Myelodysplastic syndrome (MDS) is a preleukemic condition that transforms into acute myeloid leukemia (AML) in approximately 30-40% of cases. MDS-associated leukemia is rarely cured by chemotherapy, and patient prognosis is typically poor [1]. MDS represents an excellent model of leukemic development with progressive increases in blastic bone marrow involvement; however, the genetic events underlying this evolution remain poorly understood.

MDS is separated into subtypes based on bone marrow blast and dysplasia. Among the MDS subtypes, refractory anemia with excess blast (RAEB) is the most common pre-leukemic type and is associated with poor prognosis. Thus, the focus of the present study is on the bone marrow blast subtype of MDS, specifically on different DNA methylation profiles between the subtypes morphologically affected.

A whole genome-wide DNA microarray technique was used in the present study to assess the DNA methylation profiles between RAEB and refractory cytopenia with multilineage dysplasia (RCMD). Genes displaying different DNA methylation patterns were identified using Ingenuity Pathway Analysis (IPA) tools. In this study, we aimed at identifying genes that were hypermethylated in RAEB but not in RCMD.

DNA methylation is associated with the transcriptional silencing of genes in cell differentiation, apoptosis, growth regulation, cell signaling, detoxification, and tumor invasiveness and thus confers a selective growth advantage to malignant cells [4]. AML and MDS are diseases associated with aberrant methylation in a number of genes [5, 6]. Hypermethylation of genes involved in cell-cycle control and apoptosis is observed commonly in these diseases, particularly high-risk MDS. Transcription of the p15 gene often is silenced due to abnormal methylation of its promoter region, and several studies have indicated that approximately 50% of MDS patients exhibit this alteration [7, 8]. Other genes frequently affected by hypermethylation in MDS are cancer 1 (HIC1), E-cadherin (CDH1), and estrogen receptor (ER) [9].

The hypermethylation of promoters from a set of cancer-related genes has been linked to the progression of MDS [10]. By examining 1,505 CpG sites in 807 genes using microarray technique and a high-density single nucleotide polymorphism array, Jiang et al. reported that aberrant hypermethylation is associated with disease stage in MDS. The ubiquity, extent, and progression of aberrant hypermethylation in the disease stage suggests that hypermethylation is the primary source of tumor suppressor gene silencing and clonal variation in the neoplastic evolution of myeloid disease.

The present study focuses on bone marrow blast in primary MDS because increased numbers of bone marrow blasts occur during disease progression and are linked to an increased risk for transformation to AML. The hypothesis that drove the present study was that the mechanism of increased bone marrow blast will provide clues to the neoplastic evolution of MDS and may be associated with aberrant DNA methylation. Based on the WHO classification, 9 patients of the RAEB subtype and 11 patients of the RCMD subtype were selected and used to determine the different patterns of gene methylation status between the 2 subtypes.

The use of complimentary DNA microarray technology in MDS has demonstrated a high level of gene expression heterogeneity among MDS patients [11]. However, the profiling of DNA methylation in MDS by the microarray technique has not been studied yet. The advent of microarray technology has provided new opportunities for high-throughput study on DNA methylation [12]. The whole genome-wide microarray approach has been used to study the mechanisms of cancer disease progression and prognosis. Thus, the present study investigated DNA methylation in MDS by using a commercial whole-genome-wide DNA microarray technique. The Infinium Methylation Assay is a new technology that differs from the Goldengate Methylation Assay in terms of whole-genome-wide content, but not cancer/custom content. It still is not practical to use genome-wide sequencing of individual tumors to assess the entire mutation spectrum. Similarly, research that uses the genome-wide DNA methylation assay is in its early stages and expensive for use. These considerations underlie the decision to limit the assay to 20 MDS patients in this initial study.

Among the 14,495 genes, 69 were hypermethylated in RAEB and hypomethylated in RCMD. The DNA methylation profile of the 20 samples exhibited a high level of heterogeneity. The small sample size limited identification of different patterns between subtypes based on hierarchical clustering. Furthermore, this study was conducted on mononuclear nonadherent bone marrow cells from routine aspirates. These samples represented a non-homogeneous pool of different maturation stages, which resulted in heterogeneity.

It is a daunting task to use microarray studies to discover biological links between these genes and disease progression and to integrate them within a connected network. The present study utilized IPA software to analyze the candidate genes further. The IPA-computed score of each network indicates the likelihood that input genes in the network will be found together due to random chance. The 69 candidate genes analyzed using IPA mapped in 5 networks (see Table 3). The top scoring network (network 1) was selected for further study.

Although the functions of the 5 candidate network genes (GSTM5, BIK, CENPH, RERG, and ANGPTL2) were not determined, previous studies have provided sufficient evidence that links these genes with malignancy. The methylation status of each gene was validated by MPS analysis on both subtypes.

