Sign in  |  Register
Cart

CSJ Journals

▾
  • -

2021, Vol.94, No.6

1794-1798
Selected PaperOpen Access

Computational Ab Initio Interaction Analyses between Neutralizing Antibody and SARS-CoV-2 Variant Spike Proteins Using the Fragment Molecular Orbital Method

Kazuki Watanabe ,1,†Chiduru Watanabe ,*2,3Teruki Honma ,2Yu-Shi Tian ,*1Yusuke Kawashima ,4Norihito Kawashita ,5Kaori Fukuzawa ,*4,6 and Tatsuya Takagi *1
1Graduate School of Pharmaceutical Sciences, Osaka University, 1-6 Yamadaoka, Suita, Osaka 565-0871, Japan
2Center for Biosystems Dynamics Research, RIKEN, 1-7-22 Suehiro-cho, Tsurumi-ku, Yokohama, Kanagawa 230-0045, Japan
3JST PRESTO, 4-1-8 Honcho, Kawaguchi, Saitama 332-0012, Japan
4School of Pharmacy and Pharmaceutical Sciences, Hoshi University, 2-4-41 Ebara, Shinagawa-ku, Tokyo 142-8501, Japan
5Graduate School of Science and Engineering Research, Kindai University, 3-4-1 Kowakae, Higashiosaka, Osaka 577-8502, Japan
6Department of Biomolecular Engineering, Graduate School of Engineering, Tohoku University, 6-6-11 Aoba, Aramaki, Aoba-ku, Sendai, Miyagi 980-8579, Japan
https://doi.org/10.1246/bcsj.20210104

Tools
PDF (1189K)

Abstract

Section:
Next section

The interaction energies between the receptor-binding domain of SARS-CoV-2 spike proteins and neutralizing antibody CC12.1 Fab were calculated using the fragment molecular orbital method. South African and Brazilian variants showed weaker interactions than the wild-type. Mutations, K417N/T and E484K, were considered to be responsible for escape from the antibody.

figure

The interaction energies between the receptor-binding domain of SARS-CoV-2 spike proteins and neutralizing antibody CC12.1 Fab were calculated using the fragment molecular orbital method. The South African and Brazilian variants showed weaker interactions than the wild-type. Mutations, K417N/T and E484K, were considered responsible for the escape from the antibody, because of changes in interactions with the surrounding amino acid residues.

1. Introduction
Section:
Previous sectionNext section

Pandemic caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is still ongoing worldwide.1 Vaccines developed by Pfizer, Moderna, and other pharmaceutical companies have been released;2,3 however, the global acquisition of herd immunity requires time. Recently, variant strains carrying multiple spike protein mutations (Figure S1) have become further threats.4,5 The most important variants are the South African (SA) variant, 501.V2,4 and the Brazilian (BRA) variant, lineage P1,5 carrying the mutations K417N, E484K, and N501Y and mutations K417T, E484K, and N501Y in the receptor-binding domain (RBD) of the spike protein,6 respectively. It has been reported that these variants can escape from monoclonal neutralizing antibodies7,8 and reduce the effectiveness of vaccines.9,10 Some studies have reported spike protein–antibody interaction analyses11,12 and the binding affinity of the angiotensin-converting enzyme 2–variants12,13 using the fragment molecular orbital (FMO) method.14–16

This study aims to conduct precise comparisons of RBD–neutralizing antibody interactions among wild-type (WT), SA, and BRA variants using the FMO method and provide crucial information for future antibody design.

2. Experimental
Section:
Previous sectionNext section

Variant RBD–antibody complexes were modeled from the co-crystallographic data of WT RBD (N334–P527 in the spike protein) and antibody CC12.1 Fab7,17 (PDB ID: 6XC2). Point mutations were conducted per sequence of variants using the Molecular Operating Environment. For both WT and variant complexes, sugar chains were removed, N- and C-terminals were capped with -NH2 and –COOH. All heavy atoms were fixed, and all hydrogen atoms were optimized without constraint. Subsequently, the fixed atomic coordinates of the side chain and main chain heavy atoms were gradually released. All hydrogen atoms were optimized without constraint, and heavy atoms were optimized under constraint (tether strength = 1 kcal mol−1 Å−1). The force field used was Amber10:EHT. The validity of variant RBD structures was further confirmed (see Supporting Information).

