For animal influenza viruses to cause pandemics in human populations, their hemagglutinin (HA) proteins must acquire mutations that allow human-to-human transmission. Fortunately, this barrier has so far protected us from rapid spread of H5N1, an especially pathogenic strain of avian flu. In this issue, Chandrasekaran et al.1 illuminate what would be needed for H5N1 to make the transition to a virulent human pathogen. In particular, they show that HA must acquire mutations that confer viral specificity for host receptors that adopt an umbrella-like topology, as opposed to the cone-like structure of receptors recognized by avian viruses.

The influenza subtypes responsible for pandemics appear to have arisen through reassortment between avian, swine and human strains. The Spanish flu of 1918 caused as many as 40 million deaths, and the 1957 Asian flu and the 1968 Hong Kong flu, although taking a lower toll, also reached pandemic proportions. In recent years, H5N1 has devastated poultry populations throughout Asia, Africa and Europe, and has infected >330 people worldwide, with a fatality rate of 60%2. The possibility that it will acquire the ability to spread rapidly in human populations is clear cause for concern.

Infection by influenza viruses is mediated by binding of the viral surface glycoproteins HA and neuraminidase to their host receptors, with receptor specificity and membrane fusion governed by HA. Whereas avian viruses bind preferentially to glycans terminating in sialic acid connected to galactose through an α2-3 linkage, human-adapted viruses infect cells bearing glycans with α2-6 linkages. Chandrasekaran et al. now revise this paradigm by showing that the presence of an α2-6 linkage alone is insufficient for human-to-human transmission and that the critical considerations are glycan composition and topology.

More than 25 years ago, fairly primitive studies of viral binding to sialylated erythrocytes or erythrocytes whose sialosides had been removed or altered demonstrated that closely related influenza HAs can discriminate between α2-3– and α2-6–linked sialosides3. Since then, X-ray structures of HAs of essentially all major influenza virus subtypes (H1, H2, H3, H5 and H9) in complex with human, avian or swine receptors have revealed much of the structural detail affecting receptor specificity4,5,6,7,8,9. For instance, to accommodate the more roomy α2-6 receptors, receptor binding sites of human-adapted HAs are ∼0.5 Å wider than their avian counterparts. Furthermore, in the most predictable cases, specificity can be predicted by the amino acids present at positions known to contact the receptor.

Advances in carbohydrate synthesis and microarray technologies have led to the development of glycan microarrays, some of which carry more than 200 unique cell-surface glycans10 and enable rapid profiling of the glycan specificity of any protein of interest (Fig. 1a). Glycan microarrays have been used to characterize the receptor specificity of various native and mutant HAs, including those responsible for major flu outbreaks. Similarly, advances in mass spectrometry now permit the determination of the sequence and composition of glycans present on the surfaces of diverse cell types.

Figure 1: Glycan structures required for human adaptation of influenza viruses.
figure 1

(a) Representative examples of glycans present on the glycan microarray and their respective groupings. (b) Topologies adopted by α2-3, short α2-6 and long α2-6 sialylated glycans on receptors relative to the sialic acid (SA) moiety. Within the limits of the SAα(2-3)Gal– and SAα(2-6)Gal–containing glycans printed on the microarray and the influenza hemagglutinins (HA) available for profiling, only receptors modified by long α2-6 sialylated glycans, such as those containing repeats of lactosamine (for example, Galβ(1-4)Glcβ(1-3)]n), are recognized.

Full size image

These technologies are invaluable for assigning receptor specificity. However, the relationship between receptor preference, cell tropism and host-to-host transmission often remains unclear. For instance, some human influenza virus isolates that have HAs which bind α2-6 sialosides do not transmit in human-to-human models. Such observations may be explained in part by recent findings that, in addition to host cells bearing the correct receptor, efficient transmission often requires that virus replication occur in the upper respiratory tract11. The complexity of factors contributing to influenza virus infection and transmission underscores the need to merge outputs from information-rich technologies for optimal interpretation of complementary datasets.

Chandrasekaran et al. develop just such an integrative approach. First, they use lectin costaining and mass spectrometry to demonstrate that α2-6 linkages are widely distributed on epithelial cells in human upper respiratory tract tissues and appear primarily on asparagine- or N-linked glycans. Given that agglutination assays using chicken erythrocytes were widely used to define glycan binding specificity, it is noteworthy that the cell surfaces of the human upper respiratory tract bear long α2-6 sialyl lactosamine structures—in contrast to the short α2-6 receptors present on chicken erythrocytes.

The authors then examine available HA-receptor crystal structures to explore the possibility of conformational distinctions between related receptors bound to HA. In particular, differences in glycan topology related to the fixed position of the sialic acid unit that anchors the glycan to the receptor show a striking correlation with receptor usage. In receptors modified with α2-3 sialylated glycans or short α2-6 sialylated glycans, sugars located in front of the terminal sialic acid residue occupy a space described by a cone. In contrast, reducing sugars in receptors bearing long α2-6 sialylated glycans fold back toward the plane of the sialic acid residue and contact it5, occupying a space resembling an open umbrella (Fig. 1b). Whereas the cone-like topology is associated with recognition by avian and swine HAs, a propensity to adopt the umbrella-like topology is associated with the capacity for viral transmission between humans.

Although this observation that cone-like and umbrella-like topologies define viral specificity may at first seem obvious to those familiar with the nuances of carbohydrate structure and function, the story gets better. Glycan microarrays have established the receptor specificities of HAs derived from modern and pandemic influenza viruses12. The authors mine this microarray data by assigning unique classifiers to distinct and relevant glycan fragments present in HA receptors and then analyze the pattern of HA recognition of each with respect to the parent virus' ability to transmit. The analysis suggests that specific binding of HA to, for example, long α2-6 sialyl lactosamine branches (containing repeats of Galβ1-4GlcNAc, or lactosamine) and not to short α2-6 sialyl lactosamine contributes to human-to-human transmission. Strikingly, fine receptor specificity assigned from the glycan microarray data also correlates with tissue tropism. For example, whereas strain SC18 (specific for long α2-6 receptors) localizes to the apical side of epithelial cells, Mos99 HA (able to bind multiple receptors) is observed throughout upper respiratory tract tissue.

The finding that glycan topology—and not the type of linkage alone—determines the ability of influenza to cross the species barrier should translate to improved technologies for early detection of viruses able to cause a flu pandemic. Expanding the repertoire of N- and O-linked sialosides present on glycan microarrays to incorporate these considerations should be valuable in this regard. Likewise, discovery of sialic acid–specific lectins with greater specificity than those currently available would permit more detailed characterization of cell surface receptors linked to tropism. If the results of Chandrasekaran et al.—many of which are based on recombinant HAs glycan fragments derived from even more complex oligosaccharides—hold true for viral particles, miniaturizing such platforms could enable rapid and sensitive diagnostics for detection of viruses possessing traits associated with human transmission. Ideally, these would be sufficiently robust for field use to keep tabs on flu outbreaks among livestock and other non-domesticated animals. As we hone our detection technologies, viruses will no doubt persist in finding new ways to breach cross-species barriers. Nonetheless, integrative approaches like that of Chandrasekaran et al. will be indispensable for staying ahead of the curve.