Slow Feedback Inhibition of Calcium Release-activated Calcium Current by Calcium Entry*

In many nonexcitable cells, depletion of the inositol 1,4,5-trisphosphate-sensitive store activates Ca2+ influx, a process termed store-operated Ca2+ entry. In rat basophilic leukemia cells, emptying of the stores activates a highly selective Ca2+release-activated Ca2+ current (CRAC),I CRAC. We have recently found thatI CRAC activates in an essentially all-or-none manner when the current is evoked by receptor stimulation, dialysis with inositol 1,4,5-trisphosphate via the patch pipette, or through the Ca2+ATPase inhibitor thapsigargin (Parekh, A. B., Fleig, A., and Penner, R. (1997) Cell 89, 973–980). Regulatory mechanisms must therefore operate to control the overall amount of Ca2+ that enters through CRAC channels. Such mechanisms include membrane potential and protein kinase C. In the present study, we have investigated additional inhibitory pathways that serve to determine just how much Ca2+ can enter throughI CRAC. We have directly measured the current using the whole cell patch clamp technique. We report the presence of a slow Ca2+-dependent inactivation mechanism that curtails Ca2+ entry through CRAC channels. This inactivation mechanism is switched on by Ca2+ entering through CRAC channels, and therefore constitutes a slow negative feedback process. Although it requires a rise in intracellular Ca2+ for activation, it maintains CRAC channels inactive even under conditions that lower intracellular Ca2+ levels. The inactivation mechanism does not involve store refilling, protein phosphorylation, G proteins, nor Ca2+-dependent enzymes. It accounts for up to 70% of the total inactivation ofI CRAC, and therefore appears to be a dominant inhibitory mechanism. It is likely to be an important factor that shapes the profile of the Ca2+ signal in these nonexcitable cells.

In many nonexcitable cells, depletion of the inositol 1,4,5-trisphosphate-sensitive store activates Ca 2؉ influx, a process termed store-operated Ca 2؉ entry. In rat basophilic leukemia cells, emptying of the stores activates a highly selective Ca 2؉ release-activated Ca 2؉ current (CRAC), I CRAC . We have recently found that I CRAC activates in an essentially all-or-none manner when the current is evoked by receptor stimulation, dialysis with inositol 1,4,5-trisphosphate via the patch pipette, or through the Ca 2؉ ATPase inhibitor thapsigargin (Parekh, A. B., Fleig, A., and Penner, R. (1997) Cell 89, 973-980). Regulatory mechanisms must therefore operate to control the overall amount of Ca 2؉ that enters through CRAC channels. Such mechanisms include membrane potential and protein kinase C. In the present study, we have investigated additional inhibitory pathways that serve to determine just how much Ca 2؉ can enter through I CRAC . We have directly measured the current using the whole cell patch clamp technique. We report the presence of a slow Ca 2؉ -dependent inactivation mechanism that curtails Ca 2؉ entry through CRAC channels. This inactivation mechanism is switched on by Ca 2؉ entering through CRAC channels, and therefore constitutes a slow negative feedback process. Although it requires a rise in intracellular Ca 2؉ for activation, it maintains CRAC channels inactive even under conditions that lower intracellular Ca 2؉ levels. The inactivation mechanism does not involve store refilling, protein phosphorylation, G proteins, nor Ca 2؉ -dependent enzymes. It accounts for up to 70% of the total inactivation of I CRAC , and therefore appears to be a dominant inhibitory mechanism. It is likely to be an important factor that shapes the profile of the Ca 2؉ signal in these nonexcitable cells.
In many nonexcitable cells, depletion of the inositol 1,4,5trisphosphate (InsP 3 ) 1 -sensitive intracellular Ca 2ϩ stores activates a Ca 2ϩ influx pathway in the plasma membrane (1). This mechanism was originally proposed by Putney (2) and called capacitative Ca 2ϩ influx. Patch-clamp experiments have iden-tified a variety of Ca 2ϩ -permeable channels in the plasma membrane that seem to underlie capacitative Ca 2ϩ influx (reviewed in Ref. 3). These channels differ in their biophysical properties and are generically referred to as store-operated Ca 2ϩ channels (3,4).
Of the store-operated Ca 2ϩ currents, the best characterized is I CRAC , which was originally discovered in mast cells (5). I CRAC has subsequently been shown to exist in several different nonexcitable cells including basophils, T cells and megakaryocytes (3). CRAC channels are remarkably selective for Ca 2ϩ ions and have a low single-channel conductance (3,5).
