Epigenetic Differences May Be Useful Biomarkers in AML Prognosis

Jacquelyn K. Beals, PhD

January 12, 2010

January 12, 2010 — A study of 344 patients with acute myeloid leukemia (AML) has determined that epigenetic differences in DNA methylation profiles are useful for classifying distinct AML subtypes and that a 15-gene DNA methylation classifier may effectively predict clinical outcomes in patients with AML.

Published online January 7 in Cancer Cell, the study was carried out by scientists at Weill Cornell Medical College, New York City, in collaboration with other New York colleagues and researchers in the Netherlands.

AML is both clinically and biologically heterogeneous, and it presents challenges for prognosis, risk assessment, and targeting therapy. The investigators point out that aberrant DNA methylation patterns are characteristic of cancer and that epigenetic factors such as DNA methylation can profoundly influence cells' biological behavior. The present study explored epigenetic patterns and AML subtypes.

The working hypothesis was that DNA methylation patterns in cancer, specifically AML, may cause or indicate clinically important cellular differences. To pursue this inquiry, researchers carried out DNA methylation profiling of blasts (non–terminally differentiated cells) from mononuclear cell fractions of 344 patients recently diagnosed with AML.

Statistical analysis grouped patients into 16 AML "clusters" based on their methylation profiles: 3 clusters coincided with AML subtypes classified by the World Health Organization, 8 contained a greater than expected number of AML cases with specific genetic or epigenetic aberrations (mutated NPM1, mutated CEBPA, or silenced CEBPA), and the 5 remaining clusters did not correspond to existing AML categories based on molecular or cytogenetic characteristics or morphology.

"these data indicate that DNA methylation is not randomly distributed in AML blasts but rather is organized into highly coordinated and well-defined patterns," the authors write. Although most AML subgroups were more highly methylated than normal comparison (CD34+) cells, several subgroups showed hypomethylation.

The influence of epigenetics compared with genetics is evident, for example, in DNA methylation cluster 3: Although 22 of the 31 participants have the same t(8;21) translocation, the remaining 9 patients lack this translocation. Nevertheless, survival curves of these 9 patients were identical (P = .83) to those of the other 22 cluster members, reflecting the biological significance of altered DNA methylation profiles.

The study also consistently found a "common aberrant DNA methylation signature consisting of 45 genes" in 10 or more of the 16 methylation profile groups, including 70% or more of the 344 patients. Many of these aberrantly methylated genes — genes for tumor suppression or transcription regulation — were logically associated with leukemic transformation.

However, "there are several [of these] genes that have not been functionally studied yet, and so it is hard to say one way or the other whether they would be logical AML genes," said senior author Ari M. Melnick, MD, from the Department of Medicine, Hematology Oncology Division, Weill Cornell Medical College, via email to Medscape Oncology. "The real surprise is that so many genes are so universally aberrantly epigenetically repressed in AML," he added.

"To put this in perspective: There isn't a single genetic lesion in AML that occurs as frequently as these 45 epigenetic lesions," said Dr. Melnick. "The frequency with which these genes are coordinately aberrantly methylated suggests that it might be the collective loss of the genes, rather than a single hit, that makes critical contributions to malignant transformation."

Among the 37 genes (of these 45) that were successfully analyzed, 32 showed complete silencing or significant downregulation because of their aberrant methylation.

Building on the consistent involvement of these 45 genes in AML, investigators looked for genes whose methylation state might predict clinical outcomes. They developed a predictor model based on 15 genes using a training set (n = 200), test set (n = 95), and validation set (n = 49) randomly assigned from the original AML cohort. The model successfully predicted overall survival (hazard ratio, 1.39; 95% confidence interval [CI], 1.10 - 1.75; P < .005) and event-free survival (hazard ratio, 1.53; 95% CI, 1.21 - 1.93; P < .0002) in the validation set.

DNA is stable and easily extracted from clinical specimens. The authors suggest that this 15-gene DNA methylation classifier might be used as a biomarker for decision-making in clinical trials.

However, "I think that it is early to conclude that epigenetic profiling will in fact make its way into the prognostication of AML," Martin Carroll, MD, associate professor in the Division of Hematology and Oncology, University of Pennsylvania, Philadelphia, commented in an email to Medscape Oncology. Dr. Carroll was not associated with the study.

"There are many prognostic variables in AML including age of the patient, subtype of disease...cytogenetic and molecular markers," said Dr. Carroll. "We do not yet know the robustness of DNA methylation analysis in the clinic, so there will be a technical question to be addressed as well."

Other recent studies show that DNA methylation may predict survival but not response to decitabine in myelodysplastic syndrome. Asked to comment, Dr. Carroll replied: "I think that the important take-home message...is that epigenetic (as opposed to genetic) changes are part of the pathogenesis of AML..... Is this a specific Achilles heel of the tumor or leukemia that can be targeted, or is it a marker of disease subtype that will help us [do prognosis] better? I think the latter is clearly going to be the case," he said. "Whether the former is true is an experiment still running."

However, Dr. Carroll anticipates that epigenetic patterns will be used to guide therapy. For example, 5-azacytadine and decitabine are 2 DNA demethylating agents currently approved by the US Food and Drug Administration for use in myelodysplatic syndromes.

"They have been tested in AML and have some activity, but it has not been documented that they are effective in achieving DNA demethylation in the clinic in AML," said Dr. Carroll. "This is one of the questions that the methodology shown in this manuscript can be used to address."

Dr. Melnick and Dr. Carroll have disclosed no relevant financial relationships.

Cancer Cell. Published online January 7, 2010.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: