Are reactive oxygen species always bad? Lessons from hypoxic ectotherms

Oxygen is a double-edged sword for aerobic organisms. On the one hand, O₂ is essential for oxidative phosphorylation in the mitochondria, the process that transduces substrate energy into its usable form, ATP – the energy currency of the cell. On the other hand, O₂ is the substrate for the formation of reactive oxygen species (ROS; see Glossary), such as superoxide (O₂^·−) and hydrogen peroxide (H₂O₂), which can interfere with signalling processes and even damage cellular components (Mailloux, 2020; Murphy, 2009; Sies et al., 2022) (Box 1). ‘Oxidative distress’ (or oxidative stress; see Glossary) occurs when ROS generation overwhelms protective scavenging mechanisms within the cell, leading to the accumulated ROS interfering with a broad range of non-specific targets, such as pathways that do not typically rely on redox signalling, or by damaging a number of biomolecules (Sies, 2020). In this context, ROS generation is often understood as something to avoid or minimize. Indeed, the observation that ROS can cause extensive tissue damage and pathology forms the basis of the ecological theory of oxidative stress and life history, which hypothesizes that increased ROS and oxidative damage are associated with reduced fitness (Costantini, 2019; Metcalfe and Alonso-Alvarez, 2010).

However, this dichotomy of O₂ is not as simple as it seems, and ROS are not invariably harmful in biological systems (D'Autréaux and Toledano, 2007; Hernansanz-Agustín et al., 2020; Holmström and Finkel, 2014; Murphy et al., 2011, 2022; Sies et al., 2022). Low amounts of ROS produced by routine cellular metabolism contribute to positive (beneficial) stress or physiological ‘oxidative eustress’ (see Glossary). These ROS are involved in redox signalling and are required for normal biological function (Holmström and Finkel, 2014; Sies, 2017). Oxidative eustress is kept within homeostatic range by a complex series of ROS scavenging processes (see Glossary; Box 1) including enzymatic conversion reactions – where a protein reacts with ROS to convert it into a less damaging or reactive form – and antioxidant sinks that pre-emptively react with ROS to act as a redox buffer (see Glossary) to protect more sensitive cellular components (Birben et al., 2012; Mailloux, 2020). Other reactive species such as nitric oxide and hydrogen sulfide may also be involved in redox signalling, but we refer readers to other excellent reviews on this topic (Bryan et al., 2004; Bundgaard et al., 2020b; Fago and Jensen, 2015; Kolluru et al., 2017).

The difference between pathological oxidative distress and routine oxidative eustress is, therefore, a careful balance between the rate and amount of ROS generation and the capacity of the ROS scavenging systems to dismutate or remove that generated ROS (Box 1). This balance can be disturbed by several factors, including both environmental [e.g. high altitude (Dosek et al., 2007) and cold temperatures (Schmidt et al., 2002)] and biological [e.g. sleep apnoea (Lavie, 2009), ageing (Finkel and Holbrook, 2000) and starvation (Scherz-Shouval and Elazar, 2009)]. One context where the interaction between the environment and redox biology has been investigated in detail is the organismal response to oxygen variability, particularly periods of low oxygen (hypoxia) that can be followed by subsequent periods of reoxygenation (e.g. ischaemia–reperfusion). Oxygen levels modulate oxidative stress (Magalhães et al., 2005; Schild et al., 1997), and fluctuations in tissue O₂ supply are linked to several medically relevant pathologies (e.g. sleep apnoea, stroke and cardiac arrest; Eltzschig and Eckle, 2011) in mammalian models.

In this Review, we discuss the physiological interactions between O₂ variability and ROS with special reference to ectotherms and the potential effect of hypoxia tolerance. We discuss whether ROS is harmful to ectotherms in hypoxia, suggest general strategies that ectotherms may employ when exposed to environmental O₂ variability and hypothesize how ROS may contribute to hypoxic signalling in ectotherms.

The biochemical basis for the production of ROS (Box 1) and the consequences of oxygen deprivation associated with ischaemia–reperfusion (Box 2 and Fig. 1) are relatively well understood in clinically relevant models such as humans, pigs and mice (Martin et al., 2019) that are hypoxia intolerant. But whether animals generally accumulate oxidative damage as a result of fluctuations in environmental O₂ levels, and whether oxidative damage is linked to decreased fitness and/or survival, has not been convincingly experimentally demonstrated. However, it is foreseeable that the loss of function of key biological macromolecules such as proteins, nucleic acids and lipids would be detrimental (Costantini, 2019; Metcalfe and Alonso-Alvarez, 2010).

