Preclinical evaluation of an mRNA vaccine developed from the first human isolate of bovine H5N1

Summary

Given the global threat posed by H5N1 clade 2.3.4.4b avian influenza, rapid development of effective vaccines is imperative. We design an mRNA vaccine encoding hemagglutinin (HA) from A/Texas/37/2024, the first bovine-to-human strain. In murine models, both wild-type and cleavage-site-modified HA vaccines elicit robust and durable humoral immunity, along with a balanced Th1/Th2 response, conferring complete protection against lethal homologous viral challenge. The vaccine, along with the World Health Organization (WHO)-recommended candidate (A/Astrakhan/3212/2020), elicits cross-clade binding antibody responses and demonstrates improvement against specific clades at a 1 μg dose. Pre-existing H1 immunity does not diminish H5-specific immunogenicity. In avian species, the vaccine also provides full protection against lethal clades (2.3.4.4b and 2.3.4.4h). Formulated with another ionizable lipid, the vaccine elicits responses comparable to benchmark lipid nanoparticles (LNPs) and shows a favorable safety profile in rats. This work establishes a rapidly adaptable mRNA-LNP vaccine prototype for pandemic preparedness against evolving avian influenza threats.

Graphical abstract

Highlights

• •SM102 and DB-Y ionizable lipids deliver H5 mRNA vaccine with high efficiency and safety
• •Vaccine-induced antibody and T cell response protect mice from H5N1 challenge
• •Pre-existing H1 immunity does not diminish H5-specific immunogenicity
• •Vaccine fully protects chicken against clade 2.3.4.4b/h H5 virus challenge

Confronting the global spread of H5N1 avian influenza, Li et al. develop an mRNA vaccine that elicits broad and protective immunity in murine and chicken models. This work provides a rapidly adaptable vaccine prototype for pandemic preparedness.

Introduction

Gs/Gd-lineage highly pathogenic avian influenza (HPAI) A(H5N1) virus has caused unprecedented global spread since its first detection in 1996, posing a severe threat to both wildlife conservation and the poultry industry, alongside persistent zoonotic infections in humans.^1,2 As of December 2024, a total of 954 human infections have been reported in 24 countries worldwide, with a mortality rate of approximately 48.63%.³ The World Health Organization (WHO) has classified H5N1 as a potential high-risk pathogen for the next pandemic.^4,5 Since 2016, clade 2.3.4.4b of the H5N1 virus has rapidly spread within wild bird populations and poultry, with sporadic but sustained detections in various marine and terrestrial mammalian hosts across six continents.^2,6,7,8 Its ability to spread and cause cross-species spillover has raised significant concerns in the field of public health.

In March 2024, the United States reported the first detection of clade 2.3.4.4b H5N1 in dairy cattle,^9,10 followed by the first documented case of cross-species transmission between mammals (dairy cattle) and humans (dairy worker) in April.¹¹ This discovery differs from previous cases of transmission from poultry to humans. Moreover, H5N1 viruses closely associated with dairy cattle have been isolated from various hosts, including poultry, cats, and other livestock (e.g., pigs, sheep, and goats), as well as wild animals like birds and raccoons.^10,12 The ongoing and effective interspecies transmission raises significant concern. By December 2025, 71 confirmed cases of human infections were reported in the United States, with 65 cases related to the exposure to infected dairy cattle or poultry, 2 cases to other animals (e.g., backyard chickens, wild birds, or mammals), and 3 cases with unknown exposure sources. Most human cases presented mild illness, with one fatality reported as of June 2025.¹³ As of now, 1,084 dairy herds across at least 19 states, over 185 million poultry in 51 states, and 17,258 wild birds in 51 jurisdictions have been affected,¹³ underscoring the rising risk of H5N1 spillover to humans. Virological and epidemiological data reveal that the current isolate transmitted from dairy cattle to humans (A/Texas/37/2024) retains typical phenotypic characteristics of avian influenza viruses, which has the same characteristics in receptor binding, pH fusion, or thermal stability.^14,15 There is no evidence to suggest that it has the ability to spread continuously from person to person. However, compared to other clade 2.3.4.4b strains isolated before the outbreak of dairy cattle, A/Texas/37/2024 exhibits stronger virulence, transmissibility, and airborne transmission in mice and ferret models.^6,14,16,17 Following infection, the virus rapidly disseminates throughout the body, with high viral titers found in both respiratory and non-respiratory organs. Additionally, A/Texas/37/2024 carries the E627K adaptive mutation in PB2, a key adaptation mutation linked to mammalian host compatibility, which has also been detected in humans and other mammals infected with other highly pathogenic avian influenza A viruses (such as H7N9 and H9N2).^4,11,18

Currently, 14 H5N1 HPAI vaccines have been approved for human use.⁴ Studies have shown that individuals vaccinated with commercial vaccines can produce cross-neutralizing antibody responses against the HPAI clade 2.3.4.4b isolate A/Astrakhan/3212/2020.¹⁹ However, these vaccines are antigenically distant from A/Texas/37/2024, and their effectiveness against the current epidemic H5N1 infection remains uncertain. In addition, three candidate vaccine strains (IDCDC-RG71A, IDCDC-RG78A, and NIID-002), which are genetically closer to circulating H5N1 HPAI viruses, are currently under clinical evaluation.²⁰ Given the considerable uncertainty regarding the evolution and mutation of H5N1 viruses, and the remarkable potential demonstrated by the mRNA platform in the rapid production and development of coronavirus disease 2019 (COVID-19) vaccines,²¹ mRNA vaccines are considered an ideal tool for responding to HPAI outbreaks. Previous studies have shown that mRNA vaccines targeting clade 2.3.4.4b of H5 (e.g., A/Astrakhan/3212/2020 or A/American wigeon/South Carolina/22-000345-001/2021) can induce robust neutralizing antibody responses and HA-specific CD8⁺ T cell responses in mice, providing complete protection against the 2.3.4.4b H5N1 HPAI viruses in ferret models.^22,23 However, A/Texas/37/2024 differs from A/Astrakhan/3212/2020 in four amino acid substitutions (L104M, L115Q, T195I, and V210A) in the HA1 region, and no protective comparison data are available. A recent deep mutation scan study of H5 HA showed that the mutation at the L115Q site increased the escape of the virus to polyclonal sera from mice or ferrets vaccinated or infected with 2.3.4.4b virus.²⁴ Therefore, novel H5 vaccine candidates should incorporate the aforementioned mutation sites to avoid potential viral immune evasion.

In our previous study, we successfully engineered an mRNA vaccine platform, which has shown significant promise in advancing vaccine technology. The COVID-19 vaccine developed utilizing this platform exhibited robust protective efficacy in phase III clinical trials, culminating in the granting of Emergency Use Authorization (EUA) for its distribution in China.²⁵ In view of the ongoing public health threat and pandemic potential posed by the newly emerging H5N1 HPAI virus, this study developed an mRNA candidate vaccine that matches the current epidemic strain A/Texas/37/2024 (abbreviated as Texas) HA and evaluated its immunogenicity and protective effect in mice and poultry. Our results demonstrated that the mRNA vaccines encoding the HA protein, both in its wild-type (WT) and cleavage-defective mutant (Mut) forms, induced durable and robust antibody responses. The Mut mRNA vaccine also elicited strong T cell responses and provided protection against lethal challenges, with the H5 clade 2.3.4.4b virus in both species (mice and chickens). Specifically, the Mut vaccine triggered a broader antibody response, significantly outperforming the WHO-recommended vaccine candidate (A/Astrakhan/3212/2020, abbreviated as Ast20). Pre-existing H1 immunity did not compromise H5-specific responses, and mice with prior H1 immunization achieved equivalent protective efficacy and survival against lethal H5N1 challenges. Additionally, we developed an ionizable lipid LQ-016-DB-Y (abbreviated as DB-Y) for use in lipid nanoparticles (LNPs) formulation. The vaccine formulated with DB-Y induced similar immune responses in mice. Our findings offer valuable insights for developing of vaccine reserves against potential H5N1 HPAI pandemics.

Results

H5 Texas mRNA design and mRNA-LNP vaccine production

Building on our previous COVID-19 vaccine platform, which received EUA for marketing in China, this study aims to verify whether the matched bovine H5N1 strain can induce an effective immune response. To this end, we developed an mRNA vaccine based on this platform, targeting the first human isolate of bovine H5N1 (Texas). Two mRNA vaccines were synthesized encoding either full-length WT HA or a cleavage-defective mutant HA (Mut) (Figure 1A). The highly pathogenic multi-basic cleavage site REKRRKR in Mut was modified to REKR to achieve reduced virulence and increased the stability of the protein in a vaccine context. All sequences were optimized by codon adaptation index (CAI) to promote expression. Capillary electrophoresis confirmed high-quality mRNA preparations, obtained by in vitro transcription (IVT), with RNA integrity of 96.9% for WT and 97.1% for Mut (Figure 1B). mRNAs were then transfected into BHK-21, HeLa, and HEK293T cells, and translation of the HA antigens was detected by western blotting. As expected, in vitro expression analysis showed robust HA protein production in all tested cell lines (Figure 1C). The Mut exhibited significantly reduced proteolytic processing compared to WT, confirming the expression impact of the polybasic cleavage site deletion (Figure 1C). Flow cytometry was further optimized to assess the cell surface expression of both mRNA constructs. The HA protein expression level in BHK-21 cells is correlated with mRNA transfection level. After transfection with the highest doses of WT and Mut mRNAs, 89.07% and 90.79% of the cells, respectively, exhibited positive HA staining (Figure 1D). These results confirm that full-length WT and Mut HA proteins are successfully and conformationally exposed on the cell membrane following mRNA transfection.

Figure 1.

