Metabolic cold adaptation in fishes occurs at the level of whole animal, mitochondria and enzyme

Abstract

Metabolic cold adaptation (MCA), the hypothesis that species from cold climates have relatively higher metabolic rates than those from warm climates, was first proposed nearly 100 years ago and remains one of the most controversial hypotheses in physiological ecology. In the present study, we test the MCA hypothesis in fishes at the level of whole animal, mitochondria and enzyme. In support of the MCA hypothesis, we find that when normalized to a common temperature, species with ranges that extend to high latitude (cooler climates) have high aerobic enzyme (citrate synthase) activity, high rates of mitochondrial respiration and high standard metabolic rates. Metabolic compensation for the global temperature gradient is not complete however, so when measured at their habitat temperature species from high latitude have lower absolute rates of metabolism than species from low latitudes. Evolutionary adaptation and thermal plasticity are therefore insufficient to completely overcome the acute thermodynamic effects of temperature, at least in fishes.

1. Introduction

Understanding the factors that influence the metabolic rate (MR) of animals has been a key focus of animal physiology for over 100 years [1,2]. Much effort has been expended on describing the allometric scaling of MR with body mass [2–5] and on attempting to explain why MR scales allometrically [5,6], but the relationship between MR and body mass conceals considerable variation, and species of similar mass can vary in MR by more than an order of magnitude [4]. That species vary by so much suggests that such variation is important, and inter-individual and interspecific differences in MR are correlated with differences in juvenile survival [7], starvation resistance [8], reproductive output [9] and a range of other life-history traits [10–12].

Despite the clear importance of MR, however, the ultimate factors that drive its evolution remain incompletely understood. Insects and endotherms from cold environments have been shown to have relatively high MRs [13–16], a phenomenon known as metabolic cold adaption (MCA) [17,18] or Krogh's rule [19]. MCA is an example of countergradient variation, a geographical pattern of genotypes where genetic influences on a trait oppose environmental influences, thereby minimizing phenotypic patterns along a gradient [20]. In ectotherms, the negative relationship between MR and the temperature of the environment in which an animal lives acts to partially or wholly compensate for the positive thermodynamic effect of temperature on MR. In contrast to the situation in terrestrial animals, for aquatic animals MCA has variously been described as a myth [21], an experimental artefact [18,22], and one of the most controversial hypotheses in physiological ecology [23,24].

For fishes, interspecific studies of MR show mixed support for MCA [17,18,22,23,25,26], while studies at other levels of organization have revealed patterns consistent with MCA. The activities of enzymes associated with energy metabolism are typically higher in species from polar and temperate environments than in those from tropical environments [27,28], and mitochondrial volume density is relatively high in the muscles of Antarctic fishes [29,30]. Given that mitochondrial proton leak contributes a significant proportion of resting MR [31,32], it seems reasonable to predict that MR should be higher in cold-adapted animals because of their high muscle mitochondrial densities [33], making it surprising that such a pattern is not always documented.

In the present study, we adopt an integrative approach to re-evaluate the MCA hypothesis in fishes at multiple levels of biological organization. In doing so, we address several criticisms of previous tests of MCA [22,23]. We selected data according to rigorously defined criteria and account for methodological differences between studies, and, where appropriate, account for the phylogenetic non-independence of closely related species [34–37]. We test the MCA hypothesis by relating the highest latitudinal extreme of a species' range to standard metabolic rate (SMR), the activity of isolated mitochondria, and the activity of a key metabolic enzyme (citrate synthase, which controls flux through the citric acid cycle).

2. Methods

Data for body mass (g), rate of oxygen consumption (V̇O₂, ml min⁻¹, an indirect measure of MR, n = 37 species; electronic supplementary material, table S1), white muscle citrate synthase activity (µmol g⁻¹ min⁻¹, n = 54 measurements of 49 species; electronic supplementary material, table S2), and state III rate of respiration of muscle mitochondria using pyruvate as a substrate or heart mitochondria using pyruvate and malate as a substrate (µmol g⁻¹ min⁻¹, n = 24 measurements of 23 species; electronic supplementary material, table S3) were obtained from a range of published sources (see the electronic supplementary material). Data were included only if measured at a known temperature.

