To investigate the baseline frequency of SARS-CoV-2 S-reactive B cells prior to any SARS-CoV-2 S antigen encounter and the role of these cells in the vaccination-induced anti-SARS-CoV-2 humoral immune response, we sampled peripheral blood mononuclear cells (PBMCs) from 11 SARS-CoV-2-uninfected donors prior to vaccination and matched PBMCs two weeks after the second vaccine dose ( Table S1 ). S-reactive naïve B cells (CD3^-CD19⁺CD27^-IgD⁺IgG^-) and memory B cells (CD3^-CD19⁺CD27⁺IgD^-IgG⁺) were assessed with two fluorescence-labelled SARS-CoV-2 S probes in each donor ( Figure 1A , Figure S1 ). SARS-CoV-2-reactive B cells were found in both naïve and classical memory B-cell subsets prior to vaccination, with median frequencies of 0.091% and 0.044%, respectively ( Figure 1B ). Although different fluorescent probes were used, consistent frequencies of SARS-CoV-2-reactive B cells in unexposed individuals were identified in a recent study (25) (0.11% of naïve and 0.049% of classical memory B cells). After the second vaccination dose, the frequency of SARS-CoV-2-reactive memory B cells appeared to double (median from 0.044% to 0.083%), while the frequency of SARS-CoV-2-reactive naïve B cells remained unaltered (from 0.091% to 0.095%, Figure 1B ). The amplification of SARS-CoV-2-reactive memory B cells suggests the activation of humoral immunity induced by vaccination. To understand the potential origins of the amplification of SARS-CoV-2 reactive memory B cells post-vaccination (hereafter post-M), we performed a correlation analysis of the frequency of these cells with prevaccination SARS-CoV-2-reactive naïve (pre-N) and memory B cells (pre-M). The results showed that the frequency of post-M cells was positively correlated with the frequency of both pre-N and pre-M cells, suggesting that both SARS-CoV-2-reactive naïve and preexisting memory B cells may cocontribute to the development of anti-SARS-CoV-2 humoral immunity after vaccination ( Figure 1C ). This result is in line with recent studies on serum and mAbs from unexposed healthy donors and COVID-19 convalescents, suggesting that the antibody response against SARS-CoV-2 could stem from the naïve B-cell repertoire and cross-reactive memory B cells induced by prior infection with endemic HCoVs (9, 22, 25, 29).

To understand naturally existing SARS-CoV-2-reactive B cells in more detail, we sorted single SARS-CoV-2-reactive pre-N, pre-M, and post-M cells from 9 of the 11 donors to amplify their antibody genes. We acquired 105 heavy chain (HC) and paired light chain (LC) sequences by reverse transcription-polymerase chain reaction ( Figure 2A ). Of the 105 pairs of sequences, 48, 20, and 37 were derived from pre-N, pre-M, and post-M cells, respectively. Sequence analysis showed noticeable enrichment of the IGHV4-34 (22.9% of total) and IGHV3-30 (20.8%) genes in pre-N cells, with both genes present in 5 of the 9 donors. IGHV3-30 (25.0%) was also enriched in pre-M cells ( Figure S2A ). In addition, we observed an increased presence of IGKV3-20 (24.8%) in post-M cells. The heavy chain third complementarity-determining region (HCDR3) amino acid lengths of pre-N (average ~17), pre-M (~16), and post-M (~15) cells were commonly distributed and showed no discrepancy. The LCDR3 lengths in pre-N, pre-M, and post-M cells are comparable (average ~9, Figure S2B ). The antibody genes identified in pre-N cells exhibited germ-like properties in both the variable heavy (V_H) and light (V_L) chains (100% and 99.7% median identity to the germline, respectively). However, the sequences obtained from pre-M and post-M cells exhibited more SHMs, with a median of 93.4% and 94.2% identity to the germline for V_H and 95.2% and 96.9% for V_L ( Figure S2C ), consistent with the possible maturation of pre-M cells in response to prior HCoV infection (8, 26, 29–31).