Just how depletion of the stores activates CRAC channels is still unclear. Several potential mechanisms have been proposed but the signal has not been unequivocably identified (reviewed in Refs. 3 and 4). One interesting aspect of I CRAC is that the current activates in an essentially all-or-none manner, irrespective of whether activation is evoked by dialysis with inositol 1,4,5-trisphosphate, receptor stimulation, or thapsigargin (6). One consequence of this is that, if I CRAC activates in an all-or-none manner, mechanisms must exist that control the amount of Ca 2ϩ entering the cell through CRAC channels. This is required in order to achieve graded activation of Ca 2ϩ -dependent processes like secretion that correlate with the level of cell stimulation by receptors. We have recently reported that Ca 2ϩ entry through CRAC channels in rat basophilic leukemia (RBL) cells can be graded, despite all-or-none activation, because of several regulatory mechanisms that serve to control CRAC channel activity (6,7). One way is by changing the membrane potential. Hyperpolarization increases the electrical gradient for Ca 2ϩ entry, thus favoring further Ca 2ϩ influx, whereas depolarization decreases the driving force and hence reduces Ca 2ϩ entry (6).
I CRAC is also regulated by protein kinase C. Stimulation of this enzyme inactivates I CRAC . Since protein kinase C will be activated by diacylglycerol, which is produced following stimulation of receptors that engage the phosphoinositide pathway, it constitutes an important negative feedback mechanism on Ca 2ϩ influx (7).
In mast cells and jurkat T lymphocytes, I CRAC is subjected to a fast inactivation process operating on a milliseconds time scale. This arises from Ca 2ϩ ions entering the cell through CRAC channels and then binding to sites probably located on the channels themselves (5,8).
Here we report an additional mechanism that serves to regulate CRAC channels. We find that Ca 2ϩ influx through CRAC channels exerts a slow feedback inhibition that curtails further Ca 2ϩ entry and which is dependent on a rise in intracellular Ca 2ϩ levels. Slow inactivation accounts for up to 70% of the inhibition of I CRAC . Once activated, this slow inactivation mechanism can maintain CRAC channels in an inactivated state for several minutes, even after intracellular Ca 2ϩ levels have been reduced. Hence slow inactivation appears to be a dominant inhibitory mechanism that determines the time course of Ca 2ϩ influx following store depletion in RBL cells. It also provides an additional mechanism whereby the amount of Ca 2ϩ entering through CRAC channels can be regulated despite all-or-none Ca 2ϩ entry.

EXPERIMENTAL PROCEDURES
Rat basophilic leukemia cells (RBL-2H3) cells were kindly supplied by Michael Pilot, Max Planck Institute for Biophysical Chemistry, Goettingen, Germany, and were cultured essentially as described previously (6,7). Patch-clamp experiments were conducted in the tight seal whole cell configuration at room temperature (18 -25°C) as described previously (6,7). Patch pipettes were pulled from borosilicate glass (Hilgenberg), sylgard-coated, and fire-polished. Pipettes had d.c. resistances of 2.5-4 megohms when filled with standard internal solution that contained (in mM): cesium glutamate 145, NaCl 8, MgCl 2 1, MgATP 1, InsP 3 0.03, EGTA 1.4, HEPES 10, pH 7.2, with CsOH. A correction of ϩ10 mV was applied for the subsequent liquid junction potential. In some experiments, the EGTA concentration was raised to 14 mM (indicated in text) or substituted with BAPTA. Drugs were added to this internal solution as described in the text. Extracellular solution contained (in mM): NaCl 145, KCl 2.8, CaCl 2 10, MgCl 2 2, CsCl 10, glucose 10, HEPES 10, pH 7.2 (NaOH). CsCl was present to block the activity of the inwardly rectifying potassium channel. High resolution current recordings were acquired by a computer-based patch-clamp amplifier system (EPC-9, HEKA Electronics, Germany). Capacitative currents were canceled before each voltage ramp using the automatic compensation of the EPC-9. Series resistance was between 5 and 15 megohms. Currents were filtered using an 8-pole Bessel filter at 2.5 kHz and digitized at 100 s. I CRAC was measured using either voltage ramps (Ϫ100 to ϩ100 mV in 50 ms) or voltage steps (pulses to Ϫ80 mV for 200 ms) applied every 2 s using PULSE software (HEKA Electronics) on a 9500 PowerMac. Cells were held at 0 mV between pulses. All currents were leak subtracted by averaging the first two to four ramps/steps after breaking in and then subtracting this from all subsequent traces. Several parameters (capacitance, series resistance, holding current) were displayed simultaneously on a second monitor at a slower rate (2 Hz) using the X-Chart display (HEKA Electronics). Data are presented as mean Ϯ S.E., and statistical evaluation was carried out using Student's unpaired t test.

