Influence of seasonal temperature on the repeat swimming performance of rainbow trout Oncorhynchus mykiss

An extensive literature exists on the recovery of metabolites and ions following exhaustive exercise in fish (see reviews by Driedzic and Hochachka, 1978; Milligan, 1996; Kieffer, 2000). Considerably fewer studies have measured how quickly or how well swimming performance recovers following exhaustive exercise (e.g. Stevens and Black, 1966; Randall et al., 1987; Brauner et al., 1994; Jain et al., 1998; Farrell et al., 1998, 2001, 2003). Given that metabolic recovery in skeletal muscle (muscle lactate, ATP and glycogen, but not PCr)occurs more rapidly at warm than cold temperatures in exhausted rainbow trout Oncorhynchus mykiss and Atlantic salmon Salmo salar(Kieffer et al., 1994; Wilkie et al., 1997; Kieffer, 2000), the expectation is that swimming performance is restored faster at a higher temperature. This expectation would be consistent with the known increase in both maximum oxygen uptake and maximum cardiac output with temperature (e.g. Butler et al., 1992; Farrell, 1997; Taylor et al., 1997) because an improved oxygen delivery system could support a more rapid recovery of the metabolic debt incurred with exhaustive exercise. However, when Atlantic salmon were angled rather than chased to exhaustion, muscle glycogen,intracellular pH and lactate were restored more rapidly under cold conditions than warm conditions (Wilkie et al.,1996). Therefore, given this uncertainty and the absence of any study that has directly measured how acclimation temperature affects the recovery of swimming performance, the primary objective of the present study was to test the hypothesis that the ability of rainbow trout to repeat a critical swimming speed (U_crit) test is improved with temperature.

A second objective of the present study was to search for correlations between the ability to reperform after an exhaustive U_critswim and the alteration in plasma levels of ions, metabolites and hormones during exercise. In particular, possible linkages were sought between the recovery of swimming performance and the post-exhaustion levels of plasma potassium, lactate and total ammonia concentrations(T_amm), all of which have been linked with muscular exhaustion in both mammals and fish. For example, high intensity exercise in mammals produces a potassium loss from the muscle(Sjøgaard et al., 1985; Vøllestad et al., 1994; Hallén, 1996), which could decrease the muscle membrane excitability and compromise tension development (reviewed by Sjøgaard,1991). Plasma potassium levels increase in rainbow trout just prior to U_crit and, moreover, exercise training increased U_crit while blunting and delaying the increases in plasma potassium and lactate just prior to exhaustion(Holk and Lykkeboe, 1998). Plasma lactate concentration has long been considered a useful indicator of aerobic limitations and anaerobic capabilities in exercise studies. Indeed,rainbow trout refused to perform repetitive bouts of burst exercise when plasma lactate concentration exceeded 13 mmol l^-1(Stevens and Black, 1966) and a poorer repeat U_crit was found for sockeye salmon Oncorhynchus nerka when plasma lactate concentration was >10 mmol l^-1 (Farrell et al.,1998). In mammals, elevated plasma T_amm has been implicated in exercise fatigue (reviewed by Mutch and Banister, 1983) due to inhibitory influences on anaerobic metabolism(Zaleski and Bryla, 1977; Su and Storey, 1994), aerobic metabolism (McKhann and Tower,1961; Avillo et al.,1981) and neuromuscular coordination(Binstock and Lecar, 1969;O'Neill and O'Donovan, 1979). Plasma T_amm also increases in rainbow trout during exercise (Turner et al., 1983; Wang et al.,1994; but see Beaumont et al., 1995a,b). Furthermore, when routine T_amm was elevated in brown trout Salmo trutta, as a result of exposure to acidic, copper-containing water, the subsequent U_crit performance was inversely related to pre-exercise plasma T_amm concentration(Beaumont et al., 1995a). Thus,because plasma levels of potassium, lactate and T_amm are good indicators of exhaustion in fish, we anticipated that they are potentially strong indicators of repeat swimming capability in rainbow trout. If this is the case, the expectation is that individual variation in these plasma variables prior to a second swim would be correlated with individual variation in the performance of a second U_crit test compared to the initial performance.

