Single-particle cryo-EM reveals conformational variability of the oligomeric VCC β-barrel pore in a lipid bilayer

The pore-forming toxins (PFTs) are a class of secreted virulent proteins produced by pathogenic bacteria. PFTs assemble on the host cell plasma membrane and permeabilize the target cells by punching transmembrane holes. PFTs adopt various strategies to traverse the lipid bilayer for creating pores in biomembranes. Individual protomers of PFTs are secreted as soluble protein by pathogenic bacteria that then accumulate on the surface of the target cell membranes to form a prepore (Iacovache et al., 2012). Subsequently, the pore-forming motifs of the membrane-associated prepore span through the lipid bilayer to punch holes (Dal Peraro and van der Goot, 2016). Generally, PFTs are classified as α-PFT and β-PFT based on the 3D structure of their membrane-inserted transmembrane region (Yamashita et al., 2014; Gouaux, 1997; Parker and Feil, 2005). The transmembrane channel of α-PFTs is spanned by α-helices, while the β-PFTs are able to construct the pore by a β-barrel structure (Iacovache et al., 2010). PFTs do not just puncture holes and permeabilize cells; many also hijack housekeeping functions of the host and lead to various pathophysiological conditions, as in the case of diphtheria, anthrax, or cholera.

Vibrio cholerae cytolysin (VCC) belongs to the β-PFT family, where a β-hairpin structure of the individual protomer assembles to form a β-barrel structure. VCC is encoded by the hlyA gene in enterotoxic strains of V. cholerae. It initially folds into an 80-kD protoxin, which is processed by endogenous or exogenous proteases to form the mature toxin. Proteolytic cleavage occurs in the N-terminus of the protoxin and removes ∼15 kD of “pro-domain” (Paul and Chattopadhyay, 2011). Thus, hemolytically active VCC is an ∼65-kD monomer that can bind the host cell membrane and oligomerize into a heptameric pore of an outer diameter of ∼8 nm and inner diameter between 1 and 2 nm. It has a preference to penetrate cholesterol-rich membranes and form a seven-membered ring-like oligomeric pore structure (Harris et al., 2002).

The crystal structure of the VCC oligomeric pore solved one decade ago revealed a heptameric assembly of protomers remarkably similar to the staphylococcal toxin α-hemolysin (De and Olson, 2011). Each protomer of VCC contributes β-trefoil lectin-like domain, β-prism lectin-like domain, cytolysin domain, and the membrane-proximal rim domain, which come together and build the extracellular mushroom head-like architecture. The membrane-spanning element of the toxin is a rigid barrel composed of 14 antiparallel β-strands (De and Olson, 2011). Sequence analyses of α-hemolysin, leukocidin F, VCC, and aerolysin showed that despite low sequence identity, a central 250-amino acid region of VCC was related to the three other β-PFTs (Olson and Gouaux, 2003). It was reported that VCC has its cytolytic core conserved with α-hemolysin and leukocidin F, suggesting that they follow the same pore formation mechanism (Olson and Gouaux, 2003, 2005). The crystal structure of the VCC heptamer also indicated an unexpected resemblance with anthrax toxin because of the presence of a ring of aromatic residues (W318) lining the narrowest constriction of the pore (De and Olson, 2011). In addition, various studies have impressed upon an obligatory role of cholesterol to promote lytic activity of VCC (Zitzer et al., 2000, 2001, 2003; Harris et al., 2002; Chattopadhyay et al., 2002; Ikigai et al., 2006; Paul and Chattopadhyay, 2012). Although pore-forming proteins have been studied exhaustively over the years, the repertoire of changes guiding such colossal structural rearrangements still remains an enigma. It is believed that furthering knowledge regarding the prepore intermediates may lay the foundation in understanding the precise pathway followed by these membrane-attacking oligomers. Additionally, the cryo-EM resolution revolution (Kühlbrandt, 2014) and recent developments in single-particle analysis allow us to explore the possibility to visualize this macromolecular ensemble on biomembrane in near-native state at atomic resolution. Previously, few cryo-EM structures of various detergent-solubilized PFTs have been characterized. However, only two PFTs to date have been visualized and characterized using single-particle cryo-EM and cryo-electron tomography in cholesterol-containing liposomes (Ruan et al., 2018; van Pee et al., 2017; Tilley et al., 2005; Gilbert et al., 1999).

