Acute heat tolerance of cardiac excitation in the brown trout (Salmo trutta fario)

All biological functions have strict thermal limits, making environmental temperature a decisive factor in geographical distribution of animal species (Precht et al., 1955). During more than 500 million years of evolution, fishes have experienced changes in ambient temperature, which have resulted in adaptations to a wide variety of thermal niches. Depending on the extent of temperature specialization, fishes can be classified as steno-, meso- and eurythermal, having narrow, moderate and wide thermal tolerance range, respectively (Beitinger and Bennett, 2000). Fishes living in stenothermal environments are usually specialists and tolerate only a narrow range of temperatures, the most striking examples being those of the Southern Ocean (Verde et al., 2006). Many freshwater fishes of north-temperate latitudes experience large seasonal temperature changes and are therefore adapted to operate under a wider range of temperatures, which extends from freezing point up to ~40°C (Horoszewicz, 1973; Bennett and Beitinger, 1997; Ford and Beitinger, 2005). Practically all such fishes tolerate freezing winter waters, although the upper thermal tolerance limit varies considerably among species. Salmonid fishes (family Salmonidae) usually prefer cool habitats and show an upper thermal tolerance range of 22–28°C. For example, brown trout (Salmo trutta fario Linnaeus), which manage best in cool waters with high oxygen content, have an upper incipient lethal temperature of 22–25°C, and are therefore classified as mesothermal fish (Elliott and Elliott, 2010).

Although thermal dependences of various molecular and cellular processes are known, factors that set the ultimate thermal tolerance limits of ectotherms are still poorly understood. Evolutionary thermal adaptation is expressed in mitochondrial volume density, membrane lipid composition, metabolic enzyme kinetics, functions of membrane transporters and contractile proteins (Hazel and Williams, 1990; Somero, 1995; Johnston et al., 1998). Interestingly, the limits of thermal tolerance appear first at the level of intact animals, and only later in the function of tissues, cells and molecules (Lagerspetz, 1987). For example, heat tolerance of proteins and lipid membrane structure are often higher than the upper thermal tolerance of the whole organism (Cossins and Prosser, 1978; Hochachka and Somero, 1984; Somero, 1995). Although thermal disturbances first appear in higher level functions, they ultimately reflect temperature-related deterioration or suboptimal function of some cellular and molecular processes, or mismatch of linked physiological processes.

Recent findings from both invertebrate and vertebrate ectotherms suggest that heart function could be a limiting factor for upper thermal tolerance of animals (Stillman and Somero, 1996; Seebacher et al., 2005; Farrell, 2009). Similarly, the hypothesis of an oxygen-limited thermal tolerance suggests that the circulatory system is one of the key factors in setting thermal tolerance limits of ectotherms (Frederich and Pörtner, 2000). Therefore, it would be interesting to know which molecular mechanisms might be limiting for the heat tolerance of ectothermic hearts.

Ion channels of fish cardiac myocytes are flexible entities that strongly respond to chronic temperature changes (Haverinen and Vornanen, 2004; Hassinen et al., 2007; Hassinen et al., 2008b; Haverinen and Vornanen, 2009), implying that they are intimately involved in thermal tolerance and temperature acclimation of cardiac function. Species-specific differences in channel composition and subunit assemblies may provide different thermal dependencies to electrical excitability of the fish heart. For example, the inward rectifier potassium current, I_K1, which is responsible for maintaining the negative resting membrane potential, is formed by different Kir2 channels in different species (Hassinen et al., 2007; Hassinen et al., 2008b). In cardiac myocytes of rainbow trout (Oncorhynchus

mykiss) the I_K1 is produced by Kir2.1 and Kir2.2 channels, while in cardiac myocytes of crucian carp (Carassius carassius) the same current is generated by Kir2.2 and Kir2.5 channels. The absence of Kir2.5 channels in trout is probably the reason why cold-acclimation decreases I_K1 in rainbow trout heart, while crucian carp myocytes show cold-induced increase of the I_K1 (Hassinen et al., 2007; Hassinen et al., 2008b).

Electrical excitation of the heart involves a range of molecular mechanisms that potentially could be responsible for thermal limitation of fish cardiac function. Therefore, the aim of the current study was to examine whether ion channel function of the brown trout heart is sensitive to temperatures close to the upper thermal tolerance limit of the fish, and therefore represent a limiting factor for electrical excitability of the heart. To this end acute thermal tolerances of ion currents of cardiac myocytes, contractility of spontaneously beating sinoatrial preparations and in vivo electrocardiograms were compared. Based on thermal sensitivity of cardiac function in other ectotherms, we hypothesized that electrical excitability of brown trout heart is vulnerable to temperatures that lead to heat death of the fish, and that this can be attributed to thermal deterioration or suboptimal function of one or more ion currents.

