Vaccination with antigenically complex hemagglutinin mixtures confers broad protection from influenza disease

Abstract

Current seasonal influenza virus vaccines induce responses primarily against immunodominant but highly plastic epitopes in globular head of the hemagglutinin (HA) glycoprotein. Due to viral antigenic drift at these sites, vaccines need to be updated and re-administered annually. To increase the breadth of influenza vaccine-mediated protection, we developed an antigenically complex mixture of recombinant HAs designed to redirect immune responses to more conserved domains of the protein. Indeed, vaccine-induced antibodies were disproportionally redistributed to the more conserved stalk of the HA without hindering, and in some cases improving, antibody responses against the head domain. These improved responses led to increased protection against homologous and heterologous viral challenge in both mice and ferrets compared with conventional vaccine approaches. Thus, antigenically complex protein mixtures can at least partially overcome HA head domain antigenic immunodominance and may represent a step towards a more universal influenza vaccine.

Introduction

Infection with influenza virus causes mild to severe respiratory disease with an estimated 290,000 to 650,000 deaths annually (1). The most effective countermeasure to prevent influenza disease is vaccination. Current seasonal influenza vaccines predominantly elicit antibodies that target immunodominant antigenic sites on the hemagglutinin (HA) globular head domain for neutralization, which are a strong correlate of protection against influenza in humans (2, 3). However, because of frequent antigenic variation of these immunodominant epitopes, annual influenza vaccines provide limited protection against drifted strains and typically confer poor protection during pandemic outbreaks (4, 5). Thus, improved vaccines able to induce cross-protective immunity against divergent influenza viruses by eliciting responses against more conserved viral proteins or protein domains are of high importance.

Despite head domain immunodominance, the influenza virus HA is still a favorable target for vaccine development because it is the most abundant structural protein on virions and plays a critical role during viral entry. Further, antibodies targeting the more conserved HA stalk domain can provide broad protection to divergent virus strains (6–11). However, stalk specific antibodies are generally poorly induced by inactivated influenza virus vaccines (12, 13). These results have motivated many of the current efforts in the development of universal vaccine to focus on eliciting stronger HA stalk domain responses. One limitation with this approach, however, is that most stalk specific antibodies are less potent in controlling viral spread compared with head reactive antibodies (14, 15). As such, vaccine approaches to elicit these responses must be highly immunogenic, or sometimes computationally optimized (16–18) to elicit high antibody titers. Previous approaches specifically to induce stalk responses have included rationally designed peptides and nanoparticles to present only conserved HA stalk to elicit response to stalk (9, 19–23). Additionally, increased glycosylation sites in HA head have been used to shield immunodominant epitopes to preferentially expose the conserved stalk domain to the immune system (24, 25). Sequential vaccination with chimeric HAs consisting of divergent HA head domains but a conserved HA stalk domain has also been shown to boost stalk response and elicits broader protection (26–28). The increased stalk responses with these approaches, however, typically come at the expense of what would have been protective responses against the head domain.

Our goal in this study was to develop a vaccine that would allow the post-vaccination elicitation of therapeutically protective titers of both head and stalk directed antibodies. We hypothesized that by vaccinating with an antigenically complex mixture of HA proteins with diversity in a key head antigenic site, we could change the “molarity” of key antigenic epitopes and alter the resulting composition of the antibody response. Using an oligo- based mutagenesis approach targeting “Sb,” the most immunodominant antigenic site in H1 HA head domain (29–31), we generated a protein mixture representing more than 10⁴ individual variant HA proteins. Vaccination with the antigenic mixture disproportionally increased responses to the conserved stalk domain compared with standard vaccination. This increased stalk response also led to better reactivity against divergent group 1 HA proteins. Although we initially thought that these responses may come at the expense of some of the normal head-directed responses, we found instead that the antigenic mixture induced higher quality head antibody responses. Challenge studies revealed that this altered immune response provided better protection against challenge with both homologous and heterologous influenza strains. Together, these data show that antigenically complex protein mixtures provide an approach to redirect immune responses across protein domains of different inherent antigenicity and may represent a path forward for universal influenza vaccine development.

Results

Development of an antigenically complex HA subunit vaccine

The primary goal of this research was to alter the ratio of antibodies raised against variable and conserved domains in the HA. We decided on an approach to change the relative molarity of antigenic and conserved sites so that the immune system would be more likely to repeatedly encounter sub- immunodominant, but conserved, epitopes. The Sb antigenic site of the A/Puerto Rico/9/1934 (PR8) H1 HA was therefore selected, as it is known to be a dominant antigenic site in both humans and mice (29). We selected four codons in that site that were proximal to each other in the primary sequence and, using DNA oligos with the four codons randomized, we initially generated a complex mixture of approximately 10⁵ HA genes (Fig. 1, A and B). Because not all the mutants were expected to express or fold correctly, we used a lentivirus vector to deliver the mutant HA proteins at a low multiplicity of infection (MOI) and performed a magnetic-activated cell sorting (MACS) with a conformation-specific HA stalk antibody, 6F12 (10), to only purify HA trimers that folded and trafficked appropriately (Fig. 1C). After purification, we amplified the mutant HA genes and cloned the soluble ectodomains into a protein expression vector.

Fig. 1. Design, generation, and characterization of and antigenically complex HA vaccine.

(A) Shown is a Pymol model of the H1 HA trimer based on the published crystal structure of the HA of A/Puerto Rico/8/1934, PDB number 5VLI (60). Residues that were mutated are indicated in orange. (B) Shown is a diagram of the approach to generate the HA mutant library. Ab, antibody. (C) Shown is the percentage of 6F12 positive cells (out of the lentivirus transduced population) of Sb_mut HA library by flow cytometry after MACS. SSC-A, side scatter area. (D) Sequencing results are shown for 8 clones from sorted library. Alignment was generated by Clustal Omega Multiple Sequence Alignment tool. * in (D) indicates a conserved sequence. (E) The heatmap shows the amino acid frequency of each substituted position in the DNA library. (F) Sequence logo plot showing amino acid frequency of each substituted position of DNA library. (G) The glycosylation of Sb_mut HA was evaluated by SDS-PAGE after PNGase F treatment. (H) Antigenicity of Sb_mut HA was determined by ELISA with monoclonal antibodies targeting different antigenic sites (n=6 biological replicates). Antigenic sites are indicated at the bottom. OD, optical density. Statistical differences for (H) were determined using Mann-Whitney U test (*p<0.05; ns, not significant), and bars indicate mean values. (I) Pie charts show the amino acid distribution of each mutagenized position of the protein library.

