Assembly of the Bak Apoptotic Pore

Background: Bak and Bax form pores that damage mitochondria during apoptosis. Results: Under native conditions, the basic oligomeric unit of Bak is a BH3:groove homodimer that requires the α6 helix to multimerize and mediate cell death. Conclusion: Bak forms pores via two independent requisite interfaces. Significance: Characterizing Bak pores will facilitate inhibition of damaging apoptosis, e.g. in ischemic stroke. Bak and Bax are the essential effectors of the intrinsic pathway of apoptosis. Following an apoptotic stimulus, both undergo significant changes in conformation that facilitates their self-association to form pores in the mitochondrial outer membrane. However, the molecular structures of Bak and Bax oligomeric pores remain elusive. To characterize how Bak forms pores during apoptosis, we investigated its oligomerization under native conditions using blue native PAGE. We report that, in a healthy cell, inactive Bak is either monomeric or in a large complex involving VDAC2. Following an apoptotic stimulus, activated Bak forms BH3:groove homodimers that represent the basic stable oligomeric unit. These dimers multimerize to higher-order oligomers via a labile interface independent of both the BH3 domain and groove. Linkage of the α6:α6 interface is sufficient to stabilize higher-order Bak oligomers on native PAGE, suggesting an important role in the Bak oligomeric pore. Mutagenesis of the α6 helix disrupted apoptotic function because a chimera of Bak with the α6 derived from Bcl-2 could be activated by truncated Bid (tBid) and could form BH3:groove homodimers but could not form high molecular weight oligomers or mediate cell death. An α6 peptide could block Bak function but did so upstream of dimerization, potentially implicating α6 as a site for activation by BH3-only proteins. Our examination of native Bak oligomers indicates that the Bak apoptotic pore forms by the multimerization of BH3:groove homodimers and reveals that Bak α6 is not only important for Bak oligomerization and function but may also be involved in how Bak is activated by BH3-only proteins.

Bak and Bax are the essential effectors of the intrinsic pathway of apoptosis. Following an apoptotic stimulus, both undergo significant changes in conformation that facilitates their self-association to form pores in the mitochondrial outer membrane. However, the molecular structures of Bak and Bax oligomeric pores remain elusive. To characterize how Bak forms pores during apoptosis, we investigated its oligomerization under native conditions using blue native PAGE. We report that, in a healthy cell, inactive Bak is either monomeric or in a large complex involving VDAC2. Following an apoptotic stimulus, activated Bak forms BH3:groove homodimers that represent the basic stable oligomeric unit. These dimers multimerize to higher-order oligomers via a labile interface independent of both the BH3 domain and groove. Linkage of the ␣6:␣6 interface is sufficient to stabilize higher-order Bak oligomers on native PAGE, suggesting an important role in the Bak oligomeric pore. Mutagenesis of the ␣6 helix disrupted apoptotic function because a chimera of Bak with the ␣6 derived from Bcl-2 could be activated by truncated Bid (tBid) and could form BH3:groove homodimers but could not form high molecular weight oligomers or mediate cell death. An ␣6 peptide could block Bak function but did so upstream of dimerization, potentially implicating ␣6 as a site for activation by BH3-only proteins. Our examination of native Bak oligomers indicates that the Bak apoptotic pore forms by the multimerization of BH3:groove homodimers and reveals that Bak ␣6 is not only important for Bak oligomerization and function but may also be involved in how Bak is activated by BH3-only proteins.
Bak and Bax are the pivotal effectors of the intrinsic apoptotic pathway because either is required for perturbation of the mitochondrial outer membrane (MOM) 3 to release cytochrome c, which consequently activates caspases (1). How Bak and Bax damage mitochondria is unclear. Studies indicate that Bak and Bax need to oligomerize to form either a proteinaceous or a lipidic pore in the MOM (2)(3)(4)(5)(6)(7)(8)(9)(10). Deciphering how Bak and Bax become activated and the molecular interactions involved in their self-association is pivotal in understanding how these proteins permeabilize the MOM to kill a cell.
In a healthy cell, inactive Bax is monomeric and predominantly cytosolic because of its trafficking away from mitochondria and its ability to sequester its C-terminal transmembrane domain in its hydrophobic groove (11)(12)(13). In contrast, inactive Bak is constitutively integrated in the MOM as an inactive monomer or is restrained by interactions with other proteins, including Bcl-x L , Mcl-1, or voltage-dependent anion channel 2 (VDAC2) (14 -16).
