Mitochondrial Ca2+ uptake requires sustained Ca2+ release from the endoplasmic reticulum.

We analyzed the role of inositol 1,4,5-trisphosphate-induced Ca(2+) release from the endoplasmic reticulum (ER) (i) in powering mitochondrial Ca(2+) uptake and (ii) in maintaining a sustained elevation of cytosolic Ca(2+) concentration ([Ca(2+)](c)). For this purpose, we expressed in HeLa cells aequorin-based Ca(2+)-sensitive probes targeted to different intracellular compartments and studied the effect of two agonists: histamine, acting on endogenous H(1) receptors, and glutamate, acting on co-transfected metabotropic glutamate receptor (mGluR1a), which rapidly inactivates through protein kinase C-dependent phosphorylation and thus causes transient inositol 1,4,5-trisphosphate production. Glutamate induced a transient [Ca(2+)](c) rise and drop in ER luminal [Ca(2+)] ([Ca(2+)](er)), and then the ER refilled with [Ca(2+)](c) at resting values. With histamine, [Ca(2+)](c) after the initial peak stabilized at a sustained plateau, and [Ca(2+)](er) decreased to a low steady-state value. In mitochondria, histamine evoked a much larger mitochondrial Ca(2+) response than glutamate ( approximately 15 versus approximately 65 microm). Protein kinase C inhibition, partly relieving mGluR1a desensitization, reestablished both the [Ca(2+)](c) plateau and the sustained ER Ca(2+) release and markedly increased the mitochondrial Ca(2+) response. Conversely, mitochondrial Ca(2+) uptake evoked by histamine was drastically reduced by very transient ( approximately 2-s) agonist applications. These data indicate that efficient mitochondrial Ca(2+) uptake depends on the preservation of high Ca(2+) microdomains at the mouth of ER Ca(2+) release sites close to mitochondria. This in turn depends on continuous Ca(2+) release balanced by Ca(2+) reuptake into the ER and maintained by Ca(2+) influx from the extracellular space.

mitochondrial Ca 2ϩ uptake sites. The apparent discrepancy was reconciled by the concept that high [Ca 2ϩ ] microdomains generated at mouth of IP 3 Rs during its activation are sensed by neighboring mitochondria (17), which are thus exposed to [Ca 2ϩ ], that allow efficient Ca 2ϩ uptake. Finally, the diffusion of Ca 2ϩ through the outer mitochondrial membrane creates a lag time between the initial [Ca 2ϩ ] c and [Ca 2ϩ ] m rises into mitochondria (14,18,19). Thus, the properties of mitochondrial Ca 2ϩ accumulation suggest that these organelles may represent the prototype of a cytosolic effector, which requires sustained release of Ca 2ϩ from the ER (and thus maintenance of a high [Ca 2ϩ ] microdomain at ER/mitochondria contacts) to be actively recruited in the calcium signaling pathway.
To investigate the relationship between the kinetics of ER Ca 2ϩ release and mitochondrial Ca 2ϩ accumulation, we carried out a study in the epithelial cell line HeLa, utilizing organellespecific probes and agonists that induce different ER Ca 2ϩ release patterns. These cells endogenously express the H 1 Gprotein-coupled receptor coupled to phospholipase C activation and consequent continuous IP 3 production without detectable desensitization (20). Thus, histamine generates a typical biphasic cytosolic Ca 2ϩ signal with sustained [Ca 2ϩ ] c elevation and parallel emptying of the ER until the agonist is present (21). Activation of SOC following ER emptying has been also shown after histamine stimulation (22,23). In contrast to the H 1 receptor, the group I (Ca 2ϩ -mobilizing) metabotropic glutamate receptors, such as mGluR1 or mGluR5, undergo receptor desensitization in the continuous presence of glutamate (for a review, see Ref. 24). In several cell types expressing either endogenous or recombinant mGluRs, the desensitization of these receptors was shown to be mediated by PKC (25). Feedback inhibition of the receptor by PKC phosphorylation results in inhibition of phosphoinositide hydrolysis; thus, application of glutamate induces only a transient IP 3 production (26). In the present study, we have taken advantage of the different properties of H 1 and mGluR1a receptors regarding the kinetics of IP 3 production for analyzing the role of IP 3 -induced Ca 2ϩ release in generating a sustained cytosolic Ca 2ϩ response and its efficacy in inducing mitochondrial Ca 2ϩ uptake.
