Introduction

Exercise training has been shown to increase the growth rate of many teleost fishes (see Davison 1997 for review). The majority of studies examining the effects of exercise on fish have generally used salmonids (Davison and Goldspink 1977; Grünbaum et al. 2008; Jørgensen and Jobling 1994) but other teleosts also show positive rates of growth at some level of exercise (Hammer 1994; Merino et al. 2007; Yogata and Oku 2000). Despite the occasional report of no growth response in some species (Alcaraz and Urrutia 2008; Bjørnevik et al. 2003; Hoffnagle et al. 2006), it is now widely accepted that exercise-training can improve the growth of many species (Davison 1997), particularly those with an active pelagic habit (Yogata and Oku 2000). Studies on salmonids generally show a 20–40% increase in growth when swimming on a sustained basis under exercise regimes of 0.8–1.5 BL s−1 (Christiansen et al. 1989; Jørgensen and Jobling 1993; Totland et al. 1987). For the active pelagic Japanese yellowtail (Seriola quinqueradiata), increased growth appears to be of a similar magnitude but occurs after swimming at faster speeds (i.e. 1.5–2 BL s−1) (Yogata and Oku 2000). The current literature would therefore have us believe that exercise can yield impressive gains in the productivity of certain farmed fish species. However, no study has yet examined whether consistent growth improvements can be gained at set levels of optimal swimming under different environmental conditions (e.g. temperature). Exercise may become a useful and standard aquaculture practice but it is first necessary to understand the parameters and conditions that govern growth optimisation.

Optimal sustained swimming could improve aquaculture production efficiency (Jobling et al. 1993) but how fish such as salmon and yellowtail kingfish balance the simultaneous costs of growth and exercise is currently unknown. The anabolic and catabolic costs of growth and exercise are both substantial (Clarke and Seymour 2006; Houlihan et al. 1988; McMillan and Houlihan 1988), and exercise should be reduced if specific dynamic action (SDA) and growth are prioritised (Owen 2001). However, some fish can swim and grow at the same time, suggesting either that multiple metabolic costs are balanced with ease or that some form of behavioural or physiological efficiency is derived from long-term exercise. Christiansen et al. (1991) and Adams et al. (1995) proposed that salmonids undergoing enforced exercise in currents did not engage in costly agonistic encounters and that the associated metabolic savings allowed fish to grow faster whilst swimming. However, exercise training might also lead to further metabolic savings (over and above any change in agonistic behaviour) through greater swimming efficiency. Long-term exercise does after all improve swimming efficiency in some fish species by reducing the metabolic costs of aerobic swimming in a flume respirometer (Bagatto et al. 2001; Nahhas 1981). This is compelling but there is currently no solid evidence linking swimming efficiency with enhanced rates of growth, meaning that the mechanism of exercise-induced growth is still not yet fully understood.

The primary focus of this study is to ascertain the potential benefits of exercise on the growth of the New Zealand yellowtail kingfish, Seriola lalandi, at different temperatures, with a secondary aim of probing potential sources of physiological efficiency which may feasibly underpin exercise-induced growth. S. lalandi is a circum-globally distributed carangid found in temperate and subtropical disjunct populations (Carpenter 1992). It has been adopted within the Japanese yellowtail culture industry (Nakada 2002), it is farmed in sea cages in South Australia (Love and Langenkamp 2003), and it is an excellent candidate species for New Zealand finfish culture (Moran et al. 2008). Therefore, both commercial and scientific interests require us to understand how exercise affects the productivity (i.e. growth and FCR) of this species.

Materials and methods

Two growth trials exposing kingfish to various levels of exercise in currents were carried out at 14.9 and 21.1°C using ambient seawater. The swimming metabolism of kingfish was also examined at 22°C following a long-term phase of exercise training. All fish were sourced from the National Institute of Water and Atmospheric Research (NIWA), Bream Bay Aquaculture Park, Ruakaka, New Zealand. All experiments were carried out with approval from the University of Auckland Animal Ethics Committee (AEC license number R666).

