Corticosterone and immune responses to dehydration in squamate reptiles

The survival of organisms depends in part on their adaptability to predictable changes in their environment, with alterations made to many aspects of physiology and behavior, such as immune function (Lazzaro et al., 2008), life history (Rogell et al., 2014) and foraging strategies (Manenti et al., 2013). However, evolutionary constraints may limit such adaptations or the speed at which adaptations can occur.

Organisms that inhabit environments that have periods of water limitation have various adaptations that enable them to survive. Some animals survive by maintaining stable plasma osmolality despite desiccating conditions (Bradley, 2009), while others tolerate dehydration (e.g. increased plasma osmolality and/or reduced total body water) in their bodies or eggs (Crowe et al., 1992; Peterson, 1996; Davis and DeNardo, 2009). Even if an animal can survive dehydration to some extent, this condition can affect physiological functions (Churchill and Storey, 1994; Giuliano et al., 1995; Plummer et al., 2003). Of particular interest are (1) the influence of dehydration on stress responsiveness and immune function, as these greatly influence an animal's defense mechanisms, and (2) whether effects vary with habitat type and the likelihood that the animal naturally experiences periodic bouts of dehydration.

Glucocorticoid hormones, the end-products of the hypothalamic-pituitary–adrenal (HPA) axis, are essential to homeostatic maintenance during periods with and without stressful stimuli, and play possible long-term roles in recovering from stressors (Romero and Beattie, 2022). Though physiological stressors may cause an increase in an organism's glucocorticoid levels, this trend does not always hold true, especially in wild animals (Dickens and Romero, 2013). Regardless, assessing changes in glucocorticoid levels still offers insight into homeostatic challenges within an animal. If dehydration presents a challenge to homeostasis, we might expect an increase in glucocorticoids; however, if animals are well adapted to deal with the physiological effects of dehydration, responses may instead be decoupled from any glucocorticoid response.

Dehydration has been linked to increased plasma glucocorticoids (Tsuchida et al., 2004; Maresh et al., 2006; Barsotti et al., 2019), but much of our knowledge regarding this relationship is based on domesticated animals that have not undergone selection to tolerate water limitations [mice (Tsuchida et al., 2004); rats (Sebaai et al., 2002); rabbits (Kallaras et al., 2004); cattle (Parker et al., 2004); chicken (Klandorf et al., 1984); pigeons (De and Ghosh, 1993); and quail (Cain and Lien, 1985)]. Given that a stress response should be an adaptive, beneficial response to challenging situations (see Boonstra, 2005), the scope of an individual's stress response likely depends on the environment to which it is adapted (Wingfield et al., 1995) or in which it has developed (Ellis et al., 2006).

For many animals that live in seasonally water-limited environments and rely primarily on free-standing water (versus dietary or metabolic water), plasma osmolality increases for extended periods during the dry season (Peterson, 1996; Davis and DeNardo, 2009; Reynolds et al., 2011). Increased plasma osmolality could negatively affect fitness if it causes glucocorticoid release over extended periods, possibly resulting in immunosuppression (Sapolsky et al., 2000; Dhabhar, 2014) and increased water loss (Parker et al., 2003; Jessop et al., 2013). Such a response could be maladaptive; thus, desert-dwelling animals would likely benefit from avoiding an increase in circulating glucocorticoids. Therefore, it is likely that dehydration does not ubiquitously induce a glucocorticoid increase, and glucocorticoids likely have varied downstream effects on immune function (Moeller et al., 2017). Further, these effects of dehydration may even vary among related species, as species-specific dehydration tolerance adaptations may be greatly influenced by the local environment (Stahlschmidt et al., 2011).

In the last several decades, the field of ecological immunology has grown, with a primary focus on the relationship between the environment and an organism's immune function (‘ecoimmunology’; Sheldon and Verhulst, 1996). However, the extent to which ecology and phylogeny influence or limit adaptations of the immune system has not been well studied. In many cases, the immune system is assumed to adapt in response to specific situations or conditions. For example, pathogen exposure can greatly influence immune development and function (Mazmanian et al., 2005; Minias, 2019), and this is often aligned with habitat type (e.g. the Pathogen Exposure Hypothesis; Scharsack et al., 2007). Parasite and microbe diversity are expected to be higher in mesic and freshwater environments, potentially creating more immune challenges for inhabitants of these areas and often resulting in higher parasite loads, changes in cell proliferation, and other immune activities (Horrocks et al., 2015). Studies on the effect of habitat on immune function often look at different populations of a single species across a range of habitats (Scharsack et al., 2007; Krynak et al., 2016) or at related species with varied life histories or ecologies (Mendes et al., 2005; Schneeberger et al., 2013; Horrocks et al., 2015; Heinrich et al., 2016; Minias, 2019). However, very few studies address how immune function adaptations are affected by water availability in the environment or are limited by phylogeny (e.g. Hoverman et al., 2011).

Because of the dynamic capabilities of immune responses, we hypothesized that the effects of dehydration on immunocompetence and stress responsiveness would vary with habitat. Furthermore, we hypothesized that glucocorticoids do not necessarily increase in response to dehydration. Limited studies on animals that experience seasonal drought or inhabit xeric environments suggest a depressed dehydration-induced glucocorticoid response [e.g. Gila monsters, Heloderma suspectum (Moeller et al., 2017); Children's pythons, Antaresia childreni (Dupoué et al., 2014); and Awassi sheep, Ovis aries (Jaber et al., 2004; Hamadeh et al., 2006); but see Ghanem et al., 2008]. Dehydration has also been shown to enhance immune performance in multiple reptile species (Moeller et al., 2013; Brusch and DeNardo, 2017; Moeller et al., 2017; Brusch et al., 2019), including some that inhabit mesic environments (Brusch et al., 2019, 2020). While these studies show interesting trends, immune and stress responses to dehydration have not been systematically compared among species found in habitats with different levels of water availability.

A major challenge associated with examining the effects of dehydration on immunocompetence is to avoid potential confounding effects of other immune drivers such as food restriction or heat treatment (e.g. Ohira et al., 1981). Because of this, recent studies on dehydration and immune function have focused on squamate reptiles that feed infrequently. These species have extended energy budgets and naturally undergo periods of fasting that do not appear to influence immunocompetence (Moeller et al., 2013; Brusch and DeNardo, 2017). As such, we investigated interspecific variation in the physiological responses of squamate reptiles to dehydration.

