Multiple Modes of Calcium-Induced Calcium Release in Sympathetic Neurons II: A [Ca²+]_i- and Location-Dependent Transition from Endoplasmic Reticulum Ca Accumulation to Net Ca Release

Neurons respond to depolarizing stimuli with a rise in the concentration of free cytosolic calcium ([Ca²+]_i). This rise, initiated by Ca²+ entry through voltage-gated channels, is strongly influenced by the activity of intracellular calcium stores (Pozzan et al. 1994; Berridge 1998; Verkhratsky and Petersen 1998). One important calcium store, operationally defined by its sensitivity to caffeine and ryanodine, is a locus for CICR. A large body of work in a variety of neurons indicates that CICR accelerates stimulus-evoked Ca²+ elevations (Kuba 1994; Verkhratsky and Shmigol 1996; Alonso et al. 1999; Emptage et al. 1999; Sandler and Barbara 1999; Usachev and Thayer 1999). In sympathetic neurons, evidence pointing to such a role for CICR comes from the effects of caffeine and ryanodine on stimulus-evoked [Ca²+]_i elevations and the supralinear dependence of these elevations on stimulus duration (Lipscombe et al. 1988; Thayer et al. 1988; Friel and Tsien 1992; Kuba 1994). As in other cells, CICR is thought to accelerate evoked [Ca²+]_i responses by promoting net Ca²+ release from the ER. However, we have presented evidence in the companion paper (see Albrecht et al. 2001, in this issue) for an alternative outcome during weak stimulation, namely, net Ca accumulation at a rate that is modulated by a Ca²+-release pathway. In this case, the ER acts as a Ca²+ buffer, but one whose strength is downregulated by the activation of a [Ca²+]_i-dependent, ryanodine-sensitive Ca²+ release pathway as [Ca²+]_i rises during stimulation.

In that study, a model was presented that describes the interplay between [Ca²+]_i-dependent uptake and release processes as the critical factor in determining the direction and rate of net ER Ca²+ transport. This model distinguishes three distinct “modes” of CICR; it accounts for Ca²+ buffering at low [Ca²+]_i (Mode 1), but also predicts that in certain cases buffering strength will be further attenuated, and eventually eliminated, by graded activation of the Ca²+ release pathway. If activation of the CICR pathway is sufficiently strong at high [Ca²+]_i that the rate of Ca²+ release can exceed the rate of uptake, stimulation should lead to net Ca²+ release. In one mode (moderate [Ca²+]_i, Mode 2) the transition is gradual, whereas for the other (high [Ca²+]_i, Mode 3) CICR may be regenerative, although both cases correspond to classical CICR and are expected to amplify depolarization-evoked [Ca²+]_i elevations in concert with a reduction in the Ca load of the ER.

The present study tests several predictions of this model by directly measuring changes in total calcium concentrations within the ER ([Ca]_ER) that occur in response to changes in [Ca²+]_i during depolarization and/or in the presence of agents that affect intracellular Ca²+ transport. A basic analysis of elemental composition and Ca²+-handling characteristics confirmed that the ER is the structural correlate of the ryanodine-sensitive Ca²+ store. We then examined how [Ca]_ER changes during and after steady depolarization as a function of [Ca²+]_i, time, and intracellular location. The results show that as depolarization-induced increases in [Ca²+]_i become larger, there is a switchover from Ca²+ buffering to net Ca²+ release, providing direct experimental support for Mode 2 or 3 CICR in these cells. They also reveal a dependence of ER Ca²+ handling on radial position within the cell that is reciprocal to the described previously radial gradient of mitochondrial Ca content (Pivovarova et al. 1999a), the shape and magnitude of which is refined by calcium maps obtained from electron energy loss spectrum images.

Preparations of bullfrog sympathetic neurons, as isolated cells in primary culture or as dispersed ganglia, were obtained as described previously (Pivovarova et al. 1999a; Colegrove et al. 2000); these papers also describe measurements of cytosolic free calcium concentrations ([Ca²+]_i).

