Research participants
The Institutional Review Board of Vanderbilt University Medical Center approved human studies (protocol number 8675). All participants gave written informed consent before study participation. Participants were offered compensation for study participation. All ethics regulations for human participant studies were followed. Demographic data and exposure histories for the participants profiled in this manuscript are available in Extended Data Table 2. D2102 was a 28-year-old otherwise healthy female in the United States with a previous history of three mRNA vaccinations encoding the ancestral SARS-CoV-2 sequence (Wuhan-Hu-1). This individual became infected in December 2021 with SARS-CoV-2 at a time when BA.1 and BA.1.1 variants dominated local circulation. Five days after exposure, the individual had upper respiratory illness symptoms and a positive PCR test on nasal secretions for the presence of SARS-CoV-2, suffered a mild clinical illness that resolved on the sixth day, and fully recovered with only supportive therapy. Two months later, a sample of peripheral blood was obtained by phlebotomy after written informed consent and processed for PBMC isolation. This individual subsequently underwent leukapheresis, and PBMCs were purified using a negative-selection-based magnetic enrichment kit (StemCELL Technologies, 19654). After isolation, PBMCs were cryopreserved and stored in a liquid nitrogen freezer until use. For participants D1672, D2015, D2098 and D2170, PBMCs were isolated from peripheral blood draws using the same kit.
Cell lines
Expi293F cells (Thermo Fisher, A1452) were propagated in Expi293F Expression Medium (Thermo Fisher, A1435102) at 37 °C and 8% CO2. A previously described engineered NIH3T3 fibroblast line (originating from a male mouse) constitutively expressing cell-surface human CD40L, BAFF and IL-21 (ref. 24) was kindly provided by D. Bhattacharya (University of Arizona, Tuscon, AZ). HEK-293T/17 cells were obtained from The American Type Culture Collection (ATCC, CRL-11268). HEK-293T cells stably transduced to express human ACE2 (293T-hACE2 cells) or 293T cells stably expressing both hACE2 and TMPRSS2 (293T-hACE2-TMPRSS2) were obtained from BEI Resources (NR-52511, NR-55293). Vero-TMPRSS2 cells36 were maintained at 37 °C in 5% CO2 in 1× Dulbecco’s modified Eagle medium (DMEM, Gibco, 11995-040) supplemented with fetal bovine serum (FBS, 10% v/v), HEPES (10 mM), sodium pyruvate (1 mM), non-essential amino acids, blasticidin (5 µg ml−1) and penicillin–streptomycin (100 U ml−1). A PCR-based mycoplasma detection kit (ATCC, 30–1012K) was used to test cell cultures on a monthly basis and all tests were negative during the time of study.
Viruses
The rVSV-SARS-CoV virus was a generous gift from S. Whelan (Washington University, St Louis). The WA1/2020 recombinant strain containing the D614G substitution was described previously37,38. The SARS-CoV-2 variant isolates XBB.1.5, EG.5.1 and BQ.1.1 were generous gifts from Y. Kawaoka (University of Wisconsin-Madison), M. Suthar (Emory University) and A. Pekosz (Johns Hopkins University), respectively. All authentic SARS-CoV-2 isolates were expanded once on Vero-TMPRSS2 cells and next-generation sequencing39 was used to confirm expected sequences. All experiments involving authentic SARS-CoV-2 were approved by the institutional biosafety committees of either Washington University, St Louis (protocol number 14366) or UNC-Chapel Hill (protocol numbers 202517662, 202517170). Studies were conducted in certified biosafety level 3 (BSL-3) facilities using positive-pressure respirators.
Recombinant antigen expression and purification
To express the RBD portion of the SARS-CoV-2 S protein, a synthesized cDNA encoding residues 328–531 was inserted into a mammalian expression vector downstream of a phosphatase-mu signal peptide and including C-terminal AviTag and 8×His tags (COV2-RBD_His). Some constructs contained an additional FLAG or myc tag, although these were not used and did not affect protein expression or purification. Three previously identified stabilizing mutations (Y365F, F392W, V395I)40 were introduced to improve yield and protein stability. Mutations corresponding to each Omicron subvariant were introduced in the context of these three stabilizing mutations. Expi293F cells were transfected with plasmids encoding each RBD, and expressed proteins were purified by affinity chromatography with HisTrap Excel or TALON HP columns (Cytiva). For crystallographic and binding studies, we also expressed RBD constructs containing an IL-2 signal peptide (MYRMQLLSCIALSLALVTNS), residues 319–528 of the RBD, a Prescission 3C protease cleavage site, an AviTag, a tobacco etch virus (TEV) protease cleavage site, a TwinStrep tag and an 8×His tag (COV2-RBD_Strep). We also introduced Y365F, F392W and V395 mutations to enhance stability and yield. We expressed these COV2-RBD_Strep constructs in Expi293F cells, purified using StrepTrap XT columns (Cytiva) and eluted with 100 mM biotin in 100 mM Tris-HCl pH 8 + 150 mM NaCl. Recombinant protein size, glycosylation state and purity were assessed by SDS–PAGE. For BA.1 constructs where the 444 glycan was introduced, site-directed mutagenesis was used to introduce K444N and S446T mutations, on the basis of previous observations that sequons containing threonine promote more efficient glycosylation than those containing serine41.
