Purpose: Epigenetic regulation by promoter methylation plays a key role in tumorigenesis. Our goal was to investigate whether altered DNA methylation signatures associated with oncogenic signaling delineate biomarkers predictive of endometrial cancer recurrence.

Experimental Design: Methyl-CpG-capture sequencing was used for global screening of aberrant DNA methylation in our endometrial cancer cohort, followed by validation in an independent The Cancer Genome Atlas (TCGA) cohort. Bioinformatics as well as functional analyses in vitro, using RNA interference (RNAi) knockdown, were performed to examine regulatory mechanisms of candidate gene expression and contribution to aggressive phenotype, such as epithelial–mesenchymal transition (EMT).

Results: We identified 2,302 hypermethylated loci in endometrial tumors compared with control samples. Bone morphogenetic protein (BMP) family genes, including BMP1, 2, 3, 4, and 7, were among the frequently hypermethylated loci. Interestingly, BMP2, 3, 4, and 7 were less methylated in primary tumors with subsequent recurrence and in patients with shorter disease-free interval compared with nonrecurrent tumors, which was validated and associated with poor survival in the TCGA cohort (BMP4, P = 0.009; BMP7, P = 0.007). Stimulation of endometrial cancer cells with epidermal growth factor (EGF) induced EMT and transcriptional activation of these genes, which was mediated by the epithelial cell adhesion molecule (EpCAM). EGF signaling was implicated in maintaining the promoters of candidate BMP genes in an active chromatin configuration and thus subject to transcriptional activation.

Conclusions: Hypomethylation signatures of candidate BMP genes associated with EpCAM-mediated expression present putative biomarkers predictive of poor survival in endometrial cancer. Clin Cancer Res; 19(22); 6272–85. ©2013 AACR.

Translational Relevance

Although the majority of endometrial tumors are confined to the uterus, approximately 20% of patients have identifiable metastases or occult disease that ultimately recurs within 12 to 36 months of the original diagnoses. To identify DNA methylation signatures that can be used as prognostic markers for recurrence, we performed genome-wide DNA methylation analysis in an endometrial cancer cohort. We identified a subset of the bone morphogenetic protein (BMP) gene family with low promoter methylation in primary tumors with subsequent recurrence compared with nonrecurrent tumors. These findings were validated in The Cancer Genome Atlas (TCGA) endometrial cancer cohort. Our studies in endometrial cancer cells revealed that these BMP genes are regulated by EGF, an inducer of epithelial–mesenchymal transition (EMT), and epithelial cell adhesion molecule (EpCAM), which maintains the BMP promoters in a transcriptionally permissive state. Thus, the hypomethylation status enabling transcriptional activity of candidate EpCAM-regulated BMP genes may delineate a set of putative biomarkers associated with poor prognosis of endometrial cancer.

Endometrial cancer is the leading gynecologic malignancy in new cases and its incidence is on the rise (1). It is usually confined to the inner lining of the uterus and can be removed by hysterectomy. Unfortunately, about 20% of patients have identifiable metastases or occult disease that ultimately recur within 12 to 36 months of original diagnosis. Recurrence has been implicated with high morbidity of patients with endometrial cancer (2–4). The underlying molecular mechanisms that regulate progression of endometrial cancer recurrence remain unclear. In addition, there is an unmet need to develop biomarkers predictive of recurrence and effective therapy to reduce the death rate associated with endometrial cancer.

Our previous studies and those of others suggest that gene promoter CpG islands epigenetically marked by de novo DNA methylation may serve as biomarkers in endometrial cancer and other malignancies (5–9). These epigenetic changes work hand-in-hand with histone modifications to repress transcription of genes encoding proteins for tumor suppression, cell-cycle regulation, and DNA repair (5, 10–14). There is increasing evidence that differential methylation of the same loci can be seen in different tumors, highlighting the dynamics of epigenetic plasticity during cancer development (15, 16). Because aberrant methylation patterns are stable and can inherently be transmitted from parental to daughter cells, methylation signatures acquired in candidate loci during tumorigenesis may be used as diagnostic or prognostic markers for cancer.

In particular, tumor suppressor genes are highly susceptible to increased promoter methylation (i.e., hypermethylation) in tumors compared with normal tissue. DNA hypermethylation of these loci renders the chromatin configuration in a nonpermissive state, leading to gene silencing. This transcriptional inactivation results in loss of tumor-suppressive properties that in part drives tumorigenesis (17–20). Another class of differential promoter methylation in cancer relative to normal tissue is a decreased methylation signature (i.e., hypomethylation) in tumors. Such hypomethylation targets oncogenes, leading to a permissive chromatin configuration and subsequent transcriptional induction in tumors (21). These oncogenes would be otherwise silenced in differentiated normal tissue. As an example, promoter hypomethylation of cancer/testis antigen genes, CAGE and MAGE, in tumors (relative to normal tissue) has been attributed to its derepression (22).

In this study, we address yet another class of differential methylation that targets specific loci during malignant progression, specifically comparing recurrent and nonrecurrent endometrial tumors. For this purpose, we used methyl-CpG-capture sequencing to screen global DNA methylation in primary endometrial tumors and normal endometrium. This endometrial cancer cohort consisted of primary endometrial tumors with subsequent recurrence (n = 17) and nonrecurrent (n = 50) tumors. Our data identified DNA hypomethylation in a subset of bone morphogenetic protein (BMP) genes in endometrial tumors with recurrence compared with nonrecurrent tumors. The molecular mechanisms underlying aberrant expression of BMP genes have been explored, with increased evidence pointing to epigenetic regulation (23–26). Several studies suggest that BMP genes are susceptible to epigenetic silencing by promoter hypermethylation in cancer (24–26). For example, the BMP2 and 6 promoters were found to be hypermethylated in gastric tumors and multiple myeloma, respectively (25, 27). On the other hand, induction of BMP2, 4, and 7 observed in different cancers associated with aggressive growth and metastasis raises the question how these genes are expressed during malignant progression although their loci could be highly susceptible to DNA methylation during tumor development (28–30). We hypothesize that oncogenic transcriptional activities at the promoters of BMP genes maintain these loci in an open chromatin configuration with activating histone marks, thereby decreasing the levels of DNA methylation.

Our goal in this study was to evaluate BMP promoter methylation and gene regulation in endometrial cancer. Specifically, we investigated whether BMP promoter hypomethylation distinguishes between primary endometrial tumors with subsequent recurrence and those that are nonrecurrent, thereby identifying these loci as putative biomarkers for recurrence and survival. To determine regulatory and epigenetic mechanisms involved in BMP gene expression, functional analysis was carried out by focusing on the EGF and epithelial cell adhesion molecule (EpCAM) signaling pathways implicated with oncogenesis (31, 32).

