Heat hardening in a pair of Anolis lizards: constraints, dynamics and ecological consequences

Rising global temperatures have led to an increase in the occurrence of extreme heat events (Meehl et al., 2007; Stillman, 2019; Vose et al., 2005). Extreme heat causes organisms to experience physiological stress and can push species beyond their thermal limits (Kingsolver and Buckley, 2017; Dowd et al., 2015; Schulte, 2014). For example, extreme environmental temperatures can induce selection (Rodríguez-Trelles et al., 2013), drive rapid population declines (Chandrapavan et al., 2019) and perturb mutualistic associations (Dunn et al., 2004; Kingsolver et al., 2013; Stillman, 2019). Heat waves are expected to become more and more common as climate change progresses (Diffenbaugh and Field, 2013), and therefore it is imperative that we understand the impact of extreme heat on organisms to better predict and potentially mitigate the deleterious outcomes of anthropogenic activity (Williams et al., 2016; Stillman, 2019).

Phenotypic plasticity in thermal physiology is considered a critical determinant of resilience to extreme heat (Somero, 2010; Huey et al., 2012; Gunderson and Stillman, 2015; Seebacher et al., 2015). Species that can plastically increase their heat tolerance when habitat temperatures rise should be more resilient to heat extremes than those that cannot (Huey et al., 2012). While in general the heat tolerance plasticity of ectotherms does not appear to be fully compensatory (i.e. plasticity cannot fully counteract negative effects of warming; Gunderson and Stillman, 2015; van Heerwaarden et al., 2016; MacLean et al., 2019), heat tolerance plasticity can still reduce overheating risk and potentially delay extinction (Gunderson et al., 2017; Morley et al., 2019; Rohr et al., 2018). Nonetheless, there are many aspects of the dynamics of thermal tolerance plasticity that we do not yet fully understand, hindering our ability to fully assess the consequences of plasticity in changing environments (Beaman et al., 2016).

Thermal tolerance plasticity is generally thought to manifest over the course of weeks to months (Somero et al., 2017; Angilletta, 2009). However, there are forms of plasticity that can increase thermal tolerance in minutes to hours. Known as ‘hardening’, these physiological responses arise after organisms experience a thermal shock (Bowler, 2005). For example, the heat tolerance of five notothenioid fish species increased within 4 h of experiencing their critical thermal maximum (CT_max, the temperature at which organisms lose righting ability; Bilyk et al., 2012). Heat hardening represents the first line of physiological defense against heat stress (Phillips et al., 2016), and the expression of heat hardening has been linked to increased fitness under hot field conditions (Bowler, 2005; Loeschcke and Hoffmann, 2007). Despite its potential to be a key physiological buffer from rapidly changing conditions, the expression and temporal dynamics of heat hardening have been relatively little explored compared with longer-term plastic responses (Gilbert and Miles, 2019).

A recent pattern to emerge from studies of thermal physiology is a negative relationship between basal thermal tolerance and thermal tolerance plasticity (‘trade-off hypothesis’; van Heerwarden and Kellermann, 2020). Looking across species or populations, those with the greatest heat or cold tolerance often have the lowest capacity to plastically increase tolerance when experiencing warmer or cooler conditions, respectively (Stillman, 2003; Comte and Olden, 2017; Armstrong et al., 2019; Gilbert and Miles, 2019; Nguyen et al., 2019). Trade-offs between basal thermal tolerance and plasticity have also been observed among individuals of the same species (Phillips et al., 2016; Morgan et al., 2018), raising the possibility that similar mechanisms influence the expression of plasticity at different levels of organization. Several processes could underlie the basal tolerance/plasticity trade-off, including shared genetic architecture of each trait (pleiotropy), correlational selection on each trait, and the existence of hard physiological thresholds (van Heerwarden and Kellerman, 2020). However, it is also possible that the trade-offs do not reflect biological reality and are instead a statistical artifact. Measurements of basal thermal tolerance occur in both the dependent and independent variable when testing the trade-off hypothesis, and therefore a negative relationship between the variables could simply be due to autocorrelation (van Heerwarden and Kellermann, 2020). Testing for this artifact is rarely considered in analyses of plasticity (Ghalambor et al., 2015).

