Bax and Bak promote apoptosis by modulating endoplasmic reticular and mitochondrial Ca2+ stores.

Alterations in intracellular Ca(2+) homeostasis and cytochrome c release from mitochondria have been implicated in the regulation of apoptosis, but the relationship between these events remains unclear. Here we report that enforced expression of either Bax or Bak via adenoviral gene delivery results in the accumulation of the proteins in the endoplasmic reticulum (ER) and mitochondria, resulting in early caspase-independent BCL-2-sensitive release of the ER Ca(2+) pool and subsequent Ca(2+) accumulation in mitochondria. The inhibition of ER-to-mitochondrial Ca(2+) transport with a specific inhibitor of mitochondrial Ca(2+) uptake attenuates cytochrome c release and downstream biochemical events associated with apoptosis. Bax and Bak also directly sensitize mitochondria to cytochrome c release induced by immediate emptying of ER Ca(2+) pool. Our results demonstrate that the effects of the "multidomain" proapoptotic BCL-2 family members Bak and Bax involve direct effects on the endoplasmic reticular Ca(2+) pool with subsequent sensitization of mitochondria to calcium-mediated fluxes and cytochrome c release. These effects modulate the kinetics of cytochrome c release and apoptosis.

Caspase activation during apoptosis is regulated by the release of mitochondrial polypeptide activators including cytochrome c (1), SMAC/Diablo (2,3), and AIF (4,5). Antiapoptotic members of the BCL-2 family prevent caspase activation by blocking factor release (4,6,7), whereas proapoptotic members of the family promote release (8 -11). Indeed, very recent work (12) has demonstrated that the expression of either Bax or Bak, members of the "multidomain" subfamily of proapoptotic BCL-2 homologs, is absolutely required for cytochrome c release and caspase activation in mouse embryonic fibroblasts that are exposed to a wide array of stimuli. However, the biochemical mechanisms underlying the effects of BCL-2 family polypeptides are still unclear.
Another large body of evidence implicates alterations in intracellular Ca 2ϩ homeostasis in the control of apoptosis. Early work (13) demonstrated that endogenous endonuclease activation proceeds via a Ca 2ϩ -dependent mechanism in thymocytes and certain other cell types exposed to a wide range of stimuli, and more recent studies (14 -17) suggest that BCL-2-sensitive depletion of the ER 1 Ca 2ϩ pool is an early event in apoptosis. Other studies (18,19) have shown that BCL-2 regulates mitochondrial Ca 2ϩ homeostasis and prevents Ca 2ϩ -induced cytochrome c release (20). BCL-2 constitutively associates with both the mitochondrial (21)(22)(23) and ER (24 -26) membranes via a conserved "transmembrane" domain localized within its C terminus (27,28). On the other hand, Bax appears to reside within the cytosol in resting cells but translocates to membrane fraction(s) particularly mitochondria during apoptosis (23,29,30) via a mechanism that is also dependent upon its C-terminal transmembrane domain (30). Whether or not Bax and/or other members of the BCL-2 family accumulate within the ER has not been addressed directly.
Other recent work (31) has shown that ER Ca 2ϩ release has immediate effects on mitochondrial function. Close contacts exist between mitochondria and the sites of ER Ca 2ϩ release, such that ER Ca 2ϩ release leads to rapid Ca 2ϩ accumulation in mitochondria (32,33). Under normal conditions, mitochondrial Ca 2ϩ uptake appears to serve as an activating signal to increase metabolism (32). However, one group has shown that mitochondrial Ca 2ϩ uptake promotes cytochrome c release in cells exposed to the proapoptotic agent, staurosporine (34,35). Given previous work implicating BCL-2 in the control of ER Ca 2ϩ homeostasis, we wondered whether or not proapoptotic members of the BCL-2 family might exert their effects through modulation of intracellular Ca 2ϩ stores. then diluted in normal tissue culture medium, and cells were incubated for an additional 24 h without selection. The medium was then replaced with 1 mg/ml Geneticin (G418, Invitrogen), and cells were maintained in this medium until obvious colonies appeared on the tissue culture dishes. Individual colonies were harvested and expanded, and BCL-2 expression was confirmed by immunoblotting with a monoclonal anti-BCL-2 antibody (6C8, generously provided by Dr. Timothy McDonnell, M. D. Anderson Cancer Center). Resistance to apoptosis was also verified in cells incubated for 24 -48 h with staurosporine or thapsigargin.
