Reagents and resources

All reagents, consumables, and resources, including supplier information where applicable, are provided in Supplementary Table 5.

Cells lines

All cell lines were propagated at 37 °C in a humidified atmosphere with 5% CO₂. MeWo cells (HTB-65; ATCC) were propagated in minimal essential medium (Corning Cellgro) supplemented with 10% fetal bovine serum (FBS; Invitrogen), nonessential amino acids (100 μM; Corning Cellgro), antibiotics (penicillin, 100 U/ml; streptomycin, 100 μg/ml; Invitrogen), and the antifungal agent amphotericin B (Invitrogen). CHO-DSP1 cells previously derived from the Chinese hamster ovary (CHO) K1 cell line (CCL-61; ATCC) and express the dual split protein (DSP1) R8(1–8)³² were propagated using F-12K nutrient mixture with Kaighn’s modification (Invitrogen) supplemented with 10% FBS and antibiotics (penicillin, 100 U/ml; streptomycin, 100 μg/ml; Invitrogen) and maintained under puromycin selection (8 μg/ml; Invitrogen). Mel-DSP2 cells previously derived from the MeWo cell line and express the DSP2 R8(9–11)³² were propagated as for MeWo cells but under puromycin selection (5 μg/ml).

Viruses

The VZV parental Oka strain was originally cloned into a bacterial artificial chromosome (BAC) and designated pPOKA-BAC-DX⁵². The recombinant virus pOka-TK-GFP (pOka-rTK) was generated in a previous study³⁶. All recombinant pOka-TK-GFP VZV mutants were derived from the self-excisable BAC, pPOKA-TK-GFP-BAC-DX ΔORF31³⁶. The gB-KAN cassette⁵³ was digested with BstZ171 and NaeI. The 4056-bp fragment was gel-purified and used to transform electrocompetent GS1783 Escherichia coli carrying the pPOKA-TK-GFP ΔORF31 BAC. After red recombination, the pPOKA-TK-GFP BAC was purified using a large-construct purification kit (Qiagen). BACs were digested with Hind III to verify that spurious recombination had not occurred, and successful incorporation of ORF31 mutations were verified by sequencing the BAC directly. To generate BAC-derived VZV, 10⁶ MeWo cells seeded in six-well plates (Nunc) 24 h previously were transfected with 4 µg of the pPOKA BACs using Lipofectamine 2000 (Invitrogen) following the manufacturer’s instructions. Recombinant VZV was typically recovered at 5–10 days post-transfection. All virus stocks, pOka and gB DIV mutants, were sequenced to verify that the expected ORF31[gB] sequence was present. Briefly, DNA was extracted from infected cells using proteinase K and phenol/chloroform (Invitrogen). VZV ORF31 was amplified by PCR with KOD Extreme™ (EMD Millipore) following the manufacturer’s instructions using the oligonucleotides [31]F56625-56645/[31]R59697-59717. The PCR products were gel purified and sequenced by Sanger sequencing.

Construction of VZV pOka-TK-GFP-gB-TEVV5

For the purification of gB from VZV-infected cells, a green fluorescent protein (GFP) expressing virus was generated that produced gB with a tag containing the tobacco etch virus protease cleavage site and a V5 epitope (TEVV5). A gB-Kan-TEVV5 shuttle vector was generated in a three-step cloning procedure. First, a gB-Kan-V5 vector was generated by amplifying two fragments from the gB-Kan vector⁵³ using AccuPrime™ Pfx (Invitrogen) with oligonucleotides gB-AgeI/gB931 and gB-V5/M13R, purified using a QIAquick gel purification kit (QIAGEN) following the manufacturer’s instructions, and ligated into the AgeI and SpeI site of the gB-Kan vector. Secondly, two fragments were amplified from gB-Kan-V5 using AccuPrime™ Pfx with oligonucleotides gB-AgeI/gB_cterm_Sprotein and gB_link_TEV_link/M13R, gel purified and ligated into the AgeI/SpeI site of gB-Kan. The final step deleted the S-tag from the gB-Kan-STEVV5 to generate the gB-Kan-TEVV5 shuttle vector. Two fragments were amplified from gB-Kan-STEVV5 using AccuPrime™ Pfx with oligonucleotides gB-AgeI/ΔS-tag-sense and ΔS-tag-antisense/M13R, gel purified, and ligated into the AgeI and SpeI site of the gB-Kan vector. The gB-Kan-TEVV5 shuttle vector was used to reconstitute ORF31-TEVV5 into the pOka-TK-GFP-ΔORF31 BAC to generate pPOKA-TK-GFP-gB-TEVV5 and recovery of pOka-TK-GFP-gB-TEVV5 virus was performed as described in the ‘Viruses’ section.