Taking cues from the present understanding of VCC, we aimed to resolve a high-resolution cryo-EM structure of the native toxin embedded in cholesterol-enriched liposomes. Our cryo-EM map shows good agreement with the crystal structure and adds to existing knowledge with novel insights into the transmembrane β-barrel insertion mechanism. VCC embedded in the biomembrane mimetic asolectin-cholesterol liposomes were imaged at near-physiological conditions using cryo-EM, which provides the detailed structural insight of the β-barrel domain of VCC. This unambiguously highlights the amino acid residues of the β-barrel domain that directly interact with lipid moieties in the bilayer environment. Such interactions may specifically aid in membrane lipid recognition and initiate the several transitions toward pore formation. A number of additional amino acid residues observed to be coupled with the lipid density in our cryo-EM map may have a stabilizing effect and help to tether the massive extracellular part of the heptamer firmly to the membrane. Careful inspection of our cryo-EM map indicates a minor shift starting from F288 located near the mouth of the β-barrel. We speculate that this interesting phenomenon may be attributable to a bilayer membrane that could have a role in slightly altering the barrel architecture as reported in the presence of a detergent micelle. Thus, our structural findings may provide a holistic view regarding the pore formation mechanism followed by VCC and could be extended to its structurally related staphylococcal toxins, cholesterol-dependent cytolysins, and distantly related β-PFTs when present in a lipid bilayer–bound condition.

Crystal structure of the detergent-solubilized heptameric pore state of VCC was resolved by De and Olson (2011). However, there is no high-resolution structure available for the VCC pore incorporated in the membrane lipid bilayer. Therefore, our main objective was to resolve the atomic-resolution structure of VCC embedded in the liposome. To accomplish our goal, we used negative staining room-temperature transmission electron microscopy (TEM) analysis and single-particle cryo-EM to image membrane-embedded VCC (Figs. 1 and 2). For this, ∼4,500 movies were collected using a 200-kV Talos Arctica cryo-electron microscope equipped with K2 direct electron detector. Only images where VCC-embedded liposomes were visible in the micrographs, and ∼1,173,994 membrane-embedded VCC single particles were automatically selected for image processing were considered (Figs. 2 and S2 A). Furthermore, we performed several rounds of rigorous 2D and 3D classifications to classify various states of VCC associated with the lipid bilayer. This cryo-EM–based structural study and powerful 3D classification technique allowed us to categorize the dataset into three major classes: (i) a structural state that was fully embedded in the lipid bilayer (representing the complete pore formation), (ii) a state in which only the part of the transmembrane scaffold partially embedded in the membrane lipid bilayer was documented, and (iii) a state that was anchored on the surface of the membrane and most of the transmembrane scaffold could not be documented/visualized (Fig. 3; Fig. S1; and Fig. S2, B and C). In the next sections, we discuss the structural features of these three structural states of VCC and lipid–protein interactions that we observed in the presence of the membrane lipid bilayer.

Initially, membrane-associated VCC particles were visualized using negative staining TEM, where heptameric ring-shaped complexes were observed (Fig. 1, A–D). Most of the ring-shaped complexes were tightly attached with the lipid vesicle, and no free VCC particles were observed in the background of the TEM grids. Therefore, negative staining images and 2D class averages suggested that most of the VCC heptamers were attached with the lipid vesicle and were suitable for our cryo-EM study (Fig. 1, C and D). These reference-free 2D class averages are comparable with the heptameric VCC crystal structure (Fig. 1, B–D). Previously, there has been no report of using single-particle analysis to resolve the high-resolution cryo-EM structure of small protein molecules, like VCC, embedded in the lipid vesicle. Additionally, biochemical preparation of oligomeric VCC indicates that VCC protomers were incubated with liposomes for oligomerization, and the mixture was washed with detergent-free buffer to remove the unbound VCC molecules. Thus, it is expected that only liposome-bound VCC molecules would be visible, which is supported by negative staining TEM study (Fig. 1, C and D). Therefore, our initial target was to confirm whether single-particle cryo-EM is suitable to characterize the structure of the membrane-embedded VCC molecules. We performed reference-free 2D classification of negatively stained VCC particles to identify various orientations of the VCC molecules. Reference-free 2D classification showed top, bottom, side, and partial side views of the VCC molecules (Fig. 1, C and D). These results provided us sufficient confidence to visualize the small VCC molecules in association with the lipid vesicles at cryogenic conditions. Therefore, we decided to proceed toward structural characterization of the membrane-associated forms of VCC using a single-particle cryo-EM approach.

