Drosophila, like most insects, are susceptible to low temperatures, and will succumb to temperatures above the freezing point of their hemolymph. For these insects, cold exposure causes a loss of extracellular ion and water homeostasis, leading to chill injury and eventually death. Chill-tolerant species are characterized by lower hemolymph [Na+] than chill-susceptible species and this lowered hemolymph [Na+] is suggested to improve ion and water homeostasis during cold exposure. It has therefore also been hypothesized that hemolymph Na+ is replaced by other ‘cryoprotective’ osmolytes in cold-tolerant species. Here, we compared the hemolymph metabolite profiles of five drosophilid species with marked differences in chill tolerance. All species were examined under ‘normal’ thermal conditions (i.e. 20°C) and following cold exposure (4 h at 0°C). Under benign conditions, total hemolymph osmolality was similar among all species despite chill-tolerant species having lower hemolymph [Na+]. Using NMR spectroscopy, we found that chill-tolerant species instead have higher levels of sugars and free amino acids in their hemolymph, including classical ‘cryoprotectants’ such as trehalose and proline. In addition, we found that chill-tolerant species maintain a relatively stable hemolymph osmolality and metabolite profile when exposed to cold stress while sensitive species suffer from large increases in osmolality and massive changes in their metabolic profiles during a cold stress. We suggest that the larger contribution of classical cryoprotectants in chill-tolerant Drosophila plays a non-colligative role for cold tolerance that contributes to osmotic and ion homeostasis during cold exposure and, in addition, we discuss how these comparative differences may represent an evolutionary pathway toward more extreme cold tolerance of insects.

Some cold-adapted insect species have evolved the ability to either tolerate or avoid ice crystal formation in the hemolymph and thus survive at sub-zero temperatures (Zachariassen, 1985; Bale, 1996; Ramløv, 2000). Most insects are chill susceptible, however, meaning they succumb to low temperature exposures that are unrelated to the risk of their hemolymph freezing (Zachariassen, 1985; Bale, 1996; Sinclair, 1999; MacMillan and Sinclair, 2011a; Andersen et al., 2015). The thermal sensitivity of these ‘chill-susceptible’ species is therefore not linked to damage caused by the water–ice transformation or recrystallization but instead to physiological problems induced by the low temperature, per se.

Increasing evidence indicates that chill injuries in insects accumulate in association with a loss of ion and water homeostasis at low temperatures (e.g. Koštal et al., 2004; MacMillan and Sinclair, 2011b; MacMillan et al., 2014, 2015a,c,d). Under normal conditions, extracellular ion balance is maintained by active transporters in cell membranes and epithelia, where hemolymph [Na+] is kept high and [K+] is kept low by a continuous active transport that balances the passive leak of these ions. Low temperature impairs the activity of ion pumps, causing Na+ to leak away from, and K+ toward, the hemolymph (Koštal et al., 2004; MacMillan and Sinclair, 2011a,b; MacMillan et al., 2015a,b). A number of studies have further shown that hemolymph volume is reduced during cold exposures. This reduction in volume, which is probably caused by water following Na+ osmotically, has the end result that hemolymph [K+] increases in the reduced hemolymph volume (Koštal et al., 2004; MacMillan and Sinclair, 2011b; MacMillan et al., 2015a,,d). Membrane resting potential is highly dependent on the cellular [K+] gradient (Hoyle, 1954; Wareham et al., 1974; Fitzgerald et al., 1996) and hemolymph hyperkalemia therefore causes cellular depolarization that leaves the neuromuscular system unexcitable and slows chill coma recovery time; ultimately, hyperkalemia may cause cold-induced injury and death (Boutilier, 2001; Koštal et al., 2004; Findsen et al., 2013; MacMillan et al., 2015c).

This cascade of events, starting with Na+ leaving the hemolymph and ending with a depolarization of the cell membranes due to hyperkalemia, is mechanistically linked to the initial electrochemical gradient of Na+ from the hemolymph to the gut and other tissues (see discussion in MacMillan and Sinclair, 2011b and MacMillan et al., 2015a,,d). In support of this hypothesis, a lower [Na+] gradient (driven by a lower hemolymph [Na+]) is associated with increased cold hardiness following different forms of cold acclimation in Drosophila melanogaster (MacMillan et al., 2015a) and strikingly similar observations have been made recently when comparing multiple Drosophila species with different cold tolerances (MacMillan et al., 2015b,,d). What remains unknown, however, is whether these observed differences in [Na+] contribute to general differences in hemolymph osmolality (such that chill-tolerant species generally have lower hemolymph osmolality) or whether this ‘missing’ Na+ is ‘replaced’ with other compatible osmolytes in chill-tolerant species. Measurements on D. melanogaster from different acclimation treatments suggest that the hemolymph osmolality is similar (or slightly larger) in cold-acclimated animals, and therefore reductions in [Na+] are matched or even exceeded by accumulation of other compatible osmolytes in the hemolymph (MacMillan et al., 2015a). Here, we examined whether a similar pattern is found among species of the same genus, and further we examined which osmolytes may replace Na+ in the hemolymph of cold-tolerant Drosophila species.

A number of previous studies have examined the metabolite profile of chill-susceptible insect species before and after cold exposure. Overall, these results suggest that both cold acclimation and cold exposure cause an increase in the relative abundance of some sugars (trehalose, maltose and glucose), polyols (glycerol and sorbitol) and free amino acids (proline and β-alanine) (Chen et al., 1987; Yoder et al., 2006; Michaud and Denlinger, 2007; Lalouette et al., 2007; Overgaard et al., 2007; Koštal et al., 2011a; Colinet et al., 2012; Purać et al., 2016). While these results might offer clues as to which metabolites could replace Na+ in the hemolymph of the cold-tolerant Drosophila species they are all based on whole-body samples and therefore do not provide any information on the metabolite profile of the hemolymph specifically. The main goals for the present study were: (1) to investigate the osmolality of the hemolymph in five Drosophila species: D. birchii, D. equinoxialis, D. melanogaster, D. persimilis and D. montana, before and after a cold exposure; and (2) to analyze the composition of small organic molecules in the hemolymph of the same species using NMR-spectroscopy. As the accumulation of chill injuries seems to be linked to a general failure to maintain ion gradients across membranes, we hypothesized that the cold-sensitive species suffer from larger shifts in ion and water balance and hence experience larger increases in osmolality than the more cold-tolerant species. Further, we hypothesized that chill-tolerant Drosophila species will constitutively maintain higher concentrations of compatible osmolytes (sugars, polyols and/or amino acids) compared with chill-sensitive species, in effect ‘replacing’ [Na+].

Animal stock

The Drosophila species used in the present study were D. birchii, D. equinoxialis, D. melanogaster, D. persimilis and D. montana. These species were selected from different species subgroups (obscura, melanogaster, virilis, montium and willistoni) within the Drosophila phylogeny, with four species from the subgenus Sophophora and one species from the subgenus Drosophila. The choice of species was therefore an attempt to avoid strong phylogenetic bias, but was also based on the fact that we had already generated considerable physiological and ecological information on the cold biology of these five species. The five species differ greatly in their chill tolerance, with D. birchii being very chill susceptible (LTe50=−3.3°C, where LTe50 is the temperature that causes mortality in 50% of individuals after a 2 h exposure) and D. montana being more chill tolerant (LTe50=−13.2°C; Andersen et al., 2015). The five species of Drosophila also maintain very different hemolymph [Na+] when reared under the same conditions, such that D. birchii maintain much higher [Na+] than D. montana (94 and 63 mmol l−1,respectively), and the hemolymph [Na+] correlates very tightly with cold tolerance (MacMillan et al., 2015d).

Laboratory populations of these species were kept at a constant temperature of 20°C and a 12 h:12 h light:dark cycle. Flies were maintained in 250 ml plastic bottles (ca. 50–250 per bottle) containing 50 ml of Leeds Drosophila medium (oatmeal–sugar–yeast–agar). Groups of parental flies were given an appropriate amount of time (from 4 h to 4 days) to lay eggs in the medium, after which they were tipped into a new bottle. New adult flies were collected on the day of emergence into smaller vials containing 4–7 ml of Drosophila medium; 3–5 day old adult flies were then sexed using a light CO2 anesthesia (less than 5 min) and females were kept for 3 days to recover before they were used in the experiments. Accordingly, all flies were 6–8 day old females at the time of the experiments.

