In order to maintain the survival of complex organisms, cells need a strict regulation of division and cell death programs. Members of the Bcl-2 protein family are essential factors in this regulation (Cory and Adams 2002; Kuwana and Newmeyer 2003). The Bcl-2 proteins can be divided into three subgroups: First the anti-apoptotic Bcl-2 like proteins, second the pro-apoptotic multi-domain proteins, and third the pro-apoptotic BH3 only proteins. Human Bax alpha (abbreviated Bax) is probably the best studied example within this family and belongs to the proapoptotic multi-domain proteins. Moreover, Bax is one of the key factors known to initiate apoptosis in vertebrates. Deletion of Bax and its homolog Bak in mice leads to prenatal death or severe deformation in different tissues (Lindsten et al. 2000).

Monomeric Bax protein is cytosolic and inactive. Upon activation, it can undergo a conformational change to form an oligomeric protein complex that inserts into the MOM (Antonsson et al. 2001; Youle and Karbowski 2005). By this, Bax permeabilizes the membrane for cytochrome c (Kuwana et al. 2002) and downstream apoptotic events can be initiated. The conformational change of Bax is regulated by an ever growing number of proteins [e.g. Bid and Bim (Chipuk et al. 2006; Kuwana et al. 2005; Lovell et al. 2008), Bif1 (Cuddeback et al. 2001; Takahashi et al. 2005), p53 (Chipuk et al. 2004; Mihara et al. 2003), ASC (Ohtsuka et al. 2004), Ku70 (Sawada et al. 2003), 14-3-3 (Nomura et al. 2003) or Humanin (Guo et al. 2003)]. In vitro, activation and oligomerization can be induced by mixing Bax with detergents (Antonsson et al. 2000; Antonsson et al. 2001; Hsu and Youle 1997, 1998; Kuwana et al. 2002; Suzuki et al. 2000). Obviously, in living cells Bax activation has to be strictly regulated since premature activation may leads to sudden cell death while blocked activation favors early cancer development.

The structure of monomeric Bax has been determined by NMR methods (Suzuki et al. 2000) and shows a globular fold composed of nine alpha helices, termed α1 to α9. In the structure, α5 is hidden in the hydrophobic core and α9 is attached to a hydrophobic cleft on the protein surface. In contrast, the structure of oligomeric Bax is unknown, but it was demonstrated by biochemical methods that α5, α6 and α9 may be inserted into the OMM (Annis et al. 2005; Garcia-Saez et al. 2004; Nechushtan et al. 1999; Wolter et al. 1997).

The aim of this work was to study the conformational changes upon Bax oligomerization in order to understand the underlying architecture of the oligomeric complex. In our study, we compared monomeric and oligomeric Bax (in detergents or reconstituted in liposomes) using CD-spectroscopy, tryptophan fluorescence (TF), protease digestion, size exclusion chromatography (SEC) and chemical cross-linking. Our results indicate that Bax is extraordinarly stable in the oligomeric form. During monomer to oligomer transition the protein undergoes the following conformational changes: (a) helix α9 becomes protected against protease attack whereas (b) α1 moves in a protease accessible position. (c) The tryptophans in α5 to α9 experience a more hydrophobic environment. (d) The monomer shows an unusually high energy barrier for unfolding but needs a hydrophobic environment to adopt the active conformation. (e) The oligomerization even further stabilizes the protein against unfolding. (f) In the complex, the protein is tightly packed and protected towards thermal and chemical stress.

ClustalW alignment In order to perform the alignment, amino acid sequences from different vertebrate species annotated as Bax isoform alpha [human (Q07812), rat (Q63690), mouse (Q07813), bovine (O022703)] or complete sequences having an equivalent start codon assignment [cat (Q8SQ43), dog (Q8HYUS), zebra fish (Q919N4)] as well as having only a short extension [xenopus (Q98U13)] were used. Sequences with N-terminal extensions or deletions within the sequence were excluded, e.g. the sequence from Pan troglodytes (chimpanzee), which shows only two differences in the part homologous to the human Bax alpha, but is elongated by 70 amino acids at the N-terminus.

Expression and purification of human Bax alpha We essentially followed the procedure described by Suzuki et al. (2000). The purity of the protein was analyzed by SDS-PAGE and LC-MS (calculated mass: 21,184 Da; determined mass: 21,183 Da).

