Role of the Inositol 1,4,5-Trisphosphate Receptor in Ca2+ Feedback Inhibition of Calcium Release-activated Calcium Current (I crac)*

We examined the activation and regulation of calcium release-activated calcium current (I crac) in RBL-1 cells in response to various Ca2+ store-depleting agents. With [Ca2+] i strongly buffered to 100 nm,I crac was activated by ionomycin, thapsigargin, inositol 1,4,5-trisphosphate (IP3), and two metabolically stable IP3 receptor agonists, adenophostin A andl-α-glycerophospho-d-myoinositol-4,5-bisphosphate (GPIP2). With minimal [Ca2+] i buffering, with [Ca2+] i free to fluctuateI crac was activated by ionomycin, thapsigargin, and by the potent IP3 receptor agonist, adenophostin A, but not by GPIP2 or IP3 itself. Likewise, when [Ca2+] i was strongly buffered to 500 nm, ionomycin, thapsigargin, and adenophostin A did and GPIP2 and IP3 did not activate detectableI crac. However, with minimal [Ca2+] i buffering, or with [Ca2+] i buffered to 500 nm, GPIP2 was able to fully activate detectableI crac if uptake of Ca2+intracellular stores was first inhibited. Our findings suggest that when IP3 activates the IP3 receptor, the resulting influx of Ca2+ quickly inactivates the receptor, and Ca2+ is re-accumulated at sites that regulateI crac. Adenophostin A, by virtue of its high receptor affinity, is resistant to this inactivation. Comparison of thapsigargin-releasable Ca2+ pools following activation by different IP3 receptor agonists indicates that the critical regulatory pool of Ca2+ may be very small in comparison to the total IP3-sensitive component of the endoplasmic reticulum. These findings reveal new and important roles for IP3 receptors located on discrete IP3-sensitive Ca2+ pools in calcium feedback regulation ofI crac and capacitative calcium entry.

Many extracellular stimuli act through cell surface receptors to promote generation of intracellular inositol 1,4,5-trisphosphate (IP 3 ) 1 and consequently release intracellular Ca 2ϩ stores (1). The release of stored Ca 2ϩ is commonly accompanied by influx of Ca 2ϩ from the extracellular space, through the "capacitative Ca 2ϩ entry" pathway (2)(3)(4). Although depletion of intracellular stores and activation of capacitative calcium entry are inextricably linked, the underlying mechanism remains poorly defined (4).
Store-operated currents, which are proposed to mediate capacitative calcium entry, have been measured in various cell types (5). To date, "Ca 2ϩ release-activated Ca 2ϩ current" (I crac ) is the best defined member of the store-operated current family (6). I crac , measured in mast cells, T-lymphocytes, and rat basophilic leukemia (RBL-1) cells is both highly selective for Ca 2ϩ as the permeant ion and strongly inhibited by Ca 2ϩ feedback (5,(7)(8)(9). At least two forms of Ca 2ϩ -dependent inactivation of I crac have been reported, termed fast and slow inactivation. Fast inactivation occurs on a sub-second time scale and is caused by increases in sub-plasmalemmal Ca 2ϩ concentration in the vicinity of the Ca 2ϩ channel (7,9). Slow inactivation, on the time scale of tens to hundreds of seconds, is caused by increases in the bulk cytosolic [Ca 2ϩ ] and is partly dependent on refilling of intracellular Ca 2ϩ stores (8). In order to maximize, or indeed detect, the typically small whole-cell CRAC currents, it is necessary to minimize these forms of Ca 2ϩ feedback. First, the [Ca 2ϩ ] i is tightly buffered to basal levels with high concentrations of Ca 2ϩ chelators such as BAPTA or EGTA. Second, the cell membrane potential is held at depolarized levels to decrease the driving force for Ca 2ϩ entry between current measurements. A major disadvantage of these conditions is the inability to measure simultaneous changes in [Ca 2ϩ ] i .
Recently, simultaneous measurements of [Ca 2ϩ ] i and I crac have been reported in RBL-1 cells, an immortalized mast cell line (10). In this study, the degree of Ca 2ϩ buffering was reduced from a level which effectively clamps [Ca 2ϩ ] i to basal concentrations, to a minimal level, which permitted fluctuations of [Ca 2ϩ ] i . Under these conditions, adenophostin A, an IP 3 receptor agonist with 100-fold higher affinity for the IP 3 receptor compared with the native ligand IP 3 (11) induced Ca 2ϩ release and activation of I crac (10). Several other agents that would be expected to deplete intracellular stores, including analogues of IP 3 ((2,4,5)IP 3 , 3-deoxy-3-fluoro-IP 3 ), thapsigargin (a SERCA inhibitor), and ionomycin (a Ca 2ϩ ionophore), did not activate I crac under these conditions. It was suggested that the unique ability of adenophostin A to activate I crac was perhaps due to action at the level of Ca 2ϩ -dependent inactivation of the CRAC channel.
