Hyperoxia increases maximum oxygen consumption and aerobic scope of intertidal fish facing acutely high temperatures

Temperature has pervasive impacts on biological processes and plays an important role in shaping the distribution and abundance of species (Schulte, 2015). All fish face at least some natural fluctuation in environmental temperature, but species inhabiting intertidal rock pools can be exposed to large acute (hours) increases in temperature when daytime low tide coincides with hot terrestrial conditions (Fig. 1A). Low tide thermal ramping events in rock pools may ultimately threaten the upper thermal tolerance limits of intertidal fish, but are also associated with sub-lethal impacts including increased energetic demands. These higher energetic demands are reflected in the profound influence of temperature on aerobic metabolism where rates of oxygen consumption typically increase exponentially under acute thermal ramping (e.g. Healy and Schulte, 2012b; McArley et al., 2017). The aerobic metabolic response to thermal ramping has been well characterised in fish but there is a lack of studies investigating the role of aerobic metabolism in thermal tolerance in intertidal fish species, which are routinely exposed to acute high temperatures.

Thermal tolerance has been proposed to be linked to aerobic metabolism in aquatic ectotherms through the influence of temperature on aerobic metabolic scope (MS) (Portner and Farrell, 2008; Portner et al., 2017; Schulte, 2015; Sokolova et al., 2012). MS is the difference between basal (termed the standard metabolic rate, SMR) and maximum (Ṁ_O2,max) rates of oxygen consumption and theoretically represents the portion of metabolic capacity above and beyond basal requirements (i.e. SMR) that can be directed to aerobic-dependent activities such as growth, feeding, digestion, locomotion and reproduction (Claireaux and Lefrancois, 2007; Clark et al., 2013). As thermal ramping proceeds in rock pools, the MS of intertidal fish is expected to fall because higher temperatures increase SMR (Schulte, 2015), and the capacity to support aerobically driven activities may then decline at elevated temperatures. However, for MS to decline under thermal ramping, Ṁ_O2,max has to increase less than SMR across an equivalent temperature increment. In some fish species, Ṁ_O2,max may increase more than or in parallel with SMR at higher temperatures, and therefore MS fails to decline because Ṁ_O2,max does not plateau (Healy and Schulte, 2012b; Lefevre, 2016; Norin et al., 2014). In the common triplefin (Forsterygion lapillum Hardy 1989), a species known to inhabit rock pools, Ṁ_O2,max did not increase at higher temperatures following acclimation to 15, 18, 21 and 24°C (McArley et al., 2017). Furthermore, in the same study, routine Ṁ_O2 under acute thermal ramping to critical thermal limits (>29°C) only marginally exceeded Ṁ_O2,max measured at the lower acclimation temperatures. Therefore, at least for this triplefin species, there may be a limited capacity to increase Ṁ_O2,max at higher temperatures, and this would result in vulnerability to a collapse of MS when exposed to thermal ramping in rock pools.

Although the MS of at least one triplefin species appears constrained at higher temperatures (McArley et al., 2017), these findings were determined under normoxia (100% of air saturation) and this is not always the setting within rock pools. Thermal ramping in rock pools generally occurs on hot sunny days when rates of algal photosynthesis are high and rock pool water can become hyperoxic at this time (see Fig. 1B for example and Berschick et al., 1987; Bridges, 1988; Richards, 2011; Truchot and Duhamel-Jouve, 1980). Thus, in a thermal ramping event, intertidal fish reside within normoxic water at ambient temperatures as the tide first recedes; then, rock pool seawater warms and becomes increasingly hyperoxic until the incoming tide flushes the rock pool. Ṁ_O2,max and MS were approximately doubled under hyperoxia compared with values under normoxia in European perch (Perca fluviatilis) when measured at an acclimation temperature of 23°C (Brijs et al., 2015). Furthermore, in the same study, fish exposed to hyperoxia had a greater capacity to increase routine Ṁ_O2 at temperatures approaching upper tolerance limits (>30°C). It is not known whether the effects of hyperoxia seen in perch are widespread in fish but, if present in triplefins, an expected loss of MS under thermal ramping due to constrained Ṁ_O2,max may be compensated for through the permissive effects of hyperoxia. MS could also be protected at high temperature under hyperoxia because the energetic costs of meeting oxygen demands (e.g. ventilatory costs) are diminished by excess oxygen availability (Mark et al., 2002; Portner et al., 2017). While environmental hyperoxia in rock pools may therefore offer a metabolic refuge for intertidal fish facing acute thermal stress, this possibility has not been addressed.

