Overall structures of detergent solubilized VSV G in the pre- and post-fusion states

To gain insight into the full-length structure of VSV G, i.e., the ectodomain plus the TMD and ID (Fig 1A), we developed a protocol to isolate VSV G (Indiana strain) from concentrated virus particle preparations without any affinity purification tag. Briefly, VSV G was extracted directly from viral membranes using lauryl maltose neopentyl glycol (LMNG) and subsequently purified to homogeneity through successive chromatography steps (S1 Fig). We then performed cryo-EM single-particle analysis (SPA) on VSV G incubated at pH 8.0 (VSV G pH 8.0) and at pH 5.5 (VSV G pH 5.5) (S2, S3 Figs, S1 Table).

2D class averages were generated for both VSV G pH 8.0 and VSV G pH 5.5. Analysis of the 2D classes revealed a preferential orientation of the molecule in both conditions, with approximately 80% of the particle exhibiting top views (S2-S3 Figs). Using automated particle picking to enrich the data set with sides views, data processing yielded reconstructions with an overall resolution of 2.9 Å for both the high- and low-pH structures (S1B- S1C Fig).

The overall structure of the ectodomain in both reconstructions was consistent with previous studies, showing a trimeric spike with dimensions in height of 8 nm at pH 8.0 and 12 nm at pH 5.5, corresponding to the pre- and post-fusion conformations [8,31], respectively (Fig 1B-1C) and associated with a detergent micelle at the FD tips. Notably, the presence in the sample of TMD and ID did not appear to influence the global structure of VSV G ectodomain (Fig 1B-1C). Indeed, structural overlay of the crystallographic and cryo-EM derived models yielded root-mean-square deviations (RMSD) for the Cα backbone of 0.78 Å for the pre-fusion structure and 0.69 Å for the post-fusion structure (PDB code 6TIT [15] and 5I2M [16] respectively) (S2 Table). Additionally, there was a very clear density for the starting residues of the two N-linked glycosylation chains (attached to N163 and N320) [32] in both cryo-EM structures (Fig 1B-1C). It is worth noting that comparison with the crystallographic structure of the post-fusion state of VSV G [16] revealed an additional density corresponding to part of the CTD in the ectodomain of the post-fusion cryo-EM structure (Fig 1C, in magenta).

Despite repeated attempts to visualize the detergent micelle content using several computational cryo-EM methods (such as 3D-classification, 3DVA, 3DFLEX and several masking strategies), TMDs (resp. the IDs) were not discernible within (resp. in the vicinity of) the micelle (Fig 1B-1C) in both conformations, although SDS-PAGE did not reveal any proteolysis (S1 Fig).

On the other hand, the tips of FDs were visible inside the micelles, suggesting that the absence of detection of TMDs in the density is due to the fact that TMDs are mobile and do not have a fixed position within the micelle. Measurements indicate that the penetration depth of FDs into the micelle is approximately 9 Å for both conformations (Fig 1D-1E). This is consistent with the organization of the micelle of LMNG having a hydrophobic core radius of 14 Å and a hydrophilic shell thickness of 7 Å [33]. This allows the accommodation of the hydrophobic residues in the fusion loop in the hydrophobic core as well as several polar or charged residues in the hydrophilic shell (Fig 1D-1E).

Reorganization of the C-terminal linker region

The resolution of the CTD in the post-fusion conformation is approximately 4.5 Å, allowing for its partial tracing up to residue 426 (Fig 2A). For each protomer, the traced segment extends into a predominantly hydrophilic groove (Fig 2B) located between this protomer (protomer A) and the neighboring one (protomer B). Structural analysis reveals a hydrophobic patch on the CTD, composed of residues L423 and F425, which interact closely with residue I82 located on the FD of the same protomer (protomer A), as well as residues L105 and P107 of the FD of the neighboring protomer (protomer B) (Fig 2C and S4 Fig). These interactions, which are mainly hydrophobic with very few polar contacts (S4 Fig), stabilizes the C-terminal domain against the FD. Sequence alignment of Vesiculovirus glycoproteins indicates that the hydrophobic nature of these residues is conserved among the genus (S5 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. VSV G C-terminal domain (CTD) rearrangement.

(A) Ribbon representation of VSV G trimer in its post-fusion conformation, fitted and traced up to residue 426 in the cryo-EM map obtained at pH 5.5. The same color code as in Fig 1A is used. (B) Surface hydrophobicity representation of the interaction between segment 410-426 of chain A (right) and neighboring FD (left) in the post-fusion conformation. The hydrophobicity scale is indicated. (C) Hydrophobic patch on the CTD-FD interface in the post-fusion conformation; Left panel: Overall view of VSV G trimer showing FD from protomer A (in yellow) and FD from protomer B (in green) with the CTD from protomer A in magenta. Residues displayed as sticks contribute to the stabilizing of the CTD-FD interface in the post-fusion state. Right panel: close-up on the hydrophobic patch formed at the level of the CTD (involving L423 and F425) interacting with I82 from the FD of protomer A (in yellow), and L105 and P107 from the FD of protomer B (in green). These residues stabilize the CTD-FD interface in the post-fusion conformation. Putative hydrophobic interactions are indicated by red dashed lines. (D) Ribbon representation of VSV G trimer in its pre-fusion conformation, fitted and traced up to residue 427 in the cryo-EM map obtained at pH 8.0. The same color code as in Fig 1A is used. (E) Close-up view of VSV G CTD in the pre-fusion conformation. The inset shows the folding of the CTD and a detailed view of the environment surrounding I82, L423 and F425 in VSV G pre-fusion conformation. Putative hydrophobic interactions are indicated by red dashed lines.