Generally, the effect of variable environmental O₂ levels on ROS production and oxidative damage has mostly been studied in hypoxia-tolerant species when they encounter hypoxia in poorly oxygenated water such as tide-pools (for aquatic ventilators), during diving (for air-breathing animals) or with overwintering, high altitude or burrowing. This large body of work shows variable responses to the oxidative damage markers caused by hypoxia and anoxia such as lipid peroxidation [e.g. thiobarbituric acid reactive substances (TBARS) and malondialdehyde (MDA)], protein carbonyl accumulation, DNA damage [e.g. terminal deoxynucleotidyl transferase dUTP nick end labelling (TUNEL) and comet assays] and redox status (e.g. glutathione oxidation ratio). These responses are dependent on the type of hypoxic exposure (e.g. severity, duration), the species and the tissue under investigation (reviewed in Hermes-Lima and Zenteno-Savín, 2002; Leveelahti et al., 2014). Although the literature is variable, there seems to be a trend indicating that ectotherms, in general, do not suffer from the same degree of oxidative damage after hypoxia as endothermic species (Borowiec and Scott, 2020; Bundgaard et al., 2018; Hermes-Lima and Zenteno-Savín, 2002; Ivanina and Sokolova, 2016; Wasser et al., 1992).

Whether hypoxia-tolerant species generally sustain less oxidative damage than hypoxia-intolerant species – and whether this is directly related to their hypoxia sensitivity – is also poorly understood. Furthermore, studies that directly compare hypoxia-tolerant and -intolerant species or hypoxia-acclimated and normoxia-acclimated individuals show equivocal evidence. For example, in a comparison of molluscs that differ in hypoxia tolerance, the less tolerant scallop (Argopecten irradians) showed more oxidative damage during anoxia–reoxygenation exposure than the more tolerant clam (Mercenaria mercenaria), as reflected by the concentrations of malondialdehyde (MDA) protein carbonyls in their tissues (Ivanina and Sokolova, 2016). The opposite, however, was observed in marine sculpins, where the more hypoxia-tolerant species showed a significant accumulation of TBARS in the brain during 6 h of hypoxia (that recovered to normoxic levels after 1 h reoxygenation) despite no evidence of mitochondrial H₂O₂ generation during the trial. In the same study, the less hypoxia-tolerant species showed no increase TBARS (Lau et al., 2019). This dichotomy highlights the need for rigorous comparisons between hypoxia-tolerant and -intolerant species to resolve whether hypoxia tolerance prevents oxidative damage. It also highlights the fact that ectothermy itself may be protective against oxidative stress, at least in some animals.

Even though the response to hypoxia is diverse, there is much to gain from understanding the different strategies that hypoxia-tolerant animals employ, especially as we do not know how oxidative eustress or oxidative distress factor into some of these adaptive strategies, and what distinguishes ectotherms in this regard from other animals.

While the effect of hypoxia tolerance on the avoidance of oxidative damage has been the most frequently studied, it seems clear that being an ectotherm is protective against oxidative damage following oxygen deprivation. Two different, but not mutually exclusive, strategies can be hypothesized to underlie this protection against oxidative damage in ectotherms: scavenging oxygen radicals and repairing oxidative damage to macromolecules (strategy 1) and/or low(ering) ROS generation (strategy 2) (Fig. 2).

Increasing scavenging capacity in response to hypoxia exposure [e.g. through an upregulation of antioxidant enzymes such as superoxide dismutase (SOD) and catalase; see Box 1] presumably allows for a higher threshold of ROS accumulation without leading to oxidative distress and enables the tissue to remain within the boundaries of oxidative eustress (Fig. 2B). This strategy has been the focus of studies on how antioxidant capacity changes in response to hypoxia or hypoxia–reoxygenation exposure in both endotherms – particularly diving seals and birds (Corsolini et al., 2001; Wilhelm Filho et al., 2002; Zenteno-Savín et al., 2002) – and ectotherms – including both vertebrates (e.g. freshwater turtles, cyprinid fish and frogs) and invertebrates (e.g. clams) (Hermes-Lima and Zenteno-Savín, 2002; Leveelahti et al., 2014).