Experimental protocol

Separate experiments were conducted to investigate the osmolality before and after cold exposure in the five Drosophila species, and to characterize the interspecific variance in metabolomic profile before and after cold exposure in these species. These findings were analyzed and are discussed in relation to a number of other ecological and physiological traits that we have recently characterized (within the same year) on the same fly stocks. For example, we associate the physiological parameters investigated here in relation to recent measures of species cold tolerance (Andersen et al., 2015) and furthermore we discuss the observed differences in osmotic balance in the light of recent measurements of ion content and hemolymph volume on the same fly lines (MacMillan et al., 2015d).

A common experimental protocol was used for the experiments, characterizing the osmotic and metabolomic status before and after cold exposure. Flies were either taken directly from standard conditions (20°C; control group) or exposed to a cold treatment (4 h at 0°C) by submerging them in an ice/water slurry (in plastic vials sealed with screw caps). Osmolality could be measured on hemolymph samples from individual flies (see below). For assessment of osmolality, between 9 and 16 flies were sampled for each species and treatment combination. To achieve sufficient hemolymph volume for measurements of hemolymph metabolite profile, flies were again randomly distributed to one of the two treatments (control or cold exposed) but here we pooled hemolymph from 13–15 flies for each sample to be used in the subsequent NMR spectroscopy. Six samples were prepared for each species and treatment [i.e. 90 flies (6×15) for each species and treatment combination].

Hemolymph extraction

Hemolymph was extracted using antennal ablation as described by MacMillan and Hughson (2014). Here, the fly was pushed head first into a 10 μl pipette tip using positive air pressure in a system composed of rubber tubing and three-way stopcocks. When the fly was positioned in the pipette tip, the pipette was cut directly in front of the fly's head to expose the antennae. One antenna was removed using sharp forceps and the positive air pressure then caused a clear droplet of hemolymph to emerge from the wound. The hemolymph was immediately collected for either osmolality measurements or NMR spectroscopy (see ‘Sample preparation for NMR measurements’, below, for details). Flies exposed to the cold treatment were preloaded individually into the pipette tips before cold exposure to allow for rapid sampling after cold exposure. The pipette tips containing the flies were then put in larger bottles, which were sealed before they were placed in the ice/water slurry.

Hemolymph osmolality

Osmolality measurements were obtained using a Clifton nanoliter osmometer (Clifton Technical Physics, Hartford, NY, USA) through measurements of melting point. A pulled borosilicate glass capillary was used to inject approximately 5 nl of hemolymph from individual flies into immersion oil suspended in a platinum disc. When all samples were loaded, the controller box was set to ‘freeze’ causing all sample droplets to flash freeze. The droplets were then slowly warmed while observing the ice crystals and osmolality was determined from the melting point (i.e. the temperature at which the last tiny ice crystal disappeared). We did not observe any signs of thermal hysteresis and as the melting point of an aqueous solution is decreased by 1.86°C with each mole of solute, the melting point was used to determine the hemolymph osmolality through reference to standards of known osmolality.

Sample preparation for NMR measurements

In an attempt to obtain quantitative measurements of hemolymph composition by NMR-spectroscopy, we used a modified method following the procedures developed by Ragan et al. (2013), who recently described a method for NMR-based quantification of metabolites in larval Drosophila hemolymph. However, it is possible to sample considerably more hemolymph from larvae (>200 nl per larva) while antennal ablation only renders 10–50 nl of hemolymph because the adult flies have less hemolymph volume available (depending on species and treatment). To obtain sufficient hemolymph volume, each experimental sample was composed of hemolymph pooled from 13–15 flies. A Petri dish with wax in the bottom was filled to the top with paraffin oil, and the dish was placed on a cooling plate at 5°C. A 20 μl drop of cold saline with 100 μmol l−1 sodium 13C-formate was placed in the dish using a Hamilton syringe, so that it was completely covered in oil to prevent evaporation. Hemolymph samples were prepared as described above and after antennal ablation the hemolymph droplets were immediately dipped into the oil bath and merged with the 20 μl drop of saline. Sampling of the 13–15 flies needed for each sample took maximally 90 min and after the last fly was sampled 17.5 μl of the saline/hemolymph droplet was removed using a Hamilton syringe. This sample was then added to 200 μl of deionized water containing 80 μmol l−1 trimethylsilyl propionate [TSP; note, the original experiments by Ragan et al. (2013) used 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS)]. The sample was then centrifuged through a 0.22 μm Millipore filter at ∼2800 g for 2 min in order to remove any cells. To control for any evaporation or contamination during the procedure, a 20 μl drop of saline was left under the oil for 40–90 min, after which 17.5 μl was removed and treated as described for the rest of the samples. These samples are henceforth referred to as the blank samples. After the centrifugation, a methanol–chloroform metabolite extraction was performed using the method of Bligh and Dyer (1959) in order to purify to the hydrophilic metabolites and separate them from the lipophilic metabolites. The solution was transferred to an Eppendorf tube containing 750 μl of 2:1 methanol:chloroform and vortexed thoroughly. Then, 250 μl of water and 200 μl of chloroform were added to obtain phase separation. The aqueous layer (upper layer) containing the hydrophilic metabolites was isolated and evaporated to dryness using N2 gas, after which the samples were kept in the freezer until they were run on the NMR spectrometer.

NMR measurements

Immediately before NMR measurements, samples were rehydrated in 50 µl of 50 mmol l−1 phosphate buffer (pH 7.4) in D2O, and 30 µl was transferred to 1.7 mm NMR tubes. Two samples were lost during sample transfer so the final dataset consisted of 64 samples (58 biological samples and 6 blank samples). NMR measurements were performed at 300 K on a Bruker Avance III 600 spectrometer (Bruker Biospin, Rheinstetten, Germany), operating at a 1H frequency of 600.13 MHz, equipped with a 1.7 mm TXI cold probe. 1H NMR spectra were acquired using a 1D NOESY-sequence with water signal suppression by pre-saturation during relaxation delay (4.0 s) and mixing time (100 ms). A total of 256 transients of 32k data points spanning a spectral width of 20 ppm were collected (acquisition time 1.36 s), resulting in a total acquisition time of 26 min per sample.

Data analysis

Osmolality data

Two-way ANOVA was used to test for effects of species and treatment (before/after cold) on hemolymph osmolality. Because initial analysis showed that the data were not normally distributed, the osmolality data were log-transformed. Further, a number of correlation analyses using Pearson's correlation tests were done to examine whether the osmotic differences among species were correlated to cold tolerance. In these correlations, we referenced the newly obtained data on osmolality with recent data on hemolymph Na+ and K+ concentrations (MacMillan et al., 2015d) and LTe50 (Andersen et al., 2015).

NMR data

Data reduction and statistical analysis

All spectra were referenced to the TSP signal at −0.017 ppm and baseline corrected. Unfortunately, the signal intensity of TSP showed very large variations (presumably due to interactions with larger molecules), such that we could not obtain absolute metabolite concentrations as in Ragan et al. (2013). Before analysis, we removed the remaining water signal (5.1–4.7 ppm) as well as the high- and low-field ends of the spectrum (including TSP at −0.017 ppm). The blank samples (containing no hemolymph) showed a fairly high degree of contamination (presumably from the oil), including very strong signals from an unidentified polyol. The contaminated regions (8.60–8.27, 5.63–5.59, 5.34–5.27, 4.90–4.27, 4.12–4.05, 3.90–3.56, 1.91–1.27 ppm) were removed from the entire dataset. The remaining signals in the spectra were normalized to total intensity to allow for analysis of relative concentrations.

In the initial analysis of the metabolomics data, we wanted to identify metabolic ‘fingerprints’ specific to species and/or treatment. This was done using principal component analysis (PCA), where variation in large and complex datasets can be reduced to a few components explaining a large fraction of the variation. For the PCA, the data were centered by subtracting the mean intensity of each signal, and scaled using Pareto scaling (Craig et al., 2006) such that the intensity at each frequency was divided by the square root of the standard deviation at that frequency. This analytical approach is a compromise between giving low-concentration metabolites reasonable weight and not putting too much weight on noise. The initial PCA identified one sample, an extract from D. equinoxialis kept at room temperature, as an outlier, and this was removed from all further analyses.

To further investigate species differences in the metabolite profiles, we performed an O2PLS (orthogonal partial least squares) discriminant analysis (O2PLS-DA). This analysis resembles PCA, but components are structured to maximize their correlation with group membership (i.e. species and/or treatment) rather than explaining maximal variation. The O2PLS-DA models were validated by cross validation where models were made with randomly chosen groups of samples left out one at a time, and group membership was predicted for the left out samples. All O2PLS-DA and orthogonal partial least squares (OPLS) plots show cross-validated values. As the O2PLS-DA analysis gave a clear separation of species and because this separation correlated very well with cold tolerance, we interpreted the loadings of this analysis to identify the metabolites responsible for the variation in PC1 and PC2 (and hence also cold tolerance). Metabolites were identified based on earlier assignments in hemolymph from D. melanogaster larvae (Ragan et al., 2013), whole flies (Malmendal et al., 2006; Pedersen et al., 2008) and metabolomics databases (Cui et al., 2008; Wishart et al., 2007) (see Fig. S1 and Table S1).

