Structural basis of mpox virus A30/H2 subcomplex formation

Significance

The recent mpox virus (MPXV) transmission has raised worldwide concern. MPXV A30 and H2 from fusion machinery are identified as an interaction pair playing indispensable roles in virus infection. Due to low-affinity binding between the two subunits, however, the structural basis of the A30/H2 subcomplex formation remains elusive. In this study, we solve the complex structure of MPXV A30/H2 at high resolution using single-chain H2-A30 fusion protein. Further structural and functional analyses reveal that H2-binding could induce large conformational changes in A30. We also show that H2-A30 fusion protein could elicit virus-neutralizing responses that are greater than either ectodomain-component individually or mixed. Taken together, these data enrich our knowledge on MPXV entry and facilitate the development of specific therapeutics.

Abstract

The continuous spread of mpox disease caused by mpox virus (MPXV) has posed great threat to global public health. The postattachment membrane fusion process of MPXV is mediated by a multimeric protein machinery, termed as entry-fusion complex (EFC). Among EFC components, A30 and H2 are the earliest identified interaction pair and play important roles in virus entry. Here, we determine the crystal structure of MPXV A30/H2 subcomplex via the tandem-fusion strategy, and show that A30 undergoes large conformational rearrangements upon H2 binding. Structural analysis reveals extended intersubunit interface and highly conserved intermolecular interactions. In vitro binding data further clarify key residues and elements involved in the A30/H2 subcomplex formation. Finally, we show that the H2-A30 fusion protein, superior to A30 ectodomain alone or the ectodomain-mixture of H2+A30, can induce more potent neutralizing-antibody responses which could inhibit viral infection. These data provide valuable information for the understanding of poxvirus EFC assembly and the H2-A30-based immunogen design and optimization.

Mpox virus (MPXV), which belongs to the Orthopoxvirus genus in the Poxviridae family, is the etiologic agent of the recently ongoing mpox disease (1, 2). Historically, this zoonotic pathogen was initially isolated from cynomolgus monkeys in Denmark in 1958, and its first infection in humans was reported in 1970 by the Democratic Republic of the Congo (3, 4). For a considerable period of time, MPXV mainly circulated in Central and West Africa, with only sporadic mpox cases documented in nonendemic countries (3, 4). However, in 2022, a global mpox outbreak emerged, accompanied by a dramatic increase in confirmed cases through human-to-human transmission (1, 5). As of 31 July 2025, more than 158,425 people from 138 countries or areas have suffered from MPXV infection, resulting in 399 deaths (https://worldhealthorg.shinyapps.io/mpx_global/#key-figures). The mpox epidemic has raised worldwide concern, becoming another major public health event after the notorious COVID-19 (6). Currently, three live vaccinia virus (VACV)-based vaccines (ACAM2000, MVA-BN, and LC16m8), originally developed to combat variola virus (VARV), have been authorized for MPXV prevention. However, these vaccines either pose safety concerns or require further evaluation on their protective efficacy against MPXV in human populations (7, 8). Concomitantly, tecovirimat, an antiviral agent approved for the treatment of VARV, has demonstrated limited efficacy against MPXV in recent clinical investigations (9). Therefore, to facilitate the development of specific vaccines and therapeutics targeting MPXV, it is urgent to dissect the viral surface architectures involved in virus entry, which represents a feasible target for antivirals.

MPXV is a large, enveloped, double-stranded DNA virus with a genome size of approximately 197 kb that encodes more than 190 proteins (10, 11). The biological characteristics of MPXV are similar to those of other poxviral members, including VARV and VACV (3, 12). These viruses replicate exclusively within the host cell cytoplasm and can generate two distinct infectious forms, the extracellular enveloped virions (EEVs) and intracellular mature virions (IMVs) (12–14). Despite having different numbers of membrane layers and surface proteins, both EEV and IMV require a common set of proteinaceous machinery [known as the entry-fusion complex (EFC)] that mediate virus-cell membrane fusion to complete postattachment step of entry (15, 16). Previous functional studies on VACV have shown that EFC is composed of 11 conserved transmembrane subunits, including A16 (17), A21 (18), A28 (19), F9 (20), G3 (21), G9 (22), H2 (23), J5 (24), L1 (25), L5 (26), and O3 (27) (protein nomenclatures based on VACV). Accordingly, mutant IMVs lacking individual EFC components, while capable of cell attachment, exhibit defects in fusion-machinery bipolarization and the subsequent process of hemifusion or full fusion (15, 16, 28). Interestingly, it was shown that poxvirus EFC does not share any sequence or structure homology with the widely investigated class I, II, and III viral fusion proteins, indicating that the enigmatic EFC would utilize a unique mechanism to mediate membrane fusion (29–31).

Among the multiple EFC components, three stable subcomplexes, namely A16/G9 (32), G3/L5 (33), and A28/H2 (23, 34), have been identified so far. These subcomplexes are shown to play an indispensable role in poxvirus entry (28). Thus far, the structures of the first two protein pairs [A16/G9 (35) and G3/L5 (36)] have been experimentally determined. As for the A28/H2 subcomplex, though an NMR structure for A28 (29) and a crystal structure for H2 (37) have been reported, the precise binding modes and atomic interaction details between A28 and H2 remain unknown. In addition, A28 was reported to be able to elicit neutralizing-antibody responses after immunization (12, 38). It is noteworthy that antibody responses induced by A28 were specifically enhanced in the presence of H2 (15, 39), thus making the A28/H2 complex a superior neutralizing antibody target and immunogen candidate when compared to A28 alone. However, due to the lack of high-resolution complex structure, the molecular mechanism underlying the enhanced immunogenicity of A28 in the presence of H2 remains to be investigated.

