Store-operated Ca²⁺ entry: dynamic interplay between endoplasmic reticulum, mitochondria and plasma membrane

Abstract

In eukaryotic cells, hormones and neurotransmitters that engage the phosphoinositide pathway evoke a biphasic increase in intracellular free Ca²⁺ concentration: an initial transient release of Ca²⁺ from intracellular stores is followed by a sustained phase of Ca²⁺ influx. This influx is generally store-dependent and is required for controlling a host of Ca²⁺-dependent processes ranging from exocytosis to cell growth and proliferation. In many cell types, store-operated Ca²⁺ entry is manifest as a non-voltage-gated Ca²⁺ current called I_CRAC (Ca²⁺ release-activated Ca²⁺ current). Just how store emptying activates CRAC channels remains unclear, and some of our recent experiments that address this issue will be described. No less important from a physiological perspective is the weak Ca²⁺ buffer paradox: whereas macroscopic (whole cell) I_CRAC can be measured routinely in the presence of strong intracellular Ca²⁺ buffer, the current is generally not detectable under physiological conditions of weak buffering following store emptying with the second messenger InsP₃. In this review, I describe some of our experiments aimed at understanding just why InsP₃ is ineffective under these conditions and which lead us to conclude that respiring mitochondria are essential for the activation of I_CRAC in weak intracellular Ca²⁺ buffer. Mitochondrial Ca²⁺ uptake also increases the dynamic range over which InsP₃ functions as the second messenger that controls Ca²⁺ influx. Finally, we find that Ca²⁺-dependent slow inactivation of Ca²⁺ influx, a widespread but poorly understood phenomenon that helps shape the profile of an intracellular Ca²⁺ signal, is regulated by mitochondrial Ca²⁺ buffering. Thus, by enabling macroscopic store-operated Ca²⁺ current to activate and then by controlling its extent and duration, mitochondria play a crucial role in all stages of store-operated Ca²⁺ influx. Store-operated Ca²⁺ entry reflects therefore a dynamic interplay between endoplasmic reticulum, mitochondria and plasma membrane.

An increase in intracellular free Ca²⁺ concentration is a ubiquitous signalling mechanism that regulates a broad spectrum of kinetically disparate processes ranging from exocytosis to cell growth and proliferation (Carafoli, 2002). Two major ways whereby eukaryotic cells can increase their cytosolic Ca²⁺ concentration are release of the ions from intracellular stores (mainly the endoplasmic/ sarcoplasmic reticulum) or entry of Ca²⁺ into the cell from the external solution (Berridge, 1993; Pozzan et al. 1994). Because the intracellular stores have a limited capacity, Ca²⁺ entry into the cell down an enormous electrochemical potential gradient is important for supporting the activities of numerous Ca²⁺-dependent processes. In excitable tissues, like nerve and muscle, Ca²⁺ entry is accomplished by the opening of Ca²⁺-permeable ion channels (voltage- and ligand-gated; Burnashev, 1998; Catterall, 2000). Much is known about the biophysical properties of these proteins, their stoichiometry, and even the identity of amino acids that are crucial for key channel functions like selectivity and gating (Burnashev, 1998; Catterall, 2000; Hille, 2001).

However, until relatively recently, our understanding of how Ca²⁺ enters non-excitable cells has been weak. Non-excitable cells are cells that do not fire action potentials and include the cells of the immune system, endothelia lining the blood vessels, epithelia that line the respiratory and digestive tracts, and glial cells in the brain. In these cell types, Ca²⁺ entry is essential for maintaining normal cell function but voltage-operated Ca²⁺ channels tend not to be expressed and ligand-gated Ca²⁺-permeable channels are sparse. Instead, a major route for Ca²⁺ influx in these cells is via store-operated Ca²⁺ channels (Parekh & Penner, 1997).

Store-operated Ca²⁺ influx

In an erudite paper published in 1986, James Putney proposed a model for Ca²⁺ influx in non-excitable cells, which he called capacitative Ca²⁺ entry (Putney, 1986). The fundamental tenet of this model was that the Ca²⁺ entry pathway in the plasma membrane was controlled by the Ca²⁺ content of the intracellular stores. As the store Ca²⁺ content fell, a signal was somehow sent from the stores which caused the Ca²⁺ entry pathway to open (now referred to as store-operated Ca²⁺ entry). Although indirect evidence began to accrete in support of Putney's model, it was not until 1992 that Reinhold Penner directly demonstrated, in an elegant series of experiments, that Ca²⁺ store depletion resulted in the activation of a Ca²⁺-selective current which was called I_CRAC (Hoth & Penner, 1992). Ca²⁺ entry through store-operated Ca²⁺ channels is important for regulating a host of temporally diverse processes from exocytosis and enzymatic activity to gene transcription and cell proliferation (Parekh & Penner, 1997; Putney & McKay, 1999). Aberrant store-operated entry has been reported to underlie debilitating human diseases like certain primary immunodeficiencies (Partiseti et al. 1994) and acute pancreatitis (Raraty et al. 2000; Parekh, 2000). It now seems clear that store-operated Ca²⁺ channels are not a homogeneous class but instead reflect a family with different biophysical properties. The best characterised store-operated current is I_CRAC, which is found in several types of cell including mast cells, lymphocytes, macrophages, megakaryocytes and hepatocytes (Parekh & Penner, 1997; Lewis, 1999).

