Receptor Specificity of HAs from Human H3N2 Viruses 1968 to 2011

Several reports document loss of H3 avidity to human-type receptors during antigenic drift, which has been attributed to the accumulation of N-glycosylation that impacts receptor binding, protein folding, and shielding of immunogenic sites (Gulati et al., 2013; Lin et al., 2012; Medeiros et al., 2001; Nobusawa et al., 2000; Skehel et al., 1984; Vigerust et al., 2007) (Table S2). The potential for glycans added near the receptor binding pocket to impact receptor interactions is illustrated in Figure 2A with a bi-antennary N-glycan modeled at each confirmed glycosylation site on the HA trimer. To compare the impact of antigenic drift of H3 viruses on receptor avidity and on receptor specificity, we produced representative recombinant HAs spanning 1968-2011 in mammalian cells (293F) (Figure S4).

Open in a separate window

Figure 2

Receptor Specificity of Recombinant H3 Hemagglutinins Spanning 1968 - 2011

(A) Visualization of the increase in the number of N-glycans that have accrued during H3N2 antigenic drift. A top view of HA is shown with the HA protein in gray with sialic acid (magenta) α2-6 linked to a galactose (yellow) in the receptor-binding site of each protomer of the trimer. Asialo-LacNAc-bi-antennary-N-glycans (red spheres) are modeled at each known glycosylation site on the HA surface (see also Figures S4, S5, Table S2). Strain-specific glycosylation sites modelled onto the structure of Vic/11 H3 in each case. (B) H3 isolates isolated after 2009 lose avidity to polyacrylamide (PAA) conjugates bearing NeuAcα2-3diLacNAc (3SLN₂ – open circles) or NeuAcα2-6diLacNAc (6SLN₂ – closed circles) glycans. These PAA-glycans were imprinted in a multi-well microarray and overlayed with serially diluted HA (50 to 0.4 μg/ml). After washing, bound HA was detected with mouse anti-HIS and AlexaFluor488 labeled anti-mouse-IgG as described in Experimental Procedures. (C) Recombinant H3 HA proteins exhibit specificity for a subset of human-type receptors. Glycans on the array comprise non-sialoside controls (1-10; Grey), α2-3 sialosides (11-79; Yellow) and α2-6 sialosides (80-135; Green). Glycans are grouped by structure type (see top): L, linear; O, O-linked; N, N-linked and L^x, sialyl Le^x (see also Table S1).

The avidity of the HAs for simple sialosides was assessed using an ELISA-like assay that was adapted to microarray format by immobilization of linear NeuAcα2-3diLacNAc (3SLN₂-L) or NeuAcα2-6diLacNAc (6SLN₂-L) sialosides conjugated to polyacrylamide (PAA) in micro wells of a glass slide (Figure 2B) (McBride et al., 2016; Zhang et al., 2015). All H3 proteins from 1968 to 2003 showed high avidity and specificity to 6SLN₂, with no apparent binding to 3SLN₂. By contrast, later H3 proteins, Perth/09 and Vic/11, lost all binding avidity to 6SLN₂. These results are in agreement with previous reports .

In contrast to the altered binding in the ELISA assay, HAs from 1968-2011 all exhibited remarkably similar receptor specificity on the expanded sialoside array, with exclusive binding to α2-6 sialosides (Figure 2C). On further inspection, the earliest HAs bind a broader range of glycans than the later strains. Binding of HAs to linear fragments of larger glycans (83-86) and linear O-glycans (87-91, 99-103, 109, 110) was most variable, with weakest binding to glycans with only one or two LacNAc (83, 84, 86, 87, 88, 92, 93, 99 & 100). Branched O-glycans with at least three LacNAc repeats were strongly bound by all the HAs (95-98, 106-108), while glycans with one or two repeats were bound more weakly (92, 93, 104, 105). Strikingly, the region of the array displaying N-glycans (116-131) exhibits a comb-like appearance for HAs after 1975. Strongest binding is observed to extended glycans with three to five LacNAc repeats in each core series, including biantennary glycans without (120-122) and with (124,125) core fucose and triantennary chains both without (127,128) or with (130-131) core fucose (see Figure 1B). Conversely, in each series, there is relatively poor binding to glycans with two LacNAc repeats (118, 123, 126, 129), resulting in the gaps, and a comb-like appearance. Although one N-glycan with three LacNAc repeats (119) showed less avid binding, this glycan has a single β−Asn linker that has reduced coupling to the NHS-activated slides compared to the hexapeptide linker of the other N-glycans (118, 120-131). In summary, despite the apparent loss of avidity of H3 after 2003, these HAs maintain binding to extended, branched, N- and O-glycans on the array, providing evidence of conserved receptor specificity during almost 50 years of antigenic drift.