If the assumption is that hypoxia or hypoxia–reoxygenation causes increased ROS generation, and that all effects of hypoxia-induced ROS are harmful, then one would expect animals that have adapted to O₂-variable environments to either constitutively express higher levels of antioxidants or show a consistent plastic response of increased antioxidants during hypoxic events. However, studies on diverse animals, such as freshwater turtles (Chrysemys and Trachemys species) (Hermes-Lima et al., 1995; Willmore and Storey, 1997), cyprinid fish (Lushchak et al., 2001), killifish (Du et al., 2016), garter snakes (Hermes-Lima and Storey, 1993), frogs (Hermes-Lima and Storey, 1996) and bivalves (Ivanina and Sokolova, 2016) and in various tissue types, including heart, brain, liver, kidney and skeletal muscle, have revealed that neither the first nor the second prediction is true in ectotherms. In fact, the response varies greatly across different antioxidant molecules and enzymes, animal species, tissue types and environments (Hermes-Lima and Zenteno-Savín, 2002; Leveelahti et al., 2014). Therefore, despite the research effort, there is little agreement in the literature as to how fluctuations in oxygen levels affect the activity of antioxidant enzymes in ectotherms. In contrast, diving mammals and birds have a higher antioxidant capacity than non-diving species. For example, plasma glutathione levels and the activity of antioxidant enzymes (SOD, glutathione peroxidase, glutathione reductase and catalase) are higher in seals (including elephant seals, Mirounga leonine) than in non-diving species such as pigs (Sus scrofa) (Wilhelm Filho et al., 2002; Zenteno-Savín et al., 2002). Both Adélie (Pygoscelis adeliae) and emperor (Aptenodytes forsteri) penguins maintain higher levels of total oxyradical scavenging capacity (TOSC) in the blood than non-diving birds such as south polar skuas (Catharacta maccormicki) and snow petrels (Pagodroma nivea) (Corsolini et al., 2001). This suggests that scavenging ROS and repair of oxidized macromolecules may be a more attractive strategy to avoid oxidative damage with O₂ fluctuations in endotherms with a relatively higher metabolic rate than ectotherms.

Secondly, as ROS production is often a by-product of aerobic metabolism, and ectotherms typically have much lower and more variable metabolic rates than endotherms (Nagy, 2005), baseline ROS production may be much lower in ectotherms compared with endotherms. Relatively few studies have directly measured ROS production in ectotherms, and those that have suggest that ectotherms may be less prone to ROS generation (and therefore oxidative damage) than endotherms as mentioned above (Borowiec and Scott, 2020; Bundgaard et al., 2023; Cochemé et al., 2011; Salin et al., 2017). Even with little or no change in the scavenging capacity or intrinsic changes to ROS-producing pathways, low baseline levels of ROS generation would decrease the likelihood of excess ROS accumulation in response to changes in environmental O₂ and oxidative distress would be avoided (Fig. 2C). Furthermore, animals may lower the potential amount of ROS generated when environmental O₂ levels change further by decreases in aerobic metabolism, either by intrinsic cellular changes or metabolic depression. This is the less investigated of the strategies, largely because it ideally involves the direct measurement of ROS production, which is technically challenging (Murphy et al., 2022). However, it is also a strategy that may be particularly agreeable for ectotherms, because of their low metabolic rate and a potentially much larger scope to reduce ROS generation than comparable endotherms – even at the cost of reduced aerobic metabolism.

Ectothermic metabolism can produce ROS even if the O₂ levels are lower than in endothermic animals, and consistent with the mammalian ischaemia–reperfusion model (see Box 1 and Fig. 1), succinate accumulation and ADP depletion have been observed in several ectotherms exposed to O₂ limitation (Hochachka and Storey, 1975), including anoxic turtles (Buck, 2000; Bundgaard et al., 2019a), anoxic crucian carp (Dahl et al., 2021) and hypoxic clams (Ivanina and Sokolova, 2016), albeit at much lower absolute concentrations than in mammals. However, unlike in hypoxia-intolerant mammals, succinate accumulation does not seem to be associated with a large amount of mitochondrial ROS accumulation upon reoxygenation in anoxic turtles (Bundgaard et al., 2018, 2023) or crucian carp (Lefevre et al., 2017) (G.Y.L. and Sjannie Lefevre, unpublished). One possible mechanism for this, observed in both turtles and carp, is preservation of the adenylate pool (particularly ADP; see Glossary) during anoxia exposure (Bundgaard et al., 2019a; Dahl et al., 2021). This would provide ATP synthase (complex V) with its substrate ADP and consequently would allow dissipation of the proton motive force, which then removes the potential for reverse electron transfer (Fig. 1C; see Glossary). However, in turtle hearts, inhibition of ATP synthase alone was not enough to induce reverse electron transfer upon reoxygenation (Bundgaard et al., 2023), suggesting that a low level of succinate accumulation is also essential to prevent oxidative distress.

