Structural and mechanistic basis for antibody neutralization of the measles fusion protein

Abstract

Measles virus (MeV) is a highly contagious pathogen and a major global health threat. Resurgent infections, driven by insufficient vaccine coverage, waning herd immunity, and the vulnerability of immunocompromised populations, highlight the urgent need for effective countermeasures. While most vaccine-elicited antibodies target the hemagglutinin (H) protein, antibodies against the fusion (F) protein are also potent inhibitors. However, the specific target sites on this class I fusion protein, which undergoes dramatic conformational changes during entry, remain insufficiently characterized. Here, we characterize four mAbs targeting distinct F conformations. Structural analyses map antibody interactions, revealing that three neutralizing mAbs recognize the metastable prefusion conformation, while a non-neutralizing mAb binds only the post-triggered state. Biophysical and functional assays define distinct mechanisms of action: neutralization occurs either by stabilizing the prefusion protein or by preventing the extended intermediate from completing fusion. Uniquely, we detect a novel mechanism where an antibody prematurely triggers F activation but blocks the subsequent refolding required for viral entry. These findings provide the first detailed mapping of neutralizing epitopes on MeV F, establishing a framework for the rational design of F-targeted interventions.

Monoclonal antibodies (mAbs) represent a promising strategy as antiviral agents. Here, the authors selected four mAbs against Measles virus (MeV) and characterize their epitopes, specificities, and antiviral activities

Introduction

Measles is a highly contagious viral disease that has reemerged globally in recent years. Achieving widespread vaccine-induced immunity is critical to prevent outbreaks, particularly as unvaccinated and immunocompromised individuals remain at high risk for severe disease. Although the measles vaccine has an excellent long-term safety and efficacy record, neither global nor American vaccination rates have reached the 95% threshold required for herd immunity^1–6. Consequently, the recent resurgence in measles cases reflects a failure to vaccinate rather than vaccine failure. Over the past five years, declining vaccination rates have been fueled by disruptions in routine immunization programs during the COVID-19 pandemic and by the increasing impact of vaccine hesitancy or active anti-vaccination sentiments^1–6.

These developments increase the vulnerability of individuals who cannot be vaccinated due to their early age or medical contraindications. Post-exposure prophylaxis (PEP) with total immunoglobulin remains the only recommended option for these individuals; however, its reliability is threatened by declining MeV-specific antibody titers within the donor pool^1,7–15. With measles elimination efforts stalled and a growing number of high-risk individuals, there is an urgent need for effective antiviral therapies. However, currently, there is no antiviral treatment approved for measles. Passive immunization using monoclonal or polyclonal antibodies has long been a cornerstone in preventing and treating other viral infections, particularly in individuals who are unable to receive active vaccination. However, only a limited number of monoclonal antibodies (mAbs) have received clinical approval for the treatment of those viral infectious diseases^16–18, and none are available for measles.

Antibodies can exert antiviral effects by targeting viral glycoproteins critical for host cell entry. In MeV, this process involves two envelope proteins: the H protein, which binds host receptors, and the F protein, which mediates membrane fusion. MeV F is a class I trimeric fusion protein undergoing substantial conformational rearrangements from prefusion to postfusion states, a feature shared with many other enveloped viruses^19–22. Fusion is a dynamic, multistep process: following H-F interaction and receptor engagement, F transitions through an extended “pre-hairpin” intermediate, during which its heptad repeat regions (HRN and HRC) span the viral and host membranes. These regions then fold into a six-helix bundle (6HB), completing fusion and enabling viral entry.

MAbs targeting MeV H and F offer several advantages: they can neutralize the virus early in infection, avoid the toxicity and off-target effects of small molecules, and provide protection for immunocompromised or vaccine-ineligible individuals^1,16. Historical studies suggest that vaccine-induced virus neutralizing antibodies primarily target the H protein, but natural infection elicits robust neutralizing antibodies against both H and F, indicating that F is a promising target for mAbs development^23,24. Strategies to block viral entry include targeting: (i) H protein binding to hSLAMF1 (CD150) or Nectin-4 receptors, (ii) F activation via H-F interaction, (iii) conformational transitions exposing the F fusion peptide, and (iv) F refolding and pore expansion. We have generated and characterized a panel of novel chimeric (mouse Fab/human Fc) mAbs targeting MeV F, including the previously characterized antibody 77.1²⁵. We present here the structures and functions of four antibodies selected for distinct epitope specificity and mechanisms of action. Three of these mAbs exhibit potent neutralization against the two currently circulating wild-type MeV strains, providing a strong foundation for antiviral development. Our data demonstrate that the MeV F protein contains multiple neutralization-sensitive epitopes that can be targeted by potent mAbs. By combining antibodies with distinct mechanisms, broad and synergistic inhibition can be achieved, raising the barrier to resistance. A cocktail of long-acting humanized anti-F antibodies could serve as a critical countermeasure for exposed health care workers or individuals unable to receive vaccination, for outbreak control, and as a complement to existing PEP strategies.

Results

Neutralization activity and F protein reactivity of chimeric monoclonal antibodies against MeV

We generated and characterized a panel of mouse monoclonal antibodies (mAbs) targeting the MeV F protein. The Fab regions of four of these antibodies were grafted onto a common human Fc backbone. Here, we present the functional and structural properties of these chimeric mAbs, comprising three novel antibodies (Y10F, H8, and C6) and the previously characterized 77.1 (described in ref. ²⁵) for comparison.

We first assessed the ability of these antibodies to recognize a variety of F glycoproteins, including F from two currently circulating wild-type strains (B3 and D8), the vaccine strain (Moraten), and two previously characterized B3 F variants (L454W F), which is characterized by a destabilized prefusion conformation^26,27, and E455G F, which contains a stabilized prefusion conformation^25,28 (Fig. 1A). MAbs 77.1 and Y10F recognized each of the currently circulating wild-type F proteins, the Moraten-vaccine F, and the prefusion-stabilized E455G variant. Recognition of B3 F carrying the destabilizing L454W mutation, however, was markedly reduced. In contrast, mAb H8 recognized all F proteins tested, including both pre- and postfusion forms. mAb C6 exhibited a different pattern: it did not recognize wild-type or stabilized E455G F variant and instead, it recognized only the Moraten-vaccine F and the destabilized L454W variant, consistent with its specificity for post-triggered conformations²⁶.

Fig. 1. Epitope Recognition and Neutralizing Activity of Anti-F Chimeric Monoclonal Antibodies.

A Recognition of cell-surface expressed F protein from the wild-type B3 and D8 strains (the only two currently circulating wt MeV strains), vaccine strain Moraten, destabilized mutant B3 L454W, and stabilized mutant B3 E455G. Cells expressing F variants were incubated for 24 h and analyzed for binding to chimeric monoclonal antibodies (Y10F, 77.1, H8, and C6). Data are presented as mean ± SEM of n = 3 biologically independent experiments, each performed in technical triplicate. Source data are provided as a Source Data file. B Antiviral potency of mAbs measured by plaque reduction neutralization tests (PRNT). We assessed the antiviral potency of the four mAbs in a viral entry assay using two circulating wild-type strains (B3 and D8), a vaccine strain (Moraten), and a neuropathogenic variant (B3 L454W F). Vero-hCD150 cells were infected with approximately 250 plaque-forming units (PFU)/well of each virus in the presence of increasing concentrations of mAbs 77.1 (violet circles), H8 (blue squares), Y10F (red triangles), and C6 (green diamonds) as indicated on the x-axis. After 2 h, the inoculum was removed and replaced with 1% methylcellulose. Infection was monitored for 72 h, with images acquired every 24 h using a Cytation 5 imager. Plaques were quantified at 48 h post-infection. Data represent mean ± SEM from n = 3 biologically independent experiments, each performed with technical duplicates. Source data are provided as a Source Data file. C Surface plasmon resonance (SPR) sensorgrams collected on a Carterra LSA with immobilized antibodies and prefusion F_ECTO as the analyte. Association (0-350 s) and dissociation phases (350–1300 s) were globally fit to a 1:1 Langmuir model across a concentration series to obtain k_on and k_off; equilibrium dissociation constants (K_D= k_off/k_on) are reported for each antibody. Binding kinetics were determined from a single measurement (n = 1 technical replicate) per antibody. Source data are provided as a Source Data file. D 2D class averages obtained from negative-stain electron microscopy (NS-EM) of the Fabs-F_ECTO complexes. The last column shows a graphic overlay with labels showing the locations of F_ECTO and Fabs.

Neutralization potency, measured by plaque reduction neutralization tests (PRNT) against four recombinant viruses (B3, D8, B3 F-L454W, and vaccine strain Moraten), revealed that 77.1 and Y10F were the most potent antibodies (Fig. 1B). 77.1 exhibited IC₉₀ values ranging from 0.15-0.45 nM, while Y10F values ranged from 1.3–42 nM. H8 was less potent overall (IC₉₀: 11-63 nM) but showed higher activity against the neuropathogenic F-destabilized B3 F-L454W²⁸ variant (23 nM) compared to Y10F (42 nM). In contrast, C6 exhibited no measurable neutralizing activity (using a specific mechanistic assay described in Supplementary Fig. 1 we determined that the C6 epitope is exposed prior to complete refolding into the postfusion conformation, indicating that C6 is a post-trigger-specific antibody).

We next quantified antibody binding to purified, stabilized, soluble prefusion F ectodomain (F_ECTO²⁵) using high-throughput surface plasmon resonance (SPR) on a Carterra LSA (Fig. 1C). Unlike constructs locked by disulfide bridges, this F protein relies on natural substitutions to stabilize the prefusion state, which allows it to undergo conformational transitions, but at a more gradual, controlled rate than the wild-type²⁵. Three antibodies (77.1, H8, Y10F) bound this purified, prefusion MeV F. In contrast, the post-trigger-specific control mAb C6 showed no detectable binding to this prefusion-stabilized F, as expected for its post-triggered conformation-specific recognition.

We observed distinct kinetic signatures across the three binders. mAb 77.1 showed the highest affinity (K_D = 1.71 ± 0.03 × 10⁻⁹ M). H8 showed an intermediate association and had the slowest dissociation, with a K_D of 3.17 ± 2.06 × 10⁻⁹ M. Antibody Y10F displayed the slowest association rate (k_a=1.79 × 10⁴ M⁻¹s⁻¹), and dissociated more quickly, yielding a lower overall affinity (K_D = 9.99 ± 0.18 × 10⁻⁸ M). These trends indicate that Y10F forms a shorter-lived complex than the other two characterized antibodies (Fig. 1C and Table 1).

