An H5N1 clade 2.3.4.4b virus vaccine that elicits cross-protective antibodies against conserved domains of H5 and N1 glycoproteins

Abstract

The continuous evolution and global spread of highly pathogenic avian influenza (HPAI) H5N1 viruses, particularly clade 2.3.4.4b, pose major challenges for pandemic preparedness. This study evaluates a low-dose inactivated split-virus vaccine derived from H5N1 clade 2.3.4.4b, formulated with an Alum/CpG adjuvant, in a preclinical female mouse model. The vaccine induces strong humoral and cellular immunity, generating high titers of cross-reactive antibodies against diverse H5 hemagglutinin (HA) and across different N1 neuraminidase (NA) glycoproteins. The Alum/CpG adjuvant supports substantial antigen dose sparing and promotes a balanced Th1/Th2 profile. Functional assays show potent virus neutralization, neuraminidase inhibition, and antibody-dependent cellular cytotoxicity, alongside robust antigen-specific CD4⁺ and CD8⁺ T cell responses, efficient control of lung viral replication, and reduced lung inflammation. Vaccinated mice are fully protected from lethal challenge with both homologous H5N1 clade 2.3.4.4b and heterologous clade 1 viruses, despite low hemagglutination inhibition (HAI) titers. Electron microscopy polyclonal epitope mapping shows serum antibodies recognizing multiple epitopes on homologous HA and NA, with cross-reactivity to conserved epitopes on heterologous proteins, indicating broad recognition. Together, these findings support this vaccine candidate as a promising strategy to provide broad, multifunctional, and durable immunity against current and emerging H5N1 threats.

Highly pathogenic avian influenza H5N1 viruses are of global concern. This study shows that a low-dose H5N1 clade 2.3.4.4b Alum/CpG-adjuvanted vaccine elicits broad, durable antibody and T cell responses and protects female mice against lethal homologous and heterologous H5N1 challenges.

Introduction

The ongoing global spread and rapid evolution of highly pathogenic avian influenza (HPAI) H5N1 viruses, particularly those of clade 2.3.4.4b (Fig. 1A), are a major threat to both animal and human health¹. Since their resurgence, these viruses have demonstrated remarkable geographic expansion, affecting a broad range of avian species and increasingly spilling over into mammals, including humans². The persistent circulation and genetic diversification of H5N1 clade 2.3.4.4b viruses have led to frequent outbreaks and significant economic losses in the poultry industry, and the emergence of H5N1 in dairy cattle has increased concerns about the potential for further adaptation to human hosts. These trends highlight the urgent need for improved and updated vaccines capable of providing broad and durable protection against emerging H5N1 variants.

Fig. 1. Circulation of H5N1 viruses, vaccine characterization, and vaccine dose de-escalation in mice.

A Timeline of global clade frequencies of H5N1 viruses (May-2020 to April-2025). The graph was adapted from Nexstrain.org. B Characterization of different vaccine components by SDS-PAGE and confirmation of the presence of H5 and N1 antigens by WB. The molecular weight of each band is indicated in kilodaltons. Representative results from a single SDS-PAGE and Western blot experiment are shown. C Experimental design for vaccine dose de-escalation with and without adjuvants. Groups of BALB/c mice (n = 5) were immunized either with a single dose or in a prime-boost regimen using various doses of split H5N1 vaccine (0.0015–1.5 µg HA/mouse). Adjuvanted groups received 10 µg CpG, 50 µg Alum, or both per mouse. Dose de-escalation was tested only in adjuvanted prime-boost groups. For single-dose and unadjuvanted groups, only the 1.5 µg HA/mouse condition was evaluated. All mice were subsequently challenged with the A/bald eagle/Florida/W22-134-OP/2022 (H5N1, 6:2 A/PR/8/34 (PR8)) virus, and body weight and survival were monitored for 14 days. D–F Geometric mean titers (GMT) of total IgG, IgG1, and IgG2a antibodies against the vaccine-related H5 HA from A/mallard/New York/22-008760-007-original/2022 are shown (n = 5). The black triangle indicates the different vaccine doses assessed from 1.5 to 0.0015 µg HA. Individual titers and GMTs are shown; statistical comparisons were performed within the same vaccine platform (CpG, Alum, or Alum/CpG), including  the unadjuvanted group but excluding saline controls. G Body weight loss and Kaplan–Meier survival curves of mice (n = 5) challenged with 25 × LD₅₀ of the challenge virus. Data were shown as mean ± standard deviation (SD). The dotted horizontal line indicates the humane endpoint (25% body weight loss). One mouse in the CpG (0.015 µg) and one in the Alum/CpG (0.0015 µg) groups were excluded due to early death unrelated to virus infection. AUC area under the curve, MW molecular weight, WB Western blot, SA saline, UN unadjuvanted, PO or p.o. prime only. Data were analyzed by Kruskal–Wallis test followed by Dunn’s post hoc test and FDR correction; statistically significant p values (<0.05) are shown. Source data are provided as a Source Data file.

Current pandemic preparedness strategies rely mainly on stockpiled unadjuvanted H5 vaccines or H5 vaccines formulated with MF59, AS03 or aluminum-based adjuvants³. While oil-in-water emulsions can enhance antibody responses and offer some cross-clade reactivity, their efficacy is increasingly undermined by antigenic drift and the static nature of vaccine stockpiles⁴. As the antigenic landscape of H5N1 continues to evolve, there is a growing risk that existing vaccines may provide suboptimal protection against newly emerging clade 2.3.4.4b viruses⁴. To address these challenges, research is increasingly focused on advancing H5 vaccine design by integrating updated antigens with innovative adjuvant systems that can elicit more potent and broadly protective immune responses. Unlike traditional adjuvants such as MF59⁵ and AS03⁶, which primarily drive a Th2-skewed response and are generally more reactogenic⁷, the combination of US FDA-approved aluminum hydroxide gels and a toll-like receptor 9 (TLR9) CpG adjuvant can synergistically activate humoral and cellular pathways, fostering a more balanced Th1/Th2 profile and enabling significant antigen dose sparing^8,9.

Importantly, while hemagglutination inhibition (HAI) titers remain a well-established correlate of protection for influenza vaccines, accumulating evidence indicates that HAI titers explain only a portion of the overall protective effect. Alternative mechanisms, including neutralizing antibodies that target non-HAI epitopes, neuraminidase-inhibiting (NAI) antibodies, antibody-dependent cellular cytotoxicity (ADCC), and T cell responses, are increasingly recognized as essential contributors to broad and durable immunity^10,11. There is also limited information on the specific epitopes targeted by immune responses elicited by current and experimental pandemic H5 vaccines, particularly on the H5 HA and especially on the N1 NA glycoproteins. Most studies have focused on the magnitude and breadth of antibody responses, with limited information on the epitopes targeted by cross-reactive polyclonal antibody responses^4,5,12. This lack of comprehensive epitope mapping restricts our understanding of the mechanisms underlying vaccine-induced protection and impedes the rational design of next-generation vaccines.

In this context, the present study provides a comprehensive evaluation of an H5N1 clade 2.3.4.4b inactivated split virus vaccine formulated with an Alum/CpG adjuvant in a preclinical mouse model. This vaccine was specifically developed to maintain robust immunogenicity against both the H5 HA and N1 NA antigens, thereby aiming for broader protective potential^13,14. Beyond quantifying the magnitude and breadth of antibody responses, this study explores the spectrum of cross-reactive epitopes recognized on both H5 HA and N1 NA glycoproteins across different viral clades by negative stain electron microscopy polyclonal epitope mapping (nsEMPEM). This study systematically examines alternative protective mechanisms, including virus neutralization, NAI, ADCC, and the induction of distinct T cell responses targeting different viral antigens. Through the integration of detailed functional assays with structural analyses, this work not only delineates the complex immunological landscape elicited by the vaccine but also offers critical insights to inform the rational design of next-generation pandemic H5N1 vaccines.

