A highly potent nanobody-based bispecific therapeutic provides broad-spectrum protection against ebolavirus

Abstract

The highly lethal Ebola virus species—Zaire (EBOV), Sudan (SUDV), and Bundibugyo (BDBV)—pose persistent threats to global health. Current antibody therapies target EBOV but lack broad neutralization across ebolaviruses. Recent pan-ebolavirus strategies rely on antibody cocktails. Here, we identified two camelid-derived nanobodies (1A10 and BA2) that neutralize EBOV, SUDV, and BDBV in vitro and protect female rodents against these pathogens. High-resolution cryo-EM structures of their GP complexes showed that 1A10 and BA2 bind conserved but non-overlapping epitopes near the GP1 base and GP2’s internal fusion loop (IFL), and biochemical analyses revealed their distinct neutralization mechanisms. To further improve efficacy, we engineered a bispecific antibody (BA2-1A10) via GS linker-mediated IgG-Fc fusion, which provided highly potent protection against all three viruses in female rodents model and positions it as a strong broad-spectrum anti-ebolavirus candidate. Our work demonstrates a structure-guided bispecific nanobody strategy for pan-ebolavirus therapy and highlights compact antibodies for next-generation antivirals.

Current antiviral treatment approaches target specific Ebola virus species. Here, the authors develop bispecific antibodies with pan-ebolavirus neutralization capacity, structurally resolve the epitope and neutralization mode and show potent protection in a rodent model.

Introduction

Ebolavirus, a member of the family Filoviridae, is a highly virulent pathogen that causes Ebola virus disease (EVD), a severe and often fatal haemorrhagic fever in humans and non-human primates. Of the six known species, the Zaire ebolavirus (EBOV), Sudan ebolavirus (SUDV) and Bundibugyo ebolavirus (BDBV) are responsible for the majority of human infections and mortality. Among them, EBOV is associated with the highest case fatality rates, approaching 90% in the absence of supportive care¹. Since its discovery in 1976, EBOV has caused numerous large-scale outbreaks, posing a persistent threat to global public health. In 2022, the Democratic Republic of the Congo experienced its 14th outbreak of Ebola virus disease (EVD) caused by EBOV, while Uganda reported an outbreak driven by SUDV, resulting in 164 cases (142 confirmed and 22 probable) and 55 deaths across nine districts, highlighting the ongoing threat posed by multiple Ebolavirus species and underscoring the urgent need for effective therapeutics and vaccines². The Ebola virus glycoprotein (GP), the sole viral protein exposed on the virion surface, is a key determinant of pathogenicity and the principal mediator of viral entry. As a class I fusion protein, GP orchestrates host cell attachment via receptor engagement, facilitates membrane fusion within endosomes, and contributes to virion assembly and egress. Beyond these structural functions, GP actively subverts host immune responses by modulating cellular gene expression and inducing cytotoxic effects, thereby reinforcing its status as a critical target for therapeutic intervention and vaccine design^3–5. GP biosynthesis begins with translation of a ~ 676 aa precursor that undergoes furin-mediated cleavage into two functionally distinct subunits: the receptor-binding GP1 (aa 1-312) and the fusion-mediating GP2 (aa 313-676). These remain tethered through a conserved Cys53-Cys609 disulfide bond, ultimately assembling into trimeric spike complexes that adopt a metastable, chalice-shaped conformation on virions - a structural prerequisite for membrane fusion activity^6–8. The GP1 subunit contains three modular domains that collectively facilitate immune evasion and host cell attachment: the heavily glycosylated glycan cap and mucin-like domain (MLD) form a protective shield around the underlying receptor-binding domain (RBD)^9,10. The RBD interacts with the endosomal receptor Niemann–Pick C1 (NPC1), a critical determinant of viral tropism and entry^8,11–13. The GP2 subunit harbours four functional elements essential for fusion: an internal fusion loop (IFL) that inserts into host membranes; heptad repeat regions HR1 and HR2, which refold to form a six-helix bundle driving fusion pore formation; a membrane-proximal external region (MPER) that modulates fusion kinetics; and a transmembrane domain (TMD) that anchors the spike complex and contributes to its structural integrity^14–17.

Targeting the GP with antibody-based therapeutics has emerged as a clinically validated strategy for the treatment of EVD. Following the 2013–2016 West African Ebola epidemic¹⁸ that garnered global concern, several monoclonal antibody therapies have advanced through clinical development. Among these, ZMapp—a cocktail of three chimeric mAbs (c2G4, c4G7, and 13C6) directed against non-overlapping GP epitopes—was the first to demonstrate therapeutic efficacy in nonhuman primates and compassionate use settings¹⁹. Subsequently, two antibody-based therapeutics have received regulatory approval from the U.S. Food and Drug Administration: REGN-EB3 and mAb114. REGN-EB3 consists of three fully human antibodies-REGN3470, REGN3471, and REGN3479-targeting the receptor-binding site (RBS), glycan cap periphery, and internal fusion loop (IFL), respectively^20,21. mAb114, in contrast, recognizes a quaternary epitope spanning the glycan cap and inner chalice of the GP trimer, offering potent neutralization when administered as monotherapy^22,23. Despite their success, these mAbs confer protection exclusively against EBOV, and show limited or no efficacy against other pathogenic Ebolavirus species, such as SUDV and BDBV, both of which retain epidemic potential^24,25. Moreover, conventional full-length IgG antibodies (~150 kDa) are constrained by their pharmacokinetic and tissue-penetration profiles. In particular, their poor permeability across the blood–brain barrier (BBB) hampers efficacy during late-stage infection, where viral dissemination to immune-privileged sites is common. Consequently, even with antibody therapy, case fatality rates can remain as high as 30%^26,27. To overcome these limitations, single-domain antibodies, or nanobodies (~15 kDa), derived from the variable domains of camelid or shark heavy-chain-only antibodies, have gained attention as next-generation biologics^28,29. Nanobodies combine high target specificity with exceptional biophysical properties, including robust thermal and chemical stability, low manufacturing costs, and deep tissue penetration. Their high sequence identity with human germline V gene segments facilitates humanization and reduces immunogenicity, further supporting their clinical applicability^30,31. Collectively, these features position nanobodies as versatile and scalable therapeutic candidates for broad-spectrum protection against EVD and other filoviral infections.

Despite growing efforts to develop pan-ebolavirus antibody therapies, no broadly protective therapeutic has yet been approved for clinical use. Given the unpredictable emergence and genetic divergence of ebolavirus species, cross-reactive therapeutics remain a critical unmet need. Here, we identified two naturally occurring nanobodies—1A10 and BA2—that potently neutralize EBOV, SUDV, and BDBV. In murine challenge models, 1A10 and BA2 individually conferred 100% protection against EBOV. Furthermore, 1A10 alone also achieved 100% protection against SUDV. To further enhance the therapeutic breadth, we engineered a bispecific antibody, BA2-1A10, by connecting BA2 and 1A10 with a GS linker followed by an IgG-Fc. The engineered antibody conferred highly potent protection against EBOV, SUDV, and VSV-BDBV in rodent challenge models. We also elucidated the broad-spectrum neutralization mechanisms of 1A10 and BA2 through structural biology approaches and in vitro biochemical characterization.

Results

Isolation and identification of broad-spectrum nanobodies from immunized camels

To generate natural broad-spectrum nanobodies, we expressed and purified the extracellular domains of glycoproteins (GP) from three ebolavirus species- SUDV, BDBV, and EBOV- each with a truncated mucin-like domain (dMLD-GP). Based on our analysis of the GP sequences from these three ebolavirus species, which revealed higher similarity between EBOV and BDBV and a more distant relationship with SUDV, we adopted a strategy of immunizing one camel with a mixture of SUDV and BDBV dMLD-GP, and the other with EBOV dMLD-GP alone. Following the initial immunization, all animals received two booster shots at three-week intervals. (Fig. S1). Robust immunogenicity was confirmed by enzyme-linked immunosorbent assay (ELISA), with post-immune plasma exhibiting endpoint titers exceeding 10⁴ against all three GPs (Fig. 1a, b). Furthermore, pseudovirus neutralization assays revealed broad-spectrum neutralizing activity in the sera from both animals (Fig. 1c, d). Peripheral blood mononuclear cells (PBMCs) were harvested for variable heavy chain-only heavy chain (VHH) gene amplification, and two independent phage display libraries (each with a complexity of ~10⁸ CFU) were constructed (Fig. S1). An iterative biopanning approach was employed, alternating antigen exposure among the three GPs over three rounds to enrich cross-reactive clones. Candidate binders were identified by phage ELISA (Fig. 1e, f), and sequencing analysis yielded 102 unique nanobody clones with confirmed GP reactivity. To enable functional screening, each nanobody was fused to a human IgG1 Fc domain (Nb-Fc), expressed in HEK293F cells, and purified by Protein A affinity chromatography. Initial characterization using pseudotyped recombinant vesicular stomatitis viruses expressing ebolavirus glycoproteins and GFP reporters (rVSV-EBOV/SUDV/BDBV-GP-eGFP) revealed five cross-reactive nanobodies (1A10, BA2, BC2, ZC2, ZH7) displaying distinct neutralization profiles. All antibodies effectively neutralized VSV-EBOV (IC₅₀ range: 0.6–10 nM), with ZC2-Fc being the most potent (IC₅₀ = 0.6 nM). However, ZC2-Fc showed no neutralization against VSV-BDBV and exhibited an approximately 120-fold reduction in potency against VSV-SUDV (IC₅₀ = 76.1 nM) compared to VSV-EBOV. In contrast, BC2-Fc demonstrated the strongest neutralization against VSV-BDBV (IC₅₀ = 1.3 nM) but only moderate activity against VSV-SUDV (IC₅₀ = 509.0 nM). Neutralization of VSV-SUDV and VSV-BDBV by ZH7-Fc was reduced by 20- to 50-fold relative to its activity against VSV-EBOV (IC₅₀ values: 78.0 nM and 37.8 nM, respectively). Notably, 1A10-Fc and BA2-Fc exhibited the broadest neutralization profiles, with activity against all three pseudotyped viruses (IC₅₀ = 1.5–30 nM). To further evaluate the breadth of antibody neutralization, we also tested their activity aginst Reston ebolavirus (RESTV) using the recombinant human immunodeficiency virus expressing RESTV glycoproteins (HIV-RESTV). Among all antibodies, only BA2-Fc exhibited potent neutralization against HIV-RESTV, with an IC₅₀ of 2.3 nM. None of the other antibodies showed detectable neutralizing activity against this pseudovirus. (Fig. 2a-e,).

