key: cord-0268292-7f73ioj0
authors: Beach, Sierra S.; Hull, McKenna; Ytreberg, F. Marty; Patel, Jagdish Suresh; Miura, Tanya A.
title: Molecular modeling predicts novel antibody escape mutations in respiratory syncytial virus fusion glycoprotein
date: 2022-02-26
journal: bioRxiv
DOI: 10.1101/2022.02.25.482063
sha: de096d1469c6a2b6f5fe05a1a48876a41db88280
doc_id: 268292
cord_uid: 7f73ioj0

Monoclonal antibodies are increasingly used for the prevention and/or treatment of viral infections. One caveat of their use is the ability of viruses to evolve resistance to antibody binding and neutralization. Computational strategies to predict which mutations will result in antibody resistance would be invaluable because current methods for identifying potential escape mutations are labor intensive and system-biased. Respiratory syncytial virus is an important pathogen for which monoclonal antibodies against the fusion (F) protein are used to prevent severe disease in high-risk infants. In this study, we used an approach that combines molecular dynamics simulations with FoldX to estimate changes in free energy in F protein folding and binding to the motavizumab antibody upon each possible amino acid change. We systematically selected 8 predicted escape mutations and tested them in an infectious clone. Consistent with our F protein stability predictions, replication-effective viruses were observed for each selected mutation. Six of the eight variants showed increased resistance to neutralization by motavizumab. Flow cytometry was used to validate the estimated (model-predicted) effects on antibody binding to F. Using surface plasmon resonance, we determined that changes in the on-rate of motavizumab binding were responsible for the reduced affinity for two novel escape mutations. Our study empirically validates the accuracy of our molecular modeling approach and emphasizes the role of biophysical protein modeling in predicting viral resistance to antibody-based therapeutics that can be used to monitor the emergence of resistant viruses and to design improved therapeutic antibodies. Importance Respiratory syncytial virus (RSV) causes severe disease in young infants, particularly those with heart or lung diseases or born prematurely. As no vaccine is currently available, monoclonal antibodies are used to prevent severe RSV disease in high-risk infants. While it is known that RSV evolves to avoid recognition by antibodies, screening tools that can predict which changes to the virus will lead to antibody resistance are greatly needed.

Monoclonal antibodies are increasingly used for the prevention and/or treatment of viral 25 infections. One caveat of their use is the ability of viruses to evolve resistance to antibody 26 binding and neutralization. Computational strategies to predict which mutations will result in 27 antibody resistance would be invaluable because current methods for identifying potential escape 28 mutations are labor intensive and system-biased. Respiratory syncytial virus is an important 29 pathogen for which monoclonal antibodies against the fusion (F) protein are used to prevent 30 severe disease in high-risk infants. In this study, we used an approach that combines molecular 31 dynamics simulations with FoldX to estimate changes in free energy in F protein folding and 32 binding to the motavizumab antibody upon each possible amino acid change. We systematically 33 selected 8 predicted escape mutations and tested them in an infectious clone. Consistent with our 34 F protein stability predictions, replication-effective viruses were observed for each selected 35 mutation. Six of the eight variants showed increased resistance to neutralization by 36 motavizumab. Flow cytometry was used to validate the estimated (model-predicted) effects on 37 antibody binding to F. Using surface plasmon resonance, we determined that changes in the on-38 rate of motavizumab binding were responsible for the reduced affinity for two novel escape 39 mutations. Our study empirically validates the accuracy of our molecular modeling approach and 40 emphasizes the role of biophysical protein modeling in predicting viral resistance to antibody-41 based therapeutics that can be used to monitor the emergence of resistant viruses and to design 42 improved therapeutic antibodies. The use of monoclonal antibodies (mAb) for the treatment or prevention of viral 71 infections has accelerated over the past few years. mAb are a powerful tool against viral 72 infections because of the ability to target specific epitopes to neutralize viral pathogens. The first 73 mAb approved by the FDA for the prophylaxis of viral infection was Synagis® (palivizumab), 74

which is used to prevent respiratory syncytial virus (RSV) infection in high-risk infants (1) (2) (3) . 75

One of the most notable recent uses of mAb has been the emergency use authorization of three 76 mAb treatments for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) the 77 causative agent for coronavirus disease 2019 (COVID-19) (4-7). The FDA has also approved the 78 mAb cocktail, Inmazeb™, for the treatment of Ebola virus infection (8) and Trogarzo®, a mAb 79 treatment for drug-resistant human immunodeficiency virus (HIV-1) infection (9). There are also 80 mAb therapies in development for other viruses including influenza virus (10) and herpes 81 simplex virus 1 (11, 12). The use of mAb for the prevention and/or treatment of viral infections 82 has become an important strategy for the medical community. 83

