Bax Contains Two Functional Mitochondrial Targeting Sequences and Translocates to Mitochondria in a Conformational Change- and Homo-oligomerization-driven Process*

The apoptosis gateway protein Bax normally exists in the cytosol as a globular shaped monomer composed of nine α-helices. During apoptosis, Bax translocates to the mitochondria, forms homo-oligomers, and subsequently induces mitochondrial damage. The mechanism of Bax mitochondrial translocation remains unclear. Among the nine α-helices of Bax, helices 4, 5, 6, and 9 are capable of targeting a heterologous protein to mitochondria. However, only helices 6 and 9 can independently direct the oligomerized Bax to the mitochondria. Although Bax mitochondrial translocation can still proceed with mutations in either helix 6 or helix 9, combined mutations completely abolished mitochondrial targeting in response to activating signals. Using a proline mutagenesis scanning analysis, we demonstrated that conformational changes were sufficient to cause Bax to move from the cytosol to the mitochondria. Moreover, we found that homo-oligomerization of Bax contributed to its mitochondrial translocation. These results suggest that Bax is targeted to the mitochondria through the exposure of one or both of the two functional mitochondrial targeting sequences in a conformational change-driven and homo-oligomerization-aided process.

Numerous stress signals cause apoptosis through the mitochondria-dependent pathway (1)(2)(3). During apoptosis, distinct cellular pathways are initiated in response to various stimuli and funneled to a common destiny, the outer membrane of mitochondria, which upon damage releases cytochrome c and other apoptogenic factors leading to the formation of the apoptosome and the activation of the effector caspases (3)(4)(5).
Damage to the mitochondrial outer membrane during apoptosis is mostly inflicted by Bax and Bak, two members of the Bcl-2 family, which are major regulators of the mitochondrial pathway of apoptosis (4,6,7). The Bcl-2 family proteins are classified into two main functional groups, the anti-apoptotic and the pro-apoptotic groups, which are further divided into the Bax-like proteins and the BH3-only proteins. Members of the Bcl-2 family share at least one of the four Bcl-2 homology (BH) 3 regions, BH1-4. The anti-apoptotic proteins, such as Bcl-2 and Bcl-xL, containing three or four BH regions, protect the integrity of the mitochondrial outer membrane. The BH3only proteins, containing only the BH3 domain, are sentinels that can transmit pro-apoptotic signals to the mitochondria (8). Both the anti-apoptotic and BH3-only proteins are regulators of the Bax-like proteins, which are also termed the multi-BH domain proteins due to the presence of BH1-3. Upon activation during apoptosis, Bax and Bak homo-oligomerize and permeabilize the outer membrane of the mitochondria (6,9,10).
Bax and Bak have been termed the gateway to mitochondriadependent apoptosis. Although loss of Bax or Bak individually shows no or mild apoptotic phenotypes in mice, the combined loss of Bax and Bak causes remarkably strong blocks in apoptosis (11,12). The Bax and Bak doubly deficient mouse embryonic fibroblasts (MEFs) show almost complete resistance to various stimuli, including UV light, ␥-irradiation, serum starvation, and others (12). These genetic studies have unequivocally established a critical role for Bax and Bak in mitochondria-dependent apoptosis.
Accumulating evidence suggests that Bax and Bak undergo an activation process during apoptosis (13)(14)(15). Although Bak is a mitochondrial resident, Bax resides in the cytoplasm as a monomeric protein or is loosely attached to mitochondria under normal conditions (16). Upon apoptotic stimulation, Bax undergoes a conformational change and translocates to the mitochondria (17). How Bax moves from the cytosol to the mitochondria in response to upstream activators remains one of the central questions in the understanding of mitochondriadependent apoptosis. Currently, two main competing models have been proposed. In the direct activation model, certain BH3-only proteins, including tBid, Bim, and Puma, cause Bax to change conformation and translocate to the mitochondria (18 -21). On the other hand, the indirect activation model proposes that Bax or Bak is constitutively complexed with, and therefore suppressed by, the anti-apoptotic family proteins (22)(23)(24). The selective inactivation of these anti-apoptotic Bcl-2 family proteins by multiple BH3-only proteins liberates Bax or Bak from sequestration, thus allowing it to function as an active molecule by default (22). A third model, the "embedded together" model, combines the two previous models but takes into account the involvement of the mitochondrial outer membrane in orchestrating the interactions among the Bcl-2 family proteins (15). However, for any given apoptotic pathway, the molecular details of the upstream events leading to mitochondrial translocation of Bax remain poorly understood (14).