Table 1.

Fitting parameters of antibody kinetics data from LSA

Antibody	H8	Y10F	77.1	C6
ka (M−1s−1)	23640 ± 810	17900 ± 300	121000 ± 2000	ND
kd (s−1)	(7.48 ± 4.82) × 10⁻⁵	(1.790 ± 0.009) × 10⁻³	(2.07 ± 0.09) × 10⁻⁴	ND
KD (M)	(3.17 ± 2.06) × 10⁻⁹	(9.99 ± 0.18) × 10⁻⁸	(1.71 ± 0.07) × 10⁻⁹	ND
R2	0.99967	0.99942	0.99987	ND

ND Not determined.

To visualize the epitopes of each antibody, we performed negative-stain EM (NS-EM) on Fab-F_ECTO complexes (Fig. 1D). The structure of mAb 77.1, at the base of prefusion F, was previously described²⁵. For the new antibodies, class averages showed that Y10F engages the prefusion apex of F, whereas H8 can bind both prefusion and postfusion conformations. In contrast, C6 only produced apparent postfusion complexes. Taken together, these results indicate that three mAbs that bind prefusion epitopes are neutralizing (77.1, Y10F, H8), while the mAb that binds only a post-trigger epitope is not neutralizing (C6) (see also Supplementary Fig. 1).

Stability of the F protein in antibody complexes

To assess how antibody binding affects the stability of the prefusion F_ECTO, we performed nano differential scanning fluorimetry (nanoDSF; NanoTemper Prometheus Panta) (Fig. 2A). F_ECTO was incubated with a threefold molar excess of Fab over F trimer for 30 min at room temperature. F_ECTO alone exhibited a melting temperature (T_m) of 47.2 °C, consistent with prior measurements²⁵. The presence of the post-trigger-specific C6 had a negligible effect on the F_ECTO T_m ( ~ 0.7 °C). Although prefusion-specific antibodies (Y10F and H8) altered F_ECTO thermal stability, they did so in distinct ways. Specifically, Y10F increased it by 8.3 °C to 55.5 °C. In contrast, H8 decreased the F_ECTO T_m by 2.9 °C to 44.3 °C. The Fab melting transitions could be observed at higher temperatures, above the unfolding temperature of the F_ECTO-Fab complex (Fig. 2A).

Fig. 2. Thermal stability, size, and conformational effects of F-mAb complexes.

A First-derivative nanoDSF traces for F_ECTO-Fab complexes. F_ECTO-Fab complexes (F_ECTO-Y10F, F_ECTO-H8, and F_ECTO-C6) were assembled 15 min before measurement, loaded into capillaries, and monitored for thermal unfolding by intrinsic fluorescence at 330 nm. Dashed, color-coded lines mark F_ECTO unfolding temperatures: F_ECTO alone, 47.2 °C; with Y10F, 55.5 °C; with H8, 44.3 °C; with C6, 47.9 °C. Unfolding of Fabs was observed in the range of 60–80 °C. Source data are provided as a Source Data file. B In parallel, dynamic light scattering (DLS) was recorded, and the cumulant radius was plotted as a function of temperature (°C). Source data are provided as a Source Data file. C C6 binding to post-triggered F in the presence of H8 or Y10F, measured after incubation at 4 °C or following a 50 °C temperature shift. To assess the effect of H8 and Y10F on native full-length MeV F stability, cells expressing wild-type B3 F were incubated with 200 nM of either antibody. Exposure of the post-triggered F epitope was measured using C6 after 90 min at 4 °C or following a temperature shift (30 min at 4 °C then 30 min at 50 °C). The number of C6-positive cells, reflecting post-triggered F exposure, is plotted on the y-axis. Data represent the mean ± SEM of n = 3 (mock) or n = 5 (antibody-treated) biologically independent experiments. The y-axis is on a logarithmic scale. Source data are provided as a Source Data file. D H8-induced activation of the MeV F protein was assessed in the presence of Y10F. To evaluate the effect of Y10F on H8-mediated F triggering, HEK293T cells expressing either wild-type B3 F or the B3 E455G F mutant were pre-incubated with Y10F at the indicated concentrations (room temperature), and subsequently exposed to increasing concentrations of H8 (x-axis) at 37 °C. Post-triggered F exposure was detected after 90 min using the conformation-specific mAb C6. The number of C6-positive cells, reflecting the extent of post-triggered F exposure, is plotted on the y-axis (in logarithmic scale). Data points represent the mean ± SD of n = 3 technical replicates from a single representative biologically independent experiment. Source data are provided as a Source Data file. E Size-exclusion chromatography analysis of purified MeV F ectodomain (F_ECTO). Chromatograms show F_ECTO immediately after thawing (0 h, light green), after 24 h at room temperature (dark green), and Fab H8 alone (blue). F_ECTO incubated for 24 h showed increased absorbance in the void volume (V₀), indicating aggregation and/or large conformational rearrangements (such as refolding to postfusion state). The integrated V₀ region (7-10 mL) is shaded. Source data are provided as a Source Data file. F SEC profiles of F_ECTO incubated with H8 Fab at an approximately 1:1 molar ratio (monomer: Fab) for increasing times (0.25-24 h). With longer incubation, the V₀ peak progressively increased while the peak corresponding to the intact F_ECTO-H8 complex decreased, consistent with time-dependent triggering and structural rearrangement of F. Source data are provided as a Source Data file. G Quantification of the integrated V₀ signal from panel C plotted as a function of incubation time. Fitting to a single-exponential model yielded a half-life (t_½) of 0.98 ± 0.58 h for H8-induced triggering. The residuals of the fit are shown in the inset. Source data are provided as a Source Data file. H Representative two-dimensional class averages from single-particle NS EM across six time points from the same reaction. Class 1 corresponds to prefusion, unbound F_ECTO. Class 2 shows prefusion F_ECTO bound by a single H8 Fab. Class 3 represents a top view of F bound by three H8 Fabs. Class 4 shows the postfusion conformation bound by three H8 Fabs. Class 5 contains H8 Fab bound to a fragment of F_ECTO; due to its heterogeneity, this class was not included in occupancy analyses. Classes 3 and 4 are shown separately because some datasets contained class 3 without class 4, suggesting an alternative or intermediate conformational state. I Time-resolved occupancies of the EM classes shown in panel E, determined from single-particle NS EM datasets. Class distributions were obtained using an identical processing workflow and a comparable number of micrographs at each time point. Empty cells indicate that a given class was not detected at the given time point. Source data are provided as a Source Data file. J Stacked area plot showing the evolution of class occupancies over time. The x-axis is displayed on a pseudo-logarithmic scale (ln(t + 1)) to simultaneously visualize rapid initial conformational changes and long-term equilibrium.

In parallel, we monitored the cumulant (hydrodynamic) radius as a function of temperature (Fig. 2B, Supplementary Fig. 2). The F_ECTO trimer alone maintained a radius of ~9 nm until reaching an aggregation onset temperature (T_on) of ~39.7‒41.4 °C, after which the radius increased continuously. The sample containing the post-trigger-specific mAb C6 displayed a nearly identical aggregation profile (T_on = 44.6 °C), consistent with its inability to bind the prefusion ectodomain.

In contrast, the neutralizing antibodies distinctively modulated these hydrodynamic properties. The F_ECTO-Y10F complex appeared more compact initially and maintained this stability up to 52.7 °C, delaying the onset of aggregation relative to the protein alone. The F_ECTO-H8 complex exhibited a unique profile: despite H8 lowering the melting temperature of F_ECTO (Tm = 44.3 °C, Fig. 2A), the complex did not undergo immediate aggregation. Instead, it maintained a stable radius until 67.4 °C, a temperature corresponding to the thermal unfolding of the Fab itself. Collectively, these analyses demonstrate that while C6 has no impact on stability, Y10F stabilizes the prefusion trimer, and H8 prevents the formation of large aggregates despite lowering the activation barrier.

Mechanistically, the data suggest opposing modes of action for Y10F and H8: Y10F stabilizes the prefusion conformation and raises the energy barrier for activation, thereby preventing fusion triggering, whereas H8 lowers the activation barrier and promotes premature triggering of the fusion machinery.

To validate these mechanistic insights in the context of native membrane-anchored F, we performed cell-surface staining with the post-trigger-specific mAb C6 after preincubating cells at either 4 °C or 50 °C (the latter to thermally induce F triggering) in the presence of Y10F or H8 (Fig. 2C). Because C6 recognizes an epitope exposed only upon F triggering (Supplementary Fig. 1), its binding serves as a proxy for F activation. At 4 °C, C6 bound F in the presence of H8, but not in the presence of Y10F. At 50 °C, C6 showed no binding to F in the presence of Y10F. These findings corroborate the biophysical data, supporting a model in which Y10F raises the conformational activation threshold of F, thereby maintaining the prefusion state, while H8 lowers this barrier and drives premature activation.

We next asked whether the stabilizing effect of Y10F could block H8-mediated F activation. For the experiment shown in Fig. 2D, cells expressing either wild-type B3 F or the B3 E455G F mutant (which require more energy for activation^25,28) were incubated overnight at 37 °C. The following day, Y10F was added for 30 min (at the indicated final concentration), after which H8 was added to achieve final concentrations of 100, 50, 25, 12.5, 6.25, 3.13, or 1.56 nM. Cells were then transferred to 37 °C for 60 min. Following incubation, cells were stained with the conformation-specific mAb C6 to quantify exposure of the post-trigger epitope. In the absence of Y10F, H8 induced a dose-dependent increase in F triggering, which was more pronounced for the B3 F wt than for the stabilizing E455G mutant. In contrast, at the highest concentrations of Y10F, H8-induced F triggering was also completely abolished for the B3 F wt.

