Is hypoxia vulnerability in fishes a by-product of maximum metabolic rate?

Global oceans and other aquatic habitats are changing at unprecedented rates as a result of natural and anthropogenic forces driving climate change. These changes include ocean warming, deoxygenation and acidification, all of which have been the subject of intense study by marine biologists in recent years (e.g. Keeling et al., 2010; Bozinovic and Portner, 2015; Esbaugh, 2018). Much of this work has focused on identifying and understanding the physiological traits that confer tolerance and sensitivity to environmental changes in different species. From a physiological perspective, there has been particular focus on respiratory performance, as all three climate change stressors pose specific challenges to respiratory gas transport. Respiratory performance is often quantified through the determination of aerobic scope (AS; see Glossary), which is the difference between the baseline cost of living (i.e. standard metabolic rate, SMR; see Glossary) and the maximum capacity to transport oxygen from the environment to mitochondria (maximum metabolic rate, MMR; see Glossary) (Clark et al., 2013). AS has become an important ecophysiological metric because it quantifies all the energy available for non-vital functions such as activity, reproduction and growth. Environmental stressors can act as either ‘limiting’ or ‘loading’ stressors that constrain oxygen (O₂) supply or raise metabolic costs, respectively.

The global scope of climate change has pushed researchers to seek broad-scale unifying approaches to assess the effects on species' performance and biogeography (i.e. latitudinal shifts in species distribution), the most recent of which is the metabolic index concept (see Glossary; Deutsch et al., 2015; Deutsch et al., 2020). This concept estimates the energy available to aquatic ectotherms (i.e. the metabolic index, ɸ) under varying environmental conditions using the thermal sensitivities of O₂ supply and demand, as well as the effects of O₂ partial pressure (⁠⁠) on O₂ supply. This concept has been used to estimate the energy available to aquatic ectotherms at equatorial and depth range limits, providing important information on the possible energetic limitations at these geographic boundaries (Deutsch et al., 2015; Deutsch et al., 2020). Notably, energetic limitations are not based on baseline costs of living (i.e. SMR), but instead are the cumulative result of baseline and routine activity costs across a wide variety of aquatic ectotherms. Approaches such as the metabolic index concept are critical to allow us to understand and predict the effects of climate change on ecosystems. However, it is equally important to critically evaluate whether these concepts are truly universal, and to identify species that may require alternative approaches. This seems particularly important for fishes, which are the most diverse vertebrate group on the planet and which occupy a plethora of different environments and ecological niches (Ravi and Venkatesh, 2008; Nelson et al., 2016). In this Commentary, we ask whether the metabolic index concept, in its current iteration, can be effectively applied to all fish species. Specifically, we seek to examine whether O₂ supply capacity (see Glossary) of fishes is constant across O₂ partial pressures, and whether it may be optimized to offset hypoxia vulnerability in some species.

The metabolic index concept defines α_S as the O₂ consumption rate divided by at which the O₂ consumption was measured. This can be measured as SMR/P_crit (Seibel and Deutsch, 2020). The effect of on O₂ consumption, and thus α_S, is generally described in one of two ways (Fig. 1B). Firstly, through a classical description of the line of O₂ limitation, which is often hyperbolic (e.g. Fry, 1947; McBryan et al., 2013; Claireaux and Chabot, 2016). Secondly, and more recently, models have described a linear relationship followed by a plateau (e.g. Wood, 2018; Seibel and Deutsch, 2020). The breakpoint is defined as the P_crit,max (see Glossary), the below which MMR is constrained. A comprehensive assessment of P_crit,max across a variety of animal phyla, including mollusks, arthropods and chordates, was recently performed by Seibel and Deutsch (2020), which used available SMR, MMR and P_crit values to predict P_crit,max (P_crit,max=MMR×P_crit/SMR) in a variety of species from different phyla. Of the 52 species tested – 28 of which were fish – 73% demonstrated a P_crit,max of approximately 21 kPa O₂. This led to the suggestion that hypoxia tolerance, as defined by P_crit, is simply a by-product of selective pressures on MMR (Seibel and Deutsch, 2020). This conclusion was based on the premise that maximizing MMR and AS in normoxia is a prominent evolutionary driver of respiratory performance and ecological success in aquatic ectotherms. If α_S is similar at P_crit and P_crit,max, then it is logical to suggest that hypoxia tolerance as defined by P_crit is a by-product of the evolutionary pressures on MMR. This is probably true for those species with a P_crit,max of 21 kPa; however, it is less certain for the other 27% of species.

