Pan-Influenza A Protection by Prime–Boost Vaccination with Cold-Adapted Live-Attenuated Influenza Vaccine in a Mouse Model

Influenza virus infections continually pose a major public health threat with seasonal epidemics and sporadic pandemics worldwide. While currently licensed influenza vaccines provide only strain-specific protection, antigenic drift and shift occasionally render the viruses resistant to the host immune responses, which highlight the need for a vaccine that provides broad protection against multiple subtypes. In this study, we suggest a vaccination strategy using cold-adapted, live attenuated influenza vaccines (CAIVs) to provide a broad, potent, and safe cross-protection covering antigenically distinct hemagglutinin (HA) groups 1 and 2 influenza viruses. Using a mouse model, we tested different prime–boost combinations of CAIVs for their ability to induce humoral and T-cell responses, and protective efficacy against H1 and H5 (HA group 1) as well as H3 and H7 (HA group 2) influenza viruses. Notably, even in the absence of antibody-mediated neutralizing activity or HA inhibitory activity in vitro, CAIVs provided a potent protection against heterologous and heterosubtypic lethal challenges in vivo. Heterologous combination of prime (H1)–boost (H5) vaccine strains showed the most potent cross-protection efficacy. In vivo depletion experiments demonstrated not only that T cells and natural killer cells contributed to the cross-protection, but also the involvement of antibody-dependent mechanisms for the cross-protection. Vaccination-induced antibodies did not enhance the infectivity of heterologous viruses, and prime vaccination did not interfere with neutralizing antibody generation by the boost vaccination, allaying vaccine safety concerns associated with heterogeneity between the vaccines and challenge strains. Our data show that CAIV-based strategy can serve as a simple but powerful option for developing a “truly” universal influenza vaccine providing pan-influenza A protection, which has not been achieved yet by other vaccine strategies. The promising results of potency, breadth, and safety demonstrated in the mouse model support further studies in higher animal models for clinical relevance.

Influenza virus infections continually pose a major public health threat with seasonal epidemics and sporadic pandemics worldwide. While currently licensed influenza vaccines provide only strain-specific protection, antigenic drift and shift occasionally render the viruses resistant to the host immune responses, which highlight the need for a vaccine that provides broad protection against multiple subtypes. In this study, we suggest a vaccination strategy using cold-adapted, live attenuated influenza vaccines (CAIVs) to provide a broad, potent, and safe cross-protection covering antigenically distinct hemagglutinin (HA) groups 1 and 2 influenza viruses. Using a mouse model, we tested different prime-boost combinations of CAIVs for their ability to induce humoral and T-cell responses, and protective efficacy against H1 and H5 (HA group 1) as well as H3 and H7 (HA group 2) influenza viruses. Notably, even in the absence of antibody-mediated neutralizing activity or HA inhibitory activity in vitro, CAIVs provided a potent protection against heterologous and heterosubtypic lethal challenges in vivo. Heterologous combination of prime (H1)-boost (H5) vaccine strains showed the most potent cross-protection efficacy. In vivo depletion experiments demonstrated not only that T cells and natural killer cells contributed to the cross-protection, but also the involvement of antibody-dependent mechanisms for the cross-protection. Vaccination-induced antibodies did not enhance the infectivity of heterologous viruses, and prime vaccination did not interfere with neutralizing antibody generation by the boost vaccination, allaying vaccine safety concerns associated with heterogeneity between the vaccines and challenge strains. Our data show that CAIV-based strategy can serve as a simple but powerful option for developing a "truly" universal influenza vaccine providing pan-influenza A protection, which has not been achieved yet by other vaccine strategies. The promising results of potency, breadth, and safety demonstrated in the mouse model support further studies in higher animal models for clinical relevance.
Keywords: influenza virus, cold-adapted live-attenuated vaccine, cross-protection, T cell, nK cell, antibody Pan-Influenza A Protection by CAIVs Frontiers in Immunology | www.frontiersin.org February 2018 | Volume 9 | Article 116 inTrODUcTiOn Influenza virus is an important respiratory pathogen that causes annual epidemics and occasional pandemics. In each season, the influenza epidemic results in 3-5 million cases of severe illness and 250,000-500,000 deaths worldwide (1). Due to the antigenic diversity and variability of the virus, an influenza vaccine has to be updated almost every year to match circulating strains (2). Currently used influenza vaccines provide only strain-specific protection, primarily by inducing neutralizing antibodies against surface glycoproteins, hemagglutinin (HA) and neuraminidase (NA) of the virus (2). In addition, the occurrence of influenza pandemics were often accompanied by zoonotic spillovers of the surface genes into the humaninfecting viruses (3), rendering preexisting vaccines ineffective to newly emerging viruses. In the last decade, a significant breakthrough has been made in the development of a universal influenza vaccine, triggered by the discovery of rare antibodies specific to the immunogenically subdominant but conserved stalk domain of the influenza HA.
To redirect the host immune responses from the HA globular head domain toward this conserved stalk domain, rational vaccine designs were heralded, such as headless HA and chimeric HA vaccines (2,4). The HA stalk-based approaches have been successful in inducing a broader protection than preexisting influenza vaccines. However, concerns were also raised from a practical standpoint, including the low protective efficacy against different HA group viruses, the necessity of multiple vaccinations to achieve sufficient protective efficacy, and the rare cases of adverse effects such as viral infectivity-enhancing activity of the HA stalk antibodies (5)(6)(7). Furthermore, recent studies have isolated mutant influenza viruses each showing resistance to a particular HA stalk-specific antibody (8)(9)(10). Meanwhile, a recent study has shown that the chimeric HA strategy can be extended to cold-adapted live-attenuated influenza vaccine (CAIV), in which a prime-boost vaccination with CAIV/split-virus vaccines provided superior protection against the pandemic H1N1 infection compared to two doses of split-virus vaccination in ferret model (11). This study is currently under a clinical trial to examine the feasibility of the strategy in humans (https://clinicaltrials.gov/ct2/ show/NCT03300050).
Despite the well-documented cross-protective efficacy, CAIVs remain relatively unexplored in the field of universal influenza vaccine development, most likely due to their inefficiency in inducing systemic antibody responses and the difficulties in genetic engineering to expose the HA stalk to the host immune system. On the other hand, CAIVs have many immunological advantages in terms of cross-protection over the other vaccine platforms, including the delivery of a whole set of antigens, the induction of mucosal IgA antibodies, and T-cell responses, as well as the stimulation of innate immunity (12)(13)(14)(15). Perhaps most importantly, it has been widely acknowledged that T-cell immunity, which targets viral proteins that are relatively conserved between different influenza strains, is the key to the crossprotection by natural infection or vaccination (16,17). Therefore, if all of these factors are combined properly and reinforced by a rational vaccination strategy, a CAIV is expected to serve as a powerful platform for a universal influenza vaccine. To date, many studies have reported the development and evaluation of CAIVs targeting homologous strains and, in a subset of those studies, antigenically closely related strains, addressing the issue of cross-protection (15). However, very few studies have addressed in detail the potential of CAIVs as a reliable platform for a universal influenza vaccine that provides the desired levels of breadth, potency, and safety of cross-protection against diverse influenza subtypes.
In this study, we hypothesized that prime and boost vaccinations with CAIVs would further stimulate the immunological correlates and thus display potent and broad cross-protection efficacy. To test this hypothesis, we designed prime-boost vaccinations with different strains of X-31ca-based CAIVs. Distinct to A/Ann Arbor/6/60ca (H2N2) strain that has been used as a donor strain for currently licensed type A CAIVs, the X-31ca was derived from the parent X-31 virus, a reassortant carrying the HA and NA genes of A/Hong Kong/1/1968 (H3N2) under A/Puerto Rico/8/34 (H1N1) backbone (18). We included two different boosting groups − homosubtypic but heterologous boosting (different strains within H1 subtypes) and heterosubtypic boosting (different strains carrying different HA subtypes) − to evaluate and compare their immune responses and protective efficacy against those of single or homologous boosting group. Mice were vaccinated with various combinations of prime-boost CAIVs, and vaccination-induced antibody and T-cell responses as well as protection efficacy were assessed against antigenically distant HA group 1 (H1 and H5) and group 2 influenza viruses (H3 and H7), which are closely associated with both seasonal epidemics and pandemics worldwide. Additionally, we examined whether our strategy accompanied any adverse effects such as vaccination-associated enhanced respiratory disease (VAERD) or antigenic sin-like phenomenon (19,20), to address potential safety issues related to heterogeneity between the vaccine and the challenge virus. Promising results of the breadth, potency, and safety of the CAIV-based vaccination strategy support its further development into a reliable universal influenza vaccine technology for clinical use.

