key: cord-0687403-m2t06rw5
authors: Contractor, Darshan; Globisch, Christoph; Swaroop, Shiv; Jain, Alok
title: Structural basis of Omicron immune evasion: A comparative computational study of Spike protein-Antibody interaction
date: 2022-03-15
journal: bioRxiv
DOI: 10.1101/2022.03.15.484421
sha: 02cf16efba756acda701503824a91df0c1ea2632
doc_id: 687403
cord_uid: m2t06rw5

The COVID-19 pandemic has caused more than 424 million infections and 5.9 million deaths so far. The vaccines used against SARS-COV-2 by now have been able to develop some neutralising antibodies in the vaccinated human population and slow down the infection rate. The effectiveness of the vaccines has been challenged by the emergence of the new strains with numerous mutations in the spike (S) protein of SARS-CoV-2. Since S protein is the major immunogenic protein of the virus and also contains Receptor Binding Domain (RBD) that interacts with the human Angiotensin-Converting Enzyme 2 (ACE2) receptors, any mutations in this region should affect the neutralisation potential of the antibodies leading to the immune evasion. Several variants of concern (VOC) of the virus have emerged so far. Among them, the most critical are Delta (B.1.617.2), and recently reported Omicron (B. 1.1.529) which have acquired a lot of mutations in the spike protein. We have mapped those mutations on the modelled RBD and evaluated the binding affinities of various human antibodies with it. Docking and molecular dynamics simulation studies have been used to explore the effect of the mutations on the structure of the RBD and the RBD-antibody interaction. The analysis shows that the mutations mostly at the interface of a nearby region lower the binding affinity of the antibody by ten to forty per cent, with a downfall in the number of interactions formed as a whole and therefore, it implies the generation of immune escape variants. Notable mutations and their effect was characterised by performing various analyses that explain the structural basis of antibody efficacy in Delta and a compromised neutralisation effect for the Omicron variant. Our results pave the way for robust vaccine design that can be effective for many variants. Graphical Abstract Synopsis The research study utilises comparative docking and MD simulations analyses to illustrate how mutations in delta and omicron variants affect the binding of antibodies to the spike receptor binding domain (RBD) of SARS CoV-2.

Human Coronavirus disease 2019 (COVID-19) is an extremely infectious disease which is caused by SARS coronavirus 2 (SARS-CoV-2). The disease was first reported in Wuhan, China in mid-December 2019 and has spread globally after that, resulting in the ongoing pandemic. Symptoms vary from mild in majority of the cases to severe like pneumonia and multi-organ failure in few 1, 2 .

The genome of SARS-CoV-2 is a single-stranded positive-sense RNA that codes for 10 genes and 26 proteins. The surface spike protein (S) is not only important for the viral entry but also is the main immunogenic protein in the virus 3 . The S protein is known to be cleaved into an amino-terminal S1 subunit which is involved in virus-host cell receptor binding, and a carboxyl-terminal S2 subunit that is responsible for a virus-host membrane fusion. The S1 subunit contains two domains, an N-terminal (NTD) and a C-terminal (CTD) domain where the latter called receptor binding domain (RBD) as it is involved in receptor binding. The RBD of S protein is a promising target for molecules which can bind to RBD cause viral entry inhibition 4, 5 .

Vaccines are used as a main prophylactic measure against any type of infection. They mimic the natural infection in terms of activating the host immune response, including antibody generation, to develop an immunity against that pathogen. The use of vaccines against SARS-CoV-2 has been a major success in reducing the pace of COVID-19 pandemic. Till now, more than 100 vaccines of different types have been developed, and around 26 vaccines have undergone phase III clinical trials, as per WHO 6 [8] [9] [10] [11] [12] . The specific mutations in that region of the spike protein which is the target of the antibodies leads to escaping the immune response when compared to the original Wuhan strain or D614G variant [13] [14] [15] [16] [17] [18] .

