Calcium Sparks: Release Packets of Uncertain Origin and Fundamental Role

How Big Is a Spark?

Measured spark amplitude varies widely from the lower limits of detection (defined by noise, usually ∼0.3 U F ₀) to the highest values recorded, which in our experience with cut fiber segments under voltage clamp go routinely beyond 6 F ₀ for a few sparks in every experiment. Current work with cardiac myocytes and smooth muscle also finds events of amplitude >5, much larger than in earlier reports (Wier et al., 1999). In our experience amplitudes are substantially lower in fibers permeabilized by saponin or by notches and immersed in internal solution (a preparation described by Lacampagne et al., 1998).

The reason for the differences in amplitude may be quite simple: as stated, ΔF/F ₀ is close to the ratio between increase in local [Ca²⁺] and resting [Ca²⁺]_i. The cell may be able to maintain a substantially lower internal Ca²⁺ in the voltage clamp experiments, which expose it to internal solution at the cut ends only. By contrast, in permeabilized fiber segments, free [Ca²⁺] should equilibrate rapidly within the cell at the solution value, and the dye rises rapidly, reaching values higher than the concentration in the internal solution. Therefore, the relative increase ΔF/F ₀ (or Δ[Ca²⁺]/ resting [Ca²⁺]) should be less in the permeabilized fibers due to the buffering effect of the higher [dye] (which reduces Δ[Ca²⁺]) and the higher resting [Ca²⁺]. Indeed, sparks with very large values of ΔF/F ₀ are only found early in voltage clamp experiments, when the resting fluorescence is very low (indicating low dye concentration, low [Ca²⁺], or both).

In our first study of discrete events (Tsugorka et al., 1995), we evaluated event amplitude by constructing all-points histograms of the difference records between fluorescence at triadic centers and in the neighboring sarcomeres. The resulting histograms had modes at between 0.1 and 0.3 F ₀, which were interpreted as spark amplitudes. We now believe that the all-points histogram is not an adequate tool to recognize sparks because it does not take into account their multidimensional aspects. When their frequency is low, sparks will contribute negligibly to the all-points histogram, which should be dominated by very small fluctuations that may correspond to out-of-focus sparks or to continuous, eventless release. In fact, when automatic procedures are used with an amplitude selection criterion (see next section), the resulting histograms also peak at or near the lowest allowed amplitude. Therefore, modes in the amplitude distributions will always be at very small values regardless of the true amplitude of the events. We now prefer methods of recognition by multidimensional criteria (amplitude, width, duration). Such recognition was initially done by an observer (Cheng et al., 1993; Klein et al., 1996; Shirokova and Ríos, 1997), but is now entirely automatic (Cheng et al., 1999), and yields amplitude estimates at least an order of magnitude greater.

The width of large sparks (measured as full width at half magnitude, FWHM) varies between 1.5 and 2.2 μm in skeletal muscle. In cardiac muscle, where some sparks appear to result from multiple release sources (Parker et al., 1996; Blatter et al., 1997) and can be very wide, there is a well defined subset of sparks that appear to originate from single release sites. The half width of those is similar to that in skeletal muscle, ∼2 μm. Surprisingly, and as shown in Fig. 1, in simulations of sparks resulting from a discrete Ca²⁺ source the half width is only ∼1 μm, regardless of assumptions on the properties of buffers and the dye. The width of the simulated spark could be increased to reach values near 2 μm by increasing the radius of the source (to 1 μm). The discrepancy and its resolution may be taken as evidence that the source of release is extensive (rather than a single channel), or it may mean that the models of Ca²⁺ release and removal are inadequate in ways we do not understand.

Ca²⁺ Release Flux Underlying a Spark

This has been calculated in cardiac (Blatter et al., 1997; Lukyanenko et al., 1998) and skeletal muscle (Ríos et al., 1998; manuscript submitted for publication) by generalizing global or whole-cell procedures of Baylor et al. (1983) and Melzer et al. (1984, 1987) to the spatially resolved images obtained by confocal microscopy. The earlier procedures first derive from the optical signals the transient increase in free [Ca²⁺], and then calculate the increase in total calcium necessary to account for such transient by adding contributions from different Ca²⁺ binding and removal processes. They are therefore subject to errors in the calculation of free [Ca²⁺] and in the ulterior estimate of binding and removal. The generalization to sparks has additional uncertainties regarding diffusional mobility of the different Ca²⁺ ligands and placement of fixed sites. If calculations are carried out assuming a homogeneous and unrestricted distribution of sites, with consensus values for concentrations, diffusion coefficients, and reaction rate constants, the values obtained are typically near 10 pA, and often >20 pA for the largest sparks in every image.

