Isolation and characterization of antibodies targeting MPXV M1R and B6R

In this study, we employed phage surface display technology to identify a diverse and extensive array of antibodies targeting M1R and B6R proteins on the surface of MPXV (Fig. 1A). The M1R and B6R antibodies were isolated from a fully human antibody library (ST-ST-HuNAL, Fab format, Sanyou Bio), which was constructed using peripheral blood mononuclear cells (PBMCs) derived from more than 2000 healthy donors aged between 20 and 60 years. To isolate the target antibodies, we adopted a standard solid-phase immuno tube screening strategy (Fig. S1). The recombinant M1R or B6R proteins were used as bait, and three rounds of panning were conducted (Fig. S1 and Table S1), which respectively yielded 37 and 40 clones (Table S2), each exhibiting unique sequences. Subsequently, we sequenced the V regions and assessed the length of complementarity-determining region 3 (CDR3). Notably, IGHV3-23, IGHV3-30, IGHV3-66 and IGHV3-21, as well as IGKV1-12 and IGKV1-39, were found to be significantly prevalent among the M1R binding antibodies (Fig. 1B). Similarly, antibodies targeting B6R also exhibited high frequencies in the IGHV3-23, IGHV3-30, IGKV1-12, and IGKV1-39 germlines (Fig. 1C). Further analysis of the CDR3 length in these binding antibodies revealed values within the range of 8-10 amino acids. While the antibodies targeting B6R and M1R shared a similar average CDRH3 length, a higher number of B6R antibodies were found to possess longer CDRH3 compared to M1R antibodies (Fig. 1D). Finally, we extracted plasmid DNA containing the coding sequences of the identified human antibodies and constructed full-length IgG1 constructs expression in Expi239F system.

A Diagram of the EV and MV forms of MPXV. M1R and B6R are located on the surface of the mature and enveloped virions, respectively. B and C Frequency distribution of human IGVH and IGVL in M1R (B) and B6R antibodies (C). D The length of the CDR3 at IGVH and IGVL in M1R and B6R antibodies. (E) Binding activity and neutralization potency of the M1R and B6R antibodies. The binding assay was based on ELISA while the neutralizing activity was measured using replication-deficient MPXV. The number in the box indicates the half-maximal effective concentration (EC₅₀) or half-maximal inhibitory concentration (IC₅₀) values. Blank or excluded values were represent by an X in the corresponding table cells. No detective (N.D.) indicates that no response was observed at the highest concentration, while > 1000 or > 66.7 denotes that the EC50 or IC50 value of the antibody is higher than the maximum detectable concentration. Binding and neutralization curves are shown in Figure S2. Source data are provided as a Source Data file.

We then assessed the binding and neutralizing activity of the antibodies using enzyme-linked immunosorbent assay (ELISA) and plaque reduction neutralization test (PRNT), respectively (Figs. 1E and S2). Overall, M1R antibodies showed slightly weaker binding affinities compared to B6R antibodies. Only a few M1R antibodies demonstrated remarkable binding affinity, with half-maximal effective concentration (EC₅₀) values below 1 ng/mL, including A046, A084 and A094, while the remaining antibodies displayed much lower binding activities with EC₅₀ values above 20 ng/mL. By contrast, B6R antibodies showed robust binding activity, with the majority of clones exhibiting EC₅₀ values below 10 ng/mL. Notably, B026, B046, B089, B161 and B177 exhibited particularly tight binding to B6R, with EC₅₀ values below 0.5 ng/mL (Fig. 1E). We further evaluated the in vitro neutralization capacity of M1R and B6R antibodies against replication-deficient MPXV²⁷. Specifically, MV forms of MPXV were utilized to evaluate the neutralization potency of M1R antibodies, while EV forms were employed to assess the neutralizing capacity of B6R antibodies. The potency of neutralization was determined using the half-maximal inhibitory concentration (IC₅₀). The IC₅₀ values of most antibodies were above 1 nM, but three M1R antibodies (A061, A094, and A138) and four B6R (B026, B050, B083, and B161) antibodies demonstrated more potent neutralizing activity, with IC₅₀ values below 0.5 nM. Notably, M1R antibody A138 and B6R antibody B026 stood out with the most potent neutralization activity (Fig. 1E).

