Rationally engineered cationic ferritin nanoparticles overcome pre-existing immunity to adenoviral vectors Authors: Yilong Yang, Di Wang, Yaohui Li, Xiaofan Zhao, Mengran Yi, Xiaoyan Huang, Xiaodong Zai, Zhe Zhang, Shipo Wu, Jun Zhang, Lihua Hou, Junjie Xu, Wei Chen Published: 2026 ABSTRACT AbstractViral vectors and protein nanoparticles represent important platforms in vaccine delivery, yet their potential synergy to overcome fundamental biological barriers remains largely unexplored. To address the challenge of pre-existing immunity against viral vectors, we rationally engineered a cationic human ferritin nanoparticle, termed (+)hF, designed to electrostatically assembles with adenovirus serotype 5 (Ad5) into a supramolecular complex. This construct was generated via structure-guided point mutations that introduced positive surface charges while retaining the innate self-assembly capability of the ferritin nanocage. The resulting (+)hF formed nanocomplexes with Ad5, which enhanced transgene expressions both in vitro and in vivo and effectively reduced sensitively to anti-Ad5 neutralization. In murine models with pre-existing immunity, intranasal immunization with (+)hF-complexed Ad5-based vaccines (encoding antigens for SARS-CoV-2 or Hendra virus) elicited significantly higher antigen-specific IgG and neutralizing antibody titers. A key advantage of (+)hF is its endogenous origin, which ensures high biocompatibility by preventing the induction of anti-carrier antibodies and any associated immune burden. This study presents a new paradigm that engineers programmable biointerfaces to synergize viral vectors with protein nanocages, offering a general strategy to circumvent critical biological barriers for enhanced vaccine and gene delivery.Graphical abstract Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12951-026-04704-8.Keywords: Cationic protein nanoparticles, Ferritin, Self-assembly, Pre-existing immunity, Viral vector Introduction Protein nanoparticles are extensively utilized in nanobiotechnology due to their favorable biocompatibility, unique structural and functional properties [1]. Ferritin nanocages, which self-assemble from 24 subunits into a hollow, highly symmetric, and stable globular structure, represent a prominent class of such nanoparticles and serve as excellent delivery platforms for bioactive molecules [2–5]. The exterior surface of ferritin nanoparticle can be functionalized via protein adaptors [6–8], chemical conjugation [9], or genetic fusion [10–12], enabling antigen multimerization that enhances immune recognition and potentiates immune responses [13, 14]. Meanwhile, the interior cavity allows for the encapsulation of small-molecule drugs (e.g., siRNA and doxorubicin) through a pH-mediated reversible disassembly/assembly process, facilitating tumor-targeted delivery [15–19]. Although ferritin has been widely investigated as a delivery vector, substantial potential remains for engineering this natural nanoparticle to broaden its biomedical applications by orchestrating synergistic interactions with other active molecules [20–24]. Viral vectors, engineered to deliver genetic material into target cells, constitute another major platform for vaccines and gene therapy. Human adenovirus serotype 5 (Ad5) vectors, for instance, have demonstrated notable efficacy against pathogens such as SARS-CoV-2 [25, 26] and Ebola [27]. However, widespread pre-existing immunity to adenoviruses in human populations can generate neutralizing antibodies that reduce vector transduction, suppress transgene expression, and ultimately compromise vaccine immunogenicity [28, 29]. To address this, several strategies have been explored, including the use of rare serotypes [30], capsid engineering [31, 32], and chemical modifications such as PEGylation [33, 34] or lipid envelopment [35]. Each strategy presents distinct considerations [32]: rare serotypes may differ in natural prevalence and immunogenic profile; capsid engineering requires extensive redesign efforts; and chemical modifications can sometimes affect essential vector–cell interactions, influencing transduction efficiency and biocompatibility. The superstructures formed by the electrostatic co-assembly of oppositely charged nanoparticles represent a promising platform for designing novel delivery systems in a non-covalent manner [36, 37]. Compared to direct structural or chemical modifications of single adenoviral particles, our prior research established that cationic nanocapsules can complex with adenoviral vectors via electrostatic interactions, boosting Ad5 delivery and overcoming pre-existing immunity in both in vitro and in vivo models [38]. Here, we extend this strategy by rationally engineering ferritin nanocages to biosynthesize protein nanoparticles with positively charged outer surfaces. This approach leverages the innate biocompatibility and modular self-assembly of ferritin to create a more efficient and simpler-to-produce alternative to previous cationic carriers. The resulting positively charged human ferritin nanoparticles, designated (+)hF, self-assemble with Ad5 into nanocomplexes that significantly enhance transduction efficiency in both coxsackievirus and adenovirus receptor [39] (CXADR)-positive and CXADR-negative cells. Moreover, Ad5@(+)hF robustly elevate transgene expression and enhance the immunogenicity of Ad5-based vaccines in models of pre-existing immunity, demonstrating a promising protein-based solution to this translational challenge. Materials and methods Materials HeLa, 3T3, HEK-293T, and ACE2-293T cells were laboratory-preserved. Cell culture reagents, including Dulbecco’s modified Eagle’s medium (DMEM), fetal bovine serum (FBS), penicillin-streptomycin, 0.25% trypsin, Opti-MEM, and phosphate-buffered saline (PBS) were purchased from Gibco (USA). Luciferase-expressing plasmid DNA (pDNA) was