The solution structure of full-length Bax provides critical insights into the regulation of Bax. Bax contains a central hydrophobic helix, ␣5, around which eight amphipathic ␣-helices are packed (25). BH1, -2, and -3, roughly corresponding to individual helices, form a hydrophobic groove similar to that of Bcl-xL (26). However, unlike the Bcl-xL hydrophobic groove, which has been shown to receive BH3 domains such as those from Bak and Bad, the Bax hydrophobic groove was found to be occupied by its own ␣-helix 9 (also termed the C-terminal tail or CT) (27). Because the CT has been proposed to mediate the mitochondrial targeting of Bax, it appears that a major conformational change in Bax must occur for the C-terminal tail to disengage from the groove, thereby targeting Bax to the mitochondria (25, 28 -30). In addition to the elusive nature of the conformational changes, it is not clear whether the conformational change is the cause of the mitochondrial translocation or merely its result. Furthermore, as deletion of the CT fails to abolish mitochondrial localization (31) and cytochrome c release from mitochondria (32), the role of the C-terminal tail as either a bona fide mitochondrial targeting sequence or a regulatory sequence remains controversial. In addition, it has been reported that helix 1 of Bax may serve as the mitochondrial targeting sequence (33). These studies suggest that multiple mitochondrial targeting sequences may be able to target Bax to the mitochondria.
In an earlier mutagenesis study, we identified the homo-oligomerization domain of Bax, which is composed of helices 2, 4, and 5 (34). In the current study, using a similar strategy, we have presented evidence that two mitochondrial targeting sequences (helices 6 and 9) can independently mediate the mitochondrial translocation of Bax during apoptosis. We have also demonstrated that conformational changes may play a causal role in mitochondrial targeting. In addition, we examined the functional relationship between homo-oligomerization and mitochondrial translocation of Bax.
Transfection-HeLa and DKO MEFs were plated (at a density of 2.5 ϫ 10 5 cells/35-mm plate and 7 ϫ 10 5 cells/60-mm plate, respectively) and cultured for 20 -24 h prior to transfection with Effectene reagents (Qiagen) according to the manufacturer's recommendation. For each transfection in a 60-mm plate, 100 ng of the plasmid of interest was used, and pcDNA3 was included in the mixture to maintain the total amount of DNA at 1.5 g/transfection. Transfection in 35-mm dishes included 30 ng of the plasmid of interest and pcDNA3 to maintain total DNA at 0.7 g/transfection. Where indicated, C-terminal FLAG-tagged tBid was included at 120 ng/60-mm plate transfection or 40 ng/35-mm plate transfection. Twenty hours after transfection, cells were fixed for immunostaining and stained with Hoechst dye for quantification or harvested for gel-filtration analysis.
Immunostaining-Immunostaining was performed essentially as described (35). TOM20 antibody (Santa Cruz Biotechnology) was added to the paraformaldehyde-fixed cells at a 1:200 dilution. After washing in phosphate-buffered saline, the cells were incubated with Alexa Fluor 594-conjugated goat anti-rabbit secondary antibody for TOM20 (Molecular Probes) at 1:500. GFP-positive and immunostained cells were photographed under a Nikon Eclipse 50i fluorescence microscope.
Gel-filtration Analysis-Cells were transfected in the presence of 20 M benzyloxycarbonyl-fluoromethyl ketone (Z-VAD). Whole-cell lysates were obtained 18 -20 h after transfection by lysing the cells in lysis buffer (Buffer A with 2% CHAPS). The composition of Buffer A was 20 mM HEPES-KOH (pH 7.5), 10 mM KCl, 1.5 mM MgCl 2 , 1 mM sodium EDTA, 1 mM sodium EGTA, 1 mM dithiothreitol, and 0.1 mM phenylmethylsulfonyl fluoride supplemented with protease inhibitors (5 mg/ml pepstatin A, 10 mg/ml leupeptin). The cell lysates were incubated for 1 h at 4°C with rotation before centrifugation for 1 h at 100,000 ϫ g at 4°C. The supernatant, ϳ1 mg of total protein as measured by protein assay (Bio-Rad), was loaded onto an FPLC Superdex 200 10/30 column equilibrated with 100 mM NaCl in Buffer A (with 1.8% CHAPS) and calibrated using the calibration kit provided by the manufacturer (Amersham Biosciences). Fractions were eluted at a rate of 0.5 ml/min in the same buffer. One-milliliter fractions were collected. Twenty microliters of each fraction was loaded onto SDS-PAGE and transferred to a nitrocellulose membrane after electrophoresis. Western blot analysis was performed to detect GFP fusion proteins.