After several rounds of 3D classifications, we obtained three different classes of 3D reconstruction of VCC particles associated with the membrane lipid bilayer at the local resolution range of 3.7–6.5 Å: (i) the VCC oligomeric pore fully embedded in the membrane lipid bilayer, (ii) oligomeric assembly structure in which only the part of the transmembrane scaffold partially embedded in the membrane lipid bilayer was documented, and (iii) VCC oligomer anchored at the surface of the membrane in which most of the transmembrane scaffold could not be documented/visualized (Figs. 2 C and 3). The membrane-distal mushroom-shaped head parts of the heptameric assembly in all three classes were found to be structurally similar. However, the major alterations in these classes were observed at the membrane-proximal region, mainly in their β-barrel regions and the membrane-proximal rim regions. The length of the β-barrel region and the surrounding lipid environment were found to be remarkably different in all three classes (Fig. 3, B, D, and F; and Fig. S2 C). A close investigation of the cryo-EM maps indicated that the length of the β-barrel region gradually increased in these three classes of the 3D reconstructions in the following sequences: (i) from the VCC oligomer where the structure corresponding to the β-barrel embedded in the lipid bilayer could not be visualized clearly (Figs. 3 B and S2 C), (ii) to the VCC oligomer where the part of the β-barrel embedded in the membrane lipid bilayer could be visualized (Figs. 3 D and S2 C), and (iii) to the VCC oligomeric pore assembly that was found to be fully embedded in the membrane lipid bilayer (Figs. 3 F and S2 C).

In one of the classes, 3D reconstruction of the VCC oligomeric assembly was found to remain bound only at the membrane surface, and only a limited portion of the β-barrel region was visible (Fig. 3, A and B; and Fig. S2 C). In this model, two layers of lipid were visible where the β-barrel appeared to remain tethered with the first lipid layer and the length of the visible β-barrel region was small (19 Å) compared with the complete β-barrel transmembrane region (Fig. S2 C). Furthermore, the pore opening of the β-barrel at the membrane side appeared to remain blocked, presumably by the lipid bilayer (Fig. 3 A). However, the membrane-proximal rim region was successfully affixed and inserted partially within the lipid bilayer. Structural characterization of the 3D reconstruction of this class indicated that the membrane-proximal rim region of VCC interacts with the lipid membrane to provide the robust foundation of the pore formation process. It is important to note that only a small subset of 25,564 particles formed this class where the part of the β-barrel was visible, the rim region was affixed with membrane, and the membrane-proximal region of the channel was blocked by lipid layer (Figs. 3 B, S1, and S2 C).

Our study also illustrates the 3D structure of another minor class of 36,849 particles at a resolution of 5.6 Å in which the part of the β-barrel stem of the VCC oligomeric assembly was found to be partly embedded into the single lipid layer (Fig. 3, C and D; and Figs. S1, S2 C, and S3). In this structural model, a single lipid layer was visible, and the β-barrel stem was observed to pass through the first lipid layer, while the rest of the β-barrel crossing the second lipid layer could not be detected (Fig. 3 C). As a result, the membrane-proximal region of the β-barrel channel remained obstructed by lipid membrane. The length of the β-barrel region in this class is 50 Å, which is shorter than the fully oligomeric pore assembly but longer than the membrane surface–associated VCC oligomer (Fig. S2 C). Interestingly, in this structure, the membrane-proximal rim region of VCC was tightly inserted within the lipid monolayer, whereas the β-barrel stem was partially implanted through the first lipid layer. Such continuous flexibility of the β-barrel region was also supported by the 3D variability analysis of the dataset (Figs. S1 and S2 B, Video 1, and Video 2).

As mentioned earlier, after several rounds of 3D classifications, one cryo-EM reconstruction of the VCC oligomeric assembly in association with the liposome membranes showed the elongated β-barrel stem that passed through the membrane lipid bilayer, and this model contained 117,427 particles (Fig. 3, E and F; and Figs. 4, S1, and S2 C). The local resolution calculation indicated that the lipid bilayer–embedded, fully formed pore model has a high resolution of 3.7–4.0 Å, and this is the first cryo-EM structure of the fully membrane-embedded VCC at atomic resolution (Fig. 2 C; Fig. 3, E and F; and Figs. 4, 5 and 6). The previous crystal structure of VCC oligomeric pore (Protein Data Bank [PDB] accession no. 3O44) solved by De and Olson (2011) could not visualize the detergent micelle. However, our cryo-EM reconstruction of the VCC oligomeric pore fully embedded in the membrane lipid bilayer reveals detailed information of the β-barrel region and rim region, where the lipid bilayer is distinctly detectable (Figs. 3 F and 5, A and B). In this cryo-EM reconstruction, two lipid layers are also clearly visible, and the β-barrel passes through the lipid bilayer. The cryo-EM model shows that the length of the overall molecules is ∼145 Å, and the length of the β-barrel transmembrane region is 64 Å (Fig. S2 C), which are similar to that reported in the previously published crystal structure (Fig. 6 A). Additionally, the rim region is firmly (∼24 Å) penetrated through the lipid bilayer (Fig. 5, A and B).

