Functional coupling between glycolysis and excitation—contraction coupling underlies alternans in cat heart cells

Abstract

1. Electromechanical alternans was characterized in single cat atrial and ventricular myocytes by simultaneous measurements of action potentials, membrane current, cell shortening and changes in intracellular Ca²⁺ concentration ([Ca²⁺]_i).

2. Using laser scanning confocal fluorescence microscopy, alternans of electrically evoked [Ca²⁺]_i transients revealed marked differences between atrial and ventricular myocytes. In ventricular myocytes, electrically evoked [Ca²⁺]_i transients during alternans were spatially homogeneous. In atrial cells Ca²⁺ release started at subsarcolemmal peripheral regions and subsequently spread toward the centre of the myocyte. In contrast to ventricular myocytes, in atrial cells propagation of Ca²⁺ release from the sarcoplasmic reticulum (SR) during the small-amplitude [Ca²⁺]_i transient was incomplete, leading to failures of excitation-contraction (EC) coupling in central regions of the cell.

3. The mechanism underlying alternans was explored by evaluating the trigger signal for SR Ca²⁺ release (voltage-gated L-type Ca²⁺ current, I_{Ca, L}) and SR Ca²⁺ load during alternans. Voltage-clamp experiments revealed that peak I_{Ca, L} was not affected during alternans when measured simultaneously with changes of cell shortening. The SR Ca²⁺ content, evaluated by application of caffeine pulses, was identical following the small-amplitude and the large-amplitude [Ca²⁺]_i transient. These results suggest that the primary mechanism responsible for cardiac alternans does not reside in the trigger signal for Ca²⁺ release and SR Ca²⁺ load.

4. β-Adrenergic stimulation with isoproterenol (isoprenaline) reversed electromechanical alternans, suggesting that under conditions of positive cardiac inotropy and enhanced efficiency of EC coupling alternans is less likely to occur.

5. The occurrence of electromechanical alternans could be elicited by impairment of glycolysis. Inhibition of glycolytic flux by application of pyruvate, iodoacetate or β-hydroxybutyrate induced electromechanical and [Ca²⁺]_i transient alternans in both atrial and ventricular myocytes.

6. The data support the conclusion that in cardiac myocytes alternans is the result of periodic alterations in the gain of EC coupling, i.e. the efficacy of a given trigger signal to release Ca²⁺ from the SR. It is suggested that the efficiency of EC coupling is locally controlled in the microenvironment of the SR Ca²⁺ release sites by mechanisms utilizing ATP, produced by glycolytic enzymes closely associated with the release channel.

Mechanical and electrical alternans in cardiac muscle describes the cyclical beat-to-beat alterations in twitch amplitude and action potential duration (e.g. Wohlfart, 1982). Alternans occurs in cardiac failure and during myocardial ischaemia, and is believed to be an important factor in the pathogenesis of arrhythmias (Dilly & Lab, 1988; Konta et al. 1990). Clinically, alternans becomes manifest as pulsus alternans and specific changes in the electrocardiogram (alteration of the ST segment; see e.g. Uno, 1991; Surawicz & Fish, 1992; Lab & Seed, 1993). In cardiac muscle preparations or isolated myocytes alternans can be induced experimentally by high frequency stimulation during acidosis (Orchard et al. 1991), low extracellular calcium concentration (Badeer et al. 1967) and hypothermia (e.g. Spencer et al. 1992).

Despite the various conditions under which electromechanical alternans can be observed, the underlying molecular and cellular mechanisms remain elusive. There is strong evidence, however, that alternans is ultimately linked to alterations in myocardial Ca²⁺ homeostasis, since changes in contraction amplitude are paralleled by alternating changes in the amplitude of the underlying [Ca²⁺]_i transients (e.g. Lab & Lee, 1990; Kihara & Morgan, 1991; Orchard et al. 1991; Uno, 1991). Several studies have provided evidence that Ca²⁺ handling by the SR may play a crucial role in the impaired [Ca²⁺]_i regulation during alternans (Spencer et al. 1992; Hirayama et al. 1993; Rubenstein & Lipsius, 1995). Furthermore, conditions that favour the occurrence of alternans, such as ischaemia (e.g. Lee et al. 1988; Wu & Clusin, 1997) or acidosis (Orchard et al. 1991), are also directly related to changes in the metabolic state of cardiac myocytes and affect levels of intracellular ATP. Cardiac Ca²⁺ cycling such as Ca²⁺ uptake into the SR as well as the Ca²⁺ release process are ATP dependent. In cardiac cells over 90 % of ATP is produced by oxidative phosphorylation in the mitochondria. This process maintains bulk cytosolic ATP concentrations in the millimolar range (5–10 mM). The primary purpose of mitochondrial ATP production is to provide fuel for the contractile process. In recent years, however, evidence is accumulating that ATP derived from glycolysis is utilized preferentially by membrane ion transport mechanisms. Such functional microcompartmentation is brought about by the close physical association of glycolytic ATP-generating enzymes with the ATP-dependent ion transporters and ion channels, including the Na⁺-K⁺-ATPase (Glitsch & Tappe, 1993), the ATP-dependent K⁺ channel (Weiss & Lamp, 1989), and the SR Ca²⁺-ATPase (Xu et al. 1995). Glycolytic enzymes have been found to be tightly associated with sarcolemmal and SR membranes in muscle (e.g. Pierce & Philipson, 1985). In skeletal muscle it has been shown that the triads form a functional glycolytic ATP microcompartment (Han et al. 1992). A similar microcompartment may be formed by the cardiac diads which share many structural and functional properties with skeletal muscle triads. Because the ryanodine receptor (RyR)-Ca²⁺ release channel is a potential target for phosphorylation (e.g. Hain et al. 1995), variations in glycolytic activity may affect cardiac EC coupling by altering SR Ca²⁺ release, and might have a direct effect on the development of alternans.

Thus, one of the goals of the present study was to test the hypothesis that impairment of glycolytic flux may be a factor that induces cardiac alternans. This was achieved by testing the effects of various inhibitors of glycolysis on cell shortening, action potential kinetics and [Ca²⁺]_i transients. Furthermore, we characterized the spatio-temporal organization of cardiac [Ca²⁺]_i transients during alternans and contrasted differences in the subcellular properties of Ca²⁺ release in atrial and ventricular myocytes. Finally, we set out to identify the crucial step in EC coupling that causes cardiac alternans. A preliminary report of this study was published in abstract form (Hüser et al. 1998).

METHODS

Cell isolation

Cats of either sex were anaesthetized with sodium pentobarbital (70 mg kg⁻¹i.p.) and hearts were quickly removed by thoracotomy. Cardiac myocytes were enzymatically isolated by the method described previously (Wu et al. 1991; Rubenstein & Lipsius, 1995). Experiments were carried out at room temperature (20-22°C).

Calcium measurements

[Ca²⁺]_i was measured using the fluorescent Ca²⁺ indicators indo-1 (ratiometric, spatially averaged measurements) and fluo-3 (confocal imaging). Myocytes were loaded with the Ca²⁺ indicators by exposure to 5 μM acetoxymethyl esters of the dyes (fluo-3 AM and indo-1 AM; Molecular Probes, Eugene, OR, USA) for 20 min at 20°C. For fluorescence measurements a coverslip with the cells attached was mounted on the stage of an inverted microscope. For spatially averaged single cell [Ca²⁺]_i measurements indo-1 fluorescence was excited at 357 nm. Cellular fluorescence signals were recorded simultaneously at 405 nm (F₄₀₅) and 485 nm (F₄₈₅). Changes of [Ca²⁺]_i are expressed as changes in the ratio R = F₄₀₅/F₄₈₅. For spatially resolved [Ca²⁺]_i imaging a confocal laser scanning unit (LSM 410, Carl Zeiss, Germany) was used. In these experiments fluo-3 fluorescence was excited with the 488 nm line of an argon ion laser and the emitted fluo-3 fluorescence was measured at wavelength > 515 nm. [Ca²⁺]_i images were calculated according to the formula (Cannell et al. 1994):

where R is the normalized fluorescence (F/F_rest). A value of 1.1 μM was assumed for the dissociation constant (K_d) for the Ca²⁺-fluo-3 complex (Harkins et al. 1993) and [Ca²⁺]_rest was taken as 100 nM (Hüser et al. 1996; Blatter et al. 1997).

Electrophysiological measurements

Ionic currents were recorded using an amphotericin-perforated patch whole-cell recording method. Amphotericin B was dissolved in dimethylsulfoxide at a concentration of 60 mg ml⁻¹, and then added to the internal pipette solution to yield a final amphotericin concentration of 240 μg ml⁻¹. The standard internal pipette solution contained (mM): potassium glutamate 100, KCl 40, MgCl₂ 1, Na₂-ATP 4, EGTA 0.5 and Hepes 10. The solution was titrated with KOH to a pH of 7.2. To record I_{Ca, L}, K⁺ currents were blocked by replacing K⁺ with Cs⁺ in the internal pipette solution and adding 5 mM CsCl to the external solution. A single suction pipette was used to record voltage (bridge mode) or ionic currents (discontinuous voltage-clamp mode), using an Axoclamp 2A amplifier (Axon Instruments, Inc., Foster City, CA, USA). Action potentials were elicited by stimulation through the recording pipette with 2–3 ms voltage pulses. I_{Ca, L} was activated by clamping cells from a holding potential of -40 mV (to inactivate fast Na⁺ and T-type Ca²⁺ currents) to 0 mV for 200 ms.

Measurements of cell shortening

Unloaded cell shortening was measured with a video-based edge detector system (Crescent Electronics, Salt Lake City, UT, USA), which uses a single-raster line-scanning technique to detect edge motion at one or both ends of the cell. Cell length (l) data are expressed in micrometres of detected cell shortening (Δl).

