Predicting CD4 T-cell epitopes based on antigen cleavage, MHCII presentation, and TCR recognition

Abstract

Accurate predictions of T-cell epitopes would be useful for designing vaccines, immunotherapies for cancer and autoimmune diseases, and improved protein therapies. The humoral immune response involves uptake of antigens by antigen presenting cells (APCs), APC processing and presentation of peptides on MHC class II (pMHCII), and T-cell receptor (TCR) recognition of pMHCII complexes. Most in silico methods predict only peptide-MHCII binding, resulting in significant over-prediction of CD4 T-cell epitopes. We present a method, ITCell, for prediction of T-cell epitopes within an input protein antigen sequence for given MHCII and TCR sequences. The method integrates information about three stages of the immune response pathway: antigen cleavage, MHCII presentation, and TCR recognition. First, antigen cleavage sites are predicted based on the cleavage profiles of cathepsins S, B, and H. Second, for each 12-mer peptide in the antigen sequence we predict whether it will bind to a given MHCII, based on the scores of modeled peptide-MHCII complexes. Third, we predict whether or not any of the top scoring peptide-MHCII complexes can bind to a given TCR, based on the scores of modeled ternary peptide-MHCII-TCR complexes and the distribution of predicted cleavage sites. Our benchmarks consist of epitope predictions generated by this algorithm, checked against 20 peptide-MHCII-TCR crystal structures, as well as epitope predictions for four peptide-MHCII-TCR complexes with known epitopes and TCR sequences but without crystal structures. ITCell successfully identified the correct epitopes as one of the 20 top scoring peptides for 22 of 24 benchmark cases. To validate the method using a clinically relevant application, we utilized five factor VIII-specific TCR sequences from hemophilia A subjects who developed an immune response to factor VIII replacement therapy. The known HLA-DR1-restricted factor VIII epitope was among the six top-scoring factor VIII peptides predicted by ITCall to bind HLA-DR1 and all five TCRs. Our integrative approach is more accurate than current single-stage epitope prediction algorithms applied to the same benchmarks. It is freely available as a web server (http://salilab.org/itcell).

Designing of cytotoxic and helper T cell epitope map provides insights into the highly contagious nature of the pandemic novel coronavirus SARS-CoV-2

“…The ranking was done after sorting from higher to lower immunogenicity score [ 35 ]. For helper T cell epitopes immunogenicity prediction, CD4episcore [ 36 ] and ITcell [ 37 ] were used. CD4episcore was developed using neural networks and combines HLA binding and immunogenicity prediction and outputs a list of immunogenic peptides using a combined score.…”

Peptide Epitope Hot Spots of CD4 T Cell Recognition Within Influenza Hemagglutinin During the Primary Response to Infection

“…It is reasonable to speculate that such selection is multifactorial. Selection of CD4 T cell responses in the host has been shown to reflect at least in part the sequential processes in antigen presentation, including antigen uptake into endosomal compartments of antigen presenting cells (APC), pH-induced unfolding, reduction of disulfide bonds, proteolytic release of antigenic peptide, acquisition of the peptide by host MHC class II molecules and editing by HLA-DM and, finally the export of the peptide:class II complex to the cell surface of the APC (reviewed in [44][45][46][47][48][49][50][51][52][53][54][55][56][57]). Factors that have been implicated in previous studies for selection of dominant epitopes by CD4 T cells include proteolytic processing, three dimensional structure, sensitivity to proteolytic enzymes, or biochemical features of peptide:MHC class II complexes [56,[58][59][60][61][62][63].…”

FVIII Immunogenicity—Bioinformatic Approaches to Evaluate Inhibitor Risk in Non-severe Hemophilia A

“…Hart et al acknowledge that potential confounders remain, limiting the repertoire of FVIII-derived peptides available for MHC presentation (31). Addressing these, Schneidman-Duhovny et al provide a step-change refinement in their in silico pipeline to further improve prediction accuracy (35). Specifically, a three step, "integrative structure-based" algorithm starts with a peptide cleavage prediction to account for the cleavage preferences of natural intracellular proteases, cathepsins B, H, and S (36,37).…”

“…Their data includes a validating peptide series, unrelated to FVIII, but subsequently use FVIII derived peptides as a proof of translational principle, in particular to narrow the field of likely preferred tFVIII-derived binders. Using 5 patient-derived TCR sequences reduced the number of possible 12 mer epitope cores from 2,340 to just six peptides including the correct epitope core (35). Such refinement is hypothesis generating, providing a manageable repertoire of candidate immunogenic peptides with which to work.…”

“…van Haren observes only a minority of the peptides predicted in silico to bind, actually bind to HLA class II and are retrievable in peptide elution experiments (40). The Schneidman cathepsin-cleaving modeling is likely to contribute to this more limited repertoire (35). Competition from the multitude of other intracellular self and non-self peptides vying for presentation position within the HLAII repertoire has not been accounted for, although may be an explanation, in part, for the limited retrievable repertoire.…”