To map out the epitopes of these antibodies, we performed a cross-competition assay using Bio-Layer Interferometry (BLI) (Fig. S3). Our findings revealed a spectrum of epitopes utilized by the tested antibodies. Notably, some of these epitopes overlapped with those recognized by previously reported antibodies, such as 7D11²⁶ targeting M1R and 8AH8AL²⁸ targeting B6R. However, we also identified antibodies binding to novel epitopes that have not yet been characterized.

M1R and B6R antibodies exhibited cross-neutralization activity against VACV and MPXV

Next, we evaluated the neutralizing activities of M1R and B6R antibodies against VACV and replication-deficient MPXV. Four antibodies (A138 and A129 against M1R, B026 and B019 against B6R) were chosen for the following cross-neutralization experiments considering their high expression level, differentiated neutralizing activity and varied binding epitopes. Here, we prepared EV and MV forms of VACV and replication-deficient MPXV for the neutralization assays with M1R and B6R antibodies (Fig. 2A, B). Conversely, when assessed the combination of antibodies targeting different antigens, a mixture of EV and MV forms of VACV/MPXV was used (Fig. 2C, D).

A, B Neutralizing activity of individual antibodies (A129, A138, B026, and B019) or antibody cocktail targeting the same antigens against MV and EV forms of replication-deficient MPXV (A) and VACV (B). MVs of MPXV were utilized to evaluate the neutralization potency of M1R antibodies, while EVs were employed to assess the neutralizing capacity of B6R antibodies. C, D Neutralization activity of individual antibodies (A129, A138, B026, and B019) or antibody cocktail targeting the two different antigens against a mixture of MV and EV forms of replication-deficient MPXV (C) or VACV (D). E Schematic of bispecific antibodies (bsAbs). The parental mAbs contributing to the bsAbs are color-coded: red A138 and green B026. F Binding affinity of bsAbs for M1R and B6R. M1R and B6R were immobilized on the surface of biosensors and individual antibodies were tested at various concentrations. The association and dissociation of antibodies is indicated by the dashed line, which represents the fitted curve. The apparent dissociation constants (K_D, app) are shown above each plot. G Cross-neutralizing activity of bsAbs against VACV and replication-deficient MPXV. Data from a representative neutralization experiment are shown for each antibody. The experiment was replicated twice with similar results. The data represent means ± SEM of triplicates indicated by error bars. The dashed line indicates a 50% reduction of viral infectivity. Source data are provided as a Source Data file.

The results suggested that M1R antibodies could effectively neutralize the MV form of the virus but not the EV form (Fig. S4). Conversely, the B6R antibodies demonstrated neutralization of the EV but not the MV form of MPXV (Fig. S4). The neutralizing activity of A138 or A129 against MPXV and VACV was comparable (Fig. 2A, B), in agreement with previous reports on the 7D11 antibody^29,30. Similarly, the B6R antibodies B026 and B019 also exhibited a comparable neutralizing capacity against MPXV and VACV (Fig. 2A, B). However, compared to B019, B026 exhibited stronger neutralizing activity against both MPXV and VACV (Fig. 2A, B). Furthermore, we explored the combined effect of two antibodies targeting the same antigens against VACV and replication-deficient MPXV. The results suggested that mixing two antibodies (A129 + A138 or B019 + B026) led to no obvious enhancement of the neutralizing effect (Fig. 2A, B).

Additionally, we conducted assessments on the neutralizing activities of M1R and B6R antibody combinations using a mixture of MV and EV forms of MPXV/VACV (Fig. 2C, D). Notably, none of the individual M1R or B6R antibodies could entirely neutralize the combination of MV and EV of MPXV/VACV, even at the maximum concentration of 66.7 nM (10 μg/mL). However, the four combinations of M1R and B6R antibodies exhibited potent neutralizing activity, successfully neutralizing the mixture of MV and EV in both MPXV (Fig. 2C) and VACV (Fig. 2D). Among these combinations, the pairing of A138 and B026 exhibited the most robust neutralizing activity against MPXV and VACV, with IC₅₀ values of 0.163 and 0.112 nM, respectively. Therefore, the neutralization and cross-neutralization of MPXV and VACV can be more effectively achieved by employing a combination of two antibodies targeting different antigens, rather than relying solely on a single antibody with a specific neutralization target.