Adenovirus-mediated Transduction-The construction of the Ad/GT-Bak, Ad/GT-Bax, Ad/GT-LacZ, and Ad/GV16 vectors was reported previously (36). The Ad/GT-Bak or Ad/GT-Bax vector was constructed by placing Bak or Bax cDNA downstream of the GAL4/TATA promoter (GT) to generate the shuttle plasmid pAd/GT-Bak or Bax. This plasmid was cotransfected into 293 cells along with a 35-kb ClaI fragment purified from human adenoviral type 5 to generate the Ad/GT-Bak or Ad/GT-Bax vector. Bak or Bax gene expression can then be induced in target tissues by a coadministration of the Ad/GT-Bak or Ad/GT-Bax vector with the second adenoviral vector in our system, Ad/GV16, which produces the GAL4/GV16 fusion protein. Purified Ad/GT-Bak or Ad/GT-Bax was obtained by expanding the virus in 293 cells, harvesting the supernatant of those cells, and then subjecting the supernatant to ultracentrifugation on a cesium chloride gradient. Virus titers were determined by optical absorbancy at A 260 (1 A 260 unit ϭ 1012 viral particle/ml). The transduction efficiencies of adenoviral vectors in various cancer cell lines were determined by infecting cells with Ad/GT-LacZ and then determining the titers needed to transduce at least 80% cells. These levels were achieved in PC-3 cells following treatment with Ad/GT-Bak (2000 viral particles) Ad/GV16 (1000 viral particles), Ad/GT-Bax (2000 viral particles) Ad/GV16 (1000 viral particles, and Ad/GT-LacZ (2000 viral particles) Ad/GV16 (1000 viral particles).
DNA Fragmentation Analysis-We measured DNA fragmentation by propidium iodide staining and fluorescence-activated cell sorter analysis as described previously (37). Cells were harvested, pelleted by centrifugation, and resuspended in phosphate-buffered saline containing 50 g/ml propidium iodide, 0.1% Triton X-100, and 0.1% sodium citrate. Samples were stored at 4°C for 16 h and vortexed prior to fluorescenceactivated cell sorter analysis (FL-3 channel, Becton-Dickinson FACScan, Mountain View, CA).
Cytochrome c Release Measurements-The release of cytochrome c from mitochondria was measured by immunoblotting as described previously (38). Cells were harvested by centrifugation and gently lysed for 5 min in an ice-cold buffer containing 25 mM Tris and 5 mM MgCl 2 , pH 7.4. Lysates were centrifuged for 5 min at 16,000 ϫ g, supernatants were mixed with 1ϫ Laemmli's reducing SDS-PAGE sample buffer, and extracts from equal numbers of cells (10 -20 ϫ 10 6 ) were resolved by 15% SDS-PAGE. Polypeptides were transferred to nitrocellulose membranes (0.2 M, Schleicher & Schuell), and cytochrome c was detected by immunoblotting with the monoclonal antibody clone 7H8.2C12 (Phar-Mingen, San Diego, CA).
Quantification of Intracellular Ca 2ϩ in Prostate Cancer Cells-Cells plated on 22 ϫ 30-mm glass coverslips were loaded with 10 M fura-2 acetoxymethyl ester (Molecular Probe, Inc.) for 1 h at 37°C with humidified air (5% CO 2 ). The coverslips were washed thoroughly with phosphate-buffered saline and mounted on a 1.5-ml volume chamber (cells facing upward). The chamber was placed on an epifluorescence/ phase-contrast microscope for Ca 2ϩ imaging and quantitation. Cells were bathed in 1 ml of Hank's Balanced Salt Solution without Ca 2ϩ at room temperature. After a base-line [Ca 2ϩ ] i was established, cells were then treated with thapsigargin (5 M) to empty [Ca 2ϩ ] ER stores.