In vivo heart rate (HR) of brown trout was similar or slightly lower than measured previously for this species (Altimiras et al., 2002). Acute temperature changes had a profound effect on the electrocardiogram (ECG) of brown trout. From time-domain analysis, in vivo HR showed a curvilinear increase with temperature, and the interbeat (R–R) interval showed a linear decrease (Q₁₀=2.17; −87.0±3.0 ms °C⁻¹; Fig. 1). We used the R–T (instead of the Q–T; for definitions, see supplementary material Fig. S1) interval as an index of ventricular action potential (AP) duration, due to the waveform of fish ECG preventing a consistent identification of Q-wave position, which also showed a linear change with temperature (Q₁₀=2.17; −29.2±0.9 ms °C⁻¹; Fig. 1). Similar temperature dependencies of R–R and R–T intervals suggest that systolic/diastolic duration remains fairly constant under acute temperature changes.

When approaching the upper thermal tolerance limit of the animal, irregularities appeared in cardiac rhythmicity and HR started to decline at BPT of 23.5±0.6°C. Arrhythmicity first appeared as increasing heterogeneity of interbeat intervals and later on as bursts of repetitive activity (supplementary material Fig. S1). The break-up of normal cardiac rhythm was obvious in the parameters of HR power spectra (BPT=21.6±5°C; Fig. 2). Power spectral analysis of a broader selection replicated the HR versus temperature relationship, and showed a consistent thermal sensitivity among animals (NN=−82.7±3.0 ms °C⁻¹; supplementary material Fig. S2). HR variability followed a similar trend to that of NN versus temperature (−8.68±0.92 ms °C⁻¹; Fig. 2A). Coefficient of variation, a normalized measure for variability of inter-beat intervals, showed a transient increase as animals approached their BPT (Fig. 2B, circled data), but thereafter declined (overall thermal sensitivity of −0.14±0.09 ms °C⁻¹). Although total spectral power varied among individual fish, there was a clear trend of an inverse relationship with temperature (Fig. 2C). The relative sympathovagal balance also varied among fish, with two showing a relative thermal insensitivity of low frequency/high frequency ratio and two showing a gradual rise until the BPT, after which all fish showed a clear upward shift (Fig. 2D, circled data points). With further increases in temperature, spectral analysis became difficult due to electromyograph interference from respiratory muscles, and unreliable due to heterogeneity in inter-beat intervals. This was evident around the BPT, clearly seen as deviations in the Poincare plots and period histograms (supplementary material Fig. S1).

One obvious change in ECG was an abrupt increase in QRS complex duration at high temperatures. First, the duration of the QRS slightly reduced with rising temperature up to the BPT of 21.9±2.2°C, after which it strongly increased (Fig. 3). Duration of the QRS complex is a measure for the velocity of impulse transmission over the ventricle and therefore broadening of the QRS complex indicates depression in the rate of AP spread over the heart.

Spontaneously beating sinoatrial preparations were used to determine temperature tolerance of cardiac contractility (Fig. 4). The intrinsic HR increased linearly from 25±5 beats min⁻¹ at 3°C to a maximum of 124±6 beats min⁻¹ at 25°C (overall Q₁₀=2.15), then declined at higher temperatures and often completely ceased if temperature was not immediately lowered. The BPT for intrinsic HR was 25.8±0.6°C (Table 1). The force of atrial contraction decreased linearly with rising temperature up to a similar BPT (25.6±0.7°C) and then increased at higher temperatures, i.e. the temperature-related decline in force of contraction was partly due to the negative force–frequency relationship. Atrial pumping capacity (product of rate and force of contraction) doubled between 3 and 25.7°C, and declined above the BPT (25.4±0.4°C). Kinetics of contraction accelerated in a curvilinear manner as a function of rising temperature up to the BPT, with little change or slight decline at higher temperatures.

Thus for an acute temperature rise, contractility of the sinoatrial tissue improves up to the BPT of ~25°C and declines at higher temperatures.

