Mucin-Mimetic Glycan Arrays Integrating Machine Learning for Analyzing Receptor Pattern Recognition by Influenza a Viruses

Abstract

Influenza A viruses (IAVs) exploit host glycans in airway epithelial mucosa to gain entry and initiate infection. Rapid detection of changes in IAV specificity towards host glycan classes can provide early indication of virus transmissibility and infection potential. IAVs use hemagglutinins (HA) to bind sialic acids linked to larger glycan structures and a switch in HA specificity from α2,3- to α2,6-linked sialoglycans is considered a prerequisite for viral transmission from birds to humans. While the changes in HA structure associated with the evolution of binding phenotype have been mapped, the influence of glycan receptor presentation on IAV specificity remains obscured. Here, we describe a glycan array platform which uses synthetic mimetics of mucin glycoproteins to model how receptor presentation in the mucinous glycocalyx, including glycan type and valency of the glycoconjugates and their surface density, impact IAV binding. We found that H1N1 virus produced in embryonated chicken eggs, which recognizes both receptor types, exclusively engaged mucin-mimetics carrying α2,3-linked sialic acids in their soluble form. The virus was able to utilize both receptor structures when the probes were immobilized on the array; however, increasing density in the mucin-mimetic brush diminished viral adhesion. Propagation in mammalian cells produced a change in receptor pattern recognition by the virus, without altering its HA affinity, toward improved binding of α2,6-sialylated mucin mimetics and reduced sensitivity to surface crowding of the probes. Application of a support-vector machine (SVM) learning algorithm as part of the glycan array binding analysis efficiently characterized this shift in binding preference and may prove useful to study the evolution of viral responses to a new host.