Here, we focused on MPXV A30 and H2 (which are homologs of VACV A28 and H2, respectively) and prepared a single-chain fusion protein containing both subunits for structural study. Via crystallography, we solved the complex structure of A30/H2 at atomic resolution and found that A30 showed large conformational changes upon H2 engagement. Further structural and functional analyses identified key amino acids and elements in A30 and H2 that mediate subcomplex formation. Finally, plaque reduction neutralization assays showed that the mouse sera collected after immunization with the H2-A30 fusion protein can effectively neutralize VACV IMV infection. These results suggested that our fusion protein could be efficiently targeted by neutralizing antibodies and used as a potential vaccine component for orthopoxvirus prevention.

Results

Design of Stable A30/H2 Complex for Structural Study.

Both MPXV A30 and H2 adopt a type II transmembrane topology, with an ectodomain at the C-terminus that mediates interactions between the two subunits (Fig. 1A). Thus, we first prepared full-length ectodomain proteins of A30/28-146 (spanning residues 28 to 146) and H2/52-189 (spanning residues 52 to 189) by inclusion body refolding, SUMO tag cleavage, and gel-filtration chromatography. In addition, guided by the individual structures of VACV A28 and H2 (29, 37), we further prepared two N-terminal truncated ectodomain proteins of A30/56-146 (spanning residues 56 to 146) and H2/91-189 (spanning residues 91 to 189) using the established protein-preparation strategy. On a calibrated Superdex 75 Increase column, these proteins were eluted at ~13.2 to 14.8 mL (Fig. 1B), indicating that these purified EFC components existed mainly as monomers in solution.

Fig. 1.

In vitro assembly of stable MPXV A30/H2 complex. (A) A schematic representation of the protein-engineering strategy used to yield MPXV A30 (Left panel) or H2 (Right panel) recombinant ectodomain proteins. The transmembrane domain (TM), the ectodomain, and the cytoplasmic domain (Cyto) are individually marked with the boundary-residue numbers. (B) Solution-behavior characterization of A30 and H2 ectodomain proteins in individual forms using Superdex 75 Increase 10/300 GL column. The representative 280-nm absorbance curves and the SDS-PAGE migration profiles of the pooled samples are shown. (C) Affinity determination between A30 (A30/28-146 or A30/56-146) and H2 (H2/52-189 or H2/91-189) via SPR. The representative real-time binding curves and the calculated mean K_D values ± SD from three independent experiments are shown. (D) Solution-behavior characterization of A30 and H2 ectodomain proteins in tandem-fusion states using Superdex 75 Increase 10/300 GL column. The representative 280-nm absorbance curves and the SDS-PAGE migration profiles of the pooled samples are shown.

Subsequently, we performed surface plasmon resonance (SPR) assays to measure the real-time binding kinetics between MPXV A30 and H2. Typical fast-on/fast-off binding features were recorded and the equilibrium dissociation constant (K_D) values were thus calculated via the steady-state affinity model. The results showed that full-length ectodomains of A30/28-146 and H2/52-189 readily interacted with each other but exhibited a limited binding ability (with the K_D value determined to be 16.33 ± 1.07 μM). The truncated A30/56-146 showed a significantly reduced binding capacity to H2/52-189 (with a K_D of 182.67 ± 9.29 μM, representing ~11-fold decrease in affinity), suggesting that the N-terminal region of A30 ectodomain should play an important role in subcomplex assembly. In contrast, though the H2/91-189 protein carried an N-terminal deletion, its capacity to engage A30, as demonstrated by the determined affinity values (19.07 ± 3.40 μM to A30/28-146 and 175.67 ± 11.24 μM to A30/56-146), remained essentially unchanged in comparison to that of the full ectodomain protein (Fig. 1C and SI Appendix, Table S1). Thus, the N-terminal region of H2-ecto is unlikely to interact directly with A30. The results also echo previous findings that the N-terminal element of VACV H2 ectodomain, though involved in EFC formation, is dispensable for A28 binding (37).

Given the relatively weak affinity between A30 and H2, we speculated that the subcomplex formed by the two components could be stabilized by engineering them into a single-chain fusion protein. In light of the better binding capacity recorded for A30/28-146, it was fused, via a flexible (GGGGS)₅ linker, to H2/52-189 and H2/91-189, respectively. The resultant fusion proteins were obtained by inclusion body refolding. The H2/52-189-A30/28-146 conjugate exhibited excellent homogeneity and stability in solution. The H2/91-189-A30/28-146 fusion protein, however, was unstable and partially degraded after purification (Fig. 1D). Therefore, we selected the H2/52-189-A30/28-146 complex for subsequent structural studies.

Overall Structure of A30/H2 Complex.

Through intensive crystallization screening, diffractable crystals of the H2/52-189-A30/28-146 conjugate were obtained. Initially, a dataset of 2.5-Å resolution was collected from crystals grown at pH 4.0. The structure was determined by molecular replacement and refined to R_work and R_free values of 0.211 and 0.258, respectively (SI Appendix, Table S2), revealing two 1:1 A30/H2 heterodimers in the asymmetric unit. Additionally, a second 2.5-Å-resolution dataset was collected from crystals grown under near-neutral pH condition (pH 6.8). Structure determination for this dataset yielded R_work and R_free values of 0.219 and 0.257 (SI Appendix, Table S2), with three pairs of A30/H2 subcomplexes present in the asymmetric unit. Further structural comparison demonstrated that the A30/H2 heterodimers obtained at both pHs were essentially of the same binding mode and that all the secondary structural elements from A30 and H2 could be well aligned (SI Appendix, Fig. S1). Thus, the acidic crystallization conditions do not alter the subcomplex conformation. In addition, the AlphaFold3 (40) modeling of the MPXV A30/H2 full-length proteins further support our experimental structure (SI Appendix, Fig. S2). In light of the high degree similarity between the two structures, we selected the two A30/H2 subcomplexes (designated as Complex 1 and Complex 2 hereafter) from the low-pH model for subsequent structural analyses.

