Replication in human cells supersedes serum antigenic match for the in vivo efficacy of an A/H1N1pdm09 LAIV virus

Abstract

For FluMist®/Fluenz live attenuated influenza vaccine (LAIV), efficient virus replication in human nasal epithelial cells (hNEC) is an important factor in vaccine effectiveness (VE). The necessity of antigenic similarity of LAIV viruses to the WHO-recommended strains for VE is less clear. Here, we aimed to describe the relative importance of optimal hNEC replication and antigenic match in protection from wild-type (wt) challenge in vivo. The efficacy of an A/H1N1pdm09 LAIV strain (LAIV_H1.opt) with optimised hNEC replication but reduced antigenicity relative to its parent (LAIV_H1.par) was assessed. Delivered at a low dose, animals vaccinated with antigenically matched LAIV_H1.par shed wt challenge virus and exhibited influenza-like illness post-challenge. Conversely, antigenically divergent LAIV_H1.opt provided protection from both endpoints. LAIV_H1.opt was also found to be more immunogenic in ferrets than LAIV_H1.par, generating superior anti-HA and anti-NA antibody responses. We conclude that efficient virus replication can supersede antigenic match for high LAIV efficacy.

Introduction

The influenza virus is a respiratory pathogen of the Orthomyxoviridae family responsible for seasonal influenza epidemics in humans. It is estimated to cause 3–5 million cases of severe infection and 290,000-650,000 deaths per annum^1,2. Monitoring and limiting virus spread is crucial for reducing the impact and burden of influenza on global public health.

The primary strategy for combating seasonal influenza outbreaks is annual vaccination, which protects vulnerable individuals and reduces cases of severe infection³. As influenza viruses evolve, the haemagglutinin (HA) and neuraminidase (NA) surface proteins accumulate sequence changes, leading to antigenic drift^4,5. Such changes in circulating influenza strains can reduce the cross-reactivity of neutralising antibodies induced by previous vaccinations or infections. Thus, seasonal vaccines require annual reformulation to maintain immune responses capable of protecting against currently circulating viruses^6,7.

Any candidate vaccine virus (CVV) for a commercial vaccine formulation must demonstrate antigenic match to the named wild type (wt) WHO-recommended strain, as assessed by a haemagglutination inhibition (HAI) assay⁸. The inhibitory properties of post-infection ferret antisera are assessed against the wt WHO-recommended virus, and the parent virus from which the CVV was developed. To be classified as antigenically matched, the anti-CVV ferret antisera must inhibit agglutination by the reference wt strains and the homologous CVV itself with <4-fold difference in antibody titre^9,10. The HAI assay is the gold-standard means of determining the antigenicity of CVVs across all currently licensed influenza vaccine manufacturing platforms.

FluMist® (Fluenz®) is an intranasal, live-attenuated influenza vaccine (LAIV) produced in trivalent and quadrivalent formulations for protection against seasonal influenza virus subtypes. LAIV is manufactured in embryonated hens’ eggs, with final CVV selection determined by the same WHO antigenic-match criteria as described above¹¹. LAIV CVVs are developed using reverse genetics. HA and NA genes derived from circulating wt strains are reassorted with the PB2, PB1, PA, NP, M, and NS genes from the cold adapted (ca), temperature sensitive (ts) and attenuated (att) master-donor viruses A/Ann Arbor/6/1960 or B/Ann Arbor/1/1966^12,13. Following intranasal delivery, efficient replication of LAIV strains at the respiratory mucosa leads to a multifaceted immune response, with induction of innate, cell-mediated and humoral elements^14–16.

During the 2013–2014 and 2015–2016 influenza seasons, the A/H1N1pdm09 components of LAIV exhibited sub-optimal vaccine effectiveness (VE) in quadrivalent (QLAIV) formulation in the U.S.A¹⁷. This was despite both A/H1N1pdm09 LAIV viruses in use during that period being antigenically matched to the relevant WHO-recommended strains. Subsequent investigations attributed this reduced VE to a decrease in the replicative fitness of A/H1N1pdm09 LAIV strains in primary human nasal epithelial cells (hNEC)¹⁸. This led to less-fit strains suffering from inter-strain competition and reduced protection from wt infection in multivalent formulations^19,20.

To overcome the reduced hNEC replicative fitness of A/H1N1pdm09 LAIV CVVs, a strategy for HA-protein optimisation was developed. Amino acid changes N125D, D127E, D222G, and R223Q were incorporated into the HA proteins of A/H1N1pdm09 LAIV strains isolated in 2015 (A/Slovenia/2903/2015) and 2020 (A/Victoria/1/2020) using site-directed mutagenesis. The result in both cases was a > 100-fold increase in hNEC replication, whilst maintaining high yield in eggs, HA protein thermal stability, virus genetic stability and antigenic match to WHO-recommended strains^21,22. The improved replicative fitness in hNEC was attributed to an increase in human-like α2,6-linked sialic acid binding. These optimised strains produced VE estimates of between 50 and 90% in the UK and Denmark^23,24, confirming the importance of viral replicative fitness of A/H1N1pdm09 LAIV CVVs.

Whilst these applications of HA-optimisation improved hNEC replication without impacting the antigenicity of the A/Slovenia/2903/2015 and A/Victoria/1/2020 strains, introduction of amino acid substitutions close to receptor binding and antigenic sites could cause measurable changes in antigenicity, particularly in the context of ongoing HA protein evolution^25,26. However, while hNEC replication for H1N1 LAIV CVVs has been shown to be critical for VE, the importance of antigenic match to WHO-recommended strains for LAIV is less clear. As highlighted between 2013 and 2016, reduced A/H1N1pdm09 VE was observed despite LAIV viruses being antigenically matched to WHO-recommended strains. In addition, in historical randomised controlled clinical trials, LAIV gave significant efficacy despite circulating strains being antigenically distinct from the vaccine viruses used^27–29. One of these LAIV strains, the pre-2009 H1N1 A/New Caledonia/20/1999 (A/NC99)²⁷, has since been shown to replicate to high levels in hNEC¹⁸. This indicates that antigenic match alone may not be sufficient for high LAIV VE and that optimal hNEC replication might be able to compensate for a degree of antigenic difference.

Given this, in a scenario where optimisation of hNEC replication results in a measurable change in HA antigenicity, it is possible that the benefits of optimised hNEC replication might overcome any corresponding reduction in antigenic match to circulating strains. In this case, robust LAIV VE might be best achieved by prioritisation of hNEC replication over antigenic match during CVV development.

Here, we aimed to determine whether an LAIV CVV with optimised hNEC replication but reduced antigenic match to WHO-recommended strains would provide higher levels of protection from wt challenge than its antigenically matched parent. Using a recently described ferret model able to identify highly effective LAIV CVVs²⁰, the in vivo efficacy of an HA-optimised A/H1N1pdm09 LAIV strain with enhanced hNEC replication but reduced antigenic match was assessed, relative to its antigenically matched parent strain. Increased understanding of the critical parameters contributing to LAIV VE will be invaluable in the rational design of future LAIV strains, ensuring the continued inclusion of the most optimal and effective vaccine virus components.

Results

A/H1N1pdm09 LAIV strain with enhanced human cell replication shows reduced antigenic match to wild-type virus

To evaluate the relative contributions of LAIV replication and antigenicity to protection from wt challenge in a ferret model, the previously described approach to optimising the HA protein of A/H1N1pdm09 strains²¹ was applied to A/Norway/31694/2022 (A/NOR22). The optimised H1N1pdm09 LAIV strain (LAIV_H1.opt) carried amino acid substitutions at residues 127 (D127E), 222 (D222G) and 223 (R223Q) in the HA protein, relative to the egg-derived, A/NOR22 parental LAIV (LAIV_H1.par), which maintained the wt HA sequence (Table 1).

Table 1.

Optimised A/NOR22 LAIV variant was generated by introducing 127E, 222G and 223Q amino acid substitutions in the HA protein

Strain	127	222	223
A/NOR22 egg wildtype reference	D	D	R
LAIVH1.par	D	D	R
LAIVH1.opt	E	G	Q

HA amino acid residues at positions 127, 222, and 223 (H1 numbering) of A/Norway/31694/2022 (A/NOR22) LAIV variants included in this study. Parent LAIV (LAIV_H1.par) has an identical HA sequence to the A/NOR22 egg-derived wildtype reference strain. HA amino acid substitutions were introduced by site-directed mutagenesis at residues 127, 222 and 223 to generate the optimised LAIV virus (LAIV_H1.opt).

To quantify human cell replication, both A/NOR22 variants were used to infect hNEC cultures at a low MOI of 0.01 FFU/well. Viral titres were determined daily over a four-day time course by TCID₅₀ assay. The pre-2009 A/H1N1 LAIV strain, A/NC99, was used as a positive control due to its known high levels of hNEC replication¹⁸.