An INCA work station (Intracellular Imaging, Inc.) was used to quantify [Ca 2ϩ ] i levels based on fura-2 fluorescence. The INCA software allowed the subtraction of background fluorescence. Fluorescence was monitored using a ϫ20 fluorescence objective. Cells were illuminated alternately at excitation wavelengths of 340 and 380 nm using a xenon arc lamp. The emitted fluorescence was monitored at 511 nm with a video camera, and the calculated free [Ca 2ϩ ] i was determined using the cell-free calibration curve. The data were collected with INCA software (Win 3.1 version).
Spectrofluorimetric Analysis of Mitochondrial Ca 2ϩ -Cells were pelleted and resuspended in 5 ml of complete RPMI 1640 medium. 50 g of Rhod-2 acetoxymethyl ester (Molecular Probes) was diluted to 0.5 g/ml in Me 2 SO. Cells were loaded with Rhod-2 acetoxymethyl ester for 45 min, and the washed cells were analyzed in a spectrofluorimeter (Model LS 50 B, PerkinElmer Life Sciences) at 540 nm excitation and 585 emission. The analysis of washed MitoTracker-counterstained cells by confocal microscopy confirmed that the vast majority of Rhod-2 fluorescence was associated with mitochondria. Furthermore, preincubation with the mitochondrial uncoupler CCCP reduced fluorescence levels to base line and completely blocked the increases in fluorescence normally observed following stimulation with thapsigargin. To obtain fluorescence maxima and minima, cells were sequentially incubated with FIG. 1. Kinetics of cytochrome c release induced by adenoviral Bak or Bax. A, constructs were added to PC-3 as described under "Experimental Procedures," and time-dependent expression of each polypeptide was measured by immunoblotting. Loading was controlled by immunoblotting with anti-actin antibody. Cytosolic extracts were also prepared for analysis of cytochrome c release, and DNA fragmentation was assessed in gently permeabilized (0.1% Triton X-100) cells by propidium staining and flow cytometry. Experiments were performed in triplicate. B, effects of Ad-Bax. Experiments were conducted as outlined in A.

FIG. 2.
Bak and Bax localize to the endoplasmic reticulum and mitochondria. PC-3 cells were infected with Ad-Bax or Ad-Bak as described under "Experimental Procedures." After 12 h, cells were stained for each polypeptide using phycoerythrin-conjugated secondary antibodies (red) and counterstained with either MitoTracker Green or an antibody to the ER-specific protein GRP-78/BIP with an Alexa 594conjugated (green) secondary antibody, and fluorochromes were detected by confocal microscopy (ϫ60 objective). detergent and EGTA (10 mM final concentration) in the presence of saturating concentrations of extracellular Ca 2ϩ . No differences in minimum Ca 2ϩ -dependent fluorescence were observed when CCCP was substituted for EGTA in these studies. Intramitochondrial Ca 2ϩ concentrations were calculated by the formula: Statistical Analyses-Values are the means Ϯ S.E. The numbers of experiments are shown in the legends of each figure. Statistical analysis was performed by analysis of variance with either Neuman-Keuls post-hoc comparison or an unpaired Student's t test.

RESULTS AND DISCUSSION
With the observations outlined above in mind, we hypothesized that Bak and Bax-mediated emptying of the endoplasmic reticular Ca 2ϩ pool might trigger mitochondrial Ca 2ϩ uptake and subsequent cytochrome c release. To investigate this possibility, we overexpressed Bax or Bak in human PC-3 prostate adenocarcinoma cells with adenoviral vectors Ad-Bak and Ad-Bax. We decided on this approach, because it allowed us to study the effects of Bak and Bax in isolation. In addition, we designed our experiments to reproduce the kinds of increases in protein expression (2-4-fold) observed in cells exposed to more conventional proapoptotic stimuli (i.e. DNA-damaging agents) (39 -41). Transduction of the PC-3 cells resulted in detectable increases in Bak or Bax by 10 -12 h (Fig. 1). Protein expression was followed closely by the release of cytochrome c from the mitochondria and DNA fragmentation both by 12 h (Fig. 1), consistent with previous results (36,42). To determine whether Bax and Bak accumulated within mitochondria and the endoplasmic reticulum, we characterized their subcellular localizations by immunofluorescence confocal microscopy using antibodies specific to Bax, Bak in conjunction with mitochondrial (MitoTracker), or endoplasmic reticular (antibody to GRP-78/ BIP) probes. The results confirmed that both proteins traffic to both organelles, although some diffuse cytosolic staining was also detected (Fig. 2).
