Interindividual variation in maximum aerobic metabolism varies with gill morphology and myocardial bioenergetics in Gulf killifish

Animal metabolic rates are complex phenotypes that reflect the flow of energy and materials through organisms, and they are essential features of animal physiology and ecology (Brown et al., 2004; Schmidt-Nielsen, 1997). Because most animals fuel the majority of their metabolism through the aerobic breakdown of macronutrients, the rate of oxygen consumption (Ṁ_O2) is a standard and reliable measure of metabolic rate (Schmidt-Nielsen, 1997). Much has been learned about the factors that influence Ṁ_O2 from studies of diverse species, populations and the effects of acclimation (McNab, 2002; Glazier, 2005; Schmidt-Nielsen, 1997). However, even when controlling ambient conditions and factors such as body size, activity, nutrition and reproductive status, the variation in Ṁ_O2 among individuals within a species or population is still substantial. In general, intraspecific variation in aerobic metabolism is repeatable, heritable and related to fitness. Because variation among individuals within a species reflects the raw material upon which natural selection acts, considerable attention has been given to understanding the proximate (mechanistic) and ultimate (evolutionary) causes of this variation (Bouchard et al., 1998; Burton et al., 2011; Careau and Garland, 2012; Crawford and Oleksiak, 2007; Glazier, 2005; Hoppeler, 2018; Killen et al., 2016; Konarzewski and Książek, 2013; Nespolo and Franco, 2007; Pettersen et al., 2018; Scott and Dalziel, 2021; White and Kearney, 2013).

When evaluating the causes and consequences of variation in aerobic metabolism, distinguishing between differences in metabolic rates due to activity is important. The upper limit of Ṁ_O2 by an animal is reflected in its maximum metabolic rate (MMR), typically measured during or immediately after vigorous exercise. During sustained aerobic exercise, oxygen utilization by tissues must be matched by oxygen uptake by the respiratory system and oxygen delivery by the circulatory system. Thus, studies of proximate causes of variation in MMR have examined various steps in the O₂ transport pathway from the environment to metabolically active tissues, where oxygen is ultimately consumed in mitochondrial oxidative phosphorylation. A large body of research has shown that the diffusing capacity of lungs or gills, cardiac output and the diffusing capacity of skeletal muscle vary among and within species as a function of body mass, ‘athleticism’, acclimatization or training, and habitat in a fashion that correlates with active Ṁ_O2 (Bassett and Howley, 2000; de Jager and Dekkers, 1975; Duncan, 2020; Glazier, 2005; Hillman et al., 2013; Hoppeler, 2018; Hughes, 1966; Killen et al., 2016; Scott and Dalziel, 2021; Wagner, 1996; Weibel et al., 2004, 1991; Weibel and Hoppeler, 2005). At the lower end of the Ṁ_O2 spectrum is that of a resting, post-absorptive, non-reproductive animal, known as the standard metabolic rate (SMR) for an ectothermic animal at a specified ambient temperature or the basal metabolic rate (BMR) for an endotherm within its thermoneutral zone (McNab, 1997). The Ṁ_O2 under these conditions is much less than MMR and, accordingly, the O₂ transport pathway is not likely to be limiting (Weibel et al., 1991). Rather, interspecific and intraspecific variation in SMR and BMR has been attributed to the mass or Ṁ_O2 of organs that are active in animals at rest (e.g. liver, brain, kidneys and heart) or constitute a large proportion of the body mass (e.g. skeletal muscle) (Burton et al., 2011; Killen et al., 2016; Konarzewski and Książek, 2013; Rolfe and Brown, 1997; White and Kearney, 2013).

The goal of the present study was to investigate the relationships between MMR and SMR and morphological and physiological variables related to the O₂ transport cascade in the Gulf killifish, Fundulus grandis. Fundulus grandis is a small, ecologically dominant, teleost fish of the salt marshes of the Gulf of Mexico (Nordlie, 2006). Previous research on this species has demonstrated significant, repeatable, mass-independent variation in MMR and SMR among individuals held and measured under common garden conditions (Reemeyer and Rees, 2020; Virani and Rees, 2000). Our first objective was to determine whether individual variation in MMR is related to the initial steps of the O₂ transport cascade: the capacity for O₂ uptake and distribution. Although fish are the most speciose group of vertebrate animals and demonstrate marked variation in aerobic metabolism (Killen et al., 2016; Norin and Clark, 2016), the O₂ transport cascade in this group has received less attention than in terrestrial, air-breathing vertebrates (but see Scott and Dalziel, 2021; Steinhausen et al., 2008).