elisa for antibody Titration
Vaccination-induced antibodies specific to the whole virus or the HA proteins were estimated by ELISA. Ninety-six-well plates were coated with 100 µl of 10 5 PFU of sucrose-gradient purified viruses or 1 µg/ml of the HA proteins overnight. After blocking with 150 µl of 1% BSA in PBS and washing the plates, the wells were incubated with 100 µl of twofold serial dilutions of sera or BALF for 1 h at RT. After washing the plates, the wells were incubated with 100 µl of HRP-conjugated secondary goat anti-mouse IgG antibodies or IgA antibodies for 1 h at RT. The plates were washed and supplemented with 100 µl of TMB solution and incubated for 30 min at RT in the dark. The reaction was stopped by the addition of 50 µl of 2 N H2SO4 solution, and OD450 was measured on an ELISA reader. Antibody titers were calculated as the endpoint dilution that yielded an OD value greater than the mean + 2 SD of the control group.

na inhibition assay
To determine the standard virus titer for NA inhibition (NAI) assay, viral NA activity was measured as described elsewhere (27 fetuin-coated plates and subjected to the subsequent procedures described above to measure the NA activity.

antibody-Dependent cell-Mediated cytotoxicity assay
Vaccination-induced sera antibodies were examined for their ability to mediate antibody-dependent cell-mediated cytotoxicity (ADCC) activity against virus-infected MDCK cells. We used primary spleen cells isolated from naive mice instead of natural killer (NK) cells as effector cells in ADCC assay, since pure NK cells isolated from the spleen demonstrated very weak ADCC activity in the preliminary study. Confluent MDCK cells grown in 96-well plates were infected with two multiplicities of infection of the influenza virus and incubated in MEM supplemented with 20 µM of Z-VAD-FMK (Promega), pan-caspase inhibitor, and in the absence of trypsin to minimize cytotoxicity from the multicycle viral infection. Eight hours after the infection, the supernatant was removed and 100 µl of sera (1:10 dilution) was added to the MDCK cells and incubated for 1 h at 37°C. After the incubation, 100 µl of mouse spleen cells (10 6 cells) pre-treated with RBC lysis buffer were added to each well and incubated for 2 h at 37°C. The cytotoxicity of the target MDCK cells was measured in triplicate by lactate dehydrogenase (LDH) release assay.

nP-Based Virus infectivity assay
To examine whether vaccination-induced sera antibodies promote the viral infection of heterologous influenza viruses, NP-based ELISA method (6) was used with minor modifications. Confluent MDCK or RAW264.7 cells grown in a 96-well plate were infected with 100 TCID50 of the virus. Twenty-four hours later, the infected cells were washed and fixed by an acetone fixative and 100 µl of anti-influenza A/WSN/33 (H1N1) NP antibodies (26) were added to each well and incubated for 1 h at RT. After the binding of secondary antibodies and washings, OD490 was measured by an ELISA reader. The virus infectivity was measured in triplicate and expressed as % of control infection with non-vaccination immune sera.