To characterize post-purification mutant HA diversity, we picked individual clones for Sanger sequencing and also performed deep sequencing of the mutated region. Our manual analysis revealed mutations only in the desired locations, and the next-generation sequencing detected 176,745 unique clones in pre-purification library and 86,835 unique clones in post-purification library, including clones missing codons and containing stop codons (Fig. 1, D to F). Although the frequency of amino acids at each of the mutated positions was altered slightly after selection, diversity remained high (data file S1). To understand how the diverse Sb mutant mixture would express when transfected en masse compared with the “clonal” WT HA, we transfected the expression plasmids into 293F suspension cells and purified the his-tagged HA ectodomains. Sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE) analysis of the purified proteins revealed bands of similar purity and size, and treatment with PNGase F revealed they were similarly glycosylated (Fig. 1G, fig. S1). We next validated antigenicity of our Sb mutant HA protein mixture by performing enzyme linked immunosorbent assays (ELISAs) with a panel of antigenic site monoclonal antibodies (31–35). Whereas antibodies against the other antigenic sites (Sa, Ca, Cb) bound both the WT and Sb mutant HA mixture equally, the Sb-reactive monoclonal antibody failed to bind the Sb mutant library due to the lack of an unmutated Sb site in the mixture (Fig. 1H). Finally, we developed a mass spectrometry-based approach to evaluate the amino acid diversity of the Sb mutant mixture at the protein level after purification. Although not every clone in our DNA library could be detected at the protein level due to technical limitations, the amino acid distributions of the sampled population were on par with the expectations from the input plasmid sequencing (Fig. 1I, table S1). Thus, our approach was successful in generating a recombinant HA Sb mutant protein mixture (Sb_mut HA) for further study as an influenza vaccine.

Vaccination with the WT or Sb antigenic HA mixture elicits distinct antibody responses in mice

We were next interested in understanding how the immune system would respond to the Sb_mut HA mixture compared with the homogenous WT HA. We therefore vaccinated mice (adjuvanted with AddaVax) with a prime or prime-boost regimen of one of the two HA groups or an irrelevant mCherry control group, which also harbored the same trimerization domain and protein purification tag. Post-vaccination serum samples were collected and a series of ELISAs were performed to measure antibody responses (Fig. 2A). Although we thought that it was likely the immunogenicity of the protein would be compromised by randomizing the Sb site, the Sb_mut HA post-vaccination serum samples showed significantly higher binding to purified full-length HA (post-prime, P ≤ 0.0001; post-boost, P = 0.0015) when compared with serum samples from mice immunized with the WT HA vaccine (Fig. 2B). The Sb_mut HA also induced increased responses to authentic purified virions post-prime (P ≤ 0.0001) (Fig. 2C). We next wanted to evaluate if we were successful in our original goal of shifting relative antibody responses against conserved and variable domains. We therefore cloned and purified trimerized soluble PR8 HA head and stalk domains for ELISA analysis. The HA head and stalk constructs were first checked by ELISA with monoclonal antibodies targeting different epitopes; the HA head protein was bound only by HA head-reactive monoclonal antibodies, while HA stalk protein showed binding only to HA stalk-reactive antibodies, as expected (fig. S2). Although the responses to both head and stalk domains were higher in the Sb_mut HA vaccine group (Fig. 2, D and E), we also found that the relative binding of antibodies targeting the conserved stalk domain were increased by a larger margin than the ratios of antibodies against full-length HA or the head domain suggesting that indeed, antibodies against the conserved stalk were disproportionately elicited (Fig. 2F). Additionally, to understand responses to the Sb antigenic site itself, we constructed and expressed a mutant HA protein harboring a mutated Sb antigenic site (fig. S3, A and B) and confirmed the mutation by monoclonal antibody ELISAs (fig. S3C). We then evaluated WT and Sb_mut serum binding to the mutant Sb HA. Although we could detect a decrease in binding compared with WT HA protein with WT vaccinated serum samples, we failed to detect a similar change with the Sb_mut serum samples (fig. S3D). These data suggest that Sb-directed antibodies are not major contributors to the increased HA reactivity observed after Sb_mut HA vaccination.

Fig. 2. Vaccination with antigenically complex HAs induces increased head and stalk directed antibodies.

(A) Mice received two doses of vaccine, three weeks apart and serum samples were collected 21 days post-prime or 21 days post-boost for analysis. (B to E) Antibody response to PR8 full-length HA (B), PR8 virus (C), PR8 HA head domain (D), and PR8 HA stalk domain (E) are depicted as a titration curve (left) or the calculated area under the curve (AUC, right) based on ELISAs. Arb., arbitrary. (F) Shown are relative ELISA AUC values normalized to the WT HA group against PR8 full-length (FL) HA, PR8 HA head, or PR8 HA stalk. Fold-changes are shown above each pair of bars. (G to I) Antibody avidity to PR8 full-length HA (G), PR8 HA head domain (H) and PR8 HA stalk domain (I) are represented as antibody avidity indices. Conc, concentration. n=10 mice per group. Prime and boost samples were analyzed independently; absorbance curves and AUC values cannot be directly compared across experiments. Bars indicate mean values in the right plots of (B to E) and in (F to I). Dots indicate mean values in the left plots of (B to E), where error bars represent the standard error of the means (SEM). Statistical differences in (B to E) were determined using Kruskal-Wallis test followed by a Wilcoxon rank-sum test with Bonferroni correction; statistical differences in panel G-I were determined using Mann-Whitney U test.

Finally, we tested antibody avidity to full-length HA and HA head and stalk domains to understand if our vaccine was affecting antibody quality. Antibodies elicited by Sb_mut HA showed better avidity to both full-length HA and HA head domain (Fig. 2, G and H). The avidity of the stalk domain-specific antibodies induced by Sb_mut HA was not increased after a single vaccination; however prime-boost vaccination did show an increase (Fig. 2I). Together, the data suggest that the increased antibody avidity after vaccination is primarily driven by improved responses against the head domain. Thus, vaccination with the Sb_mut HA elicits a stronger overall antibody response while simultaneously altering the antibody ratio between variable and conserved HA domains.

Sb antigenic mutant elicited antibodies are functional and prevent disease after influenza challenge in mice

Because in vitro binding does not necessarily correlate with antibody functionality, we next tested our post-vaccination serum in hemagglutination inhibition assays (HAI) and microneutralization (MN) assays. Consistent with the increased ELISA binding by the Sb_mut HA vaccination group, HAI and MN activities were significantly increased in mice that were vaccinated with the Sb_mut HA vaccine compared with WT HA vaccination when detected (HAI, post boost, P = 0.042; MN, post prime, P = 0.01; MN, post boost, P = 0.022; Fig. 3, A and B) although no HAI activity was detectable in either vaccine group after prime only. Additionally, and because stalk directed antibodies frequently do not display HAI or MN activity, we measured the ability of the post-vaccination serum to promote antibody-dependent cellular cytotoxicity (ADCC), a known function of cross-reactive antibodies including stalk-reactive antibodies (36–38). Vaccination with the Sb mutant HA mixture elicited significantly (P = 0.0005 for prime, P = 0.045 for boost) higher ADCC activity compared with WT HA vaccination (Fig. 3C). Further, and consistent with the increased neutralizing and non-neutralizing activities of the serum, we could detect higher titers of IgG isotypes associated with those functions (IgG1, IgG2b, and IgG2c, respectively) in the Sb mutant group compared to the WT HA group after prime, however those changes were not always reproducibly observed post-boost (fig. S4).

Fig. 3. Antigenically complex HA vaccination provides improved, and at least partially antibody-mediated, protection from homologous challenge.

