1. Many prey organisms exhibit adaptive phenotypic plasticity in life-history traits that facilitate a better chance of survival in the presence of predators. The evolution of such plastic traits requires that the defensive phenotype incurs a cost in the absence of predation.
2. Model systems that are used to examine predator-induced defences are often organisms with complex life histories that only induce defences during the larval stage. While many studies have detected costs of inducible defences during the larval stage, detecting the costs of larval defences after metamorphosis is also important.
3. We examine the benefits and costs of inducible larval defences in the urban mosquito, Aedes notoscriptus, by rearing them in the presence and absence of predation cues. We compared survival of larvae inducing behavioural defences, when exposed to predation cues, in predation trials with predatory fish Hypseleotris galii to that of larvae reared in the absence predation cues. We also compared life-history traits of predator-exposed larvae to larvae reared in control conditions.
4. Larvae exposed to chemical predation cues limited activity and were able to avoid predation for longer in trials with H. galii. However, predator-exposed larvae suffered retarded larval growth and development, were smaller at metamorphosis and less resistant to starvation as adults.
5. While it is difficult to understand the ‘fitness costs’ that poorer starvation resistance might confer to adult mosquitoes, it is likely that smaller adult size of predator-exposed individuals would reduce fitness, particularly for females where body size limits the size of blood meal they could take to facilitate egg production. We suggest that the demonstrable costs of inducible defences in mosquito larvae make this a good system for testing theoretical models for the evolutionary maintenance of adaptive phenotypic plasticity.
Predation is a strong selective agent that drives the evolution of defences in prey organisms. Defences can be a permanent part of the prey’s phenotype or inducible, switching on in the presence of predators and off in their absence. Organisms with inducible defences change their behaviour, morphology or physiology to increase their chances of survival in the presence of predators by increasing escape capacity (Havel & Dodson 1984; Nilsson, Bronmark & Pettersson 1995), limiting detectability (Juliano & Reminger 1992; Sih 1992; Werner & Anholt 1993), or reducing their time exposed to predators (Relyea 2007; Hagman et al. 2009). Some of the best-known examples of inducible defences are found in the larval stage of organisms that have complex life histories, transitioning from an aquatic larval stage to terrestrial at metamorphosis (Kohler & McPeek 1989; Peckarsky et al. 1993). Such systems of inducible defences are often used for empirical tests of theoretical models of how adaptive phenotypic plasticity persists in natural populations.
One of the critical requirements for the maintenance of adaptive phenotypic plasticity is a functional trade-off between alternate phenotypes across environments (DeWitt, Sih & Wilson 1998, DeWitt & Sheiner 2004, Hammill, Rogers & Beckerman 2008). For inducible defences in prey organisms, this means that the noninduced phenotype must have a fitness advantage over the induced phenotype in the absence of predators and vice versa in their presence. If one phenotype has an advantage in both environments, then plasticity should not be maintained, resulting in a fixed phenotype (Tollrian & Harvell 1999). The fitness advantage of the predator-induced phenotype in the presence of predators is well-established for most systems, as the induced individuals survive predation better (Juliano & Reminger 1992; Nilsson, Bronmark & Pettersson 1995, Van Buskirk, McCollum & Werner 1997). However, the fitness advantage of noninduced phenotypes in the absence of predators is often less clear. Behavioural and morphological responses to predators can reduce feeding time and energy acquisition or increase energy expenditure on defensive structures and escape behaviours. Energy invested in these responses, or not acquired due to them, reduces the amount of energy an organism can utilise for growth, survival and reproduction. Many studies on various taxonomic groups demonstrate costs of inducing predator phenotypes (McCollum & Van Buskirk 1996; Turner 1997, Van Buskirk 2000; Fraker 2008). For those studies on organisms with complex life histories, detecting costs of defensive phenotypes is mostly done in larval traits, with fewer studies detecting costs that carry over to the adult stage (Johansson et al. 2001; Stoks et al. 2006; Hagman et al. 2009).