Design, in vitro characterization, and LNP formulation of H5 Texas mRNA-LNP vaccines

(A) Schematic of wild-type (WT) and cleavage-defective mutant (Mut) HA mRNA vaccines. Mutations are indicated by labeled amino acids above the column.

(B) Capillary electrophoresis (Fragment Analyzer) analysis of mRNA integrity. Percentages indicate RNA integrity values (RIN).

(C) Western blotting analysis of HA protein expression in transfected BHK-21, HeLa, and HEK293T cells.

(D) Flow cytometry analysis of cell-surface HA expression in BHK-21 cell line. Cells were transfected with varying doses (50, 100, 250, and 500 ng/μL) of WT or Mut mRNA, and HA expression was detected via indirect immunofluorescence staining followed by flow cytometry (CR6261 as primary antibody). Percentages indicate HA-positive cells. Data are presented as the mean ± SD, and n = 3 biological replicates for each transfection dose.

(E) Physicochemical properties of H5 Texas-WT/Mut HA mRNA-LNP containing SM102. Dynamic light scattering (DLS) analysis of mRNA-LNPs, showing Z-average size and polydispersity index (PDI). Encapsulation efficiency (E.E.) was determined using the RiboGreen RNA assay.

(F) Cryo-EM images of LNPs (scale bars: 50 nm).

We engineered LNPs containing the SM102 ionizable lipid for our mRNA vaccines. Both mRNAs were effectively encapsulated with favorable physicochemical properties including particle sizes of 76.9 nm (WT) and 82.4 nm (Mut) with polydispersity index (PDI) of 0.075 and 0.062, respectively (Figure 1E). Encapsulation efficiencies exceeded 90% for both formulations as determined by RiboGreen assay. Cryo-electron microscopy confirmed the spherical morphology and uniform size distribution of both LNP formulations (Figure 1F).

In parallel with our evaluation using the SM102, we also aimed to develop lipid excipients that could further optimize delivery performance and safety. Based on our established delivery platform, we characterized a developed ionizable lipid,^26,27 LQ-016-DB-Y (DB-Y), which was incorporated into our LNP formulation to assess its physical properties, transfection efficiency, and immunogenicity. The physical properties of DB-Y are shown in Table S1. The pKa of DB-Y-formulated LNPs was comparable to that of SM102, indicating similar stability and delivery efficiency. A higher generalized polarization (GP) value also indicated more ordered lipid packing in DB-Y LNPs. We then compared DB-Y and SM102 using a luciferase expression model in mice. Mice were intramuscularly administered a single 5 μg dose of LNP encapsulating firefly luciferase (Fluc) mRNA. Whole-body imaging was performed for 14 days post-injection, and Fluc expression at the injection site was quantified. DB-Y consistently produced higher Fluc expression at the injection site throughout most time points (Figure S1A). These results indicate that DB-Y enables efficient in vivo delivery of mRNA. We then used DB-Y to formulate WT and Mut mRNAs into LNPs and evaluated their physicochemical properties. Particle sizes were 79.7 nm (WT) and 73.8 nm (Mut), with PDIs of 0.042 and 0.065, respectively (Figure S1B). Both formulations exhibited encapsulation efficiencies exceeding 95%. Cryo-electron microscopy (Cryo-EM) revealed that both LNPs exhibited spherical morphology and uniform size distribution (Figure S1C). These characterization data confirm the successful development of two distinct sequences but well-formulated mRNA-LNP vaccine candidates against H5N1 HAPI, with the Mut construct showing reduced HA proteolytic processing while maintaining robust expression.

The H5 Texas mRNA-LNP vaccines elicit antibody responses and protect mice against lethal challenge of an H5 clade 2.3.4.4b virus

Next, BALB/c mice were vaccinated with either 1 or 5 μg of the vaccines or a placebo, following a prime-boost schedule with a 4-week interval (Figure 2A). Serum samples were collected to quantify antibody levels using enzyme linked immunosorbent assay (ELISA) and microneutralization (MN) assays. Both types and doses of vaccines induced high levels of immunoglobulin G (IgG) antibodies that bound to the matched HA protein of Texas after the prime, with a significant enhancement following the boost (Figure 2B). Notably, a dose-dependent response was observed, with higher vaccine doses eliciting higher levels of binding IgG antibodies (Figure 2B). To evaluate the protective capacity of the elicited antibodies, we also evaluated the neutralization of a homologous 2.3.4.4b H5 strain, A/ostrich/China/HB/2024, which is 98.06% amino acid identical to HA sequence of Texas (Table 1). Both types of Texas-based mRNA-LNP vaccines induced high levels of antibodies that neutralized this homologous strain in a dose-dependent manner following the booster (Figure 2C). This suggests that our vaccines can provide protection against H5 viruses in clade 2.3.4.4b. The mRNA delivered by DB-Y also elicited a similar immune response. The serum from the immunized mice contained high levels of both binding and neutralizing antibodies against 2.3.4.4b H5 strains (Figures S2A–S2C).

Figure 2.

H5 Texas mRNA-LNP elicits robust immune responses and protect mice against lethal challenge of an H5 clade 2.3.4.4b virus

(A) Vaccination schedule. Mice were vaccinated (i.m.) with 1 or 5 μg mRNA-LNP (SM102) encoding WT, Mut, or placebo using in a prime-boost schedule with an interval of 4 weeks. Blood samples (n = 8) were collected from mice 2 weeks after prime and boost (weeks 2 and 6). Spleens of 5 μg-Mut-vaccinated mice (n = 8) were collected 1 week post-boost (week 5). Four weeks post-boost (week 8), Mut-vaccinated mice (n = 8) were challenged with 5 × LD₅₀ of an H5 clade 2.3.4.4b virus (A/ostrich/China/HB/2024). WT-vaccinated mice (n = 8) were used for long-term protective monitoring.

(B) Binding IgG of serum collected 2 weeks after prime and boost (weeks 2 and 6), reactive to the Texas recombinant secreted HA protein, was measured by ELISA.

(C) 50% neutralization titers of unmatched clade 2.3.4.4b H5 virus (A/ostrich/China/HB/2024) in serum collected 2 weeks after prime and boost (weeks 2 and 6).

(D) Long-term binding IgG of serum collected 42, 80, 120, 180, 240, 300, and 365 days post-prime, reactive to the Texas recombinant secreted HA protein, was measured by ELISA.

(E) Long-term 50% neutralization titers of unmatched clade 2.3.4.4b H5 virus (A/ostrich/China/HB/2024) in serum collected 42, 80, 120, 180, 240, 300, and 365 days post-prime.

(F) Survival curves. Data were analyzed using a log rank test.

(G) Body weight changes.

Data are presented as the mean ± SD; p values are analyzed with t test (#p < 0.05, ∗∗/##p < 0.01, ∗∗∗/###p < 0.001). ∗ (D and E) indicates significant differences between group 1 μg and 5 μg at the same sampling time. # (D and E) indicates significant differences between group 80, 120, 180, 240, 300, and 365 days and 42 days post-prime at the same immunizing dose. Titers of placebo group were below detection limit (B–E).

Table 1.

The sequence identity analysis of HA proteins in H5 virus

H5 virus	Sequence identities with HA protein of A/Texas/37/2024
	Nucleotide (%)	Amino acid (%)
A/Viet Nam/1203/2004 (clade 1)	89.34	91.39
A/Indonesia/5/2005 (clade 2.1.3.2)	89.16	91.74
A/bar-headed goose/Qinghai/5/2005 (clade 2.2)	89.36	91.40
A/Hubei/1/2010 (clade 2.3.2.1a)	88.62	91.37
A/Anhui/1/2005 (clade 2.3.4)	91.02	92.61
A/Astrakhan/3212/2020 (clade 2.3.4.4b)	97.48	98.94
A/ostrich/China/HB/2024 (clade 2.3.4.4b)	96.42	98.06
A/goose/GuangDong/1189/2023 (clade 2.3.4.4b)	97.01	98.06
A/duck/Fujian/QG4/2023 (clade 2.3.4.4h)	89.44	92.78

Since there was no significant difference in immune response between the WT and Mut vaccines, and HA-Mut proteins have been reported to reduce pathogenicity and are widely used in the inactivated H5N1 vaccine stockpile,^28,29 we selected to use the Mut vaccine for the subsequent assays. To assess the long-lasting potency of the vaccine, serum IgG titers induced by two lipid-delivered vaccines were monitored for 1 year after prime vaccination. The serum IgG titers remained at high levels throughout the observation period (Figures 2D and S2C). In the 5 μg vaccine group, IgG titers declined by 5.2-fold after the acute immune phase (80 days post-prime) but remained at superior levels for 1 year. Meanwhile, the 1 μg vaccine group maintained stable binding IgG titers in this monitoring period (Figure 2D). DB-Y-delivered vaccine produced a comparable long-lasting protective effect (Figure S2C). Although the neutralizing antibody titers elicited by two-dose vaccines against the unmatched strain declined by 4.5- to 6.3-fold at 120 days post-prime compared to those at 2 weeks post-boost, moderate titer levels were maintained for over half year (Figure 2E). These results demonstrate that our vaccines are capable of inducing and maintaining high antibody levels for an extended period after vaccination.

