Uptake and Release of Ca2+ by the Endoplasmic Reticulum Contribute to the Oscillations of the Cytosolic Ca2+ Concentration Triggered by Ca2+ Influx in the Electrically Excitable Pancreatic B-cell*

The role of intracellular Ca2+ pools in oscillations of the cytosolic Ca2+ concentration ([Ca2+] c ) triggered by Ca2+ influx was investigated in mouse pancreatic B-cells. [Ca2+] c oscillations occurring spontaneously during glucose stimulation or repetitively induced by pulses of high K+ (in the presence of diazoxide) were characterized by a descending phase in two components. A rapid decrease in [Ca2+] c coincided with closure of voltage-dependent Ca2+ channels and was followed by a slower phase independent of Ca2+ influx. Blocking the SERCA pump with thapsigargin or cyclopiazonic acid accelerated the rising phase of [Ca2+] c oscillations and increased their amplitude, which suggests that the endoplasmic reticulum (ER) rapidly takes up Ca2+. It also suppressed the slow [Ca2+] c recovery phase, which indicates that this phase corresponds to the slow release of Ca2+ that was taken up by the ER during the upstroke of the [Ca2+] c transient. Glucose promoted the buffering capacity of the ER and amplified the slow [Ca2+] c recovery phase. The slow phase induced by high K+ pulses was not affected by modulators of Ca2+- or inositol 1,4,5-trisphosphate-induced Ca2+ release, did not involve a depolarization-induced Ca2+ release, and was also observed at the end of a rapid rise in [Ca2+] c triggered from caged Ca2+. It is attributed to passive leakage of Ca2+ from the ER. We suggest that the ER displays oscillations of the Ca2+ concentration ([Ca2+]ER) concomitant and parallel to [Ca2+] c . The observation that thapsigargin depolarizes the membrane of B-cells supports the proposal that the degree of Ca2+ filling of the ER modulates the membrane potential. Therefore, [Ca2+]ER oscillations occurring during glucose stimulation are likely to influence the bursting behavior of B-cells and eventually [Ca2+] c oscillations.

release of Ca 2ϩ by the endoplasmic reticulum (ER). Thapsigargin-sensitive Ca 2ϩ -ATPases (SERCA pumps) are responsible for the sequestration process during the upstroke of the [Ca 2ϩ ] c transient, whereas the subsequent phase of release does not involve depolarization-, Ca 2ϩ -or IP 3 -mediated processes and likely results from leakage from the ER. This suggests that the Ca 2ϩ concentration within the endoplasmic reticulum ([Ca 2ϩ ] ER ) oscillates. As the filling state in Ca 2ϩ of the ER may modulate the membrane potential of B-cells (16), it is possible that [Ca 2ϩ ] ER oscillations play a role in the control of the oscillations of the membrane potential.

Solutions and Drugs
Except for patch-clamp measurements and the experiments illustrated in Fig. 4D (see below), the medium used was a bicarbonatebuffered solution that contained 120 mM NaCl, 4.8 mM KCl, 0.5-10 mM CaCl 2 , 1.2 mM MgCl 2 , 24 mM NaHCO 3 and 0 -20 mM glucose as indicated. When the concentration of KCl was increased, that of NaCl was decreased accordingly to keep the osmolarity of the medium unchanged. Ca 2ϩ -free solutions were prepared by substituting MgCl 2 for CaCl 2 and were supplemented with 0.5 or 2 mM EGTA as indicated in the legends to Figs. 2, 3, and 5.
In the experiments illustrated in Fig. 4D, it was important to minimize changes in the activity of the Na ϩ /Ca 2ϩ exchange between solutions containing various K ϩ concentrations. Therefore, KCl was not replaced with NaCl but with choline chloride to keep a similar Na ϩ concentration in all solutions. The low K ϩ solution contained: 79.8 mM NaCl, 4.8 mM KCl, 40.2 mM choline chloride, 2.5 mM CaCl 2 , 1.2 mM MgCl 2 , 24 mM NaHCO 3 , and 0.01 mM atropine, which prevented activation of muscarinic receptors by choline. The solutions containing higher K ϩ concentrations were prepared by substituting KCl for choline chloride.
All solutions were gassed with O 2 /CO 2 (94:6) to maintain a pH of 7.4 at 37°C. Except for electrophysiological recordings, they were supplemented with 1 mg/ml bovine serum albumin (fraction V; Roche Molecular Biochemicals).
Thapsigargin was obtained from Sigma or from Alomone Laboratories (Jerusalem, Israel). Ryanodine was from RBI (Natick, MA) or from Alomone Laboratories, diazoxide was from Schering-Plough Avondale (Rathdrum, Ireland), caffeine was from Merck A.G. (Darmstadt Germany), and ruthenium red was from Alexis Corp. (San Diego, CA). All other chemicals were from Sigma.