Although store-operated entry is a widespread Ca²⁺ influx pathway, it is by no means the only route for Ca²⁺ entry in non-excitable cells. In addition to a range of second-messenger operated Ca²⁺-permeable channels (activated by Ca²⁺, InsP₃/InsP₄, cyclic nucleotides; reviewed in Clapham, 1995), growing evidence indicates that a Ca²⁺ influx pathway controlled by arachidonic acid might be prominent under certain conditions (reviewed in Taylor, 2002).

In this Wellcome Prize Lecture review, I shall focus on some of our work over the past several years that has been aimed at understanding the mechanisms that control I_CRAC. It is, as befits a prize lecture, a subjective account and is not a general review of the literature. Because of space constraints, certain papers have not been cited and I apologise to those authors for this.

Basic features of I_CRAC

CRAC channels are activated by the process of emptying the intracellular Ca²⁺ stores. It is of little consequence how the stores are actually emptied, the net result is the same (Hoth & Penner, 1992; Parekh & Penner, 1997). Physiologically, stores are emptied following cell-surface receptor stimulation which increases the levels of the Ca²⁺ releasing messenger InsP₃. Experimentally, stores can be emptied by increasing net Ca²⁺ flux out of the stores following application of receptor agonists, increasing intracellular InsP₃ by dialysis with a patch pipette or by exposure to Ca²⁺ ionophores. Alternatively, suppressing Ca²⁺ re-uptake into the stores allows the endogenous Ca²⁺ leak pathway to deplete the stores. Inhibition of store refilling can occur by exposure to thapsigargin, a potent inhibitor of the sarcoplasmic/endoplasmic reticular Ca²⁺ ATPase (SERCA) pumps on the stores.

The Ca²⁺ flux through CRAC channels gives rise to the small Ca²⁺ current I_CRAC, which can be measured directly and unambiguously using the whole cell patch clamp technique. With Ca²⁺ as the charge carrier, I_CRAC activates mono-exponentially (time constant of around 20 s; Fig. 1A) and is a non-voltage-gated, inwardly rectifying current with a very positive reversal potential (> +60 mV; Fig. 1B), the latter indicating high selectivity for Ca²⁺. Although Na⁺ ions outnumber Ca²⁺ by 70:1 in the external solution, neither the extent of I_CRAC nor the positive reversal potential (still > +60 mV) appear to be altered by substituting external Na⁺ for organic cations like NMDG⁺ in the presence of physiological levels of external Ca²⁺ (1–2 mm) (Fierro & Parekh, 2000). Moreover, removal of external Ca²⁺ in the continuous presence of external Na⁺ and Mg²⁺ abolishes the current completely (Hoth & Penner, 1992; Zweifach & Lewis, 1993; Fierro & Parekh, 2000). Because of the lack of a clear reversal potential for I_CRAC with Ca²⁺ as the charge carrier, the permeability ratio of Ca²⁺ to Na⁺ (P_Ca/P_Na) has been hard to establish. Using the ‘added buffer method’, where cells are loaded with sufficient Fura-2 to capture all the incoming Ca²⁺ through CRAC channels, it has been estimated that CRAC channels are as selective for Ca²⁺ as their voltage-operated counterparts (Hoth, 1995), the latter exhibiting a P_Ca/P_Na of 1000. However, CRAC channels discriminate between various divalent cations, with Ba²⁺ and Sr²⁺ supporting less current than Ca²⁺ (Hoth & Penner, 1992; Zweifach & Lewis, 1993; Fierro & Parekh, 1999a). The single CRAC channel conductance is low. Stationary noise analysis has estimated it to be well below 1 pS, beyond the current level of detection (Hoth & Penner, 1992; Zweifach & Lewis, 1993). A consequence of this is that there are estimated to be at least 5000 functional CRAC channels in the plasma membrane following stimulation.

Figure 1. Development of I_CRAC following store depletion.

A, the upper panel shows the typical voltage ramp protocol used to monitor I_CRAC in mast cells, the related rat basophilic leukaemia (RBL-1) cell line and jurkat T-lymphocytes. The ramp spans −100 to +100 mV in 50 ms, and is applied from a holding potential of 0 mV once every 2 s. The lower panel depicts the time course of activation of I_CRAC. The current amplitude, measured at −80 mV from each ramp, has been normalised for cell size by dividing the current by cell capacitance. The thick line shows a mono-exponential fit to the activation time course. B shows the current-voltage relationship, once the current had reached steady state. Note the inward rectification and very positive reversal potential, characteristic of I_CRAC. The pipette solution is caesium glutamate based, supplemented with InsP₃ and 10 mm EGTA.

Activation mechanism

The unique gating of I_CRAC has made it an attractive target of investigation for the identification of new signal transduction pathways. CRAC channels are activated by the emptying of intracellular Ca²⁺ stores. But just how a reduction in store content is translated into opening of these channels remains unclear. A sensor within the stores presumably detects the fall in Ca²⁺ content and this then initiates a process resulting in a signal being sent from the stores to the channels in the plasma membrane. Virtually nothing is known about the identity of the sensor, other than it is probably not calreticulin (Fasolato et al. 1998), the major intraluminal Ca²⁺ binding protein of non-muscle cells. As for the activation signal, three models have been put forward (see Fig. 2; reviewed in Parekh & Penner, 1997; Putney & McKay, 1999): (1) the vesicular fusion hypothesis (in which CRAC channels are inserted into the plasma membrane upon store depletion via an exocytotic mechanism; Yao et al. 1999; Alderton et al. 2000); (2) conformational coupling (where InsP₃ receptors on the stores are physically attached to CRAC channels in the plasma membrane; Berridge, 1995; Kiselyov et al. 1999), including the related secretion-like coupling model (where peripheral endoplasmic reticulum (ER) moves to the plasma membrane upon store depletion so that coupling can take place and this movement is regulated by the peripheral cytoskeleton; Patterson et al. 1999; Rosado et al. 2000), and (3) the diffusible messenger model (in which a small mobile factor is released from the stores upon their emptying and which then directly opens CRAC channels; Ramdriamampita & Tsien, 1993; Csutora et al. 1999).