Bax reconstitution in liposomes Lipid mixtures from E. coli or bovine heart extracts (Avanti polar lipids Inc. Alabaster, AL) were dissolved in chloroform, dried by evaporation to form a thin film, and resuspended in buffer 1 (20 mM Tris-HCl, 100 mM NaCl; pH 7.5) by repeatedly vortexing, freezing and thawing. The final lipid concentration was 20 mg/ml. The solution was passed through a membrane with 200 or 400 nm pores (Avestin Inc. Ottawa, Canada) for at least 25 times. Bax pre-incubated with 0.1% DDM was mixed with the vesicles pre-incubated with 0.1% DDM to a final concentration of 0.25 mg/ml protein and 12 mg/ml lipid. Biobeads (BioRAD, Hercules, CA) were added to remove excess detergent. For protein insertion the solution was incubated for 3 h at 30 °C at 300 rpm. To remove free protein as well as aggregates, a nycodenz gradient was used. The gradient was layered as followed: the liposome/Bax solution in 10% nycodenz (dissolved in buffer 1) was overlaid with 5% nycodenz in buffer 1 and finally pure buffer 1. The gradient was spun at 200,000×g for 30 min at 10 °C. A pellet and a swimming lipid band (fuzzy when bovine heart lipids were used) were separated. The latter was mixed with buffer 1 and centrifuged under identical conditions to remove the nycodenz by sedimentation.

Secondary structure determination by circular-dichroism spectroscopy The concentration of Bax (in buffer 2; 20 mM Tris, pH 8.8) was adjusted to 0.1 mg/ml. The CD-spectra and melting curves were recorded on a Jasco J715 spectropolarimeter (Jasco, Gross Umstadt, Germany) with a Jasco PFD 350S Peltier type FDCD attachment for temperature control using a 0.1 mm quartz cuvette. Two spectra were accumulated per measurement using a data pitch of 0.1 nm, a scan speed of 20 nm s⁻¹ and 1 nm slit width. The content of secondary structure was calculated using the program CDNN (Bohm et al. 1992). Notably, samples of Bax did not show any precipitation during the temperature increase which was tested by UV-spectroscopy. If buffer 1 was used instead of buffer 2, only neglectable differences were visible in the spectra.

Protease digestion Monomeric and oligomeric Bax samples were mixed with subtilisin or proteinase K at a stoichiometry of 1:200 and incubated for one hour on ice. The reaction was stopped either by (a) treatment with the denaturing SDS-gel loading buffer and subsequent boiling for 5 min or (b) freezing in liquid nitrogen and storage at −80 °C before further analysis. Protein samples were subjected to SDS-PAGE and blotted onto a PVDF membrane (transfer buffer: 25 mM Tris-HCl, 192 mM glycine, 20% methanol) for subsequent N-terminal sequencing of individual bands (using a gas-phase sequencer Procise 492cLC, Applied Biosystems, Foster City, CA). Sample mixtures were also analyzed by ESI-MS. After Bax reconstitution in liposomes, the protein concentration was estimated by the band intensity on a SDS-gel. Consequently, the estimation was less accurate than the estimation by UV spectroscopy. Furthermore, since ESI-MS did not work with Bax liposomes only N-terminal sequencing was performed.

Tryptophan fluorescence Fluorescence emission spectra were recorded on a Perkin-Elmer spectrometer (LS50B, Waltham MA). Bax (concentrations of 0.25 to 1 µM), and free acetylated-tryptophan (4 µM) were excited at 280 nm at a slit width of 5 nm to detect the emitted fluorescence in the range between 300 and 400 nm.

Cross-link experiments Protein samples were adjusted to a concentration of about 0.5 mg/ml in the presence or absence of 0.5% DDM. These samples were slowly heated to 50–90 °C and the temperature was kept constant for 600 s. Afterwards, the samples were immediately cooled on ice. 10 µl samples were mixed with 1 µl ammonium peroxydisulfate (APS, 25 mM) and 2 µl ruthenium (II) Tris-bipyridyldication (5 mM Sigma-Aldrich) in the dark and then immediately exposed to illumination with visible light (400 to 700 nm, generated by a Xenon lamp, 100 W Leica, filters: KG4, GG 395 nm). The photo-induced reaction was stopped by the addition of 5 µl SDS-gel loading buffer.

