Human participant

Blood sample from Donor 19, a 41-year-old Asian male, was collected at Columbia University Irving Medical Center. Donor 19 was confirmed for a BA.5 infection by PCR sequencing (single-nucleotide polymorphisms) and provided written informed consent. Sample collections were performed under protocols reviewed and approved by the Institutional Review Board of Columbia University. Clinical information of Donor 19 is provided in Supplementary Fig. 1a.

Cell lines

Vero-E6 (CRL-1586) and HEK293T (CRL-3216) cells were purchased from the American Type Culture Collection (ATCC). Expi293 cells (A14527) were purchased from Thermo Fisher Scientific. Vero-E6-TMPRSS2-T2A-ACE2 (NR-54970) were obtained from BEI Resources. 293T-ACE2 were kindly provided by Dr. Jesse D. Bloom. The morphology of each cell line was confirmed visually before use. All cell lines tested mycoplasma negative. Vero-E6 and Vero-E6-TMPRSS2-T2A-ACE2 cell lines are from African green monkey kidneys. HEK293T, 293T-ACE2, and Expi293 cells are of female origin.

Plasmid construction

SARS-CoV-2 spike-expressing plasmids for D614G, Alpha, Beta, Gamma, Delta and Omicron were previously generated^13,20,21,27. Expressing constructs for the spike proteins of SARS-CoV-2-like sarbecoviruses were either previously generated⁶⁹ or newly generated by synthesizing (GenScript) spike genes and then cloned into the pCMV3 vector. To generate the expression constructs for soluble spike trimer (S2P) proteins, the ectodomains (1-1208aa, numbering based on WA1) of the spikes were PCR amplified and cloned into the paH vector and then introduced K986P and V987P substitutions, as well as a “GSAS” substitution of the furin cleavage site (682-685aa in WA1) into the spikes⁷⁰. SARS-CoV S2P was fused with an AVI tag at the C terminus and D614G and XBB.1.5 S2P spikes were tagged with a 6×His tail also at the C terminus. To make the SARS-CoV-2 RBD-expressing construct, the RBD region (319–537aa) of each variant was fused with a 6×His tag and then cloned into the p3BNC vector. All constructs were confirmed by Sanger sequencing.

Antibody-expressing constructs were generated as previously described³⁸. The variable regions of heavy and light chains for each antibody were synthesized (GenScript) and then cloned into the gWiz vector. To make spike/antibody plasmid constructs carrying individual mutations, the Q5 Site-Directed Mutagenesis Kit (NEB) was utilized following the manufacturer’s instructions.

Protein purification

To make human ACE2-Fc (hACE2) protein, pcDNA3-sACE2-WT(732)-IgG1⁷¹ (Addgene plasmid #154104, gift of Erik Procko) plasmid was transfected into Expi293 cells using 1 mg/mL polyethyleneimine (PEI) at a ratio of 1:3, and the supernatants were collected after 5 days. hACE2 was purified from the cell supernatant by using rProtein A Sepharose (GE). For antibody purification, both heavy and light chains of each antibody were transfected at a ratio of 1:1 into Expi293 cells using PEI. And the expressed antibody in the cell supernatant was purified using the same method as for hACE2 purification. For the spike trimer proteins or RBD proteins, paH-spike or p3BNC-RBD, respectively, was transfected into Expi293 cells using PEI at a ratio of 1:3, and the supernatants were collected five days later. The His-tagged and the AVI-tagged proteins were purified using Excel resin (Cytiva) and Agarose-bound Galanthus nivalis lectin (VectorLabs, AL-1243-5) according to the manufacturers’ instructions. The molecular weight and purity were checked by running the proteins on SDS-PAGE.

After purification of 19-77 and 19-77\(\triangle\)A/L/V, 100 µg of each antibody was prepared and run through a Superdex 200 Increase 10/300 GL column to generate their size exclusion chromatography (SEC) profiles. AVI-tagged SARS-CoV S2P protein were biotinylated using the BirA biotin-protein ligase standard reaction kit (Avidity LLC; BirA500) following the manufacturer’s instructions.