Preparation of Islets and Cells
All experiments were performed with tissue from fed female NMRI mice (25-30 g). Pancreatic islets were isolated aseptically after collagenase digestion of the pancreas, and when needed, they were dispersed into cells as described previously (17). Cells were allowed to attach to 22-mm circular coverslips and cultured for 2-3 days. Intact islets were maintained in culture for 1-3 days. When the membrane potential of B-cells was to be measured with an intracellular microelectrode, the islets were allowed to attach to the coverslip by a culture period of at least 2 days. The culture medium was RMPI 1640 medium containing 10 mM glucose, 10% heat-inactivated fetal calf serum, 100 IU/ml penicillin, and 100 g/ml streptomycin.

Measurements of [Ca 2ϩ ] i
Cultured islets were loaded with 2 M fura-PE3/AM (Teflabs, Austin, TX) for 90 -120 min at 37°C in a bicarbonate-buffered solution containing 10 mM glucose. Cultured cells were loaded with 1 M fura-2/AM (Molecular Probes, Eugene, OR) for 60 min in a similar bicarbonatebuffered medium. The tissue was then transferred into a temperaturecontrolled (37°C) perifusion chamber of ϳ1 ml (Intracell, Royston, Herts, United Kingdom) with a bottom made of a glass coverslip and mounted on the stage of an inverted microscope. The flow rate of the perifusion was approximately 2 ml/min. When rapid exchange of solutions was required, a ϳ250-l chamber was used and solutions were changed by Iso-Latch valves (Parker Hannifin, Fairfield, NY). [Ca 2ϩ ] i was directly measured in cells attached to the coverslip or in islets held in place close to the coverslip by gentle suction with a glass micropipette. In some experiments, cultured cells were pressure-injected with an 5242 Eppendorf microinjector (Hamburg, Germany). The injected solution contained either 6 -10 mM fura-2 K ϩ salt or 10 mM fura-dextran K ϩ salt (molecular weight, 3000) (Molecular Probes) dissolved in H 2 O, and it was supplemented or not with test substances. The techniques used to monitor [Ca 2ϩ ] c have been described previously (4).

Flash-Photolysis
Clusters of B-cells were incubated with 5 M nitrophenyl-EGTA AM and 1.5 M fura-2 AM (Molecular Probes) for 60 min at 37°C. Photolysis of nitrophenyl-EGTA was performed by two or three consecutive 1-ms UV flashes of 240 J (Xenon flashlamp system XF-10, Hi-Tech, Hamburg, Germany).

Electrophysiology
Membrane Potential Recordings-The islets were mounted in a perifusion chamber (7 ml/min at 37°C) following attachment to glass coverslips. The membrane potential of a single cell within the islet was continuously measured with a high resistance microelectrode.
Patch-Clamp Recordings-Voltage-clamp experiments were performed on single B-cells using the perforated patch-whole cell configuration and an EPC-7 patch-clamp amplifier (List Elektronik, Darmstadt, Germany). The holding potential was Ϫ70 mV, and the cells were submitted either to 100-ms depolarizations to 0 mV or to bursts of 100-ms depolarizations (2 Hz) from Ϫ50 mV to Ϫ10 mV for 12 s. The associated changes in [Ca 2ϩ ] i were measured using an IonOptix fluorescence imaging system (IonOptix, Inc., Milton, MA). The extracellular solution contained 138 mM NaCl, 5.6 mM KCl, 1.2 mM MgCl 2 , 2.6 mM CaCl 2 , 5 mM HEPES (pH 7.4 with NaOH), and 10 mM glucose. The pipette solution contained 76 mM Cs 2 SO 4 , 10 mM NaCl, 10 mM KCl, 1 mM MgCl 2 , and 5 mM HEPES (pH 7.35 with CsOH). Electrical contact with the cell interior was established by adding 0.24 mg/ml amphotericin B to the pipette solution, and the voltage-clamp was considered satisfactory when the series conductance (G series ) was Ͼ35-40 nano Siemens. All experiments were performed at 33°C, and the zero-current potential of the pipette was adjusted with the pipette in the bath solution.

Presentation of Results
The experiments are illustrated by recordings that are averaged or representative traces of results obtained with the indicated number of cells or islets from at least three different cultures. The statistical significance of differences between means was assessed by unpaired Student's t test.

RESULTS
[Ca 2ϩ ] c Oscillations Induced by Glucose Are Followed by Ca 2ϩ Release from the ER-B-cells within intact islets display a rhythmic electrical activity when perifused with a medium containing an insulin-releasing glucose concentration (10 mM) and 10 mM Ca 2ϩ (Fig. 1A). These bursts of electrical activity consist of sharp depolarizing waves of the membrane potential with superimposed spikes reflecting Ca 2ϩ influx through voltagedependent Ca 2ϩ channels (18). Under these conditions, [Ca 2ϩ ] c also oscillates, but, in contrast to the fast, monophasic repolarization of the oscillations of membrane potential, the descending phase of each Ca 2ϩ oscillation clearly displays two components (Fig. 1B). Whereas the initial fast one appears to coincide with the closure of voltage-dependent Ca 2ϩ channels following rapid repolarization of the plasma membrane, the second, much slower phase appears to occur during the repolarized intervals.
