Desiccation tolerance as a function of age, sex, humidity and temperature in adults of the African malaria vectors Anopheles arabiensis and Anopheles funestus

Persistence of any natural biological population not being continually rescued by immigration is dependent on survival and reproduction. For insects, ambient temperature and water availability are the two key extrinsic factors influencing survival, with the latter being significant especially for smaller species (Benoit and Denlinger, 2010; Chown et al., 2011; Harrison et al., 2012). How survival is influenced by their interactions is not always clear (e.g. Hayward et al., 2001; Chown and Nicolson, 2004). Investigations of survival of dry conditions are typically undertaken at a given water content of the air and at a specific temperature to obtain an indication of the desiccation resistance or tolerance of a given species or population (Hoffmann, 1990; Gibbs and Markow, 2001; Gray and Bradley, 2005). By contrast, investigations of the effects of temperature × water interactions are uncommon (How and Lee, 2010; Kleynhans and Terblanche, 2011), and understanding interactions of this kind is generally considered a significant challenge in physiology (Chown and Nicolson, 2004; Gaston et al., 2009; Hoffmann, 2010).

Interactions among environmental variables are also unlikely to be consistent among age groups and sexes, given that responses to both temperature and water availability vary with age and sex (Bowler and Terblanche, 2008; How and Lee, 2010; Weldon et al., 2013). In several insect species, as they age, they become less tolerant of temperature extremes (e.g. Bowler and Terblanche, 2008; Lyons et al., 2012; Colinet et al., 2013). Differences between sexes can also be pronounced and significant, e.g. in the mosquito vectors Anopheles arabiensis and Anopheles funestus (Lyons et al., 2012) and in males of the tropical butterfly Bicyclus anynana (Dierks et al., 2012), although this is not always the case (e.g. Mironidis and Savopoulou-Soultani, 2010). Similarly, age- and/or sex-related differences in the survival of desiccation have been found in mosquitoes, Drosophila spp. and tephritid flies (Gibbs and Markow, 2001; Gray and Bradley, 2005; Benoit and Denlinger, 2007; Fouet et al., 2012; Weldon et al., 2013), although much variation among species is typical, and investigations of their interactions is uncommon. Survival time under dry conditions is also influenced by several other characteristics such as body size, associated with initial body water content and lipid stores, and reduced water loss rate (WLR) (Chown and Klok, 2003; Gibbs et al., 2003; Gray and Bradley, 2005). In addition, desiccation may be affected by variation in tolerance to water loss (i.e. body water content when the insect succumbs to death), although for insects this is typically not as significant as variation in other traits (Edney, 1977; Chown and Nicolson, 2004). In a similar manner, starvation resistance can be influenced by sex, age, strain or variation among individuals in size or lipid content, though again no general conclusions across insects are yet possible (e.g. Gibbs et al., 1997; Hoffmann et al., 2005; Ballard et al., 2008).

For these reasons, the influence of sex, age, temperature, water availability and their interactions on the desiccation tolerance of two of the most significant vectors of Plasmodium falciparum malaria in south-eastern Africa (Gillies and Coetzee, 1987), A. arabiensis Patton and A. funestus Giles were investigated here. Although only the females are vectors of the Plasmodium spp. parasites, population persistence is dependent on male and female survival. While efforts to control malaria are ongoing and new methods of control are constantly being developed and tested (e.g. Farenhorst et al., 2009; Munhenga et al., 2011), the disease remains a major public health concern (WHO, 2013). As climates continue to change, how malaria, and other vector-borne diseases, will be affected remains uncertain (Rogers and Randolph, 2000; Hay et al., 2002; Pascual et al., 2006; Githeko, 2009). Providing detailed physiological information at the species level will therefore improve forecast models, so reducing this uncertainty (Thomson et al., 2010; Buckley and Kingsolver, 2012; Woodin et al., 2013).