Quantification of Apoptosis-To quantify apoptosis according to nuclear morphology, cells were transfected with the plas-mid of interest together with 30 ng of the GFP-expressing plasmid pEGFP-C3 in 35-mm cell culture plates. Twenty hours after transfection, cells were stained with Hoechst 33342 at 1 g/ml (Molecular Probes). Two different viewing areas were chosen randomly for each transfection. Pictures for both GFP and Hoechst staining were taken for each viewing area, which contained between 80 and 200 GFP-positive cells. The percentage of GFP-positive cells undergoing pycnotic nuclear condensation among all of the GFP-positive cells was calculated for each viewing area. At least three independent transfection experiments were performed for each construct of interest.

Helices 4, 5, 6, and 9 of Bax Are Independent Mitochondrial
Targeting Sequences-In light of the finding that the CT is not essential for mitochondrial translocation of Bax (31, 33), we systematically searched for additional mitochondrial targeting sequences in Bax. One class of mitochondrial targeting sequences has been characterized by ␣-helical structures containing varying numbers of hydrophobic residues (36). Because Bax is composed primarily of nine ␣-helices, we examined individual helices for their ability to target GFP to mitochondria (Fig. 1). DNA sequences encoding each of the nine helices of Bax were cloned individually into the pEGFP-N2 vector to generate GFP fusion proteins, which were expressed individually in HeLa cells and tested for mitochondrial targeting. Although helices 1, 2, 3, 7, and 8 failed to target GFP to the mitochondria, with helix 8 appearing to target GFP to an organelle that resembles the Golgi, helices 4, 5, 6, and 9 were found to target GFP independently to the mitochondria (Fig. 1). This result is consistent with two earlier findings that helix 9 is a bona fide mitochondrial targeting signal (28,37). However, these results are inconsistent with an earlier study in which helix 1 was reported to serve as a mitochondrial targeting sequence (33). Nonetheless, the existence of three additional mitochondrial targeting sequences (helices 4 -6) provides a plausible explanation for the observation that the CT truncation mutant of Bax can still target to mitochondria in response to apoptotic signals (31).
Mitochondrial Targeting of the Homo-oligomerization Domain of Bax by Helix 6 and Helix 9-Of the four ␣-helices capable of targeting GFP to the mitochondria, which ones can target Bax to mitochondria in vivo? One of the unique features of Bax is that it forms homo-oligomers during its activation (9). Therefore, a more stringent test for the function of these targeting sequences is to see whether they can target the oligomerized Bax to mitochondria. In our earlier study, we found that helices 2-5 constitute the homo-oligomerization domain of Bax (34). When expressed in HeLa cells, GFP-BaxH(2-5) constitutively forms homo-oligomers as monitored by gel-filtration analysis in the presence of the zwitterionic detergent CHAPS ( Fig. 2A). Surprisingly, however, this mutant failed to target to the mitochondria, despite the presence of two of the mitochondrial targeting sequences, helices 4 and 5 (Fig. 2B). This result indicates that helices 4 and 5 are not functional mitochondrial targeting sequences in the context of homo-oligomerized Bax, presumably because of a masking by other helices in the oligomers. To test the function of helices 6 and 9 in the targeting of the oligomerized Bax, we attached helices 6 and 9 to the N and C termini of GFP-BaxH(2-5), respectively. The resulting Bax mutants, GFP-BaxH(2-5)-H9 and H6-GFP-BaxH(2-5), were transfected into HeLa cells and found to form homo-oligomers and to target to the mitochondria (Fig. 2). Thus, helices 6 and 9 are both functional mitochondrial target- Each of the nine Bax ␣-helices was cloned individually into the GFN2 vector as specified under "Experimental Procedure." Individual Bax ␣-helix-GFP fusions were expressed in HeLa cells, and mitochondrial localization was monitored by co-localization with TOM20. The same phenotype was observed in all of the cells in the viewing field for any given construct.
ing sequences of Bax. Moreover, these results suggest that the oligomerization state of Bax may affect the functions of mitochondrial targeting sequences.