To characterize the variation among species in the response to the cold treatment, the median spectrum from the control (unexposed) samples from that species were subtracted from each of the cold-exposed spectra [exposed−median(control)]. This yielded a new set of spectra, which were then submitted to PCA modeling, in order to analyze the variation among species. Here, the data were not centered; hence, zero in the PC scores and loadings plots corresponds to no change in the metabolomics profile after cold exposure.

To test whether the metabolite profiles of the five Drosophila species showed correlations with their cold tolerance, two OPLS models were made, the first using the metabolite profiles from the untreated group, and the second using the response to cold exposure [exposed−median(control)]. When cross-validating the OPLS models, all samples for each species were left out one at a time, so that the cold tolerance for that ‘left out’ species could be predicted only based on the relationship between cold tolerance and metabolite concentration for the other four species. This analysis was repeated for all five species and predictability of the model (Q2), which is the correlation between predicted and actual cold tolerance, was used as a measure of the quality of the model. In this analysis, we also made loading plots in order to identify specific metabolites responsible for the change in metabolite profile associated with cold. The loadings and the correlation coefficient (R2) between intensities at the individual frequencies and the predictive component were calculated for the entire spectrum, including the regions that were not included in the model. A cutoff value for R2 corresponding to P<0.05 with Bonferroni correction for an assumed number of 30 metabolites was used for identification of significant metabolites. All multivariate analysis was performed using Simca software (Umetrics, Umeå, Sweden).

Cold tolerance, ion balance and hemolymph volume

Cold tolerance of the five Drosophila species used here was recently characterized in our laboratory (Andersen et al., 2015). Because these previously published measurements were performed on the same populations and within the same year as the present study, we can draw from these observations when relating differences in osmotic balance and metabolites to cold tolerance. We also recently characterized the ion concentrations and hemolymph volume before and after a cold exposure identical to that used in the present study (MacMillan et al., 2015d). These values are also presented here for reference to the discussion of osmotic balance and metabolite changes (Table 1). In the present study, we focused on the cold-tolerance trait LTe50, which represents the temperature that causes mortality in 50% of individuals after a 2 h cold exposure. LTe50 was chosen because it represents an ecologically important measure of cold-induced injury, which we found to be a very powerful correlate of drosophilid latitudinal distribution (Andersen et al., 2015). As seen in Table 1, LTe50 varied considerably among the five species: LTe50 was much lower for the two temperate species (D. montana and D. persimilis), intermediate for the widespread species (D. melanogaster) and high for the two tropical species (D. birchii and D. equinoxialis). Table 1 also shows that cold-tolerant species were characterized by a lower initial [Na+] in the hemolymph and a smaller perturbation of hemolymph volume and [K+] after cold exposure (see also MacMillan et al., 2015a,,d, for further discussion).

Table 1.

Cold tolerance and hemolymph characteristics of five temperate and tropical Drosophila species

Cold tolerance and hemolymph characteristics of five temperate and tropical Drosophila species
Cold tolerance and hemolymph characteristics of five temperate and tropical Drosophila species

Osmolality and ion concentration

When species were measured under benign conditions (20°C), hemolymph osmolality of the five Drosophila species ranged from 369.5±10.2 to 426.3±28.8 mOsm (Fig. 1A, Table 1) and in all species except D. persimilis the osmolality of the hemolymph increased significantly following cold treatment (two-way ANOVA: F=63.5, P<0.001). There was no significant correlation between cold tolerance and hemolymph osmolality before cold exposure (Pearson’s correlation test; R2=0.50, P=0.14; Fig. 1A), but there was a non-significant tendency for the osmolality to be higher in the less cold-tolerant species after cold exposure (Pearson’s correlation test: R2=0.64, P=0.07; Fig. 1A). This tendency was confirmed by two-way ANOVA; a significant interaction between species and treatment suggests that the osmotic response to cold varied among species (F=2.9, P=0.025).

Fig. 1.

Osmolality and ion concentration in relation to cold tolerance of five Drosophila species. (A) Hemolymph osmolality in untreated controls (circles) and after a 4 h exposure to 0°C (squares) for the five Drosophila species (D. birchii, D. equinoxialis, D. melanogaster, D. persimilis and D. montana) as a function of LTe50 (the temperature causing 50% mortality). Values are plotted as means±s.e.m. (error bars that are not visible are obscured by the symbols). (B) Proportion of osmolality accounted for by ions (before cold exposure) plotted as a function of LTe50 (see Table 1). Species are color coded according to their cold tolerance (warm colors indicate chill-susceptible species and cool colors indicate chill-tolerant species).

Fig. 1.

Osmolality and ion concentration in relation to cold tolerance of five Drosophila species. (A) Hemolymph osmolality in untreated controls (circles) and after a 4 h exposure to 0°C (squares) for the five Drosophila species (D. birchii, D. equinoxialis, D. melanogaster, D. persimilis and D. montana) as a function of LTe50 (the temperature causing 50% mortality). Values are plotted as means±s.e.m. (error bars that are not visible are obscured by the symbols). (B) Proportion of osmolality accounted for by ions (before cold exposure) plotted as a function of LTe50 (see Table 1). Species are color coded according to their cold tolerance (warm colors indicate chill-susceptible species and cool colors indicate chill-tolerant species).

The chill-sensitive species had higher hemolymph [Na+] than the chill-tolerant species and, assuming that each 1 mmol l−1 of ions corresponds to 2 mOsm of hemolymph osmolality (i.e. each cation has a corresponding anion), we calculated the proportion of total osmolality accounted for by K+ and Na+ (and their accompanying anions; Table 1). As shown in Fig. 1B, chill-tolerant species have less of their osmolality accounted for by the ions than the chill-susceptible species (Fig. 1B, Pearson's correlation test; R2=0.85 P=0.001).

Metabolomic profile of adult Drosophila hemolymph

Fig. 2 depicts a representative 1H NMR spectrum of a control group D. melanogaster hemolymph sample. The areas that were removed because of contamination from the oil (see Materials and methods) are marked in gray (see Fig. S1 for an example of the buffer sample without hemolymph). We initially analyzed the spectral variation among all samples using principal component analysis (PCA; N=57, 5–6 samples per species and treatment). This analysis separates the samples based on the variation in metabolite signal intensity only and as shown in Fig. 3 the PCA scores revealed a clear separation of the species, particularly along the first principal component (PC1). Interestingly, the untreated flies of the most cold-tolerant species (D. montana and D. persimilis) were characterized by low PC1 scores, while the most chill-sensitive species (D. birchii) was characterized by high scores, and D. equinoxialis and D. melanogaster had intermediate PC1 scores.

Fig. 2.

General 1H NMR spectrum from hemolymph of D. melanogaster. Gray shading indicates the areas that were removed in the data reduction process (see Materials and methods). A and B are from the same sample but emphasize strong (A) and weak (B) intensity signals. The most dominant metabolite signals have been assigned numbers: (1) formate, 8.44 ppm; (2) possibly guanine or xanthine, 7.67 ppm; (3) O-phosphotyrosine, 7.22, 7.17 ppm; (4) tyrosine, 7.18, 6.89 ppm; (5) fumarate, 6.50 ppm; (6) maltose, 5.39, 5.22, 3.92, 3.89, 3.68, 3.40, 3.26 ppm; (7) trehalose, 5.18, 3.84, 3.81, 3.75, 3.44 ppm; (8) residual water, 4.72 ppm; (9) malate, 4.28, 2.66 ppm; (10) O-phosphocholine, 4.15, 3.21 ppm; (11) proline, 4.12, 2.34, 2.06, 1.99 ppm; (12) O-phosphoethanolamine, 3.97, 3.21 ppm; (13) polyol, 3.61, 1.58 ppm; (14) methanol, 3.38 ppm; (15) methionine sulfoxide or sarcosine, 2.74 ppm; (16) β-alanine, 2.54 ppm; (17) glutamine, 2.44, 2.13 ppm; (18) succinate, 2.39 ppm; (19) fatty acid, 1.28, 0.85 ppm; (20) trimethylsilyl propionate, −0.017 ppm.

Fig. 2.

General 1H NMR spectrum from hemolymph of D. melanogaster. Gray shading indicates the areas that were removed in the data reduction process (see Materials and methods). A and B are from the same sample but emphasize strong (A) and weak (B) intensity signals. The most dominant metabolite signals have been assigned numbers: (1) formate, 8.44 ppm; (2) possibly guanine or xanthine, 7.67 ppm; (3) O-phosphotyrosine, 7.22, 7.17 ppm; (4) tyrosine, 7.18, 6.89 ppm; (5) fumarate, 6.50 ppm; (6) maltose, 5.39, 5.22, 3.92, 3.89, 3.68, 3.40, 3.26 ppm; (7) trehalose, 5.18, 3.84, 3.81, 3.75, 3.44 ppm; (8) residual water, 4.72 ppm; (9) malate, 4.28, 2.66 ppm; (10) O-phosphocholine, 4.15, 3.21 ppm; (11) proline, 4.12, 2.34, 2.06, 1.99 ppm; (12) O-phosphoethanolamine, 3.97, 3.21 ppm; (13) polyol, 3.61, 1.58 ppm; (14) methanol, 3.38 ppm; (15) methionine sulfoxide or sarcosine, 2.74 ppm; (16) β-alanine, 2.54 ppm; (17) glutamine, 2.44, 2.13 ppm; (18) succinate, 2.39 ppm; (19) fatty acid, 1.28, 0.85 ppm; (20) trimethylsilyl propionate, −0.017 ppm.

Fig. 3.

Principal component analysis (PCA) scatter plot for the five species before (circles) and after (squares) exposure to 0°C for 4 h. Arrows indicate the metabolomic response to cold exposure. Each point represents the mean±s.e.m. PCA score of a given species and treatment (N=5–6 from each species and treatment combination). The two first components account for 22% and 10% of the variation, respectively.

Fig. 3.

Principal component analysis (PCA) scatter plot for the five species before (circles) and after (squares) exposure to 0°C for 4 h. Arrows indicate the metabolomic response to cold exposure. Each point represents the mean±s.e.m. PCA score of a given species and treatment (N=5–6 from each species and treatment combination). The two first components account for 22% and 10% of the variation, respectively.

This initial PCA also reveals that cold exposure shifts the metabolite profile of all species. However, the metabolite response after cold exposure varies among species and the magnitude and direction of the metabolite response change with the cold tolerance (Fig. 3). Thus, cold exposure results in slightly lower PC1 and unaltered PC2 values for the cold-tolerant species (D. montana, D. persimilis) and intermediate species (D. melanogaster) while the more cold sensitive species (D. birchii and D. equinoxialis) shift toward higher PC1 and lower PC2 scores (Fig. 3).

In order to identify the metabolites that differ between untreated flies from the different species, we used O2PLS-DA, a supervised regression and prediction method that identify the differences in the multivariate dataset responsible for the group variation (importantly, this analysis is also blind to species cold tolerance). In accordance with the initial PCA analysis, this O2PLS-DA analysis effectively separated the different species (Fig. 4A). Again, we found that the species separated according to their cold tolerance such that PC1 scores correlated tightly with LTe50 (Pearson’s correlation, P=0.001, R2=0.93; Fig. 4B). A high PC1 score (and a low cold tolerance) corresponds to an increase in phosphocholine and maltose, and a metabolite signal at 7.67 ppm that is likely to represent guanine or xanthine, and is henceforth called ‘possible purine’. In contrast, a low PC1 score (and high cold tolerance) corresponds to an increase in trehalose, phosphoethanolamine, β-alanine, succinate, proline and methionine sulfoxide or sarcosine (Fig. 4C). There is also a separation of the species along PC2 that is not immediately related to cold tolerance (Fig. 4A,D).

Fig. 4.

Hemolymph metabolite variations between five Drosophila species with different cold tolerance. (A) Orthogonal partial least squares discriminant analysis (OPLS-DA) scatter plot of individual samples of untreated flies of the five different species. (B) Correlation plot for LTe50 and PC1 score. (C) Loading plot for PC1 that identifies the individual metabolites in the spectral signals associated with PC1. The signal intensity provides information on the amplitude of the metabolite changes and whether the metabolite is present in higher or lower concentration (positive and negative signal intensity, respectively). (D) Loading plot for PC2. Numbers in both loading plots correspond to the metabolites given in Fig. 2.

Fig. 4.

Hemolymph metabolite variations between five Drosophila species with different cold tolerance. (A) Orthogonal partial least squares discriminant analysis (OPLS-DA) scatter plot of individual samples of untreated flies of the five different species. (B) Correlation plot for LTe50 and PC1 score. (C) Loading plot for PC1 that identifies the individual metabolites in the spectral signals associated with PC1. The signal intensity provides information on the amplitude of the metabolite changes and whether the metabolite is present in higher or lower concentration (positive and negative signal intensity, respectively). (D) Loading plot for PC2. Numbers in both loading plots correspond to the metabolites given in Fig. 2.

Given the strong species separation in the O2PLS-DA analysis, we could further demonstrate strong positive relationships between cold tolerance and individual metabolites (especially trehalose and methionine sulfoxide or sarcosine) and negative correlations to other substances such as maltose and phosphocholine (see Fig. S2 for further details).

Metabolite response to cold exposure

To further characterize the metabolite response to cold exposure, we performed PCA on the metabolite changes, defined as the metabolite spectra of cold-exposed flies minus the median of all (N=5–6) unexposed (control) samples of that species. The data were not centered before PCA, and therefore the distance to zero in the scores plot reflects the relative magnitude of the cold response.

As shown in Fig. 5A, the metabolite response to cold exposure was largest in the chill-susceptible species (D. birchii and D. equinoxialis). The chill-susceptible species are characterized by negative scores along PC1, implying an increase in maltose and the probable purine, and simultaneous decreases in trehalose, phosphotyrosine, phosphocholine, proline and succinate (Fig. 5A,B). The cold-tolerant D. persimilis and D. montana, in contrast, show positive PC2 scores, indicating increases in the relative abundance of trehalose, glutamine and the possible purine and decreases in phosphotyrosine, phosphoethanolamine, fumarate, proline, β-alanine and succinate (Fig. 5A,C). Drosophila melanogaster, which have an intermediate cold tolerance, shows responses that are intermediate between the two extremes.

Fig. 5.

Changes in hemolymph metabolites after cold exposure in five Drosophila species with different cold tolerance. (A) PCA scatter plot for the response to cold exposure for the five species. Each point represents a cold-exposed sample minus the median(control) of the same species. The origin (0, 0) corresponds to no response relative to the control. (B) Loading plot for PC1. The loading plot identifies the individual metabolites in the spectral signals associated with PC1. The signal intensity provides information on the amplitude of the metabolite changes and whether the metabolite is present in higher or lower concentration (positive and negative signal intensity, respectively). (C) Loading plot for PC2. Numbers in both loading plots correspond to the metabolites given in Fig. 2.

Fig. 5.

Changes in hemolymph metabolites after cold exposure in five Drosophila species with different cold tolerance. (A) PCA scatter plot for the response to cold exposure for the five species. Each point represents a cold-exposed sample minus the median(control) of the same species. The origin (0, 0) corresponds to no response relative to the control. (B) Loading plot for PC1. The loading plot identifies the individual metabolites in the spectral signals associated with PC1. The signal intensity provides information on the amplitude of the metabolite changes and whether the metabolite is present in higher or lower concentration (positive and negative signal intensity, respectively). (C) Loading plot for PC2. Numbers in both loading plots correspond to the metabolites given in Fig. 2.

Prediction of cold tolerance from the metabolite profile

OPLS models were made to test the direct relationship between hemolymph metabolite profiles and cold tolerance (the model included only two components, one predictive and one orthogonal). Values were predicted with leave-one-out cross-validation, where the cold tolerance for each species was predicted using a model based on data from the metabolite profiles of the other four species. The loading plots for the control samples show that high levels of trehalose, proline, malate, succinate, fumarate and methionine sulfoxide or sarcosine and low levels of phosphocholine, formate and fatty acid are significantly and positively correlated to cold tolerance (Fig. 6A). The predictability of the model (Q2), which is the correlation between the actual cold tolerance (LTe50) and that predicted from the metabolite profile, was 0.87 (P=0.0026).

Fig. 6.