Effects of long-term exercise on kingfish growth at 21.1°C

An exercise growth trial incorporating four levels of exercise in currents (0, 0.75, 1.5 and 2.25 BL s−1) was carried out at the NIWA Bream Bay Aquaculture Park over a 6-week period in 13 m3 tanks (4 m diameter, 1 m depth). Tanks were housed in a purpose-built shed under ambient light (13L:11D) with a supply of fresh aerated seawater at 21.1 ± 0.03°C (mean ± SEM). Tanks were fitted with false mesh walls creating a one-metre-wide swimming channel around the perimeter of the tank. This was designed to limit the range of water velocities that fish would typically experience in large circular tanks. Different water velocities between tanks were achieved via combinations of directional inflows, submersible pumps and external pumps plumbed over the side of the tank. Water velocities were measured using a Höntzsch HFA hand held flow metre with a vane wheel sensor positioned 0.5 m from the edge of the tank. Three hundred and sixteen juvenile yellowtail kingfish with passive integrated transponder (PIT) tags were anaesthetised (AQUI-S® at 0.01 mL L−1 prior to 0.2 mL L−1 2-phenoxy-ethanol in a separate tank), and their weight and fork length measured prior to random allocation to one of the four tanks (n = 79 per tank). Fish were fed twice daily to satiation over the course of the trial using commercial 11-mm floating pellets (Nova, Skretting, Australia) designed for S. lalandi culture. Fish at the start of the trial were 1,591 ± 6.6 g in mass and had a fork length of 476 ± 0.8 mm.

Effects of long-term exercise on kingfish growth at 14.9°C

An exercise growth trial incorporating two levels of exercise in currents (0 and 0.75 BL s−1) was carried out over a 7-week period at the Leigh Marine Laboratory, University of Auckland, in 1 m3 tanks (1.8 m diameter, 0.8 m depth). Tanks were housed outside (10L : 14D) and received fresh aerated seawater at 14.9 ± 0.1°C (mean ± SEM). One hundred and twenty juvenile yellowtail kingfish had PIT tags implanted in the peritoneal cavity under anaesthetic (AQUI-S® at 0.01 mL L−1 prior to 0.2 mL L−1 2-phenoxy-ethanol in a separate tank) and their weight and fork length (FL) recorded. Individuals were assigned randomly to exercise and non-exercise tanks with equal numbers of fish (n = 60) in each tank. Fish in the exercised treatment were exposed to a water velocity of 0.75 BL s−1 0.5 m from the edge of the tank and thus defined as a 0.75 BL s−1 regime group. The non-exercised tank had negligible directional flow (<0.05 BL s−1). Water velocities were measured using the Höntzsch flow metre with vane wheel sensor positioned 0.5 m from the edge of the tank. Fish were fed to satiation daily over the course of the trial using commercial 5-mm floating pellets (Nova, Skretting, Australia). Fish at the start of the trial were 179 ± 4.6 g in mass and had a fork length of 236.6 ± 2 mm (mean ± SEM).

Growth, FCR and condition estimates

At the end of each trial, the weight and length of each tagged fish were once again measured under anaesthetic. Weight-specific growth rate (SGR W ) was calculated from the tagged fish according to the following:

$$ SGR_{W} = \frac{{100 \times \left( {\ln W_{2} - \ln W_{1} } \right)}}{T} $$

where W 1 = weight of fish (g) at the start of the experiment, W 2 = weight of fish (g) at the end of the experiment, and T = duration of the experiment (days). Length growth rate (GR L ) was calculated according to the following:

$$ GR_{L} = \frac{{\left( {L_{2} - L_{1} } \right)}}{T} $$

where L 1 = weight of fish (g) at the start of the experiment, L 2 = weight of fish (g) at the end of the experiment, and T = duration of the experiment (days). The condition factor (CF) of each fish was calculated according to:

$$ CF = \frac{{100 \times {\text{Weight}}}}{{\left( {{\text{length}}\,{\text{in}}\,{\text{cm}}} \right)^{3} }} $$