The GSTM5 gene encodes a GST that belongs to the mu class of enzymes, which conjugates with glutathione to detoxify electrophilic compounds, including carcinogens, therapeutic drugs, environmental toxins, and products of oxidative stress. Genetic variation in mu genes can change an individual's susceptibility to carcinogens and toxins and also affect the toxicity and efficacy of certain drugs [13]. Expression of mu class GST has been associated with a poor prognosis in acute lymphoblastic leukemia that results from cytotoxic drug resistance [14, 15]. Recently, Peng et al. [16] reported that DNA hypermethylation regulates the expression of mu-class GST in Barrett's adenocarcinoma. In the present study, MSP analysis results indicated that the GSTM5 gene was methylated more frequently in RAEB patients than in RCMD patients (55.5% vs. 20%, respectively). However, in comparison with the normal control, slightly methylated products were also noted by MSP. Although the possible role for GSTM5 in the development of cancer remains uncertain, the hypermethylation of GSTM5 may be associated with the function of other members of the mu-class GSTM family and may change individual susceptibility to carcinogens and toxins, thereby resulting in secondary MDS mutations that underlie disease progression.

The ANGPTL2 gene encodes angiopoietin-like protein 2, a secreted protein belonging to the angiopoietin family. Kikuchi et al. [17] have reported that ANGPTL2 frequently is silenced in ovarian cancer cell lines via DNA methylation of sequences around the CpG-island that exhibit promoter activity. Zhang et al. [18] reported that ANGPTL2 and ANGPTL3 proteins stimulate ex vivo expansion of hematopoietic stem cells. Here, the ANGPTL2 gene displays a slightly higher methylation status in RAEB patients than in RCMD patients (44.4% versus 10%, respectively), but both subtypes exhibited less DNA methylation. Although the physiologic and pathologic functions of ANGPTL2 remain to be discovered, epigenetic silencing by hypermethylation of the ANGPTL2 promoter may lead to a loss of ANGPTL2 function and affect hematopoietic stem cells in the bone marrow. ANGPTL2 may function differently in leukemic and myelodysplastic cells.

Ectopic expression of BIK induces apoptosis in numerous tumor types, including colon, prostate, breast, melanoma, and glioma cells [19-21]. BIK interacts with Bcl-xL and Bcl-2 and triggers apoptosis in a Bak-dependent manner [20]. BIK may be a tumor suppressor gene that is downregulated by hypermethylation or histone deacetylation of the promoter [22, 23]. In this study, a methylated allele of BIK was observed in 2 samples of RAEB, but not in any RCMD samples. In other studies, increased apoptosis has been documented within CD34⁺ bone marrow cells in patients with MDS [24]. Increased apoptosis appears to be the highest in patients with refractory anemia and is associated with early stage MDS [25, 26]. Although methylation of the BIK gene was detected less frequently in RAEB patients, it is possible that BIK expression may be downregulated by promoter hypermethylation. Decreased levels of apoptosis in advanced MDS may be associated with BIK expression.

RERG is a novel gene that was first reported in breast cancer following microarray analysis [27]. Gender-dependent expression of RERG in hepatocellular carcinoma and its regulation by histone deacetyltransferases (HDACs) have been reported previously [28]. In this study, the RERG gene was selected as a different candidate gene between RCMD and RAEB samples, but no specific pattern was noted upon MSP analysis. The Ras family has been studied in MDS and leukemogenesis, and some of these mutations may influence the course of disease. Therefore, the RERG gene may be associated with myeloid disorder and regulated by multiple pathways, including the HDAC-related signaling pathway.

The methylated primers, which were biased to detect methylated alleles, detected alleles that were 100% hypermethylated. Because MSP analysis allows for only a small number of individual CpG dinucleotides to be sampled, a negative result does not exclude methylation elsewhere in the CpG island [29]. Thus, it was necessary to conduct bisulphate sequencing analysis of the CpG island to determine the methylation status of each gene.

In this study, the whole-genome-wide DNA microarray technique and IPA were used to determine the differing states of DNA methylation between RAEB and RCMD; 16 genes were shown to exhibit increased hypermethylation in RAEB samples compared to RCMD samples. Among the 16 genes, GSTM5, ANGPTL2, and BIK promoters exhibited methylation more frequently in RAEB than in RCMD samples. Although the functions of these genes remain unknown, their state of DNA methylation may induce epigenetic silencing and contribute to increasing blasts, which results in the disease progression of MDS. Additional studies focusing on epigenetic silencing by various detection modalities are required.

Cluster pattern for all microarray gene data. Abbreviations: RCMD, refractory cytopenia with multilineage dysplasia; RAEB, refractory anemia with excess blast.

Hierarchical clustering analysis. No noteworthy hypermethylated gene cluster patterns were identified.

Network 1 gene expression, cancer, and cell cycle. The 16 candidate genes are denoted in red and burrowed with white genes to form a disease network construct.

Methylation-specific PCR results on GSTM5, ANGPTL2, BIK, and RERG genes from bone marrow samples. Patients were designated numbers 1-10 (RCMD) and 11-19 (RAEB). The PCR products were analyzed by electrophoresis on a 2% agarose gel. Abbreviations: N, normal peripheral blood; H, Human leukemic HL-60 cells; M, amplified products used as primers for the methylated sequence; U, amplified products used as primers for the unmethylated sequence.

Table 1 Methylated allele primer sequences and optimized annealing temperatures for methylation-specific PCR.

Table 2 Patient characteristics (N=20).

Table 3 Top networks based on IPA analysis.

Table 4 Statistically different gene methylation in network 1.