DNA samples

DNA samples of 67 primary endometrial tumors and 10 normal controls were obtained from a previously published endometrial cancer cohort following approval of Institutional Review Board committee (Supplementary Table S1; ref. 6). Clinical staging and tumor grade was assigned on the basis of Federation Internationale des Gynaecologistes et Obstetristes (FIGO) 1988 criteria. Tumor specimens had high neoplastic cellularity (mean 74%; median 80%), whereas normal tissue did not contain any malignant portion by direct microscopic visualization. All cancer samples tested are primary endometrial tumors (Supplementary Table S1). Recurrence is defined as a return of cancer in primary sites or distant metastasis within 3 years of initial diagnosis.

Cell culture

Endometrial cancer cell lines, RL95-2, AN3CA, and HEC1A, were obtained from American Type Culture Collection (ATCC) and grown in culture media based on ATCC recommendations. These cells were grown in Dulbecco's modified Eagle medium (DMEM) culture medium (Life Technologies) supplemented with 10% FBS and 1% penicillin/streptomycin (Life Technologies) at 37°C. Media were supplemented with 10% charcoal-stripped heat-inactivated (CSHI) FBS for treatments with EGF (Peprotech). IRESSA (gefitinib) was purchased from Tocris Biosciences. Normal primary endometrial cells were generously supplied by Dr. Robert Schenken (Department of OB/GYN, University of Texas Health Science Center at San Antonio, San Antonio, TX). To assess the role of DNA methylation on BMP expression, RL95-2 cells were treated with 5-aza-2′-deoxycytidine (DAC; 0.5 μmol/L; Sigma) for 48 hours.

Methyl-capture sequencing

Methylated DNA was eluted by the MethylMiner Methylated DNA Enrichment Kit (Life Technologies) following the manufacturer's instructions. Briefly, genomic DNA sonicated to approximately 300 bp was captured by MBD2 proteins and eluted in 1 mol/L salt buffer for precipitation. Eluted DNA (>10 ng) was used to generate libraries for sequencing following the standard protocols from Illumina. MBDCap-seq libraries were sequenced using the Illumina GAII system as per the manufacturer's instructions. Image analysis and base calling were performed with the standard Illumina pipeline. Using the ELAND algorithm, unique reads (up to 50-bp reads) were mapped to the human reference genome (hg18), with up to two mismatches (see Supplementary Table S2). Bioinformatics analysis is described in detail in Supplementary Data. We used paired t test (P < 0.05) to determine whether the methylation level of the regions in question differed significantly between these two groups. Differentially methylated loci between endometrial tumor samples and normal controls as well as between recurrent and nonrecurrent tumor groups were analyzed by the algorithm outlined in Supplementary Data. Methodology for Steiner-tree network analysis is described in Supplementary Data.

Pyrosequencing analysis

DNA (500 ng/sample) was processed for bisulfite conversion using the EZ DNA Methylation Kit (Zymo Research). Primers used to amplify specific CpG island regions are listed in Supplementary Data. Methylated CpG sites were detected by the PyroMark Q96 MD system (Qiagen). Incomplete bisulfite conversion checkpoint was set as 5%. The methylation percentage of each interrogated CpG site was calculated and visualized using MultiExperiment Viewer v4.8 (Dana-Farber Cancer Institute, Boston, MA) and GraphPad Prism (GraphPad Software).

siRNA/shRNA knockdown

The lentiviral plasmid vectors harboring specific short hairpin RNA (shRNA) sequences for human EpCAM and nontarget scrambled control were purchased from Sigma-Aldrich. Sequences for the EpCAM shRNA are outlined in Supplementary Data. HEK293T cells were seeded in Opti-MEM medium (Life Technologies) 1 day before virus production. Plasmids (shRNA EpCAM or nontarget scrambled control) were mixed with lentivirus package plasmids (psPAX2 and pMD2G; a gift from Dr. Rong Li, Department of Molecular Medicine, Institute of Biotechnology, University of Texas Health Science Center at San Antonio, San Antonio, TX) with Lipofectamine 2000 (Life Technologies) and added to HEK293T cells for 24 hours followed by change to RPMI medium (Life Technologies) for another 24 hours of incubation. Virus-containing medium was filtered with 0.45-μm filter (Thermo Fisher Scientific). Viral particles were mixed with polybrene (Millipore) to transduce RL95-2 cells. These RL95-2 cells were then cultured in DMEM medium with 10% FBS, 1% penicillin/streptomycin, and 1 μg/mL puromycin (Life Technologies) at 37°C. Stable clones were cultured for 2 months under puromycin selection and knockdown of EpCAM expression level was determined by both reverse transcriptase PCR (RT-PCR) and Western blotting.

For transient transfection, siRNAs specific to human BMP4, BMP7, and nontarget scrambled control were purchased from Thermo Fisher Scientific. RL95-2 cells were seeded 1 day before transfection in DMEM medium with 10% FBS. siRNAs were mixed with Lipofectamine 2000 in Opti-MEM medium (Life Technologies) and added to RL95-2 cells for a 6-hour incubation followed by changing to DMEM culture medium with 10% FBS. Transfected cells were harvested after 48 hours, and the expression levels of BMP4 and 7 were checked by RT-PCR.

RT-PCR analysis

RNA was isolated from treated and control cells for RT-PCR analysis. Real-time PCR was performed with SuperScript III RT (Life Technologies) using the StepOnePlus Real-Time PCR Systems (Life Technologies). To examine the effects of EGF on the expression of epithelial–mesenchymal transition (EMT)–related genes, the Biomark system (Fluidigm) was used. cDNA (48 wells, containing experimental replicates), PCR primers (48 sets, including housekeeping genes UBB and GAPDH), and reaction mixture were loaded in the Fluidigm 48.48 microfluidic chip (Fluidigm dynamic array integrated fluidic circuits) and subjected to PCR in the Biomark system. The ΔΔCt was calculated for each gene using those of GAPDH and UBB for normalization. Primer sequences for RT-PCR assays are listed in Supplementary Data.

Protein analysis

Subcellular protein fractions (membrane and nuclear extracts) were prepared with the subcellular protein fractionation kit for cultured cell as directed by the manufacturer's manual (Thermo Fisher Scientific). Western blot analysis was performed as described previously (33). Briefly, 40 to 60 μg of protein was loaded on 10% Bis–Tris polyacrylamide gels under denaturing/reducing conditions and then transferred to polyvinylidene difluoride (PVDF) membranes in an iBlot electroblotter as per the manufacturer's recommended conditions (Life Technologies). After blocking with TBS containing 0.1% Tween (TBST) and 5% milk for 1 hour and then washing with TBST, membranes were incubated with primary antibody diluted in TBST (5% bovine serum albumin) for 1 hour, 2 hours, or overnight (depending on antibody). Protein bands were visualized after 1 hour of incubation with secondary antibody (conjugated to horseradish peroxidase) in TBST (5% milk) followed by exposure to the ECL-plus detection system and X-ray autoradiography (GE Healthcare; Amersham). Antibody for EpCAM (clone E44; Abcam) detects the C-terminus of EpCAM, which corresponds to its intracellular domain, EpICD. Lamin B and α-tubulin were purchased from Santa Cruz Biotechnology. Antibodies for Na+/K+-ATPase and E-cadherin were obtained from Cell Signaling Technology and BD Bioscience, respectively. Immunofluorescence with E-cadherin antibody (BD Biosciences) was performed as described previously (33).