To assess the protective efficacy of the vaccine, we challenged mice with a phylogenetically distinct H5N1 clade 2.3.4.4b strain (A/ostrich/China/HB/2024) to evaluate the cross-protection provided by the mRNA-LNP vaccines. Compared with the vaccine strain Texas, A/ostrich/China/HB/2024 virus harbors Q122L, N158D, and N193K mutations, which are prone to antigenic escape²⁴ (Figure S3). We confirmed that it is lethal in mice and has a strong virulence (Figures S2D and S2E). Four weeks post the booster, mice immunized with either 1 μg or 5 μg doses of the Texas Mut vaccine or placebo were intranasally challenged with 5 × LD₅₀ of A/ostrich/China/HB/2024. All vaccinated mice survived in the lethal challenge, while mice in the placebo group reached the clinical endpoint by 12 days post-challenge (Figures 2F and 2G). Additionally, mRNA vaccines delivered via DB-Y similarly provided protective immunity against the H5N1 challenge (Figures S2F and S2G). These results demonstrate that our vaccines are both immunogenic and protective in mice, particularly offering cross-protection against a phylogenetically distinct H5N1 clade 2.3.4.4b strain.

Comparison of H5 Texas and Ast20 mRNA-LNP vaccines showed superior broad-spectrum antibody responses with Texas vaccine

The HA sequence of Ast20, a WHO-recommended candidate vaccine virus (CVV) for H5N1 2.3.4.4b strains, shares 98.94% amino acid identity with Texas, differing by only four residues (Table 1). And, this virus serves as the primary antigen in the currently available commercial H5 vaccines. To evaluate the advantages of Texas-HA as antigen, we compared the antibody responses in mice induced by mRNA vaccines encoding either Texas or Ast20 HA, both of which were modified with the same mutations (multi-basic cleavage site REKRRKR→REKR) and codon optimization (Figure 3A).

Figure 3.

Comparison of vaccine efficacy between H5 Texas Mut vaccine and H5 Ast20 Mut vaccine in mice

(A) Schematic of cleavage-defective mutant (Mut) HA mRNA vaccines.

(B) Western blotting analysis of HA protein expression in transfected BHK-21 cells. Cells were transfected with varying doses (50, 100, 200, and 500 ng/μL) of Texas Mut or Ast20 Mut mRNA.

(C) Mice were vaccinated (i.m.) with 1 or 5 μg Texas/Ast20 mRNA-LNP (SM102) encoding Mut HA or placebo using in a prime-boost schedule with an interval of 4 weeks. Blood samples (n = 8) were collected from mice 2 weeks after prime and boost. Binding IgG of serum, reactive to the Texas HA protein, was measured by ELISA. Titers of placebo group were below detection limit.

(D) 50% neutralization titers of pseudotyped virus (A/Texas/37/2024) in serum collected 2 weeks post-boost (week 6). Titers of placebo group were below detection limit.

(E and F) Binding IgG of 1 μg Texas vaccine group (E) and Ast20 vaccine group (F) was measured by ELISA, reactive to the HA of clade 1 virus (A/Viet Nam/1203/2004), clade 2.1.3.2 virus (A/Indonesia/5/2005), clade 2.2 virus (A/bar-headed goose/Qinghai/5/2005), clade 2.3.2.1a virus (A/Hubei/1/2010), clade 2.3.4 virus (A/Anhui/1/2005), clade 2.3.4.4b virus (A/Astrakhan/3212/2020), and clade 2.3.4.4b virus (A/Texas/37/2024).

(G) Radar chart of broad-binding IgG levels elicited by 1 μg Texas vaccine and Ast20 vaccine.

(H) Vaccination schedule. Mice were vaccinated (i.m.) with 1 or 5 μg mRNA-LNP (SM102) encoding Texas HA-Mut, Ast20 HA-Mut, or placebo. Four weeks post-vaccination (week 4), mice were challenged with 50 × LD₅₀ of A/ostrich/China/HB/2024. Virus load of organs in vaccinated mice 1, 3, 5 days post-infection determined using plaque assays on MDCK cells.

(I–L) Virus load of lung (I), heart (J), kidney (K), and brain (L). The dashed horizontal line indicates the assay limit of detection.

Data are presented as the mean ± SD; p values are analyzed with t test (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001) (C–L). ∗/∗∗∗ (E and F) indicates significant differences between other HA proteins and Texas (E) or Ast20 (F) protein. ∗ (E and F) indicates significant differences between other HA proteins and Texas (E) or Ast20 (F) protein. ∗ (I–L) indicates significant differences between placebo and vaccination groups in the same day.

Both constructs exhibited equivalent HA expression in BHK-21 cells, as confirmed by western blotting (Figure 3B). BALB/c mice were immunized with prime-boost doses (4-week interval) of either 1 μg or 5 μg of the vaccines. Both vaccine groups induced dose-dependent IgG antibodies against Texas HA post-boost, with the 5 μg Texas group achieving significantly higher binding titers against the matched Texas HA (Figure 3C). As for the neutralization, we performed pseudovirus neutralization assays using serum from mice that had received booster vaccinations. For these tests, we employed vesicular stomatitis virus (VSV) modified to express the HA and NA proteins from Texas. Texas mRNA vaccine induced higher neutralizing antibody titers against Texas pseudoviruses, indicating superior immunogenicity against circulating strain (Figure 3D).

Antibody breadth was assessed across various H5 clades. Both vaccines generated cross-reactive IgG antibodies against heterologous HAs (Clade 1, 2.1.3.2, 2.2, 2.3.2.1a, and 2.3.4 strains) (Figures 3E and 3F). However, the Texas-based vaccine, at a low 1 μg dose, induced a stronger IgG response to the clade 1, 2.2, 2.3.2.1a, and 2.3.4 strains than the Ast20 vaccine did (Figure 3G). Although the advantage in the high-dose group was diminished, it still demonstrated a significant improvement against clade 1 strain (Figures S4A–S4C). Furthermore, these antibody responses were correlated with the amino acid identity of the HAs (Table 1), showing significantly higher binding IgG levels in more closely related clades (Figures 3E, 3F, S4A, and S4B).

Next, we challenged mice with A/ostrich/China/HB/2024 to evaluate the vaccines’ abilities to limit viral replication and systemic spread. We analyzed the viral load and replication kinetics of the challenge virus across multiple organs in a mouse infection model (Figures S4D and S4E). The lung was identified as the primary site of robust viral replication, exhibiting a complete replication curve with the highest titers. In contrast, viral presence was detected in the heart, kidney, and brain only between days 5 and 7 post-infection, while no replication was observed in the liver or spleen (Figures S4D and S4E). Four weeks following the vaccination, mice were intranasally challenged with a high viral dose of 50 × LD₅₀ of A/ostrich/China/HB/2024, and detection of virus load in the lungs, heart, kidneys, and brain was carried out (Figure 3H). Both vaccines effectively suppressed viral replication in organs even with a single immunization dose of 1 μg, with all mice testing negative for infectious virus (Figures 3I–3L). The protective effects of both vaccines were further confirmed in survival rates. Mice that received boost immunization were all completely survived and maintained stable body weight after infection with A/ostrich/China/HB/2024 at a dose of 50 × LD₅₀ (Figures S4F and S4G). Moreover, in a long-term protection experiment, a vaccination dose of merely 1 μg was adequate to fully eliminate the virus in the lung by the third day after the challenge, which occurred 22 weeks after the booster (Figure S4H).

These findings confirm that Texas-based vaccine not only triggers stronger antibody responses against A/Texas/37/2024 strains but also offers superior cross-clade protection. Both vaccines offer complete protection against the same clade virus up to 6 months post-vaccination and efficiently inhibit the initial viral replication during early infection.

The H5 Texas mRNA-LNP vaccines elicit robust T cell responses

Cellular immunity plays a critical role in controlling viral infections by producing effector cytokines and eliminating virus-infected cells.³⁰ To evaluate antigen-specific T cell responses in H5-Texas-mRNA-LNP-immunized group, we quantified cytokine-secreting splenocytes in 5-μg-Mut-vaccinated mice via ELISpot assay 7 days after boost immunization. Compared to the placebo group, 5 μg Mut vaccine induced significantly elevated frequencies of interferon (IFN)-γ⁺, tumor necrosis factor (TNF)-α⁺, and interleukin (IL)-2⁺ splenocytes (Figures 4A–4C). Similar results were observed in the DB-Y lipid-formulated vaccine group (Figures S5A–S5C). We further confirmed the T cell responses by flow cytometry; the gating strategies were shown in Figure S5D. Flow cytometry further confirmed that 5 μg Mut vaccine elicited robust TNF-α⁺ and IL-2⁺ CD4⁺ (Figures 4D–4F) and IFN-γ⁺, TNF-α⁺ and IL-2⁺ CD8⁺ (Figures 4G–4I) T cell responses. The DB-Y formulation also triggered substantial TNF-α⁺ and IL-2⁺ CD4⁺ T cells and IFN-γ⁺, TNF-α⁺, and IL-2⁺ CD8⁺ T cells (Figures S5E–S5J). To assess Th1/Th2 polarization, we analyzed serum IgG subclass profiles (IgG1 and IgG2a) against HA post boost. Two doses of Mut group demonstrated high titers of HA-binding IgG2a and IgG1 antibodies, with an IgG2a/IgG1 ratio approaching 1.0 (Figures 4J and S5K), indicative of a balanced Th1/Th2 response.

Figure 4.

Cellular immune responses elicited by H5 Texas Mut mRNA-LNP (SM102) in mice

(A–C) The number of splenocytes secreting IFN-γ (a), TNF-α (b), and IL-2 (c) was detected by ELISpot.

(D–F) The percentage of CD4⁺ T cells secreting IFN-γ (D), TNF-α (E), and IL-2 (F) was quantified by flow cytometry.

(G–I) The percentage of CD8⁺ T cells secreting IFN-γ (G), TNF-α (H), and IL-2 (I) was quantified by flow cytometry.

(J) IgG1 and IgG2a subclass Abs against Texas HA were determined, and the ratio of IgG2a/IgG1 was calculated.

Data are presented as the mean ± SD, and n = 8 biological replicates for each group (A–J). p values are analyzed with t test (ns p > 0.05, ∗p < 0.05, ∗∗∗p < 0.001) (A–J).