Solutions

Cells were continuously superfused with Tyrode solution of the following composition (mM): NaCl 140, KCl 5, MgCl₂ 1, CaCl₂ 2, glucose 10 and Hepes 10; pH was adjusted to 7.3 with NaOH. In the experiments designed to evaluate the effect of metabolic substrates, glucose was replaced with pyruvate (10 mM) or β-hydroxybutyrate (10 mM). Iodoacetate was applied at a concentration of 1 mM.

Statistical analysis

Statistical significance was determined with the non-parametric two-tailed Mann-Whitney U test.

RESULTS

Electromechanical alternans in cat atrial and ventricular myocytes

Figure 1 illustrates electromechanical alternans recorded from single cat atrial (A; n = 8) and ventricular (B; n = 10) myocytes. In both cell types alternans was triggered by electrical stimulation at frequencies of 0.5–1.5 Hz at room temperature. The simultaneous recordings of action potentials (APs) and cell shortening revealed the typical characteristics of discordant mechanical and electrical alternans where contraction amplitude and AP duration alternated in a beat-to-beat fashion.

Figure 1. Electrical and mechanical alternans in atrial (A) and ventricular (B) myocytes.

Action potentials were evoked by 2 ms pulses in the current-clamp mode using the perforated patch technique (see Methods). Mechanical alternans were quantified by measuring cell shortening with a video edge detection method. In both cell types electrical stimulation at frequencies of 0.5–1.5 Hz evoked discordant electrical (top trace) and mechanical (bottom trace) alternans. To the right in A and B two APs recorded during subsequent small (○)- and large (•)-amplitude shortening are superimposed to illustrate the differences in duration and kinetics.

Spatio-temporal characteristics of [Ca²⁺]_i transients during alternans

The linescan mode of the confocal microscope was used to visualize the spatial and temporal patterns of [Ca²⁺]_i transients during alternans. A single atrial myocyte (Fig. 2A) was scanned repetitively along a line positioned perpendicular to the longitudinal cell axis (transverse scanning). Alternans was evoked by electrical field stimulation with 2 ms voltage pulses of suprathreshold amplitude applied through parallel platinum wires. The whole-cell [Ca²⁺]_i transient, averaged over the entire width of the cell (Fig. 2Aa) revealed regularly alternating amplitudes of [Ca²⁺]_i. The linescan images (Fig. 2Ab) and local [Ca²⁺]_i transients (Fig. 2Ac) showed marked [Ca²⁺]_i gradients underlying the electrically evoked [Ca²⁺]_i transients. During the early phase of both the large- and small-amplitude [Ca²⁺]_i transients, the rise of [Ca²⁺]_i occurred initially at the cell periphery (ss, subsarcolemmal space) and subsequently expanded towards the centre (ct) of the myocyte. This wave-like propagation of subsarcolemmal Ca²⁺ release towards the cell centre caused a ‘U’-shaped appearance of the onset of the [Ca²⁺]_i transient in the linescan image. In an earlier study from this laboratory (Hüser et al. 1996) we showed that the [Ca²⁺]_i gradients underlying [Ca²⁺]_i transients in atrial myocytes result from the ultrastructural arrangement of the membranes involved in the process of EC coupling, i.e. the sarcolemmal and SR membranes, and the fact that atrial cells lack t-tubules (McNutt & Fawcett, 1969; Hüser et al. 1996). During alternans marked differences in local [Ca²⁺]_i gradients were observed between the large- and small-amplitude transients in atrial myocytes. During the large-amplitude transient, Ca²⁺ release from the SR occurred throughout the entire cell. In contrast, during the small-amplitude transient, Ca²⁺ release in central regions of the cell consistently failed, leading to marked [Ca²⁺]_i gradients in centripetal directions. These spatial inhomogeneities of [Ca²⁺]_i contrasted clearly with [Ca²⁺]_i transients observed during alternans in ventricular myocytes (Fig. 2Ba). In ventricular myocytes the large- and small-amplitude [Ca²⁺]_i transient lacked the ‘U’-shaped appearance in the linescan images (Fig. 2Bb) and for a given transient the amplitude and the amount of Ca²⁺ release was homogeneous throughout the entire cell (Fig. 2Bc). This observation is in agreement with the general notion that the well-developed t-tubular system in ventricular myocytes ensures a tight coupling of transsarcolemmal Ca²⁺ flux to SR Ca²⁺ release throughout the entire cell volume.

Figure 2. Spatio-temporal characteristics of [Ca²⁺]_i transients during alternans in atrial and ventricular cells.

A, alternans (a) in atrial cells is characterized by marked variations in the spatial profile of the [Ca²⁺]_i transients (b). Local [Ca²⁺]_i profiles (c), recorded during large- and small-amplitude [Ca²⁺]_i transients, revealed [Ca²⁺]_i gradients directed from subsarcolemmal (ss) regions toward central regions (ct) of the cell. The small-amplitude [Ca²⁺]_i transients were spatially restricted to the subsarcolemmal regions. B, in ventricular myocytes during alternans (a) the [Ca²⁺]_i transients alternate in a spatially homogeneous pattern (b). During the large or the small transient the rise time and the size of the transients were similar in the subsarcolemmal (ss) and in the centre region (ct) of the cell (c).