CF was calculated at the start and end of the trial, and a change in CF value (ΔCF) was derived to gauge the change in body shape over the trial. Positive ΔCF values indicate the fish grew relatively greater in weight compared with length (i.e. they bulk out). A negative ΔCF value indicates fish grew relatively greater in length compared with weight and the fish appear more slender. Feed conversion rate (FCR) was estimated according to the following:

$$ FCR = \frac{{{\text{Total}}\,{\text{food}}\,{\text{delivered}}\,\left( {\text{kg}} \right)}}{{{\text{Total}}\,{\text{gain}}\,{\text{in}}\,{\text{weight}}\,\left( {\text{kg}} \right)}} $$

Effects of exercise training on swimming metabolism at 22°C

Kingfish were held in two 1 m3 tanks at the Leigh Marine Laboratory for at least 30 weeks and subjected to a long-term swimming regime in currents (as above) at 0 and 0.75 BL s−1. Swim flume respirometry was carried out on 20 kingfish (9 trained and 11 untrained) at a time when water temperature in the tanks was averaging 21.1 ± 0.11°C and fish were 699 ± 31 g and 362 ± 15 mm in terms of weight and fork length, respectively. There was no difference in the size of fish selected for respirometry between the trained and non-trained groups; fish weight and length data are therefore pooled. A 38.4-L Brett-style respirometer, equipped with a 12-L swimming section, was employed to measure the metabolic rate of fish at various speeds. The mass specific rate of oxygen consumption (MO2—a useful proxy of true metabolic expenditure) was measured by sealing the fish in the respirometer and recording the decline in O2 at various swimming speeds (see later for more detail). MO2 was sampled once every 10 min in high-quality water with O2 saturations maintained above 70%. This was achieved by sealing the respirometer over a 4-min “measurement” phase and following this with a 5-min flush and a 1-min “wait” period before the next sealed measurement phase began (intermittent flow respirometry) (Steffensen 1989). All test fish were also denied food for approx 47 h prior to any experimentation to eliminate any effect of specific dynamic action (see Secor 2009 for review). The 38.4-L respirometer was immersed in an even larger water bath (122 L) that held a connection with an additional 130-L reservoir of seawater. The combined volume of seawater was maintained at 100% oxygen saturation with an aerator, and water temperature was held at 22 ± 0.5°C using an in-line aquarium chiller (HC-1000A, Hailea, Guandong, China). The respirometer was surrounded in black-out material to minimise external disturbance, and fish in the test section were viewed via a CCD camera linked to a monitor. The flume respirometer was under the control of customised software that managed water velocity, the state of flushing (i.e. timing of the open “flush” and closed “measurement” phase) and O2 pressure data for MO 2 calculations. Water velocity was adjusted via a frequency-modulated induction motor that drove an impeller and produced an efficient flow of water within the confined space of the respirometer. The software incorporated fish dimensions to allow manipulation of swimming velocities in BL s−1 whilst concurrently correcting for the solid blocking effect (Steffensen 1989). Oxygen saturation was measured continuously with a Microx-TX3 oxygen metre fitted with a needle type, fibre-optic oxygen sensor (PreSens Precision Sensing GmbH, Regensburg, Germany).

