Over a decade of field physiology reveals life-history specific strategies to drought in garter snakes (Thamnophis elegans)

Abstract

Changing climates and severe weather events can affect population viability. Individuals need to buffer such negative fitness consequences through physiological plasticity. Whether certain life-history strategies are more conducive to surviving changing climates is unknown, but theory predicts that strategies prioritizing maintenance and survival over current reproduction should be better able to withstand such change. We tested this hypothesis in a meta-population of garter snakes having naturally occurring variation in life-history strategies. We tested whether slow pace-of-life (POL) animals, that prioritize survival over reproduction, are more resilient than fast POL animals as measured by several physiological biomarkers. From 2006 to 2019, which included two multi-year droughts, baseline and stress-induced reactivity of plasma corticosterone and glucose varied annually with directionalities consistent with life-history theory. Slow POL animals exhibited higher baseline corticosterone and lower baseline glucose, relative to fast POL animals. These patterns were also observed in stress-induced measures; thus, reactivity was equivalent between ecotypes. However, in drought years, measures of corticosterone did not differ between different life histories. Immune cell distribution showed annual variation independent of drought or life history. These persistent physiological patterns form a backdrop to several extirpations of fast POL populations, suggesting a limited physiological toolkit to surviving periods of extreme drought.

1. Background

Physiological processes both permit and limit the correlated evolution of life-history traits. Thus, two decades ago, Ricklefs & Wikelski [1] argued that quantifying sources of variation in physiological traits is essential to understanding patterns and trade-offs observed in the evolution of life histories. Their ‘physiology/life-history nexus' synthesis hypothesizes a prominent role for physiology in shaping pace-of-life (POL) strategies along a fast–slow continuum (sensu [2]). Resource allocation to competing life-history traits (i.e. growth, reproduction, self-maintenance, survival) is governed by physiological processes. Thus, populations experiencing variation in resource quality across generations should diverge in life-history strategies provided that both genetic variation and permissive physiological processes exist for such evolution to occur [3–5]. Fast–slow paced strategies should be accompanied by either concomitant divergence—or sufficient plasticity—in physiological traits (e.g. metabolic, energetic, endocrine, immune) owing to environmental drivers and subsequent resource allocation strategies [6,7]. Within generations, physiological plasticity, often reliant on the endocrine system, mediates how individuals respond to their environment within their lifespans [8–11].

Glucocorticoid hormones such as corticosterone function as primary mediators of energy metabolism, acting to both mobilize energy stores when needed and maintain longer-term homeostasis [12–13]. Thus, glucocorticoids often fluctuate in parallel with other biomarkers of energy expenditure, such as glucose, and act as indicators of the reactivity or sensitivity of an individual to a given environmental context, including response to stress [14–17], reviewed in [18]. Additionally, circulating glucocorticoids influence the distribution of leukocytes, with increased glucocorticoid levels resulting in the movement of lymphocytes into peripheral tissues and heterophils into circulation (i.e. increasing the heterophil-to-lymphocyte ratio, ‘HL’ hereafter; [19–21]). Increased HL in response to stress is a conserved response across vertebrates; heterophils proliferate in circulation and act as primary phagocytes, while redistributed lymphocytes function in immunoglobulin production and modulate immune defence (reviewed in [21]). In general, slow POL animals are predicted to have elevated and less variable levels of circulating glucocorticoid hormones [1,17,22]. Also, because glucocorticoids facilitate the movement of lymphocytes into tissues and heterophils into circulation, higher glucocorticoids predict higher HL [21,23], but see [24]. Moreover, slow-paced animals, with lower energy demands supporting slower growth and reproduction, are predicted to have lower circulating glucose concentrations than their fast-paced counterparts [25]. Thus, considering measures of multiple biomarkers can provide a more complete picture of an organism's response to its environment—providing information about the activity of the hypothalamic-pituitary-adrenal axis (hypothalamic-pituitary-interrenal [HPI] axis in reptiles) and its downstream effects on energy mobilization and distribution of immune cells (table 1).