Fig. 1. Immunization and Screening for Broad-Spectrum Nanobodies.

a, b ELISA detection of antibody titers in camel plasma against EBOV/BDBV/SUDV-dMLD after three immunizations. PBS was used as a negative control. Binding activity was measured at different plasma dilution gradients, with each concentration tested in triplicate (data points represent mean values; error bars indicate standard deviation). The x-axis shows the logarithm (base 10) of plasma dilution factors, while the y-axis represents absorbance at 450 nm. EC₅₀ values reflect effective antibody titers. c, d Neutralization assessment of camel plasma. A pseudovirus system (VSV-EBOV/BDBV/SUDV-GP) was employed to evaluate neutralizing activity. Plasma was serially diluted (2-fold gradients), and inhibition of VSV-EBOV/BDBV/SUDV-GP recombinant viruses was measured. Data points and error bars denote the mean and standard deviation of neutralization from triplicate wells per concentration. The x-axis shows the logarithm (base 2) of plasma dilution factors; the y-axis indicates inhibition rate (%). e Scatter plot of positive clone identification. Each point represents a monoclonal ELISA result, with SUDV-, BDBV-, and EBOV-binding clones colored blue, red, and green, respectively. The dashed line marks the positivity threshold (OD450 > 0.5). f Pie chart of positive clone distribution. Clones with OD₄₅₀ > 0.5 were considered positive, while those >1.0 exhibited high binding affinity. g Summary of screening results. A total of 102 high-potential candidate clones were obtained.

Fig. 2. Identification of broad-spectrum neutralizing antibodies.

a–d Neutralization assays using recombinant VSV-EBOV/SUDV/BDBV-GP and HIV-RESTV-GP pseudoviruses. Data points and error bars denote the mean and standard deviation (SD) of neutralization from triplicate wells per concentration. e Summary of neutralizing antibody IC₅₀ values against pseudoviruses. Values are shown as IC₅₀ (95% CI). The“ND” indicates neutralization was not detected at the highest antibody concentration tested (1000 nM). f–h Plaque reduction neutralization tests (PRNT) with live EBOV, SUDV, and BDBV. Viral inhibition was measured at various antibody concentrations, with each concentration tested in triplicate. Data points and error bars represent the mean and SD of viral neutralization from triplicate wells at each antibody concentration. i Summary of neutralizing antibody IC₅₀ values against live viruses. Values are shown as IC₅₀ (95% CI).The“ND” indicates neutralization was not detected at the highest antibody concentration tested (4000 nM, with the following exceptions: ZC2-Fc was tested up to 500 nM against BDBV). “>” indicates that the IC₅₀ exceeds the maximum test concentration (4000 nM).

To validate these findings using authentic viruses, we conducted plaque reduction neutralization tests (PRNT) in biosafety level 4 (BSL-4) facilities against live EBOV, SUDV, and BDBV. All five nanobodies demonstrated EBOV neutralization with comparable IC₅₀ values (1.9-7 nM), with ZC2-Fc retaining its superior potency. However, in the authentic virus neutralization assay, ZC2-Fc lost its activity against both SUDV and BDBV, contrasting with its neutralization of VSV-SUDV in the pseudotyped system. Likewise, BC2-Fc showed no detectable neutralization of authentic SUDV, and ZH7-Fc displayed only minimal activity against it (IC₅₀ > 40 μM). In contrast, ZH7-Fc neutralized authentic BDBV as effectively as it did EBOV (IC₅₀= 1.9 nM). These observations highlight system-dependent differences, likely attributable to variations in viral entry mechanisms and epitope presentation (Fig. 2f–i). Importantly, both assay platforms consistently identified 1A10 and BA2 as the most broadly neutralizing nanobodies, supporting their selection for downstream structural and mechanistic analyses.

Mechanisms of broad-spectrum nanobodies 1A10 and BA2 in inhibiting ebolavirus entry

Ebolavirus entry into host cells is a multistep process involving initial attachment, endocytosis, and subsequent trafficking to the late endosome^32,33, where GP undergoes proteolytic priming. Within the endosomal compartment, host cathepsins B and L cleave GP1, removing the glycan cap and MLD to unmask the RBS^11,34. The exposed RBS engages the endosomal receptor NPC1^35,36, triggering a conformational rearrangement of the GP trimer into a thermostable post-fusion six-helix bundle^37,38. This transition converts the trimer into a low-energy, thermostable post-fusion six-helix bundle structure³⁹. ultimately driving GP2-mediated fusion of the viral and endosomal membranes.

To elucidate the neutralization mechanisms of 1A10 and BA2 during ebolavirus entry, we first assessed their binding affinities to various GP forms by ELISA and SPR under neutral (pH 7.4) and acidic (pH 5.5) conditions, mimicking the extracellular and late endosomal environments, respectively. ELISA results demonstrated that both antibodies exhibited high-affinity binding to GP variants lacking the mucin-like domain (GP-dMLD) from EBOV, SUDV, and BDBV, with half-maximal effective concentration (EC₅₀) values ranging from 0.15 nM to 2.79 nM. Notably, 1A10-Fc showed approximately 10-fold stronger binding to cleaved EBOV GP (GP-CL) compared to uncleaved EBOV GP (GP-dTM) under both neutral and acidic conditions. In contrast, BA2 maintained stable nanomolar binding to both GP-dTM and GP-CL, with no significant difference in affinity between the two forms (Fig. 3a). To further characterize antibodies binding to pre- and post-cleavage GP with higher precision, we performed SPR under acidic conditions (pH 5.5), approximating the environment of late endosomes, to assess binding kinetics toward GP-dTM, GP-dMLD, and GP-CL. Consistent with ELISA findings, 1A10-Fc bound to uncleaved GP-dTM with nanomolar affinity (K_D = 1.62 nM) but exhibited picomolar affinity for cleaved GP-CL (K_D = 7.51 pM), reflecting an approximately 200-fold enhancement. This suggests that the 1A10 epitope is partially obscured on uncleaved GP and becomes more accessible after proteolytic cleavage. In contrast, BA2-Fc maintained consistent nanomolar affinities for GP-dTM, GP-dMLD, and GP-CL (K_D = 2.80 nM, 2.00 nM, and 1.48 nM, respectively), indicating its ability to recognize both pre- and post-cleavage states of GP, albeit with weaker binding to GP-CL compared to 1A10-Fc (Fig. 3b).

Fig. 3. Characterization of broad-spectrum antibodies 1A10 and BA2: in vitro binding properties and neutralization mechanisms.

a ELISA analysis of 1A10-Fc and BA2-Fc binding to various GP protein forms (including SUDV/BDBV/EBOV-dMLD, EBOV-dTM, EBOV-CL) under different pH conditions. Data points represent mean values from triplicate measurements with error bars indicating standard deviation. The EC₅₀ values are indicated in parentheses within the graphs. b The the binding kinetics of 1A10-Fc and BA2-Fc to EBOV-GP-dTM/dMLD/CL under acidic conditions by SPR. The actual responses (colored lines) and the data fitted to a 1:1 binding model (black dotted lines) are shown. The equilibrium dissociation constant (K_D), association rate constant (K_a), and dissociation rate constant (K_d) are indicated. c Competitive ELISA between 1A10 and GP-CL receptor binding. Two competition formats were tested: (1) varying concentrations of 1A10 competing against fixed NPC1-C-Fc (red sigmoidal curve), and (2) varying NPC1-C concentrations competing against fixed 1A10-Fc (blue sigmoidal curve). Both assays demonstrated moderate competition between 1A10 and receptor binding. Data points represent mean values from triplicate measurements with error bars indicating standard deviation. Glycan cap cleavage inhibition assay evaluating 1A10’s effect on proteolytic processing. Cleavage products were analyzed by: d non-reducing SDS-PAGE with Coomassie blue staining, and e Western blot detecting the His-tag on GP-dMLD. Intact GP-dMLD (~70 kDa, red boxes) showed delayed cleavage in the presence of 1A10. All experiments were performed in triplicate with consistent results. f Proposed neutralization mechanisms. Schematic illustrates distinct inhibition profiles: 1A10 potentially blocks cell attachment, GP cleavage, NPC1 binding, and fusion triggering, while BA2 primarily affects cell attachment and fusion triggering.

