Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2014 Aug;9(8):1108-19.
doi: 10.4161/epi.29315. Epub 2014 May 27.

Differential Methylation in CN-AML Preferentially Targets non-CGI Regions and Is Dictated by DNMT3A Mutational Status and Associated With Predominant Hypomethylation of HOX Genes

Affiliations
Free PMC article

Differential Methylation in CN-AML Preferentially Targets non-CGI Regions and Is Dictated by DNMT3A Mutational Status and Associated With Predominant Hypomethylation of HOX Genes

Ying Qu et al. Epigenetics. .
Free PMC article

Abstract

The extent and role of aberrant DNA methylation in promoter CpG islands (CGIs) have been extensively studied in leukemia and other malignancies. Still, CGIs represent only a small fraction of the methylome. We aimed to characterize genome-wide differential methylation of cytogenetically normal AML (CN-AML) cells compared with normal CD34(+) bone marrow cells using the Illumina 450K methylation array. Differential methylation in CN-AML was most prominent in genomic areas far from CGIs, in so called open sea regions. Furthermore, differential methylation was specifically found in genes encoding transcription factors (TFs), with WT1 being the most differentially methylated TF. Among genetic mutations in AML, DNMT3A mutations showed the most prominent association with the DNA methylation pattern, characterized by hypomethylation of CGIs (as compared with DNMT3A wild type cases). The differential methylation in DNMT3A mutant cells vs. wild type cells was predominantly found in HOX genes, which were hypomethylated. These results were confirmed and validated in an independent CN-AML cohort. In conclusion, we show that, in CN-AML, the most pronounced changes in DNA methylation occur in non-CGI regions and that DNMT3A mutations confer a pattern of global hypomethylation that specifically targets HOX genes.

Keywords: DNA methylation; DNMT3A; Homeobox gene family; acute myeloid leukemia; non-CGI region.

Figures

None
Figure 1. Differentially methylated CpG sites (DMC) between CN-AML and normal bone marrow CD34+ cells. DMC was defined by moderated Student’s T-test combined with ∆β-value between the two groups (FDR < 0.05 and ∆β > 0.17). (A) Principal component analysis (PCA) of DMC. (B) Density plots of DMC in CN-AML (upper panel in red) and NBM CD34+ (lower panel in green). (C) Distribution of DMC in different CpG regions compared with the proportions of all analyzed CpG sites on the platform. Enrichment analysis was performed by Chi-square test (χ2). D. Smoothed density scatter plot of average methylation levels in different CpG regions. Smoothed CpG densities show in gradient blue. Hypermethylated DMC is marked in red and hypomethylated DMC in green.
None
Figure 2. Association profile between global methylation patterns and mutational status of CN-AML cases. In total, the 41723 most variably methylated CpG (MVM) sites were selected by choosing sites with a standard deviation of β-values higher than 0.15 across all 62 patient samples. (A) Smoothed scatter plots and boxplots of general methylation levels according to gene mutational status (mutated vs. wild type) on MVM sites. CN-AML patients were divided according to each gene mutational status and average methylation levels of mutated vs. wild type cases were plotted in parallel to scatterplot and boxplots (DNMT3A mutated vs. wild type, Wilcoxon Sum rank P < 10−3) (B) Unsupervised consensus cluster plot showing the correlation of global methylation patterns of 62 CN-AML patients. Patients were sequentially segregated into 2 to 6 clusters (K2 to K6) according to the correlation of methylation patterns (Fig. S4). K2, K4 and K6 clusters are shown in the plot with a color code. The mutational statuses of the corresponding patients are marked (mutated cases in black, not known in gray). Statistical analysis was performed validating the distribution of indicated gene mutations among patient groups in K2 to K6 clusters by Fisher’s exact test. (C) Average methylation levels of patients according to K2 cluster are shown in the left panel (Wilcoxon sum rank test, P < 10−4) and a Kaplan-Meier diagram shows the prognostic difference of patients divided according to K2 cluster (log-rank test, P = 0.0296)
None
Figure 3. Hypomethylation in DNMT3A mutated patients. Patients were separated according to DNMT3A mutational status and, in total, 5074 DNM3A-DMC were defined by FDR < 0.05 and |∆β| > 0.17. (A) Volcano plot shows methylation differences of DMC between DNMT3A mutated vs. wild type CN-AML cases (DNMT3A-DMC; FDR < 0.05, |∆β| > 0.17). DMCs are marked with red circles. X-axis shows the FDR odd ratios (-log10 FDR) and Y-axis shows the average difference between mutated and wild type cells. (B) Heatmap of DNMT3A-DMC. The mutational status is color-coded in the column index, mutated is shown in red and wild type is shown in blue. The CpG regions of DNMT3A-DMC are presented in rows (CGI in red, shores in yellow, shelves in green and open sea in blue). (C) Hypomethylation of HOXA5 and PAX7 in DNMT3A mutated CN-AML. DNMT3A-DMC of HOXA5 and PAX7 are shown in the diagram. Methylation levels are plotted according to chromosomal positions (β-value, scale 0 to 1.0). Patients are color-coded according to DNMT3A mutational status (mutated in red and wild type in blue). Schematic graphs and annotated CpG islands (CGIs in light blue) were obtained from USCS Genome browser.
None
Figure 4. DMCs are enriched in transcription factor genes. The group of 1620 TF was compared with the total number of analyzed genes. The proportion of DMCs for each gene was tested by Fisher’s exact test vs. the proportion of total number of genes with annotated DMC. An FDR lower than 0.01 combined with a minimum of 3 DMCs per gene was used to eliminate genes with possible random methylation change. (A) Number of TF genes enriched for DMC compared with all genes with DMC (χ2, P < 10−4). (B) Differential methylation of WT1 validated by pyrosequencing. DMCs in WT1 are shown in a heatmap for the corresponding genomic locations. Genomic features are presented using the same annotation as in Figure 3C. Green indicates normal control samples and red CN-AML samples. Pyrosequencing targets are marked with dark blue blocks. Region 1 represents the intragenic region and Region 2 is in the core promoter region of WT1.
None
Figure 5. Correlation between DNA methylation and gene expression. (A) Correlation between average DNA methylation and average gene expression of all genes represented on both methylation and expression arrays. Promoter and gene body methylation was summarized for each type of CpG region (in relation to CGI). Correlation coefficient was tested by Spearman’s correlation test. (B) Association between differentially methylated and differentially expressed genes within the CGI and open sea-associated promoters, respectively. All genes with significant differential methylation were plotted in the figure. In the left panel, the X-axis shows methylation difference in average β-value (∆β-value) between CN-AML and CD34+ controls; Y-axis shows changes in expression as log 2-transformed fold change. Differentially expressed genes are marked in red circles. Proportional Venn diagrams (right panel) show the differential expression corresponding to methylation changes in different CpG regions (tested by Fisher’s exact test).

Similar articles

See all similar articles

Cited by 34 articles

See all "Cited by" articles

Publication types

MeSH terms

Feedback