collected between 0700 and 0900 h from the inside of houses from the village of Soumousso, located in the south-west of Burkina Faso (N 11° 00' 46", W 04° 03' 25"). A total of six collections were undertaken during August to November 2010. Soumousso lies within the heart of a large cotton-growing region where crops such as maize and rice are also grown. For control of agricultural pests, insecticides are sprayed up to six times per year between June and October and rotated among pyrethroids, organophosphates and carbamates. Routine insecticide resistance monitoring has been performed in this area since 2008 as part of a WHO/TDR multi-country project and resistance to all major insecticide classes available for malaria control has been reported [20]. No centrally organized malaria control is conducted in this region and insecticide treated net usage is low. The pre-dominant An. gambiae s.s. sub-species found in Soumousso during the early part of the malaria transmission season is An. gambiae s.s. S-molecular form, whereas the number of An. gambiae s.s. M-form and An. arabiensis increases towards the end of the rainy season [20].

Mosquito rearing and species identification

All mosquitoes were reared in the insectaries at the Centre National de Recherche et de Formation sur le Paludisme (CNRFP), Ouagadougou. Blood-fed female An. gambiae were placed individually in small cups for oviposition over a two to four-day period. Immediately after egg-lay, a single leg was removed from each F0 female parent and used directly for species identification following SINE-PCR which discriminates between S-form, M-form and An. arabiensis[28]. Mosquitoes which failed to lay eggs or could not be identified to species were removed from the study. Eggs and newly hatched larvae from each individual parent were then either pooled according to species status or raised separately as isofemale lines. Adult mosquitoes were provided with a sugar solution for feeding. Blood meals were provided to mosquitoes older than five days. Mosquitoes that constituted the older cohort (17-19 days old) used in the study had received two blood meals and completed at least two gonotrophic cycles.

Deltamethrin susceptibility tests

In order to demonstrate an age-effect in the progeny of field caught mosquitoes, a minimum of 100 three to five-day and 17-19-day old female mosquitoes from each of two pooled S-form populations were exposed to WHO 0.05% deltamethrin impregnated papers for one hour. Only An. gambiae s.s. S-forms were used as this was the predominant species during the study period. The number of mosquitoes 'knocked down' was recorded at the end of exposure and mortality scored 24 h after. Control assays were performed throughout the experiment with a minimum of 25 mosquitoes exposed to non-insecticide treated papers.

L1014F TaqMan assay

Genomic DNA (gDNA) was extracted from 12 An. gambiae s.s. S-form mosquitoes either exposed to deltamethrin (0.05%) or from the control assays (non-exposed). One percent of the total gDNA was used in a TaqMan allelic discrimination PCR described by Bass et al.[29] to detect the presence of the 1014F kdr allele

Long-lasting insecticide net assays

Three Permanet2.0^® LLINs were sampled from the village of Laye situated 35 km north-west of the capital Ouagadougou (N 12° 31' 35", W 1° 46' 21"). The nets had been deployed in the village for approximately 2 years prior to the study as part of the Burkina Faso NMCP and were exchanged for new LLINs with the householder upon retrieval. Cohorts of three to five- and 17-19-day old mosquitoes from individual isofemale lines were exposed to each LLIN in WHO cone assays. Batches of five mosquitoes from the two age-cohorts belonging to the same isofemale line were exposed to the same piece of LLIN for 3 minutes according to WHO guidelines [30]. Only isofemale lines, which produced more than 20 female adults were selected for testing to ensure sufficient numbers were available for bioassays on each age cohort. Knockdown was recorded 1 hour following exposure and mortality scored 24 h later.

LLIN evaluation

The LLINs were transported to the Liverpool School of Tropical Medicine (LSTM) to assess the insecticidal properties of the nets. A minimum of 24 females from a laboratory susceptible An. gambiae strain (Kisumu) were exposed for 3 minutes in WHO cone assays against each LLIN and mortality recorded 24 h later.

