Influenza mRNA vaccine lowers severity and spread of A(H5N1) virus in ferrets

A nucleoside-modified mRNA construct encoding the full-length, codon-optimized HA protein with the polybasic cleavage site deleted from A/Astrakhan/3212/2020 A(H5N8) (Ast/20), a WHO-recommended prototype candidate vaccine virus (CVV), was synthesized and formulated in lipid nanoparticles (LNPs) (Supplementary Fig. 1A). Production and quality control followed the same platform methods used for Pfizer's COVID-19 and seasonal influenza mRNA vaccines^24–26. In vitro analysis confirmed that HA protein expression was efficient and dose-dependent (Supplementary Fig. 1B).

To evaluate the efficacy of the H5 mRNA vaccine candidate, ferrets were challenged intranasally with 10⁴ TCID₅₀ of HPAI A(H5N1) A/Chile/25945/2023 (Chile/23) virus, five weeks after the second vaccination (8 weeks after the first dose). All ferrets vaccinated with H5 mRNA (Group 2) or adjuvanted H5 protein (Group 3) survived (4/4), while only one of four unvaccinated ferrets (Group 1) survived (Fig. 1C). The surviving unvaccinated ferret lost nearly 20% of its body weight but remained above the humane endpoint threshold (Fig. 1D). Despite receiving supportive care, including carnivore plus diet feeding on day 9 post-challenge, the animal failed to regain weight (supportive care was not provided to any other vaccinated or unvaccinated ferrets). In contrast, H5 mRNA-vaccinated ferrets exhibited minimal weight loss (mean maximum: 1.2%), while protein-vaccinated ferrets lost slightly more (mean maximum: 5.3%). The difference in weight loss between the mRNA and protein vaccine groups was statistically significant (P < 0.05) (Fig. 1D). These results indicate that the H5 mRNA vaccine candidate effectively protected ferrets against HPAI A(H5N1) disease, reducing severity and preventing lethal outcomes.

To evaluate viral dissemination, tissues were collected on day 5 from select animals and infectious virus was quantified (Fig. 2B). In unvaccinated ferrets, robust virus replication was detected in the trachea and lungs of all four ferrets; in the spleen and kidney of one ferret; in the liver of one ferret; and in the small and large intestines of three ferrets. In contrast, virus was detected only in the small intestine of one mRNA-vaccinated ferret and in the large intestine of one protein-vaccinated ferret. No virus was detected in the trachea, lung, spleen, liver, or kidney of vaccinated ferrets. Viral titers in the olfactory bulb and brain were below detection in all groups (Fig. 2B). Digital PCR confirmed significantly lower RNA levels in the trachea and lungs of vaccinated ferrets (P < 0.001), consistent with infectious titers (Supplementary Fig. 2). These results demonstrate that the H5 mRNA vaccine markedly reduced viral replication in both the upper and lower respiratory tracts and intestines.

Histopathologic analysis of respiratory tissues revealed marked protection from lung pathology in H5 mRNA-vaccinated ferrets compared to unvaccinated controls (Supplementary Fig.). Among unvaccinated animals, two had lung lobes with up to 50% of the parenchyma affected by pneumonic changes. Affected areas had peribronchial and peribronchiolar lymphoplasmacytic inflammation, and variable epithelial necrosis and luminal filling by degenerate inflammatory cells, neutrophils, and/or mucus. Alveoli were filled and sometimes effaced by neutrophils, macrophages, and few erythrocytes. Non-pneumonic lung regions showed mild congestion and mild interstitial and perivascular mononuclear infiltrates. The other two unvaccinated animals showed milder but similar changes affecting 10–20% of the examined lung lobes. In all unvaccinated ferrets, A(H5N1) viral antigen was detected by immunohistochemistry (IHC) in bronchial and bronchiolar epithelium, including bronchial submucosal glands, and in alveolar epithelium and inflammatory cells within the pneumonic lung tissue. 3

