Mapping the HLA Ligandome Landscape of Chronic Myeloid Leukemia (CML)—Towards Peptide Based Immunotherapy

Despite the success of targeted therapy with tyrosine kinase inhibitors (TKIs), chronic myeloid leukemia (CML) remains largely incurable. Immunotherapy, and in particular multi-peptide vaccination, may be a promising approach to eliminate residual CML cells. As of now, a multitude of potential vaccine targets have been proposed based on reverse immunology and functional genomic approaches focusing either on BCR-ABL junction peptides, which represent CML-specific neo-antigens, or on aberrantly expressed self-proteins such as WT-1, PR and hTERT. However, the results of clinical studies employing such antigens have so far not been encouraging. This might in part be due to the inherent limitations of the above mentioned approaches: evidence of natural presentation of the predicted epitopes is lacking and the correlation of transcript abundance and HLA restricted presentation of the corresponding gene product has been shown to be skewed. Modern mass spectrometry, on the other hand, enables the comprehensive analysis of the entirety of naturally presented HLA ligands on tissues of interest, termed the HLA ligandome.

Here we implemented this direct approach and comparatively mapped the HLA ligandome landscape of 16 primary CML samples and 40 healthy volunteer (HV) controls (30 blood and 10 bone marrow samples). We identified more than 30,000 different naturally presented HLA class I ligands representing ~10,000 source proteins. Regression analysis suggests source protein identifications on CML (4,337 different proteins) to attain >95% of maximum achievable coverage with the implemented analytical setup. Based on this extensive dataset, we investigated the HLA restricted presentation of established CML-associated/specific antigens and applied a novel approach defining tumor-associated antigens strictly based on exclusive and frequent representation in CML ligandomes. Strikingly, we found the vast majority of previously described antigens including wild-type BCR protein (6% CML, 5% HV), Myeloperoxidase (56% CML, 15% HV) and Proteinase 3 (38% CML, 11% HV) to be (also) represented on normal PBMC or BMNC. No evidence of naturally presented BCR-ABL junction peptides was found. However, we identified a panel of 7 LiTAAs (ligandome-derived tumor-associated antigens) represented by 16 different HLA ligands, showing CML-exclusive representation in ≥25% of CML patient ligandomes.

As CD4⁺ T cells mediate important indirect and direct effects in anti-tumor immunity, we further applied our approach to HLA class II ligandomes of 15 CML patients and 18 HV (13 blood and 5 bone marrow samples), identifying more than 9,000 different naturally presented HLA class II ligands (1,900 source proteins). Applying the same antigen-ranking strategy as described for HLA class I, we identified 7 additional HLA class II LiTAAs represented by 50 corresponding LiTAPs (ligandome-derived tumor-associated peptides). Overlap analysis of CML-exclusive source proteins revealed 6 proteins to be represented both in HLA class I and II ligandomes. Notably, for Galectin-1, which shows CML-exclusive representation in 19% of HLA class I and 13% of HLA class II ligandomes, one of the HLA class II ligands was found to contain a complete, embedded HLA class I peptide. Such naturally presented embedded HLA ligands might present optimal vaccine candidates that are recognized by both, CD4⁺ and CD8⁺ T cells.

Functional analysis of the here defined HLA class I and II LiTAPs with regard to induction of T cell responses is presently ongoing and serves to validate them as prime targets for the development of an off-the-shelf peptide vaccination in CML patients.