Parallel ionoregulatory adjustments underlie phenotypic plasticity and evolution of Drosophila cold tolerance

Low temperature tolerance is especially important for determining geographic range limits of insects (Battisti et al., 2005; Chen et al., 2011; Kellermann et al., 2012). Thus understanding the mechanisms that set thermal limits allows prediction of changes to their distribution and abundance in a changing environment (Pörtner and Farrell, 2008; Hofmann and Todgham, 2010; Andersen et al., 2015; Overgaard et al., 2014). Insect lower thermal limits manifest as entry into chill coma at the critical thermal minimum (CT_min), and the subsequent accumulation of chilling injuries that lead to sublethal effects on performance and fitness or death (MacMillan and Sinclair, 2011a; MacMillan et al., 2014). Thermal tolerance can be modulated through phenotypic plasticity or through genetic adaptation, but the former has been suggested to be constrained by the latter; this means that species that have evolved to tolerate more extreme temperatures may be less able to acclimate to changes in temperature (Nyamukondiwa et al., 2011; Stillman, 2003) (but see Calosi et al., 2008; Overgaard et al., 2011). Identifying the mechanisms underlying thermal tolerance would elucidate the reasons for this possible trade-off, and allow for a priori estimation of low temperature tolerance, and high-throughput prediction of insect susceptibility to climate change (Williams et al., 2014).

The physiological causes of chill coma and chilling injury have not been thoroughly explored in holometabolous insects, which comprise the majority of terrestrial animal species (Wiegmann et al., 2009), including many threatened by climate change (Chen et al., 2011). The cosmopolitan genus Drosophila encompasses 60 million years of evolutionary history (Clark et al., 2007) and has both a broad geographic range and well-described variation in cold tolerance (Kellermann et al., 2012). Variation in cold tolerance among species persists after rearing Drosophila species under common conditions, even for many years, which implies an underlying genetic component to cold tolerance that is resilient to laboratory selection and suffers little inbreeding depression (Gilchrist et al., 1997; Ayrinhac et al., 2004; Kellermann et al., 2012; Bechsgaard et al., 2013). Similarly, most laboratory populations of Drosophila, including D. melanogaster, retain phenotypic plasticity of thermal tolerance traits, including those related to cold (Nyamukondiwa et al., 2011; Overgaard et al., 2011). The CT_min is simple to measure in large numbers of flies, and is closely correlated to both the poleward distribution limits of Drosophila species, and the sensitivity of flies to chilling injury (Andersen et al., 2015). Drosophila are thus an ideal genus in which to explore the mechanisms of thermal tolerance variation in the Holometabola, both within- and among-species.

In hemimetabolous insects, chilling injury is associated with a disruption of ion and water balance (Koštál et al., 2004; Koštál et al., 2006; MacMillan and Sinclair, 2011b; Findsen et al., 2013). The concentration of Na⁺ is high in the extracellular fluid relative to the diet of phytophagous insects, and at low temperatures haemolymph Na⁺ leaks across the gut wall. Because water distribution is heavily dependent on Na⁺, water follows Na⁺ down its concentration gradient, thereby reducing haemolymph volume and consequently increasing extracellular [K⁺] (MacMillan and Sinclair, 2011b). Although much Na⁺ leaves the haemolymph, the concurrent loss of haemolymph water can result in little change in extracellular [Na⁺]. In crickets and locusts, this imbalance is thought to lead to chilling injury, and re-establishment of ion homeostasis is required for recovery of neuromuscular function (MacMillan et al., 2012; Findsen et al., 2014; MacMillan et al., 2014). This net leak of Na⁺ in the cold is thought to occur because the activity of temperature-sensitive enzymatic ion pumps is insufficient to balance passive leak of ions, the rate of which is dependent on the magnitude of the Na⁺ gradient (Zachariassen et al., 2004; MacMillan and Sinclair, 2011a).

Failure of Na⁺/K⁺-ATPase has been specifically hypothesized to be associated with chill coma and chilling injury, because of its primary role in maintaining Na⁺ and K⁺ homeostasis and electrogenic role in determining the membrane potential (e.g. Hosler et al., 2000; MacMillan and Sinclair, 2011a). At the cellular level, Na⁺/K⁺-ATPase in the cell membrane consumes ATP to pump 3Na⁺ out of the cell cytoplasm and 2K⁺ in during each reaction cycle, maintaining high intracellular [K⁺] and low intracellular [Na⁺] (Emery et al., 1998), and also resulting in high extracellular [Na⁺]. In insects, Na⁺ balance is primarily maintained by the renal system, composed of the Malpighian tubules and hindgut. Expression of mRNA for the α-subunit of Na⁺/K⁺-ATPase is enriched 7.9- and 5.3-fold (relative to whole body) in the Malpighian tubules and hindgut, respectively, of D. melanogaster (Chintapalli et al., 2007). Although fluid secretion at the Malpighian tubules is primarily energized by the proton-motive V-ATPase coupled to a H⁺/K⁺ exchanger, Na⁺/K⁺-ATPase in the basolateral membrane of Malpighian tubule principal cells is important in modulating rates of Na⁺ and water secretion (Linton and O'Donnell, 1999; Beyenbach et al., 2010). Indeed, inhibition of Na⁺/K⁺-ATPase in D. melanogaster Malpighian tubules elevates [Na⁺] of the secreted fluid by as much as 73%, and increases the fluid secretion rate by ~15% (Linton and O'Donnell, 1999). High Na⁺/K⁺-ATPase activity in the basal membrane of the hindgut is responsible for transporting Na⁺ back into the haemolymph of the mosquito Aedes aegypti (Diptera: Culicidae) (Patrick et al., 2006), and net Na⁺ is transported into the haemolymph in the larval hindgut of D. melanogaster (Naikkhwah and O'Donnell, 2012). Thus Na⁺/K⁺-ATPase in the Drosophila renal system appears to maintain high [Na⁺] in the haemolymph and limit Na⁺ loss to the gut lumen and faeces.