https://doi.org/10.1371/journal.ppat.1013589.g002

In the pre-fusion conformation, the overall structure of the ectodomain was similar to previously determined crystal structures [15], with the CTD modeled up to residue 427 (Fig 2D-2E). In both X-ray crystallography and cryo-EM pre-fusion structures, the CTD forms a small helix (res 409 to 416) followed by a β-turn between P418 and E421 and a short β-hairpin (res 422 to 432, only visible up to res 427 in the cryo-EM structure) forming a hook-like structure (Fig 2E). This indicates that the presence (or absence) of the TMD of VSV G does not interfere with its fold. The C-terminal part of VSV G is further stabilized by the insertion of F109 into a hydrophobic pocket composed of residues I82, A413, L417, and L423, which position the CTD in close proximity to the FD. Notably, as in the post-fusion conformation, it is also the hydrophobic residues L423 and F425 on the CTD that form a hydrophobic cluster with I82 on the FD (Fig 2E). Thus, this hydrophobic cluster is present in both the pre- and post-fusion states and, in both cases, it locks the CTD in close proximity to the FD (Fig 2C and 2D-2E). However, the organization of the CTD is completely different between the two structures although mainly involving hydrophobic interactions (Figs 2C, 2D-2E and S4). Thus, the CTD, as so R1, R2, R3, R4 and R5 segments, refolds during the low-pH induced structural transition of VSV G. In fact, the CTD and R5 are part of a single long segment that refolds during G conformational change.

Neutralization of VSV by mAb 8G5F11

The commercial monoclonal antibody 8G5F11 [29] neutralizes VSV and other Vesiculoviruses [30]. To characterize the interaction of VSV G with mAb 8G5F11, we incubated purified mAb at various pH values with magnetic beads coated with protein A, followed by the addition of purified full-length VSV G pre-incubated at pH values between 5.0 and 8.0. After an incubation of 30 minutes, beads were washed, and the associated proteins were analyzed by SDS-PAGE. These interaction experiments demonstrated that 8G5F11 is able to bind VSV G over a range of pH values from 8.0 to 5.0 (Fig 3A). This suggested that 8G5F11 recognizes several VSV G conformations.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. mAb 8G5F11 interacts with the pre- and the post-fusion conformations of VSV G.

(A) Coomassie-stained SDS-PAGE analysis (under non-reducing conditions) of interaction experiments between VSV G and mAb 8G5F11 bound to protein A coated magnetic bead, incubated at various pHs. (B) BLI sensograms showing the binding kinetics of VSV G_ect to mAb 8G5F11 at pH 8.0. Experimental curves (black) were fitted (red) using a 1:1 binding model. The calculated KD is indicated on the panel. (C) Coomassie-stained SDS-PAGE analysis (under reducing conditions) of interaction experiment between VSV G and Fab 8G5F11 bound to strep-Tactin coated magnetic beads, incubated at various pH values. (D) BLI sensograms showing the binding kinetics of VSV G_ect to 8G5F11 Fab at pH 8.0. Experimental curves were fitted using a 1:1 binding model. The calculated KD is indicated on the panel. (E) Neutralization curve of VSVe-GFP by mAb (dark blue) and Fab (light blue). VSV-eGFP was preincubated with increasing concentrations of mAb or Fab. At 5 hours p.i., the percentage of infected cells determined by counting the number of cells expressing eGFP using flow cytometry. This was used to calculate the infectious viral titer. Data depict the mean with standard error from experiments performed in triplicate. Average IC₅₀ value are indicated.

https://doi.org/10.1371/journal.ppat.1013589.g003

Additionally, we employed bio-layer interferometry (BLI) to investigate the binding parameters of coated mAb 8G5F11 to VSV G_ect at pH 8.0. For this, we used the thermolysin-generated VSV G_ect (aa residues 1–422) that is monomeric in solution [8]. Analysis of the BLI data revealed that mAb 8G5F11 exhibited a dissociation constant in the nanomolar range for VSV G_ect at pH 8.0 (Fig 3B, S3 Table).

We engineered the corresponding Fab fragment of mAb 8G5F11. For that, variable heavy (VH) and light (VL) chains of mAb 8G5F11 were sequenced and cloned into an insect cell expression vector (S6 Fig). The construction resulted in a Fab C-terminally fused to a strep-TagII. After establishment of the stable S2 cell line, recombinant Fab was purified from cell culture media by affinity chromatography followed by SEC (S6 Fig). Interaction experiments conducted under the same conditions as for the mAb 8G5F11 (except that the magnetic beads were coated with strep-tactin) show that Fab 8G5F11 binds VSV G across the same pH range, with BLI measurements also indicating a binding affinity in the nanomolar range (Fig 3C-3D).