The concentration of total calcium within individual cisternae of ER ([Ca]_ER) was measured by energy-dispersive X-ray (EDX) microanalysis of freeze-dried cryosections obtained from rapidly frozen dispersed ganglia, as described (Pivovarova et al. 1999a; see also Buchanan et al. 1993; Pozzo-Miller et al. 1997). Briefly, specimens were frozen at defined times and conditions by impact against a liquid nitrogen-cooled copper block using a modified CF-100 freezing machine (LifeCell). Cryosections (80-nm nominal thickness) were prepared from the well frozen specimen face by means of Leica Ultracut S/FCS or Ultracut E/FC-4E ultracryomicrotomes. Sections mounted on carbon- and Formvar-coated grids were cryotransferred into an analytical electron microscope (model EM912 Omega; LEO Electron Microscopy) or into an HB501 field-emission scanning transmission electron microscope (VG Instruments) and freeze-dried at about −110°C. EDX spectra were recorded at about −170°C in the LEO 912 Omega, which was equipped with a Linksystem Pentafet EDX detector (Oxford Instruments) and a ProScan slow-scan CCD camera controlled by AnalySIS software (Soft-Imaging Software GmbH). A nominal probe size of 63 nm was used; smaller probes yielded essentially similar results, confirming that the ∼63-nm probe was adequate to determine ER content without significant contamination from adjacent cytosol. To avoid overlying cytosol, only cisternae wholly contained within the section were analyzed.

Depolarization-induced changes in [Ca]_ER were measured by transferring ganglia from normal Ringer's to depolarizing media (30 K⁺ or 50 K⁺ Ringer's, isotonic substitution of Na⁺) for times specified. For 10-s depolarizations, the solution change was accomplished by continuous superfusion on the stage of the freezing machine. Repolarization experiments were performed by transferring ganglia that had been depolarized for 2 min in 50 K⁺ back to normal Ringer's for specified times. The effect of ryanodine was assayed by preincubating ganglia in normal Ringer's containing 1 μM ryanodine plus 10 mM caffeine for 5 min; the latter facilitates the use-dependent effects of ryanodine on the CICR store. For [Ca]_ER measurements in unstimulated cells, ganglia were then transferred to normal Ringers plus 1 μM ryanodine without caffeine 1 min before rapid freezing. For measurements of depolarization-evoked [Ca]_ER changes, ganglia were preincubated with caffeine plus ryanodine followed by ryanodine only as described, then transferred to high K⁺ Ringer's plus ryanodine before rapid freezing.

Spectrum images (EELSI's) are x-y arrays of pixels, i.e., images, containing a segment of an electron energy loss spectrum at every pixel. EELSI's were acquired using a Gatan model 766 DigiPEELS (parallel-detection electron energy loss spectrometer; Gatan, Inc.) mounted on the HB501 STEM, as described previously (Leapman and Rizzo 1999). A cold-field emission source in this instrument provides a probe of diameter ∼1 nm with a current of ∼1 nA. The spectrometer was equipped with a 1024-channel photodiode array detector with a read-out noise of ∼2 counts (rms; equivalent to ∼20 fast electrons). Frozen-hydrated cryosections were transferred into the STEM at liquid nitrogen temperature and were freeze-dried at −100°C before recooling to −160°C for analysis. Annular dark-field images and EELS-images were recorded using a Gatan DigiScan acquisition system together with Digital Micrograph software (V3.3.1) running on a Power Macintosh 9600/233 computer that incorporated the Gatan spectrum-imaging package of Hunt et al. 1999. The system enables EELSI's to be acquired from specimen regions defined on an initially recorded dark-field image. This software also corrects for specimen drift by acquiring at given intervals a fast dark-field STEM image from a defined region of the specimen. Each successive image is cross-correlated with the previous one and the probe position is adjusted accordingly. Drift correction was applied at the end of each scan line. An acquisition time of 0.4 s per pixel was selected, which gave typical image collection times of 10–20 min. The spectrum-images were corrected for read-out noise, dark-current, and channel-to-channel gain variation at each pixel.

Calcium and carbon maps were computed by fitting the background in a pre-edge window according to an inverse power law. The extrapolated background was subtracted from the post-edge window to obtain the integrated elemental signals for the Ca L₂₃ and C K core-edges at each pixel in the image. Post-edge windows of width Δ = 10 eV (from 345 to 355 eV) for calcium and width Δ = 50 eV (from 285 to 335 eV) for carbon were selected. Pre-edge windows of similar width were selected to perform the background extrapolation. Atomic fractions of N_Ca/N_C were computed from the ratios of signal intensities I_Ca/I_C by dividing by the ratios of partial scattering cross-sections obtained from the Gatan EL/P program. The calcium concentrations in mmol/kg dry weight were obtained from the measured atomic fraction of N_Ca/N_C using the formula: [Ca] = N_Ca/N_C × 4.5 × 10⁴ (Leapman et al. 1993). Values of the partial cross sections for the spectrometer collection semi-angle of 20 mrad were 5.96 × 10³ barns for Ca (Δ = 10 eV) and 4.99 × 10³ barns for C (Δ = 50 eV).

Population results are expressed as mean ± SEM. Statistical significance was assessed using a sequentially rejective Bonferroni procedure for multiple comparisons against a single control, FWE = 0.05 (Holm 1979). Fura-2-AM was obtained from Molecular Probes. Unless otherwise stated, all reagents were obtained from Sigma-Aldrich, whereas electron microscopy supplies were obtained from EMS or Ted Pella.

