The glycoproteins hemagglutinin (HA) and neuraminidase are the major determinants of host range and tissue tropism of the influenza virus. HA is the most abundant protein in the virus particle membrane and represents the basis of most influenza vaccines. It has been reported that influenza virus HA N-glycosylation markedly depends on the host cell line used for virus production. However, little is known about how differential glycosylation affects immunogenicity of the viral proteins. This is of importance for virus propagation in chicken eggs as well as for innovative influenza vaccine production in mammalian cell lines. In this study, we investigated the impact of the differential N-glycosylation patterns of two influenza A virus PR/8/34 (H1N1) variants on immunogenicity. Madin–Darby canine kidney cell–derived and Vero cell–derived glycovariants were analyzed for immunogenicity in a TCR-HA transgenic mouse model. Next-generation pyrosequencing validated the congruence of the potential HA N-glycosylation sites as well as the presence of the HA peptide recognized by the TCR-HA transgenic T cells. We show that differential HA N-glycosylation markedly affected T cell activation and cytokine production in vitro and moderately influenced IL-2 production in vivo. Cocultivation assays indicated that the difference in immunogenicity was mediated by CD11c+ dendritic cells. Native virus deglycosylation by endo- and exoglycosidases dramatically reduced cytokine production by splenocytes in vitro and markedly decreased HA-specific Ab production in vivo. In conclusion, this study indicates a crucial importance of HA N-glycosylation for immunogenicity. Our findings have implications for cell line–based influenza vaccine design.
Besides ideal hygiene measures, the only known efficient protection from influenza infection is annual vaccination because the virus regularly undergoes antigenic change due to gene reassortment (antigenic shift) or point mutations (antigenic drift) (1). Strategies for enhancing influenza vaccine protection are intensively investigated. One possibility is to use known epitope targets for eliciting neutralizing Abs (2, 3). A focus is the development of vaccines that stimulate the production of Abs capable of neutralizing multiple influenza subtypes (4). Another option is to optimize vaccine uptake into APCs by targeting dendritic cells (DCs) or by increasing Ag immunogenicity (5).
The major Ags of influenza virus are the envelope glycoproteins hemagglutinin (HA) (6) and neuraminidase (7, 8). HA is highly abundant in the virus particle and is able to induce strong and protective immune responses. As for other glycoproteins, the quality characteristics of HA such as activity (9), antigenicity (9–11), binding avidity (12), and receptor-binding specificity (13) strongly depend on macro- and microheterogeneity of its N-glycosylation. HA antigenicity is strongly influenced by the HA amino acid sequence with certain epitopes of HA being highly immunogenic (14, 15). Although the impact of HA N-glycosylation on influenza virus binding to host cell sialosides has been investigated in detail (16–19), less is known about its influence on immunogenicity such as recognition by and activation of APCs or subsequent T cell priming.
In mice, recent studies focused on the impact of influenza virus N-glycan structures on the interaction between the virus and the immune system. For instance, it was demonstrated that increased influenza virus glycosylation resulted in decreased virulence, which was partly mediated by surfactant protein D (SP-D)–induced virus clearance from the lung (20). In addition to SP-D, the macrophage mannose receptor on airway macrophages contributed to HA glycan recognition (21, 22). Mannose-rich glycans on the globular head of HA impacted virus sensitivity to neutralization by a mannose-specific lectin in mouse lung fluids (23). However, in a DNA-based influenza H5 HA vaccine study, three potential glycosylation sites of H5 from the influenza A H5N1 virus did not substantially influence Ab responses (24).
Little is known about the impact of influenza virus N-glycosylation on immunogenicity in humans. It was reported that HA N-glycosylation may mask antigenic epitopes, which can prevent binding of HA by neutralizing Abs or the recognition of HA epitopes by CD4+ T cells (9, 25). A recent study showed that oligosaccharides present on the influenza virus HA are a target for recognition by innate immune proteins of the collectin and pentraxin superfamilies (26). Thus, although there are some indications for recognition of influenza virus HA N-glycans by pattern recognition receptors of the innate immune system, the impact of differentially glycosylated HA on T cell activation remains to be determined.
To date, vaccine production processes of manufacturers and developers such as Baxter, GlaxoSmithKline, Kaketsuken, Novartis, and Sanofi Pasteur range from traditional embryonated egg-based to more recently established cell culture–based processes using adherent Vero and Madin–Darby canine kidney (MDCK) cells or suspension cells such as EBx, MDCK, and PER.C6 cells. Some MDCK and Vero cell–derived vaccines have already been licensed (27–31). The MDCK cell culture–based processes that have been licensed so far are the subunit vaccines Optaflu and Celtura (Novartis) as well as a split vaccine (Solvay Pharmaceuticals/Abbott) (27). The advantages of cell culture–based processes are the independence from egg supply and the possibility of rapid process adjustment to better match supply with vaccine demand during pandemics. Moreover, cell culture–derived vaccines bear no risk of anaphylactic reactions caused by egg proteins.