Aequorin Measurements-In the case of cytAEQ-or mtAEQmutexpressing cells, the coverslips with the cells were incubated with 5 M wild type coelenterazine for 2 h in Krebs-Ringer modified buffer (KRB; 135 mM NaCl, 5 mM KCl, 1 mM MgSO 4 , 0.4 mM K 2 HPO 4 , 1 mM CaCl 2 , 5.5 mM glucose, 20 mM HEPES, pH 7.4) at 37°C and then transferred to the perfusion chamber. In erAEQmut-expressing cells, the luminal [Ca 2ϩ ] of the ER was reduced during aequorin reconstitution by incubating the cells for 1 h at 4°C in KRB supplemented with 5 M coelenterazine n, the Ca 2ϩ ionophore ionomycin (5 M), and 600 M EGTA. After this incubation, cells were extensively washed with KRB without CaCl 2 , supplemented with 2% bovine serum albumin and 1 mM EGTA. The ER Ca 2ϩ store was refilled at the beginning of the experiments by perfusing the cells with KRB supplemented with 1 mM CaCl 2 .
All aequorin measurements were carried out in KRB containing either 1 mM CaCl 2 (KRB/Ca 2ϩ ) or 100 M EGTA (KRB/EGTA). Histamine (100 M) or glutamate (100 M) were added to the same medium. The aequorin experiments were terminated by lysing the cells with 100 M digitonin in a hypotonic Ca 2ϩ -rich solution (10 mM CaCl 2 in H 2 O), thus discharging the remaining aequorin pool. The light signal was collected in a purpose-built luminometer and calibrated into [Ca 2ϩ ] values as previously described (28). Chemicals and reagents were from Sigma or from Merck except for coelenterazine and coelenterazine n, which were from Molecular Probes, Inc. (Eugene, OR). Statistical data are presented as mean Ϯ S.E.
Imaging Measurement of Cytosolic [Ca 2ϩ ]-To monitor [Ca 2ϩ ] in the cytosol, mGluR1a-transfected HeLa cells were placed on 24-mm coverslips and loaded with 2 M fura-2/AM in KRB/Ca 2ϩ for 30 min at 37°C. Cells were then washed in the same solution, and [Ca 2ϩ ] c changes were determined using a high speed, wide field digital imaging microscope. A Zeiss Axiovert 200 inverted microscope was used with a ϫ40 objective. Fura-2 was excited at 340 and 380 nm using a random access monochromator (Photon Technology International). Images were acquired by a Micromax 1300YHS camera (Princeton Instruments) using 4ϫ binning in both the horizontal and vertical direction. Measurements were carried out at room temperature. Images were analyzed using the MetaFluor software (Universal Imaging Corp.).

RESULTS
Mitochondrial Ca 2ϩ Uptake Depends on the Duration of ER Ca 2ϩ Release-The start point of this work was to determine the correlation between the kinetics of the changes in cytosolic and mitochondrial Ca 2ϩ concentration ([Ca 2ϩ ] c and [Ca 2ϩ ] m , respectively) evoked in HeLa cells by the stimulation with a Ca 2ϩ -mobilizing, IP 3 -coupled agonist. For this purpose, HeLa cells were transfected with the appropriate targeted aequorin chimera (cytAEQ and mtAEQmut, respectively) (28). 36 -48 h after transfection, functional aequorin was reconstituted by adding the prosthetic group. The coverslip with the cells was then transferred to the luminometer chamber and challenged with the agonist histamine. Light emission was collected and calibrated into [Ca 2ϩ ] values, as described under "Experimental Procedures" and references therein. HeLa cells endogenously express the H 1 G-protein-coupled receptor, the stimulation of which leads to sustained IP 3 production. The results ( Fig. 1) showed that agonist stimulation (100 M histamine) evoked a cytosolic Ca 2ϩ peak (2.5 M), followed by a plateau phase (Fig. 1A). The cytosolic response was followed by an efficient mitochondrial Ca 2ϩ uptake, reaching a peak level of ϳ70 M (Fig. 1B). When the kinetics were compared, it was apparent that [Ca 2ϩ ] c peaked ϳ4 s after histamine addition (i.e. when the [Ca 2ϩ ] m rise was at Ͻ50% of the peak) and then rapidly declined toward a sustained plateau that was maintained throughout agonist stimulation. The [Ca 2ϩ ] m peak was reached when, through the activity of the SERCA and PMCA pumps, [Ca 2ϩ ] c was rapidly declining. As shown in Fig. 1, C and D, mitochondrial Ca 2ϩ uptake depended almost entirely on Ca 2ϩ release from internal stores, since stimulation of cells in Ca 2ϩ -free extracellular medium led to the same mitochondrial Ca 2ϩ uptake (64 Ϯ 2.7 M, n ϭ 5 versus 66 Ϯ 4.8 M, n ϭ 35 in control). We thus concluded that mitochondrial Ca 2ϩ uptake depends on sustained release of Ca 2ϩ from the ER.