To define the kinetics of H8-induced triggering in purified protein, we performed time-resolved size-exclusion chromatography and followed the accumulation of signal in the void volume. F_ECTO, incubated alone at room temperature for 24 h, showed only a small, gradual increase in the 7-10 mL region, indicating slow, spontaneous refolding toward the postfusion state (Fig. 2E). In contrast, F_ECTO incubated with H8 Fab displayed a much stronger, time-dependent increase in this void-volume peak, accompanied by loss of the intact complex peak (Fig. 2F). Plotting the integrated void-volume signal as a function of incubation time revealed a single-exponential behavior, yielding an estimated half-life of 0.98 ± 0.58 h for H8-induced triggering (Fig. 2G).

To visualize the structural species that arise during this reaction, we next analyzed time-resolved negative-stain electron microscopy datasets of F_ECTO-H8 mixtures. Across all six time points, we repeatedly identified four main classes: prefusion F_ECTO without antibody (class 1), prefusion F_ECTO bound by a single H8 Fab (class 2), a top view of F bound by three Fabs (class 3), and a side view of F in the postfusion conformation with three bound Fabs (class 4, Fig. 2H). A fifth, highly heterogeneous class contained H8 bound to F fragments, free Fab, and other mixed species and was therefore excluded from quantitative analyses. Notably, classes 3 and 4 did not always appear together in the same datasets, suggesting that they may represent distinct conformational endpoints or intermediates rather than 2 distinct views of a single class. Over time, the occupancy of classes 1 and 2 decreased and became negligible at later time points, whereas classes 3 and 4 increasingly dominated the particle population (Fig. 2I-J). This shift in class distribution aligns with the size-exclusion data and supports a reaction pathway in which H8 converts prefusion F into multi-Fab-bound, post-trigger-like conformations and the postfusion state.

Cryo-EM mapping of Y10F and H8 epitopes on the F protein

To define the antibody epitopes on F, we determined cryo-EM structures of the F_ECTO in complex with Y10F alone, with Y10F and 77.1 bound simultaneously, and with H8 alone.

The F_ECTO-Y10F dataset exhibited mild preferred orientation, with many particles showing a tilted top view (Supplementary Fig. 3). Despite this, we reconstructed a 2.59 Å map that revealed asymmetric binding of a single Y10F Fab at the apex of F_ECTO (Fig. 3A). Y10F binds the region in the center of the apex, surrounded by the six glycosylation sites positioned on the three F2 subunits at positions 61 and 67 (Fig. 3A). Both heavy and light chains contribute to a dense interaction network. The heavy chain forms nine hydrogen bonds distributed across three protomers (including one F2 subunit and two F1 subunits), with residues N55 and N31 accounting for six of these nine bonds (Supplementary data 1). The light chain forms five hydrogen bonds with the same two F1 chains. The buried interface area is ~870 Ų, ~70% of which is contributed by the heavy chain, with a binding free energy (ΔG) estimated by PISA of −8.6 kcal/mol, ~85% of which arises from heavy-chain contacts (Fig. 3B).

Fig. 3. Cryo-EM analysis of F_ECTO-Y10F and dual F_ECTO-Y10F-77.1 complexes.

A Cryo-EM density and atomic model of the F_ECTO trimer with a single Y10F Fab bound asymmetrically at the apex. The model is shown as ribbons: Y10F heavy chain in light red, light chain in yellow; F1 protomers in shades of green and F2 protomers in shades of gray. Upper and lower panels show side and top views, respectively. Glycans near Y10F Fab are indicated with arrows. B Epitope-paratope contacts between Y10F and F_ECTO. The top panel highlights heavy-chain interactions, and the bottom panel highlights light-chain interactions. Both chains engage an asymmetric epitope on F. Interacting residues are color-coded by chain, matching the coloring from (A). C Cryo-EM map and model of F_ECTO bound by Y10F and 77.1 Fabs. Due to a symmetry mismatch (C1 for Y10F and C3 for 77.1) and partial Fab occupancy, only one copy of 77.1 is resolved in the final locally refined reconstruction.

We also simultaneously determined the structure of the F_ECTO-bound complex with Y10F and 77.1 (Fig. 3C) to test co-occupancy and any binding-induced conformational changes resulting from simultaneous binding of these two antibodies. Although both Fabs were supplied in excess, the complex displayed partial occupancy of each Fab, with individual particles displaying one, two, or three copies of 77.1. The combination of preferred orientation, asymmetric nature of the complex, and incomplete occupancy complicated particle selection. However, through symmetry expansion and focused masking, we obtained a reconstruction in which single copies of 77.1 and Y10F were fully resolved. Superposition of the Y10F-only and Y10F + 77.1 models yielded a Cα RMSD of 0.79 Å, indicating no significant global rearrangement of F_ECTO. Local differences were restricted to the 77.1 epitope, particularly residues 433-442, which showed an RMSD of ~2 Å, consistent with previous descriptions of 77.1-induced adjustments²⁵. These data demonstrate that Y10F and 77.1 can co-bind F, supporting their use in an antibody cocktail.

To characterize H8 binding in both conformational states of F, we used two sample-preparation strategies guided by negative-stain EM (Fig. 1D) and nanoDSF (Fig. 2) results. To solve a prefusion complex (Fig. 4A), F_ECTO and H8 Fab were incubated for 15 min at room temperature before vitrification to promote complex formation while minimizing triggering; longer incubations increased the proportion of postfusion particles. To solve a postfusion complex (Fig. 4B), F_ECTO and H8 were incubated overnight at 4 °C and then applied to graphene oxide grids.

Fig. 4. Cryo-EM structures of H8 bound to the F_ECTO in pre- and postfusion states.

A Cryo-EM density and atomic model of prefusion F_ECTO symmetrically bound by three copies of Fab H8. Ribbon representation shows H8 heavy chains in light purple and light chains in light blue. Side (top) and top (bottom) views are shown. B Close-up of the prefusion interface. H8 engages the outer face of domain II exclusively through its heavy chain. C Cryo-EM density and atomic model of postfusion F_ECTO with three copies of Fab H8 bound. Side (top) and top (bottom) views are shown. D Close-up of the postfusion interface. As in the prefusion state, H8 binds exclusively via its heavy chain and targets the same region on domain II.

In the prefusion state, H8 bound symmetrically to all three protomers at an epitope on the outer face of domain II, opposite the F vertex contacted by 77.1, producing a triangular-like structure. The epitope centered on the single loop in domain II (residues 396-405). The binding was dominated by the heavy chain and was stabilized by ten hydrogen bonds. PISA also identified a single salt bridge attributed to the light chain in proximity to residue G115, the first residue resolved in F1, placing the binding site near the furin cleavage site (Supplementary data 1). H8 binding induced a local shift of the 396-405 loop, with a maximal displacement of 2.5 Å toward the Fab.

H8 recognized the same topological site on postfusion F. Because domain II rotates relative to the rest of the protein during refolding around residue G433²⁵, the Fab reoriented accordingly; whether H8 remains bound throughout this transition is unclear, as the lack of a transmembrane region in F_ECTO allows for greater degrees of freedom and facilitates the movement of the HRC domain. In the postfusion complex, contacts were again mediated mainly by the heavy chain, now comprising nine hydrogen bonds and two salt bridges, one per antibody chain (Supplementary data 1). The interface areas in pre- and postfusion were comparable (461 vs. 486 Ų). ΔG values for both conformations were estimated to be similar ( − 4.2 kcal/mol for prefusion and −3.1 kcal/mol for postfusion).

In summary, cryo-EM resolved Y10F as an apex-binding antibody that fits between glycan clusters on F2, confirmed the ability of Y10F to co-occupy F with 77.1 without major conformational changes of F, and showed that H8 targets a solvent-exposed loop in domain II that is accessible in both prefusion and postfusion conformations (see Supplementary Fig. 4 for antibody footprint on the F protein).

Ex vivo, in vitro and in vivo efficacy

We next determined whether the mAbs could inhibit MeV infection of authentic target cells (Fig. 5A-D). Vero cells stably expressing either human CD150 (Vero-hCD150) or Nectin-4 (Vero-hNectin-4), the two MeV entry receptors for wt MeV, were used initially, with parental cells serving as a negative control. As authentic immune target cells, assays were performed on human Epstein-Barr virus (EBV)-transformed B lymphoblastoid cell lines (B-LCL). 77.1 and H8 exhibited robust neutralizing activity against MeV on all receptor-positive cells; H8 with lower potency. Functionality of Y10F was variable: effective MeV neutralization on Vero-hCD150 and Vero-hNectin4 was observed (IC₅₀ of 6.25 nM and 0.34 nM, respectively), but Y10F could not prevent MeV infection of CD150-positive B-LCL (Fig. 5A). Across all conditions, 77.1 consistently outperformed Y10F and H8 in neutralizing MeV.

Fig. 5. In Vitro, Ex Vivo, and In Vivo Efficacy of Anti-F Monoclonal Antibodies.

A Neutralizing activity of Y10F, H8, and 77.1 against recombinant MeV strain rMeV-KS-Venus(3) was assessed in B-LCL, SLAMF1 (CD150)-expressing Vero cells (Vero-hCD150), and Nectin-4-expressing Vero cells (Vero-hNectin-4). Serial dilutions of each mAb (100-0.05 nM) were incubated with MeV for 1 h prior to inoculation onto target cells. Infection was monitored for 7 days, and neutralization efficacy is expressed as IC₅₀ values (nM). uLoD upper limit of detection. Data represent the mean derived from n = 4 technical replicates per condition. Source data are provided as a Source Data file. B Activity of Y10F, H8, and 77.1 in a MeV dissemination assay. Infection percentages in B-LCLs were determined 48 h post co-culture of 150 rMeV-KS-Venus(3) infected B-LCLs with 6 × 10⁵ uninfected B-LCLs. Inhibition of MeV dissemination is shown as a dose-response curve; inhibition of dissemination was expressed as half-max inhibitory concentrations (IC₅₀). Data represent the mean derived from n = 4 technical replicates per condition. Source data are provided as a Source Data file. C Airway organoids differentiated at air-liquid interface (AO at ALI) were prepared by removing mucus and mechanical disruption to expose Nectin-4. Y10F, H8, and 77.1 (1000, 100, and 10 nM) were incubated with rMeV-KS-Venus(3) for 1 h and applied to the apical compartment for 2 h. After removal of the virus-antibody mixture, ALI conditions were restored. Infection was visualized by fluorescence imaging at 3 days post-inoculation (infected areas shown in black). D Semi-quantitative interpretation of infection in AO at ALI: “+++“ indicates extensive infection, “++“ moderate infection, “+“ low-level infection, and “-“ no detectable infection. The untreated (NT) sample showed the highest infection levels among all tested conditions. E, F Antibody transfer experiment to assess in vivo protective efficacy of mAbs. Cotton rats (n = 5 biologically independent animals per group) received intraperitoneal (IP) injection of 1 mg/kg of the indicated mAbs or IgG purified from human sera (Carimune, CSL Behring) 14 h before (E) or 24 h after (F) challenge with MeV. Animals in the control group received PBS before the challenge. Values correspond to MeV titers in lung tissue collected 4 days post-infection. Significance was determined using a two-sided Mann-Whitney-Wilcoxon test with Holm-Bonferroni correction for multiple comparisons against the PBS control group. Exact P values for E): PBS vs. Y10F P = 0.0119, PBS vs. 77.1 P = 0.0097. Exact P values for F): PBS vs. Y10F P = 0.1161 (not shown), PBS vs. 77.1 P = 0.0079. P value annotation legend: ns (non-significant): p > 0.05, *: 0.01<p ≤ 0.05, **: 0.001<p ≤ 0.01. Box plots represent the distribution of viral titers: the center line represents the median (50th percentile), the box limits indicate the 25th and 75th percentiles, and the whiskers extend to the minimum and maximum values within 1.5-times interquartile range (IQR) from the box limits. Individual data points are overlaid. LoD = limit of detection. Source data are provided as a Source Data file.