Here, we used four confamilial pairs of squamate species (Table 1) to determine whether environmental water availability and dehydration influence stress and immunocompetence, and whether the response is influenced by relatedness at the familial level. One species from each of the four pairs inhabits a seasonally water-limited environment, while the other of each pair inhabits environments where water is typically always present, albeit to varying extents. In each species, we tested water loss rates and how an animal's hydration status (hydrated versus dehydrated) affected circulating corticosterone (CORT, the glucocorticoid found in reptiles) at initial and reactive levels, as well as how it affected various blood-based measures of immune performance (agglutination, lysis, bacterial killing ability and white blood cell differentials). We also tested dehydration tolerance in select species.

To test the effects of hydration state on immune function and CORT response, we collected blood samples from serially hydrated, dehydrated and, in some cases, rehydrated captive individuals of the following squamate species: Gila monster (Heloderma suspectum; n=8), Mexican beaded lizard (Heloderma horridum; n=7), Children's python (Antaresia childreni; n=8), ball python (Python regius; n=8), Sonoran gopher snake (Pituophis catenifer affinis; n=8), diamond-backed water snake (Nerodia rhombifer; n=8), western diamond-backed rattlesnake (Crotalus atrox; n=8) and northern cottonmouth (Agkistrodon piscivorus; n=8). These species represent confamilial pairs (within Helodermatidae, Pythonidae, Colubridae and Viperidae) that inhabit environments with varied seasonal water limitations (Table 1). While the species within each familial pair have varied degrees of relatedness, each is substantially more related to the confamilial species than to the other species; in fact, the range in divergence between species may provide a broader and more useful view of the effects in question, as we are not limiting our focus to species that diverged evolutionarily for any set duration. To validate differences in their adaptations to water limitation, we also measured the rate of cutaneous evaporative water loss for each species, as cutaneous evaporation can be influenced by native habitat aridity (e.g. in snakes; Lahav and Dmi'el, 1996; Dmi'el, 1998). Snout-to-vent length (SVL) was only measured in animals which were already being anesthetized, during which an accurate reading could be obtained (Gila monsters and beaded lizards), but all animals were of adult size, though the beaded lizards were on the smaller end of the adult range (Beck, 2009) and of a size similar to the adult Gila monsters. None of the subjects were siblings­. In studies with non-model vertebrates, it can be difficult to access large sample sizes. We found significant effects previously, using n=8 subjects (in Gila monsters) and wanted to minimize numbers for animal safety; thus, we had a maximum of n=8 animals per species. All research was conducted with approval from the ASU Institutional Animal Care and Use Committee (protocol 14-1044R).

Gila monsters are large lizards that inhabit desert scrub habitats in southwestern North America. In the Sonoran Desert, which encompasses most of the species' range, Gila monsters experience an extended hot, dry season prior to the monsoon season in late summer. During the dry season, Gila monsters experience a 60–80 day drought, during which there typically is no rainfall or standing water (Beck, 2009). Mexican beaded lizards are large-bodied lizards found in Mexico from Sonora to Oaxaca, mainly in tropical deciduous forests. Overall, these areas experience more rainfall and are more humid than those where Gila monsters live. While beaded lizards experience periods with little rain that can last at least 4 months (Murphy and Lugo, 1986), unlike the habitat of Gila monsters, standing water is usually available during this dry season (Beck, 2009). We used eight adult Gila monsters (SVL 274–338 mm, mass 302–556 g) received from the Arizona Game and Fish Department (AZGFD; permits SP577864, SP684760) and seven captive-bred small adult Mexican beaded lizards (SVL 347–352 mm, mass 513–719 g) of a slightly larger size compared with the Gila monsters. Both species were held under an AZGFD Holding Permit (SP598954; SP666234).

Children's pythons are small-bodied pythons that live in the wet–dry tropical rainforests of northern Australia. While their habitat receives extensive rainfall during the wet season, it also experiences a dry season that lasts for several months between May and October, during which these snakes have elevated plasma osmolality (Brusch and DeNardo, 2017). Ball pythons are heavier bodied, medium-sized pythons that live in tropical grasslands and rainforests of sub-Saharan Africa. These areas have a cool, dry season, when rain is infrequent (from November to May), but there are no extended dry periods, thus ensuring sufficient water availability throughout the year. We used eight captive-bred adult Children's pythons (mass: 362–433 g) and eight captive-born adult ball pythons (mass: 732–1058 g) in this experiment.

Sonoran gopher snakes are medium-sized snakes that inhabit the southwestern USA and northern states of Mexico, especially in the Sonoran Desert. Most of their range experiences extended hot, dry periods with no rainfall (Stebbins and McGinnis, 2018). Diamond-backed water snakes are small to medium-sized snakes that are restricted to marshes, ponds or other habitats along bodies of water in the middle southern USA. They primarily eat fish and amphibians, and thus rely heavily on constant access to water (Powell et al., 2016). We used eight wild-caught gopher snakes (mass: 223–454 g) from the Sonoran Desert in southern Arizona (AZGFD permit SP609640) and eight wild-caught diamond-backed water snakes (mass: 213–251 g) obtained by Stephen Secor in Itta Bena, MS, USA (Mississippi Department of Wildlife, Fisheries, and Parks collecting permit 1127121).

Western diamond-backed rattlesnakes are medium to large venomous snakes found in the southwestern USA and central and northern Mexico. They are widely distributed across a variety of habitat types, including the Mojave, Sonoran and Chihuahuan deserts, where they withstand extended hot, dry seasons with no access to water (Stebbins and McGinnis, 2018). Cottonmouths are medium to large venomous snakes that live in the southeastern USA. They are found in marshes, swamps and other slow-moving water habitats. Cottonmouths eat fish and amphibians, but they also eat more terrestrial prey, including mammals and birds. These snakes depend on ready access to water year round (Powell et al., 2016). We used eight wild-caught western diamond-backed rattlesnakes (mass: 292–479 g) obtained as nuisance animals near Phoenix, AZ, USA, and eight wild-caught northern cottonmouths (mass: 246–308 g) obtained from Steven J. Beaupre in Arkansas, USA. All venomous species were held under AZGFD permits (SP598954 and SP666234).