To determine the IC₅₀ values of mAb 8G5F11 and its corresponding Fab, we performed neutralization experiments by pre-incubating a recombinant VSV that express eGFP (VSV-eGFP) with serial dilutions of mAb or Fab. After 30 minutes of incubation, the inoculum was transferred onto HEK-293T cells. After 5 hours of infection, the percentage of infected cells (expressing eGFP) was measured by flow cytometry. Both mAb 8G5F11 and Fab 8G5F11 inhibit VSV infection in a dose-dependent manner. Calculated viral titers were plotted against the concentration of mAb or Fab and the dose-response curve was fitted using the Hill equation (Fig 3E). The estimated IC₅₀ value for the Fab 8G5F11 was approximately 40 nM, whereas it decreased to 0.6 nM when neutralization was performed with the mAb 8G5F11. Fab 8G5F11 neutralization curve exhibited a Hill coefficient of 1.2 ± 0.17 whereas mAb neutralization curve had a Hill coefficient of 3.35 ± 0.79 indicating some cooperativity effect (Fig 3E). This is consistent with the divalent nature of the mAb, which induces avidity-enhanced binding to VSV G, as well as with its larger size, which increases steric hindrance, potentially decreasing the number of mAbs required to neutralize a virion compared to that of Fabs.

Structural characterization of Fab 8G5F11 bond to VSV G

To better understand how mAb 8G5F11 binds VSV G at different pH values, we performed cryo-EM SPA on VSV G in complex with Fab 8G5F11 at pH 8.0 and pH 5.5 (S6-S8 Figs, S1 Table). The cryo-EM structure of VSV G in complex with the Fab revealed a stoichiometric binding of one Fab per protomer at both pHs (Fig 4A and 4C). At pH 8.0, the Fab binds each VSV G protomer at the top of the molecule (Fig 4A-4B) with an angle of approximately 45° with respect to the axis 3 of the trimer (Fig 4B). At pH 5.5 the binding site is the same but the relative orientation of the Fab is different with an angle of 135° (Fig 4D).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Structural basis for the neutralization of VSV by mAb 8G5F11.

(A) Cryo-EM density map of VSV G in complex with Fab 8G5F11 at pH 8.0 shown in side view orientation. One 8G5F11 Fab molecule was observed to bind to each protomer of VSV G. G ectodomain is in sky-blue, the LMNG micelle in grey, the starting residues of the glycosylation chains are in orange. Fusion loops are in yellow and CTD in magenta. Fab variable heavy (VH) and variable light (VL) chains are in green and in cyan respectively. (B) Ribbon diagram of VSV G trimer in the pre-fusion conformation in complex with Fab 8G5F11, viewed from the side (left panel) and from the top (right panel). The VSV G model was fitted in the density and traced up to residue 426. The VH and VL chains of the 8G5F11 Fab were modelized and fitted in the identified density. The same color code as in Fig 1A is used for VSV G. Fab VH chain is in green and VL chain is in cyan. (C) Cryo-EM density map of VSV G in complex with Fab 8G5F11 at pH 5.5, shown in side orientation. The same color code as in Fig 4A is used. (D) Ribbon diagram of VSV G post-fusion trimer in complex with Fab 8G5F11, viewed from the side (left panel) and from the top (right panel). As in the pre-fusion state, one molecule of Fab binds each VSV G protomer of the post fusion trimer. The VSV G model was fitted in the density and traced up to residue 426, while the VH and VL chains of the 8G5F11 Fab were fitted in the identified density. The same code for G as in Fig 1A is used. Fab VH chain is in green and VL chain is in cyan. (E) Interaction between 8G5F11 and VSV G. Key residues are depicted in stick representation and are highlighted in the boxed panel. For clarity only, labels of key residues on VSV G are indicated. (F) Close up on the residues involved in the antibody-antigen interaction at high pH. VSV G PHD is in orange, and the key residues interacting with 8G5F11 are depicted in sticks.

https://doi.org/10.1371/journal.ppat.1013589.g004

The binding of the Fab did not influence the structure of the ectodomain; at pH 8.0 VSV G is in its pre-fusion conformation and at pH 5.5 VSV G is in its post-fusion conformation. The complexes between G and the Fab did not allow us to determine the structure of new parts of G.