In this study, we tested the magnitude, temporal dynamics and ecological consequences of rapid heat hardening in two Anolis lizards: A. carolinensis and A. sagrei. The magnitude and temporal dynamics of hardening were tested by measuring the thermal tolerance of individuals 2, 4 or 24 h after an initial heat shock. To investigate the effect of heat hardening on performance, we estimated the threat of overheating for each species within our sampling area each day over the past 25 years. We did so using a biophysical model to translate daily temperature records into operative field temperatures for comparison against the observed thermal limits of each species. In addition, we tested for individual-level trade-offs between basal heat tolerance and heat hardening with an approach that accounts for the possibility that apparent trade-offs could be a statistical artifact. Using the same statistical approach, we extended our analysis by reanalyzing previously published data on lizard heat hardening that support the trade-off hypothesis at the individual level.

All animal protocols were approved by the Tulane University Institutional Animal Care and Use Committee (protocol #504). Anoliscarolinensis Voigt 1832 and Anolis sagrei Duméril and Bibron 1837 adult males were collected from the greater New Orleans area (29.9511°N, 90.0715°W) from February to September of 2019. Animals were transported to Tulane University, where they were housed in individual plastic cages (18.03×11.18×13.97 cm) with a wooden dowel perch in climate-controlled incubators (Percival Model I30-NL) at 32.0°C and 75% humidity on a 12 h:12 h light:dark cycle for 24 h before experiments. Snout–vent length (A. sagrei: 52.4±5.5 mm, A. carolinensis: 61.2±5.4 mm; mean±s.d.) and mass (A. sagrei: 4.30±1.46 g, A. carolinensis: 4.44±1.22 g) were also measured prior to heat tolerance testing.

After 24 h, CT_max was measured for each animal following published protocols (Leal and Gunderson, 2012; Gunderson et al., 2018). CT_max is a dynamic thermal tolerance metric taken as the point at which an organism loses neuromuscular coordination such that it cannot escape a harmful situation (Hutchison, 1961). A T-type 36-gauge wire thermocouple probe attached to a digital thermometer (Omega HH81A, accurate to 0.1°C) was inserted into the cloaca of each lizard and secured in place with a small strip of tape. We could therefore continuously monitor body temperature during trials and control the rate of warming of each individual, targeting a heating rate of 2°C min⁻¹. This allowed us to ensure that all individuals had similar heating rates, and thus avoid the effects of heating rate differences (which often result from body size differences) on heat tolerance measurements (Terblanche et al., 2007). There was no difference in heating rate between A. sagrei and A. carolinensis in our trials (2.38±0.43 and 2.32±0.38°C min⁻¹, respectively; P=0.383) based on a mixed effects linear model with heating rate as a fixed effect and individual lizard as a random effect and implemented in the R package ‘nlme’. Therefore, heating rate was not considered in subsequent analyses. After the probe was affixed, the lizard was tethered to a 30×30×16 cm cardboard enclosure by a short piece of dental floss with a loop at the end, which was secured around the waist, allowing movement but preventing escape. Each lizard was warmed under a 100 W incandescent lamp and righting response was tested at 1°C body temperature intervals beginning at 38°C, by flipping the animal onto its back. To induce righting, tweezers were used to gently pinch the hindlimbs within a 10 s window. If the lizard righted itself, it was placed back under the heat lamp and flipped again at the next test temperature. If the lizard was unable to right itself, CT_max was recorded as 0.5°C below the final test temperature (i.e. halfway between the last body temperature at which the lizard righted itself and the body temperature at which it did not). Each lizard was then placed back in the incubator under the standard conditions described above.

To measure heat hardening, the CT_max of each animal was measured a second time following the protocol described above at either 2, 4 or 24 h after the initial CT_max measurement. Animals were randomly assigned to a given time interval. The heat hardening capacity of each animal (ΔCT_max) was calculated by subtracting the initial CT_max value from the second CT_max value. All lizards were returned to their collection locality the day after their second CT_max measurement. Fifteen to twenty lizards were collected for each hour interval for each species, resulting in 97 lizards tested in total (Table 1).

