key: cord-0916904-oa9jfots
authors: Taka, E.; Yilmaz, S. Z.; Golcuk, M.; Kilinc, C.; Aktas, U.; Yildiz, A.; Gur, M.
title: Critical Interactions Between the SARS-CoV-2 Spike Glycoprotein and the Human ACE2 Receptor
date: 2020-11-24
journal: bioRxiv
DOI: 10.1101/2020.09.21.305490
sha: 7edf75e0341fe8e0401892b23d132f0bdfd6dc68
doc_id: 916904
cord_uid: oa9jfots

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) enters human cells upon binding of its spike (S) glycoproteins to ACE2 receptors and causes the coronavirus disease 2019 (COVID-19). Therapeutic approaches to prevent SARS-CoV-2 infection are mostly focused on blocking S-ACE2 binding, but critical residues that stabilize this interaction are not well understood. By performing all-atom molecular dynamics (MD) simulations, we identified an extended network of salt bridges, hydrophobic and electrostatic interactions, and hydrogen bonding between the receptor-binding domain (RBD) of the S protein and ACE2. Mutagenesis of these residues on the RBD was not sufficient to destabilize binding but reduced the average work to unbind the S protein from ACE2. In particular, the hydrophobic end of RBD serves as the main anchor site and unbinds last from ACE2 under force. We propose that blocking this site via neutralizing antibody or nanobody could prove an effective strategy to inhibit S-ACE2 interactions.

The COVID-19 pandemic is caused by SARS-CoV-2, which is a positive-sense single-stranded RNA betacoronavirus. Phylogenetic analyses demonstrated that the SARS-CoV-2 genome shares ~79% sequence identity with severe acute respiratory syndrome coronavirus (SARS-CoV), and ~52% with the Middle-East respiratory syndrome coronavirus (MERS-CoV) 1 .

Despite these similarities, SARS-CoV-2 is much more infectious and fatal than SARS-CoV and MERS-CoV together 2 .

SARS-CoV-2 consists of a 30 kb single-stranded RNA genome that is encapsulated by a lipid bilayer and three distinct structural proteins that are embedded within the lipid membrane: envelope (E), membrane (M), and spike (S). Host cell entry is primarily mediated by homotrimeric S glycoproteins located on the viral membrane ( Fig. 1a) 3 . Each S protomer consists of S1 and S2 subunits that mediate binding to the host cell receptor and fusion of the viral envelope, respectively 3, 4 . The receptor-binding domain (RBD) of S1 undergoes a large rigid body motion to bind to ACE2. In the closed state, all RBDs of the S trimer are in the down position, and the binding surface is inaccessible to ACE2. The switching of one of the RBDs into a semi-open intermediate state is sufficient to expose the ACE2 binding surface and stabilize the RBD in its up position (Fig. 1b) 5 .

The S protein binds to the human angiotensin-converting enzyme 2 (ACE2) receptor, a homodimeric integral membrane protein expressed in the epithelial cells of lungs, heart, kidneys, and intestines 6 . Each ACE2 protomer consists of an N-terminal peptidase domain (PD), which interacts with the RBD of the S protein through an extended surface (Fig. 1a , c) 6, 7, 8 . Upon ACE2 binding, proteolytic cleavage of the S protein by the serine protease TMPRSS2 separates the S1 and S2 subunits 9 . The S2 protein exposes fusion peptides that insert into the host membrane and promote fusion with the viral membrane 4 .

To prevent SARS-CoV-2 infection, there is a global effort to design neutralizing antibodies 10 , nanobodies 11 , peptide inhibitors 12 , and small molecules 13 that target the ACE2 binding surface of the S protein. Yet, only a limited number of studies were performed to investigate critical interactions that facilitate S-ACE2 binding using MD simulations. Initial studies have constructed a homology model of SARS-CoV-2 RBD in complex with ACE2, based on the SARS-CoV crystal structure 8, 14 and performed conventional MD (cMD) simulations totaling 10 ns 15, 16 and 100 ns 17, 18 in length to estimate binding free energies 15, 16 and interaction scores 18 .

More recent studies used the crystal structure of SARS-CoV-2 RBD in complex with ACE2 to perform coarse-grained 19 and all-atom 20, 21, 22, 23 MD simulations. The effect of the mutations that disrupt close contact residues between SARS-CoV-2 RBD and ACE2 on binding free energy was investigated by post-processing of the MD trajectories 15, 16, 21, 22 or by using bioinformatic methods 20 . The work required to unbind the S protein from ACE2 would provide a more accurate estimate of the binding strength, but this has not been performed under low pulling velocities using the structure of SARS-CoV-2 RBD in complex with ACE2. In addition, systematic analysis of critical residues that stabilize S-ACE2 binding and how mutagenesis of these interaction sites reduces the binding strength and alters the way the S protein detaches from ACE2 under force have not yet been performed.

