In this study, a total of 10 convalescent COVID-19 individuals infected with SARS-CoV-2 were enrolled, and the peripheral bloods were collected. The study was approved by the institutional review board of the School of Public Health in accordance with the Declaration of Helsinki, and written informed consent was obtained.

To detect the plasma titers of total antibodies (Ab), IgG and IgM against SARS-CoV-2 RBD, we performed commercial enzyme-linked immunosorbent assay (ELISA) kits (Beijing Wantai Biological Pharmacy Enterprise), according to the manufacturer’s instructions. The Ab-ELISA kit is based recombinant viral antigen using a double-sandwich reaction form. The IgG-ELISA kit is an indirect ELISA assay, and the IgM-ELISA kit is based on μ-chain capture method. The samples were initially tested undiluted, and the positive samples with the signal to a cutoff ratio (S/CO) >=10 were further diluted (1:10, 1:100, 1:1,000 and 1:10,000) by PBS buffer containing 20% newborn bovine serum (NBS) and tested again. The titers for Ab, IgG and IgM antibody were calculated via S/CO multiplied by the maximum dilution factors.

For SARS-CoV and SARS-CoV-2 S protein, a gene encoding the ectodomain of a prefusion conformation-stabilized S protein was synthesized, composed of SARS-CoV gene sequence (GenBank: ABF65836) or SARS-COV-2 gene sequence (GenBank: MN908947), a C-terminal T4 fibritin trimerization motif, an HRV3C protease and 8xHisTag. To determine the blocking capacity of mAbs, we also synthesized gene of SARS-CoV-2 S fluorescin probe comprising SARS-COV-2 gene sequence, a C-terminal T4 fibritin trimerization motif, an HRV3C protease, 8xHisTag and a C-terminal green fluorescent protein (mGamillus). Moreover, to express mutate, wildtype SARS-CoV-2 RBD, residues 319-518 fused to mouse IgG1 Fc domain. For mutate RBDs, different selected amino acid of RBD were substituted by alanine or arginine on purpose ( Table S3 ). Recombinant expressions of these proteins were performed by the ExpiCHO™ expression system (Thermo Scientific, A29133). Briefly, plasmids encoding targeted proteins were transiently transfected into ExpiCHO cells by using ExpiFectamine™ CHO transfection kit (Thermo Scientific, A29129). The cell-free supernatants were obtained 7 days after transfection by centrifugation and filtration with a 0.22 μm filter. Subsequently, the S-related proteins (SARS-CoV S, SARS-CoV-2 S and SARS-CoV-2 S fluorescin probe) were purified by Ni Sepharose Excel resin, and RBD fused to mouse IgG1 Fc domain by Protein A column, and the S and RBD proteins were stored in the PBS buffer.

RBD specific B cells were obtained in the same way as previously reported. PBMCs collected from 10 individuals were incubated with a cocktail containing live/dead-Aqua, CD3-PE-Cy7, CD19-BV786, CD27-BV650, anti-human IgM-PerCP-Cy5.5, anti-human IgG-BV421, RBD-FITC and biotinylated RBD, followed with Streptavidin-APC binding to biotinylated RBD. The RBD-specific memory B cells were identified as live+CD19+CD3-CD27+IgG+RBD+, and then the single specific memory B cells were sorted by fluorescence activated cell sorting on an Aria III sorter (BD Biosciences) into 96-well PCR plates containing 20 μL per well of lysis buffer [5 μL of 5×first strand buffer (Invitrogen), 1.25 μL dithiothreitol (Invitrogen), 0.5 μL RNase Out (Invitrogen), 0.0625 μL Igepal (Sigma)]. Plates were stored at -80°C prior to reverse transcription reaction.

Antibody variable genes (IgH, Igλ and Igκ) were amplified by RT-PCR and nested PCR reactions as previously described (33). The paired heavy and light chains were then cloned into expression vectors containing the constant regions of human IgG1 and light chain. The paired heavy and light chain expression cassettes were then transiently co-transfected into ExpiCHO cells with equal amounts of plasmids according to the manufacturer’s instructions (Life Technologies), and antibodies were purified from culture supernatant 5-7 days after transfection, using a recombinant protein-A column (GE Healthcare).

