According to the metabolic theory of ecology, metabolic rate, an important indicator of the pace of life, varies with body mass and temperature as a result of internal physical constraints. However, various ecological factors may also affect metabolic rate and its scaling with body mass. Although reports of such effects on metabolic scaling usually focus on single factors, the possibility of significant interactive effects between multiple factors requires further study. In this study, we show that the effect of temperature on the ontogenetic scaling of resting metabolic rate of the freshwater amphipod Gammarus minus depends critically on habitat differences in predation regime. Increasing temperature tends to cause decreases in the metabolic scaling exponent (slope) in population samples from springs with fish predators, but increases in population samples from springs without fish. Accordingly, the temperature sensitivity of metabolic rate is not only size-specific, but also its relationship to body size shifts dramatically in response to fish predators. We hypothesize that the dampened effect of temperature on the metabolic rate of large adults in springs with fish, and of small juveniles in springs without fish are adaptive evolutionary responses to differences in the relative mortality risk of adults and juveniles in springs with versus without fish predators. Our results demonstrate a complex interaction among metabolic rate, body mass, temperature and predation regime. The intraspecific scaling of metabolic rate with body mass and temperature is not merely the result of physical constraints related to internal body design and biochemical kinetics, but rather is ecologically sensitive and evolutionarily malleable.
The rate of biological processes or ‘pace of life’ has major consequences for the survival and reproductive fitness of organisms, including their ability to capture food and avoid becoming food. An important indicator of the pace of life is the rate of metabolism, which fuels all biological activities (Brown et al., 2004; Glazier, 2015). Given the central importance of metabolism in supporting the time-sensitive activities of life, biologists have devoted much attention to what controls its rate.
This equation, first used by Krogh (1916) to quantify metabolic scaling, describes a linear relationship in log–log space, where a is the scaling coefficient (antilog of the intercept of the log–linear regression line, at which M is 1), and b is the scaling exponent (slope of the log–linear regression line). The RTN model of West et al. (1997) specifically predicts that b should equal ¾ or nearly so, thus providing theoretical support for the classical ¾ power law (Kleiber, 1932, 1961; Hemmingsen, 1960; Savage et al., 2004; Banavar et al., 2010). In addition, the MTE assumes that increasing temperature (T) accelerates the rate of the biochemical reactions constituting metabolism in an exponential way, following the Arrhenius equation (Arrhenius, 1915; Laidler, 1984; Gillooly et al., 2001). Furthermore, the MTE assumes that mass and temperature have independent, multiplicative effects on metabolic rate.
However, several studies have shown that mass and temperature interactively affect metabolic rate, causing the body-mass scaling exponent b to covary significantly with temperature, both within and across species (reviewed in Glazier, 2005, 2014c, 2018a, 2020). Therefore, some researchers have recommended that equations describing the combined effects of mass and temperature on metabolic rate should have an interaction term (Xie and Sun, 1990; Ohlberger et al., 2012; Glazier, 2014b). A growing number of studies have also shown that b varies in response to many other biological and ecological factors, including lifestyle (Glazier, 2005, 2006, 2014c, 2018a, 2020; Killen et al., 2010; Pequeno et al., 2017; Tan et al., 2019), activity level (Weibel et al., 2004; Glazier, 2005, 2008, 2009a, 2010, 2020; White et al., 2007; Carey et al., 2013; Dlugosz et al., 2013), developmental stage (Wieser, 1984; Riisgård, 1998; Glazier, 2005; Killen et al., 2007; Czarnołęski et al., 2008; Glazier et al., 2015), growth rate (Glazier, 2005, 2014b; Czarnołęski et al., 2008; Glazier et al., 2011; Tan et al., 2019) and predation regime (Glazier et al., 2011; Gjoni et al., 2020), among others (reviewed in Glazier, 2005, 2014b,c; White and Kearney, 2014).
