Influenza virus neuraminidase regulates host CD8+ T-cell response in mice

Influenza A virus (IAV)-specific CD8+ T-cell response was shown to provide protection against pandemic and seasonal influenza infections. However, the response was often relatively weak and the mechanism was unclear. Here, we show that the composition of IAV released from infected cells is regulated by the neuraminidase (NA) activity and the cells infected by NA-defective virus cause intracellular viral protein accumulation and cell death. In addition, after uptake of NA-defective viruses by dendritic cells (DCs), an expression of the major histocompatibility complex class I is induced to activate IAV-specific CD8+ T-cell response. When mice were infected by NA-defective IAV, a CD8+ T-cell response to the highly conserved viral antigens including PB1, NP, HA, M1, M2 and NS1 was observed along with the increasing expression of IL10, IL12 and IL27. Vaccination of mice with NA-defective H1N1 A/WSN/33 induced a strong IAV-specific CD8+ T cell response against H1N1, H3N2 and H5N1. This study reveals the role of NA in the IAV-specific CD8+ T-cell response and virion assembly process, and provides an alternative direction toward the development of universal influenza vaccines. Chung-Yi Wu et al. demonstrate a role of neuraminidase (NA) for Influenza A viruse (IAV)-specific CD8+ T-cell response and virion assembly. When mice are vaccinated with the live attenuated NA-defective IAV, they are protected from the upcoming IAV infection. This study provides insights into the development strategies of universal influenza vaccines.

I nfluenza A virus (IAV) has been a major threat to human health, causing seasonal outbreaks and pandemics with high morbidity and mortality especially involving more virulent strains such as 1918 A/H1N1, H5N1, and H7N9. Vaccination has been an effective strategy to control the spread of IAV, and the protective antibodies induced by the vaccine are known to recognize mainly the viral hemagglutinin (HA) domain and partially the NA moiety 1,2 . However, seasonal vaccines may lose their efficacy because of antigenic shift or drift in the circulating virus population through genomic reassortment or antigen mutations, respectively. Interestingly, certain populations, including pregnant women infected by some novel strains of IAV were asymptomatic with little antibody response, but had a steady IAV-specific CD8 + T-cell response to IAV-infected cells 1,2 . In addition, it was shown that humans, mice, and macaques with positive IAV-specific CD8 + T-cell response had cross-protective immunity against pandemic and seasonal influenza [3][4][5][6][7] , and the targets of CD8 + T-cell response were the highly conserved proteins from various IAV strains 8 . Despite the importance of CD8 + T-cell response in protection against influenza infection, the reported cellular immune response is often relatively weak and the mechanism has not been well understood.
In our previous study, we found that mice infected by a live attenuated IAV without the stalk or catalytic domain of NA were able to strongly induce the IAV-specific CD8 + instead of CD4 + T-cell response against the virus, leading to the development of a live attenuated influenza vaccine (LAIV) 4 . In this study, we further investigate the mechanism of such a selective immune response and elucidate the role of NA in the induction of IAVspecific CD8 + T cells against a broad range of IAV strains.

Results
NA activity regulated virus release. We previously reported that immunization of mice with IAV without the stalk or catalytic domain of NA induced a strong IAV-specific CD8 + T-cell response with little antibody elicited 4 . In order to further understand the mechanism, WSN viruses with defective NA activity were generated by reverse genetics through deletion of the stalk or the catalytic domain (LAIV WSN-NA or LAIV WSN-NA-CD) or inactivation of the active site domain (LAIV WSN-NA-AS1 or LAIV WSN-NA-AS2) (Fig. 1a, b, Table 1). After A549 cells or MDCK cells were infected by defective NA viruses, only a small number of viruses was released in the supernatant compared to that from the cells infected by WSN with intact NA (Fig. 1c). In addition, the viral RNA (vRNA) (Fig. 1d) and viral proteins were accumulated inside the cells (Fig. 1e, Supplementary Fig. 7a), suggesting that the release of virus would require NA activity (Supplementary Figs. 1 and 7b).
NA activity affected the virus-mediated host immune response.
To study whether the IAV without NA activity was the cause of reduced viral release and accumulation of viral proteins inside the infected cells leading to the biased host immune response 4 , we performed an animal study (Fig. 2a). Mice infected with the NAdefective virus (LAIV WSN-NA or LAIV WSN-NA-AS1) survived well and showed no body weight loss (Fig. 2b, c). The virus titers were lower in the lungs compared to non-vaccinated or WSN control groups with less interferon alpha and gamma (IFNα/γ) induced in the bronchial alveolar lavage (BAL) fluid on day 4 after infection ( Fig. 2d-f), and little anti-HA antibody was produced (Fig. 2g).