To further evaluate efficacy, antibodies were evaluated in a dissemination assay in human Epstein-Barr virus-transformed B-lymphoblastic cells (B-LCL)²⁹. Consistent with the virus neutralization results, 77.1 inhibited MeV dissemination better than Y10F and H8, with an IC₅₀ of 12.7 nM. Interestingly, while unable to block infection in neutralization assays, Y10F could inhibit MeV dissemination in B-LCL cultures at an IC₅₀ of 22.3 nM (Fig. 5B).

To confirm functionality in a physiologically relevant epithelial cell context, bronchial ciliated human airway organoids differentiated at air-liquid interface (AO at ALI) were used. Prior to exposure to a mixture of mAbs and MeV, cultures were washed to remove mucus and mechanically disrupted to expose Nectin-4. All three mAbs inhibited MeV infection in this model (Fig. 5C). Notably, 77.1 achieved complete neutralization at all tested concentrations, whereas Y10F and H8 allowed residual infection at 10 nM and minimal infection at 100 nM (Fig. 5D).

We next evaluated the prophylactic efficacy of these mAbs in vivo using a cotton rat model of MeV infection³⁰ (Fig. 5E). Cotton rats were inoculated 14 h before infection with a single dose (1 mg/kg) of the indicated mAbs. Four days after infection, lungs were collected and viral titers quantified (y-axis). A purified polyvalent human IgG (huIgG) preparation, which is used clinically for patients with primary immune deficiency and immune thrombocytopenia purpura, exhibited a neutralization titer of 640 at 60 mg/mL and served as a control. As expected, mAb 77.1 confirmed its strong in vivo efficacy previously demonstrated²⁵. The second-most potent mAb, Y10F, reduced viral titers to levels near the limit of detection. The other two mAbs (C6 and H8) performed comparably to huIgG, each reducing viral titers relative to PBS (diluent)-treated animals but without a significant difference.

We then tested mAb efficacy when administered after infection (Fig. 5F). Cotton rats were inoculated 24 h after infection with a single dose (1 mg/kg) of the indicated mAbs. Four days post-infection, lungs were collected, and viral titers were quantified (y-axis). Under these conditions, mAb 77.1 showed significant efficacy, whereas the other mAbs did not reduce viral titers. We then asked whether the lack of activity in vivo of Y10F and H8 was due to faster clearance. We inoculated cotton rats intraperitoneally with 1 mg/kg of the three mAbs and the Human IgG. Blood was collected at 14, 38, 62, 86, and 110 h post-inoculation. We assessed the neutralizing activity for all and the presence of the three mAbs using a semi-quantitative cell-based ELISA at the time of infection and during the course of the challenge study (Supplementary Fig. 5).

The data show that only sera from animals inoculated with mAb 77.1 exhibited neutralizing activity (Supplementary Fig. 5A, expressed as mIU/mL y-axis), whereas all three mAbs remained at stable circulating levels throughout the experiment (14-110 h post-inoculation). However, at the starting serum dilution used for the neutralization assay (1:40), the other antibodies did not block viral infection. By interpolating the serum binding data onto standard curves of purified mAbs (Supplementary Fig. 5D), we estimated that serum concentrations remained in the low nanomolar range for all three antibodies (Table S2). Specifically, H8 showed the highest retention, with concentrations ranging from approximately 40 to 65 nM, while Y10F and 77.1 remained stable at 17-33 nM (Table S2). These results indicate that the lack of therapeutic efficacy observed for Y10F and H8 in the post-infection setting was not due to rapid clearance or instability of the antibodies in vivo.

Mechanism of fusion inhibition

Since Y10F and 77.1 can simultaneously bind to the F protein (Fig. 3C), we tested whether they exhibit synergistic inhibition of fusion. Fusion was measured using an α/ω β-galactosidase complementation assay, where HEK293T target cells expressing hCD150 and ω subunit were co-cultured with effector cells expressing MeV H/F and α subunit; and reconstituted enzyme activity was quantified by luminescence (Fig. 6A)²⁵. Data were analyzed using SynergyFinder (https://synergyfinder.fimm.fi/) using the BLISS model. We observed detectable moderate synergism between Y10F and 77.1 in inhibiting viral fusion.

Fig. 6. Mechanism of Action of Anti-F Monoclonal Antibodies.

A Synergistic inhibition of cell-cell fusion by 77.1 and Y10F antibodies. B3 H and F co-expressing effector cells were co-cultured with hCD150-expressing target cells in the presence of 77.1 and Y10F antibodies at the indicated concentrations (x- and y-axes). Fusion was quantified after 18 h using a β-galactosidase complementation assay, and synergy was assessed using the BLISS independence model. A moderate synergistic effect was observed, with a BLISS score of 13.22. The area of highest synergism is highlighted by a white rectangle. Source data are provided as a Source Data file. B A bar chart summarizing the results of a mechanistic assay comparing various known inhibitors, including the newly described mAbs. For each condition, cells were categorized based on their interaction with red blood cells (RBCs): reversibly bound (dark blue) where the RBCs were only adhered by chimeric H and could be easily removed by treatment with zanamivir; irreversibly bound (pink) where RBCs were initially attached by chimeric H, but subsequent F protein refolding anchored the fusion peptide in the RBC membrane, making the RBCs attachment zanamivir-insensitive and retaining RBC on the HEK293T cell surface; and fused (light blue) where RBCs underwent membrane fusion with HEK293T cells expressing the fusion complex. Data represent the mean ± SEM of n = 3 biologically independent experiments (see Supplementary Fig. 9). Source data are provided as a Source Data file. C Mechanistic assay over a range of mAb concentrations (x-axis), in which both 77.1 and Y10F efficiently block fusion (light blue triangles). 77.1 traps the F in the extended state (red squares). Y10F prevents F activation and F insertion (dark blue circles). Data represent the mean ± SEM of n = 3 biologically independent experiments. Source data are provided as a Source Data file.

To define the stage of the fusion process at which each mAb acts, we used an assay measuring fusion of red blood cells (RBCs) with HEK293T cells expressing modified versions of MeV H and F (Fig. 6B)³¹. To distinguish F-mediated anchoring from H-mediated attachment, we employed a chimeric receptor-binding protein consisting of the MeV H stalk fused to the head domain of human parainfluenza virus type 3 hemagglutinin-neuraminidase (HPIV3 HN). This H-HN chimera allows releasable attachment to sialic acid receptors (via HN) while still permitting MeV H-mediated triggering of F. As expected, RBCs bound through HN-sialic acid interactions could be released by adding Zanamivir^25,31. In untreated conditions, ~38% of RBCs fused with HEK293T cells, ~22% irreversibly bound via F (fusion initiated, but incomplete), and ~40% were reversibly bound via HN only (fusion not initiated). Zanamivir treatment, which blocks HN-mediated attachment, prevented nearly all F-mediated anchoring, confirming that receptor engagement is required for F activation. Treatment with the HRC4³¹ peptide inhibitor yielded mostly irreversibly bound RBCs, indicating that F was activated, but fusion was blocked at a post-triggering step. By contrast, the FIP-HRC4^25,31 peptide inhibitor prevented nearly all F-mediated anchoring, consistent with inhibition of F activation. MAbs 77.1 and H8 blocked fusion in the same manner as HRC4, with most RBCs irreversibly bound, confirming that these two mAbs permit (77.1) or promote (H8) the F extension and allow fusion peptide insertion into the target cell, but block completion of the F transition. The profile observed in the presence of C6 at the tested concentration (200 nM) was statistically indistinguishable from that of mock-treated samples, consistent with their lack of efficacy. In contrast, mAb Y10F behaves similarly to FIP-HRC4³¹, preventing the initiation of F activation. A dose-response version of this assay, with 77.1 and Y10F, confirmed that these antibodies act through distinct mechanisms: at higher concentrations, 77.1 drives cells into an irreversibly bound state with F trapped in an intermediate conformation, whereas Y10F drives cells into a reversibly bound state with F remaining untriggered (Fig. 6C).

Viral evolution in the presence of the mAbs 77.1 and Y10F

We first evaluated the potential for viral escape from mAb 77.1 by subjecting MeV to increasing concentrations of the antibody over five serial passages. Viral replication ceased after the fifth cycle, likely due to the accumulation of deleterious mutations. Nevertheless, viral populations from cycles four and five were harvested and subjected to deep sequencing to identify potential resistance-associated genetic changes. Remarkably, despite prolonged selective pressure from mAb 77.1, we detected no significant point mutations with allele frequencies ≥10% or sequencing depths ≥20.