Sympathetic neurons express an abundant network of ER (Fig. 1). They also contain internal Ca stores that take up Ca²+ using sarco- and endoplasmic reticulum Ca ATPase (SERCA) pumps and release this ion mainly via a caffeine- and ryanodine-sensitive CICR pathway (Kuba and Nishi 1976; Thayer et al. 1988; Friel and Tsien 1992; Cseresnyés et al. 1997). It is generally presumed that these mechanisms for Ca²+ homeostasis are invested in the ER, in which case several testable predictions follow. First, the ER should contain Ca at a high concentration under resting conditions. Second, the concentration of intraluminal total Ca should be reduced by inhibiting the SERCA pump-dependent uptake pathway. Third, the concentration of Ca should be reduced by caffeine and/or ryanodine. Finally, conditions that favor Ca accumulation should lead to an increase in the Ca content of the store.

To test these predictions, changes in the mean total Ca concentration within the ER ([Ca]_ER) were measured by EDX microanalysis of rapidly frozen cells in isolated ganglia, in parallel with measurements of free cytosolic Ca²+ ([Ca²+]_i) in fura-2–loaded cultured neurons. The first observation is that resting [Ca]_ER, spatially averaged over the cell soma, was notably high, 12.8 ± 0.9 mmol/kg dry weight (±SEM; equivalent to ∼3.6 ± 0.3 mmol/liter hydrated tissue; Table, Fig. 2 A; see Pozzo-Miller et al. 1997 and Pivovarova et al. 1999a for an explanation of the concentration unit mmol/kg dry weight, and how to convert it to mmol/liter hydrated tissue). This value is higher than in any other organelle or compartment, including the cytoplasm (3.3 ± 1.0 mmol/kg dry weight). A second set of measurements showed that inhibiting Ca²+ uptake with Tg (1 μM) significantly reduced [Ca]_ER (Table and Fig. 2 A), consistent with the observation that Tg elicits a transient rise in [Ca²+]_i (Fig. 2 B), even in the absence of extracellular Ca²+ (not shown). Additional experiments indicated that RyR activation stimulated net Ca²+ release from the ryanodine-sensitive store. Thus, exposure first to ryanodine plus caffeine and then briefly to ryanodine alone reduced [Ca]_ER (Table and Fig. 2 A, Ryan, and Fig. 2 C) to similar levels as did Tg. Finally, sympathetic ER can, as a consequence of Ca²+ entry across the plasma membrane, accumulate Ca above basal levels; for example, [Ca]_ER is elevated to 17.0 ± 1.6 mmol/kg dry weight by weak depolarization that raises [Ca²+]_i to ∼300 nM (30 mM K⁺ for 2 min; Table; see Fig. 3 in Albrecht et al. 2001, in this issue). No other cellular compartment, including the mitochondria, Golgi apparatus, or nucleus, was affected by these drug treatments (not shown). Thus, the ER and only the ER had the pharmacological signature of the CICR pool.

These results identify the ER as the principal, probably the only, Ca store responsible for CICR, but leave open the question of whether the ER Ca pool is a single, homogeneous functional entity. Arguments based on the effects of ryanodine on basal Ca²+ uptake rates and supporting a single pool model were presented in the companion paper (see Albrecht et al. 2001, in this issue). Here, additional support comes from the observation that the distributions of [Ca]_ER measurements from individual ER cisternae are normal under conditions of rest and depletion (Fig. 2A and Fig. B, insets). Although there is a measurable inositol trisphosphate (InsP₃)-gated Ca²+ release pathway in amphibian sympathetic neurons (Pfaffinger et al. 1988), the lack of evidence for discreet pools may reflect the comparatively small size of this pool. Alternatively, it may indicate that the ryanodine- and InsP₃-sensitive pools are coextensive, or at least in functional communication, so that manipulating one pool affects the other.

In the companion paper (see Albrecht et al. 2001, in this issue), we presented evidence that during weak depolarization the ER accumulates Ca²+, but net accumulation is opposed, and therefore attenuated, by activation of a Ca²+- and ryanodine-sensitive Ca²+ release pathway. A predicted property for two of the three general types of CICR discussed in that study is a transition from Ca²+-dependent Ca accumulation to net Ca²+ release—a change in the direction of net ER Ca²+ transport—as stimulus-evoked [Ca²+]_i elevations become larger. To determine if such a transition occurs, we examined changes in [Ca]_ER induced by stimuli that increase [Ca²+]_i to higher levels. Isolated sympathetic ganglia were exposed to a various high K⁺ Ringers' solutions for defined periods of time before rapid freezing and EDX analysis; [Ca²+]_i was determined in fura-2–loaded, isolated cells in parallel experiments.

