In silico evaluation of anti SARS-CoV-2 antibodies neutralization power: A blueprint with monoclonal antibody

Immune escape caused by genetic variations of SARS-CoV-2 S protein immunogenic epitopes affects the e�ciency of monoclonal antibody-based therapy of COVID-19. Therefore, predicting the effects of these variations on immune escape is important to adapt rapidly anti SARS-CoV-2 Mab therapy. We herein describe a computational method to evaluate the neutralizing power a monoclonal antibody speci�c of a given SARS-CoV-2 variant and to compare it to its potential neutralizing power of others and emergent variants. The method’s calls for building in silico complex between the spike protein of a SARS-CoV-2 variant and a neutralizing antibody, analyzing the molecular interactions pattern and calculating the binding energy. This data is assigned a neutralizing value of 100% to which can be compared the neutralization value of any SARS-CoV-2 variant determined after molecular replacement in the complex of the RBD sequence with the RBD of this variant. Application of this method to the class 3 neutralizing antibody Sotrovimab and 24 variants and subvariants showed that the a�nity binding and neutralizing power, decreased gradually with new variants. This method is of interest to adapt the use of therapeutic antibodies to the treatment of emerging variants. It could be applied to antibody-based treatment of other viral infections.


Introduction
While the world is entering at a ground level the fourth year of the severe acute respiratory syndrome pandemic caused by the newly emergent coronavirus SARS-CoV-2, this persistent virus is still lingering away.This is mainly due to the virus relatively high mutational rate with speci c mutations occurring on the spike protein affecting its immunogenicity [1,2].The battle against this virus covers several aspects ranging from prevention, mitigation, and treatment.One promising approach that is still developing with proven e ciency consists of using anti-SARS-CoV-2 monoclonal neutralizing antibodies.However, selective pressure caused by infection and/or vaccination is accelerating the emergence of new variants and sub variants, which poses a challenge not only to antibody-mediated therapy but also to vaccine use and development.Anti-SARS-CV-2 monoclonal antibodies recognize speci c epitopes mainly on the spike protein preventing target cell binding and/or fusion and accumulation of mutations in theses speci c epitopes increases the tness of the virus.Additionally, the e cacy of the available anti-SARS-CoV-2 neutralizing antibodies (NAbs) therapies varies dramatically and is di cult to foresee how useful would it be for new circulating variants [3].
Currently, the Food and Drug Administration (FDA) and the European Medicines Agency (EMA) have issued emergency use authorization for several anti-SARS-CoV-2 NAbs including Evusheld, Ronapreve and Regkirona, Sotrovimab (S309), Casirivimab and Imdevimab (REGN-COV2), and Bamlanivimab (LY-CoV555, LY3819253) [4,5] and many more are still under evaluation.Based on their binding site, theses neutralizing antibodies are classi ed into different groups.There are currently two classi cation methods [6].One is based on high-throughput surface Plasmon resonance technique combined with negative stain electron microscopy to identify speci c epitopes on the receptor-binding domain (RBD).This method groups the NAbs into seven distinct communities denoted RBD-1 to RBD-3, RBD-4 to RBD-5, and RBD-6 to RBD-7.The other method is based on considerations such as the overlap of the neutralizing antibody with the ACE2 binding site and if it recognizes activated (up) or baseline (down) states of RBD.Four different classes were described, class I-class IV, where class I compete on the ACE2 binding site and can bind the receptor binding domain (RBD) in its up position while class II binds the RBD in both states (UP and Down).Class III neutralizing antibody bind an interface that is outside the RBD domain and hence does not compete with the ACE2 receptor and will bind both forms of the RBD while class IV bind only RBDs in the up position [7,8].