Mass (mean ± s.d. mass: A. arabiensis males 0.99±0.17 mg, females 1.88±0.57 mg; A. funestus males 0.52±0.08 mg, females 0.96±0.22 mg) and sex both had significant effects on survival time in A. funestus (Table 1; for averages for all groups see supplementary material Table S1). On average, females survived longer than males, and larger individuals survived longer than smaller individuals. As expected, survival was significantly affected by temperature and humidity. Older individuals of both sexes died more quickly than younger ones under the specific stress, although this was more evident at high humidity compared with low humidity (Fig. 1). The significant sex × relative humidity (RH) interaction indicates that the response of each sex was different across humidities (Table 1). In general, males died faster than females at all humidities. The significant RH × age interaction showed that the relationship between survival time and RH differs between age groups at each RH. However, across all RH treatments, younger individuals survived longer than older ones under the specific stress. The RH × temperature interaction suggests that the slopes of survival against temperature differed between humidity treatments, with individuals at the lowest temperatures and highest humidities having the highest survival, evident even in the oldest male and female age groups for this species (Fig. 2).

Mass, sex, RH, temperature and age all significantly influenced desiccation resistance in A. arabiensis (Table 1). As well as significant main effects in the models, there were also significant two-way interactions, between sex × mass, RH × temperature, sex × age, temperature × age, sex × RH and RH × age for A. arabiensis (Table 1). The sex × mass interaction showed differing responses amongst individuals of different mass in the two sexes (i.e. different slopes of the time–mass relationship). The RH × temperature interaction showed higher survival across all temperatures at 100% RH relative to 55% or 5% RH treatments and higher survival at lower temperatures (Fig. 3). Survival showed a steady decline with increase in temperature across all humidity treatments (Fig. 3). The RH × age interaction for A. arabiensis again indicated a different relationship between RH and survival time for different age groups, with survival of younger age groups being higher across all humidities, but in different ways. The temperature × age interaction showed that high temperatures consistently led to faster death, across all ages and both sexes, especially at low (5%) RH (Fig. 4). At high temperatures, older individuals of both sexes survive for less time than younger individuals, even at 100% RH (Fig. 4). This model also had significant three- and four-way interactions between variables, which were not easily interpretable.

Results from the generalized linear model of the effect of WLR and mass on the dependent variable, time, indicated that WLR and mass both significantly influenced survival time (desiccation tolerance) across all sexes, ages and temperatures for A. funestus at 5% RH (Table 2). WLR was most often more important in contributing to survival, given that it typically had the largest effect sizes, with lower WLR significantly increasing survival time. The positive estimate values for mass indicated that increased survival was associated with increased mass (Table 2).

For A. arabiensis, WLR again contributed most to survival across sexes, ages and temperatures at the 5% RH treatment (Table 3). Reduced WLR led to significantly longer survival (negative estimate value) while increased mass (when significant) led to increased survival (positive estimate) (Table 3).

Results from the generalized linear model for the effects of species, mass, RH and temperature on survival time of the oldest groups of each species showed significant interactions and main effect involving the species term (Table 4). On average, A. funestus survived longer than A. arabiensis females across all treatments (Table 4; supplementary material Table S1). The RH × species interaction suggested that slopes of survival time versus RH differed between species, with a steeper slope in A. arabiensis. Survival of females at different RH × temperature combinations was highest at low temperatures and high humidities and became steeper at high temperature and low humidity.

When comparing survival times of males between species, no significant species main effect was observed (Table 4). However, mass, RH and temperature all significantly influenced survival (Table 4; supplementary material Table S1). Higher RH and lower temperatures increased survival for males. The two-way interactions between mass × species, RH × temperature, species × temperature and species × RH were also significant in the model. The significant mass × species interaction indicated that mass/time slopes were different between males of each species. The species × temperature and species × RH interactions showed that the species responded differently to RH and temperature in terms of their survival times, with A. arabiensis males dying faster than A. funestus males under high temperatures and low humidities.