Most studies of the ecology of metabolic scaling have focused on the effects of single environmental factors, each considered alone. Although several studies have examined the combined effects of temperature and various chemical or physical factors on the metabolism of aquatic animals (e.g. Mosser and Hettler, 1989; Claireaux and Lagardère, 1999; Finn et al., 2002; Spanopoulos-Hernández et al., 2005; Allan et al., 2006; Glover et al., 2012; Carey et al., 2016; Leiva et al., 2018; Kratina et al., 2019; Schwieterman et al., 2019; and references therein), only a few have reported size-specific, interactive environmental effects on the metabolic scaling exponent (Dehnel, 1960; Newell et al., 1972; Carey and Sigwart, 2014). Our study (along with accompanying work by Gjoni et al., 2020: see the Discussion section) is the first to show that an abiotic environmental factor (T) and a biotic environmental factor (predation regime) interactively affect metabolic scaling. These results have both scientific and applied value. First, they not only reinforce the view that the body-mass scaling of metabolic rate is highly sensitive to many kinds of external (ecological) factors and not simply the result of internal body-design constraints, as assumed by the MTE, but also show that specific ecological effects may themselves depend on other interacting ecological factors. Second, our findings extend the wide range of reported top-down effects of predators on prey traits to include the temperature-sensitive scaling of metabolic rate with body mass. Several studies have examined effects of predators on prey metabolic rates (reviewed by Glazier et al., 2020; also see Jermacz et al., 2020a,b), but few have focused on how predators affect metabolic scaling (Glazier, 2006; Glazier et al., 2011; Gjoni et al., 2020). Moreover, our work shows that predator effects on prey are temperature sensitive, not only with regard to metabolic rate (e.g. Janssens et al., 2015; Laws, 2017), but also its scaling with body mass. Gjoni et al. (2020) have examined how temperature and chemical cues of fish predators interactively affect metabolic scaling, whereas the present study examines interactive effects of temperature and the predation regime of source habitats. Metabolic scaling is not a simple result of bottom-up effects of resource supply, as assumed by the MTE, but also is influenced by top-down effects on resource-demanding processes, such as growth and behavioral activity (also see Glazier et al., 2011, 2020; Harrison, 2017; Hatton et al., 2019). Third, our results suggest that effects of climate change (T) on biological systems may not be simple results of biochemical kinetics, as assumed by the MTE, but may depend critically on the ecological context.
In the laboratory, we investigated the effects of temperature on the ontogenetic metabolic scaling of mid-Appalachian populations of the amphipod Gammarus minus Say 1818 inhabiting three freshwater springs with fish predators and three without. Amphipods are abundant keystone species in many springs, streams and groundwater ecosystems throughout temperate regions of the world (Glazier, 2014a). Their advantages for metabolic scaling studies include ease of collection and study throughout the year, lack of postembryonic developmental metamorphosis (thus simplifying ontogenetic comparisons of metabolic rate), and a body-mass range exceeding 30-fold, thus ensuring adequate power of the regression analyses (also see Materials and Methods). The study springs are also advantageous because they have similar, relatively constant water temperatures and chemical compositions, and their macroinvertebrate faunas consist mostly of amphipods (G. minus) and relatively few other species, thus representing useful, relatively simple, naturally controlled ecosystems for studying a wide variety of ecological and evolutionary questions (cf. Glazier, 2014d). In addition, the study populations of G. minus are relatively isolated geographically, thus allowing adaptive responses to differences in predation regime to evolve in the absence of substantial gene flow. Mid-Appalachian populations of G. minus exhibit significant genetic differentiation (Gooch, 1990; Gooch and Glazier, 1991; Kane et al., 1992; Carlini et al., 2009) and the study populations in particular display significant morphological, behavioral and life-history differences between springs with versus without fish predators (Glazier, 1999, 2000; Glazier and Deptola, 2011; Glazier et al., 2011, 2020; Glazier and Paul, 2017).
MATERIALS AND METHODS
The study springs in Blair and Huntingdon Counties of central Pennsylvania (USA) have similar physical and chemical conditions (i.e. clear, shallow, cool, well oxygenated, moderately flowing hard water; common occurrence of watercress, Nasturtium offinale; temperature ≈9–13°C; pH ≈6.6–7.7), but three contain the predatory fish Cottus cognatus (Ell, Blue and Williamsburg) and three do not (Petersburg, Kanesatake and Big Rock). Amphipods are a major component of the diet of C. cognatus (Newman and Waters, 1984) and other Cottus species (MacNeil et al., 1999). Other predators in the study springs are either rare (e.g. salamanders) or relatively ineffective at preying on healthy amphipods (e.g. flatworms) (Jenio, 1979; Glazier and Gooch, 1987). The macroinvertebrate faunas of the study springs are relatively species poor, with the most abundant species including the amphipod G. minus at all sample sites, and at some sites also the isopod Lirceus brachyurus (Blue Spring) and the hydrobiid snail Fontigens nickliniana (Ell, Williamsburg and Kanesatake Springs). Further details about the study springs can be found in Glazier and Gooch (1987), Glazier et al. (1992), Glazier (1999), Glazier et al. (2011), and Glazier and Paul (2017).