obtained from Absin (China). mRNA expressing luciferase was purchased from Vazyme (China). Lipofectamine 3000 and Hoechst 33,342 was purchased from Thermo Fisher Scientific (USA). Adenoviral vectors, including Ad5-Luc (luciferase-expressing), Ad5-GFP (GFP-expressing), and Ad5-Null (empty vector), were purchased from Hanbio (China). E1/E3-deleted replication-defective Ad5-based vaccines (Ad5-nCoV and Ad5-HeV), HeV pseudotyped virions, hexon protein of Ad5, and HeV-G protein were laboratory-preserved. SARS-CoV-2 XBB.1.5 Spike RBD protein and pseudotyped virus were obtained from Sino Biological (China) and Vazyme (China), respectively. Expression, purification and characterization of hF(+) The recombinant plasmids encoding wild-type human ferritin heavy chain (hF-wt; NCBI: NP_002023.2), its engineered mutants, and wild-type mouse ferritin heavy chain (mF-wt; NCBI: NP_034369.1) were synthesized (General Biol, China) and subsequently transformed into E. coli BL21 competent cells (TransGen, China) for protein expression. Cells were cultured at 37 °C until an OD₆₀₀ of 0.6–0.8 was reached, then shifted to 15 °C for protein expression induced by the addition of 1 mM isopropyl-β-D-thiogalactoside (IPTG; Sigma-Aldrich, USA) for 20 h. Bacterial cells were pelleted by centrifugation and lysed via sonication, followed by purification of the His-tagged recombinant proteins from the supernatant using a HisTrap Excel affinity chromatography column (Cytiva, USA). Lipopolysaccharide (LPS) was removed with polymyxin B gel (Thermo Fisher, USA), and the residual LPS concentration was determined by a quantitative chromogenic microplate assay (BIOENDO, China). Protein purity was assessed by SDS-PAGE and the hydrodynamic size of the purified proteins was characterized by dynamic light scattering (DLS) along with visualization using transmission electron microscopy (Thermo Fisher, USA). Electrophoretic mobility shift assays Electrophoretic mobility shift assays were conducted to investigate the electrostatic interactions between (+) hF and pDNA or mRNA. Briefly, (+) hF and pDNA or mRNA were mixed at mass ratios of 0:1, 5:1, 10:1, and 20:1, and the resultant mixtures were incubated at room temperature for 15 min. Agarose gel electrophoresis was then performed in TAE buffer at a constant voltage of 180 V for 20 min. Following electrophoresis, the gels were visualized under ultraviolet (UV) illumination (Thermo Fisher Scientific, USA) using a gel imaging system. The retardation rate of pDNA@(+)hF was calculated using the following formula: IFree and ITotal represent the integrated density values of the unbound plasmid DNA (pDNA) band and the total pDNA band, respectively, as quantified by grayscale analysis using ImageJ. Preparation and characterization of Ad5@(+)hF Ad5-Luc and (+) hF were mixed at molar ratios of 1:0, 1:1, 1:10, 1:100, and 1:1000 in 20 mM phosphate buffer (PB, pH 7.0), followed by incubation at room temperature for 15 min. The size and zeta potential of Ad5-Luc@(+) hF complexes were assessed by dynamic light scattering (DLS) performed on Zetasizer Nano (Malvern, UK). The morphological characteristics of the complexes were characterized by transmission electron microscopy (Thermo Fisher, USA). In vitro transgene expression of Ad5@(+)hF 3T3 and HeLa cells were seeded in 96-well plates (Corning, USA) at a density of 2 × 104 cells per well and incubated overnight. Cells were then transfected with Ad5-Luc or Ad5-GFP (2 × 105 IFU per well), either alone or pre-complexed with hF-wt or (+)hF at specified molar ratios (Ad5:(+)hF = 1:0, 1:1, 1:10, 1:100 and 1:1000). After 24 h of incubation, transgene expression was analyzed. Luciferase activity was quantified using the Bright-Lite Luciferase Assay System (Vazyme, China), while GFP expression was evaluated by fluorescence imaging (Cytation, USA). Internalization mechanism study To investigate the internalization pathways of Ad5@(+)hF complexes, a panel of inhibitors and a low-temperature (4°C) treatment were employed. HeLa cells were seeded into 96-well plates (Corning, USA) at 2 × 104 cells/well. After overnight incubation at 37 °C with 5% CO2, the cells were pretreated for 4 h with one of the following 200 mM 2-deoxy-D-glucose (2-DG; MCE, China), 2 mg/mL methyl-β-cyclodextrin (M-β-CD; Sigma-Aldrich, USA), 20 µg/mL amiloride hydrochloride (AM; Solarbio, China), or were pre-incubated at 4°C. Cells were then incubated for 4 h with culture medium containing Ad5-Luc (2 × 10⁵ IFU/well), either alone or pre-complexed with hF-wt or (+)hF at a 1:1000 molar ratio. After 24 h, luciferase activity was measured with the Bright-Lite Luciferase Assay System (Vazyme, China) to assess transduction efficiency. RNA interference CXADR-targeting small interfering RNA (siRNA; siCXADR, sense sequence:5’-GCUACAUCGGCAGUAAUCAdTdT-3’; antisense sequence: 5’-UGAUUACUGCCGAUGUAGCdTdT-3’) and negative control siRNA (siCon, sense sequence: 5’-UUCUCCGAACGUGUCACGUdTdT-3’; antisense sequence: 5’-ACGUGACACGUUCGGAGAAdTdT-3’) were purchased from Sangon Biotech (Shanghai, China). HeLa cells were seeded into 96-well plates at a density of 1 × 104 cells/well and cultured overnight at 37 °C with 5% CO2. Subsequently, the cells were transfected with siCXADR or siCon using Lipofectamine RNAiMAX (Thermo Fisher Scientific, USA) following the manufacturer’s protocol. Forty hours after siRNA transfection, cells were incubated for 12 h with Ad5-Luc (2 × 10⁵ IFU/well) alone or complexed with hF-wt or (+)hF (molar ratio 1:1000). Luciferase activity was subsequently quantified. Animal experiments Female BALB/c mice (5–6 weeks old) were purchased from Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China) and housed under specific pathogen-free (SPF) conditions. All animal experiments were performed in strict accordance with the Guide for the Care and Use of Laboratory Animals and were approved by the