To assess which species exhibited more heat hardening capacity, a two-way ANOVA was conducted on ΔCT_max values using species and hour interval as factors. Residuals for the delta values of all treatment groups were normally distributed (Shapiro–Wilks test, P=0.1329). A Bartlett's test found unequal variances (P=0.002286), though ANOVA is robust to violations of the assumption of homogeneous variances, particularly when groups have similar and large sample sizes, as is the case with our analysis (Glass et al., 1972; McGuinness, 2002; Oehlert, 2000). To test for differences in absolute heat tolerance between species, an unpaired t-test was conducted on CT_max values at each time point with subsequent false discovery rate correction to adjust for multiple comparisons.

To test for a trade-off between heat hardening capacity and basal thermal tolerance, we used Pearson product-moment correlation analysis of the animals at the 2 and 4 h time intervals. The standard null hypothesis (i.e. no relationship between two variables) is a correlation coefficient (r) of 0. However, 0 is not the null hypothesis when the variables being tested share the same term (Jackson and Somers, 1991). That was the case here, as basal heat tolerance was used to calculate the heat hardening value, ΔCT_max. To estimate the null expectation, we employed the randomization approach of Jackson and Somers (1991; see also Ghalambor et al., 2015). We calculated 1000 correlation coefficients with randomized data in which the initial (basal) and hardened CT_max values were randomly sampled with replacement. The null hypothesis is rejected if the empirical correlation coefficient is greater than 95% of all permuted values. In addition to our own data, we conducted the randomization test with the published dataset on heat hardening in the lizard Lampropholis coggeri, the only organism that we are aware of for which a trade-off has been shown between basal heat tolerance and heat hardening at the individual level (Phillips et al., 2016). The authors reported results from two different observers, and we report analyses with both datasets. We performed all statistical analyses using the R statistical programming language version 3.6.2 (http://www.R-project.org/) (R Development Core Team, 2019).

We estimated the threat of overheating for each species using a dynamic bioenergetic model that estimated operative temperatures experienced by lizards in both open, full sun (0% shade) and closed, deeply shaded (90% shade) microhabitats. We employed the model developed for small lizards from Buckley (2008), which we implemented in R (script available from figshare: https://doi.org/10.6084/m9.figshare.c.4955858.v4). Operative temperatures were calculated every hour over the 25 year period from 1995 to 2019 in the New Orleans region. We selected those years because they span the time frame during which A. sagrei density increased dramatically in New Orleans (Edwards, 2017). The following climatic variables were extracted from the biophysical modeling package ‘NicheMapR’ for inclusion in the model: estimated hourly air temperature, soil temperature, wind speed, solar radiation and sun zenith angle (Kearney and Porter, 2017). Operative temperatures were estimated at different perch heights for each species (1.6 m for A. carolinensis and 0.6 m for A. sagrei), reflecting differences in their microhabitat use (Edwards and Lailvaux, 2012). For each species, snout–vent length and mass were set to the mean values of the lizards used in the heat hardening analysis. For every day of every year, we calculated the number of hours operative temperatures would exceed the heat tolerance limit of each species in the sun and in the shade. Overheating hours were calculated both with and without heat hardening incorporated. Thermal tolerance without heat hardening was set as the mean initial CT_max of each species, while CT_max with heat hardening was set as the mean CT_max at the 2 h heat hardening time interval.

With respect to heat hardening, there was no significant interaction between hour and species in ΔCT_max values (ANOVA, F_2,91=2.143, P=0.1232). However, there was a significant effect of hour interval (F_2,91=3.583, P=0.0318) and species (F_1,91=48.394, P<0.001) on ΔCT_max. Mean ΔCT_max for A. carolinensis was consistently positive (0.4 to 2.1°C) and was highest at the 2 and 4 h intervals (Fig. 1). The mean ΔCT_max of A. sagrei was consistently negative (−0.4 to −1.5°C; Fig. 1).

Species differed in CT_max under all conditions (all P<0.05 after false discovery rate correction; Table 1, Fig. 1), though the sign of the difference changed before and after heat hardening. Anolis sagrei had an initial mean CT_max 0.6°C higher than that of A. carolinensis (Table 1). However, A. carolinensis had a mean CT_max 0.9 to 2.6°C higher than that of A. sagrei after heat hardening, with the biggest mean difference at the 2 and 4 h intervals (Table 1, Fig. 1).

