Rate-Dependent Changes in Cell Shortening, Intracellular Ca²⁺ Levels and Membrane Potential in Single, Isolated Rainbow Trout (Oncorhynchus Mykiss) Ventricular Myocytes

Cardiac output is the product of heart rate and stroke volume. In fish, cardiac muscle contractility is known to influence stroke volume and is, in turn, heavily influenced by heart rate (Farrell and Jones, 1992). In ventricular strips from rainbow trout and a variety of other teleost species, an increase in stimulation frequency has been reported to decrease (Driezdic and Gesser, 1985, 1988; Bailey and Driezdic, 1990; Hove-Madsen, 1992; Shiels and Farrell, 1997) or increase (Hove-Madsen and Gesser, 1989; Hove-Madsen 1992; Matikainen and Vornanen, 1992) the isometric force of contraction.

Mechanisms that underlie the contraction/frequency relationship in fish cardiac muscle may be related to the process of excitation–contraction coupling. Our understanding of excitation–contraction coupling in teleost hearts is incomplete (see Tibbits et al., 1992). Cardiac muscle contractility is largely determined by the amplitude of the systolic intracellular Ca²⁺ ([Ca²⁺]_i) transient (Lewatowski and Pytkowski, 1987; Bers, 1991). In mammals, most of this activating Ca2+ originates from the sarcoplasmic reticulum and is released via Ca²⁺-induced Ca²⁺ release (CICR; e.g. Fabiato, 1985). However, Ca²⁺ can also enter the sarcoplasm through the sarcolemma, via the L-type Ca2+ current (I_Ca) and possibly the Na⁺/Ca²⁺ exchanger (Bers, 1991). The relative contribution to the systolic [Ca²⁺]_i transient varies with species (Negretti et al., 1993; Cleeman and Morad, 1991; Tibbits et al., 1991; Varro et al., 1993).

Previous studies in teleost fish suggest that release of Ca²⁺ from the sarcoplasmic reticulum is not necessary to activate contraction, which is instead initiated by a large trans-sarcolemmal Ca²⁺ influx (e.g. Tibbits et al., 1990). This is further supported by ultrastructural studies showing that fish cardiac muscle has a sparse sarcoplasmic reticulum and lacks t-tubules (Santer, 1985). In addition, studies on multicellular preparations have failed to show any effect of ryanodine (which impairs the function of the sarcoplasmic reticulum Ca²⁺-release channels) on contraction under physiological conditions (Dreidzic and Gesser, 1988; Hove-Madsen and Gesser, 1989). Furthermore, patch-clamp analysis of I_Ca currents indicates that the trans-sarcolemmal Ca²⁺ influx may contribute significantly to the activation of contraction in carp heart muscle (Vornanen, 1997, 1998). However, other recent studies on trout atrial and ventricular myocytes concluded that I_Ca was not sufficient to activate contraction fully (Hove-Madsen and Tort, 1998). This finding, together with the measurement of sarcoplasmic reticulum Ca²⁺ uptake rates, suggested that, in trout, the sarcoplasmic reticulum may be capable of participating in Ca²⁺ regulation during excitation–contraction coupling (Aho and Vornanen, 1998; Hove-Madsen et al., 1998).

It is possible that an increase in stimulation frequency may alter the shape and duration of the action potential. Such changes have been shown to be a major determinant of the force/frequency relationship in rabbit, guinea pig and rat cardiac muscle (Szigligeti et al., 1996). An increase in stimulation frequency may also influence the levels of [Ca²⁺]_ivia effects on I_Ca (Li et al., 1999) and/or the sarcoplasmic reticulum (Smith et al., 1988). Hove-Madsen et al. (1998) demonstrated a similar effect of stimulation frequency on I_Ca in fish. Shiels and Farrell (1997) concluded that the contribution of the sarcoplasmic reticulum to activating [Ca²⁺] was inversely proportional to pacing frequency at 22 °C, but that ryanodine was without effect at stimulation frequencies above 0.2 Hz at 12 °C. Changes in membrane ionic currents and [Ca²⁺]_i are likely to affect and to be affected by changes in the time course of the action potential (see Boyett et al., 1993).

The aim of the present study was to characterise the effects of changes in stimulation frequency upon contractility, [Ca²⁺]_i transients and action potential configuration in ventricular myocytes isolated from rainbow trout (Oncorhynchus mykiss) heart. To our knowledge, the effects of stimulation frequency upon action potentials and [Ca²⁺]_i transients in fish cardiac muscle have not previously been reported. By measuring these variables within a single system we were able to compare their magnitude and temporal changes and thus gain greater insights into the interactions between the mechanisms involved. A preliminary account of this work has been presented in abstract form (Harwood et al., 1998a).