Model based on metabolomics data, to identify metabolite profiles that are predictive of Drosophila cold tolerance. See Materials and methods and Results for details. (A) Loading plot identifying the individual metabolites predicting cold tolerance from the metabolite profile of untreated Drosophila. The signal intensity provides information on whether the metabolite is present in higher or lower concentration (positive and negative signal intensity, respectively). Significant alterations were color coded from blue to red, with red representing the highest correlation between the metabolite and the cold-tolerance model. (B) Loading plot for predicting cold sensitivity from the metabolite response to cold stress. (C) Correlation between LTe50 values predicted from untreated metabolite levels using cross-validation and measured values. (D) Correlation between LTe50 values predicted from the cold response and measured values.

Fig. 6.

Model based on metabolomics data, to identify metabolite profiles that are predictive of Drosophila cold tolerance. See Materials and methods and Results for details. (A) Loading plot identifying the individual metabolites predicting cold tolerance from the metabolite profile of untreated Drosophila. The signal intensity provides information on whether the metabolite is present in higher or lower concentration (positive and negative signal intensity, respectively). Significant alterations were color coded from blue to red, with red representing the highest correlation between the metabolite and the cold-tolerance model. (B) Loading plot for predicting cold sensitivity from the metabolite response to cold stress. (C) Correlation between LTe50 values predicted from untreated metabolite levels using cross-validation and measured values. (D) Correlation between LTe50 values predicted from the cold response and measured values.

The same analysis was repeated for the metabolite changes following cold exposure. Here, the cold-tolerant species responded to cold by a higher increase or lower decrease in trehalose and phosphocholine and the possible purine, and a lower increase or higher decrease in maltose, glucose, β-alanine and glutamine (Fig. 6B). For this analysis, we found the Q2 for LTe50 was 0.68 (P=0.0015). The strong link between the changes in metabolome and chill tolerance is further emphasized by an even better prediction (Q2=0.86) when we repeated the analysis using the time to stand following the particular cold treatment used here, as an alternative cold-tolerance measure (data not shown). The metabolite changes following cold were thus also closely related to variation in other cold-tolerance traits.

Hemolymph osmolality before and after cold exposure

Previous studies have found that the cold tolerance of Drosophila spp. is tightly linked to the hemolymph [Na+], so that the more chill-tolerant species depend less on Na+ as an osmolyte in the hemolymph (MacMillan et al., 2015b,,d). This suggests that either hemolymph osmolality is lower in cold-tolerant species or cold-tolerant species have higher concentrations of other osmolytes in their hemolymph. In the present study, we found no differences in osmolality among species in benign conditions (Fig. 1A), suggesting that cold-tolerant species maintain higher hemolymph concentrations of other compatible osmolytes. This finding is further supported by a significant negative correlation between cold tolerance and the proportion of osmolality that can be accounted for by ions (i.e. the correlation between LTe50 and ion contribution is positive; Fig. 1B). We also examined hemolymph osmolality after cold exposure and found that osmolality increased in four of the five species. In this case, we found a (non-significant) trend suggesting that the cold-tolerant species had smaller increases in osmolality following cold stress, which supports the hypothesis that cold-tolerant insects may better maintain water balance in the cold (Koštal et al., 2004; MacMillan and Sinclair, 2011b; MacMillan et al., 2015a,,d).

The present study cannot directly reveal why osmolality increases with cold, but we speculate that three mechanisms could be involved. Firstly, different species may experience different degrees of respiratory, evaporative or excretory water loss during cold exposure. Indeed, tropical (and cold-sensitive) Drosophila species are generally also more susceptible to dehydration than xeric or cold-adapted species (Kellermann et al., 2013). Further, it has previously been shown that dehydration exacerbates the development of chill injury in D. melanogaster, particularly at moderate cold exposure temperatures (Kobey and Montooth, 2013). It should be mentioned, however, that the water vapor deficit is very low during the cold exposure and the flies are therefore unlikely to be severely stressed by dehydration during short-term exposures (Kobey and Montooth, 2013). Secondly, it is possible that dissipation of osmotic gradients occurs within the animal as a consequence of reduced active transport at low temperatures. While it is normally recognized that there cannot be any osmotic gradients between intracellular and extracellular spaces (Hill et al., 2008), it is possible to maintain osmotic gradients in different body compartments through continuous active transport. It has been shown, for example, that the insect hindgut is markedly hyperosmotic relative to the hemolymph and somatic cells (Edney, 1977). Hyperosmolality of the gut is dependent on continuous active ion transport in the rectal wall driving reabsorption of water from the gut to the hemocoel and we therefore speculate that cold-induced impairment of active transport in the hindgut could lead to an increase in hemolymph osmolality. Finally, it is possible that increases in hemolymph osmolality are a consequence of (adaptive or maladaptive) breakdown of protein or glycogen sources or transport to/from the hemocoel in response to cold (Edney, 1977; Hochachka and Somero, 2002). Regardless of the proximate cause of these osmotic perturbations, our data clearly indicate that cold exposure is associated with larger osmotic disturbance in the most sensitive species. For all species, we observed that the total cation concentration (sum of [Na+] and [K+]) remained fairly constant before and after cold exposure, while osmolality increased in most species (Table 1). Assuming that the concentrations of anions are also relatively constant after the volume change, we conclude that the osmolality increase must be associated with increases in other osmolytes and that these increases are more dramatic in the cold-sensitive species.

General metabolite profiles

Based on previously published data (MacMillan et al., 2015d), we knew that cold-tolerant Drosophila experience smaller disturbances in ion homeostasis during cold exposure (Table 1). This homeostatic capacity is partially attributed to smaller Na+ gradients between the gut lumen and hemolymph, which help to preserve hemolymph water balance (see MacMillan and Sinclair, 2011b; MacMillan et al., 2015a,b,,d, for further discussion; Table 1). Because cold-tolerant species have lower [Na+] but similar osmolality, we suggest that these species must replace Na+ with other osmolytes in the hemolymph (Fig. 1A,B). To investigate this hypothesis, we characterized the metabolomic profile of hemolymph samples using 1H NMR spectroscopy and found clear interspecific differences in the hemolymph metabolite profile (note, these differences represent qualitative differences in signal intensity that are not necessarily directly proportional to concentration differences). Importantly, the metabolic profiles of species segregated in very close association with their cold tolerance (Figs 3 and 4). Specifically, the characteristics of the cold-tolerant species before cold exposure were larger relative signals of sugars (trehalose and glucose) and free amino acids (β-alanine, proline and methionine sulfoxide or sarcosine) and lower signals of maltose, phosphocholine and the possible purine (Fig. 4C).

The spectra showed that trehalose is by far the most abundant sugar molecule in the hemolymph and that the greater abundance of trehalose in tolerant species more than outweighs the lower levels of maltose found in these species. Chill-tolerant species are therefore generally characterized by a higher content of sugars in their hemolymph. Trehalose is a known cryoprotectant which can contribute to cold tolerance colligatively by suppressing the super cooling point (Zachariassen, 1985). However, given the modest (normal) osmolality found in the Drosophila hemolymph, it is unlikely that the increased relative signal for trehalose found in either D. persimilis or D. montana exerts any important colligative effect. Nevertheless, trehalose (in low concentrations) has previously been found to facilitate other non-colligative functions in stressed organisms. These functions include protection of native proteins and biological membranes by preferential exclusion from their hydration shells (Arakawa and Timasheff, 1983,1985), assistance in refolding unfolded proteins by molecular chaperones (Lee and Goldberg, 1998; Viner and Clegg, 2001), scavenging for oxygen radicals (Benaroudj et al., 2001) and replacement of missing water molecules in hydration shells of proteins and phospholipid membranes during desiccation (Yancey et al., 1982; Crowe et al., 1987, 1988; Crowe, 2002). We were unfortunately unable to directly quantify trehalose content in the present study and cannot therefore conclude whether the concentrations were high enough to offer increased cold tolerance by any of these mechanisms. However, small increases in trehalose are found on a whole-organism level after both rapid cold hardening and long-term cold acclimation in larval and adult D. melanogaster (Overgaard et al., 2007; Koštal et al., 2011a; Colinet et al., 2012). Further, whole-organism levels of trehalose (and glucose) were found to increase during autumn and winter in D. montana (Vesala et al., 2012). However, interestingly, feeding with dietary sugars does not increase cold tolerance of D. melanogaster even though it increases the general level of sugars (Colinet et al., 2013).