ELISA

50 ng per well of an antigen such as S2P spike, NTD (ACROBiosystems, S1D-C52H6), RBD (ACROBiosystems, SPD-C52H1), RBD-SD1 (Exonbio, 19Cov-S130), S1 (ACROBiosystems, S1N-C52H3), or S2 (ACROBiosystems, S2N-C52H2) was coated onto ELISA plates at 4 °C overnight. The ELISA plates were then blocked with 300 μL of blocking buffer consisting of phosphate-buffered saline (PBS) with 1% bovine serum albumin and 20% bovine calf serum (Sigma-Aldrich) at 37 °C for 2 h. Afterward, 100 μL of fivefold serially diluted antibodies was added and then incubated at 37 °C for 1 h. Next, 100 μL of 10,000-fold diluted Peroxidase AffiniPure goat anti-human IgG Fcγ fragment-specific antibody (Jackson ImmunoResearch, catalog no. 109-035-170, RRID: AB_2810887) was added into each well and incubated for another 1 h at 37 °C. The plates were washed between each step with PBST (0.5% Tween-20 in PBS). Last, 3,3′,5,5′-tetramethylbenzidine (TMB) substrate (Sigma-Aldrich) was added and incubated before the reaction was stopped using 1 M sulfuric acid. Absorbance was measured at 450 nm.

Antibody pharmacokinetics in mice

About 100 µg (1 mg/mL) of each antibody was intraperitoneally (ip) injected into Balb/c mice (three mice for each antibody). Mouse blood was collected on days 2, 4, 7, and 10 post-IP injection and antibody concentrations in serum were measured by ELISA. Briefly, 100 ng goat anti-human IgG Fc antibody (Cat: 109-005-008, Jackson ImmunoResearch) was coated per well in 96-well plates overnight and the ELISA plates were then blocked with 300 μL of blocking buffer consisting of PBS with 1% bovine serum albumin and 20% bovine calf serum (Sigma-Aldrich) at 37 °C for 2 h. Afterward, 100 μL of threefold serially diluted serum or twofold serially diluted purified antibody was added and then incubated at 37 °C for 1 h. Next, 100 μL of 10,000-fold diluted Peroxidase AffiniPure goat anti-human IgG H + L HRP (Cat: 109-035-088, Jackson ImmunoResearch) was added into each well and incubated for another 1 hour at 37 °C. The plates were washed between each step with PBST (0.5% Tween-20 in PBS). Last, TMB substrate (Sigma-Aldrich) was added and incubated before the reaction was stopped using 1 M sulfuric acid. Absorbance was measured at 450 nm.

Pseudovirus production and infectivity

SARS-CoV-2 pseudoviruses were generated as previously described³⁸. In brief, HEK293T cells were transfected with a spike-expressing construct using 1 mg/mL PEI and then infected with VSV-G pseudotyped ΔG-luciferase (G*ΔG-luciferase, Kerafast) 1 day post-transfection. Two hours after infection, cells were washed three times with PBS, changed to fresh medium, and then cultured for one more day before the cell supernatants were harvested. Pseudoviruses in the cell supernatants were clarified by centrifugation, aliquoted, and stored at −80 °C.

To evaluate the infectivity of the pseudotyped escape variants of 19-77 and 19-77∆V, fresh pseudoviruses without freezing and thawing were serially titrated from 50 µL with a dilution factor of 3 and then inoculated into Vero-E6 cells. After overnight culture, Vero-E6 cells were harvested and quantified for luciferase activity using the Luciferase Assay System (Promega).

Pseudovirus neutralization assay

To normalize the viral input between assays before conducting the neutralization assays, pseudoviruses of SARS-CoV and SARS-CoV-2 variants and SARS-CoV-2-like sarbecoviruses were titrated on Vero-E6 cells and 293T-ACE2 cells. Heat-inactivated serum from Donor 19 was serially diluted starting from 1:25 with a dilution factor of four. Monoclonal antibodies were fivefold serially diluted starting from 20 µg/mL in 96-well plates in triplicate. Then, 50 µL of diluted pseudovirus was added and incubated with 50 µL serial dilutions of serum or mAb for 1 h at 37 °C. During the coculture, target cells were trypsinized, resuspended with fresh medium, and then added into the virus-sample mixture at a density of 4–10 × 10⁴ cells/well. The plates were incubated at 37 °C for ~12 h before luciferase activity was quantified using the Luciferase Assay System (Promega) using SoftMax Pro v.7.0.2 (Molecular Devices). Neutralization ID₅₀ values for sera and IC₅₀ values for antibodies were calculated by fitting a nonlinear five-parameter dose-response curve to the data in GraphPad Prism v.10.

