Rational design of an optimized ferritin nanoparticle vaccine targeting both SARS-CoV-2 and MERS-CoV Authors: Yaoming Liu, Haiyue Rao, Nana Wang, Xiaoqing Liu, Tao Chen, Bin Zhang, Jintao Lai, Jianyu Shan, Shiqi Xiao, Haojie Peng, Yiqiang Zhu, Taizhen Liang, Sen Liu, Meilin Hu, Lixiang Xie, Guochang Qiu, Xiaobo Li, Yaxin Li, Xiancai Ma Published: 2026 ABSTRACT BackgroundCoronaviruses including SARS-CoV-2 and MERS-CoV remain threats to global health. Ferritin nanoparticle-based vaccines are promising platforms for coronaviral multivalent antigen display. However, their development is often constrained by limited stability and homogeneity, which hinders scale-up manufacturing and long-term storage.ResultsHere, we employed artificial intelligence (AI)-guided structural modeling and optimization to introduce disulfide bonds into Helicobacter pylori ferritin (HPF). Cryo-EM at 2.2 Å confirmed the formation of inter-subunit disulfide bonds in the most promising variant HPF (I69C), resulting in a more homogeneous nanoparticle with enhanced thermal and pH stability, as well as improved solubility in physiological conditions. We utilized the ST003/SC003 molecular glue system to covalently conjugate receptor-binding domains (RBDs) of both SARS-CoV-2 and MERS-CoV, either as a mixture of individual RBD-HPF (I69C) particles or as a dimeric RBD displayed on a single HPF (I69C). Both bivalent nanoparticle vaccines elicited significantly higher titers of RBD-specific antibodies and neutralizing antibodies compared to monomeric and dimeric vaccines. Vaccination also increased frequencies of antigen-specific B cells and polyfunctional CD4+ and CD8+ T cells. No vaccine-related systemic abnormalities were observed. In both hACE2 and hDPP4 transgenic mice, two doses of bivalent nanoparticle vaccines provided protection against authentic SARS-CoV-2 and MERS-CoV challenges.ConclusionsOur study demonstrated that rationally engineered HPF (I69C) produced highly stable and efficiently functionalized nanoparticle vaccines capable of eliciting potent humoral and cellular immune responses against both SARS-CoV-2 and MERS-CoV infection, thereby supporting the further development of bivalent nanoparticle vaccine platforms.Graphical abstract Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12951-026-04738-y. Background Since the 20th century, seven types of coronaviruses have crossed into human populations, resulting in millions of infections. Among these, human coronavirus 229E (HCoV-229E), HCoV-OC43, HCoV-NL63, and HCoV-HKU1 typically cause mild respiratory symptoms in adults but exhibit increased pathogenicity in infants and elderly individuals [1]. These coronaviruses account for 10% to 30% of common cold infections. The remaining three coronaviruses, severe acute respiratory syndrome coronavirus (SARS-CoV), Middle East respiratory syndrome coronavirus (MERS-CoV), and SARS-CoV-2, are highly pathogenic and predominantly associated with severe respiratory illness. Between 2002 and 2003, approximately 8,096 cases of SARS-CoV infection were reported, including 774 deaths [2]. Fortunately, no new SARS cases have been reported since 2003. MERS-CoV has caused sporadic infections with limited human-to-human transmission, leading to 2,638 confirmed infection cases and 957 fatalities since its initial identification in 2012 [3]. Unlike SARS-CoV and MERS-CoV, SARS-CoV-2, the virus responsible for coronavirus disease 2019 (COVID-19), remains highly transmissible and prevalent among human populations [4]. Since its outbreak in late 2019, over 778 million cases of SARS-CoV-2 infection have been reported globally, which included more than 7 million deaths (https://data.who.int/dashboards/covid19). Therefore, developing broad-spectrum antiviral drugs or vaccines remains urgently needed to prevent infections caused by prevalent coronaviruses, or potentially, future unknown ‘X’ coronaviruses. Although 60 years have passed since the first coronavirus HCoV-229E was discovered in 1965, no widely used vaccine has been approved for HCoV-229E, HCoV-OC43, HCoV-NL63, HCoV-HKU1, SARS-CoV, or MERS-CoV [5]. Almost all approved coronavirus vaccines target SARS-CoV-2, including inactivated vaccines, mRNA vaccines, subunit vaccines, vectored vaccines, and many others (https://covid19.trackvaccines.org/vaccines/approved). The predominant immunogen used in these vaccines focuses on Spike (S) glycoprotein or receptor-binding domain (RBD) of coronaviruses [6–10]. As SARS-CoV-2 persistently acquires escape mutations, most vaccines based on the ancestral variant are unable to provide protection against current mutants [11]. Unlike SARS-CoV-2, MERS-CoV undergoes low frequency of mutation. Therefore, RBD vaccines based on original MERS-CoV have been shown to elicit neutralizing antibodies against other major strains [12]. As both SARS-CoV-2 and MERS-CoV continue to circulate in human populations, it is increasingly necessary to develop bivalent COVID/MERS vaccines against both highly pathogenic coronaviruses. The Lanying Du group and their colleagues developed two bivalent vaccination strategies by sequential immunization with chimeric S proteins expressing MERS-CoV RBD and SARS-CoV-2 variant RBD, or combined immunization with both SARS-CoV-2 and MERS-CoV RBD subunits. Both strategies induced potent neutralizing antibodies (nAbs) against SARS-CoV-2 and MERS-CoV and provided cross-protection against challenge with both authentic viruses in animal models [13, 14]. Another two groups designed mosaic RBD nanoparticle vaccines by conjugating RBD subunits of SARS-CoV-2 and MERS-CoV to self-assembling protein nanoparticles [15, 16]. These nanoparticle vaccines also elicited robust antibody responses against RBD antigens of SARS-CoV-2 and MERS-CoV and effectively prevented infection by authentic viruses in humanized mouse models, highlighting the advantages of nanoparticle-based platforms in designing multivalent vaccines. With the outbreak of COVID-19, diversified vaccine platforms have emerged and provided protection against SARS-CoV-2 with distinct advantages [17]. Among these, the innovative nanoparticle vaccines are particularly effective in inducing higher titers of nAbs by displaying dominant and spatially accessible epitopes on their surface [5]. Due to their nanometer-scale size, these vaccines can directly drain into lymphoid tissues and are subsequently presented by follicular dendritic cells (FDCs) to B cells in the form of intact nanoparticles. Commonly used nanoparticles in coronavirus vaccine development include Matrix-M (as an adjuvant), ferritin, lumazine synthase (LS), mi3, I53-50, plant-derived viral-like particle (VLP), murine leukemia virus (MLV) Gag-based VLP, and many others [18–24]. These nanoparticle vaccines have been verified in clinical trials or approved for preventing SARS-CoV-2 infection. Unlike traditional subunit vaccines, nanoparticle vaccines present multiple copies of antigens on single particles at high density, thereby more closely mimicking the antigen display pattern of viruses. The ferritin nanoparticle can present 24 copies of antigens on its surface, while LS and mi3 nanoparticles can display 60 copies of antigens [25–27]. The two-component I53-50 nanoparticle is a computationally designed 120-subunit complex composed of 12 pentamers and 20 trimers [28]. Theoretically, it can present 120 copies of antigens. In practical applications, only 20 trimers are conjugated with 60 antigens. Recently, the Neil P. King group and their collaborators developed several I53-50-based nanoparticle vaccines targeting the S, RBD, and N-terminal domain (NTD) of MERS-CoV, respectively [29]. Their research showed that I53-50 nanoparticles displaying S-derived antigens induced potent nAbs against MERS-CoV variants and prevented mouse-adapted authentic MERS-CoV infection in mice. Nanoparticles can also present multivalent antigens on single particles, resulting in the development of mosaic nanoparticle vaccines. Based on their polymeric characteristics, Alexander A. Cohen et al. developed an mi3-based mosaic nanoparticle vaccine against RBDs of 8 different sarbecoviruses [30]. Yanjun Zhang et al. developed a betacoronavirus nanoparticle vaccine by simultaneously conjugating RBDs of SARS-CoV, MERS-CoV, and SARS-CoV-2 on one ferritin nanoparticle [16]. Furthermore, Dan Bi Lee et al. utilized proliferating cell nuclear antigen (PCNA) heterotrimeric scaffolds as the nanoparticle and conjugated 6 RBDs derived from both alphacoronaviruses and betacoronaviruses [15]. These previous studies indicate that the nanoparticle-based antigen presentation strategy could be a promising platform for the development of universal vaccines against multiple viruses and their variants. We previously developed a ferritin nanoparticle vaccine targeting both RBD and heptad repeat (HR) subunits of SARS-CoV-2 Spike [19]. The RBD/HR nanoparticle vaccine elicited more potent humoral and cellular immune responses than monomers, which prevented authentic SARS-CoV-2 challenge in human angiotensin-converting enzyme 2 (hACE2) mice. It also elicited cross-reactive nAbs and cellular immunity against other coronaviruses and mutants [31]. However, during vaccine production, we observed that ferritin nanoparticles exhibited limited stability, demonstrated by visible precipitation after prolonged storage and the presence of additional oligomeric peaks upon separation with size-exclusion chromatography (SEC). Furthermore, the conjugation efficiency of the SpyTag/SpyCatcher (ST/SC) system, which covalently linked ST-tagged antigens to SC-tagged nanoparticles, was suboptimal, resulting in approximately 30% of nanoparticles remaining unconjugated [32]. In this study, we developed an optimized Helicobacter pylori ferritin (HPF) nanoparticle vaccine with enhanced stability and conjugation efficiency. Through disulfide bond (S–S) introduction at the loop regions on the HPF surface, we confirmed that the 69th amino acid isoleucine mutated to cysteine (I69C) led to the formation of intermolecular disulfide bonds between adjacent protomers. This optimization significantly improved the thermostability and homogeneity of nanoparticles. In addition, we employed the SpyTag003/SpyCatcher003 (ST003/SC003) system to covalently link antigens to nanoparticles [33]. The improved version of ST003/SC003 system markedly increased the conjugation efficiency, as demonstrated by complete conjugation of HPF (I69C) with antigens within 5 min. SARS-CoV-2 and MERS-CoV RBD antigen-conjugated HPF (I69C) nanoparticle vaccines elicited robust humoral and cellular immune responses against both viruses. The optimized HPF (I69C) nanoparticle vaccine-induced nAbs potently neutralized both pseudotyped and authentic SARS-CoV-2 and MERS-CoV viruses. Notably, the bivalent nanoparticle vaccines protected hACE2 and human dipeptidyl peptidase 4 (hDPP4) transgenic mice from authentic SARS-CoV-2 and MERS-CoV infection, respectively. Our studies provided a promising bivalent nanoparticle vaccine strategy for the prevention of multi-coronavirus infections. Methods Cell lines and microbes Antigens and ferritin nanoparticles were expressed in HEK293F (Gibco) and E. coli BL21 (DE3) (AlpalifeBio) cells, respectively. HEK293F cells were cultured in 293 F cell expression medium (SMM 293-TII-N) (SinoBiological) in a shaking incubator with 8% CO2 at 140 rpm and 37 °C, while BL21 cells were cultured in Luria-Bertani (LB) broth while shaking with 220 rpm at 37 °C. Upon isopropyl β-D-1-thiogalactopyranoside (IPTG) induction, BL21 cells were cultured at 37 °C while shaking. HEK293T (CRL-3216, ATCC), HEK293T-hACE2, HEK293T-hDPP4, Vero E6 (CRL-1586, ATCC), and Vero 81 (CCL-81, ATCC) cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) (BasalMedia) supplemented with 10% fetal bovine serum (FBS) (ExCell Bio) and 1% penicillin-streptomycin (BasalMedia). Animal vaccination Specific-pathogen-free (SPF) 6-8-week-old C57BL/6 mice (N000013, GemPharmatech Co., Ltd.), hACE2 transgenic mice (T037657, GemPharmatech Co., Ltd.), and hDPP4 transgenic mice (NM-HU-190042, Shanghai Model Organisms Center Inc.) were housed and vaccinated in an SPF facility at the Animal Center of Guangzhou National Laboratory. Mice were randomly assigned to groups by body weight prior to the first immunization. All mice received a prime vaccination at week 0, a boost vaccination at week 4, and were euthanized at week 6. Serum samples were collected every two weeks and at week 6. Vaccine groups were normalized on an equal total RBD mole basis. All vaccines were co-formulated with aluminum hydroxide (alum) adjuvant at a 1:1 volume ratio (50 µL alum plus 50 µL vaccine solution) immediately before each immunization. Four mice were subcutaneously immunized with 10 µg of RBD (Dimer)-ferritin (HPF) (I69C) nanoparticle vaccines. Mice in the nanoparticle mixture group received equal mole amounts of RBD (SARS2)-HPF (I69C) and RBD (MERS)-HPF (I69C) nanoparticle vaccines. The dimer-only group received equal moles of RBD (SARS2)-RBD (MERS) vaccines, while mice in the monomer-only group were immunized with equal moles of RBD (SARS2) and RBD (MERS) vaccines. For the negative control group, mice were immunized with equal moles of SC003-HPF (I69C) nanoparticles, whereas mice in the mock group received equal volumes of alum adjuvant. The animal ethics committees of Guangzhou National Laboratory Animal Center approved the mouse experiments in this study (Approval Number: GZLAB-AUCP-2023-03-A05). Nanoparticle optimization and molecular docking To increase the stability of Helicobacter pylori ferritin (HPF) nanoparticles, we utilized the disulfide bond introduction strategy. The HPF 24-mer structure was predicted using AlphaFold 3, and the model with the highest composite ranking score was selected as the initial structure. This model was refined using Rosetta with the REF2015 energy function, and per-residue energy contributions were calculated using the per_residue_energies tool to identify key regions for subunit interface stability. Of note, four amino acids were removed from HPF, resulting in the aspartic acid residue at position 5 (D5) being renumbered as position 1 (D1) in our analysis and optimization. The asparagine residue at position 15 (N15), which was N19 in original full-length HPF, was mutated to glutamine residue (Q) to abolish potential N-linked glycosylation. Candidate disulfide bond sites were screened based on the following geometric criteria, Cβ-Cβ distance between candidate residue pairs within 4–6 Å to permit disulfide bond formation, side-chain orientation after cysteine substitution compatible with disulfide geometry, and location at the subunit interface without disruption of native secondary structure or key functional regions. Applying these criteria, we identified the V64-H74 loop region, which exhibited both high conformational flexibility and inter-subunit spatial proximity, as the most favorable site for cysteine engineering. Within this region, four residue pairs met the geometric criteria, L66-P72 (Cβ distance, 6.5 Å), T67-A71 (4.5 Å), S68-S70 (4.3 Å), and I69-I69 (5.0 Å). The Cβ distance of L66-P72 (6.5 Å) slightly exceeded the ideal range, which was attributable to the conformational constraint of wild-type Pro72. Upon P72C mutation, the cysteine side chain regained rotational freedom, providing the structural basis for disulfide bond formation. Rosetta per-residue energy analysis further differentiated these candidates. L66 and P72 showed the lowest total energies (-3.079 and − 3.657) with weak steric clashes (fa_rep, 0.971 and 0.839) and strong van der Waals attraction (fa_atr, -6.720 and − 4.868), consistent with a low-strain, flexible peripheral interface locking mode. T67, S68, S70, and A71 displayed near-neutral total energies (-1.183 to -0.062), moderate solvent exposure (fa_sol, 1.488–3.261), and no significant steric repulsion (fa_rep, 0.070–0.422), representing moderate interface stabilization. I69 was distinguished by the strongest interfacial interaction features, including the most favorable fa_atr (-6.283) and fa_elec (-1.985), and the only notable inter-subunit backbone hydrogen bond contribution (hbond_lr_bb = -1.190), placing it at the core of the dimer interface within a highly coupled interaction network. Its relatively elevated fa_rep (4.390) suggested local conformational strain, implying that disulfide bond formation at this site would impose stronger conformational restraint, favoring a rigid core interface locking mode. Through AlphaFold 3 prediction and Rosetta optimization, we ultimately selected four candidate HPF mutants, which were HPF (L66C + P72C), HPF (T67C + A71C), HPF (S68C + S70C), and HPF (I69C). Theoretically, L66C on one protomer of HPF (L66C + P72C) could form a disulfide bond with P72C on the adjacent protomer. T67C on one protomer could form a disulfide bond with A71C on the other protomer. S68C on one protomer could form a disulfide bond with S70C on the other protomer, while I69C on one protomer could form a disulfide bond with I69C on the adjacent protomer. The side chain of D10 on ST003 formed an isopeptide bond with the side chain of K31 on SC003, resulting in the covalent conjugation of ST003-tagged antigens to SC003-tagged nanoparticles. Accordingly, RBD-ST003 was subjected to molecular docking with SC003-HPF (I69C) using AlphaFold 3. The simulated 3D structure of the 24-mer RBD-ST003-conjugated SC003-HPF (I69C) nanoparticle was further predicted by AlphaFold 3 and visualized using ChimeraX. Structures exhibiting the highest confidence scores, as indicated by predicted local distance difference test (pLDDT), predicted template modelling (pTM), and interface predicted template modelling (ipTM) values, were selected for further analysis. Protein expression and purification DNA sequences of 6× histidine (6His)-tagged HPF variants, including HPF (WT), HPF (L66C + P72C), HPF (T67C + A71C), HPF (S68C + S70C), HPF (I69C), SC003-HPF, and SC003-HPF (I69C), were codon-optimized for expression in E. coli BL21 cells and subsequently cloned into the pET28a vector. DNA sequences of 6His-tagged RBD variants, including RBD (SARS2)-ST003, RBD (MERS)-ST003, and dimeric RBD (SARS2)-RBD (MERS)-ST003, were codon-optimized for expression in HEK293F cells and cloned into the pcDNA3.4 vector. Sanger sequencing was performed to confirm the accuracy of these sequences. Constructs expressing HPF variants were transformed into BL21 cells and amplified in LB broth supplemented with kanamycin at 37 °C while shaking. When the optical density (OD) at 600 nm of bacterial solution reached 0.4, about 1 mM of IPTG solution was added to the suspension to induce protein expression. The bacterial solution was cultured for 18 h (h) at 16 °C while shaking, followed by centrifugation to acquire bacterial pellets. These bacteria pellets were lysed by high pressure homogenizer. Clarified lysates were incubated with Co2+-conjugated TALON Superflow beads (Cytiva) to enrich 6His-tagged target proteins, followed by target protein elution with imidazole. Purified proteins were concentrated and buffer-replaced with conventional tris buffer (20 mM Tris, 50 mM NaCl, pH 7.5). Protein concentrations were determined by bicinchoninic acid (BCA) assay, while purities and homogeneities were assessed via SDS-PAGE, native PAGE, and size-exclusion chromatography (SEC). Constructs expressing RBD variants were transfected into HEK293F cells. After 7-day cultivation, supernatants were harvested and incubated with Ni2+-conjugated Sepharose Excel beads (Cytiva). 