Aromaticity at spike position 456 affects functional immunogenicity of SARS-CoV-2 Omicron variant receptor-binding domain

Abstract

The recurrent emergence of the F456L substitution in SARS-CoV-2 Omicron sublineages has been associated with immune evasion, yet its impact on vaccine-induced immune responses remains poorly understood. Using a trimeric RBD-HR subunit vaccine platform, we investigated how residue 456 influences the functional immunogenicity of the receptor-binding domain (RBD). We found that the F456L substitution markedly reduced neutralizing antibody responses against multiple SARS-CoV-2 variants, despite maintaining comparable levels of total antigen-specific IgG, indicating a selective impairment of functional immunogenicity rather than overall antigenicity. Single-cell RNA sequencing and B-cell receptor (BCR) repertoire analyses further revealed that the F456L mutation weakened germinal center responses, reduced clonal expansion, and impaired repertoire remodeling of antigen-specific B cells. Structural modeling, biolayer interferometry, and calcium flux assays demonstrated that aromatic residues at position 456 promote effective germline BCR recognition and activation. Importantly, replacing the non-aromatic leucine residue with aromatic amino acids, including tyrosine or tryptophan, restored the neutralizing antibody-eliciting capacity of the antigen. Together, these findings identify aromaticity at residue 456 as a critical determinant of RBD functional immunogenicity and demonstrate how a single amino acid substitution can reshape B-cell responses and antibody quality. This work provides important insights for rational vaccine antigen design and advances our understanding of how site-specific mutations regulate antiviral immunity.

Introduction

The ongoing emergence of immune-evasive SARS-CoV-2 variants continues to challenge global vaccination strategies, even following the acute phase of the COVID-19 pandemic.¹ Since the initial outbreak, the virus has undergone extensive antigenic evolution, leading to successive waves of variants that exhibit both enhanced transmissibility and the ability to evade pre-existing immunity. While neutralizing antibodies induced by infection or vaccination have provided substantial protection against severe disease, the emergence of variants with multiple spike protein mutations has increasingly undermined the breadth and durability of humoral responses. Among the selective pressures driving SARS-CoV-2 evolution, enhanced binding affinity to human angiotensin-converting enzyme 2 (hACE2) and immune escape are widely recognized as major determinants of viral fitness.^2–4 However, despite extensive characterization of these factors, the contribution of individual amino acid changes to spike immunogenicity, particularly in the receptor-binding domain (RBD), and their influence on the quality and potency of vaccine-elicited antibody responses remain incompletely understood.

Since late 2022, Omicron XBB sublineages such as XBB.1.5 and EG.5.1 have successively dominated the global epidemic landscape, demonstrating the continued ability of SARS-CoV-2 to circumvent population-level immunity.^5–7 These variants differ by a single amino acid substitution at position 456 in the receptor-binding domain (RBD) of the spike protein: phenylalanine (F) in XBB.1.5 and leucine (L) in EG.5.1 (Fig. 1). Interestingly, a similar F456L substitution has independently re-emerged in the JN.1 (F456) and JN.1.16 (L456) variants, suggesting a potential role in antigenic evolution of the virus (Fig. 1; detailed sequences are provided in Supplementary Table 1).⁸ The repeated independent emergence and persistence of the F456L substitution across multiple Omicron sublineages strongly implies that this residue may represent an adaptive hotspot, enabling the virus to modulate its antigenic profile under strong humoral immune pressure (Fig. 1a–c). This observation aligns with previous reports highlighting that single-site substitutions within the RBD can disproportionately influence neutralizing antibody recognition, potentially altering both the breadth and potency of immune responses.

Fig. 1. Recurrent emergence and circulation of SARS-CoV-2 Omicron variants carrying F456L.

a Proportions of XBB.1.5, EG.5.1, JN.1, and JN.1.16 variants from December 2022 to February 2025 among globally reported sequences, including data from the World, United States, Europe, and Oceania (https://cov-spectrum.org). b, c Circulation dynamics of variants carrying the F456L mutation (https://cov-spectrum.org/explore). d Upper panel, schematic representation of the RBD-HR construct. Lower panel, emergence of XBB and JN.1 subvariants carrying distinct RBD mutations, particularly F456L, across multiple independent lineages (outbreak.info). SP stands for signal peptide; NTD represents the N-terminal region; HR1/HR2 denote heptad repeat regions 1 and 2; TM indicates the transmembrane segment; and CP is the cytoplasmic tail. e Structural representation of the residue 456 region on the RBDs of XBB.1.5, EG.5.1, JN.1, and JN.1.16. PDB Codes: XBB.1.5 (8WRL), EG.5.1 (8XN2), JN.1 (8Y18). The JN.1.16 RBD structure was modeled based on the JN.1 structure using PyMOL

The concept of residue 456 acting as an “aromatic switch” arises from the observation that the presence of an aromatic phenylalanine at this position appears to preserve immunogenicity, whereas substitution with a non-aromatic leucine is associated with reduced antibody recognition and functional escape. Given the critical position of residue 456 within the receptor-binding motif, even minimal changes at this site have the potential to perturb the presentation of key epitopes to B cell receptors (BCRs), thereby influencing germinal center selection, clonal expansion, and the maturation of neutralizing antibodies. Notably, while F456L has been associated with enhanced immune evasion in several neutralization studies,^9,10 the mechanistic basis by which this single substitution impacts functional immunogenicity—defined here as the ability to induce robust, high-quality B cell responses—remains poorly defined.

To systematically interrogate this question, we leveraged a previously established RBD-HR vaccine platform, in which the SARS-CoV-2 RBD is fused to the heptad repeat domains HR1 and HR2 to form self-assembling trimers. These recombinant proteins not only recapitulate the trimeric structure of the native spike but are also readily adapted to match the evolving sequences of emerging variants.¹¹ Using this platform, we observed that a recombinant RBD-HR vaccine based on EG.5 (RBD_EG.5-HR) induced significantly weaker neutralizing antibody responses compared to its XBB.1.5-based counterpart (RBD_XBB.1.5-HR), despite their minimal overall sequence divergence. The primary determinant of this discrepancy was identified as the F456L substitution, highlighting the profound influence that a single amino acid can exert on antigenicity and the elicited humoral immune response.

Building on these observations, we conducted an integrated analysis combining protein engineering, antigen-specific B cell isolation, and single-cell immune profiling to dissect the immunological consequences of the F456L mutation. Single-cell RNA sequencing (scRNA-seq) and paired BCR sequencing of RBD-specific B cells revealed that the F456L mutation not only reduces germinal center activity but also constrains the diversity of the B cell repertoire, limiting the range of potential neutralizing clones. Complementary structural modeling using Rosetta Flex ddG suggested that the aromatic character of residue 456 facilitates optimal BCR engagement, whereas non-aromatic substitutions diminish binding affinity. Functional validation through biolayer interferometry (BLI) and calcium flux assays further corroborated these findings, demonstrating impaired BCR signaling and reduced activation of antigen-specific B cells in response to L456-bearing RBDs.

Collectively, these results support a role for residue 456 as a critical determinant of RBD functional immunogenicity. By linking a single amino acid change to measurable reductions in B cell clonal expansion, germinal center responses, and neutralizing antibody production, this study provides mechanistic insight into how SARS-CoV-2 evolves under immune pressure while simultaneously informing rational vaccine design. These findings emphasize that, beyond merely tracking antigenic escape in neutralization assays, careful evaluation of functional immunogenicity is essential for updating vaccine antigens to maintain protective efficacy against rapidly evolving variants. Ultimately, understanding the nuanced interplay between spike sequence variation and host immune recognition will be pivotal for anticipating viral evolution and optimizing next-generation COVID-19 vaccines.

Results

The F456L mutation impairs neutralizing immunogenicity elicited by RBD-HR vaccines

The mutation dynamics of emerging SARS-CoV-2 variants have been closely monitored by our group, with particular attention to the development of a trimeric recombinant protein vaccine targeting these epidemic strains.^12,13 By July 2023, the EG.5 lineage, characterized by the F456L mutation within the RBD of the spike protein (Fig. 1d, e), was classified as a variant under monitoring by the World Health Organization (WHO).¹⁴ Residue 456 is located within the hACE2-binding interface.¹⁵ Despite the presence of the F456L mutation, the cell entry efficiency of EG.5.1 remained unchanged.⁹ However, the efficiency of evading neutralizing antibodies increased,^9,16 while the impact of its mutation on immunogenicity during infection has not been thoroughly investigated.