Intertidal organisms are assumed to live very close to their thermal tolerance limits (Helmuth et al., 2002). In rock pools inhabited by triplefin fish, we have measured temperatures exceeding 29°C in summer (Fig. 1A). These temperatures are close to the upper thermal tolerance limit (U_TL, 31.4°C) of F. lapillum (McArley et al., 2017), suggesting that, in at least some rock pools, this species would face temperatures close to its tolerable limit. While the exclusively intertidal triplefin [Bellapiscis medius (Günther 1861)] also appears to live close to its thermal limits (Hilton et al., 2008), the upper thermal limit of this species is unknown. An inability to match oxygen supply and demand has been proposed as a key determinant of thermal tolerance in aquatic ectotherms (Portner and Farrell, 2008; Portner, 2010; Portner et al., 2017) and experimental manipulations of tissue oxygen supply using hyperoxia have addressed the upper thermal limits of fish in this context (Brijs et al., 2015; Devor et al., 2016; Ekström et al., 2016; Healy and Schulte, 2012a; Mark et al., 2002). While these studies indicate that factors other than tissue oxygen supply may also set absolute upper thermal limits, hyperoxia did slightly increase the critical thermal maximum (+0.9°C) of European perch (Ekström et al., 2016). Rock pools are expected to become hyperoxic during thermal ramping events that should approach upper thermal limits, and for this reason, the possibility of increased maximal thermal tolerance under hyperoxia in intertidal fish requires consideration. Given predicted increases in the severity of heat wave events due to climate change (Perkins et al., 2012), and the possibility that intertidal fish already live close to their upper thermal limits, resolving thermal tolerance limits of intertidal fish under environmentally relevant conditions is pertinent.

The aims of this study were to determine: (1) whether hyperoxia increases the Ṁ_O2,max and MS of intertidal fish during thermal ramping, (2) whether hyperoxia decreases routine Ṁ_O2 (an estimate of basal metabolic costs during rapidly changing temperature) in intertidal fish under thermal ramping, and (3) whether hyperoxia increases the upper thermal limits of intertidal fish. To address these aims, the Ṁ_O2 (mass-specific oxygen consumption) and upper thermal tolerance capacities were determined for two New Zealand triplefin fishes under hyperoxia (∼200% air saturation) and normoxia (∼100% air saturation). This included an exclusively intertidal species (B. medius) and a second species (F. lapillum), which also occupies intertidal rock pools but is more commonly found in shallow subtidal habitats. To address the first and second aims, MS was measured at 21°C, routine Ṁ_O2 was measured at 21, 24, 26, 28 and 29°C under thermal ramping and MS above routine metabolism was measured at 29°C (see Fig. 2). It was predicted that Ṁ_O2,max, and therefore MS, would be higher under hyperoxia at 21 and 29°C because of a limitation of Ṁ_O2,max under normoxia (Brijs et al., 2015). Additionally, routine Ṁ_O2 under thermal ramping was predicted to increase less under hyperoxia than under normoxia because of decreased cardiorespiratory costs required to meet oxygen demands. As such, the rate and amplitude of gill ventilation were also measured under thermal ramping to determine whether the lower metabolic rates expected in hyperoxia could indeed be attributed to decreased ventilatory costs. To address the third aim (whether hyperoxia increases absolute thermal limits), the upper thermal tolerance limit (U_TL) of each species was assessed under hyperoxia and normoxia. U_TL was defined as the temperature of equilibrium loss during a standardised (2°C h⁻¹) thermal ramping exposure.

Fish used in this study were netted from intertidal rock pools (B. medius) or shallow subtidal (<5 m) habitats by divers (F. lapillum) in the vicinity of Leigh, New Zealand. Animals were housed in 30 l flow-through seawater tanks (air saturated, 200 μm filtered, 35 ppt salinity) at the Leigh Marine Laboratory (Leigh, New Zealand) and were acclimated to a 12 h photoperiod and a temperature of 21±0.5°C for at least 4 weeks prior to experimentation. During the acclimation period, fish were fed daily on a mix of crushed aquaculture feed (Skretting, Cambridge, TAS, Australia) and pilchard. Food was withheld for a period of at least 48 h prior to the start of experimental protocols. Specific dynamic action is completed in 48 h in triplefin fish (Stobie, 2017), so a stable non-digesting state should have been reached by the time experimental protocols began. All experimental techniques were performed under approval of the University of Auckland Animal Ethics Committee (approval: 001801).