Fig. 3 A illustrates a typical [Ca²+]_i response to strong depolarization (50 mM K⁺, 2 min) elicited in the absence of drugs or CICR modifiers. During depolarization, [Ca²+]_i rapidly rises to a peak and declines toward a steady level of ∼600–800 nM that is sustained for several minutes. Under these conditions, mitochondria avidly and continuously accumulate Ca (Pivovarova et al. 1999a), but [Ca]_ER does not change significantly at either 45 s or 2 min (Fig. 3 D, middle data points and Table). This finding is in contrast to the increase in [Ca]_ER induced when [Ca²+]_i is ∼300 nM under otherwise equivalent conditions (Fig. 3 D, top left data point). One interpretation is that although both Ca²+-dependent Ca²+ uptake and release by the ER are accelerated when [Ca²+]_i reaches ∼700 nM, the average rates of these processes are equal. In effect, the ER is cycling Ca²+, i.e., “spinning its wheels.”

In contrast to untreated cells, cells depolarized in the presence of 1 μM FCCP attain [Ca²+]_i levels in excess of 1 μM for ∼1 min before undergoing a rapid decay to a steady, low level of ∼200–300 nM (Fig. 3 B). After repolarization, [Ca²+]_i recovers to its prestimulus level with a plateau phase that is greatly reduced compared with the plateau observed under control conditions. The larger [Ca²+]_i response and the more rapid recovery are both consistent with suppression of mitochondrial Ca²+ uptake. Quite strikingly, depolarization in the presence of FCCP (45 s) reduces [Ca]_ER to 6.5 ± 1.0 mmol/kg (Fig. 3 D, bottom right data point), showing that when [Ca²+]_i is very high, the ER loses net Ca. Net ER Ca²+ release would also explain the time course of [Ca²+]_i, especially the >1-min-long “hump” (Fig. 3, compare B with C) seen under these conditions. The hump consists of a prolonged [Ca²+]_i elevation followed by a rapid decline (at arrow, Fig. 3 B). In cells treated with Tg, in addition to FCCP, the hump is not observed, consistent with the idea that it reflects net Ca²+ release from the ER. The basis for the accelerated recovery after ∼2 min of FCCP exposure is not known, but, regardless of specific mechanisms, the prompt return to normal resting [Ca²+]_i levels after repolarization indicates that exposure to FCCP had not depleted ATP sufficiently to affect normal Ca²+ clearance mechanisms; if it had, a return to a much higher basal [Ca²+]_i level would be expected.

To summarize, there is a biphasic relationship between the [Ca²+]_i levels reached during depolarization and parallel changes in [Ca]_ER (Fig. 3 D). Increasing [Ca²+]_i from basal levels to ∼300 nM leads to ER Ca accumulation, whereas with larger [Ca²+]_i elevations, [Ca]_ER declines. To this point, the evidence for this relationship at high [Ca²+]_i depends on measurements obtained while FCCP was present for the purpose of inhibiting mitochondrial Ca²+ uptake. FCCP is generally thought to be a highly specific protonophore without short-term effects on mitochondrial structure at the concentration and exposure times used here (Stout et al. 1998; Minamikawa et al. 1999), although uncouplers have been reported to promote mitochondrial swelling under some circumstances (Garlid and Nakashima 1983; Petronilli et al. 1993). However, we did not observe such effects and independent evidence for net ER Ca²+ release at high [Ca²+]_i, obtained under conditions of intact Ca²+ homeostasis and presented below, argues against the possibility that the [Ca]_ER reduction illustrated in Fig. 3 D is an indirect effect of inhibiting mitochondrial function.

Results just presented support the idea that with increasing [Ca²+]_i, ER Ca accumulation slows, leading to lower and eventually to no net accumulation as Ca²+ uptake and release come into balance. The finding that such a steady-state condition occurs when [Ca²+]_i is ∼600 nM (Fig. 3 D) is further reinforced in Fig. 4, a plot of [Ca]_ER as a function of time during and after strong depolarization, which shows that [Ca]_ER changes little if at all during maintained strong depolarization. (The apparent, statistically insignificant decline at 10 s is discussed below). This behavior contrasts with the very large increases in the concentration of total mitochondrial Ca ([Ca]_mito) that occur under the same conditions (Fig. 4; Pivovarova et al. 1999a).