Rainbow trout Oncorhynchus mykiss Walbaum[mass=871.49±43.34 g (mean ± standard error of the mean, s.e.m.); fork length (FL)=40.95±0.77 cm, N=15] were obtained from a local hatchery (Sun Valley Trout Farm,Mission, British Columbia, Canada). They were held outdoors in a 2000 liter round fiberglass aquarium provided with aerated and dechlorinated Vancouver municipal water, pH 6.7, hardness 5.2-6.0 mg l^-1 CaCO₃,and ambient temperature 5-17°C. Experiments were performed between November 1997 and April 1998, and September-October 1998. All experimental work conformed to the guidelines set out by the Canadian Council on Animal Care, as approved by the Simon Fraser University Animal Care Committee.

Fish were swum in a modified Brett-type swim tunnel, similar to that described by Gehrke et al.(1990). The swim chamber was 21 cm diameter and 97 cm length, with a metal grid at each end. The rear grid was equipped with an electrical pulse generator (4 V) that, when contacted by the fish, provided a mild stimulation to encourage the fish to swim forward. Water speed was uniform across the swim tunnel throughout the speed range used in these experiments. The water current in the tunnel was produced by a 3-phase induction motor and a centrifugal pump attached to a tachometer whose readings (Hz) were calibrated with known water velocities, as measured with a Valeport current meter (Valeport Marine Scientific Ltd., Dartmouth, UK).

The dorsal aorta was cannulated to permit sampling of blood prior to and during the swimming tests, and during the recovery periods. Arterial cannulation was performed under anesthesia (0.1 g l^-1 buffered MS-222; Syndel Laboratories, Vancouver, BC, Canada), using the method of Smith and Bell (1964). Fish mass,fork length, maximum width and maximum depth were also measured at this time. Cannulated fish were either placed in the swim tunnel to recover or returned to the outdoor tank, where they recovered for up to 3 days before being placed in the swim tunnel. During subsequent transfer from the outdoor tank to the tunnel, fish were lightly and briefly anaesthetized (0.05 g l^-1buffered MS-222). There was no significant relationship between post-cannulation recovery time and measured swimming performance (data not presented).

Fish recovered from anesthesia in the tunnel at a water speed of 10 cm s^-1 for at least 45 min. After this time, fish performed a 20 min practice swim, as suggested in Jain et al.(1997), during which water speed was increased in 9-10 cm s^-1 increments every 2 min to a speed of ∼41 cm s^-1. Water speed was then returned to 10 cm s^-1 for 2 min and again increased in the same fashion to a speed of either 55 or 59 cm s^-1, depending on the fish's swimming capability. The practice swim, which did not exhaust the fish, prevented the training effect often observed with naive fish on a second U_crit (Farlinger and Beamish, 1977; Jain et al.,1997). Fish then recovered overnight (14-16 h) at a water speed of 10 cm s^-1.

All experiments were started between 08:00 h and 10:00 h. Fish performed a ramp-U_crit test (Jain et al., 1997). The first U_crit test was followed by a 40 min recovery period at a water speed of 10 cm s^-1and then a second ramp-U_crit test followed by another recovery period. Each ramp-U_crit test involved increasing water speed to ∼50% of U_crit over a 5 min period,after which water speed was increased in 10 cm s^-1 increments(∼15% of U_crit) every 20 min until exhaustion. Exhaustion was taken as the point at which the fish failed to swim away from the electrified rear grid after 20 s of contact. The ramp-U_crit protocol produces similar values for U_crit to the more standard U_crittesting protocol in which the longer time intervals are used from the onset of the test (Jain et al.,1997).

Blood samples (0.9 ml) were taken through the dorsal aorta cannula to measure plasma ion and metabolite levels. Normally, samples were taken immediately prior to the swimming protocol (routine samples), at exhaustion for both swim tests (U_crit exhaustion samples), and after a 40 minrecovery for both tests (recovery samples; the recovery sample for the first U_crit swim also served as the sample taken immediately before the second U_crit swim). In 14 of the 16 fish, a blood sample was taken during aerobic swimming, i.e. after 15 min at 45 cm s^-1 (approx. 69% U_crit). (These data are not reported as they simply provided intermediate values between the routine and U_crit values.) An equal volume of physiological saline solution was used to replace all blood samples(Gallaugher et al., 1992). Routine hematocrit was never less than 23% and remained elevated throughout the swim tests (see Fig. 2D).