The ability to avoid extensive succinate accumulation and preserve the adenylate pool during oxygen deprivation in these species is likely due to several factors. As ectotherms, they have a low basal metabolic rate, leading to a slower accumulation of metabolic end-products and a slower depletion of ATP. Also, because of their ectothermy, they can tolerate lowering of their body temperature, lowering their metabolic rate further (Bickler and Buck, 2007; Else and Hulbert, 1981; Nagy, 2005). Additionally, these hypoxia-tolerant species have large energy stores in the form of glycogen and an active circulation, which will ensure continued substrate for energy production in hypoxia further preventing depletion of adenine nucleotides (Bickler and Buck, 2007; Hyvärinen et al., 1985; Warren and Jackson, 2017).

Finally, another possible mechanism to prevent ROS production is to decrease the potential for mitochondrial H₂O₂ generation further with hypoxia or anoxia acclimation by intrinsic cellular changes. This strategy has been observed in turtles (Bundgaard et al., 2019b; Milton et al., 2007), killifish (Borowiec and Scott, 2020; Du et al., 2016), sablefish (Gerber et al., 2019) and shovelnose rays (Hickey et al., 2012), and suggests that intrinsic regulation of mitochondrial function is likely also important to prevent ROS generation upon reoxygenation. The mechanism behind this decrease is less clear, but could involve post-translational modifications of mitochondrial proteins (Bundgaard et al., 2018; Mathers and Staples, 2019), differential expression of protein isoforms (Lau et al., 2017; Little et al., 2018) and/or regulation of protein interactions such as supercomplex formation (Bundgaard et al., 2020a; Hutchinson et al., 2022). Another possible mitochondrial modification involves changes in proton conductance (Galli et al., 2013; St-Pierre et al., 2000), due to changes in fatty acid composition of the mitochondrial membrane (Almeida-Val et al., 1994; Gerber et al., 2019) or activation of uncoupling proteins or K_ATP channels (Pamenter et al., 2008; Staples and Buck, 2009).

Alterations in ROS levels and/or redox status can occur for a variety of reasons that may not directly result in oxidative distress (Ayala et al., 2014; D'Autréaux and Toledano, 2007; Hernansanz-Agustín et al., 2020; Holmström and Finkel, 2014; Murphy et al., 2011, 2022; Sies et al., 2022; Wong et al., 2008; Zarkovic et al., 2013). Correspondingly, an increase (or decrease) in ROS levels may be more representative of a change in signalling, perhaps even the induction of a defence response, as opposed to a direct mechanism that damages cells. Similarly, changes in glutathione or antioxidant enzyme activity may also be reflective of an effective response by the cell (e.g. using glutathione as a sink for ROS), as opposed to an inability to cope with an oxidative challenge (Sies, 2017).

One possible example of this was observed in the different responses of estuarine killifish (Fundulus heteroclitus) exposed to a single bout of hypoxia–reoxygenation, or acclimated to either diel intermittent hypoxia or continuous hypoxia (Borowiec and Scott, 2020). Fish exposed to a single hypoxia–reoxygenation bout showed a typical oxidative stress response with modest and transient changes in ROS homeostasis (as reflected by an increase in the ROS proxy dichlorofluorescein), redox status (increased oxidized glutathione), depletion of scavenging capacity and oxidative damage to lipids in the skeletal muscle. In contrast, acclimation to hypoxia led to substantial adjustments in ROS homeostasis and redox status, but no increase in oxidative damage. This suggests that the adjustments in ROS homeostasis and oxidative status associated with hypoxia acclimation were not reflective of oxidative stress, but instead are part of the responses that killifish use to cope with chronic hypoxia. Another example is the response of pyramidal neurons from crucian carp and turtles (Chrysemys picta), where GABA signalling changes in a species-specific manner both with exposure to hypoxia and with exposure to ROS scavengers, suggesting a role for decreases in ROS levels as an oxygen sensing signal (Buck et al., 2017; Hawrysh and Buck, 2019; Hogg et al., 2015; Pillai et al., 2021).

To our knowledge, the role of ROS and redox signalling as related to whole-animal hypoxia tolerance has not been directly tested. Hypoxia-tolerant animals may be more readily able to activate some signalling pathways to coordinate their response to O₂ limitation to prevent or mitigate subsequent damage associated with variation in O₂ availability. Alternatively, there could be adaptive traits associated with hypoxia tolerance (e.g. higher constitutive antioxidant capacity, lower background ROS generation) that would make rapid, plastic and potentially costly responses to acute hypoxia superfluous. Alternatively, several of the most studied signal transduction pathways related to hypoxia, such as hypoxia inducible factors (HIFs), AMP-activated protein kinase (AMPK), nuclear factor erythroid 2-related factor 2 (NRF2), Forkhead box O proteins (FOXOs), peroxisome proliferator-activated receptor gamma coactivator 1α (PGC1α) and apoptotic signalling, are all, at least in part, under redox regulation (Sies et al., 2022). Foreseeably, these could be modulated in hypoxia-tolerant ectotherms, potentially during an acclimation period. Below, we discuss some of these potential responses that could be regulated by ROS in ectotherms.