Comparisons among the wild and laboratory strains of A. arabiensis revealed that under both sets of conditions (25°C/5% RH and 20°C/55% RH) the 10 day old adults showed no significant differences in water loss or lipid content at death (supplementary material Table S2). By contrast, time to death, the most significant variable from a fitness perspective, varied significantly among the strains, as did initial wet mass and final dry mass on death (supplementary material Table S3). When initial wet mass was included as a covariate, time to death did not vary among strains or the sexes (supplementary material Tables S2 and S3), indicating that variation in time to death among sexes and strains is a consequence of variation in mass.

Knowledge of the duration of survival of mosquitoes under different combinations of temperature and humidity, and among ages and sexes, is important for understanding population dynamics under various conditions, including those of the dry season. As might be expected (Chown and Nicolson, 2004), high humidity and lower temperature favour survival in both vector species across all ages and for both sexes, with the form of this response indicating survival times at low humidities typical of mesic/hygric insects this size, but with the interspecific variation characteristic of this trait (e.g. Hood and Tschinkel, 1990; Benoit and Denlinger, 2010).

One major concern with the current study might be that the findings reflect the situation only with the laboratory strains, given that laboratory adaptation has been recorded in several, though not all, species for several traits [see elsewhere for discussion (Chown and Terblanche, 2006; Parkash and Ranga, 2014) and for differing extents of laboratory adaptation in Anopheles species (Huho et al., 2007; Lyons et al., 2012)]. We investigated the extent of variation among laboratory and wild strains of A. arabiensis for the traits examined here. No differences were found in water loss tolerated and lipid content at death among the strains, although they differed in size (both initial wet mass and final dry mass), as did the sexes. This size difference accounted for the difference among the sexes and the strains in time to death, the key fitness trait (see Chown and Nicolson, 2004). Thus, the results presented here are considered generalizable to the field situation, once mass is taken into account as can readily be done, and indeed should be done given its key influence on survival time (see Tables 2 and 3 for relationships). Nonetheless, it is important to note the size differences among the laboratory and wild strains, and that for thermal tolerance traits of this species, laboratory strains showed significant differences from wild strains, but these differences were not large enough in effect size to render thermal tolerance work in the laboratory irrelevant to the field situation (Lyons et al., 2012). Other work has also found differences among laboratory and wild strains of A. gambiae in size and in internal nutrient resources (Huho et al., 2007). The size differences in A. gambiae males are similar to those found here for A. arabiensis males (larger individuals in the wild strain), but A. arabiensis females varied in the opposite direction [Huho et al. did not investigate females (Huho et al., 2007)]. Such trait variation among laboratory and field strains, although often small in effect, may nonetheless be important and should therefore be the subject of further investigation to improve the translation of laboratory to field outcomes, given the many laboratory studies undertaken on these species.

Given the relatively small, and readily correctable, differences among laboratory and wild strains, the effects of age and sex, as well as the differing trial conditions, on water balance found can inform both fundamental understanding of water balance in this group and its application to field population dynamics. Differences between different age groups indicated a decrease in desiccation tolerance with an increase in age, similar to previous findings confined to females for A. arabiensis and A. gambiae (Gray and Bradley, 2005), and those of some other insect species, mostly species of Drosophila (Nghiem et al., 2000; Gibbs and Markow, 2001; Shahrestani et al., 2012). This reduced tolerance of older age groups suggests a senescence response in terms of desiccation stress, similar to senescence observed in thermal tolerance traits of A. arabiensis and A. funestus (Lyons et al., 2012) and in other insect species (Bowler and Terblanche, 2008). Such a response has not been documented across a wide variety of insect groups, except for several species of Drosophila, and even here variation is typically among early and late life and also among species and sex (e.g. Matzkin et al., 2007; Shahrestani et al., 2012; Parkash and Ranga, 2013; Aggarwal, 2014). Mechanisms that might underlie a decline in tolerance with age are thought to include changes in WLR, probably a function of variation in cuticular hydrocarbon content and composition, melanization, differences in initial body water content and tolerance of dehydration (Nghiem et al., 2000; Gibbs and Markow, 2001; Benoit and Denlinger, 2007; Weldon et al., 2013). Here, we found a pronounced difference in the influence of WLR on overall survival time among age groups, suggesting that a change in cuticular resistance with age lies at the heart of the differences. Work showing that cuticular hydrocarbon amount and composition can change substantially with circumstances in Anopheles mosquitoes (Caputo et al., 2005; Reidenbach et al., 2014; Wagoner et al., 2014) bears out this idea. Irrespective of its cause, including age-related variation in tolerance of abiotic conditions is important for improving mechanistic models of population dynamics in mosquitoes (Styer et al., 2007; Beck-Johnson et al., 2013).