Following apoptotic stress, both Bak and Bax undergo changes in conformation as they adopt their active oligomeric forms (17)(18)(19). The stepwise activation of Bak and Bax is currently being delineated. However, there remains a paucity of information regarding the molecular composition of the oligomeric pore formed by Bak and Bax that permeabilizes the MOM. Although liposome studies suggest that a minimum of four molecules is necessary to mediate cytochrome c release (20), how many Bak or Bax molecules are required to constitute a functional pore in cells is unknown, with oligomeric complexes comprising potentially hundreds of Bax molecules described in dying cells (21). An important change in Bak and Bax implicated in our studies is the exposure of the BH3 domain (2,22). Cross-linking studies indicate that the exposed BH3 domain inserts into the hydrophobic groove of a partner mole-cule to form a homodimer (2,22,23). However, the symmetrical nature of the BH3:groove homodimer necessitates an additional interface, independent of both the BH3 domain and groove, to form the higher-order oligomers thought responsible for damaging the MOM. Because the ␣6 helix of Bak (and Bax) has an induced proximity during apoptosis, we proposed that an ␣6:␣6 interface may represent this secondary interface (22,24). In an alternative model, Bak/Bax oligomerize via a repeated asymmetric interface involving interaction of the BH3 domain from one molecule with a site involving the ␣6 or "rear pocket" on a partner molecule (25)(26)(27). In silico modeling suggested that this nose-to-tail association allows the formation of an octameric pore with a sufficient diameter to allow the passage of cytochrome c (28).
Blue native PAGE (BN-PAGE) has proven effective in characterizing mitochondrial complexes, including respiratory complexes, import complexes, and those formed by Bax (19,29,30). Here we used BN-PAGE to examine Bak oligomerization both before and during apoptosis. Under these native conditions, the BH3:groove homodimer was stable and represented the basic oligomeric unit of Bak. The core dimerization domain (␣2-␣5) (8) was not sufficient for Bak to mediate cell death because ␣6 was important for mediating the formation of high molecular weight oligomers and, thereby, apoptotic function. Our studies of native Bak complexes support the notion that the Bak apoptotic pore involves higher-order multimers of homodimers.

EXPERIMENTAL PROCEDURES
Immunoblotting Antibodies-For Western blotting, Bak on SDS-PAGE was detected using an anti-Bak polyclonal rabbit IgG (amino acids 23-38, catalog no. B5897, Sigma) or on native-PAGE with an anti-Bak monoclonal rat IgG (7D10, D. Huang, Walter and Eliza Hall Institute). Cytochrome c was detected with a monoclonal mouse IgG (clone 7H8.2C12, BD Biosciences) and VDAC with a rabbit polyclonal (Ab-2, Calbiochem). Secondary antibodies used were horseradish peroxidase-conjugated anti-rabbit IgG, anti-mouse IgG, and anti-rat IgG (Southern Biotech).
Cell Lines, Cell Culture, and Induction of Apoptosis-Mouse embryonic fibroblasts (MEFs) derived from Bak Ϫ/Ϫ Bax Ϫ/Ϫ C57/ Bl6 mice were transformed with SV40 large T and cultured in Dulbecco's Modified Eagles medium supplemented with 10% fetal calf serum, 250 M L-asparagine and 55 M 2-mercaptoethanol as described (2). Apoptosis was induced in MEFs by treating with etoposide (10 M) for 24 h, and cell death was assessed by flow cytometric analysis of propidium iodide uptake.
PCR Mutagenesis-Bak mutants were generated by site-directed PCR mutagenesis (primer sequences are available on request) and cloned into the retroviral expression vector pMX-IRES-GFP (internal ribosome entry site-GFP) as described previously (2). Bak mutants were retrovirally expressed in Bak Ϫ/Ϫ Bax Ϫ/Ϫ MEFs using Phoenix ecotropic packaging cells, and infected cells were selected on the basis of GFP expression.
Subcellular Fractionation and Cytochrome c Release-MEFs were harvested and permeabilized in buffer (20 mM Hepes (pH 7.5), 100 mM KCl, 2.5 mM MgCl 2 , and 100 mM sucrose) containing 0.025% digitonin and supplemented with complete protease inhibitors without EDTA (Roche). Permeabilization was confirmed by trypan blue uptake, and cytosol and membrane fractions were separated by centrifugation at 13,000 ϫ g for 5 min prior to SDS-PAGE analysis. To assess cytochrome c release, pelleted membrane fractions were resuspended in permeabilization buffer without digitonin and incubated with or without caspase 8-cleaved human Bid (tBid, 20 nM) at 30°C for 30 min. Supernatant and membrane fractions were separated by centrifugation at 13,000 ϫ g for 5 min prior to SDS-PAGE.
Disulfide Linkage of Bak Complexes-To examine Bak complexes before and after an apoptotic stimulus, following treatment of membrane fractions with or without tBid, disulfide linkage of Bak was induced by incubating with the redox catalyst copper (II)(1,10-phenanthroline) 3 (CuPhe, 1 mM) on ice for 30 min as described (2). Samples were then run under nonreducing SDS-PAGE or BN-PAGE where indicated.