We measured cutaneous evaporative water loss (CEWL) rates under standard room conditions (25.8±0.1°C, 23.3±0.1% relative humidity) for all subjects. To measure CEWL, we used an AquaFlux AF200 (Biox Systems Ltd, London, UK), which derives water vapor flux from humidity gradient measurements using a condenser within a closed chamber. Using AquaFlux version 6.2 software, we calibrated the unit prior to each use and recorded CEWL rate (g m⁻² h⁻¹) in real time. To maintain tight contact between the standard measurement cap (orifice diameter of 7 mm) and an animal's dorsolateral skin at mid-body, we fitted the measurement cap with a donut-shaped piece of closed-cell foam with the hole in the center of the foam being of the same size as, and lined up with, the orifice of the measuring cap. We maintained contact between the unit and the animal's skin until CEWL rates stabilized (±0.02 g m⁻² h⁻¹ for 10 s). We ran all measurements in duplicate to verify repeatability.

We used all individuals to serially examine the effects of hydration state on immune function and stress responses. For Gila monsters, Mexican beaded lizards, and both python species, we collected blood samples from all subjects in a hydrated state (290–305 mOsm kg⁻¹), with elevated but well-tolerated increased plasma osmolality, or when ‘dehydrated’ (315–360 mOsm kg⁻¹), and rehydrated (265–300 mOsm kg⁻¹). In the other species, because of logistical limitations (e.g. difficulty of timed bleeding), we only collected two samples from each snake – one when hydrated (either initial or rehydrated) and the other when dehydrated, but the plasma osmolality range for each hydration state was the same in all species.

To begin the experiment, we fasted animals for 2 weeks to achieve a post-absorptive state. We then placed them in individual cages (75×35×13 cm; Freedom Breeder, Turlock, CA, USA) in an environmental chamber at 30°C. We gave them 2 weeks to acclimate, with water ad libitum. For the hydrated state assessments, we then collected an initial 0.7 ml blood sample from each individual from the caudal vein or by cardiocentesis using a heparinized 1 ml syringe. All initial samples were collected within 3 min of disturbance to avoid elevation of plasma CORT associated with handling (Romero and Reed, 2005). Blood samples were usually taken in the morning, between 08:00 and 11:00 h. Following each animal's initial sample, we administered a standardized 30 min perturbation that commonly induces a CORT response in reptiles (Schuett et al., 2004; Romero and Reed, 2005; Langkilde and Shine, 2006): either restricting the animal's movement and prodding it gently with a stick or agitating the animal in a closed container. We administered the stressors for 10 out of every 30 s of the treatment period. Immediately after the perturbation (31–34 min after handling started), we collected a 0.1 ml blood sample to assess stress reactivity (i.e. the difference in CORT concentration between the pre- and post-stressor samples).

After we collected the hydrated set of samples, we immediately removed water from the animal's cage and began the dehydration treatment for this experiment. For lizards, we also removed all fluid from the urinary bladder as needed, as some lizards, including Gila monsters, use the urinary bladder as a water reservoir (Davis and DeNardo, 2007). Removing urinary fluid thus increased the rate of dehydration in the lizards. While lizards were anesthetized (using 2% isoflurane), we removed bladder fluid using transurethral catheterization according to Davis and DeNardo (2007). We then confirmed the lack of bladder fluid using ultrasonography (Concept/MLV, Dynamic Imaging, Livingston, UK). As snakes lack a urinary bladder, we only catheterized lizards.

Once animals were deprived of water, we checked them visually every day and weighed them every 1–7 days to estimate water loss and dehydration. Frequency of weighing depended on the species (e.g. semi-aquatic species dehydrate more quickly) and the time since water was last provided (i.e. frequency increased as animals dehydrated). To monitor plasma osmolality, we typically collected small blood samples (0.1 ml) after 2 weeks of dehydration and then once a week thereafter. The semi-aquatic species were an exception; we collected their first blood samples 4 days after water was removed and subsequent samples more frequently. When an animal approached an elevated but well-tolerated plasma osmolality, the animal was not disturbed for 4 days to reduce the chance of recent stress affecting CORT in initial samples. After 4 days without disturbance, we collected the dehydrated blood sample set (as above, in the hydrated state; 0.7 ml initial, 0.1 ml post-stressor), then gave the animal water.

After 7 days of free water access, we collected the rehydrated blood sample set (as above; 0.7 ml initial and 0.1 ml post-perturbation) and removed the animal from the experiment. We immediately refrigerated samples until we could process, aliquot and freeze them (within 6 h of sampling) for later assays.

If an attempt to collect a blood sample failed, we thoroughly evaluated that animal and then left it for 4 days of no disturbance before another attempt at sampling. When repeating the sample collection process for the dehydrated sample set, we gave animals controlled amounts of water (roughly equivalent to 50% of the mass lost in the previous 4 days) to slightly rehydrate them, so they would remain safely dehydrated 4 days later. If three sampling attempts were unsuccessful, we skipped the sample and removed the animal from the hydration state experiment.