Survival and persistence of malaria vector populations is determined not only by surviving females but also by the presence of males. However, in most mark–recapture studies, females are often the sex shown to persist over several months (e.g. Omer and Cloudsley-Thompson, 1970; Lehmann et al., 2010). The results from the present study clearly demonstrated a pronounced sex effect on desiccation tolerance in both A. arabiensis and A. funestus. In both cases, females survived significantly longer than males under all combinations of temperature and humidity, consistent with the expectation that the females, which overwinter, should have higher innate desiccation resistance. Our data revealed that both mass and WLR have significant influences on survival time. In consequence, two mechanisms might explain sex-related differences. First, cuticular WLR, which generally dominates loss (Chown et al., 2011), differs among the sexes, as confirmed by pronounced effect size differences between males and females for the influence of WLR on survival. In turn, these differences are likely to be a consequence of variation in cuticle lipid content and/or composition, or cuticle thickness (Parkash and Ranga, 2013; Parkash and Ranga, 2014). Although some evidence exists for thicker cuticles in female mosquitoes (Wood et al., 2010), it seems likely that variation in cuticular lipid amount and composition is more likely to account for variation in WLR (Reidenbach et al., 2014). Second, larger mass in females means higher water content, which would further contribute to enhanced survival time in females (see also Reidenbach et al., 2014). Although higher body mass should also mean higher metabolic rates, and hence greater respiratory water loss (Chown et al., 2011), no evidence is available to indicate to what extent this mechanism may offset the two others, resulting in greater tolerance in females.

Overwintering females, by contrast to those examined in this study, are likely nulliparous (Omer and Cloudsley-Thompson, 1970), and probably have a reduction in blood feeding, associated with the reduction in metabolic demand in this season, thus increasing survival (Huestis et al., 2011). Along with the tendency of unfed mosquito adults to seek cool, humid refugia (Kessler and Guerin, 2008), this would probably explain their survival in the field during the dry season for several months (Omer and Cloudsley-Thompson, 1970; see also Lehmann et al., 2010), substantially longer than survival of any individual in the present study. Furthermore, overwintering females could be under strong selection for low WLR or high water storage [both known mechanisms of desiccation tolerance (Chown and Nicolson, 2004)]. The individuals used in this trial were provided constant access to sugar water and offered blood three times weekly. Hence, their metabolic rates may exceed those normally expected of overwintering females, which could lead to significantly faster death under desiccating conditions (Huestis et al., 2011).

Although A. arabiensis is traditionally thought of as the more drought tolerant of the two vector species investigated (Gillies and Coetzee, 1987; Lindsay et al., 1998; Gray and Bradley, 2005), this species was shown consistently to survive for shorter periods than A. funestus under different combinations of temperature and humidity. Anopheles funestus is typically a behaviourally flexible species and occurs in a wide range of habitats and climatic conditions (Sinka et al., 2010). It is a highly anthropophilic species, although it does exhibit behavioural changes to this pattern in some regions (e.g. Muriu et al., 2008; Sinka et al., 2010). These differences in behaviour between populations of the same species may have led to differences in desiccation tolerance observed for A. arabiensis and A. funestus. Additionally, A. funestus adults have also been shown to be more tolerant of high temperatures than A. arabiensis adults (Lyons et al., 2012). In the present study, A. funestus survived for longer under desiccation trials than A. arabiensis, largely because of its reduced WLR, though, once corrected for size, the differences between the species (especially in survival time) are much less pronounced. Indeed, our analyses show (Table 3) that the interspecific difference disappears for males and is only marginally significant for females with relatively small effect sizes. Aestivation by adult mosquitoes over the dry season is one mechanism by which mosquitoes are thought to survive in areas where malaria has a seasonal transmission (Huestis et al., 2011). In some regions, A. funestus is thought to be more important in the persistence of malaria throughout a dry season than A. arabiensis or A. gambiae (Charlwood et al., 2000).