From our cryo-EM structural model, we could pinpoint several amino acid residues, such as Trp420, Val422, Val423, and Trp362, in the rim region that might be responsible for lipid binding (Fig. 5, A, B, D, and E). Strikingly, we noticed that the rim region is more embedded in the lipid membranes compared with that proposed in the previously determined crystal structure (Fig. 5, D–F). Furthermore, we were able to detect all the amino acids that are embedded in the lipid membranes. Several studies indicated that a two-residue motif (Thr490 and Leu491) in the membrane-interacting region of the cholesterol-dependent cytolysins is crucial in recognizing the cholesterol moiety in the lipid membranes (Farrand et al., 2010; De and Olson, 2011). Additionally, a previous study (Rai and Chattopadhyay, 2015) showed that in the case of VCC, a similar T237–L238 motif is crucial for the membrane-binding process. Consistent with such report, we observed strong lipid density visible in the cryo-EM map in the corresponding locations. Thus, the cryo-EM model and the atomic model of VCC in this study suggest that indeed, the T237–L238 motif directly interacts with lipids and faces the lipid (Fig. 5). Furthermore, as reported in a previous study (De and Olson, 2011), several extra lipid densities were observed in the regions surrounding the loop A360–L361, thus supporting the notion that this loop also plays a vital role in the membrane interaction process. We were able to visualize clear lipid moieties, and these lipid moieties appeared to be prolonged from one protomer to another protomer (Fig. 5 D). The lipid moieties are dispersed within the several amino acid residues, like Trp362-His419/His426, Trp362-Tyr417/Tyr241, Tyr420-Trp362, and Tyr235-Phe242, identified between different VCC protomers (Fig. 5 D). In the current study, several substantial lipid densities are observed within the various amino acid residues of two protomers, like protomer 1 Trp362 with protomer 2 Tyr 241, Tyr420, Tyr421, and Tyr417. Also important to mention is that lipid moieties are observed near the two valine residues (V422 and V423) at the tip of the most extended loop of the rim region at a lower threshold level.

Our local resolution calculation indicates that the cryo-EM 3D reconstruction of the VCC oligomeric pore embedded in the membrane lipid bilayer is of significant high resolution, which is appropriate to determine the atomic model structure of the VCC pore using Phenix refinement (Fig. 2 C; and Figs. 4, 6, and S3). In accordance with past report, we have noticed that Trp318 residues are located in the β-barrel transmembrane domain initiation region, and Trp318 residues are clearly visible in the high-resolution cryo-EM map. The location of the Trp318 residues in the cryo-EM structure and crystal structure are identical. No significant displacement was noticed in the initiation region (Trp318 residues) of the β-barrel transmembrane domain. However, our study could not unravel the biological significance rendered by Trp318 residues, and past Förster resonance energy transfer study as well shows that mutation of W318F does not alter the pore-forming efficiency of VCC (Rai and Chattopadhyay, 2014). Although, no significant displacement was noticed in Trp318 residues, there is a minute 2.43 Å shift in the membrane-proximal region of the β-barrel transmembrane domain between the atomic model of the lipid bilayer–embedded VCC pore and detergent-solubilized crystal structure of VCC determined earlier (Fig. 6 A). It is important to mention that the residues Ser299, Gly300, Asp301, and Gly302 at the membrane-proximal region of the β-barrel stem appear with lipid moieties and are exceptionally disordered, presumably due to the presence of the unstable lipid layer. A similar phenomenon was observed in cryo-EM reconstruction of the lysenin pore complex and anthrax toxin pore complex, where the extreme lower half of the β-barrel transmembrane region is partially disordered (Fig. S4). The similar variability was also noticed in our high-resolution, fully membrane–embedded VCC reconstruction, and the above-mentioned residues appeared as disordered. Interestingly, Val423 at the tip of the membrane-proximal rim region appears shifted (Fig. 6 A). The presence of a bilayer lipid membrane could be the reason for the extended state of the rim region. Additionally, on average, the β-trefoil domain was observed to be nominally shifted by 2.3 Å (Fig. 6 A). This structural analysis also indicates an outward movement in the β-barrel transmembrane domain and the β-trefoil domain in the lipid bilayer–embedded cryo-EM structure (Fig. 6 A). However, it could be said that overall, a strong coherence exists between the lipid bilayer–embedded cryo-EM map and the detergent micelle–purified crystal structure.

