Discovery of human antibody specific to the spike glycoprotein subunit 2 of SARS-CoV-2 and characterization of its broad-spectrum neutralizing activity Authors: Ying Chi, Huimin Yin, Chao Huang, Yiyue Ge, Jingxian Liu, Yuhan Ding, Fengjuan Shi, Zhou Cui, Shuo Ning, Yongjun Jiao, Lunbiao Cui Published: 2026 ABSTRACT Highlights•In this study, a specific neutralizing antibody targeting the S2 domain wasscreened out, which can serve as a broad-spectrum therapeutic agent against the high-frequency mutations of SARS-CoV-2.•This research provides potential targets for the development of diagnostic reagents and neutralizing antibody drugs against the frequent mutations of COVID-19.•Furthermore, this study offers strategies for the future development of vaccines against other coronaviruses (CoVs). Introduction The outbreak and prevalence of coronavirus disease 2019 (COVID-19) have exerted a tremendous impact on global public health, social stability, and economic development. Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) continues to evolve and mutate during transmission, and multiple variants of concern (VOCs) have emerged, including Alpha (B.1.1.7), Beta (B.1.351), Gamma (P.1), Delta (B.1.617.2), and Omicron (B.1.1.529).1 Viral mutations pose a challenge to the protective efficacy of both natural infection-induced immunity and vaccine-induced immunity. The resulting immune evasion has led to the significant failure of early-generation vaccines and therapeutic antibodies, contributing to multiple large-scale breakthrough infections worldwide.2, 3, 4, 5 SARS-CoV-2 belongs to the genus Beta of coronaviruses. Its genome is composed of approximately 30,000 nucleotides, which mainly encode structural proteins such as spike glycoprotein (S protein), membrane protein, and nucleoprotein. As the main surface antigen component of SARS-CoV-2, the S protein plays a crucial role in tissue tropism, cell fusion, and viral virulence.6 It mediates the entire process of viral attachment and fusion entry into host cells, can stimulate the production of neutralizing antibodies, and is essential for both viral pathogenesis and host antiviral immune response.7 Studies have found that the S protein has a trimeric structure and consists of two domains, spike glycoprotein subunit 1(S1, globular) and spike glycoprotein subunit 2 (S2, rod-shaped), which are bound together through non-covalent interactions.7 The S1 domain can bind to the angiotensin-converting enzyme 2 receptor on the surface of host cells. The S2 domain, on the other hand, anchors the entire S protein to the cell membrane and mediates the fusion of the viral envelope with the host cell membrane. This fusion process enables the delivery of the viral genome into the host cell, thereby endowing the virus with infectivity.7 Unlike the S1 domain, which is prone to mutations, the S2 domain, a typical class I viral membrane fusion protein, is more conserved in both sequence and structure.8 This higher conservation makes S2 a more suitable target for screening broad-spectrum antibodies. Consequently, the S2 domain is widely recognized as a key target for developing protective vaccines and broad-spectrum neutralizing antibodies against SARS-CoV-2.9, 10, 11, 12 At present, the prevention and treatment of severe COVID-19 cases caused by SARS-CoV-2 mainly rely on antiviral therapy and other symptomatic treatments, and there is still a lack of effective and powerful strategies. The clinical prevention and treatment strategies mediated by antibodies have shown good effects. Antibody-based preparations can not only provide emergency passive immunity to high-risk susceptible populations in affected areas, enabling them to develop protective antibodies, but also offer therapeutic benefits for clinical patients. Additionally, using neutralizing antibodies as guide molecules to identify their corresponding antigenic epitopes plays a critical role in the development of broad-spectrum recombinant vaccines.13 Based on the mutability of SARS-CoV-2, this study aims to generate fully human antibody targeting the S2 domain of SARS-CoV-2. Antibodies developed against this conserved S2 domain can inhibit different SARS-CoV-2 variants fusion with host cells, which possess broad-spectrum neutralizing activity. This research will lay the foundation for further development of broad-spectrum epitope-based vaccines against SARS-CoV-2, while also contributing to the optimization and improvement of prevention and treatment strategies targeting SARS-CoV-2. Materials and methods Samples, cells and virus strains Fifteen peripheral blood samples from convalescent COVID-19 patients were provided by the Jiangsu Provincial Center for Disease Control and Prevention. Vero-E6 cell line was purchased from American Type Culture Collection. SARS-CoV-2 wild type (WT) strain, and Omicron variants (XBB.1.19 and JN.1 strains) were obtained from the laboratory’s repository. Primer design Based on the primer sequences provided in Phage Display, 16 pairs of primers for the human-derived antibody κ light chain variable region, 27 pairs for the λ light chain variable region, 6 pairs for the immunoglobulin G (IgG) heavy chain variable region (VH), and 1 pair for overlap-PCR were synthesized.14,15 The complete list of all primers used in this study is provided in Supplementary