Dehydration tolerance trials were only conducted in animals that had already completed the hydration state experiment (above). The dehydration tolerance experiment was conducted last, as it explored critical levels of dehydration, and thus we were uncertain as to the recovery period and possible impact on other trials. Because of logistical limitations, we determined the extent of dehydration (i.e. plasma osmolality) that could be tolerated in only four of the eight species – gopher snakes (n=5), diamond-backed water snakes (n=7), western diamond-backed rattlesnakes (n=6) and cottonmouths (n=5). After completing the hydration state experiment, subjects were fed and then given at least 4 weeks to recover and reach a post-absorptive state before starting tolerance trials. Animals that had been difficult to bleed during the hydration state experiment were excluded from this part of the experiment, as obtaining blood samples at greater levels of dehydration would be even more difficult; with a limited number of animals, this reduced tolerance sample sizes. We measured the number of days until the snake showed signs of clinical dehydration, as well as the change in plasma osmolality and decrease in body mass over that time. Animals in a post-absorptive state were held in individual cages in an environmental chamber at 30°C with no food or water and were monitored daily for health and weekly (at most) for plasma osmolality (as for the hydration state experiment). When plasma osmolality reached 315–360 mOsm kg⁻¹; dehydrated), we increased health monitoring to 2–4 times daily until the animals showed the first symptoms of clinical dehydration (e.g. lethargy, loss of righting response, reduced responsiveness to stimulation; Divers, 1999). We limited blood samples to 0.1 ml and collected samples every 4–14 days (as the trials went on), for a maximum of 10 times over the course of the trial. Blood samples were usually taken in the morning, between 08:00 and 11:00 h. As soon as clinical signs appeared, we collected a final blood sample to test plasma osmolality, and we then provided the animal with access to water ad libitum. Two days after each animal was rehydrated, we removed the animal from testing altogether.

We assayed plasma from each initial sample (0.7 ml) to determine hydration state (i.e. plasma osmolality, mOsm kg⁻¹), initial plasma CORT concentration and immune function [agglutination, lysis, bacterial killing ability (BKA) and white blood cell differential]. We used post-perturbation plasma samples to determine reactive CORT concentrations.

After collecting the initial sample, we used two drops of whole blood to create blood smears for analysis of white blood cell count (described below). We centrifuged the remaining blood from the initial sample as well as the blood from the post-perturbation sample and divided the resulting plasma into 50 μl aliquots. We stored separated blood cells and plasma at −80°C until later analysis.

We analyzed each initial sample for plasma osmolality using vapor pressure osmometry (±6 mOsm kg⁻¹; model 5100C, Wescor, Inc., Logan, UT, USA) as described in Davis and DeNardo (2007).

We assayed plasma samples for CORT in duplicate using enzyme-linked immunoassay kits following the manufacturer's instructions (ADI-900-097, Enzo Life Sciences, Farmingdale, NY, USA). To validate the assay for these eight squamate species, we conducted a pilot assay to determine that the curves generated by serially diluted (5–80 times) pooled samples for each species were parallel with the standard curve (Fig. S1). Based on the pilot, we diluted the plasma samples from each species 40 times with assay buffer containing the prepared steroid displacement reagent at a volume equal to 1% of plasma volume before assaying.

We used 10 assay plates in total, with a standard curve on each plate, and ran them over 3 days. We assayed all the samples from a species on the same day and randomly assigned samples to a plate on a given day. Each plasma sample was 6.25 μl. The average inter- and intra-assay coefficients of variation were 7.1% and 13.2%, respectively. The average assay sensitivity was 6.34 pg ml⁻¹.

We measured agglutination and hemolytic ability of all initial samples following Moeller et al.’s (2013) modified protocol, originally from Matson et al. (2005). We serially diluted 20 µl of each plasma sample from 1:2 to 1:2048 in a 96-well plate and added 20 µl of diluted 50% heparinized sheep blood (SBH050, HemoStat Laboratories, Dixon, CA, USA) to each well. After incubation at 29°C [the active season mean diurnal body temperature of Gila monsters (Davis and DeNardo, 2010) and an active temperature for many of the species tested], we assayed for agglutination (marked by changes in the formation of a dense red blood cell pellet), incubated the plates at 29°C again, then assayed for hemolysis activity (marked by a lack of red pigment pellet presence).

Following French and Neuman-Lee (2012), we used initial plasma samples to conduct bacterial killing assays. We pipetted 2 µl of thawed plasma in duplicate onto 96-well round-bottom microplates. Negative-control wells held 6 µl phosphate-buffered saline and 18 µl CO₂-independent medium plus 4 mmol l⁻¹l-glutamine and no bacteria. Positive-control wells had 6 µl working bacteria solution, which contained 10⁴ colony-forming units of Escherichia coli (ATCC no. 8739), along with 18 µl medium. We also added 6 µl working bacteria solution and 16 µl of medium to samples, so all wells had a volume of 24 µl. We thoroughly mixed all wells, gently vortexed the microplates for 1 min, then incubated them at 37°C (the optimal temperature for exponential E. coli growth) for 30 min.

After incubation, we vortexed the plates for another minute, then added 125 µl sterile tryptic soy broth (Sigma-Aldrich no. T8907;15 g broth per 500 ml Nanopure water) to each well. After an additional minute of gentle vortexing, we read the absorbance of the plates at 300 nm (BioRad xMarkTM Microplate Absorbance Spectrophotometer). We incubated the plates at 37°C for 12 h, then vortexed for 1 min and read again. We compared sample absorbance before and after the 12 h of bacterial growth with the positive controls (0% bacterial killing), and calculated percentage bacteria killed as one minus the mean absorbance for each sample, which we ran in duplicate, divided by the mean absorbance for the positive control (triplicate wells containing only media and bacteria), multiplied by 100.

We created and analyzed duplicate blood smears from each initial sample. After inverting each blood sample several times to remix any settled cells, we smeared one drop of blood on a slide and stored them in a desiccator until they were preserved with methanol and re-stored for up to 6 months. We then stained with Giemsa–Wright stain for 60 min at the optimal dilution for each species to maximize visualization (1:15 dilution for helodermatids; 1:18 dilution for other species), then rinsed slides in two Nanopure water baths for 5 min each. We counted lymphocytes, heterophils, monocytes and basophils under a light microscope (BX60, Olympus Optical Co., Tokyo, Japan) at ×400 magnification, counting the number of each blood cell type (according to Cooper-Bailey et al., 2011) along a mapped grid until the first 100 white blood cells were counted on a slide. Eosinophils and azurophils were not found in most samples so were not counted.

We assessed normality using Shapiro–Wilk tests and qqnorm plots on the variables and their residuals. Within each state of hydration, stress responsiveness was measured as the difference between reactive CORT concentration and initial CORT concentration (ΔCORT=reactive CORT−initial CORT). In order to achieve a normal distribution for the models, a constant was added to ΔCORT values to ensure they were positive, and then log transformations were applied.

