Step-by-step maturation mechanism of the binary toxin pore revealed by cryo-EM analysis

Significance

Although the mode of action of pore-forming proteins (PFPs) has been increasingly elucidated, the structural mechanisms underlying their maturation into membrane-spanning pores remain unclear. This gap is important because pore-forming toxins are increasingly repurposed as biological nanopores for molecular sensing and DNA sequencing, as well as platforms for cytosolic cargo delivery. Here, we trapped a bacterial toxin PFP in a supercomplex, enabling cryo-EM determination of multiple prepore intermediate states. Comparative structural analysis revealed that the prepore-to-pore transition proceeds through stepwise, domino-like conformational changes. These structural snapshots provide a framework for understanding PFP maturation and for engineering membrane pores with controlled assembly, stability and lumen architecture properties.

Abstract

Membrane pore-forming proteins (PFPs) form ring-shaped membrane-translocating oligomers on membranes, contributing to infection, immunity, and cell death functions. Binary toxins produced by some bacteria consist of an enzymatic component that acts as a toxin and a membrane-binding component that forms a pore that delivers the enzymatic component into target cells. Cryoelectron microscopy (cryo-EM) has advanced our understanding of these translocation mechanisms by revealing several binary toxin complexes’ structures. However, the mechanisms underlying the initial pore formation remain unclear. We determined the structures of several oligomeric forms of the membrane-binding component Ib of the iota toxin from Clostridium perfringens at various stages of pore formation. Structural comparisons revealed how the symmetrically arranged soluble oligomer (prepore) asymmetrically matures into a transmembrane oligomer (pore). These findings enhance our understanding of mechanisms of PFP and provide a structural basis for developing nanodevices using membrane pores.

Membrane pore-forming proteins (PFPs) are a family of proteins produced by many organisms, including bacteria, which form homogeneous or heterogeneous multimeric pores on target biological membranes (1, 2). These proteins are involved in various vital functions, including cell death, immunity, and infection. PFPs are produced as soluble monomers that form ring-shaped oligomers and subsequently create membrane-spanning α-helices or β-barrels to form pores in various biological membranes. Specifically, toxic PFPs are known as pore-forming toxins (PFTs). These PFTs can disrupt the homeostasis of target cells by forming pores, leading to the efflux of ions, amino acids, and water and ultimately inducing cell death.

Clostridium perfringens (3), Clostridioides difficile (4), Clostridium spiroforme (5), Clostridium botulinum (6), Bacillus thuringiensis (7), and Bacillus anthracis (8) produce iota toxin (Ia and Ib), CDT (CDTa and CDTb), CST (CSTa and CSTb), C2 toxin (C2I and C2II), Vip1/Vip2 toxin, and anthrax toxin (LF, EF, and PA) (9), respectively. These binary toxins comprise an enzymatic component and a membrane-binding component, the latter of which functions as a PFP. Whereas the enzymatic components of iota toxin, CDT toxin, CST toxin, C2 toxin, and the Vip2 toxin possess ADP-ribosyltransferase activity (6, 10, 11), the enzymatic components of anthrax toxin, edema factor (EF) and lethal factor (LF), function as a calmodulin-dependent adenylate cyclase and a zinc-dependent metalloprotease (8, 12, 13), respectively. However, the membrane-binding components of these toxins are highly conserved, except for their receptor-binding domains. The addition of the enzymatic component alone to cells or organisms does not result in activity, whereas the membrane-binding component alone induces necrosis and dermonecrosis (14, 15). However, enzymatic component and binding component coaddition increases toxicity compared to their individual addition, inducing morphological changes and cell death (16).

The monomeric membrane-binding component binds to the target cell receptor and is activated by proteases to form a soluble ring-shaped oligomer (prepore). The endosomes take up the enzymatic components and prepores. As the endosome acidifies, the prepore undergoes a conformational change into a membrane-spanning β-barrel pore, allowing the enzymatic component to translocate into the cell (17, 18). The translocated enzymatic component exerts severe effects on host cells through actin depolymerization via ADP-ribosylation activity by clostridial binary toxins (6, 19, 20), or inhibition of signaling pathways through degradation of MAPKK in anthrax toxin (21, 22). At first, it was shown that the pore of anthrax toxin PA forms the narrowest constriction site (ϕ-clamp) with a diameter of only 6 Å in the center (23, 24). Understanding how the enzymatic component passes through the narrow ϕ-clamp is a key interest in studying membrane translocation mechanisms of binary toxins. We previously determined the cryoelectron microscopy (cryo-EM) structures of the enzyme–pore complexes of the iota and CDT toxins (25, 26). The constriction sites of Ib and CDTb, as named ϕ-clamp in PA, were conserved among Ib and CDTb. We found that after complex formation with Ca-edge and NSQ-loop, which are constriction-sites of the pore, the N-terminal α-helix of the enzymatic component tilts toward the ϕ-clamp, and this structure further unfolds (25, 26). Recent structural studies of the Vip1/Vip2 toxin complex have indicated that Vip1 pore induces comparable conformational changes in Vip2 through a similar mechanism (27). In the case of anthrax toxin, the first α-helix of LF and EF unfold and dock into a deep amphipathic cleft called the α-clamp (28). Thus, N-terminal unfolding is commonly observed. It is hypothesized that the disordered structure passes through the ϕ-clamp and refolds its secondary structure within the β-barrel or cytosol. As described above, binary toxin membrane pores are not only pore-forming toxins but also serve as an intriguing research subject because of their unique translocation mechanism involving the unfolding of the enzymatic component.