To express SARS-CoV-2 S proteins for ELISA binding and electron microscopy (EM) studies, we introduced the mutations of the BA.2, XBB and BQ.1.1 variants into the background of a previously described stabilized S protein construct (VFLIP) that was originally based on the Wuhan-Hu-1 sequence42. This construct contains a C-terminal T4 fibritin foldon domain, an 8×His tag and a TwinStrep tag. For expression and purification of recombinant soluble SARS-CoV S protein, we used the previously described 2P prefusion-stabilizing substitutions, and this construct also contained a C-terminal T4 fibritin foldon domain, an 8×His tag and a TwinStrep tag43. To express proteins, plasmids encoding each prefusion-stabilized S protein construct were transfected into Expi293F cells (Thermo Fisher), and at 4–5 days post transfection, culture supernatants were clarified by centrifugation and sterile filtered using a 0.2 µm filter after the addition of BioLock (IBA Biosciences). Recombinant S protein ectodomains were purified by streptactin-based affinity chromatography using StrepTrap HP or StrepTrap XT columns (Cytiva) and eluted with 25 mM desthiobiotin in 1× Dulbecco’s phosphate buffered saline (DPBS) (for StrepTrap HP) or 100 mM biotin in 100 mM Tris-HCl pH 8 + 150 mM NaCl (for StrepTrap XT). For COV2-3835 structural studies, SARS-CoV-2 Omicron BA.1 mutations were expressed in the background of Hexapro (6P), a trimer construct containing 6 previously described stabilizing mutations, and purified as previously described44. Sequences for antigen constructs used for binding assays, antigen-specific sorting and structural studies are included in Supplementary Table 8.
Modelling of native and engineered SARS-CoV-2 RBD glycans
We used a glycan modelling tool to visualize the location and relative size of RBD glycans45. To design the construct, we uploaded a crystal structure of the RBD (PDB: 6M0J) to the GlycoSHIELD web application (http://www.glycoshield.eu/). We selected a high-mannose N-linked glycan (Man9) and used the GlycoSHIELD interface to simulate conformations of the glycans and graft each glycan at positions 343 (the site of a native N-linked glycan in RBD) and 444 (the location of our engineered glycan mask). We used default grafting settings with 30 glycan conformers for display on the structure. Structures were then visualized using PyMOL v.2.5.7 (Schrödinger).
Biotinylation of SARS-CoV-2 RBD antigens and tetramer preparation
Site-specific biotinylation of COV2-RBD_His antigens was performed through labelling of the AviTag amino acid sequence with biotin by BirA according to manufacturer protocol (Avidity, BirA500). Briefly, 40 μM of AviTag-containing SARS-CoV-2 RBD, 2.5 μg of BirA biotin-protein ligase, 1× BiomixA and 1× BiomixB were mixed and incubated at room temperature for 1 h to ensure rapid biotinylation. Biotinylation of RBDs was confirmed by ELISA using avidin conjugated to horseradish peroxidase (HRP) (Thermo Fisher, 434423) and TMB substrate (Thermo Fisher, 34029). Colorimetric signal was monitored and the reaction was stopped with 1 M hydrochloric acid. Absorbance then was measured at 450 nm using a plate reader (Biotek). For preparation of tetramers, biotinylated SARS-CoV-2 RBDs were incubated with fluorophore-conjugated streptavidin. Briefly, streptavidin-PE (Thermo Fisher, SA10041) was added to biotinylated XBB or BQ.1.1 RBDs, and streptavidin-Alexa Fluor 647 (AF647; Thermo Fisher, S21374) was added to biotinylated BA.1-444glyc RBD at a molar ratio of 4:1. Fluorophore-conjugated streptavidin was added stepwise over the course of 10 different steps, with a brief mix in between steps to maximize tetramer formation. Each addition of streptavidin to the RBDs was followed by incubation at room temperature for 10 min. Reactions were then incubated overnight at 4 °C to complete tetramer formation.
B cell enrichment and flow cytometric cell sorting
Cryopreserved PBMCs (1 × 108 total cells) were thawed in a 37 °C water bath and immediately mixed with cold RoboSep buffer (StemCELL Technologies, 20104). After a brief centrifugation (250 × g, 5 min) at room temperature, the cell pellet was resuspended in cold RoboSep buffer. B cells were enriched using a negative-selection magnetic bead-based enrichment kit (EasySep Human B Cell Isolation kit, StemCELL Technologies, 17954) according to manufacturer protocol. After washing with cold RoboSep buffer (250 × g, 5 min) at room temperature, the isolated B cells were incubated with a cocktail of phenotyping antibodies, each at a 1:20 dilution, for 45 min on ice. The phenotyping antibodies included: anti-human CD19–Brilliant Violet 421 (BioLegend, clone HIB19, 302234), anti-human IgD–FITC (BioLegend, clone IA6–2, 348206) and anti-human IgM–FITC (BioLegend, clone MHM-88, 314506). The isolated B cells were then centrifuged briefly, washed with RoboSep buffer (250 × g, 3 min) and incubated with SARS-CoV-2 RBD tetramers, each at a final concentration of 1 µg ml−1, on ice for 45 min. Two separate reactions were performed using either the XBB or BQ.1.1 RBD tetramers-PE and BA.1-444glyc RBD tetramers-AF647. Cells were then washed briefly with RoboSep buffer (250 × g, 3 min) and resuspended in 500 µl of RoboSep buffer for flow cytometric analysis using an SH800 cell sorter (Sony Biotechnology). Flow cytometric data were analysed with the SH800 software (Sony Biotechnology) and FlowJo v.10 (Tree Star). XBB or BQ.1.1 RBD-reactive and BA.1-444-glyc RBD-non-reactive B cells (1,345 XBB- and 462 BQ.1.1-reactive) were sorted into Medium A (StemCELL Technologies, 03801, 1000041736) containing penicillin and streptomycin. Cells were then expanded for 8 days in the presence of CpG and irradiated 3T3 feeder cells expressing human CD40L, IL-21 and BAFF, as previously described24. Expanded ASCs were screened and confirmed by ELISA for secretion of SARS-CoV-2 RBD-specific antibodies. Expanded ASCs were separated from irradiated 3T3 feeder cells through flow cytometric cell sorting. For antibody sequence recovery from participant D2102, 40,000 expanded ASCs were stained with anti-human CD45-PE (BioLegend, 368509). After, flow cytometric sorting, ~25,000 cells were prepared for loading onto the OptoSelect 11k chip of a Beacon optofluidic system (Bruker Cellular Analysis, formerly Berkeley Lights) for single-cell analysis. The remaining cells were sequenced using the 10x Genomics Chromium sequencing method to generate paired antibody–variable gene libraries. For antibody sequence recovery from participants D1672 and D2105, ASCs were sorted into 96-well PCR plates containing lysis buffer.