Till November 2021, the delta variant of the SARS-CoV-2 was the main variant of concern as it could spread over 163 nations by August 2021 19 There are several experimental and computational studies which clearly show the cause of a high transmission of the Omicron variant and its immune evasion [22] [23] [24] [25] . The increase in the binding of the SARS-CoV-2 receptor ACE2 with the RBD is the main cause of the enhanced transmission of the Omicron variant. The RBD-ACE2 receptor interaction is very well characterised in the case of VOCs that emerged before Omicron and has been evidenced as the major cause of higher transmission in the Delta variant. Kumar et al. 23 performed computational studies to assess the effect of spike protein mutations in the Delta and Omicron variants. It was found that the Omicron variant showed higher binding affinity with the human ACE-2 receptor as compared to Delta variant. In another study, it was shown that the mutations in the Omicron RBD yields more contacts, hydrogen bonds and buried surface area at the interface of spike RBD and ACE-2 receptor, as compared to the original strain 24 . Rath et al. 22 reported 2.5 times stronger binding between the mutated residues and ACE-2 receptor together with a much relaxed dynamics of the complex, as compared to the wild type.

In this study we focus on finding the molecular basis of immune evasion by the Omicron variant, as it is already known to infect the previously infected as well as vaccinated population. Cao et al. 26 determined the escape mutations in the RBD against a panel of 247 anti-RBD Nabs and found that several single Omicron mutations lead to impaired Nabs of various epitopes. In another study, an infectious Omicron virus isolated in Belgium was tested for its sensitivity to the antibodies in the sera of 115 people which were either vaccinated or recovered from COVID infection, and against nine clinically approved monoclonal antibodies (mAbs). Omicron showed partial or total resistance to neutralisation by all mAbs. Similar results were observed for the sera from vaccinated individuals and convalescent sera collected 5 to 6 months after the vaccination or infection of individuals 27 .

In a related study, Liu et al. 24 reported diminished neutralisation of the Omicron by the convalescent sera and sera from the vaccinated individuals. Similar results were found for 17 out of the 19 mAbs specific to all known epitope clusters on the spike protein 27 .

In this article, we used computational tools like molecular docking, MD simulations, free energy calculations and thorough structure analysis to assess the effect of such mutations in the Delta and the Omicron variants on the structure of RBD and its binding with the major neutralising antibodies generated against SARS-COV-2. These antibodies include CR3022 28 , S230 29 , CC12.1 30 , REGN10987 31 and S309 32 . The results show that the mutations in the RBD and few at other positions in the spike protein lowers the binding affinity between the RBD and the various antibodies. We also looked into the original interactions which are hampered and the new interactions resulting out of these mutations. Such studies are useful not only to understand the cause of immune evasion by these VOCs at a molecular level, but also to predict the immune evasion ability of the upcoming variants and act accordingly for the preparedness against those variants.

Any protein-protein or protein-ligand docking is represented in the form of 'best fit' orientation between them which is the principle of molecular docking. The binding affinity of the various neutralising antibodies to that of the receptor binding domain of the spike protein was evaluated by molecular docking analysis.

The RBD of the spike proteins from three variants of SARS-CoV-2, the wild type, the delta (Table S1) , which was in contrast to the published data that clearly showed a significant reduction in the neutralisation of the omicron strain by corresponding antibodies.

Such observations suggest that introducing the point mutations followed by molecular docking may not give the correct picture as reported previously 27 . (Table S1 ) as the mutations can lead to a partially modified fold and performed molecular dynamic simulation studies on the spike RBD structures, which allowed them to acquire a more reliable adopted dynamic conformation. After the MD simulations of the spike RBDs, we extracted the most populated structure for all the three strains using the gromos clustering method. Such an approach overcomes the biases towards the end structure and represents the ensemble of most visited conformations during the MD. Superposition of starting and the most populated structure along with the location of the mutations are shown in Figure 2 . It is clearly visible that the relaxed (wild strain) and mutated structures (Delta and Omicron strains) exhibit some conformational changes. These changes were more prominent at the epitopes related to the binding of some of the antibodies. Overall wild to delta to omicron variants suggest an increased stability to the viral protein via the attained mutations. We also compared all the three structures after MD ( Figure 2D ) visualising the differences across them and noticed most visible changes at the mutational site. Represented structures after MD were then subjected to molecular docking analysis. We examined the binding affinity of the protein-neutralising antibody complex due to induced mutations especially after considering the conformational changes on the protein structures during MD simulations. The best conformations generated from the docking analysis obtained by the ClusPro 2.0 33 were visualised and comparative data was interpreted highlighting the phenomenon of escapism of the mutated SARS COV-2 virus particles from neutralisation by the antibodies due to mutations at the binding sites of the protein.

