Application of cryo-electron microscopy for investigation of Bax-induced pores in apoptosis

1 Apoptosis

Cells initiate a suicide program to self-destruct, following a variety of both intrinsic and extrinsic signals. This cell death mechanism, termed apoptosis, controls the cell number in organ development or homeostasis of the immune system. It also checks the uncontrollable proliferation characteristic of cancer [1]. The decision to die is most commonly made through a myriad of signaling pathways converging on the Bcl-2 family proteins that act on mitochondria (Figure 1A) [3]. These proteins share the Bcl-2 homology (BH) domains (up to four) and classified as pro- and anti-apoptotic proteins [4], [5], [6]. The current model posits that BH3 domain-only proteins act as sensors and activate the pro-apoptotic effectors (Bax and Bak) and inactivate anti-apoptotic members (Bcl-2, Bcl-xL, etc.). Bax and Bak are functionally redundant, although Bax resides in the cytoplasm whereas Bak is constitutively in the mitochondrial membrane in non-apoptotic cells. This review focuses mainly on Bax for simplicity; however, the molecular details can also be applied to Bak unless otherwise stated. Activated effectors, then, oligomerize and permeabilize the mitochondrial outer membrane to release apoptogenic proteins, such as cytochrome C, from the intermembrane space of mitochondria (Figure 1B). These proteins, in turn, activate caspase proteases, leading to eventual demise of the cell. Earlier work has established that mitochondrial outer membrane permeabilization (MOMP) is one of the most critical regulatory points in apoptosis [3], [7], [8]. A number of drugs have been developed to directly act on Bcl-2 family members and can induce cell death in cancer with proven efficacy in animal models as well as in some patients [9], [10], [11]. One drug, a Bcl-2-specific inhibitor, venetoclax (ABT-199), has recently been approved by the Food and Drug Administration to treat a certain form of chronic lymphocytic leukemia [12]. This attests to the importance and significance of studying the molecular mechanisms of MOMP.

Figure 1:

Mammalian apoptotic pathways and the current model for the Bcl-2 family protein function. (A) Depending on the death stimulus, apoptosis is activated through the extrinsic pathway (mediated through death receptors) or the intrinsic pathway, in which MOMP is critical. There is also crosstalk between the two pathways via Bid, one of the Bcl-2 family proteins. As a result of Bax/Bak activation and oligomerization in the mitochondrial outer membrane, mitochondria release intermembrane space proteins (cytochrome C, SMAC, or Omi/HTRA2), which leads to activation of caspases. (B) Current model of how Bcl-2 family proteins regulate MOMP. Effector proteins, Bax and Bak, undergo conformational change when activated by BH3 domain-only proteins and oligomerize at the mitochondrial outer membrane to permeabilize it. Anti-apoptotic members can inhibit MOMP by sequestering BH3-only proteins (mode 1 inhibition [2]) and/or activated Bax/Bak (mode 2 inhibition [2]).

2 Mysterious release of proteins from mitochondria

Biological membranes are tightly sealed; in the secretory pathway, COPI (coat protein)-coated vesicles bud from the endoplasmic reticulum, then fuse to the Golgi apparatus [13], or endosome-lysosome fusion occurs in the endocytic pathway [14], all the while the vesicle contents do not leak out. In contrast, mitochondrial membranes become permeable during apoptosis. As a result, multiple apoptogenic macromolecules are released from mitochondria that reside in the intermembrane space, including cytochrome C, SMAC/DIABLO, Omi/HTRA2, etc. [15]. These proteins act in the cytoplasm to exert their pro-apoptotic function. No specificity for the proteins released nor the size limit has been described. Earlier studies to gauge the size of the “pore” induced by Bax by monitoring the release of fluorescently labeled intermembrane space proteins in cells [16] or loading defined size dextrans in liposomes [17] showed mixed results. We developed in vitro vesicle systems to dissect the process that faithfully recapitulate MOMP in a stepwise reduction, in which we ascertained the systems to preserve the characteristic release of molecules [18]. We started with (i) isolated mitochondria from Xenopus egg extracts; (ii) isolated outer membranes from mitochondria, which resealed naturally to form vesicles (outer membrane vesicles – OMVs); and (iii) liposomes made from extracted lipids from whole mitochondria. As these mitochondria-lipid liposomes responded to Bcl-2 family proteins like isolated mitochondria, we generated (iv) liposomes, using purified lipids, consisting of the similar lipid composition to the mitochondria-lipid liposomes. The size-non-discriminatory release induced by Bax and one of its activator, a cleaved form of Bid (cBid), was preserved in this most simplified vesicle system. These data support a model that Bcl-2 family proteins either execute or inhibit apoptosis primarily at the mitochondrial outer membrane by regulating its permeability, as opposed to the hypothesis that Bcl-2 family proteins perturb the electrochemical gradient across the inner membrane (permeability transition) to regulate the release [19], [20]. The physiological relevance of the liposome model has been validated, although the kinetics of release is slightly different between OMVs and the liposomes, suggesting the presence of an unknown catalytic regulator that may be present in the native mitochondrial outer membrane [21].

