Apoptosis is the major form of programmed cell death in mammals. It is involved in the development and organ morphogenesis, and in tissular homeostasis also. Alterations of apoptosis are responsible for developmental defects and proliferative diseases. Apoptosis is also involved in the response to anti-tumoral therapies, making it one of the most investigated processes of cell biology (1, 2, for reviews).

The intrinsic pathway of apoptosis, often denoted the “mitochondrial pathway”, is activated following intracellular injuries, such as therapies targeting at DNA, and involves the action of Bcl-2 family members on mitochondria. This family is identified as proteins sharing 1 to 4 homology domains (BH1 to BH4) with the anti-apoptotic protein Bcl-2 (3). They are classically classified as multidomain anti-apoptotic proteins (such as Bcl-2, Bcl-xL, Mcl-1, et al), multidomain pro-apoptotic proteins (Bax, Bak, Bok), and BH3-only proteins (Bid, Bim, Puma, et al.) (4).

Bax and Bak form the core of the action of Bcl-2 family members on mitochondria. Both proteins are normally inactive in non-apoptotic cells. Following apoptosis induction, they are activated and are arranged as a large size pore in the outer mitochondrial membrane (OMM) that favors the release of so-called “apoptogenic factors” from the mitochondrial intermembrane space to the cytosol (5, 6). Altogether, these factors contribute to the apoptotic characteristics of the cell through, namely, the activation of caspases.

Bax is a 21kDa-protein that is expressed at a significant level in most cells. It is by far the most expressed Bcl-2 family member in mammalian tissues ( Figure 1 ). The total number of human cells in the human body is estimated to be 3.7 x 10¹³ cells, including 80% of blood cells and the daily turnover of cells is estimated to be around 3.3 x 10¹¹ cells (7). 90% of replaced cells are blood cells and 10% are from solid tissues, representing about 0.33 x 10¹¹ nonblood cells. This shows that about 0.45% of solid tissues cells are dying (and are replaced) each day, implying that Bax, Bak and Bok should be active in a significant fraction of cells (which can however significantly vary depending on the tissue). It is therefore crucial to understand how Bax function is regulated. Anti-apoptotic proteins Bcl-2 and Bcl-xL are potent inhibitors of Bax function, in the sense that they fully prevent Bax-induced MOM-permeabilization, through the formation of inactive Bax/Bcl-2 (or Bax/Bcl-xL) heterodimers. However, apart from the situation when they are overexpressed (i.e. in cancer cells) or in tissues where Bax amount is low (such as in liver), the anti apoptotic proteins are unlikely to contribute significantly to the regulation of Bax, since their relative abundance is significantly much lower ( Figure 1 ). Other regulation mechanisms of Bax function, independent from the Bcl-2 family, are therefore likely to exist.

Bax phosphorylation by the survival protein-kinase AKT has been identified early (8). Ser184, located in the C-terminal hydrophobic helix α9, was shown as a target of AKT. It was next shown that the suppression of Bax function through AKT-induced phosphorylation of Ser184 could explain the anti-apoptotic effect of nicotine (9). Other kinases could also phosphorylate Bax (10), and this residue is also a target of phosphatases (11). Ser184 had been previously identified as a key residue in the biochemical properties of helix α9 (12). Indeed, while wild-type full-length Bax does not bind spontaneously to membranes (see also 13), a mutant carrying a deletion of Ser184 (ΔS184) did, suggesting that the deletion favored a movement of α9-helix towards a conformation compatible with membrane insertion (12). As a matter of fact, while a RFP-Hα9 fusion did no bind spontaneously to mitochondria, a RFP-Hα9-ΔS184 fusion did (14). Furthermore, contrary to wild-type Bax, the mitochondrial insertion of the ΔS184 mutant was independent from the presence of Tom22 and stimulated the further insertion of native BaxWT (14). This led to the hypothesis that, once inserted, membrane Bax could serve as a receptor for cytosolic Bax, in an autocatalytic process (15).