Results

Low-dose H5N1 clade 2.3.4.4b virus vaccine is highly immunogenic and confers protection against lethal challenge with a homologous virus

An inactivated split virus vaccine (Split) based on the clade 2.3.4.4b genotype B1.1 strain A/bald eagle/Florida/W22-134-OP/2022 (H5N1, 6:2 A/PR/8/34) was developed using a bioprocess designed to maximize HA and NA antigenicity^13,14. The presence of both H5 and N1 glycoproteins was confirmed by sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and Western blot (WB) (Fig. 1B). Different doses of this vaccine (0.0015–1.5 μg) were tested in a prime or prime-boost regimen with different adjuvants including CpG 1018^® (CpG), aluminum hydroxide gel (Alum), and the combination of both (Alum/CpG, Fig. 1C). Prime-boost vaccination resulted in a ~5–10-fold increase in H5 specific immunoglobulin G (IgG) titers compared to a single dose of the vaccine (Fig. 1D). Combination with Alum/CpG yielded the highest IgG titers in mouse sera, with similar titers elicited by different vaccine doses (0.015–1.5 μg). This vaccine dose sparing effect was not as pronounced with CpG or Alum alone, especially after boost. Interestingly, 0.015–0.15 μg of vaccine with CpG or Alum, or 0.0015 μg with Alum/CpG, elicited similar antibody titers to 1.5 μg of unadjuvanted vaccine. Mice in the Alum/CpG group displayed a more balanced Th1 (IgG2a)/Th2 (IgG1) polyclonal IgG response, whereas CpG and Alum alone skewed the immune responses to Th1 and Th2 profiles, respectively (Fig. 1E, F and Supplementary Fig. 1).

The protective efficacy of these different vaccine strategies against a highly lethal challenge dose of 25× the median mouse lethal dose (LD₅₀) with the homologous virus was assessed (Fig. 1G). All mice in the saline group and the lowest vaccine dose (0.0015 μg) with CpG quickly succumbed to the viral infection (100% mortality), and mice vaccinated with the lowest dose with Alum (0.0015 μg) were partially protected (60% mortality). Mice receiving the unadjuvanted vaccine or a single dose (p.o.) of vaccine (1.5 μg) with CpG were completely protected from mortality but experienced remarkable body weight loss. The rest of mice groups vaccinated with either a single dose or in a prime-boost regimen were fully protected and exhibited little to no body weight loss.

Vaccine combination with diverse adjuvants enhances cross-reactive antibody responses against heterologous H5 and N1 glycoproteins and induces durable immunity

Considering the excellent vaccine immunogenicity profile resulting in high IgG titers in sera and complete protection at low doses, 0.15 μg of split vaccine was selected as dose for the following studies (Fig. 2A). Sera from mice vaccinated in a prime-boost regimen were analyzed for IgG cross-reactivity against different recombinant HA and NA glycoproteins (Fig. 2B). Vaccine combination with adjuvants increased the magnitude and cross-reactivity of IgG responses. This was especially apparent after a single vaccine dose with Alum/CpG. Boosting with an additional dose increased IgG responses in all groups of mice, with Alum/CpG attaining the highest titers, followed by Alum, CpG, and the unadjuvanted vaccine. High IgG cross-reactivity titers could also be detected against other clade 2 H5 HAs, including clade 2.3.4.4c, the more distant clade 2.3.2.1.c, and even against the clade 1 H5 HA from A/Vietnam/1203/2004. Little cross-reactivity was detected against the current seasonal H1 HA or to the H1 HA stalk (mini HA). Notably, this vaccine induced robust anti-N1 IgG responses across a broad range of N1 NA glycoproteins, targeting not only the matched strain but also phylogenetically distant N1 NAs, including those from A/Vietnam/1203/2004 and from recent seasonal strains such as A/Michigan/45/2015 and A/Wisconsin/67/2022.

Fig. 2. Analysis of antibody cross-reactivity and durability of antigen secreting plasma cell responses.

A Different groups of BALB/c mice (n = 5) were prime-boosted with 0.15 µg HA/mouse of split vaccine adjuvanted with 10 µg/mouse of CpG, 50 µg/mouse of Alum, or both. Spleens, lungs, and bone marrow were collected, and mice (another set of n = 5) were challenged with A/bald eagle/Florida/W22-134-OP/2022 (H5N1, 6:2 A/PR/8/34) virus and A/Vietnam/1203/2004 (H5N1, 6:2 A/PR/8/34) virus. B GMT of serum IgG responses of mice (n = 5) against different HA and NA glycoproteins analyzed by ELISA. C Quantification of IgG BMBCs in mice (n = 4) at 3 months post-vaccine prime against different HA and NA glycoproteins. One representative mouse per group is shown (left). Box plot (right) with counted spots per 10⁶ BMBCs in the ELISpot assay for different HA (top) and NA (bottom) glycoproteins. The average number of counted spots per mouse is shown, together with the group mean, as well as the mouse with the lowest mean (lower limit) and the mouse with the highest mean (upper limit). Statistical comparisons between groups were conducted for each different glycoprotein excluding the saline group. BE22: A/bald eagle/FL/W22-134-OP/2022, DC 24: A/dairy cow/Texas/24-008749-001-original/2024, CB23: A/Cambodia/NPH230032/2023, VN04: A/Vietnam/1203/2004, WI 22: A/Wisconsin/67/2022, MA 00: A/mallard/Alberta/24/2001, DA 21: A/Darwin/6/2021, Unadj.: unadjuvanted. Data were analyzed using the non-parametric Kruskal–Wallis test followed by Dunn’s post hoc test with FDR correction. Statistically significant p values (<0.05) are shown. Source data are provided as a Source Data file. Illustrations from NIAID NIH BioArt Source (bioart.niaid.nih.gov/bioart/000056, bioart.niaid.nih.gov/bioart/000243).

To analyze the capacity of this vaccine to sustain long-term cross-reactive H5 HA and N1 NA IgG responses, bone marrow antibody secreting B-cells (BMBC) were analyzed 2–3 months post vaccination (Fig. 2C and Supplementary Fig. 2). Mice receiving the unadjuvanted vaccine showed limited induction of cross-reactive BMBCs beyond the homologous H5 HA and N1 NA glycoproteins. A higher number of moderately cross-reactive BMBCs were detected in the Alum and CpG groups, but persistence of cross-reactive BMBCs against the most phylogenetically distant HA and NA antigens was only detected in mice that received the split vaccine in combination with Alum/CpG.

Enhanced antibody and cellular immune functions lead to decreased viral replication and inflammation

Analysis of HAI responses in mice sera against the homologous strain and a related clade 2.3.4.4b genotype B3.13 (dairy cow) virus isolate indicated that vaccine prime was not sufficient to elicit HAI antibodies. Low HAI titers were detected after boosting in 80% of the mice in the Alum/CpG group (Fig. 3A, C). However, neutralizing antibodies titers against the homologous virus were detected in all mice after the prime dose, regardless of adjuvanted formulation, and subsequently boosted by ~3–4-fold after the second vaccine dose (Fig. 3B). Cross-neutralizing antibody titers against the dairy cow virus isolate were the highest in the Alum/CpG group (Fig. 3D). Almost all mice (9/10) in the Alum/CpG group exhibited high NAI antibody titers after prime in comparison to the other groups. After boosting, mice in the Alum and Alum/CpG groups showed the highest NAI antibody titers (Fig. 3E). Similarly, antibodies in the sera of mice vaccinated with Alum/CpG displayed the highest level of ADCC in vitro, while little to no ADCC activity could be detected in the unadjuvanted vaccine group (Fig. 3F).

Fig. 3. Assessment of vaccine-mediated antibody functions, cellular immunity, and viral replication in lungs.

