Starvation resistance is associated with developmentally specified changes in sleep, feeding and metabolic rate

Food acquisition represents a major challenge to many animal species, and the ability to locate food, or survive in the absence of food, strongly associates with reproductive fitness (Chippindale et al., 1996; Wayne et al., 2006). The ability to resist starvation varies dramatically throughout the animal kingdom, and even between closely related species, yet surprisingly little is known about the biological basis for differences in this behavior (Gibbs and Reynolds, 2012; Matzkin et al., 2009; Rion and Kawecki, 2007). Animals have evolved diverse mechanisms for responding to acute shortages in nutrient availability, including the induction of foraging behavior, alterations in sleep and locomotor activity, and changes in metabolic rate (Schmidt, 2014; Stahl et al., 2017; Sternson et al., 2013; Yurgel et al., 2014). While starvation resistance is likely influenced by developmental processes that contribute to an organism's size, metabolic phenotypes and brain function, it is not known whether selection occurs at developmentally specified stages or is maintained throughout development. Defining the effects of selection for starvation resistance on behavior and metabolism across development is therefore critical for understanding the developmental specificity of evolved changes in these processes.

The fruit fly Drosophila melanogaster provides a powerful model for investigating the mechanistic basis of starvation resistance (Rion and Kawecki, 2007). Outbred populations of fruit flies display highly variable starvation resistance, as well as traits that are associated with starvation resistance, including developmental timing, sleep and feeding behaviors (Folguera et al., 2008; Garlapow et al., 2016; Harbison et al., 2017; Masek et al., 2014; Svetec et al., 2015; Yadav and Sharma, 2014), but little is known about how these individual traits contribute to the evolution of starvation resistance. We have implemented experimental evolution by starving outbred adult Drosophila until only 15% of the initial population remain alive, then passaging the survivors onto the next generation (Hardy et al., 2018). These populations have been independently selected over 100 generations, resulting in flies that survive up to two weeks in the absence of food, while non-selected flies survive for only 3–4 days. These starvation-selected populations provide an opportunity to examine how behavioral and physiological traits are altered by selection for starvation resistance and whether selection in adults also influences their development.

Altered life history and behavioral changes in adults are associated with evolutionarily acquired resistance to nutrient stress (Bubliy and Loeschcke, 2005; Gefen et al., 2006; Kolss et al., 2009), but the specific contributions of the many behavioral and physiological changes to starvation resistance have been difficult to test experimentally. We have previously identified increased sleep and reduced feeding in adult Drosophila selected for resistance to starvation stress (Masek et al., 2014; Slocumb et al., 2015). While these traits likely emerged as a mechanism to conserve energy in the absence of food, their precise roles in starvation resistance are unknown. In addition, both sleep and feeding are developmentally plastic behaviors, and are modulated by both shared and independent neural mechanisms during the larval and adult stages (Itskov and Ribeiro, 2013; Koh et al., 2006; Melcher and Pankratz, 2005; Pool and Scott, 2014; Szuperak et al., 2018). Drosophila eat voraciously throughout development, and this is essential for organismal growth and the generation for energy stores that persist through adulthood (Tennessen and Thummel, 2011). In addition, we have recently characterized larval sleep and found that this sleep is critical for development (Szuperak et al., 2018). Therefore, it is possible that selection for starvation resistance differentially influences adult behavior and physiological function, or that shared genetic architecture between development and adulthood results in an evolutionary constraint on developmental state-specific modification of behavior.

Here, we investigate sleep, feeding and metabolic function during development in flies selected for starvation resistance. We find that development time is extended, starting at the second instar larval stage, and persists throughout development. In addition, whole-body metabolic rate is reduced during both development and adulthood and is accompanied by an increase in mass, suggesting that reduced energy expenditure allows these starvation-resistant populations to increase their energy stores. Our findings also reveal that increased sleep and reduced feeding are specific to the adult stage, suggesting that selection for starvation resistance can target specific behaviors at different developmental time points.

Starvation-selected populations at generation 122 were obtained and then tested and maintained off-selection for a maximum of 5 generations. All populations were grown and maintained on standard Drosophila medium (Bloomington Drosophila Stock Center) and maintained in incubators (Percival Scientific, Perry, IA, USA) at 25°C and 50% humidity on a 12 h light:12 h dark cycle. For larval experiments, adult flies were maintained in population cages with access to grape juice agar and yeast paste (Featherstone et al., 2009). Unless stated otherwise, eggs were collected from the cages within 2 h of being laid and then transferred into Petri dishes containing standard food at a constant density of 100 eggs per dish.

