Methods matter: considering locomotory mode and respirometry technique when estimating metabolic rates of fishes

We estimated maximum metabolic rates (MMR) in four steady-swimming coral reef fishes - two body-caudal and two median-paired fin swimmers - using three different techniques. Differences between MMR estimates were not due to locomotory mode, but rather the method used to induce maximum performance, with swimming respirometry resulting in the highest estimates.


Introduction
The growing fields of ecological and conservation physiology (Wikelski and Cooke, 2006;Cooke et al., 2013) aim to understand the mechanisms underpinning the behaviour and fitness of organisms in changing environments. As human impacts on global ecosystems continue to increase, greater emphasis will be placed on research that aims to understand whether and how organisms respond and adapt to anthropogenic and environmental stressors (Pörtner and Farrell, 2008;Alcaraz et al., 2013). Choosing appropriate methodologies to study a given species or system is critical to ensure that data are robust and comparisons among studies valid.
The performance of an organism over a range of activities can be tightly linked to oxygen utilization. Respirometry is a tool commonly used to measure oxygen consumption rates ( MO 2 ) and estimate the metabolic performance of an organism at rest, during exposure to stressors or while performing different locomotory activities. Two important physiological parameters describe the upper and lower bounds of an organism's capacity to metabolize energy. Maximal metabolic rate (MMR) is the maximal amount of energy that can be metabolized aerobically by an organism and can be estimated by measuring an organism's MO 2 during or immediately after exhaustive exercise ( MO 2Max ; Norin and Clark, 2016). In contrast, standard metabolic rate (SMR) is the minimal amount of energy required for maintenance and is estimated by measuring MO 2 in a post-absorptive, resting state ( MO 2Min ; Nelson and Chabot, 2011;Chabot et al., 2016). The difference between MO 2Max and MO 2Min is the absolute or total scope for aerobic activity (aerobic scope; AS), which is, in essence, the capacity for aerobic metabolism, in excess of basic maintenance costs, for activities essential to support biological fitness, such as swimming, feeding and reproduction (Brett, 1964). Respirometry can therefore provide essential information about the metabolic performance of an organism and is thus rapidly becoming more widely used Norin and Clark, 2016). A particularly interesting model system is fish, where experimental techniques in respiratory physiology have been widely used for nearly a century and increasingly over recent decades (Webb, 1975;Steffensen et al., 1984Steffensen et al., , 1989Svendsen et al., 2016). Many of these techniques are used extensively today in ecological and conservation physiology studies (Nilsson et al., 2007Donelson et al., 2011;Couturier et al., 2013;McLeod et al., 2013;Rummer et al., 2013Trappett et al., 2013Binning et al., 2014).
The most established method to estimate MMR in fish uses a treadmill-like swimming respirometry chamber (hereinafter, swimming respirometry), where individuals swim against near-laminar water flow at incrementally increased speeds until fatigue is reached ( Fig. 1A; Brett, 1964). Using this protocol, SMR can be estimated indirectly by extrapolating the non-linear MO 2 -swimming speed relationship to zero swimming speed (Steffensen et al., 1984;Steffensen, 2005;Roche et al., 2013).
A second common method uses an exhaustive chase protocol (commonly referred to as the 'stick method'), during which fish are first chased to exhaustion (some protocols also incorporate air exposure) and then placed into a resting respirometer that closely matches the fish's size, such that active swimming is restricted ( Fig. 1B; Ferguson and Tufts, 1992;Donaldson et al., 2010;Clark et al., 2012Clark et al., , 2013Roche et al., 2013). The MO 2Max calculated immediately after this exhaustive exercise (which includes repayment of the oxygen debt resulting from anaerobic activity) is used as an indirect estimate of MMR. After recovery from the chase, SMR can be estimated directly once the fish has fully recovered from exercise (6-24 h depending on species; Milligan and Wood, 1986;Milligan, 1996;Milligan et al., 2000;Chabot et al., 2016).