Design and characterization of bispecific antibodies

To further increase the neutralizing potency and broaden the spectrum against MPXV and other Orthopoxviruses, we designed and synthesized seven bispecific antibodies (bsAbs) in IgG-ScFv format³¹, Bs(ScFv) format, DSD format³² and Cross-Ab format³³ based on A138 and B026 (Figs. 2E and S5). All the bsAbs were successfully obtained as determined by sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE) analysis (Fig. S5). We evaluated the binding affinity of these bsAbs for M1R and B6R using BLI. Differentiated binding patterns were found in various forms of bsAbs (Figs. 2F and S5). Notably, ScFvA138B026 and ScFvB026A138 displayed robust binding to both M1R and B6R, whereas the remaining bsAbs exhibited lower or undetectable affinity (Figs. 2F and S5). The binding affinity of ScFvA138B026 for M1R and B6R were determined to be 0.77 and 1.17 nM, respectively. For ScFvB026A138, the K_D values were 7.7 and 0.74 nM for binding to M1R and B6R, respectively. Remarkably, the binding capacity of ScFvA138B026 and ScFvB026A138 was comparable to that of their parental antibodies (K_D = 2.78 nM for A138 and M1R; K_D = 0.53 nM for B026 and B6R) (Figs. 2F and. S5).

Next, we further assessed the neutralizing efficacy of the these two bsAbs in IgG-ScFv format (ScFvA138B026 and ScFvB026A138) against VACV and replication-deficient MPXV. Interestingly, ScFvA138B026 demonstrated slightly superior neutralization compared to ScFvB026A138, with IC₅₀ concentrations of 0.270 nM and 0.694 nM (IC₅₀) for MPXV and VACV, respectively (Fig. 2G). The enhanced neutralizing activity of ScFvA138B026 might be attributed to its stronger binding affinity for M1R and B6R. Despite the marginally lower neutralizing efficacy of ScFvA138B026 compared to the mixture of parental antibodies, it exhibited significantly better neutralization effects than B026 or A138 against MPXV and VACV (Fig. 2C, D, G).

Structural basis for the recognition of M1R by the antibodies

To gain insights into the molecular mechanism underlying the neutralization activity of M1R antibodies, we attempted a structural study on the M1R-A129/A138 Fab complexes. The crystal structures of the M1R-A129/A138 Fab complexes were determined at resolutions of 2.17 and 3.49 Å, respectively (Fig. 3, Table 1 and Fig. S8). The binding approach of A138 Fab towards M1R was closely resembled that of previously reported mouse antibodies M12B9 (PDB: 4U6H)³⁴ and 7D11 (PDB: 2I9L)²⁶ binding to VACV L1R, whereas A129 represents a novel binding epitope (Fig. 3A). Upon superposition, only a minor clash was found between A138 and A129 Fab, which was consistent with their limited competitive binding characteristics (Fig. S3). Similar to the structure of VACV L1R, M1R within the M1R-Fab complex took the conformation of a bundle of α-helices packed against a pair of two-stranded β-sheets (Figs. 3 and S6). Notably, a slight structural variance between MPXV M1R and VACV L1R was observed in the β1 and β3 strands (Fig. S6). In the M1R structure, these strands are discontinuous and segmented into two shorter β strands, which we named β1, β1’, β3, and β3’, respectively. By contrast, the corresponding region of VACV L1R displayed continuous β1 and β3 strands (Fig. S6). While we observed this structural difference, we could not conclude that the slight difference between MPXV M1R and VACV L1R was a result of their inherent nature, crystal packing effects or antibody interactions.

