Rising CO₂ enhances hypoxia tolerance in a marine fish

Abstract

Global environmental change is increasing hypoxia in aquatic ecosystems. During hypoxic events, bacterial respiration causes an increase in carbon dioxide (CO₂) while oxygen (O₂) declines. This is rarely accounted for when assessing hypoxia tolerances of aquatic organisms. We investigated the impact of environmentally realistic increases in CO₂ on responses to hypoxia in European sea bass (Dicentrarchus labrax). We conducted a critical oxygen (O_2crit) test, a common measure of hypoxia tolerance, using two treatments in which O₂ levels were reduced with constant ambient CO₂ levels (~530 µatm), or with reciprocal increases in CO₂ (rising to ~2,500 µatm). We also assessed blood acid-base chemistry and haemoglobin-O₂ binding affinity of sea bass in hypoxic conditions with ambient (~650 μatm) or raised CO₂ (~1770 μatm) levels. Sea bass exhibited greater hypoxia tolerance (~20% reduced O_2crit), associated with increased haemoglobin-O₂ affinity (~32% fall in P₅₀) of red blood cells, when exposed to reciprocal changes in O₂ and CO₂. This indicates that rising CO₂ which accompanies environmental hypoxia facilitates increased O₂ uptake by the blood in low O₂ conditions, enhancing hypoxia tolerance. We recommend that when impacts of hypoxia on aquatic organisms are assessed, due consideration is given to associated environmental increases in CO₂.

Introduction

A lack of O₂ is one of the greatest challenges that most life can face. In terrestrial ecosystems conditions of low O₂ are rare. In contrast low O₂, referred to as hypoxia, is much more common in freshwater and marine ecosystems^1,2,3,4. Hypoxia occurs because high biological demand for O₂ can exceed the rate of O₂ supply to the ecosystem, leading to a reduction in environmental O₂ levels^5,6. However, the challenges of hypoxia are not solely a result of reduced O₂. Organisms must also contend with simultaneous but reciprocal changes in the other respiratory gas, CO₂.

When O₂ decreases in aquatic systems there is a corresponding increase in CO₂^7,8. This is a by-product of respiration, the same process that causes depletion of O₂. As such high CO₂ during hypoxia is ubiquitous and unavoidable. This coupling of O₂ and CO₂ has been highlighted numerous times in oceanographic sciences, most recently by Robinson⁵. Yet unaccountably, despite the known link between decreasing O₂ and increasing CO₂ during hypoxia^8,9, the issue of increased environmental CO₂ during periods of low O₂ has been relatively overlooked by biologists.

Implications of rising CO₂ during hypoxia on aquatic organisms are particularly important to address in the face of human driven climate change. Hypoxic areas are predicted to become more common and more severe, particularly in marine systems, with the de-oxygenation of the world’s oceans recently highlighted as a major component of climate change^2,3,10,11,12. In addition, there will be an increase in ambient CO₂ as rising atmospheric CO₂ is absorbed by the world’s oceans¹³. Non-linear interactive effects between higher atmospheric CO₂ and CO₂ accumulation during hypoxia will lead to increased CO₂ levels during hypoxia in future oceans¹⁴. This means that effects of rising CO₂ during hypoxia in marine systems will be amplified by climate change.

Typically, experiments which test responses to hypoxia or impacts from hypoxia on aquatic organisms create hypoxic conditions by off-gassing oxygen from water by gassing with pure nitrogen or a mix of nitrogen (N₂) and O₂ (for examples see^15,16,17,18). This creates low O₂ conditions without the concurrent CO₂ increase that would be expected in the environment. The lack of studies in which an environmentally realistic simultaneous decrease in O₂ and increase of CO₂ have been conducted may lead to mismeasurement of responses to hypoxia. Recently, several studies on marine fish and invertebrates have demonstrated interactive effects of low oxygen and increased CO₂^9,19,20, with some species exhibiting loss of equilibrium (LoE) and death at higher O₂ concentrations when CO₂ is simultaneously elevated^19,21. However, these experiments do not give insight into the physiological mechanisms underlying the influence of CO₂ on hypoxic responses of fish.

