Insect herbivores can choose microclimates to achieve nutritional homeostasis

Temperature and nutrition greatly influence life history outcomes for ectotherms through their effect on body size and age at maturity (Angilletta, 2009; Berrigan and Charnov, 1994; Stearns, 1992). These interactions are generally thought to be mediated by the effects of temperature on metabolic rate, which alters energy requirements. For example, growth efficiencies can be reduced with increasing temperatures as energy is diverted away from growth to fuel heightened metabolic processes (e.g. Angilletta, 2009; Kingsolver and Huey, 2008; Miller et al., 2009). However, this classic concept of thermal effects on growth based on energy or caloric budgets may be incomplete because it does not consider the type or balance of nutrients being ingested. Extensive research demonstrates that macronutrient balance, rather than energy per se, is a primary determinant of rates of development and growth in animals (Raubenheimer and Simpson, 1993; Raubenheimer et al., 2009; Simpson and Raubenheimer, 2000; Simpson and Raubenheimer, 2012; Simpson et al., 2004). For many herbivores, including locusts, the uptake of the macronutrients protein and carbohydrate influences growth and body mass the most (e.g. Behmer, 2009; Simpson and Raubenheimer, 2012; Simpson et al., 2004).

For many chewing herbivorous insects, final body mass does not vary with temperature in a consistent way, and temperature effects are host plant specific (e.g. Diamond and Kingsolver, 2010; Mousseau, 1997; Walters and Hassall, 2006). For example, Diamond and Kingsolver (Diamond and Kingsolver, 2010) found that the tobacco hornworm, Manduca sexta, grew larger at colder temperatures on their preferred host plant, tobacco, but the slope was reversed for the thermal reaction norm when eating devil's claw, a less preferred host plant. Emerging research has demonstrated that, when herbivores ingest plants, complex interactions between the plants and the herbivores come into play, and protein and carbohydrate are not acquired in the ratio they are ingested (reviewed by Clissold, 2007). This discrepancy is due to not yet fully understood interactions between the anatomy and cellular structure of the leaf, the morphology of the insect's mouthparts and the time food is retained within the gastrointestinal tract (Clissold et al., 2006; Clissold et al., 2009; Clissold et al., 2010). Thus, the plant-specific effects on body mass may occur because temperature affects the rate of feeding and chewing, food transit times and post-ingestive metabolic processing of food (Angilletta, 2009; Du et al., 2007; Huey and Slatkin, 1976) – all factors that have plant-specific effects on the balance of protein to carbohydrate absorped.

Experiments using synthetic diets have revealed that herbivorous insects have a well-developed capacity to adjust their feeding behaviour and post-ingestive physiology in order to regulate intake and allocation of specific nutrients to maintain nutritional homeostasis (reviewed by Harrison et al., 2012; Simpson and Raubenheimer, 2012). If temperature does alter the ratio of protein and carbohydrate absorbed from the ingested plants, then potentially, insect herbivores could make use of variations in temperature to regulate their nutrition. In an earlier study, the migratory locust, Locusta migratoria, used temperature selection to adjust the trade-off between the rate and efficiency of energy gain (Coggan et al., 2011). Locusta migratoria favoured increased metabolic efficiency when food was limited and increased growth rate when food was in excess (Coggan et al., 2011; Miller et al., 2009).

To test the hypothesis that insects can use temperature to regulate the ratio of protein to carbohydrate absorbed from plants, we used an established insect–plant system: L. migratoria feeding on two host grasses, Themeda triandra (kangaroo grass) and seedling Triticum aestivum (wheat). These grasses were chosen because L. migratoria absorb different amounts of protein (P) and carbohydrate (C) (Clissold et al., 2013) from each grass. Studies using synthetic diets revealed that L. migratoria nymphs achieve optimal growth and development when feeding on a slightly carbohydrate-biased diet (1P:1.1C) (Clissold et al., 2013; Miller et al., 2009). When feeding on kangaroo grass, protein and carbohydrate were absorbed in a ratio (1P:1.3C) very close to this optimum. We expect rates of growth and development will be reduced when feeding on seedling wheat, as protein is massively oversupplied relative to carbohydrate requirements (1P:0.5C) (Clissold et al., 2013).