Construction of the VZV pOka-TK-GFP-gB-TEVV5 gB DIV mutants

Site directed mutagenesis was performed using the pCAGGs-VZV gB (pCAGGs-gB) vector for template to generate DNA fragments using AccuPrime™ Pfx. For the single alanine substitutions at S589A, R592A, and I594A, DNA fragments were amplified using the oligonucleotide combinations pCAGGs-gB-XmaI-sense/S589A-antisense (S589A), S589A-sense/pCAGGs-gB-AgeI-antisense, pCAGGs-gB-XmaI-sense/R592A-antisense (R592A), pCAGGs-gB-XmaI-sense/I594A-antisense (I594A), and pCAGGs-gB-3617-sense/pCAGGs-gB-AgeI-antisense. For the combined alanine substitutions at S589A/R592A/I594A (⁵⁸⁹AAA⁵⁹⁴), DNA fragments were amplified using the oligonucleotides pCAGGs-gB-XmaI-sense/589AAA594-antisense and 589AAA594-sense/pCAGGs-gB-AgeI-antisense. For alanine substitutions at Q596, N597, and ⁵⁹⁶QN⁵⁹⁷, DNA fragments were amplified using the oligonucleotide combinations pCAGGs-gB-XmaI-sense/Q596A-antisense (Q596A), pCAGGs-gB-3623-sense/pCAGGs-gB-AgeI-antisense, pCAGGs-gB-XmaI-sense/pCAGGs-gB-3622-antisense, N597A-sense/pCAGGs-gB-AgeI-antisense (N597A), pCAGGs-gB-XmaI-sense/Q596A-antisense, and N597A-sense/pCAGGs-gB-AgeI-antisense (⁵⁹⁶QN⁵⁹⁷). All DNA fragments were gel purified and digested with the appropriate restriction enzymes for ligation into the XmaI/AgeI site of pCAGGs-gB. To generate the ⁵⁹²A/⁵⁹⁶QN⁵⁹⁷ substitutions the pCAGGs-gB-R592A was used as template to amplify DNA fragments using oligonucleotides pCAGGs-gB-XmaI-sense/R592A-Q596A-antisense and N597A-sense/pCAGGs-gB-AgeI-antisense. The two DNA fragments were gel purified and digested with the appropriate restriction enzymes for ligation into the XmaI/AgeI site of pCAGGs-gB. For the alanine substitution Y667A, E670A and ⁶⁶⁷A/A⁶⁷⁰, DNA fragments were amplified using the oligonucleotide combinations pCAGGs-gB-MluI-sense/Y667A-antisense (Y667A), E670A-sense/pCAGGs-gB-AgeI-antisense (E670A), pCAGGs-gB-MluI-sense/Y667-antisense and V668-sense/pCAGGs-gB-AgeI-antisense. All DNA fragments were gel purified and digested with the appropriate restriction enzymes for ligation into the MluI/AgeI site of pCAGGs-gB. To generate the ⁵⁹²A/⁵⁹⁶QN⁵⁹⁷/⁶⁶⁷A/A⁶⁷⁰ substitutions the pCAGGs-gB-592A/596AA597 was used as template to amplify DNA fragments using oligonucleotides pCAGGs-gB-MluI-sense/Y667A-antisense (Y667A) and E670A-sense/pCAGGs-gB-AgeI-antisense (E670A). All the pCAGGs vectors were sequenced to verify that only the specific mutations were incorporated and spurious mutations from the PCR had not been incorporated.

