Ca2+-induced Ca2+ Release in Sensory Neurons

Ca2+-induced Ca2+ release (CICR) is a ubiquitous mechanism by which Ca2+ release from the endoplasmic reticulum amplifies the trigger Ca2+ entry and generates propagating Ca2+ waves. To elucidate the mechanisms that control this positive feedback, we investigated the spatial and temporal kinetics and measured the gain function of CICR in small sensory neurons from mammalian dorsal root ganglions (DRGs). We found that subsurface Ca2+ release units (CRUs) are under tight local control by Ca2+ entry, whereas medullar CRUs as a “common pool” system are recruited by inwardly propagating CICR. Active CRUs often displayed repetitive Ca2+ sparks, conferring the ability to encode a “memory” of neuronal activity well beyond the duration of an action potential. Store Ca2+ reserve was able to support all CRUs each to fire ∼15 sparks, excluding use-dependent inactivation or store depletion as the major CICR termination mechanism. Importantly, CICR in DRG neurons operated in a low gain, linear regime (gain = 0.54), which conferred intrinsic stability to CICR. Combined with high Ca2+ current density (–156 pA/pF at –10 mV), such a low gain CICR system generated large intracellular Ca2+ transients without jeopardizing the stability. These findings provide the first demonstration that CICR operating in a low gain regime can be harnessed to provide a robust and graded amplification of Ca2+ signal in the absence of counteracting inhibitory mechanism.

CICR, in the form of propagating Ca 2ϩ waves, is also capable of broadcasting an otherwise local signal over the entire cytoplasm as well as nucleoplasm, such as those observed during fertilization (7). Understanding the control mechanisms of CICR is thus fundamental to understanding the regulation of intracellular Ca 2ϩ signaling.
In its simplest form, uncontrolled CICR with high gain amplification is expected to behave in an all-or-none fashion. Yet, CICR in various biological systems is often finely tuned to the magnitude and duration of the trigger Ca 2ϩ signal, mainly Ca 2ϩ influx via Ca 2ϩ currents (I Ca ). In an exemplary system found in mammalian ventricular myocytes, CICR serves to amplify the trigger I Ca at a gain of 10 -70, depending on membrane voltage (8 -10). Studies for more than a decade have established that cardiac CICR system comprises a discrete, rather than continuum, architecture, with ryanodine receptor (RyR) Ca 2ϩ release channels assembled into discrete Ca 2ϩ release units (CRUs) (11); individual CRUs operate in a digital, rather than analogous, mode, generating "Ca 2ϩ sparks" as the elementary events of CICR (12)(13)(14)(15). During cardiac excitationcontraction coupling, spark genesis is under tight control of local Ca 2ϩ influx, which is essential to achieving high gain amplification and stability simultaneously (8, 13, 16 -18). Emerging evidence from intensive recent research indicates that terminating CICR in the heart involves substantial ER Ca 2ϩ store depletion (19 -22) and strong use-dependent inactivation (18,23) or some other inhibitory mechanism (24,25).
Recently, we have demonstrated that Ca 2ϩ sparks from type 3 RyR (RyR3) constitute the elementary events of CICR in small sensory neurons from rat dorsal root ganglions (DRGs). 2 DRG Ca 2ϩ sparks, particularly those localized to surface membrane, play an important role in the regulation of vesicular secretion from the somata of these cells. 2 In the present study, we used DRG neurons as a model system to investigate possible control mechanisms of CICR. We demonstrated that CICR in DRG neurons operates in the low gain, linear amplification regime in conjunction with high I Ca density, which confers intrinsic stability and large Ca 2ϩ transient amplitude in the absence of counteracting termination mechanisms. Our results were compared and contrasted with those from heart muscle cells, a well characterized model system of CICR.