For water-breathing fish, the first step in the O₂ transport cascade is oxygen diffusion across the epithelium of the gill lamellae. The area available for gas exchange depends upon the area of individual lamellae, the density of lamellae on gill filaments, and the length and the total number of gill filaments projecting from the gill arches (Hughes, 1984). Interspecific comparisons have shown that total gill surface area tends to be greater in more active species than in less active species (Hughes, 1966; Killen et al., 2016; Palzenberger and Pohla, 1992). However, within species, the association between gill morphology and the capacity for aerobic metabolism is less clear. In the mummichog, Fundulus heteroclitus, population variation in gill surface area is positively correlated with the routine metabolic rate (RMR), a level of metabolism between SMR and MMR (McBryan et al., 2016). In laboratory-reared offspring from stream and marine populations of three-spined stickleback, corresponding to low and high MMR populations, respectively, the total gill filament length was greater in offspring from the high MMR marine population, although the difference in gill surface area was not significant (Dalziel et al., 2012). Conversely, more active, limnetic ecotypes of lake whitefish were found to have fewer gill filaments than less active, benthic ecotypes, just the opposite of what would be predicted if gill surface area was limiting aerobic metabolism (Laporte et al., 2016). These conflicting results could arise because fish gills serve multiple functions (Evans, 1980) and respond morphologically to diverse selective pressures (Chapman et al., 2008; Laporte et al., 2016). Importantly, the aforementioned studies evaluated trait differentiation among populations to infer the consequences of past evolution rather than testing whether variation in gill morphology among animals with similar evolutionary history affects their MMR. Here, we measured gill filament number, average filament length and total filament length as proxies of gill surface area (Chapman et al., 2008; Hughes, 1984; Palzenberger and Pohla, 1992). We predicted that these measures, either singly or together, would be positively correlated with MMR in F. grandis from a given location and held under similar laboratory conditions.

Once oxygen is taken up across the gills, it is circulated to the tissues by the heart and vascular system. According to the Fick principle, whole-animal Ṁ_O2 should increase with increasing cardiac output (the product of stroke volume and heart rate) for a given arterio-venous oxygen difference. Indeed, increased Ṁ_O2 of Atlantic cod during exercise is linearly related to cardiac output (Webber et al., 1998). Moreover, intraspecific variation in the capacity to elevate Ṁ_O2 during swimming in rainbow trout is associated with greater cardiac output in vivo and greater power generation by hearts in vitro (Claireaux et al., 2005). Because cardiac power generation requires ATP and is directly correlated with Ṁ_O2 by the heart itself (Farrell and Smith, 2017), we hypothesized that variation in the capacity for cardiac mitochondrial metabolism should be reflected by variation in MMR. This expectation is partly supported by measurements of enzyme activity in fish hearts. In Atlantic cod, intraspecific variation in Ṁ_O2 during swimming is positively correlated with the ventricular activity of cytochrome c oxidase, the terminal enzyme of the electron transport system (Sylvestre et al., 2007). In contrast, variation in aerobic activity among populations of three-spined stickleback or ecotypes of lake whitefish was not paralleled by changes in the ventricular activity of cytochrome c oxidase and citrate synthase, a marker of citric acid cycle activity (Dalziel et al., 2012, 2015). To directly test the hypothesis that cardiac mitochondrial metabolism is correlated with MMR within a population, we used high-resolution respirometry (HRR) to measure mitochondrial Ṁ_O2 by permeabilized myocardial tissue (Doerrier et al., 2018; Kuznetsov et al., 2008; Pesta and Gnaiger, 2012; Veksler et al., 1987). By the controlled addition of substrates and inhibitors, HRR affords the opportunity to probe various steps of mitochondrial oxidative phosphorylation in preparations that preserve some cellular structure. Moreover, HRR can differentiate Ṁ_O2 due to ATP synthesis by oxidative phosphorylation (OXPHOS) from that required to offset the dissipation of the transmembrane proton gradient in the absence of ATP synthesis (LEAK). We predicted that Ṁ_O2 coupled to ATP synthesis by permeabilized myocardial tissue would be positively related to MMR in F. grandis from a given location and held under similar laboratory conditions.

The second objective of this study was to determine whether individual variation in SMR was correlated with these same gill and heart variables. Although the O₂ transport cascade is not expected to limit oxygen uptake by animals at rest (see above), several observations suggest that traits related to O₂ uptake and transport are associated with resting Ṁ_O2. Gillooly et al. (2016) demonstrated a strong positive relationship between the respiratory diffusing capacity of lungs and gills and RMR or BMR across a broad range of vertebrates. Also, as mentioned above, in populations of F. heteroclitus, gill surface area is positively associated with RMR, a measure of metabolism that is close to SMR, but includes some low, indeterminant level of activity. Furthermore, Jayasundara et al. (2015) showed that Ṁ_O2 by intact hearts was correlated with RMR across five species, and Drown et al. (2021) recently showed that the Ṁ_O2 of whole hearts is positively correlated with SMR in F. heteroclitus, albeit under specific acclimation conditions and only with certain metabolic substrates. Also, LEAK respiration may contribute to interindividual variation in BMR in mammals or SMR in fish (Rolfe and Brown, 1997; Salin et al., 2016; but see Larsen et al., 2011). While this effect is likely to be most relevant for large organs (e.g. liver and skeletal muscle), the contribution of cardiac LEAK respiration to whole-animal Ṁ_O2 is unknown. Because the predicted outcomes of this second objective were less certain, we adopted the null hypothesis (no relationship) for this objective.