In Vivo Depletion of T cells or nK cells
For the depletion of CD4+ T cells and CD8+ T cells in vivo, 200 µg of anti-CD8 mAb (clone 2.43; BioXcell) and anti-CD4 mAb (clone GK1.5; BioXcell) were injected intraperitoneally into mice four times at days 1, 3, 5, and 7 before challenge. Non-depleted control mice were given isotype control rat IgG2b antibodies (clone LTF-2, BioXcell). Blood and lungs were taken from the mice 24 h after the last antibody injection and were subjected to flow cytometry to confirm the depletion. For flow cytometric analysis, we used anti-CD8 mAb (clone 53-6.7; Biolegend) and anti-CD4 mAb (clone RM4-5; Biolegend) directed against different epitope of CD4 and CD8 molecules to that of the depleting antibodies. For the depletion of NK cells, 20 µl of anti-asialo GM1 antiserum (Wako Pure Chemical Industries) were injected intraperitoneally into mice four times at days 1, 3, 5, and 7 before challenge. Non-depleted control mice were given normal rabbit serum (Wako Pure Chemical Industries). The spleens were taken from the mice 24 h after the last antibody injection to confirm the depletion by flow cytometry. For flow cytometric analysis, we used anti-CD3 mAb (clone 17A2; Biolegend) and anti-CD49b mAb (clone DX5; Biolegend).

statistical analysis
All data were expressed as mean ± SD of each cohort, and the difference comparison between two groups was conducted by the unpaired, two-tailed Student's t-test. The difference was considered statistically significant when P values were <0.05 (***P < 0.001; **P < 0.01; *P < 0.05).

resUlTs experimental Design of Vaccination and challenge in Mice
We previously developed X-31ca as a master strain for CIAVs, by the cold-adaptation of the parent X-31 virus (21,22). The X-31ca was used as a reliable backbone for CAIVs for the 2009 pandemic A/Korea/1/09 (H1N1) virus (ca-pH1N1), seasonal influenza A/New Caledonia/20/99 (H1N1) virus (ca-NCH1N1), and highly pathogenic H5N1 avian influenza A/Indonesia/ 5/05 (H5N1) virus (ca-IDH5N1) in subsequent studies (23-25, 28, 29). Using these three CAIVs, four different combinations of prime-boost vaccinations were designed to evaluate and compare induced immune responses and protective efficacy against antigenically distant heterologous influenza viruses in a mouse model. ca-pH1N1 was intranasaly inoculated into the single vaccination group and also used as prime strains in three boosting vaccination groups. Two weeks after the prime vaccination with ca-pH1N1, three different CAIVs were used as boosting strains: ca-pH1N1 for homologous boosting (identical to the prime strain), ca-NCH1N1 for heterologous boosting (different strain of H1N1), and ca-IDH5N1 for heterosubtypic boosting ( Figure 1A). Cross-reactive antibody responses elicited by each vaccination were analyzed through various in vitro assays as described later. In addition, to evaluate cross-protection in vivo, the following four heterologous influenza viruses were used as challenge strains: PR8 (H1N1) and MA81 (H5N2)  belonging to the HA group 1 and Phil82 (H3N2) and reNet03 (H7N1) belonging to the HA group 2 ( Figure 1B). Thus, while three CAIVs had identical six internal gene segments derived from the X-31ca and the surface genes from different viruses belonging to the HA group 1 (H1 and H5), challenge viruses encompass both HA groups. This experimental design allowed us to assess the potency and breadth of immune responses and protective efficacy against diverse influenza viruses covering both HA groups.