LAIV_H1.opt replicated to high titres under these conditions in hNEC (Fig. 1a), resembling the control A/NC99 growth curve and peaking at a titre of 6.7 Log₁₀ TCID₅₀/ml on day 4 post-infection, compared to 7.2 Log₁₀ TCID₅₀/ml for A/NC99, also at day 4 post-infection. In contrast, LAIV_H1.par did not replicate efficiently, with viral titres between 2.0 and 2.2 Log₁₀ TCID₅₀/ml across the four-day time course. Geometric mean titre per day was calculated for each virus, to describe virus replication over time as previously reported¹⁸, to aid in statistical comparison (Fig. 1b). By this measure, LAIV_H1.opt (5.6 Log₁₀ TCID₅₀/ml/day) was not statistically different from A/NC99 (6.3 Log₁₀ TCID₅₀/ml/day), while LAIV_H1.par replication was significantly lower (2.0 Log₁₀ TCID₅₀/ml/day, p < 0.01). This indicated that HA optimisation had improved A/NOR22 hNEC titres over 3000-fold in this model. Moreover, LAIV_H1.opt replicated to similar titres in eggs as LAIV_H1.par (Supplementary Fig. 1).

Fig. 1. LAIV_H1.opt has enhanced human cell replication in hNEC model but reduced antigenic match to wildtype virus using HAI assays.

a Four-day time-course for infection in human nasal epithelial cells (hNECs) for LAIV_H1.par and LAIV_H1.opt, at an MOI of 0.01 FFU/cell. Data represent three independent experiments of two hNEC transwells per experiment (total six transwells per virus), measured by TCID₅₀. The pre-2009 H1N1 LAIV strain, A/NC99, was included as a positive control for high hNEC replication. b Average virus titre per day calculated from (a), for A/NC99, LAIV_H1.par and LAIV_H1.opt. Columns show GMT of individual experiments (symbols) and error bars show standard deviation. Groups were compared with one-way ANOVA and all groups compared against each other with Tukey–Kramer HSD post-test **p < 0.01. c Antigenic analyses of LAIV_H1.par and LAIV_H1.opt viruses against the A/NOR22 egg wildtype (e-wt) and A/VIC22 WHO wildtype reference strains (WHO-wt). Numbers indicate HAI titres. Underlined values indicate homologous titres. Results provided by the Worldwide Influenza Centre, The Francis Crick Institute, UK. d Ferret serum HAI titres generated for LAIV_H1.par (n = 2) and LAIV_H1.opt (n = 3) against the relevant homologous LAIV strain, the e-wt and the WHO-wt. Antisera were produced using a high LAIV dose, reflective of the routine seasonal CVV selection process. Dotted line indicates lower limit of detection LoD. e Fold-difference in HAI titres for LAIV_H1.par and LAIV_H1.opt relative to A/NOR22 e-wt and WHO-wt, A/VIC22. Grey line indicates WHO threshold for antigenic match (<4-fold). Columns in all cases show GMT of individual ferrets, distinguished by unique symbols. Lines show geometric standard deviation. A/NC99: A/New Caledonia/20/1999; A/NOR22: A/Norway/31694/2022; A/VIC22: A/Victoria/4897/2022.

For antigenic characterisation of LAIV_H1.par and LAIV_H1.opt, groups of two (LAIV_H1.par) or three (LAIV_H1.opt) ferrets were intranasally vaccinated with a dose of approximately 6–7 Log₁₀ FFU/ferret and antisera were harvested at day 21 post-vaccination. Testing of antisera was performed at the Worldwide Influenza Centre, WHO Collaborating Centre for Influenza Virus Reference and Research, The Francis Crick Institute, by HAI assay (Fig. 1c). HAI titres were produced against the egg-derived parental wt virus, A/NOR22 egg-wt (e-wt); the WHO recommended, egg-based wt reference strain, A/Victoria/4897/2022 (WHO-wt); and the homologous vaccine virus delivered to that animal. Using the raw data generated by the Crick Institute, geometric mean HAI titre (GMT) and standard deviation (SD) was calculated against the vaccine virus and both wt strains (Fig. 1d). Fold-differences between GMT were used to establish antisera cross-reactivity (Fig. 1e).

LAIV_H1.par and LAIV_H1.opt both induced measurable serum immune responses in ferrets, as determined by HAI (Fig. 1d), with LAIV_H1.opt producing a 3-fold higher response than LAIV_H1.par (11.3 vs 8.3 Log₂ HAI). However, against e-wt, LAIV_H1.par and LAIV_H1.opt GMTs were comparable (7.3 Log₂ HAI vs 7.6 Log₂ HAI; SD, 0.6). This was also true against the WHO-wt (7.8 Log₂ HAI; SD, 0.7 vs 6.6 Log₂ HAI; SD, 0.6). As the GMTs were ≤2-fold different between LAIV_H1.par and both e-wt and WHO-wt (Fig. 1e), it was confirmed to be antigenically matched. Conversely, LAIV_H1.opt antisera had an approximately 13-fold difference in cross-reactivity against e-wt virus, and 26-fold against the WHO-wt, relative to LAIV_H1.opt itself.

In summary, due to incorporation of HA protein residue changes D127E, D222G, and R223Q, LAIV_H1.opt possessed significantly improved human replicative fitness, relative to LAIV_H1.par, but reduced antigenic match to both its parent e-wt and the reference WHO-wt.

A/NOR22 monovalent LAIV_H1.opt sheds to higher vaccine virus titres in ferrets than LAIV_H1.par

With confirmation that LAIV_H1.par and LAIV_H1.opt had contrasting human cell replication and antigenicity characteristics, a ferret study was designed to evaluate whether LAIV_H1.opt, due to its improved human cell replication, could provide increased immunogenicity or protection from wt influenza virus infection, despite its reduced antigenic match to wt reference strains.

Based on our recently published clinically translatable A/H1N1pdm09 LAIV ferret efficacy model²⁰, groups of four male/female ferrets (musetla putorius furo) aged over 5 months were intranasally vaccinated with the LAIV or Mock vaccine formulations (Fig. 2). Monovalent LAIV_H1.par (MP-4) and LAIV_H1.opt (MO-4), delivered at a ferret-optimised dose of 4.0 Log₁₀ FFU/ferret, would examine the ability of each strain to induce an immune response and protect from wt challenge in isolation, while quadrivalent (QLAIV) formulations (QP-4, QO-4) would assess immunogenicity and protection in the context of inter-strain competition, as previously described¹⁹.

Fig. 2. Study design for vaccine efficacy assessment of LAIV_H1.par and LAIV_H1.opt in ferrets.

a Timeline of in vivo study. Animals were intranasally vaccinated at Study Day 0 with monovalent or quadrivalent LAIV formulations. All groups were challenged with wt virus at day 28 post-vaccination, and study termination occurred at day 33. Vertical grey arrows highlight the major interventions during the study time period. Red boxes indicate blood collection. Blue boxes indicate nasal wash sample collection. Horizontal blue arrow indicates ferret core body temperature readings, taken hourly during the study time period. Horizontal yellow arrow indicates ferret bodyweight observations, taken daily during the study time-period. b Table outlining inocula formulations and challenge virus used for assessment of LAIV_H1.par and LAIV_H1.opt efficacy in ferrets. Rows detail vaccine composition. Ferrets were vaccinated with a defined dose (4.0 Log₁₀ FFU/ferret) of either LAIV_H1.par (shaded yellow) or LAIV_H1.opt (shaded purple). Vaccine groups included Monovalent LAIV_H1.par (MP-4) and LAIV_H1.opt (MO-4) LAIV, along with quadrivalent formulations for each LAIV strain (QP-4, QO-4), to assess any impact of inter-strain competition. Both an unvaccinated (Mock) and high-dose (7.0 Log₁₀ FFU/ferret) LAIV_H1.par (MP-7) group were included as controls (shaded white). Groups were challenged with cell-derived A/NOR22 wildtype virus (c-wt) at 5.0 Log₁₀ FFU/ferret (shaded grey)¹⁹.

As well as an unvaccinated control group (Mock), a ferret group was dosed with monovalent LAIV_H1.par at 7.0 Log₁₀ FFU/ferret (MP-7) to mimic conditions used for production of post-infection ferret antisera during CVV development³⁰, as seen in Fig. 1. All ferrets were challenged with A/NOR22 cell-wildtype (c-wt) virus at 5.0 Log₁₀ FFU/ferret at day 28 post-vaccination (Fig. 2). This was used in place of the parental e-wt, as Schewe et al. showed that egg-derived A/H1N1pdm09 wt viruses can be non-pathogenic in ferrets due to a Q223R egg-adaptation in the HA protein²⁰. A/NOR22 e-wt contained this Q223R egg-adaptation and so c-wt A/NOR22 was used as the challenge agent, ensuring induction of a symptomatic influenza infection. Nasal washes were taken daily for 5 days post-vaccination and post-challenge to measure LAIV or wt virus shedding, respectively. Serum bleeds were taken at day 14 and day 21, with data primarily generated from day 21 samples as a representative timepoint, based on previous results¹⁹. Fever and body weight were monitored throughout the study to evaluate the development of influenza-like illness (ILI). Previously, it was shown that measurement of wt shedding and fever in this model could broadly reproduce A/H1N1pdm09 LAIV clinical VE data²⁰. As such, protection from these endpoints was the primary readout of efficacy in this study. Additional clinical signs of infection were also observed for up to 5 days post-challenge.