Mutations in the F427 residue forming the ϕ-clamp of the anthrax toxin PA-pore, as well as the surrounding residues, have significant effects on the maturation from prepore to pore and the translocation activity of the enzymatic component (29–31). In anthrax PA, the D425A point mutation, located two residues upstream of F427, does not affect oligomerization but inhibits the conformational change to the pore state. Cryo-EM structural analysis of this mutant revealed that the oligomer was formed by PA protomers in multiple intermediate states between prepore and pore conformations (32). Furthermore, adjacent protomers have similar structural states, suggesting that maturation from prepore to pore proceeds in a domino-like manner within the heptamer. However, the complete heptamer structure of each intermediate in this domino-like pore maturation process has not yet been experimentally revealed.

By chance, we obtained a supercomplex in which multiple Ib oligomers were assembled radially, referred to here as the Ib-rosette, while preparing the wild-type Ib component of the iota toxin with liposomes. Single-particle structural analysis of the Ib-rosette sample revealed that it contained both Ib-pores with formed β-barrels and Ib-prepores without β-barrels. Detailed analysis of the prepore classes allowed us to determine the structure of a sevenfold-symmetric prepore in Ib, as well as multiple asymmetric prepore structures. Structural analysis of the pore-deficient mutant Ib D452A independently reproduced a prepore conformation observed in the rosette, supporting the physiological relevance of the rosette-derived intermediates. Comparison of the series of obtained overall structures suggested that the symmetrically arranged protomers underwent a partial loss of symmetry, and that this structural change proceeded step-by-step in a domino-like manner during maturation into pores.

Results

Cryo-EM Single Particle Analysis of Ib-Rosette.

While attempting to prepare Ib-pores reconstituted in liposomes, the obtained sample was observed using negative staining and cryo-EM revealing that the structure consisted of multiple radially assembled Ib oligomers (referred to as Ib-rosette) (Fig. 1 A and B and SI Appendix, Fig. S1A). Focusing on the units of the oligomers, single particle structural analysis using cryoSPARC (33) revealed that Ib-rosette contained not only particles of Ib-pores with membrane-spanning regions covered by lipid micelles, but also particles of Ib-prepores where the β-barrel was not yet formed (SI Appendix, Fig. S1 B–D). When observing the two-dimensional average images of the prepore class, an ambiguous density was found in the side view in the direction where the β-barrel would form (Fig. 1C). This indicated that other oligomers in Ib-rosette spatially hinder the formation of the β-barrel. In a previous study of C. difficile CDTb, it is known that paired heptamers form a dimer of heptamers; however, clear two-dimensional images of each heptamer are observed in these cases. In the case of Ib-rosette sample, the average images of these prepore classes were clearly different (34, 35).

Fig. 1.

Cryo-EM maps obtained from Ib-rosette analysis. (A) Micrograph of negatively stained Ib-rosette. (B) Schematic representation of the Ib-rosette. The model was manually generated in UCSF ChimeraX based on negative-stain EM images. (C and D) Representative 2D average images of the side views of the Ib-prepores and Ib-pores. (D–M) Cryo-EM maps of Ib-prepores obtained from cryo-EM single-particle analysis. Initial state protomers are shown in light purple, intermediate state protomers in brown, and pore state protomers in purple. (D–K) are views from the top of the prepore when viewed from the extracellular side. (L and M) Side views of Prepore I and Prepore VIII. (N) Cryo-EM map of Ib-pore viewed from the side (Left) and top (Right).

The prepore particle dataset was further classified into five subclasses: particles with an open central (class 1) region and particles with a constricted central structure where the base of the β-barrel has formed but is still immature (class 5) (SI Appendix, Fig. S2A). Other classes suggested that the particle dataset was composed of protomers in multiple states, because the density of some portion in the protomers were averaged out. Detailed structural analysis at the protomer level with symmetry expansion, focused 3D classification (SI Appendix, Fig. S2 B–D), refinement using appropriate reference map (SI Appendix, Fig. S2E) for each class and further focused 3D classification, revealed that the heptameric prepore state is formed by two distinct protomer states, which differ from those in the pore state: the initial state and intermediate state (SI Appendix, Fig. S2 F and G). Finally, the particle datasets for prepore class 1 to 4 were classified into eight classes, reflecting the different proportions of these two states (Fig. 1 D–M and SI Appendix, Fig. S3). For clarity, adjacent protomers are hereafter referred to as clockwise (CW) and counterclockwise (CCW) relative to a given protomer when viewed from the extracellular side. The structural characteristics of each protomer’s state are described in the following sections.

The arrangements of the protomers constituting the prepores obtained in this analysis were distinctive. From Prepore I to Prepore VIII, the number of protomers in the initial state decreased one by one, whereas the number of protomers in the intermediate state increased correspondingly. In addition, the intermediate state protomers were not scattered throughout the oligomer but were positioned adjacent to one another. These differences among the Prepore classes suggest that prepore maturation proceeds in a step-by-step and domino-like manner. Furthermore, the Ib-rosette enabled high-resolution structural determination of the pore state after β-barrel formation (Fig. 1N). However, we did not obtain density maps representing an oligomeric state in which intermediate- and pore-state protomers coexist.

Definition of the Initial and Intermediate Protomer States.

Ib consists of four domains: D1, which binds to the enzymatic component Ia; D2, which forms the pore lumen; D3, which interacts with other domains within a protomer and contributes to structural stabilization directly; and D4, which is responsible for binding to the target cell receptor (Fig. 2 A–C). D2 can be further subdivided into D2c which serves as the structural foundation of the oligomer, and D2s which form β-barrel upon pore formation. Structural changes were observed in both of D2c and D2s. The constriction, which was found in classification of Prepores, corresponds to the conformational change of D2c toward the center of the oligomer. The visibility of the density of D2s contributed as a key indicator to classify multiple prepore classes during the single particle analysis. The state of protomer with visible D2s density was termed the initial state (Fig. 2D) because its structure was similar to the previously reported monomeric structures of CDTb (PDB ID: 8DCM) (SI Appendix, Fig. S4 A and B) and PA (PDB ID: 1ACC), with which it shares the feature that D2s is retained in the same position (36, 37). In contrast, the state of protomer without D2s density had a structure more similar to the pore state than to the initial state, although it had not yet formed the β- barrel. We therefore termed this state the intermediate state (Fig. 2E), as it appeared to represent an intermediate between the initial and pore states (Fig. 2F).