Upon repolarization, intracellular Ca²+ dynamics change dramatically as [Ca²+]_i recovers. How does [Ca]_ER respond during this recovery phase? After repolarization (from 50 mM K⁺, 2-min depolarization), [Ca²+]_i recovers in several kinetically distinct phases, the first two of which are strongly influenced by mitochondrial Ca²+ transport. The earliest phase is speeded by mitochondrial Ca accumulation, whereas the long second phase is slowed by net mitochondrial Ca²+ release (Fig. 3 A; Colegrove et al. 2000). During the second phase, when mitochondria actively extrude Ca²+ via the Na⁺/Ca²+ exchanger, [Ca²+]_i is “clamped” at ∼200–300 nM (Fig. 3 A) while [Ca]_mito declines (Fig. 4, Fig. 2,Fig. 3,Fig. 4,Fig. 5 min). Over this same period of the recovery, [Ca]_ER rises significantly (Fig. 4), which is consistent with the [Ca]_ER rise observed when [Ca²+]_i is within this same range during steady weak depolarization (see Albrecht et al. 2001, in this issue).

Fig. 4 also shows that [Ca]_ER is elevated for >10 min after repolarization-induced loading. Average [Ca]_ER in many cells has begun to decline by 15 min, although that is not evident in Fig. 4, owing to substantial cell–cell variability (which also accounts for the large error bars at 7 and 17 min). In any case, [Ca]_ER recovers quite slowly. Presumably, this reflects a relatively small imbalance between the rates of ER Ca²+ uptake and release, as well as potentially slow rates of release at low [Ca²+]_i, both factors contributing to the slow tail phase of [Ca²+]_i recovery that outlasts the period of mitochondrial Ca²+ release (Fig. 3 A). These observations are reminiscent of similar findings for the ER in these and other neurons (Shmigol et al. 1994a; Friel 1995; Garaschuk et al. 1997; Pozzo-Miller et al. 1997) and suggests that long-lived Ca loads, carrying a “memory” of activity-evoked changes in [Ca²+]_i, may be a general feature of neuronal ER.

The results presented so far indicate that the ER accumulates Ca at low [Ca²+]_i, releases net Ca²+ at high [Ca²+]_i, and transports little net Ca²+ at intermediate levels. These conclusions are based on spatially averaged measurements of [Ca]_ER from multiple cells after steady depolarizations that were much longer than necessary for decay of the initially large, stimulus-evoked radial [Ca²+]_i gradients, which dissipate in only a few seconds in these cells (Hernandez-Cruz et al. 1990; Hua et al. 1993b). During strong depolarization that raises [Ca²+]_i to ∼600–800 nM, such measurements do not detect significant changes in average [Ca]_ER, but there is a tendency for (averaged) [Ca]_ER to become lower as the length of depolarization shortens (Fig. 4). Therefore, we considered factors that might quantitatively depress spatially averaged [Ca]_ER at short depolarization times. One potential contribution is spatial nonuniformity of [Ca²+]_i during initial periods of Ca²+ entry. We have provided evidence for the existence of radial gradients in the concentration of mitochondrial total Ca ([Ca]_mito) after long depolarization times (e.g., 45 s), and proposed that mitochondria maintain a record of previous, spatially heterogeneous [Ca²+]_i, signals when the stimulus raises [Ca²+]_i, high enough that mitochondria are forced into their so-called “buffering mode,” thereby accumulating Ca continuously (Pivovarova et al. 1999a). Therefore, we asked if [Ca]_ER might similarly provide a record of radial [Ca²+]_i, gradients, one that conforms to the [Ca²+]_i dependence of ER Ca²+ transport.

An analysis of [Ca]_ER as a function of distance from the plasma membrane in cells depolarized briefly (10 s) with 50 mM K⁺ reveals that net Ca²+ release does in fact occur, but only in the outermost 5-μm shell of cytoplasm (Fig. 5 A). One likely explanation for this [Ca]_ER gradient is that it reflects the transient radial gradients in [Ca²+]_i that arise from Ca²+ entry through voltage-sensitive Ca²+ channels during the initial seconds of depolarization (Hua et al. 1993b). During this period, [Ca²+]_i would reach its highest levels in the outermost cytoplasmic shells, after which channel inactivation would reduce Ca²+ entry and [Ca²+]_i gradients would dissipate by diffusion. Since [Ca²+]_i in the more interior shells would not exceed the equilibrium spatial average, which is ∼600–800 nM in this case, [Ca]_ER in these shells would change little, as observed. Furthermore, the dominance of these shells in the spatially averaged [Ca]_ER values would largely obscure any local decline of [Ca]_ER within peripheral cisternae (Fig. 4).