FMO calculations were performed with a computational level of FMO2-MP2/6-31G*18 using the ABINIT-MP program.16,19 The FMO method splits RBDs and antibodies into residual fragments, conducts ab initio calculations among fragments, and evaluates the inter-fragment interaction energy (IFIE) values. The IFIEs were further decomposed using the pair interaction energy decomposition analysis (PIEDA) method into four components: electrostatic (ES), exchange repulsion (EX), charge transfer with the mixed term (CT+mix), and dispersion (DI).20

\begin{equation*} \Delta E_{IJ} = \Delta E_{IJ}^{\text{ES}} + \Delta E_{IJ}^{\text{EX}} + \Delta E_{IJ}^{\text{CT} + \text{mix}} + \Delta E_{IJ}^{\text{DI}}. \end{equation*}

The sum of IFIE (IFIE-sum) for a particular fragment J (e.g., mutation residues on RBD) and the set of multiple fragments A (e.g., all antibody residues) can be calculated as follows:

\begin{equation*} \Delta E_{J}^{\text{A}} = \sum \nolimits_{I \in A}\Delta E_{IJ}. \end{equation*}

Furthermore, we analyzed the hydrogen bonds and XH/π interactions (X = C, N, and O) using geometric analysis to reveal the origin of the interaction energy. Final verification of these interactions by visual inspection was conducted based on the IFIE values and fragment pair geometry. The FMO calculation data of WT, SA, and BRA are available at FMODB (URL: https://drugdesign.riken.jp/FMODB/, FMODB ID: LL149, 9Y392, and MM5YZ, respectively).21,22

3. Results and Discussion
Section:
Previous sectionNext section

The RBD–antibody interaction energies were obtained as the IFIE-sum (Table 1). Compared with the WT (−566.1 kcal/mol), the SA and BRA variants showed lower IFIE-sum values (−448.4 and −430.8 kcal/mol, respectively), suggesting a decline of interaction owing to the mutations. Interaction energy analyses caused by mutations clarified that the mutations K417N/T and E484K remarkably decreased the interaction energy, whereas N501Y caused a much lower decrease (Tables 1 and 2). These results agreed with the fact that the pseudotype SARS-CoV-2, carrying the N501Y mutation, preserved the antibody sensitivity.23 Thus, we hypothesized that mutations K417N/T and N484K were responsible for escape from the CC12.1 Fab. Further structural analyses of K417N/T and E484K were conducted. WT K417 is a positively charged residue with a long side-chain, close to the antibody residues, allowing it to easily form strong electrostatic interactions (i.e., a salt bridge with D97H) and close interactions, such as hydrogen bonds and π-orbital interactions (Figure 1a, and Table S1a). In contrast, the mutated N417 and T417 have shorter side chains and are neutral residues, resulting in loss of the electrostatic interaction with D97H, and weak close interactions with the antibody (Figures 1b and 1c); the IFIE-sum values, indicating the interaction between the antibody and N417 and T417, were −51.7 and −26.2 kcal/mol, respectively (Table 2). The binding modes of the close interactions around WT K417 showed that the hydrogen atoms bound to the terminal nitrogen atom NZ formed two hydrogen bonds with the oxygen atoms (OD1 and OD2) on the side-chain of D97H and that the side-chain hydrogen atoms of K417 formed XH/π interactions with Y33H and Y52H on the CC12.1 Fab heavy chain (H chain) (Figure 1a and Tables S2a and S3). In contrast, N417 and T417 formed hydrogen bonds with Y52H, which was not observed in WT RBD, and the SA variant N417 also formed a hydrogen bond with D97H (Figures 1b and 1c, Tables S2b and 2c). No π-orbital interaction was formed between the SA variant N417, BRA variant T417, and the antibody aromatic residues. Thus, there was a change in the binding mode around K417N/T between WT RBD and variant RBDs. However, the main factors contributing to the interaction energy and changes in the contribution interaction type should be the differences in the side-chain length and charges.