Mass-specific oxygen consumption (Ṁ_O2) was assessed in B. medius (body mass 1.92±0.13 g, length 57±1.6 mm) and F. lapillum (body mass 2.13±0.11 g, length 56.4±2.7 mm) under conditions of normoxia (∼21 kPa, 100% air saturation) or hyperoxia (∼42 kPa, 200% air saturation) (N=10 for all treatments) and there were no differences in body mass between treatments groups (ANOVA, d.f.=3, F=1.21, P=0.32). Hyperoxia was achieved by bubbling oxygen into the seawater reservoir supplying the respirometers and was maintained at the required level by an oxy-guard control unit (Loligo Systems, Tjele, Denmark). The sequence of Ṁ_O2 measurements that were made in each treatment group is displayed in Fig. 2. Assessment of Ṁ_O2 started by transferring fish to a circular 30 l tank filled with normoxic or hyperoxic seawater at 21°C in which they were exhaustively exercised by chasing for a period of 5 min. All fish appeared to be fully exhausted after 5 min (i.e. no longer able to perform burst swimming in response to touch). At the conclusion of chasing, the fish were transferred to individual respirometry chambers (see ‘Respirometry methods’, below) supplied with either normoxic or hyperoxic (21°C) seawater and repeating 4 min Ṁ_O2 measurement cycles were initiated. Ṁ_O2,max (21°C) was taken as the highest Ṁ_O2 value recorded in any measurement cycle, which was in all cases the first cycle following exhaustive exercise. A 5 min exhaustive exercise protocol has previously been used to elicit Ṁ_O2,max values in triplefins (Khan et al., 2014a; McArley et al., 2017) and is suitable for assessment of Ṁ_O2,max in benthic species which will not continuously swim in a flume (Clark et al., 2013; Norin and Clark, 2016). The fish were then left undisturbed in respirometers for a period of ∼16 h under conditions of constant temperature (∼21°C) and water oxygen content (normoxia or hyperoxia), and Ṁ_O2 was measured in repeating 10 min cycles (∼100 individual Ṁ_O2 measurements). SMR at 21°C was defined as the mean of the lowest 10% of Ṁ_O2 measurements recorded in the overnight recovery period (Khan et al., 2014a; Norin et al., 2014) (Fig. 2). At ∼08:00 h, the overnight SMR measurement ceased and a thermal ramping protocol was started where temperature was raised from 21°C to 29°C at a rate of ∼2°C h⁻¹. Metabolic rate was measured over two measurement cycles (∼15 min) at 21, 24, 26, 28 and 29°C during thermal ramping and the mean of the two measurements was defined as the temperature-specific routine Ṁ_O2 (Fig. 2). After the final routine Ṁ_O2 measurement at 29°C, the fish was removed from the respirometer and transferred to a 30 l tank filled with normoxic or hyperoxic seawater preheated to 29°C. The fish was exhaustively exercised by chasing for 5 min, then returned to the respirometer for assessment of Ṁ_O2,max over four 3–4 min measurement cycles. The Ṁ_O2,max at 29°C was defined as the highest value of Ṁ_O2 recorded over these cycles. Finally, the fish was removed from the respirometer and allowed to recover prior to recording body mass and length. The Ṁ_O2 measurements described above were used to determine true MS (the difference between Ṁ_O2,max and SMR) under normoxia and hyperoxia at 21°C (Fig. 2). True MS could not be determined at 29°C because fish were not held at this temperature for the prolonged periods required to reliably estimate SMR. The difference between the lowest value of routine Ṁ_O2 at 29°C and Ṁ_O2,max at 29°C was used as an approximation of MS following thermal ramping (see Fig. 2). In four individuals, routine Ṁ_O2 at 29°C was marginally higher than Ṁ_O2,max at 29°C and these individuals were treated as having zero MS above routine Ṁ_O2. Because MS was defined differently at 21 and 29°C, no statistical comparisons were made between temperatures.

