SARS-CoV-2–Reactive Mucosal B Cells in the Upper Respiratory Tract of Uninfected Individuals

This article is featured in Top Reads, p.

Severe acute respiratory syndrome coronavirus 2 is a novel betacoronavirus identified as the causative agent of the coronavirus disease 2019 (Covid-19) pandemic. It is closely related to bat-SARS virus and thus, like the SARS virus, has been shown to use the angiotensin-converting enzyme 2 (ACE2) receptor and the cellular protease transmembrane protease serine 2 for cell entry (1). The viral Spike (S) protein is responsible for receptor binding and fusion of the viral and host membranes, and its receptor-binding domain (RBD) mediates the interaction with ACE2 (2). The SARS-CoV-2 RBD is also the most frequently targeted region on the S-protein Ag by neutralizing Abs in humans (3–5). Our understanding of the immune response to SARS-CoV-2 in general, and the Ab response in particular, is based largely on studies of blood-derived samples from acute or convalescent patients with Covid-19. Interestingly, circulating SARS-CoV-2–reactive T cells have been described in SARS-CoV-2 naive individuals, and it has been suggested that they may have arisen from encounters with related coronaviruses (6–10). Circulating Abs recognizing SARS-CoV-2 in uninfected individuals have also been observed (11, 12), and pre-existing S-protein–reactive serum Abs have neutralizing activity in vitro (13). In a recent study, it was reported that H chains of convergent clones of B cells in the blood and tissue samples of individuals unexposed to SARS-CoV-2 are shared with clones isolated from patients contracting the disease, suggesting the possibility of pre-existing SARS-CoV-2–specific B cells (14). Although SARS-CoV-2 is a respiratory pathogen, only very limited information exists with respect to mucosal immune responses to pathogenic challenge, especially with regard to mucosal B cells and Ab responses. Epidemiological studies indicate that asymptomatic infections or infections with a mild course of disease can frequently represent the majority of the study population (15). These observations suggest that the mucosal immune system of the upper respiratory tract may play a role in limiting disease severity in a significant percentage of SARS-CoV-2 infections. This prompted us to explore the presence of mucosal SARS-CoV-2–reactive B cells in tonsillar tissue specimens collected from pediatric patients 3 or more years prior to the outbreak of the Covid-19 pandemic.

In this study, we report the identification of pre-existing mucosal SARS-CoV-2–reactive Abs. These Abs are encoded by naive and Ag-experienced B cells, exhibit neutralizing potential, and do not recognize endemic human coronaviruses.

Abs recognizing human CD3 (clone SK7), CD19 (clone HIB-19), CD38 (clone HIT-2), and IgD (clone 1A6-1) Ags were obtained from BD Biosciences (San Jose, CA). Streptavidin (SA)-PE, SA-allophycocyanin, PE-conjugated anti-human, and anti-mouse secondary Abs were purchased from SouthernBiotech (Birmingham, AL). Protein A Sepharose was obtained from Sigma-Aldrich (St. Louis, MO). Human embryonic kidney (HEK) 293T cells, parental HEK293F cells, and HEK293F cells stably expressing SARS-CoV-2 S-protein or ACE2 were grown in DMEM supplemented with 10% FBS, glutamine, and penicillin/streptomycin (100 U/ml) in a humidified atmosphere at 37°C with 5% CO₂. Anti–human coronavirus (HCoV)-229E (clone 9.8E12) and anti–SARS-CoV-2 (clone DH1057-1) Abs were purified in our laboratories. Soluble trimeric S-proteins of SARS-CoV-2, and SARS-CoV, were cloned into the doxycycline-inducible expression vector PB-T-PAF and expressed in HEK293F cells, followed by purification of recombinant protein from culture supernatants using Ni-affinity chromatography and size-exclusion chromatography, as described previously (16). The quality of the purified trimeric S-proteins was monitored by negative-stain electron microscopy.

S-protein, RBD, and variable lymphocyte receptor (VLR)–negative control tetramers were generated by preincubation of 50 pmol biotinylated recombinant protein and 12.5 pmol SA-PE or SA-allophycocyanin for 2 h on ice at a molar ratio of 4:1. Potentially unoccupied biotin binding sites on SA were blocked by subsequent addition of free biotin. Allophycocyanin- and PE-coupled tetramers were mixed at a 1:1 ratio prior to addition to tonsillar lymphocytes.

