Bax-dependent Regulation of Bak by Voltage-dependent Anion Channel 2*

Many studies have demonstrated a critical role of Bax in mediating apoptosis, but the role of Bak in regulating cancer cell apoptotic sensitivities in the presence or absence of Bax remains incompletely understood. Using isogenic cells with defined genetic deficiencies, here we show that in response to intrinsic, extrinsic, and endoplasmic reticulum stress stimuli, HCT116 cells show clear-cut apoptotic sensitivities in the order of Bax+/Bak+ > Bax+/Bak– » Bax–/Bak+ » Bax–/Bak–. Small interference RNA-mediated knockdown of Bak in Bax-deficient cells renders HCT116 cells completely resistant to apoptosis induction. Surprisingly, however, Bak knockdown in Bax-expressing cells only slightly affects the apoptotic sensitivities. Bak, like Bax, undergoes the N terminus exposure upon apoptotic stimulation in both Bax-expressing and Bax-deficient cells. Gel filtration, chemical cross-linking, and co-immunoprecipitation experiments reveal that different from Bax, which normally exists as monomers in unstimulated cells and is oligomerized by apoptotic stimulation, most Bak in unstimulated HCT116 cells exists in two distinct protein complexes, one of which contains voltage-dependent anion channel (VDAC) 2. During apoptosis, Bak and Bax form both homo- and hetero-oligomeric complexes that still retain some VDAC-2. However, the oligomeric VDAC-2 complexes are diminished, and Bak does not interact with VDAC-2 in Bax-deficient HCT116 cells. These results highlight VDAC-2 as a critical inhibitor of Bak-mediated apoptotic responses.

Bax and Bak are important mediators of cell death as bax/ bak double knockout cells are resistant to multiple apoptotic inducers (1,2). In most cells, Bax is normally localized in the cytosol or loosely associated with the outer mitochondrial membrane, whereas Bak is mostly localized in the outer mitochon-drial membrane and remains inactive perhaps because of binding to VDAC-2 1 (3). BH3-only Bcl-2 proteins such as tBid and Bim, upon activation or up-regulation by apoptotic stimuli, cause translocation of Bax to the outer mitochondrial membrane and subsequent conformational changes and oligomerization of Bax and Bak. The activated Bax and/or Bak facilitate the release of some (e.g. cytochrome c and Smac) (4) but not all (e.g. AIF and Endo G) (5) proapoptotic proteins from the mitochondria by forming pores or channels or otherwise altering mitochondrial membrane permeability and structural architecture of the organelles (6,7). Recent work suggests that Bax and Bak localized on the ER also play a role in maintaining Ca 2ϩ homeostasis and in ER-mitochondria cross-talk during apoptosis induction (8).
Numerous studies have demonstrated a critical role for Bax in determining tumor cell sensitivity to apoptosis induction and in tumor development. For example, somatic mutation in bax, which occurs in colon (9,10), prostate (11,12), hematopoietic (13), and other (14) cancer cells, confers on tumor cells a survival advantage and promotes clonal expansion. Increased Bax degradation through proteosome-dependent mechanisms (15) and reduced Bax expression (12,16) have also been reported in prostate cancers. In animal models, Bax has been shown to suppress tumorigenesis (17) and to be required for tBid-induced hepatocyte apoptosis (18). Loss of Bax alters tumor spectrum (from lymphoma to carcinoma) (19) and impairs Mycinduced apoptosis during Myc-mediated lymphomagenesis (20). Even loss of a single allele of bax has been shown to accelerate mammary tumor development (21). Somatic knocking out of bax renders HCT116 colon cancer cells resistant to both intrinsic and extrinsic apoptotic inducers as well as to ER stressors (22)(23)(24)(25)(26)(27). Bax has also been shown to play an important role in apoptosis induced by death receptors (25, 28 -30), p53 (31), ceramide (32), and BH3-only protein Bik (33).
Compared with Bax, fewer studies have been dedicated to Bak. Somatic bak mutations have been reported only in gastric and colon cancer cells (34), and reduced Bak expression is observed in prostate cancer cells (16). Some studies have also implicated Bak as a determinant of drug sensitivity (35). In general, most comparative studies suggest that Bax and Bak have independent, nonredundant proapoptotic activities (29, 36 -38). However, the role of Bak in regulating cancer cell apoptotic sensitivities in the presence or absence of Bax re-mains incompletely understood. In this study, we address this issue using the siRNA-mediated ablation of Bak and also explore the potential underlying mechanisms of action.
Subcellular Fractionation and Western Blotting-Whole cell lysate, mitochondrial membrane fractions, and cytosol were prepared, and Western blotting was preformed as previously detailed (16, 39 -41).
Quantification of Apoptosis and Caspase Activity Measurement-Apoptotic cells were counted based on apoptotic nuclei upon staining cells live with DAPI (12,16). DEVDase and LEHDase activities were measured as previously described (39 -41).