A simple linear model with two independent variables (habitat, species) was used to test for differences in CEWL and dehydration tolerance (plasma osmolality, mOsm kg⁻¹) during dehydration tolerance trials. We used a linear mixed model with habitat and family as fixed factors and individual as a random factor to test for differences in body mass loss in the dehydration tolerance trial. We also used linear mixed models to test the effects of our independent variables on stress and immunity metrics. Independent variables were hydration treatment (two levels: hydrated, dehydrated), family (four levels: Colubridae, Helodermatidae, Pythonidae, Viperidae), habitat (two levels: mesic, xeric), mass (continuous variable) and sex. We did not determine the sex of six gopher snakes; thus, our model included three levels for sex (male, female, unknown). Initial hydration values were used for the model when available for a family (i.e. helodermatids and pythons) but if a hydrated blood sample was not able to be collected (i.e. vipers and colubrids; see the description of blood sampling in ‘Hydration state experiment’, above), these were replaced with a sample from the rehydrated state. The models were fitted using the lme4 package (Bates et al., 2015) and by REML's t-tests using Satterthwaite approximations to degrees of freedom. Post hoc analyses were done using Satterthwaite degrees of freedom with the package lsmeans (Lenth, 2016).

For the CORT metrics, we used the same model parameters to construct separate models for initial CORT and CORT reactivity assessments. Independent fixed factors were hydration treatment, family, habitat, sex, mass and a habitat by family interaction term. Individual was included as a random factor. For a finer-grained analysis, we used a two-factor repeated measures analysis of variance (rmANOVA) with interactions followed by post hoc t-tests with Bonferroni corrections to test for a response to perturbation treatments and hydration in each species. We used paired t-tests to test for difference in CORT reactivity between hydrated and dehydrated states within a species and Welch's t-test to compare CORT reactivity between two species within a family.

For each immune metric (agglutination, lysis, BKA, lymphocytes, heterophils, monocytes, basophils), we constructed a model according to the previously described parameters with the immune metric as the dependent variable and independent fixed factors of treatment, family, habitat, sex, mass and a habitat by family interaction term. Sex and mass were not significant in any model, so were removed from the final model. Individual was included as a random factor. A single paired t-test was used within each species to assess the effects of hydration on immune metrics. Principal components analysis (PCA) was used to assess correlated variance among the immune metrics, and habitat and family were visually mapped onto the principal components. Habitat and family were also tested for their effects on the loading variables for the first two principal components using ANOVA.

CEWL was significantly higher in species from water-abundant habitats (F_1,61=84.4, P<0.001), though the mean value for one species from a water-abundant habitat (ball pythons; mean=5.35 g m⁻² h⁻¹) was lower than the mean value for one of the species from a water-limited habitat (Gila monsters; mean=6.41 g m⁻² h⁻¹). However, in all instances, the species from water-abundant habitats had higher CEWL than did their confamilial species from water-limited habitats (Table S1).

In our full model, initial plasma CORT concentration increased during dehydration sufficiently in some species (see below) to show a significant effect (F_1,141=31.0, P<0.001; Table 2, Fig. 1); however, a family by habitat interaction (F_3,60=5.80, P<0.001; Table 2) highlighted the complexity of these relationships. Within families, there were no significant differences in initial CORT between species from water-abundant and water-limited habitats.

The results of rmANOVA within each species showed that perturbation treatment had a significant effect on plasma CORT for all species (in species from water-abundant habitats: beaded lizards F_1,6=25.2, P=0.002; ball pythons F_1,7=28.3, P=0.001; northern cottonmouths F_1,7=120, P<0.001; diamond-backed water snakes F_1,7=6.89, P=0.034; and in species from water-limited habitats: Gila monsters F_1,7=5.94, P=0.045; Children's pythons F_1,6=28.3, P=0.003; diamond-backed rattlesnakes F_1,7=12.6, P=0.009; Sonoran gopher snakes F_1,6=28.4, P=0.002).

Dehydration had a significant effect on plasma CORT for three species from water-abundant habitats (beaded lizards F_2,12=25.2, P=0.002; northern cottonmouths; F_2,12=8.95, P=0.020; diamond-backed water snakes; F_2,12=11.5, P=0.011), and post hoc analyses for these species showed that dehydration increased plasma CORT in both the initial and perturbed states, although in beaded lizards the post hoc tests were just out of the range of significance (beaded lizards: initial t₆=1.68, P=0.072, perturbed t₆=1.63, P=0.077; northern cottonmouths: initial t₇=5.12, P=0.001, perturbed t₇=2.83, P=0.026; diamond-backed water snakes: initial t₇=4.25, P=0.72, perturbed t₇=2.93, P=0.077; Fig. 1). Dehydration had a significant effect on CORT for one of the species from water-limited habitats (Children's pythons: F_2,12=6.51, P=0.014), although Gila monsters showed a significant hydration by perturbation interaction term (F_2,14=5.84, P=0.014). Post hoc analyses showed that dehydration increased plasma CORT concentrations in both the initial and perturbed states for Children's pythons (initial t₆=2.88, P=0.028, perturbed t₆=2.48, P=0.047; Fig. 1), but only in the perturbed state for Gila monsters (initial t₇=1.73, P=0.128, perturbed t₇=2.97, P=0.021; Fig. 1).

For CORT reactivity, there was a significant effect of hydration (F_1,136=7.99, P=0.005; Table 2) and family (F_1,56=17.5, P<0.001; Table 2), but no other model factors were significant. Paired t-tests within species showed that only Gila monsters had a significantly different CORT reactivity between hydration states (t₇=3.04, P=0.019; Fig. 1). Welch's t-test comparisons of CORT reactivity between species from water-abundant and water-limited habitats within the same family showed no differences in the hydrated state (beaded lizards and Gila monsters t_6,7=1.98, P=0.088; cottonmouths and diamond-backed rattlesnakes t_7,7=1.29, P=0.235; water snakes and gopher snakes t_7,6=0.638, P=0.541), although the difference between ball pythons and Children's pythons was close to significant (t_7,6=2.20, P=0.055). Results were similar in the dehydrated state, with no significant differences between species from water-abundant and water-limited habitats for most family comparisons (beaded lizards and Gila monsters t_6,7=1.66, P=0.132; cottonmouths and diamond-backed rattlesnakes t_7,7=2.03, P=0.075; water snakes and gopher snakes t_7,6=0.737, P=0.480), but there was a significant difference between ball pythons and Children's pythons (t_7,6=2.66, P=0.025).