Cellular mechanism underlying alternans

Because the amplitude and the spatial distribution of [Ca²⁺]_i is profoundly affected during electromechanical alternans we investigated several key steps in the process of EC coupling that might provide insight into the mechanisms governing alternans. The reduced release of Ca²⁺ from the SR suggested that alternans might be caused by either a reduction of trigger Ca²⁺, i.e. alternations of I_{Ca, L}, beat-to-beat alternations of the SR Ca²⁺ load or alternating changes of the gain of Ca²⁺-induced Ca²⁺ release (CICR), or efficiency of EC coupling. The following series of experiments was designed to discriminate between these possibilities.

I_{Ca, L} during alternans

To test whether trigger Ca²⁺ for SR Ca²⁺ release was changing during alternans, we simultaneously measured cell shortening and I_{Ca, L} in atrial (n = 7) as well as in ventricular (n = 4) cells. As shown in Fig. 3 electrical stimulation under voltage-clamp conditions caused mechanical alternans. In contrast to cell shortening, peak I_{Ca, L} measured simultaneously did not alternate, suggesting that the signal which triggers SR Ca²⁺ release was not the key mechanism responsible for the development of alternans. During the large-amplitude [Ca²⁺]_i transient, the inactivation of I_{Ca, L} was slightly faster, and the current component elicited upon repolarization, attributable to Na⁺-Ca²⁺ exchange, was larger. Both of these observations are explained by a greater amount of released Ca²⁺ during the large-amplitude transient, leading to stronger Ca²⁺-dependent inactivation of the current and enhanced Ca²⁺ extrusion via Na⁺-Ca²⁺ exchange. Nevertheless, the constant amplitude of peak I_{Ca, L} during alternans argues against significant alternations in trigger Ca²⁺ for SR Ca²⁺ release as a primary mechanism. In addition, the occurrence of alternans in voltage-clamped cells demonstrates that the variations in AP duration are not causing the beat-to-beat changes in [Ca²⁺]_i transient and contraction amplitude. In fact, the changes in AP duration were probably the result of alternations of [Ca²⁺]_i-dependent processes responsible for the shape of the AP.

Figure 3. I_{Ca, L} and cell shortening during alternans.

Simultaneous measurements of cell shortening (top) and I_{Ca, L} (bottom) in an atrial myocyte. Holding potential V_H=−40 mV, test potential = 0 mV. Despite large differences in cell shortening during alternans peak I_{Ca, L}remained constant.

SR Ca²⁺ load during alternans

An alternative mechanism responsible for the changes in amplitude of the [Ca²⁺]_i transient during alternans could be an alternating change in the SR Ca²⁺ load, specifically beat-to-beat variations in the amount of releasable Ca²⁺, a mechanism that has been proposed in earlier studies (for review see Euler, 1999). We tested this possibility by challenging individual cells with caffeine during alternans in order to estimate the SR Ca²⁺ content in relation to the small- and large-amplitude [Ca²⁺]_i transients. Individual cells were exposed to 10 mM caffeine after a small- and a large-amplitude transient, respectively. As shown in Fig. 4 no significant difference in the amplitude of the caffeine-induced [Ca²⁺]_i transients was observed. The average amplitude (expressed as the indo-1 ratio R) of the caffeine-induced [Ca²⁺]_i transients was 1.83 ± 0.38 (mean ±s.d.; n = 16) after the large-amplitude transient, and 1.84 ± 0.38 (n = 15) after the small-amplitude transient (difference not statistically different). When normalized to the last large-amplitude transient preceding the application of caffeine, the average amplitude of the caffeine-induced [Ca²⁺]_i transient was 114 and 113 % after the large- and small-amplitude transient, respectively. These results indicate that the differences in amplitudes of electrically evoked [Ca²⁺]_i transients were not caused by beat-to-beat alterations of SR Ca²⁺ load or the amount of releasable Ca²⁺.

Figure 4. SR Ca²⁺ content during alternans.

SR Ca²⁺ content was evaluated with 10 mM caffeine pulses in atrial myocytes. Caffeine was applied twice to the same cell. The amplitude of the caffeine-induced [Ca²⁺]_i transient did not reveal any significant differences after large- (left panel) and small-amplitude (right panel) [Ca²⁺]_i transients, indicating that SR Ca²⁺ content remained constant and did not alternate.

Effect of β-adrenergic stimulation on alternans

β-Adrenergic stimulation enhances cardiac inotropy and affects Ca²⁺ regulation and EC coupling through a number of different mechanisms (for reference see Bers, 1991, and Discussion). To test the effect of β-adrenergic stimulation on the efficiency of EC coupling we exposed cardiac cells showing electromechanical alternans (n = 4) to the β-adrenergic agonist isoproterenol (1.5 nM). As shown in Fig. 5, over the course of approximately 30 s in the presence of the β-adrenergic agonist, alternans disappeared. More specifically, the larger-amplitude contraction slightly decreased and the smaller-amplitude contraction progressively increased until the contraction amplitudes failed to alternate. The reciprocal changes in contraction amplitude suggest that both the small- and large-amplitude contractions draw Ca²⁺ from a common pool. In addition, the alternations in action potential duration subsided in concert with the change in contraction. The effects of β-adrenergic stimulation were reversible as alternations in action potential duration and contraction resumed after removal of isoproterenol.