Establishment of HCT116 Cells Stably Expressing Bak siRNA Using Retroviral Vectors-Three double-stranded 21-nt siRNAs against human bak were initially made using the Silencer TM kit (Ambion, Austin, TX). Their sequences were: bak1 (5Ј-CTCTGAGTCATAGCGT-CGG-3Ј), bak2 (5Ј-GGTGAAGTACTCATAGGCA-3Ј), and bak3 (5Ј-AA-ACAGGCTGGTGGCAATC-3Ј). Preliminary experiments revealed the strongest effect with bak3, and this siRNA was then used in most subsequent experiments. A scrambled sequence (5Ј-ACAAGAGTCGT-GGACGTCA-3Ј) for bak3 was used as control. To generate stable expression of siRNA, we first cloned the shRNA sequences corresponding to bak3 and the scrambled control into the pSuper.retro retroviral vector (42). These vectors were transfected into the Ampho Phoenix packaging cells. 48 h after transfection, the culture medium containing the infectious virions was harvested, filtered (0.2 m), and directly used to infect HCT116 cells. 48 h later, the cells were selected with puromycin (2 g/ml), and stable clones were isolated 2-3 weeks later using a cloning ring. Stables clones were analyzed by Western blotting to confirm Bak down-regulation.
Clonogenic Survival Assays-Cells (2 ϫ 10 5 /well) in 6-well tissue culture dishes were transfected with Bak-siRNApool or cont-siRNA and treated with apoptotic stimuli 24 h later. At the end of the treatment, the cells were fixed and stained with Giemsa at room temperature.
Immunofluorescence-Cells grown on coverslips were treated and, 15 min prior to the end of the treatment, incubated live with MitoTracker Orange CMTMRos and DAPI to label mitochondria and nuclei, respectively. Then cells were fixed, permeabilized, and immunolabeled (39 -41) for conformationally active Bax or Bak.
Chemical Cross-linking and Oligomerization Assay-Mitochondria (50 g) were suspended in 45 l of HIM buffer (200 mM mannitol, 70 mM sucrose, 10 mM HEPES-KOH, 1 mM EGTA, pH 7.5) followed by addition of freshly prepared bismaleimidohexane to a final concentration of 10 mM and incubated at room temperature for 30 min. The mitochondria were then mixed in protein sample buffer and subjected to SDS-PAGE and Western blotting. Dithiothreitol in the sample buffer quenched the cross-linking reaction (3).
Immunoprecipitation-HCT116 and HCT116-Bax-KO cells (treated or untreated) were lysed in CHAPS lysis buffer (10 mM HEPES, 150 mM NaCl, 1.5 mM MgCl 2 1 mM EGTA, 2% CHAPS, pH 7.4). Cell lysates were first precleared by rabbit or goat (depending on the primary antibody used) IgG-conjugated beads for 1 h at 4°C. The resulting supernatants were incubated with primary antibodies, i.e. rabbit anti-Bak NT, or goat anti-VDAC-2, respectively, for 2 h at 4°C, followed by the addition of rabbit or goat IgG beads for 1 h. Finally, the beads were centrifuged, washed thoroughly, boiled in SDS sample buffer, and analyzed by SDS-PAGE and Western blotting.
Statistics-Statistical analyses were performed using paired Student's t tests in SPSS software.

RESULTS AND DISCUSSION
Bax Deficiency Confers Apoptosis Resistance in HCT116 Colon Cancer Cells That Express Bak-To determine the role of Bak in regulating apoptosis, we first utilized isogenic colon cancer cells HCT116 and HCT116-Bax-KO (22) to investigate how Bax deficiency would affect the apoptotic sensitivity. HCT116 cells expressed Bax, whereas HCT116-Bax-KO did not, but both expressed Bak (Fig. 1A). We treated both cells side by side with three intrinsic apoptosis inducers including STS and two DNA-damaging agents (i.e. VP16 or etoposide and x-ray irradiation), one extrinsic apoptosis inducer (i.e. Fas ligation with a cross-linking antibody), and an ER stress inducer (i.e. TG).
The results demonstrated that Bax deficiency delayed and inhibited apoptosis (measured by counting apoptotic nuclei) induced by all five stimuli (Table I). For instance, STS treatment for 6 h resulted in Ͼ40% apoptosis in HCT116 wild-type cells, whereas it only killed ϳ12% HCT116-Bax-KO cells (Table  I). STS-induced HCT116 cell death was partially caspase-dependent as evidenced by 1) increased DEVDase activity (Fig.  1E) and 2) inhibition of apoptosis by Z-VAD (Table I). Similarly, when HCT116 cells were challenged with a Fas cross-linking antibody in the presence of CHX, ϳ40% cells were killed within 4 h, and the loss of Bax delayed the onset and also decreased the peak level of apoptosis (Table I). The combined treatment of anti-Fas antibody and CHX resulted in the highest level of DEVDase activity, which was inhibited by loss of Bax (Fig. 1E), suggesting that HCT116 cells behave as so-called "type II" cells, in which the death receptor-induced apoptosis requires the mitochondrial activation of the intrinsic apoptotic pathway. Not surprisingly, Z-VAD demonstrated an inhibitory effect (Table I).