Previously, we have shown production cell line–specific HA N-glycosylation patterns (32–34). Because it is unclear whether and to which extent HA N-glycosylation impacts immunogenicity, in this study we investigated the effect of different cell culture–derived influenza A virus PR/8/34 (H1N1) glycovariants on T cell activation. Two adherent cell lines, MDCK and Vero, were selected for influenza virus production, as these are so far the only ones having passed all clinical trials. In a TCR transgenic mouse model, we show that differential HA N-glycosylation markedly affected T cell activation and effector cytokine production in vitro and moderately influenced IL-2 production in vivo. Next-generation pyrosequencing validated the congruence of potential HA N-glycosylation sites as well as the presence of the HA peptide recognized by the TCR of the transgenic T cells. The difference in HA-specific T cell activation induced by the glycovariants in vitro was mediated by CD11c+ DCs. Native virus deglycosylation by a mixture of endo- and exoglycosidases dramatically reduced cytokine production by splenocytes, indicating the crucial importance of HA N-glycosylation for immunogenicity. In vivo relevance of deglycosylation was demonstrated by a marked reduction of HA-specific Ab levels upon immunization of mice with the deglycosylated virus preparations. Thus, our findings may have implications for optimized chicken egg influenza vaccine production as well as for innovative cell line–based influenza vaccine design.
Materials and Methods
Adherent MDCK (no. 841211903; European Collection of Cell Cultures, Salisbury, U.K.) and Vero (no. 88020401; European Collection of Cell Cultures) cells were cultivated in closed roller bottles at 37°C until confluence. Virus was produced using MDCK or Vero cell–adapted influenza A virus PR/8/34 (H1N1, Amp. 3138; Robert Koch Institute, Berlin, Germany) virus seeds as published before (32, 34). Virus-containing supernatant was harvested and cleared 96 h after infection by step gradient centrifugation (100 × g for 20 min, 4,000 × g for 35 min, 10,000 × g for 45 min) and inactivated using β-propiolactone (β-PL) (12, 35). Virus isolation was performed at an average of 70,714 × g (31,000 rpm, type 70Ti rotor; Beckman Coulter, Brea, CA) for 90 min at 4°C. The virus-containing pellet was washed in ∼32 ml 100 mM Tris (pH 7) and finally resuspended in 100 mM Tris (pH 7). Virus preparations were stored at −80°C.
An aliquot of the MDCK and Vero cell–adapted influenza A virus PR/8/34 (H1N1) was natively deglycosylated in solution. All buffers and enzymes used for deglycosylation were purchased from Sigma-Aldrich (Steinheim, Germany) unless otherwise stated. After ultracentrifugation, the virus pellet was resuspended in 160 μl virus infection medium, and 6.7 μl protease inhibitor (40×, no. 11777700; Roche, Mannheim, Germany), 50 μl reaction buffer (no. R9150), 10 μl endoglycosidase F2 (no. E0639), 10 μl endoglycosidase F3 (no. E2264), and 10 μl α-galactosidase (no. G8507) were added and the mixture was shaken at 450 rpm and 37°C for 24 h in the dark. Then, 10 μl reaction buffer (no. R9025) and 10 μl endoglycosidase F1 (no. E9762) were added and shaking was continued at 450 rpm at 37°C for 24 h in the dark. In the following, 10 μl reaction buffer (no. R0266), 10 μl α-mannosidase (no. M7257), 10 μl α-neuraminidase (no. N8271), 10 μl β-N-acetylglucosaminidase (no. A6805), 20 U β-galactosidase (no. G5160), 2 μl α-galactosidase, and 2 μl endoglycosidase F3 were added and the mixture was again shaken at 450 rpm and 37°C for 24 h in the dark. As before, the virus was isolated by ultracentrifugation at 31,000 rpm and 4°C for 90 min. The pelleted virus was resuspended in 100 mM Tris (pH 7) and stored at −80°C. Total virus protein was quantified by bicinchoninic acid assay (Thermo Scientific, Rockford, IL). Additionally, protein concentrations were determined by a 10% nonreducing SDS-PAGE (data not shown). Band intensities were analyzed using the open source imaging software ImageJ 1.45 (36).
N-glycosylation pattern analysis
N-glycosylation pattern analysis was performed based on a method developed by Laroy et al. (37). For analysis, samples were prepared as described previously (33, 38) with some optimizations and modifications. These comprise the substitution of 20 mM NaHCO3 (aqueous) with 50 mM NH4HCO3 (aqueous) during enzymatic in gel deglycosylation and N-glycan extraction as well as the reduction of the final sample volume for peptide-N-glycosidase F digestion to 60 μl, which resulted in an increased enzyme concentration. Furthermore, labeled samples were desalted and excess label was removed by hydrophilic interaction liquid chromatography as described before (39).
Finally, HA N-glycosylation patterns were analyzed as described before (32–34, 38) using an ABI Prism 3100-Avant genetic analyzer (Applied Biosystems), allowing multiplex capillary gel electrophoresis with laser-induced fluorescence detection (xCGE-LIF). In an HA N-glycosylation fingerprint, one peak corresponds to at least one distinct N-glycan structure. Overlays of N-glycosylation fingerprints allowed a direct comparison of different N-glycan pools. All data were processed as published previously (38, 39).
The DNA for full-length genome sequencing using the Roche/454 genome sequencer (GS) FLX (Roche) was prepared according to the protocol of Höper et al. (40). The equimolar pool of PCR amplicons was fragmented according to the manufacturer’s instructions for preparation of shotgun sequencing libraries. The fragmented DNA was then converted to a GS FLX Titanium Library with the SPRIworks Fragment Library System II (Beckman Coulter, Krefeld, Germany) using a SPRIworks Fragment Library kit II (Beckman Coulter) and a GS FLX Titanium Rapid Library molecular identifier adaptor (Roche). The resulting library was quantified with the aid of the KAPA Library Quant Roche 454 Titanium Universal Kitsystem (Kapa Biosystems, Cape Town, South Africa) on a Bio-Rad CFX96 real-time PCR system (Bio-Rad, München, Germany). The libraries were then clonally amplified in the emulsion PCR system with 0.08 copies per bead. Finally, the amplified library was sequenced with the GS FLX using Titanium chemistry and the appropriate instrument run protocol. Raw sequencing data were analyzed with the GS FLX software suite (version 2.5.3; Roche). During sequence assembly, primer sequences were trimmed off the raw data according to the protocol (40). The HA sequence is deposited in the Global Initiative on Sharing All Influenza Data EpiFlu database (http://www.gisaid.org) with the accession number EPI351614. For quasispecies analyses, a mapping of the raw sequencing reads along the reference sequence using the GS FLX reference mapper software (version 2.5.3; Roche) was performed.
Animal experiments were performed in strict accordance with the German regulations of the Society for Laboratory Animal Science and the European Health Law of the Federation of Laboratory Animal Science Associations. The protocol was approved by the Landesamt für Gesundheit und Soziales (Berlin, Germany; permit no. G0259/11). All efforts were made to minimize suffering.
Mice were kept in the animal facility of the Federal Institute for Risk Assessment under specific pathogen-free conditions. TCR-HA transgenic mice were kindly provided from Dr. W. Hansen (Essen, Germany). These transgenic mice express a TCRαβ specific for the peptide 110–120 from HA of influenza A virus A/PR/8/34 (H1N1) presented by I-Ed MHC class II molecules (41). BALB/c mice were bred at the Max Planck Institute for Infection Biology (Berlin, Germany).
Whole spleen cell stimulation
Spleens were removed, cells were flushed out, and RBCs were lysed by addition of ammonium chloride. For Ag-specific T cell stimulation, 2 × 105 spleen cells from TCR-HA transgenic mice or, as control, from BALB/c wild-type mice were stimulated with different protein concentrations of the fully glycosylated or deglycosylated influenza A virus PR/8/34 (H1N1) MDCK cell–derived (M-variant) or Vero cell–derived influenza virus (V-variant) (0.01, 0.1, and 1 μg/ml). HA110–120 peptide (0.1, 1, and 10 μg/ml) was used as positive control, and 10 μg/ml OVA323–339 peptide (Anaspec, Seraing, Belgium) was used as negative control. After 48 h, cells were analyzed for the expression of T cell activation markers (CD69, CD25) by flow cytometry. A portion of cells was first incubated with anti-CD16/32 (BD Biosciences) at 4°C for 30 min, washed, and then stained with cell marker–specific Abs at 4°C for 30 min. Data were acquired on a FACSCanto II flow cytometer (BD Biosciences) and analyzed with the FlowJo analysis software (Tree Star, Ashland, OR). The cytokine concentrations of IL-2, IFN-γ, and IL-4 in cell supernatants were quantified by ELISA (PeproTech, Hamburg, Germany) according to the manufacturer’s protocol.
LPS contamination in the β-PL–inactivated virus preparations was checked using a gel clot Limulus amebocyte lysate kit (Lonza, Cologne, Germany) to rule out potential contaminations. The endotoxin detection level of this assay was 0.06 endotoxin unit/ml. The virus preparations were negative for all concentrations used in the cell stimulation assay.
Stimulation studies with isolated spleen cell subsets
CD11c+ and CD19+ cells were isolated from spleen of BALB/c mice by MACS using CD11c or CD19 microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer’s instructions. The eluate of the CD11c+ MACS was applied on a second column to enhance cell purity. In the end, a purity of ∼80% was achieved (data not shown). TCR-HA transgenic T cells were purified from spleen by negative selection using the MACS Pan T cell isolation kit II (Miltenyi Biotec) according to the manufacturer’s protocol. T cell purity was >95% (data not shown).
For Ag-specific T cell stimulation, 1 × 104 CD11c+ cells were first pulsed at 37°C for 1 h with the Ags mentioned above. In the next step, 1 × 105 purified TCR-HA T cells were added and cells were incubated at 37°C for 48 h. T cell activation was analyzed as described before.