In the current study, the first atomic-resolution cryo-EM structure of the VCC oligomeric pore assembled in the membrane lipid bilayer of liposome was resolved using single-particle analysis. This is the only structural study where small β-PFT is characterized in the presence of the real lipid membrane and reveals the lipid–protein interactions of the rim region and the β-barrel of VCC. This structural study will assist us in following a similar pipeline to characterize the small β-PFTs to address the previously unattainable questions of the pore formation mechanism.

Previous x-ray crystallography–based structural study of VCC oligomeric pore was performed with detergent-solubilized protein preparation (De and Olson, 2011). However, the detergent micelle was not visible around the β-barrel transmembrane domain. Several cryo-EM structures of pores and prepore assembly of pneumolysin (van Pee et al., 2017; Tilley et al., 2005; Gilbert et al., 1999), lysenin (Bokori-Brown et al., 2016), and aerolysin (Iacovache et al., 2016) were reported previously. Additionally, Jiang et al. (2015) showed the atomic-resolution cryo-EM model of the pore assembly of anthrax toxin, where the extended β-barrel transmembrane region is distinctly visible. However, in most of the cases, oligomeric pore assembly of the PFT formed in the lipid bilayer has been solubilized using detergent for the subsequent cryo-EM–based structural study (Podobnik et al., 2016; Iacovache et al., 2016; Bräuning et al., 2018). Recently, anthrax toxin (Hardenbrook et al., 2020; Antoni et al., 2020) was characterized in the presence of lipid nanodiscs. To date, only one cryo-electron tomographic analysis of the pore assembly of pneumolysin (van Pee et al., 2017) and a few single-particle cryo-EM analyses of PFTs have been done in the presence of liposome (Gilbert et al., 1999; Tilley et al., 2005; Ruan et al., 2018; Piper et al., 2019; Pang et al., 2019). However, pneumolysin structures are low resolution (∼27 Å), whereas the single-particle–based gasdermin A3 membrane pore is determined at atomic resolution. Both structures, however, are incredibly massive compared with the oligomeric pore structure of VCC. In the current study, single-particle cryo-EM was used to characterize the 3D structure of VCC oligomeric pore in the membrane environment. Thus, this unique approach implements single-particle cryo-EM structural studies to resolve the atomic models of liposome-embedded small β-barrel PFTs and identify the interactions of lipid membrane with various regions of the VCC pore assembly. Previously, several studies showed that cryo-EM and 3D classifications have significant potential to identify various conformational states of biomolecules (Abeyrathne et al., 2016; Banerjee et al., 2016; Plaschka et al., 2017; Shah et al., 2020; Wrapp et al., 2020). Based on the present understanding, it could be presumed that imaging of a large unilamellar vesicle–inserted VCC oligomeric pore at native conditions using the cryo-EM technique may reveal discrete structural entities of VCC comparable to other PFTs (Figs. S4 and S5). Our cryo-EM analysis indicates the first cryo-EM structure of VCC at atomic resolution and that the majority of the VCC particles comprise a fully membrane-embedded heptamer. This structural characterization of VCC is one of the few PFTs resolved in the presence of lipid membrane at an atomic resolution (van Pee et al., 2017; Ruan et al., 2018; Piper et al., 2019; Pang et al., 2019). In this 3D reconstruction, membrane-embedded β-barrel surrounded by lipid membranes is strongly noticeable, where the β-barrel transmembrane channel appears to pass through the lipid bilayers.

