Association of Bax and Bak Homo-oligomers in Mitochondria

ATP depletion induced by hypoxia or mitochondrial inhibitors results in Bax translocation from cytosol to mitochondria and release of cytochrome c from mitochondria into cytosol in cultured rat proximal tubule cells. Translocated Bax undergoes further conformational changes to oligomerize into high molecular weight complexes (Mikhailov, V., Mikhailova, M., Pulkrabek, D. J., Dong, Z., Venkatachalam, M. A., and Saikumar, P. (2001) J. Biol. Chem. 276, 18361–18374). Here we report that following Bax translocation in ATP-depleted rat proximal tubule cells, Bak, a proapoptotic molecule that normally resides in mitochondria, also reorganizes to form homo-oligomers. Oligomerization of both Bax and Bak occurred independently of Bid cleavage and/or translocation. Western blots of chemically cross-linked membrane extracts showed nonoverlapping “ladders” of Bax and Bak complexes in multiples of ∼21 and ∼23 kDa, respectively, consistent with molecular homogeneity within each ladder. This indicated that Bax and Bak complexes were homo-oligomeric. Nevertheless, each oligomer could be co-immunoprecipitated with the other, suggesting a degree of affinity between Bax and Bak that permitted co-precipitation but not cross-linking. Furthermore, dissociation of cross-linked complexes by SDS and renaturation prior to immunoprecipitation did not prevent reassociation of the two oligomeric species. Notably, expression of Bcl-2 prevented not only the oligomerization of Bax and Bak, but also the association between these two proteins in energy-deprived cells. Using Bax-deficient HCT116 and BMK cells, we show that there is stringent Bax requirement for Bak homo-oligomerization and for cytochrome c release during energy deprivation. Using Bak-deficient BMK cells we further show that Bak deficiency is associated with delayed kinetics of Bax translocation but does not affect either the oligomerization of translocated Bax or the leakage of cytochrome c. These results suggest a degree of functional cooperation between Bax and Bak in this form of cell injury, but also demonstrate an absolute requirement of Bax for mitochondrial permeabilization.

Members of the Bcl-2 family of proteins are key regulators of programmed cell death or apoptosis (1,2). These proteins are known to affect mitochondrial function and regulate the release of apoptosis-activating factors from mitochondria (1)(2)(3). Antiapoptotic members of the Bcl-2 family (Bcl-2, Bcl-XL, Ced-9, Bcl-w, and Mcl-1) act primarily to preserve mitochondrial integrity by suppressing the release of cytochrome c. In contrast, pro-apoptotic members (Bax, Bak, Bik, Blk, Bok, Hrk, BNIP3, Bad, Bid, Bim, and EGL-1) mainly induce the release of stimulators of apoptosis and cause mitochondrial dysfunction.
Apoptosis is regulated by the subcellular localization and translocation of Bcl-2 family members. Anti-apoptotic members such as Bcl-2 and Bcl-XL and the pro-apoptotic member Bak reside predominantly in mitochondria (4,5). In contrast, proapoptotic members Bax, Bid, and Bad that reside in the cytosol are translocated to the outer mitochondrial membrane in response to stress or apoptotic stimuli and release intermembrane space proteins (4, 6 -8). It was shown through gene knockout studies that pro-apoptotic Bax and Bak have redundant function and are required for the induction of apoptosis in response to a variety of death signals (9,10). Distinct mechanisms are involved in translocation of BH3 only proapoptotic proteins Bad and Bid from cytosol to mitochondria. Phosphorylation and 14-3-3 binding regulate Bad translocation (11), whereas Bid translocation is regulated by proteolytic cleavage as part of the Fas and tumor necrosis factor-␣ signaling pathways (7,8). Regulation of Bax translocation is distinct from that of Bad and Bid and involves a conformational change that exposes the amino and carboxyl termini leading to mitochondrial translocation (12)(13)(14)(15). Nevertheless, the exact mechanisms that cause conformational changes in Bax are still unknown.
Although translocation of Bax to the mitochondrial outer membrane is required to release cytochrome c during apoptosis induced by various death stimuli, the mechanism of Bax-mediated membrane permeabilization is still being debated. Recent studies have shown that Bax forms homo-oligomeric complexes in mitochondrial membranes (16 -19). These findings and other experiments on artificial lipid membranes (17) have suggested that Bax undergoes conformational changes before and after insertion, followed by oligomerization and pore formation. However, the possibility that Bax could interact with other proteins in mitochondrial locations continues to be explored. Oligomerization and/or activation of Bak and Bax were also induced by Bid, another pro-apoptotic Bcl-2 family member, during death receptor signaling (15, 19 -22). Recent studies have indicated that non-Bid-mediated mechanisms are also involved in the oligomerization of Bax and Bak in viral-mediated apoptosis (23).
We have shown previously that severe ATP depletion of cultured rat kidney proximal tubule cells induced by hypoxia or chemical inhibitors of mitochondrial respiration triggers the translocation of cytosolic Bax to mitochondria and cytochrome c release into the cytosol (24). This was attributed to pore formation by Bax homo-oligomerization (16). Bax oligomerization as well as mitochondrial outer membrane permeabilization was prevented by Bcl-2 without forming physical complexes with Bax (16). Here we report that Bak, another proapoptotic member of the Bcl-2 family, exists as part of a large protein complex of unknown composition in mitochondria of normal cells, but undergoes rearrangement to form homooligomeric complexes following Bax insertion into mitochondria of hypoxic/ATP-depleted cells. Our data suggest functional dependence between Bax and Bak and strong association between Bax and Bak complexes in mitochondria. In addition they also demonstrate a stringent Bax dependence to affect mitochondrial permeabilization. Accordingly, selective Bak deficiency delayed the kinetics of Bax translocation, but did not prevent Bax oligomerization or cytochrome c release. On the other hand, selective Bax deficiency prevented both Bak oligomerization and cytochrome c release.
