Signatures of polycomb repression and reduced H3K4 trimethylation are associated with p15INK4b DNA methylation in AML

Abstract

DNA hypermethylation of the p15INK4b tumor suppressor gene is commonly observed in acute myeloid leukemia (AML). Repressive histone modifications and their associated binding proteins have been implicated in the regulation of DNA methylation and the transcriptional repression of genes with DNA methylation. We have used high-density chromatin immunoprecipitation-on-chip to determine the histone modifications that normally regulate p15INK4b expression in AML cells and how these marks are altered in cells that have p15INK4b DNA methylation. In AML patient blasts without p15INK4b DNA methylation, a bivalent pattern of active (H3K4me3) and repressive (H3K27me3) modifications exist at the p15INK4b promoter. AML patient blasts with p15INK4b DNA methylation lose H3K4me3 at p15INK4b and become exclusively marked by H3K27me3. H3K27me3, as well as EZH2, extends throughout p14ARF and p16INK4a, indicating that polycomb repression of p15INK4b is a common feature in all AML blasts irrespective of the DNA methylation status of the gene. Reactivation of p15INK4b expression in AML cell lines and patient blasts using 5-aza-2'-deoxycytidine (decitabine) and trichostatin A increased H3K4me3 and maintained H3K27me3 enrichment at p15INK4b. These data indicate that AML cells with p15INK4b DNA methylation have an altered histone methylation pattern compared with unmethylated samples and that these changes are reversible by epigenetic drugs.

Figure 1

p15INK4b DNA methylation and expression in AML cell lines used in study. (A) Bisulfite sequencing of the p15INK4b exon 1 CpG island. Filled circles represent percentage of DNA methylation at CpG positions from 10 independently sequenced clones (● represents ≥ 50% methylcytosine; , 10%-49% methylcytosine). (B-C) Quantitative reverse-transcriptase (RT)–PCR analysis of p15INK4b, p14ARF, and p16INK4a expression in AML cell lines. Copies of cDNA per 100 ng of total RNA were calculated from a plasmid DNA standard curve and represent the averages of 3 independent experiments with error bars of SD.

Figure 2

Distribution of H3K4me3, H3K27me3, and EZH2 at INK4b-ARF-INK4a differs in AML cell lines with DNA methylation. ChIP-on-chip analysis of enrichment of H3K4me3 (blue), H3K27me3 (red) on chromosome 9: 21800000 to 22450000 (A) and chromosome 9: 21953000 to 22003000 (B) in AML cell lines. Top axis represents H3K4me3; and bottom axis, H3K27me3. Enrichment of EZH2 (green; C) and H3K9me3 (orange; D) at the INK4b-ARF-INK4a region in AML cell lines. Normalized enrichment data (IP/input) are plotted on a log 2 scale.

Figure 3

AML patient blasts with p15INK4b DNA methylation display reduced H3K4me3 enrichment at the promoter. (A) Bisulfite sequencing of the p15INK4b exon 1 CpG island in AML patient samples. Filled circles represent the frequency of DNA methylation at CpG positions in sequenced clones (● represents ≥ 50% methylcytosine; , 10%-49% methylcytosine). (B) Quantitative RT-PCR analysis of p15INK4b expression in AML patient blasts. cDNA copies per 50 ng total RNA were calculated from a plasmid DNA standard curve and represent the averages of 3 independent experiments with error bars of SD. Quantitative PCR analysis (C) of H3K4me3 and H3K27me3 enrichment at the p15INK4b promoter. Data are averages of ΔΔC_t compared with input DNA from 3 technical replicates with error bars of SD. ChIP-on-chip analysis of enrichment of H3K4me3 (blue) and H3K27me3 (red) on chromosome 9: 21953000 to 22003000 (D) and chromosome 9: 21990000 to 22003000 (E) in AML patient samples. Top axis represents H3K4me3; and bottom axis, H3K27me3. Normalized enrichment data (IP/input) are plotted on a log 2 scale.

Figure 4

Reactivation of p15INK4b expression and loss of DNA methylation after treatment with DNMT or HDAC inhibitors. (A) Quantitative RT-PCR analysis of p15INK4b expression in KG-1, KG-1a, and AML-193 cells after reactivation with 5-aza-dC and/or TSA. (B) Quantitative RT-PCR for p15INK4b, p14ARF, and p16INK4a expression in KG-1cells after reactivation with 5-aza-dC and/or TSA. (A-B) *Increase in expression relative to control (P < .05, 1-way analysis of variance). (C) Quantitative RT-PCR for p15INK4b expression in AML patient blasts treated with 5-aza-dC or TSA. Quantitative RT-PCR data in both cell lines and clinical samples were normalized to control cells and calculated from ΔΔC_t averages of 3 independent experiments with error bars of SD. *Significant increase in expression in cells treated with a combination of 5-aza-dC and TSA compared with cells treated with 5-aza-dC alone (P < .05, 1-way analysis of variance). (D) Bisulfite sequencing of the p15INK4b CpG island in individual clones from control and treated KG-1, AML-193, and AML8 blasts after treatment with 5-aza-dC and/or TSA. Filled circles represent percentage DNA methylation at individual CpG positions from 10 independently sequenced clones (● represents ≥ 50% methylcytosine; , 10%-49% methylcytosine).

Figure 5

Reactivation of p15INK4b using DNMT and HDAC inhibitors is associated with an increase in H3K4me3 and maintenance of H3K27me3 in KG-1 and AML blasts. ChIP-quantitative PCR measured enrichment of H3K4me3 and H3K27me3 at the p15INK4b promoter after reactivation in KG-1 (A) and AML-193 cells (B). The enrichment of H3K4me3 (C) and H3K27me3 (D) measured at the p15INK4b promoter by quantitative PCR in clinical samples AML8, AML9, AML10, and AML11 patient blasts treated with DMSO (control), 5-aza-dC, or TSA. (C) *Significant increase in H3K4me3 enrichment compared with control treatment (P < .05, 1-way analysis of variance). Enrichment was calculated from ΔΔC_t averages of 3 independent experiments with error bars of SD. (E) ChIP-on-chip for enrichment of H3K4me3 (blue) and H3K27me3 (red) at INK4b-ARF-INK4a in DMSO control and 5-aza-dC plus TSA-treated KG-1 and AML-193 cells. (F) ChIP-on-chip for enrichment of H3K4me3 (blue) and H3K27me3 (red) at INK4b-ARF-INK4a in AML8 patient blasts treated with control DMSO, 5-aza-dC, or TSA. Normalized enrichment data (IP/input) are plotted on a log 2 scale.