Our study provides detailed structural information of fully lipid bilayer–embedded VCC conformer, where we notice intriguing conformational changes between the cryo-EM structure and crystal structure (De and Olson, 2011). These conformational changes are observed in the lower half of the β-barrel transmembrane region more precisely after the amino acid residue F288 in the β-barrel transmembrane channel. It is important to mention that the rim region of VCC also strongly interacts with the lipid membrane, and amino acid residues (His419–His426) of the rim region are fully incorporated into the lipid layer. Interaction of the rim region with the membrane lipid bilayer was not documented in the previously determined crystal structure of the detergent-solubilized VCC oligomeric pore. The present cryo-EM structure of VCC provides the first direct evidence that the rim region of VCC is anchored in the membrane lipid bilayer in the process of membrane pore formation. Several such residues are embedded in the outer lipid layer, which may render extra stability to the VCC pore in the membrane platform. Additionally, several other amino acid residues (Tyr235, Tyr241, Phe242, Trp362, Tyr417, Tyr420, and His426) of the cytolysin domain are tightly coupled with lipid moieties. Furthermore, our study pinpointed that the last four amino acid residues of the β-hairpin loop of the β-barrel stem are considerably disordered. A similar phenomenon is observed in the cryo-EM structure of anthrax toxin (Jiang et al., 2015) and lysenin pore (Bokori-Brown et al., 2016), where the detergent-surrounded extreme lower part of β-barrel appeared as low-resolution features and comparatively less resolved structure (Fig. S5, A–C). It is possible that some of the amino acid residues at the β-hairpin loop region of the extreme lower part of the β-barrel stem are flexible, and that might help the β-hairpin loop to cross the lipid membrane. Future studies in this direction will be required to elucidate the implications, if any, of such flexible/disordered regions of the β-barrel for pore formation in membranes.

The pore formation mechanism of the archetypical β-PFTs, including that of VCC, follows a three-step pathway to punch holes in the lipid membranes: (i) individual monomers interact with the lipid membranes and accumulate on the membrane surface, (ii) membrane-bound toxin monomers self-assemble to form intermediate prepore oligomeric structures, and (iii) prestem motif inserts through the lipid membrane to generate the stable transmembrane pore. Initiation of the membrane insertion of the β-hairpins of the pore-forming motif and transition from prepore to pore are the essential steps in the β-PFT mode of action, the mechanistic basis of which remains unclear in most cases. In addition, the exact structural disposition of the pore-forming motifs in the prepore intermediate states remains unknown for most of the β-PFTs, including VCC. Capturing the snapshots of such remarkably dynamic processes is challenging. In most of the previous studies, prepore intermediates have been trapped by introducing mutations in the protein structure and/or by altering the membrane composition. It is also possible, at least in principle, to capture the intermediate states in the process of membrane pore formation by a cryo-EM–based approach if certain fractions of the membrane-associated proteins are incapable of completing the pore-formation process.

Our cryo-EM study observed that the majority (117,427 particles) of VCC molecules could form the complete β-barrel transmembrane pores. Notably, a subpopulation of the VCC molecules represented structural states where the complete pore assembly could not be visualized. Based on visual inspection, density corresponding to the fully formed transmembrane β-barrel scaffold was not observed in these subgroups of VCC particles. Furthermore, we noticed that in one of these particle subsets (36,849 particles), some density, presumably corresponding to the part of the β-barrel transmembrane scaffold, is partially anchored in the lipid bilayer. In the other subgroup of particles (25,564 particles), structure is associated at the surface of the lipid membranes with a considerably short transmembrane barrel (Figs. S1 and S2, B and C). Similar direction of continuous flexibility in the transmembrane portion is also suggested by 3D variability analysis of the data. Based on such observation, it would be tempting to consider the possibility of whether these unique structures represent the abortive intermediates that are yet to complete the pore-formation event. However, it is important to mention that we were unable to determine the atomic-resolution models of the two auxiliary structural states, likely owing to a smaller number of particles belonging to these cohorts. Therefore, these structural models did not allow us to visualize the detailed structural disposition of the pore-forming motifs in the absence of the complete pore formation, if these are at all representing the intermediate states in the pore-formation process of VCC. Therefore, in the absence of such structural details of the pore-forming motifs, it is not possible to convincingly claim these smaller fractions of the membrane-associated VCC particles to be the intermediate states trapped in the process of membrane pore formation. Nonetheless, the current study establishes a robust foundation for possible future research where trapped intermediates of VCC (via introduction of structure-guided mutations) can be targeted for the structural characterization to elucidate the actual structural disposition of the pore-forming motifs in the prepore states.

In summary, our study elucidates for the first time the high-resolution cryo-EM structure of the oligomeric pore state of VCC in the near-physiological membrane lipid environment. From this structure, we could observe direct interaction of several amino acids of the rim region as well as the β-barrel transmembrane scaffold with the lipid moiety and noticed slight displacement of the β-hairpin loop. Therefore, this study provides important new insights regarding the assembly process of the oligomeric pore state of VCC in the membrane lipid bilayer toward execution of its membrane-damaging action.