(A to C) Mice received two doses of vaccine three weeks apart, and serum samples were collected 21 days post-prime or 21 days post-boost for HAI (n=10 per group) (A), microneutralization (n=10 per group) (B) and ADCC (n=8 for prime, n=6 for boost) (C) assays as measured against PR8 virus. ND, not detected. LOD, limit of detection. (D and E) Mice were intranasally infected with 12,500 PFU of PR8 virus three weeks after prime vaccination; body weight (D) and survival (E) were monitored and recorded for 14 days post infection (n=9 per group). (F and G) Mice were primed and then infected with 12,500 PFU of PR8 virus as in (D and E); lungs were collected at 5 days post infection for H&E staining (representative lung lobes from each group are shown, n=3 per group; scale bar in upper row, 5 mm; scale bar in lower row, 500 μm) (F) or at 4 days post infection for virus detection (n=5) (G). (H and I) Mice were intranasally infected with 160,000 PFU of PR8 virus three weeks after boost vaccination, and body weight (H) and survival (I) were monitored and recorded for 14 days post infection (n=10 per group). (J and K) Mice received 200 μL of neat serum transfer from vaccine-primed mice, then were challenged with 500 PFU of PR8 virus 24 hours later; body weight (J) and survival (K) were monitored and recorded until 14 days post infection. Data in (J) and (K) are aggregated from two independent experiments (n=9 in Sb_mut HA group, n=10 in both mCherry and WT HA group). Dotted lines in (D), (H), and (J) indicate 75% of starting weight. Fractions in (E), (I), and (K) represent the number of surviving animals of total animals per group. Horizontal bars indicate mean values in (A), (B), (C), and (G). Dots represent mean values in (D), (H), and (J), where error bars represent the SEM. Statistical differences in (A to C) and (G) were determined using Kruskal-Wallis test followed by a Wilcoxon rank-sum test with Bonferroni correction, statistical differences in (E), (I), and (K) were analyzed by log rank (Mantel-Cox) test.

To understand how the altered antibody response would affect protection from disease, we first challenged vaccinated mice with a lethal dose of the vaccine matched virus, PR8. In mice that received a single dose of either WT HA, Sb_mut HA, or the mCherry control, we observed a reduced weight loss and a significant reduction in mortality (P = 0.0099) in the Sb_mut HA vaccine group compared with the WT HA (Fig. 3, D and E). This trend held true across a range of doses tested, including at 10x the half-maximal lethal dose (LD₅₀) up to 1250x the LD₅₀ (fig. S5). Both vaccines reduced viral titers in the lungs, and H&E stained lung sections were imaged to allow comparison of the gross pathology (Fig. 3, F and G). The improved protection in the Sb mutant vaccine group compared with the WT HA vaccine without a corresponding decrease in viral titer suggested that a non-neutralizing vaccine effect was potentially responsible. In mice that received a prime-boost vaccine regimen, we observed complete protection from both morbidity and mortality from both vaccines, as expected (Fig. 3, H and I). Finally, we could demonstrate that this protection was at least partially antibody-mediated, as passive serum transfer experiments showed enhanced protection from the Sb_mut HA vaccine serum compared with the WT HA vaccine group (Fig. 3, J and K). These data together demonstrate that the Sb_mut HA vaccine elicits more functional antibodies and provides better protection against homologous virus challenge.

Vaccination of mice with the Sb_mut HA vaccine elicits heterologous, homosubtypic reactive antibody responses

Next, we were interested in understanding if the increased reactivity against more conserved HA domains like the stalk would manifest as increased binding to heterologous HA proteins. Using a Luminex-based purified HA binding assay, we could detect generally higher magnitude responses against H1 HAs from 1933–2015 in the Sb mutant post-vaccination serum compared with the WT HA vaccine group (Fig. 4A). We could not, however, detect meaningful reactivity to heterologous, cross-group H3 HAs, suggesting the breadth of the responses are likely restricted to either homosubtypic or potentially group 1 HA proteins (Fig. 4B). To validate the Luminex data, which is only semi-quantitative, we performed ELISAs against authentic virions from A/WSN/1933, A/USSR/92/1977, A/Bayern/07/1995, A/Solomon Islands/03/2006, and A/California/04/2009 (Cal/09). In agreement with previous observations, the post-prime Sb_mut HA vaccine serum displayed a higher magnitude of reactivity in nearly all of the post-boost samples compared with the WT HA group (Fig. 4, C to G). To test if this increased reactivity would extend to other group 1 HAs, we tested serum binding in an ELISA against H2, H5, H6, and H9 HAs. Similar to the heterologous H1 binding data, we found Sb_mut HA vaccination elicited higher antibody response to the other group 1 HAs (Fig. 4H).

Fig. 4. Antigenically complex vaccine induced responses against heterologous group 1 HA IAV strains.

Mice received two doses of vaccine, three weeks apart. Serum samples were collected three weeks after prime and three weeks after boost for analysis. (A) Luminex H1 HA binding assay results are shown as MFI values for HAs derived from the indicated strains (n=5 mice per group). (B) Luminex H3 HA binding assay results are shown as MFI values for HAs derived from the indicated strains (n=5 mice per group). (C to G) ELISA results are shown for serum collected after prime (top) or boost (bottom) vaccination with the indicated HA vaccine. AUC values are shown on the right. Shown are responses to A/WSN/1993 HA (C), A/USSR/92/1977 HA (D), A/Bayern/07/1995 HA (E), A/Solomon Island/03/2006 HA (F), and A/California/04/2009 (Cal/09) HA (G) (n=10 mice per group). Prime and boost samples were analyzed independently; absorbance curves and AUC values cannot be directly compared across experiments. Bars indicate mean values in the right plots of (C to G). Dots indicate mean values in the left plots of (C to G), where error bars represent the SEM. Statistical differences were determined using Kruskal-Wallis test followed by a Wilcoxon rank-sum test with Bonferroni correction. (H) ELISA AUC values are shown for prime and boost serum against H2 A/Japan/305/1957, H5 A/Vietnam/1203/2004, H6 A/Taiwan/2/2013, or H9 A/Hong Kong/33982/2009 HA; mean AUC values are shown (n=10 mice per group).

Vaccination of mice with the Sb antigenic mutant vaccine leads to improved protection from heterologous H1N1 challenge

Since we had observed improved reactivity against heterologous H1 HAs, we decided to test antibody functionality with the antigenically dissimilar, post-pandemic H1N1 Cal/09 isolate. Consistent with substantial divergence in the antigenic sites between the two viruses (fig. S6), our experimental conditions could not detect any meaningful HAI nor MN activity in any of the post-vaccination serum samples (Fig. 5, A and B). We could, however, detect improved significantly improved (P = 0.044) ADCC activity in the Sb mutant vaccine group after the prime, however, high sample variability led to a loss of the significance of this trend post-boost (Fig. 5C). We hypothesized this phenotype was likely due to higher abundance cross-reactive, but non-neutralizing antibodies elicited by the Sb mutant vaccine.