The initial step of EBOV entry into host cells involves the virus binding to non-specific receptors on the cell surface via surface glycoproteins (GPs), followed by cellular uptake through macropinocytosis. Therefore, we next investigated the neutralizing effects of antibodies 1A10-Fc and BA2-Fc on the pseudovirus VSV-EBOV-GP at different stages: before and after viral attachment, as well as 4 h post-infection. The results showed that both antibodies exhibited nanomolar level neutralization both before and after attachment, with slightly stronger neutralization prior to attachment (2.5–3.5-fold enhancement), indicating that the antibodies are effective at both stages. Surprisingly, even when the antibodies were added after the virus had entered the cells, they still demonstrated substantial neutralization. Specifically, 1A10-Fc showed neutralization potency comparable to pre-attacnment stage, while BA2-Fc exhibited even stronger neutralization—3-fold higher than pre-attacnment and 10-fold higher than post-attachment (Fig. S2c, d). These results suggest that BA2-Fc effectively restricts viral spread to neighboring cells, consistent with its ability to recognize both pre- and post-cleavage forms of GP. Given that GP-CL engagement with NPC1 is essential for viral entry^35,36, To further investigate whether 1A10 and BA2 could interfere with the receptor-GP-CL binding process, we expressed the GP protein-binding region of the classical receptor NPC1 and performed competitive ELISA assays to validate the competitive binding between antibodies and the receptor. we found that 1A10, but not BA2, partially competed with NPC1 for GP-CL binding (Figs. 3c, S2a), indicating that 1A10 may inhibit viral entry by blocking receptor engagement.

We further examined whether these nanobodies modulate proteolytic sensitivity of GP. Thermolysin, a bacterial protease mimicking endosomal cathepsin activity^34,40 was used to cleave GP-dMLD in the presence or absence of nanobodies. Non-reducing SDS-PAGE and immunoblot analyses showed that 1A10 delayed the disappearance of the GP-dMLD band in a time-dependent manner, indicating reduced cleavage efficiency (Fig. 3d, e). In contrast, BA2 had negligible impact on proteolysis, consistent with its recognition of the cleaved GP form (Fig. S2b). Our finding demonstrate that the 1A10 exert partial inhibition on the proteolytic processing of the glycan cap, whereas this effect is less pronounced with BA2.

Collectively, these findings indicate that nanobody 1A10 acts through a multifaceted mechanism during ebolavirus entry, interfering with cell attachment, GP proteolytic processing, blocking NPC1 binding, and potentially disrupting subsequent membrane fusion. In contrast, BA2 exhibits a more focused mechanism, primarily neutralizing through the inhibition of cellular adhesion and membrane fusion. (Fig. 3f).

Structural basis for broad ebolavirus neutralization of 1A10 and BA2

To elucidate the structural mechanisms underlying the broad neutralization capacity of nanobodies 1A10 and BA2, we determined the cryo-EM structures of EBOV GP-dMLD in complex with 1A10 and BA2 at 2.56 Å and 2.96 Å resolution, respectively (Figs. S3, S4, Table S1). In the GP-dMLD/1A10 complex, three 1A10 copies bound simultaneously to the GP trimer, with their binding sites centered at the IFL base and spanning both GP1 and GP2. 1A10 buried ~456 Ų of surface area on GP1 and ~396 Ų on GP2 (Fig. 4a–c). The contacts are mediated by the three complementarity determining regions (CDRs) and the loop between CDR2 and CDR3. Specifically, the long CDR3 (aa 98-105) contacts multiple regions on GP1 (residues 71–75, 77, 105–107, 136), including the 3₁₀ pocket and glycerol pocket, and on GP2 (residues 512–514, 559–560). CDR2 (aa 98-105) also contribute to contacting with GP1 by inserting L54 into the 3₁₀ pocket but mainly contacting with GP2 (aa 510-512). CDR1(aa 27, 29-31) only involved in contacting with GP2 (aa 514, 549), as well as the loop between CDR2-CDR3 (aa 73) contacting with GP2 (aa 551) (Fig. 5a). Although the glycan cap is invisible in the GP-dMLD structure, structural alignment with full-length GP (PDB: 5JQ3) reveals that 1A10 binds just beneath the glycan cap and displaces the β17–β18 loop by occupying the 3₁₀ pocket—an interaction previously observed in antibodies such as rEBOV-515 and rEBOV-520 (Fig. 5b). GP cleavage removes the glycan cap and the β17–β18 loop, thereby exposing the pocket and enhancing 1A10 binding to GP-CL, consistent with our binding data (Fig. 3a). Moreover, superposition with full-length GP shows that the CDR3 of 1A10 (residues 108–111) potentially interacts with the glycan cap (residues 249–255), suggesting a possible role in inhibiting cap displacement during viral entry (Fig. 3d, e, Fig. 5b). High-resolution maps allowed accurate modeling of side chains at the binding interface, revealing a dense interaction network comprising sixteen hydrogen bonds and salt bridges—eleven with GP1 and five with GP2. Additionally, strong cation-π stacking interactions were observed between Y101 (CDR3) and R136 (GP1), as well as Y27 (CDR1) and H549 (GP2) (Fig. 5d, e, Fig. S5a, b, Table S2). Notably, these key interacting residues are highly conserved across the genus of ebolavirus with the exception of the N514D and H549E/Q variants found in some, which do not affect the interactions involving residues 514 and 549, (Fig. 5g, Fig. S5d), providing a structural explanation for the broad neutralization potency of 1A10. Despite being a heavy-chain-only nanobody, its binding interface forms over 10 major interactions (hydrogen bonds, salt bridges, and π-stacking), comparable to full-length antibodies (which possess both heavy and light chains). This extensive interaction network ensures high binding affinity, which underlies its potent neutralization efficiency.

Fig. 4. Cryogenic electron microscopy (cryo-EM) structures of EBOV-GP-dMLD complex with 1A10 and BA2.

Color scheme: Nanobody 1A10 is shown in lime, nanobody BA2 in slate, GP1 in sky blue, and GP2 in gold. a Top view of the EBOV-GP-dMLD/1A10 complex with a transparent cryo-EM map overlay. Three 1A10 nanobodies (lime) bind to the base of the IFL in the glycoprotein (GP) trimer. b Side view of the EBOV-GP-dMLD/1A10 complex. GP-dMLD is displayed as a molecular surface, while 1A10 is shown as a cartoon model. c Interface between 1A10 and GP-dMLD in one protomer. The interactions are primarily mediated by 1A10 CDRs (solid representation) and span both GP1 and GP2 (orange-red, solid representation). d Top view of the EBOV-GP-dMLD/BA2 complex with a transparent cryo-EM map overlay. Three BA2 nanobodies (slate) bind to the fusion peptide of the IFL, spanning two adjacent GP monomers in the trimer. e Side view of the EBOV-GP-dMLD/BA2 complex. GP-dMLD is displayed as a molecular surface, while BA2 is shown as a cartoon model. f Interface between BA2 and GP-dMLD. The CDRs (solid representation) of BA2 mediate binding to GP2 from one monomer and GP1 (orange-red, solid) from an adjacent monomer. g Top view: Superposition of 1A10 and BA2 bound to GP-dMLD. Three copies of 1A10 (lime, cartoon) and three copies of BA2 (slate, cartoon) simultaneously engage the GP trimer (surface) at distinct epitopes. h Side view of 1A10 and BA2 bound to two adjacent GP monomers, highlighting their relative orientations. i Footprint analysis of 1A10 (lime) and BA2 (slate) on GP-dMLD. The combined binding sites nearly cover the entire IFL (GP2) and partially overlap with GP1.

Fig. 5. Interactions of ebolavirus (EBOV) glycoprotein (GP) with 1A10 and BA2.

