Modulation of Mitochondrial Ca2+ Homeostasis by Bcl-2*

We have investigated the role of mitochondrial Ca2+ (Ca m ) homeostasis in cell survival. Disruption of Ca m homeostasis via depletion of the mitochondrial Ca2+ store was the earliest event that occurred during staurosporine-induced apoptosis in neuroblastoma cells (SH-SY5Y). The decrease of Ca m preceded activation of the caspase cascade and DNA fragmentation. Overexpression of the anti-apoptosis protein Bcl-2 led to increased Ca m load, increased mitochondrial membrane potential (ΔΨ m ), and inhibition of staurosporine-induced apoptosis. On the other hand, ectopic expression of the pro-apoptotic protein Bik led to decreased Ca m load and decreased ΔΨ m . Inhibition of calcium uptake into mitochondria by ruthenium red induced a dose-dependent apoptosis as determined by nuclear staining and DNA ladder assay. Similarly, reducing the Ca m load by lowering the extracellular calcium concentration also led to apoptosis. We suggest that the anti-apoptotic effect of Bcl-2 is related to its ability to maintain a threshold level of Ca m and ΔΨ m while the pro-apoptotic protein Bik has the opposite effect. Furthermore, both ER and mitochondrial Ca2+ stores are important, and the depletion of either one will result in apoptosis. Thus, our results, for the first time, provide evidence that the maintenance of Ca m homeostasis is essential for cell survival.

Cellular Ca 2ϩ signaling is a complex process involving coordinated action of various subcellular compartments including endoplasmic reticulum (ER), 1 mitochondria, nucleus as well as plasma membrane (1,2). While earlier Ca 2ϩ mobilization studies have emphasized the interaction between the plasma membrane and the ER Ca 2ϩ store, recent work from several laboratories has clearly indicated the active participation of mitochondria in this intracellular Ca 2ϩ network (3)(4)(5)(6)(7)(8)(9). The availability of novel optical indicators for Ca 2ϩ ions has now enabled studies of mitochondrial Ca 2ϩ signaling in living cells. It has been demonstrated that, during Ca 2ϩ signaling events, functionally competent mitochondria can rapidly sequester cytoplasmic Ca 2ϩ (Ca c ) through a uniporter and subsequently release Ca 2ϩ slowly via the mitochondrial Na ϩ /Ca 2ϩ exchanger (10). Dynamic studies of adrenal chromaffin cells (6) and T lymphocytes (7) indicate that mitochondrial Ca 2ϩ uptake limits the rise of cytosolic Ca 2ϩ and underlies the rapid decay of Ca c signal. Furthermore, suppressing export of Ca 2ϩ by inhibition of the mitochondrial Na ϩ /Ca 2ϩ exchanger hastens final recovery of Ca c (6,7). It is suggested that mitochondria are essential for both the generation of Ca 2ϩ signals, and the modulation of store-operated or "capacitative" Ca 2ϩ entry.
Studies in excitable and nonexcitable cells have suggested several important consequences of Ca 2ϩ uptake by mitochondria. First, increases in mitochondrial Ca 2ϩ concentration (Ca m ) are believed to modulate the production of ATP. Several dehydrogenases that supply substrates for the electron transport chain, are stimulated by Ca 2ϩ (11)(12)(13). Recent in vivo studies have also verified that mitochondrial uptake of Ca 2ϩ is accompanied by increased levels of NADH or NADPH (5,14) and activation of the matrix enzyme pyruvate dehydrogenase (15). This behavior is thought to couple the rate of ATP production to demand such that, during periods of high cytoplasmic Ca 2ϩ concentration ([Ca c ]), the increased rate of ATP hydrolysis by Ca 2ϩ pumps is balanced by enhanced ATP generation. In addition, mitochondrial uptake may help to maintain normal calcium homeostasis and protect cells from apoptotic insult.