Miscellaneous To induce oligomerization, Bax was mixed with 0.5% DM (Anatrace), 0.5% DDM (Anatrace) or 2% OG (Anatrace) in buffer 1. The samples were incubated under shaking for at least 8 h at 4 °C. The oligomerization was analyzed by size exclusion chromatography on a Superdex 200 column using SMART FPLC (GE Healthcare). ESI-MS was performed on a micrOTOF LC (Bruker Daltonics, Billerica, MA). Absorption spectra were collected on a Shimadzu UV-1700 UV-visible spectrophotometer. Bax structure was illustrated with Pymol using the PDB file 1F16 of monomeric Bax (Suzuki et al. 2000). Data were plotted using origin 6.1.

Bax is involved in the intrinsic apoptotic pathway and known to exist in two distinct structural conformations: inactive Bax is monomeric, whereas the active protein forms an oligomer. In vivo it is activated by apoptotic stimuli and, thus, decides on the death or survival of a cell (Kuwana and Newmeyer 2003). During its activation Bax inserts into the MOM and forms a pore by oligomerization (Antonsson et al. 2001; Kuwana et al. 2002; Nechushtan et al. 2001). Through this pore cytochrome c is released promoting further apoptotic events (reviewed in Antignani and Youle 2006; Kuwana and Newmeyer 2003; Reed 2006 and references therein). Activation of Bax is known to involve conformational rearrangements such as the insertion of the helices α5, α6 and α9 into the MOM (Annis et al. 2005).

We address two main questions in this work: First, we intended to identify which part(s) of Bax are involved in structural rearrangements during oligomerization. Furthermore, it was important to assess whether the recombinantly produced monomeric Bax and the detergent induced Bax oligomer folded in the same way as does Bax in non-apoptotic and apoptotic mammalian cells, respectively. Since structural data on Bax conformation in living cells are difficult to obtain and, therefore, are limited we decided to focus on in vitro studies. Detergent induced oligomerization of Bax has previously been used to study conformational rearrangements of Bax (Antonsson et al. 2000; Hsu and Youle 1997). However, these studies did not convincingly reveal that the fold of detergent oligomerized Bax corresponds to that in apoptotic cells. In order to better mimic the natural environment of Bax we also reconstituted the protein into liposomes.

SEC and secondary structure analysis showed that the recombinantly produced Bax was monomeric and folded as previously reported in the literature (Suzuki et al. 2000; Yethon et al. 2003). Detergent induced Bax oligomerization was confirmed by SEC and chemical cross-linking (refer to Fig. 3c and Fig. 7b). The conformation of liposome reconstituted Bax was not accessible by SEC, however, chemical cross-linking yielded mainly dimers (shown in Fig. 7c). Our inability to detect higher number Bax oligomers in liposomes may be explained by the chemical cross-linker’s charge which renders it unlikely to move freely in the hydrophobic lipid environment. It appears that possible other oligomerization sites were buried within the lipid matrix and were, thus, inaccessible to the cross-linker.

The secondary structures of monomeric and detergent solubilized Bax were similar and showed an alpha helical content of about 60–65% (CD-spectroscopy data shown in Fig. 3b). Previously reported secondary structures of monomeric and liposome reconstituted Bax also contained approximately 65% alpha helix (Suzuki et al. 2000; Yethon et al. 2003).

In order to identify structurally important regions in Bax we initiated our studies with an in silico analysis. The alignment of Bax amino acid sequences from different vertebrate species revealed a highly conserved C-terminal half of the proteins whereas the N-terminal halfwith exception of helix α2 is less conserved (see Fig. 2). This implies that the C-terminus of the Bax proteins is more likely involved in Bax function than the N-terminal half (except α2). By projecting the conserved amino acids onto the NMR structure of Bax (Suzuki et al. 2000) a hydrophobic region on the protein surface which is composed of α2 (BH3), the loop between α4 and α5, a part of α5 (BH1) as well as the region of α7 to α9 (containing BH2) became evident (see Fig. 1). The high conservation and the accumulation of BH domains within this region implicate an important role in Bax function. For Bax to act as a cytochrome c pore the protein must run through a series of distinct events: (1) activation by an appropriate trigger, followed by (2) its membrane insertion and (3) either subsequent or simultaneous oligomerization. It appears conceivable that the highly conserved surface region is relevant for at least one of the above outlined events.