In the present study we investigated activation of I crac in RBL-1 cells under conditions of weak Ca 2ϩ buffering, while simultaneously monitoring [Ca 2ϩ ] i . We compared the ability of various store depletion agents and combinations of these agents to induce a rise in [Ca 2ϩ ] i and to activate detectable I crac . We now find that adenophostin A, thapsigargin, and ionomycin all activate I crac under low buffering conditions; however, IP 3 and a stable analogue of IP 3 , L-␣-glycerophospho-Dmyoinositol-4,5-bisphosphate (GPIP 2 ), do not. This failure was not the result of Ca 2ϩ feedback on the CRAC channel. Rather, deactivation of the current appears due to inactivation of IP 3 receptors and rapid refilling of critical Ca 2ϩ stores thereby diminishing the signal for activation of I crac . This work reveals a new mechanism that involves the IP 3 receptor for Ca 2ϩ feedback on I crac . These IP 3 receptors are likely located in spatially restricted regions within the endoplasmic reticulum and are closely coupled to activation and regulation of I crac .
Fura-2 Loading and Fluorescence Measurements-Coverslips with attached cells were mounted in a Teflon chamber and incubated at room temperature for 25 min in HEPES-buffered saline solution (HBSS; in mM, 140 NaCl, 4.7 KCl, 10 CsCl, 1.8 CaCl 2 , 1.13 MgCl 2 , 10 glucose, and 10 HEPES, pH 7.2) containing 1 M Fura-2 AM (Molecular Probes). Cells were then washed and bathed in HBSS for at least 10 min before Ca 2ϩ measurements were made.
Fluorescence was monitored by placing the Teflon chamber with the coverslip of Fura-2-loaded cells onto the stage of a Nikon Diaphot microscope (40ϫ Neofluor objective). Cells were excited by light (340 and 380 nm) from a Deltascan D101 (Photon Technology International Ltd.) light source equipped with a light path chopper and dual excitation monochromators. Emitted fluorescence (510 nm) was collected by a photomultiplier tube (Omega). All experiments were conducted at room temperature (22°C). Calibration of [Ca 2ϩ ] i was performed by reference to a look-up table created from Ca 2ϩ standards supplied by Molecular Probes.
In all experiments, upon forming the whole-cell configuration the cell membrane potential was held at ϩ30 mV (to minimize Ca 2ϩ entry and Ca 2ϩ -dependent inactivation of CRAC channels). Periodically (once every 5 s) the membrane potential was stepped to Ϫ100 mV (for 20 ms to assess I crac ), and then a voltage ramp to ϩ60 mV, over a period of 160 ms, was applied. Currents are normalized to cell capacitance. All voltages are corrected for a 10 mV liquid-junction potential. Membrane currents were amplified with an Axopatch-1C amplifier (Axon Instruments, Burlingham, CA). Voltage clamp protocols were implemented and data acquisition performed with PCLAMP 6.1 software (Axon Instruments). Currents were filtered at 1 kHz and digitized at 200-s intervals.

Activation of I crac Under Conditions of Strong and Weak
[Ca 2ϩ ] i Buffering-With pipette solutions strongly buffered (with 10 mM BAPTA) to 100 nM free [Ca 2ϩ ], and following the voltage protocol described under "Experimental Procedures," I crac was activated by a variety of store depletion agents in agreement with previous studies (6,9,10,13) (Fig. 1). Intracellular delivery of IP 3 (20 or 40 M), GPIP 2 (100 or 200 M), a non-metabolizable analogue of IP 3 , or adenophostin A (2 M), or extracellular addition of thapsigargin (1 M) or ionomycin (500 nM) activated an inward current with properties characteristic of I crac . In each case, the current decreased rapidly upon removal of extracellular Ca 2ϩ (Fig. 1B), showed inward rectification, and reversed direction at a potential positive of ϩ30 mV (Fig. 1A). There were no significant differences among the maximum amplitudes of the currents activated by these agents (Fig. 1C). The development times were also similar, although slightly prolonged for thapsigargin ( Fig. 1C; complete time course of I crac activation with 10 mM BAPTA, 100 nM [Ca 2ϩ ] i is shown by the open circles in Fig. 3).