Blue Native PAGE and Antibody Gel Shift-BN-PAGE was performed essentially as described (31). Membrane fractions were treated with tBid and, where indicated, with CuPhe. For antibody gel shift, membranes were incubated with 2 g of the indicated monoclonal antibody either after or during incubation with tBid. Membranes were pelleted and then solubilized in 20 mM Bis-Tris (pH 7.4), 50 mM NaCl, 10% glycerol, 1% digitonin with or without 10 mM DTT before centrifugation at 13,000 ϫ g to pellet insoluble debris. When performing gel shift or when in combination with induced disulfide linkage, DTT was omitted from the solubilization buffer. BN-PAGE loading dye (5% Coomassie Blue R-250 (Bio-Rad) in 500 mM 6-aminohexanoic acid, 100 mM Bis-Tris (pH 7.0)) was then added to each sample. Gels were electrophoresed in anode buffer (50 mM Bis-Tris (pH 7.0)) and blue cathode buffer (50 mM N-[2-hydroxy-1,1-bis(hydroxymethyl)ethyl]glycine and 15 mM Bis-Tris unbuffered containing 0.02% Coomassie Blue G-250). Blue cathode buffer was replaced with clear buffer (without Coomassie Blue) when the dye front was one-third of the way through the resolving gel. Gels were transferred to PVDF in Tris-glycine transfer buffer containing 20% methanol and 0.037% SDS. Blots were destained in 50% methanol and 25% acetic acid and washed with TBS prior to immunoblotting with anti-Bak 7D10 or VDAC (Ab-2).
Clear Native-PAGE (CN-PAGE)-CN-PAGE was performed essentially as described (32). Briefly, digitonin-solubilized membranes generated as above were supplemented with 0.01% Ponceau S prior to electrophoresis. Instead of Coomassie Blue, 0.05% deoxycholate and 0.01% n-dodecyl ␤-D-maltopyranoside were added to the cathode buffer.
Limited Proteolysis-Membrane fractions were isolated and incubated with recombinant tBid as described above but with protease inhibitors omitted. Samples were then prechilled on ice prior to incubation with proteinase K (30 g/ml) for 20 min on ice. Proteinase K was quenched with 100 mM PMSF prior to boiling samples in SDS-PAGE sample buffer. Cleavage fragments were detected by immunoblotting with an internal antibody recognizing the Bak BH3 domain (4B5) (2).

Bak Is Monomeric or in a Large Complex Involving VDAC2 in
Healthy Cells-When examined by BN-PAGE, endogenous Bak in Bax Ϫ/Ϫ MEFs or human Bak ectopically expressed in Bak Ϫ/Ϫ Bax Ϫ/Ϫ MEFs were observed predominantly in a low molecular weight (Ϸ60 kDa) form and a large molecular weight complex (Ϸ400 kDa) (Fig. 1A), as noted by others (22,34,35). Consistent with previous reports implicating an interaction of Bak with VDAC2 (15,16,36), the large Bak complex comigrated with a complex containing VDAC (Fig. 1A) and has been shown previously to be absent in Vdac2 Ϫ/Ϫ MEFs (16). To confirm that this large complex indeed contained both Bak and VDAC2, we expressed FLAG-VDAC2 in Vdac2 Ϫ/Ϫ MEFs and performed an antibody gel shift assay (Fig. 1B). It was necessary to express a FLAG-tagged form to circumvent the lack of antibodies that specifically and efficiently detect VDAC2. The Bakand FLAG-immunoreactive complexes both gel-shifted with an anti-FLAG antibody, confirming that the large complex contains both Bak and VDAC2 (15,16). The low molecular weight form was monomeric Bak because its migration was not affected by denaturation with urea, in contrast to the complex with VDAC2 that was dissociated completely (Fig. 1C).
Because certain detergents can cause conformation change and oligomerization of Bak and Bax, we tested whether digitonin, the detergent of choice to retain supramolecular membrane complexes on BN-PAGE (29,30), might also affect Bak conformation. Digitonin did not cause aberrant conformation change, as indicated by retention of the intramolecular disulfide linkage between the endogenous cysteines that is diagnostic of the inactive fold of Bak (M X , Fig. 1D) (15). Consistent with this, digitonin, but not other detergents tested, could discriminate clear Bak complexes and, importantly, changes in Bak complexes in response to an apoptotic stimulus on BN-PAGE (data not shown). 1% w/v digitonin was sufficient to extract most of the Bak from the MOM (Fig. 1E), and similar complexes were detected when mitochondria were solubilized with increasing digitonin concentration or at increasing temperature (data not shown), suggesting that the solubilization conditions accurately reflected the distribution of Bak between the VDAC2 complex and monomer. Therefore, native PAGE indicates that, in healthy cells, Bak is monomeric unless in complex with VDAC2.
