Finding the best estimates of metabolic rates in a coral reef fish

Ecophysiology is the study of how organisms respond physiologically to environmental stressors (Fry, 1947; Fry, 1971; Schurmann and Steffensen, 1997; Claireaux and Lefrançois, 2007). Given the prevalence of anthropogenic stressors in natural systems, conservation physiology is rapidly growing as a discipline that aims to better understand and predict organisms' responses to these environmental changes (Wikelski and Cooke, 2006; Kieffer, 2010; Cooke et al., 2012). Respirometry, in particular, is increasingly used by ecophysiologists as advances in technology and equipment accessibility are facilitating studies (Kieffer, 2010), especially in less studied groups such as tropical fishes (e.g. Donelson et al., 2011; Munday et al., 2012). However, as conservation physiology and respirometry continue to grow in popularity, standardized methods must be used to ensure that physiological data are robust and comparisons among studies are valid.

In aquatic respiratory physiology, two types of respirometry chambers are commonly used to conduct either swimming (Fry and Hart, 1948; Blazka et al., 1960; Brett, 1964; Steffensen et al., 1984) or resting respirometry (Teal and Carey, 1967; Hemmingsen and Douglas, 1970; Fry, 1971). Resting respirometry is sometimes also referred to as static respirometry (e.g. Reidy et al., 2000; Brick and Cech, 2002; Barnes et al., 2011), but this terminology is much less common. Despite differences in their complexity and ease of use, both methods allow measuring oxygen consumption rates () to estimate metabolic rates during or following varying levels of activity (e.g. resting versus active swimming) (Reidy et al., 1995; Peake and Farrell, 2006; Killen et al., 2007). Different calculations can also be used within each method to compute the same estimates of metabolic rate. This probably introduces variation in metabolic rate estimates, but studies have yet to carefully examine whether data obtained in different ways produce comparable results (but see Reidy et al., 1995; Reidy et al., 2000).

Two key physiological parameters characterize the upper and lower bounds of a fish's capacity to uptake oxygen: standard (resting) metabolic rate (SMR or ) and maximum metabolic rate (MMR or ). SMR corresponds to the minimum maintenance metabolism of a resting fish in a post-absorptive state (Fry, 1971; Brett and Groves, 1979; Schurmann and Steffensen, 1997), whereas MMR corresponds to a fish's maximum rate of oxygen consumption (Fry, 1971; Beamish, 1978; Schurmann and Steffensen, 1997; Korsmeyer and Dewar, 2001; Clark et al., 2011). During exercise, MMR is measured at a fish's maximum swimming speed during prolonged swimming (Bushnell et al., 1994; Schurmann and Steffensen, 1997; Korsmeyer and Dewar, 2001), which requires anaerobic metabolism and typically ends in fatigue within 200 min (Beamish, 1978; Peake and Farrell, 2004). In contrast, active metabolic rate (AMR) is a term describing the oxygen consumption rate of fish at their maximum sustained swimming speed (U_max). Unlike prolonged swimming, sustained swimming can be maintained for >200 min and is powered solely by aerobic metabolism (Beamish, 1978; Peake and Farrell, 2004). Beyond U_max, fish generally engage in burst-and-coast swimming and begins to asymptote (Sepulveda and Dickson, 2000; Claireaux et al., 2006). As a result, MMR often slightly exceeds AMR because fish are forced to swim beyond their maximum sustained swimming speed for a limited time (Bushnell et al., 1994; Schurmann and Steffensen, 1997). Once measured, SMR and MMR can be used to calculate a fish's aerobic scope for activity, which determines the range of metabolic energy available for aerobic activities (Fry, 1947; Bushnell et al., 1994; Cutts et al., 2002; Claireaux and Lefrançois, 2007; Clark et al., 2011). SMR and MMR exclude metabolic activities powered anaerobically because anaerobic metabolism cannot be measured directly through oxygen consumption at the time of exercise (Korsmeyer and Dewar, 2001).