Single-cell optofluidic assay selection of SARS-CoV-2 RBD-reactive B cells
Expanded ASCs were screened using a Beacon optofluidic system (Bruker Cellular Analysis). First, cells were loaded onto OptoSelect 11k chips in a plasmablast survival medium, which promotes antibody secretion and maintains cell viability46. Next, thousands of ASCs were transferred into individual nl-volume chambers (NanoPens) across the chip using opto-electropositioning (OEP). To screen ASCs for binding reactivity, 6–8-µm (BLI assay beads, 520-00053) and 10–14-µm (Spherotech, SVP-100-4) beads were prepared by coupling biotinylated XBB or BQ.1.1 RBDs to streptavidin-coated polystyrene particles. Conjugated beads were prepared at final concentrations of 5% (w/v) for 6–8 μm beads and 0.5% (w/v) for 10–14 μm beads. These conjugated beads were mixed with AF568-labelled anti-human IgG (H + L) cross-adsorbed secondary antibodies (Thermo Fisher, A-21090) at a 1:2,500 dilution for detection of secreted RBD-reactive antibodies. This mixture was imported onto the OptoSelect 11k chips for an on-chip, fluorescence-based assay. In this assay, positive SARS-CoV-2 RBD binding reactivity was detected through antibody binding to the conjugated beads and sequestration of fluorescent signal (AF568) from the secondary antibodies. Fluorescent signal on beads adjacent to individual NanoPens was used to identify B cells secreting XBB or BQ.1.1 RBD-reactive antibodies. An on-chip in-pen assay was also performed to select for antibodies that blocked hACE2 binding to either XBB or BQ.1.1 RBDs. In this surrogate hACE2 blocking assay, 10–14-µm XBB or BQ.1.1 RBD-conjugated streptavidin-coated beads (Spherotech) were loaded into NanoPens containing individual ASCs and incubated to allow saturation of the RBD with secreted antibodies. Then, a mixture containing recombinant hACE2 with a FLAG tag (10 µg ml−1, Sigma-Aldrich, SAE0064), a rat anti-FLAG tag AF647 antibody (1:100 dilution, BioLegend, clone L5, 637315, B265929) and anti-human IgG (H + L) cross-adsorbed AF568 antibodies (1:200 dilution, Thermo Fisher, A-21090) was perfused throughout the OptoSelect 11k chip for diffusion into the NanoPen chambers. Cells secreting antigen-reactive antibodies were identified by fluorescent signal (AF568) from XBB or BQ.1.1 RBD-conjugated beads using the Beacon TRED filter cube. Simultaneously, AF647 signal was detected using a Cy5 filter cube as a measure of hACE2 binding. NanoPen chambers that contained fluorescent XBB or BQ.1.1 RBD-conjugated beads in both fluorescence channels were considered to contain B cells secreting XBB or BQ.1.1 RBD-reactive antibodies that were unable to block hACE2 binding at the concentrations tested. In contrast, NanoPens that contained fluorescent XBB or BQ.1.1 RBD-conjugated beads in the TRED channel but not in the Cy5 channel were classified to contain B cells secreting XBB-RBD-reactive and hACE2-blocking antibodies. Cells with activities of interest were exported from specific NanoPens by OEP into lysis buffer in individual wells of 96-well RT–PCR plates for antibody sequencing.
Sequencing of B cells following optofluidic functional screening or single-cell sorting
After export of D2102 ASCs that bound to XBB or BQ.1.1 from the Beacon instrument, components of the Opto Plasma B Discovery cDNA Synthesis kit (Berkeley Lights, 750-02030) were used to amplify and recover antibody heavy- and light-chain sequences. The Opto Plasma B Discovery Sanger Prep kit (Berkeley Lights, 750-02041) was used to perform further rounds of sequence amplification, and amplicons were sequenced using Sanger sequencing (Azenta Life Science). This same amplification and sequencing workflow was used to recover the heavy- and light-chain sequences of ASCs from D1672 and D2015. Sequence data were analysed using a Python-based antibody variable-gene analysis tool (PyIR; https://github.com/crowelab/PyIR)47 to control the quality of the sequences selected. Subsequent steps were performed only for those clones for which full-length, unambiguous, paired heavy- and light-chain variable-gene sequences were obtained.
Chromium sequencing of single B cells through 10x Genomics
Expanded ASC populations derived from XBB- or BQ.1.1-binding B cells were sequenced using the Chromium single-cell sequencing platform following vendor protocol (10x Genomics). Briefly, ~40,000 XBB-reactive or 20,000 BQ.1.1-reactive ASCs were split evenly into replicates of 10,000 cells each and separately added to a reverse-transcription reaction mix (following vendor protocol), which was then loaded directly onto a Chromium chip for cell and barcoded bead encapsulation (10x Genomics, PN-1000263). Following reverse transcription, sequencing libraries were prepared following the User Guide for Chromium NextGEM Single Cell 5′ Reagent Kits v2 (CG000331 Rev C). Briefly, total cDNA was amplified, after which BCR sequences were amplified using the Chromium Single Cell Human BCR Amplification kit (10x Genomics, PN-1000253) according to the User Guide for Chromium Single Cell 5′ Reagent Kits v2 (CG000331 Rev C). The enriched libraries were sequenced using a NovaSeq 6000 S4 Reagent kit (Illumina) on a NovaSeq sequencer for 300 cycles. As suggested by the Chromium User Guide, a sequencing depth of greater than 5,000 raw reads was selected for each sample on the basis of input cells.