In the present study, we first determined the host-plant-specific responses to temperature. In laboratory experiments, the effects of temperature on life history variables were ascertained and gravimetric measures were used to quantify the rates of protein and carbohydrate absorbed over the final nymphal (fifth) instar for L. migratoria feeding on either kangaroo grass or wheat. Results from the first experiment revealed that the relative rates of protein and carbohydrate absorption depended on temperature when feeding on kangaroo grass, whereas the absolute rate of absorption of protein and carbohydrate, but not the ratio, were temperature dependent when feeding on wheat. Thus, the potential existed for locusts feeding on kangaroo grass but not wheat to make use of thermal heterogeneity over fine scales to adjust for nutrient-specific imbalances. We tested two predictions: (1) locusts fed kangaroo grass should respond to an imposed imbalance in protein–carbohydrate status by selecting a higher temperature when deficient in protein, and a lower temperature when deficient in carbohydrate; and (2) locusts fed wheat should select higher temperatures irrespective of their protein–carbohydrate status, as higher rates of nutrient acquisition and growth occur at such temperatures. In further trials, we assessed whether temperature selection could improve with experience.

Locusta migratoria (Linnaeus 1758) (Orthoptera: Acrididae) came from a long-term culture at The University of Sydney (originally collected from the Central Highlands of Queensland, Australia). Between 500 and 1000 nymphs were reared on seedling wheat and wheat germ in large plastic bins (56×76×60 cm) under a 14 h:10 h light:dark photoperiod in a room kept at 30°C. During the ‘lights on’ phase, each bin had an additional heat source (250 W heat lamp) mounted on the mesh roof of the bin allowing locusts to thermoregulate by moving vertically on perches and the sides of the bins.

Leaf blades were harvested from mature Themeda triandra (Poales: Poaceae) and from 2-week-old Triticum aestivum (Poales: Poaceae) grown in a glasshouse. Three treatment diets were used, with the percentages of protein and digestible carbohydrate as follows: 21%P:21%C (21P:21C, balanced); 35%P:7%C (35P:7C, low carbohydrate); and 7%P:35%C (7P:35C, low protein). Treatment diets consisted of dry granular synthetic foods differing in the ratio of protein (a 3:1:1 mixture of casein, bacteriological peptone and egg albumen) and carbohydrate (a 1:1 mixture of sucrose and dextrin). All diets contained 4% micronutrients (salts, vitamins and sterols) and the remainder consisted of indigestible cellulose (54%), prepared as described previously (Simpson and Abisgold, 1985).

We measured the protein and carbohydrate (digested and absorbed) from the gut (measured as nutrient in ingesta less nutrient in faeces), and subsequent protein and carbohydrate growth (a measure of utilization) by locusts at two temperatures (see below for details). We used 32°C, where nutrient utilization efficiency is highest, and 38°C, where development rate is fastest (Miller et al., 2009). Freshly moulted (within 3.5 h) fifth instar male locusts, weighing between 360 and 420 mg, were randomly assigned to treatments (N=20 per treatment), with treatments being temperature (32 or 38°C) and grass species (wheat or kangaroo grass). Locusts were killed within 3.5 h of moulting into adulthood and lyophilized to a constant mass. A known amount of fresh grass was provided daily to allow ad libitum feeding and all remaining grass and faeces were removed and stored at −80°C prior to lyophilisation. Two experimental rooms were maintained at either 32 or 38°C (±0.1°C) under a 14 h:10 h light:dark photoregime. Each locust was individually housed within a clear plastic container (7×13.5 cm, diameter × height) so that consumption and faecal output could be determined for each nymph for the duration of the stadium. To minimize differences in leaf chemistry and biomechanical properties over the 24 h that the locusts had access to the grasses, blades of the three oldest fully expanded leaves per stem were detached at the ligule and placed in a vial as previously described (Clissold et al., 2004). Each day, a subset of the harvested grass blades were put aside to determine the ratio of fresh mass to dry mass and the amount of protein and non-structural carbohydrate (supplementary material Table S1). During the experiment, water loss and metabolic changes to the grass blades during the 24 h in which the grass blades were available to the locusts were quantified and values used to correct intake if required (supplementary material Table S2) (Bowers et al., 1991).