Based on the role of Na⁺/K⁺-ATPase in ion balance in Drosophila and the patterns of ion and water balance disruption in other cold-exposed insects, cold tolerance could potentially be improved by reducing Na⁺/K⁺-ATPase activity at high or ‘normal’ temperatures, which would decrease retention of Na⁺ under resting conditions, and consequently reduce resting haemolymph [Na⁺]. This decrease in haemolymph [Na⁺] would reduce the transmembrane and transepithelial Na⁺ gradients that drive Na⁺ leak and water balance disruption at low temperatures, and thereby improve cold tolerance. This hypothesis is supported by the repeated observation that cold acclimation of insects results in less-polarized muscle Na⁺ equilibrium potentials (Koštál et al., 2004; Koštál et al., 2006; Coello Alvarado, 2012).

Although low Na⁺/K⁺-ATPase activity at relatively high temperatures might reduce resting Na⁺ gradients and improve cold tolerance, a failure of this pump in the cold will probably cause net ion leak across cell membranes and epithelia, which would similarly lead to increased haemolymph [K⁺] and chilling injury. Reductions in enzyme thermal sensitivity are often associated with adaptation to low temperatures (Somero, 2004; Dong and Somero, 2009; Galarza-Muñoz et al., 2011; Garrett and Rosenthal, 2012). For example, polar and temperate octopus species have Na⁺/K⁺-ATPase α-subunits that differ in thermal sensitivity; at temperatures greater than 25°C the maximal rate of Na⁺/K⁺-ATPase does not differ between polar and temperate octopuses, but that of polar octopuses is 4-fold higher at 10°C (Galarza-Muñoz et al., 2011). Thus, cold tolerance may also be improved in Drosophila through similar reductions in the thermal sensitivity of Na⁺/K⁺-ATPase.

In this study, we therefore hypothesized that reduced Na⁺/K⁺-ATPase activity and decreased haemolymph [Na⁺] at rest underlie improvements in insect cold tolerance, both within species (phenotypic plasticity) and among species (genetic adaptation). To test this hypothesis, we used thermal acclimation to induce variation in cold tolerance via phenotypic plasticity in D. melanogaster, and 24 species of Drosophila whose cold tolerance naturally varies when reared under common conditions, with the CT_min used as an index of cold tolerance. We predicted that cold-acclimated D. melanogaster and cold-tolerant species would have low haemolymph [Na⁺], thereby reducing transmembrane and transepithelial Na⁺ gradients at rest, which would consequently protect against Na⁺ leak in the cold. We also predicted that cold-tolerant flies would maintain low haemolymph [K⁺], which would reduce the impact of any cold-induced loss of Na⁺ and water balance on cell survival. Lastly, we hypothesized that cold-tolerant Drosophila better maintain their ion gradients when exposed to low temperatures and predicted that cold-adapted and cold-acclimated flies would display reduced thermal sensitivity of Na⁺/K⁺-ATPase.

Here we show that cold exposure below the CT_min causes elevated haemolymph [K⁺] in D. melanogaster, as in other insects, and that within- and among-species improvements in cold tolerance are accompanied by low haemolymph cation concentrations. Despite low concentrations of cations, osmotic homeostasis is maintained, with the changes in extracellular ion balance at low temperatures probably compensated by accumulating other compatible osmolytes. This change in cation homeostasis appears to be associated with reduced Na⁺/K⁺-ATPase activity at rest (particularly with cold acclimation), but we observed no differences in the thermal sensitivity of the enzyme.

We measured the CT_min of adult males and females of 24 Drosophila species (supplementary material Table S1), reared under common laboratory conditions and fed a similar diet; and of adult male D. melanogaster acclimated for 5 days to ‘warm’ (i.e. a typical laboratory culture temperature; 21.5°C) and ‘cold’ (6°C) temperatures. We used phylogenetically independent contrasts (PICs; which account for phylogeny in statistical analyses), to test whether among-species relationships between physiological traits and cold tolerance can be explained by the phylogenetic relationships of our sample species (Garland et al., 1992).

The CT_min varied by more than 11°C among species (Fig. 1; supplementary material Fig. S1; Table S2) and had significant phylogenetic signal (K=0.77, P=0.002), meaning closely related species share similar critical thermal minima (Fig. 1). We showed previously that acclimation to 6°C reduced the CT_min of male D. melanogaster in our laboratory population from 3.4±0.2 to 0.9±0.1°C (Ransberry et al., 2011).

We measured ion concentrations in haemolymph extracted from warm- and cold-acclimated D. melanogaster, both at their acclimation temperatures and following a 6 h exposure to 0°C. Acclimation temperature and cold exposure interacted to significantly affect haemolymph [K⁺] (F_1,71=4.6, P=0.036, Fig. 2A). Haemolymph [K⁺] did not differ significantly between warm- and cold-acclimated D. melanogaster at their respective acclimation temperatures (Tukey's HSD, P=0.291; Fig. 2A), and a 6 h exposure to 0°C significantly increased haemolymph [K⁺], but only in warm-acclimated flies (Fig. 2A; Tukey's HSD; P=0.004). By contrast, cold-acclimated D. melanogaster maintained low haemolymph [K⁺] during exposure to 0°C (Tukey's HSD; P=0.441; Fig. 2). Haemolymph [Na⁺] of cold-acclimated D. melanogaster was ~25% lower than that of warm-acclimated flies at their respective acclimation temperature (F_1,56=14.1, P<0.001), and exposure to 0°C for 6 h had no effect on haemolymph [Na⁺] in either acclimation group (F_1,56=0.4, P=0.544; Fig. 2B).