Bioinformatic analysis of antibody–variable gene libraries
Following next-generation sequencing, all samples were demultiplexed and the 10x Genomics Cell Ranger software (v.6.1.2) was used to process FASTQ files. Following the Cell Ranger bioinformatic processing, we collected all the heavy- and light-chain paired sequences for which there was a 1:1 pairing and processed those sequences using PyIR47. We next excluded sequences from downstream processing if they: (1) contained a stop codon, (2) did not encode an intact heavy- or light-chain CDR3 and (3) did not contain an in-frame junctional region. In a second phase of processing, we deduplicated sequences to exclude those with identical amino acid sequences. Paired antibody sequences that were assigned as IgM on the basis of Cell Ranger analysis were not considered. PyIR was used to determine germline gene usage, define CDRs and identify somatic mutations relative to germline sequences for each paired antibody sequence.
Bioinformatic downselection of antibody panel using paired sequence clustering
Paired, antigen-specific antibody sequences were clustered on the basis of genetic similarity to identify clonal families. Sequences were first binned together if they were encoded by the same inferred heavy-chain V and J gene and had the same HCDR3 length. Next, sequences were clustered according to 80% identity on the HCDR3 nucleotide sequence using the single-linkage clustering algorithm48, implemented by SciPy49. Then, the antibody sequences were grouped together on the basis of shared V and J gene and LCDR3 length. Finally, sequences were again clustered according to 80% identity on the LCDR3 nucleotide sequence using the single-linkage algorithm. We defined these clusters of sequences as clonal families. The most somatically mutated sequence from each clonal family, as determined by PyIR, was selected for synthesis and expression. If there were multiple members of a clonal family, the sequence closest to the consensus sequence was also selected for synthesis and expression.
Microscale and ‘midi-scale’ expression of recombinant mAbs and Fabs
To express a large panel of ~300 mAbs, a microscale expression platform was used. Briefly, ~1 ml per well of ExpiCHO cell cultures was transfected using the Gibco ExpiCHO expression system in deep 96-well plates (Thermo Fisher), as previously described25,26. For high-throughput microscale mAb purification, supernatants were clarified by centrifugation and incubated with MabSelect SuRe resin (Cytiva). The resin was then washed with 1× DPBS and filtered deionized water, followed by elution and neutralization with 0.1 M sodium acetate pH 3.3 buffer and 5 M Tris-HCl pH 8.0, respectively. Eluted mAbs were then buffer exchanged into 1× DPBS using Zeba spin desalting plates (Thermo Fisher) and stored at 4 °C until functional analyses were performed. To generate larger amounts of recombinant mAbs, ‘midi-scale’ expressions were performed. In this case, ~15–30 ml ExpiCHO cell cultures were transfected using the Gibco ExpiCHO expression system as described by vendor protocol. For high-throughput ‘midi-scale’ mAb purification, mAbs were purified from clarified cell culture supernatants using HiTrap MabSelect SuRe (Cytiva) columns on a 24-column parallel protein chromatography system (Protein BioSolutions). Following purification, mAbs were buffer exchanged into 1× DPBS, concentrated using Amicon Ultra 50-kDa-cutoff centrifugal filter units (Millipore Sigma) and stored at 4 °C until use. For recombinant Fab production for high-resolution structural studies, constructs encoding the heavy chain variable region and CH1 domain, and the light chain variable and constant regions were transfected into ExpiCHO cells using the same procedure as for recombinant mAb production. Recombinant Fabs were purified from culture supernatant using an anti-CH1 CaptureSelect column (Thermo Fisher).
High-throughput mAb quantification
High-throughput quantification of recombinantly expressed mAbs was performed using purified mAbs in a 96-well plate. The Cy-Clone plus kit was used according to vendor protocol and analysis performed on an iQue Plus Screener flow cytometer (Sartorius). Purified mAbs were measured at a single dilution (final dilution of 1:10, 2 μl of mAb per reaction). The concentration of mAb was interpolated on the basis of the relative competition signal relative to a standard curve of a control human IgG. Data were analysed using the ForeCyt software v.6.2 (Sartorius).
ELISA dose–response binding assays
Microtitre plates (384-well) were coated with purified recombinant SARS-CoV-2 RBD proteins (diluted 2 µg ml−1 in 1× DPBS) in a volume of 25 µl well−1 at 4 °C overnight. The next day, plates were washed with 1× DPBS containing 0.05% Tween-20 (DPBS-T) and blocked with blocking buffer (2% (w/v) non-fat dry milk and 2% (v/v) normal goat serum in DPBS-T) for 1 h at room temperature. Each mAb was diluted in blocking buffer at a starting concentration of 10 µg ml−1 and serially diluted 3-fold for a 12-point dilution series. Plates were then washed and 25 µl of each mAb dilution were added to each well and incubated for 1 h at room temperature. Plates were washed and 25 µl of blocking buffer and goat anti-human IgG secondary antibody conjugated with HRP (Southern Biotech, 2014-05, L2118-VG00B, 1:5,000 dilution in blocking buffer) were added to each well. Plates were then incubated for 1 h at room temperature. After plates were washed, 25 µl of 3,3′,5,5′-tetramethylbenzidine (TMB) substrate (Thermo Fisher) was added and the plates incubated at room temperature to develop the signal. Following sufficient signal development, the reaction was stopped by addition of 25 µl 1 M hydrochloric acid and absorbance was measured at 450 nm using a spectrophotometer (Biotek). ELISA dose–response binding assays were performed with technical triplicates in at least 2 independent experimental replicates.
