Hyperoxia disproportionally benefits the aerobic performance of large fish at elevated temperature Free

Aquatic heatwaves are increasing in prevalence and intensity as the global climate continues to change (IPCC, 2023). Although there is widespread recognition that different ectothermic species will have differential resilience to acute warming, there is also a growing acknowledgement that different sized individuals of the same species may be differently impacted. Indeed, observations for a range of fish species suggest that larger individuals have lower heat tolerance than smaller individuals (Clark et al., 2008; Di Santo and Lobel, 2017; Messmer et al., 2017; McPhee et al., 2023). In a review of empirical studies by McKenzie et al. (2021), the overarching tendency was for upper critical temperature limits to decrease with body size in post-metamorphic fishes.

A scaling exponent less than 1 indicates ‘hypoallometry’, and this tends to apply to both standard (baseline) metabolic rate (SMR) as well as maximum metabolic rate (MMR; ≈maximum oxygen uptake capacity), but not necessarily to the same degree for each of these metabolic extremes. Any difference between the scaling exponents of SMR and MMR will result in a progressive change in the aerobic scope (MMR minus SMR) of animals as they increase in size, and it is this relationship between aerobic scope and body mass that some argue to be the driver of size-dependent thermal tolerance (Pauly and Cheung, 2018).

Temperature is considered a ‘master factor’ in the biology of ectothermic animals, as environmental temperature sets body temperature and inextricably regulates biochemical reaction rates. Although the thermal dependence of aerobic scope has received substantial attention for many decades (Fry, 1947; Clark et al., 2013; Pörtner et al., 2017), and has been suggested to play some role in thermal performance (Pauly, 1981; Pörtner, 2010), comparatively little empirical attention has focused on the allometry of aerobic scope and its potential role in size-dependent thermal tolerance. Of the few empirical studies that exist, Atlantic cod (Gadus morhua) and European perch (Perca fluviatilis) held at warmer temperatures have been reported to suffer from a decline in aerobic scope as they increase in size, with increased mortality of the largest Atlantic cod individuals (Tirsgaard et al., 2015; Christensen et al., 2020). A study on coral grouper (Plectropomus leopardus) suggested that an observed decline in critical thermal maximum (CT_max) with body mass may have been linked to a concomitant decline in mass-specific oxygen uptake capacity at the highest test temperature (Messmer et al., 2017).

Environmental oxygen supersaturation (i.e. hyperoxia) is increasingly acknowledged to be ecologically relevant for several species of aquatic ectotherms, with levels of ≥150% air saturation frequently being observed in shallow, low-flow habitats with high photosynthetic output (Giomi et al., 2019; McArley et al., 2021). If it is indeed the case that an oxygen uptake limitation could help to explain size-dependent aerobic performance, we would expect that these patterns would be circumvented by providing supplemental environmental oxygen (Fig. 1). Specifically, we would anticipate that any temperature-induced contribution to metabolic hypoallometry under normoxia could be largely ‘corrected’ under hyperoxic conditions, given that hyperoxia has been demonstrated to enhance oxygen uptake capacity (Brijs et al., 2015; Skeeles et al., 2022).

To address some of the abovementioned knowledge gaps, we used 130 rainbow trout (Oncorhynchus mykiss) to investigate the allometry of SMR, MMR and aerobic scope in fish acclimated to 17°C and acutely warmed to 25°C in either normoxia (100% air saturation) or hyperoxia (150% air saturation). We hypothesised that: (1) the metabolic scaling exponent for MMR and aerobic scope would decrease at the higher temperature in normoxia, (2) hyperoxia at the higher temperature would increase the MMR and aerobic scope scaling exponents back to those observed in normoxia at 17°C, and (3) hyperoxia would have a clear benefit to the thermal resilience (survivability) of large fish at 25°C.

All experiments were conducted in accordance with the guidelines set by the Deakin University Animal Ethics Committee (#B31-2022), which complies with the Australian Code for the Care and Use of Animals for Scientific Purposes set out by the Australian Federal Government.