The computational method we describe in this paper was developed to evaluate the interaction between a given neutralizing antibody of a speci c SARS-CoV-2 variant, compare it to the interaction with a different variant and thus predict a possible immune escape.It uses as a model of the interaction of the neutralizing monoclonal antibody Sotrovimab (S309) with SARS-CoV-2 Wuhan variant.This monoclonal antibody (Mab) was rst isolated from the memory B cells of a SARS-CoV survivor patient [9,10].It has been reported to have neutralization potencies for SARS-CoV, SARS CoV-2 and SARS-like coronaviruses.Currently, it is one of only two approved therapeutic monoclonal antibodies for newly emerged Omicron subvariants [7,11,12].S309 is a recombinant human monoclonal antibody used under the generic name (Xevudy®).In May 2021, it was granted rst emergency use for the early treatment of COVID-19 [13].S309 belongs to class III antibodies that are characterized by their binding site on the spike protein, as they do not compete with the receptor binding domain (ACE2) [7].While ACE2 binds to the SARS-CoV-2 spike residues between K417 and Y505 [14], S309 recognizes a distinct proteoglycan epitope at the opposite of the ACE2 binding site involving residues N334, E340, N343, T345, R346, K356 and a structural loop (443-450) that can be accessed on both states of the RBD (up and down).These key glycan residues are not affected by mutations of the new omicron subvariants [7,15].However, other mutations found on the structural loop seems to have a signi cant effect on the neutralization capacity of S309.Since S309 does not compete with ACE2 binding site, its neutralization mechanism does not depend on direct blocking of RBD.Though, binding of S309 to the SARS-CoV-2 -2 spike receptor-binding domain induces rather an antibody-dependent cell cytotoxicity (ADCC) and antibody-dependent cellular phagocytosis (ADCP) [16].
Several experimental and clinical reports described the neutralization effect of Mab S309 with early Wuhan strain and its effect in reducing disease progression [10,17,18].Therefore, in the computational method we report in this paper, the estimated interaction a nity of Mab S309 to the Wuhan strain is assigned a value of 100%.Comparison of the estimated a nities of S309 to each SARS-CoV-2 variants to this reference value allows the evaluation of the neutralization e ciency and the prediction of possible immune escape for each existing or newly emerging variant.This straightforward computational method can rapidly give valuable insights on the eventual e ciency of existing neutralizing therapeutic antibodies in treating newly emergent variants prior to the experimental methods.Since immune escape is a major criterion retained by WHO and the CDC in their labelling systems of new variants particularly the variants of concern (VOC) [19], this method can also be considered to label new variants early after their emergence.

Results
Method development work ow. Figure 1 outlined the methods for the anti SARS-COV-2 antibody neutralization potential of S309.The blueprint of the method we developed using Mab S309 an experimentally proven neutralizing Mab for SARS-CoV-2 and its variants is described in Fig. 1a.We proceeded by modifying available model PDB ID: 7YAD to generate a reference model that can be used to measure the neutralization potential in terms of binding a nity ΔG [Fig.1b].Several in silico 3D models representing spike monomer chain of each variant were generated.The quality of the generated 3D model was evaluated based on the homology modeling report and Swiss-model structural assessment.The generated models showed a QMEAN Z-Score between − 1.0 and − 3.2 indicating a good quality model where Z-scores around 0.0 is ideal and any value below − 4.0 indicates low quality model [20].The QMEANDisCo Global value represent the combined scoring of global (for the entire structure) and local (per residue) absolute quality estimates of a single model [21].Our models QMEANDisCo scores were ranging from 0.64 to 0.76 ± 0.05.These values re ect a good quality model (any value below 0.6 represent low quality model).Each complex was built by molecular replacement of chain M of the reference model with the extracted RBD followed by binding a nity and interaction analysis.
Analysis of the molecular interaction pattern of Sotrovimab (S309) with nine main SARS-CoV-2 Variants.The generated complexes were energy minimized and polar and hydrophobic interactions were analyzed.Several interactions identi ed between of S309 Fv domain and spike RBD with more interactions towards the heavy chain.
Interacted residues of the spike protein include residue 321-428 of SARS-CoV spike and 334-441 of SARS-CoV-2 and its variants.SARS-CoV showed four polar interactions compared to the Wuhan SARS-CoV-2 that shares a total of three polar interactions with S309.Interestingly, variant Kappa shows the highest number of polar interactions of six while Delta 21J shows the lowest with only one polar interaction.Kappa variant shows a unique two salt bridges between Arg346 and Lys356 with S309 heavy chain amino acid Glu108.All the variants share the same polar interaction between Glu340 and S309 heavy chain Ala104 except for Delta 21J variant.All Omicron subvariants showed the same interaction pattern except for BA.2.75 with one missing polar interaction between Thr343 and S309 heavy chain Ser109.GAMMA variant showed more hydrophobic interactions with the light chain of S309.All polar interactions are represented in Fig. 2 and detailed interactions are listed in the Supplementary Table S2.