To investigate the impact of the F456L substitution on immunogenicity, recombinant trimeric RBD-HR proteins derived from the XBB.1.5(F456) strain (RBD_XBB.1.5-HR) and the EG.5(L456) lineage (RBD_EG.5-HR) were produced (Fig. 2a, b). The RBD sequences of these two vaccine antigens are identical except for the F-to-L substitution at position 456 (F456L). The vaccines were formulated by mixing the RBD-HR protein with an MF59-like adjuvant. Female NIH mice received two intramuscular doses of RBD_XBB.1.5-HR(F456) or RBD_EG.5-HR(L456) vaccine on days 0 and 21. Neutralizing antibodies in sera were measured on day 35. The RBD_EG.5-HR vaccine carrying the F456L substitution induced markedly lower neutralizing antibody titers. The geometric mean titers (GMTs) for 50% neutralization against authentic XBB.1.5 (F456) and JN.1 (F456) viruses were 2438 and 768, respectively, in the RBD_XBB.1.5-HR(F456) group, but both titers fell to the assay detection limit in the RBD_EG.5-HR(L456) group (Fig. 2c). Compared to the RBD_XBB.1.5-HR(F456) group, neutralization of pseudoviruses representing XBB.1.5 (F456), XBB.1.9.1 (F456), XBB.1.16 (F456), XBB.1.16.6 (L456), EG.5.1 (L456), FL.1.5.1 (L456), HV.1 (L456), HK.3 (L456), BA.2.86 (F456), and JN.1 (F456) was reduced by 184-, 423-, 205-, 13-, 46-, 49-, 51-, 85-, 46-, and 93-fold, respectively, in the RBD_EG.5-HR(L456) group (Fig. 2d). Cellular analyses further revealed that the EG.5-based(L456) vaccine induced significantly fewer T follicular helper (Tfh) cells, germinal center (GC) B cells, memory B cells, antigen-specific antibody-secreting cells, and IFN-γ⁺ CD8⁺ T cells, compared to the XBB.1.5-based(F456) vaccine (supplementary Fig. 1). Collectively, these findings confirm that the F456L mutation substantially impairs the adaptive immune responses induced by RBD-HR vaccines.

Fig. 2. F456L-containing RBD-HR vaccine elicits reduced cross-neutralizing antibody responses.

Biochemical characterization of the recombinant RBD_XBB.1.5-HR (a) and RBD_EG.5-HR (b) trimers. Representative size-exclusion chromatography (SEC) profile depicting the purified RBD-HR trimer resolved on a calibrated Superdex 200 Increase analytical column. The corresponding SEC elution fractions were subsequently evaluated via SDS-PAGE. Samples 1-3 correspond to the ascending shoulder, peak apex, and descending shoulder of the main elution peak, respectively. M marker. The left Y-axis represents UV absorbance at 280 nm (blue line), and the right Y-axis represents UV absorbance at 260 nm (red line). c Neutralizing antibody titers against authentic XBB.1.5 and JN.1 viruses following vaccination with the RBD_XBB.1.5-HR and RBD_EG.5-HR vaccines. d Neutralizing antibody titers against XBB.1.5, XBB.1.9.1, XBB.1.16, XBB.1.16.6, EG.5.1, FL.1.5.1, HV.1, HK.3, BA.2.86, and JN.1 pseudoviruses following vaccination with the RBD_XBB.1.5-HR and RBD_EG.5-HR. Biochemical characterization of the recombinant RBD_JN.1-HR (e) and RBD_JN.1.16-HR (f) trimers, as described in panels a and b. g Neutralizing antibody responses induced by full-length spike protein vaccines. h Evaluation of neutralizing antibodies against BA.2.86, JN.1, JN.1.1, JN.1.13, JN.2, JN.3, KP.2, and KP.3 pseudoviruses following vaccination with the RBD_JN.1-HR and RBD_JN.1.16-HR vaccines. The letters in parentheses following the variant names indicate the residue present at position 456 within the RBD (F phenylalanine, L leucine). Data are presented as GMTs with 95% CIs. Statistical significance was determined using two-tailed unpaired Student’s t-tests. A P-value below 0.05 was defined as indicating statistical significance (ns denotes non-significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001)

To assess whether the F456L mutation altered protein stability or structural integrity, such as misfolding, aggregation, or poor solubility, the structural and functional integrity of the RBD-HR variants was rigorously characterized (Supplementary Fig. 2). HPSEC and DSF analyses confirmed high homogeneity and comparable thermal stability between RBD_XBB.1.5-HR (F456) and RBD_EG.5-HR (L456). Additionally, both proteins showed high and comparable binding affinities to hACE2 and various conformation-sensitive monoclonal antibodies, suggesting that the F456L mutation does not substantially disrupt overall folding or the presentation of key functional epitopes.

The recently emerged JN.1.16(L456) variant also carries the F456L mutation in the RBD compared to its parental JN.1(F456) strain (Fig. 1d). To further evaluate the impact of this mutation, recombinant RBD_JN.1-HR and RBD_JN.1.16-HR proteins were expressed. Consistent with previous observations, the F456L-containing RBD_JN.1.16-HR elicited significantly lower neutralizing antibody titers (Fig. 2e, f, h) and weaker cellular immune responses (Supplementary Fig. 3).

To determine whether these effects were restricted to the RBD-HR system, neutralizing antibody responses induced by full-length spike-XBB.1.5 (F456) and spike-EG.5 (L456) recombinant protein vaccines were also examined. spike-EG.5 (L456) exhibited a downward trend in eliciting neutralizing antibodies compared to spike-XBB.1.5 (F456) (Fig. 2g), although this difference did not reach statistical significance (P > 0.05). The attenuated impact of the F456L mutation in the context of the full-length spike protein likely results from a broader and more heterogeneous antibody repertoire elicited by the full-length spike protein.

The F456L substitution impairs neutralizing antibody responses induced by prototype RBD-HR vaccination

The Omicron XBB.1.5 variant contains over 30 amino acid substitutions in the RBD compared to the ancestral SARS-CoV-2 strain (referred to here as WT/prototype).¹⁷ To determine the specific effect of the F456L substitution independent of other co-mutations, this single mutation was introduced into the WT-RBD. Recombinant RBD-HR proteins were expressed, containing either the isolated F456L substitution (RBD_F456L-HR) or no mutation (RBD_WT-HR) (Supplementary Fig. 4a). RBD_WT-HR with the L452R mutation (RBD_L452R-HR) was also included as a control, given its established role in maintaining or enhancing RBD-induced neutralizing antibodies.¹⁸

Neutralization assays showed that the RBD_L452R-HR vaccine elicited a strong response against pseudoviruses carrying the L452R substitution (Fig. 3b, d). In contrast, the F456L mutation significantly impaired the neutralizing response. The RBD_F456L-HR vaccine consistently induced lower neutralizing antibody titers across all tested prototype-based pseudoviruses (Fig. 3a–d), including the prototype, prototype (L452R), prototype (F456L), and prototype (L452R + F456L). These findings indicate that the F456L mutation markedly reduces neutralizing antibody titers even within an otherwise WT antigenic background.

Fig. 3. The single F456L mutation reduces neutralizing antibody responses following prototype RBD vaccination.

Neutralizing antibody titers were evaluated against Prototype (a), Prototype (L452R) (b), Prototype (F456L) (c), and Prototype (L452R + F456L) (d) pseudoviruses. e A radar chart summarizes the neutralizing antibody response profiles elicited by each vaccine, based on Log10 GMTs from panels a–d. The distinct chart shapes illustrate the antigen-specific breadth and potency of neutralization across prototype-based pseudoviruses. Data are presented as GMTs with 95% CIs. Comparison was performed by one-way analysis of variance (ANOVA). A P-value below 0.05 was defined as indicating statistical significance (ns denotes non-significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001)

This pattern is further illustrated by the radar chart (Fig. 3e). Compared to the reduced neutralization profile of RBD_F456L-HR (red line), both RBD_WT-HR (blue line) and RBD_L452R-HR (green line) elicited substantially higher antibody titers. These results highlight the antigen-specific nature of vaccine-induced responses and emphasize the markedly reduced immunogenicity of the F456L-modified antigen. Collectively, these results indicated that the F456L substitution alone is sufficient to significantly impair neutralizing antibody generation, indicating its potential contribution to diminished vaccine-induced neutralizing responses.

Although the single F456L substitution substantially reduces neutralizing titers, the magnitude of this effect is smaller than that observed for the Omicron variants (Fig. 2 vs. Fig. 3). This difference likely reflects the cumulative impact of additional co-occurring RBD mutations, such as N450D, N460K, or R346T, present in the natural variant, which may collectively alter the antigenic surface and contribute to immune evasion. Therefore, F456L is a key, but not the sole, determinant of the reduced neutralizing titers observed in circulating Omicron subvariants.