Haemolymph [K⁺] at rest ranged from 16.8±1.4 mmol l⁻¹ (D. auraria) to 37.6±2.8 mmol l⁻¹ (D. willistoni; Fig. 3A) among the Drosophila species and had significant phylogenetic signal (K=0.70, P=0.014), as some cold-tolerant species groups (such as the obscura and virilis groups) had a consistently low haemolymph [K⁺] (supplementary material Table S2). There was a significant positive relationship between the CT_min and extracellular [K⁺] among species (r=0.72, d.f.=20, P<0.001; Fig. 3A). Resting haemolymph [Na⁺] of Drosophila species (at 21.5°C) ranged from 47.6±4.7 mmol l⁻¹ (D. immigrans) to 154.1±6.7 mmol l⁻¹ (D. sechellia; supplementary material Table S2), and did not have significant phylogenetic signal (K=0.43, P=0.125; Fig. 3B). Drosophila immigrans had remarkably low [Na⁺], and when this species was included in the correlation of [Na⁺] against the CT_min the relationship was non-significant (r=0.20, d.f.=20, P=0.371). However, if D. immigrans is excluded, there was a significant positive relationship between the CT_min and haemolymph [Na⁺] (r=0.48, d.f.=19, P=0.029). Thus cold-tolerant species had low concentrations of both Na⁺ and K⁺ in their haemolymph, and these relationships between ion concentrations and the CT_min both remained significant following PIC regression ([K⁺]: F_1,20=12.3, P=0.002; [Na⁺] not including D. immigrans: F_1,19=8.5, P=0.008).

We measured the osmolality of haemolymph collected under control conditions to determine whether the observed reductions in ion concentrations in cold-tolerant flies drive reductions in extracellular osmolality. Despite the variation observed in haemolymph ion concentrations, haemolymph osmolality did not significantly differ between warm- (420±23 mOsm) and cold-acclimated (396±23 mOsm) D. melanogaster (t₁₁=0.67, P=0.526), and there was no relationship between haemolymph osmolality and CT_min among Drosophila species before (r=0.16, d.f.=20, P=0.485; Fig. 3C), or after PIC regression (F_1,20=1.5, P=0.240; supplementary material Fig. S2).

To examine whether variation in haemolymph ion homeostasis is associated with modulation of Na⁺/K⁺-ATPase, we measured maximal activity of this ion transporter during a temperature ramp. This approach yielded a complete data set of enzyme V_max between 3 and 22°C from each biological replicate prepared from whole-fly homogenates. Na⁺/K⁺-ATPase activity (V_max), the temperature–activity inflection point (IP), and thermal sensitivity (Ts) of activity were extracted from logistic models fitted to Na⁺/K⁺-ATPase V_max (Fig. 4; see Materials and methods for further details).

When measured at a common temperature (21.5°C) Na⁺/K⁺-ATPase activity of cold-acclimated D. melanogaster was 55% lower than that of warm-acclimated flies (t₈=4.4, P<0.001, Fig. 5A). Similarly at 6°C (the cold acclimation temperature) Na^+/K⁺-ATPase activity was 54% lower in the cold-acclimated flies. Comparing the groups at their respective acclimation temperature reveals that cold-acclimated D. melanogaster had 93% lower Na⁺/K⁺-ATPase V_max than warm-acclimated flies. Thus cold acclimation suppresses Na⁺/K⁺-ATPase activity beyond that caused by the passive effects of temperature (Fig. 5A).

The V_max of Na⁺/K⁺-ATPase at 21.5°C ranged among species from 14.4±1.7 U g⁻¹ (μmol ATP min⁻¹ g⁻¹; D. funebris) to 75.4±13.3 U g⁻¹ (D. yakuba; supplementary material Table S2). Although there was a trend for more cold-tolerant species to have low Na⁺/K⁺-ATPase V_max, there was no significant relationship between Na⁺/K⁺-ATPase activity and CT_min among species (r=0.30, d.f.=21, P=0.162; Fig. 6A). Excluding D. immigrans, the relationship between Na⁺/K⁺-ATPase activity and the CT_min approached statistical significance (r=0.42, d.f.=20, P=0.053; Fig. 6A), but we note that here D. immigrans did not substantially deviate from the other species. Na⁺/K⁺-ATPase activity had strong phylogenetic signal (K=1.20, P=0.001), but accounting for phylogeny did not resolve any relationship between Na⁺/K⁺-ATPase activity at 21.5°C and CT_min among Drosophila species (F_1,21=0.2, P=0.670; supplementary material Fig. S2). We did, however, find a positive linear relationship between haemolymph Na⁺ concentration and log₁₀-transformed Na⁺/K⁺-ATPase activity among the Drosophila species (r=0.47, d.f.=19, P=0.030; supplementary material Fig. S3).

Contrary to our hypothesis that variation in cold tolerance is driven by changes in the thermal sensitivity of Na⁺/K⁺-ATPase, neither the inflection point (t₉=1.8, P=0.101) nor thermal sensitivity (t₉=1.2, P=0.261) of Na⁺/K⁺-ATPase activity differed significantly between warm- and cold-acclimated D. melanogaster (Fig. 5B). Similarly, the CT_min correlated with neither the thermal sensitivity (r=182, d.f.=21, P=0.403; Fig. 6C), nor the inflection point (r=0.19, d.f.=21, P=0.385; Fig. 6B) of Na⁺/K⁺-ATPase activity among Drosophila species. There was also no relationship between the PICs of CT_min and the inflection point (F_1,21=1.3, P=0.262) or thermal sensitivity (F_1,21=3.3, P=0.083; Fig. 6C), and neither the Na⁺/K⁺-ATPase inflection point (K=0.17, P=0.841) nor thermal sensitivity (K=0.20, P=0.727) had significant phylogenetic signal. Thus we detected no alterations in the shape or position of the in vitro activity–temperature curve for Na⁺/K⁺-ATPase, within or among Drosophila species, but instead observed wholesale reductions in pump activity (particularly in cold-acclimated D. melanogaster).

Plasticity of Na⁺/K⁺-ATPase activity can result from transcriptional, translational or post-translational changes (Bertorello et al., 1991; McDonough and Farley, 1993). We examined whether the observed reduction in Na⁺/K⁺-ATPase activity with cold acclimation in D. melanogaster was related to a reduction in mRNA transcript or protein abundance of the Na⁺/K⁺-ATPase subunits, using high-throughput mRNA sequencing (RNA-seq) and western blotting, respectively. Transcript abundance of the primary Na⁺/K⁺-ATPase α-subunit gene (Atpα) did not differ between acclimation groups (t₉=1.0, P=0.723; supplementary material Fig. S4). Drosophila melanogaster has three genes that code for the Na⁺/K⁺-ATPase β-subunit (nrv1, nrv2 and nrv3). Cold-acclimated flies had significantly higher expression of nrv2 mRNA than warm-acclimated flies (t₉=3.7, P=0.003), but the relative abundance of either nrv1 (t₉=0.7, P=0.897) or nrv3 (t₉=1.9, P=0.257; supplementary material Fig. S4) did not differ between warm- and cold-acclimated flies. Similarly, cold acclimation did not change the protein abundance of either the α-subunit (F_2,11=0.9, P=0.379) or β-subunit (F_2,6=0.2, P=0.835; supplementary material Fig. S4) of Na⁺/K⁺-ATPase.