A third and more recent method uses a circular chamber and a magnetic stir-bar to create a vortex-like water flow, against which a fish must swim facing either clockwise or anticlockwise Gardiner et al., 2010;Couturier et al., 2013;Rummer et al., 2013;Trappett et al., 2013;Pope et al., 2014;hereinafter, circular chamber;Fig. 1C, online supplementary material, Fig. S1, but referred to as a 'swim respirometer' by Nilsson et al., 2007). The circular chamber combines aspects of both swimming and resting respirometers, but instead the fish swims in tight circles. As in a traditional swimming respirometer, this method allows direct estimates of MMR during exertion. In contrast to traditional swimming respirometry, direct estimates of SMR can also be obtained, as in resting chambers, when the revolutions of the stir-bar are reduced to a level that gently mixes the water but does not induce swimming. All three methods have been used interchangeably to estimate metabolic performance in fishes. However, studies suggest that the choice of method could significantly affect the accuracy (Reidy et al., 1995;Roche et al., 2013;Norin and Clark, 2016;Svendsen et al., 2016) and, perhaps, repeatability of MMR and SMR estimates.
Swimming performance is crucial for nearly every aspect of a fish's ecology, including predator-prey interactions, reproductive behaviour and habitat selection (Breder, 1926;Webb, 1975Webb, , 1984Blake, 2004), and engages a variety of body regions and fin appendages (referred to hereinafter as swimming modes; Breder, 1926). Functional differences in swimming modes may have important implications for determining the best method of obtaining accurate metabolic rate estimates. For instance, body-caudal fin (BCF) swimming (e.g. carangiform) is powered by movements of the caudal (tail) fin and the posterior half of the body (Fig. 2) and is often (but not always) coupled with a long, streamlined body ideal for long-distance and fast-propulsive swimming in open environments (Videler, 1993;Blake, 2004). Alternatively, median-paired fin (MPF; e.g. labriform; pectoral fin) swimming is powered by movements of the median or paired fins, such as the pectoral or dorsal-ventral fins, while maintaining a rigid body (Fig. 2), and is thought to promote manoeuvrability (Korsmeyer et al., 2002). These two swimming modes differ kinematically in how thrust is produced and physiologically in terms of which muscles are used during swimming (Blake, 2004). Although BCF swimming is used by many species during high-speed escapes or chases, species from many families of coral reef fishes (e.g. Acanthuridae, Labridae, Pomacentridae and Scaridae) regularly use an MPF swimming mode during daily activities. Currently, no study has evaluated whether functional differences in routine swimming mode affect which respirometry method should be used to obtain the best metabolic rate estimates.
Previous studies that compare methods or calculations for estimating metabolic rates in fishes are restricted to one or two species and/or a single swimming mode (e.g. Reidy et al., 1995;Schurmann and Steffensen, 1997;Korsmeyer et al., 2002;Roche et al., 2013). This limits our ability to generalize results across taxa with different swimming modes, swimming durations and/or life histories. Additionally, time-sensitive and field-based studies would benefit from more portable and faster protocols for estimating MMR and SMR. The chase and circular chamber methods may therefore provide more rapid options for estimating MMR than traditional swimming respirometry, which is expensive and can require upwards of 12 h per individual. Therefore, the aims of this study were as follows: (i) to compare metabolic rate estimates obtained using three common methods, i.e. a prolonged swim trial using a swimming respirometer, a short-duration exhaustive chase protocol with air exposure followed by resting respirometry, and a short-duration exhaustive swimming trial in a circular chamber followed by resting respirometry; and (ii) to determine whether metabolic rate estimates obtained with each method vary among four fish species, all of which are prolonged swimmers but exhibit different swimming modes (BCF and MPF).

Experimental animals
We used four species of coral reef fishes, all of which are abundant on the Great Barrier Reef, Australia, to explore differences in MMR and SMR estimates across three respirometry methods and two swimming modes. We chose two predominantly BCF swimming caesionid (fusilier) species [Pterocaesio marri (sample size n = 5; standard length, SL, 85.2 ± 11.4 mm; wet mass 9.89 ± 1.86 g; means ± SD) and Caesio teres (n = 11; 82.2 ± 6.7 mm; 15.07 ± 4.46 g)] that form mixed-species shoals that cruise mid-water along the reef edge (Fig. 2). We also chose two predominantly MPF swimming pomacentrid (damselfish) species [Acanthochromis polyacanthus (n = 11; 74.2 ± 4.8 mm; 16.88 ± 3.38 g) and Chromis atripectoralis (n = 10; 62.5 ± 2.2 mm; 8.10 ± 1.51 g)] that are site attached as adults and territorial (Randall et al., 1997;Fig. 2). The MPF swimmers we chose generate lift-based thrust by flapping their fins, in contrast to other MPF swimmers that produce drag-based thrust by rowing their fins like paddles, a behaviour better suited for low-speed manoeuvring (Vogel, 1994;Walker and Westneat, 2002;Binning and Roche, 2015;Fig. 2). All four species co-occur on the mid-shelf reef crest and feed primarily on plankton in the water column (Randall et al., 1997). Despite differences in their habitat use, all four species are considered steady/prolonged swimmers.