The anti-apoptosis protein Bcl-2 is localized to the mitochondrial membrane as well as ER and nuclear membranes (16). The co-localization of Bcl-2 with Ca 2ϩ pumps and channels on ER and nuclear membrane has raised the possibility for a role of Bcl-2 in the maintenance of Ca 2ϩ homeostasis in these compartments. Previously, several groups including ours have reported modulation of ER (17,18) or nuclear Ca 2ϩ stores by Bcl-2 (19). However, the role of Bcl-2 on mitochondrial Ca 2ϩ homeostasis has been unclear. In the present study, we have examined the effect of Bcl-2 overexpression on mitochondrial Ca 2ϩ load and the relationship between mitochondrial Ca 2ϩ load and the maintenance of mitochondrial membrane potential (⌬⌿ m ). We find that Bcl-2 enhances mitochondrial Ca 2ϩ load and enables cells to maintain a stable ⌬⌿ m to offset the insult by pro-apoptotic agents such as Bik or staurosporine which promote the decrease of ⌬⌿ m and cause depletion of the mitochondrial Ca 2ϩ store.

MATERIALS AND METHODS
Cells and Viruses-SH-SY5Y human neuroblastoma cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, 2 mM L-glutamine, and 25 g/ml gentamycin at 37°C. Recombinant adenovirus type 5 expressing human Bcl-2 and Escherichia coli LacZ have been described (20). Construction of recombinant adenovrius expressing human Bik tagged with HA epitope will be described elsewhere. To overexpress Bcl-2, LacZ, or Bik, cells were infected with the appropriate virus at 100 plaque-forming units/cell in DMEM with 5% serum. After 6 h of infection, the medium was replaced with DMEM growth medium containing 10% serum. Cells were used 24 h later after virus infection. The expression of the Bcl-2 or HA-Bik was verified by Western blot analysis.
Western Analysis-For detecting Bcl-2 or Bik protein expression, 70 -80% confluent cells were lysed and protein samples were separated on 12% polyacrylamide gel for Western blotting as described previously (18). The Bcl-2 antibody was obtained from Dako. A monoclonal antibody 12CA5 that recognizes HA epitope was used for Bik detection.
Measurement of Cytosolic Ca 2ϩ -Cytosolic Ca 2ϩ concentration was measured using the Photon Technology International instrument and the fluorescent dye Fura-2 as described previously (21). SH-SY5Y cells grown on coverslip (approximately 5 ϫ 10 4 cells/coverslip) were loaded with Fura 2/AM (4 M) in loading buffer A containing (in mM) 5.4 KCl, 137 NaCl, 0.44 KH 2 PO 4 , 4.2 NaHCO 3 , 0.34 Na 2 HPO 4 , 1 MgCl 2 , 5 Hepes (pH 7.4), 11.1 D-glucose, 2 CaCl 2 , 0.1% bovine serum albumin. Fura 2 fluorescence was recorded at excitation wavelengths of 340 and 380 nm and an emission wavelength of 500 nm. To begin the fluorescence measurement, cells were rinsed (after 30 min of dye loading) in loading buffer A without Ca 2ϩ . After establishing the base line, cells were then treated with either ionomycin (10 M) plus 0.1 mM EGTA or thapsigargin (1 M) plus 0.1 mM EGTA. A standard curve was used to convert the fluorescence ratio to Ca 2ϩ concentration.
Measurement of Mitochondrial Ca 2ϩ -Rhod2-AM was used to measure mitochondrial Ca 2ϩ according to the procedure of Hajnoczky et al. (14). Rhod2-AM has a net positive charge, which facilitates its sequestration into mitochondria due to membrane potential-driven uptake. The use of dihydro-rhod2-AM enhances the selectivity for mitochondrial loading because this dye exhibits Ca 2ϩ -dependent fluorescence only after it is oxidized and this occurs preferentially within mitochondria. Cells were loaded with 4 M dihydro-rhod2-AM for 60 min. The residual cytosolic fraction of the dye was eliminated when the cells were kept in culture for an additional 18 h after loading, whereas the mitochondrial dye fluorescence was maintained. Cellular fluorescence was measured by the Photon Technology International system with excitation at 550 nm and emission at 580 nm or by image acquisition using the Meridian ACAS laser confocal microscope. The fluorescence units (counts/s) or the average pixel intensity/cell was determined. Rhod2 fluorescence was not calibrated in terms of [Ca 2ϩ ], since it is not a radiometric dye and is localized in only a small compartment within the cell (14).