Table 1.

Summary of predictions. Pace-of-life strategy (slow- versus fast-living) predictions for measures of baseline corticosterone, glucose and heterophil-to-lymphocyte ratios (HL) and stress-induced reactivity (Δ) of corticosterone and glucose for years classified as either average precipitation or severe drought. (Result noted below prediction: indicates that prediction was supported. indicates that prediction was not supported.)

	corticosterone	glucose	HL	citations
baseline
average precipitation	slow > fast	slow < fast	slow > fast	Ricklefs & Wikelski [1]
				Hau et al. [22]
				Dupoué et al. [26]
				Tomasek et al. [25]
drought	slow = fast	slow < fast	slow = fast	Davis et al. [21]
stress-induced reactivity (Δ)
average precipitation	slow > fast	slow = fast	—
drought	slow = fast	slow < fast	—

In response to extreme events, physiological parameters may deviate from their functional range as a means of maintaining organismal homeostasis [8–10]. In the short-term, this reactivity may increase an individual's chances of survival through increasing or enhancing physiological function and behaviour while temporarily shunting resources away from other, non-vital, processes (e.g. feeding, digestion, growth, reproduction, immunity [9,17,18]). However, if sustained, this increased allocation to immediate needs can reduce an individual's fitness, induce pathology, or lead to population-level decline [8,18,27]. Thus, the magnitude of increase in circulating biomarkers in response to a given perturbation reveals the reactivity or sensitivity of an individual, and by extension a population. For long-lived species, long-term studies are needed to test for associations between the fast-slow life-history continuum and physiology (e.g. [28–30]). Such longitudinal studies provide across-generation data that encompass patterns of variation in resource availability and climate. By quantifying physiological biomarkers across these time scales, we can quantify the stability of the physiology/life-history nexus, and subsequently, answer how responsivity/plasticity can buffer populations across generations—i.e. whether the structure of the nexus lends resiliency or susceptibility to population extirpation [31–33].

Our study focuses on populations of western terrestrial garter snakes (Thamnophis elegans) exhibiting slow and fast-paced life-history strategies, which have been the subject of over 45 years of research on ecology, behaviour and evolution (e.g. [34–36]). Notably, these populations have experienced several droughts over that period. This affords us the opportunity to address how physiology, climate, and life-history interact. We do so by quantifying baseline and stress-induced levels of physiological biomarkers (corticosterone, glucose, HL) and test for differences in the baseline and stress-induced reactivity of these biomarkers over more than a decade. We test predictions in table 1 and extend these predictions to encompass drought conditions (table 1). Specifically, we test whether the association between life history and physiology has been impacted by historic droughts that have occurred in this region over the past decade, and whether certain combinations of life-history and physiology confer resiliency to population extirpation (e.g. [26]).

2. Material and methods

(a) . Study system

We conducted research on populations of garter snakes in the vicinity of Eagle Lake, Lassen County, CA, USA. Populations either inhabit areas of rocky shoreline around the lake or higher elevation mountain meadows. Meadow habitats have cooler air temperatures, lower prey and water availability, and lower predation rates relative to lakeshore sites [35,37,38]. Lakeshore and meadow snakes exhibit contrasting life-history strategies, characterized as either fast- or slow-POL ecotypes, respectively (L-fast and M-slow). M-slow snakes grow slower to smaller adult size, mature later, reproduce less frequently, and live longer relative to L-fast snakes (summarized in [39], see also [40]). Even so, heterogeneity exists among populations of the same ecotype in physiological traits (e.g. metabolism [41]; immune function [42]) and can vary with annual precipitation patterns [43,44]. For this study, we focused on three replicate L-fast and four replicate M-slow populations (electronic supplementary material, figure S1).