The ⁵⁸⁹AAA⁵⁹⁴, ⁵⁹⁶AA⁵⁹⁷, ⁵⁹²A/⁵⁹⁶AA⁵⁹⁷, and ⁵⁹²A/⁵⁹⁶AA⁵⁹⁷/⁶⁶⁷A/A⁶⁷⁰ combined and single mutations were transferred into the gB-Kan-TEVV5 shuttle vector by restriction digest of the pCAGGS-gB mutant vectors and cloned into the NdeI/AgeI site of the gB-Kan-TEVV5. All the gB-Kan-TEVV5 vectors were sequenced to verify that only the specific mutations were incorporated and spurious mutations from the PCR had not been incorporated. The gB-Kan-TEVV5 shuttle vectors were used to reconstitute ORF31-TEVV5 DIV mutants into the pOka-TK-GFP-ΔORF31 BAC to generate pPOKA-TK-GFP-gB-TEVV5 DIV mutant BACs and recovery of pOka-TK-GFP-gB-TEVV5 DIV mutant viruses was performed as described in the ‘Viruses’ section. Critically, sequencing of virus stocks confirmed that none of the gB β23 mutants had unexpected nucleic acid substitutions in ORF31 (Supplementary Fig. 6).

Purification of native full-length gB from VZV-infected cells

MeWo cells were infected with the pOka-gB-TEVV5 virus and replication was allowed to proceed until cytopathic effect was observed across 90% of the cell monolayer. Infected cells were scrapped into ice cold PBS and pelleted at 424 RCF for 5 mins. Cells were lysed in glycoprotein extraction buffer (0.1 M Tris-base[pH7.2], 0.1 M NaCl, 5 mM KCL, 1 mM CaCl₂, 0.5 mM MgCl₂, 1% sodium deoxycholate, and 1% NP40) plus an EDTA-free protease inhibitor cocktail (Roche, CA, USA)⁵⁴. Cell lysates were clarified at 3000 RCF for 10 mins and 5 ml of clarified lysate was incubated with 250 μl of anti-V5 agarose beads (Sigma) for 2 h at room temperature. The beads were washed extensively in PBS + 0.1% Triton and a final wash with PBS then incubated with PBS containing TEV protease (55 µg/ml) for 20 h +4 °C. The TEV cleaved gB was eluted from the beads with TBS pH7.4 containing 1 mg/ml lauroylsarcosine (Sigma) and 1 mg/ml Amphipol 8–35 (Anatrace). Buffer exchange into TBS pH7.4 and 1 mg/ml Amphipol 8–35 using Amicon® Ultra-4 centrifugation filters with a 100 kDa cutoff (Millipore). The concentration of Amphipol 8–35 was brought up to 35 mg/ml and incubated at room temperature for 4 h then Bio-Beads™ SM-2 (Bio-Rad) were added and incubated for 16 h at 4 °C. The purified native, full-length gB was resolved on either Native PAGE or denaturing SDS–PAGE and either stained with Coomassie (Native PAGE) or Gel Code Blue (SDS–PAGE) following the manufacturer’s instructions. To determine that the purified protein was native, full-length gB western blot was performed by transferring proteins to Immobilon-P membranes (Millipore Biosciences, Temecula, CA) and blocked with 5% BSA. The mouse mAb SG2, the human mAb 93k and a rabbit polyclonal antiserum, 746–868, which recognizes the peptide sequence ⁸³³PEGMDPFAEKPNAT⁸⁴⁶ in the cytoplasmic region of pOka gB³³, were used to detect gB. Horse radish peroxidase conjugated antibodies that detect either mouse, human, or rabbit IgG (GE Healthcare Bio-Sciences Corp., Piscataway, NJ) were used and HRP activity detected using ECL plus (GE Healthcare Bio-Sciences Corp., Piscataway, NJ). The native, full-length gB was further purified on a Superose-6 column (GE Healthcare Life Sciences) into TBS pH7.4 to remove aggregates.

Isolation of mAb 93k and preparation of 93k Fab

B-lymphocytes from a VZV immune individual were used to generate triomas that secrete antibodies that cross reacted with VZV gB and had neutralizing activity against 11 clinical isolates⁵⁵. Triomas that secreted gB reactive antibodies were subcloned by limiting dilution resulting in clone 93kA9. The 93k Fab coding sequences for the variable heavy (VH) chain and variable light (VL) were subsequently sequenced and cloned into the pRS5a mammalian expression vector (Novartis AG, Basel, Switzerland), which expresses a generic constant region of IgG1 heavy chain (HC) and light chain (LC). The complete heavy chain has a cleavable double strep tag at the C-terminus. Fab fragments were generated from mAb 93k then purified by affinity and size-exclusion column.