Hematocrit was measured in microcapillary tubes after centrifugation at 2000g for 3 min. The remainder of the blood was centrifuged at 10 000g for 5 min to obtain plasma, which was stored at-80°C. Within 1 week of testing, plasma lactate and glucose concentrations were measured on 25 μl samples using a YSI 2300 lactate/glucose analyzer(Yellow Springs, OH, USA) that calibrated automatically every five samples. Plasma potassium and sodium concentrations were measured using a model 510 Turner flame photometer (Palo Alto, CA, USA). Plasma (5 μl) was diluted 1:200 with a prepared 15 mEq l^-1 lithium diluent for analysis. The machine was calibrated prior to use and checked against a standard approximately every six samples. The measurement was repeated if there was disagreement between duplicates beyond 2% of absolute value. Osmolality was measured on duplicate 10 μl samples using a calibrated Wescor Vapour Pressure Osmometer, Model 5500 (Wescor, Logan, UT, USA). The measurement was repeated if there was disagreement between duplicates beyond 3% of absolute value. The thermocouple heads were cleaned periodically in order to maintain consistency. Plasma cortisol concentration was measured using a commercial radioammunoassay kit (ICN Biomedicals, Inc., Costa Mesa, CA, USA), with a detection limit of 1.5 ng ml^-1. Plasma ammonia concentration(T_amm) was measured spectrophotometrically on 0.1 ml plasma samples (Sigma Diagnostics kit no. 171, St Louis, MI, USA) with a calibration every seven samples.

All plasma metabolites and ions were measured in duplicate and averaged for individual data. Fish were subdivided into two temperature acclimation groups based on their swimming performance (see Results) and values (mean ± s.e.m.) are presented for cold- and warm-acclimated fish. One warm-acclimated female fish that was overtly gravid was not included in the statistical analysis to eliminate any confounding effect, because reproductive maturity is known to negatively affect U_crit performance in salmon (Williams et al.,1986). Statistical comparisons within temperature groups were made with a one-way repeated measures analysis of variance (ANOVA) followed by a post hoc Tukey test. With this test, the values associated with each fish were compared to other levels at other sampling times for the same fish to determine whether either swimming speed or metabolite levels changed throughout testing. Comparisons of swimming performance and metabolite levels between temperature groups were made using t-tests. U_crit1 was compared to U_crit2 using a Bland-Altman plot. Bland and Altman(1986, 1995) introduced this method of graphical analysis to assess the equivalency of two testing approaches (here U_crit1 and U_crit2), while removing the bias that comes from assuming that one method represents the true value(independent variable). The Bland-Altman plot uses the mean of both methods as the independent variable and the difference between the two testing methods as the dependent variable. If the linear regression of the points is non-significant, then the two testing procedures (i.e. U_crit1 and U_crit2 here) can be considered to be equivalent testing procedures. Sub-groups can be identified within a data set in a Bland-Altman plot by demonstrating different significant linear regressions from each other. In the present study,different regressions would identify sub-groups with different relationships between U_crit1 and U_crit2. Relationships between U_crit values and plasma variables were fitted with the best-fitting regressions using the options provided in Sigma-Plot (SPSS Inc.; Chicago, IL, USA). P<0.05 was used to establish statistical significance.

As water speed increased, fish progressed from a steady swimming mode to one that included periods of burst-and- glide swimming. In conjunction with higher speeds, fish ram-ventilated their gills, except during burst-and-glide swimming when active ventilation was observed. Visually, swimming behavior did not appear to be different for the first and second U_crittests.

A Bland-Altman plot revealed that U_crit1 and U_crit2 were equivalent testing procedures(P=0.98), but visual inspection of the plot revealed that overall the fish could be divided into two sub-groups each with a different and significant linear relationship (Fig. 1A). Each of the two sub-groups corresponded to different acclimation temperatures and hereafter are termed warm- and cold-acclimated fish (14.9±1.0°C and 8.4±0.9°C, respectively; see Table 1).

