Proteomic data reveal a physiological basis for costs and benefits associated with thermal acclimation

Climate change has raised a great deal of interest in how animals and plants can adapt to expected future climates (Araújo et al., 2013; Chown et al., 2010). Numerous studies have suggested that changes could occur so fast that many species will not be able to adapt through evolutionary changes (Araújo et al., 2013; Hoffmann et al., 2013; Kristensen et al., 2015). Plastic responses, e.g. through physiological adaptation within the lifetime of an individual, is one alternative way to cope with extreme and fluctuating temperatures (Angiletta, 2009; Chown and Terblanche, 2006; Hoffmann and Parsons, 1997; Hoffmann et al., 2003). Studies on ectotherms investigating thermal acclimation and the physiological changes induced in response to acclimation typically do so by exposing adult organisms to short-term cold or hot temperatures and subsequently studying the consequences on thermal and molecular phenotypes (Overgaard et al., 2005; Sørensen et al., 2005; Malmendal et al., 2006; Teranishi and Stillman, 2007; Nota et al., 2010; Teets et al., 2012; Colinet et al., 2013; Jansen et al., 2013; Wei et al., 2015). Such studies have provided novel insights into the molecular aspects of acclimation responses and have added important components to our knowledge on associations between the genotype, the molecular phenotype and thermal resistance. The molecular fingerprints of short-term exposure of adult ectotherms to variable and sometimes stressful temperatures are important if we are to understand the background for thermal acclimation responses. However, in their natural habitats, Drosophila melanogaster and other ectotherms are exposed to thermal stress for extended periods e.g. during cold winters or hot summers (Izequierdo, 1991; Rodríguez-Trelles et al., 2013; Schou et al., 2015), making the molecular consequences of long-term exposure of different life stages highly ecologically relevant.

Analysis of the molecular response to temperature changes can be completed by profiling of the functional proteins and can thus provide insight into the regulation of biochemical pathways governing the functional phenotype (Feder and Walser, 2005). One of the first proteomic studies investigating thermal acclimation responses was the pioneering work on the Antarctic archaeonin Methanococcoides burtonii (Goodchild et al., 2004). Since then, technologies have been developing fast and state of the art methods now enable identification and quantification of sometimes thousands of proteins in organisms such as Escherichia coli (Soufi et al., 2015). In spite of this, studies on thermal acclimation responses in arthropods have rarely identified more than a few hundred proteins (Colinet et al., 2007,, 2013; Pedersen et al., 2010; Wei et al., 2015), making investigation of complex physiological changes in response to thermal acclimation difficult.

Here, we expose D. melanogaster to three different thermal regimes (12, 25 or 31°C) throughout development from egg to adult and at the first 2-3 days of their adult life, and use label-free quantitative shotgun proteomics on flies from the different thermal regimes. This approach enables us to identify and quantify a large number of differentially expressed proteins affected by the different long-term temperature exposures. Long-term acclimation responses have received little attention in previous proteomic studies, and with the possibility of identifying a large number of proteins, we can take previous proteomic analyses of thermal acclimation to the next level by performing enrichment analyses and thereby identify complex physiological changes. Results from the proteomic analysis were linked to functional phenotypes obtained on flies from the three temperatures, enabling us to discuss biochemical causes of costs and benefits associated with thermal acclimation.

The laboratory population used in this study was established from 589 inseminated females caught in Denmark in 2010 (for details, see Schou et al., 2014). Prior to the experiment, the population was maintained at 20°C at 12 h light:12 h dark photoperiod and reared with a standard Drosophila medium composed of yeast, oatmeal, sugar and agar. The parental flies used for egg production were density controlled during development (∼40 eggs per vial with 7 ml medium). When the parental flies were 4 days of age, they were transferred to bottles with the standard Drosophila medium but with a high agar concentration (3%). The following day, eggs were washed off the surface of the medium and distributed into vials in groups of exactly 40 eggs. This methodology was used to ensure random grouping of eggs into vials (Schou, 2013). Vials were allocated to each of three developmental temperatures: 12°C (12 vials), 25°C (8 vials) or 31°C (10 vials). All developmental regimes had a 12 h light:12 h dark photoperiod. On the day of emergence, flies were anaesthetized with CO₂ and counted to obtain estimates of egg-to-adult viability and developmental time. Then, flies were separated into sexes, females were discarded, and males, which were used for proteomics and for thermal resistance measurements, were relocated to the temperature at which they emerged.