Both HIF-1α stability and expression can be induced by ROS to initiate mechanisms that help animals cope with hypoxia, such as blood vessel growth and increased glycolytic rate (Chandel et al., 2000). Interestingly, there is a functional divergence in the HIF-prolyl hydroxylase interactions across teleost fishes (Rytkönen et al., 2011), which could be associated with variance in hypoxia tolerance. HIF-1α mediates isoform switching in cytochrome c oxidase in mammals (Fukuda et al., 2007) to better maintain electron transport system (ETS) flux during O₂ limitation, subsequently limiting ROS generation. This isoform switch was not consistently observed in several teleosts and reptiles exposed to hypoxia/anoxia (Kocha et al., 2015), though the constitutively expressed COX4 isoform differed between tissues, which could be associated with better matching of ETS flux with changes in O₂ variability. Conversely, the absence of an induction of HIF-1α gene expression with hypoxia and anoxia in turtles suggests that HIF-1α signalling may not be beneficial in all hypoxia-tolerant organisms (Sparks et al., 2022).

ROS can indirectly lead to the activation of the cellular energy sensor and major metabolic regulator AMPK (Hinchy et al., 2018). AMPK and/or its downstream phosphorylation targets likely play an important role in the response to hypoxia exposure by targeting processes important in maintaining cellular energy balance (e.g. inhibiting protein synthesis and stimulating glycolysis; Emerling et al., 2009). AMPK phosphorylation occurs in anoxia-exposed crucian carp (Stenslokken et al., 2008) and hypoxia-exposed goldfish (Jibb and Richards, 2008), but not in molluscs exposed to hypoxia–reoxygenation (Ivanina et al., 2016), or in the heart tissue of hypoxia-exposed tilapia (Speers-Roesch et al., 2010). These varying responses in AMPK activation may reflect different uses of this signalling pathway in different organisms.

In mammals, increased ROS induces apoptosis in cells as a response to oxidative stress caused by the higher ROS production during reoxygenation (Circu and Aw, 2010; D'Autréaux and Toledano, 2007; Eefting et al., 2004; Wu et al., 2018). If an organism does not experience significant oxidative stress during hypoxia–reoxygenation, then these apoptotic pathways are not activated. For example, hypoxia-tolerant Pacific oysters (Crassostrea gigas) show blunted and delayed responses in transcripts involved in apoptosis and autophagy after exposure to hypoxia–reoxygenation than hypoxia-intolerant blue mussels (Mytilus edulis) (Falfushynska et al., 2020). Alternatively, moderate increases in ROS production (oxidative eustress) could activate apoptotic and cell proliferation processes, which may play an adaptive role during reoxygenation or during the preceding hypoxia response. For example, anoxic crucian carp showed signs of cell death during reoxygenation concomitant with an increase in proliferating cell nuclear antigen mRNA levels in the brain, suggesting an increase in neurogenesis processes upon recovery (Lefevre et al., 2017). Apoptosis also aids in gill remodelling in goldfish exposed to severe hypoxia, which results in a drastically increased surface area that presumably enhances oxygen extraction from the water (Sollid et al., 2005).

Encountering variable oxygen levels causes changes in cellular redox status in ectotherms, but whether this also translates to oxidative damage and a reduction in fitness is less clear. Here, we propose that ectothermy inherently protects against oxidative distress, most likely due to the low inherent potential for production of ROS. We argue that hypoxia-tolerant ectotherms may further reduce their capacity for mitochondrial ROS production by a combination of preventing the metabolic changes that drive oxidative distress (succinate accumulation and ADP depletion) and intrinsic changes to mitochondrial function. Furthermore, we suggest that changes in ROS production and oxidative damage markers with hypoxia may reflect oxidative signalling/eustress, which could be driven by classical signalling pathways such as AMPK, HIF-1α and NRF-2. We hope that this framework of thinking of ROS generation with variable O₂ levels not only as harmful but also as a potential signalling pathway will encourage further studies on the effect of ROS on the ecophysiology of animals. Increasingly accurate techniques for ROS measurements, combining in vivo and in vitro studies, and broader and more direct species comparisons will also help reveal the impact of ROS generation on hypoxia tolerance.

We thank Dr Rashpal Dhillon (University of British Columbia) for graphical design of the figures.