Fig. 5. Antigenically complex hemagglutinin vaccination provides better protection against Cal/09 challenge.

(A to C) Mice received two doses of vaccine, three weeks apart and serum samples were collected 21 days post-prime or 21 days post-boost for HAI (n=10 per group) (A), microneutralization (n=10 per group for prime, n=6 in mCherry group and 7 in WT HA and Sb_mut HA group for boost) (B) and ADCC (n=10 per group) (C) assays as measured against Cal/09 virus. ND, not detected. LOD, limit of detection. (D and E) Mice were intranasally infected with 24,000 PFU of Cal/09 virus three weeks after prime vaccination, and body weight (D) and survival (E) were monitored and recorded for 14 days post infection (n=5 per group). (F and G) Mice were primed as in (D and E) and then infected with 24,000 PFU of Cal/09 virus; lungs were collected at 7 days post infection for lung H&E staining (representative lung lobes from each group are shown, n=3 per group; scale bar in upper row, 5 mm; scale bar in lower row, 500 μm) (F) or at 5 days post infection for virus detection (n=5 per group) (G). (H and I) Mice were intranasally infected with 24,000 PFU of Cal/09 virus three weeks after boost vaccination, and body weight (H) and survival (I) were monitored and recorded for 14 days post infection (n=7 per group). (J and K) Mice were intranasally infected with 24,000 PFU of Cal/09 virus three weeks after boost vaccination as in (H and I); lungs were collected at day 7 for H&E staining (representative lung lobes from each group are shown, n=3 per group; scale bar in upper row, 5 mm; scale bar in lower row, 200 μm) (J) or for virus detection (n=5 per group) (K). Horizontal bars indicate mean values in (A), (B), (C), (G), and (K). Dots represent mean values in (D) and (H), where error bars represent the SEM. Dotted lines in (D) and (H) indicate 75% of starting weight. Fractions in (E) and (I) represent the number of surviving animals of total animals per group. Statistical differences in (A to C), (G), and (K) were determined using Kruskal-Wallis test followed by a Wilcoxon rank-sum test with Bonferroni correction, statistical differences in (E) and (I) were analyzed by log rank (Mantel-Cox) test.

We next tested how the differential vaccine-elicited responses would affect protection from heterologous virus challenge. Prime-only mice were challenged with a high dose of Cal/09, conditions where little protection from a standard vaccine would be expected, and morbidity and mortality were monitored. The mCherry vaccine control group rapidly lost weight after infection, and, although both HA-vaccinated groups also lost weight, the majority of Sb_mut HA-vaccinated animals were able to recover from infection prior to reaching the humane endpoint whereas the WT HA vaccine group did not (Fig. 5, D and E). We were again unable to detect a change in lung viral titers, and H&E stained lung sections were imaged to allow evaluation of the gross lung pathology (Fig. 5, F and G). Because we observed clinical signs of disease with the prime-only strategy, we also evaluated protection in the prime/boost model to see whether our vaccine could provide full protection from disease. Both HA-vaccinated groups displayed improved protection compared with control, but the Sb_mut HA vaccine group generally lost less weight than the WT HA group (Fig. 5H). Whereas all control vaccinated animals eventually succumbed to infection, both vaccine groups provided complete protection from mortality (Fig. 5I). In contrast to the prime-only regimen, we could detect decreased viral titers in lung for both vaccine groups, and H&E stained lung sections were imaged to allow comparison of the effects on lung disease (Fig. 5, J and K). Thus, Sb_mut HA vaccination elicited protection in mice in the context of infection with a highly antigenically divergent strain, but likely not primarily by eliciting neutralizing antibody responses.

Vaccination of ferrets with the Sb_mut HA confers protection from heterologous H1N1 challenge

Although the Sb site is immunodominant in mice and humans, it has previously been shown that, in ferrets, the Sa site is immunodominant (29). We hypothesized that in this animal model, altered ratios of vaccine-induced head and stalk antibodies would again be observed given that the abundance of different epitopes between the two vaccines would remain; the previously observed enhanced responses to the head domain, however, could be lost due to the different head epitope dominance hierarchy. To test this, ferrets were primed and boosted with the PR8-based vaccine followed by a series of blood draws and then viral challenge with the heterologous A/California/07/2009 virus (Fig. 6A). We first analyzed vaccine-induced serum antibody responses against different WT HA epitopes by ELISA. Both vaccines induced detectable responses against those antigens after a single-dose of vaccination, and responses were boosted after a second vaccination (Fig. 6, B to D). The Sb_mut HA vaccine elicited significantly (P =0.0286) more HA stalk-reactive antibodies after the boost, as expected from the earlier mouse experiments. However, different from what was observed in mice, there was no difference in serum responses against the full-length HA or HA head domain between two HA vaccine groups suggesting that the antigenic variability must be in the immunodominant site to enhance the overall protein immunogenicity. Correspondingly, there were no differences detected in head domain antibody-mediated HAI or MN activity against the vaccine-matched PR8 virus between two groups (Fig. 6, E and F).

Fig. 6. Antigenically complex hemagglutinin vaccination reduces illness during Cal/09 challenge in ferrets.

(A) Ferrets received two doses of the indicated vaccine, then animals were challenged with Cal/09 virus. Serum samples and nasal wash were collected at indicated time points for analysis. (B to D) Shown are ELISA results serum antibody response against PR8 full-length HA (B), PR8 HA head domain (C) and PR8 HA stalk domain (D). (E and F) Shown are analyses of serum HAI (E) and neutralization (F) activities against PR8 virus. (G) Results of a luminex binding assay show serum responses to the indicated H1 HAs reported as MFI values. USSR/77 indicates A/USSR/90/1977, TX/91 indicates A/Texas/36/1991, Bayern/95 indicates A/Bayern/07/1995, Bris/07 indicates A/Brisbane/59/2007, and MI/15 indicates A/Michigan/45/2015. (H) Serum antibody responses against Cal/09 full-length HA were analyzed by ELISA. (I) Viral load was measured in nasal wash collected at the indicated time points before and after challenge. Values were normalized to 18s rRNA values. (J) Maximum body temperature of individual ferrets are shown. Bars represent individual animals. Averages are shown above each group. (K) Respiratory scores are shown at day 2 post infection; 0 means no symptoms, 1 means nasal rattling or sneezing, and 2 means nasal discharge on external nares. n=4 ferrets per group. Bars indicate mean values in the right plots of (B), (C), (D), and (H) and in (I). Horizontal bars indicate mean values in (E) and (F). Dots represent mean values in the left plots of (B), (C), (D), and (H), where error bars represent the SEM. Statistical differences were determined using Mann-Whitney U test.