Combined Mutations in Helices 6 and 9 Abolish Mitochondrial Translocation of Bax-To investigate the role of helices 6 and 9 in the mitochondrial translocation of Bax, we generated deletion and proline mutations of these two helices individually or in combination. These mutants were tested for mitochondrial translocation in HeLa cells in the absence or presence of tBid, a known inducer of Bax mitochondrial translocation and homo-oligomerization (38,39). As shown in Fig. 3, deletion of helix 6 caused the protein to target to the mitochondria regardless of whether tBid was included. Deletion of helix 9 (G-Bax⌬H9) displayed a cytosolic staining pattern in HeLa cells when transfected alone but localized to the mitochondria in the presence of tBid. These results indicate that neither helix 6 nor helix 9 is essential for mitochondria translocation. In contrast, a compound mutant lacking both helices 6 and 9 (G-Bax⌬H6/ ⌬H9) completely lost the ability to translocate to the mitochondria even in the presence of tBid, suggesting that helices 6 and 9 are the only two functional mitochondrial targeting sequences in Bax. However, one might suggest that the loss of mitochondrial targeting of G-Bax⌬H6/⌬H9 may be because of the inabil-ity of this mutant to respond to the attacks from the BH3-only proteins. Relevant to this possibility, a recent study found that helix 1 and helix 6 of Bax constitute the surfaces that interact with the BH3 domain of Bim (40). We therefore tested the requirement of helix 6 for mitochondrial translocation with a proline substitution mutation instead of a deletion (G-BaxH6P). Although this proline mutant is cytosolic when transfected alone, co-transfection with tBid caused the protein to translocate to the mitochondria, similar to the wild-type Bax, indicating that the proline mutation in helix 6 did not abolish the response to activators (Fig. 3). Similar to the deletion mutant Bax⌬H6/⌬H9, the compound mutant G-BaxH6P⌬H9 lost the ability to target to the mitochondria in the presence of tBid, supporting the conclusion that helix 6 and helix 9 are the only two functional mitochondrial targeting sequences of Bax.
As a further test for the functions of these Bax mutants, we examined their apoptotic activities in DKO cells. Wild-type Bax, Bax⌬H9, Bax⌬H6, BaxH6P, BaxH6P⌬H9, and Bax⌬H6⌬H9 were transfected individually or co-transfected with tBid into Bax Ϫ/Ϫ Bak Ϫ/Ϫ DKO cells (Fig. 4). Apoptosis was monitored by Hoechst staining. In the absence of tBid, wild-type Bax and all mutants except Bax⌬H6 displayed no apoptotic activity. Bax⌬H6, on the other hand, displayed strong apoptotic activity without tBid co-transfection, consistent with a constitutive mitochondrial translocation as shown in Fig. 3. In the presence of tBid, G-Bax, G-Bax⌬H9, and G-BaxH6P all displayed strong apoptotic activity (Fig. 4). In contrast, the compound mutants G-BaxH6P⌬H9 and G-Bax⌬H6/⌬H9 remained inactive even in the presence of tBid. Taken together, these results indicate that helix 6 and helix 9 are the only two functional mitochondrial targeting sequences of Bax.