Gulf killifish, Fundulus grandis Baird and Girard 1853 (n=39; mean mass 5.71 g, range 3.73–9.51 g at the time of experiments), were collected from the Grand Bay National Estuarine Research Reserve in August 2018. These fish were maintained and used between August 2018 and May 2019, as described in Reemeyer and Rees (2020), after which they were held for an additional 6 weeks on a 12 h:12 h light:dark photoperiod in filtered, aerated and dechlorinated municipal water made to a salinity of 9–12 using artificial sea salt (Instant Ocean) at 25°C. Fish were fed dried flake food (TetraMin) once per day. Food was withheld for 24 h before whole-animal respirometry or euthanasia for tissue sampling. All procedures on live animals were approved by the University of New Orleans Institutional Animal Care and Use Committee.

Fish used in this study were previously subjected to intermittent-flow respirometry (Svendsen et al., 2016), as described in Reemeyer and Rees (2020). For the current study, the same procedures were repeated to determine MMR and SMR within 2 weeks of euthanasia for tissue harvesting (see below). To induce MMR, fish were chased individually by hand to exhaustion (3 min) and immediately transferred to the respirometer (<30 s). Consistent with previous observations (Reemeyer and Rees, 2020), the highest Ṁ_O2 for many fish occurred several hours after chasing. For only 8 fish was Ṁ_O2 highest immediately following chasing: 25 fish displayed their highest Ṁ_O2 when the room lights went out (ca. 4 h later); 3 had their highest Ṁ_O2 when the lights came on the next morning; and 3 had their highest Ṁ_O2 at some other time during the respirometry trials. These rates were up to 77% higher than the Ṁ_O2 measured immediately after chasing (average 35% higher). Accordingly, the measurement interval with the highest Ṁ_O2 over the entire respirometry trial (ca. 20 h) was retained for analysis. In addition, we used the iterative slope analysis described by Zhang et al. (2020) to ensure that we captured the highest reliable Ṁ_O2 for each fish. In brief, for the retained measurement intervals, which were 2 min long during the first hour after chasing and 4 min thereafter, the slope of oxygen concentration versus time was determined over all possible periods of 30, 60, 90 and 120 s, each advancing by 1 s (Zhang et al., 2020). As reported elsewhere (Zhang et al., 2020), a period of 60 s resulted in statistically higher Ṁ_O2 than those estimated over the entire measurement interval (average increase 11%), while still achieving high coefficients of determination from least-squares linear regression (average r²>0.98). SMR was determined as the 20% quantile of all Ṁ_O2 measurements collected between 20:00 h and 06:00 h during a given trial (Chabot et al., 2016; Reemeyer and Rees, 2020). All Ṁ_O2 data were corrected for microbial respiration as previously described (Reemeyer and Rees, 2020).

Fish were euthanized by rapid chilling followed by cervical transection (Larter and Rees, 2017), and heart ventricles were quickly dissected (<2 min) and placed in 1 ml relaxing and biopsy preservation solution (BIOPS, in mmol l⁻¹: 50 K⁺-MES, 20 taurine, 0.5 dithiothreitol, 6.56 MgCl₂, 5.77 ATP, 15 phosphocreatine, 2.77 CaK₂ EGTA, 7.23 K₂ EGTA and 20 imidazole, adjusted with 5 mol l⁻¹ KOH to pH 7.1 at 0°C; Pesta and Gnaiger, 2012) on ice. Hearts were held at 0–2°C for 1 or 2 days prior to determination of myocardial Ṁ_O2 (see Supplementary Materials and Methods). Immediately after the removal of the heart, the right gill basket was removed. Individual gill arches were dissected, fixed overnight in neutral buffered formalin (4% formaldehyde, 33 mmol l⁻¹ NaH₂PO₄, 46 mmol l⁻¹ Na₂HPO₄), and kept in 100% ethanol. Total filament number (TFN), average filament length (AFL) and total filament length (TFL) were determined as described previously (Rees and Matute, 2018).

