A Mechanistic Model for Predicting Cell Surface Presentation of Competing Peptides by MHC Class I Molecules

A Mechanistic Model for Predicting Cell Surface Presentation of Competing Peptides by MHC Class I Molecules

Denise S. M. Boulanger (1), Ruth C. Eccleston (2,3), Andrew Phillips (4), Peter V. Coveney (2,3,)
Tim J. Elliott (1) and Neil Dalchau (4)

1 Centre for Cancer Immunology and Institute for Life Sciences, Faculty of Medicine, University of Southampton,Southampton, United Kingdom,
2 Centre for Computational Science, Department of Chemistry, University College London, London, United Kingdom,
3 CoMPLEX, University College London, London, United Kingdom,
4 Microsoft Research,Cambridge, United Kingdom
Frontiers in Immunology (2018) doi.org/10.3389/fimmu.2018.01538

The processing and presentation of cellular antigens to cytotoxic T lymphocytes (CTL) lies at the heart of protective immune responses to infections and cancer, be they natural or induced by vaccination.
In a previous collaboration, the Elliott group at the Centre for Cancer Immunology at the University of Southampton, and the Phillips/Dalchau group at Microsoft Research Cambridge (both Associate Partners of CompBioMed) have built and experimentally validated a computational model that describes the antigen processing and presentation pathway. The team has subsequently hooked up with the Coveney group at UCL to develop the model in a manner that can be used to predict the likelihood of a viral or cancer antigen being presented to CTL. This capability is a very valuable addition to the armoury of predictive tools that underpin vaccine development for infection and cancer.

CTL recognise peptide fragments of proteins (selected for presentation from a vast pool of candidates generated from the natural turnover of cellular proteins, including viral proteins in infected cells) and presented at the surface of cells bound to MHC I molecules. Inside the cell therefore, the function of MHC I is to select peptides for presentation, and this is assisted by by the peptide-loading complex (PLC) in the ER which comprises newly assembled MHCI:beta2-microglobulin (b2m), tapasin, ERp57, calreticulin and TAP (the transporter associated with antigen presentation).

The paper, which appears in Frontiers in Immunology this month (July 2018) models the contribution to antigen presentation of a) the intracellular abundance of the source-protein, b) competition between peptides, c) the binding affinity of individual peptides for MHC I and d) the action of intracellular cofactors known to assist peptide loading of MHC I. Computationally intense model-fitting to experimental data delivered a predictive, mechanism-based algorithm as well as a mathematical abstraction that is well suited to transcriptomic big-data.

This is important because currently there is no quantitative model to predict the relative abundance of different peptide:MHC I complexes at the surface of antigen presenting cells, virus infected cells or cancer cells – and the Elliott group have previously shown that it is the abundance of specific peptide:MHC I that determines CTL immunodominance to both vaccines and cancer.

The model will therefore augment correlative algorithms designed to predict the presentation of specific peptides, and which are currently the only instrument available to vaccinologists resolved to provoking CTL responses to viral epitopes, and more recently to epitopes generated from the cancer mutanome – the thousands of coding mutations that are found in individual tumours that could be seen by patient CTL. Because the Elliott/Dalchau model incorporates mechanistic knowledge of the antigen processing pathway, it could be used to predict the outcome (levels of presentation) under different physiological conditions – such as cytokine milieu. In fact the study accurately predicted the antigen presentation outcome when antigen presenting cells were cultured in the presence of interferon-gamma: a signature cytokine of cancers that respond well to immunotherapy.