Animals commonly face combinations of thermal and nutritional stress in nature, which will intensify under climate change. While genetic adaptation is necessary to buffer long-term stress, it’s unclear whether adaptation to combined stress can occur without compromising viability and thermal plasticity. We tested larval thermotolerance and thermal plasticity in Drosophila melanogaster selected under different temperatures (18°C, 25°C, and 28°C) and diets (standard, diluted, and low-protein:high-carbohydrate [P:C]). Basal larval cold tolerance was affected by both protein concentration and temperature; larvae evolved higher basal cold tolerance on the diluted and low P:C diets at 18°C and 28°C. Hardening increased cold tolerance for most lines, except those selected at 18°C and 28°C on low P:C diets and at 25°C on standard diets. Basal larval heat tolerance was affected by selection temperature; selection at 25°C increased heat tolerance. An interaction between selection temperature, selection diet, and hardening treatment affected larval heat tolerance; hardening reduced heat tolerance in most lines, except those selected at 25°C on low P:C diets and at 28°C on standard diets. Our results suggest that adaptation to combined stress allows basal cold tolerance and its plasticity to co-evolve, but not heat tolerance, highlighting ectotherm’s vulnerability to long-term climate change.

Fly stock – Collection and initiation of experimental evolution lines

Two hundred female D. melanogaster were collected in Duranbah, Queensland (28.3° S, 153° E) in 2018 to initiate isofemale lines (Alton et al. 2024). Flies were returned to the laboratory and were treated with tetracycline to remove Wolbachia (Min and Benzer 1997). Two generations after tetracycline treatment, five males and five virgin females from each line were used to create a mass-bred base population. A large number of isofemale lines were used to maximise genetic variation in the founding mass-bred population. The base population was maintained under control conditions — at 25°C on the standard diet with 12:12 hour light:dark cycles (see below) — across 60 bottles each containing approximately 750–1000 flies, for two generations before the initiation of selection. It is worth noting that, because the mass-bred population was maintained for only two generations before the selection experiment, it was most likely not fully adapted to laboratory conditions and was therefore not at equilibrium. This, in turn, may have contributed to linkage disequilibrium in the base (starting) population (e.g. Kellermann et al. 2015). That said, we believe that the control lines at least would have reached equilibrium by the time of the experiments described in this study, given laboratory adaptation occurs relatively quickly, and is expected to have stabilised by 7–10 generations (Harshman and Hoffmann 2000; Hoffmann et al. 2001). Because we used laboratory natural selection (rather than directional selection), it is also likely that the selected lines had reached equilibrium by the time the experiments were conducted (Sgrò and Blows 2004).

Experimental evolution regimes and selection protocol

Eggs were collected from the base population and were divided among nine combined thermal-nutritional selection treatments with five replicate lines per treatment (Alton et al. 2024). These treatments consisted of three temperatures (18°C, 25°C, and 28°C) and three diets (standard, diluted, and low P:C) (Table S1) under a 12:12 light:dark cycle. Only larvae were exposed to selection. All diets contained varying amounts of potato flake, inactive yeast, agar, and dextrose, with the addition of nipagin and propionic acid to prevent bacterial and fungal growth (Holleley et al. 2008) (Table S1). The standard diet had a 1:3 protein-to-carbohydrate (P:C) ratio (yeast 36.36 g/L) and contained 320 kcal/L; the diluted diet had a 1:3 P:C ratio (yeast 9.09 g/L) and contained 80 kcal/L; and the low P:C diet had a 1:12 P:C ratio (yeast 8.18 g/L) and contained 320 kcal/L (Table S1).

These selection conditions reflect combinations of thermal and nutritional changes that may be expected in the future under climate change; 18°C and 25°C represent the current average Australian winter and summer temperatures respectively (Alton et al. 2024), while 28°C represents a 3°C increase of the average summer temperature projected in twenty years (Australian Academy of Science 2021). The diluted diet and the low P:C diet simulate a reduction of protein and varying carbohydrates in plant composition projected under climate change (DaMatta et al. 2010), and span the range of nutritional contexts experienced by developing larvae in rotting fruit (Matavelli et al. 2015; Silva-Soares et al. 2017). The 25°C standard diet selection regime was used as a control and a reference point for further analysis.

At the start of each generation of selection, each selection replicate line was established with three bottles containing 200–250 eggs on 70 mL of selection diets to equalize density across all selection regimes. Animals were allowed to develop on their respective diet and temperatures until adult eclosion. We avoided inadvertent selection on development time by collecting all eclosing adults over a 5-day period across selection lines (all adults had emerged over a 5-day period, even those lines that took the longest to develop). Adults from the same replicate line were mixed and separated into two bottles containing the standard diet at 25°C for three days to allow mating. On the fourth day, adults were transferred to 250 mL egg-laying chambers containing 11 mL of egg-laying medium, the standard diet with double the concentration of agar of the standard fly diet and a layer of autoclaved yeast for acclimation. We doubled the concentration of agar to prevent females from embedding their eggs into the food, making egg collection easier. After the 24-hour acclimation, fresh medium was provided for egg-laying overnight. The eggs were collected the next morning to establish the next generation as described above. This means that adults were between 5–10 days old at the time of egg collection, and we may therefore have inadvertently selected for early life fecundity.