When exposed to 0°C for 6 h, warm-acclimated D. melanogaster lost the ability to maintain ion and water homeostasis, leading to an increase in haemolymph [K⁺]. This increase in haemolymph [K⁺] also occurred during cold exposure in chill-susceptible cockroaches (Koštál et al., 2006), crickets (MacMillan and Sinclair, 2011b), locusts (Findsen et al., 2013) and firebugs (Koštál et al., 2004), which suggests that cold exposure leads to a loss of [K⁺] balance in both holometabolous and hemimetabolous chill-susceptible insects. Muscle and nerve resting potentials depend mostly on extracellular [K⁺] (Hoyle, 1953; Armstrong et al., 2012). Therefore high haemolymph [K⁺] is likely to depolarize cell membranes, prevent neuromuscular signal transmission (Hosler et al., 2000), slow chill-coma recovery (MacMillan et al., 2012; MacMillan et al., 2014) and lead to chilling injury (Koštál et al., 2006; MacMillan and Sinclair, 2011b).

In contrast to the warm-acclimated flies, D. melanogaster acclimated to 6°C maintained low extracellular [K⁺] after 6 h at 0°C. Phenotypic plasticity in cold tolerance has been consistently associated with an enhanced ability to maintain ion and water balance during cold exposure; cold-acclimated crickets, tropical cockroaches and adult firebugs also maintain haemolymph [K⁺] at low levels during cold exposure (Koštál et al., 2004; Koštál et al., 2006; Coello Alvarado, 2012). Thus, like other insects, acclimation to low temperatures appears to improve the ability of Drosophila to maintain low haemolymph [K⁺] during cold exposure. This ability is probably driven by maintenance of Na⁺ and water balance in the haemolymph.

We observed that cold-tolerant Drosophila species maintain lower concentrations of both Na⁺ and K⁺ in their haemolymph, and that modifications to cold tolerance were generally accompanied by inverse modifications in haemolymph ion concentrations across the Drosophila phylogeny (because these relationships remained significant following PIC analysis). Similarly, cold-acclimated D. melanogaster constitutively maintain low extracellular [Na⁺], but acclimation has no effect on haemolymph [K⁺] at rest.

Decreased haemolymph [Na⁺] could improve cold tolerance by reducing the influence of Na⁺ gradients on water balance, but only if other osmolytes fill the ‘osmotic gap’ left by Na⁺ and thereby maintain haemolymph osmotic pressure (Wyatt, 1961; Pierce et al., 1999). Adult Lepidoptera maintain low haemolymph [Na⁺], in some cases low enough to reverse the muscle Na⁺ gradient (Fitzgerald et al., 1996). To maintain haemolymph osmotic balance in the absence of Na⁺, Lepidoptera maintain high haemolymph concentrations of carbohydrates (Wyatt and Kalf, 1957; Wyatt, 1961), and, as a group, maintain muscle excitability to lower temperatures than members of Diptera and Hymenoptera, which have more ‘conventional’ (high) extracellular [Na⁺] (Natochin and Parnova, 1987; Goller and Esch, 1990). Although we observed low cation concentrations in more cold-tolerant flies, we observed no differences in haemolymph osmolality between warm- and cold-acclimated D. melanogaster, nor among Drosophila species. Thus, reductions in haemolymph [Na⁺] in Drosophila must be paired with accumulations of other extracellular osmolytes.

Small accumulations of extracellular osmolytes in chill-susceptible insects may protect organismal water balance at low temperatures. A variety of organic solutes act as cryoprotectants at very high concentrations, and are central to the overwintering success of freeze-tolerant and freeze-avoidant insects (Lee, 1991; Storey, 1997). Relatively modest amounts of such compounds, at concentrations unlikely to yield cryoprotection, have been noted to accumulate following cold-exposure in chill-susceptible insects, including D. melanogaster (Lee et al., 1987; Overgaard et al., 2007). For example, brief cold exposures that improve subsequent cold tolerance in flesh flies (rapid cold-hardening) increase haemolymph glycerol concentrations from 28 to 81 mmol l⁻¹, a change that is too small to significantly change the freezing point (Lee et al., 1987). Many cryoprotectants are also osmoprotectants (Yancey, 2005; Sinclair et al., 2013; Teets et al., 2013). Thus we propose that these relatively small accumulations of organic solutes that have been previously observed serve to maintain or increase haemolymph osmotic pressure in the place of highly permeable osmolytes (particularly Na⁺), and thereby permit a decoupling of Na⁺ and water balance that promotes low temperature survival. We hypothesize that such a change in osmotic balance may have facilitated the evolution of more extreme cold tolerance phenotypes (such as freeze avoidance and freeze tolerance) associated with accumulations of low molecular weight cryoprotectants.

Cold-acclimated D. melanogaster had significantly reduced Na⁺/K⁺-ATPase activity, and the low haemolymph [Na⁺] observed in cold-adapted Drosophila species was positively correlated with maximal activity of Na⁺/K⁺-ATPase in vitro. The ubiquitous Na⁺/K⁺-ATPase plays important roles in both cellular and whole-organism ion balance, but importantly functions in the Drosophila renal system to maintain high extracellular [Na⁺] (see Introduction). Our observations are thus consistent with our expectation that Na⁺/K⁺-ATPase activity influences (but does not wholly determine) haemolymph [Na⁺] and may play a role in determining thermal tolerance through modification of resting [Na⁺] gradients. Given the complexity of insect renal physiology and the fact that our estimates of Na⁺/K⁺-ATPase activity were necessarily based on whole-organism homogenates, the observation that Na⁺/K⁺-ATPase activity does not directly correlate with the CT_min among species is perhaps not surprising, and suggests that selection for cold tolerance does not act directly on resting Na⁺/K⁺-ATPase activity.