cross-reactive antibody responses induced by Vaccination
Considering the roles of the HA as a major protective antigen in the protection against influenza viral infection, we first examined whether vaccination-induced sera antibodies bind to the HA proteins of heterologous influenza viruses with ELISA using the insect cell-expressed recombinant HA proteins derived from seven different influenza viruses. All the vaccinations elicited high levels of antibodies cross-reactive to antigenically close group 1 HAs including H1, H2, and H5 (Figure 2A). The robustness of the cross-reactivity of the antibody responses extended to antigenically distant H9 HA, although the antibody titers were significantly reduced as compared to those to the closer HAs. Specific antibody binding was observed even in group 2 HAs (H3 and H7), although noticeable antibody binding seen only in sera from heterologous boosting vaccinations (ca-pH1N1 + ca-NCH1N1) (Figure 2A). Next, we estimated vaccination-induced antibody titers directed to the conserved HA stalk domain. For this purpose, we expressed recombinant HA full-length and stalk proteins derived from A/Korea/1/09 (pH1N1) and PR8 (H1N1) viruses, as soluble proteins fused with the E. coli lysyl tRNA synthetase (LysRS) ( Figure S1A in Supplementary Material). In ELISA, vaccination-induced antisera strongly bound to the fusion proteins, but not to the LysRS, enabling the measurement of HA-specific antibody titers ( Figure S2B in Supplementary Material). The results also suggest that the sequential epitopes present in the HA proteins expressed in E. coli can be recognized by specific antibodies. A single vaccination with ca-pH1N1 resulted in a pH1N1-HA-specific antibody titer of 100, and boosting by ca-pH1N1, ca-NCH1N1, and ca-IDH5N1 increased the antibody titers to 210, 160, and 150, respectively ( Figure 2B). The most significant increase in the HA-specific antibodies compared to single vaccination was seen in homologous boosting with ca-pH1N1, while heterologous boosting with ca-NCH1N1 and heterosubtypic boosting with ca-IDH5N1 resulted in only modest increases in the HA-specific antibody titers. Vaccinations also resulted in pH1N1-HA stalk-specific antibodies. A single vaccination yielded the HA stalk-specific antibody titer of 192, and each boosting increased the antibody titers to 250−520.
In parallel, the cross-reactivity of the antibodies was examined against the PR8 (H1N1) HA protein. A single vaccination induced the PR8 HA antibody titer of 56, and each boosting resulted in a significant increase in the antibody titers ranging 210−260, an approximate fourfold increase of the single vaccination. The PR8 HA stalk-specific antibody levels were very similar between the single vaccination and homologous boosting groups (mean antibody titer of 384 in both). However, heterologous boosting and heterosubtypic boosting resulted in more than a twofold increase in the HA stalk-specific antibody titer compared to that of homologous boosting. The results suggest that heterologous or heterosubtypic boosting could be more effective way of inducing the HA stalk-specific antibodies than the homologous boosting. We also estimated the NA-specific antibodies induced by the vaccination. All the vaccinations generated high levels of antibodies reactive to closely related N1 NA proteins, and the antibody binding reactivity became weak against heterologous N2, N6, N7, N8, and N9 NA proteins ( Figure S2 in Supplementary Material). Additionally, vaccination-induced anti-influenza antibodies were measured using purified whole viruses as coating antigens in ELISA. A single vaccination with ca-pH1N1 induced high levels of serum IgG antibodies against not only homologous pH1N1 but also four heterologous influenza viruses, PR8 (H1N1), MA81 (H5N2), Phil82 (H3N2), and reNet03 (H7N1) ( Figure S3A in Supplementary Material). Boosting by homologous, heterologous, or heterosubtypic CAIV induced antibody levels similar to, or greater than, that of the single vaccination. IgA antibody levels in the brochoalveolar lavage fluid (BALF) were also estimated against the same set of viruses. Similar to the results of serum IgG antibodies, BALF IgA antibodies    Figure 3A). However, none of vaccinationinduced antibodies yielded detectable levels of MN antibodies against the four heterologous influenza viruses. HI assay with the sera showed the same trend as the MN assay results; sera antibodies could inhibit the hemagglutination activity of the pH1N1 but not the heterologous influenza viruses ( Figure 3B). Additionally, we performed an NAI assay with the same set of viruses to examine whether the sea antibodies could block the enzyme activity of the NAs of the heterologous influenza viruses. As expected, homologous boosting yielded the highest level of NAI antibodies against the homologous pH1N1, and heterologous or heterosubtypic boosting resulted in the levels of NAI antibodies similar to that of the single vaccination ( Figure S4 in Supplementary Material). However, none of the vaccination-induced antibodies demonstrated NAI activities that were greater than 50% against any of the heterologous viral NAs, with only partial inhibition seen against the heterologous viruses at the least diluted sera. Recent studies have provided considerable evidence that influenza-specific antibodies have Fc-effector functions, such as ADCC, which target the infected cells for killing by immune cells such as NK cells, monocytes, neutrophils, and macrophages (30)(31)(32). We examined whether the vaccination-induced antibodies exert such ADCC activity to MDCK cells infected with the heterologous influenza viruses. All sera from the four vaccination groups resulted in a significant increase in the cytotoxicity of PR8 (H1N1)-infected cells, as compared to the non-vaccination Pan-Influenza A Protection by CAIVs Frontiers in Immunology | www.frontiersin.org February 2018 | Volume 9 | Article 116 control ( Figure 3C). Additionally, sera from the three boosting vaccination groups led to the lysis of MA81 (H5N2)-infected or reNet03 (H7N1)-infected cells. Significant lysis of Phil82 (H3N2)-infected cells was seen by sera from heterologous boosting and heterosubtypic boosting vaccination groups. These results suggest that although the antibodies induced by CAIVs could not neutralize heterologous influenza viruses, they exerted a weak but distinct ADCC activity against the heterologous viruses.