Larvae of several mosquitoes adopt low-risk behaviours when exposed to chemical cues of conspecific predation (Juliano & Gravel 2002; Kesavaraju & Juliano 2004; Ferrari, Messier & Chivers 2007). Larvae limit motility and spend a greater proportion of time at the surface of the water column. Juliano & Reminger (1992) demonstrate the adaptive significance of low-risk behaviour in Aedes triseriatus larvae exposed to predation by larger predatory larvae of another mosquito, Toxorhynchites rutilus. Active A. triseriatus below the surface of the water column were preyed upon preferentially to those that were inactive and at the top of the water column. Juliano & Gravel (2002) further demonstrated that predation by T. rutilus can drive the evolution of lower activity levels in A. triseriatus, reinforcing that inactivity is adaptive for mosquito larvae in predator environments.
The domestic container-breeding mosquito A. notoscriptus is an urban pest throughout Australia. Aedes notoscriptus are competent vectors for Ross River virus and Barmah Forest virus and are the major transmission vector of canine heartworm, Dirofilaria immitis (Watson & Kay 1998, 1999). Females oviposit in natural habitats such as tree holes, fallen palm fronds, bamboo stems and rock pools, as well as in artificial containers such as tyres, roof-gutters, tins, bottles and domestic tanks. Aedes notoscriptus are an ideal model species to investigate the costs of inducible defences as they are amenable to rearing and laboratory techniques (Watson, Marshall & Kay 2000) and are known to induce ‘low-risk’ behaviours in the presence of predator cues.
Here, we determine (i) the benefit of the inducible defences in the presence of a predatory fish and (ii) the costs of behavioural defences in the absence of an actual risk of predation, during the larval and adult life stages. We test the idea that predator-induced responses in mosquito larvae can result in substantial costs during both larval development and to the adult mosquito. We used the firetail gudgeon, Hypseleotris galii, a native Australian freshwater fish common along the eastern coast of Australia as a model predator of A. notoscriptus. These fish are amenable to laboratory keeping and highly motivated predators in the laboratory setting.
Materials and methods
Colonies of A. notoscriptus were established at The University of Queensland from eggs sourced from colonies at the Mosquito Control Laboratory at the Queensland Institute of Medical Research (QIMR). The QIMR colony was established from wild caught mosquitoes collected in Brisbane, Australia in 2007. Each generation was hatched from eggs from several hundred blood-fed females that were able to mate with several hundred males to maximise the maintenance of genetic diversity in the colony. However, the level of genetic diversity in our colony is not known. Hatching was induced in a vacuum flask filled with reverse osmosis (RO) water with 0·15 g crushed Tetramin algae pellets (Tetra, Melle, Germany) placed on a vacuum for c. 1 h. The following day, larvae were placed into rearing pans filled with 3 L of RO water and fed c. 0·15 g of crushed algae pellets per day during the entire larval stage. Larvae were kept at a density of c. 150 individuals per pan to avoid effects of density on growth and development. Mosquitoes were placed in cups filled with RO water upon pupation and allowed to metamorphose into breeding cages (50 cm3). Adults were fed on 10% sucrose solution and blood fed on a single human volunteer each week to facilitate egg production. Eggs were collected in oviposition cups lined with scoured brown paper and filled with RO water conditioned with leaf litter. Eggs were allowed to dry for several hours then sealed in plastic ziplock bags with moist paper towel to avoid desiccation, then hatched a minimum of 4 days later to start the next generation of the colony. Colonies were kept at 25 ± 1 °C under an approximate 12L : 12D photoperiod in large plastic bags to maintain high humidity.
Benefits of Inducible Defences
Larvae were reared in either the absence (control) or presence (treatment) of chemical predation cues. All larvae were reared in 5-L tubs filled with 2 L of RO water conditioned with marine salts (c. 30 g per 100 L) for fish survival. There were 60 replicates (each 5 L tub) in each of the control and treatment groups. A fish cage (WLPET, China; 16 × 12 × 13 cm) with fine mesh was partially submerged within each replicate tub in both the control and treatment groups. One H. galii was placed into the fish cage in treatment replicates, while fish cages in control replicates remained empty (mass of largest fish = 0·41g). The fish within the treatment replicates were fed 5–10 3rd/4th instar A. notoscriptus larvae to initiate the production of chemical predation cues. Experimental A. notoscriptus were hatched on the day that H. galii were placed into the treatment tubs. The following day larvae were randomly assigned to control or treatment with two larvae placed on the outside of the breeder net in each replicate tub. Two experimental larvae were used initially to avoid loss of replication from incidental death. Experimental larvae were then fed 0·2 mL solution of crushed Tetramin algae pellets in RO water (0·4 g per 100 mL) daily. Hypseleotris galii in treatment replicates continued to be fed 5–10 4th instar A. notoscriptus larvae daily to maintain a high concentration of chemical predation cues.