Animal Care and Welfare Ethics Committee of the Academy of Military Medical Sciences (approval number: IACUC-SWGCYJS-2023-007). Pre-existing anti-Ad5 immunity (Pre-Ad5) model To establish a pre-existing immunity model, BALB/c mice received a single intramuscular injection of empty Ad5 vector (1 × 107 IFU per mouse). Four weeks later, serum was collected via retro-orbital bleeding under anesthesia. The titer of Ad5-specific antibodies in the serum was quantified by enzyme-linked immunosorbent assay (ELISA). In vivo bioluminescence imaging (BLI) BALB/c Mice with or without pre-existing anti-Ad5 immunity were intranasally administered 20 µL of Ad5-Luc (5 × 106 IFU per mouse), either alone or complexed with hF-wt or (+)hF at designated molar ratios of 1:10, 1:100 and 1:1000. Bioluminescence imaging was conducted at indicated days after administration. Before in vivo imaging using the IVIS Spectrum system (PerkinElmer, USA), mice were injected intraperitoneally with D-Luciferin (Promega, USA) at a dose of 150 mg/kg and anesthetized with isoflurane. The resulting bioluminescent signal was quantified within defined regions of interest to analyze expression kinetics and tissue distribution. Immunization BALB/c mice with pre-existing anti-Ad5 immunity were intranasally immunized with 5 × 10⁵ IFU of Ad5-nCoV or Ad5-HeV, either alone or complexed with (+)hF at molar ratios of 1:10 and 1:100. Serum was collected on days 14 and 28 post-immunization, and bronchoalveolar lavage fluid (BALF) was harvested on day 28. Levels of Ad5-specific, SARS-CoV-2-RBD-specific, and HeV-G-specific immunoglobulins (IgG and IgA) in serum and BALF were measured by Enzyme-linked immunosorbent assays (ELISA). Additionally, pseudovirus neutralization assays against SARS-CoV-2 XBB.1.5 and HeV were performed on day 28 serum to determine neutralizing antibody titers. ELISA 96-well plates were coated overnight at 4 °C with 1 µg/mL of the specified antigens (RBD of SARS-CoV-2 XBB.1.5, HeV-G, Ad5 hexon, hF-wt, (+)hF variants, or mF-wt) in 50 mM carbonate buffer (pH 9.6). Plates were then blocked with 2% BSA in PBS for 1 h at 37 °C. After washing with PBST, serially diluted serum or BALF samples were added and incubated for 1 h at 37 °C. Following another wash, plates were incubated for 45 min at 37 °C with horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG or IgA antibody (1:10,000 dilution). After a final wash, the reaction was developed with TMB substrate for 6 min, stopped with stop solution, and absorbance was measured at 450 nm. Pseudovirus neutralization assays Serum samples were serially diluted 3-fold and mixed with an equal volume of pre-titrated SARS-CoV-2 XBB.1.5 or HeV pseudotyped virions. After incubation at 37 °C for 1 h, the mixture was transferred to ACE2-293T or HEK-293T cells (seeded at 2 × 104 cells/well in 96-well plates). Following a 48-h incubation, cells were lysed and luminescence was measured. The 50% neutralization titer (NT₅₀) was calculated as the highest serum dilution that reduced luminescence by 50% relative to the virus-only control (set as 0% neutralization). Wells containing cells only were used as background controls (set as 100% neutralization). Samples with NT₅₀ values below the lowest tested dilution (1:30) were considered negative for neutralizing activity. Safety assessments BALB/c mice were intranasally immunized with PBS, 5 × 10⁵ IFU of Ad5-nCoV or Ad5-nCoV@(+)hF (at molar ratios of 1:10 and 1:100). Seven days post-administration, whole blood was collected for biochemical analysis, which included quantification of alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), urea, creatinine (CREA), triglycerides (TG), lactate dehydrogenase (LDH), and hemoglobin (HGB). In parallel, major organs (liver, lung, and kidney) were harvested, fixed, embedded in paraffin, sectioned, and stained with hematoxylin and eosin (H&E) for histopathological examination. BALB/c mice were intranasally immunized with PBS, 5 × 10⁵ IFU of Ad5-HeV or Ad5-HeV@(+)hF (at molar ratios of 1:10 and 1:100). Serum was collected 28 days post-immunization, and IgG antibody responses against mouse ferritin (mF-wt), human ferritin (hF-wt), and the engineered variants (+)hF-R2 and (+)hF-A1 were measured by ELISA. Statistical analysis Statistical analysis was performed with GraphPad Prism 9.4.1. Data are presented as means ± SEM. The statistical significance was analyzed by one-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test for single-factor comparisons and by two-way ANOVA for comparisons involving two independent factors. Statistic differences were labeled as *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; ns, not significant. Results Design and characterization of cationic ferritin nanoparticles Ferritin spontaneously assembles into 24-subunit nanocages. We delineated the sequence of the heavy chain of human ferritin (hF), encompassing the N-terminus, C-terminus, five α-helices, and the loop interspersed between them. To achieve positive charge modification on the ferritin surface with minimal effects on its self-assembly, we focused on surface-exposed residues outside critical self-assembly domains as potential mutation sites. Based on the crystal structure of ferritin nanocage (PDB:3ajo), we pinpointed the surface-exposed amino acid residues (S1) and their corresponding locations within the protein sequence (Fig. 1A and Fig. S1). Previous studies have shown that the C4 and C3 interfaces (S2) are essential for ferritin nanocage assembly [4], mainly corresponding to the HC and LD−E/HE regions, respectively. Therefore, the set difference between S1 and S2 generated a pool of potential mutation sites. Structure-guided design and characterization of cationic human ferritin nanoparticles. (A) Structure-guided workflow for surface engineering of human ferritin (hF). The hF monomer architecture and