In both structures, the interface between the Fab and VSV G is the same and principally involves the variable heavy chain (VH) of the Fab binding to the VSV G PHD (Fig 4B, 4D–4F). The interaction is mainly hydrophilic. Three VSV G residues (D241, K242, and D243) are located at the core of the interface between G and 8G5F11 paratope (Fig 4E). The carboxylic group of D241 establishes polar contact via its side chain with residue H53 of the Fab (Fig 4F). The side chain of D243 establishes a cluster of polar contacts with the hydroxyl groups of Y94 (on the VH chain) and Y107 (on the VL chain) as well as with the carbonyl of the main chain of G100 (on the VH chain) (Fig 4F). VSV G K242 ε-amino group interacts with the hydroxyl group of Y94 carboxyl group of D32 and the main chain of Y91 (on the VL chain) with the hydroxyl group of Y107 (on the VH chain) (Fig 4F). Two additional residues on VSV G (Q42 and R249) located outside of the core also contribute to the interaction. The amide group of Q42 side chain establishes hydrogen bonds with the carbonyl of the main chain of G103 and the hydroxyl group of Y101 on the VH chain of the Fab. Finally, the guanidinium group of R249 on VSV G establishes H-bonding with N59 side chain and the T58 main chain (both on the VH chain of the Fab) (Fig 4F).

Characterization of key residues for 8G5F11 binding

To investigate the contribution of residues constituting the mAb 8G5F11 binding site on VSV G (Fig 4E), we generated alanine mutants at positions Q42, D241, K242, D243, R249 of G. HEK-293T cells were transfected with plasmids encoding either wild type (WT) VSV G or VSV G alanine mutants (G_mut: Q42A, D241A, K242A, D243A or R249A). Surface expression of the VSV G mutants was confirmed using a conformational probe, GST-CR3 (glutathione S-transferase fusion with the CR3 domain of the LDL-R [2]), labeled with a fluorescent dye ATTO⁵⁵⁰ (Fig 5A). Indeed, GST-CR3 binds VSV G in its pre-fusion conformation and its binding site [2,15] is distinct from the binding site of mAb 8G5F11 (S9 Fig). Flow cytometry analysis confirmed that all mutants were correctly expressed at the cell surface, as evidenced by their efficient binding to the CR3-GST conformational probe (Fig 5B). Then, we assessed the ability of VSV G mutants to bind mAb 8G5F11 and its corresponding Fab on the cell surface. Mutations D241A and D243A induced a significant reduction in binding for both the mAb and Fab 8G5F11 (Fig 5C). Afterwards, we evaluated whether these mutant glycoproteins (VSV G_D241A and VSV G_D243A) could maintain viral infection. We employed a recombinant VSV (VSVΔG/G) lacking the G envelope gene, replaced by the green fluorescent protein (GFP) gene pseudotyped with VSV glycoprotein (Fig 5D). The incorporation of VSV G mutants into the membrane of VSV pseudotypes was verified by Western-blot analysis (Fig 5E). Indeed, the VSVΔG/G_D241A and VSVΔG/G_D243A pseudotypes exhibit an equivalent level of glycoprotein incorporation to that observed with VSVΔG/G_WT. Infectivity of VSV pseudotypes in HEK-293T cells was determined by flow cytometry, measuring the percentage of infected cells 16 hours p.i. In the absence of antibody, both VSVΔG/G_WT and VSVΔG/G_mut pseudotypes exhibited comparable infectious titers, approximately 2 10⁷ pfu/ml (Fig 5F, left panel). We then performed neutralization experiments by pre-incubating VSVΔG/G_WT and VSVΔG/G_mut inoculum with serial dilutions of mAb or Fab 8G5F11 (Fig 5F). As expected, VSVΔG/G_WT pseudotype was efficiently neutralized by mAb and Fab 8G5F11, with IC₅₀ values of 0.2 nM and 10 nM respectively (Fig 5F, right panel). In contrast, VSVΔG/G_D241A and VSVΔG/G_D243A pseudotypes were not neutralized by Fab or mAb 8G5F11 at either concentration, indicating that these residues play an important contribution to mAb 8G5F11 binding and neutralization activity (Fig 5F, left panel). It is noteworthy that the acidic character of the residues corresponding to D241 and D243, is widely conserved among vesiculoviruses (S5 Fig). This explains why 8G5F11 is broadly neutralizing [30].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Residues D241 and D243 are essential for the interaction of 8G5F11 with VSV G and for neutralization.

(A) Schematic description of the binding assays used to assess the capacity of VSV G WT and VSV G alanine mutants to bind CR domains and 8G5F11 mAb or Fab. (B) Surface expression test for VSV G WT and VSV G alanine mutants by measuring their ability to bind the conformational probe (CR3-GST labelled by and ATTO⁵⁵⁰ fluorophore). The histogram indicates the mean fluorescence intensity of ATTO⁵⁵⁰ positive cells for each G construct. Three independent experiments were performed. Error bars represent the standard deviation. (C) Antibody recognition assay for VSV G WT and mutants G by measuring their ability to bind 8G5F11 mAb (left panel) or Fab (right panel). Statistically significant differences with WT are indicated by stars (**p < 0.005, ***p < 0.0005). (D) Schematic description of the neutralization assay using VSVΔG/G pseudotypes. (E) Incorporation of VSV G WT and VSV G alanine mutants in VSVΔG viral particles, assessed using a polyclonal anti-VSV G and an anti-VSV M antibodies. (F) Neutralization of VSVΔG/GWT, VSVΔG/G_D241A, and VSVΔG/G_D243A by mAb (left panel) or Fab 8G5F11 (right panel). VSV pseudotypes were preincubated with increasing concentrations of mAb or Fab. At 16 hours p.i., the percentage of infected cells was determined by counting the number of cells expressing eGFP using flow cytometry. This was used to calculate the infectious viral titer. Data depict the mean with standard error for experiments performed in triplicate. Average IC₅₀ are indicated.