The warmest operative temperatures are found during the summer months in sunny microhabitats (Fig. 2). Operative temperatures in a sunny microhabitat were estimated to exceed the CT_max of A. carolinensis by 757±107 h year⁻¹ per year assuming no heat hardening, and by 495±102 h year⁻¹ with heat hardening factored in (Fig. 3A). For A. sagrei, operative temperatures in a sunny microhabitat were estimated to exceed CT_max by 702±114 h year⁻¹ without heat hardening, and by 875±118 h year⁻¹ with heat hardening factored in (Fig. 3A). Both species had a greatly diminished overheating threat in shaded microhabitats (Fig. 2, Fig. 3B). Here, operative temperatures were predicted to exceed the CT_max of A. carolinensis by 3.8±10 h year⁻¹ without heat hardening and by 0.07±0.4 h year⁻¹ with heat hardening. For A. sagrei, operative temperatures were estimated to exceed CT_max by 0.44±2 h year⁻¹ without heat hardening and by 2.28±7 h year⁻¹ with heat hardening.

Based on correlation analysis, basal heat tolerance was negatively associated with heat hardening (ΔCT_max) in both A. carolinensis (r=−0.65; P<0.001; Fig. 4A) and A. sagrei (r=−0.55; P<0.001; Fig. 4B). However, the empirical correlation coefficients did not exceed the 95th percentile of permuted values for either species (Fig. 4C,D). Therefore, the null hypothesis that there is no relationship between basal heat tolerance and heat hardening capacity cannot be rejected. We found the same result in our reanalysis of published trade-off data in the skink Lampropholis coggeri (Phillips et al., 2016), as the empirical correlation coefficients did not exceed the 95th percentile of permuted values for their observer 1 (r=−0.77; 95th percentile of permuted values=−0.90) or observer 2 (r=−0.52; 95th percentile of permuted values=−0.84; see Fig. S1).

Thermal tolerance plasticity can buffer organisms from stressful climatic conditions that are predicted to become more common in the future (Vose et al., 2005). Heat hardening is a rapid form of plasticity that can increase the survival of organisms subsequently challenged with high temperatures (Loeschcke and Hoffmann, 2007; Sejerkilde et al., 2003; Sørensen et al., 2009; Folk et al., 2006), and therefore may be important under global change. Nonetheless, data on heat hardening are relatively rare compared with data on other forms of thermal plasticity, especially in vertebrate ectotherms.

We found that A. carolinensis exhibited significant heat hardening (Fig. 1). The heat hardening response of A. carolinensis (ΔCT_max of 2.1°C by 2 h post-heat shock) is notably rapid and of high magnitude for a vertebrate ectotherm. For example, the mean response in several fish, salamander and other lizard species (ΔCT_max ∼1°C) is about half of that observed for A. carolinensis over similar time frames (Abayarathna et al., 2019; Bilyk et al., 2012; Gilbert and Miles, 2019; Hutchison and Maness, 1979; Phillips et al., 2016). In contrast, A. sagrei did not exhibit heat hardening, and in fact their thermal tolerance tended to decrease after heat shock (Fig. 1, Table 1). Heat hardening capacity has therefore evolved to different states among Anolis lizards. More broadly, A. sagrei lacks putatively adaptive plasticity in several other thermal traits (Rogowitz, 1996; Gunderson et al., 2020; but see Kolbe et al., 2014) for which A. carolinensis does exhibit plasticity (Campbell-Staton et al., 2018; Gatten et al., 1988; Ryan and Gunderson, 2021). Why these species have evolved different levels of thermal plasticity is unknown and any proposed mechanism is purely speculative (Garland and Adolph, 1994), but the difference may be related to A. carolinensis experiencing greater thermal variation over evolutionary time because their historical range is centered at higher latitudes than that of A. sagrei (Bozinovic et al., 2011; but see Gunderson and Stillman, 2015).

