Highly Pathogenic Influenza A(H5Nx) Viruses with Altered H5 Receptor-Binding Specificity

Emergence and intercontinental spread of highly pathogenic avian influenza A(H5Nx) virus clade 2.3.4.4 is unprecedented. H5N8 and H5N2 viruses have caused major economic losses in the poultry industry in Europe and North America, and lethal human infections with H5N6 virus have occurred in Asia. Knowledge of the evolution of receptor-binding specificity of these viruses, which might affect host range, is urgently needed. We report that emergence of these viruses is accompanied by a change in receptor-binding specificity. In contrast to ancestral clade 2.3.4 H5 proteins, novel clade 2.3.4.4 H5 proteins bind to fucosylated sialosides because of substitutions K222Q and S227R, which are unique for highly pathogenic influenza virus H5 proteins. North American clade 2.3.4.4 virus isolates have retained only the K222Q substitution but still bind fucosylated sialosides. Altered receptor-binding specificity of virus clade 2.3.4.4 H5 proteins might have contributed to emergence and spread of H5Nx viruses.

Emergence and intercontinental spread of highly pathogenic avian influenza A(H5Nx) virus clade 2.3.4.4 is unprecedented. H5N8 and H5N2 viruses have caused major economic losses in the poultry industry in Europe and North America, and lethal human infections with H5N6 virus have occurred in Asia. Knowledge of the evolution of receptor-binding specificity of these viruses, which might affect host range, is urgently needed. We report that emergence of these viruses is accompanied by a change in receptor-binding specificity. In contrast to ancestral clade 2. 3 (1)(2)(3)(4) and shown unprecedented intercontinental spread. Whereas highly pathogenic avian influenza A(H5N1) viruses previously detected in poultry in the Americas are believed to have descended from low pathogenicity viruses in this region (5), clade 2.3.4.4 viruses were the first highly pathogenic avian influenza virus of the A/goose/Guangdong/1/1996 lineage to appear in North America (2)(3)(4) during the winter of 2014-2015. Emergence of these novel viruses resulted in culling of >7 million turkeys and 42 million chickens in the United States (6). In Asia, 13 (mostly lethal) cases of human infection with clade 2.3.4.4 H5N6 viruses have been reported (7). Emergence of new viruses with clade 2.3.4.4 hemagglutinin (HA) that infect poultry and humans emphasizes the need for detailed characterization of molecular properties of these viruses.
HA proteins bind to sialoside receptors on the host cell surface. Avian and human influenza A viruses prefer binding to sialic acids linked to a penultimate galactose by an α2-3 or α2-6 linkage, respectively (18). Type and number of internal monosaccharides and their linkages determine fine specificity of virus receptors (19,20). NA removes sialic acids from glycans, which enables virus particles to be released from the cell surface after assembly and from decoy receptors (e.g., in mucus). The balance between activities of HA and NA proteins has a critical role in optimal viral fitness, tropism, and transmission (21).
Changes in HA receptor-binding properties might affect virus host range and within-host virus properties. These changes might have contributed to the remarkable spread of clade 2.3.4.4 viruses. Although a recent study (22) reported enhanced avidity of H5N6 viruses for human-type receptors, recombinant clade 2.3.4.4 highly pathogenic influenza A virus H5 proteins from virus isolates in North America show a strict avian receptorbinding preference (23). We compared receptor-binding properties for clade 2.3.4.4 H5 proteins from an H5N8 virus from Europe with those for an early ancestral clade 2.3.4 H5 protein from an H5N1 virus to identify differences in these properties.

Receptor-Binding Properties of Different H5 Proteins
All H5 proteins efficiently bound fetuin. However, only H5N8 virus HA showed limited binding to transferrin, which indicated that H5N1 2.3.4 and H5N8 HA proteins prefer binding to α2-3-linked sialic acids similarly as H5N1 1 HA (27). The H1 protein bound fetuin to a lower extent than the H5 proteins and bound transferrin, which was consistent with this protein preferentially binding α2-6-linked sialosides (24).
Receptor fine specificity of H5N1 2.3.4 and H5N8 HAs was determined by using glycan array analysis. The H5 proteins bound to a range of mono-and bi-antennary α2-3−linked glycan structures corresponding to N-and O-linked sialosides ( Figure 2). Both proteins did not bind to α2-6−linked sialosides. The specificity of the H5N1 2.3.4 HA was similar to that of its H5N1 1 ancestor (27) (online Technical Appendix Figure, https://wwwnc.cdc.gov/EID/ article/23/2/16-1072-Techapp1.pdf). However, glycan array analysis showed a fucosylation-specific change in receptor fine specificity for H5N8 HA.