EXPERIMENTAL PROCEDURES
Cell Preparation-Male Sprague-Dawley rats (200 -250 g) were rendered unconscious by exposure to CO 2 and decapitated. The DRGs (C5-L5) were harvested and treated with collagenase (1.5 mg ml Ϫ1 ) and tripsin (1 mg ml Ϫ1 ) at 37°C, as described. 2 Cells were used 2-10 h after preparation. Only the small (15-25 m, C-type) neurons without apparent processes were used. * This work was supported by the Chinese Natural Science Foundation and Major State Basic Science Development Program and by the intramural research fund of the National Institutes of Health (to H. C.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Confocal Ca 2ϩ Imaging-DRG neurons were loaded with the Ca 2ϩ indicator, Fluo-4, either via dialysis through the patch pipette in the broken-in whole-cell patch clamp configuration or incubation with the acetoxymethyl ester form of the indicator (5-10 M, 20 -30 min) in the perforated patch clamp experiments. Indicator-stained DRG neurons were imaged at 488 nm excitation and Ͼ510 nm emission with a Zeiss 510 inverted confocal microscope (Carl Zeiss), in the line scan or curve scan mode. The horizontal and axial resolutions were 0.4 and 1.5 m, respectively, achieved with 40ϫ oil immersion lens (NA 1.3), and the line scan rate was 3.07 ms per scan. In some experiments, 5 mM caffeine was added to nominal zero Ca 2ϩ solution and delivered to cells by locally placed glass pipette. Acquisition of confocal Ca 2ϩ imaging was usually synchronized to electrophysiological commands in whole-cell voltage or current clamp experiments. Image processing and data analysis were performed using custom-devised algorithms coded in Interactive Data Language (Research Systems, Inc., Boulder, CO).
Measurement of the Gain of CICR-The gain of CICR was defined as the ratio between the amount of released Ca 2ϩ for the ER and the trigger Ca 2ϩ influx via I Ca . The I Ca and the subsurface Ca 2ϩ transient were directly measured with whole-cell voltage clamping and confocal curve scan imaging, respectively. A linear relationship between Ca 2ϩ transient and Ca 2ϩ flux via I Ca (Q ICa ) alone was first established in cells which CICR was abrogated by ryanodine pretreatment (10 M, 20 min) (see Fig. 3C). The total Ca 2ϩ flux (Q total ) from both release and influx in cells with intact ER function was then determined from this calibration curve, and the release component was determined by subtracting Q ICa from its corresponding Q total .
Immunolabeling-To visualize RyR3 localization, dissociated DRG neurons were fixed and incubated overnight at 4°C with primary monoclonal antibodies that recognize both RyR1 (not expressed in this type of cells; see Ref. 27) and RyR3 (28). Cells were then incubated with Cy3-conjugated anti-mouse antibodies (Jackson ImmunoResearch Laboratories, Inc.) for 1 h in the dark. Immunofluorescence was detected by the Zeiss confocal microscope with 543 nm excitation, Ͼ560 nm emission, and 1.0 m axial resolution. Confocal Z-stack imaging was performed at Z intervals of 0.5 m. Zeiss LSM software (Carl Zeiss) was used for three-dimensional reconstruction and animation.
Statistics-Data were given as mean Ϯ S.E. The significance of difference between means was determined, when appropriate, using Student's t test and paired t test. A p Ͻ 0.05 was considered statistically significant. Fig. 1A shows typical immunofluorescent staining of RyR3 in a small sensory neuron (diameter ϭ 15-25 m) from rat DRGs, illustrating that intensely stained spots were enriched along a ring right beneath the surface membrane. The bulk cytoplasm was diffusively stained at a reduced intensity, and the nucleus was devoid of any specific immunofluorescence. The average distance between adjacent RyR3 spots was measured to be 1.84 Ϯ 0.10 m (n ϭ 113), and about 400 subsurface RyR3 spots could be identified in a typical cell. These results are in good agreement with our previous report. 2 To determine whether these RyR3 spots are indeed functional CRUs, we employed the so-called Ca 2ϩ spike measurement (9) combined with the "curve scan" imaging technique (Fig. 1B). Specifically, the fast Ca 2ϩ indicator, Fluo-4, along with a slow Ca 2ϩ chelator (EGTA, 10 mM), was dialyzed into the cell through the patch pipette under whole-cell voltage clamp conditions. Because of kinetic disparity between Fluo-4 and EGTA, a fraction of released Ca 2ϩ ions will first bind to the indicator (of fast kinetics) before being captured by the nonfluorescent EGTA at excess. As a result, this experimental setting allows for pinpointing the spatial localization and tracking the time course of Ca 2ϩ release fluxes (9). While holding the membrane potential at Ϫ70 mV, rapid local application of 5 mM caffeine, which sensitizes CICR, elicited transient Ca 2ϩ releases from discrete sites. These functional release sites were, on average, separated 1.7 m apart, similar to those between adjacent RyR3 spots (Fig. 1A). Thus, the Ca 2ϩ spike measurement strongly suggests that RyR3 spots visualized in Fig. 1A represent CRUs in DRG neurons.

