Studies of Tropical Tuna Swimming Performance in a Large Water Tunnel: II. Thermoregulation

This paper reports experiments, conducted in a large water tunnel, testing the capacity of yellowfin tuna (Thunnus albacares) to control rates of heat gain and loss in response to changes in ambient water temperature (T_a) and swimming velocity (U). Tunas swim continuously in search of prey and depend upon locomotion to ventilate their gills (Brown and Muir, 1970; Roberts, 1978) and to maintain hydrostatic equilibrium (Magnuson, 1973, 1978). Tuna aerobic locomotor (red) muscle, which powers sustained swimming, is unique among teleosts in that the supporting vasculature forms counter-current heat exchangers (retia mirabilia) that conserve metabolic heat. Large bluefin tuna can elevate body temperature (T_b, red muscle temperature) to as much as 20°C above T_a (Carey and Lawson, 1973).

As in mammals and birds, the elevated T_b of tunas enhances physiological and biochemical functions (Graham, 1975; Stevens and Neill, 1978; Dizon and Brill, 1979; Stevens and Carey, 1981). Endothermy also affords tunas a level of independence from the effects of acute changes in T_a such as those encountered during feeding sojourns into deeper, cooler water (Carey and Lawson, 1973; Neill and Stevens, 1974; Neill et al. 1976; Stevens and Neill, 1978; Dizon and Brill, 1979; Graham, 1983; Holland et al. 1992).

Although many endotherms (i.e. birds and mammals) maintain a nearly constant T_b by precisely balancing rates of heat gain and loss, the T_b and T_x (=T_b-T_a) of tunas is variable and is affected by both T_a and by U (Stevens and Fry, 1971; Carey and Lawson, 1973; Stevens and Neill, 1978; Dizon and Brill, 1979; Graham, 1983). This has raised questions about the potential for T_b control (Carey and Teal, 1966; Stevens and Fry, 1971; Carey and Lawson, 1973). Do tuna have thermoregulatory capabilities similar to, but less precise than, those of birds and mammals or is heat balance unregulated with the T_x values of the tuna being the consequence of a constant, though elevated, thermal inertia (Neill and Stevens, 1974; Stevens and Neill, 1978)?

Because there are circumstances in which an elevated T_b and thermal inertia may be a disadvantage, the ability to modulate thermoconservation would be useful. At high activity levels or in warm T_a, a tuna may risk overheating and thus would need to curtail heat retention to avoid thermal limits to activity and distribution (Neill et al. 1976; Stevens and Neill, 1978; Dizon and Brill, 1979). Also, a fish ascending to the surface following a period of foraging at depth, may have a T_b less than T_a and would be able to gain heat faster by increasing heat transfer (Dizon and Brill, 1979; Holland et al. 1992; Brill et al. 1994).

The water tunnel and the protocols for maintaining tuna in captivity and during an experiment are described in Dewar and Graham (1994). An important feature for thermoregulation studies was the ability to bring about rapid (1°C min⁻¹) thermal changes by the admixture of large reservoirs (3000 l each) of hot and cold water.

Yellowfin selected for study were dip-netted from the holding tank, placed immediately in a plastic cradle filled with hyperoxic water, and quickly fitted with a 30 gauge copper–constantan thermocouple that penetrated the red muscle (Fig. 1A,B). To facilitate implantation and prevent thermocouple damage during insertion, the wire was housed in a 17 gauge hypodermic needle and its tip bent to form a hook. The needle and thermocouple were inserted through a preformed hole in the corselet (at the leading edge of the first dorsal fin, approximately 1.5 cm from the dorsal mid-line; Fig. 1A) and directed ventro-posteriorly. The thermocouple was connected to a Physi-Temp digital thermometer (model BAT 12) and, during insertion, the temperature readout was used to position the thermocouple tip in the warmest sector of red muscle. Once the thermocouple position had been set, the 17 gauge needle was carefully withdrawn while T_b was monitored to ensure patent thermocouple position. The wire was then disconnected from the thermometer and secured in place by ‘self-tying’ loops through the corselet. Thermocouple implantation generally required less than 2 min, following which the fish was placed into the water tunnel. Thermocouple position was verified post mortem.

Immediately after placing the fish in the water tunnel’s working section, the lid was secured and the thermocouple wire was fed out and reconnected to the digital thermometer. For the first 2 h, U and illumination were adjusted to induce stable swimming as detailed by Dewar and Graham (1994). Throughout the experiment T_b, T_a, O₂ concentration and U were monitored continuously.