6His-tagged target proteins were eluted with the imidazole buffer, followed by concentrating and buffer-replacing with conventional tris buffer. Protein concentrations were determined via BCA assay, while SDS-PAGE, native PAGE, and SEC were conducted to determine the purity and homogeneity of these proteins. Protein samples subjected to densitometric analysis included purified HPF (WT) and HPF (I69C) proteins before and after freeze-thaw cycles, pH incubation, and long-term storage, as well as RBD-conjugated nanoparticle vaccines following thermal stress and freeze-thaw treatment. SDS-PAGE and native PAGE gels were imaged and subjected to densitometric analysis using Image J software. For SDS-PAGE, the intensity of the target protein band was quantified relative to the total lane intensity, and the percentages of intact protein and degraded protein were calculated. For native PAGE, the intensity of the 24-mer nanoparticle band was quantified and expressed as a percentage of the total lane signal to assess the proportions of intact nanoparticle assembly and abnormal aggregates. Size-exclusion chromatography (SEC) According to the molecular weight of different proteins, about 500 µL of concentrated protein solutions were subjected to SEC on protein purification system (ACHROM Firin Plus) with different columns. For RBD (SARS2)-ST003 (28 kDa), RBD (MERS)-ST003 (26 kDa), and RBD (SARS2)-RBD (MERS)-ST003 (54 kDa), Superdex 200 Increase column (Cytiva) was used to separate target proteins. RBD (SARS2)-ST003, RBD (MERS)-ST003, and RBD (SARS2)-RBD (MERS)-ST003 were eluted at retention volumes of 14–18 mL, 15–17 mL, and 11–16 mL, respectively. For HPF variants and SC003-HPF (I69C)-conjugated nanoparticles, Superose 6 Increase column (Cytiva) was used to separate target proteins. The major peaks of HPF (L66C + P72C), HPF (T67C + A71C), HPF (S68C + S70C), and HPF (I69C) appeared at retention volumes of 13–20 mL, 13–22 mL, 8–22 mL, and 13–16 mL. SC003-HPF (I69C) was eluted at retention volumes of 10–14 mL, while RBD (SARS2)-ST003-, RBD (MERS)-ST003-, and RBD (Dimer)-ST003-conjugated SC003-HPF (I69C) were eluted at retention volumes of 8–13 mL, 8–13 mL, and 8–13 mL, respectively. All the eluted proteins were concentrated by ultrafiltration device. The concentration was determined by BCA assay, while the purity was further confirmed by SDS-PAGE- and native PAGE-based Coomassie blue staining. Negative staining HPF-related nanoparticles were characterized by negative staining. Briefly, about 4 µL of each sample (0.025 µg/µL) was applied to the glow-discharged copper grids covered by a thin layer of continuous film four times. After each application, excess samples were quickly blotted away with filter paper. Grids were then stained with 2% uranyl acetate three times, followed by removing excess staining solution with filter paper after each staining step. Samples were imaged on Talos L120C microscope (ThermoFisher) operating at an acceleration voltage of 120 kV. Images were captured at a magnification of 73,000× and a defocus value of 1.5 μm. 2D class averages were generated from automatically picked particles using cryoSPARC to assess particle size, morphology and homogeneity. Cryo-electron microscopy (Cryo-EM) sample preparation and data collection Cryo-electron microscopy (Cryo-EM) grids for the HPF (I69C) structure determination were prepared using a Vitrobot Mark IV (ThermoFisher). 3-µL aliquots of purified HPF (I69C) protein at 9 mg/mL were applied to glow discharged holey-carbon gold grids (Quantifoil Au R1.2/1.3), blotted with filter paper (Ted Pella) under 100% humidity at 4 °C, and then plunged into the liquid ethane. Grids screening and data collection were carried out using a Titan Krios G4 transmission electron microscope (ThermoFisher) operating at 300 kV. Movie stacks were acquired using a Falcon 4 direct electron detector in super-resolution mode, controlled by EPU software (ThermoFisher). Data were collected at a nominal magnification of 165,000× with a defocus range of -0.6 to -1.6 μm. Each micrograph was dose-fractionated into 40 frames, with a total accumulated dose of 50 e⁻/Ų. Detailed data collection statistics were provided in Supplementary material, Table S1. Cryo-EM data processing and model building All steps of the 3D reconstruction for HPF (I69C) were performed using cryoSPARC [34]. A total of 1,532 movie stacks were acquired. The drift correction and dose-weighting were performed using patch motion correction. Contrast transfer function (CTF) parameters were estimated via patch CTF estimation. Following manual curation to remove low-quality micrographs, a set of 1,501,674 particles was automatically picked using templates derived from preliminary 2D classification following blob picker. All extracted particles underwent one round of 2D classification, resulting in 405,737 particles that were retained for subsequent ab-initio 3D reconstruction to generate an initial model. One class from the ab-initio analysis, comprising 392,558 particles, was selected and subjected to the homogeneous refinement, yielding a density map at an overall resolution of 2.56 Å. Finally, non-uniform refinement was performed with O symmetry, yielding a final density map at an overall resolution of 2.20 Å, as determined by the gold-standard Fourier shell correlation (FSC) threshold of 0.143. This map was used for subsequent atomic model building and structural analysis. For model building of HPF (I69C), the cryo-EM density map was initially fitted with the crystal structure of HPF (PDB: 3EGM) using UCSF Chimera [35]. The model was subsequently manually adjusted in Coot [36]. All isoleucine residues at position 73 were manually mutated to cysteine, corresponding to the I69C mutations described herein. Final real-space refinement was performed using Phenix to optimize the model against the density map [37]. Figures were prepared using UCSF ChimeraX [38]. Surface electrostatic potential, SASA, MLP, and interface analyses The cryo-EM structure of HPF (I69C) and the crystal structure of HPF (PDB: 3EGM) were preprocessed by removing solvent molecules and heteroatoms while retaining the complete 24-mer assembly. Surface electrostatic potential calculations were performed in UCSF ChimeraX (version 1.x) using the default atomic charges and residue parameters of the Amber20 force field. The average electrostatic potential on the solvent-excluded surface of each of the 24 subunits was extracted, and the fluctuation amplitude was defined as the difference between the maximum and minimum mean subunit values. Solvent-accessible surface area (SASA) and molecular lipophilicity potential (MLP) analyses were also conducted in UCSF ChimeraX. MLP values were calculated based on the Ghose-Crippen atomic lipophilicity parameters. The MLP fluctuation amplitude was defined as the range of mean MLP values across the 24 subunits. Inter-subunit interface energetics were evaluated using Rosetta InterfaceAnalyzer on the adjacent subunit interface, with key output parameters including interface separation free energy (dG_separated), buried interface surface area (dSASA_int), shape complementarity score (sc_value), and interfacial hydrogen bond energy fraction (hbond_E_fraction). Dynamic light scattering (DLS) The hydrodynamic diameter and zeta potential of nanoparticles were analyzed via Zetasizer Ultra (Malvern Panalytical). Briefly, nanoparticle samples were diluted with tris buffer to a final concentration of 0.1 mg/mL, followed by transferring into quartz cuvettes. After One minute (min), samples were evaluated five times at 25 °C and 633 nm, 30 s (sec) each time. Scattered light was detected at specific angles. Fluctuation data were analyzed using autocorrelation functions and diffusion coefficient was extracted to calculate diameters of nanoparticles. Zeta potentials of nanoparticles were measured at a protein concentration of 0.5 mg/mL and 25 °C. The electric field was applied to the sample to cause particle movement. The scatter light fluctuations caused by the movement of charged particles in the electric field were analyzed. The zeta potential was calculated based on the velocity of particles determined through the electrophoretic mobility via Zetasizer Ultra ZS Xplore software, which was further visualized by GraphPad Prism 10.1.2. Tm and Tagg determination HPF (WT) and HPF (I69C) proteins stored at -80 °C for 1 day or 10 months were subjected to thermal stability analysis using the Uncle system (Unchained Labs, USA). Protein samples were clarified by ultracentrifugation and diluted to 1 mg/mL in tris buffer (20 mM Tris, 50 mM NaCl, pH 7.5). A 9 µL aliquot was loaded into the sample chamber. The temperature ramp was programmed from 20 °C to 95 °C at 0.7 °C/min with a 180 s equilibration period and 45 s data acquisition intervals. Excitation light intensity was adjusted to ensure peak fluorescence signal intensity of no less than 1000 counts. Data were processed using Uncle Analysis software in Tm & Tagg mode. The barycentric mean (BCM) fluorescence signal within 300–375 nm was used to derive the thermal denaturation profile. The melting temperature (Tm), defined as the temperature at which 50% of the protein is unfolded, was determined from the peak of the first derivative of the BCM curve. The aggregation onset temperature (Tagg), defined as the temperature at which the protein begins to undergo significant aggregation, was determined from the SLS266 channel as the temperature corresponding to 10% of the peak of the first derivative of the static light scattering signal. Turbidity The turbidity of SC003-conjugated nanoparticles was performed to assess the stability of HPF variants. SC003-HPF and SC003-HPF (I69C) were expressed in BL21 cells for 18 h. Lysis buffers containing various concentrations (50, 100, 150, 200, and 500 mM) of sodium chloride (NaCl) were utilized to resuspend bacteria pellets, followed by high pressure homogenization to acquire protein lysates. The resulting lysates were transferred into spectrophotometer cuvettes, followed by measuring absorbance at OD600. Subsequently, lysates were centrifuged to separate supernatants from precipitates, which were visually examined to compare precipitate sizes. Target proteins in both supernatants and sediments were subjected to SDS-PAGE and Coomassie blue staining. Greater amounts of target proteins in supernatants and fewer in sediments indicated higher stability. Biolayer interferometry (BLI) assay BLI experiments were performed on an Octet R8 system (ForteBio). For receptor-RBD interaction analysis, RBD proteins including RBD (SARS2), RBD (Dimer), RBD (SARS2)-HPF (I69C), and RBD (Dimer)-HPF (I69C) were immobilized onto AR2G biosensors via amine coupling. Briefly, sensors were pre-equilibrated in PBST buffer (PBS containing 0.02% Tween-20), followed by EDC/NHS activation of the carboxylated sensor surface. RBD proteins were loaded and covalently immobilized through primary amine coupling, and unreacted sites were quenched with ethanolamine. Following baseline equilibration, the loaded sensors were incubated with serial concentrations of recombinant hACE2, and association/dissociation kinetics were recorded. hDPP4 binding was measured analogously, with RBD (MERS), RBD (Dimer), RBD (MERS)-HPF (I69C), and RBD (Dimer)-HPF (I69C) immobilized onto AR2G biosensors and incubated with serial concentrations of recombinant hDPP4. For antibody-RBD interaction analysis, REGN10933 (casirivimab) (kindly provided by Prof. Jincun Zhao at Guangzhou Medical University) was captured onto Protein A biosensors. After baseline re-equilibration in PBST, the sensors were incubated with serial concentrations of RBD (SARS2), RBD (Dimer), RBD (SARS2)-HPF (I69C), or RBD (Dimer)-HPF (I69C), and association/dissociation kinetics were recorded. Data acquisition and analysis were performed using Octet BLI Discovery software (version 13.0) and Octet BLI Analysis software (version 13.0), respectively. Binding kinetics for monovalent and dimeric RBD proteins were fitted to a 1:1 binding model to derive the equilibrium dissociation constant (KD). For nanoparticle-displayed RBDs, the reported apparent KD values reflected avidity-enhanced binding arising from multivalent antigen presentation. Endotoxin measurement Endotoxin levels were minimized through multiple established measures. Buffers were prepared with certified endotoxin-free water, and stock reagents were pre-screened prior to use. All glassware was depyrogenated at 250 °C for ≥ 4 h, and single-use endotoxin-free plasticware was used throughout. Affinity (TALON Co2+) and SEC columns were sanitized with 0.5 M NaOH between runs. The SEC step itself provided effective separation of endotoxin micelles from the 24-mer nanoparticle fraction based on their different hydrodynamic radii. Following SEC, five cycles of buffer exchange with endotoxin-free tris buffer through 100 kDa-cutoff ultrafiltration further depleted residual low-molecular-weight contaminants. Endotoxin levels in purified HPF-related protein preparations were quantified using a chromogenic Tachypleus amebocyte lysate (TAL) assay (End-point Chromogenic TAL Kit, EC80545S, Xiamen Bioendo Technology Co., Ltd.). The assay is based on endotoxin-mediated activation of the Factor C enzymatic cascade, which converts proclotting enzyme to clotting enzyme. The activated clotting enzyme cleaves a synthetic chromogenic substrate, releasing p-nitroaniline (pNA), which is subsequently diazotized and measured at 545 nm. The amount of pNA produced is proportional to the endotoxin concentration in the sample. Briefly, 100 µL of each protein sample, endotoxin standard (0.1, 0.25, 0.5, and 1.0 EU/mL), or endotoxin-free water (negative control) was added to pyrogen-free tubes, followed by 100 µL TAL reagent, mixed, and incubated at 37 °C for T1 min (per lot-specific quality control report). Then, 100 µL chromogenic substrate solution was added, mixed, and incubated at 37 °C for T2 min. The reaction was terminated by sequential addition of 500 µL each of diazotization reagents 1, 2, and 3, with mixing after each step. After standing for 5 min, absorbance at 545 nm was measured using a microplate reader. A standard curve (Y = bX + a, where Y is A545 and X is endotoxin concentration) was constructed. Assays were considered valid when |r| ≥ 0.980 and the A545 of the negative control was below that of the lowest standard. Endotoxin concentrations in test samples were interpolated from the standard curve. The pharmacopoeial limit for parenteral products specified by USP < 85>, Ph. Eur. 2.6.14, and ChP 1143 is 5 EU per kg of body weight per h, corresponding to 1.25–1.5 EU/mL in the 100 µL injection volume administered to mice in this study. All measured endotoxin levels were below this threshold. Enzyme linked immunosorbent assay (ELISA) Purified RBD proteins (1 µg/mL) of SARS-CoV-2 and MERS-CoV were coated on high-binding ELISA plates overnight at 4 °C, respectively. After washing with PBS/T for 3 times, each well was blocked with blocking buffer (5% non-fat dry milk in PBS) for 1 h at 37 °C. Plates were washed again with PBS/T following the removal of the blocking buffer. Serum from each mouse was 10-fold serially diluted with PBS in U-bottom plates. All the diluted sera were added into each well in duplicate, followed by incubating for 1 h at 37 °C. After incubating, plates were washed with PBS/T for 3 times. Then, horseradish peroxidase (HRP)-conjugated goat anti-mouse secondary antibodies at dilution of 1:10,000 were added into each well and incubated for 1 h at 37 °C. After washing with PBS/T for 5 times, HRP substrate TMB solution (eBioscience) was added, followed by stopping reaction with 1 M H2SO4. Absorption of each well was measured at 450 nm within 5 min. The absorption values were analyzed using GraphPad Prism 10.1.2 for non-linear regression to calculate the endpoint titer of each serum. Histopathology and immunohistochemistry C57BL/6 mice were euthanized one week after the boost vaccination. Major tissues including hearts, livers, spleen, lungs, kidney, cerebrum, intestines, and bladders were harvested and fixed in 4% paraformaldehyde (PFA). Then, these tissues were embedded within paraffin, followed by sectioning to obtain tissue slices for further analysis. These sections were stained with hematoxylin and eosin (H&E) for histopathology analysis. Images of H&E samples were captured with an inverted optical microscope (Sunny Optical Technology). Lungs of SARS-CoV-2- or MERS-CoV-challenged mice were fixed with 4% PFA for 7 days in the animal biosafety level 3 (ABSL-3) facility of Guangzhou National Laboratory, followed by H&E staining as above. H&E-stained infected-lung sections were imaged at 10× magnification under an optical microscope with blinded field selection. Three trained pathologists independently scored each section using a double-blind method. Scoring evaluated the bronchioles, pulmonary blood vessels, and alveolar regions, integrating both the extent of tissue damage and the degree of inflammatory cell infiltration. A semi-quantitative grading scale (0–4) was applied: 0, within normal limits for the animal’s age, sex, and strain; 1, slight changes above the normal range; 2, lesions easily identifiable but limited in severity; 3, lesions relatively obvious with significant potential for worsening; 4, lesions occupying the majority of the organ. The final pathological score for each section was the arithmetic mean of the three pathologists’ scores. Based on the final score, lesion severity was classified as: 0–1, normal lung tissue structure; 1–2, mild lesions; 2–3, moderate lesions; 3–4, severe lesions. Lungs of authentic virus-challenged mice were also subjected to the immunohistochemistry (IHC) assay. PFA-fixed lung tissues were deparaffinized and rehydrated with xylene and gradient alcohol, followed by microwave retrieval with citric acid buffer. The endogenous peroxidases were quenched with 3% H2O2 for 10 min. The non-specific binding regions were blocked by BSA, followed by incubating with 1,000-fold diluted primary antibodies targeting SARS-CoV-2 N (40143-T62, SinoBiological) and MERS-CoV N (40068-RP02, SinoBiological) for lung tissues of hACE2 and hDPP4 mice, respectively. Subsequently, the samples were incubated with HRP-conjugated secondary antibodies and stained with 3,3’-diaminobenzidine (DAB) and hematoxylin. The finalized samples were dehydrated with gradient ethanol and covered with neutral balsam. Images of IHC samples were captured in blind field selection with an