For each developmental temperature, 20 male flies 2–3 days old (60±12 h) were tested for their critical thermal minimum (CT_min) and their critical thermal maximum (CT_max) (Overgaard et al., 2012). In these assays, flies were placed individually into sealed 6 ml glass vials and submerged in a water bath at 20°C. When assessing CT_max, the bath contained water, in which the temperature was increased at a rate of 0.1°C min⁻¹. Conversely, when assessing CT_min, the bath contained a mixture of ethylene glycol and water (1:1, v:v), in which the temperature was decreased at a rate of 0.1°C min⁻¹. Once a fly was immobile when provoked (with a flashlight or gently knocking on the vials with a stick), the temperature of the hot or cold bath was noted as the upper thermal limit (CT_max) or lower thermal limit (CT_min), respectively.

To investigate the effect of average temperature during development on CT_max and CT_min, we constructed linear models with temperature as a continuous variable and performed F-tests. There were no violations of the assumptions of normality of residuals and homogeneity of variances. Egg-to-adult viability was modelled with a logistic regression, and because D. melanogaster has a quadratic performance curve for this trait (e.g. Schou et al., 2015), temperature was modelled as such. We observed overdispersion in the model and corrected the standard errors using a quasi-generalized linear model (Zuur et al., 2009). Developmental time exponentially decreases with increasing temperature (David et al., 1983) and was modelled using a Poisson generalized linear mixed model with a log link and with ‘vial’ as a random effect. We detected underdispersion in the model, but found it appropriate not to perform any corrections (Zuur et al., 2009). P-values for the effect of temperature on egg-to-adult viability and developmental time were obtained by likelihood ratio tests. All statistical analyses were performed in R (v3.2.2; R Core Team, 2015).

Three replicate samples with 13 male flies at 2-3 days of age (60±12 h) per sample were obtained from each of the three rearing temperatures (12, 25 and 31°C). Proteins were extracted from each of the 9 samples using an optimized protocol based upon the B-PER reagent (Thermo Scientific; as described by Hansen et al., 2014). Proteins were denatured by heating samples at 95°C for 10 min in SDS-PAGE sample buffer supplemented with dithiothreitol (DTT). After boiling, samples were cooled to room temperature prior to loading them onto a pre-cast 4–15% gradient SDS-gel (BioRad). Samples were separated for 30 min at 160 V. Proteins were then digested in-gel (Shevchenko et al., 2007). Briefly, each gel lane was excised and pieces were then washed, reduced and alkylated prior to being digested with trypsin. Digested peptides were extracted, dried down and finally resuspended in 0.1% (v/v) trifluoroacetic acid (TFA), 2% acetonitrile (v/v). Peptides were desalted using the StageTip protocol (Rappsilber et al., 2007). The tryptic digests were resuspended in 0.1% (v/v) TFA, 2% acetonitrile (v/v).

Tryptic digests were analysed by an automated LC-ESI-MS/MS with an UltiMate 3000 RSLCnano system on-line coupled to a Q Exactive mass spectrometer via a Nanospray Flex ion source (Thermo Fisher). Settings were as previously described (Kjeldal et al., 2014), with the following modifications: samples were loaded at a flow rate of 4 μl min⁻¹ and eluted at a constant flow rate of 300 nl min⁻¹ in a 120 min linear gradient, ranging from 12% to 40% solvent B (100% acetonitrile). Survey scans were acquired at 400–1200 m/z, and no lock mass was used. The maximum ion injection time was set to 100 ms for MS and 75 ms for MS/MS scans. The automatic gain control for MS and MS/MS were 1E6 and 1E5, respectively and a dynamic exclusion of 30 s was used for minimizing repetitive selection of the same ions.

