Oligomerization-driven avidity correlates with SARS-CoV-2 cellular binding and inhibition

Abstract

Cellular processes are controlled by the thermodynamics of the underlying biomolecular interactions. Frequently, structural investigations use one monomeric binding partner, while ensemble measurements of binding affinities generally yield one affinity representative of a 1:1 interaction, despite the majority of the proteome consisting of oligomeric proteins. For example, viral entry and inhibition in SARS-CoV-2 involve a trimeric spike surface protein, a dimeric angiotensin-converting enzyme 2 (ACE2) cell-surface receptor and dimeric antibodies. Here, we reveal that cooperativity correlates with infectivity and inhibition as opposed to 1:1 binding strength. We show that ACE2 oligomerizes spike more strongly for more infectious variants, while exhibiting weaker 1:1 affinity. Furthermore, we find that antibodies use induced oligomerization both as a primary inhibition mechanism and to enhance the effects of receptor-site blocking. Our results suggest that naive affinity measurements are poor predictors of potency, and introduce an antibody-based inhibition mechanism for oligomeric targets. More generally, they point toward a much broader role of induced oligomerization in controlling biomolecular interactions.

Keywords: SARS-CoV-2; avidity-based neutralization potency; label-free single-molecule tracking; mass photometry; receptor oligomerization.

Fig. 1.

ACE2 induces oligomerization of spike trimers in solution. (A) Schematic of the multivalent interaction partners at the SARS-CoV2 virus–cell interface, containing viral trimeric spike glycoproteins and dimeric ACE2 on the surface of the host cell. (B) The detection principle of solution-based mass photometry, relying on nonspecific binding of soluble proteins to a glass surface. (C) Resulting MP images of individual complexes from a spike–ACE2 mixture. (Scale bar: 1 μm.) (D) Mass histograms of spike–ACE2 mixtures. Spike trimers at 0.55 μM were incubated in the presence of 0.33 to 3.3 μM ACE2 for 10 min at ambient room temperature, and then rapidly diluted to the working concentration of MP just before data acquisition (see SI Appendix, Fig. S1 for measurement of incubated spike at the same conditions without added ACE2). The final concentration of spike trimer was 16.7 nM. Vertical lines indicate the masses of the expected molecular complexes. Histograms were generated by combining 3 to 6 technical replicates. The total number of particles were 12,430, 28,246, and 86,390 for samples containing 10, 50, and 100 nM ACE2, respectively. (E) Mass histograms, for mixing spike trimers with ACE2 (Top) and monomeric ACE2 (mACE2, Bottom), presented on a semilogarithmic scale, showing the increase in the solution concentration of large spike–ACE2 complexes. The probability density was calculated using kernel width of 100 kDa and included all particles larger than 450 kDa to exclude the varying contribution of free ACE2. Insets show representative frames of the recorded MP video of spike with 50 nM ACE2 and 50 nM mACE (Scale bar: 1 μm).

Fig. 2.

Single-particle tracking and mass measurement reveal ACE2-induced oligomerization of spike trimers on supported lipid bilayers. (A) Schematic of the assay with bilayer-tethered spike trimers and ACE2 binding from solution. The supported lipid bilayers also serve as a passivation layer keeping the concentration of ACE2 constant in solution. (B) Representative frame from a median ratiometric video, including trajectories for a few particles. The color gradient corresponds to trajectory propagation time. (C) Representative recorded single-particle mass trace, and the corresponding mass histogram of an individual spike trimer measured at 270 Hz (gray). The black curve corresponds to the intrinsic measurement noise distribution in mass units. The distribution was shifted to be centered at the average particle mass. (D) Cumulative probability distribution of the distance traveled by a single spike trimer during a single frame within its measured trajectory. The corresponding time interval for particle displacement was 3.7 ms. The blue curve corresponds to the best-fitted model used to extract the diffusion coefficient (SI Appendix, section Generating 2D Mass-Diffusion Plots). (E–G) Two-dimensional plots of the measured diffusion coefficient vs. average mass of individual trajectories (scatter points) for (E) wtSpike following equilibration for a few hours on the supported lipid bilayer, (F) wtSpike and (G) omSpike, both following equilibration with 100 nM ACE2. The expected masses for different stoichiometries are indicated by vertical lines. The numbers above the vertical lines indicate the number of bound ACE2 to spike trimer and to cross-linked dimer of trimers. The number of trajectories recorded for each condition ($N_{\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{j}}$) and detected initial surface density of spike (${\rho }_0$) are stated in each panel. Arrows indicate the regions in the two-dimensional plot where the measured diffusion and mass values are expected for trajectories that correspond to complexes of oligomerized spike trimers. (H) Measured diffusion coefficient vs. the inverse of spike cluster size for tethered wtSpike. Circles indicate the diffusion coefficient estimate based on a two-dimensional Gaussian mixture model from eight independent experiments (SI Appendix, Fig. S6), black bars and error bars correspond to the averages and their SD. Error bars for individual circles are not shown for clarity. The red line corresponds to a linear fit with the intercept set to the origin.

Fig. 3.