In swimming respirometry, the most common means of estimating a fish's metabolic rate is using a critical swimming speed (U_crit) protocol such as the one initially developed by Brett (Brett, 1964) (Reidy et al., 1995; Plaut, 2001; Farrell, 2007). Fish are made to swim against a laminar water flow in a swimming respirometer while water velocity is increased incrementally, at regular intervals, until the fish fatigues. The U_crit is the swimming speed at which fish become exhausted and stop swimming. Because oxygen consumption is measured continuously while fish are exercised to exhaustion, swimming respirometry is thought to provide a very accurate estimate of MMR (Farrell and Steffensen, 1987; Plaut, 2001; Shultz et al., 2011). In contrast, SMR is not directly measured using this method, but can be calculated by extrapolating the non-linear swimming speed– relationship to a swimming speed of zero (Bushnell et al., 1994; Reidy et al., 2000; Korsmeyer and Dewar, 2001; Korsmeyer et al., 2002; Binning et al., 2013). Despite many advantages of this method for measuring MMR, U_crit protocols can be time consuming and species that are poor swimmers (e.g. ambush predators) often lack the motivation to swim in a respirometer (Peake and Farrell, 2006).

To circumvent the limitations of swimming respirometers, exhaustive chase protocols have been developed to estimate MMR whereby fish are manually chased to exhaustion (Black, 1958; Milligan, 1996; Kieffer, 2000) and immediately placed into a resting respirometer (Cutts et al., 2002; Jordan and Steffensen, 2007; Norin and Malte, 2011). Variations of this method also exist in which fish are briefly held out of the water after chasing (Ferguson and Tufts, 1992; Donaldson et al., 2010; Clark et al., 2012). Air exposure contributes to increasing metabolic demands and has been argued to simulate exercise stress associated with catch-and-release fisheries, where fish are temporarily held out of the water to allow hook removal (Donaldson et al., 2010; Clark et al., 2012). Because the volume of resting respirometers is generally small relative to the size of the fish, individuals tend to remain immobile in the chamber and MMR is measured during recovery from exercise/chasing (Steffensen, 2005). This method relies on post-exercise oxygen consumption rates and MMR therefore corresponds to the sum of the fish's routine metabolic rate (RMR; during activities that elevate SMR) (Schurmann and Steffensen, 1997; Steffensen, 2005) and excess post-exercise oxygen consumption (EPOC) to repay the oxygen debt incurred from anaerobic metabolism during chasing (Killen et al., 2007). One major advantage of using resting respirometry is that SMR can be measured while fish have remained inactive in the chamber for several hours (typically between 2 and 24 h, depending on the species), thus allowing both SMR and MMR to be calculated in one trial (Cutts et al., 2001; Brick and Cech, 2002; Cutts et al., 2002; Nilsson and Ostlund-Nilsson, 2004; Nilsson et al., 2009; Nilsson et al., 2010; Donelson et al., 2011; Norin and Malte, 2011; Clark et al., 2012).

Some studies suggest or anecdotally report that similar MMR measurements can be obtained using both exhaustive chase protocols and U_crit protocols (e.g. Killen et al., 2007; Gingerich et al., 2010). However, a comprehensive comparison of key metabolic parameters measured with different respirometry methods has yet to be conducted. Here, we compare data obtained using three common methods of measuring SMR and MMR in fishes: (1) a traditional U_crit protocol, (2) an exhaustive chase protocol by manual chasing and (3) an exhaustive chase protocol by manual chasing followed by brief (1 min) air exposure. measurements for protocol 1 were carried out in a swimming respirometer whereas measurements for protocols 2 and 3 were performed in resting respirometers.