Measurement of Mitochondrial Membrane Potential-The dye tetramethylrhodamine methyl ester (TMRM) was used to measure mitochondrial membrane potential (⌬⌿ m ) according to Ichas et al. (22). Cells infected with recombinant adenovirus (for 24 h) were loaded with 0.2 M TMRM for 20 min in culture medium. Cells were then rinsed and transferred to a chamber on the microscope stage in the same loading buffer A (containing glucose and no Ca 2ϩ ) as the Fura-2 experiments. Images were collected at every 10 s using Meridian ACAS laser confocal microscope with excitation at 514 nm and emission at 575 nm, and the average pixel intensity/cell was determined.
Apoptotic Nuclei-Cells were fixed in 4% paraformaldehyde and stained with the DNA-binding dye Hoechst 33258 according to Jacobson and Raff (23). Apoptotic nuclei were visualized and counted under epifluorescence illumination (340 nm excitation and 510 nm barrier filter) using a 40ϫ oil immersion objective (more than 200 cells/culture were counted, and counts were made in at least four to six separate cultures).
DNA Ladder Assay-Cells were seeded at a density of 4 ϫ 10 6 cells/plate in 100-mm plate and grown for 24 h in medium containing 10% serum. Cells were then treated with various concentrations of ruthenium red in medium without serum for 24 h. Cells were then scraped and pelleted. The cell pellet was lysed in buffer containing 0.5 mg/ml proteinase K. DNA was extracted and analyzed on 1% agarose gel as described by Loo and Rillema (24).
Caspase Activity Assay-Cells were lysed in buffer containing 50 mM Hepes, pH 7.4, 0.1% CHAPS, 1 mM dithiothreitol, 0.1 mM EDTA, and 1% Trition x-100. After centrifugation, the supernatant was used for protein determination and caspase assay (Biomol kit). Caspase 3-like activity was determined by continuously monitoring the proteolysis of the substrate N-acetyl-Asp-Glu-Val-Asp-p-nitroanilide at 405 nm in a microtiter plate (Bio-Rad). Data were collected at 30-s intervals for the entire assay period up to 30 min. An inhibitor N-acetyl-Asp-Glu-Val-Asp-aldehyde was used to measure the nonspecific activity. The specific activity was obtained by subtracting the nonspecific activity from the total activity as described by the protocol from the Biomol kit.
Statistical Analysis-Values are mean Ϯ S.D. Significance was de-termined by Student's t test. A value of p Ͻ 0.05 was considered to be significant.

Increased Capacity of Internal Ca 2ϩ
Store by Bcl-2-We have examined the effect of Bcl-2 overexpression on the Ca 2ϩ load of the internal store in the neuroblastoma cells (SH-SY5Y). For this purpose, cells infected with recombinant adenovirus expressing Bcl-2 (Ad-Bcl2) were compared with cells infected with a recombinant adenovirus expressing E. coli LacZ (Ad-LacZ). Fig. 1A shows the Western analysis of Bcl-2 protein in the cell lysate prepared 24 h after infection. Results from three experiments indicated that there was a low level of endogenous Bcl-2 present in control cells infected with Ad-LacZ, while the level of Bcl-2 was enhanced by more than 2-fold in cells infected with Ad-Bcl2. To estimate the loading state of the internal stores, Bcl-2 overexpressing cells and control cells were treated with ionomycin, a calcium ionophore that releases Ca 2ϩ from ER store as well as mitochondrial store. The addition of ionomycin (10 M) resulted in the release of Ca 2ϩ into the cytosol, where it was measured by Fura-2 as an increase in intracellular Ca 2ϩ (21). To block Ca 2ϩ influx due to the release of internal Ca 2ϩ store, i.e. the capacitative Ca 2ϩ entry (25), these experiments were