H1 pre-existing immunity does not compromise H5 Texas mRNA-LNP vaccine immunogenicity or protective efficacy in mice

Given the widespread circulation of seasonal influenza viruses and routine vaccination programs, most individuals have pre-existing immunity to seasonal influenza, which could potentially impact the immune response and protective efficacy of H5 mRNA vaccines through original antigenic sin (OAS).³¹ To investigate this, we tested the Texas Mut vaccine in mice with and without prior immunization using an mRNA vaccine encoding the HA protein of the A/California/07/2009 (H1N1) virus. Mice were first vaccinated with the H1 vaccine for the first 4 weeks, followed by H5 vaccine immunization and an intranasal challenge with 5 × LD₅₀ of A/ostrich/China/HB/2024 4 weeks later (Figure 5A). As expected, mice previously immunized with the H1 vaccine exhibited binding antibodies against the HA protein of A/California/07/2009 (Figure 5B). The Texas Mut vaccine induced serum-binding IgG antibodies that recognized the Texas HA, with antibody levels similar between mice with and without prior H1 immunization (Figure 5C). Moreover, the immunity induced by H5 also promoted the antibodies against H1, though this effect is not significant (Figure 5B). These results indicate that pre-existing immunity to the H1 virus does not hinder the immune response generated by the H5 Texas vaccine in mice. Furthermore, both H1-primed and unprimed groups were fully protected against lethal H5 virus challenge (Figures 5D and 5E). This protection was obtained after a single dose of H5 mRNA-LNP vaccine administration, which has very important reference value.

Figure 5.

Effect of H1 prior immunization on the immunogenicity and protective ability of H5 Texas Mut mRNA-LNP vaccines

(A) Vaccination schedule. Mice were vaccinated (i.m.) with 5 μg mRNA-LNP (SM102) encoding the HA protein of A/California/07/2009 H1N1 (CA07) virus (n = 5) or placebo (n = 8) to set up the prior immunization model. After 4 weeks, five mice from H1 group and three mice from placebo group were vaccinated with 5 μg Texas Mut mRNA vaccine. Blood samples were collected from mice 2 weeks post-vaccination (weeks 2 and 6). After 4 weeks of Texas Mut vaccination, all mice were challenged with 5 × LD₅₀ of an H5 clade 2.3.4.4b virus (A/ostrich/China/HB/2024) and then monitored for 14 days post-infection.

(B) Binding IgG of serum collected 2 weeks post-vaccination (weeks 2 and 6), reactive to the A/California/07/2009 recombinant secreted HA protein, was measured by ELISA.

(C) Binding IgG of serum collected 2 weeks post-immunization (weeks 2 and 6), reactive to the Texas recombinant secreted HA protein, was measured by ELISA.

(D) Survival curves. Data were analyzed using a log rank test.

(E) Body weight changes. Data are presented as the mean ± SD.

Data are presented as the mean ± SD; p values are analyzed with t test (ns p>0.05, ∗∗∗p < 0.001) (B and C). Titers of placebo group were below detection limit (B and C).

The H5 Texas mRNA-LNP vaccine is immunogenic and cross-protective in chicken

H5N1 HPAIV has spread widely among both poultry and cattle, causing millions of domestic poultry deaths.⁸ To evaluate the protective efficacy of our vaccine in poultry, specific pathogen-free (SPF) chickens were administered 30 μg of the Mut mRNA vaccine or a placebo, following a prime-boost regimen with a 1-week interval (Figure 6A). Serum samples were collected to assess antibody levels using ELISA and MN assays. The results show that the vaccine induced IgG antibodies that bound to the HA protein of Texas 1 week post the prime, with a significant increase observed 2 weeks post the boost (Figure 6B). Additionally, we tested serum collected post-boost for its ability to neutralize the 2.3.4.4b H5 strain, A/ostrich/China/HB/2024. The vaccine demonstrated neutralizing activity against this H5N1 strain (Figure 6C). These results indicate that the vaccine successfully induces an effective antibody response in chickens.

Figure 6.

H5 Texas Mut mRNA-LNP vaccine is immunogenic and protective against clade 2.3.4.4 viruses in chickens

(A) Vaccination schedule. SPF chickens were vaccinated (i.h.) with 30 μg mRNA-LNP (SM102) encoding Mut or placebo using in a prime-boost schedule with an interval of 1 week. Blood samples (n = 10) were collected from chickens at 1 week post-prime (day 7) and 2 weeks post-boost (day 21). Vaccinated chickens (n = 10) were challenged (i.n.) with a H5N1 clade 2.3.4.4b virus (A/goose/GuangDong/1189/2023) and a H5N6 clade 2.3.4.4h virus (A/duck/Fujian/QG4/2023) at 2 weeks post-boost (day 21) and then monitored for 9 days post-infection. Oropharynx and cloacal swab samples were collected at 5 days post-challenge (day 26).

(B) Binding IgG of serum, reactive to the Texas recombinant secreted HA protein, was measured by ELISA.

(C) 50% neutralization titers of unmatched H5 clade 2.3.4.4b virus (A/ostrich/China/HB/2024) in serum collected at 2 weeks post-boost (day 21).

(D and E) At 2 weeks post-boost (day 21), all groups of chickens (n = 10) were challenged (i.n.) with 10⁶ EID₅₀ of two viruses respectively. Survival rates of chickens were measured daily for 9 days post-challenge.

Data are presented as the mean ± SD; p values are analyzed with t test (∗∗p < 0.01, ∗∗∗p < 0.001) (B and C). Titers of placebo group were below detection limit (B and C).

H5N1 clade 2.3.4.4b and H5N6 clade 2.3.4.4h are distinct genetic lineages of avian influenza viruses that exhibit significant variability and are widely circulating among wild birds and poultry populations (Table 1). Clade 2.3.4.4h exhibited higher proportions of documented human infections (8.89%) compared to the most frequently sampled strain 2.3.4.4b (0.19%).³² To assess the cross-protective efficacy of the vaccine, chickens that had been immunized were challenged with the H5N1 clade 2.3.4.4b virus (A/goose/GuangDong/1189/2023) and the H5N6 clade 2.3.4.4h virus (A/duck/Fujian/QG4/2023). The HA sequences of these two viruses are 98.06% and 92.78% amino acid identical to Texas, respectively (Table 1). Compared with the vaccine strain Texas, A/goose/GuangDong/1189/2023 has mutations of Q122L and N158D, while A/duck/Fujian/QG4/2023 has mutations such as I121T, Q122R, P128S, and N193K, which are prone to antigenic escape (Figure S3). Following a lethal viral challenge, all chickens in the vaccine group survived, whereas all placebo chickens succumbed to infection within 2 days when challenged with the clade 2.3.4.4b virus (Figure 6D) and within 3 days when challenged with the clade 2.3.4.4h virus (Figure 6E). Oropharyngeal and cloacal swabs were collected 5 days post-challenge to monitor virus shedding. No swabs were collected from chickens that died following the challenge. Only one chicken in the vaccine group, challenged with both viruses, showed detectable virus in both oropharyngeal and cloacal swab samples, highlighting the vaccine’s potential to reduce disease severity (Table 2). These findings collectively demonstrate that the vaccine is effective across different species, stimulating antibody responses and providing cross-protection against diverse H5 clade 2.3.4.4 viral challenges in chickens.

Table 2.

Virus shedding after H5 virus challenge of chickens

Group	Challenge virus	Oropharyngeal and cloacal swab in 5 dpc (virus shedding number/total number)a
H5 Texas-Mut	A/goose/GuangDong/1189/2023 (H5N1 2.3.4.4b)	1/10
Placebob		NA
H5 Texas-Mut	A/duck/Fujian/QG4/2023 (H5N6 2.3.4.4h)	1/10
Placebob		NA

^aThe oropharyngeal and cloacal swab samples were collected at 5 days post-challenge. Virus positivity or shedding was determined by inoculating each swab solution into three eggs of 10-day-old specific-pathogen-free chicken embryos. dpc, days post-challenge.

^bNA, not applicable due to the death of chickens.

Toxicological evaluation of H5 Texas mRNA-LNP vaccine in SD rats

The safety of vaccine candidates is a critical consideration for their potential clinical application. The toxicological evaluation of two H5N1 mRNA vaccine candidates (DB-Y and SM102) was systematically evaluated in Sprague-Dawley (SD) rats by monitoring body weight dynamics, local reactogenicity of injection site. (Figure 7A). All vaccinated groups exhibited a transient suppression of weight gain within the first 24 h post-vaccination, followed by a recovery starting at 48 h (Figure 7B). Local reactogenicity showed dose-dependent swelling at the injection site. Swelling was transient, fluctuating over the first 3 days, but resolved entirely by day 7, indicating complete resolution of local reactions (Figure 7C). Notably, even at high doses, neither vaccine formulation induced persistent physiological disturbances, highlighting a favorable safety profile and indicating that the local inflammatory response was self-limiting.

Figure 7.

Acute toxicological evaluation of two lipid-delivered vaccines in SD rats

(A) Vaccination schedule. SD rats were vaccinated (i.m.) with 50 or 300 μg SM102, or DB-Y delivered mRNA vaccines encoding Mut Texas HA protein or placebo. Blood samples (n = 8) were collected at 6 h and 2 weeks post-injection (day 14).

(B) The body weight growth trend.

(C) Local swelling grading of injection site.

(D and E) White blood cell (D) and granulocyte (E) counts were analyzed in automatic animal blood cell analyzer.

(F and G) IL-6 (F) and IFN-γ (G) cytokines were measured using multiplex immunoassay technology.

(H) Binding IgG of serum, reactive to the Texas recombinant secreted HA protein, was measured by ELISA.