Bak Forms a Single Stable Oligomeric Species during Apoptosis-Following treatment of cells with etoposide ( Fig.  1F) or membrane fractions with tBid (A and G), Bak dissociated from the VDAC2 complex and formed two predominant complexes (designated ␣ and ␤) that correlated with cysteine linkage of Bak homo-oligomers on non-reducing SDS-PAGE, cytochrome c release (G), and cell death (F). These two Bak complexes were even more clearly resolved by highresolution clear native PAGE (Fig. 1H) and were consistent with those observed during TNF-induced apoptosis of HeLa cells (34). However, when membranes were solubilized in the presence of DTT to prevent disulfide linkage or when a Bak variant lacked its N-terminal cysteine (C14S) ( Fig. 1H and data not shown), Bak migrated only as the ␣ form. Thus, the ␤ form was a consequence of solubilization and BN-PAGE under non-reducing conditions, leading to disulfide linkage of proximal N-terminal cysteines, but is unlikely to form in cells because of the reducing environment of the cytosol. Together, these data indicate that the Bak oligomer formed during apoptosis resolves as a stable single species on BN-PAGE.
Bak BH3:Groove Homodimers Are the Stable Oligomeric Unit Formed during Apoptosis-Bak activation involves exposure of its N terminus and BH3 domain (2,37). To investigate the conformation of Bak in its native complexes in healthy or apoptotic cells, we performed a gel shift analysis with conformation-specific antibodies recognizing the Bak N terminus (23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37)(38) or BH3 domain (4B5). As a control, an antibody that recognizes both inactive and active human Bak (7D10) (24) gel-shifted all forms of Bak both before and after tBid treatment ( Fig. 2A, third and fourth lanes). The N-terminal antibody gel-shifted the nascent Bak complex induced by tBid but failed to gel-shift Bak in untreated mitochondria either in its monomeric form or when associated with VDAC2 ( Fig. 2A), supporting Bak as inactive in healthy cells even when complexed with VDAC2 (37). The BH3 domain antibody only gel-shifted Bak when the antibody was added during the tBid incubation and not before or after, indicating that the BH3 domain was exposed during Bak activation but reburied upon oligomerization (Fig. 2B).
To further characterize the apoptotic Bak oligomer, we analyzed three loss-of-function Bak mutants that fail to self-associate following an apoptotic stimulus (2). Like wild-type Bak, the mutants dissociated from VDAC2 in response to an apoptotic stimulus ( Fig. 2C) but did not form the nascent Bak oligomer or mediate cytochrome c release even after extended incubation with tBid (60 min, C). Moreover, the loss-of-function mutants gel-shifted with 4B5 when the antibody was added after tBid treatment, indicating that the BH3 domain became exposed but was not reburied, further supporting the involvement of the BH3 domain in the oligomer interface (Fig. 2D, lanes 5 and 6). The D83G mutant failed to gel-shift with 4B5 because the mutation abrogated binding of the antibody (data not shown) (2).
To test whether the stable oligomeric form of Bak comprised Bak BH3:groove homodimers, we employed BN-PAGE in combination with disulfide linkage. Following tBid-induced oligomerization, induced disulfide linkage between cysteines introduced into the Bak BH3 domain (M71C) and hydrophobic groove (K113C) allows only a single complex to be trapped that can be accurately sized as a dimer on non-reducing SDS-PAGE (data not shown) (2). Notably, disulfide linkage of the BH3: groove interaction in Bak M71C/K113C did not alter the migration of the nascent Bak oligomer on BN-PAGE (Fig. 2E, compare lanes 2 and 3), indicating that the oligomer indeed comprised Bak in a BH3:groove conformation. Furthermore, because oxidation of the oligomer formed by the M71C/K113C variant, but not that formed by a Bak Cys-null variant, prevented urea denaturation (Fig. 2E, compare lane 4 with lane 5 and lane 9 with lane 10), the stable oligomer did not involve interfaces outside of the BH3:groove interaction. Together, our data indicate that, during apoptosis, Bak forms a BH3:groove homodimer under native conditions that correlates with mito- chondrial perturbation and is stable upon solubilization from the MOM.
Bak ␣6 Peptide Blocks Bak Oligomerization and Function-Although the stoichiometry of the Bak oligomers required to permeabilize the MOM is unknown, a Bak BH3:groove homodimer would likely be insufficient to mediate mitochondrial permeabilization. To determine which regions of Bak may constitute the requisite second interface for the multimerization of BH3:groove dimers, we screened an overlapping array of 15 amino acid peptides derived from mouse Bak (spanning amino acids 1-185) for their ability to inhibit Bak pore formation. A peptide spanning ␣6 most efficiently inhibited Bak-mediated cytochrome c release from mouse liver mitochondria (MLM, Fig. 3A) and also blocked Bak function in Bax Ϫ/Ϫ MEF membrane fractions (Fig. 3B). Peptides corresponding to ␣1 and ␣5 also perturbed cytochrome c release, albeit to a lesser extent than the ␣6 peptide (Fig. 3A). A peptide incorporating ␣2/BH3 residues of Bak (amino acids 61-75) also slightly inhibited cytochrome c release in mouse liver mitochondria, consistent with the role of the BH3 domain in mediating the BH3: groove interaction (Fig. 3A) (2). Consistent with the blockade in cytochrome c release, the ␣6 peptide inhibited Bak oligomerization (Fig. 3C). However, the ␣6 peptide reduced the conversion of monomer to dimer and inhibited dissociation from the VDAC2 complex (Fig. 3C), suggesting that the ␣6 peptide interfered with Bak activation rather than downstream higher-order oligomer formation. This observation is consistent with reports that Bax can be activated by Bim via a binding site involving Bax ␣6 (38,39) and supports that a similar mechanism may be involved in Bak activation by tBid.