Antibody gene repertoire was analyzed for the variable region of IgG heavy and light chains using the IMGT V-quest webserver (http://www.imgt.org/IMGT/vquest). Phylogenetic analysis of antibody gene was performed by ggtree R package (34).

The binding activity of the mAbs against SARS-CoV-2 S protein were determined using an indirect ELISA. The mAbs were added to antigen-coated microwell plates, and incubated at 37°C for 30 min. Then, incubation of HRP-conjugated anti-human antibody at 37°C for 30 min to detect the bound mAbs, followed by washing five times. Finally, substrate solution was added to the wells for 15 min at 37°C, and reaction was stopped by adding 50 μL of 2 M H₂SO₄. The optical density (OD) was measured at 450 nm with a reference wavelength of 630 nm. In addition, binding activity of the mAbs against SARS-CoV-2 RBD and SARS-CoV S protein were determined by same method.

To determine the critical residues, mAbs were conjugated with horse radish peroxidase (HRP). Microwell plates were pre-coated with mutate SARS-CoV-2 RBD and wildtype RBD at 100 ng per well. mAbs-HRP were performed two times gradient dilution with 4 μg/mL begin and incubated at 37°C for 30 min followed by washing five times. Substrate solution was incubated for 15 min at 37°C and stopped by 50 μL of 2 M H₂SO₄. OD was determined at 450 nm with a reference wavelength of 630 nm. The binding activity of mAbs against wildtype and mutated RBD was calculated by area under the curve (AUC), and the influence of residues was assessed by binding activity reduction against corresponding mutate SARS-CoV-2 RBD. Residues reducing binding activity by more than 75% are identified as critical residues.

For SARS-CoV-2 S protein-blocking assay, mAbs were pre-made as 2-fold serial dilutions using DMEM containing 10% FBS. Aliquots (44 μL per well) of diluted samples and S protein probes (11 μL per well) were mixed in a 96-well plate with U shaped bottom. Half of the culture medium (50 μL) of 293T-ACE2iRb3 cell plate were gently removed, and 50 μL of sample/probe mixtures were added to each well. Cell image acquisitions performed with Opera Phenix (green, red and near-infrared channels in confocal mode) using a 20x water immersion objective at 1-hour after probe incubation in wash-free and live-cell conditions.

All quantitative image analyses were based on images that acquired by Opera Phenix. All image data were transfer to Columbus system (version 2.5.0, PerkinElmer Inc) for analysis. Multiparametric image analysis was performed as described in the following. The signals of blue channel or near-infrared channel were used to detect the nucleus. As the ACE2 is a membrane protein, the signals of ACE2-mRuby3 (red channel) were used to determine the cell boundary. Then, the cells were further segment into the regions of membrane (outer border: 0%, inner border: 15%), cytoplasm (outer border: 20%, inner border: 45%), and nucleus (outer border: 55%, inner border: 100%). The MFI of probe channel (Ex488/Em525) in the cytoplasmic region (cMFI). The MFI of ACE2-mRuby3 (Ex561/Em590) on the membrane were also calculated for inter-well normalization. The cMFI inhibition ratio (%) of the test sample was calculated using the following equation: [(cMFIpc-cMFItst)/(cMFIpc-cMFIblk)]×100%. In this formula, the cMFIpc is the cMFI value of probe-only well (as positive control), the cMFItst is the cMFI value of test well and the cMFIblk is the cMFI value of cell-only well. For each plate, five replicates of probe-only well and one cell-only well were included. The blocking capacity of mAbs were expressed as IC50.

SARS-CoV-2 pseudovirus using VSV carrying the SARS-CoV-2 spike protein were produced according to our previous study. Briefly, SARS-CoV-2 S gene was codon optimized for expression in human cells and truncated with 18 amino acids at the C-terminal, then was cloned into the eukaryotic expression vector pCAG to obtain pCAG-nCoVSde18. The plasmid pCAG-nCoVSde18 was transfected into Vero-E6. VSVdG-EGFP-G (Addgene, 31842) virus was inoculated into cells expressing SARS-CoV-2 Sde18 truncated protein and incubated for 1 hour. Then the VSVdG-EGFP-G virus was removed from the supernatant and anti-VSV-G rat serum was added to block the remaining VSVdG-EGFP-G infection. The progeny virus will carry SARS-CoV-2 Sde18 truncated protein. After VSVdG-EGFP-G infection, supernatant was collected, centrifuged and filtered (Millipore, SLHP033RB) to obtain the SARS-CoV-2 pseudovirus without debris. SARS-CoV pseudovirus was constructed by the same method. Finally, pseudovirus was stored for use at -80°C.