We chose the coral reef fish Scolopsis bilineata (Bloch 1793) (Nemipteridae) for this study due to its high abundance on the Great Barrier Reef (Boaden and Kingsford, 2012; Roche et al., 2013), adequate size relative to the respirometry equipment used, and amenable behaviour in the swimming respirometer (Binning et al., 2013). Adult fish were collected by divers using barrier and hand nets between February and March 2012 from reefs surrounding Lizard Island, on the northern Great Barrier Reef, Australia (14°40′S, 145°28′E). Fish were transported live in buckets to the aquarium facilities at the Lizard Island Research Station within 2 h of capture and held in individual aquaria (40.0×29.0×18.0 cm, width × length × height) with a flow-through water system directly from the reef. Fish were fed once daily with pieces of raw prawn (mean wet mass ~1 g) and maintained in aquaria for a minimum of 3 days before the respirometry trials. Length measurements for individual fish were obtained by holding each fish in a plastic bag half-filled with water and measuring total length (TL), body width and body depth with handheld callipers. Body mass (M) was measured directly on a balance. Fish were fasted for 24 h prior to the experimental trials (Johansen and Jones, 2011; Shultz et al., 2011) to evacuate the digestive tract and standardize a post-absorptive state that maximizes energy availability for swimming (Niimi and Beamish, 1974). Ten fish (TL=17.6±0.4 cm, M=97.0±7.7 g; means ± s.d.) were subjected to each of three protocols in a random order: a U_crit trial (Brett, 1964; Beamish, 1978; Johansen and Jones, 2011), a 15 min exhaustive chase trial (see Cutts et al., 2002; Killen et al., 2007; Fu et al., 2009; Norin and Malte, 2011; Shultz et al., 2011) and a 3 min exhaustive chase followed by a 1 min air exposure trial (Ferguson and Tufts, 1992; Donaldson et al., 2010; Clark et al., 2012). The same fish (N=10) were subjected to each protocol following a repeated-measures design (Reidy et al., 1995) to minimize inter-individual variation in metabolic rates. Fish were fed and allowed a minimum of 48 h to recover between trials. Prior to trials, individuals were starved for at least 24 h, but never more than 36 h. In all three protocols, was measured using intermittent-flow respirometry (Steffensen et al., 1984; Steffensen, 1989).

We used a linear mixed effects model (LMM; lme function in R) to compare values of maximum metabolic rate (MMR_swim, MMR_chase, MMR_air). We used a second LMM to compare SMR estimates obtained in resting respirometry (SMR_{rest_low}, SMR_{rest_last}, SMR_{rest_hist}) with those obtained in swimming respirometry (SMR_{swim_low}, SMR_{swim_last}, SMR_{swim_hist}) using three different functional forms to describe the relationship between swimming speed and (i.e. a two-parameter exponential function, a three-parameter exponential function and a three-parameter power function). LMMs can be used to reduce inter-individual variation in metabolic rates and control for the non-independence of data points obtained on the same individuals (Bolker et al., 2009). Diagnostic plots and Shapiro–Wilk's test were used to ensure that the data met the assumptions of the models. We compared the fit of non-linear relationships by computing the proportion of variance explained. All analyses were performed in R v2.11.1 (R Development Core Team, 2010).

The mean (±s.e.m.) critical swimming speed for all fish was 3.76±0.10 BL s⁻¹, whereas the mean maximum swimming speed at which fish completed at least one 10 min determination was 3.85±0.10 BL s⁻¹. At 4.25 BL s⁻¹, only three out of 10 fish completed one determination. MMR differed according to the respirometry method employed (LMM; F_2,18=19.2, P<0.001; Fig. 1): MMR_swim was 36% higher than MMR_chase (estimate=−167.69, 95%CI=−221.35 to −114.03, t=−6.13, P<0.001) and 23% higher than MMR_air (estimate=−107.15, 95% CI=−160.81 to −53.49, t=−3.91, P=0.001), and MMR_air was significantly higher than MMR_chase (estimate=60.54, 95% CI=6.88 to 114.20, t=2.21, P=0.04; Fig. 2B, Fig. 3A).

SMR estimates obtained in resting respirometry differed based on the calculation method used: SMR_{rest_low} was significantly lower than SMR_{rest_last} (estimate=23.75, 95% CI=6.71 to 40.78, t=2.73, P<0.01) but not different from SMR_{rest_hist} (estimate=7.84, 95% CI=−9.19 to 24.87, t=0.90, P>0.3); there was no significant difference between SMR_{rest_last} and SMR_{rest_hist} (estimate=−15.91, 95% CI=−32.94 to 1.12, t=−1.83, P=0.07; Fig. 2). In contrast, when calculated for each of the three functional forms, SMR estimates obtained in swimming respirometry did not differ significantly, irrespective of the calculation method employed (LMM; all P>0.05; Fig. 2B).