inverted optical microscope (Sunny Optical Technology). All images were processed and analyzed using Image J software (National Institutes of Health, USA). In the IHC staining procedure, DAB specifically precipitated at antigen sites to produce a brown positive signal, while hematoxylin counterstained cell nuclei blue. To eliminate interference from nuclear gray signals while preserving the specific DAB-positive stained regions, each original image was split into three independent channels (red, green, and blue). Since DAB absorbs strongly in the blue spectrum while hematoxylin transmits blue light, the blue channel was selected to preferentially capture the DAB signal over the hematoxylin counterstain. The blue channel was converted to an 8-bit grayscale image and used for subsequent quantitative analysis. The grayscale analysis sensitivity threshold was set to 0–60 (total grayscale range 0-255), and the grayscale values of all pixels within the image were analyzed to obtain the integrated density value, which was used to quantify the extent of positive staining in each section and presented as integrated density of N per field. Pseudotyped virus neutralization assay To quantify neutralizing antibodies (nAbs) within serum, we generated pseudotyped SARS-CoV-2 and MERS-CoV viruses to perform neutralization assays. Briefly, lentivirus-based pseudotyped viruses were produced by co-transfecting HEK293T cells with a luciferase-expressing plasmid (pHIV-Luciferase, Addgene), a packaging plasmid (psPAX2, Addgene), and a plasmid expressing Spike proteins of SARS-CoV-2 (GenBank: QHD43416.1) or MERS-CoV (GenBank: AFS88936.1). About 6 h post-transfection (h.p.t.), the culture medium was replaced with freshly prepared DMEM. Another 48 h later, supernatants containing pseudotyped viruses were harvested and filtered with a 0.45 μm filter. For the detection of nAbs against pseudotyped SARS-CoV-2, serial 10-fold dilutions of serum were mixed with pseudotyped SARS-CoV-2 viruses, followed by incubating with HEK293T-hACE2 cells. At 48 h post-infection (h.p.i.), cells were lysed and measured for relative luminescence units which represented the relative expression of Luciferase. The half-maximal nAb titer (NT50) against pseudotyped viruses, defined as the reciprocal of half-maximal neutralizing dilution, was analyzed via GraphPad Prism 10.1.2. Serum nAb titers against pseudotyped MERS-CoV viruses were measured as above, except that the target cells used for infection were HEK293T-hDPP4 cells. Authentic virus neutralization assay Titers of nAbs targeting authentic viruses were quantified via focus reduction neutralization test (FRNT). Vero E6 cells were seeded into the 96-well plate one day prior to infection. Two hundred focus-forming units (FFU) of authentic SARS-CoV-2 D614 viruses (NC_045512.2, Wuhan-Hu-1) were mixed with 10-fold serially diluted serum, followed by incubating with Vero E6 cells at 37 °C for 1 h. The serum/virus mixture was removed from the 96-well plate. Cells were overlaid with DMEM containing 2% FBS and 1.6% sodium carboxymethyl cellulose (CMC). Another 24 h post-incubation at 37 °C, supernatants were completely removed. Cells were fixed with 4% PFA for 24 h, followed by washing with PBS 3 times. Subsequently, cells were permeabilized and blocked with PBS containing 0.2% Triton X-100 and 1% BSA for 30 min. After washing with PBS for 3 times, cells were incubated with diluted primary antibodies targeting SARS-CoV-2 N (40143-T62, SinoBiological) for 1 h at 37 °C. Then, cells were washed with PBS/T 3 times, followed by incubating with diluted HRP-conjugated secondary antibodies (SSA004, SinoBiological) at 37 °C. One h post-incubation, cells were washed with PBS/T for 3 times again. The TrueBlue (5510-0030, KPL) reagents were added into each well and incubated for 5 min. Subsequently, cells were washed with ddH2O twice, followed by imaging with ImmunoSpot microanalyzer to count SARS-CoV-2 N-specific spots. The FRNT50 titer was defined as the reciprocal of serum dilution at which nAbs achieved 50% inhibition of viral infection in cultured cells. The FRNT50 titers of nAbs in serum against authentic MERS-CoV viruses (KT006149.2, ChinaGD01) were determined as above, except that Vero 81 cells were used as target cells and antibodies targeting MERS-CoV N (40068-RP02, SinoBiological) served as primary antibodies. All authentic SARS-CoV-2 (Wuhan-Hu-1) and MERS-CoV (ChinaGD01) experiments were conducted in ABSL-3 facility of Guangzhou National Laboratory. Authentic virus challenge assay Both hACE2 and hDPP4 transgenic mice were respectively employed in authentic SARS-CoV-2 and MERS-CoV challenge assays. Briefly, all mice were prime/boost vaccinated as described for C57BL/6 mice. Serum samples were collected at week 0, 2, 4, and 5, which were utilized for the quantification of RBD-specific antibodies and nAbs. One week after the boost vaccination, mice were transferred to the ABSL-3 facility. hACE2 mice were anaesthetized with isoflurane and intranasally challenged with 5 × 104 FFU of authentic SARS-CoV-2 (Wuhan-Hu-1). Vaccinated hDPP4 mice were intranasally inoculated with 5 × 103 FFU of authentic MERS-CoV (ChinaGD01). The body weight of each mouse was monitored daily after infection. Mice were euthanized early if they lost more than 20% of baseline body weight or developed severe clinical signs such as hunched posture, piloerection, labored breathing, or impaired ambulation. Another week after challenge, all mice were euthanized, followed by the collection of lung tissues. One lung lobe of each mouse was subjected to H&E staining and IHC with antibodies against SARS-CoV-2 N or MERS-CoV N proteins. The remaining lung tissues were homogenized in 500 µL of PBS. About 50 µL of the homogenates were used for total RNA extraction and viral RNA quantification. The primers and probes specific to SARS-CoV-2 N gene used for reverse transcription-quantitative polymerase chain reaction (RT-qPCR) were SN-F (5’-CAGTAGGGGAACTTCTCCTGCT-3’), SN-R (5’-TCTGTCAAGCAGCAGCAAAG-3’), and SN-P (5’-FAM-CTGGCAATGGCGGTGATGCTGC-BHQ1-3’). The MERS-CoV ORF5 gene-specific primers and probes were MO-F (5’-GCAACGCGCGATTCAGTT-3’), MO-R (5’-GCCTCTACACGGGACCCATA-3’), and MO-P (5’-FAM-CTCTTCACATAATCGCCCCGAGCTCG-BHQ1-3’). The remaining homogenates were subjected to a focus-forming assay (FFA). Homogenates from SARS-CoV-2-challenged hACE2 mice were incubated with Vero E6 cells, while homogenates from MERS-CoV-challenged hDPP4 mice were incubated with Vero 81 cells. The subsequent procedures of FFA were the same as those of FRNT. All authentic viral challenge assays and viral titer quantification assays were carried out in the ABSL-3 facility of Guangzhou National Laboratory. Antigen-specific B cell detection RBD-specific mouse B cells were evaluated via flow cytometry (FCM). Briefly, Avi-tagged RBD (SARS2) and RBD (MERS) were purified from HEK293F cells, followed by the biotinylation of purified proteins utilizing biotin protein ligase (BI001, GeneCopoeia). Biotinylated RBD (SARS2) and RBD (MERS) proteins were further labelled with PE (SA10044, Invitrogen)- and AF488 (S32354, Invitrogen)-conjugated streptavidin, respectively. To quantify the percentages of RBD-specific B cells, about 1.5 × 106 splenocytes were incubated with 3 µg of PE-RBD (SARS2) or AF488-RBD (MERS) for 1 h at 4 °C, followed by incubating with APC/Cy7-CD3 (100222, BioLegend), BV605-B220 (103244, BioLegend), BV421-CD73 (127217, BioLegend), and APC-CD80 (104714, BioLegend) antibodies for 30 min. The CD3−/B220+/RBD+ splenocytes were gated and represented RBD-specific B cells, while CD3−/B220+/CD73+/CD80+/RBD+ splenocytes were gated and represented RBD-specific memory B cells (MBCs). Tfh cell and antigen-specific T cell quantification Splenic Tfh cells were quantified for each vaccinated mouse. Briefly, 1.5 × 106 splenocytes were stained with FITC-CD3 (100306, BioLegend), AF700-CD4 (100536, BioLegend), BV605-PD-1 (135220, BioLegend) and BV650-CXCR5 (563981, BD Biosciences) antibodies for 30 min, followed by FCM analysis. The CD3+/CD4+/CXCR5+/PD-1+ splenocytes were gated and represented Tfh cells. Antigen-specific polyfunctional CD4+ and CD8+ T cells were analyzed by intracellular cytokine staining (ICCS). Briefly, 2 × 106 splenocytes were stimulated with 50 µg/mL peptide pools of SARS-CoV-2 RBD and MERS-CoV RBD, respectively. Both peptide pools of SARS-CoV-2 RBD (53 peptides) and MERS-CoV RBD (58 peptides) were 15-mers with 11-amino-acid overlap, covering the entire RBD sequence. Cells were co-stimulated with anti-mouse CD28 antibodies (102116, BioLegend) at 37 °C with 5% CO2 for 1 h, followed by the incubation with Brefeldin A (S7046, Selleck) for 5 h at 37 °C with 5% CO2. FITC-CD3, AF700-CD4, and PE/Cy7-CD8a (100722, BioLegend) antibodies were further utilized to stain splenocytes. The dead cells were excluded via the Fixable Viability Dye eFluor 506 kit (65-0866-18, Invitrogen). Cells were fixed and permeabilized with Cytofix/Cytoperm Fixation/Permeabilization Kit (554714, BD Biosciences). Subsequently, intracellular staining was performed utilizing APC-TNF-α (506307, BioLegend), BV421-IFN-γ (505830, BioLegend), and BV605-IL-2 (503829, BioLegend) antibodies for 30 min. The TNF-α+, IFN-γ+, and IL-2+ T cells were analyzed by FCM for both CD3+/CD8+ and CD3+/CD4+ T-cell populations. Blood biochemical index and urinalysis in mice To verify the safety of nanoparticle vaccines, whole blood and urine samples were collected from each mouse one day prior to euthanasia for complete blood count (CBC), blood biochemical analysis, and urinalysis. At necropsy, we also did not observe any visible injection-site abnormalities across any group, including no erythema, swelling, induration, alopecia, or granuloma formation. Therefore, the injection-site reactogenicity was not statistically indicated. Blood samples were collected using the retro-orbital sinus method, and subsequently subjected to cell counting and blood biochemistry via automatic hematology analyzer and chemistry analyzer, respectively. The numbers of white blood cells (WBCs), platelets (PLTs), and red blood cells (RBCs) were counted. WBC counts ranging from 0.8 × 109 to 10.6 × 109 cells/L were considered within the normal reference range. PLT counts ranging from 400 × 109 to 1,600 × 109 cells/L were considered normal, while RBC numbers ranging from 6.5 × 1012 to 11.5 × 1012 cells/L were also considered normal. Hemoglobin (HGB) concentrations ranging from 110 to 165 g/L were defined as within the normal range, and hematocrit (HCT) values between 35% and 55% were considered normal. Biochemical indexes representing liver and kidney functions were measured using enzymatic colorimetric assays. Major indexes and corresponding normal ranges were shown as follows: alanine aminotransferase (ALT), 10.06 to 96.47 U/L; aspartate aminotransferase (AST), 36.31 to 235.48 U/L; alkaline phosphatase (ALP), 22.52 to 474.35 U/L; total bilirubin (TBIL), 6.09 to 53.06 µmol/L; gamma-glutamyl transferase (γ-GT), 0 to 7.78 U/L; total bile acids (TBA), 0 to 8.51 µmol/L; urea (UREA), 3.9 to 12.4 mmol/L; creatinine (CREA), 10.91 to 85.09 µmol/L. Urine samples were obtained through manual bladder expression method. WBC counts marked with “–” represented 0 cells/L, “+-” represented ≤ 15 cells/L, “+1” represented ≤ 70 cells/L, “+2” represented ≤ 125 cells/L, and “+3” represented ≤ 500 cells/L. Blood (BLD) counts marked with “–” represented 0 cells/L, “+-” represented ≤ 10 cells/L, “+1” represented ≤ 25 cells/L, “+2” represented ≤ 80 cells/L, and “+3” represented ≤ 200 cells/L. The presence of nitrite (NIT) was marked with “+”, while NIT absence was marked with “–“. Ketone (KET) concentrations marked with “–” represented 0 mmol/L, “+-” represented ≤ 0.5 mmol/L, “+1” represented ≤ 1.5 mmol/L, “+2” represented ≤ 4.0 mmol/L, and “+3” represented ≥ 8.0 mmol/L. Urobilinogen (URO) concentrations marked with “normal” represented 0 µmol/L, “+1” represented ≤ 33 µmol/L, “+2” represented ≤ 66 µmol/L, and “+3” represented ≥ 131 µmol/L. Bilirubin (BIL) concentrations marked with “–” represented 0 µmol/L, “+1” represented ≤ 8.6 µmol/L, “+2” represented ≤ 33 µmol/L, and “+3” represented ≥ 100 µmol/L. Specific gravity (SG) values with 1.005, 1.010, 1.015, 1.020, 1.025, and 1.030 represented the concentration of solutes in urine. The pH values ranging from 5.0 to 9.0 represented urine acidity/alkalinity. CREA concentrations were measured and categorized into the following ranges: ≤ 0.9, ≤ 4.4, ≤ 8.8, ≤ 17.6, and ≥ 26.4 mmol/L. Protein (PRO) concentrations marked with “–” represented 0 g/L, “+-” represented ≤ 0.15 g/L, “+1” represented ≤ 0.3 g/L, “+2” represented ≤ 1.0 g/L, and “+3” represented ≤ 3.0 g/L. Glucose (GLU) concentrations marked with “–” represented 0 mmol/L, “+-” represented ≤ 2.8 mmol/L, “+1” represented ≤ 5.5 mmol/L, “+2” represented ≤ 14 mmol/L, “+3” represented ≤ 28 mmol/L, and “+4” represented ≥ 55 mmol/L. All the analysis data were presented as semi-quantitative tables. Statistical analysis Statistical details, including specific statistical tests, sample sizes, mean values, standard errors of the mean (SEM), and P values, have been indicated in figure legends and shown in figures. All statistical analyses were conducted via GraphPad Prism 10.1.2, Microsoft Excel, or Image J. Histopathological scoring and viral load quantification were performed with investigators blinded to group allocation. Triplicate, quadruplicate, pentaplicate, sextuplicate and other replicates were presented as mean ± SEM. For comparisons between two groups, Student’s t-tests were conducted. For comparisons between more than two groups, ANOVA or Kruskal-Wallis analyses were performed. Statistical significance was defined as follows: ns (not significant), P ≥ 0.05. * P < 0.05, ** P < 0.01, *** P < 0.001. Results Design of optimized ferritin nanoparticles incorporating disulfide bonds The self-assembling 24-mer ferritin nanoparticle, particularly ferritin from Helicobacter pylori (HPF), has been widely used for drug and vaccine delivery [39]. However, ferritin-based nanoparticle vaccines often encounter challenges related to limited solubility, homogeneity, and long-term stability, which significantly hinder scale-up manufacturing, in vivo safety evaluation, and regulatory approval [40]. We utilized artificial intelligence (AI)-based rational design strategies to optimize HPF nanoparticles. Four disordered residues including methionine (M), leucine (L), serine (S), and lysine (K) at the N-terminal of HPF were removed to facilitate subsequent antigen conjugation. Therefore, the aspartic acid residue at position 5 (D5) was renumbered as position 1 (D1) in our optimization. Additionally, we mutated the asparagine residue at position 15 (N15) (corresponding to N19 in original full-length HPF) to glutamine residue (Q) to abort the potential N-linked glycosylation [25]. The formation of supramolecular HPF assemblies and subsequent sediments results from the unstable assembly strength between 24 protomers and the strong hydrophobic interactions between nanoparticles. We have tried to substitute hydrophobic residues on the HPF surface with hydrophilic residues utilizing Rosetta optimization, however, all the attempts were unsuccessful, resulting in the de-assembly of 24-mer nanoparticles. Subsequently, we introduced disulfide bonds between adjacent protomers to covalently strengthen intermolecular interactions within nanoparticles. Candidate disulfide bond sites were screened based on the following geometric criteria, Cβ-Cβ distance between candidate residue pairs within 4–6 Å to permit disulfide bond formation, side-chain orientation after cysteine substitution compatible with disulfide geometry, and location at the subunit interface without disruption of native secondary structure or key functional regions. Two loops spanning the 64th valine residue (V64) to H74 between adjacent protomers exhibited high flexibility and spatial proximity. Through AlphaFold 3 prediction and Rosetta optimization, the region from L66 to P72 on one loop, which was within a distance of less than 5.0 Å from an adjacent loop on another protomer, was identified as a potential site for cysteine (C) mutation. As two loops were in an antiparallel arrangement, we introduced paired and centrosymmetric mutations to facilitate disulfide bond formation. These paired mutations included L66C + P72C, T67C + A71C, and S68C + S70C (Fig. 1A C). The isoleucine (I) at position 69 was precisely symmetric with I69 of the other loop, which formed backbone hydrogen bonds between I69 residues. Therefore, only one I69C mutation was introduced in this variant (Fig. 1D). We overexpressed and purified these optimized 6× histidine (6His)-tagged HPF mutants in Escherichia coli BL21 (DE3) cells. All mutants were successfully expressed (Fig. 1E and H). However, the protein yields of HPF (L66C + P72C) and HPF (S68C + S70C) were slightly lower than those of HPF (T67C + A71C) and HPF (I69C) upon elution with imidazole. Redundant contaminant proteins with high molecular weight also appeared in HPF (S68C + S70C) eluates. During ultrafiltration concentration, HPF (T67C + A71C) proteins were precipitated, resulting in low soluble fractions. Significantly, the HPF (I69C) mutant exhibited the highest yield, with no precipitates observed upon ultrafiltration. Fig. 1The optimization of ferritin nanoparticles by disulfide bond introduction. (A-D) Helicobacter pylori ferritin (HPF) 24-mer, with four-amino-acid N-terminal deletion and the N15Q mutation, was predicted by AlphaFold 3 based on the determined crystal structure (PDB: 3EGM). The surface representation of each HPF model showed locations of engineered cysteine pairs (left panel) and the amplified inter-subunit interfaces (right panel). Subunits were colored individually to highlight inter-subunit interfaces. Engineered cysteine residues were labeled in the zoomed panels. (E-H) SDS-PAGE analysis of lysate and Ni-affinity purification fractions for each HPF construct. Lanes were labeled (left to right) as crude lysate, flow-through, and imidazole elution gradient. (I-L) Size-exclusion chromatography (SEC) profiles for purified HPF variants with a Superose 6 Increase column, monitored by absorbance at 280 nm (A280). (M) Eluates from major peaks of SEC profiling, which contained HPF variants, were analyzed by SDS-PAGE. Red dashed boxes highlighted redundant proteins. (N) Native PAGE (non-denaturing) of purified HPF variants run in parallel (same lane order as in M). HPF (S68C + S70C) contained redundant proteins with high and low molecular weight. (O) SDS-PAGE analysis of HPF (WT) and HPF (I69C) under reducing (+ DTT) and non-reducing (-DTT) conditions. The 40-kDa band observed for HPF (I69C) under non-reducing conditions corresponded to the disulfide bond-linked dimer The optimization of ferritin nanoparticles by disulfide bond introduction. (A-D) Helicobacter pylori ferritin (HPF) 24-mer, with four-amino-acid N-terminal deletion and the N15Q mutation, was predicted by AlphaFold 3 based on the determined