Comparison with Previous Specificity Results of 2006-2007 H3N2 Viruses

Previous glycan array analysis of influenza viruses isolated in 2006 and 2007 reported widely varying receptor specificity for different isolates, and inconsistent preference for ‘human-type’ α2-6 sialosides (Gulati et al., 2013; Stevens et al., 2010). Using aliquots of eight of the same viruses that were isolated from respiratory secretions and produced in MDCK cells, we evaluated receptor specificities on the expanded array (Figure 3). It is apparent that these viruses have a very similar binding pattern compared to one another, and to recombinant H3 HAs (Figure 2). All H3N2 viruses consistently bound to branched α2-6 sialosides with three to five LacNAc repeat unit extensions including, for most viruses, the comb-like pattern resulting from reduced binding to di-LacNAc extended glycans for each N-glycan core (116-131).

Open in a separate window

Figure 3

Receptor Specificity of 2006 Human Influenza A H3N2 Viruses

Intact H3N2 viruses exhibit consistent human-type receptor specificity on the expanded sialoside array. Viruses tested were from frozen aliquots of human influenza isolates grown on MDCK cells and were identical to those previously found to have varied specificity on a glycan array without poly-LacNAc extended N-linked and O-linked glycans (Hatakeyama et al., 2005; Oh et al., 2008). The signals highlighted in red correspond to the glycan structures on the microarray used in the previous study. See also Table S1.

These results are in contrast to glycan array analyses reported previously for the same viruses (Stevens et al., 2010), where it was concluded that there was no consistent receptor specificity and that some viruses bound preferentially or equally to α2-3 sialosides. In light of data presented here, this discrepancy now clearly results from a lack of extended receptors on previously available glycans arrays (Figure 1B). Our expanded array library also features glycans employed in these earlier studies and are shown highlighted in red in Figure 3. Looking at the red highlighted signals, it is apparent that binding to these receptors is weak relative to the robust binding to extended α2-6 sialylated glycans. Moreover, if the highlighted red signals in Figure 3 are normalized to full scale (as normally done in array experiments), and only these signals are considered in assigning receptor specificity, specificity of the various viruses do indeed appear varied as reported previously (Stevens et al., 2010). However, on this expanded array that includes glycans with extended glycan chains, the receptor specificities of the viruses are remarkably similar.

Specificity of H5N1, Avian H3N8, Human H3N2, and Pandemic H1N1 HAs

To place the preference of H3 HAs for extended α2-6 sialosides into broader context, we compared receptor specificities from an H5N1 virus (A/VN/1203/04, isolated from human but with avian-type receptor specificity), the avian H3 pandemic progenitor virus (A/Duck/UKR/63 (H3N8)), a human 2010 H3N2 virus sequenced directly from the clinical isolate without prior growth in laboratory hosts (A/HK/6934/10), and the H1N1 2009 pandemic virus (A/Cal/04/09) (Figure 4). The H5 and avian H3 progenitor HAs recognized only α2-3 sialosides, with preference for branched N- and O-glycans and little selectivity for length, thus confirming the specificity of these avian viruses is maintained across a diverse range of glycans (Figure 4A, B). In contrast, the 2010 H3N2 clinical isolate exhibited the most restricted specificity, binding exclusively to branched N- and O-linked α2-6 sialosides (Figure 4C).