Intact heart ventricles were weighed and prepared for high-resolution respirometry following protocols for permeabilized human skeletal muscle fibers (Doerrier et al., 2018; Pesta and Gnaiger, 2012) with minor modification (see Supplementary Materials and Methods). Permeabilized myocardial tissue, roughly corresponding to one-quarter to one-half of the ventricle mass (mean mass 1.25±0.41 mg, range 0.64–2.32 mg), was transferred to the chamber of an Oroboros O2k respirometer (Oroboros Instruments) containing 2.0 ml MiR05 respirometry medium (in mmol l⁻¹: 110 sucrose, 60 K⁺-lactobionate, 0.5 EGTA, 3 MgCl₂, 20 taurine, 10 KH₂PO₄, 20 Hepes adjusted to pH 7.15 with KOH at 25°C, with 1 g l⁻¹ essentially fatty acid free BSA; Pesta and Gnaiger, 2012), supplemented with 20 mmol l⁻¹ creatine and 25 μmol l⁻¹ blebbistatin. Myocardial tissue respiration was measured at 25°C using a substrate–uncoupler–inhibitor titration (SUIT) modified from previous studies of fish heart metabolism (Fig. 1) (Baris et al., 2016; Iftikar and Hickey, 2013). The titration and corresponding respiratory states followed the sequence: NADH-generating substrates (5 mmol l⁻¹ pyruvate, 2 mmol l⁻¹ malate and 10 mmol l⁻¹ glutamate) to measure leak respiration due to complex I activity (LEAK_N); 5 mmol l⁻¹ ADP to stimulate phosphorylation-coupled respiration due to complex I activity; 10 μmol l⁻¹ cytochrome c to check the integrity of the outer mitochondrial membrane (OXPHOS_N); 10 mmol l⁻¹ succinate to saturate complex II (OXPHOS_NS); 0.5 μmol l⁻¹ oligomycin to inhibit ATP synthesis to measure leak respiration due to complex I and complex II activity (LEAK_NS); carbonyl cyanide m-chlorophenyl hydrazone (CCCP), in steps of 0.5 μmol l⁻¹, to uncouple mitochondria and elicit maximum electron transport system activity (ETS_NS); and 2.5 μmol l⁻¹ antimycin A to completely inhibit respiration due to mitochondrial electron transport and measure residual oxygen flux (ROX). The assay medium was gassed with pure oxygen as needed to keep the oxygen concentration between 120 and 400 µmol l⁻¹. Data acquisition, rate analyses and instrument calibration were performed with DatLab, version 7 (Oroboros Instruments) as previously described (Doerrier et al., 2018; Pesta and Gnaiger, 2012). Up to four O2k respirometers were used at a time (eight chambers), allowing the simultaneous measurement of four hearts in duplicate.

Pilot experiments were conducted to validate the use of ventricles after storage in BIOPS at 0–2°C for up to 2 days, along with validation of uncoupler and substrate concentrations (Supplementary Materials and Methods, Table S1, Figs S1–S4). Of the 39 fish used in the main study, ventricles from 22 fish were permeabilized and assayed after 1 day of storage, and ventricles from 17 fish were permeabilized and assayed after 2 days of storage. As observed in the pilot experiments, there was no significant effect of time stored at 0–2°C on any metric of mitochondrial performance (P>0.15 for all rates, flux control ratios and cytochrome c effects). Accordingly, data from all preparations were pooled without regard for the time ventricles were stored in BIOPS.

In addition, for a given ventricle, variation between oxygen flux determinations of duplicate permeabilized preparations was low: the coefficient of variation between duplicates averaged 11–12% for Ṁ_O2 during oxidative phosphorylation and maximum electron transport system activity (OXPHOS_NS and ETS_NS, respectively). In 8 out of 78 total replicates, however, one replicate assay had much lower oxygen flux (<50%), slower kinetics and higher noise than the other replicate from the same heart. This included two replicates for which the addition of 10 µmol l⁻¹ cytochrome c resulted in >20% increase in OXPHOS_N respiration, indicating damage to the outer mitochondrial membrane, likely arising from tissue dissection and permeabilization (see Supplementary Materials and Methods). The respiratory control ratios (RCR, OXPHOS_NS/LEAK_NS) for these replicates (4.2±1.1, mean±s.d., n=8) were significantly lower (P=0.001, paired t-test) than the RCR for the other replicates from the same hearts (10.3±2.5, mean±s.d., n=8), indicating poorly coupled mitochondria. These replicates were removed from the analyses. In addition, the ventricle of one of the smaller fish yielded only one sample of myocardial tissue >0.5 mg. Thus, oxygen flux was from a single ventricle preparation for nine fish. For the remaining 30 fish, oxygen flux from duplicate ventricle preparations was averaged and considered a single determination for statistical analyses.

General linear modeling was used to assess the relationship between whole-animal aerobic metabolism, gill morphology and myocardial Ṁ_O2. Separate models were fitted for MMR and SMR with body mass, ventricle mass, TFL, LEAK_N, LEAK_NS, OXPHOS_N, OXPHOS_NS and ETS_NS as predictor variables. Including body mass in the models simultaneously accounted for mass effects on whole-animal Ṁ_O2, as well as on other predictor variables, such that all reported results are independent of variation in body mass among individuals. Variables were log₁₀ transformed prior to analysis because of the known allometric relationships of metabolic and morphological data, although analyses of untransformed data yielded essentially identical results. All predictor variables were included in the initial models, and individual variables were removed in a stepwise fashion when doing so improved model fit, as judged by a decrease in Akaike's information criterion, corrected for small sample size (AIC_c) (Burnham and Anderson, 2002). The strength of the evidence supporting the final models was determined as the ratio of their Akaike weight to that of competing models. Visualization of the relationships between specific predictor variables and either MMR or SMR and the determination of Pearson's correlation coefficients used mass-independent residuals of ordinary least-squares linear regression of log–log plots of the variables of interest and body mass. Although the fish used in this study included both male and female fish collected from two nearby drainages, initial analyses of variance showed that neither sex nor collection site affected any metabolic or morphological variable, consistent with previous work on this cohort (Reemeyer and Rees, 2020). Thus, fish sex and collection site were not included in these models. Analyses were performed in GraphPad Prism, v.7.01, and SYSTAT, v.13. Data collected during this study are available from Dryad (doi:10.5061/dryad.jsxksn0bx).