U_TL was measured under hyperoxia (200% air saturation) and normoxia (100% air saturation) in both species (N=9–10, B. medius: body mass 2.03±0.24 g, length 56.6±2.1 mm, F. lapillum: body mass 1.52±0.12 g, length 55±1.2 mm). Assessment of U_TL was carried out in a 60 l reservoir filled with either normoxic or hyperoxic seawater pre-heated to 21°C. The reservoir accommodated four fish per run (∼20 h) and two fish of each species were included in each run. The fish were introduced into the reservoir and allowed to acclimate overnight (∼16 h) under conditions of normoxia or hyperoxia and stable temperature (∼21°C). Following overnight acclimation, the temperature of the reservoir was raised at a rate of ∼2°C h⁻¹ until the fish lost equilibrium for a period of at least 10 s. The loss of equilibrium temperature was taken as the U_TL (Beitinger et al., 2000; Brijs et al., 2015; Mora and Maya, 2006; Sandblom et al., 2016). Temperature and water oxygen content were measured throughout the U_TL assessment using a FireSting O₂ meter (PyroScience).

Ventilation frequency and amplitude were measured by video analysis in both species under normoxia or hyperoxia and thermal ramping (21, 24, 26, 28 and 29°C) (N=8, B. medius: body mass 2.29±0.35 g, length 50±1.9 mm, F. lapillum: body mass 1.8±0.15 g, length 49.4±1 mm). There were no differences in body mass between treatment groups (ANOVA, d.f.=3, F=0.73, P=0.54). The measurements were carried out in an insulated 60×45×30 cm tank which was filled to a depth of 7 cm and continuously supplied with normoxic or hyperoxic seawater from a temperature-controlled 60 l reservoir. The depth of the tank allowed video cameras to be mounted directly above pens made of cylindrical acrylic tubing, which maintained the position of the fish throughout the measurement period. The experimental setup accommodated four fish and two of each species were used in each run. The fish were introduced into the pens and then allowed to recover overnight (∼16 h) under constant temperature (∼21°C) and oxygen conditions (normoxia or hyperoxia). Following overnight recovery, a thermal ramping protocol was initiated where temperature was increased from 21 to 29°C at a rate of ∼2°C h⁻¹ and videos were recorded at 21, 24, 26, 28 and 29°C. A single video recording (∼2 min) was taken at each temperature set point and thermal ramping continued immediately after the video had been taken.

Ventilation frequency and amplitude were quantified from videos using Image J imaging software (US National Institutes of Health, Bethesda, MD, USA). Ventilation frequency was counted over 60 s in each video by using the Plot Z-axis function to measure the change in pixel intensity of an area adjacent to the operculum as it opened and closed. A 60 s time period has been used to estimate ventilation frequency in other fish species (Frisk et al., 2012). Ventilation amplitude was assessed by comparing the maximum width (mm) of the operculum when open to the width of the operculum when closed. The opening operculum was observed frame by frame during an opercula beat and width was measured in the frame where the maximum separation between the right and left operculum was observed and where the operculum was closed. Three randomly selected individual opercula beats were measured in each video and were converted to a percentage increase in operculum width. The mean of these three measurements was taken as the ventilation amplitude at each temperature. A plastic mesh grid of known dimensions attached to the base of the fish pens was used to calibrate the distance measurements in each video.

All statistical analysis was carried out using the SPSS Statistics 24 or Sigma Plot 13.0 software packages. Two-way analysis of variance (ANOVA) with Holm–Sidak post hoc comparisons was used to determine the effect of species and water oxygen content on SMR at 21°C, MS at 21°C, MS above routine Ṁ_O2 at 29°C and U_TL. For the analysis of MS above routine Ṁ_O2 at 29°C, the data were log transformed. The impact of species, water oxygen content (hyperoxia versus normoxia) and thermal ramping temperature on routine Ṁ_O2 was assessed using a three-way mixed ANOVA. For this analysis, species and water oxygen content were set as between-subjects factors and thermal ramping temperature as a within-subjects factor. For the tests of within-subjects effects, a violation of sphericity was assumed and epsilon (Huynh–Feldt)-adjusted F-tests were run. For all post hoc tests, Holm–Sidak-adjusted significance levels were used. Mixed three-way ANOVA was also used to assess the impact of species, water oxygen content and temperature on Ṁ_O2,max. No three-way interactions were significant (P>0.05) and only the two-way interactions and main effects are reported in the Results. Ventilation frequency, amplitude and estimated ventilation volume were compared within each species using repeated-measures ANOVA with temperature set as a within-subjects factor and water oxygen content set as a between-subjects factor. For F. lapillum, the natural logarithm of ventilation frequency was used in the analysis.