Figure 2. The three contemporary models proposed to account for the activation mechanism of I_CRAC.

The vesicular fusion model does not provide any insight into the activation signal itself. It postulates instead that this signal results in the insertion of CRAC channels into the plasma membrane. The fusion model therefore could be accommodated within either of the other two contemporary models. However, we found that dialysing the cytosol with small GTP-binding proteins known to be involved in vesicular trafficking (Parekh & Penner, 1996), interfering with GTP-binding protein function (Fierro & Parekh, 1999b) or exposing cells to either clostridial neurotoxins or to recombinant proteins that suppress exocytosis (in collaboration with Professor Burgoyne, Liverpool) all fail to affect the activation of I_CRAC, findings that are not readily reconcilable with a fusion mechanism. The secretion-like conformational coupling model predicts that, first, intimate interactions between the ER and plasma membrane are essential for I_CRAC to activate, second, InsP₃ receptors are required for all stages of store-operated Ca²⁺ entry, and third, that interfering with the peripheral cytoskeleton should impair the ability of I_CRAC to activate. We have tested these predictions in rat basophilic leukaemia (RBL-1) cells (Bakowski et al. 2001). Inflating the cell by applying positive pressure to the patch pipette (cell ballooning) did not interefere with the ability of CRAC channels to activate, in spite of some rather marked changes in the cellular architecture (Bakowski et al. 2001). A variety of InsP₃ receptor inhibitors did not affect I_CRAC development, in spite of occupying InsP₃ receptors, and pharmacological tools directed towards re-shaping the cytoskeleton consistently failed to interfere with the activation of I_CRAC (Bakowski et al. 2001). Hence, a secretion-like coupling model is probably not a viable candidate for the activation of CRAC channels in RBL-1 cells. As for the diffusible messenger hypothesis, no convincing candidate has yet been identified. Recently, we have found that inhibition of the lipoxygenase family of enzymes impairs the activation of I_CRAC (Glitsch et al. 2002a). The inhibitors are much less effective if administered after CRAC channels have been activated. Such findings raise the intriguing possibility that a lipoxygenase might be involved in the activation mechanism, although, as discussed in Glitsch et al. (2002), such a conclusion is not unequivocal as it is based on a pharmacological approach.

Clearly, the ability to study single CRAC channel currents in an inside-out patch would greatly facilitate identification of key regulatory molecules involved in channel gating, as would the identification of the CRAC channel gene(s). It was recently suggested that the protein TRPV6 comprises all or part of the CRAC channel pore (Yue et al. 2001). But work by us and others suggests that this is probably not the case (Bakowski & Parekh, 2002a,b; Hermosura et al. 2002; Prakriya & Lewis, 2002). Understanding the molecular identity of the CRAC channel as well as the activation signal remain outstanding questions in the field.

The weak buffer paradox

Another major problem concerning I_CRAC is the weak Ca²⁺ buffer paradox. Whereas I_CRAC can be activated maximally following whole cell dialysis with a patch pipette containing InsP₃ and strong buffer (10 mm EGTA), the current generally fails to develop in the presence of weak Ca²⁺ buffer (0.1 mm EGTA; Fig. 3, modified from Bakowski & Parekh, 2001; see also Broad et al. 1999; Fierro & Parekh, 2000). Such low concentrations of EGTA are only slightly larger than the levels of endogenous mobile Ca²⁺ buffers (Zhou & Neher, 1993). InsP₃ is able to release Ca²⁺ from the stores under these conditions (Parekh et al. 1997). Hence the inability of InsP₃ to activate I_CRAC in weak Ca²⁺ buffer does not reflect a failure to mobilise Ca²⁺ from the stores. Clearly, resolution of why InsP₃ is so effete in weak buffer is essential if we are to place I_CRAC in any kind of physiological context.

Figure 3. The weak buffer paradox.

A, InsP₃ generally fails to activate I_CRAC in weak intracellular Ca²⁺ buffer (0.1 mm EGTA) whereas it is effective in strong buffer (10 mm EGTA). B, current-voltage relationships for the two recordings from A, taken at steady state. C, the mean amplitude of I_CRAC (left-hand panel) and the fraction of cells responding (right-hand panel) for each condition is shown. In this and all subsequent figures, data from RBL-1 cells are shown. ^*P < 0.01.

The first explanation that we considered was that the inability of InsP₃ to activate I_CRAC in weak Ca²⁺ buffer was an idiosyncrasy of our recording conditions. Like most electrophysiologists, we routinely work at room temperature. To see whether InsP₃ was effective at physiological temperature, we compared the effects of InsP₃ in weak versus strong buffer at 36 °C. However, InsP₃ was still ineffective in weak buffer (Fig. 4A and B, modified from Fierro et al. 2000). We also systematically altered the composition of the pipette solution to see whether we could render InsP₃ more effective. However, InsP₃ was still largely ineffective in weak buffer in either a Cs⁺- or K⁺-based pipette solution, when Cl⁻ was the dominant anion or when EGTA was replaced by BAPTA or dimethyl BAPTA (Fierro & Parekh, 2000; Gilabert & Parekh, 2000). Non-metabolisable analogues of InsP₃ were also incapable of routinely generating the current, as was receptor stimulation, the latter being used as a physiological route to generate InsP₃.