Hydration state affected all immunity variables except heterophils (Table 2). There was a habitat by family interaction for each immune variable (Table 2), and post hoc analyses showed complex patterns (see below for details and see Table S2 for values).

Agglutination was significantly affected by hydration state in our model (F_1,94=45.7, P<0.001; Table 2). Post hoc analyses showed that dehydration increased agglutination in three species from water-limited habitats: Gila monsters (t₆=4.46, P=0.004), Children's pythons (t₇=2.51, P=0.040) and Sonoran gopher snakes (t₇=3.31, P=0.013), as well as in two species from water-abundant habitats: northern cottonmouths (t₇=3.43, P=0.011) and diamond-backed water snakes (t₇=3.19, P=0.015) (Fig. 2). Habitat did not significantly affect hydrated comparisons of agglutination within any family, but vipers from the different habitats did differ in their agglutination response to dehydration (t₁₀=2.22, P=0.048).

Hemolysis was significantly affected by hydration state in our model (F_1,92=44.0, P<0.001; Table 2). Post hoc analyses showed that dehydration significantly increased hemolysis in two species from water-limited habitats: Gila monsters (t₆=4.89, P=0.003) and Children's pythons (t₇=2.97, P=0.021), and two from water-abundant habitats: ball pythons (t₇=2.94, P=0.023) and northern cottonmouths (t₇=2.78, P=0.027) (Fig. 2). The Viperidae were the only family in which the two species differed in hydrated lysis metrics (t₇=2.97, P=0.021).

BKA was significantly affected by hydration state (F_1,132=38.6, P<0.001; Table 2, Fig. 2). Post hoc analyses showed that BKA increased significantly during dehydration in three species from a water-limited habitat: Gila monsters (t₆=4.46, P=0.004), Children's pythons (t₇=2.70, P=0.031) and western diamond-backed rattlesnakes (t₇=3.36, P=0.012), but there was no significant effect in any species from a water-abundant habitat (Fig. 2). In three families, hydrated BKA significantly differed between the species from water-limited and water-abundant environments: Helodermatidae (t₆=2.71, P=0.035), Pythonidae (t₇=2.39, P=0.048) and Viperidae (t₇=7.00, P<0.001). No family had species that differed in their BKA response when dehydrated, although Viperidae trended towards significance (t₇=2.31, P=0.054).

Lymphocytes varied significantly with hydration state (F_1,92=6.51, P=0.012) (Table 2). However, post hoc analyses paired t-tests were not statistically powerful enough to capture an effect within species, although during dehydration beaded lizards (t₆=2.37, P=0.056) and ball pythons (t₇=1.99, P=0.087) both decreased their lymphocytes close to significance. For three families, the hydrated lymphocyte count differed between the species from water-limited and water-abundant habitats: the Pythonidae (t₇=7.73, P<0.001), the Viperidae (t₇=3.94, P=0.011) and the Colubridae (t₇=4.21, P=0.001).

Heterophils did not vary significantly among hydration states (F_1,88=0.13, P=0.716) (Table 2). Post hoc comparisons of heterophil counts between species within each family from water-abundant and water-limited habitats showed no significant differences, regardless of hydration state.

Monocytes varied with hydration state (F_1,95=6.42, P=0.013) (Table 2). Post hoc analyses showed that ball pythons were driving this effect, having a significant increase when dehydrated (t₇=3.01, P=0.019). Within families, there were no differences in hydrated monocyte counts between species from the two habitats. However, the Viperidae species differed in their monocyte counts during dehydration (t₅=4.36, P=0.007).

Basophils were significantly affected by hydration state (F_2,95=4.60, P=0.034) (Table 2), with all species except gopher snakes decreasing basophil count during dehydration (Table S1). However, similar to lymphocytes, post hoc analyses paired t-tests were not statistically powerful enough to capture an effect within species, and no trends were apparent. Hydrated basophil counts were not significantly different within any family. Within the Viperidae, western diamond-backed rattlesnakes trended towards a significantly higher basophil count than cottonmouths during dehydration (t₅=2.54, P=0.052).

The first two principal axes contained 55.5% of the variance among the immune response metrics (Fig. 3), while the first four principal axes contained 84.2% of the variance. Habitat did not show visual clustering when mapped onto the first two axes, but family did show clustering along the axes (Fig. 3). The first principal component was positively correlated with lysis and agglutination, and negatively correlated with lymphocytes and basophils. The second principal component was positively correlated with monocytes and negatively correlated with BKA and heterophils. For the first principal component, family, but not habitat, was a significant predictor for the two largest loadings of agglutination (family: F_3,152=60.5, P<0.001; habitat: F_1,152=0.519, P=0.473) and lysis (family: F_3,152=47.1, P<0.001; habitat: F_1,152=0.733, P=0.393). Similarly, for the second principal component, only family was a significant predictor for all of the loadings: heterophils (family: F_3,152=13.5, P<0.001; habitat: F_1,152=3.14, P=0.078), monocytes (family: F_3,152=129, P<0.001; habitat: F_1,152=3.40, P=0.067) and BKA (family: F_3,152=6.51, P<0.001; habitat: F_1,152=2.48, P=0.118). Both family and habitat together were significant predictors for only the two weakest loadings in the first principal component: lymphocytes (family: F_3,152=43.4, P<0.001; habitat: F_1,152=48.0, P<0.001) and basophils (family: F_3,152=10.0, P<0.001; habitat: F_1,152=6.34, P=0.013).

Dehydration tolerance (plasma osmolality at which an animal showed clinical signs of dehydration) was similar between habitat groups (F_1,17=0.35, P=0.56). Body mass loss did not differ between habitat groups (F_1,16=1.70, P=0.21) or species (F_2,16=2.12, P=0.15), and the number of days that it took species from water-limited environments to reach clinical dehydration was significantly greater than that of species from water-abundant environments (F_1,18=117, P<0.001) (Table S1). Sample collection was repeated for a few samples in two species: cottonmouths and water snakes. In cottonmouths, three hydrated, one dehydrated and one rehydrated sample were retaken. For the one dehydrated cottonmouth repeat sample, failure to sample occurred 3 times, so the animal was fed and given 2 weeks to recover before re-entering the trial and being dehydrated again. In water snakes, four hydrated and one dehydrated sample were retaken, but were all collected on the second attempt.