With pipette solutions buffered with 0.1 mM BAPTA, [Ca 2ϩ ] i was weakly buffered and free to fluctuate. Under these conditions, each of the store-depleting agents caused an increase in [Ca 2ϩ ] i in Fura-2-loaded cells (Fig. 2, A-E upper traces). Adenophostin A ( Fig. 2A), IP 3 (Fig. 2B), and GPIP 2 (Fig. 2C) each produced a biphasic increase in [Ca 2ϩ ] i as follows: first a peak then a lower plateau, sustained slightly but significantly above basal levels. Thapsigargin (Fig. 2D) and ionomycin (Fig. 2E), agents that bypass the IP 3 receptor, produced sustained and monophasic increases in [Ca 2ϩ ] i . Importantly, all the sustained increases in Ca 2ϩ were reversed by removal of extracellular Ca 2ϩ and therefore reflected Ca 2ϩ entry. Ca 2ϩ entry presumably occurs mostly during the brief hyperpolarizing pulses. In support of this idea, expansion of the Ca 2ϩ traces revealed episodic Ca 2ϩ peaks, with a 5-s periodicity, that correspond to membrane hyperpolarizations. These Ca 2ϩ peaks were absent when extracellular Ca 2ϩ was removed (data now shown).
Simultaneous measurements of membrane currents in the same Fura-2-loaded cells revealed that adenophostin A, thapsigargin, and ionomycin, but neither IP 3 nor GPIP 2 , induced detectable activation of I crac (Fig. 2, A-E, lower traces). The currents activated by adenophostin A (Fig. 2A), thapsigargin (Fig. 2D), and ionomycin ( Fig. 2E) in weakly Ca 2ϩ -buffered cells displayed properties characteristic of I crac . In each case, the current decreased rapidly upon removal of extracellular FIG. 1. All store depletion agents activate I crac to a similar extent when [Ca 2؉ ] i is strongly buffered to basal levels. All data are from RBL-1 cells patched in the whole-cell mode with 10 mM BAPTA in the pipette (free Ca 2ϩ ϳ100 nM). A and B, cells were held at 30 mV, and I crac was measured once every 5 s with voltage steps to Ϫ100 mV (20 ms) followed by voltage ramps to 60 mV (160 ms) (A, top panel). A, a and b show raw traces in response to these voltage ramps and correspond to the times labeled in B. A, lower panel, the leak current measured in Ca 2ϩ -free HBSS has been subtracted from the current measured in the presence of Ca 2ϩ -HBSS to give I crac . B, activation of I crac , measured at Ϫ100 mV, is plotted against time. Current is normalized against cell capacitance and the current density (pA/pF) is plotted. 2 M adenophostin A was included in the patch pipette, and the whole-cell configuration was Ca 2ϩ . The currents also showed inward rectification and reversed direction at a potential positive of ϩ30 mV (not shown).
Following break in with pipettes containing adenophostin A, or following application of ionomycin, activation of I crac occurred after short delays of 49 Ϯ 11 and 88 Ϯ 9 s, respectively (Fig. 2, A and E). However, there was a considerably longer delay between the delivery of thapsigargin and subsequent activation of I crac (181 Ϯ 19 s) (Fig. 2D). I crac was activated by thapsigargin with a mean development time (time from the initial increase in current to development of current; see Ref. 13)) of 375 Ϯ 35 s, considerably slower than that for adenophostin A (132 Ϯ 25 s) or ionomycin (145 Ϯ 22 s) (Fig. 2F, open bars). The initial rise in Ca 2ϩ induced by thapsigargin indicates blockade of endoplasmic reticulum Ca 2ϩ -ATPases and the subsequent leak of Ca 2ϩ from intracellular stores. The delay before detectable I crac activation and the slow time course of development presumably reflect the time it takes to empty, to the necessary degree, the stores linked to I crac activation. It is because of this delayed and slow activation of I crac with thapsigargin that our earlier study failed to detect activation of I crac by thapsigargin in RBL-1 cells (10). Also, in our earlier study (10) ionomycin failed to activate I crac . However, a substantially higher concentration of ionomycin was used (5 M, as opposed to 500 nM in the current study), and at this concentration ionomycin raises [Ca 2ϩ ] i to extremely high levels, in excess of 1 M, which likely caused a strong Ca 2ϩ -dependent inactivation of I crac .
Activation of I crac at Elevated [Ca 2ϩ ] i -We next considered how minimal Ca 2ϩ buffering might prevent IP 3 and GPIP 2 from activating detectable I crac . Direct effects of global [Ca 2ϩ ] i on CRAC channels is unlikely to be responsible, because the agents that activated I crac actually raised steady-state [Ca 2ϩ ] i to somewhat higher levels than did IP 3 or GPIP 2 . Thus, either IP 3 and GPIP 2 are capable of raising Ca 2ϩ to higher levels than other agents in small, discrete regions close to the CRAC channels or, alternatively, I crac activated by IP 3 and GPIP 2 is for some reason more sensitive to Ca 2ϩ inhibition. Thus, we studied the effect of increased cytosolic Ca 2ϩ on I crac under conditions whereby both sub-plasmalemmal and global [Ca 2ϩ ] i were strongly buffered with 10 mM BAPTA. We buffered [Ca 2ϩ ] i to either a basal value ([Ca 2ϩ ] i ϳ100 nM) or a value ([Ca 2ϩ ] i ϳ500 nM) similar to that recorded in weakly buffered cells after activation with store depletion agents (Fig. 2E).