Figure 4. InsP₃ still fails to activate I_CRAC in weak Ca²⁺ buffer at physiological temperature.

Experiments were carried out at 36 °C. A, whereas InsP₃ evoked a large I_CRAC at 36 °C in 10 mm EGTA, no current was detectable in 0.1 mm EGTA. B, aggregate data are summarised. ^*P < 0.01 (Student's t test).

Passive activation of I_CRAC

An important clue came from experiments designed to try to understand the mechanism underlying activation of I_CRAC following passive depletion of stores (Fierro & Parekh, 1999c). Dialysis with a high concentration of Ca²⁺ chelator alone (10 mm EGTA or BAPTA) is sufficient to maximally activate I_CRAC, a process referred to as passive depletion of stores (Hoth & Penner, 1992). Passive depletion is thought to rely only on the endogenous leak of Ca²⁺ from the stores. In strong buffer (several millimolar chelator), cytoplasmic Ca²⁺ is strongly clamped at very low levels, so as Ca²⁺ leaks out of the stores it is captured by the chelator and cannot therefore be taken back up. Stores will be drained gradually of Ca²⁺ and hence empty sufficiently for I_CRAC to activate (passive activation). Figure 5A–C shows the pattern of development of I_CRAC following dialysis with 10 mm of EGTA, BAPTA or dimethyl BAPTA. Superimposed on the EGTA trace in Fig. 5A is a recording from a cell in which InsP₃ was also included in the pipette. With InsP₃, I_CRAC activated rapidly and mono-exponentially. In marked contrast, the pattern of development of the current to passive depletion was much slower than that seen with InsP₃. With EGTA (Fig. 5A), BAPTA (Fig. 5B) or dimethyl BAPTA (Fig. 5C), the development of the current was kinetically complex: after a sizeable delay of around 60 s, I_CRAC initially activated slowly (first phase of development), followed by a faster developing phase (second phase of development). We analysed the various features of the current (delay, rate of development of the two phases as well as their durations, their relative contributions to total current amplitude) and these were not significantly different between the chelators (Fierro & Parekh, 1999b). These chelators have apparent equilibrium dissociation constants of 150, 225 and 150 nm, respectively, and on-rates of 1.5 × 10⁶m⁻¹ s⁻¹ for EGTA and 5 × 10⁸m⁻¹ s⁻¹ for BAPTA and dimethyl BAPTA at pH 7.2. Hence neither the speed of binding Ca²⁺ nor the equilibrium affinity of the chelators accounts for the complex kinetic profile.

Figure 5. Passive activation of I_CRAC.

A, comparison of the time course of development of I_CRAC following dialysis with either 10 mm EGTA alone or 30 μm InsP₃+ EGTA. Note the initial slow development of I_CRAC followed by a second, faster phase in response to EGTA alone. B, development of I_CRAC following dialysis with 10 mm BAPTA or dimethyl BAPTA (panel C).

Importance of SERCA pumps

We found that SERCA pumps were extremely influential in shaping the pattern of development of I_CRAC following passive store depletion (Fierro & Parekh, 1999c). When SERCA pumps were inhibited with either cyclopiazonic acid or thapsigargin, the kinetics of I_CRAC changed dramatically compared with the current activated in 10 mm EGTA alone (Fig. 6A). The biphasic nature of the current (initial slow phase followed by a second faster phase), which was always seen with passive depletion using 10 mm EGTA, was not present when SERCA pumps were blocked. Now, the intial slow component was lost and instead the rapid phase developed in isolation. The slow initial activation therefore reflects a dynamic interplay between Ca²⁺ leakage from the stores and SERCA pump-mediated Ca²⁺ reuptake. The Ca²⁺ pumps help shape the profile of I_CRAC development, even in 10 mm EGTA.

Figure 6. SERCA pumps shape the pattern of passive activation of I_CRAC.

A, inclusion of cyclopiazonic acid (CPA) eliminates the initial slow phase of current development, leaving the second component intact. B, following dialysis with 2.5 mm EGTA, I_CRAC activated partially in this cell. Subsequent application of thapsigargin increased the extent of the current, as well as its rate of development.

For lower concentrations of EGTA (1–2.5 mm), the effects of SERCA pumps were even more striking. Following dialysis with EGTA < 2.5 mm, I_CRAC activated slowly and mono-phasically to reach a submaximal steady-state amplitude (Fig. 6B). Once the small current had reached steady-state, application of the SERCA pump blocker thapsigargin increased the size of the current further. This demonstrates that the submaximal I_CRAC reflected a steady state between Ca²⁺ efflux from the stores and reuptake by SERCA pump activity (Fierro & Parekh, 1999b). Moreover, even in the presence of a relatively high concentration of EGTA (2.5 mm), SERCA pumps were still active and were able to resequester Ca²⁺ into the stores.