Table S1. Isolation of peripheral blood mononuclear cells (PBMCs) Peripheral blood collected from patients in the convalescent phase of COVID-19 was mixed with phosphate-buffered saline at an equal volume. The mixture was slowly layered onto the lymphocyte separation medium (Solarbio, Beijing, China). After centrifugation, the intermediate buffy coat layer was aspirated, which contained the PBMCs. The cells were then washed for subsequent nucleic acid extraction. Construction of phage scFv library Following the instructions of the total RNA extraction kit (Qiagen, Hilden, Germany) and reverse transcription kit (Invitrogen, Carlsbad, USA), RNA was extracted from PBMCs and reverse-transcribed into cDNA. Using the obtained cDNA as a template, PCR amplification (Invitrogen) was performed to synthesize human-derived antibody VH and VL chains. For the overlap-PCR method, the recovered and purified VH and VL chain products were mixed at an equimolar ratio as templates, and overlap-PCR primers were used to splice the scFv gene. The pComb3XSS (Agilent, Santa Clara, USA) phagemid vector and the purified scFv gene were digested with Sfi Ⅰ restriction enzyme (Roche, Basel, Switzerland). The digested products were ligated using T4 ligase (Roche), and the ligation product was electro-transformed into E. coli XL1-Blue (Agilent). The transformation efficiency was calculated by serial dilution. Screening of single-chain antibodies specific to the SARS-CoV-2 S2 protein The prepared library bacterial solution was inoculated into 2 yeast extract-tryptone medium (containing ampicillin and tetracycline) to amplify the phage antibody library. Helper phage VCSM13 (Agilent) and kanamycin (Sigma-Aldrich, St. Louis, USA) were added. After cultivation, a polyethylene glycol/NaCl solution was added to precipitate the phage antibodies. The precipitate was collected by centrifugation to obtain purified phage antibodies. The SARS-CoV-2 S2 protein (Youai Biotechnology, Nanjing, China) was coated onto Nunc PolySorb immunotubes (Qiagen). The amplified phage library was added to the Nunc PolySorb immunotubes, and after incubation for 1–2 h, a 0.1 M glycine-hydrochloric acid solution was added for elution, followed by the addition of 1 M Tris solution for neutralization. The neutralized eluate was added to E. coli XL1-Blue in the logarithmic growth phase for cultivation. The phages before and after each round of screening were collected and subjected to gradient dilution to determine the titer. The purified phage antibodies were subjected to 4 rounds of screening following the above steps. The overall success of the affinity selection was quantified by calculating the enrichment fold, defined as the ratio of the output/input phage titer from the final round (Round 4) to that of the first round (Round 1) [Formula (1): (Output₄/Input₄) / (Output₁/Input₁)]. Phage-enzyme-linked immunosorbent assay (ELISA) for specific single-chain antibodies Single colonies were randomly picked from the plate and inoculated into 2YT medium (containing 100.0 µg/mL ampicillin and 12.5 µg/mL tetracycline) for expansion culture. After 12 h, helper phage VCSM13 and kanamycin at a final concentration of 50 µg/mL were added, followed by shaking culture overnight to collect the supernatant. An ELISA plate was coated with the SARS-CoV-2 S2 protein at a concentration of 0.1 µg per well. The supernatant was subjected to phage-ELISA identification using a horseradish peroxidase-labeled anti-M13 antibody diluted at 1∶5000, and the optical density at 450 nm value was measured. A sample was determined to be positive when the positive/negative ratio ≥ 2.1. The bacterial solution of positive clones was selected for sequencing. Antibody sequence analysis and human-derived IgG antibody expression DNASTAR software V11.1 (DNAStar Inc., Madison, USA) was used to analyze the antibody sequences obtained from sequencing. For complete and correct sequences, the gene sequences of antibody variable regions were aligned against the International ImMunoGeneTics information system (IMGT) database, and their complementarity-determining region (CDR) fragments were identified. The scFv antibody sequences obtained from the analysis were used for the expression of full-length human-derived IgG antibodies (Biointron, Shanghai, China). In addition, the complete nucleotide and deduced amino acid sequences of the heavy chain variable region (VH) and light chain variable region (VL) of the S2 antibody have been deposited in the GenBank database under accession numbers PZ381121 and PZ381122, respectively. Detection of binding affinity of anti-SARS-CoV-2 S2 protein-specific full-length IgG antibodies Using the ELISA method, an ELISA plate was coated with the SARS-CoV-2 S2 protein at a concentration of 0.1 µg per well. The synthesized human-derived anti-SARS-CoV-2 S2-specific full-length IgG antibodies were diluted to different concentrations. A horseradish peroxidase-labeled anti-human IgG antibody was used as the secondary antibody for detection, and the binding activity of the antibodies was evaluated by measuring the optical density at 450 nm value. Binding kinetics measurements were performed using biolayer interferometry on a ForteBio Red Octet 96 instrument (Sartorius, Inc., Göttingen, Germany). Briefly, individual human antibodies were diluted to 