ATP Depletion, Subcellular Fractionation, and Chemical Crosslinking-RPTC were cultured in serum-supplemented Ham's F-12/ DMEM as described before (24), HCT116 Bax(Ϫ/Ϫ) and Bax(ϩ/Ϫ) cells (26), BMK cells, and HeLa cells were cultured in McCoy's 5A medium, DMEM, and minimum essential medium supplemented with 10% fetal bovine serum, respectively. Cells were plated at 1-2 ϫ 10 5 cells/cm 2 in 60-or 100-mm dishes. After overnight growth, cells in glucose-free Krebs-Ringer bicarbonate buffer (in mM: 115 NaCl, 1 KH 2 PO 4 , 4 KCl, 1 MgSO 4 , 1.25 CaCl 2 , and 25 NaHCO 3 ) for RPTC, HeLa, and HCT116 or glucose-and serum-free DMEM for BMK were subjected to ATP depletion induced by a mitochondrial inhibitor (uncoupler CCCP) or hypoxia (incubated in an anaerobic chamber) for the indicated times. Necrotic injury in ATP-depleted cells was prevented by inclusion of 5 mM glycine in the buffer to simulate glycine contents of tissues in vivo (27). Cytosolic and membrane fractionation and chemical cross-linking were done as described before (16). Briefly, cells were permeabilized at room temperature with 0.015-0.02% digitonin for 1-2 min in isotonic buffer A (10 mM HEPES, 150 mM NaCl, 1.5 mM MgCl 2 , 1 mM EGTA, pH 7.4) containing protease inhibitors (1 mM 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride, 0.8 mM aprotinin, 50 mM bestatin, 15 mM E-64, 20 mM leupeptin, 10 mM pepstatin A). The permeabilized cells were shifted to 4°C, scraped, and collected into centrifuge tubes. The supernatants (digitonin/cytosol) were collected after centrifugation at 15,000 ϫ g for 10 min at 4°C. The pellet was further extracted with ice-cold lysis buffer (2% CHAPS in buffer A containing protease inhibitors) for 60 min at 4°C to obtain membrane fraction. Cells permeabilized with digitonin or membranes extracted with CHAPS were incubated with bifunctional cross-linkers (1 mM dithiobis(succinimidyl propionate) or EGS with linker lengths of 12 and 16 Å, respectively) on a head-to-head rocker for 30 min at room temperature. After quenching the unreacted cross-linkers with 1/10 volume of 2 M Tris-HCl (pH 7.4), cells or extracts were incubated for another 30 min at room temperature with rocking. After cross-linking, membranes were extracted with 2% CHAPS in buffer A. In some experiments, membrane extracts were prepared in 0.35% SDS plus 1 mM DTT in buffer A instead of CHAPS. Mitochondria from Bax(Ϫ/Ϫ) and Bax(ϩ/Ϫ) cell homogenates were obtained by differential centrifugation (16).
Assessment of Mitochondrial Membrane Potential-To examine changes in the mitochondrial membrane potential, experimental cells were loaded with rhodamine 123 (50 M), a membrane-permeant mitochondria-specific tracer dye, for 10 -15 min at 37°C in a CO 2 incubator. Cells were washed with phosphate-buffered saline and observed under a fluorescent microscope.
Size Exclusion Chromatography-Size exclusion chromatography was performed using the AKTA purifier 10 (Amersham Biosciences) at room temperature with a Superose 6 (10/30) gel filtration column equilibrated in column buffer (buffer A supplemented with 2% (w/v) CHAPS and 1 mM DTT) at a flow rate of 0.4 ml/min. For gel filtration experiments, whole cells were extracted with column buffer containing protease inhibitors. Extracts were centrifuged at 500,000 ϫ g for 15 min before loading onto the Superose 6 column. After loading the column with 200-l samples, fractions of 0.5 ml were collected and concentrated with trichloroacetic acid/acetone precipitation before analysis by Western blotting. Gel filtration protein standards were used to calibrate the column. Protein standards were thyroglobulin (669 kDa), 11 ml; ferritin (440 kDa), 12.6 ml; catalase (232 kDa), 14.3 ml; bovine serum albumin (67 kDa), 15.1 ml; ovalbumin (43 kDa), 15.6 ml; chymotrypsinogen A (25 kDa), 17.5 ml; and cytochrome c (12.4 kDa), 19.1 ml. The void volume of the column was 7.5 ml.
Immunoanalysis-Western blotting was done as described before (16,24). Briefly, proteins were resolved by SDS-PAGE in Xcell II mini cells using 10 or 4 -12% (gradient) NuPAGE gels (Invitrogen) with MES running buffer in the presence (EGS cross-linking) or absence (dithiobis(succinimidyl propionate) cross-linking) of DTT. Proteins in the gels were transferred to polyvinylidene difluoride membranes (0.2 m) by electroblotting. For Western blot analysis, appropriate primary antibodies and peroxidase-conjugated secondary antibodies were used. Chemiluminescent substrates (Pierce) were used to detect antigenantibody complexes on the polyvinylidene difluoride membrane. Molecular weights of cross-linked Bax and Bak complexes were calculated by plotting their migrations against migrations of molecular weight standards in semilogarithmic plots.
Immunoprecipitation was carried out as described earlier (16). Solubilized extracts (100 -500 g) in lysis buffer were precleared and the resultant supernatants were incubated with primary antibody (2 g) at 4°C for 2 h. Immunoprecipitates were collected by incubating with protein G-Sepharose for 1 h, followed by centrifugation for 2 min at 4°C. The pellets were washed with lysis buffer three times. The immunoprecipitates dissolved in SDS/sample buffer were analyzed by Western blotting as described above.
Immunocytochemistry was performed as described before (24). Cells plated on coverslips were fixed with a modified Zamboni's fixative followed by exposure to primary antibodies, anti-Bax (mouse monoclonal) and anti-Bak (rabbit polyclonal) or anti-cytochrome c (monoclonal, clone 2G8.B6 kindly provided by Dr. R. Jemmerson of University of Minnesota Medical school, MN) followed by Alexa-conjugated antirabbit and CY3-conjugated anti-mouse antibodies.