Total intake of protein and carbohydrate were calculated from daily food consumption based on the percentage of protein and carbohydrate in the grass. Absorption of protein and carbohydrate was determined as the difference between the mass of nutrient ingested and that remaining in the faeces (Fig. 1) (Clissold et al., 2006; Clissold et al., 2009). Using the Bradford assay, we are able to determined exogenous protein and remove the effect of endogenous nitrogenous waste products, such as amino acids and uric acid in the faeces. The majority of carbohydrates absorbed are respired, thus, those in the faeces are mostly of exogenous origin (the peritrophic membrane is not hydrolysed by our extraction technique). Total growth was defined as final dry mass less initial dry mass, with the initial dry mass being calculated from a regression of dry mass against wet mass derived from a subset of 20 freshly moulted fifth instar nymphs. Protein and lipid growth was measured as the increase in carcass protein and lipid, respectively, estimated from regressions for the protein and lipid content of the subset of 20 freshly moulted fifth instar locusts. Digestive efficiency (absorption/intake) was calculated using ANCOVA-corrected means for nutrient absorption expressed relative to nutrient intake, and growth efficiency (growth/initial biomass) was calculated from ANCOVA-corrected means for nutrient specific growth expressed relative to initial nutrient biomass. For example, protein digestibility at 38°C=100(ANCOVA-corrected mean of protein absorbed at 38°C/mean of protein intake across both temperatures).

Protein and non-structural carbohydrates (available carbohydrate) were determined from lyophilized samples of the grasses and faeces that had been finely ground prior to analysis. Protein was extracted from replicate 10 mg samples with 0.1 mol l⁻¹ NaOH and determined using the Bio-Rad Protein Microassay for Microtiter Plates based on the Bradford assay (Bio-Rad Laboratories, Hercules, CA, USA). Total non-structural carbohydrate was determined colorimetrically from 15 mg replicate samples following extraction with 0.1 mol l⁻¹ H₂SO₄ (Smith et al., 1964) using the phenol-sulphuric assay (DuBois et al., 1956). Carcass lipid was determined gravimetrically following three 24 h extractions with 100% chloroform. Previous regression analysis had found that, regardless of diet, the lipid-free dry matter contains the same proportion of protein (0.81) (F.J.C., personal observation) and thus protein was determined by multiplying the lipid-free dry matter by this value.

Freshly moulted fifth instar male locusts were randomly allocated to either the 32 or 38°C constant temperature room (rearing temperature) and provided with an optimally balanced synthetic diet (21P:21C) (Miller et al., 2009) from Day 0 (day of moult) to Day 3. On Day 3, half the locusts from each temperature were confined to a diet containing 35P:7C and the other half to 7P:35C for 4 h at 35°C, which rendered the locusts either carbohydrate or protein deprived, respectively (see  Appendix for methods; supplementary material Fig. S1). Protein- or carbohydrate-deprived locusts were provided with fresh blades of either wheat or kangaroo grass and observed on a thermal gradient plate as described below. This experiment was repeated (trials 1 and 2, ca. 5–10 per treatment, with a total of eight treatments – 2× rearing temperature, 2× nutrient state and 2× grass species – per trial, N=122).

Forty-eight freshly moulted fifth instar male nymphs were provided with kangaroo grass and randomly assigned to either 32 or 38°C (‘experienced’ locusts) or reared as for Experiment 2 on an optimally balanced synthetic diet (21P:21C) (‘naive’ locusts). Locusts feeding on kangaroo grass at 32°C were supplied with a 1:1 ratio of protein and carbohydrate, which is higher than optimal, and at 38°C a lower than optimal diet was supplied (1P:1.5C). Regardless of diet treatment locusts were switched between the two temperatures every 3.5 h [the average time interval at which they switch between a diet high in protein to one high in carbohydrate if allowed a choice (Chambers et al., 1995)] from Day 0 (day of moult) to Day 3. On Day 3, locusts were rendered protein or carbohydrate deprived (at 35°C) as previously described and then placed in a thermal gradient with a meal of kangaroo grass. This experiment was repeated (N=94).