Previous experiments have shown that intracellular Ca 2ϩ stores of whole islets are efficiently emptied by thapsigargin (TG), a specific inhibitor of the SERCA pump (19), but that this emptying requires preincubation of the islets with the drug (17). In islets pretreated with 1 M TG, the amplitude of [Ca 2ϩ ] c oscillations was much larger than in control islets, and the descending phase of each [Ca 2ϩ ] c oscillation was surprisingly very fast with no slow second phase (Fig. 1C). This suggests that the slow phase observed in control islets results from a release of Ca 2ϩ from the ER, rather than from a slow Ca 2ϩ extrusion from the cytosol.
The effect of intracellular Ca 2ϩ store depletion on [Ca 2ϩ ] c oscillations was also investigated in clusters of islet cells, a preparation in which the SERCA pump can be blocked by an acute addition of TG or cyclopiazonic acid (CPA). CPA is an inhibitor structurally unrelated to TG (19) and has also been shown to empty the ER of Ca 2ϩ in pancreatic B-cells (20). In the presence of 15 mM glucose and 2.5 mM Ca 2ϩ , [Ca 2ϩ ] c oscillated slowly and regularly (Fig. 1D). Addition of 50 M CPA to the medium accelerated the oscillations, which increased in amplitude and frequency and became sharper mainly because of the disappearance of the slow recovery phase. Similar results were obtained in clusters of islet cells treated by TG (not shown).
Ca 2ϩ Release from the ER Can be Detected after Pulses of High K ϩ -In this series of experiments, glucose-induced [Ca 2ϩ ] c oscillations were inhibited by diazoxide, which, by opening K ATP channels, clamps the membrane potential at a hyperpolarized level. [Ca 2ϩ ] c oscillations were then reinduced by rhythmically depolarizing the plasma membrane with high K ϩ .
In control islets, high K ϩ pulses (for 30 s) induced [Ca 2ϩ ] c oscillations characterized by a descending phase that displayed an initial fast component concomitant with rapid repolarization of the plasma membrane, followed by a slow decline ( Fig. 2A, solid line). In TG-pretreated islets, [Ca 2ϩ ] c oscillations were higher than in control islets (467 Ϯ 21 versus 352 Ϯ 11 nM, n ϭ 10, p Ͻ 0.01) and devoid of a slow recovery phase. Similar results were obtained after pretreatment of the islets with 50 M CPA.
The effects of TG on voltage-dependent Ca 2ϩ current were evaluated in single B-cells using the perforated patch configuration. Under control conditions, a 100-ms voltage-step from Ϫ70 to 0 mV elicited a peak Ca 2ϩ current of 52 Ϯ 6 pA (n ϭ 8) that was not significantly affected by a 5-min exposure to 1 M TG (48 Ϯ 3 pA). The integrated whole-cell Ca 2ϩ current was similarly unaffected by TG (data not shown). This excludes the possibility that the larger rise in [Ca 2ϩ ] c induced by high K ϩ in TG-treated islets results from an increased Ca 2ϩ current.
We also verified that the slow [Ca 2ϩ ] c recovery is not a peculiarity observed only in a medium containing 10 mM Ca 2ϩ . To this end, 30-s pulses of 45 mM K ϩ were applied in the presence of various concentrations of external Ca 2ϩ (0.5-10 mM) (Fig. 2B). The amplitude of the resulting [Ca 2ϩ ] c peaks clearly depended on the Ca 2ϩ concentration of the medium, but a slow decaying phase, prevented by TG pretreatment, was observed at all external Ca 2ϩ concentrations tested. The observation that TG suppresses the slow recovery after [Ca 2ϩ ] c os- cillations of various amplitude also excludes the possibility that TG might increase the rate of Ca 2ϩ extrusion from the cytosol due to a high Ca 2ϩ signal.
The slow [Ca 2ϩ ] c recovery phase could be artifactual and reflect changes in the Ca 2ϩ concentration within the ER if fura-PE3 is compartmentalized. To exclude this possibility, single B-cells were microinjected with fura-dextran, a Ca 2ϩ probe that is exclusively localized in the cytosol (21). High K ϩ pulses induced [Ca 2ϩ ] c oscillations with a slow recovery phase (Fig. 2C). Addition of TG to the medium induced a transient increase in [Ca 2ϩ ] c reflecting intracellular Ca 2ϩ pool emptying. Subsequent depolarization by pulses of high K ϩ triggered [Ca 2ϩ ] c oscillations that were of much larger amplitude than before addition of the SERCA pump inhibitor and that lacked a slow [Ca 2ϩ ] c recovery phase. These observations strongly support the conclusion that the slow decaying [Ca 2ϩ ] c phase results from release of Ca 2ϩ from the ER.
