Epigenetic Profiling of Primary CLL Reveals Novel DNA Methylation-Based Clusters and Novel Mechanisms of Leukemogenesis

Chronic Lymphocytic Leukemia (CLL) is the most common leukemia in the Western world with nearly 15,000 new cases diagnosed every year in the USA. The characterization of CLL has resulted in the identification of important disease biomarkers: these include the recurrent genomic deletions del17p and del11q, genomic complexity, TP53 mutations, the expression level of ZAP70 and the mutational status of IgV_H. While genomic and transcriptional profiling of CLL identified clinically and biologically relevant markers, there is still significant uncertainty about the pathobiology and the origin of CLL. It is increasingly clear that epigenetic deregulation plays an important role in the biology of all lymphomas/leukemias including CLL.

We hypothesized that DNA methylation profiling would allow us to identify new, biologically significant CLL subtypes and yield greater insight into the biology of this disease. We therefore examined the DNA methylation of over 240 patients with CLL using the HELP assay and hybridization to high density custom microarray that reports on the methylation status of more than 250,000 CpGs corresponding to 20,401 genes. Gene expression profiling and SNP array-based copy number assessments and targeted gene resequencing were available on most of these cases. We performed unsupervised analysis on the most variable probesets (standard deviation > 1.3) using K-means consensus clustering.

The experimental approach reproducibly identified three robust CLL subtypes based on epigenetic profiles. To identify the genes that define these three subtypes we next performed unequal variance t-test of the CLL subtypes comparing them to Peripheral Blood CD19+ B cells as a normal control, and identified that clusters are defined by differential methylation of 3719, 6145 and 3349 genes (selected probes displayed changes in methylation of at least 30% and FDR corrected p-value < 0.05), The three clusters featured respectively i) aberrant methylation of MYC and WNT target genes, ii) aberrant methylation of NOTCH1 targets and iii) aberrant methylation of bcl6 and inflammatory cytokines. There was inverse correlation between gene expression and cytosine methylation, suggesting that DNA methylation had an impact on the transcriptional programming of these CLL cases. Strikingly the CLL MYC/WNT cluster displayed poorer prognosis as opposed to the CLL BCL6 cluster (HR=0.14 95% CI: 0.07–0.30). The CLL NOTCH1 cluster had an intermediate prognosis. It was also notable that all CLL patients exhibited deregulation of the B-cell receptor pathway as compared to normal CD19+ B-cells, consistent with the notion that this pathway plays a critical role in CLL pathogenesis. Finally, we divided the cohort into training and testing cohorts and used a machine learning BDVAL algorithm to identify DNA methylation outcome classifiers. This procedure identified a 40-probeset classifier that accurately predicted outcome (Area Under the ROC Curve of 0.77; performance was assessed with 10 fold cross-validation in a training set with 76 patients; validation on an independent set of 105 samples).

This large epigenetic profiling study in CLL identifies aberrant epigenetic regulation as a core part of the pathobiology of CLL and identifies novel CLL clusters with distinct effects on survival. MYC-WNT pathway inhibitors are warranted for use in clinical trials for patients belonging to this aggressive epigenetically defined subtype.

No relevant conflicts of interest to declare.