10 µg/mL, loaded onto Protein A biosensors (Sartorius, Gottingen, Germany) for 30 s, and washed to remove any unbound material before conducting measurements in phosphate-buffered saline with tween-20 (PBST). The immobilized antibodies were incubated with varying concentrations of S2 to capture the kinetic data. All binding data were collected at 30 °C. A total of 4–6 concentrations of antigens were used. The baseline and dissociation steps were carried out in a buffer only. The kinetic data were fit to a simple 1∶1 binding model to determine the dissociation constant (KD) using the association and dissociation rates. Binding assays were performed in triplicate, and average KD values were calculated. Detection of neutralizing activity of anti-SARS-CoV-2 S2 protein-specific full-length IgG antibodies S2-Specific full-length immunoglobulin G (IgG-S2) antibodies were serially diluted in a 4-fold dilution series starting at an initial concentration of 200 µg/mL, and eight concentration gradients were prepared in total. In the biosafety level 3 (BSL-3) laboratory, three SARS-CoV-2 strains (WT, XBB.1.19 and JN.1) each at 100 50% tissue culture infective dose, were mixed with the correspondingly diluted antibodies at an equal volume, followed by incubation at 37 °C for 1.5 h. A volume of 100 µL of antigen-antibody complex was dispensed into Vero-E6 cell culture 96-well plates, with four replicate wells assigned to each concentration. An irrelevant IgG isotype antibody, a virus-only control without antibody treatment and a cell-only blank control were established simultaneously. All experimental procedures were performed in three independent replicates. The cells were cultured in an incubator at 37 °C with 5%CO₂ for 48 h. Cell culture supernatants from each well were collected, inactivated at 90 °C, and then used for nucleic acid extraction (BioPerfectus, Taizhou, China). The PCR reaction system was prepared according to the instructions of 2019-nCoV Nucleic Acid Detection Kit (BioPerfectus). The viral RNA load of the ORF gene of SARS-CoV-2 was determined to evaluate the viral replication level. Values of 2−ΔCt were calculated according to the Ct value measured from the PCR instrument, to represent the relative virus copies of the antibody-treated group relative to the virus control group. The virus replication inhibition rate (%) was calculated as Formula (2): (1–2−ΔCt) × 100%. The half maximal inhibitory concentration (IC₅₀) values were plotted according to viral RNA copies and the antibody concentrations using GraphPadPrism5 (GraphPad Software, San Diego, USA). Prediction of antibody fragment antigen-binding (Fab) region structures The three-dimensional structures of antibody Fab regions were predicted using tFold (https://github.com/TencentAI4S/tFold), AlphaFold3 (https://alphafoldserver.com/), and HelixFold3 (https://paddlehelix.baidu.com/app/all/helixfold3/forecas), respectively. The amino acid sequences corresponding to the variable domains of the heavy chain and light chain were used as inputs for structure prediction. For each antibody, the predicted Fab structures generated by different algorithms were subjected to structural alignment and comparison using PyMOL 3.1.1 (Schrodinger, Inc., New York, USA) . Structural superposition and visualization were performed to evaluate overall conformational consistency, with particular attention paid to the complementarity-determining regions (CDRs). Results Construction of a human-derived SARS-CoV-2 scFv library cDNA extracted from PBMCs of 15 COVID-19 convalescent patients was used as the template. PCR amplification with 6 pairs of specific primers targeting the variable region of IgG heavy chain yielded the target band with a molecular weight of approximately 400 bp (Fig. 1, 1–4). PCR amplification with 16 pairs of specific primers targeting the variable region of light chain resulted in the target band of approximately 350 bp (Fig. 1, 5–11). Using VH and VL genes as templates, overlap extension PCR was performed with one pair of overlapping primers to splice scFv genes, resulting in target bands of approximately 750 bp (Fig. 2, 1–7). The scFv gene and pcomb3xss plasmid were digested with restriction enzymes and then ligated. The ligation product was transformed into XL1-Blue competent cells by electroporation. The calculated library capacity of the constructed library was 1.6 × 10⁷ CFU, indicating that a human-derived SARS-CoV-2 scFv library with a capacity of 1.6 × 10⁷ CFU was obtained through a single electroporation.Fig. 1Electrophoretograms of PCR-amplified VL and VH genes.Fig 1 dummy alt textNotes: M: DNA marker, 1–4: PCR amplification product of heavy chain, 5–11: PCR amplification product of light chain.Abbreviations: PCR, polymerase chain reaction; VH, heavy chain variable region; VL, light chain variable region.Fig. 2Electrophoretogram of PCR-amplified scFv genes.Fig 2 dummy alt textNotes: M: DNA marker; 1–7: overlap-PCR Amplification Product of scFv Genes.Abbreviation: PCR, polymerase chain reaction; scFv, single-chain variable fragment. Electrophoretograms of PCR-amplified VL and VH genes. Electrophoretogram of PCR-amplified scFv genes. Screening of single-chain antibodies specific to the SARS-CoV-2 S2 protein The SARS-CoV-2 S2 protein was coated onto immunotubes, and four rounds of affinity screening