A Overall structure of antibodies in complex with M1R. MPXV M1R or VACV L1R is colored in cornflower blue; Fab of A138, M12B9, 7D11 and A129 is shown in green, brown, pink and golden, respectively. The heavy chain is shown in a darker color, and all the complex structures are positioned with M1R/L1R in the same orientation. B A129 mainly uses its heavy chain to bind the β-sheet of M1R. The buried surface area (BSA) between the A129 heavy chain and M1R is 431 Ų while that of the light chain is 192 Ų. The β-sheet of M1R is shown in red and the loops connecting β-strands with helices are shown in green. The epitope of A129 on M1R is labeled by a golden line. C A138 mainly uses its heavy chain to bind to the loops connecting β-strands and helices of M1R. The buried surface area (BSA) between A129 heavy chain and M1R is 540 Ų while that of the light chain is 258 Ų. The β-sheet of M1R is shown in red and the loops connecting β-strands with helices are shown in green. The epitope of A129 on M1R is labeled by a green line. D Sequence and secondary structures of M1R. Contacting residues in M1R for A129 and A138 are indicated by golden and green dots, respectively.

A138 Fab utilized its three CDRs of the heavy chain and CDR3 of light chain to bind to the loop region connecting the α helices and β sheets of M1R (540 Ų for the heavy chain and 258 Ų for the light chain), which is located on the opposite side of the amino and carboxyl termini of M1R (Fig. 3C). Within the loop region of M1R (A30-E37), interactions occurred with CDR3 residues (T91-F94, L96) of the A138 light chain, as well as with CDR1 (Y33), CDR2 (S52-S54, T56-T57), and CDR3 (Y105) residues of the heavy chain. Additional scattered interaction residues, including N55, K127, and I128 predominantly engaged with residues from CDR1 (R30, D31) and CDR2 (R53, S54, and T57) from the heavy chain of A138 (Figs. 3D, 4B and Table S3). These interactions included 14 hydrogen bonds and two salt bridges (Fig. 4B and Table S3). Notably, the sequence of the A138 heavy chain originates from human germline IGHV3-11, whereas both 7D11 and M12B9 stem from the mouse germline IGHV1S26, which is the closest match to human germline IGHV1-8*02. Additionally, all these antibodies, including A138, 7D11, and M12B9, displayed robust neutralizing activity against VACV. These findings suggested that although these antibodies are from diverse germlines, they could also bind to a shared epitope, reinforcing the notion that this epitope represents a conserved vulnerable site among Orthopoxviruses.

Detailed interactions of the A129 Fab (A) and A138 Fab (B) with M1R. Hydrogen bonds and salt bridges are indicated by dashed lines. M1R is shown in cornflower blue. The heavy and light chains of A129 are shown in dark goldenrod and yellow, respectively. The heavy and light chains of A138 are shown in spring green and green yellow, respectively.

The analysis of the complex structure of M1R-A129 Fab revealed that A129 mainly binds to the β sheets of M1R, specifically the parallel β1-β1’ and β3 strands, as well as the connecting loop between α4 and β3 (Figs. 3B and 3D). The heavy chain of A129 Fab comprises the majority of the recognition sites for M1R (Fig. 4A and Table S3). The interface between the heavy chain and M1R buries a total surface area of 431 Ų, whereas the area between the light chain and M1R spans only 192 Ų (Fig. 3B). The β1 and β1’ strands of M1R (residues N40-N46) interact with CDR2 (W52, N57, T58, and D59), CDR3 (N104-F108) of the A129 heavy chain, as well as CDR3 (A91 and N92) of the A129 light chain. In addition, several sporadic residues on or near the β3 strand also participate in interactions with the heavy chain CDR2 (E62 and D59) and light chain CDR3 (N92, S93, and F94) (Figs. 3D and 4A). These interactions include a total of 11 hydrogen bonds, with six involving the heavy chain and two involving the light chain (Fig. 4A and Table S3).