Eggs were transferred into Petri dishes containing standard food and green food coloring, which allowed for easier viewing of the larvae, at a density of 25 eggs per dish. Larvae were then scored every 4 h for their transition through the first, second and third instar stages. Each larval stage was distinguished by the size and complexity of their mouthparts (transition from first to second instar), as well as the branching pattern of the anterior spiracles and the size of the dark orange ring on the posterior spiracles (transition from second to third instar; JoVE Science Education Database). The time at which at least 50% of the larvae within each Petri dish had transitioned through each developmental stage was recorded. Time to pupariation and eclosion was measured independently from each larval instar stage. Eggs were collected within 2 h of being laid and placed individually into glass test tubes, each containing 2 ml standard Drosophila medium. Tubes were then scored every 4 h.

Short-term food intake in adult flies was measured as previously described (Wong et al., 2009). Briefly, sets of five 3- to 4-day-old female flies were either transferred to vials containing a damp Kimwipe and starved, or maintained on standard food for 24 h. At zietgeber time (ZT) 0, flies from both treatments were transferred to food vials containing 1% agar, 5% sucrose and 2.5% blue dye (Federal Food, Drug and Cosmetic Act, blue dye no. 1, Spectrum Chemical Manufacturing Corp., New Brunswick, NJ, USA). After 30 min, flies were flash frozen and stored for subsequent analyses. For food consumption measurements in larvae, eggs were obtained as previously described. Eggs were transferred to Petri dishes containing standard food at a larval density of 100 larvae per dish. Food consumption was measured at 60 and 96 h after egg laying for second and third instar larvae, respectively, as previously described (Kaun et al., 2007). Briefly, larvae were transferred to Petri dishes containing a thin layer of 1% agar and yeast paste with 2.5% blue dye. After 15 min of feeding, larvae were collected and then washed in ddH₂O three times. The larvae were then flash frozen in groups of 10 and 5 for second and third instar larvae, respectively. Each larval and adult sample was homogenized in 400 μl phosphate-buffered saline (PBS) and then centrifuged at 4°C at 15,710 g. The supernatant was then extracted and its absorbance at 655 nm was measured in triplicate using a 96-well plate absorbance spectrophotometer (Bio-Rad Laboratories). Baseline absorbance was determined by subtracting the absorbance obtained from flies/larvae not fed blue dye from each experimental sample. The amount of food consumed was then determined from a standard curve. To assess feeding rate in second and third instar larvae, the number of mouth hook contractions were counted (Shen, 2012). For each group, second or third instar larvae were placed onto a Petri dish containing agar and yeast paste. After a 1 min acclimation period, larvae were video recorded and the number of mouth hook contractions within a 30 s period were counted.

For adults, 3- to 5-day-old female flies were isolated and placed on fresh medium for 24 h, and then the mass of groups of 10 flies were determined. For second and third instar larvae, mass was measured at 60 and 96 h after egg laying and was determined using groups of 10 and 20 larvae, respectively.

In adults, individual 3- to 5-day-old mated female flies were placed into tubes containing standard food and allowed to acclimate to experimental conditions for at least 24 h. Sleep and activity were then measured over a 24 h period starting at ZT0 using the Drosophila Locomotor Activity Monitor System (DAMs) (Trikinetics, Waltham, MA, USA) as previously described (Hendricks et al., 2000; Shaw et al., 2000). The DAM system measures activity by counting the number of infrared beam crossings for each individual fly. These activity data were then used to calculate bouts of immobility of 5 min or more using the Drosophila sleep counting macro (Pfeiffenberger et al., 2010), from which sleep traits were then extracted. In larvae, sleep and activity was measured as described (Szuperak et al., 2018). Briefly, individual freshly molted second instar larvae were loaded into wells of custom-made polydimethylsiloxane (PDMS) microplates (‘larva lodges’) containing 3% agar and 2% sucrose with a thin layer of yeast paste. Larva lodges were loaded into incubators at 25°C and time-lapse images were captured every 6 s under dark-field illumination using infrared LEDs. Images were analyzed using custom-written MATLAB software and activity/quiescence determined by pixel value changed between temporally adjacent images. Total sleep was summed over 6 h beginning 2 h after the molt to second instar. Sleep bout number and average sleep bout duration was calculated during this same period.