Evaluation of the binding a nity of Sotrovimab (S309) with 9 SARS-CoV-2 Variants by comparing their binding a nity with the Wuhan reference.The thermodynamic stability of the generated complexes was measured by computational prediction of Gibbs free energy (ΔG) using CSM-AB tool.Gibbs free energy re ects energy differences between coupled and decoupled antibody-antigen protein complex.This difference in energy (ΔG) indicate complex stability where a negative normalized energy ΔG < 0 indicates spontaneous and exergonic reaction and hence more stable complex and more e cient protein-ligand interaction.So the lower the value of (ΔG), the more sable the complex (antibodyantigen).In our model, we found that the neutralizing antibody S309 has the binding a nity of -8.26 Kcal/mol with SARS-CoV and − 7.13.26Kcal/mol with SARS-CoV-2 indicating a loss in binding a nity.However, comparing SARS-CoV-2 variants to the binding a nity of the rst Wuhan strain showed an improvement of the binding a nity of S309 with variants Alpha, Beta, Gamma and Kappa.This improvement in a nity, when compared to the interaction pro le, can be related to the increased number of polar and hydrophobic interactions and more similar interactions' pro le to SARS-CoV than to Wuhan strain.In the other hand, Delta variant showed a substantial decrease in the binding a nity as it exhibits only one polar interaction.In the case of Omicron subvariants, they all share similar interaction pro le; however, they exhibit different binding a nities.Although they show signi cant decrease in binding a nity compared to Wuhan strain, they can be clustered in to two groups; those with G339H mutation (BA.2.75, XBB, XBB.1) and those with G339D mutation (BA.1, BA.2, BA.4/5, BQ.1, BA.2.12.1) (Fig. 3, Supplementary Table S3 and Supplementary Fig. S1).The data showed that histidine residue on position 339 slightly enhance the binding a nity compared to aspartic acid substitution.This residue is located in the middle of the interaction loop and hence plays a great role on the complex stability and binding a nity.In addition, our results are aligned with reported effect of the G339D mutation and its role in escaping antibody neutralization [22][23][24].Furthermore, to test the impact of G339 mutation, we analyze the effect of reverse mutagenesis.We used the generated models and in silico tools to test the effect of reverse mutation at residue 339 on the complex stability of subvariants (BA2.75,XBB, and XBB.1).They have n aspartic acid on the position 339.By reversing it back to either Glycine or Histidine (G339 or H339), we calculated the effect in the form of ΔG value.The results showed an increase of the stability of the SARS-CoV-2/S309 complex and hence enhanced binding a nity with the Glycine residue.However, reverse mutagenesis to Histidine has none to a very low effect except for the variant BA.2.12.1 where there was a slight increase on the binding a nity Table 1.
Evaluation of Sotrovimab (S309) binding a nity to experimentally tested and some hypothetic SARS-CoV-2 variants.
The effect of several amino acids exchange in the NAb Sotrovimab epitope have been tested experimentally by ELISA and/or pseudovirus neutralization assays.These mutations showed to be resistance to inhibition by S309 leading to an antibody escape.These key residues include R346S and P337L, G339D, N440K and S371L [23,25].Here we used our developed method to evaluate this effect computationally.By generating models with the new reported mutations and CSM-AB tool we predict the effect of the reported mutations on the binding a nity of the complex and hence on the S309 antibody neutralization effect.Interestingly, our computational results came comparable with the experimentally reported effect of these residues mutations on the S309 monoclonal antibody escape.Additionally, we predicted a possible effect of hypothetical mutations on some of the proteoglycan epitopes Table 2.