Complex 1 contained A30 residues P53-L146 and H2 amino acids P54-N187, while Complex 2 was composed of A30 residues P53-L146 and H2 amino acids I65-S183 (SI Appendix, Fig. S3 A and B). The architectures of both A30 and H2 are composed of five anti-parallel β strands (β1–β5 in A30, β1′–β5′ in H2) with four helices (η1–η2 and α1–α2 in A30, α1′–α4′ in H2) surrounding them. Two conserved intradomain disulfide bonds (C75/C112 and C129/C139 in A30, C102/C148 and C162/C182 in H2) could be observed in each subunit for structural stabilization. Overall, A30 and H2 are sterically juxtaposed to each other, with each subunit using one structural side to contact the other molecule. In addition, the N-terminal elements (helices η1–η2 and its intervening loops) of A30 extend into the core region of H2, half-covering H2 and further stabilizing the subcomplex. The N-terminal α1’ helix within H2, however, is spatially distant from A30 and does not participate in subcomplex formation (Fig. 2A). Such structural observations were consistent with our in vitro binding data showing that deletion of the N-terminal region of A30, but not that of H2, impacted the interactions between the two subunits (Fig. 1C and SI Appendix, Table S1). We also compared Complex 1 and Complex 2 within the asymmetric unit, which revealed highly similar protein folding structures and intermolecular binding patterns. However, the N-terminal region of H2 showed a slight difference in orientation (Fig. 2B and SI Appendix, Fig. S4).

Fig. 2.

Structure of MPXV A30/H2 heterodimer at pH 4.0. (A) Cartoon representation of the subcomplex architecture formed between A30 (cyan) and H2 (violet). The secondary structural elements and the intramolecular disulfide linkages are labeled. (B) Structural comparison of individual A30 and H2 from the two complexes (Complex 1 and Complex 2) observed in the crystallographic asymmetric unit. (C) Superposition of the component structures from Complex 1 onto the previously reported VACV A28 (PDB code: 8GQO) (29) and H2 (PDB code: 8INI) (37) structures. Those elements exhibiting variant conformations are highlighted and marked.

We further compared our A30/H2 heterodimeric structure with the unbound A28 and H2 structures derived from VACV reported previously (29, 37). For amino acid sequence, both A30 and H2 of MPXV exhibit high sequence identities with other orthopoxviral homologs (~84 to 97% for A30 and ~92 to 100% for H2) (SI Appendix, Fig. S5 A and B). Accordingly, we found that, regardless of the long N-terminal α1’ helix which was not observed in the free H2 structure of VACV, the two H2 structures could be well aligned, showing an RMSD of ~0.8 Å for all the equivalent Cα pairs. Thus, the conformation of H2 remained unchanged pre- and post-A30-binding (Fig. 2C). Unexpectedly, superimposition of the A30/A28 structures revealed large conformational difference for the N-terminal loop. In addition, the η1-helix present in our structure was not observed in the free A28 structure of VACV (Fig. 2C). The results therefore highlight an interesting structural rearrangement in A30, especially at the N-terminal region, for subcomplex formation with H2.

Detailed Atomic Interactions Along A30/H2 Interface.

We then performed a detailed characterization of the atomic interactions between A30 and H2 to decipher the structural basis of the subcomplex assembly. On the whole, the two components buried a surface area of ~1,027 Ų in A30 and ~1,029 Ų in H2, respectively. Based on the H2 elements participated in subcomplex formation, we further allocated the extended intermolecular binding interface to three binding patches (Patch1, 2, and 3) (Fig. 3A). The first patch (Patch1) mainly involves two spatially adjacent loops (β3′/β4′ intervening loop and the C-terminal loop) in H2, which are positioned to the vicinity of the N-terminal loop in A30. H2-residues R126, T128-D130 and Y172, and A30-amino acids P53-A54, D56, R58-V59, and D64 cluster within this patch, forming an extensive network of hydrophilic and van der Waals (vdw) interactions. Five interchain hydrogen bonds (H2-protein R126 with A30-protein V59, T128 with D56, T128 with R58, D130 with A54, and Y172 with A54) and one salt bridge (H2-protein R126 with A30-protein D64) are observed to form, further consolidating the engagement (Fig. 3B). The second patch (Patch2) involves strand-β4′ and the β3′/β4′ and β4′/α2′ intervening loops in H2. Amino acids W132, K135, A137, and D141 from H2 and residues D61, N63, D64, V69, and K72 from A30 are packed together, providing three hydrogen bonds (H2-protein W132 with A30-protein D61, K135 with D64 and K135 with K72), one salt bridge (H2-protein D141 with A30-protein K72), and multiple vdw interactions (Fig. 3C). The third patch (Patch3) encompasses the largest number of residues, forming diverse types of intermolecular contacts (with four hydrogen bonds, and multiple hydrophobic and vdw contacts). These include the H2-residues S143, Q146-F147, Y150, and K153-H154 located in helix-α2′ engaging with the A30-amino acids K72-C75, R110, I113-D114, and F117 (Fig. 3D).

Fig. 3.