The ␣6:␣6 Interface Stabilizes Bak Higher-order Oligomers-Inhibition of Bak apoptotic function with an ␣6 peptide suggested that the ␣6 helix could be an important site for Bak oligomerization, consistent with our previous disulfide linkage studies (24). If ␣6 was involved in the multimerization of Bak dimers, we hypothesized that linkage at this interface should be able to link native BH3:groove dimers to higher-order complexes. Indeed, a single disulfide linkage between ␣6 helices (H164C:H164C) was sufficient to trap higher-order Bak com-plexes on BN-PAGE (Fig. 4, A and B). As before (Fig. 2E), disulfide linkage of the BH3:groove interaction did not alter migration of the dimer (Fig. 4, A and B). This indicates that the ␣6 helices are proximal in multimers of the stable BH3:groove homodimer but that this induced proximity is labile and so is disrupted upon detergent solubilization of the MOM. Consistent with this, when tBidinduced Bak oligomers were solubilized with detergent (Triton X-100) prior to the induction of disulfide linkage and analysis on non-reducing SDS-PAGE, ␣6:␣6 linkage of higher-order oligomers (M71C/K113C/H164C) was inhibited, whereas linkage of the BH3:groove dimer (M71C:K113C) was unaffected (Fig. 4C). Furthermore, the tBid-induced proximity of cysteines introduced at other positions in the ␣6, but not at two exposed positions in the N terminus, was prevented by detergent (Fig.  4D). That the proximity of the N termini persisted following detergent solubilization suggests that the N termini are proximal because of the BH3:groove interaction. These data indicate that the proximity of the ␣6 helices is disrupted following solubilization of the MOM.
Having shown that ␣6:␣6 linkage (H164C:H164C) was sufficient to trap higher-order Bak oligomers on BN-PAGE (Fig. 4B, lane 8), we tested other ␣6 mutants for their ability to link Bak oligomers. This indicated a periodicity in the efficiency of linkage, with cysteine introduced at Phe-157 and His-164 most able to trap high molecular weight complexes, placing these residues at the oligomer interface (Fig. 4E). The efficiency of linkage also suggests that the ␣6 helices may exhibit a parallel rather than an antiparallel association in a membrane-integrated Bak pore. Disulfide linkage of the N terminus at positions Gly-4, Cys-14 and Ser-23 (Fig. 4E) caused faster migration of the native BH3:groove homodimer, consistent with proximity within the dimer rather than between dimers (as with the ␤ form in Fig. 1H). However, linkage of higher-order oligomers was also detectable. This promiscuity in being able to link within a dimer or between dimers suggests that, following its exposure, the N terminus is flexible and mobile in a BH3:groove dimer, with the flexibility likely afforded by the long loop between the ␣1 and ␣2/BH3 domains. As a control, Cys introduced at an exposed position in the ␣4, Asn-124, failed to link either within or between dimers. Therefore, stabilizing the induced proximity of the ␣6 helices was sufficient for multimerization of native BH3:groove dimers.
Bak ␣6 -8 Are Important for Mitochondrial Targeting and Apoptotic Function-A previous study on Bax suggested that ␣6 was not necessary for oligomerization or apoptotic function because Bax ␣-helices 2-5 (comprising the BH3 and hydrophobic groove) targeted to mitochondria via the transmembrane ␣9 helix were sufficient for oligomerization and induction of cell death (8). To test this in Bak, we tested the function of two deletion mutants, ⌬6 -8 or ⌬1/6 -8 (Fig. 5A) following stable expression in Bak Ϫ/Ϫ Bax Ϫ/Ϫ MEFs. These mutants were not constitutively active because the MEFs stably expressing these mutants expressed GFP from the bicistronic vector (data not shown). In addition, neither deletion mutant mediated apoptosis in response to etoposide, despite stable expression comparable with the level of wild-type Bak (Fig. 5B). Subcellular fractionation indicated that a failure to be targeted to mitochondria likely contributed to the loss of function of the deletion mutants ( Fig. 5C). Because the Bak ␣9 transmembrane domain is sufficient to target GFP to membranes (40), the mitochondrial targeting defect was likely a consequence of improper protein folding. On BN-PAGE, Bak⌬6 -8 in the membrane was already evident as a dimer prior to an apoptotic stimulus (Fig. 5D). After treatment with etoposide, Bak⌬6 -8 expression was lost, suggesting that the protein is somewhat unstable. Therefore, Bak lacking its ␣6 -8 helices could form a dimer, yet this dimer could not mediate cell death. Thus, the Bak ␣6 -8 helices are not necessary for dimerization but are important for apoptotic function via stabilizing the protein and facilitating mitochondrial localization.