VP16, a DNA topoisomerase inhibitor used to treat many epithelial cancers, killed nearly all cells within 72 h (Table I). VP16-induced apoptosis was similarly delayed and inhibited in HCT116-Bax-KO cells (Table I). A lack of Bax inhibited DEVDase activity (Fig. 1E), suggesting that VP16-induced apoptosis in these cells involved Bax-mediated caspase activation. Indeed, Z-VAD inhibited apoptosis in both cells (Table I).
Another DNA-damaging agent, x-ray irradiation, killed ϳ90% HCT116 cells within 72 h (Table I). Bax deficiency also partially inhibited x-ray-induced cell death (Table I). Interestingly, x-ray-induced death in either HCT116 or HCT116-Bax-KO cells was not inhibited by Z-VAD (not shown), suggesting that it might be caspase-independent. In support, DEVDase activity was not detected in either cell type upon irradiation (Fig. 1E). Furthermore, x-ray-induced death in these cells morphologically resembled necrosis (not shown).
Finally, when HCT116 cells were stimulated with TG, an irreversible inhibitor of the ER outer membrane pump SERCA whose function is to maintain the concentration of ER Ca 2ϩ several orders of magnitude above that of the cytosol, Ͼ40% of the cells died within 12 h, and ϳ85% underwent apoptosis by 36 h (Table I). Again, loss of Bax delayed the onset and also decreased the peak level of apoptosis (Table I). As expected, Z-VAD significantly inhibited TG-induced cell death (Table I), and Bax deficiency led to lower DEVDase activity (Fig. 1E).
To determine whether the Bax deficiency-conferred resistance is at least partially mediated via decreased release of mitochondrial cytochrome c, we treated HCT116 or HCT116-Bax-KO cells with one extrinsic (i.e. Fas ligation) and one intrinsic (i.e. STS) inducer. As shown in Fig. 1B, the loss of Bax reduced the apoptotic stimuli-induced release of the mitochondrial holocytochrome c, which was detected by an antibody that specifically recognizes holo-but not apo-cytochrome c (39).
Altogether, the above results have revealed the following points. First, the loss of Bax delays the apoptosis initiation in that the inhibitory effect of Bax deficiency appears more dramatic at earlier time points (Table I). These observations are consistent with Bax functioning as a critical gateway in the mitochondrial pathway of apoptosis (1,2). Second, at later time points, the Bax deficiency-rendered apoptosis resistance is somewhat diminished (Table I), likely because of compensation from Bak (see below). Third, Bax might also play a role in caspase-independent cell death (43-45) because HCT116-Bax-KO cells are resistant to x-ray irradiation, which does not activate DEVDase and induces Z-VAD-insensitive cell death (Table I and Fig. 1E).
Restoration of Bax Expression Enhances Apoptosis in Baxdeficient, Bak-expressing Du145 Prostate Cancer Cells-Baxdeficient Du145 cells, compared with other prostate cancer cells, are relatively more resistant to apoptosis by some inducers such as starvation (12) and ceramide (32). Restoration of Bax expression using stable retroviral transduction confers sensitivity in Du145 cells to ceramide-induced cell death (32). As shown in Fig. 1C, Du145-mock cells (i.e. Du145 cells subjected to mock infection) (32) were devoid of Bax expression, whereas Du145-Bax cells expressed Bax but both expressed Bak. Bax re-expression enhanced the apoptotic sensitivities of Du145 cells to all five apoptotic stimuli, with different kinetics and to different degrees (Table I). For example, Du145-Bax cells were more sensitive to STS, VP16, and x-ray throughout the treatment period (Table I). By contrast, Du145-Bax cells showed increased sensitivity to Fas ligation only early upon  (Table I). By 8 h after Fas ligation, similar levels of apoptosis were observed in both Du145-mock and Du145-Bax cells (Table I), probably because caspases became activated by Bax-independent mechanisms. On the other hand, Du145-Bax cells did not show enhanced apoptotic responses to TG until 36 h after stimulation (Table I).
