Partitioning the metabolic scope: the importance of anaerobic metabolism and implications for the oxygen- and capacity-limited thermal tolerance (OCLTT) hypothesis

Climate change is predicted to affect the aerobic capacity available for sustainable performances and fitness in fishes. This study reveals trade-offs between aerobic and anaerobic components of swimming performance and metabolic scope, and highlights the possibility of overestimating sustainable aerobic performances of fishes in relation to climate change.


Introduction
Effects of climate change (e.g. increased temperature, ocean acidification and hypoxia) are predicted to have profound effects on the physiology of aquatic ectotherms, including fishes (Pörtner and Peck, 2010;IPCC, 2014;Deutsch et al., 2015). This has raised conservation concerns for the persistence of fish populations and increased the interest in developing predictive models for effects of climate change on different species (Jørgensen et al., 2012;Farrell, 2016). The oxygenand capacity-limited thermal tolerance (OCLTT) hypothesis uses the metabolic scope (MS) to derive a range of tolerable temperatures (thermal window) of organisms with an optimal temperature for oxygen supply to sustain life, including growth, foraging, migration and reproduction (Pörtner and Farrell, 2008;Pörtner and Peck, 2010;Holt and Jørgensen, 2015;Motyka et al., 2016;Verberk et al., 2016). The MS is defined as the difference between the maximal metabolic rate ( MO 2max ) and standard metabolic rate ( MO 2stand ). Importantly, OCLTT is concerned solely with the aerobic component of MS (Pörtner and Farrell, 2008;Eliason et al., 2011;Clark et al., 2013), essentially making the assumption that any anaerobic component is insignificant. Anaerobic metabolism within the MS is important because it depletes substrates (e.g. glycogen), accumulates metabolic waste (e.g. lactate) and results in fatigue (Alexander, 1989;Goolish, 1991a;Sänger and Stoiber, 2001;Hedrick et al., 2015). The fraction of the MS that is influenced by anaerobic metabolism is therefore not available for sustainable activity. Given that the OCLTT hypothesis concerns only aerobic performance (Pörtner, 2010), it is imperative that any anaerobic component of the MS is known and corrected for to ensure accurate predictions. Although previous studies have acknowledged an anaerobic component within MS (Goolish, 1991a;Reidy et al., 2000;Svendsen et al., 2012;Careau et al., 2014;Norin et al., 2014), the component has, to date, received little quantitative attention in relation to OCLTT. Fish acquire energy to accomplish different types of physiological work, such as biosynthesis (e.g. somatic growth), maintenance (e.g. circulation, respiration and osmoregulation) and generation of external work to allow locomotion (Careau et al., 2014). If all these functional traits were running at maximal rate, the oxygen requirements would exceed the available supply, forcing individual organisms to prioritize their oxygen budget within the finite size of MS (Killen et al., 2007;Guderley and Pörtner, 2010;Holt and Jørgensen, 2015). For example, the metabolic costs of locomotion and digestion exhibit a trade-off in several fish species (Priede, 1985;Jordan and Steffensen, 2007;Altimiras et al., 2008;Li et al., 2010), which has ecological and evolutionary implications in relation to important performance traits (e.g. predator evasion, foraging, growth, migration and reproduction; Reidy et al., 2000;Oufiero and Garland, 2009). In some fish species, however, metabolism associated with digestion may be additive to the metabolism associated with locomotion (Jourdan-Pineau et al., 2010). Performance trade-offs often play important roles in relation