The raw data was analysed with MaxQuant (v1.5.1.2; Cox and Mann, 2008), applying the label-free quantification (LFQ) feature but otherwise using standard settings (data are available via ProteomeXchange with identifier PXD003755). These standard settings included peptide and protein false discovery rates (FDRs) of 1%, tryptic digestion with two allowed miss cleavages, oxidation of methionine as a variable modification and carboxymethylation of cysteine as a fixed modification. Data were cross-referenced with a database containing all proteins from D. melanogaster downloaded from UniProt (accessed 6 January 2015).

The generated proteingroups.txt file was imported into Perseus (v1.5.0.31) (part of the MaxQuant package), LFQ values were log₂-transformed and filtered for contaminants and reverse hits before being subjected to statistical analysis. Differences in the expression levels of single proteins between the three temperature regimes were investigated using one-way analysis of variance (ANOVA) followed by a multiple testing correction using a permutation-based FDR of 0.05. The individual P-values between groups were calculated using a Student's t-test. Statistically significant P-values (P≤0.05) were corrected for multiple testing using the permutation-based FDR of 0.05. Hierarchical clustering in Perseus (v1.5.0.31) was performed with Pearson correlation distance with average linkage after Z-scoring of LFQ values. A multi scatter plot was constructed using the log₂-transformed LFQ values of each protein.

Statistically significant differentially regulated proteins (only proteins repetitively found in at least two out of three replicates were included in the analyses) were uploaded to DAVID bioinformatics resources (v6.7) (Huang et al., 2009a,,b) and functionally annotated. The output from DAVID was visualized in Cytoscape (v3.2) (Shannon et al., 2003) using the EnrichmentMap plugin (v2.0.1.) (Merico et al., 2010) using a P-value cut off of 0.005 and a FDR Q-value cut off of 0.05.

The effects of developmental temperature on the four traits investigated were strong and all highly significant (Table 1). Thermal tolerance traits increased linearly with increasing developmental temperature; however, at very different rates (Table 1; Fig. 1A,B). CT_min increased (cold tolerance decreased) with ∼0.5°C per degree increase in developmental temperature, whereas CT_max increased (heat tolerance increased) with only 0.2°C per degree increase in developmental temperature (Table 1). Development from egg to the adult stage at 12°C lasted 50.4±1.6 days, whereas development at 25 and 31°C were faster and lasted 8.9±0.1 and 7.9±0.1 days, respectively (Fig. 1C). Egg-to-adult viability at 12 and 31°C was 66.6% and 92.9%, of the viability at 25°C, respectively (Fig. 1D).

From the proteomic analysis, a total of 1440 proteins could be quantified from D. melanogaster developed and reared at 12°C (1066), 25°C (963) and 31°C (1292), respectively (for full protein lists, see Table S1). Replicates (n=3) within each temperature correlated well despite a modest overlap (Fig. 2 and Fig. 3A–C, Fig. S1). Approximately 54% (789) of all the quantified proteins were shared between the three different rearing regimes (Fig. 3A). Overall, only a small overlap (19 proteins) of the differentially expressed proteins was observed between the three contrasts (25°C vs 12°C, 31°C vs 25°C, and 31°C vs 13°C; Fig. 3B). With increasing temperature, both the number of identifications and number of differentially expressed proteins increased. Contrasting flies reared at 12°C and 25°C revealed 48 proteins (minimum no. unique peptides ≥1) that were differentially regulated (FDR-corrected P≤0.05; Fig. 3B). At 25°C vs 31°C, 132 proteins were differentially regulated (FDR-corrected P≤0.05) and, finally, when comparing 12°C and 31°C, 223 proteins were differently regulated (FDR-corrected P≤0.05; Fig. 3B).

The proteomic data was subjected to a functional annotation enrichment analysis in order to identify biological processes, molecular functions and cellular components that were statistically enriched (P≤0.005, FDR Q≤0.05) in response to temperature change (Fig. 4). Among the proteins that were differentially upregulated when increasing the temperature from 12°C to 25°C (38 proteins), several functional clusters related to general metabolism/catabolism, energy production, structural components and proteins related to specific enzyme families (hydrolases and oxidoreductases) were enriched (Fig. 4). Proteins that were downregulated in flies from 25°C compared with flies from 12°C (10 proteins), did not enrich for any statistical significant functional cluster (i.e. P≥0.005, FDR Q≥0.05). The differentially downregulated proteins included a heat shock protein (P29844) and a CAMP-dependent kinase (P12370), involved in biological processes related to behavioural responses to stimuli (GO:0007610) and adult behaviour (GO:0001654), respectively. Complete protein and GO annotation lists, as well as functional enrichment analysis can be found in Table S1.