Thermal gradients were established in a 28°C constant temperature room on six steel plates. Each plate was 700×100×10 mm (length × width × depth), and a thermal gradient was established by the use of 150 and 75 W ceramic heat lamps beneath either end of each plate, producing horizontal temperature gradients ranging from 27 and 45°C, both at the surface and 1 cm above the surface of the plate (Coggan et al., 2011). Temperature profiles of each plate were measured using infrared images (IR Camera S65, FLIR Systems, Wilsonville, OR, USA) and measurements validated with a 0.5 mm beaded wire K-type thermocouple placed on and 1 cm above the plate (YC-747D, Yu Ching Technology, Taiwan). Locusts were confined to a region on the plate between approximately 27 and 45°C using an elliptical wall, 300 mm high and coated in liquid Teflon (Fluon, The Herp Shop, Ardeer, Vic, Australia), as previously described (Coggan et al., 2011). A vial of fresh grass blades was placed on each plate at 35°C and perches of wire mesh were placed at 32 and 38°C, which were equidistant (~12 cm) from the food. The water-filled vial containing the grass blades was an inverted 1.5 ml microfuge tube cut to a height of 15 mm and attached to a 40 mm diameter Petri dish lid. Locust behaviour was captured on digital video (JVC Everio, Yokohama, Japan) and then quantified visually. Locusts alternate between periods of movement and remaining relatively stationary (Coggan et al., 2011), and during the stationary periods, locusts are either feeding or processing ingesta. Thermal preference was recorded once the locust had settled (within 5 min of completing the ingestion of a meal of the grass blades) and remained stationary for a minimum of a further 15 min. To quantify thermal preferences, the plate was divided into five ‘zones’ of width 3°C; ~27–30.5°C, 30.5–33.5°C, 33.6–36.5°C, 36.6–39.5°C and 39.5–~43°C, with locusts being scored as being at 29, 32, 35, 38 or 41°C, respectively. As found previously (Coggan et al., 2011), locusts did not go below 30°C or above 43°C; i.e. they were not constrained in their upper or lower temperature selection by the walls confining them within the thermal gradient. The position of the thorax of the locusts was used as the point of reference. The steel plates were randomly oriented in the room to control for any effects of position.

Life history outcomes and nutritional behaviours were analysed separately for each grass using ANOVA and ANCOVA, as there were complex interactions between diet and temperature. Nutrient-specific intake, digestion and total growth were compared with ANCOVA using initial biomass as the covariate. In all cases the assumptions for ANCOVA and ANOVA were met. Temperature selection behaviour was compared using ANOVA and χ² tests. Temperature selection from Experiment 2 was initially modelled with trial, rearing temperature, grass species and nutrient state as fixed factors; as trial and rearing temperature were non-significant (P>0.5, thus, there was a minimal chance of a type II error), these terms were removed from the final model. To investigate the effect of learning on temperature selection for locusts being offered kangaroo grass (Experiment 3), we used ANOVA with knowledge state and nutritional state as fixed factors. Approximately 10% of locusts were excluded from the temperature preference analyses as they failed to ingest a meal of grass within 2 h.

Regardless of grass eaten, locusts developed ca. 35% faster at 38°C than at 32°C (5.4 vs 7.6 days, respectively; F_1,76=450.9, P<0.001). However, growth of fifth instar L. migratoria nymphs differed with temperature in a host plant-specific way, due to differences in the rate and balance of nutrients absorbed from the gut. For nymphs fed kangaroo grass, although this was ingested 38% faster at 38°C than at 32°C, temperature did not affect the total amount of nutrients (P + C), and hence energy, absorbed (F_1,37=0.39, P=0.580; Fig. 2A, inset). However, the ratio of protein to carbohydrate absorbed differed markedly (Fig. 2A). At 32°C, more carbohydrate than protein was absorbed than at 38°C (1P:1.5C vs 1P:1.1C, respectively; P: F_1,37=36.96, P<0.001; C: F_1,37=7.14, P=0.011; ratio P:C: F_1,37=4.77, P=0.035; Fig. 2A,B). Once absorbed from the gut, both protein and carbohydrate were converted to biomass with higher efficiency at 32°C than at 38°C (Fig. 2B), and locusts were heavier at 32°C than at 38°C (Fig. 2C).