Attempts to visualize the membrane changes in OMVs, using conventional transmission EM (TEM), failed to detect any specific features in their membranes [18]. It was very puzzling, considering the fact that Bax can release 2000-kD dextrans, which are ~60 nm in diameter, well within the resolution limit of electron microscopy (EM). Neither did we see changes in the outer membrane of isolated mitochondria in scanning EM or electron tomography [22], [23], deepening the mystery. Our conclusion from these studies was that Bax-induced membrane changes are too subtle to be preserved through dehydration and staining required for the above EM techniques.

3 Cryo-EM applied on liposomes and OMVs

To avoid the dehydration and staining steps necessary for conventional TEM, we utilized cryo-EM, as it preserves the native and hydrated state of the membrane. Liposomes are smaller in size (50–200 nm) than OMVs (~500 nm) and therefore easier to be accommodated in the thin ice film required for cryo-EM inspection, whereas thicker ice is needed to freeze large OMVs. The thicker ice is not optimal, however, for it compromises the contrast. The vesicle suspension is deposited on a grid, which is blotted in the humidity chamber, and plunged into liquid ethane to freeze the sample instantaneously in vitreous ice. The blotting and plunging can be performed in an automated system (Vitrobot; FEI, Hillsboro, OR, USA) or manually by gravity. Blotting could produce artifacts such as flattening of vesicles, if too much fluid is removed. The fluid movements during blotting deformed these “soft” vesicles in unpredictable ways. Deformation was worse with larger objects such as OMVs. Therefore, we carefully chose the areas to inspect where vesicles appeared not deformed.

This methodology identified Bax-dependent pore-like openings in the liposomes in projection images [24], [25], [26] (Figure 2A). These pores are variable in size and the edges are often jagged, as if nibbled. We did not see obvious protein density around the pore edges, leaving us to wonder where Bax molecules were. The pore size was variable at ~50–150 nm or even larger (very small pores may have been missed due to the resolution limit), but is consistent with the experimental data showing size-non-discriminatory release and the release of very large dextrans, such as 2000-kD [18]. We also visualized more physiological OMVs and found morphologically very similar pores to those in liposomes (Figure 2B). This suggests that fundamental pore-forming mechanisms are conserved between liposomes and OMVs [27], and these pores may explain the protein release from mitochondria. Curiously, we only saw one pore per OMV, although multiple pores were visible in a liposome [24]. Even though liposomes faithfully capture the membrane-permeabilizing functions of Bcl-2 family proteins, it appears that there is an additional, highly organized process in the mitochondrial outer membrane, involving proteins other than Bcl-2 family (outer membrane proteins?). This finding might also relate to the different kinetics OMVs exhibit in dextran release. However, we could conclude that these pores mediate protein release from mitochondria.