For V̇O₂, care was taken to include data conforming to the conditions specified for the measurement of SMR (all animals were post-absorptive and non-reproductive). Prior to measurement of SMR, animals were fasted for a median of 3 days (range: 3 h to 14 days) and acclimated to a known temperature, usually close to the temperature at which they were measured, for a median of 14 days (range 1 day to 1 year). These fast and acclimation durations were considered appropriate by the authors of the primary data, and we test for a relationship between SMR and fast duration or acclimation duration to ensure that short durations do not compromise the analyses (see §3). Where data for a range of temperatures were provided, a value close to habitat temperature was selected if possible, or a value at the mid-point of the temperature range. Data for anaesthetized, restrained or surgically instrumented individuals were excluded, as were SMR data estimated by extrapolating a relationship between swimming speed and V̇O₂. V̇O₂ data measured using closed-box respirometry were excluded, because the measurements may overestimate V̇O₂ [22,38]. The temperature at which fishes were maintained prior to SMR measurement was provided for all species, and was strongly correlated with the temperature at which they were measured (r > 0.99, mean absolute difference between measurement and acclimation temperatures = 0.2°C, range 0–1.5°C). For each species, the highest latitudinal extreme of the range was determined using data from FishBase [39]. Salinity was scored on an ordinal scale (freshwater = 0, brackish = 1, marine = 2) with fishes assigned to multiple categories in FishBase given the mean category value. The environment usually occupied by each species was also noted (bathypelagic, bathydemersal, benthopelagic, demersal, reef-associated or pelagic), as was the hemisphere into which their range extends the furthest (Northern or Southern). The coarse relationship between latitude and the temperature to which animals are exposed may limit the reliability of conclusions about MCA based upon the relationship between SMR and latitude owing to variation in temperature between habitats at the same latitude [40]. However, in a broad-scale comparison, such as the present study, it is important only that, on average, higher latitudes are colder than warm ones, not that latitude is a precise estimate of temperature [41].

Where data for SMR were available for individuals of a range of body sizes of a single species, only data for the largest were included to minimize the influence of growth on SMR. Multiple measurements of mitochondrial respiration or citrate synthase activity for a single species were included by placing them in very short branches (0.001) [42]. Where data for a given trait were available for a range of temperatures, temperature dependence was determined as Q₁₀ = e^10b, where b is the slope of the line relating ln(Trait) and temperature. A mean Q₁₀ value for each trait was then calculated (mitochondrial respiration: 2.2 ± 0.5 (s.d.), n = 14; Q₁₀ for citrate synthase was assumed to be 2 and that of SMR to be 2.4 [23]). Data for SMR and mitochondrial respiration were then normalized to a temperature of 20°C and data for citrate synthase activity were normalized to a temperature of 10°C (most measurements of citrate synthase activity were made at this temperature), using the appropriate value of Q₁₀. SMR at temperature 1 (T₁, 10 or 20°C) was calculated from SMR at measurement temperature (T_m) according to:

This temperature normalization procedure has been used previously in other comparative studies of MCA [13], and, in the present study, was preferred to a multiple regression approach, including both measurement temperature and absolute latitude as predictors of log(trait), because fishes tend to be acclimated to, and measured at, temperatures broadly representative of their habitat temperature, so measurement temperature and latitude are strongly correlated (SMR: r = −0.85, p < 0.0001; mitochondrial respiration: r = −0.77, p < 0.0001). Such collinearity can result in spurious conclusions about the relationship between the dependent and independent variables, because the partial regression coefficients associated with the independent variables may not be representative of the relationship that exists in the population [43].