In contrast, for locusts feeding on wheat, although intake rates were 63% higher at 38°C than at 32°C, there was no difference in the ratio of nutrients extracted at 32 and 38°C (F_1,37=1.28, P=0.266). Approximately 90 and 60% of the ingested protein and carbohydrate, respectively, was absorbed. However, more of both protein and carbohydrate were extracted at the higher temperature (P: F_1,37=25.6, P<0.001; C: F_1,37=12.8, P=0.001), leading to more nutrients in total being absorbed over the fifth stadium at 38°C compared with 32°C (F_1,37=30.3, P<0.001; Fig. 3A, inset). Temperature had no effect on the efficiency with which nutrients were either extracted from wheat or subsequently converted to biomass (Fig. 3B), and locusts were bigger at 38°C than at 32°C (F_1,38=14.68, P<0.001; Fig. 3C).

For locusts reared on 21P:21C for the first 72 h of the fifth stadium, following the manipulation of the locust's nutritional state, all nymphs feeding on wheat, irrespective of state (protein or carbohydrate deprived), selected an average temperature of 37.8±0.2°C (χ²=0.06, d.f.=4, P=0.861; Fig. 4A, Table 1), with over 85% of insects selecting temperatures within the 38 or 41°C ‘zones’ (Fig. 4A). In contrast, locusts fed kangaroo grass chose the temperature at which they were most efficient at gaining from that grass the nutrient for which they were specifically deficient (Fig. 2A). Protein- and carbohydrate-deprived insects selected significantly different temperatures (38.3±0.4 and 34.1±0.4°C, respectively; F_1,38=37.24, P<0.001; Fig. 4B). Eighty-five per cent of protein-deprived locusts selected 38 or 41°C (Fig. 4B) – the temperature zone at which more protein than carbohydrate is extracted from kangaroo grass (Fig. 2A). In contrast, only 20% of carbohydrate-deprived locusts selected these higher temperatures. However, fewer than 50% selected 32°C (Fig. 4B), at which they would have obtained more carbohydrate than protein (Fig. 2A) and therefore redressed their nutritional imbalance.

Following the 3 day period where locusts were able to experience the nutritional outcomes when ingesting kangaroo grass at both temperatures, before being rendered protein or carbohydrate deprived, the capacity of carbohydrate-deprived insects ingesting kangaroo grass to choose 32°C improved substantially (Table 2, Fig. 4B,C), with over 85% of locusts now choosing 32°C – a 70% improvement from that of naive locusts. Protein-deprived locusts ingesting kangaroo grass also improved marginally, from an 85% success rate at choosing 38°C without having previously ingested kangaroo grass to 95% after experience (Table 2). The temperature regime (starting at either 32 or 38°C) experienced during the rearing period from Day 0 to 3 had no effect (P>0.5) on temperature selection and this term was removed from the analysis.

Although temperature-specific nutrient digestion has been recorded in other ectotherms (Bendiksen et al., 2003; Zhang et al., 2009), and ectotherms are known to use temperature to ameliorate the effects of starvation (Angilletta, 2009; Coggan et al., 2011), to our knowledge this is the first example of an ectotherm using temperature selection to redress specific nutrient imbalances. Additionally, locusts improved their ability to address nutrient imbalances by temperature selection following experience. This provides a rare example of a learned nutrient-specific regulatory response. The discovery that an insect is able to select temperatures to meet the specific nutritional challenges posed by different host plants has implications for interpreting host-plant choice and has consequences for our understanding of how changed environmental temperatures may affect animal–plant interactions.