SERCA and InsP₃

Although the preceding results demonstrate that SERCA pumps are very powerful and their activity needs to be overcome in order for I_CRAC to activate fully, they do not address the key question as to whether the pumps can sculpt the ability of the natural trigger, namely InsP₃, to activate I_CRAC under physiological conditions of weak Ca²⁺ buffering. Because the current cannot be measured when cells are dialysed with InsP₃ in weak Ca²⁺ buffer (Fig. 3), we therefore determined the lowest concentration of EGTA that would still enable us to reliably measure macroscopic I_CRAC and, having found this, we examined the effects of SERCA pump block on the features of the current (Glitsch & Parekh, 2000). Figure 7A summarises results from experiments in which cells were dialysed with a maximally effective concentration of InsP₃ and different concentrations of EGTA. Concentrations of EGTA less than 0.35 mm generally failed to support any macroscopic activation of I_CRAC whereas close to maximal activation was seen with 1 mm EGTA. However, with 0.6 mm EGTA, I_CRAC clearly had a sub-maximal amplitude in spite of the cells being continuously exposed to a high concentration of InsP₃ via the patch pipette. We reasoned that the stores might nevertheless refill partially in the presence of such a moderate Ca²⁺ chelator concentration and this would therefore reduce the overall extent of I_CRAC activation. If true, then one might expect that the size of I_CRAC would increase under these conditions if SERCA pumps were to be blocked since this would suppress store refilling. The results of Fig. 7B show that this was indeed the case. Inclusion of thapsigargin in the recording pipette together with InsP₃ and 0.6 mm EGTA significantly increased the size of the current compared with InsP₃ and 0.6 mm EGTA alone. Moreover, the delay before the current activated was also reduced by blocking SERCA pumps, although the time to peak was unaffected (Fig. 7C).

Figure 7. SERCA pumps function in the presence of InsP₃.

A, the relationship between intrapipette EGTA concentration and amplitude of I_CRAC (evoked by dialysis with 30 μm InsP₃) is plotted. Note that I_CRAC was clearly submaximal in the presence of 0.6 mm EGTA. Each point is the mean ±s.e.m. of at least 5 cells. B, recordings are shown for a cell dialysed with 30 μm InsP₃ and 0.6 mm EGTA (•) and for one dialysed with this solution but supplemented with 2 μm thapsigargin, a SERCA blocker (○). Note the increase in the size of I_CRAC following inhibition of the pumps. C, the amplitudes, delays and times to peak are summarised for experiments as in B. Filled circles denote InsP₃+ 0.6 mm EGTA whereas open cirlces represent InsP₃+ 0.6 mm EGTA + thapsigargin.

Collectively, these results demonstrate that SERCA pumps can refill stores, even in the continuous presence of InsP₃, and to the extent that I_CRAC activates only partially.

InsP₃ activates I_CRAC in physiological buffer provided SERCA pumps are blocked

The preceding results provide an explanation as to why InsP₃ is largely ineffective in activating I_CRAC under physiological conditions of weak Ca²⁺ buffering. The SERCA pumps are so effective that they can prevent InsP₃ from depleting the stores sufficiently or long enough for I_CRAC to activate. If true, then one would expect InsP₃ together with a SERCA pump blocker to activate I_CRAC in weak buffer. Figure 8 shows experiments designed to test this (Fierro & Parekh, 2000). Whereas InsP₃ was ineffective in weak Ca²⁺ buffer, the combination of InsP₃ and thapsigargin evoked a large current (Fig. 8A). The effects of thapsigargin were mimicked by the different SERCA pump blockers cyclopiazonic acid (Fig. 8B) and thapsigargicin (which is almost as potent as thapsigargin but less hydrophobic). Inhibition of SERCA pumps alone, in the absence of InsP₃, also activated I_CRAC (Fig. 8B; Fierro & Parekh, 2000). This occurred at a much slower rate than that seen in the presence of InsP₃ because store depletion to thapsigargin alone occurs via the background leak pathway.

Figure 8. InsP₃ activates I_CRAC in weak buffer provided SERCA pump activity is compromised.

A, InsP₃ together with thapsigargin activates I_CRAC in weak buffer whereas InsP₃ alone is ineffective. B, InsP₃ with cyclopiazonic acid, a structurally distinct SERCA pump blocker, is also effective. SERCA pump block alone (2 μm Thap + 0.1 mm EGTA) can also activate I_CRAC, albeit at a much slower rate.

In fact, provided SERCA pumps were blocked, I_CRAC could activate even in the absence of any exogenous Ca²⁺ chelator (Fierro & Parekh, 2000). Figure 9 shows the development of I_CRAC following dialysis with InsP₃, thapsigargin, 100 μm Ca²⁺ and no added Ca²⁺ chelator. Ca²⁺ in the pipette solution was weakly buffered by ATP (2 mm) and glutamate (main anion in the patch pipette). The current activated at the normal rate (trace marked InsP₃+ Thap + Ca²⁺ in Fig. 9A). Dialysis with thapsigargin and 100 μm Ca²⁺ (in the absence of InsP₃) also activated I_CRAC, with the typical slow kinetics expected for thapsigargin alone (trace marked Thap + Ca²⁺). Aggregate data are summarised in Fig. 9B. Although the free Ca²⁺ concentration in the cells is not known precisely under these conditions (due to exogenous buffering/removal), in experiments with thapsigargin alone the slow development of I_CRAC would ensure that reasonable dialysis with micromolar Ca²⁺ concentrations would have occurred as I_CRAC activated. I_CRAC can therefore activate when cytosolic Ca²⁺ is in the micromolar range. Moreover, exogenous Ca²⁺ chelators are not required for I_CRAC to be activated.