Data were analysed using phylogenetic generalized least squares (PGLS) [44–46] in the Analysis of Phylogenetics and Evolution package within R [47,48] according to established procedures and using published code [49,50]. An appropriate tree is not available for the species in the present study, so informal supertrees were constructed using data from established sources (see the electronic supplementary material). Because the branch lengths in the informal supertrees are unknown, a range of branch length transformations were compared: star (non-phylogenetic), punctuated (all branches set equal and equal to 1), Grafen's [44], Nee's [51] and Pagel's [52]. For each of these models, a measure of phylogenetic correlation, λ [36,53], was estimated by fitting PGLS models with different values of λ and finding the value that maximizes the log likelihood. The degree to which trait evolution deviates from Brownian motion (λ = 1) was accommodated by modifying the covariance matrix using the maximum-likelihood value of λ, which is a multiplier of the off-diagonal elements of the covariance matrix (i.e. those quantifying the degree of relatedness between species). Alternative models were compared on the basis of Akaike's information criterion (AIC) as a measure of model fit [54,55]. The relative support of alternative models was compared on the basis of its Akaike weight (w_i, the relative-likelihood of the model compared with all others: the likelihood of the model divided by the sum of the likelihoods of all other models) and Δ_i (i.e. AIC – minimum AIC; models having Δ_i ≤ 2 have substantial support, those where 4 ≤ Δ_i ≤ 7 have considerably less support, while models having Δ_i > 10 have essentially no support [54]).

For all traits, the approach taken was to compare statistical models that incorporated absolute latitude as a predictor with those that did not. All traits were log-transformed for analysis. Log(body mass) was included as a covariate for SMR, but not for mitochondrial respiration and citrase synthase activity. For those species for which body mass data were available, log(body mass) was not a significant predictor of log(mitochondrial respiration) or log(citrate synthase activity; p > 0.05 in both cases). For citrate synthase activity, log-transformed maximum depth of occurrence was also included as a covariate [56]. For mitochondrial activity, tissue type (muscle or heart) was included as a fixed factor.

3. Results

(a). SMR

The best-fit model for Q₁₀-normalized SMR included body mass and absolute latitude (table 1; electronic supplementary material, table S4) and modelled evolution on a tree with Nee's arbitrary branch lengths and λ set equal to 1 (table 2; electronic supplementary material, table S4). The scaling exponent of SMR was 0.80, similar to other values for fishes [4,57], and SMR was positively related to latitude (table 3 and figure 1a). Models excluding latitude were less well supported than equivalent phylogenetic or non-phylogenetic models, including latitude; the change in AIC (Δ_i) associated with removing latitude from a given model was always greater than 4. AIC changed by less than 2 when acclimation temperature (Δ_i = −1.3), log-transformed acclimation duration (Δ_i = −0.7), acclimation temperature + log-transformed acclimation duration (Δ_i = 0.3), log-transformed fast duration (Δ_i = 1.5), measurement method (intermittent or continuous flow, Δ_i = −0.7) and salinity (Δ_i = 1.2) were added to the best (evolutionary + statistical) model. Adding environment increased AIC (Δ_i = 7.5), whereas including hemisphere of origin (Northern or Southern Hemisphere) decreased AIC (Δ_i = −3.0; Southern Hemisphere species had a lower SMR than Northern Hemisphere species; table 3) but the effect appeared to be driven by low SMRs in just two Southern Hemisphere species (figure 1a): short-finned eels Anguilla australis [58] and spotted stargazers Genyagnus monopterygius [59]. Including hemisphere of origin and an interaction between latitude and hemisphere of origin increased AIC (Δ_i = −1.1).

Table 1.

Akaike weights for statistical models for standard metabolic rate (SMR), mitochondrial respiration and citrate synthase activity, summed over all evolutionary models.

statistical model	Σwi
Q10-normalized SMRa
logM	0.01
logM + latitude	0.99
SMR
logM + temp	0.59
logM + temp + latitude	0.41
Q10-normalized mitochondrial respiration
intercept only	0.11
tissue	0.12
latitude	0.31
tissue + latitude	0.46
mitochondrial respiration
temp	0.39
temp + tissue	0.32
temp + latitude	0.15
temp + tissue + latitude	0.14
citrate synthase activity
log(depth)	0.001
latitude	0.001
log(depth) + latitude	0.10
log(depth) + latitude + environment	0.89

^aNote that AIC decreased further when hemispere of origin was added to the best single model for Q₁₀-normalized SMR (see text for details).