cross-Protection against heterologous lethal challenges In Vivo
There has been substantial experimental and clinical evidence that CAIVs provide varying degrees of cross-protection against heterologous influenza viruses even without detectable antibody-mediated neutralizing activities (28,33,34).
To evaluate the breadth and potency of this cross-protection in vivo, vaccinated mice were challenged with a lethal dose of each heterologous influenza virus. Non-vaccinated control mice infected with 10 MLD50 of PR8 (H1N1), MA81 (H5N2), Phil82 (H3N2), or reNet03 (H7N1) virus showed rapid weight loss and death upon the challenge ( Figure 4A). By contrast, a single vaccination with ca-pH1N1 protected the mice from mortality associated with lethal infections, causing mild weight loss of 5−8% upon challenge with MA81 (H5N2) or reNet03 (H7N1), and less than 1% upon challenge with PR8 (H1N1) or Phil82 (H3N2) (Figure 4B). Homologous boosting also resulted in excellent protection against PR8 (H1N1) and Phil82 (H3N2) challenge and resulted in the weight loss of 8.9% following reNet03 (H7N1) challenge, a comparable protective efficacy to that from a single vaccination ( Figure 4C). Interestingly, however, MA81 (H5N2) challenge caused weight loss of 12.9% in the homologous boosting group, which was higher than in the single vaccination. Heterologous boosting with ca-NCH1N1 protected the mice from the morbidity associated with the challenges, except for the temporal weight loss of 5.8% at 2 days postchallenge (dpc) with MA81 (H5N2) ( Figure 4D). Remarkably, heterosubtypic boosting with ca-IDH5N1 did not resulted in any weight loss after the challenges, showing the most potent cross-protection efficacy among the four vaccination groups ( Figure 4E). The virulence represented by weight loss are summarized in Figure 4F. Evidently, challenge with PR8 (H1N1) or Phil82 (H3N2) caused only mild weight loss of less than 3% in all vaccination groups, making it difficult to discriminate the crossprotective efficacy between prime-boost combinations. On the other hand, challenge with MA81 (H5N2) or reNet03 (H7N1) yielded variable weight loss ranging 0−13% depending on the vaccination combination. Homologous boosting demonstrated seemingly lessened protective potency against MA81 (H5N2) or reNet03 (H7N1) as compared to the single vaccination, but the potency was strengthen by heterologous boosting and became complete by heterosubtypic boosting causing weight loss less than 1% upon the challenge. Protective efficacy was further substantiated by estimating challenge virus titers in the lungs of vaccinated mice. All the challenge viruses replicated to greater than 10 6 PFU in the lungs of the non-vaccinated control group, whereas vaccinations reduced the viral titers to 10 2 −10 4 PFU (Figure 4G), showing potent sterile immunity. As compared to the single vaccination, each boosting vaccination resulted in an approximate 10-fold reduction of the viral titers of PR8 (H1N1) and Phil82 (H3N2). However, the viral titers of MA81 (H5N2) and reNet03 (H7N1) were higher in the homologous boosting group than the single vaccination group. Again, heterologous boosting and heterosubtypic boosting further decreased the viral titers of MA81 (H5N2) and reNet03 (H7N1). Overall, the protective efficacy evaluated by the weight loss was mirrored in the lung viral titer (Figures 4F,G). The results of protection tests in vivo suggest two important points regarding the breadth and potency of our vaccination strategy. First, despite the absence of antibody-mediated viral neutralizing activities, prime-boost vaccination with CAIVs comprising only the HA group 1 strains provided a broad and potent cross-protection covering both HA group 1 and group 2 influenza viruses in vivo. Second, the potency of cross-protection against more virulent strains such as H5 and H7 avian influenza viruses could be enhanced through the heterologous or heterosubtypic boosting.

contribution of T cells and nK cells to cross-Protection
To further define immunological correlates responsible for the cross-protection, we focused on T cells which have been considered as the key factor for cross-protection. We first examined whether CD8+ cytotoxic T lymphocytes (CTLs) directed to the conserved NP147−155 epitope could be recalled in vaccinated mice upon a heterologous challenge. For this purpose, separate groups of mice were primed and boosted with CAIVs and then challenged with PR8 (H1N1) or Phil82 (H3N2), and the specific CTLs in the blood and lungs were analyzed by flow cytometry (Figure 5A). In the peripheral blood, a variable but slightly increased recall of NP147−155+ CTLs was observed through six days after the challenge with PR8 (H1N1) or Phil82 (H3N2) (Figures 5B,C). In the lungs, a much more significant increase in the NP147−155+ CTLs were observed in the vaccinated groups upon PR8 (H1N1) challenge, in which NP147−155+ CTLs accounted for 2.7−3.7% of the whole CD8+ T-cell population, corresponding to a 7.3−9.8-fold increase relative to that of the non-vaccinated control ( Figure 5D; Figure S5A in Supplementary Material). Phil82 (H3N2) challenge also recalled the NP147−155+ CTLs in the lungs of vaccinated mice, albeit not significantly (Figure 5E; Figure S5B in Supplementary Material). However, IFN-γ producing CTLs were significantly increased upon Phil82 (H3N2) challenge in all three boosting vaccination groups (Figure 5F; Figure S5C in Supplementary Material), suggesting that the boosting vaccination generated CTLs with a specific memory, which then can be re-stimulated to produce IFN-γ cytokine upon the heterologous challenge.
To further determine cell-mediated immunity for crossprotection in vivo, a particular subset of the T cells or NK cells were depleted by injecting anti-CD4, anti-CD8, or anti-asialo GM1 antibodies into mice (Figure 6A). Injection of the antibodies to vaccinated mice resulted in the depletion of >98% of CD4+ T cells and CD8+ T cells in the peripheral blood and the lung, and the depletion of 70−80% of NK cells in the spleen ( Figures S6A,B in Supplementary Material). First, the lethality of non-vaccinated mice ( Figure 6B). Next, separate mice were vaccinated with ca-pH1N1 (single vaccination) or ca-pH1N1 + ca-NCH1N1 (heterologous boosting) and their T cells or NK cells were depleted before challenge with either PR8 (H1N1) or Phil82 (H3N2). Challenge with PR8 (H1N1) caused little morbidity in the single vaccination group that was given isotype control antibodies (Figure 6C). Depletion of CD8+ T cells resulted in a weight loss of ~11% upon the challenge without death. When both CD8+ T cells and CD4+ T cells were depleted, PR8 (H1N1) challenge caused a weight loss of ~13% and death in one of the five mice, and the recovery of the survived mice was significantly delayed (Figure 6C). In a single vaccinated mice, NK cell-depleted group and CD8+ T-cell-depleted group demonstrated similar levels of weight loss upon PR8 (H1N1) challenge, and depletion of both T-cell populations as well as NK cells led to a substantial weight loss of ~20% and the death of three mice (Figure 6C). In heterologous boosting group (ca-pH1N1 + ca-NCH1N1), depletion of both T-cell populations and NK cells resulted in a weight loss of ~10%  and the death of one mouse ( Figure 6D). Phil82 (H3N2) was also used as a challenge strain. In the single vaccination group, Phil82 (H3N2) challenge caused a weight loss of ~10% and the death of one mouse when CD8+ T cells, CD4+ T cells, and NK cells were depleted ( Figure 6E). Heterologous boosting vaccination group did not develop any morbidity upon the Phil82 (H3N2) challenge even in the absence of both T cells and NK cells (Figure 6F).
Of note, depletion of CD4+ T cells, CD8+ T cells, and NK cells did not completely abolish cross-protection against the heterologous challenges. These results suggest not only that T cells and NK cells contribute significantly to the cross-protection, but also that there remained other protection mechanisms operating robustly.