Organization and Operation of Subsurface CRUs-
Measurement of Store Ca 2ϩ Capacity-Local Ca 2ϩ release functions at five representative sites are shown to the bottom of Fig. 1C. During the onset of the caffeine-induced Ca 2ϩ release, discrete Ca 2ϩ spikes, which reflect solitary Ca 2ϩ sparks, were clearly discernible. However, at increasing rate of occurrence, they were quickly fused into an apparently continuous release pattern. Nearly all DRG CRUs underwent repetitive activation, indicating a large store Ca 2ϩ reserve. To quantify the capacity of store Ca 2ϩ reserve, we divided the total signal mass of the release function (space-time integral of ⌬F/F 0 ) with the average signal mass for individual spikes (8.3 Ϯ 1.1 ⌬F/F 0 ⅐ms⅐m, n ϭ 68) (see supplemental Fig. s1A). We found that, on average, each CRU discharges about 15 Ca 2ϩ spikes or sparks in a caffeine-elicited Ca 2ϩ release (Fig. s1B). Since excessive EGTA should have largely "clamped" the cytosolic free Ca 2ϩ level and thereby minimized the replenishment of the ER during an ongoing release (9,23), this result suggests that the caffeine-liable Ca 2ϩ store does not undergo substantial depletion in individual sparks.
Action Potential Evoked Non-inactivating Ca 2ϩ Spark Activity-To examine the ability of neural action potentials, typically of millisecond duration, to evoke Ca 2ϩ sparks, cells were loaded with Fluo-4 via incubation with the acetoxymethyl ester form of the indicator and were then subjected to perforated patch clamp-ing in conjunction with confocal curve scan imaging. In response to brief action potentials (APD 50 ϭ 4.4 Ϯ 0.1 ms, n ϭ 10), subsurface Ca 2ϩ rose sharply to 1.16 Ϯ 0.13 (⌬F/F 0 unit) within 21.0 Ϯ 2.5 ms and then declined with a half-decay time of 165 Ϯ 31 ms. The rise time of Ca 2ϩ transients outlasted APD 50 but was similar to those of Ca 2ϩ sparks, 2 consistent with a synchronous activation of sparks that evolve autonomously. Moreover, trains of Ca 2ϩ sparks were discernible in the wake of the action potential-induced Ca 2ϩ transient, lingering for more than 2000 ms. This is in sharp contrast to the use-dependent inactivation and refractoriness in heart muscle cells (18,23). We conclude that subsurface Ca 2ϩ sparks serve to decode membrane excitation into cytosolic Ca 2ϩ signal and retain a "memory" of membrane activity well beyond the duration of an action potential.
Recruitment of Medullar CRUs-The immunostaining in Fig. 1A suggests that medullar CRUs stand free of contact with the surface Ca 2ϩ channels, forming a common pool system. To determine whether medullar CRUs participate in excitaion-Ca 2ϩ release coupling in DRG neurons, and, if so, to investigate the mechanism responsible for the recruitment of medullar CRUs, we examined subsurface and medullar Ca 2ϩ transients in response to trigger I Ca under whole-cell voltage clamp conditions. At near-threshold voltages (Ϫ55 mV), Ca 2ϩ release was exclusively restricted to discrete subsurface sites, each displaying repetitive spark activity ( Fig. 2A, right). When medullar Ca 2ϩ was elevated due to Ca 2ϩ diffusion from the subsurface, medullar Ca 2ϩ sparks became evident at intermediate voltages (Ϫ45 mV, Fig. 2B). During full-fledged excitation-Ca 2ϩ release coupling at Ϫ10 mV, coordinated radial CICR took the form of inwardly propagating, unmitigated Ca 2ϩ waves, rendering the crescent wavefront seen in the linescan image (Fig. 2C). Similar inward propagation of Ca 2ϩ signals upon excitation has been seen previously in sympathetic ganglion cells (29). After ryanodine pretreatment (  Ϫ151 Ϯ 23 pA/pF in CICR-deficient cells, n ϭ 11, p Ͼ 0.05). The amplitude ratio of subsurface and medullar (5 m into the cytosol) Ca 2ϩ transients was also decreased from 0.96 Ϯ 0.05 to 0.76 Ϯ 0.03 (n ϭ 11) in the absence of CICR. Hence, while subsurface CRUs directly amplify the trigger Ca 2ϩ entry, medullar CRUs relay subsurface Ca 2ϩ signals and ensure a speedy and uniform broadcast of the signal over the entire cytoplasm.