To test this possibility, we stimulated HeLa cells through a G-protein-coupled receptor that rapidly undergoes PKCdependent inactivation and thus causes a transient production of IP 3 and Ca 2ϩ release from the ER. The receptor employed was mGluR1a (27), which was co-transfected with the targeted aequorin probes. Fig. 2 shows the functional properties of the transfected mGluR1a receptor. First, we analyzed the kinetics of Ca 2ϩ release from the ER evoked by histamine and glutamate in parallel batches of mGluR1a-expressing cells, cotransfected with erAEQmut (28) (Fig. 2A). The data obtained highlight the fundamental difference between the two agonist responses: (i) stimulation of the endogenous histamine receptor caused a rapid initial [Ca 2ϩ ] er drop followed by a continuous and slower decrease reaching, ϳ2 min after the start of agonist stimulation, a low [Ca 2ϩ ] er steady-state value of ϳ50 M, and (ii) in the case of glutamate, the initial rapid [Ca 2ϩ ] er decrease was followed by the refilling of the Ca 2ϩ store, which was almost complete in about 2 min even in the continuous presence of glutamate.
Next, we applied another approach to explore the relationship between the stores controlled by the two agonists. We perfused cells co-transfected with cytAEQ and mGluR1a with a Ca 2ϩ -free solution (KRB/EGTA; i.e. KRB supplemented with 100 M EGTA instead of 1 mM CaCl 2 ) in order to prevent refilling of the stores, and we applied repetitive histamine pulses in order to deplete its intracellular Ca 2ϩ store and then glutamate. It is apparent, as shown in Fig. 2C, that glutamate evoked substantially lower further release of Ca 2ϩ , compared with that observed when glutamate was applied as first stimulus (see Fig. 2B). The reverse protocol (i.e. applying histamine after depleting the glutamate releasable pool) showed similar results (Fig. 2B). These experiments demonstrated that most of the glutamate-and histamine-releasable pools are overlapping; thus, the different kinetic behavior of the Ca 2ϩ signal evoked by the two agonists cannot be ascribed to the use of separate intracellular stores.
We  (Fig. 3A). The difference in [Ca 2ϩ ] m response was even more dramatic; the peak rise evoked by glutamate stimulation was drastically reduced (16.5 Ϯ 1.8 M (n ϭ 28) versus 66 Ϯ 4.8 M (n ϭ 35) during the histamine challenge) (Fig. 3B). Moreover, the difference between the time to peak of [Ca 2ϩ ] c and [Ca 2ϩ ] m responses in this case was reduced (4.1 s for histamine versus 2.4 s for glutamate). In order to show that the same Ca 2ϩ pools were used by both agonists to feed mitochondrial Ca 2ϩ uptake, we applied histamine after 30 s of glutamate-induced Ca 2ϩ release, before refilling of the ER (see Fig. 2A). In this case, the histamine-induced mitochondrial Ca 2ϩ uptake was markedly reduced (23 Ϯ 4.5 M, n ϭ 5) (Fig. 4A). Apparently, further depletion of the ER Ca 2ϩ pool by histamine is responsible for the remaining [Ca 2ϩ ] m increase after glutamate stimulation as measured by erAEQmut (Fig. 4B).
The Kinetics of Initial Ca 2ϩ Release Induced by Glutamate and Histamine Are Identical-We thus investigated the various possible reasons for the drastic reduction of the [Ca 2ϩ ] m rise in the glutamate response. The first possibility is that the kinetics of Ca 2ϩ release from the ER is faster in histaminestimulated cells, an effect that could be overlooked by the aequorin measurements. Indeed, given that the high rate of mitochondrial Ca 2ϩ uptake depends on the exposure of Ca 2ϩ microdomains generated at the mouth of IP 3 Rs, one would envision that a faster release through the IP 3

FIG. 1. Cytosolic (A and C) and mitochondrial (B and D) [Ca 2؉ ] responses of HeLa cells stimulated with 100 M histamine. [Ca 2ϩ ] c and [Ca 2ϩ
] m was measured in cytAEQ-or mtAEQmutexpressing cells, respectively, 36 h after transfection. Aequorin luminescence was collected and calibrated into [Ca 2ϩ ] values as described under "Experimental Procedures." Where indicated, HeLa cells were stimulated with the agonist, added to the perfusion medium (KRB in experiments shown in A and B; Ca 2ϩ -free KRB plus 100 M EGTA in C and D). The dotted lines indicate the start of stimulation and peak of cytosolic and mitochondrial Ca 2ϩ signals, respectively. The traces are representative of Ͼ30 trials. 712 Ϯ 86 ms for glutamate; n ϭ 12). Thus, the differences in mitochondrial Ca 2ϩ uptake do not originate from differences in the velocity of the initial Ca 2ϩ release. This conclusion is compatible with the preceding observations in cell populations transfected with the cytosolic and mitochondrial targeted aequorin probes (see above), where the mitochondrial Ca 2ϩ uptake rate was slower than the cytosolic rise, and the mitochondrial uptake continued even after the cytosolic peak (see Fig. 1).