In the H5N1 challenge study (Fig. 2h), mice vaccinated with the NA-defective WSN also survived well without significant body weight loss (Fig. 2i, j) and were able to clear different strains of viruses from the lungs (Fig. 2k). Interestingly, the mice challenged with H5N1 virus induced less IFN-α ( Fig. 2l) but more IFN-γ in BAL (Fig. 2m), indicating the activation of IAV-specific T-cell response. These results suggest that the activity of NA is an important factor in virulence, immune response, and probably is a key factor to be considered for the development of live attenuated influenza vaccines (LAIVs).
NA-defective viruses, expression of MHC, and activation of IAV-specific CD8 + T cells. To understand how responses to NAdefective viruses might activate IAV-specific CD8 + T cells, we first investigated the interaction of dendritic cells (DCs) with IAVs and IAV-infected cells to understand how they were processed by DCs for presentation to T cells and other immune cells 9,10 . After A549 cells were infected by NA-defective WSN viruses, the cells showed increased viability (Fig. 3a) and released fewer viruses into the supernatant as compared to the WSNinfected cells (Fig. 3b). It was also found that the distribution of some viral proteins inside the cell was not significantly affected, probably due to the different processes of cell death in different IAV-infected cells and viral proteins accumulated in NAdefective virus-infected cells ( Supplementary Fig. 2). The green fluorescent signals were relatively weak compared to WSN (Fig. 3c) in DCs co-cultured with mouse B cells labeled with 5(6)carboxyfluorescein diacetate N-succinimidyl ester (CFSE) and infected with NA-defective WSN that had similar vRNA levels at 6-h post-infection to WSN-infected B cells (Fig. 3d), suggesting that the NA activity might affect the uptake of IAV antigens by DCs. When DCs were cultured with the cells infected by WSN, more DCs with green fluorescent signal and more viral protein M1 inside the DCs (Fig. 3e, Supplementary Fig. 7c) were observed. After further incubation at 37°C for 24 h, more viruses were released to the supernatant (Fig. 3f), suggesting that WT viruses replicated in DCs.
To further study the effects of DCs infected by IAV, the major histocompatibility complex class I (MHC I) and class II (MHC II), which are essential for presentation of the internalized molecules after processing, were measured by flow cytometry 9 . In general, we observed that when DCs were infected by WSN or cocultured with the cells infected by WSN, MHC II was more expressed than MHC I (Fig. 3g, h, j, k), and the virus replicated more in DCs (Fig. 3i). However, when DCs were co-cultured with the cells infected by NA-defective WSN viruses, MHC I was more expressed whereas MHC II was not expressed on DCs (Fig. 3j, k) and little virus was released to the medium (Fig. 3l). We also found that when DCs were incubated with inactivated IAVs, more MHC II than MHC I was expressed ( Supplementary Fig. 3a, b); however, when DCs were incubated with the cells treated with the inactivated virus, the expression of MHC I/II was not increased ( Supplementary Fig. 3c, d).
MHC I-expressed DCs can prime and activate the proliferation of CD8 + T cells and secretion of IFN-γ 9 . LAIV WSN-NA treated mice would induce IAV-specific CD8 + T cells 4 . When DCs were co-cultured with the B cells infected by NA-defective viruses, the isolated DCs were incubated with CD8 + T cells from mice treated with different IAVs. As expected, the DCs induced a proliferation of CD8 + T cells (Fig. 3m) accompanied with IFN-γ secretion ( Supplementary Fig. 4a), especially when DCs were incubated with the CD8 + T cells isolated from the mice vaccinated with LAIV WSN-NA produced from MDCK cells. This result suggested that the uptake of antigens by DCs from the cells infected by NA-defective viruses would prime and activate the IAV-specific CD8 + T cells.
