Unravelling the Role of O-glycans in Influenza A Virus Infection

The initial stage of host cell infection by influenza A viruses (IAV) is mediated through interaction of the viral haemagglutinin (HA) with cell surface glycans. The binding requirement of IAVs for Galβ(1,4)Glc/ GlcNAc (lactose/lactosamine) glycans with a terminal α(2,6)-linked (human receptors) or α(2,3)-linked (avian receptors) N-acetylneuraminic residue commonly found on N-glycans, is well-established. However the role and significance of sialylated Galβ(1,3)GalNAc (core 1) epitopes that are typical O-glycoforms in influenza virus pathogenesis remains poorly detailed. Here we report a multidisciplinary study using NMR spectroscopy, virus neutralization assays and molecular modelling, into the potential for IAV to engage sialyl-Galβ(1,3)GalNAc O-glycoforms for cell attachment. H5 containing virus like particles (VLPs) derived from an H5N1 avian IAV strain show a significant involvement of the O-glycan-specific GalNAc residue, coordinated by a EQTKLY motif conserved in highly pathogenic avian influenza (HPAI) strains. Notably, human pandemic H1N1 influenza viruses shift the preference from ‘human-like’ α(2,6)-linkages in sialylated Galβ(1,4)Glc/GlcNAc fragments to ‘avian-like’ α(2,3)-linkages in sialylated Galβ(1,3)GalNAc without involvement of the GalNAc residue. Overall, our study suggests that sialylated Galβ(1,3)GalNAc as O-glycan core 1 glycoforms are involved in the influenza A virus life cycle and play a particularly crucial role during infection of HPAI strains.


Results
Here, we report the role of Sialyl-Galβ(1-3)GalNAc in IAV infection at a molecular and atomic level by a multidisciplinary experimental approach. Significant variations between virus strains were detected including the importance of a non-sialic acid glycan residue. This study provides more detailed insights into the complexity of influenza virus-host-interactions and suggests that Sialyl-Lac/LacNAc are not the only influenza virus receptors. First, a histochemical analysis using antibodies to evaluate the expression of sialyltransferases (ST), especially ST6GalNAc1 and ST6GalNAc2 that are essential in the biosynthesis of core 1 O-linked glycans ( Fig. 1C) 11,12 , was carried out. Both, type 1-like pneumocytes and A549 human lung cell lines, showed strong expression of ST6GalNAc1 but little ST6Gal1 or ST6GalNAc2 expression ( Supplementary Fig. 1A), indicating that binding of human IAV would preferentially use Sialyl-O-glycan motifs such as 6-sialyl-Galβ(1-3)GalNAc and 3,6-disialyl-Galβ(1-3)GalNAc. The low expression of ST6Gal1, predominantly involved in the biosynthesis of sialylated N-glycans, suggests that O-glycans might play a role in IAV cell recognition. Immunohistochemical analysis of normal adult lung and bronchial tissues also showed an elevated expression of ST6GalNAc1 in bronchus and bronchioles, with ST6Gal1 expression in the bronchial epithelium. Upper respiratory nasopharyngeal tissue staining showed positive punctate staining for ST6GalNAc1 and ST6Gal1 in the cytoplasm while ST6GalNAc2 was mainly confined to infiltrating inflammatory cells ( Supplementary Fig. 1B).

Influenza virus neutralization assays reveal a significant strain variation between H1 and H3 subtypes with regard to O-glycan binding.
To investigate the role of Sialyl-Galβ(1,3)GalNAc fragments during IAV cell attachment and infection, virus neutralization assays were performed in the presence of soluble glycans at various concentrations. The glycans act as decoy receptors by occupying the receptor binding sites of the HA protein, affecting the binding ability of the virus to the cell receptors depending on the affinity of the soluble glycans. Using a modified focus forming assay protocol, we were able to determine significant differences in IC 50 values for 3-sialyl-Galβ(1,3)GalNAc, 6-sialyl-Galβ(1,3)GalNAc and 3,6-disialyl-Galβ (1,3) GalNAc between the two influenza strains H1N1sea and H3N2 ( Fig. 5

Inhibition of O-glycan biosynthesis in Lec1 cells affects 3045/H5N1 influenza virus infection.