In vivo studies
To analyze the immune stimulatory effect of HA N-glycosylation in vivo, female 6- to 8-wk-old BALB/c wild-type mice were immunized i.p. with 10 μg fully glycosylated or deglycosylated M- and V-variant or with PBS as control. A boost immunization with the same amount of Ag was done on day 14. On day 28, CD4+ spleen T cells were analyzed for IL-2 and IFN-γ production by ELISPOT (Abs were purchased from BD Biosciences) after restimulation with 20 μg/ml HA110–120 peptide for 18 h. A freshly prepared 3-amino-9-ethylcarbazole substrate solution (Sigma-Aldrich) was used for development. Spots were recorded and analyzed using Bioreader 5000 Pro-E (BioSys, Karben, Germany). HA-specific IgG/IgM Ab levels in sera were measured by ELISA on days 14 and 28. Plates were coated with 10 ng/well recombinant influenza A virus PR/8/34 H1N1 HA (Sino Biological, Beijing, China). Sera were also tested for HA inhibition (HAI) activity as previously published (42).
For further analysis of the immunogenic properties of the M- and V-variant, TCR-HA transgenic T cells (purified by MACS) were adoptively transferred into female BALB/c wild-type mice. Before injection, cells were labeled with the cell proliferation dye eFluor 670 (eBioscience). On day 0, 1.5 × 107 cells in 150 μl PBS were adoptively transferred by i.v. injection in the lateral tail vein. Mice were immunized i.p. with 50 μg M- or V-variant on the following day. On day 5, CD4+eFluor 670+ spleen cells were analyzed for proliferation by flow cytometry. Furthermore, cytokine production of spleen cells was measured by ELISPOT as described earlier.
Statistical analyses were performed with Student t test. An unpaired t test was used for analyzing data within one experiment. A paired t test was used when results were compared across different experiments. All statistical analyses were performed with the Prism software (GraphPad Software).
Genomic composition of Ags
To verify the presence of the HA110–120 peptide from influenza A virus PR/8/34 (H1N1), the MDCK and Vero cell–adapted virus seeds were sequenced by next-generation pyrosequencing. The sequence for the HA110–120 peptide was detected for both virus seeds and it was not altered in any detected subpopulation. Because pyrosequencing allows for the characterization of the quasispecies composition, it was also applied to check for homogeneity of virus seeds. As published before (34), the MDCK cell–adapted virus seed was homogeneous for HA composition (i.e., no other HA sequences were detected), indicating that the seed only consists of one population. The Vero cell–adapted virus seed (32) had the same consensus sequence as did the MDCK cell–adapted virus seed. Pyrosequencing revealed a subpopulation of 40% with a deletion of isoleucine at position 338 (I338–) and a subpopulation of 11% with a substitution of valine by methionine at position 459 (V459M). Furthermore, a silent point mutation at position 364 (G364G) was present in a subpopulation of 12% (Table I). Both the deletion I338– and the substitution V459M are located in the stem region of the HA molecule within the fusion subdomains (43). The deletion I338– is located only a few residues before the HA1–HA2 cleavage site whereas the substitution V459M is located within the big α helix of the HA2 subunit. No potential N-glycosylation site was altered by virus adaptation from MDCK to Vero cells, neither in the HA (Table I) (34) nor in the neuraminidase molecule (data not shown). These results confirm that the altered amino acids in MDCK and Vero cell–derived HA did not affect HA N-glycan recognition by host immune cells.
|Segment .||Coded Protein .||Base Pair Substitution .||Amino Acid Substitution .||Vero Cell–Adapted Virus Seed (%) (32) .|
|Segment .||Coded Protein .||Base Pair Substitution .||Amino Acid Substitution .||Vero Cell–Adapted Virus Seed (%) (32) .|
The deletion I338– is located in the HA1 molecule, only a few amino acids before the HA1–HA2 cleavage site between amino acid 344(R)–345(G). The silent point mutation at G364G and the substitution V459M are located in the HA2 molecule. The deletion I338– as well as the substitution are both located in the stem region of the HA molecule.