Seibel and Deutsch (2020) suggested that species with a P_crit,max below approximately 21 kPa are adapted to hypoxic environments, but argue that adaptations to these environments are based on maximizing α_S at P_crit,max (i.e. maximizing MMR). However, this conclusion was reached by extrapolating only a single measurement of α_S (i.e. SMR/P_crit) per species to the measured MMR of that species. Here, we compiled datasets from seven different fish species that allow the calculation of α_S across a range (MMR/⁠; Fig. 2; Table S1). These seven species have P_crit,max values that range from 8.35 to 17.2 kPa, and P_crit values that range from 2.3 to 6.2 kPa. Of these seven, the red drum (Sciaenops ocellatus) (Ackerly and Esbaugh, 2020) and pumpkinseed (Lepomis gibbosus) (Crans et al., 2015) generally follow the pattern of a constant α_S until P_crit,max. In contrast, four of the species – the largemouth bass (Micropterus salmoides), rock bass (Ambloplites rupestris), blue gill (Lepomis macrochirus) (Crans et al., 2015) and common sole (Solea solea) (Lefrancois and Claireaux, 2003) – deviate from the described pattern with a gradual decline in α_S as increases from P_crit. The pattern of the last species, the black-axil chromis (Chromis atripectoralis) (Ern et al., 2017), is ambiguous but it is included for completeness. All four species without a constant α_S have at least one data point between P_crit and P_crit,max (Fig. 2). The physiological explanation for the patterns described above is uncertain. It is possible that fish species with a calculated P_crit,max below 21 kPa have an α_S that is selected for specifically on the basis of hypoxia tolerance (i.e. α_S at P_crit), not based on MMR at P_crit,max. Regardless of the mechanism, it is clear that α_S should not be assumed to be constant across values in all fishes.

The first component of the Fick equation is cardiac output, which is a combination of heart rate and stroke volume and refers to the total amount of blood pumped by the heart per unit time. When fish are exposed to hypoxia, they respond with a series of physiological adjustments intended to maintain O₂ delivery to the heart and brain. In most fishes, hypoxia exposure causes significant bradycardia (see Glossary) accompanied by increased stroke volume (Farrell, 2007; Gamperl and Driedzic, 2009) stimulated by elevated venous pressure (Sandblom and Axelsson, 2005) – although, in some species, this response does not occur until the fish approaches P_crit (e.g. Atlantic cod, Gadhus morhua) (McKenzie et al., 2009; Petersen and Gamperl, 2011). The observed change in cardiac output under hypoxia depends on whether increased stroke volume fully compensates for the bradycardia (Farrell, 2007; Gamperl and Driedzic, 2009). In some species, such as Adriatic sturgeon (Acipenser naccarii; Agnisola et al., 1999) and rainbow trout (Oncorhynchus mykiss; Sandblom and Axelsson, 2005), cardiac output is fully maintained. In other species, such as sea bass (Dicentrarchus labrax; Axelsson et al., 2002) and tilapia (Oreochromis hybrid sp.; Speers-Roesch et al., 2010), stroke volume does not compensate for the bradycardia, reducing cardiac output under hypoxia. Because these responses occur in conjunction with reduced arterial blood pressure (Sandblom and Axelsson, 2005; Speers-Roesch et al., 2010) it is hypothesized that their purpose is to reduce the cardiac power (see Glossary), and thus reduce O₂ demand. Put simply, the acute hypoxia strategy of fish prioritizes reduced cardiac O₂ demand as opposed to increased work to support uptake and delivery via perfusion.