Alteration of the Duration of ER Ca 2ϩ Release Modifies Mitochondrial Ca 2ϩ Uptake-To further test the hypothesis that the [Ca 2ϩ ] m rise depends on a sustained release of Ca 2ϩ from the ER, we decided to modify the kinetics of the cytosolic responses induced by the two agonists. As to glutamate, we applied the PKC inhibitor staurosporine, which was shown to prevent mGluR1a phosphorylation and to reduce the consequent receptor desensitization, ensuring sustained IP 3 production during glutamate stimulation. Preincubation of the cells with 400 nM staurosporine reversed the kinetics of both the ER and cytosolic Ca 2ϩ signal; glutamate stimulation caused continuous Ca 2ϩ release from the ER (Fig. 6B), similarly to the Ca 2ϩ signal observed during histamine stimulation. As to [Ca 2ϩ ] c , no difference was observed in the peak, but the [Ca 2ϩ ] c decrease after the peak was significantly slower, and a sustained plateau (maintained throughout agonist stimulation) was reached (Fig. 6A). We have to note that the addition of staurosporine does not prevent mGluR1a desensitization completely (24), as indicated by the slower kinetics of continuous ER Ca 2ϩ release and the lower sustained [Ca 2ϩ ] c (see, for comparison, the sustained plateau following histamine stimulation) (Fig. 1A). Still, importantly, converting the glutamateinduced transient Ca 2ϩ release into a more continuous response by staurosporine application, the [Ca 2ϩ ] m peak was markedly increased (25.6 Ϯ 2.9 M, compared with 16.5 Ϯ 1.8 M cells not treated with staurosporine; n ϭ 28) (Fig. 6C).
The opposite experiment was also performed (i.e. the histamine-evoked release was made more transient) by reducing the duration of the agonist challenge to 2 s (Fig. 7). Under those conditions, the peak [Ca 2ϩ ] c response was marginally reduced (2.27 Ϯ 0.1 versus 2.45 Ϯ 0.1 M cells receiving a 2-min histamine challenge; n ϭ 16 for both groups) (Fig. 7A), but [Ca 2ϩ ] c rapidly returned to basal values, with disappearance of the sustained [Ca 2ϩ ] c plateau. Importantly, using the same protocol for [Ca 2ϩ ] m measurements, matching with the reduction of Ca 2ϩ release time, the [Ca 2ϩ ] m peak was substantially reduced (Fig. 7B, from 66 Ϯ 4.8 M, n ϭ 35, for 2-min stimulation to 29.1 Ϯ 2.9 M, n ϭ 16, for 2-s stimulation), and [Ca 2ϩ ] m reached its peak earlier (9.6 Ϯ 0.29 s control versus 4.5 Ϯ 0.35 s for 2 s stimulation) (Fig. 7B).
Continuous Ca 2ϩ Release from the Intracellular Ca 2ϩ Stores Is Balanced by SOC and Ca 2ϩ Recycling by SERCA-We thus concluded that prolonged release of Ca 2ϩ from the ER is necessary to achieve maximal mitochondrial responses. But how can this release and the ensuing microdomains at ER/mitochondria contacts be maintained, given that PMCA efficiently reduces [Ca 2ϩ ] c by extruding Ca 2ϩ in the extracellular space? To this end, it is necessary that ER be continuously refilled by Ca 2ϩ entry and redistribute Ca 2ϩ to the release sites, in keeping with the pathway of Ca 2ϩ studied in depth in pancreatic acinar cells (29,30). In other words, the steady-state phase of agonist-stimulated Ca 2ϩ release from the ER, which is necessary for transferring the Ca 2ϩ signal to mitochondria and other cytosolic effectors, must be sustained by the process of Ca 2ϩ entry through the plasma membrane.