to phenotypic variation found between individuals (Oufiero et al., 2011;Seebacher and Walter, 2012;Svendsen et al., 2015) and may take place when two antagonistic traits cannot be optimized simultaneously because of conflicting demands on the same capacity (Priede, 1985;Roff and Fairbairn, 2007;Svendsen et al., 2015), such as the oxygen budget (Farrell, 2007;Altimiras et al., 2008). This implies that excellence in one trait comes at the cost of performance in a different trait (Vanhooydonck et al., 2014;Walker and Caddigan, 2015), which is classically exemplified by the conflicting relationship between sprinters and endurance athletes (Reidy et al., 2000;Van Damme et al., 2002;Marras et al., 2013). To date, evidence of the corresponding locomotory trade-off in fishes remains inconclusive, with some studies finding support (Reidy et al., 2000;Ojanguren and Braña, 2003;Langerhans, 2009;Oufiero et al., 2011;Ellerby and Gerry, 2011;Yan et al., 2012), whereas others have not Oufiero and Garland, 2009;Seebacher and Walter, 2012;Marras et al., 2013).
In many fish species, locomotor performance is powered by the myotomal musculature, consisting of segmented red and white muscle fibres. Red oxidative muscles are slow contracting, fuelled by aerobic metabolism and power steady, sustainable swimming (Webb, 1993(Webb, , 1998Kieffer, 2000;Sänger and Stoiber, 2001). When approaching swimming speeds that exceed the power capacity and contraction speed of the red muscle fibres, a gait transition to unsteady, unsustainable, burst-assisted swimming occurs with the activation of fast white muscle fibres. White fibres are mainly fuelled by anaerobic metabolism, and their activation depletes substrates (e.g. glycogen), accumulates metabolic waste (e.g. lactate) and results in fatigue (Webb, 1993;Kieffer, 2000;Sänger and Stoiber, 2001).
Anaerobic metabolism may occur at submaximal exercise levels (Goolish, 1991a;Svendsen et al., 2010) and before reaching the maximal aerobic metabolic rate (Burgetz et al., 1998;Lee et al., 2003;Hinch et al., 2006;Teulier et al., 2013). Hence, the gait transition to burst-assisted swimming can be used to partition swimming performance and the MS (Peake and Farrell, 2004;Peake, 2008;Marras et al., 2013;Svendsen et al., 2015) into a sustainable aerobic component and an unsustainable component strongly influenced by anaerobic metabolism. While trade-offs related to sustainable (aerobic) and unsustainable (anaerobic) swimming performances have been examined (Ellerby and Gerry, 2011;Yan et al., 2012;Marras et al., 2013), to date no study has tested for trade-offs related to sustainable and unsustainable components of the MS.
Based on existing data on fish swimming performance and metabolism (Svendsen et al., , 2015, the present study examined the OCLTT assumption that the fraction of the MS which is influenced by anaerobic metabolism, is insignificant. At increasing swimming speeds, the gait transition speed (U GT ) to burst-assisted swimming was used to partition swimming performance into a sustainable and strictly aerobic component (U sus ) and an unsustainable component influenced by anaerobic metabolism (U unsus ). This partitioned MS into the MS associated with sustainable swimming speeds ≤ U GT (sustainable metabolic scope; MS sus ) and the MS associated with unsustainable swimming speeds > U GT (unsustainable metabolic scope; MS unsus ). Using these data, we tested for trade-offs between the two swimming performance measures (U sus and U unsus ) and between the two measures of the MS (MS sus and MS unsus ). We predicted negative correlations between both groups of measures, implying that individuals cannot optimize U sus and U unsus or MS sus and MS unsus simultaneously.