After sleep assessment, the same flies were also used to measure starvation resistance. Following 24 h of testing on standard food, flies were transferred to tubes containing 1% agar (Fisher Scientific) and starvation resistance was assessed. The time of death was manually determined for each individual fly as the last bout of waking activity.

Metabolic rate was measured using indirect calorimetry by measuring CO₂ production (Stahl et al., 2017). Staged larvae were placed in groups of five (second instar) or individually (third instar) onto a small dish containing standard food medium. Each dish was placed in a behavioral chamber where larvae were acclimated for 30 min, which is approximately the time required to purge the system of ambient air and residual CO₂. Metabolic rate was then assessed by quantifying the amount of CO₂ produced in 5 min intervals for 1 h. All experiments were conducted during ZT0–ZT6 to minimize variation attributed to circadian differences in sleep, feeding or metabolic rate. Metabolic rate in adults was assessed as described previously (Stahl et al., 2017). Briefly, adult flies were placed individually into behavioral chambers containing a food vial of 1% agar and 5% sucrose. Flies were acclimated to the chambers for 24 h and then metabolic rate was assessed by quantifying the amount of CO₂ produced in 5 min intervals during the subsequent 24 h. Metabolic data for each group were normalized for body weight by dividing metabolic rate by mass, measured as described above.

To assess differences in survivorship between starvation-selected and control populations, starvation resistance was analyzed using a log-rank test. Log-rank tests were also used to assess differences in development time, from first instar to eclosion. A two-way ANOVA was performed on measurements of metabolic rate, mass, food consumption, mouth hook contractions and sleep traits (factor 1: selection regime; factor 2: replicate population). If significant differences were observed, Sidak's multiple comparisons test was performed to identify significant differences within each replicate population. All statistical analyses were performed using GraphPad Prism 7.0. In all figures, individual data points are shown, along with bars indicating mean values and error bars showing s.e.m.

To assess developmental correlates of increased starvation stress, we utilized outbred populations that were subjected to laboratory natural selection for starvation resistance (Rose et al., 1996). Flies were selected for starvation resistance by placing adult flies on agar and passaging starvation-resistant populations onto food when only ∼15% of flies remained alive. Three parallel starvation-resistant groups were generated (S_A, S_B and S_C) as well as three controls that were continuously passaged on food (F_A, F_B and F_C). Experiments in this study utilized flies maintained on this selection protocol for 122 generations (Fig. 1A). In agreement with previous studies performed on flies selected for 60–80 generations (Hardy et al., 2018; Masek et al., 2014), this selection protocol robustly increased starvation resistance. All three S populations survived on agar for an average of 9–13 days compared with 2–3 days for the F populations (Fig. 1B,C), confirming that selection for starvation resistance results in a ∼4-fold increase in survival under starvation conditions.

It has been previously shown that starvation selection is associated with a larger body size in adult flies (Masek et al., 2014; Slocumb et al., 2015). To investigate whether this increase in body mass also occurs during development or is restricted to adults, we measured body mass during the second and third instar stages. Overall, we found that starvation-selected populations weighed significantly more than control populations at the second instar stage. However, we did not observe this effect when directly comparing each replicate F and S group individually (Fig. 1D). In the third instar stage, starvation-selected populations weighed significantly more than fed control populations, and each individual S group replicate weighed significantly more than their respective F control group (Fig. 1E). This increase in mass for all three starvation-selected replicate groups was maintained into adulthood (Fig. 1F). These findings suggest that starvation selection is accompanied by an increase in mass that occurs during the third instar stage and persists through adulthood.