Data are presented as the mean ± SD; p values are analyzed with t test (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001) (B–H). Titers of placebo group were below detection limit (H).

Hematological analysis revealed characteristic alterations in blood parameters, consistent with vaccine-induced immune activation. Specifically, a transient leukopenia was observed within 6 h post-vaccination, with a subsequent recovery in white blood cell (WBC) counts by 14 days (Figure 7D), consistent with leukocyte migration to immune activation sites. Additionally, a pronounced early granulocytic response was observed across all vaccine groups, (Figure 7E), reflecting rapid engagement of the innate immune system. These hematological changes align with known immune responses to mRNA-based vaccines, including those developed for COVID-19.³³ Cytokine profiling demonstrated typical immune activation, with significantly elevated IL-6 levels and a dose-dependent increase in IFN-γ at 6 h post-vaccination (Figures 7F and 7G). These cytokine responses normalized by day 14, further confirming that the immune activation produced robust yet transient systemic effects. Other cytokines, such as IL-2, IL-4, IL-5, IL-10, IL-13, TNF-α, and granulocyte-macrophage colony-stimulating factor (GM-CSF), showed no significant changes (data not shown), consistent with immune signatures observed in COVID-19 mRNA vaccine platforms. Serological analysis conducted at day 14 confirmed the successful induction of antigen-specific IgG antibodies by both vaccine formulations, validating their capacity to elicit protective humoral immunity (Figure 7H).

Further histological evaluation of the injection site and major organ systems, including heart, spleen, lung, kidney, and liver tissues, revealed no pathological abnormalities (Figure S6). Notably, a reduction in lymphocytes in the spleen was observed, which may be attributed to immune activation by the mRNA and its expressed antigen, leading to the migration of immune cells to lymph nodes or inflammatory sites. The observed splenic modifications were consistent with safety data from approved mRNA vaccine platforms for COVID-19,³⁴ suggesting biological relevance and a favorable safety margin for these vaccine candidates under the conditions tested.

Discussion

The unprecedented cross-species transmission of H5N1 clade 2.3.4.4b between dairy cattle and human has raised significant public health concerns, necessitating the rapid development of vaccines tailored to emerging strains.¹¹ The efficacy of current HPAI vaccines, primarily based on historical HA antigens such as A/Viet Nam/1203/2004 (clade 1, VN04) and A/Astrakhan/3212/2020 (clade 2.3.4.4b, Ast20),⁴ against the newly emerged Texas strain is a cause for concern due to antigenic divergence (91.39% and 98.94% HA amino acid identity, respectively) (Table 1). Notably, prior studies demonstrated that heterologous protection was absent in mice immunized with VN04-based vaccines, highlighting the risk of immune evasion by contemporary strains.³⁵ Furthermore, while Ast20-based vaccine showed partial cross-reactivity, their neutralizing titers were lower against phylogenetically Texas-closer isolates (e.g., A/pheasant/New York/22-009066-001/2022).²² A deep mutation scan of HA of H5 clade 2.3.4.4b showed that the L115Q mutation promoted immune escape of the virus, which is located within antigenic region.²⁴ A/Texas/37/2024 contained this mutation compared to A/Astrakhan/3212/2020, indicating the possibility of antigenic drift in Texas. These limitations underscore the critical need for antigenically matched vaccines to address the evolving H5N1 threat. In addition, position 131 underwent revertant mutation during evolution (Q > L > Q).³² The HA antigens from other clades tested in this study all carry 131Q, which helps explain the superior antibody induction capability of the Texas vaccine.

In response to the potential pandemic, we developed mRNA-LNP vaccines encoding the Texas HA antigen, utilizing either the full-length WT or the cleavage-defective mutant (Mut) design, delivered by two distinct LNP systems, each containing different ionizable lipids: SM102 and DB-Y (Figure 1A). All mRNA vaccines elicited robust humoral immunity in mice, with neutralizing antibody titers against A/ostrich/China/HB/2024 persisting for over 8 months post-vaccination (Figures 2B–2E and S2A–S2C). Notably, the Texas vaccine demonstrated superior cross-clade protection and higher neutralizing antibody titer against Texas pseudovirus compared to the Ast20-based counterpart (Figures 3C–3G, S4A, and S4C). These findings suggest that the mRNA vaccines incorporating the Texas antigen may offer enhanced efficacy in addressing a potential H5 pandemic.

The protective efficacy of H5 vaccines across various species, including birds, bovines, and other potential intermediate hosts to humans, is a key consideration.⁸ The H5 HPAIV has been widely circulating in domestic poultry and wild birds globally. To assess the effectiveness of our mRNA vaccine, we evaluated its performance in SPF chickens, where it successfully induced a robust antibody response (Figures 6B and 6C). Following a lethal viral challenge with H5 clades 2.3.4.4b and 2.3.4.4h, all chickens in the vaccine group survived, suggesting cross-protection of our vaccine against multiple H5 virus clades in poultry (Figures 6D and 6E).

T cell responses play a crucial role in vaccine effectiveness, and T cell immune mechanisms are essential for understanding antiviral responses and developing effective vaccines.^36,37,38,39 CD8⁺ T cells can protect mice against lethal virus challenge in the absence of neutralizing antibodies.^36,40 The T cell epitopes of influenza A virus are relatively conserved, compared to B cell epitopes.⁴¹ However, HA is not typically regarded as a classic, conserved T cell epitope antigen.^22,42 More than  64% of the CD8⁺ T cell epitopes are conserved in the majority of H5N1 clade 2.3.4.4b viruses assessed but are mainly concentrated in NP, M1, PB2, NS1, and PB1 proteins (>90% sequence identity) rather than HA proteins (an average of 54% sequence identity).⁴² In our evaluation of T cell immunity induction using ELISpot, we observed an accumulation of T cell cytokines, suggesting a role for cellular immunity in protection (Figures 4A–4C and S5A–S5C). However, when using flow cytometry to assess specific T cell populations at designated time points, the fold change was minimal (Figures 4D–4F and S5E–S5J). We reached the same conclusion when examining the T cell immune response induced by HA from other viral subtypes, including H1, H3, and influenza B (data not shown). We also attempted to investigate the T cell response mechanism of HA based on mRNA platform by stimulating spleen cells of mice immunized with mRNA-PR8 HA using known T cell epitope peptides. However, no effective CD8⁺ T cell response was detected by flow cytometry (Figures S7A and S7B). Taken together, our data support the involvement of T cell function in the immune response, demonstrating a balanced Th1/Th2 response characterized by comparable IgG1 and IgG2a titers (Figures 4 and S5). The elevated expression of IgG2a effectively promotes antibody-dependent cell-mediated cytotoxicity⁴³ and opsonophagocytosis by macrophages.²⁷ Further investigations into T cell immune mechanisms are planned, focusing on more suitable antigens such as NP and NA.

The impact of influenza virus immune history on mRNA vaccination is also of concern. In humans, older individuals who were exposed to H1 or H2 subtype viruses in their youth were less susceptible to avian H5N1 viruses.⁴⁴ And, antibody titers to both historical and recent H5N1 strains are highest in older individuals and correlate strongly with birth year, consistent with immune imprinting.⁴⁵ This protective effect may be attributed to the fact that H5 and H1 viruses belong to the same group 1 influenza virus HA, which share highly conserved epitopes, particularly in the stem region of the HA protein. In our research, the Texas vaccine elicited similar antibody levels and provided complete protection against heterologous strain in mice, regardless of prior H1 HA mRNA vaccine immunization (Figure 5). Studies have shown that ferrets with prior H1N1 immunity were protected from mortality and severe clinical disease when intranasally infected with bovine H5N1.⁴⁶ Interestingly, immunization with the avian influenza virus H5N1 can also induce cross-reactive H1N1 HA stem-specific antibodies.⁴⁷ Therefore, prior H1 exposure may provide protection in a potential H5 pandemic without affecting the production of H5-specific antibodies, supporting the feasibility of deploying H5 vaccines in populations with diverse immune histories.

In summary, our research demonstrates the development of a safe and reliable mRNA vaccine platform capable of rapidly adapting to emerging viral variants. The mRNA vaccine, which encodes the HA protein of a current circulating strain (Texas), elicits robust humoral and T cell immune responses across multiple animal models, including mice and chickens, and provides effective protection against heterologous strains. The Texas-based vaccine outperformed the Ast20-based candidate, inducing higher antibody titers against A/Texas/37/2024 strain and offering broader cross-clade protection. Future clinical trials are warranted to validate its immunogenicity and safety in humans.

Limitations of the study

The protective efficacy of our vaccine against the homologous Texas strain could not be directly assessed due to biosafety constraints. However, analogous RNA vaccines targeting bovine H5N1 isolates (A/bovine/OH/B24OSU-342/2024) have demonstrated complete protection in murine models, supporting the generalizability of our findings.³⁵ Given the generally limited efficacy of mRNA vaccines in inducing respiratory mucosal immunity, particularly via intramuscular injection,^48,49,50 we have not evaluated mucosal immunity here, despite its critical role in combating respiratory pathogens. However, existing reports indicate that vaccines based on the self-amplifying (saRNA) platform have induced mucosal immunity,⁵¹ and this platform will also serve as a future research direction for us. In addition, we examined the role of pre-existing immunity in the efficacy of our vaccine. The pre-existing immunity model was developed using the H1-HA mRNA vaccine. However, in humans, the most commonly administered influenza vaccines are inactivated whole-virus vaccines, which contain both HA and neuraminidase (NA) antigens. The NA component may play a significant role in providing protection against H5N1 infection.⁵² As such, inactivated virus vaccines may be a more suitable option for constructing a model of pre-existing immunity. Furthermore, further research should be undertaken to investigate the potential contributions of the NA antigen to vaccine efficacy.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Qiong Zhang (zhang_qiong2@gzlab.ac.cn).