Phylogenetic Analysis of Clade 2.3.4.4 H5 Proteins
To identify residues responsible for these differences in receptor binding, we mapped the amino acid substitutions along the trunk of an HA phylogenetic tree (  Table 1). Substitutions K222Q and S227R (H3 numbering used here, corresponding to K218Q and S223R in H5 numbering), which have rarely occurred (0.08% and 0.9%, respectively), were found in the receptor-binding site. In the course of further evolution, a glycosylation site at the head domain at position 160 is lost, and there are 2 other unique substitutions (K193N, T193D) in the vicinity of the receptor-binding site. The Ser residue at position 227 present in H5N1 2.3.4 HA was again found in more recent clade 2.3.4.4 H5-containing viruses that have been detected in North America.

Receptor-Binding Properties of Mutant H5 Proteins
We introduced the amino acid substitutions in H5N8 HA into H5N1 2.3.4 HA and analyzed for their effects on receptor  binding. Substitutions T160A or S227R did not affect fetuin binding ( Figure 6, panel A), but K222Q reduced binding ≈2-fold. Reduction of binding was not observed when substitution K222Q was combined with S227R. Introduction of additional substitutions at positions 160, 193, and 199 in the order in which they occurred during evolution of clade 2.3.4.4 HAs did not affect fetuin binding. We studied the receptor fine specificity of these proteins by using glycan array analysis. Substitution T160A did not change receptor fine specificity (online Technical Appendix Figure). In comparison with wild-type H5N1 2.  The branch of clade 2.3.4.4 H5 proteins that contains viruses from Taiwan and North America contains reverse substitution R227S ( Figure 5). Therefore, we studied how residues at positions 222 and 227 affects receptor binding in the background of H5N8 HA. Substitutions R227S or Q222K hardly affected fetuin binding ( Figure 6, panel  B). In the glycan array, the R227S substitution did not affect binding of H5N8 HA to fucosylated sialosides (Figure 7, panel E), but the Q222K substitution, alone or in combination with the R227S substitution, inhibited binding of fucosylated receptors (Figure 7, panels D, F). We conclude that the identity of residue 222 plays a crucial role in binding of fucosylated sialosides, regardless of the background of highly pathogenic influenza A virus H5 protein. However, although R227 is required for binding

Modeling of Amino Acid Substitutions in H5 Protein Structure
We analyzed the interaction of the H5N1 2.3.4 and the H5N8 HAs with a fucosylated and sialylated tetrasaccharide (3′SLe X ; glycan no. 3) (Figure 3) by modeling on the structure of a clade 1 highly pathogenic influenza A(H5N1) virus H5 protein (H5N1 1 A/Vietnam/1194/2004) that was cocrystalized with 3′SLe X (28). The 3′SLe X ligand and major parts of the receptor-binding site (190-helix, 130-loop, and 220-loop) is shown in Figure 8. Poor binding of A/Vietnam/1194/2004 HA to 3′SLe X was explained (29) by steric hindrance between Lys at position 222 and fucose ( Figure  8, panel A). This steric hindrance is maintained in H5N1 2.3.4 HA but is lost after the K222Q substitution in H5N8 HA (Figure 8, panels B, C).
The effects of aa 227 on binding of fucosylated sialosides might result from the possibility of forming a hydrogen bond between the 220-loop and the amino-terminal end of the 190-helix through the side chains of R227 and N186, thereby influencing the flexibility of the receptor-binding site. At 2 positions that differ between H5N1 2.3.4 and H5N8 HAs (S137A and S185P), we observed changes in the potential to form hydrogen bonds between major elements of the receptor-binding site ( Figure 8). Such changes might affect the interaction of HA with sialic acid-containing glycans and might explain the background-dependent effect of the residue at position 227.