Lec1 CHO cells, which lack the N-acetylglucosaminyl-transferase I (GlcNAcT-I) activity necessary for the formation of complex or hybrid N-glycans 32 , were used to further elucidate the role of O-linked glycans in avian IAV infection. Lec1 cells were also treated with GalNAc-αO-Bn to inhibit their O-glycan glycoprotein but not glycolipid biosynthesis. By comparing the influenza infection rate in the parental CHO (Pro − 5) cells and the Lec1 cells, the significance of O-linked glycans could be ascertained. Fig. 6 and Table 2 show that there was minimal infection by WSN33/H1N1 in the Lec1 cells (0.57% ± 0.34%), but significant infection by 3045/H5N1 (10.57% ± 3.54%) and 1203/H5N1 (10.73% ± 1.58%), similar to previous findings 33     reveals a potential bound conformation of 3-sialyl-Galβ(1,3)GalNAc (Fig. 7a). The Neu5Ac and Gal residues of 3-sialyl-Galβ(1-3)GalNAc maintain key interactions with the protein when compared to 6′SLN ( Supplementary  Fig. 10). In particular, the methyl protons of the acetamido group of Neu5Ac are engaged in strong hydrophobic contacts with the indole ring of Trp-104 ( Supplementary Fig. 11) similar to the carbohydrate-aromatic interactions previously reviewed 34 . In contrast, the GalNAc residue was involved in some interaction with the protein.
The methyl protons of the acetamido group of the GalNAc residue established no direct contact with the protein, which is in excellent agreement with our STD NMR results. This result further suggests that, despite the difference in the Sia glycosidic linkages, the HA protein from the H1N1pdm can accommodate a sialylated Galβ(1,3)GalNAc core without any steric hindrance. A combined docking and MD simulation of the H1N1pdm HA in complex with 6-sialyl-Galβ(1,3)GalNAc was also performed illustrating a 'loose' bound conformation of 6-sialyl-Galβ(1,3) GalNAc with the HA protein (Fig. 7b). The major 6-sialyl-Galβ(1,3)GalNAc-HA interactions can be attributed to the Neu5Ac residue, whereas the remainder of the glycan core does not make any significant contact with the protein. This loose binding conformation of 6-sialyl-Galβ(1,3)GalNAc is likely the reason that H1N1pdm has binding preference for 3-sialyl-Galβ(1,3)GalNAc which is in excellent agreement with the STD NMR experiments demonstrating that 3-sialyl-Galβ(1,3)GalNAc is clearly the preferred bound sialylated Galβ(1,3)GalNAc fragment. We have also docked 3-sialyl-Galβ(1,3)GalNAc into the binding site of the HA of A/Vietnam/1194/2004 (H5N1) (PDB: 3ZNK) 35 . The 3-sialyl-Galβ(1,3)GalNAc structure with the best superimposition of the Neu5Ac residue with that of the crystal structure bound ligand 6-O-sulfated 3′sialyllactosamine [Neu5Acα(2,3)Galβ(1,4) (6-HSO 3 )GlcNAc], was chosen for a subsequent 20 ns MD simulation, which illustrated that GalNAc of 3-sialyl-Galβ(1,3)GalNAc is significantly engaged in interaction with the HA protein (Fig. 7c). In particular, the acetamido group of the GalNAc residue makes a strong polar interaction with Lys-189 as well as a hydrophobic interaction with Leu-190 (Lys-193 and Leu-194 according to H3 numbering). This finding is also in excellent agreement with the STD NMR results showing that the methyl protons of the acetamido group of the GalNAc residue are important for 3-sialyl-Galβ(1,3)GalNAc binding. Likewise, an MD structure for the bound conformation of 6-sialyl-Galβ(1,3)GalNAc was obtained (Fig. 7d). Although its Neu5Ac moiety maintained interactions with the HA protein similar to α(2,3)-linked O-and N-glycans, both the Gal and GalNAc portions of 6-sialyl-Galβ(1,3) GalNAc were more solvent-exposed and established less contacts with the protein than those of 3-sialyl-Galβ(1,3) GalNAc. These observations are consistent with the overall weak absolute STD enhancements of 6-sialyl-Galβ(1,3) GalNAc ( Supplementary Fig. 5d). Our analysis suggests that sialylated Galβ(1,3)GalNAc fragments are potential receptors for avian H5N1 influenza virus due to additional strong interaction with the acetamido group of the GalNAc residue. Our discovery that sialylated Galβ(1,3)GalNAc fragments are potential receptors for H5N1 has promoted us to compare the amino acid sequence of the HA of several HPAI strains (Fig. 8). We identified a sequence motif EQTKLY consisting of six amino acids including the crucial GalNAc-binding residues Lys-193 and Leu-194 that is highly conserved in H5N1, H7N3, H7N7 and H7N9. We postulate that the EQTKLY-motif is essential in coordinating the GalNAc residue in O-glycans for successful binding to HPAI strains. Additionally, we have performed docking studies of the 3,6-disialyl core 1 α-linked to serine that shows that the α-serine residue is completely solvent exposed, supporting the NMR experiments ( Supplementary Fig. 12).