Rainbow trout, Oncorhynchus mykiss (Walbaum 1792), of both sexes ranging from 12 to 358 g, were obtained from the Snobs Creek hatchery in Victoria, Australia (N=180 fish obtained, n=130 fish used in the present study). Fish consisted of three age classes: fingerlings (∼20 g), 1-yr-olds (∼100 g) and 2-yr-olds (∼200–300 g). On 21 April 2023, the fish were shipped in an aerated 600 l tank (∼10°C) 250 km south to the Queenscliff Marine Research Station (Deakin University, Victoria, Australia).

Upon arrival, fish were split across four circular 500 l tanks in one of the facility's outside research aquarium systems (RAS4×500 l, Fresh by Design, Moss Vale, Australia). All n=100 fingerlings were netted into one tank, whereas n=50 1-yr-olds and n=30 2-yr-olds were mixed across the remaining three tanks. A digital sensor recorded the holding tank temperature every 10 min (Web ID temperature sensor, IOTMSS, Perth, WA, Australia). The temperature of the tanks was lowered to ∼12°C for fish introduction and then gradually raised over 18 h to ambient conditions of ∼17°C. Fish were acclimated to this regime for 20 days before experiments commenced, with temperature averaging 16.9±1.4°C (mean±s.d.) over the period. Post-transport mortality was low (n=2 fingerlings; 1.1%), and all other fish displayed normal behaviour (i.e. swimming, schooling, feeding) within 48 h and until experiments ended on 15 June 2023.

The holding system consisted of a 300 l sump which supplied the four vigorously aerated holding tanks with UV-sterilized, circulating freshwater. Water within the 500 l tanks was exchanged at a rate of approximately 50 l min⁻¹, particles were removed by a perforated central pipe, and water returned to the sump via an external overflow pipe. The inflow pipe was directed at a 45 deg angle into each tank to create a gentle current, enabling both swimming and resting behaviour of the rainbow trout. The direction of the inflow was alternated weekly to prevent one-sided muscular training. Fish behaviour was monitored every 2–3 days during the acclimation period via underwater cameras (GoPro11, GoPro, San Mateo, CA, USA). The sump received ∼1000 l day⁻¹ flow-through of clean, dechlorinated freshwater. Within the sump, water was treated with particle filters, a foam fractionator and a biofilter. Each tank was fed 2% of its total biomass every other day, using the same food used at Snobs Creek hatchery (2.5 mm semi-floating pellets for fingerlings, 4 mm semi-floating pellets for bigger fish) (Skretting Australia, Cambridge, Tasmania). Initially, the holding system water quality was tested daily for pH, ammonia, nitrite and nitrate (API freshwater test kits, Mars, Inc., Chalfont, PA, USA). After 2 weeks, the feeding regime was established, the biofilters had stabilized, and water chemistry testing was reduced to ≥2 times per week. The natural Victorian day:night cycle determined photoperiod in the outdoor holding system, maintaining identical conditions to the hatchery. This ranged from 12 to 10.5 h of daylight during the study.

Our custom-built respirometry system consisted of four opaque reservoir trays (dimensions ∼900×500×200 mm) housing submerged, transparent respirometry chambers with watertight clip-down lids. Fingerlings were measured in 1350 ml chambers (dimensions ∼160×105×70 mm), and all larger fish were measured in 8200 ml chambers (dimensions ∼320×230×100 mm). Each chamber had a recirculation loop, which ran at a rate of ∼3 l min⁻¹ (pump type 1005, Eheim, Deizisau, Germany), as well as an integrated fibre optic oxygen sensor that measured dissolved oxygen inside the chamber in mg l⁻¹ (Robust Oxygen Sensors, PyroScience, Aachen, Germany). The chambers were additionally connected to a flush pump (∼3 l min⁻¹ per chamber) on a digital, 10 min:10 min flush:seal timer (Smart_shifter, National Instruments, Austin, TX, USA), and each chamber had an overflow pipe that extended above the water surface. Before experiments started, we ensured that the chambers were leakless by injecting deoxygenated or oxygen-enriched water down the overflow pipe and sealing it from the flush system for >30 min. Water temperature in each reservoir tray and oxygen levels within each respirometer were recorded every 2 s to a laptop running Oxygen Logger software (PyroScience).