From 2006 to 2019, California experienced two severe multi-year droughts (2007–2009, 2012–2015) defined by a Palmer's hydrological drought index PHDI, [45] of less than or equal to −2, owing to decreased winter snowfall and snowpack, decreased spring precipitation, and, during the latter drought, record high temperatures [46–48]; electronic supplementary material, figure S2. Unlike the drought of 2007–2009, and earlier droughts (e.g. 1980s), the most recent drought (2012–2015) not only affected water and prey availability in meadow locations but also altered the habitat structure along the lakeshore, which impacts retreat site availability for L-fast animals. Decreased lake levels created a 2–3 m gap between the water and rocky shoreline (Lassen County Public Works), devoid of retreat sites, which impacted foraging ability and thermoregulatory options for lakeshore animals [49]. Several L-fast populations declined or were extirpated from 2010 to 2015 (electronic supplementary material, figure S1), probably through effects on adult survival [43,50].

(b) . Field work

Garter snakes in this region of California emerge from hibernacula in late spring; forage and mate over the summer (June–August), decrease activity in autumn (September–October); and return to hibernacula in the autumn for overwintering—a pattern which generally follows seasonal precipitation and temperature (electronic supplementary material, figure S3). Free-ranging garter snakes were hand-captured in grasses or under rocks in M-slow and L-fast populations from late May through to early July. Beginning in 2006, we collected a blood sample from the caudal vein with a heparin rinsed syringe immediately upon capture to assess baseline measures of corticosterone (2006–2010, [17] 2010–2019, this study). We used a drop of fresh whole blood to prepare blood smears (2007–2019) and, beginning in 2013, to measure glucose (2013–2019; see below).

In 2010 and from 2013 to 2019, we collected an additional blood sample at 3 h post-capture to assess reactivity (the difference between an individual's baseline and stress-induced maximal measures) of corticosterone and glucose, following established restraint protocols (snakes were held in individual cloth bags for the duration; [15]). All blood samples were kept on ice until centrifugation where we separated packed red blood cells from plasma. Blood components were flash-frozen in liquid nitrogen, shipped to Iowa State University on dry ice, and stored at −80°C until time of assays. After blood collection, we sexed, weighed (g), and measured (snout-to-vent (SVL), mm) all snakes; females were palpated to determine number of developing embryos. We excluded immature snakes in the current study ([43], i.e. less than 300 mm SVL or low body weight (L-fast less than 20 g; M-slow less than 15 g)). After processing, we released snakes at their point of capture.

Based on the corticosterone handling-time response curves from 2010 and 2013 ([17] and electronic supplementary material, figure S4), peak corticosterone and glucose levels in this species occurs at 3 h post-capture, while no response was seen in HL over 3 days (electronic supplementary material, table S1). Thus, for the present study, we have baseline measures for corticosterone (n = 2557), glucose (n = 2 586) and HL (n = 1 073), and measures of stress-induced reactivity for corticosterone (n = 1481) and glucose (n = 1269) (See the electronic supplementary material, table S2 for sample sizes by year and ecotype).

(c) . Corticosterone, glucose and heterophil-to-lymphocyte ratio measurements

We measured plasma corticosterone concentration using a double-antibody radioimmunoassay (ImmuChem Double Antibody Corticosterone I-125 RIA kit, MP Biomedicals, Irvine, CA, USA). We conducted the assay following previously described protocols adapted for use with garter snakes [17]. We measured all baseline plasma samples at either a 1 : 40 dilution (n = 1995) or a 1 : 80 dilution (n = 562) to fit within the bounds of the standard curve. We measured all stress-induced samples at 1 : 160 dilution. We ran a subset of samples (n = 20) at a dilution factor of 1 : 40, 1 : 80 and 1 : 160 to test for an effect of dilution. We found an average coefficient of variation (CV) across these three dilutions of 20.5%, with better agreement between 40-fold and 80-fold dilutions (CV 14.4%). Final concentrations from 160-fold dilutions tended to be lower than 40-fold and 80-fold, thus the stress-induced samples represent a conservative estimate of corticosterone concentration. All runs included from 1 to 3 pooled samples (inter-assay CV 13.8%). We re-ran sample duplicates with an intra-assay CV greater than 10% and any samples outside the standard curve (average intra-assay CV for all final samples 4.15%). Most samples from 2010 (published in [17]; figure 1a) were rerun in this study with pooled samples to allow for across-year comparisons of corticosterone concentrations. As circulating glucocorticoids increase after 10 min of handling [17], we used this time as a cut-off for inclusion as a baseline measure.