safety issues
As described earlier, VAERD has been reported in the cases of influenza virus vaccines, in which non-neutralizing HA stalk antibodies enhanced viral infectivity and aggravated diseases upon heterologous infection (6), raising a concern on the HA stalk-based approaches for cross-protection. Considering that prime-boost vaccinations with CAIVs induced high levels of HA stalk antibodies (Figure 2B), we examined whether the antibodies caused enhanced viral infectivity of the heterologous viruses. Each virus was pre-incubated with the serial dilutions of sera for binding and absorbed into MDCK or RAW264.7 cells, and the viral replication was monitored by NP-based ELISA (6). In both cell lines, vaccination-induced sera antibodies effectively inhibited the replication of the homologous pH1N1 virus, as compared to the non-vaccinated control sera (Figures 7A,B). The sera antibodies showed partial inhibition of the replication of the four heterologous influenza viruses, resulting in the infectivity ranging 50−85%, as compared to the control. However, none of the sera antibodies promoted the viral infectivity of heterologous influenza viruses. The results, along with in vivo protection tests described above, suggest that our strategy provides a safe crossprotection against heterologous influenza viruses without causing VAERD by non-neutralizing antibodies. Another potential safety issue of prime-boost vaccination with CAIVs is related to the phenomenon of original antigenic sin. The original antigenic sin theory, first described in 1953 by Thomas Francis (19), refers the phenomenon in which sequential exposure to antigenically different virus strains results in preferential antibody responses to the first strain and impaired immune responses to the second strain. In line with this, recent studies have demonstrated that sequential infections of mice with two different influenza virus strains resulted in almost an exclusive neutralizing antibody response to the first strain and a severely impaired protective immunity to the second strain (20,35). Thus, we examined if a similar phenomenon occurred in the prime-boost vaccination strategy, in which antigenically heterologous CAIVs were sequentially given to mice. ca-NCH1N1 and ca-IDH5N1 were used as the second vaccine strain in heterologous boosting (ca-pH1N1 + ca-NCH1N1) and heterosubtypic boosting (ca-pH1N1 + ca-IDH5N1) vaccinations, respectively. Neutralizing antibodies to each homologous wild-type virus were estimated by MN and HI assays and were compared to those induced  Figure 7C). Consistently, these two vaccination groups produced similar HI antibody titers (341 and 363) against the same virus. In parallel, a single vaccination with ca-IDH5N1 induced the mean HI and MN antibody titers of 128 and 310, respectively, which were similar to those developed by heterosubtypic boosting vaccination with ca-pH1N1 + ca-IDH5N1 ( Figure 7C). The results show that prime vaccination did not prevent the generation of neutralizing antibody responses to the second antigenically heterologous vaccination. Taken together, our data demonstrate that prime-boost vaccination with CAIVs does not accompany adverse effects from heterogeneity not only between the vaccine strains (VAERD) but also between the vaccine and challenge virus (antigenic sin-like phenomenon), supporting the safety of our vaccination strategy.