The response of T_b to rapidly implemented cyclic changes in T_a was monitored while the fish swam at a constant U. Fig. 2 illustrates the sequence of temperature changes used in all tests. Once the fish had been swimming stably in the water tunnel for at least 2 h, T_a was rapidly cooled from initial, ambient levels (23–25°C) to 18°C. After T_b had fallen to at least 22°C the first cyclic T_a change was initiated; T_a was increased by 7°C then, once T_b had risen to just 0.5°C below T_a, T_a was rapidly reduced by 4°C. The same sequence and timing were repeated two more times, but each subsequent thermal cycle was shifted upward by 3°C. Thus, the complete experiment consisted of three thermal cycles staggered by 3°C (Figs 2, 3 and 4).

After the final cyclic change in T_a, the influence of increased U on heat balance was examined by monitoring the response of T_b to two incremental increases in U at T_a values of 30, 25 and 20°C (see Fig. 5). At each T_a, the yellowfin was first held at a U of 0.9 L s⁻¹ (where L is fork length) and T_b was recorded. Then U was increased to 1.5 L s⁻¹ and held there for 10 min while changes in T_b were recorded. Following this, U was returned to 0.9 L s⁻¹. After T_b had cooled to control levels, U was elevated to 2 L s⁻¹ for 5 min, and T_b records were obtained. Then U was returned to 0.9 L s⁻¹ while cooling was again monitored.

Using equation 2 and a MATLAB curve-fitting program, k was calculated from the T_b(t) curves associated with each change in T_a. The program was applied after the latency, or lag time, when T_b began to change exponentially (Fechhelm and Neill, 1982; Weller et al. 1984; Brill et al. 1994). To remove the latency period from each data set, successive points were eliminated from the beginning of the data set until the predicted values closely approximated the actual data over the region where T_b changed most rapidly [minimizing the square root of the mean of the sum of the squared residuals (RMS error)]. The T_e (equation 2) was estimated from the T_x measured at the end of the initial 2 h recovery period and was then adjusted for each estimation of k to minimize the RMS error.

The effects of cyclic T_a changes on the T_b of three yellowfin tuna are shown in Figs 2, 3 and 4 and the k values for the warming and cooling phases are summarized for nine fish in Table 1. The high k for cooling during cycle 3 (Fig. 2) occurred in all fish (Table 1) and this value is significantly higher [P<0.05; non-parametric a posteriori, multiple-comparison test (Nemenyi, 1963)] than k values in cycles 1 and 2, which do not differ from one another. No significant differences exist between the k values for warming.

The thermal latency (the time interval between the reversal of the thermal gradient and the onset of the exponential decline in T_b) provides another indication of the yellowfin’s capacity to regulate heat transfer. Fig. 3 illustrates the effect of T_a on the latency for yellowfin no. 5. The reduction in cooling latency between cycles was observed for all yellowfin and the values in cycle 3 (mean 2.9 min) are significantly less (P<0.05; statistics as above) than in both cycles 1 and 2 (mean values 6.0 and 6.4 min). Latencies during heating were brief (1.5–4 min) and did not differ among cycles. This pattern of thermal dependence is the same as that observed for k.

The T_b traces for yellowfin no. 8 (Fig. 4) are distinguished by the ‘thermal notches’ that occurred immediately after the thermal gradient reversal during cycles 1 and 2 but not during cycle 3. The thermal notch is characterized by an abrupt reduction and then rise in T_b, followed by a more gradual decline. The same pattern of notches was observed for yellowfin no. 9 and a notch is also seen in Fig. 5.

The effect of increased U on T_b at three T_a values is shown in Fig. 5. At 30°C, the initial steady-state T_x is low (0.7°C) and the two step changes in U have a minimal effect (1,+0.2°C; 2,+0.3°C). The steady-state T_x at both 25°C (1.7°C) and 20°C (1.8°C) is greater than that at 30°C, and the elevation of T_b associated with increases in U is three-to fourfold higher at 25°C (1,+0.7°C; 2,+1.1°C) and 20°C (1,+0.9°C; 2,+1.2°C) than at 30°C. It is also noteworthy that a thermal notch occurred with the reduction of T_a from 30–25°C.

Several previous studies examined the potential of tunas to control heat retention physiologically. Long-term ocean tracks by Carey and Lawson (1973) showed that large bluefin tuna can maintain stomach temperature within 5°C during 15°C changes in T_a. Holland et al. (1992) analyzed T_b/T_a relationships for a bigeye tuna tracked at sea: they determined that it could adjust k to retard heat loss during excursions into cool water and to allow rapid heat gain upon its return to warmer surface waters (when T_a>T_b). Modulation of k was also inferred during laboratory studies by Dizon and Brill (1979), where long-term changes in T_a from 20 to 30°C, while increasing swimming speed and consequently Hp, had a negligible effect on the steady-state T_x of yellowfin tuna.