Animals
A total of 13 gilthead sea bream (Sparus aurata; unknown sex; mean ± SEM body mass 79.8 ± 2.4 g and length 14.8 ± 0.2 cm) from a fish farm (Ferme Marine de Douhet) in France were maintained in a holding tank (0.7 m 3 ) with seawater (salinity of 30‰) at 10°C. In addition, 18 guppies (Poecilia reticulata; female) (body mass 0.296 ± 0.009 g and length 3.0 ± 0.0 cm) were captured in Trinidad and maintained in freshwater holding tanks (30 l) at 26°C. Fish were acclimated to the laboratory for at least 2 weeks and fed daily on commercial fish food.

Respirometry
Two swimming respirometers (8.4 and 0.17 l) were used to measure oxygen consumption rate ( MO 2 ; in milligrams of oxygen per kilogram per hour) as a function of swimming speed (U; in centimetres per second) in S. aurata and P. reticulata. Temperature-controlling instruments (TMP-REG; Loligo Systems, Tjele, Denmark) were employed to maintain temperatures at 10 and 26°C (±0.1°C), respectively. Oxygen partial pressure (in kilopascals) inside the respirometers was measured using fibre-optic sensor technology (PreSens, Regensburg, Germany). Intermittent flow respirometry (Forstner, 1983) was applied in accordance with previous studies (Steffensen, 1989;Peixoto et al., 2016). The software AutoResp (Loligo Systems Aps, Tjele, Denmark) was used to collect data and calculate MO 2 from measurements of oxygen content inside the respirometers (Peixoto et al., 2016).

Experimental protocol
Poecilia reticulata were fasted for 24 h, whereas S. aurata were fasted for 48 h to ensure post-absorptive states and then transferred to the respirometers. Fish were acclimated to the respirometers for 8-12 h (overnight) while swimming at a speed of 0.5 body lengths (BL) s −1 (S. aurata) and 2 BL s −1 (P. reticulata) prior to collection of data. These speeds were the minimal swimming speeds that ensured positive rheotaxis. Critical swimming speed protocols were then used to measure MO 2 at increasing swimming speeds until fatigue (Brett, 1964;Svendsen et al., 2013Svendsen et al., , 2015. The time interval at each swimming speed was 30 min for S. aurata and 12 min for P. reticulata, both including 15 s of speed increment to reach each new test speed. Unlike steady swimming, burst-assisted swimming is partly fuelled by anaerobic metabolism (Peake and Farrell, 2004;Peake, 2008;Marras et al., 2013). The number of bursts correlates positively with excess post-exercise oxygen consumption and therefore anaerobic metabolism (Svendsen et al., 2010), suggesting that the onset of burst-assisted swimming can be used as a reliable indicator of anaerobic metabolism (Svendsen et al., 2015). Hence, in parallel with the measurements of MO 2 , the onset of burst-assisted swimming was recorded. The gait transition speed, U GT , was defined as the highest swimming speed that was supported using only a steady, undulatory locomotory gait (equivalent to U STmax of Peake, 2008). The U sus was defined as the range of sustainable swimming speeds between zero speed and maximal swimming speed maintained by steady swimming (U GT ), whereas U unsus was defined as the range of unsustainable swimming speeds higher than U GT until the maximal swimming speed (U max ; Fig. 1A).