At the time of the experiment described below, generations of selection were as follows: 61 for the 18°C standard diet, 51 for the 18°C diluted and low P:C diets, 89 for the 25°C standard diet, 75 for the 25°C diluted and low P:C diets, 100 for the 28°C standard diet, 81 for the 28°C diluted diet, and 84 for the 28°C low P:C diet. Due to logistical limitations, three replicate lines were randomly selected from the five set up for each selection regime to perform the experiment described below. Prior to the experiment, all lines were placed in common garden conditions at 25°C on a standard diet for at least two generations to minimise parental effects (De Villemereuil et al. 2016).

Plastic and evolved shifts in larval thermotolerance

Experimental set up

After two generations of common garden rearing, adults were transferred to laying plates containing egg-laying medium and a surface layer of autoclaved yeast to induce oviposition. Laying plates were placed at 25°C for 14 hours to allow egg-laying. Eggs were collected from laying plates and transferred to vials containing 7.2 mL of standard diet, at a density of twenty eggs per vial (Agnew et al. 2002). Five replicate vials were set up for each replicate line per selection regime. First instar larvae (hatched 24 hours after egg transfer) were exposed to the hardening and acute stress treatments. Vials were maintained at 25°C under 12:12 hour light:dark cycles throughout the experiment, except when exposing larvae to the hardening and acute stress treatments described below. This common garden experimental design enabled us to investigate the effects of genetic adaptation to combined thermal-nutritional stress on larval thermotolerance and thermal plasticity.

Assessing induced (plastic) and basal thermotolerance

Larval cold plasticity was assessed by subjecting 24-hour-old first instar larvae, reared from embryos at 25°C on a standard diet, to a two-hour hardening treatment in a 0°C water bath, followed by a 24-hour recovery period at 25°C. After recovery, these 48-hour-old larvae were exposed to a four-hour acute cold shock at 0°C in a water bath and then returned to 25°C under 12:12 hour light: dark cycles to complete development. Basal cold tolerance was assessed by exposing 48-hour-old larvae, reared from embryos at 25°C on a standard diet, to a four-hour acute cold shock at 0°C in a water bath, without prior hardening treatment. The larvae were then allowed to complete development at 25°C under 12:12 hour light:dark cycles. 48-hour-old larvae were exposed to the acute cold shock treatment to equalise the stage at which larvae were exposed to the acute stress across the hardened and basal treatments.

Larval heat plasticity was assessed by subjecting 24-hour-old first instar larvae, again reared from embryos at 25°C on a standard diet, to a one-hour hardening treatment in a 35°C water bath, followed by a 24-hour recovery period at 25°C. After recovery, these 48-hour-old larvae were exposed to an acute heat shock of 39°C for 30 minutes. Those larvae were then allowed to develop at 25°C under 12:12 hour light:dark cycles until adult eclosion. Basal heat tolerance was assessed by exposing 48-hour-old larvae, reared from embryos on the standard diet at 25°C, to an acute heat shock treatment of 39°C for 30 minutes without any hardening treatment. They were then allowed to complete development at 25°C under 12:12 hour light:dark cycles.

A control treatment was also included, where vials contained larvae that were not exposed to any heat or cold stress treatments and were maintained on control food at 25°C under 12:12 hour light:dark cycles until adult eclosion.

The temperatures and durations for both heat and cold shock hardening and acute treatments were determined by pilot experiments (details in supplementary materials). Throughout these treatments, all vials were temporarily sealed with plastic caps and parafilm to prevent water leakage. After adult eclosion, larval-adult viability was subsequently measured as the ratio of the number of emerged adult flies to the initial egg number deposited in the vials.

Statistical analyses

All data analyses and visualisations were performed in R (R Core Team 2024). The larval-adult viability data from all treatments were fitted with generalised linear mixed-effects models with a binomial distribution using the lme4 package (Bates et al. 2015). All model weights were set to the number of eggs deposited per vial (20 eggs).

To test for the effects of long-term selection under combined thermal-nutritional stress on larval-adult viability under control (25°C standard diet) conditions, selection treatments (combinations of selection temperature and diet) were included as a fixed effect, while block (the day that eggs were picked to set up the experiment) and replicate line were included as non-nested (crossed) random effects to account for variation between blocks or variation among replicate lines; these random effects were included to improve model accuracy but were not tested for statistical significance (Bates et al. 2015).

To test for the effects of selection on larval basal thermotolerance and thermal plasticity, selection temperature, selection diet, hardening treatment (hardened and non-hardened), and their three-way interaction were included as fixed effects. Block and replicate line were treated as non-nested (crossed) random effects to account for any effect of block or variation among replicate lines; these random effects were not tested for statistical significance (Bates et al. 2015). Interactions between the random and fixed effects were also not tested because they were not core to our question.

Significance of the fixed effects was assessed using ANOVA using the car package (Fox et al. 2001), which uses the Wald Chi-square test to generate p-values for the fixed effects (Bolker et al. 2009; Rebolledo et al. 2023). Post hoc multiple comparisons were performed using the emmean package (Lenth 2023) to test for significant differences in larval-adult viability across levels of each factor. To quantify the amount of variation in the data explained by our models, we estimated conditional R² values which take both fixed and random effects in the model into account (Nakagawa et al. 2017) using the MuMIn package (Nakagawa and Schielzeth 2013), since traditional measures of effect size are not applicable to generalised linear mixed models (Nakagawa et al. 2017).