Slow Inactivation of I CRAC in the Presence of Moderate intracellular calcium buffering-Previous
work has demonstrated that I CRAC inactivates partially when RBL cells are dialyzed with a patch pipette solution containing a high concentration of the slow Ca 2ϩ chelator EGTA (10 mM), and this is due to a kinase-mediated phosphorylation (7). In the present study, we have examined the effects of more moderate calcium buffering on the properties of I CRAC by including 1.4 mM EGTA in the recording pipette. Fig. 1A(i) shows an experiment in which the internal solution contained 30 M InsP 3 (a supramaximal concentration) (6) and 1.4 mM EGTA. Cells were voltage clamped at 0 mV and I CRAC was monitored using voltage ramps applied every two seconds (shown in Fig. 1A(ii)). The current was measured at Ϫ80 mV. Fig. 1A(i) depicts the time course of the current during the experiment. Following the onset of whole cell recording, I CRAC activated as InsP 3 diffused into the cell from the recording pipette and depleted the stores. The time constant for activation () was 15.2 Ϯ 1.6 s, similar to our previous measurements (19.4 Ϯ 1.2 s) (9). The current peaked after 50 -80 s and then inactivated substantially with time. For the cell shown in Fig. 1A(i), the current had fallen by 70% (relative to the peak amplitude) after 300 s. Fig. 1A(ii) shows A, dialysis of an RBL cell with InsP 3 and 1.4 mM EGTA generates I CRAC that inactivates over a time course of tens of seconds (i). The current/voltage relationship, measured using voltage ramps applied every two seconds, revealed that the current was I CRAC : voltage-independent, inwardly rectifying with a reversal potential Ͼϩ60 mV (ii). The data points in (i) were measured from the ramps at Ϫ80 mV, and have been normalized to cell size by dividing the current by membrane capacitance. B, time course of I CRAC in a different cell (i) in which the current was measured using hyperpolarizing steps to Ϫ80 mV, applied every 2 s (ii). The current amplitude was measured at the arrow, when it had reached a steady-state value. The raw currents shown in the right-hand panels correspond to the labeled data points on the left (a and b, respectively). The rate and extent of slow inactivation was similar using either voltage protocol, as shown in C. Number of cells was six each for the step and ramp experiments respectively. D, slow inactivation is reduced in the presence of a higher EGTA concentration. Number of cells was 10 for 14 mM and 8 for 1.4 mM EGTA respectively. All experiments were carried out as paired recordings as described in the text. The difference in current amplitude between 1.4 and 14 mM EGTA at longer times (Ͼ150 s) was significant (p Ͻ 0.01). The inset shows current traces obtained in either 1.4 or 14 mM EGTA. These were taken at 200 s, and the current in 1.4 mM EGTA has been scaled so that it has the same peak as that in 14 mM. The horizontal (time) axis is 40 ms and the vertical axis, Ϫ0.2 pA/pF. Fig. 1A(i) (a and b). The current measured in the ramps exhibited the classic features of I CRAC : voltage-independent activation, inward rectification, and a reversal potential close to ϩ60 mV.

I-V relationships for the two time points indicated in
We also monitored activity of I CRAC by applying voltage steps to Ϫ80 mV for 200 ms every 2 s from a holding potential of 0 mV. In Fig. 1B(i), the time course of I CRAC is plotted using this protocol, and Fig. 1B(ii) shows the inward current measured during these hyperpolarizing pulses. During the step, I CRAC inactivated by up to 40% with a time constant in the range of 50 ms ( Fig. 1B(ii)). This fast inactivation has been described in detail in T cells and arises from local feedback inactivation by Ca 2ϩ ions in the vicinity of the channel pore (8). We measured the steady-state current during the step (indicated by the arrow in 1B(ii)), and this amplitude is plotted against time in Fig.  1 B(i). As with the ramp protocol employed in Fig. 1A, I CRAC inactivated substantially during the experiment. For the cell shown in Fig. 1B, the current amplitude at 300 s had fallen by 68% relative to the peak value. Fig. 1C plots the time course of I CRAC measured using either the ramp or step methods (six cells each). These experiments were carried out such that ramps or steps were applied to alternate cells on the same coverslips (henceforth referred to as paired experiments). There was no significant difference between ramp or step protocols with respect to activation (15.2 Ϯ 1.6 versus 17.7 Ϯ 2.0 s, respectively) peak amplitude (Ϫ2.98 Ϯ 0.24 versus Ϫ2.58 Ϯ 0.19 pA/pF) or in the extent of steady-state inactivation (fallen by 53.4 Ϯ 8.9% versus 56.1 Ϯ 9.3%, respectively, Fig. 1C).
This inactivation in the presence of moderate Ca 2ϩ buffering will henceforth be referred to as slow inactivation to distinguish it from the fast inactivation seen during the hyperpolarizing step.
Variability in the Overall Extent of Slow Inactivation between Cell Preparations-In some cells, slow inactivation was very prominent and contributed to an almost complete inactivation of I CRAC . Typical examples of this type of response are shown in Figs. 3B and 5A. In some other cells, slow inactivation contributed to around a 50% decrease in the amplitude of I CRAC (e.g. Fig. 5C). Overall, there was little difference between prep-arations, with the average decrease being 64.8 Ϯ 7.9% (n ϭ 35, measured using the step protocol after 300 s).
In experiments where drugs were used to interfere with slow inactivation, control responses were always obtained from the same preparation and the data were compared with these controls.