During exercise, fishes prioritize increased O₂ delivery to skeletal muscle through changes to blood flow (e.g. Kolok et al., 1993). To facilitate the increased skeletal muscle O₂ demand, the heart rate will increase (tachycardia; see Glossary), as will stroke volume and cardiac output (reviewed by Farrell, 1991). In most fishes, an increase in stroke volume is more important than tachycardia – but this is species dependent. For example, yellowfin tuna (Thunnus albacares) exhibit a reduced stroke volume with exercise, yet cardiac output is still elevated because of substantial tachycardia (Korsmeyer et al., 1997). Exercise also has been shown to elevate arterial blood pressure (Stevens and Randall, 1967; Kiceniuk and Jones, 1977). Combined with elevated cardiac output, these changes work to substantially increase cardiac power, a critical limiting factor for maximal swim performance in fishes (e.g. Cox et al., 2017). O₂ delivery is also a crucial determinant of cardiac power, which is most evident in fishes with coronary circulation (e.g. Ekström et al., 2018). Yet, coronary circulation is generally considered as an adaptation to an athletic lifestyle, and we are unaware of any suggestion of hypoxia-induced evolution of this trait in fishes. As such, it appears that the selective pressures of aerobic exercise and acute hypoxia on cardiac performance are fundamentally different in fishes, with exercise-adapted fish prioritizing increased O₂ delivery and hypoxia-adapted fish prioritizing decreased O₂ demand.

The second component of the Fick equation is the arteriovenous difference in total O₂ content, which is a combination of and hemoglobin (Hb)–O₂ (i.e. O₂ bound to Hb). Hypoxia exposure and exercise both induce a number of similar physiological responses in the blood that protect total blood O₂ content, including elevated hematocrit (reviewed by Gallaugher and Farrell, 1998) and the initiation of a catecholamine response to protect red blood cell intracellular pH (Primmett et al., 1986; Tetens and Christensen, 1987). The former raises the total blood O₂-carrying capacity, whereas the latter protects Hb–O₂ binding affinity during periods of systemic acidosis. But these similarities between the responses to hypoxia and exercise belie a number of important differences that can be the focus of selective pressure. The first notable difference is that although both hypoxia and exercise result in decreased blood ⁠, only hypoxia significantly reduces the total blood O₂ content, at least with respect to the species studied to date (e.g. Holeton and Randall, 1967; Kiceniuk and Jones, 1977; Bushnell and Brill, 1992; Furimsky et al., 2003; McKenzie et al., 2004). The best datasets for illustrative purposes come from rainbow trout, where fish exercised at maximal swimming speed do not exhibit significantly reduced arterial total oxygen content (Brauner et al., 2000; McKenzie et al., 2004). Conversely, in response to hypoxia (4 kPa, 30 mmHg), the arterial blood of rainbow trout is significantly reduced to 37% saturation, as compared with ≥95% saturation under normoxia (Holeton and Randall, 1967). Differences between the responses to exercise and hypoxia also extend to venous blood, where evidence suggests that maximal exercise does not fully exhaust the available blood oxygen content, whereas hypoxia-exposed fish generally deplete venous oxygen reserves to maintain cardiac function. For example, the venous blood total O₂ content of maximally exercised rainbow trout is ∼23% that of the arterial blood (Brauner et al., 2000). Conversely, exposure to a of 4 kPa – a value approximately 1.5 kPa above the reported P_crit for rainbow trout (Svendsen et al., 2011) – lowers venous to 6 mmHg with only 3% Hb–O₂ saturation (Holeton and Randall, 1967).