Thermodynamic analysis of the spike–ACE2 interaction in solution reveals variant-specific cooperativity. (A–C) Normalized mass distributions for mixtures of 25 nM wtSpike with increasing concentrations of (A) mACE2 and (B) ACE2, and 25 nM omSpike with increasing concentrations of ACE2 (C). Histograms represent cumulative counts of 3 to 5 technical repeats per mixture. The black solid curve corresponds to the modeled distribution based on a fitted sum of Gaussian functions. For overlapping peaks in (A), the individual Gaussian functions are shown. Gaussian functions were constrained to the expected masses of the individual complexes and to their expected experimental mass SD. The expected positions of the different spike:ACE2 stoichiometries are indicated at the Top of each panel. (D–F) The resolved occupancy probabilities (scatter points) based on the fitted Gaussian functions for spike with 0 (free spike, $P_0$ and gray symbols), 1 to 3 bound ($P_{1-3}$, labeled as red, blue, and green, respectively) (D) mACE2, (E) ACE2 to wtSpike and (F) ACE2 to omSpike as a function of ACE2 concentration. Individual scatter symbols and error bars correspond to the average values of 3 to 5 technical replicates and their SD, respectively. For wtSpike and omSpike interactions with dimeric ACE2, the scatter points include results from two biological replicates (Material and Methods section). Solid lines correspond to the expected occupancies based on the globally best fitted thermodynamic model. The model takes into account a fundamental standard free energy change for interaction between individual RBD site and ACE2 monomer, $\mathrm{\Delta }{G}^{\circ }$, the degeneracy of the multivalent subunits, and an effective free energy term to account for cooperativity between RBDs on the same spike trimer, $\delta \mathrm{\Delta }{G}^{\circ }$ (SI Appendix, Distribution Analysis of Solution Mass Measurement). Broken lines represent the best-fitted model, assuming no cooperativity in binding. For omSpike (F) we assumed a maximum coverage of 2 ACE2 given no evidence for three bound ACE2 on a single spike trimer. Dashed lines in panels B and C (50 and 100 nM ACE2) show the expected shape of the mass distribution in the absence of cooperativity ($\delta \mathrm{\Delta }{G}^{\circ }=0$), showing that a simple binding process cannot represent the data well.

Fig. 4.

Induced oligomerization by patient-derived antibodies correlates with enhanced inhibition of infection by full IgGs. (A and B) Side and Top view of the spike trimer based on PDB entry 6vsb (8, 42) with the NTD (yellow) and RBD (light blue) domains highlighted. The binding locations of the different antibodies are highlighted based on their identified interactions (5). Gray, dark blue, and red correspond to COVOX384 (down state of the RBD), COVOX150 (up state of the RBD), and COVOX159, respectively. (C–E) Two-dimensional plots of diffusion coefficient vs. mass for equilibrated tethered wtSpike (stabilized by two proline substitutions, 2P) with added 5 nM of (C) COVOX150, (D) COVOX159, and (E) COVOX384. The number of trajectories sampled for each solution condition and the measured initial surface density of spike on the supported lipid bilayer are indicated for each plot. Dashed lines in (C) correspond to the expected masses of spike and spike bound to one antibody. (F) COVOX384 interacting with tethered wtSpike, stabilized by six proline substitutions (HexaPro). (G) Mole fraction of spike trimers in higher oligomeric states as a function of the concentration of added antibody to the solution. Values were extracted by counting the number of trajectories for each oligomeric state relative to the total number of detected trajectories while considering the stoichiometry of each state. Symbols and error bars indicate the average values and SD of two independent measurements (except for the case of COVOX150 at 20 nM that includes one repetition). Colored dashed lines correspond to our fitted two-dimensional thermodynamic model (SI Appendix, Eqs. S1–S18). (H) Representative normalized mass distribution of equilibrated, tethered wtSpike, with 10 nM COVOX159, fitted to a sum of individual Gaussian functions constrained around the expected mass of the different spike:antibody stoichiometries. The black line corresponds to the sum of the Gaussian functions and red dotted lines to the individual Gaussian fits. Dashed vertical lines indicate the masses of the expected wtSpike:antibody complexes, where blue, orange, and green correspond to complexes containing one, two, or three spike trimers, respectively. (I) Mole fractions of individual complexes (gray symbols), calculated from the fitted relative areas of the individual Gaussian functions from H. The x-axis corresponds to the number of bound COVOX159 antibodies to an individual spike trimer (orange), to two trimers (cyan), or to three trimers (pink). Error bars indicate the variation between two independent experiments. Red symbols correspond to the predicted mole fractions based on our globally fitted two-dimensional thermodynamic model. (J) Ratio of binding affinity from solution, $K_{\mathrm{D}}^{(1)}$, to its two-dimensional affinity for cross-linking, ${\mathrm{K}}_D^{(2)}$. Both values were calculated from the fitted interaction free energies (${ϵ}_{1,2}^{\circ }$ in SI Appendix, Eqs. S15 and S17). Average values for the different antibodies are indicated by the height of each column, where the statistical variation of the fitted ratio values (blue symbols) was calculated by repeating the fitting procedure (global fitting for 5,10, and 20 nM of antibody simultaneously) for different sets of concentrations and repetitions (three different concentrations of antibody per set and two independent repetitions for each concentration, except for 20 nM COVOX150). SD are indicated by black error bars.

Fig. 5.

Induced oligomerization enhances binding of multivalent ligands to their surface multivalent receptors. (A and B) Molar fraction of spike trimers bound to at least one ACE2 (A) or COVOX159 antibody (B) for different spike surface densities calculated from experimentally determined affinities. The monomer binding curve (black) corresponds to no induced oligomerization. (C) Required ligand solution concentration resulting in half of spike bound to at least one ligand as a function of spike surface density. (D) Normalized surface density of spike bound to a surface containing different densities of diffusive ACE2 dimers as a function of spike solution concentration. (E–H) Mechanisms of binding and inhibition of SARS-CoV-2 to its host cell-surface. (E) Induced oligomerization of spike and ACE2 during cell-surface binding. (F) Inhibition of SARS-CoV-2 binding by blocking the ACE2 binding site by competitive antibodies. (G) Blocking cell-surface attachment without affecting the ACE2 binding site by spike oligomerization alone. (H) The most potent antibodies combine ACE2 binding site blocking with spike cross-linking in their mechanism of action.