Among the four H5 mRNA-vaccinated animals, two had patchy foci of peribronchial, and more prominent peribronchiolar, and dense perivascular inflammation. Inflammation was comprised mostly of lymphocytes, plasma cells, and histiocytes, but neutrophils were occasionally present within and around bronchiolar lumens. Alveolar walls contained low numbers of mononuclear cells, but alveolar lumina were free of inflammation. Lung from a third H5 mRNA-vaccinated animal had only bronchial submucosal inflammation, and the lung from the fourth animal was unremarkable. No viral antigen was detected by IHC in conducting airways or lung parenchyma of H5 mRNA-vaccinated animals. Ferrets vaccinated with the adjuvanted H5 protein displayed similar or slightly greater inflammation than mRNA-vaccinated animals, without viral antigen detection.

Consistent with Fig. 1B, H5 mRNA-vaccinated ferrets developed a robust antibody response, with HI titers significantly higher than protein-vaccinated animals (Fig. 3B). Five weeks after the second vaccination, ferrets assigned as inoculated animals were challenged intranasally with 10⁶ PFU of A/Michigan/90/2024 (MI/24), a clade 2.3.4.4b HPAI A(H5N1) virus shown to transmit efficiently (100%) in a direct contact setting among cohoused ferrets^29,30. Twenty-four hours post-inoculation, a contact ferret was co-housed with each inoculated ferret (three pairs per group).

It is possible that NW viral titration may miss transmission events between inoculated and contact pairs, particularly if the contact ferret rapidly cleared the infection due to limited exposure. To confirm transmission, neuraminidase (NA)-specific antibodies were measured by ELISA at week 11 (Fig. 3A). Since neither vaccine included NA, the presence of NA antibodies served as a marker of infection. In Group 2 (mRNA-vaccinated inoculated/naïve contact), although infectious virus was not detected in NW samples from all three inoculated ferrets, NA antibodies were detected in all, confirming infection. In contrast to Group 1 (naïve inoculated/naïve contact), where all contact ferrets seroconverted, NA antibodies were detected in only 2 of 3 contact ferrets in Group 2, confirming a reduced rate of transmission (Fig. 4).

In Group 3 (protein-vaccinated inoculated/naïve contacts), virus shedding from the inoculated ferrets was lower than in Group 1 at 3 dpi (P < 0.01), but continued through day 5, and all three contact ferrets became infected (Fig. 4). Vaccination with the adjuvanted H5 protein vaccine did not prevent transmission to any of the contact ferrets in Group 3. In Group 4 (naïve inoculated/mRNA-vaccinated contacts), transmission was completely blocked. No virus was detected in NW samples, and no NA antibodies were detected in any contact ferrets (Fig. 4).

Collectively, these results show that H5 mRNA vaccination significantly reduced viral shedding and limited transmission from vaccinated, infected ferrets. Importantly, mRNA-vaccinated contact ferrets were fully protected from infection despite direct exposure to shedding, unvaccinated animals.

Sera from H5 mRNA-vaccinated ferrets efficiently neutralized nearly all clade 2.3.4.4b HPAI A(H5) viruses, with titers comparable to those observed against the reference candidate vaccine virus, Ast/20. Neutralization of an A/Missouri/121/2024-like virus was reduced 9-fold, likely due to the HA A156T substitution identified in viral RNA from a clinical specimen. This virus was generated by reverse genetics, as isolation of live virus was unsuccessful³². Sera from H5 mRNA-vaccinated ferrets showed variable neutralization against clade 2.3.2.1e HPAI A(H5N1) viruses, with no detectable activity against one strain. This likely reflects over 40 amino acid differences in HA between Ast/20 and clade 2.3.2.1e viruses, as well as strain-specific substitutions (Supplementary Tables 1–3). Sera from protein-vaccinated ferrets showed a similar neutralization pattern, although overall titers were lower than those induced by H5 mRNA vaccination (Fig. 5).