Temperature sensitivities of resting membrane potential (RMP) and AP were measured from enzymatically isolated ventricular myocytes. RMP increased linearly upon warming from −60.6±1.5 mV at 4°C to −89.1±3.4 mV at ~33°C (Fig. 5), being then essentially equal to the theoretical equilibrium potential of K⁺ ions (−86.9 mV). The BPT of RMP was 29.6±1.2°C. The amplitude of AP first increased with increasing temperature between 4 and 20°C, and then levelled off between 20°C and the BPT of 26.4±1.3°C. The duration of ventricular AP (APD₅₀) decreased with temperature in a curvilinear manner from 776±124 ms at 4°C to 36±6 ms at 36°C. The shortening of AP was much stronger at low temperatures (Q₁₀=3.20 between 4 and 19°C) in comparison with high temperatures (Q₁₀=2.16 between 19 and 36°C).

BPTs for AP amplitude, RMP and APD₅₀ of the ventricular myocytes were somewhat higher (26.4–29.6°C) than BPTs for atrial contractile parameters, suggesting that none of the three membrane potential parameters is directly causative for malfunction of sinoatrial contractility. A clear exception was the upstroke velocity of ventricular AP, which showed a much lower BPT (21.7±1.2°C) in comparison with any other parameter of ventricular AP or sinoatrial contractility (Fig. 5F). The thermal response curve of AP upstroke velocity had the shape of an inverted V, with a peak at the BPT and minimum values at 4 and 35°C.

Outward K⁺ currents are repolarizing, i.e. they promote shortening of APD. Similar to other teleosts the major cardiac K⁺ currents of brown trout cardiac myocytes are the inward rectifier current (I_K1) and the rapid component of the delayed rectifier current (I_Kr) (Vornanen et al., 2002) (Fig. 6). Temperature dependence of I_Kr was measured from atrial myocytes, where this current is larger in comparison with ventricular myocytes. Respectively, I_K1 was measured from ventricular myocytes where the I_K1 is much larger than in atrial myocytes.

The tail current density of I_Kr increased 6.3-fold between 4°C and BPT of 27.3±0.6°C (Q₁₀=2.22), while the outward I_K1 increased 2.22-fold between 4 and 32°C (Q₁₀=1.33; Fig. 6). Thermal tolerance of I_K1 was better than that of I_Kr, without a clear BPT at temperatures below 32°C. Collectively, these results show that heat tolerances of both I_Kr and I_K1 are much higher than the upper thermal tolerances of sinoatrial contractile parameters or ECG in vivo.

Density and current–voltage relationship of the nifedipine-sensitive L-type I_Ca is similar as in rainbow trout (Oncorhynchus mykiss) (Fig. 7) (Vornanen, 1998; Shiels et al., 2000). The density of I_Ca increased with an overall Q₁₀ of 1.7 between 10°C and a BPT of 30.1±0.5°C, above which I_Ca steeply declined. These findings indicate that I_Ca of the brown trout heart is fairly resistant to high temperatures.

I_Na of the brown trout was similar to that of the rainbow trout with regard to voltage dependence and current density (Fig. 7) (Haverinen and Vornanen, 2004). I_Na increased with an average Q₁₀ of 2.3 between 4°C and a BPT of 20.9±0.5°C, above which the current strongly decreased. The V-shaped temperature dependence curve of I_Na is a mirror image to the inverted V-shaped curve of AP upstroke velocity (Fig. 5F). I_Na density at 35°C is only slightly higher than at 5°C. The BPT of I_Na was significantly lower than the BPTs of I_Kr, I_K1 and I_Ca (P<0.05), and was the most sensitive molecular function to high temperatures among the measured electrical or contractile parameters of the brown trout heart.

To explain the thermal sensitivity of a particular organ function or an intact organism, temperature dependencies of the underlying molecular processes need to be described. Ideally, it would be illustrative to compare the BPT of an organ function with BPTs of all those molecular entities that generate it. Electrical excitability of the heart (ECG, AP, HR) is a thoroughly characterized higher level function and its underlying molecular entities (ion channels/currents) are also well known for fish cardiac myocytes (Vornanen, 1998; Haverinen and Vornanen, 2004; Hassinen et al., 2008b; Haverinen and Vornanen, 2009). Therefore, the fish heart offers a well-defined model system to test heat tolerance mechanisms of organ function in ectotherms.