Fishes were collected from reef crest sites around Lizard Island (14°40′08″S; 145°27′34″E) using monofilament barrier and hand nets under Marine Parks Permit #G10/33239.1. Fishes were maintained in flow-through aquaria directly from the reef at the research station laboratory at ambient temperature (∼27°C) and fed to satiation daily (NRD pellets; INVE Aquaculture, Salt Lake City, UT, USA) until being transferred to the James Cook University Marine Aquarium Research Facilities Unit in Townsville, Queensland, Australia, ∼14 days after collection. At James Cook University, fishes were evenly distributed among five aquaria supplied with well-aerated seawater (28.5 ± 0.5°C) and fed to satiation daily (NRD pellets) for a minimum of 19 days before the experiments commenced. Throughout the duration of the project, fishes were maintained under James Cook University Animal Ethics Committee regulations (permit #A1722, approved for this study) according to the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes and the Queensland Animal Care and Protection Act 2001.
Prior to experimental procedures, individual fish were fasted for 36 h in separate aquaria to ensure a post-absorptive state (Niimi and Beamish, 1974). Fishes were tested in the three different respirometry methods in a random order following a repeated-measures design. All experiments were performed between 22 October and 14 November 2012.

Swimming respirometry
Oxygen consumption rates ( MO 2 ) were measured for solitary individuals swimming in a 4.8 l custom-built Steffensen-type Plexiglas swimming respirometer ( Fig. 1A; Steffensen et al., 1984;Johansen and Jones, 2011;Roche et al., 2013), a system that allows oxygen consumption rates and swimming performance to be measured simultaneously while fish are being exercised.
The working section of the swimming chamber was 7.0 cm × 36.0 cm × 7.0 cm (width × length × depth), but the entire respirometer was immersed in a temperature-controlled bath and maintained at 28.5 ± 0.1°C (mean +/− SD) (Fig. 1A). Flow straighteners were used to create laminar flow within the working section, and flow was calibrated from 0 to 86.6 ± 0.5 cm s −1 (mean ± SD) using a digital vane wheel flow sensor (Höntzsch GmbH, Waiblingen, Germany; model #ZS30GFE-md20T). Solid blocking effects of the fish in the working section were corrected following Bell and Terhune (1970) and did not exceed 5% for any individual.
At the beginning of each trial, the respirometer was filled with temperature-controlled, filtered, ultraviolet-sterilized, well-aerated seawater. Next, a fish was placed in the working section of the chamber and left to habituate for 4-8 h at a swimming speed of 0.5 BL s −1 (body length, taken as the standard length of the fish, per second) until the fish had settled into a continuous swimming rhythm and MO 2 stabilized. Preliminary trials conducted on all four species in the same chamber demonstrated that all reached gait transitions and critical swimming speeds (U crit ) well above 4 BL s −1 . Specifically, gait transitions occur when MPF swimmers transition from using paired fins to caudalassisted swimming (U p-c ; Johansen and Jones, 2011;Binning et al., 2014) or when BCF swimmers transition from steady swimming to anaerobic burst and coast motions (U burst ; Svendsen et al., 2010) as speeds increase. As is common in other fast, steady-swimming coral reef fish species (Johansen and Jones, 2011;Roche et al., 2013), increases in MO 2 are only marginal at swimming speeds <40% of U crit . Therefore, after the initial habituation period at 0.5 BL s −1 , swimming protocols commenced at 4 BL s −1 , which was substantially below U pc , U burst , and U crit for these species. All fish were swum during daylight hours.