Data analyses
Applying data from individual fish, the relationships between U and MO 2 were described using an exponential equation (Brett, 1964;Fig. 1A). The analyses were limited to steady swimming speeds because the relationship between U and MO 2 tends to vary at speeds faster than U GT (Schurmann and Steffensen, 1997;Svendsen et al., 2013Svendsen et al., , 2015Killen et al., 2015). Extrapolating the exponential equation to zero speed provided an estimate of the standard metabolic rate ( MO 2stand ; Brett, 1964;McKenzie et al., 2003;Arnott et al., 2006;Fig. 1A). Maximal metabolic rate ( MO 2max ) was defined as the highest MO 2 measured during the complete swimming protocol (i.e. until fatigue; Binning et al., 2014). The MS was calculated as the difference between MO 2max and MO 2stand in individual fish (Fig. 1A).
The metabolic rate corresponding to maximal sustainable swimming speed was defined as the maximal metabolic rate that was maintained aerobically at speeds up to, and including, U GT without the accumulation of anaerobic metabolic products that contribute to performance (U) and negatively impact endurance (Hillman et al., 2014). Hence, swimming performance and MS were partitioned into the following two different components: (i) sustainable components supported by aerobic metabolism alone (U sus and MS sus ); and (ii) unsustainable components strongly influenced by anaerobic metabolism (U unsus and MS unsus ; Fig. 1A).
To test for trade-offs in S. aurata and P. reticulata at the intraspecific level, components were correlated using leastsquares linear regression (Zar, 2010). Regressions were multivariate, with the unsustainable components (i.e. U unsus and MS unsus ) as the dependent variables and the sustainable component (i.e. U sus and MS sus ) and body size (i.e. length and mass) as the independent variables. This was done to test whether predicted correlations were revealed independently of factors related to body size. Multivariate regressions were carried out using stepwise backward elimination. The trade-off related to swimming performance was tested by correlating U sus and U unsus , whereas the trade-off related to the MS was tested by correlating MS sus and MS unsus . In both cases, a trade-off would be revealed by a significant negative relationship.
In addition, we calculated the fractions (as percentages) of U max and MS that were constituted by the unsustainable components (i.e. U unsus and MS unsus , respectively). The unsustainable fractions (percentages) of U max and MS were then correlated with the corresponding sustainable components (i.e. U sus and MS sus ). The correlations were tested using leastsquares linear regressions. showing the metabolic rate as function of swimming speed, including metabolic scope (MS; (black double-headed arrow) and gait transition speed (U GT ) as the highest sustainable swimming speed, equivalent to U STmax of Peake (2008). Using U GT , swimming performance is partitioned into sustainable (ranging from zero speed to U GT ) and unsustainable (swimming speeds higher than U GT until U max ) components. The metabolic rate at U GT is used to distinguish between sustainable metabolic scope (MS sus ; blue double-headed arrow) and unsustainable metabolic scope (MS unsus ; red double-headed arrow). (B and C) Raw data showing oxygen consumption rate ( MO 2 ; in milligrams of oxygen per kilogram per hour) as a function of swimming speed (in centimetres per second) in an individual Trinidadian guppy (Poecilia reticulata; B) and gilthead sea bream (Sparus aurata; C) used in this study. Data are adapted from Svendsen et al. (2013Svendsen et al. ( , 2015. Grey symbols represent MO 2 when no burst-assisted swimming occurred, whereas red symbols represent MO 2 when burst-assisted swimming occurred. Tests were carried out using the software MATLAB 8.5 (MathWorks, 2015), and results were considered significant at P < 0.05. All values are reported as means ± SEM unless otherwise stated.

Burst-assisted swimming
Burst-assisted swimming was evident as forward movement in the chamber ( Fig. 2A), elevated tail beat frequency and amplitude (Fig. 2B) and increased swimming speed (Fig. 2C). Between bursts, fish swam slower than the test speeds (Fig. 2C), leading to backwards movement in the chamber ( Fig. 2A) and subsequent bursts until fatigue. Burst-assisted swimming was observed in all fish, except three P. reticulata.

Inconsistent correlations between sustainable and unsustainable swimming performances
Although a significant negative linear relationship between sustainable swimming performance (i.e. U sus ) and unsustainable swimming performance (i.e. U unsus ) was evident in S. aurata (P = 0.02, R 2 = 0.41; Fig. 3A), no correlation was found for P. reticulata (P = 0.86, R 2 = 0.002; Fig. 3B). For both species, the regression analyses indicated no effects of fish body length. The relationship between U sus and the fraction (percentage) of U max constituted by U unsus was significant for S. aurata (P < 0.01, R 2 = 0.61; Fig. 3C) but not for P. reticulata (P = 0.33, R 2 = 0.06; Fig. 3D). Given that correlations were significant in only one species, these results suggested inconsistent relationships between sustainable and unsustainable swimming performances.
In contrast to the swimming performance data (Fig. 3), MS data revealed consistent significant negative correlations between MS sus and MS unsus in both S. aurata (P = 0.02, R 2 = 0.41; Fig. 4A) and P. reticulata (P = 0.02, R 2 = 0.30; Fig. 4B), indicating a trade-off where individuals exhibiting superior MS sus exhibit inferior MS unsus . The regression analyses indicated no effects of body mass for either species. The relationships between MS sus and the fraction (percentage) of MS constituted by MS unsus were significant for both S. aurata (P < 0.01, R 2 = 0.64; Fig. 4C) and P. reticulata (P < 0.01, R 2 = 0.45; Fig. 4D).