In this study, we performed a comprehensive set of all-atom MD simulations totaling 16.5 µs in length using the recently-solved structure of the RBD of the SARS-CoV-2 S protein in complex with the PD of ACE2 7 . Simulations were performed in the absence and presence of external force to investigate the binding characteristics and estimate the binding strength. These simulations showed additional interactions between RBD and PD domains to those observed in the crystal structure 7 . An extensive set of alanine substitutions and charge reversal mutations of the RBD amino acids involved in ACE2 binding were performed to quantify how mutagenesis of these residues weaken binding in the presence and absence of force in simulations. We showed that the hydrophobic end of RBD primarily stabilizes S-ACE2 binding, and targeting this site could potentially serve as an effective strategy to prevent SARS-CoV-2 infection.

Interaction sites between the S protein and ACE2. To model the dynamic interactions of the S-ACE2 binding interface, we used the co-structure of RBD of the SARS-CoV-2 S protein in complex with the PD of human ACE2 7 (Fig. 1c) . The structure was solvated in a water box that contains physiologically-relevant salt (150 mM NaCl) concentration. Two sets of cMD simulations, each of 100 ns in length, were performed to determine the formation of a salt bridge 24 and a hydrogen bond, as well as electrostatic and hydrophobic interactions between RBD and PD (Supplementary Table 1) . A cutoff distance of 6 Å between the basic nitrogens and acidic oxygens was used to score a salt bridge formation 24 . For hydrogen bond formation, a maximum 3.5 Å distance between hydrogen bond donor and acceptor and a 30° angle between the hydrogen atom, the donor heavy atom, and the acceptor heavy atom was used 25 . Interaction pairs that satisfy the distance, but not the angle criteria were analyzed as electrostatic interactions. For hydrophobic interactions, a cutoff distance of 8 Å between the side chain carbon atoms was used 26, 27, 28 . Using these criteria, we identified eleven hydrophobic interactions (Fig. 2a) , eight hydrogen bonds (Fig. 2b) , two salt bridges and six electrostatic interactions (Fig. 2c ) between RBD and PD. Observation frequencies were classified as high and moderate for interactions that occur in 49% and above and between 15-48% of the total trajectory, respectively. F486 and Y489 of RBD formed hydrophobic interactions with F28, L79, M82, and Y83 of PD, while L455, F456, Y473, and A475 of RBD formed hydrophobic interactions with T27 of PD at high frequencies (Fig. 2d) . Salt bridges between K417-D30 (RBD-PD) and E484-K31, and hydrogen bonds between N487-Y83, T500-D355, and Q493-E35 were observed at high frequencies, whereas hydrogen bonds Y449-D38, Q498-K353, T500-Y41, Y505-E37, and Q493-E35 were observed at moderate frequencies (Fig. 2d) .

Residue pairs Y453-H34, N487-Q24, T500-Y41, N501-K353, Q493-K31, and Y449-Q42 exhibited electrostatic interactions throughout the simulations (Fig. 2d) .

The interaction network we identified in our cMD simulations were mostly consistent with reported interactions in the RBD-PD crystal structure 7 . However, our simulations identified four hydrogen bonds (Q498-K353, T500-D355, Y505-E37, and Q498-Q42), one hydrophobic interaction (L455-T27), and two electrostatic interactions (Y453-H34 and N501-K353) that are not present in the crystal structure. In turn, we did not detect frequent hydrogen bonding between G446-Q42, G502-K353, and Y505-R393 and an electrostatic interaction between G496-K353 observed in the crystal structure 7 . This discrepancy may be due to radically different thermodynamic conditions between crystallization solutions and cMD simulations 29 .

We divided the RBD-PD interaction surface into three contact regions (CR1-3, Fig 2A-C) 23 .

The contact region 2 (CR2) comprised significantly fewer interactions than the ends of the RBD binding surface (CR1 and CR3). Remarkably, 10 out of 13 interactions we detected in CR1 were hydrophobic, which were proposed to play a central role in anchoring of RBD to PD 23 .

Unlike CR1, CR2 formed only a single hydrophobic interaction with PD, whereas CR3 did not form any hydrophobic interactions.