An estimate of the concentration of deltamethrin on each net was performed using HPLC. For each net, five random samples (30 mm in diameter) were cut from different areas of the net (total area of 35.5 cm²). The five pieces were cut into smaller pieces into a 50 ml Falcon tube, 40 μl of DDT 2 mg/m² was added as an internal standard (normalise extraction), and acetone extracted by vortexing 3 × with 5 ml acetone, pooled and evaporated to dryness under a stream of nitrogen. The insecticide residue was recovered in 1 ml acetonitrile and filtered with a 17 mm PTFE 0.2 μm syringe filter (Chromacol LTD, UK) before analysis. HPLC analysis was performed by injection of 20 μl aliquots onto a reverse-phase Dionex Acclaim C18 column (120Ǻ, 250 × 4.6 mm, 5 μ, Dionex, Camberley, UK). A mobile phase of methanol/water 90:10 was used at a flow rate of 1 ml/min^-1.

The quantity of deltamethrin remaining on the net (mg/m²) was calculated from the standard curve established with known concentrations of an authenticated deltamethrin peak detected using an Ultimate 3000 UV detector and analysed with Dionex Chromeleon software.

Bendiocarb IRS assays

Three houses from the village of Laye were sprayed with bendiocarb (200 mg/m²) on 27 August, 2010 using Hudson sprayers. Two houses were composed of mud walls and one built of cement. Between three and 8 weeks later, approximately 10 mosquitoes from each age cohort from each isofemale line described above were exposed to the bendiocarb sprayed wall for 3 min using WHO cone tests. A maximum of five mosquitoes were used per cone (two cones) and knockdown and mortality were recorded as described above. The same area of each wall was used to test the different age groups from the same isofemale line. Mosquitoes from both age groups were exposed to a non-sprayed wall in a different house as a control throughout all bioassays.

The above 3 min exposure experiment was designed to maximise the chances of observing age related changes in mortality. However, as the WHO protocol for evaluating IRS involves a 30 min exposure, we repeated the experiment in May 2011. Three houses were sprayed as described above and mosquitoes collected from Soumousso were used for the bioassays. The only difference in methodology was that mosquitoes were exposed to the surfaces for 30 min, as opposed to 3 min.

Statistical analysis

The exact 95% confidence intervals for mortality in bioassays were calculated according to the Binomial distribution using R version 2.12.2 for Windows [31]. A Fisher exact test (p < 0.05) was used to assess the statistical significance of differences in mortality between age groups.

Mortality from the LLIN and IRS isofemale experiments was modelled using binomial generalized linear mixed models (GLMM) with a logit link function in the R package "lme4" [32]. A separate model was derived for each experiment with the aim of estimating whether age had an effect on mosquito mortality (binary response variable) upon exposure to either a LLIN or a wall sprayed with bendiocarb. Mortality was primarily modelled as a function of three fixed factors; 'Age' (young (three to five days) or old (17-19 days)), 'Species' (An. arabiensis, An. gambiae s.s. S-form or An. gambiae s.s. M-form) and 'Treatment' (LLIN or sprayed wall labelled T1-T3). Furthermore, in both models, 'Family' was included as a random effect to allow the log odds of mortality to vary between families. Testing for interactions between categorical explanatory variables is problematic when there are zero frequencies within the contingency table (i.e. zero mortality within a subgroup). This results in parameter estimates whose likelihoods increase towards ± infinity so that maximum likelihood estimates do not exist [33]. For this reason, and to test for evidence of possible interactions, a small weighted constant (total = 0.5) was added to the mortality rates contributing to the zero cell count (no mortality) in the younger age-group exposed to LLIN#1 [33]. Akaike's Information Criterion (AIC) was used to provide a strength of evidence for each model tested [34]. The estimated odds ratio with 95% confidence intervals was calculated relevant to the control treatment. More detail on the modelling approaches can be found in Additional file 1. All GLMM analyses were performed in R version 2.12.2 for Windows [31]. The R-script is available from the author on request.

Mosquito collections from Soumousso

A total of 582 blood-fed Anopheles gambiae s.l. were collected from Soumousso and identified to species. An. gambiae s.s. S-form pre-dominated throughout all collections. Between August and September over 73% of all An. gambiae s.s. identified were An. gambiae s.s. S-form. The proportion of An. arabiensis and An. gambiae s.s. M-form was higher in October and November (Figure 1).