A Chimera of Bak with the ␣6 from Bcl-2 Can Form BH3: Groove Dimers but Cannot Form Higher-order Oligomers or Mediate Apoptosis-Because deletion of ␣6 -8 resulted in defects in Bak protein stability and mitochondrial localization, a more conservative mutagenesis approach was used to examine the role of ␣6 in mediating higher-order oligomerization. Because Bak and Bax are distinct from their prosurvival relatives in their ability to multimerize to high molecular weight oligomers, we hypothesized that if Bak ␣6 mediated multimerization of homodimers, then a chimera of Bak with the Bcl-2 ␣6 would lack apoptotic function. We generated chimeric proteins of Bak with substitutions of all or part of Bcl-2 ␣6 (Fig. 6A). When stably expressed in Bak Ϫ/Ϫ Bax Ϫ/Ϫ MEFs, the partial ␣6 swaps (⌬N, ⌬C, and ⌬N/C) could still target mitochondria and mediate cytochrome c release in response to tBid and cell death in response to etoposide (Fig. 6, B-D). These Bak/Bcl-2␣6 chimeras oligomerized in response to an apoptotic stimulus, consistent with their apoptotic function (Fig. 6D).
In contrast, swapping the entire Bcl-2 ␣6 resulted in loss of function, as indicated by failure to mediate cell death and cytochrome c release despite expression comparable with the functional mutants and an ability to target mitochondria (Fig. 6,  A-D). The ␣6 helix of Bax has been implicated in its interaction with Bim (38) and, potentially, for tBid activating Bak (Fig. 3C). Therefore, to test whether the loss of function of the Bak/Bcl-2␣6 was due to impaired direct activation by BH3-only proteins, we tested the function of the Bak variants in response to a BH3-only independent stimulus: heat. Incubating mitochondria at 43°C can activate Bak (41), forming oligomers similar to those induced by tBid (Fig. 6D). Bak/Bcl-2␣6 failed to release cytochrome c upon heat treatment, consistent with a role for ␣6 in Bak apoptotic function independent of BH3-only proteinmediated activation (Fig. 6D).
BN-PAGE of the mitochondrial population revealed that, similar to Bak lacking ␣6 -8 (Fig. 5D), Bak/Bcl-2␣6 was significantly dimeric prior to an apoptotic stimulus, suggesting that a proportion of the protein was constitutively activated (Fig. 6D). Consistent with this, Bak/Bcl-2␣6 exhibited a reduced association with the VDAC2 complex (Fig. 6D) and a proteolysis profile consistent with at least a proportion being constitutively activated (E). Prior to an apoptotic stimulus, neither the monomer nor the dimer of Bak/Bcl-2␣6 gel-shifted with the BH3 domain antibody 4B5 (Fig. 6F, lane 3), whereas only the dimer gel-shifted with the N-terminal antibody 23-38 (lane 4). This indicates that the BH3 domain is buried in the monomer and dimer but that the N terminus is exposed in the dimer, consistent with the BH3:groove homodimers formed by wild-type Bak during apoptosis (Fig. 2, A and B). That the monomeric population of the chimera did not gel-shift with either antibody supports that the chimera is properly folded in its monomeric form. Introduction of a loss-of-function mutation in the BH3 domain, I81T, prevented the constitutive dimerization of the chimera as it does wild-type Bak after an apoptotic stimulus (Figs. 2C and 6G). An M71C/K113C variant of the chimera also disulfide-linked before an apoptotic stimulus (Fig. 6H). Together, the data indicate that Bak/Bcl-2␣6 can form a BH3: groove homodimer in untreated cells. Following tBid treatment, BakBcl-␣6 became activated, as indicated by proteolysis ( Fig. 6E), and the monomeric form of the chimera gel-shifted with both 4B5 and 23-38 (with antibody added during the tBid incubation). This indicates that the chimera could still be activated by tBid to expose its BH3 domain and N terminus (Fig. 6F,  lanes 7 and 8).