Preparation of mAb 93k Fab fragments bound to the native full-length VZV gB

Native full-length VZV gB was purified from infected cells as described in the ‘Purification of native full-length gB from VZV-infected cells’ section except the mAb 93k Fab fragments were added in molar excess immediately after the TBS pH7.4 and 1 mg/ml Amphipol 8–35 buffer exchange. The native full-length gB plus Fab fragments were incubated overnight at +4 °C on a rotary mixer. The complexes were concentrated using Amicon Ultra 10 kDa filter units following the manufacturer’s instructions. The concentration of Amphipol 8–35 was brought up to 35 mg/ml and incubated at room temperature for 4 h then Bio-Beads™ SM-2 (Bio-Rad) were added and incubated for 16 h at 4 °C. The purified native, full-length gB-Fab complexes were evaluated by Native PAGE and purified on a Superose-6 column (GE Healthcare Life Sciences) into TBS pH7.4 to remove aggregates.

Grid freezing

Lacey carbon copper 400 mesh grids with an ultrathin layer of carbon or Quantifoil R 1.2/1.3 gold 300 mesh grids were used for specimen freezing. EM grids were glow discharged for 25 s. To each grid, 3 µl of purified protein was dispensed and immediately plunge frozen into liquid ethane using a Leica EM GP. Optimum chamber humidity and blotting times were determined empirically for each sample and ranged from 95 to 99% and 1.8–2.5 s.

Cryo-EM data collection

Micrographs for native, full-length VZV gB in complex with 93k Fab fragments were captured on a 300-kV Titan Krios (FEI) controlled by SerialEM⁵⁶ to automate the data collection procedure. Movie data (11,283 total stacks) were captured with a Gatan K2 Summit (5 μm/pixel) in counted mode with a dose rate of ~1.335 e⁻/Ų/s per frame and 200 millisecond exposure time per frame and 12-s total exposure time at a nominal magnification of ×130,000 and a pixel size of 1.06 Å/pixel on the specimen. The defocus range was 1.5–2.0 µm.

Map reconstruction of full-length VZV gB in complex with mAb 93k Fab

The motion correction and damage compensation for all movie-mode data were performed using MotionCor2⁵⁷. CTFFIND4 was used to estimate the contrast transfer function parameters⁵⁸. Initially, the first 100 micrographs were selected to box out particle images using EMAN2’s e2boxer.py⁵⁹, followed by Relion’s^60,61 2D classification which generated a set of 2D class averages. The good 2D class averages were selected as templates to box out the particle images from all micrographs using Relion’s auto-picking. EMAN2’s e2initialmodel.py⁵⁹ or Relion’s 3D initial model was utilized to build the initial model. A couple of Relion’s 2D classifications were first performed to remove junk, and the good classes were selected from 3D classification to do the final 3D auto-refine, for which 856,068 particles were selected. The C3 symmetry was imposed during the 3D auto-refine of Relion.

Structure and visualization of the VZV gB-93k complex

The resolution of the cryo-EM map and model of gB-93k was determined using Fourier shell correlation overall resolution estimate⁶². A protomer model was built by fitting our X-ray structure (PDB 6VLK; unpublished) and the 93k VH and VL chains were built de novo. The gB-93k structure models were generated and refined using Coot⁶³ and Phenix^64,65,66,67. ResMap was used to calculate local resolution variation³⁴. A newly developed Q scoring tool was applied to calculate feature resolvability³⁵. Interactions between amino acids were calculated using the Find Contacts tool in UCSF Chimera 1.13.1 using the default settings⁶⁸. Surface electrostatic potential was calculated using APBS (Adaptive Poisson-Boltzmann Solver)⁶⁹. All images and movies were generated using the Animation tool in UCSF Chimera 1.13.1.