Comparative docking scores of all the three variants were tabulated in Table 1 percentage reductions in their neutralisation capability respectively. Similar reduction was reported in many experimental reports 36 but as per best of our knowledge none of them explained the molecular mechanism in atomistic level which is discussed below. It is important to highlight that for all the antibodies either other epitopes and/or a different binding pose was observed compared with the crystal structures ( Figure S1 ). The S230, S309

and CR3022 antibodies were binding at different locations while the remaining two displayed different binding poses. Such significant changes for the antibodies in terms of docking scores and binding locations correlated very well with their epitopes and positions of mutated residues in Omicron strain. 1 1 CC12.1 exhibits more than 40% reduction in the docking score mainly due to major mutations happening near the receptor binding epitope of spike RBD, thus, aiding mutated variants to escape the neutralising antibodies.

A closer look at the spike RBD/CC12.1 interface ( (Table 1 ) could be noticed. It was observed that for the omicron strain, spike RBD residues that participate in interactions with neutralising antibodies were either mutated residues (blue circle) or in the close proximity with the mutated residues (orange circle) thus having a significant influence of mutations ( Figure 4B ). Such observation was majorly absent for the delta strain. Thus, suggested a probable easy escape of the mutated omicron spike protein from coming in recognition by the antibodies and forming a neutralisation complex. A similar trend was observed in the case of the antibodies S309, REGN10987, CR3022 and to some extent for S230 as shown in (Fig S2 to S8) . Overall a thorough structural analysis explains the effect of mutations on the antibody-antigen interaction interface and is very well correlated with the docking data. 

Our findings motivated us to analyse the representative complexes via more sophisticated the Delta variant also exhibited flexibility to some extent that allowed it to be neutralised by many antibodies however in some cases its affinity decreased as observed in the molecular docking results. Importantly, the Omicron variant displayed relatively rigid structures that must be the after-effect of the large number of mutations. As antigen-antibody interaction is governed by a delicate balance of rigidity and flexibility 35 , this was lost in the case of the Omicron variant leading to a significant reduction of the binding score for the omicron variant (Table 1) .

To characterise the interaction interface of the spike RBD and the antibody, Solvent

Accessible Surface Area (SASA) of the RBD/antibody interface was calculated. Averaged distributions from all the simulations are displayed in Figure 5B and 5C. It is clearly evident that WT and Delta strain had higher SASA available from both interface sides compared to the Omicron strain. It indeed confirms that both strains can interact with the antibodies with higher affinity further helping in neutralisation due to a large available interface to form antigen-antibody complexes.

Additional quantification of the interaction strength is provided by the number of contacts between the spike RBD and the antibody. As expected both, WT and Delta strain, form almost similar numbers of contacts with marginally edge to WT strain ( Figure 5D ). On the other hand, the Omicron strain displayed a ~50% reduction in total number of contacts, which correlated well with the docking score. It clearly depicts the effect of the large number of mutations for the Omicron variant. In the next step, we classified the residues involved in these interactions from the RBD and the antibody perspective. Residues which form contacts for at least 50% of the simulation time during the last 75% of simulations (100 to 300 ns) are illustrated in Figure 6A and 6B. Depicted are residues involved from the RBD and antibody interface along with their probability of being in contact with the counterpart protein respectively. With no surprise, both WT and Delta strains were majorly involved in very stable contacts as largely blue and green coloured dots were observed in the range of 80-100% stable contacts.

It is important to mention that residues involved in such contacts were distinct in WT and Delta strain. It suggests that due to mutations in the spike RBD the antibody binds at the 

Finally we have quantified the binding energy of antibody-antigen complexes using two approaches. In the first approach we have calculated the average MM-PBSA free energy throughout the trajectory by considering structures at every 10 ns (Table 2 ). In the other approach we have identified the most populated structure during the MD simulation using the gromos clustering method and calculated the free energy for all three complexes averaged over all three sets of simulations (Table S3 ). The in-depth analysis of the mutations at the interfacial residues and its effect on the binding with the neutralising antibodies across the major dominant variants helps in the designing of the consensus based immunogens where the highly mutable and critical residues could be excluded in the peptide based immunogen sequence. Such immunogens are expected to elicit broadly neutralizing antibodies which may work against the future variants as well 38 .

Similarly, the analysis of antibody-antigen contact surfaces using computational tools could be used to guide for the choice of mutations for modelling the antigen-antibody complexes and the rational affinity engineering of therapeutic antibodies 39 .

The 

Spike Protein. The receptor binding domain (RBD) of the spike protein (S1) was selected for the study, as the major population of the serum neutralising antibodies target the RBD domain of the spike protein, supported by various clinical and serological studies done on the COVID-19 infected patients 40 .