A lower bound of released Ca²⁺ can be obtained without calculating local [Ca²⁺], based on the rate of Ca²⁺ binding to the dye that is needed to account for the observed fluorescence. Take as numeric example a permeabilized fiber equilibrated with a 100 nM free [Ca²⁺], 200 μM fluo-3 solution, at a time when the internal dye concentration has risen to 300 μM. In this condition, the resting fluorescence corresponds to ∼30 μM of the dye:Ca complex (assuming that the dye reacts inside the fiber with a K _d of 1 μM, the low end of the range estimated by Harkins et al., 1993). A large spark under these conditions may have an amplitude of 2.5 F ₀, requiring an additional 75 μM of dye-bound Ca²⁺. The quantity of dye-bound Ca²⁺ can be found by volume integration of concentration, roughly the product of peak value by the volume of the sphere that intersects the spark at half magnitude, say 1.8 μm. Thus, 75 μM × 3.05 μm³ or 0.23 amol Ca²⁺, equivalent to 43.4 fC, must be released within the rise time of the spark, 5 ms. Therefore, an average current close to 10 pA must flow for 5 ms just to account for the Ca²⁺ bound to the dye in this example. Such values are typical of large sparks in experiments with permeabilized fibers.

The free [Ca²⁺] in cardiac SR was estimated using ¹⁹F NMR at 1.5 mM (Chen et al., 1996). Mejía-Alvarez et al. (1999) used the result to estimate the physiologic current-carrying capability of cardiac release channels, measuring their Ca²⁺ current in bilayers, with a luminal side [Ca²⁺] of 2 mM and in the presence of concentrations of Cs⁺ comparable to physiologic [K⁺]. Under those conditions, the average unitary current was 0.35 pA. Extrapolated to the case of skeletal muscle, this estimate indicates that tens of channels should be open to account for large sparks.

Though lower than in earlier work (Tinker et al., 1992), these estimates of current per channel are consistent with whole-cell measures of release flux. The maximum flux density (expressed as rate of rise of total calcium concentration in the accessible myoplasmic volume) under voltage clamp in the frog is between 180 and 200 mM/s (Pape et al., 1995; Shirokova and Ríos, 1996). The volume density of release channels can be gathered from morphometry. The ratio of transverse tubule length to fiber volume is 0.82 μm⁻² in frog muscle (Eisenberg and Peachey, 1975). If 70% of this length is junctional and contains a double row of release channels at 30 nm spacing on each side (Block et al., 1988), the number of channels per liter is 0.82 μm⁻² (tubule length/fiber volume) × 0.7 (triad length/tubule length), × 2 (junctions/triad), × 2 (rows/junction), × 33 (channels/row/μm), or 1.08 × 10¹⁷ channels per liter of fiber. At 100% activation, such channels passing 0.35 pA would generate a flux density of 180 mmol/liter of fiber per s, or 260 mM/s in terms of accessible aqueous myoplasmic volume, surpassing the highest whole-cell estimates.

The Distribution of Spark Amplitudes

The appeal of sparkology, as we stated, is to some extent esthetic—images with sparse sparks, in a dark experimental room, look very much like a starry sky. Much as stars in the sky, small sparks greatly outnumber the large ones, and the reasons for this are not very different for sparks and stars. Indeed, sparks are recorded in line scans, but may originate anywhere within the junctions of a Z disk. The smoothly decaying point spread function of the imaging system implies that the sparks originating farther from the scanned line will appear smaller in the record. And the regions of Z disk that lie “far” are greater than those that are “near,” hence the probability density function of spark amplitudes is expected to be decaying, even if sparks as objects are all of the same size.

The above reasoning led two groups to derive, using different formal approaches, the distribution of amplitudes expected in a line scan image for identical sparks that originate with homogeneous probability in triadic locations (Izu et al., 1998; Cheng et al., 1999). This function is a simple inverse proportionality (pdf ∝ a ⁻¹, where a is recorded amplitude).

In agreement with theory, the amplitude histograms of events identified by a program that locates them without human intervention (Cheng et al., 1999) are monotonically decreasing. The dependence of frequency on amplitude, however, is not the inverse function predicted for identical sparks, being consistent instead with a widely spread distribution of real spark amplitudes. The distribution of real spark amplitudes (derived by a method of González et al., 1999, which corrects for the effect of line scanning) exhibits in the presence of caffeine a mode or preferred amplitude at between 2 and 3 F ₀.

