Ferret model to mimic the sequential exposure of humans to historical H3N2 influenza viruses

Abstract

Mutations accumulate in influenza A virus proteins, especially in the main epitopes on the virus surface glycoprotein hemagglutinin (HA). For influenza A(H3N2) viruses, in particular, the antigenicity of their HA has altered since their emergence in 1968, requiring changes of vaccine strains every few years. Most adults have been exposed to several antigenically divergent H3N2 viruses through infection and/or vaccination, and those exposures affect the immune responses of those individuals. However, animal models reflecting this ‘immune history’ in humans are lacking and naïve animals are generally used for vaccination and virus challenge studies. Here, we describe a ferret model to mimic the serial exposure of humans to antigenically different historical H3HA proteins. In this model, ferrets were sequentially immunized with adjuvanted recombinant H3HA proteins from two or three different H3HA antigenic clusters in chronological order, and serum neutralizing antibody titers were examined against the homologous virus and viruses from different antigenic clusters. For ferrets immunized with a single HA antigen, serum neutralizing antibody titers were elevated specifically against the homologous virus. However, after immunization with the second or third antigenically distinct HA antigen in chronological order, the ferrets showed an increase in more broadly cross-reactive neutralizing titers against the antigenically distinct viruses and against the homologous virus. Sequentially immunized animals challenged with an antigenically advanced H3N2 virus showed attenuated virus growth and less body temperature increase compared with naïve animals. These results suggest that sequential exposure to antigenically different HAs elicits broader neutralizing activity in sera and enhances immune responses against more antigenically distinct viruses Our findings may partly explain why adults who have been exposed to antigenically divergent HAs are less likely to be infected with influenza virus and have severe symptoms than children.

Introduction

Influenza A viruses evolve rapidly with the accumulation of mutations especially in their hemagglutinin (HA) surface glycoprotein, which contains the main epitopes recognized by host humoral immunity. This results in antigenic alteration of the viruses, or ‘antigenic drift’, which, in turn, leads to virus escape from host adaptive immune responses. Influenza A(H3N2) viruses first emerged in human populations in 1968. The HA of the circulating H3N2 strains remain antigenically similar for a couple of years, forming ‘antigenic clusters’, and then antigenic drifts occur¹. Since these antigenic alterations or ‘cluster transitions’ result in antigenic escape of the viruses from the immune memories in the human population, which are shaped in response to previous infection and/or vaccination, candidate vaccine virus strains for seasonal vaccines are updated periodically to match the circulating strains. Most adults are considered to have been exposed to several antigenically distinct H3N2 viruses through vaccination, infection, or both. Pre-exposure to influenza HAs is imprinted as host adaptive immunity and affects immune responses to subsequent infections with more recent viruses and antibody responses to later vaccinations^2-8. As such, pre-immunity in the human population to influenza viruses may complicate the evaluation of seasonal vaccine efficacy. To better understand influenza pre-immunity in humans and its effects on vaccine efficacy or immune responses to virus infection, numerous studies have developed influenza pre-immune animal models utilizing ferrets, which are the best-established animal species for influenza virus infection and transmission studies, and serological analyses including antigenic monitoring of the circulating virus strains^9,10. In most of the previous H3N2 pre-immune ferret models, the animals mount immune memory through active virus infection rather than vaccination, and serum antibody titers were analyzed in in vitro assays that focused on anti-HA antibodies (i.e., hemagglutination inhibition assays and/or neutralization assays, or ELISA)^10-13. In an H3N2 virus infection model utilizing ferrets, the phenomenon of “original antigenic sin”, in which higher serum titers are elicited against the first virus variant encountered and much lower titers are induced against later infecting viruses, was observed after chronologically sequential H3N2 virus infections, similar to that seen in humans, suggesting the usability of ferret pre-immune models¹¹. However, these models in which pre-immunity is mounted in response to active virus infection have the potential to include immune responses against other viral proteins when the pre-immune animals are subjected to in vivo virus challenge. Therefore, there is a need for animal models that solely focus on the effects of repeat exposure to historical HAs on virus replication and/or clinical symptoms in infected individuals upon subsequent exposure to the virus. Here, we established a ferret model that reflects the sequential exposure in the human population to antigenically distinct H3HAs through immunization with adjuvanted recombinant HA proteins. Using this animal model, we then evaluated virus replication and body temperature increase upon challenge with antigenically distinct, chronologically recent H3N2 virus.

