A versatile H5N1-VSV platform for safe influenza virus research applications

ABSTRACT

The H5N1 strain of influenza A virus (IAV) continues to cause severe infections in a range of avian and mammalian species, including sporadic but concerning cases in humans. There is growing concern that circulating H5N1 strains could lead to widespread human outbreaks. Research with highly pathogenic H5N1 viruses is restricted to Biosafety Level 3 (BSL-3) laboratories. Vesicular stomatitis virus (VSV)-based vaccine vectors expressing heterologous viral proteins from Ebola, SARS-CoV-2, Lassa virus, etc., have previously been shown to be safe and effective in animal models and human clinical trials. Here, we report the development of a recombinant VSV expressing the hemagglutinin (HA) and neuraminidase (NA) genes of H5N1 IAV (H5N1-VSV), which serves as a versatile platform to study various aspects of H5N1 IAV biology. H5N1-VSV replicated robustly to titers comparable to those of the full H5N1 virus in multiple cell lines. In mice, H5N1-VSV vaccination was safe, elicited strong immunity, and conferred protection against a circulating H5N1 strain. Notably, we found that polymorphisms in antigenic site Sa of circulating strains emerged under immune selection pressure in cattle, resembling the evolution of pandemic IAV in humans. These findings suggest that H5N1-VSV can serve as a safe, adaptable platform for influenza research.

LETTER

To construct a VSV vector expressing H5N1 HA and NA, we replaced the endogenous VSV G ORF with the HA ORF and inserted the NA gene as a separate transcriptional cassette flanked by gene start (GS) and gene end (GE) signals (Fig. 1A, HA/NA based on the A/Michigan/90/2024 H5N1 isolate) (1, 2). To alleviate any potential concerns about recombination or reassortment, we codon-deoptimized the HA and NA sequences and excluded all viral non-coding regions. The VSV backbone also included a GFP reporter as the first gene cassette following the leader RNA, allowing easy visualization of infection by microscopy. Recombinant H5N1-VSV was rescued by co-transfecting 293T cells (pre-infected with vaccinia virus expressing T7 polymerase) with the H5N1-VSV antigenome and helper plasmids encoding the VSV N, P, G, and L proteins (3). Rescue of recombinant virus was confirmed by immunofluorescence for HA and GFP expression (Fig. 1B). H5N1-VSV replicated efficiently in both human lung epithelial cells (A549) and Madin-Darby canine kidney (MDCK) cells, reaching titers of 10⁷–10⁸ PFU/mL (plaque-forming units: PFU; Fig. 1C) (4). Expression of HA and NA in infected MDCK cells was confirmed by Western blot (anti-HA2 antibody, NR-4539, BEI Resources; anti-NA antibody Reference (5)). In addition, HA₀ was properly cleaved into HA1 and HA2 at the multibasic site (Fig. 1D). As controls, lysates from MDCK cells infected with a low-pathogenic H5N1 strain (A/Texas/37/2024, H5N1/Tx24) were included. These results confirm the successful generation of recombinant H5N1-VSV.

Fig 1.

Generation of recombinant VSV expressing HA and NA from the 2024 H5N1. (A) Genome organization of VSV-GFP and H5N1-VSV. (B) Immunofluorescence of HA and GFP in A549 cells infected with H5N1-VSV. Cells were infected with indicated viruses (MOI = 1), and at 16 hours post infection (hpi), cells were fixed and subjected to immunofluorescence analysis. (C) Replication kinetics in A549 and MDCK cells. Cells were infected with H5N1-VSV (MOI = 0.01) and viral loads in the supernatants were measured. Data are shown as mean titer (PFU/mL) of triplicate samples ± SD. (D) Western blot for HA and NA in infected MDCK cells. Cells were infected with indicated virus (MOI = 1), and at 16 hpi, cell lysates were collected for Western blot analysis. Low-pathogenic H5N1 is included as control.

Next, we compared the replication kinetics of H5N1-VSV with the low-pathogenic H5N1/Tx24 strain in A549 and MDCK cells. H5N1-VSV replicated to similar levels as H5N1/Tx24 and VSV-GFP with viral titers exceeding 10⁸ PFU/mL (Fig. 2A and B). As a recent report indicated that a human isolate of bovine H5N1 (2024) strain shows reduced sensitivity to the neuraminidase inhibitor oseltamivir (Tamiflu), we assessed replication of different IAV strains in A549 cells in the presence of oseltamivir carboxylate (6). Older H5N1 (A/Vietnam/1203/2004; H5N1/VN04) and H7N7 (A/Netherlands/219/2003; H7N7/NL03) strains showed marked sensitivity at lower drug concentrations (Fig. 2C). In contrast, both H5N1-VSV and H5N1/Tx24 showed reduced sensitivity, with inhibition occurring only at higher concentrations. Together, these results support the use of H5N1-VSV as a surrogate model to study circulating H5N1 strains.

Fig 2.