The data described above suggest that the O₂-supply limitations that govern O₂ uptake during exhaustive exercise (i.e. MMR) and O₂ uptake at low ambient O₂ levels (i.e. P_crit) are fundamentally different in fishes. During exercise, increased cardiac output and hyperventilation (Stevens and Randall, 1967) maintain arterial blood O₂ saturation and augment delivery to muscles, suggesting that aerobic constraints relate to O₂-extraction efficiency, particularly in the heart. Conversely, hypoxia limitations are clearly governed by O₂ uptake, which is benefited by hyperventilation (e.g. Ern and Esbaugh, 2016) in combination with high Hb–O₂ binding affinity. This dichotomy is further supported by our understanding of the contribution of different Hb characteristics to O₂ uptake and delivery. Several studies have highlighted the mechanistic link between high Hb–O₂ affinity, whereby a low or reduced P₅₀ (i.e. the O₂ tension when Hb is 50% saturated with O₂) improves aerobic performance in hypoxic environments (e.g. Mandic et al., 2009; Speers-Roesch et al., 2012; Pan et al., 2017). Conversely, recent data have highlighted an important role for the Root effect (see Glossary) in driving O₂ extraction at the tissues when combined with plasma-accessible carbonic anhydrase activity (Rummer et al., 2013; Alderman et al., 2016; Harter et al., 2019), red blood cell Na⁺/H⁺ exchange (Rummer and Brauner, 2011) and cytoplasmic carbonic anhydrase activity (Dichiera and Esbaugh, 2020; Dichiera et al., 2020). This system has been specifically highlighted in the eye (Damsgaard et al., 2020), heart (Alderman et al., 2016) and red muscle (Rummer et al., 2013) of fishes, but has also been suggested to act to benefit systemic O₂ delivery more broadly (Randall et al., 2014; Rummer and Brauner, 2015; Shu et al., 2018; Harter et al., 2019). Interestingly, an 8 day hypoxia acclimation significantly reduced the magnitude of the Root effect in red drum, while simultaneously improving Hb–O₂ affinity. This further supports the notion of divergent evolutionary pressures exerted by hypoxia and exercise on fishes.

We would like to close this section by clarifying that we are not suggesting that exercise-based adaptations cannot benefit hypoxia tolerance, and vice versa. Elevated hematocrit, cardiac morphology and hyperventilatory responses are obvious examples that would benefit O₂ supply capacity under any circumstance, and are known to respond positively to exercise and hypoxia acclimation in fishes (e.g. Yamamoto, 1987; Anttila et al., 2015). We fully support the notion that fish not specifically adapted to hypoxia will likely have a hypoxia tolerance that is a by-product of the selective pressures on MMR (Seibel and Deutsch, 2020). However, we suggest that hypoxia adaption can also include a suite of unique respiratory traits subject to selection, which may optimize α_S at P_crit to avoid an over-reliance on unsustainable anaerobic metabolism.

The effects of hypoxia on ɸ are specific to the supply (S) side of Eqn 1, which can be simplified for a reference body mass and temperature as ⁠. In those species where α_S is not constant, there is a trend of gradually declining α_S as the environment approaches normoxia (Fig. 2). The rate of change in α_S is the determining factor in the accuracy of the ɸ estimate. To illustrate this, we have evaluated three different datasets as examples (Fig. 3). Red drum generally conform to the bisecting line model and have a P_crit,max relatively close to normoxia (16.6 kPa; 79% air saturation). In this case, the estimated ɸ deviates from the measured values by <5%, and this only occurs near the P_crit,max point. Common sole have a lower P_crit,max and a more pronounced 26% deviation between measured and estimated ɸ values at 13.3 kPa. A more extreme example is shown for rock bass, which reveals a 51% overestimate of ɸ when compared with the measured value at 9 kPa. We would stress that the existence of a P_crit,max below normoxia is not necessarily indicative of a species with an α_S optimized for P_crit, as evident in pumpkinseed (Fig. 2). However, P_crit,max will represent the point of greatest deviation between estimated and measured ɸ in such fish species.