Materials and Methods

Cells.

Madin-Darby Canine Kidney (MDCK) cells and hCK cells¹⁴ were maintained in Eagle’s minimal essential medium (MEM) containing 5% newborn calf serum. Human embryonic kidney 293T cells were cultured in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum (FBS). SIAT1-MDCK cells were propagated in the presence of 1 mg/ml geneticin (G418; Invivogen) in DMEM containing 5% FBS and antibiotics. These cell lines were incubated at 37 °C with 5% CO₂. Expi293F cells (ThermoFisher Scientific) were maintained in Expi293 expression media (ThermoFisher Scientific) in spinner flasks, at a speed of 120 rpm, at 37 °C under 8% CO₂. All cell lines were regularly tested for mycoplasma contamination by using PCR and were confirmed to be mycoplasma-free.

Viruses.

The H3N2 influenza viruses A/Netherlands/399/1993 (Wuhan 1995 cluster; WU95), A/Netherlands/312/2003 (Fujian 2002 cluster; FU02), A/Perth/16/2009 (Perth 2009 cluster; PE09), and A/Wisconsin/15/2009 (PE09) were kindly provided by Dr. Ron A M Fouchier at Erasmus Medical Centre and were propagated in MDCK cells. A/Kansas/14/2017_PR8HY (clade 3c3a1) was generated by reverse genetics¹⁵ with its HA and NA segments from the homologous virus and the other segments from the PR8 high-yield backbone¹⁶. None of the viruses used in this study were isolated or amplified in chicken embryonic eggs.

Recombinant protein expression and purification.

Soluble-form recombinant HAs (rHAs) consisting of the HA signal peptide and ectodomain (amino acid residues 1–505; H3 numbering) with stabilizing mutations to form disulfide bonds (T30C and Q376C), a T4 foldon trimerization domain, and a hexa-histidine tag at the C-terminus (Supplementary Figure 1A)¹⁷ were cloned into pCAGGS plasmids. Proteins were expressed in Expi293F cells (Thermo Fisher Scientific) and purified by using TALON metal affinity resin (TaKaRa Clontech).

Micro-neutralization assay.

Virus neutralizing antibodies against the H3N2 influenza viruses A/Netherlands/399/1993, A/Netherlands/312/2003, A/Perth/16/2009, A/Wisconsin/15/2009, and A/Kansas/14/2017_PR8HY were evaluated in serum samples. Microneutralization assays in SIAT1-MDCK cells were utilized instead of hemagglutination inhibition assays for serum analysis throughout this study, because of the lower hemagglutination efficiency of recent H3N2 viruses compared with older strains due to different sialic acid receptor-binding preference¹⁸. Serum samples were treated with receptor-destroying enzyme (RDE; Denka seiken) at 37 °C for 20 h, inactivated at 56 °C for 1 h, and diluted 1:10 in PBS. Two-fold serial dilutions of sera were prepared in MEM, and each dilution was incubated for 1 h at room temperature with the same volume of virus diluent (100 TCID₅₀/50 μL) containing 1 μg/mL TPCK-trypsin. The serum/virus mixture was added to 100% confluent SIAT1-MDCK cells that were plated one day prior in 96-well plates. The cells were incubated for 5 days at 33 °C and then cytopathic effect was microscopically detected by eye. Virus neutralization titers were determined as the reciprocal of the highest serum dilution that completely prevented cytopathic effects. Each sample was analyzed in duplicate for the determination of geometric mean titers.

Animal experiments.

All experiments with ferrets were performed in accordance with the guidelines set by the Institutional Animal Care and Use Committee at the University of Wisconsin-Madison. The protocol was approved by the Animal Care and Use Committee of the University of Wisconsin-Madison (protocol numbers V00806 and V6426).

Immunization and virus challenge of ferrets.