Replication kinetics and oseltamivir sensitivity of H5N1-VSV. (A and B) Replication of H5N1-VSV, H5N1/Tx24, and VSV in A549 and MDCK cells. Cells were infected with indicated viruses (MOI = 0.01), and at various times post-infection, viral loads in the supernatants were measured. Data are shown as mean titer (PFU/mL) of triplicate samples ± SD. (C) Sensitivity of various IAV strains to oseltamivir carboxylate. A549 cells were infected with indicated viruses (MOI = 0.01) and incubated with varying concentrations of oseltamivir carboxylate. At 48 hpi, viral loads in the supernatants were measured. Data are shown as mean percentage inhibition relative to DMSO control of triplicate samples ± SD. Statistical significance was determined by one-way analysis of variance (ANOVA). ∗ denotes P < 0.05 or lower, and ns, non-significant.

As VSV-based Ebola vaccines have proven effective in humans, we next evaluated whether H5N1-VSV could protect mice from a low pathogenic H5N1/Tx24 challenge (7, 8). C57BL/6J mice were vaccinated intranasally or intramuscularly with H5N1-VSV and boosted 2 weeks later (doses of 1.25 × 10⁷ PFU). Mice vaccinated with VSV-GFP served as the control group. Four weeks after the initial vaccination, serum was collected for neutralization assays against H5N1-VSV and H5N1/Tx24. Mice vaccinated with H5N1-VSV via either route showed high-titer neutralizing antibodies in the serum as compared to VSV-GFP vaccinated mice (Fig. 3A and B). Upon lethal challenge with 10 LD₅₀ of H5N1/Tx24 (Dose = 2370 PFU), all of the H5N1-VSV vaccinated mice survived without displaying any weight loss or clinical signs of infection (Fig. 3C and D). This protection correlated with an undetectable virus in lung tissues at day 7 post-challenge (Fig. 3E). In contrast, the majority of VSV-GFP vaccinated mice and all of the mock-vaccinated mice succumbed to H5N1/Tx24 challenge and showed high lung viral loads. These results demonstrate that the H5N1-VSV vaccine is safe and elicits strong protective immunity against a circulating H5N1 strain.

Fig 3.

Protective efficacy of H5N1-VSV vaccination in mice. (A and B) Neutralization titers in the sera post-vaccination. Sera from vaccinated mice were treated with RDE-II prior to use in neutralization assays. (C through E) Protection in vaccinated mice following lethal H5N1 challenge: (C) weight loss (n = 4–5), (D) survival (n = 4–5), and (E) lung titers (n = 4–6). Statistical significance was determined by ANOVA. ∗ denotes P < 0.05, **P < 0.01, and ns, non-significant.

IAV evolution in humans is often associated with the accumulation of immune-driven mutations in major HA antigenic sites, including the acquisition of N-linked glycosylation sequon (N-X-S/T) (9, 10). We had previously shown that descendants of the 1918 H1N1 virus acquired N-glycosylation at antigenic site Sa upon circulation in humans (11). Interestingly, sequence analysis of circulating H5N1 strains revealed an A172T mutation in cattle isolates, which introduces a potential N-glycosylation site (Fig. 4A). To determine whether this A172T change is due to immune pressure, we passaged H5N1-VSV in MDCK cells in the presence of immune serum (1:100 dilution) from vaccinated mice (n = 5 independent replicates). At passage 5, we observed robust H5N1-VSV replication in all five replicates, suggesting the potential emergence of escape variants. Bulk sequencing of viral RNA from passage 5 supernatants revealed that four of five independent replicates acquired the A172T mutation (NDA→NDT), while one had an E142K mutation, both occurring within antigenic site Sa. This remarkable finding further validates the H5N1-VSV vector as a safe and effective tool for investigating antigenic drift in influenza viruses. Critically, it highlights that continued H5N1 circulation in cattle promotes this drift, increasing the risk of new strains emerging with an altered host range—a scenario with significant public health consequences. It should be noted that during passaging of H5N1-VSV, we observed the emergence of mutants with a truncated HA tail (~13 amino acid deletion), as recently reported (12). These findings strongly suggest that immune selection is a key driver of H5N1 IAV evolution within cattle. As human IAV strains initially escape preexisting immunity through antigenic drift at the Sa site, it is likely that H5N1 strains carrying the A172T mutation will evade immunity induced by vaccines based on ancestral H5 strains, underscoring the importance of considering variability at the Sa antigenic site when selecting future vaccine strains.

Fig 4.

Immune selection at antigenic site Sa in cattle H5N1 isolates. (A) HA sequence comparison of currently circulating H5 isolates and neutralization escape mutants. Hu—human, Ca—cattle, and Ms Ab—mouse polyclonal sera selected variant. GeneBank sequences of indicated strains are Hu[MI]: XBE32674.1, Ca[MI]: XFE99892.1, and Ca[TX]: XAJ06472.1. (B and C) Structural modeling of site Sa mutations on HA trimer. (B) Top view. At passage 5, A172T mutation was present in four out of five independent replicates, and E142K mutation was present in one replicate. (C) Side view.

In summary, our study demonstrates that the H5N1-VSV construct is a powerful and safe tool for studying contemporary H5N1 viruses. This platform allows for research outside of BSL-3 conditions and can support vaccine development, viral evolution studies, and antiviral screening.