Open in a separate window

Figure 4

Receptor Specificity of HA Proteins from a Human H5N1 infection, an Avian H3N8, a Clinical H3N2 Isolate and 2009 Pandemic H1N1

Glycan microarray analysis of (A) recombinant H5 HA from A/Vietnam/1203/04 H5N1, and (B) an avian H3 precursor to the 1968 human pandemic strain (A/Duck/Ukraine/1/1963 (H3N8)), show typical avian-type specificity for α2-3 sialosides. (C) Recombinant H3 from an A/HK/6934/2010 (H3N2) patient isolate, and (D) H1 Cal/04/09 HA show binding to almost exclusively to N-glycans with three or more LacNAc repeats.

The 2009 H1N1 pandemic virus Cal/04/09 is well documented to exhibit low avidity in receptor binding assays, precluding analysis of receptor specificity on glycan microarrays due to inefficient binding (Stevens et al., 2010). Remarkably, on this expanded array, Cal/04/09 HA exhibited robust and restricted binding to extended branched α2-6 sialosides with preference for N-glycans with three or more LacNAc repeats (120-122, 124, 125, 127, 128, 130, 131), and a single branched O-glycan with five LacNAc repeats (108). Thus, as exemplified by the H3 HAs (Figures 2,​,3)3) and this low-avidity 2009 pandemic virus (Figure 4D), human influenza HAs consistently bind branched N- and O-glycans with three or more LacNAc repeats with higher avidity than linear glycans and branched glycans with shorter chains.

Extended Glycan Receptors Increase Avidity and Support H3N2 Infection of MDCK Cells

To directly assess the biological impact of glycan length, we coupled linear and N-glycans with one to four LacNAc repeats to biotin for ELISA-type assays, and to lipids to assess their ability to serve as receptors for virus infection in MDCK cells (Figure 5A). The avidity of H3 HAs from HK/68 and Vic/11, representative of early and late strains, were assessed by overlaying purified HA onto biotinylated-glycans adsorbed to streptavidin-coated plates. For HK/68 H3, binding to linear glycans (6SLN_1-4-L) was weak and had little dependence on glycan length. Binding to shorter branched N-glycans (6SLN_1-2-N) showed similarly low avidity to linear receptors, while binding to longer N-glycans (6SLN_3-4-N) was dramatically stronger (Figure 5B, left). In contrast, for Vic/11 HA, binding was strongly dependent on glycan length for both linear and branched N-glycans. Indeed, there was no detectable binding to shorter receptors (6SLN_1-2-L and 6SLN_1-2-N) over the avian virus receptor control (3SLN₂-L), and strong binding to longer glycans (6SLN_3-4-L and 6SLN_3-4-N) amounting to > 100-fold higher avidity relative to the non-binding shorter glycans (Figure 5B, right).

Open in a separate window

Figure 5

Glycan Binding Specificity of H3N2 HK/68 and Vic/11 for Host Cell Receptors

(A) α2-6 Sialosides with 1-4 LacNAc extensions used in ELISA analysis of HA binding specificity (biotinylated glycans) or virus infection study (lipidated glycans). L, Linear glycans; N, N-glycans. (B) ELISA analysis of H3N2 HK/68 (left) and Vic/11 (right). HK/68 HA shows weak binding to all linear α2-6-sialosides regardless of length, and binds preferentially to branched N-glycans with tri- and tetra-LacNAc extensions. Apparent K_d for each glycan was calculated using the Prism 6 software package: 6SLN₃-N = 1.83 ± 0.16 μg/ml, 6SLN₄-N = 2.09 ± 0.22 μg/ml. Vic/11 HA shows specificity toward α2-6 receptors with tri- and tetra-LacNAc extensions both in Linear and N-glycan structures. K_d for respective glycans were: 6SLN₃-L = 2.95 ± 0.21 μg/ml, 6SLN₄-L = 0.57 ± 0.04 μg/ml, 6SLN₃-N = 0.46 ± 0.05 μg/ml, 6SLN₄-N = 0.53 ± 0.07 μg/ml. Linear, biotinylated NeuAcα2-3diLacNAc (3SLN₂-L) was used as a negative control (gray diamonds) in all experiments. (C) Infection of glycolipid-engineered MDCK cells (linear & N-linked) with both HK/68 (left) and Vic/11 (right) H3N2 viruses. Plots are average virus titers, measured by plaque assay, at 48 HPI. From left to right in each panel, MDCK monolayers were treated with PBS only (mock infected), 50-80 pfu virus (positive control), CPN then 50-80 pfu virus (background), or treated with CPN, overlaid with SLN-glycolipids, then virus. Linear NeuAcα2-3triLacNAc (3SLN₃-L) glycolipid was used as a negative control (gray diamonds). See also Figure S7.