Phenotypic plasticity of Na⁺/K⁺-ATPase activity in D. melanogaster does not appear to be achieved through changes in mRNA or protein abundance, as we observed no differences in the abundance of either the α- or β-subunit proteins and no changes in the expression of subunit transcripts except for a modest increase in one β-subunit isoform (nrv2). Alternatively, modulation of Na⁺/K⁺-ATPase activity may involve differential expression of the 11 tissue-specific Atpα alternative transcripts that are known (but could not be distinguished accurately in our RNA-seq experiment because of the close similarity in sequence among isoforms) (Marygold et al., 2013), or post-translational modification. For example, suppression of Na⁺/K⁺-ATPase activity though reversible phosphorylation is associated with winter diapause in the goldenrod gall fly (McMullen and Storey, 2008). We suggest that post-translational modification or alternative isozyme expression reduce Na⁺/K⁺-ATPase activity and improve organismal cold tolerance through reduced Na⁺ gradients.

Reductions in the effects of temperature on enzyme activity can allow for comparatively higher rates of catalysis at low temperatures (Galarza-Muñoz et al., 2011). However, we found no evidence to support the hypothesis that cold-acclimated and cold-adapted Drosophila better maintain ion balance at low temperatures through reductions in the thermal sensitivity of Na⁺/K⁺-ATPase. Although the estimates of Na⁺/K⁺-ATPase V_max we obtained in vitro are reliable indicators of relative differences in Na⁺/K⁺-ATPase activity, they may not approximate rates of ion transport in vivo, where local substrate and co-factor concentrations and the immediate membrane environment could have a substantial impact on ion transport rates and thermal sensitivity. Adaptation and acclimation to low temperatures in Drosophila have been associated with decreased saturation of phospholipid fatty acids, which would increase membrane fluidity and maintain ion pump function at low temperatures (Ohtsu et al., 1993; Ohtsu et al., 1998; Overgaard et al., 2005; Overgaard et al., 2008) (but see MacMillan et al., 2009). Comparative analyses of Na⁺/K⁺-ATPase activity functioning within the membrane bilayer are needed to elucidate the role of these influences on ion balance in the cold. Because the physiological mechanisms underlying insect ionoregulation are diverse (O'Donnell, 2008), the reductions in Na⁺/K⁺-ATPase activity observed here may be one of several mechanisms of ionoregulation that contribute to cold tolerance. The roles, for example, of other ion-motive pumps (i.e. H⁺-ATPase and Ca²⁺-ATPase) and exchangers, ion channels and aquaporins, paracellular routes of ion and water transport, and the hormones that regulate them remain to be explored in the context of insect cold acclimation and adaptation.

Cold-acclimated D. melanogaster and cold-tolerant Drosophila species have reduced reliance on Na⁺ as an extracellular cation and more cold-tolerant species also maintain lower extracellular [K⁺]. Cold-tolerant flies probably maintain osmotic balance by accumulating other organic and inorganic osmolytes in their haemolymph to replace Na⁺ as the primary determinant of water balance. These changes to ionic and osmotic homeostasis would limit Na⁺ and water migration, maintain K⁺ balance in the cold, and thereby improve cold tolerance. Thus similar ionoregulatory mechanisms appear to underlie both phenotypic plasticity and cold tolerance evolution, which could lead to trade-offs between basal and plastic responses to cold, and suggest an evolutionary route for the cryoprotectant-mediated cold tolerance strategies of freeze tolerance and freeze avoidance.

We examined a total of 24 species from the genus Drosophila. Complete information on stock origins is presented in supplementary material Table S1. All species were maintained on the same banana, barley malt and yeast-based medium (Nyamukondiwa et al., 2011), except that the diet of four cactophilic species (D. mojavensis, D. obscura, D. persimilis and D. pseudoobscura) was supplemented with 2.1 g l⁻¹ Opuntia ficus-indicta powder (OroVerde Export, Morelos, Mexico). With the exception of cold-acclimated D. melanogaster (described below), all flies were maintained at a constant temperature of 21.5±0.5°C, and at 50±5% relative humidity with a 13 h:11 h light:dark cycle. Newly eclosed adult Drosophila were transferred, without anesthesia, to fresh 35 ml vials containing food medium. Following transfer, adults were returned to 21.5°C for 5 days before use in experiments. For the D. melanogaster acclimation experiments, virgin males were collected under light CO₂ anesthesia (<10 min) on the day of their emergence, divided randomly into two groups which were placed at either 21.5°C (warm-acclimated) or 6±0.5°C conditions (cold-acclimated) for 5 days to acclimate.

The CT_min of each Drosophila species was quantified as previously described (Ransberry et al., 2011). Adult flies were transferred into a custom-built, temperature-controlled 150×25 cm glass column containing aluminium baffles to which the flies cling [similar to the design of Huey et al. (Huey et al., 1992)]. The temperature of the column was controlled by circulating an ethylene glycol:water (1:1) mixture through the column jacket from a refrigerated circulating bath (model 1157P, VWR International, Radnor, PA, USA). The temperature inside the column was independently monitored by four type-T thermocouples (two at both the top and bottom of the column) and a TC-08 interface connected to a computer running Picolog version 5.20.1 (Pico Technology, St Neots, UK). Adult flies were released into the column where they clung to the baffles, and the temperature was held at 21°C for 15 min before being reduced at 0.1°C min⁻¹. At their CT_min, flies lost the ability to cling to the baffles and fell into a collecting tube containing soapy water that was changed every 1°C (10 min). Flies collected from the column were frozen, and later sorted according to sex and counted. For the interspecific comparisons a mean of 206 (range: 78–695) flies were used to determine the CT_min of each sex in each Drosophila species (supplementary material Fig. S1).

