To the Editor:

Since it first emerged in North America in mid-February 2009, the novel influenza A (H1N1) virus has spread to most other regions of the world, causing the World Health Organization (Geneva) to declare an emergent pandemic1. H1N1 is a reassortant virus with six of its eight genes, including the hemagglutinin (HA) gene, originating from “North American triple reassortant” swine H1 viruses2. Although these swine viruses have caused sporadic human infection in recent years, onward human-to-human transmission of infection was limited3. In relation to the increased transmissibility of the novel 'swine' virus and its establishment in the human population, the receptor specificity of the HA is a key determinant. The receptors are sialyl glycans, which vary in distribution in tissues of different species and determine host range and tissue tropism, as well as the capacity of animal viruses to initiate a human pandemic4.

In the June issue, Soundararajan et al.5 made a prediction concerning the receptor-binding specificity of a representative pandemic H1N1 2009 virus, A/California/4/2009 (Cal/09). The authors took account of the differences in amino acid residues in the receptor-binding site of Cal/09 from those of previous human H1N1 HAs and constructed homology-based HA-glycan structural complexes with 'human-type' oligosaccharide receptors, namely sialyl oligosaccharides terminating with N-acetylneuraminic acid α2-6-linked to galactose (Neu5Acα2-6Gal), and 'avian-type' terminating with Neu5Acα2-3Gal. From these models, Soundararajan et al. predicted that Cal/09 would be able to make optimal contacts with α2-6-linked sialyl glycans, a feature shared with other human H1N1 HAs and, in addition, make contacts with α2-3-linked sialyl glycans. However, features governing the receptor specificity of HA are complex and it is not always possible at present to draw definitive conclusions merely from sequence analysis and homology modeling studies.

We have compared directly, by carbohydrate microarray analysis, the receptor-binding characteristics of two isolates of the novel pandemic H1N1 virus, Cal/09 and A/Hamburg/5/2009 (Ham/09), with those of a seasonal human H1N1 virus, A/Memphis/14/96-M (Mem/96), as representative of a virus well adapted to humans6. As the HA of the novel H1N1 pandemic virus originated from a virus similar to triple reassortant swine H1N1 viruses, we compared one such example, A/Iowa/1/2006 (Iowa/06), isolated from a human infection3, and an older close relative of classical swine H1N1 viruses, A/New Jersey/76 (NJ/76), the human isolate that initiated the concern of a pandemic threat in 1976 (ref. 7). Information on the viruses and their analyses is provided in Supplementary Methods.

Our analysis system8,9 is based on the neoglycolipid (NGL) technology, which has been validated as an effective approach for presenting oligosaccharides for carbohydrate ligand assignments, not only for soluble carbohydrate-binding proteins but also for whole cells, bacteria10 and virus-like particles11 that express carbohydrate-binding proteins, many of which have low affinities. The special advantages of the lipid-linked probes noncovalently presented on a nitrocellulose matrix are their clustered state and an element of mobility that confer high avidity. Eighty-six sequence-defined oligosaccharide probes were incorporated (Supplementary Methods and Supplementary Table 1). These included eighty sialyl-terminating oligosaccharide probes with differing backbone types, chain lengths and branching patterns, also various sialylation, fucosylation and sulfation patterns, representative of N- and O-glycans and glycolipids. Six neutral probes served as negative controls.

We observed a clear distinction between the receptor-binding repertoire of the pandemic H1N1 viruses Cal/09 and Ham/09 and that of the seasonal virus Mem/96 (Fig. 1). The Cal/09 and Ham/09 viruses bound not only to the majority of α2-6-linked sialyl sequences included irrespective of the backbone chain length and type, but also to a considerable range of α2-3-linked sialyl sequences. In contrast, Mem/96 bound exclusively to α2-6-linked sialyl sequences, and the binding was almost always to those with tetrasaccharide or longer backbones irrespective of the backbone type (type 1, Galβ1-3GlcNAc or type 2, Galβ1-4GlcNAc; Supplementary Table 1 and Supplementary Fig. 1). Even at a high virus concentration, no binding to the α2-3-linked sialyl sequences was observed (Supplementary Fig. 2). Features of the differential binding between the pandemic H1N1 viruses and the seasonal H1N1 virus Mem/96 are highlighted in Table 1 with selected probes that have closely related backbone sequences.

Figure 1: Carbohydrate microarray analyses of the six viruses investigated.
figure 1

Numerical scores for the binding signals are shown as means of duplicate spots at 5 fmol per spot (with error bars). The microarrays consisted of eighty sialylated and six neutral lipid-linked oligosaccharide probes, printed on nitrocellulose-coated glass slides. These are listed in Supplementary Table 1 and arranged according to sialic acid linkage, oligosaccharide backbone chain length and sequence. The various types of terminal sialic acid linkage are indicated by the colored panels as defined at the bottom of the figure.

Table 1 Features of binding to selected sialyl sequences.