Fish were transferred to the swim flume at around 1800 h daily and left to swim overnight at 0.5 BL s−1. Once the fish started swimming in the correct orientation with minimal bursts of sporadic activity, intermittent flow respirometry was initiated and MO 2 measurements commenced. The three lowest MO 2 values obtained between 1800 h and 0700 h were assumed to be the metabolic rate of settled (i.e. non-stressed) fish swimming steadily at 0.5 BL s−1. At 0700 h, a critical swimming velocity (U crit ) test (Brett 1964) was started. The U crit test used a velocity increment of 0.25 BL s−1 for a duration of 30 min. This meant three values of MO 2 could be measured and averaged at each velocity (Korsmeyer et al. 2002). Fish were considered to crash when either ≈1/3 of the posterior body was against the rear mesh of the test section, or erratic burst activity in all directions was observed. U crit was determined using Brett’s (1964) formula:

$$ U_{crit} = U_{i} + \left( {U_{ii} \cdot Ti \cdot T_{ii}^{ - 1} } \right) $$

where U i = the highest velocity maintained for the entire incremental time, U ii = velocity increment, T i = time into final velocity before crash, and T ii = time at each speed increment. Following a crash, the fish was removed and their weight and volume measured. Following the removal of each fish, background changes to oxygen saturation were recorded and subtracted from corresponding MO2 values. The rate of change in % O2 saturation (% sat s−1) was converted to MO2 (mg O2 kg−1 min−1) according to the following:

$$ M\hbox{O}_{2} = \frac{{\left( {\frac{{\Updelta {\text{sat}}}}{100}} \right) \times {\text{PO}}_{{2\left( {@100\% } \right)}} \times \beta wo_{2} \times V_{w} \times 60}}{m} $$

where Δsat is the recorded change in water saturation per second, PO2(@100%) is the measured partial pressure of oxygen at 100% saturation, βwo 2 is the capacitance coefficient of oxygen in water at a certain salinity (36 ppt for this experiment) in mg O2 L−1 kPa−1, V w is the volume of water (volume of the respirometer after subtraction of fish volume), 60 is the number of seconds in a minute, and m is the live mass of the fish (modified from; Steffensen 1989). The MO2 data from an individual fish were extrapolated to 0 BL s−1, using an exponential regression to estimate the minimum rate of metabolism (i.e. standard metabolic rate, SMR). The top 10% of overnight MO2 values provided the best estimate of maximum metabolic rate (MMR) because these values showed little intra-individual variation and were consistently higher than MO2 at U crit . Aerobic metabolic scope (AMS) was calculated by subtracting SMR from MMR. The gross cost of transport (GCOT, mg O2 kg−1 m−1) was calculated by dividing the MO2 values by the corresponding swimming velocity (m s−1). The speed returning the lowest GCOT value was taken at the optimal swimming speed of the fish (U opt ).

Statistical analyses

Statistical analysis was carried out using SigmaPlot® 11 and JMP® 7. Mann–Whitney rank sum tests were used to compare each of the growth parameters at 14.9°C whilst a Kruskall–Wallis ANOVA, using the Dunn’s Method for specific pairwise comparisons, compared the productivity data at 21.1°C. A two-way ANOVA was used to compare MO2, GCOT and NCOT between exercised and non-exercised fish over the range of swimming velocities used in the respirometer. Exponential regressions (of the form: \( y = c + ae^{kx} \)) were used to describe MO2 over the range of swimming velocities. Minima of polynomial regressions (of the form: \( y = ax^{2} + bx + c \)) were used to determine U opt from GCOT for every fish, and U opt was compared between treatments with a t-test. The same polynomial method was used to calculate the mean U opt and illustrates differences in pooled GCOT over different swimming velocities for each treatment group. Two sample t-tests compared the different exercise regimes in terms of U crit , SMR, MMR and AMS. Significance was accepted at P ≤ 0.05.

Results

Effects of long-term exercise on kingfish growth at 21.1°C

There was a significant difference in SGR W amongst the different levels of exercise at 21.1°C (H3 = 10.528; P < 0.05). However, only the 0.75 BL s−1 group showed a significant 9.6% increase in SGR W over the low flow control (P < 0.05) (Table 1). No other significant difference was found between the exercise groups in terms of GR L (H3 = 3.593; P > 0.05) or ΔCF (H3 = 7.049; P > 0.05) (Table 1). Tanks subjected to the 0.75, 1.5 and 2.25 BL s−1 regimes had more efficient FCRs than the control 0 BL s−1 group (8, 6 and 3.5% difference respectively) (Table 1), but no statistical test could be performed to test whether a true difference exists.