This preclinical study evaluated the protective effects of an influenza mRNA vaccine against clade 2.3.4.4b highly pathogenic avian influenza (HPAI) A(H5N1) viruses. Ferrets were chosen as the animal model due to their relevance for studying influenza virus pathogenicity, transmission, and vaccine efficacy. Two in vivo experiments were conducted. In both, ferrets were immunized with either an mRNA vaccine candidate, a recombinant protein control, or phosphate-buffered saline (PBS). The first experiment assessed vaccine-mediated protection against severe disease using a lethal dose of a clade 2.3.4.4b virus isolate from Chile. Clinical signs (body weight), nasal viral shedding, and viral loads in tissues were monitored. Each group consisted of eight animals: four were monitored for clinical outcomes and nasal shedding over 14 days, and four were euthanized at 5 days post-infection for tissue viral load quantification and histopathological analysis. The second experiment evaluated the effect of vaccination on viral transmission. Ferrets were infected with a clade 2.3.4.4b human isolate from the United States (associated with the second confirmed U.S. human case in 2024). Virus shedding and direct-contact transmission were assessed. Each transmission condition included three donor-contact pairs (six animals per group), as illustrated in Fig. 3. Different virus strains were selected for each experiment to suit their respective goals: a lethal isolate for evaluating disease severity and a more transmissible isolate for assessing transmission potential.

Group sizes were selected to balance statistical power with logistical feasibility in a high-containment laboratory setting. Animals were randomly assigned to experimental groups. Blinding was not applied to the animal studies. All cell culture and serological assays followed CDC standard operating procedures, with appropriate technical replicates and controls to ensure statistical robustness.

All research involving HPAI A(H5N1) viruses was conducted within Biosafety Level 3 enhanced (BSL-3E) or Animal Biosafety Level 3 (ABSL-3) facilities at the CDC. All personnel engaged in virus-related activities underwent comprehensive training in relevant safety protocols and specific procedural techniques. Their competence to perform tasks within BSL-3E/ABSL-3 laboratories was rigorously assessed. All ferret experiments were conducted with the approval of the CDC's Institutional Animal Care and Use Committee (IACUC) and were conducted in an Association for Assessment and Accreditation of Laboratory Animal Care International-accredited ABSL-3 facility with enhancements as required by the U.S. Department of Agriculture⁵⁷. The principle of using the minimum number of animals required to achieve scientific validity was strictly adhered to⁵⁸. Procedures designed to minimize animal discomfort and pain were implemented. A clinical scoring system was used to assess the euthanasia endpoints as follows: 2 points each, hunched back/huddling or ruffled coat/piloerection/hair loss; 3 points each, dehydration, or fluctuation of body temperature; 5 points each, abnormal breathing (dyspnea, tachypnea, rales/audible breaths) anemia; 10 points each, paralysis/torticollis/moribund behavior/unresponsive, weight loss reaching 20% of baseline, reduced body temperature for greater than 3 days, frank hemorrhage/bleeding, unrelated trauma where continued exposure to cage mates is likely to lead to death, or other immediate veterinary concerns. If the clinical scoring scale reached 10 total points or above, the animals were humanely euthanized.