Aromaticity at position 456 enhances neutralizing antibody responses

The data indicate that loss of aromaticity at position 456, specifically the substitution of F456 with L456, results in markedly reduced neutralizing antibody titers. These findings prompted us to investigate whether replacing the non-aromatic L456 with other aromatic amino acids, such as tyrosine (Y) or tryptophan (W), could restore neutralizing immunogenicity. The chemical structures of Y and W differ in their side chains: Y contains a phenolic hydroxyl group attached to a benzene ring, while W possesses a larger and more complex indole ring.¹⁹

To test this hypothesis, RBD-HR proteins carrying single amino acid substitutions at position 456 in the EG.5 background were generated: RBD_EG.5-HR(L456Y) and RBD_EG.5-HR(L456W) (Fig. 4a, Supplementary Fig. 4b, c and Supplementary Table 1). Vaccine formulation, immunization, and serum collection were performed according to established protocols. Substitution with either aromatic residue significantly enhanced neutralizing antibody responses compared to the non-aromatic L456 group. The 50% GMTs against a panel of pseudoviruses, including XBB.1.5 (F456), XBB.1.9.1 (F456), XBB.1.16 (F456), XBB.1.16.6 (L456), XBB.2.3 (F456), EG.5.1 (L456), HV.1 (L456), and FL.1.5.1 (L456), were all significantly elevated in mice vaccinated with RBD_EG.5-HR(L456Y) and RBD_EG.5-HR(L456W) (Fig. 4b–i).

Fig. 4. Aromaticity at residue 456 is a critical determinant for eliciting potent neutralizing antibody responses.

a The chemical structures of amino acids F, L, Y, and W at position 456 are shown in the context of the RBD (PDB ID: 8WRL). b–i Evaluation of neutralizing antibody titers against XBB.1.5 (b), XBB.1.9.1 (c), XBB.1.16 (d), XBB.1.16.6 (e), XBB.2.3 (f), EG.5.1 (g), HV.1 (h), and FL.1.5.1 (i) pseudoviruses. Panel a was generated using BioRender and PyMOL. The letters in parentheses following the variant names indicate the amino acid residue at position 456 of the RBD (F phenylalanine, L leucine). Data are presented as GMTs with 95% CIs. Comparison was performed by one-way analysis of variance (ANOVA). A P-value below 0.05 was defined as indicating statistical significance (ns denotes non-significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001)

Restoration of neutralizing antibody-eliciting capacity through aromatic substitutions (Y or W) at position 456 was not restricted to variants carrying F456. Instead, enhanced neutralizing activity was observed broadly, including those with the L456 mutation. These results indicate that the immunogenic enhancement conferred by aromatic residues at position 456 is broadly applicable, independent of the original 456 residue.

Single-cell BCR repertoire analysis reveals impaired clonal selection associated with the F456L mutation

To comprehensively assess the influence of distinct RBD-based vaccines on antigen-specific B cell responses, single-cell RNA sequencing and BCR profiling were performed on CD3⁻CD19⁺RBD⁺ B cells isolated by flow cytometry from spleens pooled from six mice per group (six in one) (Fig. 5a). This approach enabled integrated analysis of transcriptional states and clonal architecture at single-cell resolution. A total of 38369 single cells were clustered into 11 transcriptionally distinct populations (Supplementary Fig. 5). Eleven distinct B cell clusters were identified in the scRNA-seq analysis. As shown in the lower panel of supplementary Fig. 5a, clusters were annotated using established markers (such as Apoe and Aicda for GC B_LZ cells, and Ighm for B_mem cells) and nomenclature from previous studies.²⁰ These annotations enabled classification of the 11 clusters into major functional subsets, including germinal center (GC) B cells, memory B cells, and plasma cells.

Fig. 5. The F456L mutation impairs B cell repertoire remodeling and clonal expansion in germinal centers.

a Workflow of antigen-specific B cell isolation. b CDR3 amino acid length distribution in Immunized-F456 and Immunized-L456 groups. c IGHV gene usage frequencies across the four groups. d Morisita overlap indices based on CDR3 amino acid sequences (bottom right) and V(D)J gene usage (top left) between immunized and naive repertoires. The heatmap displays the Morisita overlap index, ranging from 0 (no overlap) to 1 (complete overlap). The top-left triangles represent similarity at the V(D)J gene level, while the bottom-right triangles represent similarity at the CDR3 amino acid level. The color scale ranges from purple/blue (lower similarity) to yellow (higher similarity). The heatmaps indicate that F456 immunization (left) leads to greater repertoire divergence (predominantly purple), whereas L456 immunization (right) retains greater similarity to the naive baseline (yellow)

Single-cell transcriptomic analysis identified substantial differences in B cell subset composition across experimental groups (Supplementary Fig. 5b). In unimmunized mice (Naive), the RBD-specific B cell populations were predominantly composed of naïve B cells, with minimal representation of activated or differentiated subsets. In contrast, both immunized groups exhibited robust expansion of plasmablast (PB), germinal center B cells (dark zone, GCB_DZ; light zone, GCB_LZ), and memory B cells (B_mem). Notably, mice immunized with the RBD_XBB.1.5-HR vaccine (Immunized-F456) exhibited higher proportions of these differentiated subsets than those immunized with the RBD_EG.5-HR vaccine, indicating a more vigorous response elicited by the F456-containing immunogen.

Despite these compositional differences, both immunized groups exhibited comparable distributions of CDR3 lengths and broadly similar overall IGHV gene usage profiles (Fig. 5b, c), indicating that the two immunogens recruited B cell precursors from largely overlapping naive B cell pools. However, individual gene usage demonstrated functionally relevant changes. Notably, IGHV1-81, the most frequently used V gene in naive mice, decreased to 12th place following F456 immunization, underscoring the impact of antigen-driven clonal selection in reflecting antigen-driven reshaping of the naive repertoire. Additionally, although overall IGHV usage rankings were broadly conserved between the Immunized-F456 and Immunized-L456 groups, the Immunized-F456 repertoire exhibited a more clonally focused pattern, characterized by greater expansion of dominant clones. In contrast, the Immunized-L456 group displayed a more diffuse polyclonal distribution, consistent with weaker antigen-driven clonal selection (Fig. 5c).

Morisita overlap analysis showed that immunization with the F456-containing immunogen, but not the L456-containing immunogen, induced substantial repertoire remodeling and divergence from the naive repertoire baseline (Fig. 5d). Clone tracking further supported this observation, revealing that F456 immunization resulted in the highest number of expanded clones, particularly within GCs (supplementary Fig. 5c). Enhanced recognition and uptake of the F456-containing antigen were also observed in naïve B cells, as well as in professional antigen-presenting cells, including bone marrow-derived dendritic cells and macrophages (supplementary Fig. 6). However, this effect may be largely diminished given that a sufficient amount of protein (10 μg) was used for immunization. Collectively, these findings support a role for F456 in promoting clonal selection and affinity maturation in the GC microenvironment.

These findings indicate that the aromatic residue at position 456 enhances neutralizing immunogenicity and contributes to the formation of a major neutralizing epitope that promotes focused B cell clonal selection. To directly test this hypothesis, we next examined how the F456L mutation affects antibody binding to this epitope using monoclonal antibodies with defined footprints.

F456 shapes the neutralizing antibody repertoire

Although the F456L substitution significantly reduced neutralizing antibody titers, it did not alter total binding antibody levels (Supplementary Fig. 7). These findings indicate that F456L does not significantly affect the overall immunogenicity of the RBD backbone but specifically impairs the induction of neutralizing antibody populations.

To determine whether this site contributes to a dominant neutralizing epitope, monoclonal antibody (mAb) binding footprint analysis (data from PDB) and serum competition assays were conducted (Fig. 6a). A panel of 20 mAbs known to bind the F456 site was selected from public databases (e.g., PDB and CoV-AbDab). The relative energetic contribution of F456 to antibody binding was calculated using CCP4 software and compared to the mean contribution of all epitope residues (dashed line in Fig. 6a). In most cases, the contribution of F456 was similar to the average. Several antibodies exhibited below-average dependence on F456, with only one displaying markedly low reliance. Conversely, a substantial subset demonstrated significantly higher-than-average dependence on this residue. These findings indicate that F456 is a critical hotspot in antigen–antibody interactions and a major determinant of neutralizing epitope recognition.