Despite a somewhat altered vaccine-induced immune response in ferrets compared with what we previously observed in mice, we wanted to determine if the enhanced Sb_mut HA stalk-reactive antibodies would lead to improved binding to, and protection from, heterologous viral strains. Both a Luminex-based panel with antigenically drifted H1 HAs (Fig. 6G) and ELISAs against Cal/09 HA (Fig. 6H) showed improved antibody responses in the Sb_mut HA vaccine group after boosting compared with the WT HA. Post-challenge, analysis of viral loads in the nasal wash showed both vaccines reduced viral load; however, there was no difference between two vaccine groups as previously observed in the mouse experiments (Fig. 6I). Although our ferret challenge dose was sub-lethal, we did observe clinical disease symptoms that were different between the vaccine groups. For example, some animals in the control and WT HA vaccine groups displayed higher fevers (> 41°C in body temp) than the Sb_mut HA vaccine group, which had an average maximum temperature of 40.2°C (Fig. 6J). Similarly, measurement of respiratory clinical scores at day 2 post-challenge revealed clinical systems in all of the control and WT HA vaccine; in contrast, half of the Sb_mut HA vaccinated animals did not have evidence of clinical disease (Fig. 6K). Thus, the Sb_mut HA vaccine elicited a protective immune response in ferrets, despite being an animal model where the B-cell antigenic immunodominance hierarchy was suboptimal for our vaccine design.

Discussion

Influenza vaccines capable of providing broader or more durable protection would considerably decrease the public health burden of these viruses. Vaccines that increase HA stalk-directed responses while simultaneously retaining neutralizing head responses represent one approach to potentially accomplish that goal. In this work, we developed an HA antigenic mixture-based vaccine with four positions of the Sb antigenic site randomized within the backbone of an H1 HA. We showed that the Sb_mut HA vaccine elicited a stronger overall antibody response, likely with more and higher affinity antibodies, while concurrently altering the relative amounts of the response directed against the HA stalk and head. The Sb_mut HA post-vaccination serum samples showed better reactivity against HAs of H1N1 virus strains spanning 82 years, as well as heterosubtypic group 1 (non-H1) HAs. Functionally, the Sb_mut HA increased HAI, neutralization and ADCC activities against the homologous strain and increased ADCC activity against a heterologous, homosubtypic strain after prime vaccination. As a result, Sb_mut HA provided better protection during both homologous and heterologous challenge studies compared with WT HA.

Although the immune response is clearly altered by vaccinating with an antigenically complex mixture compared with a “clonal” WT HA, many questions as to the mechanisms underlying this differential response remain. One of the most prominent remaining questions is to understand the basis for the observed head:stalk antibody ratio alterations. It has been previously reported that exposure to divergent HAs can increase frequencies of B cells targeting conserved and sub-immunodominant epitopes, and therefore elicit broadly neutralizing antibodies (39–42). Similarly, recent studies show that heterotypic antigens can preferentially engage B cells with cross-reactive B cell receptors when distinct antigens are present (43–46). It is therefore possible that, because the sub-immunodominant stalk domain is constant across different trimers while the immunodominant Sb site is highly variable, we may be tapping into similar immunological phenomena. It is important to note, however, that many of the aforementioned studies analyze vaccine responses in the context of pre-existing immunity to influenza viruses. As our experiments were performed in influenza naïve animals, it is also possible that our head:stalk antibody ratio phenotype may not be driven by the antigenic divergence per se, but rather fundamental changes to the immunogenicity of different HA domains. It is also worth noting that although we are using the head and stalk domains as general surrogates of variable and conserved domains, we may also be eliciting more antibodies to conserved, sub-dominant regions of the HA head. More targeted mapping of sub-domains in the HA, as well as testing the hypothesized mechanisms underlying altered antibody ratios, remain important areas of future study.

Another open question is which antigenic sites are the “best” locations for the introduction of antigenic complexity. The Sb site is generally immunodominant on the H1 HA head in both humans and mice and much of neutralizing antibody response induced by vaccination or infection is thought to bind to this site (29–31). When we introduced antigenic variability into this site and administered the vaccine to mice, we observed increased avidity in head-directed antibodies, suggesting that antigenic complexity in the most immunodominant location may be affecting the initiation of germinal center reactions with our antigenic mixture. In ferrets, the Sb site is not the dominant antigenic site (29), and we failed to see the same enhanced head-directed responses in that system. The differences between mice and ferrets in inducing higher quality head-directed antibodies suggests that the interplay between the immunodominance hierarchy of mutated locations and links to the immunogenicity of antigenic mixture should be studied further.

Despite the head antibody avidity differences across different experimental systems, the antigenic mixtures were uniformly successful in inducing more stalk-directed responses. With respect to antigenic library composition, targeting additional antigenic sites in combination with Sb has the potential to enhance epitope re-focusing effect and increase responses to other sites in the HA such as the stalk. However, the protection from infection may not always be correlated with decreasing classical antigenic site-directed responses. It is also possible that the specific nature of the mutation in an antigenic site (e.g. number of amino acids mutated) may be important. Further, our antigenic mixture may also contain some partially misfolded or less “tight” trimers, which may expose additional protective epitopes, including those in the trimer interface (25, 47, 48). In the future, structural analysis of trimers in the library and more precise mapping of the HA-reactive antibody footprints elicited by the antigenically complex vaccine will help to answer these questions. Additionally, the minimum HA mutant numbers required to change the canonical immunodominance hierarchy remains undefined. Although we detected 86,835 unique clones in our HA antigenic library, we cannot assess currently if the full complement of clones is required. In fact, it may be the case that more or less complexity is ultimately more beneficial compared with our observed vaccination phenotypes.

There are also several limitations to the current study which are important to note. As mentioned above, our vaccine was only evaluated in naïve animals; how it functions in animals with pre-existing immunity to influenza virus has not been studied. This is an important caveat, as essentially every non-infant human has preexisting immunity to influenza viruses and thus, pre-immune individuals would represent the major target population for a Sb_mut HA vaccine. Additionally, pre-existing immunity could affect not only how the immune system recognizes an antigenically complex vaccine, but immune imprinting to a non-vaccine-matched HA subtype could also limit the magnitude of the antibody response, as has been observed for some vaccines and after infection (49–51). It is also important to note that we only performed studies with a single H1 HA, whether this approach is generalizable to other HAs (either homo- or hetero-subtypic) remains to be seen. Finally, the mechanisms of protection of our vaccine also remain incompletely defined. Although protection from the vaccine-matched strain was almost certainly mediated predominantly by improved head antibody binding and classical antibody effector functions such as virus neutralization, we also saw increased protection from disease after heterologous challenge when viral titers were not changed. We found that the Sb_mut HA vaccine induced increased serum ADCC activity and accordingly, we also observed higher titers of IgG2 isotypes against WT full-length HA were elicited by Sb_mut HA vaccine. Although both cross-group stalk antibodies and conserved head epitopes, including trimer interface antibodies, can function by interaction with Fc receptor (37, 38, 52, 53), additional studies mapping the binding epitopes of the ADCC-mediating antibodies and defining their contributions to protection against heterologous viral challenge will be informative.

In conclusion, we developed an antigenically complex HA vaccine that increase the responses to the conserved stalk domain while maintaining and sometimes improving head-directed antibody responses. This altered antibody response led to improved protection from challenge with homologous and heterologous influenza strains compared with a WT HA vaccine approaches and thus represents a broader and potentially more durable vaccine at least for group 1 HA subtype viruses. Expansion of this approach to group 2 HA subtype viruses, such as H3N2 viruses, as well as influenza B viruses could, in combination, represent a more broadly-protective influenza vaccine. Moreover, generating antigenically complex protein mixtures potentially represents a strategy to redirect immune responses to protective but sub-immunodominant epitopes in other pathogens and may lead to new classes of vaccines with enhanced efficacy in the future.