A, B Sera from vaccinated mice (n = 10) were tested for HAI and virus neutralization in a microneutralization (MNT) assay against A/bald eagle/Florida/W22-134-OP/2022 (H5N1, 6:2 A/PR/8/34) and C, D A/dairy cow/New Mexico/A240920343-93/2024 (H5N1), respectively. For the cow isolate, n = 5 post-boost sera were randomly selected for HAI and MNT under BSL-3 conditions. Positive controls included an anti-A/bald eagle/Florida/W22-134-OP/2022 head monoclonal antibody 1A1³⁶ (30 µg/mL) in HAI, mouse anti-A/bald eagle/Florida/W22-134-OP/2022 sera in MNT, and ferret antisera for both HAI and MNT against the cow isolate. Individual titers and GMT (gray bars) are shown. E NAI activity of sera (n = 10) against A/bald eagle/Florida/W22-134-OP/2022 (H5N1, 6:2 A/PR/8/34), expressed as the serum dilution inhibiting 50% NA activity (ID₅₀); individual and geometric mean ID50 values are shown. F ADCC activity of sera (n = 10) against A/bald eagle/Florida/W22-134-OP/2022 (H5N1, 6:2 A/PR/8/34), reported as fold induction of reporter signal over blanks; individual values and GMT are shown. The positive control is the anti-HA stalk monoclonal antibody KB2⁴⁰ (30 µg/mL). G Lung viral loads in BALB/c mice vaccinated in a prime-boost regimen and challenged with 0.5 × LD₅₀ of A/bald eagle/Florida/W22-134-OP/2022 (H5N1, 6:2 A/PR/8/34); lungs were collected on days 3 (n = 5) and 6 (n = 5) post-challenge. The limit of detection was 50 PFU/mL; negative samples were assigned 25 PFU/mL. Individual titers and GMT (gray bars) are shown. H, I Frequency of nucleoprotein (NP)-specific CD4⁺ effector memory T cells in lungs and spleens of vaccinated BALB/c mice (n = 5) on day 5 post-challenge with 0.5 × LD₅₀, following ex vivo stimulation with an NP peptide pool from A/California/04/2009 (H1N1); mean ± SD is shown. J Star plot of mean log₁₀ cytokine concentrations in lungs on days 3 (n = 5) and 6 (n = 5) post-challenge. Data were analyzed using the non-parametric Kruskal–Wallis test followed by Dunn’s post hoc test with FDR correction. Statistical comparisons between vaccine groups excluded the saline group, except for viral titers and T cell analyses. Only p values (<0.05) are shown. Pos. control positive control. Source data are provided as a Source Data file. Illustration from NIAID NIH BioArt Source (bioart.niaid.nih.gov/bioart/000243).

Viral replication in the lungs was effectively controlled in all groups receiving vaccines combined with Alum, CpG, or Alum/CpG, with minimal to no detectable virus even at the earliest infection stages (Fig. 3G). This control was accompanied by robust CD4⁺ T cell responses targeting NP, H5 HA, and N1 NA, particularly in the Alum/CpG group (Fig. 3H, I and Supplementary Figs. 3–6), together with a marked increase in H5 HA-specific CD8⁺ T cells in the lungs of the same mice (Supplementary Fig. 7). Mice receiving adjuvanted vaccines, especially Alum/CpG, also showed reduced lung inflammation after infection, as evidenced by lower levels of interleukin (IL)-6, interferon (IFN)-α, IFN-γ, tumor necrosis factor (TNF)-α, C-X-C motif chemokine ligand 10 (CXCL10), and other pro-inflammatory mediators compared with saline and unadjuvanted controls, which showed persistently high cytokine levels from early (day 3) through later (day 6) stages of infection (Fig. 3J). Notably, the unadjuvanted split vaccine, and to a lesser degree the Alum group, induced partial Th2 polarization characterized by IL-4, IL-5, and IL-13 production, whereas the Alum and Alum/CpG groups showed a transient, modest shift toward a Th17 phenotype, with increased IL-22 and IL-17 early after infection. These immunological differences between adjuvant groups were most pronounced during the early phase of infection (Fig. 3J and Supplementary Fig. 8).

Mice vaccinated with an H5N1 clade 2.3.4.4b vaccine in combination with Alum/CpG are fully protected against lethal challenge with a heterologous clade 1 virus

The vaccine’s capacity to elicit diverse antibody responses and confer protection against phylogenetically distant H5N1 viruses was evaluated using the clade 1 A/Vietnam/1203/2004 (H5N1, 6:2 A/PR/8/34) virus (Fig. 4A, green). HAI titers were undetectable after two doses of the split vaccine, regardless of adjuvant formulation (Fig. 4B). Nevertheless, low levels of neutralizing antibodies were detected after priming in sera from all mice in the Alum and Alum/CpG groups, and all mice developed neutralizing antibodies after two 0.15 μg HA doses, with the highest titers in these two groups (Fig. 4C). Following priming, 60% of mice in the Alum/CpG group developed NAI antibodies, a response absent in the other groups. Booster immunization led to a tenfold increase in NAI titers in the Alum/CpG group, with all Alum-vaccinated mice also seroconverting (Fig. 4D). Notably, 50% of the mice in the Alum/CpG group also developed NAI antibodies against the more genetically divergent seasonal N1 NA from A/Michigan/45/2015 (Supplementary Fig. 9). Sera from mice vaccinated with CpG and Alum/CpG exhibited the strongest ADCC activity in vitro against A/Vietnam/1203/2004, with lower ADCC activity in the Alum group and minimal activity in the unadjuvanted group (Fig. 4E).

Fig. 4. Cross-clade antibody functions and protection against a phylogenetically distant H5N1 virus.

A Cladogram of HAs from representative H5N1 influenza viruses. The HA from the clade 2.3.4.4b vaccine strain is highlighted with an asterisk. The HA from the clade 1 A/Vietnam/1203/2004 (H5N1, VN04) virus is highlighted in green. The tree was constructed using amino acid sequences aligned in Clustal Omega and visualized with FigTree. BE22: A/bald eagle/Florida/W22-134-OP/2022, DCNM24: A/dairy cow/New Mexico/A240920343-93/2024, DCTX24: A/dairy cow/Texas/24-008749-001-original/2024, CG22: A/Canada goose/New York/NYCVH 22-9190/2022, BC24: A/British Columbia/PHL-2032/2024, CK14: A/chicken/Netherlands/14015531/2014, IN05: A/Indonesia/05/2005, CB23: A/Cambodia/NPH230032/2023. B, C HAI and neutralizing activity of mouse (n = 10) sera were evaluated against A/Vietnam/1203/2004 (H5N1, 6:2 A/PR/8/34) virus. A mouse anti-A/Vietnam/1203/2004 serum was used as the positive control for the HAI and MNT assays. Individual and GMT (gray bars) values for HAI and MNT titers are shown. D NAI assay of sera (n = 10) against A/Vietnam/1203/2004 (H5N1, 6:2 A/PR/8/34) virus. Individual and geometric mean ID₅₀ values are shown. E Serum of vaccinated mice (n = 10) was analyzed for ADCC activity against the A/Vietnam/1203/2004 (H5N1, 6:2 A/PR/8/34) virus using a reporter assay. The fold induction of the reporter signal from individual mouse serum over those from blanks were analyzed and plotted as individual and GMT values. The positive control is the monoclonal antibody KB2 (30 µg/mL). F–I Percentage body weight loss and Kaplan–Meier survival plots of mice (n = 5) challenged with 5 × LD₅₀ and 25 × LD₅₀ of the A/Vietnam/1203/2004 (H5N1, 6:2 A/PR/8/34) virus. Average body weight values and SD are shown. Transverse dotted line denotes the humane endpoint (25% of body weight loss). Statistical comparisons between groups were conducted excluding the saline group. Data were analyzed using the non-parametric Kruskal–Wallis test followed by Dunn’s post hoc test with FDR correction. Statistically significant p values (<0.05) are shown. Pos. control: positive control. Source data are provided as a Source Data file.

Challenge with a lethal dose (5× LD₅₀) of the clade 1 A/Vietnam/1203/2004 (H5N1, 6:2 A/PR/8/34) virus resulted in complete protection and minimal morbidity in mice receiving the split vaccine formulated with Alum, CpG, or Alum/CpG, whereas saline and unadjuvanted vaccine groups either succumbed to infection or were only partially protected (Fig. 4F, G). To further evaluate protective capacity, a new set of mice was challenged with a fivefold higher lethal dose (25× LD₅₀) of the same virus. Mice in the Alum/CpG and Alum groups were fully protected from mortality, but 3/5 mice in the Alum group lost substantial body weight, while no sign of morbidity was observed in mice vaccinated with Alum/CpG. Eighty percent of the mice in the CpG group were protected from mortality and exhibited low signs of morbidity. Mice vaccinated with saline or the unadjuvanted vaccine succumbed to the infection or were barely protected (1/5), respectively (Fig. 4H, I).