https://doi.org/10.1371/journal.ppat.1013589.g005

8G5F11 binding interferes with membrane fusion

As mentioned above, the binding site of 8G5F11 is distinct from that of the receptor. The antibody is therefore expected to have only a limited impact on receptor engagement. This is consistent with the findings of Munis et al. [30] showing that this 8G5F11 blocks infection after receptor binding and endocytosis. We therefore investigated whether the antibody interferes with membrane fusion using a previously described cell-cell fusion assay [34].

BSR cells expressing transiently VSV G were incubated in presence or absence of 10 nM 8G5F11 30 minutes before exposition to low pH. After incubation for 1 hour in DMEM buffered at pH 7.4, cells were fixed and analyzed (Fig 6A). As shown in Fig 6B, the presence of 8G5F11 resulted in a drastic inhibition of the fusion activity and virtually, no syncytia were observed at low pH, demonstrating that 8G5F11 efficiently blocks membrane fusion.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Inhibition of VSV G mediated cell-cell fusion by 8G5F11.

(A) Schematic description of cell-cell fusion assay. BSR cells were co-transfected with plasmids encoding VSV G and a cytoplasmic fluorescent marker (P-GFP). Twenty-four hours later, cells were incubated in fusion buffer at different pH values in the absence or presence of 10 nM 8G5F11 and syncytium formation was assessed. (B) Cell-cell fusion assay of VSV G with or without 8G5F11 after incubation at the indicated pH values. Nuclei were stained with DAPI. All images are representative examples from three independent experiments. (C) Schematic representation of 8G5F11 mechanism of action. There are no constraints for 8G5F11 to bind its epitope on the glycoprotein in its pre-fusion conformation when G is inserted into the viral membrane. However, the density of spikes on the surface of the viral particle prevents the structural transition due to the size of the antibody and its orientation when bound to the post-fusion state, which induce significant steric hindrance (indicated by arrows).

https://doi.org/10.1371/journal.ppat.1013589.g006

Viruses and cells

BSR, a clone of BHK21 (Baby Hamster Kidney cells, ATC CCCL-10) and HEK-293T (human embryonic kidney cells expressing simian virus 40T antigen, ATCC CRL-3216) were cultured in DMEM supplemented with 10% FCS, 50 U/mL of penicillin and 50 μg/mL of streptomycin at 37°C in a humidified incubator with 5% CO2. Drosophila Schneider 2 (S2) cell line (Invitrogen) were grown in ESF 921 medium (Clinisciences) supplemented with 50 U/mL of penicillin and 50 μg/mL of streptomycin at 28°C.

VSV MS (Mudd-Summer strain, Indiana serotype), VSVeGFP (a gift from Denis Gerlier) were propagated on BSR cells.

VSVΔG/G was propagated on HEK-293T cells that had been previously transfected with pCAGGS VSV G.

Plasmid and cloning

Point mutations were introduced starting from the cloned VSV G gene (Indiana Mudd-Summer strain; Indiana serotype) in the pCAGGS plasmid. Briefly, forward and reverse primers containing the desired mutation were combined separately with one of the primers flanking the G gene to generate two PCR products. These two G gene fragments overlapped in the region containing the mutation and were assembled into pCAGGS vector, which had been linearized using EcoRI, using Gibson assembly reaction kit (New England Biolabs).

Purification of VSV G

VSV G was directly solubilized from concentrated stocks of VSV MS using 2% w/v laurylmaltose neopentyl glycol (LMNG, Anatrace) for 22 hours at 37°C. Insoluble material was removed by centrifugation (20 minutes at 20 000g, 4°C). The supernatant was then loaded onto an anion exchange chromatography column (Resource Q 1 mL, Cytiva) equilibrated with 20 mM Tris-HCl pH 8.0, 2 mM EDTA, 0.01% w/v LMNG. VSV G was eluted with a linear gradient of NaCl (from 0 to 0.5 M NaCl). An additional purification step was performed on a size exclusion chromatography column (Superdex S200 Increase HR10/300 GL, Cytiva) equilibrated in 20 mM Tris-HCl pH 8.0, 150 mM NaCl, 2 mM EDTA, 0.005% w/v LMNG. Pure fractions containing VSV G were concentrated up to 5 mg/ml (Amicon Ultra 30 kDa cutoff, Millipore) and stored at -80°C until further use.