Materials availability

Plasmids generated in this study are available through request. Recombinant proteins and antibodies are available from their respective sources.

Data and code availability

• •All data are available in the manuscript or the supplemental information. Data reported in this paper will be shared by the lead contact Qiong Zhang (zhang_qiong2@gzlab.ac.cn) upon request.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This work was supported by grants from Major Projects of Guangzhou National Laboratory (grant no. GZNL2023A01006, MP-GZNL2025C01003-01, YW-JCYJ0612, and SRPG22-002), the Natural Science Foundation of China (grant no. 82402135 and 82502214), the Pearl River Talent Recruitment Program (grant no. 2023QN10Y266 and 2023QN10Y469), Young Scientists Program of Guangzhou Laboratory (grant no. QNPG23-05 and QNPG24-13), and Modern Agricultural Research System Innovation Team Project of Guangdong Province (grant no. 2024CXTD15).

Author contributions

Q. Z. and H. Li conceptualized the study. Q.Z., H. Li, M.H., S.F., J. Lin., W.J., K.K.-W. T., and Y.Q. designed the experiments. Y.L. synthesized and characterized DB-Y. H.G. and T.J. prepared mRNA vaccines. H. Li and S.F. performed mouse experiments and ELISA experiments. M.Z. performed pseudovirus neutralization assay. M.H. and D.W. performed MN experiments. H. Li, S.F., W.L., and Q.P. performed T cell experiments. D.W., X.Z., and Y.Y. performed chicken experiments and virus challenge. X.L. performed toxicological analysis. Q.Z., H. Li, M.H., S.F., W.L., D.W., and X.L. drafted the manuscript. L.C., Y.Q., J. Lin, W.S., J.Y., H. Li, K.K.-W.T., W.J., J. Lu, and Q.Z. reviewed and edited the manuscript. All authors contributed to the preparation of the manuscript. All authors gave final approval before submission.

Declaration of interests

J. Lu, Y. Liu, T. Jiang, and H. Gu are employees of Shanghai RNACure Biopharma Company, which collaborated in the development of the vaccine. J. Lin, J. Lu, Y. Liu, and T. Jiang are inventors of patent CN119264001A, which is owned by Shanghai RNACure Biopharma Company. Q. Zhang, Y. Quan, H. Li, and S. Feng. are inventors on patent application CN202510125661.5, which covers the mRNA sequence design described in this manuscript. The remaining authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
CR6261	Synthesized from Biointron	N/A
4R5	Synthesized from Biointron	N/A
Bacterial and virus strains
DH5a Competent E.coli	Ktsmo	KTSMCC600
A/ostrich/China/HB/2024 (H5N1)	Obtained from the Lab of Professor Weixin Jia	N/A
A/goose/GuangDong/1189/2023 (H5N1)	Obtained from the Lab of Professor Weixin Jia	N/A
A/duck/Fujian/QG4/2023 (H5N6)	Obtained from the Lab of Professor Weixin Jia	N/A
VSV pseudovirus expressed HA and NA protein of A/Texas/37/2024 (H5N1)	Obtained from the Lab of Professor Youchun Wang	N/A
Chemicals, peptides, and recombinant proteins
HA recombinant protein of A/Viet Nam/1203/2004	Synthesized from Novoprotein	N/A
HA recombinant protein of A/Indonesia/5/2005	Synthesized from Novoprotein	N/A
HA recombinant protein of A/bar-headed goose/Qinghai/5/2005	Synthesized from Novoprotein	N/A
HA recombinant protein of A/Hubei/1/2010	Synthesized from Novoprotein	N/A
HA recombinant protein of A/Anhui/1/2005	Synthesized from Novoprotein	N/A
HA recombinant protein of A/Astrakhan/3212/2020	Synthesized from Novoprotein	N/A
HA recombinant protein of A/Texas/37/2024	Synthesized from Novoprotein	N/A
HA peptide pool of A/Texas/37/2024	Synthesized from GenScript	N/A
Critical commercial assays
ELISpot Plus: Mouse IFN-γ (HRP)	MabTech	3321-4HST-2
ELISpot Plus: Mouse IL-2 (HRP)	MabTech	3441-4HPW-2
ELISpot Plus: Mouse IL-4 (HRP)	MabTech	3311-4HPW-2
ELISpot Plus: Mouse TNF-α (HRP)	MabTech	3511-4HPW-2
LEGENDplex™ Rat Th1/Th2 Panel V02 (9-plex) w/FP kit	BioLegend	741227
Experimental models: Cell lines
BHK-21 cells	Laboratory preservation	ATCC-CCL-10
HeLa cells	Laboratory preservation	ATCC-CCL-2
HEK293T-17 cells	Laboratory preservation	ATCC-CRL-11268
MDCK cells	Laboratory preservation	ATCC-CCL-34
Recombinant DNA
pUC57-kan:hHBBv2 mRNA::H5N1(A/Texas/37/2024) HA:delA	This study	Addgene ID: 247357
pUC57-kan:hHBBv2 mRNA::H5N1(A/Texas/37/2024) Mut HA (343-345del):delA	This study	Addgene ID: 247358
pUC57-kan:hHBBv2 mRNA::H5N1(A/Astrakhan/3212/2020) HA:delA	This study	Addgene ID: 247360
pUC57-kan:hHBBv2 mRNA::H5N1(A/Astrakhan/3212/2020) Mut HA (343-345del):delA	This study	Addgene ID: 247359
Software and algorithms
ProSize data analysis software	Agilent	https://www.agilent.com/zh-cn/product/automated-electrophoresis/fragment-analyzer-systems/fragment-analyzer-systems-software/fragment-analyzer-software-1149185
GraphPad Prism 10.2.3	Graphpad Software	www.graphpad.com/scientific software/prism/
NovoExpress	Agilent	https://www.agilent.com/zh-cn/product/research-flow-cytometry/flow-cytometry-software

Experimental model and study participant details

Cell culture

BHK-21, HeLa, HEK293T-17, and MDCK cells were grown and maintained in DMEM supplemented with 10% fetal bovine serum (FBS), 1% PS (100 units/mL penicillin, 100 mg/mL streptomycin) at 37°C and 5% CO₂. All cell lines were identified by short tandem repeat sequences (STR) typing and compared with the ATCC database, and none were contaminated with Mycoplasma.

Virus culture

The A/ostrich/China/HB/2024 (H5N1), A/goose/GuangDong/1189/2023 (H5N1) and A/duck/Fujian/QG4/2023 (H5N6) strains were propagated in 9- to 10-day-old SPF embryonated chicken eggs (Wens). Eggs were candled to confirm embryo viability before inoculation. Virus stock (1,000 FFU/egg) was injected into the allantoic cavity using a sterile needle. Inoculated eggs were incubated at 37°C for 24–48 h, then chilled at 4°C for ≥4 h to facilitate harvesting. The allantoic fluid was collected, clarified by centrifugation (3,000 ×g, 10 min, 4°C), aliquoted, and stored at −80°C. Research involving HPAI viruses was carried out exclusively within BSL-3 containment facilities. All laboratory personnel involved in viral research underwent comprehensive biosafety training before study initiation.

Animals and biosafety

Female BALB/c mice (5-7 week-old) were obtained from GemPharmatech (Guangzhou, China) and randomly assigned to experimental groups. Male and female SD rats (6-8 week-old) were obtained from Charles river (Shanghai, China) and randomly assigned to experimental groups. The mice and rats were acclimatized for 5 days in an SPF animal facility with strict environmental controls. Temperature was maintained at 20 ± 2°C with 50 ± 10% relative humidity, and a controlled 12-h light/dark cycle (7:00 a.m. to 7:00 p.m. light phase) was followed. Mice had ad libitum access to food and water throughout the study period.

10-day-old SPF chickens were sourced from Wen’s Dahuanong Biotechnology Co., Ltd. (Guangzhou, China) and randomly allocated to experimental groups. These chickens were quarantined for 5 days in SPF isolators equipped with heat lamps. The temperature was maintained at 26 ± 2°C, and humidity at 60 ± 10%. The isolators were under negative pressure (100 ± 10 Pa), with an 8-h light/16-h dark cycle (9:00 a.m. to 5:00 p.m. light phase). Chickens were provided with ad libitum access to food and water during the entire experimental period.

All studies were conducted in full compliance with the ethical regulations set forth by the Guangzhou National Laboratory and South China Agricultural University, Guangzhou, China. The ethical approval number for the animals is GZLAB-AUCP-2025-09-A09. Animal experiments were carried out under protocols approved by the Animal Care and Use Committee of Guangzhou National Laboratory. All experimental procedures involving mice and chickens were conducted under appropriate anesthesia to ensure animal welfare. The experimental design adhered to the principle of using the minimum number of animals necessary to obtain statistically valid results, and measures were implemented throughout the study to minimize discomfort and pain for all animals.

Method details

mRNA vaccine design and synthesis

The cDNA sequence of the HA gene from A/Texas/37/2024 (H5N1) (GenBank: PP577943) was codon-optimized for efficient mammalian expression. Two constructs were generated: the wild-type HA (WT) and a cleavage-defective mutant (Mut), where the polybasic cleavage site (RKR) was deleted. Both constructs were synthesized as DNA fragments incorporating 5’ and 3’ untranslated regions (UTRs) derived from human β-globin, along with a 100-nucleotide poly(A) tail added during PCR amplification. The DNA templates were linearized by PCR and purified using a gel extraction kit (Promega) prior to in vitro transcription. mRNA synthesis was performed using T7 RNA polymerase (APExBIO), with the modified uridine analog pseudouridine (Ψ) (APExBIO) replacing uridine triphosphate to reduce immunogenicity. The Cap1 analog (APExBIO) was incorporated into the IVT reaction to add a Cap1 structure to the 5’ end of the mRNAs. Following transcription, template DNA was removed by DNase I treatment (APExBIO), and the mRNAs were purified using oligo-dT chromatography. The final mRNA products were resuspended in nuclease-free water (Beyotime) and quantified spectrophotometrically using a NanoDrop One (Thermo Fisher). RNA integrity was assessed by capillary electrophoresis (5200 Fragment Analyzer, Agilent Technologies).