In this study, we used a comparative approach to examine how adaptation to varied environments affects the response to dehydration in terms of CORT production and immunocompetence among eight squamate species from four families. From each family, we included a species found in environments where water is seasonally limited, and one found in more water-abundant environments. While the degree to which chosen species are adapted to water limitations is likely quite variable, our categorization of species as being from water-abundant or water-limited habitats is supported by our results related to water balance (Table S1). CEWL rates in the species from the water-abundant habitats were, on average, 2 times greater than the values for confamilial species from more water-limited environments. This result supports our habitat assignments for the species pairs, as reptile species from water-limited habitats have reduced water loss compared with confamilial species living in more mesic conditions (Lahav and Dmi'el, 1996; Dmi'el, 1998; Cox and Cox, 2015). One water-limited species (Gila monsters) had higher CEWL rates than one of the mesic species (ball pythons); however, helodermatids are known for their high rates of EWL, which is thought to support thermoregulation and is mitigated by using the urinary bladder as a water reservoir (Davis and DeNardo, 2007). While only tested in two families (Colubridae, Viperidae), snake species from water-limited and water-abundant environments had similar dehydration tolerance but, consistent with the differing CEWL rates, the former took longer to reach clinical dehydration.

Our cumulative dataset provides initial assessments of the relative importance of phylogeny and habitat type in contributing to or limiting interspecific variation in the effects of dehydration on CORT circulation and immune function. We found that in some species, hydric state influenced initial CORT, CORT reactivity and various measures of immunocompetence, but that variation among species reflected complex interactions between family and habitat.

Circulating glucocorticoid concentrations are often used as proxies of stress levels (Norris and Carr, 2013), though such logic may be misleading (Romero and Beattie, 2022). Here, we assessed changes in CORT concentrations to understand dehydration as a potential challenge to an organism's homeostasis. In this study, we detected no significant difference in initial CORT between species from water-abundant and water-limited habitats (F_1,59.07=3.12, P=0.082), though there was a trend toward significance (Table 2, Fig. 1). The lack of a difference is not surprising considering all individuals were housed in similar, unchallenging captive conditions. Interestingly, however, both helodermatid species had relatively low initial CORT levels compared with most other species, suggesting a possible phylogenetic influence on CORT in some species (Fig. 1).

Each species' CORT response to dehydration was measured by a comparison of initial CORT concentrations when hydrated and dehydrated. While our model showed that initial CORT was significantly higher when individuals were dehydrated (Table 2, Fig. 1), it was not able to detect a difference based on habitat, although three of the species with higher levels of initial CORT were from water-abundant habitats (Fig. 1; beaded lizards are just out of range of significance). Strangely, pythons exhibited the reverse of our expectations, with xeric Children's pythons showing increased CORT when dehydrated, while mesic ball pythons showed no increase in initial CORT. Thus, while the three other family comparisons lend some support to our hypothesis that adaptations to abundant water availability increase sensitivity to dehydration, python physiology is not adequately captured by our model, either because our classification of their habitat was incorrect or because they have other factors driving their CORT response. Interestingly, two of the species from water-abundant habitats were semi-aquatic snakes, and although we did not include enough semi-aquatic species to run statistics on semi-aquatic versus terrestrial species patterns, our results suggest physiological access to water, either standing water in the environment or water reserves in the urinary bladder, may drive the increase of initial CORT during dehydration. While further studies are needed to validate this possibility, a clear result of this study is that it should not be assumed that dehydration always leads to an elevation of CORT, though the results suggest it does in some species (Fig. 1; Children's pythons, cottonmouths and water snakes).

In addition to measuring initial CORT, we also measured CORT responsiveness to an acute perturbance, and did so across various hydration states. Gila monsters were the only species to have a significantly higher ΔCORT when dehydrated (hydrated ΔCORT=−0.15±1.56 ng ml⁻¹, dehydrated ΔCORT= 38.8±15.8 ng ml⁻¹), although beaded lizards and cottonmouths also trended to higher ΔCORT with dehydration. Although our full statistical model showed a significant effect of family and hydration on CORT reactivity (Table 2), it appears to be largely driven by the high response of Gila monsters. Thus, unlike initial CORT, CORT responsiveness appears to be uncoupled from both family and habitat in these species. That is, CORT responsiveness was approximately constant across hydration states; however, it is possible that the animals sufficiently habituated to the stressor by the second treatment so that their stress reactivity was blunted, which would obscure any effect of dehydration.

In cases where a CORT response could have negative fitness effects (e.g. cause reduced parental care or decrease immunocompetence), either the CORT response can be attenuated in general (e.g. Wingfield and Hunt, 2002) or the effects of increased CORT can be dampened. As an example of the latter, during a test of the immunocompetence handicap hypothesis (that there is a trade-off between sexual signal development and immune function), Roberts and colleagues (2007) showed that increased CORT is not necessarily immunosuppressive in zebra finches. As CORT is a broad physiological driver, its production is also under constant selection, and thus the CORT response often aligns with species-specific needs based on social, developmental or environmental conditions (Wingfield and Sapolsky, 2003; Boonstra, 2005; Bókony et al., 2009). Our data suggest a similar phenomenon – of the three species that experienced increases in circulating CORT during dehydration, two showed increases in agglutination (northern cottonmouths and diamond-backed water snakes), one showed an increase in hemolysis (northern cottonmouths), and one showed a decrease in lymphocyte numbers (beaded lizards). These three species showed no significant change in any other immune metrics, supporting a complex relationship between CORT and immunocompetence. Further testing would be required to determine whether increased CORT levels are uncoupled from certain physiological responses and whether the decoupling mechanism involves the binding of CORT in plasma (e.g. to CORT-binding globulin; Breuner et al., 2013) or modification of responses within target tissues.