One caveat of using mAb against viral infections is the increased selective pressure for 84 antibody escape mutations to occur. The current methodology for monoclonal antibody-resistant 85 mutants (MARMs) discovery involves sequencing samples taken from patients or serially 86 passaging virus with the antibody of interest to select for evolved mutations. This presents 87 challenges in that the escape mutation is already circulating in the population and passaging 88 experiments can be time consuming and are biased in only detecting mutations that can arise 89 over limited replication time in cell culture. Given that it is now common to monitor the genetic 90 variation in a population, an aspirational goal in the fight against infectious disease is the ability 91 to predict when viral evolution is about to outpace the effectiveness of protective mAb. 92 Paramount to this goal is the ability to predict amino acid changes that will significantly disrupt 93 binding of antibodies to viral proteins; a collection of these predicted MARMs would constitute a 94 watch list. An increase in the frequency of MARMs on the watch list in an infected population 95 could signal reduced effectiveness of antibody-based therapies and the possible adaptation to 96 these therapies. 97

Protein biophysical models can be used to predict protein stability and the disruption of 98 interactions with other biomolecules due to mutations. In our previous studies, we used FoldX 99 software combined with molecular dynamics (MD) simulations (MD+FoldX) to estimate folding 100 and binding stabilities of Ebola virus (EBOV) envelope glycoprotein (GP) and mAbs KZ52, 101

Antibody 100 (Ab100), Antibody 114 (Ab114) and 13F6-1-2. Using this approach, we 102 generated a watch list comprised of 127 mutations that were predicted to disrupt binding between 103 GP and four mAbs but would not disrupt the ability of the glycoprotein to fold and form a trimer 104 (13) (14) (15) (16) . Three mutations from our watch list have already been seen in humans or are 105 experimentally known to reduce efficacy of the antibody treatment (17, 18). While these 106 previous studies showed the predictive capability of protein biophysical modeling to help guide 107 empirical science in examining phenotypes in viral evolution, these results were not empirically 108 validated to test the accuracy in predicting MARMs. 109

MARMs for palivizumab have been identified from cell culture, animal models, and 110 clinical samples (19-27). RSV is an important pathogen for infants, elderly, and immune 111 compromised that causes severe lower respiratory infections and is the second leading cause of 112 infant death in the world (2, 3). The only targeted treatment for RSV is the prophylactic mAb 113 palivizumab and additional mAbs are in development, including nirsevimab, which is currently 114

in Phase 3 clinical trials (28-32). Palivizumab and its derivative, motavizumab, target the site II 115 epitope on the fusion glycoprotein (F protein), the protein responsible for the fusion of viral and 116 host cell membranes during viral entry (28, 33-38). F protein is a desirable target for anti-viral 117 treatments given the function of the protein for viral entry, the response to the antigen by the host 118 immune system, and the conservation of antigenic sites including site II among RSV strains (38-119 43). F protein is also advantageous to use to test molecular modeling given the multiple co-120 crystal structures with mAbs (44-47). For this study we used the co-crystal structure of 121 motavizumab-F protein because there is no crystal structure with palivizumab. Motavizumab is 122 also of interest as only one MARM, K272E, has ever been derived from clinical and empirical 123 lab studies (21). 124

In this study we describe the application and empirical validation of our MD+FoldX 125 approach to RSV F protein and motavizumab complex to predict MARMs by analyzing changes 126 in relative binding affinities due to all possible amino acid changes in the F protein. We 127 empirically tested the fitness, mAb neutralization, and binding characteristics of eight predicted 128

MARMs using an infectious clone. Our modeling approach was able to predict the already MD+FoldX approach was used to predict RSV F protein escape mutations that would 136 allow the F protein to fold correctly but disrupt its recognition by motavizumab. We estimated 137 the effect of all possible mutations of RSV F protein for all 448 amino acids and 19 possible 138 substitutions at each site for both F protein monomer folding (ΔΔGFold) and its binding with 139 motavizumab (ΔΔGBind) (Figure 1 and Appendix). Mutations with ΔΔGFold value less than 2 140 kcal/mol are not predicted to affect the stability of F protein and therefore have the potential to 141 arise under selective pressure (15). To identify MARMs and test the accuracy of the MD+FoldX 142 approach, we selected eight mutations with ΔΔGBind values ranging from 0.5 to 5.5 kcal/mol (low 143 to high disruption of binding to motavizumab) and ΔΔGFold values less than 2 kcal/mol ( Figure 1 , 144 To confirm that ΔΔGFold values less than 2 kcal/mol did not disrupt F protein function, we 150 first assessed the ability of the eight variants to replicate. To test the growth kinetics of the 151 variants, site directed mutagenesis was used to create the variants, which were subcloned into a 152 Previously published data determined that the IC50 for K272E was 30.04  611.35 g/mL (21). 172 S275H, L258K, and S275R demonstrated a greater than 5-fold increase of IC50 when compared 173 to WT and L258E and N262D had a greater than 3-fold increase in IC50 ( Figure 3 , Table 2 ). 174 Variants N262Y and N276G had similar neutralization kinetics to WT ( Figure 3 , Table 2 ). In 175 addition to confirming K272E as a motavizumab-resistant mutation, we identified 5 novel 176