Wild-type pro-VCC sequence was cloned between NdeI and BamHI restriction enzyme sites into pET-14b vector (Novagen; Merck Millipore [forward primer 5′-TGT​AGT​CAT​ATG​AAT​ATC​AAT​GAA​CCA​AGT​GGT​GA-3′ and reverse primer 5′-ACA​AGT​GGA​TCC​TTA​GTT​CAA​ATC​AAA​TTG​AAC​CCC​T-3′]) and transformed into Escherichia coli Origami B cells (Novagen) for recombinant protein expression and purification (Paul and Chattopadhyay, 2011). Concisely, cells were grown to mid-log phase, and protein expression was induced in the presence of 1 mM IPTG at 30°C for 3 h. 6× histidine-tagged recombinant pro-VCC was purified by Ni-nitrilotriacetic acid (QIAGEN) affinity chromatography followed by Q-Sepharose anion-exchange chromatography (GE Healthcare). To obtain the mature form of VCC, proteolytic cleavage of the pro domain was performed by treating the purified protein with trypsin in a ratio of 2,000:1 (wt/wt) for 5 min at 25°C. Following trypsin digestion, mature VCC toxin was further subjected to another round of Q-Sepharose anion-exchange chromatography.

Asolectin-cholesterol liposomes (1:1 wt/wt) were prepared by a method described previously (Paul and Chattopadhyay, 2012). Nearly 6 μM VCC was incubated with 1,000 μg of liposomes at 25°C for 1 h. Treatment was followed by ultracentrifugation at 105,000 g for 30 min. Supernatant was discarded, and the pellet was resuspended in 200 μl buffer containing 10 mM Tris-HCl (pH 8.0) and 150 mM NaCl and stored at −20°C for subsequent negative staining TEM imaging.

Wild-type VCC reconstituted in liposomes was assessed for overall sample homogeneity and particle distribution using negative staining TEM. Carbon-coated Cu grids were glow discharged for 30 s in a GloQube (Quorum Technologies Ltd.) glow discharge system before sample addition. 3.5 μl of 0.1 mg/ml VCC-liposome sample was added to the Cu TEM grids and incubated at room temperature for 1.5 min. Excess sample was mildly blotted off, and negative staining was performed with freshly prepared 1% uranyl acetate solution. Data were acquired at room temperature on a 120-kV FEI Tecnai T12 electron microscope equipped with Velita (2k × 2k) side-mounted TEM CCD camera (Olympus) at ×80,000 magnification and calibrated pixel size of 2.54 Å/pixel.

The micrographs were imported into EMAN2.1 software (Tang et al., 2007) for further evaluation. Approximately 1,850 liposome-embedded VCC particles were manually picked, and the coordinates were extracted using e2boxer.py in EMAN2.1. Reference-free 2D classification of the projections was performed using simple_prime2D of SIMPLE 2.0 software (Elmlund and Elmlund, 2012).

QUANTIFOIL R1.2/1.3 300 mesh gold grids were glow discharged in a GloQube (Quorum Technologies Ltd.) glow discharge system for 130 s at 20 mA before sample addition. 3 μl of liposome-embedded VCC sample was added to freshly hydrophilized cryo-EM grids and incubated for 10 s. Excess sample was blotted for 5.5 s at 100% humidity and rapidly plunged into liquid ethane using an FEI Vitrobot IV plunger. Cryo-EM data were acquired at cryogenic temperature using a Thermo Fisher Scientific 200-kV Talos Arctica TEM equipped with K2 Summit direct electron detector (Gatan Inc.). Automated image acquisition was performed using LatitudeS (Kumar et al., 2021) automatic data collection software (Gatan Inc.) at nominal magnification of ×42,000 and pixel size 1.17 Å at specimen level. Total electron dose of ∼40 e⁻/Ų between the defocus range of −0.75 μm and −2.5 μm at a calibrated dose of ∼1 e⁻/Ų per frame was subjected to the sample. Data were recorded for 8 s for a total of 40 frames. Approximately 4,500 movies were collected for data processing.