Fig. 6. F456 is crucial for the generation of F456-targeting neutralizing antibodies.

a The contribution of the residue F456 was analyzed by CCP4 software. The dashed line represents the average contribution of epitope residues and serves as a reference rather than a strict cutoff. Antibodies with values on the dashed line show an F456 contribution equal to the average, whereas those above or below the line exhibit higher or lower contributions, respectively. F456 contribution was calculated as: (total atomic contacts between F456 residue and antibody/total atomic contacts between S-RBD and antibody)×100%, the distance cutoff is 4.5 Å. Average contribution was calculated as: (1/total number of amino acids in S-RBD which interact with antibody)×100%, using a distance cutoff is 4.5 Å. Results of the competitive ELISA using F456-targeting antibodies, including Omi-42 (b), CC12.3 (c), and S2K146 (d). The blue, red, and green lines represent PBS, RBD_WT-HR, and RBD_F456L-HR sera, respectively. e Relative contributions of F456 and L456 to the generation of 456-targeting antibodies quantified based on the competitive data in b–d. Results of the competitive ELISA using non-F456-targeting antibodies, including S309 (f) and P2B-2F6 (g). Data are presented as mean ± SEM. All graphs were created using GraphPad Prism 9.0. Comparison was performed by one-way ANOVA. A P-value below 0.05 was defined as indicating statistical significance (ns denotes non-significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001)

To evaluate the functional impact of F456 compared to L456 on neutralizing antibody-eliciting capacity, two F456-dependent mAbs with strong F456 dependency (Omi-42 and CC12.3) and one with a lower score (S2K146) were selected and expressed (Supplementary Fig. 8a–c). Competitive binding assays were performed using sera from mice immunized with either RBD_WT-HR or RBD_F456L-HR (Fig. 6b–d). Previous studies have demonstrated that antibodies generated during vaccination preferentially recognize the same antigen.¹¹ Sera from RBD_WT-HR-immunized mice showed significantly greater competition with F456-binding mAbs (Omi-42, CC12.3, and S2K146) than sera from RBD_F456L-HR-immunized mice against the F456-containing antigen. For the L456-containing antigen, enhanced competition by RBD_F456L-HR-immunized sera might be expected. However, the data revealed no significant differences between the two groups (right panels, Fig. 6b–d). To further quantify the effect of the F456L mutation, the relative contribution of F456- versus L456-targeting responses was calculated from the competitive neutralization data (Fig. 6e). This analysis demonstrated that F456 contributes significantly more to overall neutralizing activity than L456, supporting the conclusion that the F456L substitution substantially impairs the induction of antibodies targeting key immunogenic epitopes. As controls, mAbs that did not contact residue 456 (S309 and P2B-2F6) showed no significant differences in serum competition under either antigen-coating condition (Fig. 6f, g), confirming the site-specific nature of these observations.

These patterns indicate a shift in the functional composition of the antibody repertoire rather than a global change in antibody quantity. Although total binding antibody levels were consistent across groups, the RBD_F456L-HR vaccine did not elicit a high-titer neutralizing antibody response targeting the 456-associated epitope. This localized impairment underscores the pivotal role of residue 456 as a determinant of neutralizing immunogenicity within the receptor-binding motif (RBM).

Aromaticity at position 456 enhances antigen-BCR binding and promotes B cell activation

The affinity between antigen and germline-encoded BCR is the primary determinant for naive B cell recruitment and GC initiation.^21–23 To investigate how position 456 modulates initial antigen-BCR recognition, we first employed in silico modeling using AlphaFold3 and Rosetta Flex ddG. Substitution of phenylalanine (F456) with non-aromatic residues, such as leucine (L) or serine (S), significantly destabilized the antibody–antigen interface, whereas aromatic substitutions, such as tyrosine (Y) or tryptophan (W), largely preserved interface stability (Fig. 7a–e). Structural analysis revealed that F456 functions as an aromatic interaction hotspot, mediating π–π stacking interactions with heavy-chain residue F99 in the CC12.3 germline BCR (Fig. 7f, g). These results collectively indicate that aromatic residues at position 456 stabilize antigen–BCR binding.

Fig. 7. Aromatic residues at position 456 promote germline BCR engagement.

Calculated ΔΔG (kcal/mol) values for residue 456 in RBD interactions with germline antibodies, including CC12.3 (a), S2K146 (b), Omi-42 (c), S309 (d), and P2B-2F6 (e). All energy changes are expressed relative to the native F456 residue. The interactions between the residue 456 (aromatic residues in (f) and non-aromatic in (g) and the germline-CC12.3 antibody). The residues modeled to be involved in the antibody–antigen interaction at position 456 are shown as sticks. Gray dashed line: Hydrophobic Interaction; Dark blue dashed line: hydrogen bonds; Green dashed line: π-stacking; Orange dashed line: π–cation interaction. BLI sensorgrams of germline-encoded CC12.3 (h) and S309 (i) binding to RBD-HR variants containing either F456 or L456. Calcium flux in splenic B cells from mice immunized with either RBD_XBB.1.5-HR (j) or RBD_EG.5-HR (k) vaccines following stimulation with recombinant RBD-HR proteins. Representative curves are shown

To directly test the computational predictions, biolayer interferometry (BLI) assays were conducted using germline-encoded monoclonal antibodies. In agreement with the computational models, Germline-CC12.3 exhibited a 3.6-fold increase in the K_D value for RBD_F456L-HR compared to RBD_WT-HR (F456), confirming impaired germline antibody recognition caused by the F456L substitution (Fig. 7h). In contrast, Germline-S309, a Class 3 antibody that binds distal to position 456, exhibited nearly identical binding affinities for both RBD_WT-HR (F456) and RBD_F456L-HR. The measured K_D values were 1.66 × 10⁻⁹ and 1.75 × 10⁻⁹ M, respectively (Fig. 7i).

BLI analysis demonstrated preferential interaction with the aromatic F456 residue. To determine whether this preference affects functional B cell responses, calcium flux was measured following stimulation of splenic B cells from mice immunized with either RBD_XBB.1.5-HR (F456) or RBD_EG.5-HR (L456).²⁴ Across both immunization backgrounds, antigens containing F456 consistently induced greater calcium mobilization than those containing L456 (Fig. 7j, k). These findings indicate that aromatic residues at position 456 enhance BCR engagement and promote more robust B cell activation, consistent with the computational and BLI results.

Discussion

The ongoing evolution of SARS-CoV-2 continues to pose a persistent public health challenge, underscoring the urgent need for next-generation antiviral strategies and vaccines with broad and durable efficacy. Not only antigenic drift but also the potential impact of viral evolution on the intrinsic immunogenicity of spike antigens, a dimension that has received comparatively less attention. Central to this challenge is the spike glycoprotein, whose rapid diversification under immune pressure drives both enhanced hACE2 binding affinity and increased immune escape. These evolutionary adaptations enable emerging variants to partially evade pre-existing humoral immunity, thereby reducing the effectiveness of vaccines and therapeutic antibodies developed against earlier strains.^2–4 In addition to antigenic escape, accumulating evidence suggests that spike mutations may also modulate the intrinsic immunogenic properties of viral antigens, thereby influencing the magnitude, quality, and durability of vaccine-induced immune responses.^9,10,25 However, the functional relationship between immune evasion and immunogenicity remains insufficiently understood.

Among recently emerged Omicron sublineages, the recurrent appearance of the F456L substitution across genetically distinct lineages highlights its evolutionary importance. Notably, this mutation does not significantly alter ACE2 binding affinity (Supplementary Fig. 2),^26,27 suggesting that its selective advantage is more likely associated with immune modulation rather than receptor engagement. F456L has been frequently observed across emerging Omicron sublineages and has been reported to reduce susceptibility to neutralizing antibodies in prior studies.^9,10 Consistent with these observations, our data suggest that F456L is associated with reduced functional immunogenicity.

Using a structure-guided RBD-HR vaccine platform, we found that the RBD-HR vaccine carrying the F456L mutation elicited lower neutralizing antibody titers than the RBD-HR vaccine with F456, and also induced reduced germinal center responses. Furthermore, we determined that the immunological consequences of this mutation were associated with restoration of aromatic residues at position 456, including tyrosine (Y456) and tryptophan (W456). These substitutions restored the neutralizing immunogenicity of the RBD, indicating that the aromaticity of residue 456 is crucial for neutralizing antibody induction. This discovery provides important insights for rational antigen design, particularly in the development of RBD-based vaccines.

Single-cell V(D)J sequencing of antigen-specific B cells revealed that the F456-containing immunogens elicited higher proportions of germinal center B cells, plasma cells, and memory B cells, and more effectively expanded the dominant IGHV clonotypes compared to L456. Structural modeling and calcium flux assays further indicate that aromatic residues at position 456 may stabilize π–π interactions with germline BCRs, thereby enhancing B-cell activation. In addition to intrinsic B-cell activation, the initial interaction between the antigen and the innate immune system is critical for priming the adaptive response. Notably, our data also suggest that F456L may influence early antigen recognition during immune priming by innate and adaptive immune interfaces (Supplementary Fig. 5), further reinforcing its role as a key immunological determinant.

Although the F456L mutation reduced neutralizing antibody titers, total antigen-specific IgG levels remained largely unchanged (Supplementary Fig. 7). This indicates that the mutation does not broadly impair B cell activation or antibody production, but rather selectively affects the quality of the antibody response. This conclusion was further supported by competitive binding assays, which demonstrated that the F456L substitution selectively impairs the elicitation of neutralizing antibodies targeting residue F456. These results highlight the critical role of residue F456 within a major neutralizing epitope and its substantial influence on shaping the neutralizing antibody response. Importantly, they also reveal a key distinction between antigen-binding immunogenicity and functional immunogenicity, the latter being more closely associated with protective efficacy.