Color scheme matches Fig. 4. Interactions are denoted as follows: hydrogen bonds/salt bridges (black dashed lines), cation-π interactions (red dashed lines). a Interface between GP-dMLD and 1A10 CDRs. GP-dMLD is shown as a molecular surface; 1A10 CDRs are displayed as cartoon. CDR3 and CDR2 localize at the GP1-GP2 interface near the 3₁₀ pocket and glycerol pocket. Key interactions: Y105 (CDR3) and L54 (CDR2) extend into the hydrophobic 3₁₀ pocket. Y101 (CDR3) inserts into the glycerol pocket. CDR1 and the loop between CDR2-CDR3 primarily engage GP2. b Interface between GP-dTM (PDB: 5JQ3) and 1A10. 1A10 binds beneath the glycan cap (gray). CDR3 and CDR1 occupy the 3₁₀ pocket region, displacing the β17-β18 loop (blue) in unbound GP. The C-terminus of CDR3 (aa108-111, pink) interacts with the glycan cap. c Interface between GP-dMLD and BA2 CDRs. BA2 CDRs bind the fusion peptide of GP2 IFL, spanning two protomers. CDR2 additionally contacts GP1 of the adjacent monomer. d Contact residue details of 1A10-GP1. 1A10 forms 12 hydrogen bonds/salt bridges with GP1 via CDR3 residues: D98, S100, Y101, A107, S108 and GP1 residues: T77, E102, A105, R136 (see Table S2). Y101 (CDR3) and R136 (GP1) exhibit a strong cation-π interaction. e Contact residue details of 1A10-GP2. Six hydrogen bonds/salt bridges and a cation-π interaction (Y27 in CDR1 and H549 in GP2) are present. (f) Contact residue details of BA2-GP. Fourteen hydrogen bonds/salt bridges and one aromatic interaction (W104 in CDR3 and F535 in GP2) are observed (see also Table S3). g Conservation analysis of 1A10/BA2 epitopes across Ebolavirus species (EBOV, SUDV, BDBV, TAFV, BOMV, RESTV). Top: Sequence logo (WebLogo) showing residue conservation. Bottom: Multiple sequence alignment. Green triangles represent 1A10 contact sites; purple triangles represent BA2 contact sites. Almost residues are conservative in Ebolavirus species.

In the GP-dMLD/BA2 structure, three copies of BA2 bound to the GP trimer at the highly conserved hydrophobic fusion peptide of the IFL. Unlike 1A10, the footprint of BA2 spanned two adjacent GP monomers, covering both GP1 and GP2 (Fig. 4d–f). This binding mechanism suggests that BA2 likely neutralizes EBOV by blocking fusion-essential conformational changes. Each BA2 buried ~536 Ų of surface area on one GP protomer and ~164 Ų on another. The majority of the contacts are between the nanobody complementarity determining regions (CDRs) and the hydrophobic fusion peptide of EBOV GP2. Eleven hydrogen bonds and one aromatic interactions (W104 in CDR3 and F535 in GP2) were formed between BA2 and GP2 IFL. Additionally, CDR2 also forms three hydrogen bonds and two salt bridges between GP1 (Fig. 5c, f, Fig. S5c, Table S3). The key interacting residues primarily include S90 and K155 on GP1, along with Q521, A526, G528, Y534, F535, E540, G541, Y543 and E545 on GP2. Notably, all these residues except S90 and Q521 are highly conserved across the genus of ebolavirus (Fig. 5g), rationalizing BA2’s cross-species neutralization activity.

Together, these structures reveal that 1A10 and BA2 recognize distinct yet overlapping conserved epitopes within the IFL (residues 510-551), which constitute a critical functional domain for membrane fusion (Fig. 4g–i). 1A10 shares partial epitope overlap with the broadly neutralizing antibodies rEBOV515 and rEBOV520^41,42, while BA2 binds an epitope similar to that targeted by pan-ebolavirus antibody 1C11⁴³. These findings reinforce the IFL as a conserved and highly effective broad-neutralization hotspot (Fig. S6). Strikingly, the single-domain 1A10 and BA2 engage GP with high affinity and structural breadth comparable to conventional antibodies, highlighting their promise as minimalist, broadly neutralizing therapeutics for ebolavirus infection.

Engineering of 1A10 and BA2 to enhance neutralization efficiency and resistance to viral escape

Previous neutralization assays demonstrated that both 1A10 and BA2 neutralized EBOV and BDBV with comparable potency (IC₅₀: 6–7 nM), while showing reduced efficacy against SUDV (IC₅₀: 31.28 nM for 1A10; 17.15 nM for BA2) (Figs. 2, 6a–c, f). Structural analyses revealed that these antibodies recognize distinct, non-overlapping epitopes on the EBOV glycoprotein, suggesting the potential for functional synergy (Fig. 4g–i). To enhance neutralization potency and broaden activity across ebolavirus species, we engineered a bispecific antibody by genetically linking the variable domains of BA2 and 1A10 in tandem with a flexible glycine-serine (GS) linker. To determine the optimal GS linker length, we performed comparative structural analysis of the 1A10 and BA2 complexes. The results showed that the C-terminal residue (S127) of BA2 and the N-terminal residue (Q1) of 1A10 were separated by a linear distance of approximately 40 Å, with no steric hindrance from other residues between them, making them suitable for connection via a GS linker. Based on this structural analysis, we ultimately connected 1A10 to the C-terminus of BA2 using four repeats of GGGGS (4G₄S), followed by an Fc domain to construct a single-chain bispecific antibody (BA2-1A10-Fc) (Fig. 6e).

Fig. 6. Binding reactivity and neutralization efficiency evaluation of monospecific and bispecific antibody.

a–d Neutralization activity of 1A10-Fc, BA2-Fc, and BA2-1A10-Fc against live EBOV, SUDV, BDBV or TAFV was assessed by plaque reduction assay. Neutralization of EBOV, SUDV, BDBV, and TAFV by serially diluted antibodies. Data points and error bars represent the mean and standard deviation of viral neutralization from triplicate wells at each antibody concentration. The engineered BA2-1A10-Fc antibody exhibited enhanced neutralization potency, with particularly notable improvement against SUDV. e Schematic of bispecific antibody design. BA2 and 1A10 were linked head-to-tail via four repeats of GGGGS, followed by a human IgG1 Fc fragment at the C-terminus of 1A10 to form a single-chain bispecific antibody. f Summary table of Half Maximal Inhibitory Concentration (IC₅₀) values for each antibody, rEBOV515 is the positive control. g Summary table of binding affinities measured by ELISA and surface plasmon resonance (SPR).

Subsequently, the binding activity of 1A10-Fc, BA2-Fc, and BA2-1A10-Fc were assessed by ELISA and SPR. The ELISA results showed that the bispecific antibody BA2-1A10-Fc exhibited a modest (~1- to 2-fold) enhancement in binding activity to GP-dMLD compared to the individual antibodies (BA2-Fc and 1A10-Fc). This observation is unsurprising, given the already high binding affinity of the single antibodies, both of which reached picomolar-level EC₅₀ values—with the exception of BA2-Fc binding to BDBV-dMLD, which showed slightly weaker activity (EC₅₀ = 2.79 nM) (Figs. 6g, S7b). The SPR results were largely consistent with the ELISA data, showing that all antibodies exhibited picomolar-level affinity for GP-dMLD. In particular, 1A10 demonstrated the highest affinity for BDBV-dMLD, with a K_D value of 0.4 pM. It is worth noting that BA2-1A10-Fc showed approximately 20-fold enhanced binding to EBOV-dMLD compared to the individual antibodies, while its affinity for SUDV-dMLD was similar to that of BA2-Fc and 1A10-Fc. In contrast, its affinity for BDBV-dMLD (K_D = 38.4 pM) was about 100-fold lower than that of 1A10-Fc (Figs. 6g, S7a).

To further evaluate the broad-spectrum neutralizing activity of the antibodies, we performed plaque reduction neutralization tests (PRNT) to assess the neutralization efficacy of 1A10-Fc, BA2-Fc, and BA2-1A10-Fc against EBOV, SUDV, BDBV, and another human- and primate-infecting Tai Forest ebolavirus (TAFV). The previously reported broad-spectrum antibody rEBOV-515, which shares an overlapping epitope with 1A10, was used as a positive control. PRNT assays demonstrated that the individual antibodies 1A10-Fc and BA2-Fc neutralized EBOV more effectively than the control antibody rEBOV-515 (~6-fold enhancement), while their neutralization capacities against other viruses were comparable. However, the engineered bispecific antibody achieved IC₅₀ values of 1.9 nM, 2.21 nM, 0.81 nM and 1.68 nM against EBOV, BDBV, SUDV, and TAFV, respectively. Compared to the 1A10-Fc alone, the bispecific antibody showed approximately 3.6-fold, 3-fold, 38-fold and 5.7-fold improvement in neutralization efficacy. When compared to BA2-Fc alone, the improvements were approximately 3.5-fold, 2.7-fold, 21-fold and 3.4-fold, respectively (Fig. 6a–d, f). In addition, we evaluated the broad-spectrum neutralization capacity of Fc-free nanobodies (1A10-VHH, BA2-VHH, and BA2-1A10-VHH). As expected, the nanobodies also demonstrated potent broad-spectrum neutralizing activity, though their efficacy was reduced compared to their Fc-fused counterparts. Specifically, the neutralizing activity of 1A10-VHH was 3- to 4-fold lower than that of 1A10-Fc, while BA2-VHH showed a more pronounced reduction of 14- to 75-fold compared to BA2-Fc. Surprisingly, the bispecific nanobody exhibited significantly enhanced neutralization potency, comparable to that of BA2-1A10-Fc, with IC₅₀ values below 5 nM (Fig. 6f,S7c–f). Our results indicate that although the engineered bispecific antibody does not show a dramatic increase in binding affinity, its neutralization potency is markedly enhanced and exhibits synergistic effects. This improvement is particularly notable against SUDV, suggesting that neutralizing activity is not strictly correlated with binding affinity.