RESULTS
Bax Translocation, and Reorganization/Oligomerization of Bax and Bak during ATP Depletion-ATP depletion by hypoxia or treatment with CCCP, a mitochondrial uncoupler, was shown to cause Bax translocation to mitochondria and release of cytochrome c into cytosol in cultured rat kidney proximal tubule cells (16,24). The BMK cells deficient in both Bax and Bak showed resistance to tumor necrosis factor-␣-induced cell death (25). However, cells deficient in Bax or Bak alone were susceptible to tumor necrosis factor-␣-induced cell death suggesting that Bak can substitute for Bax in forming mitochondrial pores to release cytochrome c. This prompted us to further investigate the role of Bax and Bak in outer membrane permeability of mitochondria in energy-deficient cells. Double immunolabeling of RPTC with anti-Bak (green) and anti-Bax (red) revealed that Bak is constitutively present in mitochondria whereas Bax is predominantly localized in the cytosol of normal cells (Fig. 1A, panel 1). ATP depletion by CCCP resulted in co-localization of both Bak and Bax in the mitochondria (Fig.  1A, panel 2). We have shown earlier that following Bax translocation, cytochrome c is released into cytosol (16,24). To find out whether the development of permeability transitions can explain cytochrome c release after prolonged exposure to CCCP, cells depleted of ATP by CCCP were allowed to recover in complete growth medium without CCCP, but in the presence of z-VAD, a caspase inhibitor that blocks downstream apoptotic events. During recovery, cells were loaded with rhodamine 123, whose accumulation in mitochondria is dependent on the presence of the potential gradient across the mitochondrial inner membrane. The results presented in Fig Inset shows Bak (green) and Bax (red) now co-localized to mitochondria. B, RPTC were exposed to CCCP for 4 h (CCCP). Another group of CCCP-treated cells were allowed to recover in full growth medium without CCCP in the presence of z-VAD, a caspase inhibitor to block downstream events of apoptosis (Recovery). Both groups of cells as well as control cells (Control) were loaded with rhodamine 123 for 10 -15 min and photographed using fluorescence microscopy. C, following incubation of RPTC with 15 M CCCP for 0, 1, 2, 3, or 4 h, cells were treated with the cleavable membrane-permeable cross-linker dithiobis(succinimidyl propionate) (1 mM). Membrane fractions were obtained as described under "Experimental Procedures," and analyzed for Bax (lanes [1][2][3][4][5] or Bak (lanes 6 -10) by Western blotting under nonreducing conditions. Prolonged exposure to CCCP resulted in progressive accumulation of Bax in the membrane fraction as slow moving complexes (lanes [3][4][5]. Similarly, slow moving Bak complexes appeared after ATP depletion in parallel with Bax accumulation in mitochondria (lanes 6 -10). Closed arrowheads indicate Bax oligomers and open arrowheads indicate Bak oligomers. D, membrane extracts of ATP-depleted cells (4 h CCCP) after incubation with the noncleavable cross-linker EGS were analyzed by Western blotting under reducing conditions. The molecular weights of Bax and Bak complexes were calculated by plotting their migrations against migrations of molecular weight standards in semilogarithmic plots from several experiments. Two representative lanes of Bak ladders and Bax ladders are shown and they show homo-oligomers of Bak and Bax, respectively. E, time course analysis of Bid distribution by Western blotting in the cytosol and membranes of ATP-depleted cells.
both control and recovering cells were able to a mount potential showing of perinuclear mitochondrial distribution of rhodamine fluorescence (Fig. 1B, panels 1 and 3). On the other hand, in CCCP-treated cells, rhodamine 123 remained in the cytosol indicating loss of mitochondrial potential (Fig. 1B, panel 2). Experiments done in parallel showed that cells treated similarly with CCCP had leaked cytochrome c into cytosol without or with an additional period of recovery (not shown). The ability of mitochondria to accumulate rhodamine 123 despite having leaked cytochrome c precluded the possibility that permeability transitions had occurred, because the development of transitions is inconsistent with the ability of membranes to maintain a barrier to the free diffusion of protons.
To identify Bax-Bak interactions, cells were fractionated into cytosolic and membrane fractions (16) followed by chemical cross-linking with dithiobis(succinimidyl propionate) or EGS. Western blotting of cross-linked proteins showed that progressively greater amounts of Bax translocated to mitochondria during increasing durations of ATP depletion and oligomerized into dimers and higher oligomers in multiples of ϳ21 kDa, the monomer (16) (Fig. 1C, lanes 1-5). Concurrently, mitochondrial Bak, normally present as monomers as well as very large protein complexes, rearranged to form Bak oligomeric "ladders" (Fig. 1, C and D). As in the case of Bax (16), the calculated molecular weights of Bak complexes in these ladders correspond to multiples of ϳ23 kDa, the monomer (Fig. 1D, lane 4). This suggests that these newly formed complexes of Bak are homo-oligomers. The mobility and molecular sizes of Bak oligomers were quite distinct from those of Bax oligomers (Fig. 1,  C and D). The large Bak complexes greater than 250 kDa that were observed in normal cells at the top of the gel (Fig. 1D, lane 3) were diminished in amount during ATP depletion with concomitant appearance of Bak oligomers (Fig. 1D, lane 4). It has been reported that Bak oligomerization is associated with t-Bid translocation to mitochondria after Bid cleavage by caspase-8 (17). Analysis of Bid protein in RPTC during ATP depletion (Fig. 1E) indicated no significant change of Bid protein levels in the cytosol. The small amount of Bid detected in membrane fractions of control cells was unchanged during ATP depletion. Furthermore, we did not detect cleaved products of Bid such as t-Bid (ϳ15 kDa) in the membrane fraction, suggesting that Bid is not involved in Bax translocation and oligomerization.
Bax translocation was observed also in HeLa cancer cells that had been subjected to hypoxia or ATP depletion. The data presented in Fig. 2A show Bax translocation and release of cytochrome c from mitochondria into cytosol during hypoxia in HeLa cells. As in the case of RPTC, hypoxia did not lead to Bid cleavage ( Fig. 2A) or affect its localization. Chemical crosslinking of membrane proteins in HeLa cells showed oligomerization of translocated Bax (Fig. 2B, lane 4Ј) and dimerization of Bak (Fig. 2B, lane 4). Bak reorganization was mainly seen as Bak dimer in HeLa cells; this appears to be because of overshadowing of Bak oligomers by very large amounts of Bak containing protein complexes of heterogeneous molecular size in control as well as in hypoxic cells (Fig. 2B, lanes 2 and 4).