Chromatin immunoprecipitation-PCR

Chromatin immunoprecipitation-PCR (ChIP-PCR) was carried out as previously described (34). Briefly, cells were fixed with 1% formaldehyde, and DNA cross-linked to protein complexes was sonicated to approximately 300 to 500 bp fragments. Immunoprecipitation was carried out using Dynabeads protein G (Life Sciences) coated with antibodies for H3K9/14 ac (Diagenode), H3K27 me3 (Diagenode), RNA polymerase II (Covance), EpCAM (Santa Cruz Biotechnology), and control immunoglobulin G (IgG; Diagenode). Coprecipitated DNA was used for PCR assays. PCR primers (listed in Supplementary Data) targeting regions upstream of transcriptional start site (TSS) for BMP target genes were used in PCR mixture containing SYBR Green (Life Technologies) to amplify the DNA. The $2^{{\rm \Delta \Delta}C_{\rm t}}$ values relative to the input DNA were calculated.

Cell invasion assay

Endometrial cancer cells (~50,000) were seeded onto the top insert (layered with Matrigel) of an invasion chamber (BD Biosciences). The invasion chambers were then incubated at 37°C in 5% CO2 for 20 hours. Cells that did not invade through the Matrigel, those on the upper surface of the insert membrane, were mechanically removed with cotton tip applicators and several washes with PBS. Invaded cells on the bottom of the coated membranes were visualized using a fluorescence microscope with a ×20 objective after incubation with Hoechst stain (Life Technologies). Images were obtained from four standardized, nonoverlapping fields. Invaded cells were counted using the ImageJ software (http://rsbweb.nih.gov/ij/). Invasion assays were performed in triplicates and images of four fields per well (covering about 85% of the well) were taken for counting of invaded cells.

Statistical analysis

Student t test was used to compare pyrosequencing, RT-PCR, and invasion assay results in different treatment and control groups. Statistical significance was assigned as *, P < 0.05; **, P < 0.01; or ***, P < 0.001.

Methyl-capture screening identifies differentially methylated CpG islands in primary endometrial tumors

DNA hypermethylation of promoter CpG islands has previously been reported in more than 400 transcriptionally silenced genes in cancer cells (35). Although this epigenetic silencing is usually linked to tumor initiation, it is unclear how differential DNA methylation plays a role in the progression of endometrial cancer. Our goal was to investigate whether aberrantly methylated genes are subject to regulation by oncogenic pathways that promote endometrial cancer recurrence. For this purpose, we used methyl-capture sequencing (MBDCap-seq) to survey differentially methylated CpG islands in 67 primary endometrial tumors relative to 10 normal controls (i.e., uninvolved endometrium or myometrium; Supplementary Table S1). Approximately, 5 billion sequence reads were processed, 49.3% of which were mapped to unique genome locations (Supplementary Fig. S1A and Supplementary Table S2). Because DNA methylation is expected to occur in GC-rich regions, the coverage of 20 million unique reads was reported to provide sufficient sequence depth for comprehensive methylation mapping of a genome (36, 37). On the basis of this estimate, we calculated that on average more than 85% of all promoter regions and more than 79% of the whole genome were covered by at least one unique sequence read in the endometrial MBDCap-seq dataset (Supplementary Fig. S1B). Technical repeats of tumor samples and validation analysis of candidate genes were conducted, and the results suggest that MBDCap-seq can reliably identify hypermethylated regions in tumors relative to control samples (Supplementary Figs. S2 and S3).

Of 13,081 promoter CpG islands analyzed by pair comparison between tumors and normal controls based on Student t test, we found that 2,302 (17.6%) loci were hypermethylated in endometrial tumors (P < 0.05; Fig. 1A, left). DNA hypermethylation was frequently observed in the TSS regions, or CpG island cores, of candidate genes (Fig. 1A, middle or representative examples shown in the right). This aberrant event also occurred in flanking regions of TSSs, or so-called CpG shore regions previously reported in the literature (20, 38). Furthermore, hypermethylation was detected in inter- and intragenic CpG islands and non-CpG island promoters (Supplementary Fig. S4), which is consistent with previous observations that DNA methylation is not limited to promoter CpG islands in tumors (10, 20, 39).

Figure 1.

MBDCap-seq of endometrial carcinomas and normal endometrial tissue reveal differential DNA methylation between recurrent and nonrecurrent tumors. A, differential methylation is more evident in cores of CpG islands. Differentially methylated loci (n = 2,302; corresponding to 17.6% of CpG islands tested) are shown in a pie chart; the majority of differentiation is evident in CpG island cores (left). Average methylation of differentially methylated loci in CpG island cores, right shores, left shores, and both shores present in normal and tumor samples (middle) and representative examples (right). B, average DNA methylation of 1,082 loci that seem to be less methylated in primary tumors with subsequent recurrence compared with nonrecurrent tumors. Low-methylation of individual loci is shown in (left). Average methylation is shown for core, left shore, right shore, and both shores of CpG islands (right).

Figure 1.

MBDCap-seq of endometrial carcinomas and normal endometrial tissue reveal differential DNA methylation between recurrent and nonrecurrent tumors. A, differential methylation is more evident in cores of CpG islands. Differentially methylated loci (n = 2,302; corresponding to 17.6% of CpG islands tested) are shown in a pie chart; the majority of differentiation is evident in CpG island cores (left). Average methylation of differentially methylated loci in CpG island cores, right shores, left shores, and both shores present in normal and tumor samples (middle) and representative examples (right). B, average DNA methylation of 1,082 loci that seem to be less methylated in primary tumors with subsequent recurrence compared with nonrecurrent tumors. Low-methylation of individual loci is shown in (left). Average methylation is shown for core, left shore, right shore, and both shores of CpG islands (right).

Close modal

Promoter hypomethylation found in a subset of BMP loci is associated with tumor recurrence and decreased disease-free survival

To identify genes prone to transcriptional deregulation during aggressive tumor growth, we compared DNA methylation levels of the aforementioned 2,302-hypermethylated loci in 50 primary nonrecurrent versus 17 primary endometrial tumors with subsequent recurrence by pairwise comparison using t test analysis. More than half (n = 1,232) of these loci displayed differential methylation between the two groups of tumors (P < 0.05). Interestingly, 1,082 of the loci appeared less methylated in the recurrent group compared with the nonrecurrent group (P < 0.05; Fig. 1B; Supplementary Table S3). Using a Steiner-tree computational algorithm (40), we determined the network topology of these 1,082 loci using the Human Protein Reference Database (Release 9; ref. 41). These low-methylated loci were first used as seeds to identify connector loci that can link the seed genes to a connected sub-network. Although connectors are not necessarily methylated themselves, these loci and their associated networks were significantly enriched for pathways involved in aggressive growth and metastasis, including EGF receptor (EGFR; P < 10−12, Fisher exact test), mitogen—activated protein kinase (MAPK; P < 10−10), Wnt (P < 10−8), and Gap junction (P < 10−12) pathways (Supplementary Fig. S5). At least 39 low-methylated loci and connectors were identified in this interactive network, including BMP, ephrin/ephrin receptor (EPH), and cadherin (CDH) gene families known to encode proteins linked to EMT (Supplementary Fig. S5; refs. 42–45).