Fig. 2.

Ib protomer structures. (A) Schematic of the N-terminal cleaved Ib domains with densities observed in cryo-EM maps. (B–F) Cryo-EM maps of Ib protomers, generated by cropping the density within 3 Å around a single protomer using UCSF ChimeraX and colored according to the domains shown in panel A. (B and C) Schematic showing the arrangement of protomers within the oligomer. (D) The initial state map was extracted from the Prepore I. (E) The intermediate state map was extracted from the Prepore VIII. (F) The pore state map was extracted from the Pore.

Structural Basis of Prepore Maturation and ϕ-Clamp Assembly.

The major structural changes among the initial, intermediate, and pore states were observed in D2 (Fig. 3A). Among these changes, α5 (aa. 463 to 470) underwent pronounced rearrangements during each transition, and the other residues showing substantial conformational changes were located in the vicinity of α5 (Fig. 3B). During the transition from the initial to the intermediate state, α5 shifted markedly from the outer side of the oligomer toward the central axis. This movement was accompanied by a concomitant displacement of the β-sheet located above α5, which is composed of β7, β8, and β10 (aa. 406 to 422, 476 to 483) (Fig. 3C).

Fig. 3.

Significant structural transition between the protomer states. (A) Summary of the major structural changes associated with transitions between protomer states. (B) Arrangement of the structural elements involved in the protomer state transition. The model was generated using the initial state. Because the ϕ-clamp was not modeled, it is indicated by a dashed line. (C) Comparison of cartoon models with tube α-helix between the initial state (light purple) vs. the intermediate state (brown). (D and E) Close-up side views of the protomer interfaces, comparing the initial–initial interface in Prepore I with the initial–intermediate interface in Prepore II. (F) Comparison of cartoons between the intermediate state (brown) and the pore state (purple). (G) Atomic model of pore state Ib. The dashed lines show the hydrogen bonds. (H) Close-up view of the central constriction site ϕ-clamp. Each map was calculated using C7 symmetry.

Comparison of the dimeric interfaces within the oligomer showed that the rearrangement of the β7–β8–β10 sheet was directly associated with the formation of the interface between adjacent protomers. A gap was observed at the initial–initial interface, whereas this gap was closed at the intermediate–initial interface in which an intermediate state protomer is positioned clockwise (CW) of an initial state protomer (Fig. 3 D and E). This closure was caused by movement of the β7–β8–β10 sheet in the CW protomer toward the counterclockwise (CCW) protomer, resulting in a tighter protomer–protomer interface. This rearrangement establishes an interface that closely resembles that observed in the pore state (SI Appendix, Fig. S4 C and D).

The most apparent difference between the intermediate and pore states is whether D2s has formed a β-barrel; however, α5 was also found to undergo a further slight shift in the pore state (Fig. 3F). In the pore state, the region around α5 acquires interactions through its proximity to the base of the β-barrel (aa. 382 to 388) (Fig. 3G). On the other hand, no corresponding density for the foundation of the β-barrel was observed in Prepore VIII.

The membrane-binding components of binary toxins are known to form a hydrophobic constriction site, termed the ϕ-clamp, at the center of the pore through the symmetric arrangement of highly conserved phenylalanine residues (Fig. 3H). Comparison of the structures of Prepore I (sevenfold-symmetric initial state), Prepore VIII (sevenfold-symmetric intermediate state), and the pore state showed that the center of the oligomer remains widely unconstricted in Prepore I. Moreover, no density corresponding to residues surrounding F454 was observed in the initial state, indicating that the ϕ-clamp is not formed in Prepore I. In Prepore VIII, a constricted density was observed at the center of the oligomer, although it was weaker than that in the pore state, suggesting that this structure remains unstable. Unlike Prepore I, however, the density around F454 in Prepore VIII allowed tracing of the main chain, indicating that an immature ϕ-clamp is formed. In the pore state, the ϕ-clamp was clearly resolved allowing to observe side-chain density, which suggests it is stable matured structure. Because F454 is located immediately upstream of α5, which shifts during the course of these structural transitions, these observations suggest that ϕ-clamp formation depends on α5 movement.

Mutation Near the ϕ-Clamp Stabilizes a Monodispersed VIII-Like Prepore.

The structural determination and comparison of the initial and intermediate states provided insight into the structural basis of maturation toward the pore state. However, prepore particles derived from Ib-rosettes were extremely limited in abundance and were further distributed across multiple prepore classes, precluding more detailed analysis. In addition, it is known that when wild-type Ib oligomers were purified following cleavage of the N-terminal propeptide that inhibits oligomer formation, they turned into an irreversible pore state (25). As a result, a monodisperse prepore had not previously been obtained.

Previous studies of anthrax protective antigen (PA) showed that mutation of the D425 which is located near ϕ-clamp forming F427 inhibits pore formation (32) (Fig. 4A). The homologous residue in Ib D452 contributes to stabilization of the ϕ-clamp loop by forming hydrogen-bonding interactions within the loop (Fig. 4B). Following this precedent, we prepared the D452A mutant of Ib as a conventional monodisperse oligomer and subjected it to cryo-EM analysis, which enabled us to determine the structure of a monodisperse Ib-prepore (SI Appendix, Fig. S5A) that had not previously been obtained from wild-type Ib. Unexpectedly, however, the D452A dataset also contained pore particles: unlike the anthrax PA mutation, which fully blocks maturation to the pore state, the homologous mutation in Ib only partially impaired this transition, yielding a mixture of prepore and pore particles.

Fig. 4.