SMR was significantly higher in B. medius than in F. lapillum (ANOVA, d.f.=1, F=44.09, P<0.001) and in both species it was significantly higher under hyperoxia than under normoxia (ANOVA, d.f.=1, F=9.03, P=0.005; Fig. 4). The influence of hyperoxia on Ṁ_O2 during thermal ramping was explained by an interaction between thermal ramping temperature and water oxygen content (ANOVA, d.f.=3.65, F=3.78, P=0.008). In both species, Ṁ_O2 was significantly higher under hyperoxia than under normoxia but only at temperatures of 26, 28 and 29°C (Fig. 4). There was also an interaction between thermal ramping temperature and species (ANOVA, d.f.=3.65, F=12.65, P<0.001). While Ṁ_O2 did not differ between the species at 21°C, at temperatures of 24, 26, 28 and 29°C, B. medius had a significantly higher Ṁ_O2 than F. lapillum. Moreover, Ṁ_O2 increased significantly from 21 to 24°C and from 28 to 29°C in B. medius (Fig. 4A), but there was no significant difference in Ṁ_O2 over these temperature changes for F. lapillum (Fig. 4B).

The impact of species, temperature and water oxygen content on Ṁ_O2,max was explained by interactions between temperature and species (ANOVA, d.f.=1, F=4.15, P=0.049), and temperature and water oxygen content (ANOVA, d.f.=1, F=22.4, P<0.001). Under all conditions, Ṁ_O2,max was significantly higher in B. medius than in F. lapillum but the difference between the species was greatest at 29°C (Fig. 4). Ṁ_O2,max was significantly higher under hyperoxia than under normoxia in all treatments; however, the difference in Ṁ_O2,max between hyperoxia and normoxia was larger at 29°C than at 21°C (Fig. 4). Interestingly, in both species, Ṁ_O2,max increased significantly between 21 and 29°C under hyperoxia whereas under normoxia there was no change in Ṁ_O2,max between these temperatures (Fig. 4).

At 21°C, MS was significantly greater under hyperoxia in both species (ANOVA, d.f.=1, F=13.12, P<0.001) and higher in B. medius than in F. lapillum (ANOVA, d.f.=1, F=27.71, P<0.001) (Fig. 4). Following thermal ramping to 29°C, MS above routine Ṁ_O2 was significantly higher under hyperoxia than under normoxia in both species (Fig. 4). Bellapiscis medius had significantly greater MS above routine Ṁ_O2 than F. lapillum at 29°C (ANOVA, d.f.=1, F=5.46, P=0.025).

There was a significant interaction between the effect of species and water oxygen content on U_TL (ANOVA, d.f.=1, F=4.19, P=0.048). Although the differences were small (<0.6°C), the intertidal specialist B. medius had a significantly higher U_TL than F. lapillum under both normoxia and hyperoxia (Fig. 5). There was no effect of hyperoxia on U_TL in B. medius but U_TL did increase (0.43°C) under hyperoxic conditions in F. lapillum (Fig. 5).

In B. medius and F. lapillum, ventilation frequency during thermal ramping was significantly lower under hyperoxia than under normoxia (B. medius: ANOVA, d.f.=1, F=13.74, P=0.002 and F. lapillum: ANOVA, d.f.=1, F=10.65, P=0.006; Fig. 6A) but there was no effect of hyperoxia on ventilation amplitude (Fig. 6B). Estimated ventilation volume during thermal ramping was also significantly lower under hyperoxia in both species (B. medius: ANOVA, d.f.=1, F=44.1, P<0.001 and F. lapillum: ANOVA, d.f.=1, F=47.93, P<0.001; Fig. 6C). During thermal ramping, ventilation frequency increased significantly at higher temperatures in B. medius (ANOVA, d.f.=4, F=11.44, P<0.001; Fig. 6Ai) but there was no effect of temperature in F. lapillum (ANOVA, d.f.=4, F=1.04, P=0.37; Fig. 6Aii). Contrastingly, there was no change in ventilation amplitude across temperature in B. medius (ANOVA, d.f.=4, F=0.96, P=0.43; Fig. 6Bi) but, in F. lapillum, ventilation amplitude increased at higher temperatures (ANOVA, d.f.=4, F=28, P<0.001; Fig. 6Bii). In both species, estimated ventilation volume increased with thermal ramping (B. medius: ANOVA, d.f.=4, F=21.98, P<0.001 and F. lapillum: ANOVA, d.f.=4, F=29.9, P<0.001; Fig. 6C).