Compared with the isogenic pair of HCT116 cells, isogenic Du145 cells showed some interesting differences. For example, Du145 cells were more sensitive to Fas ligation than the HCT116 pair (Table I) in that the former cells, with or without Bax, were nearly completely killed by 8 h, whereas within the same time frame only ϳ50% of the HCT116 cells were dead. On the other hand, Bax appeared to play a more prominent role in HCT116 cells than in Du145 cells in determining their sensitivities to TG (Table I), suggesting that the Bax-regulated response to ER stress might be cell type-dependent. Furthermore, Bax re-expression did not result in increased DEVDase activities in Du145-Bax cells treated with TG (Fig. 1F), suggesting that Bax restoration may lead to TG-induced activation of caspases other than caspase-3/7. Indeed, TG has been shown to activate the death-receptor pathway leading to caspase-8 activation (25,27).
To determine whether the Bax restoration-conferred apoptosis sensitivity is at least partially mediated via increased release of mitochondrial cytochrome c, we treated isogenic Du145 cells with Fas ligation or STS. As shown in Fig. 1D, Bax re-expression increased the release of the mitochondrial cytochrome c.
siRNA-mediated Down-regulation of Bak in Bax-expressing Cells Only Slightly Inhibits Mitochondrial Protein Release, Caspase Activation, and Apoptosis-The preceding experiments support a critical role of Bax in determining cancer cell apoptotic sensitivities, as previously reported by others (e.g. [22][23][24][25][26][27][28][29][30][31][32][33]. Nevertheless, both HCT116-Bax-KO and Du145 cells still undergo apoptosis in response to these stimuli except with slower kinetics, suggesting that other molecules, in particular, Bak, are also functional apoptosis regulators. To test this, we carried out reciprocal experiments to examine how ablation of Bak in Bax-expressing cells might affect cancer cell apoptotic sensitivity. To that end, we designed and synthesized three siRNA oligonucleotides targeting human bak. When tested in Du145 cells, all three bak siRNAs down-regulated Bak protein levels in the order of bak3 Ͼ bak1 Ͼbak2 (supplemental Fig.  S1A). These bak siRNAs also down-regulated Bak protein expression in HCT116 and HCT116-Bax-KO cells (not shown). Unfortunately, although these siRNAs dramatically down-regulated Bak expression by 24 h, their inhibitory effects essentially disappeared by 72 h (supplemental Fig. S1A). To circumvent this problem, we constructed retroviral vectors encoding either bak3 shRNA or a scrambled control shRNA, which were used to infect HCT116 cells. Of seven stable clones, clones 3 and 4 showed the lowest Bak protein expression (supplemental Fig. S1B). We thus treated scrambled clone 3 (as control) and Bak clone 4 HCT116 cells with STS, Fas/CHX, and TG. Much to our surprise, both clones showed overall similar levels of release of cytochrome c and Smac as well as caspase-3 activation, i.e. generation of the p20/p17 fragments of active caspase-3 (40) (Fig. 2A). Clonogenic survival assays also did not reveal significant differences in cell death and clonogenicity (not shown).
To further establish the role of Bak in Bax-expressing cells, we obtained a new generation of siRNAs from Dharmacon, which have longer lasting and more powerful silencing effects as a pool of target siRNAs is used. A pool of bak siRNAs (containing four siRNA oligonucleotides) or control siRNAs were transfected into HCT116 or HCT116-Bax-KO cells, and Western blot analysis revealed a time-dependent down-regulation of Bak in both cell types by the bak siRNAs compared with the control siRNAs, whereas levels of Bax were not affected ( Fig. 2B). By 72 h, Bak was down-regulated to barely detectable levels (Fig. 2B). Subsequently, we transfected HCT116 cells with either control or bak siRNA pool, and, 48 h later, the cells were treated with STS, Fas/CHX, or TG. Several repeat experiments demonstrated that release of cytochrome c and Smac, caspase-9 activation (i.e. LEHDase activity), caspase-3 activation (i.e. generation of the p20/p17 bands and DEVDase activity), and cell death were only slightly inhibited by Bak downregulation (Figs. 2C and 3C, below). These results together suggest that Bak ablation in Bax-expressing cells, surprisingly, does not significantly inhibit apoptotic stimuli-induced mitochondrial events or cell death.

Down-regulation of Bak in Bax-deficient Cells Significantly Inhibits Mitochondrial Release of Proapoptotic Proteins as Well as Caspase Activation and Apoptosis-To determine whether
Bak even plays a role in regulating apoptosis in the absence of Bax, we transfected HCT116-Bax-KO cells with either the control or bak siRNA pool, the latter of which time-dependently down-regulated Bak levels (Fig. 2B). Forty-eight h after transfection, the HCT116-Bax-KO cells were treated with STS, Fas/ CHX, or TG. As shown in Fig. 3A, Bak down-regulation completely inhibited the apoptotic stimuli-induced cytochrome c and Smac release as well as caspase activation. Little active caspase-3 fragments or activity was detected (Fig. 3A). Consequently, the clonogenic survival of Bak-ablated HCT116-Bax-KO cells was significantly enhanced (Fig. 3B).