To test whether the loss of function of the BakBcl-2␣6 chimera correlated with a failure to form higher-order oligomers, we tested whether disulfide linkage could trap higher-order oligomers of the BakBcl-2␣6 chimeras on BN-PAGE. To do this, we introduced a cysteine into the Bcl-2 ␣6 at Arg-183 (residue numbering on the basis of Bcl-2) at an equivalent position to Bak H164, linkage at which traps higher-order oligomers on BN-PAGE (Figs. 4E and 6I). Without induction of disulfide linkage, the R183C mutants migrated similarly to their parent proteins on BN-PAGE (Fig. 6, D and I). When combined with disulfide linkage, high molecular weight complexes were observed in the functional chimera, Bak/Bcl-2␣6 ⌬N . In contrast, the loss-of-function chimera Bak/Bcl-2␣6 failed to link to higher-order complexes but, rather, was limited to a tetramer (Fig. 6I). Therefore, the Bak/Bcl-2␣6 chimera retained an ability to expose its BH3 domain and to form BH3:groove dimers but could not form high molecular weight oligomers or mediate cell death.

DISCUSSION
How Bak and Bax self-associate to form the pores that damage mitochondria and kill the cell remains unclear. An interaction between the BH3 domain and a rear pocket on Bax involving its ␣6 helix has been implicated (25,27), supporting an asymmetrical "nose-to-tail" oligomerization of Bax (26). Furthermore, recent modeling of a Bak octamer involved the interaction of monomers via a repeated asymmetric interface (28). However, our examination of Bak oligomers under native conditions indicated that the basic oligomeric unit of Bak is a homodimer. This observation excludes such an asymmetric model of pore formation. Instead, our data indicate that Bak pores form via at least two distinct interfaces. Upon activation, Bak forms symmetrical BH3:groove homodimers that represent the basic oligomeric unit. The homodimer is stable following isolation from the MOM, likely because of the reciprocal BH3:groove interaction. An interface involving Bak ␣6, independent of both the BH3 and groove, allows homodimers to multimerize. Disulfide linkage of this second interface is sufficient to stabilize large Bak oligomers, and mutagenesis supports an important role for ␣6 in Bak apoptotic function. This model of Bak/Bax pore formation is supported by our recent studies showing that Bax can also adopt a symmetrical BH3:groove structure and that Bax ␣6 is juxtaposed during apoptosis (22,42).
Upon activation, Bak and Bax undergo a major reconfiguration, including exposure of the N terminus (37), exposure of the BH3 domain (2), and dissociation of ␣1-5 ("core") from ␣6 -9 ("latch") (42). Further, although not shown for Bak, Bax inserts its ␣5/6 helices into the MOM during activation (43). Our proposed model of Bak oligomerization is potentially consistent with such structural changes. Our gel shift experiments of the native Bak dimer indicate that it forms downstream of both N-terminal and BH3 domain exposure, with the exposure of the BH3 domain necessary for the consequent proximity with the groove of a partner Bak molecule. Activation-induced movement of the Bak N terminus (37) or dissociation of the core and latch domains, as reported for Bax (42), could expose ␣6 to facilitate higher-order oligomerization. Because our data indicate that a BH3:groove homodimer persists when mitochondria have completely released cytochrome c, presumably the ␣5/6 can insert into the MOM without significantly disrupting the groove. The reciprocity of the BH3:groove interaction involving ␣2-4 may be able to compensate for extrusion of ␣5 in this context.
It has been reported that the Bak N terminus is constitutively exposed analogous to activated and translocated Bax (39). Our gel-shift/BN-PAGE analysis using digitonin, a detergent that does not alter Bak conformation, indicates that prior to an apoptotic stimulus, Bak is in an inactive conformation whether associated with VDAC2 or monomeric and that during apoptosis, the Bak N terminus becomes permanently exposed to be a marker of Bak in its lethal conformation.
We show that Bak lacking ␣-helices 6 -8 can still form dimers, consistent with ␣2-5 (incorporating both BH3 and groove) being the core dimerization domain. However, this mutant lacks apoptotic function when stably expressed in Bak Ϫ/Ϫ Bax Ϫ/Ϫ MEFs, with loss of function likely because of reduced targeting to mitochondria and/or protein instability. In Bax, it has been shown that ␣-helices 2-5/9 is the minimum domain required for oligomerization and apoptotic function because mutants lacking ␣-helices 1, 6, 7, and 8 were constitutively active upon transient transfection in Bak Ϫ/Ϫ Bax Ϫ/Ϫ MEFs (8). This suggested that the induced proximity of the Bax ␣6 observed during apoptosis is not critical for Bax apoptotic function (22,25). The difference between Bak and Bax may suggest that they employ different mechanisms of pore formation. However, in our hands, two mutants of Bax lacking ␣-helices 6 -8 did not mediate cell death in Bak Ϫ/Ϫ Bax Ϫ/Ϫ MEFs because of failure to stably express (data not shown). As with the Bak deletion mutants, cells infected with Bax⌬6 -8 constructs remained GFP-positive from the bicistronic vector, indicating that the failure to detect protein was not because Bax⌬6 -8 is constitutively active and inducing cell death (data not shown). The reason for the conflicting findings is unclear but may relate to the previous study employing transient overexpression of Bax mutants with an N-terminal GFP fusion that may stabilize an otherwise unstable protein. Nevertheless, our data indicate that the core ␣2-5 helices of Bak are sufficient for dimerization but that the ␣6 -8 helices are essential for Bak function.