Figure 5. Effect of β-adrenergic stimulation on electromechanical alternans.

Discordant electromechanical alternans were elicited in a single ventricular cell by electrical stimulation (1.1 Hz) under current-clamp conditions at room temperature. Top panel: application of isoproterenol (1.5 nM) reversed alternans. Upon wash-out of isoproterenol electromechanical alternans resumed. Bottom panels, action potential recordings and cell shortening at expanded time scale recorded at times indicated by the letters a to e in the top panel.

From the experiments above we could rule out two crucial steps in EC coupling as the primary cause for alternans: beat-to-beat variations in I_{Ca, L} and the amount of releasable Ca²⁺ from the SR. Consistent with the positive inotropic effect of β-adrenergic stimulation, we then hypothesized that alternans appeared to be the result of variations in the efficacy of EC coupling, i.e. the efficiency of a given trigger signal (I_{Ca, L}) to initiate SR Ca²⁺ release. Because many of the conditions which facilitate the occurrence of alternans, such as ischaemia, low pH and hypothermia, are accompanied by or are indicative of profound changes in cellular energy metabolism we set out to test specifically whether the occurrence of alternans is linked to alterations in cellular energy production.

Relationship between cellular energy metabolism and alternans: impairment of glycolysis

Based on the notion that glycolytically derived ATP is preferentially used by certain ion transport mechanisms (see Introduction) we hypothesized that impairment of glycolytic ATP formation might be a crucial causal factor in the development of alternans. Several protocols were used to selectively reduce glycolysis without affecting oxidative metabolism. With the first protocol the metabolic substrate in the bathing solution was changed from glucose to pyruvate. Pyruvate is readily taken up by the cell and effectively metabolized by mitochondria. In fact, many authors have reported positive inotropic effects under these conditions which were explained by the increased overall cytosolic phosphorylation potential in the presence of pyruvate (for reference see e.g. Martin et al. 1998). With pyruvate as metabolic fuel the flux through glycolysis is markedly inhibited. This inhibitory effect is thought to be caused by inhibition of the phosphofructokinase (PFK) reaction by increased cytosolic concentrations of citrate and ATP (Newsholme et al. 1962). The effects of inhibition of glycolytic flux by exchanging glucose with pyruvate in the bath solution are illustrated in Figs 6 and 7. Electrical stimulation under current-clamp conditions in the presence of 10 mM pyruvate resulted in typical electromechanical alternans in atrial (n = 6) and ventricular cells (Fig. 6A; n = 2) which was reversible upon removal of pyruvate. Figure 6B shows that during pyruvate-induced alternans the voltage-gated L-type Ca²⁺ current (I_{Ca, L}) was not affected (n = 4 in atrial myocytes; n = 3 in ventricular cells), further suggesting that the trigger Ca²⁺ signal for EC coupling did not appear to be primarily involved in the generation of alternans. This result also shows that peak I_{Ca, L} was unaffected during alternans independent of the protocol applied to induce it, i.e. whether alternans was triggered by pacing (Fig. 3) or by inhibition of glycolysis (Fig. 6B). Figure 7A illustrates that pyruvate-induced electromechanical alternans was paralleled by [Ca²⁺]_i transient alternans.

Figure 6. Alternans of cell shortening, action potentials and I_{Ca, L} induced by pyruvate.

A, discordant electromechanical alternans elicited by external application of 10 mM pyruvate in a ventricular myocyte under current-clamp conditions. B, mechanical alternans recorded from voltage-clamped atrial myocytes superfused with pyruvate. During alternans I_{Ca, L} remained constant. Holding potential V_H=−40 mV, test potential = 0 mV.

Figure 7. [Ca²⁺]_i transient alternans induced by inhibition of glycolysis.

Inhibition of glycolysis by exposure to 10 mM pyruvate (A) and 1 mM iodoacetic acid (IAA; B) caused [Ca²⁺]_i transient alternans in an atrial myocyte. Alternans were reversible upon wash-out of the inhibitors of glycolysis. The bottom traces show the [Ca²⁺]_i transient at expanded time scales.

The second protocol used iodoacetate (iodoacetic acid, IAA) to directly inhibit glycolysis at the level of glyceraldehyde 3-phosphate dehydrogenase (GAPDH). Because IAA treatment typically causes irreversible cell damage when applied for prolonged periods of time (e.g. Weiss & Hiltbrand, 1985), cells were challenged with this drug for only short durations (< 30 s). Similar to pyruvate, a short exposure to 1 mM IAA (Fig. 7B) elicited reversible alternans of [Ca²⁺]_i transients (n = 11). Furthermore, the fatty acid β-hydroxybutyrate, another metabolic substrate which inhibits glycolysis through accumulation of intermediates of the TCA (tricarboxylic acid) cycle and ATP in the cytosol (Newsholme et al. 1962), also was effective in triggering alternans (data not shown).