Interestingly, while EG.5-based mRNA or multivalent protein vaccines have been reported to elicit robust neutralization responses.^28,29 Our RBD-HR subunit platform exhibited a more pronounced sensitivity to the F456L substitution, resulting in relatively lower titers. This discrepancy likely stems from differences in antigenic breadth: the full-length spike protein incorporates distal epitopes within the N-terminal domain (NTD) and S2 subunit, which may provide a structural and immunological buffer that mitigates the impact of single-residue substitutions. In contrast, our focused RBD-HR model serves as a high-resolution tool by isolating the RBM–BCR interface.³⁰

Despite these findings, several limitations should be noted. Our analyses were primarily based on murine models, which may not fully reflect the human condition. In addition, while most SARS-CoV-2 variants that emerged in recent years carry only the F456L mutation, we observed that a small proportion of variant viruses from late 2023 (https://cov-spectrum.org) carried a combination of L455F and F456L (also referred to as the “FLip” mutations).³¹ It would be interesting to explore in future studies whether the aromatic substitution at position 455 could compensate for the detrimental effects of 456L on RBD immunogenicity. Moreover, the RBD-HR platform focuses on RBM–BCR interactions, and the observed effects may be less pronounced in the context of the full-length spike protein. Further studies using diverse vaccine platforms will help clarify how these structural features contribute to protective immunity in more physiologically relevant settings. Furthermore, the effect of F456L on functional immunogenicity may be influenced by the host’s immune background. We found that, in RBD_XBB.1.5-HR-primed mice, boosting with either RBD_XBB.1.5-HR or RBD_EG.5-HR antigens elicited statistically comparable binding antibody levels and neutralization titers (Supplementary Fig. 11), suggesting that immune imprinting may partially obscure the immunogenic differences of single-site mutations during recall responses. Consequently, in response to every new variant, prioritizing antigenic potency and quality over achieving a perfect sequence match may represent a better strategy for vaccine development. These findings suggest that prioritizing antigenic potency and preservation of critical structural anchors, rather than achieving exact sequence matching alone, may represent a valuable strategy for future vaccine design. This approach may provide a more robust means of sustaining durable neutralizing immunity.

In conclusion, our study highlights the significant impact of the substitution of an aromatic amino acid at position 456 with a non-aromatic residue on the functional immunogenicity of RBD. The F456L mutation reduces the generation of neutralizing antibodies targeting residue 456 and impairs BCR recognition of the antigen. However, this reduction can be compensated for by replacing non-aromatic amino acids with aromatic ones. This work provides valuable insights into how mutations at specific sites influence antigen functional immunogenicity, which should inform future vaccine design and contribute to our understanding of immune response mechanisms.

Methods

Cell lines

Human embryonic kidney 293T (HEK-293T-hACE2) cells were constructed by our group as previously reported.³² Cells were cultivated in Dulbecco’s modified Eagle medium (DMEM, Gibco), supplemented with 10% fetal bovine serum (FBS, Gibco) and 1% penicillin–streptomycin (Gibco) at 37 °C in 5% CO₂. HEK293F suspension cells (Thermo Fisher Scientific, R79007) were cultured in serum-free SMM 293-TII Expression Medium (Sino Biological Inc.) at 37 °C under 5% CO₂ with shaking at 120 rpm.

Animals

Female NIH mice (6–8 weeks old) and female C57BL/6 mice (4–6 weeks old) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. Mice were acclimated for 7 days before the experiments. All animal procedures complied with the guidelines approved by the Experimental Animal Ethics Committee of the State Key Laboratory of Biotherapy, Sichuan University (Approval No. 20230307059).

Expression and purification of RBD-HR

The RBD-HR proteins employed in this investigation were synthesized using the Bac-to-Bac baculovirus expression system (Invitrogen), following previously established protocols.¹¹ Detailed amino acid substitutions for each variant are presented in Table S1. The construct design involved the direct fusion of the S-RBD sequence (amino acids 320–545) with the HR1 (amino acids 916–966) and HR2 sequences (amino acids 1157–1203) of the SARS-CoV-2 S2 subunit. To ensure optimal expression and folding, the N-terminus of the construct was fused with a GP67 signal peptide, a Thioredoxin (Trx) tag, and a 6×His tag, followed by an enterokinase (EK) cleavage site (Fig. 1d). Notably, all N-terminal fusion tags were enzymatically removed via EK cleavage after purification, yielding the final tag-free antigen for subsequent evaluation.³³

Gene cloning

To construct the expression plasmids, the target coding sequences were amplified and inserted into the pFastBac1 backbone utilizing BamHI and HindIII cloning sites. Once the sequences were confirmed to be accurate, the plasmids were introduced into Escherichia coli DH10Bac competent cells, enabling the generation of recombinant bacmids.

Protein expression

For the synthesis of target proteins, purified bacmids were transfected into Sf9 insect cells. After an incubation period of approximately six days, the culture medium containing the packaged recombinant baculoviruses was collected. This viral supernatant was subsequently used to infect fresh Sf9 cultures for large-scale protein production.

Protein purification, the supernatants containing the target protein were collected about 72 h after infection and subjected to primary purification using Ni-NTA resin (Qiagen). The recovered proteins were buffer-exchanged into a cleavage buffer using Amicon Ultra-15 centrifugal filters through repeated concentration and dilution cycles. Subsequently, the recovered proteins underwent cleavage with EK protease at a ratio of 1 U:1 mg protein, incubated at 16 °C for ~18 h. The cleaved tag was removed through affinity chromatography using Ni-NTA resin. The resulting flow-through solution, enriched with S-RBD-HR protein, was then collected, concentrated, and loaded onto a Superdex 200 Increase 10/300 GL column (Cytiva) for further purification. The protein’s purity was assessed through SDS–PAGE and visualized by Coomassie blue staining. The point mutations introduced into the RBD-HR constructs are listed in Table S1.

Vaccine formulation and immunization

Spike-XBB.1.5 and spike-EG.5 were purchased from ARCO. Trimeric RBD-HR proteins were mixed with an MF59-like adjuvant at a 1:1 volume ratio to a final protein concentration of 100 μg/mL. A total volume of 100 μL was administered intramuscularly into the hind leg of each mouse.¹¹ A total of six mice were intramuscularly immunized with two doses of recombinant protein vaccines on days 0 and 21, comprising the following formulations: (1) RBD_XBB.1.5-HR and RBD_EG.5-HR; (2) RBD_JN.1-HR and RBD_JN.1.16-HR; (3) spike-XBB.1.5 and spike-EG.5; (4) RBD_WT-HR, RBD_F456L-HR, and RBD_L452R-HR; and (5) RBD_XBB.1.5-HR, RBD_EG.5-HR, RBD_EG.5-HR(L456Y), and RBD_EG.5-HR(L456W). Serum samples were collected on day 35, and neutralizing antibody responses against pseudoviruses were assessed. A comprehensive schematic of the immunization and sampling schedule is illustrated in Supplementary Fig. 9.

Live SARS-CoV-2 virus neutralization assay

The neutralization capacity of sera against authentic JN.1 and XBB.1.5 strains was determined utilizing a live-virus neutralization assay. We first prepared serial dilutions of the sera, which were then allowed to interact with live SARS-CoV-2 (adjusted to a 50% tissue culture infectious dose, TCID50) for 60 min at 37 °C. Following this pre-incubation step, the resulting virus–serum mixtures were inoculated onto Vero cell monolayers (ATCC CCL-81) seeded at 50,000 cells per well. After a 72-h cultivation period, the plates were visually inspected under a microscope for cytopathic effects (CPE). The final neutralizing titer was recorded as the maximum dilution factor capable of suppressing 50% of the viral-induced CPE.

SARS-CoV-2 pseudovirus-based neutralization assay

Luciferase-expressing pseudoviruses representing various SARS-CoV-2 strains were procured from Genomeditech (China, Shanghai). Sera were serially diluted and incubated with the pseudoviruses in 96-well dark plates (WHB) at 37 °C. After one hour, 1.5 × 10⁴ cells were added to each well and incubated for 48 h. Subsequently, the culture medium was removed, and 100 μL of the firefly luciferase reporter gene assay kit (Beyotime, China) was added. The relative light units were measured using a multi-mode microplate reader (PerkinElmer, USA) with Kaleido 3.0 software. The 50% inhibitory concentration (IC₅₀) was analyzed through GraphPad Prism 9.0 software using the log (inhibitor) vs. normalized response model.

Kinetics by bio-layer interferometry (BLI)

The binding kinetics and affinity of RBD-HR variants to recombinant human ACE2 and germline neutralizing antibodies (CC12.3 and S309) were characterized using the Gator system (Gator Bio) at 25 °C with a shaking speed of 1000 rpm. All reagents and proteins were prepared in PBS containing 0.1% Tween 20 (PBST) buffer (1 × PBS, 0.05% Tween-20, pH 7.4).