The IAV-specific CD8 + T cells induced by NA-defective WSN were able to recognize the epitopes from different IAV proteins such as PB1, NP, HA, M1, M2, and NS1, as shown by their secretion of granzyme B (GrzB) by flow cytometry (Fig. 3n). In addition, the CD8 + T cells were more activated when incubated with different combinations of viral peptides ( Supplementary  Fig. 4b). Furthermore, when the cells infected by NA-defective WSN were incubated with different ratios of CD8 + T cells, the lysis of the IAV-infected cells (Fig. 3o)  NA-defective viruses and expression of cytokines in host immune response. The mechanism of immune response to IAV infection that causes tissue injury in the host is complicated, and it is known that the cellular activities of innate and adaptive immune response can be affected by cytokines 11,12 . Here we found that mice infected by WSN induced higher levels of IFNα, IFNγ, IL-2, IL-4, and IL-6 known to activate the antibody response, while the NA-defective WSN induced higher levels of IL-10, IL-12, and IL-27 known to trigger the cellular immune response (Fig. 4a-h) [13][14][15][16][17][18] . In a recent study, IL-17D was shown to be a critical cytokine during IAV infection to suppress the activity of CD8 + T cells through regulation of dendritic cells 19 . We also found that mice immunized with NA-defective WSN induced a very low level of IL-17D (Fig. 4i), likely minimizing  Table 1) for analysis of the intracellular viral proteins HA, NP, and M1 by western blot. The filter was probed with anti-NA, anti-NP, anti-M1, and anti-β-actin antibodies. b Mean ± SD of five independent experiments. c, d Mean ± SD of three independent experiments. *P < 0.001.  its suppression of the CD8 + T-cell induction (Fig. 4j, k) and activation ( Fig. 4l, m). This result was supported by the observation that IL-17D was able to reduce the secretion of INFγ from CD8 + T cells (Fig. 4j, k) or IAV-specific CD8 + T cells (Fig. 4l, m) 19 . We speculate that when IAV was taken by DCs, MHC II, and the cytokines including IL-2, IL-4, and IL-6 were expressed to activate CD4 + T cells, and IL-17D was induced to suppress the CD8 + T-cell activity. However, the expression of MHC I on DCs to prime the IAV-specific CD8 + T cells required the uptake of IAV-infected cells and the cooperation of cytokines such as IL-10, IL-12, and IL-27 ( Fig. 4n).
Severe influenza infection causes a cytokine storm that originates from the hyperinduction of proinflammatory cytokine production 20,21 . Interestingly, the mice with robust IAV-specific CD8 + T-cell activation could avoid the cytokine storm caused by highly pathogenic IAV infection. When mice were vaccinated with LAIV WSN-NA followed by infection with a high lethal dosage of H5N1, the mice produced less total protein in the BAL fluid and lower levels of IFN-α, IL-2, IL-4, IL-6, and IL-17D, but higher levels of IFNγ, IL-12, and IL-27 to activate the IAV-specific CD8 + T-cell response, suggesting that the NA activity would affect the course of host immune response to IAV infection (Fig. 5).
NA activity and composition of viruses. Unexpectedly, we also found that the composition of released viruses from IAV-infected cells was affected by NA-defective strains. At the late stage of NAdefective IAV infection in MDCK cells (on 18-h post-infection), the cells released some particles (LAIV WSN-NA(p) and LAIV WSN-NA-AS1(p)) that had different distributions of proteins ( Supplementary Fig. 6a). These particles contained more M2 (ion channel) protein and less HA protein (Fig. 6a, Supplementary  Fig. 7d), but still had the viral genome (vRNA) (Supplementary Figs. 6b, 7e) albeit with lower HA-sialic acid binding ability ( Supplementary Fig. 6c), and were mainly elongated, filamentous, and irregular-shaped particles ( Fig. 6b-d).
Application to universal vaccine design. After A549 cells were infected by LAIV WSN-NA(p) and LAIV WSN-NA-AS1(p) produced from MDCK cells, there were fewer viruses released into the supernatant (Fig. 6e), and the intracellular viral genome (Fig. 6f) and protein M1 (Supplementary Figs. 6d, 7f) were accumulated, along with a reduction in viable cells (Fig. 6g). When mice were infected with 1 × 10 6 pfu of LAIV WSN-NA(p) or LAIV WSN-NA-AS1(p) via oral administration, they survived well with no body weight loss, and were able to effectively clear the infection particles ( Supplementary Fig. 6e-g). In addition, the mice induced less IFN-α/γ in the BAL fluid ( Supplementary  Fig. 6h, i) and produced very low levels of anti-HA antibodies (Fig. 6h), indicating that these particles could be used as effective LAIVs with low pathogenicity. In the H5N1 challenge study, the immunized mice were able to survive well without body weight loss (Fig. 6i, Supplementary Fig. 6j), clear the virus from the lungs ( Supplementary Fig. 6k), and induce less IFN-α ( Supplementary  Fig. 6l) and more IFN-γ (Fig. 6j), suggesting that the IAV-specific CD8 + T-cell response was activated. Since the immune responses of LAIV WSN-NA(p) and LAIV WSN-NA-AS1(p) were similar to that of NA-defective WSN, we concl uded that the induction of IAV-specific CD8 + T-cell response does not require the release of much infective virus, further supporting the important role of NA in the process of immune response.