We next determined the effects of Bak or Bax expression on ER Ca 2ϩ pool content. Substantial reductions (Ͼ50%) in ER Ca 2ϩ levels were detected as early as 10 h after transduction with Ad-Bax or Ad-Bak, whereas infection with a control vector Ad-␤-galactosidase had no effect (Fig. 3, A and B). Bax and Bak were not expressed until 10 h, and [Ca 2ϩ ] ER calcium was not detectable after 12 h. Therefore, even though [Ca 2ϩ ] ER was significantly depleted in cells treated with Ad-Bax or Ad-Bak compared with controls at 10 h, there were still measurable amounts of Ca 2ϩ within the ER at this time point that were equal to or greater than the concentration of Ca 2ϩ within the mitochondria. The loss of ER Ca 2ϩ induced by Bax (data not shown) or Bak was unaffected by pretreatment with a pancaspase inhibitor ZVADfmk (Fig. 3C) but was prevented by the overexpression of BCL-2 in PC-3 transfectants (Bcl-2.9) (Fig.  3D). The overexpression of BCL-2 also blocked the release of cytochrome c from the mitochondria and delayed DNA fragmentation (Fig. 4).
As noted above, an efflux of Ca 2ϩ from the ER can lead to ] ER was measured as described above. C, ER calcium depletion is caspase-independent. PC-3 cells were pretreated with the pan-caspase inhibitor ZVAD (20 M) for 30 min. Cells were then exposed to Ad-Bak for 12 h, and [Ca 2ϩ ] ER was quantified as described above. D, ER calcium depletion is Bcl-2-sensitive. PC-3 and PC-3-Bcl-2.9 cells were treated with Ad-Bak and Ad-Bax for 10 h, and [Ca 2ϩ ] ER was measured by the thapsigargin release method as described above. coupled increases in Ca 2ϩ levels within mitochondria. Therefore, we assessed the effects of Bak and Bax on mitochondrial Ca 2ϩ concentrations in cells loaded with the Ca 2ϩ -sensitive dye, Rhod-2. The expression of either protein induced a significant increase in mitochondrial Ca 2ϩ accumulation, although Bax produced significantly more Ca 2ϩ uptake than Bak (p Ͻ 0.01) (Fig. 5A). The increases in mitochondrial calcium levels were almost completely suppressed in cells pretreated with a specific inhibitor of mitochondrial Ca 2ϩ uptake (RU-360) (43) (Fig. 5A). In contrast, mitochondrial Ca 2ϩ levels were unchanged following infection with the control vector Ad-␤-galactosidase or in the BCL-2 transfectants infected with Ad-Bax or Ad-Bak (Fig. 5B). Importantly, RU-360-mediated inhibition of mitochondrial Ca 2ϩ uptake attenuated the cytochrome c release and DNA fragmentation induced by Ad-Bax or Ad-Bak (Fig. 5, C and D), demonstrating that cytochrome c release occurred via a Ca 2ϩ -sensitive mechanism.