Impact of N-glycosylation on immunogenicity
Strict host cell specificity of HA N-glycosylation was confirmed. As reported previously (33), HA N-glycosylation from MDCK and Vero cell–derived virus differs significantly (Fig. 1Ai, 1Bi). Whereas the M-variant HA of influenza A virus PR/8/34 (H1N1) contains tri- and tetra-antennary as well as bisecting N-acetylglucosamine structures, the V-variant HA exhibits fewer N-glycan structures and mainly with a low molecular mass. Furthermore, the V-variant HA carries terminal β-galactose and high-mannose structures, whereas the M-variant HA displays terminal α- and β-galactose and no high-mannose glycan structures (Fig. 1Ai, 1Bi, and Ref. 33). To analyze whether this difference in the HA N-glycosylation patterns influences T cell stimulation, TCR-HA transgenic spleen cells were incubated with β-PL–inactivated preparations of the influenza A virus PR/8/34 (H1N1) M- or V-variant. T cells from TCR-HA transgenic mice have a TCRαβ specific for the HA110–120 peptide presented by I-Ed MHC class II molecules (41). Therefore, it was possible to investigate specifically how CD4+ T cell stimulation was affected by differential HA N-glycosylation. Flow cytometry revealed that a higher frequency of splenic T cells expressed the activation marker CD69 when stimulated with the V-variant (Fig. 2A). This cell surface glycoprotein is upregulated very early during lymphoid activation and represents a costimulatory molecule involved in lymphocyte proliferation (44). The observed difference in CD69 expression on T cells stimulated with the M- or V-variant was statistically significant when the results of three independent experiments were combined (Fig. 2B). In contrast, there was no significant difference in the expression of the T cell activation marker CD25 (data not shown). Next, we analyzed cytokine production by splenocytes to investigate whether the influenza A virus PR/8/34 (H1N1) M- and V-variant induced distinct cytokine profiles. Consistent with the CD69 expression analysis, IL-2 was produced at a significantly higher level by splenocytes incubated with the V-variant than cells incubated with the M-variant (Fig. 3A). Thus, HA N-glycosylation of the V-variant had a marked impact on IL-2 production. For IFN-γ, the same tendency as for IL-2 was observed, albeit to a lesser extent, whereas no difference was observed for IL-4 (Fig. 3B, 3C). These findings suggest that the initiated T cell response is markedly influenced by differential HA N-glycosylation.
Next, we investigated the kinetics of T cell activation by the two glycovariants. The difference in CD69 expression was already detectable after 24 h but no longer after 72 h (Supplemental Fig. 1). The production of IL-2 did not yet differ after 24 h (Supplemental Fig. 2A). However, splenocytes stimulated with the V-variant produced still significantly higher IL-2 levels after 72 h compared with stimulation with the M-variant (Supplemental Fig. 2B, 2C). This indicates that the influenza A virus PR/8/34 (H1N1) V-variant promotes a faster T cell activation than does the M-variant and thus enhances T cell proliferation.
Differential T cell activation is mediated by CD11c+ DCs
To identify the APC population responsible for the distinct stimulatory effect of the influenza A virus PR/8/34 (H1N1) M- and V-variant, CD11c+ cells were separated from spleen, incubated with the M- or V-variant, and cocultivated with TCR-HA transgenic T cells. The same effects in CD69 expression (Fig. 4A, Supplemental Fig. 3), IL-2 production (Fig. 4B), as well as IL-4 and IFN-γ production (Supplemental Fig. 4) were observed as in the whole spleen cell assay. This finding indicates that CD11c+ DCs play a role in the recognition of the two glycovariants, in CD4+ T cell activation, and in acquisition of T cell effector functions. We also investigated whether other subsets of spleen cells contributed to differential T cell stimulation. However, cocultivation of TCR-HA transgenic T cells with other spleen cell subsets led to no or very weak T cell activation (data not shown). Thus, the differences in T cell activation induced by the influenza virus glycovariants were indeed mediated by CD11c+ DCs.
Deglycosylation of the influenza virus HA variants dramatically affects immunogenicity
To further investigate the effect of HA N-glycosylation on T cell stimulation, the MDCK and Vero cell–derived glycovariants of influenza A virus PR/8/34 (H1N1) were natively deglycosylated using a variety of endo- and exoglycosidases. HA bands were shifted to lower molecular masses in the SDS-PAGE, indicating successful deglycosylation (data not shown). Bands of fully glycosylated HA monomers were found at ∼70 kDa (MDCK-derived just above and Vero-derived just below). In contrast, deglycosylated preparations exhibited a more diffuse and broader band pattern just below the fully glycosylated band reaching to the nucleoprotein at ∼55 kDa (data not shown). N-glycosylation pattern analysis confirmed that both variants were deglycosylated to most parts (Fig. 1Aii–iii, 1Bii–iii). The significant reduction of signal intensity from ∼450–500 relative fluorescence units (RFU) to <50 RFU confirmed efficient protein deglycosylation. Although no complete deglycosylation was achieved, glycan analysis showed a reduction of N-glycosylation by at least a factor of 10. Moreover, glycan analysis suggested that Vero cell–derived N-glycan structures were removed more efficiently (Fig. 1Bii) than MDCK cell–derived structures (Fig. 1Aii) (maximum RFUVero < maximum RFUMDCK). Interestingly, deglycosylation resulted in multiple truncated glycan structures of lower molecular mass on the Vero cell–derived HA (marked with an asterisk in Fig. 1Bii). In contrast, on the MDCK cell–derived HA only one shorter glycan structure was detected after deglycosylation (marked with an asterisk in Fig. 1Aii).