Unbinding of the S protein from ACE2 under force. To estimate the binding strength of the S protein to ACE2, we performed steered MD (SMD) simulations to pull RBD away from PD at a constant velocity of 2 Å −1 along the vector pointing away from the binding interface ( Fig. 3a) . Steering forces were applied to the Cα atoms of the RBD residues on the binding interface, whereas Cα atoms of PD residues at the binding interface were kept fixed. These simulations were also repeated in the absence of ACE2 to account for the work done against viscous drag of water ( Supplementary Fig. 1 ). The calculated average work against viscous drag was subtracted from all SMD work values reported in this study. Because part of the work applied is lost to the irreversible processes as we pull RBD away from PD at a finite velocity, the second law of thermodynamics indicates that unbinding free energy difference between the initial and final states cannot be larger than the average work required for unbinding. Therefore, our calculations report relative changes in the binding free energy of wild-type (WT) and mutant RBD under the same velocity and thermodynamic conditions. In 20 SMD simulations (each 15 ns, totaling 300 ns in length, Supplementary Table 1) , the average work applied to unbind SARS-CoV-2 RBD from PD was 64.0 ± 6.0 kcal/mol (mean ± s.d.), demonstrating that the S protein binds stably to ACE2 (Fig. 3b) .

To investigate the contribution of each of the 16 interactions we identified to the overall binding strength, we introduced point mutations on the RBD. Salt bridges were eliminated by charge reversals (K417E and E484K). We also replaced each interaction amino acid (except A475) with alanine (Supplementary Table 1 Table 1) .

F486A, Y489A, N487A, Y473A, E484K and Y505A mutations decreased the work requirement to unbind RBD-PD by 9-15% (Fig. 3e, f and Supplementary Fig. 3 ). We note that most of these mutations also led to the largest increase in residue fluctuations on the binding surface (Fig. 3c) . 5 of these residues (F486, Y489, N487, Y473, and E484) are located in CR1, whereas Y505 is located in CR3. These results highlight the primary role of hydrophobic interactions in CR1 to stabilize the S-ACE2 binding.

To further characterize critical interactions of the S-ACE2 binding interface, we introduced double mutants to neighboring residues of RBD that form critical interactions with PD. We performed a total of 2.8 µs of cMD and 4.2 µs of SMD simulations for 14 double mutants (Supplementary Table 1 ). In particular, double mutants in CR1 resulted in 4 out of the 6 highest increase in RMSF ( Fig. 4a and Supplementary Fig. 2 ). The F486A/N487A mutation at CR1 resulted in the largest increase in fluctuations in both CR1 and CR3 ( Fig. 4a and Supplementary   Fig. 2 ). In SMD simulations, 12 out of 14 double mutations also further decreased the average work to unbind RBD from PD (Fig. 4c, d and Supplementary Fig. 4 ). Similar to the RMSF analysis, double mutants in CR1 (F486A/N487A, E484A/Y489A, E484A/F486A, and L455A/F456A) resulted in 4 out of the 6 largest decreases in average work (Fig. 4d) . A charge reversal of K417E in combination with either Q493A or Y453A also resulted in a large decrease in work values (Fig. 4d) . We also used Jarzynski equality 32, 33 to construct the free energy profiles as a function of a reaction coordinate, referred to as the potential of mean force (PMF) 34 .

Based on the estimated PMF ( Supplementary Fig. 5 ), double mutants in CR1 resulted in the largest decrease in the binding energy by 31-21% compared to WT. Collectively, these results show that two salt bridges (E484-K31 and K417-D30) and the network of hydrophobic interactions in CR1 involving F486, Y489, Y473 and F456 residues are the most significant contributors of binding strength between the S protein and ACE2.