Affinity between RBD-HR and hACE2: To assess the receptor-binding capacity, Fc-tagged hACE2 (Arco, AC2-H5257) was captured onto Gator Fc probes (Gator) to reach a stable loading response of ~0.5–1.0 nm. Following a 60 s baseline step in PBST, the sensors were dipped into wells containing serially diluted RBD-HR variants (RBD_WT-HR or RBD_F456L-HR) as the analytes. The concentrations of RBD-HR ranged from 25 to 1.56 nM (two-fold serial dilution). The association phase was monitored for 180 s, followed by immersion in PBST buffer for 300 s to monitor the dissociation phase.

Affinity between RBD-HR and germline antibodies: To evaluate antibody–antigen interactions, germline neutralizing antibodies (CC12.3 and S309) were immobilized onto Gator Fc probes via their Fc domains. Subsequently, the sensors were moved into wells containing serially diluted RBD-HR variants at a starting concentration of 100 nM. The association and dissociation timings were consistent with the hACE2 binding assay. The probes were regenerated between cycles using 10 mM glycine (pH 3.0) to restore the baseline. The association rate (k_on), dissociation rate (k_off), and equilibrium dissociation constant (K_D) were calculated by globally fitting the data to a 1:1 Langmuir binding model using the Gator Analysis Software. It should be noted that the RBD-HR constructs used in these experiments are trimeric. The 1:1 Langmuir binding model provides apparent values rather than exact measurements and was employed for simplicity, allowing comparative analysis across different RBD-HR variants.

HPSEC

The homogeneity and oligomeric distribution of purified RBD-HR proteins were analyzed using high-performance size-exclusion chromatography (HPSEC) on a Waters e2695 system. Chromatographic separation was performed using an Agilent SEC-300 column (5 μm, 4.6 × 300 mm) at 30 °C. The mobile phase was 50 mM phosphate buffer (PB) supplemented with 200 mM arginine hydrochloride, delivered at a flow rate of 0.35 mL/min. Protein elution was monitored by UV absorbance at 280 nm. Samples were centrifuged prior to injection to improve clarity. The relative proportions of trimers, aggregates, and fragments were determined by integrating peak areas using the system’s workstation software.

Differential scanning fluorimetry (DSF) assay

The thermal stability of RBD_XBB.1.5-HR and the RBD_EG.5-HR variants was measured by differential scanning fluorimetry (DSF) as previously described.^34,35 SYPRO Orange dye (Sigma-Aldrich) was used for protein thermal denaturation. Recombinant RBD-HR proteins were diluted to 5 μM in a buffer with 20 mM HEPES (pH 7.5) and 150 mM NaCl. For DSF, 20 μL of each protein–dye mix was loaded into a 96-well PCR plate and heated from 25 to 95 °C at 1 °C/min over 70 min using a CFX connect real-time PCR system (Bio-Rad). Fluorescence signal was monitored continuously as a function of temperature. Melting temperature (T_m), defined as the midpoint of unfolding, was calculated by fitting the first derivative of the fluorescence curve (−dRFU/dT) to a Gaussian function using Bio-Rad CFX Manager software. Further analysis was performed with GraphPad Prism 9.0.

Antibody expression and purification

Genes encoding the antibody variable regions (heavy and light chains) were cloned into the pCAGGS vector. This vector contained the signal peptide sequence (METDTLLLWVLLLWVPGSTGD) and the constant region of human IgG1 (for the heavy chain), Igκ (for the κ chain), or Igλ (for the λ chain), thereby generating antibody expression plasmids. Heavy- and light-chain plasmids were transiently co-transfected into Expi293F cells (ATCC). In brief, the HEK293F cells were cultured in serum-free SMM 293-TII Expression Medium (Sino Biological Inc.) with 1% penicillin–streptomycin (Gibco). Antibody-encoding DNA plasmid was transfected into HEK293F cells using PEI, with a plasmid-to-PEI mass ratio of 1:3. Additionally, plasmid DNA was used at 1 mg/L culture volume. Four days later, supernatants were collected and antibodies purified with protein G beads (Smart-Lifesciences). Purity and molecular weight of antibodies were analyzed by SDS–PAGE (Supplementary Fig. 4d and e).

Contact analyses of antibodies with the RBD or spike protein

The contacts between antibodies and the RBD or spike protein were analyzed using the CCP4 suite. CB6, PDB is 7c01; S2K146, PDB is 7tas; Omi-42, PDB is 7zr7; CC12.3, PDB is 6xc4; CV07-250, PDB is 6xkq; COVOX-88, PDB is 7bel; BD-771, PDB is 7ey5; XBB-6, PDB is 8qrf; XBB-2, PDB is 8qrg; CV2.6264, PDB is 7qf1; P2C-1F11, PDB is 7cdi; C102, PDB is 7k8m; B38, PDB is 7bz5; C1A-F10, PDB is 7kfy; COVA2-04, PDB is 7jmo; CV30, PDB is 6xe1; C105, PDB is 6xcn; BD-236, PDB is 7x63; BD-629, PDB is 7CH5; LY-CoV488, PDB is 7KMH. It is important to clarify that this antibody pool corresponds to established structural data curated from public databases (e.g., PDB and CoV-AbDab) and is distinct from the experimental results generated from the immunized mice in this study.

Enzyme-linked immunosorbent assay (ELISA)

Antigen-specific IgG titers were measured using a standard indirect ELISA. Briefly, a 96-well microtiter plate was coated with 100 μL/well of RBD-HR (1 μg/mL) in coating buffer and incubated overnight at 4 °C. After three washes with PBST, wells were blocked with 200 μL of 1% bovine serum albumin (BSA; Invitrogen) for 1 h at 37 °C. Serially diluted serum samples (100 μL) were added and incubated for 1 h at 37 °C. Following three washes, 100 μL of HRP-conjugated anti-mouse IgG (1:10,000; Thermo Fisher Scientific) was added and incubated for 1 h at 37 °C. After five final washes, 100 μL of 3,3’,5,5’-tetramethylbenzidine (TMB; Thermo Fisher Scientific) was added and incubated at room temperature for 10 min. The reaction was stopped with 100 μL of TMB stop solution (Beyotime), and absorbance at 450 and 630 nm was measured using a microplate reader (Spectramax ABS, Molecular Devices) with SoftMax Pro 7.1 software.

Competitive ELISA (cELISA)

For the competitive enzyme-linked immunosorbent assay (cELISA) experiments, all antibodies were biotinylated using the Super Biotin Rapid Labeling Kit (ARL0020S, Frdbio). The antibody S309 (ATMA10183Mo) was obtained from AtaGenix. A 96-well microtiter plate was coated with 100 μL of either RBD_WT-HR or RBD_F456L-HR (0.5 μg/mL) per well and incubated overnight at 4 °C. Following three washes with PBST, the wells were blocked with 200 μL of 1% bovine serum albumin (BSA) (Invitrogen) for 1 h at 37 °C. After another three washes, 100 μL of biotin-labeled antibodies at varying concentrations (100, 50, 25, 12.5, 6.25, 3.125, 1.5625 ng/mL) was added to the wells and incubated for 1 h at 37 °C. Post-incubation, the wells were washed three times, followed by the addition of 100 μL of Streptavidin–HRP (1:5000 dilution, A0305, Beyotime), and incubated at 37 °C for 1 h. The concentration of each biotinylated antibody corresponding to an A_450–630 value of ~1.0 was selected for subsequent competitive assays.

For the cELISA, the RBD_WT-HR or RBD_F456L-HR protein was coated onto wells overnight at 4 °C as previously described. After washing, the wells were blocked with 200 μL of 1% BSA at 37 °C for 1 h. Following each incubation, wells were washed five times with PBST. Serial dilutions of serum (100 μL) were added to the wells, followed by a 1h incubation at 37 °C. After washing, the biotin-labeled antibodies were added and incubated at 37 °C for an hour. Subsequently, 100 μL of Streptavidin–HRP (1:5000 dilution) was added and incubated at 37 °C for 1 h. The subsequent colorimetric development and measurement were performed as described for the standard ELISA. The percentage contribution of the 456-residue to the overall competitive inhibition was calculated using the following formula:

For a plate coated with RBD_WT-HR protein:

$$\frac{({\mathrm{A}}_{450-630}\mathrm{PBS}-{\mathrm{A}}_{450-630}{\mathrm{RBD}}_{\mathrm{WT}}-\mathrm{HR})-({\mathrm{A}}_{450-630}\mathrm{PBS}-{\mathrm{A}}_{450-630}{\mathrm{RBD}}_{\mathrm{F}456\mathrm{L}}-\mathrm{HR})}{A_{450-630}\mathrm{PBS}}\times 100\mathrm{\%}$$

For a plate coated with RBD_F456L-HR protein:

$$\frac{({\mathrm{A}}_{450-630}\mathrm{PBS}-A_{450-630}{\mathrm{RBD}}_{\mathrm{F}456\mathrm{L}}-\mathrm{HR})-({\mathrm{A}}_{450-630}\mathrm{PBS}-{\mathrm{A}}_{450-630}{\mathrm{RBD}}_{\mathrm{WT}}-\mathrm{HR})}{{\mathrm{A}}_{450-630}\mathrm{PBS}}\times 100\mathrm{\%}$$

Note: To estimate the relative contribution of residue 456 to serum antibody competition, we operationally separated the competitive inhibition attributable to residue 456 from that attributable to the remaining residues within the same epitope. When RBD_WT-HR (F456) was used as the coating antigen, the difference between the PBS control and the RBD_WT-HR immune serum (A_450–630 PBS- A_450–630 RBD_WT-HR) was considered to represent the total contribution of antibodies recognizing both residue 456 and surrounding epitope residues. In contrast, under the assumption that RBD_F456L-HR immune serum does not efficiently induce antibodies recognizing the F456 residue, the difference between the PBS control and the RBD_F456L-HR immune serum (A_450-630 PBS- A_450-630 RBD_F456L-HR) was considered to primarily reflect antibodies recognizing surrounding epitope residues independent of F456. Therefore, the additional competitive inhibition observed with RBD_WT-HR immune serum relative to RBD_F456L-HR immune serum was interpreted as the relative contribution of the F456 residue to antibody competition. A reciprocal analysis was performed using RBD_F456L-HR as the coating antigen to estimate the relative contribution of the L456 residue.

