Estimating maximum oxygen uptake of fishes during swimming and following exhaustive chase – different results, biological bases and applications

The maximum rate at which an aquatic animal can take up oxygen from its environment is its Ṁ_O2,max (see Glossary). This rate reflects the upper limit for the extraction and delivery of oxygen to tissues to support aerobic metabolism and it is positively associated with the capacity for sustained aerobic performance (Claireaux et al., 2005; Norin and Clark, 2016). In aquatic ectotherms, especially fishes, Ṁ_O2,max is commonly used as a proxy for maximum metabolic rate (MMR; see Glossary), specifically the aerobic component of MMR (Box 1). In addition, Ṁ_O2,max sets the upper boundary of an animal's scope for metabolic activity (Fry, 1947) or aerobic scope (AS; see Glossary) (Clark et al., 2013). AS reflects the animal's capacity to aerobically fuel activities such as foraging, digestion, predator avoidance, migration, growth and reproduction (Fry, 1947; Priede, 1985; Claireaux and Lefrancois, 2007; Clark et al., 2013). A higher AS, indicating a greater capacity to support these activities, may enhance Darwinian fitness in natural populations (Claireaux and Lefrancois, 2007). Conversely, if AS is reduced by a decrease in Ṁ_O2,max (or an increase in standard metabolic rate, SMR; see Glossary), energetic trade-offs among processes that ‘use’ AS could ensue (Killen et al., 2007; Farrell, 2009; Holt and Jørgensen, 2015; Eliason and Farrell, 2016; Farrell, 2016; Metcalfe et al., 2016). Although the time an animal spends at or near its Ṁ_O2,max is likely to be short and species dependent (Priede, 1977; Farrell, 2016), its activities during this time (e.g. predator avoidance, migration or active foraging) might be disproportionately important to its survival. Hence, a high Ṁ_O2,max and, by extension, an expanded AS are thought to benefit individuals, populations and species.

The Fick equation (see Glossary) states that Ṁ_O2 (see Glossary) is a function of cardiac output and the difference in oxygen content of arterial and venous blood. Thus, methods to measure Ṁ_O2,max should elicit rates of tissue oxygen consumption that reach the cardiorespiratory limits for oxygen extraction and delivery to the tissues (Jones and Randall, 1978; Rummer and Brauner, 2015; Scott and Dalziel, 2021; Rees et al., 2022). The two most common methods to estimate Ṁ_O2,max in fishes are to use intermittent-flow respirometry (see Glossary) to measure Ṁ_O2 either (1) as a fish swims at increasing speeds against an imposed current in a swim tunnel or (2) during recovery after an exhaustive chase (Clark et al., 2013; Rummer et al., 2016; Svendsen et al., 2016; Norin and Clark, 2016; Killen et al., 2017). Although we focus on these two methods, higher Ṁ_O2 values can be attained under other conditions in certain species (see ‘Matching method to biology’, below).

Swim tunnel respirometry measures the Ṁ_O2 required to support the elevated aerobic metabolism of vigorous, sustained swimming, primarily powered by contraction of red skeletal muscle and largely fuelled by oxidative phosphorylation. The rate of mitochondrial oxygen consumption, and therefore whole-animal Ṁ_O2 (assuming mitochondrial and organismal oxygen uptake are in a steady state; Box 1), is largely to replenish ATP that is consumed by skeletal muscle cross-bridge cycling and calcium ion regulation (Hoppeler, 2018). In addition, heart rate and ventilation rate are elevated, although ram ventilation reduces the energetic costs of gill ventilation during high-speed swimming in some species (Steffensen, 1985). Ṁ_O2,max is estimated as the highest Ṁ_O2 (peak Ṁ_O2; see Glossary) as the fish swims (i.e. peak Ṁ_O2,swim) at increasing water speeds until exhaustion (U_crit test; Brett, 1964) or during more rapid and dynamic increases in water speed (U_max test; Clark et al., 2011, Raby et al., 2020). Forced locomotor activity was the first technique used to measure Ṁ_O2,max in fishes (Blazka et al., 1960; Brett, 1964; Steffensen et al., 1984), and it is still considered by many to be the ‘gold standard’ for determining Ṁ_O2,max. However, swim tunnel respirometry is time consuming and requires continuous monitoring, which reduces experimental throughput. It is also cumbersome to deploy in remote field conditions, and some fishes either are poor sustained swimmers or cannot be coaxed to perform in a swim tunnel (Clark, 2022).