The conserved interaction details between A30 and H2 ectodomains. (A) An overview of the subcomplex interface, that is further subdivided into three patches (Patch1–Patch3) based on the H2 elements involved in A30 engagement. A30 and H2 subunits are shown in surface and cartoon representations, respectively. (B) Atomic binding details in Patch1. (C) Atomic binding details in Patch2. (D) Atomic binding details in Patch3. Residues providing ≥10 vdw contacts (the distance cutoff is 4.5 Å), hydrophilic bonds (the distance cutoff is 3.2 Å), or hydrophobic interactions are shown and labeled. Hydrogen bonds or salt bridges are indicated as dashed lines. (E) Multiple sequence alignment of the A30 homologs from orthopoxviruses. Key residues in MPXV A30 that interact with H2 are marked with blue squares. (F) Multiple sequence alignment of the H2 homologs from orthopoxviruses. Key residues in MPXV H2 that interact with A30 are marked with violet squares. Abbreviations: VACV (vaccinia virus), VARV (variola virus), CPXV (cowpox virus), CMLV (camelpox virus), AKMV (Akhmeta virus), VPXV (volepox virus), ECTV (ectromelia virus), Abatino (abatino macacapox virus), RCNV (raccoonpox virus), SKPV (skunkpox virus), TATV (taterapox virus).

It is noteworthy that all the 15 amino acids in H2 and 15 out of the 17 residues in A30 involved in subcomplex assembly are identical among the representative orthopoxviral members (Fig. 3 E and F, and SI Appendix, Fig. S6 A and B). These results indicated that the characterized MPXV A30/H2 binding modes, as well as the atomic intermolecular interactions, should be shared by the other orthopoxviruses.

Key Residues and Elements Mediating A30/H2 Subcomplex Formation.

To verify the A30/H2 assembly pattern observed in our structure, we focused on representative interface residues that are shown to form hydrogen bond/(s) and salt bridge or to contribute a large number of hydrophobic stacking interactions via their side chains (Fig. 4A). These amino acids include R126, W132, H154, F147, Y150, and Y172 in H2, and D114 and F117 in A30, which are replaced in the H2/52-189 and A30/28-146 constructs with alanines. A series of single- or double-residue mutant proteins were prepared (SI Appendix, Fig. S7) and subsequently tested for their real-time binding kinetics via SPR.

Fig. 4.

Biochemical analysis validating the critical residues and elements at the A30/H2 interface. (A) Two-dimensional diagrams [modified from LigPlot+ (41)] of key amino-acid interactions (hydrophilic bonds or hydrophobic contacts) between subunits, with the dashed arrows denoting hydrogen bonds or salt bridge. (B) An SPR assay characterizing the binding kinetics between the paired A30 and H2 proteins. Gradient concentrations of A30 (A30/28-146, A30/28-146-D114A, A30/28-146-F117A, A30/52-146, or A30/70-146) were flowed through H2 (H2/52-189-R126A, H2/52-189-W132A, H2/52-189-H154A, H2/52-189-Y172A, H2/52-189-F147A/Y150A, or H2/52-189) that immobilized on the sensor-chip surface. The representative real-time binding curves and the calculated mean K_D values ± SD from three independent experiments are shown.

For the five H2 mutant proteins, the calculated affinity values were 399.33 ± 11.68 μM for H2/52-189-R126A, 532.67 ± 92.40 μM for H2/52-189-W132A, 199.33 ± 29.57 μM for H2/52-189-H154A, 176.67 ± 5.51 μM for H2/52-189-Y172A and 657.33 ± 236.77 μM for H2/52-189-F147A/Y150A, respectively. Compared with the affinity between wild-type H2/52-189 and A30/28-146 (16.33 ± 1.07 μM), these values represented >10-fold decrease in binding capacity toward A30. For the two A30 mutant proteins, the K_D values were determined to be 94.93 ± 19.17 μM for A30/28-146-D114A and 1184.00 ± 300.25 μM for A30/28-146-F117A, respectively (Fig. 4B and SI Appendix, Table S1). The data therefore also featured a significant decrease in affinity upon introduction of alanines to A30-residues D114 or F117. These concerted affinity reductions demonstrate the importance of these residues in A30/H2 subcomplex formation.

It is worth noting that in the H2-A30 fusion protein, the N-terminus of A30 is initiated from E28 (Fig. 1A). However, in the crystal structure we solved, the N-terminal loop of A30, whether in Complex 1 or Complex 2, starts from P53 (SI Appendix, Fig. S3). This untraceable density of the E28-T52 region in A30 may result from intrinsic flexibility, protein truncation, alternate conformations not visible at current resolution, or partial unfolding/misfolding induced by crystallization conditions (e.g., pH, salt concentration). In contrast, the N-terminal elements of A30 (P53-V69), with definitive densities, extend into the center of H2 protein and provide multiple intermolecular interactions with H2 (Fig. 3E and SI Appendix, Fig. S8). Therefore, we further designed and prepared two truncated versions of A30, namely A30/52-146 (eliminating the untraceable N-terminal loop of A30) and A30/70-146 (eliminating the whole N-terminal elements including those interact with H2) (SI Appendix, Fig. S7), to investigate the role of the deleted structural elements in A30/H2 engagement. As expected, A30/52-146 showed similar binding capacity to H2/52-189 (7.32 ± 0.72 μM in K_D) as A30/28-146, while the K_D value between A30/70-146 and H2/52-189 was reduced to only 702.33 ± 234.02 μM (Fig. 4B and SI Appendix, Table S1). It is noteworthy that A30/56-146 (only four-residue short in comparison to A30/52-146) exhibits a significantly reduced activity for binding with H2/52-189 (Fig. 1C and SI Appendix, Table S1). This decrease is due to the deletion of A54 residue, which eliminates the hydrogen-bond interactions between the peptide bond of A54 and the side chains of D130 and Y172 in H2 protein (Fig. 4A).

In summary, our in vitro binding results are highly consistent with the atomic binding details delineated in the subcomplex structure, in turn confirming the observed interaction mode between A30 and H2.

Enhanced Serum Neutralization in Mice Immunized with H2-A30 Fusion Protein.