Materials and Methods

Study design

This study was to designed to generate and evaluate an improved influenza A virus vaccine based on the hypothesis that the immune response can be directed to conserved but less immunodominant epitopes of HA if the immunodominant epitopes were highly variable. Vaccine efficacy was evaluated in mice and ferrets by morbidity, mortality, viral load, lung pathology, and clinical scores. ELISAs to measure antibody binding and antibody functional assays, including HAI assays, viral neutralization assays and ADCC reporter assays, were also used. In all experiments, animals were assigned randomly to each experimental group. The HA binding Luminex assay was performed by blinded investigators, whereas other experiments were not conducted in a blinded manner. No exclusion criteria were applied for any data except the passive transfer data and ferret clinical disease data. In the mouse passive transfer experiment, one mouse in Sb_mut HA group was excluded because there was insufficient immune serum available to administer a full dose. The ferret challenge data shown are derived only from experiments in which the viral infection induced clinical disease. All serological assays were performed at least twice with samples from different animal experiments, except for the HA binding custom luminex assay. All animal experiments were repeated. Library DNA sequencing and protein mass spectrometry analysis were performed twice with two sets of samples from different preparations. Statistical analysis was done independently by R.S. using the methods described in figure legends and no statistical power calculations were used to inform group sample size prior to experimentation.

Library construction and purification

To make a HA mutant library with antigenic site Sb mutated, four positions in Sb site of PR8 HA (GenBank: CY083950.1) including Q192, N193, Q196, and E198 (H3 numbering, based on a previous study (54)) were chosen for mutagenesis. The fragment was cloned to pCMV-IRES-GFP version 3 (pCIG3, Addgene, 78264) lentiviral cloning vector, then lentivirus was packaged. Lentivirus titers on 293T cells (American Type Culture Collection, ATCC; CRL-3216) were determined by flow cytometry for GFP expression. Samples were acquired on FACSCanto II using FACSDiva software (BD Bioscience), data were analyzed with FlowJo (Tree Star). To sort viable mutants, 293T cells were transduced with lentivirus at MOI of 0.3 and properly folded mutants were isolated by MACS with monoclonal 6F12 (10) (a kind gift from Dr. Peter Palese) and anti-mouse IgG MicroBeads (Miltenyi Biotec, 130048401). To check the purity of the mutant HA libraries, cells were stained with 6F12 followed by an allophycocyanin (APC)-conjugated anti-mouse antibody (Invitrogen, A865) and analyzed by flow cytometry as described above.

DNA library sequencing and analysis

The DNA fragments containing mutated positions were amplified and sequencing was performed on a NextSeq 500 device (Illumina) using NextSeq 500 mid Output kit (150 bp paired end). Geneious (version 2023.0) was used to process sequencing reads for analysis. This included trimming low quality ends using the BBDuk Geneious plugin, followed by mapping the trimmed reads to their respective reference sequence. Next, a custom Python script was used to generate a matrix of amino acid frequencies at each position. Data visualization was performed with R (version 4.2.2) using ggplot2 and sequence logo plots were generated using the ggseqlogo R package. All scripts used in this analysis have been made publicly available at Zenodo (DOI: 10.5281/zenodo.10810877).

Recombinant protein expression and purification

For proteins used in animal vaccination, the ectodomains of viable mutant or WT HAs (or mCherry) were cloned into a mammalian expression vector containing a TEV protease cleavage site and a GCN4pII trimerization domain with a His tag and Ser-Gly linkers at the C terminus. The expression vectors were transiently transfected into Expi293F cells (Gibco, A14527) using ExpiFectamine293 Transfection Kit (Gibco, A14635) following the manufacturer’s instruction. Protein was collected on day 3 post transfection and was purified by chromatography on Ni-NTA agarose (Qiagen, 30250) and then was dialyzed against phosphate-buffered saline (PBS) using Slide-A-Lyzer dialysis cassettes (Thermo Scientific). Proteins were quantified by bicinchoninic acid (BCA) assay (Thermo Scientific, A53226). Normalized proteins were visualized by SDS-PAGE gel to check normalization and purity. The PR8 full length HA, PR8 HA head, and PR8 HA stalk proteins used in ELISAs were fused to a T4 fibritin foldon to avoid measuring false-positive cross reactivity toward GCN4pII used in the vaccine constructs. These designs were based on PR8 HA (GenBank: CY083950.1). Amino acids numbering start at the initiator methionine and all numbers are based on the full-length HA. The PR8 full length HA construct was PR8 HA amino acids 1–530. The PR8 HA head construct had an artificial signal peptide (MKAILVVLLYTFATANAGS) preceding PR8 HA amino acids 59–291. The PR8 HA stalk construct was based on the Gen6 HA-SS design generated in a previous study with PR8 HA amino acids 518–566 removed (23). To adapt the design to the PR8 HA sequence, HA between L49 and L328 was replaced with a GSG linker and HA between M402 to T436 was replaced with a GSGGSG linker. Four additional substitutions (K394M, Y437D, N438L, E446L) in HA were introduced in Gen6 HA-SS (relative to the full-length A/New Caledonia/20/1999 HA) to promote correct folding of the stalk. The proteins were expressed, purified, and quantified as described above.

SDS-PAGE electrophoresis and deglycosylation of recombinant HA

Samples mixed with 2x Laemmli samples buffer (Bio-Rad, 1610737) containing 5% 2-mercaptoethanol and incubated at 95°C for 5 min. Treated samples were loaded onto Mini-PROTEAN TGX precast gels (Bio-rad) for electrophoresis. Gels were stained with SimplyBlue SafeStain (Invitrogen, 465034) and were imaged on ChemiDoc MP Imaging System (Bio-Rad). To check glycosylation of recombinant proteins, WT HA and Sbmut HA were treated with PNGase F (New England Biolabs, P0704S) following the manufacturer’s instruction under denaturing reaction conditions. Separation of reaction products were visualized by SDS-PAGE electrophoresis.