Fitting a two-parameter exponential function produced SMR estimates 25% lower, on average, than the lowest SMR estimate obtained in resting respirometry (LMM; contrast group=SMR_{rest_low}; SMR_{swim_low} estimate=−27.49, 95%CI=−44.59 to −10.39, t=−3.15, P=0.004; SMR_{swim_last} estimate=−24.29, 95% CI=−41.40 to −7.19, t=−2.78, P=0.01; SMR_{swim_hist} estimate=−22.37, 95% CI=−39.47 to −5.27, t=−2.56, P=0.016). Alternatively, SMR_{rest_low} did not differ from SMR_{swim_low} when we fit a three-parameter exponential function or a three-parameter power function to the swimming speed– relationship (LMM; all P>0.05; Table 1, Fig. 2B). When using either a three-parameter exponential or power function, most differences between SMR obtained in resting versus swimming respirometry occurred between SMR_{rest_last} and SMR_{swim_last} (Table 1, Fig. 2B); there were very few significant differences between SMR_{rest_hist} and SMR_{swim_hist} (Table 1, Fig. 2B). Including measurements at speeds that induced bursting and coasting (U=3.75 and 4.25 BL s⁻¹) into the swimming speed– relationship did not change these results qualitatively.

We found notable differences in MMR, a key metabolic rate parameter, measured using different respirometry methods (Fig. 1). Previous studies have suggested that resting and swimming respirometry produce similar MMR estimates (Gingerich et al., 2010), with some support from data on the lumpfish Cyclopterus lumpus (Killen et al., 2007). Although there was overlap in SMR estimates obtained with different methods, MMR estimated using a U_crit protocol was significantly higher than MMR obtained using two different exhaustive chase protocols combined with resting respirometry.

Using swimming respirometry, SMR is indirectly measured by extrapolating the swimming speed– relationship to U=0 BL s⁻¹ (Brett, 1964; Bushnell et al., 1994; Schurmann and Steffensen, 1997). Following this approach, we used three common calculations to estimate SMR, by averaging (1) the lowest three measurements at U=0.75 BL s⁻¹ (SMR_{swim_low}), (2) the last three measurements at U=0.75 BL s⁻¹ (SMR_{swim_last}) or (3) all values in the lowest mode of an frequency distribution at U=0.75 BL s⁻¹ (SMR_{swim_hist}). Despite different calculations, the three SMR estimates did not significantly differ from each other (Fig. 2) and considerably overlapped SMR_{rest_low} and SMR_{rest_hist} estimates (Table 1) when extrapolated based on a three-parameter exponential or power function (Fig. 3B,C). SMR_{rest_last} differed from the two other SMR estimates in resting respirometry because spontaneous activity elevated values in the early morning, towards the end of the trials. This was not the case in swimming respirometry as U_crit trials began shortly before sunrise.

In a study on the Atlantic cod, Gadus morhua, Schurmann and Steffensen (Schurmann and Steffensen, 1997) found similar results when comparing SMR estimated using both swimming and resting respirometry. Our findings also suggest that SMR can accurately be estimated by extrapolating the swimming speed– relationship obtained from U_crit protocols. Importantly, however, when we fit a simpler, two-parameter exponential function to this data, SMR values estimated with the U_crit protocol were ~25% lower than SMR_{rest_low}, irrespective of the calculation employed (Table 1, Fig. 3C). This finding is in stark contrast with those of Korsmeyer et al. (Korsmeyer et al., 2002), who recommend using the traditional two-parameter exponential function. While this simpler function requires deriving only two constants (Korsmeyer et al., 2002), it may not be the most reliable functional form to extrapolate beyond the range of swimming speed values measured. In contrast, the hydrodynamics-based power function is believed to overestimate SMR because it places more weight on higher swimming speed values (Videler and Nolet, 1990; Korsmeyer et al., 2002). Our SMR estimates from the hydrodynamics-based power function were higher than estimates from the three-parameter exponential function, but this difference was not significant (Fig. 2).