This study shows that hyperoxia increases Ṁ_O2,max and MS in intertidal triplefin fish. Ṁ_O2,max was higher with hyperoxia than under normoxia at ambient temperature (21°C) and also after thermal ramping to ecologically relevant temperatures (29°C). Consequently, despite higher SMR and routine Ṁ_O2, MS at 21°C and MS above routine Ṁ_O2 at 29°C increased under hyperoxia. At the start of thermal ramping events in rock pools, intertidal fish are first exposed to normoxia and ambient temperature, but, as temperature increases, rock pools can also become super-saturated with oxygen (see Fig. 1B for an example). Therefore, comparisons of MS between ambient temperature normoxic conditions and high-temperature hyperoxic conditions are relevant. In hyperoxia, 79% and 64% of the MS available under normoxia and ambient temperature (21°C) was retained after thermal ramping to 29°C in B. medius and F. lapillum, respectively. This contrasts with the near-complete collapse of MS in both species at 29°C under normoxia or, more specifically, without hyperoxia (Fig. 4). While hyperoxia did not provide a meaningful increase in the absolute thermal tolerance limits of either species, the observed retention of MS may allow intertidal fish to continue performing aerobically demanding activities, such as feeding and digestion, to higher temperatures than would otherwise be possible. These findings indicate that environmental hyperoxia may provide a ‘metabolic refuge’ for intertidal fish exposed to acute thermal stress during low tide ebbs if photosynthesis is active.

MS is thought to be a key determinant of fitness as it represents the boundaries within which aerobic activities can be performed (Claireaux and Lefrancois, 2007; Clark et al., 2013). Indeed, positive correlations between growth performance and MS have been demonstrated in several fish species (Claireaux et al., 2000; Claireaux and Lefrancois, 2007; Jobling, 1981; Khan et al., 2014b), including the triplefin F. lapillum (McArley et al., 2017). If increased availability of MS does indeed help to mitigate a prioritisation or constraint on simultaneous energetic processes (Claireaux and Lefrancois, 2007; Clark et al., 2013; Lefrançois and Claireaux, 2003), then an expansion of MS under hyperoxia may offset expected losses in performance during acute high-temperature exposure. For example, if rock pools become hyperoxic and MS is expanded, feeding and digestion may be better supported at high temperatures, despite increased basal costs. Interestingly, Hilton et al., (2008) showed that the intertidal triplefin B. medius was more abundant in rock pools with a high algal coverage, suggesting a preference for rock pools likely to become hyperoxic during low tide. Greater MS under hyperoxia may also allow rock pool fish to avoid an anaerobic acidosis, which may develop during thermal ramping as anaerobic metabolism increases to meet ATP demands (e.g. Clark et al., 2008). Indeed, Devor et al. (2016) found that hyperoxia mitigated the rise in plasma and muscle lactate levels in two Antarctic notothenioid fishes exposed to thermal ramping. So, whether hyperoxia helps intertidal fish recover from heat stress by avoiding a metabolic acidosis or whether hyperoxia assists other aspects of performance (feeding, digestion, growth, etc.) during thermal ramping would of course be interesting avenues for future research.

The exclusively intertidal triplefin B. medius has a higher MS than F. lapillum, a species found more commonly in shallow sub-tidal habitats. Bellapiscis medius is also found in higher intertidal zone rock pools than F. lapillum (T.J.M., personal observation) and occupies habitat more prone to acute high-temperature events and, possibly, night-time hypoxia too. Whether it is increased basal metabolic costs during thermal ramping or constrained maximum metabolic capacity during hypoxia, intertidal fish regularly face environmental conditions that limit MS. The high MS of B. medius may therefore reduce the occurrence of situations where limiting environmental conditions result in the prioritisation or constraint of aerobic activities (e.g. growth, reproduction, swimming, etc.).