Previously observed impacts of hypoxia-associated rises in CO₂ on hypoxia tolerance of fish could be a result of changes in O₂ uptake, as CO₂ has been shown to impact upon several aspects of organismal biology that are involved in O₂ uptake and transport^22,23. We aimed to assess whether concurrent increases in CO₂ during decreases in O₂ affect O₂ uptake in a marine fish, the European sea bass (Dicentrarchus labrax) by conducting a standard critical O₂ level test (O_2crit). Under normal O₂ (normoxic) conditions fish maintain a minimum level of O₂ consumption rate (${\dot{{\rm{M}}}{\rm{O}}}_{2}$), referred to as the standard metabolic rate (SMR), in order to meet maintenance energetic demands of essential processes through aerobic respiration²⁴. As the level of O₂ in water drops fish deploy a number of responses (i.e. increased ventilatory water flow and cardiac output, increased haematocrit, functional changes in gill morphology, changes in Hb-O₂ affinity) in order to maintain and regulate this minimum level of ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ ^25,26,27. If environmental O₂ continues to drop there comes a point at which fish are unable to regulate ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ to meet minimal energy demands, referred to as the critical O₂ level (O_2crit). At O₂ levels below O_2crit fish become oxy-conformers (where ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ is directly proportional to environmental O₂ availability) and fish become increasingly reliant on anaerobic metabolism which is unsustainable in the medium to long term. In the past the measure of O_2crit has been used as a proxy for overall hypoxia tolerance but recently this approach has been questioned^28,29. Nevertheless O_2crit does provide information related to the ability of fish to maintain O₂ uptake and supply during hypoxia, and its prevalence in the literature allows comparison of responses between species³⁰. Furthermore, we investigated whether any changes in O_2crit could be linked to changes in blood acid-base chemistry and blood gas transport via alteration of Hb-O₂ binding caused by rising environmental CO₂. Our hypothesis was that the simultaneous increase in CO₂ during a progressive decrease in O₂ would decrease hypoxia tolerance (increase O_2crit) and that this response may be a result of blood acid-base disturbance decreasing Hb-O₂ affinity and O₂ transport.

Results

O_2crit tests

There was evidence of enhanced tolerance to hypoxia in sea bass exposed to rising compared to constant CO₂ conditions. This was indicated by measurements of O_2crit in European sea bass being significantly different between fish exposed to either constant, or rising CO₂ levels during O_2crit tests when accounting for variation in SMR (Fig. 1; ANCOVA, F_1,12 = 7.525, p = 0.0178). A CO₂ increase during O_2crit tests resulted in a 20% reduction of O_2crit (3.88 ± 0.19 kPa O₂, 18.7 ± 0.9% air saturation, mean ± S.E.) when compared to tests in which CO₂ levels were maintained at ambient levels (4.87 ± 0.22 kPa O₂, 23.4 ± 1.1% air saturation, mean ± S.E.).

Figure 1

Calculated critical oxygen level (O_2crit) of European sea bass, Dicentrarchus labrax, when O_2crit tests include a constant ambient CO₂ level (~500 μatm, N = 8) or an ecologically realistic rise in CO₂ (~500–2500 μatm, N = 7) during the test. *Indicates significant difference between CO₂ regimes (p < 0.05). Boxes represent median value and inter-quartile range, whiskers represent minimum and maximum values.

Blood chemistry analysis

A comparison of blood chemistry parameters between the two treatment groups indicated that sea bass fully compensated for the rise in CO₂ during hypoxia within 5 hours (the period of exposure prior to blood sampling). Blood pH (pH_e) was not different between fish exposed to constant ambient CO₂ (7.87 ± 0.03) and fish exposed to progressively rising CO₂ (7.88 ± 0.02) (Fig. 2A, GLM, F_1,14 = 0.23, p = 0.64). The acidifying effect of the ~79% rise in blood pCO₂ levels in the rising CO₂ regime (0.272 ± 0.019 kPa CO₂) compared to the constant ambient CO₂ regime (0.152 ± 0.015 kPa CO₂) (Fig. 2C, GLM, F_1,13 = 23.9, p < 0.001) was fully compensated by elevating blood HCO₃⁻ (Fig. 2D, GLM, F_1,13 = 40, p < 0.001). Plasma HCO₃⁻ was 88% higher under the rising CO₂ regime (6.76 ± 0.29 mM) when compared to the constant ambient CO₂ (3.60 ± 0.43 mM). There were no differences in haematocrit (Fig. 2B, general linear model, F_1,13 = 0.69, p = 0.42) or plasma lactate (Fig. 2E, general linear model, F_1,14 = 1.48, p = 0.24) between fish sampled under a constant ambient CO₂ regime (haematocrit = 39.4 ± 0.8%, lactate = 0.88 ± 0.26 mM) or rising CO₂ regime (haematocrit = 38.2 ± 1.1%, lactate = 0.53 ± 0.12 mM). Blood glucose levels were ~26% lower in fish exposed to a progressively rising CO₂ regime (4.50 ± 0.53 mM) when compared to a constant ambient CO₂ regime (6.09 ± 0.31 mM) (Fig. 2F, GLM, F_1,14 = 6.74, p = 0.021).

