Structural Insight into Antibody Evasion of SARS-Cov-2 Omicron Variant

Data Retrieval and Structure Preparation

We collected the sequence information for surface glycoprotein of SARS-CoV-2 WT and the B.1.1.529 variant (Omicron) from the NCBI database with the Gen Bank IDs YP_009724390.1 and UFO69279.1, respectively [25, 26]. A pairwise alignment was performed for the two spike protein sequences using the EMBOSS Needle program [27]. Further, the structures of various antibodies bound to the spike protein were retrieved from the RCSB Protein Data Bank (http://www.rcsb.org/pdb/), and their epitope information was collected from the IEDB database (http:// www.iedb.org/) [28, 29]. For our study, we selected CB6 (PDB ID: 7C01), REGN10933 (PDB ID: 6XDG), S309 (PDB ID: 7R6X) and S2X259 (PDB ID: 7M7W) antibody complexes. The B.1.1.529 variant mutations were introduced in the selected RBD-antibody crystal structures with the maestro interface of the Schrodinger suite (v2019.1) [30]. These raw structures were further processed in the protein preparation wizard for the addition of missing amino acid side chains, addition of missing hydrogen atoms, and optimization of the hydrogen bond network. The structures were finally energy minimized with the OPLS3e force field and saved for further calculations.

Molecular Dynamics Simulation

We carried out the MD simulation of the RBD- antibody structures and their respective mutant (including B.1.1.529 mutations) complexes. Hence, a total of 8 individual simulation systems were prepared with identical parameters. The simulation was performed with the GROMACS package v2019.4 [31]. The systems were parameterized with the AMBER99SB force field in a cubic box filled with TIP3P water molecules. Neutralization was achieved by adding the required number of counter ions (Na+/Cl−), and energy minimization was carried out using the steepest descent minimization algorithm. Before the final MD run, the systems were equilibrated in two consecutive steps, NVT (particles, volume, and temperature kept constant) and NPT (particles, pressure, and temperature kept constant) for 100 picoseconds (ps) each. Finally, each system was simulated for 100 nanoseconds (ns), and energies were saved every 10 ps.

Interaction Analysis and Binding Free Energy Calculation

The MD simulation trajectory of each RBD-antibody complex was analyzed, and coordinates of the lowest energy conformation of the complex were extracted for further interaction analysis and energy calculations. We used the web application COCOMAPS (biocomplexes Contact MAPS) for evaluating the interface and interaction statistics of the predicted conformations of the RBD-antibody complexes [32, 33]. Further, we calculated the binding free energies of all the complexes with the Molecular Mechanics/Generalized Born Surface Area (MMGBSA) approach using the HawkDock server [34, 35].

Class 1: CB6 (Etesevimab)

The antibody CB6, also known as etesevimab, is an approved therapeutic antibody against SARS-CoV-2. The structure of CB6 in complex with RBD (PDB ID: 7C01) and its omicron mutation-induced complex was simulated for

100 ns. MD simulation revealed the overall stability of the complexes. The root mean square deviation (RMSD) graph of the WT complex was stable; however, the backbone of the omicron mutant complex experienced fluctuations throughout the trajectory (Figures 3g).

a b

Click to enlarge

Figure 3: Molecular dynamics simulation results and interface statistics of RBD-CB6. (a) Wild type RBD and CB6 interface interactions. Residues involved in interactions are highlighted. Interface residues forming hydrogen bonds are labeled and shown as sticks. Hydrogen bonds are sown as black dashed lines. (b) Omicron induced mutant RBD and CB6 interface interactions. Mutant interface residues are shown in cyan and labeled in blue color. (c) Polar and non-polar interface area of the complexes. The values are in Angstrom (Å) (d) Solvent accessible surface area (SASA) for the CB6-RBD/WT and CB6-RBD/ omicron complex. (e) Plot for number of hydrogen bonds for the complexes. (f) Interface interaction count (HPhil; Hydrophilic, HPho; Hydrophobic). (g) Root mean square deviation (RMSD) plot for the complexes. (h) Radius of gyration (Rg) plot.

The radius of gyration (Rg) for the CB6-RBD/omicron complex deviated between 3.2 to 3.3 nm, whereas the average Rg for the CB6-RBD/WT was 3.2 nm reflecting a higher degree of compactness (Figure 3h). The minimum energy conformation of both complexes was extracted from the trajectory for comparative interaction analysis. The total interface area between the RBD (WT) and CB6 is 1018.8 Å, which is reduced to 841.8 Å when CB6 interacts with the omicron RBD. Both polar and non-polar interface areas are reduced for the CB6 and omicron RBD. There is a steep decline observed in the solvent accessible surface area (SASA) of the CB6-RBD/WT complex towards the end

Click to enlarge

of the trajectory. An increased SASA for the omicron mutant complex may reflect conformational changes in the complex leading to a higher number of solvent exposed residues. The number of interacting residues at the interface for the WT RBD was 34; however, they were reduced to 26 in the case of omicron RBD. As a result, the total number of interface interactions also declined. The omicron mutated residues at the interface S496, R498, and Y501 fail to form hydrogen bond with CB6. Therefore, the number of hydrogen bonds were reduced from 18 to 11, resulting in an overall decrease in hydrophilic interactions at the interface.