Female rainbow trout [Oncorhynchus mykiss (Walbaum)] (140±5 g, mean ± S.E.M., N=20) were purchased from Washburn Valley Trout Farm, North Yorkshire, UK. They were held for up to 12 weeks in indoor, 2 m diameter, fibreglass tanks containing filtered, recirculated aerated fresh water. The water temperature was maintained at 12–15 °C. The trout were exposed to a 16 h:8 h L:D photoperiod and were fed commercial trout pellets ad libitum.

Ventricular myocytes were obtained by enzymatic digestion using a modified version of the method described by Vornanen (1998) and techniques for isolating mammalian cardiac myocytes (e.g. White et al., 1995). Trout were killed by a sharp blow to the head, and the spinal cord and brain were destroyed. The heart was carefully removed and placed in a Petri dish containing a nominally Ca²⁺-free Tyrode solution (for composition, see below). A glass cannula was inserted through the bulbus arteriosus into the ventricle (Moon et al., 1996) and held in place by suture thread. It was found to be important to penetrate the ventricle to obtain a good yield of cells. The atrium was cut to prevent fluid build-up in the ventricle (Milligan, 1994). The heart was mounted on a Langendorff perfusion apparatus and perfused from a height of 10 cm with a nominally Ca²⁺-free, low-Na⁺ Tyrode solution to lower intra- and extracellular [Ca²⁺] and to close intercalated disc gap junctions. This initial perfusion lasted for 7 min and was at a flow rate of 11 ml min⁻¹. The composition of the Ca²⁺-free Tyrode solution was (in mmol l⁻¹): NaCl, 100; KCl, 10; KH₂PO₄, 1.2; MgSO₄, 4; taurine, 50; pyruvate, 20; Hepes, 10; pH was adjusted to 6.9 with KOH. Milli-Q ultrapure water was used because water quality greatly affects the yield of myocytes. The heart was then perfused for 40 min with a recirculating, Ca²⁺-free Tyrode solution additionally containing proteolytic enzymes to dissolve intracellular connective tissue (0.75 mg ml⁻¹ collagenase, Worthington type II, activity 200–230 units mg⁻¹; 0.1 mg ml⁻¹ protease, Sigma type XXIV), 1.0 mg ml⁻¹ bovine serum albumin (BSA) and 32 μmol l⁻¹ Ca²⁺.

Following superfusion, the bulbus arteriosus and the atrium were removed. The ventricle was placed in Ca²⁺-free Tyrode solution and cut into small pieces, and the tissue was gently dispersed using forceps and repeated aspiration with a plastic Pasteur pipette. The suspension of cells was filtered through nylon gauze (pore size 250 μm) into a round-bottomed test-tube. The remaining tissue was resuspended, and the dispersion, aspiration and filtering processes were repeated. After filtration, the cells were allowed to settle to the bottom of the test-tube. The supernatant was poured off, and the cells were resuspended in the Ca²⁺-free Tyrode solution containing progressively higher concentrations of Ca²⁺ up to a final concentration of 750 μmol l⁻¹. The cells were stored in a Petri dish and refrigerated at 4–6 °C, the storage procedure used for mammalian myocytes in our laboratory. Cells were used within 8 h of isolation. All isolation procedures and subsequent experiments were carried out at 15±1.0 °C, using a Grant LTD water bath (Grant, Cambridge, UK) as a cooler.

The myocytes were allowed to settle on the glass coverslip forming the bottom of a chamber (volume 0.7 ml) mounted on the stage of a Nikon Diaphot inverted microscope (White et al., 1995). The cells were superfused with a control solution containing (in mmol l⁻¹): NaCl, 124.1; KCl, 3.1; CaCl₂, 2.5; MgSO₄, 0.9; pyruvate, 5.0; Tes sodium salt, 11.8; Tes free acid, 8.2; pH 7.8 at 15 °C), which is routinely used in experiments on in situ (e.g. Olson et al., 1994) and in multicellular (e.g. Harwood et al., 1998b) fish cardiac muscle preparations, thus giving consistency with previous experiments. Solution flowed through the chamber by means of gravity feed, and the level was controlled by suction. Cells were stimulated using 10 ms current pulses delivered by external platinum electrodes connected to a Grass (SD9) stimulator. Stimulation frequency was progressively increased in increments of 0.2 Hz from 0.2 to 1.4 Hz and then decreased back to 0.2 Hz in similar steps. This range of frequencies covers the physiological heart rates found in trout (Priede, 1974). Each rate was maintained until steady-state shortening was reached.