Cell-free VZV neutralization assay

Cell-free VZV stocks were prepared as described⁷⁰. MeWo cells in 100 mm culture dishes infected with pOka-TK-GFP were washed with cold PBS then incubated at room temperature with PBS + 0.1% EDTA. The cells were dislodged by pipetting, centrifuged at 424 RCF for 5 mins then resuspended in PSGC buffer (PBS + 145 M sucrose + 6 mM L( + )-glutamic acid + 10% FBS). The resuspended cells were transferred to a Dounce homogenizer (KONTES^®) and disrupted with 15 strokes of pestle A and 15 strokes of pestle B. The resulting homogenate was centrifuged at 3000 RCF to removed cell debris and the supernatant was stored in 1 ml aliquots under liquid nitrogen. The titers of cell-free VZV were typically in the 3.5 log₁₀/ml range. To test antibodies for VZV neutralization capabilities, 100 pfu of cell-free pOka-TK-GFP was incubated with 10 μg antibody (93k, SG2 or 206) at room temperature for 30 mins. MeWo cells seeded 24 h previously at 1 × 10⁵/cm² in 12-well plates were inoculated with the pOka-TK-GFP/antibody mixtures and incubated for 24 h. The media was changed, and the plates were incubated for a further 72 h then fixed with 4% paraformaldehyde. Plaques were detected by immunohistochemical staining using an anti-VZV mixed mouse mAb followed by detection with a streptavidin conjugated goat anti-mouse IgG (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) and alkaline phosphatase conjugated avidin (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA). Enzyme activity was detected using a fast red substrate (0.1MTris - pH 8.0, 5 µM Naphthol AS-Mx phosphate (Sigma), 80 µM Fast Red TR (Sigma)).

Quantitation of cell surface gB for the DIV mutants

CHO-DSP1 cells (8 × 10⁵ cells/well) in six-well plates were transfected with 5 μg WT or mutant pCAGGS-gB expression vectors. Cells were dislodged at 24 h post transfection using an enzyme-free cell dissociation buffer (Life Technologies, Grand Island, NY), washed with PBS then fixed with 1% paraformaldehyde. The fixed cells were washed with PBS then resuspended in FACS staining buffer (DPBS (Dulbecco’s Phosphate-Buffered Saline; Cellgro, Manassas, VA) with 0.2% IgG-free BSA (Jackson ImmunoResearch, West Grove, PA) and 0.1% NaN3 (Sigma Aldrich, St. Louis, MO)) for cell surface staining with anti-VZV gB mAb SG2-2E6 or 93k. A donkey anti-mouse IgG-Alexa Fluor 555 antibody (SG2-2E6) or goat anti-human IgG-Alexa Fluor 488 antibody (93k) (Life Technologies, Grand Island, NY) was used to detect bound anti-VZV gB mAb. Total gB expression was determined by using the same staining protocol except cells were permeabilized using Cytofix/Cytoperm (BD Biosciences, San Jose, CA) before adding the primary antibody and during the staining procedure. Stained cells were analyzed using a FACSCalibur with CellQuest Pro (BD Biosciences, San Jose, CA). FlowJo (TreeStar, Ashland, OR) was used to determine the quantity of total and cell surface gB on the transfected cells. The quantities for gB mutants were normalized to WT gB, which was set at 100%. Experiments were performed with at least two gB mutant clones, each tested in duplicate.

VZV stable reporter fusion assay

The stable reporter fusion assay for the VZV glycoproteins gB/gH-gL has been reported previously³² but was adapted for use with a 96-well plate format. CHO-DSP1 cells seeded at 8 × 10⁵ per well in six-well plates 20 h previously were transfected with 1.6 μg each of pCAGGs-gB, pME18s-gH[TL], and pCDNA-gL plasmids with Lipofectamine 2000 following the manufacturer’s instructions. At 6 h post transfection, the transfected CHO-DSP1 cells were trypsinized, collected by centrifugation at 424 RCF, and resuspended in 1 ml of medium, of which 250 μl of cells were mixed with 0.75 ml of Mel-DSP2 cells at 10⁶ cells/ml. To test fusion inhibition properties of mAbs, 93k, SG2 and 206, 10 µg of antibody was added to the cell mixtures. The cells were mixed by inversion and 75 µl of the suspension was dispensed to at least triplicate wells of 96-well blacked sided optical bottom plates culture plates (Thermo Scientific). At 40 h post seeding, 50 µl membrane permeable coelenterazine-H (5 μM, Nanolight Technology) substrate for five mins at room temperature. Fusion was quantified by measuring luminescence using a Synergy H1 Multi-mode reader (Biotek). A minimum of two clones were tested in duplicate experiments.