To determine the neutralizing capacity, mAbs with 2-fold serial dilutions with 10% FBS-DMEM from 2 μg/mL were mixed with diluted SARS-CoV or SARS-CoV-2 pseudovirus (MOI = 0.05), incubated at 37°C for 1 hour. A mixture of 80 μL was added to the precoated BHK21-hACE2 cells. After incubation for 12 hours, post-infection cells were fluorescently imaged using Opera phenix or Operetta CLS (Perkinelmer), and quantitatively analyzed by Columbus image management analysis software to detect the number of green, fluorescent positive cells. The inhibition rate was calculated by reduction of GFP positive cells with presence of mAbs compared with the untreated control wells.

Briefly, the unlabeled mAbs (50 μg per well) or PBS were added to RBD-coated 96-well microplates and then incubated for 30 min at 37°C. Next, HRP-conjugated mAbs were added at selected dilutions, at which OD readings was ~1.5 with PBS present. After incubation for 30 min at 37°C, the microplates were rinsed, and the color was developed. The competitive ability was measured quantitatively by comparing OD in the presence and absence of competitor mAbs and transformed using the formula log₂ (OD_inhibited/OD_original). For mAbs to be clustered by competitive ability, clustering distance was calculated by Euclidean, and mAbs were clustered by ward. D2 method using pheatmap R package (version: 1.0.12).

For antibody response evaluation of RBD_{Glycan420,475}, RBD_{Glycan458,475}, RBD_{Trucation455-491} and RBD_{Trucation470-491}, BALB/c mice were immunized with various RBD proteins at 20 μg/dose with FH002C through intramuscular injection. Serum samples were collected at Week 0 and 2 via retro-orbital bleeding to measure the antibody titers.

Microplates were pre-coated with recombinant antigens of SARS-CoV or SARS-CoV-2 S protein. For detections, serial-diluted (2-fold) serum samples (100 μL per well) were added into the wells, and the plates were incubated at 37°C for 30 min, followed by washing with PBST buffer (20 mM PB7.4, 150 mM NaCl and 0.05% Tween 20). Then, HRP-conjugated anti-mouse IgG solutions (100 μL per well) were added. After a further 30 min incubation followed by washing, TMB chromogen solution (100 μL per well) was added into the well. 15 min later, the chromogen reaction was stopped by adding 50 μL of 2 M H₂SO₄, and the OD450-630 was measured. The IgG titer of each serum was defined as the dilution limit to achieve a positive result (>median+3×SD of ODs of negative controls).

To compare continuous variables, the Mann-Whitney U test and non-paired t test were performed. Linear regression model and Spearman test were used for correlation analyses. For difference analysis, p values less than 0.05 are considered statistically significant. GraphPad Prism (version 8.0.1) was used for all statistical calculations. Analysis of protein structure was performed by PyMOL Molecular Graphics System (version 2.3.0).

We collected blood samples from 10 convalescent COVID-19 patients ( Table S1 ). The humoral immune response was efficiently elicited in these convalescent individuals, as indicated by high plasma titers of RBD-specific IgG with differences of less than an order of magnitude ( Table S2 ). Since the plasma neutralizing capacity strongly correlates with RBD-specific total antibody and IgG titers, indicating that the RBD of the S protein is the dominant target of nAbs elicited by SARS-CoV-2 infection ( Figure S1 ). The proportion of RBD-specific B cells in memory B cells ranged from 0.03% to 0.18%, and RBD-specific memory B cells contained a higher percentage of the IgG subtype than the IgM subtype, revealing that SARS-CoV-2 infection efficiently promoted B cell receptor (BCR) class switching and affinity maturation in convalescent individuals, thus eliciting strong humoral immune response against SARS-CoV-2 ( Figures S2A–C ). Then, 77 RBD-specific monoclonal antibodies (mAbs) were obtained from the 10 convalescent individuals for comprehensive feature description ( Figure S2D ).