Analysis of epitopes recognized by B026 and B019 antibodies targeting B6R

We first attempted to solve crystal structures in order to decipher the binding epitopes of neutralizing antibodies targeting B6R. Unfortunately, no crystals appeared for either the B6R-B019 or B6R-B026 Fab complexes. As an alternative, we employed biochemical studies to determine the potential epitopes of these antibodies on B6R. Employing AlphaFold2, we predicted the three-dimensional structure of B6R, revealing that its extracellular domain to consists of a flexible N-terminal region (residue 1-19), followed by four sushi domains (residues 20 to 239), followed by a stalk region (residues 240-279). The latter is composed of a connecting loop and a partial alpha helix juxtaposed to transmembrane domain (Fig. 5A). Based on the predicted domain, we designed a series of constructs with a 6×His-tag at the C-terminus (Fig. 5B). Utilizing the BLI assay, we assessed the binding affinity of these truncated B6R proteins for B026 and B019. The results indicated that B026 bound to the entire extracellular domain of B6R (residues 19-279; clone 1). By contrast, B026 failed to bind to truncated B6R variants containing only domains 1-4 (clone 2), indicating that the stalk region of B6R (residues 240-279) is indispensable for B026 binding. To validate this specific interaction, we subsequently generated constructs with or without the stalk region. B026 exhibited robust binding activity with the constructs containing the stalk region (clones 7, 9, 11 and 12), while showing no binding with fragments lacking this region (clones 3, 4, 5, 6, 8 and 10). The results further confirmed the critical role of the stalk region in mediating B026-B6R interactions (Fig. 5C). In addition, strong binding reactivity was observed between B019 and all truncated B6R proteins containing domain 4 (clones 1, 2, 6, 7, 8, 9, 10 and 11), whereas minimal or no binding occurred with constructs lacking this domain (clones 3, 4, 5 and 12), which suggests the specific recognition of domain 4 by B019 (Fig. 5D).

A Modeled structure of the B6R protein predicted using AlphaFold2. The four domains of B6R are shown in red (domain 1), blue (domain 2), yellow (domain 3), and green (domain 4). The stalk region is shown in brown. B Schematic diagrams representing the structures of a series of truncated B6R proteins variants. All these truncated B6R constructs contain a 6×His tag at the C-terminus, which was used for protein purification. C, D Binding of B026 (C) and B019 (D) to different fragments of B6R. Different fragments of B6R were loaded onto the surface of biosensors, and individual antibodies were tested at three concentrations (75 nM, 150 nM, 300 nM). The association and dissociation curves are shown. The dashed lines represent the fitted curves based on the experimental data. Source data are provided as a Source Data file.

Prophylactic and therapeutic efficacy of B026 and A138 in mouse models of VACV and MPXV infection

Given the remarkable neutralizing capabilities of B026 and A138, we first evaluated their prophylactic and therapeutic efficacy of VACV infection in a mouse model. Intranasal administration of VACV resulted in significant weight loss and mortality in C57BL/6 mice. Subsequently, mice were administered either B026, A138, or a combination of both either one day before (prophylactic group) or one day after (therapeutic group) VACV challenge (Figs. 6A and S7A). Over the course of 15 days, the animals were closely monitored for changes of body weight and survival.

A Schematic diagram of the prophylactic and therapeutic models. C57BL/6 mice were randomly divided into 4 groups (n = 6 per group). For the prophylactic model, mice were injected intraperitoneally (i.p.) with PBS, individual antibodies (A138 or B026) or antibody cocktail (A138 and B026) at 24 hrs before challenged with a high-dose (10⁹ PFU) of VACV. For the therapeutic model, mice were injected i.p. 24 hrs after challenge with 10⁹ PFU VACV. Body weight (B, D) and survival curves (C, E) were recorded. B, C show data for prophylactic model while (D) and (E) for the therapeutic model. F, G Viral titers in the lungs and spleen of mice in the prophylactic (F) and therapeutic groups (G). The data represent means ± SEM, indicated by error bars. The significance of differences between groups was assessed using one-way ANOVAs followed by Dunnett’s multiple comparisons test for multiple comparisons. Source data are provided as a Source Data file.