Ferrets (4–6-month-old females; Triple F farm) were confirmed to be seronegative in hemagglutination inhibition assays with turkey red blood cells (seasonal H1N1, pandemic H1N1, and Flu B Victoria lineage) or neutralization assays (H3N2) with recent circulating influenza viruses before being immunized. Animals were intramuscularly immunized with 5, 15, or 45 μg of purified rHA protein (diluted in 100 μL of PBS) adjuvanted with AddaVax (100 μL; InvivoGen) or with Alhydrogel (2% solution, 100 μL; InvivoGen). For all sequential immunizations, the antigen was injected in the right shoulder–right brachial muscle of the ferrets, split into 2–3 sites to reduce the injection volume/site. Three different rHA doses (5, 15, or 45 μg) were tested due to potential dose-dependent responses. The groups were combined for the analysis since no statistically significant difference was seen in serum titers, clinical symptoms, or nasal swab titers, with the different antigen amounts. The serum titer data (Figure 2) and the clinical symptoms upon virus challenge (Supplementary Fig. 4) are shown with differentiated symbols (×, immunized with 5 μg of rHA; +, immunized with 15 μg of rHA; *, immunized with 45 μg of rHA) and gray shading (light gray, immunized with 5 μg of rHA; gray, immunized with 15 μg of rHA; dark gray, immunized with 45 μg of rHA) respectively, depending on the rHA amount used for immunization. Animals were serially bled via the jugular vein under anesthesia by ketamine and dextomitor. Then, the animals were intranasally inoculated with 10⁶ plaque-forming units (pfu) of A/Kansas/14/2017_PR8HY virus under anesthesia. Nasal swabs were taken daily under anesthesia, alternately from the left and right nostrils, for 8 days. During this period, the body temperature was measured by using an implantable subcutaneous temperature transponder (Bio Medic Data Systems), and their body weights were also measured. Nasal swabs were resuspended in 1 mL of MEM containing antibiotics and kept at −80°C until titration by plaque assays in hCK cells.

Figure 2. Serum reactivity of serially immunized ferrets against viruses from different antigenic clusters.

The FU02-PE09 ferret group (A) and the WU95-FU02-PE09 ferret group (B) were serially immunized with rHAs from the FU02 and PE09 clusters, or with rHAs from the WU95, FU02, and PE09 clusters, respectively. All ferrets were bled 3 weeks after each immunization and at the time of virus challenge (pre-challenge). Neutralizing antibody titers of their sera against A/Netherlands/399/1993 (Neth399; WU95 cluster), A/Netherlands/312/2003 (Neth312; FU02 cluster), A/Perth/16/2009 (Perth; PE09 cluster), A/Wisconsin/15/2009 (Wis15; PE09 cluster), and A/Kansas/14/2017_PR8HY (Kansas; clade 3c3a) were analyzed in a micro-neutralization assay. Dots and bars show the values of individual animals and medians of the groups, respectively. Animals were immunized with 5, 15, or 45 μg of purified rHA protein; data from these groups were combined for the analysis because no statistically significant difference was seen in serum titers with the different antigen amounts (animals immunized with 5 μg of rHA (×), 15 μg of rHA (+), or 45 μg of rHA (*)). Data for each animal are shown as the geometric mean of duplicates. Dashed lines represent the detection limit of the assay. Statistical analysis is described and shown in Supplementary Figure 2.

Antibody titer plots.

Mean neutralizing titer for each group of sera against each antigen was computed using a Bayesian statistical model that correctly handles censored (i.e. less than 10) values. Neutralizing antibody titers, t, were transformed as y = log₂(t/10), and then standardized to have a combined mean of 0 and standard deviation of 1 prior to inference. Titer i from group g, y_ig, was modelled as y_ig ~ Norm(μ_g, σ). The following priors were used: σ ~ Exp(1); μ_g ~ Norm(μ_μ, σ_μ); μ_μ ~ Norm(0, 1); σ_μ ~ Exp(1). The likelihood for censored observations was taken from the cumulative Normal probability density function. PyMC was used to run four chains of 1,000 samples which converged well. X-axis positions of viruses were based on distances between viruses in the Fonville et al.-basemap⁸. When a virus was not present the centroid of its corresponding antigenic cluster was used instead. Clade 3c3a1 viruses are not in the Fonville et al.-basemap, so the mean distance between consecutive antigenic clusters in the basemap was used to represent the PE09–3c3a1 distance.

Results

rHA antigen preparation to mount H3HA immune history in naïve ferrets

The aim of this study is to establish an animal model that reflects the immune history induced by sequential exposure to historical H3N2 influenza viruses that is seen in the human population. To specifically focus on the effects of HAs on pre-immunity, without any contributions from other viral proteins, we immunized ferrets with recombinant HA (rHA) proteins rather than inactivated influenza whole virion vaccine or active virus infection. To retain the posttranslational modifications (i.e., glycosylation) on the HAs of the human seasonal viruses, which may affect the antigenicity of the viruses^19,20, soluble-form rHA proteins were expressed in human-derived cell line Expi293F cells, purified (Supplementary Figure 1), and used as antigens.