Discussion
Using two teleost species, this study tested for locomotory performance trade-offs between U sus and U unsus ; however, contrary to our prediction, we found no evidence of a consistent negative correlations. Nevertheless, data indicated significant trade-offs within the MS between MS sus and MS unsus in both species. Earlier studies have acknowledged that anaerobic metabolism may occur within the MS (Goolish, 1991b;Reidy et al., 2000;Svendsen et al., 2012;Norin et al., 2014), but the fraction of the MS influenced by anaerobic metabolism remains largely unknown (Burgetz et al., 1998;Farrell, 2007). The present study is the first to report intraspecific variation in MS unsus in fish, and we found that MS unsus may comprise up to 61% of MS in individual fish. This fraction of MS was associated with burst-assisted swimming and therefore partly fuelled by anaerobic metabolism. On average, MS unsus constituted 24.3 and 26.1% of MS in S. aurata and P. reticulata, respectively, highlighting the importance of anaerobic metabolism within the MS. The fact that a significant fraction of the MS is influenced by anaerobic metabolism, and therefore not available for sustainable activity, could have important implications when MS is used to predict effects of environmental variation, particularly in relation to climate change and OCLTT.

Inconsistent correlations between U sus and U unsus
The negative correlation between U sus and U unsus observed in S. aurata indicates a locomotor performance trade-off. Conversely, we found no evidence for the same trade-off in P. reticulata. We recognize the fact that the significant correlation for S. aurata may be coincidental and may not hold true if a future study investigates the same relationship using a substantially increased sample size. Numerous studies have tested the conflicting nature of aerobic and anaerobic swimming performance, with equivocal results (Reidy et al., 2000;Ojanguren and Braña, 2003;Claireaux and Lefrançois, 2007;Oufiero et al., 2011;Seebacher and Walter, 2012;Yan et al., 2012;Marras et al., 2013). Reidy et al. (2000) found a negative relationship between aerobic and anaerobic swimming performance in Atlantic cod (Gadus morhua). This was supported by Oufiero et al. (2011), who presented evidence of a similar trade-off in Trinidadian killifish (Rivulus hartii).

6
Conservation Physiology • Volume 4 2016 Research article Figure 3: (A and B) Relationships between sustainable (U sus ) and unsustainable (U unsus ) swimming performances (see Fig. 1A for details). (C and D) Relationships between U sus and the fraction (percentage) of U max constituted by U unsus . Significant relationships for gilthead sea bream (S. aurata) were found between U sus and U unsus (A), and between U sus and the fraction (percentage) of U max constituted by U unsus (C). (B and D) No significant relationships were found in Trinidadian guppy (P. reticulata).
In contrast, Marras et al. (2013) did not detect a negative correlation between aerobic and anaerobic swimming performances in European sea bass (Dicentrarchus labrax). Comparing relationships between aerobic and anaerobic swimming performances across studies is difficult, because previous studies have employed a diversity of methods to determine swimming performance in many different species and in disparate environmental conditions (e.g. temperatures).
It has been argued that the spatial separation of red and white muscle fibres in most teleosts (Webb, 1993(Webb, , 1998Sänger and Stoiber, 2001;Johnston et al., 2011) is consistent with the absence of a locomotor trade-off because the aerobic and anaerobic muscle structures are recruited separately Marras et al., 2013). According to this conjecture, optimization of one type of muscle fibre does not come at cost to the other (Marras et al., 2013). However, assuming morphological constraints, a greater proportion of red muscle fibres would presumably occupy the space available for white muscle fibres and thus, suppress anaerobic performance. Consequently, variation in the proportion of white muscle fibres could constitute the mechanistic basis of negative relationships between aerobic and anaerobic swimming performances in individual fish.
Although metabolism is important for swimming performance (Priede, 1985;Binning et al., 2015), physiology is not the sole determinant of swimming performance. Several intraspecific studies have shown that swimming performance may be unaffected by variation in metabolism (Anttila et al., 2014;Svendsen et al., 2015) and strongly affected by morphology (Drucker and Lauder, 2000;Domenici et al., 2008;Langerhans, 2009) and biomechanics (Langerhans, 2009;Shadwick and Goldbogen, 2012;Svendsen et al., 2013). Consequently, intraspecific variation in morphology and biomechanics may mask physiological variation (and trade-offs) and could explain the inconsistent correlations between U sus and U unsus observed in the present study.