The ER Ca 2ϩ ATPase inhibitor, thapsigargin, is often em-ployed to induce rapid pharmacologic emptying of the ER Ca 2ϩ pool (reviewed in Ref. 44). Thapsigargin does induce apoptosis in a variety of different cell types, but a significant lag period is invariably observed between thapsigargin-induced emptying of ER Ca 2ϩ pool and the first biochemical end points of apoptosis (12,(45)(46)(47). To test whether direct effects of Bax and/or Bak on mitochondria might sensitize them to cytochrome c release induced by emptying of ER pool, we measured cytochrome c release in cells that had been pretreated with Ad-Bax or Ad-Bak for 10 h. At this time point, Bax and Bak protein accumulation was readily detected, but ER Ca 2ϩ release was submaximal and cytochrome c release had not yet occurred. Cytochrome c release was not observed in untreated cells or in cells infected with Ad-␤-galactosidase (Fig. 5E). Strikingly, however, exposure to Ad-Bax-or Ad-Bak-sensitized mitochondria to Ca 2ϩ induced cytochrome c release, such that the release was induced by thapsigargin within 5 min (Fig. 5E). Together with the results presented above, the data demonstrate that Bax and Bak regulate Ca 2ϩ -induced cytochrome c release by promoting the release of Ca 2ϩ from the ER pool and by sensitizing mitochondria to the effects of this release. Although not absolutely required for cytochrome c release, ER Ca 2ϩ release appears to augment the kinetics of cytochrome c release. The regulation of SMAC/Diablo and other proteins may further refine this process, which was set up initially by ER Ca 2ϩ release from the multidomain proapoptotic Bcl-2 family members, Bak and Bax. The release of protein factors from mitochondria plays a central role in cellular commitment to apoptosis, but the biochemical mechanisms controlling the release remain obscure. A growing consensus suggests that proapoptotic and antiapoptotic members of the BCL-2 family are involved. The results of this study support this conclusion and identify one candidate mechanism for their effects. Specifically, our data demonstrate that two proapoptotic members of the BCL-2 family, Bak and Bax, promote early emptying of the endoplasmic reticular Ca 2ϩ pool and subsequent accumulation of Ca 2ϩ within the mitochondria. These effects are blocked by the overexpression of BCL-2 but are not dependent upon caspase activation. Importantly, a selective inhibitor of mitochondrial Ca 2ϩ uptake attenuated the cytochrome c release and DNA fragmentation induced by either protein, providing direct evidence for a causal role for Ca 2ϩ uptake in apoptosis. Furthermore, cells overexpressing Bax or Bak were sensitized to cytochrome c release induced by thapsigargin, a compound that directly stimulates direct and complete emptying of ER Ca 2ϩ pool through the inhibition of the ER Ca 2ϩ ATPase. Together, our data indicate that Bax and Bak simultaneously promote alterations in intracellular Ca 2ϩ compartmentalization and sensitize mitochondria to cytochrome c release induced by these alterations. A role for mitochondrial Ca 2ϩ uptake in promoting cytochrome c release is consistent with other recent studies. For example, Hajnoczky and coworkers (34,35) showed that staurosporine sensitized mitochondria to cytochrome c release induced by inositol 1,4,5trisphosphate-induced emptying of the ER Ca 2ϩ pool, and independent work by Gogvadze and colleagues (48) demonstrates that Bax promotes Ca 2ϩ -mediated cytochrome c release in isolated liver mitochondria.
Although BCL-2 did not affect steady-state levels of Ca 2ϩ in the ER or mitochondria, it did inhibit the effects of Bax and Bak on ER Ca 2ϩ release and subsequent mitochondrial Ca 2ϩ uptake. These observations suggest that BCL-2 acts primarily to antagonize the effects of Bax and Bak in our cells and does not exert direct effects of its own. However, it should be stressed that other laboratories have reported different effects of BCL-2 on intracellular Ca 2ϩ pools (16,19,49,50). For example, at least three groups have shown that BCL-2 lowers the steady-state level of Ca 2ϩ within the ER, which they argued inhibits apoptosis by reducing Ca 2ϩ efflux across the ER membrane (16,19,49). In contrast, another group (50) showed that the overexpression of BCL-2 increased steady-state levels of ER Ca 2ϩ . On the other hand, Murphy et al. (18) concurred that BCL-2 had no significant effect on steady-state Ca 2ϩ levels in their cells, but they found that BCL-2 potentiated Ca 2ϩ uptake by mitochondria, an observation that stands in opposition to our results. Although we cannot explain these contrasting findings at present, they are probably related to the use of different cellular model systems. Hematopoietic cells and excitable cells (neurons) are likely to regulate intracellular Ca 2ϩ compartmentalization very differently from epithelial cells. Other groups (51) have argued that BCL-2 is capable of interacting with a variety of different proteins, some of which are involved in Ca 2ϩ -associated signal transduction (i.e. the Ca 2ϩ -dependent protein phosphatase, calcineurin). It is conceivable that cellular context dictates precisely how BCL-2 will influence intracellular Ca 2ϩ pools. Importantly, the connection between ER Ca 2ϩ pool emptying and apoptosis is not disputed.