Figure 2

Blood chemistry characteristics of European sea bass sampled at ~8.4 kPa O₂ (~40% air saturation) following a progressive O₂ decline accompanied by either constant CO₂ (~650 μatm CO₂, N = 8) or a progressive increase in CO₂ (sampled at ~1770 μatm CO₂, N = 8). Blood pH (A), haematocrit (B), plasma lactate (E) and plasma glucose (F) were directly measured whilst blood pCO₂ (C) and plasma HCO₃⁻ (D) were calculated (see Methods for details). *Indicates statistical significance with p < 0.05 and ***indicates statistical significance with p < 0.001. Boxes represent median value and inter-quartile range, whiskers represent minimum and maximum values.

Haemoglobin affinity for oxygen

Oxygen affinity of haemoglobin was increased in fish sampled under the progressively rising CO₂ regime (Fig. 3A, GLM, F_1,12 = 10.42, p = 0.0073). Haemoglobin P₅₀ was decreased by ~32.5% in fish under the progressively rising CO₂ regime (1.64 ± 0.15 kPa O₂, 7.9 ± 0.7% air saturation) compared to fish sampled with a constant ambient CO₂ (2.43 ± 0.20 kPa O₂, 11.7 ± 1% air saturation). There was no significant change in Hills number between treatments (Fig. 3B, GLM, F_1,12 = 0.50, p = 0.494).

Figure 3

Haemoglobin P₅₀ (A) and Hills number (B) for fish sampled at ~8.4 kPa O₂ (~40% air saturation) following a progressive O₂ decline accompanied by either constant CO₂ (~650 μatm CO₂, N = 7) or a progressive increase in CO₂ (sampled at ~1770 μatm CO₂, N = 7). Measurements were made using a gas mix which matched the calculated blood pCO₂ of each individual fish blood sample. **Indicates statistical significance with p < 0.01. Boxes represent median value and inter-quartile range, whiskers represent minimum and maximum values.

Discussion

Our results highlight the biological importance of simultaneously rising CO₂ under conditions where O₂ levels in water are depleted – a scenario that reflects the natural conditions during hypoxia which will be exacerbated by climate change – by demonstrating that ecologically relevant changes of CO₂ impact physiological performance of a marine fish at both the molecular and whole organism level. We hypothesised that rising CO₂ during progressive O₂ decreases would lead to an increase in O_2crit as a result of increased blood CO₂, decreased blood pH and the associated Bohr/Root effect of fish haemoglobin (Hb), in which Hb-O₂ affinity (Bohr effect) and the total capacity of Hb for O₂ (Root effect) are reduced when pH falls. In contrast, we show that increasing CO₂ as O₂ declined led to enhanced hypoxia tolerance of sea bass with a 20% lower critical oxygen level (Fig. 1). This change in whole organism hypoxic response was accompanied by an increase in Hb-O₂ affinity of blood cells in fish exposed to concurrent CO₂ rises (Fig. 3). The change in Hb-O₂ affinity was not accompanied by a change in blood pH (although the one hour acclimation to treatment pCO₂ prior to blood sampling may have contributed to this result). This provides a potential mechanistic basis to explain improved O_2crit, enabling sea bass to enhance O₂ uptake during hypoxia and thus maintain normal aerobic metabolism to lower environmental O₂.

The driver of increased Hb-O₂ affinity in sea bass exposed to concurrent O₂ decline and CO₂ rise is not clear from our results. Several allosteric factors that can modulate the affinity of haemoglobin for O₂ could be involved, including pH, organic phosphates and inorganic ions. Fish haemoglobin is highly sensitive to pH, which modulates Hb-O₂ affinity and carrying capacity via the Bohr and Root effects³¹, but we found no differences in blood pH of sea bass between treatment groups. In addition, the in-vivo increase in pCO₂ in fish exposed to concurrent CO₂ rises during hypoxia led to an opposite response of Hb-O₂ affinity than would be expected by an in-vitro rise in pCO₂ which would result in pH induced Bohr/Root effects. Increased Hb-O₂ affinity could result from increased intracellular pH of erythrocytes³², as acute hypoxic exposure has been shown to stimulate a β-adrenergic stimulated increase in intracellular erythrocyte pH in rainbow trout³³. Alternatively, increased Hb-O₂ affinity could be due to decreased red cell nucleoside triphosphates (NTPs)^31,34, a known hypoxia adaptation, but this can take more than 6 days to complete³⁵. Sea bass may also have a particularly strong β-adrenergic response and/or fast NTP response, although there is little evidence to suggest this because P₅₀ measurements from fish exposed to hypoxia at ambient CO₂ levels do not differ from P₅₀ measurements in normoxic fish from the same population (Montgomery et al., Unpublished data). It is possible that rising CO₂ during hypoxia may modulate the β-adrenergic response and/or the red cell NTP response within the time frame (~4–6 hours) of our treatments. However, direct measurements of red cell pH_i and NTP content of sea bass in each treatment group would be needed to confirm this.