Figure 1.

Annual baseline plasma corticosterone levels (a) and stress-induced corticosterone reactivity (b). Panel (a) includes data from 2006 to 2010 redrawn from Palacios et al. [17,40] samples remeasured in the present study (see the electronic supplementary material, methods). Asterisks denote years in which M-slow populations exhibited higher values. The slopes in (b) represent mean annual stress-induced reactivity and illustrate the two groupings of variation in corticosterone reactivity. Years classified by severe drought underlined. Data are back-transformed least squares means ± s.e. from statistical models in table 2. (Online version in colour.)

We measured the concentration of glucose at the time of blood collection using a Freestyle Lite Glucometer (Abbott Diabetes Care, Alameda, CA) from a drop of fresh whole blood (for method validation see [51]). The lower limit of this glucometer is 20 mg dl⁻¹; lower concentrations return a value of ‘low’. We assigned a value of 19 mg dl⁻¹ to ‘low’ measures (n = 630).

We measured HL from blood smears prepared from freshly drawn blood. Smears were fixed in methanol and stained with Wright-Giemsa stain. Relative abundances of leukocytes (lymphocytes, heterophils, monocytes and basophils) were estimated by classifying the first 100 leukocytes encountered with a compound microscope under oil submersion with the 100× objective (i.e. a total of 1000× magnification; described in detail in [40]).

(d) . Statistical methods

For analyses of physiology, we analysed baseline measures of corticosterone, glucose and HL, and the stress-induced increase (hereafter ‘reactivity’, depicted as Δ) of corticosterone and glucose. All analyses were conducted in SAS (v. 9.4, SAS Institute, Carey, NC, USA) using Proc MIXED for repeated-measures mixed-effect linear models. We assessed significance of fixed effects using type III sums of squares and estimated denominator degrees of freedom for F-tests using the Kenward-Roger degrees of freedom approximation [52]. Figures were made in the programming language R [53] with package ggplot2 [54]. K-means clustering analyses were performed on least squares means estimates from mixed-effect models to assess the number of groupings into which measures of mean annual physiological reactivity are partitioned using the R package NbClust [55]. All physiological measures (corticosterone, glucose, HL) were log₁₀-transformed to meet assumptions of normality of model residuals [56].

Models included the fixed effects of ‘sex’ (mature snakes of three levels: male (M), female (F) and pregnant female (G)); ‘ecotype’ (two POL levels: L-fast and M-slow); ‘year,’ modelled as a categorical variable to allow for post hoc linear contrasts (corticosterone, nine levels: 2010–2019 excluding 2011, earlier years published previously [17]; glucose, seven levels: 2013–2019; HL, 10 levels: 2007–2019, excluding 2009, 2011 and 2012); and ‘population-nested-within-ecotype’ (L-fast, three levels: L2, L4, L8; M-slow, four levels: M1, M2, M3, M5). Covariates included day-of-year to account for variation across the sampling season (ordinal day-of-year where our earliest sampling date 17 May = day 138, and last date 12 July = day 193); and body size ‘zSVL’ which is Z-transformed SVL to mean of 0 with unit variance within each sex-by-ecotype group because size varied across sex-by-ecotype categories (SVL range L-fast (mm): M = 305–655, F = 315–724, G = 456–780; range M-slow (mm): M = 307–591, F = 316–627, G = 351–619). This size standardization allows us to assess the variation among relatively small and relatively large animals within each sex-by-ecotype group. We also included the repeated measures random effect of capture incidence with individual nested in within-year recaptures (five levels: 1, 2, 3, 4, 5) and among-year recaptures (four levels: 1, 2, 3, 4) to account for the non-independence of repeated samples collected from the same individual over seasons (up to five times) and years (up to four times). Thus, our models took the following form where Y is corticosterone, glucose, HL, Δcorticosterone or Δglucose; μ represents the grand mean; and ε is the error term:

$$\begin{array}{rl}Y\sim \mu & +\mathrm{zSVL}+\mathrm{day}+\mathrm{sex}+\mathrm{year}\\ & +\mathrm{ecotype}+(\mathrm{year}\hspace{0.17em}\times \hspace{0.17em}\mathrm{ecotype})+\mathrm{population}\hspace{0.17em}(\mathrm{ecotype})\\ & +|\mathrm{ID}\hspace{0.17em}(\mathrm{within}\hspace{0.17em}\mathrm{year}\hspace{0.17em}(\mathrm{among}\hspace{0.17em}\mathrm{year}))+\epsilon .\end{array}$$

Additional models were considered. To assess whether aspects of climate predicted variation in physiology, we tested effects of precipitation and temperature. We tested whether body condition and, for pregnant females, whether the number of developing embryos was a better predictor of physiological trends than size (zSVL). Additionally because we have witnessed extirpation of populations during the 45 years of study and aspects of physiology can differ among populations of the same ecotype [41,44], we also ran models to assess variation in measures of physiology substituting ‘population’ for ‘ecotype’ as a fixed effect. None of these approaches provided additional insight into physiological variation beyond those gained from post-hoc contrasts among years and therefore we do not consider these models further (see the electronic supplementary material, for additional information).

3. Results

(a) . Annual fast–slow pace-of-life-associated variation in physiology

We found significant heterogeneity in baseline measures of corticosterone and glucose across years and between ecotypes within years (table 2, figures 1a and 2a, for corticosterone and glucose, respectively). Within a given year when baseline corticosterone differed between ecotypes, M-slow animals always had higher concentrations than L-fast animals (figure 1a). Interestingly, this pattern was observed in all years with normal precipitation (electronic supplementary material, figure S2), but this pattern was largely absent in drought years. Our primary motivation was to assess annual variation, as we thought drought duration may affect physiology as much as broad categorization of drought/non-drought. The physiological effects of one ‘bad’ year may be negligible, but several consecutive drought years could result in dramatic physiological responses or consequences. Nonetheless, least-squares means comparisons of ecotype from analyses categorizing years as drought versus non-drought supported results from models including the effect of year (electronic supplementary material, table S3 and figure S5). Additionally, we found that L-fast snakes had more variable baseline measures of corticosterone when compared with M-slow snakes (F_649,1906 = 1.19 p = 0.004). By contrast, baseline glucose was consistently higher in L-fast animals than in M-slow animals, and therefore showed no pattern with drought severity (figure 2a). HL also varied across years, but with no effect of ecotype or drought (table 2 and figure 3).

Table 2.

Repeated-measures mixed linear model analysis of physiology in wild-caught western terrestrial garter snakes (Thamnophis elegans). (Physiological measures of baseline plasma corticosterone (CORT), blood glucose (GLUC) and the ratio of circulating heterophils-to-lymphocytes (HL), as well as measures of physiological reactivity, or the change between baseline and stress-induced measures of physiology (ΔCORT and ΔGLUC), are log₁₀-transformed.)

	CORT	GLUC	HL	ΔCORT	ΔGLUC
zSVL
Fd.f.n,d.f.d	23.261,2135	19.611,2180	0.651,038	1.971,1222	10.541,1133
Pr > F	<0.0001	<0.0001	0.421	0.161	0.001
ordinal day
Fd.f.n,d.f.d	4.451,2495	71.531,2498	16.441,1012	1.541,1453	32.081,1222
Pr > F	0.035	<0.0001	<0.0001	0.215	<0.0001
sex category
Fd.f.n,d.f.d	7.472,2188	15.322,2245	2.702,1022	4.252,1323	7.122,1162
Pr > F	0.001	<0.0001	0.068	0.015	0.001
year
Fd.f.n,d.f.d	11.938,2506	6.626,2537	7.269,195	6.707,1447	2.155,1233
Pr > F	<0.0001	<0.0001	<0.0001	<0.0001	0.057
ecotype
Fd.f.n,d.f.d	25.661,2446	43.621,2462	0.491,171	2.821,1430	0.221,1199
Pr > F	<0.0001	<0.0001	0.485	0.093	0.642
ecotype × year
Fd.f.n,d.f.d	2.797,2505	3.336,2539	0.749,171	0.846,1444	1.995,1237
Pr > F	0.007	0.003	0.675	0.543	0.078
population (ecotype)
Fd.f.n,d.f.d	11.125,1921	3.735,1923	1.065,424	5.225,1187	13.645,1036
Pr > F	<0.0001	0.002	0.379	<0.0001	<0.0001