Discussion
Antibody-based therapies has proven its e ciency against SARS-CoV-2 virus and appears to be the most promising approach to control COVID-19 pandemic.A number of neutralizing monoclonal antibodies used in clinical setting showed very good results particularly in stopping the disease progression [26,27].However, constant emergence of new virus' variants hinder the potency of available anti-SARS-CoV-2 antibodies and urged the continuous development of improved more effective neutralizing antibody.In this paper, we describe an in-silico method we 've developed to predict a possible effect of newly emerged mutations on the e cacy of available neutralizing anti-SARS-CoV-2 antibody.We used Mab Sotrovimab (S309) as an example.Sotrovimab recognizes a proteoglycan epitope embedded in a structural loop located on the outer side the SARS-CoV-2 protein and covering residues 333-441.This speci c epitope location permits the binding to both con gurations RBD up and down without affecting the binding to ACE2.Indeed, this epitope does not overlap with the ACE2 binding site.However, several newly emerged RBD mutations were reported to have an impact on the neutralization effect of S309.To further explore this, we developed this computational method to evaluate and compare the neutralization potential of the antibody Sotrovimab against different SARS-CoV-2 variants and possible new emerging mutations as outlined in Fig. 1.
Using bioinformatics tools, we developed a spike models for several SARS-CoV-2 new variants and evaluated the effect of several emerged mutations on the interaction with the neutralizing monoclonal antibody Sotrovimab (S309) used for the treatment of mild-to-moderate coronavirus disease.In addition, by applying this method, we foresee the effect of some predicted or not yet been observed mutations.Interestingly, the predicted signi cantly decreased computational neutralization values of Mab Sotrovimab (from 10 to 50%) for some new omicron variants are con rmed by the newly published clinical results indicating reduction in effectiveness against these same Omicron new variants and possible immune evasion [28][29][30][31][32]. Early on, Sotrovimab has been clinically considered to be one of the most effective monoclonal antibodies against all SARS-CoV-2 variants [7].However, this statement has proven wrong as recent convergent evolution of Omicron and its sub variants has led to a new set of spike mutations within the Sotrovimab epitope and consequently the new sub variants became more and more resistant [33].several mutations were identi ed to be critical and others are yet to be investigated.For example, mutation of the nonpolar glycine 339 located in the mid of the antibody epitope to the acidic charged aspartic acid (G339D), showed to have a remarkable impact on the binding a nity of Omicron's subvariants [22,34] with predicted neutralization power reduction of respectively 30% for BA.1, 45% for BA.4,BA.5 BQ.1, 50% for BA.2.12.1,BA.2 and 60% for BF.7, BQ.1.1.We reported similar effect in our proposed computational method and we showed that the impact was less intense with the G339H mutation (Table 1 and Supplementary Table S3).However, the combination of multiple mutations in the Omicron's sub variants has more profound effect on the binding a nity indicating increased antibody resistance.This effect was clearly detected in the possible next dominant new subvariants BM.1.1.1,BA.2.3.20 and CH.1.1 (Orthrus) [35] and Table 2. Furthermore, and to test our method, we examined some experimentally evaluated mutations in residues P337, R346, G339, and S371 that are located in the S309 epitope and once more our computational method was compatible with the experimental results (Table 2).This reduced susceptibility of Sotrovimab with P337, R346 and other mutations has been experimentally recognized [13,23,25].Considering the clinical observations of the e ciency of Sotrovimab in neutralization of SARS-CoV, SARS-CoV-2 variants and Omicron sub variants, a 50% reduction of the binding a nity as compared to the reference model might be taken as cut off for considering if a monoclonal antibody will neutralize a new variant using the method described in this paper.The comparison of the predicted values of the evaluation of neutralizing power with a larger number of clinical observations about the e ciency of a neutralizing Mab would help re ne this theoretical cut off value.
In conclusion, this in silico method gives good insights on possible antibody-escape following emergence of new SARS-CoV-2 mutations and helps in evaluating the usefulness of existing neutralizing antibodies in ghting new emerging variants and sub variants.This method is straightforward, rapid and applicable ahead of obtaining statistically signi cant clinical observations.In addition, this method highlights the advantages of computational approaches in viral surveillance and for the development of novel Mab therapies.Table 1 Gibbs free energy (ΔG) analysis of the effect of D339 reverse mutation on the binding a nity of Omicron's SARS-Cov-2 subvariants with S309 NAb.▲: Increase, ▼: Decrease.