A third possible driver of increased Hb-O₂ affinity in sea bass exposed to concurrent CO₂ rise during hypoxia could be decreased erythrocyte chloride (Cl⁻)³⁶. Although erythrocyte Cl⁻ was not directly measured in our study, plasma HCO₃⁻ was approximately 3 mM higher in fish exposed to concurrent CO₂ rises during hypoxia than fish which experienced constant ambient CO₂ during hypoxia (Fig. 2D). The higher HCO₃⁻ in fish exposed to rising ambient CO₂ during hypoxia is likely a result of rapid compensation for a respiratory acidosis due to rising blood pCO₂. This change in plasma HCO₃⁻ is typically mirrored by a reciprocal change in plasma Cl^− 37,38 which is likely to be followed by a similar decline in erythrocyte Cl⁻.

Analysis of O_2crit is a common measure of hypoxia tolerance in fish but concurrent CO₂ increases during hypoxia have been generally unaccounted for. A recent meta-analysis by Rogers et al.³⁹ constructed a database of O_2crit research of fish (both freshwater and marine). This analysis identified two broad methods employed in O_2crit measurements:

• 1. Closed respirometry where O₂ is usually reduced by the O₂ consumption of the fish (52 identified studies) or;

• 2. Intermittent or flow-through respirometry in which O₂ is usually reduced via gassing with pure N₂ or combined N₂ and O₂ mixes (32 identified studies).

The use of closed respirometry in the majority of studies would result in concurrent CO₂ rises as O₂ is depleted by fish O₂ consumption. The increase in ambient CO₂ during closed respirometry is well known and often used as a criticism of this respirometry technique⁴⁰. In contrast, use of intermittent-flow respirometry in O_2crit trials normally necessitates the reduction of O₂ in the water by aeration with N₂ or a mix of N₂ & O₂. As a result CO₂ would likely decrease during the time course of hypoxia induction (as the gas mixture would contain zero CO₂, rather than ~400 µatm present in atmospheric air). Such a change in CO₂ during the O_2crit trial would be the opposite of that seen in nature. Therefore it may be considered that closed respirometry provides conditions which give a more environmentally relevant measure of O_2crit²⁸.

As our results indicate that rising CO₂ during hypoxia directly affects the ability of sea bass to maintain O₂ uptake, it could be expected that the use of closed respirometry methods would result in lower measurements of O_2crit than intermittent-flow methods for the same species. This effect has not been documented for species in which a direct comparison has been made – with either there being no effect of respirometry method on O_2crit^39,41 or higher O_2crit measurements when closed respirometry is used⁴¹. However, such comparisons are complicated by differences in the rate of hypoxia induction (RHI) by different studies, which in turn will influence how much time fish have to regulate blood pH when ambient CO₂ is rising. For example, Regan and Richards⁴¹ have demonstrated that the faster rates of hypoxia induction (RHI) typical of closed respirometry O_2crit trials lead to higher values of O_2crit (i.e. lower hypoxia tolerance) when compared with longer trials using slower RHI’s typical of the intermittent-flow method. The effect of RHI on O_2crit was proposed by Regan and Richards⁴⁰ as a potential explanation of the results of Snyder et al. (i.e. higher O_2crit in closed respirometry compared to intermittent-flow respirometry)⁴². The speed of RHI during closed respirometry will also effect the speed of CO₂ rise. Almost all studies using closed respirometry to measure O_2crit do not report changes in CO₂ over the course of measurement period. When accumulation of CO₂ during a closed respirometry O_2crit trial was measured by Regan and Richards, CO₂ levels were ~8,000 μatm after ~90 minutes⁴¹. However, we should note that this pCO₂ level was measured after the O_2crit point, when anaerobic metabolism continues to produce CO₂ in the absence of O₂ consumption, but also metabolic acid production and excretion further drives up water pCO₂ in the respirometer due to excess H⁺ ions titrating ambient HCO₃⁻ to CO₂. Regardless, at O₂ levels above O_2crit the rate of CO₂ onset will be faster than those used in our current study as a result of the faster RHI. Increased speed of CO₂ onset in closed respirometry trials may ‘outstrip’ the ability of fish to acid-base regulate, causing an uncompensated respiratory acidosis during the time of the trial, which in turn would decrease Hb-O₂ affinity via the Bohr & Root effects and potentially increase O_2crit. Similarly, fish species which have reduced ability to acid-base regulate may have an increased O_2crit when rising CO₂ is included in trials.