A further possible reason for these species differences might be the influence of laboratory adaptation on desiccation tolerance. Some physiological traits are known to be more affected by laboratory conditions than others (Hoffmann et al., 2001; Huho et al., 2007; but see Terblanche et al., 2006; Parkash and Ranga, 2014). Our data suggest that the maintenance of A. arabiensis under colony conditions over several decades has had an effect on body size, which in turn affects survival time, but that other traits such as water loss tolerated are not affected. Moreover, the 12.5 h survival time of 10 day old females at 25°C was indistinguishable statistically (P>0.05 by t-test) from the 13.4 h found by Gray and Bradley (Gray and Bradley, 2005) for the same strain investigated at similar temperatures (28°C) but maintained under different conditions for 37 generations. Alternatively, among-population variation in the two species could account for the unexpected finding of lower desiccation resistance in A. arabiensis than expected, given that population-level variation in physiological traits is commonly found (e.g. Hoffmann and Harshman, 1999; Terblanche et al., 2006; Rocca et al., 2009; Simard et al., 2009). Anopheles gambiae s.s., the sister taxon to A. arabiensis, exhibits chromosomal polymorphic inversions, one of which confers an advantage on the species in arid environments (Coluzzi et al., 1979; Gray et al., 2009; Rocca et al., 2009). Anopheles funestus has also recently been shown to consist of different chromosomal polymorphic inversion forms occurring in different regions of the African continent (Guelbeogo et al., 2009; Sinka et al., 2010). Furthermore, A. arabiensis in Sudan has been shown to survive during the dry season where temperatures spike to over 50°C (Omer and Cloudsley-Thompson, 1970), in contrast to lethal temperature estimates of only ~34°C over a 4 h period for the southern African strain of this species (Lyons et al., 2012). In addition, the A. arabiensis colony was originally collected from a population occurring in the Zambezi River Valley close to a permanent tributary of the Zambezi River (R. Hunt, personal communication). This original population is unlikely to be as desiccation tolerant as populations from more arid habitats based solely on their proximity to this humid refuge. The existence of different phenotypes of A. arabiensis probably accounts for the seasonality of malaria in certain regions (Hay et al., 1998; Tanser et al., 2003) and the overwintering of females in some populations (Taylor et al., 1993; Huestis et al., 2011). In consequence, further investigations should focus on among-population variation in desiccation resistance in both species. Importantly, though, the direction of evolution cannot be inferred from our investigation of the two species (see Garland and Adolph, 1994). Thus, for any evolutionary conclusions to be reached, differences among the species and among populations will have to be examined in a broader comparative context.

To survive throughout the dry season, it is clear that both species must seek out refuges, and that additional downregulation of water loss is likely to occur. The former is in keeping with what is known of the behaviour of Anopheles species (Kessler and Guerin, 2008) and suggests that a trapping method based on humidity manipulation might be developed, or at the very least where other control methods might be targeted. The latter provides grounds for further investigations of whether such downregulation takes place, whether it can be induced under laboratory and/or field situations. To date, investigations of overwintering have largely met with little success, but promising data are now starting to appear (Lehmann et al., 2010; Huestis et al., 2011). A key new set of work should investigate whether an aestivation response (Hahn and Denlinger, 2011) can be elicited, what physiological mechanisms might be involved, and what the population dynamics consequences thereof might be. Alternatively, pockets of individuals displaying a greater tolerance for desiccation may also be able to persist through the dry season; and indeed, this seems to be the case, especially given the seasonality of malaria in some areas (e.g. Patz et al., 1998; Charlwood et al., 2000; Tanser et al., 2003; Ndiath et al., 2012).