Cell shortening was measured on-line from a video image of the cell using an edge detection device (Crystal Biotech, Northborough, MA, USA) which sampled cell length at 50 Hz. The cell image could be optimally aligned for measurement by rotating a CCD camera attached to the side-arm of the microscope. Fish cardiac myocytes are long and narrow, making it difficult to record consistently from both edges of the cell. When this was possible, cell shortening was expressed as a percentage of resting cell length. When it was only possible to record from one edge of the cell, contractility was expressed in terms of the relative amplitude of contraction. The mean length of cells measured in this study was 143±7 μm (mean ± S.E.M., N=20 cells).

Changes in [Ca²⁺]_i were recorded in Fura-2-loaded myocytes using a spectrophotometer (Cairn Research, Faversham, UK) as previously described in detail by Frampton et al. (1991). Briefly, ventricular myocytes were loaded with the acetoxymethyl ester (AM) of the Ca²⁺-selective fluorescent dye Fura-2 (Molecular Probes, Eugene, OR, USA). To load the cells, 50 μg of Fura-2 AM was dissolved in 50 μl of dimethyl sulphoxide (DMSO) to give a 1.0 mmol l⁻¹ stock solution. A sample (6.25 μl) of this stock solution was then added to 2.5 ml of cells, giving a final Fura-2 concentration of 3.0 μmol l⁻¹. Cells were gently shaken for 4 min and then left to settle to the bottom of the test-tube for a further 6 min. The supernatant was removed, and the cells were resuspended and stored in a Tyrode solution containing 750 μmol l⁻¹ Ca²⁺.

Fura-2-loaded myocytes were alternately illuminated (every 2 ms) by 340 and 380 nm light using a rotating filter wheel. This excitation light was passed to the cell under study via a 430 nm dichroic mirror beneath the microscope nosepiece and a ×40 oil-immersion objective lens. The resultant fluorescence emission was collected by the objective lens and transmitted to the microscope side-port via a 580 nm dichroic mirror and a 510 nm emission filter for detection by a photomultiplier tube. A variable diaphragm was used to ensure that only fluorescence from the cell under study was collected. The output of the photomultiplier tube was passed to the spectrophotometer. The ratio of the emitted fluorescence at the two excitation wavelengths (340 nm:380 nm ratio) was calculated to give an index of [Ca²⁺]_i. Cells were also illuminated with red light, which passed via the diochroic mirror to the camera in the side-arm to give a video image of the cell.

To monitor membrane potential and membrane currents in fish myocytes (see Vornanen, 1997, 1998; Hove-Madsen and Tort, 1998), experiments were performed in patch-clamped myocytes using the whole-cell configuration. Experiments were performed using an Axoclamp-2B amplifier (Axon Instruments Inc.) controlled by a 1401-plus CED interface and software (Cambridge Electronic Design Ltd, Cambridge, UK). Glass pipettes with a resistance of 2.5–5.0 MΩ were pulled from non-heparinized haematocrit tubes using a vertical pipette puller (List Medical; type LM3PA). The pipette solution contained (in mmol⁻¹): potassium aspartate, 110; KCl, 10; NaCl, 10; MgCl₂, 8; K₂ATP, 8; Hepes, 10; EGTA, 0.05, pH 7.1. The level of intracellular Ca²⁺ buffering by this solution was sufficiently low that, under whole-cell conditions, cells contracted vigorously in response to triggered membrane depolarisations.

Gigaohm seals sometimes formed spontaneously on contact with the myocytes but more usually following application of suction to the pipette via a syringe. Once the whole-cell configuration had been achieved, action potentials were recorded in bridge mode in the absence of added hyperpolarising current. I_Ca was measured by voltage-clamping cells at −70 mV. The cells were then step depolarised to −40 mV for 150 ms to inactivate Na⁺ current, then depolarised to 0 mV for 500 ms to invoke inward I_Ca (e.g. White et al., 1995) (see Fig. 5). The amplitude of I_Ca, after steady state had been reached, was measured as the difference between the peak inward current and the current at the end of the depolarisation to 0 mV.