To investigate whether HA deglycosylation affected immunogenicity, again TCR-HA transgenic splenocytes were stimulated with the deglycosylated influenza virus variants (Fig. 5). Deglycosylation led to a dramatically reduced T cell activation as measured by the frequency of CD4+ T cells expressing CD69 (Fig. 5A) and cytokine production (Fig. 5B–D) after stimulation with the V-variant. For splenocyte stimulation with the M-variant, the same tendency was observed, albeit to a lower extent (Fig. 5). This might be due to the fact that HA deglycosylation of the M-variant reduced the HA N-glycosylation level by ∼90% but without modifying most N-glycan structures. In contrast, N-glycans of the V-variant HA were truncated during deglycosylation, leading to multiple new glycan structures in addition to the reduced level of glycosylation (Fig. 1Bii). However, for both glycovariants of influenza A virus PR/8/34 (H1N1) a significantly diminished frequency of CD4+CD69+ T cells and decreased levels of IL-2, IL-4, and IFN-γ were observed upon deglycosylation, particularly at lower protein concentrations (Fig. 5). Interestingly, the reduction of the T cell proliferation cytokine IL-2 was more pronounced compared with the effector cytokines IL-4 and IFN-γ. This finding suggests that the N-glycosylation pattern of the V-variant may lead to a faster recognition and virus uptake by APCs and influences the kinetics of T cell activation rather than changing the quality of the initiated immune response.
To confirm the crucial role of CD11c+ DCs, we analyzed the effect of the HA deglycosylation in the CD11c+/TCR-HA T cell cocultivation assay. Consistent with the results in the whole spleen cell assay, IL-2, IFN-γ, and IL-4 levels were dramatically reduced after HA deglycosylation (Supplemental Fig. 4). The reduction in cytokine levels was even more pronounced than in the whole spleen cell assay. This finding further confirmed the crucial role of CD11c+ DCs for HA N-glycan recognition and T cell activation. In conclusion, these observations clearly indicate that HA N-glycosylation of the influenza A virus PR/8/34 (H1N1) has a marked impact on virus immunogenicity.
In vivo studies
To investigate the immune stimulatory effect of HA N-glycosylation in vivo, wild-type mice were immunized with the fully glycosylated or deglycosylated M- or V-variant. On days 14 and 28, HA-specific Ab levels in sera were analyzed by ELISA. Consistent with the in vitro results, anti-HA Ab levels were dramatically reduced in sera of mice immunized with the deglycosylated preparations compared with mice immunized with the fully glycosylated M- or V-variant (Fig. 6A, left). An HAI assay performed with serum samples collected on day 28 confirmed these results (Fig. 6A, right). Accordingly, the frequency of IFN-γ–producing CD4+ T cells was slightly lower for spleen cells from mice immunized with the deglycosylated virus preparations upon restimulation with the HA110–120 peptide (data not shown). The number of IFN-γ–producing CD4+ T cells in spleens of mice immunized with the V-variant was higher compared with the M-variant although values did not reach statistical significance (data not shown). Thus, the crucial effect of HA N-glycosylation was demonstrated by a dramatic reduction of HA-specific Ab levels in mice immunized with the deglycosylated virus variants. Interestingly, HA-specific Ab levels between the M- and V-variant differed on day 14 when the M-variant induced significantly higher Ab levels (Fig. 6A, left). This effect was still observed on day 28 when sera of immunized mice were tested for hemagglutination inhibition activity by HAI assay (Fig. 6A, right). This finding might suggest that the glycosylation pattern of the M-variant induces a bias toward a humoral immune response whereas the V-variant promotes cellular immune responses as described before.
To address the role of HA N-glycosylation on T cell proliferation and cytokine production further in vivo, TCR-HA transgenic T cells were labeled with the cell proliferation dye eFluor 670 and were adoptively transferred into BALB/c wild-type mice. This allows for analysis of the fate of HA110–120 peptide–specific T cells in natural conditions. Subsequently, mice were immunized with the β-PL–inactivated virus glycovariants and the proliferation of the transferred T cells was analyzed 5 d after adoptive T cell transfer (Fig. 6B, 6C). Furthermore, spleen cells were restimulated with HA110–120 peptide and the frequency of cytokine-producing cells was determined using ELISPOT assay (Fig. 6D). Although proliferation of T cells was similar after immunization with the two glycovariants (Fig. 6C), the frequency of IL-2–producing splenocytes was increased in mice immunized with the V-variant (Fig. 6D). In conclusion, the findings observed in the TCR-HA transgenic system and upon immunization of wild-type mice indicate that differential HA N-glycosylation impacts immunogenicity in vitro and in vivo.
This study showed that differences in influenza HA N-glycosylation caused by virus production in different host cell lines have a marked impact on immunogenicity in vitro and in vivo. Furthermore, this study emphasizes the advantage of combining next-generation pyrosequencing, high-throughput multiplex CGE-LIF–based HA N-glycosylation pattern analysis, and immunogenicity assays to identify optimal (i.e., highly immunogenic) HA N-glycosylation patterns or single N-glycan structures for targeted influenza vaccine design.