To characterize SARS-CoV-2 RBD-specific mAb repertoire usage, we compared sequences with the well-defined naïve repertoire of the IMGT database to obtain the assigned germline V region. Based on the exclusion of clonal expansion in P01, P03 and P04, 67 unique clonotypes were identified. Notably, 7 out of 8 antibody sequences obtained from P03 were highly conserved, except for P03-3B1, illustrating that this BCR clonotype was the immunodominant clone in P03 induced by SARS-CoV-2 infection ( Figure 1A and Figure S3 ). Furthermore, enrichment of multiple VH and VK/VL sequences was observed. In addition, VK1-39-derived light chains were most often combined with the heavy chain of various types of VHs to form antibodies, accounting for 22.4% (15/67) ( Figure S4 and Figure 1A ). The mean somatic hypermutation (SHM) rate of the heavy chain V region was similar among individuals, and the mean levels (2%) were comparable to those detected in the context of infections with other respiratory viruses ( Figure 1B ) (35–37). Additionally, even though the average length of CDRH3 of RBD-specific mAbs was consistent with that of the naïve repertoire (~15 amino acids), we observed significant enrichment of shorter CDRH3 sequences (11 amino acids) in VH3-53/66-derived mAbs, differing from influenza virus and human immunodeficiency virus (HIV)-1 mAbs ( Figure 1C and Figure S8A ) (38–40).

Subsequently, we assessed the binding activity of these mAbs to recombinant S protein and RBD protein fragment of SARS-CoV-2 using enzyme-linked immunosorbent assay (ELISA). mAbs presented diverse binding activity to SARS-CoV-2 S protein, of which 61.0% showed strong binding activity (EC₅₀< 1 μg/mL); this result suggested that infection with SARS-CoV-2 can effectively stimulate humoral immune response to produce a large number of specific high-affinity mAbs ( Figure S2A and Figure 1D ). Interestingly, while there was a correlation between binding activity to SARS-CoV-2 RBD and to S protein, some mAbs bound more strongly to the RBD, implying that their target epitopes were poorly presented in S protein to be recognized, due to coverage by another RBD monomer or the NTD domain ( Figure 1E ) (41). Next, we assessed the neutralizing activity of these mAbs using a vesicular stomatitis virus (VSV) pseudovirus model carrying the SARS-CoV-2 S protein, and the neutralization IC₅₀ potencies are shown in Figure S5C . In total, 80.5% (62/77) of these mAbs displayed neutralization of SARS-CoV-2, characterized as nAbs, among which 16 were identified as potent neutralizers with an IC₅₀< 0.1 μg/mL, 22 as moderate neutralizers with an IC₅₀ of 0.1-1 μg/mL, and 24 as weak neutralizers with an IC₅₀ of 1-100 μg/mL ( Figure 1F ). Surprisingly, none of the antibodies isolated from the convalescent individual P03 had a potent neutralizing capacity, although clonal expansion was efficiently elicited ( Figure S6 ).

Next, we investigate whether blocking S protein binding to ACE2 is the key neutralizing mechanism. The results showed that 61.0% of the specific mAbs had the ability to block entrance of the S protein into cells, with IC₅₀ values ranging from 4 ng/mL to 100 μg/mL ( Figure S5D and Figure 1G ). These blocking and neutralizing capacities were well correlated with the binding capacity ( Figures 1H, I ). In terms of the neutralization and blocking data, we found a good correlation, indicating that blocking the attachment of the SARS-CoV-2 S protein to receptor ACE2 was critical for the inhibition of SARS-CoV-2 infection by nAbs targeting RBD ( Figure 1J and Figure S7 ). However, some nAbs neutralized SARS-CoV-2 with weak blockade of SARS-CoV-2 S protein binding to ACE2 (28). Although the corresponding neutralizing mechanism has not been explained clearly, it is putative that binding of these nAbs may impede sequential conformational changes in the S protein.

To determine whether the affinity of SARS-CoV-2 RBD-specific mAbs efficiently evolves, we analyzed the association between the binding activity of mAbs and the duration of the immune response. The binding activity and neutralizing capacity of specific mAbs correlated with days after symptom onset, illustrating that the affinity of these specific mAbs can continuously evolve ( Figures 1K, L and Figure S9 ). However, the plasma anti-RBD IgG titer did not correlate with days after symptom onset for individuals, which might resulted from the limitation of the plasma samples ( Figure S1A ) (42). Taken together, humoral immune response is efficiently elicited by SARS-CoV-2 infection, and affinity of functional mAbs is constantly evolving.