In this study, we focused on a subset of BMP genes that were differentially methylated in primary tumors with subsequent recurrence compared with nonrecurrent endometrial tumors. To begin to understand BMP regulation in endometrial cancer, we surveyed the DNA methylation landscape in an 8-kb region surrounding the TSS of the BMP loci. BMP1, 2, 3, 4, and 7 were frequently hypermethylated in endometrial cancer, whereas BMP6 exhibited hypomethylation in endometrial tumors compared with normal control (Fig. 2A, differentially methylated regions are designated by the dashed-line squares). Differential methylation between tumor and normal samples was mostly seen in CpG islands of BMP2, 3, 4, and 7 (Fig. 2A, designated by the dashed-line squares). However, differential methylation in BMP1 and 6 was further downstream of the TSS (Fig. 2A). We also found that when comparing primary endometrial tumors with recurrence to tumors without recurrence, BMP2, 3, 4, and 7 were significantly hypomethylated in the recurrence group (P = 0.006, P < 0.001, P < 0.001, P = 0.04, respectively; Fig. 2B). Attesting to the role of BMP2, 3, 4, and 7 in aggressive tumor growth, their hypomethylation status was associated with poor survival in our endometrial cancer cohort (P = 0.02; Fig. 2C). Progression-free survival analysis was conducted by using the third quantile as the cutoff value to dichotomize patients into high- and low-methylation groups. Kaplan–Meier curve analysis of progression-free survival with combined BMP2, 3, 4, and 7 showed that patients with higher number of “high-methylated” BMP genes exhibit better survival (Fig. 2C). In this analysis, 92.3% survival rate of patients with more than one high-methylated gene is significantly higher than 65.1% survival rate of patients with zero or one high-methylated gene (P = 0.02; Fig. 2C). Patients with more than one high-methylated gene showed significant prolonged survival by univariate Cox regression analysis [HR, 0.13; 95% confidence interval (CI), 0.018–0.977].

Figure 2.

Differentially methylated members of the BMP families and high methylation in BMP loci give better survival. A, DNA methylation of BMP family is shown for normal endometrium (N), nonrecurrent tumors (NR), and tumors with recurrence (R). Dashed-line squares highlight differentially methylated regions in the BMP loci between normal and tumor endometrial samples. B, average methylation of BMP candidate loci spanning 4-kb upstream and downstream the TSS. C, Kaplan–Meier curve of progression-free survival analysis of BMP2, 3, 4, and 7 combined. Dotted line represents a group of patients with less than one high-methylated gene among the four BMP loci. The solid line represents a group of patients with more than two high-methylated BMP loci.

Figure 2.

Differentially methylated members of the BMP families and high methylation in BMP loci give better survival. A, DNA methylation of BMP family is shown for normal endometrium (N), nonrecurrent tumors (NR), and tumors with recurrence (R). Dashed-line squares highlight differentially methylated regions in the BMP loci between normal and tumor endometrial samples. B, average methylation of BMP candidate loci spanning 4-kb upstream and downstream the TSS. C, Kaplan–Meier curve of progression-free survival analysis of BMP2, 3, 4, and 7 combined. Dotted line represents a group of patients with less than one high-methylated gene among the four BMP loci. The solid line represents a group of patients with more than two high-methylated BMP loci.

Close modal

To validate hypomethylated BMP2, 3, 4, and 7 in recurrent compared with nonrecurrent tumors, we performed in silico analysis of DNA methylation panels of The Cancer Genome Atlas (TCGA) database (http://tcga-data.nci.nih.gov/tcga/; more detail is provided in Supplementary Methods) endometrial cancer cohort (46). These BMP loci were significantly hypomethylated in primary endometrial tumors with recurrence/progression (Fig. 3A; P < 0.05 for BMP2, 3, and 7; and P < 0.01 for BMP4). Furthermore, higher levels of DNA methylation of BMP4 and 7 were significantly associated with increased survival (Fig. 3B; P < 0.01). Because only DNA samples were available from our cohort (6), we conducted in silico analysis to determine the relationship between DNA hypermethylation and gene silencing using the endometrial cancer cohort from TCGA. Taking median expression in normal samples as a cutoff, we divided the endometrial cancer samples to high expression and low expression groups for each of BMP2, 3, 4, and 7 (Fig. 4A). We observed overall that samples exhibiting higher DNA methylation corresponded to lower BMP expression, which was statistically significant for BMP4 and 7 (Fig. 4A and B; P < 0.05 and P < 0.001, respectively). We also observed an association between lower DNA methylation levels and higher clinical stages (stages III/IV compared with I/II), as shown in Supplementary Fig. S6.

Figure 3.

DNA methylation distribution and survival analysis of BMP genes in the TCGA cohort. A, top, average DNA methylation level of BMP2, 3, 4, and 7 across 2-kb region of TSS or differentially methylated regions near TSS in all endometrial tumor samples. Map of TSS location and CpG island (green) is shown for each of the BMP above corresponding graphs. Vertical lines indicate CpG probes used in the TCGA analysis. Bottom, representative scatter plots showing the distribution of DNA methylation level in each patient within the recurrence (red) and nonrecurrence (blue) groups in one representative probe. *, P < 0.05 or **, P < 0.01. B, Kaplan–Meier curves of progression-free survival analysis of BMP4 and 7. Solid line represents high methylation group and the dotted line represents low-methylation group. The low methylation of BMP4 and 7 showed significant association with poor survival (P < 0.01).

Figure 3.

DNA methylation distribution and survival analysis of BMP genes in the TCGA cohort. A, top, average DNA methylation level of BMP2, 3, 4, and 7 across 2-kb region of TSS or differentially methylated regions near TSS in all endometrial tumor samples. Map of TSS location and CpG island (green) is shown for each of the BMP above corresponding graphs. Vertical lines indicate CpG probes used in the TCGA analysis. Bottom, representative scatter plots showing the distribution of DNA methylation level in each patient within the recurrence (red) and nonrecurrence (blue) groups in one representative probe. *, P < 0.05 or **, P < 0.01. B, Kaplan–Meier curves of progression-free survival analysis of BMP4 and 7. Solid line represents high methylation group and the dotted line represents low-methylation group. The low methylation of BMP4 and 7 showed significant association with poor survival (P < 0.01).