After steady-state [Ca²⁺]_i transients had been obtained, stimulation was stopped for 10 s and the cells were exposed to 10 mmol l⁻¹ caffeine using a rapid solution changing device (Levi et al., 1996). Briefly, cells were locally superfused with normal Tyrode solution via a small delivery tube positioned within 100 μm of the cell. This solution could be rapidly exchanged for a caffeine-containing solution. Caffeine is thought to release Ca²⁺ from the sarcoplasmic reticulum, and the amplitude of the induced [Ca²⁺]_i transient can be used as an index of sarcoplasmic reticulum Ca²⁺ content (Bassani et al., 1995). Electrical stimulation was resumed in the presence of caffeine until steady-state transients were achieved, before caffeine was washed out with control solution.

To complement the studies with caffeine, experiments were also performed with ryanodine and thapsigargin (Calbiochem). Ryanodine, at concentrations up to 10 μmol l⁻¹, irreversibly locks the Ca²⁺-release channels in a partially open (but sub-conductive) state (Rousseau et al., 1987). This facilitates a continuous leak of Ca²⁺ from the sarcoplasmic reticulum and prevents the sarcoplasmic reticulum from acting as a functional Ca²⁺ store. Thapsigargin is an irreversible inhibitor of the sarcoplasmic reticulum Ca²⁺-ATPase pump and will prevent sarcoplasmic reticulum Ca²⁺ uptake. After steady-state [Ca²⁺]_i transients had been obtained, ryanodine (10 μmol l⁻¹) and thapsigargin (2 μmol l⁻¹) were applied to the cell using the rapid switching method described above. Measurements were taken after the decline of the [Ca²⁺]_i transient had reached a stable level (usually within 5 min).

Analogue signals were passed to a Neuro-corder DR-890 (Cygnus Technology, Inc, PA, USA) A/D converter for storage on digital tape. All data are presented as means ± S.E.M. Statistical significance of the results was tested using Spearman rank order correlation, Mann–Whitney rank sum tests and Student’s t-tests (Sigmastat Statistical Software, SPSS) as appropriate. Significance levels were set to P<0.05. Normalised data are expressed relative to the value at 0.2 Hz.

We observed rate-dependent effects on both diastolic and systolic variables. Responses to increased stimulation frequency were observed to reverse when stimulation frequency was decreased. Fig. 1 shows, on a slow-time-base, recordings of cell length (Fig. 1A), [Ca²⁺]_i (Fig. 1B) and membrane potential (Fig. 1C) from representative ventricular myocytes. Increasing the stimulation frequency caused a progressive decrease in diastolic length (diastolic length was 154±9 μm at 0.2 Hz and decreased by 7.5±1.7 μm at 1.4 Hz). Consistent with the decrease in diastolic length was an increase in diastolic [Ca²⁺]_i (from a fluorescence ratio of 0.22±0.02 at 0.2 Hz to 0.26±0.03 at 1.4 Hz). The resting membrane potential also depolarised as stimulation frequency was increased (diastolic potential was −82.2±1.9 mV at 0.2 Hz and depolarised by 0.57±0.08 mV at 1.4 Hz) (N=7–9 cells for each of these variables). Spearman rank correlation on normalised data showed significant (P<0.05) rate-dependent decreases in diastolic length and resting membrane potential and a significant rate-dependent increase in diastolic [Ca²⁺]_i.

Fig. 2A shows averaged (up to 20 traces) fast-time-base recordings of single contractions from a trout ventricular myocyte. In this cell, an increase in stimulation frequency from 0.2 Hz to 1.4 Hz resulted in a graded decrease in cell shortening from a maximal amplitude of 14.3 μm at 0.2 Hz to 8.6 μm at 1.4 Hz. Mean cell shortening, expressed as a percentage of resting length, was 7.6±0.4 % at 0.2 Hz; this decreased to 5.5±0.4 % (N=5 cells) at 1.4 Hz.

Consistent with the rate-dependent decrease in cell shortening was a rate-dependent decrease in the amplitude of the [Ca²⁺]_i transient. Fig. 2B illustrates fast-time-base recordings of averaged [Ca²⁺]_i transients from a ventricular myocyte. Mean [Ca²⁺]_i transient amplitude fell from 0.075±0.012 at 0.2 Hz to 0.063±0.009 at 1.4 Hz (N=9 cells).

Fig. 2C illustrates more clearly how the decrease in contractility closely resembles the decline in the [Ca²⁺]_i transient as stimulation frequency is increased. The rate-dependent decreases in both contractility and [Ca²⁺]_i transient amplitude were significant (P<0.001, Spearman rank correlation on normalised data).