The three dimensional structure of the wild strain was retrieved from Protein Data Bank 41 The interaction residues and binding sides were selected based on the literature 28-32 , together with visualisation and identification of the molecular interactions among the spike RBD and the antibodies using UCSF Chimera and LigPlot+ 45 . During the docking process, antibody mode was selected for generating the best clusters after docking analysis and complementarity-determining regions (CDRs) were masked for effective antigen-antibody complex formation 46 . Best models were selected according to the highest Cluster member size generated and therefore having the optimum score. The collective scores of all the docked complexes were compared and the results were interpreted.

Since the static structures provide only superficial information as the mutations can lead to a partially modified fold, the wild and the mutant spike RBD structures were subjected to molecular dynamic simulations studies to provide more reliable adopted dynamic conformations. All the strains were subjected to 100 ns of MD simulations (as discussed below) and representative structures were extracted for further docking studies. These representative structures were obtained by clustering of the MD simulations with the gromos algorithm together with a cut off distance of 0.2 nm and correspond to the cluster centroid of the biggest obtained cluster.

Complex structures after the molecular docking were visualised using UCSF-Chimera 43 and 2D plots were generated using Lig-Plot + 45 .

All molecular dynamics simulations were performed with GROMACS version 2021.4 47 . We used the CHARMM36m 48,49 ; forcefield for the antibody protein CC12.1 (PDB ID: 6XC2) 50, 51 together with the TIP3P water model 52 . The force field parameters for the system have been generated with the Input generator tools in CHARMM-GUI 51,53,54 using Solution Builder.

The two missing fragments (three and four residues) in the chain H of the antibody part of the structure were built with modeller using the loop modeller routine 55, 56 . The initial structures for the Delta and Omicron mutant were obtained by docking the antibody into a representative structure of the RBD, obtained from a MD simulation, with ClusPro 2.0 as discussed above.

The simulation box was set to dodecahedron and defined in such a way that the minimum distance of the structure and the box was at least 1.5 nm for the initial RBD simulation and 2.0 nm for the docked complexes and subsequently solvated with water and neutralised with potassium chloride together with an additional concentration of 150 mmol/L.

The following settings have been applied. The Leapfrog integrator was utilised together with all bonds being constrained by the LINCS algorithm 57 in order to enable a time-step of 2 fs.

We used a modified cut off for short-ranged electrostatic and Lenard Jones interactions of 1.2 nm and applied a switching function to smoothly approach the cut off between 1.0 and 1.2 nm.

Particle mesh Ewald (PME) 58 method was applied to calculate Long-range Coulomb interactions. The neighbour list was updated every 10 steps. In a first step, all systems were conducted to energy minimisation with the steepest-descent algorithm for 50000 steps.

Subsequently two consecutive equilibration simulations followed (100 ps each) in a canonical cluster. This approach mimics the molecular docking approach as was considering a single, most populated structure for the energy calculation.

Evaluation, and Treatment of Coronavirus

Clinical Characteristics of Laboratory Confirmed Positive Cases of SARS-CoV-2 Infection in Wuhan, China: A Retrospective Single Center Analysis

Cryo-EM Structure of the 2019-NCoV Spike in the Prefusion Conformation

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor

Structural and Functional Properties of SARS-CoV-2 Spike Protein: Potential Antivirus Drug Development for COVID

World Health Organization Covid-19. Home/Publications/Overview/COVID-19

Peiffer-Smadja, N. Comparing COVID-19 Vaccines for Their Characteristics, Efficacy and Effectiveness against SARS-CoV-2 and Variants of Concern: A Narrative Review

Assessing Transmissibility of SARS-CoV-2 Lineage B.1.1.7 in England

Model-Based Estimation of Transmissibility And

Increased Mortality in Community-Tested Cases of SARS-CoV-2 Lineage B.1.1.7.Nature2021

Risk of Mortality in Patients Infected with SARS-CoV-2 Variant of Concern 202012/1: Matched Cohort Study

Genomics and Epidemiology of the P.1 SARS-CoV-2 Lineage in Manaus

Sensitivity of Infectious SARS-CoV-2 B.1.1.7 and B.1.351 Variants to Neutralizing Antibodies

Neutralizing Antibodies Against SARS-CoV-2 Variants After Infection and Vaccination