With this improved understanding of the meaning of recorded spark amplitude, it is natural to use large sparks, rather than those of average size, to estimate release flux, because the large ones are more likely originated near the scanned line, where recorded amplitudes are greatest. Because there is a spread of actual amplitudes, the largest sparks, whose underlying release current was estimated at >20 pA, may be outliers of unusually large amplitude as objects. For the sparks of the most common size, one third of the above estimate should apply, or ∼7 pA.

Beyond Sparks

The above considerations, plus the structural evidence that release channels are clustered in closely packed arrays, in skeletal (Block et al., 1988) as well as cardiac muscle (Sun et al., 1995), present a problem if single channels gate individually to produce sparks. How can a release channel pass the spark current (somewhere between 1 and 30 pA) without activating other channels that face the same junctional gap, where [Ca²⁺] would rapidly rise beyond 100 μM (Langer and Peskoff, 1996; Stern et al., 1997)? The argument was bolstered by the demonstration in cardiac muscle of coupled fluorescence events in scans perpendicular to the fiber axis (Parker et al., 1996) and of multiple sources of release under large sparks of cat atria (Blatter et al., 1997). Both observations show that the activation of cardiac release channels, presumably by Ca²⁺, occasionally reaches across distances of 1 μm or more, stressing that single Ca²⁺-activatable channels within a junction can hardly operate independently.

A many-channel origin for sparks was supported by the observation in ventricular myocytes of a diffuse increase in fluorescence, devoid of sparks, when trigger Ca²⁺ was delivered by photolysis of DM-nitrophen (Lipp and Niggli, 1996). The authors hypothesized that thus triggered SR Ca²⁺ release was composed by contributions (termed quarks) that could not be individually detected. Although the terminology is intriguing, skeletal muscle soon forced consideration of another alternative, continuous release.

In skeletal muscle, an increase in fluorescence not composed of sparks can be readily demonstrated under a number of conditions. When slightly depolarized, some frog fibers show at triads release not constituted by discrete events, while typical sparks appear at higher voltages (Shirokova and Ríos, 1997). In tetracaine at 200 μM, which blocks release channels in bilayers (Xu et al., 1993), depolarization triggers continuous release but no sparks (Shirokova and Ríos, 1997).

In mammalian muscle of adult rats, voltage-clamp depolarization caused continuous Ca²⁺ release and contraction, but no sparks (Shirokova et al., 1998). The triadic origin of the release was demonstrated by the existence of a triadic gradient of fluorescence (which is nearly proportional to release flux; Tsugorka et al., 1995). Because the gradients of fluorescence associated with continuous release in both frogs and rats could be much lower than in sparks, it appears that continuous release may involve smaller currents, again indicating the participation of channel groups in the production of sparks.

Shirokova and Ríos (1997) interpreted the continuous release observed in frogs at very low voltages or in the presence of tetracaine as flux through channels directly operated by voltage sensors. They proposed that sparks are the consequence of opening of multiple channels, caused by a local increase in [Ca²⁺]. This increase would be started in frog skeletal muscle by release channels operated by voltage-sensing DHPrs, and reinforced by the opening of additional channels. The directly voltage-operated release component could appear alone, without triggering CICR and sparks, if the concentration of trigger Ca²⁺ was insufficient (e.g., at low voltage depolarization) or the sensitivity to Ca²⁺ was inhibited (in the tetracaine experiment). This hypothesis also explains why it is so difficult to demonstrate nonspark release in cardiac muscle, where presumably there is no direct activation of release by voltage.

Rat muscle provides an example of eventless release (Shirokova et al., 1998) that is most interesting for various reasons. The failure to observe sparks there could simply evidence difficulties in working with smaller, weaker cells, cut (Hollingworth et al., 1996), and at far from physiologic temperature. We believe instead that the absence of sparks in adult rats reflects a fundamental difference in the E-C coupling mechanism, consistent with differences found in whole cell experiments (Shirokova et al., 1996). These differences include a lesser peak of Ca²⁺ release flux, a lower ratio between the peak and steady levels of release under voltage clamp, the absence in the rat of the characteristic voltage dependence of this ratio observed in the frog, and the presence in the rat (Suda and Penner, 1994), but not in the frog (Shirokova and Ríos, 1996), of the RISC phenomenon, a stop by repolarization of the release induced by caffeine.