Flow cytometry

On day 35 post-immunization, NIH mice were euthanized, and lymphocytes were isolated from the inguinal lymph nodes, spleen, and bone marrow. Flow cytometry was used to identify specific lymphocyte populations. The gating strategy is shown in Supplementary Fig. 10. Single-cell suspensions were incubated with the indicated fluorochrome-conjugated antibodies (1:100 dilution) for 30 min at 4 °C. For the identification of T follicular helper (Tfh) cells, cells were stained with PerCP/Cyanine5.5-conjugated anti-mouse CD3 (BioLegend, 100328), Brilliant Violet 605-conjugated anti-mouse CD19 (BioLegend, 115539), FITC-conjugated anti-mouse CD4 (BioLegend, 130308), PE/Cyanine7-conjugated anti-mouse CXCR5 (BioLegend, 145516), and Brilliant Violet 421-conjugated anti-mouse PD-1 antibodies (BioLegend, 135218). To detect antigen-specific GC B cells, MBCs, and plasma cells, cells were incubated with biotinylated SARS-CoV-2 RBD protein at room temperature for 40 min. Subsequently, the following panels were used:

GC B cells: PerCP/Cyanine5.5-conjugated anti-mouse CD3, Brilliant Violet 605-conjugated anti-mouse CD19 (BioLegend, 115539), Brilliant Violet 421-conjugated anti-mouse GL7 (BioLegend, 144614), PE/Cyanine7-conjugated anti-mouse CD95 antibodies (BioLegend, 152618), and PE-conjugated Streptavidin (BioLegend, 405204).

MBCs: PerCP/Cyanine5.5-conjugated anti-mouse CD3, FITC-conjugated anti-mouse CD45R/B220 (BioLegend, 103206), Brilliant Violet 605-conjugated anti-mouse CD19, Brilliant Violet 510-conjugated anti-mouse IgD (BioLegend, 405723), PE/Cyanine7-conjugated anti-mouse CD38 antibodies (BioLegend, 165612), and Brilliant Violet 421-conjugated anti-mouse GL7 (BioLegend, 144614), and PE-conjugated Streptavidin (BioLegend, 405204).

Plasma cells: FITC-conjugated anti-mouse CD45 (BioLegend, 147710), Brilliant Violet 510-conjugated anti-mouse IgD (BioLegend, 405723), PE/Cyanine7-conjugated anti-mouse B220/CD45R (BioLegend, 103222), Brilliant Violet 650-conjugated anti-mouse CD138 (BioLegend, 142518), and PE Streptavidin (BioLegend, 405204). For viability discrimination, cells were incubated with DAPI staining solution (Beyotime, C1005) for 5 min prior to flow cytometric acquisition.

For intracellular cytokine staining (ICS), splenic lymphocytes were cultured in RPMI 1640 complete medium supplemented with 10% fetal bovine serum (FBS), 100 μg/ml streptomycin, 100 U/mL penicillin, 1 mM sodium pyruvate (Gibco, USA), 50 μM β-mercaptoethanol, and 20 U/mL IL-2 (Sigma-Aldrich). Cells were stimulated with RBD peptide pools (each peptide pool at a final concentration of 1 μg/mL), each consisting of 15-mer peptides overlapping by 11 amino acids and covering the spike protein of XBB.1.5, EG.5.1, or JN.1 variants (custom-designed by GenScript). To allow intracellular accumulation of cytokines, a protein transport inhibitor (BD GolgiStop™, BD, 554715) was added prior to surface staining. Cells were then stained with PerCP/Cyanine5.5-conjugated anti-mouse CD3, PE-conjugated anti-mouse CD8 (BioLegend, 100708), FITC-conjugated anti-mouse CD4 (BioLegend, 130308), Brilliant Violet 510-conjugated anti-mouse CD44 (BioLegend, 103044) for 25 min at 4 °C. DAPI was added to exclude dead cells. Cells were then fixed and permeabilized using the Fixation/Permeabilization Kit (BD, 554715) in combination with BD GolgiStop™, followed by intracellular staining with PE/Cyanine7-conjugated anti-IFN-γ (BioLegend, 505826) for 1 h at room temperature in the dark. Samples were acquired on a NovoCyte flow cytometer (ACEA Biosciences), and data were analyzed using NovoExpress software (v1.4.1).

Generation and characterization of BMDCs and BMDMs

Bone marrow-derived dendritic cells (BMDCs) and macrophages (BMDMs) were generated from 4 to 6-week-old female C57BL/6 mice. Bone marrow cells were flushed from the femurs and tibias and filtered through a 70 μm cell strainer to obtain single-cell suspensions.

For BMDC preparation, cells were cultured in complete RPMI 1640 medium (Gibco) supplemented with 10% FBS, 1% penicillin–streptomycin, 1 mM sodium pyruvate, and 50 μM β-mercaptoethanol. Recombinant murine GM-CSF (20 ng/mL) and IL-4 (10 ng/mL) were added to induce differentiation. On day 3, half of the medium was replaced with fresh medium containing the same cytokine concentrations. Cells were collected on day 7 for subsequent assays. For BMDMs, bone marrow cells were cultured in complete RPMI 1640 medium containing recombinant murine M-CSF (20 ng/mL). After a 7-day culture, cells were collected as BMDMs.

To assess antigen recognition, BMDCs or BMDMs were incubated with FITC-labeled RBD-HR variants (2 μg/mL) at 37 °C for 1 h. Cells were then washed three times with cold PBS to remove unbound antigens. BMDCs were stained with anti-mouse CD11c antibody (BioLegend, 117306), while BMDMs were stained with CD11b (BioLegend, 101212) and F4/80 (BioLegend, 123108) antibodies to identify macrophage populations. The frequency of antigen-positive cells was analyzed using a NovoCyte flow cytometer (ACEA Biosciences), and data were analyzed using NovoExpress software (v1.4.1).

ELISpot assay

The frequency of antigen-specific B cells was determined using an enzyme-linked immunospot (ELISpot) assay.^12,36 Briefly, a 96-well 0.45 μm hydrophobic high protein-binding Immobilon-P membrane was activated with 15 μL of 70% ethanol, followed by pre-coating with 3 μg/mL RBD-HR protein in PBS. The plates were incubated overnight at 4 °C. Afterward, they were blocked with RPMI 1640 medium supplemented with 10% FBS for 2 h at 37 °C.

Single cell suspensions were prepared and plated at a density of 2 × 10⁵ cells per well. The plates were incubated overnight at 37 °C in a 5% CO₂ atmosphere. After incubation, cells were washed five times with PBS. Next, goat anti-mouse IgG (H + L) secondary antibody conjugated to HRP (Invitrogen, 31430) was added at a 1:10,000 dilution and incubated for 2 h at room temperature. After washing, ELISpot substrate (TMB for HRP, MabTech, 3651-10) was applied for 15 min to develop the spots. The color reaction was stopped by extensive washing with deionized water. Once dried, spots were counted using an automated ELISpot reader (Cellular Technology Ltd.).

Single-cell RNA and BCR sequencing

To ensure a sufficient yield of rare antigen-specific B cells for downstream library construction, splenic mononuclear cells from six mice per immunization group (pooled from six mice per group) were pooled prior to fluorescence-activated cell sorting (FACS). This pooling strategy was a technical necessity to meet the cell-input thresholds required for the 10× Genomics platform, given the extremely low frequency of RBD-specific B cells within the total lymphocyte population. Spleens were collected from mice immunized with either the RBD_XBB.1.5-HR or RBD_EG.5-HR recombinant protein vaccine. For each group, single-cell suspensions were prepared. RBD-specific B cells were enriched using biotinylated RBD probes (biotinylated RBD probes followed by streptavidin-PE staining), followed by FACS. Live CD3⁻CD19⁺RBD⁺ B cells were sorted using a FACSAria^TM Fusion cytometer (BD Biosciences).