In this study, we analyzed β-PL–inactivated influenza A virus PR/8/34 (H1N1) glycovariants of two host cell lines, MDCK and Vero, for their HA N-glycosylation pattern and immunogenicity. First and foremost, these cell lines were selected because they are either already licensed for human influenza vaccine manufacturing or have at least successfully passed clinical trial III studies. Although we demonstrate the impact of HA N-glycosylation on immunogenicity for only two cell lines, screening for highly immunogenic N-glycosylation patterns can easily be extended to a variety of other virus expression systems by using the platform presented in this study. In previous studies, we showed that the host cell line indeed influences HA N-glycosylation patterns (32–34). Whereas the influenza A virus PR/8/34 (H1N1) M-variant abundantly displays larger, tri- or tetra-antennary, α- and β-galactose–terminated N-glycan structures, the V-variant carries smaller, exclusively β-galactose–terminated glycan structures and also some high-mannose N-glycans (33).
Whole spleen cell stimulation using a TCR-HA transgenic system revealed that HA N-glycosylation significantly affected early T cell activation and proliferation in vitro. In this assay, the V-variant induced a markedly higher frequency of CD4+ T cells expressing the activation marker CD69 than did the M-variant. Moreover, splenocyte stimulation with the V-variant led to significantly increased IL-2 levels. Interestingly, the production of the effector cytokines IFN-γ and IL-4 was less affected than IL-2 release. IL-2 is a cytokine that is typically produced by Th0 cells, but also by Th1 cells, after Ag stimulation and promotes T cell growth, differentiation, and survival (45). That the change in IL-2 production was more pronounced than levels of Th1/Th2 effector cytokines suggests that the V-variant HA N-glycosylation patterns facilitate faster recognition and virus uptake by APCs. In tendency, the V-variant also induced a higher production of the Th1 cytokine IFN-γ by splenocytes. This finding might be of importance because it is known that induction of a potent Th1 immune response is essential for viral clearance (46).
Furthermore, we demonstrated in DC/T cell cocultivation assays that the difference in T cell priming based on differential HA N-glycosylation was mediated by CD11c+ DCs. As a bridge to adaptive immunity, DCs transport Ags from the site of infection to the regional lymph nodes or to the spleen where they activate Ag-specific CD4+ T cells and CD8+ T cells by cross-presentation (47). CD4+ T cells generally lead to optimal activation of CD8+ T cells and promote maturation of Ab-producing B cells (48). In this study, the crucial role of DCs in recognition of the glycovariants was further confirmed by DC/T cell cocultivation assays with deglycosylated virus preparations where T cell activation was almost completely abrogated.
A number of DC subsets are involved in innate and adaptive immunity and thus might be responsible for the different effect on T cell activation by the two influenza A virus PR/8/34 (H1N1) glycovariants. A previous study indicated that CD103+ DCs and CD11bhigh DCs uptake influenza virus particles in the lung and transport them to the draining mediastinal lymph nodes (49, 50). CD8α+ DCs were reported to play a crucial role in CTL priming during influenza infection (51). However, this might rather be relevant for influenza infections of mice, as only a very limited number of CTL epitopes within HA are known in humans (52). Recently, a study using a human challenge model revealed that memory CD4+ T cell populations correlated with less severe illness (53). Thus, further experiments with human DC populations and PBMCs are needed to determine the relevance of HA N-glycosylation on influenza virus immunogenicity in humans.
Additionally, DCs are known to express high levels of pattern recognition receptors, particularly C-type lectin receptors (CLRs) (54). Our finding that in vitro the influenza A virus PR/8/34 (H1N1) V-variant exhibits a more immunogenic N-glycosylation pattern than does the M-variant suggests differential glycan recognition by DCs. Because some of the V-variant N-glycans are of the high-mannose type (33), mannose-rich HA N-glycan structures may be recognized by CLRs expressed by splenic CD11c+ DCs. This is in accordance with studies elegantly demonstrating the involvement of the macrophage mannose receptor (MMR) in macrophage infection by the influenza virus (21, 55). In a recent study, the macrophage galactose-type lectin (MGL) was identified as a second receptor besides MMR for virus recognition by host cells (56). Binding of influenza virus to the CLRs MMR and MGL was independent of sialic acid and was mediated by Ca2+-dependent recognition of viral glycans by the carbohydrate recognition domains of MMR and MGL (56). An additional lectin reported to play a role in influenza virus clearance from body fluids is SP-D that may be involved in glycan recognition (20, 22, 57). An alternative mechanism that may also account for differential immunogenicity of the influenza virus glycovariants is masking of antigenic epitopes by HA N-glycosylation (9, 25). In all of these studies, except for one (9), live influenza virus was used for the experiments. However, because glycan-binding receptors on the APC surface may contribute to virus endocytosis, but may also serve as first attachment receptors for virus entry (51), we used β-PL–inactivated virus preparations in this study to decipher the role of N-glycosylation for immunogenicity. In a recent study published during revision of this manuscript, immunogenicity of influenza HA N-glycosylation was investigated by using HA recombinantly expressed in HEK or in insect cells (58). The authors showed that the N-glycan structure impacted HAI Ab titers in chicken and mice.