Collectively, the data in Figs. 1-3 demonstrate that: 1) Bax plays an obvious role in regulating cancer cell apoptotic sensitivities whether Bak is present or not; 2) Bak down-regulation in Bax-expressing cells does not significantly affect the apoptotic events; and 3) however, Bak plays a critical role in mediating apoptosis in the absence of Bax.
Bak N Terminus Exposure in Both Bax-expressing and Baxdeficient Cells-In the foregoing experiments, we explored the potential mechanisms that might explain why Bak seems to play differential roles in Bax-expressing and Bax-deficient cells. One possibility is that Bak might not be efficiently activated by apoptotic stimuli in Bax-expressing cells, whereas it is activated in the absence of Bax. We therefore first examined its N terminus exposure because Bak activation, like Bax activation, is thought to involve at least two important events: conformational changes (e.g. exposure of the N terminus) to expose the buried BH3 domains and oligomerization to form putative "channels" to release proapoptotic molecules (8,46,47). We treated the two pairs of cells with various inducers and immunostained with an antibody that preferentially recognizes the conformationally active Bak or Bax. Cell death was assessed by staining cells live with the plasma membrane-impermeable dye DAPI, whereas mitochondria were identified using the membrane potential-sensitive dye, MitoTracker (40,41). As illustrated in Figs. 4 and 5, both Bak and Bax underwent conformational changes evidenced by the exposure of their N termini. For example, control Du145-Bax cells did not show Bax or DAPI labeling, and cells demonstrated heterogeneous mitochondrial labeling (Fig. 4, A-D). Twelve h post STS treatment, ϳ40% of the cells showed activated Bax (Fig. 4E) and apoptotic nuclei (Fig. 4G), which could be more clearly seen under higher magnification (supplemental Fig. S2). Interestingly, most of the STS-treated Du145-Bax cells had lost the mitochondrial membrane potential as revealed by loss of MitoTracker labeling, but surprisingly, the majority of Bax-activated, apoptotic cells still retained prominent MitoTracker labeling such that the activated Bax clearly co-localized with the mitochondria in these cells (Fig. 4, E-H, and supplemental Fig. S2). In cells treated with VP16 or Fas ligation, more Bax-activated cells were observed than DAPI-positive cells (Fig. 4 and supplemental Fig. S2), suggesting that Bax activation, as expected, preceded cell death. Again, activated Bax in most cases co-localized with the mitochondria (Fig. 4 and supplemental Fig. S2). Different from STS, VP16 and Fas did not cause widespread mitochondrial potential collapse ( Fig. 4 and supplemental Fig.  S2). We observed similar results in stimulated HCT116 cells with respect to Bax N terminus exposure (not shown).
Likewise, in both Bax-expressing as well as Bax-deficient Du145 cells, we observed similar levels of Bak staining indicative of N terminus exposure in response to STS (Fig. 5), Fas (Fig. 5), or VP16 (not shown). In both cell types stimulated with STS, most cells with Bak staining were DAPI-positive and also retained MitoTracker labeling (Fig. 5), just as in Bax-activated cells (Fig. 4). Similarly, in Fas (Fig. 5) or VP16 (not shown) stimulated Du145-mock or Du145-Bax cells, many Bak-activated cells had not shown apoptotic nuclei, suggesting that, like Bax (Fig. 4), Bak activation precedes nuclear apoptosis. Similar results on Bak N terminus exposure were obtained in HCT116 and HCT116-Bax-KO cells (data not shown). Collectively, the observations in Figs. 4 and 5 suggest that Bak (and Bax) underwent apoptotic stimuli-induced N terminus exposure in both Bax-expressing and Bax-deficient cells.
Different Bak and Bax Oligomerization Status in Unstimulated Cells and during Apoptotic Stimulation-Next, we examined the second event critical in Bak (and Bax) activation, i.e. oligomerization. Bax is generally thought to exist as monomers in the cytosol (or loosely attached to the outer mitochondrial membrane), and apoptotic stimulation induces its oligomerization in the mitochondria. Indeed, when we utilized gel filtration analysis to characterize Bax, we found that Bax in unstimulated HCT116 cells was mainly expressed as monomers eluted at fractions 14 -16 (Fig. 6A). Fas ligation in HCT116 cells resulted in a shift of the protein toward high molecular mass fractions (fractions 2-11) (Fig. 6A) peaking at ϳ158 kDa (fractions 9 and 10), indicating the formation of Bax oligomers. To provide additional support for Bax oligomerization, we analyzed the formation of a higher order Bax complex by chemical cross-linking. To that end, freshly isolated mitochondria were incubated with bismaleimidohexane, a noncleavable, membrane-permeable homobifunctional maleimide that covalently and irreversibly cross-links sulfhydryl groups (1,3). As shown in Fig. 7 (A and B), Fas ligation induced formation of homooligomeric Bax proteins in the mitochondria as early as 4 h (7A), and Bax oligomers further increased by 6 h (Fig. 7B). As expected, Bax oligomerization was not observed in HCT116-Bax-KO cells (Fig. 7, A and B) or the cytosol of Fas-treated HCT116 cells (not shown). STS treatment similarly induced Bax oligomerization in HCT116 but not in HCT116-Bax-KO cells (see Fig. 7G, lanes 3 and 9). Likewise, Bax oligomerization was observed in Du145-Bax but not Du145-mock cells treated with either Fas or STS (Fig. 7C and supplemental Fig. S3). Interestingly, apoptotic stimulation slightly increased the Bax protein levels in some experiments (Fig. 7, A-C, and supplemental Fig. S3), which resulted, most likely, from increased translocation to the mitochondria. These data altogether suggest that Bax is activated in our apoptotic systems by both exposing the N terminus and undergoing homo-oligomerization as identified by bismaleimidohexane.
