Tolerance of an acute warming challenge declines with body mass in Nile tilapia: evidence of a link to capacity for oxygen uptake

Global change is having profound effects on aquatic ecosystems worldwide, with poorly known consequences for fish populations. Global warming is causing a rise in average seasonal temperatures and, more importantly, a higher frequency and intensity of acute summer heatwaves (Smale et al., 2019; Stillman, 2019). This has stimulated a wave of research into the mechanisms that underly tolerance of warming in fishes, especially into the Fry paradigm and its mechanistic elaboration, the Oxygen and Capacity Limited Thermal Tolerance hypothesis (here abbreviated for convenience as Fry-OCLTT; reviewed in Lefevre et al., 2021). The Fry-OCLTT posits that upper thermal tolerance in fishes is defined by physiological capacity to provide oxygen for the demands of aerobic metabolism as these are increased by warming, a capacity that ultimately reaches a physiological ceiling, leading to functional collapse (Eliason et al., 2011; Fry, 1971; Pörtner, 2010; Pörtner and Farrell, 2008; Schulte, 2015). Although the Fry-OCLTT paradigm seems intuitive, empirical evidence for it is mixed and therefore its universality remains open to question (Farrell, 2016; Jutfelt et al., 2018; Lefevre, 2016; Lefevre et al., 2021; Pörtner, 2021).

A major biological phenomenon that has been correlated with ongoing global warming is a progressive reduction in the adult size of many fish species (Audzijonyte et al., 2020; Baudron et al., 2014; Daufresne et al., 2009). This may be a consequence of the temperature–size rule (TSR; Atkinson, 1994), which derives from laboratory and field observations that adult body size declines when aquatic ectotherms are grown at warmer temperatures (Audzijonyte et al., 2020; Hume, 2019; Loisel et al., 2019). The mechanisms underlying the TSR remain to be understood (Verberk et al., 2021; Wootton et al., 2022) but it has been suggested that, at warmer temperatures, larger individuals may be prone to limitations in their capacity for oxygen uptake, so that species may have evolved to modify their growth trajectories and avoid constraints on aerobic activities in adults (Atkinson et al., 2006; Hoefnagel and Verberk, 2015; Leiva et al., 2019; Rubalcaba et al., 2020; Verberk et al., 2021). Modelling across species provides some support for potential oxygen supply limitation of larger fishes at warm temperatures (Rubalcaba et al., 2020) but studies of how body size and temperature interact to affect oxygen supply capacity within species are very rare (Tirsgaard et al., 2015).

If we accept that oxygen supply limitation could define upper thermal tolerance in fishes and that larger individuals might be more prone to suffer from oxygen limitation at warm temperatures, a prediction would be that tolerance will decline with increasing body size. If true, this would be important in projecting how global warming will affect fish populations. Leiva et al. (2019) provided evidence that, in aquatic ectotherms generally, upper thermal tolerance declines with body and cell size of species. Within fish species, studies using the critical thermal maximum (CT_max) protocol have found that upper thermal tolerance drops with increasing body size in some species, although not in all (reviewed in McKenzie et al., 2020). The CT_max protocol is the standard method to evaluate acute thermal tolerance in fishes and ectotherms in general; animals are warmed incrementally until loss of equilibrium (LOE), the tolerance endpoint (Beitinger and Lutterschmidt, 2011). Although the CT_max protocol is easy to perform and is a repeatable trait within fish individuals and populations (Claireaux et al., 2013; McKenzie et al., 2020; Morgan et al., 2018), it has the weakness that the mechanisms for LOE are not understood. They presumably reflect acute loss of function of critical organs such as the heart and central nervous system, but may differ among species and with warming rate (Ekstrom et al., 2018; Jutfelt et al., 2019; Lefevre et al., 2021; McKenzie et al., 2020; Morgenroth et al., 2021; Rezende and Bozinovic, 2019; Rezende et al., 2011). In particular, there is evidence that LOE may not always be attributable to problems with systemic oxygen supply, which has been a major source of contention in the debate surrounding the universality of the Fry-OCLTT paradigm (Blasco et al., 2020; Brijs et al., 2015; Ekström et al., 2016; Ern et al., 2016; Lefevre et al., 2021; Wang et al., 2014). Some studies have considered whether size-dependent warming tolerance might be linked to the capacity for oxygen uptake within species (Christensen et al., 2020; Messmer et al., 2017) but this remains to be tested explicitly.