SARS-CoV-2 S-pseudotyped lentivirus generation and titration
Single-cycle lentiviruses pseudotyped with the S glycoproteins of SARS-CoV-2 variants were generated using a previously described protocol50. Briefly, HEK-293T/17 cells (ATCC, CRL-11268) were seeded in a T-225 cm2 culture flask targeting 50–70% confluency the following day. The morning of the transfection, the medium was changed to fresh DMEM supplemented with sodium pyruvate (Thermo Fisher, 11995073), Ultra-Low IgG FBS (10% v/v, Gibco, 16250078) (DMEM + 10%), HEPES (25 mM, Gibco, 15630-080) and penicillin/streptomycin (Gibco, 15140122). To generate lentiviral-based reporter pseudotyped viruses, we used components of a SARS-CoV-2 lentiviral pseudotyping kit (BEI NR-53817). Briefly, 22.5 µg of the lentiviral genome plasmid pHAGE-CMV-Luc2-IRES-ZsGreen-W (BEI Resources, NR-52516), 4.95 µg each of the packaging plasmids pRC-CMV- Rev (BEI Resources, NR-52519), HDM-Hgpm2 (BEI Resources, NR-52517) and HDM-tat1b (BEI Resources, NR-52518), and 7.65 µg of a CMV-driven plasmid encoding a codon-optimized SARS-CoV-2 S variant gene with a 21-amino-acid deletion were added to 1 ml of serum-free DMEM. After mixing, 45 µl of BioT transfection reagent (Bioland Scientific, B01-00) was added and the plasmid:transfection reagent mixture was mixed gently and incubated at room temperature for 5 min. Following incubation, the transfection mixture was added dropwise to the flask while swirling gently. Approximately 16 to 18 h later, the medium was removed and fresh DMEM supplemented with sodium pyruvate (Thermo Fisher, 11995073), penicillin/streptomycin (Gibco, 15140122), 25 mM HEPES (Gibco, 15630-080) and 2% (v/v) Ultra-Low IgG FBS (Gibco, 16250078) (DMEM + 2%). Supernatants were collected ~48 h after transfection, centrifuged to remove cells and filter sterilized to further remove cellular debris. Single-use aliquots of pseudotyped virus stocks were prepared and stored at −80 °C.
Lentiviral pseudotyped virus neutralization assays
Lentivirus-based pseudotyped viral neutralization assays were performed according to a previously described protocol50. One day before, poly-D-lysine-coated 96-well tissue culture plates (Thermo Fisher, A3890401) were seeded with HEK-293T cells stably expressing human ACE2 (293T-hACE2 cells, BEI Resources NR-52511) at a density of 1.25 × 104 cells per well in DMEM + 10%. The following day, 4-fold serial dilutions of mAbs were prepared in DMEM + 2% in a 96-well polypropylene microtitre plate. mAbs dilutions were then mixed with pseudotyped virus for 1 h at 37 °C. Polybrene (5 µg ml−1, EMD Millipore) was present to enhance pseudovirus infection. After the incubation, pseudotyped virus–mAb mixtures were added to 293T-hACE2 cell monolayers. Plates were incubated at 37 °C for 48–60 h, at which point cells were lysed using the Bright-Glo Luciferase Assay System (Promega). Luciferase activity was then quantified using a CLARIOStar plate reader (BMG LabTech). Wells in which neither pseudotyped virus nor mAb was added were used to determine the average background luminescence signal from each well, which was then subtracted from readings. The percent infection of each well was then determined relative to the average signal from wells in which only pseudotyped virus was added. Four-parameter (inhibitor) vs response curves were fit to the data using nonlinear regression in Prism v.9.5 (GraphPad). IC50 values were defined by constraining the top or bottom values of curve fits to 100 or 0, respectively. Each neutralization assay was executed in technical duplicate for each mAb and performed in at least two independent experiments. For neutralization assays with KP.3, 293T-hACE2-TMPRSS2 cells were used instead of 293T-hACE2 cells, with no other alterations to the assay protocol.
rVSV-SARS-CoV virus neutralization assay
To screen for the neutralizing activity of SARS-CoV S (S2P)-reactive mAbs, we used a previously described impedance-based real-time cellular analysis (RTCA) assay and an xCelligence RTCA HT Analyzer (Agilent). RTCA measures changes in electrical impedance that are associated with changes in cell physiology, including the virus-induced cytopathic effect (CPE)51. Briefly, 50 µl of cell culture medium (DMEM + 2%) was added to each well of a 96-well plate containing gold electrodes (E-plate, Agilent) and the plate was measured to establish background impedance. Vero CCL81 cells were then added to each well (18,000 cells per well in a volume of 50 μl DMEM + 2%) and the plate was placed back on the analyser. At 17–20 h after seeding, 60 µl of a recombinant VSV expressing the S protein of SARS-CoV (rVSV-SARS-CoV, 5,000 p.f.u.s) was mixed with 60 µl of 3-fold serially diluted antibodies to make a total volume of 120 μl using DMEM + 2% diluent and incubated for 1 h at 37 °C in 5% CO2. After incubation, the virus–mAb mixtures were added to 96-well E-plates without removal of culture medium. Wells containing virus only (in the absence of mAb or negative control mAb) and wells containing only Vero CCL81 cells in the medium were included as controls. Plates were measured for 68–72 h to assess virus neutralization. Sensograms were visualized using RTCA HT software v.1.0.1 (ACEA Biosciences). Antibodies were assessed in technical duplicates at a starting concentration of 10 µg ml−1. Neutralization was calculated as a percentage of cell index in control wells where no virus was added after subtraction of the cell index of virus-only wells that exhibited maximal CPE ~40–48 h after addition to cells.