Thus, in the next set of experiments, we analyzed the contribution of SOC to the generation of the sustained Ca 2ϩ signal. For this purpose, we released Ca 2ϩ from the intracellular pools by applying either histamine (Fig. 8A) or glutamate (Fig. 8B) in KRB/EGTA to cells transfected with cytAEQ and mGluR1a. Then we evoked Ca 2ϩ influx by changing the perfusion medium from KRB/EGTA to KRB/Ca 2ϩ (in the continuous presence of the agonist). Ca 2ϩ release from intracellular Ca 2ϩ stores produced a transient peak upon stimulation with both agonists, which returned to the base line, showing that the presence of extracellular Ca 2ϩ is essential to achieve sustained Ca 2ϩ signal. Furthermore, depletion of the stores by histamine activated SOC, as observed from the increase of [Ca 2ϩ ] c after the readdition of external Ca 2ϩ (see Fig. 8A). In contrast, Ca 2ϩ reintroduction into KRB caused only a slight elevation in the presence of glutamate, probably due to smaller depletion of the stores caused by the transient IP 3 production. However, glutamate, after staurosporine preincubation, was able to activate SOC, since the readdition of Ca 2ϩ caused a [Ca 2ϩ ] c elevation comparable with that of caused by histamine (see Fig. 8B). Thus, according to these data, we confirmed that both Ca 2ϩ influx and continuous IP 3 production are necessary for maintaining the cytosolic Ca 2ϩ signal.
Then we tested the effect of Cd 2ϩ , a blocker of Ca 2ϩ entry pathways of the plasma membrane, including store-operated channels, on [Ca 2ϩ ] er at different stages of cell stimulation (Fig.  9). As expected, if Cd 2ϩ is added after the stimulation with histamine (i.e. when no Ca 2ϩ release occurs and the ER is actively reaccumulating Ca 2ϩ ), the process of refilling is blocked, and a [Ca 2ϩ ] er steady state lower than in unstimulated cells is reached. Conversely, if Cd 2ϩ is added in the presence of the agonist (i.e. when the ER is largely depleted and a steady state [Ca 2ϩ ] er value of ϳ50 M is maintained), a rapid, further emptying of the ER is observed, leading to almost complete emptying of the ER. Thus, importantly, an equilibrium between refilling and Ca 2ϩ release through IP 3 Rs is maintained throughout the process of agonist stimulation.
If this is the case, one would expect that both blocking Ca 2ϩ release through IP 3 Rs or the process of ER refilling though store-operated channels should be equally effective in reducing the sustained [Ca 2ϩ ] c rise observed during agonist stimulation. Moreover, the termination of the [Ca 2ϩ ] c signal in the former way should be more rapid, whereas the effect of Ca 2ϩ entry blockade should occur only after the ER is depleted of the remaining Ca 2ϩ . This was directly investigated in the experiment of Fig. 10, where we compared the kinetics of terminating the sustained cytosolic signal after histamine stimulation with two experimental protocols. In the first, we applied Cd 2ϩ to block Ca 2ϩ influx still in the presence of histamine and IP 3induced Ca 2ϩ release. In the second, we washed out histamine rapidly in order to terminate IP 3 production (i.e. to stop Ca 2ϩ release). As shown in Fig. 9A, the half-decay time of the cytosolic Ca 2ϩ signal was about 2 times longer during blockade of influx by 1 mM Cd 2ϩ (t1 ⁄2 ϭ 15.4 Ϯ 0.98 s, n ϭ 12) than in the case of histamine washout (t1 ⁄2 ϭ 8.5 Ϯ 0.49 s, n ϭ 11, p Ͻ 0.01). These results show that Ca 2ϩ release from the internal stores has the major role in the generation of [Ca 2ϩ ] c elevation, whereas the Ca 2ϩ entry maintains the state of filling of the Ca 2ϩ store, thus counteracting the forces of Ca 2ϩ extrusion by PMCA.