Next, we cloned and recombinantly expressed 49 immunoglobulin G (IgG) mAbs, 33 of which were derived from pre-N cells, 11 from pre-M cells, and 16 from post-M cells ( Figure 2A , Figure S4A ). We identified 19 mAbs from multiple donors that bound to the SARS-CoV-2 S protein, which were used further evaluated ( Figure S3 , Figure S4A ). Among these mAbs, 4 were derived from pre-N cells, 5 from pre-M cells, and 10 from post-M cells. The pre-N cell-derived mAbs (4/33, 12.12%) bound SARS-CoV-2 with a substantially lower positive rate and binding activity than the pre-M cell (5/11, 45.5%) and post-M cell-derived mAbs (10/16, 62.5%), consistent with the fact that they were derived from naïve B cells that had never encountered SARS-CoV-2 antigen ( Figure 2B , Figure S4A ). We further defined the targeted domain of these SARS-CoV-2-binding mAbs by ELISA using the subunits of the SARS-CoV-2 S protein. Among the 4 pre-N cell-derived mAbs, 3 targeted the S2 subunit, and 1 targeted a non-RBD/NTD region on the S1 domain. Among the 5 pre-M cell-derived mAbs, 3 targeted the S2 subunit, and 2 targeted the undetermined domain on S protein. While among the 10 post-M cell-derived mAbs, 5 targeted the RBD, 2 targeted the S2 subunit, and the remaining 3 mAbs targeted the NTD, a non-RBD/NTD region on the S1 domain and the undetermined domain on S protein ( Figure 2C , Figure S4B ). Previous reports have found that the B-cell response in the late stages of vaccination showed amplification of antibodies targeting the S1 domain. In contrast, the early stages of the response to the vaccine mainly included the recall of antibodies targeting the S2 domain (32).

We then compared the sequence characteristics of the SARS-CoV-2-binding mAbs from pre-N cells and post-M cells. Compared with the mAbs from post-M cells, the mAbs from pre-N cells exhibited a higher degree of V_H germline identity ( Figure 2E ). However, no discrepancy in V_L germline identity or CDR3 length was observed ( Figure 2D ). Among the 4 mAbs from the pre-N cells, 3 mAbs employed the IGHV4-34 gene, while no propensity for V-gene pairing was evident among them ( Table 1 ). However, mAbs from post-M cells exhibited diversity in V gene usage ( Table 1 ).

A previous study has shown that germline antibodies are more likely to exhibit polyreactivity compared to mature antibodies with high levels of SHM, and a study of SARS-CoV-2-reactive naïve B cells revealed the presence of polyreactivity in these cells that may be involved in the physiopathology of COVID-19 (25, 33). We then further assessed the polyreactivity of SARS-CoV-2-binding naïve mAbs, no obvious polyreactivity was observed (Data not show).

Whether ancestral SARS-CoV-2 S-reactive naïve B cells prior to antigen exposure possess binding reactivity to the newly emerging variants remains to be further characterized. Therefore, we tested the binding reactivities of SARS-CoV-2 S-reactive mAbs derived from naïve B cells to the S of circulating viral variants, including Alpha (B.1.1.7), Beta (B.1.351), Gamma (P.1), Delta (B.1.617.2), and Omicron subvariants [BA.1 (B.1.1.529.1), BA.2 (B.1.1.529.2), BA.3 (B.1.1.529.3), and BA.4/5 (B.1.1.529.4/5)]. We also evaluated their binding to the S of other 6 HCoVs in circulation prior to SARS-CoV-2 emergence: SARS-CoV, MERS-CoV, HCoV-OC43, HCoV-HKU1, HCoV-NL63, and HCoV-229E. The results revealed that the 4 anti-SARS-CoV-2 mAbs showed broad-spectrum reactivity to at least 2 variants of SARS-CoV-2. VA021-44, the only mAb targeting the S1 domain, showed the best reactivity against all tested variants except for Omicron BA.1 ( Figure 3A ). In addition, VA021-44 also cross-bound to SARS-CoV, similar to the binding profile of the broad-spectrum nAb S309, although VA021-44 typically exhibited lower binding activity. However, none of the mAbs exhibited cross-reactivity to MERS-CoV or the 4 endemic HCoVs tested. We further tested the binding reactivities of SARS-CoV-2 S-reactive mAbs derived from preexisting memory B cells to these S proteins. And 3 showed broad-spectrum reactivity to most of the variants of SARS-CoV-2 ( Figure S4C ). Among them, 2 mAbs exhibited cross-binding to all 4 endemic HCoVs and other HCoVs. The results indicate that the naïve and preexisting memory mAbs exhibited binding to primarily SARS-CoV-2 and its variants, but cross-reactivity of the naïve mAbs was detected only for SARS-CoV but not to MERS-CoV or other endemic HCoVs. The lack of binding to other HCoVs suggests that isolated SARS-CoV-2-reactive naïve B cells are more likely to have originated from the SARS-CoV-2-reactive naïve repertoire rather than have been induced by prior HCoV infection. Finally, using pseudotyped SARS-CoV-2 S virus, we performed a neutralization analysis of the SARS-CoV-2-binding mAbs derived from naïve B cells. Unfortunately, the results showed that none of the 4 mAbs displayed potent neutralizing activity ( Figure 3B ).