Of the amino acids, we found proline to be higher in cold-tolerant species. Proline has also been found to have non-colligative effects that increase insect cold tolerance. Like trehalose, it is unclear from this study whether the relatively large abundance of proline in cold-tolerant species has any direct cryoprotective role but proline may also help to protect proteins and cell membranes by preferential exclusion from hydration shells (Arakawa and Timasheff, 1983, 1985). Several other studies have found increases in free amino acid levels in relation to different cold-acclimation treatments. In cold-acclimated adult D. melanogaster, several amino acids increased (Colinet et al., 2012) and in D. melanogaster larvae, long-term cold acclimation increased levels of proline in particular (Koštal et al., 2011a). Furthermore, in D. montana, proline varies seasonally, showing a maximum in the winter (Vesala et al., 2012), while fluctuating thermal regimes and diapause lead to increased levels of many amino acids (e.g. lysine, isoleucine and leucine) in the tropical beetle Alphitobius diaperinus and codling moth Cydia pomonella, respectively (Lalouette et al., 2007; Rozsypal et al., 2013). Finally, proline seems to be a very potent (and dominant) cryoprotectant in the freeze-tolerant drosophilid Chymomyza costata (Koštal et al., 2011b). Most convincingly, increased dietary proline allows C. costata to survive freezing in liquid nitrogen (Koštal et al., 2011b), and (when combined with cold acclimation) a proline-augmented diet even allows D. melanogaster larvae to survive freezing of almost 50% of their body water (Koštal et al., 2012).

The small sample volumes and sensitive techniques used in this study caused a fairly high degree of contamination in the samples. As a result of this contamination, we excluded several areas of the spectra from the analysis (Fig. 2). These removed areas could potentially contain important information on other important metabolites such as glycerol and other polyols whose chemical shifts lie in these areas. Further, NMR spectroscopy generally has lower sensitivity than, for example, mass spectroscopy-based techniques and it is therefore quite possible that our analysis has missed compounds that are in low concentration or are obscured by the contamination found in some regions of the spectra. Glycerol is a very important cryoprotectant in many freeze-avoiding insect species (Zachariassen, 1985) and has also been found to be associated with increased cold tolerance in cold-sensitive species like the flesh fly and the tropical beetle A. diaperinus (Chen et al., 1987; Yoder et al., 2006; Michaud and Denlinger, 2007; Lalouette et al., 2007; Teets et al., 2012). Glycerol has not previously been associated with cold tolerance in Drosophila (Kelty and Lee, 1999; Overgaard et al., 2007; Koštal et al., 2011a; Colinet et al., 2012), but D. montana does accumulate large amounts of another polyol, myo-inositol, as a cryoprotectant (Vesala et al., 2012). However, analysis of the entire spectra without removal of the contaminated areas did not show any strong signals for these compounds.

In summary, the metabolite profiles of cold-tolerant species of Drosophila show relatively high levels of sugars and amino acids compared with more sensitive species. These compounds (including trehalose and proline) are likely to replace Na+ to preserve osmolality, but they could also potentially contribute to the higher cold tolerance of these species. Although absolute concentrations were not measured in the present study, the magnitude of the differences in the metabolite profile are unlikely to offer cryoprotection by means of colligative effects (Zachariassen, 1985) and instead we suggest that these compounds offer cryoprotection by means of non-colligative effects. In support of this, we found that the metabolite profiles were very exact in their prediction of the cold tolerance of the species (Fig. 6C). This indicates that the metabolomic profile is highly important (and predictive) for species-specific cold tolerance. Cryoprotectants are classically associated with adaptations to avoid or tolerate ice formation (Zachariassen, 1985; Ramløv, 2000) but in these chill-sensitive species we demonstrate that classical cryoprotectants also play an important role for chill-susceptible insects. Thus, we propose that these compatible sugars and amino acids are important for osmotic balance and non-colligative cellular protection, and suggest that this mechanistic role of cryoprotectants (in low concentrations) might represent a new/alternative view on the pathway that has led to the evolution of freeze tolerance and freeze avoidance.

Metabolite response to cold exposure

The metabolome of the very chill-sensitive species D. birchii responded vigorously to cold exposure (Fig. 5A) with massive increases in maltose levels and smaller decreases in phosphocholine and trehalose. Drosophila equinoxialis and D. melanogaster showed smaller but qualitatively similar responses with the addition of an increase in the possible purine. Increases in maltose levels have previously been found in response to cold shock treatments in D. melanogaster (Overgaard et al., 2007; Pedersen et al., 2008) and maltose has also been found to increase in D. melanogaster selected for high chill susceptibility (Williams et al., 2014). Maltose has been linked to the protection and function of proteins and membranes during temperature stress (Kaplan and Guy, 2004; Pereira and Hünenberger, 2006), and maltose accumulation has been linked to a stress-induced increase in amylase activity in plants (Kaplan and Guy, 2004). The accumulation of maltose in the chill­-susceptible species could therefore function to reduce the likelihood of chill injury following cold exposure, but may also be a more general and non-specific response to stressful conditions. Increased levels of maltose have previously been linked to a wide range of different stressors like inbreeding, thermal stress, starvation and artificial expression of the amyloid β-peptide (Pedersen et al., 2008; Malmendal et al., 2013; Ott et al., 2016).

Following cold exposure, D. persimilis and D. montana showed very little change along the trajectory of the more chill-sensitive species, while both showed increases in the relative abundance of trehalose and the possible purine (Fig. 5B,C). The relatively small response to cold exposure suggests that these species maintain their metabolite homeostasis far better than the chill-susceptible species. Similarly, the metabolomes of cold-selected, cold-acclimated and rapidly cold-hardened D. melanogaster are less responsive to cold exposure than warm-acclimated flies or flies directly exposed to a cold shock (Overgaard et al., 2007; Colinet et al., 2012; Williams et al., 2014). The increases in trehalose and the possible purine found in response to cold in D. persimilis and D. montana seem to be adaptive for coping with the cold stress. Indeed, in a parallel study, we found that these two species were able to reverse the cold-induced drift of ions down their gradients and began to recover neuromuscular function and stand while still being held at 0°C (MacMillan et al., 2015d), which suggests that these species are capable of acclimating at this temperature.

Conclusions

This study has provided the first examination of hemolymph-specific metabolite profiles and osmolality measurements both before and after cold exposure in five Drosophila species with different cold tolerances. We found that there were no osmolality differences among species before cold exposure but that a larger proportion of the hemolymph osmolality was accounted for by organic osmolytes in chill-tolerant species. 1H NMR spectroscopy on the hemolymph revealed that the chill-tolerant species had higher levels of trehalose and proline and that their metabolite profiles correlated remarkably well with cold tolerance. The response to cold exposure was increased osmolality in all species, suggesting that all species had an accumulation of osmolytes during cold exposure. However, during cold exposure, the chill-sensitive species lost hemolymph water and accumulated the stress-related metabolite maltose, while the chill-tolerant species accumulated primarily trehalose (which may be a more adaptive response). High levels of trehalose and proline in the hemocoel may offer cryoprotection through non-colligative and osmotic effects in chill-sensitive insects. Further, we suggest that the small increases found in the more tolerant species might represent a step in an evolutionary pathway to true cold tolerance where these compounds are accumulated to such high levels that they offer cryoprotection from both colligative and non-colligative effects. In this respect, it would also be interesting to examine the somewhat larger increases in cryoprotectants found in diapausing drosophilids such as D. montana as metabolite changes reported under these conditions (Vesala et al., 2012) are considerably larger, albeit still too low to confer significant colligative effects.

We are grateful to Martin Holmstrup for use of equipment and to Kirsten Kromand for excellent technical assistance.

Author contributions

T.O., H.M.A., A.M. and J.O. contributed to the design of the experiments. T.O, N.N. and D.S. performed the experiments. T.O. H.M.A. and A.M analyzed the data. T.O, A.M. and J.O. wrote the manuscript, and all of the authors revised the manuscript.

Funding

This research was funded by a research grant from Det Frie Forskningsråd [EK15] | Natur og Univers [501100004836] to J.O.