U_crit1 performance was temperature dependent(Fig. 1B;r²=0.74, P<0.05). U_crit1 (78.9±1.0 cm s^-1) for warm-acclimated fish was significantly greater (P<0.05) than that for cold-acclimated fish (59.1±2.5 cm s^-1; Table 1). However, U_crit2 did not show any temperature dependency. Unexpectedly, U_crit2 performance (65.8±2.70 cm s^-1) for warm-acclimated fish was significantly lower than U_crit1, whereas U_crit2 for cold-acclimated fish (58.0±4.2 cm s^-1) was the same as their U_crit1 values (Table 1). As a result, the overall relationship between U_crit1 and U_crit2 was best described by a polynomial equation(y=-204.3+7.57x-0.05x²; P<0.001; Fig. 1C),with cold-acclimated fish lying close to the line of identity and warm-acclimated fish lying below the line of identity. Thus, while warm acclimation conferred a faster U_crit1, a similar swimming speed could not be attained after a 40 min recovery period, as shown by recovery ratios that are less than unity for warm-acclimated fish(Fig. 1D).

There were no significant differences between the cold- and warm-acclimated groups of fish in terms of routine values for plasma levels of lactate,potassium, T_amm, sodium, glucose, cortisol and osmolality and hematocrit. When cold-acclimated fish were exhausted at U_crit1, plasma levels of lactate, potassium and T_amm, as well as hematocrit, all increased significantly(Fig. 2A-D). Plasma cortisol(Fig. 2E) and sodium(Fig. 2F) levels were unchanged at exhaustion for U_crit1. After a 40 min recovery from U_crit1, plasma lactate increased significantly beyond the level observed at exhaustion, plasma T_amm decreased to the routine level, and plasma potassium and hematocrit remained elevated at the same level. As a result, plasma lactate and potassium levels, and hematocrit were all significantly elevated at the outset of the U_crit2 test.

For cold-acclimated fish exhausted at U_crit2, plasma levels of lactate, potassium, sodium and T_amm, and hematocrit, were again significantly elevated compared with the routine values, but no more so than for U_crit1. In fact, compared with the recovery values for U_crit1, plasma lactate levels had decreased significantly (Fig. 2A) at exhaustion for U_crit2, while T_amm had increased significantly(Fig. 2C). Similar to U_crit1, plasma lactate increased during the 40 min recovery from U_crit2 to a level significantly higher than that observed at exhaustion, plasma T_amm decreased to the routine level, and plasma potassium and hematocrit remained elevated at the same level. As a result, none of the recovery values for U_crit2 in cold-acclimated fish were significantly different to those for U_crit1. Plasma levels of cortisol,glucose and osmolality remained unchanged throughout both swimming protocols(data not shown). Therefore, the second swim for cold-acclimated fish had no additive effects on any of the measured plasma variables.

When warm-acclimated fish were exhausted at U_crit1,plasma T_amm and hematocrit increased by the same amount as for cold-acclimated fish (Fig. 2C,D). In contrast, the faster U_crit1 of the warm-acclimated fish was associated with significantly larger increases in plasma levels of lactate and potassium(Fig. 2A,B) compared with cold-acclimated fish. Furthermore, warm-acclimated fish significantly increased plasma sodium and cortisol levels at exhaustion for U_crit1 (Fig. 2E,F), unlike cold-acclimated fish. After a 40 min recovery from U_crit1, the levels of plasma lactate, potassium, T_amm, sodium and cortisol, as well as hematocrit all remained significantly elevated in warm-acclimated fish, whereas only plasma levels of lactate, potassium and hematocrit remained elevated in cold-acclimated fish (Fig. 2). In addition, plasma lactate, potassium, sodium and cortisol remained elevated in warm-acclimated fish at levels that were significantly greater than those observed in cold-acclimated fish during recovery. In fact, the plasma lactate level was about threefold higher and plasma potassium almost twofold higher. These results suggest that the higher U_crit1 of warm-acclimated fish may have been partly due to a greater anaerobic swimming effort compared with cold-acclimated fish, and (or) lactate and potassium were released from muscle to plasma to a greater extent.