Characteristics and Kinetics of Ca 2ϩ Exchanges between the ER and the Cytosol-Application of high K ϩ pulses every 5 min triggered a train of [Ca 2ϩ ] c oscillations with a slow decaying phase, which indicates that the phenomenon is not a transient one (Fig. 3A). The experiments depicted in Fig. 3B were designed to explore the temporal requirements for refilling the intracellular Ca 2ϩ stores responsible for the slow decay in [Ca 2ϩ ] c . The islets were repetitively depolarized by 30-s pulses of high K ϩ . Extracellular Ca 2ϩ (10 mM) was present before and during the depolarization (first and last pulses) or only during the depolarization (second to seventh pulses). The slow recovery phase was present and not attenuated by Ca 2ϩ omission during the repolarization phases (compare Fig. 3, A and B). This shows first that it does not result from Ca 2ϩ influx, and second that Ca 2ϩ entry during depolarization is sufficient to refill the pools from which Ca 2ϩ is slowly released.
However, no slow recovery phase was observed when high K ϩ pulses were applied in the continuous presence of acetylcholine (ACh), a potent IP 3 -producing agent in pancreatic Bcells (Fig. 3C). This is likely due to the fact that Ca 2ϩ cannot accumulate into the ER because it immediately exits from the ER into the cytosol through IP 3 receptors that are maintained opened by the continuous presence of ACh.
The ability of the ER to take up Ca 2ϩ rapidly was next tested (Fig. 3D). Islets perifused with a Ca 2ϩ -free medium were submitted to three pulses of 100 M ACh applied at 12.5-min intervals. A 30-s pulse of high K ϩ /high Ca 2ϩ was applied between the second and the third pulses of ACh. Whereas the first application of ACh triggered a large [Ca 2ϩ ] c peak, the second one induced only a small rise in [Ca 2ϩ ] c suggesting that intracellular Ca 2ϩ stores were nearly completely emptied already by the first application of ACh. However, the third application of ACh in a Ca 2ϩ -free medium after the short pulse with high K ϩ /high Ca 2ϩ induced a transient rise in [Ca 2ϩ ] c that was much larger than that seen after the second application of ACh. This indicates further that intracellular Ca 2ϩ pools rapidly refill during the large [Ca 2ϩ ] c rises triggered by high K ϩ pulses.
If the slow recovery phase reflects release of Ca 2ϩ from the ER, its characteristics should depend on the filling state of the ER. This was tested by emptying the ER with ACh between two series of 3 pulses of high K ϩ /high Ca 2ϩ of 20 s duration (Fig.  3E). The first three [Ca 2ϩ ] c oscillations were all characterized by a slow recovery phase. In contrast, the first two oscillations following intracellular Ca 2ϩ pool depletion by ACh were of lower amplitude and displayed a much smaller slow recovery phase than before ACh application. Because the pulses were of constant duration, the lower amplitude of [Ca 2ϩ ] c oscillations post-ACh is unlikely to result from a decreased Ca 2ϩ influx. It may rather be explained by a more avid sequestration of Ca 2ϩ into an emptied than into a filled ER. Because the first high K ϩ /high Ca 2 pulse did not carry enough Ca 2ϩ to fully refill the ER, no slow recovery phase could be seen, and three pulses were needed to refill the ER enough to see a slow recovery phase of an amplitude similar to that observed at the end of the first series of [Ca 2ϩ ] c oscillations. These data demonstrate that the buffering capacity of the ER permits a rapid control of [Ca 2ϩ ] c and that its ability to release Ca 2ϩ is affected by its filling state.
Comparison of the [Ca 2ϩ ] c changes induced by a pulse of high K ϩ in control and TG-treated islets permits estimation of the kinetics of Ca 2ϩ uptake and release from the ER (Fig. 4A)  islets (Fig. 4B). The downward deflection of the curve reflects Ca 2ϩ uptake by the ER, whereas the upward deflection reflects release from the ER. This shows that the uptake is very fast, whereas the release is comparably slow and lasts several minutes.  (Fig. 4D), suggesting that the Ca 2ϩ loading of the ER is directly proportional to the level of [Ca 2ϩ ] c . This did not result from a K ϩ effect, as the amplitude of the rise in [Ca 2ϩ ] c was similar in clusters perifused with a Ca 2ϩ -free medium containing 4.8 or 45 mM K ϩ . The large rise in [Ca 2ϩ ] c produced by TG in the presence of high K ϩ and Ca 2ϩ is in agreement with the large slow [Ca 2ϩ ] c decay observed after depolarizing pulses with high K ϩ .