Formation of Oligomeric Complexes of Bax and Bak-We used size exclusion chromatography to characterize Bax and Bak protein complexes in RPTC. Total cell lysates in 2% CHAPS lysis buffer from normal and ATP-depleted RPTC were fractionated on a Superose 6 gel filtration column and eluted fractions were analyzed by Western blotting. Results presented in Fig. 3A indicated that Bax is monomeric in normal cells with a peak elution at about 25 kDa. In contrast, Bax from ATPdepleted cells eluted in fractions with apparent molecular weights between 25,000 and 1340,000 with a peak at about 440,000 consistent with the formation of large Bax complexes (Fig. 3A). On the other hand, Bak eluted in a broad range between 43 to 5000 kDa even in normal cells, with a peak at about 232 kDa (Fig. 3B). The distribution of Bak as part of a large complex in mitochondria of normal cells was evident also after protein cross-linking; Bak adducts of Ͼ250 kDa size were observed (Fig. 1C). At present, the identity of protein(s) associated with Bak in mitochondria is unknown. ATP depletion induced a small but significant alteration in Bak elution profile with a peak at about 440 kDa (Fig. 3B). The overlapping elution profiles of Bax and Bak suggest possible association between these proteins in the mitochondria of ATP-depleted cells.
Immunoelectron microscopy confirmed the translocation and localization of Bax in mitochondrial membranes in the form of complexes. Bax was seen exclusively in the cytosol of normal cells but clustered around mitochondria in apoptotic cells after ATP depletion and repletion induced by reincubation in growth  (Fig. 3C, panels 1 and 2, respectively). Localization of Bax clusters largely on the surfaces of mitochondria is consistent with our proposal that translocated and oligomerized Bax forms channels resident in mitochondrial outer membranes (16). We attempted immunoelectron microscopy of Bak also; although suitable for Western blotting, the antibodies currently available caused unacceptable background signals in immuno-EM images (not shown).
Oligomerization of Bax and Bak Is Inhibited by Bcl-2-We have shown earlier that Bcl-2 overexpression prevents Bax oligomerization and cytochrome c release in RPTC after hypoxia or ATP depletion (16). We have now tested the effect of Bcl-2 on Bak oligomerization using chemical cross-linking and gel filtration. As shown in Fig. 4A, Bcl-2 overexpression prevented the rearrangement of both translocated Bax and mitochondrial Bak during ATP depletion induced by CCCP. By gel filtration, alterations in the elution profile of Bax in ATPdepleted Bcl-2 1 cells were only modest (Fig. 4B). On the other hand, the Bak elution profile was compressed following CCCP treatment of Bcl-2 1 cells (Fig. 4C, bottom panel) as in the wild type RPTC (Fig. 3B, bottom panel). The compression of elution profiles corresponded to reduction in the amounts of normally present large Bak adducts retained at the top of the gel after chemical cross-linking (Fig. 4A, lanes 6 and 8). Nevertheless, smaller Bak oligomers did not form in ATP-depleted Bcl-2 overexpressing cells relative to controls (Fig. 4A, lane 8). These results suggest that Bcl-2 inhibits not only the oligomerization of Bax following translocation of the protein, but also the rearrangement of mitochondrial Bak to form small homo-oligomers.
Bax Oligomers Are Associated with Bak Oligomers-Whereas gel filtration studies have helped to identify the formation of large protein complexes, they failed to characterize the organization of these complexes. Therefore, chemical crosslinking along with immunoprecipitation was undertaken to clarify how Bax and Bak complexes are organized during ATP depletion. For immunoprecipitation, CHAPS extracts were used because nonionic detergents such as Nonidet P-40 and Triton X-100 induce conformational change and oligomerization of Bax (16,29). The anti-Bax antibody (1D1) recognizes a buried epitope of Bax in normal cells and precipitates little or no Bax (Fig. 5A, lane 7 and 9, bottom panel) (24). However, Bax is precipitated by the anti-Bax antibody in ATP-depleted RPTC and Bcl-2 1 cells (Fig. 5A, bottom panel, compare lanes 8 and 10  with lanes 7 and 9) because of Bax conformational changes that result in membrane translocation. A modest but significant increase of Bak precipitation was seen with anti-Bak antibodies also, following ATP depletion (Fig. 5A, upper panel, compare  lanes 1 and 3 with 2 and 4). Although chemical cross-linking yielded distinct nonoverlapping ladders of Bak and Bax (Fig. 1,  B and C), Western blotting of anti-Bax immunoprecipitates of ATP-depleted wild type RPTC revealed the presence of Bak in the precipitate (Fig. 5A, top panel, lane 8). Similarly, anti-Bak immunoprecipitates from ATP-depleted cells contained Bax (Fig. 5A, bottom panel, lane 2). Therefore, co-immunoprecipitation results indicated possible association between Bak and Bax oligomers in the mitochondria of ATP-depleted wild type RPTC (Fig. 5A, lanes 2 and 8). As expected, overexpression of Bcl-2 inhibited co-immunoprecipitation of Bax and Bak from ATP-depleted cells (Fig. 5A, lanes 4 and 10). Furthermore, Bcl-2 was not co-precipitated with either Bax antibodies or Bak antibodies (not shown) suggesting lack of physically stable association between pro-apoptotic Bak or Bax with anti-apoptotic Bcl-2. To detect Bak-Bax oligomers with increased sensitivity, chemically cross-linked extracts were also subjected to co-immunoprecipitation. Interestingly, distinct ladders of Bax dimers, trimers, and higher order oligomers were co-precipitated with anti-Bak (Fig. 5B, lane 6), and Bak dimers, trimers, and tetramers were co-precipitated with anti-Bax in ATP-depleted RPTC (Fig. 5B, lane 2) confirming the association between Bax and Bak oligomers.