Following vaccination, to determine whether the enhanced hNEC replication of LAIV_H1.opt would translate to improved replication in the ferret upper-respiratory tract, shedding of the monovalent LAIV viruses post-vaccination was assessed. Monovalent viral titres in nasal washes collected daily for 5 days post-vaccination were measured using TCID₅₀ assay. LAIV_H1.par shedding was not detectable in MP-4 ferrets, while MO-4 ferrets produced consistent, detectable virus shedding across multiple days post-vaccination, with a peak mean viral titre of 2.5 Log₁₀ TCID₅₀/ml at day 3 post-vaccination (Fig. 3a). Interestingly, even LAIV_H1.par delivered at a 1000-fold increased dose (MP-7) did not produce detectable virus shedding. Only LAIV_H1.opt (MO-4 group) demonstrated average geometric mean shedding per day above the assay’s limit of detection (1.8 TCID₅₀/ml/day, Fig. 3b). These data confirmed that LAIV_H1.opt replicated more efficiently than LAIV_H1.par in the ferret nasal-epithelial environment, corroborating observations from the hNEC infection assay in vitro.

Fig. 3. Increased virus shedding was detected by TCID₅₀ assay for LAIV_H1.opt in vivo compared to LAIV_H1.par.

Animals were vaccinated at a dose of 4.0 Log₁₀ FFU/ferret with monovalent LAIV_H1.par (MP-4), LAIV_H1.opt (MO-4), or with a high dose (7.0 Log₁₀ FFU/ferret) of monovalent LAIV_H1.par (MP-7) control. Nasal washes were taken at days 1–5 post-vaccination and virus shedding was determined by TCID₅₀ assay in MDCK cells. a Virus titre measurements for each time point generated from four ferrets (symbols) per vaccination group. b LAIV virus geometric mean shedding per ml per day calculated from data in (a). Individual ferrets are indicated by unique symbols. Columns show group geometric mean with error bars representing geometric standard deviation. Horizontal dotted lines indicate the assay’s LoD. Statistical comparisons were not applied due to two of three vaccination groups generating data at the LoD. LoD limit of detection.

LAIV_H1.opt produces higher serum antibody titres than LAIV_H1.par but with reduced cross-reactivity to wt viruses

To confirm whether differences in vaccine replication properties impacted the ferret humoral immune response, the immunogenicity and antigenicity of LAIV_H1.par and LAIV_H1.opt variants were measured. Magnitudes of anti-HA serum antibody responses were assessed from bleeds taken at day 21 post-vaccination using two assays: HAI (Fig. 4a–d) and Microneutralisation (MN, Fig. 4e–h). Vaccine immunogenicity was measured against the homologous A/NOR22 LAIV strains and antisera cross-reactivity was assessed against both the parental e-wt and the c-wt challenge virus.

Fig. 4. LAIV_H1.opt produces higher magnitude of serum antibody responses than LAIV_H1.par but shows reduced HA cross-reactivity to egg and cell wildtype viruses.

Serum immune responses were measured post-vaccination at day 21 through HAI, MN and NAI ELLA. For HAI and MN, titres were measured against LAIV_H1.par and LAIV_H1.opt, along with e-wt and c-wt viruses. For NAI ELLA, groups were assessed against a recombinant H2N1_NorNA virus. HAI (a–d). a HAI GMT for monovalent LAIV_H1.par and LAIV_H1.opt antisera and (b) fold-difference in HAI titres in (a) between homologous LAIV strains and wt viruses. Grey line and arrow represent WHO threshold for antigenic match with HAI assay (<4-fold difference vs homologous virus). Mock group HAI titres were not applicable (N/A) in HAI fold-difference calculations. c HAI GMT for quadrivalent LAIV_H1.par (QP-4) and LAIV_H1.opt (QO-4) antisera and (d) fold-difference in quadrivalent HAI titres between homologous LAIV strains and wt viruses. MN (e–h). e Neutralisation GMT for monovalent LAIV_H1.par and LAIV_H1.opt antisera and (f) fold-difference in monovalent MN titres between homologous LAIV strains and wt viruses. g Neutralisation GMT for quadrivalent LAIV_H1.par and LAIV_H1.opt antisera and (h) fold-difference between homologous LAIV strains and wt viruses. Cross-reactivity against virus strains where antisera failed to neutralise classified as ‘Did Not Neutralise’ in MN fold-difference calculations. i NAI GMT for LAIV_H1.par and LAIV_H1.opt in monovalent and quadrivalent formulation. Groups were assessed for equal variance by Levene’s test and intra-group normality using the Shapiro-Wilk test. The geometric mean NAI₅₀ of each group was compared to the geometric mean of the Mock vaccination control using the Welch t-test with the Holm-Šídák correction for multiple comparisons. ****p < 0.0001. Bars show GMT from individual animals (unique symbols) with SD. Dotted horizontal line indicates LoD. Values that failed to elicit an antibody response were plotted as 0.5 x lower LoD. Values that exceeded the upper LoD of assays denoted with #. HAI, MN and NAI GMT for high dose (7.0 Log₁₀ FFU/ferret) monovalent LAIV_H1.par (MP-7) and mock-vaccinated ferrets included as controls. HAI: hemagglutination inhibition; MN: microneutralisation; NAI: Neuraminidase inhibition; ELLA: Enzyme-linked Lectin Assay; LoD: limit of detection; GMT: geometric mean titres; SD: Geometric standard deviation.

LAIV_H1.par vaccinated MP-4 ferrets gave generally low HAI GMT, with those against both homologous LAIV_H1.par (3.6 Log₂ HAI; SD, 1.5) and c-wt challenge virus (<3 Log₂ HAI) approaching or below the limit of detection of the assay. In contrast, LAIV_H1.opt vaccinated MO-4 animals produced the highest HAI antibody GMTs recorded across all groups (Fig. 4a); against homologous LAIV_H1.opt (≥12 Log₂ HAI), e-wt (10.5 Log₂ HAI; SD; 0.6) and c-wt (10.25 Log₂ HAI; SD; 1.0). LAIV_H1.opt homologous GMTs were saturated, meaning that true GMTs may have been underestimated in these data. The MP-7 high dose control group confirmed that a greatly increased dose of LAIV_H1.par in this study produced GMTs against LAIV_H1.par (10.0 Log₂ HAI; SD; 1.4) and both wt viruses (e-wt; 9.5 Log₂ HAI; SD, 1.3, c-wt; 8.8, Log₂ HAI, SD; 1.7) that resembled those seen during initial strain characterisation (Fig. 1). Overall, these HAI results suggested that a low dose of LAIV_H1.opt generated a robust antibody response, whereas LAIV_H1.par at the same dose elicited minimal antibody responses.

While values for the cross-reactivity of MP-4 antisera against e-wt and c-wt viruses were calculated (Fig. 4b), the low and variable titres measured make these fold-difference values unreliable. The high dose LAIV_H1.par control group (MP-7), which induced higher GMTs, remained antigenically matched to both e-wt (1.5-fold) and c-wt (2.75-fold) viruses, despite two ferrets reaching the WHO antigenicity threshold (4-fold difference) against the c-wt. MO-4 antisera were less cross-reactive with e-wt (2–4-fold) and c-wt (2–8-fold) viruses (Fig. 4b). However, due to the saturated homologous HAI titres for LAIV_H1.opt, the fold differences calculated relative to wt strains were likely to have been underestimated.

In quadrivalent formulation (Fig. 4c), LAIV_H1.opt (QO-4) again produced markedly higher HAI GMTs than LAIV_H1.par (QP-4), with both groups comparable to their monovalent equivalents (MP-4 and MO-4, Fig. 4a). Antigenic similarity was also comparable, with LAIV_H1.opt giving fold-differences of 4–8-fold (e-wt) and 4–16-fold (c-wt). Again, these values were likely to be underestimated due to saturated homologous QO-4 GMTs against LAIV_H1.opt.

To give a clearer picture of the magnitude and cross-reactivity of antibodies raised by LAIV_H1.par and LAIV_H1.opt, functional antibodies induced by both LAIV variants were assessed by MN, with half-maximal inhibitory-dilution values (MN₅₀) measured against the relevant homologous LAIV strain and both wt strains (Fig. 4e–h).