Human embryonic kidney 293T (HEK-293T) cells (CRL-3216, ATCC) were maintained in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum. Madin-Darby canine kidney (MDCK) cells (ATCC, CCL-34) were cultured in Minimum Essential Medium (MEM) with 5% fetal bovine serum. The cells were cultured at 37 °C with 5% CO₂. HPAI viruses A/Chile/25945/2023 (H5N1, clade 2.3.4.4b, genotype B3.2, Chile/23), A/Michigan/90/2024 (H5N1, clade 2.3.4.4b, genotype B3.13, MI/24), A/Missouri/121/2024-like (H5N1, clade 2.3.4.4b), A/British Columbia/PHL-2032/2024 (H5N1, clade 2.3.4.4b, genotype D.1.1), A/Wyoming/01/2025 (H5N1, clade 2.3.4.4b, genotype D.1.1), A/Cambodia/NPH230032/2023 (H5N1, clade 2.3.2.1e), A/Cambodia/NPH230776/2023 (H5N1, clade 2.3.2.1e), A/Cambodia/SVH240441/2024 (H5N1, clade 2.3.2.1e) and the candidate vaccine virus (CVV) of A/Astrakhan/3212/2020 (H5N8, clade 2.3.4.4b, Ast/20) were propagated in MDCK cells and titrated by plaque assay or 50% tissue culture infection dose (TCID₅₀) assay with standard procedures^59,60. The virus stocks were sequenced to confirm the absence of unwanted mutations and tested for exclusivity to rule out the presence of other subtypes of influenza virus.

Codon-optimized mRNA encoding a full-length HA protein from Ast/20 virus with the polybasic amino acids at HA cleavage site mutated from "REKRRKR" to "RETR" (GenBank, UJS29065) was in vitro transcribed with modified nucleosides as previously described⁶¹. Following in-vitro transcription, the mRNA was formulated into lipid nanoparticles using an ethanolic lipid mixture of ionizable cationic lipid as previously described⁶¹. Briefly, the lipid nanoparticle contains mRNA, an ionizable lipid, ((4-hydroxybutyl) azanediyl)bis(hexane-6, 1-diyl)bis(2-hexyldecanoate)), a PEGylated lipid, 2-((polyethylene glycol)-2000]-N,N-ditetradecylacetamide, a phospholipid (1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC]) and a sterol lipid (cholesterol). The expression of HA protein from the mRNA vaccine was examined in cells transfected with LNP-formulated H5 mRNA by flow cytometry. H5 mRNA vaccine was 2-fold serially diluted and added to an HEK-293T cell monolayer in 12 well plates (4.0 × 10⁵ cells/well and 0.97 to 500 ng of mRNA/well). HA protein expression was measured using the broadly reactive, stem-specific anti-HA human monoclonal antibody FI6 (Creative Biolabs). The percentage of live cells expressing the HA protein (% HA+ cells) was enumerated by quantifying the number of cells that had a positive signal for bound anti-HA antibody.

The recombinant H5 protein ectodomain of Ast/20 HA was codon-optimized for expression in insect cells. The protein was expressed as trimers with a C-terminal T4 Foldon motif and His-tag in a baculovirus expression system as described previously^62–64. The secreted recombinant HA protein was recovered from the cell culture supernatant by tangential flow filtration through a 30 kDa molecular weight cut-off membrane, followed by metal affinity chromatography and size exclusion chromatography (SEC). Stock concentration of recombinant H5 HA protein was diluted in PBS and mixed with AddaVax adjuvant (InvivoGen) to achieve the desired dose for vaccine administration in animals.

H5 mRNA was diluted in Opti-MEM (Thermo Fisher Scientific) and added to an HEK-293T cell monolayer at increasing RNA dose levels (0.97 to 500 ng/well). Opti-MEM media alone was used as a negative control. Protein expression was measured with a flow cytometer (BD FACS Fortessa) using the broadly reactive, stem-specific anti-HA human monoclonal antibody FI6 (Creative Biolabs) followed by a secondary anti-rabbit antibody conjugated to Alexa-Fluor 488 (Thermo Fisher Scientific). The percentage of cells expressing the H5 HA protein was enumerated by quantifying the number of live cells that had a positive signal for bound anti-HA antibody.

Male ferrets (Triple F Farms), which were confirmed serologically negative to circulating seasonal influenza A and B viruses, aged 7–12 months, were used in this study. The animals were housed in Duo-Flow Mobile units (Lab Products, LLC) throughout the experiments. Thirty µg of H5 mRNA vaccine was diluted in physiological saline to a final volume of 300 µL. Recombinant H5 HA protein was diluted in PBS and mixed at a 1:1 ratio with AddaVax adjuvant (InvivoGen) to a final volume of 300 µL. All animals were vaccinated with 300 µL of vaccine or PBS intramuscularly in a thigh. Serum samples were collected at the time points indicated in Figs. 1A, 3A, from the cranial vena cava of ferrets while under anesthesia (ketamine/xylazine).

