Aberrant CpG island methylation in acute myeloid leukemia is accentuated at relapse

Abstract

DNA methylation of CpG islands around gene transcription start sites results in gene silencing and plays a role in leukemia pathophysiology. Its impact in leukemia progression is not fully understood. We performed genomewide screening for methylated CpG islands and identified 8 genes frequently methylated in leukemia cell lines and in patients with acute myeloid leukemia (AML): NOR1, CDH13, p15, NPM2, OLIG2, PGR, HIN1, and SLC26A4. We assessed the methylation status of these genes and of the repetitive element LINE-1 in 30 patients with AML, both at diagnosis and relapse. Abnormal methylation was found in 23% to 83% of patients at diagnosis and in 47% to 93% at relapse, with CDH13 being the most frequently methylated. We observed concordance in methylation of several genes, confirming the presence of a hypermethylator pathway in AML. DNA methylation levels increased at relapse in 25 of 30 (83%) patients with AML. These changes represent much larger epigenetic dysregulation, since methylation microarray analysis of 9008 autosomal genes in 4 patients showed hypermethylation ranging from 5.9% to 13.6% (median 8.3%) genes at diagnosis and 8.0% to 15.2% (median 10.6%) genes in relapse (P < .001). Our data suggest that DNA methylation is involved in AML progression and provide a rationale for the use of epigenetic agents in remission maintenance.

Figure 1

DNA methylation in AML at diagnosis and relapse. Methylation in normal peripheral blood controls is shown as box plots with bars indicating 5% to 95% confidence intervals. Methylation densities in individual patients with AML and their changes are shown in the second and third columns. PGRA, PGRB, CDH13, and OLIG2 genes were methylated above the cutoff at diagnosis in most patients and their methylation densities further increased at relapse. LINE-1 repetitive element showed a higher methylation in AML compared with normal peripheral blood controls and a slight decrease in methylation at relapse compared with diagnosis.

Figure 2

DNA methylation in patients with AML at diagnosis and relapse. Patients are arranged horizontally, genes vertically. Average methylation densities of 0% to 10% are shown in green (0%-15% for PGRB), 10% to 50% methylation in yellow (15%-50% for PGRB); and methylation over 50% in red. NR, no result. We observed increased methylation in AML at diagnosis in 144 of 295 data points and in 175 of 294 data points at relapse.

Figure 3

Changes in methylation between relapse and diagnosis. Changes in methylation z scores between relapse and diagnosis are shown as box plots. The boxes represent the 25% to 75% interquartile range, horizontal lines inside the boxes show the medians, and vertical bars show the 5% to 95% range. Gray boxes denote a significant increase in methylation in relapse (P < .05, Wilcoxon signed rank test). Mean z score was calculated as the mean of individual gene z scores. All genes showed increase in methylation at relapse with the exception of the LINE-1 repetitive element.

Figure 4

Changes in methylation between diagnosis and relapse in patients with AML. (A) Categorical changes. Red boxes denote changes from unmethylated at diagnosis to methylated at relapse. Green boxes show genes methylated at diagnosis and unmethylated at relapse. White boxes denote no change in methylation status between diagnosis and relapse. NR, no result available. Genes were considered methylated if their methylation density exceeded the 95% confidence interval established in normal controls. (B) Differences in methylation densities. Red boxes denote an increase in methylation density at relapse of more than 10% when compared with diagnosis; green boxes show decrease of methylation density at relapse over 10%, and white boxes denote methylation changes within the 10% interval. Blasts, bone marrow blast count changes between the diagnosis and relapse. Red boxes denote an increase in blasts at relapse of more than 10% when compared with diagnosis; green boxes show a decrease of blasts at relapse over 10%; and white boxes denote changes of blasts within the 10% interval. Changes in blast counts did not correlate with methylation changes (Spearman nonparametric test not significant).

Figure 5

Microarray analysis of genes hypermethylated in AML. The Venn diagram shows the overlap and differences in genes methylated at the time of diagnosis and the first relapse in at least 3 of 4 patients with AML studied. A total number of 4710 genes was analyzed by 16 475 microarray probes recognizing CpG islands near gene transcription start sites.