Adult Drosophila were positioned for haemolymph sampling using a custom-made apparatus previously described (MacMillan and Hughson, 2014). Briefly, adult flies were moved directly from their rearing vial and positioned head-first in a 10 μl pipette tip through a system of rubber tubing by air flow. The end of the pipette tip was then removed to expose the antennae, an antenna was amputated at its first segment, and a clear droplet of haemolymph was secreted. The pipette tip, with the fly and droplet attached, were immediately removed from the rest of the device and the droplet was placed under hydrated paraffin oil for measurement of Na⁺ and K⁺ concentrations by the ion-selective microelectrode technique or osmolality by nanoliter osmometry (see below). The time from removal of a fly from its acclimation temperature or from 0°C to measurement of ion concentration or osmolality of the haemolymph was less than 2 min.

Droplets of haemolymph from adult flies were used to measure extracellular ion concentrations and osmolality. Different flies were used to measure each trait and a single droplet was taken from a single fly. Haemolymph ion concentrations and osmolality were measured in 4–8 (K⁺), 3–5 (Na⁺) and 3–8 (osmolality) droplets from flies of each species and osmolality was measured in 5–7 droplets from D. melanogaster males from each acclimation group. Ion concentrations were measured in 14–16 (Na⁺) and 15–21 (K⁺) droplets per treatment (control and 6 h at 0°C) from flies in each acclimation group of D. melanogaster. The D. kanekoi and D. algonquin lines were lost to mould before haemolymph ion concentrations and osmolality were measured, and so were not included in this analysis. Haemolymph Na⁺ and K⁺ concentrations were measured in all Drosophila at rearing temperature (21.5±1°C). For the cold exposure, D. melanogaster were transferred to microcentrifuge tubes and submerged in an ice-water slurry (0°C) for 6 h. Flies that received a cold exposure were sampled immediately following removal from the cold to 21.5°C.

Ion concentration was measured using an ion-selective microelectrode (ISME) technique using pulled glass microelectrodes front-filled with ionophore cocktails (Jonusaite et al., 2011). Ion-selective microelectrodes were constructed by pulling borosilicate glass capillaries [TW-150-4, World Precision Instruments (WPI), Sarasota, FL, USA] to a tip diameter of ~5 μm using a P-97 Flaming Brown micropipette puller (Sutter Instruments, Novato, CA, USA). Pulled micropipettes were silanized at 300°C with N,N-dimethyltrimethylsilylamine vapour for 1 h and backfilled with 100 mmol l⁻¹ KCl or NaCl. Microelectrodes were then front-filled with ionophore cocktails for either K⁺ (K⁺ ionophore I, cocktail B, Sigma-Aldrich, St Louis, MO, USA) or Na⁺ (Na⁺ ionophore X; Messerli et al., 2008) and dipped in a solution of polyvinylchloride (Sigma-Aldrich) in tetrahydrofuran (Sigma-Aldrich). A borosilicate glass (IB200F-4, WPI) reference electrode back-filled with 0.5 mol l⁻¹ KCl was used to complete the circuit. Voltage was recorded using a ML 165 pH amplifier and PowerLab 4/30 data acquisition system connected to a computer running LabChart 6 software (AD Instruments, Colorado Springs, CO, USA).

Haemolymph osmolality was measured using a Clifton Nanoliter Osmometer (Clifton Technical Physics, Hartford, NY, USA). Small droplets obtained as described above (~20 nl) were suspended in wells filled with type B immersion oil under a Nikon SMZ 1500 microscope equipped with a Nikon Digital Sight DS-Fil camera connected to software NIS-Elements D2.30 SP4 Laboratory Imaging software (Nikon Corporation, Tokyo, Japan) and rapidly cooled until frozen. The droplets were then warmed slowly until the temperature at which one last crystal remained visible, before the crystal was warmed again to determine the melting point (i.e. the temperature at which the last crystal disappeared). The melting point was used to determine osmolality, as one mole of solute will decrease melting point by 1.86°C.

To quantify maximal Na⁺/K⁺-ATPase activity, whole flies (~80 mg of pooled adults; 20–80 flies, depending on species) were transferred to 1.7 ml microcentrifuge tubes without anaesthesia, snap frozen in liquid nitrogen vapour, and stored at −80°C. Na⁺/K⁺-ATPase activity was measured in six biological replicates of each temperature acclimation group of D. melanogaster, and 3–6 biological replicates of each species were used for the interspecific analysis. Drosophila arawakana was excluded from this analysis because of insufficient sample size (N=1). Frozen Drosophila were weighed to obtain pooled fresh mass in pre-weighed 2 ml microcentrifuge tubes before being homogenized on ice in 1 ml of homogenization buffer (25 mmol l⁻¹ imidazole, 0.2% w/v Na⁺-deoxycholate, 10 mmol l⁻¹ β-mercaptoethanol, 2 mmol l⁻¹ EDTA, pH 7.5) with a Tissue-Tearor (Biospec Products, Bartlesville, OK, USA) using four 10 s bursts each followed by 20 s rest on ice. Homogenized samples were sonicated (Virsonic 100; VirTis, Gardiner, NY, USA) following the same timing of bursts and rests, and centrifuged at 7000 g for 5 min at 4°C. Size-exclusion filtration columns, which permit the passage of proteins larger than ~50 kDa, were prepared by plugging the tip of a 3 ml plastic syringe barrel with glass wool and adding 3 ml of Sephadex G-50 (GE Healthcare, Waukesha, WI, USA). Columns were stored at 4°C for a maximum of 2 weeks before use, and were conditioned by eight passes of 300 μl of homogenization buffer and 1 min centrifugations (500 g). A conditioned column was placed into a clean 5 ml plastic test tube, a 300 μl aliquot of supernatant derived from the homogenate was added to the column, and the column within the tube was centrifuged at 500 g for 1 min to draw the sample through the column and into the tube.

Filtered fly homogenates were diluted 7-fold in homogenization buffer immediately before a 20 μl aliquot was added to a 1 ml cuvette containing 880 μl of assay buffer initially at 23°C. The reaction was initiated by the addition of 100 μl of a 50 mmol l⁻¹ ATP solution. Final conditions for the assay were: 70 mmol l⁻¹ imidazole (pH 7.5), 140 mmol l⁻¹ NaCl, 30 mmol l⁻¹ KCl, 7 mmol l⁻¹ MgCl₂, 4 mmol l⁻¹ phosphoenolpyruvic acid, 300 μmol l⁻¹ NADH, 5 mmol l⁻¹ ATP, 50 U ml⁻¹ pyruvate kinase (EC: 2.7.1.40) and 50 U ml⁻¹ lactate dehydrogenase (EC: 1.1.1.27).