Andersen
,
J. L.
,
Manenti
,
T.
,
Sørensen
,
J. G.
,
MacMillan
,
H. A.
,
Loeschcke
,
V.
and
Overgaard
,
J.
(
2015
).
How to assess Drosophila cold tolerance: chill coma temperature and lower lethal temperature are the best predictors of cold distribution limits
.
Funct. Ecol.
29
,
55
-
65
.
Arakawa
,
T.
and
Timasheff
,
S. N.
(
1983
).
Preferential interactions of proteins with solvent components in aqueous amino acid solutions
.
Arch. Biochem. Biophys.
224
,
169
-
177
.
Arakawa
,
T.
and
Timasheff
,
S. N.
(
1985
).
The stabilization of proteins by osmolytes
.
Biophys. J.
47
,
411
-
414
.
Bale
,
J. S.
(
1996
).
Insect cold hardiness: a matter of life and death
.
Eur. J. Entomol.
93
,
369
-
382
.
Benaroudj
,
N.
,
Lee
,
D. H.
and
Goldberg
,
A. L.
(
2001
).
Trehalose accumulation during cellular stress protects cells and cellular proteins from damage by oxygen radicals
.
J. Biol. Chem.
276
,
24261
-
24267
.
Bligh
,
E. G.
and
Dyer
,
W. J.
(
1959
).
A rapid method of total lipid extraction and purification
.
Can. J. Biochem. Physiol.
37
,
911
-
917
.
Boutilier
,
R. G.
(
2001
).
Mechanisms of cell survival in hypoxia and hypothermia
.
J. Exp. Biol.
204
,
3171
-
3181
.
Chen
,
C.-P.
,
Denlinger
,
D. L.
and
Lee
,
R. E.
(
1987
).
Cold-shock injury and rapid cold hardening in the flesh fly Sarcophaga crassipalpis
.
Physiol. Zool.
60
,
297
-
304
.
Colinet
,
H.
,
Larvor
,
V.
,
Laparie
,
M.
and
Renault
,
D.
(
2012
).
Exploring the plastic response to cold acclimation through metabolomics
.
Funct. Ecol.
26
,
711
-
722
.
Colinet
,
H.
,
Larvor
,
V.
,
Bical
,
R.
and
Renault
,
D.
(
2013
).
Dietary sugars affect cold tolerance of Drosophila melanogaster
.
Metabolomics
9
,
608
-
622
.
Craig
,
A.
,
Cloarec
,
O.
,
Holmes
,
E.
,
Nicholson
,
J. K.
and
Lindon
,
J. C.
(
2006
).
Scaling and normalization effects in NMR spectroscopic metabonomic data sets
.
Anal. Chem.
78
,
2262
-
2267
.
Crowe
,
L. M.
(
2002
).
Lessons from nature: the role of sugars in anhydrobiosis
.
Comp. Biochem. Physiol. A Mol. Integr. Physiol.
131
,
505
-
513
.
Crowe
,
J. H.
,
Crowe
,
L. M.
,
Carpenter
,
J. F.
and
Wistrom
,
C. A.
(
1987
).
Stabilization of dry phospholipid bilayers and proteins by sugars
.
Biochem. J.
242
,
1
-
10
.
Crowe
,
J. H.
,
Crowe
,
L. M.
,
Carpenter
,
J. F.
,
Rudolph
,
A. S.
,
Wistrom
,
C. A.
,
Spargo
,
B. J.
and
Anchordoguy
,
T. J.
(
1988
).
Interactions of sugars with membranes
.
Biochem. Biophys. Acta
947
,
367
-
384
.
Cui
,
Q.
,
Lewis
,
I. A.
,
Hegeman
,
A. D.
,
Anderson
,
M. E.
,
Li
,
J.
,
Schulte
,
C. F.
,
Westler
,
W. M.
,
Eghbalnia
,
H. R.
,
Sussman
,
M. R.
and
Markley
,
J. L.
(
2008
).
Metabolite identification via the Madison Metabolomics Consortium Database
.
Nat. Biotechnol.
26
,
162
-
164
.
Edney
,
E. B.
(
1977
).
Water Balance in Land Arthropods
.
Berlin
:
Springer
.
Findsen
,
A.
,
Andersen
,
J. L.
,
Calderon
,
S.
and
Overgaard
,
J.
(
2013
).
Rapid cold hardening improves recovery of ion homeostasis and chill coma recovery time in the migratory locust, Locusta migratoria
.
J. Exp. Biol.
216
,
1630
-
1637
.
Fitzgerald
,
E. M.
,
Djamgoz
,
M. B. A.
and
Dunbar
,
S. J.
(
1996
).
Maintenance of the K+ activity gradient in insect muscle compared in Diptera and Lepidoptera: contributions of metabolic and exchanger mechanisms
.
J. Exp. Biol.
199
,
1857
-
1872
.
Hill
,
R. W.
,
Wyse
,
G. A.
and
Anderson
,
M.
(
2008
).
Animal Physiology
, 2nd edn.
Sunderland, MA
:
Sinauer Associates
.
Hochachka
,
P. W.
and
Somero
,
G. N.
(
2002
).
Biochemical Adaptation: Mechanism and Process in Physiological Evolution
.
Oxford
:
Oxford University Press
.
Hoyle
,
G.
(
1954
).
Changes in the blood potassium concentration of the African migratory locust (Locusta migratoria migratorioides) during food deprivation, and the effect on neuromuscular activity
.
J. Exp. Biol.
31
,
260
-
270
.
Kaplan
,
F.
and
Guy
,
C. L.
(
2004
).
b-Amylase induction and the protective role of maltose during temperature shock
.
Plant Physiol.
135
,
1674
-
1684
.
Kellermann
,
V.
,
Overgaard
,
J.
,
Loeschcke
,
V.
,
Kristensen
,
T. N.
and
Hoffmann
,
A. A.
(
2013
).
Trait associations across evolutionary time within a drosophila phylogeny: correlated selection or genetic constraint?
PLoS ONE
8
,
e72072
.
Kelty
,
J. D.
and
Lee
,
R. E.
, Jr
. (
1999
).
Induction of rapid cold hardening by cooling at ecologically relevant rates in Drosophila melanogaster
.
J. Insect Physiol.
45
,
719
-
726
.
Kobey
,
R. L.
and
Montooth
,
K. L.
(
2013
).
Mortality from desiccation contributes to a genotype–temperature interaction for cold survival in Drosophila melanogaster
.
J. Exp. Biol.
216
,
1174
-
1182
.
Koštal
,
V.
,
Vambera
,
J.
and
Bastl
,
J.
(
2004
).
On the nature of pre-freeze mortality in insects: water balance, ion homeostasis and energy charge in the adults of Pyrrhocoris apterus
.
J. Exp. Biol.
207
,
1509
-
1521
.
Koštal
,
V.
,
Korbelová
,
J.
,
Rozsypal
,
J.
,
Zahradnicková
,
H.
,
Cimlová
,
J.
,
Tomcala
,
A.
and
Simek
,
P.
(
2011a
).
Long-term cold acclimation extends survival time at 0°C and modifies the metabolomic profiles of the larvae of the fruit fly Drosophila melanogaster
.
PLoS ONE
6
,
e25025
.
Koštal
,
V.
,
Zahradníčková
,
H.
and
Šimek
,
P.
(
2011b
).
Hyperprolinemic larvae of the drosophilid fly, Chymomyza costata, survive cryopreservation in liquid nitrogen
.
Proc. Natl. Acad. Sci. USA
108
,
13041
-
13046
.
Koštal
,
V.
,
Šimek
,
P.
,
Zahradníčková
,
H.
,
Cimlová
,
J.
and
Štětina
,
T.
(
2012
).
Conversion of the chill susceptible fruit fly larva (Drosophila melanogaster) to a freeze tolerant organism
.
Proc. Natl. Acad. Sci. USA
109
,
3270
-
3274
.
Lalouette
,
L.
,
Koštal
,
V.
,
Colinet
,
H.
,
Gagneul
,
D.
and
Renault
,
D.
(
2007
).
Cold exposure and associated metabolic changes in adult tropical beetles exposed to fluctuating thermal regimes
.
FEBS J.
274
,
1759
-
1767
.
Lee
,
D. H.
and
Goldberg
,
A. L.
(
1998
).
Proteasome inhibitors cause induction of heat shock proteins and trehalose, which together confer thermotolerance in Saccharomyces cerevisiae
.
Mol. Cell. Biol.
18
,
30
-
38
.
MacMillan
,
H. A.
and
Hughson
,
B. N.
(
2014
).
A high-throughput method of hemolymph extraction from adult Drosophila without anesthesia
.
J. Insect Physiol.
63
,
27
-
31
.
MacMillan
,
H. A.
and
Sinclair
,
B. J.
(
2011a
).
Mechanisms underlying insect chill-coma
.
J. Insect Physiol.
57
,
12
-
20
.
MacMillan
,
H. A.
and
Sinclair
,
B. J.
(
2011b