All fish cages and fish were removed on day six of the experiment. Where two larvae had survived in a single tub, one was randomly chosen to remain in the experiment and the other discarded. Larvae were removed from the replicate tub and body length measured using an eyepiece micrometre calibrated with digital callipers. We then divided each tub into two compartments using a nontransparent divider and placed two H. galii on one side of the divider. Larvae were then returned to the vacant side of the divided tub and allowed to settle for a minimum of 5 min in the water in which they were reared (i.e. containing chemical predation cues for treatment individuals and without cues for control individuals). Each larva was then observed for 2 min to measure the total time active defined as any time spent in a thrashing motion. After the behavioural trials, the divider was gently removed to not disturb the fish or mosquito larvae. The H. galii were allowed to prey freely on the mosquito larvae for 20 min with larval survival checked every minute.
Costs of Inducible Defences
Larvae were again reared in control and treatment conditions identical to those described earlier. However, in this experiment, four experimental larvae were placed into each tub to minimise loss of replication because of incidental death and to maximise the chances that each tub will have at least one female and one male emerge as adults. Accordingly, we doubled the amount of Tetramin algae pellet solution added to each replicate (0·4 mL of 0·4 g per 100 mL solution) than in the experiment previously described. Larvae were checked daily until the first individual had pupated, at which time fine mesh was secured over the top of the replicate tub so that emerged adults could not escape. From that point forward, pupae were checked every 3 h for emergent adults. The first male and female from each tub were taken at the time of emergence and placed in a 20-mL vial secured with fine mesh. All others that emerged were discarded. For many replicates, only one sex emerged. Cotton wool was placed in each vial and saturated with water to prevent dehydration of the adult mosquito. Mosquitoes remained unfed throughout their adult life with survival checked every 3 h on a continuous cycle. Upon death, mosquitoes were placed in 2-mL Eppendorf tubes and dried in a desiccating oven. Both wings were then removed and affixed to glass microscope slides. Images of the lateral view of the wingless mosquitoes as well as images of each wing were taken using a digital camera (model B686CF; PixeLINK, Ottawa, ON, Canada) mounted on a dissecting microscope and connected to the digital imaging software program PixeLINK Capture SE 3.1 (PixeLINK). Measurements of thorax length and wing length were taken using digital imaging software (ImageJ; NIH, Bethesda, Maryland, USA) with the mean length of both wings taken as wing length.
A follow-up experiment was performed to determine the potential for differences in water quality between control and treatment replicates to contribute to any life-history changes observed. Ammonia content in five control and five treatment tubs identical to those described earlier was measured using a commercial aquarium ammonia test kit (Aquasonic, Wauchope, NSW, Australia), after 6 days. Ammonia content did not markedly differ between control and treatment tubs with ammonia accumulating in both groups. Four tubs in each of the control and treatment groups accumulated ammonia after the 6 days. However, none accumulated more than 0·1 p.p.m. of ammonia. The remaining single tub in each group did not accumulate any detectable amount of ammonia.