Structural analysis of D452A Ib oligomer. (A) Alignment of amino acid sequences surrounding the ϕ-clamp region among binary toxins. UniProt IDs are as follows: Ib, Q46221; CSTb, O06498; CDTb, O32739; C2II, O86171; Vip1, A0A075BFL2; PA, P13423. (B) Model in map representation of the ϕ-clamp in the wild-type Ib pore state. (C) Cryo-EM map of D452A Ib prepore viewed from the side (Left) and top (Right). (D) Cryo-EM map of the protomer in D452A Ib prepore, generated by cropping the density within 3 Å around a single protomer using UCSF ChimeraX and colored according to the domains shown in Fig. 2A. (E) Cryo-EM map of D452A Ib pore viewed from the side (Left) and top (Right). (F) Cartoon representations of the ϕ-clamp structures in the wild-type Ib pore (salmon) and the D452A Ib pore (purple). (G) Model in map representation of the ϕ-clamp in the D452A Ib pore state.

For D452A prepore, single particle analysis was performed using the same workflow as that applied to the Ib-rosette dataset, but no prepore classes corresponding to Prepore I–VII were obtained. Instead, all particles corresponded to a Prepore VIII-like class in which D2s had been released from all protomers (Fig. 4 C and D). These results suggested that, under these conditions, the intermediate state is more stable than the initial state in monodisperse D452A oligomers in solution. Notably, the density around the ϕ-clamp was even less well defined than that in Prepore VIII derived from Ib-rosettes, and the main chain could not be traced. Furthermore, the density around α5 in the D452A prepore was less well defined than that in Prepore VIII, and the corresponding atomic model exhibited higher B-factors (SI Appendix, Fig. S6), indicating that this region is unstable. These features suggest that ϕ-clamp formation is coupled to the stabilization of α5.

For the D452A pore, by contrast, a ϕ-clamp was observed at the center of the structure, although its main-chain conformation differed from that of the wild type (Fig. 4 E and F). Interestingly, the amide group of the less conserved N455 main chain was found to be stabilized through a hydrogen bond with another N455 residue side chain in the adjacent clockwise (CW) subunit (Fig. 4 A and G). In addition, the density corresponding to the F454 side chain was weak in D452A pore, suggesting that this side chain is flexible or asymmetrically arranged (Fig. 4 B and G). These observations indicated that the ϕ-clamp in the D452A mutant is stabilized by a mechanism clearly distinct from that in the wild type.

Discussion

In this study, cryo-EM analysis of the Ib-rosette, a supercomplex formed by a bacterial PFT, revealed step-by-step conformational changes during the transition from a soluble prepore to a membrane-spanning pore. Although the general mode of action of PFTs is widely recognized, detailed insights into their structural changes have mostly been based on comparisons of a few stable states and biochemical experiments with mutants. Supported by cryo-EM density maps (Fig. 1 D–K), we identified multiple structural states, ranging from an early immature symmetric prepore composed of soluble-state protomers (Prepore I), through asymmetric prepores containing structurally rearranged protomers (Prepore II–VII), to the fully matured symmetric Prepore VIII and the pore state with a membrane-spanning β-barrel. These structural comparisons allowed us to propose a stepwise and domino-like maturation model.

In our previous cryo-EM studies of the iota toxin, we determined the structures of both the Ib-pore and the Ia-bound Ib-pore complex as well as the CDTa-bound CDTb-pore complex, revealing the docking mode of the enzymatic component as the initial step of its translocation through the pore. These studies showed that the N terminus of the enzymatic component undergoes unfolding, allowing it to pass through the ϕ-clamp, the narrowest constriction site of the pore. Similar enzymatic component-bound pore structures have also been reported for Vip1/Vip2 toxin complexes, further highlighting the importance of understanding how unfolded enzymatic components are threaded through the narrow pore lumen. This study extends these findings upstream by revealing structural intermediates that precede β-barrel formation and pore maturation.

Scott et al. previously reported structural heterogeneity among protomers in oligomeric anthrax toxin PA using cryo-EM analysis of the pore-formation-deficient D425A mutant (32). In that study, distinct protomer states were classified mainly on the basis of receptor-binding domain density, and they reported that protomers in similar states tended to be adjacent within the oligomer. Extending this concept to wild-type C. perfringens iota toxin Ib, we classified protomer states based on structural features of D2c and D2s, and defined three states: initial, intermediate, and pore states (Fig. 2 D–F). In addition, we identified multiple oligomeric prepore structures that differed in the ratio and arrangement of these protomer states. The adjacent distribution of initial- and intermediate-state protomers further suggests that prepore maturation follows an ordered, locally propagated transition rather than random structural fluctuation.

Aggregates similar to the Ib-rosette have been observed in oligomers of anthrax toxin PA and other binary toxins. It has been reported that such structures can be obtained by incubating oligomers bound to a carrier under acidic conditions (38). However, unlike in the present study, structural analysis of the constituent oligomers has not been achieved. The antigen component of the influenza virus vaccine forms a radial structure composed of HA proteins, referred to as the HA rosette (39, 40). Following this convention, we refer to the aggregated Ib structure as the Ib-rosette. Although Ib-rosette formation was unexpected, it provided a unique opportunity to trap otherwise transient prepore intermediates.

We identified a series of eight prepores that differed in the proportion of protomer states (Fig. 1 D–K). These structures indicate that maturation of the Ib pore does not proceed through a simultaneous transition of the entire heptamer, but rather through stepwise conformational changes in individual protomers. Moreover, intermediate state protomers were not randomly distributed within the oligomer, but were positioned adjacent to one another. This arrangement suggests that the conformational transition may propagate locally through interactions between adjacent protomers, consistent with a domino-like maturation process.