crystal structure (PDB: 3EGM). The surface representation of each HPF model showed locations of engineered cysteine pairs (left panel) and the amplified inter-subunit interfaces (right panel). Subunits were colored individually to highlight inter-subunit interfaces. Engineered cysteine residues were labeled in the zoomed panels. (E-H) SDS-PAGE analysis of lysate and Ni-affinity purification fractions for each HPF construct. Lanes were labeled (left to right) as crude lysate, flow-through, and imidazole elution gradient. (I-L) Size-exclusion chromatography (SEC) profiles for purified HPF variants with a Superose 6 Increase column, monitored by absorbance at 280 nm (A280). (M) Eluates from major peaks of SEC profiling, which contained HPF variants, were analyzed by SDS-PAGE. Red dashed boxes highlighted redundant proteins. (N) Native PAGE (non-denaturing) of purified HPF variants run in parallel (same lane order as in M). HPF (S68C + S70C) contained redundant proteins with high and low molecular weight. (O) SDS-PAGE analysis of HPF (WT) and HPF (I69C) under reducing (+ DTT) and non-reducing (-DTT) conditions. The 40-kDa band observed for HPF (I69C) under non-reducing conditions corresponded to the disulfide bond-linked dimer To evaluate the nanoparticle assembly capabilities of these HPF mutants, we employed size-exclusion chromatography (SEC) using a Superose 6 Increase column to separate target proteins. HPF (L66C + P72C), HPF (T67C + A71C), and HPF (I69C) samples exhibited sharp peaks at retention volumes of 13–16 mL, while the major peak of HPF (S68C + S70C) samples appeared at retention volumes of 16–20 mL (Fig. 1I and L). Notably, samples from HPF (L66C + P72C), HPF (T67C + A71C), and HPF (S68C + S70C) also showed peaks broadening at retention volumes of 16–20 mL, 16–22 mL, and 8–16 mL, respectively, while only one peak was observed for HPF (I69C) sample, resembling higher homogeneity of HPF (I69C) compared to the other mutants. Eluates from major peaks were collected and concentrated, followed by both SDS-PAGE and native PAGE to identify target proteins. The Coomassie blue staining of SDS-PAGE gel indicated that all samples contained target HPF mutant proteins, which migrated to a position similar to that of the wild-type (WT) HPF (Fig. 1M). The HPF (WT) sample also contained a higher quantity of redundant proteins with low molecular weight compared to HPF mutants, which might be attributed to the delayed and long-term incorporation of host proteins into nanoparticles during the assembly process. Notably, samples from HPF (T67C + A71C) and HPF (S68C + S70C) also contained considerable amounts of redundant proteins with high molecular weight, which might result from improper assembly between HPF mutants and host proteins. Only the HPF (L66C + P72C) and HPF (I69C) samples showed higher purity. The native PAGE results further confirmed that these HPF mutants formed homogeneous nanoparticles, migrating to positions similar to that of HPF (WT) (Fig. 1N). The molecular weight of the observed protein band corresponded to that of a 24-fold HPF, indicating the successful assembly of 24-mer nanoparticles. However, the HPF (S68C + S70C) sample also contained other types of protein polymers, which was consistent with the presence of redundant proteins identified by SDS-PAGE. Considering that the HPF (I69C) mutant was characterized by higher yield, purity, and homogeneity, we selected this variant for further analysis. Through both reduced and non-reduced SDS-PAGE, we confirmed that the majority of HPF (I69C) migrated as a 40 kDa band under non-reducing conditions, corresponding to the inter-subunit disulfide bond-linked dimer, whereas HPF (WT) migrated as a single 20 kDa monomeric band under both reducing and non-reducing conditions (Fig. 1O). A minor fraction of HPF (I69C) monomers was also observed under non-reducing conditions. This was likely attributable to the stochastic nature of inter-subunit disulfide bond formation within the 24-mer assembly, a phenomenon commonly observed in engineered disulfide bond-containing multimeric proteins. Structure determination confirmed disulfide bond formation within HPF (I69C) To characterize the biophysical properties of HPF (I69C), we measured its size and zeta potential distributions via dynamic light scattering (DLS). The results demonstrated that both HPF (WT) and HPF (I69C) showed similar size distributions, with mean hydrodynamic diameters centered around the same size of 14 nm (Fig. 2A). However, upon calculating the polydispersity index (PDI), HPF (WT) nanoparticles exhibited slightly broader distribution, with a PDI value of 0.2588. In contrast, HPF (I69C) nanoparticles demonstrated a lower PDI value of 0.1746, indicating a narrower distribution. Both nanoparticles were negatively charged on surface, with a median zeta potential centered around the same value of -10 mV (Fig. 2B and D). Nanoparticles with moderate negative-charge surface are less likely to disrupt cell membranes or induce membrane-associated toxicity, representing favorable biocompatibility. These nanoparticles can still be internalized by dendric cells and macrophages, thereby facilitating antigen presentation without causing significant cytotoxicity. To visualize the shape and measure the size of 24-mer HPF (I69C), we utilized the negative staining to characterize these nanoparticles. The negative staining image showed that all HPF (I69C) nanoparticles were uniformly covered on the copper grid (Fig. 2E). The average diameter of these nanoparticles was approximately 12 nm (Fig. 2F). The size of HPF (I69C) nanoparticle measured by DLS was larger than that determined by negative staining visualization, which was attributed to the different physical properties being measured by each technique. Negative staining measured the projected physical diameter of individual particles, whereas the hydrodynamic diameter measured by DLS represented the effective size of these particles plus any solvent layer. Fig. 2The structure of optimized ferritin with I69C mutation. (A) Overlaid size distributions for HPF (WT) (red) and HPF (I69C) (blue) were evaluated via dynamic light scattering (DLS). Recorded polydispersity indexes (PDIs) for each sample were indicated on the panel. (B-C) Zeta potential distribution histograms for HPF (WT) (B) and HPF (I69C) (C) were shown. (D) Bar chart showing median zeta potential for both nanoparticles. (E) Representative negative staining image of HPF (I69C). Scale bar represented 100 nm. (F) Representative 2D class averages derived from negative staining data were shown, demonstrating consistent ring-like morphology and internal cavities. (G) The overall structure of HPF (I69C) was depicted in ribbon representation. Each subunit was displayed in a unique color. (H) A magnified view of the electron density corresponding to a representative S–S bond formed between I69C residues. Chain A and chain B were colored purple and pink, respectively. (I) The distribution of engineered S–S bonds formed between cysteine residues at position 69 (I69C) on adjacent chains. The overall structure was shown as gray ribbons and S–S bonds were highlighted as yellow spheres. (J-K) Electrostatic surface renderings of HPF (PDB: 3EGM) (J) and HPF (I69C) (PDB: 25HO) (K) nanoparticles. Negative patches (red) and positive patches (blue) were highlighted and repeated by icosahedral symmetry. Right panels of each map showed zoomed regions of inter-subunit interfaces The structure of optimized ferritin with I69C mutation. (A) Overlaid size distributions for HPF (WT) (red) and HPF (I69C) (blue) were evaluated via dynamic light scattering (DLS). Recorded polydispersity indexes (PDIs) for each sample were indicated on the panel. (B-C) Zeta potential distribution histograms for HPF (WT) (B) and HPF (I69C) (C) were shown. (D) Bar chart showing median zeta potential for both nanoparticles. (E) Representative negative staining image of HPF (I69C). Scale bar represented 100 nm. (F) Representative 2D class averages derived from negative staining data were shown, demonstrating consistent ring-like morphology and internal cavities. (G) The overall structure of HPF (I69C) was depicted in ribbon representation. Each subunit was displayed in a unique color. (H) A magnified view of the electron density corresponding to a representative S–S bond formed between I69C residues. Chain A and chain B were colored purple and pink, respectively. (I) The distribution of engineered S–S bonds formed between cysteine residues at position 69 (I69C) on adjacent chains. The overall structure was shown as gray ribbons and S–S bonds were highlighted as yellow spheres. (J-K) Electrostatic surface renderings of HPF (PDB: 3EGM) (J) and HPF (I69C) (PDB: 25HO) (K) nanoparticles. Negative patches (red) and positive patches (blue) were highlighted and repeated by icosahedral symmetry. Right panels of each map showed zoomed regions of inter-subunit interfaces To further confirm the formation of disulfide bonds (S–S) between adjacent protomers on HPF (I69C) nanoparticles, we determined the structure using cryo-electron microscopy (cryo-EM) (Fig. S1; Table S1). The representative micrograph of HPF (I69C) nanoparticles showed a uniform distribution and characteristic circular projections (Fig. S1A). The final cryo-EM reconstruction was achieved at a global resolution of 2.2 Å, with central regions reaching a higher local resolution ranging from 1.7 to 1.9 Å (Fig. S1A and S1B). The gold-standard Fourier shell correlation (FSC) curve confirmed the 2.2 Å global resolution at the 0.143 threshold (Fig. S1C). Euler angle distribution analysis demonstrated comprehensive orientation sampling of particles, supporting the reliability of the reconstruction (Fig. S1D). At this resolution, side-chain densities were clearly resolved for nearly all amino acid residues, as exemplified by the density map across residues E59-T80 of HPF (I69C) chain A, allowing unambiguous atomic model building (Fig. S1E). The cryo-EM structure of HPF (I69C) nanoparticles revealed a symmetric and spherical assembly comprising 24 repeating protomers (Fig. 2G). The overall internal cavity diameter of these nanoparticles was 8 nm, whereas the outer diameter was about 12 nm, which was consistent with the diameter characterized by negative staining and reported by others [41]. Of note, a magnified view revealed electron density belonging to side chains of two adjacent engineered cysteines at position 69, which supported the formation of a covalent S–S bond (Fig. 2H). Across the entire HPF (I69C) assembly, 12 such engineered S–S bonds were uniformly distributed on the surface, stabilizing the nanoparticle structure through inter-subunit crosslinking (Fig. 2I). This structural observation aligned with our biochemical data, which indicated enhanced nanoparticle stability and homogeneity. Additionally, electrostatic surface renderings of both nanoparticles showed that the introduction of S–S bond slightly increased the net charge zeta potential at the interface, although the overall zeta potential remained nearly unchanged (Fig. 2J and K). This charge alteration at the interface might alter the pairing mode and the distribution of interfacial water molecules, consequently increasing the solubility and suppressing non-specific aggregation of nanoparticles. To further elucidate the mechanism underlying these improvements, we performed quantitative surface and interface analyses. Surface electrostatic potential analysis revealed that the I69C mutation reduced the fluctuation amplitude of mean electrostatic potential among subunits from 0.27 to 0.10 kcal/(mol·e), corresponding to an approximately 63% reduction, indicative of a more homogeneous charge distribution (Table S2). Solvent-accessible surface area (SASA) and molecular lipophilicity potential (MLP) analyses further demonstrated that HPF (I69C) adopted a more compact conformation with a 45% reduction in surface lipophilicity fluctuation (Table S3). Rosetta InterfaceAnalyzer confirmed that the I69C mutation enhanced inter-subunit binding stability, with interface separation free energy (dG_separated) increasing from − 42.826 to -58.002 kcal/mol and shape complementarity improving from 0.545 to 0.594 (Table S4). Collectively, the cryo-EM structure of HPF (I69C) delineated its quaternary architecture, confirmed the formation of disulfide bond between protomers, and provided direct evidence for disulfide-mediated crosslinking that likely contributed to the increased stability of nanoparticles. Optimized ferritin with I69C exhibited enhanced pH and freeze-thaw stability Previous reports have shown that ferritin proteins from different species exhibit remarkable structural stability over a pH range of 2 to 12 [42]. To verify whether HPF (I69C) also maintained stability under extreme pH conditions, we assessed the size distribution of both HPF (WT) and HPF (I69C) at pH values of 3.0, 5.0, 7.0, 9.0, and 11.0, respectively. The DLS results showed that the median hydrodynamic diameters of both nanoparticles at various pH were still around 14 nm (Fig. S2A and S2B). However, the PDI values at pH 3.0 and 11.0 were markedly higher than those at pH 5.0, 7.0, and 9.0, ranging from 0.2 to 0.3, indicating broader particle size distribution and reduced homogeneity. These nanoparticles still maintained a negatively charged surface at various pH. However, the median zeta potential of both HPF (WT) and HPF (I69C) samples increased at both low and high pH values, approaching zero net charge, with the minimum values observed at pH 7.0 (Fig. S2C and S2D). The SDS-PAGE analysis of HPF (WT) samples revealed that the pH alteration led to the presence of additional low-molecular-weight fragments (Fig. S2E). HPF (I69C) samples maintained high protein stability across all tested pH levels, as indicated by a single protein band (Fig. S2F). Densitometric analysis confirmed that HPF (I69C) retained substantially higher protein integrity across all pH conditions compared to HPF (WT), which exhibited marked accumulation of degradation fragments at pH extremes (Fig. S2G and S2H). Nonetheless, both HPF (WT) and HPF (I69C) samples analyzed by native PAGE contained stable and homogeneous 24-mer nanoparticles at various pH conditions (Fig. S2I-S2L). The redundant fragments observed in HPF (WT) samples might have migrated out of the native PAGE gel or been noncovalently associated with nanoparticles. Taken together, the optimized ferritin with I69C mutation demonstrated enhanced pH stability compared to the wild-type protein. A promising nanoparticle-based vaccine is expected to maintain high stability during freeze-thaw cycles. To assess this property, we evaluated the biochemical and structural integrity of both HPF (WT) and HPF (I69C) nanoparticles following one and five freeze-thaw cycles, with temperatures transitioning from − 80 °C to 37 °C. The DLS analysis showed that both nanoparticles maintained similar diameters of 14 nm after one-cycle and five-cycle freeze-thaw processes (Fig. S3A and S3B). Consistent with our previous observations, HPF (I69C) maintained a narrow size distribution, with PDI values ranging from 0.1 to 0.2 after multiple freeze-thaw cycles. In contrast, HPF (WT) displayed higher PDI values, ranging from 0.2 to 0.3, suggesting a slightly broader size distribution following freeze-thaw cycles. Both nanoparticles maintained similar median zeta potential after freeze-thaw tests, showing only a minor increase from − 10 mV to -8 mV after five-cycle treatment (Fig. S3C and S3D). We also compared the protein composition stability via SDS-PAGE analysis upon freeze-thaw. Unexpectedly, a small number of aggregates with high molecular weight was observed in HPF (WT) samples after five freeze-thaw cycles, whereas the HPF (I69C) still maintained compositional stability between one and five freeze-thaw cycles, as evidenced by the presence of a single protein band corresponding to the monomer (Fig. S3E-S3H). Conversely, after one and five freeze-thaw cycles, both HPF (WT) and HPF (I69C) in native PAGE migrated to positions similar to those observed for freshly prepared samples in our previous study (Fig. S3I-S3L). These investigations indicated that the optimized HPF (I69C) nanoparticle exhibited improved freeze-thaw stability, with no observable aggregate formation. To evaluate the long-term stability of HPF (I69C), we compared nanoparticles stored at -80 °C for 1 day vs. 10 months. DLS analysis revealed that all samples maintained monodisperse size distributions, with HPF (I69C) exhibiting consistently lower PDI values than HPF (WT) at both time points (Fig. S4A). Zeta potential values remained approximately − 10 mV across all samples (Fig. S4B). SDS-PAGE analysis revealed that HPF (WT) exhibited substantial degradation across both time points, whereas HPF (I69C) retained predominantly intact protein (Fig. S4C). Native PAGE showed that both HPF (WT) and HPF (I69C) maintained intact 24-mer nanoparticles with no detectable aggregates at both time points (Fig. S4D). Densitometric quantification of SDS-PAGE revealed that HPF (WT) retained only 70–88% intact protein compared with 93–96% for HPF (I69C), while native PAGE densitometry demonstrated 100% intact 24-mer nanoparticles for both proteins (Fig. S4E and S4F). The apparent discrepancy between SDS-PAGE and native PAGE results was likely attributable to the fact that degraded protein fragments migrated out of the native PAGE gel and thus were not visible under native conditions. Thermal stability was further assessed using the Uncle system. Both proteins exhibited melting temperature (Tm) values exceeding 95 °C, the upper detection limit of the instrument (Fig. S4G). However, HPF (I69C) displayed a higher initial barycentric mean (BCM) fluorescence value and a shallower unfolding trajectory, indicating a more compactly folded native state with greater resistance to thermal perturbation. The aggregation onset temperature (Tagg) of HPF (I69C) was approximately 8 °C higher than that of HPF (WT) (63.09 °C vs. 54.92 °C at day 1, 61.76 °C vs. 54.73 °C at 10 months) (Fig. S4H). Taken together, these results demonstrated that the inter-subunit disulfide bonds in HPF (I69C) conferred superior long-term storage stability, as evidenced by higher protein integrity and elevated aggregation resistance relative to HPF (WT) after 10 months of -80 °C storage. HPF (I69C) exhibited enhanced freeze-thaw stability upon conjugating antigen Given the improved pH and freeze-thaw stability of HPF (I69C), we subsequently utilized the molecular glue conjugation system to covalently display coronaviral antigens on HPF (I69C) nanoparticles. One of the most efficient protein-based molecular glues is the improved version of ST003/SC003 conjugation system, which covalently links two proteins by forming an intermolecular isopeptide bond between an aspartic acid (D) residue on the ST003 peptide and a lysine (K) residue on the SC003 domain [33]. Our previous attempt on HPF (WT) has shown that both HPF and SC003-conjugated SC003-HPF