Rate-dependent changes in the time course of the contraction and [Ca²⁺]_i transients were also observed (Fig. 2A,B). Changing the stimulation frequency progressively from 0.2 to 1.4 Hz resulted in a significant decline (P<0.001; N=9 cells, Spearman rank correlation on normalised data) in the time from stimulation to the peak (t_peak) of both. Time to peak contraction decreased from 418±16 ms at 0.2 Hz to 346±10 ms at 1.4 Hz (N=9), while time to peak [Ca²⁺]_i decreased from 271±17 ms at 0.2 Hz to 157±9 ms at 1.4 Hz (N=9) (Fig. 3A). There was no significant change in the time taken for the cell to re-lengthen from maximal shortening to half diastolic length (⁠⁠); however, the time taken for the [Ca²⁺]_i transient to fall from peak to half diastolic level (⁠⁠) decreased significantly (P<0.001; Spearman rank correlation on normalised data) with increasing stimulation frequency (from 251±11 ms at 0.2 Hz to 186±6 ms at 1.4 Hz; N=9) (Fig. 3B).

Increasing the stimulation frequency stepwise from 0.2 to 1.4 Hz induced rate-dependent changes in the action potential duration that were consistent with the rate-dependent changes in contractility and [Ca²⁺]_i (see Discussion). Both the time taken from the peak of the action potential upstroke to 20 % repolarisation (APD₂₀) and the time taken from the peak of the action potential upstroke to 90 % repolarisation (APD₉₀) decreased significantly (P<0.01, Spearman rank correlation on normalised data) with increasing stimulation frequency (Fig. 4A). APD₂₀ was 110±26 ms at 0.2 Hz and decreased to 62±18 ms at 1.4 Hz (N=7), while APD₉₀ was 311±23 ms at 0.2 Hz and decreased to 241±36 ms at 1.4 Hz (N=7). Similarly, the relative action potential amplitude decreased significantly (P<0.01, Spearman rank correlation on normalised data) with increasing stimulation frequency (Fig. 4B). Action potential amplitude was 126±7 mV at 0.2 Hz and this decreased to 119±9 mV at 1.4 Hz (N=7 cells).

The overall effects of these changes on the shape of the action potential are illustrated in Fig. 4C, which shows fast-time-base recordings of action potentials from a trout ventricular myocyte. At 0.2 Hz, there was a long plateau phase during which a high membrane potential was maintained. When stimulation rate was increased to 1.4 Hz, the plateau phase became shorter and membrane potential fell more steeply.

The magnitude and time course of contraction of cardiac muscle are influenced by several factors, including action potential duration and [Ca²⁺]_i. These variables are, in turn, influenced by I_Ca. Because both action potential duration and [Ca²⁺]_i were found to be rate-dependent, the influence of stimulation frequency on I_Ca amplitude was also examined. Fig. 5A illustrates the protocol used to activate I_Ca. The membrane potential was held at −70 mV before being depolarised to −40 mV for 150 ms to inactivate the Na+ current and then further depolarised to 0 mV for 500 ms to activate I_Ca. Fig. 5B shows representative averaged (up to 20 traces) I_Ca recordings from a trout ventricular myocyte. The amplitude of I_Ca decreased significantly (P<0.01, Spearman rank correlation on normalised data) with increasing stimulation frequency (Fig. 5C) from 150±37 pA at 0.2 Hz to 116±32 pA at 1.0 Hz (N=3). These observations in trout ventricular myocytes are consistent with the recent findings in trout atrial myocytes by Hove-Madsen and Tort (1998).

The rate-dependent decrease in the amplitude of [Ca²⁺]_i transients might in part be due to decreased loading (and subsequent release) of Ca²⁺ from the sarcoplasmic reticulum. Fig. 6A,B illustrates the experimental protocol used to assess the effects of stimulation frequency on the Ca²⁺ content of the sarcoplasmic reticulum using 10 mmol l⁻¹ caffeine. The cell was stimulated to contract at 0.6 Hz in control solution, and [Ca²⁺]_i transients were measured. Stimulation was then stopped for 10 s, after which caffeine was applied to the cell. This resulted in a rapid rise in [Ca²⁺]_i that was greater in amplitude than the [Ca²⁺]_i transient elicited by steady-state electrical stimulation (Fig. 6A). Stimulation was resumed in the presence of caffeine, and the cell was then returned to the control solution (not shown in Fig. 6). Experiments were performed at 0.6 Hz and 1.0 Hz in the same cells.