Blasco et al. (2020) proposed a measure of upper thermal tolerance whereby fish are warmed incrementally while they perform intense but steady aerobic exercise in a swimming respirometer (Steinhausen et al., 2008), with fatigue as the tolerance endpoint. In two tropical freshwater species, warming caused a progressive and profound increase in oxygen uptake up to a maximum rate, at which both species transitioned from steady aerobic swimming to an unsustainable anaerobic swimming gait, which rapidly led to fatigue. Fatigue in both species occurred at temperatures significantly lower than their CT_max based on LOE (Blasco et al., 2020). The protocol evaluates the capacity for oxygen uptake because it challenges a fish to meet the oxygen demands of a constant metabolic load, aerobic exercise, plus an incremental metabolic load, progressive warming; fatigue presumably occurs when a fish can no longer meet these combined demands (Blasco et al., 2020; Steinhausen et al., 2008).

We used this swimming protocol (called CT_swim in Blasco et al., 2020, but see later) to investigate whether tolerance of acute warming declined with increasing body mass in Nile tilapia, Oreochromis niloticus, and whether this could be related to a reduced capacity for oxygen uptake with increasing mass. The Nile tilapia is a relatively eurythermal tropical freshwater teleost (natural thermal range 14–33°C) that provides important fisheries in tropical and sub-tropical countries across the globe (De Silva et al., 2004; Schofield et al., 2011). We studied a group of fish that ranged over one order of magnitude in body mass; we first measured key traits of respiratory metabolism and performance at the acclimation temperature (26°C) by swimming respirometry, in order to better interpret any effects of mass on thermal tolerance. We then investigated the effects of mass on thermal tolerance using the CT_max and the critical thermal maximum for aerobic swimming (CTS_max, previously called CT_swim) protocols. We correlated variation in CTS_max with variation in maximum rates of oxygen uptake at fatigue, as evidence of a causal relationship.

Experiments were performed according to Brazilian National Council for Control of Animal Experimentation (CONCEA) regulations; the protocol was approved by the Ethics Committee on Animal Use of the Federal University of São Carlos (CEUA/UFSCAR), protocol number CEUA 3927151016.

Oreochromis niloticus (Linnaeus 1758) of unknown sex but of different sizes and age classes were obtained from the population at the Piscicultura Pollettini fish farm in Mogi Mirim (SP) Brazil, and held in recirculating biofiltered water at 25±1°C in outdoor tanks (1000 l volume) at the Department of Physiological Sciences, UFSCar, São Carlos (SP), Brazil. Eighteen animals were selected with a mass range greater than one order of magnitude (20.6 g to 313.0 g), studied in three sequential relatively size-matched groups of six individuals, with the treatment protocol shown in Fig. 1. Initially, they were tagged for identification (PIT under benzocaine anaesthesia) then allowed to recover for at least 96 h in an indoor tank (100 l volume) in the biofiltered water system with routine feeding. Fish were fasted 24 h prior to experimentation, and weighed prior to overnight recovery in each treatment (Fig. 1). They did not change in mass between sequential tests of performance and tolerance (see below).