DiscUssiOn
In this study, we presented prime-boost vaccination with CAIVs as a reliable universal influenza vaccination strategy that conferred a broad protection against diverse influenza viruses. Although the vaccination developed cross-reactive systemic and mucosal antibodies against heterologous influenza viruses, HI and MN assays showed that the antibodies were able to neutralize the homologous virus but not heterologous viruses. Considering that the ELISA using the whole viruses represented the collective results of antibodies to multiple surface proteins including HA, NA, and M2 proteins, further study on the profiles of specific antibodies toward the individual surface proteins would be needed with respect to pan-influenza A protection including HA group 2 viruses. Even without antibody-mediated neutralizing activities, the vaccination provided a complete protection against heterologous lethal challenges and restricted the viral replication in the lungs. It was unexpected that homologous boosting would demonstrate diminished protective efficacy against MA81 (H5N2) and reNet03 (H7N1) as compared to a single vaccination (Figure 4), considering that homologous boosting generally guarantees improved protection against homologous challenge. It is likely that homologous boosting negatively affected the quality of cross-protection, which could become noticeable when a challenge virus was a highly pathogenic strain such as MA81 (H5N2) or reNet03 (H7N1). This phenomenon merits further investigation into the underlying mechanisms to establish a safe vaccination strategy against virulent strains. Heterologous boosting and heterosubtypic boosting demonstrated a more potent cross-protection against virulent H5 and H7 influenza viruses than the homologous boosting vaccination. Based on the results, we suggest that prime-boost vaccination with CAIVs with genetically and immunologically different HA and NA surface antigens would be a strategy to increase the potency of cross-protection. It is likely that upon a boosting vaccination with heterologous CAIV, potently neutralizing antibodies specific to the immunologically dominant but variable regions are rarely boosted, whereas non-(or less) neutralizing antibodies or CTLs cells directed to the conserved regions are preferentially increased. In support of this assumption, significant increase in the HA stalk antibodies was achieved by heterologous boosting and heterosubtypic boosting but not by homologous boosting (Figure 2B). Furthermore, ADCC activity to antigenically distant H3 and H7 viruses was significant only in the sera from heterologous boosting and heterosubtypic boosting ( Figure 3C). Consistent with this, heterologous boosting and heterosubtypic boosting vaccinations demonstrated a higher ability to generate IFN-γ-producing CTLs upon Phil82 (H3N2) challenge than homologous boosting ( Figure 5F). PR8 (H1N1) challenge was able to stimulate NP147−155+ CTLs although noticeable differences among vaccination groups were not observed ( Figure 5D). Parallel experiments with the Phil82 (H3N2) challenge only barely stimulated the NP147−155+ CTLs (Figure 5E). These results suggest that the repertoire of activated CTLs can vary according to the challenge virus strain even under the same vaccination conditions. Clearly, further studies are needed on the potential enhancement of CTLs by boost immunization and epitopespecific CTLs responsive to the challenge virus.
The most contrasting feature of our strategy from previously developed HA stalk-based universal vaccines is that the crossprotection elicited by CAIVs did not depend on the neutralizing activities of HA stalk antibodies. Although prime-boost vaccinations with CAIVs induced high levels of HA stalk antibodies, the antibodies exhibited no viral neutralizing activity against the heterologous influenza viruses. HA stalk antibodymediated inhibition of membrane fusion has been presented as a key mechanism for eliciting a broad protection in vaccine approaches that target the HA stalk (8,9). It should be noted, however, that most of the HA stalk-based approaches were based on the recombinant HAs carrying the same stalk domains to enable directed boosting toward the conserved region. The current vaccine strategy relied on attenuated viruses carrying phylogenetically distinct HAs carrying relatively conserved and yet different stalk domains. Furthermore, vaccination with CAIVs generally induces a mixture of polyclonal antibodies against the surface antigens including the HA, NA, and M2 as well as the internal proteins, each of which carries a different profile of antigenic epitopes. It remains to be determined whether a heterogeneous profile of antibodies negatively affects the neutralizing activity of the HA stalk antibodies generated in the present study. The induction of polyclonal antibody mixture presents a beneficial effect on cross-protection. There is increasing evidence that ADCC-mediating antibodies directed toward the surface HA, NA, and M2 as well as the internal proteins including NP and M1 were correlated strongly with cross-protection against heterologous influenza viruses (32,36). Therefore, prime-boost vaccination with CAIVs is likely to induce the whole set of ADCC-mediating antibodies, further increasing the potency of cross-protection. Our data showed that vaccination-induced non-neutralizing antibodies had weak but distinct ADCC activity against heterologous influenza virus-infected MDCK cells (Figure 3C). Consistent with this, depletion of NK cells increased morbidity from PR8 (H1N1) challenge in the single vaccinated mice (Figure 6C). While in vivo depletion experiments clearly showed that CD8+ T cells, CD4+ T cells, and NK cells contributed to the cross-protection, the depletion did not completely eliminate the cross-protection elicited by vaccinations, with boosting vaccination providing better protection than the single vaccination (Figure 6). These results suggest that other protective mechanisms are in operation with CAIVs, which can be augmented by boosting vaccination.
To explain possible mechanisms for the cross-protection shown in the present study, careful considerations on various immunological factors potentially induced by a CAIV should be given in a comprehensive manner, since our data did not address all of them individually. One of possible factors is the mucosal IgA antibody-mediated protection in the respiratory tracts. It has been well-known that CAIVs mimic natural infection and thus induce secretory IgA antibodies at the upper and lower respiratory tracts (37)(38)(39). Considerable reports have suggested that the secretory IgA antibodies are more cross-protective against influenza virus infections than systemic humoral antibody responses (40)(41)(42). Although we could not detect antibody-mediated neutralizing activity in the BALF or nasal turbinates in vitro assays, the vaccination resulted in high levels of cross-reactive IgA antibodies ( Figure S3B in Supplementary Material), which are likely to play a protective role in vivo. CAIVs also develop antibody responses to the M2 external (M2e) domain that is highly conserved across influenza A viruses, and the M2e has long been an attractive target for developing broadly protective universal influenza vaccines (30). In our vaccination strategy, the M2e-specific antibodies also likely play an important role for the cross-protection. Additionally, it has been shown that non-neutralizing antibodies can mediate a number of protective functions, such as complement-dependent cytotoxicity (CDC) Pan-Influenza A Protection by CAIVs Frontiers in Immunology | www.frontiersin.org February 2018 | Volume 9 | Article 116 and antibody-dependent phagocytosis (ADCP) (43,44), in addition to ADCC. There is also evidence that innate immune effector cells such as macrophages, monocytes, and neutrophils are capable of inducing ADCC against influenza virus (45)(46)(47). In our data, the depletion of CD8+ T cells, CD+ T cells, and NK cells did not completely abolish the protection against heterologous challenges, especially against Phil82 (H3N2) challenge ( Figure 6F). It is therefore likely that various types of antibodydependent protective functions are in operation in vivo even after the depletion of CD8+ T cells, CD4+ T cells, and NK cells.