secondary-structure annotation (helices HA–HE) are shown. Based on the ferritin nanocage crystal structure (PDB: 3AJO), surface-exposed residues (S1) were mapped. To minimize disruption of self-assembly, residues belonging to the key nanocage assembly interfaces (S2; C4 and C3 interfaces) were excluded, and the resulting set (S1 − S2) was defined as the pool of candidate mutation sites for charge engineering. (B) Net charge comparison of wild-type hF (hF-wt) and engineered variants generated under different design modes, including (+)hF-R1/R2/R3 and (+)hF-A1/A2/A3. (C) Electrostatic surface representations of hF-wt and (+)hF variants, illustrating redistribution toward a more positively charged outer surface for engineered nanocages. Color scale indicates electrostatic potential (red, negative; blue, positive). (D) Dynamic light scattering (DLS) size distributions of hF-wt and (+)hF variants, confirming self-assembly into nanoscale particles. The mean hydrodynamic diameters are indicated. (E) Transmission electron microscopy (TEM) images of hF-wt, (+)hF-R2 and (+)hF-A1. Scale bar, 100 nm Within this S1-S2 pool, we utilized the Rosetta Online Server to generate supercharged [40] candidate protein sequences (Fig. S2) with tuned positive charges under Rosetta-mode and AvNAPSA-mode [41, 42], designated as (+)hF-R1/R2/R3 and (+)hF-A1/A2/A3, respectively. Net charge analysis revealed that, relative to wild-type hF (hF-wt), the (+)hF-R1/R2/R3 and (+)hF-A1/A2/A3 variants displayed distinct degrees of positive charge (Fig. 1B). AlphaFold3 [43] analysis indicated that the reengineered proteins can self-assemble into nanocages. Moreover, charge distribution analysis further showed that mutant (+)hF nanocages featured positively charged outer surfaces (Fig. 1C). Subsequently, the recombinant hF-wt and (+)hF variants were prepared utilizing the E. coli expression system, and their LPS levels were determined to be ≤ 1 EU/mg (Fig. S3). In agreement with AlphaFold3 predictions, dynamic light scattering (DLS) analysis confirmed that the (+)hF variants could self-assemble into nanoparticles (Fig. 1D). The average sizes of (+)hF-R1, (+)hF-R2, (+)hF-R3, (+)hF-A1, (+)hF-A2, and (+)hF-A3 were 37.6, 17.7, 20.5, 32.3, 79.9, and 37.6 nm, respectively, which were larger than the 13.1 nm size of hF-wt. However, transmission electron microscopy (TEM) revealed that only (+)hF-R2 and (+)hF-A1 self-assembled into protein nanoparticles. (+)hF-R2 formed well-defined, homogeneous nanostructures reminiscent of the hF-wt nanocage, whereas (+)hF-A1 exhibited lower structural integrity and morphological disorder (Fig. 1E). Self-assembly of (+)hF with anionic biomacromolecules (+)hF was engineered to have positively charged outer surfaces, enabling it to self-assemble with negatively charged biomacromolecules via electrostatic interactions. Electrophoretic mobility shift assays (EMSA) revealed that (+)hF could form complexes with negatively charged nucleic acids (plasmids and mRNA), as evidenced by a dose-dependent retardation of mobility (Fig. S4). Among the variants, (+)hF-R2 and (+)hF-A1 exhibited the highest binding efficiency, achieving up to 90% retardation of plasmid (Fig. 2A), a level comparable to protamine, one of nature’s most highly charged proteins. As a control, the absence of detectable binding with hF-wt and HSA confirmed the specificity of the nucleic acid interaction to the engineered (+)hF variants. Electrostatic assembly of cationic ferritin with nucleic acids and adenoviral vectors. (A) Electrophoretic mobility shift assays showing dose-dependent complex formation between engineered (+)hF variants and negatively charged pDNA. (B) Zeta potential measurements of Ad5-Luc alone and after incubation with control proteins or engineered (+)hF variants. (C, D) Dynamic light scattering (DLS) analysis of size distributions of Ad5-Luc@(+)hF complexes formed at different molar ratios. Size distribution profiles (C) and the average hydrodynamic diameter (D) are shown. (E) Transmission electron microscopy (TEM) images of Ad5-Luc and Ad5@(+)hF-R2 at representative molar ratios. Scale bar, 100 nm (white line), 500 nm (black line). Data in B are presented as means ± SEM and were analyzed by one-way ANOVA with Tukey’s multiple comparisons test against the Ad5 control (**p < 0.01, ****p < 0.0001) Moreover, the capability of (+)hF-R2 and (+)hF-A1 to complex with Ad5 was also validated. DLS analysis indicated that HSA did not alter the Ad5 surface charge, whereas protamine and (+)hF could shift it to less negative values (Fig. 2B). The hydrophobic interactions between HSA or hF-wt and Ad5 did not induce a measurable change in the viral particle size. In contrast, the electrostatic driving force of the (+)hF variants, particularly (+)hF-R2 and (+)hF-A1 at high molar ratios, enabled their robust assembly with Ad5. At a molar ratio of 1:1000, the diameter of Ad5@(+)hF-R2 reached approximately 1100 nm (Fig. 2C-D). Notably, protamine was not able to promote Ad5 nanocomplex formation. It is plausible that this lack of efficacy of protamine stems from a steric insufficiency relative to (+)hF nanoparticles. Despite its high charge density, protamine’s small molecular size likely precludes the multivalent electrostatic cross-linking necessary to bridge adjacent Ad5 particles and drive extensive aggregation. Consistent with the DLS data, TEM imaging revealed that (+)hF-R2 induced minimal changes to Ad5 at lower molar ratios, whereas it caused readily observable and extensive aggregation of the viral particles at a 1:1000 ratio (Fig. 2E). In vitro transgene expressions of Ad5@(+)hF We evaluated transgene expressions mediated by Ad5@(+)hF in both HeLa and 3T3 cells, which are CXADR-positive and CXADR-negative, respectively. Compared to the Ad5-Luc (expressing luciferase) control, both (+)hF-A1 and (+)hF-R2 enhanced transgene expressions of