Consistent negative correlations between MS sus and MS unsus
Although this study provides inconsistent evidence for a trade-off related to swimming performance, our data show unambiguous support for a physiological trade-off between MS sus and MS unsus in two teleost species, suggesting that MS sus and MS unsus are two antagonistic traits that cannot be optimized simultaneously.
The mechanistic basis for the apparent trade-offs between MS sus and MS unsus or U sus and U unsus remain uncertain and poorly described. However, a general prediction is that a functional trade-off exists between sustainable and unsustainable locomotion for fishes using a coupled locomotor system, where the same body parts are used for propulsion in both sustainable and unsustainable locomotion (Yan et al., 2012). The red muscles are oxidative tissues and cannot function without the supply of oxygen in order to yield ATP (Webb, 1998;Seebacher and Walter, 2012). The supply of oxygen to sustain aerobic activity may depend on the ability of the fish to extract oxygen from the water (Priede, 1985;Davison, 1997;Sänger and Stoiber, 2001) and cardiac performance (Farrell, 2007;Eliason et al., 2011;Eliason and Farrell, 2015). In fact, it has been proposed that U GT is limited by the oxygen supply to the heart rather than to the red muscles themselves, implying that fish with superior cardiac performance also exhibit higher U GT (McKenzie and . The present study used the recruitment of white muscles (i.e. gait transition) to partition MS into MS sus and MS unsus , suggesting that MS sus could also be limited by the capacity of the red muscles (e.g. contraction speed; Alexander, 1989;Johnston and Altringham, 1991;Drucker and Lauder, 2000) and not by the oxygen supply alone. The contractile properties of the red muscle fibres may determine a peak power production corresponding to the maximal swimming speed that can be achieved before the white muscles are recruited (i.e. U GT ; Drucker and Lauder, 2000). For example, if a fish is constrained by the contraction speed of its red muscles, it may recruit the white muscles at a slow swimming speed, even if the oxygen supply is sufficient. Jayne and Lauder (1994) found that white muscles are recruited before the oxidative capacity of the red muscles is fully exploited, which potentially leaves an aerobic component within the red muscle to be used during unsustainable locomotion (i.e. burst-assisted swimming). Given that red muscles may remove catabolites produced in the white muscles (Wittenberger and Diaciuc, 1965;Johnston and Moon, 1980;Milligan and Girard, 1993;Richards et al., 2002), unexploited aerobic capacity would be available to metabolize anaerobic waste products and elevate MS unsus . Hence, if a fish is using only 70% of the red muscle capacity when the white muscles are recruited, then it would have a substantial aerobic component (30%) within the red muscles to metabolize lactate from unsustainable swimming. Consequently, supporting our findings, anaerobically influenced performance (i.e. U unsus and MS unsus ) would be enhanced at the expense of aerobic performance (i.e. U sus and MS sus ) and vice versa.
The fate of lactate, however, is not determined by the aerobic capacity of the red muscles alone, and other oxidative tissues, such as the liver, gills and heart (Milligan and Farrell, 1991;Milligan and Girard, 1993;Milligan, 1996;Omlin and Weber, 2013), may also contribute to the metabolism of lactate. Yet, studies have indicated that white muscles have a limited ability to export lactate in rainbow trout (Oncorhynchus mykiss;Weber, 1991;Omlin and Weber, 2013) and suggest that the separation of red and white muscle structures precludes intramuscular lactate shuttles (Weber, 1991;Teulier et al., 2013). If so, the suggestion that limiting contractile properties of the red muscles may leave an aerobic fraction for removal of anaerobic waste products would not underpin the metabolic trade-off observed in the present study.
Assuming that MS sus most directly depends on red muscles, expansion of the aerobic capacity presumably involves increasing the proportion of red muscles and the mitochondrial content of individual fibres (Johnston and Altringham, 1991;Wieser, 1995). Red muscle fibres comprise 10-30 times higher volume densities of mitochondria compared with white muscle fibres (Johnston and Altringham, 1991;Sänger and Stoiber, 2001;Moyes and Genge, 2010), implying that the ratio between the two fibre types could influence the magnitude of MS sus and MS unsus , because aerobic production of ATP takes place in the mitochondria (Johnston and Altringham, 1991;Sänger and Stoiber, 2001). This suggests that a fish with a bigger proportion of red muscle fibres could have more mitochondria per body volume, hence a greater MS sus (Moyes and Genge, 2010). However, it has been argued that the volume density of mitochondria is not always a reliable descriptor of the aerobic potential, owing to differences in mitochondrial cristae density and aerobic enzyme activity (Johnston and Altringham, 1991). Instead, several studies have scaled the content of mitochondrial and glycolytic (i.e. anaerobic) enzymes in relation to body size, and although exceptions exist (Siebenaller et al., 1982;Norton et al., 2000), data have revealed negative scaling relationships between the two types of enzymes (Childress and Somero, 1990;Wieser, 1995;Moyes and Genge, 2010), supporting our findings of a metabolic trade-off between MS sus and MS unsus . Nevertheless, the mechanistic basis of the apparent trade-off between MS sus and MS unsus within the MS needs further attention before firm conclusions can be drawn (Crans et al., 2015).