While all of these findings support the hypothesis that tuna can regulate heat balance physiologically, each study is problematic in that swimming speed and consequently Hp (equation 1) were either unknown or uncontrolled (Brill et al. 1994). Hp will influence both steady-state and non-steady-state T_b and therefore estimated k values (Graham, 1983; Brill et al. 1994). In addition, inferences about physiological regulation based solely on comparisons of rates of heating and cooling (Holland et al. 1992) are complicated by the absence of Hp data. Because Hp can retard cooling and increase the rate of heating (Graham, 1983; Brill et al. 1994), it may produce results similar to those expected through the physiological regulation of heat balance. The differential influences of Hp on heating and cooling may be further confounded by the increased swimming effort associated with swimming upward as opposed to downward.

Our studies with stably swimming yellowfin are the first investigations of tuna thermoregulation in which the key variables influencing T_b (T_a and U) could be controlled. The ability to work with fish several hours after handling and to implement cyclic changes in T_a, approximating those experienced during foraging excursions (Holland et al. 1990), enabled us to mimic natural conditions and thereby to increase the probability of triggering ‘normal’ thermoregulatory responses.

The water tunnel, in conjunction with our experimental design, also enabled us to factor out effects that changes in T_a may exert on Hp (controlled in our tests) and cardiac output, both of which affect k. Cardiac performance affects tuna heat balance because retial heat transfer is inversely related to blood velocity (Dizon and Brill, 1979; Graham, 1983). Thus, alterations in cardiac output, in response to changes in either U or T_a, could affect heat exchange and produce results that mask or mimic physiological thermoregulation. The heart-rate reduction caused by cooler water would retard heat flux whereas warmer water would increase it.

Our experimental design allowed for separate comparisons of the k values estimated for heating and cooling during each of the thermal steps. This eliminated the need to correct for the influences of Hp and cardiac output within one thermal cycle. Also, the potential contribution of Hp and cardiac output between cycles can be estimated. If only thermally induced changes in Hp and cardiac output were responsible for the observed patterns, we would have expected a gradual increase in k with the rise in T_a from cycle 1 to cycle 3. Also, k should have changed maximally during the heating phases when both Hp and cardiac output would synergistically augment warming. Thus, because the k values for heating in cycle 2 are not intermediate to those of cycles 1 and 3, and because the most dramatic effects are observed during cooling (when Hp and cardiac output act in opposition, Figs 2, 3, 4 and Table 1), changes in k are due to the active modulation of heat flux.

We interpret the thermal notches as evidence for the active regulation of heat transfer. Fig. 4 suggests that during the warming phase (i.e. T_b<T_a, and T_x is negative), heat transfer was increased to allow the influx of ambient heat to augment warming. Then, when T_a was dropped (T_x became positive), the fish was momentarily trapped in a ‘high-k mode’ which led to the rapid but short-lived fall in T_b. To prevent continued heat loss at the lower T_a in cycles 1 and 2, heat flux was abruptly curtailed, leading to the levelling off and subsequent increase in T_b. The absence of a notch in cycle 3 indicates that thermoconservation was not increased after T_a was reduced and that heat transfer is maximized at high T_a.

Given the constraints of equation 1, and because Hp should be constant over this short time span, the best explanation for the ‘notch’ is the active modulation of heat transfer. If this is the case, how can the subsequent rise in T_b be explained? The most parsimonious explanation for the rise in T_b is that the rapid induction of thermoconservation curtailed convective heat loss sufficiently to elevate T_b. Also, with the heat exchangers ‘turned off’ during the warming phase, the abrupt onset of cool water may have channelled unheated blood directly into the yellowfin’s red muscle to cause localized convective cooling. Because the adjacent white muscle is also warm but less vascularized (Carey, 1973; Bone, 1978; Dickson, 1988), its temperature would not decline as rapidly. Thus, conductive heat flux from the white to the red muscle may have also contributed to the elevation of T_b.

The dramatic reversal of dT_b/dt suggests that tuna can control heat transfer to permit the influx of ambient heat if T_a>T_b. This confirms the contention of Holland et al. (1992), who reported that the heating and cooling k values, calculated during rapid vertical excursions of a bigeye tuna, were different.

Fig. 5 demonstrates that while changes in U can affect T_b, the magnitude of the change is dependent on T_a. The extent that an increase in U (=Hp) affects T_b is determined by the level of counter-current heat exchange. Thus, the slight increase in T_b at 30°C in comparison with that at 20 and 25°C verifies that thermoconservation is reduced at 30°C. That heat transfer is increased at high T_a is also indicated by the elevated T_x at both 25 and 20°C and by the ‘thermal notch’ that occurred when T_a was dropped from 30°C to 25°C. These findings corroborate earlier conclusions that heat is lost more rapidly at high T_a and show that the risk of overheating during high activity can be alleviated.