Perturbation of the Conformation of Individual Helices Can Cause Mitochondrial Translocation of Bax-How Bax translocates to the mitochondria in response to upstream signals remains unclear. Although a conformational change of Bax has been associated with its mitochondrial translocation during apoptosis (30), the causal relationship between these two events remains unclear. We therefore used a "proline scanning" analysis (34) to test whether a conformational change was sufficient to cause mitochondrial targeting of Bax. Because Bax is composed primarily of nine ␣-helices, we reasoned that the introduction of a proline residue, a known helix breaker, in any one of these helices might cause a conformational change in the overall structure of Bax, thus allowing us to test directly the functional relationship between a conformational change and mitochondrial targeting. Altogether, nine proline substitution mutants were generated on the background of the full-length Bax, with each mutant carrying a single proline substitution mutation in the middle of a different helix (Fig. 5, A and B). These Bax proline mutants were fused to GFP and transfected into the Bax Ϫ/Ϫ Bak Ϫ/Ϫ DKO cells. Although none of the proline mutations in helix 4, 6, 7, or 9 affected the subcellular localization of Bax, a proline mutation in any of the remaining five helices, ␣1, ␣2, ␣3, ␣5, or ␣8, led to a dramatic mitochondrial translocation (Fig. 5C). Together, these results suggest that conformational changes are sufficient for mitochondrial translocation. It is worth mentioning that the same set of mutants had been used to identify constitutively homo-oligomerizing mutants of Bax (34). In that study, proline mutations in either ␣3 or ␣8 caused the spontaneous homo-oligomerization of Bax. These results highlight the distinct regulation of mitochondrial targeting and homo-oligomerization of Bax by its helices.

Homo-oligomerization of Bax Contributes to Its Mitochondrial Translocation-What is the relationship between mitochondrial translocation and the homo-oligomerization of Bax?
This question has remained difficult to answer because of the lack of mutants defective for homo-oligomerization, especially those with a conformation similar to that of wild-type Bax. A systematic search for such mutants was carried out in a previous study in which three homo-oligomerization-defective mutants of Bax were identified (34). One of these mutants, G-Bax92-94A, was chosen for further analysis because it displayed a cytosolic staining when expressed alone, indicating that the protein maintained a conformation similar to that of wild-type Bax. To examine homo-oligomerization, plasmids expressing GFP, wild-type Bax, or G-Bax92-94A were transfected into HeLa cells in the presence or absence of tBid expression. Homo-oligomerization was monitored in a gel-filtration analysis. As shown in Fig. 6, none of these proteins formed homo-oligomers without tBid, whereas in the presence of tBid, wild-type Bax, but not GFP or G-Bax92-94A, formed homooligomers. We next examined the mitochondrial translocation of this mutant either in the presence of tBid or following UV treatment. Although the co-transfection of G-Bax with tBid showed predominantly mitochondrial staining, G-Bax92-94A showed a much less efficient mitochondrial localization in the presence of tBid, with the majority of the cells remaining cytosolic. We also subjected the wild-type Bax and 92-94A mutant-expressing cells to UV treatment. Although wild-type Bax responded to UV treatment within 3 h by translocating efficiently to the mitochondria, the mutant G-Bax92-94A responded very poorly. In particular, 4 h after UV treatment, G-Bax92-94 in most cells remained cytosolic (Fig. 6). Even 20 h after UV treatment, there was still a significant percentage of cells expressing G-Bax92-94A that maintained cytosolic staining, whereas all of the cells expressing wild-type G-Bax showed mitochondrial staining (data not shown). These results demonstrate that homo-oligomerization plays a positive role in mitochondrial translocation induced by apoptotic stimuli.

DISCUSSION
The mitochondrial translocation of Bax is a crucial step in the mitochondria-dependent apoptotic pathway. The mechanisms of this translocation, however, remain one of the most challeng-  ing questions in the regulation of apoptosis (6). This study was aimed at answering three related questions regarding Bax mitochondrial translocation. First, can a conformational change cause Bax mitochondrial translocation? Second, what are the sequences that mediate Bax mitochondrial translocation? Third, what is the functional relationship between homo-oligomerization and mitochondrial translocation? Our results suggest that a conformational change of Bax is sufficient to induce exposure of two functional mitochondrial targeting sequences and that mitochondrial translocation is at least facilitated by homo-oligomerization.
Two Functional Mitochondrial Targeting Sequences in Bax-The location of the mitochondrial targeting sequences of Bax has been actively debated (28,30,33,37). In our systematic search, we found that helix 6, in addition to the previously identified helix 9, is a functional mitochondrial targeting sequence in Bax. In support of this proposition, mutation of both helices 6 and 9 completely abolished the mitochondrial translocation in HeLa cells. Interestingly, helix 6 has been proposed to function as part of a pore-forming domain, a hairpin structure that resembles that of diphtheria toxin (26). However, helix 6 was not an essential part of the minimum homo-oligomerization domain, which was shown to induce apoptosis in Bax Ϫ/Ϫ Bax Ϫ/Ϫ DKO cells when targeted to the mitochondria (34). Results from the present study suggest that helix 6 plays an active role in targeting Bax to the mitochondria. Consistent with this finding, helix 6 was found to span the mitochondrial outer membrane in a cross-linking analysis of the mitochondrial Bax (41).