In summary these experiments showed that selective inhibition of glycolysis was capable of triggering alternans, suggesting that availability of glycolytically derived ATP plays a crucial role in the development of electromechanical alternans.

DISCUSSION

The present study was undertaken to gain insight into the subcellular mechanism underlying alternans in cardiac myocytes, and to investigate the relationship between cellular energy metabolism and the occurrence of electromechanical alternans. The most important result of the present study was the novel observation that alternans could be induced specifically by inhibition of glycolysis. In addition, the distribution of [Ca²⁺]_i during alternans was different in ventricular and atrial myocytes and revealed distinct spatial inhomogeneities, particularly in atrial myocytes. We also established evidence that eliminated alternations of trigger Ca²⁺ (i.e. I_{Ca, L}) and SR Ca²⁺ load as the primary cause of alternans. Furthermore, conditions that enhance cardiac inotropy such as β-adrenergic stimulation are able to overcome alternans. These results suggest that electromechanical alternans originates in regular alternations of the gain of CICR or the efficiency of EC coupling. The mechanism underlying these alternations appears to be linked to the cellular energy metabolism as will be discussed below.

Spatial characteristics of [Ca²⁺]_i transients during alternans

Spatially resolved [Ca²⁺]_i imaging during electromechanical alternans revealed distinct differences in the distribution of [Ca²⁺]_i between ventricular and atrial myocytes. In ventricular cells the subcellular distribution of [Ca²⁺]_i during the small and large transients was homogeneous throughout the entire cell and no spatial inhomogeneities could be observed. This observation was in sharp contrast to the [Ca²⁺]_i transients in atrial cells, which revealed a typical U-shaped [Ca²⁺]_i profile in the linescan image. During the small-amplitude transients subsarcolemmal [Ca²⁺]_i elevations frequently failed to elicit further release in more central regions of the cell.

The differences in [Ca²⁺]_i distribution between atrial and ventricular cells during normal activation (Berlin, 1995; Hüser et al. 1996) and during alternans can be explained by differences in the mechanism of EC coupling in these two cardiac cell types. In ventricular cells a small amount of Ca²⁺ entering the cell through voltage-controlled surface membrane Ca²⁺ channels (i.e. dihydropyridine receptors, DHPRs) triggers regenerative release of Ca²⁺ from the SR via CICR (Fabiato, 1983). The extensive three-dimensional transverse-tubular system in ventricular myocytes ensures close physical association of surface and SR membranes throughout the entire cell volume. As a result, depolarization triggers virtually instantaneous SR Ca²⁺ release throughout the cell (Fig. 2B). In contrast, atrial myocytes lack t-tubules (McNutt & Fawcett, 1969; Hüser et al. 1996) and are thought to possess two types of SR termed junctional and non-junctional (or corbular) SR. During EC coupling Ca²⁺ release is initiated at the peripheral junctional SR by a mechanism virtually identical to ventricular muscle. However, the increase of peripheral [Ca²⁺]_i subsequently diffuses to central regions of the cell to trigger Ca²⁺ release from non-junctional SR, resulting in an inhomogeneous distribution of [Ca²⁺]_i (Fig. 2A; see also Hüser et al. 1996). Because ventricular cells depend significantly less on diffusion and the trigger signal for Ca²⁺ release is more uniform, during alternans the changes in [Ca²⁺]_i during both the small- and large-amplitude transients are relatively homogeneous. During alternans in atrial cells, however, the small-amplitude [Ca²⁺]_i transient initiated at the periphery provides a relatively small diffusional gradient for further non-junctional Ca²⁺ release. As a result, non-junctional SR Ca²⁺ release is decrementally diminished to the point where in most central regions of the cell, failures of Ca²⁺ release and EC coupling occur. Despite these differences in the subcellular patterns of intracellular Ca²⁺ release between atrial and ventricular myocytes, the molecular mechanisms causing alternans of [Ca²⁺]_i transients are conserved in both cell types. On the other hand, these subcellular differences in Ca²⁺ release may affect other parameters of alternans, not examined in the present study, such as the threshold cycle length at which alternans is initiated or the susceptibility of each tissue type to develop alternans under a given set of conditions. The fact that alternans occurred in both cell types even though they exhibit differences in spatial [Ca²⁺]_i gradients during EC coupling as well as structural properties (t-tubules) indicates that these factors do not contribute to the primary mechanisms underlying alternans. Therefore we looked at other steps during the EC coupling process that both cell types have in common.

Mechanism underlying alternans: what alternates?

