Modeling of the Epigenome of the Cell-of-Origin Identifies Cancer-Specific DNA Methylation Patterns in CLL

Normal B cells undergo extensive epigenetic programming during normal differentiation and distinct B cell differentiation stages represent unique DNA methylation patterns. Chronic Lymphocytic Leukemia (CLL) originates from rapidly differentiating B cells and their DNA methylation signature is stably propagated in CLL. Consequently, CLL methylome data can be used to infer the putative cell-of-origin (COO) for each individual CLL case. We define the COO of CLL as the cell that has acquired a first oncogenic hit and which will initiate tumorigenic growth if one or more additional hits have been acquired. This means that two factors contribute to the epigenetic profile of CLL cells: first, the epigenetic profile of the founder B cell at the time of malignant transformation and second, CLL-specific epigenetic alterations that are acquired during leukemogenesis and progression of the disease. Previous studies using peripheral blood CD19⁺ B cells as a reference for aberrant methylation calls completely neglected the massive epigenetic programming that occurs during normal B cell differentiation. Thus, novel strategies aiming at identifying truly CLL-specific methylation changes considering the highly dynamic methylome during normal B cell differentiation were urgently needed.

Here we outline a new analytical framework to delineate CLL-specific DNA methylation. We demonstrate how this approach can be applied to detect epigenetically deregulated transcripts in CLL. Firstly, we modeled the epigenome dynamics occurring during normal B cell differentiation using linear regression. The DNA methylomes of CLL cells were then precisely positioned onto the normal B cell differentiation trajectory to define the closest normal B cell methylome for every CLL patient, the COO. The epigenome of the COO then served as a reference for aberrant DNA methylation calls. We dissected two categories of CLL-specific methylation events: those occurring at sites undergoing epigenetic programming during B cell differentiation and those that normally do not change during B cell differentiation. The first group was further subdivided into class A and B, displaying exaggerated methylation loss or gain, respectively, and class C showing both hyper- and hypomethylation relative to the normal differentiation. The second group was classified into class D displaying hypo- and class E showing hypermethylation. Overall, only 1.6% of the CpG-sites (7,248 CpGs) represented on the Illumina 450k array were affected by disease-specific methylation programming, mostly hypomethylation (6,680 CpGs).

Next, the molecular programs underlying the CLL-specific methylation patterns were investigated. We tested enrichment of chromatin states and of transcription factor binding sites (TFBS) as identified in an immortalized B cell line (GM12878). This indicated that disease-specific methylation events target transcriptionally relevant cis-regulatory elements in CLL (enhancers, weak and poised promoters and insulator regions). In line with this, CLL-specific differentially methylated regions affected TFBS associated with signaling pathways known to be important in normal B-cell differentiation (i.e. BATF, EBF1). We also observed altered methylation at CTCF binding sites suggesting their involvement in CLL pathogenesis.

In the present work, we dissected CLL methylomes to distinguish between normal B cell differentiation-associated methylation patterns and CLL-specific methylation events. We showed that this approach is indispensable to identify key pathogenic events driving CLL pathogenesis. The relevance of our approach was demonstrated by contrasting the number of epigenetically deregulated miRNAs and protein-coding genes to those determined with a classic analysis using CD19⁺ B cells as controls. This highlights the extent of overcalling of CLL-specific methylation patterns in previous studies (~30-fold for protein-coding genes and ~10-fold for miRNAs) and stresses the importance to consider normal differentiation trajectories for the identification of aberrant DNA methylation events. Here we propose 11 protein-coding genes (e.g. DOK2, CLLU1) and 4 miRNAs (e.g. miR-486, miR-195) as being epigenetically deregulated in CLL. Our analytical approach provides a general framework for the identification of disease-specific epigenomic changes that should be applicable to other cancers in the future.