Figure 9. SERCA pump block is sufficient to activate I_CRAC even in high intracellular Ca²⁺.

A, InsP₃ and thapsigargin can activate I_CRAC in weak buffer even in the presence of high intracellular Ca²⁺ (100 μm total CaCl₂ in the pipette solution, free Ca²⁺ concentration estimated to be in the micromolar range). Thapsigargin alone was also able to evoke I_CRAC under these conditions, but after a delay and at a slower rate. B, aggregate data are summarised.

SERCA and secretion

Upon stimulation, mast cells and basophils degranulate, thereby releasing a host of factors into the extracellular medium which are important in shaping a local inflammatory response. We measured secretion from individual RBL-1 cells using the capacitance method to track vesicular fusion (Artalejo et al. 1998). We found that exocytosis required both an increase in Ca²⁺ and a GTP-dependent step, at least in the whole cell configuration of the patch clamp technique. Strikingly, dialysis with InsP₃ in weak Ca²⁺ buffer failed to evoke any clear secretory response unless thapsigargin was present (Fig. 10). No doubt, part of the effects of thapsigargin are likely to involve reduced Ca²⁺ removal by SERCA pumps. But it seems likely that Ca²⁺ entry following the development of I_CRAC, as a consequence of exposure to the combination of InsP₃ and thapsigargin, will also promote a larger secretory response.

Figure 10. The combination of InsP₃ and thapsigargin evokes a robust secretory response in weak buffer, whereas InsP₃ alone is largely ineffective.

InsP₃ alone (•) or in combination with GTPγS (▵) did not trigger a detectable increase in membrane capacitance, whereas InsP₃ together with thapsigargin and GTPγS produced a prominent secretory response (○). In these experiments, cells were clamped at −60 mV.

Central role for mitochondria in development of I_CRAC

SERCA pump activity therefore appears to be one of the main determinants of whether I_CRAC activates or not under physiological conditions. Only when SERCA pump activity is reduced does I_CRAC develop. How might SERCA pump activity be reduced under physiological conditions of weak Ca²⁺ buffering? We reasoned that an increase in the rate of Ca²⁺ removal from the cytosol by another clearance mechanism might be particularly effective because this would compete effectively with SERCA pumps and hence reduce the rate and extent of store refilling. Stores would therefore empty more and I_CRAC should activate. Moreover, enhanced Ca²⁺ clearance away from the endoplasmic reticulum might also reduce Ca²⁺-dependent inactivation of InsP₃ receptors (Taylor & Traynor, 1995). Combined, this would enable InsP₃ to deplete stores sufficiently for I_CRAC to activate. We focused on mitochondria for three reasons. First, much recent evidence has established that they can and do take up significant amounts of Ca²⁺ following cytosolic Ca²⁺ increases under physiological conditions (Rizzuto et al. 1993, 1998; Jouavaille et al. 1995; Babcock et al. 1997; Hajnoczky et al. 1999; Tinel et al. 1999; Montero et al. 1999; Park et al. 2001; Arnaudeau et al. 2001) and this Ca²⁺ uptake, which occurs via a relatively low affinity ruthenium red-sensitive uniporter, is driven by the large inner mitochondrial membrane potential (Duchen, 2000). Second, mitochondria are often physically very close to the endoplasmic reticulum (Rizzuto et al. 1998; Park et al. 2001; Csordas & Hajnoczky, 2001), an arrangement that would greatly facilitate interaction between the two Ca²⁺-handling organelles. Third, mitochondria can take up Ca²⁺ that has entered via the store-operated pathway (Lawrie et al. 1996; Hoth et al. 1997; Hartmann & Verkhratsky, 1999).

To examine the effects of mitochondria on the ability of InsP₃ to activate I_CRAC in weak Ca²⁺ buffer, we dialysed cells with a pipette solution containing weak buffer and no thapsigargin but supplemented with a cocktail (malate, pyruvate, NaH₂PO₄, cAMP and GTP) known to be important for sustaining respiring mitochondria in the whole cell recording configuration (Gunter & Pfeiffer, 1990; Herrington et al. 1996). The results were dramatic (Fig. 11). InsP₃ in the absence of cocktail usually failed to activate any detectable I_CRAC (Fig. 11A–C) and, in the small fraction of cells that responded (Fig. 11D), the current was very small. Inclusion of the cocktail had striking effects (Gilabert & Parekh, 2000). A robust I_CRAC could now be activated (Fig. 11A–C) and the vast majority of cells responded (Fig. 11D), although the size of the current was smaller than that seen in strong buffer. The effects of the cocktail were prevented by pre-treatment with the combination of antimycin A and oligomycin or by including ruthenium red in the pipette together with InsP₃ and cocktail (Fig. 11C). Oligomycin alone was without effect (Glitsch et al. 2002b). Hence mitochondrial Ca²⁺ uptake is required for InsP₃ to activate I_CRAC under physiological conditions.

Figure 11. InsP₃ activates I_CRAC in weak buffer provided mitochondria are energised.

A, time course of activation of I_CRAC in the presence of the mitochondrial cocktail (○). I_CRAC does not activate in weak Ca²⁺ buffer in response to InsP₃ alone (•). B, steady-state I-V relationships are shown for the cells in A. C, the effects of cocktail are suppressed by preventing mitochondrial Ca²⁺ uptake by either depolarising mitochondria (pre-treatment with antimycin A (5 μg ml⁻¹) and oligomycin (0.5 μg ml⁻¹)) or inhibiting the uniporter with ruthenium red. D, the presence of cocktail dramatically increases the fraction of cells that respond to InsP₃ in weak Ca²⁺ buffer. ^*P < 0.01.