According to this general idea, radial [Ca²+]_i gradients drive spatial heterogeneities of [Ca]_ER, which, therefore, should be increasingly prominent as the strength of depolarization increases. In addition, for a stimulus of given strength, gradients would be most pronounced during the initial moments of stimulation. As [Ca²+]_i gradients dissipate, [Ca]_ER would relax, independently of location, toward a value that depends on (now spatially uniform) [Ca²+]_i. In contrast, such a relaxation would not occur for [Ca]_mito if [Ca²+]_i remained above the level where mitochondria continuously accumulate Ca²+. Thus, it appears that different regimes of CICR can coexist in different parts of the cell at a given time, with consequent effects on the direction of net Ca²+ transport. Viewed another way, the radial gradient of [Ca]_ER observed after short depolarization recapitulates in a single cell the [Ca²+]_i dependence of CICR first deduced from spatially averaged [Ca²+]_i measurements after long depolarizations in cell populations (Fig. 3 D).

We have previously reported that mitochondria within a similarly sized ∼5-μm peripheral shell of cytoplasm accumulate disproportionately large amounts of Ca (Pivovarova et al. 1999a), and suggested that the resulting radial [Ca]_mito gradient follows from differences in local [Ca²+]_i to which central and peripheral mitochondria are exposed during the history of stimulation. Modeling studies are consistent with this idea, and suggest that mitochondria, owing to the steep [Ca²+]_i dependence and integrative nature of mitochondrial Ca²+ transport, should serve as highly sensitive detectors of spatial nonuniformities in [Ca²+]_i. However, the precision of the [Ca]_mito measurements supporting these conclusions was limited by the punctate nature of mitochondrial Ca sequestration. Specifically, we and others (David 1999; Pivovarova et al. 1999a,Pivovarova et al. 1999b) have reported preliminary evidence that evoked mitochondrial Ca²+ uptake leads to the transient formation of mainly small (<20 nm) calcium- and phosphorus-rich inclusions within the mitochondrial matrix. It has been speculated that these inclusions reflect a high capacity, chemical mechanism of mitochondrial Ca sequestration, but, regardless, their formation results in substantial heterogeneity in the distribution of intramitochondrial Ca, and therefore also to scatter for EDX point-probe analyses.

Here, we have applied a novel analytical approach—electron energy loss spectrum imaging (EELSI; Leapman et al. 1994)—that largely overcomes this difficulty because it comprehensively maps elemental distributions at better than 20-nm resolution (Leapman and Rizzo 1999). To generate an EELSI, a high resolution EELS spectrum, with its inherently high Ca sensitivity, is acquired at every pixel over a selected field. Major advantages are that the specimen is exhaustively sampled, with an energy loss spectrum at each pixel that is suitable for rigorous quantitation. As a result, EELSI can detect microheterogeneity at the single pixel level, and provide a dataset suitable for determining the shape and magnitude of the depolarization-induced [Ca]_mito gradient with much improved resolution and precision. The refined radial gradient for [Ca]_mito (Fig. 5 B, compare Fig. 8 in Pivovarova et al. 1999a) was determined after stimulation of the same strength and duration (50 mM K⁺, 45 s) used previously; qualitatively similar gradients can be detected as early as 10 s after depolarization. The results are consistent with the idea that radial [Ca²+]_i gradients lead to differences in the rate of mitochondrial Ca accumulation and, therefore, to the mitochondrial Ca load, and thus provide additional evidence that peripheral ER are exposed to a significant, spatially heterogeneous [Ca²+]_i signal. Finally, it is clear (the time difference between Fig. 5A and Fig. B notwithstanding) that the gradient in [Ca]_mito is reciprocal to that for [Ca]_ER. Therefore, the spatial distribution of ER Ca²+ release, like the temporal distribution of Ca²+ uptake, is reciprocal and complementary to that of mitochondrial Ca²+ transport.

The goal of this study was to characterize the functional contributions of the ryanodine-sensitive Ca store to depolarization-evoked changes in [Ca²+]_i, with a particular eye toward determining whether in sympathetic neurons this store acts as a Ca²+ source or as a sink. The study was prompted by our seemingly paradoxical observation that, although this store exhibits many of the hallmarks of classical CICR (Kuba 1994; Verkhratsky and Shmigol 1996; Usachev and Thayer 1999), under conditions of weak depolarization, it actually accumulates Ca and slows evoked [Ca²+]_i elevations (Albrecht et al. 2001). Therefore, we extended those studies to include larger evoked [Ca²+]_i elevations, and find that although the ER is a graded Ca²+ buffer over the lower range of [Ca²+]_i, progressive [Ca²+]_i-dependent activation of its Ca²+ release pathway leads to weaker buffering, followed by a transition to net Ca²+ release, until finally the ER becomes a triggered Ca²+ source at high [Ca²+]_i. These findings provide a conceptual framework for understanding previous studies of ER Ca²+ handling in these and other cells (Friel and Tsien 1992; Hua et al. 1993a; Shmigol et al. 1995).