Our results indicate improved hypoxia tolerance during rising CO₂ in European sea bass. This contrasts with previous research investigating interactive effects of CO₂ and hypoxia on O_2crit of fish. Woolly sculpin, Clinocottus analis (an intertidal species that can breathe air), exposed to ~1100 μatm CO₂ showed no impact on O_2crit after 7 days acclimation but after 28 days had O_2crit measurements ~34% higher than fish held in ambient (~400 μatm) conditions⁴³. Higher O_2crit after 28 days corresponded with higher RMR and Na⁺, K⁺, ATPase activity. This contrast in results could indicate that the beneficial effect of acute rises in CO₂ associated with natural hypoxia documented in our study are potentially reversed when fish are exposed to long term constantly high CO₂ associated with anthropogenic climate change. Moreover, acute changes in CO₂ had no effect on O_2crit of the estuarine fish species mummichog, Fundulus heteroclitus, and Norfolk spot, Leiostomus xanthurus when they were exposed to ~8,000–10,000 μatm CO₂ immediately prior to an O_2crit trial⁴⁴. As such the effect of CO₂ on O_2crit will likely depend on differences in physiological responses to CO₂ and O₂ between species.

Simultaneously rising CO₂ also shows variable impacts on non-metabolic responses to hypoxia of several species. Cycling CO₂ had no effect on aquatic surface respiration (ASR), the use of the thin surface layer of water for aquatic respiration⁴⁵, or survival in juvenile Menidia menidia, Fundulus majalis, Fundulus heteroclitus or Morone saxatalis exposed to short term cycles of O₂⁴⁶. In contrast, combined hypoxia and acidification resulted in an increase in the O₂ level at which Menidia menidia and Menidia beryllina first performed ASR, consistently performed ASR, exhibited LoE, and finally died²¹. Additionally, combined high CO₂ (~2,000 μatm) and hypoxia had no effect on survival of larval Cyprinodon variegatus, an additive negative effect on larval Menidia beryllina, and a synergistic negative effect on larval Menidia menidia⁴⁷. This variation in effect of CO₂ on hypoxia responses could be a result of methodological differences (e.g. constant high CO₂ in Dixon et al.⁴⁶, cycling DO/pH in DePasquale et al.⁴⁷, and concurrent CO₂ rise/O₂ decrease in Miller et al.²¹), the level of CO₂ used in studies (e.g. CO₂ levels used by Miller et al.²¹ were ~23,000 μatm which is much higher than levels likely to be commonly found in the environment during hypoxia and may have contributed to the negative effects of rising CO₂ noted in the study), differences in species and life stages used (changes in physiological tolerance across life stages have been noted for thermal tolerance by Komoroske et al.⁴⁸), or possibly variability in response as a result of differences in previous environmental experience⁴⁹. The role of environmental variability in species sensitivities to CO₂ has recently been outlined in the proposed Ocean Variability Hypothesis (OVH)⁵⁰ and warrants testing on various species in the future.

Overall our results indicate that the environmentally realistic, simultaneous rises in CO₂ during a hypoxic event increased the hypoxia tolerance (i.e. reduces O_2crit) of European sea bass which is at least partly explained by an enhanced ability of fish to uptake O₂ via increased Hb-O₂ affinity. Miller et al.²¹ also demonstrated impacts of concurrent CO₂ rise on measurements of hypoxia tolerance, although in an opposite direction to that noted in our study. As concurrent CO₂ rises during hypoxia are the norm in nature, evidence that this affects physiology of organisms exposed to hypoxia highlights an important shortcoming of research to predict tolerances to hypoxia of fish. More research on this issue is needed to clarify how common this modifying effect of CO₂ on the response to hypoxia is and whether such measurements in the lab are ecologically relevant. A greater understanding of this issue may allow more accurate assessments of the impacts of hypoxic events on marine fish in nature, aiding management and conservation of fish species. With specific regard to measurements of O_2crit we believe future studies should include concurrent rising CO₂ in the following ways:

• 1. Intermittent-flow respirometer studies should include increases of CO₂ relevant to hypoxic events that organisms may experience,

• 2. Closed respirometry studies should report the start and end CO₂ levels in the respirometer.

In addition, both methods should aim to create an environmentally relevant rate of hypoxia induction/CO₂ increase for the species studied, and consistently report CO₂ levels measured. By incorporating these recommendations we believe that future studies of O_2crit will give more representative estimations of species hypoxia tolerance.

Materials and Methods

Fish collection and husbandry

We collected juvenile sea bass from estuaries and coastal lagoons on the south Dorset coast and Isle of Wight in June 2017 (Marine Management Organisation permit #030/17 & Natural England permit #OLD1002654). Prior to experimentation, these fish were held in the University of Exeter’s Aquatic Resource Centre in an aerated recirculating aquaculture system and fed a commercial pellet at a ration of ~1–2% body weight per day three times a week (for system water chemistry see Table 1). All fish were starved for a minimum of 72 hours prior to the start of all measurements to ensure their metabolism was not affected by digestion (i.e. specific dynamic action²⁴). All experimental procedures were carried out under home office licence P88687E07 and approved by the University of Exeter’s Animal Welfare and Ethical Review Board.