Much to our surprise, the elution profile of Bak in size exclusion chromatography showed a very different pattern from that of Bax in unstimulated HCT116 cells (Fig. 6A). In contrast to Bax, Bak was eluted in fractions 1-14 with 3 relative peaks, i.e. peak 1 (complex I, ϳ700 -2,000 kDa) at fractions 1-3, peak 2 (complex II, ϳ230 kDa) at fractions 7-9, and peak 3 (monomers, Յ67 kDa) at fractions 11-13 (Fig. 6A). These results suggest that most Bak molecules in unstimulated HCT116 cells pre-exist in at least two protein complexes, i.e. complex I and II. Fas ligation in HCT116 cells resulted in the reduction in complex I and monomers and a corresponding increase in complex II, which partially overlapped with the ϳ158-kDa Bax-containing protein complex (Fig. 6A). These results suggest that: 1) Fas ligation induced increased Bak oligomerization in complex II, 2) the ϳ230 kDa Bak-containing complex II may represent an apoptosis-related protein complex, and 3) the Bak complex II might contain both Bax and Bak homo-oligomers as well as Bax-Bak hetero-oligomers (see below).
The Bak elution profiles in unstimulated and Fas-activated HCT116-Bax-KO cells were overall similar to those observed in the corresponding wild-type HCT116 cells (Fig. 6A, lower  panel). However, we observed relatively less dramatic increase in the Bak-containing complex II formation accompanied by less obvious decrease in complex I in Fas-stimulated HCT116-Bax-KO cells compared with in the HCT116 cells (Fig. 6A). These observations were sort of opposite to what we had expected and suggest that Bax might actually be involved in facilitating Bak oligomerization, as recently suggested by others (26). Indeed, when we analyzed the oligomerization status of Bak using bismaleimidohexane cross-linking, we observed Bak oligomerization in HCT116 (Fig. 7, E and G, lanes 1-3) or Du145-Bax (Fig. 7F, lanes 4 -6) cells treated with STS or Fas ligation. By contrast, little (Fig. 7, E and G) or significantly reduced (Fig. 7F) Bak oligomerization was observed in HCT116-Bax-KO or Du145-mock cells. It is unclear at the moment why chemical cross-linking could not efficiently detect the slightly increased Bak oligomers observed by gel filtration analysis in stimulated Bax-deficient cells (Fig. 7, E-G). Perhaps the cross-linked high molecular mass complexes could not be resolved well on regular polyacrylamide gel. Interestingly, Bak oligomerization in Fas-stimulated cells was observed at 6 h but not 4 h (Fig. 7, D and E), when Bax oligomerization was already apparent (Fig. 7A), suggesting that Bax oligomerization might also precede Bak oligomerization. By contrast, when Bak expression was down-regulated by siRNA, similar levels of Bax oligomerization were still observed in HCT116 cells treated with STS or Fas ligation (Fig. 7G, lanes 1-6). As expected, no significant Bak oligomerization was observed in bak siRNA-transfected HCT116 cells treated with STS or Fas ligation (Fig. 7G, lanes 4 -6). It was interesting to note that apoptotic stimulation sometimes also up-regulated Bak protein expression (Fig. 7, E and F).

Different Bak and Bax Oligomerization Statuses: Relationship to the Changes in VDAC-1 and VDAC-2-Both Bak and
Bax have been shown to interact with some resident mitochondrial proteins, and these interactions have been proposed to regulate the apoptotic responses of the cell. For example, Bax has been proposed to interact with some components of the permeability transition pore, such as VDAC-1, to mediate mitochondrial protein release during apoptosis (reviewed in Refs. 46 and 47), although some recent studies have challenged this view (48,49). By contrast, Bak has been shown to bind to VDAC-2 in unstimulated mouse embryonic fibroblasts and apoptotic signals disrupt this "sequestration" and thus release Bak to carry out proapoptotic functions (3). On these considerations, we reasoned that perhaps Bak interactions with these proteins might explain why Bak seems to play a more prominent role in regulating apoptosis in Bax-deficient cells than in Bax-expressing cells.