The effect of glucose was also tested. 30-s pulses of high K ϩ induced a larger [Ca 2ϩ ] c rise in the absence of glucose than in the presence of 20 mM glucose (Fig. 4E). By contrast, the slow [Ca 2ϩ ] c recovery phase was more pronounced in a glucosecontaining than in a glucose-free medium. It was prevented by TG pretreatment (not shown). To estimate the amplitude and the kinetics of Ca 2ϩ release from the ER in glucose-containing and glucose-free medium, the averaged [Ca 2ϩ ] c oscillation of TG-treated islets was subtracted from the averaged [Ca 2ϩ ] c oscillation of control islets in the presence and in the absence of glucose (Fig. 4F). This revealed a much larger [Ca 2ϩ ] c release phase in the presence of 20 mM glucose than in its absence.
Mechanisms of the Slow Ca 2ϩ Release Process from the ER-In skeletal muscle cells, depolarization of the plasma membrane alone can trigger release of Ca 2ϩ from intracellular stores (22). However, this process does not seem to be operative in pancreatic B-cells, as no [Ca 2ϩ ] c increase could be detected when the islets were depolarized by pulses of high K ϩ in a Ca 2ϩ -free medium supplemented with 2 mM EGTA (Fig. 5A). This lack of effect of high K ϩ did not result from exhaustion of intracellular Ca 2ϩ pools by EGTA present in the medium because ACh could still trigger Ca 2ϩ mobilization after 2 inefficient pulses of high K ϩ .
Depolarization is not sufficient and even not necessary, as shown by the following experiment. Pancreatic B-cells were loaded with the caged Ca 2ϩ compound, nitrophenyl-EGTA, whereas [Ca 2ϩ ] c was kept at basal levels by the presence of 4.8 mM K ϩ and 250 M diazoxide in the medium. Flashes of UV light induced a large rise in [Ca 2ϩ ] c , which then decreased in two phases, an initial fast one followed by a slow recovery to basal levels (Fig. 5B). Brief depolarization with 45 mM K ϩ was followed by a similar biphasic response. After addition of TG, a second series of UV flashes triggered a new increase in [Ca 2ϩ ] c followed by a rapid decrease that now lacked the slow recovery phase. These experiments clearly demonstrate that the rise in [Ca 2ϩ ] c is sufficient to induce a slow recovery phase, even in the absence of membrane depolarization.
Two well characterized mechanisms can trigger Ca 2ϩ release from intracellular Ca 2ϩ stores: Ca 2ϩ -induced and IP 3 -induced Ca 2ϩ release (2, 23). High concentrations (5-20 mM) of caffeine are known to induce or to potentiate Ca 2ϩ -induced Ca 2ϩ release and to inhibit IP 3 -induced Ca 2ϩ -release (24). However, 10 mM caffeine was without effect on the slow recovery phase observed at the end of a 30-s pulse of 45 mM K ϩ (data not shown). Ryanodine is a potent modulator of Ca 2ϩ -induced Ca 2ϩ release. It activates this process at low concentrations (Յ1 M) but blocks it at high concentrations (Ն10 M) (24,25). Here, the islets were treated acutely, preincubated, or cultured with different concentrations of ryanodine (1-100 M). In no case did we observe an effect of ryanodine on basal [Ca 2ϩ ] c , on peak [Ca 2ϩ ] c -induced by high K ϩ pulses, or on the subsequent slow recovery phase (data not shown). Microinjection of B-cells with 10 mM ryanodine was also ineffective (n ϭ 3, data not shown).
In many tissues, including pancreatic B-cells, phospholipase C can be activated directly by a rise in [Ca 2ϩ ] c or by depolarization of the plasma membrane (8,26). It is therefore possible that a [Ca 2ϩ ] c rise increases IP 3 levels, which in turn trigger Ca 2ϩ release. To test this hypothesis, pancreatic B-cells were microinjected with heparin, a blocker of the IP 3 receptor in various tissues, including pancreatic B-cells (27). In B-cells microinjected with fura-2 free acid alone, a pulse of high K ϩ induced a large [Ca 2ϩ ] c oscillation characterized by a slow recovery phase (Fig. 5C). Subsequent addition of 100 M ACh triggered a large and transient increase in [Ca 2ϩ ] c . Emptying the ER with TG induced a further rise in [Ca 2ϩ ] c and prevented the slow recovery phase of the [Ca 2ϩ ] c oscillation induced by a subsequent pulse of high K ϩ . Heparin microinjection (molecular weight 6000; 200 mg/ml) completely prevented the AChinduced [Ca 2ϩ ] c rise without affecting the response to TG (Fig.  5C). As heparin did not affect the slow recovery phase of the [Ca 2ϩ ] c oscillation induced by high K ϩ , it is clear that this phase is not induced by an IP 3 -mediated Ca 2ϩ release.