Gain of CICR in DRG Neurons-We reckoned that, by examining subsurface Ca 2ϩ transient arising from a known I Ca , the so-determined relationship between the Ca 2ϩ transient amplitude and Q ICa could then be used to calibrate the release flux. In cells where CICR was abrogated by ryanodine pretreatment (10 M, 20 min), I Ca and the ensuing subsurface Ca 2ϩ transient were directly measured with whole-cell voltage clamping and confocal curve-scan imaging, respectively. Two interesting results came from this experiment. First, DRG neurons displayed a very high I Ca density, with the peak of Ϫ156 Ϯ 24 pA/pF at Ϫ10 mV (n ϭ 11) (Fig. 3B). This value is at least an order of magnitude higher than those in heart muscle cells (ϳϪ10 pA/pF at Ϫ10 mV) (6, 8 -10). Moreover, the subsurface Ca 2ϩ transient amplitude (⌬F/F 0 ) was linearly correlated with Q ICa over wide ranges of ⌬F/F 0 (0.21-3.44) and Q ICa (0.07-0.97 ϫ 10 Ϫ9 Coulomb) (r 2 ϭ 0.85, Fig. 3C).
In cells with the ER function intact, we measured the I Ca and the subsurface Ca 2ϩ transient arising from both Ca 2ϩ entry and release. As summarized in Fig. 3, A and B, subsurface Ca 2ϩ transient amplitude exhibited a bell-shaped voltage dependence, mirroring the voltage-I Ca relationship. This finding corroborates the notion that CICR is smoothly graded to the trigger Ca 2ϩ entry. To quantify the release component, we first calculated the total Ca 2ϩ flux, Q total , based on the calibration curve in Fig. 3C, and then the release flux by subtracting from it the I Ca component, i.e. Q total Ϫ Q ICa . From this, the gain of CICR was then defined as the ratio between the release and I Ca components.
Surprisingly, the average value of the gain of CICR in DRG cells was merely 0.54 (n ϭ 35 cells). This value is not only far below those in cardiac cells (10 -70 depending on voltage) (6, 8 -10) but also considerably smaller than unity, a criterion for intrinsic stability of CICR (16). Furthermore, the gain function was virtually constant over test voltages ranging from Ϫ30 mV to ϩ50 mV, independently of either whole-cell I Ca density or singlechannel current amplitude (i Ca ). This is also in stark contrast to cardiac CICR, which gain function monotonically declines from ϳ70 to about ϳ10 between Ϫ40 and ϩ40 mV (8 -10). DISCUSSION Uncontrolled CICR, when operating in the regime of amplification gain Ͼ1.0, is expected to be intrinsically unstable, unless it is under tight local control (16). Even with local control, a stable, high gain CICR, such as that in heart muscle cells, still necessitates counter mechanisms that negate the positive feedback to terminate the release (30). To this end, the present study has uncovered several important features on the control of CICR in a mammalian sensory neuron, which are summarized in Table I along with cardiac CICR as the reference system. First, neural CICR comprises a hybrid architecture; the subsurface CRUs are under tight local control by Ca 2ϩ influx, whereas the medullar CICR is more appropriately categorized as a common-pool system (in which all RyRs sense a more-or-less uniform level of global Ca 2ϩ ). In addition to regulating membrane-delimited Ca 2ϩ -dependent events (e.g. somatic secretion), 2 subsurface CICR also amplifies the trigger Ca 2ϩ for the recruitment of medullar CRUs. The medullar CICR, on the other hand, appears to be indispensable for a speedy and unmitigated relay of Ca 2ϩ signal over the entire cell (Fig. 2). Second, virtually all CRUs manifest repetitive spark activity, indicative of lacking inactivation and refractoriness (Figs. 1, C and D, and 2, A and B). Although unexpected, this is consistent with the report that RyR3 from DRGs displays no Ca 2ϩ -depndent inactivation in planar lipid bilayer even at 10 mM Ca 2ϩ on the cytosolic side (27). The large store Ca 2ϩ reserve capacity ( Fig. 1) further excludes store depletion as a major determinant for the termination of CICR. At the same time, we found that DRG Ca 2ϩ transient amplitude is smoothly graded by the trigger I Ca (Fig. 3, A and B), indicating the stability of CICR.