Under these conditions (10 mM (Fig. 3C) did not. Thus, IP 3 and GPIP 2 apparently fail to activate I crac because with these agents the signaling mechanism is more sensitive to inhibition by elevated [Ca 2ϩ ] i.
Adenophostin A (2.18 Ϯ 0.03 pA/pF), thapsigargin (2.57 Ϯ 0.65 pA/pF), and ionomycin (2.95 Ϯ 0.45 pA/pF) caused similar increases in I crac . These values were also similar to measurements made with free Ca 2ϩ buffered to 100 nM (Fig. 1C), being 90 Ϯ 3, 104 Ϯ 26, and 105 Ϯ 20%, respectively. It is worth noting, however, that although the peak magnitude of I crac measured with free Ca 2ϩ set to 100 or 500 nM was similar, a slow inactivation of the current was prominent at the higher cytosolic [Ca 2ϩ ]. I crac activated by adenophostin A fell to 90 Ϯ 12 and 56 Ϯ 9% of the peak after 300 s, at 100 and 500 nM free Ca 2ϩ , respectively. Over the same period, I crac activated by ionomycin also declined more at 500 nM, falling to 62 Ϯ 9% of the peak compared with 87 Ϯ 11% with 100 nM free Ca 2ϩ .
The mean development time of I crac for thapsigargin at 500 nM [Ca 2ϩ ] i (415 Ϯ 27 s) was again much slower than for adenophostin A (94 Ϯ 8 s) and ionomycin (68 Ϯ 5 s). There was also a considerable delay before thapsigargin induced detectable activation of I crac (295 Ϯ 34 s) compared to when free [Ca 2ϩ ] i was set to 100 nM (91 Ϯ 14 s) (Fig. 3D). For both adenophostin A and ionomycin, activation of I crac was seen after only a short delay (42 Ϯ 12 and 88 Ϯ 9 s, respectively).
Role of the IP 3 Receptor and SERCA Pumps-Buffering subplasmalemmal [Ca 2ϩ ] gradients (10 mM BAPTA), and raising bulk cytosolic [Ca 2ϩ ] (500 nM), mimicked some of the effects of low Ca 2ϩ buffering, namely failure of GPIP 2 and IP 3 to activate I crac and slow activation of I crac by thapsigargin. Given that adenophostin A, ionomycin, and thapsigargin all activate I crac under conditions where IP 3 and GPIP 2 are ineffective, direct Ca 2ϩ -dependent inactivation of the CRAC channel seems unlikely to be the cause. Rather, an involvement of the IP 3 receptor itself would be indicated. Thus, I crac could be inactivated if, when IP 3 and GPIP 2 are used, the IP 3 receptors were desensitized, and as a result the stores signaling I crac activation were refilled. Increases in [Ca 2ϩ ] i are known to promote dissociation of IP 3 from the IP 3 receptor leading to faster IP 3 receptor inactivation (14). Adenophostin A, by virtue of its high affinity for the IP 3 receptor, may be less susceptible to this increased rate of dissociation. Adenophostin A would therefore prolong IP 3 receptor activation, relative to the lower affinity agonists (GPIP 2 and IP 3 ), and maintain depleted Ca 2ϩ stores. If indeed IP 3 receptor inactivation does lead to rapid store refilling and turns off detectable I crac , then prevention of Ca 2ϩ re-uptake (with thapsigargin) should prevent the re-uptake of Ca 2ϩ that occurs with IP 3 and GPIP 2 , and thus these agents would act more like adenophostin A. To test this hypothesis, thapsigargin was applied to cells loaded with Fura-2 and patchclamped in the cell-attached mode (Fig. 4A). Patch pipettes contained 100 M GPIP 2 and 0.1 mM BAPTA. After 50 -70 s, a slight increase in [Ca 2ϩ ] i was detected indicating SERCA inhibition (Fig. 4A, upper trace). The whole-cell mode was then established in order to deliver GPIP 2 and measure I crac (Fig.  4A, lower trace). GPIP 2 caused a further, more substantial increase in [Ca 2ϩ ] i which was sustained (compare with GPIP 2 alone, Fig. 2C). Importantly, the increase in [Ca 2ϩ ] i that occurred upon delivery of GPIP 2 was accompanied by rapid activation of I crac . I crac was activated after only a short delay (33 Ϯ 8 s) and with a time course of 86 Ϯ 14 s. Had the current been activated by thapsigargin alone, the delay from break-in would have been in excess of 3 min and the development time in excess of 5 min (inferred from Fig. 2D). Development of I crac in the presence of GPIP 2 is therefore triggered by the rapid release of Ca 2ϩ through activated IP 3 receptors and independent of the slow leak of Ca 2ϩ induced by thapsigargin alone. The peak amplitude of I crac was not significantly altered by the presence (1.53 Ϯ 0.23 pA/pF) or absence of GPIP 2 (1.05 Ϯ 0.14 pA/pF, Fig. 2F). 2 This is to be expected if GPIP 2 increases the rate of release, through opening IP 3 receptors, but not the overall extent.