Dissociation and Reassociation of Bax and Bak Oligomers-
The presence of nonoverlapping ladders of Bax and Bak raised the question of how Bax homo-oligomers would interact with Bak homo-oligomers. To address this question, in some experiments, mitochondrial membranes from ATP-depleted cells were extracted with SDS buffer (0.35%) with or without prior chemical cross-linking (labeled ϩSDS in Fig. 6). The SDS extracts were heated to 70°C for 10 min to dissociate noncovalently interacting molecules. Proteins were renatured by 10fold dilution of SDS in the presence of 2% CHAPS at room temperature. After SDS dilution, proteins were immunoprecipitated with either anti-Bak or anti-Bax antibodies. In another group of experiments, immunoprecipitation was performed without prior SDS treatment (labeled ϪSDS in Fig. 6). Western blotting analysis of Bak (or Bax) immunoprecipitates of noncross-linked mitochondrial membranes revealed that little or no Bax (or Bak) is co-precipitated if they had been exposed to SDS prior to treatment with the precipitating antibody (shown for Bak immunoprecipitate in Fig. 6A). Thus, in the absence of chemical cross-linking, all complexes of Bax and Bak were separated into monomers during SDS treatment. Failure to co-precipitate Bax with Bak from dissociated complexes suggests that Bak and Bax monomers did not reassociate with each other after exposure to SDS. On the other hand, immunoprecipitation of Bak from chemically cross-linked mitochondrial proteins after SDS denaturation and dilution in CHAPS yielded Bax oligomers containing three or more molecules (Fig. 6B, lanes 5 and 6). These immunoprecipitates contained little or no Bax monomers and dimers (Fig. 6B, lane 6) relative to conditions where they had not been previously exposed to denaturation with SDS (Fig. 6B, lane 5). Of interest, there were modest amounts of Bax-Bak heterodimers in the precipitates that resisted dissociation by SDS. However, higher order Bax-Bak oligomers could not be identified. Anti-Bax antibodies precipitated all forms of Bak including monomers and dimers, with or without SDS denaturation and dilution in CHAPS (Fig. 6B, lanes 3 and 4). In the absence of interaction between monomeric forms of Bax and Bak, this result suggests that all forms of Bak interact with oligomeric Bax.
The most important inference from these observations on the effect of SDS exposure prior to immunoprecipitation of crosslinked Bax and Bak is that regardless of prior SDS treatment, co-precipitated ladders of Bax and Bak are nonoverlapping with respect to the molecular sizes of the respective oligomers (Fig. 6B, lanes 1, 2 and 3, 4). The results pose a paradox with respect to how Bax and Bak oligomers are formed and how they associate with each other. It seems apparent that the observations can only be explained by the formation of largely homogeneous Bax and Bak homo-oligomers in ATP-depleted mitochondrial membranes. The propensity of these oligomers to co-precipitate may indicate a degree of affinity between Bax and Bak oligomers that cannot be preserved by cross-linking, but one which may be compatible with a role for Bax translocation and homo-oligomerization in subsequently inducing the recruitment and formation of Bak oligomers. Thus, oligomerization of translocated Bax may be a pre-requisite to interact with Bak and cause Bak reorganization.
Functional Interdependence of Bax and Bak-To investigate the role of Bax in Bak reorganization and cytochrome c release, we have subjected Bax(ϩ/Ϫ) and Bax(Ϫ/Ϫ) colon cancer cells to FIG. 5. Co-immunoprecipitation of Bak and Bax oligomers. A, immunoprecipitation of 2% CHAPS extracts of whole cells (ϳ300 g) from control and ATP-depleted RPTC and Bcl-2 1 cells with anti-Bax (mouse monoclonal, 1D1) and anti-Bak (rabbit polyclonal) antibodies were carried out as described under "Experimental Procedures." The immunoprecipitates were analyzed under reducing SDS-PAGE, followed by immunoblotting with anti-Bak and anti-Bax antibodies. These blots were subsequently re-probed for either Bax or Bak. Both Bak (lanes 1 and 3) and Bax (lanes 7 and 9) antibodies failed to co-precipitate Bax and Bak, respectively, from control cells. On the other hand, ATP-depleted (4 h CCCP) cells showed signals for the presence of Bax (lane 2, bottom) and Bak (lanes 8, top), respectively, in Bak and Bax immunoprecipitates. Bax in cytosol is poorly precipitated by anti-Bax antibody in CHAPS extracts from normal cells (lanes 7 and 9, bottom panel). Only Bax translocated to mitochondria is recognized well by this antibody in CHAPS extracts from ATP-depleted RPTC and Bcl-2 1 cells (lanes 8 and 10, bottom panel). A moderate increase in Bak recognition by polyclonal anti-Bak antibody is also noted (compare lanes 1 and 3 with lanes 2 and 4, top panel). There is little or no cross-precipitation of Bax or Bak with each other in ATP-depleted Bcl-2 1 cells (Lanes 4 and  10). B, digitonin insoluble, 2% CHAPS extracted membrane fractions (ϳ100 g) after chemical cross-linking with EGS from both ATP-depleted and control cells were subjected to immunoprecipitation as described under "Experimental Procedures" with monoclonal anti-Bax (lanes 1-4) and polyclonal anti-Bak antibodies (lanes 5-8). Immunoblotting of Bax immunoprecipitates with anti- Bak (lanes 1-4) and Bak immunoprecipitates with anti-Bax antibodies (lanes 5-8) was performed after SDS-PAGE under reducing conditions. Distinct ladders of Bak in Bax immunoprecipitates and Bax in Bak immunoprecipitates were observed in lysates from ATP-depleted RPTC but not in control or Bcl-2 1 cells with or without ATP depletion.

