Glycan remodeled erythrocytes facilitate antigenic characterization of recent A/H3N2 influenza viruses

During circulation in humans and natural selection to escape antibody recognition for decades, A/H3N2 influenza viruses emerged with altered receptor specificities. These viruses lost the ability to agglutinate erythrocytes critical for antigenic characterization and give low yields and acquire adaptive mutations when cultured in eggs and cells, contributing to recent vaccine challenges. Examination of receptor specificities of A/H3N2 viruses reveals that recent viruses compensated for decreased binding of the prototypic human receptor by recognizing α2,6-sialosides on extended LacNAc moieties. Erythrocyte glycomics shows an absence of extended glycans providing a rationale for lack of agglutination by recent A/H3N2 viruses. A glycan remodeling approach installing functional receptors on erythrocytes, allows antigenic characterization of recent A/H3N2 viruses confirming the cocirculation of antigenically different viruses in humans. Computational analysis of HAs in complex with sialosides having extended LacNAc moieties reveals that mutations distal to the RBD reoriented the Y159 side chain resulting in an extended receptor binding site.


Synthesis of N-glycans
Sialylglycopeptide (SGP) was extracted from egg yolk and further enzymatically modified to yield compound 1, which was used as a starting material for the synthesis 3,4 . The terminal galactose of compound 1 was sialylated with an α2,6 specific sialyltransferase mutant P34H/M144L from Pasteurella multocida and CMP-Neu5Ac providing compound 2 2 . This compound was subsequently extended with N-acetyllactosamine (LacNAc) repeats by using mammalian β1,4-galactosyltransferase 1 (B4GalT1) and β1,3-N-acetylglucosamine transferase (B3GnT2) with their corresponding nucleotide sugars UDP-Gal and UDP-GlcNAc, respectively. As a result, compound 3, 4 and 6 with one, two and three consecutive LacNAc repeats on the MGAT1 (Mannose-3) branch and a terminal α2,6 linked N-acetylneuraminic acid (Neu5Ac) on the MGAT2 (Mannose-6) branch were obtained. These intermediates were used to synthesize the bisialylated compounds 5 and 7 by sialylating the extended MGAT1 (Mannose-3) branch with the α2,6 specific sialyltransferase mutant from Pasteurella multocida and CMP-Neu5Ac. Compounds 9 and 11, modified with a single terminal Neu5Ac on either branch, were prepared by first quantitatively desialylating the intermediates 4 and 6 in an aqueous solution of acetic acid. Afterwards, sialyltransferase ST6Gal1 and CMP-Neu5Ac were used to install a single terminal Neu5Ac moiety providing the products 9 and 11.
The acceptor and UDP-GlcNAc (1.5 eq) were dissolved in a HEPES buffer (50 mM, pH 9.6, 0.1 wt% BSA) containing DTT (1 mM) and MnCl2 (20 mM) to obtain a concentration of 5 mM. B3GnT2 (30 µg per µmol acceptor) and CIAP (1 u µL -1 , 1 u per µmol of added nucleotide) were added to the reaction mixture and it was incubated overnight at 37 °C with gentle shaking. The progress of the reaction was monitored by LCMS. In case of incomplete conversion after 18 h, additional UDP-GlcNAc (0.5 eq), CIAP (1 u µL -1 , 1 u per µmol of added nucleotide) and B3GnT2 (15 µg per µmol acceptor) were added and the reaction mixture incubated at 37 °C for an additional 24 h. After completion the reaction mixture was lyophilized and applied to size exclusion S4 chromatography. Carbohydrate-containing fractions were lyophilized and used without further purification.
General procedure for the removal terminal Neu5Ac: The substrate was dissolved in an aqueous solution of acetic acid (2 M) and kept at 65 °C for 24 h. The solvent was removed in an N2 flow and the reaction mixture was applied to size exclusion chromatography. Carbohydrate-containing fractions were lyophilized and used without further purification.
General procedure for the installation of α2,6-linked Neu5Ac using ST6Gal1: The acceptor and CMP-Neu5Ac (1.1 eq) were dissolved in a Tris buffer (50 mM, pH 7.3, 0.1 wt% BSA) to obtain a concentration of 2 mM. ST6Gal1 (42 µg per µmol acceptor) was added to the reaction mixture and it was incubated overnight at 37 °C with gentle shaking. The progress of the reaction was monitored by LCMS. In case of incomplete conversion after 18 h, additional ST6Gal1 (20 µg per µmol acceptor) was added and the reaction mixture incubated at 37 °C for 24 h. After completion the reaction mixture was lyophilized and applied to size exclusion chromatography. Carbohydrate-containing fractions were purified by HPLC (9: 68%B-64%B in 80 min, 3.4 mL min -1 ; 11: 67%B-62%B in 80 min, 3.4 mL min -1 ) providing the product as a white powder (9: 47 µg, 17%, 11: 50 µg, 10%).