The preceding data suggest that when Ca 2ϩ is free to increase, re-uptake of Ca 2ϩ prevents detectable activation of I crac by GPIP 2 . If this assumption is correct, then delivery of thapsigargin before GPIP 2 , with Ca 2ϩ clamped to 500 nM with 10 mM BAPTA, should also allow rapid activation of I crac (Fig. 4B). As predicted, in cells pretreated with thapsigargin for 50 -70 s, delivery of GPIP 2 led to a rapid activation of I crac . Activation was detected after only a short delay (34 Ϯ 11 s) and had a development time of 81 Ϯ 13 s, compared with 415 Ϯ 27 s for thapsigargin alone. The amplitude of the current induced by GPIP 2 and thapsigargin combined (3.16 Ϯ 0.28 pA/pF) was once again not significantly greater than that that seen with thapsigargin alone (2.57 Ϯ 0.65 pA/pF).
These results reveal that SERCA pumps need to be blocked if GPIP 2 is to activate detectable I crac when [Ca 2ϩ ] i is weakly buffered. Refilling of stores by SERCA pumps must occur very quickly, and therefore, a decrease in IP 3 receptor activity must also occur quickly to explain the complete lack of detectable 2 Although the mean value for the current in the presence of GPIP 2 and thapsigargin was almost 50% larger than that for thapsigargin alone, the difference was not statistically significant by either t test or analysis of variance, owing to the variability in values of I crac from one preparation to another. However, we note that the value for GPIP 2 plus thapsigargin (1.53 Ϯ 0.23 pA/pF) was similar to the value for ionomycin under the same conditions (1.29 Ϯ 0.13 pA/pF). This suggests that if GPIP 2 plus thapsigargin does induce a current larger than for thapsigargin alone, it is likely due to increased release of Ca 2ϩ rather than a specific effect of GPIP 2 on the channel activation mechanism. I crac activation by GPIP 2 or IP 3 . GPIP 2 is successful in activating I crac after thapsigargin addition because despite rapid IP 3 receptor desensitization, Ca 2ϩ cannot be re-accumulated into the critical stores. This predicts that if the addition of the SERCA inhibitor is delayed for even a short interval following addition of GPIP 2 , I crac activation will still fail. As shown in Fig. 4C, this is indeed the case. When thapsigargin was added 50 -70 s after break-in with GPIP 2 , no rapid activation of I crac was observed; rather I crac activation occurred slowly (311 Ϯ 7 s) and after a latency of 222 Ϯ 21 s, similar to activation of I crac by thapsigargin alone (see Fig. 2D).

FIG. 3. Adenophostin A, thapsigargin, and ionomycin, but not IP 3 and GPIP 2 , activate I crac when [Ca 2؉ ] i is strongly buffered to activated levels (500 nM). A-E, cells were patched with
Collectively these results indicate that IP 3 and GPIP 2 activate IP 3 receptors and allow a rapid release of Ca 2ϩ from intracellular stores, but this receptor activity and enhanced Ca 2ϩ release are transient. Hence, when re-uptake of Ca 2ϩ is blocked prior to GPIP 2 exposure, the release induced by GPIP 2 is sufficiently sustained to activate I crac rapidly (Fig. 4A). However, if Ca 2ϩ re-uptake is allowed to proceed for even a minute after GPIP 2 delivery, blockade of SERCA pumps at this stage does not rapidly activate I crac , because the IP 3 receptors have already desensitized and Ca 2ϩ has been re-accumulated (Fig.  4C).
The Size of the Critical Stores Regulating I crac -Finally, therefore, we attempted experiments designed to demonstrate more directly the re-uptake of Ca 2ϩ into pools regulating I crac and to determine their size relative to the presumably larger IP 3 -and thapsigargin-sensitive stores. To this end, we activated signaling in cells with either GPIP 2 or adenophostin A, and we examined the size of the [Ca 2ϩ ] i signal on application of thapsigargin in Ca 2ϩ -free media (Fig. 5, lower panels). On delivery of adenophostin A or GPIP 2 , cells were exposed to either Ca 2ϩ -free or Ca 2ϩ -containing media before assessing the thapsigargin-sensitive store content in order to determine the contribution of Ca 2ϩ influx to the content of the stores. A residual, thapsigargin-sensitive, Ca 2ϩ store was detected in cells after exposure to adenophostin A (Fig. 5A) or GPIP 2 (Fig.  5B) in the absence of Ca 2ϩ influx (dotted lines). When Ca 2ϩ influx was allowed before store assessment (dashed lines), some refilling of stores occurred after both agents, although slightly more refilling appeared to occur in the presence of GPIP 2 . In each case, however, Ca 2ϩ stores were not completely refilled, because control cells exposed only to thapsigargin showed the largest release of Ca 2ϩ (solid lines).
