Thermal limits of wild and laboratory strains of two African malaria vector species, Anopheles arabiensis and Anopheles funestus

Two long-established laboratory colonies held at the Vector Control Reference Unit in Johannesburg were used for all investigations of thermal tolerance. Anopheles arabiensis was taken from the KGB colony, originally established in 1975 from Kanyemba in the Zambezi Valley, Zimbabwe (R.H. Hunt, pers. comm.) and An. funestus from the FUMOZ colony established in 2000 from southern Mozambique[30]. These colonies are maintained under an insectary temperature of 25 °C (± 1 °C) and 80% relative humidity (verified using repeated measures with a Masons Hygrometer, Brannan, UK) with 12:12 light/dark cycle and 45 min dusk/dawn simulation. Larvae are fed a mixture of ground-up dog biscuits and yeast extracts and females are offered a blood meal three times weekly and allowed to lay eggs two to three times weekly. All adults are provided with a 10% sugar water solution ad libitum.

Anopheles arabiensis females were collected from Malahapanga in the Kruger National Park, South Africa (22° 53.23 S, 31° 02.22 E) in October 2010. Wild An. funestus females were collected from villages surrounding the Maragra Sugar Estate in southern Mozambique (25° 27.41 S, 32° 46.59 E) in April 2011. Adult anophelines were collected using active-search techniques from inside huts or houses or from indoor animal dwellings using a flashlight and 30 cm glass aspirator. Females were transported back to the laboratory within three days for egg-laying in polystyrene cups with rough surfaces at a density of 20 females per 250 ml and were provided with a ball of cotton wool moistened with 10% sugar water solution. Egg batches from these females were kept separate until positive species identifications of the wild adults were made using standard PCR methods[31, 32]. The progeny of at least 80 individual females was used to establish a laboratory colony of the wild strains, with the fifth to seventh generations being used in experiments on An. arabiensis, and the first generation used in experiments on An. funestus. Different generations were used as a result of the inherent difficulties associated with establishing An. funestus colonies compared with An. arabiensis colonies (R.H. Hunt, pers. comm.). These colonies were kept under the same conditions as the laboratory strains.

Three age groups for each of the laboratory strains were used. An. arabiensis adults were 10-, 15- and 20-day olds, while An. funestus adults were 10-, 20- and 30-day olds. These ages were chosen because of the different lengths of the gonotrophic cycle and different adult longevities of the two species[30, 33]. Only two adult age comparisons for the wild An. arabiensis strain (10- and 15-day olds) and wild An. funestus strain (10- and 20-day olds) were possible due to low colony numbers and the requirement to make assessments before 10 generations in the laboratory.

Between 20 and 40 individual males and females from all age groups were exposed to each of three acclimation treatments prior to CT determinations. Adult mosquitoes were acclimated for a period of five to seven days at 20 °C, 25 °C or 30 °C and a RH > 80% at either insectary conditions (25 °C) or using PTC-1 Peltier portable temperature control cabinets (Sable Systems, Las Vegas, Nevada, USA, 20 ± 1 °C and 30 ± 1 °C). Humidity in the insectary was checked using a Masons hygrometer (Brannan, UK). At 20 °C and 30 °C, humidity was maintained through the use of distilled water (checked using a Hygrochron i-button, DS 1923-F5, Maxim/Dallas Semiconductor, Sunnyvale, CA, USA). Each acclimation treatment was maintained on a 12L:12D cycle for the five or seven day period. Most insect species show acclimation responses in less than seven days[34]. Following these acclimation treatments, 10 individuals (comprising five individuals of each sex) per individual trial were placed into a double-jacketed insulated chamber connected to a programmable water bath (Grant LTC-12 Series, Grant Instruments, Ltd., Cambridge, UK). For each age group and acclimation treatment of each species, a total of four replicate trials were completed. CTmin experiments started at 20 °C while CTmax started at 25 °C, decreasing or increasing at a rate of 0.25 °C/min, respectively after an equilibration period of 10 min. While it has been shown that rate of temperature change can significantly alter the upper thermal tolerances of various insect species, the current rate was chosen as one comparable with many other studies[35]. The CTmin was regarded as the point where individuals displayed reduced motor function (ie, onset of spasms) and could not cling to the tip of a paint brush, while CTmax was regarded as the point where individuals displayed reduced motor function following a period of rapid flight[36]. At no point were individuals removed from the trial chambers for assessments of motor function (ie, individuals were continuously subjected to the thermal assay).