In the current study, SMR and routine Ṁ_O2 both increased under hyperoxic conditions but not to the extent that Ṁ_O2,max increased. Thus, the observed increase in MS under hyperoxia was driven primarily by the response of Ṁ_O2,max and this is in line with the observations of Brijs et al. (2015), who found that the MS of European perch approximately double at 23°C under hyperoxia (∼200% air saturation) as a result of expansion in Ṁ_O2,max. This is not universal among all fish species, however, because hyperoxia did not increase Ṁ_O2,max or MS in either rainbow trout (Duthie and Hughes, 1987) or common sole (Lefrançois and Claireaux, 2003). Brijs et al., (2015) also suggested a higher Ṁ_O2,max under hyperoxia at temperatures approaching thermal limits (∼34°C) but the results of the present study are the first to directly demonstrate an increase in Ṁ_O2,max under hyperoxia following thermal ramping.

So, if MS expansion under hyperoxia results from increased Ṁ_O2,max, how does this expansion occur? The oxygen consumption rate of an intact animal is the sum of aerobic metabolic demands of all tissues, which is mostly determined by the oxygen carrying capacity of blood and the rate at which the circulatory system delivers oxygen to respiring tissues. While hyperoxia probably increases Ṁ_O2,max by modulating these factors, the exact mechanisms by which this occurs remain unclear. European perch lacks a coronary circulation and, therefore, relies on the venous blood return to supply oxygen to the heart (Ekström et al., 2016; Santer and Walker, 1980). Although we do not know whether triplefin fish also lack a coronary circulation, it is not present in another blennioid fish species (Blennius pholis) (Santer and Walker, 1980). If triplefin fish do indeed lack a coronary circulation like European perch, one possibility is that cardiac output is improved under hyperoxia following exhaustive exercise as a result of an increased supply of oxygen to the heart via the venous blood return. Indeed, elevated partial pressure of oxygen in venous blood has been demonstrated in fish exposed to hyperoxia (Ekström et al., 2016; Takeda, 1990; Wilkes et al., 1981). This could also explain why there was no effect of hyperoxia on Ṁ_O2,max in rainbow trout (Oncorhynchus mykiss) (Duthie and Hughes, 1987) because salmonids possess a well-developed coronary circulation which supplies the heart directly with oxygenated blood (Ekström et al., 2017; Santer and Walker, 1980) and this could theoretically override the influence of hyperoxia on venous blood oxygen content and hence cardiac output. An alternative suggestion is that cutaneous oxygen uptake is increased in hyperoxia (Brijs et al., 2015; Ekström et al., 2016; Wang et al., 2014). There is evidence that the contribution of the skin to gas exchange is upregulated following exercise in fish (Rummer et al., 2014) and cutaneous respiration provides a relatively greater contribution to the total Ṁ_O2 of carp (Cyprinus carpio) under hyperoxia (Takeda, 1989).

While increased Ṁ_O2,max in hyperoxia may be beneficial for rock pool fish through enabling MS expansion, hyperoxia does appear to elevate basal metabolic costs. Here, we hypothesised that basal metabolic costs decrease under hyperoxia, yet the SMR at ambient temperature (21°C) and routine Ṁ_O2 at high temperatures (26–29°C) during thermal ramping were higher under hyperoxia than under normoxia. This observation is somewhat paradoxical because, as expected, our measures showed that ventilation effort (and presumably the energetic cost of ventilation) was decreased with hyperoxia (Fig. 6). This cost pattern was also apparent for the rainbow trout, which showed a transient increase in Ṁ_O2 despite a 50% reduction in ventilatory flow rate under hyperoxia (Wood and Jackson, 1980), and spotted wolfish (Anarhichas minor) also showed higher routine Ṁ_O2 under hyperoxia (Foss et al., 2003). A similar decrease in ventilation with hyperoxia has also been seen in carp (Takeda, 1990), the white sucker (Catostomus commersonii) (Wilkes et al., 1981) and an intertidal goby (Gobius cobitis) (Berschick et al., 1987), but in these species Ṁ_O2 was unchanged, suggesting a relatively smaller increase in basal costs with hyperoxia. Taken together, these findings suggest the basal metabolic costs of fish are raised under hyperoxia and these costs appear to override the reduced costs of ventilation. While higher basal metabolic costs under hyperoxia put a performance benefit of hyperoxia in question, in the current study they occurred alongside comparatively larger increases in Ṁ_O2,max, such that MS was still expanded with hyperoxia. Clearly, further work is required to determine whether intertidal fish facing thermal ramping are able to derive a performance benefit from expanded MS under hyperoxia.