Figure 2:

Cryo-EM detects visible pores in liposomes and OMVs following Bax activation. (A) Liposomes were incubated with recombinant Bax and cBid (cleaved Bid; an active form cleaved at the caspase-8 site [22]) for 1 h and plunge-frozen with a Vitrobot and inspected in a Tecnai F20 (FEI; Hillsboro, OR, USA) electron microscope equipped with a field emission gun at 120 keV. The images were recorded with a CCD camera. Most liposomes exhibit visible pores with variable sizes (blue arrows denote side view pores with frayed membrane edges; red arrows, face view pores). By tilt images, they are confirmed to be pores, not overlapping vesicles or osmotic effects [24]. Often multiple pores are seen in one vesicle. No pores are seen in the absence of activated Bax [24]. (B) OMVs were also incubated with Bax and cBid and frozen manually using a plunger. The grids were inspected in a Tecnai G2 (Sphera; FEI) scope with LaB₆ filament as electron source at 200 keV, and the images were recorded by a CCD camera. The ice needed to be thick to accommodate OMVs and they were often present on the carbon area of the grid, which compromised the resolution. We only saw one pore per vesicle (arrows), but pore features such as variable sizes and shapes are very similar to the liposome pores. Refer to Gillies et al. [27] for the control images.

4 Constituents of Bax-induced pores

There is a long-standing debate about the constituents of Bax-induced apoptotic pore: whether it is “proteinaceous” (a pore lined completely by protein) or “lipidic” (part of the pore walls consist of lipids) [28]. One example of the former pore is the one formed by pneumolysin, which is visualized by cryo-EM and averaged to high resolution owing to its regular size and shape [29]. Bax pores are too variable in size and not amenable to this analysis. Bax has been suspected to form a lipidic pore, because of the similarity of the structure fold of Bax-type proteins to diphtheria toxin and colicin [30], [31]. Some helices (α5, α6, or α9) of Bax, when taken as peptides, are capable of permeabilizing the membrane and form lipidic pores [32], [33], [34], [35], [36]. Our cryo-EM images of the pore could support either model (proteinaceous or lipidic) at this time; even if protein density is invisible on the pore edges, this is not proof that the pore is lipidic, because the protein could exist unfolded and embedded in the bilayer, for example.

5 Bax conformational change

Bax has been known to undergo conformational change upon activation and membrane integration. There is a conformation-specific antibody, 6A7, which detects an activated form of Bax, indicating that the N-terminus, where its epitope is located, is exposed upon activation [37]. There is also an equivalent antibody for Bak [38], and Bax and Bak essentially have the same structure fold [31], [39] and undergo very similar conformational change, particularly after constitutively cytosolic Bax associates with the membrane. Bax requires an additional step to insert into the membrane [40], [41], whereas Bak resides in the mitochondrial outer membrane, without the need for this step. Recent studies have revealed Bax/Bak conformational change upon membrane integration and oligomerization in more detail [42]. According to the models, Bax/Bak unfolds its helices, when activated by BH3-only proteins and/or liberated from anti-apoptotic proteins [43], [44], [45], [46]. The functionally important BH3-domain helix then inserts itself in the hydrophobic groove of another Bax/Bak molecule to form a symmetric dimer [47]. It appears that inter-dimer interactions mediated by multiple sites in the molecule would lead to the formation of higher-order oligomers [48], [49], [50], [51]. Whether these inter-dimer interactions play a critical functional role in MOMP is yet unclear. The idea that Bax and Bak are unfolded and embedded in the membrane [43] is consistent with the cryo-EM image of the pore showing no obvious protruded protein density near the pore.