performed in the absence of extracellular Ca 2ϩ (see "Materials and Methods"). Under this condition, the increase of cytoplasmic Ca 2ϩ (Ca c ) triggered by ionomycin is largely due to release from the ER and the mitochondrial stores. After reaching a peak within 60 s, the Ca c started to decline to near basal level. Fig. 1B shows that the resting level of Ca c was similar for Bcl-2 expressing cells and control cells (ranging from 66 Ϯ 2.4 nM for control, and 70 Ϯ 6.5 nM for Bcl-2 cells, n ϭ 8). However, the ionomycin-induced release of Ca 2ϩ was significantly higher in Bcl-2 expressing cells (151 Ϯ 22 nM) as compared with controls (87 Ϯ 9.7, p Ͻ 0.001, n ϭ 5). The results suggest increased loading of the internal store in the Bcl-2 overexpressing cells. Since the ionomycin-sensitive stores encompass both the ER and the mitochondria compartments, we then examined the effect of Bcl-2 on the Ca 2ϩ load of the ER store, which is sensitive to thapsigargin (Tg, 1 M), a specific inhibitor of the ER calcium ATPase. Fig. 1C shows that Tg induced a larger increase of Ca c in Bcl-2 overexpressing cells than controls. Results from five independent experiments indicated that the Tg-releasable ER Ca 2ϩ was 90 Ϯ 7.8 nM for Bcl-2 and 73 Ϯ 4.6 nM for controls (p Ͻ 0.01). These results are consistent with our previous report of enhanced ER Ca 2ϩ load in the breast epithelial cells by Bcl-2 overexpression (18).
Increased Loading of the Mitochondrial Calcium Store by Bcl-2-To further test if the mitochondria Ca 2ϩ store is also increased by Bcl-2, we evaluated the mitochondrial store in SH-SY5Y cells infected with Ad-Bcl2 or Ad-LacZ by using rhod2 as the specific probe for mitochondria Ca 2ϩ (see "Materials and Methods"). Fig. 2A shows the time-resolved measurement of mitochondrial Ca 2ϩ (Ca m ). The concentration of Ca m was represented in fluorescence units. Initially, at the resting state, in the absence of external Ca 2ϩ , Ca m was at a basal level. Addition of FCCP (4 M, see arrow), a protonophore led to collapse of the mitochondria membrane potential (⌬⌿ m ), and release of Ca m as shown by a rapid reduction in rhod2 fluorescence. Since de-energized mitochondria no longer take up Ca 2ϩ , the magnitude of the fluorescence reduction was used to estimate the Ca 2ϩ load in the mitochondrial store. Fig. 2A shows that the addition of 4 M FCCP caused larger release of Ca m in Bcl-2 cells than control cells. Calculation of the percentage of release (defined as the difference between the initial fluorescence and the final fluorescence that occurred within 100 s after FCCP addition, divided by the initial fluorescence) indicated 6.7% for Bcl-2 cells and 2.4% for controls. Results from four independent experiments indicated that the percentage of release (representing Ca m load) was approximately 3 times higher in the Bcl-2 group as compared with the control group (Bcl-2/control ϭ 3.4 Ϯ 1.0, n ϭ 4). In addition, comparison of basal fluorescence intensity under the same loading condition (i.e. same day and same cell density), also indicated an average value of 9245 Ϯ 600 counts/s for Bcl-2 group versus 7875 Ϯ 475 for the control group (two independent experiments). The results are consistent with an enhanced load of Ca m in the Bcl-2-overexpressing cells. This conclusion is further supported by the Fura-2 experiment that monitors the change in cytoplasmic Ca 2ϩ concentration (Ca c ). Fig. 2B shows that addition of 4 M FCCP led to a gradual and sustained increase of Ca c as a consequence of mitochondrial release. FCCP triggered a larger increase of Ca c in Bcl-2 cells than controls.