In vitro expression analysis

Transfection experiments were carried out in three mammalian cell lines: BHK-21, HeLa, and HEK293T. Cells were seeded in 12-well plates and transfected at 50–60% confluency with 500 ng of mRNA using Lipo8000 transfection reagent (Beyotime) in Opti-MEM reduced serum medium (Gibco). For Western blotting analysis, cells were harvested 24 h post-transfection and lysed in RIPA buffer containing protease inhibitors. Protein samples were resolved on 4–20% gradient SDS-PAGE gels (ACE Biotechnology) and transferred to PVDF membranes (Millipore). Membranes were blocked with 5% non-fat milk and probed with either the anti-HA antibody, which is a monoclonal antibody targeting the HA stem region (Sino Biological; 1:2000 dilution) or HRP-conjugated anti-GAPDH antibody (Proteintech; 1:5000 dilution) for 1 h at room temperature. After incubation with HRP-conjugated goat anti-human secondary antibody (Sangon Biotech; 1:5000 dilution) for 1 h at room temperature, protein bands were visualized using enhanced chemiluminescence substrate (CoWin Biotech) and imaged using the iBright CL750 imaging system (Thermo Fisher). For flow cytometry analysis, BHK-21 cells were transfected with varying doses (50, 100, 250, or 500 ng) of WT or Mut mRNA. 24 h post-transfection, cells were detached with 0.25% trypsin-EDTA, washed with PBS, and stained with 1 μg/mL CR6261 primary antibody (diluted in PBS), followed by Alexa Fluor 488-conjugated secondary antibody (Invitrogen; 1:500 dilution in PBS). Mock-transfected cells were used as negative controls for gating. Samples were analyzed using a Novocyte Advanteon flow cytometer (Agilent Technologies), with at least 200,000 events collected per sample. Data analysis was performed using NovoExpress software (Agilent Technologies), and results were expressed as the percentage of HA-positive cells.

LNP formulation and characterization

SM102 (Shenzhou Pharmaceutical) or DB-Y (Shanghai RNACure Biopharma), DSPC (Nippon Fine Chemical), cholesterol (Nippon Fine Chemical), and PEG2000-DMG (JenKem Technology) were dissolved in 100% ethanol at molar ratios of 50:10:38.5:1.5. The lipid mixture was then combined with the mRNA solution in 25 mM sodium acetate buffer (pH 5.0) using the Nano Precision (Shanghai RNACure Biopharma). The formulations were immediately diluted 2-fold with the same buffer and dialyzed against Tris-acetate buffer (pH 7.5) overnight at 4°C using Slide-A-Lyzer dialysis cassettes. After dialysis, the formulations were concentrated using Amicon Ultra Centrifugal Filters (Millipore) and passed through a 0.22 μm filter (Merck). Particle size distribution and zeta potential were measured with a Zetasizer Nano ZA (Malvern Instruments Ltd, UK). RNA encapsulation efficiency (EE) was determined using the Quant-iT RiboGreen Assay Kit (Thermo Fisher Scientific, USA) according to the manufacturer’s protocol. For cryo-electron microscopy, LNP samples were applied to freshly glow-discharged Quantifoil grids, blotted for 4 s (blot force: 0, wait time: 30 s), and plunge-frozen in liquid ethane. The samples were then stored in liquid nitrogen until imaging. Data collection was performed on a Glacios transmission electron microscope (Thermo Fisher) operating at 200 kV, equipped with a Ceta-D camera. Images were captured using EPU software (Thermo Fisher Scientific).

2-p-Toluidinylnaphthalene-6-sulfonate (TNS) binding assay

The apparent pKa of mRNA-LNP was analyzed by TNS binding assay. The mRNA-LNP samples were tested at a concentration of 50 μg/mL (by mRNA) in 300 μM TNS dissolved in the base buffer (10 mM sodium phosphate, 10 mM sodium borate, 10 mM sodium citrate, 150 mM sodium chloride). Fluorescence intensity was detected using a multifunctional microplate reader with fluorescence excitation at 320 nm and emission at 445 nm.

Quantitation of generalized polarization (GP) of laurdan

The fluorescent dye Laurdan was dissolved in DMSO and incubated with LNP at room temperature, where the molar ratio of lipid and Laurdan was 600:1, the volume ratio of DMSO to PBS was 1:500. It was then detected using a multifunctional microplate reader with fluorescence excitation at 340 nm and emission at 490 nm.

In vivo imaging of luciferase mRNA expression

To evaluate in vivo gene expression, mRNA encoding luciferase was encapsulated in LNPs (SM102/DB-Y) and injected into the right thigh muscle of BALB/c mice (5 μg/mouse). For bioluminescence imaging, each mouse was administered 100 μL of XenoLight D-Luciferin Potassium Salt (PerkinElmer #122799) at a concentration of 30 mg/mL as the luciferase substrate. Mice were anesthetized with isoflurane gas anesthesia during the imaging procedure. Bioluminescence was then measured using the IVIS Spectrum Imaging System (Guangzhou Biolight Biotechnology, China) at 6 h, 24 h, 48 h, 7 days, and 14 days post injection. Imaging was performed 20 min after substrate injection to allow for optimal substrate-luciferase reaction. Luminescence signals (expressed in photons/sec, p/s) were quantified using the system’s built-in software (Living Image or equivalent). Regions of interest (ROIs) were drawn around the injection site to measure total flux.

Vaccination in mice

The vaccination procedures were conducted in strict accordance with the approved animal protocol (GZLAB-AUCP-2025-09-A9) from the Animal Care and Use Committee of Guangzhou Laboratory, China. 7-week-old female BALB/c mice were administered intramuscular immunizations with either 1 μg or 5 μg of mRNA vaccine or a PBS placebo control, following a prime-boost schedule with an interval of 4 weeks.

Blood samples were collected through retro-orbital bleeding at 2 weeks post prime and post boost for immunological analysis. An additional group of mice was humanely euthanized at 1 week post boost to collect spleens for the evaluation of antigen-specific T cell responses. All procedures were performed under appropriate anesthesia to minimize animal discomfort, in strict compliance with institutional animal welfare guidelines.

Viral challenge in mice

To determine the optimal challenge dose, a dose-escalation study was performed using the H5N1 clade 2.3.4.4b strain A/ostrich/China/HB/2024. 7-week-old female BALB/c mice (n = 5 per group) were intranasally inoculated with 50 μL of viral suspension containing serial 10-fold dilutions (ranging from 10⁻¹ to 10⁶ EID₅₀) under isoflurane anesthesia. Mice were monitored daily for 14 days post challenge, with body weight recorded as a primary clinical endpoint. Mice showing ≥25% weight loss from baseline were euthanized. The median lethal dose (LD₅₀) was calculated to be 10^1.319 EID₅₀ using GraphPad Prism, which was used as the benchmark for future challenge studies.

4 or 22 weeks post boost, vaccinated mice were anesthetized and challenged with 5×LD₅₀ or 50×LD₅₀ (50 μL, i.n.) of the A/ostrich/China/HB/2024 strain. Clinical monitoring continued for 14 days, including daily weight measurements and observation for signs of disease. The humane endpoint of ≥25% weight loss was strictly enforced, with all mice reaching this endpoint being immediately euthanized. Surviving animals were humanely sacrificed at study termination for comprehensive analysis. All procedures were carried out in ABSL-3 containment with appropriate personal protective equipment.

Enzyme linked immunosorbent assay (ELISA)

Binding antibody titers against recombinant H5 hemagglutinin (rHA) were measured in mouse and chicken serum samples using a standardized ELISA protocol. 96-well plates (Costar 3599, Corning) were coated with 1 μg/mL rHA in carbonate/bicarbonate coating buffer (pH 9.6) and incubated overnight at 4°C. After three washes with PBS containing 0.05% Tween 20 (PBS-T), plates were blocked with 5% non-fat milk in PBS for 1 h at 37°C. Serum samples were initially diluted 1:200 in PBS, followed by 3-fold serial dilutions. After incubation with diluted sera for 1 h at 37°C, plates were washed and incubated with horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG secondary antibody (Sangon Biotech, #D110087) at a 1:2000 dilution for 1 h at 37°C. Following four additional washes, color development was initiated with TMB substrate (Beyotime, #P0209) for 10 min, and the reaction was stopped with 2 M H₂SO₄ (Beyotime, #P0215). Absorbance was measured at 450 nm using an Ensight Multimode Plate Reader (PerkinElmer). Endpoint titers were calculated as the reciprocal of the highest serum dilution yielding an optical density ≥2-fold above background levels. For IgG subclass analysis, the same protocol was followed, with the use of subtype-specific secondary antibodies: HRP-conjugated goat anti-mouse IgG1 (Abcam, ab97240; 1:5,000) and IgG2a (Abcam, ab97245; 1:3,000).