Increasing the stimulation frequency from 0.6 to 1.0 Hz led to a decrease in the amplitude of the electrically stimulated Ca²⁺ transients but an increase in the amplitude of the caffeine-induced [Ca²⁺]_i transient (Fig. 6B). Similar results were obtained from three cells. Caffeine-induced Ca²⁺ release from the sarcoplasmic reticulum can be used as an index of sarcoplasmic reticulum Ca²⁺ content (Bassani et al., 1995). The ratio of the amplitude of electrically stimulated Ca²⁺ transients to the amplitude of caffeine-induced Ca²⁺ transients was significantly greater (P<0.05, Student’s t-test) at 0.6 Hz (0.67±0.07) than at 1.0 Hz (0.35±0.03) (Fig. 6C). These observations may indicate that, as stimulation frequency increases, sarcoplasmic reticulum Ca²⁺ load actually increases but that the fraction of this Ca²⁺ released by electrical stimulation falls (but see Discussion for an alternative explanation).

There has been much debate about the contribution made by the sarcoplasmic reticulum to the [Ca²⁺]_i transient in teleost fish. We therefore attempted to assess this contribution under our experimental conditions. [Ca²⁺]_i transients were evoked by electrical stimulation (0.6 Hz) in the presence and in the absence of 10 mmol l⁻¹ caffeine. In the presence of caffeine, the decline in the amplitude of the [Ca²⁺]_i transient was slowed (see representative [Ca²⁺]_i transients in Fig. 7) and the amplitude was reduced to 83.6±3.1 % (N=12) of the pre-caffeine control level. After washing out the caffeine, the decay of the [Ca²⁺]_i transient speeded up and the amplitude increased to 100.6±3.7 % of the pre-caffeine control level. In each of three cells stimulated at 1.0 Hz, the difference between the amplitude of the [Ca²⁺]_i transient in the presence of caffeine (90.7±0.9 % of pre-caffeine control) was less than that found at 0.6 Hz (9 % compared with 17 %).

To complement the caffeine experiments, ryanodine (10 μmol l⁻¹) and thapsigargin (2 μmol l⁻¹) were used in combination to disrupt sarcoplasmic reticulum function. Exposure of myocytes to ryanodine and thapsigargin caused a decline in the amplitude of the [Ca²⁺]_i transient (at 0.6 Hz) and, as seen in the caffeine experiments, the transient became slower (see representative [Ca²⁺]_i transients in Fig. 7). Visual inspection of the cells (although not measured) revealed a decrease in cell shortening. The amplitude of the [Ca²⁺]_i transient decreased significantly (P<0.01; Mann–Whitney rank sum test) to 86.7±1.3 % of the pre-caffeine control (a reduction of 13 %, Fig. 7), a result very similar to that found during the caffeine experiments.

Cell shortening, our index of contractility, is used routinely to assess contractility in mammalian myocytes (e.g. White et al., 1995). Studies comparing the responses to inotropic interventions (including stimulation frequency) in mechanically loaded and unloaded multicellular (Lee and Allen, 1989) and single myocyte (White et al., 1995) cardiac preparations have concluded that the responses are qualitatively the same but quantitatively smaller in the unloaded preparations.

Increasing the stimulation frequency caused an increase in diastolic [Ca²⁺]_i, possibly by reducing the time available for removal of Ca²⁺ from the myoplasm via sarcoplasmic reticulum uptake (see Aho and Vornanen, 1998; Hove-Madsen et al., 1998) and Na⁺/Ca²⁺ exchange extrusion (Hove-Madsen et al., 1998). This may be linked to a rate-dependent increase in intracellular [Na⁺] as described in mammalian myocytes by Harrison and Boyett (1995). It seems likely that the rise in resting [Ca²⁺]_i explains the frequency-dependent fall in diastolic cell length. The decrease in resting membrane potential was small and, we feel, is unlikely to be the cause of the rise in [Ca²⁺]_i (Fig. 1).

We have demonstrated a rate-dependent decrease in cell shortening as stimulation frequency is increased from 0.2 to 1.4 Hz, a range that includes physiological frequencies (Priede, 1974). This negative contraction/frequency relationship can be related to the negative force/frequency relationship seen in heart muscle from many teleost fish (e.g. Dreidzic and Gesser, 1985, 1988; Hove-Madsen, 1992; Shiels and Farrell, 1997).

One possible explanation for a negative force/frequency relationship in multicellular or whole heart preparations is anoxia or metabolic substrate depletion in the deeper fibres (Koch-Weser and Blinks, 1963; Blinks and Koch-Weser, 1963; Schouten and ter Keurs, 1986). Previous work examining the effect of stimulation frequency on contractility in the fish heart has been carried out on such preparations (e.g. Dreidzic and Gesser, 1985, 1988; Farrell et al., 1989). However, it is very unlikely that this explains the findings in our experiments because teleost ventricular myocytes have a small diameter (8 μm; Vornanen, 1998) and single myocytes are not thought to suffer from such diffusion-related limitations.