Lethal temperature (LT) determinations of larvae and pupae, most appropriate for less mobile stages[37], for both species were carried out on six groups of 10 individuals each, per life stage (n = 60 per exposure temperature). The plunging technique was used instead of a ramping protocol[37, 38]. Each replicate (ie, group of 10 individuals) was exposed for a period of two hours to temperatures ranging from −12 °C to 8 °C for LLT (lower lethal temperature) and from 34 °C to 44 °C for ULT (upper lethal temperature) in 2 °C increments to ensure that 0% and 100% survival of test individuals was recorded. A water temperature of 24 ± 0.5 °C was used as a control and survival at this temperature was 100%. Temperatures were maintained through the use of programmable water baths (Grant LTD-20 and GR150 R4 Series, Grant Instruments, Ltd., Cambridge, UK). Following the two-hour exposure, experimental groups were returned to water at 24 °C (± 1 °C) and survival was scored every 24 hours until either eclosion to adulthood or complete mortality occurred. Percentage survival was then scored as the percentage of the 10 individuals that eclosed. Larvae were fed daily on the same larval food as the colony strains.

Adult lethal temperature experiments were performed on five groups of 10 individual males and females each (n = 50 individuals per sex, per temperature), acclimated at only one temperature (25 °C, RH 80%). The three age groups (An. arabiensis: 10-, 15- and 20-days; An. funestus 10-, 20- and 30-days), were used in the upper lethal temperature and the LLT experiments, with the exception of the LLT determinations for An. arabiensis adults where only two age groups (10- and 15-day olds) were used due to unexpected mortality in the colony. Each replicate (ie, group of 10 individuals) was exposed to a given temperature in the range −6 °C to 16 °C for LLT determinations and 24 °C to 38 °C for ULT determinations, for a period of four hours to ensure that 0% and 100% survival temperature was measured for both LLT and ULT. This four-hour temperature exposure was chosen as an estimate of the length of time of the hottest period in the day, to which mosquitoes would be exposed, based on generalized daily temperature profile data which show that for many regions, including those of tropical Africa, high daytime temperatures are maintained for approximately four hours[39, 40]. Experiments were conducted in a SANYO incubator (MIR-154, SANYO Electric Co. Ltd., Osaka, Japan). A temperature of 25 ± 1 °C was chosen as a control and survival at this temperature was close to 100%. Adults were immediately removed from the exposure temperature following the four-hour period, given sugar water and left to recover at 25 °C (± 1 °C) and relative humidity of 80%. Survival was scored as the percentage of the 10 adults still living, 24 hours after the experiment concluded.

Normality and homogeneity of variances were examined using Shapiro-Wilk’s tests and Levene’s tests, respectively (Statistica v. 11, StatSoft, Tulsa, Oklahoma, USA). Some deviations from normality were observed, but the model assumptions were generally met (supplementary materials, Additional file1) and the sample sizes sufficiently large to allow for the use of parametric general linear models (GLM)[41], as implemented in R (v. 2.13.1) (R Foundation for Statistical Computing, Vienna, Austria). The first model examined the effects of age, acclimation, sex and strain on the variables CTmin and CTmax for each species. Because significant effects of strain or an interaction with strain were found for CTmin/max, models were then run separately for each strain incorporating age, sex and acclimation as predictor variables. As an estimate of effect size, the mean percent deviation in CTmin/max from the grand mean per group was calculated by subtracting from each factor mean, the grand mean, and dividing this by the grand mean, multiplied by 100 to obtain a percentage (Additional file2). The sign of this % deviation from the mean provides an indication of whether or not each factor had on average a lower (negative) or higher (positive) CTmin/max than the grand mean.

The mean (± S.E.) lethal temperature at which 50% of the sample population died (LT₅₀) for each species and life stage, in relation to age and sex (for adults) was determined through the use of logistic regression with binomial distributions (logit link) in R (v.2.13.1). Using Hochberg’s GT-2 method as described in[42], lower and upper 95% confidence limits for each group were calculated using the means and standard errors obtained from logistic regression analyses. Mean LT₅₀ (± 95% C.I.) for each group was plotted. Overlapping confidence intervals indicate no significant difference between groups.