Figure 2.

Annual baseline blood glucose levels (a) and stress-induced glucose reactivity (b). Asterisks in (a) denote years in which L-fast populations exhibited higher values. The slopes in (b) are included to visualize the lack of annual variation in glucose reactivity; for all years, baseline and stress-induced measures were elevated in L-fast animals. Years classified by severe drought underlined. Data are back-transformed least-squares means ± s.e. from statistical models in table 2. (Online version in colour.)

Figure 3.

Annual baseline circulating heterophil-to-lymphocyte ratio (HL) levels in free-ranging western terrestrial garter snakes (Thamnophis elegans). Years classified by severe drought underlined. Data are back-transformed annual least-squares means ± s.e. (ecotypes combined) from models in table 2. (Online version in colour.)

We found significant heterogeneity in corticosterone reactivity, such that the relative change in corticosterone between baseline and stress-induced maximal measures differed among years irrespective of ecotype (table 2 and figure 1b; electronic supplementary material, figure S6). Based on post hoc analyses, there were two clusters of slopes—years with greater reactivity (2013, 2014, 2019 both ecotypes, and 2010 for M-slow) and years with lower reactivity (2015, 2016, 2017, 2018 both ecotypes, and 2010 for L-fast) with no association with drought. Both baseline and stress-induced maximal glucose were greater in L-fast individuals. The relative change between baseline and stress-induced measures did not differ between ecotypes or across years (table 2 and figure 2b; electronic supplementary material, figure S6). Of note, the maximum value of stress-induced corticosterone was 867 ng ml⁻¹ in an M-slow snake; and the maximum value of stress-induced glucose was 212 mg dl⁻¹ in an L-fast snake.

(b) . Effects of covariates on physiology

Measures of baseline corticosterone increased with increasing body size (zSVL) in M-slow animals but was unaffected by size in L-fast animals. Baseline glucose decreased with increasing body size in both life-history ecotypes. HL was unaffected by body size. Corticosterone decreased across the sampling period (17 May–12 July), whereas glucose and HL increased with advancing day in the season. Interestingly, the analyses of each dependent variable demonstrate patterns of variation among sex categories. Males (M) and pregnant females (G) had higher baseline corticosterone than females (F) (i.e. (M, G) > F), and lower levels of baseline glucose (i.e. (M, G) < F). Whereas for HL, males had higher HL than pregnant females (i.e. (M, F) > (F, G)). When assessing the effect of sex categories within years designated as drought or non-drought, we find that non-pregnant females from M-slow populations exhibited a greater decrease in baseline corticosterone than any other ecotype-by-sex combination in drought years.

4. Discussion

Our goals in this study were to characterize the relationship between physiology and life-history, and to test for the stability of this relationship over time and in the context of drought and non-drought years. We found significant annual variation in both baseline levels and the stress-induced reactivity of corticosterone and glucose in a pattern consistent with POL strategies and predicted by life-history theory (table 1). Importantly, however, these differences were observed primarily in resource-abundant (non-drought) years, suggesting that plasticity in physiological traits maintains functionality similarly between these contrasting life-history ecotypes during periods of resource limitation. Furthermore, we found significant annual variation in HL but with no support for POL specific ecotype differences. We discuss our findings in the context of POL trade-offs and population sensitivity to climate and explore the ramifications for population persistence.