Each reservoir tray received continuous flow-through water at ∼8 l min⁻¹ (pump type 1100, Eheim) from a 125 l sump. The sump was instrumented with (1) a vigorously bubbling airstone to achieve normoxic conditions, (2) the sensor of an oxygen control system (‘Pacific’, OxyGuard, Farum, Denmark) regulating a pump that dispersed pure oxygen to achieve hyperoxic conditions, (3) a chiller (HC-1000A, Hailea, Guangdong, China) keeping the temperature at 17°C when testing fish at the acclimation temperature, and (4) a custom-made submersible heater with two 1 kW heating elements to raise the temperature for acute warming treatments.

The experiments were conducted with n=130 rainbow trout (12 to 358 g) between 11 May and 15 June 2023. Each fish was assigned to one of four main treatments in a 2×2 factorial design: normoxia (100% air saturation [hereafter ‘100% O₂’]), hyperoxia (150% air saturation [hereafter ‘150% O₂’]), acclimation temperature (17°C) and acute warming temperature (25°C). Sample sizes varied slightly across the treatments as maximization and homogeneity of the mass ranges were prioritized (Fig. 2, Table 1). The 25°C treatment was chosen to push the fish to the upper limits of their tolerable range (McCullough et al., 2001) and maximise the potential for oxygen to become limiting. To enable a higher resolution understanding of how temperature affected metabolism between the two main thermal treatments, we assigned a smaller subset of fish to a fifth treatment of 21°C normoxia (Table 1).

Each experiment commenced in the morning, when 5 to 15 rainbow trout of one size class, fasted for 24 to 48 h, were netted from their holding tanks into an aerated 75 l transport tank containing water from the holding system. Fish were moved from the outside research aquarium system to the temperature-controlled laboratory (∼3 min using trolley cart). Then, fish were individually placed into the size-appropriate respirometry chambers at 17°C and a preset oxygen treatment (100 or 150% O₂). Regardless of treatment, all fish were naïve to the respirometers and experienced the same amount of laboratory exposure. For the acute warming treatments, the chamber temperature was increased at ∼1°C h⁻¹ to achieve 25°C over ∼8 h (or 21°C over ∼4 h). Thus, all fish had at least one entire night in a respirometer at their treatment conditions, during which time oxygen depletion rates were recorded on a 10 min:10 min flush:seal cycle to capture resting levels of oxygen uptake rate (Ṁ_O2). The first 8 h of all trials, including those without a temperature ramp, were considered a habituation period and Ṁ_O2 data were not used in the analyses. The laboratory's light cycle was kept at 12 h:12 h light:dark with a gradual increase in light intensity for the first hour starting at 06:30 h (sunrise), and a gradual decrease during the last hour starting at 18:30 h (sunset).

The next morning, all fish were individually netted from the respirometers and placed into a circular exercise arena (diameter 400 mm, water depth 220 mm), filled with treatment-specific water from the sump, to undergo a two-minute exhaustive chase protocol (Norin et al., 2014). All fish across the size range appeared distinctly exhausted after the chase. The exhausted fish were moved back into their respective chamber by hand, and the chamber was sealed for 10 min to record maximum Ṁ_O2. This was always conducted in less than 30 s. Finally, fish were removed from the respirometers and the chambers were closed for ≥20 min to quantify background oxygen fluxes (i.e. microbial respiration and the effects of minor temperature changes). Fish were euthanised with a sharp blow to the back of the head before being weighed and dissected to check for sex. To control for potentially confounding effects between body size and life stage, fish used in the present study were post-metamorphic, and sexual maturity was assessed for all individuals. Signs of sexual maturity were seen in n=6 fish, where maturity was taken as gonads greater than stage 3 (Gomez et al., 1999), so these have been indicated in figures for transparency.