No significant differences in CTmax were found between the wild and laboratory strains of An. funestus mosquitoes, and the interactions involving strain were generally not significant, except in a single case (Table 1, Figure 1). Only acclimation affected CTmax values significantly in both strains (Table 1), although the effect size was typically ≤2 °C, with the significant 3-way interaction between strain, age and acclimation not being clearly interpretable (Figure 1). However, it is clear that the overall acclimation response is less in the laboratory than wild strain in both males and females, explaining the significant two-way interaction between strain and acclimation. Clearly, some difference in the effects of acclimation, sex and age exists among strains and therefore the models were run separately for each strain (Table 2). Acclimation and age have much greater effects on CTmax in the wild than in the laboratory strains, with some differences in the interactions too. However, the total variation in CTmax was c. 3 °C (Figure 1, Additional file2). Generally, higher acclimation treatments resulted in higher CTmax, and younger adults and females tend to have higher CTmax values (Figures 1 and2). CTmin values differed between the An. funestus strains, which also showed significant differences in response to acclimation (Table 1). The strongest acclimation response was found in the colony strain and specifically in 10-day old males and females, whereas by comparison differences among other ages and among genders and acclimation treatments at other ages were much reduced (Figure 1). Maximum effect size of c. 4 °C was found following different acclimation treatments in 10-day old colony females (Figure 1, Additional file2). When the models were run independently for the two strains it became clear that in each strain, sex, age and acclimation temperature had significant effects, but in somewhat different ways among strains, with the effects tending to be most pronounced in the laboratory strains (Table 3). Across the full set of treatments, the maximum difference in CTmin was c. 6 °C (Figure 1).

In An. arabiensis, with the exception of two-way interactions between sex and age, and sex and acclimation treatment, as well as several three-way interactions between strain, sex, age and acclimation treatment, no other effects on CTmax were significant, and especially not the main effects in the model (Table 4). It does appear that females have higher CTmax values than males (Figure 2), but these effects were not readily distinguished in the full model. When the models were implemented separately for each strain, the sex effect was significant (Table 5, Figure 2), as was the effect of acclimation for the wild strain, largely reflecting the large effect of the 30 °C acclimation treatment on 15-day old males (Figure 3). By contrast, age and various interactions did not have significant effects on the laboratory strain. The overall range of CTmax values was c. 3 °C (Figure 3).

CTmin responded strongly to acclimation treatments, and age, sex and strain were also all significant in An. arabiensis (Table 4). The wild strain tended to have lower CTmin values than the laboratory strain, while 10-day old females in the laboratory colony showed the strongest response to acclimation (Table 6, Figure 3), just as was the case in An. funestus. In the wild strain, females tended to have a lower CTmin than males (Figure 2), and acclimation had a strong, generally linear effect on CTmin (Table 6, Figure 3). However, in the laboratory strain, although all of the main effects and interactions were significant (Table 6), the responses were non-linear among acclimation treatments, and the variation among age groups at a given acclimation ≤ 1.5 °C (Figure 3). Overall, among strains, ages, sexes and acclimation treatments the variation in CTmin was c. 5 °C (Figure 3, Additional file2).

Lower lethal temperature (LLT) in An. funestus was approximately −1 °C to −2 °C for all stages and age groups examined, with the exception of the larvae (mean ± 95% C.I., 1.94 °C ± 0.62 °C), and 30-day old adult males (mean ± 95% C.I., 0.68 °C ± 0.83 °C), which were less tolerant of low temperature (Figure 4). In An. arabiensis, the situation was similar, with larvae likewise showing the least tolerance of low temperatures (mean ± 95% C.I., 1.59 °C ± 0.71 °C), and adult males being the least resistant of all groups (10-day old males mean ± 95% C.I., 3.66 °C ± 0.98 °C; 15-day old males mean ± 95% C.I., 3.48 °C ± 0.83 °C). Lower lethal limits in the adults were generally 8-11 °C less than the CTmin. The full range of LLT values spanned c. 6 °C (Figure 4).

Upper lethal temperatures (ULT) across the full range of stages, ages and species varied by c. 11 °C. In both species, larvae and pupae had the highest ULT, with An. arabiensis having more tolerant immature stages than An. funestus (Figure 5). Females of both species tended to have higher ULT than males, with the most heat sensitive group being the males of An. arabiensis. The lethal temperature estimates were typically 8-10 °C lower than the CTmax estimates, indicating a much reduced scope for long-term tolerance of high temperature in the adults.