Due to the 76% sequence identity between SARS-CoV S protein and SARS-CoV-2 S protein, there may be conserved epitopes. Accordingly, humoral immunity could utilize these conserved epitopes to produce cross-reactive mAbs response. In this study, 50.6% of SARS-CoV-2-specific antibodies also bound to SARS-CoV S protein, of which 11 (14.3%) recognized SARS-CoV S protein with strong binding activity (EC₅₀< 1 μg/mL) ( Figure 2A and Figure S5B ). As expected, these cross-binding mAbs also showed continuous affinity maturation ( Figure 2B ). In terms of genetic characteristics, it was observed that mAbs encoded by some VH germline genes, such as VH1-69, VH3-13, VH3-30 and VH4-46, had a tendency to broadly react with SARS-CoV and SARS-CoV-2 ( Figure 2C ). Additionally, cross-binding mAbs showed a tendency for lower SHM in the VH and JH regions, which might limit their affinity maturation, putatively due to the lower exposure of corresponding conserved antigenic sites ( Figures 2D, E ). Overall, these results confirmed an efficient cross-binding antibody response during SARS-CoV-2 infection and the presence of conserved antigenic sites inducing the maturation of cross-reactive mAbs.

Notably, the majority of mAbs with potent neutralizing activity against SARS-CoV-2 showed no reactivity with SARS-CoV, confirming that the unique antigenic sites in SARS-CoV-2 RBD, as immunodominant sites, more efficiently elicit high-affinity mAbs with potent neutralizing capacity than the conserved sites ( Figure 2F ). Subsequently, cross-binding mAbs were tested for their ability to neutralize SARS-CoV. Our results indicated that only P10-6G3, P07-4D10, P05-6H7 and P05-5B6 were identified as cross-neutralizing mAbs against both SARS-CoV and SARS-CoV-2, and P10-6G3 displayed higher binding activity to the SARS-CoV-2 S protein than the SARS-CoV S protein ( Figures 2F, G ). Combined with the weak binding activity, low affinity for SARS-CoV S protein is likely to be the reason why a large number of cross-binding antibodies could not neutralize SARS-CoV ( Figure 2G ). As the antigenic sites recognized by these cross-neutralizing mAbs were common between SARS-CoV and SARS-CoV-2, they could become the key targets for design of universal vaccines against SARS-like coronaviruses and selection of broad therapeutic antibodies.

In order to further explore the function of each antigenic sites in RBD, nAbs were classified into six clusters (C1-6) using competition binding assay, and the antigenic sites for C1-6 nAbs were termed as sites S1-6, respectively ( Figures S10 , S11 and Figure 3A ). Site S1 should be immunodominant antigenic sites, because the C1 nAbs accounted for relatively higher proportion ( Figure 3A ). Notably, nAbs in different clusters were efficiently induced for majority of convalescent SARS-CoV-2 patients, which embodied common immunogenic characteristics of these antigenic sites in RBD and indicated that SARS-CoV-2 infection can elicit antibody response using similar model ( Figure 3B ).

