HLA Ligandome Analysis of Different Hematological Malignancies Identifies a Small Panel of "Pan-Leukemia"-Associated Antigens

Effective antigen-specific T-cell-based cancer immunotherapy requires exact knowledge of tumor-associated epitopes that can act as rejection antigens. While the current paradigm views mutation-derived neoantigens as the most promising targets, we have recently demonstrated that tumor-specific T-cell responses target panels of non-mutated tumor-associated self antigens in patients with hematological malignancies. Using the approach of direct HLA ligandome analysis by mass spectrometry, we were able to identify and characterize multiple immunogenic and naturally presented tumor-associated antigens for chronic lymphocytic leukemia (CLL, Kowalewski et. al., PNAS 2015), acute myeloid leukemia (AML, Berlin/Kowalewski et. al., Leukemia 2014), multiple myeloma (MM, Walz/Stickel et. al., Blood 2015) and chronic myeloid leukemia (CML, unpublished data).

In this project we performed a comprehensive meta-analysis of our HLA ligandome data from different hematological malignancies (HM) to screen for the existence of "pan-leukemia" antigens for the broad application in T-cell based immunotherapy approaches in hematological malignancies.

In a first step we performed unsupervised cluster analyses to identify similarities and differences in the HLA ligandome landscape of HM. To avoid skewed clustering due to HLA types of the samples, these analyses were performed specifically for the most common HLA allotypes in our datasets (A*02 (n=46 HM), A*03 (n=28 HM)). Distinct clustering was shown for the different entities (CLL, MM, CML, AML) as well as for the lymphoid versus myeloid malignancies on the HLA ligandome level. To identify leukemia-exclusive HLA ligands we compared the HLA ligandomes of CLL (HLA class I, n=35; HLA class II, n=30), AML (HLA class I, n=19; HLA class II, n=20), MM (HLA class I, n=15; HLA class II n=12) and CML (HLA class I, n=16; HLA class II n=15) with our normal tissue database including 153 HLA class I and 82 HLA class II ligandomes of various normal tissues (including normal blood, bone marrow and spleen). Cluster analysis of the leukemia-exclusive antigens showed identical clustering of the different entities and lymphoid/myeloid malignancies as shown before for the whole HLA ligandome and the respective source proteins. Overlap analysis revealed only 0.6% (16/2,716) and 0.3% (10/3,141) of the identified leukemia-exclusive HLA class I and class II antigens, respectively, to be represented across all analyzed hematological malignancies. These "pan-leukemia" antigens (n=26) include candidate antigens associated with T-cell activation (HSH2D), lymphoid development (IL2RF) and oncogenesis (LYN protooncogene, RAB5A). However, none of these "pan-leukemia" antigens shows frequent representation (>20%) across all 4 entities (CLL, AML, MM, CML). Furthermore, none of the "pan-leukemia" source proteins yielded corresponding peptides represented in all entities. To identify "pan-leukemia" HLA ligands, overlap analyses were performed in an allotype-specific fashion for the most frequent HLA allotypes (HLA-A*01, -A*02, -A*03, -A*24, -B*07, -B*08, -B*18) in our cohort. 0% (0/92) of HLA-A*01-, 1.6% (12/744) of HLA-A*02-, 1.4% (8/561) of HLA-A*03-, 0% (0/331) of HLA-A*24-, 0.1% (1/830) of HLA-B*07-, 0% (0/472) of HLA-B*08- and 0.8% (5/600) of the HLA-B*18-restricted peptides showed representation in all four entities. Out of these 26 "pan-leukemia" HLA ligands, only two (1 HLA-A*02-, 1 HLA-A*03-restricted peptide) showed frequent representation (>20%) in all entities. These peptides represent "pan-leukemia" targets that might be used for immunotherapeutic approaches in patients expressing the respective HLA allotype.

Taken together, our approach of direct HLA ligandome analysis of hematological malignancies identified a small panel of "pan-leukemia"- proteins and peptides that show cancer-exclusive representation across all 4 included hematological malignancies. However, due to the low presentation frequencies of the candidate targets within the different entities, target discovery and compound development for the immunotherapy of HM may be more effectively achieved in an entity-specific or even patient-individualized manner.