Authentic virus neutralization assays
Serial dilutions of mAbs at a starting concentration of 10 μg ml−1 were incubated with 102 focus-forming units (f.f.u.s) of WA1/2020 D614G, XBB.1.5, EG.5.1 or BQ.1.1 for 1 h at 37 °C. Antibody–virus mixtures were added to Vero-TMPRSS2 cells seeded the day before in 96-well plates. Antibody–virus mixtures were incubated with cells at 37 °C for 1 h to allow virus internalization. Next, an overlay of 1% (w/v) methylcellulose in MEM was added to cells. Methylcellulose overlays were removed 30 h (WA1/2020 D614G) or 72 h (XBB.1.5, EG.5.1, BQ.1.1) later and cell monolayers were fixed with paraformaldehyde (4% in PBS, v/v) for 20 min at room temperature (r.t). Plates were then washed and incubated with a pool of anti-S antibodies of mouse origin52 (SARS2-02, -08, -09, -10, -11, -13, -14, -17, -20, -26, -27, -28, -31, -38, -41, -42, -44, -49, -57, -2, -64, -65, -67 and -71). This pool included antibodies with cross-reactivity to SARS-CoV. After washing, an HRP-conjugated goat anti-mouse IgG secondary antibody (Sigma, A8924) was added to detect bound anti-S antibody. Staining steps were performed in PBS supplemented with 0.1% saponin and 0.1% bovine serum albumin. After washing, SARS-CoV-2-infected cell foci were detected by addition of TrueBlue peroxidase substrate (KPL). Foci were imaged and quantitated on an ImmunoSpot microanalyzer (Cellular Technologies).
Competition-binding ELISA
Wells of 384-well microtitre plates were coated with 1 μg ml−1 of purified SARS-CoV-2 BA.2 VFLIP ectodomain protein at 4 °C overnight. Plates were blocked with blocking buffer (2% BSA in DPBS-T) for 1 h at r.t. Unlabelled mAbs were diluted 10-fold in blocking buffer and added to wells (20 μl per well). The plates were then incubated for 1 h at room temperature. Biotinylated preparations of recombinantly expressed reference mAbs rLY-CoV1404, rS309 or rCR3022 were then added to each respective mAb at 2.5 μg ml−1 in a volume of 5 μl per well (final concentration of 0.5 μg ml−1) without previous washing of the unlabelled mAbs. Plates were then incubated for 1 h at r.t. Plates were washed with DPBS-T and incubated with HRP-conjugated avidin (Sigma-Aldrich, A3151) for 1 h at r.t. Bound mAbs then were detected by addition of 25 μl of a TMB substrate and the reaction was stopped by addition of 25 μl 1 M HCl. Background signal was subtracted, and binding signal was normalized to the binding of each biotinylated reference mAb in the absence of competing mAbs. The following criteria were used to determine competition with the reference mAbs: <33% of the maximal binding signal of the reference mAb indicates full competition, 33–67% indicates partial competition, and >67% indicates no competition.
hACE2 competition-binding ELISA
Assays to measure the ability of mAbs to compete with hACE2 for binding to S were completed as previously described18. Briefly, 384-well microtitre ELISA plates were coated with 2 μg ml−1 purified recombinant SARS-CoV2-S_VFLIP proteins in a total volume of 25 μl at 4 °C overnight. The following day, plates were washed with DPBS-T and blocked with 2% non-fat dry milk and 2% normal goat serum in DPBS-T (blocking buffer) for 1 h at r.t. After blocking, plates were washed with DPBS-T and 2-fold serial dilutions of mAbs at a starting concentration of 10 μg ml−1 were added to the wells (20 μl well−1 in blocking buffer) and incubated for 1 h at r.t. Recombinant hACE2 with a C-terminal FLAG tag (Sigma-Aldrich, SAE0064) was added to wells at 2 μg ml−1 in a 5 μl volume of blocking buffer (final 0.4 μg ml−1 concentration of hACE2 following addition to each well) without washing. Plates were incubated for 40 min at ambient temperature. Plates were then washed with DPBS-T and an HRP-conjugated anti-FLAG antibody (Sigma-Aldrich, A8592, 1:5,000 dilution in blocking buffer) was added to detect bound hACE2. After 1 h incubation at r.t, plates were washed with DPBS-T and signal was developed by addition of 25 µl TMB substrate followed by 25 µl 1 M HCl. ACE2 binding without competing antibody served as a control. The signal obtained for hACE2 binding at each dilution of tested antibody was expressed as a percentage of the hACE2 binding signal in wells without antibody. IC50 values for mAb inhibition of hACE2 binding to COV2-S_VFLIP proteins were determined after log transformation of antibody concentration using sigmoidal dose–response nonlinear regression analysis (Prism software, GraphPad Prism v.8.0). Each mAb dilution series was performed in triplicate except for the SA55 positive control mAb, which was performed in duplicate. Each assay was repeated in 2 independent experiments.
MAb passive-transfer protection studies in mice
Animal studies were carried out in accordance with the Institutional Animal Care and Use Committee at UNC-Chapel Hill (protocol no. 23-085). The hACE2-K18 mice used for this study were bred at UNC and original breeding pairs were obtained from the Jackson Laboratory (034860). Twelve-month-old hACE2-K18 mice of both sexes were treated with 200 μg of mAb or isotype-matched control mAb via intraperitoneal injection 24 h before viral challenge. Mice were anaesthetized with a mixture of ketamine/xylazine, inoculated intranasally with 104 p.f.u.s of SARS-CoV-2 (XBB.1.5) and monitored daily for clinical signs of disease, weight loss and mortality. At the indicated times after infection, mice were euthanized via isoflurane overdose, and the inferior lung lobe was collected in PBS with glass beads and stored at −80 °C for viral titre determination via plaque assay as previously described53. Briefly, lung homogenates were serially diluted in PBS and 200 µl of diluted homogenate were added to Vero E6 cells, followed by the addition of an agarose overlay. Three days following infection, wells were stained with neutral red and plaques were quantified.