In contrast to the HAI data, MP-4 antisera did not detectably neutralise any test virus and gave responses comparable to the Mock vaccinated group (Fig. 4e). LAIV_H1.par antisera from the MP-7 control group produced detectable but variable neutralising antibody GMTs against the homologous LAIV_H1.par virus (10.0 Log₂ MN₅₀; SD, 1.5), the e-wt (9.5 Log₂ MN₅₀; SD, 2.0) and c-wt (8.6 Log₂ MN₅₀; SD, 1.8). MO-4 antisera had markedly higher GMTs than MP-4 against LAIV_H1.opt (14.4 Log₂ MN₅₀; SD, 0.4) as well as e-wt (8.6 Log₂ MN₅₀; SD, 0.7) and c-wt (9.2 Log₂ MN₅₀; SD, 1.0). MN₅₀ neutralisation curves with detectable GMT are shown in Supplementary Fig. 2.

Cross-reactivity could not be assessed for LAIV_H1.par based on MP-4 antisera, due to the absence of measurable neutralisation. When measured by MN, MO-4 antisera exhibited clearly reduced neutralisation capacity against both e-wt (61.4-fold difference) and c-wt (43.3-fold difference), relative to the homologous LAIV_H1.opt virus (Fig. 4f). These cross-reactivity differences were considerably higher than those observed in the HAI assay and were above the 4-fold threshold commonly applied to antigenic comparisons. In contrast, LAIV_H1.par MP-7 antisera were similarly cross-reactive against LAIV_H1.par and both e-wt (1.5-fold) and c-wt (2.8-fold) viruses, with fold-differences of <4-fold in both cases, as seen for HAI data. Finally, QP-4 and QO-4 MN data corroborated monovalent vaccination observations (Fig. 4g). QP-4 antisera did not neutralise LAIV or wt strains, while QO-4 antisera gave elevated MN titres relative to QP-4 but with reduced cross-reactivity against e-wt (58.1-fold different) and c-wt (25.1-fold different) viruses when compared to the homologous LAIV_H1.opt strain.

Taken together, these data indicated that LAIV_H1.opt was considerably more immunogenic than LAIV_H1.par in both monovalent and quadrivalent formulations, with a 4 Log₁₀ dose of LAIV_H1.opt generally inducing higher levels of serum antibodies than a 7 Log₁₀ dose of LAIV_H1.par. In addition, LAIV_H1.opt in this study was confirmed to be antigenically distinct from the parental e-wt virus, as observed during initial CVV characterisation (Fig. 1), as well as the c-wt virus used for ferret challenge. These antigenic differences appeared most pronounced when measuring functional, neutralising antibody responses by MN assay, as compared to HAI.

Higher titres of A/NOR22 NA-specific antibodies were detected in ferrets vaccinated with LAIV_H1.opt compared to LAIV_H1.par

Alongside analysis of anti-HA antibody responses by HAI and MN assays, alternative, non-HA-mediated mechanisms of protection were also investigated. NA-inhibiting (NAI) antibody titres have been demonstrated to correlate with protection against flu infection^31,32. All LAIV and wt viruses in this study shared the same NA sequence, so it was hypothesised that more efficient induction of anti-NA antibodies by LAIV_H1.opt could also contribute to protection from the ensuing challenge.

Assessment of serum anti-NA antibody responses in the same day 21 post-vaccination antisera as for HAI and MN was performed using an NAI ELLA. The source of NA protein used in the assay was a recombinant LAIV virus, A/H2N1_NorNA, carrying the A/NOR22 NA protein sequence common to all LAIV and wt viruses in the study, combined with the antigenically distinct H2 HA protein of A/Ann Arbor/6/60 to isolate anti-NA antibody responses. In brief, the ability of serially diluted ferret antisera to inhibit NA-induced cleavage of fetuin, expressed as a percentage of virus-only positive control wells, was measured. NAI titres were calculated as the reciprocal of the dilution required to produce 50% inhibition (here referred to as NAI₅₀) using a three-parameter logistic curve. Example raw data curves used to calculate the NAI₅₀ values are shown in Supplementary Fig. 3. The magnitude of NAI titres for all LAIV_H1.par and LAIV_H1.opt study groups were then compared to those for Mock vaccinated animals (Fig. 4i).

Antisera from mock-vaccinated control animals showed a low level of NA inhibition, with a mean NAI titre of 5.0 Log₂ NAI₅₀ (SD, 0.7), indicating the presence of background levels of non-specific NA-inhibitors.

The LAIV_H1.par vaccinated MP-4 group gave detectable but variable NAI responses between individual ferrets (GMT 7.3 Log₂ NAI₅₀; SD, 2.9) that were not statistically different from Mock animals. By comparison, the LAIV_H1.opt vaccinated MO-4 group showed significantly increased NA inhibition compared to the Mock group (p < 0.0001). The mean MO-4 NAI titre of 14.5 Log₂ NAI₅₀ (SD, 1.2) was >2-fold higher than that of MP-4 ferrets.

The LAIV_H1.par quadrivalent group QP-4 also failed to display any significant NAI antibody response, with an NAI titre comparable to that of mock-vaccinated ferrets. In contrast, LAIV_H1.opt continued to show induction of NAI antibodies in the QO-4 group, with a significantly increased NAI titre compared to the Mock group of 13.5 Log₂ NAI₅₀ (SD, 0.8; p < 0.0001). This showed that, unlike LAIV_H1.par, the LAIV_H1.opt virus delivered at a low, ferret-optimised dose, induced a substantial NA-specific immune response even in quadrivalent formulation.

Similarly to the results seen for HAI and MN, the LAIV_H1. The MP-7 control group induced a significantly higher NA-specific immune response than Mock animals, with a mean NAI titre of 12.3 Log₂ NAI₅₀ (SD, 0.7; p < 0.0001). This confirmed that when administered at a sufficiently high dose, LAIV_H1.par was also able to induce a robust anti-NA serum immune response. Even so, this was approximately 4-fold lower than the titres seen for MO-4 ferrets.

In summary, the LAIV_H1.opt virus induced higher NAI titres than LAIV_H1.par in both monovalent and quadrivalent formulations as measured by ELLA, with LAIV_H1.par only able to induce robust NAI antibody responses at a greatly increased dose.

LAIV_H1.opt provided improved protection against challenge virus shedding and influenza-like illness

Following confirmation of differences in vaccine shedding and serum antibody responses from ferrets vaccinated with the LAIV_H1.par and LAIV_H1.opt, protection from c-wt challenge was assessed (Fig. 5).

Fig. 5. LAIV_H1.opt provided superior protection from challenge virus shedding and influenza-like illnesses than LAIV_H1.par.

Following vaccination, groups were challenged with A/NOR22 c-wt virus 28 days post-vaccination. Nasal washes were taken at days 1–5 post-challenge and viral titres measured by TCID₅₀ assay. a Virus titre measurements from each time point post-challenge from four ferrets (symbols) per vaccination group. b Geometric mean wt shedding per day of virus titre recorded over 5 days. For statistical analysis the MO-4 vaccination group excluded from group analysis as it did not measure values above LoD. The remaining groups were compared to the Mock control by Welch t test with Holm-Šídák adjustment for multiple comparisons. MO-4 was compared to the Mock control with a one-sample t test, using LoD values. **p < 0.01. c Influenza-like illness in ferrets was monitored through measurement of Fever, as previously described¹⁹. Fever is calculated as the average increase from the pre-challenge baseline body temperature during a defined ‘fever period’, post-challenge. d Changes in animal bodyweight were also assessed. Weight changes are expressed as differences relative to a ferret’s average pre-challenge weight. Columns in all cases represent mean value, with error bars representing standard deviation. Each treatment group was assessed for equal variance using Levene test and intra-group normality using Shapiro–Wilk test. Fever and bodyweight changes for each treatment group were compared to the Mock control using Welch t test with Holm–Sidak’s post-test correcting for multiple comparison. *p < 0.05, **p < 0.01, ***p < 0.001. c-wt: Cell-derived wildtype; LoD: Limit of Detection.

Unprotected mock-vaccinated animals shed c-wt virus detectably for 5 days post-challenge, with shedding peaking at approximately day 2 (Fig. 5a). MP-4 and QP-4 groups, vaccinated with LAIV_H1.par, gave c-wt shedding profiles similar to Mock. In contrast, LAIV_H1.opt vaccinated groups MO-4 and QO-4 produced almost no detectable c-wt shedding post-challenge. Statistical comparison of geometric mean shedding per day confirmed these differences (Fig. 5b), with MP-4 (2.8 Log₁₀TCID₅₀/ml/day) and QP-4 (2.1 Log₁₀TCID₅₀/ml/day) groups indistinct from Mock (2.6 Log₁₀ TCID₅₀/ml/day). Ferrets from the MO-4 group with titres at the LoD gave significantly reduced c-wt shedding (p < 0.01) compared to the Mock. Reduced c-wt geometric mean shedding was also observed for QO-4 (1.2 Log₁₀ TCID₅₀/ml/day), but this change was not statistically significant.