Table 2.

Values of Akaike's information criterion (AIC) for evolutionary models (with λ = 0,1, or taking the maximum likelihood (ML) value) for Q₁₀-normalized standard metabolic rate (SMR), Q₁₀-normalized mitochondrial respiration, and Q₁₀-normalized citrate synthase (CS) activity. (Evolutionary models are for the best-fit statistical models from tables 1 and 3.)

evolutionary model	SMR		mitochondrial respiration		CS activity
	λ	AIC	λ	AIC	λ	AIC
star	0	−2.98	0	6.01	0	−6.5
Grafen	ML = 0	−0.98	ML = 0	8.01	ML = 0.39	−6.0
	1	6.27	1	32.9	1	18
Nee	ML = 1	−2.33	ML = 0	8.01	ML = 0.40	−7.5
	1	−4.33	1	17.0	1	−3.5
Pagel	ML = 0.97	−2.03	ML = 0	8.01	ML = 0.37	−6.0
	1	−3.97	1	24.6	1	0.1
punctuated	ML = 0	−0.98	ML = 1	7.04	ML = 0	−4.5
	1	−2.61	1	5.04	1	4.2

Table 3.

Parameter estimates for the best-fit (statistical + evolutionary) models for standard metabolic rate (SMR), mitochondrial respiration and citrate synthase activity.

term	estimate	s.e.
Q10-normalized SMR
intercept	−1.14	0.17
logM	0.80	0.03
latitude	0.008	0.002
hemisphere: southern	−0.22	0.10
SMRa
intercept	−1.03	0.18
logM	0.78	0.03
temperatureb	0.016	0.005
Q10-normalized mitochondrial respiration
intercept	1.56	0.22
latitude	0.006	0.003
tissue: red muscle	−0.19	0.11
mitochondrial respiration
intercept	1.49	0.08
temperature	0.017	0.005
citrate synthase activity
intercept	−0.30	0.17
log(depth)	−0.29	0.06
latitude	0.015	0.003
environment: pelagic	0.28	0.10
environment: demersal	−0.12	0.08

^aThe best (statistical + evolutionary) model for SMR modelled evolution on a tree with all branch lengths set equal to 1, with λ set equal to 1 (w_i = 0.16).

^bA parameter estimate of 0.016 for the temperature dependence of log-transformed SMR is equivalent to an across species value of Q₁₀ of 1.4 (95% confidence interval: 1.11–1.85).

Figure 1.

Relationships between absolute latitude and (a) standard metabolic rate (SMR), (b) mitochondrial respiration, and (c) citrate synthase activity. Data are normalized to measurement temperatures of 20°C (SMR and mitochondrial respiration) or 10°C (citrate synthase activity). Solid lines are the associations between the various traits and latitude from table 3. Because Q₁₀-normalized SMR is positively related to body mass (table 3), and citrate synthase activity is positively related to maximum depth of occurrence (table 3), residuals are presented (i.e. measured – predicted). The best model for SMR included hemisphere of origin (table 3), so SMR data are shown for Northern Hemisphere (filled circles) and Southern Hemisphere (unfilled circles).

If the collinearity of latitude and measurement temperature was ignored and models including both temperature and latitude considered as predictors of non-Q₁₀-normalized SMR, latitude was not included in the best model for SMR (tables 1 and 3; electronic supplementary material, table S5).