Compared with cold-acclimated fish, warm-acclimated fish clearly began the second U_crit test with a greater plasma ionic and metabolic disruption and, as a result in these fish, U_crit2 was significantly lower than U_crit1. In addition, while U_crit2 for warm-acclimated and cold-acclimated fish was the same, warm-acclimated fish displayed a significant, further increase in plasma potassium levels(Fig. 2B) at exhaustion and a significant, further increase in plasma lactate levels(Fig. 2A) during the recovery from U_crit2. However, plasma T_amm did not recover to a routine level, as it did in the cold-acclimated fish(Fig. 2C). Therefore, the second U_crit swim of warm-acclimated fish produced significant additive effects on some of the plasma variables, unlike in cold-acclimated fish where there were none.

The initial swimming performance of individual fish was related to the appearance of lactate in the plasma. Plasma lactate concentrations measured at U_crit1 and after a 40 min recovery were both linearly related to U_crit1 (Fig. 3; r²=0.73, P<0.05 and r²=0.79, P<0.05, respectively). As might be expected from Fig. 3, plasma lactate concentrations were highly correlated with each other (2^ndexhaustion with 1^st exhaustion: r²=0.95, P<0.05; 1^st exhaustion with 1^st recovery: r²=0.92, P<0.05; 2^nd exhaustion with 1^st recovery: r²=0.94, P<0.05;2^nd recovery with 2^nd exhaustion: r²=0.94, P<0.05).

The difference in swimming performance between U_crit1and U_crit2 was significantly related to the plasma lactate concentration prior to the second U_crit test(Fig. 4). This relationship was described by either a polynomial (r²=0.74), or a 2-parameter power (r²=0.65) regression. Both types of analysis suggest that the reduction in U_crit2 relative to U_crit1 occurred when fish reached a plasma lactate of 12.2 mmol l^-1 (95% confidence intervals of 7.9 and 16.5 mmol l^-1) 40 min after being exhausted by an initial U_crit swim test. Only warm-acclimated fish reached this threshold plasma lactate level.

Swimming effort in the initial swim was also related to the appearance of potassium in the blood. Plasma potassium concentration measured at U_crit1 was linearly related to U_crit1(r²=0.60, P<0.05). However, there was no significant correlation between plasma potassium levels and performance on the second swim. Plasma T_amm at exhaustion was not significantly related to U_crit1, but T_amm values for the 1^st recovery were related to U_crit1 (Fig. 5; r²=0.34, P<0.05). There were no other significant correlations for plasma T_amm.

The influence of acclimation temperature on the plasma ionic and metabolic responses to exercise is illustrated by the significant linear correlations that existed between plasma lactate, cortisol and potassium levels and temperature (Table 2). There were no significant correlations with temperature and the other parameters measured (T_amm, [sodium] and hematocrit).

One overtly gravid, warm-acclimated female fish was treated as an outlier,based on its slow swimming performance, and was not used for any correlation analysis. However, it is important to note that all the plasma changes observed in this fish were consistent with the slower swimming performance of the cold-acclimated fish.

This study tested the hypothesis that warm-acclimated rainbow trout would perform better in repeated U_crit swimming tests than cold-acclimated fish. The present findings, however, do not support this hypothesis because U_crit2 was significantly lower than U_crit1 in warm-acclimated fish than in cold-acclimated fish. At U_crit1, the warm-acclimated fish showed a greater metabolic disturbance in the plasma compared with cold-acclimated fish and also showed additive effects for the second U_crit, unlike the cold-acclimated fish. Therefore, although warm-acclimated fish swam better than cold-acclimated fish for U_crit1, as expected, the consequence of this faster U_crit1 was a reduced performance on the second U_crit test. If anything, it appeared that warm-acclimated fish, by apparently swimming harder and possibly more anaerobically, were unable to recover sufficiently well during the fixed recovery period to repeat this initial level of performance. For cold-acclimated fish, however, the 40 min recovery period was sufficient for adequate recovery and allowed swimming performance to be repeated. Therefore,we are left with the conclusion that overall recovery, as it pertains to repeat swimming capabilities and time allowed for recovery, was superior for the cold-acclimated compared with the warm-acclimated group of rainbow trout.