Substitution of BaxS184 by negatively charged residues D or E, mimicking the negative charge of phosphoserine, prevents the mitochondrial localization of Bax, not only in mammalian cells (8, 9, 16), but also following heterologous expression in yeast cells (17, 18). However, rather unexpectedly, the expression of BaxS184D in yeast cells led to a much higher level of cytochrome c release than the expression of BaxWT, a phenomenon that has never been observed in mammalian cells. In addition, the mutant BaxS184D was highly sensitive to proteolytic degradation, but was protected by the co-expression of Bcl-xL (18). Also, Bcl-xL appeared to have a greater affinity for BaxS184D than for BaxWT (or a mutant carrying a S184V substitution) (19). We hypothesized that BaxS184D, although mostly cytosolic, was under a conformation able to permeabilize OMM, and that the small fraction located at the mitochondria was sufficient to promote a large release of cytochrome c. This did not take place in mammalian cells, likely because of the presence of anti-apoptotic proteins, namely Bcl-xL, that forces (non-oligomeric) Bax retrotranslocation (20, 21) or possibly other yet unidentified factors inhibiting Bax. Furthermore, it has been shown that Bax-S184E could bind cBid and remained in solution, thus “depleting” BH3-only proteins in cells and leading to some unexpected anti-apoptotic effect (16).

The substitution S184K also prevents Bax mitochondrial localization in yeast cells (17) and is defective to permeabilize liposomes (22). In this case, the introduction of a large positively charged residue in the middle of Hα9 likely breaks its hydrophobic nature, thus fully preventing Bax membrane insertion and its further capacity to form a pore.

Studies have also been done on mutants where S184 is replaced by a non-phosphorylatable and non-charged residue, A or V. These substitutions stimulate the mitochondrial localization of Bax in mammalian cells (8, 9, 16) and in yeast cells (17). This is associated to an increase of the release of cytochrome c. These observations led investigators to identify small molecules having the capacity to prevent S184 phosphorylation, thus activating Bax and stimulating apoptosis, as possible anti-tumorogenic agents (23, 24).

In the present study, we found that, in spite of having a strong mitochondrial localization in cellulo, in accordance to previous studies, Bax carrying Ala or Val substitutions of Ser184 does not efficiently bind to isolated mitochondria. However, these mutants displayed a significant exposure of the 6A7 epitope, suggesting that a conformational change of the N-terminus of the protein took place. The Ser184>Val mutant only formed small-size oligomers, likely tetramers, while a constitutively fully active and membrane-inserted mutant (Pro168>Ala) formed large-size oligomers, likely decamers or dodecamers. This shows that the dephosphorylation of Ser184 is not sufficient to promote a fully active conformation of Bax, and that its capacity to permeabilize the MOM in cellulo might be rather related to its high mitochondrial residency (25).

Data reported in this paper show that the substitution of S184 by a non-phosphorylatable residue A or V only induced a partial activation of Bax. When expressed in cellulo (in mammalian cells or in yeast), these non-phosphorylatable mutants displayed a greater mitochondrial localization (16, 18), that is associated to a low rate of retrotranslocation (21, 25). However, when tested in vitro on isolated mitochondria or in liposomes, their binding was not significantly greater than the binding of non-activated BaxWT and much weaker than the binding of the constitutively membrane inserted mutant BaxP168A (this study). Their increased capacity to release cytochrome c in cellulo might therefore be more related to their high mitochondrial content than to their activation. This is in line with observations showing that a greater mitochondrial localization of Bax, even poorly active, is sufficient to promote the release of cytochrome c and subsequent apoptosis (25).

However, these mutants displayed some degree of activation. BaxS184V was able to form small-size oligomers, likely tetramers (85kDa), that might be able to promote some membrane permeabilization, although less efficiently than the larger oligomers formed by BaxP168A (>220kDa). Besides, both BaxS184V and BaxS184A exhibited a large exposure of the 6A7 epitope, indicating that a large conformational change of the N-terminal end of the protein did occur. In the pioneering study by Hsu and Youle (32), the 6A7 monoclonal antibody was not able to immunoprecipitate Bax interacting with anti-apoptotic proteins Bcl-xL or Bcl-2. Many investigators (including us) have often interpreted this observation as the 6A7 antibody being able to fully recognize active Bax. However, in the same set of experiments, the authors showed that the 6A7 antibody did not recognize heterodimeric Bax (in a complex with Bcl-2 or Bcl-xL) in the presence of Triton-X100 or Nonidet-P40, that have been shown to actually activate Bax (i.e. to stimulate its capacity to permeabilize membranes) (35). It is therefore inaccurate to state that “6A7 epitope recognition” equals “every active conformation” of Bax, as also shown by the absence of 6A7 exposure of the constitutively active mutant BaxP168A.