While VACV A28 has been recognized as a target for neutralizing antibodies (38), it is interesting that neutralizing antibody response induced by A28 is shown to be specifically enhanced in the presence of H2 (39). Echoing these studies, our structural analyses revealed large conformational differences for A30 N-terminal elements prior to and after H2 binding (Fig. 2C). This has raised an interesting notion that H2 binding could prepare and stabilize A30 to elicit better protective antibody responses. In light of the relatively weak binding affinity between A30 and H2 ectodomains, we speculated that the stable H2-A30 fusion protein likely represented a better immunogen. Thus, purified proteins, including single H2/52-189 and A30/28-146, the H2/52-189+A30/28-146 mixture and the H2/52-189-A30/28-146 fusion antigen, were emulsified with Freund’s adjuvant (38, 42,43, 44) and administered intramuscularly to BALB/c mice. Immunizations were delivered in three doses at 3-wk intervals, with each dose consisting of either 10 µg (low dose) or 20 µg (high dose) protein. Serum samples were collected at day 40 (after dose 2) and day 54 (after dose 3) since the initial immunization (Fig. 5A). Subsequently, an enzyme-linked immunosorbent assay (ELISA) was performed to calculate the antigen-specific IgG titers.

Fig. 5.

Enhanced serum neutralization in mice immunized with H2-A30 fusion protein. (A) The schematic representation of mouse immunization and serum collection protocol [created by BioRender (https://www.biorender.com/)]. Mice (n = 5/group) were vaccinated intramuscularly on days 0, 21, and 42 with either 10 µg (low dose) or 20 µg (high dose) of the antigen proteins, including single H2/52-189 and A30/28-146, the H2/52-189+A30/28-146 mixture and the H2/52-189-A30/28-146 fusion protein, or with PBS as control. The sera were collected on days 40 and 54. (B) The binding of immune sera (collected after the second or third dose) to the indicated antigens was quantified using ELISA. Endpoint IgG titers are determined and presented as the mean ± SD. (C) The neutralizing activity of the sera collected on day 54 after the third-dose immunization was evaluated by a VACV IMV-based PRNT assay. Data are presented as mean ± SD. Data were analyzed by one-way ANOVA with a multiple-comparison test. P < 0.05 is considered statistically significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Regardless of immunization dose (10 or 20 µg) or timepoint (postsecond or postthird dose), serum samples from all immunization groups consistently exhibited potent antigen-binding capacity (Fig. 5B). In contrast, no antigen-binding activity was detected in sera from the PBS control group, thereby confirming that the robust antibody responses observed are specifically elicited by the H2 and A30 ectodomain components, either in their free or complexed forms.

We further carried out a plaque reduction neutralization test (PRNT) using VACV IMV to evaluate the neutralizing activity of the sera collected after the third-dose immunization. For each immunogen, both high- and low-dose vaccination groups showed comparable virus neutralization, with clearly dose-dependent responses (Fig. 5C). Immunization with single H2 ectodomain protein elicited high-titer binding antibodies but with largely negligible neutralizing capacity (Fig. 5 B and C). In contrast, enhanced neutralizing efficacy was observed in the single A30 ectodomain and the H2+A30 ectodomain-mixture groups. Notably, the H2-A30 fusion protein group displayed the most robust viral inhibition among all groups (Fig. 5C). Taken together, these immunological data suggest that the H2-A30 fusion protein, superior to A30 ectodomain alone or the ectodomain-mixture of H2+A30, induces more potent neutralizing-antibody responses capable of inhibiting viral infection. It is also notable that the H2-specific sera, though with high binding titers, exhibit no obvious neutralizing efficacy. We thus speculate that neutralizing antibody responses induced by the fusion protein should largely target the A30 component. Consistently, previous and recent studies on VACV A28/H2 immunization have similarly shown that A28-specific sera exhibit much stronger neutralization of viral infection than H2-specific sera (38, 39, 45).

Discussion

The recent mpox outbreak has led the World Health Organization (WHO) to declare twice a Public Health Emergency of International Concern (PHEIC) for the viral infection in the past 3 y (46). The pandemic potential of MPXV highlights the importance of pathogenic studies on the virion attachment and fusion process, which is a prerequisite for infection. Unlike typical class I, II, and III viral fusion proteins, which rely on single or a few surface component/(s) for membrane fusion, the fusion associated process following poxvirus attachment involves the assembly and regulation of numerous protein subunits (up to a dozen) (15, 30). Thus, the conserved EFC of poxviruses may possess a complex molecular architecture, an exquisite conformational-rearrangement mode, and a unique membrane fusion mechanism. However, the insufficiency of studies on poxviral EFC, coupled with the absence of structural data, have hindered the molecular understanding of the machinery’s functionality as well as the development of drugs to block its activity. A30/A28 (designated as A30 in MPXV and A28 in VACV) and H2 are among the earliest identified EFC components, and the subcomplex formed by the two subunits is one of the three known subcomplexes present in EFC (15, 19, 23). We believe the atomic structure of MPXV A30/H2 subcomplex reported in the current study, together with our previously determined VACV A16/G9 and G3/L5 heterodimeric structures (35, 36), should pave the way for future studies aiming to delineate the mechanistic details of poxviral EFC, facilitating the development of antivirals. Previous studies on other enveloped viruses (e.g., coronavirus, flavivirus, herpesvirus, etc.) have demonstrated that targeting viral fusion protein/(s) and the associated virus-cell membrane fusion process for inhibitor screening and vaccine design is a promising antiviral countermeasure (47–51).

