Ectothermic organisms respond to rapid environmental change through a combination of physiological and behavioral adjustments (Bennett and Huey, 1990; Angilletta et al., 2006; Williams et al., 2008; Huey et al., 2012). As these traits are often functionally linked (Careau et al., 2008; Biro and Stamps, 2010; Mathot and Dingemanse, 2015), an effective response to environmental change will depend on the magnitude and direction of their association. For example, metabolic rate fuels functions and processes underlying locomotion, while locomotor activity is involved in both energy expenditure (e.g., mate searching) and energy gain (e.g., foraging) (Mathot and Dingemanse, 2015). Accordingly, locomotor activity level contributes to the acquisition and allocation of energy into maintenance, growth, reproduction, and survival, and thus to individual fitness in a given habitat (Burger et al., 2021). Information on behavioral and physiological covariation in a thermal context is required to understand the mechanistic link between individual traits and population dynamics under environmental change (Johnston et al., 2019); however, this issue has received surprisingly limited attention.

Current theory predicts that the relationship between metabolic rate and locomotor activity will depend on how an organism manages its energy budget (Careau and Garland, 2012; Mathot and Dingemanse, 2015). Under the “allocation model,” which assumes a fixed daily energy expenditure, an individual with a high standard metabolic rate can invest less energy into locomotor activity than an individual with low maintenance costs, and vice versa. The “performance model,” on the other hand, assumes a positive relationship (slope > 1; Halsey et al., 2019) between maintenance metabolic rate and daily energy budget. Under this scenario, locomotor activity should be positively related to standard metabolic rate, irrespective of its influence on the energy budget. Finally, the “independent model” assumes no relationship between maintenance metabolic rate and energy available for other behavioral (e.g., locomotor activity) and physiological tasks (e.g., digestion) under non-restricted total energy expenditure. In this case, a positive relationship is predicted between energy-gaining locomotor activity and maintenance metabolic rate because of a positive association between maintenance metabolic rate and daily energy expenditure (Mathot and Dingemanse, 2015).

A recent meta-analysis revealed no relationship between resting metabolic rate (RMR) and locomotor activity across taxa (Mathot et al., 2019), which was interpreted as being the result of interspecific variation in net energy contribution of this behavioral trait to the total energy budget. However, there are at least two alternative explanations for this result in ectotherms. First, that metabolic-behavioral relationships are affected by environmental stress (e.g., temperature), which may mask trait covariation due to individual variation in the slope or shape of their thermal reaction norms (Killen et al., 2013). Second, trait covariation may be attenuated by low repeatability (i.e., the relative amount of among-individual variation) of one or both traits at a given body temperature. Unfortunately, most studies on metabolic-activity relationships have been performed at one body temperature only and have used phenotypic correlations, which assume thermal independence of the link, perfect repeatability (R = 1) of both traits, and/or identical directions for among- and within-individual correlation. However, available studies have shown that the metabolic-activity link will change with temperature in an endotherm (Chappell et al., 2004), that the average repeatability of behavioral and physiological traits is 0.4 (Bell et al., 2009; White et al., 2013; Holtmann et al., 2017) and that metabolic-behavioral trait correlations vary at different variation levels (Gifford et al., 2014; Videlier et al., 2019), which suggests that temperature and within-individual (residual) variation may attenuate or mask trait correlation at the whole-phenotypic level. The empirical support for this prediction is missing (Dingemanse and Dochtermann, 2013; Killen et al., 2013; Careau and Wilson, 2017).

In this study, we undertook repeated measurements of RMR (energy consumption of post-absorptive and non-active, but immature, individuals during their typical inactivity period) and exploratory locomotor activity (ELA; distance covered in a new open-field arena) in juvenile newts at two body temperatures in the autumn and the following spring. While both traits are repeatable and thermally dependent in newts (Gvoždík and Kristín, 2017; Baškiera and Gvoždík, 2019, 2021), their covariation level is unknown. A recent analysis of metabolic traits has provided support for the allocation model of energy management in newts (Baškiera and Gvoždík, 2020), implying that energy-consuming ELA should be negatively associated with RMR. However, among-individual variation in the slopes of their thermal reaction norms for ELA is negligible (Baškiera and Gvoždík, 2020), suggesting that the masking or attenuating effect of its within individual variation modifies the RMR-ELA relationship between temperatures. To demonstrate the joint effect of body temperature and among-individual variation on the association of RMR and ELA, we first examined trait relationships at two body temperatures in addition to the commonly used “single temperature” approach. As a next step, we analyzed thermal reaction norm parameters for these traits with respect to among- and within-individual variation.

We obtained 496 repeated measures of RMR and ELA from 62 individuals (Figures 1A,B). Both traits were affected by body mass [October: 0.76 ± 0.12 (SD) g; May: 1.04 ± 0.24 g] at both body temperatures, with RMR increasing with body mass and ELA decreasing (Table 1). Mass-adjusted mean RMRs were similar between months at both body temperatures (Figures 1A,B). While mass-adjusted mean ELA showed similar values between months at 12°C, however, newts increased their average locomotor activity at 22°C in May (Figures 1C,D).

