Decitabine (DAC) Treatment Induces Very Early In Vivo DNA Methylation and Transcriptome Changes in Primary AML Peripheral Blood Blasts

The in vitro effects of the DNA hypomethylating agents 5-aza-C (5-azacytidine) and 5-aza-dC (DAC) upon the epigenome and transcriptome of AML cell lines have been investigated by numerous groups, demonstrating induction of gene-specific and global hypomethylation, as well as up- and downregulation of multiple genes. However, their in vivo mechanisms of action are not well-studied, with only very sparse data on primary blasts from patients (pts) with myeloid neoplasias treated with these compounds (Daskalakis et al., Blood 2002, Yang et al., Cancer Research 2006, Fandy et al., Blood 2009). This is primarily due to the substantial methodological challenges of obtaining sufficient cell numbers of bona fide malignant cells during treatment. To study very early in vivo effects, we isolated and analyzed primary cells from AML pts with high numbers of peripheral blood blasts treated with the hypomethylating agent DAC.

Of 48 consecutive pts aged >60 years with newly diagnosed AML treated with DAC as described (Lübbert et al., 2011) at a single centre within a multicenter phase II study (trial 00331), successful sequential sampling of peripheral blood blasts in sufficient numbers and purity was feasible in 8 pts. The median WBC at AML diagnosis was 18,400/μl (range 4,800–241,000), the median age was 80 years, and 6/8 had cytogenetic abnormalities. The methylation state of LINE1 and 35 genes (5' regions, represented by a total of 635 evaluable CpG dinucleotides) was quantified by MALDI-TOF mass spectrometry. Gene selection criteria were based on previous identification as potentially silenced tumor suppressors in genome-wide methylation analyses, and on reports as epigenetic targets in myeloid neoplasia. Transcriptional changes were assessed by array-based transcriptome profiling using the HG-U133plus 2.0 GeneChip array.

Isolation and purification of sufficient numbers of blasts was done both immediately before and at day 6 (median, range 4–7) after start of DAC infusion. Methylation quantification of pre-treatment blasts yielded patterns clearly distinguishing them from CD34-positive normal hematopoietic precursors of healthy donors (n=10). Specifically, P73, P15 and CDH1 exhibited the strongest and most consistent methylation gains (9–18%, p<0.015) in the malignant vs. normal cells. Early after DAC treatment, a striking decrease in LINE1 methylation was noted in 7/8 pts (p = 0.02), the most consistent effect among all investigated genomic regions. Overall, DAC-induced methylation changes (hypo- and/or hypermethylation) occurred in all eight pts. Hypomethylation was observed in the CpG islands of PBX1 (median 11%, range 5.5–25%, p<0.069), ESR1 (median 1.5%, range 0–9%, p<0.1) and MPO (median 4.3%, range 1–19%, p<0.1). For P73 and TLX3, substantial hypomethylation was observed in a subset of patients (maximum changes of 36% and 34% respectively), while other pts showed unchanged or increased DNA methylation levels. Unsupervised hierarchical clustering revealed that the highest similarities were intra- but not inter-individual. When interrogating unbiased methylation changes at all 635 CpGs, significant hypomethylation was induced in 4/8 pts, significant hypermethylation in 1/8 pts, and non-significant methylation changes seen in 3/8 pts. DAC-induced mRNA expression changes also occurred in all eight pts, without clear correlations between hypomethylation and restored expression.

DAC administered at a clinically effective dose and schedule resulted in a consistent and pronounced in vivo decrease of LINE1 methylation already at early time points after treatment start, with significant hypomethylation events in 50% of the pts This might be indicative of effective, early Dnmt1 depletion. Induction of multiple transcriptional changes (but up- and downregulation) was also noted in all pts. This is, to the best of our knowledge, the first report of an integrated methylation/transcriptome analysis in primary leukemia cells isolated at very early time points during therapy with hypomethylating agents, i.e. before clonal selection or even replacement and the occurrence of secondary events. Our observations are compatible with a mechanism of action of DAC that is distinct from those of a non-hypomethylating cytosine analogues such as cytarabine, in line with in vitro results (Flotho et al, Leukemia 2009).

No relevant conflicts of interest to declare.