Table 1 Differences in the productivity (growth and FCR) of fish exposed to various exercise regimes (0.0–2.25 BL s−1) at two different temperatures (14.9 and 21.1°C)
Full size table

Effects of long-term exercise on kingfish growth at 14.9°C

No difference was found between the 2 exercise groups (0 and 0.75 BL s−1) in terms of SGR W (U = 474.5; P > 0.05), GR L (U = 629; P > 0.05) or ΔCF (U = 292; P > 0.05) (Table 1).

Effects of exercise training on swimming metabolism at 22°C

Metabolic costs increased with increased swimming velocity in both the exercise-trained and non-trained groups as expected (Clarke and Seymour 2006) (F = 56.258; P < 0.001), but a significant difference was also found in MO2 across the exercise regimes over the range of swimming speeds used (F = 4.757; P < 0.05). Specific pairwise comparisons revealed that MO2 was significantly lower in exercise-trained fish at 2 and 2.25 BL s−1 (P < 0.01) (Fig. 1a). No interaction between exercise regime and swimming velocity was detected (F = 1.498; P > 0.05). As a result of the difference in MO2 at speed (above), GCOT also differed significantly between the exercise treatments across different swimming speeds (F = 9.255; P < 0.01). This difference was driven by GCOT also being lower in the exercise-trained group at 2 and 2.25 BL s−1 (P < 0.05) (Fig. 1b). GCOTmin was significantly lower in the exercise-trained fish (0.11 mg O2 kg−1 m−1) than in non-exercise-trained fish (0.14 mg O2 kg−1 m−1) (t = 2.228; P < 0.05), but no difference was detected in U opt between the exercised (2.2 ± 0.3 BL s−1) and non-exercised groups (2.3 ± 0.1 BL s−1) (t = −0.036; P > 0.05) (Fig. 1b). No difference was detected in either SMR (t = 0.249; P > 0.05) or MMR (t = −0.791; >0.05) and concurs with the lack of difference in AMS (t = −1.059; P > 0.05) (Table 2). U crit also did not differ between the exercise groups (t = 0.0294; P > 0.05).

Fig. 1
figure 1

MO2 (a), and GCOT (b) of exercised (solid points and lines) and non-exercised (hollow points and dashed lines) fish at different relative swimming velocities. Regressions in a are \( y = 1.5081 + 1.2274e^{0.6298x} \) and \( y = - 1.2769 + 3.8925e^{0.3595x} \) for exercise and non-exercised treatments, respectively. Regressions in b are \( y = 0.0457x^{2} - 0.2148x + 0.3658 \) and \( y = 0.0364x^{2} - 0.1797x + 0.3658 \) for exercised and non-exercised treatments, respectively. Regression lines are for observational purposes only; GCOTmin and U opt were determined from individual data. Vertical drop lines represent U opt , and horizontal drop lines represent GCOTmin. Drop lines intersect at 2.185 BL s−1 and 0.1134 mg kg−1 m−1 (exercised) and 2.298 BL s−1 and 0.144 mg kg−1 m−1 (non-exercised). Velocities at which MO2 (a) and GCOT (b) differed significantly between treatments are labelled with an asterisk (P < 0.05)

Full size image
Table 2 Measures of three metabolic parameters and swimming performance for non-exercised (control) and exercise-trained yellowtail kingfish (Seriola lalandi)
Full size table

Discussion

Effects of exercise on the growth of yellowtail kingfish

Exercise training is shown to improve the productivity (i.e. growth and FCR) of the New Zealand kingfish, S. lalandi, which has commercial relevance and is consistent with work on other active species (Davison 1997). From a practical standpoint in tank and land-based culture, water flows could be manipulated to hold fish swimming at their optimal swimming speed over sustained periods. However, before exercise can be employed as a useful aquaculture practice for S. lalandi culture, there are important issues that must be addressed. For example, it is essential to know the level of exercise that supports maximal growth but it is equally important to know whether the characteristics of exercise-induced growth (i.e. expected gains in growth, required optimal level of swimming etc.) are sensitive to biotic and abiotic factors such as body size and temperature. If exercise-induced growth is proven to be context dependent, optimal productivity will obviously not occur with a single fixed level of exercise and swimming regimes will probably require adjustment for season and/or the various stages of grow out.