At 12°C, we detected repeatable among-individual variation in ELA, but not in RMR (Table 2 and Figures 2A,C). Association between both traits was statistically non-significant at all variation levels (Table 2 and Figure 3A). At 22°C, we found support for repeatability of RMR, but not ELA (Table 2 and Figures 2B,D). Further, we found no evidence for trait association at this body temperature (Table 2 and Figure 3B).

The fixed structure of random intercept and slope model showed that RMR increased with body mass, while ELA decreased (Table 3). Intercepts of thermal reaction norms for both traits and slopes for RMR contained detectable amounts of among-individual variation (Table 3 and Figures 2E–H). There was no significant covariation between intercepts and slopes among individuals and between months in either trait (Table 3). Intercepts of thermal reaction norms for RMR and ELA were positively associated at the individual level, but not at the whole-phenotypic or residual levels (Table 2 and Figure 3C). As the credible interval for r_ind was broad, we were unable to obtain any firm conclusion on the exact magnitude of this association. Slope values indicated a disparate effect of body temperature on RMR and ELA (Table 3), with positive values indicating that RMR increased with temperature (Figures 1A,B), while ELA decreased (Figures 1C,D). We found no support for the repeatability of slopes for ELA and its between-trait association at any variation level (Table 3 and Figure 3D).

Despite their fundamental importance in behavioral and physiological ecology (Careau et al., 2008; Biro and Stamps, 2010; Careau and Garland, 2012; Mathot and Dingemanse, 2015; Killen et al., 2016), factors affecting metabolic-behavioral relationships remain poorly understood. We identified body temperature and among-individual trait variation (repeatability) as important modifiers of the relationship between RMR and ELA in juvenile newts and showed that detection of among-individual variation in RMR and ELA varies with body temperature. At 12°C, we only detected repeatability in ELA, while at 22°C, we detected repeatability in RMR but not ELA. Accordingly, we were only able to obtain poor estimates for their association at each body temperature. In contrast, intercepts of thermal reaction norms across these temperatures showed repeatability in both traits, with the traits being positively associated at the individual, but not the whole-phenotypic, level.

Our results provide evidence for a positive association between RMR and ELA in juvenile newts at the individual level. While a recent meta-analysis found no support for this relationship across taxa, which was interpreted as the result of the ambiguous impact of locomotor activity on individual energy budget (Mathot et al., 2019), our results suggest a previously unconsidered explanation, i.e., that the inconclusive link between metabolic rate and locomotor activity results from the opposing effect of body temperature on among-individual variation in each trait. In our study, the variation pattern produced poor estimates for correlation coefficients at both temperatures; however, “averaging” trait values across temperatures (intercepts of thermal reaction norm) stabilized individual rank for each trait, which improved estimates of individual correlation coefficient. This demonstrates the necessity to examine metabolic-behavioral relationships using repeated measurements from at least two body temperatures to obtain representative values for ectotherm taxa.

We observed that the relationship between RMR and ELA remained inconclusive at both body temperatures in juvenile newts. The body temperatures used here were within the range of preferred body temperatures for this species (Gvoždík and Kristín, 2017), suggesting that the newts in our study were not thermally stressed. According to a recent theory, the absence of thermal stress should not affect physio-behavioral associations (Killen et al., 2013). However, our results showed that even in the absence of stress, temperature changes altered the amount of among-individual variation in opposite directions for each trait, which resulted in similarly weak support for their association. At present, it is not known how thermal stress alters the association between metabolism and behavior in other ectotherms. While it is known that experimental temperature is an important modifier of the relationship between metabolism and behavior in endotherms (Chappell et al., 2004), the results are not directly comparable between endotherms and ectotherms due to the disparate influence of environmental temperature on their metabolic rate (McNab, 2002). Although further empirical studies are needed to fully understand the effects of environmental stressors on trait (co)variation, it is already clear that trait (co)variation is more sensitive to environmental changes than has previously been thought.

In our study, unlike intercepts, slopes contained significant among-individual variation in RMR only. Previously published results provide mixed support for this finding (Careau et al., 2014; Mitchell and Biro, 2017; Baškiera and Gvoždík, 2019; Kar et al., 2021), suggesting that the amount of among-individual variation in thermal sensitivity varies substantially among taxa and traits. Among-individual variation in slopes appears to be affected by the disparate effects of temperature on physiological and behavioral traits (Abram et al., 2017). The thermal dependence observed in RMR is a typical example of a kinetic effect, which is determined by the amount of kinetic energy in a physiological system. In contrast, ELA is mostly affected by the integrated effect of temperature, which results from sensed thermal information integrated by the neural system. Accordingly, the shape of reaction norms for ELA varies between newt life stages and differs substantially from thermal performance curves for forced locomotion (Baškiera and Gvoždík, 2019; Gvoždík and Boukal, 2021). Given the presence of among-individual variation is the basic assumption for evolution by natural selection, different amount of this variation in slopes for RMR and ELA suggests that thermal sensitivities of physiological and behavioral traits evolve in disparate rates in newts (Gvoždík, 2015) as in other long-lived ectotherms (Muñoz et al., 2014; Bodensteiner et al., 2021).

