DNA Methylation Changes In Patients With Chronic Lymphocytic Leukemia Relapsing After Treatment Are Not Stochastic But Rather Selectively Affect Critical Pathways For B-Cell Physiology

Even with the advent of highly efficacious therapies for CLL, like the FCR regimen, many initially responding patients will eventually relapse, underscoring a characteristic resistance of the disease to current treatment options. The prognosis and management of patients relapsing after treatment differs significantly based on the nature of 1^st line therapy and the quality and duration of remission to that therapy, as well as on other prognostic factors, especially 17p and 11q deletions and IGHV mutational status. However, robust and specific markers predictive of response to treatment are still lacking and, therefore, understanding the mechanisms underlying clinical aggressiveness and resistance to treatment constitutes a research priority. Aberrant DNA methylation is increasingly recognized as relevant for CLL with strong correlations between promoter methylation and transcriptional silencing for critical genes. There is very limited information on DNA methylation changes in CLL patients relapsing after treatment, with the single published study reporting results on 9 patients treated with alkylating agents. With this in mind, we profiled the DNA methylation of paired samples from 14 CLL cases before treatment and relapsing after treatment with FCR (n=10), Rituximab-Bendamustine (n=3) and FC (n=1). DNA from tumor samples (≥70% CD19⁺ B cells) was bisulfite-converted and analyzed with the Infinium HumanMethylation450 BeadChip array. The methylation level of each CpG site was calculated in GenomeStudio Methylation module, while analysis for differentially methylated CpG sites (DMCpGs) was performed using the Genomestudio software and in-house developed Perl applications. Comparison between all samples before treatment versus all the relapsed samples revealed no DMCpGs; similar results were obtained for the FCR-treated subgroup. However, intra-individual analysis of the before treatment sample versus the relapsed sample revealed significant differences in most patients. The total number of DMCpGs was found to vary among patients from 0 to 58,648 out of ∼485,000 studied CpG sites. We next identified the genes related to DMCpGs across the gene region (promoter, 5’ UTR, first exon, gene body, 3’ UTR), ranging from 8 to 12,003 genes/case. In order to search for distinctive gene patterns among differentially methylated genes, we performed pathway enrichment analysis (KEGG Pathway database) with the WebGestalt bioinformatics tool, and using as input the total number of genes related to DMCpGs per case without subdividing them into hypo- and hypermethylated, after noticing that different CpGs of the same gene displayed different methylation status after relapse (some hypomethylated, others hypermethylated). Gene lists from 10 out of the 14 analyzed patients showed significant (p≤0.001) enrichment for several KEGG pathways (ranging from 17 to 173). Following, we searched for common pathways among patients and observed 4 common pathways (Focal Adhesion, Cell adhesion molecules, Calcium signaling pathway, Arrhythmogenic right ventricular cardiomyopathy) among all 10 patients, and 16 pathways (including MAPK, Notch and Wnt signaling pathways) common among 9/10 patients. In parallel we performed transcription factor (TF) target enrichment analysis (MsigDB database) for the 10 gene lists showing enrichment for KEGG pathways and observed significant (p≤0.001) enrichment for genes sharing common TF binding sites. We focused on the top-10 statistically significant TF binding sites for each case and found that 6/10 TF binding sites were common among all cases corresponding to 5 known TFs (TCF3, LEF1, MAZ, FOXO4, NFATC) and one unknown. Most of these TFs are predicted to target critical genes of significant B-cell signaling pathways e.g. TCF3 (E2A) and LEF1 targeting the MAPK, Notch, Wnt and Calcium signaling pathways. In conclusion, our analysis indicates that deregulation of DNA methylation in CLL cases relapsing after treatment is not stochastic but rather it selectively affects distinct pathways critical for B-cell/CLL biology. Intriguingly, on the basis of the present in silico findings, it could be argued that specific TFs acting as master regulators of B-cell differentiation may target several molecules of the epigenetically deregulated signaling pathways, creating a possible dynamic network.