Selection of escape mutations

SARS-CoV-2 isolate Omicron JN.1 (BEI NR-59693) was mixed with serial fivefold dilutions of 19-77∆V antibody at an MOI of 0.2 and incubated for 1 h. Following incubation, the mix was overlaid on a 24-well plate bearing a monolayer of Vero-ACE2-TMPRSS2 cells (BEI NR-54970) to a final volume of 1 mL. Plates were incubated at 37 °C/ 5% CO₂ for 70 h till cytopathic effect (CPE) was complete (100%) in virus control wells bearing no antibody. At this time, all wells with antibody dilutions were scored to determine the 50% inhibition titer (EC₅₀) and supernatant collected from this well was used for the subsequent round of selection. Passaging of the progeny over new Vero-ACE2-TMPRSS2 cells continued till each of the virus variant was able to form CPE in the presence of 50 µg/mL of the antibody. The resulting supernatant was then collected, and RNA was extracted using QiaAMP Viral RNA kit (Qiagen 57704). cDNA was obtained using the Superscript IV enzyme (Thermo Scientific 18090010). Spike gene from the cDNA was amplified using limiting dilution nested PCR and sequenced using Sanger sequencing (Genewiz). Multiple clones from limiting dilution nested PCR were sequenced to confirm the dominant mutants in the pool of the resulting progeny viruses, and a percentage of their prevalence was calculated from the total number sequenced. At least eight clones were sequenced from each of the passages reported in Fig. 7.

To generate recombinant replication-competent VSV-ΔG bearing the JN.1 spike protein (VSV-ΔG-JN.1), the pVSVΔG-SARS-CoV-2-S_nLucP plasmid was purchased from Kerafast (Cat# EGA292) and its encoding SARS-CoV-2-S gene was replaced with the JN.1 spike gene to create pVSVΔG-JN.1_nLucP. The pVSVΔG-JN.1_nLucP plasmid and a set of helper plasmids, including VSV-N, VSV-P, VSV-L, and VSV-G (Kerafast, Cat# EH1012), were then transfected into 293T cells at a ratio of 5:3:5:1:8 using 1 mg/mL PEI-MAX. Before transfection, 293T cells were rinsed with serum-free DMEM, incubated with Vaccinia vTF7-3 (Imanis Life Sciences, Cat# REA006) at an MOI of 5 for 45 min, and then replaced with fresh medium. Two days post-transfection, the supernatant was harvested and filtered through a 0.22 µm filter to remove cell debris and Vaccinia vTF7-3. The rVSV-ΔG-JN.1 generated from 293T cells was then serially diluted with a dilution factor of 5 and inoculated into Vero-E6-TMPRSS2-T2A-ACE2 cells in 24-well plates for 1 h. The virus was then washed away, and the cells were cultured at 37 °C/5% CO₂ for 16–24 h. Vero-E6-TMPRSS2-T2A-ACE2 cells were monitored, and the virus was harvested from wells in which only one plaque was observed, then filtered and stored at −80 °C. The rVSV-ΔG-JN.1 generated from Vero-E6-TMPRSS2-T2A-ACE2 cells was titrated on Vero-E6-TMPRSS2-T2A-ACE2 cells before use.

To select escape viruses, rVSVΔG-JN.1 was incubated with 0.4 µg/mL of 19-77∆V for 1 h before being added to Vero-E6-TMPRSS2-T2A-ACE2 cells in 24-well plates at an MOI of 0.01. A total of 66 replicates were set up. Two days after coculture, cell supernatants from wells with CPE were harvested, and 100 µL of each supernatant was incubated with 4 µg/mL of 19-77∆V for 1 hour before another round of infection in pre-seeded Vero-E6-TMPRSS2-T2A-ACE2 cells in 24-well plates. Two days later, supernatants containing escape viruses were further harvested and selected the same way using 20 µg/mL, and then 50 µg/mL of 19-77∆V in a stepwise manner. mRNAs of the escape viruses in the supernatants were then extracted using the viral RNA/DNA purification kit (MACHEREY-NAGEL, Cat# 740643) and reverse-transcribed to cDNA using the SuperScript VILO cDNA Synthesis Kit (Invitrogen, Cat# 11754). The RBD genes of the escape viruses were then amplified using primers 5’-GGGCATCTACCAGACCAGCAACTTCA-3’ and 5’-GAACACATTGCTGCCTGTGCTGT-3’ and sequenced.