Epitope mapping identifies conserved sites on H5 and N1 glycoproteins as targets of cross-reactive antibodies

To dissect the epitopes targeted by the polyclonal antibody (pAb) response induced by H5N1 vaccination, nsEMPEM was performed on serum pooled from five mice per group 3–4 weeks after the second boost. For the vaccine strain-matched A/bald eagle/Florida/W22-134-OP/2022 H5 HA, HA-specific fragment antigen-binding (Fab) pAbs were evident in 2D class averages, and pAb–HA complexes were reconstructed in 3D for all vaccinated groups (Fig. 5A and Supplementary Fig. 10). All vaccinated groups exhibited pAbs targeting the HA head, specifically the side of the head, the non-receptor binding site (RBS) upper head, and the vestigial esterase epitopes (Fig. 5B), whereas no HA-specific response was detected in the saline group (Fig. 5B). When complexed with NA, vaccinated groups showed pAbs directed to the top, side, and underside of NA (Fig. 5C). For A/Vietnam/1203/2004, 2D class averages revealed HA stem-specific pAbs in all adjuvanted groups (Fig. 5A and Supplementary Fig. 10), and pAbs recognizing the vestigial esterase epitope were also observed in the Alum group. No HA-specific response was detected in the saline or unadjuvanted group (Fig. 5B). When complexed with NA, vaccinated groups showed pAbs targeting the top and side regions of NA (Fig. 5C). For analysis of particle distribution profiles, responses to the central stem and anchor epitopes were grouped as “Stem”, whereas responses to the non-RBS upper head, side head, and vestigial esterase epitopes were grouped as “Head”. For A/bald eagle/Florida/W22-134-OP/2022, a robust HA head-directed response was observed, with no unbound HA detectable in the Alum or Alum/CpG groups (Fig. 5D). A similar pattern was seen for the NA, with a higher proportion of Fab-antigen complexes in the Alum and Alum/CpG groups, corresponding to 92.2% and 76.8% bound NA, respectively (Fig. 5E). For A/Vietnam/1203/2004, fewer HA- and NA-targeting pAbs were detected than for the vaccine strain-matched antigens. When comparing particle distributions between HA and NA antigens from each strain, A/bald eagle/Florida/W22-134-OP/2022 showed higher frequencies of HA-pAb than NA-pAb complexes, whereas the opposite pattern was observed for A/Vietnam/1203/2004, with more pAbs directed against NA than HA (Fig. 5D, E). Overall, pAbs recognized multiple head epitopes on the homologous HA at high abundance but targeted the stem of the heterologous HA at lower frequency, while pAbs against homologous and heterologous NAs predominantly bound the top and side regions of NA.

Fig. 5. nsEMPEM analysis of polyclonal IgG Fab responses in mice 3–4 weeks after second boost.

A Negative stain electron microscopy reconstructions of purified polyclonal IgG Fab bound to A/bald eagle/Florida/W22-134-OP/2022 H5 and N1 and A/Vietnam/1203/2004 H5 and N1 glycoproteins. Representative, false-colored 2D classes are presented for epitopes that could not be reconstructed. 3D reconstructions that illustrated epitopes targeted but were poorly resolved are presented with cartoon Fabs. B, C Summary of epitopes targeted by pAbs against the H5 HA (B) and N1 NA (C) of A/bald eagle/Florida/W22-134-OP/2022 and A/Vietnam/1203/2004. D, E Particle distribution bar charts of free antigen or immune complexed particles targeting HA (D) or NA (E) observed by nsEMPEM. BE22: A/bald eagle/Florida/W22-134-OP/2022, VN04: A/Vietnam/1203/2004. Source data are provided as a Source Data file.

Discussion

The present study demonstrates that an inactivated split virus vaccine based on H5N1 clade 2.3.4.4b, when formulated with optimized adjuvant combinations, is highly immunogenic and confers robust protection against both homologous and heterologous lethal H5N1 challenges in mice. The combination of Alum and CpG as adjuvants enabled substantial dose sparing, with as little as 0.0015 μg of vaccine antigen eliciting IgG titers comparable to higher doses of unadjuvanted vaccine. The Alum/CpG combination not only increased the magnitude of the antibody response but also promoted a balanced Th1/Th2 profile, as evidenced by the IgG2a/IgG1 ratio, in line with the ability to synergistically activate both humoral and cellular arms of the immune system, a feature that has been shown to be critical for broad and durable protection against rapidly evolving influenza viruses¹⁵.

A key strength of this vaccine formulation is its capacity to induce cross-reactive antibodies against a range of H5 HA and N1 NA antigens, including those from phylogenetically distant clades such as 2.3.2.1c (A/Cambodia/NPH230032/2023)¹⁶ and clade 1 (A/Vietnam/1203/2004). The Alum/CpG vaccine elicited the highest titers of cross-reactive IgG and, importantly, only this group showed persistent BMBCs against the most divergent HA and NA antigens months after vaccination. The robust anti-N1 response is particularly noteworthy since NA is increasingly recognized as a critical target for cross-protective immunity against influenza A viruses^17,18. The capacity of the vaccine to elicit anti-N1 antibodies that cross-react with both homologous and heterologous N1 antigens, including those from clade 1 viruses, suggests that this formulation could provide a broader layer of protection, potentially limiting viral replication even in the face of antigenic drift and shift in the HA and NA glycoproteins.

The Alum/CpG vaccine induced not only high titers of binding antibodies but also potent functional responses, including neutralizing antibodies detectable after a single dose and further boosted after the second dose, with the highest titers in the Alum/CpG group. NAI antibodies were rapidly induced and maintained at high levels, particularly in the Alum/CpG and Alum groups. Sera from Alum/CpG‑vaccinated mice showed the strongest ADCC activity, a mechanism increasingly recognized as important for heterosubtypic protection¹⁹. These humoral responses were complemented by strong cellular immunity, with increased frequencies of antigen-specific CD4⁺ and CD8⁺ T cells, especially in the lungs. T cell responses were most robust against HA and NP, which likely reflects greater antigen abundance and availability for presentation. Polyfunctional T cell states were primarily observed following NP peptide pool stimulation, whereas stimulation with recombinant HA and NA proteins required longer incubation and may have underestimated polyfunctional responses due to protocol limitations. The Alum/CpG group exhibited the most robust T cell responses against HA, NA, and NP, which likely contributed to rapid control of viral replication and reduced lung inflammation after challenge, consistent with recent work linking broad T cell responses to cross-protection against diverse influenza virus strains²⁰. Remarkably, the Alum/CpG vaccine conferred complete protection against lethal challenge with a heterologous clade 1 H5N1 virus, even in the absence of detectable HAI titers, with protection associated with cross-neutralizing and cross-NAI antibodies²¹ as well as strong ADCC activity, highlighting the critical role of non-HAI mechanisms in mediating cross-clade immunity²².

A key innovation of this study is the structural mapping of pAb serum responses using nsEMPEM, which revealed that vaccination, particularly with Alum/CpG, elicited antibodies targeting conserved regions of both H5 HA and N1 NA glycoproteins. Within HA, responses focused on the vestigial esterase site^23,24, a critical and relatively conserved region among H5 strains that can mediate cross-neutralization despite being generally non-HAI active, as well as other non-HAI epitopes such as the side head and non-RBS upper head domains. Although epitope mapping of pAb responses to A/Vietnam/1203/2004 H5 HA was limited, likely due to low-affinity or low-resolution Fab-antigen interactions, cross-reactive antibodies against the central stem were consistently detected across all adjuvanted groups, potentially explaining the observed cross-neutralizing, non-HAI activity.^25,26. In NA, multiple epitopes were identified, including sites near the catalytic pocket and along the lateral face. Notably, epitopes along the NA side region dominated cross-reactive recognition of A/Vietnam/1203/2004 N1, in line with monoclonal antibody studies showing that conserved lateral-face epitopes confer broad NAI activity and in vivo protection^18,27. Antibodies targeting N1 NA may also have contributed to the high neutralization titers by limiting viral spread.

