HLA molecules presenting peptides derived from tumor-associated self-antigens (self-TAA) are attractive targets for T-cell-based immunotherapy of cancer. However, detection of such epitopes is hampered by self-tolerance and limitations in the sensitivity of mass spectrometry. Here, we used T cells from HLA-A2-negative donors as tools to detect HLA-A2-bound peptides from two leukemiaassociated differentiation antigens; CD20 and the previously undescribed cancer target myeloperoxidase. A high-throughput platform for epitope discovery was designed using dendritic cells cotransfected with full-length transcripts of self-TAA and HLA-A2 to allow presentation of all naturally processed peptides from a predefined self-protein on foreign HLA. Antigen-reactive T cells were directly detected using panels of color-coded peptide-HLA multimers containing epitopes predicted by a computer algorithm. Strikingly, cytotoxic T cells were generated against 37 out of 50 peptides predicted to bind HLA-A2. Among these, 36 epitopes were previously undescribed. The allorestricted T cells were exquisitely peptide-and HLA-specific and responded strongly to HLA-A2-positive leukemic cells with endogenous expression of CD20 or myeloperoxidase. These results indicate that the repertoire of self-peptides presented on HLA class I has been underestimated and that a wealth of self-TAA can be targeted by T cells when using nontolerized T-cell repertoires.C ytotoxic T cells (CTLs) selectively kill target cells that express defined peptides in complex with MHC class I molecules on the cell surface. Most tumor-associated antigens (TAA) are wild-type self-proteins, and T cells that recognize peptides from these antigens with high affinity are deleted during thymic development. Thus, the utility of T cells for detection of self-peptides presented on self-HLA is limited by tolerance. This may be one reason why the number of epitopes identified from TAA after 2 decades of intense research amounts to less than 600 (1). The ability to rapidly identify new CTL epitopes would likely facilitate the development of effective immunotherapeutic strategies against cancer.MHC molecules can be isolated from cells and the associated peptides eluted for identification by MS. Ultimately, this approach may provide a description of the entire MHC-bound peptide repertoire: the immunopeptidome (2, 3). This is, however, a daunting task, and it is unclear whether current MS-based protocols already provide the required sensitivity. Indeed, although 100,000-750,000 peptide-MHC class I complexes are expressed for each allelic product on the cell surface (for HLA-A and HLA-B loci) (3, 4), the largest HLA ligandome identified to date contains 14,065 peptides (5). In contrast, the predicted number of different HLA class I ligands would be 352,000 using the well-renowned computer algorithm NetMHCpan, considering that on average, 4.4% of all nonamers bind HLA class I (6) and that a cell contains at least 8 × 10 6 distinct nonamers (7). Thus, there is a very large gap between the numb...