Following tagging, each individual's metabolic and performance phenotype was measured at 26±0.1°C (Fig. 1) with a Steffensen-type swim-tunnel, exactly as described in Blasco et al. (2020). Briefly, fish were placed in the tunnel (13.4 l volume) supplied with vigorously aerated, biofiltered water, and left overnight at a low swimming speed equivalent to 1 body length s⁻¹ (BL s⁻¹, corrected for solid blocking effect; Blasco et al., 2020). The following morning, swimming speed was increased each 30 min in steps of either 1 BL s⁻¹ (mass range 20.6–86.5 g, n=12) or 0.5 BL s⁻¹ (204–313 g, n=6), up to either 5 or 2.5 BL s⁻¹, respectively. Speeds differed for the size groups because relative swimming speed declines with size (length) in fishes (Bainbridge, 1958; Beamish, 1978), and the objective to was obtain a range of speeds at which all individuals exclusively engaged steady aerobic body-caudal swimming (Webb, 1998). Measurements of oxygen uptake rate (Ṁ_O2) were made by stopped-flow respirometry at each speed and calculated as both absolute (mmol fish⁻¹ h⁻¹) and mass-corrected (mmol kg⁻¹ h⁻¹) rates, to subsequently derive standard metabolic rate (SMR) (Chabot et al., 2016) and active metabolic rate (AMR) (Norin and Clark, 2016) as described in Blasco et al. (2020), and then calculate aerobic scope (AS) as AMR−SMR.

Following this element of the swim test, the swimming speed was increased by 0.1 BL s⁻¹ every 10 s until fatigue, with this speed noted as maximum swimming speed (U_max; Blasco et al., 2020; Marras et al., 2013). When the fish fatigued and fell back against the rear screen of the swim tunnel, speed was immediately reduced to 1 BL s⁻¹ and, after 30 min, the fish was returned to a second holding tank for at least 96 h under normal rearing conditions, prior to testing thermal tolerance. Blank measurements were made before and after each trial and, if they increased over a trial, were presumed to do so linearly. Background respiration did not exceed 10% of fish Ṁ_O2 and all data were corrected accordingly.

During the constant acceleration element of the protocol, gait transition speed (U_GT) was identified as when an individual began to engage an unsteady anaerobic gait, which presaged fatigue (Blasco et al., 2020; Marras et al., 2013). The aerobic gait comprised steady body-caudal swimming (rhythmic beating of the tail) that relies upon slow-twitch oxidative muscle (Hachim et al., 2021; Webb, 1998). The unsteady anaerobic gait comprised intermittent powerful tailbeats that propelled the fish forward in the swim tunnel, after which it coasted back and then repeated the powerful tailbeat, the so-called ‘burst and coast’ swimming, which relies on recruitment of fast-twitch glycolytic muscle (Hachim et al., 2021; Webb, 1998). The net difference between U_max and U_GT (U_max−U_GT) was calculated as a potential indicator of relative anaerobic swimming capacity.

After suitable recovery in their holding tank, fish were measured for thermal tolerance by CT_max and CTS_max. The order of the two tests was reversed for 50% of animals (Fig. 1) to avoid any potential systematic effect of exposure history on responses to the second test. CT_max was measured on groups of tilapia that had acclimated overnight in a 68 l tank containing vigorously aerated water at 26°C. The following morning, water was warmed by 1°C every 30 min until LOE. Individual CT_max was then recorded as the highest temperature step fully completed plus the proportion of the last step that the fish endured prior to LOE (Blasco et al., 2020). Immediately upon LOE, fish were recovered at 26°C for at least 30 min, in a 68 l tank of aerated water. If this was the first test applied (Fig. 1), the animal was then returned to a holding tank for at least 96 h prior to the CTS_max.

For CTS_max, each individual recovered overnight in the swim-tunnel at 1 BL s⁻¹ then, the next day, speed was increased over 30 min until 85% of its own U_GT (Blasco et al., 2020). That is, each fish was exercised at a fixed proportion of their own aerobic swimming capacity, which was expected to impose a metabolic load that represented a similar proportion of their AS. After 30 min at that speed, temperature was increased by 1°C every 30 min until fish fatigued, resting against the rear screen (Blasco et al., 2020). The CTS_max was calculated as for CT_max but using fatigue as the endpoint (Blasco et al., 2020). The fish was immediately placed in a recovery tank at 26°C for 30 min. If this was the first test applied (Fig. 1), the animal was then returned to a holding tank for at least 96 h prior to the CT_max.