Our results show that sialylated Galβ(1,3)GalNAc glycoforms are not only recognized and bound by several IAV strains, but also affect viral infection cycles. Sialylated core 1 O-glycans have been found previously in the respiratory tract tissues of ferrets and humans 10,11 . In the current study we show expression of ST6GalNAc1, a key enzyme in α(2,6)sialylated O-glycan biosynthesis, in A549 cells and type I pneumocytes, as well as in human nasopharyngeal and lower respiratory tract tissues including bronchus, bronchioles and alveoli.
The observation that the acetamido group of the GalNAc residue of sialyl-core 1 O-glycans is bound with similar affinity to the acetamido group of the Neu5Ac moiety by avian H5, suggests that sialylated Galβ(1,3)GalNAc fragments might be essential receptors for this subtype. This outcome correlates very well with binding data from glycan microarray assays 36,37 and the results reported by Gaunitz and co-workers showing strong binding of H5Vn protein to 3-sialyl-Galβ(1,3)GalNAc by SPR technology 38 . Our structural investigation presented here shows that the GalNAc moiety of the O-glycan core 1 structure is coordinated by . Both amino acid residues lie within the EQTKLY-motif that is highly conserved in HPAI strains but not in human virus HA proteins of the H1 and H3 subtypes. This structural observation correlates well with our NMR finding that proteins of human strains that lack the EQTKLY-motif show only weak, if any, interaction with the acetamido group of the GalNAc moiety and therefore reduced binding. The question of why avian H5N1 influenza viruses are less effective in human transmission may also be explained by the high prevalence of sialylated motifs in mucins. It is likely that the sialylated O-glycan motifs present in mucins restrict access to apical receptors expressed on the respiratory epithelium, suggesting a dual role for exposed mucin O-glycans in masking specific cell-surface receptors and acting as a decoy. It has been reported that mucin O-glycans prevent bacterial adhesion by limiting access to the epithelial cell surface 39 .  The second significant finding of our study is that the presence of the Galβ(1,3)GalNAc disaccharide in sialylated O-glycan motifs causes the human H1N1pdm virus to switch its preference from α(2,6)-linked Neu5Ac to α(2,3)-linked Neu5Ac ('avian receptor'). The different impact of the neighbouring glycan residues on recognition of sialic acid as a receptor for influenza virus has been observed before 40 .  and Leu-194 (H3 numbering) (★) of several HPAI strains. The two residues can be found within a sequence motif consisting of six amino acids EQTKLY (shown in bold) being highly conserved in all HPAI strains used in this alignment (see Supplementary Fig. 13 for the entire sequence alignment). This sequence motif is not conserved in the human IAV strains H1N1pdm, H1N1sea and H3N2 (alignment not shown). The virus neutralization assays also revealed strain-specific differences in the biological importance of sialylated core 1 O-glycan oligosaccharides to influenza H1N1sea and H3N2 subtypes. All investigated O-glycan oligosaccharides showed strong inhibitory potency towards human H3N2 influenza virus. In contrast, H1N1sea infection of MDCK.2 cells was less adversely affected by sialylated core 1 structures; for example the soluble O-glycan 6-Sialyl-Galβ(1,3)GalNAc fragment was approximately 14-fold less effective than 6-SL/SLN in competing with glycan receptors on the host cell surface. However, sialylated O-glycoforms clearly bound to H1N1sea suggesting that they may act like attachment factors to support initiation of the infection cycle. We have demonstrated for the first time that particular avian H5 influenza strains can infect CHO Lec1 cells via O-glycan receptors. Chu and Whittaker postulated that influenza viruses could not infect CHO Lec1 cells due to the lack of complex and sialylated N-glycans 33 , however in that study influenza viruses A/WSN/33(H1N1), Udorn/307/72(H3N2) and influenza B/Yamagata/78 were used, but no avian influenza strains were tested.