MARMs that have not been previously identified by traditional approaches. 177 178

Flow cytometry was used to assess motavizumab/F binding in a high-throughput format. To further evaluate motavizumab binding to MARMs we analyzed binding kinetics of the 206 variants we identified to have the highest increase in IC50 compared to WT: K272E, S275H, and 207 L258K. Surface plasmon resonance (SPR) was used to quantify binding kinetics of purified 208 virions to motavizumab. A decrease in the binding on-rate (ka) was observed for all three 209 variants, notably a >6-fold reduction of binding for S275H ( Figure 5A , Table 3 ). There was no 210 significant increase in the off-rate (kd) of antibody binding to the three variants ( Figure 5B , Table  211 3). The overall equilibrium disassociation constant (KD) for the variants demonstrates a lower 212 affinity between the MARMs and motavizumab when compared to WT ( Figure 5C , Table 3 ). 213

The low KD for WT is expected as once the antibody binds it is unlikely to be disrupted. There 214 was a correlation between GBind and on-rate, as GBind increased the on-rate decreased 215 ( Figure 5D ). No correlation was observed between GBind and off-rate, which is expected as 216 there were no significant differences between the off-rate values ( Figure 5E ). A strong linear 217 correlation was observed between GBind and overall affinity of motavizumab binding, as 218 GBind increased, KD increased ( Figure 5F ). The escape mechanism for all three variants 219 appears to be a decrease in association with motavizumab rather than an increase of 220 disassociation from motavizumab. 221 222

The MD+FoldX approach provides rapid and testable predictions of MARMs. In this 224 study, we applied this modeling approach to identify MARMs of RSV F protein and 225 motavizumab antibody complex. This also enabled us to empirically validate the accuracy of our 226 modeling approach. We selected eight variants that were predicted to allow the F protein to fold 227 correctly but disrupt its ability to bind to motavizumab. All the selected variants were found to for motavizumab (21). This study examined mutations for motavizumab given the availability of 236 the co-crystal structure and revealed novel escape variants. The binding kinetics of K272E have 237 been previously studied using purified F protein and found that binding on rate rather than off 238 rate appeared to be the mechanism of escape (27). In our study we were able to replicate those 239 findings with purified virions as well as establish the same mechanism for escape for two 240 additional MARMs that we identified. Previous studies have also found that antibody association 241 rates correlate well with viral neutralization (37). 

In silico predictions of MARMs using MD+FoldX approach. To predict RSV F protein escape 295 mutations against motavizumab mAb, we applied our approach from previous studies that 296 combines classical molecular dynamics (MD) and FoldX software (MD+FoldX) (15, 16) . To 297 designate a mutation as an escape mutation it requires: 1) disrupt binding to a mAb, and 2) leave 298 the F protein monomer stable thus allowing it to fold and assemble. It is thus necessary to 299 determine how amino acid mutations alter stabilities (ΔΔG values) for F protein monomer folding 300 (ΔΔGFold) and binding to Motavizumab (ΔΔGBind). Therefore, we used our MD+FoldX approach 301 to estimate the folding stability of F protein monomer and F protein trimer/Motavizumab complex 302 binding affinities due to all possible single mutations. 303 Structure preparation. The X-ray crystal structure of RSV F glycoprotein bound to motavizumab 304 was downloaded from Protein Data Bank. (PDB ID:4ZYP) 3D coordinates file was first modified 305 to remove all but F protein trimer and three copies of heavy and light chains of motavizumab 306 bound to each F protein monomer (44). The MODELLER software was then used to alter 307 engineered residues and build the missing residues in all the chains (60). Missing amino acid 308 residues 96 to 137 in F protein monomer represent liberated glycopeptide because of proteolysis 309 by furin like proteases. These residues were ignored in our simulations. 