Prior to this study, the atomic structures of free VACV A28 and H2 have been experimentally determined (29, 37). Via structure-guided mutagenesis, several residues or residue-combinations that are important for VACV A28/H2 engagement have been identified (29, 37). E.g., mutations of K72A, D56A/R58A/D114A/D119A, D61A/V62A/N63A/D64A, D68A/K72A/R74A, and N122A/Y126A/N127A in A28 would significantly compromise A28’s binding capacity for H2 and lead to impaired EFC-mediated membrane-fusion activity (29). Reciprocally, mutations of D130A, W132A, D141A, R125A/R126A, and G127A/G129A in H2 could also dramatically jeopardize the A28/H2 binding (37). Notably, these residues proposed to be involved in complex formation of VACV A28/H2, including A28-D56, -R58, -D61, -N63, -D64, -K72, -R74 and -D114, and H2-R126, -G129, -D130, -W132 and -D141, are also observed to locate along the binding interface in our structure (Fig. 3 B–D). Given the high amino acid conservation of A30/A28 and H2 between VACV and MPXV (~97% sequence identity for A30/A28, and ~99% sequence identity for H2), we believe the two viruses, and likely other orthopoxviral members, share the same interface for subcomplex assembly between A30/A28 and H2.

Upon receptor recognition or low-pH induction, the fusion protein on the surface of the enveloped virus can undergo a series of conformational changes, inducing the insertion of its hydrophobic fusion peptide or fusion loop into the host membrane for hemifusion, then the virus-cell membrane fusion could be accomplished (30, 31). VACV IMV carrying H2 defect fails to complete hemifusion after attachment, indicating that H2 is involved in the initial stage of membrane fusion (28). Further studies have found that there are two fully conserved fusion peptide-like sequences (¹²⁵RRGTGDAW¹³² and ¹⁷⁰LGYSG¹⁷⁴) in VACV H2, and that these two elements are important for H2-A28 engagement and EFC-mediated membrane-fusion activity (34, 37). These two elements, though individually positioned between strands β3′ and β4′, and helices α3′ and α4′, are spatially close to each other (SI Appendix, Fig. S9). In our structure, the ¹²⁵RRGTGDAW¹³² peptide is fully concealed by the N-terminal loop of A30. This embedded motif is well stabilized by several hydrogen bonds, involving multiple residues in the motif including R126, T128, D130, and W132 (Fig. 3 B and C). For the ¹⁷⁰LGYSG¹⁷⁴ peptide, it is also shielded, though partially, by A30. The Y172 residue is shown to form a hydrogen bond with A54 of A30 (Fig. 3B), thereby stabilizing the conformation and position of the motif. Therefore, these fusion peptide-like elements of functional significance in H2 are clearly buried by A30, thereby preventing premature exposure (SI Appendix, Fig. S9).

Although the crystal structures solved in this study are determined using an H2-A30 fusion protein connected by an exogenous linker, our subsequent structural analysis allowed us to design a series of mutant and truncated proteins based on the interaction interface revealed in the structure. As expected, further SPR assays show that the affinity values between these mutated/truncated proteins and their wild-type partners match well with the key residues and elements identified in the structure. Though it can not be excluded that the linker might still pose impact/(s) on the complex structure away from the interface, it is unlikely that the linker would have perturbed the assembly of A30/H2 and affected the subcomplex binding mode. In support of this, the AlphaFold3 modeling of the subcomplex further supports our experimental structure (SI Appendix, Fig. S2). In fact, the strategy of converting multicomponent proteins into a single-chain format via an exogenous linker has been commonly used for the structural and functional investigations on viral envelope glycoprotein complexes (52,53, 54).

Notably, it has been suggested in a previous study that the A28/H2 interaction would stabilize the immunogenic form of A28, thereby significantly enhancing A28’s ability to induce neutralizing-antibody responses (39). Our binding data, however, shows that the binding capacity of A30 and H2 ectodomains in solution is rather limited, with a K_D value of 16.33 ± 1.07 μM. In this case, the immunogenic conformation of A30 ectodomain might not be stably maintained when it exists alone or simply in a mixture with H2 ectodomain. Accordingly, our immunological analyses demonstrate that the A30-specific binding-antibody titers are largely comparable among immunization groups with single A30 ectodomain protein, the H2+A30 ectodomain-mixture, and the H2-A30 fusion protein. However, sera from the group immunized with the fusion protein exhibits significantly enhanced neutralizing efficacy relative to the other two groups. These findings suggest that the covalent linkage of H2 ectodomain with A30 ectodomain may eliminate the defect of low-affinity binding between the two subunits and help lock A30 and H2 ectodomains together for A30 stabilization. While the A30/A28 subunit has been selected in the antigen recipe for the development of vaccines (12, 38, 39), we would propose to involve a H2-A30 fusion protein, rather than the single A30 ectodomain-antigen or the H2+A30 ectodomain-mixture, as a vaccine component for MPXV prevention.

It is also notable that previous studies (23, 34) showing interactions between VACV A28 and H2 are performed with the full-length proteins that include the transmembrane (TM) domains. In the full-length A30/H2 model predicted by AlphaFold3, we find that the TMs are arranged in close proximity (SI Appendix, Fig. S2), which might provide further intermolecular interactions. Thus, the TM domains might contribute to the stability of the subcomplex and increase the immunogenicity of the heterodimer. Accordingly, the previous neutralizing antibody studies are carried out with plasmids encoding proteins containing the TM domains (39). Therefore, our findings regarding the superiority of the fusion protein to the A30 and H2 homologs may only apply to ectodomains, not full-length proteins. Expression by mRNA or other methods of the full-length A30 and H2 proteins might be equal or superior to the fusion protein. In addition, the immunization assay in our study is conducted only with Freund’s adjuvant, a potent research-grade formulation. The immunogenicity of our fusion antigen when combined with other adjuvants remains to be tested in the future.

Materials and Methods

Cells and Viruses.

All Escherichia coli strains were cultured in lysogeny broth (LB) medium (0.5% w/v yeast extract, 1% w/v tryptone, and 1% w/v NaCl) using a nonhumidified shaker at 37 °C. BHK-21 (ATCC, CCL-10) and BS-C-1 (ATCC, CCL-26) cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM, Gibco) supplemented with 10% fetal bovine serum (FBS, ExCell) at 37 °C.