Conversely, the resolution of the structure of the Fab fragment of the 6A7 antibody interacting with the Bax epitope showed that the 6A7 recognized a conformation that strongly differs from the known structure of soluble inactive Bax (12) with a large displacement of this epitope by 9.5Å away from the rest of protein (36). However, as underlined by the authors, this conformational change does not recapitulate the whole process leading to Bax activation, that depends on other domains of the protein. For example, a BaxΨ mutant (in which the whole 6A7 epitope is deleted) binds to mitochondria and can permeabilize the OMM to cytochrome c, that is consistent with the loss of the regulatory “ART” domain (37). However, substitutions in the C-terminal half of the α1-helix (residues 26 and 27) fully prevented both BaxΨ-binding and activity, showing that this part of α1 (that is outside from the 6A7 epitope) is required for Bax activation. It was later shown that this might involve an interaction between Bax and Tom22 (14).

It follows that the recognition of Bax by the 6A7 antibody is an early step in the process of Bax activation (that is observed following the interaction with cBid, for example) but does not recapitulate the whole extent of conformational changes supported by Bax along its activation.

Considering the large recognition of both BaxS184A and BaxS184V, it can be suggested that these substitutions initiate the conformational change leading to the exposure of the 6A7 epitope. However, this does not appear to be sufficient to promote a strong membrane insertion in vitro. A possible explanation could be that the substitution S184V induced an alteration leading to the inability to display the fully active conformation. However, in flotation assays, the addition of cBid rendered the protein more able to bind to liposomes. This shows that, although the 6A7 epitope of BaxS184V is exposed, the rest of the protein did not spontaneously display the conformational change that is initiated by cBid, but nevertheless still responded to the activator BH3-only protein and might subsequently be fully activated, as shown by experiments in apoptotic mammalian cells (16).

The spontaneous oligomerization of BaxS184V leads to small size oligomers, likely tetramers, while that of the constitutively full active and membrane inserted mutant BaxP168A leads to objects having a size that likely corresponds to decamers or dodecamers (between 200 and 250kDa). This might explain why, in spite of a strong mitochondrial localization and a large exposure of the 6A7 epitope, the BaxS184V mutant remained poorly active, as compared to the BaxP168A mutant, when heterologously expressed in yeast cells (17 and data herein). It has been shown that membrane inserted Bax spontaneously forms dimers, that further self-assemble into oligomers of different sizes, including tetramers (38). The fact that the S184 substitution limited the formation of larger oligomers, would imply that a correct conformation of Hα9 is needed to form these largest oligomers. It is interesting to note that the P168A substitution, that is expected to change the mobility of Hα9 without modifying its conformation, strongly favored the formation of large oligomers ( Figure 8 ), in accordance with its high ability to release cytochrome c in mammalian cells (28), in yeast (27) and in liposomes (29), in spite of the fact that it did not expose the 6A7 epitope ( Figure 7 ).

From data reported in this study, the non-phosphorylatable mutants of Bax on the S184 residue represent an intermediate activation state of the protein. The fact that this conformation is highly recognized by the 6A7 antibody previously led to the conclusion that it represents a fully activated form of Bax. However, as we show in the present study, it forms small-size oligomers and, contrary to a fully activated mutant, it is not spontaneously inserted on isolated mitochondria or liposomes. S184 is a target of several protein kinases including AKT (8) and PKCζ (9). This led investigators to design small molecules able to interfere with S184 phosphorylation (23, 24). Such molecules might be useful to maintain a high level of Bax activation in cells where these kinases are overactivated, as it happens in PTEN-deficient tumors. Our data suggest that these molecules, although initiating the activation of Bax (or preventing its inhibition by anti-apoptotic kinases) might not be sufficient and would require additional steps to favor the formation of full-size oligomers, which should be designed to improve the effects of these molecules. Interestingly, it has been shown that the activation status of Bax was influencing the response to BH3-mimetics: indeed, the presence of tBid or Bim modulated their effects, even though their primary target are anti-apoptotic proteins (5). This suggest that a combination of this type of molecules, of which the effect against tumoral cells is now well-established (39, 40) and molecules such as the ones developped to specifically target Bax phosphorylation, could form interesting combinations in anti-tumoral treatments. Similarly, screening approaches identified molecules inducing Bax conformational changes related to those induced by its interaction with Bim and/or tBid (41, 42). Bax (and its closely related functional homolog Bak) are at the core of the pro-apoptotic process. Yet, they have for long been overlooked by pharmacology research, likely because their expression often remains unchanged or poorly changed in cancer cells. Hopefully, current advances in the knowledge of both structural and functional aspects of these proteins are likely to stimulate their identification as trackable targets in cancer and other apoptosis related diseases.

The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author.

LS, MG, JH, JF, KD, HA, AE, TD and SC performed and analysed the experiments, and critically read the manuscript. SM and LD integrated data, constructed the figures, and wrote the manuscript. All authors contributed to the article and approved the submitted version.