Proteins differentially upregulated at 31°C compared with 25°C (68 proteins), were functionally enriched for a major cluster consisting of proteins related to the proteasome and a minor cluster consisting of phosphoproteins (Fig. 4). The differentially downregulated proteins (56 proteins) formed several clusters functionally related to carbohydrate metabolism/catabolism, energy production, cofactor binding and coenzyme production (pyridoxal phosphate) (Fig. 4).

Contrasting flies developed and reared at 12°C and 31°C, respectively, revealed that the increase in temperature resulted in the differential upregulation of proteins (185 proteins) related to the proteasome, energy production, chemical and cellular homeostasis production of precursor metabolites, and calcium binding proteins (Fig. 4). The differentially downregulated proteins (38 proteins) were mainly related to cofactor binding (Fig. 4).

Several proteins were unique (only identified and quantified) in flies from one of the three temperatures, 12°C (5 proteins), 25°C (28 proteins) and 31°C (123 proteins). Statistically significant enrichment clusters (P≤0.005, FDR Q≤0.05) could only be formed for proteins unique to 31°C, revealing five functional clusters comprising proteins related to the proteasome, cytoplasm, hydrolases, thiol proteases and ribonucleoproteins (Fig. S2).

An enrichment analysis was also performed on proteins that were shared between the three different temperatures and for which no statistical significant change in abundance was observed (460 proteins). This enrichment analysis of shared functionality between the three temperatures revealed a multitude of functional clusters most notably clusters related to common metabolic processes, including carbohydrate metabolism, oxidation/reduction reactions and processes related to structural components (Fig. 5). A cluster of several stress-related proteins was also enriched among the proteins for which the abundance was not modulated by changing the developmental and rearing temperature (Fig. 5).

In this study, we investigated proteomic profiles, coupled with thermal resistance and life history traits, of D. melanogaster flies developed and reared at cold, benign and hot temperatures. Not surprisingly egg-to-adult viability and developmental time were affected by developmental temperature. Development was slower at lower temperatures and egg-to-adult viability was highest at benign temperatures, which is in accordance with well-established results (David and Clavel, 1969; David et al., 1983). Furthermore, we showed that flies reared at 12°C had a CT_min that was 5.3 and 7.0°C lower than flies reared at 25 and 31°C, respectively (Fig. 1). By contrast, flies developed at 31°C had a CT_max that was 0.6 and 2.9°C higher than flies developed at 25 and 12°C, respectively (Fig. 1). These results are in accordance with other studies showing marked benefits and costs across tolerances with thermal acclimation (Chidawanyika and Terblanche, 2011; Chown and Terblanche, 2006; Kristensen et al., 2008; Loeschcke and Hoffmann, 2007; Schou et al., 2015). Consequences of climate change include higher and more variable temperatures in many parts of the world (IPCC, 2013). Thus the strong trade-off between tolerances observed here and in other studies clearly suggests that thermal fluctuation constitutes a challenge to ectotherms in nature (Kristensen et al., 2008; Schou et al., 2015) and will potentially be even more of a challenge in the predicted future thermal environment.

Our data also provide novel insights into the underlying physiology of the strong acclimation responses observed in the functional phenotypes. Replicates within temperatures clustered neatly and the three rearing temperatures were clearly separated, with flies from 12°C and 25°C having a more similar proteomic profile compared with flies reared at 31°C (Figs 2, 3 and Fig. S1). Thus despite cold acclimation inducing the strongest functional phenotypic response in terms of increased cold resistance, decreased egg-to-adult viability and developmental time (Fig. 1), effects on the proteomic level were small compared with effects of heat acclimation (Figs 2–4). This finding is similar to that of previous studies on the transcriptional and proteomic levels (Colinet et al., 2013; Overgaard et al., 2014) and suggests a non-linear relationship between cellular acclimation responses and functional phenotypic responses.