For calculations of Ṁ_O2 (mg O₂ min⁻¹), slopes of O₂ over time (5 min sections in the middle of each 10 min sealed cycle) were first extracted by loading the raw respirometry text files into the software LabChart 7 (ADInstruments Pty Ltd, Bella Vista, NSW, Australia) and accounting for respirometer volume. Further data analysis was conducted using R (version 4.1.3, https://www.r-project.org/). After removal of the habituation period (∼8 h, including any temperature ramp), this resulted in an average of ∼43 Ṁ_O2 data points for each fish. The standard metabolic rate (SMR) for each individual was calculated as the lowest 10% of Ṁ_O2 data points after outlier removal (>2 s.d. from the mean of the lowest 10% of Ṁ_O2 values). We decided on this arithmetical approach after a visual confirmation of each entire respirometry trial, which we encourage as a general practice (see Fig. S1). Maximum metabolic rate (MMR) was taken as the highest Ṁ_O2 value overall. Post-chase Ṁ_O2 was extracted across three consecutive 3-min intervals starting from the moment when chambers were sealed. For 126 of 130 fish (97%), MMR corresponded to the first 3-min period immediately after the chase. Microbial respiration and minor temperature changes within the system were accounted for by matching time-specific, linearly regressed values of pre- and post-experiment background oxygen fluxes (see Fig. S1). Absolute aerobic scope (AAS) was calculated as the difference between maximum metabolic capacity and baseline metabolic demands, i.e. MMR−SMR.

For the statistical analysis, Ṁ_O2 response variables (i.e. SMR, MMR and AAS) and body mass were log₁₀-transformed. A linear model that allowed intercepts and slopes to vary by treatment (i.e. defining all interactions of body mass, temperature and oxygen treatment) was formulated for each Ṁ_O2 response variable. Linearity, normality of residuals, homoscedasticity and independence of residual error terms were assessed visually. The R package ‘lsmeans’ (Lenth, 2016) was used to evaluate the estimated marginal means (least-squares means) of the three linear models. This allowed for comparison of the 95% confidence intervals (CI) across all treatment slopes (scaling exponents) and mass-standardised intercepts. Tukey's honestly significant difference (HSD) tests were used to assess the difference of slopes and the difference of intercepts at zero log-transformed body mass (corresponding to an antilog body mass of 1 g for the determination of scaling factors) and at the grand mean body mass of 88.5 g.

To allow for an easy observation of metabolic thermal trajectories excluding mass effects, Ṁ_O2 response variables were mass-standardised to the grand mean body mass of 88.5 g. To achieve this, power functions of the response variables as a function of body mass were formulated for each treatment, and individual residuals were calculated and added to the regression-predicted value for an 88.5 g fish.

At the cooler temperature of 17°C, hyperoxia did not change the scaling exponent (slope, b) of the relationship between body mass and any of the three metabolic parameters (Fig. 2, Table 2). The scaling exponents for SMR averaged 0.77 and were higher for MMR (0.86), resulting in scaling exponents for aerobic scope of 0.87.

Although hyperoxia also did not change the elevation (a) of the relationship between body mass and SMR compared with normoxia at 17°C (a=0.0043 versus 0.0047, respectively; Fig. 3), it did modestly elevate the curve for MMR and aerobic scope (Fig. 3, Table 1). At the grand mean body mass of 88.5 g, hyperoxia at 17°C caused a 14.6% elevation in MMR and a 17.8% elevation in aerobic scope.

Increasing the temperature to 21°C and 25°C caused an overall increase in SMR regardless of oxygen treatment (Fig. 3, Table 1). The scaling exponent for SMR at 25°C in normoxia (0.90) was significantly higher than those at 17°C in each of the oxygen treatments (mean 0.77), but hyperoxia at 25°C reestablished a lower scaling exponent (0.78; Table 1).

MMR and aerobic scope in normoxia increased with temperature between 17°C and 21°C (by 21.9% and 14.0%, respectively, at the grand mean mass) and decreased again as temperature continued up to 25°C (no difference between 17°C and 25°C at the grand mean mass) (Fig. 3, Table 1). The scaling exponent for aerobic scope at 25°C normoxia (0.74) was significantly lower than at 17°C (0.87), driven by a higher scaling exponent for SMR (noted above) and a numerically lower (yet not significantly different) scaling exponent for MMR (Table 2). Similar to the allometry of SMR described above, hyperoxia at 25°C reestablished the scaling exponents for MMR and aerobic scope (0.84 and 0.86, respectively) that were similar to those at 17°C. Although hyperoxia at 25°C had little effect on SMR, it significantly increased MMR and aerobic scope across the mass range (32.8% and 48.2% increase, respectively, at the grand mean mass), elevating both parameters above those measured at 17°C.