After the 0.5 BL s −1 habituation period, the swimming trial commenced by increasing the water flow speed to 4 BL s −1 over a period of 4 min. Then, starting at 4 BL s −1 , MO 2 was measured, and the swimming speed was incrementally increased (increments of 0.5 BL s −1 ) following a standard U crit protocol (Jain et al., 1997). Fish swam at each speed for three 8 min cycles (i.e. 24 min at each speed). Each 8 min cycle consisted of a 5 min measurement period and a 3 min flush period to replenish the chamber with filtered, well-aerated seawater. Fish were continuously monitored throughout the swimming trial, and the trial was considered finished when the fish could no longer swim against the flow, being swept downstream onto a retaining grid for >5 s (U crit ; Roche et al., 2013). At this point, the total swimming time and flow speed were recorded, the experiment was terminated, and the fish was returned to its holding tank. Maximal metabolic rate (MMR Swim ) was estimated from the MO 2 at the maximal swimming speed where fish completed at least one full 8 min cycle.
Critical swimming speed (U crit ) was calculated following Brett (1964): where U is the penultimate swimming speed before the fish fatigued and stopped swimming (U crit ); U i is the swimming speed at which the fish was unable to continue swimming (i.e. swimming speed at increment i); t is the length of time the fish swam at the final swimming speed where fatigue occurred; and t i is the amount of time fish were swum at each speed interval in the trial (24 min).

Exhaustive chase protocol and resting respirometry
Individual fish were placed into a 0.6 m (diameter) round aquarium containing well-aerated and temperature-controlled seawater (0.15 m deep) maintained at the same temperature as the respirometer (28.5 ± 0.1°C). Fish were then chased continuously by hand for 3 min, during which time the experimenter would touch the tail of the fish if it slowed down or stopped swimming. All species swam primarily with their caudal fin when chased, repeatedly bursting away from the stimulus. Fish were deemed exhausted when they became unresponsive to chasing, which always occurred by the end of the 3 min chase period. Fish were then scooped into a rubber mesh net and maintained out of the water for 1 min Roche et al., 2013). After air exposure, fish were immediately placed in a 1615 ml closed-loop, recirculating, resting respirometry chamber submerged in a temperaturecontrolled water bath ( Fig. 1B; reviewed by Clark et al., 2013). The MO 2 measurement period commenced within 10 s of placing the fish in the chamber and continued for a period of 5 min. The MO 2Max was calculated from the steepest 1 min slope during this 5 min interval and used to estimate MMR (MMR Chase ). Standard metabolic rate (SMR Rest ) was estimated from MO 2 values obtained after leaving the fish in the chamber for an additional 6-12 h. This time period was deemed sufficient to ensure that MO 2 stabilized and no longer decreased in each species investigated here (online supplementary material, Fig. S2; also see Roche et al., 2013;Rummer et al., 2013).

Circular chamber respirometry
Individual fish were placed into a 2654 ml (17.2 cm internal height, 14.8 cm internal diameter) o-ring sealed cylinder (Perspex, 0.3 cm thickness) connected via tubing to a submersible pump in a temperature-controlled water bath (28.5 ± 0.1°C; Fig. 1C, online supplementary material, Fig.  S1). A magnetic stir plate below the chamber activated a stir bar (1 cm × 6 cm) to create a circular water motion in the chamber. The speed of the rotating stir bar was increased over a period of ∼1 min to the maximal speed at which the fish could only just maintain its position in the chamber (Fig. 1C, online supplementary material, Fig. S1; see also Nilsson et al., 2007). If the fish could no longer hold position in the chamber, the speed was decreased slightly until it was able to maintain position while swimming. The MO 2Max was calculated from the steepest 1 min slope of the change in O 2 depletion recorded during the first 5-7 min of measuring (between 5 and 10 min used in previous studies; Nilsson et al., 2009;Gardiner et al., 2010;Couturier et al., 2013;Rummer et al., 2013), which corresponds to MMR Circle . The chamber was then flushed, and the rotational speed of the stir bar was decreased to a minimal speed, at which the fish was able to stop swimming and rest on the false mesh bottom (see Fig. 1C). Similar to the chase protocol, SMR was estimated from MO 2 values obtained after leaving the fish in the circular chamber for an additional 6-12 h until MO 2 stabilized and no longer decreased (SMR Circle ; online supplementary material, Fig. S2).