The central issue is then what mechanism confers the stability to CICR in the absence of counteracting termination mechanisms? Quantitative measurement of the gain function indicates that CICR in DRG neurons is a linear, low gain amplification system, whose gain value is always smaller than unity (Fig. 3D). According to the theory put forward by Stern (16), such a low gain CICR system is endowed with intrinsic stability, i.e. CICR is self-limiting in the absence of additional termination mechanisms other than spontaneous closure of the release channel. The intrinsic stability of CICR is also consistent with our preliminary observation that CICR in DRG neurons appears to be unconditionally stable; it is unable to support spontaneous propagating Ca 2ϩ waves even when the cells were challenged with high extracellular Ca 2ϩ (10 mM, data not shown). It is noteworthy that, despite the low gain amplification, neural Ca 2ϩ transient magnitude is not compromised, because DRG cells are equipped with a compensating high I Ca density (Table I). Taken together, the present study provides the first demonstration that CICR operating in a low gain regime can be harnessed to provide a robust and graded amplification of Ca 2ϩ signal in the absence of counteracting termination mechanism. Another interesting finding is that the CICR gain function in DRG neurons is voltage-independent (Fig. 3D), rendering neural CICR a linear amplification mechanism. By contrast, the gain of cardiac CICR is a monotonic decreasing function of voltage (8 -10). The cardiac result was initially taken as a signature feature in favor of the local control model of CICR (8), because the gain is not uniquely determined by whole-cell I Ca but also by microscopic properties of single-channel i Ca . Later it has been shown to reflect the fact that RyR activation is a nonlinear, power function of i Ca with the power of 2 (31). Intuitively, the linear amplification in DRG neurons may further ensure the stability of neural CICR.
The present results indicate that CRUs retain a memory of membrane excitation, in the form of hyperactive sparks, well beyond the duration of an action potential. One possible explanation is that Ca 2ϩ influx via voltage-gated Ca 2ϩ channels results in overfilling of subsurface Ca 2ϩ stores above normal, which, in turn, promotes elevated spontaneous spark activity (memory). However, direct measurement of store Ca 2ϩ capacity demonstrated that this is not the case, because we failed to detect any significant change in store Ca 2ϩ capacity prior to and 1 s after the action potential (see supplemental Fig. s2). Alternatively, the hyperactive sparks might be activated due to the elevation of background cytosolic Ca 2ϩ . This hypothesis is not tenable, either, because hyperactive sparks are restricted to only a number of CRUs, while all CRUs are amid the elevated cytosolic Ca 2ϩ . Moreover, hyperactive sparking sites are observed during small depolarization (Fig. 2, A and B), when little elevation of cytosolic or ER Ca 2ϩ is expected. Hence we propose that the memory-encoding hyperactive Ca 2ϩ sparks reflect an intrinsic property of CRU operation. For instance, a CRU at an excited state may undergo many open-close transitions before returning to its basal state.
Using DRG sensory neurons as a model system, we have uncovered mechanisms fundamental to the control of CICR, which are distinctly different from what we have thus far learned from muscle cells. Among others, CICR in DRG neurons is characterized by the hybrid local control and common pool architecture, the apparent lack of counteracting termination mechanisms, and the low gain, linear amplification combined with the high density Ca 2ϩ entry. In this way, the neural CICR mechanism is capable of generating large intracellular Ca 2ϩ transients without jeopardizing stability. The striking differences as well as similarities between neural and cardiac CICR (Table I) are instructive as to how the same signaling mechanism can be adaptive and plastic in different biological systems. Since similar hybrid architecture is common to many types of cells, including sympathetic ganglion neurons (29), neurons of the central nervous system (4), and smooth muscle myocytes (32), the insights gleaned from DRG neurons may prove to be valuable in understanding Ca 2ϩ regulation and signaling in diverse physiological contexts.