Our results warrant consideration of the possible mechanisms used by yellowfin to modulate heat transfer. Among the factors affecting k (i.e. internal and external conductive and convective processes), conductive heat loss is dependent upon fixed variables (surface area, tissue properties affecting molecular heat diffusion) and is thus not subject to short-term regulation. Convective heat loss is, by contrast, affected by blood flow, and rapid circulatory modifications are employed by many animals to alter heat flux. Also, changes in latency, which in this study show the same thermal dependence as the k values, are suggested to reflect variations in blood flow (Fechhelm and Neill, 1982). Thus, it is likely that alterations in blood flow to or through the retia are used to regulate heat transfer. There is presently no anatomical evidence for vascular shunts that affect tuna heat transfer. It is, nevertheless, possible to obtain insight into this question by comparing the general circulation patterns of tuna heat exchangers.

In addition to the yellowfin, there are 12 other endothermic tuna species in the scombrid tribe Thunnini. It is therefore useful to consider how factors such as retial configuration (which varies considerably) may influence both T_b and thermoregulatory capabilities. Among the Thunnini, there is a range of reliance on the central and lateral retia (Fig. 1B; Graham, 1975). At one extreme, Auxis and Euthynnus rely primarily on a central rete and have only one small pair of lateral retia. At the other extreme, the three species comprising the ‘bluefin group’ of Thunnus (T. thynnus, T. alalunga and T. maccoyii), as well as the bigeye tuna (T. obesus), have two pairs of lateral retia and have completely lost the central rete (Graham, 1975; Stevens and Neill, 1978). The yellowfin tuna is intermediate in this morphology, having a central rete and two pairs of well-developed lateral retia. It has been speculated that the frequent occurrence of the bluefin-group species and the bigeye tuna in cool waters at high latitude and below the thermocline relates to their sole reliance on the lateral retia (Graham, 1975; Stevens and Neill, 1978).

The results of preliminary studies with kawakawa (Euthynnus affinis) support the hypothesis that the lateral heat exchangers enhance thermoregulatory capabilities. In experiments similar to those described for yellowfin, we found the mean k for warming for three kawakawa to be 0.04 min⁻¹ (N=7), which is less than half that for the yellowfin (Table 1). This reduced capacity for regulating heat flux may be attributable to a primary dependence on blood flow through the dorsal aorta and post-cardinal vein and the adjacent central rete.

Three facts demonstrate the ability of yellowfin tuna to regulate heat transfer in response to changes in T_a and the thermal gradient. First, the abrupt increase in heat loss between cycles 1 and 2 and cycle 3, and the maximal changes during cooling (Figs 2, 3, 4 and Table 1), are indicative of active control of heat flux rather than passive changes in Hp and cardiac output. Second, the thermal notches (Figs 4 and 5) indicate that both T_a and the direction of the thermal gradient trigger a thermoregulatory response. Finally, the ability to regulate k is confirmed by the differential effects of U (Hp) and T_a on steady-state and non-steady-state T_x (Fig. 5).

Tuna, like homeothermic vertebrates, affect T_b through alterations of both k and Hp. However, while Hp can be altered independently of locomotion in birds and mammals, it is logical to assume that Hp cannot be controlled independently of swimming speed in tunas. Thus, the physiological, energetic and hydrodynamic constraints on tuna swimming limit the contribution of Hp to the control of T_b. Consequently, regulating heat balance is dependent on altering heat transfer, most probably through changes in retial blood flow. The physiological modulation of heat retention is thus a powerful complement to behavioural thermoregulation in tunas. Enhanced warming during an ascent in concert with reduced cooling during the subsequent descent will augment foraging time at depth. An elevated heat flux will also reduce the risk of overheating. Although tunas have variable T_b and are only regionally endothermic, physiological thermoregulation reduces potential limitations to distribution and activity and thus facilitates the effective exploitation of cooler waters.

We thank Drs J. R. Hunter, R. H. Rosenblatt, R. Shabetai, R. E. Shadwick and Chin Lai for their valuable comments on this manuscript. We also thank W. Crawford, S. Zimmermann, P. Fields, S. Arce, C. Lowe, G. Spencer, D. Reisinger, K. Korsmeyer and the staff at the NMFS Kewalo Basin Research Facility for assistance. This research was supported by the US National Science Foundation (OCE-89-15927 and 91-03739). Supplemental funding was provided by the William H. and Mattie Wattis Harris Foundation, the Academic Rewards for College Scientists Foundation, Inc., the Maurice Massarini Charitable Trust and the SIO Tuna Endowment Fund. We thank the Southwest Fisheries Science Center NMFS for facilitating the Kewalo Basin research.