Binding of H5 Proteins to Avian Tissues
We studied binding of H5N1 2.3.4 and H5N8 H5 proteins to avian tissues that differ in the presence of fucosylated sialosides. An antibody to 3′SLe X bound strongly to epithelial cells of chicken trachea, but not to duck intestinal tissue (Figure 9, panel A), which is consistent with previous findings (36). Removal of sialic acids by Vibrio cholerae neuraminidase inhibited binding. Tissue derived from another Anseriformes species (Greylag/Canada goose) also did not display 3′SLe X , and differential staining results were obtained for intestinal tissues of different Galliformes species ( Table 2). The H5 proteins tested efficiently bound chicken trachea (Figure 9,   than for H5N8 HA. We conclude that 3′SLe X −containing sialosides might affect binding of H5N8 HA to avian tissues but is not essential.  (Figure 5). Structural analysis of a clade 1 HA indicates that the close proximity of the conserved K222 side chain and the fucose moiety of 3′SLe X most likely destabilizes their interaction (29). Modeling indicates that such destabilization is still present in H5N1 2.3.4 HAs but is absent in H5N8 HAs that contain Q222 and R227 residues (Figure 8).

Discussion
Introduction of K222Q into H5N1 2.3.4 HA, which removes the competition with the fucose-moiety (29), in itself does not enable binding to fucosylated receptors ( Figure 7). However, additional introduction of an arginine at position 227 (double substitution K222Q/S227R) in H5N1 2.3.4 HA was sufficient to cause a glycan array ). An HA protein tree was constructed by using the PHYLIP neighbor-joining algorithm (https://ugene.net/wiki/display/UUOUM/PHYLIP+Neighbor-Joining) and the F84 distance matrix. This tree was used to construct a guide tree with 52 HA sequences representing all branches of the tree. These sequences were used to construct a summary tree of similar topology as the guide tree. Items above the branches indicate key residues that differ between different branches. Items in red above the branches indicate mutations introduced in this study. The HA protein tree is rooted by an early clade 2. binding profile nearly identical to that of H5N8 HA. Substitution S227R might result in the 220-loop interacting with the conserved loop at the N terminus of the 190-helix through 2 hydrogen bonds between R227 and N186 ( Figure 8, panel C). This interaction, which potentially limits mobility of the 220-loop, might contribute to stabilizing the receptor-binding site in a conformation that enables binding of fucosylated receptors. R227 was required for binding of fucosylated sialosides in the background of H5N1 2.3.4 HA, but not H5N8 HA (Figure 7). Consistent with our findings, HAs of the Taiwanese/North American branch of clade 2.3.4.4 viruses, which obtained reverse substitution R227S (Figure 5), also efficiently bind fucosylated receptors (23).
Analysis of all HA sequences for all avian genotypes available in GenBank showed that the frequency of Q222 and R227, depending on the genotype, is 0 or extremely low. An exception is that Q222 is highly conserved in H7 and H10 genotypes, and R227 is the dominant amino acid in H6 and H13 genotypes. The combination Q222/R227 is present in a 1 clade of low pathogenicity H5N2 viruses represented by A/chicken/Ibaraki/1/05 (H5N2) (38), which was shown to be able to bind to fucosylated receptors, and for which residues 222 and 227 were shown to be essential (39). The HA of this virus appeared to show reduced binding to nonfucosylated glycans, in contrast to the clade 2.3.4.4 proteins analyzed in this study. However, the contribution of individual residues at these positions for receptor binding was not evaluated.
The isolates of this H5N2 clade in Japan originated from H5N2 viruses in Central America (38). Intercontinental transfer of avian influenza viruses is a rare event, drawing a remarkable parallel between 2 viral clades (highly pathogenic H5Nx and low pathogenicity H5N2), both of which have acquired the ability to bind fucosylated receptors and managed to spread intercontinentally. Similar to highly pathogenic H5Nx viruses, low pathogenicity H5N2 viruses might be spread by wild birds, although the possibility could not be ruled out that a virus or vaccine was illegally introduced into Japan from Central America (40). The ability of some low pathogenicity avian influenza A viruses to bind to fucosylated receptors of the 3′SLe X type has been considered a poultry-specific adaptation (20,36,37,39,41) although extensive studies on sufficient numbers of isolates across the complete range of HA genotypes are