were performed. The overall success of the affinity selection was quantified by calculating the enrichment fold, defined as the ratio of the output/input phage titer from the final round (Round 4) to that of the first round (Round 1). After each round of screening, the input and output titers of the phage antibodies were measured, with the results shown in Table 1. This calculation yielded an 18.5-fold enrichment, indicating a successful selection process. During the screening of single-chain antibodies specific to the SARS-CoV-2 S2 protein, there was a transient decrease in the output/input ratio during Round 3, which can occur in phage display panning due to factors such as increased washing stringency in later rounds to select for higher-affinity clones, and intense competition among the enriched phage population for a fixed amount of target antigen. Nonetheless, the recovery rate increased markedly in the final round, and the cumulative enrichment confirmed the effective isolation of specific binders.Table 1Enrichment effect of phage antibody titers.Table 1 dummy alt textScreening roundTiter of anti-S2 protein phage antibodiesInputOutputInput/Output11.8 × 10111.1 × 1066.11 × 10–622.8 × 10109.0 × 1063.21 × 10–437.0 × 10101.7 × 1062.43 × 10–545.5 × 1096.2 × 1051.13 × 10–4Abbreviations: S2, spike glycoprotein subunit 2. Table 1Enrichment effect of phage antibody titers.Table 1 dummy alt textScreening roundTiter of anti-S2 protein phage antibodiesInputOutputInput/Output11.8 × 10111.1 × 1066.11 × 10–622.8 × 10109.0 × 1063.21 × 10–437.0 × 10101.7 × 1062.43 × 10–545.5 × 1096.2 × 1051.13 × 10–4Abbreviations: S2, spike glycoprotein subunit 2. Enrichment effect of phage antibody titers. Abbreviations: S2, spike glycoprotein subunit 2. Analysis of the characteristics of heavy and light chains and CDRs of specific single-chain antibody The SARS-CoV-2 S2 protein was coated onto ELISA plates, and phage-ELISA was performed using phage antibody supernatant as the primary antibody and anti-M13 antibody as the secondary antibody. Positive clones were analyzed by sequencing, and an S2 protein-specific scFv antibody (designated S2) was ultimately identified. Analysis of the antibody sequence via the IMGT database revealed that the heavy chain of the S2 antibody belongs to the VH3-30 family, while the light chain originates from the VL3-19 germline gene of the Lambda chain (Table 2). The amino acid sequences of the CDRs in the heavy and light chains of the antibody are presented in Table 3.Table 2Characteristics of the heavy and light chains of the SARS-CoV-2-S2 scFv antibody.Table 2 dummy alt textAntibodyHeavy chainsLight chainsH-Allele-VH-Allele-DH-Allele-JL-Allele-VL-Allele-JS2IGHV3-30×03IGHD5-12×01IGHJ3×02IGLV3-19×01IGLJ2×01Abbreviations: S2, spike glycoprotein subunit 2; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; scFv, single-chain variable fragment.Table 3Amino acid sequence alignment of CDRs in the heavy and light chains of the SARS-CoV-2-S2 scFv antibody.Table 3 dummy alt textAntibodyHeavy chainsLight chainsCDR1CDR2CDR3CDR1CDR2CDR3S2GFSFSSYDISSDGGFKAREGRYGGNDSGAFDTILRSYYGKNNSRDSSGNHVVAbbreviations: CDR, complementarity-determining region; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; scFv, single-chain variable fragment. Table 2Characteristics of the heavy and light chains of the SARS-CoV-2-S2 scFv antibody.Table 2 dummy alt textAntibodyHeavy chainsLight chainsH-Allele-VH-Allele-DH-Allele-JL-Allele-VL-Allele-JS2IGHV3-30×03IGHD5-12×01IGHJ3×02IGLV3-19×01IGLJ2×01Abbreviations: S2, spike glycoprotein subunit 2; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; scFv, single-chain variable fragment. Characteristics of the heavy and light chains of the SARS-CoV-2-S2 scFv antibody. Abbreviations: S2, spike glycoprotein subunit 2; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; scFv, single-chain variable fragment. Table 3Amino acid sequence alignment of CDRs in the heavy and light chains of the SARS-CoV-2-S2 scFv antibody.Table 3 dummy alt textAntibodyHeavy chainsLight chainsCDR1CDR2CDR3CDR1CDR2CDR3S2GFSFSSYDISSDGGFKAREGRYGGNDSGAFDTILRSYYGKNNSRDSSGNHVVAbbreviations: CDR, complementarity-determining region; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; scFv, single-chain variable fragment. Amino acid sequence alignment of CDRs in the heavy and light chains of the SARS-CoV-2-S2 scFv antibody. Abbreviations: CDR, complementarity-determining region; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; scFv, single-chain variable fragment. Analysis of antigen binding and affinity of SARS-CoV-2 S2 protein-specific IgG antibody To further confirm the binding specificity and affinity of the humanized IgG-S2 antibody to the SARS-CoV-2 S2 protein, indirect ELISA and biolayer interferometry assays were performed. Indirect ELISA analysis showing the reactivity of serum antibodies against the S2 antigen across serial dilutions, as measured by absorbance at 450 nm (A₄₅₀) (Fig. 3A). The kinetic curve of the IgG-S2 antibody was recorded in Fig. 3B. The dissociation constant (KD) value of the IgG-S2 antibody was 4.780 × 10–12 M, the association constant (KA) was 3.005 × 104/Ms and the equilibrium dissociation constant was 1.437 × 10–7/s as shown in Table 4. From these results we found that the IgG-S2 antibody exhibited tight binding affinity for the