With these illustrative examples in mind, we have two recommendations for those applying the ɸ concept at an organismal level. The first is that a calculated P_crit,max approximating normoxia (i.e. >80% air saturation) is suitable evidence that a species has a constant α_S. The second recommendation is that when a P_crit,max below 80% air saturation is observed (e.g. pumpkinseed), researchers should experimentally validate α_S at the equivalent P_crit,max oxygen tension. This will allow easy characterization of the consistency of α_S, and quantify the maximum estimate error as a result of a non-linear α_S.

A hyperbolic relationship between O₂ supply capacity and is unlikely to directly impact the thermal sensitivity coefficients (E; see Glossary) of aerobic parameters, and the data of Lefrancois and Claireaux (2003) support this assumption. Their dataset measured MMR at multiple O₂ tensions at three different temperatures, as well as SMR and P_crit for those temperatures. Using these data, we calculated relatively similar values for thermal sensitivity of oxygen supply (E_S) across the experimental range. Specifically, the E_S values were 0.23 at P_crit, 0.22 at 8 kPa, 0.23 at 13.3 kPa and 0.27 at 18.6 kPa (all values are equal to the mean of the reported measurement intervals).

A more important consideration with respect to warming was observed by Seibel and Deutsch (2020), who noted that the relative relationship between E_S and the thermal sensitivity of oxygen demand (E_D) may result in SMR exceeding MMR at very high temperatures, and more specifically that E_S could not effectively predict MMR beyond a breakpoint that the authors designated as CT_max. This breakpoint is defined by the occurrence of a thermal plateau, or decline, in MMR that deviates from the exponential increase observed at lower temperatures. The authors used this relationship to suggest that impaired α_S is not related to the decline in MMR because such an impairment would also be observed in P_crit. This conclusion assumes that α_S is constant at normoxic MMR and P_crit, which as described above may not be true for a number of fish species. In fact, the available evidence suggests that α_S limitations with warming in fishes are related to the ability to maintain cardiac output (Steinhausen et al., 2008; Eliason et al., 2011; Eliason et al., 2013), which, as described above, is not a major driver of P_crit. It is important to note the practical implications of these findings with respect to predicting ɸ under warming conditions. The numerator of the ɸ formula is equivalent to MMR (e.g. ⁠). This suggests that care should be taken when estimating ɸ at any temperature beyond which a thermally induced plateau in MMR is observed, as such extrapolation would result in an overestimate of ɸ.

As with the situation described above for hypoxia, the potential to overestimate ɸ is likely to affect only a subset of fishes. For example, species such as goldfish (Carassius auratus) (Fry and Hart, 1948), killifish (McBryan et al., 2013) and black sea bass (Centropristis striata) (Slesinger et al., 2019) all have documented thermal plateaus in MMR. Conversely, many species, such as Atlantic halibut (Hippoglossus hippoglossus) (Grans et al., 2014) and barramundi (Lates calcarifer) (Norin et al., 2014), show a consistent increase in MMR with warming. A thorough examination of the effects of warming on metabolic rate across a wide range of species can be found in Lefevre (2016). In the context of ɸ it is important that focus is placed on defining whether MMR collapses with warming, and identifying the temperature at which such breakpoints occur in species. To date, reduced cardiac performance has been highlighted as a likely cause for warming-induced collapses in MMR (e.g. Eliason et al., 2013); however, we should continue to explore additional causative factors that may contribute to this phenomenon. Intraspecies variation in such traits will ultimately be the focus of the selective processes that define the trajectory of species with climate change.

Thus far we have demonstrated that respiratory performance in fishes is not consistent with the assumption that α_S is singularly evolved to meet the demands of exercise in all fishes. Instead, we hypothesize that the evolutionary pressures that have shaped respiratory performance in fishes can be divided into two categories. The first category contains those species adapted for exercise in their environment, which would be represented by a constant α_S between P_crit and P_crit,max (e.g. red drum; Fig. 2C) (see Seibel and Deutsch, 2020). The second category contains those fishes whose α_S is adapted to meet O₂ demands at P_crit, which is identifiable by declining α_S as increases from P_crit to a theoretical P_crit,max (e.g. rock bass; Fig. 2H). It is important to recognize these patterns when studying the physiological mechanisms related to hypoxia tolerance and vulnerability in fishes. When α_S is adapted for performance under normoxia, hypoxia vulnerability is likely to be a by-product of the selective pressures on MMR. Conversely, fishes with an α_S that increases as O₂ declines are likely to have specific adaptations to reduce hypoxia vulnerability.