We next repeated the viral evolution experiment with both 77.1 and Y10F (since these two antibodies were the most potent in vivo, as shown in Fig. 5E), using a different strategy. Instead of progressively increasing antibody concentrations, viruses were passaged for over 50 days in alternating concentrations of mAbs to permit ongoing replication (see “Methods”). This approach yielded eight distinct viral populations: four selected in the presence of 77.1 and four in the presence of Y10F. Three of the four populations grown with 77.1 showed no mutations in the F protein (Supplementary Fig. 6), while one population contained a single amino acid substitution (M46V) (Fig. 7A). In contrast, all four populations selected with Y10F acquired one or more amino acid substitutions in F (Fig. 7A). Among these, only the F375S variant reached complete fixation (100% allele frequency), whereas the others displayed partial penetrance. Nevertheless, all Y10F-selected populations were resistant to Y10F inhibitory activity but retained sensitivity to 77.1, albeit with reduced potency compared to the parental virus (Supplementary Fig. 7). Similarly, M46V was also resistant to Y10F but retained sensitivity to 77.1, even though it was selected in the presence of 77.1. None of these mutations (Fig. 7B) mapped directly to the binding interface defined in Figs. 3 and 4. This suggests that viral evolution under antibody pressure favors allosteric changes that functionally alter the F protein dynamics, rather than mutations within the epitope itself.

Fig. 7. Viral Evolution Under Selective Pressure of 77.1 and Y10F mAbs.

A Viral evolution assay results from control (mock), as well as passages in the presence of Y10F and 77.1, with four replicates per mAb, were evaluated for mutations in the F gene. Only mutations with an allele frequency greater than 10% and a minimum read depth of 20 were included in the analysis. Source data are provided as a Source Data file. B Location of point mutations identified in A) mapped onto a single protomer of the prefusion F structure. Mutated residues are highlighted, including I498T, which is located in the membrane proximal region and shown schematically (in Supplementary Fig. 10 the MeV F protein with domains color-coded).

Discussion

Measles vaccines are highly effective but cannot be given to infants below the age of 9 months, or to immunocompromised or pregnant individuals, leaving them dependent on herd immunity^1,32,33. Population immunity to measles is fragile due to suboptimal vaccine coverage and the extreme transmissibility of MeV (R₀ = 12-18), driving resurgences in the U.S. and worldwide. MeV still causes 10-20 million infections and >200,000 deaths annually, with poor outcomes compounded by the post-measles “immune amnesia” that erodes immunity to other pathogens^{1–6,34–37}. No licensed antivirals exist, and current post-exposure prophylaxis with immune sera shows declining potency^7–15. While vaccine-induced neutralizing antibodies primarily target H, natural infection elicits potent antibodies against both the H and F proteins, highlighting F as a promising therapeutic target^{23,24,38–40}. Two recent preprint manuscripts from our group further characterize the human antibody response to MeV^41,42. In the first, we show that anti-F protein antibodies are found predominantly in older individuals, suggesting that wild-type infection elicits a stronger humoral response to the F protein than vaccination, which is consistent with prior observations in the literature^23,24. The second preprint examines the neutralizing activity of anti-F and anti-H antibodies isolated from human PBMCs. Notably, in that study, the most potent neutralizing antibodies target the F protein, supporting our conclusion that anti-F antibodies can be highly effective in vivo.

In the current study, we investigated the neutralizing activity of chimeric mAbs targeting the MeV F protein. These mAbs were generated in mice, with their variable (Fab) regions grafted onto human constant (Fc) domains. One mAb, 77.1, which was previously described²⁵, binds an epitope present in both the prefusion and extended conformations of the F protein and arrests the F conformational transition at that intermediate step. Here, we evaluated its potential for use in combination therapies and its barrier to viral resistance. A second mAb, Y10F, has previously been shown to possess neutralizing activity^43,44, but its sequence, epitope specificity, and mechanism of action were unknown. The other two mAbs, H8 and C6, also with unknown epitope structures, were selected for their distinct binding properties: H8 recognizes both pre- and post-triggered forms of F, whereas C6 binds only a post-triggered epitope.

The three neutralizing mAbs (Y10F, 77.1, and H8), which target three distinct sites, collectively demonstrate that the MeV F protein contains multiple vulnerable epitopes that can be effectively targeted. Y10F binds an apical prefusion epitope, stabilizes the prefusion conformation, and neutralizes infection both in vitro and in vivo. However, for Y10F, ex vivo efficacy is reduced and viral escape can arise through allosteric mutations outside the binding interface. 77.1 recognizes F in the prefusion state but stays bound after triggering and inhibits fusion at a late stage (as we have previously shown²⁵). 77.1 blocks MeV infection with high potency in authentic human tissue ex vivo, remains efficacious in vivo even when administered 24 h post-infection, and displays a high barrier to resistance. H8 engages epitopes accessible in both the pre- and post-triggered conformations and, uniquely, promotes F activation upon binding, revealing a novel mode of antibody-mediated fusion regulation. Interestingly, H8 binds the 396-405 loop, a region critical for the H-F interaction and subsequent fusion triggering. In particular, this region includes residue Leu-394 (L394), which is notable as a known L394S mutation results in reduced fusion function and a 35-50% decrease in H-protein interaction, even though the protein itself is fully processed⁴⁵. While EM revealed H8-bound postfusion structures, RBC assays indicate that H8 arrests the process at a refolding step. We attribute this apparent discrepancy to the soluble ectodomain’s lack of transmembrane constraints, which allows it to complete refolding transitions that are energetically or sterically restricted in the membrane-embedded, full-length protein. Thus, combining these structural insights with our cell-based triggering, time-resolved SEC, and negative-stain data, we hypothesize that H8 may bind this signal-receiving site, mimic the H protein, and activate F through a similar mechanism. In addition, H8 binding to F might influence the H-F interaction, thereby disrupting the glycoprotein complex on the surface. This binding profile of H8 is similar to that of the 1H1 antibody, which targets the Nipah virus F protein at an analogous epitope on the outer side of domain II⁴⁶. Like H8, 1H1 is also a neutralizing antibody and recognizes both the prefusion and postfusion states of the F protein. However, no data are currently available to support the 1H1 F triggering mechanism. This confirms that the premature activation induced by H8 represents a unique mechanism among this class of antibodies. In addition, H8 neutralizes a broad spectrum of MeV strains, including hyperfusogenic variants from neuropathogenic isolates²⁸ and blocks fusion mediated by all the F variants obtained from the viral evolution with Y10F and 77.1 (Supplementary Fig. 8). By contrast, C6, which recognizes only a post-triggered epitope in F (Supplementary Fig. 1), failed to block viral entry or fusion, underscoring the importance of antibodies that target pre- and intermediate conformations.

Collectively, these findings establish the MeV F glycoprotein as a rich and underexplored source of neutralizing epitopes. A cocktail of long-acting, humanized anti-F antibodies, such as Y10F and 77.1, could provide durable protection with a high barrier to resistance, complement vaccination, and address a critical therapeutic gap for MeV-exposed seronegative health care workers, and to individuals who are unvaccinated, vaccine-ineligible (either due to early age or medical conditions), or have waning immunity^{1–6,23,24,38–40}.

Methods

Ethics statement

All studies received approval by the Institutional Animal Care and Use Committee of The Ohio State University. Inbred cotton rats (Sigmodon hispidus) were purchased from Inotiv (formerly Envigo; West Lafayette, IN) and maintained in a University Laboratory Animal Resources barrier facility at The Ohio State University. Cotton rats were housed in NexGen Rat 900 polysulfone microisolator cages (Allentown Inc., Allentown, NJ) with 12:12 h light cycles, 22 ± 2 °C, and 20–70% relative humidity. Age ranges were 5-7 weeks. Both sexes were used in all experimental groups.

Plasmids

The genes of MeV genotype B3, D8, or A (vaccine) H and F proteins, and the cellular receptor hCD150 were sub-cloned into the mammalian expression vector pCAGGS. The identification codes of the different F proteins are: B3-F (GS63325-1), D8-F (GS77420-3), Moraten-F (GS55588-1), L454W-F (GS55608), and E455G-F (GS74446-1). A codon-optimized F protein sequence derived from the Measles virus strain Ichinose-B95a (Taxonomy ID: 645098) was synthesized by GeneScript and subcloned into the pcDNA3.1 vector. Stabilizing point mutations (E170G and E455G) previously characterized in SSPE and MIBE Measles strains^27,28,47 were sequentially introduced using the Q5 Site-Directed Mutagenesis kit. The ectodomain construct was generated by removing the transmembrane region and cytoplasmic tail residues (amino acids 496-550) using the Q5 Site-Directed Mutagenesis kit. To generate Drosophila S2 stable cell lines, F protein ectodomain construct E170G/E455G was subcloned from the pcDNA vector to the pMT-Puro vector, including a C-terminal enterokinase cleavage site-twin-strep-tag via NEB HiFi DNA assembly.

Antibodies

MeV F-specific murine monoclonal antibodies 77.4 and Y10F were generated as described previously⁴⁴. Chimeric antibody 77.1 has been functionally and structurally characterized²⁵. The remaining two antibodies, H8 and C6, were derived from murine hybridomas obtained after immunization with stabilized MeV F⁴¹, cell-based ELISA was used to detect binding to pre- and postfusion state of the F protein; their Fab regions were subsequently sequenced and cloned into a human Fc backbone, as described for 77.1²⁵

Cells

HEK293T (human kidney epithelial), Vero (African green monkey kidney), Vero-hCD150 (SLAMF1)⁴⁸ (kindly provided by Dr. Y. Yanagi) and Vero-hNectin-4⁴⁹ (kindly provided by Dr. C. Richardson) cells were grown in Dulbecco’s modified Eagle’s medium (DMEM; Life Technologies; Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS, Life Technologies; Thermo Fisher Scientific) and antibiotics at 37 °C in 5% CO₂. The Vero-hCD150 culture media were supplemented with 1 mg/mL Geneticin (Thermo Fisher Scientific). Drosophila S2 cells were cultured in complete Schneider’s Drosophila Medium (Gibco) supplemented with 10% heat-inactivated fetal bovine serum (FBS) and 1% penicillin-streptomycin (100 U/mL penicillin and 100 µg/mL streptomycin) at 25 °C without CO₂ or Lonza Insect Xpress Medium. ExpiCHO cells were maintained in ExpiCHO Expression Medium (Gibco) supplemented with 8 mM L-glutamine in a humidified incubator at 37 °C with 8% CO₂.