Close modal
Figure 4.

Higher methylation levels correspond to lower expression of BMP2, 3, 4, and 7 in the TCGA endometrial cancer cohort. In silico analysis was performed to determine whether methylation levels correspond to expression of the four BMP genes in the TCGA endometrial cancer database. A, on the basis of median BMP expression of normal endometrium as a cutoff (top), tumor samples were divided into “high” and “low” expressing groups for each of the BMP genes (bottom). Methylation levels were examined at different sites (corresponding to a specific probe) in TSS regions of each BMP gene. Map of TSS location and CpG island (green) is shown for each of the BMPs above the corresponding graphs. Vertical lines indicate CpG probes used in the TCGA analysis. B, average methylation levels of all probes shown in (A) for each BMP gene in “high” and “low” expressing tumor groups is shown in a scatter plot. *, P < 0.05; ***, P < 0.001.

Figure 4.

Higher methylation levels correspond to lower expression of BMP2, 3, 4, and 7 in the TCGA endometrial cancer cohort. In silico analysis was performed to determine whether methylation levels correspond to expression of the four BMP genes in the TCGA endometrial cancer database. A, on the basis of median BMP expression of normal endometrium as a cutoff (top), tumor samples were divided into “high” and “low” expressing groups for each of the BMP genes (bottom). Methylation levels were examined at different sites (corresponding to a specific probe) in TSS regions of each BMP gene. Map of TSS location and CpG island (green) is shown for each of the BMPs above the corresponding graphs. Vertical lines indicate CpG probes used in the TCGA analysis. B, average methylation levels of all probes shown in (A) for each BMP gene in “high” and “low” expressing tumor groups is shown in a scatter plot. *, P < 0.05; ***, P < 0.001.

Close modal

EGF induces expression of BMP2, 3, and 7 and EMT

On the basis of network analysis of the differentially methylated loci described above (Supplementary Fig. S5), we detected enrichment of the EGF/EGFR signaling pathway, which is commonly activated in aggressive endometrial cancer (47, 48). To functionally verify the involvement of EGFR signaling in BMP regulation, we examined whether EGF (10 ng/mL) induces the expression of BMP genes in endometrial cancer, using the endometrial cancer cell lines RL95-2, AN3CA, and HEC1A (Supplementary Fig. S7A). RL95-2 was most responsive exhibiting significant upregulation in BMP2, 3, and 7 (P < 0.05, P < 0.001, and P < 0.01, respectively) but not in BMP4 (Fig. 5A). This induction by EGF was further confirmed by treating RL95-2 cells with the EGFR inhibitor IRESSA (gefitinib) that led to decreasing EGF-induced expression of the BMP genes (Fig. 5A). We also observed decrease in BMP gene expression by IRESSA in the absence of EGF, which could be attributed to blocking EGFR activation by autocrine actions (Fig. 5A). Similar effects were seen when treating these cells with inhibitors to EGFR intracellular effectors (Supplementary Fig. S7B).

Figure 5.

Responsiveness to EGF induction of gene expression and cellular invasion in endometrial cancer cells may partly depend on the methylation levels of BMP loci. A, expression of BMP loci in the presence or absence of EGF and/or EGFR inhibitor IRESSA in RL95-2 endometrial cancer cells. Cells were incubated for 24 hours in growth media supplemented with 10% CSHI FBS before EGF (10 ng/mL) treatment. IRESSA was added at a concentration of 100 nmol/L. Treatment was carried out for 72 hours followed by RT-PCR, which was performed in triplicates. Error bars for RT-PCR data indicate SD. *, P < 0.05 **, P < 0.01; and ***, P < 0.001. B, methylation status of CpG islands and regions surrounding the TSS of BMP2, 3, 4, and 7 (left). Methylation analysis was carried out by pyrosequencing as described in Materials and Methods. Effects of DNA demethylation on BMP expression was determined by DAC (0.5 μmol/L) treatment (right). C, cell invasiveness and EMT phenotype with and without EGF treatment in RL95-2 endometrial cancer cells. EGF treatment was carried out as described above, and its effects on cellular invasion (left), cell morphology (middle), and E-cadherin expression (right) were examined. E-cadherin protein expression was determined by Western blotting (top) and immunocytochemistry (bottom). *, P < 0.05; **, P < 0.01; or ***, P < 0.001.

Figure 5.

Responsiveness to EGF induction of gene expression and cellular invasion in endometrial cancer cells may partly depend on the methylation levels of BMP loci. A, expression of BMP loci in the presence or absence of EGF and/or EGFR inhibitor IRESSA in RL95-2 endometrial cancer cells. Cells were incubated for 24 hours in growth media supplemented with 10% CSHI FBS before EGF (10 ng/mL) treatment. IRESSA was added at a concentration of 100 nmol/L. Treatment was carried out for 72 hours followed by RT-PCR, which was performed in triplicates. Error bars for RT-PCR data indicate SD. *, P < 0.05 **, P < 0.01; and ***, P < 0.001. B, methylation status of CpG islands and regions surrounding the TSS of BMP2, 3, 4, and 7 (left). Methylation analysis was carried out by pyrosequencing as described in Materials and Methods. Effects of DNA demethylation on BMP expression was determined by DAC (0.5 μmol/L) treatment (right). C, cell invasiveness and EMT phenotype with and without EGF treatment in RL95-2 endometrial cancer cells. EGF treatment was carried out as described above, and its effects on cellular invasion (left), cell morphology (middle), and E-cadherin expression (right) were examined. E-cadherin protein expression was determined by Western blotting (top) and immunocytochemistry (bottom). *, P < 0.05; **, P < 0.01; or ***, P < 0.001.

Close modal

Because the expression of BMP4 was not induced by EGF stimulation in RL95-2 cells, we determined whether this nonresponsiveness could be attributed to epigenetic repression. Pyrosequencing analysis revealed extensive methylation in regions surrounding the TSS of BMP4, whereas other responsive loci, BMP2, 3, and 7 showed little or no methylation in TSS regions (Fig. 5B, left). When treating cells with the DNA-demethylating agent DAC (0.5 μmol/L), we observed enhanced expression of BMP4, suggesting a release of methylation-mediated repression on the locus (Fig. 5B, bottom left). Interestingly, increased expression of unmethylated BMP2, 3, and 7 was seen in DAC-treated cells compared with vehicle controls (Fig. 5B). We speculate that in addition to the demethylating effect, DAC treatment resulted in chromatin remodeling by changing histone-modifying protein profiles in the 5′-end of these loci, leading to their transcriptional activation (49, 50).

