Missteps in “tango” for epigenome targeting

In this issue of Blood, Fandy and colleagues describe the results of a trial in which leukemia patients were treated with 5-azacytidine (5-azaC) and the HDAC inhibitor, entinostat.¹  Their results confirm the therapeutic efficacy of these agents but also demonstrate that we have much to learn about the mechanism of action of agents that target the epigenome.

Deregulated epigenetic marks, represented by methylation of clusters of CpG dinucleotides in the promoter DNA along with histone methylation and deacetylation, result in silencing of tumor suppressor genes (TSGs), which play a significant role in the pathogenesis and biology of myelodysplastic syndrome (MDS) and acute myeloid leukemia.²  Sequential treatment with DNA methyltransferase 1 (DNMT1) followed by a histone deacetylase (HDAC) inhibitor derepressed TSGs, as well as induced cell-cyclin dependent kinase inhibitors, growth arrest, differentiation, and apoptosis of transformed cells.^3,4  Administration of entinostat on the 3rd and 10th days of 5-azaC treatment resulted in clinical responses in 46% of the patients, most often after 4 to 6 cycles of therapy. While confirming the efficacy of these agents, the findings raise more questions than answers as to the molecular mechanism of action of agents that target epigenetic mechanisms. In the study, Fandy et al analyzed DNA methylation levels in the bone marrow and purified CD34⁺ cells obtained during cycle 1 of treatment by methylation-specific polymerase chain reaction (MSP) at 4 commonly methylated TSGs, specifically p15INK4b, CDH-1, DAPK-1, and SOCS-1. The analyses revealed that, although gene demethylation occurred in some patients, there was no correlation between demethylation of any of these 4 genes and the observed clinical response. The MSP results were largely qualitatively confirmed through the use of high-resolution bisulfite genomic sequencing (BGS). However, the BGS results raise some interesting conceptual and methodological issues. In most samples in which a positive result for methylation by MSP agreed with elevated DNA methylation by BGS, the total methylation level approached only 20% in most cases. Although this may be the result of cellular heterogeneity (as evidenced by some clones that are almost fully methylated and others with no methylation), there are many examples of partially methylated clones. The consequences of such low-level DNA methylation on gene transcription are largely unknown and may, in fact, be insufficient to repress gene activity. In addition, there are situations in which a sample scoring positive for methylation by MSP had little or no DNA methylation when analyzed by BGS. These results, therefore, stress the need for carefully correlating DNA methylation levels with transcription, as well as highlighting the need for more quantitative and less subjective methods than MSP (eg, pyrosequencing) for interrogating DNA methylation levels in clinical specimens. Global analysis of gene expression also failed to reveal patterns of gene reactivation that characterized responders and nonresponders. Given that aza-nucleosides are known to induce DNA damage, Fandy et al also determined the levels of γH2AX, a well-known marker of DNA damage, in their patient samples. γH2AX is a phosphorylated form of the histone variant H2AX that becomes rapidly and specifically localized to regions of DNA containing double strand breaks. This aspect of aza-nucleosides has not been examined in earlier studies and, in fact, revealed a robust induction of γH2AX even at the lowest doses of 5-azaC. Like the other parameters examined in this study, however, there was no correlation between γH2AX induction and clinical response as it occurred in responders and nonresponders.

So where do we go from here with epigenetic-based cancer therapies? Although there is little doubt the aza-nucleosides induce demethylation in patient tumor cells, the exact therapeutic molecular target(s) remains an open question. At the most simplistic level, one could argue that we simply have not found the one or few critical genes that, when demethylated and reexpressed, result in differentiation or cell death. For these studies, use of more sensitive microarray platforms than the one employed by Fandy and colleagues, along with more quantitative confirmatory methods such as quantitative RT-PCR will likely help to address this possibility. Coupling these methods with genome-wide tiling arrays or high-throughput sequencing DNA methylation analyses would also be expected to shed significant light on whether a few critical genes are being targeted by epigenetic therapies. Compared with cytotoxic combination chemotherapy, the slower development of response occurring after 4 cycles of therapy may suggest that the epigenetic effects and ensuing potential differentiation/cytotoxic effects were occurring in the small subset of early progenitor or stem cells.⁵  Therefore, as much as possible, the correlative analyses should be performed in these cells rather than in the bulk progenitors. Other parameters worthy of future consideration that could influence or predict clinical responses to epigenetic therapies include DNMT expression levels and expression of modifiers of aza-nucleoside bioavailability: nucleoside transporters like hCNT1⁶  or cytosine deaminase.⁷  Finally, it should be considered that aza-nucleosides and HDAC inhibitors are exerting effects indirectly on other epigenetic marks, such as histone methylation, and that these may be more predictive of clinical response. It is well known that DNA methyltransferases interact not only with HDACs but also with various histone methyltransferases and, as such, it is reasonable to expect that interfering with one epigenetic mark will impact other marks.^4,5,8  The full extent of this interaction in vivo in human patients has yet to be explored.

Overall, this study creates a sobering realization about our lack of full understanding of several important dimensions of epigenetic targets and therapies. Identifying the predictive biomarkers of clinical response and elucidating the mechanisms of resistance to epigenetically targeted agents looms as a big challenge, but a compelling necessity for learning and refining the steps of this tango!