exhibited reduced solubility in sodium chloride (NaCl) concentrations exceeding 100 mM. However, the physiological salt concentration in human serum typically ranges from 140 to 150 mM, raising concerns that HPF-derived nanoparticle vaccines may precipitate at the injection site. Therefore, we firstly evaluated the stability of SC003-HPF (I69C) in various NaCl concentrations. Our results demonstrated that upon lysing nanoparticle-expressing bacteria in buffers containing increased concentrations of NaCl (50, 100, 150, 200, and 500 mM), SC003-HPF (I69C) exhibited enhanced solubility compared to SC003-HPF, as evidenced by reduced turbidity and decreased precipitation (Fig. S5A and S5B). The Coomassie blue staining of the supernatant and sediment under varying NaCl concentrations revealed that SC003-HPF (I69C) contained higher proportions of soluble fractions and correspondingly lower levels of insoluble aggregates compared to SC003-HPF (Fig. S5A). Although small amounts of SC003-HPF (I69C) precipitates still formed at NaCl concentrations of 200 and 500 mM, the finalized vaccines seldom employed buffers containing such high salt concentrations. Subsequently, we utilized the ST003-conjugated receptor-binding domain (RBD) of Middle East respiratory syndrome coronavirus (MERS-CoV) to assess the conjugation efficiency of the ST003/SC003 molecular glue. SC003-HPF (I69C) nanoparticles were incubated with RBD (MERS)-ST003 at RBD: Nano molar ratios of 1:1, 2:1, and 3:1, and conjugation efficiency was quantified at defined time points by SDS-PAGE densitometry. At a 1:1 ratio, conjugation reached approximately 95% within 30 min and was complete by 60 min (Fig. S5C and S5D). At higher ratios, conjugation was markedly accelerated, achieving completion within 10 min at a 2:1 ratio and within 5 min at a 3:1 ratio, demonstrating the high efficiency and dose-dependent kinetics of the ST003/SC003 conjugation system. To evaluate whether HPF (I69C) nanoparticles maintained enhanced freeze-thaw stability upon conjugating RBD (MERS) antigens, both RBD (MERS)-HPF and RBD (MERS)-HPF (I69C) were subjected to one and five cycles of freeze-thaw. The DLS quantifications revealed that both RBD (MERS) nanoparticles exhibited an average diameter of 25 nm after various freeze-thaw cycles (Fig. 3A and B). However, RBD (MERS)-HPF (I69C) showed smaller PDI values than RBD (MERS)-HPF, suggesting enhanced monodispersity following the I69C mutation. Both nanoparticles presented similar alterations of zeta potentials after freeze-thaw treatment, with a median value of -10 mV (Fig. 3C and D). The SDS-PAGE analysis demonstrated that both nanoparticles exhibited protein composition stability, with no aggregates or degraded fragments presented after one or five cycles of freeze-thaw processes (Fig. 3E and H). However, the native PAGE results showed that RBD (MERS)-HPF formed a small quantity of abnormal aggregates with high molecular weight, which were not observed in RBD (MERS)-HPF (I69C) (Fig. 3I and L). Fig. 3The freeze-thaw stability of HPF (I69C) nanoparticle vaccine against MERS-CoV. (A-B) MERS-CoV RBD-conjugated HPF and HPF (I69C) nanoparticle vaccines were subjected to a single freeze-thaw cycle (Freeze-Thaw×1, red) and five repeated freeze-thaw cycles (Freeze-Thaw×5, blue). Overlaid size distribution curves of both nanoparticle vaccines were evaluated by DLS. PDI values were indicated on the panels. (C-D) Median zeta potential of RBD (MERS)-HPF (C) and RBD (MERS)-HPF (I69C) (D) after Freeze-Thaw×1 and Freeze-Thaw×5. (E-F) SDS-PAGE analysis of both nanoparticle vaccines upon Freeze-Thaw×1 and Freeze-Thaw×5. (G-H) Densitometric quantification of SDS-PAGE bands in (E) and (F), respectively. Data were presented as proportions of intact proteins and degraded proteins relative to total lane intensity. (I-J) Native PAGE analysis of assembled nanoparticle vaccines after different freeze-thaw cycles. The red dashed box highlighted aggregates of RBD (MERS)-HPF nanoparticles. (K-L) Densitometric quantification of native PAGE bands in (I) and (J), respectively. Data were presented as proportions of intact nanoparticles and abnormal aggregates relative to total lane intensity. (M-N) Approximately 10 µg of RBD (MERS)-HPF and RBD (MERS)-HPF (I69C) subjected to 1 or 5 freeze-thaw cycles were administered to C57BL/6 mice as a single immunization. The serum samples were collected at 4 weeks post-immunization. The MERS-CoV RBD-specific binding IgG antibodies in serum were evaluated via enzyme-linked immunosorbent assay (ELISA). The endpoint titers were evaluated utilizing non-linear regression analysis. (O-P) Neutralizing antibodies (nAbs) within each serum were quantified via MERS-CoV pseudotyped virus (PsV) neutralization assay. The relative luciferase activity indicated the extent of MERS-CoV PsV infection (O). The half-maximal nAb titer (NT50) was defined as the reciprocal of the half-maximal neutralizing dilution (P). Data in (C-D) and (M-P) were presented as mean ± SEM. P values were determined by one-way ANOVA with Dunnett’s multiple comparisons tests. ns, not significant The freeze-thaw stability of HPF (I69C) nanoparticle vaccine against MERS-CoV. (A-B) MERS-CoV RBD-conjugated HPF and HPF (I69C) nanoparticle vaccines were subjected to a single freeze-thaw cycle (Freeze-Thaw×1, red) and five repeated freeze-thaw cycles (Freeze-Thaw×5, blue). Overlaid size distribution curves of both nanoparticle vaccines were evaluated by DLS. PDI values were indicated on the panels. (C-D) Median zeta potential of RBD (MERS)-HPF (C) and RBD (MERS)-HPF (I69C) (D) after Freeze-Thaw×1 and Freeze-Thaw×5. (E-F) SDS-PAGE analysis of both nanoparticle vaccines upon Freeze-Thaw×1 and Freeze-Thaw×5. (G-H) Densitometric quantification of SDS-PAGE bands in (E) and (F), respectively. Data were presented as proportions of intact proteins and degraded proteins relative to total lane intensity. (I-J) Native PAGE analysis of assembled nanoparticle vaccines after different freeze-thaw cycles. The red dashed box highlighted aggregates of RBD (MERS)-HPF nanoparticles. (K-L) Densitometric quantification of native PAGE bands in (I) and (J), respectively. Data were presented as proportions of intact nanoparticles and abnormal aggregates relative to total lane intensity. (M-N) Approximately 10 µg of RBD (MERS)-HPF and RBD (MERS)-HPF (I69C) subjected to 1 or 5 freeze-thaw cycles were administered to C57BL/6 mice as a single immunization. The serum samples were collected at 4 weeks post-immunization. The MERS-CoV RBD-specific binding IgG antibodies in serum were evaluated via enzyme-linked immunosorbent assay (ELISA). The endpoint titers were evaluated utilizing non-linear regression analysis. (O-P) Neutralizing antibodies (nAbs) within each serum were quantified via MERS-CoV pseudotyped virus (PsV) neutralization assay. The relative luciferase activity indicated the extent of MERS-CoV PsV infection (O). The half-maximal nAb titer (NT50) was defined as the reciprocal of the half-maximal neutralizing dilution (P). Data in (C-D) and (M-P) were presented as mean ± SEM. P values were determined by one-way ANOVA with Dunnett’s multiple comparisons tests. ns, not significant Subsequently, we evaluated the immunogenicity of both nanoparticles upon various freeze-thaw treatments. These nanoparticle vaccines were subcutaneously administered to C57BL/6 mice as a single immunization, followed by euthanasia to harvest serum samples after 4 weeks. All vaccines used in this study were consistently co-formulated with aluminum hydroxide (alum) adjuvant at a 1:1 volume ratio (50 µL alum plus 50 µL vaccine solution) immediately before each immunization. Antigen-specific antibodies and virus-specific neutralizing antibodies (nAbs) within each serum were quantified by the enzyme linked immunosorbent assay (ELISA) and pseudotyped virus (PsV) neutralization assay, respectively. The ELISA results showed that both RBD (MERS)-HPF and RBD (MERS)-HPF (I69C) nanoparticle vaccines, following either one-cycle or five-cycle freeze-thaw treatment, elicited high titers of MERS-CoV RBD-specific IgG antibodies, with endpoint titers exceeding 103 (Fig. 3M and N). Similarly, these nanoparticle vaccines also induced high titers of nAbs capable of neutralizing MERS-CoV PsV, with half-maximal nAb titers (NT50) approximately 103 (Fig. 3O and P). These results demonstrated that antigen-conjugated HPF (I69C) nanoparticle vaccines exhibited enhanced freeze-thaw stability, which consequently preserved their immunogenicity. HPF (I69C) exhibited enhanced thermal stability upon conjugating antigen We next evaluated the thermal stability of RBD (MERS)-conjugated HPF (WT) and HPF (I69C) under conditions of short-term deviations from the recommended storage temperature and duration. The DLS results demonstrated that the average diameters of both RBD (MERS) nanoparticles stably maintained 25 nm under storage conditions of 4 °C, 25 °C, and 37 °C over durations of 1 h, 7 days, and 14 days (Fig. 4A and B). Nonetheless, the PDI values gradually increased under storage conditions characterized by higher temperatures and extended durations, indicating broader size distributions. The polydispersity of RBD (MERS)-HPF increased more significantly at 37 °C for 7 and 14 days compared to that of RBD (MERS)-HPF (I69C), with PDI values exceeding 0.4. RBD (MERS)-HPF (I69C) nanoparticles still maintained low polydispersity under 37 °C for 7 days. However, the PDI increased to above 0.2 upon storing at 37 °C for 14 days. Storage of RBD (MERS)-HPF nanoparticles at 37 °C for 7 and 14 days also resulted in an increase in the net charge zeta potential, with a value of exceeding − 20 mV (Fig. 4C). In contrast, the increase in net charge zeta potential for RBD (MERS)-HPF (I69C) was observed only after storage at 37 °C for 14 days (Fig. 4D). We utilized both SDS-PAGE and native PAGE to evaluate the stability and homogeneity of both nanoparticles at various conditions. Consistently, storage for longer times at higher temperatures led to the degradation of nanoparticles, although the native PAGE showed that the remaining nanoparticles maintained homogeneity (Fig. 4E and L). The degradation of RBD (MERS)-HPF at 37 °C was more significant compared to that of RBD (MERS)-HPF (I69C). RBD (MERS)-HPF (I69C) nanoparticles still maintained thermal stability and homogeneity at 37 °C for 7 days (Fig. 4F, H, J, and L). Fig. 4The thermal stability of HPF (I69C) nanoparticle vaccine against MERS-CoV. (A-B) MERS-CoV RBD-conjugated HPF and HPF (I69C) nanoparticle vaccines were subjected to different storage conditions, which included 25 °C for 1 h (h), 4 °C for 7 days (D), 4 °C for 14 D, 25 °C for 7 D, 25 °C for 14 D, 37 °C for 7 D, and 37 °C for 14 D. Size distributions of both RBD (MERS)-HPF (A) and RBD (MERS)-HPF (I69C) (B) after various storage conditions were assessed by DLS. PDI values were shown on each panel. (C-D) Median zeta potentials of RBD (MERS)-HPF (C) and RBD (MERS)-HPF (I69C) (D) across various conditions were presented as bar charts. (E-F) SDS-PAGE analysis of both nanoparticle vaccines subjected to various storage conditions. Red dashed boxes highlighted degraded proteins. (G-H) Densitometric quantification of SDS-PAGE bands in (E) and (F), respectively. Data were presented as proportions of intact proteins and degraded proteins relative to total lane intensity for each condition. (I-J) Native PAGE analysis of nanoparticle vaccines after storage at different conditions. The protein bands with high molecular weight corresponded to intact 24-mer nanoparticles. (K-L) Densitometric quantification of intact nanoparticles and abnormal aggregates as demonstrated by native PAGE in (I) and (J). (M-N) About 10 µg of RBD (MERS)-HPF (I69C) nanoparticle vaccines stored at various conditions were administered to C57BL/6 mice as single immunization, followed by the serum collection at 4 weeks post-vaccination. The endpoint titers of MERS-CoV RBD-specific binding IgG antibodies in serum were evaluated via ELISA. (O-P) RBD (MERS)-HPF (I69C) nanoparticle vaccines were administered as described for (M). NAbs from each serum were quantified via MERS-CoV PsV neutralization assay (O). The NT50 was defined as the reciprocal of half-maximal neutralizing dilution (P). Data in (C-D) and (M-P) were presented as mean ± SEM. P values were determined by one-way ANOVA with Dunnett’s multiple comparisons tests. ns, not significant The thermal stability of HPF (I69C) nanoparticle vaccine against MERS-CoV. (A-B) MERS-CoV RBD-conjugated HPF and HPF (I69C) nanoparticle vaccines were subjected to different storage conditions, which included 25 °C for 1 h (h), 4 °C for 7 days (D), 4 °C for 14 D, 25 °C for 7 D, 25 °C for 14 D, 37 °C for 7 D, and 37 °C for 14 D. Size distributions of both RBD (MERS)-HPF (A) and RBD (MERS)-HPF (I69C) (B) after various storage conditions were assessed by DLS. PDI values were shown on each panel. (C-D) Median zeta potentials of RBD (MERS)-HPF (C) and RBD (MERS)-HPF (I69C) (D) across various conditions were presented as bar charts. (E-F) SDS-PAGE analysis of both nanoparticle vaccines subjected to various storage conditions. Red dashed boxes highlighted degraded proteins. (G-H) Densitometric quantification of SDS-PAGE bands in (E) and (F), respectively. Data were presented as proportions of intact proteins and degraded proteins relative to total lane intensity for each condition. (I-J) Native PAGE analysis of nanoparticle vaccines after storage at different conditions. The protein bands with high molecular weight corresponded to intact 24-mer nanoparticles. (K-L) Densitometric quantification of intact nanoparticles and abnormal aggregates as demonstrated by native PAGE in (I) and (J). (M-N) About 10 µg of RBD (MERS)-HPF (I69C) nanoparticle vaccines stored at various conditions were administered to C57BL/6 mice as single immunization, followed by the serum collection at 4 weeks post-vaccination. The endpoint titers of MERS-CoV RBD-specific binding IgG antibodies in serum were evaluated via ELISA. (O-P) RBD (MERS)-HPF (I69C) nanoparticle vaccines were administered as described for (M). NAbs from each serum were quantified via MERS-CoV PsV neutralization assay (O). The NT50 was defined as the reciprocal of half-maximal neutralizing dilution (P). Data in (C-D) and (M-P) were presented as mean ± SEM. P values were determined by one-way ANOVA with Dunnett’s multiple comparisons tests. ns, not significant The above results have demonstrated the superior thermal stability of engineered HPF (I69C) upon conjugating antigens. We subsequently assessed the immunogenicity of RBD (MERS)-HPF (I69C) nanoparticles when stored under varying conditions. C57BL/6 mice received a single subcutaneous immunization of 10 µg RBD (MERS)-HPF (I69C) nanoparticle vaccines, followed by euthanasia and serum harvesting at 4 weeks post-immunization. Serum from each mouse was subjected to both ELISA and PsV neutralization assay. The ELISA results revealed that RBD (MERS)-HPF (I69C) nanoparticle vaccines stored at various conditions generally induced high titers of MERS-CoV RBD-specific IgG antibodies, with endpoint titers over 103 (Fig. 4M and N). No significant differences were observed in titers induced by vaccines stored under different conditions. Consistently, antibodies induced by these vaccines also efficiently neutralized MERS-CoV PsV, with NT50 around 103 (Fig. 4O and P). Altogether, these results demonstrated that the RBD (MERS)-HPF (I69C) nanoparticle vaccine exhibited high thermal stability, thereby preserving its immunogenicity. Notably, although the immunogenicity of the nanoparticle vaccine remained consistently high after being stored at 37 °C for 14 days, we recommend against storing vaccines under such conditions due to the potential degradation of vaccines. HPF (I69C) nanoparticle vaccine conferred enhanced immunogenicity independent of anti-scaffold immunity To evaluate whether the pre-existing immunity against the scaffold or conjugation domains would interfere with nanoparticle vaccine boosting, C57BL/6 mice were primed with alum (Adjuvant), ST003-SC003 (STSC), HPF (I69C) (Core), RBD (MERS) monomer (Mono), or RBD (MERS)-HPF (I69C) (Nano) at week 0. All groups were subsequently boosted with RBD (MERS)-HPF (I69C) (Nano) at week 4, followed by euthanasia at week 6 (Fig. 5A). ELISA analysis of week 6 serum revealed that priming with alum, ST003-SC003, or HPF (I69C) did not impair the RBD (MERS)-specific antibody response to the nanoparticle boost, with all three groups achieving comparable endpoint titers of approximately 103 (Fig. 5B). Priming with RBD (MERS) monomer yielded higher titers, while two doses of RBD (MERS)-HPF (I69C) elicited the highest anti-RBD (MERS) IgG titers. Notably, the two-dose nanoparticle regimen induced RBD-specific antibody titers exceeding 103 as early as week 2, reaching peak titers above 104 at week 6 (Fig. 5C). MERS-CoV PsV neutralization assays using week 6 sera yielded results consistent with ELISA findings, with NT50 titers reaching approximately 103 in the alum-, ST003-SC003-, and HPF (I69C)-primed groups (Fig. 5D). These results demonstrated that the pre-existing immunity against the scaffold or conjugation domains did not detectably interfere with subsequent nanoparticle vaccine boosting. Fig. 5The impact of optimized scaffold on the immunogenicity of vaccine. (A) Schematic of the prime-boost vaccination groups. C57BL/6 mice were primed at week 0 with alum (Adjuvant), ST003-SC003 (STSC), HPF (I69C) (Core), RBD (MERS) monomer (Mono), or RBD (MERS)-HPF (I69C) (Nano), and boosted with RBD (MERS)-HPF (I69C) (Nano) at week 4. Serum was collected at weeks 0, 2, 4, and 6. (B) Endpoint titers of MERS-CoV RBD-specific IgG antibodies in week 6 serum from the corresponding groups in (A). (C) Time-course analysis of MERS-CoV RBD-specific IgG titers in each group at weeks 0, 2, 4, and 6. (D) The NT50 titers of MERS-CoV PsV neutralization using week 6 serum from the corresponding groups in (A) were shown. (E) Schematic table of the vaccination regimen comparing HPF (WT)-based (WT-Nano) and HPF (I69C)-based (I69C-Nano) RBD (MERS) nanoparticle vaccines. C57BL/6 mice were immunized with 5–10 µg of each vaccine at weeks 0 (prime) and 4 (boost). Serum was collected at weeks 0, 2, 4, and 6. (F) Endpoint titers of MERS-CoV RBD-specific IgG antibodies in week 6 serum from mice immunized with 5–10 µg of WT-Nano or I69C-Nano vaccines. (G) Time-course analysis of MERS-CoV RBD-specific IgG titers at weeks 0, 2, 4, and 6. (H) Endpoint titers of HPF and HPF (I69C) scaffold-specific IgG antibodies in week 6 serum. (I) Time-course analysis of HPF and HPF (I69C) scaffold-specific IgG titers at weeks 0, 2, 4, and 6. (J) NT50 titers from MERS-CoV PsV neutralization assays using week 6 serum. (K) NAbs against authentic MERS-CoV (ChinaGD01) using week 6 serum were evaluated via the focus reduction neutralization test (FRNT). The FRNT50 titer was defined as the reciprocal of the serum dilution at which nAbs achieved 50% inhibition of viral infection. Data in (B-D) and (F-K) were presented as mean ± SEM. P values were determined by one-way ANOVA with Tukey’s multiple comparisons tests. ns, not significant. * P < 0.05, *** P < 0.001 The impact of optimized scaffold on the immunogenicity of vaccine. (A) Schematic of the prime-boost vaccination groups. C57BL/6 mice were primed at week 0 with alum (Adjuvant), ST003-SC003 (STSC), HPF (I69C) (Core), RBD (MERS) monomer (Mono), or RBD (MERS)-HPF (I69C) (Nano), and boosted with RBD (MERS)-HPF (I69C) (Nano) at week 4. Serum was collected at weeks 0, 2, 4, and 6. (B) Endpoint titers of MERS-CoV RBD-specific IgG antibodies in week 6 serum from the corresponding groups in (A). (C) Time-course analysis of MERS-CoV RBD-specific IgG titers in each group at weeks 0, 2, 4, and 6. (D) The NT50 titers of MERS-CoV PsV neutralization using week 6 serum from the corresponding groups in (A) were shown. (E) Schematic table of the vaccination regimen comparing HPF (WT)-based (WT-Nano) and HPF (I69C)-based (I69C-Nano) RBD (MERS) nanoparticle vaccines. C57BL/6 mice were immunized with 5–10 µg of each vaccine at weeks 0 (prime) and 4 (boost). Serum was collected at weeks 0, 2, 4, and 6. (F) Endpoint titers of MERS-CoV RBD-specific IgG antibodies in week 6 serum from mice immunized with 5–10 µg of WT-Nano or I69C-Nano vaccines. (G) Time-course analysis of MERS-CoV RBD-specific IgG titers at weeks 0, 2, 4, and 6. (H) Endpoint titers of HPF and HPF (I69C) scaffold-specific IgG antibodies in week 6 serum. (I) Time-course analysis of HPF and HPF (I69C) scaffold-specific IgG titers at weeks 0, 2, 4, and 6. (J) NT50 titers from MERS-CoV PsV neutralization assays using week 6 serum. (K) NAbs against authentic MERS-CoV (ChinaGD01) using week 6 serum were evaluated via the focus reduction neutralization test (FRNT). The FRNT50 titer was defined as the reciprocal of the serum dilution at which nAbs achieved 50% inhibition of viral infection. Data in (B-D) and (F-K) were presented as mean ± SEM. P values were determined by one-way ANOVA with Tukey’s multiple comparisons tests. ns, not significant. * P < 0.05, *** P < 0.001 To directly quantify the contribution of scaffold stabilization to vaccine immunogenicity, C57BL/6 mice were immunized with 5–10 µg of either HPF (WT)-based (WT-Nano) or HPF (I69C)-based (I69C-Nano) RBD (MERS) nanoparticle vaccines in a prime-boost regimen (Fig. 5E). ELISA analysis of week 6 serum showed that I69C-Nano induced significantly higher MERS-CoV RBD-specific IgG titers than WT-Nano at both doses (Fig. 5F). Time-course analysis revealed that the 10 µg dose elicited higher titers than the 5 µg dose at weeks 2 and 4 within each vaccine group, whereas a significant scaffold-dependent advantage of I69C-Nano over WT-Nano emerged at week 6 (Fig. 5G). Notably, the 10 µg dose did not further increase antibody titers compared to the 5 µg dose at week 6 within each vaccine group, suggesting that both doses could saturate the B-cell response capacity in this model. Anti-scaffold antibody responses were comparable between WT-Nano and I69C-Nano at both doses, confirming that the enhanced immunogenicity of I69C-Nano originated from structural stabilization rather than increased scaffold immunogenicity (Fig. 5H and I). Furthermore, I69C-Nano elicited significantly higher neutralizing antibody titers than WT-Nano against both pseudotyped MERS-CoV and authentic MERS-CoV (ChinaGD01) (Fig. 5J and K). Taken together, the enhanced immunogenicity of the HPF (I69C) nanoparticle vaccine was attributable to scaffold stabilization conferred by the I69C mutation, rather than to differences in scaffold-directed antibody responses. Construction of HPF (I69C) nanoparticle vaccine against SARS-CoV-2 and MERS-CoV Due to the continued prevalence of severe acute respiratory syndrome coronavirus 2 (SARS‑CoV‑2) and the epidemic spread of MERS-CoV, we employed the HPF (I69C) as an inner core to construct bivalent nanoparticle vaccines targeting both viruses. The ST003 tag was expressed in-frame with RBD antigens of both SARS-CoV-2 and MERS-CoV, followed by their respective covalent conjugation to SC003-HPF (I69C). Therefore, the first constructed bivalent nanoparticle vaccine was a mixture of both RBD (SARS2)-HPF (I69C) and RBD (MERS)-HPF (I69C) (Fig. 6A and E). Simultaneously, we constructed a dimeric RBD antigen tagged with ST003, which contained in-frame fused RBDs from both viruses (Fig. 6A). The finalized second bivalent nanoparticle vaccine displayed the RBD (Dimer) on a single HPF (I69C) particle (Fig. 6C and F). ST003-tagged RBD antigens were purified from HEK293F cells and subjected to SEC for further purification and assessment of homogeneity. Using a Superdex 200 Increase column, SEC profiles showed that RBD (SARS2)-ST003, RBD (MERS)-ST003, and RBD (Dimer)-ST003 were eluted as single peaks with retention volumes of 14–18 mL, 15–17 mL, and 11–16 mL, respectively (Fig. 6G). These purified antigens were conjugated with equal moles of SC003-HPF (I69C) via ST003-SC003 conjugation. The SEC analysis using a Superose 6 Increase column demonstrated that SC003-HPF (I69C) nanoparticles were eluted at retention volumes of 10–14 mL as a single peak (Fig. 6H). Conjugation with corresponding antigens led to a shift toward earlier elution, with nanoparticles eluting at retention volumes of 8–13 mL (Fig. 6H). Both SDS-PAGE and native PAGE analyses revealed that SC003-HPF (I69C), the mixture of RBD (SARS2)-HPF (I69C) and RBD (MERS)-HPF (I69C), and RBD (Dimer)-HPF (I69C) migrated as single protein bands, demonstrating their high purity and homogeneity (Fig. 6I and J). Fig. 6The construction of bivalent HPF (I69C) nanoparticle vaccines against SARS-CoV-2 and MERS-CoV. (A-C) Schematics of vaccine constructs used in this study. SARS-CoV-2 RBD, MERS-CoV RBD, and dimeric RBD from both SARS-CoV-2 and MERS-CoV were expressed in-frame with ST003 tag (A), while SC003 was fused to the nanoparticle scaffold HPF (I69C) (B). ST003-tagged RBD (SARS2), RBD (MERS), and RBD (Dimer) were conjugated to SC003-HPF (I69C), respectively (C). The isopeptide bond icon indicated the covalent SpyTag/SpyCatcher linkage. (D-F) Surface renderings of SC003-HPF (I69C) (D), RBD (SARS2)-HPF (I69C), RBD (MERS)-HPF (I69C) (E), and RBD (Dimer)-HPF (I69C) (F). HPF (I69C) nano cores were colored in gray. RBD (SARS2) antigens were colored in blue, while RBD (MERS) antigens were colored in orange. (G) SEC profiles (A280) of RBD (SARS2)-ST003, RBD (MERS)-ST003, and RBD (Dimer)-ST003 with a Superdex 200 Increase column. Peaks corresponded to monomeric and dimeric antigens. (H) SEC profiles (A280) of SC003-HPF (I69C), RBD (SARS2)-HPF (I69C), RBD (MERS)-HPF (I69C), and RBD (Dimer)-HPF (I69C) with a Superose 6 Increase column. (I) SDS-PAGE analysis of SC003-HPF (I69C) and RBD-conjugated nanoparticles. Bands at expected monomer sizes indicated the presence of corresponding components. Shifted bands with higher molecular weight indicated covalent ST003/SC003 conjugates. (J) Native PAGE analysis of assembled nanoparticles. Lanes corresponded to SC003-HPF (I69C) scaffold alone and RBD-conjugated nanoparticles. Mobility shifts confirmed the formation of self-assembled particles after conjugation. (K) Morphologies of SC003-HPF (I69C) particles, monomeric RBD nanoparticles, and dimeric RBD nanoparticles were characterized by negative staining. SC003-HPF (I69C) particles displayed hollow spherical structures, while RBD-conjugated particles exhibited crown-like structures. (L-M) Biolayer interferometry (BLI) analysis of recombinant hACE2 binding to RBD (SARS2)-HPF (I69C) (L) and RBD (Dimer)-HPF (I69C) (M) immobilized on AR2G biosensors. Equilibrium dissociation constants (KD) were derived from global fitting to a 1:1 binding model. (N-O) BLI analysis of recombinant hDPP4 binding to RBD (MERS)-HPF (I69C) (N) and RBD (Dimer)-HPF (I69C) (O). Reported KD values for nanoparticle-displayed RBDs reflected avidity-enhanced binding The construction of bivalent HPF (I69C) nanoparticle vaccines against SARS-CoV-2 and MERS-CoV. (A-C) Schematics of vaccine constructs used in this study. SARS-CoV-2 RBD, MERS-CoV RBD, and dimeric RBD from both SARS-CoV-2 and MERS-CoV were expressed in-frame with ST003 tag (A), while SC003 was fused to the nanoparticle scaffold HPF (I69C) (B). ST003-tagged RBD (SARS2), RBD (MERS), and RBD (Dimer) were conjugated to SC003-HPF (I69C), respectively (C). The isopeptide bond icon indicated the covalent SpyTag/SpyCatcher linkage. (D-F) Surface renderings of SC003-HPF (I69C) (D), RBD (SARS2)-HPF (I69C), RBD (MERS)-HPF (I69C) (E), and RBD (Dimer)-HPF (I69C) (F). HPF (I69C) nano cores were colored in gray. RBD (SARS2) antigens were colored in blue, while RBD (MERS) antigens were colored in orange. (G) SEC profiles (A280) of RBD (SARS2)-ST003, RBD (MERS)-ST003, and RBD (Dimer)-ST003 with a Superdex 200 Increase column. Peaks corresponded to monomeric and dimeric antigens. (H) SEC profiles (A280) of SC003-HPF (I69C), RBD (SARS2)-HPF (I69C), RBD (MERS)-HPF (I69C), and RBD (Dimer)-HPF (I69C) with a Superose 6 Increase column. (I) SDS-PAGE analysis of SC003-HPF (I69C) and RBD-conjugated nanoparticles. Bands at expected monomer sizes indicated the presence of corresponding components. Shifted bands with higher molecular weight indicated covalent ST003/SC003 conjugates. (J) Native PAGE analysis of assembled nanoparticles. Lanes corresponded to SC003-HPF (I69C) scaffold alone and RBD-conjugated nanoparticles. Mobility shifts confirmed the formation of self-assembled particles after conjugation. (K) Morphologies of SC003-HPF (I69C) particles, monomeric RBD nanoparticles, and dimeric RBD nanoparticles were characterized by negative staining. SC003-HPF (I69C) particles displayed hollow spherical structures, while RBD-conjugated particles exhibited crown-like structures. (L-M) Biolayer interferometry (BLI) analysis of recombinant hACE2 binding to RBD (SARS2)-HPF (I69C) (L) and RBD (Dimer)-HPF (I69C) (M) immobilized on AR2G biosensors. Equilibrium dissociation constants (KD) were derived from global fitting to a 1:1 binding model. (N-O) BLI analysis of recombinant hDPP4 binding to RBD (MERS)-HPF (I69C) (N) and RBD (Dimer)-HPF (I69C) (O). Reported KD values for nanoparticle-displayed RBDs reflected avidity-enhanced binding Because of the flexible linkers between antigens and nanoparticles, we were unable to determine the cryo-EM structures of these nanoparticle vaccines. However, the negative staining results confirmed the monodispersity and structural homogeneity of RBD-conjugated nanoparticles (Fig. 6K). Notably, SC003-HPF (I69C) nanoparticles exhibited hollow spherical structures, whereas monomeric RBD- and dimeric RBD-conjugated nanoparticles displayed crown-like morphologies, indicating the successful surface arrangement of antigens on nanoparticles. To verify that RBD antigens retained functional receptor-binding capacity and neutralizing epitope accessibility following conjugation to HPF (I69C), we performed biolayer interferometry (BLI) assays measuring binding to recombinant hACE2, hDPP4, and the clinical-grade neutralizing monoclonal antibody REGN10933 (casirivimab) [43]. Recombinant hACE2 bound to RBD (SARS2) and RBD (Dimer) with equilibrium dissociation constant (KD) values of 2.54 nM and 6.66 nM, respectively (Fig. S6A and S6B). Following conjugation, hACE2 bound to RBD (SARS2)-HPF (I69C) and RBD (Dimer)-HPF (I69C) with comparable KD values of 7.68 nM and 8.95 nM, indicating that nanoparticle conjugation did not occlude the receptor-binding motif of SARS-CoV-2 RBD (Fig. 6L and M). Recombinant hDPP4 bound to RBD (MERS) and RBD (Dimer) with KD values of 65.6 nM and 81.3 nM, whereas binding to RBD (MERS)-HPF (I69C) and RBD (Dimer)-HPF (I69C) was markedly enhanced, with KD values of 3.66 nM and 3.03 nM, representing 17.9- and 26.8-fold improvements, respectively (Fig. 6N and O and Fig. S6C-S6D). BLI analysis using REGN10933 (casirivimab), which targets the receptor-binding motif of SARS-CoV-2 RBD, revealed that RBD (SARS2) and RBD (Dimer) bound REGN10933 with KD values of 4.29 nM and 9.3 nM, whereas RBD (SARS2)-HPF (I69C) and RBD (Dimer)-HPF (I69C) exhibited substantially enhanced binding, with KD values of < 0.001 nM and 0.53 nM (Fig. S6E-S6H). The markedly enhanced binding of nanoparticle-displayed RBD to hDPP4 and REGN10933 was consistent with avidity effects arising from multivalent presentation of correctly oriented antigens on HPF (I69C) nanoparticles. Collectively, these results demonstrated that RBD antigens conjugated to HPF (I69C) nanoparticles retained a functionally native conformation with preserved receptor-binding capacity and neutralizing epitope accessibility. Multivalent display by HPF (I69C) nanoparticles conferred avidity-enhanced binding to hDPP4 and REGN10933, while preserving steric accessibility of the receptor-binding motif for hACE2. HPF (I69C) nanoparticle vaccines demonstrated robust immunogenicity To evaluate the immunogenicity and safety of HPF (I69C) nanoparticle vaccines, we firstly confirmed that endotoxin levels in all E. coli-produced nanoparticle preparations were below 1.25 EU/mL, falling under the pharmacopoeial limit of 5 EU per kg of body weight per hour (corresponding to 1.25–1.5 EU/mL for the mice used in this study) (Fig. S6I). C57BL/6 mice were then immunized with equal moles of vaccines adjuvanted with alum. Vaccine groups were normalized on an equal total RBD mole basis. The adjuvant-only group and the nanoparticle core (nano core) group received equal volumes of alum and equal moles of SC003-HPF (I69C), respectively (Fig. 7A). The RBD monomer mixture (mono mix) group was immunized with RBD (SARS2) and RBD (MERS), whereas mice from the dimer group received a dimeric RBD (SARS2)-RBD (MERS). Another two groups of mice were immunized with two bivalent nanoparticle vaccines. In the monomer-nanoparticle mixture (Mono_Nano Mix) group, mice received equal moles of RBD (SARS2)-HPF (I69C) and RBD (MERS)-HPF (I69C), while the dimer nanoparticle (Dimer_Nano) group was vaccinated with RBD (Dimer)-HPF (I69C). The vaccine used in Dimer_Nano group was 10 µg of RBD (Dimer)-HPF (I69C), whereas all the other groups received formulations containing equimolar amounts equivalent to the molar quantity of 10 µg RBD (Dimer)-HPF (I69C). All mice were subcutaneously immunized twice, with the prime vaccination at week 0 and boost vaccination at week 4 (Fig. 7B). Serum of each mouse was collected every 2 weeks. Mice were euthanized at week 6, followed by the harvesting of mouse tissues for further analysis. The ELISA results of serum from week 6 revealed that both bivalent nanoparticle vaccines elicited higher titers of SARS-CoV-2 RBD- and MERS-CoV RBD-specific IgG antibodies compared to the monomer mixture and dimer, with endpoint titers higher than 104 (Fig. 7C and D). Notably, although the nanoparticles also induced high titers of HPF (I69C)- and ST003-SC003-specific antibodies, the production of RBD-specific antibodies remained unaffected (Fig. 7E and F). Additionally, we measured RBD-specific antibodies at week 2 and week 4 post-vaccination. The results showed that titers of both SARS-CoV-2 RBD- and MERS-CoV RBD-specific antibodies reached 103 as early as 2 weeks after the prime vaccination, which were significantly higher than those induced by the monomer mixture and dimer groups (Fig. 7G and J). A second vaccination further boosted the levels of RBD-specific antibodies. The above results indicated that the bivalent HPF (I69C) nanoparticle vaccines exhibited significantly enhanced immunogenicity against both SARS-CoV-2 and MERS-CoV compared to monomeric and dimeric vaccines. Fig. 7The immunogenicity and safety of HPF (I69C) nanoparticle vaccines. (A) Study groups for mouse vaccination. Group 1 (G1) represented the alum-only adjuvant group. The SC003-HPF (I69C)-formulated G2 was the nanoparticle core (Nano Core) group. G3 represented the mixture of RBD (SARS2) and RBD (MERS) monomers (Mono Mix). G4 was dimeric RBD (SARS2)-RBD (MERS). G5 received the mixture of both RBD (SARS2)-HPF (I69C) and RBD (MERS)-HPF (I69C) nanoparticles (Mono_Nano Mix), while G6 represented RBD (Dimer)-HPF (I69C) nanoparticles (Dimer_Nano). (B) Schematic of vaccination timeline in C57BL/6 mice. Each mouse was vaccinated at week 0 (prime) and at week 4 (boost), followed by euthanasia at week 6. Serum was collected every two weeks for antibody quantification. Systemic safety examinations were performed at week 6. (C-F) Endpoint titers of SARS-CoV-2 RBD-, MERS-CoV RBD-, HPF (I69C)-, and ST003-SC003-specific antibodies were quantified via ELISA using week 6 serum samples. (G-J) Anti-SARS-CoV-2 RBD, anti-MERS-CoV RBD, anti-HPF (I69C), and anti-ST003-SC003 titers at each time point (week 0, 2, 4, and 6) were determined by ELISA and presented as time-course curves. (K-O) Complete blood counts (CBCs) including concentrations of WBC, PLT, RBC, HGB, and HCT levels were analyzed utilizing week 6 serum samples. (P-W) Serum biochemical analysis of ALT, AST, ALP, TBIL, γ-GT, TBA, UREA, and CREA levels was performed on mouse serum collected at week 6. The normal reference ranges of each parameter were indicated in methods. Data in (C-W) were presented as mean ± SEM. P values in (C-F) were calculated by one-way ANOVA with Tukey’s multiple comparisons tests. ** P < 0.01, *** P < 0.001 The immunogenicity and safety of HPF (I69C) nanoparticle vaccines. (A) Study groups for mouse vaccination. Group 1 (G1) represented the alum-only adjuvant group. The SC003-HPF (I69C)-formulated G2 was the nanoparticle core (Nano Core) group. G3 represented the mixture of RBD (SARS2) and RBD (MERS) monomers (Mono Mix). G4 was dimeric RBD (SARS2)-RBD (MERS). G5 received the mixture of both RBD (SARS2)-HPF (I69C) and RBD (MERS)-HPF (I69C) nanoparticles (Mono_Nano Mix), while G6 represented RBD (Dimer)-HPF (I69C) nanoparticles (Dimer_Nano). (B) Schematic of vaccination timeline in C57BL/6 mice. Each mouse was vaccinated at week 0 (prime) and at week 4 (boost), followed by euthanasia at week 6. Serum was collected every two weeks for antibody quantification. Systemic safety examinations were performed at week 6. (C-F) Endpoint titers of SARS-CoV-2 RBD-, MERS-CoV RBD-, HPF (I69C)-, and ST003-SC003-specific antibodies were quantified via ELISA using week 6 serum samples. (G-J) Anti-SARS-CoV-2 RBD, anti-MERS-CoV RBD, anti-HPF (I69C), and anti-ST003-SC003 titers at each time point (week 0, 2, 4, and 6) were determined by ELISA and presented as time-course curves. (K-O) Complete blood counts (CBCs) including concentrations of WBC, PLT, RBC, HGB, and HCT levels were analyzed utilizing week 6 serum samples. (P-W) Serum biochemical analysis of ALT, AST, ALP, TBIL, γ-GT, TBA, UREA, and CREA levels was performed on mouse serum collected at week 6. The normal reference ranges of each parameter were indicated in methods. Data in (C-W) were presented as mean ± SEM. P values in (C-F) were calculated by one-way ANOVA with Tukey’s multiple comparisons tests. ** P < 0.01, *** P < 0.001 To address the potential safety of HPF (I69C) nanoparticle vaccines, systemic