Because of these challenges, Ṁ_O2,max has also been estimated as the highest Ṁ_O2 during recovery from an exhaustive chase (hereafter, peak Ṁ_O2,recovery) (Soofiani and Priede, 1983; Reidy et al., 1995). Typically, fish are chased by the experimenter in a circular arena and, in some cases, held in air briefly (e.g. 1 min) prior to commencing intermittent-flow respirometry in a chamber with limited volume and minimal water movement (‘static’ chambers). The fish's activity during the chase is supported by both oxidative phosphorylation and anaerobic metabolism (glycolysis and creatine phosphate hydrolysis). During recovery, very little Ṁ_O2 is used for locomotion, but rather, Ṁ_O2 remains elevated as a result of reoxygenation of internal oxygen stores (haemoglobin and myoglobin), persistently elevated cardiac activity and ventilation (without the benefit of ram ventilation), re-establishment of pH and ion balance and the clearance of anaerobic end-products (Wood, 1991; Moyes et al., 1992; Scarabello et al., 1992). Although Ṁ_O2 is expected to decline exponentially after the chase, in some cases peak Ṁ_O2,recovery is not achieved until hours later (Clark et al., 2012; Soofiani and Priede, 1983; Andersson et al., 2020; Brieske et al., 2024). Moreover, the extent to which Ṁ_O2 is elevated depends upon how the chase is performed (Reidy et al., 1995; Roche et al., 2013; Zhang et al., 2018), presumably reflecting different degrees of metabolic and cellular disturbance. Nevertheless, the chase method is typically easier and quicker to conduct than swim tunnel respirometry, allowing several individuals to be tested in parallel, thereby increasing experimental throughput (Norin and Malte, 2011; Salin et al., 2016; Reemeyer and Rees, 2020). In addition, the apparatus is more portable, facilitating experiments in remote locations (e.g. Little et al., 2020).

Despite the differences in the biological processes measured and apparatus used by these two methods, a meta-analysis of data from 121 species of fishes differing in lifestyle (i.e. benthic, benthopelagic, pelagic) did not detect a systematic difference between peak Ṁ_O2,swim and peak Ṁ_O2,recovery (Killen et al., 2017). This conclusion was largely based on values of Ṁ_O2,swim and Ṁ_O2,recovery measured on different individuals and, in some cases, determined in different studies. This conclusion differs from those of a number of studies, many of which used both approaches on the same individuals, showing that peak Ṁ_O2,swim can be substantially higher than peak Ṁ_O2,recovery (Roche et al., 2013; Rummer et al., 2016; Hvas and Oppedal, 2019; Slesinger et al., 2019; Raby et al., 2020; Eisenberg et al., 2024; Brieske et al., 2024). In other species, peak Ṁ_O2,swim is markedly lower than peak Ṁ_O2,recovery (Soofiani and Priede, 1983). Thus, the degree to which these methods yield similar estimates of Ṁ_O2,max remains unclear.

Additionally, several studies have documented that an individual's peak Ṁ_O2 is repeatable when determined in two or more trials of either swim tunnel respirometry or recovery from an exhaustive chase (Reidy et al., 2000; Marras et al., 2010; Norin and Malte, 2011; Killen et al., 2016; Norin et al., 2016; Reemeyer and Rees, 2020; Brieske et al., 2024). These results are important for two reasons. First, the repeatability of the technique is a measure of its precision. Second, repeatability of peak Ṁ_O2 over time suggests that it is a stable feature of the individual, and thus potentially subject to evolution by natural selection (see Roche et al., 2016). However, whether an individual's peak Ṁ_O2 is repeatable when measured by different techniques has received considerably less attention (but see Zhang et al., 2020; Brieske et al., 2024; Eisenberg et al., 2024).

In this Commentary, therefore, we directly compare estimates of peak Ṁ_O2 when the same individuals were used both in swim tunnel respirometry and during recovery from an exhaustive chase (i.e. using repeated-measures protocols). We asked (1) whether mean estimates of peak Ṁ_O2 determined by these two approaches were the same, and (2) whether an individual's peak Ṁ_O2 determined in swim tunnel respirometry was correlated with its peak Ṁ_O2 after an exhaustive chase (i.e. repeatable across techniques). We extracted data from 10 species representing various fish lineages and habitats, used in 14 experimental trials, conducted over a range of experimental conditions (temperature, salinity and apparatus) (Table S1). All data were published previously (references in Table S1), except for those for pumpkinseed sunfish (Lepomis gibbosus). Sunfish were collected, housed and acclimated for 4 weeks to one of three temperatures (20, 25 and 30°C; n=12 each) as described in De Bonville et al. (2024). At each temperature, peak Ṁ_O2,swim and peak Ṁ_O2,recovery were determined essentially as described in Binning et al. (2013) and Guitard et al. (2022), respectively. These experiments were approved by Université de Montréal's animal care committee (Comité de déontologie de l'expérimentation sur les animaux; certificate number 22-025).