(b). Mitochondrial respiration

The best-fit model for Q₁₀-normalized mitochondrial respiration modelled evolution on a phylogeny with all branch lengths equal to 1, with λ set at 1 (table 2; electronic supplementary material, table S6), and included latitude and tissue type as predictors (table 1; electronic supplementary material, table S6). The relationship between latitude and Q₁₀-normalized mitochondrial respiration was positive (table 3 and figure 1b). Removing latitude from a given model always increased AIC (Δ_i = 0.9–4.8). Removing latitude from the best model increased AIC by 1.8. AIC increased when salinity (Δ_i = 6.1) or environment (Δ_i = 8.5) were added to the best (evolutionary + statistical) model for Q₁₀-normalized mitochondrial respiration. If the collinearity of latitude and measurement temperature was ignored and models including both temperature and latitude considered as predictors of non-Q₁₀-normalized mitochondrial respiration, the best model included only measurement temperature and modelled evolution on a star phylogeny (tables 1 and 3; electronic supplementary material, table S7). AIC increased when environment was added to the best (evolutionary + statistical) model for non-Q₁₀-normalized mitochondrial respiration (Δ_i = 3.7), when hemisphere of origin was added to the best model (Δ_i = 0.7), and when hemisphere of origin and an interaction between hemisphere of origin and latitude was added (Δ_i = 7.6), but decreased when salinity was added (Δ_i = −0.6), although the effect of salinity was not significant (p = 0.13).

(c). Citrate synthase activity

Summed over all evolutionary models, the best statistical model for Q₁₀-normalized citrate synthase activity included log-transformed maximum depth of occurrence, latitude and environment (table 1; electronic supplementary material, table S8). To determine which of the environmental categories explained variance in Q₁₀-normalized citrate synthase activity, the categories were coded as dummy variables and added stepwise to a model including log(depth) and latitude. Only those categories that decreased AIC were retained in the final model, which modelled evolution on a tree with Nee's arbitrary branch lengths with maximum-likelihood λ = 0.35 (table 3). The relationship between latitude and citrate synthase activity was positive (table 3 and figure 1c) and that between citrate synthase activity and log(depth) was negative (table 3). Pelagic species had relatively high citrate synthase activities, whereas demersal species had relatively low citrate synthase activities (table 3); other species were intermediate. Adding salinity to the best model increased AIC (Δ_i = 1.9), as did adding hemisphere of origin (Δ_i = 1.7); adding hemisphere of origin and an interaction between hemisphere of origin and latitude decreased AIC (Δ_i = −0.2) and the terms themselves were not significant (p > 0.05).

4. Discussion

The present study supports the MCA hypothesis at multiple levels of biological organization in fishes. The SMR of fishes from high latitudes is higher than that of fishes from low latitudes (figure 1), and this relationship is matched with positive relationships between absolute latitude and mitochondrial respiration (figure 1b), and a positive relationship between latitude and the activity of a key metabolic enzyme (figure 1c). Support for MCA is strongest for SMR, because removing latitude as a predictor of SMR increased AIC by more than 4 over equivalent phylogenetic and non-phylogenetic models including latitude, indicating that models without latitude have considerably less support than those that include latitude [54]. Removing latitude from the best model for SMR increased AIC by 10, indicating that the model without latitude has essentially no support relative to the equivalent model that included latitude. Similarly, removing latitude as a predictor of citrate synthase activity increased AIC by at least 5, indicating that models without latitude have considerably less support than those including latitude [54]. Support for a relationship between latitude and mitochondrial respiration was more equivocal. Summed over all evolutionary models, the best statistical model for Q₁₀-normalized mitochondrial respiration included latitude as a predictor (table 1), but excluding latitude from the best single (statistical + evolutionary) model (table 3) increased AIC by only 1.8, indicating that both models are well supported (the model with latitude is 2.5 times more likely to be a better model than the equivalent without latitude). In fishes, mitochondrial volume density varies with water temperature [60,61], and thermal acclimation can induce changes in the cristae surface density of mitochondria as well as mitochondrial volume density of muscle [62,63]. Thus, changes in tissue ultrastructure may be more important evolutionary responses to cold than changes in mitochondrial respiration. Data for rates of mitochondrial respiration, however, are relatively scarce for tropical fishes restricted to low latitudes (figure 1b), and data for additional species from these regions would be valuable in further investigating the relationship between latitude and mitochondrial respiration.