Upregulation of these BMP loci by EGF can lead to an invasive phenotype of RL95-2 cells in a Matrigel invasion assay (Fig. 5C, left; see also Supplementary Fig. S8 for two other cancer lines). Single or double knockdown of BMP genes attenuated this aggressive behavior (Supplementary Fig. S9A and S9B). This phenotypic alteration is likely mediated by autocrine BMP actions through BMP receptors, BMPR1A, BMPR1B, and BMPR2 expressed in endometrial epithelial cells (Supplementary Fig. S10). Disruption of cell–cell contact and changes in their cobblestone epithelial morphology to a more fibroblastic mesenchymal appearance by EGF was observed in RL95-2 cells (Fig. 5C, middle). This EMT was further confirmed by the reduction of E-cadherin in EGF-treated cells (Fig. 5C, right; ref. 51). Extending this observation, we conducted microfluidic RT-PCR system to simultaneously assess expression profiles of 33 EMT-related genes in RL95-2 cells (Supplementary Fig. S11). The data showed that EGF treatment resulted in a significant increase in the expression of EMT-related transcription factors SNAIL1, TWIST1, FOXC2, TCF3, and ZEB2 and the mesenchymal marker vimentin (Supplementary Fig. S7C, top; refs. 51–53). We also observed decrease in the expression of other mesenchymal markers, suggesting that they are likely not involved in this EMT-mediated induction by EGF (Supplementary Fig. S11, bottom; refs. 51, 54).

EpCAM mediates EGF signaling and maintains activated chromatin of BMP2, 3, and 7

Interestingly, we found that EGF strongly induced the expression of the EpCAM in RL95-2 but not in the other endometrial cancer cells (Fig. 6A). Although EpCAM has been studied mainly as a cell surface marker, it can be cleaved by membrane-bound proteolytic enzymes, resulting in the release of its intracellular domain (EpICD; refs. 31, 32, 55). EpICD subsequently translocates into the nucleus, where it binds target promoters for transcriptional activation (31, 32, 55). Our data show that EGF treatment resulted in increasing the membrane fraction of EpCAM (Fig. 6B, middle). Analysis of nuclear extracts of EGF-treated RL95-2 cells revealed that EGF stimulated EpICD nuclear internalization starting at 12-hour posttreatment, suggesting that EGF induces the translocation of EpICD into the nucleus (Fig. 6B, middle-bottom). Furthermore, ChIP analysis with a C-terminal antibody for EpCAM showed that EGF stimulation enhanced EpICD binding to the promoter regions tested of at least BMP3 and 7 (Fig. 6C, left). Although not statistically significant, we also observed slightly increased EpICD binding to the BMP2 locus (Fig. 6C, left).

Figure 6.

EpICD mediates open chromatin configuration of BMP genes and their transcriptional induction by EGF. A, RNA expression of EpCAM was determined by RT-PCR as described in Materials and Methods. Cells were incubated 24 hours in growth media supplemented with 10% CSHI FBS before EGF (10 ng/mL) treatment. Treatment was carried out for 72 hours. PCR was performed in triplicates. B, left, schematic drawing of EpCAM showing that it consists of two domains, EpEx (N-terminal, extracellular domain) and EpICD (C-terminal, intracellular domain; top). Time-course EGF (10 ng/mL) stimulates EpCAM membrane expression (middle) and EpICD nuclear translocation (bottom) in RL95-2 cells. Na+/K+-ATPase was used as a loading control, Ctrl, of membrane fraction (middle); and lamin B1 was used as a loading control, Ctrl, of nuclear fraction (bottom). EpICD nuclear levels relative to loading Ctrl Lamin B1 was quantified using ImageJ software (http://rsbweb.nih.gov/ij/), showing 0.2 relative expression at 0 hr and 0.1, 0.5, 0.5, and 0.8 relative expression at 6, 12, 24, and 72 hours, respectively, post-EGF treatment (bottom). Right, mRNA expression (top) and protein levels (bottom) of EpCAM-knockdown and scrambled control in RL95-2 cell line. C, graph (left) showing binding of EpICD to the promoters of BMP genes with and without EGF treatment, using ChIP-quantitative PCR (qPCR) assay. Graph (middle) showing recruitment of RNA polymerase II (Pol II), H3K9/14 acetylation, and H3K27-trimethylation to TSS regions in scrambled control or EpCAM shRNA expressing RL95-2 cells, using ChIP-qPCR assay. Graph (right) showing RNA expression of BMP genes with and without EGF in EpCAM-knockdown and scrambled control transfected RL95-2 cells. D, DNA methylation of BMP genes was determined by pyrosequencing of RL95-2 cells expressing scrambled control or EpCAM shRNA. Error bars for RT-PCR and ChIP-PCR data indicate SD. *, P < 0.05; **, P < 0.01; and ***, P < 0.001.

Figure 6.

EpICD mediates open chromatin configuration of BMP genes and their transcriptional induction by EGF. A, RNA expression of EpCAM was determined by RT-PCR as described in Materials and Methods. Cells were incubated 24 hours in growth media supplemented with 10% CSHI FBS before EGF (10 ng/mL) treatment. Treatment was carried out for 72 hours. PCR was performed in triplicates. B, left, schematic drawing of EpCAM showing that it consists of two domains, EpEx (N-terminal, extracellular domain) and EpICD (C-terminal, intracellular domain; top). Time-course EGF (10 ng/mL) stimulates EpCAM membrane expression (middle) and EpICD nuclear translocation (bottom) in RL95-2 cells. Na+/K+-ATPase was used as a loading control, Ctrl, of membrane fraction (middle); and lamin B1 was used as a loading control, Ctrl, of nuclear fraction (bottom). EpICD nuclear levels relative to loading Ctrl Lamin B1 was quantified using ImageJ software (http://rsbweb.nih.gov/ij/), showing 0.2 relative expression at 0 hr and 0.1, 0.5, 0.5, and 0.8 relative expression at 6, 12, 24, and 72 hours, respectively, post-EGF treatment (bottom). Right, mRNA expression (top) and protein levels (bottom) of EpCAM-knockdown and scrambled control in RL95-2 cell line. C, graph (left) showing binding of EpICD to the promoters of BMP genes with and without EGF treatment, using ChIP-quantitative PCR (qPCR) assay. Graph (middle) showing recruitment of RNA polymerase II (Pol II), H3K9/14 acetylation, and H3K27-trimethylation to TSS regions in scrambled control or EpCAM shRNA expressing RL95-2 cells, using ChIP-qPCR assay. Graph (right) showing RNA expression of BMP genes with and without EGF in EpCAM-knockdown and scrambled control transfected RL95-2 cells. D, DNA methylation of BMP genes was determined by pyrosequencing of RL95-2 cells expressing scrambled control or EpCAM shRNA. Error bars for RT-PCR and ChIP-PCR data indicate SD. *, P < 0.05; **, P < 0.01; and ***, P < 0.001.