Nonetheless, the current work provides information on the way in which different intrinsic and extrinsic factors interact to determine the survival of adults of A. arabiensis and A. funestus. Such information can considerably improve population dynamic models of these vectors and of the likelihood of malaria under a range of conditions (e.g. Mordecai et al., 2013; Beck-Johnson et al., 2013), especially given that even recent models consider adult mortality parameters uncertain (e.g. Tompkins and Ermert, 2013). Thus, in conjunction with information on thermal effects on development in the immatures and survival in adults (Lyons et al., 2012; Lyons et al., 2013), this information will improve current malaria forecasting abilities, especially in southern Africa.

Anopheles arabiensis mosquitoes from the KGB-strain, originally established from individuals caught in the Zambezi River Valley in 1975, and A. funestus mosquitoes from the FUMOZ-strain, originally established from individuals caught in Mozambique in 2000, were used for desiccation tolerance and starvation resistance experiments. Prior to experiments, colonies were maintained at insectary conditions (25±2°C, ±80% RH, 12 h:12 h light:dark cycle), checked with a Masons Thermohygrometer (Brannan, Cleator Moor, Cumbria, UK). During this time, all mosquitoes were provided with a 10% sugar water solution and females were offered a blood meal three times weekly. Three age groups for each species were used during desiccation resistance experiments. Age groups for A. arabiensis were 10, 15 and 20 day old adults, while ages for A. funestus were 10, 20 and 30 days old. Ages differed between the species because A. funestus adults are typically longer lived than those of A. arabiensis (Hunt et al., 2005; Munhenga et al., 2011).

Because laboratory populations may show adaptation to this situation (reviewed in Chown and Terblanche, 2006) [for Anopheles see Huho et al. and Lyons et al. (Huho et al., 2007; Lyons et al., 2012)], we sought to explore the extent of differences between the laboratory populations and a wild population in A. arabiensis, to which we had ready access in the field. The field population was collected from Malahapanga in the Kruger National Park, South Africa (22°53.23S, 31°02.22E), and experiments were undertaken (as described below) on the F1 generation [for further details of field collections and animal maintenance see above and Lyons et al. (Lyons et al., 2012)].

Individual males and females from each of these age groups were exposed to different combinations of three RH treatments and three different experimental temperatures. The lowest humidity, ~5%, was maintained through the use of silica gel, the ~55% treatment through saturated Mg(NO₃)₂ solution (Winston and Bates, 1960), and the ~100% humidity treatment (i.e. a starvation assessment) through the use of double-distilled water. Temperature was controlled using PTC-1 cabinets (Sable Systems, Las Vegas, NV, USA) or a SANYO incubator (MIR-154, SANYO Electric Co. Ltd, Osaka, Japan). Temperature and humidity were checked using hygrochron i-buttons (DS 1923-F5, Maxim/Dallas Semiconductor, Sunnyvale, CA, USA) accurate to 5% RH and 0.5°C.

At the start of each trial, each individual mosquito was anaesthetized by brief CO₂ exposure (<10 s) so that initial mass could be obtained to the nearest 0.0001 mg (using a Mettler Toledo UMX2 microbalance, Greinfensee, Switzerland). Each individual was then placed into a clear, double open-ended 10 ml vial, closed on either end with 1 mm gauze mesh. Each of the 20 vials containing an individual mosquito was placed into one of four replicate clear containers (230×160×100 mm) containing the silica gel, Mg(NO₃)₂ solution or distilled water (four replicate containers of each at each temperature). Each replicate container was then sealed and placed at one of three temperatures (20, 25, 30°C) with a 12 h:12 h light:dark cycle. Mosquitoes were checked every 2–3 h for the first 24 h and then every 6 h until death or visible knockdown without any sign of recovery occurred. Following each experiment at each temperature, containers were opened to remove mosquitoes that were visibly knocked down and showed signs of desiccation stress. Knocked down or dead mosquitoes were weighed immediately after removal from the experimental conditions to obtain a wet mass at death for each individual. The difference between initial mass and wet mass at death provided an indication of mass lost, which was attributed to desiccation, acknowledging that some mass was lost via metabolism [see Chown and Nicolson (Chown and Nicolson, 2004) for a discussion of various estimates of water loss]. Dividing this difference by the time each mosquito took to die provided an indication of WLR (in mg h⁻¹). In total, 1120 and 1060 A. funestus and A. arabiensis individuals, respectively, were used in these trials.