Specific neutralizing antibodies are induced by a single HA antigen and broader neutralizing antibodies are elicited by further serial immunization

To mimic infections with H3N2 influenza viruses several times in humans, two groups of ferrets were sequentially immunized with H3 rHAs of viruses from 2 or 3 different antigenic clusters in chronological order (Figure 1). For the immunization with the first antigen, a prime-boost regimen with AddaVax was carried out (Figure 1B). In a study conducted in parallel, we found that Alhydrogel induced higher neutralizing titers even after a single immunization, and therefore used this adjuvant for the immunization with the second and third antigens (Figure 1B; see Discussion for more information).

Figure 1. Ferret model reflecting sequential exposure to historical H3HAs.

(A) Timeline of ferret immunization with recombinant HAs (rHAs) from different H3HA antigenic clusters, Wuhan 1995 (WU95), Fujian 2002 (FU02), and Perth 2009 (PE09). (B) Immunization regimens. Each group of ferrets was immunized with the first antigen twice 9 weeks apart, followed by a single immunization with the second or third antigen.

One group of ferrets (group ‘FU02-PE09’) was intramuscularly immunized with the first antigen from the rHA of A/Netherlands/312/2003 virus (Neth312; FU02 cluster), and then 12 weeks later, immunized with the second antigen, rHA of A/Wisconsin/15/2009 virus (Wis15; PE09 cluster). Sera at 3 weeks post-immunization were analyzed in a microneutralization assay to examine the reactivity against H3N2 viruses from different antigenic clusters (Figure 2A). After the immunization with the first antigen, neutralization titers against the homologous virus increased, whereas the titers against a virus from an older antigenic cluster (WU95) and against viruses from more recent clusters (PE09 and clade 3c3a1) remained at background levels (Figure 2A, middle left panel, 3 wk post-FU02 Imm.). After immunization with the second antigen Wis15 virus rHA (PE09), the titers significantly increased against two viruses in the homologous antigenic cluster [A/Perth/16/2009 (Perth) virus and Wis15 virus] (Figure 2A, middle right panel, 3 wk post-PE09 Imm.). Of note, the neutralization titers against the more recent A/Kansas/14/2017 (Kansas) virus from clade 3c3a1 were also significantly increased after immunization with the second antigen, which was not observed after immunization with the first antigen. The serum median titers against Neth312 (FU02) virus, the antigen to which the animals were first exposed, were also slightly elevated although the difference was not statistically significant. The titers against A/Netherlands/399/1993 (Neth399) virus (WU95 cluster), an older virus to which the ferrets had not been exposed, remained at background levels. Thirty-five weeks after the second exposure, the serum titers waned against the homologous viruses that were the source of the first and second antigens, whereas the titers against the Kansas virus did not significantly change (Figure 2A, right panel, Pre-challenge).

Another group of ferrets (group ‘WU95-FU02-PE09’) was immunized first with rHA of Neth399 virus (WU95), then with Neth312 virus (FU02) rHA, and thirdly with Wis15 virus (PE09) rHA (Figure 1). After the first immunization with Neth399 (WU95) rHA, the neutralization titers increased against the homologous virus (Figure 2B, middle left panel, 3 wk post-WU95 Imm.), whereas only background level titers were seen against viruses from other antigenic clusters, consistent with our observations of group FU02-PE09 (Figure 2A). After the immunization with the second antigen Neth312 (FU02), the serum neutralizing titers against the homologous virus were significantly elevated, as anticipated; the titers against Perth and Wis15 viruses from the more recent antigenic cluster PE09 and the titers against the Kansas virus from clade 3c3a1 also increased, although to a lesser extent than against the homologous virus (Figure 2B, middle right panel, 3 wk post-FU02 Imm.). Finally, after the third immunization [with Wis15 virus rHA (PE09)], the serum neutralization titers were further elevated against the viruses in the homologous antigenic cluster, Perth and Wis15 viruses (Figure 2B, right panel, 3 wk post-PE09 Imm.). Notably, the serum titers against the Kansas virus also significantly increased after the immunization with the third antigen PE09 rHA. The neutralizing titers against Neth312 virus (FU02), the antigen for the second exposure, also increased slightly, although the difference between before and after PE09 rHA immunization was not statistically significant.