Amphipod collection, temperature acclimation and measurements of metabolic rate
Amphipods with a wide range of sizes (∼0.06 to 8 mg dry body mass) were collected with dip nets during the spring, summer and autumn months of 2010, 2018 and 2019, and acclimated in 5 liter basins containing aerated water and natural detrital leaf food from their native spring at one of three temperatures (4, 10 and 16°C) in an environmental control room for at least 7 days before their metabolic rate was measured. The normal mean temperature of our study springs throughout the year is near 10°C, which we used as a reference point for choosing symmetrically lower and higher experimental temperatures at levels causing minimal mortality during acclimation. Before a respirometry run, experimental amphipods (totaling 710 individuals) were fasted for 24 h (to remove the metabolic effects of feeding) in porous 120 ml cups placed in a 5 liter basin of aerated native spring water at each of the experimental temperatures (4, 10 and 16°C). During starvation, a gammarid gut is evacuated within 24 h at temperatures ranging from 5 to 15°C (Marchant and Hynes, 1981). Each coprophagous amphipod was placed in the upper compartment of each cup separated by nylon mesh from the lower compartment where feces could collect without being eaten.
where PO2,E is the partial pressure (1 Torr≈1 mm Hg) of oxygen in an experimental syringe, PO2,C is the partial pressure of oxygen in a control syringe, S is the solubility coefficient of oxygen in water (2.57, 2.22 and 1.82 μmol l−1 Torr−1 at 4, 10 and 16°C, respectively), A is the volume of 1 mol O2 at standard temperature and pressure (22.414 l mol−1), V is the volume of water in a syringe (liters), and t is the incubation time (h). Immediately after a respiration rate measurement was made, each amphipod was stored at −70°C, and to determine dry body mass, dried in an oven at 60°C for 60 h, and weighed (±l μg) on a Cahn C-31 microbalance (Cahn Instruments, Cerritos, CA, USA). All raw data on metabolic rate in relation to body mass of G. minus in six populations are available in Table S1.
We used least squares regression (LSR) to relate metabolic rate to body mass (following Glazier et al., 2011). This method is preferred over reduced major axis (RMA) analyses when the independent variable (body mass) is measured with much less error than the dependent variable (metabolic rate) (Smith, 2009), as is typically the case for metabolic scaling relationships (Glazier, 2010, 2014b; White, 2011). Log10-transformation normalized the data variation, and permitted proportional, linear relationships to be readily discerned (Kerkhoff and Enquist, 2009; Glazier, 2013). The significance of differences among scaling exponents (slopes) and intercepts was estimated by ANCOVA (with body mass as a covariate), and by comparing these parameters using 95% confidence intervals. If a mean value was outside the 95% CI of another mean value and vice versa, they were considered significantly different (P<0.05), following Smith (1997). We used SYSTAT 10 software (SPSS Inc., Chicago, IL, USA) for all statistical analyses.
We deemed the sampled body-mass ranges as sufficient for statistical testing because they were greater than the one order of magnitude often used as the minimal acceptable range in experimental and comparative studies of intraspecific metabolic scaling (Xie and Sun, 1990; Glazier, 2009b, 2020; Vaca and White, 2010; Carey et al., 2013; Hirst et al., 2014; Fossen et al., 2019; Tan et al., 2019). In addition, they were similar to those used in other amphipod studies that successfully detected significant population or species differences in the scaling of metabolism and other traits (Glazier et al., 2011, 2020; Glazier and Deptola, 2011; Glazier and Paul, 2017; Shokri et al., 2019; Gjoni et al., 2020).
Individual variation in resting metabolic rate (RMR) of Gammarus minus was large, as often reported in other studies of amphipods (Sutcliffe, 1984; Glazier et al., 2011; Galic and Forbes, 2017; Semsar-kazerouni and Verberk, 2018; Kratina et al., 2019; Shokri et al., 2019; Gjoni et al., 2020) and other animals (Phillipson and Watson, 1965; Kremer, 1977; Chown et al., 2007; Careau et al., 2008; Gudowska et al., 2017; also see the Discussion section). Nevertheless, we found highly significant relationships between RMR and body mass (M) at all three experimental temperatures for aggregate amphipod samples from the three springs with fish predators and the three without fish predators (Tables 1 and 2; Figs 1 and 2). Significant relationships between RMR and body mass also occurred in all individual population samples, except for the Williamsburg Spring population samples at 10 and 16°C (Tables 3 and 4; Fig. 3). Lower correlation coefficients appear to be related to smaller available body-mass ranges in populations exposed to size-selective fish predation (Table 3; Figs 1–3), especially in Williamsburg Spring where fish predators are most numerous (Glazier et al., 2011), and where it is difficult to find amphipods, most of which hide in the sediment. Stubbington et al. (2017) describe the difficulty of sampling all size classes of amphipods in sediments.