examinations were performed on euthanized mice, including complete blood counts (CBCs), serum biochemical analyses, urinalysis, and histopathological evaluation. The CBC analysis revealed that all cell-related parameters, including concentrations of white blood cells (WBCs), platelets (PLTs), red blood cells (RBCs), hemoglobin (HGB), and hematocrit (HCT) levels, remained within the normal reference ranges for all vaccinated groups (Fig. 7K and O). These CBC results demonstrated that nanoparticle vaccines did not produce detectable adverse effects on the hematologic system or provoke significant systemic inflammatory responses. Additionally, we performed serum biochemical analysis for vaccinated mice. These analyses showed that all measured parameters, including concentrations of alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), total bilirubin (TBIL), gamma-glutamyl transferase (γ-GT), total bile acids (TBA), urea (UREA), and creatinine (CREA), were also within normal reference ranges in vaccinated mice (Fig. 7P and W). These quantifications indicated that HPF (I69C) nanoparticle vaccines did not induce detectable hepatobiliary and renal toxicity after two-dose vaccination. Furthermore, we harvested mouse urine for urinalysis tests. The dipstick and microscopic findings revealed that WBCs, blood (BLD), and nitrite (NIT) were negative in all samples, indicating no evidence of pyuria or urinary tract inflammation, hematuria, and bacterial infection, respectively (Fig. S7A). The metabolic markers, including ketones (KET), urobilinogen (URO), bilirubin (BIL), protein (PRO), and glucose (GLU), were either negative or trace in urine samples with no group-specific pattern, indicating no evidence of hepatobiliary dysfunction, hemolysis, or renal abnormalities attributable to vaccination. The other physicochemical measurements, including specific gravity (SG), pH levels, and CREA, were within normal reference ranges, indicating normal urinary concentrating ability and no evidence of impaired renal excretory function across all vaccinated mice. We also performed histopathological analyses for major organs isolated from vaccinated mice. The hematoxylin and eosin (H&E) results demonstrated no vaccine-related pathological changes in any of the examined organs, including myocardium, liver, spleen, lung, kidney, intestine, bladder, and cerebrum (Fig. S7B). Tissue morphology in the nanoparticle vaccine groups was comparable to the adjuvant control group, supporting the absence of acute or subacute organ toxicity after vaccination. Overall, our results indicated that HPF (I69C)-based bivalent nanoparticle vaccines targeting SARS-CoV-2 and MERS-CoV demonstrated robust immunogenicity and favorable safety profiles in vaccinated mice. HPF (I69C) nanoparticle vaccines elicited potent immune responses To verify whether HPF (I69C) nanoparticle vaccines elicited protective immune responses against both SARS-CoV-2 and MERS-CoV infection, we conducted both PsV neutralization assays and authentic virus neutralization assays. Results from PsV assays revealed that the mixture of RBD (SARS2)-HPF (I69C) and RBD (MERS)-HPF (I69C), as well as RBD (Dimer)-HPF (I69C) induced over 104 titers of nAbs targeting SARS-CoV-2 PsV, which were significantly higher than those induced by the monomer mixture and dimer vaccines (Fig. 8A). To evaluate neutralization breadth against SARS-CoV-2 variants, we further tested sera against Omicron sublineages BA.2, BA.5, XBB, and JN.1. Both nanoparticle vaccines elicited significantly higher NT50 titers against all four sublineages compared to monomeric and dimeric vaccines, with NT50 values ranging from 101 to 102 for BA.2, BA.5, and XBB, and exceeding 102 for JN.1 (Fig. S8A-S8D). Neutralization by control and subunit vaccine groups was nearly undetectable. These results indicated that multivalent nanoparticle presentation conferred measurable cross-neutralization breadth against antigenically divergent Omicron sublineages. Both nanoparticle vaccines also elicited higher titers of nAbs against MERS-CoV PsV compared to the monomeric and dimeric vaccines, with nAb titers ranging from 103 to 104 (Fig. 8B). Consistently, the bivalent nanoparticle vaccines also induced high titers of nAbs against the infection of both authentic SARS-CoV-2 (Wuhan-Hu-1) and MERS-CoV (ChinaGD01), with antibody titers reaching approximately 103 (Fig. 8C and D). NAbs-elicited by nanoparticle vaccines achieved nearly complete neutralization of authentic viruses at serum dilutions of 1:10 and 1:100 (Fig. 8E and F). However, both monomeric and dimeric vaccines only induced nAbs at titers ranging from 101 to 102 against authentic viral infection, with residual viral infection observed even at serum dilutions of 1:10 and 1:100 (Fig. 8C and F). Fig. 8The humoral and cellular immunity elicited by bivalent HPF (I69C) nanoparticle vaccines. (A-B) NAbs against SARS-CoV-2 and MERS-CoV PsVs were quantified via PsV neutralization assays. The NT50 titer of nAbs induced by each vaccine was defined as the reciprocal of half-maximal neutralizing dilution. (C-D) FRNT50 titers of nAbs against authentic SARS-CoV-2 (Wuhan-Hu-1) (C) and MERS-CoV (ChinaGD01) (D) were evaluated via the FRNT assay. (E-F) Representative 96-well images of FRNT assays for control and vaccine groups at the indicated serum dilutions of 1:10 and 1:100. Immunospots corresponding to authentic SARS-CoV-2 (Wuhan-Hu-1) (E) and MERS-CoV (ChinaGD01) (F) were detected with SARS-CoV-2 N- and MERS-CoV N-specific antibodies, respectively. The number of spots was shown alongside each well. (G-H) Percentages of antigen-specific B cells in each group were quantified by gating on the CD3−/B220+/RBD+ splenocytes. (I-J) Percentages of antigen-specific memory B cells (MBCs) were analyzed by gating on the CD3−/B220+/CD73+/CD80+/RBD+ splenocytes. (K-L) Splenic Tfh cells in vaccination groups were determined by FCM, wherein CD3+/CD4+/CXCR5+/PD-1+ splenocytes were gated. Percentages of Tfh cells were normalized relative to total CD4+ T cells (K), while absolute numbers of Tfh cells per million splenocytes were also recorded (L). (M-O) Antigen-specific polyfunctional CD8+ T cells were stimulated by incubating with RBD peptide pools, followed by intracellular cytokine staining (ICCS). The percentages of TNF-α+ (M), IFN-γ+ (N), and IL-2+ (O) CD8+ T cells were analyzed by FCM and normalized relative to total CD8+ T cells. (P-R) The percentages of TNF-α+ (P), IFN-γ+ (Q), and IL-2+ (R) CD4+ T cells within each vaccination group were analyzed as in (M). Data in (A-D) and (G-R) were presented as mean ± SEM. P values were determined via one-way ANOVA with Tukey’s multiple comparisons tests. * P < 0.05, ** P < 0.01, *** P < 0.001 The humoral and cellular immunity elicited by bivalent HPF (I69C) nanoparticle vaccines. (A-B) NAbs against SARS-CoV-2 and MERS-CoV PsVs were quantified via PsV neutralization assays. The NT50 titer of nAbs induced by each vaccine was defined as the reciprocal of half-maximal neutralizing dilution. (C-D) FRNT50 titers of nAbs against authentic SARS-CoV-2 (Wuhan-Hu-1) (C) and MERS-CoV (ChinaGD01) (D) were evaluated via the FRNT assay. (E-F) Representative 96-well images of FRNT assays for control and vaccine groups at the indicated serum dilutions of 1:10 and 1:100. Immunospots corresponding to authentic SARS-CoV-2 (Wuhan-Hu-1) (E) and MERS-CoV (ChinaGD01) (F) were detected with SARS-CoV-2 N- and MERS-CoV N-specific antibodies, respectively. The number of spots was shown alongside each well. (G-H) Percentages of antigen-specific B cells in each group were quantified by gating on the CD3−/B220+/RBD+ splenocytes. (I-J) Percentages of antigen-specific memory B cells (MBCs) were analyzed by gating on the CD3−/B220+/CD73+/CD80+/RBD+ splenocytes. (K-L) Splenic Tfh cells in vaccination groups were determined by FCM, wherein CD3+/CD4+/CXCR5+/PD-1+ splenocytes were gated. Percentages of Tfh cells were normalized relative to total CD4+ T cells (K), while absolute numbers of Tfh cells per million splenocytes were also recorded (L). (M-O) Antigen-specific polyfunctional CD8+ T cells were stimulated by incubating with RBD peptide pools, followed by intracellular cytokine staining (ICCS). The percentages of TNF-α+ (M), IFN-γ+ (N), and IL-2+ (O) CD8+ T cells were analyzed by FCM and normalized relative to total CD8+ T cells. (P-R) The percentages of TNF-α+ (P), IFN-γ+ (Q), and IL-2+ (R) CD4+ T cells within each vaccination group were analyzed as in (M). Data in (A-D) and (G-R) were presented as mean ± SEM. P values were determined via one-way ANOVA with Tukey’s multiple comparisons tests. * P < 0.05, ** P < 0.01, *** P < 0.001 To elucidate the immunological foundation of the potent adaptive immunity elicited by HPF (I69C) nanoparticle vaccines, we quantified antigen-specific B cells in isolated splenocytes (Fig. S9A). The flow cytometry (FCM) analysis revealed that both nanoparticle vaccines elicited significantly higher percentages of RBD (SARS2)- and RBD (MERS)-specific B cells compared to monomeric and dimeric vaccines, which was consistent with the higher nAb titers induced by the nanoparticle formulations (Fig. 8G and H). Additionally, nanoparticle vaccines also induced higher percentages of RBD-specific memory B cells (MBCs), although no statistically significant difference was observed between the dimeric nanoparticle vaccine and the dimer-only vaccine (Fig. 8I and J). The establishment of robust adaptive immunity relies on the cognate interaction between antigen-specific B cells and T follicular helper (Tfh) cells, which undergo substantial differentiation and proliferation to collectively promote B-cell affinity maturation [44]. Therefore, we also quantified the percentages of Tfh cells after two-dose vaccination (Fig. S9B). The FCM results showed that nanoparticle vaccines induced significantly higher percentages and absolute numbers of Tfh cells within the CD4+ T-cell population compared to both monomeric and dimeric subunit vaccines, demonstrating enhanced Tfh-cell signaling (Fig. 8K and L). RBD proteins of both SARS-CoV-2 and MERS-CoV contain multiple T-cell immunogenic epitopes [45, 46]. To determine the T-cell immune responses induced by vaccines, we quantified the proportions of antigen-specific CD4+ and CD8+ T cells following exposure to RBD peptide pools (Fig. S9C). Our results demonstrated that compared to monomeric and dimeric vaccines, nanoparticle vaccines elicited higher percentages of antigen-specific polyfunctional CD8+ T cells, as evidenced by higher percentages of TNF-α+, IFN-γ+, and IL-2+ CD8+ T cells (Fig. 8M and O). Furthermore, nanoparticle vaccines also induced greater proportions of antigen-specific TNF-α+, IFN-γ+, and IL-2+ CD4+ T cells, demonstrating enhanced T-cell responses, particularly Th1-biased immune profiles, which were associated with improved protection against intracellular viruses and stronger activation of cell-mediated immunity (Fig. 8P and R). Taken together, the above results indicated that HPF (I69C)-based bivalent nanoparticle vaccines elicited potent immune responses against the infection of both SARS-CoV-2 and MERS-CoV. HPF (I69C) nanoparticle vaccines demonstrated protective efficacy against authentic coronavirus challenges Due to the potent immune responses elicited by HPF (I69C) nanoparticle vaccines, we hypothesized that HPF (I69C)-based bivalent nanoparticle vaccines might protect humanized mice from authentic coronavirus infections. hACE2 or hDPP4 transgenic mice were randomly assigned into six groups, including alum adjuvant only (G1), SC003-HPF (I69C) Nano Core (G2), Mono Mix of RBD (SARS2) and RBD (MERS) (G3), Dimer of RBD (SARS2)-RBD (MERS) (G4), Mono_Nano Mix of RBD (SARS2)-HPF (I69C) and RBD (MERS)-HPF (I69C) (G5), and Dimer_Nano of RBD (Dimer)-HPF (I69C) (G6) (Fig. 9A). All mice were prime (week 0) and boost (week 4) vaccinated in the same manner as the C57BL/6 mice. One week after the boost vaccination, hACE2 mice and hDPP4 mice were intranasally challenged with 5 × 104 focus-forming units (FFU) of authentic SARS-CoV-2 (Wuhan-Hu-1) and 5 × 103 FFU of authentic MERS-CoV (ChinaGD01), respectively (Fig. 9B). The body weight of each mouse was monitored every day post-infection. All mice were euthanized one week after the challenge. Subsequently, lung tissues were collected for the quantification of viral titers, H&E staining, and IHC analyses. Fig. 9The protective efficacy of bivalent HPF (I69C) nanoparticle vaccines. (A) hACE2 mice or hDPP4 mice were randomly allocated to six groups with similar initial average body weights. All mice were vaccinated as in Fig. 7A. G1: alum-only adjuvant. G2: SC003-HPF (I69C)-formulated Nano Core. G3: Mono Mix of RBD (SARS2) and RBD (MERS). G4: Dimer of RBD (SARS2)-RBD (MERS). G5: Mono_Nano Mix of RBD (SARS2)-HPF (I69C) and RBD (MERS)-HPF (I69C). G6: Dimer_Nano of RBD (Dimer)-HPF (I69C). (B) Schematic of vaccination in hACE2 or hDPP4 transgenic mice. Two doses of vaccines were administered to mice at week 0 (prime) and week 4 (boost). Sera were collected at week 0, 2, 4, and 5. One week after boost vaccination, hACE2 mice were challenged intranasally with 5 × 104 focus-forming units (FFU) of authentic SARS-CoV-2 (Wuhan-Hu-1), while hDPP4 mice were challenged intranasally with 5 × 103 FFU of authentic MERS-CoV (ChinaGD01). All mice were euthanized at week 6, followed by viral titer quantification in lung tissues. (C) Endpoint titers of SARS-CoV-2 RBD-specific antibodies (Abs) were quantified by ELISA using serum samples obtained from vaccinated hACE2 mice at week 5. (D) The FRNT50 titers of nAbs against authentic SARS-CoV-2 (Wuhan-Hu-1) were quantified by FRNT assay, utilizing serum samples collected at week 5 from vaccinated hACE2 mice. (E-F) SARS-CoV-2 RNA copies in lung tissues of euthanized hACE2 mice were quantified by RT-qPCR and presented as log10 copies per gram (g) of lung tissues (E), while infectious viral titers were evaluated by focus-forming assay (FFA) and presented as log10 FFU per g of lung tissues (F). (G-H) H&E staining (G) and IHC with antibodies against SARS-CoV-2 N proteins (H) were conducted on lung tissues of euthanized hACE2 mice. (I-J) Serum samples of vaccinated hDPP4 mice were collected at week 5 and subjected to Ab titer evaluation. MERS-CoV RBD-specific IgG titers were assessed by ELISA (I), while FRNT50 titers of nAbs against authentic MERS-CoV (ChinaGD01) were quantified using FRNT assay (J). (K-L) One week after MERS-CoV challenge, lung tissues of euthanized hDPP4 mice were homogenized and utilized for viral titer quantification. MERS-CoV RNA copies were quantified by RT-qPCR (K), while infectious viruses were evaluated by FFA (L). (M-N) One lung lobe from each mouse was subjected to H&E staining (M) and IHC using antibodies against MERS-CoV N (N). Scale bars in (G-H) and (M-N) represented 50 μm. Data in (C-F) and (I-L) were presented as mean ± SEM. P values were calculated via one-way ANOVA with Tukey’s multiple comparisons tests. ** P < 0.01, *** P < 0.001 The protective efficacy of bivalent HPF (I69C) nanoparticle vaccines. (A) hACE2 mice or hDPP4 mice were randomly allocated to six groups with similar initial average body weights. All mice were vaccinated as in Fig. 7A. G1: alum-only adjuvant. G2: SC003-HPF (I69C)-formulated Nano Core. G3: Mono Mix of RBD (SARS2) and RBD (MERS). G4: Dimer of RBD (SARS2)-RBD (MERS). G5: Mono_Nano Mix of RBD (SARS2)-HPF (I69C) and RBD (MERS)-HPF (I69C). G6: Dimer_Nano of RBD (Dimer)-HPF (I69C). (B) Schematic of vaccination in hACE2 or hDPP4 transgenic mice. Two doses of vaccines were administered to mice at week 0 (prime) and week 4 (boost). Sera were collected at week 0, 2, 4, and 5. One week after boost vaccination, hACE2 mice were challenged intranasally with 5 × 104 focus-forming units (FFU) of authentic SARS-CoV-2 (Wuhan-Hu-1), while hDPP4 mice were challenged intranasally with 5 × 103 FFU of authentic MERS-CoV (ChinaGD01). All mice were euthanized at week 6, followed by viral titer quantification in lung tissues. (C) Endpoint titers of SARS-CoV-2 RBD-specific antibodies (Abs) were quantified by ELISA using serum samples obtained from vaccinated hACE2 mice at week 5. (D) The FRNT50 titers of nAbs against authentic SARS-CoV-2 (Wuhan-Hu-1) were quantified by FRNT assay, utilizing serum samples collected at week 5 from vaccinated hACE2 mice. (E-F) SARS-CoV-2 RNA copies in lung tissues of euthanized hACE2 mice were quantified by RT-qPCR and presented as log10 copies per gram (g) of lung tissues (E), while infectious viral titers were evaluated by focus-forming assay (FFA) and presented as log10 FFU per g of lung tissues (F). (G-H) H&E staining (G) and IHC with antibodies against SARS-CoV-2 N proteins (H) were conducted on lung tissues of euthanized hACE2 mice. (I-J) Serum samples of vaccinated hDPP4 mice were collected at week 5 and subjected to Ab titer evaluation. MERS-CoV RBD-specific IgG titers were assessed by ELISA (I), while FRNT50 titers of nAbs against authentic MERS-CoV (ChinaGD01) were quantified using FRNT assay (J). (K-L) One week after MERS-CoV challenge, lung tissues of euthanized hDPP4 mice were homogenized and utilized for viral titer quantification. MERS-CoV RNA copies were quantified by RT-qPCR (K), while infectious viruses were evaluated by FFA (L). (M-N) One lung lobe from each mouse was subjected to H&E staining (M) and IHC using antibodies against MERS-CoV N (N). Scale bars in (G-H) and (M-N) represented 50 μm. Data in (C-F) and (I-L) were presented as mean ± SEM. P values were calculated via one-way ANOVA with Tukey’s multiple comparisons tests. ** P < 0.01, *** P < 0.001 Following authentic SARS-CoV-2 (Wuhan-Hu-1) challenge, no mortality occurred in any group at a virus dose of 5 × 104 FFU. However, mice in the adjuvant-only and Nano Core control groups exhibited progressive body weight loss, whereas mice vaccinated with monomeric, dimeric, or nanoparticle vaccines maintained stable body weight throughout the monitoring period (Fig. S10A). The ELISA analysis, which utilized serum samples obtained from hACE2 mice one week after the boost vaccination, demonstrated that both bivalent nanoparticle vaccines induced higher titers of SARS-CoV-2 RBD-specific antibodies than the monomer mixture and dimer vaccines, with endpoint titers around 104 (Fig. 9C). These antibodies also demonstrated potent neutralization of authentic SARS-CoV-2, with nAb titers approximately 103 (Fig. 9D). The quantification of SARS-CoV-2 RNA copies in lung tissues of euthanized hACE2 mice revealed that the administration of bivalent nanoparticle vaccines effectively suppressed viral replication in lung tissues, resulting in significantly lower viral loads than those in adjuvant and nanoparticle core groups (Fig. 9E). Consistently, the focus-forming assay (FFA) demonstrated that lung tissues of hACE2 mice immunized with bivalent nanoparticle vaccines contained no infectious SARS-CoV-2 (Fig. 9F). In contrast, the adjuvant-only and nanoparticle core-only groups contained approximately 104 FFU of authentic viruses per gram