Table 1 Water chemistry parameters of the recirculating aquaculture system in which sea bass were held prior to experimental work (means ± S.E. shown).

Measuring hypoxia tolerance

We determined oxygen consumption rates (${\dot{{\rm{M}}}{\rm{O}}}_{2}$) of sea bass using an intermittent-flow respirometer system. The respirometer system set up followed recommendations set out by Svendsen et al.⁵¹. Briefly, the system comprised of a sealed 4.515 L respirometer chamber connected to a recirculating loop, including an in-line recirculating pump (Eheim universal 600, Deizisau, Germany), and a measurement chamber into which a temperature-compensated fibre optic oxygen optode (Firesting O₂ oxygen meter, Pyroscience GmBH, Germany) was placed. Oxygen optodes were calibrated in water at the start of experiments at 100% air saturation and 0% air saturation according to manual instructions. Respirometry was conducted in a semi-closed system consisting of three 100 L experimental tanks fed by a 100 L sump, with overflowing water from the experimental tanks recirculating back to the sump. A second pump was used to periodically flush the respirometer system with water from the surrounding tank. This pump was controlled by an automated computer program (AquaResp 3, AquaResp®) to intermittently flush the respirometer. Five respirometer chambers were distributed between the three experimental tanks (maximum of 2 chambers per experimental 100 L tank). The sump was temperature controlled (18.27 ± 0.02 °C, mean ± S.E.) using a heater/chiller unit (Grant TX150 R2, Grant Instruments, Cambridge, UK) attached to a temperature exchange coil. Together these tanks formed a 400 L system with the same temperature, oxygen and water chemistry parameters for all respirometers.

Individual sea bass (average mass = 131.2 ± 7.5 g), chosen at random, were placed inside the respirometers and allowed an overnight recovery period, for a minimum of 13 hours, before O_2crit tests began. While sea bass were in the respirometers measurements of ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ were conducted every 10 minutes, including a flush period of 300 s, a wait period of 60 s and a measurement period of 240 s. During the wait and measurement period the chamber was sealed by switching off the flush pump and the decline in dissolved O₂ within the chamber was continuously measured by the fibre optic O₂ electrode.

Following the overnight recovery period a standard O_2crit test was conducted. Oxygen levels in the respirometer system were reduced from ~100% air saturation to ~15% air saturation over the course of 6 hours (decline in O₂ was ~20% air saturation per hour between ~100% air saturation and ~40% air saturation and ~10% air saturation per hour from ~40% air saturation until the end of trials). Oxygen was regulated by gassing the sump and experimental tanks with a mix of N₂ and O₂ (G400 Gas mixing system, Qubit Biology Inc.) at a rate of 10 L min⁻¹ following a pre-set automated protocol (Flowvision, Alicat software). Levels of CO₂ within the system were controlled under one of two treatments (with 8 fish exposed to Treatment 1 and 8 separate fish exposed Treatment 2):

• Treatment 1 (Constant CO₂) - ambient levels of CO₂ were maintained by including 0.04% CO₂ as part of the gas mix delivered to the respirometer system.

• Treatment 2 (Rising CO₂) - the proportion of CO₂ in the gas mix was gradually increased as O₂ was decreased. This increase in CO₂ was designed to reflect environmentally realistic increases in CO₂ predicted as a result of depletion of O₂ by bacterial respiration (assumed respiratory quotient of 1)⁵², using the seawater carbonate chemistry calculator CO2sys (see supplementary material for predictions of increased CO₂ during hypoxia).

Water chemistry of treatment 2 was monitored once per hour by measuring pH_NBS, temperature and salinity as well as taking a 12 mL water sample to measure total CO₂ (TCO₂)/Dissolved Inorganic Carbon (DIC). Water chemistry of treatment 1 was monitored at the start and end of the treatment to ensure no change in water pCO₂ occurred over the time course of the O_2crit trial. Seawater DIC analysis was conducted using a custom built system described in detail by Lewis et al.⁵³. These four parameters were then input into the seawater carbon calculator programme, CO2SYS to calculate pCO₂ based on the NBS pH scale, equilibration constants from Mehrbach et al. refitted by Dickson and Millero, and KSO₄ dissociation constants from Dickson. The reciprocal changes in O₂ and CO₂ during O_2crit tests for each treatment are illustrated in Fig. 4. O_2crit tests were stopped once a minimum of 3 ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ measurements showed a transition from an oxy-regulating to oxy-conforming state for each fish or fish showed a large drop in ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ and signs of distress in the respirometer. No fish exhibited LoE during trials. Following completion of O_2crit trials experimental tanks were aerated with ambient air to swiftly restore O₂ and CO₂ levels.