This study aimed to determine whether variation in MMR and SMR by the Gulf killifish, F. grandis, held under common-garden conditions correlates with steps in the O₂ transport cascade from the environment to the tissues. As expected, MMR, SMR, gill morphology and heart size were strongly influenced by body mass (Table 1). Nevertheless, considerable variation remained after accounting for body size. In particular, mass-specific MMR and SMR varied approximately 2-fold (mean mass-specific MMR 0.336 µmol g⁻¹ min⁻¹, range 0.222–0.437 µmol g⁻¹ min⁻¹; mean mass-specific SMR 0.073 µmol g⁻¹ min⁻¹, range 0.052–0.106 µmol g⁻¹ min⁻¹). We asked whether the variation in MMR and SMR not accounted for by body size could be explained by size-adjusted gill morphology, ventricle mass or myocardial mitochondrial metabolism.

Protocols for tissue storage, cell permeabilization and high-resolution respirometry of myocardial mitochondria were optimized (Supplementary Materials and Methods, Table S1, Figs S1–S4) and used to determine Ṁ_O2 by ventricular tissue in various respiratory states. Oxygen flux in the LEAK state with complex I (PMG) and complex II (PMGS) substrates was low, whereas oxygen flux upon addition of ADP (OXPHOS) was high and sustained (Fig. 1, Table 2). The RCR (OXPHOS/LEAK) is commonly used to assess the quality of mitochondria. The RCR averaged 9.2±2.4 (mean±s.d.) with complex I and II substrates, indicating well-coupled mitochondria. Moreover, the addition of cytochrome c only had a small stimulatory effect on OXPHOS_N (5.7±3.7% increase, mean±s.d.), indicating minimal damage to the outer mitochondrial membrane. Finally, OXPHOS respiration with complex I and II substrates (OXPHOS_NS) routinely exceeded 90% of the oxygen flux during maximum electron transport capacity (ETS_NS). These characteristics of mitochondrial respiration are consistent with high-quality myocardial mitochondria.

While variation between replicate oxygen flux determinations for a given individual was low (see Materials and Methods), there was considerable variation in myocardial Ṁ_O2 among individuals in all respiratory states (Table 2). This variation was of similar magnitude to the individual variation in mass-specific MMR and SMR. For example, OXPHOS_NS and ETS_NS varied from 1.5- to 1.6-fold among fish. This variation was also evident in the flux control ratio (FCR) (Table 2), which standardizes the Ṁ_O2 in each state by the Ṁ_O2 during maximum electron transport system activity (e.g. 1.6-fold variation in FCR for OXPHOS_NS). Oxygen flux expressed per milligram of permeabilized ventricular tissue was independent of body mass (r²<0.03 for all respiratory states). Values of Ṁ_O2 for intact ventricles in each respiratory state, calculated as the product of respiration rate per milligram of tissue and ventricle mass, were positively related to body mass, which was a direct consequence of the strong allometric relationship of ventricular mass (Table 1).

General linear modeling was used to determine whether variation in whole-animal MMR and SMR was explained by variation in gill morphology, ventricular mass or myocardial bioenergetics (Table 3). Body mass was included in these models because of its strong effects on metabolic and morphological variables. The best model describing variation in MMR included body mass, TFL and myocardial Ṁ_O2 during oxidative phosphorylation with complex I and complex II substrates (OXPHOS_NS). The ΔAIC for this model was 1.60 compared with the next-best model. The corresponding Akaike weights indicated the final model was >2.2 times more likely to describe the MMR data than competing models (Table S2). The best model describing variation in SMR included body mass, TFL and myocardial Ṁ_O2 during maximum electron transport system activity (ETS_NS). It should be noted that the final model describing variation in SMR was only marginally better (ca. 1.4 times more likely) than models that retained ventricle mass (ΔAIC=0.65) or excluded ETS_NS (ΔAIC=0.70) (Table S2). Nevertheless, the final models explained 84% and 72% of the variation in MMR and SMR, respectively.

The correlations between MMR and SMR and the explanatory predictor variables are shown in Fig. 2. The correlations were highest for body mass (Fig. 2A,B), followed by TFL (Fig. 2C,D) and myocardial Ṁ_O2 (Fig. 2E,F). The effects of TFL and myocardial Ṁ_O2 are presented as mass-independent residuals of least-squares linear regression of log₁₀ of the indicated variable with log₁₀ body mass. After accounting for the effects of body mass, variation in TFL explained approximately 30% of the variation in MMR, whereas variation in myocardial OXPHOS_NS explained 13% of the variation in MMR. However, variation in TFL explained only 21%, and variation in myocardial ETS_NS only 7%, of the mass-independent variation in SMR.