In the prophylactic group, we initially conducted experiments using a dose of 10⁷ plaque-forming units (PFU). The results demonstrated that either a single dose of B026 or the combination of B026 and A138 conferred complete protection (Fig. S7B). To further evaluate the protective efficacy of the antibodies, we increased the challenge dose to 10⁹ PFU. Under this heightened challenge scenario, control mice administrated PBS experienced severe illness, with all animals succumbing by day 10 post-inoculation (p.i.), while a single dose of either B026 or A138 only conferred partial protection (Fig. 6B, C). Specifically, all mice in the A138 and B026 treated groups experienced significant weight loss, with 83% and 66% survival rates, respectively. By contrast, mice pre-treated with the combination of B026 and A138 displayed only slight weight loss during the initial 10 days after viral challenge, followed by subsequent body weight recovery. Meanwhile, those in the control group continued to lose weight without any signs of improvement (Fig. 6B, C). The viral load in the lungs and spleen was assessed via the PRNT assay at four days post-infection (dpi). In the combination treatment group, infectious viral levels were extremely low or undetectable in both the lungs and spleen, even after challenge with 10⁹ PFU (Fig. 6F).

Mice mock-treated with PBS in the treatment group showed similar outcomes to those in the prophylactic group, with all succumbing by day 10 p.i. (Fig. 6D, E). Treatment with either B026 or A138 alone resulted in considerable reductions of body weight, with survival rates of approximately 66% and 50% in the A138 and B026 treated groups, respectively (Fig. 6D, E). The levels of infectious virions in the lungs and spleen of these two groups ranged from 10² to 10³ PFU/g, which was lower than in the control group (Fig. 6G). By contrast, animals treated with the combination of B026 and A138 maintained relatively stable body weights and were fully protected from VACV infection (Fig. 6D, E). No virions were detected in the two organs of mice treated with the combination of B026 and A138, whereas the PBS mock-treated mice exhibited viral levels ranging from 10⁴ to 10⁵ PFU/g (Fig. 6G). Collectively, these results underscored the superior protective efficacy of the combination of B026 and A138 against authentic VACV infection in vivo, using either prophylactic or therapeutic interventions.

To further investigate the protective and therapeutic efficacy of the identified antibodies in a mouse model of MPXV infection, we conducted animal studies using live clade IIb MPXV (MPXV-B.1-China-C-Tan-CQ01)³⁵ in BALB/c mice³⁶. Since the MPXV strain we used does not cause lethality in BALB/c mice, here we measured viral loads in various organs to assess antibody efficacy through qPCR analysis. The results revealed that viral loads were highest in the lungs (3.17 × 10⁵ copies/mL, PBS group), while being lower in the liver and spleen (Fig. 7A, B). In both prophylactic and therapeutic groups, antibody-treated mice exhibited reduced viral loads in the lungs, liver, and spleen compared to the control group (PBS group) (Fig. 7A, B). This demonstrating that the antibodies used in our study effectively provided protection against MPXV infection. Notably, co-administration of the A138 and B026 antibodies resulted in significantly lower viral titers in all organs, suggesting that the combination of two antibodies targeting different antigens enhanced protective efficacy. However, the bispecific antibody ScFvA138B026 containing both A138 and B026 binding regions exhibited a similar, or slightly reduced, protective effect compared to A138 or B026 alone.

A Viral titers in the lungs, liver and spleen of mice in the prophylactic model. B Viral titers in the lungs, liver and spleen of mice in the therapeutic model. BALB/C mice were randomly divided into 6 groups (with n = 4). For the prophylactic model, mice were injected intraperitoneally (i.p.) with PBS, individual antibodies (A138, B026, ScFvA138B026, or 7D11) or antibody cocktail (A138 and B026) at 24 hrs before challenged with 10⁶ PFU of MPXV. For the therapeutic model, mice were injected i.p. 24 hrs after challenged with 10⁶ PFU of MPXV. The data represent means ± SEM, indicated by error bars. The significance of differences between groups was assessed using one-way ANOVAs followed by Dunnett’s multiple comparisons test for multiple comparisons. Source data are provided as a Source Data file.