Cryopreserved single-cell suspensions of tonsillar lymphocytes were thawed and subjected to density gradient centrifugation using Lymphocyte Separation Medium. Cells were washed with PBS and resuspended in PBS/1% BSA and incubated with 5 μg/ml SA to reduce potential nonspecific subsequent binding of S-protein or RBD tetramers. The cells were washed with PBS and incubated with allophycocyanin- and PE-coupled S-protein or RBD tetramers mixed at a 1:1 ratio for 30 min on ice. Subsequently, anti-IgD (V605), anti-CD27 (allophycocyanin-H7), anti-CD38 (V450), anti-CD19 (allophycocyanin-Cy7), and anti-CD3 (PerCP-Cy5.5) were added to the cells followed by incubation on ice for additional 20 min. Cells were washed twice with PBS, resuspended in PBS containing Live/Dead dead cell exclusion dye (Thermo Fisher Scientific, Waltham, MA), and signals acquired using a BD LSR II instrument (BD Biosciences). Flow cytometry data were analyzed using the FlowJo software package (Tree Star, Ashland, OR). The gating strategy for the analysis of tonsillar B cell populations is depicted in Supplemental Fig. 2. Alternatively, cells double-positive for tetramer binding on PE and allophycocyanin were purified in bulk by FACS for single-cell transcriptome and immunoreceptor analyses or individually sorted for single-cell RT-PCR of Ag receptor-encoding transcripts.

Individual Ag-reactive B cells were sorted into PCR plates containing 10 μl catch buffer supplemented with RNasin and immediately frozen at −80°C. Primary RT-PCR reactions were performed using the One-Step RT-PCR system (Qiagen, Hilden, Germany), and the secondary PCR reactions were performed using the KOD enzyme system (EMD Millipore, Billerica, MA) as described earlier (17). PCR-amplified gene fragments were cloned into the IgG1 H chain and κ L chain vectors and the plasmids transiently transfected into HEK293T cells at a 1:1 ratio. HEK293F cells stably expressing SARS-CoV-2 S-protein were incubated for 30 min with culture supernatants 48 h after transfection followed by incubation with PE-conjugated goat anti-human IgG secondary reagents and analysis of Ab binding using a Guava easyCyte flow cytometer (EMD Millipore). Amplified V(h) and V(l) sequences were verified by DNA sequencing and Abs reactive with HEK293F cells expressing SARS-CoV-2 S-protein, and nonreactive with HEK293F parental control cells were selected for purification from culture supernatants using protein A affinity chromatography and further analysis. Inferred germline revertant (iGR) Abs were generated by DNA custom synthesis or site-directed mutagenesis following reversal of mutations in the V(h) and V(l) genes identified using the IMGT tool (18).

Single-cell library preparation and sequencing were performed by the Princess Margaret Genomics Centre (Toronto, ON, Canada) using 10x Genomics Chromium controller to prepare libraries for gene expression (GEX) (Single Cell 5′ R2-only Chemistry) and BCR [Single Cell V(D)J R2-only]. We prepared libraries for six samples: three tonsillar samples enriched for the S-protein–reactive cells by FACS sorting and three corresponding samples of nonenriched tonsillar B cells. Libraries were sequenced on an Illumina-2500 for single-cell gene expressions (5000 cells, 50,000 reads/cell) and the single-cell BCR sequences (5000 cells, 5000 reads/cell). Our data consisted of single-cell GEX and BCR sequencing data for three S-reactive enriched tonsil samples and their corresponding pairwise total tonsil samples. Data for the single-cell BCR sequencing of three tonsillar tissues were deposited at the Sequence Read Archive with accession number PRJNA741134 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA741134/).

Our computational Ab discovery was based on the single-cell RNA-sequencing (scRNAseq) gene expression and BCR sequencing data. Due to the low numbers of purified S-protein–reactive cells, we added purified T cells as carriers to facilitate processing.