Production and purification of recombinant 8G5F11 Fab

The heavy and light chains (HC and LC respectively) of 8G5F11 mAb were sequenced (Absoute antibody, https://absoluteantibody.com/) and then cloned into an insect cell expression vector derived from the pMT-puro plasmid (Addgene #17923) using Gibson Assembly (New England Biolabs), as described in [49,50]. Briefly, the LC was placed under the control of a metallothionein (MT) promoter and Drosophila BiP secretion signal (BiP ss) at the N-terminus, with a polyA tail at the C-terminus. This LC part was followed by the Fd portion of the HC, also flanked by a MT promoter and BiP ss at the N-terminus, and fused to C-terminal Strep-tag II, followed by a poly(A) tail (S5 Fig).

Pseudostable S2 cell lines expressing 8G5F11 Fab were generated by transfecting S2 cells with pMT-puro plasmid encoding 8G5F11 Fab and carrying the puromycin resistance gene. For large scale production, S2 cells were cultured in suspension at 28°C under agitation (120 rpm) in medium supplemented with 7 µg/ml puromycin, until reaching 1.5 10⁷ cells/ml. The cultures were then induced with 500 µM CuSO4. Five days post-induction, the supernatant was harvested and concentrated by tangential flow filtration (Vivaflow 200, VWR). Proteins were purified via their Strep-tag II using a streptavidin affinity column (StrepTrap, Cytiva) equilibrated in 20 mM Tris-HCl pH 8.0, 150 mM NaCl, 2 mM EDTA. Elution was performed in the same buffer supplemented with 3 mM α-Desthiobiotin. An additional purification step was carried out by size exclusion chromatography (Superdex 200 Increase 10/300 GL, Cytiva) equilibrated in 20 mM Tris-HCl pH 8.0, 150 mM NaCl, 2 mM EDTA. Pure fractions containing 8G5F11 Fab were concentrated (Amicon Ultra, 30 kDa cutoff, Millipore) and stored at -80 °C until further use.

Binding assay of VSV G to mAb and Fab 8G5F11 at different pH values

8G5F11 antibody was immobilized on protein A-coated magnetic beads (Protein A Magnetic Beads, Biorad). Bead-bound antibody was incubated for 30 minutes  at room temperature in buffers adjusted to different pH values ranging from 8.0 to 5.0. In parallel, purified VSV G was pre-incubated 30 minutes in the same pH-adjusted buffers. Antibody-coated beads were then mixed with the pre-incubated VSV G samples and incubated for 30 minutes at room temperature with gentle agitation. After incubation, beads were washed three times with the corresponding buffer, and bound VSV G-Fab complex was eluted by boiling in SDS sample buffer. The samples were analyzed by SDS-PAGE under reducing conditions. Interaction experiments with Fab 8G5F11 were conducted under the same conditions except that magnetic beads were coated with Strep-Tactin (MagStrep XT, IBA) instead of protein A.

Purification of G-Fab complex

VSV G-Fab complexes were prepared by mixing VSV G with 8G5F11 Fab at a molar ratio of 1:1.2 in a total volume of 50 µL, aiming for a protein concentration of 5 mg/mL. The mix was then incubated for 30 minutes at 37 °C under gentle agitation (100–200 rpm).

Following incubation, the sample volume was adjusted to 1 mL with buffer C (10 mM Tris-HCl pH 8.0, 100 mM NaCl, 2 mM EDTA, 0.005% w/v LMNG) and loaded onto size exclusion chromatography (Superdex 200 Increase 10/300 GL, Cytiva) equilibrated in the same buffer. Fractions corresponding to the G-Fab complex were pooled and concentrated (Amicon Ultra, 30 kDa cutoff, Millipore) before downstream use.

Preparation of cryo-EM grids and data collection

For VSV G “alone” samples, grids were prepared by applying 4 µL of VSV G concentrated at 1.95 mg/ml to negatively glow discharged (25 mA, 45 seconds) UltraFoil Au grids 300 R1.2/1.3 (QUANTIFOIL) prior to plunge freezing using a Vitrobot Mark IV (FEI) (20°C, 100% humidity, 5.5 seconds blot time and 0 blot force). For VSV G in the post-fusion conformation, MES 1 M pH 5.5 was added to VSV G to reach final concentration of MES of 100 mM in the sample 30 minutes at 37°C just before freezing. Grids were pre-screened on a 200 kV Glacios and automated data collection was performed at the beamline CM01 of ESRF (Grenoble, France) on a 300 kV Titan Krios equipped with a K3 direct electron detector, coupled to an energy filter (Bioquantum LS/967, Gatan Inc, USA) [51]. 8,840 and 6,864 movies were recorded with EPU (Thermofisher) for the pre- and post-fusion of VSV G respectively, with a pixel size of 1.05 Å/pixels, a total dose of 40.4 e-/Å and a defocus range of -0.8 to -2.2 µm.

