Hierarchical Multivalent Effects Control Influenza Host Specificity

Abstract

Understanding how emerging influenza viruses recognize host cells is critical in evaluating their zoonotic potential, pathogenicity, and transmissibility between humans. The surface of the influenza virus is covered with hemagglutinin (HA) proteins that can form multiple interactions with sialic acid-terminated glycans on the host cell surface. This multivalent binding affects the selectivity of the virus in ways that cannot be predicted from the individual receptor-ligand interactions alone. Here, we show that the intrinsic structural and energetic differences between the interactions of avian- or human-type receptors with influenza HA translate from individual site affinity and orientation through receptor length and density on the surface into virus avidity and specificity. We introduce a method to measure virus avidity using receptor density gradients. We found that influenza viruses attached stably to a surface at receptor densities that correspond to a minimum number of approximately 8 HA-glycan interactions, but more interactions were required if the receptors were short and human-type. Thus, the avidity and specificity of influenza viruses for a host cell depend not on the sialic acid linkage alone but on a combination of linkage and the length and density of receptors on the cell surface. Our findings suggest that threshold receptor densities play a key role in virus tropism, which is a predicting factor for both their virulence and zoonotic potential.

Figure 1

The MAP chip is a platform to quantitate IAV binding to a host cell-mimicking surface. (a) Cartoon showing how glycans immobilized by streptavidin onto supported lipid bilayers (SLBs) were used to mimic the host cell architecture (not to scale). (b) Structure of the biotinylated glycans. 2,3-S(LN)_n are avian-type receptors and 2,6-S(LN)_n human-type. (LN)₂ has no sialic acid and was used to control receptor density on QCM and BLI platforms. (c) Schematic representation of the MAP platform, where viruses only bind above the threshold receptor density for a particular virus–sialoglycan combination. (d) Methodology to create the gradient platform and to study virus binding on receptor gradients. Electrophoretic gradients of biotinylated lipids are formed in SLBs in a microfluidic device at elevated temperature and subsequently cooled down to lock the gradients (i). The biotin gradient is modified with fluorescently labeled streptavidin (SA) and biotinylated sialoglycans (ii). The fluorescently labeled influenza virus is passed over the SLBs with the sialoglycans acting as receptors (iii). (e) Cartoons showing how length and linkage of the sialoglycans can affect virus binding in the case of PR8. The contact area is indicated in peach. If both receptor length and linkage (2,3 or 2,6) are favorable, the virus binds at high and low receptor density (i). If linkage is unfavorable (ii), PR8 may still bind at high and low densities. Other influenza viruses may show a stronger preference for one receptor type and bind only at high densities. If length is unfavorable (iii), the virus binds only at high receptor density. If both are unfavorable, the virus may not bind at all (iv).

Figure 2

Multivalent affinity measurements using different glycan receptors. (a) Fluorescence micrograph of receptor gradients with labeled SA. (b) Fluorescence micrograph of adsorbed labeled virus, binding selectively at high receptor densities. (c) Fluorescence intensity profiles of labeled SA (green) and virus (red) as a function of distance along the horizontal direction. The black lines show the median along the vertical direction. The variance is given by the red and green contours which indicate the percentage of data points between each contour line and the median. (d) Multivalent affinity profiles of influenza PR8 virus as a function of receptor density. The data to which the curves were fitted are shown in Figure S1. (e) Threshold receptor densities for S(LN)_n. Error bars show the 95% confidence interval of the fit. n.s., not significant; *, p < 0.05; ***, p < 0.001.

Figure 3

Theoretical model for multivalent binding of influenza. (a) Cartoon of a virus bound to a glycan-coated surface with model parameters indicated. σ_L is the density of RBDs on the virus. V_ex is the excluded volume of a bound virus, which is a function of its diameter. Ñ is the average number of HA–glycan interactions that can form simultaneously. σ_R is the density of receptors on the surface. A_contact is the receptor surface area that can be reached by RBDs on the virus. (b) zoom-in of a single HA–glycan interaction. K_i,eff is the monovalent receptor–ligand equilibrium of a surface-bound glycan and virus-bound RBD. V_explore is the volume accessible to a glycan. (c) Values of Ñ_threshold for different glycan types and lengths, calculated from the threshold receptor density and maximum length of each glycan. (d) Fitted values of K_i,eff/N_AV_explore, the contribution of individual interactions to the avidity.

Figure 4

Glycan shapes and HA–SA angles that lead to binding in molecular dynamics. (a) The crystal structure of influenza HA in complex with 2,3-S(LN)₁ (gray surface, binding sites in cyan) superimposed onto the biotinylated 2,3-S(LN)₃ bound by SA via the penultimate Gal residue of the glycan (shown as 3D-SNFG). This superimposition was repeated for each of the four biotinylated glycans in each glycan system, and for each snapshot taken from the MD simulations. Any structures with atomic overlaps between the HA and SA were removed from consideration. The angle between the two center lines of the HA and SA was measured (180° for the structure shown here). (b) Histogram plots of the frequency of shapes that did not result in atomic overlap between HA and SA as binned every 10° of the angle between the center lines of SA and HA. The total percentage of glycan shapes without atomic overlap and the percentage of shapes accessible if the rotations of HA and SA are restricted are shown under each glycan structure (n = 20 000 for each system). (c) Range of acceptable angles (ϕ) between the HA axis and the normal to the flat surface as a function of virion radius (R) and glycan length (L).