We analyzed the antibody responses in serum samples of 12 individuals after receipt of three doses of ZF2001 vaccine from a cohort recruited previously (Supplementary Fig. S1)24. Six individuals had received three doses of vaccine at month 0, 1, and 2 (short-interval group). The other six individuals had received three doses of vaccine at month 0, 1, and 4–6, with an extended time interval between the second and third dose (long-interval group). Serologic binding and neutralizing antibodies were measured at the following time points: 0.5–2 months, 4–5 months, and 7–8 months after the third dose. Consistent with our previous reports4,24, binding antibodies to RBD from prototype SARS-CoV-2 (ancestral virus identified in Wuhan) and its VOCs (Beta, Delta, and Omicron BA.1) were higher in the long-interval group in comparison to the short-interval group at all time points (Fig. 1a). Consistently, the titers of neutralizing antibodies against VSV-pseudotyped viruses which bear S proteins of prototype SARS-CoV-2 or its VOCs on the surface of virions were also higher in the long-interval group at all time points (Fig. 1b). Notably, the titer gap between these two groups widened when serum samples were tested for the binding and neutralization of pseudovirus from the cognate prototype SARS-CoV-2 to VOCs at all time points (Fig. 1a, b).
To profile the BCR repertoires elicited by the RBD-based protein subunit vaccine ZF2001 with either long- or short-dosing interval, antigen-binding CD19+IgG+ B cells from both groups were sorted from PBMCs (Fig. 2a; Supplementary Fig. S2). The mRNA from each single B cell was extracted and reversed to cDNA. Variable regions of heavy (VH) and light (VL) chains were individually amplified and cloned into IgG1 backbone for antibody production. The supernatant of the cells that were transfected with DNA fragments of each mAb was screened for binding to prototype SARS-CoV-2 RBD. The positive supernatants were further tested for their neutralization breadth against pseudotyped HIV bearing S protein of prototype, Beta, Delta, and Omicron (BA.1), respectively.
As a result, 1604 mAbs obtained from B cell clones bound prototype RBD, with 1000 and 604 mAbs derived from the long- and short-interval individuals, respectively (Fig. 2b). Among them, 533 (long-interval group) and 262 (short-interval group) mAbs showed detectable neutralizing activity against prototype SARS-CoV-2 pseudovirus or its variants (Fig. 2b, c). Therefore, the long-dosing interval developed an increased number of B cells encoding antibodies with binding and neutralizing activities to SARS-CoV-2. In the long-interval group, 453 mAbs neutralized prototype pseudovirus, 272 neutralized Beta variant, 283 neutralized Delta variant, and 128 neutralized Omicron variant (BA.1). In contrast, in the short-interval group, 219 mAbs neutralized prototype pseudovirus, 152 neutralized Beta variant, 104 neutralized Delta variant, and 95 neutralized Omicron variant (BA.1). Substantial proportion (13.9%) of ZF2001-elicited RBD-binding antibodies neutralized Omicron variant. Compared with the short-interval individuals, the long-interval individuals were observed with higher geometric mean numbers of mAb in the neutralization of SARS-CoV-2 pseudoviruses, including its variants (Beta, Delta, and Omicron BA.1) (Fig. 2d, e). These results suggested the advantages of the extended dosing interval of ZF2001 vaccination in B cell quantity and quality.
Next, all antigen-binding mAbs obtained from B cell clones were Sanger-sequenced. As a result, we obtained 1500 paired antibody sequences from these 12 individuals (Fig. 2b). Expanded clonotypes of antigen-binding B cells were detected in all individuals (Fig. 3a). Notably, the long-interval individuals showed a higher percentage of expanded clonotypes compared to the short-interval individuals (32.6% vs 25.6%) (Fig. 3b). This result suggested that the prolonged dosing interval of vaccination was prone to increase the proportion of antigen-binding B cells in clonality from single clonotypes to expanded clonotypes, indicating a more advanced B-cell evolution.