For the challenge experiment, vaccinated ferrets were anesthetized and intranasally inoculated with 10⁴ TCID₅₀ of the Chile/23 HPAI A(H5N1) virus at 5 weeks post-second vaccination. Clinical signs of disease, body weight, and body temperature were monitored daily for 14 days. Nasal wash (NW) samples were collected from all remaining animals in each group on days 2, 4, 6, 8, and 10 post-challenge. Organ samples (lung, trachea, nasal turbinates, spleen, liver, kidney, small intestine, large intestines, brain tissues, and olfactory bulb) were collected for virus titration and histopathological analysis on day 5 post-challenge. NW samples and organ samples were titrated by TCID₅₀ assay on MDCK cells as described previously²⁷.

Male ferrets (Triple F Farms) were intranasally administered 10⁶ PFU of MI/24 A(H5N1) virus in 1 mL PBS (0.5 mL per nostril) on week 8, 5 weeks following the second vaccination. To establish contact, at 24 hours post-inoculation, an uninfected ferret (naïve or vaccinated depending on group) was placed in the same cage as each inoculated ferret. Four separate groups of ferrets (three ferret pairs per group) were established to assess the effect of vaccination on infection, virus shedding, and transmission (Fig. 3A): Group 1. Naïve inoculated ferrets housed with naïve contacts. Group 2. H5 mRNA-vaccinated inoculated animals housed with naïve contacts. Group 3. Adjuvanted H5 protein-vaccinated inoculated ferrets housed with naïve contacts. Group 4. Naïve inoculated ferrets were housed with H5 mRNA-vaccinated contacts.

NW samples were collected on days 1, 3, 5, 7, 9, 11, and 13 post-inoculation and post-contact by sedating the ferrets (ketamine/xylazine) followed by rinsing the nares with 1 mL PBS (0.5 mL each side) containing penicillin/streptomycin, gentamicin, and 0.1% bovine serum albumin (BSA). NW samples were titrated by TCID₅₀ assay on MDCK cells. Convalescent serum was obtained by collecting 1 mL of blood from the cranial vena cava on days 22 or 21 post-inoculation/contact, respectively. Whole blood was transferred to an SST tube (Becton Dickinson), centrifuged per the manufacturer instructions, transferred to a cryotube and stored at −80 °C until assays could be conducted.

HI assays were performed against the candidate vaccine virus (CVV) of Ast/20 or MI/24 virus with 1% horse red blood cells (HRBCs). Briefly, ferret serum samples were treated with receptor-destroying enzyme (RDE; Denka Seiken Co., Ltd) at 37 °C for 16 to 20 hours followed by inactivation at 56 °C for 30 min. After adsorption of non-specific agglutinins with HRBCs, the RDE-treated serum samples were serially diluted 2-fold with PBS and mixed with eight hemagglutination units of virus, followed by incubation at room temperature for 30 min. 1% HRBCs were then added to the mixtures and further incubated at room temperature for 60 min. HI titers are expressed as the highest reciprocal dilution that resulted in inhibition of agglutination.