Slow Inactivation Is Reduced by a Higher EGTA Concentration in the Recording Pipette-To confirm that slow inactivation is dependent on a rise in intracellular Ca 2ϩ levels, we dialyzed cells with a pipette solution in which the EGTA concentration had been increased 10-fold (to 14 mM). Pooled data from 10 cells in 14 mM EGTA and eight cells in 1.4 mM EGTA are summarized in Fig. 1D (paired recordings). Slow inactivation was reduced by almost 2-fold in the presence of the higher EGTA concentration and appears therefore to require a rise in intracellular free Ca 2ϩ because it can be suppressed by increasing the concentration of a slow mobile Ca 2ϩ buffer in the cytosol. The inset of Fig. 1D shows current traces taken from the voltage steps from two cells dialysed with 1.4 and 14 mM EGTA. Fast inactivation was the same in both cells, yet slow inactivation was less pronounced in the presence of higher EGTA. Fast and slow inactivation therefore reflect distinct processes.
Slow Inactivation Is Dependent on Ca 2ϩ Entry into the Cell-To assess the contribution of Ca 2ϩ influx through CRAC channels to the slow Ca 2ϩ -dependent inactivation, we carried out a series of experiments in which the electrochemical gradient for Ca 2ϩ influx was altered. First, we modified the electrical gradient by applying voltage pulses to potentials that either reduced Ca 2ϩ entry (Ϫ40 mV) or enhanced it (Ϫ120 mV), relative to the responses obtained on pulsing to Ϫ80 mV. Typical results are shown in Fig. 2A, and the data are summarized in Table I. Stepping the voltage to Ϫ40 mV reduced the rate and extent of inactivation almost 2-fold ( Fig. 2A(i) and Table I).
The current at Ϫ40 mV in Fig. 2A(i) has been scaled so that it has the same amplitude as that at Ϫ80 mV in order to clearly show the slower inactivation. Voltage pulses to Ϫ120 mV, however, did not increase the level of steady-state inactivation compared with that seen on stepping to Ϫ80 mV ( Fig. 2A(ii), where the current at Ϫ80 mV has been scaled, see also Table I),

FIG. 2. Slow inactivation of I CRAC is dependent on Ca 2؉ influx.
A, effects of changing the electrical gradient for Ca 2ϩ entry on slow inactivation. In (i), reducing the driving force from Ϫ80 to Ϫ40 mV reduces slow inactivation, whereas increasing the driving force to Ϫ120 mV (ii) has no effect. The current recording at Ϫ40 mV in (i) has been scaled to match the response at Ϫ80 mV in peak amplitude, as has been the recording at Ϫ80 mV in (ii) so that it matches the response at Ϫ120 mV. The absolute values for the currents in (i) are Ϫ1.6 pA/pF (Ϫ80 mV) and Ϫ1.1 pA/pF (Ϫ40 mV). For (ii) they are Ϫ1.4 pA/pF (Ϫ80 mV) and Ϫ3.2 pA/pF (Ϫ120 mV). All cells in Fig. 2A were from the same coverslip. B, reducing the concentration gradient for Ca 2ϩ influx from 10 mM to 4 mM at Ϫ80 mV results in less inactivation. C, the fast Ca 2ϩ chelator BAPTA (1.4 mM) does not prevent slow inactivation from occurring (six cells). See Table I  which might suggest that sufficient Ca 2ϩ enters at Ϫ80 mV to maximally activate the inhibitory process.
In a second set of experiments, we reduced the concentration gradient for Ca 2ϩ influx, while leaving the electrical driving force unchanged. This was accomplished by lowering the external Ca 2ϩ concentration in the bath to 4 mM (from 10 mM). A typical example is shown in Fig. 2B, in which the cell was hyperpolarized to Ϫ80 mV at 2-s intervals. Steady-state inactivation was less pronounced (22% for this cell, see also Table  I). The fact that slow inactivation was reduced following a decrease only in the concentration gradient for Ca 2ϩ entry while the electrical gradient was unchanged suggests that it is predominantly a Ca 2ϩ -dependent rather than a voltage-dependent phenomenon.
Ca 2ϩ -dependent Slow Inactivation Is Not Affected by a Fast Ca 2ϩ Chelator-The preceding results demonstrate that I CRAC is subject to a negative feedback mechanism in which Ca 2ϩ influx through CRAC channels induces a slow Ca 2ϩ -dependent inactivation. Fast negative feedback inactivation of CRAC channels operates on a milliseconds time-scale and can be substantially reduced by inclusion of the fast Ca 2ϩ chelator BAPTA in the recording pipette, whereas it is not altered by increasing the concentration of the slower Ca 2ϩ chelator EGTA (8). Replacing EGTA with the fast chelator BAPTA (both at 1.4 mM) did not significantly affect the rate or extent of inactivation (six cells, Fig. 2C and Table I). However, rapid inactivation of I CRAC was slightly slowed by this concentration of BAPTA (1.2-1.4-fold, data not shown), reinforcing the notion that the inactivation we observe is distinct from the fast inactivation process.