To further investigate whether the bispecific antibody can reduce the risk of viral escape from neutralization. We performed viral escape assays using pseudotyped viruses VSV-EBOV and VSV-SUDV. Antibodies were serially diluted in a five-fold gradient and mixed with the virus, starting at an initial concentration of 50 μg/mL. After ten consecutive passages, in VSV-EBOV infections, cytopathic effect (CPE) was observed at concentrations below 0.4 μg/mL for 1A10-Fc and BA2-Fc, and below 2 μg/mL for the bispecific BA2-1A10-Fc. Sequencing of viral supernatants from wells showing CPE revealed an E304A mutation in the GP protein (located in the glycan cap) across all conditions. In VSV-SUDV infections, CPE emerged at concentrations below 2 μg/mL for 1A10-Fc and the bispecific BA2-1A10-Fc, and below 0.4 μg/mL for BA2-Fc. Sequencing results showed that the monospecific antibodies 1A10-Fc and BA2-Fc each acquired two mutations: E106K/L122F for 1A10-Fc and K155Q/L122F for BA2-Fc, where E106K and K155Q are located within the respective antibody binding sites. In contrast, no mutations were detected in viruses passaged under bispecific antibody pressure (Fig. S7g). These findings suggest that the bispecific antibody may confer an advantage in resisting SUDV immune escape.

Protective efficacy of BA2-Fc, 1A10-FC, bispecific BA2-1A10-Fc against EBOV, SUDV and VSV-BDBV in rodents

To investigate the in vivo protective potential of engineered nanobody-Fc fusion proteins, we assessed BA2-Fc, 1A10-Fc, and the bispecific BA2-1A10-Fc in rodent models of EBOV, SUDV, and BDBV infection.

In a mouse model challenged with a lethal dose of mouse-adapted Ebola virus (MA-EBOV), a single administration of either BA2-Fc or 1A10-Fc (5 mg/kg, intraperitoneally) at 1day post-infection (dpi) conferred 100% protection from death, with no signs of weight loss. In contrast, BA2-1A10-Fc afforded partial protection, with 83% survival and mild weight loss observed in one animal. In contrast, PBS-treated mice exhibited only 17% survival. When treatment was delayed to 2dpi, BA2-Fc and 1A10-Fc still achieved 100% protection, while BA2-1A10-Fc provided 60% survival, whereas PBS-treated animals exhibited 100% mortality (Fig. 7a).

Fig. 7. Protective efficacy of nanobodies with Fc segment against EBOV, SUDV, VSV-BDBV infection in rodents.

a Mice were treated with 1A10-Fc, BA2-Fc or BA2-1A10-Fc at a 5 mg/kg dose after a 24 h or 48 h exposure to 300 TCID₅₀ of mouse-adapted EBOV. A PBS-treated group was used as a control. Animal survival was monitored daily for 14 days post-infection. n = 6 or n = 5 mice were studied per treatment condition. b Syrian hamsters were treated with 1A10-Fc, BA2-Fc or BA2-1A10-Fc at a 5 mg/kg dose after 1 day or 2 days exposure to 1000 TCID₅₀ of guinea pig-adapted Sudan ebolavirus. c IFNAR⁻/⁻ mice were treated with 1A10-Fc, BA2-Fc or BA2-1A10-Fc at a 25 mg/kg dose after 1 day or 2 days exposure to 100 PFU of recombinant vesicular stomatitis virus expressing Bundibugyo ebolavirus glycoprotein (VSV-BDBV). Kaplan–Meier survival curves of 1 day and 2 days, body weights are shown. For analysis of survival, each group was compared to PBS mock control (Mantel-Cox test). p values are indicated.

We next evaluated antibody efficacy in a Syrian hamster model of Sudan ebolavirus (SUDV) infection. Administration of either 1A10-Fc or BA2-1A10-Fc (5 mg/kg at 1dpi) provided 100% protection against mortality and weight loss. Notably, animals treated with the bispecific antibody exhibited continuous weight gain, indicative of rapid recovery and minimal disease burden. BA2-Fc treatment at 1dpi conferred only partial protection, with 50% survival. When treatment was delayed to 2dpi, both 1A10-Fc and BA2-1A10-Fc still conferred substantial protection, with 83% survival, while BA2-Fc-treated animals showed modest protection (67% survival). PBS-treated controls showed high mortality across all groups (Fig. 7b).

Finally, therapeutic efficacy was assessed in IFNAR⁻/⁻ mice challenged with a lethal dose of recombinant VSV expressing BDBV glycoprotein (VSV-BDBV). In the recombinant VSV-BDBV mouse model, a higher antibody dose (25 mg kg⁻¹) was administered compared with that used in the mouse-adapted EBOV and SUDV challenge studies (5 mg kg⁻¹). This adjustment reflects the distinct biological characteristics of the VSV-based surrogate system, in which disease progression is substantially more rapid, with earlier onset of clinical signs and a compressed therapeutic window. To ensure adequate systemic exposure of the nanobodies during the critical early phase of viral dissemination, an increased dose was therefore employed. A single dose of BA2-Fc, 1A10-Fc, or BA2-1A10-Fc (25 mg/kg, intraperitoneally) administered at 1dpi conferred 100% protection from death. PBS-treated animals exhibited 20% survival. Delayed treatment at 2dpi resulted in 100%, 60%, and 80% survival for BA2-Fc, 1A10-Fc, and BA2-1A10-Fc, respectively, whereas all PBS-treated mice succumbed to infection by Day 5 (Fig. 7c).

Collectively, these findings demonstrate that BA2-Fc, 1A10-Fc, and BA2-1A10-Fc possess potent therapeutic activity across multiple ebolavirus challenge models. Notably, the bispecific BA2-1A10-Fc consistently exhibited enhanced and broad-spectrum efficacy against diverse ebolaviruses, including EBOV, SUDV, and BDBV. BA2-1A10-Fc not only provided robust protection when administered early but also maintained substantial efficacy even under delayed treatment conditions, underscoring its translational potential for post-exposure therapy. These results highlight the added value of bispecific antibody engineering in achieving cross-species protection against genetically distinct ebolaviruses.

Discussion

Ebolavirus, members of the family Filoviridae, have caused recurrent outbreaks of viral hemorrhagic fever with high case fatality rates, posing a persistent global health threat. In 2022, the Democratic Republic of Congo (DRC) experienced its 14th Ebola virus disease (EVD) outbreak, underscoring the ongoing public health challenge. Monoclonal antibody therapies represent the most effective countermeasure against EVD. Two antibody therapeutics, Ebanga (mAb114) and Inmazeb (REGN-EB3), have received regulatory approval, yet both are primarily efficacious against EBOV and show limited activity against SUDV and BDBV—species that have also caused human outbreaks. The genetic and antigenic diversity of ebolaviruses, along with their sporadic and geographically unpredictable emergence, highlights the urgent need for broadly neutralizing therapeutics. Several full-length antibodies with cross-reactive potential—such as ADI-15878, CA45, rEBOV-520, rEBOV-515, and 1C11—have demonstrated partial cross-neutralization of EBOV, SUDV, and BDBV when administered in cocktail formats^42–46. However, these antibodies require combination strategies to achieve pan-ebolavirus coverage, limiting their clinical utility and manufacturing feasibility.

Camelid-derived single-domain antibodies (nanobodies) offer an attractive alternative in antiviral development due to their small size (~15 kDa), modularity, high solubility, ease of engineering and cost-effective in production. In this study, we identified two nanobodies, 1A10 and BA2, with potent cross-neutralizing activity following llama immunization and phage display selection. Monotherapy with either nanobody with IgG1 Fc provided 100% protection in a murine EBOV challenge model. Notably, early administration of 1A10, given as soon as one day post-challenge, fully protected mice from lethal SUDV infection, whereas BA2 conferred only partial protection with a survival rate of 50%.

Capitalizing on the inherent modularity of nanobodies, we engineered a biparatopic antibody by fusing 1A10 and BA2. The resulting construct exhibited significantly enhanced broad-spectrum neutralization, with synergistic effects particularly evident against SUDV—achieving a 20-fold increase in neutralization potency. This biparatopic nanobody combines high expression yields (20 mg/L in HEK293F cells), excellent solubility, and robust protective efficacy in both SUDV and BDBV mouse models, solidifying its potential as a leading broad-spectrum therapeutic candidate. Beyond 1A10 and BA2, we identified three additional nanobodies (ZC2, BC2, and ZH7) with comparatively weaker but still detectable cross-neutralizing activity (Fig. 2). Given the versatility of nanobody engineering, further multiepitope constructs incorporating these antibodies with 1A10-BA2 could potentially broaden neutralization coverage and enhance protective efficacy. In addition, concatenating individual nanobodies to achieve multivalent forms, thereby enhancing their neutralization efficacy, represents a promising direction for future research. Cryo-EM structural analysis revealed that 1A10 and BA2 bind adjacent but non-overlapping epitopes near the IFL, enabling their synergistic combination (Fig. 4g–i). Together, their epitopes cover nearly the entire conserved IFL region (explaining the exceptional neutralization and protection efficacy, representing the most potent broad-spectrum anti-ebolavirus antibody reported to date). Biochemical assays showed that 1A10 partially inhibits receptor binding, though structural analysis indicated no direct steric clash with NPC1-C, suggesting allosteric modulation of GP-CL/NPC1-C interaction—a hypothesis requiring further validation. Additionally, 1A10 suppressed GP cleavage, despite lacking observable interactions with the cleavage site (β13-β14 loop) or even visible density for aa190–227⁴⁷. We propose that 1A10 may modulate glycan cap conformation to indirectly impair protease access, given its direct contacts with the glycan cap. Further structural studies of 1A10 bound to GP-dTM are needed to elucidate this mechanism.