A delayed, slow return of [Ca 2ϩ ] c to basal level after a depolarization-induced [Ca 2ϩ ] c rise has been observed in neurons and chromaffin cells (28 -31) and attributed to a slow release of Ca 2ϩ from mitochondria. No similar process seems to be operative in B-cells. Thus, microinjection of B-cells with 1 mM ruthenium red (giving a final cytosolic concentration Ͼ1 M) did not affect the slow recovery [Ca 2ϩ ] c phase after a pulse of high K ϩ (n ϭ 3, data not shown), although the drug is regarded as a potent and selective inhibitor of the mitochondrial Ca 2ϩ uniporter at ϳ 1 M (32).
These experiments indicate that the release of Ca 2ϩ observed after a rise in [Ca 2ϩ ] c originates from the ER; that it is not triggered by depolarization-, IP 3 -, or Ca 2ϩ -induced Ca 2ϩ release; and that none of these three mechanisms contribute to the [Ca 2ϩ ] c rise induced by high K ϩ .
Effect of Intracellular Ca 2ϩ Pool Depletion under More Physiological Conditions-Ideally, release of Ca 2ϩ at the end of [Ca 2ϩ ] c oscillations induced by glucose in a physiological medium containing 2.5 mM Ca 2ϩ should now be sought for. Unfortunately, this was not possible because emptying of intracellular Ca 2ϩ pools by TG transformed oscillations of the membrane potential of whole islets induced by 10 mM glucose into a sustained depolarization with continuous spike activity (Fig. 6A).
We therefore tested the effect of intracellular Ca 2ϩ pool depletion in voltage-clamped single B-cells subjected to trains of 100-ms depolarizations (2 Hz for 12 s) from Ϫ50 to Ϫ10 mV (holding potential, Ϫ70 mV) designed to mimic glucose-induced bursts of action potentials (Fig. 6B). This induced a large rise in [Ca 2ϩ ] c that was followed by a slow recovery to basal levels upon repolarization to Ϫ70 mV. Once [Ca 2ϩ ] c had returned to basal levels, TG was applied for 5 min, and the cell was again subjected to a burst of depolarizations. This raised [Ca 2ϩ ] c to a higher level than before TG addition (776 Ϯ 64 versus 577 Ϯ 53 nM, respectively; p Ͻ 0.05; n ϭ 8). Importantly, the time constant of the falling phase was much shorter (3.1 Ϯ 2.1 versus 13 Ϯ 2.4 s, respectively; p Ͻ 0.01). This suggests that intracellular Ca 2ϩ stores play a role in the oscillations in [Ca 2ϩ ] c during bursts of action potentials induced by glucose.

DISCUSSION
The present study demonstrates that rapid uptake and release of Ca 2ϩ by the ER contributes to [Ca 2ϩ ] c oscillations induced by Ca 2ϩ influx through voltage-dependent Ca 2ϩ channels in pancreatic B-cells.

Nature of the Intracellular Ca 2ϩ Store Taking up and Releasing Ca 2ϩ in Response to a Rise in [Ca 2ϩ ] c Triggered by Ca 2ϩ
Influx-A fast and strong [Ca 2ϩ ] c buffering has been documented in patch-clamped B-cells (33). Two observations in the present study ascribe this property to the ER. First, a 30-s pulse of high K ϩ could replenish nearly completely emptied ACh-sensitive stores. Second, the rise in [Ca 2ϩ ] c induced by depolarization was faster and larger in TG and CPA-treated islets than in controls, although the Ca 2ϩ current was not increased. This is in agreement with the recent report that 20 mM K ϩ raises [Ca 2ϩ ] ER in INS-1 rat insulinoma cells expressing aequorin in the ER (34). The ER is also the source of Ca 2ϩ that is released into the cytoplasm after a rise in [Ca 2ϩ ] c , because inhibition of the SERCA pump by TG or CPA or opening of IP 3 receptors by ACh completely abolished the slow [Ca 2ϩ ] c recovery phase. Furthermore, the filling state of the ER profoundly affected the characteristics of the slow recovery phase.
Uptake of Ca 2ϩ by mitochondria during a rapid rise in [Ca 2ϩ ] c has been documented in various cell types, including neurons, chromaffin cells, and insulin-secreting cells (28 -31, 35, 36). However, we found that the slow decay of [Ca 2ϩ ] c following an abrupt rise was unaffected by ruthenium red, a blocker of the mitochondrial Ca 2ϩ uniporter (32). These observations clearly establish that mitochondria are not primarily responsible for the slow [Ca 2ϩ ] c decay.
Regulation of Ca 2ϩ Uptake-Uptake of Ca 2ϩ by the ER directly depends on [Ca 2ϩ ] c . Indeed, the amount of Ca 2ϩ that was released from the ER by TG was proportional to the steadystate level of [Ca 2ϩ ] c in clusters of cells depolarized with various concentrations of K ϩ . Such a Ca 2ϩ dependence of the uptake has been clearly demonstrated in various cells (14,37,38). Pancreatic B-cells express SERCA-2B and SERCA-3 isoforms (39), SERCA-2B being the most sensitive to Ca 2ϩ among all SERCA isoforms (40).