Significant effects of temperature on RMR occurred for both aggregate samples from the springs with fish predators and those without (Table 2) and for all six populations, as well (Table 4). However, the RMR responses to cooling (10 to 4°C) and warming (10 to 16°C) differed significantly between the aggregate samples from springs with versus without fish predators. Although their body-mass adjusted log10 RMR means (±95% C.I.) did not differ significantly at the baseline native temperature of 10°C (−0.473±0.082 versus −0.393±0.088), they did differ significantly at 4°C (−0.346±0.078 versus −0.782±0.083) and 16°C (0.062±0.069 versus −0.345±0.077) (based on the ANCOVA analysis in Table 2; also compare Fig. 1A–C). Furthermore, the RMR scaling exponents (b) differed significantly with temperature (T) in the aggregate sample from springs without fish predators, as indicated by a significant T×M interaction, though this was not the case for the aggregate sample from springs with fish predators (Table 2). Four individual population samples (all three of the population samples from springs without fish predators, and one population sample from a spring with fish) showed significant T×M interactions, whereas two population samples from springs with fish predators did not (Table 4). In addition, the effect of temperature on the elevation of the RMR scaling relationships varied among the various populations and with predation regime (Figs 2 and 3). For example, the RMR of amphipods from springs with fish predators increased more substantially between 10 and 16°C than between 4 and 10°C, but the opposite occurred for amphipods from springs without fish predators (Fig. 2). Other kinds of inter-population variation in temperature effects also occurred, including unexpected decreases in overall RMR with increasing temperature in two populations (4 to 10°C in the Ell Spring population: Fig. 3A; and 10 to 16°C in the Big Rock Spring population: Fig. 3F).
Although b tended to decrease with increasing temperature in the populations from springs with fish predators (Ell, Blue and Williamsburg: Fig. 2A, Fig. 3A–C and Fig. 4A), the opposite pattern was observed in populations from springs without fish predators (Petersburg, Kanesatake and Big Rock: Fig. 2B, Fig. 3D–F and Fig. 4B). Furthermore, at the lowest experimental temperature (4°C), populations from springs with fish predators showed significantly steeper RMR scaling relationships than those of populations from springs without fish predators (Fig. 1A and Fig. 4), whereas the opposite was observed at the highest temperature (16°C) (Fig. 1B and Fig. 4). As a result, in populations from springs with fish, small juveniles tended to show large increases in RMR with increasing temperature, whereas large adults showed less change (Fig. 2A and Fig. 3A–C). In contrast, in populations from springs without fish, large adults tended to exhibit large increases in RMR with increasing temperature, whereas small juveniles showed less change (Fig. 2B and Fig. 3D,F).
Interactive effects of temperature and predation regime on metabolic scaling
Both temperature and predation regime affected the scaling of metabolic rate in populations of the amphipod G. minus (Tables 1–4; Figs 1–4). Furthermore, the effect of temperature on metabolic scaling differed markedly between populations from springs with versus without fish. As a result, at 4°C, RMR scaling exponents (b) were always higher in populations from springs with versus without fish predators (0.718–0.912 versus 0.440–0.575, respectively), whereas the opposite occurred at 16°C (0.374–0.612 versus 0.923–1.078; see Fig. 4).
These results challenge various theoretical models of metabolic scaling. For example, they contradict the MTE, which assumes that b is fixed and unrelated to temperature and other environmental factors (Brown et al., 2004). The cell-size model (Davison, 1955; Kozłowski et al., 2003) and classical dynamic energy budget (DEB) theory (Kooijman, 2000) do not consider effects of temperature and predators on metabolic scaling, though potentially they could be modified or extended to do so. Only the metabolic-level boundaries hypothesis (MLBH) (Glazier, 2005, 2010, 2014c; Killen et al., 2010), the viscosity hypothesis (VH) (Verberk and Atkinson, 2013), the acclimation hypothesis (AH) (Fossen et al., 2019) and a modified version of DEB theory applied to some colonial animals (White et al., 2011; Kearney and White, 2012; cf. Glazier, 2018a) predict that temperature should affect metabolic scaling. According to the MLBH, the RMR scaling exponent should correlate negatively with the overall level of metabolism, as indicated by the elevation of a scaling relationship (Glazier, 2005, 2008, 2010). This should occur because volume-related tissue demand should chiefly influence RMR at low metabolic levels (b approaching 1), but surface-related resource supply and waste removal should chiefly influence RMR at high metabolic levels (b approaching 2/3 in isomorphic organisms). Therefore, since increasing temperature increases the overall level of RMR, b should correlate negatively with temperature, as often observed in ectothermic organisms (Glazier, 2005, 2014c, 2018a), and especially in sedentary species (Glazier, 2020). The VH and AH also predict a negative relationship between b and temperature. However, although the inverse correlations between b and temperature observed in amphipod populations from springs with fish predators (Fig. 2A, Fig. 3A–C and Fig. 4A) are consistent with the MLBH, VH and AH, the positive correlations between b and temperature observed in populations from springs without fish (Fig. 2B, Fig. 3D–F and Fig. 4B) seem to contradict all of these hypotheses. However, as supported by evidence reported by Glazier (2020), the MLBH predicts that positive associations between b and temperature may arise because increasing temperature may increase various volume-related, energy-demanding processes such as growth and activity, thus causing b to increase toward isometry. The specific hypothesis proposed below follows this line of reasoning.
