Hormonal and metabolic responses to upper temperature extremes in divergent life-history ecotypes of a garter snake

Temperature constrains organismal performance in a wide array of traits (Angilletta, 2009; Huey, 1982; Huey et al., 2012). Although ectotherms can tolerate a broad range of body temperatures, many species might soon experience temperatures above those optimal for performance, including growth and reproduction, but below lethal limits as a consequence of ongoing climate change (Diamond et al., 2012; Visser, 2008). Temperatures outside of an organism's optimal range challenge homeostasis and result in reduced performance and the suspension of normal activity (Huey and Stevenson, 1979). These effects in turn reduce individual fitness and often impact population growth and persistence (Huey et al., 2012; Lemoine and Burkepile, 2012; Pörtner, 2002; Pörtner and Knust, 2007). Because physiology largely mediates organismal responses to stressful environments, understanding the physiological mechanisms responsible for declining fitness at environmental temperatures above thermal optima is crucial to quantifying organismal responses to environmental change (Angilletta, 2009). Additionally, characterizing the processes that set thermal tolerance limits in ectotherms, in terms of both acute and long-term maintenance of energy balance, is essential for developing predictions of how organisms will respond to projected future environments (Williams et al., 2016). Our goals in the present study are to quantify the metabolic and hormonal responses to high temperatures, test for local adaptation in the thermal stress response, and test the hypothesis that oxygen limitation is a proximate mechanism setting thermal tolerance limits.

Central to all three of these goals are measures of organismal energetics, such as metabolic rate, which are highly temperature dependent and fundamental to recent models of organismal response to climate change (Bozinovic and Pörtner, 2015; Dillon et al., 2010; Gillooly et al., 2001; Glazier, 2015; Pörtner and Farrell, 2008). Even so, vertebrate ectotherms such as squamate reptiles (i.e. snakes and lizards) exhibit intraspecific variation in metabolic rate beyond temperature dependence, suggesting that additional physiological mechanisms can modulate metabolic rates. For example, metabolic rate can be temperature insensitive within a range (Aleksiuk, 1971; Davies et al., 1981; Seidel and Lindeborg, 1973), be dependent on ecological role (e.g. foraging strategy; Andrews and Pough, 1985; Angilletta, 2001; Beaupre, 1993; Sears, 2005; Taylor and Davies, 1981; Zaidan, 2003; Zari, 2013) or exhibit temperature-dependent scaling with body size (Buikema and Armitage, 1969; Davies and Bennett, 1981; Dmi'el, 1972; Gangloff et al., 2015; Vinegar et al., 1970). Both the standard effect of temperature to increase metabolic rate and the capacity of organisms to adjust that relationship (for example, via the effects of glucocorticoids) will determine the energetic cost of high-temperature exposure.

Characterizing organismal responses to increased temperatures requires examining the endocrine system, which likely mediates metabolic plasticity in squamates. For example, glucocorticoids, such as corticosterone (CORT) in reptiles, can increase metabolic rate independent of temperature (Bradshaw, 2003; DuRant et al., 2008; Preest and Cree, 2008). In general, we expect that ambient temperatures that approach critical limits will induce an acute endocrine-driven stress response as an individual attempts to both survive and avoid permanent damage (McEwen and Wingfield, 2010; Romero et al., 2009). Glucocorticoids are major mediators of energy balance during the stress response in vertebrates, serving to both mobilize energy stores and maintain longer-term energetic homeostasis (Johnstone et al., 2012; Wingfield and Kitaysky, 2002). Even so, relatively few studies have examined the temperature dependence of glucocorticoids in reptiles (but see Dupoué et al., 2013; Sykes and Klukowski, 2009; Telemeco and Addis, 2014). Moreover, CORT and insulin generally antagonistically regulate circulating glucose concentrations (Dallman et al., 1995; Kaplan, 1996; Strack et al., 1995). In birds, typical CORT and insulin functions are maintained throughout a capture–restraint stress response (Remage-Healey and Romero, 2001). However, heat stress seems to de-couple this counter-regulatory relationship between glucocorticoids and insulin in many mammals. Despite reduced nutrient intake, the glucocorticoid response is accompanied by increased insulin levels, which may aid in the activation and upregulation of heat shock proteins (reviewed in Baumgard and Rhoads, 2013; Li et al., 2006). Few studies have measured insulin in reptiles and none have categorized the insulin response to temperature extremes. By measuring both CORT and insulin concentrations, and their subsequent effects on glucose concentration, we can quantify the impact of heat stress on a specific aspect of energy regulation.