Since an important aspect of our approach depended on directly measuring changes in the Ca content of the ryanodine-sensitive store, it was necessary to know the corresponding subcellular structure. The Ca content of the ER ([Ca]_ER), as well as its pharmacological profile, directly identifies the ER as the principal, if not the only, structural correlate of this Ca store. Although this result is not surprising, it was essential to demonstrate, considering that other organelles may function as Ca stores (Verkhratsky and Petersen 1998). Furthermore, even if the identity of the store is taken for granted, the amount of Ca it sequesters is critical for understanding the magnitude and direction of CICR. Note that other functions for the ER, such as InsP₃-gated Ca²+ release, spatially segregated or not, are not precluded by these observations, but, for reasons outlined above, regions of ER representing these other functions must be fairly small and/or in communication with RyR-rich regions.

The companion article (see Albrecht et al. 2001, in this issue) introduced the concept that the particular complement of Ca²+ transport systems in a given cell specifies the availability of three distinct modes of CICR. One main experimental finding of the present study is that when [Ca²+]_i is maintained in the range ∼600–800 nM, there is little net change in [Ca]_ER, suggesting that at these [Ca²+]_i levels, enhanced CICR via RyRs largely balances Ca²+ uptake, even though both inward and outward fluxes are larger than basal levels. Therefore, we conclude that sympathetic neurons express either Mode 2 or Mode 3 CICR, since these are the types characterized by a transition to net Ca²+ release.

After repolarization (from 50 K⁺), [Ca]_ER increased during the plateau phase of recovery, a period during which [Ca²+]_i is maintained at ∼300 nM by net mitochondrial Ca²+ release (Pivovarova et al. 1999b; Colegrove et al. 2000). Since this is a permissive range for ER Ca²+ accumulation (see Albrecht et al. 2001, in this issue) the observation implies that Ca²+ release from mitochondria serves as a stimulus for ER Ca accumulation, thus, effectively determining which CICR mode is active. There is good evidence for stimulus-driven increases in [Ca]_ER or [Ca²+]_ER in other neuronal cell types (Shmigol et al. 1994b; Markram et al. 1995; Garaschuk et al. 1997; Pozzo-Miller et al. 1997). One striking common feature of these increases is their longevity; recovery of [Ca]_ER and/or [Ca²+]_ER to basal levels is extremely slow, typically on the order of tens of minutes. Long-lived changes in [Ca]_ER, with attendant effects on [Ca²+]_i, would undoubtedly influence cellular processes, e.g., electrical excitability or synaptic activity, and, as mentioned, might be a mechanism for storing “memory traces” of previous activity over time frames suitable for processes like protein synthesis and transport. The longevity of [Ca]_ER elevations also suggests future experiments directed toward understanding the extent to which the Ca load of the ER influences the mode and direction of ER Ca²+ transport.

The ER of sympathetic neurons appears to differ from certain other neuronal cell types in at least two important respects. First, the distribution of [Ca]_ER measurements, at rest and in response to depolarization and a variety of drugs, is generally normal. Therefore, the results are consistent with the idea that the somatic ER of sympathetic neurons is a single functional entity, and provide no evidence for the existence of discrete ER domains that differ in their capacity to transport Ca²+. Although this agrees with results from some nonmuscle cell types (Khodakhah and Armstrong 1997), it contrasts with several others, (Golovina and Blaustein 1997; Pozzo-Miller et al. 1997; for review see Meldolesi and Pozzan 1998b). Second, the range of changes in [Ca]_ER of sympathetic neurons, between 5 and 20 mmol/kg, is small, especially in comparison to the huge loads accumulated by, for example, the dendritic ER of CA3 pyramidal cells (Pozzo-Miller et al. 1997). Taken together, these observations imply significant functional diversity of ER from one cell type (or region) to another, so that it remains a goal of future research to explain such heterogeneity, or lack thereof.

Given the acceleration of ER Ca²+ release that occurs with stronger stimuli, we tested whether even larger [Ca²+]_i elevations lead to net ER Ca²+ release. After disabling mitochondrial Ca²+ uptake (with FCCP), the same strong depolarizing stimuli (50 mM K⁺) raised [Ca²+]_i to >1 μM, with a concomitant decline in [Ca]_ER. This decline reflects Ca²+ release as expected during classical CICR, which, in turn, appears to have synergistically amplified the [Ca²+]_i response (Fig. 3 B) beyond that attributable to just Ca²+ entry.