Discussion
It is known that infection with one subtype of IAV may induce IAV-specific CD8 + T cells to target different subtypes of IAV 6,8,22 , or IAV-specific CD4 + T cells to target that specific subtype of IAV 8,23 . However, patients with severe IAV infection usually have a limited or severely compromised CD8 + T-cell proliferation and differentiation response to eliminate the virus 6,24,25 . Our study indicated that though CD4 + T cells were generally induced in response to IAV infection, and IL-17D was expressed to suppress the CD8 + T-cell activity, MHC I was expressed to activate CD8 + T cells in the presence of NAdefective IAV. In addition, the mice infected with WT IAV did not generate higher CD8 + T cells through induction of cytokines such as IL-10, IL-12, and IL-27 18,26 .
Recent studies have suggested strategies for the design of LAIV to elicit CD8 + T cells for cross protection 4,27-30 , and uptake of large amounts of target antigens by DCs will stimulate an efficient activation of CD8 + T cells [31][32][33][34] . However, our study showed that the viral particles released from MDCK cells infected by NAdefective WSN would provide sufficient antigens for DCs to prime the IAV-specific CD8 + T cells and eliminate IAV-infected cells through recognition of conserved cytotoxic T lymphocyte epitopes of various IAV antigens such as PB1, NP, HA, M1, and NS1. During the process of influenza infection, the cytoplasmic tails of HA and NA play an important role in the budding process. The concentration of HA and NA in lipid raft provides a local alteration in membrane curvature resulting in the interaction of the cytoplasmic tails of HA and NA with M1 protein to start the budding process. M1 recruits M2 to the budding site for membrane scission and NA then cleaves the sialic acid linkage on the cell surface to release the budding virion [35][36][37][38] . In our study, the NA-defective viruses can still express the cytoplasmic tails of NA and produce infection particles in MDCK cells but not in A549 cells, probably because the MDCK cell is highly polarized and susceptible for IAV replication 39 . Surprisingly, these infection particles have different compositions, especially with increase in M2 and decrease in HA expression, suggesting that NA is used not only for the cleavage of sialic acid ligands to release virions but its ectodomain also affects virion composition in the budding process.
In summary, this study has demonstrated that the NAdefective WSN or the viral particles produced from MDCK cells infected with NA-defective WSN can be used as LAIV to effectively induce IAV-specific CD8 + T-cell response to protect against different influenza strains such as H1N1, H3N2, and H5N1. The LAIV particles can be easily produced and this approach may lead to a new direction toward the development of a universal influenza vaccine.

Methods
Cell lines and viruses. Madin-Darby canine kidney cells (MDCK) and human embryonic kidney cells (HEK293T) were maintained in Dulbecco's modified Eagle's medium (DMEM) (Invitrogen, Rockville. MD). A549 human adenocarcinoma alveolar basal epithelial cells were kept in F-12K medium (Invitrogen, Rockville, MD). All media were supplemented with 10% heat-inactivated fetal bovine serum (FBS) (Thermo Scientific) and antibiotics (100 U/ml penicillin G and 100 gm/ml streptomycin). The MDCK cells with stable NA expression were prepared by cloning the full-length NA (from WSN strain) into a cDNA expression lentivector (pLAS2w.Ppuro) to generate NA-expression lentivirus using the protocol provided by the National RNAi Core Facility, Academia Sinica, Taiwan (rnai. genmed.sinica.edu.tw) and were maintained in DMEM medium 4   Generation of recombinant viruses. Eight fragments of A/WSN/33 viral genome were amplified by RT-PCR. As shown in Table 1, we mutated the N residue to A in the putative sequon N-X-S/T to delete the glycosites 44, 72, and 219 (i.e., N44A mutation designated as 44-G), or added a stop codon to the first amino acid residue of the stalk or catalytic domain in the NA genome to remove the stalk and/or catalytic domain (designated as LAIV WSN-NA (S31stop) and LAIV WSN-NA-CD (L75stop)) of NA as previously described 4 , or changed the designated amino acid in the active site (AS1: R102A (designated as LAIV WSN-NA-AS1); AS2: D135A (designated as LAIV WSN-NA-AS2)) by site-directed mutagenesis. The viral cDNAs were inserted into pcDNA3.1 containing the pol I and CMV promoter with the method similar to the way pHW2000 was generated. The recombinant viruses were generated by the 8-plasmid co-transfection method into MDCK/ 293T cells with stable NA expression following the method described previously by our group 4 . Supernatants were collected, titrated, and frozen at −80°C until use.