S2 protein.Fig. 3Antigen-binding and affinity characteristics of S2 protein-specific IgG antibody. (A) ELISA detection of binding activity of anti-S2 antibody under gradient dilution (A₄₅₀). (B) Binding kinetics measurements were performed using Bio-Layer Interferometry on ForteBio Red Octet 96 instrument. The KD value of IgG-S2 antibody was 4.780 × 10–12 M.Fig 3 dummy alt textAbbreviations: IgG-S2, S2-specific full-length immunoglobulin G; KD, dissociation constant; S2, spike glycoprotein subunit 2.Table 4Kinetic parameters of IgG-S2 antibody determined by BLI assay.Table 4 dummy alt textSample IDLoading Sample IDKD (M)Ka (1/Ms)Kdis (1/s)Full R^2AbmAbS24.780 × 10–123.005 × 1041.437 × 10–70.9957Abbreviations: BLI, biolayer interferometry; IgG-S2, S2-Specific full-length immunoglobulin G; Ka, association constant; KD, dissociation constant; Kdis, dissociation constant. Antigen-binding and affinity characteristics of S2 protein-specific IgG antibody. (A) ELISA detection of binding activity of anti-S2 antibody under gradient dilution (A₄₅₀). (B) Binding kinetics measurements were performed using Bio-Layer Interferometry on ForteBio Red Octet 96 instrument. The KD value of IgG-S2 antibody was 4.780 × 10–12 M. Table 4Kinetic parameters of IgG-S2 antibody determined by BLI assay.Table 4 dummy alt textSample IDLoading Sample IDKD (M)Ka (1/Ms)Kdis (1/s)Full R^2AbmAbS24.780 × 10–123.005 × 1041.437 × 10–70.9957Abbreviations: BLI, biolayer interferometry; IgG-S2, S2-Specific full-length immunoglobulin G; Ka, association constant; KD, dissociation constant; Kdis, dissociation constant. Kinetic parameters of IgG-S2 antibody determined by BLI assay. Abbreviations: BLI, biolayer interferometry; IgG-S2, S2-Specific full-length immunoglobulin G; Ka, association constant; KD, dissociation constant; Kdis, dissociation constant. Evaluation of the neutralizing activity of SARS-CoV-2 S2 specific antibody Further in vitro neutralization assays were conducted to evaluate the efficacy of the IgG-S2 antibody against three SARS-CoV-2 strains: WT, XBB, and JN.1. Based on viral inhibition rates detected by quantitative PCR 48 h post-infection, all dose-response curves of the antibody were recorded and are depicted in Fig. 4. Our results showed that the antibody exhibited broadly neutralizing activities against the WT, XBB and JN.1 strains, but the neutralizing activities differed among the three SARS-CoV-2 strains. The half maximal inhibitory concentration (IC50) of IgG-S2 antibody against the WT, XBB, and JN.1 strains were 4.93 μg/mL, 0.97 μg/mL, and 3.10 μg/mL respectively.Fig. 4Detection of the neutralizing ability of IgG-S2 antibody against SARS-CoV-2 virus by RT-PCR. (A–C) Dose-response curves of the antibody were detected by quantitative PCR. The half maximal inhibitory concentration (IC50) of IgG-S2 antibody against the WT, XBB and JN.1 strains were 4.93 μg/mL, 0.97 μg/mL, and 3.10 μg/mL respectively.Fig 4 dummy alt textAbbreviations: IC50, half maximal inhibitory concentration; IgG-S2, S2-specific full-length IgG; PCR, polymerase chain reaction; RT-PCR, reverse transcription polymerase chain reaction; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; WT, wild type. Detection of the neutralizing ability of IgG-S2 antibody against SARS-CoV-2 virus by RT-PCR. (A–C) Dose-response curves of the antibody were detected by quantitative PCR. The half maximal inhibitory concentration (IC50) of IgG-S2 antibody against the WT, XBB and JN.1 strains were 4.93 μg/mL, 0.97 μg/mL, and 3.10 μg/mL respectively. Prediction of the structural characteristics and paratope physicochemical analysis of the SARS-CoV-2 S2 protein-specific IgG antibody Fab regions To further investigate the structural characteristics of the predicted antibody Fab regions, we generated Fab models using tFold, AlphaFold3, and HelixFold3 (Fig. 5A). These predicted structures provided detailed spatial information on the framework regions and CDR loops, facilitating comparative analysis of model consistency across different prediction platforms. To quantify the reliability of our structural modeling, we performed pairwise structural alignments among models generated by tFold, AlphaFold3, and HelixFold3. The global root mean square deviation (RMSD) values ranged from 0.184 to 0.384 Å, indicating high conformational consensus in the framework regions. Although higher fluctuations were observed in the CDR-H3 loop (RMSD: 0.293–1.100 Å), the core paratope geometry remained consistent across different algorithms (Table 5). Comparison with the deposited antibody crystal structure Protein Data Bank 8yzk, a co-crystallized anti-S2 antibody sharing 80.89% sequence identity with our antibody variable region, further confirmed the structural plausibility of our models. Superimposition of the three independently generated Fv models (tFold, AlphaFold3, and HelixFold3) revealed a conserved framework architecture with divergence primarily confined to the CDR-H3 (Fig. 5A), consistent with the RMSD analysis. Surface electrostatic analysis of the predicted Fv model revealed a polarized charge distribution: a distinct positively charged patch at the CDR-H3 and CDR-L2 interface, and a contrasting negatively charged region localized near CDR-H2 (Fig. 5B).Fig. 5Physicochemical characterization of the predicted anti-S2 antibody paratope. (A) Structural