Uptake of Ca 2ϩ by the ER is also modulated by the glucose concentration of the medium. A smaller influx of Ca 2ϩ through voltage-dependent Ca 2ϩ channels cannot explain why high K ϩinduced [Ca 2ϩ ] c oscillations were smaller in the presence of 20 mM glucose than 0 mM glucose because glucose enhances voltagedependent Ca 2ϩ currents in B-cells (41). The difference rather results from an increased buffering capacity of the ER in the presence of glucose. This is in agreement with the observation that stimulation of B-cells with glucose causes an initial drop in [Ca 2ϩ ] c that is blocked by TG (42). Other studies have shown that the amount of Ca 2ϩ taken up by the ER of permeabilized RINm5F insulinoma cells depends on the ATP/ADP ratio (37). The longer [Ca 2ϩ ] c recovery phase observed in the presence of glucose therefore reflects a larger release of Ca 2ϩ from the ER into which glucose has promoted Ca 2ϩ sequestration during influx of the ion.
Mechanism of Ca 2ϩ Release-Three mechanisms of Ca 2ϩ release from intracellular Ca 2ϩ stores have been described: depolarization-, Ca 2ϩ -, and IP 3 -induced Ca 2ϩ release (2,22,23). We did not detect any depolarization-induced Ca 2ϩ release from filled Ca 2ϩ stores in mouse pancreatic B-cells, but we showed that a rise in [Ca 2ϩ ] c is sufficient to induce slow release of the ion from the ER when [Ca 2ϩ ] c decreases. Ca 2ϩ -and/or IP 3 -induced Ca 2ϩ release has been suggested to play a role in glucose-induced [Ca 2ϩ ] c changes (6 -10). Pancreatic B-cells express very low levels of type 2 ryanodine receptors (9) responsible for Ca 2ϩ -induced Ca 2ϩ release (2)  receptors. We used heparin, caffeine, and ryanodine, three established modulators of Ca 2ϩ -or IP 3 -induced Ca 2ϩ release (24,25,27), to investigate the possible contribution of these processes to [Ca 2ϩ ] c oscillations induced by high K ϩ pulses. None of these compounds affected the oscillations. Moreover, if a depolarization-, Ca 2ϩ -, or IP 3 -induced Ca 2ϩ release participated in high K ϩ -induced [Ca 2ϩ ] c rise in pancreatic B-cells, the latter would be reduced by depletion of intracellular pools with TG. The results show exactly the opposite. Taken together, our experiments suggest that the release of Ca 2ϩ observed after a rise in [Ca 2ϩ ] c is not triggered by depolarization-, IP 3 -or Ca 2ϩinduced Ca 2ϩ release and that none of these three mechanisms contribute to the [Ca 2ϩ ] c rise induced by high K ϩ . Ca 2ϩ release from the ER at the end of [Ca 2ϩ ] c oscillations more likely corresponds to a slow release of Ca 2ϩ from the organelle, which slowly adapts its Ca 2ϩ concentration to [Ca 2ϩ ] c . Release of Ca 2ϩ through the same pathway may explain the rise in [Ca 2ϩ ] c that occurs upon blockade of the SERCA pump with TG. Because of its high Ca 2ϩ permeability, this pathway has often been referred to as leak from the ER. Although it has been observed in many cell types, its exact nature has not been determined (14,43 (38,43). The Ca 2ϩ -ATPase of the ER is proportionally more stimulated by cytosolic Ca 2ϩ than the PMCA (38), but it is inhibited by luminal Ca 2ϩ (14). Ca 2ϩ leakage through the ER seems only mildly stimulated by luminal Ca 2ϩ (14). On the other hand, there is an interplay between the rate at which Ca 2ϩ -ATPases work and the available energy. It is indeed possible that changes in the pumping rate of Ca 2ϩ -ATPases during [Ca 2ϩ ] c oscillations modulate the ATP/ADP ratio, which in turn modulates the activity of Ca 2ϩ -ATPases (44). We have recently shown that the ATP/ADP ratio drops rapidly when [Ca 2ϩ ] c is raised and increases when [Ca 2ϩ ] c falls (45). Periodic release and reuptake of Ca 2ϩ from the ER of permeabilized RINm5F cells supplemented with an oscillating glycolytic cell-free muscle extract have been reported (46).