Ad5-Luc at higher molar ratios. Specifically, at a 1:1000 ratio, Ad5-Luc@(+)hF-A1 showed a 1.3-fold increase in HeLa cells, while in 3T3 cells, Ad5-Luc@(+)hF-R2 and Ad5-Luc@(+)hF-A1 achieved 3.6-fold and 2.1-fold higher expressions than Ad5-Luc, respectively (Fig. 3A). Moreover, similar enhancements were observed for Ad5-GFP (expressing GFP) complexed with (+)hF-A1 or (+)hF-R2 at a molar ratio of 1:1000 (Fig.S5). To assess whether self-assembly with (+)hF could reduce sensitively to pre-existing immunity of Ad5, a cellular model was established by pre-treating cells with Ad5-neutralizing antibodies. While the transgene expression of Ad5-Luc was significantly reduced in both HeLa and 3T3 cells under these conditions, Ad5-Luc@(+)hF-A1 and Ad5-Luc@(+)hF-R2 (at a 1:1000 molar ratio) maintained or even enhanced compared to untreated Ad5-Luc (Fig. 3B-C). In contrast, neither HSA, protamine nor cationic lipid (Lipofectamine 3000, Lipo3000) enhanced the transgene expression of Ad5-Luc in HeLa or 3T3 cells (Fig.S6). In vitro transgene expressions of Ad5@(+)hF and mechanistic investigation of cellular entry pathways. (A) Luciferase expression in CXADR-positive HeLa cells (top) and CXADR-negative 3T3 cells (bottom) following transduction with Ad5-Luc complexed with the indicated (+)hF variants at different Ad5-Luc@(+)hF molar ratios. (B, C) In vitro evaluation of resistance to pre-existing anti-Ad5 neutralizing antibodies in HeLa (B) and 3T3 (C) cells, comparing transgene expressions mediated by Ad5-Luc, Ad5-Luc@hF-wt and Ad5-Luc@(+)hF. (D) Cellular transport inhibitors using amiloride, 2-deoxy-D-glucose, methyl-β-cyclodextrin, and low-temperature (4 °C) treatment prior to transduction with Ad5-Luc alone or complexed with hF-wt or (+)hF at a molar ratio of 1:1000. (E) Receptor-dependence analysis based on siRNA-mediated knockdown of CXADR in HeLa cells, assessing the impact on transgene expressions mediated by Ad5-Luc alone or complexed with hF-wt or (+)hF at a molar ratio of 1:1000. Data are presented as means ± SEM. Data in A-C were analyzed by one-way ANOVA with multiple comparisons. Data in D-E were analyzed by two-way ANOVA with multiple comparisons. Pre-Ad5: pre-existing anti-Ad5 immunity. **p < 0.01, ****p < 0.0001; ns, not significant Endocytic pathways of Ad5@(+)hF were investigated using transport inhibitors and low-temperature (4 °C) incubation (Fig. 3D). Specific inhibitors included amiloride (AM, an inhibitor of macropinocytosis), 2-deoxy-d-glucose (2-DG, an inhibitor of energy-dependent pathways), and methyl-β-cyclodextrin (M-β-CD, an agent that disrupts lipid raft-dependent endocytosis). Compared to the untreated control, transgene expressions of Ad5-Luc and Ad5-Luc@hF-wt were significantly reduced by the energy inhibitor 2-DG and at 4 °C, consistent with canonical clathrin-mediated endocytosis [44]. In contrast, the transgene expression of Ad5-Luc@(+)hF-R2 was not impaired by AM, 2-DG, M-β-CD, or 4 °C, suggesting reduced dependence on canonical uptake pathways. For Ad5-Luc@(+)hF-A1, while resistant to AM, its transgene expression was significantly reduced by 2-DG, 4 °C, and M-β-CD, indicating that internalization remains energy-dependent but appears to shift towards a cholesterol-sensitive, lipid raft-associated mechanism. Receptor-mediated endocytosis is critical for viral vector entry. To evaluate the CXADR [39] dependence of Ad5@(+)hF entry, we knocked down CXADR expression in HeLa cells via siRNA. The knockdown significantly reduced transgene expression mediated by Ad5-Luc, Ad5-Luc@hF-wt, and Ad5-Luc@(+)hF-A1, whereas the efficiency of Ad5-Luc@(+)hF-R2 remained largely unchanged (Fig. 3E). This result aligns with the endocytosis inhibition assay, which indicated that Ad5-Luc@(+)hF-R2 utilizes a non-canonical entry pathway. The CXADR-independent transduction of Ad5-Luc@(+)hF-R2 may underlie its enhanced tropism in certain cell types and its ability to maintain efficacy under pre-existing immunity, warranting further study to elucidate its precise cellular uptake mechanism. In vivo transgene expressions of Ad5@(+)hF To determine if the enhanced transgene expression observed in vitro could be replicated in vivo, mice were intranasally administered with Ad5-Luc@(+)hF and monitored via in vivo imaging. The results revealed a ratio-dependent effect in vivo that diverged from the data in vitro. At lower molar ratios (1:10 and 1:100), Ad5-Luc@(+)hF exhibited an increased bioluminescent signal compared to Ad5-Luc. Conversely, at a high molar ratio (1:1000), the signal intensity in both nasal and lung tissues was notably reduced (Fig.S7). This discrepancy may result from physiological barriers in the respiratory tract, including physical entrapment or rapid clearance of the larger Ad5@(+)hF formed at high molar ratios (Fig. 2), which likely impedes efficient mucosal penetration and pulmonary delivery. Subsequent in vivo studies therefore focused on lower molar ratios, tracking transgene expressions over time after intranasal administration (Fig. 4). While the control (Ad5-Luc@hF-wt) showed no enhancement, both Ad5-Luc@(+)hF-A1 and Ad5-Luc@(+)hF-R2 significantly boosted transgene expressions. In nasal tissue, Ad5-Luc@(+)hF-A1 at a 1:100 ratio produced a substantial enhancement, peaking at 4.8-fold over Ad5-Luc on day 6. In contrast, effective pulmonary delivery was achieved primarily at a 1:10 ratio. At this ratio, Ad5-Luc@(+)hF-A1 reached a peak enhancement of 4.4-fold, while Ad5-Luc@(+)hF-R2 showed a significant increase of up to 2.7-fold. The tissue-specific enhancement of Ad5-Luc@(+)hF is likely mediated by the biophysical properties of the complexes, which vary with the molar ratio. In vivo transgene expressions following intranasal delivery of Ad5@(+)hF. BALB/c Mice (5–6 weeks old, female, n = 4–5) were intranasally administered with Ad5-Luc (5 × 106 IFU per mouse) alone or complexed