To assess the length dependence of sialosides as receptors for infection, we reasoned that MDCK cells could be engineered to carry synthetic sialoside receptors coupled to lipid anchors that would spontaneously insert into the cell membrane, as shown for glycolipid receptors of Sendai virus (Markwell and Paulson, 1980; Markwell et al., 1981). To this end, MDCK cells were treated with C. perfringens neuraminidase (CPN) to destroy endogenous receptors, overlaid with lipid-conjugated glycans and incubated briefly with low-titer virus samples. After washing, bound viruses were allowed to replicate for 48 hours, and final titers assessed by plaque assay. For both HK/68 and Vic/11, CPN pre-treatment of cells effectively abolished susceptibility to infection, while cells treated with the lipid-linked avian type receptor (3SLN₃) produced no virus. However, for HK/68, susceptibility to infection was fully restored by lipid-linked human-type receptor glycans (6SLN_1-3-L and 6SLN_1-3-N) regardless of length (1-3 LacNAc repeats) or branching (Figure 5C, left). In contrast, for Vic/11, susceptibility to infection was restored more selectively, with strong preference for longer glycans. For the linear series 6SLN₃-L restored replication equivalent to native MDCK cells, with 6SLN₂-L somewhat less, and virus did not replicate with 6SLN₁-L (Figure 5C, right). For the bi-antennary N-glycans, only the longest human-type receptor (6SLN₃-N) restored replication, recapitulating the binding activity results in ELISA assays and glycan arrays.

Preference for Extended Branched N-linked Receptors is Not Due to the HA Glycan Shield

One possible explanation for the striking preference of the exemplary Vic/11 HA for extended glycan receptors is that they are of sufficient length to allow the sialic acid to reach the receptor binding site without being sterically restricted by N-glycans added around the HA receptor binding site. To directly assess this possibility, we expressed the A/Vic/11 and A/HK/68 HAs in HEK 293S cells that produce only high mannose glycans, allowing them to be removed by treatment with endoglycosidase H (Endo H). As shown in Figure 6A, Endo H removed the glycans efficiently, yielding an apparent molecular weight (63.2 kDa), equivalent to that of Endo H treated HK/68 HA (Figure S7A). Remarkably, glycan arrays showed no change in specificity as a result of deglycosylation (Figures 6B, S7B). In particular, Vic/11 maintains its extreme preference for human-type branched N-glycans with at least three LacNAc repeats, resulting in the familiar ‘comb-like’ appearance for binding to N-glycans 115-131

Open in a separate window

Figure 6

HA deglycosylation does not restore binding to shorter glycan receptors in Vic/11

(A) Victoria/11 HA was expressed in 293S cells (left) and natively deglycosylated using Endo H (center). As a control, a small HA sample was denatured prior to Endo H treatment (right) to ensure complete deglycosylation of the folded protein. (B) Receptor specificities of glycosylated (upper) and Endo H-treated (lower) Vic/11 HA were analyzed by glycan microarray. See also Figure S7.