The hydrophobic end of RBD serves as the main anchor site for ACE2 binding. To test whether CR1 anchors RBD to PD 23 , we investigated the order of events that result in detachment of RBD from PD in SMD simulations. The unbinding process appears to perform a zipper-like detachment starting from CR3 and ending at CR1 in 85% of the simulations (Fig.   5a ). In only 15% of the simulations, CR3 released last from PD (Fig. 5a) . Because unbinding simulations can reveal features characteristic for the reverse process of binding 35, 36, 37, 38, 39 , these results suggest that CR1 binding is the first and critical event for the S protein binding to ACE2. Mutagenesis of the critical residues in CR1, in general, resulted in a substantial decrease in the percentages of unbinding events that terminate with the release of CR1 from PD. In alanine replacement of the hydrophobic residues (F486A, F487, and Y489), CR1 was released last for 60%, 55%, and 65% of the SMD simulations, respectively (Fig. 5b) . The probability of CR1 to release last under force was further reduced in double mutants of E484A/F486A (50%) and L455A/F456A (55%) (Fig. 5b) . Unlike these mutants, F456A and E484K mutants in CR1 increased the probability of CR1 to release last, but this could be attributed to a large increase in fluctuations in CR3 upon these mutations (Supplementary Fig. 2b) . These results indicate that single and double mutants of the critical residues in CR1 substantially reduce the binding free energy of this region to ACE2. (Fig. 6a) , six hydrogen bonds (Fig. 6b) , and seven electrostatic interactions (Fig. 6c) Table 1 ). The average total unbinding work of SARS-CoV (64.5 ± 12.2 kcal/mol, mean ± s.d., Fig. 6e ) was identical but more broadly distributed than that of SARS-CoV-2 (64.0 ± 6.0 kcal/mol, Fig. 3b ). Unlike SARS-CoV-2, CR1 released last from PD in only 40% of the unbinding events of RBD of SARS-CoV, whereas the unbinding of CR3 was the last event in the remaining 60% (Fig. 6f) . These results indicate that the S protein binds stably to ACE2 in both SARS-CoV and SARS-CoV-2 and the higher infectivity of SARS-CoV-2 cannot be explained by an increase in binding strength. Higher variability in unbinding work values and the absence of a clear order in unbinding events of RBD of SARS-CoV suggest that SARS-CoV has a more variable binding mechanism to ACE2 than SARS-CoV-2.

We performed an extensive set of in silico analysis to identify critical residues that facilitate binding of the RBD of the SARS-CoV-2 S protein to the human ACE2 receptor. Mutagenesis of these residues and pulling the RBD away from PD at a low velocity enabled us to estimate the free energy of binding and the order of events that result in the unbinding of RBD from PD.

Our simulations showed that the PD interacting surface of RBD can be divided into three contact regions (CR1-3). Hydrophobic residues of CR1 strongly interact with the hydrophobic pocket of PD in both SARS-CoV and SARS-CoV-2. CR1 of SARS-CoV-2 also forms a salt bridge with ACE2 that is not present in SARS-CoV. Based on our SMD simulations, we did not observe a major difference in binding strength of the S protein to ACE2 between SARS-CoV and SARS-CoV-2, indicating that higher infectivity of SARS-CoV-2 is not due to tighter binding of S to the ACE2 receptor. These results are consistent with a recent MD simulation that applied the generalized Born and surface area continuum solvation approach (MM-GBSA) 22 , coarse-grained simulations 19 , and biolayer interferometry 2 .

Our analysis suggests that CR1 is the main anchor site of the SARS-CoV-2 S protein to ACE2, and blocking the CR1 residues F456, Y473, E484, F486, N487, and Y489 could significantly reduce the binding affinity. Consistent with this prediction, llama based nanobodies H11-H4 and H11-D4 neutralize SARS-CoV-2 11 by each interacting with 50% of the critical residues we identified in each of CR1 and CR2. Furthermore, alpaca based nanobody Ty1 was shown to neutralize SARS-CoV-2 40 and among its primary interactions are E484 in CR1, and Q493 and Y449 in CR3, which were also determined as critical residues in our study. Similarly, the human neutralizing antibodies CV30 41 , B38 41 , CB6 41 , and VH3-53 42, 43 are interacting with 50-100% of the critical residues identified in our study.

Experimental studies revealed that antibodies against SARS-CoV induce limited neutralizing activity against SARS-CoV-2 10, 23 . This may be attributed to the low sequence conservation of the CR1 region between SARS-CoV and SARS-CoV-2. In particular, the S protein of SARS-CoV-2 contains critical phenylalanine (F486) and glutamate (E484) residues not present in SARS-CoV, that form hydrophobic interactions and a salt bridge with ACE2, respectively. It remains to be determined whether this difference plays a role in higher infectivity of SARS-CoV-2 than SARS-CoV.

Our simulations show that single and double mutants of CR1 are not sufficient to disrupt the binding of RBD to ACE2, but reduce the binding free energy of this region. Because RBD makes multiple contacts with ACE2 through an extended surface, small molecules or peptides that target a specific region in the RBD-ACE2 interaction surface may not be sufficient to prevent binding of the S protein to ACE2. Instead, blocking of a larger surface of the CR1 region with a neutralizing antibody or nanobody is more likely to introduce steric constraints to prevent the S-ACE2 interactions. This puts a 50 Å water cushion between the RBD-PD complex and its periodic image in the xdirection, creating enough space for unbinding simulations. Ions were added to neutralize the system and salt concentration was set to 150 mM to construct a physiologically relevant environment. The size of each solvated system was ~164,000 atoms. All system preparation steps were performed in VMD 44 .