Negative-stain electron microscopy sample and grid preparation, imaging and processing of S-Fab complexes
To perform negative-stain electron microscopy (nsEM), Fab molecules were produced by digesting purified IgG molecules using a resin-immobilized cysteine protease enzyme (Genovis, A2-AFK-100). The digestion occurred in 100 mM sodium phosphate, 150 mM NaCl pH 7.2 (PBS) for ~16 h at ambient temperature. Following digestion, the reaction mix was incubated with CaptureSelect IgG-Fc (multispecies) resin (Thermo Fisher) for 30 min at ambient temperature in PBS buffer to remove intact IgG and cleaved Fc.
For screening and data collection of XBB or BQ.1.1 VFLIP S proteins in complex with Fab molecules, the proteins were incubated at a molar ratio of 4:1 (Fab:S) for ~1 h at r.t. and ~3 µl of the sample at concentrations of ~10–15 µg ml−1 was applied to a glow-discharged grid with continuous carbon film on 400-square-mesh copper EM grids (Electron Microscopy Sciences). Uranyl formate (2%) was used for staining54. Images were collected using an FEI TF20 (TFS) transmission electron microscope equipped with a Gatan US4000 4k × 4k CCD camera operated at 200 keV and controlled with SerialEM55. All imaging was performed at ×50,000 magnification, resulting in a pixel size of 2.18 Å pixel−1 in low-dose mode at a defocus of 1.5–1.8 μm. The total dose for the micrographs was ~30 e− Å−2. The cryoSPARC software package was used for image processing56. Images were imported, and micrographs were contrast transfer function (CTF)-estimated and particles autopicked. The particles were extracted with a box size of 256 pixels and binned by 2–128 pixels (4.36 Å pixel−1). Class averages (2D) were performed and good classes were selected for ab initio model and non-uniform (NU) refinement with or without symmetry depending on the occupancy of the Fab molecules. The final resolution of the maps was ~20–26 Å. ChimeraX57 was used for model docking, segmentation of the nsEM map and figure preparation. PDB IDs 7LRT and 12E8 were used for the S trimer protein and Fab molecules, respectively. All nsEM data have been deposited in the EMDB with accession codes EMD-43882 through EMD-43888.
Cryo-EM sample and grid preparation for the BA.1 spike protein in complex with COV2-3835 Fab molecules
The BA.1 spike protein was mixed to a final concentration of 0.65 mg ml−1 with a 1.25× molar excess of COV2-3835 Fab molecules in buffer containing 2 mM Tris pH 7.5, 200 mM NaCl and 0.02% NaN3. Of the complex, 4 μl was applied to UltrAufoil R 1.2/1.3 300-mesh gold TEM grids (Electron Microscopy Sciences) that had been glow discharged using a PELCO easiGlow (Ted Pella) at a current of 20 mA for a total of 60 s. Using a Vitrobot Mark IV (Thermo Fisher), a blot force of 2 was applied for 3 s to blot away excess liquid before plunge freezing into liquid ethane. Samples were blotted in 100% humidity at 22 °C. A total of 3,096 videos were collected from a single grid using a Titan Krios G3 TEM (Thermo Fisher) equipped with a Biocontinuum Imaging Filter (Gatan), set to a slit width of 20 eV and a K3 direct electron detector (Gatan). SerialEM v.4.0.10 software was used for automatic collection of all videos55. Particles were imaged with an exposure of 14 eps for 3.8 s (total exposure of 60 e Å−2) at a calibrated magnification of 0.84 Å pixel−1. Additional details about data collection can be found in Extended Data Fig. 6 and Supplementary Table 3.
Cryo-EM data processing and structure building for the BA.1:COV2-3835 Fab complexes
Motion correction, CTF estimation, particle picking and preliminary two-dimensional (2D) classification were performed using cryoSPARC v.4.1.1 live processing56 (Extended Data Fig. 6 and Supplementary Table 3). After data collection was completed, 1,127,216 extracted particle picks (box size: 512 pixels, Fourier cropped to 160 pixels) were sorted into 100 2D class averages using an uncertainty factor of 1 to determine and eliminate ‘junk’ classes. From this, 522,475 particles from 25 classes were selected and a second round of 2D classification was performed using an uncertainty factor of 2 to increase diversity of higher-quality classes. A total of 498,988 particles were selected, and 200,000 were used to perform a 4-class ab initio reconstruction. All particles were subsequently used to perform a heterogeneous refinement of the 4 ab initio volumes. Particles from the 2 highest-quality classes were selected and another ab initio reconstruction was performed using 5 classes to aid in sorting out any remaining ‘junk’ particles. From the subsequent refinement of these volumes, particles from the 2 highest-quality classes were further sorted into 3 classes using 3D classification limited to 8 Å resolution to observe subtle heterogeneity in the particles due to flexibility and/or mixed Fab occupancy. A total of 153,719 particles from a single 3D class were refined using homogeneous refinement with no applied symmetry, followed by non-uniform refinement. After re-extracting the particles without binning, an additional non-uniform refinement was performed with no applied symmetry, and with refined per-particle defocus and per-group CTF parameters58. The resulting map reached a global resolution of 2.8 Å, although resolution of the RBD–Fab interface was limited due to RBD flexibility. The refined volume was imported into ChimeraX59 to generate a mask encompassing a single Fab molecule bound to the RBD and most of the N-terminal domain (NTD) of the neighbouring protomer. The mask was then imported back into cryoSPARC to perform focused refinement of the final global reconstruction to yield a 3 Å reconstruction of the RBD bound to COV2-3835 Fab molecules. To improve map quality, the focused refinement volume was processed using DeepEMhancer within COSMIC2 science gateway60. An initial model of the complex was generated by fitting the RBD (residues 329–529) from a high-resolution model of the SARS-CoV-2 Omicron BA.1 spike protein (PDB ID: 7TM0 (ref. 61)). A model of the COV2-3835 Fab molecule was generated using SAbPred ABodyBuilder62 and fit into the local refinement volume via ChimeraX59 The structure was iteratively refined and completed using a combination of Phenix (v.1.21.1)63,64, Coot (v.0.9.2)65 and ISOLDE (v.1.8)66. Data collection and refinement statistics are available in Supplementary Table 4.