For locusts feeding on both grasses, the differences in mass with temperature were primarily a consequence of differences in the digestion and absorption of protein and carbohydrate and secondarily a consequence of losses due to post-absorptive metabolism (Figs 1, 2 and 3). When feeding on synthetic diets, regardless of temperature or the ratio of protein and carbohydrate, both nutrients were absorbed with equal efficiency (Miller et al., 2009). When ingesting intact leaves, locusts are unable to absorb all the ingested protein and carbohydrate (reviewed by Clissold, 2007) and we found the digestion and absorption of protein and carbohydrate occurred in a host-plant-specific way with temperature (Fig. 2A, Fig. 3A).

For locusts feeding on wheat, temperature did not alter the percentage of protein and carbohydrate absorbed nor allocated to body mass (Fig. 3B). Rather, intake rate increased with increasing temperature, and as a consequence locusts absorbed ca. 30% more energy (protein + carbohydrate) at 38°C than at 32°C (i.e. Fig. 1Bi vs Biii). This contrasts with our previous results where we found temperature had a profound effect on nutrient utilization efficiency of synthetic diets, with locusts reared at 38°C retaining substantially less of absorbed energy than those reared at 32°C (Miller et al., 2009). At 32°C, locusts obtained nutrients from wheat at a much reduced rate, which would increase metabolic costs (Clissold et al., 2009), potentially nullifying the improved nutrient utilization previously recorded at 32°C.

Although a difference in the total energy absorbed due to differences in intake rates of total energy (protein plus carbohydrate) explains differences in body mass with temperature for locusts feeding on wheat, this was not the case for locusts feeding on kangaroo grass. Similar amounts of kangaroo grass were ingested by locusts at both temperatures, but protein was absorbed with greater efficiency at 38°C and carbohydrate with greater efficiency at 32°C (Fig. 2B, Fig. 1Bi vs Bii). This resulted in a lower ratio of P:C being absorbed at 32°C than at 38°C (Fig. 2A). Increased body mass when eating diets in which carbohydrate is oversupplied relative to protein has previously been recorded for L. migratoria (Raubenheimer and Simpson, 1999). The much poorer efficiency with which protein was retained as body mass at 38°C was fully consistent with previous work when locusts are ingesting synthetic diets high in protein (Raubenheimer and Simpson, 1999). Previous research has shown that locusts are able to use a variety of post-absorptive mechanisms (Fig. 1) to maintain homeostasis when ingesting protein and carbohydrate in sub-optimal ratios (Raubenheimer and Simpson, 1999).

Among ectotherms, size at maturity is typically reduced with increasing temperatures, a form of phenotypic plasticity known as the temperature–size rule (TSR) (Atkinson, 1994). The robustness of this association has been challenged, as the TSR can be readily reversed by altering diet quality (Berrigan and Charnov, 1994; Diamond and Kingsolver, 2010; Kingsolver and Huey, 2008; Mousseau, 1997; Walters and Hassall, 2006). ‘Poor’ diet is often provided as an explanation of TSR reversal (e.g. Diamond and Kingsolver, 2010). Locusts feeding on wheat are such an example of TSR reversal, with locusts feeding on kangaroo grass following the typical TSR. In neither case is diet ‘quality’ a useful or sufficient explanation. Rather, by determining protein and carbohydrate ingested, absorbed and subsequent allocation to growth, we have been able to describe the nutritional underpinning of growth differences with diet and temperature for L. migratoria when feeding on these two grasses. Further research is required to understand the mechanisms responsible for these nutritional differences, before general predictions regarding the effects of temperature on insect herbivores can be made. Such mechanisms could include differential changes in the activities of enzymes and transporters, and interactions with secondary metabolites (Angilletta, 2009; Stamp, 1990; Stamp and Yang, 1996). Regardless of the mechanism, temperature altered the ratio of protein and carbohydrate absorbed from kangaroo grass and thus offered locusts the opportunity to select a temperature and prioritize optimizing nutrient uptake and maximizing body mass or prioritize development rate at the expense of growth.