6 Bax localization in the pore

To elucidate further the composition of the Bax-induced pore, we undertook the study of Bax localization. We decided to use liposomes, as they are more suited for cryo-EM and their biochemical and morphological similarities to OMV pores support their physiological relevance. We labeled Bax in two ways to compensate for each other’s drawbacks: (i) direct nanogold labeling at the relatively flexible N-terminus, through the conjugation of maleimide-nanogold (Nanoprobes Inc., Yaphank, NY, USA) to the engineered cysteine residue (gBax) and (ii) labeling N-terminally histidine-tagged Bax (His-Bax) with Ni⁺⁺-conjugated nanogold also from nanoprobes. Direct conjugation would eliminate the accessibility issue of nanogold to Bax, whereas His-Bax would not suffer the artifact from all the introduced mutations and attached nanogold in gBax. gBax and His-Bax behaved exactly the same as the wild type in our dextran release assays and induced morphologically identical pores in liposomes to that of the wild type-Bax, assuring that these genetically altered Bax molecules are physiological. We found that nanogold particles were present precisely at the pore edges in both labeling schemes (Figure 3) [52]. This argues for the validity of Bax localization on the pore edges and suggests that Bax deforms the membrane at a close range to form pore edges. Even though these images are projections, clear visibility of the pore edges and simple geometry of the pore made interpretation possible. Significantly, other investigators succeeded in visualizing “Bax rings” in mitochondria in apoptotic cells, using super-resolution microscopy [53], [54]. These striking images unambiguously show that the pore is indeed present in apoptotic cells and the pore edges are decorated by Bax, authenticating our findings in liposomes and OMVs. Furthermore, we found that the density of nanogold (=Bax) along the pore edges is constant regardless of the size of the pore, suggesting that the pore enlarges as more and more Bax molecules join the preexisting Bax oligomers. In the liposome system, Bax associated with the membrane is all oligomeric [52]; therefore, Bax molecules on the pore edges must also be oligomers. It seems that Bax deforms the membrane and generates the pore, while attracting more and more Bax molecules to the pore edges to enlarge it in a feed-forward manner. It remains to be seen whether Bax oligomerization is a cause or a consequence of pore enlargement.

Figure 3:

Nanogold labeling demonstrates that Bax is densely localized to the edges of the pore in liposomes. (A) Bax is directly conjugated with nanogold at the flexible N-terminal helix and incubated with liposomes together with cBid. The sample was plunge-frozen manually and inspected with a Tecnai G2 (Sphera) scope. The images were captured in a CCD camera. The particles densely decorate the pore edges. The orange dots denoting gold particles and the yellow lines showing the pore edges are added manually to aid the interpretation of the raw and low-contrast images on the left. Specific localization of the gold particles on the pore edges was confirmed in an objective image analysis [52] (also applied to B). (B) His-Bax and cBid were incubated with liposomes and N⁺⁺-nanogold was added to label His-Bax. Nanogold particles localize to the pore edges just like gBax. Its density was not as high as gBax, hinting that there may be steric hindrance to the His-tag, or the labeling was not as efficient.

7 Possible mechanism of Bax-induced pore formation

What mechanisms of pore formation could we envisage from the data so far? Although there are still missing pieces, we think that unfolding and embedding of Bax helices in the membrane hold the key to pore formation mechanism. There are hints in the studies of antimicrobial peptides, of which some are known to permeabilize the biomembrane, forming “lipidic pores” [28], [55], [56]. Bax and Bak also possess amphipathic helices similar to these short α-helical peptides. According to the pioneering work by Huey Huang [32], [57], [58], [59], [60], it is hypothesized that antimicrobial peptides bind the lipid bilayer “horizontally” at the interface of the hydrophilic head groups and the hydrophobic carbohydrate chains in the outer leaflet, for this is most energetically favorable. When more peptides bind to the membrane, the outer leaflet would expand, due to the presence of the peptides, while the inner leaflet would not. This creates curvature stress, and to alleviate the stress, the peptides would start to orient perpendicular to the bilayer plane, for the pore formation expands the outer leaflet surface area. As a result, the pore edges are formed, consisting of the peptides as well as lipids. The common feature in both the antimicrobial peptides and Bax associations with the membrane is the increase in the density of amphipathic helices on the membrane surface. One could imagine that when Bax unfolds and embeds itself on the surface of the membrane, its amphipathic α-helices (at least there are three of them per molecule) could exert curvature stress very similar to that of antimicrobial peptides. By visualizing nanodisc-incorporated Bax by cryo-EM, Volkmann and colleagues saw a small (~3 nm) pore [61]. It is presumed to be a single Bax molecule per nanodisc; therefore, it suggests that the helices in one Bax molecule could be enough to deform the membrane to form a pore. Even if one molecule supply of helices is not enough in vivo, Bax can form dimers and oligomers, by which we could easily envisage the increase of local concentration of amphipathic helices in the membrane. The clear difference between the anti-microbial peptides and Bax is the size of the pores they induce; for antimicrobial peptides, they are very small (several nanometers in diameter) and uniform [62], [63], whereas Bax pores can become very large (>100 nm) and variable. This could be explained by the propensity of Bax to form oligomers, and our data support this idea. Then, what are the constituents of Bax pores? The above model is consistent with the lipidic pore, involving curvature stress, but the density of Bax molecules we saw in Bax labeling studies is rather high. The inter-particle distances were ~5 nm with gBax and ~10 nm with gHis-Bax [52]. How Bax is lined along the pore edges is still unknown, but I attempt to present a speculative model from the cross-linking and structure studies (Figure 4). This model presumes the position and helicity of the N-terminus helix (α1) (Figure 4A). If true, the inter-molecular distance would be several nanometers (Figure 4B). Considering the incomplete Bax labeling in our data, our numbers are close to this estimation. Are the pores “proteinaceous” with membrane-embedded molecules surrounding the pore? How do new Bax molecules join the existing Bax oligomers? If the contrast in cryo-EM is improved and high resolution is achieved, it may aid answering these questions.

