Systematic Identification of Personal Mutated Tumor-Specific Neoantigens in CLL

Abstract 954

A major goal of cancer immunotherapy is to target tumor cells with high specificity and minimal toxicity. Tumor neoantigens are attractive targets for immunotherapy as they are generated from cancer-driven gene mutations and are expressed exclusively in tumor cells. Neoantigens have not previously been systematically investigated in human tumors due to technical challenges in identifying them. To address this, and to identify immunogenic epitopes personal to an individual tumor, we sought to exploit results from next-generation DNA sequencing to detect somatic mutations together with algorithms that predict peptide binding to HLA class I. We tested this approach in chronic lymphocytic leukemia (CLL), a B cell malignancy associated with a rich source of mutations (Wang et al, NEJM 2011).

In a published study, we sequenced 91 CLL samples using whole-genome (n=3) or whole-exome (n=88; WES) DNA sequencing. To detect mutations, sequences in each tumor sample were compared with corresponding normal sequences, and the frequency of various mutation classes that lead to amino acid alterations were determined. On a per-tumor level, we detected a median of 17 (range: 2–72) missense mutations; 1 (range: 0–6) splice-site mutation; 1 (range: 0–1) read-through mutation, and 1 (range: 0–5) frameshift mutation. Frameshift mutations were predicted to generate potential neo-open reading frames with a median length of 40 amino acids (range: 0–180).

Only peptides binding HLA molecules have the potential to elicit immune responses. We therefore focused on mutations generating autologous HLA binding peptides. Using a well-validated MHC class I prediction algorithm, pan-NetMHC, we systematically evaluated the binding potential of 9- and 10-mers tiled around each mutation. For 31 of 91 samples with available HLA typing, a median of 22 strongly binding peptides (range: 9–75) was predicted to be generated from a median of 26 (range: 7–87) mutations per sample. Using a competitive HLA binding assay, we experimentally validated the predicted high-binding capacity of 60 of 112 (53.5%) synthesized peptides, generated from 3 patients.

To determine if tumor neoantigens are naturally recognized by cytotoxic T lymphocytes, we focused our analysis on CLL patients who had undergone successful hematopoietic stem cell transplantation (HSCT) followed by repeated vaccinations with irradiated autologous whole CLL cells, as normal donor T cell reconstitution following HSCT can overcome endogenous immune defects of the host and since donor T cells in this setting are already primed against host leukemia cells in vivo. Analysis of one patient identified 31 coding mutations, which generated 47 peptides predicted to bind autologous HLA. Of these, 24 were experimentally confirmed to bind autologous HLA alleles and thereafter were screened for inducing T cell reactivity. Using IFN-g ELISPOT reactivity following 2 ex vivo peptide stimulations, memory CD8+ T cell responses could be detected against a HLA A2-binding peptide derived from mutated FNDC3B (VVMSWAPPV). These cells were preferentially reactive to the mutated (but not wildtype) peptide. Results were confirmed with peptide-reactive T cell clones and staining with HLA-A2+/mutated FNDC3B peptide-specific tetramers. Mutated FNDC3B-reactive T cells also recognized HLA-A2+ expressing cells transfected with a minigene encompassing 300 base pairs surrounding the FNDC3B mutation, consistent with endogenous processing and presentation of the peptide. Ongoing studies focus on exploring the tumor cytolytic potential of mutation-specific T cells, and the kinetics of developing neoantigen-specific T cell reactivity in relation to immune reconstitution.

Our studies provide proof-of-concept for systematic identification of tumor neoantigens by integrating information from next-generation sequencing of tumors with predictive HLA binding algorithms. These studies now set the stage for the implementation of clinical studies to explore the therapeutic potential of targeting tumor neoantigens.