4,500 movies were imported into RELION 3.1 (Scheres, 2012), and beam-induced motion correction with all the frames was performed using MotionCor2 (Zheng et al., 2017). The motion-corrected micrographs were screened in cisTEM (Grant et al., 2018), and data with the best signal-to-noise ratio were retained for further processing. However, multiple rounds of beam-induced motion correction were performed to remove initial and last frames to improve the drift of huge liposomes. Finally, the first three frames were removed to perform motion correction using MotionCor2 to gain the maximum resolution. CTFFIND 4.1.13 (Rohou and Grigorieff, 2015) was used to estimate the contrast transfer function. Initially, particles were manually picked in RELION 3.1, and reference-free 2D class averages were calculated. The best 2D class averages were selected as a template for autopicking 3,733 final micrographs. A total of 1,173,994 particles were obtained, and 2D classification was run iteratively to clean the dataset. Approximately 406,275 best particles were selected and split into 12 classes with C7 symmetry (Fig. S2). 3D classification identified several distinct classes with respect to the lipid environment. The same particles were imported into cryoSPARC 3.2 (Punjani et al., 2017), and 3D variability analysis (Punjani and Fleet, 2021) was run over 20 frames. The resolutions of individual classes were observed from 2D cross sections of 3D models and metadata of 3D classification. High-resolution models were selected for final refinement. 3D autorefinement for each class was performed with a soft mask created in RELION 3.1. Contrast transfer function refinement for each particle subset was performed with correcting anisotropic magnification, beamtilt, per-particle defocus, and per-particle astigmatism. Particles were subjected to Bayesian polishing; the “shiny” particles were imported into cryoSPARC 3.2; and homogeneous refinement was performed to yield final reconstructions at 5.3 Å, 5.6 Å, and 4.0 Å (Fourier shell correlation [FSC] 0.143) with 25,564, 36,849, and 117,427 particles, respectively. Local resolution was estimated using Blocres (Heymann, 2018). Density modification of bilayer-bound VCC was done with the ResolveCryoEM program (Terwilliger et al., 2020). All 3D structural visualizations were performed in UCSF Chimera (Pettersen et al., 2004) and UCSF ChimeraX (Goddard et al., 2018).

The fully lipid membrane–embedded reconstruction resolved at 4.0 Å was docked with PDB accession no. 3O44 in UCSF Chimera to perform rigid body fitting. PDB accession no. 3O44 was further taken as an initial model for phenix.dock_in_map in Phenix (Adams et al., 2010) followed by a refinement with respect to the cryo-EM map using phenix.real_space_refine. The atomic models of VCC were next compared in UCSF Chimera.

Fig. S1 shows extended 2D class averages highlighting the membrane–protein interaction of VCC heptamers and the different 3D structural states obtained in the context of a lipid bilayer. Fig. S2 provides an overview of the cryo-EM data processing pipeline followed for 3D reconstruction and the continuous flexibility observed in the transmembrane channel of lipid bilayer–bound VCC. Fig. S3 includes the gold standard FSC calculation, local resolution estimation, and Euler angle sampling plot of the structural states of VCC. Fig. S4 shows a comparative analysis of the disorder observed in membrane proximal structural elements of PFTs. Fig. S5 represents the association of lipid moieties with the membrane-interacting domains of PFTs. Video 1 demonstrates the 3D variability of lipid bilayer–embedded VCC as visualized from the side of the protein oligomer. Video 2 demonstrates the 3D variability of lipid bilayer–embedded VCC as visualized from the top of the protein oligomer.

The cryo-EM density maps of the three structural states of membrane bilayer-bound VCC are available from the Electron Microscopy Data Bank (under accession nos. EMD-31973 [membrane surface-adhered VCC], EMD-31974 [partially membrane-bound VCC], and EMD-31972 [fully membrane-embedded VCC]).

We thank Anil Kumar for preparing cryo-EM grids and for cryo-EM data collection.

We acknowledge the Department of Biotechnology, Ministry of Science and Technology, India (DBT); Department of Science and Technology, Ministry of Science and Technology, India (DST); and Ministry of Human Resource Development, India, for funding the cryo-EM facility at the Indian Institute of Science. Indian Institute of Science Education and Research Mohali is also acknowledged for funding support. We acknowledge the DBT-BUILDER program (BT/INF/22/SP22844/2017) and DST-FIST program (SR/FST/LSII-039/2015) for funding the cryo-EM facility at the Indian Institute of Science. We acknowledge financial support from the Ministry of Human Resource Development (grant no. STARS-1/171), Science and Engineering Research Board (grant nos. SB/S2/RJN-145/2015, SERB-EMR/2016/000608), and DBT (grant no. BT/PR25580/BRB/10/1619/2017) for consumables.

The authors declare no competing financial interests.

Author contributions:N. Sengupta: data curation, investigation, formal analysis, validation, visualization, writing–original draft, and review and editing. A.K. Mondal: data curation and writing–review and editing. S. Mishra: formal analysis, investigation, validation, visualization, and writing–review and editing. K. Chattopadhyay: conceptualization, methodology, supervision, writing–original draft, and review and editing. S. Dutta: conceptualization, methodology, funding acquisition, data curation, formal analysis, project administration, supervision, validation, visualization, writing–original draft, and review and editing.