The growth of S. lalandi was improved by moderate non-stop swimming but our data also demonstrate a reasonable degree of context-dependency with respect to the expected growth gains and the optimal level of exercise. Firstly, exercise-induced growth was only observed at 21.1°C and not at 14.9°C. Secondly, most studies using active fish species generally show that moderate sustained exercise boosts growth by more than 20%, which is clearly greater than the 10% rate of increase observed in the current study. A more detailed comparison of Seriola species highlights a similar discrepancy between the growth rate gain of juvenile S. lalandi (10%) and S. quinqueradiata (40%) (Yogata and Oku 2000). The optimal current speed of S. lalandi (0.75 BL s−1) is also considerably slower than the currents which led to maximal growth in S. quinqueradiata (1.5–2 BL s−1. Yogata and Oku 2000). Given these two species share a similar ecotype and close phylogenetic relationship, it is unlikely that S. lalandi and S. quinqueradiata would differ markedly in their response to exercise under standardised conditions. We, therefore, suggest that temperature imposed a bioenergetic constraint on S. lalandi, which limited the growth potential of our fish to slower speeds. Indeed, 21.1°C in the current study might have been marginally suboptimal, given the superior growth of S. quinqueradiata at 22–24.6°C (Yogata and Oku 2000). From a bioenergetics perspective, suboptimal temperatures might not allow the exorbitant costs of growth and exercise to be accommodated within the available range of total metabolic scope (Brett 1979). This would effectively explain why swimming fish showed no extra growth at 14.9°C, but it does also imply that S. lalandi might have capacity to swim and grow faster at slightly higher temperatures (i.e. 22–25°C). Temperature is obviously important but the size of our fish may also contribute to variability in the growth response of S. lalandi to exercise. The 180 and 1,600 g fish used in the current study would certainly have been on a slower growth trajectory to the 4 g fingerlings of Yogata and Oku (2000). Although most other studies utilise smaller fish, Totland et al. (1987) did examine the growth of adult Atlantic salmon, Salmo salar, swimming in currents <0.5 BL s−1. Exercise regimes were low but Totland et al. (1987) did obtain a 38% increase in growth at this level of swimming suggesting that larger fish do also respond very well to exercise. With uncertainties and discrepancies unresolved, further studies should attempt to disentangle the effects of body size and temperature by examining the bioenergetics of swimming kingfish in more detail. Running trials on similarly sized fish at different temperatures would help in this regard because there is currently no data on the growth or FCR of small S. lalandi at higher temperatures (Moran et al. 2009).

Balancing the costs of growth and exercise: evidence of improved swimming efficiency

Swimming and growth are in direct competition for cellular energy which undoubtedly places a demand on fish bioenergetic budgets if both activities occur simultaneously. Balancing the costs of growth, feeding and exercise is simply not possible in some species, and costly swimming activities are reduced through behavioural downregulation (Owen 2001). In other species, however, efficient cost partitioning is apparently achieved through a reorganisation of metabolic profiles (Arnott et al. 2006). Whilst behavioural downregulation was not an option for our exercise-trained kingfish, we provide evidence that kingfish might have the capacity to reconcile the simultaneous costs of growth and exercise through improved swimming efficiency.