Protein library LC-MS/MS mass spectrometry analysis

Samples in PBS were brought to 4% SDS, reduced with 10 mM dithiothreitol (DTT) for 20 min at 55°C, alkylated with 25mM iodoacetamide for 45 min at room temperature and then subjected to S-trap (Protifi) trypsin digestion using manufacturer recommended protocols. Digested peptides were lyophilized to dryness and resuspended in 50 μL of 0.2% formic acid/2% acetonitrile. 1 μg of each sample was subjected to chromatographic separation on a Waters MClass ultra-performance liquid chromatography (UPLC) equipped with a 1.7 μm high strength silica (HSS) T3 C18 75 μm inside diameter (I.D.) × 150 mm reversed-phase column (Waters). The mobile phase consisted of (A) 0.1% formic acid in water and (B) 0.1% formic acid in acetonitrile. 3 μL was injected and peptides were trapped for 3 min on a 5 μm Symmetry C18 180 μm I.D. × 20 mm column at 5 μl/min in 99.9% A. The analytical column was then switched in-line and a linear elution gradient of 5% B to 40% B was performed over 90 min at 400 nL/min. The analytical column was connected to an Exploris 480 Orbitrap mass spectrometer (Thermo Scientific) through an electrospray interface operating in a data-independent acquisition (DIA) mode. The instrument was set to acquire a precursor MS scan from m/z 390–1010 at R=120,000 (target AGC 1000%, max IT 60 ms) followed by 75 × 8 m/z wide DIA windows collected from m/z 404.4337 to 1000.7047 at r=15,000 (target AGC 1000%, max IT 20 ms) at higher collisional dissociation (HCD) energy of 30%. Raw data files were processed through DIA_NN (https://github.com/vdemichev/DiaNN), open-source DIA proteomics processing software. Default settings were used and searched against a custom HA database containing each possible combination of amino acid residues within four hypervariable points within a single peptide and as well as additional contaminant proteins. The automated precursor matrix report that contains the normalized intensity for each peptide/precursor was used for further data analysis. Peptides that spanned the variable residue locations of 9, 10, 13 or 15 within “HPPNSKEQXXLYXNXNAYVSVVTSNYNRRFTPEIAERPKVRDQAGRMNY” were identified. For each variable location, the peptide intensity was binned for each amino acid. The binned total intensity values were used to describe the expression at each variable location.

ELISAs

96-well microplates were coated with 200 ng of antigen per well in sodium carbonate-bicarbonate buffer and then blocked with 1% bovine serum albumin (BSA) in PBS at room temperature for 2 hours. Serial dilutions of serum samples were added into each well and incubated at 4°C overnight. After incubation of horseradish peroxidase (HRP)-coupled second antibody at room temperature for 1 hour, color was developed by using tetra-methyl-benzidine (TMB) substrate (Thermo Scientific, 34029), and reactions were stopped with 2 M sulfuric acid. Absorbance was measured at 450 nm on a Varioskan Lux microplate reader (Thermo Scientific). Avidity ELISA was performed using ammonium thiocyanate (NH₄SCN) displacement as described previously, with some modifications (55). 96-well microplates were coated as described above followed by incubation of diluted serum samples. Plates were then incubated with 0–8 M NH₄SCN for 25 min. Bound antibodies were detected as described above. The results were expressed as the molar concentration required to displace 50% of the bound antibodies. Most ELISA and antibody functionality prime and prime-boost experiments were conducted with serum samples from the same experiment; however serum samples from a replicate experiments were sometimes used if there was insufficient serum. Specifically, in the following locations, the prime and boost samples shown were collected from independent replicate experiments: Fig. 2B to F, Fig. 3B and C, Fig. 5A to C, and fig. S3C.

Monoclonal antibodies used in determination of antigenicity of recombinant proteins including Y8–10C2 (Sa site), Y8–1C1 (Sb site), H37–80 (Ca site), H2–4C2 (Cb site) were kindly gifted by Dr. Scott Hensley. CR9114 was purchased from Creative Biolabs (PABL-593). In IgG isotyping, goat anti-mouse IgG1-HRP, IgG2b-HRP, IgG2c-HRP (Southern Biotech, 1071-05, 1-91-05, 1078-05) in 1:5000 dilution were used as second antibodies. To detect serum and other monoclonal antibodies, goat anti-human IgG-HRP (Invitrogen, A18805) in 1:5000 dilution, goat anti-mouse IgG-HRP (Invitrogen, A16072) in 1:10000 dilution or goat anti-ferret IgG-HRP (Novus Biologicals, NB7224) in 1:5000 dilution were used as second antibodies.

Viruses and group 1 HAs

Viruses used in mouse experiments and antibody functionality analyses were propagated in embryonated chicken eggs and were sequenced and described previously (56). The LD₅₀ of the PR8 virus used in C57BL/6 mice is approximately 50 PFU. H1N1 viruses used in ELISAs were propagated in embryonated chicken eggs and were sequenced and described previously (57). Influenza A virus A/California/07/2009 ((H1N1)pdm09 Antiviral Resistance (AVR) - Reference Virus M2: S31N NA: wild type (WT), FR-458) used in the ferret experiment was obtained through the Influenza Reagent Resource, Influenza Division, World Health Organization Collaborating Center for Surveillance, Epidemiology and Control of Influenza, Centers for Disease Control and Prevention. Group 1 HA with T4 foldon including HA of A/Vietnam/1203/2004 (H5) (GISAID: EPI25595), A/Japan/305/1957 (H2), A/Taiwan/2/2013 (H6) and A/Hong Kong/33982/2009 (H9) (plasmids were provided by Dr. Scott Hensley) were purified by Protein Production Facility Core, Duke Human Vaccine Institute.

Immunization, passive transfer, and challenge

Mice were purchased from Jackson Laboratory. For immunization experiments, six- to ten-week-old C57BL/6 female mice were administered 10 μg of recombinant protein intramuscularly in 100 μL of 50% (v/v) mixture of AddaVax (InvivoGen, vac-adx-10). For the passive transfer experiment, mice received 200 μL of pooled primed serum intraperitoneally from each group and were challenged 24 hours later. Data from mice that received insufficient serum due to limited availability were excluded. For infection, mice were administered 40 μL of virus intranasally after anesthesia with a ketamine-xylazine mixture (12,500 PFU of PR8 or 24,000 PFU of Cal/09 for challenge after prime vaccination, 160,000 PFU of PR8 or 24,000 PFU of Cal/09 for challenge after prime-boost vaccination, 500 PFU of PR8 for challenge after passive transfer). Mice were weighed daily and euthanized once their body weight reached <75% of the starting weight as a humane endpoint. Euthanasia was performed using CO₂ as the primary method, and a bilateral thoracotomy was performed as the secondary method. Ferrets were purchased from Marshall Bioresources. For immunization experiments, six-month-old male ferret were administered 10 μg of recombinant protein intramuscularly in 250 μL of 50% (v/v) mixture of AddaVax (InvivoGen, vac-adx-10). For infection, ferrets were administered 500 μL of virus (10⁷ PFU of A/California/07/2009, intranasally after inhalation of isoflurane). The ferret disease data shown in Fig. 6 are from the only replicate experiment in which the infection successfully induced clinical disease; the serology experiments are representative of all experiments performed. Animals were housed at ambient temperature and humidity with free access to food and water in ABSL-2 laboratory on 12 hours:12 hours light:dark cycle. All procedures were performed according to protocols (mouse experiment: A142-21-07, ferret experiment: A077-23-03) approved by the Duke University IACUC.