The initial stock of vaccinia virus (VACV, Western Reserve strain), which is deficient in the thymidine kinase (TK, J2R) gene and harbors an insertion of the enhanced green fluorescent protein (EGFP) gene, was commercially obtained from BrainVTA (Cat# VV01002). The virus was propagated in BHK-21 cells cultured in DMEM supplemented with 2% FBS. The IMV form of VACV was obtained from lytic cells. Briefly, BHK-21 cells were infected with VACV at a multiplicity of infection (MOI) of 0.5. At 72 h postinfection, the cells were harvested and subjected to three freeze–thaw cycles. The lysates were then centrifuged at 4,000 rpm for 15 min to remove cell debris. The resulting supernatant, containing the VACV IMV form, was collected and stored at −80 °C until use. All experimental procedures involving VACV were performed in a Biosafety Level 2 facility.

Plasmid Construction.

The coding sequences for the ectodomains of MPXV A30/28-146 (residues E28-L146, GenBank: URK20578.1) and H2/52-189 (residues D52-E189, GenBank: URK20529.1) were synthesized [General Biol (Anhui)] and individually subcloned into pET-28a-SUMO vector (an engineered pET-28a plasmid that contains a SUMO tag and a PreScission Protease cleavage site). The coding fragments for the modified A30 (including A30/52-146, A30/56-146, A30/70-146, A30/28-146-D114A, and A30/28-146-F117A), H2 (including H2/91-189, H2/52-189-R126A, H2/52-189-W132A, H2/52-189-H154A, H2/52-189-Y172A, and H2/52-189-F147A/Y150A), and H2-A30 [including H2/52-189-A30/28-146 and H2/91-189-A30/28-146 which were constructed by tandemly fusing H2/52-189 or H2/91-189 to A30/28-146 via a (Gly-Gly-Gly-Gly-Ser)₅ linker] were generated via a standard PCR-based strategy and also inserted into pET-28a-SUMO vector.

Protein Expression and Purification.

All the above proteins were expressed as inclusion bodies in E. coli and refolded in vitro, followed by removal of the SUMO tag. The recombinant plasmids were transformed into E. coli BL21 (DE3) for expression. The cells containing plasmids were then grown in LB medium supplemented with 50 µg/mL kanamycin at 37 °C and induced with 500 µM isopropyl-β-D-thiogalactopyranoside (IPTG) at 37 °C for about 6 h. Inclusion bodies were then extracted and refolded as previously described (55). Briefly, aliquots of inclusion bodies were diluted dropwise into a stirring refolding buffer consisting of 100 mM Tris-HCl (pH 8.0), 400 mM L-Arg HCl, 2 mM EDTA, 5 mM reduced glutathione and 0.5 mM oxidized glutathione, and incubated overnight to complete the refolding process. Next, the refolded proteins were concentrated using an Amicon Stirred Cell concentrator with a 10 kDa cutoff membrane and then adjusted to size-exclusion chromatography (SEC) buffer containing 10 mM HEPES-NaOH (pH 7.5) and 150 mM NaCl. Subsequently, the protein samples were cleaved by PreScission Protease with the protein-to-enzyme weight-ratio (w/w) of 50:1 in SEC buffer supplemented with 1 mM DTT at 4 °C for about 2 h. To eliminate the SUMO tag, the digested proteins were then loaded onto an anion-exchange column (SOURCE 15Q, Cytiva) and eluted with an increasing concentration of NaCl. Finally, the target proteins were collected and further purified by gel-filtration chromatography in SEC buffer using the Superdex 75 Increase 10/300 GL column (Cytiva).

Crystallization.

Commercial crystallization kits (Molecular Dimensions and Hampton Research) were used for initial crystallization screenings by the vapor-diffusion sitting-drop method. In brief, 1 µL protein sample (H2/52-189-A30/28-146 fusion protein) was mixed with 1 µL reservoir solution, and the resultant drop was then equilibrated against 70 µL reservoir solution at 18 °C. High-quality crystals for subcomplex (with protein concentration of 9 mg/mL) were grown in the following conditions with two different pHs. Condition 1 (pH 4.0, P2₁2₁2₁ space group): 0.2 M Ammonium acetate, 0.1 M Sodium acetate and 15% w/v PEG4,000. Condition 2 (pH 6.8, P22₁2₁ space group): 0.2 M Potassium nitrate and 20% w/v PEG3,350.

Data Collection and Structure Determination.

For data collection, crystals were flash-cooled in liquid nitrogen after a brief soaking in reservoir solution supplemented with 20% (v/v) glycerol. Diffraction data were collected at Shanghai Synchrotron Radiation Facility (SSRF) beamline BL18U1 (56). The collected data of pH-4.0 crystal were processed with HKL-2000 v706e (57) for indexing, integration, and scaling. The collected data of pH-6.8 crystal were processed by XDS v20230630 (58) and scaled with Aimless (59) from CCP4 Software Suite v8.0.002 (60). The subcomplex structures were solved by molecular replacement using Phaser program (61) from CCP4, with the search templates (MPXV A30 and H2) generated by AlphaFold2 (62). Initial restrained rigid-body refinement was performed using Refmac5 (63), which was followed by manual rebuilding and adjustment in Coot v0.9.8.1 (64), and water molecules were automatically added in Phenix.refine from PHENIX v1.19.2-4158 (65). Final statistics for data collection and structure refinement are summarized in SI Appendix, Table S2. All structural figures were generated using PyMOL v2.5.0 (https://pymol.org/).

Surface Plasmon Resonance (SPR) Assay.