Assessment of the Ca 2ϩ stores by inhibition of SERCA pumps revealed Ca 2ϩ release kinetics for adenophostin A (Fig.  5A, dashed lines) which were distinct from that in control cells (Fig. 5, solid line) or GPIP 2 -treated cells (Fig. 5B, dashed lines). For adenophostin A-treated cells, the rate of Ca 2ϩ release was faster than in control cells, with [Ca 2ϩ ] i returning to base line within 514 Ϯ 31 s, compared with 1058 Ϯ 60 and 966 Ϯ 92 s for control or GPIP 2 -treated cells, respectively. This finding fits with our prediction that, in the presence of adenophostin A, IP 3 receptors remain more active and this maintains critical stores depleted and I crac -detectable. GPIP 2 did not significantly increase the kinetics of Ca 2ϩ release over that in control cells, suggesting minimal residual activation of IP 3 receptors.
Although the residual Ca 2ϩ store was similar in cells treated with adenophostin A or GPIP 2 , simultaneous current measurements revealed I crac activation in response to adenophostin A only (Fig. 5, upper panels). This indicates that the pool of Ca 2ϩ involved in controlling I crac is very small in comparison to the total IP 3 -or thapsigargin-sensitive Ca 2ϩ pool. In support of this, thapsigargin, when added after adenophostin A, releases the remaining Ca 2ϩ store but does not further increase I crac amplitude (0.98 Ϯ 0.26 versus 0.96 Ϯ 0.12 pA/pF) (Fig. 6A). Also, adenophostin A and ionomycin activate I crac to a similar amplitude (0.96 Ϯ 0.12 pA/pF versus 1.16 Ϯ 0.14 pA/pF), but only the latter entirely depletes thapsigargin-sensitive Ca 2ϩ stores (Fig. 6B). Thus while neither adenophostin nor GPIP 2 release the entire thapsigargin-sensitive pool, adenophostin A releases and maintains empty, a small pool closely coupled to I crac activation, which is not maintained empty by GPIP 2 .
Our data strongly indicate that small specialized sub-compartments of the endoplasmic reticulum are coupled to activation of I crac . These sub-compartments are drained of Ca 2ϩ by adenophostin A and GPIP 2 , but when [Ca 2ϩ ] i rises, only adenophostin A keeps them in a sufficiently depleted state to activate detectable I crac , apparently because its high receptor affinity prevents Ca 2ϩ -dependent IP 3 receptor desensitization. Ionomycin and thapsigargin release Ca 2ϩ from all compartments, but release of these additional compartments is not coupled to I crac activation. DISCUSSION Our laboratory previously established a method for the simultaneous measurement of I crac and [Ca 2ϩ ] i in RBL-1 cells (10). Prevention of Ca 2ϩ entry between current measurements (by holding the cell membrane at a positive voltage) was the key to minimizing Ca 2ϩ -dependent inactivation of I crac . In that study we showed that adenophostin A, but not IP 3 congeners, was an effective activator of I crac when [Ca 2ϩ ] i was weakly buffered and free to rise. We now show that adenophostin A is effective because it keeps IP 3 receptors active, maintaining specific, critical stores sufficiently depleted of Ca 2ϩ to signal I crac activation. In contrast, other IP 3 receptor agonists, including IP 3 itself and GPIP 2 , fail to maintain active IP 3 receptors, allowing refilling of the critical stores. Depletion of Ca 2ϩ stores represents the sine qua non for capacitative calcium entry, and so the signal for I crac is substantially diminished and the current is not activated to a detectable level. This is a novel mode of Ca 2ϩ -dependent inactivation of I crac , in that the Ca 2ϩ -feedback occurs at the level of the IP 3 receptor, rather than the CRAC channel. The possible involvement of direct Ca 2ϩ -dependent inactivation of the CRAC channel being responsible for the failure of GPIP 2 or IP 3 to activate I crac is ruled out by the ability of other store-depleting agents, including thapsigargin and ionomycin, to activate I crac under similar conditions.
Our data reveal how IP 3 receptor activity can strongly influence the activity of CRAC channels. The IP 3 receptor participates "indirectly" in regulating CRAC channels in RBL-1 cells, because its activity determines the state of filling of a subset of Ca 2ϩ stores intricately linked to I crac activation. When the IP 3 receptor inactivates, these stores rapidly refill, through the activity of SERCA pumps, minimizing the signal for CRAC channel activation. Recently, the IP 3 receptor was also suggested to have a "direct" role in the regulation of Htrp3 channels stably expressed in HEK293 cells (15). The Htrp3 protein has previously been shown to form a channel activated by phospholipase C-linked agonists (16 -18). Kiselyov and colleagues (20) now report that overexpressed Htrp3 channels can also be activated by store depletion but that the IP 3 receptor directly interacts with the Htrp3 channel and that occupation of the IP 3 receptor is a requirement for Htrp3 activation. Whether the IP 3 receptor works to modulate directly endogenous Ca 2ϩ channels is unclear, although single channels that could be activated by either store depletion (in cell attached patches) or IP 3 (in excised patches) have been reported (19,20).