Maximal activity of Na⁺/K⁺-ATPase was measured across a range of temperatures using a thermally dynamic, spectrophotometric assay. The absorbance of NADH at 340 nm was recorded using a Cary 100 Bio spectrophotometer with a Cary Peltier-effect temperature controller (Agilent Technologies, Santa Clara, CA, USA) connected to a computer running WinUV Thermal Application version 3.0 (Agilent Technologies). Four replicate cuvettes of each Drosophila sample were run, two of which contained 1 mmol l⁻¹ ouabain (a specific inhibitor of Na⁺/K⁺-ATPase). Temperature inside a dummy cuvette was monitored by a ceramic temperature probe (Agilent Technologies) interfaced with the spectrophotometer, which controlled the rate of temperature change. A type-T thermocouple was also suspended in the dummy cuvette and connected to a TC-08 interface (Pico Technology), which measured the temperature inside the cuvette every second for the duration of each sample run. The rate of temperature change inside the cuvette was consistent throughout the temperature ramp (supplementary material Fig. S5A). Activity of Na⁺/K⁺-ATPase was determined as the difference between the rates of cuvettes containing 1 mmol l⁻¹ ouabain and cuvettes that did not contain ouabain (supplementary material Fig. S5B). For each replicate, rates of change in absorbance (OD min⁻¹) were smoothed using a 15-point sliding window in LoggerPro (version 3.8.4, Vernier Software Inc., Beaverton, OR, USA).

Logistic functions provided a good fit to the data (Fig. 4). As Max in this equation represents the enzyme reaction rate at T=∞, and enzymes denature at high temperatures, this parameter is not biologically relevant and was not used in further analyses. Estimates of Ts and IP were obtained for each sample from the fitted models, and rates across the range of temperatures (3–21.5°C) were extracted from models using the predict() function in R (which allowed us to standardize all rates to common temperature intervals).

To confirm that the thermally ramped assays described above did not yield different results from a more traditional method, the activity of Na⁺/K⁺-ATPase was also measured at five static temperatures (6, 11, 13.3, 17.6 and 21.5°C) in N=4 samples of male D. melanogaster acclimated to 21.5°C (supplementary material Fig. S5C). The methods of sample preparation and the final assay conditions for the static-temperature Na⁺/K⁺-ATPase assays were the same as those used for the thermally dynamic assay. Once the temperature inside the dummy cuvette was stable at the set temperature (determined from the thermocouple trace), the reaction was initiated by the addition of ATP as in the thermally dynamic assay. Rates of Na⁺/K⁺-ATPase activity in OD min⁻¹ over a 10 min period of recording were converted to moles of ATP consumed per minute using the Beer–Lambert law. Static and dynamic methods of the Na⁺/K⁺-ATPase assay yielded similar estimates of Na⁺/K⁺-ATPase activity in male D. melanogaster acclimated to 21.5°C (supplementary material Fig. S5C).

The abundance of the Na⁺/K⁺-ATPase α- and β-subunit proteins was quantified by western blot in seven (α) and four to five (β) biological replicates of each acclimation group of D. melanogaster. Twenty adult male flies were freshly homogenized using a glass Dounce homogenizer in 1 ml of buffer containing detergent and protease inhibitors (160 mmol l⁻¹ Tris-HCl, 0.2 % w/v Na⁺-deoxycholate, 1 mmol l⁻¹ phenylmethanesulfonylfluoride, 5 μg ml⁻¹ aprotinin and leupeptin, pH 6.8). Homogenized samples were centrifuged at 7000 g and 4°C for 5 min. A 200 μl aliquot was reserved for protein quantification, and a second 200 μl aliquot was added to an equal volume of a 2× concentrated loading buffer to yield final loading conditions [4 mol l⁻¹ urea, 4% w/v sodium dodecyl sulfate (SDS), 4% v/v β-mercaptoethanol, 4% v/v glycerol, 0.005% w/v bromophenol blue)]. Samples were loaded (15 μg soluble protein) on SDS-polyacrylamide gels (10% w/v resolving gel, 5% w/v stacking gel) in a SE600 electrophoresis unit (Hoefer Inc., Holliston, MA, USA). Loaded samples were separated by electrophoresis (120 V for ~2.5 h) in running buffer (25 mmol l⁻¹ Tris-base, 192 mmol l⁻¹ glycine, 0.3% w/v SDS, pH 8.6) and resolved proteins were wet transferred at 4°C to polyvinylidene fluoride (PVDF) membranes (Biotrace, Pall Corporation, Port Washington, NY, USA) at 15 V overnight (Newington et al., 2011). The following morning membranes were soaked in blocking buffer (20 mmol l⁻¹ Tris-base, 150 mmol l⁻¹ NaCl, 0.05% v/v Tween 20, 3% w/v BSA, 1.5% w/v dry blotting milk, pH 7.5) for 1 h at 22°C. Blocked membranes were washed (3×5 min) in wash buffer (20 mmol l⁻¹ Tris-base, 150 mmol l⁻¹ NaCl, 0.05% v/v Tween 20, pH 7.5). Membranes were incubated in primary antibody buffer (20 mmol l⁻¹ Tris-base, 150 mmol l⁻¹ NaCl, 0.05% v/v Tween 20, 3% w/v BSA, 0.05% w/v sodium azide, pH 7.5) for 1 h at 22°C. The primary antibody buffer contained mouse antibodies targeting either the α- [a5; 1:500; antigen species: chicken; Developmental Studies Hybridoma Bank (DSHB), Iowa City, IA, USA] or β-subunit (nrv5F7; 1:50; antigen species: Drosophila; DSHB) of Na⁺/K⁺-ATPase (Lebovitz et al., 1989; Sun and Salvaterra, 1995). Probed membranes were then washed (3×5 min, as above), incubated in blocking buffer containing goat anti-mouse IgG horseradish peroxidase conjugate secondary antibody (1:10,000; Bio-Rad Laboratories Inc., Hercules, CA, USA), and washed again. Following 1 min incubation in SuperSignal West Pico chemiluminescent substrate (Thermo Fisher Scientific, Rockford, IL, USA), images of immunoblots were produced using a ChemiDoc digital imaging system (Bio-Rad). After probing, membranes were stained for total protein as a loading control following described methods (Welinder and Ekblad, 2011). Membranes were washed (2×5 min), stained with 0.1% Coomassie R-350 in methanol:water (1:1) for 1 min, and destained with acetic acid:ethanol:water (1:5:4) for 20 min. Following destaining, the membranes were air-dried and scanned at 600 d.p.i. on a flatbed scanner. Densitometric quantification of immunoblots and total protein images were completed by measuring band and total lane intensity in ImageJ.