An additional goal of this study is to assess the role of oxygen limitation in setting thermal limits. The recent oxygen and capacity-limited thermal tolerance (OCLTT) hypothesis posits that ectotherms suffer from a mismatch between oxygen demand and supply at high temperatures, leading to reduced energetic capacity and, ultimately, physiological collapse (Pörtner, 2002). A number of physiological indicators identify this limit, including reduction of whole-organism metabolic rate and an increase in circulating anaerobic metabolites, such as lactate (Frederich and Pörtner, 2000; Pörtner, 2002). Thus, by simultaneously measuring oxygen consumption rate and markers of anaerobic respiration across increasing temperatures, we may elucidate the proximate cause of organismal failure (i.e. death) at critically high temperatures.

The divergent life-history ecotypes in populations of the western terrestrial garter snake [Thamnophis elegans (Baird & Girard 1853)], around Eagle Lake, CA, USA, provide an excellent system in which to study physiological response to high temperatures and thermal adaptation. Here, replicate fast-living lakeshore populations (hereafter, ‘L-fast’) and slow-living meadow populations (hereafter, ‘M-slow’) share a common ancestral source (Manier and Arnold, 2005) but have evolved distinct life-history ecotypes on the pace-of-life continuum (sensu Ricklefs and Wikelski, 2002), diverging in growth, reproduction, lifespan, immune investment and endocrine function (Bronikowski, 2000; Bronikowski and Arnold, 1999; Palacios et al., 2011; Schwartz and Bronikowski, 2011; Sparkman et al., 2007). The selective forces driving these ecotypic differences likely include thermal environment, with air temperature at the lakeshore habitats averaging 5–10°C warmer during the active season than the montane meadows (Bronikowski, 2000; Bronikowski and Arnold, 1999) and corresponding differences in retreat-site temperatures (Huey et al., 1989; Peterson et al., 1993). Additionally, stress physiology pathways likely contribute to the divergent life-history ecotypes (Schwartz and Bronikowski, 2013). This system of locally adapted populations provides a unique opportunity to test the potential of the thermal environment to shape aerobic performance curves, shifts in the temperature dependence of energetic trade-offs, and divergences in endocrine pathways shaping the evolution of life histories (Andrews and Pough, 1985; Angilletta, 2001; Beaupre, 1995; Dunham et al., 1989).

We exposed snakes from both ecotypes to ecologically relevant ambient temperatures ranging from moderate to near-lethal, and measured metabolic rate (oxygen consumption rate) and physiological markers of energy regulation and processing (plasma CORT, insulin, glucose and lactate concentrations). We tested hypotheses focusing on three aspects of the physiological response to high temperatures. First, we hypothesized that temperatures above natural activity temperatures (35°C and above) induce a hormonal stress response (hypothesis 1). Specifically, we predicted that both circulating CORT and glucose concentrations would increase with temperature (glucose: Baumgard and Rhoads, 2013; CORT: Schwartz and Bronikowski, 2013). Additionally, we predicted that insulin concentration would increase with increasing temperatures independent of hyperglycemia, similar to the pattern observed in mammals. That is, we predicted that insulin and glucose concentrations would respond similarly in overall pattern, but exhibit different within-individual dynamics. Second, we hypothesized that the thermal tolerances of the ecotypes are locally adapted to their divergent thermal environments, as are other aspects of their biology, and that this divergence would be maintained in adult snakes even after an extended period in common-garden conditions (hypothesis 2). Thus, we predicted that L-fast snakes would tolerate higher temperatures before displaying physiological stress and would show an attenuated stress response at high temperatures in comparison with M-slow snakes. Finally, we tested the OCLTT hypothesis for the mechanism setting thermal limits in our T. elegans populations (hypothesis 3). If thermal tolerance results from oxygen limitation, we predict that oxygen consumption rate will plateau at near-critical high temperatures, plasma glucose concentration will drop, and lactate concentration will increase (Frederich and Pörtner, 2000; Verberk et al., 2016). In quantifying both the hormonal effects on energy mobilization and the whole-organism metabolic response across a range of high temperatures, this study provides insight into the impact of acute exposure to high temperatures on the regulation of homeostasis and energy stores.