The targeting of conserved epitopes within both the HA head and NA by vaccine-induced antibodies further underscores the potential of this vaccine platform to provide broad and durable immunity against a spectrum of emerging H5Nx and HxN1 viruses. This breadth of protection is especially significant considering the ongoing evolution and geographic expansion of H5N1 clade 2.3.4.4b viruses, which have recently spilled over into mammals and humans, raising concerns about pandemic risk. Consequently, the inclusion of the N1 NA component alongside HA, rather than focusing solely on H5 HA as in many current vaccine strategies^28,29, may offer superior control of pandemic influenza virus transmission and disease by leveraging the complementary protective roles of both HA- and NA-specific antibodies.

Despite these advances, this study has several limitations. All experiments were conducted in mice, necessitating validation in more predictive models such as ferrets or non-human primates to confirm the durability and breadth of protection. Moreover, although multiple protective mechanisms were identified, the individual contributions of NAI, ADCC, T cell responses, and antibodies targeting specific nsEMPEM-defined epitopes remain to be fully dissected. Future work should also examine the impact of pre-existing immunity on vaccine efficacy and assess how rapidly this adjuvant platform can be adapted to counter emerging pandemic influenza threats.

In summary, this study shows that an H5N1 clade 2.3.4.4b split vaccine, particularly when adjuvanted with Alum/CpG, is highly immunogenic and compatible with dose-sparing strategies. It elicits broad, durable, and multifunctional immune responses that extend beyond traditional HAI readouts. By providing a comprehensive structural and functional roadmap, this work underscores the importance of targeting conserved epitopes on both HA and NA glycoproteins and offers valuable insights for the rational design of next-generation pandemic influenza vaccines.

Methods

Cell lines

Baculovirus generation and amplification was performed in Sf9 cells (CRL-1771, ATCC), grown in Trichoplusia ni medium-formulation Hink insect cell medium (TNM-FH, Gemini Bioproducts) supplemented with 10% v/v fetal bovine serum (FBS, Gibco), penicillin (100 U/mL) and streptomycin (100 μg/mL) solution (Gibco), and 0.1% v/v Pluronic F-68 (Gibco). High Five cells (BTI-TN-5B1-4, B85502, Thermo Fisher Scientific) were grown in Sf900 II SFM (Gibco) and used for recombinant HA and NA production. Both cell lines were grown at 27 °C. HEK293F cells (Thermo Fisher Scientific) cultured at 37 °C, 8% CO₂, shaken at 125 rpm in FreeStyle 293 expression medium (Gibco) and used for recombinant protein production.

Madin–Darby canine kidney (MDCK) cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% v/v FBS and penicillin (100 U/mL) and streptomycin (100 μg/mL) solution in a humidified incubator at 37 °C and 5% CO₂.

Recombinant proteins

Recombinant HA and NA proteins were produced in High Five cells³⁰, and the N1 from A/Vietnam/1203/2004 was expressed and purified from transiently transfected HEK293F cells³¹. Cell-culture supernatants were purified using Ni²⁺ -nitrilotriacetic acid agarose beads (Qiagen) after incubation at room temperature (RT) for 2 h or overnight (O/N) at 4 °C. Samples were subsequently run over a gravity flow column (Qiagen) and washed with phosphate-buffered saline (PBS) or tris-buffered saline (TBS) and 20 mM imidazole, pH 8.0. Proteins were eluted with 300–500 mM imidazole and buffer-exchanged into PBS or TBS three times on 30-kDa Amicon ultra centrifugal filters (Merck Millipore). Affinity-purified N1 from A/Vietnam/1203/2004 was further processed using size exclusion chromatography over a Superdex 200 Increase 10/300 column (Cytiva). Fractions corresponding to tetrameric NA were pooled, concentrated, and buffer-exchanged to TBS using a 30-kDa Amicon concentrator.

Viruses

A/bald eagle/Florida/W22-134-OP/2022 (H5N1, 6:2 A/PR/8/34)³², A/Vietnam/1203/2004 (H5N1, 6:2 A/PR/8/34), and H7N1_{A/Michigan/45/2015} viruses were generated by reverse genetics. The H5 and N1 glycoproteins were derived from the wildtype A/bald eagle/Florida/W22-134-OP/2022 (H5N1) and A/Vietnam/1203/2004 (H5N1) viruses, with removal of the H5 polybasic cleavage site. The H7 and N1 from the H7N1_{A/Michigan/45/2015} virus were derived from A/Shanghai/1/2013 (H7N9) and A/Michigan/45/2015 (H1N1) viruses. The internal gene segments of these three viruses belong to the donor vaccine strain A/Puerto Rico/8/1934 (H1N1, A/PR/8/34). The A/dairy cow/New Mexico/A240920343-93/2024 (H5N1) was isolated in MDCK cells from a milk sample provided by the Texas A&M Veterinary Medical Diagnostic Laboratory³³.

The A/bald eagle/Florida/W22-134-OP/2022 (H5N1, 6:2 A/PR/8/34), A/Vietnam/1203/2004 (H5N1, 6:2 A/PR/8/34), and H7N1_{A/Michigan/45/2015} viruses were grown in 10-day-old embryonated chicken eggs at 37 °C for 48 h, and cooled at 4 °C O/N. The A/dairy cow/New Mexico/A240920343-93/2024 (H5N1) was grown on MDCK cells. Cell debris was removed by low-speed centrifugation at 4000×g (TX-1000 rotor, Thermo Fisher Scientific), 4 °C and for 20 min. Viruses were aliquoted, stored at −80 °C, and titrated by the plaque assay method on MDCK cells³⁴.

Production of inactivated split influenza virus vaccines

A/bald eagle/Florida/W22-134-OP/2022 (H5N1, 6:2 A/PR/8/34) vaccine production was conducted as previously described^13,14. Briefly, virus inactivation was performed with 0.025% v/v beta-propiolactone (Millipore Sigma) prepared in ice-cold water for injection (Gibco) for 30 min after pH buffering with 0.01 M disodium hydrogen phosphate (Millipore Sigma) and stopped by incubation at 37 °C for 1 h. Then, the inactivated virus sample was centrifuged at 4000×g (TX-1000 rotor), 4 °C for 30 min. The clarified supernatant was loaded on 5 mL of 30% w/v sucrose cushion prepared in 1X NTE buffer consisting of 1 M NaCl, 100 mM Tris-HCl, 10 mM ethylenediaminetetraacetic acid (EDTA) in water for injection, with the pH adjusted to 7.4. The supernatant containing the inactivated virus was concentrated by high-speed centrifugation at 76,800×g (SW-32Ti rotor, Beckman Coulter), 4 °C for 2 h, and the pelleted virus was resuspended in TBS (pH 7.5). The resuspended virus was split with 1% v/v Triton X-100 (Fisher Bioreagents), and the detergent was removed by incubation with 0.2 g of Bio-Beads SM-2 (Bio-Rad) per mL of inactivated split virus. The supernatant was collected, and the total protein concentration was adjusted to 0.5 mg/mL in TBS (pH 7.5) using the Bradford assay (Bio-Rad). Vaccine samples were aliquoted and stored at −80 °C until use. The concentration of HA in the final A/bald eagle/Florida/W22-134-OP/2022 (H5N1, 6:2 A/PR/8/34) vaccine was quantified in an indirect enzyme-linked immunosorbent assay (ELISA) using the CR9114 human monoclonal antibody²⁶. An H5 recombinant protein standard of known concentration was also included for approximate absolute HA quantification.

SDS-PAGE and WB

An SDS-PAGE was performed to characterize the H5N1 inactivated split vaccine. Before running the SDS-PAGE, samples were deglycosylated with rapid PNGase F (New England Biolabs) according to the manufacturer’s instructions. After deglycosylation, 20 μL of the sample was mixed with 4X Laemmli buffer (Bio-Rad) containing 50 mM NuPAGE sample reducing agent (dithiothreitol, Thermo Fisher Scientific). Samples were then incubated at 90–95 °C for 10–15 min and run on a 4–20% precast polyacrylamide Mini-PROTEAN TGX gel (Bio-Rad) at 100 V for 10 min followed by 180 V for 35 min (30 μL/well). The gel was stained with InstantBlue Coomassie protein stain solution (Abcam) for 15 min before visualization. Images were taken in a Chemidoc MP Imaging System with Image Lab (Bio-Rad, v.6.1).