General respirometry information and calculations
We used intermittent-flow respirometry (Steffensen, 1989;Svendsen et al., 2016) for all three methods. A digital relay timer (MFRT-1 Multi Function Recycling Timer; Xiamen SUPERPRO Technology Co., Ltd, Xiamen, Fujian, China) was connected to submersible pumps to repeat an 8 min cycle that began with a 5 min measurement period followed by a 3 min flush period. The measurement period was short enough to ensure that oxygen within the chambers remained above 80% air saturation at all times. This is important to ensure that oxygen consumption rates are not influenced by the adrenergic stress response or other metabolic changes associated with hypoxia (Hughes, 1973;Tetens and Lykkeboe, 1985;Boutilier et al., 1988). The flush period was long enough to ensure that oxygen levels returned to 100% air saturation. During the experiments, test fish were shielded from outside stimuli by a dark cloth to avoid unwanted stress. However, there was a small viewing window in each chamber for the researcher to check on the fish and so that the fish could still experience photoperiod (Fig. 1B). Temperature-and barometric pressurecompensated O 2 concentration (in milligrams per litre) in the water were continuously measured at 0.5 Hz using oxygensensitive REDFLASH dye on contactless spots (2 mm) adhered to the inside of each chamber. Spots were linked to a Firesting Optical Oxygen Meter (Pyro Science e. K., Aachen, Germany) via fibre-optic cables. Each sensor was calibrated using fully aerated seawater (as 100%) prior to each trial and with sodium sulphite (as zero) weekly or as needed.
Text files were imported into LabChart v. 6.1.3 (ADInstruments, Dunedin, New Zealand), and MO 2 (in milligrams of O 2 per kilogram per hour) was calculated as the slope of the linear regression of oxygen concentration decline over time during the measurement period of each cycle using the following equation: modified from (Bushnell et al., 1994;Schurmann and Steffensen, 1997), where S is the slope (in milligrams of O 2 per litre per secondl), V resp is the volume of the respirometer minus the volume of the fish (in litres), and M is the mass of the fish (in kilograms). We subtracted the proportional background O 2 consumption rate (measured as O 2 depletion in the empty respirometer before and after each trial, assumed linear) from each MO 2 measurement. To limit background respiration rates to <5% of a fish's SMR, chambers and pumps were rinsed daily with a 10% bleach solution and fresh water and allowed to dry overnight prior to commencing trials on the next day. The SMR (in milligrams of O 2 per kilogram per hour) was estimated from the average of the lowest 10% of MO 2 values Rummer et al., 2013Rummer et al., , 2014.

Statistical analyses
We used linear mixed-effects models (LMM; 'nlme' package in R) to compare estimates of MMR and SMR obtained with different respirometry methods on the same individuals. This repeated-measures design minimized inter-individual variation in metabolic rate estimates. Species and respirometry method were specified as fixed factors, and fish identity was included as a random factor to control for the non-independence of data points collected using the same individual (Bolker et al., 2009). Post hoc multiple comparisons were done using the R function 'ghlt' in the package 'multcomp'. This model also allowed for individuals to be included even if they did not complete all three methods. We used two general linear models (LM) and Tukey's tests to examine differences in absolute and relative U crit among swimming modes and species. Diagnostic plots were used to ensure that the data met the assumptions of the models. Non-significant interactions were removed for model simplification and fit. Analyses were performed in R v3.0.2 (R Core Team, 2013).

Results
Estimates of MMR differed significantly between respirometry methods (LMM, F 2,62 = 11.38, P < 0.001; Fig. 3A and Table 1) and fish species (LMM, F 3,33 = 4.13, P = 0.014; Fig. 3B and Table 1). However, the effect of method was the same across all species (species × method interaction not significant, F 6,56 = 0.74, P = 0.62; online supplementary material, Fig. S3A). The value of MMR Swim was consistently higher than MMR Chase (z = 2.95, P < 0.01) and MMR Circle (z = 4.79, P < 0.001) for all species (BCF and MPF swimmers; Fig. 3A). Specifically, MMR Swim was 20% higher than MMR Chase based on model predictions computed with the R package effects (Fox et al., 2014); this difference ranged from 6.3 to 35.3% across the four species. Likewise, MMR Swim was 25% higher than MMR Circle ; this difference ranged from 15.5 to 38.3% across the four species. On average, MMR Chase was numerically higher than MMR Circle , but this difference was not statistically significant (z = 1.84, P = 0.15; Fig. 3A).