with hF-wt or (+)hF at molar ratios of 1:10 and 1:100. Bioluminescence imaging was conducted at specific days after administration. PBS was used as a blank control. (A, B) Representative in vivo bioluminescence images. (C, D) Quantification of bioluminescent signals in nasal and lung regions over time based on region-of-interest (ROI) analysis. Data are presented as means ± SEM and were analyzed by two-way ANOVA with multiple comparisons test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 The transgene-enhancing effect was further evaluated under conditions of pre-existing anti-Ad5 immunity, established by administration with empty Ad5 vector (Ad5-null) four weeks prior to intranasal delivery of Ad5-Luc@(+)hF. In this model, Ad5-Luc@(+)hF-A1 significantly enhanced transgene expression in nasal tissue on days 2 and 6, with a peak enhancement of 3.9-fold over the Ad5-Luc control observed at a 1:100 ratio on day 2. Ad5-Luc@(+)hF-R2 also mediated a notable enhancement, achieving a 4.0-fold increase relative to the control on day 10 (Fig.S8). Immune response induced by Ad5@(+)hF under pre-existing immunity The immunogenicity of adenovirus-vectored COVID-19 (Ad5-nCoV) and Hendra virus (Ad5-HeV) vaccines was assessed in a pre-existing immunity model to evaluate the potential of Ad5@(+)hF to enhance immune responses (Fig. 5A). The model was validated by confirming that pre-immunization with Ad5-null elicited high levels of Ad5-specific antibodies by day 30, prior to intranasal administration of Ad5-nCoV or Ad5-nCoV@(+)hF (Fig. 5B). Compared to the naïve (non-pre-immunized) group, pre-existing immunity significantly reduced the antibody titers induced by Ad5-nCoV, with a 6.7-fold decrease observed at day 14 post-vaccination (Fig. 5C). Immune profiling of Ad5-vectored vaccines complexed with (+)hF in mice with pre-existing immunity. BALB/c mice (5–6 weeks old, female, n = 6) with pre-existing immunity were intranasally administered with Ad5-vectored vaccines (5 × 105 IFU per mouse), including Ad5-nCoV and Ad5-HeV, alone of complexed with hF-wt, (+)hF-A1 or (+)hF-R2 at molar ratios of 1:10 and 1:100. (A) Schematic overview of the pre-existing anti-Ad5 immunity model and intranasal immunization strategy. (B) Ad5-specific antibody titers measured at indicated time points after immunization with Ad5-nCoV or Ad5-nCoV@(+)hF in mice with pre-existing immunity. (C) A comparison of SARS-CoV-2 RBD-specific IgG titers between naïve BALB/c mice and mice with pre-existing immunity. Both groups (5–6 weeks old, female, n = 6) were intranasally administered with Ad5-nCoV. (D) Serum SARS-CoV-2 RBD-specific IgG responses induced by Ad5-nCoV complexed with or without (+)hF. (E) Neutralizing antibody titers against SARS-CoV-2 pseudovirus. (F, G) RBD-specific IgA responses in serum (F) and bronchoalveolar lavage fluid (BALF) (G). (H) Serum HeV-G–specific IgG responses induced by Ad5-HeV complexed with or without (+)hF. (I) Neutralizing antibody titers against HeV pseudovirus. (J, K) HeV-G-specific IgA responses in serum (J) and bronchoalveolar lavage fluid (BALF) (K). Data are presented as means ± SEM. Data in C, D and H were analyzed by two-way ANOVA with multiple comparisons test. Data in E-G and I-K were analyzed by one-way ANOVA with multiple comparisons. Pre-Ad5: pre-existing anti-Ad5 immunity. *p < 0.05, **p < 0.01, ***p < 0.001 In the pre-existing immunity model, serum anti-RBD IgG titers were significantly enhanced by Ad5-nCoV@(+)hF, whose size distribution profiles (Fig. S9) resembled those of Ad5-Luc@(+)hF (Fig. 2C and D). Both at 1:10 and 1:100 molar ratios, Ad5-nCoV@(+)hF-A1 elicited a consistent 3.6-fold increase over Ad5-nCoV at days 14 and 28. Ad5-nCoV@(+)hF-R2 (1:100) also boosted IgG levels, showing 3.0-fold and 3.6-fold enhancements at the respective time points (Fig. 5D). These binding antibody responses were corroborated by pseudovirus neutralization assays. At day 28, neutralization titers were 3.0-fold higher for Ad5-nCoV@(+)hF-A1 (1:100) and 5.7-fold higher for Ad5-nCoV@(+)hF-R2 (1:10) compared to the Ad5-nCoV control (Fig. 5E). Mucosal immunity was also evaluated at day 28 by measuring IgA antibody levels in serum and bronchoalveolar lavage fluid (BALF) following intranasal immunization. The IgA titers elicited by Ad5-nCoV@(+)hF were comparable to those induced by Ad5-nCoV (Fig. 5F and G). For Ad5-HeV encoding the viral glycoprotein, (+)hF similarly enhanced antigen-specific antibody responses in the pre-existing immunity model. At day 28, Ad5-HeV@(+)hF-R2 (1:10) elicited a 7.6-fold increase in G-specific antibody titers compared to Ad5-HeV (Fig. 5H). Neutralizing antibody responses were further elevated, with pseudovirus neutralization titers reaching 5.1-fold higher for Ad5-HeV@(+)hF-A1 (1:100) and 8.3-fold higher for Ad5-HeV@(+)hF-R2 (1:10) compared to Ad5-HeV alone (Fig. 5I). Notably, Ad5-HeV@(+)hF-R2 (1:10) enhanced mucosal immunity, eliciting significantly higher serum IgA and an 11.5-fold increase in bronchoalveolar lavage fluid IgA titers at day 28 (Fig. 5J and K). In vivo biocompatibility of Ad5@(+)hF The biocompatibility profile of Ad5@(+)hF was evaluated in a mouse model. Serum biochemical and hematological analyses conducted 7 days post-treatment assessed important physiological markers: alanine aminotransferase (ALT), aspartate aminotransferase (AST), and alkaline phosphatase (ALP) for liver function; urea and creatinine (CREA) for renal function; triglycerides (TG) for lipid metabolism; hemoglobin (HGB) for hematological status; and lactate dehydrogenase (LDH) as a general indicator of tissue or cardiac stress. No statistically significant differences were observed in measured parameter among the groups receiving Ad5-nCoV, Ad5-nCoV@(+)hF-A1 (at 1:10 or 1:100 molar ratios), Ad5-nCoV@(+)hF-R2 (at 1:10 or 1:100 molar ratios), or PBS control (Fig. 6A). These results demonstrated