Female T. elegans were wild-caught from three fast pace-of-life lakeshore populations (L-fast; N=13) and four slow pace-of-life meadow populations (M-slow; N=15) around Eagle Lake, Lassen County, CA, USA in June 2010. All snakes were caught as reproductive-sized adults [snout–vent length (SVL); M-slow range: 416–568 mm; L-fast range: 460–692 mm]. Snakes were transported to Iowa State University and individually housed in 40 gallon (=150 liters) glass aquaria with ground corn-cob substrate and a plastic bowl that served as both water dish and retreat site. They were placed on a thermal gradient for 24 h day⁻¹ (range: 25–34°C) and kept on a 12 h:12 h light:dark schedule. We offered frozen/thawed mice once per week, recording the amount eaten at each feeding and increasing the amount offered weekly according to the amount eaten and in relation to body size. For each of the 4 years in captivity, snakes were hibernated uniformly in the dark at 4°C from January to May. Four days before the experiment began, we measured body mass (in g; M-slow range: 35.6–86.2, s.e.m.=3.3; L-fast range: 63.2–120.5, s.e.m.=5.3) and SVL (in mm; M-slow range: 454–590, s.e.m.=9.6; L-fast range: 545–736, s.e.m.=13.4). Body condition was calculated as the residual of the log-mass on log-SVL regression (Palacios et al., 2013; Weatherhead and Brown, 1996) and did not differ between the ecotypes at the start of the experiment (t-test, P=0.88). Utilizing animals maintained under identical conditions allowed us to control for any immediate external effects on physiology, which is known to vary seasonally and with environmental conditions (Palacios et al., 2013, 2012). Fieldwork was conducted under permit from the California Department of Fish and Game, and treatment of all experimental animals was in accordance with Iowa State University Institutional Animal Care and Use Committee protocol 3-2-5125-J.

We conducted the experiment across 15 consecutive days beginning 18 May 2014, with snakes recently emerged from hibernation (5 May 2015) and fasted throughout hibernation and the experiment. Although this time roughly corresponds to the reproductive period and vitellogenesis in wild snakes, which can be energetically costly (e.g. Van Dyke and Beaupre, 2011), females likely did not vary in reproductive status because of similarities in age, body condition and time in captivity. Snakes were divided into five blocks of four to six snakes and each block was subjected to a 3 h treatment at each of five ambient temperatures (24, 28, 32, 35 and 38°C) with temperature order randomized, allowing for a complete, randomized, repeated-measures statistical design. Thus, there were 28 snakes×5 temperatures for N=140 measures of metabolism and physiology. Thamnophis elegans demonstrates a preferred activity range in the wild of 26–32°C (Scott et al., 1982), with a mean field-active body temperature of 29.9°C (Peterson, 1987). Snakes become inactive at 35°C (Scott et al., 1982), with a maximum recorded field body temperature of 36.0°C (Stevenson et al., 1985). Scott et al. (1982) report a critical thermal maximum (muscle spasms and failure of righting response) of between 43 and 44°C, and Stevenson et al. (1985) identified a median survival temperature of 42.8°C. It is not known, however, how snakes from these ecotypes differ in preferred or maximal temperatures.

Originally, the intent of our experiment was to measure responses at temperatures up to 40°C; however, two snakes exposed to this temperature during preliminary trials died within several days of treatment. We therefore concluded that an upper extreme of 38°C represents a stressful treatment near the lethal limit for fasted, post-hibernation subjects. Repeated measures were separated by 72 h, with two blocks subjected to temperature treatments from 08:30 to 11:30 h and three blocks from 12:30 to 15:30 h. We chose to randomize order of temperature exposure but not time of day within blocks to avoid confounding the effects of time of day and temperature on individual snakes. The evening preceding each treatment, snakes were moved from their home cages into 1.5 liter plastic boxes fitted with a rubber stopcock and tubing, covered with a mesh lid, and kept at 24°C overnight. At the time of the experiment, each container was sealed with an airtight plastic lid and in- and out-flow tubing was attached to the respirometry equipment. Containers were then placed in an incubator (Econotemp incubator, Thermo Fisher Scientific, Waltham, MA, USA) at the specified temperature treatment. Snakes were maintained in containers in the incubator, undisturbed and in the dark, for 3 h during oxygen consumption rate measurements. Qualitatively, we observed little movement of snakes immediately before and after the experimental procedure. Incubators were equipped with customized fans to ensure air flow and uniform temperature distribution. Incubator temperature was monitored using thermistor probes (model SENT-TH, Sable Systems, Las Vegas, NV, USA), corroborated with a calibrated electric thermometer.