At each temperature step during the CTS_max, measurements of Ṁ_O2 were made over the last 20 min and the highest rate achieved was denoted as Ṁ_O2,max (Blasco et al., 2020). Fatigue in the CTS_max was not simply due to lack of endurance; Nile tilapia acclimated to 26°C can swim steadily at 85% of U_GT for at least 9 h, exceeding the duration of the CTS_max (Blasco et al., 2020). Blank measurements were made before all trials and at the highest temperature achieved after all trials; increases in background were presumed to be linear over temperature. Background respiration did not exceed 10% of fish Ṁ_O2 and all data were corrected proportionally.

No animals were excluded from analyses; data are available from Mendeley (doi:10.17632/npmchftsjf.1). Statistics were performed with SigmaPlot 11. Relationships with body mass of metabolic, performance and tolerance variables were assessed by least squares regression of log-transformed data. Correlations between variables (CT_max versus CTS_max; CTS_max versus Ṁ_O2,max) were assessed by Pearson product-moment correlation. Mean CT_max and CTS_max were compared by paired t-test assumptions of normality and equal variance was tested by Shapiro–Wilks and F-test, respectively. P<0.05 was the limit for statistical significance.

Mass-independent SMR, AMR and AS showed highly significant positive relationships with mass (Table 1; Fig. S1). Slopes for SMR and AMR were less than 1, indicating an allometric decline in metabolic rate with increasing mass (Table 1). By contrast, the slope for AS was very close to 1, indicating that this trait scaled isometrically with mass (Table 1). In order to better evaluate how metabolic traits depended on body mass (especially the Ṁ_O2,max during CTS_max, see below), data are presented here as plots of mass-specific Ṁ_O2 (Table 1). At 26°C, log mass-specific SMR and AMR both showed a clear decline with increasing log body mass, with significant negative linear relationships (Table 1, Fig. 2). However, log mass-specific AS did not show any significant relationship to mass, indicating that it was maintained unchanged as mass increased (Table 1, Fig. 2). Relative swimming performance declined significantly with mass; log U_GT and U_max both fell significantly with log mass, but U_max−U_GT was independent of mass (Table 2).

As fish were warmed during CT_max they first exhibited erratic behaviour, rolling sideways, followed by complete LOE. During CTS_max, fish swam using a steady aerobic gait up to a certain temperature, beyond which they progressively engaged unsteady burst and coast anaerobic swimming that led to fatigue. No fish lost equilibrium during CTS_max. All fish from both protocols recovered normal swimming behaviour within 10 min at 26°C. The two thermal thresholds were not correlated (r=0.371, P=0.129) but overall mean (±s.e.m.) CT_max (40.7±0.23°C) was significantly higher than CTS_max (38.1±0.29°C). Log CT_max had no relationship with log body mass whereas log CTS_max declined significantly with mass (Fig. 3, Table 2).

At the outset of the CTS_max test, when individuals were swimming at 85% of their U_GT at acclimation temperature, absolute and mass-specific Ṁ_O2 declined with body mass, with slopes that were similar to those for AMR (Table 1; Fig. S2). When each individual's mass-specific SMR was subtracted from their mass-specific Ṁ_O2 at 85% of their U_GT, the net difference represented 71±3% (mean±s.e.m.) of mass-specific AS across all 18 individuals, with no significant relationship to their mass (Fig. S3). During the CTS_max test, Ṁ_O2,max occurred at the temperature step that was either the penultimate or final step endured, in all individuals except for one (mass 204 g; see data in Mendeley: doi:10.17632/npmchftsjf.1). There was a highly significant negative relationship between mass-specific Ṁ_O2,max and mass (Fig. 4A, Table 1; see Fig. S1 for the reciprocal allometric scaling relationship). There was a significant positive correlation between log Ṁ_O2,max and log CTS_max (Fig. 4B).