The serum neutralizing antibody titers of the pre-immune ferrets were visualized based on the relative HA antigenic distance (Figure 3). The antibody titer plots obtained in this pre-immune ferret model show similar trends to those observed in hemagglutination inhibition (HAI) titer landscapes of human individuals with infection and vaccination history⁸. First, a “back-boost” titer increase on the preexisting landscape was observed upon the second or third immunizations, and second, the back-boost titer increase was larger against antigenically closer strains. Notably, the serum neutralization titer fold-change detected in this ferret model was comparable to that observed with human sera after either virus infection or vaccination (typically 1- to 32-fold), although there was a larger deviation with human sera. These findings underscore the potential utility of the ferret pre-immune model.

Figure 3. Neutralizing antibody titer changes following serial immunization of ferrets.

The antibody titer plots were generated based on the neutralization titers shown in Figure 2. The titers against the two PE09 cluster viruses (Perth and Wis15 viruses) were combined. The gray shading in the top panels represents the naïve background titers, whereas the gray shading in the other panels shows the heights of pre-immunization. The titer increase post-immunization compared to the previous naïve/immunization status is shown in red for the FU02-PE09 ferret group (left) and blue for the WU95-FU02-PE09 ferret group (right). The black circles and gray circles on the x-axis indicate the antigens used for the most recent immunization and the previous immunization, respectively. The scale bar denotes 2 antigenic units (AU; 1 unit = a 2-fold change in HAI titer). Statistical analysis is described and shown in Supplementary Figure 2.

H3HA pre-immunity affects virus replication and clinical symptoms in ferrets

Finally, all of the immunized ferrets and a group of naïve animals were challenged with the antigenically advanced A/Kansas/14/2017 virus (H3N2; clade 3c3a1). The ferrets were intranasally inoculated with 10⁶ pfu/animal of the Kansas virus. Nasal swab samples were collected daily for 8 days. The naïve control animals showed infectious virus titers in their nasal swabs in the range of 10⁵ to 10³ pfu/swab on Day 1 through Day 6, eliminating the virus by Day 7 (Figure 4A, Supplementary Figure 4A). The FU02-PE09 pre-immunized ferrets showed comparably high virus titers on Days 1 to 4, but significantly lower titers compared to the naïve animals on Days 5 and 6 (Figure 4A, Supplementary Figure 4B). The WU95-FU02-PE09 pre-immunized ferrets showed statistically significantly lower virus titers (1.5–2 log lower) than those of the naïve animals on all sampling days except Day 4 (Figure 4A, Supplementary Figure 4C). Body temperature and body weight were monitored after the challenge. Naïve animals showed a >2 °C body temperature elevation that peaked on Day 2 (Figure 4B, Supplementary Figure 4D). The FU02-PE09 ferrets showed a 1-degree milder mean peak temperature, which returned to normal on Day 3 (Figure 4B, Supplementary Figure 4E). The WU95-FU02-PE09 ferrets did not show any obvious body temperature changes for the 8-day post-infection observation period (Figure 4B, Supplementary Figure 4F).

Figure 4. Virus replication and clinical symptoms in ferrets with immune history.

Naïve ferrets (N=3) and the groups of pre-immunized ferrets (FU02-PE09, N=9; WU95-FU02-PE09, N=8) were intranasally infected with 10⁶ pfu of A/Kansas/14/2017_PR8HY virus. Virus titers in nasal swabs (A), and body temperature (B) and body weight (C), as clinical symptoms, were monitored for 8 days post-challenge. Data represent the means + SD of each group. Statistical analyses were performed by using a one-way analysis of variance (ANOVA) and corrected for multi-group comparison by using Dunnett’s test. (*P<0.05; **P<0.01; ***P<0.001; ****P<0.0001). Naïve control group animals were confirmed to be seronegative pre-challenge (Supplementary Figure 3). Data of individual animals are shown in Supplementary Figure 4.

The body weights of the naïve ferrets decreased by 5% for 3 days post-challenge and showed the most severe loss on Day 6 (Figure 4B, Supplementary Figure 4G). The FU02-PE09 ferrets showed slightly more severe weight loss than the naïve group at the earlier timepoints (Figure 4B, Supplementary Figure 4H). The WU95-FU02-PE09 ferret group lost weight comparably to the naïve group until Day 5, at which point they recovered faster than the other groups (Figure 4B, Supplementary Figure 4I). Collectively, the groups of animals with serial pre-immunizations with historical H3HAs showed reduced virus replication and clinical symptoms upon challenge with an antigenically distinct recent H3N2 virus compared to naïve animals with the exception of body weight loss in the FU02-PE09 group. These data suggest that previous exposures to H3HAs attenuate antigenically more advanced H3N2 virus infection.