Immediately following each temperature treatment, we collected blood samples (∼150 μl) from the caudal vein of each snake using heparin-rinsed syringes. Blood was immediately transferred to cold centrifuge tubes and kept on ice for a short time (<50 min), then centrifuged at 3000 g for 10 min at 4°C. We then aliquoted the plasma into separate tubes for each assay (CORT, insulin, glucose and lactate), immediately snap-froze them in liquid nitrogen, and stored them at −80°C until analysis. We recorded the time from the end of the experimental treatment to the time each snake was bled and the total handling time for each snake. All handling times were less than 10 min, which is before circulating CORT is elevated as a result of handling stress in T. elegans (Palacios et al., 2012).

Instantaneous oxygen consumption rate (V̇_O2) and carbon dioxide production rate (V̇_CO2) were measured using a push mode flow-through respirometry system (Lighton, 2008; Lighton and Halsey, 2011). V̇_O2 is a proxy for resting metabolic rate, which is a measure of steady-state maintenance in post-absorptive individuals at rest during normal activity periods (Andrews and Pough, 1985; Bennett and Dawson, 1976). Standard operating protocols were used for this system: air flow (400 ml min⁻¹) was maintained by an MFS-2 mass flow generator and cycled through eight chambers at 5-min intervals with an RM-8 flow multiplexer. One chamber in each block served as a blank to control baseline values during the experiment. Each chamber was sampled for four 5-min intervals, with the blank chamber sampled for eight intervals. Air outflow was scrubbed of water using Drierite desiccant (Hammond Drierite, Xenio, OH, USA) and then measured for oxygen and carbon dioxide content with CA-10 and FC-10 analyzers, respectively (Sable Systems). Data were recorded using a UI-2 analog–digital interface. Oxygen and carbon dioxide levels were corrected for variation in barometric pressure and then converted to V̇_O2 and V̇_CO2 using ExpeData 1.7.30 software (Sable Systems). Measurements from the first intervals (i.e. the first 45 min of the treatment) were excluded from analysis to allow time for snake body temperature to equilibrate with incubator temperature. For analysis of V̇_O2, we mathematically corrected for V̇_CO2 (following equations in Lighton, 2008) and averaged across the entire 5 min interval, which accounts for any potential periods when the snake was not breathing. Values from the three 5-min readings were then averaged, resulting in a single value of V̇_O2 for analysis for each individual at each temperature. Analysis of V̇_CO2 demonstrated the same qualitative patterns as the analysis of V̇_O2, so we present only the latter here for simplicity. During analysis, we identified a malfunction whereby the multiplexer machine did not switch between channels as programmed, leaving us with N=82 observations. Because the order of individuals was randomized within each trial, the resulting lost data are equally distributed between ecotypes and among temperature treatments (χ² test P=0.75 and P=0.54, respectively).

The concentration of CORT in blood plasma was measured using double-antibody radioimmunoassay kits (kit 07120102, MP Biomedical, Orangeburg, NY, USA). We followed previously described protocols (Robert et al., 2009), except samples were diluted 1:160 (instead of 1:40) to account for high CORT concentrations. All experimental measurements were made with a single kit over 2 days. The kit-provided controls were used to assess overall variability in measurements. Each control was run seven times on each day, yielding coefficients of variation (CV) of 6.67% and 6.48% for the low and high controls, respectively. All samples (N=140) were run in duplicate with a CV<10%. To compare measures from this experiment with baseline laboratory CORT values, we also included measures made on blood samples from these snakes (N=26) in normal husbandry conditions from August 2013. Because these samples were measured with a different CORT kit, we transformed the calculated concentrations using the equation from a linear regression of eight samples that were measured with both kits, which were strongly correlated (R²=0.93).

The relative change in blood plasma insulin concentration was measured using an enzyme-linked immunosorbent assay (ELISA) with a monoclonal insulin antibody assay kit (catalog no. MBS269923, MyBioSource, San Diego, CA, USA) following the manufacturer's protocols with slight modification to sample and reagent volumes. To validate this kit for measures of snake insulin, we tested the parallelism of kit-provided controls with a serially diluted pooled sample from a congeneric garter snake species, Thamnophis marcianus (heterogeneity of slopes: F_1,14=2.04, P=0.17). All samples were run in duplicate on three plates. A pooled sample was run in duplicate on each plate to assess inter-assay variability, resulting in a CV of 11.62% across plates. Because of the large amount of plasma necessary for this assay, insulin concentration could only be measured in a subset of samples, and duplicate samples with a CV>15% (N=8) were excluded, yielding N=85 measurements for analysis. Caution is warranted when concluding absolute insulin concentrations from a heterologous assay; however, in interpreting patterns of relative change, this approach is valid.