Four fish died in each of the 25°C normoxia and 25°C hyperoxia treatments prior to the exercise protocol (Table 1), yet there was no indication that the mortalities were dependent on body mass (mean body mass of mortalities 84.47 g; Fig. S2). Mortalities appeared across n=6 separate experimental trials (n=3 separate trials for each oxygen treatment) throughout the duration of the experiment. Firstly, this indicates that 25°C was a suitably challenging temperature to address the objectives of the study, and secondly, this suggests that hyperoxia did not provide any clear benefit to the survivability of fish undergoing a thermal challenge.

The aerobic scope of rainbow trout in the present study increased as temperature rose from 17°C to 21°C before dropping to its lowest level at 25°C. In general, the values obtained here were consistent with those reported previously for rainbow trout at similar temperatures (Chen et al., 2015; Verhille et al., 2016). Although we did not aim to provide a high-resolution estimate of the optimal temperature for aerobic scope (T_opt,AS), it is clear that T_opt,AS was closest to 21°C out of the three temperatures tested. We take two main points from this observation. First, given that rainbow trout are known to perform best at temperatures around 14–18°C (Grabowski, 1973; Myrick and Cech, 2000; McCullough et al., 2001), our results add to a growing database showing that T_opt,AS does not closely approximate the true optimal temperature for the species (Clark et al., 2013; Jutfelt et al., 2018). Second, it is apparent from both the decline in aerobic scope and the appearance of mortalities at 25°C that the fish were being pushed towards their thermal limits at 25°C.

The deterioration of aerobic performance at the highest temperature was exacerbated as body mass increased from ∼10 to 360 g, as the scaling exponent for aerobic scope was significantly lower at 25°C (0.74) than at 17°C (0.87). This pattern in aerobic scope was a result of steeper SMR scaling at 25°C compared with 17°C, not matched by an increase in MMR scaling, perhaps resulting from oxygen uptake limitation in larger fish at the highest test temperature (Pauly and Cheung, 2018). We explicitly addressed these ideas through oxygen manipulation in the present study and we evaluate the evidence below.

Hyperoxia has been shown to increase arterial and, to a lesser extent, venous blood oxygen partial pressure in a range of fish species owing to an increased diffusion gradient between the water flowing across the gills and the blood perfusing the gills (Takeda, 1990; Soncini and Glass, 1997; McArley et al., 2021). In the present study, hyperoxia elevated MMR and aerobic scope at both temperatures relative to normoxia, while SMR stayed constant. This aligns with previous literature showing that oxygen supplementation causes an increase in MMR with little or no increase in SMR (Kaufmann and Wieser, 1992; Brijs et al., 2015; McArley et al., 2018; Skeeles et al., 2022). We also observed that the increase in MMR with hyperoxia was temperature dependent, with the increase being less at 17°C (14.6% at the grand mean body mass) than at 25°C (32.8%). This temperature dependence may explain why a previous study on rainbow trout at 15°C failed to show an increase in MMR under hyperoxia (190% air saturation; Duthie and Hughes, 1987). Indeed, a comparison of the available literature investigating MMR in normoxia and hyperoxia shows that across a range of species with similar maximum temperatures encountered in their natural environment (Table S1; Froese and Pauly, 2010), the choice of experimental temperature might help to explain the amplitude of the observed effects of hyperoxia on MMR (Fig. 4). Studies of European perch (Brijs et al., 2015) and two intertidal triplefin fishes (Bellapiscis medius and Forsterygion lapillum) (McArley et al., 2018) investigated aerobic performance at multiple temperatures, including beyond the species' natural temperature ranges, and showed greater hyperoxia-induced increases in MMR at higher temperatures. This difference in utilisation of the supplemented oxygen with increasing temperature could result from the increasing difference between oxygen supply capacity and oxygen demand (Clark et al., 2008), changes in the potential of hyperoxia to increase oxygen supply owing to temperature-dependent changes in the blood oxygen-carrying capacity (Weber and Campbell, 2011), or other factors.