Our original hypothesis, which we now reject, was based on the established temperature dependence of post-exercise metabolic and ionic recovery when salmonids are chased to exhaustion to produce similar levels of intracellular acidosis, lactate accumulation and glycogen depletion in white muscle regardless of temperature (Kieffer et al.,1994; Wilkie et al.,1997). However, when Atlantic salmon were angled to exhaustion at a warmer temperature, essentially the opposite effect of temperature on post-exercise muscle recovery was obtained; they displayed a greater depletion of muscle glycogen, a greater intracellular acidosis and a slower recovery of muscle metabolites at the warmer temperature compared with colder temperatures(Booth et al., 1995; Wilkie et al., 1996). The present findings for U_crit swim tests are more in line with data obtained when fish are angled rather than chased to exhaustion because the metabolic disturbances were higher and performance recovery slower at warmer temperatures. We suggest that the disparity among studies could simply reflect differences in the degree of exhaustion and the methods used to exhaust the fish, with fish becoming more exhausted because they perceive the chasing protocol as more of a threat or provocation than either angling or U_crit testing. Given this possibility, cold-acclimated fish could opt to stop swimming sooner than warm-acclimated fish to preserve glycogen reserves.

A U_crit value, like time-to-exhaustion at a prescribed water speed (e.g. Facey and Grossman,1990; Mitton and McDonald,1993), allows quantification of the swimming effort, something that is not easily done when fish are chased or angled to exhaustion. U_crit tests also encompass a spectrum of swimming speeds,with the aerobic demands of swimming up to maximum oxygen uptake being met by cardiorespiratory adjustments, while white muscle recruitment and anaerobic metabolism increasingly supports the higher muscular power output near U_crit (Burgetz et al.,1998), culminating in exhaustion(Brett, 1964; Beamish, 1978). The simplest explanation for the higher U_crit1 values obtained for warm-acclimated compared with cold-acclimated fish is a greater involvement of anaerobic swimming, given the significantly larger alterations in plasma metabolites observed for warm-acclimated rainbow trout. Certainly, the warm-acclimated fish were more stressed than the cold-acclimated fish, as judged by the greater elevation in plasma cortisol levels. However, since muscle metabolites were not measured here, we cannot exclude other possibilities. The higher levels of plasma potassium, lactate and T_amm, as well as the additive effects of the second swim,could simply reflect a greater release of lactate and potassium into the plasma because the release of lactate and hydrogen ions from white muscle to the blood is known to be temperature dependent (see Kieffer, 2000). Nevertheless,it is unlikely that different muscle glycogen levels were a factor since these are unaffected by acclimation temperature(Kieffer, 2000).

Rome et al. (1985) showed that acutely exposing warm-acclimated carp Cyprinus carpio to cold water resulted in white muscle fibres being recruited at a lower swimming speed, and this `compression of recruitment order' led to earlier fatigue and a reduced sustained swimming speed. However, when the carp were cold-acclimated, they recruited white muscle at a higher swimming speed than warm-acclimated fish, presumably because cold temperature acclimation had improved the mechanical performance of the red muscle. The present findings are consistent with this earlier work with carp in that the cold-acclimated rainbow trout appeared to rely less on anaerobic white muscle than warm-acclimated fish, but the two studies differ in that cold-acclimated rainbow trout had a lower U_crit than warm-acclimated rainbow trout whereas cold-acclimated carp swam to the same maximum speed as warm-acclimated fish (Rome et al.,1985). Rome et al.(1985) suggested three possible physiological differences in cold-acclimated fish compared with warm-acclimated fish: (1) a higher mechanical power output from aerobic muscle, (2) limitations on the neural control of locomotory muscle and (3)limitations of the respiratory and circulatory systems in supplying oxygen. The present findings suggest a fourth possibility: fish may opt to swim to different states of exhaustion depending on either the temperature or a resulting physiological condition. One benefit of limiting the level of exhaustion under cold conditions appears to be a more reasonable recovery rate, which allows for repeated performance. At warm temperatures, fish benefit from a higher initial level of performance but, by exhausting themselves to a relatively greater degree, have the disadvantage of a more prolonged recovery period. An additional disadvantage, but for unknown reasons, is that exhaustive exercise at warm, but not at cold temperatures,can result in appreciable levels of postexercise mortality (see Kieffer, 2000).