In RBL-1 cells, when Ca 2ϩ is free to fluctuate, adenophostin A, GPIP 2 , and IP 3 each readily release Ca 2ϩ from the specialized domains linked to I crac activation. However, only adenophostin A keeps these stores sufficiently empty to signal detectable I crac activation, because only it maintains IP 3 receptor activity. The distinct abilities of these agonists to maintain IP 3 receptor activity may reflect the very different affinities they display for the IP 3 receptor. Adenophostin A is ϳ100-fold more potent than 1,4,5-IP 3 (11), whereas GPIP 2 is ϳ10-fold weaker (21). Increases in Ca 2ϩ are reported to increase the rate of dissociation of IP 3 from the IP 3 receptor (14). The lower affinity agonists would dissociate from the IP 3 receptor more quickly than adenophostin A, leading to a more rapid receptor inactivation. In the continued presence of adenophostin A some slow inactivation of I crac is seen with higher cytosolic [Ca 2ϩ ] i (Fig.  3A). This may reflect either the slow dissociation of adenophostin A from the IP 3 receptor, reducing the signal for I crac activation, or a slow Ca 2ϩ -dependent inactivation of the CRAC channel (8,22).
RBL cells have been shown to express all three forms of the IP 3 receptor (23), with the predominating species being type 2. Because the number of IP 3 receptors involved in this subcompartment of the endoplasmic reticulum is likely small compared with the total number of receptors, any of the three types or possibly all three could be located there. There is evidence for specific involvement of the type 3 IP 3 receptor in regulating capacitative calcium entry (24). However, it is only the type-1 and type-2 receptors that appear to be regulated negatively by Ca 2ϩ (25,26), and thus type 1 and/or type 2 receptors would more likely be involved in the negative regulation seen here.
In addition to the findings discussed above, our results also suggest that the stores linked to I crac activation in RBL-1 cells reside within a specialized subcompartment of the endoplasmic reticulum. Adenophostin A activates I crac by releasing only a fraction of the thapsigargin-and ionomycin-releasable Ca 2ϩ stores, yet further release of residual Ca 2ϩ stores does not augment I crac (Figs. 2F and 6, A and C). Hofer et al. (27) have also shown that in RBL-1 cells, full activation of I crac occurs with only partial depletion of intracellular stores by ionomycin. We cannot determine if the failure of ionomycin and thapsigargin to induce a larger activation of I crac than adenophostin A reflects subcompartments of the endoplasmic reticulum, some of which are and some of which are not linked to I crac activation, or more simply a requirement for only partial depletion of all (or of critical) stores for maximal I crac activation. However, we also observed that with minimal [Ca 2ϩ ] i buffering, GPIP 2 depletes nearly as much of the thapsigargin-sensitive store as does adenophostin A, yet GPIP 2 fails to activate detectable I crac and adenophostin A activates it maximally. This striking distinction in the action of GPIP 2 and adenophostin A therefore must reflect the filling state of a pool that is very small in comparison to the total IP 3 -sensitive Ca 2ϩ stores.
There is already some evidence to suggest that specialized domains of the endoplasmic reticulum are coupled to activation of I crac in RBL-1 cells (13,28). Parekh et al. (28) observed differences in the concentration-effect relationships for Ca 2ϩ release and I crac activation which they interpreted as reflecting heterologous Ca 2ϩ pools, only some of which were involved in I crac regulation. Huang and Putney (13) observed different latencies for (2,4,5)IP 3 , thapsigargin, and ionomycin activation of Ca 2ϩ release and I crac . A significant delay was observed between the initiation of Ca 2ϩ stores depletion and the activation of I crac for (2,4,5)IP 3 and thapsigargin. However, with ionomycin, little or no delay was observed. A small compartment of the endoplasmic reticulum was suggested to regulate Ca 2ϩ entry, which was relatively resistant to store depletion by IP 3 and thapsigargin but not by ionomycin. This could result from a non-homogenous distribution of IP 3 receptors and leak channels. The more direct Ca 2ϩ -transporting action of ionomycin would not be affected by such distributions. There is evidence that sub-compartments of the endoplasmic reticulum regulate Ca 2ϩ influx in other cell types. Treatment of NIH-3T3 cells with the phorbol ester phorbol 12-myristate 13-acetate caused a loss of Ca 2ϩ (ϳ70%) from thapsigargin-and IP 3 -sensitive pools, but this depletion did not lead to activation of capacitative Ca 2ϩ entry (29).