Fish behaviour in swim tunnels
Since the early work of Brett (1964), swim tunnels have become widely used tools to elucidate patterns of hydrodynamics (Liao, 2007), swimming performance (Schurmann and Steffensen, 1997;Tudorache et al., 2007Tudorache et al., , 2010 and physiology (Peake and Farrell, 2004;Clark et al., 2013;Svendsen et al., 2013Svendsen et al., , 2015 of fishes. However, the ecological relevance of swimming performance measures obtained from forced swimming experiments has been questioned because the uniform hydraulic conditions in swim tunnels may be rare in nature (Kemp et al., 2011;Maddock et al., 2013). Instead, spontaneously moving fish perform frequent changes in speed and direction (Tudorache et al., 2009;Steinhausen et al., 2010). Importantly, fish swimming behaviour may be influenced by tunnel dimensions (Tudorache et al., 2007), suggesting that fish swimming behaviour could have been influenced by the tunnel designs used in the present study. It is possible that the correlations between MS sus and MS unsus observed here are partly explained by behavioural responses to the swim tunnels and perhaps not relevant in natural settings. Likewise, it is not known whether intraspecific variation in MS unsus is dependent on the protocol used for data collection. It is possible that the chase protocol to estimate MS (Reidy et al., 2000;Clark et al., 2013;Gräns et al., 2014) would have resulted in less variation between individuals, because this methodology might be less affected by behavioural differences between individuals, including behavioural responses to a swim tunnel. The present study used swimming respirometry because this methodology typically provides the highest measures of MO 2max (Roche et al., 2013).

