Molecular dynamics analysis of a flexible loop at the binding interface of the SARS-CoV-2 spike protein receptor-binding domain

Abstract

Since the identification of the SARS-CoV-2 virus as the causative agent of the current COVID-19 pandemic, considerable effort has been spent characterizing the interaction between the Spike protein receptor-binding domain (RBD) and the human angiotensin converting enzyme 2 (ACE2) receptor. This has provided a detailed picture of the end point structure of the RBD-ACE2 binding event, but what remains to be elucidated is the conformation and dynamics of the RBD prior to its interaction with ACE2. In this work, we utilize molecular dynamics simulations to probe the flexibility and conformational ensemble of the unbound state of the receptor-binding domain from SARS-CoV-2 and SARS-CoV. We have found that the unbound RBD has a localized region of dynamic flexibility in Loop 3 and that mutations identified during the COVID-19 pandemic in Loop 3 do not affect this flexibility. We use a loop-modeling protocol to generate and simulate novel conformations of the CoV2-RBD Loop 3 region that sample conformational space beyond the ACE2 bound crystal structure. This has allowed for the identification of interesting substates of the unbound RBD that are lower energy than the ACE2-bound conformation, and that block key residues along the ACE2 binding interface. These novel unbound substates may represent new targets for therapeutic design.

Keywords: SARS-CoV-2; coronavirus; molecular dynamics simulation; protein conformation; protein dynamics; protein modeling; spike glycoprotein.

FIGURE 1

X‐ray crystal structures of CoV1 and CoV2 receptor binding domains (RBD) used in MD simulations. X‐ray crystal structures of the SARS‐CoV RBD in complex with a neutralizing antibody (PDB ID: 2dd8) and the SARS‐CoV2 RBD in complex with the ACE2 receptor (PDB ID: 6m0j). The RBDs from these structures are used as starting structures in this work. The RBD is shown in color and the binding partner is in gray. The loop regions are in blue, and the secondary structure elements are in purple, highlighting the large degree of unstructured regions in the RBD. Enlarged inset of the CoV2 RBD‐ACE2 binding interface shown on the right. Residue sidechains on the RBD (blue) and ACE2 (gray) that participate in the binding interaction are shown in stick configuration

FIGURE 2

Four loops in the SARS‐CoV‐2 RBD form the binding interface with ACE2 and harbor several single amino acid substitutions identified during the COVID‐19 pandemic. (A) The CoV2‐RBD (PDB 6m0j) showing the four different loops that make up the binding interface (green and pink). Residues 438–450 (CoV2) or residues 425–437 (CoV) make up Loop 1, residues 455–470 (CoV2) or residues 442–457 (CoV) make up Loop 2, residues 471–491 (CoV2) or residues 458–477 (CoV) make up Loop 3, and residues 495–508 (CoV2) or residues 481–494 (CoV) make up Loop 4. (B) Prediction of natively disordered regions using the Protein Disorder prediction System (PrDOS) webserver for CoV‐RBD (left) and CoV2‐RBD (right). PrDOS was used without template‐based prediction and thus reports only on the disorder probability of the local amino acid composition. A prediction false positive rate of 5% was used, and values above the 50% threshold (dotted line) indicate regions of predicted disorder. (C) The five most common mutations of the RBD identified during the first 6 months of the COVID‐19 pandemic, with four of these being located in the loop regions of the binding interface. (D) Depictions of the side‐chains of the mutant residues contained in Loop 3 that are studied in the current work

FIGURE 3

Microsecond timescale MD simulations of wild‐type SARS‐CoV and SARS‐CoV2 RBD. (A) Per residue root‐mean‐square fluctuation (RMSF) of all backbone (N, CA, and C) atoms of SARS‐CoV (black) and SARS‐CoV2 (red) RBDs from 4 μs MD trajectories. The sequences were aligned, and the CoV2 residue numbering is used as reference for the x‐axis. The small inset shows the large fluctuation of the Loop 2 and Loop 3 regions near the binding interface with ACE2, where several mutations in the RBD are clustered (dotted lines). (B) Conformational snapshots throughout the 4 μs MD trajectories. (C) Average conformations of the SARS‐CoV (black) and SARS‐CoV2 (red) RBDs, with a focus on the disulfide (yellow)‐containing loop region that shows large fluctuations over the 4 μs simulation. The colors of the data and models are kept consistent throughout the figure

FIGURE 4

Diverse conformational sampling of Loop 3 from different loop models of the RBD. (Left Side) Snapshots from 750 ns of MD simulations of five different loop model structures of the SARS‐CoV RBD are shown on the periphery, with the average structure from each simulation overlaid in the center. The bottom box shows the sampling of Loop 3 conformational space from overlaid average structures. (Right Side) Snapshots from 750 ns of MD simulations of five different loop model structures of the SARS‐CoV‐2 RBD are shown on the periphery, with the average structure from each simulation overlaid in the center. The bottom box shows the sampling of Loop 3 conformational space from the overlaid average structures. The different colors of the models in both cases are used to differentiate between the different starting structures used for each simulation

FIGURE 5

MD simulations of four mutants in the flexible Loop 3 region of the SARS‐CoV2 RBD binding interface. (A) Snapshots of conformations sampled during 2 μs MD simulations of mutants G476S (green), S477N (purple), T478I (red), and V483A (blue). The wild‐type snapshots (black) are the same as shown in Figure 3B, reproduced here for comparison with the mutants. (B) Per residue RMSF of all backbone (N, CA, and C) atoms from the five different models. The inset shows the Loop and Loop 3 regions. RBD mutations are clustered in this region (dashed lines: G476S, S477N, T478I, and V483A). (C) Average conformations of the five different models, showing the high similarity between all of the models. The colors of the data and models are kept consistent throughout the figure

FIGURE 6

Representative conformations from MD simulations that show the greatest similarity or difference from the starting crystal structure of CoV1‐RBD or CoV2‐RBD. (A) The conformations that are most similar (blue, small RMSD) or most different (pink, large RMSD) of the binding interface loops from the wild‐type (left) or loop‐modeled (right) CoV1‐RBD. The crystal structure (PDB: 2dd8) is shown in dark gray. (B) The conformations that are most similar (blue, small RMSD) or most different (pink, large RMSD) of the binding interface loops from the wild‐type (left) or loop‐modeled (right) CoV2‐RBD. The crystal structure (PDB: 6m0j) is shown in dark gray