Glycan Array Printing and Quality Control

Glycan arrays were custom printed on a MicroGridII (Digilab) using a contact microarray robot equipped with StealthSMP4B microarray pins (Telechem) as previously described (Xu et al., 2012). Briefly, samples of each glycan were diluted to 100 μM in 150 mM Na₃PO₄ buffer, pH 8.4. Aliquots of 10 μl were loaded in 384-well plates and imprinted on NHS-activated glass slides (SlideH, Schott/Nexterion), each containing 6 replicates of each glycan. Remaining NHS-ester residues were quenched by immersing slides in 50 mM ethanolamine in 50 mM borate buffer, pH 9.2, for 1 hr. Blocked slides were washed with water, centrifuged dry, and stored at room temperature until use.

Expression and Purification of Recombinant HA

Recombinant H3 subtype HAs from 1968 to 2011 viral isolates, A/HK/1/68 (HK/68), A/Victoria/3/77, A/Beijing/353/89, A/Wyoming/3/03, A/Perth/16/09 and A/Victoria/361/11 (Vic/11) were expressed in 293F freestyle expression system as described previously (Lee et al., 2014). H1 A/Cal/04/09 (Cal/04/09), H3 A/HK/6934/10, H3 A/Duck/UKR/63 and H5 A/VN/1203/04, and for some experiments Vic/11 HAs were expressed in HEK293S (GnTI^−/−) cells and purified from the cell culture supernatants as described previously (de Vries et al., 2013).

Preparation of H3N2 Viruses

The H3N2 viruses used in this study, A/Florida/2/06, A/Georgia/4/06, A/Honduras/3112/06, A/New Hampshire/3/06, A/New Jersey/2/06, A/New York/2/06, A/New York/3/06, A/Pennsylvania/4/07 and Vic/11 were obtained from the Centers for Disease Control and Prevention (CDC) and the CDC-Influenza Reagent Resource (CDC-IRR). They were grown in MDCK cells, purified and frozen in aliquots for glycan array analysis as previously described (Stevens et al., 2010).

Glycan Array Analysis Experiments

Purified, soluble trimeric HA (50 μg/ml) was pre-complexed with either an anti-HIS or anti-Strep mouse antibody and an Alexa647-linked anti-mouse IgG was added (4:2:1 molar ratio) for 15 min on ice in 100 μl PBS-T, and incubated on the array surface in a humidified chamber for 90 min. Array analysis of inactivated virus stocks was performed as described previously (Stevens et al., 2010).

ELISA with Biotinylated Glycans

For glycan ELISA, purified HA trimers were concentrated to 500 μg/ml, precomplexed with anti-HIS mouse antibody and HRP-conjugated anti-mouse IgG, then diluted in series to appropriate assay concentrations (typically 40 – 0.05 μg/ml final). Preparation of streptavidin-coated plates with biotinylated glycans, incubation and washing of pre-complexed HA dilutions was exactly as previously described (Chandrasekaran et al., 2008).

Virus Infection Study

MDCK cells were grown to confluency in 6-well tissue culture trays and digested with 50 units (per well) of commercial CPN (New England BioLabs). Cells were then incubated with 3 μM (final) glycolipid (6SLN_1-3-L/N-lipid) for 20 min, followed by 50-80 pfu virus for 10 min. Bound virus was allowed to grow for 48 hours at 37^oC, with final solution titers assessed by plaque assay. Full virus infection protocols are described in supplementary experimental procedures.

This work was funded in part by grants from National Institutes of Health, AI099274 and AI114730 (to J.C.P.), CA207824 and GM103390 (to R.J.W.) and R56 AI117675 (to I.A.W.), and the Kuang Hua Educational Foundation (to J.C.P.). R.P.d.V. is a recipient of Rubicon and VENI grants from the Netherlands Organization for Scientific Research (NWO). A.J.T. is the recipient of a Long-term Fellowship from the European Molecular Biology Organization (EMBO ALTF 963-2014). We thank Li-Mei Chen and Ruben Donis of the Centers for Disease Control (Atlanta, GA) for samples of 2006 H3N2 viruses, Jennifer Pranskevich for preparation of enzymes used in synthesis of glycans, and Anna Crie for help in preparation of figures and tables. The Consortium for Functional Glycomics (http://www.functionalglycomics.org/) funded by NIGMS grant GM062116 (J.C.P.) provided several glycans used in this study.