In addition to elevating the allometric curves for MMR and aerobic scope, hyperoxia ‘reestablished’ an aerobic scope scaling exponent at 25°C (0.86) that had decreased in 25°C normoxia (0.74) relative to 17°C normoxia (Fig. 2, Table 2). This provides support for our hypothesis that a breakdown of ‘normal’ metabolic hypoallometry at high temperature in normoxia could be corrected under hyperoxic conditions. That is, the aerobic scope of larger fish disproportionately benefited from hyperoxia. However, our expectation was that the benefit of hyperoxia on aerobic scope at high temperature would be driven exclusively by an increase in MMR scaling (Fig. 1). Instead, we found that a decrease in SMR scaling between normoxia and hyperoxia played a significant role. The contribution of MMR to the hyperoxia-induced improvement of aerobic scope at 25°C at the largest body size was ∼85%, while SMR contributed the other ∼15%. We discuss the implications of these directional shifts in MMR and SMR below.

The decrease in the SMR scaling exponent from normoxia to hyperoxia at 25°C suggests a disproportionate reduction of baseline metabolic demands in bigger fish. This may be due to the hyperoxia-induced increase in blood oxygenation causing a size-dependent decrease in ventilatory and circulatory effort (Wood and Jackson, 1980; Takeda, 1990; Soncini and Glass, 1997; Mark et al., 2002; McArley et al., 2021). Although we are not aware of any studies that have looked into such size dependence, the costs of ventilation in normoxia have been reported to constitute ∼10% of total energetic costs both in ∼250 g rainbow trout resting at 4–12°C (Jones and Schwarzfeld, 1974), and in ∼750 g fish at 14–16°C shifting from buccal pumping to ram ventilation at a constant, high swimming speed (Steffensen, 1985).

Although the numerical increase in the scaling exponent of MMR between normoxia and hyperoxia at 25°C did not reach statistical significance (Table 2), it was responsible for ∼85% of the significant improvement in aerobic scope of large fish at 25°C in hyperoxia compared with normoxia, and therefore warrants some discussion. As the gill surface area of fish typically scales hypoallometrically within species (median b=0.85 in Scheuffele et al., 2021) and metabolic rates increase with temperature, it is possible that fish face a progressive mismatch between oxygen supply and demand as they increase in body mass at warmer temperatures (Clark et al., 2008). Although there is increasing evidence that gill surface area is plastic and acclimates to meet metabolic demands (Scheuffele et al., 2021; Skeeles and Clark, 2024), diminishing arterial oxygen partial pressure and haemoglobin–oxygen saturation have been shown in rainbow trout (including sea-run ‘steelhead’) following acute challenges such as exhaustive exercise and a temperature ramp (Primmett et al., 1986; Milligan and Wood, 1987; Keen and Gamperl, 2012; McArley et al., 2022). Indeed, McArley et al. (2022) showed that hyperoxia ‘counteracted’ a decline in arterial oxygenation in rainbow trout immediately following exercise, and MMR increased through an increase in the (Fick-estimated) arteriovenous oxygen content difference and an increase in cardiac output. It could also be that a hyperoxia-induced increase in cardiac output, perhaps owing to improved cardiac contractility (Ekström et al., 2016), disproportionally benefits bigger fish via size-dependent improvements in venous oxygen partial pressure (Clark et al., 2008). Much remains to be discovered in this field of research and we encourage the inclusion of large body mass ranges in future studies, as well as attempts to control for the confounding effects of size versus age.

Our findings of size-dependent effects of hyperoxia add to a growing discussion surrounding the potential for hyperoxia to provide a survival advantage for aquatic ectotherms facing thermal challenges (Giomi et al., 2019; McArley et al., 2021). Given that periods of high photosynthetic output typically coincide with high solar activity (and elevated temperature), our results suggest that hyperoxia may provide a ‘metabolic refuge’ to fish experiencing thermal stress. However, although our results suggest that hyperoxia could enable larger fish to better maintain aerobic processes such as digestion, protein synthesis and predator evasion, the linkage between aerobic scope and thermal tolerance remains unclear and debated (Pörtner et al., 2017; Jutfelt et al., 2018).

