In nature, multiple different environmental factors vary, potentially interacting in nonlinear and nonintuitive ways. Our factorial rearing experiment showed that different levels of food, water depth, and temperature produced interaction effects in age and size at emergence of the malaria vector mosquito An. gambiae s.s. (Tables 1 and 2; Figs. 5 and 8). These results demonstrate that how individual factors affect this species depends on the levels of other factors in the environment. This finding is consistent with other studies of mosquitoes and other insects (Colbo and Porter 1981; Lyimo et al. 1992; Leonard and Juliano 1995; Juliano 1996; Walker et al. 1997a; Baltzley et al. 1999; Agnew et al. 2002; Knight et al. 2004; Stillwell et al. 2007; Triggs and Knell 2011).
Population dynamics of insects are shaped heavily by both development time and adult size, and interactions affecting these life-history traits pose a challenge for modeling how populations of socially and economically important insects will respond to new environments. It is not feasible to experimentally assess life-history responses under all relevant climate and land-use scenarios; there are too many combinations of levels among too many factors. A more fruitful approach would be to identify and better understand the underlying causes of interactions. Such mechanistic insight would inform useful models of mosquito dynamics and malaria risk (Parham et al. 2012). How environmental conditions affect insect life histories is a consequence of how developing insects manage their growth and development. It is unlikely that mosquitoes and other insects maintain separate rules governing development for every possible set of conditions; more likely they respond to internal states that are determined by environmental conditions, physiology, and their evolved developmental programs. Thus we discuss our results in the context of energy budgets and the relationship between age and size.
Before proceeding, it is important to consider that the observed effects on age and size may have resulted from differential survival among treatment groups. There is good evidence for plastic development in An. gambiae s.s. and other mosquito species (Briegel 2003), and we generally interpret our results in the context of how larvae adaptively adjust development. However, survival was affected by food level, and both larval development time and body size show substantial heritability in An. gambiae s.s. (Lehmann et al. 2006). Thus, differential survival among phenotypes offers an alternate explanation of observed effects. However, if differential survival was wholly responsible for the observed treatment effects, the ranges of responses under low-food treatments would be subsets of the high-food treatments – and they are not. From a graphical inspection of the wing length data, the low end of the high-food distribution did not overlap with the low end of the low-food range. Similarly, the longest developmental durations under high food did not overlap with the extremes produced under low food. Furthermore, food level was the only factor found to significantly affect survival. Thus treatment effects can reasonably be attributed to plastic development.
Also, larvae reared individually in vials may poorly represent processes of natural systems. However, our experimental findings provide insight into how patterns of plastic growth and development can contribute to complex environmental effects.
Separate environmental variables
Food has a relatively large effect on growth and development because it provides the matter that gets incorporated during growth and the energy used in biological processes. Of the three experimental factors, food produced the greatest effects on age at emergence, wing length, and survival (Figs. 4, 7, and 2). Larvae with more food developed faster and reached larger final sizes. These effects were clearly evident in the overall pattern of age and size at emergence; in Figure 9, the high-food treatment groups all had distinctly greater mean wing lengths and lower mean ages than the low-food treatments. These results were consistent with other studies (Timmermann and Briegel 1998; Gimnig et al. 2002). Furthermore, from C. Phelan unpubl. data, higher food availability can produce as much as a 250% increase in body weight and 50% reduction in development time of An. gambiae s.s. In the factorial experiment, food also contributed to four of the six significant interaction effects (Tables 1 and 2).
Deeper water slowed down development and reduced adult body size (Figs. 4C and 7C). This pattern is the inverse of increased food. Previous publications suggest that An. gambiae s.s. feed exclusively at the water surface (Kaufman et al. 2006; Klowden 2007). However, larvae in culture spend substantial time at the bottoms of their rearing containers where food accumulates, and they have been shown to ingest this food (A.G. Hoi, unpubl. data). Furthermore, Timmermann and Briegel (1993) found that rearing An. gambiae s.s. in water deeper than 2 cm increased mortality, and in a study of larval diving behavior of An. gambiae s.s., Tuno et al. (2004) similarly found greater mortality with deeper water. Animals that dive use up energy and oxygen getting to depth (Leeuw 1996; Tuno et al. 2007). The experimental effects of water depth on age and size at emergence may reflect an energetic cost of diving to forage.
With respect to temperature, larvae in warmer water emerged earlier and smaller (Figs. 4B and 7B). These temperature effects differed from the food and depth effects in that age and size at emergence both changed in the same direction (increased or decreased together). This special effect of temperature is common across taxa and is reflected in the temperature–size rule (Angilletta et al. 2004). For poikilothermic organisms, higher rearing temperatures both speed up development and increase metabolic rate (Kooijman 2009). Temperature's effect on developmental rate distinguishes it from other factors. Higher temperature, like less food, produces adults with smaller wings. However, high temperature also speeds up development to reduce age at emergence, an effect that is inconsistent with a simple reduction in available energy.
Interactions among variables
Separately, the effects of each of the three experimental factors are straightforward, but statistical interactions show that development is context-dependent. Changes in a given environmental variable can have substantial effects at one level of a second variable but none, or opposite effects, at another. Furthermore, which life-history trait is affected, age or size at emergence, is similarly context-dependent.