Cryo-EM sample and grid preparation for the BQ.1.1 spike protein in complex with COV2-3891 Fab molecules
After incubating for 2 h at room temperature, the BQ.1.1 VFLIP S-COV2-3891 Fab mixture was purified by gel filtration on Superose 6 Increase 10/300 column (GE Healthcare) equilibrated in a buffer containing 20 mM HEPES, 150 mM NaCl and 1 mM EDTA (pH 8.0). Of the purified mixture at a concentration of ~0.3 mg ml−1, 2.2 µl was applied to glow-discharged grids (40 s at 25 mA) (carbon grid, 300-mesh 1.2/1.3, Quantifoil). The grids were blotted for 3 s before plunging into liquid ethane using Vitrobot MK4 (Thermo Fisher) at 20 °C and 100% relative humidity. Grids were screened on a Glacios (Thermo Fisher) microscope operated at 200 keV and equipped with a Falcon 4 (Thermo Fisher) DED detector. Data were collected and imaged on a Krios (Thermo Fisher) microscope operated at 300 keV and equipped with a K3 and GIF energy filter with a 20 eV slit (Gatan) DED detector using counting mode. Datasets were collected using EPU and videos were collected at a nominal magnification of ×130,000 and pixel size of 0.647 Å pixel−1, with a defocus range of 0.8–1.8 µm. Grids were exposed at ~1.27 e Å−2 frame−1, resulting in a total dose of ~63 e Å−2 (Extended Data Fig. 7 and Supplementary Table 5).
Cryo-EM data processing and structure building for BQ.1.1:COV2-3891 Fab complexes
To overcome the problem of preferred orientation, multiple datasets were collected with varying tilt angles (0° and 30°). These datasets were preprocessed separately and combined at a later stage. Videos were preprocessed with Relion Motioncor2 (ref. 67) and CTFFind4 (ref. 68) on the fly. Micrographs with low resolution, high astigmatism and high defocus were removed from the data set. The datasets were first picked manually and extracted in a binned box (3.0328 Å pixel−1). The particles were used for 2D classification, and good classes were used for training, repicking with Topaz69 and extraction. These particles were subjected to multiple rounds of 2D class averages, 3D initial maps and 3D classification to obtain a clean, homogeneous particle set. At this point, the datasets were combined and re-extracted at a pixel size 0.9705 Å pixel−1. The data were subjected to 3D autorefinement and post processing, resulting in a 3.7 Å map. The data were then subjected to CTFrefine within Relion70,71, and one more 3D autorefinement and post-processing step resulted in a final map at 3.6 Å resolution. Focused refinement of the RBD–Fab complexes was done on subtracted particles in Relion and cryoSPARC56,70,71, and the 3D autorefinement and post processing at this step resulted in a ~4.1 Å map. Sharpening was carried out using DeepEMhancer60. Reported resolutions are based on the gold-standard Fourier shell correlation (FSC) criteria of 0.143. Workflow schematics and detailed statistics are provided in Extended Data Fig. 7 and Supplementary Table 5. PDB ID 8FXC ref. 72 was used as an initial model for the RBD. ImmuneBuilder73 prediction was used as an initial model for the Fv of the Fab molecule. The models were first docked into the map with ChimeraX57. To improve the coordinates, the model was subjected to iterative refinement comprising manual building in Coot65 and real space refinement with Phenix63. The model was validated with Molprobity74 (Supplementary Table 5). Figures were generated using PyMOL75, and the EM maps and models have been deposited in the EMDB and PDB repositories (Supplementary Table 5).
Crystallization of the XBB.1.5 RBD in complex with COV2-3906 Fab molecules
The Prescission 3C protease tag was removed from purified XBB.1.5 RBD by adding Prescission Protease (Trialtus Biosciences) to the RBD at a 1:200 mass ratio and incubating for 24 h at 4 °C. Untagged XBB.1.5 RBD was mixed with purified, recombinantly expressed COV2-3906 Fab molecules at a 1.5:1 molar ratio (RBD:Fab) and incubated for ~1 h before being loaded onto a HiLoad 16/600 Superdex 200 size exclusion column (Cytiva Life Sciences) for purification. The purified complex was concentrated to 10 mg ml−1 in 25 mM HEPES pH 7.3 and 100 mM NaCl, and crystallization screens were set up using a mosquito XTAL3 crystallization robot (SPT Labtech). Further optimization used the dragonfly crystal liquid handler (SPT Labtech). The XBB.1.5 RBD/COV2-3906 Fab complex was crystallized in 1.4 M ammonium sulfate, 0.1 M sodium acetate pH 5.1 and 2 mM DL-panthenol, and was cryoprotected using well solution:glycerol at a 7:3 ratio. Diffraction data were collected at the European Synchrotron Radiation Facility (ESRF) beamline ID30A-3. The data were integrated using XDS76 and scaled using AIMLESS77. The crystal structure was solved by molecular replacement with the software Phaser78, using Alphafold2 (refs. 79,80,81)-generated models of the XBB.1.5 RBD and COV2-3906 Fab molecule as the search models. Iterative refinement of the structure was completed using Phenix63 and Coot65. Data collection and refinement statistics are shown in Supplementary Table 6.
Statistics and reproducibility
A non-parametric Kruskal–Wallis test with Dunn’s post hoc correction for multiple comparisons was used for comparisons between experimental conditions and the isotype-matched negative control for animal protection studies, and associated P values are reported without assigning thresholds for statistical significance. Data distributions were assumed to be normal but this was not formally tested. No data were excluded from analyses. Statistical methods were not used to predetermine sample sizes, but animal study sample sizes were powered on the basis of previous publications18. Experiments were not randomized, and investigators were not blinded to sample or participant allocation during experiments and outcome assessment.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.