The effects of the mitochondrial cocktail solution were Ca²⁺-dependent, because no enhancing effect was seen in strong Ca²⁺ buffer (Fig. 12A). Strong Ca²⁺ buffer prevents a rise in intracellular Ca²⁺ by chelating Ca²⁺ both released from the stores and entering the cell, and hence a requirement for mitochondrial Ca²⁺ uptake would be obviated. Importantly, neither antimycin A and oligomycin nor ruthenium red impaired the ability of I_CRAC to activate in strong buffer, arguing against a direct channel blocking action of these drugs (Glitsch et al. 2002b).

Figure 12. Physiological antagonism between SERCA pumps and mitochondria.

A, the potentiation of I_CRAC by cocktail is suppressed by strong intracellular Ca²⁺ buffer (10 mm EGTA), indicating that the effect is Ca²⁺-dependent. B, the extent of I_CRAC to InsP₃ in weak buffer alone, in mitochondrial cocktail and cocktail plus thapsigargin is compared. ^*P < 0.01.

SERCA pumps compete with mitochondria for removal of cytosolic Ca²⁺

In the presence of energised mitochondria, the extent of I_CRAC following dialysis with InsP₃ is only around 35 % the size of that seen in strong buffer. This smaller current in weak buffer could reflect some Ca²⁺-dependent store refilling, which would imply that not all the Ca²⁺ that has been released from the stores has been taken up by mitochondria but instead that some has been resequestrated into the stores by the SERCA pumps. To examine this, we measured the size of I_CRAC evoked by dialysis with InsP₃ in weak buffer in the absence and presence of thapsigargin in the cocktail (Gilabert et al. 2001). Results are summarised in Fig. 12B. Inclusion of thapsigargin resulted in an almost threefold increase in the size of the current, and the amplitude now was similar to that seen in strong buffer. No such potentiating effects were seen when these experiments were repeated in strong buffer (Fig. 12A). The results are consistent with the notion that mitochondria and SERCA pumps compete for removing Ca²⁺ and this form of physiological antagonism between two major Ca²⁺ clearance mechanisms, which is particularly acute under physiological conditions of weak intracellular Ca²⁺ buffering, impacts upon the extent of store depletion and subsequent activation of I_CRAC. In RBL cells, there is good morphological evidence documenting a very close apposition between mitochondria and SERCA pumps on the endoplasmic reticulum (Csordas & Hajnoczky, 2001), which would greatly promote competition for Ca²⁺ removal between these two organelles.

Respiring mitochondria reduce the threshold for activation of I_CRAC by InsP₃

The previous experiments have all involved supra-maximal concentrations of InsP₃. To investigate whether mitochondrial Ca²⁺ uptake potentiates I_CRAC to more moderate levels of InsP₃, we compared the size of I_CRAC over a range of InsP₃ concentrations in the presence and absence of mitochondrial cocktail (Gilabert et al. 2001). The results are summarised in Fig. 13. Whereas 5 μm InsP₃ generally failed to activate I_CRAC in weak Ca²⁺ buffer (Fig. 13A and B), it evoked a sizeable current in the presence of the cocktail (Fig. 13A and B). Aggregate data are depicted in Fig. 13C. Comparing the results in the absence and presence of mitochondrial cocktail revealed several striking differences. First, the size of I_CRAC was potentiated in cells exposed to mitochondrial cocktail over a range of concentrations of InsP₃ (5–30 μm; Fig. 13C). Second, the fraction of cells that responded over this concentration range increased substantially in the presence of energised mitochondria (Fig. 13D). Third, whereas 3 μm InsP₃ consistently failed to evoke any current and is therefore a subthreshold concentration in the absence of the cocktail, around half the cells responded to this dose when mitochondria were energised (Fig. 13C and D). Mitochondrial Ca²⁺ uptake therefore increases the sensitivity of store-operated Ca²⁺ influx to InsP₃, and lowers the threshold concentration of InsP₃ that is required to activate I_CRAC. Mitochondrial Ca²⁺ buffering increases the dynamic range of concentrations over which InsP₃ is able to function as the physiological messenger that triggers activation of store-operated Ca²⁺ influx (Gilabert et al. 2001).

Figure 13. Energised mitochondria increase the ability of InsP₃ to activate I_CRAC in weak Ca²⁺ buffer.

A, whereas 5 μm InsP₃ is ineffective in weak buffer (○), it evokes a prominent I_CRAC in the presence of energised mitochondria (•). B, I-V relationships for the cells shown in A, taken at steady state. C, concentration-response curve to InsP₃ in the absence (○) and presence (•) of cocktail. D, the fraction of cells responding versus concentration of InsP₃ is plotted in the absence (○) and presence (•) of energised mitochondria.