Discussion

In this study, we present a ferret pre-immune model that mimics serial exposure to antigenically divergent H3N2 influenza viruses in the human population, focusing on HAs alone through immunization with adjuvanted recombinant HAs. In this model, we focused on serum neutralizing antibody titers to evaluate the mounted pre-immunity, which are proven to correlate with protective efficacy against virus infection^21,22 and have been used to estimate the efficacy of seasonal influenza vaccines.

In our model, sequential immunizations with antigenically distinct historical HAs in chronological order resulted in different trends in neutralizing antibody titers between the initial immunization with a single HA antigen and the subsequent second and third HA antigen immunizations. Upon the initial immunization with rHA (i.e., the first exposure to H3HA for the naïve ferrets), the neutralizing antibody titer was specifically increased against the homologous virus, without cross-reactive neutralizing titers raised against viruses from other antigenic clusters (Figure 2A, middle-left panel; Figure 2B, middle-left panel). However, further immunization(s) with HA from later antigenic clusters (i.e., following exposures to more recent H3HAs) led not only to neutralizing antibodies being elicited against the homologous virus, including a back-boosting titer rise against the virus of the first immunization (albeit without a significant difference), but also against antigenically further advanced virus strains to which the animals had never been exposed (Figure 2A, middle-right panel; Figure 2B, middle-right panel). These observations are consistent with the serum HAI and neutralizing titer changes shaped by infection and/or vaccination histories in humans^6,8. Back-boosting describes the phenomenon in which exposure to second or further antigens increase titers to a previously encountered antigen(s)⁸. Consistent with this definition, the antibody titer rise was only observed for HAs to which the animals were previously exposed (Figure 2A middle-right), while no back-boost was observed against antigens that had not been used for immunization, which also mirrors the phenomenon seen in humans; that is, individuals produce cross-reactive antibodies against historical strains to which they have been infected in childhood^23,24. Collectively, these results suggest that this model could be used to mimic immune history in the human population.

We challenged the pre-immune ferrets with antigenically distinct, chronologically recent H3N2 (Kansas; 3c3a1) influenza virus. In the animals pre-immunized with H3HAs from the antigenic clusters WU95, FU02, and PE09, Kansas virus replication in the upper respiratory tract was significantly reduced and the animals did not show an increase in body temperature. In the animals with pre-immunity against WU95 and FU02 HA, which had lower neutralization titers against the Kansas virus than the 3-pre-immunization group, virus replication was also limited at later timepoints (Days 5 and 6 post-challenge), and earlier body temperature recovery was observed in these animals than in the naïve group, although the difference was less pronounced than with the 3-pre-immunization group. Of note, the changes in virus replication and body temperature were solely caused by pre-immunity to HAs without any effects of other viral proteins (i.e., NP, NA, and M2 proteins), which is a key feature of this model.

This pre-immune ferret model has other strengths and advantages over existing models. By generating rHA antigens in human-derived cell lines, the glycosylations of our rHAs should be more similar to the HAs on virions than on those generated in other expression systems such as yeast or insect cells or antigens extracted from egg-grown viruses. Also, sequential immunizations with rHAs can be given within a relatively short time period compared with pre-immune models based on active virus infection, where intervals of several months are needed between immunizations, because the immune responses induced by the earlier infection tend to reduce efficient virus replication in animals in the following infections, which may result in only limited immune history being mounted in the later immunization.

We utilized adjuvants in our model to effectively induce neutralizing antibodies. AddaVax was used in the immunization with the first antigen, and we detected specific neutralizing antibodies only after two immunizations with the same antigen without cross-reactive titers against viruses from other antigenic clusters. We also compared the effects of three different adjuvants, AddaVax, Quil-A, and Alhydrogel, on neutralizing antibody titers in the serial immunization ferret model as an independent study (Supplementary Figure 5), and found that Alhydrogel, the typical Th2 response-inducing adjuvant, most effectively raised the neutralizing antibody titers after a single immunization. To focus on neutralizing antibody titer outcomes, in this study, Alhydrogel was used for the subsequent immunizations with the second and third antigens. Although the mechanism of action for some adjuvants is not fully clarified, increasing evidence suggests that the direction of the adaptive immune responses (namely, the balance between Th1 and Th2 cell differentiation and CD8 T cell responses) is determined by the adjuvants and by the immunization routes, since activation of different innate immune receptors and signaling pathways by different adjuvants results in a different spectrum of cytokines to differentiate T cells and B cells²⁵. The pre-immune model we propose here might, therefore, be applicable to other animal models to mimic exposures to different pathogens, through immunization with different adjuvants, reflecting the immune responses induced by natural infection.