For GEX data, we used 10x Genomics CellRanger (version 3.1.0) with default settings (19) to demultiplex unique molecular identifiers of gel bead-in emulsion barcodes and embedded STAR (20) to align GEX data to the human genome version GRCh38-3.0.0 and used gene count and filtered barcodes as the default output of the pipeline for further analysis. We used the Seurat R package (21) as the main framework for single-cell transcriptome analysis. Cells with coverage on <200 genes and >6000 genes and cells with a mitochondrial RNA count of >20% were removed. We normalized cell counts using Seurat’s NormalizeData with “LogNormalize” parameters and a scale factor of 10,000 and used 5,000 most variant genes to perform principal component analysis (PCA). The first 20 principal components were used to identify clusters in the UMAP space and further visualization. We used standard cell markers to identify carrier T cells (CD3E, CD4, CD8A, and CD8B) and removed them from further transcriptome analysis. Specifically, we used FindIntegrationAnchors (30 dimensions, default parameters) to identify pairwise anchors among 6 samples and used IntegrateData with (all genes set for features.to.integrate, and otherwise default parameters) to integrate our 6 data sets. We then scaled the data using ScaleData, computed the PCA using RunPCA, found clusters in 20 lower dimensions using FindNeighbors and FindClusters (resolution = 1), and used RunUMAP for the visualization. To address whether S-protein enrichment altered the samples’ B cell composition, we computed the ratio of the fraction of cells in each cluster normalized by the total number of cells in each sample after clustering the combined tonsillar GEX data (three enriched and their three corresponding nonenriched tonsils) and clustering in UMAP space.

We used 10x Genomics CellRanger (ver.4.0) with default settings to demultiplex unique molecular identifiers of gel bead-in emulsion barcodes and used the pipeline to align the contigs to IMGT BCR loci germline reference (22). Only BCR data of cells with productive H and L chain sequences were included for further analysis; BCR records that did not have a corresponding gene-expression record in our GEX data set were removed.

After preprocessing the single-cell BCR sequencing data, our data sets consisted of 737, 606, and 83 Ab sequences for the three SARS-CoV-2 S-protein–enriched samples and 1717, 1009, and 835 sequences for the nonenriched samples, respectively. To address whether S-protein enrichment of tonsillar B cells resulted in repertoire changes of enriched samples compared with the original total tonsil samples, we compared several BCR repertoire properties, including V-J gene-usage and amino acid sequence changes (in-house scripts) as well as CDR3 loop length and physicochemical properties [e.g., grand average of hydropathicity index, bulkiness, charge, polarity, length, charge, and amino acid usage using Alakazam (23)].

Our reference Ab data set for sequence homology search consisted of 1854 publicly available Covid-19 Abs (24) and 15 mAbs that we isolated using a single-cell RT-PCR approach (p-series Abs, Fig. 3). We used MAFFT to compute pairwise alignments between the BCR data to the reference data set Abs and selected any sequences for further analysis if they passed the sequence homology criterion of matched V and J gene of the H or L chains with at least 60% similarity (hamming distance) in the CDR3 loops of heavy and light sequences. We further conducted a structural alignment of their complete H and L chain sequences to the three-dimensional crystal structures of Abs in CoV-AbDab (24) using an in-house pipeline based on Modeler (25) to identify BCR sequences that were similar in the structure at the binding site to Covid-19 RBD or S-protein. The Modeler native alignment algorithm was used to align BCR sequences to the reference Ab sequences (auto_align), to model the BCR data (automodel), and to perform the alignment with loop optimization (dope_loopmodel). Alignments were assessed by computing the Discrete Optimized Protein Energy and selected Ab candidates with negative scores for further selection, DNA custom synthesis, and functional analysis.

ELISA plates were coated with 100 μl trimeric S-proteins of SARS-CoV or SARS-CoV-2 (1 μg/ml), SARS-CoV-2 RBD (2 μg/ml), or heat-inactivated microbial ecosystem therapeutic 1/2 (MET-1/2) commensal microbiota (OD₆₀₀ 0.2) at 4°C overnight. The plates were blocked with PBS/5% BSA prior to incubation with primary human mAbs (1 μg/ml) or PBS as background control. Bound Abs were detected using HRP-conjugated rabbit anti-human IgG and tetramethylbenzidine ELISA substrate. Reactions were quenched using 2N H₂SO₄, and colorimetric values were determined using a SpectraMax i3 (Molecular Devices, San Jose, CA) instrument at the 450-nm setting. All OD values represent the means of at least three independent experiments following background subtraction.