Analytical data
VnmrJ 4 and TopSpin 4 were used to collect NMR data. NMR data was obtained at room temperature on a 600 MHz instrument from Bruker. The chemical shift δ is given in parts per million (ppm) and refers to tetramethylsilane and the residual solvent peak [ 1 H-NMR: δ(D2O) = 4.79 ppm]. NMR data is given as follows: 1 H-NMR: chemical shift (multiplicity, coupling constants, relative integral, functional group); 13 C data are extracted from HSQC spectra and given as follows: chemical shift. Multiplicity is defined as follows: s = singlet; d = doublet; t = triplet; m = multiplet. Signals were assigned by numbering the monosacharide units starting at the reducing end of the oligosaccharide. Monosacharides attached to the mannose-6 branch are indicated by a " ' " (prime) and those attached to the mannose-3 branch without any mark. The assignment was S5 done by using corresponding 2D-NMR spectra (COSY, HSQC). Due to the use of ammonium formate containing buffers during final purification, several spectra show residual formic acid (8.46 ppm, not shown in the NMR spectra) contamination. The yield/concentration of the final products was determined by NMR spectroscopy, using n-propanol as an internal standard. High resolution masses were measured on an Agilent 6560 Ion Mobility Q-TOF LC-MS system.

Structural studies
It has been found that during the transition from late 90s to early-2000, HA of A/H3N2 viruses have a reduced affinity for the prototypic human receptor, the 6'-SLN 5 . Analysis of protein sequence alignment shows that the main sequences differences reside at the 130-loop (residues 131, 135 and 137), the 150-loop (residues 155-159), the 220-loop (residues 222, 225 and 226), the E190D mutation at the 190-helix, and the appearance of a new glycosylation site at residue 144 and 158. Other residues which contribute to sialic acid binding are highly conserved among HAs including Y98, H183, Y195 and W153 (Table S3). Although X-ray structural studies 5-8 have provided an understanding of the structural basis of changes in receptor binding specificity, it has not uncovered interactions with extended glycan receptors. Such structural data is difficult to obtain by X-ray crystallography, and therefore, we

Modeling of A/NL/816/91 (NL91)
. NL91 recognized most of the human-type receptors, including compounds that have an a2,6-sialoside on a mono-LacNAc residue (glycans 7-9, Figure 1A). A somewhat higher responsiveness was observed for receptors having an a2,6-sialoside onto di-LacNAc residue (glycans 10-13, Figure 1A), while tri-LacNAc containing receptors did not further improve HA binding (glycans 14-17). Collectively, the results showed that the optimal receptor for NL91 is an N-glycan having two repeating LacNAc moieties modified by a 2,6-linked sialoside (glycan 10 and 13). Thus, we performed all atoms MD simulations of the complex between the NL91 and the receptor (LacNAc)2a2-6Neu5Ac. The MD simulation showed that the glycan receptor binds the HA protein almost exclusively through the terminal sialic acid. In fact, the analysis of the MD derived trajectory revealed a stable binding pose for the a2,6-sialoside, while the underling di-LacNAc chain explores multiple orientation along the simulation, where only transient intermolecular interactions exist. Comparing X-ray data and the molecular modeling derived structures indicates that the interaction network is preserved in HK68 and NL91. This is consistent with the high structural homology of the sialic binding site of HK68 and NL91 for which only two mutations exist, G135E and N137Y. Actually, these two residues contribute to receptor binding through their backbone atoms ( Figure 3A). Specifically, all the hydrophobic interactions, such as those with Y98, H183, Y195 and W153, and H-bond interactions with residues 135-137 and S228 are preserved. In line with the X-ray studies 7 , we found that in NL91 the Glu190 engages the Sia-1 O9 through H-bond interaction with an average distance of ~2.9 Å, while L226 makes hydrophobic interactions with the C-6 of the galactose-2 ( Figure 3A). For the underling di-LacNAc chain, the MD simulation showed that only the Gal-2 contributes to binding by either engaging E190 or G225 in an H-bond interaction, however, these interactions populate only 10% of the whole MD simulation, which results in a barley defined binding pose for the di-LacNAc chain. The results demonstrate that the HA protein of NL91 recognizes the human receptor through the terminal a2,6-sialoside while the underling glycan does not significantly contribute to binding.