Once Initiated, Slow Inactivation Is Not Strongly Dependent on Ca 2ϩ Entry-Although Ca 2ϩ entry is important for initiating slow inactivation, we set out to determine whether slow inactivation still required Ca 2ϩ entry even after the inactivation process had started to develop. Fig. 3 describes two types of experiment which were designed to address this. In Fig. 3A, hyperpolarizing steps to Ϫ80 mV were repetitively applied every 2 s from a holding potential of 0 mV. Once inactivation had clearly developed (at 200 s), the cell was held continuously at ϩ20 mV for 30 s. At this positive potential, very little Ca 2ϩ enters through CRAC channels (6). After 30 s, the cell was held again at 0 mV and hyperpolarizing steps were resumed. The amplitude of I CRAC increased only slightly relative to the level it had reached prior to clamping the cell at ϩ20 mV (four of nine cells). In the other five cells, no recovery was observed at all. Fig. 3B shows an experiment in which I CRAC had inactivated fully by 300 s. Hyperpolarizing pulses were stopped and the cell was held at 0 mV for 120 s. Hyperpolarizing pulses were then resumed. No recovery of I CRAC occurred (three of four cells). In one cell, the amplitude of I CRAC recovered by 9%. Taken together, these results suggest that, once initiated, Ca 2ϩ -depend-ent slow inactivation of I CRAC is sustained despite using maneuvers that reduce Ca 2ϩ influx.
Investigation into the Mechanism of Ca 2ϩ -dependent Slow Inactivation-The following set of experiments were aimed at elucidating the molecular mechanism that gave rise to the slow inactivation of I CRAC .
Refilling of Intracellular Ca 2ϩ Stores-In the presence of moderate concentrations of EGTA, it is conceivable that Ca 2ϩ entry through CRAC channels might enable the intracellular Ca 2ϩ stores to refill, a process that would turn off I CRAC . If such a mechanism were responsible for the inactivation of I CRAC , then one would predict that maneuvers directed toward reducing Ca 2ϩ uptake into the stores should prevent slow inactivation from occurring. Thapsigargin is a specific inhibitor of the Ca 2ϩ ATPase on the endoplasmic reticulum and prevents store refilling (10). We therefore carried out paired recordings in which six control cells were dialyzed with internal solution containing InsP 3 and 1.4 mM EGTA, whereas six other cells from the same preparations were dialyzed with this solution supplemented with 2 M thapsigargin. Fig. 4A shows the averaged results. Slow inactivation was only slightly reduced by thapsigargin, but this was not statistically significant. Thapsigargin is lipophilic, so it is possible that the drug diffuses out of the cell into the bath solution. Because we included a high concentration of thapsigargin in the pipette (2 M), one would be surprised if the steady-state cytoplasmic concentration was less than a few hundred nanomolar, a concentration that is sufficient to reduce Ca 2ϩ ATPase activity in a variety of cell types (10). In two cells, we applied thapsigargin (1 M) from the outside just after the onset of I CRAC (thapsigargin was also included in the recording pipette). The current still inactivated (by 53 and 61%). Nevertheless, we sought additional ways to probe the effects of compromised store refilling on slow inactivation. One method would be to employ the Ca 2ϩ ionophore ionomycin. Ionomycin increases the permeability of the store membrane to Ca 2ϩ , and this would enable any Ca 2ϩ that had been pumped into the stores to diffuse back into the cytosol thereby preventing stores from refilling. We included 2 M ionomycin in the recording pipette together with InsP 3 and 1.4 mM EGTA. Fig. 4B shows the effects of ionomycin. I CRAC activated slightly faster in ionomycin-treated cells ( of 16.1 Ϯ 1.7 s versus 22.7 Ϯ 3.0 s in control paired cells), which might indicate that ionomycin is diffusing into the cells rather quickly and accelerating store depletion in combination with InsP 3 . Slow inactivation of I CRAC was still apparent in the presence of ionomycin, and the current declined to a value similar to that seen in the absence of ionomycin in paired recordings (six cells for ionomycin, five for controls).
Protein Kinase-mediated Phosphorylation-We tested the possible involvement of a kinase-mediated phosphorylation reaction in the Ca 2ϩ -dependent slow inactivation in RBL cells in two independent ways. First, we examined the effects of remov- In all experiments, 30 M InsP 3 was included in the recording pipette. 1.4 mM BAPTA/EGTA refers to 1.4 mM of total chelator. The voltage represents the potential to which the cells were stepped. All experiments were carried out in 10 mM external Ca 2ϩ , except the last row in which the bath solution contained 4 mM instead. The difference in steady-state current between 1.4 EGTA, Ϫ40 mV and 1.4 EGTA, Ϫ80 mV was significant (p Ͻ 0.02), whereas the difference between 1.4 EGTA, Ϫ80 mV and 1.4 EGTA, Ϫ120 mV was not (p Ͼ 0.4). The difference between 1.4 EGTA, Ϫ80 mV and 1.4 BAPTA, Ϫ80 mV was also not significant (p Ͼ 0.2). There were no significant differences between the activation values. ing ATP from the pipette solution and then replacing it with ATP␥S, and second, we tested the effects of broad kinase and phosphatase inhibitors on the inactivation process. Slow inactivation was still present when ATP was omitted from the pipette solution (three cells, not shown), suggesting that global ATP levels are probably not important for the inactivation process. We then replaced ATP with ATP␥S in the pipette solution. ATP␥S is a nonhydrolyzable analogue of ATP that is readily used by a variety of protein kinases resulting in "irreversible phosphorylation" of the target (11). Slow inactivation was slightly increased by ATP␥S (Table II). Importantly, it was not reduced, which one might have expected if recovery from inactivation required a phosphorylation (12).