Although the epitopes targeted by 1A10 and BA2 reside within the GP base and internal fusion loop (IFL), regions previously identified as sites of vulnerability for broadly neutralizing antibodies, our findings provide new mechanistic and conceptual insights into how this conserved region can be therapeutically exploited. Prior studies have largely emphasized synergistic protection achieved by antibody combinations targeting spatially distinct GP domains, such as the glycan cap and base. In contrast, our data demonstrate that two nanobodies engaging closely related sites within the GP base can also synergize, revealing a previously underappreciated mode of cooperative neutralization within a single functional epitope cluster. This observation expands current models of antibody synergy against ebolavirus GP.

Structural analyses further show that, despite lacking light chains, 1A10 and BA2 establish extensive hydrogen bond and salt bridge networks with GP, achieving an epitope engagement footprint comparable to that of full-length IgG antibodies. The partial epitope overlap of 1A10 with rEBOV-515/520 and of BA2 with ADI-15878 and 1C11 reinforces the internal fusion loop as a highly conserved and druggable site of vulnerability, while independently validating its functional importance through a distinct antibody format.

In this study, Fc fusion was employed to extend serum half-life and enable rigorous in vivo efficacy assessment. Importantly, neutralization assays using nanobody constructs without Fc demonstrated comparable antiviral potency, indicating that the core neutralizing activity is intrinsic to the nanobody architecture rather than dependent on Fc-mediated effects. Beyond pharmacokinetic optimization, nanobodies offer intrinsic advantages—including small molecular size, modularity, and ease of engineering—that may enable improved tissue penetration compared to conventional IgGs. Recent studies have suggested that nanobodies can access immune-privileged sites such as the central nervous system⁴⁸, a key consideration for treating advanced-stage Ebola virus disease. While pharmacokinetics and biodistribution were not directly addressed here, these properties highlight the broader translational potential of nanobody-based therapeutics and warrant further investigation.

Despite a marked enhancement in vitro neutralization potency, the superior efficacy of the bispecific BA2-1A10-Fc was not uniformly translated into improved survival outcomes in vivo. This likely reflects a ceiling effect in stringent rodent challenge models, where potent antibody treatments already achieve near-maximal protection. In such settings, further gains in neutralization potency may not result in measurable differences in survival. In addition, in vivo protection depends on multiple factors beyond neutralization, including pharmacokinetics, tissue accessibility, and viral dynamics. The principal advantage of the bispecific design therefore, lies in its synergistic activity, expanded breadth, and potential to increase the genetic barrier to escape, rather than in further enhancing survival under optimal treatment conditions. To better assess the functional advantage of the bispecific format, we evaluated viral escape using serial passaging of VSV-based pseudoviruses. While escape-associated GP mutations readily emerged under selective pressure from the individual monospecific antibodies, no mutations were detected following passaging in the presence of BA2–1A10-Fc in the VSV-SUDV system. This suggests that simultaneous targeting of two conserved GP epitopes imposes a higher genetic barrier to viral escape.

While non-human primate models are the most predictive system for Ebola virus disease, practical and biosafety constraints precluded their use in the present study. Instead, we employed a tiered preclinical strategy integrating structural characterization, mechanistic neutralization analyses, and therapeutic testing across multiple small-animal models that recapitulate key aspects of ebolavirus pathogenesis. Importantly, the robust protection conferred by 1A10, BA2, and the bispecific BA2–1A10-Fc across EBOV, SUDV, and BDBV challenge systems therefore provides strong justification for subsequent evaluation in non-human primates, which will be essential to fully assess pharmacokinetics, dosing strategies, and therapeutic window in a translational context.

Methods

Ethics statement

All animal experiments were conducted in strict accordance with institutional and national guidelines for the care and use of laboratory animals. Study protocols involving mice were approved by the Laboratory Animal Welfare and Ethics Committee of the University of Science and Technology of China (approval ID: USTCACUC24120124057). Experiments involving live virus challenge in mice and Syrian hamsters were performed at the Wuhan Institute of Virology, Chinese Academy of Sciences, under protocols approved by its Institutional Animal Care and Use Committee (approval No. WIVAF45202302). All experiments involving live EBOV or SUDV were conducted in certified biosafety level 4 (BSL-4) facilities at the Wuhan Institute of Virology, Chinese Academy of Sciences, following institutional biosafety regulations. Experiments using VSV-BDBV were performed under biosafety level 2 (BSL-2) conditions. All personnel involved in animal experiments received appropriate biosafety training and adhered strictly to institutional and national biosafety guidelines.

Female BALB/c mice (6–8 weeks old) were used for mouse-adapted Ebola virus (MA-EBOV) challenge experiments. Female IFNAR⁻/⁻mice (6–8 weeks old) were used for evaluation of antibody efficacy against Bundibugyo ebolavirus (BDBV), including experiments employing recombinant VSV-BDBV challenge models. Female Syrian hamsters (5–6 weeks old) were used for Sudan ebolavirus guinea pig–adapted strain (SUDV-GP) infection experiments. Animals were randomly assigned to experimental groups to minimize experimental bias. All animals were housed in individually ventilated cages under specific pathogen-free conditions with controlled temperature (20–24 °C), humidity (40–60%), and a 12 h light/dark cycle, and were provided with food and water ad libitum. Animals were monitored daily for clinical signs, body weight changes, and overall health status.

Cells and viruses

Expi293F (Thermo Fisher, A14528) cells were cultured in CD 293 01 398 medium (JSBio). HEK293T (ATCC: ACS-4500) cell lines and Vero E6 (catalog no. ATCC® CRL-1586™) were cultivated at 37 °C and 5% CO₂ in Dulbecco’s modified Eagle’s medium (DMEM) (Gibco) supplemented with 10% fetal bovine serum (FBS), 100 μg/mL streptomycin, and 100 U/mL penicillin (Gibco). Recombinant vesicular stomatitis viruses (rVSVs) were generated by replacing the VSV glycoprotein with the enhanced green fluorescent protein (eGFP) and full-length GPs from Ebola virus (Genbank NO. AF086833.2), Sudan virus (Genbank NO. NC_006432.1) or Bundibugyo virus (GenBank: MT680253.1), termed as rVSV-EBOV GP-eGFP, rVSV-SUDV GP-eGFP, and rVSV-BDBV GP-eGFP. The mouse-adapted Zaire ebolavirus (MA-EBOV) and guinea pig-adapted Sudan ebolavirus (SUDV-GP) strains were handled and stored under biosafety level 4 (BSL-4) conditions at the State Key Laboratory of Virology, Wuhan Institute of Virology, Chinese Academy of Sciences, in accordance with approved safety protocols and regulatory guidelines.

Protein expression and purification

The ectodomain constructs of EBOV GP-dTM (residues 1–652; Makona strain; GenBank: AIW47462.1), EBOV GP-dMLD (with deletion of residues 313–468; Makona strain; GenBank: AIW47462.1), BDBV GP-dMLD (with deletion of residues 313–470; 200706291 Uganda strain; GenBank: YP_003815435.1), and SUDV GP-dMLD (with deletion of residues 313–470; Gulu strain; GenBank: ACR33190.1) were transiently expressed in Expi293F cells using the pTT5 plasmid vector, each containing a C-terminal His-tag. Following expression, culture supernatants were collected and subjected to initial purification using 5 mL HisTrap HP columns (GE Healthcare), followed by further purification and buffer exchange into PBS using Superdex200 (GE Healthcare) size-exclusion chromatography. Following size exclusion chromatography, the homogeneous peak fractions were analyzed via room-temperature negative stain electron microscopy to confirm protein conformation. Samples displaying intact trimeric structures were subsequently chosen for immunization.

For nanobody production, the coding sequences were cloned into either pTT5-His vector (containing a C-terminal His-tag for structural determination) or pTT5-Fc vector (containing a C-terminal human IgG1 Fc fragment for in vivo characterization and animal protection studies). The His-tagged constructs were purified using nickel affinity chromatography, while the Fc-fused constructs were purified using Protein A affinity chromatography (Anhui Genebiol). The engineered bispecific antibody BA2-1A10 was established by connecting BA2 and 1A10 with four tandem GGGGS repeats. All antibody preparations were finally buffer-exchanged into PBS through ultrafiltration.

For NPC1 production, the NPC1-C domain (residues 374–620) was cloned into pTT5-Fc vector incorporating a TEV protease cleavage site between the NPC1-C domain and Fc fragment. Following Protein A purification and buffer exchange into PBS, the Fc tag was removed by overnight digestion with His-tagged TEV protease at 4 °C. The cleaved product was subsequently processed through sequential purification steps using Ni-NTA chromatography to remove the His-tagged TEV protease, followed by Protein G (Anhui Genebiol) chromatography to eliminate the cleaved Fc fragments, yielding purified NPC1-C domain.