Modeling of A/NL/109/03 (NL03).
NL03 recognized far fewer glycans and did not bind to structures having their a2,6-sialosides at a mono-LacNAc moiety (glycans 7-9, 12 and 16 Figure  1A). It recognized structures having the sialoside on at least one di-LacNAc moiety (glycans 10 and 13), although it shows a significant improvement in responsiveness when interrogated against S11 structures which contain tri-LacNAc residue (glycans 14, 15 and 17). Thus, we performed all atoms MD simulations of the complex between the NL03 and the receptor (LacNAc)3a2-6NeuAc. The MD simulation showed that the extended LacNAc chain contributes to HA binding. The analysis of the MD derived trajectory revealed a stable binding pose for both the a2,6-sialoside and the underlying tri-LacNAc chain. In agreement with previously reported X-ray studies 7 , the analysis of the MD simulations showed that in NL03, the D190 does not participate in sialic acid binding. Instead, the D225 engages the Gal-2 O3 through a H-bond interaction which results in a change in a dihedral angle of the sialic acid-galactose glycosidic bond. The rotation of the Gal-2 residue places all subsequent moieties toward the 190-helix 8 . The MD simulations showed that the Asp190 engages the Gal-4 O2 through H-bond interaction, while the Asn193 provides a Hbond interaction with the acetamide moiety of GlcNAc-3 with an average distance of ~2.5 Å ( Figure 3E). Inspection of X-ray crystal structures 7,8 of post 2003 HAs shows that distal mutations to the RBD (A131T, H155T and E156H) reoriented the side chain of Y159, which resulted in an extended receptor binding site (Supplementary Figure 9). The molecular modeling study showed that Gal-6 makes a CH-p interaction with the aromatic ring of Y159 ( Figure 3E). The contribution of Y159 in further stabilizing receptor binding was analyzed by monitoring the distance of Gal-6 H3, H4 and H5 against the aromatic ring along the MD simulation (Supplementary Figure 10), which showed a stable interaction. These additional interactions support the hypothesis that post-2000 strains have undergone mutations that compensated for the reduced affinity for the terminal sialoside. Interestingly, the extensive epistatic network which correlates mutations at the RBS with those occurring at distal antigenic site, such as the 150-loop and 190-helix, has been demonstrated by large-scale mutagenesis experiments 7 .
Modeling of A/NL/1797/17 (NL17). NL17 and NL19 (3C.2a) showed strong responsiveness only to glycans 14, 15 and 17 ( Figure 1A). These glycans have in common that at least one of the arms is extended by three consecutive LacNAc units that is further modified by an a2,6-sialoside. Thus, a glycan having four LacNAc units arranged in an asymmetrical manner (15) represents the minimal receptor for these viruses. Mono-sialylated derivative 15 gave a similar responsiveness compared to the bis-sialosides 14 and 17 indicating that a bidentate binding event does not substantially contribute to recognition as previously suggested 9 . We docked the structure of the NL17 HA protein in complex with the receptor (LacNAc)3a2-6NeuAc based on the results from MD simulation of NL03 in which we replaced the mutated residues by using the mutagenesis tool implemented in PyMOL. The resulting structure was minimized by using the steepest descent algorithm implemented in the Amber MD program. The orientation of the glycan receptor is very similar in NL03 and NL17. Specifically, NL17 does not present additional amino acids mutation at the sialic acid binding site, leading to the same binding interactions as for NL03 at this site. The Gal-2 is also bound in a similar manner, thus orienting the GlcNAc-3 against the 190-helix. However, in NL17 residue 193 is substituted by Phe for which H-bond interaction is not possible, suggesting that F193 does not contribute to receptor binding ( Figure 3G, H and I). The missing interaction with the GlcNAc-3 may be the reason for the lack of binding of di-LacNAc containing glycans observed for NL17 with respect to the reduced binding of NL03 ( Figure 1A). Instead, the Y159 is preserved and allows for CH-p interaction with the Gal-6 of the internal LacNAc moiety. The importance of Y159 in extended N-glycans binding is confirmed by the receptor specificity of 3C.3a clade which present the Y159S and F193S mutations. It is expected that these viruses lost the ability to engage the Gal-6 through CH-p interactions while compensate by H-bond interaction between the GlcNAc-3 and the S193, which reflect the re-gained ability of binding sialilated di-LacNAc glycans (glycans 10 and 13). The molecular modeling study herein presented support a notion that A/H3N2 viruses have undergone mutations to create and extended binding site. S13 NMR spectra 1 H 2; 600MHz; D2O.

. Correlation of hemagglutination inhibition and focus reduction titers by recent A/H3N2 viruses.
Shown is the correlation between every HI titer and FRA titer for all viruses and sera as depicted in Table 1 and Supplementary Table 1. The titers of the focus reduction assay duplicates were averaged for the graph. Spearman's rank correlation coefficient was measured using Prism 8.3.1 (Graphpad).    Table 2. Sequence analysis. Alignment of A/H3N2 HA sequence spanning from 1968 to 2019. Amino acid mutation refers to the A/HK/1/68 sequence. Amino acids substitutions are labelled in color code according to the changes in polarity: mutations that led to amino acids (aa) of similar polarity are labelled in gray, mutations that led to less polar aa are in pink, mutations to more polar uncharged residue are labeled in light blue, mutations to amino acids with negative charged side chain are labeled in red, mutations to amino acids with positive charged side chain are labeled in dark blue. New occurring glycosylation sites are labeled in green.