Plasmids encoding the wild-type S-proteins of SARS-CoV, SARS-CoV-2, HCoV-OC43, or HCoV-229E were expressed from the PB-T-PAF vector backbone. SARS-CoV-2 S-protein variants B.1.117 and B.1.351 were obtained from InvivoGen (San Diego, CA) and recloned into the PB-T-PAF expression vector (26) prior to use in transient transfection assays. Expression constructs were transiently cotransfected with a GFP fluorescent marker into HEK293F cells at a ratio of 4:1. Cells were cultured for 48 h prior to harvesting using PBS/0.5 mM EDTA, filtration using 40-μm cell strainers, and resuspension in PBS/1% BSA. Cells were incubated with purified Abs at a concentration of 1 μg/ml for 25 min on ice, washed with PBS/1% BSA followed by incubation with PE-conjugated goat anti-human IgG, and detected by flow cytometry using a Guava easyCyte instrument. Cell surface expression of the various S-proteins was independently validated and is shown in Supplemental Fig. 2. Flow cytometry data were analyzed using the FlowJo software package. Specific binding of mAbs to transfected cells was calculated by normalizing median fluorescence intensity (MFI) values of S-protein–transfected cells to MFI values of negative control cells transfected only with GFP expression plasmids.

Purified mAbs (2.5 μg) were preincubated with biotinylated SARS-CoV-2 S-protein (0.5 μg) at room temperature for 30 min in a volume of 40 μl. Subsequently, the reaction was added to an equal volume of HEK293F cells stably expressing human ACE2, resuspended in PBS/1% BSA, and incubated for an additional 30 min. The cells were washed with PBS/1% BSA, incubated with SA-PE for 20 min on ice, and S-protein binding assessed by flow cytometry. Relative S-protein binding was obtained by normalizing MFI to MFI values obtained with negative control experiments of SA-PE without prior S-protein incubation, and Ab-mediated inhibition was calculated by dividing MFI values obtained for S-protein binding with preceding Ab incubation by values obtained without Ab incubation. Separately, RBD-dependent Ab inhibition was performed with RBD tetramers generated by precoupling biotinylated RBD to SA-PE and proceeding with the experiment as described above.

Statistical analysis was performed using Friedmann tests with Dunn multiple-comparison post hoc test, Mann–Whitney U tests, and one-sample Wilcoxon signed-rank tests for surrogate neutralization assays.

Tonsillar tissue was obtained from The Hospital for Sick Children in Toronto, Ontario, Canada from pediatric patients undergoing tonsillectomy in 2015 and 2016 with informed consent and approval of the Institute’s Ethics Review Board in accordance with the Declaration of Helsinki.

The mucosal surface of the respiratory tract is the primary point of contact with airborne pathogens. In an effort to investigate pre-existing SARS-CoV-2–reactive mucosal B cells of the upper respiratory tract that may influence the course of disease, we explored S-protein binding of cryopreserved cells from prepandemic tonsillar tissues. Flow cytometric assessment of tonsillar B cells using fluorescently labeled, stabilized trimeric S-protein showed the presence of reactive cells in all analyzed samples (Fig. 1A, top and bottom row, left panels). RBD-reactive cells were also detected, although at lower frequencies that did not reach statistical significance. The majority of S-protein–reactive cells were in the naive B cell compartment with reduced numbers of reactive cells among memory B cells and germinal center (GC) B cells (Fig. 1A, top row, right panel, and bottom row, center panel). This distribution mirrored the B cell populations detected in the unfractionated tonsil samples (Fig. 1A, bottom row, right panel). Using an additional set of 10 tonsil samples, we included recombinant VLR B39 molecules as irrelevant control Ag (27). Furthermore, we included Abs to CD27 in the flow cytometry panel because unswitched IgD⁺ memory B cells can be found in the IgD⁺/CD38⁻ naive B cell gate (28, 29). These experiments showed that binding of recombinant VLR-coupled SA-PE/SA-allophycocyanin did not differ from SA-PE/SA-allophycocyanin without bound protein Ag (Fig. 1B, left panel). Additionally, we also observed a noticeable increase in S-protein binding to IgD⁺/CD27⁺ cells relative to their frequency in total tonsillar B cells (Fig. 1B, center and right panels).