Hypothetical explanation based on size-specific variation in predation risk
We propose a novel working hypothesis to explain why predation regime adaptively alters the effect of temperature on the metabolic scaling of G. minus populations (Fig. 5). Although this hypothesis focuses only on our specific results, one may modify it so that it may potentially apply to other species with various patterns of age- and size-specific mortality in diverse ecological contexts. Our specific hypothesis relies on two well-supported assumptions. First, it assumes that visually hunting fish (C. cognatus) are more of a threat to large conspicuous amphipods than smaller less visible amphipods. Four observations support this assumption. First, large amphipods occur more frequently in the guts of fish predators (including C. cognatus) than in the surrounding habitat (Newman and Waters, 1984; Allan and Malmqvist, 1989). Second, the relative frequency of small versus large amphipods is higher in our study springs that contain fish predators versus those that do not (Glazier et al., 2011). Third, in the presence of fish predators, large adult amphipods tend to reduce their behavioral activity more than do smaller juvenile amphipods (Newman and Waters, 1984; Andersson et al., 1986; Allan and Malmqvist, 1989; Friberg et al., 1994). Fourth, large adult amphipods tend to have significantly lower fat levels in springs with versus without fish predators, whereas smaller juvenile amphipods do not show this difference or an opposite effect (Glazier et al., 2011, 2020). This size-specific pattern is likely due to reduced foraging activity by large conspicuous amphipods in springs with fish. Many studies have shown that the presence of fish predators reduces feeding activity in amphipods (e.g. Newman and Waters, 1984; Friberg et al., 1994; Åbjörnsson et al., 2000; Beermann et al., 2018).
Second, our hypothesis assumes that, given the need or opportunity, large amphipods will cannibalize smaller, more vulnerable amphipods. Amphipods are opportunistic omnivores that often engage in size-specific cannibalism (Dick, 1995; MacNeil et al., 1997, 1999; Lewis et al., 2010). Cannibalism should be especially severe on small juvenile amphipods when large adults are numerous and likely also highly voracious, because of the depletion of other food resources at high population densities. Amphipod cannibalism increases when other food resources are not available (Dick, 1995). The presence of large adult amphipods can also alter the behavior of smaller juvenile amphipods, apparently as an anti-predation response (MacNeil et al., 1997, 1999; McGrath et al., 2007).
Based on these assumptions, we propose that habitat differences in predation regime have caused adaptive evolutionary, size-specific shifts in the responses of metabolic rate to changes in temperature. According to our hypothesis (Fig. 5), in populations inhabiting fish springs, young small inconspicuous amphipods may significantly increase their activity and growth (and thus metabolic rate) in response to increasing temperature without significantly increased mortality from fish predators, whereas older larger, more conspicuous amphipods cannot do this without suffering heavy mortality. As a result, these size-specific, adaptive responses in metabolically costly activity and growth cause the metabolic scaling lines at different temperatures to converge at larger body sizes.
However, in populations from fishless springs, large adult amphipods may significantly increase their activity and growth (and thus metabolic rate) in response to increasing temperature without increased mortality risk, whereas younger smaller amphipods cannot do this without suffering substantial mortality from numerous actively foraging cannibalistic adults (made possible by absent fish predation). Therefore, in fishless spring populations, the metabolic scaling lines at different temperatures tend to converge at smaller body sizes. In short, relative mortality on juveniles versus adults may drive size-specific, energy-expensive growth rates and activity levels, and thus the rate of metabolism supporting them.
Our hypothesis (Fig. 5) is applicable to not only field metabolic rates, but also resting metabolic rates observed in the laboratory. This is because the cost of growth contributes significantly to the resting metabolic rate of ectothermic animals with low maintenance requirements (Parry, 1983; Wieser, 1994; Glazier, 2015; Glazier et al., 2020). In addition, although we attempted to minimize activity during our respiration measurements, we could not eliminate it altogether, and thus low levels of spontaneous activity may have contributed somewhat to the metabolic rates that we estimated, a hypothesis that we are currently testing. Amphipods exhibit high levels of inter-individual variation in behavioral activity (Peeters et al., 2009; Augusiak and Van den Brink, 2015), which may have affected the metabolic rates that we measured. Accordingly, size-specific thermal effects on activity (and growth) may underlie the shifts in metabolic scaling that we have observed. In support, Glazier (2014c, 2020) has provided evidence that temperature and activity can interactively affect intraspecific metabolic scaling in ectothermic animals.