ATP depletion by CCCP. Bax was totally absent in Bax(Ϫ/Ϫ) cells. Unlike RPTC or HeLa cells, Bax(ϩ/Ϫ) cells contained
ϳ30 -40% of total Bax in mitochondria, and the remainder was present in the cytosol of normal cells (Fig. 7A, lanes 1 and 2). When mitochondria isolated from Bax(ϩ/Ϫ) cells were incubated for 30 min at 30°C, the majority of mitochondrially associated Bax was released into the medium (Fig. 7A, lane 4). This observation indicates a loose association of Bax with mitochondria in normal Bax(ϩ/Ϫ) cells. Moreover, mitochondrially associated Bax in normal cells did not occur in oligomeric form as shown by chemical cross-linking (Fig. 7B, lane 10). Upon incubation of Bax(ϩ/Ϫ) cells with CCCP, formation of Bax oligomers was evident following chemical cross-linking, with concomitant release of cytochrome c (Fig. 7, B, lanes 11  and 12, C, lanes 5 and 6). Bak was abundant in both Bax(ϩ/Ϫ) and Bax(Ϫ/Ϫ) cells (Fig. 7B, lanes 1-6). As in the case of HeLa cells (Fig. 2B), CCCP-induced Bak oligomers were seen largely in the form of dimers after cross-linking in Bax(ϩ/Ϫ) cells (Fig.  7B, lanes 5 and 6). With ATP depletion, Bak dimers increased significantly in Bax(ϩ/Ϫ) cells, but not in Bax(Ϫ/Ϫ) cells (Fig.  7B, lanes 1-6). These results are most consistent with a role for Bax in the reorganization/oligomerization of Bak that takes place during ATP depletion. Regardless of these considerations, the results also showed that there is an absolute requirement for Bax expression to release cytochrome c in significant amounts during ATP depletion (Fig. 7C, compare lanes 2,  3 and lanes 5, 6). The small amount of cytochrome c release seen in Bax(Ϫ/Ϫ) cells can be attributed to the fragility of energy-deprived cells subjected to plasma membrane permeabilization methods. A potential role of permeability transitions in the release of cytochrome c during ATP depletion by CCCP or hypoxia was ruled out by the ability of affected mitochondria to mount potential and accumulate potentiometric dyes after removal of CCCP (Fig. 1C) or reoxygenation (24) in complete growth medium. Moreover, our results showing little or no release of cytochrome c in Bax(Ϫ/Ϫ) cells after prolonged treatment with CCCP (Fig. 7C) also rule out a role for the permeability transition in this phenomenon.
To determine whether Bax can independently oligomerize and release cytochrome c in the absence of Bak, we used transformed baby mouse kidney cells (25) derived from wild type (Bax(ϩ/Ϫ)/Bak(ϩ/ϩ)), Bax (Bax(Ϫ/Ϫ)), Bak (Bak(Ϫ/Ϫ))-deficient and Bax/Bak double deficient (Bax(Ϫ/Ϫ)/Bak(Ϫ/Ϫ)) mice. We observed progressive translocation of Bax to mitochondria with increasing durations of ATP depletion in wild type BMK cells (Fig. 8A, lanes 1-5). However, Bak(Ϫ/Ϫ) cells showed delayed kinetics of Bax translocation (Fig. 8A, lanes 6 -10). After 3 h of ATP depletion in wild type cells, the vast majority of cytosolic Bax had translocated to mitochondria and Ͼ90% of cells had released cytochrome c into the cytosol (Fig. 8B, panel  1). On the other hand, in Bak knockout cells, Bax translocation was significantly delayed (Fig. 8A, lanes 6 -10). Corresponding to this, less than 20% of cells showed cytochrome c release (Fig.  8B, panel 3). In contrast, both Bax knockout and Bax/Bak double knockout cells showed resistance to cytochrome c release (Fig. 8B, panels 2 and 4). Chemical cross-linking of membranes revealed that translocated Bax is oligomerized into dimers and higher order oligomers in energy-deprived wild type BMK cells (Fig. 8C, lane 2). Similarly treated Bak(Ϫ/Ϫ) cells also showed Bax oligomerization (Fig. 8C, lane 6). These results clearly suggest that Bax does not require Bak to oligomerize after translocation to mitochondria. However, Bak deficiency seems to affect the kinetics of Bax translocation probably because of molecular changes in mitochondria that could have occurred because of Bak deficiency. Similarly, mitochondrial Bak, normally present as monomers as well as larger protein complexes bigger than dimers, rearranged to form Bak homo-oligomeric ladders in wild type BMK cells during energy deprivation (Fig. 8D, compare lanes 1 and 2). As in the case of Bax, the calculated molecular weights of rearranged Bak complexes in these ladders correspond to dimers, trimers, and higher order homo-oligomers. In contrast, Bax knockout cells failed to show Bak rearrangement to form homooligomers even after prolonged incubation under ATP-depleted conditions (Fig. 8D, lanes 5 and 6). These results clearly sug-FIG. 6. Dissociation and reassociation of Bax and Bak. A, membrane fractions without cross-linking (ϪCL) from ATP-depleted RPTC (5 ϫ 10 6 cells) were extracted with either CHAPS (ϪSDS) or 0.35% SDS (ϩSDS) plus 1 mM DTT in buffer A. The samples extracted with SDS were heated at 70°C for 10 min to remove noncovalent interactions. The samples were renatured by diluting with 2% CHAPS buffer to a final SDS concentration of 0.035%. Diluted samples were immunoprecipitated with anti-Bak antibodies and precipitates were analyzed for Bax and Bak by Western blotting. SDS treatment increased Bak precipitation by anti-Bak antibodies, but it inhibited Bax co-precipitation. B, membranes were treated with EGS cross-linker (ϩCL) before extracting with CHAPS or SDS buffer. The SDS extracts were renatured in CHAPS as described above and subjected to immunoprecipitation with anti-Bak or Bax antibodies. The immunoprecipitates were separated on SDS-PAGE and analyzed by Western blotting. Bax was immunoprecipitated with 1D1 monoclonal antibody and probed with polyclonal anti-Bak antibody (lanes 3 and 4) followed by polyclonal anti-Bax (N-20) antibody (lanes 1 and 2) after stripping Bak antibody of the membrane. Bak was immunoprecipitated with polyclonal antibody and probed with monoclonal anti-Bax antibody (lanes 5 and 6). Mainly Bax trimers and higher oligomers are co-precipitated with Bak. All forms of Bak co-precipitated with Bax probably by associating with Bax oligomers. Small amounts of Bax-Bak dimers were seen when signal from Bax dimers is suppressed (lane 6).