To validate the specific recognition of the SARS-CoV-2 S-protein, we generated mAbs from these B cells for further detailed analysis. The process involved either single-cell RT-PCR (scRT-PCR) or scRNAseq immune profiling of the tonsillar B lineage cells that were enriched for SARS-CoV-2 S-protein reactivity, followed by a novel computational strategy to identify candidate Ag receptor H and L chain gene sequence pairs with SARS-CoV-2 S-protein reactivity (Fig. 2A). This computational strategy involved sequence and structural alignment approaches using Ab sequences from the SARS-CoV-2–enriched data sets with publicly available SARS-CoV-2–reactive Abs (24) and Ab sequences we obtained by scRT-PCR; also incorporated was gene usage information and B cell subsets obtained from single-cell transcriptome analysis. Exploration of H chain variable gene usage revealed shared sequences from all tissue samples with published Ab sequences isolated from SARS-CoV-2–infected individuals, an observation that was maintained when the additional constraint of >65% CDR3 homology was included and allowed detection of several sequences under very stringent conditions of >90% CDR3 homology (Fig. 2B).

We generated 87 matching H and L chain sequences; 78 H and L chain pairs secreted Abs following transfection of HEK293T cells, 40 of which displayed at least a 2-fold increase in binding to HEK293F cells expressing the stabilized S-protein trimer on the cell surface, relative to untransfected negative control cells (Fig. 3A). Two Abs that bound SARS-CoV-2 S-protein–expressing cells as well as untransfected control cells, indicative of polyreactivity, were excluded from further analysis. Fifteen of the mAbs originated from naive B cells, whereas the remaining 25 Abs originated from Ag-experienced memory B cells and GC B cells as well as plasma cells (PC) (Fig. 3B, 3E) (Supplemental Fig. 1) that consisted of unswitched IgM and class-switched IgG and IgA isotypes (Supplemental Table I). The most frequently used V(h) gene sequences belonged to the V(h) 1-69, V(h) 3-9, V(h) 3-30, and V(h) 1-46 gene families, consistent with the increased frequency of these V(h) gene families among SARS-CoV-2–targeting Abs isolated from patients with Covid-19 (4, 5, 30–32). In an independent series of experiments using ELISAs, we validated Ag binding and specificity using each of the following: 1) recombinant trimeric S-proteins from SARS-CoV-2 and SARS-CoV, 2) recombinant SARS-CoV-2 RBD, 3) commensal microbiota as a source of prevalent environmental Ag, and 4) BSA as a negative control. These experiments confirmed the SARS-CoV-2 recognition observed by flow cytometry. They also showed that a small subset of the Abs strongly reacted with the S-protein of SARS-CoV, whereas a second distinct subset of Abs reacted moderately with a combination of the MET-1 and MET-2 communities of commensal microbiota (Fig. 3C) (33, 34). None of the Abs reacted with negative control BSA (Fig. 3C), and strong SARS-CoV-2 binding did not correlate with strong SARS-CoV or MET-1/2 recognition (Fig. 3C, 3E), indicating that SARS-CoV-2 S-protein recognition is not the result of overt polyreactivity. Only three mAbs recognized the recombinant SARS-CoV-2 RBD Ag (Fig. 3C, 3E), an observation consistent with the low frequency of RBD-reactive cells observed in all of the analyzed tonsil samples. Combined, these experiments demonstrate the presence of pre-existing SARS-CoV-2 S-protein–reactive B cells in the upper respiratory tract.

Recent studies showing S-protein reactivity of sera collected from SARS-CoV-2–negative individuals suggest that a previous encounter with endemic HCoV may be a potential source of pre-existing circulating S-protein–reactive Abs (11, 13). To explore whether endemic HCoV recognition could underpin the observed SARS-CoV-2 recognition shown by our panel of tonsillar Abs, we conducted transient transfection experiments in which wild-type S-proteins of various coronaviruses were analyzed for Ab recognition. This is in contrast to the experimental results depicted in (Fig. 3A and 3C, which used modified S-proteins containing two mutations (K986P and V987P) inserted to stabilize the protein in a prefusion conformation and three amino acid mutations that eliminate the S1/S2 protease cleavage site (R682S, R683S, and R685S) (16). We observed that our panel of Abs readily recognized the wild-type SARS-CoV-2 S-proteins with a small subset of Abs also binding to the S-protein of SARS-CoV but that none of the Abs showed reactivity to the S-proteins of the endemic HCoV-OC43 and HCoV-229E coronaviruses (Fig. 3D, 3E).