superimposition of the Fv models predicted by tFold (blue), AlphaFold3 (orange), and HelixFold3 (pink). The framework regions show high structural convergence, while the complementarity-determining regions (CDRs), particularly CDR-H3 (indicated by arrows), exhibit localized conformational variations. (B) Solvent-accessible surface area of the predicted Fv model colored by electrostatic potential (red, negative; blue, positive; scale: −5 to + 5 kbT/e). A prominent negatively charged patch is observed across the paratope area formed by the CDR loops. (C) Structural and electrostatic landscape of the SARS-CoV-2 Spike protein fragment. Left: Ribbon representation of the Spike protein fragment (residues 247–816), with color-coded domains: N-terminal domain part (green), Receptor-binding domain (RBD, cyan), Receptor-binding motif (RBM, red), and S1/S2 cleavage region (yellow). Right: Electrostatic surface potential map of the corresponding fragment, highlighting the heterogeneous distribution of charge clusters. Magnified view of the RBM and adjacent RBD surface, identifying key residues available for antibody paratope engagement. Negatively charged residues (Glu 484 and Asp 420, red) and positively charged residues (Arg 403, Lys 417, and Arg 454, blue) form a complementary topographic map.Fig 5 dummy alt textAbbreviations: CDR, complementarity-determining region; RBD, receptor binding domain; RBM, receptor binding motif; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; VH, heavy chain variable region; VL, light chain variable region.Table 5Quantitative validation and structural comparison of the anti-S2 antibody variable region (Fv) models.Table 5 dummy alt textComparison pairGlobal RMSD (Å)CDR-H3 RMSD (Å)Key insightstFold vs. AF30.3841.100High framework consistencyAF3 vs. HelixFold30.1840.293Strong consensus in CDRsAF3 vs. PDB [8yzk]0.8801.747High homology to IgG-S2 AntibodyAbbreviations: AF3, AlphaFold3; CDR, complementarity-determining region; IgG-S2, S2-specific full-length immunoglobulin G; PDB, Protein Data Bank; RMSD, root mean square deviation. Physicochemical characterization of the predicted anti-S2 antibody paratope. (A) Structural superimposition of the Fv models predicted by tFold (blue), AlphaFold3 (orange), and HelixFold3 (pink). The framework regions show high structural convergence, while the complementarity-determining regions (CDRs), particularly CDR-H3 (indicated by arrows), exhibit localized conformational variations. (B) Solvent-accessible surface area of the predicted Fv model colored by electrostatic potential (red, negative; blue, positive; scale: −5 to + 5 kbT/e). A prominent negatively charged patch is observed across the paratope area formed by the CDR loops. (C) Structural and electrostatic landscape of the SARS-CoV-2 Spike protein fragment. Left: Ribbon representation of the Spike protein fragment (residues 247–816), with color-coded domains: N-terminal domain part (green), Receptor-binding domain (RBD, cyan), Receptor-binding motif (RBM, red), and S1/S2 cleavage region (yellow). Right: Electrostatic surface potential map of the corresponding fragment, highlighting the heterogeneous distribution of charge clusters. Magnified view of the RBM and adjacent RBD surface, identifying key residues available for antibody paratope engagement. Negatively charged residues (Glu 484 and Asp 420, red) and positively charged residues (Arg 403, Lys 417, and Arg 454, blue) form a complementary topographic map. Table 5Quantitative validation and structural comparison of the anti-S2 antibody variable region (Fv) models.Table 5 dummy alt textComparison pairGlobal RMSD (Å)CDR-H3 RMSD (Å)Key insightstFold vs. AF30.3841.100High framework consistencyAF3 vs. HelixFold30.1840.293Strong consensus in CDRsAF3 vs. PDB [8yzk]0.8801.747High homology to IgG-S2 AntibodyAbbreviations: AF3, AlphaFold3; CDR, complementarity-determining region; IgG-S2, S2-specific full-length immunoglobulin G; PDB, Protein Data Bank; RMSD, root mean square deviation. Quantitative validation and structural comparison of the anti-S2 antibody variable region (Fv) models. Abbreviations: AF3, AlphaFold3; CDR, complementarity-determining region; IgG-S2, S2-specific full-length immunoglobulin G; PDB, Protein Data Bank; RMSD, root mean square deviation. This dipolar surface architecture suggests a highly coordinated mode of electrostatic complementarity targeting the receptor binding domain (RBD) (Fig. 5C). Specifically, the positively charged CDR-H3/L2 patch is ideally suited to engage negatively charged residues within the receptor binding motif (RBM), such as Glu484 or Asp420. Concurrently, the negatively charged CDR-H2 region is spatially positioned to interact with adjacent cationic clusters on the RBD surface, including Lys417, Arg403, or Arg454. This dipolar locking’ mechanism, characterized by the simultaneous alignment of opposite charges at two distinct paratope-epitope interfaces, likely provides both high binding affinity and precise structural orientation during the docking process. The observed electrostatic steering suggests that the antibody-antigen interaction is governed by synergistic ionic attractions, which may reduce the entropic penalty of binding. These findings further support the functional relevance of our predicted model, demonstrating how the antibody’s