Phage display library establish and broad-spectrum screening

The phage display library was constructed as previously described. Briefly, camels were immunized three times with 500 μg of GPs-dMLD in PBS, emulsified with an equal volume of Freund’s adjuvant (Sigma-Aldrich). Peripheral blood lymphocytes were isolated via Ficoll (Sigma-Aldrich) density gradient centrifugation after the final boost. Total RNA was extracted using the Total RNA Kit (Omega Bio-Tek) and reverse-transcribed into cDNA following the manufacturer’s protocol. The variable domain of heavy-chain-only antibodies (VHH) and the phagemid pR2 were PCR-amplified, and recombinant VHH-pR2 vectors were assembled using Gibson assembly. These constructs were then transformed into TG1 competent cells to generate the phage display library. Finally, VHH-displaying phage particles were rescued and amplified from the library using KM13 helper phage.

Nanobodies targeting EBOV-GP were isolated from the library through three rounds of ELISA-based biopanning using a sequential screening strategy (Fig. S1c). Briefly, Immuno MaxiSorp plates (Nunc) were coated overnight at 4 °C with 0.1 mL (100 μg/mL) of recombinant EBOV-GP (BDBV-dMLD in the second round and SUDV-GP-dMLD in the third round). Control wells without antigen coating were included in each panning round. After blocking with MPBS (PBS containing 5% skim milk) for 2 h at room temperature (RT), the phage library 1 × 10¹¹ PFU for the first round; 1 × 10⁸ PFU for subsequent rounds) was added for selection. Unbound phages were removed by 20–30 washes with PBST (PBS containing 0.1% Tween-20). Specifically bound phages were eluted via trypsin digestion (1 h, RT) and subsequently used to infect TG1 cells for titer determination and amplification.

Enzyme-linked immunosorbent assay (ELISA)

For Nb-Fc binding assays, Immuno MaxiSorp plates (Nunc) were coated overnight at 4 °C with purified recombinant EBOV, BDBV, or SUDV glycoproteins (GPs; dMLD, dTM, or CL forms) at a final concentration of 2 μg/mL. After washing with PBST (PBS + 0.1% Tween-20), plates were blocked with MPBS (PBS + 5% skim milk) for 2 h at room temperature. Serially diluted Nb-Fc solutions (1:4 dilution series) were added and incubated for 1 h. Following four PBST washes, bound Nb-Fc was detected using an HRP-conjugated anti-human IgG1 Fc monoclonal antibody (Sino Biological, Cat#105327-MM02T-H; 1:10,000 dilution) and TMB substrate (Beyotime). Reactions were terminated with 1 M H₂SO₄, and absorbance at 450 nm was measured using a Synergy H1 plate reader (BioTek). Data were analyzed with GraphPad Prism. For pH 5.5 binding assays, PBS was replaced with MES buffer (pH 5.5, 150 mM NaCl) throughout the procedure.

To assess the receptor-blocking activity of antibodies (1A10/BA2) against NPC1-C, competitive ELISAs were performed. EBOV-CL-coated wells were incubated with either: serially diluted 1A10/BA2-His (0.02–5000 nM, 1:4 dilutions) premixed with 30 nM NPC1-C-Fc, or serially diluted NPC1-C (0.45–7500 nM, 1:4 dilutions) premixed with 5 nM 1A10-Fc. After standard washes, bound 1A10-Fc or NPC1-C-Fc was detected using the anti-IgG1 Fc-HRP antibody (1:10,000) and TMB. TMB substrate (100 μL/well) was incubated for 5 min in the dark before stopping the reaction with 50 μL/well of 1 M H₂SO₄.

Surface plasmon resonance (SPR)

Surface plasmon resonance (SPR) experiments were performed on a Biacore 8 K instrument (Cytiva) at 25 °C with a flow rate of 30 μL/min. To mimic acidic and neutral conditions, buffer A (50 mM MES, pH 5.5, 150 mM NaCl, 0.02% Tween-20) or buffer B (50 mM HEPES, pH 7.4, 150 mM NaCl, 0.02% Tween-20) was used, respectively. Purified antibodies were exchanged into the corresponding buffer (A or B) via ultrafiltration prior to analysis. Various forms of GP were diluted to a concentration of 10 μg/mL and covalently immobilized on a CM5 sensor chip. A concentration series of the antibodies (ranging from 400 nM to 25 nM via two-fold serial dilution) was injected over the chip surface. Kinetic parameters were determined using a single-cycle kinetics (SCK) approach. The sensor chip surface was regenerated with 50 mM NaOH for 60 s after each cycle. Sensograms were globally fitted to a 1:1 binding model using the Biacore Evaluation Software (Cytiva).

Glycan cap cleavage

The glycan cap cleavage assay was performed as follows: Purified EBOV-GP-dMLD trimer (50 μg) was incubated with 50 μg of nanobodies (1A10-His or BA2-His) for 2 h at room temperature. Subsequently, 0.10 μg of thermolysin L (Sigma-Aldrich) was added, and the reaction mixture was incubated at 37 °C for varying durations (5, 15, 30, or 45 min). A control sample containing 50 μg of EBOV-GP-dMLD alone (without nanobodies) was processed in parallel. At each time point, aliquots were immediately quenched by mixing with SDS-PAGE loading buffer (non-reducing conditions) and boiling for 10 min. Samples were then resolved by SDS-PAGE and analyzed via Coomassie blue staining or Western blot. For Western blot analysis, GP-dMLD was detected using an anti-His tag antibody (Proteintech; 1:1000 dilution) followed by an anti-Fc fluorescent secondary antibody (Proteintech; 1:10,000 dilution).

Pseudovirus neutralization assay

The half-maximal inhibitory concentration (IC₅₀) of nanobody candidates were analyzed by an EBOV-pseudoviruses neutralization assay. Briefly, serial dilutions of nanobody preparations were added into 96-well plates. After that, 100 μL pseudoviruses (rVSV-EBOV GP-eGFP, rVSV-SUDV GP-eGFP, or rVSV-BDBV GP-eGFP.) with concentration of 400 TCID50/100 μL were added into the plates, followed by incubation at 37 °C for 1 h. Afterward, the mixture of virus and nanobody was added to Vero E6 monolayer cells to a final volume of 50 μL per well in 96-well plates, in triplicate, and incubated at 37 °C for 1 h. The pseudoviruses without antibodies were used as the control group. After incubation, the mixture was replaced with DMEM containing 2% FBS. After 24 h, the plates were imaged using an MetaXpress high-content imaging system (Molecular Devices), and the number of infected cells per well were calculated using the associated Harmony 3.5 software. The quantification of eGFP expression was used to calculate the inhibition of infection relative to the vehicle control group, and the IC₅₀ values were generated from inhibition curves. For HIV-RESTV  pseudoviruses neutralization assay, HuH7 cells were seeded in 96-well plates and cultured overnight to reach approximately 80% confluence prior to infection. Antibodies were serially diluted three- fold in DMEM starting from 1000 nM, with three replicates per group. The HIV-RESTV viral stock was diluted and mixed with the antibody dilutions, followed by incubation at 37 °C for 1 h. The virus-antibody mixture was then added to HuH7 cells and incubated at 37 °C with 5% CO₂ for 48 h. The cells were washed twice with PBS (180 μL/well), followed by the addition of Luciferase Cell Culture Lysis reagent diluted in ddH2O (25 μL/well). The plates were then shaken at room temperature for 5 min to facilitate lysis. The lysate from each well of the 96-well plate was transferred to a 96-well white plate (20 μL/well), and 100 μL of Luciferase Cell Culture Lysis reagent was added to each well. Relative luciferase activity (RLU) per well was measured using a multi-mode microplate reader. For pre-attachment neutralization, a virus-antibody mixture was pre-incubated for 1 h at 4 °C and then applied to Vero cells. For post-attachment neutralization, prechilled Vero cells were first infected with rVSV-EBOV GP-eGFP at 4 °C for 1 h, followed by the addition of serially diluted antibodies. For post-infection neutralization, the virus was first allowed to infect Vero cells at 37 °C for 4 h. Subsequently, the virus inoculum was removed, and the cells were gently washed three times with PBS (5 min per wash) before the application of diluted antibodies. Following these initial steps, the subsequent procedures for all assays were identical to those described previously.

Plaque reduction neutralization test (PRNT) using live EBOV, BDBV, SUDV

A live ebolavirus neutralization assay was conducted to evaluate the ability of antibodies to block virus attachment using a plaque reduction method in a biosafety level 4 (BSL-4) laboratory. Briefly, Vero E6 cells were seeded at a density of 1.5 × 10⁵ cells per well in 24-well culture plates and incubated overnight. Antibody samples were serially diluted threefold in DMEM supplemented with 2.5% fetal bovine serum (FBS). Equal volumes of the diluted antibody and authentic ebolavirus (600 PFU/mL) were mixed and incubated at 37 °C for 1 h to allow antibody-virus interaction. Following incubation, the mixture was added to Vero E6 cell monolayers and incubated at 37 °C for 1 h to allow viral adsorption. The inoculum was then removed, and the cells were overlaid with 1 mL of DMEM containing 2.5% FBS and 0.9% carboxymethyl cellulose (CMC). Plates were incubated at 37 °C with 5% CO₂ for 14 days. Plaques were visualized by fixing the cells with 10% formaldehyde for 30 min, followed by thorough washing. Plaques were manually counted by visual inspection. The neutralization curve was fitted to a sigmoidal model, and the half-maximal inhibitory concentration (IC₅₀) was calculated to determine the neutralizing potency of the antibody.

