Tumor Associated Antigens Harbor Readily Defined and Universally Immunogenic Regions Relevant For Cancer Immunotherapy

Abstract

Recent advances in cancer immunology, highlighted by immune checkpoint inhibitors, have demonstrated that immunotherapy is a viable option in the oncologist’s armamentarium. Despite these advances, many patients are nonresponders. Preliminary studies have suggested that non-responders lack a de-novo anti-tumor antigen immune response that can be unmasked by checkpoint blockade; thus, strategies to induce anti-tumor immune responses are needed. We hypothesized that many tumor associated antigens (Ag) are readily susceptible to immune attack, but only in the context of identifying the tumor antigen epitopes that can reliably initiate an immune response, regardless of individual patient human leukocyte antigen (HLA) haplotype restrictions. We further hypothesized that epitope prediction strategies which seek to identify pan- or highly promiscuous-HLA binding epitopes would reduce the number of potential candidates and be more likely to accurately identify high-priority tumor Ag epitopes. Utilizing known HLA-serotype frequencies and setting a threshold of ninety percent of population coverage, regardless of race or ethnicity, twenty-nine different HLA-DRB1 haplotypes were chosen for antigen prediction utilizing the open source epitope prediction algorithm netMHCIIpan. Predictions were also performed for HLA-A serotypes utilizing the open source algorithm netMHCpan. Predicted epitopes were synthesized in the form of synthetic long peptides and tested in immune system sensitization assays involving unfractionated peripheral blood mononuclear cells (PBMC). Briefly, PBMC were subjected to a two-step culture, first synchronizing their exposure to the long peptides with aggressive surrogate activation of innate immunity, followed by IL-7-modulated T-cell hyperexpansion. Predictions resulted in identification of highly promiscuous-HLA binding epitopes. Unexpectedly, these epitopes clustered together forming high priority regions: unique “hot spots” with high densities of promiscuous HLA-binding epitopes from the widely expressed oncoproteins MUC1, HER2/neu and CMV-pp65 (p<0.0001, for predicted HLA-DRB1 binding affinities, compared to non-hot spot regions). Added synthetic long peptides (>20aa) derived from “hot spot” regions of MUC1, HER2/neu, and CMVpp65 reliably produced selective and sustained expansion of both CD4+ and CD8+ peptide-specific, interferon-γ (IFNγ)-producing Tcells when synchronized with step 2 exposure to exogenous IL-7 (p<0.0001 and p=0.0048, for CD4+ and CD8+ Ag-specific T-cells, respectively, compared to T-cells directed against peptides from non-hot spot regions). “Hot spot” peptide Ag-specific T-cells preferentially recognized endogenous tumor derived MUC1, either in MUC1 expressing tumor cell killing assays (p=0.038, compared to non-peptide Ag-specific T-cells) or as MUC1 tumor lysate when pulsed onto restimulatory PBMC (p=0.022 and 0.025, for CD4+ and CD8+ T-cells, respectively, compared to T-cells directed against peptides from non-hot spot regions). This mechanistically rational antigen selection sequence, effective even for unvaccinated donors, regardless of HLA-haplotype, enables rapid identification of tumor protein regions relevant for cancer immunology, including adoptive immunotherapy, vaccines, and even identification of tumor neo-antigens unique to each patient.