In short, we hypothesize that conservative metabolic rates showing low sensitivity to increases in temperature (as in large adults in springs with fish, and small juveniles in springs without fish, but with many cannibalistic adults) are adaptive evolutionary responses to size-specific mortality. Conservative energy expenditures may also be advantageous when predators inhibit foraging activity and thus energy intake (Wellborn, 1994; Handelsman et al., 2013; Glazier et al., 2020), as has been observed in amphipods exposed to fish predators or their chemical cues (e.g. Friberg et al., 1994; Åbjörnsson et al., 2000; Beermann et al., 2018).
Various processes related to resource uptake and use (rates of food-energy assimilation, growth and metabolism, and oxygen-uptake capacity, as indicated by gill surface area) show parallel differences in ontogenetic body-mass scaling between G. minus populations from springs with versus without fish predators (Glazier et al., 2011, 2020; Glazier and Paul, 2017). Multi-generational exposure of G. minus populations to fish predators has led to similar decreases in the scaling exponents of all of the above energetically significant traits. Therefore, further understanding may result from holistically examining the interactive effects of temperature and predation regime on the whole-organism energetic economy of spring-dwelling populations of G. minus.
Effects of fish predators on the metabolic scaling of amphipods may be the result of not only adaptive (genotypic) evolution, but also phenotypically plastic responses. It is possible that amphipods increase/decrease their metabolic rate physiologically as an anti-predator stress response (Hawlena and Schmitz, 2010), and/or in anticipation of increased/decreased activity (Glazier et al., 2020). In support of immediate phenotypic effects, Gjoni et al. (2020) have shown that fish kairomones (chemical cues) significantly affected the temperature-sensitive, ontogenetic body-mass metabolic scaling of two populations of amphipods in habitats containing fish predators, G. minus from Blue Spring, and Gammarus insensibilis from an Italian saltwater lagoon. This companion study has revealed that the metabolic scaling exponent (b) decreases at higher temperature in the presence of predator cues, but not in their absence. This result suggests that predation risk causes large conspicuous adults to resist temperature-related increases in RMR more than that of smaller less conspicuous juveniles, which parallels the apparent adaptive responses shown by G. minus populations from springs with fish predators, as reported in the present study. Currently, we are studying whether these phenotypically plastic effects depend on the history of exposure of amphipods to fish predators in their native habitats.
Relevance of ecological variation in intraspecific metabolic scaling to metabolic scaling theory
Most studies showing effects of ecological factors on metabolic scaling have examined intraspecific relationships, rather than interspecific relationships that are the chief focus of the MTE (but see Glazier, 2010, 2014c, 2018a; Pequeno et al., 2017). Brown et al. (2004) originally presented the MTE as a broad-scale view of biological scaling relationships based on body-mass ranges encompassing many orders of magnitude and many kinds of species (also see Sibly et al., 2012). Intraspecific scaling relationships often involve relatively narrow body-mass ranges that may be more prone to statistical error and the effects of various idiosyncratic extraneous factors than are broader scale interspecific relationships. Some investigators have further claimed that different factors or mechanisms affect intra- versus inter-specific metabolic scaling relationships (Heusner, 1982; Wieser, 1984; Kooijman, 2000; Maino and Kearney, 2014). Therefore, one might argue that studies of intraspecific metabolic scaling are not relevant to the MTE and the theory underlying it. However, we disagree with this view for four major reasons (also see Glazier, 2006). First, Kleiber (1961), a founder of the 3/4-power law upon which the MTE is based, regarded this law as applicable to both intra- and inter-specific metabolic scaling. Second, proponents of the MTE have argued that the RTN model underlying this theory (West et al., 1997) is universally applicable, applying to metabolic scaling relationships both within and among species (Brown et al., 1997). Indeed, growth models developed by proponents of the MTE assume that intraspecific metabolic scaling relationships obey RTN theory, as do interspecific relationships (West et al., 2001; Hou et al., 2008; Moses et al., 2008). Third, although intraspecific scaling relationships often include relatively narrow body-mass ranges encompassing less than an order of magnitude, thus increasing the error of their estimated slopes and intercepts (Bokma, 2004; Moses et al., 2008; Gudowska et al., 2017), many are based on large body-mass ranges encompassing multiple orders of magnitude (e.g. numerous kinds of animals and plants with indeterminate growth; Wesemeier, 1960; Paine, 1971; Larson, 1987; Bokma, 2004; Killen et al., 2007; Moran and Wells, 2007; Peng et al., 2010). Fourth, another major mechanism underlying the MTE, i.e. universal physical kinetic effects of temperature on the rates of various biological processes, should also operate similarly within and among species (Clarke, 2004; Gillooly et al., 2006; Dell et al., 2011).