Viral rescue

HEK293T cells were transfected using Lipofectamine 2000 on 6-well plates coated with poly-D-lysine. The transfection mixture included purified plasmid constructs coding for viral proteins N (GS58900-1), P (GS58943), and L (GS58944), all sourced from Epoch Life Science according to our design specifications. Additionally, purified plasmids for T7 polymerase (GS58929) and the genome for the MeV B3 mCherry wt virus bearing either the B3 H and F (GS69677-10), or the D8 H and F (GS77420-2), or the H and F from the Moraten vaccine strain (GS77420-1), or the B3 H and F-L454W (GS69677-1) were included. Cells were incubated overnight at 37 °C in Opti-MEM medium. The medium was then replaced with warm DMEM supplemented with 10% FBS. After 1 h, cells were subjected to a heat shock treatment by incubating them at 42 °C in a water bath for 3 h. Subsequently, the cells were returned to a 37 °C incubator for 48 h. After this incubation, the HEK293T cells were detached and transferred to Vero-hCD150 cells in a 6-well plate and incubated at 32 °C for 5 days. RNA from the virus sample was then sequenced, and the virus stock was used for subsequent in vitro and in vivo experiments (see²⁵).

In vitro viral entry

Vero-hCD150 cells were plated in 96-well plates (1 × 10⁴ cells/well). The following day, cells were treated with serial dilutions of the indicated antibodies. Cells were then infected with the indicated viruses for 3 h at 37 °C (200–300 pfu/well). After 3 h, the medium was replaced with medium containing 2% of methylcellulose 2 × complete medium (1:1). After 24 h, fluorescent/infected cells were counted using a Cytation 5 (BioTek) and analyzed with Gen5 3.11 software (min object size 25 µm, max object size 135 µm, threshold 3000). The samples were tested in three separate experiments performed with technical duplicates in five-fold dilutions in 96-well plates. Data obtained from each plate was analyzed in OriginPro 2023, where all biological and technical replicates were fitted with a logistic regression function with shared IC₅₀ and Hill slope values, and A1 and A0 (highest and lowest signals) were used to normalize the data. Based on the normalized data, the percent of inhibition was calculated using the following equation: [1 - X]*100, where X is the fraction of the normalized signal.

Beta-Galactosidase (Beta-Gal) complementation-based fusion assay

The Beta-Gal complementation-based fusion assay was performed as previously described³¹. In brief, HEK293T cells transiently transfected with the hCD150 receptor and expressing the ω subunit of β-galactosidase were co-cultured with cells co-expressing the MeV glycoproteins F and H together with the α subunit of β-galactosidase. Cell fusion results in α-ω complementation, which was measured in the presence or absence of inhibitory compounds. After 18 h, cells were lysed to terminate the fusion process, and fusion levels were quantified using the Tropix Galacto-Star™ chemiluminescent reporter assay system (Applied Biosystems) and read on a Tecan M1000 microplate reader. Biological replicates were recorded and globally fitted with a logistic function in OriginPro 2023. The obtained parameters were used to normalize the data per sample and plotted with standard error calculated from all data points.

Cell surface staining with conformation-specific anti-F mAbs

HEK293T cells transiently transfected with viral glycoprotein constructs were incubated overnight at 37 °C in complete medium (DMEM supplemented with 10% FBS). The following day, cells were shifted to indicated temperatures and incubation times (as shown in the figures). To detect specific MeV F conformations, cells were incubated on ice for 1 h with chimeric monoclonal antibodies, followed by a 1 h incubation on ice with Alexa Fluor 594 or Alexa Fluor 488-conjugated anti-human secondary antibody (1:1000; Life Technologies). Cells were then fixed on ice with 4% paraformaldehyde for 10 min and counterstained with DAPI (1:1000; Thermo Fisher). Imaging was performed using a Cytation 5 system, and relative fluorescence of antibody-bound cells was quantified with BioTek Gen5 software.

Cell surface staining with conformation-specific anti-F C6 mAb: competition between H8 and Y10F mAbs

HEK293T cells were transiently transfected with viral glycoprotein constructs (either wt B3 or B3 E455G F) and incubated overnight at 37 °C in complete medium (DMEM supplemented with 10% FBS: CM). The following day, cells were brought to room temperature, and 50 µl of 2× Y10F (final concentrations: 200, 20, or 2 nM in CM) was added for 30 min. Subsequently, 50 µl of 2× H8 was added to achieve final concentrations of 100, 50, 25, 12.5, 6.25, 3.125, or 1.5625 nM in CM, resulting in diluted Y10F concentrations of 100, 10, and 1 nM, respectively. Cells were then transferred to 37 °C for 60 min. Following incubation, cells were placed at 4 °C, the supernatant was removed, and 50 µl of murine C6 mAb 1:500 dilution from an 8 mg/ml stock in CM) was added for 60 min. Cells were washed three times with CM and incubated on ice for 30 min with Alexa Fluor 488-conjugated anti-mouse secondary antibody (1:500; Life Technologies). Cells were then fixed on ice with 4% paraformaldehyde for 10 min and counterstained with DAPI (1:1000; Thermo Fisher). Imaging was performed using a Cytation 5 system, and C6-positive cells were quantified using BioTek Gen5 software.

Virus neutralization assay in cell lines

77.1, Y10F, and H8, were serially diluted two-fold, starting at 100 nM, and incubated with 60 TCID₅₀ rMeV-KS-Venus(3)⁵⁰ (kindly provided by Dr. W.P. Duprex) per well. Following a 1 h incubation at 37 °C with 5% CO₂, 1×10⁴ cells per well were added (Vero-hCD150, Vero-Nectin4, or Vero). Plates were incubated for 7 days and infection was monitored by Venus fluorescence. After 7 days, plates were imaged using an Amersham Typhoon scanner under predefined settings: channel Cy2 (GFP,488), pixel size 10 μm, scan speed normal, and PMT auto. The assay was performed in quadruplicate. In the Vero-hCD150 assays, wells were scored as positive or negative based on the presence or absence of Venus (endpoint titration), whereas in the Vero-hNectin4 assay the Venus positive surface area was calculated with Fiji software. Wells were scored positive if the infection covered >10%. For virus neutralization assay in B-LCLs. mAbs dilutions were prepared as described above and incubated with 10⁶ TCID₅₀ rMeV-KS-Venus(3) per well. Following a 1 h incubation at 37 °C with 5% CO₂, 1×10⁵ B-LCLs per well were added and incubated for 7 days. Plates were imaged using an Amersham Typhoon scanner under predefined settings: channel Cy2 (GFP,488), pixel size 10 μm, scan speed normal, and PMT auto. The assay was performed in quadruplicate, and wells were scored as positive or negative for infection based on the presence or absence of Venus (endpoint titration).

MeV dissemination assay in B-LCLs

A previously established MeV dissemination assay was used to assess the potential of the mAbs to prevent MeV dissemination⁵¹. Briefly, B-LCLs were infected with rMeV-KS-Venus(3) for 72 h. Infection rates in live cells were quantified by flow cytometry (BD FACS Lyric), based on Venus expression. Next, 150 MeV-infected B-LCLs were co-cultured with 6 × 10⁴ uninfected B-LCLs per well in the presence of a two-fold dilution series of mAbs starting at 100 nM. MeV dissemination was determined 48 h post co-culture by measuring the percentage of Venus-positive cells by flow cytometry.

Virus neutralization assay in Airway organoids at air-liquid interface (AO at ALI)

AO differentiated at ALI for >6 weeks, exhibiting ciliation and mucus production, were generated as previously described⁵². Prior to infection, AO at ALI were rinsed with Ca²⁺/Mg²⁺-containing DPBS to remove mucus and scratched in a cross-pattern to expose Nectin-4 and facilitate viral entry. Monoclonal antibodies (1000, 100, and 10 nM) were pre-incubated with rMeV-KS-Venus(3) (MOI = 0.1) for 1 h prior to application to the apical surface. After 2 h, the inoculum was removed to restore ALI conditions. Organoids were imaged using the Typhoon scanner at 3 days post-infection, and AO at ALI were semi-quantitatively scored for MeV infection.

Viral evolution assay

Selection of Measles Virus Escape Mutants in the Presence of Monoclonal Antibodies

Cell culture and virus infection

Phase 1: cell culture and virus infection

Vero-hCD150 cells were seeded in two 96-well plates and infected with 100 plaque-forming units (PFUs) per well of recombinant Measles virus (MeV) B3 wild-type expressing mCherry. Infections were carried out in Opti-MEM (Gibco, 31985070) supplemented with 10% penicillin-streptomycin. After 2 h, the infection medium was replaced with serially diluted concentrations of monoclonal antibodies Y10F and 77.1, starting at 10 μg/mL and diluted 1:2 down to 0.04 μg/mL.

Phase 2: passaging under antibody pressure

On day 10 post-infection, wells showing evidence of ongoing viral replication in the presence of highest (i.e., 10 μg/mL) antibody concentration (1-77.1: B1 and G1; 1-Y10F: F1 and E1) were selected. The contents of each well were transferred to T75 flasks containing Vero-hCD150 cells in Opti-MEM supplemented with 10% penicillin-streptomycin and 10 μg/mL of the corresponding antibody.

Due to limited infection after 3 days, additional wells from the antibody selection plates were chosen and expanded: wells 2-77.1 (A3, C3, E3) and 2-Y10F (B2, D2, F2) were transferred to T75 flasks and cultured in Opti-MEM with 10% pen-strep, containing reduced antibody concentrations (1.25 μg/mL for 77.1; 2.5 μg/mL for Y10F).

On day 15, the medium from the initial T75 flasks (i.e., 1-77.1: B1 and G1; 1-Y10F: F1 and E1) was replaced with Opti-MEM + 10% pen-strep without antibody. Virus supernatants were collected and frozen at −80 °C on day 16. Antibody concentrations in the second batch flasks were further reduced on day 17 due to a lack of visible infection (0.625 μg/mL for 77.1; 1.25 μg/mL for Y10F). Viral supernatants from these cultures were collected on days 23 and 28 (2-Y10F: B2, D2, F2; 2-77.1: A3, C3, E3), frozen, and subsequently used for further passages.