Single-cell RNA sequencing and V(D)J repertoire libraries were prepared using the 10× Genomics Chromium Single Cell 5’ Reagent Kit v3 according to the manufacturer’s instructions.³⁷ Briefly, barcoded single-cell cDNA was generated and amplified to simultaneously profile the transcriptome and V(D)J regions. Libraries were sequenced on an Illumina NovaSeq 6000 platform. Raw sequencing reads were processed with Cell Ranger (v6.1, 10x Genomics) using the Cell Ranger Count and Cell Ranger V(D)J pipelines, aligning to the mm10 reference genome. Subsequent analyses, including dimensionality reduction, clustering, B cell subset identification, and clonal tracking, were conducted using Seurat (v4.3).³⁸

Unsupervised identification of cell states from scRNA-seq data

The combined expression data were analyzed using the Seurat pipeline.²⁰ After scaling, principal component analysis was performed on the top 2000 variable genes. The first 30 principal components were used for dimensionality reduction via UMAP, resulting in a two-dimensional visualization of cellular heterogeneity. Cell clusters were identified using the FindClusters function (Seurat v4.3) across resolutions ranging from 0.1 to 1.0. A resolution of 1.0 was selected, which maximized the average silhouette width and yielded 11 distinct clusters. After filtering out low-quality and doublets, 11 high-confidence clusters comprising 38,369 cells were retained for further analysis.

Identification and annotation of cluster-specific marker genes

To identify cluster-specific marker genes, we applied the FindAllMarkers and FindMarkers functions using the Wilcoxon rank-sum test. Genes were considered significant if they were expressed in more than 20% of cells within a cluster and had an adjusted P-value < 0.05 (Benjamini–Hochberg correction). Cell type annotation was performed based on the expression of canonical marker genes, and manual curation was conducted to ensure biological accuracy. In addition, a custom panel of curated marker genes was used to refine the classification of distinct cell populations.

BCR repertoire analysis

Raw sequencing data from single-cell V(D)J libraries were processed using Cell Ranger (10× Genomics, v6.1) to assemble full-length BCR sequences and assign clonotypes. The resulting contig annotations were further analyzed using the Immcantation framework, which enabled a detailed assessment of clonal relationships. To explore the distribution and dynamics of BCR clonotypes across different B cell subsets, we utilized the scRepertoire R package, which provides a suite of functions for single-cell immune receptor profiling and visualization. All analyses were conducted using default parameters unless otherwise specified.

Calculation of binding free energy changes (ΔΔG) between BCR and RBD

To evaluate the impact of specific amino acid mutations on the stability of BCR–antigen binding, we predicted the germline V(D)J gene segments of antibodies using the IMGT/V-QUEST server.³⁹ Then, antibody complex structures were predicted using AlphaFold3.⁴⁰ Subsequently, the binding free energy changes (ΔΔG) induced by the mutations were calculated using the Flex_ddG protocol implemented in Rosetta. The computational protocol began with energy minimization of the initial structure under geometric constraints to establish reference conformations for both WT and mutant systems. Backbone flexibility at the protein–protein interface was systematically sampled using the Backrub Mover to generate structural ensembles. Following conformational sampling, point mutations were introduced, followed by side-chain repacking and subsequent energy minimization to optimize local geometries. The resulting conformations were scored using the fa_talaris 2014 energy function, with ensemble-averaged binding energies calculated to account for backbone-dependent energetic fluctuations.⁴¹

All simulations were executed using the flex_ddG.xml script provided in the supplementary materials of Barlow et al.⁴¹ executed on a computational system. For each mutation site, the protocol performed local conformational sampling on backbone and sidechain residues within an 8 Å radius of the mutation site, followed by global sidechain optimization to minimize energy. The Backrub simulations used up to 35,000 Monte Carlo steps, generating 50 output models each for both WT and mutant complexes. A maximum of 5000 minimization iterations was implemented with an absolute convergence threshold of 1.0 and a backrub trajectory stride of 35,000. Binding energy differences (ΔΔG) were calculated as

$${\mathrm{\Delta }\mathrm{\Delta }G}_{\mathrm{bind}}={\mathrm{\Delta }G}_{\mathit{bind}}^{\mathit{MUT}}-{\mathrm{\Delta }G}_{\mathit{bind}}^{\mathit{WT}}$$

where more negative ΔΔG values indicate increased binding stability, final ΔΔG values were determined by averaging energy calculations across all 50 structural models to ensure statistical reliability.

Intracellular calcium measurement

B cells from the spleens of mice immunized with the RBD_XBB.1.5-HR or RBD_EG.5-HR vaccine were isolated using the EasySep™ Mouse B Cell Isolation Kit (STEMCELL, #19854 A). These cells were then incubated with the cell-permeant calcium indicator Fluo-4, AM (Beyotime, S1060) at a final concentration of 2 μM in pre-warmed Hanks’ Balanced Salt Solution (without Ca²⁺ and Mg²⁺). The incubation was carried out at 37 °C for 30 min in the dark. After washing, the cells were resuspended in complete RPMI 1640 medium supplemented with 10% FBS, 100 μg/ml streptomycin, 100 U/ml penicillin, 1 mM sodium pyruvate (all from Gibco, USA), 50 μM β-mercaptoethanol (Sigma-Aldrich, USA), 100 ng/ml IL-2, 10 ng/ml IL-4 (both from Novoprotein), and 5 μg/ml recombiMAb anti-mouse CD40 (BioXcell, CP133).⁴²

The cells were then treated with 100 μg/ml of RBD_XBB.1.5-HR, RBD_EG.5-HR, RBD_EG.5-HR(L456Y), or RBD_EG.5-HR(L456W) protein. As a control, a separate group of cells was treated with 100 μg/ml ovalbumen (OVA) protein (Sigma, A5503). Real-time calcium mobilization was tracked utilizing a PerkinElmer microplate reader (excitation at 488 nm; emission at 516 nm). Baseline fluorescence (F₀) was recorded for 1 min prior to stimulation, followed by continuous signal acquisition every 3 s for 3 min post-stimulation. Data were normalized as ΔF/F₀ (where ΔF = F−F₀). All procedures were performed under light-protected conditions to prevent photobleaching.

Statistical analysis

Statistical analyses were conducted using GraphPad Prism 9.0. Data are presented as geometric means ± geometric SD. The half-maximal inhibitory concentrations (IC₅₀) of neutralizing antibodies were determined through the log (inhibitor) vs. normalized response model. Unpaired Student’s t-test, one-way, and two-way analysis of variance (ANOVA) were used for statistical comparisons. P-values < 0.05 were considered significant, ****P < 0.0001; ***P < 0.001; **P < 0.01; *P < 0.05; ns, not significant.

Author contributions

W.X. conceptualized and supervised the study. W.X., L.G., and W.Y. designed the experiments and performed the investigation. C.Z. prepared the recombinant protein. P.D. and H.C. prepared the vaccination formulation and administered the vaccination. Z.H. performed the Single-cell RNA-seq data analysis. D.M. carried out computational energy analyses. P.D., C.Y., H.C., F.M., D.X., H.W., and L.H. collected serum samples and performed pseudovirus neutralization experiments in mice. Y.J., S.J., H.X., Y.C., Q.F., Z.Y., H.Y., W.B., R.W., C.L., L.J., A.A., S.H., and L.Y performed flow cytometry analysis. Y.J. and L.S. performed the authentic virus neutralization assays. P.D. and H.W. wrote the original draft of the manuscript. L.S. designed the monoclonal antibodies. P.D. expressed the antibodies and conducted the cELISA experiments. C.Z. and P.D. expressed the germlined antibodies and performed experiments characterizing protein structure and stability. P.D. and F.M. conducted the BLI assays. P.D., H.C., and H.W. reviewed and edited the manuscript. All authors have read and approved the article.

Data availability

Data supporting the findings of this study are available within the article and its supplementary materials. All computational data generated in this study, including Rosetta simulation files and analysis scripts, are available at Zenodo (https://doi.org/10.5281/zenodo.15516632). The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive (Genomics, Proteomics & Bioinformatics 2025) in National Genomics Data Center (Nucleic Acids Res 2025), China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA: CRA045057) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa.^32,43 To facilitate data retrieval, please note that the sample designations used in this manuscript correspond to the following GSA database sample IDs: Naive-F456 (PDD_XBB_1_5), Naive-L456 (PDD_EG_5_1), Immunized-F456 (PDD_130), and Immunized-L456 (PDD_168).

Competing interests

Although Xiawei Wei serves on the editorial board of Signal Transduction and Targeted Therapy, she was entirely excluded from the editorial handling and peer-review process for this submission. The remaining authors report no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41392-026-02939-3.