In the present study, cardiac function was measured at different levels of biological organization ranging from the intact fish down to the molecular function of ion channels. Heat resistance of heart function was weakest in the intact animal; electrical activity of the brown trout heart in vivo started to deteriorate at a temperature around 21.6°C (HR variability), resulting in depression of HR above the BPT of 23.5°C. This is consistent with the findings from rainbow trout (Oncorhynchus mykiss) and Atlantic cod (Gadusmorhua), where HR started to fall 2–4°C below the incipient lethal temperature (Heath and Hughes, 1973; Gollock et al., 2006). The first signs of disturbed cardiac function in the ECG of brown trout appeared as a progressive increase in HR variability slightly below the actual BPT for HR, indicating that the heart was losing control over the inter-beat interval. Beat-by-beat variations in the ECG usually reflect efferent vagal (parasympathetic) activity. The lower in vivo HR in comparison with intrinsic in vitro HR suggests that the parasympathetic tone restrains HR in trout in vivo (Priede, 1974), while the increase in low/high frequency ratio, and reduced total power with increasing temperature, are consistent with a relative increase in sympathetic drive contributing to temperature-induced tachycardia (Wood et al., 1979). However, the relative increase in sympathetic tone is at odds with suggestions that heat-related reductions in HR represent vagal bradycardia, i.e. a parasympathetically controlled adaptive HR response against hypoxia or cellular Ca²⁺ overload (Heath and Hughes, 1973; Rantin et al., 1998). The increase in low/high frequency ratio could be sooner seen as a protective mechanism to maintain high HR and cardiac output, since isoprenaline, a beta-adrenergic agonist, not only increases HR but also provides protection against heat-dependent depression of HR in rainbow trout (Aho and Vornanen, 2001). It is unlikely that the abrupt increase in sympathetic drive at elevated temperature is the immediate cause of large HR variability, arrhythmicity of the heartbeat or depression of HR.

Another large temperature-related change in the ECG of brown trout was the widening of the QRS complex at high temperatures. This is indicative for heat-dependent depression of the velocity of impulse conduction over the heart, which could result in increases in HR variability and depression of HR. Experiments on isolated cardiac myocytes provide support for this hypothesis.

In vitro cardiac function, measured from spontaneously beating sinoatrial preparations was slightly more resistant to high temperatures than in vivo ECG. BPTs for the contractile parameters of the trout sinoatrial preparations centred on 25°C (range 24.3–25.8°C), probably due to interdependence of the measured parameters. Changes in beating frequency will affect force and time course of twitches (rise time, decay time, half-width) (Shiels et al., 2002), and therefore heat sensitivity of HR is reflected in all parameters of atrial contractility. This indicates the prime importance of HR in the temperature sensitivity of fish heart function. A temperature-dependent increase in HR is able to compensate for the simultaneous decrease in force development, but pumping capacity of the heart starts to decline when heat depresses HR.

As expected, increases in temperature were associated with decreases in AP duration (Talo and Tirri, 1991; Haverinen and Vornanen, 2009). This was evident both in vivo (R–T interval of ECG) and in vitro in isolated ventricular myocytes. There was no apparent limit to the shortening of AP in vitro, although reduction in APD₅₀ became slower at higher temperatures. Because the duration of the AP plateau is determined by the balance between outward (K⁺) currents and inward (Ca²⁺ and Na⁺) currents, shortening of APD with temperature indicates that at high temperatures the ion current balance changes in favour of repolarizing K⁺ currents. This is necessary for cardiac function, because shortening of AP plateau makes room for diastolic filling of the heart when cardiac cycles are abbreviated. Indeed, in vivo ECG suggests that the duration of ventricular AP occupies a similar fraction of the total cardiac cycle at all temperatures.

A temperature-dependent increase in K⁺ currents is also seen in hyperpolarization of RMP, which reaches its theoretical maximum at 29.6°C. Hyperpolarization of RMP has a stabilizing effect on electrical excitability, which may protect the heart against ectopic beats. However, at high temperatures the very negative RMP could make myocytes electrically unexcitable and completely prevent the heartbeat, in particular if the inward I_Na current is simultaneously reduced (see below).

In general, electrical activity of enzymatically isolated cardiac myocytes was more resistant to high temperatures than either in vivo ECG or in vitro contractility of sinoatrial preparations. As ECG is produced by composite activity of all ion channels of cardiac myocytes, heat-related disturbances in ECG must be due to temperature-dependent failure of one or more ion channels or imbalance between depolarizing or repolarizing ion currents due to their different temperature dependencies. The failure could occur either in the pacemaker cells that initiate the heartbeat or alternatively in the working atrial and ventricular myocytes or in the conducting pathway between the atrium and the ventricle.