gest that Bax is required to induce Bak reorganization and cytochrome c release during hypoxia or ATP depletion. DISCUSSION The release of cytochrome c from mitochondria is a crucial step in apoptotic signaling through the activation of caspases (30 -32). Several studies point to a major role for Bax in cytochrome c release based on its ability to form channels in artificial lipid membranes (33) and large oligomeric complexes in the mitochondrial outer membrane (16 -18). Although the formation of transmembrane channels by Bax oligomers is a likely explanation, the question of how Bax triggers cytochrome c release after its translocation to mitochondria continues to be debated. Here, we demonstrate that following Bax redistribution from the cytosol to mitochondria, formation of Bax oligomers in the mitochondrial outer membrane is also accompanied by reorganization of resident Bak molecules to homooligomerize. Analysis of the molecular sizes of Bax and Bak oligomers by cross-linking and SDS-PAGE indicated that both Bax and Bak oligomers are largely if not exclusively homogeneous. Our results show also that these homo-oligomeric complexes of both Bak and Bax co-precipitated with each other during immunoprecipitation.
Unlike Bax, Bak was found to be constitutively present in the mitochondria of RPTC (Fig. 1A, panel 1), in the form of large complexes by gel filtration (Fig. 3A) and chemical crosslinking (Fig. 1C). After hypoxia or ATP depletion, both Bax and Bak were co-localized in mitochondria (Fig. 1A, panel 2). Following translocation of Bax to mitochondria, Bak reorganized to form smaller oligomeric complexes that are absent in normal cells (Fig. 1, C and D). The molecular weights of Bax and Bak oligomers indicated that these complexes are homogeneous and are therefore homo-oligomers (Fig. 1D). By analogy to mechanisms that lead to pore formation by bacterial toxins, conformational changes may occur in Bax that lead to oligomerization following membrane insertion (34,35). The observation that both Bax and Bak exhibit conformational changes during hypoxia or ATP depletion raises the question whether these two proteins are causally linked to the release of apoptogenic cytochrome c and Smac proteins from mitochondria. Based on previous reports that Bak oligomerization is primarily mediated by Bid, we explored the role of Bid in hypoxia or ATP depletioninduced apoptosis. Previous studies have shown that death stimuli by Fas ligand or tumor necrosis factor-␣ activate caspase-8 to cleave Bid, a BH3 only protein, to a truncated form (t-Bid) that is targeted to mitochondria, inducing Bak to oligomerize (21). However, in both RPTC and HeLa cells, Bid did not undergo either proteolytic cleavage or mitochondrial translocation during ATP depletion by CCCP or hypoxia. Moreover, failure to prevent Bax translocation and cytochrome c release from mitochondria of ATP-depleted cells by z-VAD (24,36), a broad spectrum caspase inhibitor, is also indicative of the noninvolvement of t-Bid. Therefore, it seems likely that Bax triggers a conformational change in Bak to homo-oligomerize following mitochondrial insertion.
Protein complex formation in mitochondrial membranes was also analyzed by size sieving chromatography. We chose CHAPS to solubilize membrane proteins because of its smaller aggregation number and its inability to induce conformational change in the Bcl-2 family of proteins (16,29). Although copurification of proteins in fractions separated by gel filtration does not necessarily mean interactions between the proteins, the converse should be true. That is, proteins with tight interactions between them should co-purify. In normal cells, both Bax and Bak have distinct elution profiles after gel filtration. The broader elution profile (43-5000 kDa) of Bak in normal cells (Fig. 3) suggests possible association of Bak with a protein or protein complex in the mitochondrial outer membrane. In contrast, the elution profile of Bax, which is cytosolic in normal cells, is sharper and indicative of its monomeric nature (Fig. 3). However, the elution profiles of both Bak (ϳ43-1340 kDa) and Bax (ϳ25-1340 kDa) were found to overlap significantly in ATP-depleted cells (Fig. 3). Our immunoelectron microscopy data on translocated Bax in mitochondria of ATP-depleted cells are in agreement with results reported by Youle's group (37) in terms of the clustering of Bax on mitochondrial surfaces. However, we did not detect large clusters of Bax molecules outside  11 and 12). Similarly, slow moving Bak dimers appeared after ATP depletion in parallel with Bax oligomers (lanes 5 and 6). Note that there is a significant increase in Bak dimers only in ATP-depleted Bax(ϩ/Ϫ) cells but not Bax(Ϫ/Ϫ) cells. The identities of slow moving bands intermediate in size between monomer and dimer in control cells (lanes 1 and 4) and corresponding to trimer in both ATP-depleted Bax(Ϫ/Ϫ) and Bax(ϩ/Ϫ) cells (lanes 2, 3, 5, and 6) are not known. C, cytosolic digitonin extracts of Bax(Ϫ/Ϫ) and Bax(ϩ/Ϫ) colon cancer cells after ATP depletion (7.5 h CCCP) were analyzed for released cytochrome c (Cyt.c in cytosol) by Western blotting. the mitochondria in cells affected by ATP depletion-induced apoptosis (Fig. 3C, panel 2). The immuno-EM findings are also consistent with our previous studies (16,24) and current results showing Bax translocation, oligomerization, and cytochrome c release.
By several types of analysis including cross-linking, gel filtration, and immunoprecipitation, our results have further clarified the inhibitory effect of Bcl-2 expression on the deleterious effects of pro-apoptotic Bax and Bak. Bcl-2 prevented the relatively close association of Bax and Bak molecules with themselves to form chemically cross-linkable homo-oligomers, as well as the possibly looser association between these two different species of homo-oligomers to form larger coimmunoprecipitable complexes that could not be stabilized by crosslinking. More Bax and Bak were precipitated by their respective antibodies from ATP-depleted cells than from normal cells (Fig. 5). This can be attributed to increased antibody recognition of these antigens caused by unfolding of the COOH terminus of Bax and translocation of the protein (13,29), and dissociation of Bak from large Bak containing complexes of otherwise unknown composition, native to mitochondrial membranes (Fig. 4). Relevant to the formulation of strategies to address still unresolved questions regarding how Bcl-2 protects cells, expression of this protein prevented neither the ATP depletion-induced mitochondrial translocation of Bax (16,24), nor the associated dissolution of large Bak containing complexes (Figs. 4 and 5). However, Bcl-2 did prevent or markedly inhibited Bax and Bak homo-oligomerization and cross-immunoprecipitation of Bax and Bak that could otherwise be demonstrated easily in the absence of Bcl-2 expression (Figs. 4 and 5). The molecular interactions that underlie these remarkable results will need close attention in future studies.