lacking. Except for low pathogenicity H7 viruses, which all bound efficiently to 6S-3′SLe X irrespective of their host species (42), duck viruses were suggested to bind poorly to 3′SLe X −type receptors (20,37,39,41,43). However, adaptation to binding of 3′SLe x −type receptors has not been reported for high pathogenicity H5N1 viruses isolated from poultry. A recent study showed poor binding of highly pathogenic H5 and H7 viruses to 3′SLe X (44). In another study, binding of a few avian influenza A viruses to fucosylated receptors correlated with their binding to α2-6−linked sialosides (41). However, we found only weak binding of H5N8 virus HA to α2-6−specific trans-ferrin ( Figure 1, panel B) and no measurable binding to α2-6−linked sialosides in the glycan array or by biolayer interferometry (Figure 4). HAs of the Taiwanese/North American branch of clade 2.3.4.4 viruses did not bind α2-6−linked sialosides (23). Therefore, these results do not support the hypothesis that increased binding to  (Figure 4) have shown the evolvement into several branches/subclades often harboring multiple H5Nx virus genotypes. Substitutions K222Q and S227R were present in the earliest H5N5 virus isolates (A/duck/Guangdong/wy24/2008) and have been maintained in all branches except the Taiwanese/ North American branch of clade 2.3.4.4 viruses, which obtained reverse substitution R227S. These findings strongly suggest that the capacity to bind to 3′SLe X −type receptors has arisen in wild birds.
The potential role of altered receptor specificity in extended host range and the contribution to the rapid worldwide spread of influenza viruses is still unknown. The presence of 3′SLe x −type receptors on intestinal epithelial cells varies between different avian species (Table 2) (36) and does not appear to be required for infection of these birds, but their presence in other tissues and species requires further investigation. Apart from determining host-range, altered receptor specificity might also influence other factors involved in spreading, such as virus titers, shedding, and pathogenesis in infected birds.
Outbreaks of clade 2.3.4.4 viruses in poultry might have contributed to enhanced spreading (45)(46)(47), However, adaptations in HA leading to evolvement of poultry-specific clades have not yet been detected in HA-based phylogenetic analyses. Reassortments of the 6 internal gene segments are continuously associated with the further evolution of H5Nx viruses, but the potential contribution of the internal gene constellation to (poultry-specific) spreading remains to be determined.
Of particular interest are the recent outbreaks of influenza caused by H5N6 viruses in poultry and ducks in Southeast Asia, which might resulted in nonavian infections, including 13, mostly lethal, cases in humans (7). Although enhanced avidity of these H5N6 viruses for human-type receptors (carrying α2-6−linked sialosides) has been reported (22), the amino acid combination Q222/R227, which all H5N6 viruses have in their HA, is unlikely to be responsible, and other amino acid substitutions, which have been shown to contribute to binding of α2-6−linked sialosides by highly pathogenic H5N1 viruses, have not been detected in H5N6 viruses (48,49). Two clades of H5N6 viruses have been identified by phylogenetic analysis (50), one harboring an NA with a truncated stem and the other harboring a fulllength stem. Truncation of the stem has been considered a poultry-specific NA adaptation. However, both H5N6 virus clades appear to have acquired their N6 segment in independent events from H6N6 viruses (50), one of which already contained the stem deletion. Also, both clades have caused infections in wild birds, poultry, and humans, but evidence for species-specific adaptions in NA is lacking.
A longstanding paradigm in influenza A virus biology is the requirement for an optimal balance between HA binding and NA cleavage. HA binding displays a clear receptor fine specificity, but substrate fine specificity of NAs has not been extensively investigated. A recent report showed that all NA genotypes (only N4 was not tested) displayed relatively poor digestion of fucosylated receptors (44). Possibly because of tight binding of such receptors by clade 2.3.4.4 viruses, N1 of highly pathogenic H5 viruses might have lost an unknown advantage over other NA genotypes in creating an optimal HA/NA balance, which lead to the remarkable success of novel H5Nx virus reassortants within this clade. †Subtypes identified in different bird species on the basis of data from GenBank and GISAID (http://platform.gisaid.org).