Figure 4:

Model for Bax lining on the pore edges. (A) Schematic diagram of Bax symmetric dimer with conjugated nanogold. Nanogold is attached at the tip of the flexible N-terminus helix (α1) not to disturb its native conformation. The BH3 domain from two Bax molecules (distinguished in green and orange) bind to each other’s pocket (BH3:groove) to form a symmetric dimer. The domain-swapped dimer in the core region is taken from the structure study by Czabotar et al. [45], and the diagram was drawn with the help of Dirk Zajonc (La Jolla Institute for Allergy and Immunology). The estimated length of the dimer core is 2.5×5 nm. The position and helicity of the N-terminus is pure speculation, and it only reflects the reported data that it is not part of the core dimer [46], [64]. (B) Two hypothetical ways to align Bax higher-order oligomers in the pore edges. (I) Two molecules of domain-swapped dimer of Bax (orange and green) line without lipids in between. In this case, the inter-particle distance would be 5 or 2.5 nm depending on the dimer orientation (see A). If all conjugated Bax molecules participated in pore formation, we would expect to see two lines of gold particles with 5- or 2.5-nm interval. Our labeling was ~80% at best, and nanogold-labeled Bax was less active than the wild type; therefore, we suspect that non-labeled species may preferentially participate in pore formation. (II) Lipids come in between the dimers to form lipidic pore. The inter-particle distance would be larger than (I). These models are based on the assumption that the gold particles are located on the tips of the α1 helix and that the N-terminus is flexible. It remains to be seen whether these assumptions are correct.

8 Nanodisc as a promising tool for future investigation

Solving Bax structure in the membrane has been a holy grail of apoptosis research, as this holds a key to cell death and would open many opportunities for therapeutic intervention. The field has experienced a number of breakthroughs and advanced a great length; however, the molecular model is still missing some critical details necessary for effective drug design. Part of the difficulty of studying Bax is its variability that defies structural analysis: variable shapes and sizes of the pore and the heterogeneous size of the oligomers. This heterogeneity poses a great hurdle for the use of liposomes to solve the Bax structure in the membrane by cryo-EM, which needs averaging for high resolution. The nuclear magnetic resonance (NMR) structure of monomeric Bax in solution was determined earlier [31], and was rather a feat at that time, considering the difficulty of producing large amounts of Bax protein. However, the monomeric structure is not sufficient to understand its main function that occurs in the membrane. Nanodiscs are small bilayer discs held by scaffold proteins. As described above, Xu et al. [61] succeeded in loading Bax into the nanodiscs in the presence of activator Bid BH3 peptide. The loading was confirmed by native gel electrophoresis. The requirement of Bid BH3 peptide strongly suggests that Bax undergoes physiological conformational change to incorporate into nanodiscs. Using cryo-EM, the authors were able to image and average the discs and found a small pore in them. Although the resolution was not sufficient to discern the helical arrangement of Bax, this approach could be applied to NMR structure analysis. Indeed, solution structure of anti-apoptotic Bcl-xL loaded in nanodiscs has been successfully solved recently and its functionality confirmed [65], reiterating the promising future of this approach. It may even be possible to incorporate oligomerized Bax, if larger nanodiscs were successfully generated and sorted out by the computerized cryo-EM image analysis [61]. New tools, such as nanodiscs, and technical advances in cryo-EM scopes may finally reveal Bax structure in the membrane.