In addition to the relationships between latitude and SMR, mitochondrial respiration and citrate synthase activity presented here (figure 1), MCA in fishes is also supported by the negative relationship between mitochondrial density and water temperature [60,61]. These findings are consistent with countergradient variation in metabolism, and demonstrate that the geographical patterns of these traits match the significant countergradient variation that has been observed in a range of other traits in fishes, including growth rate [64,65], coloration [66,67], body shape [68] and digestive performance [69]. However, acclimation to cold can also cause increases in SMR [26,70], mitochondrial respiration [71] and citrate synthase activity [72]. Thus, the present study is unable to distinguish between genetic and phenotypic associations between latitude and SMR, mitochondrial respiration and citrate synthase activity because high latitude species were acclimated to, and measured at, low temperatures, and low latitude species acclimated to, and measured at, high temperatures.

The present study is unable to establish why SMR, mitochondrial respiration and citrate synthase activity are associated with latitude, because latitude is not a cause of variation in these traits per se; latitude merely acts as a proxy for the environmental variables that do. For example, temperature is on average lower at the poles than the equator, but climate variability is also associated with latitude and has been proposed as a driver of metabolic adaptation in fishes [73], mammals [14,74] and birds [15]. The best evidence for a causal association between temperature and metabolic traits comes from studies of laboratory evolution, but such studies are rare (see Garland & Rose [75] for a review of experimental evolution). For example, lines of Drosophila melanogaster allowed to evolve in the laboratory for over 100 generations at low temperature have higher MRs than lines allowed to evolve at high temperature, suggesting that the association between MR and latitude demonstrated by common garden experiments in D. melanogaster arises as a consequence of an evolutionary increase in MR for populations adapted to low temperature ([76], see also [77] for a discussion of potential confounds between water stress and high temperature).

The present study identifies a significant negative association between citrate synthase activity and maximum depth of occurrence, and differences in citrate synthase activities between pelagic and demersal fish species (table 3). Such relationships have been identified previously for citrate synthase activity [56,78–80], and are associated with differences in locomotor activity and aerobic capacity between pelagic, demersal and deep-sea species. Given the association between citrate synthase activity and environment, it is perhaps surprising that there was no association between environment and SMR, or between environment and mitochondrial respiration. However, few data that met our criteria are available for pelagic species (two species for SMR and three for mitochondrial respiration), so the lack of an effect might arise as a consequence of low statistical power.

The findings of the present study contradict a number of recent studies that reject the MCA hypothesis for fishes [22,23]. For example, Clarke & Johnston [23] concluded that evolutionary adaptation reduced the overall thermal sensitivity of resting metabolism across species, but rejected the MCA hypothesis because polar notothenioids (Perciformes) have resting MRs that are not detectably different from those predicted from the relationship between MR and temperature of non-polar perciforme fishes [23]. We argue, however, that the reduced thermal sensitivity of metabolism across species reported by Clarke & Johnston [23] and others [4], including the present study (table 3), actually represents compelling evidence in favour of MCA. When assessed across species, the reduced temperature sensitivity occurs because fishes tend to be measured at temperatures broadly representative of the temperatures at which they occur, and because fishes from cold climates have relatively high MRs, whereas those from warm climates have relatively low MRs. Thus, because of MCA, the value of Q₁₀ calculated for interspecific data is lower than that calculated from the intraspecific relationship between temperature and MR observed for eurythermal species.

A consequence of the relatively low thermal sensitivity of metabolism across species is that metabolic compensation for the global temperature gradient is not complete. Thus, despite species from high latitudes having high MRs relative to measurement temperature and body size, when measured at their habitat temperature species from high latitudes have lower absolute rates of metabolism than species from low latitudes. This indicates that evolutionary adaptation and thermal plasticity compensate for, but are insufficient to, completely overcome the acute thermodynamic effects of temperature, at least in fishes.