To further understand the evolutionary characteristics of these SARS-CoV-2-reactive naïve B cells after antigenic stimulation, we analysed the V genes used by mAbs isolated from a single donor. Interestingly, we found that a class of mAbs using the same IGHV4-4/IGKV3-20 gene was present at both pre- and post-vaccination and that all of these mAbs bound SARS-CoV-2 S. A pre-N cell-derived mAb (VA021-31) and 3 corresponding mature post-M cell-derived mAbs (VC021-49, VC021-62, and VC021-72) isolated from donor V021 that used identical IGHV4-4 and IGKV3-20 genes were then selected and cloned for evolutionary analysis ( Figure 4A , top). All 4 cloned IGHV4-4/IGKV3-20 mAbs targeted the S2 domain. These mAbs also shared similar CDR3 amino acid sequences, especially the 3 mature post-M cell-derived mAbs, which had only 2 and 3 amino acid differences in HCDR3 and LCDR3, respectively. Compared to VA021-31, the mature mAbs isolated from post-M cells had accumulated more SHMs as a result of being more mature after antigenic stimulation. The mature mAbs all bear equal-length HCDR3 and LCDR3, with a shorter HCDR3 and a slightly longer LCDR3 than VA021-31 ( Figure 4A , top). We inferred their sequence-based evolutionary characteristics by constructing a molecular phylogenetic evolutionary tree analysis of such mAbs. The bootstrap values were 85% and 55% for VH and VL, respectively ( Figure 4A , bottom).

We then evaluated the binding activity of these IGHV4-4/IGKV3-20-derived mAbs to ancestral SARS-CoV-2 and its variants, as well as other preexisiting HCoVs, by ELISA. VA021-31 engaged all 15 CoVs we tested, which showed stronger binding to S of ancestral SARS-CoV-2 and several variants. Regarding mature VC021-49, VC021-62, and VC021-72, their binding activities to multiple SARS-CoV-2 variants and HCoVs were enhanced, but their binding activities to several newly emerging variants were lost ( Figure 4B ). This result seems to indicate that VA021-31 lineage maturation upon vaccination promoted convergent evolution under vaccine antigen stimulation ( Figure 4C ). Collectively, these data indicate that SARS-CoV-2-reactive naïve B cells, after antigenic activation, were able to undergo SHM and thus enhance their binding activity to multiple related viruses, but there was a reduction in their binding breadth. Although rather weak, we observed broad cross-reactivity of these IGHV4-4/IGKV3-20-derived mAbs to all preexisting HCoVs. This finding may suggest that such a naïve B-cell germline is a broad precursor commonly shared by HcoVs.

A total of 11 donors who had received two doses of SARS-CoV-2 inactivated vaccines (Sinovac, Beijing, China) were recruited from the First People’s Hospital of Chenzhou, Hunan Province, China ( Supplementary Table 1 ). None of the vaccinees had subsequent SARS-CoV-2 or variant exposure, as determined by frequent PCR testing and questionnaire administration during the observation period. Peripheral blood samples from the donors were collected before vaccination (day 0) and 14 days after the second vaccine dose (approximately equivalent to 45 days after the first dose). Each participant in this study had reached the age of 18 and was free from any underlying illnesses. PBMCs and plasma were isolated and then frozen in liquid nitrogen and a −80°C freezer, respectively. Each participant signed a written consent form. The study protocol was approved by the First People’s Hospital of Chenzhou (V.3.0, 2021001).