The concentration of glucose in blood plasma was measured using an ELISA with a glucose assay kit (Autokit Glucose C2, Wako Chemicals, Richmond, VA, USA) following the manufacturer's protocols with slight modification of reagent volumes. All samples (N=140) were run in duplicate on eight plates and re-run if CV>10%. A pooled sample was run on each plate to assess inter-plate variability, resulting in a CV of 9.9% across plates.

The concentration of lactate in blood plasma was measured using an ELISA L-lactate assay kit (kit A-108L, Biomedical Research Service & Clinical Application, Buffalo, NY, USA) following the manufacturer's protocols, except that sample volumes and deproteinization reagent (polyethylene glycol) were reduced to 20 µl. All samples were run in duplicate on four plates and values were accepted if CV<20% (per the manufacturer's recommendation). We did not have sufficient blood plasma to re-run 11 samples that exceeded this threshold, leaving N=129 values for analysis. Samples with values outside the standard curve (N=9) were assigned to the extreme ends of the standard curve. Inter-plate variability was assessed with a pooled sample run on each plate, resulting in a CV of 4.4% across plates.

where μ represents the grand mean and ε represents the error term. Only the models for V̇_O2 and CORT concentrations included the term for ecotype-specific mass and only the model for lactate concentration included the term for handling time.

To ensure that our measures represent the effect of temperature treatments and not the effect of handling stress or placement in metabolic chambers, we compared values of CORT concentration from blood samples taken during normal husbandry conditions 9 months prior to the experiment. We compared least-squares means from a repeated-measures ANOVA of log₁₀-transformed CORT concentration with sampling time (baseline or temperature treatment) as a categorical independent variable and individual as a repeated-measures random effect with an ar(1) covariance structure. To further elucidate the relationship between insulin and glucose concentrations within each temperature, we conducted an additional mixed-model analysis of log-transformed and normalized insulin concentration with glucose concentration, temperature and their interaction as independent factors and individual as a repeated-measures random effect with an ar(1) covariance structure. For all mixed models, we estimated denominator degrees of freedom for F-tests using the Kenward–Roger degrees of freedom approximation (Kenward and Roger, 1997).

Oxygen consumption rate increased with temperature, with no indication of plateau (linear effect of temperature P<0.0001). The main effects of condition and ecotype, and the interaction of ecotype and test temperature, did not influence V̇_O2 (Table 1, Figs 1A and 2). Mass did not exhibit a significant effect on V̇_O2, but we retained this as a covariate in the model because of the expected influence of mass on metabolic rates (see Discussion).

Temperature linearly increased CORT, insulin and glucose concentrations (all P≤0.01) and non-linearly affected CORT, glucose and lactate concentrations (P<0.0001, P<0.0001 and P=0.046, respectively; Table 1, Figs 1B–E and 2). Body condition did not affect hormone or metabolite concentrations (all P>0.53; Fig. 2, Table 1). Ecotype only affected glucose concentration, with M-slow snakes maintaining higher concentrations across all temperatures (P=0.048; Figs 1D and 2, Table 1). Neither the main effect of ecotype nor its interaction with temperature significantly influenced other physiological variables (all P>0.11; Table 1, Figs 1B–E and 2). Ecotype-specific mass negatively influenced CORT concentrations (P=0.0076; Table 1), while there was significant heterogeneity among populations within each ecotype (P=0.022), largely driven by very low CORT concentrations from one of the M-slow populations. In the analysis of lactate concentration, the handling time to bleed snakes was marginally significant in initial models of (P=0.07), which is not surprising given how quickly activity can induce increases in lactate concentrations in squamate reptiles (e.g. Bennett and Licht, 1972), and so was retained the final model.

Baseline CORT concentrations from normal husbandry conditions differed significantly only from CORT concentrations at 38°C (back-transformed least-squares mean: 44.1 ng ml⁻¹; Tukey–Kramer adjusted difference between baseline and 38°C treatment, P<0.0001; all other comparisons with baseline, P>0.13; Fig. 1B). Additionally, glucose concentration significantly influenced insulin concentration (P=0.0040), and this relationship was independent of temperature (interaction of glucose and temperature: P=0.90; Table 2, Fig. 3).

