S843 THE PROLIFERATIVE HISTORY SHAPES THE DNA METHYLOME OF B‐CELL TUMORS AND PREDICTS CLINICAL OUTCOME

• Published in: HemaSphere 2019-06, Vol.3 (S1), p.376-n/a
• Main Authors: Duran‐Ferrer, M., Clot, G., Nadeu, F., Beekman, R., Baumann, T., Nordlund, J., Marincevic‐Zuniga, Y., Rivas‐Delgado, A., Ordoñez, R., Castellano, G., Kulis, M., Queirós, A., Seung‐Tae, L., Wiemels, J., Royo, R., Puiggrós, M., Torrents, D., Giné, E., Beà, S., Jares, P., Agirre, X., Prosper, F., López‐Otín, C., Puente, X.S., Delgado, J., López‐Guillermo, A., Campo, E., Martín‐Subero, J.I.
• Format: Article
• Language: English
• ISSN: 2572-9241; 2572-9241
• DOI: 10.1097/01.HS9.0000561652.34864.a1

Background: Beyond the known role of DNA methylation in gene regulation, recent studies indicate that most DNA methylation changes in cancer accumulate in silent regions without affecting gene expression. To obtain new insights into the biological and clinical implications of DNA methylation, we systematically analyzed genetic, epigenetic, transcriptional, and clinical data of five categories of B‐cell tumors including 1553 samples. This analysis unraveled that mitotic activity shapes the overall genetic and epigenetic landscapes of B cell tumors and affects the clinical behavior of patients. Aims: To provide a holistic view on the biological and clinical implications of DNA methylation variability in B‐cell tumors, with particular emphasis in non‐regulatory regions. Methods: We used 450k methylation arrays from normal B cell subpopulations as well as ALL, MCL DLBCL, CLL and MM patients (n = 67, 830, 74, 27, 490 and 65, respectively). We integrated ChIP‐seq data of 6 histone marks from normal B cells, MCL, CLL and MM (n = 15, 5, 7 and 5, respectively). This allowed us to construct the epigenetic Cumulative Mitosis (epiCMIT) score. We validated the epiCMIT using whole‐genome sequencing and gene expression data of the same CLL patients (n = 135). We additionally used 477 CLL patients with whole exome sequencing to find genetic driver alterations associated with high epiCMIT. From the clinical perspective, we used a machine‐learning approach to build an accurate pan B‐cell cancer diagnostic tool and related the epiCMIT to clinical outcome of ALL, MCL and CLL patients. Results: We dissected the DNA methylome of B‐cell tumors at different levels, including entity‐specific, subentity‐specific and patient‐specific variability. We showed that virtually the entire methylome (87%) is modulated in normal and/or neoplastic conditions, and subsequently dissected the sources of such variability. We initially identified the presence of differential methylation patterns among B‐cell tumors and their subentities. This analysis allowed us to construct an accurate pan B‐cell cancer diagnostic tool for 14 subentities of tumors requiring different clinical management. Furthermore, of particular interest were the hypomethylation patterns in ALL, MCL and CLL, which we could relate to differential binding of transcription factors important for the pathogenesis of each disease, such as ETS1 and MEIS1A in ALL, TCF/ZEB in MCL and NFAT in CLL. Despite the importance of DNA methylatio