For WB, SDS-PAGE gels were transferred to nitrocellulose membranes using an iBlot 2 transfer device (Invitrogen) at 25 V for 7 min. After, membranes were blocked in 3% (w/v) milk powder in PBS containing 0.1 % v/v Tween 20 (PBS-T) blocking solution at 4 °C in mild rocking conditions O/N. Membranes were then washed three times in PBS-T for 5 min, and an anti-H5^35,36 or an anti-N1^18,37,38 monoclonal antibody cocktail  was added at ~10 µg/mL and incubated at RT for 1–2 h. After washing three times in PBS-T for 5 min, a 1:1 mix of horseradish peroxidase (HRP)-conjugated secondary anti-human IgG (Fc-specific, A0170, Invitrogen) and anti-mouse IgG (H&L, 610-603-002, Rockland) polyclonal antibodies were added at a 1:10,000 dilution at RT for 1 h. Blots were again washed with PBS-T and developed by adding KPL TrueBlue peroxidase substrate (Seracare). Images were taken in a Chemidoc MP Imaging System using Image (v.6.1).

ELISA

Immulon 4 HBX 96-well plates (Thermo Fisher Scientific) were coated with 2 µg/mL of recombinant protein (50 µL per well) in PBS (pH 7.4) at 4 °C O/N. The next day, plates were washed three times with PBS-T and blocked in blocking solution (3 % v/v goat serum, 0.5 % w/v non-fat dry milk in PBS-T) for 1 h at RT. After blocking, mouse serum was added to the first well at a 1:30 dilution (150 µL/well) and serially diluted 1:3 in blocking solution and incubated for 2 h at 20 °C. Plates were washed three times with PBS-T before adding the secondary antibody (100 µL/well). For total IgG quantification, a 1:3000 dilution of anti-mouse IgG (H&L) peroxidase conjugated in blocking solution was added. For IgG1 and IgG2a quantification, a 1:20,000 and 1:2000 dilution in blocking solution of rabbit anti-mouse IgG1 (PA1-86329, Invitrogen) or rabbit anti-mouse IgG2a (61-0220, Invitrogen) was added, respectively. Afterwards, plates were incubated for 1 h at 20 °C and then washed four times with PBS-T with shaking. To develop plates, 100 µL of O-phenylenediamine dihydrochloride (OPD) substrate (SigmaFast OPD, Millipore Sigma) was added to each well. After a 10 min incubation, the reaction was stopped by adding 50 µL of 3 M hydrochloric acid (HCl) to each well. The optical density at 490 nm (OD₄₉₀) was measured on a Synergy H1 microplate reader (BioTek). A cut-off value of the average of the OD₄₉₀ values of blank wells plus three times the standard deviation (SD) was established for each plate and used for calculating the area under the curve (AUC). AUC values were determined using GraphPad Prism (v.10.5).

ADCC reporter assay

ADCC activity in mouse sera was assessed using an FcγRIV cell-based ADCC reporter assay according to the manufacturer’s instructions (M1211, Promega). Briefly, white 96-well plates (Corning) were seeded with 2 × 10⁴ MDCK cells per well and incubated O/N at 37 °C and 5% CO₂. After 24 h, MDCK cells were washed with PBS and infected with A/bald eagle/FL/W22-134-OP/2022 (H5N1, 6:2 A/PR/8/34) or A/Vietnam/1203/2004 (H5N1, 6:2 A/PR/8/34) viruses at a multiplicity of infection of 5 at 37 °C and 5% CO₂ for 1 h. After this, the virus medium was aspirated, replaced with warm DMEM, and incubated O/N at 37 °C and 5% CO₂. The following day, the cell-culture medium was removed, and 25 µL of assay buffer [Roswell Park Memorial Institute (RPMI) 1640 medium supplemented with 4% v/v low IgG FBS, Gibco] was added to each well. Mouse sera previously heat-inactivated at 56 °C for 1 h were serially diluted twofold in RPMI 1640 medium and added to the infected MDCK cells (25 µL/well). The sera were incubated with MDCK cells at 37 °C for 30 min. Then, 7.5 × 10⁴ Jurkat cells expressing the mouse FcγRIV with a luciferase reporter gene under transcriptional control of the nuclear factor-activated T cell promoter were added per well (25 µL/well) and incubated at 37 °C for 6 h. After incubation, 75 µL of Bio-Glo luciferase assay reagent was added per well and incubated at RT in the dark for 10 min. The luminescence signal was measured using a Synergy H1 microplate reader. The fold induction was calculated as follows: (RLU_induced-RLU_background)/(RLU_{no antibody control}-RLU_background), where RLU is relative luminescence units. The AUC values of the resulting fold-induction values were calculated using GraphPad Prism (v.10.5).

NAI assay

The NA activity of A/bald eagle/Florida/W22-134-OP/2022 (H5N1, 6:2 A/PR/8/34), A/Vietnam/1203/2004 (H5N1, 6:2 A/PR/8/34) and H7N1_{A/Michigan/45/2015} viruses was assessed on Immulon 4 HBX 96-well plates coated with 100 µL of fetuin (Millipore Sigma) at 25 µg/mL in PBS at 4 °C O/N. Fetuin-coated plates were washed three times with PBS-T and blocked with PBS + 5% v/v bovine serum albumin (BSA, MP Biomedicals). On a separate plate, the virus was serially diluted 1:2 in PBS + 1% w/v BSA, and 75 µL of pre-diluted virus samples were added to fetuin-coated plates already containing 75 µL of PBS + 1% w/v BSA. The fetuin-coated plates were incubated at 37 °C O/N. Afterwards, plates were washed four times with PBS-T with shaking, and 100 µL per well of peroxidase-labeled peanut agglutinin from Arachis hypogaea (Millipore Sigma) at 5 µg/mL in PBS + 1% v/v BSA was added to the plates. Plates were incubated at 20 °C for 1.5 h before washing four times with PBS-T with shaking. To develop the plates, 100 µL of OPD substrate were added per well, incubated for 10 min at RT, and the reaction was stopped by adding 50 µL of 3 M HCl per well. The OD₄₉₀ was measured on a Synergy H1 microplate reader, and the half-maximal effective concentration (EC₅₀) was determined using GraphPad Prism (v.10.5).

For the NAI assay, Immulon 4 HBX 96-well plates were coated with 100 µL of fetuin at 25 µg/mL in PBS at 4 °C O/N. Fetuin-coated plates were washed three times with PBS-T and blocked with PBS + 5% v/v BSA. In parallel, heat-inactivated sera at 56 °C for 1 h were serially diluted 1:2 in PBS + 1% v/v BSA with a starting dilution of 1:30 in non-fetuin-coated 96-well plates (75 µL/well). Then, 75 µL of each virus corresponding to 2× EC₅₀ was added per well to the pre-diluted sera plates and incubated at 20 °C for 1.5 h. After incubation, 100 µL of the virus/serum mixture were transferred per well to fetuin-coated plates and incubated at 37 °C O/N. After incubation, the rest of the assay was performed as described above for the NA assay. No serum (virus only) and background controls (PBS + 1% v/v BSA only) were also included to measure the NAI. OD₄₉₀ was measured on a Synergy H1 microplate reader, and the half-maximal inhibitory concentration (IC₅₀) was calculated as: 1 – (OD_measured-OD_background)/(OD_{no serum control}-OD_background) in GraphPad Prism (v.10.5).