Table
Table 1. IFIE-sum and energy decomposition values (kcal/mol) between whole RBD residues and CC12.1 Fab
Table 1. IFIE-sum and energy decomposition values (kcal/mol) between whole RBD residues and CC12.1 Fab
Complex IFIE-sum ES EX CT+mix DI
WT −566.1 −516.7 325.1 −141.2 −233.3
SA −448.4 −390.4 297.7 −131.5 −224.2
BRA −430.8 −366.6 278.8 −122.9 −220.1
Table
Table 2. IFIE-sum and energy decomposition values (kcal/mol) between each specific residue and CC12.1 Fab in different complexes
Table 2. IFIE-sum and energy decomposition values (kcal/mol) between each specific residue and CC12.1 Fab in different complexes
Position Residue Complex IFIE-sum ES EX CT+mix DI
417 K WT −101.1 −99.6 39.1 −15.8 −24.8
N SA −51.7 −55.9 26.7 −9.7 −12.8
T BRA −26.2 −22.7 10.4 −4.4 −9.5
484 E WT −52.2 −47.6 0.9 −2.9 −2.6
K SA 0.1 0.6 4.4 −2.2 −2.7
K BRA 2.0 2.6 4.3 −2.2 −2.7
501 N WT −32.2 −32.9 22.0 −9.6 −11.7
Y SA −25.2 −19.7 19.1 −8.3 −16.3
Y BRA −20.4 −14.5 16.0 −6.8 −15.1

Subsequently, we demonstrated the presence of the E484K mutation. Similarly, in this mutation, there was also a reversal in the residue charges. Currently, K484 can generate stronger electrostatic interactions with Y99H compared to WT E484 (Table S1b). However, owing to the presence of other positive residues (e.g., R94 and Y45), the increased interaction was canceled. When the overall influence of electrostatic interaction changes caused by this mutation was established, it was found that the total IFIE-ES values were significantly decreased (IFIE-ES values of WT, SA, and BRA were −52.2, 0.1, and 2.0 kcal/mol, respectively, Table 2). For close interactions, such as hydrogen bonds and π-orbital interactions, WT E484 formed a hydrogen bond between its side-chain and Y99H of the H chain of CC12.1 Fab (Figure 2a). In contrast, no hydrogen bonds were formed between Y99H and K484 in the SA and BRA variants (Figures 2b and 2c). However, the IFIE between E484 or K484 and Y99H did not change substantially, regardless of whether hydrogen bonds were formed or not (WT: −21.6 kcal/mol, SA: −23.2 kcal/mol, −22.9 kcal/mol. Table S1). Therefore, we concluded that the significant decrease in IFIE-sum with the whole antibody of the E484K mutation (SA variant: 0.1 kcal/mol; BRA variant: 2.0 kcal/mol, Table 2) was primarily caused by the changes in electrostatic interactions associated with the reversal charges of the residues.

In summary, using FMO calculations, we demonstrated that the mutations K417N/T and E484K resulted in a significant decrease in the RDB–antibody interaction energy and a change in the interaction types, suggesting that these mutations were the main factors mediating viral immune escape.

4. Conclusion
Section:
Previous section

This study compared the RBD–antibody CC12.1 Fab interactions among WT, SA variant, and BRA variants, based on the interaction energies obtained from FMO calculations and structures. The mutations, K417N/T and E484K, in the SA/BRA variant RBD caused significant energy disadvantages for antibody interactions. Mutation N501Y also negatively contributed to antibody interactions. However, the magnitude was much less. Therefore, we assumed that K417N/T and E484K, but not N501Y, were the major causes of antibody escape, which agreed with previous experimental results.7 Here, we successfully evaluated the binding energies between the antibody and SARS-CoV-2 WT and variant RBDs using the FMO method, and highlighted the important interactions around the mutations. These results may assist in the future design of SARS-CoV-2 antibodies.