Antibody Resistance of SARS-CoV-2 Variants B.1.351 and B.1.1.7.Nature2021

Escape of SARS-CoV-2 501Y.V2 from Neutralization by Convalescent Plasma

SARS-CoV-2 501Y.V2 Escapes Neutralization by South African COVID-19 Donor Plasma

Effectiveness of Covid-19 Vaccines against the B.1.617.2 (Delta) Variant

Delta Variant: What Is Happening with Transmission, Hospital Admissions, and Restrictions?BMJ2021

World Health Organization Covid-19. Classification of Omicron (B.1.1.529): SARS-CoV-2 Variant of Concern

Omicron (B.1.1.529) Variant

Scanning the RBD-ACE2 Molecular Interactions in Omicron Variant

Omicron and Delta Variant of SARS-CoV-2: A Comparative Computational Study of Spike Protein

Mutations on RBD of SARS-CoV-2 Omicron Variant Result in Stronger Binding to Human ACE2 Receptor

SARS-CoV-2 Omicron Spike Glycoprotein Receptor Binding Domain Exhibits Super-Binder Ability with ACE2 but Not Convalescent Monoclonal Antibody

Structures of the Omicron Spike Trimer with ACE2 and an Anti-Omicron Antibody

Considerable Escape of SARS-CoV-2 Omicron to Antibody Neutralization

A New Crystal Form of the SARS-CoV-2 Receptor Binding Domain: CR3022 Complex-An Ideal Target for In-Crystal Fragment Screening of the ACE2 Binding Site Surface

Unexpected Receptor Functional Mimicry Elucidates Activation of Coronavirus Fusion

Structural Basis of a Shared Antibody Response to SARS-CoV-2

Studies in Humanized Mice and Convalescent Humans Yield a SARS-CoV-2 Antibody Cocktail

Cross-Neutralization of SARS-CoV-2 by a Human Monoclonal SARS-CoV Antibody

New Additions to the ClusPro Server Motivated by CAPRI

Rapid and Accurate Identification of SARS-CoV-2 Variants Containing E484 Mutation

Function and Properties of Antibody Binding Sites

Structural Basis of SARS-CoV-2 Omicron Immune Evasion and Receptor Engagement

SARS-CoV-2 Omicron Variant: Antibody Evasion and Cryo-EM Structure of Spike Protein-ACE2 Complex

SARS-CoV-2 Consensus-Sequence and Matching Overlapping Peptides Design for COVID19 Immune Studies and Vaccine Development.Vaccines

Structural Consensus among Antibodies Defines the Antigen Binding Site

SARS-CoV-2 Variants, Spike Mutations and Immune Escape

Announcing the Worldwide Protein Data Bank

Structure of the SARS-CoV-2 Spike Receptor-Binding Domain Bound to the ACE2 Receptor

UCSF Chimera-A Visualization System for Exploratory Research and Analysis

Recognition of the SARS-CoV-2 Receptor Binding Domain by Neutralizing Antibodies

Multiple Ligand-Protein Interaction Diagrams for Drug Discovery

Application of Asymmetric Statistical Potentials to Antibody-Protein Docking

GROMACS: High Performance Molecular Simulations through Multi-Level Parallelism from Laptops to Supercomputers

Implementation of the CHARMM Force Field in GROMACS: Analysis of Protein Stability Effects from Correction Maps, Virtual Interaction Sites, and Water Models

CHARMM36m: An Improved Force Field for Folded and Intrinsically Disordered Proteins

A Web-Based Graphical User Interface for CHARMM

Automated Sugar Identification and Simulation Preparation for Carbohydrates and Glycoproteins

Comparison of Simple Potential Functions for Simulating Liquid Water

CHARMM: The Biomolecular Simulation Program

Simulations Using the CHARMM36 Additive Force Field

Comparative Protein Structure Modeling Using MODELLER

Modeling of Loops in Protein Structures

LINCS: A Linear Constraint Solver for Molecular Simulations

A Smooth Particle Mesh Ewald Method

Canonical Dynamics: Equilibrium Phase-Space Distributions

Polymorphic Transitions in Single Crystals: A New Molecular Dynamics Method

Gmx_MMPBSA: A New Tool to Perform End-State Free Energy Calculations with GROMACS

On the Role of the Crystal Environment in Determining Protein Side-Chain Conformations

A Hierarchical Approach to All-Atom Protein Loop Prediction

The authors acknowledge support by the High Performance and Cloud Computing Group at 

The supporting information contains the molecular docking data, the antibody-RBD complexes for other antibodies, the details of interacting residues and the results from MM/GBSA calculations.