Rates of both growth and development were maximized for locusts feeding on wheat at 38°C (Fig. 3), and they selected 38°C or higher regardless of deprivation state following a meal of wheat (Fig. 4A). However, locusts fed kangaroo grass adopted a temperature consistent with optimizing the gain of the nutrient that was in shortage (Fig. 4). That is, locusts used temperature choice to correct a nutritional imbalance (Fig. 2). Previously it has been demonstrated that this same locust species is able to associate odour (Simpson and White, 1990) and colour (Raubenheimer and Tucker, 1997) cues with the protein and carbohydrate content of artificial diets, and to be attracted by the appropriate cue only when in a specific state of nutrient imbalance (Raubenheimer and Tucker, 1997). In such cases, the subsequent redressing of the nutritional imbalance occurs upon eating the food associated with the cue. Here we have shown, for the first time in any animal species as far as we are aware, that locusts associate temperature cues not with the presence of a nutritionally appropriate food, but with differential effects of temperature on digestion of nutrients from an already ingested meal.

Many insects, including grasshoppers and locusts, can rapidly associate specific nutrients with olfactory or visual cues (Dukas, 2008), and such learning has been shown to allow a grasshopper, Schistocerca americana, to achieve higher growth rates in a laboratory setting (Dukas and Bernays, 2000). However, when feeding on kangaroo grass and carbohydrate deprived, locust temperature choice favoured maximizing body size at maturity and compromising rates of development. Life history outcomes and subsequent fitness of insects are strongly influenced both directly and indirectly by rates of development and final size at maturity (e.g. Häggström and Larsson, 1995; Roff, 1992; Rowe and Ludwig, 1991; Stamp, 1990). Prolonged juvenile development has been associated with time-dependent mortality risks and the inability of insects to exploit temporally short resources (e.g. Creel and Christianson, 2008; Feeny, 1970). But it has been demonstrated that learning can be beneficial in optimising performance in heterogeneous environments, especially when dealing with features unique to a location and time (Dukas, 2008; Dukas and Bernays, 2000). Insects typically have rapid rates of development, and making use of experience to capitalize on the fine-scale thermal variation present in the natural environment (Willmer, 1982) may decrease risks associated with moving between food patches when foraging (Bernays, 1997; Gotthard, 2000), or allow insects that specialise on single host plants to meet nutritional requirements as the nutritional quality of leaves change over time (e.g. Read et al., 2003). Field studies are required to understand how locusts balance requirements for growth (optimizing nutritional needs) and rates of development given temperature and the available host plants.

The discovery that an insect is able to select temperatures to meet the specific nutritional challenges posed by different host plants highlights the pitfalls of trying to interpret an animal's preferences for foods or temperatures. Recent evidence suggests that the probability of a leaf being eaten depends on a complex suite of plant and herbivore traits (Agrawal, 2011; Carmona et al., 2011) and we suggest also abiotic conditions. As small ectotherms have abundant opportunities to make adaptive adjustments to body temperature by small-scale habitat selection (Willmer, 1982), host-plant use must be considered in a temperature context. Therefore, micro-climatic effects may be of more importance than previously appreciated in predicting the ecological consequences of climate change and altered land-use practices (Kearney et al., 2009; Suggitt et al., 2011).

We thank N. Soran T. Dodgson, B. Panayotakos and J. Popple for general assistance, and members of the University of Sydney Behaviour, Physiology and Ecology laboratory for locust rearing. This manuscript was improved by insightful comments from A. Cease and two anonymous reviewers.

The aim was to rapidly make the locusts either protein or carbohydrate deprived. While the treatment diets have rather extreme ratios, they were offered to the locusts over the shortest period of time possible so that ca. 95% of the locusts would subsequently select foods containing higher amounts of the nutrient in which they were deficient. To confirm that locusts were rendered protein or carbohydrate deprived by the 4 h treatment on unbalanced synthetic food, 20 locusts, not otherwise used for experiments, were randomly selected in a design balanced by experiment and rearing conditions, and placed individually in clear plastic containers (11×17×5 cm, length × width × height) with water and two small feeding bowls, one containing food high in protein, 35P:7C, and the other containing food high in carbohydrate, 7P:35C. Behaviour was videoed and the first meal to be eaten was scored.