It is possible that delayed development contributes to starvation resistance by allowing flies to accumulate energy stores during the larval stages. To determine whether the rate of development is altered by starvation selection, we measured the time from egg laying to each developmental transition. Overall, development rate was delayed across all S groups, confirming that starvation selection increases development time (Fig. 2A). A direct comparison of each developmental stage revealed no difference between F control groups and starvation-selected S groups in the transition from egg to first instar larvae, suggesting the selection protocol does not affect the earliest stages of development (Fig. 2B). Development time was significantly delayed at all subsequent developmental stages (from first instar to pupariation) when the three replicate starvation-selected populations and three control populations were pooled (Fig. S1). However, post hoc analyses on development time at each of these developmental stages revealed population-specific effects on the time spent within each stage. As such, a direct comparison of each replicate group revealed no differences in the duration of time spent as first instar larvae (Fig. 2C). For the duration of time spent as second instar larvae, a similar comparison of each replicate group revealed that significant differences were only observed between the F_B and S_B populations (Fig. 2D). In contrast, all three S group replicate populations spent significantly longer in the third instar stage (Fig. 2E). During pupariation, significant differences in development time were again only observed between the F_B and S_B populations (Fig. 2F). Therefore, delayed development time is present across all starvation-selected populations, but is particularly robust in the S_B population. Overall, these findings raise the possibility that increased body size and starvation resistance are related to delayed development.

In addition to delayed development, reduced metabolic rate provides a mechanism for conserving energy (Dulloo and Jacquet, 1998; Ma and Foster, 1986). Animals, including Drosophila, reduce metabolic rate under starvation conditions (Blaxter, 1989; McCue, 2010; Wang et al., 2006), suggesting that modulation of metabolic rate may promote starvation resistance. To determine the effect of starvation selection on metabolic rate, we used indirect calorimetry to determine CO₂ release, a proxy for metabolic rate, in both larvae and adults. Measurements of metabolic rate were then normalized to body mass in order to account for differences in body size between the F control groups and starvation-selected S groups. The system used to measure metabolic rate is highly sensitive, and has previously been used to detect CO₂ release from single flies (Fig. 3A; Stahl et al., 2017). In second instar larvae, no changes in metabolic rate were detected between F control groups and starvation-selected S groups (Fig. 3C). However, metabolic rate was reduced in third instar larvae when the three replicate starvation-selected populations and three control populations were each pooled (Fig. S1). Post hoc analyses of each replicate population revealed a significant decrease in metabolic rate in the S_A and S_C populations compared with their respective F control populations (Fig. 3D). In adult flies, metabolic rate was significantly decreased in all three starvation-selected populations (Fig. 3E). Therefore, selection for starvation resistance results in reduced metabolic rate that commences during the third instar stage and persists into adulthood.

We have previously shown that food consumption is reduced in fasted starvation-selected adult flies (Masek et al., 2014). However, the effects of selection on larval feeding remain unknown. To quantify feeding in second and third instar larvae, we measured food intake by placing flies on yeast paste laced with blue dye. The amount of food consumed over a 15 min period was then measured based on spectrophotometric analysis of dye consumed during this time period. Food consumption was significantly increased among second and third instar larvae when the three replicate starvation-selected populations and three control populations were each pooled (Fig. S1). However, during the second instar stage, post hoc analyses revealed that this effect was only significant in the F_C and S_C groups (Fig. 4A,B). During the third instar stage, food consumption was increased in all three starvation-selected replicate populations (Fig. 4C,D), suggesting that starvation selection promotes larval feeding. In contrast to larval feeding behavior, no differences were observed in food consumption across all three populations of starvation-selected adult flies in the fed state (Fig. 4E,F). However, when animals were food deprived, food consumption was significantly reduced across all three starvation-selected populations to induce a robust feeding response (Fig. 4G,H). These findings suggest that selection for starvation resistance has different effects on food consumption during the larval and adult stages.

It is possible that increased food consumption in starvation-selected larvae is a result of increased feeding drive or is secondary to their overall larger body size. To differentiate between these possibilities, we measured feeding rate by calculating the number of mouth hook contractions over a 30 s period. The number of mouth hook contractions did not differ between starvation-selected and control populations for second or third instar larvae (Fig. S2).

We previously reported that selection for starvation resistance increases sleep in adults (Masek et al., 2014). Here, we confirmed these results, finding that sleep duration was increased in all three starvation-selected populations, which is a consequence of increased bout length but not bout number (Fig. 5A–D). These results raise the possibility that energy conservation as a result of increased sleep may also occur during the larval stages. Recently, sleep has been characterized in second instar Drosophila larvae, allowing for the characterization of changes in sleep throughout development (Fig. 5E; Szuperak et al., 2018). Overall, we found that sleep increases among starvation-selected second instar larvae when the three replicate starvation-selected populations and three control populations were each pooled (Fig. S1). However, post hoc analyses revealed that when sleep was assessed in each replicate population, an increase in sleep was only observed among the F_C and S_C populations, suggesting that these differences in sleep are present throughout development (Fig. 5F). No significant differences in bout length or bout number were detected between replicate populations of second instar larvae, although a trend towards increased bout length in the S_C population was detected, suggesting that sleep architecture is largely unaffected by selection for starvation resistance (Fig. 5G,H). Although we found no differences in sleep in second instar larvae, it is possible that additional sleep differences exist in third instar larvae; however, it remains unknown whether third instar larvae exhibit sleep states.