Body-caudal fin swimming fishes exhibited a significantly higher absolute U crit (in centimetres per second) than MPF swimming fishes (LM, F 1,31 = 16.1, P < 0.001; Table 2), and absolute U crit differed slightly among species within swimming mode (LM, F 2,31 = 1.1, P = 0.03; Table 2). Values of absolute and relative U crit for all species are presented in Table 2.

Discussion
This study demonstrates the importance of choosing the appropriate method when estimating metabolic rates in fishes. Swimming respirometry involves incrementally increasing swimming speeds over several hours until the animal reaches a maximal swimming speed that is unsustainable. This technique has been traditionally referred to as the most accurate means of estimating MMR for steady/prolonged swimmers (Steffensen et al., 1984;Farrell and Steffensen, 1987;Plaut, 2001). The incremental increase in speed allows for the routine gait of a species to be used at low speeds, with transitions to caudal-assisted (U p-c for MPF swimmers; Johansen and Jones, 2011;Binning et al., 2014;Johansen et al., 2015) and/ or anaerobic burst-and-coast swimming (U burst ; Svendsen et al., 2010) as speeds increase, during which time substrate limitations will result in fibre and therefore muscle exhaustion. Additionally, MMR is estimated directly while the fish is maximally exercised. Indeed, we found that swimming respirometry consistently gave the highest estimates of MMR in both BCF and MPF swimmers. The two short-duration protocols, where fish are challenged for 3 min with 1 min air exposure (chase protocol) or 5-7 min (circular chamber protocol) both underestimated MMR by as much as 38% for a single species compared with swimming respirometry (online supplementary material, Table S1). This may be because MPF swimmers would be forced to transition immediately to their final swimming mode (caudal burst swimming), which involves different kinematics from their routine, labriform swimming mode (Webb, 1975(Webb, , 1984Blake, 2004), in order to undergo the challenge. In contrast, for BCF swimmers, the kinematics of movement between routine and burst swimming are very similar. Therefore, in theory, all methods for estimating MMR should have resulted in similar values for BCF swimmers and different values for MPF swimmers, but this prediction did not hold.
For these prolonged swimmers, swimming mode did not influence the best choice of respirometry method used to estimate MMR. However, other aspects may be important when considering respirometry methods. For example, the duration of the challenge and protocol may play a key role (e.g. incremental increases in speed over several hours in the swimming protocol vs. a few minutes in the chase and circular chamber protocols). Even within a long-duration protocol, such as with swimming respirometry, the time between the incremental increases in water velocity can affect how fish respond in terms of oxygen consumption rates (Kiceniuk and Jones, 1977;Farrell, 2007). The size of the animal and flow encountered in their natural habitat should also be taken into consideration (Nelson, et al., 2002). It should be noted that the effect on unsteady swimming species that perform poorly in a swim tunnel (e.g. burst swimmers and ambush predators; Killen et al., 2007) has not been thoroughly tested across species (see examples for cod in Soofiani and Priede, 1985;Reidy et al., 1995).

Locomotory modes and duration of physiological performance in fishes
Swimming is characterized by the structures required (e.g. fins and muscles) and duration. Body-caudal fin swimmers rely on caudal body musculature and a caudal fin to generate thrust, which can support steady movements over prolonged periods of time in some species (Blake, 1983) and unsteady movements over short periods of time in others (e.g. ambush predators; Fig. 2). In contrast, MPF swimmers employ a lift-based (wing) or rowing (oar) paired fin movement, while the body remains rigid to reduce drag and save energy ( Fig. 2; Webb, 1984;Blake, 2004). Median-paired fin swimmers are generally optimized for maintaining position, hovering and manoeuvring in complex environments, especially for those with more rounded fins, and were not traditionally viewed as high-performance endurance swimmers (but, see Fulton et al., 2013;Binning and Roche, 2015). Sustained swimming (via BCF or MPF) is generally considered to last >200 min; in contrast, prolonged swimming (20 s to 200 min) and burst-andcoast swimming (<20 s) both end in fatigue (reviewed by Blake, 2004; but also see Sfakiotakis et al., 1999;Korsmeyer et al., 2002). Sustained or endurance swimming, which can be measured using traditional swimming protocols, is predominantly aerobic, using red muscle and not limited by fuel supply because lipids, proteins and carbohydrates can all be oxidized aerobically (Peake and Farrell, 2004;Fangue et al., 2008). Then, as fish approach maximal swimming speeds, they recruit fast-twitch, glycolytic, white muscle, which generates more thrust (Rome et al., 1996;Korsmeyer et al., 2002;Peake and Farrell, 2004;Svendsen et al., 2010). Consequently, the short-duration protocols (chase or circular chamber) may be exploiting only this final burst-and-coast gait, explaining why MMR can be significantly underestimated.