Influenza virus neutralization assay

The neutralization capacity of immune sera was evaluated using a focus reduction neutralization test (FRNT). Prior to the assay, MDCK cells were seeded in 96-well flat-bottom plates (Costar 3599, Corning) at a density of 3×10⁵ cells per well and cultured for 24 h at 37°C with 5% CO₂. Serum samples from immunized mice and chickens were pretreated with receptor-destroying enzyme (RDE; Denka-Seiken 340122) at a 1:3 serum-to-RDE ratio, followed by incubation at 37°C for 18 h. Afterward, the samples were heat-inactivated at 56°C for 30 min and serially diluted 3-fold in virus diluent (DMEM supplemented with 25 mM HEPES and 2 μg/mL TPCK-treated trypsin), starting at a dilution of 1:24. Each diluted serum sample was mixed with 100 focus-forming units (FFU) of the virus in MEM medium and incubated at 37°C for 1 h to allow the antibody-virus interaction. After incubation, the serum-virus mixtures were removed from the cell monolayers, which were then washed twice with PBS. The cells were overlaid with 100 μL of carboxymethyl cellulose (CMC) medium containing 1% CMC, 1 μg/mL TPCK-treated trypsin, and 25 mM HEPES in MEM. The plates were incubated at 37°C for 24 h. Following incubation, the cells were fixed with 4% paraformaldehyde for 1 h at room temperature. Permeabilization and blocking were performed using 0.2% Triton X-100 in 1% BSA for 30 min at room temperature. Viral nucleoprotein was detected by incubating with a mouse anti-NP monoclonal antibody (4R5) at 37°C for 1 h, followed by a secondary HRP-conjugated goat anti-mouse IgG antibody (Sangon Biotech, #D110087) at a 1:2,000 dilution. Viral foci were visualized using TrueBlue TMB substrate (KPL 5510-0030), with a 10-min development time, and quantified using an ELISpot reader (AT-Stop1100/CTL). The neutralization titer (FRNT₅₀) was determined as the reciprocal of the highest serum dilution that resulted in a ≥50% reduction in viral foci compared to the virus-only control wells.

Pseudovirus neutralization assay

The pseudovirus for vesicular stomatitis virus (VSV) was obtained from a collaborating research group. This pseudovirus has a deleted G protein gene and incorporates firefly luciferase as a reporter gene. For the pseudovirus neutralization assay, MDCK cells were seeded at a density of 2×10⁵ cells/mL, with 100 μL added to each well in 96-well plates, and cultured overnight at 37°C. Serum samples were diluted starting from 1:900 in 3-fold serial dilutions. Each well was mixed with 1.3×10⁴ TCID₅₀ of the virus and incubated at 37°C for 1 h. Following the incubation, the cells were digested, and a 100 μL aliquot of a 2×10⁵ cells/mL cell suspension was added to the 96-well plate. After 48 h, results were read using a multimode reader (PerkinElmer EnSight). EC₅₀ values were calculated by Reed-Muench method.

Virus load

Assays were performed in the BSL-3 biocontainment facility. Prior to the assay, MDCK cells were seeded in 96-well flat-bottom plates (Costar 3599, Corning) at a density of 2.5×10⁵ cells per well and cultured for 18 h at 37°C with 5% CO₂. After 18 h, cells were washed once with 1×PBS, and after homogenization of heart, liver, spleen, lung, kidney, and brain tissues from infected mice, the supernatant was collected by centrifugation. The supernatant was serially diluted 10-fold in DMEM and then added to plates for 1 h at 37°C with 5% CO₂. After 1 h, virus dilutions were aspirated and replaced with CMC overlay. The subsequent steps are consistent with the virus neutralization assay.

Enzyme-linked immunospot (ELISpot) assay for T cell analysis

The frequencies of antigen-specific cytokine-secreting cells (TNF-α, IFN-γ, IL-2) in splenocytes from H5-immunized mice were quantified at 7 days post-boost using T-cell ELISpot assays. Sterile 96-well ELISpot plates (MabTech, Sweden) were pre-coated with capture antibodies specific for mouse TNF-α (MabTech #3511-4HPW-2), IFN-γ (MabTech #3321-4HST-2) and IL-2 (MabTech #3441-4HPW-2). After washing the plates once with PBS, they were blocked with RPMI-1640 medium containing 10% fetal bovine serum (FBS) for 30 min at room temperature. Following removal of the blocking solution, 2.5 × 10⁵ splenocytes were added to each well and stimulated with 2 μg/mL of an H5 peptide pool, which consisted of 136 overlapping 15-mer peptides (with adjacent peptides overlapping by 11 amino acids) spanning the HA sequence of Texas (≥90% purity, GenScript). The plates were incubated at 37°C for 36 h. After incubation, the cells were removed, and the plates were washed five times with 200 μL of PBS per well. Biotin-conjugated detection antibodies specific for TNF-α, IFN-γ and IL-2 (1 μg/mL in PBS with 0.5% FBS) were added to each well (100 μL/well) and incubated at room temperature for 2 h. After further washing, streptavidin-HRP (100 μL/well) was applied for 1 h at room temperature. The spots were developed using TMB substrate and counted using a plate reader (PerkinElmer).

Flow cytometry for T cell analysis

Antigen-responsive T cells from immunized mice were analyzed by in vitro antigenic restimulation. Splenocytes (1×10⁶ cells) were isolated 7 days post boost immunization. The cells were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS) and 5 μg/mL of the H5 Texas peptide pool. The culture was also supplemented with GolgiPlug (BD Biosciences #555029) and GolgiStop (BD Biosciences #554724) protein transport inhibitors at a 1:1,000 dilution. Concanvalin A (10 μg/mL; InvivoGen #inh-cona) was used as a positive control. The cells were incubated for 6 h at 37°C. After incubation, cells were washed with PBS and surface-stained for 30 min at 4°C in the dark using the following antibody cocktail: anti-CD3e FITC (BD Biosciences #553061), anti-CD4 PE-Cy7 (BD Biosciences #552775), anti-CD8a BV510 (BD Biosciences #563068), and Fixable Viability Stain 780 (BD Biosciences #565388), all diluted in MACS buffer (PBS with 2% BSA and 1 mM EDTA). Following surface staining, cells were fixed and permeabilized using Cytofix/Cytoperm solution (BD Biosciences #554722) for 20 min at 4°C. Intracellular cytokines were then stained with the following antibodies: anti-IFN-γ BV711 (BD Biosciences #564336), anti-TNF-α BB700 (BD Biosciences #566510), anti-IL-2 PE (BD Biosciences #554428), all diluted 1:100 in Perm/Wash buffer. The stained cells were resuspended in MACS buffer and analyzed on a Novocyte Advanteon flow cytometer (Agilent Technologies) using NovoExpress software. Gating strategies are shown in Figure S5D.

Vaccination in chickens

10-day-old SPF chickens were administered hypodermic immunizations with 30 μg of the mRNA vaccine or PBS placebo control following a prime-boost schedule with an interval of 1 week. Venous blood samples were collected for immunological analysis at 1 week post prime and 2 weeks post boost.

Viral challenge in chickens

2 weeks post the boost, chickens were intranasally challenged with 100 μL of 10⁶ EID₅₀/0.2 mL of either H5N1 clade 2.3.4.4b virus (A/goose/GuangDong/1189/2023) or H5N6 clade 2.3.4.4h virus (A/duck/Fujian/QG4/2023). Chickens were monitored for clinical signs and mortality for 9 days post challenge. Oropharyngeal and cloacal swab mixed samples were collected at 5 days post challenge. Virus positivity or shedding was assessed by inoculating the swab solutions into three 10-day-old SPF chicken embryos. All surviving chickens were humanely sacrificed at the end of the monitoring period.

Toxicological evaluation in SD rats

In a single-dose toxicity study evaluating different dose levels, male and female Sprague-Dawley (SD) rats (n = 4 per sex per group) were randomly allocated into five treatment groups. On Day 0, animals received intramuscular injections of either sterile saline (placebo) or lipid-encapsulated mRNA formulations at doses of 50 μg or 300 μg. Body weights were recorded post-dosing. Concurrently, the local irritation of injection sites were evaluated at the same time point as the body weight measurement. The local irritation was divided into 3 grades according to the swelling degree. Slight swelling at the injection site muscle was scored as 1 point, obvious swelling was scored as 2 points, and severe swelling with thickening of the injection muscle was scored as 3 points.

The blood samples (anticoagulated whole blood and serum) were collected from rats jugular vein at the time points of pre-dose, 6 h post-dose, and 14 days post-dose. The complete blood counts (CBC) were analyzed with EDTA-K2 anticoagulated whole blood on a Mindray BC-2800 Vet automatic animal blood cell analyzer. The cytokine profiling of H5N1 vaccine acute toxicity test was conducted using multiplex immunoassay technology with LEGENDplex Rat Th1/Th2 Panel V02 (9-plex) w/FP kit (Cat# 741227, BioLegend) at 6 h and 14 days post injection following the manufacturer’s instructions with the serum samples. The kit panel could quantify IL-10, IFN-γ, IL-5, IL-2, TNF-α, GM-CSF, IL-4, IL-13, and IL-6. The processed samples were analyzed on a CYTEK Northern Lights full spectrum cell analyzer (NL-CLC V16B14R8).

After the recovery period of these toxicity tests, animals were humanely euthanized, followed by complete necropsy. Major organs including the heart, liver, spleen, lungs, kidneys were excised and preserved in 4% paraformaldehyde fix solution. The fixed tissues were used for routine histopathological treatment (paraffin embedding, sectioning, extracting slices, and HE staining) and microscopic observation. Grading of lesions was performed using a 5-point scale (mildest, mild, moderate, marked and severe).

Quantification and statistical analysis

Data were analyzed using GraphPad Prism version 10.2.3. Data are presented as the mean with standard deviations as indicated. Means are calculated on the value of n biological replicates, with n dependent upon experiment goals. The difference between two groups was calculated by t test. All tests were performed two-tailed. Statistically significant results were determined when p < 0.05. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Published: April 8, 2026