Insight into some of the observed interaction effects is given by how the different factors affect larval energy usage and the nonindependent relationship between age and size. Under an L-shaped age–size reaction norm (see 'Introduction'), more energy for growth generally means earlier emergence at a larger size. An important consequence of this L-shaped relationship is that when growth conditions are good (i.e., energy is abundant such that larvae end up in the vertical part of the L) small differences in conditions among environments will yield responses in adult size but not time to emergence, which is minimized. Conversely, if growth conditions are generally poor (the horizontal part of the L), small differences yield a response in time to emergence but not body size, which is minimized. In other words, for medium-to-good conditions age at emergence is fixed at some minimum value, whereas for medium-to-poor conditions size becomes fixed, again at a minimum. These properties of the age–size reaction norm offer explanations of the food-by-depth interactions observed for both age and size at emergence. These and other interaction effects are discussed below.
For age at emergence, there was a food-by-depth interaction (Fig. 5A). Under low food, larvae emerged substantially later when they were in deeper water. Possibly, depth only affected days to emergence under low food because, under the L-shaped reaction norm, age is fixed at a minimum when growth conditions are generally good (under high food). Alternately, the food-by-depth interaction could be explained by bottom foraging. Diving in deeper water may require more energy because more dives (or longer) are needed to get food from the bottom when it is less abundant: individuals must do more to get more (Boyd 1997). Possibly, under high food this extra activity yields enough resources to compensate for the extra cost of foraging, whereas under low food it does not.
A food-by-depth interaction was also present for wing length, but it followed an inverse pattern to that of age at emergence. A difference in wing lengths occurred between depths only under high food (Fig. 8B). Again, this is likely because the L-shaped age–size reaction norm allows flexibility in size (i.e., wing length) only under nutritionally rich conditions. The effect of water depth was expressed by wing length only under high food because the high-food treatments fall out along the vertical (size) dimension of the L-shaped age–size curve.
Thus, to get a full picture of how water depth effects mosquito development across food levels both age and size at emergence must be considered; depth affects age at emergence under high food and wing length under low food. In the context of the age–size reaction norm, food-by-depth interactions can be explained in terms of supply and usage of metabolic energy.
For age at emergence, there was also an interaction between food and sex (Fig. 5B). Under low-food females took substantially longer to develop than males. A simple explanation of this is that females are generally larger than males (Fig. 7D) and under low food it takes longer to achieve this size difference.
For wing length, there was an interaction between food and temperature (Fig. 8A). Wing length was much smaller at high temperature than low, but only under low food. Temperature speeds up developmental rate and increases food requirements such that it can be considered equivalent to a transformation of time (Kooijman 2009). Under high food, larvae may have had enough energy to become as large as possible despite greater energetic costs at higher temperature. Under low food, however, larvae at the higher temperature may have been forced to adjust their size to the limited available energy.
There was another wing length interaction between depth and temperature (Fig. 8C). Under low temperature wing length was greater at low depth, whereas there was no significant difference at high temperature. This pattern lacks an obvious explanation. Experimental treatment levels for temperature and water depth had very similar overall effects on wing length (Fig. 7B and C), and it remains unclear why low temperature would produce a larger difference in wing lengths between the two water depths (or vice versa). Possibly the increase in rate of development from the higher temperature limited time available for growth, and because growth is a nonlinear process a longer developmental time under low temperature allowed for relatively large gains in size at the more favorable shallow depth.
Finally, for wing length there was an interaction between water depth and sex (Fig. 8D). In shallow water only, there was a large difference in wing length between sexes. Females are larger than males (Fig. 7D) except, it seems, when both are reared in deep water. This result is also difficult to explain. If this were a consequence of simple differences in energy costs between living in deep and shallow water there should be a similar food-by-sex interaction, but there is not. Perhaps an effect of body size on oxygen budgets is at work: a higher cost of being underwater for larger larvae might produce such a result (Verberk and Bilton 2011).
A temperature-dependent age–size reaction norm
Temperature appears to fundamentally differ from food and water depth in how it affects the L-shaped bivariate age–size reaction norm. In Figure 9, the positions of the four food-depth treatment level combinations are suggestive of two distinct L-shaped age–size curves within each temperature. We used polynomial regressions to test whether the data supported a model with separate curves for each temperature. A two-curve model (ANCOVA) explained the data substantially better than a single-curve one. If the age–size reaction norm was simple (Stillwell et al. 2007) – that is, if the slope or shape of the age–size curve were unaffected by changes across different variables – then a change in any given factor effectively adjusts a single universal meta-factor – “environmental quality” or net energy available for growth. A complex scenario, as these data for An. gambiae s.s. suggest, indicates context-dependence that needs to be unraveled to understand mosquito development under novel environments. Recall that only temperature caused age and size to both change in the same direction (e.g., lower temperature produced greater development time and larger size). Temperature's distinctive effects appear to extend to the age–size reaction norm. To explain L-shaped age–size reaction norms, Day and Rowe (2002) proposed a formal model with a minimum size threshold. Our results suggest that temperature adjusts the level of such a threshold. However, the eight treatment combinations from this study merely suggest such a pattern. This idea could be tested by rearing larvae across several food levels at two or more temperatures.