Hemagglutination Inhibition Assays (HAI)

Serum samples were treated with receptor-destroying enzyme (RDE II, Denka Seiken, 370013) at a 1:4 dilution at 37 °C for 20 hours followed by inactivation at 56 °C for 30 min and further dilution to 1:10 with PBS. Samples were two-fold serially diluted in v-bottom microtiter plates. Virus adjusted to 4 HA units in 25 μL was added to each well. The plates were incubated at room temperature for 15 min followed by the addition of a 50 μL of 0.5% chicken (for PR8) or turkey (for Cal/09) erythrocytes (both from Lampire Biologicals). The reaction mixture was settled for 30 min at room temperature. Wells were examined visually for inhibition of HA. HAI titers are reported as the reciprocal of the highest dilution of serum that completely inhibited HA.

Micro neutralization assay

Virus neutralization assays were performed as described previously with some modifications (45). Serum samples were serial diluted in 96-well microplates. Virus suspensions (1500 PFU of PR8 or 900 PFU of Cal/09) were added to each well, and the plates were incubated at 37°C for 1 hour. Following incubation, 1.5 × 10⁴ suspended MDCK cells (ATCC, CRL-2936) were added to each well, and the microplates were incubated at 37°C in an incubator with 5% CO₂ for 20 hours. Cells were then fixed by 4% paraformaldehyde (PFA) in PBS and stained for HA protein with CR9114 in 1:4000 dilution at 4°C overnight and goat anti-human IgG-HRP in 1:5000 dilution at room temperature for 1 hour. Color was developed by TMB and absorbance was measured as described in ELISA section above. When calculating percent of viral neutralization, control wells of virus alone and medium alone were constrained as 0% and 100%, respectively. A four-parameter nonlinear regression was generated by Prism 9 (GraphPad) and half-maximal inhibitory concentrations were obtained based on the curve.

ADCC reporter assay

The ADCC reporter assays were performed using a Mouse FcγRIV ADCC Bioassay followed the manufacturer’s instruction (Promega, M1215). Briefly, 96-well white tissue culture plate (Perkin Elmer, 6005680) were seeded with 1.25 × 10⁴ MDCK cells in 100 μL per well. After 16 to 20 hours of incubation, cells were infected with 5 MOI of virus and incubated at 37°C for 20 hours. Then the medium was replaced with 25 μL per well of assay buffer (RPMI-1640 with 4% (V/V) low IgG fetal bovine serum). Then 25 μL of serial dilutions of samples was added to each well and incubated at 37°C for 30 min. Then mouse FcγRIV effector cells at a concentration of 7.5 × 10⁴ in 25 μL were added to each well and incubated at 37°C for 6 to 10 hours. A volume of 75 μL of Bio-Glo reagent was added to each well and luciferase was measured by a EnSpire 2300 Multilabel Reader (Perkin Elmer). For analysis, serial dilutions of serum samples were tested and the lowest serum dilution in the linear range of the assay was shown and used for statistical analysis.

Lung histology

Mouse lungs tissues from the indicated treatment groups were fixed in 4% PFA/PBS at 4°C for more than 48 hours. Samples were embedded in paraffin and sectioned after dehydration and wax immersion. Slides were stained with hematoxylin and eosin (H&E) and then were scanned by HistoWiz. Uninfected control images are representative of one or two independent animals.

Plaque assay

Viral titers in lungs were determined using a standard plaque assay protocol on MDCK cells. Lung homogenates was serially diluted, and was incubated with the cells for 1 hour at 37°C. Then supernatant was removed and a 1% agar overlay containing TPCK-trypsin was applied. After incubation at 37°C for 48 to 72 hours, cells were fixed in 4% PFA for 3 hours. Polyclonal PR8 virus-specific antibody from WT PR8- or WT Cal/09-infected mice and goat anti-mouse IgG-HRP antibody were used to stain plaques. Plaques were visualized by TrueBlue peroxidase substrate (SeraCare, 5510–0030) and were manually counted.

HA binding custom multiplex assay

Soluble ectodomain of HA of the influenza strains (table S2) used in the assay was cloned into pFastBac expression constructs containing a T4-fibitin trimerization domain and His-tag as described previously (58). Recombinant protein was purified from sf9 cells (CRL-1711) by GenScript. BSA (Sigma, negative control) or recombinant HA were conjugated to magnetic microspheres (Luminex Corp) by carbodiimide coupling using Luminex’s recommended protocol. Recombinant HA- and BSA-coated microspheres (1500 of each) were mixed with samples diluted 1:100 in PBS plus 1% BSA pH 7.4 (assay buffer; Life Technologies), incubated for 1 hour at room temperature on an orbital shaker and then washed in assay buffer. Microspheres were incubated with 4 μg/mL phycoerythrin (PE)-conjugated polyclonal goat anti-mouse Ig (Southern Biotech, 1010–09) or PE-conjugated goat anti-ferret IgG (Abcam, ab112768) for 30 minutes, washed in assay buffer, and the PE median fluorescence intensity (MFI) of each microsphere population was measured using a Bio-Plex 200 analyzer (BioRad). Data were analyzed using Bio-Plex Manager v.6.2 and are reported as background-corrected MFIs from duplicate wells. Mean of each group was used to generate graphs.

RT-qPCR

Total RNA was extracted from ferret nasal wash with QIAamp Viral RNA Kits (QIAGEN, 52906). Eukaryotic 18S rRNA (Applied Biosystems, 4319413E) was used as the endogenous control. TaqMan probes targeting viral RNA of A/California/07/2009 NP gene (forward primer: CCTGGAACTGAGAAGCAGATAC, reverse primer: GAATGTAGGCTGCACACTGA, Probe: /56-FAM/AGGACCAGG/ZENAGTGGAGGAAATACCA/3IABkFQ/) were synthesized by Integrated DNA Technologies. EXPRESS One-Step Superscript qRT-PCR kit (Invitrogen, 11781200) was used to perform one-step RT-qPCR on the Applied Biosystems QuantStudio 3 Real-Time PCR System. Target and internal control were reverse transcribed at 50°C for 30 min, followed by 50 cycles of amplification at 95°C for 15 sec, 58°C for 24 sec and 72°C for 24 sec.

Statistical Analysis

Individual-level data are presented in data file S2. In this study, all statistical analyses were performed using either Prism 9 (GraphPad Software) or R statistical software (R Foundation for Statistical Computing). Pairwise comparisons of two groups were performed using Mann-Whitney U tests while comparisons of three groups were performed using Kruskal-Wallis test followed by a Wilcoxon rank-sum test with Bonferroni correction, and comparisons of survival curves were performed using log-rank tests, as indicated in each figure legend. All tests were performed at an alpha level of 0.05.

Supplementary Material

List of the Supplementary Materials

Figs. S1 to S6

Tables S1 and S2

Data files S1 and S2

Data and materials availability:

All data associated with this study are in the paper or supplementary materials. Materials can be requested through an MTA with Duke University. Raw data and analyzed data of DNA library sequencing have been deposited to the NCBI GEO database (accession number GSE252606). All computer code used in the analysis of the DNA library sequencing data has been deposited to Zenodo (DOI: 10.5281/zenodo.10810877; https://zenodo.org/records/10810878). The mass spectrometry proteomics data of protein library have been deposited to the ProteomeXchange Consortium through the PRIDE (59) partner repository with the dataset identifier PXD048318.