Given that total filament length (TFL) is the product of total filament number (TFN) and average filament length (AFL), the relationships between mass-independent variation in MMR and TFN and AFL were assessed. After accounting for the effects of body mass, both TFN and AFL were positively correlated with MMR (Fig. 3), suggesting that increases in both component variables are associated with higher MMR. Interestingly, ventricle mass was not retained in the final models describing variation in MMR or SMR. This is probably because ventricle mass is highly correlated with body mass and did not explain any variation in MMR or SMR other than that explained by body size.

Finally, general linear modeling was also performed on mass-specific MMR and SMR and on data without logarithmic transformation. These analyses yielded final models that retained the same predictor variables for MMR and SMR as shown in Table 3. In addition, analyses of the variation in absolute aerobic scope (AAS), the difference between MMR and SMR, yielded essentially identical results to those shown for MMR (Table S2). This was expected because variation in MMR is quantitatively more important than that in SMR in determining the AAS in this species (Reemeyer and Rees, 2020).

Here, we demonstrate that interindividual variation in MMR and SMR of Gulf killifish, Fundulus grandis, correlates with variation in gill morphology and myocardial bioenergetics. These relationships were observed among individuals collected from a given location, held under common garden conditions, and after accounting for variation in body mass. Thus, the current study complements and extends earlier interspecific, interpopulation and acclimation studies on the relationship between the O₂ transport cascade and animal aerobic metabolism.

We found that within a given cohort of F. grandis, variation in TFL was positively correlated with MMR (Table 3). After mass correction, variation in TFL explained 30% of the variation in MMR (Fig. 2), which was accounted for by increases in filament number and the length of individual filaments (Fig. 3). It is important to note that TFL is only a proxy of gill surface area, albeit one that is commonly used (Palzenberger and Pohla, 1992) and validated by parallel measurements of TFL and gill surface area in other species (Chapman et al., 2008). Hence, the current results support the hypothesis that variation in gill surface area is linked to variation in maximum Ṁ_O2 among individuals within a species.

Our results largely agree with comparisons among species of fish that have shown that more-active species generally have larger gills than less-active species, and that this variation correlates with differences in whole-animal Ṁ_O2 (de Jager and Dekkers, 1975; Duncan, 2020; Hughes, 1966; Palzenberger and Pohla, 1992; Killen et al., 2016). In studies of population variation, however, the relationship between gill surface area and whole-animal Ṁ_O2 is less clear: some studies have shown a positive relationship between gill surface area and aerobic metabolism (McBryan et al., 2016), whereas other studies show no relationship, or even a negative relationship, when comparing populations or ecotypes having different levels of aerobic activity (Dalziel et al., 2012; Laporte et al., 2016). An important factor contributing to the divergence of gill morphology is the ambient oxygen level, where species or populations experiencing hypoxia typically have larger gills (Chapman et al., 2002, 2008). Accordingly, Laporte et al. (2016) proposed that the larger gill surface area in a less-active, benthic ecotype of lake whitefish compared with a more-active, limnetic ecotype was a response to the lower oxygen that the benthic ecotype faces in nature, rather than a consequence of differing levels of aerobic activity. Thus, comparisons among species or populations assess the consequences of past selection, genetic drift and other processes, while studies on individuals from similar environments and having similar evolutionary histories (such as the current one) can illuminate the range of variation upon which selection may act in the future.

We also showed that interindividual variation in MMR by F. grandis was positively correlated with the maximum ATP-generating respiration (OXPHOS_NS) by myocardial mitochondria (Table 3). In particular, variation in OXPHOS_NS explained 13% of the mass-independent variation in MMR (Fig. 2). Because the heart's capacity to pump blood through the circulation is supported by aerobic metabolism, at least under conditions of normal oxygenation (Farrell and Smith, 2017), it stands to reason that whole-animal Ṁ_O2 might vary with the capacity for myocardial oxidative phosphorylation. While the connection between cardiac energy metabolism and mechanical performance has long been appreciated (Driedzic and Gesser, 1994; Farrell and Smith, 2017; Moyes, 1996; Rodnick and Gesser, 2017), the current results extend this link to an organismal trait, maximum whole-animal aerobic metabolism.

Because the heart ventricles used in this study were approximately 0.1% of the total body mass (Table 1), we do not attribute the positive relationship between OXPHOS_NS and MMR to the quantitative contribution of ventricular Ṁ_O2 to whole-animal Ṁ_O2. The Ṁ_O2 of a ventricle of average mass (5.95 mg) having an average rate of OXPHOS_NS respiration (232 pmol mg⁻¹ s⁻¹) represents only 4.3% of the average MMR (1.92 µmol min⁻¹), a value that falls within the range reported for other species (Farrell and Smith, 2017). Variation among individuals within this small fraction of MMR cannot explain the overall variation we measured in MMR. Instead, we propose that the relationship between myocardial OXPHOS_NS and MMR is due to the requirement for mitochondrial ATP production to sustain cardiac output and support oxygen delivery to the tissues. Given that myocardial Ṁ_O2 is linearly related to cardiac power generation (Farrell and Smith, 2017), the 1.5-fold variation we observed in myocardial OXPHOS_NS could translate to an equally large effect on cardiac power generation. Using the conversion factor of 0.3 µl O₂ s⁻¹ mW⁻¹ g⁻¹ ventricular mass (Farrell and Smith, 2017), the range of OXPHOS_NS respiration measured here corresponds to a range in cardiac power of 14–21 mW g⁻¹, comparable to the power generated by isolated fish hearts in normal oxygen conditions at similar temperatures (Farrell and Smith, 2017; Rodnick and Gesser, 2017).