Dialysis of cells with H-7 (200 M) did not affect slow inactivation compared with control cells in which the inhibitor was not present. A control response and one in which H-7 was present are shown in Fig. 5A. Both cells were from the same coverslip. Although H-7 seemed to slow the rate of inactivation in these paired recordings, in other cells control responses decayed at a slower rate than H-7 treated ones. The data from several cells are summarized in Table II. There was no significant difference in inactivation between H-7-treated and control cells. In three cells that were dialyzed with H-7, we stopped the application of hyperpolarizing steps after 300 s and waited FIG. 3. Once initiated, slow inactivation does not require continuous Ca 2؉ entry. A, the development of slow inactivation was monitored using hyperpolarizing steps to Ϫ80 mV applied every 2 s. Once inactivation had clearly developed, the cell was held at ϩ20 mV for 30 s to reduce further Ca 2ϩ influx into the cytoplasm. No hyperpolarizing steps were given during this time. After 30 s, the pulses were resumed. Little recovery of I CRAC occurred relative to the levels it had shown prior to reducing Ca 2ϩ influx. In B, slow inactivation completely turned off I CRAC . Hyperpolarizing steps were stopped and the cell was held at 0 mV for 2 min. Resumption of the steps did not result in any recovery of the current. Inclusion of the structurally distinct kinase inhibitor bisindolylmaleimide (2 M) did not prevent slow inactivation from occurring. In the presence of the drug however, inactivation was less pronounced compared with control cells (Table II). This reduction in inactivation is probably is due to the inhibition of protein kinase C by bisindolylmaleimide, since the former partially inactivates I CRAC (7).
The protein phosphatase inhibitor okadaic acid has been reported to inhibit slow inactivation of I CRAC in jurkat T cells (13). We therefore pretreated RBL cells for at least 20 min with 1 M okadaic acid and then included the same concentration of the inhibitor in the pipette solution. Slow inactivation still occurred, demonstrating that it was not sensitive to okadaic acid ( Fig. 5B and Table II).
GTP-dependent Proteins and Ca 2ϩ -activated Proteases-To assess the role of G proteins, we included 300 M GDP␤S in the recording pipette together with InsP 3 . Slow inactivation still occurred ( Fig. 5C and Table II) and it was not significantly different from control cells. Dialysis with 250 M GTP␥S did not affect slow inactivation either (inactivation of 77 Ϯ 9%). Later on in these same cells (after a 5-10-min recording), the GTP␥S-dependent Na ϩ current activated (14), demonstrating that the GTP␥S was active under these conditions.
To examine whether a Ca 2ϩ -dependent protease was responsible for slow inactivation, as is the case for voltage-gated calcium channels (15), we preincubated cells for 30 min with 2 M calpeptin, a potent Ca 2ϩ -dependent protease inhibitor. In some experiments, calpeptin was also added to the internal solution. Slow inactivation still occurred (Fig. 5D) and Table  II), and this was indistinguishable from control cells taken from different coverslips from the same preparations.
Because of the variability in the kinetics of slow inactivation, small effects by the agents used would not be detected. Although the findings therefore do not rule out a contribution from these signaling pathways, they demonstrate that these mechanisms are not dominant.

DISCUSSION
Negative Feedback by Ca 2ϩ Ions Entering through CRAC Channels Gives Rise to Slow Inactivation-Our observation that slow inactivation could be significantly reduced by increasing the concentration of the Ca 2ϩ chelator EGTA in the patch pipette solution (Fig. 1D) suggests that the inactivation mechanism requires a rise in cytoplasmic Ca 2ϩ concentration. This rise in Ca 2ϩ is accomplished by Ca 2ϩ entry through CRAC channels because maneuvers that reduce the electrochemical gradient for Ca 2ϩ influx result in less inactivation (Fig. 2).