Mean values for peak Ṁ_O2,recovery are plotted against peak Ṁ_O2,swim in Fig. 1A. For this comparison, Ṁ_O2 was normalized to a common body size (1 kg) and temperature (20°C) (Killen et al., 2017). Although the slope of the relationship was not significantly different from 1.0 (95% confidence intervals, 0.82 to 1.17), the intercept differed from zero (95% confidence intervals, −332 to −11 mg O₂ h⁻¹) and the large majority of individual Ṁ_O2 measurements fell below the line of unity, suggesting that peak Ṁ_O2,recovery was less than peak Ṁ_O2,swim. This suggestion was reinforced when each individual's peak Ṁ_O2,recovery was expressed as a proportion of its peak Ṁ_O2,swim (Fig. 1B). The mean ratios of peak Ṁ_O2,recovery to peak Ṁ_O2,swim ranged from 0.57 to 1.01 (median 0.765), and for 10 of 14 experimental trials the 95% confidence intervals for this ratio did not overlap one, meaning that peak Ṁ_O2,recovery was significantly less than peak Ṁ_O2,swim (P<0.05, paired t-tests; see Table S1 for individual P-values). Thus, for most species and experimental conditions examined here, peak Ṁ_O2,recovery was substantially lower than peak Ṁ_O2,swim (up to ∼40% lower; median 23.5%). Because these comparisons were paired (i.e. made on the same individuals under defined conditions), this conclusion is not influenced by differences in body size, temperature or other experimental variables.

When peak Ṁ_O2,swim and peak Ṁ_O2,recovery were measured in the same individuals, they were generally weakly related or unrelated to one another (Fig. S1). Values for Pearson's correlation coefficients comparing these two metrics within an experiment ranged from −0.25 to 0.93 (median 0.32), and in only two of 14 experimental trials were the correlations between peak Ṁ_O2,swim and peak Ṁ_O2,recovery significant (Fig. 2A; see Table S1 for individual P-values). The same pattern arose when each individual's rank within the group was assessed. Spearman's rank order correlation coefficients (ρ) ranged from −0.10 to 0.88 (median 0.29), and in only one experimental trial was the correlation between individual ranks statistically significant (Fig. 2B; see Table S1 for individual P-values). Perhaps peak Ṁ_O2 is simply poorly repeatable for the species included in these analyses. This possibility was directly addressed by Brieske et al. (2024), who found that the repeatability of peak Ṁ_O2,swim of Gulf killifish, Fundulus grandis, was high in two trials of swim tunnel respirometry (Pearson's r=0.67, P<0.05), as was the repeatability of peak Ṁ_O2,recovery in replicate trials of recovery after an exhaustive chase (Pearson's r=0.79, P<0.01). For the same fish, however, peak Ṁ_O2,swim was unrelated to peak Ṁ_O2,recovery (r<0.30, P>0.25; Fig. 2A; Table S1, Fig. S1). This study clearly showed that peak Ṁ_O2 may be consistent when assessed by a given method yet be unrelated across methods. Thus, for the species and contexts studied here, an individual's peak Ṁ_O2,swim has little to no bearing on the same individual's peak Ṁ_O2,recovery.

Several experimental design considerations affect the accuracy, precision and temporal resolution of Ṁ_O2 measurements by intermittent-flow respirometry (Clark et al., 2013; Svendsen et al., 2016). Accordingly, we tabulated several features of the experimental designs used in the studies compiled here (Table S1), and asked whether these features were related to either the agreement between mean values or the repeatability of an individual's peak Ṁ_O2 when assessed by these two approaches (Fig. S2).