Finally, we compared [Ca 2ϩ ] c changes during ER refilling in the presence and in the absence of IP 3 . As shown in Fig. 11A, after emptying the intracellular Ca 2ϩ pools by 100 M histamine in the absence of extracellular Ca 2ϩ , Ca 2ϩ readdition to the medium exerted a significant [Ca 2ϩ ] c increase only if histamine (i.e. IP 3 ) was present. If histamine was washed out 30 s before Ca 2ϩ readdition, refilling of the Ca 2ϩ stores was accompanied by only a small [Ca 2ϩ ] c elevation. The efficiency of refilling in this case was demonstrated by a second application of histamine, which elicited a Ca 2ϩ release comparable with the one exerted by the first stimulation (Fig. 11A, filled circles). Furthermore, prolonged histamine stimulation and IP 3 -induced release caused further depletion of the Ca 2ϩ stores even after Ca 2ϩ readdition and following robust Ca 2ϩ influx. This is illustrated by the small amount of releasable Ca 2ϩ remaining in the pools even after a 20 s washout of histamine before its reapplication (Fig. 11A, open circles). Similarly, the readdition of 1 mM extracellular Ca 2ϩ after depleting the Ca 2ϩ stores by the reversible SERCA inhibitor 2,5-di-(tert-butyl)-1,4-benzohydroquinone (tBHQ; 10 M) led to efficient store refilling without significant [Ca 2ϩ ] c elevation (Fig. 11B, filled circles). In contrast, in the continuous presence of tBHQ, Ca 2ϩ influx was conveyed to the cytosol, as shown by the vast [Ca 2ϩ ] c elevation following the readdition of Ca 2ϩ (Fig. 11B, open circles). These data clearly indicate that Ca 2ϩ , after entering the cells, is taken up robustly by the ER without significantly elevating [Ca 2ϩ ] c , and then Ca 2ϩ is distributed by the ER to the entire cytosolic space. In this way, SOC is not directly responsible for the generation of the sustained cytosolic Ca 2ϩ signal; nevertheless, it is necessary for its maintenance, by ensuring the continuous refilling of the internal Ca 2ϩ stores. Furthermore, Ca 2ϩ cycling at the vicinity of Ca 2ϩ release sites maintains an equilibrium on both sides of the ER membrane (i.e. the steady state in the [Ca 2ϩ ] er and the sustained phase of the [Ca 2ϩ ] c elevation); thus, ER plays an important role as a redis-tribution pathway between plasma membrane Ca 2ϩ entry and other intracellular organelles, such as the mitochondria. DISCUSSION In recent years, much information has been acquired on how specific spatio-temporal patterns of Ca 2ϩ signaling can control different cell functions (or deleterious effects in pathological conditions). In particular, attention has been drawn to the properties and functional significance of local gradients (microdomains) and thus to the importance of the source and intracellular route of the [Ca 2ϩ ] rise. Subplasmalemmal high [Ca 2ϩ ] microdomains appear to regulate the activity of plasma membrane (PM) ion channels (such as voltage-dependent Ca 2ϩ and Na ϩ channels, Ca 2ϩ -activated K ϩ channels, and SOC), polarity and excitability of the PM (neurons, smooth muscle), secretory responses, and neurotransmitter release. Ca 2ϩ entry, by means of voltage-, ligand-, or store-operated channels, in most of the cases is necessary and sufficient to generate such a high [Ca 2ϩ ] microdomain below the PM, also for longer periods of cell stimulation. Conversely, other long term processes regulated by sustained Ca 2ϩ signals take place deeper in the cell interior. Among them, a well known example is the regulation of mitochondrial enzymes involved in ATP production or steroid synthesis, where Ca 2ϩ taken up from microdomains generated at the mouth of ER Ca 2ϩ release channels plays a fundamental role (15,31). A transient [Ca 2ϩ ] m peak was shown to exert a long term effect at the level of ATP synthesis (32), and continuous mitochondrial [Ca 2ϩ ] elevation, even at relatively lower extramitochondrial Ca 2ϩ levels, was shown to increase the activity of dehydrogenases of the Krebs cycle (13,33,34), thus elevating the NADH level and the activity of the electron transport chain. At the same time, Ca 2ϩ uptake by mitochondria has been shown to be involved in a radically different process (i.e. the release of proapoptotic factors and thus the induction of cell death) (35).
Mitochondria have thus recently emerged as key decoders of calcium signals, and the mechanism and timing of their recruitment control key decisions in cell life and death. In this contribution, we have analyzed the correlation between the kinetics of [Ca 2ϩ ] c increase and its different components and the [Ca 2ϩ ] m rises occurring in agonist-stimulated cells. For this purpose, we utilized a low affinity probe for [Ca 2ϩ ] m (that allows us to fully appreciate the large [Ca 2ϩ ] m swing) and two different agonist stimulations, through the endogenous histamine receptor and through a transfected metabotropic glutamate receptor, which undergoes rapid desensitization and thus causes transient IP 3 production. The first observation is that, upon cell stimulation, [Ca 2ϩ ] m peaks well after [Ca 2ϩ ] c . As previously observed by various groups, there is a short delay in the upstroke, possibly due to the time needed for the diffusion of Ca 2ϩ released by IP 3 receptors through the outer mitochondrial membrane, thus reaching the transport systems (uniporter) of the inner membrane (14,18,19). Then [Ca 2ϩ ] m rises and reaches its maximal value after ϳ10 s (i.e. when the [Ca 2ϩ ] c increase, through the activity of the Ca 2ϩ pumps, is rapidly declining).