Close modal

To examine the role of EpCAM in BMP gene regulation, we used specific EpCAM shRNA sequences to stably knockdown its expression in RL95-2 cells. Several EpCAM-specific shRNA sequences were evaluated, two of which (clones 1 and 2) showed at least 80% knockdown at the RNA levels (Supplementary Fig. S12). EpCAM knockdown at the protein level was confirmed for shRNA clone 1 (Fig. 6B, right). We then examined whether EpCAM affects the occupancy of RNA polymerase II on the promoters of BMP2, 3, and 7 by ChIP-PCR. A significant decrease in RNA polymerase II occupancy was observed in the TSS sites of BMP2 and 3, but not BMP7 in RL95-2 cells expressing EpCAM shRNA (Fig. 6C, middle). We also investigated the role of EpCAM in epigenetic modification of the promoters of BMP2, 3, and 7 (Fig. 6C, middle). RL95-2 cells stably expressing EpCAM shRNA or scrambled control shRNA were subjected to ChIP analysis using antibodies directed against the active histone mark H3K9/14Ac and the repressive mark H3K27me3. A significant decrease in the presence of H3K9/14Ac due to EpCAM knockdown was observed in the TSS of BMP3 and 7 (P < 0.01 and 0.05, respectively) but not BMP2 (Fig. 6C, middle). A 6-fold increase in H3K27me3 was observed in the TSS of BMP7 but not BMP2 or 3 (Fig. 6C, middle).

To confirm the role of EpCAM in mediating EGF-induced BMP2, 3, and 7 expression, we examined the response to EGF in RL95-2 cells expressing EpCAM shRNA or scrambled shRNA. The EGF-induced expression of these BMP genes observed in the scrambled control RL95-2 cells was attenuated in the EpCAM knockdown cells (Fig. 6C, right). We then examined whether EpCAM also modulates DNA methylation of promoter CpG islands for the BMP2, 3, and 7 loci. We observed that long-term (2 months) stable EpCAM knockdown leads to a slight increase in DNA methylation of CpG island sites of BMP2 and 7, suggesting that EpICD occupancy of these sites may in part block DNA methylation (Fig. 6D). Combined, the data suggest that EGF-regulated EpCAM not only mediates the induction of BMP expression but also maintains an active chromatin configuration in the target loci.

The BMP gene family is known to have an extensive range of actions in cancer (56). It has been suggested that BMPs function as promalignant growth factors for advanced tumor development (29, 57, 58). For example, aberrant expression of BMP2 and 4 induced motility and invasiveness in prostate and gastric cancer cells (59, 60). Activated BMPs also play a role in regulating EMT, a process in which epithelial cancer cells acquire a more fibroblastic phenotype, enhanced motility, and metastatic potential (61). Interestingly, promoter hypermethylation of BMP genes has been observed in several types of cancer, indicating that this methylation-mediated silencing may actually render a less aggressive phenotype in tumor cells (25, 27).

In line with this prior observation, our present study indicates that promoter hypermethylation of five of 11 BMP genes (i.e., BMP1, 2, 3, 4, and 7) was a frequent event in primary endometrial tumors relative to normal controls. Notably, hypomethylation (or less methylation) was detected in a subset of primary tumors that recurred within 3 years after the diagnosis, whereas the same loci were hypermethylated in nonrecurrent tumors. A recent study by the TCGA network indicated that low DNA methylation in tumors was associated with high-grade and aggressive endometrial cancer (46). On the basis of our evidence, we suggest that aberrantly methylated BMP2, 3, 4, and 7 can be used as prognostic markers in endometrial cancer, and their hypomethylation is predictive of recurrence and/or tumor progression.

Given that recurrence is associated with more aggressive disease, the presence of active chromatin configuration of BMP2, 3, 4, and 7 is a distinct possibility in this subset of primary tumors. We speculate that their induction by oncogenic signaling contributes to malignant development in recurrent tumors. Such oncogenic signaling would be absent in normal tissue, leaving these BMP loci in a low transcriptional state despite low or no methylation in their TSSs. This speculation stems from our in vitro evidence implicating EGF/EGFR signaling in BMP induction as well as from previous findings demonstrating predominant occurrence of this oncogenic signaling in aggressive endometrial cancer (48, 62–64). However, some of these BMP genes (e.g., BMP4 in RL95-2 cells; see Fig. 5B) were less responsive to EGF stimulation. This is likely attributed to promoter hypermethylation that prevents the binding of transcription factors to BMP target loci. Interestingly, our mechanistic studies revealed that EpCAM plays an important role in this transcriptional regulation. EGF oncogenic signaling not only upregulated EpCAM expression but also induced nuclear translocation of its intracellular domain EpICD, and thus enhanced binding of the EpICD transcription factor to BMP promoters for gene expression. Of note, EpICD-mediated transcriptional activation of candidate metastasis-promoting genes (e.g., MYC and cyclin E) has previously been implicated for malignant growth (55, 65).

Combining the data presented in this study, we propose a model for the EpCAM-mediated transcriptional regulation of BMP genes (Supplementary Fig. S13). In the absence of EpICD binding, transcriptionally inactive BMPs can subsequently acquire DNA methylation in their CpG islands for stable repression, attributed to frequent overexpression of DNA methyltransferases and polycomb repressor complex 2 in cancer cells (39, 66). In aggressive cells, activation of EGF signaling leads to EpICD binding to specific BMP targets. By blocking access to DNA methyltransferase, EpICD occupancy may lead to diminished accumulation of DNA methylation in target promoters. Another possibility is that EpICD is involved in removing DNA methylation; however, such a mechanism cannot be confirmed by our present data. In addition, histone modifications of BMP promoters by EpICD, favoring an open transcriptional configuration, is suggested by our data. EPICD promoter interaction leads to increase in the levels of permissive histone marks while decreasing repressive marks. We suggest that enrichment of EpICD occupancy on BMP promoters by EGF signaling leads to transcriptional induction of the corresponding genes, which then contributes to tumor recurrence. Surrounding regions unprotected by EpICD and other transcription factors may still be susceptible to DNA methylation.

In conclusion, our data suggest that transcriptional induction of hypomethylated BMP2, 3, 4, and 7 mediates aggressive growth, associated with EMT, in endometrial cancer cells. Future correlative studies between low-level DNA methylation and transcription upregulation of BMPs are needed to confirm their oncogenic roles in promoting malignant development. When independently validated in a large cohort, these BMP loci may be verified to be putative markers for predicting endometrial cancer recurrence.

No potential conflicts of interest were disclosed.

Conception and design: Y.-W. Huang, N.B. Kirma, T.H.-M. Huang

Development of methodology: Y.-T. Hsu, F. Gu, Y.-W. Huang, J. Liu, N.B. Kirma

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y.-T. Hsu, Y.-W. Huang, C.-M. Wang, C.-L. Chen, H.-C. Lai, D.G. Mutch, P.J. Goodfellow

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y.-T. Hsu, F. Gu, J. Ruan, R.-L. Huang, C.-L. Chen, R.R. Jadhav, H.-C. Lai, D.G. Mutch, N.B. Kirma, T.H.-M. Huang

Writing, review, and/or revision of the manuscript: Y.-T. Hsu, D.G. Mutch, N.B. Kirma, T.H.-M. Huang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y.-T. Hsu, C.-M. Wang, D.G. Mutch, I.M. Thompson, T.H.-M. Huang

Study supervision: D.G. Mutch, I.M. Thompson, N.B. Kirma, T.H.-M. Huang

The authors thank Dr. Subrata Haldar for article editing and staff at the Genomic Sequencing Facility of the Greehey Children's Cancer Research Institute, University of Texas Health Science Center at San Antonio for next-generation sequencing.