In this experiment, we measured whole-organism response to acute exposure to ecologically relevant temperatures and compared norms of reaction between two recently diverged lineages adapted to differing thermal environments (Bronikowski, 2000; Bronikowski and Arnold, 1999). Our results revealed a universal temperature dependence of the physiological systems we measured involved in energy processing and allocation. Oxygen consumption rate, as well as CORT, insulin and glucose concentrations, all increased monotonically with increasing temperature, without evidence of plateauing at the highest temperatures. Body condition, as a metric of present energy stores, did not influence any physiological parameter (Fig. 2). We found support for hypothesis 1: a 3-h exposure to temperatures ≥38°C constitutes a physiological stress. Insulin concentration increased in response to hyperglycemia (as indicated by the significant effect of glucose on insulin) and in response to temperature in a glucose-independent manner (as indicated by the shape difference in the reaction norms for insulin and glucose; Fig. 1). Even so, insulin and glucose maintained a positive relationship at all temperatures, suggesting that insulin maintains its role in glucose regulation but perhaps plays additional roles in other functional pathways at high temperatures. Hypothesis 2 was not supported: our results did not demonstrate evidence of local adaptation or canalized developmental plasticity in the thermal reaction norms of these two ecotypes (i.e. no ecotype×temperature interaction) in the measured indicators of physiological response to upper temperature extremes. Finally, our results refute the hypothesis that thermal limits in T. elegans result from oxygen limitation at high temperatures (OCLTT hypothesis; hypothesis 3).

In this experiment, individuals maintained physiological functionality at temperatures just below the lethal limit. Such extreme temperatures, however, dramatically affected regulation of important aspects of energy metabolism, specifically, glucose regulation pathways. CORT concentration was thermally dependent, with temperatures near the upper thermal limit inducing elevation above baseline. Somewhat surprisingly, CORT concentrations at all temperatures were well below maximum levels attained by wild-caught Eagle Lake T. elegans under handling stress (CORT concentration mean±s.e.m.: 338.9±44.7 ng ml⁻¹; Palacios et al., 2012). Though we are unable to make a direct comparison of absolute values across studies, the large difference in maximal values from the two studies as well as the proportional increase from baseline indicate that the capture–restraint protocol of Palacios et al. (2012) induced a much greater CORT response than the temperature treatments of the present experiment. That the CORT response was elevated but not maximal indicates that, even at the highest temperatures, individuals have the capacity to mount a response to additional stressors and that exposure to high temperatures for this time period did not bring snakes across a critical threshold. Even at the highest temperatures, snakes were likely in the ‘reactive homeostasis’ range of Romero et al. (2009) or in ‘type I allostatic overload’ of McEwen and Wingfield (2010). However, prolonged exposure to the highest temperatures that we measured (38°C), exposure to higher temperatures or exposure to additional stressors could potentially continue to elevate CORT concentration beyond the capacity for further response. Furthermore, we found that larger snakes exhibited lower CORT concentrations across the temperature range (Table 1), though post hoc analysis revealed no mass × temperature interaction (F_4,91=0.72, P=0.58). Because size can be used as a proxy for age in garter snakes (Sparkman et al., 2007), this result implies that either individual snakes reduce concentrations of circulating CORT as they age or that larger snakes have been able to obtain larger sizes by virtue of reduced hormonal reactivity. This distinction – whether hormonal profiles change with age or whether certain profiles permit greater growth – has important implications for understanding the role of hormonal regulation of life-history strategies (e.g. Reding et al., 2016).