SARS-CoV-2 S labelled with Alexa Fluor 488 and Alexa Fluor 647 (Thermo Fisher Scientific, Waltham, MA, USA), as probes, was used to identify SARS-CoV-2-reactive B cells. PBMCs in liquid nitrogen were rapidly thawed in a 37°C water bath before being incubated overnight in RPMI 1640 medium containing 10% foetal bovine serum (FBS) in a 37°C, 5% CO₂ incubator (Thermo Fisher Scientific, Waltham, MA, USA). Following two washes with staining buffer (2% FBS and 1% HEPES in PBS), PBMCs were stained for 30 min with the LIVE/DEAD Fixable Blue Dead Cell Stain Kit (Thermo Fisher Scientific, Waltham, MA, USA) and then treated with an anti-ACE2 antibody and rabbit polyclonal antibody (Sino Biological, Beijing, China) to block the S-ACE2 binding site. After being washed, PBMCs were stained with the SARS-CoV-2 S probe mixture for 30 min at 4°C while avoiding light. PBMCs were washed two times to remove excess reagents. The treated PBMCs were stained with antibodies that had been pretitrated to an optimized dilution and fluorescently labelled in 96-well V-bottom plates at 4°C for 30 min. The fluorescently labelled antibodies were as follows: BUV737 mouse anti-human CD3 (SK7), PE mouse anti-human IgG (G18-145) (BD Biosciences, Franklin Lake, NJ, USA), PerCP/Cyanine5.5 mouse anti-human CD19 (HIB19), PE-cy7 mouse anti-human CD27 (M-T271), and PE/Dazzle 594 mouse anti-human IgD (IA6-2) (BioLegend, San Diego, CA, USA). After washing, single SARS-CoV-2 S-reactive naïve B cells (CD3^-CD19⁺CD27^-IgD⁺IgG^- SARS-CoV-2 S-Alexa Fluor 488⁺ SARS-CoV-2 S-Alexa Fluor 647⁺) and single memory B cells (CD3^-CD19⁺CD27⁺IgD^-IgG⁺ SARS-CoV-2 S-Alexa Fluor 488⁺ SARS-CoV-2 S-Alexa Fluor 647⁺) were sorted using a MoFlo XDP (Beckman Coulter, Brea, CA, USA) flow cytometer into a 96-well PCR plate (Thermo Fisher Scientific, Waltham, MA, USA) containing lysis buffer (10×PBS, 5 mM DTT, 1% IGEPAL, 20 ng/μl random primer, 2 U/μl RNasin Ribonuclease Inhibitor), which was placed at -80°C before subsequent processing. Flow cytometry data were analysed using FlowJo software version 10.8.1 (Tree Star, San Carlos, CA, USA).

Antibody variable gene mRNA transcripts, VH, variable kappa (Vκ), and variable lambda (Vλ), were amplified by RT−PCR in a previously reported manner (45). In brief, cDNA was synthesized using SuperScript IV Reverse Transcriptase (Thermo Fisher Scientific, Waltham, MA, USA), followed by two rounds of nested PCR. The PCR products were purified and sequenced to determine the heavy and kappa/lambda chain sequences, and the variable region sequences of the heavy and light chains were synthesized and inserted separately into Ig heavy- or light-chain expression vectors: AbVec2.0-IGHG1, AbVec1.1-IGKC, and AbVec1.1-IGLC2 (Addgene, Watertown, MA, USA). IgBLAST and IMGT/V-QUEST were used to assess antibody CDR3 length, variable region gene usage, and mutation information.

To express recombinant mAbs, heavy chain and matched light chain antibody expression vectors (0.5 μg/mL) were chemically cotransfected at a 1:2 ratio into 2×10⁶ cells/ml FreeStyle 293-F cells using 8 μg/mL polyethyleneimine (Polysciences, USA) transfection reagent. Transfected cells were propagated at 37°C and in 8% CO₂ under constant shaking at 105 rpm in serum-free FreeStyle 293 Expression Medium (Thermo Fisher Scientific, Waltham, MA, USA) for 5-7 days. Cell supernatants were harvested by centrifugation at 4,000 × g and filtered through a 0.22 μm filter after culture. The suspension was purified using HiTrap Protein A columns HP (Cytiva, Marlborough, MA, USA) with an ÄKTA pure system (Cytiva, Marlborough, MA, USA), washed with PBS, and then eluted with glycine (pH 3.0) into a collection tube containing Tris-HCl buffer (pH 8.0). Buffer exchange was accomplished by dialyzing three times in PBS, and the final concentration of purified antibodies was determined using a Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA). Purified antibodies were then kept at -40°C until further use.