The abundance of transcripts coding for Na⁺/K⁺-ATPase subunits was quantified using high-throughput mRNA sequencing (RNA-seq). Frozen warm- and cold-acclimated male D. melanogaster (five biological replicates per acclimation temperature, each containing 25 flies) were homogenized by pestle and mortar over liquid nitrogen in 1 ml of TRIzol reagent. Homogenized samples were centrifuged at 12,000 g for 5 min at 4°C and the top 200 μl (lipids) discarded. Total RNA extraction from samples was completed using a RNeasy lipid tissue mini kit (Qiagen, Hilden, Germany). Chloroform (200 μl) was added to the tubes, samples were shaken for 15 s, and left at 22°C for 3 min. Samples were centrifuged at 12,000 g for 15 min at 4°C, the upper (aqueous) phase transferred to a new tube with 1 volume of 70% ethanol, and vortexed (30 s). A 700 μl aliquot of the sample was centrifuged (8000 g for 15 s at 22°C) through an RNeasy Mini spin column (Qiagen) in a 2 ml collection tube. The effluent was discarded and the sample (bound to the column) was washed with 700 μl of RW1 buffer (Qiagen), and 2×500 μl of RPE buffer (Qiagen) using the same centrifuge settings. The column was placed into a new collection tube and RNase-free water (50 μl) was centrifuged through the column (8000 g for 1 min at 22°C) twice to release the RNA from the column. Total mRNA was purified from RNA, and cDNA libraries were prepared using a TruSeq Stranded mRNA preparation kit (Illumina, San Diego, CA, USA) following the TruSeq RNA Sample Prep version 2 protocol (Illumina). Briefly, mRNA was purified by poly-A selection using magnetic beads bound to poly-T oligo-nucleotides and chemically fragmented. Fragmented mRNA was reverse-transcribed into single-stranded cDNA by reverse transcriptase and the RNA on the opposite strand was replaced with DNA to yield double-stranded (ds) cDNA. Overhangs at the 3′- and 5′-ends of ds cDNA were cleaved and repaired by exonuclease and DNA polymerase, respectively.

All data analyses were completed in R version 3.1 (R Development Core Team, 2013). The temperature at which 80% of flies had fallen from the temperature-controlled column (CT_min) was determined using accelerated failure time models (Therneau and Grambsch, 2000) and the aggregate() function. Haemolymph ion concentrations and osmolality were compared among treatments and acclimation groups of D. melanogaster by ANOVA with group and treatment as factors, followed by Tukey's HSD. The relationships between haemolymph ion concentration or osmolality and the CT_min among Drosophila species were determined by Pearson's product-moment correlation, with line equations for plotting provided by ranged major axis model II regressions using the lmodel2 package (Legendre, 2013). Reaction rates (V_max) at 21.5°C, Ts and IP of Na⁺/K⁺-ATPase, were compared between warm- and cold-acclimated D. melanogaster using t-tests. Among Drosophila species, the relationships between these variables and the CT_min were tested by correlation, with RMA model II linear regression used for plotting purposes, as was the relationship between haemolymph [Na⁺] and Na⁺/K⁺-ATPase activity. The abundance of each Na⁺/K⁺-ATPase subunit was compared between warm- and cold-acclimated D. melanogaster using an ANCOVA, with total protein abundance included as a covariate.

Transcriptome data manipulation and statistical analyses were completed in Galaxy (Goecks et al., 2010) following the analysis workflow of Trapnell et al. (Trapnell et al., 2012). Illumina adapter sequences were clipped from reads using FastQ Clipper (part of the FastX Toolkit), and clipped reads were aligned to the genome of D. melanogaster (Ensembl build 5.25) using TopHat (version 2.0.8; Trapnell et al., 2009), with alignment limited to known splice junctions. Cufflinks was used to assemble and count transcript reads for each sample and Cuffmerge was used to merge assembled transcripts for all biological replicates into a single reference transcriptome (version 2.0.2) (Trapnell et al., 2010). Na⁺/K⁺-ATPase subunit transcript expression was compared between warm- and cold-acclimated flies using Cuffdiff (Trapnell et al., 2013).

We calculated PICs using a phylogeny constructed by combining two recently published Drosophila trees (Fig. 1). A recent comprehensive phylogeny of the family Drosophilidae (van der Linde et al., 2010) contains all but four of the species used in this study, and was used as the base for our tree. Extraneous species were trimmed from the van der Linde et al. tree (van der Linde et al., 2010), and the four additional species (D. borealis, D. kanekoi, D. nepalensis and D. triauraria) were added from a second phylogeny (Strachan et al., 2011) with branch lengths standardized to the rest of the tree using the ratio of nearest-neighbour distance. Node ages were standardized using a semi-parametric method for use in statistical analyses.

PICs of species trait means were generated in R using the pic() function in the ape package (Paradis et al., 2004). Tests of relationships between the CT_min and physiological traits while controlling for phylogeny were conducted using linear regressions of PICs forced through the origin (Garland et al., 1992) (supplementary material Fig. S2). Species traits were also tested for phylogenetic signal – a measure of the tendency for related species to have similar trait values – by the K-statistic (Blomberg et al., 2003).

The authors wish to thank Robert Cumming and Jordan Newington for assistance with the western blots, Chris Guglielmo for access to the spectrophotometer, and Johannes Overgaard and two anonymous reviewers for their constructive criticism on an earlier version of the manuscript.