Spontaneous use of metabolic scope
How frequently are fish spontaneously using 100% of their MS? Although the answer to this question is largely unknown, previous studies have indicated that this metabolic level may be engaged rarely (Priede, 1985;Lucas et al., 1993;Murchie et al., 2011;Genz et al., 2013;Marras et al., 2013). For example, Lucas et al. (1993) found that the northern pike (Esox lucius) rarely works at the upper limits of metabolism in the wild. Likewise, Murchie et al. (2011) documented that the Bahamas bonefish (Albula vulpes) typically operates at metabolic rates between 40 and 60% of the MS. It is unclear why fish rarely exploit the full extent of their metabolic capacity; however, it is possible that fish refrain from a level of aerobic metabolism that also incurs anaerobic metabolism. The present study found that on average, anaerobic metabolism was present in ∼25% of the MS. Given that the use of anaerobic metabolism is highly inefficient (Johnston and Moon, 1980;Goolish, 1991b), curtails prolonged locomotor performance (Reidy et al., 2000) and is energetically expensive to recover from (Goolish, 1991b;Lucas et al., 1993;Lee et al., 2003;Svendsen et al., 2010), it is likely that fish minimize the use of the MS that includes anaerobic metabolism (Goolish, 1991a;Lucas et al., 1993). This highlights the importance of accounting for MS unsus when the MS is used to estimate the effects of environmental stressors (e.g. temperature and hypoxia) on fish physiology and performance.

Relevance for the oxygen-and capacitylimited thermal tolerance hypothesis and conservation physiology
In the present study, MS unsus occupied between 0 and 61% of the MS in different individuals, suggesting that sustainable metabolic performance may differ substantially between two individuals, even if they exhibit similar MS. Likewise, Lee et al. (2003) indicated that the onset of burst swimming occurred at between 59 and 62% of the critical swimming speed in two salmon species (Oncorhynchus nerka and Oncorhynchus kisutch), leaving ∼40% of the swimming capacity influenced by anaerobic metabolism. Likewise, Burgetz et al. (1998) measured anaerobic metabolism at 70% of the critical swimming speed based on fuel store depletion and accumulation of anaerobic waste. Although the exact contribution of anaerobic metabolism to the total energy consumption (e.g. ATP synthesis rate) remains unknown, the mean percentage (25%) of the MS influenced by anaerobic metabolism observed in the present study still emphasizes the importance of MS unsus when assessing the sustainable component of the MS.
The MS of individual fish is often translated into capacities for fitness-related performances (e.g. growth and locomotion; Guderley and Pörtner, 2010;Eliason et al., 2011;Khan et al., 2014) and related to habitat use (Claireaux and Lefrançois, 2007;del Raye and Weng, 2015) and migratory patterns (Cooke et al., 2013;Eliason et al., 2013) in a number of fish species. Therefore, MS provides an important tool within the field of conservation and climate change management because it acts as a filter between environmental conditions and impacts on population level Jørgensen et al., 2012;Seebacher and Franklin, 2012). For example, the concept of OCLTT has rapidly gained popularity within climate change research of ectotherms, and it hypothesizes that MS, which is considered an aerobic capacity, can be used to predict how local effects of environmental conditions will affect the physiology of fish (Pörtner and Farrell, 2008;Pörtner and Peck, 2010). However, while OCLTT is applied to estimate the physiological persistence of fish, no study measuring MS in relation to the OCLTT hypothesis has, to our knowledge, accounted for the impact of anaerobic metabolism MacMillan et al., 2012;Seth et al., 2013;Norin et al., 2014;Holt and Jørgensen, 2015), although it may be prevalent within the MS as demonstrated by the present study. Therefore, the assumption that the MS, assessed as the difference between MO 2max and MO 2stand , is purely aerobic and available for sustainable activities (Reidy et al., 2000;Clark et al., 2013;Farrell, 2016) may be questioned. To improve conservation physiology of fishes, studies of MS in relation to environmental stressors and OCLTT may reveal better predictive value if they take into account that a significant part of MS is influenced by anaerobic metabolism and is unavailable for sustainable performances, as suggested by the present study. For instance, the relationship between temperature and MS sus might differ from the relationship between temperature and MS, although this hypothesis is yet to be to be tested. Comparing normoxic and hypoxic treatments, Dutil et al. (2007) found that MS unsus declines in hypoxia whereas MS sus remains unchanged at least down to 50% air saturation, indicating that MS unsus and MS sus are not necessarily affected equally by environmental variation. The possibility remains that temperature variation affects MS sus and MS unsus differently, warranting further study of aerobic and anaerobic metabolism across environmental temperatures in relation to the OCLTT.