Besides NK cells, the expression of asialo GM1 was also observed in multiple cellular subsets including NKT cells, CD8+ T cells, CD4+ T cells, γδ T cells, macrophages, eosinophils, and basophils (48)(49)(50)(51)(52)(53). Thus, it is likely that the injection of the anti-asialo GM1 antibodies affects in vivo fates of those cells. Particularly, γδ T cells were shown to contribute to heterosubtypic immunity influenza A virus in the knockout mouse model (12). In our experimental condition, however, the treatment of anti-asialo GM1 antibodies in mice rarely affected the frequencies of the cells expressing CD3 antigen that is expressed in all T-cell populations including NKT cells, CD8+ T cells, CD4+ T cells, and γδ T cells ( Figure S6B in Supplementary Material). Still, there remains the possibility that macrophages, eosinophils, and basophils that are important components of cell-mediated innate immune system might be influenced by the anti-asialo GM1 antibodies treatment. These issues underline the need for careful interpretation of phenotypes demonstrated in mice treated with anti-asialo GM1 antibodies.
A truly universal influenza vaccine should be able to guarantee both the potency and the breadth of cross-protection without inducing any adverse effects (5,54). In this study, we used ca-pH1N1, ca-NCH1N1, and ca-IDH5N1, which are antigenically distinct but close to each another belonging to the HA group 1 (Figure 1B). And yet, prime-boost vaccination with those CAIVs showed such a broad spectrum of protection covering both HA group 1 (H1 and H5) and group 2 (H3 and H7), comprising the viruses of human-infecting and zoonotic potentials. Furthermore, a single vaccination was sufficient to protect mice from death upon the lethal challenge, and boosting substantially improved the potency of protection. Given that these four influenza subtypes present major threats to humans health associated with influenza A viruses, it can safely be concluded that our vaccination strategy confers pan-influenza A protection, which has not been achieved yet by other universal influenza vaccine approaches. It is worth mentioning that, besides the excellent levels of breadth and potency of cross-protection, our vaccination strategy satisfactorily addressed vaccine safety issues, as it did not accompany adverse effects such as VAERD and the antigenic sin phenomenon, both of which sometimes skew the immunogenicity profiles. The promising results of paninfluenza A protection raise an optimistic prospect of developing a pan-influenza universal vaccine that is protective against both influenza A and B viruses simultaneously. Considering the cross-reactive B-cell and T-cell epitopes between influenza A and B viruses (55), it may well be possible to induce such antibodies and T-cell responses through a rational vaccination strategy with CAIVs.
The current study was conducted in a mouse model where the host immune responses to vaccination or infection can be studied in depth (56). It merits further evaluation in higher animal models such as ferrets or swines to better address the clinical relevance and practicality of our strategy. It should also be mentioned that the naive mice had no preexisting immunity to an influenza virus. It has been reported that in humans, preexisting memory B-cell or T-cell immunity affects the recall responses upon the subsequent influenza infection or vaccination (57)(58)(59). Our data showed that heterologous prime and boosting vaccinations with an interval of 2 weeks were capable of inducing robust neutralizing antibodies to both strains without any interference between the two vaccines. Additionally, neither heterologous nor heterosubtypic challenge after a month of vaccination accompanied any signs of VAERD. An experimental design with a much longer interval between vaccinations or between vaccination and challenge could provide a greater insight into the relationship between preexisting immunity and protection efficacy against heterologous/heterosubtypic infections.
There remains a possibility that our results of cross-protection were, in part, due to innate immunity such as temporary nonspecific immunity or the phenomenon of viral interference by vaccination with a live attenuated virus. Non-specific viral interference between related or non-related viruses has long been observed in humans, and many epidemiological studies have reported clinical cases potentially explained by the phenomenon (60,61). However, it was reported that non-specific immunity against non-influenza respiratory viruses in children vaccinated with a live-attenuated influenza vaccine (LAIV) was short-lived with the duration of as long as 1−2 weeks (62). Furthermore, many animal studies on cross-protection by LAIVs have presented protection data by performing in vivo challenge at 2−4 weeks after the vaccination (11,17,33,63,64). Considering these observations, our experimental design of the challenge at 5 weeks days after the boosting vaccination cannot be explained by non-specific mechanism. Finally, in our data, in vivo protection efficacy against the same challenge strains was significantly different among vaccination regimens (Figure 4), which is unlikely to be mediated by non-specific immunity. In the four vaccinated groups, mice were boosted with the same dose (10 5 PFU) of CAIV, but their weight loss and the viral replication in the lungs after the challenge suggested the different potency of protection depending on the boosting strain, implying specific adaptive immunity controlling the protection.
There has been an increasing need for a universal influenza vaccine with high levels of breadth, potency, and safety. Harnessing such ideal traits has been proven difficult by using strategies targeting small epitopes or domains alone. Our data demonstrate that prime-boost vaccination with the X-31cabased CAIV presents as a potentially powerful universal influenza vaccine that provides a broad and potent cross-protection by activating multiple immune arms including antibodies and T cells. Considering that the X-31ca provides the internal backbone of H1N1 strain, it is likely that the X-31ca-based CAIVs demonstrate different profiles of antibody responses and T-cell responses to the internal proteins, as compared to the A/Ann Arbor/6/60ca (H2N2) strain. Whether this difference could bring Pan-Influenza A Protection by CAIVs Frontiers in Immunology | www.frontiersin.org February 2018 | Volume 9 | Article 116 significant changes in the quality of cross-protection should be addressed by further studies. Whether cross-protection could be extended to influenza B viruses, thus covering all humaninfecting viruses, remains ultimate technical challenge for "truly" universal influenza vaccine development.

eThics sTaTeMenT
All experiments. YJ analyzed antibody responses using ELISA, in vitro neutralizing assays, and ADCC assay. AS produced recombinant HA proteins using E. coli system. JK and JL performed flow cytometry to measure CTL responses and to confirm the depletion of T cells and NK cells. JC provided technical assistance and contributed reagents. BS supervised all process of experiments and preparation of the manuscript.