charge landscape compensates for the intrinsic conformational flexibility of the CDR-H3 loop. Discussion Since its emergence at the end of 2019, COVID-19 has caused hundreds of millions of infections and millions of deaths worldwide.16 Currently, neutralizing antibodies have been widely used in the treatment of COVID-19 infections.17, 18, 19, 20 The majority of therapeutic neutralizing antibodies for COVID-19 target the S1 domain of the SARS-CoV-2 spike protein, particularly the RBD, to prevent the virus from binding to host receptors.21 However, numerous VOCs have frequently acquired immune escape mutations in the RBD, leading to the reduced efficacy of many previously authorized vaccines and therapeutic antibodies.22, 23, 24 Compared with the S1 domain of the SARS-CoV-2 spike protein, the S2 domain has a highly conserved sequence. It can induce the body to produce neutralizing antibodies with broad-spectrum neutralizing efficacy against different variants.9,10,25, 26, 27 These antibodies exert antiviral effects by inhibiting the fusion of the viral envelope with host cells, thereby depriving the virus of its ability to infect human cells.27 Therefore, in this study, a specific IgG antibody (IgG-S2) was screened and prepared via phage display technology, targeting the highly conserved S2 region of SARS-CoV-2. The capacity of a phage library can directly affect the screening efficiency of the specific antibodies. The scFv antibody library constructed in this study has a relatively low capacity, which may limit the diversity of high-affinity candidates available for selection, potentially resulting in suboptimal neutralization potency. In future studies, the capacity of the phage library can be improved by selecting competent cells with high transformation efficiency, optimizing the input amount of helper phage, and adopting in vivo recombination technology. In vitro neutralization experiments with live viruses confirmed that IgG-S2 exhibited significant inhibitory activity against the SARS-CoV-2 WT strain and Omicron subtypes (XBB, JN.1). In particular, it maintained high neutralizing efficacy against the XBB and JN.1 variants, which were once thought to potentially reduce antibody effectiveness. This result directly confirms the “broad-spectrum advantage” of S2 protein-targeting antibodies and provides a potential solution for addressing future emerging SARS-CoV-2 variants. To further elucidate the molecular recognition mechanism of the obtained IgG-S2 antibody, this study conducted a systematic analysis of its sequence using the IMGT database. The results showed that its heavy chain belongs to the VH3-30 family, and the light chain is derived from the Lambda chain VL3-19 germline gene, suggesting that the formation of this antibody may be closely related to the evolutionary selection of specific germline genes. In addition, we constructed the Fab region models using tFold, AlphaFold3 and HelixFold3, which provided detailed spatial information on the framework regions and CDR loops of the antibody and laid a theoretical foundation for subsequent analyses of antibody-antigen interactions, epitope mapping and experimental validation. However, our current research lacks wet-lab experimental verification. The predicted binding mode and steric hindrance mechanism in the S2 stalk-helix region remain computationally inferred and still need further experimental confirmation. We also attempted to model the antibody-antigen complex using AlphaFold3 and ZDOCK. However, the resulting interface confidence scores were relatively low, precluding a reliable depiction of the bound complex. This phenomenon is frequently observed in studies of SARS-CoV-2 S2-targeting antibodies, as the S2 subunit—particularly the heptad repeat regions HR1 and HR2—undergoes dramatic conformational transitions during the membrane fusion process.28 The low confidence scores likely reflect the intrinsic disorder of the paratope-epitope interface in the pre-fusion state, or an induced-fit binding mechanism that cannot be adequately captured by single-state static modeling. Collectively, we will carry out alanine scanning mutagenesis, Hydrogen/Deuterium Exchange Mass Spectrometry (HDX-MS) and Cryo-Electron Microscopy (Cryo-EM) structural analysis to precisely define the authentic epitopes and verify the proposed steric hindrance mechanism in the future. In this study, the neutralizing effects of IgG-S2 against different SARS-CoV-2 strains varied. The inhibition rate of IgG-S2 against the XBB strain was significantly higher than that against the WT strain. Although its inhibition rate against the JN.1 strain was lower than that against XBB, it was still higher than that against WT. This phenomenon may be associated with conformational differences in the S2 protein among different strains. Although the S protein of Omicron subtypes (XBB, JN.1) has accumulated a large number of mutations in the S1 region, the key sites for fusion function in the S2 region have not undergone destructive changes.29 Furthermore, the increased conformational exposure of the S2 region in these variants might facilitate its binding to IgG-S2, thereby enhancing the antibody’s neutralization efficiency.30 However, this speculation still needs to be further verified through protein structure analysis and antibody binding site