The proteomic profile of flies reared at 31°C compared with flies reared at 25°C was characterized by downregulation of proteins and multiple clusters associated with carbohydrate catabolism, carbohydrate metabolism, energy production and transport, and by upregulation of the proteasome (Fig. 4, Table S1). Thus, the data suggest that rearing at high and stressful temperatures leads to reduced conversion of carbohydrates and therefore probably lower energy production (such as ATP). In contrast to metabolic proteins, proteasome proteins involved in repair and degradation of oxidative or structural (misfolding) damaged proteins are upregulated in flies reared at 31°C compared with flies reared at 25°C (Fig. 4, Table S1). Among the upregulated proteins are several heat shock proteins known for being part of the well-described heat shock response (Colinet et al., 2013; Kültz, 2005; Sørensen et al., 2003), as well as proteins not previously associated with thermal stress responses (Table S1).

Despite the observed downregulation of proteins involved in key metabolic processes, basal metabolic rate increases in accordance with expectations for Q10 when exposing D. melanogaster to temperatures within the range investigated here (Jensen et al., 2014). Upregulation of the proteasome, including molecular chaperones is energetically costly (Kültz, 2005; Sørensen et al., 2003). Specifically, upregulation of heat shock proteins (Hsps) has been shown to be associated with downregulation of many other proteins, partly because it is energetically demanding (Sørensen et al., 2003). Thus, flies reared at high and stressful temperatures are likely to be in a state of energetic deficiency, i.e. there is a mismatch between supply and demand (Schulte, 2015). Whether downregulation of proteins involved in key metabolic processes in flies exposed to 31°C is part of an adaptive cellular stress response or a maladaptive consequence of exposure to stressful high temperatures is unresolved. It is, however, well known that induction of Hsps has a negative impact on development and survival, cell division and energy balance in Drosophila (Feder et al., 1992; Krebs and Feder, 1997; Krebs and Loeschcke, 1994; Silbermann and Tatar, 2000). We therefore propose that upregulation of proteasome proteins is an important adaptive process at stressful high temperatures, which increases heat tolerance by ridding cells of potentially harmful substances that arise as a result of misfolding of proteins and post-transcriptional and post-translational changes due to thermal damage. This occurs at the expense of downregulation of basal metabolic functions. The explanation fits in well with observations that longevity and reproduction decrease sharply at temperatures above 29–30°C in D. melanogaster (David et al., 1983; Linford et al., 2013; Miquel et al., 1976; Schnebel and Grossfield, 1984).

Previous studies have proposed a general adaptive stress response in which the same mechanisms are invoked across different stressors (e.g. Kültz, 2005). Our observation that very few proteins overlap when differentially expressed proteins are compared in the 25°C vs 31°C, 31°C vs 12°C and 25°C vs 12°C groups (Fig. 3B) provide evidence against such common physiological plastic responses in cold and hot thermal environments (see also Bubliy et al., 2012). Despite this general observation, we did identify a small number of proteins that were differentially expressed in all temperature comparisons (Fig. 3B). These proteins are likely to be under strong selection in a more variable climate in which the organism is faced with interchanging heat and cold stress, and are thus candidate genes of strong evolutionary interest that should be investigated more thoroughly in future studies. This group of proteins included proteins related to signal transduction (P61209), energy (O01666, Q9VQ29 and Q9W2X6) and carbohydrate metabolism (P07486, Q9VKX2, Q9W1G0, Q9W3Z3 and X2JF40), as well as proteins related to various stress responses, including salvage pathways (Q8SXQ1 and Q8IGX3), protein folding (Q9VFN5) and proteasomal degradation (Q9V5C6).