This work was supported by JSPS KAKENHI Grant Number (JP20K06987 and JP17K082350). Part of this research was conducted using the FMO drug design consortium (FMODD, https://fmodd.jp/). The FMO calculations were performed using the Oakforest-PACS supercomputer (project ID: hp200101). PIEDA calculations were performed using the MIZUHO/BioStation software package. This research was partially supported by the Platform Project for Supporting Drug Discovery and Life Science Research [Basis for Supporting Innovative Drug Discovery and Life Science Research (BINDS)] from AMED under Grant Number JP20am0101113. C.W. acknowledges the support from JST PRESTO, Grant Number JPMJPR18GD, Japan. Y. S. T. acknowledges the support from Hirose International Scholarship Foundation research grant.

Supporting Information

The SI file includes experimental details, three tables, and five figures. Table S1, Interaction energies between K417N/T or E484K and its important surrounding residues; Table S2, XH–Y hydrogen bonds between K417 (N/T) or E484 (K) on SARS-CoV-2 RBD and the residue on CC12.1 Fab; Table S3, XH/π interactions between K417 on the WT RBD and CC12.1 Fab detected by the CHPI program; Fig. S1, Structure of RBD and CC12.1 Fab complex and conformation around mutant residues; Fig. S2, Superposes of X-ray WT, SA, and BRA RBD structures (7NX6, 7NXA, and 7NXB, respectively) and RMSD (Å) matrices of all heavy atoms; Fig. S3, Superposes of X-ray WT RBD complexed with COVOX-222 and EY6A Fabs and CC12.1 Fab (7NX6 and 6XC2, respectively), the site views of mutation points (N501, K417, and E484), and RMSD (Å) matrices of all heavy atoms; Fig. S4, Superposes of X-ray structure of SA RBD complexed with COVOX-222 and EY6A Fabs and our model of those complexed with CC12.1 Fab (7NXA_SA and 6XC2_SA, respectively), the site views of mutation points (Y501, N417, and K484), and RMSD (Å) matrices of all heavy atoms; Fig. S5, Superposes of X-ray structure of BRA RBD complexed with COVOX-222 and EY6A Fabs and our model of those complexed with CC12.1 Fab (7NXB_BRA and 6XC2_BRA, respectively), the site views of mutation points (Y501, T417, and K484), and RMSD (Å) matrices of all heavy atoms; This material is available on https://doi.org/10.1246/bcsj.20210104.

Tatsuya Takagi

Tatsuya TAKAGI graduated from School of Pharmaceutical Sciences, Osaka University in 1979. After spending 13 years at School of Pharmaceutical Sciences, Osaka University as a Research Associate since 1980, he became an Associate Professor at Genome Information Research Center, Osaka University. While he was a Research Associate, he received his Ph. D. from Osaka University in 1988. In 1998, he was promoted to a Professor of Graduate School of Pharmaceutical Sciences, Osaka University.

<< Previous IR Intensities of CO Molecules Adsorbed on Atop and Low-Coordinate Sites of Pd Nanoparticles: Analysis Using Natural Perturbation Orbitals Next >> Elucidation of the Mechanism of Quantum Dot Arrangement Based on Self-Assembly of an Azobenzene Derivative

Select Issues

▾
  • -

Search for Articles

Advanced

Saved Search

Please login to view your saved searches.

Top Accessed Articles

Most Cited

Related Articles

  1. Difference in the Potential Energy Surfaces from the Fragment MO Method and Conventional Ab Initio SCF-MO Method. A Case of a Surface for Ring Rotation of (−)-Epicatechin Gallate Using the STO-3G Basis Set
  2. Ab Initio MO-MD Simulation Based on the Fragment MO Method. A Case of (−)-Epicatechin Gallate with STO-3G Basis Set
  3. SARS-CoV-2 Proteins Microarray by Photoimmobilization for Serodiagnosis of the Antibodies
  4. Molecular Recognition Mechanism between Mesohemin and the Antibody 11D1
  5. Fragmentation Position Dependency of the Total Energy and Atomic Charge Difference between the Fragment MO Method and Conventional Ab Initio SCF-MO Method. A Case of (−)-Epicatechin Gallate with STO-3G Basis Set
MANAGE YOUR ALERTS
RECOMMEND TO LIBRARY