Our second take home message is that a variable α_S violates a central assumption of the metabolic index concept, and results in overestimated ɸ. The significance of such an error depends on the species, the scope of the error and how ɸ is used by researchers. For example, the recent work by Deutsch et al. (2020) used biogeographical and physiological data to predict the energetic limitations that constrain animal species distributions. Note that this work included a wide variety of marine animal phyla, and reported an average ɸ_crit (i.e. the value of ɸ found at the equatorial or depth limit of a species distribution) of 3.3±0.3 (mean±s.e.m.; N=22) for marine fish species. For context, a subset of those species with both ɸ_crit estimates and FAS data across a thermal range are shown in Fig. 4 (Table S2; data from Seibel and Deutsch, 2020; Ackerly and Esbaugh, 2021). Note that ɸ_crit is an estimate of FAS, and that ɸ_crit is quite high when compared with the FAS at the upper thermal range of laboratory studies. If we assume that the thermal ranges employed for the studies on FAS (Fig. 4A) are environmentally relevant, then the estimates of ɸ_crit (Fig. 4B) suggest that the energetic limitations for many of these fishes equal the entirety of the theoretical AS. However, in cases where fish exhibit flexible α_S across O₂ tensions, or exhibit a thermally induced MMR collapse, the estimated ɸ_crit is likely to be lower than reported.

An overestimated ɸ_crit in a subset of fishes will have little bearing on the overall conclusions of Deutsch et al. (2020); namely, that the energetic limitations that define aquatic ectotherm species ranges are due to the combined effects of exercise, temperature and hypoxia. However, overestimates can change our understanding of the absolute energetic minimums that species experience, and how ɸ is used to predict the effects of climate change. In some ways, ɸ_crit is an estimate of the buffer a species has against further warming and deoxygenation before they are forced to rely on anaerobic metabolism to meet their energetic demands. Species such as the black-axil chromis (ɸ_crit=5.2) have a large buffer against respiratory stress, whereas Atlantic salmon (ɸ_crit=2.2) have relatively little capacity for further O₂ supply constraints. It is also important to recognize that species’ responses to climate change are complex, and survival may depend on poleward migration, phenotypic plasticity or other adaptations that allow species to maintain their existing ɸ_crit. If we are to properly understand the role that ɸ_crit may have in shaping biogeography and performance in fishes, it is important that we work with the best available estimates.

The metabolic index concept has placed a new and important significance on α_S and the relationship between O₂ consumption and ⁠, particularly at intermediate levels of hypoxia. In this Commentary, we reinforced the premise that this relationship can vary among species (Fry, 1947), and that many species show hyperbolic curves representative of an elevated α_S as the animal approaches P_crit. Although these findings have clear implications for the calculation of ɸ and ɸ_crit, we would like to stress that this critique should not be viewed as a broad criticism of the metabolic index concept overall. We view any framework that can effectively predict the metabolic scope of a majority of aquatic ectotherms in response to environmental change as an incredibly powerful tool. But unifying theories are challenging, and inevitably there will be species that defy expectation. We argue that it is important to acknowledge such species and to develop alternative approaches that ensure a complete understanding of the effects of environmental change on animal performance. In particular, we hope this work places renewed focus on those species that may be adapted to hypoxic environments. This should include continued efforts to understand the physiological traits that define α_S in these species, while also exploring the compounding effects of warming and deoxygenation on hypoxia vulnerability.

We would like to thank Brad Seibel for providing Excel spreadsheets that facilitated calculation of the thermal sensitivity of metabolic traits.