Cryo-EM data collection

EBOV-GP-dMLD trimer was mixed with nanobodies (1A10 or BA2) at a molar ratio of 1:3.2 and then incubated at 4 °C overnight to form a complex. The complex was further purified by Superdex 200 column (GE Healthcare). 3 μl of purified complexes of at 0.5 mg/mL were used for grid preparation. Each of the complexes was applied to freshly glow-discharged Quantifoil R1.2/1.3 300-mesh copper grids (EM Sciences) and blotted for 4.5 s at 4 ˚°C under 100% chamber humidity and plunge-frozen in liquid ethane using a Vitrobot Mark IV (FEI). GP-dMLD/1A10 complex data was collected using CRYO ARM 300 electron microscope (JEOL, Japan) operating at 300 kV equipped with a K3 direct electron detector (Gatan, USA). while GP-dMLD/BA2 complex data was collected using a Titan Krios G3 operated at 300 kV (Thermo Fisher Scientific), equipped with a Gatan K3 detector. movies were collected at a nominal magnification of 50,000x, 105,000x for 1A10/BA2-GP complexes, respectively. Parameters for data acquisition was showed in Table S1.

Cryo-EM data processing, model building and refinement

Cryo-EM data were processed using cryoSPARC v4.6.0⁴⁹, dose-fractionated movies were subjected to Patch motion correction and Patch CTF estimation^50,51. Particles were first picked using the Blob picker and Junk particles were removed through three rounds of 2D classifications. Particles from the good 2D classes were used for ab-initio reconstruction to generate five initial models. The initial models were set as the starting references for heterogeneous refinement (3D classification). The good 3D classes were then subjected to further non-uniform, and CTF refinements to generate the final maps with C3 symmetry applied. Map resolutions were determined using gold-standard Fourier shell correlation (FSC) at 0.143 between the two halfmaps (Figs. S3, S4).

The initial model of the EBOV GP/nanobody complex was constructed in Chimera v1.18, using the EBOV GP structure (PDB: 5JQ3) as the antigen template and an AlphaFold2-predicted models as the starting structures for the nanobodies⁵². Then these initial models were rigidly fitted into the cryo-EM map, followed by molecular dynamics flexible fitting (MDFF)⁵³. Models containing an asymmetric unit (a protomer containing one GP1, one GP2 and one nanobody) were manually adjusted using Coot-0.8.9 and refined using phenix.real-space refinement for multiple rounds until the final reliable models were obtained^54,55. The final coordinates for deposition were generated by building the biological unit (full trimer) in UCSF Chimera from the protomer model using the C3 symmetry⁵⁶. Then the trimer was refined in PHENIX with NCS constraints to ensure proper protomer-protomer interfaces. The glycan structures were validated with Privateer. Statistics of the map reconstruction and model building are shown in Table S1. Figures were generated using UCSF Chimera X v1.9⁵⁷. The interface between EBOV-GP-dMLD and nanobodies were analyzed using QtPISA v2.1.0.

Identification of viral escape mutants

Escape studies were conducted using recombinant VSV-EBOV GP and VSV-SUDV GP as previously described^21,58,59. Escape variants were selected by incubating the viruses under antibody pressure ranging from 0.016 μg/mL to 50 μg/mL. After 1 h of incubation, the mixture was used to infect a monolayer of Vero E6 cells in a 24-well plate at a multiplicity of infection (MOI) of 1. The cells were then monitored for cytopathic effect (CPE) over 72–96 h. Once extensive CPE was observed, supernatants were collected and stored at −80 °C.

Total RNA was extracted from the cell population using a Virus DNA/RNA Extraction Kit 2.0 (Vazyme) on an Automated Nucleic Acid Extractor, VNP-32P (Vazyme), and then subjected to gene sequencing to identify variants in the EBOV/SUDV GPs. Each subsequent passage was performed by incubating 100 μL of the supernatant under the same antibody concentration until resistance to ≥10 μg/mL antibody was observed in at least one passage, or for up to 10 passages. Total RNA was reverse transcribed into cDNA using the HiScript III 1st Strand cDNA Synthesis Kit (+gDNA wiper). The filovirus glycoprotein (GP) gene was then amplified and sequenced using cDNA as the template.

Therapeutic evaluation of monoclonal and bispecific antibodies against MA-EBOV in mice

To investigate the therapeutic efficacy of candidate monoclonal and bispecific antibodies against mouse-adapted Zaire ebolavirus (MA-EBOV), female BALB/c mice (6–8 weeks old) were challenged intraperitoneally (IP) with 300 TCID₅₀ of MA-EBOV in a volume of 200 μL. Body weight was recorded on the day of infection (Day 0). At 24 h post-infection (Day 1), animals received a single IP administration of one of the following antibodies-BA2-Fc, 1A10-Fc, BA2-1A10-Fc-at a dose of 5 mg/kg (0.5 mg/mL, 200 μL) at either 24 or 48 h post-infection (Day 1 or Day 2). Mice treated with phosphate-buffered saline (PBS) served as negative controls. Mice were monitored daily for 14 days post-infection for changes in body weight and survival, with therapeutic efficacy assessed based on these parameters.

Assessment of antibody therapeutics in a Syrian hamster model of SUDV infection

To assess the protective efficacy of monoclonal and bispecific antibodies against guinea pig-adapted Sudan ebolavirus (SUDV-GP), groups of Syrian hamsters (5–6 weeks old; n = 6 per group) were inoculated intraperitoneally (IP) with 1000 TCID₅₀ of SUDV-GP in 500 μL on Day 0. Animals were randomized to receive a single IP dose of BA2-Fc, 1A10-Fc, or the bispecific antibody BA2-1A10-Fc (5 mg/kg; 1 mg/mL; 500 μL) at either 24 or 48 h post-infection (Day 1 or Day 2). A PBS-treated group served as the negative control. Body weight and survival were monitored daily for 14 days post-infection. Animals exhibiting a ≥ 20% reduction in baseline body weight were considered moribund and humanely euthanized in accordance with ethical guidelines.

Therapeutic evaluation of monospecific and bispecific antibodies against BDBV in mice

IFNAR⁻/⁻ mice (n = 5 per group) were challenged intraperitoneally with 100 PFU of recombinant vesicular stomatitis virus expressing Bundibugyo ebolavirus glycoprotein (VSV-BDBV) in a total volume of 200 μL. Antibody treatment was administered via intraperitoneal injection at either 24 or 48 h post-infection. Mice received a single dose of BA2-Fc, 1A10-Fc, or BA2-1A10 fusion antibody (2.5 mg/mL, 25 mg/kg, 200 μL) or PBS control (200 μL). Animals were monitored daily for body weight loss and survival for 14 days.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

S.C. and T.C.J. conceived this study. S.C., T.C.J., P.G., M.H.W., and X.H.Z designed and directed the research; M.H.W, M.Y.L, J.H., and C.B.Y. immunized the camel, screened the Nbs, and analyzed sequences; M.H.W, M.Y.L., and Y.H.S. constructed the plasmids and produced antibodies; M.H.W performed the antibody characterization in vitro and engineering assay; L.X.Z and M.H.W performed the surface plasmon resonance (SPR) experiments; X.H.Z., E.T.L., M.H.W, C.P., B.Y.Z., and J.G.Z. performed the virus neutralization assay; M.H.W., Y.X.G., S.L.L., and Z.L.Z. performed the cryo-EM data collection, analysis, and deposition with the supervision of Tengchuan Jin (BA2 complex) and Peng Gong (1A10 complex). X.H.Z., Y.F.Y, W.J.L, F.X.L., M.Y.C., H.L., and F.H.Y. performed the rodents challenge studies; S.C., T.C.J, P.G., M.H.W., and X.H.Z. edited and revised the manuscript; all authors read and approved the paper.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

Source data are provided as a Source Data file. All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary information. Atomic coordinates and structure factors for the reported cryo-EM structures have been deposited with the Protein Data Bank and Electron Microscopy Data Bank under accession numbers 9VTS/EMD-65343 (https://www.rcsb.org/structure/unreleased/9VTS; https://www.ebi.ac.uk/emdb/EMD-65343), 9VT4/EMD-65314 (https://www.rcsb.org/structure/unreleased/9VT4; https://www.ebi.ac.uk/emdb/EMD-65314). Other structures for analysis, including 5JQ3, 7M8L, 6PCI, and 7SWD were obtained from the PDB. No new sequencing data has been generated as part of in this study. Source data are provided with this paper.

Competing interests

T.C.T., M.H.W., S.C., and E.T.L. are named inventors on two patents covering the sequences of the 1A10 and BA2 nanobodies described in this study. These patents pertain to the use of said nanobodies for the prevention, treatment, or diagnosis of Ebola virus infection and impose no restrictions on data publication. All other authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-70464-6.