We obtained higher MMR estimates in the swimming respirometry protocol compared with either exhaustive chase protocol. Several factors may explain this difference. Chasing by an experimenter may not have induced complete exhaustion in S. bilineata even if fish became unresponsive towards the end of the chase. However, the duration of our 15 min protocol greatly exceeded that of typical 1–5 min chases in the published literature (Cutts et al., 2002; Fu et al., 2009; Gingerich et al., 2010; Norin and Malte, 2011; Shultz et al., 2011; Clark et al., 2012). Scolopsis bilineata uses its pectoral and caudal fins for swimming and has intermediate to high sustained swimming abilities (Fulton, 2007; Binning et al., 2013). In contrast, many of the fishes that have been subjected to exhaustive chase protocols thus far are body-caudal fin swimmers with high unsteady (burst) swimming performance, such as trout (Ferguson and Tufts, 1992; Norin and Malte, 2011), Pacific salmon (Donaldson et al., 2010; Clark et al., 2012), bonefish (Shultz et al., 2011) and bass (Gingerich et al., 2010). Studies suggest that 1–2 min chases are sufficient to achieve complete fatigue in these species (Gingerich et al., 2010; Norin and Malte, 2011; Clark et al., 2012). Rapid exhaustion most likely occurs because manual chasing induces repetitive burst swimming (Clark et al., 2012), which is powered by white muscle fibres and anaerobic metabolism (Milligan, 1996; Kieffer, 2000). The use of fast, glycolytic muscles for escape swimming explains why these fishes fatigue rapidly and incur large oxygen debts, which can be measured as post-exercise metabolism in the resting respirometry chamber. Alternatively, some body-caudal fin swimmers, such as Atlantic salmon (Cutts et al., 2002), carp and catfish (Fu et al., 2009), can sustain unsteady swimming for longer periods and require chases up to 5 min to reach full exhaustion. Pectoral (e.g. Labridae, Scaridae, Pomacentridae, Cichlidae, Embiotocidae) and pectoral-caudal (e.g. Chaetodontidae, Nemipteridae) swimmers may require even longer exhaustive chases, however, as we observed in the case of S. bilineata. Fish that use their median-paired fins for swimming may burst less frequently during chases (e.g. Gotanda et al., 2009) and utilize both red (aerobic) and white (anaerobic) muscle fibres to power their escape. As such, increased use of red muscle could lead to lower oxygen debts and reduced EPOC required to clear metabolites resulting from anaerobic activity. Because the magnitude of EPOC directly influences measurements of MMR in resting respirometry (Reidy et al., 1995), fish that escape using a combination of white and red muscles will likely display lower values than fish relying predominantly on white muscle and anaerobic metabolic pathways. Although swimming respirometry appears to be a better method for measuring MMR in fish that are good steady swimmers, the opposite may be true of fish with better unsteady swimming performance (see Peake and Farrell, 2006). For example, in a study of post-exercise metabolic rates in Atlantic cod, Reidy et al. (Reidy et al., 1995) found that MMR during recovery after exhaustive chasing significantly exceeded measurements at U_crit, which is contrary to our results.

Lastly, the short 3 min exhaustive chase and 1 min air exposure protocol yielded higher estimates of MMR than the prolonged 15 min chase. Brief periods of air exposure have been used in a number of studies to simulate fisheries encounters (Ferguson and Tufts, 1992; Donaldson et al., 2010; Clark et al., 2012), and likely push fish beyond their anaerobic threshold, leading to increased EPOC. Given the considerable variability in the duration of fish responses to chase protocols and the likelihood of additional variation from using different chasing techniques and intensity (e.g. tail pinching versus manual or stick chasing), air exposure may provide a very effective method of standardizing exhaustive chase protocols and improving the accuracy of MMR estimates across species.

Our experiment demonstrated that, for S. bilineata, the U_crit swimming protocol provided a more accurate estimate of MMR than chase protocols combined with resting respirometry. However, because swimming respirometry is impractical for some species (Reidy et al., 1995; Jordan and Steffensen, 2007), chasing followed by air exposure likely provides the best alternative. Furthermore, we found that SMR can accurately be estimated from data obtained using swimming respirometry. However, extrapolating the oxygen consumption curve depends on the functional form used to describe the swimming speed– relationship. As such, resting respirometers provide a reliable measure of SMR with which to compare estimates from U_crit protocols and should be used whenever possible. Additional studies are required to test how data produced with various respirometry methods compare across fish species with different life histories (e.g. demersal versus pelagic, predatory versus herbivorous) and swimming behaviours (e.g. pectoral, pectoral-caudal and caudal swimming). However, caution is warranted when comparing results obtained with different approaches, particularly in the case of MMR, unless cross-validation has been performed on a species-specific basis. Bearing in mind these results and the limitations of the methods used, researchers should carefully choose the apparatus and method most appropriate for their species and specific research questions before conducting respirometry studies.

We thank C. Layton, C. Juan, L. Strong and staff at the Lizard Island Research Station for field support. Tim Clark and two anonymous reviewers provided useful comments on a previous version of the manuscript.