Stressful stimuli elevate both glucocorticoids, such as CORT, and catecholamines, such as epinephrine and norepinephrine (Akbar et al., 1978; Sapolsky et al., 2000). These hormones trigger pathways to mobilize glucose stores, produce glucose via glycogenolysis and gluconeogenesis (reviewed in Sapolsky et al., 2000), and inhibit glucose uptake into non-essential tissues (Kaplan, 1996). Accordingly, we observed similar patterns in the increases of CORT and glucose concentrations, with both compounds maintaining baseline levels at moderate temperatures and elevating dramatically in response to near-critical temperatures, as demonstrated by the significant quadratic effect of temperature (Table 1, Fig. 1B,D). Under homeostatic circumstances, insulin modulates the glucose-relevant effects of both catecholamines and glucocorticoids by reducing glucose production and promoting uptake into tissues (Foster and McGarry, 1996; Strack et al., 1995). With increasing heat stress, we observed both elevated insulin and glucose concentrations (Table 1, Fig. 1C,D). We also found evidence that insulin was responding to hyperglycemia and that the general relationship between glucose and insulin concentrations did not change across temperatures (Table 2, Fig. 3). Notably, insulin concentration rose linearly across the range of test temperatures, while glucose concentration increased non-linearly, as indicated by the significant quadratic effect of temperature (Table 1, Fig. 1C,D). This result suggests that that insulin may play a role under heat stress in addition to glucose regulation, similar to that in mammals under conditions of heat stress (Baumgard and Rhoads, 2013). For example, insulin may enhance protection from heat stress via its effect on the transcription of heat-shock proteins (Li et al., 2006) and potentially decreased muscle proteolysis (Foster and McGarry, 1996). The temperature dependence of insulin and its role in energy balance maintenance in ectotherms is an important avenue of future research.

Our data demonstrate that garter snakes from these populations of divergent life-history ecotypes adjust their physiology and deliver the required oxygen to their tissues at sufficient rates to maintain metabolic and physiological competence when acutely exposed to near-lethal temperatures. Neither a metabolic-rate plateau at high temperatures nor a temperature-insensitive range was apparent, in contrast to a temperature-insensitive range of oxygen consumption rate from 28 to 34°C reported in a previous study of T. elegans (Seidel and Lindeborg, 1973). Notably, the steep increase in oxygen consumption rate by L-fast snakes at the highest temperature (38°C), as well as the large variance associated with the least-squares mean estimate (Fig. 1A), was driven by two snakes with an extreme response in oxygen consumption rate to high temperatures. The importance of individual variation in temperature sensitivity of these physiological traits in shaping the evolution of the stress response to high temperatures represents another important direction for future research (Angelier and Wingfield, 2013; Cockrem, 2013). Surprisingly, the dependence of oxygen consumption rate on ecotype-specific body mass was not statistically significant. In an analysis of data across all temperatures, oxygen consumption rate scales positively with body mass (in grams) to an exponent of 0.37±0.28 (estimate±s.e.). Though low, this estimate does not differ significantly from estimates for scaling exponents in males of this species (0.59; Bronikowski and Vleck, 2010) or from squamates generally (0.7–0.8; Andrews and Pough, 1985; McNab, 2002). We attribute the large variance around this estimate, rendering it insignificant in our models, to a combination of small sample sizes, a limited range of body masses sampled from each ecotype, and the incomplete V̇_O2 data.

In contrast to the temperature dependence of oxygen consumption rate, we found no evidence that the concentration of lactate, a by-product of anaerobic metabolism, increased with increasing temperature, as might be expected if high temperatures disrupt oxygen supply and induce anaerobic metabolism (Frederich and Pörtner, 2000). We did find a significant quadratic effect of temperature on lactate, however, because of the high lactate concentrations at the low and high experimental temperatures (24 and 38°C). The increases we observed in lactate concentrations at these temperatures are likely not driven by energetic demand outpacing aerobic capacity, given that the relative increases are far below lactate increases resulting from exercise in squamates (e.g. Bennett and Licht, 1972). Rather, this pattern might result from the temperature dependence of lactate dehydrogenase (e.g. Johansson, 1969; Mendiola and De Costa, 1990), though the activity of this enzyme this has not been quantified in T. elegans. Given that neither V̇_O2 nor glucose concentration plateaued at high temperatures and that lactate concentration did not escalate with temperature, these snakes are not limited by oxygen acquisition or transport at temperatures near their thermal maximum. This result is in accordance with recent studies that refute the OCLTT hypothesis in free-living stages of terrestrial ectotherms (Fobian et al., 2014; He et al., 2013; McCue and Santos, 2013; Overgaard et al., 2012; but see Shea et al., 2016). Our data do not address other possible mechanisms that may contribute to thermal tolerance limits, including protein denaturation, membrane fluidity, enzyme–substrate interactions, mitochondrial failure and failure of neural processes (reviewed in Angilletta, 2009; Schulte, 2015). Regardless of the mechanism setting thermal tolerance limits, our results demonstrate that high temperatures increase energetic expenditure, increase resting oxidative metabolism, and affect energy-regulating pathways, which in turn could influence individual activity patterns, response to additional stressors and, ultimately, fitness.