To assess fever development, changes in core temperature were monitored hourly up to 5 days post-challenge, using intraperitoneal data loggers. Full temperature traces for all animals are shown in Supplementary Fig. 4. From these data, a value for ‘Fever’ was calculated for each animal. As previously described^19,20, Fever was calculated by first defining a ‘fever period’: a window post-challenge during which Mock-vaccinated animals experienced temperatures elevated by >1.5 standard deviations vs pre-challenge baseline. For all study animals, the average temperature difference vs pre-challenge baseline was then calculated during this fever period. An average Fever of + 1.3 °C was observed in the Mock group post-challenge (Fig. 5c), in keeping with prior A/H1N1pdm09 challenge viruses^21,22. The Fever responses measured for both LAIV_H1.par groups, MP-4 (+1.0 °C) and QP-4 (+1.2 °C), were comparable to the Mock group. Conversely, ferrets vaccinated with LAIV_H1.opt, in groups MO-4 and QO-4, both gave average Fever values of −0.1 °C, which was significantly reduced relative to Mock animals (p < 0.01).

In alignment with temperature data, no loss of bodyweight was measured for LAIV_H1.opt MO-4 (+0.8 g/day) and QO-4 (+2.3 g/day) groups (Fig. 5d). This was significantly different from the average bodyweight loss of −24.1 g/day for the Mock group (MO-4; p < 0.05, QO-4; p < 0.01). In contrast, the loss of bodyweight observed from both LAIV_H1.par groups, MP-4 (−22.2 g/day) and QP-4 (−17.6 g/day), was comparable to Mock-vaccinated animals. Full bodyweight data are shown in Supplementary Fig. 5.

For the MP-7 control group, a limited degree of detectable c-wt virus shedding was observed (1.35 Log₁₀ TCID₅₀/ml/day), along with a minor temperature increase post-challenge (Fever + 0.2 °C, p < 0.05) that was still significantly reduced relative to Mock. Similarly, MP-7 animals were significantly protected from weight loss (−3.2 g/day, p < 0.05), albeit with a greater loss of weight than LAIV_H1.opt vaccinated animals.

In summary, at a 4 Log₁₀ FFU ferret-optimised dose, ferrets vaccinated with both monovalent and quadrivalent LAIV_H1.opt formulations received significant protection from ILI. LAIV_H1.par did not provide significant protection from challenge at an equivalent dose despite being antigenically matched to wt reference strains. However, a 1000-fold increase in dose was able to overcome the limited protection conferred by LAIV_H1.par.

Discussion

Annual seasonal influenza vaccine development is predicated on the production of CVVs with antigenic match to wt WHO-recommended named viruses⁷. This is driven by the understanding that serum antibodies against the HA protein are a correlate of protection against influenza virus infection^33–35. CVVs that fail to meet the defined antigenic match criteria - where ferret serum HAI titres must be <4-fold different between the CVV itself, the parent wt virus it was derived from, and the wt WHO-recommended strain—are not eligible for inclusion in seasonal vaccine formulations. While serum antibodies against the HA protein are likely to mediate protection for systemically delivered, HA-based vaccines such as inactivated influenza vaccine (IIV), historical clinical data suggest that serum antibody responses do not show the same correlation with efficacy for LAIV^33,36. This is supported by instances of LAIV showing efficacy in seasons where circulating influenza viruses were thought to have undergone antigenic drift^27,28,37.

More recently, in attempting to differentiate LAIV from IIV, a necessity for efficient LAIV virus replication in human nasal epithelial cells was described, with the ability of A/H1N1pdm09 LAIV strains to replicate in hNEC correlating with both clinical VE and efficacy in ferrets^18,19. Rational design approaches have since been successfully applied to the optimisation of hNEC replication for higher VE^22,38. In these instances, enhanced hNEC replication was achieved whilst also maintaining the antigenic characteristics of the CVVs in question^21,22. However, when combined with ongoing HA protein evolution, the use of a standardised set of amino acid changes to produce this effect might be expected to result in sporadic changes to antigenic properties. This could result in the failure of an otherwise optimised LAIV strain to meet the prescribed antigenic match criteria.

Given the lack of a clear correlate of protection for LAIV and the necessity of efficient virus replication in hNEC for high levels of A/H1N1pdm09 VE, it was hypothesised that the key attributes required for an effective LAIV CVV may be distinct from those of systemically delivered, HA-based vaccines. Here, we aimed to determine whether an LAIV CVV with optimised hNEC replication but reduced antigenic match might be able to provide higher levels of efficacy in vivo than its antigenically matched parent.

To address this, the A/H1N1pdm09 LAIV strains LAIV_H1.par and LAIV_H1.opt, based on the wt strain A/NOR22, were compared for their ability to protect from wt challenge in both monovalent and quadrivalent formulations. LAIV_H1.opt was shown to possess significantly enhanced hNEC replication but reduced antigenic match to wt WHO reference strains, relative to LAIV_H1.par. LAIV_H1.opt failed to meet the <4-fold antigenic match threshold, demonstrating 8–32-fold reduced HAI titres against the parental e-wt and the WHO-wt.

Previously, we described a ferret efficacy model for LAIV that was able to reproduce clinical VE data for A/H1N1pdm09 strains²⁰. This relied on the use of a low, ferret-optimised dose of LAIV and the primary protection endpoints of fever generation and wt virus shedding post-challenge. Applying this model here, LAIV_H1.opt vaccination provided complete protection from both endpoints, with no detectable fever or wt shedding post-challenge. Additional measures of ILI taken to supplement the fever data, including weight loss and activity scores, corroborated these findings (see Supplementary Fig. 6). This was true in both MLAIV and QLAIV formulations, indicating that LAIV_H1.opt replication was sufficiently robust to effectively compete in a multivalent vaccine, avoiding the potential reductions in efficacy due to competition described for historical strains^19,20. In addition, improved LAIV_H1.opt replication did not significantly reduce the immunogenicity of the H3N2 or influenza B vaccine strains in the QLAIV formulation (see Supplementary Fig. 7).

Conversely, LAIV_H1.par, which met the regulatory serum-HAI antigenic match requirements for inclusion in seasonal influenza vaccines, provided minimal protection from wt challenge. However, it should be stressed that this observation does not equate to an expectation of ‘no protection’ for LAIV_H1.par in a real-world setting. One significant caveat to the use of the described ferret model is that, while it is valuable for identification of strains with the potential for high effectiveness, it is a stringent model, and its ability to differentiate strains with more moderate VE is less clear²⁰. This also applies to the experiments conducted in hNEC, where a low (0.01) MOI was used to provide the greatest differentiation between strains. Recent real-world VE data support this, confirming that A/NOR22 LAIV_H1.par produced significant, moderate VE of 50% and 42% in the UK in the 2023–24 and 2024–25 seasons, respectively^39,40. The continued accumulation of such paired data in vivo and clinically can be used to continually refine the predictive capacity of the ferret model.

The observation that LAIV_H1.par was able to provide significant VE in the clinic raises exciting possibilities for the real-world potential of a CVV such as LAIV_H1.opt, given its clear superiority in protection in ferrets. However, the wider acceptance within the community, and subsequent implementation of such an approach commercially would require changes to current seasonal influenza CVV development and approval for use processes. HAI assays have been the gold standard method for assessing antigenic characteristics of influenza viruses for decades, within a robust, well-developed and understood infrastructure, and within this, LAIV is still a comparatively new influenza vaccine platform. To support any potential changes in approach for LAIV, data confirming the increased efficacy of LAIV_H1.opt in humans would likely be required. Recent studies in controlled human infection models (CHIM), such as that published by Thwaites et al., showing that adult LAIV recipients can be segregated by their production of either mucosal or systemic responses to vaccination⁴¹, suggest that a CHIM study could provide the ideal setting for a clinical comparison of LAIV_H1.par and LAIV_H1.opt.

While the potential for superior protection from a CVV such as LAIV_H1.opt offers exciting possibilities, it remains that there is currently no defined correlate of protection for LAIV. To reassess the role of serum antigenicity in LAIV CVV selection, alternative mechanisms of protection driven by high levels of LAIV virus replication will need to be understood.