Major ion current systems of the vertebrate heart are largely the same in ectothermic and endothermic vertebrates, although isoforms of channel proteins and molecular assemblies of multi-protein channel complexes may differ significantly. I_Na and I_Ca are the major inward currents, while I_K1, I_Kr and I_Ks are the main outward currents of fish cardiac myocytes (Vornanen, 1998; Haverinen and Vornanen, 2004; Hassinen et al., 2007; Hassinen et al., 2008a; Hassinen et al., 2011). In mammals, these currents operate optimally at the body temperature of 36–39°C. In most fishes, body temperature is less than in endotherms, and importantly, it is often highly variable. Consistent with this fact, ion currents of the brown trout cardiac myocytes (perhaps with the exception of I_K1) started to decline much below the body temperature of endothermic vertebrates. The lower thermal tolerance of brown trout ion currents/channels in comparison with those of endotherms is probably a trade-off between catalytic activity and thermal stability as a consequence of molecular adaptation to optimise function at the lower body temperatures of ectothermic fishes. Catalytic activity and thermal stability are mutually exclusive properties of a protein molecule, in that high catalytic activity at low temperature means low thermal stability at higher temperatures (Závodszky et al., 1998; Fields, 2001). Whether the relatively low heat tolerance is purely a property of ion channel proteins or is also dependent on the lipid environment of channels remains to be shown. Large variability in heat tolerance between different ion currents (see below) suggests that thermal properties of the bulk lipid membrane are not decisive. However, the contribution of the immediate lipid environment of the ion channel, the lipid annulus, cannot be excluded (Zheng et al., 2011).

The four ion currents of the brown trout heart measured in this study showed different heat resistances. The most resilient to high temperatures was I_K1 followed by I_Ca and I_Kr. I_Na had clearly the lowest threshold for temperature-dependent deterioration, with a 6.4°C lower BPT than the next heat-sensitive current, the I_Kr. Opening and closing of ion channels requires conformational changes in the ion channel proteins known as ‘gating’ between conducting and non-conducting states (Bezanilla, 2005). Similar to enzyme reactions, gating of ion channels is strongly dependent on temperature (Collins and Rojas, 1982). Considering the implicit kinetic compromise, the high sensitivity of I_Na for thermal inactivation may reside in its high catalytic activity. The kinetics of sodium channel gating is very fast, resulting in an almost instantaneous opening of the channels upon small membrane depolarization, followed by large Na⁺ influx and rapid inactivation during a maintained depolarization (Patlak, 1991). The high catalytic activity of Na⁺ channels probably requires high molecular flexibility, which may come with the trade-off of low thermal stability.

The BPTs of I_Na density and upstroke velocity of AP are in excellent agreement, differing only 0.8°C. Furthermore, the BPT of the velocity of impulse conduction in vivo also shows low heat tolerance. Because the density of I_Na is the main determinant for the velocity of impulse conduction, heat inactivation of I_Na is expected to compromise the rate of AP propagation over the heart. This is in keeping with the mammalian models, where the loss of cardiac Na⁺ channel function is associated with slowing of sinoatrial conduction and frequent sinoatrial or atrioventricular conduction blocks (Derangeon et al., 2012). Therefore, the large increases in HR variability, missed beats and depression of HR in brown trout at high temperatures could be due to slowed or impaired AP conduction between cardiac compartments rather than caused by heat inactivation of the impulse generation in the pacemaker centre. Indeed, the slightly better heat tolerance of the sinoatrial tissue in vitro suggests that conductive pathways and the ventricle are more sensitive to heat than the pacemaker tissue. Examination of temperature modulation of isolated pacemaker cells is needed to exclude putative contribution of direct heat inactivation of the pacemaker mechanism to thermal deterioration of cardiac contractility.

The threshold potential is the critical level to which the membrane potential must be depolarized in order to initiate an AP, i.e. at that voltage the density of the inward I_Na exceeds the total density of the outward K⁺ currents (mainly the I_K1). Therefore, large increases in repolarizing K⁺ currents or decrease in I_Na can result in AP failure (Huxley, 1959; Guttman, 1962; Golod et al., 1998; Rosenthal and Bezanilla, 2002). Under a rising temperature regime, a cardiac myocyte is expected to become electrically unexcitable if the Q₁₀ value of the repolarizing currents is greater than that of the depolarizing currents, i.e. if the total density of K⁺ currents increases faster than the density of I_Na. In this respect it is notable that in brown trout myocytes I_Na started to decline above 20.9°C, while I_Kr and I_K1 still continue to increase with temperature. With increasing temperature RMP becomes increasingly negative, while depolarizing power of I_Na decreases; the resultant imbalance of repolarizing and depolarizing currents prevents reaching the threshold potential, i.e. AP fails (Fig. 8). This may be the reason why heartbeat in brown trout completely ceases at temperatures slightly above 25°C.