Because of differences in heat hardening, which of the two species was more heat tolerant depended on the environmental context. Anolis sagrei had higher basal heat tolerance than A. carolinensis; however, the heat tolerance of A. carolinensis exceeded that of A. sagrei post-heat shock (Table 1, Fig. 1). To assess the consequences of these differences in heat hardening and heat tolerance, we used a mechanistic biophysical modeling approach to estimate the threat of overheating for both species using daily temperature records over the past 25 years in our sampling area. Heat hardening provided a clear benefit to A. carolinensis in sunny microhabitats, decreasing their annual overheating hours by an average of 35% compared with values without heat hardening (Fig. 2A). Therefore, a relatively minor plastic change in heat tolerance (∼2°C) can have significant performance consequences (Gunderson et al., 2017; Morley et al., 2019). Conversely, heat exposure decreased the heat tolerance of A. sagrei and led to an increase in their susceptibility to subsequent heat exposure and a decrease in their performance relative to that of A. carolinensis (Fig. 3).

Shade availability is critical for most ectotherms to avoid overheating, as operative thermal conditions in the sun often reach stressful levels (Gunderson and Leal, 2012; Logan et al., 2013; Sunday et al., 2014). Shaded microhabitats are clearly important for both A. sagrei and A. carolinensis, providing operative conditions that rarely exceed their heat tolerance limits (Fig. 2, Fig. 3B). That said, there are costs associated with thermoregulatory behavior (Huey and Slatkin, 1976). For example, energy is expended when shifting between the sun and shade (Sears and Angilletta, 2015), movement can increase the risk of predation (Downes, 2001), and favorable microhabitats may not be as suitable for feeding, mating and territory defense (Gunderson and Leal, 2016; Sinervo et al., 2010). Because of their lack of heat hardening, A. sagrei may need to invest more heavily in thermoregulation than A. carolinensis, and this difference could be exacerbated by global warming. In the 25 year period addressed in this study, there were few exceptionally hot summers in which operative temperatures in shaded microhabitats exceeded CT_max. However, as global temperatures continue to increase, exceptionally hot summers will become more common (Diffenbaugh and Field, 2013) and shaded microhabitats will provide less of a refuge (Gunderson and Leal, 2012; Huey et al., 2009). This should disproportionately affect A. sagrei, which cannot physiologically adjust to warmer conditions.

Thermal plasticity itself can evolve as environments change (Lande, 2014; Chevinet al., 2010; Bodensteiner et al., 2021), and trade-offs between basal thermal tolerance and tolerance plasticity may be key to understanding the mechanisms and constraints that shape adaptation to past and future thermal regimes (van Heerwaarden and Kellermann, 2020). That said, some approaches to testing the trade-off hypothesis leave open the possibility that apparent trade-offs are actually statistical artifacts. We initially found a significant negative relationship between basal thermal tolerance and hardening in both A. carolinensis and A. sagrei (Fig. 4A,B). However, the evidence ultimately does not support the trade-off hypothesis when null expectations that take the potential for statistical artifact into account are considered (Jackson and Somers, 1991; Fig. 4C,D). Furthermore, when we reanalyzed previously published data showing a significant trade-off between basal tolerance and hardening in a species of skink (Phillips et al., 2016), we again found no evidence for a trade-off based on our null expectations (Fig. S1).

In sum, the best available evidence supports statistical artifact, rather than biological reality, as the explanation for apparent individual-level trade-offs between basal heat tolerance and heat hardening in lizards. The degree to which this problem might extend to other organisms, to other forms of plasticity (e.g. acclimation, developmental plasticity) and across different levels of biological organization is unknown. While compelling examples of tolerance/plasticity trade-offs certainly exist (reviewed in van Heerwaarden and Kellerman, 2020), it is also clear that tests that do not incorporate appropriate null expectations are likely to find spurious support for the trade-off hypothesis. Addressing this issue will be critical for assessing the true importance of tolerance/plasticity trade-offs for thermal adaptation and global change resilience.

With the predicted increase in the frequency and severity of extreme heat events in the future, the importance of thermal tolerance plasticity will likely increase. Our results show that for some, but not all, lizard species, heat hardening can be an important component of the plastic response to warming. Still, our knowledge of constraints on the expression and evolution of thermal tolerance plasticity is still limited and requires additional attention, especially with respect to the design of analyses to test leading hypotheses.

We would like to thank two anonymous reviewers for helpful feedback on this manuscript. We would also like to thank Wayne Wang, Lane Pierson, Akhila Gopal, Lucy Ryan and Annie Huang for assistance with this project.