The Dependence of Ca m Release on FCCP Concentration-We have determined the optimal concentration of FCCP required for the maximal release of Ca m for Bcl-2 cells and control cells. Decreased Loading of the Mitochondrial Ca 2ϩ Store by Proapoptotic Protein Bik-In order to understand the functional significance of mitochondrial Ca 2ϩ loading, we proceeded to examine the status of the Ca m in cells that overexpress Bik, a pro-apoptotic protein that has antagonistic function to Bcl-2 (29). Neuroblastoma cells were infected with the adenovirus vector expressing human Bik insert (see "Materials and Methods"). After 24 h, the cells were lysed and analyzed for Bik expression by Western blot. The results indicated that there was no HA-Bik expression in the control cells (infected with Ad-LacZ) while HA-Bik expression was prominent in cells infected with the Ad-Bik (Fig. 4A). The status of the Ca m load was examined in a laser cytometer (Meridian Instrument), where changes of fluorescence image were followed at the single cell level for Bcl-2, control and Bik groups (Fig. 4B). Comparison of basal fluorescence indicated that the average pixel intensity per cell was 822 Ϯ 151 (n ϭ 13) for Bcl-2, 580 Ϯ 110 (n ϭ 5) for control, and 413 Ϯ 100 (n ϭ 10) for Bik-expressing cells. More quantitative evaluation of the Ca m load by FCCP-sensitive release (Fig. 4B) also indicated that the addition of 20 M FCCP caused 14.6 Ϯ 4.5% (n ϭ 8) of Ca m release in Bcl-2, 9.0 Ϯ 1.7% (n ϭ 6) for control, and 5.1 Ϯ 4% (n ϭ 9) for Bik cells. Thus the Ca m load was highest in the Bcl-2 group, intermediate in the control group, and lowest in the Bik group. Interestingly, this study at the single cell level also revealed that Bcl-2 cells were more resistant to 20 M FCCP treatment than Bik-expressing cells (compare Bcl-2 panel with the Bik panel in Fig. 4B). While the release of Ca m by FCCP was completed in 30 s for the Bik-expressing cells, it required almost 300 s to deplete the Ca m in the Bcl-2 cells.
Association of Mitochondrial Calcium Load with Membrane Potential-We also measured mitochondrial membrane potential (⌬⌿ m ) in cells that overexpress Bcl-2 or Bik. A cationic fluorophore, TMRM, was used as an indicator of mitochondrial membrane potential ⌬⌿ m (30,31). After loading (20 min) and removal of excess TMRM, the fluorescence in the cells were compared at t ϭ 0. There was stronger TMRM fluorescence (warmer pseudocolored image) in a typical Bcl-2-expressing cell than a typical Bik-expressing cell (data not shown), suggesting higher membrane potential in the former than in the latter. Furthermore, the fluorescence was stable in the Bcl-2-expressing cell during the measurement when image scan was continuously monitored at every 10-s interval for 3 min. In contrast, the fluorescence of Bik-expressing cells was continuously declining indicating the loss of ⌬⌿ m . In fact, it required the addition of oligomycin, a mitochondrial ATPase inhibitor, to prevent the hydrolysis of cellular ATP and thereby stabilize the ⌬⌿ m temporarily. Hence, the measurement of TMRM fluorescence in the presence of oligomycin represented an overestimate of ⌬⌿ m for the Bik cells. Results from three independent experiments indicated that the Bcl-2 cells had TMRM fluorescence of 719 Ϯ 159 units versus 283 Ϯ 12 units (n ϭ 3, p Ͻ 0.02) for the Bik cells. Therefore the ⌬⌿ m of the Bcl-2 cells was at least 2.6 Ϯ 0.4-fold higher than Bik cells. Taken together, these experiments indicate that higher ⌬⌿ m is associated with increased Ca m load in Bcl-2 cells and lower ⌬⌿ m is associated with decreased Ca m load in Bik cells.