Here, we report on the ontogenetically specified changes in behavior and metabolic rate induced by selection for starvation resistance. We found that starvation selection extends larval development beginning in the second instar stage, with a concomitant decrease in metabolic rate and increase in food consumption beginning in the third instar stage. In adults, however, metabolic rate remains low, while food consumption remains unchanged and sleep is increased. These results suggest that starvation selection has differential effects on behavioral and metabolic traits as development progresses from the larval stages into adulthood and is consistent with a strategy where starvation-selected larvae prioritize growth, while adults prioritize energy conservation.

Previous studies have found that selection for starvation resistance results in slower development in Drosophila (Chippindale et al., 1996; Hoffmann and Harshman, 1999; Masek et al., 2014; Reynolds, 2013), suggesting that extended larval development represents a mechanism for developing starvation resistance as adults. Here, we showed that this reduced development rate begins as early as the second instar stage and persists until eclosion. During development, standard laboratory strains of D. melanogaster larvae increase their body size ∼200-fold during the ∼4 days of larval development (Church and Robertson, 1966), raising the possibility that even subtle changes in development rate may significantly affect adult body size and energy stores. Progression through each larval transition is regulated by the steroid hormone 20-hydroxyecdysone, a master regulator of developmental timing (Riddiford, 1993; Liu et al., 2017; Yamanaka et al., 2013) and it is possible that starvation selection acts to modify expression of this hormone, thereby delaying the onset of each larval transition. Additionally, several genetic factors have been identified that regulate nutrient-dependent changes in developmental timing, including the target of rapamycin signaling pathway (Colombani et al., 2003; Layalle et al., 2008) and insulin-like peptides (Ikeya et al., 2002; Slaidina et al., 2009). Although significant advances have been made in elucidating the mechanisms underlying larval development and growth, our understanding of how environmental conditions, including starvation, can modulate these factors remain poorly understood. Our study reveals that environmental stressors can be potent selective forces that have a strong impact on the timing of larval growth and development.

It is proposed that animals develop resistance to starvation stress by reducing energy expenditure (Aggarwal, 2014; Hoffmann and Parsons, 1989; Marron et al., 2003; Rion and Kawecki, 2007). Here, we show that metabolic rate is reduced in third instar larvae, as well as in adults, across all starvation-selected populations tested. While to our knowledge, the metabolic rate of Drosophila starvation-resistant larvae has not previously been studied, earlier reports examining metabolic rate in adults from different populations of D. melanogaster selected for starvation stress found no effect of selection on metabolic rate in flies (Baldal et al., 2006; Djawdan et al., 1997; Harshman and Schmid, 1998). However, there is evidence that selection for starvation resistance results in the production of different metabolic enzymes in response to starvation (Harshman and Schmid, 1998) as well as an accumulation of energy stores (Masek et al., 2014; Schwasinger-Schmidt et al., 2012; Slocumb et al., 2015). In our study, we measured the metabolic rate of adult flies over a 24 h period, thereby including any potential variation in the circadian effects of feeding, sleep and metabolic rate. It is possible that the independent origins of the selected populations resulted in selection on metabolic rate-dependent and -independent pathways, leading to enhanced starvation resistance. However, our finding that metabolic rate is reduced in multiple independent lines of starvation-selected populations suggests that these differences may be attributed to the initial outbred populations of flies used to evolve starvation resistance.