that intranasal delivery of Ad5@(+)hF did not induce detectable hepatotoxicity, nephrotoxicity, or adverse metabolic or hematological effects under the experimental conditions. Furthermore, hematoxylin and eosin (H&E) staining revealed no significant pathological changes, such as abnormal inflammatory cell infiltration, necrosis, hyperplasia, or edema, in the lung, liver, or kidney tissues from any of the groups (Fig.S10). In vivo biocompatibility of Ad5@(+)hF. (A) BALB/c mice were intranasally immunized with PBS, 5 × 10⁵ IFU of Ad5-nCoV or Ad5-nCoV@(+)hF (at molar ratios of 1:10 and 1:100). Seven days post-administration, whole blood was collected for biochemical analysis. (B) BALB/c mice were intranasally immunized with PBS, 5 × 10⁵ IFU of Ad5-HeV or Ad5-HeV@(+)hF (at molar ratios of 1:10 and 1:100). Serum was collected 28 days post-immunization, and IgG antibody responses against mouse ferritin (mF-wt), human ferritin (hF-wt) and the engineered variants (+)hF were measured by ELISA. Data are presented as means ± SEM. Data in A was analyzed by one-way ANOVA with multiple comparisons Human and murine wild-type ferritin heavy chains (hF-wt and mF-wt) share 92.3% sequence homology. The engineered variants (+)hF-A1 and (+)hF-R2 retain high identity to hF-wt (92.3% and 94.0%, respectively), raising the question of whether these modifications could break immune tolerance and induce autoantibodies against the conserved wild-type scaffold. To evaluate this, we measured serum antibody responses against mF-wt, hF-wt, (+)hF-A1, and (+)hF-R2 at day 28 after intranasal immunization. All groups, including the PBS control, Ad5-HeV, and Ad5-HeV complexed with (+)hF-A1 or (+)hF-R2 at 1:10 or 1:100 ratios, showed similarly low antibody levels. In ELISA performed at a 500-fold serum dilution, optical density (OD) values remained below 0.5, with no significant intergroup differences (Fig. 6B). Discussion Our study establishes a synergistic design paradigm that converges engineered endogenous protein nanocages with viral vectors, effectively preserving transduction efficiency of Ad5 both in vitro and in vivo despite pre-existing immunity. By engineering ferritin nanoparticles to present a cationic surface, we developed a modular and biocompatible scaffold that electrostatically assembles with adenoviral vectors into a supramolecular complex (Ad5@(+)hF). This strategy integrates the complementary properties of both components: the ferritin provides programmable surface properties that shield viral epitopes from neutralizing antibodies and modulate cellular entry, while the Ad5 core maintains its intrinsic gene delivery efficiency and immunogenicity. The resulting coherent, synergistic biointerface of Ad5@(+)hF significantly enhances the transgene expression and immunogenicity under pre-existing immunity. This work provides proof of concept that engineering endogenous protein nanoparticles to modulate intermolecular interactions can substantially improve the efficacy of viral vectors. The cationic ferritin nanocage, (+)hF, was engineered through a structure-guided rational design strategy. The central challenge was to introduce substantial surface positive charges while maintaining the protein’s ability to self-assemble into a stable nanocage, as excessive charge repulsion or disruption of key interfacial residues can cause structural instability. To address this, we integrated structural biology data with computational analysis, pinpointing mutable surface residues located away from critical assembly domains. This enabled the genetic encoding of cationic amino acids, which confer the target positive charge while preserving the structural integrity of the assembled architecture. This strategy of intrinsic charge programming contrasts with other established protein modification paradigms. Alternative approaches, such as chemical conjugation of cationic groups [45] or the application of extrinsic polymeric coatings [46], offer valuable routes to alter surface properties. In contrast, the (+)hF design achieves a structurally homogeneous, single-component nanoparticle with a stable and precisely defined cationic surface. A key translational advantage of (+)hF lies in its high biocompatibility and potency. As an engineered variant of the endogenous human ferritin, it exhibits favorable immune tolerance, minimizing the risk of anti-carrier antibody responses [17]. Notably, this platform operates with high efficiency at low doses. For instance, intranasal administration of Ad5-nCoV (5 × 10⁵ IFU per mouse) complexed with 2 ng of (+)hF-R2 at the molar ratios of 1:100, was sufficient to significantly enhance antibody responses in the pre-existing immunity mouse model. The biomimetic design, inherent biocompatibility, and low-dose efficacy of the (+)hF platform collectively highlight its considerable potential for advanced vaccine and delivery systems. Notably, mechanistic studies in cellular models revealed that the two lead variants, (+)hF-A1 and (+)hF-R2, mediated Ad5 entry through distinct cellular pathways. Ad5@(+)hF-A1 remained dependent on energy, cholesterol-sensitive pathways, and CXADR, whereas Ad5@(+)hF-R2 exhibited reduced energy requirements and a diminished reliance on the CXADR-mediated route. The well-defined spherical nanostructures of (+)hF-R2, in contrast to the amorphous morphology of (+)hF-A1, demonstrate that the supramolecular architecture formed with Ad5 directly influences the viral vectors’ endocytic pathway. The substantial size increase of Ad5@(+)hF at high molar ratios may further enhance cellular engagement through increased avidity or by inducing local phase separation at the cell membrane, a phenomenon increasingly recognized in nanoparticle–cell interactions [47]. Unlike (+)hF, protamine cannot form supramolecular complexes with Ad5, indicating that electrostatic interaction alone