The factors leading to the emergence of a pandemic influenza virus are complex and poorly understood. An understanding of the molecular and biologic requirements for efficient transmissibility is critical for the early identification of a potential pandemic virus and the application of optimal control measures. Despite the fact that an H2N2 virus was responsible for a major pandemic in the 20th century, the absence of this subtype in the human population in the last 40 years has resulted in the H2 subtype receiving little attention. Moreover, the proportion of humans that lack serologic immunity to the H2 subtype is ever expanding. Thus, if a contemporary avian H2 virus were to jump the species barrier and acquire the ability to undergo efficient and sustained transmission among humans, the conditions could be set for an influenza epidemic or even pandemic. Our study is the first to characterize the transmission of H2N2 pandemic viruses in mammals and reestablish the relationship between transmissibility and glycan-receptor binding properties of these viruses.
The efficient transmission of Alb/58 virus was consistent with ferret transmission results obtained for other known human influenza viruses, including the 1918 (SC/18:H1N1) pandemic virus 
. In contrast, ElSalv/57, another H2N2 virus isolated from human host did not transmit efficiently. However, isolation and characterization of the natural variant (ElSalv/57-Q226L) isolated from the nasal wash of the contact ferret revealed a receptor binding site (Q226L) mutation in the HA. The amino acids at positions 226 and 228 have been implicated to play an important role in distinguishing α2,3 and α2,6 glycan receptors 
. Furthermore, when this natural variant obtained from nasal wash of the contact ferret was plaque purified and further tested this virus demonstrated substantially improved transmission via respiratory droplet as compared to the parental ElSalv/57 virus.
In the ferret model, efficient respiratory droplet transmission of avian influenza viruses with a α2,3 SA receptor binding preference has not been observed 
. Furthermore, mutations in HA that enable a switch from the human α2,6 to the avian α2,3 SA receptor binding preference result in a virus incapable of respiratory droplet transmission 
. Characterization of the glycan-binding preference of the H2N2 viruses showed that the avian H2N2 virus, Mallard/78, and the ElSalv/57 virus preferentially bound to α2,3 glycan receptors. The dual binding of Mallard/78 to both types of receptors could be explained by recent results that showed the ability of avian H2 HA to bind to both human and avian receptors ( and 
). The inefficient transmission of these viruses is consistent with the previously established relationship between α2,3 glycan-receptor binding preference and poor transmission of avian-adapted H1, H5, H7 and H9 viruses 
. The Alb/58 virus showed dual binding to both α2,6 and α2,3 glycan receptors with its α2,6-binding affinity substantially higher than its α2,3-binding affinity (as observed in dose-dependent direct glycan array binding assay). This binding pattern is different from the one obtained with the 1918 (SC/18:H1N1) pandemic strain, which binds exclusively to a specific subset of α2,6 sialylated oligosaccharides (including 6′SLN-LN) 
. Nevertheless, the efficient transmission of Alb/58 virus was consistent with ferret transmission results of human influenza viruses, including the 1918 (SC/18:H1N1) pandemic virus 
. While some mixed binders such as the 1918 HA variant A/New York/1/1918 or A/Hong Kong/213/03 (H5N1) virus do not transmit efficiently 
, other mixed binders like A/Texas/36/91 (H1N1) virus do spread efficiently 
thereby deomonstrating that mixed binding does not necessarily disqualify a virus from transmission. Analysis of the glycan-binding specificity of the natural variant of the ElSalv/57 virus that carries the Q226L mutation in the HA showed a dramatic shift in the glycan binding specificity to α2,6 glycan receptors in comparison with its parental virus. These results contrast with the recent results of Xu et al. 
, which observed that H2 HA 226L/228G produce weak binding to both α2,6 and α2,3 linked glycans. In the current study, the increase α2,6 glycan-binding specificity of ElSalv/57 virus correlated with the increased respiratory droplet transmission to naïve ferrets.
Based on the transmission data obtained on multiple avian and human-adapted viruses over the past several years it is improbable that an influenza virus possessing a strong α2,3 SA receptor preference would spread efficiently among humans 
. Nevertheless, the question still remains of whether H2N2 viruses with α2,3 SA receptor binding specificity were circulating early in the pandemic 
. Assessing this possibility includes providing an understanding of whether and how these early-circulating viruses could switch receptor binding specificity. Such studies may provide additional information on the mechanisms of how influenza could adapt to a new host. Notably, Alb/58, a human-adapted and high passage virus, seemingly possesses α2,3 glycan-binding, in addition to α2,6 SA binding; this double binding feature has previously been noted for H2N2 viruses 
. Although a number of H2N2 pandemic viruses possessed α2,3 SA receptor preference, internal genes from these viruses (including the natural variant ElSalv/57-Q226L identified here) possess characteristics of human-adapted viruses. In particular, the H2N2 pandemic virus possessed a human adapted PB2 gene. Adaptation of the PB2 also appears to be critical for efficient aerosolized respiratory transmission of influenza virus 
. Specifically, a single amino acid substitution from glutamic acid to lysine at amino acid position 627 supports efficient influenza virus replication at the lower temperature (33°C) found in the mammalian airway, and contributes to efficient transmission in ferrets and guinea pigs.
In our study, many of the plaque-purified viruses obtained from the three inoculated ferrets nasal washes collected later in the course of infection did not show a homogeneous population at amino acids 226 and 228. Thus, the different residues within receptor binding site of the H2 HA appear to confer a selective advantage by enabling a receptor specificity switch allowing for transmission of a variant virus. Given these results, the receptor binding site of the H2 HA appears to be genetically unstable allowing the virus to switch its binding preference during replication. Finally, because the avian and human-adapted H2N2 viruses that were isolated early in the pandemic share similarities in their glycan-binding amino acids (), we hypothesize based on our results that these amino acids provide a molecular framework that allows the easy switch of their receptor binding preferences. Together with changes in PB2, these HA changes allow for efficient transmission of virus in a human host.