Increased body size is a fitness-related trait that promotes tolerance to stress (Ewing, 1961). As such, environmental perturbation and food shortages may uniquely affect fitness depending on the developmental stage in which these selective pressures occur. The selection protocol used in this study selectively applied nutrient shortages during adulthood only, limiting selection pressures to traits that enhance adult starvation resistance. However, we found that starvation selection increases body size as early as the second instar stage and persists throughout the rest of development. Similarly, we identified larval-specific effects on food consumption and sleep. We found that food intake is increased in larvae and reduced in adults, and that sleep is reduced in adults but unchanged in larvae. Our sleep analysis was limited to second instar larvae, therefore we were unable to determine whether the feeding and metabolic phenotypes observed in third instar larvae also extend to sleep. To date, sleep characterization is limited to second instar larvae, and technical challenges, including the size and mobility of third instar larvae provide technical impediments to sleep analysis at this developmental stage. However, our findings in second instar larvae provide further evidence that selection for starvation resistance results in ontogenetically specified behavioral phenotypes. It has been previously shown that selective stresses imposed during development contribute to altered behavioral states as adults. In several Drosophila species, for example, thermal stress applied during larval development confers resistance to thermal stress in adulthood (Levins, 1969; Goto, 2000; Horu and Kimuro, 1998; Maynard Smith, 2005). Therefore, selection for starvation resistance during a defined developmental window can impact a variety of traits at multiple stages throughout development.

Although we found that changes in development rate, metabolic function and sleep differ in starvation-selected populations, the genetic contribution of each trait to starvation resistance, especially at each stage of development, is unknown. We observed increased sleep and decreased starvation-induced feeding in starvation-selected adults, traits that are not present during larval development. Although these traits provide a potential mechanism for energy conservation in adult flies, in larvae, development rate slows, sleep does not differ and food consumption actually increases. Although food intake was increased in starvation-selected larvae, their feeding rate remained unchanged. It is possible, although not assessed in this study, that this may be a result of decreased activity. However, it is likely that this increase is related to their larger body size and results from increased food intake per mouth hook contraction. This, together with our findings that metabolic rate is reduced in both larvae and adults of starvation-selected populations, supports a model by which a slower development provides increased time to grow and accumulate energy stores as larvae, while reducing foraging-related behaviors in adulthood allows for animals to conserve energy as adults when food is not present during the selection process. This model suggests that distinct genetic architecture regulates sleep and feeding during the larval and adult stages. For instance, the mechanisms controlling larval sleep are partially distinct from that of adult sleep (Szuperak et al., 2018). Overall, these findings provide proof of principle for ontogeny-specific correlated behaviors during the applied starvation-selection process.

The identification of developmental, metabolic and behavioral differences in Drosophila populations selected for resistance to starvation suggest that multiple mechanisms likely contribute to the etiology of starvation resistance. Towards this end, association mapping in Drosophila identified a wide range of genes associated with starvation resistance, including those that are known regulators of development, metabolism and nutrient response (Harbison et al., 2004; Nelson et al., 2016). The complex genetic architecture underlying these traits, and their inter-relationship, suggests the evolution of starvation resistance is likely to be highly pleiotropic. A previous study examining the genetic divergence between these populations identified 1796 polymorphisms that significantly differed between starvation-selected and control populations. These polymorphisms mapped to a set of 382 genes, including genes associated with a wide variety of metabolic and physiological processes (Hardy et al., 2018). Indeed, there is experimental evidence for this in a naturally occurring population of Drosophila, in which a seasonally fluctuating molecular polymorphism in the Insulin-like receptor (InR) has been linked to increased starvation resistance, decreased fecundity and increased lifespan (Paaby et al., 2014). Given that we observed population-specific differences in mass, metabolic rate and sleep, it is also possible that distinct mechanisms contribute to starvation resistance in each of the three replicate populations. As such, genomic sequencing of these populations revealed a low correlation of allele frequencies between the starvation-selected populations, suggesting that there are indeed multiple mechanisms of adaptation. While these studies provide an initial framework for identifying genetic factors regulating traits contributing to starvation resistance, a typical limitation of studying selected populations is a lack of accessible genetic tools that can be applied to validate the phenotypic contributions of single genes. The recent application of gene-editing approaches to outbred and non-Drosophila melanogaster populations raises the possibility of examining the contributions of these candidate genes in the future. These findings reveal evidence of an ontogenetic shift associated with selection for starvation resistance in Drosophila melanogaster. This work highlights the contribution of several energy-saving traits that are modulated throughout development, including changes in metabolic rate, size, sleep and food consumption to confer resistance to starvation. The development-specific differences in sleep and feeding of the starvation-selected populations set the stage for elucidating the genetic basis of starvation resistance over the course of development.

We thank members of the Keene lab for technical assistance and helpful discussions.