A joint statement with the Fetus and Newborn Committee. 1999. Palivizumab and 431 respiratory syncytial virus immune globulin intravenous for the prophylaxis of respiratory 432 syncytial virus infection in high risk infants

Respiratory Syncytial Virus: Infection, Detection, 434 and New Options for Prevention and Treatment

Past, Present and Future Approaches to the Prevention and 437 Treatment of Respiratory Syncytial Virus Infection in Children

Antibody cocktail to SARS-CoV-2 spike protein prevents rapid mutational escape seen with 441 individual antibodies

Bamlanivimab plus Etesevimab in Mild or Moderate Covid-19

Early Treatment 451 for Covid-19 with SARS-CoV-2 Neutralizing Antibody Sotrovimab

SARS-CoV-2 Monoclonal Antibodies. COVID-19 Treatment Guidelines

REGN-EB3: First Approval. Drugs 1-4

Ibalizumab: First Global Approval

Safety and efficacy of monoclonal antibody VIS410 in 460 adults with uncomplicated influenza A infection: Results from a randomized, double-blind, 461 phase-2, placebo-controlled study

Development and Validation of an LC-MS-Based Quantification Assay 464 for New Therapeutic Antibodies: Application to a Novel Therapy against Herpes Simplex 465 Virus

A Randomised, Double-Blind Phase II Trial 467 of Topical HDIT101 Versus Placebo in Patients With Chronic Recurrent HSV-1 Infection 468 and Orolabial Lesion. NCT04539483. Clinical trial registration

The FoldX web 470 server: an online force field

Molecular Dynamics Simulation for All

Initiating a watch list for Ebola virus antibody escape mutations

Expanding the watch list for potential 477 Ebola virus antibody escape mutations

Transmission Events and Rapid Decline of Central African Wildlife

Emergence of Ebola Virus Escape Variants in Infected 484

Nonhuman Primates Treated with the MB-003 Antibody Cocktail

In Vivo Selection of Respiratory Syncytial Viruses Resistant 487 to Palivizumab

Respiratory syncytial virus escape mutant derived 489 in vitro resists palivizumab prophylaxis in cotton rats

492 Analysis of Respiratory Syncytial Virus Preclinical and Clinical Variants Resistant to 493 Neutralization by Monoclonal Antibodies Palivizumab and/or Motavizumab

Vitro and In Vivo Fitness of Respiratory

Syncytial Virus Monoclonal Antibody Escape Mutants

Transient opening of trimeric prefusion RSV F proteins. 1

Alternative 576 conformations of a major antigenic site on RSV F

Refining the Balance of Attenuation and 578 Immunogenicity of Respiratory Syncytial Virus by Targeted Codon Deoptimization of 579 Virulence Genes

Structure of RSV Fusion Glycoprotein Trimer Bound to a Prefusion-Specific 583 Neutralizing Antibody

The 589 monoclonal antibody combination REGEN-COV protects against SARS-CoV-2 mutational 590 escape in preclinical and human studies

A Syncytial Virus Associated with 592

Epidemic Disease of the Lower Respiratory Tract in Infants and Young Children

Structure 595 of a Major Antigenic Site on the Respiratory Syncytial Virus Fusion Glycoprotein in 596 Complex with Neutralizing Antibody 101F

Genetic Variability of Human Respiratory Syncytial Virus A Strains Circulating in Ontario: 600 A Novel Genotype with a 72 Nucleotide G Gene Duplication

Reverse genetics systems for contemporary isolates of respiratory syncytial virus 604 enable rapid evaluation of antibody escape mutants

Prevalence and Significance of Substitutions in the Fusion Protein of Respiratory Syncytial 608 Virus Resulting in Neutralization Escape From Antibody MEDI8897

Assessment of software 611 methods for estimating protein-protein relative binding affinities

616 Reduced sensitivity of SARS-CoV-2 variant Delta to antibody neutralization. 7871

An infectious SARS-CoV-2 B.1.1.529 Omicron virus 620 escapes neutralization by therapeutic monoclonal antibodies

Key Substitutions in the Spike Protein of SARS-CoV-2 Variants 623

Can Predict Resistance to Monoclonal Antibodies, but Other Substitutions Can Modify the 624 Effects

Comparative Protein Structure Modeling Using MODELLER

628 Motional timescale predictions by molecular dynamics simulations: Case study using proline 629 and hydroxyproline sidechain dynamics

632 GROMACS: Fast, flexible, and free

Generation of Bovine Respiratory Syncytial 634

Virus (BRSV) from cDNA: BRSV NS2 Is Not Essential for Virus Replication in Tissue 635

A simplified plaque assay for respiratory syncytial virus-638 direct visualization of plaques without immunostaining

A 641 flow cytometry-based assay to assess RSV-specific neutralizing antibody is reproducible, 642 efficient and accurate

645 Structural characterization of respiratory syncytial virus fusion inhibitor escape mutants: 646 homology model of the F protein and a syncytium formation assay