To determine the functional characteristics of nAbs targeting different sites, we analyzed the biochemical properties of these nAbs. The C1, C2, C4, C5 and C6 nAbs bound to SARS-CoV-2 S protein with comparable EC₅₀ values; however, the nAbs elicited by site S3 displayed lower binding activity to SARS-CoV-2 S protein ( Figure 3C ). Furthermore, the majority of C2, C3 and C4 nAbs showed cross-reactivity with SARS-CoV S protein and SARS-CoV-2 S protein, suggesting that these sites were conserved between SARS-CoV and SARS-CoV-2 ( Figures 3D, E ). Nevertheless, these conserved sites in the context of natural infection of SARS-CoV-2 just induced four potent cross-neutralizing mAbs (P10-6G3 targeting site S2, P07-4D10 targeting site S3, and P05-5B6 and P05-6H7 targeting site S4) against SARS-CoV and SARS-CoV-2, since most cross-binding nAbs showed obviously weaker affinity for SARS-CoV S protein than SARS-CoV-2 S protein, leading to inability to achieve cross-neutralization ( Figures 3C–G ). Notably, the conserved sites-directed nAbs showed lower neutralization capacity against SARS-CoV-2 than nAbs recognizing SARS-CoV-2 unique antigenic sites (sites S1, S5 and S6), except those targeting conserved site S3; based on the premise that all nAbs possess similar binding activity to the SARS-CoV-2 S protein, these nAbs targeting conserved sites are implied to have a disadvantage in blocking S protein binding to receptor ACE2, putatively due to less overlap between their antigenic sites and ACE2 footprint ( Figures 3C, F ). Markedly, C1 nAbs isolated from the overwhelming majority of convalescent individuals recognized antigenic site S1, and potently and specifically inhibited SARS-CoV-2 infection by blocking S protein attachment to ACE2, revealing that site S1 is an immunodominant antigenic site in SARS-CoV-2 RBD that efficiently elicits a strong NAb response during natural SARS-CoV-2 infection ( Figures 3B, F and Figure S12A ). The immunodominance of site S1 may result from either its accessibility in different conformations of SARS-CoV-2 S protein or the innate affinity of the corresponding C1 nAbs derived from the naïve B cell repertoire (VH 3-53/66 germline) ( Figure 3H ); the latter factor would lead to rapid affinity maturation of C1 mAbs without the need for a high level of SHM (31, 43, 44). In contrast, while large amplification of the same antibody clone against site S3 was exhibited in the convalescent individual P03, the nAbs derived from this antibody clone poorly inhibited SARS-CoV-2 infection, revealing that site S3 is possibly an immunodominant, but weakly neutralizing site ( Figure 3F , Figure S6 , Figure 1A and Figure S4 ). The characteristics of nAbs targeting different antigenic sites is further mapped to plasma neutralizing function. nAbs recognizing conserved antigenic site S3 showed significant clonal amplification in P03, which resulted in weaker plasma neutralization compared to the other COVID-19 patients. High proportion of specific antibodies against site S1 can ensure stronger neutralization activity against SARS-CoV-2 in COVID-19 patient plasma, but its cross-neutralization potency decreased significantly, and P04 was the most representative among these convalescent patients ( Figures 3I, J ). On the contrary, P05 and P08 plasma showed higher cross-neutralization potency, which might result from the numerous cross-neutralizing mAbs binding to the conserved antigenic sites. Thus, our findings confirm that conserved antigenic sites can broadly induce antibody response in COVID-19 patients, while cross-neutralization potency varies for different patients.

Taken together, sites S2, S3 and S4 were identified as conserved antigenic sites between SARS-CoV-2 and SARS-CoV that could induce cross-neutralizing antibody response and should be considered in the rational design of universal SARS-like coronavirus vaccines, and the remaining sites were unique antigenic sites for SARS-CoV-2. The difference in binding location possibly confer nAbs elicited by conserved antigenic sites might show weaker neutralization of SARS-CoV-2 than those elicited by unique antigenic sites.

It is more helpful to understand the functional characteristics of corresponding antibodies by analyzing the structural characteristics of each antigenic site. To determine the spatial position of sites S1-6, we performed a mutagenesis study by substituting ACE2-interactive and noninteractive residues with alanine or arginine in RBD, and then assessed the decreased binding activity of representative nAbs to mutant RBDs compared to the reference RBD (45). The reduction in binding activity to each mutant RBD is shown in Figure S14 . The spatial positions of all antigenic sites in RBD were simultaneously displayed to demonstrate their relative locations, which is critical to elucidate the functional characteristics of all neutralizing antigenic sites in RBD ( Figure S15A and Figure 4A ). Fortunately, numerous key epitopes in RBD have been identified by antigen-antibody complex structure and could be classified into five classes, which provides an important reference for this study (4, 28, 46). Site S1, S2 (and S3), S4 and S5 (and S6) sites are similar to Class 1, Class 4, Class 5 and Class 2/3 sites, respectively, according to their interactive residues and neutralizing characteristics of corresponding mAbs ( Figures 4A, B ).