Mitochondrial Calcium Load Is Decreased by Apoptosis-inducing Agent Staurosporine-To further demonstrate that Ca m load is important for normal mitochondrial function and cell survival, we tested the effect of staurosporine (STS), a commonly used apoptosis inducer. It has been shown in SH-SY5Y cells that treatment with 0.5 M STS leads to translocation of endogenous Bax from cytosol to mitochondria at 15 min, release of cytochrome c at 1 h, activation of caspase at 2-4 h, and initiation of apoptosis at 4 h (32). We have also confirmed these observations. 2 In addition, we demonstrate here that STS triggers an immediate release of Ca m in these cells (which precedes Bax translocation and cytochrome c release). Fig. 5A shows that STS caused an immediate release of Ca m similar to the effect of FCCP shown in Fig. 2A. The STS-induced Ca m release was also concomitant with an increase of Ca c (data not shown). Comparing control cells (Fig. 5A) with Bcl-2-overexpressing cells (Fig. 5B) indicated that Bcl-2 cells were more resistant to STS treatment. While it was sufficient to use 0.5 M STS to completely deplete the Ca m store in control cells, the same STS concentration caused only partial depletion of the store in Bcl-2 cells. Presumably this remaining Ca m load may be sufficient for cell survival. We have also determined caspase activity using cell extracts that was prepared after 4 h of treatment with 0.5 M STS (see "Materials and Methods"). Fig. 5C shows that STS treatment for 4 h led to prominent activation of caspase in control cells but no activation in the Bcl-2 cells (data were taken continuously at 30-s intervals during the 30-min assay). The caspase 3-like activity was 1.8 OD/min/mg of cell extract for control cells, while Bcl-2 cells had zero activity. Since at 0.5 M, STS caused complete depletion of Ca m in control cells and only partial depletion in Bcl-2 cells, the results suggest that maintenance of Ca m load at a threshold level by Bcl-2 relates to its anti-apoptotic activity.
The Pro-apoptotic Effect of Ruthenium Red-To demonstrate a direct relationship between mitochondrial Ca 2ϩ uptake and cell survival, we then treated normal SH-SY5Y cells with ruthenium red, an inhibitor of the mitochondrial uniporter. Prior control study has indicated that ruthenium red (25 M) was effective to block the mitochondrial Ca 2ϩ uptake (data not shown). The ability of ruthenium red to promote apoptosis was judged by the Hoechst dye staining of the apoptotic nuclei (see "Materials and Methods"). It should be mentioned that this assay gave a conservative estimate of apoptosis because only cells attached to the coverslip were counted, while the dead cells that became detached from the coverslip were not counted. Results from four independent experiments (Fig. 6A), indicated that while the untreated control group had negligible number of apoptotic nuclei per 200 counted cells (0.4 Ϯ 0.2, n ϭ 10), treatment with ruthenium red (25 M, 24 h) led to significant increase in the number of apoptotic nuclei (3.3 Ϯ 1.8, p Ͻ 0.001, n ϭ 8). This effect was concentration-dependent such that increasing ruthenium red to 100 M produced more apoptotic nuclei (6.3 Ϯ 0.5, p Ͻ 0.001, n ϭ 8). This result was confirmed by the DNA ladder assay (Fig. 6B), where increasing ruthenium red concentration from 25 M to 1 mM produced a dosedependent increase in the ladder formation as compared with untreated control. These data demonstrate that inhibition of mitochondrial Ca 2ϩ uptake by ruthenium red results in efficient apoptosis.
The Effect of Extracellular Calcium on Cell Viability-We reasoned that if depletion of Ca m leads to cell death, then strategies to increase or maintain the mitochondrial Ca 2ϩ store should promote cell survival. Therefore we examined the effect of extracellular calcium (Ca o ) on cell viability. Cells were grown on coverslips (1 ϫ 10 5 ) in culture medium. For this experiment, the regular growth medium was replaced with serum-free DMEM that contained variable concentration of Ca o from zero to 3 mM. After 6 h or 16 h of incubation, cells were fixed and stained for apoptotic nuclei with the Hoechst 33258 dye. Comparison between two groups was made using 1.8 mM Ca o as the reference because it approximates the normal growth condition. Fig. 7A shows a time-dependent induction of apoptosis when cells were incubated in zero Ca o for 6 h and 16 h. There was a significant number of apoptotic nuclei per 200 counted cells (4.5 Ϯ 1.3 at 6 