RISC is interpreted as indicating a preponderance of control by membrane voltage (via the DHPr). Its presence in the rat, and failure to appear in the frog, are consistent with our failure to demonstrate sparks (which presumably are mediated by CICR) in rats, and the ease with which continuous, voltage-operated release can be demonstrated in the mammal. In all, it appears that CICR is fundamental in the frog, where it determines sparks and the pronounced, steeply voltage-dependent peak of the release waveform, while it is much less important, perhaps absent, in the rat. Because sparks were present in myotubes from embryonic or neonatal mice, which unlike adult cells express RyR isoform 3, Shirokova et al. (1998) proposed that the production of sparks could require RyR3 (or the corresponding β isoform in frog muscle). This possibility is now bolstered by the observation of sparks in dyspedic mouse cells expressing RyR3 but not RyR1 (see Schneider, 1999, in this issue).

Finally, the molecular makeup of frog triads is usually assumed to be the same as that in mammals, but this may not be the case. The ratio of ryanodine-binding sites to DHPrs expected from the structural model of Block et al. (1988), in which four voltage sensors face every other release channel, is 0.5. This value is found approximately in membrane fractions of rabbit or human muscle (Bers and Stiffel, 1993; Margreth et al., 1993; Anderson et al., 1994). In frogs, the value is ∼1.5 (Margreth et al., 1993; Anderson et al., 1994), which would suggest an excess of release channels, either outside the double row or violating the pattern in some other way. The possibility of major structural differences must be kept in mind, as the morphological alignment of mammalian and fish junctions has never been confirmed in the frog (Franzini-Armstrong and Jorgensen, 1994).

Models of Spark Generation

The “couplon” model (Stern et al., 1997) is the latest of a class that assumes interactions within arrays of channels. The couplon, consisting of the contiguous array of RyRs associated with one side of a junctional segment of transverse tubule, includes two rows of RyRs that alternate between those associated with a DHPr tetrad, presumed to be controlled only by voltage (V channels), and those lacking contact with DHPrs, which are assumed to be activated and inactivated by Ca²⁺ alone (C channels). With suitable parameters, this model reproduces many of the global features of frog skeletal muscle E-C coupling (including its “quantal” aspects; Pizarro et al., 1997), which would be difficult to explain if release resulted from channels gating independently.

The couplon model normally generates spark-like release packets. Openings of V channels trigger, by CICR, openings of adjacent C channels, which in turn may recruit other C channels. The whole couplon, or a part of it, may thus activate to generate a phenomenon similar to a spark. Given the assumed C channel unitary current, 0.3 pA, and the size of a couplon (10–30 C channels), the simulated spark sizes are consistent with the observations. Most interestingly, the couplon sparks have a preferred amplitude (Stern et al., 1997), in agreement with amplitude distributions obtained by González et al. (1999) in the presence of caffeine.

Could single channel openings produce sparks with a preferred amplitude? As shown in Fig. 1, the amplitude of a spark due to a single channel opening should be nearly proportional to the opening duration. For a single Markovian channel, gating reversibly, the distribution of open times is a sum of positively weighted decaying exponentials (Colquhoun and Hawkes, 1995), so a mode in the distribution of amplitudes would not be expected. However, RyRs of striated muscle can be activated by its own permeating Ca²⁺ under suitable conditions in lipid bilayers (Tripathy and Meissner, 1995; Xu and Meissner, 1998). This would couple the free energy of Ca²⁺ permeating down its electrochemical gradient to the gating process, permitting irreversible gating kinetics and open time distributions with a mode (Colquhoun and Hawkes, 1995).

The couplon simulations reproduce the sharp peak observed in whole-cell determinations of Ca²⁺ release in the frog as essentially a sum of sparks, while the steady level that follows the peak is accounted for as a combination of sparks and continuous, directly voltage-operated release. This picture of the peak of release as composed of sparks and mediated by CICR justifies that tetracaine eliminates sparks together with the peak (Shirokova and Ríos, 1997), and that the peak can be reconstructed by superposition of sparks whose timing is determined in fibers with partially inactivated voltage sensors (Klein et al., 1997).

The couplon, and related multichannel models, explain the increase of the peak relative to the steady level as voltage is increased, as a result of cooperation of multiple voltage-operated sources within the same couplon to produce a level of Ca²⁺ that will elicit sparks. Such models predict that partial inactivation of voltage sensors or interference with CICR may be compensated by recruiting more voltage sensors, hence explaining that DHPr antagonists (Shirokova and Ríos, 1997), high intracellular Mg²⁺ (Kirsch et al., 1999), or BAPTA (Brum and Pizarro, 1998), inhibit sparks or the peak phase of release at low, but not at high voltages.