An important question is whether the transition from the initial to the intermediate state occurs only after formation of a complete heptamer, or whether it can also occur during oligomer assembly. In the Bacillus thuringiensis binary toxin Vip1/Vip2, density maps and structures have been reported not only for the pore complex of the membrane-binding component Vip1 bound to the enzymatic component Vip2, but also for oligomerization intermediates ranging from trimers to heptamers (27). In these Vip1 intermediates, the protomers possessed partially formed β-barrel corresponding to the pore state. By contrast, oligomerization intermediates reported for CDTa-unbound C. difficile CDTb protomers retain D2s within individual protomers corresponding to the initial state (41). These observations suggest that binary toxin pore maturation may occur either after initial state protomers assemble into a prepore such as Prepore I, or during oligomer formation. The Prepore I–VIII series observed here demonstrates that stepwise maturation can occur at least after heptamer formation, although the timing of this structural transition remains to be clarified.

Comparison of the initial, intermediate, and pore states revealed that the major structural rearrangements associated with Ib pore maturation are concentrated in D2 (Fig. 3 A and B). In particular, the movement of α5 appears to represent a central structural event that links remodeling of the protomer–protomer interface (Fig. 3 D and E), mobilization of D2s, and assembly of the ϕ-clamp. During the transition from the initial to the intermediate state, α5 shifted toward the central axis of the oligomer, accompanied by displacement of the β-sheet containing β7, β8, and the downstream β10 strand (Fig. 3 B and C). This movement closed the gap between adjacent protomers and established a protomer–protomer interface resembling that observed in the pore state. The absence of D2s density in the intermediate state may therefore reflect increased mobility of D2s as it becomes ready to form the β-barrel.

The ϕ-clamp forming residues are located near α5, suggesting that α5 movement is also coupled to ϕ-clamp assembly. In fact, the ϕ-clamp is not formed in Prepore I, whereas an immature ϕ-clamp is observed in Prepore VIII, and a mature ϕ-clamp is formed in the pore state (Fig. 3H). These observations suggest that inward displacement of α5 brings the ϕ-clamp forming residues toward the center of the oligomer, inducing ϕ-clamp assembly (Fig. 3 B and C). The subtle structural differences in α5 and the ϕ-clamp between Prepore VIII and the pore state further support a close relationship between these structural elements. Thus, the ϕ-clamp may not simply function as a constriction site for enzymatic component translocation, but may also participate directly in maturation of the prepore into the pore.

This interpretation is further supported by analysis of the D452A mutant. In wild-type Ib, the side chain of D452 is oriented toward the interior of the ϕ-clamp and contributes to its stabilization through hydrogen-bonding interactions. Whereas monodisperse wild-type Ib oligomers were obtained exclusively in the pore state (25), alanine substitution of this residue stabilized the prepore state, suggesting that disruption of ϕ-clamp formation impairs pore maturation and β-barrel formation. Cryo-EM analysis of Ib D452A revealed a prepore structure corresponding to Prepore VIII but lacking a clearly formed ϕ-clamp (Fig. 4 C and D), as well as pore particles in which the ϕ-clamp adopted a conformation distinct from that of the wild-type pore (Fig. 4 E and F). Whereas wild-type Prepore VIII contains an immature ϕ-clamp (Fig. 3H), the ϕ-clamp was not clearly formed in D452A, consistent with impaired ϕ-clamp stabilization caused by the mutation. In addition, the α5 density in the D452A prepore was less well defined than that in wild-type Prepore VIII (SI Appendix, Fig. S6), supporting a coupling between ϕ-clamp formation and α5 stabilization. On the other hand, the presence of the ϕ-clamp in the D452A pore suggests that ϕ-clamp formation constitutes a key structural checkpoint in the transition from Prepore VIII to the pore state. Together with previous findings that multiple ϕ-clamp mutations in anthrax toxin PA inhibit the prepore-to-pore transition, these results support a close mechanistic relationship between ϕ-clamp assembly and pore formation.

Nevertheless, it is important to consider that the multiple prepore states observed here were captured in the context of Ib-rosette formation. The spatial arrangement of neighboring Ib oligomers within the Ib-rosette may have sterically restricted β-barrel formation and stabilized intermediates that are normally transient and unstable. Thus, the possibility that some of the observed states were artificially induced by rosette formation cannot be excluded. To address this concern, we also determined a prepore structure of monodisperse Ib oligomers using the D452A mutant, independently of Ib-rosette formation. Its close agreement with Ib-rosette-derived Prepore VIII confirms that Prepore VIII is not an artifact of rosette formation. Thus, rather than creating entirely aberrant structural states, the Ib-rosette may have acted as a scaffold that trapped unstable intermediates arising during the normal prepore-to-pore transition.

On the basis of these structural snapshots, we propose a model in which Ib pore maturation proceeds in a stepwise and locally propagated manner. Jiao et al. showed for perforin that directional conformational changes may proceed clockwise, based on time-resolved atomic force microscopy observations of the same molecule combined with molecular dynamics simulations (42). In the present study, the number of intermediate state protomers increased stepwise, and these protomers tended to be adjacent to one another, raising the possibility that a similar locally propagated maturation process occurs in Ib. However, cryo-EM analysis alone cannot unambiguously trace the correspondence of individual protomers between different prepore classes. Therefore, whether the initial–intermediate transition proceeds in the clockwise or counterclockwise direction remains unresolved. It also remains unclear whether maturation is initiated at a single site, as observed in Prepore II, or at multiple sites. Although no structure corresponding to the latter scenario was identified, its existence cannot be excluded. Future studies combining single-molecule observations with molecular simulations should help define the directionality and initiation mechanism of pore maturation in greater detail.