Method
Method's outline.This work a computational method to evaluate the effect of different SARS-CoV-2 mutations on the stability of the complex and binding a nity with available neutralizing antibodies.As a working pattern, we developed a reference complex model between the neutralizing antibody Sotrovimab (S309) and the SARS-CoV-2 Wuhan strain.We evaluated the others variants and subvariants based on the differences of their speci c molecular interactions and binding energy (ΔG) with Sotrovimab. Figure 1   were extracted from the generated models and the complexes with S309 was constructed by molecular replacement.
The reference crystalized RBD chain M of PBD 7YAD was replaced with the modeled RBD.The complex was saved and energy minimized.Energy minimization was done in one-steps using Swiss-pdb Viewer 4.1.0(http://www.expasy.org/spdbv/)[43].This was applied to all the generated models.
Interactions and complex binding a nity analysis.The interactions between the RBD of spike of SARS-CoV, SARS-CoV-2 variants and Omicron subvariants with neutralizing antibody S309 were analyzed based on polar and hydrophobic interaction using the LigPlot + software [44].The stability and a nity were assisted based on thermodynamic measure of the formed complex energy, Gibbs free energy, (ΔG), this was performed using an antibodyantigen binding a nity online tool, CSM-AB (https://biosig.lab.uq.edu.au/csm_ab/prediction)[45].Binding a nity percentage was calculated in reference to Wuhan /S309 complex binding a nity.
Testing the generated method by analyzing newly reported Omicron and some experimentally tested mutations.Several reports have been discussing the neutralization effect of NAbs and possible antibody escape of some new Omicrons subvariants [32, 36-38, 40, 41].Here we used our developed method to evaluate the binding a nity of several of these new subvariants including AY. mutations showed to be resistance to inhibition by S309 leading to an antibody escape.These key residues include R346 and P337, G339, N440 and S371 [23,25].Therefore, we apply our method to computationally test the effect of some mutations on these residues.As we already generated parent's RBD sequences, newly emerged mutations were introduced, new models and complexes were built and the mutation's effect on binding energy with the NAb was predicted by recalculating complex's Gibbs free energy (ΔG) in reference to parent's complex and Wuhan binding a nity.

Declarations
The authors declare that they have no known competing nancial interests or personal relationships that could have appeared to in uence the work reported in this paper.
outlined the methods for the anti SARS-COV-2 antibody neutralization potential of S309.Models and complexes constructionBuilding the NAb/SARS-CoV-2 RBD reference model.We used model ID: 7YAD downloaded from RCSB Protein Data Bank (RCSB PDB) (https://www.rcsb.org/structure/7YAD), to generate our reference model representing the interaction of Sotrovimab variable domain (Fv) with Omicron SARS-CoV-2 spike RBD.The selection criteria of 7YAD model[15] are the generation of the 3D structure by electron microscopy, the high resolution of 2.66Å and the relatively good validation report.In addition, it represent the interaction with SARS-COV-2 RBD in the open state.The 7YAD structure shows two RBD-S309 units (S309-RBD-RBD-S309).Upon downloading the structure, only one unit was selected representing one S309 Fv domain binding to one spike RBD (Chains A, B, M).The complex was extracted, cleaned from any heteroatoms and used as reference model to generate the different variant complexes by RBD replacement.SARS-CoV and SARS-CoV-2 variants sequence retrieval, modi cations and modeling.The amino acid sequence of the extracellular domain of SARS-CoV and SARS-CoV-2 spike protein were acquired from the National Center for Biological Information (NCBI) protein ID: YP_009825051.1 and ID: YP_009724390.1 respectively.SARS-CoV-2 variants-speci c mutations were introduced to the collected sequence to generate the different variant sequences based on published mutations on databases such as CoVariants (https://covariants.

Figure 1 a
Figure 1

Table 2
Prediction of the effect of reported new SARS-CoV-2 subvariants and some experimentally tested spike mutations on the binding a nity with S309.▲: Increase, ▼: Decrease, ▬: No change.