The most parsimonious model (release entirely composed of sparks reflecting opening of one or two channels operated by one voltage sensor) is ruled out by its prediction that disabling some voltage sensors would just scale down release and spark numbers. Instead, the waveform reconstructed in partially inactivated fibers (Klein et al., 1997) showed a much smaller steady component of release flux at high voltage than in fully primed fibers (Pape et al., 1995; Shirokova et al., 1996), which again suggests the existence of a voltage-operated release not in discrete events, and required higher voltages to produce a peak of release (compared with fully primed fibers; Klein et al., 1996), which again indicates cooperation among multiple voltage sensors to activate the same release unit.

Terminating the Spark

As we stated at the outset, sparks were not expected. Their spatial discreteness could have been predicted on structural grounds (in fact, spatially segregated release at triads had been demonstrated earlier by Escobar et al., 1994), but their temporal brevity was startling. Rise times of sparks, which roughly measure release duration, are quite independent of triggering voltage, partial inactivation (Klein et al., 1997), or Ca²⁺ channel blockers (López-López et al., 1995; Santana et al., 1996), which suggests a termination mechanism intrinsic to the release unit, rather than the trigger.

The models of spark generation described above underscore the difficulties in devising a mechanism with robust termination. Local depletion of SR Ca²⁺ has been ruled out in cardiac myocytes because Ca²⁺ sparks may last up to a few seconds (Cheng et al., 1993; Parker and Wier, 1997; Xiao et al., 1997; Lukyanenko et al., 1998). In skeletal and cardiac muscle, images of sparks arising repetitively from the same unit also indicate that depletion does not determine spark termination. Sham et al. (1998) examined the issue in cardiac myocytes, where sparks are triggered by Ca²⁺ influx through membrane channels. Observing that reopenings of membrane Ca²⁺ channels, even with duration greatly prolonged by FPL64176, failed to trigger Ca²⁺ release, they concluded that the termination mechanism includes desensitization to the trigger. Furthermore, they showed that this loss of sensitivity is not an adaptation (in the sense proposed by Györke and Fill, 1993, whereby channels that lose sensitivity can be reactivated by higher [Ca²⁺]) because Ca²⁺ release activated by the tail membrane current at the end of a pulse involved only channels that were not activated at the beginning of the pulse. Taken together, these results indicate that Ca²⁺ sparks are terminated primarily by inactivation of RyRs. It remains to be established whether the inactivation is a Ca²⁺-dependent process (Fabiato, 1985; Schneider and Simon, 1988; Jong et al., 1995), or is “fatal” (Pizarro et al., 1997), an obligatory coda to channel opening.

In any case, the implementation of such mechanisms is more difficult if sparks are multichannel events, simply because it is more difficult to turn-off many channels synchronously. Inter-channel allosteric interactions (Marx et al., 1998) might help, channel closing could be synchronized by negative interactions between RyRs in the quasi-crystal of the junction (Stern et al., 1999).

In summary, sparks appear to constitute the totality of physiologic Ca²⁺ release in cardiac muscle, and a major portion of it in frog skeletal muscle. The early peak of release during a voltage pulse is largely constituted by a superposition of sparks, a conclusion that can probably be extrapolated to release in response to an action potential. Release not constituted by separable events is found under various conditions in skeletal muscle; it might be the trigger of sparks in frogs, and the sole form of release in mammals. The origin of sparks is unsettled. The release flux estimated for the largest sparks seems too high to be carried by just one channel. The shift of voltage dependence of spark activation upon partial inactivation of voltage sensors and other complex properties of the whole-cell release waveform are better understood if channels engage in group interactions. Finally, if sparks were one-channel events, it would be difficult to explain why most release channels would consistently fail to activate when facing the high triadic gap [Ca²⁺] associated with a typical spark. On the other hand, the rapid and effective termination of individual sparks can be more easily justified if sparks result from the opening of individual release channels. The distribution of spark amplitudes corrected for the distorting effects of line scanning may exhibit a modal amplitude, which is more easily explained with multichannel models of sparks, but may also be a feature of single channels that gate irreversibly. In general, it is very difficult to account for the properties of sparks, or whole-cell Ca²⁺ release, by simply extrapolating the properties of individual channels in bilayers, which suggests that interactions with other triadic proteins, including release channels, and other forms of local modulation, may crucially influence physiologic gating.

Many questions therefore remain unanswered. Going back to the stars metaphor, the trek is far from over.