The present conclusions are also in line with the results of McKenzie et al. (1996) working with Nile tilapia Oreochromisnilotica. They found that warm-acclimated fish had a greater cost of recovery (a higher and more prolonged post-exercise oxygen consumption) after being chased to exhaustion than cold-acclimated fish. Interestingly, white muscle lactate accumulation was similar for 16°C-acclimated and 23°C-acclimated tilapia,suggesting that muscle lactate may not always be a reliable measure of post-exercise recovery. However, 23°C-acclimated tilapia excreted over twice the amount of ammonia post-exercise than 16°C-acclimated fish. Kieffer et al. (1998)similarly found that ammonia excretion at 75% U_crit was almost threefold higher for 15°C-acclimated than 5°C-acclimated rainbow trout, while protein utilization at 75% U_crit was 30% at 15°C versus 15% at 5°C. Likewise, in the present study, we observed a significantly higher plasma T_amm in warm-acclimated rainbow trout. As discussed by McKenzie et al.(1996), the elevated ammonia production could be a result of either increased protein metabolism to fuel locomotion or increased protein degradation from tissue damage. Since elevated T_amm is thought to have inhibitory actions on neural and muscle activity in fish (Beamount et al., 1995a), the larger elevation in plasma T_amm in warm-acclimated fish is perhaps critical to survival post-exhaustion. On the other hand, tissue damage might negatively affect U_crit2.

U_crit values were comparable to those reported earlier by Jain et al. (1997) for rainbow trout of the same size in the same swim tunnel [1.64-1.66 body lengths(BL) s^-1] and higher than those reported for 822-1118 g rainbow trout (0.94 BL s^-1 and 0.53 BLs^-1 at 11°C and 18°C, respectively; Taylor et al., 1996). Comparisons also can be made with studies on smaller rainbow trout, which are expected to attain slightly higher U_crit values(Brett, 1964) than the 879 g fish used here. U_crit values of 1.8 to 2.0 BLs^-1 are reported for 530-730 g rainbow trout at 18-19°C(Gallaugher et al., 1992) and 2.13 BL s^-1 for 431-483 g rainbow trout at 7-11°C(Burgetz et al., 1998). For 320-520 g brown trout, U_crit values were 2.2 BLs^-1 at 15°C and 1.85 BL s^-1 at 5°C(Butler and Day, 1993; Butler et al., 1992).

As anticipated, a 40 min recovery period allowed full recovery of swimming performance for cold-acclimated fish. Originally it was suggested that salmonids be given 4 hbetween U_crit tests(Brett, 1964) to ensure a return to routine O₂ consumption but not necessarily to routine glycogen levels. Subsequently, recovery times of 2 h(Brauner et al., 1994), 1 h(Randall et al., 1987), 45 min(Farrell et al., 1998, 2003) and 40 min(Jain et al., 1998) have all been shown to be sufficient for salmonids to repeat U_crittests without any significant decline in performance. Here fish were provided with a low speed water current during recovery and this may have aided their recovery, since recent studies with rainbow trout(Milligan et al., 2000) and coho salmon Oncorhynchus kisutch(Farrell et al., 2001) have shown that low to moderate swimming post-exhaustion greatly aids metabolic recovery through a warm-down effect. In contrast, recovery time without a warm-down is >2 h for optimal performance on a time-to-exhaustion test(Mitton and McDonald, 1994). Wang et al.(1994) reported that muscle phosphocreatine and ATP levels were restored within 30 min of rainbow trout being chased to exhaustion, while the post-exercise decline of oxygen consumption lasted 3-3.5 h (Scarabello et al., 1991). However, routine oxygen consumption does not have to be restored before adult sockeye salmon can repeat a second U_crit test (Farrell et al., 1998, 2003).

There was generally good agreement between the routine plasma variables reported here and those reported in previous studies(Butler and Day, 1993; Eros and Milligan, 1996; Pagnotta et al., 1994;Thorarensen et al., 1994; Wang et al.,1994). However, the plasma lactate concentrations at U_crit in this study, especially those for the warm-acclimated fish (7.3 mmol l^-1), were at the high end of literature values for U_crit swimming (1.5-5.5 mmol l^-1) (Butler and Day,1993; Gallaugher et al.,1992; Thorarensen et al.,1993; Holk and Lykkeboe,1998; Farrell et al.,1998). Milligan(1996) cites a range for plasma lactate levels of 2-13 mmol l^-1 immediately after chasing,increasing to peak values of 12-20 mmol l^-1 at 2 h post-exercise. The values reported here for cold-acclimated fish of 4.3 mmol l^-1at U_crit and 8.9 mmol l^-1 40 min later are at the low end of this range, whereas those for the warm-acclimated fish (7.3 mmol l^-1 at U_crit and 16.6 mmol l^-140 min later) are at the upper end of the range and approached the level reached (17.8 mmol l^-1) approximately 90 min after a hypoxic U_crit test (Farrell et al., 1998).