Therefore, we believe that studies of intraspecific metabolic scaling are useful not only for specifically evaluating the MTE, but also for testing metabolic scaling theory in general. Important advantages of studies of intraspecific metabolic scaling include that they are less affected by phylogenetic complications that plague interspecific analyses because they involve single populations or comparisons of conspecific populations that are genetically more similar than are different species or higher taxa (Bokma, 2004; Glazier, 2005); and they are more readily amenable to incisive experimental tests and comparative analyses involving individuals or populations exposed to different biological or environmental conditions (Glazier, 2006; Vaca and White, 2010; Glazier et al., 2011).
Causes of high inter-individual variability in metabolic rate
Additional research is also required to understand why metabolic rate varied greatly among individual amphipods in our study, even after accounting for the effects of body mass, temperature and predation regime. Unmeasured intrinsic and extrinsic factors, such as individual differences in activity, growth and body condition, varying environmental effects occurring during the ontogeny of individuals in their natural habitats, and varying plasticity of individual responses to environmental manipulations in the laboratory may all have contributed to the high individual variation in metabolic rate that we observed. We suspect that variable levels of activity and growth are major causes of this variation (also see Peeters et al., 2009; Glazier et al., 2011, 2020; Glazier, 2020), but many other phenotypic and genotypic causes are also possible. Metabolic rate is a complex trait, composed of many underlying biochemical and physiological processes, involving numerous kinds of regulatory control and varying sensitivities to environmental factors (Darveau et al., 2002; Clarke, 2004; Suarez and Darveau, 2005; Glazier, 2005, 2014b; Pettersen et al., 2018). Therefore, it should not be surprising that metabolic rate varies substantially within not only amphipod species (also see the Results section), but also many other animal species (Speakman et al., 2004; Careau et al., 2008; Burton et al., 2011; Konarzewski and Książek, 2013; Norin and Metcalfe, 2019). The phenotypic plasticity of metabolic rate in response to environmental changes may also vary immensely among individuals (Norin and Metcalfe, 2019), as observed for temperature changes in amphipods (Réveillon et al., 2019) and insects (Terblanche et al., 2007). Given the central importance of metabolism for the functioning and evolutionary fitness of organisms, understanding its intraspecific variation is now becoming a major focus of research by many kinds of biologists (Careau et al., 2008; Burton et al., 2011; Konarzewski and Książek, 2013; Pettersen et al., 2018; Mathot et al., 2019; Norin and Metcalfe, 2019).
Our results add further support to the view that the body-mass scaling of metabolic rate is not simply the result of internal body-design constraints on resource supply, as assumed by the MTE, but is highly responsive to many kinds of external (ecological) factors that affect resource demand. The MTE assumes that resource supply dictates metabolic rate, which in turn acts as a pacemaker for other biological and ecological processes (Brown et al., 2004). In contrast, our results support an alternative (or complementary) view – an ecological theory of metabolism (ETM) – that focuses on how ecological factors influence various kinds of resource demand, which in turn modulate the rate of supporting metabolism and its scaling with body mass (Glazier, 2006, 2014b, 2015; Glazier et al., 2011; Harrison, 2017; Hatton et al., 2019). Both resource demand and supply may affect metabolic scaling (Glazier, 2014b, 2018b,c). This view is consistent with modern theory concerning the regulation of biochemical pathways, which posits that both resource supply and demand control the rates of metabolic reactions. Much evidence supports this view (reviewed in Glazier, 2015). In addition, our results add to growing evidence that specific ecological effects on metabolic scaling may themselves depend on other interacting biological and ecological factors. Abiotic environmental factors (e.g. temperature) may affect metabolic scaling interactively with not only other abiotic environmental factors (e.g. salinity: Dehnel, 1960; acidity: Carey and Sigwart, 2014; and exposure to air versus water: Newell et al., 1972) and intrinsic biological factors (e.g. nutrition: Newell et al., 1976; activity level: Glazier, 2014c, 2018a, 2020; and mode of thermoregulation: Glazier, 2018a), but also with biotic environmental factors (e.g. predation regime), as we show in this study (also see Gjoni et al., 2020).