S proteins and S subunits used in the ELISA were obtained commercially (Sino Biological, Beijing, China), including: 1) the S proteins of HCoVs: SARS-CoV-2, SRAS-CoV, MERS-CoV, HCoV-OC43, HCoV-HKU1, HCoV-NL63 and HCoV-229E (40589-V08B1, 40634-V08B, 40069-V08B, 40607-V08B, 40606-V08B, 40604-V08B, and 40605-V08B, respectively); 2) the S proteins of SARS-CoV-2 variants: Alpha (B.1.1.7, 40589-V08B6), Beta (B.1.351, 40589-V08B11), Gamma (P.1, 40589-V08B10), Delta (B.1.617.2, 40589-V08B16), and Omicron subvariants BA.1 (B.1.1.529.1, 40589-V08B33), BA.2 (B.1.1.529.2, 40589-V08H28), BA.4/5 (B.1.1.529.4/5, 40589-V08H32) and BA.2.75 (B.1.1.529.2.75, 40589-V08H36); and 3) the SARS-CoV-2 S1, S2, RBD and NTD subunits (40591-V08H, 40590-V08B, 40592-V08B and 40591-V49H, respectively). The reactivity of mAbs to S proteins or subunits of HCoVs or variants was assayed by ELISA. Briefly, 96-well ELISA plates (Corning, NY, USA) were coated with S proteins of corresponding CoVs or SARS-CoV-2 S subunits (2 μg/ml) in PBS at 100 μl/well and incubated overnight at 4°C. The plates were washed five times with PBS containing 0.5% Tween-20 (PBS-T solution) and blocked with 200 μl of blocking buffer per well (PBS containing 2% FBS and 2% bovine serum albumin (BSA) (Sigma-Aldrich, St. Louis, MO, USA)) for 2-4 hours at room temperature (RT). The plates were washed five times with PBS-T, and 100 μl of isolated mAbs (10 μg/ml) per well in blocking buffer were added to the wells and incubated for 1 hour at 37°C. The plates were washed five times with PBS-T, and then 100 μl per well Peroxidase-AffiniPure goat anti-human IgG (H+L) antibody (Jackson ImmunoResearch, PA, USA) was added at a 1:5,000 dilution in blocking buffer for 1 hour at 37°C. The plates were washed five times with PBS-T and developed with 100 μl per well 3,3’,5,5’-tetramethylbenzidine (TMB) substrate (Thermo Fisher Scientific, Waltham, MA, USA) for 5 min at RT. The reaction was terminated by addition of 50 μl per well 1 M H₂SO₄. Absorbance was measured using an RT-6100 (Rayto, Shenzhen, China) enzyme analyser at 450 nm. All ELISA experiments were performed in duplicate, and the data were averaged over three independent experiments. Positive binding of a mAb was defined by OD450 nm (optical density at 450 nm) ≥ 0.40. BSA and an unrelated hepatitis C virus (HCV) antigen were used as negative antigen controls for the ELISA binding test. An approved anti-CoV broad-spectrum neutralizing monoclonal antibody, S309 (46), was used as a positive control, and an unrelated anti-HCV antibody HNC-5 was used as a negative control.

SARS-CoV-2 neutralization by the mAbs was measured by a reduction in luciferase activity after infection of ACE2-expressing 293T cells (ACE2-293T) with the S pseudovirus, as previously described (47, 48). Briefly, ACE2-293T cells were seeded in a 96-well plate (Corning, NY, USA) (2×10⁴ cells/well) and cultivated for 24 hours at 37°C with 5% CO₂. Before the experiment, 165 μl/well of diluted mAbs (1 μg/ml and 10 μg/ml) in DMEM with 50 μg/mL streptomycin was applied in triplicate to a 96-well plate. Then, 75 μl of pseudovirus diluted to 10,000 TCID₅₀/ml was added and incubated at 37°C for 1 hour. After cell incubation, the supernatant was removed, and 70 μl of the virus-antibody mixture was added and incubated at 37°C for 24 hours before being replenished at 100 μl/well and incubated for another 24 hours. After discarding the supernatant, 50 μl of cell lysis buffer was added and incubated at RT for 30 minutes on a shaking table, 30 μl of which was then transferred to an optical white bottom plate. Each well received 50 μl of luciferase reagent to measure luciferase activity using relative light units (RLU). Negative (VC, virus+cells) and blank (NC, cells) controls were included in parallel. The percentage of neutralization was calculated by the following equation:

Positive neutralization of mAbs was defined by ≥50% neutralization of the pseudovirus. S309 and HNC-5 were the positive and negative controls, respectively.