Why does Bax contain two functional mitochondrial targeting sequences? It is conceivable that the dual mitochondrial targeting sequences increase the accuracy and efficiency of mitochondrial translocation. Furthermore, the presence of helix 6 might allow a tighter binding to the mitochondrial outer membrane. A recent study found that the targeting of Bax to the mitochondrial outer membrane is initially reversible before an  (25). B, schematic representation of the proline mutants of Bax. P represents the proline mutation. The amino acid residues for each proline mutant, specified under "Experimental Procedures," were generated to occur at the midpoint of each helix. C, subcellular localization of Bax and its proline mutants. Each of the GFP-Bax proline mutants was transfected into Bax Ϫ/Ϫ Bak Ϫ/Ϫ cells, and mitochondrial localization was determined by co-staining with TOM20. The same phenotype was observed in all of the cells in the viewing field for any given construct. Mito., mitochondrial.
irreversible association and integration (42). Accordingly, helix 9 may be more important in mediating the binding to the mitochondria in an initial, reversible stage, whereas irreversible binding may require both helices 6 and 9. Another possibility is that different apoptosis stimuli may induce different conformational changes. Therefore, helices 6 and 9 may be utilized preferentially in response to different upstream signals. Interestingly, the presence of multiple mitochondrial targeting sequences is not a unique feature of Bax. Bid, for example, contains multiple internal helices that help tBid target to and integrate into the mitochondrial outer membranes (43,44). This feature may help explain the extraordinarily strong and quick binding of tBid to the mitochondria (45).
Conformational Changes Induce Bax Mitochondrial Translocation-It is well established that conformational changes are associated with Bax mitochondrial translocation. However, it is not clear whether the conformational changes are merely a result of the translocation. In this study, we addressed this question by artificially inducing conformational changes by intro-ducing proline mutations in different helices. The 9 ␣-helix conformation of monomeric Bax allowed us to create conformational changes in multiple locations of the molecule. Because five of the nine proline mutations resulted in mitochondrial translocation, it is reasonable to suggest that the perturbation in the structure somehow removed an inhibition to a common driving force for mitochondrial translocation. Consistent with this hypothesis, both helices 6 and 9 are concealed in the inactive conformation (Suzuki et al. (25)). In particular, helix 6 is an amphipathic helix with the charged residues facing the outside, whereas the hydrophobic residues of helix 9 mostly fit into the hydrophobic groove of Bax. Considering the number of helices in which bending by the proline substitution resulted in mitochondrial translocation, we further suggest that at least one of the functions of the helices of Bax is to maintain a globular shaped conformation, concealing the mitochondrial targeting sequences.
Of note, a proline mutation in either helix 6 or helix 9 did not result in mitochondrial translocation (Fig. 5). This may be B, mitochondrial localization of Bax and Bax92-94A mutants in HeLa cells. Expression plasmids expressing GFP, G-Bax, or G-Bax92-94A were transfected into HeLa cells. These cells were either co-transfected with tBid expression plasmid or treated with UV light (20 J/m2). Four hours after UV treatment, these cells were immunostained with TOM20 antibody and visualized under a fluorescence microscope. The same phenotype was observed in all of the cells in the viewing field for any given construct.
because of the tolerance of these two helices to the conformational changes induced by the proline residue. Alternatively, specific conformational changes are necessary for the mitochondrial translocation of Bax.
It is known that nonionic detergents are able to cause the activation of Bax by inducing conformational changes, which can be detected by a conformation-specific monoclonal antibody raised against an N-terminal region covering part of helix 1 (46). Pertinently, the mutation in helix 1 was able to cause the mitochondrial translocation (Fig. 5C), supporting the notion that detergent-mediated activation of Bax is achieved through a conformational change in Bax helices. We speculate that these detergents are able to induce conformational changes in other helices in which proline mutations lead to mitochondria translocation.