Larval activity was analysed using a Mann–Whitney rank sum test as data were not normally distributed. The effect of treatment on larval body length was determined by one-way anova. Larval survival time in predation trials with H. galii was compared using parametric survival analysis, which included treatment, activity and larval body length as main effects, to determine whether any differences between control and treatment groups in time to predation were because of different activity levels or differences in larval body length between treatment groups. The simplest model had an exponential distribution (a constant rate of predation over time) rather than the default Weibull distribution. Parametric survival analysis was also used to determine treatment effects on the time taken to reach metamorphosis with sex included in the model to test for sex-specific effects. Differences in thorax and wing length between control and treatment adults were determined using manova, again with sex included in the model. The effects of treatment and sex on adult size are interpreted for wing and thorax length separately from the univariate anovas. Uncorrected P-values are reported for each response variable as a Bonferroni correction would not change the significance of any main effects. Finally, time to starvation was analysed using parametric survival analysis testing for a treatment effect with sex and wing length3 included in the model to determine any sex-specific effects and to determine whether any difference in starvation resistance was attributable to mass differences between treatment and control groups. Studies suggest that species-specific mass–wing length regressions are better to estimate mass from wing length (Siegel et al. 1992; Siegel, Novak & Ruesink 1994); however, these data are not available for this species. The full model, with sex included, showed a significant interaction between treatment and sex (z = 2·274; P = 0·0230). Therefore, our final analysis of starvation resistance was performed separately for each sex. Starvation resistance in females required a post hoc analysis of the control and treatment individuals separately to determine the effect of wing length3 on starvation resistance. The Mann–Whitney rank sum test of activity data was performed in statistical software JMP, and all other analyses were run in R statistical software (R Development Core Team, 2008). Activity, larval body length, and wing and thorax length data are presented as the mean ± standard error. All survival analyses underwent stepwise model simplification to determine the minimum adequate model (Crawley 2007).
Aedes notoscriptus larvae exposed to predation cues spent 2·11 ± 0·46% of their time thrashing during the 2-min behaviour trial on day six, whereas control larvae were significantly more active, spending 7·48 ± 1·19% of their time thrashing (T1,94 = 323; P < 0·001; Fig. 1). While treatment larvae appear to survive for longer in predation trials with H. galii, the effect is only approaching significance (z =1·79; P = 0·0799; Fig. 2). Incorporating activity and larval body length into the model showed no effect of body length (z =1·12; P = 0·2642), but those that were less active survived predation for significantly longer (z = −2·00; P = 0·0458). None of the interactions between treatment, activity or body length were significant and were therefore removed from the model.
Control larvae grew faster than treatment larvae with control individuals 7·04 ± 0·11 mm in length on day six of the experiment, and treatment larvae only 5·09 ± 0·16 mm (F1,94 = 23·6, P < 0·0001). Control individuals metamorphosed at 178·03 ± 1·69 and 160·37 ± 1·43 h after hatching in females and males, respectively. In contrast, treatment individuals metamorphosed at 191·78 ± 3·52 and 171·01 ± 2·33 h after hatching in females and males, respectively. The delay in metamorphosis in individuals exposed to predation cues was significant (z =4·16; P < 0·0001; Fig. 3) with females metamorphosing later than males (z = −5·24; P < 0·0001) but no significant interaction between treatment and sex (z = −0·25; P = 0·803).
Differences between the treatment and control individuals were evident after metamorphosis in both their size and starvation resistance. Forty-seven males and 35 females emerged from the control group and 34 males and 27 females reached metamorphosis in the treatment group. Treatment larvae were smaller as adults (F2,125 = 69·00; P < 0·0001); males were smaller than females (F2,125 = 337·13; P < 0·0001), and the effect of treatment on size was different between sexes (Fig. 4; F2,125 = 5·16; P = 0·0070). Univariate anova allows scrutiny of each response variable independently. Treatment individuals had shorter wings than control individuals (F1,126 = 119·52, P < 0·0001; Fig. 4a) with the wing length of treatment individuals, 2·56 ± 0·01 and 2·44 ± 0·02 mm in females and males, respectively. In comparison, wing length in control individuals was 3·05 ± 0·01 and 2·82 ± 0·03 mm in females and males, respectively. The wings of males were smaller than females (F1,126 = 648·73, P < 0·0001; Fig. 4a), and the effect of treatment on wing size was greater in females than in males (F1,126 = 8·95, P = 0·0070; Fig. 4a). Thorax length was shorter in treatment individuals (1·25 ± 0·02 and 1·16 ± 0·01 mm in females and males, respectively) compared with control individuals (1·48 ± 0·01 and 1·38 ± 0·02 mm in females and males, respectively; F1,126 = 111·7726, P < 0·0001; Fig. 4b). Males were significantly smaller than females (F1,126 = 444·3972, P < 0·0001; Fig. 4b), but the effect of treatment on thorax length was not different between the sexes (F1,126 = 1·1845, P < 0·2785; Fig. 4b).