It is increasingly clear that intracellular Ca²+ signaling depends on both time and space (Meldolesi and Pozzan 1998a; Pozzan and Rizzuto 2000). In sympathetic neurons, the spatial component is dominated by an initial Ca²+ flux across the plasma membrane, leading to pronounced radial gradients in [Ca²+]_i (Hernandez-Cruz et al. 1990; Hua et al. 1993b). Similar [Ca²+]_i gradients have been observed in several other cell types, including, for example, Purkinje cells (Eilers et al. 1995) and chromaffin cells (Neher and Augustine 1992; Alonso et al. 1999). In view of the [Ca²+]_i dependence of CICR shown here, we asked whether transient gradients in [Ca²+]_i might influence the spatial profile of [Ca]_ER. In fact, a [Ca]_ER gradient was observed at short times (10 s) during strong depolarization as a decline in [Ca]_ER in the outermost shell adjacent to plasma membrane Ca²+ channels. A plausible explanation is that even though globally averaged [Ca²+]_i rises to only submicromolar levels, [Ca²+]_i just beneath the plasmalemma is high enough to drive CICR into the net release mode. We propose that this is another manifestation of the [Ca²+]_i dependence of net ER Ca²+ transport, and that triggered Ca²+ release from the ER can occur in the absence of drugs that modify Ca²+ homeostasis. Furthermore, this observation reveals that different modes of CICR can coexist in the same cell at the same time, potentially providing a driving force for directional intracisternal Ca²+ diffusion. Although the radial dependence of [Ca]_ER may simply reflect the [Ca²+]_i dependence of CICR, there could well be additional factors. For example, recent reports (Akita and Kuba 1999; McDonough et al. 2000) indicate that RyR's are preferentially expressed in the subplasmalemmal (and perinuclear) region of these cells. A spatial difference in the density of RyR's would undoubtedly affect, at least quantitatively, local expression of CICR (McDonough et al. 2000). Additionally, preferential depletion of peripherally placed cisterns would be an ideal spatial arrangement for activating store-operated channels (although there is presently no evidence for such channels in these neurons). Finally, we note again the likely possibility that the size of the preexisting ER Ca load influences stimulus-evoked changes in [Ca]_ER.

An important component of this study is the confirmation and quantitative refinement of the depolarization-induced radial gradient in [Ca]_mito. We previously reported such a gradient at 45 s after strong depolarization, but the EDX data underlying this analysis was scattered owing to the punctate distribution of mitochondrially sequestered Ca (Pivovarova et al. 1999a) and the point-probe nature of EDX analysis. Here, we report the shape and magnitude of this [Ca]_mito gradient based on calcium maps obtained by EELSI (Leapman et al. 1994; Leapman and Rizzo 1999). The refined EELSI profile of [Ca]_mito versus distance from the plasmalemma (Fig. 5 B) presented here is consistent with previous results from these and other cells (Bito et al. 1996; Wang et al. 1999; Montero et al. 2000) and supports speculation that mitochondrial Ca loading, because of its steep dependence on [Ca²+]_i, retains a spatial record of [Ca²+]_i gradients.

The cytoplasmic shell where maximal mitochondrial Ca²+ uptake occurs coincides with the shell of maximal ER Ca release. This and other observations—for example, the rapid [Ca²+]_i decline followed by a plateau seen after caffeine/ryanodine-evoked dumping of the ER Ca store (Fig. 2 C)—suggest functional coupling between these two organelles and are consistent with the idea of direct Ca transfer within organized, functional complexes (Rizzuto et al. 1994, Rizzuto et al. 1998; Montero et al. 2000). However, they are also consistent with coupling through bulk cytosolic Ca²+. Regardless of mechanistic details, it is clear that over a range of time, space, and [Ca²+]_i, mitochondria and the ER engage in complementary Ca²+ transport, implying some degree of coupling. The functional consequences of such coupling are unknown, but could well be profound. As a specific example, the transformation of the ER into a triggered Ca²+ source when mitochondrial Ca²+ uptake is disabled suggests that mitochondria may be important regulators of CICR mode transitions. On a more general level, either or both organelles could serve as memory reservoirs for important functions such as synaptic potentiation or facilitation (Pozzo-Miller et al. 1997; Berridge 1998; Zucker 1999). Perhaps the ER and mitochondria share such functions, cooperating through overlapping, coordinated time domains within a common local spatial domain (Pozzo-Miller et al. 2000). If this occurs, it remains to be determined whether the ER senses mitochondrial Ca²+ transport, and vice versa, through local or global mechanisms.

We thank Drs. T.S. Reese and E.S. Ralston for helpful discussions and critical evaluation and Dr. C.A. Brantner for excellent technical assistance.

This work was supported by a grant (No. NS-33514) from the NIH to D.D. Friel and by the NIH Intramural Research Program.