The present study enhances our understanding of the complex interplay of genetic background and developmental environments in shaping physiology in T. elegans from Eagle Lake. First, a previous study found ecotypic differences in metabolic rate in wild-caught adult males after 4 months in captivity (contrasting with the 4 years in captivity in the present study). Male snakes from the L-fast ecotype had higher mass-independent oxygen consumption rates across a range of non-stressful temperatures (15–32°C; Bronikowski and Vleck, 2010). Second, studies in laboratory-born offspring from these two ecotypes indicate no difference in oxygen consumption rates between neonates at 28°C (Robert and Bronikowski, 2010), but identified ecotype-specific thermal response curves in older snakes (Gangloff et al., 2015), suggesting that genetic background exerts developmental effects on resting metabolic rates. Captive-born snakes from both ecotypes have similar CORT thermal response curves, but differ in their responses to high ambient temperatures for both liver mitochondrial H₂O₂ production and erythrocyte superoxide concentration, two measures related to mitochondrial respiration (Schwartz and Bronikowski, 2013). The present study detected no ecotypic differences in the responses of metabolic rate and hormone concentration to increasing temperature in wild-caught, laboratory-acclimated adult female snakes under heat stress, though we did identify an ecotypic difference in glucose concentrations across the range of measured temperatures. Given that the glucose concentration differences were parallel across temperature treatments (i.e. there is no ecotype × temperature interaction), the data suggest that there is variation in the maintenance of available energy between the ecotypes and that this difference is maintained under both normative and stressful conditions. After being caught as adults, these snakes were housed under identical laboratory conditions for several years, which can affect endocrine profiles (Sparkman et al., 2014). Thus, either adult females from these ecotypes do not differ in resting metabolic rates or environmental effects on thermal reaction norms remain plastic throughout adulthood (Angilletta, 2009). Although previous studies in garter snakes have showed that stress response and endocrine function are affected by current energetic stores (Dayger et al., 2013; Palacios et al., 2013), we found that body condition, which was largely homogeneous among individuals in our study, did not impact physiological variables. These animals, maintained in favorable conditions with ample food stores for 4 years, may not represent enough variation in body condition to reveal trade-offs as expected under resource-limited conditions (Glazier, 1999; van Noordwijk and de Jong, 1986).

Snakes from these locally adapted ecotypes did not differ in response to stressfully high ambient temperatures, despite being adapted to differing thermal environments in life-history traits (Bronikowski, 2000). This fits observations that upper thermal tolerance limits in terrestrial ectotherms display limited variance, and thus may have little capacity to adapt to changing local conditions (Huey et al., 2012; Sunday et al., 2011). That upper limits are constrained, however, does not mean that all thermal tolerances are fixed: in studies of other snakes in this genus, the critical thermal minima vary among and within species and are tied to temperatures experienced by populations in the field (Burghardt and Schwartz, 1999; Doughty, 1994). Garter snakes behaviorally avoid thermal extremes (Scott et al., 1982), which reduces selection on thermal physiology at high temperatures (Buckley et al., 2015; Huey et al., 2003). Although behavioral thermoregulation may provide relief from extreme high temperatures, increasing environmental temperatures could still reduce the number of suitable retreat sites (Huey et al., 1989), increase the distance between retreat sites and increase the resting metabolic rates of inactive snakes. Nonetheless, it is possible that snakes might benefit from the increased foraging capacity and energy input of warmer temperatures (Huang et al., 2013; Weatherhead et al., 2012), especially in habitats that are less restricted by high temperatures (i.e. meadow populations). In general, studies that quantify long-term energy budgets at increased ambient temperatures and test for temperature-induced mismatches between metabolic rate and consumption (e.g. Iles, 2014; Lemoine and Burkepile, 2012) are needed to assess the possible effects on individual fitness and population persistence. By utilizing an integrative, mechanistic approach in a broader energetic and ecological context, we can further develop our understanding of the processes that set thermal limits and the consequences of exposure to extreme but sub-lethal temperatures.

We thank S. Lei, W. Richardson at MP Biomedical, and B. Joos at Sable Systems for advice on laboratory techniques; G. Palacios, M. Manes and T. Schwartz for fieldwork; C. Corwin and A. Pottebaum for experimental assistance; L. Follett for statistical advice; and G. Burghardt, B. Danielson and F. Janzen for manuscript feedback.