The variation in myocardial mitochondrial respiration reported here could arise from variation in mitochondrial quantity, quality or both. Mitochondrial density in skeletal muscle has been proposed to be a major determinant of MMR among mammals differing in body size, lifestyle and training (Weibel et al., 2004; Weibel and Hoppeler, 2005). However, Jacobs and Lundby (2013) argued that training-induced increases in MMR in humans involve both increased mitochondrial density and increased capacity for oxidative phosphorylation per mitochondria. Similarly, differences in muscle mitochondrial respiration in populations of deer mice living at different elevations correlate with increased mitochondrial density, altered intracellular mitochondrial distribution and greater surface area of cristae (Mahalingam et al., 2017). Factors that could impact the OXPHOS capacity of mitochondria include altered expression of mitochondrial enzymes (Oleksiak et al., 2005) or their interactions with membrane lipids; in particular, cardiolipin (Lau et al., 2017; Moyes and Hood, 2003; Paradies et al., 2014). Further research is needed to fully elucidate the molecular bases of interindividual variation in myocardial aerobic metabolism in fish and in animals in general.

As mentioned above, it is well established that variation in the capacity for mitochondrial metabolism of skeletal muscle is linked to interspecific and intraspecific variation in MMR (Karakelides et al., 2010; Hoppeler, 2018; Jacobs and Lundby, 2013; Scott and Dalziel, 2021; Weibel et al., 1991, 2004; Weibel and Hoppeler, 2005). This association follows from the observation that skeletal muscle may account for as much as 90% of the oxygen consumed during exercise (Hoppeler, 2018). Thus, an intriguing possibility is that certain individuals have a greater capacity for mitochondrial respiration in both heart and skeletal muscle, the former in support of cardiac output and the latter in support of elevated locomotor activity. In humans, exercise training, age and disease affect mitochondrial content or respiratory capacity of skeletal muscle and potentially other organs (Memme et al., 2021), suggesting that mitochondrial function of several tissues may vary in a coordinated fashion among individuals.

Given that cardiac output is the product of heart rate and stroke volume, the latter of which is influenced by heart size, it was surprising, perhaps, that variation in ventricle mass was not related to variation in MMR, after accounting for body mass. In mammals, the increase in maximum Ṁ_O2 due to training is associated with increased heart mass (Bassett and Howley, 2000), ‘athletic’ species have larger hearts than sedentary species of the same body size (Hoppeler and Weibel, 1998), and individual variation in MMR of mice is correlated with ventricle mass (Rezende et al., 2006). Among fishes, population and ecotype variation in aerobic metabolism has been positively correlated with body mass-adjusted ventricle mass in three-spined stickleback, lake whitefish and sockeye salmon (Dalziel et al., 2015, 2012; Eliason et al., 2011). In Trinidadian guppies, in contrast, Ṁ_O2 measured during swimming was not correlated with heart mass after accounting for body size (Odell et al., 2003). Similarly, Norin and Malte (2012) reported that mass-corrected ventricle size was unrelated to variation in MMR, SMR or AAS in brown trout. Together, these results suggest that the contribution of ventricle mass to variation in MMR in fishes may depend upon species or study design (e.g. among versus within population variation, see above).

Traits related to the O₂ transport cascade that may limit maximum Ṁ_O2 are expected to exceed the much lower oxygen requirements at rest (Weibel et al., 1991) and, therefore, they may be poorly correlated or uncorrelated with SMR. Nevertheless, we found individual variation in SMR by F. grandis was positively correlated with TFL and with maximum uncoupled electron transport system respiration by myocardial mitochondria (ETS_NS) (Table 3). Notably, these relationships were both weaker for SMR than for MMR (Table 3) and accounted for 21% and 7%, respectively, of the mass-independent variation in SMR (Fig. 2). Three models were approximately equal in explaining variation in SMR (Table S2), and all included TFL as an important explanatory variable. Thus, we can reject the null hypothesis of no correlation between TFL and SMR. An association between TFL and SMR might have a mechanistic basis. For example, larger gills represent a greater surface area for passive ion and water flux, which in turn requires active mechanisms to offset, thereby potentially elevating SMR (Gonzalez and Mcdonald, 1992). Alternatively, TFL could be mechanistically linked with MMR (see above) and also correlated with SMR because MMR and SMR are strongly, positively related in this species (Reemeyer and Rees, 2020). Further experiments are needed to distinguish between these possibilities; for example, assessing correlations between SMR and gill morphology or metabolism (Dawson et al., 2020) at various salinities.