The results demonstrate that tolerance of acute warming declines with increasing body mass in a teleost fish, when measured as the temperature at which individuals fatigue from sustained aerobic swimming (CTS_max). The decline in CTS_max with mass was significantly correlated with a decline in Ṁ_O2,max, which may indicate a causal relationship.

The tilapia exhibited the significant mass-specific declines in SMR and AMR at acclimation temperature that reflect the allometric mass-scaling relationships of metabolic rate found in all fish species. The underlying mechanisms remain a matter of debate (Glazier, 2020; Killen et al., 2010, 2016). An absence of any effect of mass on AS has also been observed in multiple species, the implication being that, although SMR and AMR decline with mass, the animal maintains the same absolute scope to provide its tissues with oxygen (Killen et al., 2016; Lefevre et al., 2017). At the outset of the CTS_max protocol, when individuals were swimming at 85% of their U_GT at acclimation temperature, their associated Ṁ_O2 declined with mass but nonetheless represented a constant proportion of their AS (∼70%). The fact that relative U_GT and U_max, measured in BL s⁻¹, declined with mass at acclimation temperature was of course a direct reflection of the tight relationship between mass and length in the tilapia. It is well established that relative swimming performance declines with length in most fish species (Bainbridge, 1958; Beamish, 1978).

The finding that CT_max did not decline with increasing body mass is consistent with a previous study on Nile tilapia (Recsetar et al., 2012). It may indicate that LOE during CT_max in tilapia is not due to limitations in oxygen supply but instead occurs as a result of another mechanism. Although LOE presumably reflects loss of function of critical organs such as the heart and central nervous system, further study is required to understand the exact mechanism in tilapia or, for that matter, in any fish species (Jutfelt et al., 2019; Lefevre et al., 2021). Given the absence of any dependence of CT_max upon mass, it was interesting to find a clear negative dependence of CTS_max. This dependence of CTS_max on mass in Nile tilapia should be investigated in other fish species, to establish the generality of the phenomenon.

The profound decline in Ṁ_O2,max with body mass, and the significant correlation of Ṁ_O2,max with CTS_max, are also worthy of further investigation to establish whether capacity for oxygen uptake is indeed a mechanism underlying size dependence of CTS_max. This could, for example, include measurements of cardiac performance, blood flow and blood gases during a CTS_max test in individuals of different sizes, in species that are amenable to both blood vessel catheterisation and placement of flow probes. These results in the tilapia lend support to the Fry-OCLTT paradigm whereby thermal tolerance in ectotherms can be limited by the capacity for oxygen uptake (Blasco et al., 2020; Fry, 1971; Pörtner, 2010). However, the fact that size-related variation in Ṁ_O2,max only accounted for about 60% of the variation in CTS_max indicates that other mechanisms must have contributed to fatigue in the tilapia. All individuals engaged unsustainable anaerobic swimming prior to fatiguing (Blasco et al., 2020) and the capacity for anaerobic exercise varies among individuals in fish species (Marras et al., 2010, 2013). If, however, the difference between U_max and U_GT (U_max−U_GT) is considered an indicator of relative anaerobic swimming capacity (Marras et al., 2013), this was independent of mass and length.