Cryo-EM grids of VSV G in complex with 8G5F11 Fab at pH 8.0 were prepared by applying 4 µL of the complex at a concentration of 0.8 mg/mL to negatively glow discharged (25 mA, 45 seconds) UltraFoil Au grids 300 R1.2/1.3 (QUANTIFOIL) prior to plunge freezing using a Vitrobot Mark IV (FEI) (30°C, 100% humidity, 5.5 seconds blot time and blot force 0). SerialEM based automated data collection was performed on a 200 kV Glacios electron microscope equipped with a K2 summit direct electron detector. 15947 movies were recorded with EPU (ThermoFisher) with pixel size of 1.145 Å/pixels, a total dose of 38.5 e-/Å and a defocus range of -0.8 to -2.2 µm.

MES 1 M pH 5.5 was added to G-Fab complex to reach a final concentration of MES of 100 mM in the sample 30 minutes at 37°C just before freezing. Cryo-EM grids of G-Fab complex at pH 5.5 were then prepared by applying 3.5 µL of the complex at 0.8 mg/mL to negatively glow discharged (25 mA, 45 seconds) UltraFoil Au grids 300 R1.2/1.3 (QUANTIFOIL) prior to plunge freezing using a Vitrobot Mark IV (FEI) (30°C, 100% humidity, 5.5 seconds blot time and blot force 0). The grids were pre-screened on a 200 kV Glacios electron microscope at IBS (Grenoble, France) and automated data collection was performed at the beamline CM01 of ESRF (Grenoble, France) on a 300 kV Titan Krios equipped with a K3 direct electron detector, coupled to an energy filter (Bioquantum LS/967, Gatan Inc, USA) [51]. 19423 movies were recorded with EPU (Thermofisher) with pixel size of 0.84 Å/pixels, a total dose of 49.71 e-/Å and a defocus range of -0.8 to -2.2 µm.

Cryo-EM data processing and model building

Image processing was performed using cryoSPARC software [52] and is summarized in S2-S3 and S6-S7 Figs. Recorded movies were dose-weighted and motion-corrected using cryoSPARC Patch Motion Correction program. CTF parameters were determined using cryoSPARC Patch CTF program. Particles were at first automatically picked using circular and elliptical blobs with a diameter between 100 and 230 Å for VSV G pH 5.5, between 100 Å and 250 Å for VSV G pH 8.0, between 150 Å and 300 Å for VSV G-Fab pH 8.0 and 80 Å and 300 Å for VSV G-Fab pH 5.5. A first round of particles picking was performed for each data set to select correct side views 2D classes. Topaz neural network was then used to enrich particles data set in particles side views and visible micelle [53]. Several rounds of 2D classification were then performed in order to clean particle data sets. Ab-initio reconstruction was used to generate at least 2 models to classify particles among them. Several rounds of Homogeneous refinement and Non-uniform refinement [54] were performed without imposed symmetry. The best models were selected for global and local (per-particle) CTF refinement to improve particle, map quality and resolution. C3 symmetry was only imposed on VSV pH 8.0 and VSV-Fab pH 5.5 reconstructions. Reported maps resolutions were estimated by gold-standard Fourier shell correlation (FSC) of 0.143 criterion.

Model building for protein chains was performed using COOT v.0.9.6 EL [55]. Model refinement and validation was performed using Phenix v.1.20.1 [56,57] and ISOLDE [58]. Fab model building was based on an initial model generated from AlphaFold2 [59] then fitted into the cryo-EM map and manually refined in COOT v.0.9.6 EL [55]. Geometry was validated using MolProbity [60].

Biolayer interferometry analysis (BLI)

BLI experiments were performed using an Octet RED96e (Sartorius, Fremont, CA, USA) instrument operated at 20°C under 1000 rpm stirring with Octet BLI Discovery software. The 8G5F11 mAb was captured on Protein A-functionalized sensors at 20 µg/mL during 40 seconds in acetate 10 mM pH 4.0. Fab 8G5F11 were covalently immobilized with random orientation on Amine Reactive 2nd Generation sensors (AR2G) at 20 µg/mL during 600 seconds in acetate 10 mM pH 4.0 using standard EDC-NHs chemistry. Interactions with VSV G_ecto, from 6.25 to 150 nM in running buffer (Tris-HCl 20 mM pH 8.0, NaCl 100 mM, EDTA 2 mM), were monitored during 300 seconds (association phase). Dissociation step was set at 600 seconds on the mAb surfaces and 1200 seconds on the Fab surfaces respectively. A bare sensor was always added as a reference surface to control the non-specific binding.

Data analysis was performed on Octet Analysis Studio software and kinetics constant (ka and kdis) were extracted by fitting the data with a 1:1 binding model and equilibrium dissociation constant KD was calculated.

Binding assay of 8G5F11 mAb and Fab to HEK-293T expressing VSV G WT or VSV G mutants at their surface

CR3-GST was produced in E. Coli (C41(DE3); Lucigen) and labeled with the fluorescent dye ATTO⁵⁵⁰ NHS ester (Sigma Aldrich) following the instruction of the manufacturer, as previously described (Nikolic et al., 2018). The labelled probe was then diluted at a concentration of 50 mM and stored at -80°C for further use. The labelling ratio was estimated to be around 2 dyes per CR3-GST molecules.