The gene usage of the immunoglobulin variable region (V region) in these individuals vaccinated with RBD-based vaccine was similar to those vaccinated with full-length S-based vaccine (Fig. 3c; Supplementary Fig. S3a)18. IGHV1-69, IGHV3-9, and IGHV4-39 were the most frequent VH genes used in the long-interval group, and IGHV1-69, IGHV3-9, and IGHV3-23 were the most frequent VH used in the short-interval group (Fig. 3c). IGKV1-39, IGLV3-21, and IGKV3-20 were the most frequent VL detected in both groups (Supplementary Fig. S3a).
The analysis of BCR sequences showed pronounced differences in somatic hypermutations (SHMs) between the long- and short-interval groups. The B cell-derived binding antibodies from the long-interval individuals had significantly (Wilcoxon-rank sum test P < 0.0001) elevated nucleotide mutations of VH and VL compared with those from the short-interval individuals, with a shift of the SHM number (Fig. 3d, e; Supplementary Fig. S3b). This result indicated that the prolonged dosing interval of vaccination promoted the maturation of antigen-binding B cells.
We next analyzed the SHM of BCR sequences of B cell-derived binding antibodies with different neutralization breadths to SARS-CoV-2. The nucleotide mutations in heavy chain of antibodies elevated sequentially from those with non-neutralizing activity (average 13.0) to those with monovalent (average 14.2), bivalent (average 14.8), trivalent (average 15.2), and tetravalent (average 17.0) activities (Fig. 3e). Significantly (P < 0.0001) higher number of SHMs in heavy chain was found in neutralizing antibodies vs non-neutralizing antibodies, in multivalent antibodies vs monovalent antibodies, and in tetravalent antibodies vs mono-to-tri-valent antibodies (Fig. 3e, f). In contrast, the nucleotide mutations in light chain were less pronounced between the B-cell-derived antibodies with different neutralization breadths, except the tetravalent antibodies (Fig. 3e, f). We also analyzed the complementary determining region 3 (CDR3) length of these antibodies and found a negative correlation between antibody breadth and CDR3 length of heavy chain (Supplementary Fig. S3e, f). Accordingly, the mean CDR3 length of heavy chain was lower (P = 0.04) in the long-interval group than that in the short-interval group (Supplementary Fig. S3c, d). Therefore, our data suggested the increased B-cell maturation, indicated by the higher number of SHMs and lower CDR3 length of heavy chain, was associated with the extended neutralization breadth to SARS-CoV-2. Accordingly, we found a significantly higher number of SHMs of both heavy and light chains for neutralizing antibodies in the long-interval group than that in short interval group against all four SARS-CoV-2 strains (Fig. 3g; Supplementary Fig. S3g). Also, a shorter CDR3 of heavy chain was found in the long-interval group (Supplementary Fig. S3h). These results explained the greater B cell breadth after prolonged dosing-interval of vaccination.
The analysis of VH gene usage of the antigen-specific B cells with different neutralization breadths to SARS-CoV-2 revealed that IGHV3-33, IGHV4-4, IGHV1-24, IGHV1-58, IGHV3-53, and IGHV3-66 were more frequently used in the multivalent clones (Fig. 3h; Supplementary Fig. S3i).
We next cloned and expressed 65 tetravalent bnAbs as human IgG1, and successfully obtained 57 bnAbs. Among them, 53 bnAbs with fifty percent pseudovirus neutralization titer (pVNT50) value less than 50 μg/mL against prototype pseudovirus were further analyzed. The neutralizing activity for this panel of bnAbs was measured against pseudotyped VSV-expressing S protein from SARS-CoV-2 or its variants. Since a succession of Omicron sub-variants surged one another after the emerging of the BA.1, we tested the antibody neutralization to pseudotyped SARS-CoV-2 representing ancestral SARS-CoV-2 (prototype, Beta, and Delta) and Omicron sub-variants (BA.1, BA.2, BA.2.12.1, BA.3, BA.4/5, BF.7, XBB, and BQ.1). As a result, most of the bnAbs maintained the neutralizing potency to BA.2, and a few showed substantial reduction in neutralization of BA.2.12.1 and BA.3 (Fig. 4a). In contrast, most of the bnAbs largely reduced the neutralizing activity to BA.4/5, and only several of them still well neutralized (IC50 < 2000 ng/mL) the currently circulating BF.7, XBB, and BQ.1, suggesting the significant immune evasion. Notably, 10 of 34 (29.4%) bnAbs in the long-interval group neutralized XBB with IC50 < 2000 ng/mL, however, this ratio was decreased to 3/19 (15.8%) in the short-interval group. 6 of 34 (17.6%) bnAbs in the long-interval group neutralized BQ.1 with IC50 < 2000 ng/mL, with 3/19 (15.8%) in the short-interval group (Fig. 4a).