Evidence from this as well as earlier studies indicates that electromechanical alternans is ultimately linked to alterations in myocardial Ca²⁺ homeostasis. During alternans the changes in contraction amplitude are paralleled by changes in the amplitude of the underlying [Ca²⁺]_i transient (e.g. Lab & Lee, 1990; Kihara & Morgan, 1991; Orchard et al. 1991; Uno, 1991). Furthermore, during alternans mechanical restitution is significantly slowed after the large-amplitude twitch (e.g. Spencer et al. 1992; Rubenstein & Lipsius, 1995), indicative of an impaired [Ca²⁺]_i regulation (Hirayama et al. 1993). Mechanical restitution, however, is the result of a sequence of intracellular processes involving Ca²⁺ uptake into the SR, intra-SR redistribution of Ca²⁺, and recovery of SR release channels from inactivation (or adaptation). Thus, the mechanism of alternans may reside anywhere in this multi-step process. Despite the fact that many reports have pointed toward an involvement of intracellular Ca²⁺, the mechanism underlying electromechanical alternans remains elusive to this point.

From the present experiments we can eliminate two pivotal components of EC coupling as a primary cause for alternans: trigger Ca²⁺ in the form of I_{Ca, L} and SR Ca²⁺ load. Experiments using caffeine to estimate the amount of releasable Ca²⁺ within the SR revealed that the SR Ca²⁺ load was the same after the small- and large-amplitude [Ca²⁺]_i transient. It has been proposed that alternans could be explained by a mechanism of EC coupling that is based on two separate SR compartments, a Ca²⁺ uptake compartment and a release compartment (see Euler, 1999). According to this model alternans could occur under conditions where insufficient time is permitted between contractions for the transport of Ca²⁺ from the uptake to the release sites. Such a delay between uptake and release could account for [Ca²⁺]_i transient alternans when the time interval between individual contractions is too short to allow for sufficient recycling of Ca²⁺ to the release compartment (Lab & Lee, 1990; Kihara & Morgan, 1991). Our caffeine results (Fig. 4), however, argue against such a mechanism. As shown in Fig. 4 caffeine caused the same elevation of [Ca²⁺]_i whether it was applied after the small- or the large-amplitude [Ca²⁺]_i transient. This novel result indicates that the amount of Ca²⁺ available for release from the SR from one beat to the next is unchanged during alternans, and eliminates the simple explanation that alternans of [Ca²⁺]_i transients might be due to incomplete replenishing of the SR with Ca²⁺ after the large transient.

Furthermore, the observation that peak I_{Ca, L} failed to undergo significant changes during the small and large [Ca²⁺]_i transients implied that the primary mechanism responsible for alternans had to reside downstream of the initial step of Ca²⁺ entry. Therefore, the caffeine experiments and the I_{Ca, L} measurements suggested that the gain of EC coupling or the efficiency with which a given amount of ‘trigger Ca²⁺’ releases Ca²⁺ from the SR undergoes regular alternations.

β-Adrenergic stimulation prevents alternans

In our experiments alternans was reversed during β-adrenergic stimulation with isoproterenol, an observation that has also been reported in intact dog heart (see Euler, 1999). The present results indicate that the primary effect of β-adrenergic stimulation to prevent alternans operates at the cellular level, rather than through alterations in other parameters of heart function (Euler et al. 1996). The positive inotropic effect on cardiac function resulting from β-adrenergic stimulation involves several different cellular mechanisms (for review see Bers, 1991) which result in an increase in Ca²⁺ current and enhanced SR Ca²⁺ uptake. It is generally believed that these effects involve phosphorylation processes that are mediated by cAMP-dependent protein kinase (PKA). At least two mechanisms, which could even act synergistically, are plausible explanations for the relationship between cell metabolism and alternans observed in our study. It has been shown that the RyR-Ca²⁺ release channel is a substrate for phosphorylation (including by PKA) which modulates RyR function, enhances channel activity and appears to be required for Ca²⁺ release (Hain et al. 1995; Valdivia et al. 1995). In addition, it has been shown that in the heart β-adrenergic stimulation via a cAMP-dependent protein kinase mechanism affects glucose metabolism by enhancing glycogen breakdown (Hayes & Mayer, 1981), resulting in an increase of the amount of glucose available to processes that critically depend on glycolytically derived ATP. Interestingly, exposure of cardiac cells exhibiting alternans to the phosphodiesterase inhibitor IBMX (data not shown) also reversed electromechanical alternans. This is consistent with the observations made with isoproterenol and suggests an important role of the cAMP-PKA system.

Tight link between SR Ca²⁺ release, EC coupling and glycolysis

The notion that conditions which lead to profound changes in the metabolic state of cardiac cells with low levels of ATP such as ischaemia (e.g. Lee et al. 1988; Wu & Clusin, 1997) can also induce alternans, suggests that the occurrence of alternans might be linked directly to changes in cellular energy metabolism. Indeed, experimental conditions designed to inhibit glycolytic flux and therefore glycolytic ATP production reliably triggered alternans in our study. Inhibition or impairment of glycolysis at the level of the phosphofructokinase reaction (by pyruvate or β-hydroxybutyrate; Newsholme et al. 1962) or GAPDH (by IAA) rendered cells more susceptible to alternans. Impairment of glycolytic flux in cells stimulated at 0.8–1.2 Hz at room temperature by these experimental manoeuvres frequently induced alternans. The effect of impaired glycolytic flux was reversible upon return to normal Tyrode solution containing only glucose as metabolic substrate. Notably, increasing the temperature to 30–35°C counteracted the occurrence of metabolically induced alternans, an observation consistent with previous findings that hypothermia itself is a condition which facilitates the development of alternans (e.g. Spencer et al. 1992). Higher temperatures result in higher rates of ion transport (e.g. intra-SR redistribution of Ca²⁺) and acceleration of enzymatic processes (e.g. energy-producing metabolic pathways, phosphorylation processes) which in turn could rescue cardiac cells from alternans.