identification experiments to clarify the interaction mechanism between IgG-S2 and the S2 protein of different strains. Previous studies have reported several monoclonal antibodies targeting the S2 domain of SARS-CoV-2, such as S2P6, 1249A8, CC40.8, and CV3-25 isolated from COVID-19 convalescent patients. These antibodies exhibit broad-spectrum neutralizing activity against SARS-CoV-2 variants, with some capable of cross-reacting with other coronaviruses due to their highly conserved epitopes.11,12,31, 32, 33 Among them, 1249A8 and its derivative antibodies, as well as CC40.8, possess the widest neutralizing spectrum, covering SARS-CoV-2 Omicron variants and middle east respiratory syndrome coronavirus (MERS-CoV). In contrast, S2P6 shows no efficacy against Omicron, and CV3-25 does not neutralize MERS-CoV. These differences are attributed to variations in epitope sequences and binding modes. For instance, while both our antibody and S2P6 target the stem-helix region, they arise from different germline genes (IGHV3-30/IGLV3-19 for S2 vs. IGHV1-69/IGKV3-20 for S2P6) and possess unique CDR sequence lengths and compositions. This suggests they are products of convergent evolution, recognizing the same region of vulnerability through distinct molecular solutions.11 Both S2P6 and 1249A8 are associated with previous seasonal coronavirus infections. Consistent with previous reports, the S2 antibody identified in this study exhibits broad-spectrum neutralizing activity against both SARS-CoV-2 WT and Omicron strains. However, its neutralizing titer is relatively lower compared to antibodies targeting S1 or the RBD.11,28,31,34 We speculate that the S2 stalk-helix region resides at the base of the S-protein trimer, inducing substantial steric hindrance and thereby reducing antibody binding efficiency relative to the surface-exposed RBD. Potential follow-up experiments, including affinity measurement, binding analysis under different pH conditions and conformational locking/unlocking assays are needed to validate this hypothesis in our future studies. Therefore, further in-depth investigations are warranted to address two key questions: how to enhance the potency of antibodies specific to the S2 stalk-helix region, and whether such antibodies can cross-recognize other SARS-CoV-2 VOCs as well as other coronaviruses including SARS-CoV and MERS-CoV. Additionally, developing combination therapies that pair S2 conserved epitope-targeting antibodies with one or more potent neutralizing antibodies against other SARS-CoV-2 epitopes could improve the breadth and stability of antiviral efficacy. This approach is critical for SARS-CoV-2 prevention and represents a key future research direction for our team. Furthermore, in vivo pharmacokinetic assessment and animal challenge assays are warranted in our future work to verify its in vivo protective efficacy and translational potential. Here, we screened a broad-spectrum neutralizing antibody targeting the conserved S2 domain, which exhibits potent neutralizing efficacy against multiple SARS-CoV-2 variants. However, the capacity of the library constructed in this study is relatively limited, which may restrict the diversity of available high-affinity candidate antibodies and lead to suboptimal neutralizing activity. The present study lacks wet-lab experimental validation of the predicted binding mode and steric hindrance mechanism within the S2 stalk-helix region, as well as corresponding in vivo experiments to further confirm the in vivo neutralizing efficacy of this antibody. Conclusions In conclusion, we identified an S2-targeting specific antibody with potent broad-spectrum neutralizing activity against diverse SARS-CoV-2 variants. This work lays a preliminary molecular basis for the development of diagnostic tools and neutralizing antibody therapeutics to combat continuously evolving COVID-19 variants. CRediT authorship contribution statement Ying Chi: Writing – review & editing, Writing – original draft, Investigation, Formal analysis, Data curation, Conceptualization. Huimin Yin: Investigation, Data curation, Writing – review & editing. Chao Huang: Investigation, Validation, Data curation, Resources. Yiyue Ge: Funding acquisition, Investigation. Jingxian Liu: Investigation. Yuhan Ding: Investigation, Data curation. Fengjuan Shi: Validation. Zhou Cui: Investigation. Shuo Ning: Investigation. Yongjun Jiao: Resources. Lunbiao Cui: Funding acquisition, Project administration, Supervision. Informed consent Peripheral blood lymphocytes were obtained from 15 convalescent COVID-19 patients provided with written informed consent. Organ donation Not applicable. Ethical statement Studies on patients were approved by ethics committee of Jiangsu Provincial Disease Control and Prevention (JSJK2021-B007-01). Animal treatment Not applicable. Generative AI No generative AI or AI-assisted technologies were used in the writing of this manuscript. Funding This study was supported by the 10.13039/501100004608Natural Science Foundation of Jiangsu Province (BK20221413, BK20231374), the Nanjing Important Science & Technology Specific Projects (2021-11005) and the 10.13039/501100012154Scientific Research Project of Jiangsu Commission of Health (H2023037). Declaration of competing interest No potential conflict of interest was reported by the authors.