Further, we identified the total binding free energy and the contribution of Vander Waals, electrostatic, and solvation energies of both complexes (Table 2). The electrostatic energy of the WT complex was -160.97 kcal/mol, which was reduced to +26.59 kcal/mol for the mutant complex. Although, there was only a slight decrease in the Vander Waals energy contribution upon mutation. However, the total binding free energy of the complexes revealed a stronger affinity between CB6 and the WT RBD than the omicron mutation- induced RBD. The residue mutations, N440K, G446S, S477N, T478K, Q493K, Q496S, Q498R, and N501Y at the CB6-RBD interface contributed negatively to the total binding free energy leading to an overall decrease in the binding affinity (Figure 4). The mutations may also affect the flexibility or rigidity of the residues at the binding interface governing the conformational dynamics and structural rearrangements of the RBD-antibody complex. The free energy landscapes (FEL) shown in (Figure 5) depicts the transition between various conformational ensembles of the RBD-antibody complexes. We evaluated and compared the local minima of the CB6 in complex with WT and omicron RBD. The FEL of the CB6-RBD/WT complex shows two distinct populations of conformations at the same energy basins. However, the ensembles of CB6-RBD/omicron complex have two different clusters at distinct energy basins separated by a transition barrier of 4.0 kcal/mol (Figure 5a). Hence, the omicron mutations in the RBD substantially affect the stability and the binding affinity of the CB6 antibody to its target epitope at the RBD.

Click to enlarge

Class 2: REGN10933 (Casirivimab)

We selected casirivimab as a class 2 antibody to evaluate the change in its binding with omicron mutation-induced RBD. REGN10933 antibody occupies the ACE2 binding region with a slightly different orientation than the CB6 antibody, as shown in (Figure 2c). Hence, the interface area and RBD residues involved in interaction largely differ. The antibody REGN10933 when binds to the WT RBD has an interface area of 663.25 Å, which is only reduced to 644.0 Å in the case of omicron RBD (Figure 6). Hence, no significant change was observed in the polar and non-polar interface area of the WT and mutant complexes. The SASA of these complexes remained relatively similar throughout the 100 ns MD trajectory (Figure 6e). However, the interface interaction statistics are altered upon the omicron mutations in RBD. The numbers of hydrophilic interactions at the interface were reduced, whereas a slight increase in the hydrophobic interactions was observed. The average number of hydrogen bonds between the antibody and RBD also reduced from 10 to 5 upon the mutations. No hydrogen bonds were formed at the interface region with a continuous stretch of mutant residues, K493, S496, R498, and Y501. Although, the overall stability of both complexes revealed fluctuations in the backbone RMSD throughout the simulation trajectory. A higher degree of compactness was observed for the WT complex with Rg between 3.4 to 3.6 nm, which was observed to be greater than 3.6 nm for the omicron mutant complex.

Click to enlarge

Figure 6: Molecular dynamics simulation results and interface statistics of RBD-REGN10933. (a) Wild type RBD and REGN10933 interface interactions. Residues involved in interactions are highlighted. Interface residues forming hydrogen bonds are labeled and shown as sticks. Hydrogen bonds are sown as black dashed lines. (b) Omicron induced mutant RBD and REGN10933 interface interactions. Mutant interface residues are shown in cyan and labeled in red color. (c) Polar and non-polar interface area of the complexes. The values are in Angstrom (Å) (d) Interface interaction count (HPhil; Hydrophilic, HPho; Hydrophobic). (e) Solvent accessible surface area (SASA) for the REGN10933-RBD/WT and REGN10933-RBD/omicron complex. (f) Plot for number of hydrogen bonds for the complexes. (g) Root mean square deviation (RMSD) plot for the complexes. (h) Radius of gyration (Rg) plot.

The calculated binding free energy of the REGN10933- RBD/WT complex was -52.97 kcal/mol and -30.00 kcal/ mol for the REGN10933-RBD/omicron complex (Table 2).