Starvation resistance data of adults were analysed separately for each sex as the interaction between sex and treatment was significant. Analysing separately allowed further analysis to determine the effect of wing length3 (as a proxy for mass) on starvation resistance. Individuals reared as control larvae died of starvation 79·97 ± 2·73 and 71·61 ± 2·21 h after metamorphosis in females and males, respectively, whereas treatment individuals died of starvation 49·71 ± 3·83 and 49·91 ± 1·96 h after metamorphosis, in females and males. The interactive effect between treatment and wing length3 on starvation resistance was not significant for males and was therefore removed from the model (z =1·363: P = 0·173). The final model for males, including only main effects of treatment and wing length3, suggests that treatment individuals and smaller individuals (shorter wings) were less resistant to starvation (treatment: z = −4·910; P < 0·0001; Fig. 5; wing length: z =3·190; P < 0·0013). In the final analysis of starvation resistance in females, the effect of treatment was significant (z = −3·76; P = 0·0002; Fig. 5) and the effect of wing length3 was not (z = −0·58; P = 0·561). However, the interactive effect of treatment and wing length3 was significant (z =3·323; P = 0·0009). A post hoc analysis of the control and treatment groups separately shows that time to starvation is independent of wing length3 in control females (z = −0·57; P = 0·571), but in treatment females, smaller individuals starved to death sooner (z =6·95; P ≤ 0·0001).
Inducible defences should benefit prey organisms in the presence of predators but be costly when predators are absent. Our data show that A. notoscriptus larvae reduce activity in the presence of chemical predation cues and, in doing so, are able to avoid predation by the insectivorous fish H. galii for longer. Our data also suggest that reducing activity in the presence of predators may be costly to the larvae. Larvae reared in the presence of chemical predation cues, but at no actual risk of predation, grew slower and metamorphosed later than those in control conditions. Those same larvae metamorphosed into smaller adults that were less resistant to starvation. For prey with complex life histories, this examination of the costs of inducible larval defences manifesting in adult traits is more indicative of lifetime fitness costs than those only shown in larval traits.
Reducing activity is a common strategy employed by prey organisms to avoid predation (Skelly & Werner 1990; Juliano & Reminger 1992; Lima 1998; Stoks et al. 2005a). Limiting activity reduces the likelihood of being detected by visual predators or encountering ambush predators. Juliano & Reminger (1992) showed that larvae of the mosquito, A. triseriatus, are more frequently captured by larvae of the predatory mosquito, T. rutilus, when ‘thrashing’ or active. The adaptive significance of reducing activity in mosquito larvae is best demonstrated by Juliano & Gravel (2002). They show that colonies of A. triseriatus subject to predator selection by T. rutilus during the larval stage have reduced basal activity levels of larvae after just two generations. Analysis of our data shows that the effect of exposure to chemical predation cues significantly increased the survival time of A. notoscriptus larvae in predation trials with H. galii. Incorporating activity and body length into the model suggests that this effect is due primarily to the level of activity of the individual larvae and is not because of differences in larval size.
Reducing activity to limit the chances of being predated is generally assumed to limit foraging time in prey organisms (Sih 1992; Werner & Anholt 1993; Turner 1997; Beketov & Liess 2007). Several studies have demonstrated that prey inducing antipredator defences spend less time feeding than control individuals (Turner 1997; Stoks et al. 2005a). Although we did not directly measure foraging time in our study, we suggest that A. notoscriptus reducing activity in the presence of predation cues would have spent less time foraging. Juliano & Reminger (1992) show that foraging (both browsing and filtering) is reduced in larval A. triseriatus mosquitoes when adopting low-risk behaviours in the presence of predation cues. In our experiment, the algae solution that A. notoscriptus were fed settled to the bottom of the rearing tub. The most efficient form of feeding in this scenario would require active foraging on the bottom of the rearing pan. We believe that by reducing activity, the rate at which A. notoscriptus larvae encounter food when reared in predation cues is also reduced.