Antigen-specific memory B cell sorting and single-cell B cell receptor sequencing

Peripheral blood mononuclear cells from Donor 19, and two healthy donors were stained with the LIVE/DEAD Fixable Yellow Dead Cell Stain Kit (Invitrogen) at ambient temperature for 20 min, followed by washing with RPMI 1640 complete medium [RPMI 1640 + 10% fetal bovine serum (FBS) + penicillin/streptomycin (P/S) (100 U/mL)] and incubation with 10 μg/mL XBB.1.5 S2P protein and biotinylated SARS-CoV S2P at 4 °C for 45 min. Afterwards, the cells were washed again and incubated with a cocktail of flow cytometry and Hashtag antibodies, consisting of CD3 PerCP-Cy5.5, CD19 APC/Cyanine 7, CD27 APC, IgM FITC, anti-His PE/DazzleTM 594, Streptavidin BV421, and human Hashtag 3 at 4 °C for 1 h. Stained cells were then washed, resuspended in RPMI 1640 complete medium, and sorted for SARS-CoV and/or XBB.1.5 S2P trimer-specific memory B cells (CD3 − CD19 + CD27+IgM−antigen+ live single lymphocytes) by flow cytometry. The sorted cells were mixed with spike-in CD3+ cells and loaded into a 10X Chromium chip of the 5′ Single Cell Immune Profiling Assay (10X Genomics) at the Columbia University Single-Cell Analysis Core. Library preparation and quality control were performed according to the manufacturer’s protocol and sequenced on a NextSeq 500 sequencer (Illumina).

Identification of spike-specific antibody transcripts

Antibody transcripts specific to the XBB.1.5 S2P spike and SARS-CoV spike trimers were identified following previously established methods³⁸. The assembly of full-length antibody transcripts was performed utilizing the Cell Ranger V(D)J analysis software (version 3.1.0, 10X Genomics), employing default settings with the GRCh38 V(D)J germline sequence version 2.0.0 as the reference genome. To differentiate between cells captured in the antigen-specific sorting process and those added as spike-ins, the count module of Cell Ranger was used to quantify the presence of all hashtag oligonucleotides in each cell based on Next Generation Sequencing (NGS) raw data. Identification of high-confidence antigen-specific cells was achieved using the following criteria: (1) A minimum of 100 copies of the antigen-specific hashtag was required for a cell to be classified as antigen-specific, (2) Given that hashtags might detach from their original cells and attach to others within the sample, a cell was considered truly antigen-specific only if the copy number of its specific hashtag was at least 1.5 times higher than that of any non-specific hashtag present, (3) Cells deemed to be of low quality were excluded based on the cell quality assessment algorithm used by Cell Ranger, (4) Only cells expressing both productive heavy and light chain antibody gene pairs were retained, (5) In instances where a cell exhibited more than two transcripts for heavy and/or light chains, transcripts supported by fewer than three unique molecular identifiers (UMIs) were discarded, and 6) Cells sharing identical heavy and light chain sequences, potentially indicative of mRNA contamination, were consolidated into a single cell entry.

Antibody transcript annotation

Transcripts specific to the antigen were analyzed and annotated with SONAR version 2.0, following previously established procedures. Assignment of V(D)J gene segments to each transcript was conducted via BLASTn, employing specialized parameters against a germline gene repository sourced from the International ImMunoGeneTics (IMGT) information system database. The identification of the complementarity-determining region 3 (CDR3) utilized BLAST alignments of the V and J segments, focusing on the conserved second cysteine within the V segment and the WGXG (for heavy chains) or FGXG (for light chains) motifs in the J segment, with “X” indicating any amino acid. Isotype determination for heavy chain transcripts was achieved by analyzing Constant domain 1 (CH1) sequences against a human CH1 gene database from IMGT, using BLASTn with standard parameters. The CH1 allele presenting the lowest E-value was selected for precise isotype classification, adhering to a BLAST E-value cutoff of 10e-6. Transcripts with incomplete V(D)J segments, frameshifts, or extraneous sequences beyond the V(D)J region were discarded. The filtered transcripts were then aligned to their corresponding germline V gene using CLUSTALO, and levels of somatic hypermutation were quantified through the Sievers method. In instances where cells possessed multiple high-quality heavy or light chains, potentially indicative of doublets, combinations of all H and L chains were generated.

Crystallization and data processing

19-77 Fab was produced by digestion of IgG with immobilized Endoproteinase Lys-C (Sigma-Aldrich) equilibrated with 25 mM Tris pH 8.5 and 1 mM EDTA for 3 h. The resulting Fab was further purified from the cleaved Fc domain by cation exchange chromatography. Fab purity was analyzed by SDS-PAGE and buffer-exchanged into 20 mM Tris, 150 mM, pH 7.4 prior to cryo-EM/Crystallization experiments.