We next sought to determine the antigenic landscape of these bnAbs elicited by the RBD-based vaccine ZF2001. These bnAbs were first sorted by the competition profiles to nine benchmark mAbs that target seven major antigenic sites of SARS-CoV-2 RBD as defined previously (Fig. 4b; Supplementary Fig. S4)26,27. As a result, the epitopes of 19 bnAbs were mapped to the RBD-2 site, a place overlayed with receptor-binding motif (RBM) from the valley to the peak (similar to mAb REGN-10933); 21 were mapped to the RBD-5 site at the outer face beneath the mesa (similar to mAb C110 and S309); 6 were binned to the RBD-7 site at the inner face beneath the mesa (similar to mAb CR3022). In contrast, no epitope of bnAbs was mapped to RBD-1, -3, -4, and -6 sites. It is notable that four bnAbs (L3.87, L6.92, S3.19, and S3.34) were discovered not to compete with any of these nine benchmark mAbs, suggesting their different antigenic distributions. Nevertheless, they were all observed to compete with a pan-sarbecovirus mAb S2H9728,29, suggesting a new antigenic site (namely RBD-8 site) at the lateral ridge beneath the peak (Fig. 4b). Of note, the epitope binning only elucidates the bnAbs against four strains mentioned in our study, while the whole antibody repertoires may show different epitope maps.
Two bnAbs, L4.65 and L5.34, showed ultrapotent neutralizing activities (IC50 ≤ 10 ng/mL) against almost all the pseudoviruses tested (Fig. 4a). L4.65 and L5.34 were mapped to RBD-5 and RBD-2, respectively, the two major antigenic sites of bnAbs elicited by ZF2001 (Fig. 4b). To explore the molecular basis of their epitopes, we next determined the cryo-EM structures of these two bnAbs in complexed with S-trimer of prototype (resolution 2.65 Å), Omicron BA.2 (resolution 2.75 Å), and BA.4/5 (resolution 2.85 Å), respectively (Supplementary Table S1 and Figs. S5–S7).
Notably, an S-trimer bound three pairs of fragments antibody-binding (Fabs) of L4.65 and L5.34 in its open state without steric clash (Fig. 5a). Three Fab pairs engaged three up RBDs with the same binding mode in different S-timers. L5.34 and L4.65 interact with the inner and outer face of the RBD, respectively (Fig. 5b). The footprint of L5.34 locates at the typical RBD-2 site towards RBD peak, largely overlapping with the RBM, while the footprint of L4.65 is at the typical RBD-5 site beneath the mesa in the outer face of RBD (Fig. 5c–e). The epitopes of L5.34 harbor more mutations in Omicron BA.2 (D405N, K417N, S477N, T478K, E484A, Q493R, N501Y, and Y505H) and BA.4/5 (D405N, K417N, S477N, T478K, E484A, F486V, N501Y, and Y505H), compared with the epitopes of L4.65 (N440K and 498R) in both Omicron sub-variants (Fig. 5c–f). This explained the largely decreased binding affinity of L5.34, but not L4.65, to RBD of Omicron BA.2 or BA.4/5, when compared with its binding to prototype RBD (Supplementary Fig. S8). Additionally, we analyzed SARS-CoV-2 sequences randomly selected from the Global Initiative on Sharing Avian Influenza Data (GISAID) database representing the virus strains up to Jan 2023, and found that the epitopes of these two bnAbs in prototype, Omicron BA.2, and BA.4/5 had covered most of amino acid variations occurred before (Fig. 5f), suggesting two vulnerable sites for broad neutralization of SARS-CoV-2.