Virus replication and body temperature rise upon virus challenge were significantly lower in the group immunized with three antigenically distinct HAs than in the group immunized with two different HAs. It is hard to compare the two immunization groups side-by-side since the intervals between the immunizations and virus challenge are not identical between the groups (3 wks vs 35 wks) because the study was disrupted by the coronavirus pandemic; the shorter interval might have non-specifically and additionally contributed to the anti-viral state upon virus challenge. Nevertheless, the results suggest that higher protective effects are elicited by additional exposures to HAs, resulting in reduced virus growth and severity of disease. Notably, in the human population, children, especially young children, are at higher risk than adults of both seasonal influenza attack rate and influenza-associated respiratory hospitalization^26-31. Virus shedding from infected adults is shorter than from infected children, and older children tend to shed virus for a shorter period than younger children³². Our findings suggest that repetitive exposures to HAs in human adults throughout their lives contribute to protective immune responses against the viruses.

The ferret model is invaluable for transmission readouts of virus, as well as pathogenesis assessments. Previous studies have shown that human influenza viruses are transmitted via the air from the upper respiratory tract of infected ferrets^33,34 Also, another study showed that in a direct-contact transmission model, immunization of donor ferrets with seasonal influenza vaccine prior to homologous H3N2 virus challenge shortens the virus shedding period from the animals, which results in delayed transmission to naïve contact animals, compared to between naïve donor and naïve contact animal pairs, although the contact animals were not protected from virus transmission³⁵. The reduced virus replication in the nasal epithelium and faster virus elimination observed in the pre-immune ferrets in this study could contribute to a reduced chance and/or delay in transmission, when the donor animals have pre-immunity.

The main limitation of this study is that only neutralization titers were analyzed as pre-immunity mounted to HAs. Although serum neutralizing antibody titers have been proven to correlate with protective efficacy against influenza viruses^21,22 and are the primary method used to evaluate seasonal influenza vaccine efficacy in humans, increasing numbers of studies have shown that antibodies that do not show neutralizing activity in vitro may contribute to protective efficacy in vivo, mainly through Fc-FcγR engagement^36-38. In addition, the cellular arm of adaptive immunity has been shown to reduce the severity of influenza virus infection in mouse models^39,40. Further studies are needed to establish animal models that dissect the entire scope of the immune responses to historical influenza HAs in the human population.

A PR8-HY reassortant virus was used for virus challenge in this study, because we previously found that H3N2 viruses with PR8-HY internal genes replicate efficiently in the respiratory organs of Syrian hamsters⁴¹. Replication efficiency and pathogenicity of 3c3a1 viruses with wild-type internal genes were not compared with those of viruses with PR8-HY internal genes side-by-side in ferrets; although the comparison between the different pre-immunization groups in this study should not be compromised, different statistical significance could be observed in clinical symptoms with the virus with wild-type internal genes.

Also, we did not include a homologous control immunization group in this study, side-by-side with the pre-immune groups with heterologous immunization. The rationale for this was that in a previous study, a group of ferrets that received the 2019–2020 Fluzone quadrivalent seasonal influenza vaccine twice followed by homologous Kansas/14/2017 virus challenge, eliminated the virus in nasal washes as early as 3 days post-infection³⁵, whereas the pre-immune ferrets in this study were shedding virus in nasal swabs until 6 days post-infection (Figure 4A). Although the results from different studies cannot be compared because of different experimental conditions, such as infection dose and internal virus genes, earlier virus elimination than that seen with the pre-immune groups would be anticipated if a homologous immunization group had been included in this study.

Lastly, post-challenge serum titers were not analyzed, because, in this study, we focused on serum titer changes with serial immunization with rHA and their effects on clinical results upon virus challenge.

In conclusion, sequential immunization of ferrets with adjuvanted recombinant HA may be a useful model to mimic influenza serology in the human population.