The global cytosolic [Ca²⁺]_i transients we recorded are similar to those reported in mammalian and amphibian single myocytes (e.g. Bers, 1991). The responses of these [Ca²⁺]_i transients and contractility to changes in stimulation frequency suggest that the relationship between these transients and activating [Ca²⁺] is close despite the possible localisation of the myofilaments (Farrell and Jones, 1992) at the periphery of teleost myocytes. Consistent with the effects on cell shortening, there was a statistically significant decrease in the amplitude of the [Ca²⁺]_i transient as stimulation frequency was increased from 0.2 to 1.4 Hz. The time course of the contraction and the [Ca²⁺]_i transient changed in a similar way when frequency was increased. We found significant rate-dependent decreases in t_peak of the contraction and [Ca²⁺]_i transient. There was also a significant abbreviation of in the [Ca²⁺]_i transient, although this effect only became prominent at frequencies above 1.0 Hz (Fig. 3B). Although the decrease in the of the contraction with frequency was not statistically significant, there was a trend towards more rapid relaxation at higher frequencies (1.0–1.4 Hz). From our observations, it seems likely that the rate-dependent decrease in cell shortening can be explained in terms of the rate-dependent decrease in the [Ca²⁺]_i transient amplitude.

Our measurements of action potential configuration in single cells are consistent with those previously recorded from multicellular preparations (e.g. Hove-Madsen and Gesser, 1989). We observed a rate-dependent reduction in the duration and amplitude of the action potential plateau (Fig. 4). These effects would be predicted to have several repercussions with respect to Ca²⁺ handling (Schouten, 1986; for a review, see Boyett et al., 1993). Earlier repolarisation would modulate the Na⁺/Ca²⁺ exchanger by increasing the rate of extrusion of Ca²⁺ (in Ca²⁺ extrusion mode), which could limit the amplitude of the [Ca²⁺]_i transient and speed the decay of the falling transient (Janvier and Boyett, 1996). Alternatively, it could decrease the rate of Ca²⁺ influx (in Ca²⁺ influx mode). Recent work by Vornanen (1999) has suggested that the Na⁺/Ca²⁺ exchanger may play a role in Ca²⁺ influx in teleost fish. It can be seen in Fig. 2B that the fall in the [Ca²⁺]_i transient amplitude appears to arise from an abbreviation of the rising phase of the transient. It might also be predicted that these effects would reduce sarcoplasmic reticulum Ca²⁺ loading (however, this was not the case; see below). A reduction in action potential duration would also be predicted to reduce Ca²⁺ influx via I_Ca as a result of a more rapid return to voltages at which the channels are not conducting Ca²⁺ (see below). The change in shape of the action potential is consistent with reductions in inward Na+ current (I_Na) and I_Ca. However, it is known that K⁺ currents are affected by changes in rate (e.g. Nabauer et al., 1996), and these could also explain our observations. Further characterization of the electrophysiology of teleost cardiac myocytes, particularly with respect to K⁺ channels, is required.

A rate-dependent decrease in the amplitude of I_Ca would contribute to a fall in the [Ca²⁺]_i transient. Under voltage-clamp conditions, we observed a rate-dependent decrease in I_Ca amplitude (Fig. 5). These results agree with those seen in trout atrial myocytes by Hove-Madsen and Tort (1998), who also found that the fall in the amplitude of I_Ca was associated with a fall in Ca²⁺ charge. The decline in I_Ca amplitude could be explained in terms of the rate-dependent increase in diastolic [Ca²⁺]_i, leading to Ca²⁺-dependent inactivation or an increase in steady-state inactivation (Lee et al., 1985).