Despite MS increasing at temperatures approaching critical limits, hyperoxia did not provide a biologically relevant increase in the absolute thermal limits of either triplefin species. This finding is in agreement with other recent studies, which found no or only a small increase in the critical thermal maximum of fish under hyperoxia (Brijs et al., 2015; Devor et al., 2016; Ekström et al., 2016; Healy and Schulte, 2012a). Furthermore, the U_TL of B. medius was only 0.55 and 0.23°C higher than that of F. lapillum under normoxia and hyperoxia, respectively, even though Ṁ_O2,max in B. medius was substantially higher (32–42% greater Ṁ_O2,max) under all conditions. These findings therefore contribute to a growing body of literature suggesting that the absolute thermal limits of fish are not limited by the ability of the cardio-respiratory system to deliver oxygen to tissues (Brijs et al., 2015; Devor et al., 2016; Ekström et al., 2016; Ern et al., 2016; Healy and Schulte, 2012a; Wang et al., 2014). These findings also suggest that the constraints that set the absolute thermal limits of these species develop at a similar temperature under normoxia and hyperoxia. What these constraints are and when they develop in relation to loss of equilibrium is unknown, but beyond this temperature, any expansion of MS under hyperoxia would probably be inconsequential as death would rapidly ensue.

The most extreme rock pool temperature observed over a 3 week period in summer was 30.1°C (Fig. 1A). While this temperature is below the U_TL of each triplefin fish species, it is within 2°C of their absolute thermal limits even under hyperoxic conditions. Moreover, the rate of temperature increase used to determine U_TL (2°C h⁻¹) was faster than the rate at which peak temperatures observed in rock pools were reached (1.3°C h⁻¹). Therefore, as U_TL is often reduced under slower rates of warming (Mora and Maya, 2006), the absolute thermal limits of these species may be even closer to observed maximum rock pool temperatures under fully replicated rock pool conditions. This presents a cause for concern because climate change is expected to result in more frequent and extreme transient heat waves (Perkins et al., 2012), which could push temperature maximums of many rock pools beyond the tolerance limits of temperate intertidal triplefin fishes. In this scenario, intertidal fishes may experience habitat loss and increased risk of overheating, unless intergenerational adaption can select for more thermally tolerant phenotypes.

A limitation of the current study was that both species were acclimated to constant environmental conditions prior to the assessment of U_TL. In the wild, rock pool fish are exposed to constantly fluctuating physico-chemical conditions which could potentially condition U_TL beyond the levels observed in the present investigation. The influence of prior exposure to fluctuating temperature and other relevant environmental conditions on U_TL in intertidal fishes should be considered in future studies.

In the intertidal triplefins B. medius and F. lapillum, hyperoxia increased Ṁ_O2,max and MS both at ambient temperature and following acute thermal ramping to temperatures observed in rock pools. This finding suggests that algae-mediated hyperoxia in rock pools mitigates the collapse of MS in intertidal fish which would otherwise occur under normal air saturated oxygen conditions and thermal ramping. Retention of MS under hyperoxia offers several potential benefits to intertidal fish facing acute high-temperature exposure including less severe constraint of aerobically demanding activities and mitigation of anaerobic stress. However, as hyperoxia did not meaningfully increase the absolute thermal limits of either species, further studies are required to establish whether there are benefits of hyperoxia for intertidal fish, and at what temperatures they occur relative to loss of equilibrium. The U_TL of both B. medius _­and F. lapillum was less than 2°C above maximal observed rock pool temperatures, even with hyperoxia. This confirms that these species, particularly the exclusively intertidal B. medius, reside in a habitat where they are periodically exposed to temperatures close to their thermal tolerance limits. With more extreme heat waves predicted with climate change, intertidal triplefins may face loss of habitat and increased risk of heat death as rock pool temperature maximums increase. With this understanding, a comprehensive analysis of the likely impact of climate change on the availability of thermally suitable rock pool habitats is warranted.

Technical staff Peter Browne and Errol Murray are acknowledged for their assistance with design and fabricating the respirometry setups, and John Atkins is acknowledged for design of the respirometry software. We thank Stefan Andreef for providing the illustration used in Fig. 3.