Sequential Antibody Reduction and Virus Harvesting

Phase 3: On day 28, all remaining samples were subsequently used to infect T75 flasks of Vero-SLAM cells at 10-fold lower antibody concentrations (0.125 μg/mL for 2-Y10F, 0.0625 μg/mL for 2-77.1, and 0.5 μg/mL for first-round isolates). Viruses were all harvested in the following 10-12 days.

Phase 4: On day 39, these eight virus stocks were then used to infect new T75 flasks of Vero-hCD150 cells. After three h, antibody was reintroduced at two-fold lower concentrations (0.0625 μg/mL for 2-Y10F, 0.0313 μg/mL for 2-77.1, and 0.25 μg/mL for 1-Y10F and 1-77.1).

Phase 5: On day 54, each virus was collected, titered, and sequenced.

In vivo experiments: cotton rats

Inbred cotton rats (Sigmodon hispidus) were purchased from Envigo, Inc., Indianapolis. Both male and female cotton rats aged 5 to 7 weeks were used. For i.n. infection, 10⁵ TCID₅₀ of MeV (strain B3 wt EGFP) was inoculated intranasally to isoflurane-anesthetized cotton rats in a volume of 100 µL. Animals received either the indicated amount of the indicated antibodies, purified human IgG (Carimune, CSL Behring, measles neutralizing titer of 640 at 60 mg/mL), or vehicle 14 h prior to infection or 24 h post-infection. Four days after infection, the animals were euthanized by CO₂ inhalation, and their lungs were collected and weighed. Lung tissue was homogenized with a glass Dounce homogenizer. Serial 10-fold dilutions of supernatant fluids were assessed on Vero-hCD150 cells for the presence of infectious virus in 48-well plates using cytopathic effect (CPE) as the endpoint. Plates were scored for CPE microscopically after 7 days, and the tissue culture infectious dose 50 (TCID₅₀) was calculated. The results are presented as TCID₅₀/g of lung tissue.

RBC fusion assay

RBC fusion assays were performed using HEK293T cells transiently expressing the MeV H_Y17H-HPIV3_T193A chimera and MeV F (T461I). Cell monolayers were washed three times with serum-free medium and incubated at 4 °C for 30 min in CO₂-independent medium containing 1% RBCs. Cells were then exposed to varying concentrations of inhibitors and shifted to 37 °C for 60 min.

To quantify RBC binding and fusion, zanamivir was added at a final concentration of 10 mM, and the cells were incubated at 4 °C for 30 min with gentle agitation. The supernatant was collected into V-bottom 96-well plates to measure reversibly bound RBCs. The cells were subsequently incubated at 4 °C for 30 min in ACK lysing buffer, and the supernatant was collected into V-bottom 96-well plates to quantify irreversibly bound RBCs. Finally, the remaining cells were lysed in lysis buffer and transferred to flat-bottom 96-well plates to determine fused RBCs. Absorbance at 410 nm was measured using a Tecan My1000PRO microplate reader to quantify RBCs in each category. Assays were performed in CO₂-independent media at pH 7.2 for the bar graphs in Fig. 6B and Supplementary Fig. 9, and at pH 7.6 for the dose response curves in Fig. 6C.

F_ECTO purification

F_ECTO was purified as previously described from Drosophila S2 cells²⁵. Protein concentration was estimated using the molar extinction coefficient at 280 nm (48,820 M⁻¹ cm⁻¹).

Antibody expression

Monoclonal antibodies were produced in ExpiCHO cells using the manufacturer’s high-titer transfection protocol with pTRIOZ-mAb plasmids encoding the antibody heavy and light chains. After seven days, cells were removed by centrifugation at 4000 × g, and the clarified medium was adjusted to pH 8.0 with 1 M HEPES (pH 8.0). Antibodies were captured by incubating the medium overnight at 4 °C with protein A agarose. The next day, beads were washed with 20 mL HBS, and bound antibody was eluted with 15 mL 0.2 M glycine (pH 2.0). The eluate was immediately neutralized by adding 10% (v/v) 1 M HEPES (pH 8.0).

For the Fab generation, purified IgG at 1 mg/mL was digested with 2% (w/v) cysteine-activated papain for 4 h at 37 °C in a water bath. The reaction was quenched by adding 0.5 M iodoacetamide to a final concentration of 50 mM. The mixture was diluted tenfold with 10 mM HEPES (pH 8.0) to reduce ionic strength and loaded onto a 1 mL Mono S column pre-equilibrated with 20 mM HEPES (pH 8.0). Fab fragments were collected in the flow-through, whereas Fc fragments were eluted in a single step with 20 mM HEPES (pH 8.0) containing 2 M NaCl. Isolated Fabs were concentrated, buffer-exchanged into HBS, and stored at 4 °C. Protein concentrations were determined from A280 using the calculated molar extinction coefficients.

Surface plasmon resonance

Binding experiments were performed on a Carterra LSA instrument. Purified monoclonal antibodies (mAbs) were covalently immobilized onto an HC30M sensor chip at a concentration of 2.5 µg mL⁻¹, with each antibody injected in duplicate. A serial dilution of the analyte (F_ECTO) was applied to the chip in ascending order: 3.125 nM, 6.25 nM, 12.5 nM, 25 nM, 50 nM, and 100 nM. The resulting binding responses were recorded over time. Kinetic parameters were extracted by globally fitting the sensorgrams to Eqs. 1–6 using OriginPro software.

$$\mathrm{Observed}\mathrm{rate}\mathrm{and}\mathrm{equilibrium}{\mathrm{k}}_{\mathrm{obs}}={\mathrm{k}}_{\mathrm{a}}\mathrm{C}+{\mathrm{k}}_{\mathrm{d}}{R}_{eq}=\frac{k_a\mathrm{C}}{k_{a}C+k_d}{R}_{\max }$$ 1

$$\mathrm{Association}{S}_a\left(t\right)=R_{eq}+\left(S_0-R_{eq}\right)e^{-k_{obs}\left(t-t_s\right)}$$ 2

$$\mathrm{Value}\mathrm{at}\mathrm{the}\mathrm{switch}\mathrm{S}{@}_{{\mathrm{t}}_{\mathrm{d}}}={\mathrm{R}}_{\mathrm{eq}}+\left({\mathrm{S}}_0-{\mathrm{R}}_{\mathrm{eq}}\right){\mathrm{e}}^{-k_{obs}\left({\mathrm{t}}_{\mathrm{d}}-{\mathrm{t}}_{\mathrm{s}}\right)}$$ 3

$$\mathrm{Dissociation}{S}_d\left(t\right)S{=}_{t_d}{e}^{-k_d\left(t-t_d\right)}$$ 4

$$\mathrm{Baseline}B\left(t\right)=b+m\left(t-t_s\right)$$ 5

$$\mathrm{Final\; fit}R\left(t\right)=\left\{\begin{array}{c}\hfill B\left(t\right)+S_a\left(t\right),t<t_d\hfill \\ \hfill B\left(t\right)+S_d\left(t\right),t>t_d\hfill \end{array}\right)$$ 6

Where t time (s). t_s association start (s). t_d dissociation start (s), C analyte concentration (M), k_a association rate constant (M⁻¹ s⁻¹), k_d: dissociation rate constant (s₋₁), k_obs: observed association rate (s⁻¹), R_max: maximum response at full occupancy (response units), R_eq: equilibrium response at C (response units), S₀: residual bound signal at t_s (response units), S_@td: signal at the switch to buffer (response units), b: baseline offset (response units), m: linear drift (response units s⁻¹)

Time-resolved size-exclusion chromatography

F_ECTO was mixed with H8 Fab at a 1:1 molar ratio (monomer: Fab). The mixture was incubated at 25 °C for 15, 90, 180, or 1440 minutes. At each time point, 100 μL of the reaction was injected onto a Cytiva Superdex 200 Increase 10/300 GL column. The column was pre-equilibrated with HBS buffer containing 20 mM HEPES (pH 8.0) and 150 mM NaCl. Absorbance at 280 nm was recorded throughout the elution. Chromatograms were exported to OriginPro 2025 and baseline-corrected using the spline method. Corrected traces were used to calculate the trapezoidal area between 7 and 10 mL elution volume in Python. This region includes the column void volume and contains F_ECTO-H8 species associated with triggering. The resulting area values were plotted as a function of incubation time. Data were fitted with a single-exponential function in OriginPro 2025 to estimate kinetic parameters.

Time-resolved negative-stain electron microscopy

Complexes for negative-stain analysis were prepared in the same way as for size-exclusion chromatography. Aliquots were removed from the reaction after 0, 15, 60, 150, 240, and 1140 minutes. Each aliquot was immediately applied to a grid and processed for negative staining. For every time point, approximately 176 movies were collected and processed in cryoSPARC version 4.7.1. Patch motion correction and CTF estimation were performed using default settings. Particles were selected using the blob picker with a diameter range of 120-250 Å. A single round of two-dimensional classification was run with 100 classes and 500 particles per class. The protocol consisted of 30 online iterations, followed by two full iterations. Class occupancy was calculated from the number of particles assigned to each selected class. Class 5 was excluded from the final analysis because it contained heterogeneous particles such as H8 Fab bound to F fragments, Fab alone, and other mixed species. The final class occupancy values were plotted over time using Python, Matplotlib, and Seaborn.

Grid preparation

Negative stain

Negative-stain electron microscopy was performed using 400-mesh copper grids coated with a continuous carbon film (CF400-Cu; Electron Microscopy Sciences). Prior to sample application, the grids were rendered hydrophilic by glow-discharging them in an easyGlow Glow Discharge Cleaning System (Pelco) for 15 s at 15 mA. Protein samples were diluted to an absorbance of 0.01 at 280 nm; 3 µL of this solution was then deposited onto the grid and incubated for 1 minute. The excess sample was removed by blotting, and the grid was washed three times with 3 μL of deionized water, with blotting performed between each wash. Negative staining was performed by applying 3 µL of a 1% uranyl acetate solution to the grid, blotting twice, then adding an additional drop of stain and incubating for one minute before a final blot to remove excess stain. The grids were allowed to air-dry prior to imaging.