PFPs are major components of bacterial toxins and the immune system, and have become intriguing targets in both fundamental research and applications, such as their use as nanodevices in nanopore sequencing. However, these complicated proteins are produced as soluble monomers that mature into functional pores through dynamic structural changes. Binary toxins possess unique characteristics among pore-forming toxins that enable the translocation of enzymatic components into target cells via pores, making them highly applicable. Research is being conducted to create carriers by conjugating proteins to the binding domain of the enzymatic component, facilitating intracellular delivery of desired molecules. This study reports a wild-type structural series during the pore-forming process of a binary toxin, obtained through detailed cryo-EM classification that had previously proven difficult. Furthermore, we expect that the findings of this study will facilitate our understanding of the maturation processes of PFPs and contribute to the advancement of nanodevice development based on this structural foundation.

Methods

Expression and Purification of Wild-Type Ib and Ib D452A.

Wild-type Ib was expressed with an N-terminal 20-kDa pro-sequence, which prevents oligomer formation, and a C-terminal His₆ tag. The gene encoding Ib (UniProt ID: Q46221; residues 40 to 875), lacking the native signal sequence, was cloned into the pET23a vector without any additional tags. For Ib D452A, the same Ib coding region except for the D452A mutation was cloned into the pGEX-4T-1 vector as an N-terminal GST fusion in the order GST tag–pro-sequence–Ib D452A. Both constructs were transformed into Escherichia coli BL21 (DE3) cells and cultured in LB medium supplemented with 50 μg/mL ampicillin at 37 °C. When the cultures reached an OD₆₀₀ of 0.6, protein expression was induced by adding IPTG to a final concentration of 1 mM, followed by incubation for an additional 16 h at 23 °C. The cells were then harvested by centrifugation.

For purification of wild-type Ib, the harvested cells were resuspended in buffer A containing 20 mM Tris-HCl (pH 8.0), 100 mM NaCl, 1.0 mM CaCl₂, and 20 mM imidazole, supplemented with one cOmplete ULTRA protease inhibitor tablet (Roche, Basel, Switzerland), and lysed by sonication on ice. The lysate was clarified by ultracentrifugation at 180,000×g for 40 min, and the soluble fraction was applied to a Ni-NTA affinity column. The column was washed with 10 column volumes of buffer A, and the bound protein was eluted with buffer B containing 20 mM Tris-HCl (pH 8.0), 100 mM NaCl, 1.0 mM CaCl₂, and 500 mM imidazole. The eluate was concentrated using a centrifugal ultrafiltration device with a 50-kDa molecular-weight cutoff and buffer-exchanged into buffer C containing 20 mM Tris-HCl (pH 8.0), 100 mM NaCl, and 1.0 mM CaCl₂.

For purification of Ib D452A, the harvested cells were resuspended in buffer D containing 20 mM Tris-HCl (pH 8.0), 150 mM NaCl, 2 mM CaCl₂, and 5 mM dithiothreitol, supplemented with 1 mM phenylmethylsulfonyl fluoride, and lysed by sonication on ice. The lysate was clarified by ultracentrifugation at 180,000×g for 40 min, and the soluble fraction was applied to a glutathione Sepharose 4B column. The column was washed with buffer D, and the bound protein was eluted with buffer E containing 20 mM Tris-HCl (pH 8.0), 5 mM dithiothreitol, and 10 mM reduced glutathione. The eluted fractions were concentrated using a centrifugal ultrafiltration device with a 50-kDa molecular-weight cutoff and buffer-exchanged into buffer C.

Sample Preparation of Ib-Rosette.

Monomeric Ib was incubated at room temperature for 1 h with chymotrypsin at a chymotrypsin-to-Ib mass ratio of 1:1,000. The reaction was stopped by adding phenylmethylsulfonyl fluoride (PMSF) to a final concentration of 1 mM to inhibit chymotrypsin activity. Ethanol was then added to a final concentration of 10% (v/v), and the mixture was incubated at 37 °C for 1 h to promote Ib oligomerization. The Ib oligomer solution (450 μL) was applied to Ni-NTA resin, and the resin was washed with buffer C to remove ethanol. Prepared liposomes (540 μL), generated by drying E. coli polar lipid extract (Sigma-Aldrich) under N₂ gas, resuspending the lipids in buffer C, and extruding the suspension through a 100-nm filter using an extruder (Sigma-Aldrich), were then added together with 360 μL of buffer C and 100 μL of 1 M MES (pH 5.5), followed by incubation at 4 °C for 16 h. The resin was washed with ten column volumes of buffer A, and the bound protein was eluted with buffer B. The eluate was subjected to gel filtration using a HiLoad 16/600 Superdex 200 pg column equilibrated with buffer C at a flow rate of 0.8 mL/min. Fractions containing Ib oligomer bands were identified by SDS–PAGE, pooled, and concentrated to 1.0 mg/mL using a 50-kDa ultrafiltration membrane.

Sample Preparation of Ib D452A Oligomer.

Monomeric Ib was incubated at room temperature for 1 h with chymotrypsin (Sigma-Aldrich) at a 1/1,000 mass ratio. The reaction was stopped by adding PMSF (phenylmethylsulfonyl fluoride) at a final concentration of 1 mM to inhibit chymotrypsin activity. Ethanol and LMNG were then added to final concentrations of 10% (v/v) and 0.03% (w/v), respectively, and the mixture was incubated at 37 °C for 1 h to promote Ib oligomerization. It was loaded onto a 10 to 30% glycerol density gradient prepared using Buffer F (10% glycerol, 50 mM HEPES (pH 7.5), 100 mM NaCl, 1 mM CaCl₂, and 0.03% (w/v) LMNG) and Buffer G (30% glycerol, 50 mM HEPES (pH 7.5), 100 mM NaCl, 1 mM CaCl₂, and 0.03% (w/v) LMNG) and ultracentrifuged at 230,139×g for 16 h. After measuring the absorbance of each fraction at 280 nm, the peak fractions were collected and replaced with buffer F (10 mM HEPES (pH 7.5), 1 mM CaCl₂, and 0.003% (w/v) LMNG) using PD10. Finally, Ib D452A oligomers were concentrated to 1.2 mg/mL using a 50 K ultrafiltration membrane.