It is likely that Ca 2ϩ entry is the main source of Ca 2ϩ for activating the slow inhibitory pathway with little contribution from Ca 2ϩ release from the stores (e.g. Ca 2ϩ entry evoked Ca 2ϩ release) for the following reasons. First, dialysis with InsP 3 and either ionomycin or thapsigargin, agents that would reduce the Ca 2ϩ content of the stores, did not alter slow inactivation. Second, cells were dialyzed with a supramaximal concentration of InsP 3 which is likely to maintain the stores in a Ca 2ϩdepleted state (6). Although the InsP 3 receptor might desensitize, thereby enabling stores to refill, we feel this process is not particularly active under our conditions. The InsP 3 receptor that is involved in the activation of I CRAC in RBL cells does not seem to desensitize much because Ca 2ϩ influx can still be activated more than 10 min after dialyzing with high InsP 3 levels (6). Finally, slow inactivation still occurs in the presence TABLE II Effects of various treatments on slow inactivation Steady-state current in control recordings (i.e. absence of drug) (% peak), ATP␥S 38.3 Ϯ 9.8 (4 cells); H-7, 32.0 Ϯ 11.2 (4 cells); bisindolylmaleimide, 32.5 Ϯ 4.7 (4 cells); okadaic acid and calpeptin (same preparations), 36.7 Ϯ 7.1 (6 cells); GDP␤S, 49.0 Ϯ 10 (3 cells). Only steady-state current in the presence of bisindolylmaleimide was significantly different to control (p Ͻ 0.05).   Table II. B, preincubation for 30 min with the broad phosphatase inhibitor okadaic acid (1 M) was ineffective on slow inactivation. Okadaic acid was also included in the recording pipette. Okadaic acid enhanced Ca 2ϩ influx in Xenopus oocytes after application of lysophosphatidic acid (n ϭ 2), as described in Petersen and Berridge (19). Hence okadaic acid was active. C, dialysis with the nonhydrolyzable GDP analogue GDP␤S (300 M) did not interfere with slow inactivation relative to control cells. GDP␤S prevented activation of I CRAC by the adenosine receptor agonist NECA in two cells, indicating it was active (7). D, pretreatment with the Ca 2activated protease inhibitor calpeptin (2 M), as well as its inclusion in the recording pipette, failed to alter slow inactivation. of 2 mM ATP␥S and no exogenous ATP (Table II). Ca 2ϩ pumps cannot use ATP␥S as a substrate and therefore will be much less effective in taking up Ca 2ϩ to replenish the stores.
Because an increase in the concentration of a slow Ca 2ϩ chelator like EGTA can reduce inactivation (Fig. 1D), and that there is little difference between BAPTA (a fast chelator) and EGTA (Fig. 2C), the Ca 2ϩ -dependent inactivation step seems to have slow kinetics. Furthermore, Ca 2ϩ buffering, rather than the speed of binding, is important in determining the rate and extent of the slow inactivation.
Mechanism of Slow Inactivation-Just how a rise in intracellular Ca 2ϩ switches on the slow inactivation process is not clear. It is unlikely to reflect refilling of the stores because inclusion of thapsigargin or ionomycin in the pipette solution did not prevent the inactivation from occurring. Direct involvement of the Ca 2ϩ -binding protein calmodulin in the inactivation process is also unlikely because inclusion of the specific calmodulin inhibitory peptide fragment from Calbiochem failed to prevent slow inactivation (three cells, not shown).
One important clue to the mechanism of slow inactivation was the finding that, once activated, it did not require the continuous presence of elevated Ca 2ϩ (Fig. 3). Such long-lasting effects are often mediated by protein phosphorylation reactions. However, several treatments designed to interfere with phosphorylation failed to affect slow inactivation of I CRAC . Involvement of Ca 2ϩ -activated proteases was also unlikely since a specific and potent inhibitor of this class of enzyme did not alter inactivation. It also seems unlikely that CRAC channels are being endocytosed slowly from the membrane in a Ca 2ϩdependent manner, because membrane capacitance did not change despite substantial inactivation of the Ca 2ϩ current (data not shown). Finally, during whole cell recording small soluble components diffuse out of the cell and this washout may cause a general loss of CRAC channel activity. The fact that increasing the concentration of EGTA in the recording pipette reduced slow inactivation suggests that loss of cytoplasmic factors might not explain the inactivation. However, we cannot rule out the possibility that a rise in intracellular Ca 2ϩ accelerates washout of an important component, perhaps by promoting its Ca 2ϩ -dependent dissociation from a bound to a diffusible form.
A slow Ca 2ϩ -dependent inactivation of store-operated Ca 2ϩ influx has been described in jurkat T cells (13) and NIH-3T3 fibroblasts (16). In T cells, it was reported that slow inactivation was inhibited by okadaic acid and by the kinase inhibitor H-7 (11,13). It was concluded that recovery from slow inactivation occurred through the actions of a protein kinase. In NIH-3T3 cells, Ca 2ϩ influx was measured using the fluorescent dye fura-2. Unlike T cells, slow inactivation was not affected by kinase or phosphatase inhibitors nor by Ca 2ϩ chelators (loaded via the membrane-permeable acetoxymethyl ester form). Slow inactivation was suggested to arise from an action of Ca 2ϩ not subject to Ca 2ϩ chelation, perhaps by binding to an external part of the channel. The slow inactivation pathway described in this report exhibits some notable differences from these other inactivation ones. Unlike T cells, it was not sensitive to H-7 or okadaic acid, nor to conditions designed to facilitate protein phosphorylation. Slow inactivation was reduced by increasing the concentration of Ca 2ϩ chelator in the recording pipette, demonstrating an intracellular action of Ca 2ϩ rather than an extracellular site as proposed for the NIH-3T3 cells (16). Ca 2ϩdependent slow inactivation of I CRAC might therefore arise through different mechanisms in different nonexcitable cells.