Cell surface expression levels of G WT or mutant G were estimated by adding GST-CR3 ATTO⁵⁵⁰ at a concentration of 500 µM on transfected cells 24 hours post-transfection. The fluorescence of cells was measured using Beckman CytoFlex S cytometer.

For binding assays, HEK-293T cells were transfected with pCAGGS plasmids encoding G WT or G mutants using the calcium phosphate transfection method. Twenty-four hours after transfection, cells were collected and incubated with either 8G5F11 mAb (at a concentration of 1.3 10^-3 mM) or 8G5F11 Fab (at a concentration of 2.5 10^-5 mM) for 45 minutes at RT. Goat anti-mouse Alexa fluor 488 (at 1.3 10^-3 mM) was added to the cells to accessed the binding of mAb 8G5F11. Anti-strep-tag coupled to PE fluorophore was added to the cells to accessed the binding of Fab 8G5F11. The fluorescence of cells was measured using a Beckman CytoFlex S cytometer.

VSVeGFP neutralization assay

HEK-293T cells at 70% confluence in 96-well plates were infected at MOI 1 with VSVeGFP pre-incubated for 30 minutes with increasing concentrations of 8G5F11 mAb or Fab. The percentage of infected cells (eGFP-positive) was determined using a Beckman CytoFlex S cytometer, 5 hours p.i.. Experimental MOIs were determined using the equation MOI=−ln[p(0)], where MOI is the multiplicity of infection and p(0) is the proportion of non-infected cells. Viral titer was deduced from the MOI, considering the volume and dilution of the viral inoculum and the total number of cells.

VSVΔG-GFP pseudotypes production

HEK-293T cells at 70% confluence were transfected with pCAGGS plasmids encoding G WT or mutant G using the calcium phosphate transfection method. Twenty-four hours after transfection, cells were infected with VSV∆G pseudotyped by VSV G WT at MOI 1. Two hours p.i., cells were washed to remove the inoculum. Then, at 16 hours p.i., supernatants containing the pseudotypes were collected. The infectious titers of the pseudotyped viruses was determined by counting cells expressing the GFP using a Beckman CytoFlex S cytometer 16 hours p.i.. Experimental MOIs were determined using the equation MOI=−ln[p(0)], where MOI is the multiplicity of infection and p(0) is the proportion of non-infected cells. Viral titer was deduced from the MOI, considering the volume and dilution of the viral inoculum and the total number of cells.

VSVΔG pseudotypes neutralization assay

HEK-293T cells at 70% confluence in 96-well plates were infected at MOI 1 with VSVΔG pseudotypes pre-incubated for 30 minutes with increasing concentrations of 8G5F11 mAb or Fab. The percentage of infected cells (GFP-positive) was determined using a Beckman CytoFlex S cytometer, 16 hours p.i.. Experimental MOIs were determined using the equation MOI=−ln[p(0)], where MOI is the multiplicity of infection and p(0) is the proportion of non-infected cells. Viral titer was deduced from the MOI, considering the volume and dilution of the viral inoculum and the total number of cells.

Western blot analysis

G WT and G mutants incorporation into the membrane of the pseudotyped VSVΔG particles was assessed, after concentrating viral supernatants, by Western blot analysis using rabbit-polyclonal anti-VSV G antibody (inhouse) (1/5000) and a mouse-monoclonal anti-VSV M antibody (1/5000). Goat anti-rabbit DyLight 800 conjugate and Goat anti-mouse DyLigth 680 conjugate were used as secondary antibodies. Blots were imaged using a LI-COR Biosciences Odyssey.

Cell-cell fusion assay

Cell-cell fusion was assessed as previously described [2,22]. Briefly, BSR cells were plated in 6 well plates (~70% confluence) and co-transfected with pCAGGS encoding VSV G and a P-GFP plasmid (Rabies virus phosphoprotein fused to GFP that diffuse in the cytoplasm). At 24 hours post-transfection, cells were incubated for 10 minutes at 37°C with 0.5 mL of DMEM containing 10 nM 8G5F11 mAb, followed by exposure to fusion buffer (DMEM supplemented with 10 mM MES) adjusted to the indicated pH (5.0-7.0). After a single wash, cells were incubated for 1 hour at 37°C in DMEM containing 10 mM HEPES pH 7.4 and 1% BSA. Cells were then fixed with 4% paraformaldehyde in PBS for 15 minutes, nuclei stained with DAPI, and syncytium formation was examined using a Leica SP8 confocal microscope (40 × objective).

Statistical analysis

All numerical data were calculated and plotted with mean ±SD resulting from three independent biological replicates. Results were analyzed using Prism v8.0.1 (GraphPad Software).

For the recognition assay, results are presented as normalized MFI. To do so, the geometric mean of the measured MFI for each condition of each experimental replicate was calculated. The geometric means were then normalized to the one of the WT. Results were compared using T-tests to compare means of biological replicates. Experiments were performed in triplicates.

For neutralization and pseudotypes assays, experimental titers were determined based on experimental MOIs. Dose-response parameters, namely Hillslope and IC₅₀, were determined using a log(inhibitor) vs response - variable slope model.