Immunofluorescence staining of MeWo cells

To each well of a 12-well plate (Nunclon™ Delta Surface; Thermo Scientific) a sterile 18 mm coverslip (Fisher Scientific) was placed and 2 ml of MeWo cells at 2 × 10⁵/ml was dispensed and incubated overnight. MeWo cells were transfected with 2ug of pPOKA-TK-GFP BACs carrying gB mutants using Lipofectamine 2000 following the manufacturer’s instructions. At 72 h post transfection, the media was aspirated, the coverslips washed with PBS and fixed with 4% paraformaldehyde for 10 mins. Immunofluorescence was performed by blocking the cells with PBS + 10% normal donkey serum (NDS) + 0.1% Triton X-100 then adding a mouse mAb to the immediate early protein IE62 in PBS + 1% NDS + 0.1% Triton X-100. The anti-IE62 mAb was detected with the donkey anti-mouse IgG-Alexa Fluor 555 and nuclei were stained with Hoechst 33342 in PBS + 1% NDS + 0.1% Triton X-100. Coverslips were mounted on glass slides (Selectfrost; Fisher Scientific) using Fluoromount-G (SouthernBiotech) and a minimum of five images were captured for each transfection using a Keyence fluorescence microscope using a ×20 objective.

Quantitation of plaque sizes for the VZV gB DIV mutants

MeWo cells were seeded at 10⁶ cells/well 24 h prior to inoculation with 50 pfu of either wild type pOka or gB DIV mutants. Each well of the six-well plate was fixed at 4 days post inoculation with 4% formaldehyde and stained by immunohistochemistry. Images of stained plaques (n = 40) were digitally captured, the stained plaque was outlined, and the area (mm²) was calculated using ImageJ (National Institute of Mental Health). Statistical analyses were performed using Prism (GraphPad Software).

Immunoprecipitation of VZV gB DIV mutants

CHO-DSP1 cells seeded in six-well plates were transfected with 5 µg/well of pCAGGS-gB vectors carrying the DIV mutations using Lipofectamine 2000 following the manufacturer’s instructions. At 24 h post transfection cells were lysed with glycoprotein lysis buffer, the same buffer used for the purification of native, full-length VZV gB, and snap frozen in liquid nitrogen and stored at −20 °C. The SG2 or 93k mAbs were cross-linked to immobilized protein A (Pierce, Rockford, IL)⁷¹. Each 20 µg of mAb was incubated with 30 µl protein A beads for 1 h at room temperature on a rotary mixer. The beads were washed with DPBS then mAbs were cross-linked to the beads with 0.2 M sodium borate [pH9.0] and 20 mM DMP for 30 mins. The cross-linking reaction was quenched with 0.2 M NaCl and 0.2 M ethanolamine [pH 8.0] for 2 h at room temperature. The cross-linked beads were washed with DPBS. Lysates from the pCAGGs-gB transfected CHO-DSP1 cells were divided equally and incubated overnight at +4 °C with either the SG2 or 93k cross-linked beads. The beads were washed extensively with DPBS + 0.1% Triton X-100 and a final wash of DPBS to remove the Triton X-100. Bound proteins were eluted into sodium dodecyl sulfate (SDS) sample buffer (Bio-Rad) containing 5% 2-mercaptoethanl (Sigma) by incubating the beads at 100 °C for 5 min. Denatured samples were resolved on SDS-polyacrylamide gel electrophoresis precast gels (Bio-Rad, Hercules, CA) and western blot was performed using the 746–868 rabbit poly clonal IgG.

Immunoprecipitation of VZV gB DIV mutants in complex with gH-gL

CHO-DSP1 cells were transfected as described in the previous section with pCAGGS-gB vectors carrying the DIV mutations, pME18s-gH[V5] and pCDNA3.1-gL (1.6 µg of each vector). At 24 h post transfection cells were lysed with glycoprotein lysis buffer and snap frozen in liquid nitrogen and stored at −20 °C. The gB/gH-gL complexes were immunoprecipitated with anti-V5 agarose (Sigma). Wash steps, protein elution, and SDS–PAGE were performed as outlined in the previous section. Western blots were performed using either mouse anti-V5 tag (Bio-Rad), 746–868 rabbit poly clonal IgG, or mAb 93k.

Statistics and reproducibility

All quantitative data were analyzed with two-way ANOVA to determine statistical significance using Prism (GraphPad Software). All statistical analyses are presented in the Source Data file. Images of gB-93k class averages in Fig. 2a are representative of 23 classifications. Confocal micrographs in Figs. 5c and 6c are representative images of n = 10 from two independent experiments. Images of gels and western blots in Figs. 1a, 5e, f, and 6e, f and Supplementary Figs. 2a, b and 7 are representative of at least two independent experiments.

Reporting summary

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