h, and 9.3 Ϯ 2.9 at 16 h, n ϭ 10, p Ͻ 0.001) as compared with that in 1.8 mM Ca o (0.1 Ϯ 0.3 at 6 h, and 0.5 Ϯ 1.0 at 16 h, n ϭ 10). Increase [Ca o ] to 0.4 mM led to reduced number in apoptotic nuclei (0.9 Ϯ 0.9 at 6 h, and 4.0 Ϯ 1.9 at 16 h, n ϭ 10) which was still significantly higher than the 1.8 mM group (p Ͻ 0.001). There was also complete protection of cells grown in medium containing 3 mM Ca o (0.3 Ϯ 0.4 at 6 h, and 0.5 Ϯ 0.8 at 16 h). This result was confirmed by the quantitative analysis of apoptotic cells using a flow cytometric method (data not shown). The results suggest that external calcium (1.8 mM) is important for cell survival. This supposition is consistent with recent reports showing that reduced capacitative calcium entry correlates with apoptosis (33)(34)(35). In a separate experiment, the level of mitochondrial calcium (Ca m ) in cells treated with various [Ca o ] was determined using the rhod2 dye. After 16 h of incubation, the Ca m load was measured as the amount of Ca 2ϩ releasable by 20 M FCCP. The percent of release was calculated to estimate the Ca m load in these cells. Fig. 7B shows a linear relationship between Ca o and Ca m . The results (Fig. 7, A and B) indicated that both zero and 0.4 mM Ca o led to decreased Ca m , and more reduction of Ca m is associated with more apoptosis. It is worth mentioning that this experiment does not exclude the possibility that external Ca 2ϩ may affect the levels of other internal stores such as ER.

DISCUSSION
It is now recognized that there are dynamic interactions among various intracellular Ca 2ϩ stores such as mitochondria and ER (36 -38). The close physical interaction between the mitochondria and ER network has been demonstrated by using high speed imaging system that allows a three-dimensional fluorescence image of high resolution (37). The consequence of such interaction is the modulation of mitochondrial Ca 2ϩ by ER (38) and the influence of ER Ca 2ϩ release by mitochondria Ca 2ϩ uptake (8,22). It has been demonstrated in isolated cardiomyocytes that focal SR calcium release results in calcium microdomains sufficient to promote local mitochondrial calcium uptake (38). This has suggested a tight coupling of calcium signaling between SR release sites and nearby mitochondria. Previously, another laboratory and ours have reported the modulation of ER Ca 2ϩ load by Bcl-2 (17,18). The tight coupling between ER and mitochondria compartments has suggested the coordinated regulation of both the ER Ca 2ϩ and mitochondrial Ca 2ϩ by Bcl-2.
Indeed, we found that ectopic expression of Bcl-2 results in elevated loading of Ca 2ϩ in the mitochondria in addition to enhanced loading of the ER Ca 2ϩ (Figs. 1 and 2). This phenomenon is not restricted to the neuroblastoma cells (the present study) but also true for breast epithelial cells (18) and cardiomyocytes (data not shown). Therefore, the enhanced loading of both mitochondrial and ER Ca 2ϩ stores appears to be a general phenomenon. While we have shown that the increased ER Ca 2ϩ is due to the increased SERCA gene expression by Bcl-2 (18), the reason for the enhanced mitochondrial Ca 2ϩ load is not clear. One possibility is that the Bcl-2-expressing cells have more mitochondria and therefore more Ca 2ϩ uptake. This possibility has been ruled out by the Western blot experiment showing no increase in the cytochrome c protein in the Bcl-2expressing cells (data not shown). Another possibility is that higher mitochondrial membrane potential in the Bcl-2-expressing cells allows the ⌬⌿ m -dependent uptake of Ca 2ϩ by the uniporter. Alternatively, the expression of the uniporter gene could be up-regulated by Bcl-2 similar to the up-regulation of SERCA gene expression by Bcl-2 in epithelial cell (18). While this possibility cannot be ruled out, the data in Fig. 3 and 4 are consistent with the interpretation that increased Ca m load by Bcl-2 is related to an increase of ⌬⌿ m . Still an even more intriguing possibility is that increased Ca m load may be a consequence of increased ER Ca 2ϩ load in the Bcl-2-expressing cells. The tight coupling of Ca 2ϩ signaling between the ER and mitochondria compartment (38) suggest that local Ca 2ϩ release from ER could lead to local Ca 2ϩ uptake into mitochondria. Interestingly, in parallel studies of Bik-expressing cells, we found a decreased ER Ca 2ϩ load as compared with controls (data not shown). This decrease in ER Ca 2ϩ could contribute to the decreased Ca m load in Bik cells.