What triggered the conformational change? Many studies suggest that Bax may be attacked by a subset of BH3-only proteins, which uses a hit and run mechanism to induce a conformational change in Bax (18,19,38,47). Also, tBid has been demonstrated by fluorescence resonance energy transfer (FRET) analysis to interact with Bax on the mitochondrial outer membrane and induce homo-oligomerization (45). In support of the direct activation model, two studies have identified the Bax helices responsible for the interaction. Cartron el al. (47) suggested that the BH3 domain of tBid interacts with Bax helix 1. Using NMR, it was recently demonstrated that a stapled BH3 helix of Bim binds to the helix 1 and 6 regions of Bax (40). There are other proteins that may trigger Bax mitochondrial translocation in a similar fashion, for example, p53, Bif-1, and ASC (48 -50). The attacks from other proteins may induce a profound conformational change that eventually exposes helix 6 or helix 9. Given our results from the proline scanning experiment, it is also possible that different locations of Bax may be attacked upon different stimuli. Alternatively, post-translational modifications of Bax, for example, phosphorylation, may regulate its mitochondrial translocation (51,52). Of particular interest, the phosphorylation site of Bax (Thr 167 ) by c-Jun N-terminal kinase (JNK) is adjacent to helix 8, where a proline mutation caused Bax translocation (Fig. 5). Therefore, it will be interesting to explore the connection between the activating phosphorylation and the proline mutations that caused mitochondrial translocation of Bax.
Our results do not exclude the possibility that Bax is normally sequestered by the anti-apoptotic Bcl-2 family proteins until they are inactivated during apoptosis (53). It will be interesting to test whether the mitochondria-localizing proline mutations of Bax would abolish interaction with the anti-apoptotic family proteins. It is also possible that the cytosolic environment may play a role in causing conformational changes of Bax. For example, the alkaline pH was reported to induce a conformational change of Bax (54). It has been reported that a cysteine residue of Bax becomes oxidized and forms disulfide bond between two monomers of Bax (55). Moreover, heat has been demonstrated to trigger mitochondrial translocation of Bax (56). It will be interesting to see whether our helix 6 and 9 double mutant is defective in heat-induced mitochondrial translocation.
The Functional Relationship between Homo-oligomerization and Mitochondrial Translocation-Because of the rapid kinetics of mitochondrial translocation and homo-oligomerization (30), it is difficult to determine whether mitochondrial translocation precedes homo-oligomerization or vice versa. Because the minimum homo-oligomerization domain H2-5 can form homo-oligomers without targeting to the mitochondria, it appears that mitochondrial translocation is not required for homo-oligomerization (Fig. 3). On the other hand, the much slower kinetics of mitochondrial translocation of the oligomerization-defective mutant G-Bax92-94A strongly suggests that homo-oligomerization contributes to mitochondrial translocation. Consistent with this finding, the enforced dimerization of Bax by a cross-linker induces mitochondrial targeting (57). It is conceivable that homo-oligomerization could further expose helices 6 and 9. Alternatively, homo-oligomerization may help maintain or sustain the altered conformation induced by the initial attacks, thereby sustaining the exposure of the targeting sequences (Fig. 7). These scenarios are consistent with a model in which Bax mitochondrial translocation is achieved by shifting a series of equilibria from the initial conformational change to the irreversible integration into the mitochondrial outer membrane (15,41). However, it is worth mentioning that at least one of the homo-oligomerization domain mutants of Bax (H2P) localized to the mitochondria constitutively, suggesting that this mutation caused a conformational change strong enough to bypass the requirement for homo-oligomerization (Fig. 1).
Overall, our results suggest that Bax contains two functional mitochondrial targeting sequences (helices 6 and 9). This study also suggests that the conformational changes, presumably induced by attacks on different helices of the cytosolic Bax during apoptosis, are sustained or facilitated by the homo-oligomerization process and are sufficient to mediate mitochondrial translocation (Fig. 7). The results as well as the approach used in this study should be instrumental in future efforts to visualize FIGURE 7. Model for the mitochondrial translocation of Bax. In this model, although conformational changes and mitochondrial translocation are reversible, homo-oligomerization is an irreversible process, which helps stabilize and maintain the active conformation of Bax and therefore drives the equilibrium toward mitochondrial translocation. different conformations of Bax prior to mitochondrial targeting and integration into the mitochondrial outer membrane.