Interestingly, five of six tryptophans in the Bax sequence are present in the highly conserved surface region (all except Trp139). TF spectroscopy revealed alterations in the environment of the tryptophans upon Bax oligomerization in both, detergent and liposomes. Dependent on the detergent or the lipids, the hydrophobicity of the tryptophan environment increased as indicated by an up to 9 nm blue shift of the TF emission maximum (see Fig. 4b). Moreover, the concomitant increase in fluorescence intensity suggests a diminished accessibility of the tryptophan residues to oxygen (Fig. 4c). Consequently, at least some of the tryptophans are shielded from oxygen which quenches TF. Oxygen concentrations are usually higher in lipid bilayers and in aqueous solution than in the protein core (Altenbach et al. 1994; Dzikovski et al. 2003; Marsh et al. 2006)). Thus, our observations strongly support the notion that at least some of the tryptophan residues take part in the formation of the core oligomeric structure. This idea was also confirmed by the results of the proteinase K digestions discussed below.

The changes in TF emission wavelength and intensity upon Bax oligomerization varied between different detergents and liposomes (see Fig. 4b, c). Whereas the TF of oligomeric Bax in DDM, DM and liposomes was comparable, Bax oligomerization in OG showed a smaller blue shift and a lower increase in TF intensity. This might be due to the smaller alkyl chain and possibly the differ in micellar shape of OG in contrast to the other detergents and the lipids, causing a less effective protection of the hydrophobic parts of the protein (Lipfert et al. 2007).

As mentioned before protease digestion experiments indicated changes in the conserved surface region upon oligomerization. NMR data suggest that in monomeric Bax, α9 is buried in a hydrophobic cleft on the protein surface (Suzuki et al. 2000). In this study, we found that α9 is subject to proteinase K and subtilisin degradation (see Fig. 5) and, thus, must be at least temporarily solvent exposed. In contrast, α9 was not degraded in the oligomeric protein (see Fig. 5). α9 is a very hydrophobic helix and its movement away from the protein surface could enable its insertion into the lipid membrane. Thus, helix α9 may function as a membrane sensor and, subsequently, membrane anchor that promotes Bax activation (also proposed by Nechushtan et al. 1999).

In summary, our data obtained by TF and protease digestion suggest that at least a part of the conserved Bax surface region changes its conformation upon oligomerization and is likely to form the core of the oligomeric protein. This is in line with the data of others who showed that α5 and α9 (as well as α6, which was not included in the region investigated) become membrane inserted upon Bax activation (Annis et al. 2005; Garcia-Saez et al. 2005; Garcia-Saez et al. 2006).

In order to figure out whether our proteinase K cleavage patterns of Bax mirror the situation in vivo or are merely in vitro artifacts, we compared them to proteinase K digestion patterns Goping et al. obtained of Bax in mammalian cell extracts before and after induction of apoptosis (Goping et al. 1998). We were able to show that recombinantly produced monomeric and detergent induced oligomeric Bax had similar cleavage patterns as Bax in cells before and after apoptosis induction, respectively (see Fig. 5 and Goping et al. 1998). This finding indicates a comparable fold of the recombinantly produced Bax and Bax in mammalian cells.

The less conserved N-proximal part of Bax before α2 (see Fig. 2) most noticeably changed its conformation and became solvent exposed upon oligomerization. This region was cleaved by proteinase K and subtilisin from oligomeric Bax in detergent and liposomes but only to a minor extent from the monomeric form (see Fig. 5). The same region was unfolded during heat treatment, which will be discussed later.

Cartron et al. (2003) showed that α1 is important in membrane binding but not membrane insertion (Cartron et al. 2003). α1 deletion (Bax Δ1–37) inhibits membrane insertion of Bax (Cartron et al. 2003) and the mutant protein is unable to fold into the active conformation. Additionally, it is unclear whether it can adopt the normal monomeric conformation. C-terminal to α1 a long (15 amino acids), flexible (shown by NMR (Suzuki et al. 2000)) loop is found in all solved animal Bcl-2 structures. The loop seems to have a regulatory function (reviewed in (Petros et al. 2004) and references therein). In our studies, proteinase K and subtilisin cleaved oligomeric Bax mainly in this loop (see Fig. 5b and Table 1). In vivo Bax is cleaved in this loop by calpain, which leads to increased apoptosis (Altznauer et al. 2004; Gao and Dou 2000). Taking all findings together, α1 seems crucial for Bax to adopt its correct monomeric conformation, but as soon as folding is completed α1 can be removed, thereby enhancing Bax activation.