That a chimera of Bak with an ␣6 derived from Bcl-2 can, like wild-type Bak, be activated by tBid to expose its BH3 domain, can form BH3:groove homodimers, and yet cannot mediate cytochrome c release and cell death support that BH3:groove dimerization is not sufficient for apoptosis. The ␣5/6 region of Bax has been shown to insert into membranes as a membranespanning hairpin (43), and, therefore, it is possible that the chimera lacked apoptotic activity because of a failure to insert its ␣5/6 hairpin into the MOM. However, of note, Bcl-2 shares the ability of Bax to insert its ␣5/6 helices (44). Furthermore, antibody gel shift analysis suggests that the Bak/Bcl-2␣6 adopts a BH3:groove dimer like activated wild-type Bak, arguing against a defect in ␣5/6 insertion, which occurs (in Bax) prior to oligomerization (43). A plausible explanation for the loss of function is that the Bak/Bcl-2␣6 chimera fails to mediate cell death because it cannot multimerize its BH3:groove homodimers to form an apoptotic pore. Consequently, we propose that BH3: groove homodimers of Bak/Bcl-2␣6 that form as the cells are stressed during routine passaging, or potentially because of the chimera being slightly unstable, can persist without causing cell death and so are detectable prior to an apoptotic stimulus. That the Bcl-2 ␣6 could not mediate higher-order oligomerization is consistent with prosurvival Bcl-2 proteins inhibiting Bak and Bax by binding their activated (BH3-exposed) conformers, thereby acting as dominant negative forms preventing multimerization (45).
How the ␣6 helices become juxtaposed in Bak (2) and Bax (22,25) oligomers is unclear. Although our disulfide linkage may suggest a parallel association of ␣6 helices, we cannot exclude the alternative interpretation that their association is highly flexible and mobile. Regardless of the mechanism by which the ␣6 helices are juxtaposed, it is clear that a significant conformation change from the inactive conformer is required to expose ␣6 residues to facilitate the linkage observed, particularly at Phe-157, which is largely buried in inactive Bak (Fig.  4E) (46). Such an exposure could be facilitated by the complete dissociation of ␣6 -9 from ␣1-5, as suggested by our recent structural studies with Bax (42). It is possible that a BH3:groove homodimer induces proximity of the amphipathic ␣6 helices at the solvent-exposed surface of a pore following insertion into the MOM (43) or, potentially, parallel to the surface of the MOM (23). ␣6-mediated multimerization of symmetric homodimers is more consistent with an open chain of dimers rather than a closed proteinaceous pore. Although it is conceivable that such a linear assembly of Bak dimers may destabilize the MOM, leading to permeability, an alternative model whereby multimerized BH3:groove homodimers are intercalated with MOM lipids to form a lipidic pore can be envisaged. Biophysical studies have supported lipidic pore formation by Bax (3,4,47). The intercalated lipids may afford flexibility between the aggregated homodimers and explain why higherorder Bak oligomers are not stable when solubilized from the MOM (Fig. 4D). In general, the heterogeneous and potentially very large size of Bak (and Bax) complexes observed during cell death, that higher order oligomers are labile once removed from the MOM, and the reported biophysical effects of Bak and Bax oligomers on membranes are more consistent with lipidic pores than proteinaceous pores (3,4,48).
The hydrophobic groove of Bak (49,50) and Bax (39,42) acts as a receptor site for interactions with BH3-only proteins. Additionally, a site involving ␣6 and ␣1 has also been implicated as a receptor site in Bax (38,39). Because an ␣6 peptide could block an early event in Bak activation upstream of BH3:groove dimerization and dissociation from VDAC2, this suggests that a receptor site involving the ␣6 may also be important in Bak. However, the molecular mechanism remains to be elucidated because mutation of the Bak ␣6 with that from Bcl-2 was seemingly not sufficient to prevent activation by tBid.
Elucidating the structure of the Bak/Bax apoptotic pore is not only important to understand how Bak and Bax damage the MOM during apoptosis, it is necessary for the rational design of inhibitory compounds that would have therapeutic potential in the treatment of conditions associated with excessive apoptosis, such as ischemic stroke, acute neurodegenerative disorders, and sepsis (51)(52)(53). We have shown previously, using antibody blockade, that the BH3:groove interaction is essential for Bak function and is, therefore, a valid target to inhibit activity (2). We now show that targeting ␣6 may also represent an effective avenue to inhibit Bak apoptotic function. Because our data suggests that targeting ␣6 may block Bak activation as well as oligomerization, targeting ␣6 may be an effective therapeutic strategy to inhibit apoptosis.