Ca²⁺-dependent inactivation of CRAC channels is reduced by mitochondria

We have identified three mechanisms whereby a rise in cytosolic Ca²⁺ can inhibit CRAC channel activity. Ca²⁺-dependent fast inactivation occurs when permeating Ca²⁺ ions feed back to partially inactivate the channel through which they permeated (Fierro & Parekh, 1999a). Inactivation occurs with time constants of 10 and 100 ms, is largely independent of the macroscopic current (consistent with local feedback) and is more effectively reduced by BAPTA than EGTA, suggesting that the intracellular Ca²⁺ binding site is within a few nanometres of the pore (Fierro & Parekh, 1999a). This is very similar to fast inactivation first described in mast cells and jurkat T-cells (Hoth & Penner, 1993; Zweifach & Lewis, 1995a). Importantly, fast inactivation is unaffected by either energising mitochondria (Gilabert & Parekh, 2000) or depolarising them (Glitsch et al. 2002b), indicating that mitochondrial Ca²⁺ uptake does not seem to interfere with this form of Ca²⁺-dependent inactivation. Two slower Ca²⁺-dependent regulatory pathways also exist, operating over a time frame of several tens of seconds: Ca²⁺-dependent store refilling (Bakowski et al. 2001) and Ca²⁺ entry-dependent but store-independent inactivation (Parekh, 1998). These pathways can be separated by the use of thapsigargin, as was originally described in jurkat T cells (Zweifach & Lewis, 1995b). By inhibiting store refilling, thapsigargin eliminates the store-dependent component and enables the slow Ca²⁺-dependent inactivation mechanism to be studied in relative isolation, although the latter's contribution is probably exaggerated somewhat by thapsigargin because a major Ca²⁺ removal mechanism has been inhibited. Figure 14A shows an experiment in which a cell was dialysed with a pipette solution containing InsP₃ and thaspigargin in weak Ca²⁺ buffer. I_CRAC activated but then gradually decayed with a half-time of around 100 s, as Ca²⁺-dependent slow inactivation developed (• in Fig. 14A). Strikingly, if mitochondria were maintained in an energised state by including the cocktail, the extent of this slow inactivation was reduced significantly and the rate at which it developed was slowed (○ in Fig. 14A). Pooled data are summarised in Fig. 14B and C. Blocking mitochondrial Ca²⁺ uptake with ruthenium red prevented the cocktail from reducing the extent of slow inactivation (Fig. 14B and C). Hence mitochondrial Ca²⁺ uptake can prolong the time course of store-operated Ca²⁺ influx by reducing Ca²⁺-dependent slow inactivation (Gilabert & Parekh, 2000). A similar process has recently been reported in jurkat T cells (Hoth et al. 2000).

Figure 14. Mitochondrial Ca²⁺ uptake reduces the rate and extent of Ca²⁺-dependent slow inactivation.

A, dialysis with InsP₃ and thapsigargin in weak buffer (control) activates I_CRAC but then the current inactivates slowly (•). The extent of inactivation is reduced if mitochondria are energised (○). B and C summarise aggregate data comparing the extent (B) and rate (C) of inactivation of I_CRAC in control cells, in those in which cocktail was added to the solution and in cells exposed to cocktail and ruthenium red, to inhibit mitochondrial Ca²⁺ uptake. ^*P < 0.01. t_1/2, time at which I_CRAC had inactivated by 50 %.

Dual actions of mitochondria on I_CRAC

It is particularly striking that the same mechanism, Ca²⁺ uptake by mitochondria, should have such pronounced effects on both the activation and inactivation of I_CRAC. By facilitating the ability of InsP₃ to activate I_CRAC (by promoting more extensive store emptying) and, at the same time, reducing Ca²⁺-dependent slow inactivation, mitochondria will both increase the extent of I_CRAC as well as its duration to a much greater extent than through an action on activation or inactivation alone. A diagram summary depicting this dual action is shown in Fig. 15.

Figure 15. A diagrammatic scheme for the role of mitochondria in controlling I_CRAC under physiological conditions of weak intracellular Ca²⁺ buffering.

A, the resting state, where I_CRAC is not functioning, is shown. Stores are largely full and any Ca²⁺ that leaks from the stores is taken back up by the SERCA pumps. B, increasing InsP₃ levels in the absence of active mitochondrial Ca²⁺ uptake, releases Ca²⁺ from the stores. However, the SERCA pumps are able to resequester sufficient Ca²⁺ so only a very small fraction of CRAC channels are activated (undetectable in whole cell mode). Furthermore, the rise in cytosolic Ca²⁺ results in Ca²⁺-dependent slow inactivation of CRAC channels, and possibly InsP₃ receptors as well. C, in the presence of respiring mitochondria, InsP₃ activates macroscopic I_CRAC. Ca²⁺ released from the stores by InsP₃ is taken up by mitochondria. This reduces the amount of Ca²⁺ available to the SERCA pumps and in the vicinity of open InsP₃ receptors such that the stores are depleted sufficiently for macroscopic I_CRAC to activate (less refilling by SERCA pumps and less inactivation of InsP₃ receptors). Some refilling does occur because inclusion of thapsigargin enhances the size of the current. Furthermore, mitochondrial Ca²⁺ buffering reduces the rate and extent of Ca²⁺-dependent slow inactivation, thereby increasing the size and duration of the current. D, a simplified gating scheme for CRAC channels summarising the role of mitochondrial Ca²⁺ buffering. Mitochondria facilitate opening (Closed to Open transition) whilst, simultaneously, reducing inactivation (Open to Inactivated transition). In this way, mitochondria have a much larger impact on I_CRAC than through either transition alone.

Under physiological conditions therefore, mitochondria play a pivotal role in all stages of store-operated Ca²⁺ entry, determining whether the whole cell current develops or not, to what extent and for how long it stays operational. We are now trying to understand how mitochondrial Ca²⁺ buffering impacts upon physiologically relevant responses like exocytosis.