Basal [Ca 2ϩ ] c is set by a balance between processes that increase [Ca 2ϩ ] c (leak entry of Ca 2ϩ from the extracellular space and leak release of Ca 2ϩ from the ER) and processes that decrease [Ca 2ϩ ] c (extrusion mechanisms that remove Ca 2ϩ from the cytosol) (Fig. 7B). When [Ca 2ϩ ] c increases abruptly as a result of Ca 2ϩ influx through voltage-dependent Ca 2ϩ channels, Ca 2ϩ extrusion out of the cell by the PMCA is stimulated, and Ca 2ϩ uptake by the ER occurs at a higher rate than Ca 2ϩ release (Fig. 7C). As [Ca 2ϩ ] c rises, the ATP/ADP ratio decreases (45), perhaps because of ATP consumption by these ATPases. When repolarization of the plasma membrane closes voltagedependent Ca 2ϩ channels, Ca 2ϩ influx abruptly stops. This, together with rapid uptake by the ER and extrusion out of the cell, produces a fast drop in [Ca 2ϩ ] c that corresponds to the first phase of [Ca 2ϩ ] c decrease (Fig. 7D). As [Ca 2ϩ ] c decreases, the PMCA is less stimulated. In the ER, Ca 2ϩ release predominates over Ca 2ϩ uptake because the latter is inhibited by high luminal [Ca 2ϩ ] (14) and by the lowering of the ATP/ADP ratio. This corresponds to the beginning of the slow phase of [Ca 2ϩ ] c decrease that lasts until release and uptake reach a new equilibrium. As the PMCA extrudes Ca 2ϩ less efficiently than the SERCA pumps it into the ER (38), Ca 2ϩ could even cycle between the cytosol and the ER, which would prolong the slow [Ca 2ϩ ] c decrease.
Previous studies have clearly demonstrated that glucoseinduced [Ca 2ϩ ] c oscillations of single pancreatic B-cells or clusters of islet cells do not require the participation of the ER (17,47,48). Our present data do not contradict these observations, but they strongly support the hypothesis that [Ca 2ϩ ] ER oscillations occur synchronously with and in parallel to glucose-induced [Ca 2ϩ ] c oscillations. The parallel nature of these oscillations strikingly contrasts with the antiparallel changes of IP 3induced [Ca 2ϩ ] c and [Ca 2ϩ ] ER oscillations (13,49,50).
Physiological Implications-In many cell types, emptying of intracellular Ca 2ϩ pools activates various types of currents with distinct ion selectivity (Ca 2ϩ , K ϩ , or Na ϩ ) through a family of channels that have been termed store-operated channels (51)(52)(53)(54). It has recently been demonstrated that the magnitude of the current activated by intracellular pool emptying correlates with the extent of store depletion and that it can be activated even by small decreases in [Ca 2ϩ ] ER (55). In pancreatic B-cells, intracellular Ca 2ϩ pool depletion activates a small Ca 2ϩ entry and induces a depolarizing current, possibly carried by Na ϩ , that potentiates the activation of voltage-dependent Ca 2ϩ channels (17,56). This was evidenced in the present study by the observation that in the presence of 2.5 mM Ca 2ϩ in the medium, TG transformed oscillations of the membrane potential into a sustained depolarization. It is therefore likely that oscillations of the Ca 2ϩ concentration within the ER rhythmically activate a depolarizing current.
The model that we suggest is presented in Fig. 7. The membrane potential of B-cells is controlled by several conductances, among which the K ATP current (I KATP ) and the store-operated current (I SOC ) play major roles. In the presence of a nonstimulating concentration of glucose, I SOC is too small to counteract the overwhelming repolarizing current through K ATP channels. At stimulating glucose concentrations, I KATP is much reduced and can be counteracted by I SOC (Fig. 7B). When [Ca 2ϩ ] c increases, the ER fills with Ca 2ϩ , leading to decrease of I SOC and repolarization of the plasma membrane (Fig. 7C). Ca 2ϩ influx through voltage-dependent Ca 2ϩ channels then stops, [Ca 2ϩ ] c decreases, and the ER starts to release more Ca 2ϩ than it takes up (Fig. 7D). The subsequent slow emptying of intracellular Ca 2ϩ pools reactivates I SOC and the plasma membrane depolarizes again (Fig. 7B). This increases [Ca 2ϩ ] c , and a new cycle starts again. The model depicted in Fig. 7 thus suggests that glucose controls the membrane potential of B-cells partly indirectly, by modulating the buffering capacity of the ER. It is fully compatible with our recent model (45) indicating that I KATP might also oscillate during [Ca 2ϩ ] c oscillations. Before the onset of [Ca 2ϩ ] c oscillations, I KATP would be small, when I SOC is large. At the end of [Ca 2ϩ ] c oscillations, it would be large, when I SOC is small. Some mathematical models of the bursting electrical activity of the pancreatic B-cell (57) also include a mechanism by which oscillations of the Ca 2ϩ concentration within the ER concomitant with [Ca 2ϩ ] c oscillations control slow waves of the membrane potential. In contrast, our model differs from that proposed by Dukes et al. (58), which involves a depolarization-or IP 3