Preparation of WSN and NA-defective WSN viruses for immunogenicity test.  IAV-infected cells regulate DCs to affect the proliferation of CD8 + T cells.
After DCs were incubated with the B cells infected by WSN or LAIV WSN-NA virus, DCs were isolated and co-cultured with CFSE-labeled CD8 + T cells (1 μM CFSE for 10 min at room temperature, then washed twice with complete RPMI 1640) (ratio: 1:2) from mice treated with PBS, WSN, or LAIV WSN-NA in complete RPMI 1640 with 0.5 ng/mL IL-7 (R&D Systems), 30 U/mL IL-2 (R&D Systems), and 50 μM 2-ME for 24 h at 37°C. CD8 + T cells were then isolated by Dynabeads Untouched Mouse CD8 Cells kit (Invitrogen) following the procedure from the company and the number of cells with FITC signal was analyzed by flow cytometry.
Suppression of CD8 + T-cell activity by IL-17D. After DCs incubated with the B cells infected by LAIV WSN-NA virus, the DCs were isolated and co-cultured with CD8 + T cells (from PBS or LAIV WSN-NA treated mice) (ratio: 1:2) and IL-17D (0-200 ng/mL, R&D Systems). After 3 days, the supernatants were collected to measure the IFN-γ production by ELISA kit. Heat inactivated IL-17D was prepared by heating at 56°C for 30 min.
CD8 + T cytotoxicity assay. CD8 + T cells were isolated from the mice immunized with WSN, LAIV WSN-NA, or PBS using Dynabeads Untouched Mouse CD8 Cells kit (Invitrogen) and co-cultured with CFSE-labeled B cells infected with LAIV WSN-NA at an MOI of 3 in complete RPMI 1640. The co-cultures were run at different cell ratios (i.e., 1:1, 2:1, and 4:1 CD8 + T cell/IAV-infected cell) in 200 μl of RPMI 20% in U-bottom 96-well plates. Afterward, the mortality of IAV-infected cells was scored by flow cytometry. The value of cell mortality was normalized to that of IAV-infected cells without co-culturing with CD8 + T cells as IAV-infected cells would cause cell apoptosis. Cell-binding assay. Turkey erythrocytes were pretreated with different amounts (0-60 μg/mL) of Vibrio cholerae neuraminidase [receptor destroying enzyme (RDE)] (Sigma) for 60 min at 37°C, then washed once with PBS and divided into 2% (vol/vol) erythrocyte solutions in PBS. Twenty-five microliters of each 2% solution were added to a solution of virus containing the same amount of M1 protein (determined by western blot) to give a total volume of 125 μL. The virus and RDE-treated erythrocytes were incubated for 1 h at room temperature, and agglutination was then measured. Data were expressed as the maximal concentration of RDE that gave full agglutination 4 .
Transmission electron microscopy. MDCK cells were grown on ACLAR embedding film with 7.8-mil thickness (E, M, S) for 1 day followed by infection with influenza virus at an MOI of 5. At 24 hpi, the virus-infected cells were rinsed with 0.1 M cacodylate buffer and fixed with 2.5% glutaraldehyde in 0.1 M cacodylate buffer at 4°C for 30 min. Then cells were postfixed with 1% osmium tetroxide in 0.1 M cacodylate for 30 min, stained with 1% uranyl acetate and lead citrate for 1 h, dehydrated with ethanol, and embedded with resin. After baking, the sample was cut into 80-nanometer thin sections using ultramicrotome and examined with Tecnai G2 Spirit TWIN (FEI Company) 36 .
Statistics and reproducibility. All data were presented as means ± standard error of the mean. The numbers of sample and replicates of experiments were shown as mentioned in the figure legends. Comparisons between groups were determined using Student's t test. Differences were considered significant at *P < 0.001, **P < 0.05. All data were analyzed using GraphPad Prism 6 software.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All relevant data are available from the authors upon request and the corresponding author will be responsible for replying to the request. Source data underlying plots shown in figures are provided in Supplementary Data 1. Full blots are shown in Supplementary  Fig. 7 of Supplementary Information.