Regarding the correlation between SMR of F. grandis and myocardial ETS_NS, the final model describing variation in SMR was only marginally more likely than a model that excluded ETS_NS (Table S2). Hence, the evidence supporting a correlation between myocardial metabolism and SMR by F. grandis is weak. Nevertheless, two recent studies in the closely related F. heteroclitus support a positive association between the Ṁ_O2 of whole hearts and RMR or SMR. Jayasundara et al. (2015) found that RMR measured at 28°C was significantly correlated (Spearman's rank=0.40) with ‘basal’ Ṁ_O2 of whole hearts (the average of three lowest rates with pyruvate as a substrate). Drown et al. (2021) investigated the relationships between SMR of F. heteroclitus acclimated to 12°C or 28°C with cardiac Ṁ_O2 in the presence of various metabolic substrates. After correcting for multiple comparisons, SMR at 12°C was positively correlated with cardiac Ṁ_O2 in the presence of lactate, ketones and ethanol, whereas SMR at 28°C was positively correlated with cardiac Ṁ_O2 fueled by endogenous substrates (no added substrates). However, because both studies used intact hearts, it is uncertain whether these relationships are due to individual variation in mitochondrial metabolism, substrate transport across the plasma membrane, or the amount of stored metabolic fuels.

Finally, our data allowed us to evaluate whether LEAK respiration was correlated with either SMR or MMR of F. grandis. Rolfe and Brown (1997) estimated that LEAK respiration by liver and skeletal muscle could account for up to 20–25% of BMR in mammals. In brown trout, Salin et al. (2016) found that interindividual variation in SMR correlated with LEAK respiration by liver mitochondria and variation in MMR correlated with LEAK respiration by skeletal muscle mitochondria. In addition, Mortensen and Gesser (1999) suggested that a significant portion of the Ṁ_O2 of rainbow trout cardiomyocytes at rest is due to LEAK. However, LEAK respiration is measured in the absence of ATP synthesis when the protonmotive force (Δp) is high. During ATP synthesis, the magnitude of Δp decreases across a range where LEAK respiration declines sharply (Nobes et al., 1990). Hence, the contribution of LEAK respiration to myocardial respiration in vivo, when the heart continuously synthesizes ATP at high rates, is likely to be minimal. Combined with the fact that the heart represents a small fraction of total body mass (Table 1), a quantitative relationship between the low rates of myocardial LEAK respiration (Fig. 1, Table 2) and whole-animal Ṁ_O2 seems unlikely. In the current study, myocardial LEAK respiration with either complex I substrates or complex I and II substrates was not found to be a significant predictor of either SMR or MMR. Hence, our results do not support an association between myocardial LEAK respiration and interindividual variation in whole-animal Ṁ_O2 by F. grandis.

As with any study that correlates in vitro results to in vivo performance, our interpretations come with caveats. First, the rates reported here should be considered maximum respiratory capacities because respiratory substrates, including oxygen, were maintained at saturating concentrations. Nevertheless, the values we report for myocardial Ṁ_O2 and the calculated cardiac power generation are comparable to values determined for intact, working fish hearts (Farrell and Smith, 2017; Rodnick and Gesser, 2017). Second, it is possible that differences in mitochondrial oxygen affinity (Cardinale et al., 2019; Chung et al., 2017; Larsen et al., 2011, 2020; Lau et al., 2017), ATP synthesis per mole of oxygen consumed (Salin et al., 2015, 2019) or respiratory substrates (Drown et al., 2021; Oleksiak et al., 2005) also contribute to interindividual variation in tissue and whole-animal Ṁ_O2. Third, this study focused on gill morphology and myocardial bioenergetics, which together explain 43% and 28% of the mass-independent variation in MMR and SMR, respectively. Metabolic rates are complex, polygenic traits (Bouchard et al., 1998; Hoppeler, 2018; Pettersen et al., 2018), and variables other than those measured here certainly contribute to variation among individuals.

Despite these caveats, however, this study is significant for several reasons. By demonstrating that variation in whole-animal Ṁ_O2 correlates with gill morphology and myocardial bioenergetics, the results support the idea that variation in aerobic metabolic rate is related to variation in subordinate traits comprising the O₂ transport cascade (Scott and Dalziel, 2021). Also, the results provide a link between myocardial bioenergetics and MMR, and by extension AAS, which represents the capacity for energetically expensive activities such as growth, foraging, digestion and locomotion (Claireaux and Lefrancois, 2007). Given that cardiac metabolism may be negatively impacted by anthropogenic climate alterations (Iftikar and Hickey, 2013) and contaminant releases (Kirby et al., 2019), the current results suggest that such impacts may extend to MMR, AAS and fitness-related activities. Furthermore, this study contributes to the growing appreciation of potential roles of mitochondria in animal energetics and ecology (Heine and Hood, 2020; Chung and Schulte, 2020; Koch et al., 2021; Sokolova, 2021). Finally, there are likely to be genetic bases for both gill morphology (Chapman et al., 2008) and myocardial bioenergetics (Baris et al., 2017); hence, variation in these traits among individuals within a population may respond to natural selection and influence the evolution of aerobic metabolism.

We thank T. E. Murphy, B. H. Price and J. Stampley for technical assistance and S. C. Hand for thoughtful discussions and critically reviewing the manuscript.