In addition to wild-type SARS-CoV-2, several variants of SARS-CoV-2 were identified during the course of the pandemic with mutations in the viral genome, including S-protein encoding sequences and the potential to change virus transmissibility and recognition by the adaptive immune system (Refs. 35, 36, and D.P. Martin, S. Weaver, H. Tegally, E.J. San, S.D. Shank, E. Wilkinson, J. Giandhari, S. Naidoo, Y. Pillay, L. Singh, et al., manuscript posted on medRxiv, DOI: 10.1101/2021.02.23.21252268). We transfected HEK293F cells with expression plasmids encoding the B.1.1.7 variant first detected in the United Kingdom, or the B.1.315 variant first detected in South Africa, to explore recognition of these variant S-proteins by our panel of pre-existing Abs. These experiments showed that recognition of variant S-proteins was significantly reduced compared with the wild-type S-protein (Fig. 3D, 3E). Control experiments in which transfected cells were analyzed using antiserum from two SARS-CoV-2–vaccinated individuals showed reduced signals for the SARS-CoV-2 variants relative to wild-type SARS-CoV-2. This observation is likely due to Ab responses biased toward SARS-CoV-2 wild-type S-protein following vaccination because cell surface expression of wild-type SARS-CoV-2 and the B.1.1.7 variant was comparable following assessment using the RBD-specific human mAb C23 (Supplemental Fig. 2).

We were intrigued that pre-existing SARS-CoV-2–reactive Abs were identified among Ag-experienced B cell populations. To explore the influence of somatic mutations incorporated into the Ag receptors, we generated iGR Abs, in which nonsynomymous v-gene mutations were replaced with codons found in unmutated germline sequences, for selected Abs from memory B cells, PC, and GC B cells. The majority of these iGR Abs were no longer capable of binding to SARS-CoV-2 S-protein–expressing cells (Fig. 4A), and only a single iGR Ab showed noticeable binding to control HEK293F cells. Several of the isolated Abs, particularly among those derived from naive B cells, showed reactivity to MET-1/2 commensal microbiota. We therefore tested whether iGR clones would gain microbiota reactivity. However, recognition of MET-1/2 by parental clones, including somatic mutations and iGR Abs, remained largely unchanged (Fig. 4B). These experiments indicate that S-protein reactivity was a de novo acquired characteristic.

Sera from SARS-CoV-2–negative individuals recognizing endemic HCoV do not provide protection from SARS-CoV-2 infection (11, 37), although in some cases, pre-existing polyclonal serum Abs were reported to inhibit SARS-CoV-2 infection in vitro (13). The specific recognition of the S-protein of SARS-CoV-2 by our panel of isolated Abs prompted us to explore the potential protective function of these reagents. We investigated the ability of the various mAbs to interfere with S-protein binding to HEK293F cells expressing the ACE2 receptor in a modified surrogate neutralization assay (16). Using this approach, preincubation of the S-protein with our Ab panel revealed five mAbs that consistently showed >20% inhibition of S-protein binding (Fig. 5A). This compared with ∼50% inhibition observed with the RBD-specific mAb C23. Performing this surrogate neutralization assay with the SARS-CoV-2 RBD, we observed that only Ab s025 interfered with RBD binding to cell surface ACE2 (Fig. 5B), indicating that the potential protective function of these Abs may arise from targeting S-protein epitopes distinct from the RBD. Analysis of Ag receptor sequences from tonsillar B cells enriched for SARS-CoV-2 S-protein recognition with matching samples of nonenriched B cells did not reveal differences in variable gene usage, the level of somatic mutations, or the nature of the CDR3 loop residues (Supplemental Fig. 3). However, Abs from pre-existing B cells that interfered with SARS-CoV-2 S-protein binding to ACE2 tended to have increased usage of aromatic residues in their CDR3 loops, and 25% of our panel of mAbs used VH1-69 variable gene sequences, a finding consistent with observations of increased frequency of VH1-69 variable genes in Abs from Covid-19 convalescent samples (Supplemental Fig. 3).