One could argue that our estimates of among-individual variation are confounded by the experimental design used. Thermal reaction norms, for example, are often non-linear (Huey and Stevenson, 1979; Amarasekare, 2015; Little and Seebacher, 2021), whereas our measurements at two body temperatures assume a linear thermal dependence of the traits studied. Although the thermal dependence of ELA is U-shaped across the broad body temperature gradient, both ELA and RMR are linear within the limited temperature range used in the present study (Gvoždík and Kristín, 2017; Baškiera and Gvoždík, 2019). Accordingly, our “two-temperature” approach provided unbiased estimates of the variation in the slopes of thermal reaction norms. Another source of bias could arise from neglecting the maternal effect on examined traits (Bernardo, 1996; Mousseau and Fox, 1998; Uller, 2008). For logistical reasons, we could not raise newts from each clutch separately, and therefore consider maternal identity as an additional random factor. Therefore, we cannot exclude the possibility that among-individual variation in the traits studied was overestimated compared to values obtained from randomly selected individuals in a large population. Despite this shortcoming, the phenotypic similarity between offspring should not affect the main result of our study, i.e., the differential effect of body temperature on among-individual variation in physiological and behavioral traits.

The positive association between RMR and ELA has both ecological and evolutionary implications. From an ecological point of view, the positive RMR-ELA link suggests that juvenile newts manage their energy budget according to the performance model (Careau and Garland, 2012; Mathot and Dingemanse, 2015). This is contrary to recent findings showing a negative relationship between maintenance metabolic rate and energy available for other tasks, but is in accordance with the positive association between maintenance and maximum metabolic rate, in adult newts across body temperatures (Baškiera and Gvoždík, 2020). The likely explanation for this discrepancy is the context-dependency of energy management (Halsey et al., 2019). Juvenile newts, for example, show variation in their RMR and growth rates in the presence of intra- and interspecific competition (Janča and Gvoždík, 2017), which may corroborate this view. However, further experimental support is needed for a definite conclusion. From an evolutionary point of view, this finding suggests that phenotypic selection should act on RMR and ELA in a correlative manner. Theory suggests that the adaptive significance of RMR is highly context-dependent (Careau et al., 2008; Burton et al., 2011; Norin and Metcalfe, 2019). Positive RMR-ELA covariation means that if resources are widely available, selection should favor metabolically fast individuals with high locomotor activity level, while low metabolic rate and locomotor activity should be beneficial under a scarcity of resources. In ectotherms, context-dependent adaptive significance is complicated by variation in the thermal dependency of both traits. For example, under limited resources, high body temperatures disproportionally increase mandatory metabolic costs relative to locomotor activity level, which provides a further penalization to metabolically-fast individuals. As such, climate warming is likely to erode metabolic-behavioral variation within ectotherm populations in the near future.

To sum up, we demonstrated that body temperature and individual variation hide metabolic-behavioral relationships in an ectotherm, with body temperature modifying the link not through disparate thermal dependency of mean trait values but the amount of among-individual variation. By focusing on parameters of reaction norms rather than trait values at a single body temperature, we were able to identify the link between metabolic rate and locomotor activity. This clearly indicates the importance of a repeated measures design across at least two body temperatures for obtaining meaningful information on metabolic-behavioral links in ectotherms (Killen et al., 2016). The minimalistic approach could be further improved by adding more temperatures and repeated measures to increase the accuracy of parameter estimates (van de Pol, 2012; Dingemanse and Dochtermann, 2013; Mitchell and Houslay, 2021); however, finding an optimal solution among the statistical needs, logistical issues, and confounding effects of habituation or acclimation provides a challenging task in this field. Despite these issues, we believe that adopting the repeated measurements approach over two or more temperatures will substantially improve our understanding of the role of energy budget (Careau and Garland, 2012), pace of life syndrome (Goulet et al., 2017), and trade-offs (Careau and Wilson, 2017) in shaping ectotherm responses to environmental change.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://doi.org/10.6084/m9.figshare.17013449.v1.

All experimental procedures were approved by the Expert Committee for Animal Conservation of the Institute of Vertebrate Biology of the Czech Academy of Sciences (research protocol no. 135/2016). Permission to capture newts was provided by the Agency for Nature Conservation and Landscape Protection of the Czech Republic (1154/ZV/2008).

LG conceived the ideas, designed the methodology, and analyzed the data. SB and LG collected the data and wrote the manuscript. Both authors contributed critically to the drafts and gave final approval for publication.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.