The primary function of the larval stage for organisms with complex life histories is to acquire energy and nutrients to facilitate metamorphosis into the reproductive adult. Any interruption to larval feeding should then have an impact on larval growth and development as well as adult traits. While we are unable to test the link between activity/foraging and the life-history traits in A. notoscriptus at the level of the individuals, our data show that A. notoscriptus larvae grow and develop slower, are smaller at metamorphosis and are less resistant to starvation as adults, when exposed to predation cues. Other studies show that exposure to predation cues retards larval growth and development as well as reduces pupal mass and wing length of adults in other mosquito species (Hechtel & Juliano 1997; Beketov & Liess 2007). The simplest interpretation would suggest that reduced size of the adult is indicative of less acquisition of food during the larval stage. However, Stoks et al. (2005a), show that reduced size in adult damselflies when exposed to predation cues is not solely because of a reduction in activity but also because of a reduced capacity for larvae to assimilate food during digestion. It is possible that digestive physiology of A. notoscriptus could be affected in a similar way when reared in predation cues.
Our data suggest that the physiology of the adult mosquito is affected by larval exposure to predation cues. While starvation resistance of adult males was partially explained by differences in mass (wing length3), the effect of treatment remained significant. This suggests that some other mechanism is contributing to the reduced starvation resistance in treatment males. Presumably, those individuals exposed to predator cues during larval development had less stored energy after metamorphosis. Stoks et al. (2006) showed that damselflies exposed to predation cues as larvae have less stored fat after metamorphosis. It may well be a similar case for male A. notoscriptus exposed to predation cues as larvae. The cause of reduced starvation resistance in females, however, is less clear as starvation resistance was not significantly affected by wing length3 in control females, but was in treatment females.
To fully understand the costs of larval defensive phenotypes in metamorphic species, it is important to show that the costs carry over to the reproductive adult. The size and age of the adult at metamorphosis is thought to have a strong influence on fitness (Rowe & Ludwig 1991; Hechtel & Juliano 1997; Taylor, Anderson & Peckarsky 1998). Exposure to predation cues has been shown to reduce adult size and correlate with reduced fecundity or reproductive effort in several metamorphic species (Kohler & McPeek 1989; Hammill & Beckerman 2010). In A. notoscriptus, the adults feed on sugars and blood and are therefore not solely reliant on energy stores from the larval period. However, it is well-established in other mosquito species that larger female mosquitoes (with longer wings) produce more eggs and live longer, and that larger males produce more sperm (Briegel 1990; Armbruster & Hutchinson 2002; Ponlawat & Harrington 2007). We suggest that the smaller size in A. notoscriptus adults reared in predation cues is likely to limit their reproductive output and perhaps their ability to disperse (Briegel, Knusel & Timmermann 2001; Briegel 2003), but see the study by Maciel-De-Freitas, Codego & Lourenco-De-Oliveira (2007). Furthermore, if those larvae reared in predation cues do indeed have less stored energy after metamorphosis, we might also expect this energy deficit to influence survival and reproduction.
Here we show that A. notoscriptus induce a defensive behavioural phenotype that is beneficial in the presence of predators, but this benefit may come at a cost that is detectable at the adult stage. Several studies have gone further to demonstrate that reducing activity limits foraging time and that this is correlated with a reduction in adult size. To equivocally show that such costs are because of reduced activity requires a selection experiment. If populations (colonies) are subject to predation cues that induce costly defences and there is sufficient genetic variance in inducible defences, those individuals that have a limited response to predator cues should have a fitness advantage. If the costs are because of the reduction in activity directly, then populations constantly reared in predation cues (but at no actual risk of predation) as larvae should lose their capacity to reduce activity in response to predation cues after successive generations. Inducible defences in mosquito larvae offer an opportunity to ask such questions as the generation time of mosquitoes is sufficiently short, and mating and egg production are simple to induce.
We thank Kay Marshall from the Queensland Institute of Medical Research for assistance with colonising Aedes notoscriptus. We also thank Lesley Alton and Ben Barth for assistance with data collection. Constructive comments from Professor Thomas DeWitt and at least two anonymous reviewers improved the quality of this manuscript immensely. All experiments were conducted within the guidelines of The University of Queensland Animal Ethics and Welfare Committee, the Queensland Government Fisheries Act 1994 (Permit no: 95992) and the Medical Research Ethics Committee (Approval number: 2009001078).