When Ca 2ϩ stores are depleted by the SERCA inhibitor, thapsigargin, the release of Ca 2ϩ depends on poorly understood "leak" channels in the endoplasmic reticulum. Interestingly, and consistent with our earlier conclusions (13), our data suggest that the specialized stores linked to I crac activation are resistant to rapid emptying by the endogenous leak pathway. First, we observe a long delay between the initiation of Ca 2ϩ store depletion and activation of I crac by thapsigargin. Second, the current activated by thapsigargin in weakly buffered RBL-1 cells or in cells with [Ca 2ϩ ] i set to 500 nM has very slow development kinetics. These sub-compartments therefore either possess a low number of leak channels or lack them entirely or possess thapsigargin-insensitive pumps which act to slow the net Ca 2ϩ leak (30,31). The cause of the large increase in both the delay to onset and development kinetics of I crac activation by thapsigargin after reduction of Ca 2ϩ buffering from 10 to 0.1 mM BAPTA or on clamping [Ca 2ϩ ] i to 500 nM is unknown but presumably results from elevated [Ca 2ϩ ] i . There are at least three explanations. 1) The endogenous leak of Ca 2ϩ from intracellular stores is slower at higher cytoplasmic [Ca 2ϩ ], for example if basally active IP 3 receptors account for some of the endogenous leak. 2) In the presence of elevated [Ca 2ϩ ] i thapsigargin inhibition of SERCA pumps is slowed (32); therefore increased Ca 2ϩ is protective. As a result, blockade of the SERCA pumps may be delayed, allowing the pumps to continue working for a longer period in the presence of thapsigargin. 3) RBL-1 cells may possess thapsigargin-resistant Ca 2ϩ pumps, which are more active when cytoplasmic Ca 2ϩ is allowed to rise and therefore slow the net loss of Ca 2ϩ from intracellular stores (30). This is the first study demonstrating that Ca 2ϩ regulation of the IP 3 receptor plays a significant role in Ca 2ϩ -dependent negative feedback of I crac . Furthermore, these findings provide a mechanism explaining the previously documented failure of IP 3 and its congeners to activate measurable I crac in RBL-1 cells in the absence of maximal intracellular Ca 2ϩ buffering. In Jurkat T-cells, with [Ca 2ϩ ] i moderately buffered (1.4 mM EGTA) and Ca 2ϩ stores passively depleted (by the removal of extracellular Ca 2ϩ ), a slow turn-off of I crac , partially dependent on store refilling, accompanied the delayed rise in [Ca 2ϩ ] i induced by Ca 2ϩ influx (8). In this setting, store refilling would be expected to occur as soon as extracellular Ca 2ϩ is reapplied because nothing is actively keeping stores depleted. Another study in RBL cells used the low affinity Ca 2ϩ chelator TPEN to rapidly chelate and reset stored [Ca 2ϩ ] (27). I crac was assessed after removal of TPEN, when stores were completely refilled. These authors observed a CRAC current, which turned off over a period of ϳ70 s, confirming that store refilling turns off I crac but not immediately. Hence, both turn on and turn off of I crac are rather slow (ϳ1-2 min).
Because of these slow kinetics, in our experiments with IP 3 and GPIP 2 as activators, the stores linked to I crac activation may never be depleted sufficiently or for a long enough period to activate I crac to a detectable level. Importantly, in our experiments, despite some refilling of Ca 2ϩ stores in the presence of GPIP 2 , stores were not completely refilled (Fig. 5B), and although the currents underlying Ca 2ϩ entry for GPIP 2 and IP 3 were beneath the level of detection, Ca 2ϩ entry still occurred as evidenced by a sustained increase in [Ca 2ϩ ] i that reversed upon removal of extracellular Ca 2ϩ (Fig. 2, B and C). Hence, the partially empty stores must signal some Ca 2ϩ entry, but this is not detectable as a current. I crac may be active under all conditions in which capacitative calcium entry occurs, but it may be below the threshold for detection. In the absence of specific pharmacological probes for these channels, it cannot be disproved that an as yet unidentified but distinct pathway contributes to entry under these conditions. Interestingly, Zhang and McCloskey (33) using the nystatin perforated patch technique and with no added intracellular calcium buffer were able to detect an inwardly rectifying Ca 2ϩ current in RBL-2H3 cells in response to either immunoglobulin E (presumably acting through IP 3 ) or thapsigargin. This current was seen at 37°C but not at room temperature. Hoth et al. (34) have demonstrated that in T-lymphocytes the ability to detect I crac under conditions of low physiological calcium buffering in the patch pipette depends on endogenous calcium buffering by mitochondria. Thus, the ability to measure a calcium current associated with capacitative calcium entry may depend on a number of technical factors, including both experimentally applied as well as physiological calcium buffering. But it is clear from the data in this report and others in the literature that significant entry of calcium can occur through store-operated channels, whatever their nature, when the magnitude of the associated current is below the level of detection with presently available methodologies.