The second objective of the present study was to determine whether any of the measured metabolites displayed threshold levels that, if surpassed in the first swim challenge, were indicative of a metabolic condition that negatively affected subsequent swimming performance. Plasma lactate level was the only candidate: the plasma lactate level before U_crit2 was significantly correlated to the subsequent swimming performance(U_crit2). The threshold plasma lactate level of approximately 12.2 mmol l^-1 (95% CI 7.9-16.4) agrees with that of 13 mmol l^-1 reported by Stevens and Black(1966) for burst exercise with rainbow trout and 10 mmol l^-1 for sockeye salmon(Farrell et al., 1998). In the earlier studies, fish refused to swim if the lactate threshold was surpassed. However, no fish refused to swim outright in the present study and instead U_crit performance was reduced by 8-31%. Thus, because anaerobic metabolism is increasingly required to support swimming speeds greater than 70% U_crit(Burgetz et al., 1998),elevated levels of lactate above the lactate threshold is probably indicative of a failure to fully recruit anaerobic metabolism in white muscle (e.g. through decreased muscle pH and glycogen stores). This idea needs further study, however, because plasma lactate dynamics are complex, reflecting rates of production in the muscle, rates of release from the muscle and rates of clearance from the blood. While the present study suggests that production may be greater at warmer temperatures, release rate is dependent on temperature(Keiffer et al., 1994) and clearance rate is inversely related to temperature(Kieffer and Tufts, 1996).

Beaumont et al.(1995a,b)reported that copper-exposed brown trout in water of low pH had poor U_crit values and suggested that the elevated plasma T_amm inhibited white muscle activity either directly or through CNS inhibitory mechanisms, because elevated plasma T_amm levels were correlated with the reduced U_crit values. In the present study, we found no significant correlations between swimming performance and plasma T_amm. However, our data are not necessarily at odds with the suggestion of Beaumont et al.,(1995a,b)because the plasma T_amm levels reported in the present work were half those measured in copper-exposed brown trout and, in the earlier studies, U_crit was not reduced appreciably until plasma T_amm reached levels >200 μmol l^-1. A plasma T_amm level >600 μmol l^-1resulted in fish refusing to swim. In the present study, T_amm reached only 100 μmol l^-1 and was restored between U_crit tests for cold-acclimated fish,although not for warm-acclimated fish (Fig. 2).

Several studies report a temperature optimum for U_crit. For sockeye salmon, 15°C was the optimum temperature for U_crit, metabolic scope(Brett, 1964) and cardiac performance (Brett, 1971; Davis, 1968). The preferred temperature for sockeye salmon, however, appears to be slightly cooler(10-12°C; Birtwell et al.,1994; Spohn et al.,1996). Garside and Tait(1958) suggested a preferred temperature range for rainbow trout of 11-16°C, which coincides with the optimum temperature range suggested for cardiac performance(Farrell et al., 1996; Taylor et al., 1997; Farrell, 2002). The present experiments show that the shift in responses to repeated swimming for cold-and warm-acclimated fish occurred at around 12°C. Therefore, the fish's preferred temperature may reflect sub-maximal rates for certain activities because of negative consequences in terms of rates of recovery.

In summary, we provide evidence that warm-acclimated rainbow trout have a higher U_crit than cold-acclimated fish, but associated with this higher U_crit is a greater metabolic and ionic disturbance. A consequence of this elevated disturbance is that warm-acclimated fish do not recover well enough after a 40 min rest to perform a second test at the same level as the first one, whereas cold-acclimated fish do. Elevations in plasma lactate (but not plasma potassium, T_amm and cortisol) were significantly correlated with the poorer repeat swimming performance.

This work was funded by NSERC Canada. We thank Dr Joe Cech for the generous loan of the glucose/lactate analyzer.