Microneutralization assay

Mouse sera were treated with receptor-destroying enzyme (RDE) II (Denka Seiken) and incubated in a 37 °C water bath for 18–20 h. The same day, MDCK cells were seeded in 96-well cell-culture-treated plates (Corning) at 1.8 × 10⁴ cells per well (100 µL/well) and incubated at 37 °C with 5% CO₂ O/N. The following day, the RDE activity was stopped by the addition of a 2.5% w/v sodium citrate solution and incubation at 56 °C for 1 h. RDE-treated sera were initially diluted 1:10 and serially diluted 1:2 in infection medium consisting of minimum essential medium (MEM) with 10 mM of 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 2 mM L-glutamine (Gibco), 3.2% w/v sodium bicarbonate (Corning), 1.2% w/v BSA, 100 U/mL penicillin and 100 µg/mL streptomycin (Gibco). Next, 120 µL of 100 × tissue culture infectious dose (TCID₅₀) of virus prepared in infection medium and 120 µL of serially diluted sera were incubated on a shaker at RT for 1 h. MDCK cells were washed with 220 µL of PBS and incubated with 100 µL of the incubated serum-virus mixture at 37 °C with 5% CO₂ for 1 h. Afterwards, the virus inoculum was carefully aspirated, MDCK cells were washed with PBS, and 100 µL of the serially diluted sera containing 1 µg/mL of N-tosyl-L-phenylalanine chloromethyl ketone (TPCK)-treated trypsin (Millipore Sigma)  was added to the cells and incubated at 37 °C with 5% CO₂ for 48 h. As a readout, the presence of the virus was assessed by hemagglutination assay. In brief, 50 µL of cell supernatant was added to 96-well V-bottom plates (Nunc) and serially diluted 1:2. Then, 50 µL of 0.5% v/v turkey red blood cells (RBCs, Lampire Biological Laboratories) in PBS were added to each well, and plates were incubated on ice or at 4 °C for 45 min. The HA titer was calculated as the endpoint titer at which no RBC tear drop formation could be detected.

HAI assay

A hemagglutination assay was initially performed to determine the hemagglutination titer units (HAU) of the viruses. Mouse sera were treated with RDE II and incubated in a 37 °C water bath for 18–20 h. The following day, the RDE activity was stopped by the addition of a 2.5% w/v sodium citrate solution and incubation at 56 °C for 1 h. RDE-treated sera were initially diluted 1:10 and serially diluted 1:2 in PBS in V-bottom 96-well plates. Twenty-five µl of serum dilutions were incubated with 25 µl of viruses diluted to 8 HAU at RT for 1 h. Following this, 50 µl of 0.5% v/v turkey RBCs diluted in PBS were added to the wells and incubated at 4 °C for 45 min, and the HAI titer was calculated as the endpoint titer at which no RBC tear drop formation could be detected.

Animal studies

All animal experiments were performed under protocols approved by the Icahn School of Medicine at Mount Sinai Institutional Animal Care and Use Committee (IACUC-2014-0254). For all animal experiments conducted, 6–8-week-old female BALB/c mice (Jackson Laboratories) were used unless otherwise mentioned. The CpG adjuvant (Dynavax Technologies) was used at 10 μg/mL, and the aluminum (Alum) hydroxide gel adjuvant (Alhydrogel 2%, InvivoGen) at 50 μg/mL. Mice were vaccinated via the intramuscular route with 0.0015–1.5 μg HA of H5N1 split vaccine prepared in Tris saline solution (20 mM Tris, 100 mM NaCl, pH 7.5) in a volume of 50 µL with or without adjuvants. A negative control (saline) consisting of Alum and CpG in Tris saline solution was also included. The vaccination regimen consisted of two sequential vaccinations 4 weeks apart. Four to six weeks after vaccination, mice were anesthetized and intranasally infected with 50 µL of influenza virus containing 0.1×, 5×, or 25×LD₅₀ depending on the experiment. Additionally, mice were bled via submandibular bleeding for serological analysis at this time point. Blood was incubated at RT for 1 h and centrifuged at 5000×g (FA-45-24-11 rotor, Eppendorf) for 30 min. Serum was separated from the pellet and stored at 4 °C until analysis. After virus challenge, body weight loss was monitored for 14 days, and mice showing a weight loss of ≥25% as compared with their initial body weight were humanely euthanized.

For the duration of the experiments, mice were housed in individually ventilated cages on a 12 h dark/light cycle with controlled temperature/humidity. Food and water were provided ad libitum.

Lung titers

Virus titers in the lungs of mice challenged with 0.5×LD₅₀ of the A/bald eagle/Florida/W22-134-OP/2022 (H5N1, 6:2 A/PR/8/34) virus at days 3 and 6 after challenge were quantified by the plaque assay method. Briefly, harvested lungs were homogenized in two disruption cycles (10 sec/cycle) using tubes that contained high-impact zirconium beads (Andwin Scientific) and 1 mL of PBS. For the plaque assay, lung homogenates were serially diluted 1:10 in PBS. Samples were incubated for 1 h with MDCK cells seeded at 3 × 10⁵ cells per well (1 mL/well) the day before in 12-well plates. After the 1 h incubation, an agarose overlay containing a final concentration of 0.64% w/v agarose (Oxoid) in MEM supplemented with 2 mM L-glutamine, 0.1% w/v of sodium bicarbonate, 10 mM HEPES, 100 U/mL penicillin, 100 µg/mL streptomycin, 0.2% w/v BSA, 1 µg/mL TPCK-treated trypsin, and 0.1% w/v diethylaminoethyl-dextran was added to the cells. The cells were incubated at 37 °C for 48 h, and visible plaques were counted after fixation with 3.7% v/v formaldehyde in PBS and visualization by immunostaining. All virus titers are presented as plaque-forming units (PFU)/mL.

T cell analysis

Mice were euthanized, the chest opened, and the right ventricle of the heart perfused with 10 mL of cold PBS. Lungs were processed using gentleMACS Octo Dissociator with Heaters (130-134-029, Miltenyi Biotech) and incubated in a collagenase-I (150 U/mL, 17100017, Gibco)/DNAse-I (50 U/mL, D5025-15KU, Millipore Sigma) solution on a 37 °C water bath for 30 min. Following digestion, tissue was filtered through a 70-µm cell strainer (229484, Celltreat). Spleens were mechanically disrupted with pestle homogenizers for 1.5 mL tubes (Fisher Scientific) and filtered as described above. Erythrocytes were lysed with RBC lysis buffer (420302, BioLegend) in accordance with the manufacturer’s instructions. Cells were washed with 2% v/v FBS in PBS and reconstituted in 5 mL of complete media (RPMI 1640 supplemented with 10% v/v FBS, 100 U/mL penicillin, and 100 µg/mL streptomycin). Cells were counted in C-Chip™ Disposable Hemacytometers (22-600-107, Thermo Fisher Scientific) using a 0.4% v/v Trypan blue solution (15250061, Thermo Fisher Scientific) to discriminate between live and dead cells. 2 × 10⁶ live cells per well were seeded in 96-well cell-culture-treated plates. For antigen-specific T-cellular immune response analysis, cells were stimulated with NP peptide pool PepTivator® Influenza A/California/04/2009 (H1N1) NP (130-097-278, Miltenyi Biotec) at 5 µg/mL, or with recombinant H5 and N1 glycoproteins (5 µg/mL) from A/bald eagle/Florida/W22-134-OP/2022 (H5N1). Stimulation with the NP peptide pool was performed at 37 °C with 5% CO₂ for 6 h in the presence of 5 µg/mL BD GolgiPlug (555029, BD Biosciences), 2 µM Monensin (420701, BioLegend), and 25 µg/mL of co-stimulatory anti-CD28 antibodies (102121, BioLegend). Stimulation with the H5 and N1 glycoproteins was performed at 37 °C and 5% CO₂ for 18 h, with 5 µg/mL Brefeldin A, 2 µM Monensin, and 25 µg/mL anti-CD28 antibodies added 8 h after initial stimulation with the glycoproteins.

Following stimulation, cells were stained with fluorescently labeled antibodies against surface markers to discriminate naïve, effector memory, and central memory T‑lymphocyte subsets. Each sample was stained with 0.5 µL of CD3-BV711 (100241, BioLegend), CD4-PerCP/Cy5.5 (65-0041-U100, Tonbo Biosciences), and CD62L-APC/Cy7 (104427, BioLegend), as well as 0.25 µL of CD8-BV785 (100750, BioLegend) and CD44-PE/Cy7 (600441U100, Tonbo Biosciences). Dead cells were excluded using 1 µL per sample of Zombie Aqua viability dye (423101, BioLegend). Intracellular cytokine staining was performed with the Cytofix/Cytoperm Fixation/Permeabilization Solution Kit (554714, BD Biosciences) according to the manufacturer’s instructions. 0.25 µL per sample of TNFα-AF488 (506313, BioLegend), 1 µL of IFNγ-BV421 (502532, BioLegend), and 0.5 µL of IL-2-PE (503808, BioLegend) antibodies were used to identify the corresponding cytokines. Data were collected on a BD FACSymphony™ A5 SE flow cytometer (BD Biosciences) using BD FACSDiva 9.0 and analyzed with FlowJo (v.10.8.1). To calculate statistical differences between groups, the percentage of cytokine-producing cells in non-stimulated samples was subtracted from the corresponding values for peptide-stimulated samples.