This study shows that function of the cardiac I_Na is compromised at high temperatures, which might be a contributing factor to heat-dependent depression of cardiac contractility in the brown trout. However, Na⁺ channels are not restricted to cardiac myocytes, but occur in many other cell types, in particular in neurons and muscle cells. At least eight alpha subunits of voltage-gated Na⁺ channels exist in teleost fishes with different tissue distribution (Widmark et al., 2011). Therefore, it is possible that similar thermal sensitivity of I_Na as shown here for the trout cardiac myocytes, might also compromise Na⁺ channel function in nervous system and skeletal muscles with possibly dangerous consequences to behaviour, locomotion and other vital functions. The trout cardiac I_Na is mainly produced by the Na_V1.4a subunits (Haverinen et al., 2007), and it remains to be shown whether other Na⁺ channel isoforms show similar thermal sensitivity as the cardiac isoform. An interesting topic for further research is the molecular mechanism, which makes the fish cardiac Na⁺ channels sensitive to heat inactivation, i.e. which amino acid sequences or ion channel domains are involved in heat sensitivity of the ectothermic Na⁺ channels. The significance of lipid annulus around the channel protein also needs to be clarified.

The present findings show that ECG in vivo and contractility of sinoatrial preparations in vitro are in general more sensitive to temperature-dependent deterioration than ion channel function in single isolated myocytes. These findings are in agreement with the generalization that body functions at higher levels of biological organization (physiology) are more thermally sensitive than most cellular and molecular mechanisms upon which the higher level functions are based (Lagerspetz, 1987). In the brown trout heart, a clear exception to this rule is the upper thermal tolerance of the I_Na, which shows a lower BPT than the higher level functions, HR in vivo and HR in vitro.

Cardiac output in fishes increases with increasing temperature mainly via elevations in HR and with little or no changes in stroke volume (Gollock et al., 2006; Steinhausen et al., 2008; Mendonça and Gamperl, 2009). The current study suggests that the upper thermal tolerance of cardiac function in the brown trout might be set by heat sensitivity of the cardiac Na⁺ channels, because I_Na is thermally the weakest link in electrical excitation of the brown trout cardiac myocyte. Above 20.9°C impaired Na⁺ channel function results in depression of I_Na density, AP upstroke velocity and conduction spread over the heart with consequent disturbance in cardiac rhythmicity and depression of HR. With further increases in temperature, electrical excitability may be completely lost, because increasing amplitude of the I_K1 overwhelms the decreasing I_Na so that the threshold voltage for the regenerative opening of Na⁺ channels is not achieved with an outcome of cardiac standstill.

Experiments were conducted on cultivated brown trout (Salmo trutta fario) (113.3±10.8 g, N=37) that were obtained from a local fish farm (Kontiolahti, Finland). In the animal house of the university the fish were maintained in 500-litre metal aquaria at a water temperature of 12±1°C, and fed commercial food pellets (EWOS, Turku, Finland) five times per week. Photoperiod was a 12 h:12 h light:dark cycle. All experiments were authorized by the national Animal Experimental Board in Finland (permissions STH998A and PH472A).

ECG recordings were made as previously described (Campbell et al., 2004; Campbell et al., 2006). Trout were anaesthetized in tricaine methanesulfonate (MS-222, 0.3 mg l⁻¹, Sigma, St Louis, MO, USA) and placed ventral side up on an operating table, and the gills were irrigated with tap water. Recording electrodes (7-strand Teflon-coated wire, length 40 cm, diameter 0.23 mm; A-M Systems, Carlsborg, WA, USA) were hooked into the end of a 24-G hypodermic needle and obliquely inserted from the ventral surface at the level of the pectoral fins forward, close to the pericardium. The trailing wires were attached by a suture to the belly of the fish and by a second suture in the front of the dorsal fin. Whilst still docile, the fish was placed into a respiratory chamber (1 litre, initial O₂ content ~9 mg l⁻¹). Bipolar ECG signals were recorded using a bioamplifier (ML 136) interfaced with a digital recording system (PowerLab, ADInstruments, Colorado Springs, CO, USA). After implantation of electrodes the fish were allowed to recover from the operation for about 2 days before the thermal challenge. ECG recordings were started at a temperature of about 10°C, and temperature was raised at a rate of ~1.5°C h⁻¹ until disturbances appeared in the ECG, with uninterrupted recordings of ECG and temperature made throughout (LabChart, ADInstruments). The amplified signal was plotted in real time, and inter-beat intervals extracted from the raw trace, defined as the period between the R waves of successive heart beats.