ELISA was performed as described previously with slight modification^65,66. Recombinant monomer NA protein of clade 2.3.4.4b HPAI A(H5N1) virus, A/American wigeon/South Carolina/22-000345-001/2021, was obtained from Sino Biological. 96-well ELISA plates were coated with NA protein overnight at 4 °C. The plates were blocked with 1% BSA in PBS and then were incubated with four-fold serial diluted ferret sera. Anti-ferret IgG Primary Ab (ImmunoSpot) and a horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG (MilliporeSigma) were used as the secondary and tertiary antibody, respectively. The signal was developed using 3,3',5,5'-tetramethylbenzidine (TMB) (MilliporeSigma) as the substrate and stopped with TMB Stop Solution (SeraCare). The plates were read at a wavelength of 450 nm. The antibody titers are expressed as the highest reciprocal dilution that gave an OD above the cut-off value derived from normal ferret serum.

dPCR for the quantification of viral RNAs in ferret organ samples was performed as described previously²⁷. Briefly, RNA was extracted from organ homogenate in MagNA Pure LC RNA isolation tissue lysis buffer (Roche) with a MagNA Pure Cellular RNA Large Volume Kit (Roche) on the MagNA Pure 96 system (Roche). For absolute RNA quantification, the QIAcuity digital PCR System (Qiagen) was used. Reactions were prepared using the QIAcuity OneStep Advanced Probe Kit with InfA primer/probes from CDC rRT-PCR Flu Panel: Influenza A/B Typing Kit (VER 2; 510 (k) number K200370). Each RNA sample was loaded on QIAcuity 8.5k x 96-well plates with default parameters for priming and cycling conditions. Raw copy/µL numbers for the reaction mixture were converted to copy/mg tissue. A stock virus with a known titer was used as a positive control, while a buffer-only negative control was included during RNA extraction and carried through the quantification process.

Ferret tissues were fixed in 10% neutral buffered formalin and processed as previously described²⁷. Hematoxylin and eosin (H&E) staining and immunohistochemistry were performed by the Department of Pathology, College of Veterinary Medicine, University of Georgia. For immunohistochemistry, sample slides were pre-treated with citrate and pressure cooked at 110 °C for 15 min. They were then blocked with 3% hydrogen peroxide for 20 min and then Power Block for 5 min. The slides were incubated with goat polyclonal influenza A virus antibody (ab20841, Abcam) for 1 hour, with biotinylated rabbit anti-goat IgG secondary antibody (BA-5000-1.5, Vector Laboratories) for 10 min, and then with 4plus Streptavidin AP Label (AP605, Biocare Medical) for 10 min. Warp Red (Biocare Medical) was used as chromogen. The slides were evaluated qualitatively in a blinded manner by light microscopy by a board-certified veterinary pathologist using a Zeiss Axio Imager M2m microscope. Slides were scanned using a Leica Aperio AT2 digital pathology slide scanner, and photomicrographs were captured from whole slide images using Aperio ImageScope (v12.3.2.8013).

The focus reduction assay was performed as previously described²⁷. Briefly, 2-fold serially diluted antiserum samples and 600 focus-forming units of virus were added to 96-well plates containing confluent MDCK cells and incubated at 37 °C for 2 hours. An equal volume of overlay, consisting of 1.2% Avicel RC/CL in 2X MEM with 1 μg/mL TPCK-treated trypsin, 0.1% BSA fraction V, and antibiotics, was then added. After 14 hours incubation, the cells were fixed with 70% cold ethanol at 4 °C for 30 min. After fixation, the fluorescent NA substrate (BTP3-Neu5Ac Na) was added to washed reaction plates to visualize the Foci quantity, replacing the immunostaining. The values of 50% neutralization (NT₅₀) were calculated in GraphPad Prism 10.3.

Statistical analyses of HI titers and body weight changes were performed using two-way analysis of variance (ANOVA), with timepoint and treatment group as independent variables. Log-transformed HI titers were analyzed to compare responses across vaccine groups over time. For post hoc comparisons following ANOVA, Šídák's multiple comparisons test was applied. For bar graph data, pairwise t-tests with Tukey-Kramer's adjustment for multiple comparisons were conducted on log-transformed viral titers and neutralizing antibody titers. Animal survival was analyzed using the log-rank (Mantel-Cox) test. All analyses were performed using GraphPad Prism version 10.3, and p values < 0.05 were considered statistically significant.