Figure 4

Changes in partial pressure of O₂ (expressed as kPa O₂) and CO₂ (μatm) during (A) two O_2crit trials representing treatment 1 in which O₂ was reduced with no change in CO₂; and (B) two trials representing treatment 2 where O₂ was reduced with a corresponding rise in CO₂. Data presented are means ± S.D.

Sea bass were left to recover in respirometers, for a minimum of 1 hour post-trial, until O₂ levels reached ~21 kPa O₂ (~100% air saturation). Fish were then removed from respirometers and background respiration was measured for a minimum of 1 hour (6 measurement cycles) for all respirometers immediately post trial.

Oxygen consumption rate (${\dot{{\rm{M}}}{\rm{O}}}_{2}$) analysis

Following each 240 s measurement period ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ was automatically calculated by the AquaResp3 software. A linear regression was fitted to the O₂ versus time data for each measurement period. The slope of this regression (s, kPa O₂ h⁻¹) was then used to calculate ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ (mg O₂ kg⁻¹ h⁻¹) using the equation outlined by Svendsen et al.⁵¹:

$$\dot{{\rm{M}}}{{\rm{O}}}_{2}=s{V}_{resp}\,\alpha {m}^{-1}$$

where V_resp is the respirometer volume minus the volume of the fish (L), α is the solubility of O₂ in water (mgO₂ L⁻¹ kPa⁻¹) for the relevant salinity and temperature, and m is the mass of the fish (kg). Calculations of ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ where s had a R² of <0.98 were removed from subsequent analysis. For the purpose of establishing O_2crit values from a plot of ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ versus ambient O₂ level, the oxygen saturation of each measurement period was defined as the average of the dissolved O₂ measurement over the measurement period. The average background respiration, over the 1 hour post-trial measuring period, for each respirometer (average background respiration was <2% of fish ${\dot{{\rm{M}}}{\rm{O}}}_{2}$) was subtracted from ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ measurements to correct fish ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ for background respiration.

We calculated the standard metabolic rate (SMR) in R v3.5.3⁵⁴ using function calcSMR in package fish ${\dot{{\rm{M}}}{\rm{O}}}_{2}$²⁴. All ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ values from the overnight recovery period and beginning of the O_2crit trial in which average dissolved O₂ saturation was >80% air saturation were included for SMR calculations. This resulted in approximately 14–16 hours of ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ data to calculate SMR from for each fish. We estimated SMR for all fish using the mean of the lowest 10 ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ measurements during the overnight period. Although Chabot et al.²⁴ recommend use of the mean of the lowest normal distribution (MLND) or quantile (q = 0.2) methods to calculate SMR we chose the mean of the 10 lowest values as it produced a value of SMR that most accurately matched the consistent low ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ measurements of oxy-regulating fish at values of pO₂ above the O_{2crit_}point. Additionally, for each fish the coefficient of variation in the mean lowest normal distribution (MLND) was assessed using a ROUT test (Q = 1%) to check whether variation in low ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ measurements was consistent between fish (to account for potential differences in activity in the respirometer prior to O_2crit trials). The ROUT test removed one fish from the rising CO₂ treatment which displayed a CoV of 34.6. Mean CoV of the remaining 15 fish was 11.85 ± 1.26 (±S.E.M).

We then used function calcO2crit from the package to calculate the O_2crit for each individual fish, using the estimated SMR of each individual, as detailed in the supplementary material of Claireaux & Chabot⁵⁵. This function identifies the portion of the O_2crit test where metabolic rate data follows an oxygen conforming relationship and fits a linear regression line through this data, O_2crit is then calculated as the oxygen level at which this regression line crosses the calculated SMR of the individual fish. Plots of ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ against O₂ during O_2crit trials showing calculated O_2crit for each individual can be seen in the supplementary material (Supplementary Material Figs 1–16). Calculations were conducted using a gap limit of 0.83 kPa O₂ (4% air saturation) and a maximum number of 7 ${\dot{{\rm{M}}}{\rm{O}}}_{2}$ points to fit the regression line through the oxygen conforming component of the data used to estimate O_2crit.

