Investigating Substitutions in Antibody-Antigen Complexes Using Molecular Dynamics: A Case Study with Broad-spectrum, Influenza A Antibodies.

William D Lees,Adrian J Shepherd,Lenka Stejskal,David S Moss

doi:10.3389/fimmu.2017.00143

Abstract

In studying the binding of host antibodies to the surface antigens of pathogens, the structural and functional characterization of antibody–antigen complexes by X-ray crystallography and binding assay is important. However, the characterization requires experiments that are typically time consuming and expensive: thus, many antibody–antigen complexes are under-characterized. For vaccine development and disease surveillance, it is often vital to assess the impact of amino acid substitutions on antibody binding. For example, are there antibody substitutions capable of improving binding without a loss of breadth, or antigen substitutions that lead to antigenic escape? The questions cannot be answered reliably from sequence variation alone, exhaustive substitution assays are usually impractical, and alanine scans provide at best an incomplete identification of the critical residue–residue interactions. Here, we show that, given an initial structure of an antibody bound to an antigen, molecular dynamics simulations using the energy method molecular mechanics with Generalized Born surface area (MM/GBSA) can model the impact of single amino acid substitutions on antibody–antigen binding energy. We apply the technique to three broad-spectrum antibodies to influenza A hemagglutinin and examine both previously characterized and novel variant strains observed in the human population that may give rise to antigenic escape. We find that in some cases the impact of a substitution is local, while in others it causes a reorientation of the antibody with wide-ranging impact on residue–residue interactions: this explains, in part, why the change in chemical properties of a residue can be, on its own, a poor predictor of overall change in binding energy. Our estimates are in good agreement with experimental results—indeed, they approximate the degree of agreement between different experimental techniques. Simulations were performed on commodity computer hardware; hence, this approach has the potential to be widely adopted by those undertaking infectious disease research. Novel aspects of this research include the application of MM/GBSA to investigate binding between broadly binding antibodies and a viral glycoprotein; the development of an approach for visualizing substrate–ligand interactions; and the use of experimental assay data to rescale our predictions, allowing us to make inferences about absolute, as well as relative, changes in binding energy.

Highlights

The identification and characterization of broadly neutralizing antibodies to highly mutable pathogens such as HIV [1] and influenza [2, 3] has important consequences both for treatment and for vaccine development, but the structural understanding of antibody/antigen interactions is far from complete
We demonstrate that molecular dynamics (MD) and other computational techniques, running on widely available computer hardware, can be used to complement experimental results in order to gain a deeper insight into molecular interactions and can predict the structural and chemical impact of substitutions
We find general correspondence between MM/ GBSA and ANCHOR in identifying such contacts (Figure 6), some highly favorable contacts identified by molecular mechanics with Generalized Born surface area (MM/GBSA) are not identified by ANCHOR: in particular, CR6261 CDR3 Y98 is predicted by ANCHOR to be have a lower contact free energy than CDR2 F54 (−2.0 and −3.0 kcal/mol, respectively) in experimental assays Y98A reduced binding significantly more than F54A (ΔΔGbind −5.1 and −4.0 kcal/mol)

Summary

Introduction

The identification and characterization of broadly neutralizing antibodies (bnAbs) to highly mutable pathogens such as HIV [1] and influenza [2, 3] has important consequences both for treatment and for vaccine development, but the structural understanding of antibody/antigen interactions is far from complete. Key residues are typically identified by means of alanine substitution assays, which determine ΔΔGbind, the change in the Gibbs free energy of binding caused by the substitution. These assays can be expensive and time consuming to conduct, if the required antibody is not to hand, and can only provide a limited understanding of the contribution that individual residues make to the interaction. There are known examples of substitutions outside such contacts that significantly affect ΔΔGbind, and examples of contact residues that are not identified by alanine substitution [5].

Methods

Results

Conclusion