This difference in the binding energy was observed due to a notable decrease in the electrostatic energy at the interface. The Vander Waals energy term for the WT and omicron mutant complex was -68.08 kcal/mol and -71.15 kcal/ mol, respectively. Hence, the hydrophobic contacts at the interface were higher than the polar contacts for the mutant complex. The residue mutations S477N and T478K lead to more negative binding free energy, hence contributing positively to the total binding free energy of the complex (Figure 4). Moreover, the mutations E484A, Q493K, Q496S, and Q498R have positive values for the binding free energy, ultimately increasing the total binding free energy and hence decreasing the binding affinity of the REGN10933- RBD/omicron complex. The impact of these mutations on the lowest energy structural ensembles of each complex was observed through the FEL (Figure 5b). The FEL plot for both complexes shows majorly two distinct conformational ensembles with a transition barrier of 2.0 kcal/mol. These observations collectively suggest that there is no significant difference in the conformational stability of REGN10933 with RBD upon omicron mutations. Although, few mutations impact the binding affinity of the RBD with the REGN10933 antibody.

Class 3: S309 (Sotrovimab)

The NAb S309, aka sotrovimab is a therapeutic antibody approved for emergency use against the SARS-CoV-2 infection. Pre-clinical studies have suggested that sotrovimab retains activity against previously identified VOCs [36].

Click to enlarge

Figure 7: Molecular dynamics simulation results and interface statistics of RBD-S309. (a) Wild type RBD and S309 interface interactions. Residues involved in interactions are highlighted. Interface residues forming hydrogen bonds are labeled and shown as sticks. Hydrogen bonds are sown as black dashed lines. (b) Interface interaction count (HPhil; Hydrophilic, HPho; Hydrophobic). (c) Omicron induced mutant RBD and S309 interface interactions. Mutant interface residues are shown in cyan and labeled in blue color. (d) Polar and non-polar interface area of the complexes. The values are in Angstrom (Å) (e) Solvent accessible surface area (SASA) for the S309-RBD/WT and S309-RBD/omicron complex. (f) Plot for number of hydrogen bonds for the complexes. (g) Radius of gyration (Rg) plot for the complexes. (h) Root mean square deviation (RMSD) plot.

The antibody S309 targets an epitope in the conserved core domain of the RBD. We evaluated the binding stability and interaction of this class 3 antibody with the omicron mutation-induced RBD. The lowest energy conformation of the S309 bound to the two RBDs was extracted from the MD simulation trajectory. The total interface area between S309 and the WT RBD was 856.6 Å which reduced to 688.7 Å for the omicron RBD (Figures 7a and 7c).

Along with the polar and non-polar interface area, the overall number of interactions at the interface also decreased. The average number of hydrogen bonds at the interface for the WT and omicron complex was 12 and 8, respectively (Figure 7f). The SASA for both complexes was relatively equal up to 80 ns trajectory, and a slight decrease in the last 20 ns was observed for the WT complex. A similar trend in the trajectory was observed for the Rg of these complexes. Hence, a slightly more compact binding may occur in the case of the S309 interacting with the WT RBD. The backbone RMSD of the whole complex revealed stable conformation with lower fluctuations observed for the S309-RBD/omicron.

Further, binding free energy prediction revealed the difference in the affinity of the S309 to the WT RBD and omicron mutation-induced RBD. The total binding free energy of the S309-RBD/WT and S309-RBD/omicron complex was -76.6 kcal/mol and -52.3 kcal/mol, respectively (Table 2). This free energy change was predominantly due to a decrease in the contribution of electrostatic energy in the S309-RBD/ omicron complex. The residue mutation G339D leads to a positive binding free energy, contributing negatively to the total binding free energy of the complex (Figure 4). The energy contribution of the adjacent residue E340 was also reduced from -7.27 kcal/mol to -2.93 kcal/mol in the mutant complex. The binding energy and interactions were also affected by the mutation N440K. The residue N at position 440 could interact with S309 at the interface contributing its energy to the total free energy of the complex, whereas K at 440 fails to interact with S309. These changes could also be responsible for the difference in the local minima achieved by the conformation ensembles of these complexes. The FEL plot for the S309-RBD/ WT complex shows that the populations of conformations are clustered in two elongated ensembles occurring at the same energy basins (Figure 5c). On the other hand, multiple populations of S309-RBD/omicron are distributed along large conformational space with a single wide cluster occurring at the local minima. Hence, these observations altogether suggest that the omicron mutations may have a limited effect on the binding of class 3 antibody S309 to the RBD.

