Human infections by avian influenza viruses, including H7 subtypes, continue to pose a major public health threat. Although the species barrier prevents avian influenza viruses from widespread infection of the human population, the molecular determinants of efficient interspecies transmission and pathogenicity are still poorly understood. The viral coat protein HA however, is perhaps a critical molecule since previous pandemic viruses modified their receptor specificity and overcame the interspecies barrier to spread in the human population. Although HA structures alone and in complex with receptor analogs provide considerable insight into receptor binding, it is clear that HAs from different species and subtypes have significant structural variation. Indeed, low-pathogenic H7N2 avian influenza viruses with an 8 amino acid deletion within its RBS started to circulate in live-bird markets in the northeast United States in 1996. Despite what one would consider a debilitating mutation, these viruses have been reported as the predominant isolate 
. Whether such a deletion contributed to their evolutionary success and how are an important questions, especially in light of NY107's ability to produce respiratory illness in humans 
, as well as its reported increased affinity for human-type receptors and ability for contact transmission in ferrets 
. To try to help answer these questions, we have analyzed the molecular structures of NY107 and its complexes with receptor analogs to explain receptor specificity at the molecular level.
The crystal structures of NY107 and its complexes with both avian and human receptor analogs describe a mechanism as to how an influenza virus might adapt by dramatically altering its RBS, and still be functional. Arg220 of the HA1 chain of NY107 compensates for the loss of the 220-loop, by forming hydrogen bonds with Gal-2 from the avian analog (binding was not observed in either of the structures complexes with the human analogs). However, in the LSTb complex, branched Gal-5 forms extra interactions with the 130-loop, thus improving the binding preference for this particular glycan. Consistent with the structural evidence, glycan microarray analyses of NY107 revealed a strong binding preference for the branched α2-6 sialoside, LSTb. Except for the absence of the 220-loop, other key residues within the RBS are conserved in NY107 and thus, direct interactions with sialic acid are maintained.
The 220-loop is recognized as one of the three crucial structural elements in the RBS. Aside from the North American lineage H7N2 viruses, which have been circulating with a deletion (221–228) in this loop, there has been one other report describing a seven amino acid deletion (224–230) in a laboratory generated H3N2 escape mutant which was reported to have a slightly increased affinity for α2-3-linked glycans by hemagglutination assay 
. Meanwhile, the equivalent region in the hemagglutinin-esterase-fusion (HEF) protein of influenza C virus reveals a rearrangement resulting in a truncated 260-loop in its RBS (Figure S5
. However, without structural data with appropriate receptor analogs, it is not possible to compare the role of these loop variants in receptor binding to the H7 HA structure described here.
When compared to NL219, another co-circulating H7 avian virus HA (), overall binding to α2-3-linked glycans was markedly reduced, while increased binding to α2-6-linked receptors was only marginal. However, these results focus attention on only 2 sub-classes of human-type receptors that may be important for infection (and transmission in ferrets). The NY107 virus interaction with biantennary glycans (), although weak (not seen in with recHA), is a possible route for virus entry as biantennary structures are common on tissues, i.e.
glycan profiling data from human lung tissue on the Consortium for Functional Glycomics (CFG) web site. In addition, the internal sialoside, LSTb, was observed in both virus and recHA microarray data, suggesting this type of glycan has good affinity for this HA. The significance of this is unknown since LSTb has only been described in human milk 
Interestingly, NY107 and NL219 virus receptor binding and specificity has been addressed previously using glycan microarray analysis that reported a significantly increased preference for α2-6 and decreased preference for α2-3-linked sialosides 
. In addition, the same viruses were also included in a recent study from Gambaryan et al.
using a competitive solid-phase binding assay 
. Our findings confirm and extend the receptor binding specificity reported by these authors in that they reported both viruses binding to sulfated sialylglycans with a lactosamine (Galβ1-4GlcNAc core and reported only a moderate binding affinity for α2-6-sialyllactosamine, the human-type receptor analog used in their assay.
The 220-loop is an integral feature of the receptor binding site, and thus one would predict that such a deletion might have compromised this strain to be deleted from the population of circulating viruses. However, this was not the case 
and its existence appears to be in part due to the additional mutations at positions 186 and 205. Restoration of the loop with either or both residues mutated back to the pre-1994 consensus sequence resulted in a classic avian influenza virus binding profile. The emergence of the Glu186Gly mutation in the HA of the NY107-ins-205 mutant after only two passages of the rescued virus in eggs, also indicates the importance of these positions for HA functionality/glycan specificity. Analysis of the structural data reveals that positions 186 and 205 are on opposite sides of a monomer but are both close to the 220-loop deletion region in the trimeric form. The Glu at position 186 is close to Arg220 and may interact with Arg220 when binding avian receptors. Position 205 in the neighboring monomer may be important in trimer stability and maintaining RBS functionality. If one models the pre-1996 220-loop restored into the NY107 structure, Arg205, Glu186 and the loop all clash, thus explaining the Glu186Gly mutation that emerged in the NY107-ins-205 virus HA after limited egg passage.
The NY107 RBS with its more restricted α2-3 glycan binding preference and weak/moderate increase in α2-6 binding may have given the virus a selective advantage to be maintained in poultry at live bird markets and supplying farms. Certain terrestrial birds, such as quails and chickens, have recently been shown to present both human and avian types of receptors in the trachea and intestine 
. Although it is not known what specific glycans are presented in these animals, it is conceivable that a virus with mixed specificity might have a distinct advantage over avian viruses that have specific avian receptor requirements, particularly in bird markets where multiple species coalesce. Previous results with H7N2, H9N2 and H5N1 viruses all highlight the fact that an increase in α2-6-binding preference is not sufficient for efficient transmission of avian influenza viruses to humans 
. Although it remains to be seen whether prolonged circulation of viruses in terrestrial birds, such as domestic chickens, can provide a possible route for viruses to adapt for efficient human infection 
, continued surveillance of influenza viruses from avian and other animal reservoirs is urgently needed to define their zoonotic potential.