Deltamethrin susceptibility tests

Pools of An. gambiae s.s. S-form from approximately 50 families were exposed to deltamethrin (0.05%) in WHO susceptibility tests. This was repeated on two separate occasions (Assay 1 and Assay 2). In each experiment two age cohorts were exposed. The percentage mortality in three to five-day old mosquitoes ranged from 28.2% (N = 103, 19.7-37.9%) to 56.1% (N = 107, 46.2-65.7%). Mortality significantly increased in the 17-19-day old mosquitoes for both experiments (p < 0.0001) reaching 97.1% (N = 104, 91.8-99.4%) and 98.1% (N = 106, 93.4-99.8%) (Figure 2). The percentage knockdown after 60 min exposure was high throughout all testing (between 92.4 to 100%) and did not differ between the age-cohorts. No control mortality was observed throughout the experiment. This experiment on pooled mosquitoes was performed to confirm that the age dependent resistance we had observed in earlier laboratory studies [25] was reproduced on field populations. The numbers of M-form and An. arabiensis were too low to repeat this experiment using these species/forms.

A sub-sample of 17-19-day old S-form An. gambiae s.s. was analysed for the 1014F kdr allele from individuals dead following deltamethrin exposure and from non-exposed mosquitoes from the control bioassay. The frequency of 1014F was 95.8% in both the test and control samples (N = 12 for each test). The high frequency of 1014F, supplemented by recent data collected from Soumousso as part of the WHO/TDR project, suggests that 1014F is at a very high frequency in this population and therefore variation in 1014F frequency is unlikely to contribute to the differences in mortality seen here between age-groups.

Long-lasting insecticide net assays

The mortality of the isofemale An. gambiae s.s. from Soumousso varied between each of the LLINs used in the experiment and on the whole, the performance of the nets was poor. A boxplot showing the range of mortality within isofemale lines for each age cohort exposed to LLINs is given in Figure 3. All 12 of the isofemale lines tested on LLIN#1, and seven out of nine on LLIN#3, gave zero mortality when tested against the younger cohort (Figure 3). Against 17-19-day old mosquitoes, the performance was slightly better, with mortality ranging from 0-55.5% for LLIN#1 and 0-14.2% for LLIN#3. LLIN#2 performed better,, but even with this net, mortality rates ranged from 87.5% at best, to 0% at worst, against three to five-day old mosquitoes. The performance of each LLIN from the field was reflected in the mortality rates of the laboratory susceptible colony (Table 2). The concentration of deltamethrin on each of the nets varied between 1.0-2.0 mg/m² and is considerably lower than the original concentration on the net (55 mg/m²).

The relationship between age and resistance was not expected to differ between members of the An. gambiae complex and hence isofemale lines were not assigned to individual nets according to species. However, post hoc analysis did not indicate any evidence of a confounding effect of species (see Additional File 1 for more information).

According to the model, there was strong evidence that mosquito age had an effect on mortality. The older age group (17-19 days) were more likely to be killed on encountering the LLINs used in this experiment, compared with the younger cohort (three to five days) (Estimated OR = 5.28, 95% CI 2.81-9.92, p < 0.0001). It should be noted that this is a pooled estimate that assumes that the odds ratio is the same for each LLIN and species.

Bendiocarb IRS assays

To determine the susceptibility of An. gambiae s.s. to bendiocarb sprayed on the walls in Laye, three isofemale lines from Soumousso (An. gambiae s.s. M-form) were exposed in WHO cone tests for 30 min. All mosquitoes were killed, suggesting complete susceptibility to bendiocarb from this area and this was confirmed in WHO susceptibility tests against both M- and S-form An. gambiae s.s. (mortality 100% 24 h after 1 hour exposure, N = 50 for M-form, N = 25 for S-form).

There was strong evidence that the age had an effect on mortality and older mosquitoes (17-19 days) were more likely to be killed on exposure to the sprayed-walls used in this experiment (OR = 3.39, CI 95% = 2.35-4.90). There was little evidence that the species status of the isofemale line had any effect on mortality rates (Table 3).