Data Collection With Cryo-EM.

For the Ib-rosette and Ib D452A samples, 3 μL of the prepared sample was applied to hydrophilized Quantifoil R1.2/1.3 300-mesh Cu grids. The grids were blotted for 7 s for Ib-rosette and 3.5 s for Ib D452A using a Vitrobot Mark IV (Thermo Fisher Scientific) and rapidly frozen in liquid ethane. Grid preparation conditions were screened by single-particle analysis using a Glacios microscope (Thermo Fisher Scientific) equipped with a Falcon 4 direct electron detector and operated using EPU software at the Institute for Life and Medical Sciences, Kyoto University.

Final datasets were collected using SerialEM (43) on a Titan Krios microscope (Thermo Fisher Scientific) equipped with a K3 direct electron detector (Gatan), a BioQuantum energy filter (Gatan) with a 20-eV slit width, and a Cs corrector (CEOS GmbH) at the Institute for Protein Research, Osaka University. For the Ib-rosette sample, 11,549 movies were acquired at a pixel size of 0.88 Å per pixel. For Ib D452A, 4,073 movies were acquired at a pixel size of 0.87 Å per pixel with the specimen stage tilted to 30° to mitigate preferred particle orientation. Detailed imaging conditions are provided in Table 1.

Table 1.

Cryo-EM data collection statistics

EMDB ID	Pore EMD-61772	Prepore I EMD-61771	Prepore II EMD-61774	Prepore III EMD-61773	Prepore IV EMD-61775	Prepore V EMD-61776	Prepore VI EMD-61777	Prepore VII EMD-61778	Prepore VIII EMD-61779	D452A Prepore EMD-82472	D452A Pore EMD-82471
Magnification	81,000
Voltage (kV)	300
Electron exposure (e−/Å2)	50
Defocus range (μm)	−0.6 to −1.6									−0.8 to −1.4
Pixel size (Å)	0.88									0.87
Symmetry imposed	C7	C7	C1	C1	C1	C1	C1	C1	C7	C7	C7
Initial particle images (No.)	6,223,563									830,786
Final particle images (No.)	161,249	10,511	15,199	22,352	8,878	11,285	7,672	23,305	17,097	93,962	248,921
Map resolution (Å)	2.47	3.04	3.35	3.32	3.57	3.59	3.73	3.32	2.90	2.85	2.56
FSC threshold	0.143

Single Particle Analysis of Ib-Rosette.

A series of analyses of the cryo-EM images was conducted using cryoSPARC (ver. 4.5.3) (33). Motion correction was performed using Patch Motion Correction, and CTF estimation was performed using Patch CTF Estimation (SI Appendix, Fig. S1A). Particle images were picked using a Template Picker with templates generated from a low-pass-filtered, previously reported Ib-pore density map (EMD-0721). The particle images were extracted and downscaled to 1/4 of their original size. These extracted particles were roughly classified through Heterogeneous Refinement with C7 (sevenfold rotational) symmetry applied (SI Appendix, Fig. S1B), resulting in a distinction between the prepore class, which lacks a β-barrel (SI Appendix, Fig. S1C), and the pore class which has a β-barrel (SI Appendix, Fig. S1D). The detailed analysis of the prepore class is described later in the “Classification of Ib-prepore” section. More homologous particle images were selected for the pore class through 3D Variability Analysis (3DVA) (44) and 3D classification without alignment sampling. The resulting particle set was re-extracted without rescaling, and a 3D density map was reconstructed using Homogeneous Refinement. Additionally, Local Motion Correction (45) was performed to correct for per-particle motion, followed by Homogeneous Refinement of images with defocus correction (46) (Table 2).

Table 2.

Atomic model refinement and validation statistics

PDB ID	Pore 9JSG	Prepore I 9JSF	Prepore II 9JSI	Prepore III 9JSH	Prepore IV 9JSK	Prepore V 9JSL	Prepore VI 9JSM	Prepore VII 9JSN	Prepore VIII 9JSO	D452A Prepore 44BW	D452A Pore 44BV
Refinement
Model resolution (Å)	2.6	3.2	3.6	3.6	3.8	3.8	4.0	3.6	3.2	3.0	2.6
FSC Threshold	0.5
Model composition
Nonhydrogen atoms	28,539	27,433	26,406	26,034	25,393	25,314	24,760	24,626	24,619	23,863	26,957
Protein residues	3,710	3,563	3,427	3,374	3,287	3,275	3,198	3,179	3,171	3,080	3,451
Ligands	21	21	21	21	21	21	21	21	21	21	21
B factor
Protein (Å2)	57.22	10.83	71.90	42.04	29.51	31.24	27.34	52.07	68.30	81.16	55.87
Ligand (Å2)	60.40	10.59	62.36	31.29	21.57	18.72	16.44	37.65	49.94	61.75	41.99
RMS deviations
Bond lengths	0.008	0.004	0.003	0.003	0.003	0.002	0.002	0.003	0.004	0.004	0.002
Bond angles	1.096	0.968	0.510	0.526	0.506	0.481	0.517	0.561	0.925	0.966	0.463
Validation
MolProbity score	1.36	1.53	1.64	1.66	1.67	1.59	1.59	1.53	1.55	1.76	1.46
Clash score	3.63	4.73	6.69	6.61	7.11	6.15	7.20	5.77	7.09	6.62	4.50
Poor rotamers (%)	0.00	0.22	0.07	0.07	0.10	0.21	0.18	0.53	0.00	1.39	1.09
Ramachandran plot
Favored (%)	96.67	95.83	96.02	95.69	95.92	96.28	96.80	96.65	97.10	95.85	96.73
Allowed (%)	3.33	4.17	3.98	4.31	4.08	3.72	3.20	3.29	2.90	4.15	3.27
Disallowed (%)	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.06	0.00	0.00	0.00

Classification of Ib-Prepore.