Class 4: S2X259

S2X259 is a human mAb which neutralizes SARS-

CoV-2, including all the previously identified VOCs, B.1.1.7, B.1.351, P.1, and B.1.427 [24]. It targets an epitope in the RBD overlapping some parts of the RBM as well as the core domain. We identified the effect of omicron mutations on this broadly neutralizing antibody. The total interface area between the RBD and the antibody was 829.85 Å, which only reduced to 812.0 Å upon mutations (Figures 8a and 8b). The SASA plot for both complexes shows stable values throughout the 100 ns simulation, with a slightly higher area for the S2X259-RBD/omicron complex (Figure 8c). The antibody S2X259 fails to make hydrogen bond interactions near the RBM region in the omicron mutation- induced RBD (Figure 8b). Therefore, the average hydrogen bonds for the complex S2X259-RBD/WT and S2X259-RBD/ omicron were 8 and 5, respectively. The residue mutation S375F resulted in the loss of hydrogen bond formation. An overall decrease in the number of hydrophilic interactions at the interface were also observed. Both RMSD and Rg plots for the S2X259-RBD/omicron complex experienced higher fluctuation throughout the MD simulation trajectory. Hence, the mutations may affect the overall stability of the RBD-antibody complex.

The difference in the binding free energy of the S2X259- RBD complexes reflects a slight decrease in the total binding affinity of the S2X259 towards the omicron RBD. Moreover, as observed for other antibodies described in the previous section, the electrostatic energy for the S2X259-RBD/ omicron complex is also highly reduced (Table 2). Hence, the total binding free energy for the WT and omicron mutant complex was -98.84 kcal/mol and -75.49 kcal/ mol, respectively. The residue mutations at the binding interface affect the mutant as well as adjacent residues binding energy contribution. The substitution, S371L, decreased the total binding free energy contribution of the adjacent residues N370 and A372. The residue mutations S373P and N501Y resulted in the loss of interaction with S2X259 at the interface. The substitution Y505H also contributed negatively to the total binding free energy of the S2X259 and the RBD complex (Figure 4). Further, the lowest energy conformational ensembles of the antibody and RBD complexes were evaluated through the FEL plots (Figure 5d). The FEL plot for the S2X259-RBD/WT complex shows two distinct populations of conformations confined to different energy basins separated with a transition barrier of > 4.0 kcal/mol. However, for the S2X259-RBD/ omicron complex, we observed multiple ensembles at wide conformational space with energy difference >5.0 kcal/ mol. Finally, the results suggest that the omicron mutations may particularly affect the binding stability of the S2X259 antibody and the RBD complex.

Å a b

$$ \mathrm {E} = \frac {1}{2} \mathrm {A} ^ {2} + \mathrm {B} ^ {2} $$

Interface area

$$ \mathrm {E} = \frac {1}{2} \mathrm {A} ^ {2} + \mathrm {B} ^ {2} $$

$$ \mathrm {E} = \mathrm {E} _ {1} + \mathrm {E} _ {2} + \dots + \mathrm {E} _ {n} $$

829.85

Click to enlarge

$$ \mathrm {E} = \frac {1}{2} \mathrm {A} ^ {2} + \mathrm {B} ^ {2} $$

S2X259

$$ \mathrm {E} = \frac {1}{2} \mathrm {A} ^ {2} + \mathrm {B} ^ {2} $$

RBD

$$ \mathrm {E} = \frac {1}{2} \mathrm {A} ^ {2} + \frac {1}{2} \mathrm {B} ^ {2} + \frac {1}{2} \mathrm {C} ^ {2} $$ Interface area $$ \mathrm {E} = \frac {1}{2} \mathrm {A} ^ {2} + \frac {1}{2} \mathrm {B} ^ {2} + \frac {1}{2} \mathrm {C} ^ {2} $$ $$ \mathrm {E} = \frac {1}{2} \mathrm {A} ^ {2} + \mathrm {B} ^ {2} $$

812.0 Å

$$ \mathrm {E} = \frac {1}{2} \mathrm {A} ^ {2} + \mathrm {B} ^ {2} $$

S2X259

$$ \mathrm {E} = \frac {1}{2} \mathrm {A} ^ {2} + \mathrm {B} ^ {2} $$

RBD

$$ \mathrm {E} = \frac {1}{2} \mathrm {A} ^ {2} + \mathrm {B} ^ {2} $$

(0micron)

Figure 8: Molecular dynamics simulation results and interface statistics of RBD-S2X259. (a) Wild type RBD and S2X259 interface interactions. Residues involved in interactions are highlighted. Interface residues forming hydrogen bonds are labeled and shown as sticks. Hydrogen bonds are sown as black dashed lines. (b) Omicron induced mutant RBD and S2X259 interface interactions. Mutant interface residues are shown in cyan and labeled in red color. (c) Solvent accessible surface area (SASA) for the S2X259-RBD/WT and S2X259-RBD/omicron complex. (d) Polar and non-polar interface area of the complexes. The values are in Angstrom (Å) (e) Plot for number of hydrogen bonds for the complexes. (f) Interface interaction count (HPhil; Hydrophilic, HPho; Hydrophobic). (g) Radius of gyration (Rg) plot for the complexes. (h) Root mean square deviation (RMSD) plot.