One explanation for the improved protection conferred by LAIV_H1.opt is that, due to its enhanced replication in the nasal epithelia, it was simply more immunogenic than LAIV_H1.par. Despite LAIV_H1.opt serum antibodies being considered ‘mismatched’ to reference strains, the absolute levels of LAIV_H1.opt serum antibodies measured against both the parental e-wt and the challenge c-wt were considerably higher than those for LAIV_H1.par, when delivered at a comparable dose. Only with a 1000-fold increase in LAIV_H1.par dose were serum antibody titres against e-wt and c-wt comparable to those of LAIV_H1.opt. This was particularly evident when looking at functional, virus-neutralising antibodies. Even then, levels of infectious LAIV_H1.par virus in nasal washes remained undetectable, highlighting the difference in its replicative efficiency relative to LAIV_H1.opt. This could indicate that the MP-7 immunogenic response was linked more to the magnitude of initial antigen exposure, rather than subsequent virus replication. It is possible that induction of a greater overall magnitude of antibody responses by LAIV_H1.opt might have been sufficient to overcome its relative reduction in strain-specific antibodies.

Alternatively, anti-NA antibody responses have been proposed as a contributing factor in protection from influenza virus infection^42,43. While anti-NA antibodies have been shown to be induced in the sera of children and adults receiving LAIV⁴⁴, their role in protection is unclear. Here, LAIV_H1.opt was found to induce significantly higher levels of serum anti-NA antibodies than LAIV_H1.par, at an equivalent dose. With the NA proteins of all LAIV and wt strains in this study being conserved, changes in anti-NA antibody magnitude could feasibly be associated with changes in the level of protection observed. However, further investigation will be required to define the contribution of these anti-NA responses to LAIV-mediated protection in this model, particularly their presence and function at the mucosal surface.

Outside the endpoints measured here, there are additional mechanisms that could potentially contribute to the protective effects of LAIV_H1.opt. Existing data support the involvement of T-cells and mucosal antibodies in the LAIV immune response^15,45–47, with T-cell responses having been shown to be broadly cross-reactive and durable in children^46,48. This cross-reactivity could potentially compensate for antigenic mismatch of the HA protein. In addition, Marriott et al. showed that LAIV-induced T-cell responses contributed to reduced disease severity in the ferret model⁴⁹. However, the lack of complete protection observed suggests that additional mechanisms are also likely to be required for the success of LAIV_H1.opt.

Otherwise, the role of mucosal antibodies in protection from influenza remains relatively unexplored. At the mucosal surface, antibodies that are targeted to non-HA proteins, such as NA, or are more cross-reactive than those measured systemically, could account for reduced need for serum antigenic match. Alternatively, antibodies that mediate their effects through Fc effector functions might offer an explanation⁵⁰. Such antibodies were recently implicated in the protection of seasonally vaccinated children from infection by wt A/H1N1pdm09 viruses upon their emergence⁵¹, and have been shown to be induced in response to LAIV⁵². However, to our knowledge, these observations have not yet been extended to the mucosal compartment for LAIV.

As the aim of the work presented here was to determine the relative contributions of hNEC replication and antigenic match to the ability of A/H1N1pdm09 LAIV strains to protect from wt challenge, a detailed investigation of the immunology involved was considered out of scope of the study. However, the investigation of immune responses to LAIV and identification of alternative mediators of protection, particularly at the mucosal surface, is the subject of ongoing investigation and will be addressed in future publications.

Other limitations of the work described include the fact that these data were based on a single A/H1N1pdm09 LAIV strain. Reproducing these effects for additional strains with different genetic backgrounds will be important for building an understanding of the universality of this approach. In addition, there is currently no clear threshold for the level of hNEC replication required to achieve protection in the presence of HA mismatch. The degree of antigenic difference that a CVV such as LAIV_H1.opt could tolerate before losing these benefits is also unknown. Further work will be required to better understand such factors.

In summary, the data presented show that the optimised hNEC replication of LAIV_H1.opt led to superior protection from wt challenge, relative to LAIV_H1.par, despite its reduced antigenic similarity to WHO reference strains. This raises the possibility of a change of paradigm for LAIV CVV selection, with optimal hNEC replication prioritised over serum antigenic match as the critical determinant of high VE. However, further investigation will be required to understand the breadth of these effects and to determine which elements of the immune response are involved.

Methods

Cells and viruses

Madin–Darby canine kidney (MDCK) cells were obtained from American Type Culture Collection (ATCC) and maintained in Eagle’s minimum essential medium (EMEM) with non-essential amino acids (VWR, Cat. No. 392-0423) at 37 °C and 5% CO₂. Cultures below 20 passages were used for the titration of LAIV and cell-propagated wt (c-wt) material by focus forming assay (FFA) as previously described¹⁹.

Fully differentiated human nasal epithelial cells (hNECs) were purchased from Epithelix (Epithelix Sàrl, Switzerland) and maintained in transwells at the air-liquid interface using serum-free MucilAir^TM culture media (Epithelix Sàrl, Switzerland) as previously described¹⁸. Cells were cultured at 37 °C and 5% CO2 with basal media replacement every 3-4 days. The apical surface was washed once every 7 days using Phosphate buffer PBS pH 7.4 (Gibco™; Thermo Fisher Scientific; Cat. No. 675 10010023) to remove mucous.

Vaccine viruses used in this study were representative of the 2023–2024 FluMist® Quadrivalent LAIV (QLAIV) formulation and generated from non-commercial material. The parental (LAIV_H1.par) A/Norway/31694/2022 HA and NA gene sequences were derived from GISAID Isolate ID: EPI_ISL_16043978 and cloned into a bi-directional transcription vector (pAD3000)⁵³. The optimised A/Norway/31694/2022 HA (LAIV_H1.opt) was produced using site-directed mutagenesis to introduce 127E, 222 G, 223Q (H3 numbering)⁵⁴ substitutions into the parental A/Norway/31694/2022 HA protein. A/H3N2; A/Norway/16606/2021 (GISAID Isolate ID: EPI_ISL_5946220), B-Vic; B/Austria/1359417/2021 (GISAID Isolate ID: EPI_ISL_1519459) and B-Yam; B/Phuket/3073/2013 (GISAID Isolate ID: EPI_ISL_165594) HA and NA gene sequences were also cloned into the pAD3000 vector⁵³. All 6:2 reassortant LAIV viruses were generated using an eight-plasmid based reverse genetics system as previously described⁵³. Viruses were subsequently propagated in the allantoic cavity of 11-day old embryonated hen eggs (Charles River; USA). Eggs were incubated at 33 °C with 70% humidity for 72 h and chilled at 4 °C for a minimum of 5 h prior to harvest of the allantoic fluid. Vaccine virus sequence identity was confirmed via Sanger sequencing of the HA and NA to ensure a match to their reference plasmids.

A/NOR22 LAIV_H1.par and LAIV_H1.opt were then purified using a 30% sucrose cushion as previously described³⁸. The A/H3N2, B-Vic and B-Yam viruses were purified using a sucrose gradient as previously described⁵⁵. Final purified viruses were titrated using a focus-forming assay¹⁹. A/NOR22 LAIV_H1.par and A/NOR22 LAIV_H1.opt vaccines were formulated with A/H3N2, B-Vic and B-Yam strain components targeting a 4.0 or 7.0 Log₁₀ FFU/0.2 ml dose. Vaccine was formulated in PBS with 1x sucrose phosphate (ThermoFisher Scientific SH3A1795) and 1x gelatine-arginine-glutamate (ThermoFisher Scientific: custom product, Cat. No. AC10207676).

For the challenge virus, A/Norway/31694/2022 c-wt HA and NA gene sequence were derived from GISAID (Isolate ID: EPI_ISL_15728546) and cloned into the pAD3000 vector. PB2, PB1, PA, NP, M, and NS gene sequences were derived from A/Victoria/4897/2022 wt (GISAID Isolate ID: EPI_ISL_17102775). 6:2 reassortant A/NOR22 c-wt was generated using reverse genetics as described previously⁵³. Virus was expanded by propagation in MDCK cells at 37 °C with virus infected at a multiplicity of infection (MOI) of 0.01. Identity of HA and NA genes of progeny virions were confirmed using Sanger sequencing. Two amino acid changes were present in the HA protein of A/NOR22 c-wt, compared to the A/NOR22 e-wt reference. A common Q223R egg-adaptation acquired during WHO e-wt isolation was present in the receptor-binding site (H3 numbering, Supplementary Table 1)⁵⁴. A second change, in the fusion domain (A9T) occurred during LAIV virus rescue in cell culture. Previous studies indicated a risk of attenuation of HA fusion domain mutants in hNEC and ferrets^56,57. To address this, replication of A/NOR22 c-wt was tested in hNEC prior to in vivo use and confirmed to be comparable to a previously published challenge c-wt (Supplementary Fig. 8)³⁸.