Cryo-EM sample preparation and data acquisition

Complex stoichiometries were guided by NS-EM. Except where noted, complexes were incubated with a twofold molar excess of Fab relative to the number of F binding sites. Accordingly, samples were prepared at trimer:Fab ratios of 1:2 for F_ECTO:Y10F (one F trimer to two Y10F Fabs) and 1:2:6 for F_ECTO:Y10F:77.1 (twofold excess of 77.1). For H8, Fab was added at a 1:1 ratio per F binding site (1:3 trimer:H8), because this complex was not purified before imaging.

F_ECTO:Y10F and F_ECTO:Y10F:77.1 complexes were purified by size-exclusion chromatography (SEC) before grid preparation. The F:H8 complex was applied directly to grids after a 15-minute incubation to minimize postfusion conformational transition that could occur during SEC. Data were collected on graphene oxide (GO) grids as described previously²⁵. All F_ECTO-Fab complexes were recorded on Quantifoil R2/1 or R2/2 grids covered with GO grids prepared as described previously²⁵. Protein with 0.25-35 absorbance values at 280 nm was placed on grids and blotted for 2-3 s. Grids were loaded to the FEI Titan Krios G3 autoloader and screened for ice quality, graphene oxide layer quality, and particle distribution. Movies were recorded with an exposure rate of 70 e/Ų, with 50 frames per movie saved as non-gain normalized LWZ TIFF files. Collection statistics per dataset are reported in Table S1.

Cryo-EM data processing

All datasets were processed in cryoSPARC v4.5 or v4.6 using a standard workflow with dataset-specific adjustments during late refinement. Movies were corrected for beam-induced motion with Patch Motion, and defocus values were estimated with Patch CTF in cryoSPARC Live. Particles were initially picked with the blob picker using an intentionally broad particle diameter (80-120 Å) to over-pick F_ECTO-Fab complexes. Particles were first extracted and Fourier-cropped to 64 pixels to speed up early processing.

Particle stacks were extensively cleaned by 2D classification following the Deep2D protocol described previously²⁵, and class averages showing well-defined protein features were retained. Two independent ab initio reconstructions were run per dataset, one imposing C1 and one C3 symmetry, each with five classes; for the Y10F dataset, only C1 was used. The best ab initio models were then used for heterogeneous refinement against all classes, iterating until discrete 3D classes emerged. The best-looking class or classes were pooled and subjected to non-uniform refinement (NUR), after which particles were re-extracted with larger Fourier-cropped boxes (160-256 pixels) for further NUR and additional 3D classification to improve compositional and conformational homogeneity.

For final map generation, only particles contributing to the best-resolved class were retained and re-extracted with a Fourier crop chosen so that the 0.143 FSC resolution fell within approximately three-quarters of the box frequency (to avoid box-size-limited resolution). Final refinements included NUR, symmetry expansion where appropriate, focused/local refinement with masks, and targeted 3D classification, followed by a final round of local refinement. The final particle set was subjected to reference-based motion correction and both global and local CTF refinement, after which a final local refinement produced the maps reported.

Model building

Model building proceeded by rigid-body fitting of the previously reported F_ECTO structure ²⁵ (PDB 8UUP) into each cryo-EM map. Antibody Fab fragments were modeled with AlphaFold2 and docked into the density using UCSF ChimeraX^53,54. The combined F_ECTO-antibody models were then iteratively adjusted manually in Coot⁵⁵, followed by cycles of real-space refinement in PHENIX⁵⁶ and additional manual corrections in Coot.

Interface interactions determination

Interface contacts, buried surface area, and estimated binding energy were calculated with PISA (CCP4 suite;⁵⁷). Interaction counts are reported as unique contacts from all symmetry-related interactions.

Formation of F_ECTO-Fab Complexes for Determining Hydrodynamic Radius and Differential Scanning Fluorimetry

F_ECTO (F_ECTO E170G, E455G, residues 1-495) was mixed to a final concentration of ~1.7 µM (trimer) and incubated with Fab at room temperature for 15-30 min. Fab was added at a threefold molar excess for 77.1, H8, and C6, and at a onefold molar excess for Y10F (ratios calculated relative to the F_ECTO trimer). The resulting mixtures were used directly for downstream analyses.

nanoDSF and DLS measurements

Thermal unfolding of F_ECTO with or without Fab, and of Fab alone, was measured on a NanoTemper Prometheus Panta. F_ECTO-Fab complexes were assembled 30 min before acquisition and incubated at room temperature. For each experiment, three capillaries contained the F_ECTO-Fab complex, two capillaries contained the corresponding Fab alone, and one capillary contained HBS as a buffer control. Data were collected with a 1 °C/min temperature ramp and 100% excitation laser power. Unfolding profiles are reported as the first derivative of the intrinsic fluorescence at 330 nm. Dynamic light scattering was recorded in parallel, and the cumulant hydrodynamic radius was calculated and plotted as a function of temperature.

Next generation sequencing and variant calling

Viral whole-genome sequencing of MeV samples was performed using a metagenomic next-generation sequencing approach and a previously described protocol⁵⁸. Libraries were sequenced on NextSeq 2000 with 2x150bp or 1x100bp read format. Raw reads were trimmed and quality filtered with fastp (v0.23.4)⁵⁸ with the parameters --cut_mean_quality 20 --cut_front --cut_tail --length_required 20 --low_complexity_filter --trim_poly_g --trim_poly_x. For variant calling, filtered reads were processed with the RAVA workflow (default parameters) using the MeV plasmid map as a reference (https://github.com/greninger-lab/RAVA_Pipeline/tree/2025-08-18_DZ_MP)⁵⁹, (10.1101/2019.12.17.879320). Raw sequencing data are publicly available in NCBI BioProject PRJNA1307578.

Statistics & reproducibility

Data analysis and visualization were performed in Python using NumPy, pandas, SciPy, Matplotlib, seaborn, and statannotations. Unless stated otherwise, group comparisons used two-sided Mann-Whitney-Wilcoxon tests. When multiple pairwise comparisons were made within a figure panel, p-values were adjusted with the Holm-Bonferroni method to control the family-wise error rate. Exact p values and test statistics (U) are reported in the figure legends. P-value annotations on plots follow this legend: ns, 5.00 × 10⁻² <p ≤ 1.00 × 10⁰; *, 1.00×10⁻² <p ≤ 5.00 × 10⁻²; **, 1.00 × 10⁻³ <p ≤ 1.00 × 10⁻²; ***, 1.00 × 10⁻⁴ <p ≤ 1.00 × 10⁻³; ****, p ≤ 1.00 × 10⁻⁴.

In animal experiments, individual cotton rats represented distinct biological replicates (per group). This sample size was selected based on historical data from this infection model to achieve adequate statistical power for a non-parametric analysis (Mann-Whitney U test) without exceeding the necessary number of animals.

Chemicals and peptides

N-(3-cyanophenyl)-2-phenylacetamide (also known as 3 G) was commercially acquired from ZereneX Molecular Limited (UK). The purity of 3 G was tested by high-pressure liquid chromatography (HPLC) and shown to be >95% pure.

Peptides were purchased from CPC, Ltd. Bromoacetyl cholesterol and bismaleimide cholesterol derivatives were custom-made by Charnwood Molecular Ltd. Dimethylsulfoxide (DMSO), tetrahydrofuran (THF), N,N-diisopropylethylamine (DIPEA) were purchased from Sigma Aldrich, bismaleimide-PEG₁₁ from Quanta BioDesign. HPLC purification was performed on a 1100 Series Agilent HPLC system equipped with a UV diode array detector and a fraction collector using a reverse phase (RP.) Phenomenex Jupiter C4 LC column 300 Å (150 × 21.2 mm, particle size 5 μm). MALDI-TOF analysis was performed on a Bruker UltrafleXtreme MALDI-TOF instrument.

Reporting summary

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

Author contributions

D.S.Z.: Cloned and expressed proteins; recorded all negative stain and cryo-EM data; processed and solved structures; wrote the code for analysis; analyzed all data presented in the manuscript; prepared figures; wrote the manuscript. R.D.M.: Performed virological and cell biology experimentation; monoclonal antibodies development; analyzed data; prepared figures; contributed to manuscript writing. D.L.: Performed cell-based and virological experiments; analyzed the data; prepared figures. G.N.: Assisted with screening, protein purification, and negative stain electron microscopy (NS-EM). G.Z.: Performed virological and cell biology experimentation; analyzed data. L.D.C.: Rescued recombinant viruses; performed virological experiments; analyzed data. M.H.L.V.: performed in vitro and ex vivo experiments live virus experiments. L.L.A.V.D.: performed in vitro and ex vivo live virus experiments. O.M.C.: time-resolved size exclusion and protein expression. G.J.T.: monoclonal antibodies development. G.K.: hybridoma production and monoclonal development. M.A.: Performed monoclonal antibodies binding analysis. G.L.: Performed virological studies; analyzed data. D.V.: sequencing of viral population; data analysis. C.L.: in vivo experiments. E.P.: in vitro and in vivo experiments. K.M.H.: Analyzed data; contributed to manuscript writing and figure preparation. B.H.: Monoclonal antibodies development; analyzed data. R.L.D.S.: supervision, data analysis and manuscript writing. R.D.V.: supervision, data analysis, and manuscript writing. A.L.G.: data analysis and manuscript writing. S.N.: Conducted in vivo experimentation, data analysis, and interpretation; supervised the work; contributed to manuscript writing. E.O.S.: Analyzed and interpreted results; supervised the work; secured funding; contributed to manuscript writing. M.P.: Designed experiments; monoclonal antibodies development; performed virological and cell biological experimentation; analyzed and interpreted results; supervised the work; secured funding; wrote the manuscript.

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

Cryo-electron microscopy density maps and atomic coordinates have been deposited in public repositories. The Y10F-F ectodomain complex is available in the PDB (9Q5M) and EMDB (EMD-72240); the 77.1-Y10F-F ectodomain complex is available in the PDB (9Q5V) and EMDB (EMD-72242); the H8-F ectodomain complex in the prefusion state is available in the PDB (9Q7B) and EMDB (EMD-72293); and the H8-F ectodomain complex in the postfusion state is available in the PDB (9Q71) and EMDB (EMD-72283). Source data are provided with this paper.

Competing interests

B.H., M.P., D.S.Z., and E.O.S. have a provisional patent related to the antibodies as passive immunotherapy. All the authors declare that they have no competing interests in relation to the work described in this manuscript.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-71373-4.