In conclusion, our study sheds light on the involvement of the Sialyl-Galβ(1,3)GalNAc O-glycoforms during IAV infection, suggesting a similar important role as Sialyl-Lac/LacNAc motifs for certain influenza strains. Furthermore, it provides the first evidence of a non-sialic acid residue, namely O-glycan-specific GalNAc, playing a crucial role in HPAI pathogenesis. This outcome opens a new direction for structure-assisted drug design and may facilitate the development of potent anti-influenza drugs especially towards HPAI strains that threaten the human population.

Methods
Immunocytochemistry and image analysis. Coverslips were washed with TBS and 0.1% Tween 20 (Sigma P-1379) for 30 min at room temperature (RT) then blocked with 10% normal rabbit serum for 10 min. They were then incubated with 1/25 HB65 for 30 min at RT followed by 1/200 biotinylated rabbit anti-mouse (Dako Cytomation E-0354) for 30 min at RT and incubated with Elite-ABC kit (Vectorlab PK-6100) diluted 1/50 for 30 min at RT. Colour was developed with the Vector Red substrate kit (Vector Labs SK-5100) and counterstaining with Mayer's hematoxylin. Slides were scanned using Aperio Scanscope CS-S microscopic slide scanning system and proprietary image analysis counting the number of positive cells as a percentage of the total cell population with on average 6,000 to 12,000 cells counted/region and three regions per coverslip assessed (mean ± standard deviation). For staining with sialyltransferase antibodies coverslips were transferred to PBS then treated with 0.1% Tween 20 in TBS for 30 min RT then blocked with 2.5% normal horse serum (NH) for 5 min at RT. Primary antibody was applied (ST6GAL1 (R&D System AF5924), ST6GalNAc1 (Novus NBP1-87043), and ST6GalNAc2 (Novus NBP2-13392)) followed by biotinylated horse anti-goat IgG, then streptavidin, alkaline phosphatase conjugated 1/50, and development with VECTOR Red alkaline phosphatase (AP) substrate kit with levamisole solution added followed by counterstaining with Mayer's hematoxylin. Sections were stained using the same primary antibodies following antigen retrieval using microwave sections in 10 mM citrate buffer pH 6.0 at 95 °C for 15 min.
H5Vn-VLP preparation. HEK293T cells were transfected with pCMVdR8.91 (coding for HIV-Gag/Pol gene) and pcDNA-synH5 from a H5N1 known clinical isolate grown in the presence of soluble Vibrio cholerae sialidase (4 mU/mL; Roche). Supernatant was harvested 24 h post-transfection, filtered and concentrated on sucrose cushion (20% w/v of sucrose (Sigma, S-7903) filter sterilized on 0.45 um). After ultra-centrifugation (at 28,000 × g for 2.5 h on Optima L80 XP from Beckman Coulter equipped with rotor SW 32 TI), the pellet was re-suspended into 1/100th of the initial volume of the complete Dulbecco's Modified Eagle's Medium (DMEM/ HIGH, Invitrogen #10569), 5% foetal bovine serum (FBS, Invitrogen #10500) and 1% Penicillin/ streptomycin (Invitrogen #15140). As the haemagglutinin construct contains a flag-tag on its C-terminal end, the content of HA in these VLP's could be established by western-blot using anti-flag M2 peroxidase conjugate monoclonal antibodies (Sigma, A-8592) with a calibrating serial dilution of BAP-Flag protein (Sigma, P-7582).