In this study, we used natively deglycosylated virus preparations of two glycovariants and showed in vitro that T cell activation significantly diminished compared with fully glycosylated HA. This effect was observed for both variants as seen in a markedly decreased expression of CD69 as well as in reduced cytokine levels, albeit more pronounced for the V-variant. Glycoanalysis revealed that both influenza A virus PR/8/34 (H1N1) glycovariants exhibited a markedly reduced N-glycosylation (by at least 90%), but were not completely deglycosylated. Although similar N-glycan structures were still present on the deglycosylated MDCK cell–derived HA, new, truncated glycan structures occurred on the deglycosylated Vero cell–derived HA. Because T cell activation was almost completely abrogated upon stimulation with the deglycosylated V-variant, the removed terminal carbohydrate moieties such as β-galactose or high-mannose structures seem to be crucial for the immune stimulatory effect. Overall, deglycosylation had a dramatic effect on T cell activation, thus confirming the essential role of influenza virus HA N-glycosylation for immunogenicity.
The differential effect of the M- and V-variant on T cell priming observed in vitro was also relevant in vivo. Consistent with the in vitro results, HA-specific Ab levels were dramatically reduced in serum of mice immunized with the deglycosylated virus preparations. These findings revealed that HA N-glycosylation is also crucial for the initiation of the humoral immune response. Analysis of CD4+ T cell responses by restimulation of splenic T cells with the HA110–120 peptide also showed that native virus deglycosylation led to a reduced number of IFN-γ–producing CD4+ T cells. In contrast, CD8+ T cell responses were generally very low (data not shown). Interestingly, 14 d after the prime immunization, mice immunized with the M-variant had higher HA-specific Ab levels than mice immunized with the V-variant. HA inhibition activity was still significantly higher upon immunization with the M-variant on day 28. This might reflect a differential influence of HA N-glycosylation on humoral and cellular immune responses. This finding is in accordance with a recent study showing that HA carrying complex glycan structures or only single N-acetylglucosamine residues induces higher HAI Ab titers than does HA carrying high-mannose glycan structures (58).
To address T cell proliferation and cytokine production further in vivo, we adoptively transferred TCR-HA transgenic T cells into wild-type mice and immunized them with the M- or V-variant. In this set-up, a higher frequency of transferred TCR-HA transgenic T cells produced IL-2 when mice were immunized with the V-variant. Overall, these findings confirm the in vitro results in that the glycans of the V-variant led to a faster activation of T cells in vivo than did the ones of the M-variant. In contrast, the M-variant HA N-glycosylation pattern showed a stronger effect on the humoral immune response. The effects of HA N-glycosylation on immunogenicity in vivo are remarkable, but its impact on vaccine potency will need to be addressed more in detail in further studies. The focus of this study was to demonstrate the utility of the platform presented and to provide evidence that HA N-glycosylation potentially affects influenza vaccine efficacy.
Without doubt, there is urgent need for new production processes of potent influenza vaccines, as the classical influenza vaccine production in embryonated chicken eggs exhibits some drawbacks such as a long production time, dependence on egg supply, and the risk of anaphylactic reactions caused by egg proteins. Because production upscale in a timely fashion is a key challenge of today’s influenza vaccine manufacturing, particularly during pandemics, cell culture–based influenza vaccines indeed represent an efficient alternative. Our findings have important implications for cell-based influenza vaccine design. The platform presented in this study allows for a rapid and easy screening of N-glycosylation patterns and correlates them with immunogenicity. Thus, appropriate host cell lines can be selected for virus production that may even be glyco-designed for optimal N-glycosylation patterns by genetic engineering approaches (59).
We thank Uwe Vogel and Moctezuma Reimann for expert technical assistance and Prof. Dr. Thomas Schüler from the Institute for Immunology of the Charité Berlin for fruitful discussions. We also thank Dr. Boris Hundt from IDT Biologika (Dessau, Germany) for constructive collaboration and kind support as well as Sandra Meißner and Dr. Ralf Dürrwald from IDT Biologika for performing the HAI assays. We are also very grateful to Dr. Wiebke Hansen, University Hospital Essen, for providing the TCR-HA transgenic mice.
This work was supported by the Max Planck Society, German Federal Ministry of Education and Research Grant 0315446 (to B.L.), funding from the European Union's Seventh Framework Programme (FP7-Health-F5-2011) under grant agreement 278535 “HighGlycan” (to E.R.), and Collaborative Research Centre Grant 765 (to P.H.S. and B.L.).
The sequences presented in this article have been submitted to the Global Initiative on Sharing All Influenza Data EpiFlu database (http://www.gisaid.org) under accession number EPI351614.
The online version of this article contains supplemental material.
Abbreviations used in this article:
capillary gel electrophoresis with laser-induced fluorescence detection
C-type lectin receptor
Madin–Darby canine kidney
macrophage galactose-type lectin
macrophage mannose receptor
Madin–Darby canine kidney cell–derived influenza virus
relative fluorescence unit
surfactant protein D
Vero cell–derived influenza virus.
The authors have no financial conflicts of interest.