How can the slow kinetics of mitochondrial Ca 2ϩ accumulation be reconciled with the notion that the low affinity of the mitochondrial uniporter requires, for rapid uptake, the high [Ca 2ϩ ] gradient generated upon cell stimulation by the opening of IP 3 receptors? The most logical explanation is that, for maximal mitochondrial Ca 2ϩ uptake, prolonged Ca 2ϩ release from the ER must occur. Different experiments support this notion. Indeed, not only in the case of glutamate stimulation (in which Ca 2ϩ release from the ER is short lived) is the mitochondrial Ca 2ϩ response drastically reduced, but the effect of the two stimuli (glutamate and histamine) on mitochondria can be reversed by modifying the time course of the Ca 2ϩ release process. If desensitization of the mGluR is prevented by PKC inhibitors, ER release becomes sustained (with ensuing large amplitude emptying of the ER and activation of store-dependent Ca 2ϩ influx), and mitochondrial responses are greatly enhanced. Conversely, a short (2-s) histamine pulse causes transient emptying of the Ca 2ϩ store and drastically reduces the [Ca 2ϩ ] m rise evoked by the agonist. These data imply that in the late phases of agonist stimulation (i.e. when the activity of the pumps (SERCA and PMCA) counteracts the release of Ca 2ϩ through the IP 3 Rs), an equilibrium is attained between the two processes, as demonstrated by the direct measurement of [Ca 2ϩ ] er ; blocking Ca 2ϩ release (e.g. by terminating cell stimulation) causes the rapid refilling of the store, while conversely interrupting the reaccumulation of Ca 2ϩ in the ER (e.g. by blocking Ca 2ϩ entry through SOC; see below) allows IP 3 Rs to rapidly deplete the ER of Ca 2ϩ .
Mitochondria, and possibly other cytosolic effectors, appear thus to be activated through the kinetic behavior of the ER release process. Conversely, Ca 2ϩ entry, which is invariably essential for sustained Ca 2ϩ signals, might not provide a direct supply for these localized Ca 2ϩ regulatory events. Two considerations support this view. First, these processes rely on specific patterns of IP 3 -dependent release of Ca 2ϩ from the ER store, as discussed above. Second, Ca 2ϩ entering the cytosol is strongly buffered by Ca 2ϩ -binding proteins, such as parvalbumin, calbindin D 28K , and calretinin, rendering the diffusion rate of Ca 2ϩ in the cytosol rather low (36). Thus, Ca 2ϩ coming from the extracellular medium would reach slowly if at all the deeper regions of the cytosol. An ingenious solution for this challenge has been recently shown in a polarized cell model, the pancreatic acinar cells (30). At the initial phase of the physiological activation of these cells, focal Ca 2ϩ release occurs exclusively at the secretory pole, serving as trigger for exocytosis and leading to local emptying of the Ca 2ϩ store (30). The presence of Ca 2ϩ signaling components and particularly TRPlike channels at the vicinity of the apical pole has been suggested to provide a straightforward route for local refilling of the depleted stores (37). On the other hand, given that the ER in these cells forms a continuous network (4), the resulting luminal [Ca 2ϩ ] gradient has been shown to cause rapid diffusion of Ca 2ϩ from the basolateral part of the cells, in a model where SOC is restricted to this area (38). The ground of this arrangement of Ca 2ϩ signaling is the relatively high Ca 2ϩ mobility in the ER tunnel compared with the cytosol, depending on the much lower binding capacity of the ER lumen (ϳ20 versus ϳ2000 bound/free Ca 2ϩ of the cytosol of mouse pancreatic acinar cells (39)). Similarly, in neurons, which is still a highly polarized cell type, it was proposed that subplasmalemmal ER cisternae of the cell body may be responsible for Ca 2ϩ refilling from the extracellular space, and a continuous ER network would transport Ca 2ϩ to the site of release in dendritic spines (40). In accordance with this idea, it has been shown that the cytosolic buffering capacity of Purkinje neurons is as high (ϳ2000) as that of pancreatic acinar cells (41). On the other hand, in other cell types such as chromaffin cells (42), lower values of binding capacity have been found (ϳ40), but we should emphasize that even if Ca 2ϩ diffusion may occur at a similar velocity in the cytosol and the ER, another important advantage in the use of ER for distributing Ca 2ϩ signals to the cell interior is that it may ensure localized Ca 2ϩ release. Thus, in smooth muscle cells, even if the cytosolic buffering capacity is comparable with that of the ER (ϳ30 -40) (43), a superficial layer of SR rapidly buffers Ca 2ϩ entering from the extracellular space, which is then distributed into the cell interior, causing contraction after its directed release (44).