Of four attributes related to the species employed in these studies (taxonomic group, habitat, life stage and swimming ability) (Fig. S2A–D), only taxonomic group was related to how well mean peak Ṁ_O2 estimates agreed between techniques. The four salmonid species, on average, had a higher ratio of peak Ṁ_O2,recovery to peak Ṁ_O2,swim than non-salmonids (P=0.047, Mann–Whitney U-test; Fig. S2A). This observation is consistent with Little et al. (2020) who found that swim tunnel respirometry and exhaustive chase generally produced similar estimates of peak Ṁ_O2 in Coho salmon (Oncorhynchus kisutch). While the salmonids studied here had somewhat better agreement between peak Ṁ_O2 determined by these methods than non-salmonids, it is important to note that peak Ṁ_O2,recovery was, nevertheless, statistically lower than peak Ṁ_O2,swim in two of the four trials with salmonids (Fig. 1; Table S1). Also, the repeatability of peak Ṁ_O2 across these methods was poor for salmonids and non-salmonids alike (Fig. S2A).

Of six attributes related to experimental design (Fig. S2E–J), only the ratio of respirometer chamber volume to fish mass during the chase method was related to the agreement between peak Ṁ_O2 estimates. When the static chamber used in intermittent-flow respirometry was larger relative to the size of the fish (i.e. chamber volume to fish mass ratio >50), peak Ṁ_O2,recovery and peak Ṁ_O2,swim agreed better than when smaller chambers were used (P=0.047, Mann–Whitney U-test; Fig. S2I). It could be that larger static chambers allow greater room for fish locomotion, thus elevating peak Ṁ_O2,recovery (Peake and Farrell, 2004). However, this comparison is tempered by three considerations. First, three of four trials with salmonids used larger respirometer chambers, thus confounding chamber volume with taxonomic group (see above). Second, chambers with volume to fish mass ratios <50 are generally recommended for intermittent-flow respirometry because of better temporal resolution and signal to noise ratios (Svendsen et al., 2016). Third, the repeatability of peak Ṁ_O2 across these methods was equally poor regardless of the chamber volume to fish mass ratio (Fig. S2I).

Other experimental design features reported to influence peak Ṁ_O2,recovery [e.g. time elapsed between the chase and the start of respirometry (Fig. S2G) and whether fish are exposed to air (Fig. S2H)] failed to explain the differences noted in either the ratio of peak Ṁ_O2,recovery to peak Ṁ_O2,swim or the repeatability of an individual's peak Ṁ_O2. Finally, in nine of 14 experiments, the two methods were applied in a random order (Table S1), suggesting that these results were not biased by factors such as training, fatigue or duration of laboratory maintenance. Therefore, differences in the mean and repeatability of peak Ṁ_O2 determined by swim tunnel respirometry compared with exhaustive chase were not obviously related to the species or experimental conditions employed in the studies examined here. Rather, we attribute these differences to the fact that peak Ṁ_O2,swim and peak Ṁ_O2,recovery reflect different physiological states, each with their own underlying determinants and ranges of variation among individuals.

Because peak Ṁ_O2 is dynamic and context dependent, devices and analytical techniques must be able to capture transiently elevated rates. Given the high sampling frequency of oxygen sensors and the development of ‘rolling’ or ‘sliding-window’ regressions in analytical software, it is now possible to evaluate multiple measurement intervals and select the shortest interval that captures the highest Ṁ_O2 without sacrificing precision (Box 2; Zhang et al., 2019, 2020; Little et al., 2020; Prinzing et al., 2021). Thus, ‘rolling’ regression is more likely to capture transiently elevated Ṁ_O2 than determinations made over longer measurement intervals, and it should be incorporated into analyses whose goal is to estimate peak Ṁ_O2.

Another innovation is to modify a static respirometer chamber by introducing a chasing device, allowing the fish to be motivated to swim vigorously while Ṁ_O2 is simultaneously recorded (Norin and Clark, 2016; Zhang et al., 2019, 2020). Such modification allows determination of peak Ṁ_O2 during the chase itself (Ṁ_O2,chase). When this modification was combined with ‘rolling’ regression, peak Ṁ_O2,chase of juvenile rainbow trout (Oncorhynchus mykiss) was higher than Ṁ_O2,recovery but not different from peak Ṁ_O2,swim of the same individuals (Zhang et al., 2020). Moreover, peak Ṁ_O2,chase and peak Ṁ_O2,swim of individual fish were correlated (i.e. repeatable; Pearson's r and Spearman's ρ ≥0.77; P<0.05; Zhang et al., 2020), whereas neither was significantly correlated with Ṁ_O2,recovery. In contrast, a study on juvenile barramundi (Lates calcarifer) showed that Ṁ_O2,chase was not as high as Ṁ_O2,recovery measured immediately post-chase (Norin and Clark, 2016, 2017). Clearly, more validation is required, but using a modified ‘chase respirometer’ might offer researchers a tool that is higher throughput, lower cost and more portable than traditional swim tunnel respirometry, while also providing estimates of peak Ṁ_O2 that may be similar to Ṁ_O2,swim in mean and repeatability.