3DVA was performed on the prepore class, resulting in the classification of circular oligomers into five distinct subclasses ranging from an expanded state to a more constricted state (SI Appendix, Fig. S2A). Class 5 which is the most constricted class, exhibited a structure partially forming a β-barrel, similar to the pore state; thus, further analysis was not conducted for this class. The remaining classes 1 to 4 were subjected to Symmetry Expansion with C7 symmetry, where the particle images were symmetrically replicated seven times. A mask was created focusing on one protomer, and the 3DVA of the protomer was used to distinguish between classes with and without an observable density corresponding to the D2s within the protomer (SI Appendix, Fig. S2 B and C). For Classes 1 and 2, the maps were reconstructed using Homogeneous Reconstruction Only with the particle sets where D2s were not visible, whereas Classes 3 and 4 used particle sets where D2s were visible. These reconstructions were references for refining the original particle sets before replication using Non-uniform Refinement. Among the resulting density maps, protomers where D2s were ambiguously visible, were subjected to focused 3D classification without alignment sampling to separate classes where D2s were visible from those where they were not; this process was iteratively repeated (SI Appendix, Fig. S2 E–G). Ultimately, the prepore states were classified into eight distinct classes, each differing in the number of protomers with clearly visible D2s. Finally, Non-uniform Refinement (47) and Local refinement were performed for each of the eight prepore classes. The local resolution was estimated with the RELION (48) Local Resolution job (SI Appendix, Fig. S3).

Single Particle Analysis of D452A Prepore and Pore.

Cryo-EM movies were processed using cryoSPARC (ver. 4.5.3) (33). Motion correction was performed using Patch Motion Correction, followed by CTF estimation using Patch CTF Estimation (SI Appendix, Fig. S5A). Initial particles were picked using Blob Picker, extracted with downscaling, and subjected to 2D classification. Selected particles were used for ab initio reconstruction and Heterogeneous Refinement. The resulting particle images were used to train a Topaz model, and particles were repicked using Topaz Extract. The repicked particles were extracted with downscaling and subjected to 2D classification, followed by Heterogeneous Refinement to separate the prepore and pore classes (SI Appendix, Fig. S5B). Particles assigned to both classes were then re-extracted without downscaling. The prepore and pore particle sets were independently subjected to 2D classification (SI Appendix, Fig. S5 C and D), Heterogeneous Refinement, and Homogeneous Refinement. Reference-based Motion Correction was then performed, followed by further Homogeneous Refinement and Non-uniform Refinement with optimization of per-particle defocus (SI Appendix, Fig. S5 E and F). For the prepore class, C7 symmetry expansion was additionally applied, followed by 3D Variability Analysis using a focused mask encompassing a single protomer. However, unlike the Ib-rosette prepore dataset, the particles could not be further separated into multiple distinct conformational classes.

Model Building.

As the initial state where D2s are folded within the protomer, have not been previously reported in Ib, the predicted models were used as the initial model. The structure predicted from the AlphaFold Database was edited to start the residue numbering from 1 after the signal sequence based on the amino acid sequence of Ib (UniProt ID: Q46221), and used as the initial atomic model for the initial state. The predicted model was fitted as a rigid body into the most clearly observed initial state density in the Prepore I map using UCSF ChimeraX (ver. 1.2.4) (49), followed by manual model refinement using Coot (ver.0.9.5) (50) and automated structural refinement using PHENIX (ver. 1.19.2) (51). For intermediate state and pore state, previously revealed Ib-pore structure (PDB ID: 6KLW) was used as the starting model and refinement was conducted against the respective density maps of Prepore II and Prepore VIII, and Pore in the same manner as described above. Although the D4-I structure of Ib has not been previously reported, its density was well observed in Prepore I. Therefore, the refined D4-I structure from Prepore I was used as the initial model for other oligomers and structural refinement was conducted again using PHENIX. Finally, the chain orientations were validated using MolProbity (52) for the models obtained using the above process.

For the D452A prepore and pore structures, the Ib-rosette Prepore VIII and Pore models, respectively, were used as starting models and rigid-body fitted into the corresponding D452A maps using UCSF ChimeraX. For D452A pore, at first, the ϕ-clamp forming residues which had different structure from wild-type pore were replaced to Ala to refine the main chain coordinates tracing the map. After iterative model refinement by real-space refinement in PHENIX and manual rebuilding in Coot, replaced Ala residues were reverted to the original residues. For the D452A prepore, the D4-I region was additionally adjusted using the dock-in-map procedure in PHENIX. Residues 451 to 463 in the prepore model and residues 328 to 364 in the pore model were omitted from the final models because no corresponding density was observed. The stereochemical quality of both final models was assessed using MolProbity.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

The Cryo-EM maps and coordinates were deposited to the Electron Microscopy Data Bank (EMDB) and Protein Data Bank (PDB) with the accession codes EMD-61772 (53) and PDB ID 9JSG (54) for pore state, EMD-61771 (55) and PDB ID 9JSF (56) for Prepore I, EMD-61774 (57) and PDB ID 9JSI (58) for Prepore II, EMD-61773 (59) and PDB ID 9JSH (60) for Prepore III, EMD-61775 (61) and PDB ID 9JSK (62) for Prepore IV, EMD-61776 (63) and PDB ID 9JSL (64) for Prepore V, EMD-61777 (65) and PDB ID 9JSM (66) for Prepore VI, EMD-61778 (67) and PDB ID 9JSN (68) for Prepore VII, and EMD-61779 (69) and PDB ID 9JSO (70) for Prepore VIII, EMD-82472 (71) and PDB ID 44BW (72) for the D452A Prepore, and EMD-82471 (73) and PDB ID 44BV (74) for the D452A Pore. The cryo-EM images are deposited to Electron Microscopy Public Image Analysis (EMPIAR) with the accession code EMPIAR-12333 (75).