).
The role of the gut in insect chilling injury: cold-induced disruption of osmoregulation in the fall field cricket, Gryllus pennsylvanicus
.
J. Exp. Biol.
214
,
726
-
734
.
MacMillan
,
H. A.
,
Findsen
,
A.
,
Pedersen
,
T. H.
and
Overgaard
,
J.
(
2014
).
Cold-induced depolarization of insect muscle: differing roles of extracellular K+ during acute and chronic chilling
.
J. Exp. Biol.
217
,
2930
-
2938
.
MacMillan
,
H. A.
,
Andersen
,
J. L.
,
Loeschcke
,
V.
and
Overgaard
,
J.
(
2015a
).
Sodium distribution predicts the chill tolerance of Drosophila melanogaster raised in different thermal conditions
.
Am. J. Physiol. Regul. Integr. Comp. Physiol.
308
,
R823
-
R831
.
MacMillan
,
H. A.
,
Ferguson
,
L. V.
,
Nicolai
,
A.
,
Donini
,
A.
,
Staples
,
J. F.
and
Sinclair
,
B. J.
(
2015b
).
Parallel ionoregulatory adjustments underlie phenotypic plasticity and evolution of Drosophila cold tolerance
.
J. Exp. Biol.
218
,
423
-
432
.
MacMillan
,
H. A.
,
Baatrup
,
E.
and
Overgaard
,
J.
(
2015c
).
Concurrent effects of cold and hyperkalaemia cause insect chilling injury
.
Proc. R. Soc. B Biol. Sci.
282
,
20151483
.
MacMillan
,
H. A.
,
Andersen
,
J. L.
,
Davies
,
S. A.
and
Overgaard
,
J.
(
2015d
).
The capacity to maintain ion and water homeostasis underlies interspecific variation in Drosophila cold tolerance
.
Sci. Rep.
5
,
18607
.
Malmendal
,
A.
,
Overgaard
,
J.
,
Bundy
,
J. G.
,
Sørensen
,
J. G.
,
Nielsen
,
N. C.
,
Loeschcke
,
V.
and
Holmstrup
,
M.
(
2006
).
Metabolomic profiling of heat stress: hardening and recovery of homeostasis in Drosophila
.
Am. J. Physiol. Regul. Integr. Comp. Physiol.
291
,
R205
-
R212
.
Malmendal
,
A.
,
Sørensen
,
J. G.
,
Overgaard
,
J.
,
Holmstrup
,
M.
,
Nielsen
,
N. C.
and
Loeschcke
,
V.
(
2013
).
Metabolomic analysis of the selection response of Drosophila melanogaster to environmental stress: are there links to gene expression and phenotypic traits?
Naturwissenschaften
100
,
417
-
427
.
Michaud
,
M. R.
and
Denlinger
,
D. L.
(
2007
).
Shifts in the carbohydrate, polyol, and amino acid pools during rapid cold-hardening and diapauseassociated cold-hardening in flesh flies (Sarcophaga crassipalpis): a metabolomic comparison
.
J. Comp. Physiol. B.
177
,
753
-
763
.
Ott
,
S.
,
Vishnivetskaya
,
A.
,
Malmendal
,
A.
and
Crowther
,
D. C.
(
2016
).
Metabolic changes may precede proteostatic dysfunction in a Drosophila model of amyloid β peptide toxicity
.
Neurobiol. Aging.
41
,
39
-
52
.
Overgaard
,
J.
,
Malmendal
,
A.
,
Sørensen
,
J. G.
,
Bundy
,
J. G.
,
Loeschcke
,
V.
,
Nielsen
,
N. C.
and
Holmstrup
,
M.
(
2007
).
Metabolomic profiling of rapid cold hardening and cold shock in Drosophila melanogaster
.
J. Insect Physiol.
53
,
1218
-
1232
.
Pedersen
,
K. S.
,
Kristensen
,
T. N.
,
Loeschcke
,
V.
,
Petersen
,
B. O.
,
Duus
,
J.Ø.
,
Nielsen
,
N. C.
and
Malmendal
,
A.
(
2008
).
Metabolomic signatures of inbreeding at benign and stressful temperatures in Drosophila melanogaster
.
Genetics
180
,
1233
-
1243
.
Pereira
,
C. S.
and
Hünenberger
,
P. H.
(
2006
).
Interaction of the sugars trehalose, maltose and glucose with a phospholipid bilayer: a comparative molecular dynamics study
.
J. Phys. Chem. B
110
,
15572
-
15581
.
Purać
,
J.
,
Kojić
,
D.
,
Petri
,
E.
,
Popović
,
Z. D.
,
Grubor-Lajšić
,
G.
and
Blagojević
,
D. P.
(
2016
).
Cold adaptation responses in insects and other arthropods: an “Omics” approach
. In
Short Views on Insect Genomics and Proteomics, Insect Proteomics
, Vol.
2
(ed.
C.
Raman
,
M. R.
Goldsmith
and
T. A.
Agunbiade
).
Heidelberg
:
Springer Verlag
.
ISBN 978-3-319-24244-6
.
Ragan
,
T. J.
,
Bailey
,
A. P.
,
Gould
,
A. P.
and
Driscoll
,
P. C.
(
2013
).
Volume Determination with two standards allows absolute quantification and improved chemometric analysis of metabolites by NMR from submicroliter samples
.
Anal. Chem.
85
,
12046
-
12054
.
Ramløv
,
H.
(
2000
).
Aspects of natural cold tolerance in ectothermic animals
.
Hum. Reprod.
15
,
26
-
46
.
Rozsypal
,
J.
,
Koštál
,
V.
,
Zahradníčková
,
H.
and
Šimek
,
P.
(
2013
).
Overwintering strategy and mechanisms of cold tolerance in the codling moth (Cydia pomonella)
.
PLoS ONE
8
,
e61745
.
Sinclair
,
B. J.
(
1999
).
Insects cold tolerance: how many kinds of frozen?
Eur. J. Entomol.
96
,
157
-
164
.
Teets
,
N. M.
,
Peyton
,
J. T.
,
Ragland
,
G. J.
,
Colinet
,
H.
,
Renault
,
D.
,
Hahn
,
D. A.
and
Denlinger
,
D. L.
(
2012
).
Combined transcriptomic and metabolomic approach uncovers molecular mechanisms of cold tolerance in a temperate flesh fly
.
Physiol. Genomics
44
,
764
-
777
.
Vesala
,
L.
,
Salminen
,
T. S.
,
Koštal
,
V.
,
Zahradnicková
,
H.
and
Hoikkala
,
A.
(
2012
).
Myo-inositol as a main metabolite in overwintering flies: seasonal metabolomic profiles and cold stress tolerance in a northern drosophilid fly
.
J. Exp. Biol.
215
,
2891
-
2897
.
Viner
,
R. I.
and
Clegg
,
J. S.
(
2001
).
Influence of trehalose on the molecular chaperone activity of p26, a small heat shock/a-crystallin protein
.
Cell Stress Chaperones
6
,
126
-
135
.
Wareham
,
A. C.
,
Duncan
,
C. J.
and
Bowler
,
K.
(
1974
).
The resting potential of cockroach muscle membrane
.
Comp. Biochem. Psysiol. A Physiol.
48
,
765
-
797
.
Williams
,
C. M.
,
Watanabe
,
M.
,
Guarracino
,
M. R.
,
Ferraro
,
M. B.
,
Edison
,
A. S.
,
Morgan
,
T. J.
,
Boroujerdi
,
A. F. B.
and
Hahn
,
D. A.
(
2014
).
Cold adaptation shapes the robustness of metabolic networks in Drosophila melanogaster
.
Evolution
68
,
3505
-
3523
.
Wishart
,
D. S.
,
Tzur
,
D.
,
Knox
,
C.
,
Eisner
,
R.
,
Gou
,
A. C.
,
Young
,
N.
,
Cheng
,
D.
,
Jewell
,
K.
,
Arndt
,
D.
,
Sawhney
,
S.
, et al. 
(
2007
).
HMDB: the human metabolome database
.
Nucleic Acids Res.
35
,
D521
-
D526
.
Yancey
,
P. H.
,
Clark
,
M. E.
,
Hand
,
S. C.
,
Bowlus
,
R. D.
and
Somero
,
G. N.
(
1982
).
Living with water stress: evolution of osmolyte systems
.
Science
217
,
1214
-
1222
.
Yoder
,
J. A.
,
Benoit
,
J. B.
,
Denlinger
,
D. L.
and
Rivers
,
D. B.
(
2006
).
Stress-induced accumulation of glycerol in the flesh fly, Sarcophaga bullata: evidence indicating anti-desiccant and cryoprotectant functions of this polyol and a role for the brain in coordinating the response
.
J. Insect Physiol.
52
,
202
-
214
.
Zachariassen
,
K. E.
(
1985
).
Physiology of cold tolerance in insects
.
Am. Phys. Soc.
65
,
799
-
832
.

Competing interests

The authors declare no competing or financial interests.

Supplementary information