19-77/SARS-CoV-2-RBD and 19-77/EG5.1-CoV-2-RBD complexes were prepared by mixing each of the protein components at an equimolar concentration and incubating overnight at 4 °C. Protein complexes were then isolated by gel filtration on a Superdex 200 column (Cytiva, GE Healthcare). Fractions containing complexes were pooled and concentrated to 12.0 mg/mL in SEC buffer. Screening for initial crystallization conditions was carried out in 96-well sitting drop plates using the vapor-diffusion method with a mosquito crystallization robot (TTP LabTech) using various commercially available crystallization screens: MSCG-1 (Anthracene), Proplex and LMB (Molecular dimensions). Diffraction quality crystals were obtained after 7 days in the following conditions for 19-77/SARS-CoV-2-RBD: 0.1 M NaCl, 0.1 M Tris pH 7.5, 12% w/v PEG 4000, and the following conditions for 19-77/ EG5.1-CoV-2-RBD: 75% MPD and 0.1 M HEPES pH 7.5.

Prior to data collection, crystals were cryoprotected with 40% ethylene glycol supplemented in mother liquor and flash frozen in liquid nitrogen. X-ray diffraction data extending to 2.8 Å (19-77/SARS-CoV-2-RBD) and 3.2 Å (19-77/ EG5.1-CoV-2-RBD) resolution were collected at 100 K on beam line 17-ID-1 (AMX) at Brookhaven National Laboratory. Diffraction data were processed with XDS⁷² and scaled using AIMLESS⁷³ from the CCP4 software suite (Collaborative Computational Project Number 4, 1994)⁷⁴. Molecular replacement was performed with PHASER⁷⁵, using a previously reported RBD structure (PDB 7L5B) and for 19-77 Fab, heavy chain (PDB 7XIK), light chain (3FIK) used as search models. Manual rebuilding of the structure using COOT⁷⁶ was alternated with refinement using Phenix refine⁷⁷. The Molprobity server was used for structure validation⁷⁸ and PyMOL (version 2.1, Schrödinger, LLC) for structure visualization. A summary of the X-ray data collection and refinement statistics are shown in Supplementary Fig. 3c.

Cryo-EM sample preparation

Fab fragments of antibodies were produced by digestion of IgG with immobilized Endoproteinase Lys-C (Sigma-Aldrich) equilibrated with 25 mM Tris, pH 8.5, and 1 mM EDTA for 3 h. The resulting Fabs were purified by ion-exchange chromatography on a mono-Q column.

For the structure of 19-77 bound to D614G spike, the complex was made by mixing purified SARS-CoV-2 S2P D614G spike protein with Fab in a 1:3 molar ratio (spike potomer:Fab) in PBS, pH 7.4, such that the final concentration of spike was 1 mg/mL. This mixture was incubated on ice for 1 h.

For the structures of the ternary complex of 19-77, RBDs, and S309, complexes were made by mixing purified SARS-CoV-2 RBD with Fabs in a 1:1.2 molar ratio in PBS, pH 7.4, and further purified using size exclusion chromatography (SEC) using a Superdex 200 Increase column. The resulting complex peak was then concentrated to 4 mg/mL and held on ice until vitrification.

Before freezing, 0.005% (w/v) n-dodecyl β-D-maltoside (DDM) was added to deter preferred orientation and aggregation during vitrification. Cryo-EM grids were prepared by applying 3 µL of sample to a freshly glow-discharged carbon-coated copper grid (CF 1.2/1.3 300 mesh); the sample was then vitrified in liquid ethane using a Vitrobot Mark IV with a wait time of 30 s, a blot time of 3 s, and a blot force of 0.

Cryo-EM data collection and analysis

Cryo-EM data for single particle analysis were collected at the Columbia Cryo-EM Facility on a Titan Krios electron microscope operating at 300 kV, equipped with a Gatan K3-BioQuantum detector and energy filter, using the Leginon7 software package. Exposures were taken at a magnification of 105,000x (pixel size of 0.83 Å), using a total electron flux of 58 e-/Å2 fractionated over 50 frames with an exposure time of 2.5 s. A random defocus range of −0.8 to −2.0 µm was used.

