Identification of the PML-RARα Associated DNA Methylome Using Reduced Representation Bisulfite Sequencing in Primary Patient Samples

Abstract 2436

The origin of altered DNA methylation in cancer and more specifically in leukemia remains incompletely understood. One favored hypothesis suggests that aberrant methylation is recruited to direct genomic target regions of transcriptional repressor oncogenes. PML-RARα, as a transcriptional repressor oncogene, mediates a repressed chromatin structure at its genomic targets and leads to a rather uniform leukemia subtype. In the current study we aimed to identify patterns and variability of DNA methylation changes in PML-RARα associated acute promyelocytic leukemia (APL).

We used reduced representation bisulfite sequencing to investigate a large number of samples at single CpG-site resolution with a focus on CpG islands and related promoters. Overall, 20 APL patient samples were sequenced from patients at primary diagnosis. As controls we used 10 remission bone marrow samples and three sets of MACS-purified CD34+ hematopoetic progenitor cells from healthy donors as well as in vitro differentiated promyelocytes. Using Illumina next generation sequencing we obtained about 1 × 10⁷ uniquely mapped bisulfite sequencing reads per patient in total. More than 10 sequencing reads were available for on average 1.2 million distinct CpG sites per sample.

Unsupervised hierarchical clustering revealed that APL samples clustered separately from remission bone marrow, CD34+ progenitors and promyelocytes. Using Poisson-Regression (p<10⁻⁵) and at least a 2-fold change in methylation as threshold for analysis, primary diagnosis samples where markedly hypomethylated versus controls. 2160 promoter regions were more than 2-fold hypomethylated in primary diagnosis samples compared to controls. However, 716 promoter regions were also 2-fold hypermethylated.

A microarray data analysis (www.leukemia-gene-atlas.org) was carried out. The top hypomethylated genes, as sorted by fold changes in methylation and p-values, showed differential expression.

Using published PML-RARα binding sites, we also examined methylation differences within these. A total of 11026 CpG-sites comprising 368 promoters could be taken into account. Notably, there was a high degree of hypomethylation at PML-RARα binding sites when comparing primary diagnosis samples with controls, while only a few binding sites appeared hypermethylated. Differential hypomethylation at PML-RARα binding sites was accentuated in both CpG islands and single CpGs.

We also carried out a pathway analysis using Ingenuity IPA Software. Several canonical pathways appeared differentially methylated. This included, for example, 32 genes of the Wnt-signaling pathway as well as CDK5-signaling genes. Cancer-related genes such as CTNNA2 were also differentially methylated.

Taken together, our data provide evidence of a specific PML-RARα associated APL-Methylome. Unexpectedly, primary patient samples exhibited a high degree of relative hypomethylation within published PML-RARα binding sites and across promoter regions in general. Methylation differences were associated with deregulated canonical pathways and especially Wnt-signaling.