is insufficient and must be complemented by appropriate size and multivalent spatial organization for effective viral particle bridging. Moreover, the cationic lipid transfection reagent Lipo3000 markedly reduced Ad5-Luc transduction in vitro, suggesting that excessive shielding may instead impair Ad5 transduction. A critical finding was that the formulation parameters conferring optimal Ad5@(+)hF activity in vitro proved suboptimal in vivo. While high molar ratios (1:1000) maximized transduction in cell culture, lower ratios (1:10 and 1:100) were required for effective respiratory delivery. Physicochemical characterization revealed no significant size increase at these lower ratios, and TEM showed only sparse binding of (+)hF on the Ad5 surface without forming a dense coating or large aggregates. Therefore, a classical shielding mechanism that extensively blocks virus receptor binding or antibody neutralization is unlikely. Instead, we propose that this minimal surface modification subtly alters the properties of Ad5, thereby improving its ability to penetrate the respiratory mucosal barrier. In contrast, at a high molar ratio (1:1000), (+)hF induced extensive cross linking of Ad5 particles into large aggregates. This size increase has profound physical implications: while monodispersed Ad5 primarily exhibit Brownian motion [48] and remains suspended in solution, the large aggregates of nanoparticles are prone to gravity-driven sedimentation [49], which likely enhances their interaction with adherent cells and facilitates cellular uptake. However, when delivered intranasally in vivo, the larger size of Ad5@(+)hF are likely retained in the upper airways and cleared by mucociliary action. This discrepancy underscores the necessity of tailoring nanocomplex properties to specific physiological environments. Nevertheless, several limitations of the current (+)hF platform warrant consideration. First, given that the protective efficacy of adenoviral vectors relies on coordinated humoral and cellular immunity, while our findings elucidate the humoral profile of Ad5@(+)hF in the presence of pre-existing immunity, its effect on T-cell immune responses remains to be characterized, which represents a critical avenue for subsequent investigation. Second, from a biophysical perspective, maintaining the stable conformation of the engineered nanocage requires buffer conditions containing high concentrations of salts, such as 300 mM NaCl and 500 mM arginine, which were identified through screening various buffers with adjusted salt concentrations, including HEPES (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid), phosphate buffers and Tris (tris(hydroxymethyl)aminomethane). These components preclude the direct measurement of the particle’s zeta potential using DLS. Third, the cellular delivery mechanisms of Ad5@(+)hF warrant further investigation. It remains unclear whether large Ad5@(+)hF aggregates promote non-specific cellular uptake, whether alternative endocytic pathways are involved, and how modification with a low ratio of (+)hF facilitates intranasal delivery of Ad5. Looking ahead, these constraints delineate clear paths for future development. The integration of AI-driven protein design tools, including deep learning-guided optimization of charge distribution, computational stability prediction, and de novo folding simulation, holds considerable promise for generating next-generation (+)hF variants. Such computationally designed variants could possess enhanced assembly fidelity and colloidal stability, potentially functioning optimally in physiologically relevant buffers. Moreover, the inherent modularity of the ferritin scaffold allows for its fusion with functional motifs, such as cell-targeting ligands or immune modulators. This strategy enables the construction of multifunctional supramolecular delivery carriers with tailored specificity and enhanced potency for diverse applications. Conclusion Employing a structure-guided rational design, we engineered an endogenous ferritin nanocage featuring a precisely programmed cationic surface. This platform serves as a modular scaffold that electrostatically assembles with adenoviral vectors into nanocomplexes capable of circumventing pre-existing antiviral immunity. Our work thus establishes a generalizable blueprint for designing cooperative biomaterial interfaces to overcome fundamental barriers in gene delivery and vaccinology. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements We thank Dr. Ying Yin and Dr. Ruihua Li for experimental assistance and discussions of results. Author contributions Conceptualization: Y.L.Y., J.J.X. and W.C. Supervision: W.C. and J.J.X. Experiments: D.W., X.F.Z. and X.Y.H. Methodology: Y.H.L., M.R.Y., Z.Z. and S.P.W. Formal analysis: Y.L.Y., D.W., Y.H.L., X.D.Z., J.Z. and L.H.H. Writing original draft: Y.L.Y., D.W. and Y.H.L. Writing revision and editing: Y.L.Y., J.J.X. and W.C. All authors reviewed the manuscript. Funding This work was supported in part by National Natural Science Foundation of China (82171818). Data availability The data used and analyzed to support the current study are available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate Ethics Animal housing and experimental procedures were conducted according to the Animal Care and Welfare Ethics Committee of the Academy of Military Medical Sciences. Consent for publication All authors agree for publication. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Yilong Yang, Di Wang and Yaohui Li contributed equally to this work. Contributor Information Yilong Yang, Email: yyl_pku@126.com. Junjie Xu, Email: xujunjie@sina.com. Wei Chen, Email: cw0226@foxmail.com. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The data used and analyzed to support the current study are available from the corresponding author upon reasonable request.