Measuring blood chemistry

Following O_2crit trials sea bass were moved to individual 7 L chambers, which were aerated and supplied with seawater from the aquarium re-circulating system at a rate of about 4 L min⁻¹. After an overnight acclimation period we then exposed the fish to a decrease in O₂ levels to ~6.4 kPa O₂ (30% air saturation) over a period of 4 hours (equivalent to the rate of decrease in O₂ used in the previous O_2crit tests). We chose to blood sample fish at an O₂ level above O_2crit to ensure that anaerobic metabolism did not influence blood chemistry. This was combined with the same CO₂ regime each fish experienced during the O_2crit test (i.e. either constant or reciprocally rising CO₂). Once an O₂ level of ~6.4 kPa O₂ (30% air saturation) was achieved the fish were allowed to acclimate for 1 hour before being individually anaesthetised in situ using a dose of 100 mgL⁻¹ of benzocaine. Once fish were judged to be sufficiently anaesthetised (cessation of gill ventilation and lack of response to a pinch of the anal fin) they were immediately transferred within 5 seconds to a gill irrigation table (with the same pO₂ and pCO₂ levels of the respective treatment), where anaesthesia was maintained with a dose of 37.5 mgL⁻¹ of benzocaine. Gill ventilation was artificially maintained by a micro-pump, so that the operculum were just open and exhalant water flow could just be visualised. Once a stable gill water flow was established blood was sampled by caudal vessel puncture using a 1 ml heparinised syringe. This method has been demonstrated to obtain accurate measurements of blood chemistry parameters comparable to those achieved using cannulation (Davison & Wilson, University of Exeter, personal communication). At the time of blood sampling water pCO₂ was 656 ± 44 μatm (mean ± S.E.) for fish in the constant ambient CO₂ regime and 1763 ± 43 μatm (mean ± S.E.) for fish in the progressively rising CO₂ regime. Water pO₂ was 8.1 ± 0.2 kPa (mean ± S.E., ~38% air saturation) for fish in the constant ambient CO₂ regime and 8.7 ± 0.2 kPa (mean ± S.E., ~41% air saturation) for fish in the progressively rising CO₂ regime. Following blood sampling fish were transferred to seawater isolation tanks containing ~20.8 kPa O₂ (100% air-saturated) to recover from the anaesthetic. They were then monitored over a period of 24 hours before we returned them to their original holding tanks.

Immediately after sampling, whole blood pO2 was measured at 18 °C in a temperature-controlled system (Strathkelvin 1302 electrode and 781 meter; Strathkelvin Instruments Ltd, Glasgow, UK). We measured extracellular pH on 30 µL of whole blood using an Accumet Micro pH electrode and Hanna HI8314 pH meter at 18 °C calibrated to pH_NBS 7.04 and 9.21 specific buffers. Three 75 µL micro capillary tubes were then filled with whole blood and sealed with Critoseal capillary tube sealant (Fisher) and paraffin oil and centrifuged for 2 minutes at 10,000 rpm. Haematocrit was measured using a Hawksley micro-haematocrit reader. Plasma was extracted from these tubes for analysis of TCO₂ using a Mettler Toledo 965D carbon dioxide analyser. Plasma pCO₂ and HCO₃⁻ were then calculated from TCO₂, temperature and blood pH using the Henderson-Hasselbalch equation with values for solubility and pK¹_app based on Boutilier et al.^56,57. Haemoglobin content of the blood was also assessed by the cyanmethemoglobin method (using Drabkin’s reagent, Sigma). Half the remaining whole blood was then centrifuged at 10,000 rpm for 2 minutes at 4 °C. The resulting plasma was separated and snap frozen in liquid nitrogen and stored at −80 °C before later being used to measure plasma glucose and lactate using a YSI 2900D Biochemistry Analyzer (Xylem Analytics, UK). All measurements or storage of blood occurred within 10 minutes of blood sampling.

Measuring Hb-oxygen binding

We measured Hb-O₂ affinity using a Blood Oxygen Binding System (BOBS, Loligo systems), detailed in Oellermann et al.⁵⁸. A sample of the same whole blood used for blood chemistry measurements was diluted at a ratio of 1:4 in its own plasma. 1 μL of this blood was then used for measurements. The BOBS exposed this blood sample to gas mixes with a progressive increase in O₂ whilst measuring absorbance of light across a spectrum ranging from 200 to 800 nm. For each individual fish the gas mix that blood was exposed to matched the calculated pCO₂ of the blood sample. The change in absorption of light at a wavelength of 435 nm was used to assess changes in oxygenation of Hb, as previously used by Verhille & Farrell⁵⁹. Background changes in absorption of the blood sample were corrected using the isosbestic wavelength of 390 nm⁵⁹. Following measurements the BOBS calculated the oxygen equilibrium curve of the sample using Hill’s formula before estimating P₅₀⁵⁸.

Statistical analysis

We conducted all statistical analysis in R v3.5.3⁵⁴. There was no significant difference in mass of fish between the treatment groups (One-Way ANOVA, F_1,13 = 2.821, p = 0.117) or of estimated SMR of fish between treatment groups (Unpaired t-test, t = 0.8455, d.f. = 13, p = 0.413). There was, however, a significant correlation between estimated SMR and calculated O_2crit for all fish (Pearson’s correlation, df = 13, t = 3.32, R = 0.68, p = 0.0056). As such the effect of CO₂ treatment on O2_crit was assessed using an ANCOVA with SMR as a covariate. All other analyses were conducted using general linear modelling (GLM). All values in the text are reported as mean ± standard error (S.E.).