Infection of primary human nasal epithelial cells (hNEC)

LAIV_H1.par and LAIV_H1.opt replication was modelled in hNEC as previously described¹⁸. Prior to infection, the apical surface was washed to remove accumulated mucous. 200 μl of PBS pH 7.4 (Gibco™; Thermo Fisher Scientific; Cat. No. 675 10010023) was applied to the apical surface and incubated at 37 °C and 5% CO₂ for 20 min. Apical supernatant was removed, and cells were washed an additional three times. hNEC were infected at an MOI of 0.01 using LAIV_H1.par or LAIV_H1.opt for 1 hour at 33 °C and 5% CO₂. Inocula were removed and cells washed once with PBS pH 7.4 (Gibco™; Thermo Fisher Scientific; Cat. No. 675 10010023) to remove unbound virus. hNECs were incubated at 33 °C and 5% CO₂ for four days with apical samples taken every 24 h. 200 μl PBS pH 7.4 (Gibco™; Thermo Fisher Scientific; Cat. No. 675 10010023) was applied to the apical surface of each transwell and incubated for 20 min at 33 °C and 5% CO₂. Virus-containing apical supernatants were harvested and stored at -80 °C until subsequent titration by TCID₅₀ in MDCK cells.

Quantification of infectious virus using tissue culture infectious dose 50 (TCID₅₀)

Fully infectious virus particles were measured by TCID₅₀ assay as previously described¹⁹. Briefly, MDCK cells in 96-well tissue culture plates were inoculated with ten-fold serial dilutions of hNEC infection supernatant samples in the presence of 1:400 10xTrypLE^TM select enzyme (Life Technologies, Cat. No. A12177). Plates were incubated for 6 days at 33 °C for LAIV_H1.par and LAIV_H1.opt. Following incubation, 50 µl of 40 µM 2′-(4-Methylumbelliferyl)-α-D-N-acetylneuraminic acid sodium salt hydrate (MUNANA) (Sigma, Ref. M8639) was applied to MDCK cells and incubated for one hour at 37 °C. MUNANA-neuraminidase interaction was stopped using H₂O with 0.1 M Glycine (Sigma, Cat. No. G7126) containing 25% Ethanol (VWR chemicals, Cat No. 437433 T) and adjusted to pH 10.7 using 6 M NaOH (Avantor, Cat. No. 5672-02). Fluorescence (λ_exc = 355 nm, λ_em = 450 nm) was read from the plates immediately (SpectraMax M5, Molecular Devices) and TCID₅₀ titres determined by the Spearman-Karber method.

Immunogenicity of LAIV_H1.par and LAIV_H1.opt in the ferret model

All animal work was conducted at the UK Health Security Agency (UKHSA), Porton Down. Experimental work was conducted under the authority of a UK Home Office-approved project license subjected to local ethical review at UKHSA, Porton Down by the Animal Welfare and Ethical Review Body (AWERB) as required by the Home Office Animals (Scientific Procedures) Act 1986. Ferrets were pair housed at Advisory Committee on Dangerous Pathogens (ACDP) containment level 2. Cages met with the UK Home Office Code of Practice for the Housing and Care of Animals Bred, Supplied or Used for Scientific Procedures (December 2014). Access to food and water was ad libitum and environmental enrichment was provided. Ferrets were confirmed seronegative for circulating A/H1N1pdm09, A/H3N2, and influenza B viruses by hemagglutination inhibition (HAI) assay.

Four healthy, female ferrets (Mustela putorius furo) aged 4–5 months were obtained from a UK Home Office accredited supplier (Marshall’s Biosciences, UK). Ferrets were subcutaneously chipped with wand-activated ID/temperature chips. Vaccine (LAIV_par or LAIV_opt) was intranasally administered at a high dose (≥7.0 Log₁₀ FFU/animal) in a 0.2 ml dose (~0.1 ml/naris) per ferret, following light sedation with isoflurane. 14 days post-vaccination, ferrets were anaesthetized with intramuscular injection of ketamine/xylazine (17.9 mg/kg and 3.6 mg/kg bodyweight) and exsanguination was affected via cardiac puncture, followed by injection of an anaesthetic overdose (sodium pentobarbitone Dolethal, Vetquinol UK Ltd, 140 mg/kg). Sera were prepared from whole blood collected prior to termination.

Serum immune responses were assessed using HAI assays, which were performed as per standard methodology. Ferret antiserum (100 µl) was added to 150 µl 2x receptor-destroying enzyme (RDE) reagent (Deben Diagnostics, Cat. No. 370013) and incubated at 37 °C for 18–20 h. Sodium citrate (150 µl of 2% w/v; Sigma Aldrich, St Louis, MO, USA: Cat. No. W302600) was added and the reaction was heat inactivated at 56 °C for 45 min. Treated antiserum was then serially diluted two-fold and added to a standardised concentration of virus (8 HAU/well). Virus-antibody complexes were incubated at room temperature for 30–40 min before addition of 0.5% turkey red blood cell suspension per well. The virus-sera-red blood cell mix was incubated at room temperature for 60 min and HAI endpoints recorded as the reciprocal of the highest dilution of antiserum able to fully prevent hemagglutination. If no endpoint was detectable, a reading of 4 was recorded (1/2 lower LoD) for statistical analysis.

Vaccine efficacy of LAIV_H1.par and LAIV_H1.opt in the ferret model

Sixteen healthy male and sixteen healthy female ferrets (Mustela putorius furo) aged over 5 months were obtained from a UK Home Office accredited supplier (Marshall’s Biosciences, UK). Ferrets were randomly assigned to study groups by equal distribution of male and female ferrets in all study groups with 4 animals per group as per Fig. 2. On study day -16 and day -15, ferrets in all groups were surgically implanted with Star-Oddi DST nano-T temperature loggers (Star-Oddi, Iceland) into the peritoneal cavity and subcutaneously chipped wand-activated ID/temperature chips under anaesthesia (ketamine/xylazine at 17.9 mg/kg and 3.6 mg/kg bodyweight). On study day 0, 0.2 ml of vaccine was intranasally administered (~0.1 ml/naris) per ferret, following light sedation with isoflurane. Vaccine was dosed at a previously defined¹⁹ ferret dose (4.0 Log₁₀ FFU/strain/animal) of monovalent LAIV_H1.par (MP-4) and monovalent LAIV_H1.opt (MO-4), as well as their FluMist® quadrivalent formulation (QP-4, QO-4). A high-dose (7.0 Log₁₀ FFU/animal) monovalent LAIV_H1.par (MP-7) was included alongside an unvaccinated group (Mock) as negative controls. On study day 28, ferrets were lightly sedated with isoflurane and intranasally challenged with a 0.2 ml ( ~ 0.1 ml/naris) dose of A/NOR22 c-wt strain at 5.0 Log₁₀ FFU/animal. Following c-wt challenge, ferrets were observed twice daily, up to 5 days post-challenge, for the development of influenza-like-illness symptoms and activity monitoring. (see Supplementary Fig. 6). All ferrets were euthanised at study day 33. Ferrets were anaesthetised and exsanguinated as previously described¹⁹.

Temperature loggers recorded data hourly from day -16 until day 33. From day -2, ferret’s weight was recorded once daily and clinical observations monitored twice daily over the study duration. Clinical observations were subjectively assigned by competent and experienced animal technicians. Following sedation, nasal washes were collected in 2 ml 1x PBS at day -2, day 1 to 5 and day 29 to 33. On study days 14 and 21, ferrets were sedated and blood was taken from the cranial vena cava vessel for serum preparation. Data generation focussed on day 21 timepoint for the representative timepoint of serum response¹⁹.

Infectious virus shedding in groups vaccinated with monovalent A/NOR22 (day 1–5 post-vaccination samples) and c-wt virus (day 1–4 post-challenge samples, or day 29–33 of study) were quantified using the same TCID₅₀ assay method as described in the previous section titled ‘Quantification of Infectious Virus using Tissue Culture Infectious Dose 50 (TCID₅₀)’.

Statistical comparisons were made on geometric mean titres derived from the 5-day post-vaccination shedding data. No titres above the limit of detection (LoD) were obtained for the MO-4 group and were therefore excluded from initial statistical analysis. Titres for groups with c-wt shedding above LoD were tested for normal distribution (Shapiro–Wilk test) and equal variance (Levene test). All results were not significantly different (ns, p > 0.05). Differences between groups were tested by one-way ANOVA (p < 0.0001) and groups above LoD were compared to the Mock group (Dunnett’s test). For samples with values below LoD, the Mock was compared to the TCID₅₀ LoD threshold (1.2 Log₁₀ TCID₅₀/mL) by one-sample Student’s t-test (p = 0.006). This was a proxy test in lieu of any detectable virus.

Serum immune response of LAIV_H1.par and LAIV_H1.opt from the MP-4, MO-4, QP-4, QO-4 and MP-7 vaccinated ferrets were assessed using the same HAI assay as described in the previous section titled ‘Immunogenicity of LAIV_H1.par and LAIV_H1.opt in the ferret model’. The Mock-vaccinated group was used as a negative control.