All MD simulations were performed in NAMD 2.13 45 using the CHARMM36 46 force field with a time step of 2 fs. MD simulations were performed under N, P, T conditions. The temperature was kept at 310 K using Langevin dynamics with a damping coefficient of 1 ps -1 . The pressure was maintained at 1 atm using the Langevin Nosé-Hoover method with an oscillation period of 100 fs and a damping time scale of 50 fs. Periodic boundary conditions were applied. 12 Å cutoff distance was used for van der Waals interactions.

Long-range electrostatic interactions were calculated using the particle-mesh Ewald method.

For each system; first, 10,000 steps of minimization followed by 2 ns of equilibration was performed by keeping the protein fixed. The complete system was minimized for additional 10,000 steps, followed by 4 ns of equilibration by applying constraints on Cα atoms.

Subsequently, these constraints were released and the system was equilibrated for an additional 4 ns before initiating the production runs. The length of the equilibrium steps is expected to account for the structural differences due to the radically different thermodynamic conditions of crystallization solutions and MD simulations 29 . MD simulations were performed in Comet and Stampede2 using ~8 million core-hours in total.

RMSF calculations. RMSF values were calculated as 〈∆ 2 〉 1/2 = 〈( − 〈 〉 ) 2 〉 1/2 , where, 〈 〉 is the mean atomic coordinate of the i th Cα atom and is its instantaneous coordinate. SMD simulations. SMD 47 simulations were used to explore the unbinding process of RBD from ACE2 on time scales accessible to standard simulation lengths. SMD simulations have been applied to explore a wide range of processes, including domain motion 5, 48 , molecule unbinding 49 , and protein unfolding 50 . In SMD simulations, a dummy atom is attached to the center of mass of 'steered' atoms via a virtual spring and pulled at constant velocity along the 'pulling direction', resulting in force F to be applied to the SMD atoms along the pulling vector 45 ,

where is the guiding potential, is the spring constant, is the pulling velocity, is time, and 0 are the coordinates of the center of mass of steered atoms at time t and 0, respectively, and is the direction of pulling 45 . Total work (W) performed for each simulation was evaluated by integrating F over displacement along the pulling direction as = ∫ ( ) 0 .

In SMD simulations of SARS-CoV-2, Cα atoms of ACE2 residues S19-S43, T78-P84, Q325-N330, G352-I358, and P389-R393 were kept fixed, whereas Cα atoms of RBD residues K417-I418, G446-F456, Y473-A475, and N487-Y505 were steered (Fig. 3a) . Steered atoms were selected as the region comprising the interacting residues. For SARS-CoV SMD simulations the same ACE2 residues were kept fixed. However, two slightly different steered atoms Table 1 MD1-33 c-d).

Potential of mean force for unbinding of RBD. Work values to unbind RBD from ACE2 at low pulling velocities along the reaction coordinate were analyzed using Jarzynski equality, which provides a relation between equilibrium free energy differences and the work performed through non-equilibrium processes 32, 33, 34 :

where ΔF is the Helmholtz free energy, kB is the Boltzmann constant and T is the temperature.

Because work values sampled in our SMD simulations differ more than1 kBT ( Supplementary   Figs. 3,4) , the average work calculated in equation (3) 

Data and the analysis software are available from the corresponding author upon request. a In SMD simulations, Cα atoms of PD residues (yellow) were fixed, whereas Cα atoms of RBD (purple) were steered. The orange arrow on the RBD (green) shows the SMD pulling vector, which was taken as the reaction coordinate. 