Life-history theory is based on the premise of trade-offs, positing that organisms must make energetic decisions to allocate resources between survival and reproduction in response to variation in their environment [57,58]. Highly variable environments are predicted to drive variation in reproductive success, through the availability of resources, and ultimately select for slow POL [59]. However, relatively few empirical studies have examined the effects of habitat on intraspecific life-history variation (but see, [60,61]). Additionally, with increasing environmental variation, the probability of population extinction is greatly reduced in species that exhibit survival-fecundity trade-offs, compared to species that exhibit either positive covariation between survival and fecundity or no covariation [62]. In this system, snakes from the more annually variable meadow habitats fall on the slow-living end of the POL continuum with slower rates of growth and low fecundity, but relatively high adult survival [35,43].

Physiology is the link mediating life-history traits (the physiology/life-history nexus; [1]), and specifically, the relationship between survival and reproductive trade-offs [63,64]. In our study, we found that slow-POL strategy populations (M-slow) have relatively high circulating levels of baseline corticosterone, high HL, and relatively low circulating glucose as compared to their fast-POL counterparts (L-fast). Interestingly, these patterns were only apparent in non-drought years. We further found that stress-induced reactivity (the amount of change between baseline and stress-induced measures) did not differ between the ecotypes, regardless of whether it was a drought or non-drought year.

In this garter snake metapopulation, environmental variability has produced divergent ecotypes on the slow-fast POL continuum, with glucocorticoids playing a role in mediating life-history strategies. Specifically, slow-POL populations exhibit reduced among-year variability in corticosterone relative to fast-POL populations, suggesting the former may not have the capacity to accommodate as wide a range of extrinsic conditions (i.e. a narrower allostatic range or reactive scope; reviewed in [6]; figure 1). Slow-POL populations, on average, also exhibit higher corticosterone levels relative to their fast-POL counterparts (table 2 and figure 1). This finding bolsters previous work documenting the role of glucocorticoids in shaping ecotypic differentiation in this system [17]. In years where ecotypes differed, L-fast populations exhibited higher baseline and stress-induced measures of glucose than M-slow populations. This is probably owing to more stable resources, as L-fast populations are characterized by having access to continuous food and water, while resources for M-slow populations are more variable and depend on annual precipitation [35,43]. Previous work in this system demonstrates that food availability across generations has been a potent agent of selection for fast versus slow growth [35], such that L-fast populations, which have more consistent access to food, grow more quickly, attain larger body sizes, and in turn exhibit higher levels of circulating glucose. Additionally, L-fast populations overall were more variable in their baseline measures of corticosterone. This pattern of increased baseline glucose in L-fast populations may also give context to the greater variability in circulating levels of corticosterone. If the HPI axis functions to maintain homeostasis through regulation of energy substrates (i.e. glucose), we expect more variability in circulating levels of corticosterone to accommodate resource fluctuation and maintain available glucose within a narrow range.

To understand impacts of stress physiology on population dynamics in the face of changing climates, longitudinal studies of physiology and demography are needed. While many acknowledge the importance of long-term physiological studies on natural populations to understand how changing climates and habitats affect survival [65], empirical studies remain rare. Long-term studies have typically focused on population range shifts and decline or extinction (e.g. [66–70] or changes in phenology (e.g. [33,71–73]) and morphology (e.g. [67]). Far fewer have quantified behavioural (but see, [74–76]) or physiological plasticity (but see, [28–30]). Parallels can be drawn between our study and those that have measured biomarkers in wild populations over multiple years (e.g. [28]) in that there is significant annual variation within and among populations across years, suggesting physiological plasticity owing to variation in environmental conditions. This region of California has experienced several multi-year droughts since the study of these populations of garter snakes began (1987–1989, 2001–2003, 2007–2009, 2013–2015), with 2013–2015 exhibiting the driest and hottest years on record [46,77,78]. Extreme drought has been linked to reductions in body condition across populations of California newts (Taricha torosa; [67]) and rapid species decline in anurans across much of the Sierra Nevada mountains [79,80]. Additionally, Prugh et al. [81] quantified responses of sympatric plants, arthropods, reptiles, and mammals to the most recent drought in the Carrizo Plain (central valley of California), finding that more than 25% of species experienced appreciable declines in abundance. Given these dramatic drought responses, we were additionally interested in the effect of drought on life-history specific patterns of physiology. We found that during years characterized by extreme drought life-history specific patterns for measures of corticosterone were absent. Additionally, we found no evidence that specific aspects of the local climate (temperature, precipitation) predicted physiology (see the electronic supplementary material, for details). Together these findings suggest that the indirect effects of climate on food and water availability are mediating stress physiology. In our study, structural changes to habitat and impacts on resources owing to drought drive variation in physiology.