As species can be optimized for steady or short-term burst performance, the most relevant assessments of metabolic performance during swimming must consider the swimming behaviour of the species in their natural habitat and the duration over which it occurs. For instance, some species cannot or will not swim for extended periods and may therefore not be amenable to a protocol where water velocity is incrementally increased over several hours to assess swimming performance 8 Conservation Physiology • Volume 4 2016 Research article and estimate MMR (Reidy et al., 1995;Claireaux et al., 2006;Jordan and Steffensen, 2007;Killen et al., 2007). We investigated steady/prolonged swimmers from both BCF and MPF swimming modes and found that a longer duration protocol provides the most accurate estimates of MMR for these species that can sustain high swimming speeds for long periods.
Other methods for estimating MMR may provide more cost-and time-effective alternatives to swimming respirometry. In chase protocols, the fish is forced to use burst, anaerobic swimming until fatigue. However, it is important to consider that after an exhaustive chase challenge, MO 2 measurements come from excess post-exercise oxygen consumption (EPOC) or repayment of oxygen debt incurred from anaerobic metabolism. There are at least two assumptions underpinning this protocol. The first assumption is that EPOC represents the highest rate of oxygen consumption and is therefore equal to MO 2Max and thus estimates of MMR. The second assumption is that EPOC occurs immediately after the exercise challenge, and recovery takes long enough for a reliable estimate to be obtained. Neither of these assumptions has been rigorously verified across species (see example for barramundi, Lates calcarifer, by Norin and Clark, 2016). If EPOC occurs immediately after a challenge, but recovery is quick, there is a risk of missing the measurements between the time the fish is challenged and when the measurement begins, as with the chase protocol. Theoretically, chase protocols may also delay EPOC by elevating plasma glucose concentrations and delaying glycogen resynthesis and lactate clearance because of stress (Milligan et al., 2000;Peake and Farrell, 2004). In both of these situations, MMR may be underestimated. Nonetheless, exhaustive chase methods may be ideal for unsteady swimmers, such as ambush predators, and future studies should compare methods in such species and others with more burst-type lifestyles.
Recent studies have used another type of exhaustive chase method with a circular chamber protocol Donelson et al., 2010Donelson et al., , 2011Gardiner et al., 2010;Couturier et al., 2013;Rummer et al., 2013;Pope et al., 2014) as an alternative to swimming respirometry (e.g. Trappett et al., 2013). However, we generally do not recommend the use of a circular chamber, especially if the goals of the study are to compare estimates of AS or MMR with other studies where other methods have been used or to report some measure of swimming speed. This recommendation is based on several lines of reasoning, as follows: (i) velocity across the diameter of the circular respirometer increases significantly towards the edges of the vortex (also mentioned by Nilsson et al., 2007), meaning that the swimming speed of the individual cannot be measured accurately or compared reliably within/among studies or to field conditions; (ii) fish are constantly swimming in either a clockwise or anticlockwise direction, which will result in an imbalanced use of their musculature, may prevent recruitment of all red muscle and therefore achievement of maximal aerobic metabolism, and is likely to cause premature fatigue of right-or left-side muscles, but not complete exhaustion; and (iii) because measurements are made continuously during the exercise period, the flush pump cycle, which periodically provides the fish with clean, well-aerated seawater, cannot be started until after the fish has fatigued, risking chamber O 2 concentrations falling below 80% and initiating a hypoxic stress response (Hughes, 1973;Tetens and Lykkeboe, 1985;Boutilier et al., 1988).