HA antigenicity of LAIV_H1.par and LAIV_H1.opt using in vitro microneutralisation (MN) assay

Working solutions of A/NOR22 e-wt, c-wt, LAIV_H1.par and LAIV_H1.opt were prepared to a final infectivity of 1 – 10% per well using an NP based infectivity assay. Briefly, infected cells were washed twice with PBS (Gibco™; Thermo Fisher Scientific; Cat. No. 675 10010023) and fixed for 20 min in ice-cold Methanol (VWR Chemicals Cat No.; 83638.320). Cells were washed twice using PBS (Gibco™; Thermo Fisher Scientific; Cat. No. 675 10010023) and incubated at room temperature for 1 hour in a 1% BSA blocking solution (Thermo Scientific Cat No.; A3733). A working solution (1:1000 in 1% BSA) of Influenza A NP antibody (BioRad, Cat. No. MCA400) conjugated with DyLight MicroScale Cy5 (λ_ex: 652 nm, λ_em: 672 nm, Thermo, Cat. No. 84536) was incubated with cells for 1 hour at room temperature. Cells were washed twice with PBS (Gibco™; Thermo Fisher Scientific; Cat. No. 675 10010023) and Hoescht solution (1:10,000 in PBS, Thermo Scientific Cat No. 62249) was subsequently applied. Virus-positive cells were measured through an NP count using cell imaging reader (Cytation 7, Bio Tek).

For MN assays, 150 µl of ferret sera was treated with 225 µl of 2x RDE reagent (Deben Diagnostics, Cat. No. 370013) and incubated at 37 °C for 18–20 h. Following RDE treatment, 225 µl of EMEM (VWR, Cat. No. 392-0423) was added to each serum and samples were heated at 56 °C for 45 min, then immediately stored at 4 °C. Heat-treated antisera was serially diluted two-fold and 50 µl of sera was incubated with 50 µl of working stocks of A/NOR22 e-wt, c-wt, LAIV_H1.par or LAIV_H1.opt for 1 hour at room temperature. MDCK cells were subsequently infected for 18-h with sera-virus mixture and incubated at 33 °C, 5% CO₂ (sera-LAIVs neutralisation mixture) or 37 °C, 5% CO₂ (sera-c-wt / sera-e-wt neutralisation mixture). Virus neutralisation was measured using the NP infectivity assay described above. The half-maximal inhibitory sera dilution (MN₅₀) values were interpolated using a three-parameter dose-response curve with a robust fit method (Prism v10.1.2).

Neuraminidase neutralisation in vitro: Enzyme-linked Lectin Assay (ELLA)

NA-specific immune response was measured using enzyme-linked lectin assay (ELLA) to assess anti-NA antibody levels adapted from Gao et al.⁵⁸. A reassortant A/H2N1 LAIV virus containing the NA gene of A/NOR22 and the antigenically distinct HA and internal genes from A/Ann Arbor/6/1960 MDV (A/H2N1_NorNA) was generated to ensure NA-specific immune response with minimal HA activity was measured. This was produced using reverse genetics as described in the previous section and used to determine NA-specific immune response from vaccinated ferrets.

To determine a standardised virus working solution, fetuin-coated 96-well plates were made by diluting fetuin (Sigma-Aldrich, St Louis, MO, USA; Cat. No. F3385) in coating buffer (Bio-Rad Laboratories LTD, Hercules, CA, USA; Cat. No. BUF030B) to 25 μg/ml, before dispensing 100 μl into each well of a 96-well Maxisorp plate (Life Technologies, Carlsbad, CA, USA; Cat. No. 442404) and storing at 5 °C for 18-h. Virus dilutions were determined by serially diluting virus samples 3-fold in DPBS (Life Technologies, Cat. No. 14040-182) with 1% BSA and 0.5% Tween-20 (Sigma-Aldrich, Cat. No. P1379), before being added to fetuin-coated plates and incubated at 37 °C for 18–20 h. Plates were developed by washing with PBS containing 0.05% Tween-20 (PBS-T) before adding a peanut-agglutinin horseradish peroxidase conjugate (PNA-HRPO - Sigma-Aldrich, Cat. No. L7759-1MG) diluted in DPBS with 1% BSA. Plates were incubated at room temperature for 2 h, then washed with PBS-T. TMB-ELISA substrate (Life Technologies, Cat. No. 34029) was added for 20 min before the reaction was stopped with the addition of sulfuric acid stop solution (Life Technologies, Cat. No. N600) and OD₄₅₀ values were measured. The virus dilution required for 80% of the maximum NA activity output was measured as the virus working solution.

To measure immune responses, ferret serum samples were heat-inactivated at 56 °C for 45–60 min. Heat-inactivated antisera were serially diluted in DPBS (Life Technologies, Cat. No. 14040-182) with 1% BSA and 0.5% Tween-20, before being mixed with the virus working solution. Plates were developed as described above. NA activity in each well was determined as a percentage of the mean of the virus-only positive control wells. The NAI titre was calculated as the reciprocal of the dilution required to produce 50% inhibition (referred to as NAI₅₀). NAI₅₀ values were derived for each ferret using a three-parameter logistic curve using a robust fit method (Prism v10.1.2).

Analysis of fever and bodyweight clinical end points

Intraperitoneal data loggers recorded core body temperatures for individual study animals hourly from the time of vaccination to termination. Animal handling records identified all subsequent inoculation or sampling intervention times requiring anaesthesia. The mean temperature profile of the unvaccinated control group (four animals) was used to define the post-challenge ‘fever period’ for A/NOR22 c-wt virus. The fever period start was defined as the first of three post-challenge time-points where the Mock group mean body temperature exceeded >1.5 °C standard deviations above the mean pre-challenge baseline body temperature of all animals. The end of the fever period was defined as the first of three control group average temperature measurements returning below the same limit, or study end if this occurred first. The mean of each animal’s hourly body temperature values during the fever period was calculated. The mean of each animal’s baseline body temperature was subtracted from the mean fever period body temperature to yield a mean Δ°C value. This represented mean deviation from normal body temperature during the period that c-wt virus caused fever in Mock group.

The Δ°C values were compared between groups. No within-group outliers were detected (Huber test, K Sigma-3). Values were assessed for equal variance (Levene test, p < 0.05) and within-group normality (Shapiro–Wilk test, all ns). A Welch ANOVA (p < 0.0001) was followed by comparison of each treatment group to the Mock group (Welch t test with Holm–Šídák correction for multiple comparisons).

Post-challenge weight data for each animal (6 days, including the day of vaccination as Study Day 0) were subjected to linear regression. No outliers detected in the regression residuals (Huber test, K Sigma-3). Slope values were compared using the same procedure as described for temperature data (Levene p < 0.01, Shapiro-Wilk all ns, Welch ANOVA p < 0.01).

Statistical analysis software

Data were prepared in Excel 365 (Microsoft Inc.) and statistical analyses were performed in JMP v17.2 (SAS Institute) and Prism v10.1.2 (Graphpad Software LLC). Curve fittings were additionally performed in Prism v10.1.2 (Graphpad Software LLC).

Institutional Review Board Statement

AstraZeneca’s Council for Science & Animal Welfare (C-SAW) provided ethical approval for all study plans. All experimental work was conducted under the authority of a UK Home Office-approved project licence subjected to local ethical review at the UKHSA by the Animal Welfare and Ethical Review Body, as required by the Home Office Animals (Scientific Procedures) Act 1986.

Author contributions

S.L.l.: Writing—original draft, investigation, data curation, visualisation. K.E.S.: Writing—original draft, investigation, data curation. S.A.: Writing—original draft, validation. D.B.: Writing—original draft, investigation, data curation, visualisation. S.C.: Investigation. J.C.: Writing—original draft, methodology, software, formal analysis. S.D: Investigation, validation. K.R.: Writing—original draft, investigation, resources, data curation, project administration. J.S.: Investigation, resources, data curation. K.T.: Investigation. L.P: Conceptualisation, resources, supervision. O.D: Conceptualisation, methodology, formal analysis, writing—original draft, visualisation, supervision and project administration. All authors: Writing—review & editing.

Data availability

All data associated with this study are available in the main text or supplementary materials. Further information or requests for resources and reagents should be directed to and will be fulfilled by the corresponding author, Oliver Dibben ([oliver.dibben@astrazeneca.com](mailto:oliver.dibben@astrazeneca.com)).

Competing interests

At the time of data generation, S.L.l., K.E.S., S.A., D.B., S.C., J.C., S.D., K.T., L.P. and O.D. may have been employees and shareholders of AstraZeneca, the manufacturer of FluMist® (Fluenz) influenza live virus vaccine. K.R. and J.S. declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41541-025-01281-z.