An index of short-term HR variability in the time domain was calculated as the standard deviation of successive inter-beat intervals, with normalized variability given as the coefficient of variation, using manual identification of components from 20–30 consecutive beats. The underlying physiological control of heart rate was examined using power spectral analysis to reveal the frequency of oscillatory components due to autonomic modulation of intrinsic cardiac pacemaker activity, after converting a selected ECG trace into a tachogram of 256 consecutive inter-beat (R–R) intervals (supplementary material Fig. S2) (Campbell et al., 2004). A fast Fourier transformation was then applied using a Hanning window to minimize spectral leakage, and the resulting output plotted graphically. The ratio of low frequency to high frequency components of the spectra was calculated as an index of sympathovagal balance. Four fish gave sufficiently stable recordings across the whole temperature range.

Velocity of impulse conduction over the ventricle was determined from the width of the QRS complex at the zero voltage level. Q–T interval was used as a measure for the average duration of the ventricular AP.

Atrial and ventricular myocytes were isolated with enzymatic digestion using the same solutions and enzymes as in our original method for fish hearts (Vornanen, 1997). In brief, 7-min perfusion of the heart with Ca²⁺-free saline was followed by a 15-min perfusion with solution containing collagenase (Sigma Type IA; 0.75 mg ml⁻¹), trypsin (Sigma Type III; 0.5 mg ml⁻¹) and fatty acid-free bovine serum albumin (BSA; Sigma, 1 mg ml⁻¹). A small aliquot of dissociated cells were placed in a 150 μl chamber (RC-26, Warner Instruments, Hamden, CT, USA) mounted on the stage of an inverted microscope. Cells were allowed to adhere to the bottom of the chamber and then superfused continuously with the external solution precooled to 4±1°C. Voltage (ion currents, I) and current (action potentials, AP) clamp recordings were made using an Axopatch 1D amplifier (Axon Instruments, Saratoga, CA, USA) equipped with a CV-4 1/100 head-stage.

The external solution for AP and K⁺ current recordings contained (in mmol l⁻¹) 150 NaCl, 5.4 KCl, 1.5 MgSO₄, 0.4 NaH₂PO₄, 2.0 CaCl₂, 10 glucose, and 10 HEPES [4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid] at pH 7.82 at 4°C. This solution, without any blockers, was used for AP recordings. For measurement of K⁺ currents 1 μmol l⁻¹ tetrodotoxin, 10 μmol l⁻¹ nifedipine and 30 μmol l⁻¹ glibeclamide were added to block Na⁺, Ca²⁺ and ATP-sensitive K⁺ currents, respectively. The pipette solution for AP recordings contained (in mmol l⁻¹) 140 KCl, 5 Na₂ATP, 1 EGTA (ethylene glycol tetraacetic acid) and 10 HEPES at pH 7.2, and for K⁺ currents 140 KCl, 4 MgATP, 1 MgCl₂, 5 EGTA and 10 HEPES at pH 7.2. I_K1 was measured as a Ba²⁺ (0.2 mmol l⁻¹)-sensitive current and I_Kr as an E-4031 (N-[4-[1-[2-(6-methylpyridin-2-yl)ethyl]piperidine-4-carbonyl]phenyl]) (2 μmol l⁻¹)-sensitive current. Temperature sensitivity of the peak outward I_K1 was measured from currents elicited by 1-s voltage ramp pulses. Temperature dependence of I_Kr was measured for the peak tail current at −20 mV.