Co-immunoprecipitation of Bax oligomers with Bak antibodies and Bak oligomers with Bax antibodies (Fig. 5) and the nonoverlapping nature of Bax and Bak ladders after co-precipitation (Fig. 5B) together suggest that these different homooligomeric species interact in the membrane. Interestingly, dissociation and reassociation studies have thrown light on the nature of interaction between Bax and Bak. Regardless of treatment with SDS prior to renaturation and cross-immunoprecipitation of cross-linked membranes, Bax as well as Bak containing oligomers (ladders) were largely, if not exclusively, homogeneous, i.e. they were homo-oligomeric. This suggests that molecular interactions between individual monomers in either Bax or Bak ladders were close, to the extent that they could be stabilized by cross-linking. The failure to cross-link proteins does not necessarily mean they are not bound to each other. Steric factors related to the chemical nature of the cross-  1 and 2) and Bak-deficient cells (lanes 5 and 6) shown by immunoblotting of EGS cross-linked membrane proteins after ATP depletion (4 h CCCP). Bax-deficient cells do not show any signal (lanes 3 and 4). D, chemical cross-linking and immunoblotting show anti-Bak reactive slow moving Bak adducts in control wild type BMK cells. Note that these adducts disappear and new Bak complexes that resemble multimeric Bak molecules are formed during ATP depletion (4 h CCCP) in wild type cells (compare lanes 1 and 2). Bax-deficient cells (Bax(Ϫ/Ϫ)/Bak(ϩ/ϩ)) failed to show Bak reorganization during the same length of exposure to ATP depletion (4 h CCCP; compare lanes 5 and 6).
linkers and/or protein interactions are involved in cross-linking of proteins, and negative results need to be interpreted with caution. This was addressed by using a nonspecific heterobifunctional cross-linker SANPAH (N-succinimidyl-6-[4Ј-azido-2Ј-nitrophenylamino] hexanoate), which cross-links proteins through an amine reactive and a photoactivable nitrene that reacts nonspecifically with any atom within the reach of the spacer arm, and we obtained predominantly homo-oligomers (16). On the other hand, interactions between Bax and Bak must have been relatively loose, because they permitted coimmunoprecipitation, but could not be stabilized by cross-linking to any significant extent (for the exception see Fig. 6B, lane 6). If Bax-Bak interactions had been tight to the degree that both molecules contributed to the formation of hetero-oligomeric complexes (Bax-Bak pore), Bax and Bak ladders resolved by SDS-PAGE should have shown overlapping patterns. Obviously, this was not the case.
Our results show that trimers and higher order cross-linked oligomers of Bax reassociate with Bak during CHAPS renaturation after SDS denaturation to a much greater extent than do Bax-Bax dimers or Bax monomers (Fig. 6B, lane 6). The appearance of Bax-Bak heterodimers only as a minor component of these reassociated complexes (Fig. 6B, lane 6) further reinforces the concept that pores in the membrane are mainly constituted of homo-oligomeric Bax and Bak. Considered in their entirety, our results suggest that interaction between Bax and Bak takes place after Bax has undergone conformational change to oligomerize. Therefore, Bax oligomerization probably precedes Bak reorganization and oligomerization.
Gene knockout studies with Bax, Bak, and Bax plus Bak have indicated that cells with single gene knockout are capable of releasing cytochrome c, but those with double knockout are not (10,25). Indeed, mice deficient of both Bax and Bak died perinatally and suffered multiple developmental defects that were more severe than in animals deficient for only Bax or Bak (9). This finding was attributed to suppression of developmentally regulated apoptosis caused by the double knockout. It has been reported that Bak knockout mice are developmentally normal and reproduce normally (9), whereas Bax knockout mice are reproductively defective and display some phenotypic abnormalities (38,39). With respect to our findings, one possibility is that Bax and Bak are capable of forming pores independent of each other. On the other hand, Bak may require Bax or a similar molecule to form pores optimally. The latter argument is based on the observations that some colon cancer cells do not exhibit redundancy with respect to the apoptogenic roles of Bak and Bax (26,40). A third explanation may involve formation by both Bak and Bax of pores larger than those formed by either Bax or Bak alone. Because heteromeric Bax-Bak complexes were not seen after chemical cross-linking, our studies do not favor a hetero-oligomeric pore, at least in the hypoxic model of apoptosis in kidney epithelial, HeLa, and HCT116 cells. Research to date suggests that Bak reorganization to form homo-oligomers requires another Bcl-2 family protein such as t-Bid (21) or Bax (current study). In unpublished work, 2 we have investigated the permeabilization of isolated mitochondria from Bax(Ϫ/Ϫ) and Bax(ϩ/Ϫ) cells exposed to recombinant t-Bid in vitro. The results showed that release of cytochrome c and Smac by t-Bid from isolated mitochondria has an absolute requirement of Bax. Bak alone was not effective. 2 In the current work with cells selectively deficient in Bax or Bak, we demonstrate existence of complex relationships between Bax and Bak. In the absence of Bax, Bak alone not only failed to undergo reorganization but also failed to induce the release of cytochrome c upon ATP depletion (Figs. 8, B and D  and 7C). On the other hand, homo-oligomerization of Bax or cytochrome c release does not require Bak (Fig. 8C). However, optimal Bax translocation does seem to need the presence of Bak (Fig. 8A). Whether this should be attributed to Bak alone or a Bak associated common target for Bax and Bak needs to be determined.