Overall, variation in CORT responsiveness across hydration states was driven by family, with no interaction from habitat (Table 2). In previous studies, Dupoué et al. (2014) and Moeller et al. (2017) observed increased CORT reactivity during dehydration in Children's pythons and Gila monsters, respectively. In our study, Gila monsters showed increased ΔCORT when dehydrated, and although Children's pythons increased their average CORT reactivity during dehydration (ΔCORT_hydrated=84.5±30.9 ng ml⁻¹ versus ΔCORT_dehydrated=115±21 ng ml⁻¹), the variance was too high to reveal a significant effect. However, colubrids and vipers did not show a significant change in CORT reactivity with changes in hydric state, demonstrating the variability that can exist across species and that generalizations must be made with extreme caution.

Dehydration can increase multiple metrics of immune function in various squamate species (Moeller et al., 2013; Brusch and DeNardo, 2017; Brusch et al., 2017; Moeller et al., 2017). Overall, our study similarly shows that agglutination, lysis and BKA are enhanced during dehydration in some species, and this is likely not due to a concentration effect (Moeller et al., 2013; though dilution trials were not done on CORT levels). When the animals were hydrated, there were differences in immune metrics between species from the water-limited and water-abundant environments in all four families. Vipers were the only family to show strong habitat-driven variation in immune metrics in response to dehydration, which differentially affected their agglutination, BKA, and monocyte and basophil counts. Differences between species from water-abundant or water-limited habitats were thus unpredictable and family specific, and all immune assays showed habitat by family interactions.

Because of the habitat–family interactions prevalent in our model outcomes, we used a PCA to visualize the correlations in variance between the immune variables (the principal axes) and how habitat and family map onto these axes (Fig. 3). Differences in scattering and clusters between plots clearly show that variation in immunocompetence is best clustered by family. This is supported by our analysis of the loadings on the principal components, which shows that family, but not habitat, is a significant predictor of the main sources of variance on both axes (i.e. agglutination and hemolysis on the first axis and heterophils and BKA on the second axis). Additionally, this analysis suggests interesting correlations between immunocompetence in these families. For example, agglutination and hemolytic ability show a strong positive correlation in their variance on the first axis, while BKA and heterophils show a strong positive correlation on the second axis (Fig. 3). However, as the first two principal components only explain 55.5% of the variance, these results highlight the need for further studies that expand the list of possible mechanisms driving this variance.

When looking at species-specific immune responses, some of our current findings align with previous work (Moeller et al., 2013). In Gila monsters, we observed a significant increase in hemolysis, agglutination and BKA with dehydration (Fig. 2), which were all enhanced in Moeller et al. (2013) as well. The outcome is similar between studies even though in this study, dehydration was at a more moderate level than in our previous study. Immune response may vary with the duration and severity of dehydration (Guseinov and Guseinova, 2008), but in this case, less extreme levels of dehydration did not seem to reduce the extent of immune enhancement observed. Brusch and DeNardo (2017) found that rattlesnakes with an elevated plasma osmolality did not show enhanced hemolysis or agglutination. They also did not find any difference in BKA, which increased significantly during moderate dehydration in our experiment. Future studies may benefit from adjusting target ‘dehydrated’ plasma osmolality ranges closer to the physiological limits of water balance for a particular species to ensure the ‘dehydration’ treatment is a homeostatic challenge.

Previous studies on the effects of dehydration on white blood cell count have indicated that dehydration may activate granulocytes (Tsuchida et al., 2004; Penkman et al., 2008), decrease lymphocytes (in mice: Tsuchida et al., 2004; and rats: Guseinov and Guseinova, 2008), increase overall white blood cell count (in snakes; Brusch et al., 2020), or have little to no effect on cell numbers (in humans; Laing et al., 2008; Mitchell et al., 2002). However, most of these studies were conducted on mammals, which brings into question possible confounding effects, as they were fasted during the studies and mammals often have energy budgets that must be balanced over relatively short time frames. In our study, we found that, overall, blood cell counts were more variable than other assay results. Dehydration generally increased monocytes, but decreased lymphocytes and basophils, and there was no effect on heterophils (Table 2). Also, while family affected cell count for all cell types, habitat type only affected lymphocyte count. Clearly, more in-depth studies of leukocyte production and release are needed to fully understand the physiological triggers of immune cell redistribution, as well as potential impacts on animal ecology.

Our study provides the first comparative assessment of the effects of dehydration on circulating glucocorticoids and a comprehensive suite of blood-based immune metrics. Our study design enabled us to initially assess whether the interspecies variation was driven by environment (water-abundant or water-limited habitat) or constrained by evolutionary history (taxonomic family). Based on these data, we posit that the effects of dehydration on CORT and immune function are quite varied among squamate reptiles.

The lack of a clear habitat effect does not lend support to the Pathogen Exposure Hypothesis, or to adaptation to the water limitations of the species' current environments. Whether the habitats in which these family lines first evolved have resulted in specific families relying on different strategies (e.g. utilizing plasma proteins more than cell proliferation) has yet to be investigated.

By comparing the effects of dehydration on CORT and immunocompetence across four taxonomic families, each represented by two species that live in environments with relatively contrasting water availability, our data show that (1) dehydration, often thought to increase CORT levels, only increased initial CORT in the two semi-aquatic species; (2) when CORT did increase, it did not necessarily result in a corresponding decrease in immune function for those species; and (3) effects of dehydration on these CORT and immunity measures were not mainly driven by adaptations to current environments, but instead appear to be more heavily related to taxonomic family. While our results cannot explain the mechanisms or evolutionary significance of the findings, they provide insight and encouragement for additional studies to further our understanding.

Special thanks to Martin Feldner, Roger Repp, Christian Wright, Megan Murphy, Matthew Harris, Jason Ortega, Steven Beaupre, Doug Price, Stephen Secor, Ron Rutowski, Susannah French, Michael Angilletta and John Sabo, as well as to the two anonymous reviewers who helped us greatly improve the presentation of this work. Additional thanks to the Deviche and Blattman labs at ASU for use of equipment. A few lines from the Results and Discussion in this paper are reproduced from the PhD thesis of K.T.M. (Moeller, 2016).