Na⁺ current was measured in Cs⁺-based, low-Na⁺ saline solution that contained (in mmol l⁻¹): 20 NaCl, 120 CsCl, 1 MgCl₂, 0.5 CaCl₂, 10 glucose and 10 HEPES (pH adjusted to 7.82 with CsOH). In addition, 10 mmol l⁻¹ nifedipine (Sigma) was added to both solutions to block L-type Ca²⁺ currents. The pipette solution contained (in mmol l⁻¹): 5 NaCl, 130 CsCl, 1 MgCl₂, 5 EGTA, 5 MgATP and 5 HEPES (pH adjusted to 7.2 with CsOH). First, the myocytes were perfused with normal K⁺-based saline so that gigaseal and whole-cell patch-clamp recording of the myocyte were established. Internal perfusion of the myocytes with pipette solution was continued in this solution for at least 3 min in order to allow buffering of intracellular Ca²⁺ with 5 mmol l⁻¹ EGTA. Then, solution flow could be switched to a low-Na⁺ external solution without inducing contracture of the patched myocyte. To ensure adequate voltage control a minimum of 80% series resistance compensation was applied. I_Na was elicited from the holding potential of −120 mV (Haverinen and Vornanen, 2004). Effect of temperature on I_Na was measured at the pulse potential of −30 mV.

The composition of the physiological solution used for recording of I_Ca was as follows (in mmol l⁻¹): 130 NaCl, 5.4 CsCl, 1.5 MgSO₄, 0.4 NaH₂PO₄, 1.8 CaCl₂, 10 glucose and 10 HEPES (adjusted to pH 7.82 with CsOH). The pipette solution contained (in mmol l⁻¹): 130 CsCl, 5 MgATP, 15 TEACl, 1 MgCl₂, 5 oxaloacetate, 5 EGTA, 0.03 Na₂GTP and 10 HEPES (adjusted to pH 7.2 with CsOH). The difference between the peak current and the current at the end of 300 ms depolarizing pulse from the holding potential of −60 mV to +10 mV was taken as I_Ca at different temperatures.

All current and voltage recordings were started at 4°C and then temperature was gradually raised at the rate of ~3°C min⁻¹ until ion current or membrane voltage showed evident decline. The temperature after which they started to decline is termed the break point temperature (BPT) and is regarded as the upper thermal tolerance limit of the corresponding membrane function. Effects of temperature on ionic currents are shown relative to the initial current density at 4°C, which was set as 1 (outward current) or −1 (inward current). Electrophysiological experiments were mainly performed on ventricular myocytes with the exception of the I_Kr current that was measured from atrial cells. I_Kr is much bigger in atrial than ventricular myocytes, while the opposite is true for I_K1 current. The dichotomy in the densities of I_Kr and I_K1 currents between the two cell types enables easy and clean measurement of I_Kr and I_K1 from atrial and ventricular myocytes, respectively.

For measuring intrinsic HR, force and time course of atrial contraction, sinoatrial preparations consisting of the sinus venosus and the whole atrium were dissected free and gently fixed from one atrial corner with insect pins on a Sylgard-coated 10 ml recording chamber filled with continuously oxygenated (100% O₂) physiological saline (in mmol l⁻¹): 150 NaCl, 3 KCl, 1.2 MgSO₄, 1.2 NaH₂PO₄, 1.8 CaCl₂, 10 HEPES and 10 glucose adjusted to pH 7.82 with NaOH at 4°C. From another corner the atrium was fixed via a small hook and a braided silk thread to the force transducer (Grass FT03, Grass Products, Warwick, RI, USA) (Vornanen, 1989). The muscle was slightly stretched and allowed to spontaneously beat at the intrinsic HR. Force signals were digitized (Digidata-1340 AD/DA, Axon Instruments) with a sampling rate of 2 kHz before storing on a computer with the aid of Axotape (Axon Instruments) acquisition software. HR and contractile variables were analysed with Clampfit software (Axon Instruments) and graphs were constructed in SigmaPlot. The preparation was allowed to equilibrate at 3°C for approximately 1 h to reach a stable beating rate before responses to rising temperature were determined. Temperature was gradually raised at the rate of ~1°C min⁻¹ until increasing HR clearly reversed direction or when the heartbeat ceased. BPT of HR contractile parameters was determined as in patch-clamp experiments.

Results are given as means ± s.e.m. Statistically significant differences (P<0.05) between different variables, obtained by each research method (in vivo ECG, in vitro sinoatrial contractility, current-clamp of single myocytes, voltage-clamp of single myocytes) were assessed using one-way ANOVA after checking normality of distribution and making necessary transformation of variables. Paired comparisons between two means were done by Tukey's honestly significant difference post hoc test.

Anita Kervinen is acknowledged for skilful technical assistance.