Although overall, the strongest binding of the H1N1 pandemic viruses was to α2-6-linked sialyl sequences, binding of Cal/09 to some of the α2-3-linked sialyl sequences was comparable to that of the corresponding α2-6-linked sialyl probes, notably probes 18 and 20, based on disaccharide backbones, and 40, based on a branched hexasaccharide-backbone. Also in this relatively high binding category were sialyl Lewisx-related probes 22 and 29, with di- and tetrasaccharide backbones, and the sulfated sialyl-Lewisx probe 35 with sulfate on N-acetylglucosamine (Supplementary Table 1). Thus, our results using viruses are in accord with the prediction5 of dual specificity based on modeling of the HA protein of the Cal/09 virus. They contrast with those reported by Maines et al.12, who examined the binding of the soluble recombinant HA of Cal/09 to a limited set of biotinylated sialyl (poly)N-acetyllactosamine probes presented on immobilized streptavidin. Maines et al. detected binding to an α2-6-linked sialyl sequence with tetrasaccharide backbone, but little or no binding to α2-3-linked sialyl probes with di-, tetra- or hexasaccharide backbones or to an α2-6-linked probe with a disaccharide backbone. As the affinities of individual HA molecules for their oligosaccharide receptors are low, multivalent interaction with receptors is necessary for high-avidity cooperative binding of virus. Thus, differences in clustering of the probes10 on the one hand, and of the HAs, presented on virus rather than as antibody complexes12, on the other, are likely to account in part for the apparent higher avidity of binding in the present experiments.

Some differences between the two pandemic H1N1 isolates were, however, apparent in our analyses (Fig. 1, Table 1, Supplementary Table 1 and Supplementary Fig. 1). Cal/09 bound more strongly overall to α2-3-linked sialyl sequences and it bound to several short sequences (disaccharide backbones), for example, probes 8, 10, 12 and 20, that are bound weakly by Ham/09. Moreover, Cal/09 bound to α2-8-linked polysialyl sequences as found on brain ganglioside GD3 (probes 70/71) and at the outer arms of N-glycans of the neural cell adhesion molecule (N-CAM; probes 75–84). These differences were less apparent when lower concentrations of virus were used, whereupon binding was predominantly to α2-6-linked sialyl probes and only weak binding to the α2-3-linked probes 27–29 could be discerned (Supplementary Fig. 3). There are three amino acid differences between the HAs of Ham/09 and Cal/09—S83P, A197T and V321I—which may account, at least in part, for these differences in binding. Whereas the residues in Cal/09 are less common, the sequence of Ham/09 is representative of the consensus sequence for the majority of recently circulating H1N1 pandemic viruses.

The patterns of receptor binding of the pandemic H1N1viruses, Cal/09 and Ham/09, were similar overall to those observed for the triple-reassortant H1N1 virus Iowa/06 and the classical swine H1N1 virus NJ/76 (Fig. 1, Table 1, Supplementary Table 1 and Supplementary Fig. 1), except that binding to α2-3-linked 4-O-acetylated sialyl probes 15/16 was observed only with the latter two viruses. This form of sialic acid has been identified in a number of animal species and in trace amounts in humans. The binding pattern of X31, a reassortant virus containing the HA and neuraminidase (NA) of A/Aichi/2/68 (H3N2) from the 1968 pandemic, also showed similarities to those of the two pandemic H1N1 isolates, with preferential binding to α2-6-linked and lower binding to the α2-3-linked sialyl sequences (Fig. 1, Table 1, Supplementary Table 1 and Supplementary Fig. 1). The pattern of binding to the α2-3-linked sequences was more reminiscent of Ham/09, whereas the binding to α2-8-linked sequences resembled Cal/09, but was more apparent even at low concentrations of virus (Supplementary Fig. 4).

These results indicate that no major change in receptor-binding specificity of the HA was required for the emergent pandemic virus to acquire human-like characteristics and become established in the human population. Other factors are likely to have contributed to the sustained human-to-human transmission of the pandemic H1N1 viruses in contrast to the sporadic infections by swine viruses. For example, acquisition of a novel NA by genetic reassortment2 may have provided better complementarity between the functional characteristics of the HA and NA of the emergent virus13.

The broader specificity, namely, the ability to bind to α2-3- in addition to α2-6-linked receptors is also pertinent to the greater virulence of the pandemic virus than seasonal influenza viruses observed in animal models, and its capacity to cause severe and fatal disease in humans, despite the generally mild nature of most infections. Binding to α2-3-linked receptors is thought to be associated with the ability of influenza viruses to infect the lower respiratory tract where there is a greater proportion of α2-3- relative to α2-6-linked sialyl glycans14, although long chain α2-3-linked sialyl (poly-N-acetyllactosamine) sequences are present in ciliated bronchial epithelial cells in humans where they are the receptors for another human pathogen, Mycoplasma pneumoniae15,16. The differences in receptor binding between the 2009 pandemic and seasonal H1N1 viruses may therefore account, at least in part, for the higher virus replication and greater pathology reported in the lungs of ferrets, mice and nonhuman primates infected with pandemic viruses, than observed with contemporary seasonal viruses12,17,18.

Note: Supplementary information is available on the Nature Biotechnology website.