Although Covid-19 continues to inflict significant stress on public health systems, the majority of infections are either asymptomatic or result in a mild course of disease. Our detection of pre-existing SARS-CoV-2–reactive Abs in the mucosal immune system of the upper respiratory tract that, although lacking the potency of neutralizing Abs isolated from Covid-19 convalescent individuals, exhibit the ability to partially block ACE2 binding in vitro provides an as-yet-unexplored contribution of the mucosal adaptive immune system to pathogen protection. In addition to in vitro neutralization of SARS-CoV-2 by mAbs isolated from convalescent patients with Covid-19, in vivo experimentation strategies indicated Fc-mediated effector functions of neutralizing IgG Abs (Ref. 38 and A. Schafer, F. Muecksch, J. C. C. Lorenzi, S. R. Leist, M. Cipolla, S. Bournazos, F. Schmidt, A. Gazumyan, R. S. Baric, D. F. Robbiani, manuscript posted on bioRxiv, DOI: 10.1101/2020.09.15.298067). It is conceivable that additional Fc-mediated Ab effector functions such as complement activation by pre-existing unswitched IgM Abs may contribute to rapid clearance of the pathogen in patients with asymptomatic or mild course of disease while contributing to the severe disease manifestations of Covid-19 in the context of dysregulated immune responses (39–41).

Prepandemic serum samples were reported to cross react to SARS-CoV-2, indicating that humoral immune responses to endemic HCoV may engage the viral pathogen following infection but they do not correlate with protection from SARS-CoV-2 (11, 13). In our analysis, we did not observe cross-reactivity of the mAbs with the endemic HCoV-OC43 and HCoV-229E. Although SARS-CoV-2–reactive mAbs that cross react with endemic HCoV were isolated from patients with Covid-19 (42, 43), <2% of entries in the CoV-AbDab Ab database are reactive with HCoV-229E and <1% are reactive with HCoV-OC43 (24), suggesting that cross-reactivity to these HCoV occurs only at low frequencies. Furthermore, it remains to be investigated whether B cells encoding cross-reactive Abs will be observed with different frequencies in circulating and tissue-bound B cells.

We isolated SARS-CoV-2–reactive Abs not only from naive B cells, but also from the GC, memory, and PC compartments, indicating that GC responses contribute to but do not represent the exclusive mechanism leading to the generation of these cells. The observation that iGR Abs lost the ability to bind SARS-CoV-2 S-protein is indicative of de novo Ag recognition of SARS-CoV-2 generated during the course of GC affinity maturation. It will be of particular interest to explore whether Ags promoting this process are encountered during natural infection or vaccinations targeting distinct pathogens. The majority of the mAbs generated in this study specifically recognized the S-protein of SARS-CoV-2. Unlike investigations on pre-existing SARS-CoV-2–reactive T lineage cells, which may result from previously encountered endemic coronaviruses, our data do not support a comparable mechanism but may rather implicate involvement of commensal microbiota. Pre-existing Abs to viral pathogens with cross-reactivity to commensal microbiota, including somatically mutated Abs from memory B cells, were reported in the context of HIV-1 infection (44–46), suggesting that an immune response to viral Ags may be primed by microbiota. In HIV-1 infection, this pre-existing immune repertoire was associated with a diversion of the immune response to a nonneutralizing Ab response. In our study, we observed predominantly a loss of S-protein reactivity of iGR Abs without changes to recognition of commensal microbiota. It is imperative to note that the MET-1/2 microbiota used in our experiments represent model communities of intestinal commensals. We would anticipate observing increased frequencies of reactivity to autologous commensal microbiota of the oral cavity and upper respiratory tract. Although encountering SARS-CoV-2 S-protein either by natural infection or by vaccination results in potently neutralizing Ab responses, it will be important to explore whether engagement of populations of pre-existing memory B cells has consequences for the duration of protective Ab titers.

SARS-CoV-2 infections occur less frequently and with less severe manifestations in children compared with the adult population (47, 48). Our observation of pre-existing SARS-CoV-2–reactive B cells from tonsillar tissue of pediatric patients combined with reported decreased tonsillar B cell counts during aging (49) raises the possibility that age-related changes of the cellular composition of the mucosal immune system in the upper respiratory tract may contribute to the immunopathology of SARS-CoV-2.

We thank Dr. Nathalie Simard (Temerty Faculty of Medicine Flow Cytometry Facility, University of Toronto) for assistance with the purification of SARS-CoV-2–reactive B cells. We also thank Dr. Emma Allen-Vercoe (Department of Molecular and Cellular Biology, University of Guelph) for the generous gift of MET-1 and MET-2 commensal microbiota.

The authors declare no conflict of interest.