We first examined the oligomerization status of VDAC-1 and VDAC-2 in our gel filtration experiments. As shown in Fig. 6A, VDAC-1 in both HCT116 and HCT116-Bax-KO cells pre-existed in gigantic protein complexes peaking at ϳ2000 kDa (i.e. fraction 1). Apoptotic stimulation induced a clear-cut increase in the low molecular mass VDAC-1-containing complexes and monomers in both cell types without a corresponding decrease in the high molecular mass protein complex (Fig. 6A), suggesting that apoptotic stimulation may up-regulate the VDAC-1 protein levels. Indeed, Western blotting using whole cell lysates revealed increased VDAC-1 in HCT116 cells treated with both Fas ligation and STS (not shown). Different from VDAC-1, VDAC-2 in unstimulated cells existed in two populations, i.e. a major population representing monomers eluted at fractions 10 -13 (Յ67 kDa) and a minor population representing oligomers eluted at fractions 7-9 (ϳ230 kDa) (Fig. 6A). Fas liga- tion caused formation of larger VDAC-2-containing complexes eluted at fractions 7-9, which coincided with the Bak complex II (Fig. 6A), suggesting that the two proteins might co-exist in this complex. Of interest, the Fas ligation-induced formation of larger VDAC-2 complex in fractions 7-9 in HCT116-Bax-KO cells was much less obvious than in HCT116 cells (Fig. 6A).
To help elucidate the complex composition, we carried out co-immunoprecipitation (co-IP) experiments in both HCT116 (Fig. 6B) and HCT116-Bax-KO (Fig. 6C) cells either untreated or stimulated with Fas ligation or STS. We focused on the potential interactions between Bak and VDAC-2 because conflicting data have been presented about the role of VDAC-1 in regulating apoptosis (46 -49). In all of our co-IP experiments, we first cleared cell lysates with Rb or goat IgG conjugated to the agarose beads (see "Materials and Methods"), thereby stringently eliminating any nonspecific interactions. In unstimulated HCT116 cells, IP using the anti-Bak antibody that preferentially recognizes the N terminus-exposed Bak (but also some native Bak protein) pulled down prominent VDAC-2 but not Bax (Fig. 6B, lane 1). Reciprocal co-IP using anti-VDAC-2 antibody similarly pulled down Bak but not Bax in unstimulated HCT116 cells (Fig. 6B, lane 4). These results are consistent with gel filtration data showing co-elution of Bak with VDAC-2 but not Bax in fractions 7-9 (Fig. 6A) and with earlier studies showing the Bak sequestration by VDAC-2 in unstimulated cells (3).
Fas stimulation, and more prominently, STS stimulation, resulted in increased levels of active Bak (Fig. 6B, lanes 2 and  3), confirming that these apoptotic stimuli induce conformational changes and activation of Bak in HCT116 cells revealed in immunolabeling experiments. The anti-Bak antibody pulled down prominent amounts of Bax in Fas-stimulated HCT116 cells (Fig. 6B, lane 2), suggesting that during Fas activation Bak heterodimerizes with Bax, consistent with the gel filtration data showing that the oligomerized Bax in fractions 9 and 10 partially overlaps with the Bak complex II (Fig. 6A, top  panel). Interestingly, STS stimulation of HCT116 cells induced much less association between active Bak with Bax (Fig. 6B,  lane 3), suggesting that STS stimulation might cause more Bak homo-oligomers with less Bak/Bax hetero-oligomers. Anti-Bak IP during apoptotic stimulation pulled down reduced amounts of VDAC-2 (Fig. 6B, lanes 2 and 3), consistent with the data of others showing that VDAC-2 normally inhibits the proapoptotic activity of Bak and apoptotic signals induce dissociation (or "desequestration") of Bak from VDAC-2 (3). Interestingly, apoptosis stimulation induced increased VDAC-2 levels as revealed by anti-VDAC-2 IP (Fig. 6B, lanes 5 and 6). Because the total VDAC-2 protein levels did not change upon apoptotic stimulation (not shown), these results suggest that the apoptotic stimuli most likely cause increased VDAC-2 oligomerization, as supported by the gel filtration results (Fig. 6A, top  panel). VDAC-2 did not associate with Bax even during apoptosis (Fig. 6B, lanes 5 and 6).