Comparing the proteomic profiles of flies reared at 12°C and 31°C revealed that multiple clusters associated with cellular homeostasis, the proteasome, generation of precursor metabolites and energy, and calcium binding were upregulated in flies reared at 31°C (Fig. 4). By contrast, some proteins involved in co-factor binding were downregulated in flies reared at 31°C compared with 12°C (Fig. 4). The majority of the significant clusters in this contrast are thus upregulated at 31°C (Figs 2 and 4, Table S1). This is unlike the 25°C and 31°C comparison, where most clusters were downregulated at 31°C (Fig. 4). However, common to the two contrasts is an upregulation of the proteasome and a downregulation of cofactor binding at 31°C (Fig. 4). In temperate climates, D. melanogaster is known to survive winters in the adult state, where they are often exposed to temperatures below 10°C for several months (Izequierdo, 1991; Schou et al., 2015), resulting in a longer lifespan (Miquel et al., 1976). This suggests fundamental differences in the type of stress induced by 12°C and 31°C, which is supported by the proteomic data (Figs 2–4 and Table S1). For example, flies at 12°C are unlikely to suffer from protein damage, whereby the advantage of inducing the proteasome stress response is small (Fig. 4, Table S1). Other significant clusters and proteins upregulated in flies developed at 31°C compared with 12°C are involved in maintaining cellular homeostasis, generation of energy and calcium binding (Fig. 4). The changes in such processes are likely to be related to the increased metabolism in flies reared at 31°C compared with 12°C (Jensen et al., 2014; Schulte, 2015).

Comparison of proteomic profiles of flies reared at 12°C and 25°C revealed upregulation of all identified clusters in flies from 25°C, including clusters with proteins involved in major metabolic processes (Fig. 4), especially proteins involved in glycoside hydrolase, carbohydrate metabolism and catabolism, as well as structural components, oxidoreductase and energy production (Fig. 4). At 12°C, we detected few upregulated proteins (and no clusters) and none of these was part of the classical stress response (Kültz, 2005; Sørensen et al., 2003). The number of proteins uniquely expressed at 12°C was also low (Fig. 3A, Table S1). This suggests that although development at 12°C is stressful because only a fraction of the flies reach the adult stage, the adult life stage in itself is not under the same degree of stress. Thus the proteomic effect of rearing at low temperatures is minor and the functional phenotypic effects in terms of increased cold resistance, decreased heat resistance and increased developmental time are strong (Fig. 1). The lower basal metabolic rate in ectotherms at low compared with higher temperatures (Jensen et al., 2014; Schulte, 2015) may be important for understanding the costs and benefits associated with cold acclimation (Schulte, 2015). Flies acclimated to low temperatures have a generally low metabolic rate, which might provide these flies with benefits at low temperatures. Transferring the same flies to high temperatures will result in strong costs in survival as a result of reduced heat resistance (and vice versa with flies reared at high temperatures). Distinguishing cause from effect is a challenge that we cannot solve here, but our data support the idea that metabolic rate is key to understanding the strong costs and benefits typically observed in studies on thermal acclimation (Angiletta, 2009; Chidawanyika and Terblanche, 2011; Kristensen et al., 2008; Schou et al., 2015; Schulte, 2015).

Three lines of evidence in this study suggest that the physiological acclimation responses to cold and hot temperatures are clearly distinct (Bubliy et al., 2012): (1) the proteomic differentiation between flies from 12°C and 31°C; (2) the very low overlap observed when comparing differentially expressed proteins at 25°C vs 31°C with those differentially expressed at 25°C vs 12°C; and (3) the fact that 258 unique proteins are identified in flies reared at 31°C and only 38 unique proteins are identified in flies reared at 12°C. This provides a physiological explanation for the costs observed when exposing ectotherms to temperatures that are different to acclimation temperatures (Chidawanyika and Terblanche, 2011; Chown and Terblanche, 2006; Kristensen et al., 2008; Loeschcke and Hoffmann, 2007; Sørensen et al., 2013), as well as the trade-off between CT_min and CT_max observed in this study. More specifically, we found evidence that upregulation of proteasome proteins is an important adaptive process at stressful high temperatures, but this is at the expense of downregulation of basal metabolic functions. Thus, the interplay between metabolism or metabolic rate and the proteasome is key to the adaptive response across the thermal range.

We are grateful to Volker Loeschcke for his unfailing encouragement and support and to Doth Andersen and Helle Blendstrup for technical assistance.