Data processing was performed using cryoSPARC v3.3.1.8. Raw movies were aligned and dose-weighted using patch motion correction, and the micrograph contrast transfer function (CTF) parameters were estimated using patch CTF estimation. Micrographs were picked using a blob or template picker, and an initial particle set was selected using 2D classification. Further heterogeneous refinement was used to 3D-classify particles and remove debris. The resulting curated particle sets were corrected for local motion and refined using homogenous refinement. Local refinement was performed for the spike dataset using a mask enveloping the RBD+Fab variable region. The default cryoSPARC auto-sharpened maps were then used to build the models. Cryo-EM data collection and consensus refinements are summarized in Supplementary Fig. 3.

Model building and refinement

Initial molecular models for Fabs were generated using Alphafold Multimer⁷⁹ using paired heavy and light sequences. An RBD from PDB 7KNI, an RBD-up spike structure, was used as a starting model. For the initial 19-77 structure, the 14-7 structure (PDB 8F89), RBD (PDB 8IOV), and S309 Fab (PDB 7XSW) were used. The initial models were rigid body docked into the density map using UCSF Chimera’s “fit to map” tool and combined. The Fab CDR loops were manually fitted to the density map using Coot real-space refinement. The models were fit to density using the ISOLDE package in ChimeraX⁸⁰. Ramachandran outliers were corrected using ISOLDE’s “flip peptide bond” feature. Real-space refinement in Phenix⁸¹ was performed to remove geometry outliers. The remaining manual adjustments were performed in Coot. Models were validated using MolProbity16 in Phenix and the PDB validation server and deposited to the PDB with accession codes: 19-77 + SARS-CoV-2 D614G RBD (PDB 9CFE), 19-77 + HK.3 RBD (PDB 9CFF), 19-77∆V + HK.3 RBD (PDB 9CFG), and 19-77∆V + JD.1.1 RBD (PDB 9CFH). A summary of data collection, processing, and model refinement statistics is shown in Supplementary Fig. 3.

In silico antibody engineering

The energy changes for the binding of both the mutant and original 19-77 to the HK.3 RBD were calculated using FoldX software⁴⁰. Initially, the 19-77 and HK.3 complex was repaired and optimized using the “Optimize” function. Subsequently, every position in the heavy and light chains was subjected to saturation mutagenesis to all other amino acids using the “BuildModel” function. This process generated both an unmutated model and a mutant model for each mutation. The binding energies were estimated with the “AnalyseComplex” function, and the changes in Gibbs free energy (ΔΔG) were calculated based on the energy difference between each mutant and its corresponding original antibody-RBD complex.

Molecular dynamics analysis

Antibody 19-77, along with other VH3-53/66 class antibodies, was meticulously aligned through the RBD within each complex to ensure structural consistency. The R71A mutant antibodies were precisely engineered using the “Mutagenesis” function in PyMOL version 2.5.4, provided by Schrödinger, LLC. Subsequently, the Fab regions of both the original and mutant antibodies underwent molecular dynamics simulations employing GROMACS on the WebGro server. The preprocessing was executed using the GROMOS96 54a7 force field and incorporated an SPC water model within a cubic simulation box, supplemented with 0.15 M NaCl to mimic physiological ionic strength. The initial energy minimization step was set to 5000 iterations to stabilize the system. This was followed by equilibration and simulation phases under NVT/NPT conditions, conducted at a physiological temperature of 310 K for 10 ns, with all other parameters set to default. To ensure the reliability of the simulation, the root mean square deviation (RMSD) of each system was manually monitored to verify that each simulation reached equilibrium. Additionally, the per-residue root-mean-square fluctuation (RMSF) analysis was conducted to assess the flexibility of residues within each Fab region.

Structure comparison analysis

The paratope and epitope residues for 19-77 was identified using PISA with the default parameters. The gene-specific substitution profiles (GSSP) for 19-77 germline genes were obtained from the cAb-Rep database (https://cab-rep.c2b2.columbia.edu/). The 19-77 and its mutant complexes were first superimposed by using the “align” function in PyMOL 2.5.4. The distance between the Cα from identical residues within the 19-77 heavy chains were then determined using the rms_cur function in PyMOL by in-house Python script.

Data analysis

IC₅₀ values were determined by fitting the data to five-parameter dose-response curves in GraphPad Prism v.10.3. GraphPad Prism v.10.3 was also used for data visualization. PISA was used for identifying antibody-spike interface residuals. PyMOL V.2.5.4 was used to perform mutagenesis and to generate structural plots.

Reporting summary

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