SARS-CoV-2 unique sites S1, S5 and S6 largely overlapped with ACE2 footprint, supporting potent neutralizing activities of corresponding nAbs by efficiently blocking S protein binding to receptor, whereas conserved sites S2, S3 and S4 distant from ACE2 footprint is not conducive for neutralization ( Figure S15A ). However, site S4-specific nAbs, similar to broadly neutralizing mAb S2H97, might neutralize SARS-CoV-2 and SARS-CoV by promoting direct shed of S1 subunit, rather than blocking RBD attachment to ACE2. Site S2, S3 and S4 are highly conserved among SARS-CoV, SARS-CoV-2 and even other SARS-related viruses ( Figure 4C and Figure S15B ). Thus, nAbs targeting these antigenic sites retain strong neutralization potency against SARS-CoV-2 variants (B.1.1.7, B.1.351, B.1.1.28 and B.1.617.2), nevertheless, these variants escape neutralization of representative nAbs targeting SARS-CoV-2 unique sites ( Figure 4D ). As some neutralizing antigenic sites are hidden by the RBD in the lying-down state, this masking becomes an important immune escape mechanism of SARS-CoV-2 (41). The conserved sites, especially site S3, were highly concealed in the RBD lying-down state by adjacent RBD monomers. Even when the RBD was in the standing-up state, site S3 was not sufficiently exposed, and this inadequate exposure failed to improve the affinity maturation of S3-directed mAbs ( Figure S16 ). Similarly, insufficient space for nAbs binding to conserved site S4 was found on the closed S protein, and only the RBD in the standing-up state could improve the accessibility of site S4.

Thus, our findings indicate that conserved antigenic sites display less accessibility than SARS-CoV-2 unique epitopes and that poor accessibility hinders the affinity maturation of site S3-directed mAbs in natural infection and even might decrease the cross-neutralizing antibody response after SARS-CoV-2 vaccination.

The majority of mAbs recognizing antigenic site S1 were VH3-53/66 mAbs with a short (mostly 11 residues) CDRH3 sequence. Moreover, many studies have also reported the same enrichment of VH3-53/66 mAbs targeting antigenic sites similar to site S1 and found that they share structural similarities with each other (8, 29, 46–48). These VH3-53/66 mAbs showed native binding to SARS-CoV-2 RBD residues using CDRH1 and CDRH2 by forming many hydrogen bonds (L455, Y473, A475 and N487 bound by CDRH1, and D420, Y421 and R457 bound by CDRH2) ( Figure 5A and Figure S17 ). Therefore, the native binding advantage between antigenic site S1 and VH3-53/66 was the cardinal cause of site S1 immunodominance, which could result in massive amplification of antigenic site S1-specific B cells and competition to inhibit the proliferation of B cells directed against conserved antigenic sites. To indirectly enhance the competitiveness of conserved antigenic sites in the immune response, it is essential to decrease the immunodominance of site S1. To this end, we designed a variety of SARS-CoV-2 RBD variants with antigenic site S1 silencing using either protein truncation or glycan modification ( Figures 5B–E ). Glycan modification at positions K458 and A475 within site S1, termed RBD_{Glycan458,475}, was successful in destroying the binding of S1-directed mAbs, including P02-3C11 derived from VH3-66 and P05-5C4 derived from VH3-53, and could maintain conserved antigenic sites S2, S3 and S4 ( Figure 5F ). To preliminarily investigate whether RBD_{Glycan458,475} could improve the cross-binding antibody response, BALB/c mice were immunized with 20 μg/dose RBD_{Glycan458,475}. Two weeks after immunization, the mice receiving RBD_{Glycan458,475} or the reference RBD all presented detectable serum anti-SARS-CoV S IgG and anti-SARS-CoV-2 S IgG; however, RBD_{Glycan458,475} induced a significantly higher cross-binding IgG titer than the reference RBD ( Figure 5G ). Additionally, the results also revealed that RBD_{Glycan458,475} could induce higher cross-binding IgG titer against SARS-CoV-2 and the Omicron variant (BA.1) than the reference RBD ( Figure 5G ). Hence, universal vaccines based on such glycan modification of the SARS-CoV-2 RBD have the potential to induce a stronger cross-binding antibody response that could efficiently protect against infection by SARS-like coronaviruses and emerging SARS-CoV-2 variants.