Comparison of Slow Inactivation with Other Inhibitory Mechanisms That Regulate I CRAC -Several inhibitory pathways have been found to regulate CRAC channels in RBL cells, and it might be instructive to ascertain their relative contributions to controlling Ca 2ϩ entry. I CRAC is subject to a fast inactivation process (8). The peak current declines by up to 40% following a hyperpolarization to Ϫ80 mV (Fig. 1B). Because we have measured I CRAC at steady state (i.e. once the rapid inactivation component was over, Fig. 1B), fast inactivation is not contributing to the inactivation that we have measured and is therefore not relevant to the discussion that follows.
In the presence of high EGTA in the recording pipette, I CRAC is also subject to a slow inactivation pathway that involves protein phosphorylation, probably through protein kinase C (7). Because this protein kinase-mediated inactivation is dependent on ATP in the recording pipette and is sensitive to protein kinase C blockers, it occurs through a mechanism distinct to that involved in the Ca 2ϩ -dependent slow inactivation described in this report. The kinase-mediated action accounts for up to 30% (Fig. 1D) and 45% of the inactivation (7).
Since depletion of stores activates I CRAC , one might expect that refilling of stores turns the current off. Using fluorescent dyes, it has been reported that the duration of Ca 2ϩ influx correlates with the Ca 2ϩ uptake into the stores, the latter being assessed through the extent of Ca 2ϩ release by an agonist in Ca 2ϩ -free solution (17). To date, only one study has suggested that store refilling turns off I CRAC . In T cells, Zweifach and Lewis (13) reported that I CRAC fully inactivated within 100 s in the presence of a moderate EGTA concentration (1.2 mM) and the inactivation could be partially reduced by thapsigargin. In our experiments, we have found no evidence that store refilling contributes to inactivation of I CRAC (Fig. 4), despite dialyzing cells with a similar concentration of EGTA to that used in T cells (1.2 versus 1.4 mM). The reason why store refilling contributes significantly to slow inactivation in T cells whereas it seems to play little, if any, role in RBL cells under similar conditions is unclear. Differences in current density are unlikely to explain this because increasing the RBL cell I CRAC (by stepping to Ϫ120 mV, Fig. 2B) did not enhance slow inactivation, despite the current now exhibiting a similar current density to that in the T cells. It is conceivable that local saturation of the Ca 2ϩ buffers might contribute to these differences in the role of store refilling, especially if stores were situated close to such areas of buffer depletion. It is interesting to note that InsP 3 -sensitive stores are physically adjacent to the plasma membrane in T cells (18). Ca 2ϩ buffers just below the plasma membrane are likely to be saturated following activation of I CRAC , thereby facilitating the refilling process. We have suggested previously that the InsP 3 -sensitive stores involved in activation of I CRAC in RBL cells are close to the plasma membrane (9), but no morphological data is available to indicate just how close they are. Finally, Ca 2ϩ ATPase activity might differ between jurkat T and RBL cells in whole cell recording, such that Ca 2ϩ uptake is less effective in the RBL cells. Clearly, further work is needed to assess the contribution of refilling to the deactivation of I CRAC in RBL cells.
In this report, we have found a new Ca 2ϩ -dependent negative feedback inactivation pathway that turns off I CRAC . In the presence of moderate Ca 2ϩ buffering (1.4 mM EGTA), I CRAC inactivates by between 50 and 100%. With high buffering (14 mM EGTA), inactivation is reduced to 30% (Fig. 1D). Hence the slow Ca 2ϩ -dependent inactivation accounts for between 20 and 70% of the inactivation of I CRAC and operates on a time scale of tens of seconds. Although it is activated by Ca 2ϩ entry through CRAC channels, it does not require the continuous presence of elevated Ca 2ϩ (Fig. 3). Slow inactivation is therefore endowed with the important property of maintaining CRAC channels inactive, even when the signal that switches on the inactivation pathway (a rise in Ca 2ϩ ) has been removed. Variable levels of activation of this slow inhibitory pathway will lead to different amounts of Ca 2ϩ entry through CRAC channels, thereby providing a further mechanism for generating graded Ca 2ϩ influx despite all-or-none activation of I CRAC . This feature, as well as its slow time course, suggests that slow inactivation plays an important role in shaping the profile of the Ca 2ϩ signal, and hence it will be an important factor that regulates slow Ca 2ϩdependent processes like exocytosis and gene transcription in nonexcitable cells.