Although the circular chamber protocol cannot be compared directly with swimming respirometry and does not allow manual chasing for species that will not swim for prolonged periods (e.g. serranids and gadids), there may be situations where this protocol can be used effectively. Advantages include the following: (i) fish are not moved between chambers, and the investigator is therefore less likely to 'miss' the MO 2Max values because O 2 is being monitored continuously; and (ii) unlike chase methods, the water velocity can be increased incrementally to allow fish to transition between gaits gradually in order to achieve maximal swimming speeds. As a result, this protocol may work well for some species that cannot be manipulated easily by the researcher or are too delicate for manual chasing protocols (e.g. larval fishes, as in Nilsson et al., 2007). However, researchers should be aware of the difficulties in comparing and interpreting results obtained with a circular chamber protocol and use this technique only if other options have proved unsuccessful or are not feasible.
The ecology of a species may also be important when assessing SMR. Some species that are optimized for fast, sustained swimming (e.g. mackerel and tuna) have difficulty maintaining very low speeds and can exhibit highly variable oxygen consumption rates owing to stress while trying to maintain position (Korsmeyer and Dewar, 2001). Additionally, species using ram ventilation (e.g. tuna and sharks) have to swim continuously to survive; therefore, it may be nearly impossible to estimate SMR accurately using conventional resting methodologies (Clark and Seymour, 2006). Rather, SMR in these species should be estimated indirectly by extrapolation of the non-linear MO 2 -swimming speed relationship using swimming respirometry (Bushnell et al., 1994;Reidy et al., 2000;Korsmeyer and Dewar, 2001;Korsmeyer et al., 2002;Clark and Seymour, 2006;Roche et al., 2013;Binning et al., 2014). In other fishes, however, SMR can be estimated directly in a non-swimming, relaxed state using a resting respirometer (reviewed by Clark et al., 2013 andSvendsen et al., 2016; but also see Steffensen, 1989;Roche et al., 2013). Indeed, the two resting protocols trialled (post-chase resting chamber and circular chamber) provided consistent estimates of SMR in the present study.

Conclusions and recommendations
In recent decades, there has been a focus on studies linking swimming mode, physiological performance and environmental stress, emphasizing the importance of choosing appropriate methods for addressing a specific research question. The present study highlights that swimming respirometry appears to provide the most accurate estimates of MMR in fish that are steady/prolonged swimmers, regardless of swimming mode. In addition, swimming respirometry provides additional valuable information about swimming performance beyond critical swimming speed and MMR, such as gait transitions, burst speed, optimal swimming speed and cost of transport, all of which are ecologically relevant and could be influenced by changing environmental conditions. If swimming respirometry is not possible for fishes that are good steady/prolonged swimmers, short-duration chase or circular chamber protocols could be used, but with caution, because they may significantly underestimate MMR and therefore AS (see also Roche et al., 2013). Fish that predominantly use burst swimming (e.g. ambush predators) or unsteady MPF swimming (e.g. labriforms that use 'rowing' instead of 'flapping'; see Binning and Roche, 2015) may do better in short-duration challenges, but these predictions remain to be investigated. Nevertheless, the two short-duration alternative methods investigated here did provide direct measurements of MO 2Min and therefore provided equally reliable SMR estimates.
From a technical perspective, both swimming respirometry and the chase protocol can easily be standardized and automated (e.g. using commercially available software) to maintain oxygen concentrations above 80% saturation at all times while calculating oxygen consumption rates at regular intervals (Svendsen et al., 2016). These tools can ease user implementation and could minimize the subjectivity or bias in data collection (i.e. manually selected oxygen consumption rate slopes, duration of slope to use, r 2 of slopes, etc.). The circular chamber, however, cannot be automated as easily because the point at which maximal swimming performance and fatigue will occur cannot be anticipated easily. Some methods may also be financially less costly (e.g. chase protocol), and with shorter protocols, more fish can be tested in a limited amount of time. However, if the method chosen is not a good match for the species, then the quality of measurements 10 Conservation Physiology • Volume 4 2016 Research article will be sacrificed for quantity. In particular, this may pose a significant problem if the goal of the study is to estimate AS, because our data suggest that using chase or circular chamber protocols for prolonged swimmers can underestimate MMR (online supplementary material, Table S1) and therefore AS. Additionally, because general correction factors cannot be applied to results post hoc owing to inter-individual, speciesspecific and/or temperature-driven differences in the degree of underestimation, caution is warranted if MMR and aerobic scope estimates are compared across studies that use different methods. Indeed, this has been a topic that has recently received a lot of attention and discussion, given that these measurements are being extensively used to understand organismal responses to contemporary issues, such as climate change (see