Identification of a novel coronavirus causing severe pneumonia in human: a descriptive study

Structure, Function, and Antigenicity of the SARS-CoV-2 Spike Glycoprotein

Mechanisms of coronavirus cell entry mediated by the viral spike protein

Stabilized coronavirus spikes are resistant to conformational changes induced by receptor recognition or proteolysis

Conformational transition of SARS-CoV-2 spike glycoprotein between its closed and open states

Structural basis for the recognition of the SARS-CoV-2 by full-length human ACE2

Structure of the SARS-CoV-2 spike receptor-binding domain bound to the ACE2 receptor

Structure of SARS coronavirus spike receptor-binding domain complexed with receptor

SARS-CoV-2 cell entry depends on ACE2 and TMPRSS2 and is blocked by a clinically proven protease inhibitor

Key residues of the receptor binding motif in the spike protein of SARS-CoV-2 that interact with ACE2 and neutralizing antibodies

Neutralizing nanobodies bind SARS-CoV-2 spike RBD and block interaction with ACE2

Computational Design of ACE2-Based Peptide Inhibitors of SARS-CoV-2

Repurposing approved drugs as inhibitors of SARS-CoV-2 S-protein from molecular modeling and virtual screening

Cryo-EM structure of the SARS coronavirus spike glycoprotein in complex with its host cell receptor ACE2

Molecular Mechanism of Evolution and Human Infection with SARS-CoV-2

SARS-CoV-2, an evolutionary perspective of interaction with human ACE2 reveals undiscovered amino acids necessary for complex stability

Computational Prediction of Mutational Effects on the SARS-CoV-2 Binding by Relative Free Energy Calculations

The SARS-CoV-2 exerts a distinctive strategy for interacting with the ACE2 human receptor

Critical Differences Between the Binding Features of the Spike Proteins of SARS-CoV-2 and SARS-CoV

Effect of mutation on structure, function and dynamics of receptor binding domain of human SARS-CoV-2 with host cell receptor ACE2: a molecular dynamics simulations study

Dynamics of the ACE2-SARS-CoV-2/SARS-CoV spike protein interface reveal unique mechanisms

The MERS-CoV receptor DPP4 as a candidate binding target of the SARS-CoV-2 spike

Enhanced receptor binding of SARS-CoV-2 through networks of hydrogenbonding and hydrophobic interactions

Zipping and unzipping of adenylate kinase: atomistic insights into the ensemble of open↔ closed transitions

HBonanza: a computer algorithm for molecular-dynamics-trajectory hydrogen-bond analysis

Direct and quantitative AFM measurements of the concentration and temperature dependence of the hydrophobic force law at nanoscopic contacts

A study of the preferred environment of amino acid residues in globular proteins

Molecular Dynamics Simulation of

Antimicrobial Peptide Arenicin-2: b-Hairpin Stabilization by Noncovalent Interactions

Why protein conformers in molecular dynamics simulations differ from their crystal structures: a thermodynamic insight

Rapid mapping of protein functional epitopes by combinatorial alanine scanning

Comparing experimental and computational alanine scanning techniques for probing a prototypical protein-protein interaction

Equilibrium free-energy differences from nonequilibrium measurements: A master-equation approach

Nonequilibrium equality for free energy differences

Free energy calculation from steered molecular dynamics simulations using Jarzynski's equality

Molecular dynamics simulations suggest that electrostatic funnel directs binding of Tamiflu to influenza N1 neuraminidases

Molecular dynamics study of unbinding of the avidin-biotin complex

Steered molecular dynamics simulations reveal the likelier dissociation pathway of imatinib from its targeting kinases c-Kit and Abl

Unbinding of nicotine from the acetylcholine binding protein: steered molecular dynamics simulations

Computational insights into the mechanism of ligand unbinding and selectivity of estrogen receptors

An alpaca nanobody neutralizes SARS-CoV-2 by blocking receptor interaction

Structural basis for potent neutralization of SARS-CoV-2 and role of antibody affinity maturation

SARS-CoV-2 neutralizing antibody structures inform therapeutic strategies

An alternative binding mode of IGHV3-53 antibodies to the SARS-CoV-2 receptor binding domain

VMD: visual molecular dynamics

Scalable molecular dynamics with NAMD

Optimization of the additive CHARMM all-atom protein force field targeting improved sampling of the backbone ϕ, ψ and side-chain χ1 and χ2 dihedral angles

Steered molecular dynamics and mechanical functions of proteins

Steered molecular dynamics simulation of the Rieske subunit motion in the cytochrome bc1 complex

and electrostatic interactions (green) are shown in the right panels. Lines with colored numbers represent maximum cutoff distances for these interactions. d The frequencies and mean distances of the pairwise interactions of the SARS-CoV S protein and ACE2 binding interface. e Distribution of work applied during unbinding of RBD from PD

Right) Displacement of the critical residues in CR1 (yellow), CR2 (blue), and CR3 (red) along the reaction coordinate

We gratefully acknowledge the support of the COVID-19 HPC Consortium (Grant number: TG-MCB200070), and Extreme Science and Engineering Discovery Environment (XSEDE).

The authors declare no competing interests.