Although we did find support for our predictions that the aerobic scope scaling exponent would decline at the high temperature and would be ‘corrected’ by hyperoxia, we did not find clear evidence for our hypothesis that survivability would benefit from oxygen supplementation. Specifically, we did not see any evidence of large fish succumbing to 25°C in normoxia but not in hyperoxia. Mortality was low in general, preventing us from any definitive conclusions, but it is notable that (1) mortality rates were equivalent in normoxia (12.1%) and hyperoxia (13.8%), (2) there was no clear size dependence to the mortalities, and (3) the hyperoxia mortalities appeared to have enhanced aerobic capacities relative to normoxia mortalities immediately prior to thermal death (see Fig. S2).

There are surprisingly few empirical studies that have attempted to manipulate aerobic scope to assess its effects on thermal tolerance. Studies that experimentally induced anaemia showed that reduced blood–oxygen carrying capacity did not translate to reduced acute thermal tolerance in European perch and European sea bass (Dicentrarchus labrax) (Wang et al., 2014; Brijs et al., 2015). In the limited studies investigating the effect of oxygen supplementation on the acute thermal tolerance of fish, many reveal no change relative to normoxia (Rutledge and Beitinger, 1989; Healy and Schulte, 2012; Brijs et al., 2015; Devor et al., 2016). However, there has, been a minority of studies reporting that hyperoxia can benefit acute thermal tolerance (Ekström et al., 2016; McArley et al., 2018; Giomi et al., 2019), and Messmer et al. (2017) reported a correlation between mass-specific MMR (at the highest test temperature of 33°C) and size-dependent acute thermal tolerance in coral grouper under normoxia. Overall, the landscape of contradictory results within the literature calls for more empirical research combining thermal tolerance tests with aerobic scope manipulation.

This study provides evidence that the aerobic scope of rainbow trout is progressively compromised in larger fish at high temperature relative to more benign temperatures. Although this suggests that larger fish will have a compromised capacity to supply simultaneous oxygen-demanding processes at high temperatures, our successful efforts to reestablish ‘normal’ scaling of aerobic scope by supplying hyperoxia did not appear to change thermal resilience as gauged by mortalities.

A mismatch between oxygen supply across the gills and demand at the tissues has been implicated in driving declines in fish size with climate warming (Pauly, 1981; Pauly and Cheung, 2018). We propose that a path forward in this field must involve: (1) identification of common characteristics across model species that show warming-induced disruption to metabolic allometry; (2) empirical measurements of aerobic scope at multiple temperatures covering the natural and forecasted ranges, across the full larval/post-larval mass range of the species and disentangling size versus age effects; (3) experimental manipulations of oxygen availability to quantify impacts on metabolic allometry across temperature; and (4) acute and chronic thermal tolerance tests under all treatment conditions to quantify relationships between growth rate, body mass, aerobic scope and thermal tolerance (e.g. loss of equilibrium or survival). Further studies delving into finer-scale mechanisms would also be valuable, such as allometric investigations of limiting steps in the oxygen transport system (e.g. gill diffusion and arterial saturation, blood–oxygen carrying capacity dynamics, cardiac function, ATP production efficiency) across temperature and oxygen levels. This field of study is ‘theory rich’ but ‘data poor’; we hope that our findings and suggestions stimulate more robust empirical studies to address some of the wide but exciting knowledge gaps outlined above.

We acknowledge the Wadawurrung peoples, the traditional custodians of the area where this research was conducted. We thank Rhiannon Atkinson, Neil Hyatt, Brett Ingram, Christopher Mullins and Dylan White from the Victorian Fisheries Authority, Snobs Creek hatchery, for rainbow trout supply and advice. Further, we thank Peter Biro and Michael Skeeles for their statistical guidance, as well as Lisa Grubb, Tess Hoinville and Sam Wines from Deakin University for their technical support.