All the SPR experiments were performed with the BIAcore 8 K system (Cytiva). Purified H2 proteins (H2/91-189, H2/52-189, H2/52-189-R126A, H2/52-189-W132A, H2/52-189-H154A, H2/52-189-Y172A, H2/52-189-F147A/Y150A) were individually immobilized onto the CM5 sensor chip (Cytiva) using the Amine Coupling Kit (Cytiva). Two-fold serially diluted concentrations of analytes (A30/28-146, A30/52-146, A30/56-146, A30/70-146, A30/28-146-D114A, A30/28-146-F117A) were flowed over the chip in the running buffer containing 10 mM HEPES-NaOH (pH 7.5), 150 mM NaCl, and 0.05% Tween-20 at a rate of 30 μL/min. The obtained kinetic data were analyzed with the Biacore Insight Evaluation Software v5.0.18.22102 (Cytiva) and the equilibrium dissociation constant (K_D) values were calculated using steady-state affinity model. For each binding pair, three independent assays were conducted, and the K_D values are summarized in SI Appendix, Table S1.

Mice Immunizations.

Forty-five 6- to 8-wk-old female BALB/c mice (free of specific pathogens) were purchased from Beijing Vital River Laboratory Animal Technology (licensed by Charles River) and were housed with five companions per cage at random. All mice used in this study were in good health and were not involved in other experimental procedures. Mice were allowed free access to water and standard chow diet as well as provided with a 12-h light and dark cycle (temperature of 20 to 25 °C, humidity of 40 to 70%). All animal studies were approved by the Institutional Animal Care and Use Committee of Sichuan University (Chengdu, Sichuan, China).

For immunization of mice, all mice were randomly divided into nine groups of five mice each. 10 µg (low dose) or 20 µg (high dose) purified proteins, including H2/52-189, A30/28-146, H2/52-189+A30/28-146 mixture (with the mass ratio of 1:1), or H2/52-189-A30/28-146 fusion antigen, were individually diluted in PBS and emulsified with equal volumes of complete Freund’s adjuvant (for first-dose immunization, Sigma) or incomplete Freund’s adjuvant (for second- or third-dose immunization, Sigma) to produce eight vaccine groups. Then BALB/c mice were vaccinated via the intramuscular injection on days 0, 21, and 42. At the same time, PBS control mice were injected intramuscularly with PBS. Venous bloods were collected through the infraorbital plexus on days 40 and 54, followed by centrifugation and serum collection. The serum samples were heat-inactivated at 56 °C for 30 min prior to further use.

Enzyme-Linked Immunosorbent Assay (ELISA).

The recombinant proteins, including H2/52-189, A30/28-146, H2/52-189+A30/28-146 mixture (with the mass ratio of 1:1), or H2/52-189-A30/28-146 fusion antigen, were used to coat flat-bottom 96-well plates (Corning) at a final concentration of 1 μg/mL in 50 mM carbonate coating buffer (pH 9.6) at 4 °C overnight, and blocked with PBST containing 5% nonfat powdered milk (Sangon Biotech) for 1 h at room temperature. Then, serially diluted serum samples were added to the plates and incubated for 1 h at 37 °C. Subsequently, the plates were washed three times with PBST and incubated with goat anti-mouse IgG-horseradish peroxidase antibody in a 1:5,000 dilution. After incubation for 1 h at room temperature, the plates were washed five times with PBST and reacted with 3,3′,5,5′-tetramethyl biphenyldiamine (TMB) for about 5 min. Reactions were stopped by 2 M hydrochloric acid and the absorbance was measured at 450 nm on a microplate reader (BioTek). The results were analyzed by using GraphPad Prism v9.5.1.

Plaque Reduction Neutralization Test (PRNT).

BS-C-1 cells were seeded 1 d prior to the experiment in 12-well plates at a density of 1 to 1.5 × 10⁵ cells per well. The sera obtained after third-dose vaccination were 20-, 40-, 80-, and 160-fold diluted by DMEM and further mixed with ~150 plaque-forming units (PFU) of VACV IMV for 1 h at 37 °C. Then, the virus-serum mixtures were added onto preseeded BS-C-1 cell monolayers and incubated for another 1 h at 37 °C in 5% CO₂ incubator. Subsequently, the supernatants were removed and the cell monolayers were overlaid with DMEM containing 2% FBS and 1% low-melting-point agarose (Solarbio). After 48 h, the plates were fixed with an equal volume of 10% neutral formalin for more than 1 h. Finally, the cell monolayers were stained with 0.5% crystal violet, and the plaques were counted and photographed. The results were analyzed by using GraphPad Prism v9.5.1.

Sequences Used in This Study.

The GenBank accession numbers of the sequences (A30 and H2 homologs) used for multiple alignment and conservation analysis are as follows: MPXV Clade I, GenBank: WZB48398.1 and WZB48351.1; MPXV Clade II, GenBank: URK20578.1 and URK20529.1; VACV (vaccinia virus), GenBank: AAO89430.1 and AAO89379.1; VARV (variola virus), GenBank: AAA60883.1 and AAA60833.1; CPXV (cowpox virus), GenBank: ADZ30341.1 and ADZ30293.1; CMLV (camelpox virus), GenBank: AAL73854.1 and AAL73805.1; AKMV (Akhmeta virus), GenBank: AXN74944.1 and AXN74893.1; VPXV (volepox virus), GenBank: AOP31837.1 and AOP31789.1; ECTV (ectromelia virus), GenBank: AAM92435.1 and AAM92388.1; Abatino (abatino macacapox virus), GenBank: AYN64714.1 and AYN64666.1; RCNV (raccoonpox virus), GenBank: AKJ93777.1 and AKJ93729.1; SKPV (skunkpox virus), GenBank: AOP31626.1 and AOP31578.1; TATV (taterapox virus), GenBank: ABD97718.1 and ABD97668.1.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Atomic coordinates and structure factors for the reported crystal structures have been deposited into the Protein Data Bank under accession codes 9L7W (66) and 9WLP (67). All other data are included in the manuscript and/or SI Appendix.