Virus neutralization assay. The neutralization assay was performed using a modified focus forming assay protocol 31 . 3 × 10 4 MDCK.2 cells were seeded into 96-wells and incubated at 37 °C for 24 hrs. The following steps were performed at 4 °C to inhibit the neuraminidase activity, as the addition of oseltamivir carboxylate (OC) to block neuraminidase was not suitable in these assays because an efficient spread of virus progeny to the neighbouring cells is essential to obtain countable foci for the analysis. Confluent MDCK monolayers were washed with cold DMEM without additives. 100 mM stock solution of all glycans were prepared in ultrapure water and sterilized by irradiating under UV light on ice for 20 min. 3′SL, 6′SL, 3′SLN and 6′SLN were used as positive controls and sucrose and core 1 O-glycan motif Galβ(1,3)GalNAc as negative controls. Concentration of 0, 0.2, 0.6, 1, 2, 5 and 10 mM of each glycan were initially tested followed by a more refined concentration range of 0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 1 mM if required. Virus (150-200 FFU/well, to be determined for each strain) and glycan were mixed in infection media (DMEM + 2 mM L-glutamine + 0.1% BSA) and incubated for 30 min on ice before addition to cells monolayers for 3 hrs at 4 °C with regular rocking. As a positive control, cell monolayers were incubated with the same virus titer minus glycans. After removal of unbound virus by carefully pipetting off the supernatant and washing of the cell monolayer with cold DMEM without additives, the cells were covered with the Avicell-overlay (1.25% Avicell RC-591 + DMEM + 2 mM L-glutamine + 0.1% BSA + 1 μg/ml TPCK-trypsin) and incubated at 35 °C for 20 hrs. After propagation of virus, the Avicell-overlay was carefully removed and the monolayer was fixed with 4% paraformaldehyde, MetOH free (Pierce # 28908) at RT for 15 min. The wells were washed 3 × 5 min with PBS and permeabilised with 0.5% Triton X-100 for 10 min was used for permeabilisation followed by another washing step. H 2 O 2 treatment with 0.3% H 2 O 2 in PBS for 30 min in dark was performed to inactivate endogenous peroxidase and reduce background. Primary antibody anti-Influenza A NP mab (final 1 μg/ml; AbD Serotec Influenza A NP mAb, MCA400) in PBS + 1% BSA + 0.05% Tween 20 was then incubated for 1 hr before washing with PBS + 0.05% Tween 20. Secondary antibody goat-anti-mouse IgG-HRP (final 1 μg/ml; BioRad 172-1011) in PBS + 1% BSA + 0.05% Tween 20 was incubated for 1 hr and washed off 4 × 5 min with PBS + 0.05% Tween 20. IAV infected cells were visualised with TrueBlue reagent (KPL 70-00-64) according manufacturer's instructions and blue foci containing five or more infected cells were manually counted under an Olympus CKX31 inverted microscope. Data analysis and calculation of virus neutralization IC 50 values (non-linear regression (curve fit), dose-response inhibition, four parameter logistic) were carried out using GraphPad Prism 6 (GraphPad Software, Inc., La Jolla, CA). The IC 50 value was considered as the concentration of glycan that reduced the focus forming ability by 50% compared to a non-treated infected cell monolayer. Two individual experiments with duplicates were performed. The data represents two biologically independent experiments in duplicate, and the error bars represent the standard error of the mean.

Virus NMR experiments.
For STD NMR experiments, purified viruses were UV-inactivated for 10 min and buffer exchanged to 20 mM phosphate buffer pH 7.2 and 70 mM NaCl in D 2 O using a 100 kDa Amicon (Millipore) centrifugal device. The glycans of interest were solved in D 2 O. Influenza virus preparations were incubated with 50 μM oseltamivir carboxylate for 10 min at RT to block the neuraminidase activity, followed by the addition of 3 mM of glycan. All STD NMR spectra were acquired in Shigemi tubes (Shigemi, USA) with a Bruker 600 MHz Advance spectrometer at 283 K using 1 H/ 13 C/ 15 N cryoprobe equipped with z-gradients. Virus particles were saturated at −1.0 ppm (on-resonance) and 300 ppm (off-resonance) using a cascade of 60 selective Gaussian-shaped pulses of 50 ms duration. A 100 µs delay between each pulse was applied, resulting in a total saturation time of 3 s. A relaxation delay of 4 s was used. A total of 1512 scans per STD NMR experiment were acquired and a WATERGATE sequence was used to suppress the residual HDO signal. Spin-lock filter with 5 kHz strength and duration of 10 ms was applied to suppress protein background. On-and off-resonance spectra were stored and processed separately, and the final STD NMR spectra were obtained by subtracting the on-and off-resonance spectra. Control STD NMR experiments were performed with an identical setup but with sucrose as a non-binding control glycan as well with heat-inactivated virus (20 min at 70 °C). Before and after the STD NMR experiment, 1 H NMR spectra were recorded as control to ensure the integrity of the samples.
Carbohydrates and reagents. Deuterium oxide was purchased from Sigma Aldrich (Australia).
Sialylα ( 35 were used as receptors for molecular docking. Amber ff14SB 42 and Glycam06 43 force fields were applied to the protein and glycans, respectively. MD simulations of 20 ns length were carried out using GROMACS 44 . The final snapshot of MD trajectories was used for analyses.