B-cell enzyme-linked immunosorbent spot (ELISpot) assay

After euthanizing the mice, femurs and tibias were aseptically harvested, cleaned of muscle tissue, both ends of each bone cut, and placed in cold RPMI 1640 medium containing 2% v/v FBS. Bone marrow cells were collected by centrifugation at 400×g (TX-1000 rotor) for 7 min using a nested Eppendorf tube setup, in which the inner tube contained a small hole at the bottom to allow cell flow‑through. The resulting cell suspension was gently pipetted up and down to disperse cell clumps and passed through a 70-μm cell strainer to obtain a single-cell suspension. RBCs were lysed using RBC lysis buffer, washed twice with PBS containing 2% v/v FBS, and resuspended in complete RPMI 1640 medium supplemented with 10% v/v FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin. Cell viability and concentration were assessed by Trypan blue exclusion using a hemocytometer³⁹.

MultiScreen-IP plates (MSBVN1210, Millipore) were prewetted with 20 μL of 35% v/v ethanol, washed with water for injection (A1287301, Thermo Fisher Scientific), and coated with 25 μg/mL of recombinant antigen in PBS (100 μL/well) at 4 °C O/N. The following day, plates were washed with PBS and blocked with RPMI 1640 containing 10% v/v FBS at 37 °C for 1 h. After blocking, the single-cell suspension obtained as described in the previous section were seeded at 2 × 10⁵ cells per well and incubated at 37 °C with 5% CO₂ for 18–20 h to allow antigen-specific antibody secretion. After incubation, plates were washed thoroughly with PBS-T to remove cells and unbound antibodies. Biotinylated anti-mouse IgG (3825-6-250, Mabtech) detection antibody diluted in PBS with 0.5% v/v FBS was added to each well and incubated at RT for 2 h. Plates were washed again with PBS, followed by incubation with streptavidin-alkaline phosphatase (3825-2 A, Mabtech) diluted in PBS with 0.5% v/v FBS at RT for 1 h. After a final wash, spots were developed using 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium (BCIP/NBT, 3650-10, Mabtech) substrate solution until distinct spots appeared, typically within 10–20 min. The reaction was stopped by rinsing plates extensively with tap water, and plates were air-dried  O/N in the dark. Spots representing individual antibody-secreting B-cells were imaged using an automated ELISpot reader (ImmunoSpot) and quantified using ImageJ. Results were expressed as the number of spot-forming cells per 10⁶ bone marrow cells.

Cytokine analysis

A mouse LEGENDplex kit (741043 and 740621, BioLegend) was used to measure the concentration of pro-inflammatory and Th1/Th2/Th17-associated cytokines in lung homogenates following the manufacturer’s instructions. Data were collected using an Attune flow cytometer and processed with LEGENDplex™ online data analysis software (legendplex.qognit.com).

nsEMPEM

Serum samples were heat-inactivated at 55 °C for 1 h. For IgG isolation, 0.5 mL of mouse serum was mixed with 0.5 mL of pre-washed CaptureSelect IgG-Fc Affinity Matrix resin (Thermo Scientific) and 4 mL of PBS and incubated with gentle rotation at 4 °C for 48 h. The resin was washed three times with 5 mL of PBS, and the IgG was eluted using 2.5 mL of 0.1 M glycine (pH 2.0) into tubes containing 2 mL of 1 M Tris (pH 8) for immediate neutralization. For IgG digestion, papain was activated in 100 mM Tris-EDTA, 10 mM L-cysteine, and papain 1 mg/mL at 37 °C for 15 min. The activated papain was incubated with 1 mg of IgG at 37 °C for 5 h. The reaction was quenched with 0.05 M iodoacetamide. To purify Fabs, the digested mixture was incubated with CaptureSelect IgG-Fc affinity matrix rotating at 4 °C for 1 h. The unbound supernatant containing the Fab was collected, buffer-exchanged into TBS, and concentrated using 10 kDa Amicon ultra centrifugal filters.

Immune complexes were prepared in a molar proportion of ~1:10 (antigen:Fab), using 2 μg of antigen and 60 μg of polyclonal Fab incubated at RT O/N. The unbound Fab was washed using TBS in a 100 kDa Amicon Ultra 0.5 mL centrifugal filter. The nsEMPEM grids were prepared using 3 μl of Fab-antigen complexes in a concentration of 0.02 mg/mL and applied for 10 sec to 400 mesh Cu²⁺ grids that were carbon coated, and glow discharged at 15 mA for 25 sec. The Fab-antigen complex was negatively stained with 2% v/v uranyl formate for 1 min. Data were collected using a Talos F200C electron microscope with a Ceta 16 M camera at 200 kV and magnification of 73,000, pixel size of 2 Å/pixel. The defocus range was set between −2.5 and −2 mm and the electron dose was 27.1 e-/Å2. Automated data collection was carried out using EPU (Thermo Scientific).

After the automated data collection, Relion-4.0 was used for processing. Particles in the range of 190 to 280 Å were picked in the micrographs, and 2D classification was performed using a box size of 208 pixels. Particles that contained trimer/tetramer only or trimer/tetramer-Fab complexes were selected for 3D analysis. The 3D reference for 3D classifications and refinements was a low-resolution model of a non-liganded HA and NA. Particles were then classified into 10 classes, and classes with similar features were combined and refined.

Statistical analysis

Statistical analyses were performed in GraphPad Prism v.10.5 using the non-parametric Kruskal–Wallis test followed by Dunn’s post hoc pairwise comparisons. To adjust for multiple comparisons, the two-stage linear step-up false discovery rate (FDR) procedure of Benjamini, Krieger, and Yekutieli was applied, with the desired FDR (Q) set to 0.1. Significance was considered with p values ≤0.05(*), ≤0.01(**), ≤0.001 (***), ≤0.0001 (****).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

E.P.-M. and F.K. conceived and designed the study. E.P.-M., T.G.A., M.J.S., K.V., H.A., and A.J.R. generated laboratory data. E.P.-M., T.G.A., K.V., H.A., and J.H. analyzed the data. J.Y., D.B., J.D.C., D.Y., and R.J.W. provided adjuvants, proteins, and viruses. E.P.-M., Y.K., G.N., J.H., A.B.W., and F.K. supervised the research. E.P.-M. wrote the original manuscript draft. All authors critically reviewed and approved the final version of the paper for submission.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data associated with this work can be found in the article, supplementary materials, and Source Data file, and will be available via Immport. Negative stain electron microscopy maps representing each HA and NA epitope are deposited in the Electron Microscopy DataBank (EMDB) under accession IDs EMD-72062, EMD-72063, EMD-72064, EMD-72065, and EMD-72066. All other data associated with this work can be found in the article or in the supplementary materials. All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Reagents and antigens described in the manuscript can be provided by Florian Krammer’s laboratory pending scientific review and a completed material transfer agreement and any required shipping/handling permits for viruses. Requests for the reagents and antigens should be submitted to florian.krammer@mssm.edu. Source data are provided with this paper.

Competing interests

The Icahn School of Medicine at Mount Sinai has filed patent applications regarding influenza virus vaccines on which E.P.-M. and F.K. are listed as inventors. F.K. has consulted for Merck, GSK, Gritstone, Sanofi, Curevac, Seqirus, and Pfizer and is currently consulting for 3rd Rock Ventures and Avimex. The laboratory of F.K. is also collaborating with Dynavax on influenza vaccine development and with VIR on influenza virus therapeutics. A.B.W. has received royalty payments for the licensure of a prefusion coronavirus spike stabilization technology for which he is a co-inventor. A.B.W. and J.H. are currently consulting for Third Rock Ventures and Merida Biosciences. The laboratory of A.B.W. received unrelated sponsored research agreements from Third Rock Ventures during the conduct of the study. The authors declare that they have no other competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68457-6.