Although our studies establish a role for Bax and Bak in regulating ER and mitochondrial Ca 2ϩ fluxes, they do not directly identify the proximal mechanism(s) involved. Most of the available information on the biochemical mechanisms underlying the actions of BCL-2 family members has come from studies describing their effects on mitochondria and planar lipid bilayers. Specifically, proapoptotic and antiapoptotic members of the family appear to bind to and regulate components of the permeability transition pore (9,10), and permeability transition pore opening has been implicated in cytochrome c release in many model systems (reviewed in Ref. 52). In addition, a caspase-processed form of the BH3-only family member, Bid (tBID), can promote oligomerization of Bax or Bak to form transmembrane channels capable of directly allowing for passage of cytochrome c into the cytoplasm (53,54). Finally, in vitro studies (55-58) (reviewed in Ref. 59) have shown that BCL-2, BCL-X L , Bax, and Bak can form smaller ion-selective channels in planar membranes. It is conceivable that large pore formation induced by oligomerization of Bax or Bak within the ER membrane directly stimulates Ca 2ϩ release and/or that Ca 2ϩ release is regulated by the ion channel properties of the two proteins. Our data also do not formally rule out the possi- bility that ER Ca 2ϩ release occurs secondarily to earlier alterations within mitochondria. Importantly, the characterization of BCL-2 family channel conductance properties suggested that BCL-2 and BCL-X L form pores that are selective for monovalent cations (Na ϩ and K ϩ ), whereas Bax and Bak form pores selective for monovalent anions (Cl Ϫ ) (58). Unless these properties are dramatically modified in vivo, these observations argue against the idea that Bax and Bak directly form Ca 2ϩselective channels in the ER membrane.
Our results also demonstrated that Bax produced significantly higher increases in mitochondrial Ca 2ϩ than did Bak. To date, the multidomain members of the BCL-2 family have been thought of as largely redundant, but hints of biochemical selectivity have emerged recently. For example, studies in mouse embryo fibroblasts lacking Bax, Bak or both proteins demonstrated that one of them must be expressed for cytochrome c release to occur following exposure of the cells to a panel of common proapoptotic stimuli (12). These results suggest that other members of the multidomain subfamily are not capable of compensating for the loss of Bax or Bak in the cells. In separate studies, we found that Bax-null cells (DU-145 prostate cancer cells and BaxϪ/Ϫ mouse embryo fibroblasts) display almost no mitochondrial Ca 2ϩ uptake following exposure to staurosporine, a defect that is corrected by the reintroduction of wild-type Bax. 2 Further work is required to confirm these observations and to identify the biochemical mechanisms involved.
Although adenoviral gene delivery allowed us to isolate the effects of Bax and Bak on intracellular Ca 2ϩ fluxes, the results presented here do not address the role of intracellular Ca 2ϩ fluxes in cytochrome c release induced by "endogenous" proapoptotic stimuli. Especially important is the recent observation that almost all intracellular Bax are localized to the cytosol in resting cells but move to a rigid membraneassociated compartment in cells undergoing apoptosis in response to staurosporine, X-irradiation, or glucocorticoid hormone (23,29,30). As discussed above, we designed our experiments to reproduce the magnitudes of the increases in Bax or Bak that are observed in cells exposed to some endogenous stimuli. Furthermore, although the localization studies clearly demonstrate that staurosporine and other agents promote the translocation of Bax from the cytosol to membranes, it should be emphasized that a significant fraction of introduced Bax associated constitutively with membranes. In a separate study, 2 we found staurosporine and a cancer chemotherapeutic agent (doxorubicin), not anti-Fas antibody, also induced ER Bax/Bak accumulation, ER Ca 2ϩ pool depletion, and mitochondrial Ca 2ϩ uptake, and in these systems mitochondrial Ca 2ϩ uptake was dependent on the expression of Bax. These results support the conclusions generated in the present study and indicate that Ca 2ϩ -dependent and -independent pathways for cytochrome c release exist within cells. A direct role for mitochondrial Ca 2ϩ uptake in cytochrome c release was also demonstrated in myotubes exposed to a variety of different proapoptotic stimuli (35), although in these excitable cells, the mitochondrial Ca 2ϩ uptake occurred as a result of ER-mediated Ca 2ϩ waves. Calcium-dependent mechanisms may also dramatically alter the structural properties of the ER. Previous work (60) has shown that increases in Ca 2ϩ cause a dramatic decrease in protein mobility within the ER, and this may in part explain the rigidity of the membrane Bax compartment observed within apoptotic cells.