The rate-dependent fall in the amplitude of the [Ca²⁺]_i transient could be explained by reduced loading (and subsequent release) of Ca²⁺ from the sarcoplasmic reticulum, and the reduction in action potential duration might be predicted to cause such a fall in sarcoplasmic reticulum load. However, the experiments using caffeine (Fig. 6) clearly show that, rather than being reduced, and despite a smaller electrically evoked [Ca²⁺]_i transient, the sarcoplasmic reticulum Ca²⁺ load is increased at higher stimulation frequencies. Thus, we can conclude that the reduction in the amplitude of the electrically evoked transients is not the result of a decreased sarcoplasmic reticulum load. A decrease in electrically evoked transients and increase in caffeine-induced transients is usually interpreted as a decrease in fractional sarcoplasmic reticulum Ca²⁺ release (Bassani et al., 1995). In mammalian cardiac muscle, the amount of Ca²⁺ released from the sarcoplasmic reticulum depends on the size of I_Ca, the trigger for release, and the sarcoplasmic reticulum Ca²⁺ load (Fabiato, 1985; Bassani et al., 1995). In trout ventricular myocytes, we observed a decrease in the amplitude of I_Ca, which might decrease the trigger for Ca²⁺ release. Another possibility is that sarcoplasmic reticulum Ca²⁺ channels are inactivated because of the higher diastolic [Ca²⁺]_i (Fabiato, 1985). However, we have not excluded the possibility that the reduction in the [Ca²⁺]_i transient is independent of the sarcoplasmic reticulum. It may be that it is simply the result of decreased Ca²⁺ influx because the contributions of Ca²⁺ influx and sarcoplasmic reticulum Ca²⁺ release to the triggered Ca²⁺ transient were not directly measured.

There has been much debate about the role of the sarcoplasmic reticulum in generating [Ca²⁺]_i transients in teleost species. Differences in the findings of previous studies may well be the result of differences in experimental conditions such as temperature, stimulation frequency and tissue type. Under the conditions of our experiments (15 °C, 0.6 Hz), disruption of sarcoplasmic reticulum function with either caffeine or ryanodine and thapsigargin led to a reduction in amplitude of the electrically stimulated transients of approximately 15 %. In each of three cells tested, the fall in the amplitude of the [Ca²⁺]_i transient in response to prolonged caffeine exposure was smaller at 1.0 Hz than at 0.6 Hz. One interpretation of these observations is that the relative contribution to the [Ca²⁺]_i transient made by the sarcoplasmic reticulum may be decreasing with increasing stimulation frequency, and this is consistent with the observations of Sheils and Farrell (1997), who concluded that the sarcoplasmic reticulum is less important at higher frequencies in trout trabeculae at 12 °C.

These observations might suggest that the sarcoplasmic reticulum makes only a small contribution to the [Ca²⁺]_i transient, in agreement with Vornanen (1998) and Sheils and Farrell (1997, 1998), but in contrast to recent measurements of sarcoplasmic reticulum Ca²⁺ uptake in trout ventricular muscle that suggest that the sarcoplasmic reticulum contributes significantly to the activation of contraction (Hove-Madsen et al., 1998). However, our experiments should be interpreted with some caution and may underestimate the role of the sarcoplasmic reticulum in generating the [Ca²⁺]_i transient. In addition to its effects on the sarcoplasmic reticulum, caffeine inhibits the enzyme phosphodiesterase (for a review, see O’Neill et al., 1993) which, in turn, may increase cyclic AMP levels and the amplitude of I_Ca and subsequently [Ca²⁺]_i transients. Using ryanodine and thapsigargin to inhibit sarcoplasmic reticulum function could also lead to an underestimation of the importance of the sarcoplasmic reticulum if compensatory mechanisms (such as an increase in I_Ca) occur when the sarcoplasmic reticulum is inhibited (Bassani et al., 1995). In addition, low temperatures may reduce the action of ryanodine and thapsigargin. The temperature-dependence of ryanodine binding in cardiac muscle is not known, but it has been shown that ryanodine binds to bullfrog skeletal muscle sarcoplasmic reticulum Ca²⁺ channels with a lower affinity at colder temperatures (Ogawa and Harafuji, 1990). Furthermore, the ‘on’ rate constant of thapsigargin shows a temperature-dependence indicative of a conformational change in the protein (Davidson and Varhol, 1995).

In conclusion, we have found that, at 15 °C, single trout ventricular myocytes demonstrate a negative contraction/ frequency relationship, decreases in the amplitude of the systolic [Ca²⁺]_i transient, in action potential duration and in the amplitude of I_Ca, but an increase in the Ca²⁺ load of the sarcoplasmic reticulum. We conclude that the rate-dependent fall in contractility is due to the fall in the amplitude of the [Ca²⁺]_i transient. Reduced influx of Ca²⁺via I_Ca may directly influence the [Ca²⁺]_i transient or trigger reduced sarcoplasmic reticulum Ca²⁺ release. Reduced sarcoplasmic reticulum Ca²⁺ release as a consequence of reduced sarcoplasmic reticulum Ca²⁺ load is not thought to be a contributing mechanism.

The authors would like to thank Dr Iain Young for helpful comments on the manuscript. C.L.H was supported by a BBSRC Studentship.