The present study suggests that maintenance of both mitochondrial and ER Ca 2ϩ stores is important for cell survival. It has been well documented that depletion of ER Ca 2ϩ by thapsigargin treatment leads to apoptosis in all cell types tested (33, 34, 39 -41). Here we show for the first time that depletion of mitochondrial Ca 2ϩ is the earliest event that occurs during staurosporine-induced apoptosis. The STS-induced efflux of Ca m is followed by cytochrome c release at 1 h, caspase activation at 2 h, and initiation of apoptosis at 4 h ( Fig. 5 and data not shown). Furthermore, this STS-induced caspase activation and apoptosis in SH-SY5Y cells are inhibited by Bcl-2 (Fig. 5C). A direct relationship between Ca m load and apoptosis is demonstrated by the use of ruthenium red, an inhibitor of the uniporter. Blocking Ca 2ϩ uptake into mitochondria by treatment with ruthenium red leads to apoptosis in 24 h (Fig. 6). Moreover, incubation of SH-SY5Y cells in medium containing no calcium also leads to apoptosis in 24 h, and cell death can be blocked by the addition of 1.8 mM calcium to the medium (Fig.  7). Thus, for cell survival, it is crucial to maintain a threshold level of Ca m ; below that cells will die. The results suggest that Bcl-2 by maintaining Ca m at the threshold level can protect neuroblastoma cells from STS-induced caspase activation and apoptosis (Fig. 5C). On the other hand, Bik is antagonistic to Bcl-2; it promotes apoptosis by lowering the Ca m load.
Our study showing the ability of ruthenium red to promote apoptosis in SH-SY5Y cells is in direct contrast to a recent report indicating the prevention of STS-induced apoptosis by ruthenium red in PC12 cells (42). The discrepancy between these two studies on the ruthenium red effect is not clear. In the PC12 cell study, the authors have observed a mitochondrial Ca 2ϩ overload during STS-induced apoptosis. However, this increase in Ca m is a delayed event, occurring at 8 h, well after the activation of caspase (42). According to the emergent view on the role of mitochondria in apoptosis, once the cell releases cytochrome c and activates caspases, it is committed to die by either a rapid apoptotic mechanism or a slower necrotic process (43). A recent confocal study on STS-induced apoptosis in PC-6 cells also indicates that mitochondrial depolarization (and permeability transition) accompanies cytochrome c release (30). These observations suggest that the increase of Ca m in PC-12 cells (42) is a consequence but not the cause of mitochondrial permeability transition (MPT) and apoptosis.
How depletion of mitochondria Ca 2ϩ triggers apoptosis is not clear. One possibility is that depletion of Ca m directly accompanies MPT, thereby leading to cell death. Our results (Fig. 4) of a correlation of mitochondrial potential (⌬⌿ m ) with Ca m load are consistent with this hypothesis. Furthermore, this hypothesis is also supported by a report showing a close relationship between mitochondrial calcium homeostasis with the MPT (22). Although less likely, the depletion of Ca m could indirectly lead to apoptosis via the increase of cytoplasmic Ca 2ϩ . A recent study (44) has shown that increased cytoplasmic Ca 2ϩ by thapsigargin treatment can lead to activation of the calcium-dependent phosphatase (calcineurin), translocation of proapo- ptotic protein Bad to the mitochondria, which then causes cytochrome c release and caspase activation. Mounting evidence now indicates that translocation of BH3-domain protein such as Bad, Bax, Bid, or Bik can directly induce apoptosis (32, 44 -48), and Bcl-2 protects the cell by inhibiting the MPT (49,50). It is apparent that mitochondria provide the center stage for the initiation of apoptotic events. Further study on the relationship between mitochondrial calcium signaling, membrane potential, and permeability transition is required to elucidate the mechanism that initiate the release of Ca m and opening of the permeability transition pore.