It is also important to understand why Ṁ_O2,max declined so significantly with mass, and to establish whether this is also a general phenomenon across fish species. A regression coefficient of 0.93 on 18 fish indicates that the relationship is unlikely to be spurious in the tilapia, which is therefore a stimulus to investigate the phenomenon in other fish species. The slope for the decline in log mass-specific Ṁ_O2,max with log mass in the tilapia of −0.33 would be expected if the capacity for oxygen uptake were subject to surface to volume geometric constraints (Glazier, 2018, 2020); notably, if fishes were subject to the simple geometric constraints of spherical bodies (Pauly, 1981). Fishes are not, of course, spherical bodies; their gills are folded surfaces which, as an individual grows, can perfectly compensate increases in body volume with increases in respiratory surface area (Lefevre et al., 2017). This is why, at a given acclimation temperature, AS scales isometrically with body mass across multiple species (Lefevre et al., 2017), as it did with the tilapia in this study. It is known, however, that gill functional respiratory surface area declines with mass in fishes; in tilapia, it does so with a slope of −0.30 (Kisia and Hughes, 1992). This decline in gill surface area with mass may be linked to the decline in mass-specific SMR, whereby basal metabolic costs decline and the fish gills are then ‘tailored’ to maintain a constant scope for aerobic performance at any given mass. Maintaining a greater gill respiratory surface area than is needed for AS, at any given acclimation temperature, could have costs in terms of, for example, regulating water and ion balance (Lefevre et al., 2017; Taylor, 1998). As a consequence, however, the fish also show a decline in their maximum mass-specific oxygen uptake capacity at acclimation temperature, estimated as AMR. This design characteristic, of reduced gill surface area with increasing mass, would then become a constraint when large individuals face the profound challenge to oxygen uptake that the CTS_max protocol imposed. Although the established mass-related decline in gill respiratory surface area in the Nile tilapia seems like a coherent explanation for the decline in Ṁ_O2,max with mass, there may be multiple contributing factors, including other elements of the oxygen transport cascade. This is a topic that deserves further investigation, especially on species that are amenable to catheterisation and placement of flow probes.

Extreme warming events are predicted to increase in frequency and severity in aquatic habitats worldwide as a consequence of accelerating global climate change (Field et al., 2012; Frölicher et al., 2018; Stillman, 2019). There are clear ecological implications to the fact that, if tilapia populations experience an acute warming event, the ability of larger fishes to perform a given level of aerobic work would be constrained at lower temperatures than that of smaller conspecifics. Although we concentrated on aerobic swimming, larger tilapia would also presumably be constrained in their ability to perform other activities at warm temperatures; for example, the specific dynamic action response that follows ingestion of a meal and is believed to be in large part the cost of tissue anabolism and growth (McCue, 2006). The law of averages would predict that larger fishes are more likely to encounter temperatures sufficiently warm to limit their performance (Field et al., 2012; Stillman, 2019). Therefore, stochastic extreme warming events could threaten ecological performance, fitness and survival of fishes in a size-dependent manner (Habary et al., 2017; Stillman, 2019). If fitness and survival of larger animals is compromised more frequently, this could have important consequences for Nile tilapia fisheries, an important source of dietary protein in many countries (De Silva et al., 2004).

Overall, the data indicate that the capacity for oxygen uptake may be a factor limiting thermal tolerance in fishes, and also that global warming may ‘favour the small’ (Daufresne et al., 2009). Further work is required to understand whether this size dependence in temperature tolerance is true in other fish species. The evidence we provide, that larger individuals are more prone to limitations in their capacity for oxygen uptake at warmer temperatures, is coherent with a mechanism suggested to underly the TSR (Verberk et al., 2021). That is, that species may have evolved to modify their growth trajectories at warmer temperatures, to avoid constraints on aerobic activities in adults (Atkinson et al., 2006; Hoefnagel and Verberk, 2015; Leiva et al., 2019; Rubalcaba et al., 2020; Verberk et al., 2021). This, in turn, might be linked to ongoing declines in maximum body size of fishes globally (Audzijonyte et al., 2019; Baudron et al., 2014; Daufresne et al., 2009). The rate of global warming is too fast for adaptive evolutionary responses by most fish species, so survival and resilience will depend upon the options for migration, the prevailing plasticity in tolerance and the existence of tolerant genotypes (Bell, 2013; Gunderson and Stillman, 2015; Habary et al., 2017; Stillman, 2019). The current data indicate that the capacity for migration may be impaired at lower temperatures in larger fish, and that smaller animals are more tolerant. This could also contribute to a decline in maximum body size in future populations.

We are grateful to Cesar Polettini, of Piscicultura Polettini, for donating the tilapia.