We then carried out similar experiments in HCT116-Bax-KO cells (Fig. 6C). Surprisingly, although the same anti-Bak antibody pulled down some Bak in untreated and increased amounts of Bak in stimulated cells, it did not pull down any VDAC-2 (Fig. 6C, lanes 1-3). Reciprocal co-IP experiments using anti-VDAC-2 antibody pulled down VDAC-2 but not Bak (Fig. 6C, lanes 4 -6). These results suggest that Bak, in the absence of Bax, somehow does not appreciably interact with VDAC-2. Interestingly, decreased VDAC-2 was pulled down upon STS stimulation (Fig. 6C, lane 6). As expected, IP using Rb IgG did not result in any specific products (Fig. 6C, lanes 7-9). In summary, we have shown that in HCT116 colon and Du145 prostate cancer cells the overall apoptotic sensitivities to a spectrum of apoptotic stimuli rank in the order of Bax ϩ / Bak ϩ Ͼ Bax ϩ /Bak Ϫ Ͼ Ͼ Bax Ϫ /Bak ϩ Ͼ Ͼ Bax Ϫ /Bak Ϫ . Thus, in cancer cells that have both wild-type Bax and Bak, Bax appears to play a major role in that loss of Bax expression confers certain apoptotic resistance, whereas Bax re-expression in cancer cells that lack endogenous Bax enhances apoptotic sensitivity. By contrast, ablation of Bak in Bax-expressing cells only slightly inhibits the apoptotic stimuli-induced mitochondrial protein release, caspase activation, and cell death. On the other hand, in cells without Bax, Bak plays a pivotal role in dictating apoptotic sensitivity. Therefore, ablation of Bak in these cells renders them much resistant to mitochondrial dysfunction and apoptosis induction.
It is not very clear at present why in Bax-expressing cells, Bak knockdown only confers a subtle apoptosis-resistant phenotype because in these cells Bak does undergo the N terminus exposure and homo-and hetero-oligomerization upon apoptotic stimulation. Presumably, activated Bax oligomers are sufficient to release enough apoptogenic proteins from the mitochondria to mediate apoptosis. Therefore, although apoptotic stimuli activate Bak, ablation of Bak will not significantly affect the Bax-mediated apoptotic events as supported by similar levels of Bax oligomerization and activation in bak siRNAtreated cells.
It is also unclear why the absence of Bax renders Bak critical in mediating the apoptotic responses. One possibility is that in the absence of Bax, more activated Bak might be available for apoptosis induction because of reduced inhibitory effect of VDAC-2. This possibility is supported by the gel filtration data showing significantly reduced formation of the high molecular mass VDAC-2 complex that co-elutes with the proapoptotic Bak complex II in HCT116-Bax-KO cells (Fig. 6A). More importantly, in Bax-expressing cells, not all VDAC-2 dissociates from Bak, and some VDAC-2 remains bound to Bak during apoptosis induction (Fig. 6B). However, in Bax-deficient cells, there is no significant interaction between Bak and VDAC-2 either with or without apoptotic stimulation (Fig. 6C). Because VDAC-2 has recently been shown to be a major inhibitor of the proapoptotic functions of Bak (3), our observations potentially explain why the siRNA-mediated down-regulation of Bak manifests such a dramatic inhibitory effect on apoptosis. Absence of Bak-VDAC-2 interactions in Bax-deficient cells is unlikely caused by the Bax deficiency per se because VDAC-2 does not directly interact with Bax (Fig. 6B). Nevertheless, the absence of Bak-VDAC-2 interactions and thus the lack of inhibitory effects of VDAC-2 may underlie the critical role of Bak in dictating the apoptotic sensitivity of Bax-deficient cells.
The mechanistic studies here also shed some interesting light on how Bax and Bak may be activated and differentially involved in apoptosis. Although both Bak and Bax activation involves N terminus exposure, the two proteins differ significantly with regards to oligomerization. Bax exists mainly as monomers in unstimulated cells, and apoptotic stimulation induces apparent oligomerization on the mitochondria. In contrast, most Bak pre-exists in several protein complexes of different sizes in unstimulated cells, and apoptotic stimulation results in increased formation of the ϳ230-kDa complex II with corresponding decreases in the monomers and the Ն700-kDa complex I. That Bak preexists in several protein complexes of different sizes raises the possibility that the constitutively expressed, mitochondrially localized, and VDAC-2-bound Bak (Ref. 3 and this study) might also play a nonapoptotic function(s) in unstimulated cells. Recent studies showing a neuroprotective function for Bak (50) supports this possibility. This possibility also resonates well with recent demonstrations that several proapoptotic molecules including Omi (51), FasL (52), and Bad (53, 54) also play important physiological or even pro-survival functions. In this sense, the increase in VDAC-2 complex formation in stimulated HCT116 cells (Fig. 6, A and B) might be considered as a prosurvival mechanism activated by apoptotic stimuli, as we have recently shown in several other systems (55). In HCT116-Bax-KO cells, this putative prosurvival mechanism (i.e. VDAC-2 oligomerization) is diminished upon apoptotic stimulation (Fig. 6, A and C), thus allowing Bak to play a full-throttle role in mediating apoptosis.