To determine the influence of sex, mass (mg), RH (%), age (days) and temperature (°C), on desiccation resistance (response), measured as survival time (h), a generalized linear model with quasipoisson error distribution [to correct for overdispersion (Crawley, 2007)] and log link function was implemented in R (v. 2.15.1; R Foundation for Statistical Computing, Vienna, Austria), for each species. Because data were zero-bounded on the left and showed positively skewed distributions, the quassipoisson error distribution was chosen (Crawley, 2007). Temperature, age and RH were used as ordered factors. The highest order interactions were removed from the model sequentially if they were not significant, so results present the minimal adequate model (Crawley, 2007). To graphically present the influence of age and temperature on desiccation resistance (survival time) of each species, a distance-weighted least squares 3D contour plot using the residuals from a regression of time on mass, against temperature and age was plotted for each species. No residuals were used for statistical analyses, only for graphical representations of the interactions among the factors.

To determine the relative contributions of WLR and initial mass (body size) to desiccation resistance of A. arabiensis and A. funestus, a generalized linear model with quasipoisson distribution of errors (to correct for overdispersion) and log link function, using survival time of each group as the dependent variable and WLR and initial mass as the independent variables was implemented in R (v. 2.15.1). This analysis was performed only for the 5% RH treatment per temperature, age and sex category. Only one humidity treatment was chosen to provide an indication of possible mechanisms underlying desiccation resistance in these species.

To graphically present the influence of temperature and humidity on survival time for both species, the residuals from a mass versus time regression of the oldest groups for both species (20 days old for A. arabiensis and 30 days old for A. funestus) were plotted on a distance-weighted least squares 3D contour plot against temperature and RH.

How desiccation resistance (measured as survival time) compared between species was determined through the use of a generalized linear model with quasipoisson distribution of errors (Crawley, 2007) and log link function implemented in R (v. 2.15.1). Only one age group (the oldest for each species) was chosen for this statistical comparison, because of different longevities experienced by these species (Hunt et al., 2005; Munhenga et al., 2011). Initial mass (mg) was included as the continuous predictor in the model, owing to substantially different masses between these species. Sexes were analysed separately and RH and temperature were input as ordered factors.

To understand potential differences between field and wild strains of A. arabiensis, the following comparisons were made. First, using the methods described above, initial mass, time to death, dry mass at death, water loss tolerated (measured as the difference between initial wet mass and wet mass at death) and residual lipid content (measured as dry mass at death minus lipid free dry mass at death, with the latter determined after breaking individuals up and maintaining them for 72 h in a 1:1 chloroform:methanol solution exchanged every 24 h and then drying to constant mass) were determined for wild and laboratory populations. This was not done for the full suite of interactions, owing to limited numbers of wild-caught individuals, but rather for 25°C and 5% RH, a relatively extreme temperature × humidity treatment, and for 20°C and 55% RH, a less extreme situation, for 10 day old adults. Much of the data for the laboratory strain were drawn from the original experiments.

Generalized linear models (as above, conducted in R v. 3.0.2) were then used to examine the effects of population and sex on initial mass (to assess size differences among the strains), water lost and residual lipid content (as a measure of desiccation tolerance), and time to death (an integrated fitness measure). The analyses were then repeated, where significant differences among strains were found, to investigate the impacts of initial mass on these differences (using initial wet mass as the covariate).

Thanks to staff and students of the Vector Control Reference Laboratory of the Centre for Opportunistic and Hospital Infections, National Health Laboratory Service (NHLS) in Johannesburg for colony maintenance and upkeep. Four anonymous reviewers made very helpful suggestions for improvements.