Exercise training is known to enhance fish swimming performance, as determined by U crit and endurance measures (Farlinger and Beamish 1978; Houlihan and Laurent 1987; Young and Cech 1993). The current study, however, shows no evidence of any change to U crit , MMR or AMS, suggesting that the costs of growth and exercise in kingfish are not supported by enhanced maximal performance. Without any observable change in SMR, our results also do not concur with the results of Nahhas (1981) and those of Bagatto et al. (2001) who show lower costs of locomotion with a shift in maintenance metabolism (SMR). However, our exercise-trained fish did have lowered MO2 and GCOT at swimming velocities centred on U opt (2.2–2.3 BL s−1). Compared with data presented in the current study, Clarke and Seymour (2006) reported lower values of U opt and GCOTmin within their study of S. lalandi, but this difference may be attributable to the twofold difference in body size and/or surgical procedures carried out. We, therefore, suggest that the lowered costs of swimming (i.e. improved swimming efficiency) may allow kingfish to grow and swim at the same time by accommodating the exorbitant costs of growth and exercise within their available metabolic scope. However, for this change in efficiency to convey a functional benefit, our kingfish must have swam at ~2.2–2.3 BL s−1 on a routine basis, a range of speed clearly in excess of the 0.75 BL s−1 regime. Routine swimming behaviour was not recorded in the current trial, but kingfish never held a stationary static position against the current because they schooled steadily into the flow and around the tank at ~0.5–0.75 BL s−1. Fish swimming in circular tanks incur an extra cost as a result of the centripetal force, required to maintain a curved (vs. straight line) swimming path (Weihs 1981). For our size of fish and swimming path radius, we estimate that our fish could have swum 50–60% faster using the same propulsive force as fish swimming linearly in a flume respirometer (He and Wardle 1988). This was calculated according to the methodology of Domenici et al. (2000) using equation 11 and Table 1 from He and Wardle (1988) but making corrections for the typical swimming path radius (85 cm) and the size of our fish over the course of the training protocol. The equivalent straight line “routine” swimming speed of our fish may therefore have been in the vicinity of 1.9–2.4 BL s−1, a speed which is remarkably close to our observed U opt value. However, without behavioural recordings to support this estimation, further research is required to resolve whether exercise-trained fish can indeed reconcile the costs of growth and exercise by swimming at U opt where swimming costs are minimised.

Exercise training and other possible routes of efficiency

It has been proposed that salmonids forced to swim in currents do not engage in costly agonistic encounters and that their improved rates of growth might be associated with reduced metabolic expenditure (Adams et al. 1995; Christiansen et al. 1991). Whilst observed in stream-dwelling salmonids, agonistic encounters are far less common and relevant to schooling kingfish. From a physiological perspective, exercise training can improve cardiorespiratory efficiency (Gallaugher et al. 2001) and muscle aerobic capacity (Farrell et al. 1991), which may all feasibly contribute to our observations of improved swimming efficiency. Swimming efficiency and enhanced growth performance may also be derived from fuel-use shifts that generate a protein-sparing effect (Kaushik and Seiliez 2010; Ozório 2008). At least in freshwater species, proximate analysis of whole body composition (Davison 1989), quantification of metabolic enzyme activity (Farrell et al. 1991; McClelland et al. 2006) and respirometric measures of instantaneous fuel-use (Kieffer et al. 1998; Lauff and Wood 1996) all support an upregulation of lipid mobilisation and catabolism during exercise. Therefore, any developments in respirometric techniques that record instantaneous fuel-use (i.e. respiratory quotients) in marine species may prove important for future work in this area.

Conclusions

Exercise training enhances the swimming efficiency of S. lalandi which feasibly allows improved productivity (growth and FCR) at higher temperatures. With evidence of temperature and body size contributing to the context-dependency of growth through swimming, this study encourages researchers and farm workers alike to consider all relevant factors before implementing exercise as a standard aquaculture practice. Indeed, more detailed research should probe the physiological basis of exercise-induced growth for the purpose of building a bioenergetics model that not only quantifies but unlocks the benefits of exercise for commercial aquaculture.