From the literature it is known that monomeric and oligomeric Bax are very heat stable (Yethon et al. 2003), and that heating to 43 °C activates Bax in mammalian cell extracts (Pagliari et al. 2005). The latter indicates that the conformational change of Bax is facilitated by heat in the presence of membranes. The melting curves we obtained for monomeric and detergent induced oligomeric Bax were very similar to those recorded by Yethon et al. on monomeric Bax and Bax inserted in liposomes (see Fig. 6 and (Yethon et al. 2003)). Bax resisted thermal denaturation in both conformations: The monomer contained 35% residual α-helices at 90 °C (sigmoidal curve shape; see Fig. 6) while the oligomer was even less affected by heat (residual α-helical content >40%, non-sigmoidal curve shape; Fig. 6). This observation raised a couple of new questions: Which parts of the monomeric protein were the most stable ones? Can we identify differences in the protein structure before and after heat treatment? Can the protein also resist chemical unfolding? And why is a protein of a mesophilic, endothermic organism so stable against heat?

Detergent oligomerized Bax showed a melting curve comparable to that observed by Yethon et al. (2003) with Bax reconstituted in lipids. The refolding efficiency surpassed theirs (Yethon et al. 2003) and complete refolding occurred at lower temperatures. Similar to Yethon et al. (2003), we found that monomeric Bax irreversibly unfolded upon heating. After heating to at least 70 °C and subsequent cooling to 4 °C (heat treatment), monomeric Bax showed an aberrant fold, distinct from the normal monomeric or oligomeric Bax conformations. Instead, “monomeric” Bax formed a “megaoligomer” which was never obtained after heat treatment of the oligomeric form (shown in Fig. 7b). In the Bax megaoligomer, the N-terminus preceding α2 became accessible to proteases; whereas α9 at the C-terminus became inaccessible. This resembles the protease accessibility of the oligomeric Bax (see Table 1). Normal Bax oligomers are only formed in the presence of detergents or liposomes. We assume that in the absence of solubilizing molecules, hydrophobic patches emerging at high temperatures due to partial denaturation of the protein cannot be shielded. Hence, upon cooling the hydrophobic parts of the Bax molecules avoid exposure to the hydrophilic environment by assembly into megaoligomers. The hydrophobic α9 (see Fig. 1d) may form part of these patches.

On the contrary, the better secondary structure recovery of Bax in detergents as compared to Bax in liposomes after heat treatment might be due to the lower size and the higher flexibility of the detergent micelles. Although CD spectroscopy does not provide direct evidence, liposomes are expected to be destroyed at elevated temperatures whereas the detergent molecules maintain their solubilizing function. The shape of the melting curve indicated partial unfolding of the detergent induced Bax oligomers. Nevertheless, a part of the oligomer (maybe to hydrophobic core) seems to remain stable allowing complete refolding of the oligomer during cooling with the detergent molecules shielding the hydrophobic patches. Pagliari et al. (2005) also showed by cross linking oligomer assembly after heating Bax monomers lacking α9 to 43 °C. This shows that without α9 Bax resistance against heat seemed to be weakened.

Bax is not only very heat stable but revealed a high resistance against chemical denaturation as well. Again, the monomeric form was affected (see Fig. 8) whereas the oligomer was almost completely stable under all tested conditions. These findings raised the question why a human protein would exhibit such a high stability rarely known from other human proteins. Cancer or virus infected cells seek to suppress apoptosis. Bax represents an ideal target for apoptosis suppression, but its high stability might put up the necessary resistance.

In summary, we showed that Bax is a very indestructible protein, although substantial conformational changes occur during its activation. On the one hand, we detected significant changes in the region preceding α2. Here, α1 seems to have an important function in adopting the monomeric conformation, but loses importance as soon as the folding is complete. In living cells it is cut by calpain leading to Bax activation. On the other hand, a highly conserved, hydrophobic surface region composed of α2, the loop between α4 and α5, parts of α5, and α7 to α9 and, therefore, significant parts of all three BH domains of Bax seem to be important for and involved in the conformational change. Upon oligomerization this region is exposed to a more hydrophobic environment and α9 becomes inaccessible for proteases. In order to understand the oligomeric structure of Bax in more detail, we are currently performing further analyses on the structural properties of the protein in the monomeric as well as oligomeric form.