Our results suggest that electromechanical alternans arises from beat-to-beat variations in the efficiency with which Ca²⁺ entry triggers SR Ca²⁺ release. We suggest that the sensitivity of the Ca²⁺ release process is regulated by the flux through glycolysis, and that the availability of glycolytically generated ATP may represent the functional link between glycolysis and the sensitivity of CICR. The structural and functional prerequisite that would allow the preferential use of glycolytically derived ATP over ATP generated by oxidative phosphorylation could reside in functional microcompartments in which ATP-dependent proteins (such as RyR-Ca²⁺ release channel) are tightly regulated by the activity of co-localized glycolytic enzymes and endogenous protein kinases (reviewed by Goldhaber, 1997). The organization of sequential enzymes into functional complexes (‘metabolon’) has been suggested for several biochemical cascades, including glycolysis (Masters, 1977; Srere, 1987; al-Habori, 1995). Such metabolic microcompartments might preferentially be formed at membrane surfaces or associated with cytoskeletal elements, both of which have been shown to bind glycolytic enzymes (e.g. Entman et al. 1976; Pierce & Philipson, 1985). Moreover, biochemical (Srere, 1987) and functional experiments (e.g. Han et al. 1992; Xu et al. 1995) support the notion that sequential enzymes of the glycolytic chain associate to form complexes allowing metabolic substrate channelling. Similar to skeletal muscle triads (Han et al. 1992), we suggest that the diadic cleft of cardiac myocytes forms such a glycolytic microcompartment. In this microcompartment the activity of the RyR-Ca²⁺ release channel is modulated by locally generated ATP either directly or through phosphorylation reactions. Indeed the release channel has been shown to be the substrate for phosphorylation by protein kinases A and G as well as by an endogenous Ca²⁺-calmodulin-dependent kinase (see e.g. Witcher et al. 1991; Hohenegger & Suko, 1993; Strand et al. 1993; Hain et al. 1995). Phosphorylation results in increased channel activity, altered channel kinetics and is generally believed to stimulate EC coupling. In an alternative model EC coupling might not be linked to glycolysis through ATP itself but rather through intermediates of the glycolytic reaction. In a recent study Kermode et al. (1998) provided evidence that glycolytic sugar phosphate intermediates could directly activate cardiac ryanodine receptor channels incorporated into lipid bilayers.

The notion of metabolic microcompartments is not unique to muscle. For example, a microcompartment with similar properties is formed by the postsynaptic density (PSD) in CNS neurons (Wu et al. 1997) which contains glycolytic enzymes. The locally generated ATP is consumed by protein kinases in the PSD which control the activity of various PSD proteins, including ion channels (e.g. ionotropic glutamate receptors). Therefore, the formation of discrete cytosolic microcompartments controlled by localized glycolytic ATP generation might provide a general scheme for the regulation of various cellular processes.

Conclusion

In the intact heart electrical and mechanical alternans is most frequently observed during acute myocardial ischaemia, a condition which is characterized by impaired energy metabolism. Within the model of compartmentalized glycolytic ATP production the proposed mechanism of cardiac alternans can be summarized as follows. Cardiac alternans is explained by an impairment of glycolytic flux due to restricted substrate availability. As discussed previously the number of available and active RyR-Ca²⁺ release channels (which ultimately determines the magnitude of the [Ca²⁺]_i transient and contraction) is governed by the phosphorylation state of RyRs which in turn depends critically on ATP available to protein kinases. Thus reducing the flux through glycolysis will reduce the local phosphorylation potential. Consequently, the rate of phosphorylation of release channels will be slowed and the equilibrium between phosphorylated (available) and non-phosphorylated channels will be shifted towards the non-phosphorylated state. The smaller number of available release channels results in a smaller [Ca²⁺]_i transient. The small-amplitude [Ca²⁺]_i transient is followed by a large-amplitude [Ca²⁺]_i transient because a larger number of release channels, not available during the previous beat, have now recovered from inactivation and have become available to participate in release. Due to the slowed recovery of the release channels, a smaller number of channels is available for the following beat, resulting again in a [Ca²⁺]_i transient of smaller amplitude. Consequently, this mechanism sets up regular oscillations in the magnitude of SR Ca²⁺ release. In this way, alternans is the result of alternating changes of the gain, or efficiency, of EC coupling due to alternations in ATP levels in the microdomain of the SR Ca²⁺ release channel.