Because swim tunnel respirometry and exhaustive chase protocols measure different physiological states, it follows that they are not equivalent methods to estimate Ṁ_O2,max in fishes. Rather, each method is useful to explore a different set of biological questions and, potentially, better suited for a given species.

For questions related to the aerobic costs of locomotion, swim tunnel respirometry is the obvious choice. Swim tunnels and swimming protocols can be modified to accommodate fishes with different swimming styles and abilities (Priede and Holliday, 1980; Van den Thillart et al., 2004; Clark et al., 2011; Rummer et al., 2016). However, some species are not strong, sustained swimmers, and using swim tunnel respirometry to determine the Ṁ_O2,max of such species might not be appropriate. We also acknowledge that even for strong, sustained swimmers, the dimensions of swim tunnels might constrain activity and underestimate a fish's swimming capacity in more naturalistic settings (Peake and Farrell, 2006; Kern et al., 2017).

An exhaustive chase protocol could help define the metabolic costs of recovering from vigorous activity such as predator avoidance or capture. This could be more ecologically relevant than the cost of sustained swimming in certain conservation or fisheries studies. For example, following the chase with a period of air exposure was originally developed as a way of mimicking the stress and handling of catch-and-release sport fishing (Donaldson et al., 2010; Clark et al., 2012).

It is also possible that Ṁ_O2,max is reached during states other than sustained swimming or recovery from an exhaustive chase. Ingestion of a meal brings about an increase in Ṁ_O2, which is thought to reflect the cost of food handling, breakdown, assimilation and somatic growth (Brett and Groves, 1979; Goodrich et al., 2024). The magnitude of the post-prandial Ṁ_O2 depends upon the species of fish, the quantity and quality of food consumed, and other factors (e.g. temperature; Chabot et al., 2016a). In some circumstances, the post-prandial Ṁ_O2 may approach, or even exceed, the Ṁ_O2 measured during swimming or after an exhaustive chase, especially for poor-swimming predatory fishes that consume large meals (i.e. ambush predators; Soofiani and Hawkins, 1982; Fu et al., 2009; Steell et al., 2019). Yet other species might achieve their highest Ṁ_O2 during spontaneous activity as a result of light changes associated with photoperiod (Andersson et al., 2020).

Using paired comparisons of Ṁ_O2 determined for a diverse group of fishes measured under an array of experimental conditions, we show that peak Ṁ_O2 depends upon the method used. This is true for mean values of peak Ṁ_O2 determined for a group of individuals, as well as for the repeatability of an individual's peak Ṁ_O2. These results reinforce the need to carefully consider the biological context of experiments that measure peak Ṁ_O2 (Roche et al., 2013; Clark et al., 2013; Norin and Clark, 2016; Farrell, 2016; Rummer et al., 2016; Raby et al., 2020) and lead to several recommendations for studies of elevated aerobic metabolism in fishes (Box 3).

We hope that these recommendations will be integrated into the determination of peak Ṁ_O2 in future studies of fish physiology, behavioural ecology, conservation and management. Such studies might explore how certain biotic or abiotic variables differentially affect peak Ṁ_O2 measured by diverse methods. For example, exposure to elevated temperature might have different effects on peak Ṁ_O2,swim and peak Ṁ_O2,recovery or the ranking of individual Ṁ_O2 determined by these two methods. Such outcomes would suggest that the underlying physiological processes differ in their thermal sensitivities, which could have implications for a fish's capacity for sustained swimming versus recovery from burst swimming in the context of climate warming (Clark et al., 2017; Johansen et al., 2021). Furthermore, parasites that impair sustained swimming (e.g. Palstra et al., 2007) might alter the group mean and individual variation of peak Ṁ_O2,swim but not peak Ṁ_O2,recovery. It would also be valuable to compare the repeatability of peak Ṁ_O2,swim and peak Ṁ_O2,recovery over the lifespan of individuals to assess the influences of ontogeny (e.g. Downie et al., 2023) and acclimation to changing environments (e.g. Auer et al., 2018; Reemeyer and Rees, 2020).

By appreciating that these methods measure different biological processes and address different biological questions, we hope to enhance our understanding of both the biology of fishes and the impacts of human-induced changes to aquatic habitats.