Since our aim was to disclose the role of the ER in generating a sustained Ca 2ϩ signal and in recruiting cytosolic effectors, in our work we did not characterize the exact nature of the Ca 2ϩ entry pathway. However, some information on Ca 2ϩ influx and store refilling can be obtained from the experimental data. Recently, a receptor-activated, arachidonic acid-mediated, noncapacitative mechanism for Ca 2ϩ entry has been demonstrated that appears to operate in a PM domain distinct from that in which SOC operates in HEK293 cells (45). Two arguments suggest that this pathway does not contribute to the Ca 2ϩ signals observed in our experiments. (i) We used maximal agonist concentrations for prolonged periods, producing a substantial depletion of Ca 2ϩ stores; thus, we fully activated the SOC mechanism, which inhibits arachidonate-regulated channels (46). (ii) Ca 2ϩ entry activated by store depletion in our system was clearly necessary for store refilling, in contrast to Ca 2ϩ entering the cells by arachidonate-activated Ca 2ϩ entry, which rather plays a role in potentiating the Ca 2ϩ release induced by IP 3 , thus increasing the frequency of oscillations (47).
The other issue concerns the way Ca 2ϩ refills the Ca 2ϩ stores. We demonstrated that in the case of transient IP 3 production (i.e. during glutamate stimulation), ER refilling occurs without detectable rise of [Ca 2ϩ ] c (compare Figs. 2A and 3B). However, based on our data, we cannot distinguish between refilling from the extracellular space through SOC channels and direct Ca 2ϩ reuptake from the cytosol by SERCA. It appears that the buffering capacity of the cytosol determines the route by which Ca 2ϩ released from the ER is eliminated, since increased buffering allows ER refilling by SERCA even in the absence of extracellular Ca 2ϩ (48). Thus, in cells with inherent high cytosolic buffer capacity, such as the above mentioned pancreatic acinar cells or neurons, the SERCA pumps appear to dominate over PMCAs in rapidly reducing the [Ca 2ϩ ] c peak (49,50). Thus, the oscillatory Ca 2ϩ signals evoked by colecystokinin in these cells were shown to occur in the absence of Ca 2ϩ entry, given that Ca 2ϩ is taken back almost entirely from the cytosol after the Ca 2ϩ spikes (see above). However, the situation differs with other types of stimulation; the Ca 2ϩ oscillations evoked by carbachol strongly depend on Ca 2ϩ influx in the proximity of the Ca 2ϩ release sites of the apical pole (37). Up to now, there are no data concerning the cytosolic buffering capacity of HeLa cells, but evidence from the similar Chinese hamster ovary cell line shows that PMCA overexpression leads to larger reduction and faster termination of cytosolic Ca 2ϩ signal compared with SERCA (51). Thus, it seems likely that the glutamate-induced Ca 2ϩ transient is rapidly extruded from the cell, and the Ca 2ϩ source of refilling in this case is the extracellular space.
In conclusion, the data presented in this paper indicate that mitochondria, important transducers of the Ca 2ϩ signal, depend on the process of ER Ca 2ϩ release, which in turn is sustained by continuous release through IP 3 receptors and refilling by SERCAs (with a primary role of Ca 2ϩ influx in counteracting the extrusion of Ca 2ϩ by PMCAs). Altogether, these data suggest that the ER provides a fast route for tunneling and releasing Ca 2ϩ in the deeper portions of the cytoplasm (where mitochondria are only one of the numerous Ca 2ϩ effectors) not only in the polarized pancreatic acinar cell (as proposed by Petersen et al. (30)) but also in different cell types. This provides an additional mechanism by which the selective placement and differential activation of Ca 2ϩ channels in the ER and plasma membrane provide flexibility to the Ca 2ϩ trans-duction system, allowing this second messenger to play a key role in the modulation of virtually all cellular processes.