Evaluating Effects of Host Cell Membrane Inhomogeneity on Influenza Viral Interaction: A Modeling Approach

Despite influenza’s continued presence as a global health threat, no treatment yet exists that provides lasting and comprehensive protection against all viral strains. The constant genetic change of influenza surface proteins prevents development of a universal vaccine, increasing the attractiveness of therapeutics that target host cells rather than the virus. It is of particular interest to identify host cell membrane features that influence viral attachment and downstream infection. In this regard, increased membrane concentration of cholesterol has been shown to increase viral binding. Subsequent molecular dynamics (MD) simulations reveal transient clustering of influenza receptors, and further stabilization of these clusters by cholesterol. These results suggest that membrane receptor and cholesterol composition influence viral attachment by modulating membrane spatial patterning and dynamics.

Here we present a novel computational approach that allows mechanistic study of the molecular behavior of host membrane components, while also capturing the relationship between large-scale membrane organization and multivalent influenza interaction. We reparametrize an existing viral binding and dynamics module for influenza, but add a new membrane module featuring receptors responsive to both thermodynamics and viral attachment. Validation analyses suggest that our Langevin dynamics membrane model displays membrane characteristics reflective of higher-resolution MD simulations. Specifically, we observe a statistically equivalent receptor cluster size distribution and receptor dissociation rate comparable or greater than MD simulations. Our approach thus maintains essential aspects of molecular-level membrane physics while increasing efficiency to allow physiological-scale study of factors influencing influenza attachment.

However, simulations of virus-membrane interaction demonstrate that our model underestimates multivalent viral binding compared to experiment. Although metrics such as mean number of bound receptors and proportion time bound tended to increase with receptor concentration, a statistically significant increase in binding was only observed between the most extreme receptor concentrations. Similarly, our simulations did not produce binding trends responsive to GD1a self-interaction with high fidelity. In conjunction with lack of sensitivity to varied cholesterol composition, these results suggest our model is not sufficiently powered to detect differences in localized receptor spatial patterning and dynamics. However, inconsistent agreement between simulated and experimental observations of influenza binding behavior provide an opportunity for us to reexamine fundamental assumptions such as the relative impact on viral binding of membrane deformation and other forces. The current model provides a well-defined framework to implement these future modifications and additions.