Physiological reactivity in this system probably reflects adaptation to habitats characterized by seasonal variation and repeated exposure to extreme climatic events (i.e. droughts) rather than a stress response per se. Notably, during droughts in the 1980s and 2000s, meadow habitats experienced pronounced local extirpation. For snakes inhabiting meadows, droughts limit availability of food and water resources [35,38]. The most recent drought, however, has been accompanied by record declines in lake levels: minimum lake levels in 2015 fell to 1551.4 m above sea level, nearly 4 m lower than levels during the droughts of the 1980s and early 2000s (1555.1 and 1555.6 m, respectively, per Lassen County Public Works). This dramatic reduction in the water line left exposed shoreline with no vegetation or retreat rocks. This exposed stretch between suitable habitat and the lake could render snakes at higher risk of predation as they traverse the shoreline to their food source [37]. Altered resource availability, such as microhabitat structure and access to retreat sites [49], can leave individuals more vulnerable to their local climate, more movement-challenged [82], or more exposed to predators and ultimately lead to decreases in population density [66,68]. At present, no snakes reside in historically large L-fast populations (L1, L3, L5 and L6; electronic supplementary material, figure S1) where vegetation and retreat sites are now absent, whereas demographically healthy L-fast populations reside in locations with appropriate habitat structure—notably, the most robust lakeshore population exists in anthropogenically created habitat (i.e. population L2 which inhabit the jetty protecting the local marina). It appears that habitat structure, rather than food and water per se, limits L-fast snakes. It is doubtful that dispersal of snakes from meadow sites to lakeshore sites will result in recolonization given that stream corridors remain dry [41].

Overall, our results reveal considerable inter- and intra-annual variation in snake physiology, with slow- and fast-POL populations exhibiting variation in biomarkers predicted by life-history theory. Notably, within-year POL-specific variation in measures of corticosterone was only evident in years experiencing normal precipitation. Still, populations have disappeared during the last decade, probably in response to structural habitat change resulting from severe drought events. Understanding how physiology, demography and genetic structure interact to predict population persistence is therefore an ongoing pursuit of import.

Ethics

Fieldwork was conducted under permits administered by the California Department of Fish and Wildlife (Scientific collection DFW1397b; SC-007553), and all procedures were approved by the Iowa State University Institutional Animal Care and Use Committee (protocol no. 3–2-5125-J).

Data accessibility

Data from this study are available on DataShare: the Open Data Repository of Iowa State University (https://doi.org/10.25380/iastate.16725439).

Authors' contributions

K.G.H.: conceptualization, data curation, formal analysis, investigation, writing—original draft; E.J.G.: investigation, writing—review and editing; D.A.W.M.: conceptualization, funding acquisition, investigation, supervision, writing—review and editing; A.R.H.: investigation, writing—review and editing; C.D.: investigation, writing—review and editing; A.B.: investigation, writing—review and editing; G.K.: investigation, writing—review and editing; A.M.B.: conceptualization, funding acquisition, investigation, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Competing interests

We declare we have no competing interests.

Funding

This research was supported by grants from the National Science Foundation (IOS-1558071, IOS-0922528 and DEB-0323379).