Furthermore, our findings demonstrate that relationships between metabolic rate and temperature are not the simple result of biochemical kinetics, as assumed by the MTE (also see Schulte, 2015; DeLong et al., 2017; Moffett et al., 2018; Norin and Metcalfe, 2019; Réveillon et al., 2019). Predation regime has such a powerful effect on metabolic rate that it can significantly dampen its response to increasing temperature (as in adult G. minus from springs with fish predators: Fig. 2A, Fig. 3A–C and Fig. 5; and juveniles from springs with many cannibalistic adults made possible by the absence of fish: Fig. 2B, Fig. 3D–F and Fig. 5). Simple biochemical kinetic temperature effects also cannot explain why amphipods from springs with fish predators show significantly different responses to cooling (10 to 4°C) and warming (10 to 16°C) than do amphipods from springs without fish predators (see Results section and Fig. 2). In addition, other inter-population differences in temperature effects (including absent or even reversed effects where RMR decreased with increasing temperature: Figs 2 and 3) defy expectations based on biochemical kinetic models. Minimal or reversed effects of temperature on metabolic rate occur over various (not just upper) temperature ranges in other amphipod species (Issartel et al., 2005; Foucreau et al., 2014; Galic and Forbes, 2017; Semsar-kazerouni and Verberk, 2018; Kratina et al., 2019; Jermacz et al., 2020a) and many other kinds of animals, as well (Vernberg, 1959; Dehnel, 1960; Newell and Northcroft, 1967; Newell et al., 1976; Penick et al., 1998; Seibel et al., 2007; Marshall and McQuaid, 2011; Watson et al., 2014; Thyrring et al., 2015; Harianto et al., 2018; Cooper et al., 2019; Sargent, 2019; Scheffler et al., 2019; da Silva Vianna et al., 2020; and references cited therein). Whether these thermal responses are adaptive or stress related requires further research (Schulte, 2015; DeLong et al., 2017).
Ecological effects on the intraspecific scaling of metabolic rate with body mass and temperature deserve more consideration in theoretical models, most of which focus on the internal anatomy, physiology and biochemistry of organisms. In addition, our results have important implications for understanding the effects of climate (T) change on biological and ecological systems. Biological effects of global warming may not be simple emergent functions of biochemical kinetics, but may depend critically on the ecological context. As a result, predicting biological and ecological effects of global warming becomes complicated (also see Tseng and O'Connor, 2015; Lau and terHorst, 2020; Luhring et al., 2019; Tseng et al., 2019; Truong et al., 2020). As shown in our study, warming may increase the metabolic rates of small juveniles more than that of large adult amphipods in springs with fish predators, but do the opposite in springs without fish. These habitat-related differences in how warming affects the metabolic rate (and associated biological processes) of an abundant keystone species may have important consequences for the dynamics and structure of populations and communities in springs and other aquatic ecosystems. Our results and the associated hypothetical explanation that we have proposed allow us to make three testable predictions. First, warming should accentuate harmful direct and indirect effects of fish predators (i.e. increased mortality and risk-reducing inhibition of feeding, growth and metabolism) on large adult amphipod prey in springs with fish predators, but in contrast accentuate harmful effects of cannibalistic adult amphipods on small juveniles in springs without fish predators. Second, warming should affect amphipod population dynamics and structure differently in habitats with versus without fish predators. For example, warming should induce higher rates of juvenile growth and maturation (and thus higher rates of production of new breeding adults) in populations inhabiting springs with versus without fish predators. In addition, warming should cause the size-structure of amphipod populations to become more skewed toward smaller body sizes in springs with versus without fish predators. Third, warming should increase litter decomposition (detrital leaf shredding) rates by adult amphipods (and thus rates of nutrient cycling and loss of coarse detrital food) more in springs without versus with fish predators. As a result, greater competition for coarse particulate detrital food in detritivore communities, greater challenges to maintaining energy balance in amphipods and other shredding detritivores, and greater availability of dissolved nutrients and fine particulate food for other biota should occur in springs without versus with fish predators. However, these predicted outcomes may be altered by other ecological factors not considered here. As we have emphasized, the biological effects of warming may depend critically on the ecological context.
The authors thank Alexander B. Weimer, Shaina Robinett, Natalie E. Donofrio and Jacob M. Owings for providing field and laboratory assistance.
Conceptualization: D.S.G., V.G.; Methodology: D.S.G., V.G.; Formal analysis: D.S.G., J.P.G., J.R.H., V.G.; Investigation: D.S.G., J.P.G., J.R.H., V.G.; Resources: D.S.G.; Data curation: D.S.G.; Writing - original draft: D.S.G.; Writing - review & editing: D.S.G., J.P.G., J.R.H., V.G.; Visualization: D.S.G.; Supervision: D.S.G.; Project administration: D.S.G.; Funding acquisition: D.S.G., V.G.
This study was supported by the Kresge Foundation, Klag-Meoni Endowment and the Italian Society of Ecology (SItE), including the Luigi and Francesca Brusarosco Award to V.G.
The authors declare no competing or financial interests.