of lung tissue. Further H&E staining and IHC analyses of lung tissues indicated that SARS-CoV-2 challenge led to remarkable lung lesions in the adjuvant and nanoparticle core groups, resulting in thickened alveolar septa, inflammatory cell infiltration, and dispersed foci expressing viral nucleocapsid (N) proteins (Fig. 9G and H). Conversely, no significant pathological changes were observed in lung tissues of mice immunized with monomer, dimer, or bivalent nanoparticle vaccines. Blinded histopathological scoring confirmed that lung-pathology scores in the adjuvant and Nano Core groups ranged from 2 to 4, indicating mild-to-severe lesions, whereas nanoparticle vaccine groups displayed near-normal lung histology with scores close to 0 (Fig. S10B). IHC quantification further demonstrated that nanoparticle vaccine groups exhibited markedly reduced viral N protein density compared to control groups (Fig. S10C). Similarly, following authentic MERS-CoV (ChinaGD01) challenge at a dose of 5 × 103 FFU, no mortality occurred in any group, whereas the adjuvant and Nano Core groups exhibited progressive body weight loss and all RBD (MERS)-containing vaccine groups maintained stable body weight (Fig. S10D). Bivalent nanoparticle vaccines elicited elevated titers of MERS-CoV RBD-specific antibodies in hDPP4 mice, which were significantly higher than those elicited by the monomer and dimer vaccines (Fig. 9I). These antibodies also potently neutralized authentic MERS-CoV (Fig. 9J). Furthermore, the immunization of bivalent nanoparticle vaccines significantly reduced viral loads in lung tissues of MERS-CoV-challenged hDPP4 mice. Both RNA copies and FFU of MERS-CoV in lung tissues of bivalent nanoparticle vaccine groups were significantly lower than those in adjuvant-only and nanoparticle core-only groups (Fig. 9K and L). Additionally, lung tissues of MERS-CoV-challenged mice in the adjuvant and nanoparticle core groups showed severe pathological alterations and scattered MERS-CoV N-expressing cells, which were not observed in those of nanoparticle vaccine-immunized mice (Fig. 9M and N). Consistent with the SARS-CoV-2 challenge results, blinded histopathological analysis of MERS-CoV-challenged lungs yielded scores of 2 to 4 in the adjuvant and Nano Core groups, contrasting with scores close to 0 in nanoparticle vaccine groups (Fig. S10E). Similarly, IHC quantification confirmed a marked reduction of viral N protein density in the lungs of nanoparticle-immunized mice (Fig. S10F). Altogether, these results demonstrated that HPF (I69C)-based bivalent nanoparticle vaccines exhibited potent protective efficacy against authentic SARS-CoV-2 and MERS-CoV challenges in humanized mice. Discussion In this study, we reported a rational design of Helicobacter pylori ferritin (HPF) nanoparticle by introducing inter-subunit disulfide bonds (S–S) between adjacent protomers. Through substituting centrosymmetric amino acid residues on antiparallel loops with cysteine (C) residues, we verified four artificial intelligence (AI)-based optimizations. Our screening results revealed that the isoleucine (I) at position 69 substituted with C (I69C) exhibited the highest yield and solubility. The engineered HPF (I69C) served as a robust scaffold for constructing bivalent nanoparticle vaccines against both SARS-CoV-2 and MERS-CoV. Structural characterization by cryo-EM at 2.2 Å directly confirmed the formation of inter-subunit S–S crosslinks. Multiple biophysical and biochemical analyses, which included DLS, negative staining, SEC, and turbidity assays, revealed that HPF (I69C) was substantially more homogeneous, soluble and stable to freeze-thaw cycles, pH extremes and elevated temperature compared with wild-type HPF. Utilizing the high-efficiency ST003/SC003 conjugation system, we achieved rapid and complete covalent conjugation of antigens to HPF (I69C). Two HPF (I69C)-based nanoparticle vaccines, including the mixture of monomeric RBD nanoparticles and dimeric RBD nanoparticles, elicited markedly higher RBD-specific binding antibody and neutralizing antibody (nAb) titers compared with monomeric and dimeric subunit vaccines. Nanoparticle vaccine-elicited nAbs effectively neutralized authentic SARS-CoV-2 (Wuhan-Hu-1) and MERS-CoV (ChinaGD01) viruses. Besides, these two nanoparticle vaccines induced higher percentages of Tfh cells, antigen-specific B cells and memory B cells (MBCs), demonstrating enhanced humoral immune responses. HPF (I69C) nanoparticle vaccines also significantly enhanced polyfunctional CD4+ and CD8+ T cell responses, indicating stronger activation of cell-mediated immunity and Th1-biased immune profiles against invading viruses. Additionally, bivalent HPF (I69C) nanoparticle vaccines protected humanized mice against authentic coronavirus challenges. We also conducted systemic safety examinations including complete blood counts (CBCs), serum biochemical analyses, urinalysis, and histopathological evaluation, and confirmed that these vaccines produced no detectable systemic toxicity in mice. Collectively, these results indicated that disulfide bond-stabilized HPF (I69C) coupled with optimized covalent conjugation system constituted a practically viable and immunologically potent platform for bivalent coronavirus vaccines. Previous studies on vaccine stabilization attempts have demonstrated that the introduction of disulfide bond can efficiently stabilize the prefusion conformation of respiratory syncytial virus (RSV) F protein trimer, which potentially avoids the formation of antibody-dependent enhancement (ADE) induced by F vaccines [47]. Similar disulfide bond-introduction designs also have been successfully applied to stable Zika virus (ZIKV) E protein dimeric vaccine, which prevents the exposure of non-neutralizing epitopes and inhibits ADE induced by Dengue virus (DENV) infection [48]. The envelope glycoprotein (Env) of human immunodeficiency virus (HIV) serves as the major antigen of HIV [49]. The stabilization of Env trimer by disulfide bond introduction has facilitated the development of SOSIP, which prompts clinical trials of HIV vaccines [50]. In this study, we have successfully applied the disulfide bond introduction strategy to stabilize nanoparticle vaccines. The resulting HPF (I69C) nanoparticles form 12 pairs of covalent S–S bonds between I69C residues of adjacent protomers, which is confirmed by the high-resolution cryo-EM reconstruction. Based on both biochemical analysis and structure determination, the observed S–S bond strengthens inter-subunit contacts, reduces conformational heterogeneity, and limits the dynamic exposure of hydrophobic patches that could drive aggregation or oligomerization. The local alterations in electrostatic surface potential at the engineered interface are likely to have modified the interfacial water structure and charge pairing, a phenomenon we hypothesize contributes to the enhanced solubility under physiological ionic strength conditions and the reduced turbidity observed in high NaCl buffer environments. These structural and surface chemistry changes likely contribute to the enhanced thermal, freeze-thaw and pH stability observed in HPF (I69C), which also account for the preservation of particle integrity and immunogenicity in antigen-conjugated HPF (I69C) following long-term storage at elevated temperatures. The induction of effective nAbs against viruses relies on the proper presentation of neutralizing epitopes on vaccines [51]. Considering the improved immunogenicity of HPF (I69C) nanoparticle vaccines, we believe that both the multivalent architecture and the enhanced stabilization of RBD spatial conformation contribute to the enhanced immune responses. By displaying RBD antigens of both SARS-CoV-2 and MERS-CoV in a homogeneous, correctly folded state with accessible receptor-binding motifs, HPF (I69C) maintains the structural integrity and proper presentation of conformational neutralizing epitopes, which significantly increases the probability that B-cell receptors (BCRs) recognize and clonally expand B cells targeting virus-neutralizing sites rather than non-protective regions. Consequently, nanoparticle vaccines skew the antibody repertoire toward higher neutralization potency and breadth. Conformational stabilization also reduces epitope masking or heterogeneity caused by flexible or misfolded RBD. Uniform orientation and consistent epitope geometry across the nanoparticle surface enhance cooperative BCR engagement and favor the selection of B-cell clones that recognize conserved surfaces, which strengthen the affinity maturation of B cells within germinal centers (GCs). Additionally, conformationally stable RBD antigens on HPF (I69C) exhibit resistance to premature proteolytic degradation, thereby preserving intact epitopes during antigen-presenting cell (APC) processing and enhancing the presentation of intact nanoparticle vaccines to B cells by Follicular dendritic cells (FDCs). The enhancement of antigen presentation consequently leads to the observed increase in Tfh cells and antigen-specific B cells within the spleen. The enhanced monodispersity and homogeneity of RBD-conjugated HPF (I69C) nanoparticle vaccines may also account for the improved cellular immune responses. Our data reveal that HPF (I69C) nanoparticle vaccines significantly augment polyfunctional CD4+ and CD8+ T cell responses, suggesting more efficient antigen uptake and presentation by APCs. Taken together, these advantages suggest a synergistic model in which HPF (I69C) nanoparticles integrate geometric multivalency with conformational fidelity. By displaying antigens on HPF (I69C) nanoparticles, both physical multivalency and biophysical stabilization are achieved, thereby maintaining the proper immunogenic surfaces throughout the immune response. Previous reports have established ferritin and many other protein-based nanoparticles as effective platforms for displaying coronavirus antigens and raising potent neutralizing responses [5]. Ferritin proteins from both Helicobacter pylori and Pyrococcus furiosus have been widely used to display 24 copies of antigens [19, 52, 53]. Archaeal lumazine synthase (LS), bacterial E2p, as well as computationally designed mi3 and I53-50 can display 60 copies of antigens [20–22, 54]. Due to the incompatibility of folding and assembly requirements in mammalian cells, nanoparticles and antigens are often expressed in prokaryotic and eukaryotic cells, respectively. They are subsequently assembled in vitro through conjugation systems. Commonly used conjugation systems include covalent bond formation via the SpyTag/SpyCatcher isopeptide linkage, as well as noncovalent binding through the FagyTag affinity interaction [55]. Although the separated expression and in vitro assembly strategy potentially addresses the challenge of antigen presentation, their complex manufacturing, stability and conjugation efficiency still hinder the scale-up production and broad application. Our work extends these findings by addressing these limitations. The engineered HPF (I69C) exhibits enhanced homogeneity and stability. Compared to the earlier ferritin designs that rely on non-covalent assembly or less efficient conjugation systems, the combination of an inter-subunit disulfide bond and the ST003/SC003 molecular glue yields complete and rapid antigen conjugation while maintaining homogeneity and stability in physiologically relevant conditions [19, 32, 33]. Additionally, our bivalent nanoparticle designs generate robust humoral and cellular immune responses against both SARS-CoV-2 and MERS-CoV, offering flexibility for vaccine formulation as a multivalent regimen, and eliciting broad nAbs across both sarbecovirus and merbecovirus subgenera. In the broader landscape of MERS-CoV vaccine development, the ChAdOx1 MERS adenoviral vector vaccine achieved neutralizing antibody seroconversion in only 44% of high-dose recipients in a Phase 1 trial [56]. A recently reported MERS-CoV RBD-mRNA vaccine induced potent neutralizing responses in mice and demonstrated protection against viral challenge [57]. Among nanoparticle platforms, the I53-50-based SARS-CoV-2 RBD nanoparticle vaccine GBP510 (SKYCovione) received WHO emergency use listing and marketing authorization, and a companion I53-50 MERS-CoV RBD nanoparticle vaccine induced potent neutralizing antibodies in mice [29]. Mosaic ferritin nanoparticles displaying multiple coronavirus RBDs, including SARS-CoV-2 and MERS-CoV [16, 30], have elicited broad neutralizing responses in preclinical models, highlighting the promise of multivalent nanoparticle vaccines against coronaviruses. Our HPF (I69C) platform, while not yet evaluated clinically, offers distinct advantages in thermal stability, long-term storage, and near-instantaneous covalent conjugation, which collectively address manufacturing and cold-chain bottlenecks relevant to both pandemic and endemic settings. The stabilized, highly conjugation-efficient ferritin platform such as HPF (I69C) is broadly applicable to multivalent vaccine design beyond coronaviruses. The enhanced thermal stability and ionic tolerance of the I69C scaffold could aid manufacturing scalability and cold-chain resilience, facilitating vaccine deployment in resource-limited settings where maintaining continuous refrigeration is logistically challenging. Several additional quality attributes of the HPF (I69C) platform merit consideration for translational development. Although anti-scaffold and anti-ST003-SC003 antibody responses were detected, our prime-boost experiments demonstrated that pre-existing immunity against the scaffold or conjugation domains did not measurably impair RBD-specific antibody or neutralizing responses to subsequent nanoparticle vaccination, supporting the feasibility of homologous boosting with this platform. Additionally, endotoxin levels in all E. coli-produced nanoparticle preparations were confirmed to be below the clinical threshold for parenteral products, ensuring that the observed immunogenicity was attributable to the vaccine design rather than to endotoxin-driven innate activation. Furthermore, multiple independent batches of HPF (I69C) and its antigen conjugates exhibited consistent SEC elution profiles and DLS size distributions across this study, supporting batch-to-batch reproducibility. Notably, the N-linked glycan profiles of mammalian-expressed RBD antigens have not been characterized in this study. Comprehensive glycosylation analysis would further strengthen the regulatory package and should be addressed during process development. Although we have systematically assessed the stability, immunogenicity, and safety of the engineered bivalent HPF (I69C) nanoparticle vaccines, several limitations of this study should be acknowledged. Firstly, although the cryo-EM validation confirmed the formation of disulfide bonds in the HPF (I69C) core, the structural determination of antigen-conjugated nanoparticles was not achieved due to flexible linkers connecting antigens to nanoparticle cores. More detailed structural information regarding the antigen orientation and spacing on the nanoparticle surface would provide valuable insights for optimizing immunogen design and should be prioritized in future research. Secondly, our study only used C57BL/6 mice, hACE2 mice, and hDPP4 mice to evaluate immunogenicity and safety of nanoparticle vaccines. More relevant animal models, such as non-human primates, should be included to demonstrate protective efficacy against viral challenges. Thirdly, we have not evaluated the breadth of nAbs against diverse SARS-CoV-2 and MERS-CoV variants due to the limited availability of authentic virus strains. A comprehensive understanding of their broad-spectrum neutralizing capacity could facilitate the development of universal vaccines capable of protecting against future viral mutants. Fourthly, although alum was selected for its well-established clinical safety, its inherent Th2 bias may have limited the breadth of cellular responses. Exploring Th1-skewing adjuvants such as CpG or saponin-based formulations represents a logical next step. Fifthly, head-to-head immunogenicity comparisons with other leading nanoparticle platforms such as I53-50, mi3, and LS were not performed, and such benchmarking would further contextualize the HPF (I69C) platform within the broader nanoparticle vaccine landscape. Conclusions In summary, we demonstrate that the introduction of a rationally designed inter-subunit disulfide bond between I69C residues in Helicobacter pylori ferritin significantly enhances nanoparticle homogeneity, solubility, and stability. By coupling this stabilized scaffold with the ST003/SC003 molecular glue system, we have developed two bivalent RBD nanoparticle vaccines targeting both SARS-CoV-2 and MERS-CoV. These vaccines exhibit robust immunogenicity, potent protective efficacy, and favorable preclinical safety profiles in various mouse models. Our approach provides a scalable platform for the development of multivalent coronavirus vaccines and may have broad applicability to other antigen systems where enhanced nanoparticle stability and efficient covalent antigen presentation are required. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1: Fig. S1. Cryo-EM 3D reconstruction of HPF (I69C). Fig. S2. The pH stability of HPF (I69C). Fig. S3. The freeze-thaw stability of HPF (I69C). Fig. S4. The long-term stability of HPF (I69C). Fig. S5. The turbidity and conjugation efficiency of SC003-HPF (I69C). Fig. S6. The binding affinity of receptors and antibodies for RBD vaccines. Fig. S7. Urinalysis and histopathological analyses of vaccinated mice. Fig. S8. The neutralization breadth against SARS-CoV-2 variants. Fig. S9. Gating strategies used for cellular immune analyses. Fig. S10. Body-weight monitoring and lung-pathology scoring of authentic virus-challenge mice. Table S1. Cryo-EM data collection, processing and refinement statistics. Table S2. Surface electrostatic potential of HPF and HPF(I69C) subunits. Table S3. SASA and MLP analysis of HPF and HPF(I69C). Table S4. Rosetta InterfaceAnalyzer evaluation of HPF and HPF (I69C). Supplementary Material 1: Fig. S1. Cryo-EM 3D reconstruction of HPF (I69C). Fig. S2. The pH stability of HPF (I69C). Fig. S3. The freeze-thaw stability of HPF (I69C). Fig. S4. The long-term stability of HPF (I69C). Fig. S5. The turbidity and conjugation efficiency of SC003-HPF (I69C). Fig. S6. The binding affinity of receptors and antibodies for RBD vaccines. Fig. S7. Urinalysis and histopathological analyses of vaccinated mice. Fig. S8. The neutralization breadth against SARS-CoV-2 variants. Fig. S9. Gating strategies used for cellular immune analyses. Fig. S10. Body-weight monitoring and lung-pathology scoring of authentic virus-challenge mice. Table S1. Cryo-EM data collection, processing and refinement statistics. Table S2. Surface electrostatic potential of HPF and HPF(I69C) subunits. Table S3. SASA and MLP analysis of HPF and HPF(I69C). Table S4. Rosetta InterfaceAnalyzer evaluation of HPF and HPF (I69C).