This work was supported by P50 CA134254 (Endometrial Cancer SPORE), U54CA113001 (Integrative Cancer Biology Program), and P30CA054174 (Cancer Center Support Grant) of the NIH and by generous gifts from the Cancer Therapy and Research Center Foundation and the Voelcker Fund.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Sorosky
JI
. 
Endometrial cancer
.
Obstet Gynecol
2012
;
120
:
383
97
.
2.
Aalders
JG
,
Abeler
V
,
Kolstad
P
. 
Recurrent adenocarcinoma of the endometrium: a clinical and histopathological study of 379 patients
.
Gynecol Oncol
1984
;
17
:
85
103
.
3.
Awtrey
CS
,
Cadungog
MG
,
Leitao
MM
,
Alektiar
KM
,
Aghajanian
C
,
Hummer
AJ
, et al
Surgical resection of recurrent endometrial carcinoma
.
Gynecol Oncol
2006
;
102
:
480
8
.
4.
Reddoch
JM
,
Burke
TW
,
Morris
M
,
Tornos
C
,
Levenback
C
,
Gershenson
DM
. 
Surveillance for recurrent endometrial carcinoma: development of a follow-up scheme
.
Gynecol Oncol
1995
;
59
:
221
5
.
5.
Balch
C
,
Matei
DE
,
Huang
TH
,
Nephew
KP
. 
Role of epigenomics in ovarian and endometrial cancers
.
Epigenomics
2010
;
2
:
419
47
.
6.
Huang
YW
,
Luo
J
,
Weng
YI
,
Mutch
DG
,
Goodfellow
PJ
,
Miller
DS
, et al
Promoter hypermethylation of CIDEA, HAAO and RXFP3 associated with microsatellite instability in endometrial carcinomas
.
Gynecol Oncol
2010
;
117
:
239
47
.
7.
Powell
MA
,
Mutch
DG
,
Rader
JS
,
Herzog
TJ
,
Huang
TH
,
Goodfellow
PJ
. 
Ribosomal DNA methylation in patients with endometrial carcinoma: an independent prognostic marker
.
Cancer
2002
;
94
:
2941
52
.
8.
Zuo
T
,
Liu
TM
,
Lan
X
,
Weng
YI
,
Shen
R
,
Gu
F
, et al
Epigenetic silencing mediated through activated PI3K/AKT signaling in breast cancer
.
Cancer Res
2011
;
71
:
1752
62
.
9.
de Assis
S
,
Warri
A
,
Cruz
MI
,
Laja
O
,
Tian
Y
,
Zhang
B
, et al
High-fat or ethinyl-oestradiol intake during pregnancy increases mammary cancer risk in several generations of offspring
.
Nat Commun
2012
;
3
:
1053
.
10.
Jones
PA
,
Baylin
SB
. 
The epigenomics of cancer
.
Cell
2007
;
128
:
683
92
.
11.
Rodriguez-Paredes
M
,
Esteller
M
. 
Cancer epigenetics reaches mainstream oncology
.
Nat Med
2011
;
17
:
330
9
.
12.
Tsai
HC
,
Baylin
SB
. 
Cancer epigenetics: linking basic biology to clinical medicine
.
Cell Res
2011
;
21
:
502
17
.
13.
Kagan
J
,
Srivastava
S
,
Barker
PE
,
Belinsky
SA
,
Cairns
P
. 
Towards clinical application of methylated DNA sequences as cancer biomarkers: a joint NCI's EDRN and NIST workshop on standards, methods, assays, reagents and tools
.
Cancer Res
2007
;
67
:
4545
9
.
14.
Hsiao
SH
,
Huang
TH
,
Leu
YW
. 
Excavating relics of DNA methylation changes during the development of neoplasia
.
Semin Cancer Biol
2009
;
19
:
198
208
.
15.
Szyf
M
. 
The role of DNA hypermethylation and demethylation in cancer and cancer therapy
.
Curr Oncol
2008
;
15
:
72
5
.
16.
Szyf
M
,
Pakneshan
P
,
Rabbani
SA
. 
DNA demethylation and cancer: therapeutic implications
.
Cancer Lett
2004
;
211
:
133
43
.
17.
Kaiser
MF
,
Johnson
DC
,
Wu
P
,
Walker
BA
,
Brioli
A
,
Mirabella
F
, et al
Global methylation analysis identifies prognostically important epigenetically inactivated tumour suppressor genes in multiple myeloma
.
Blood
2013
;
122
:
219
26
.
18.
Rodriguez-Rodero
S
,
Fernandez
AF
,
Fernandez-Morera
JL
,
Castro-Santos
P
,
Bayon
GF
,
Ferrero
C
, et al
DNA methylation signatures identify biologically distinct thyroid cancer subtypes
.
J Clin Endocrinol Metab
2013
;
98
:
2811
21
.
19.
De Faveri
LE
,
Hurst
CD
,
Platt
FM
,
Taylor
CF
,
Roulson
JA
,
Sanchez-Carbayo
M
, et al
Putative tumour suppressor gene necdin is hypermethylated and mutated in human cancer
.
Br J Cancer
2013
;
108
:
1368
77
.
20.
Baylin
SB
,
Jones
PA
. 
A decade of exploring the cancer epigenome—biological and translational implications
.
Nat Rev Cancer
2011
;
11
:
726
34
.
21.
Belkhiri
A
,
El-Rifai
W
. 
5-Methylcytosine hydroxylation-mediated LINE-1 hypomethylation: a novel mechanism of proto-oncogenes activation in colorectal cancer?
Gut.
2013 Jun 27
.
[Epub ahead of print]
.
22.
Kim
R
,
Kulkarni
P
,
Hannenhalli
S
. 
Derepression of Cancer/testis antigens in cancer is associated with distinct patterns of DNA hypomethylation
.
BMC Cancer
2013
;
13
:
144
.
23.
Tandon
M
,
Gokul
K
,
Ali
SA
,
Chen
Z
,
Lian
J
,
Stein
GS
, et al
Runx2 mediates epigenetic silencing of the bone morphogenetic protein-3B (BMP-3B/GDF10) in lung cancer cells
.
Mol Cancer
2012
;
11
:
27
.
24.
Yacqub-Usman
K
,
Duong
CV
,
Clayton
RN
,
Farrell
WE
. 
Epigenomic silencing of the BMP-4 gene in pituitary adenomas: a potential target for epidrug-induced re-expression
.
Endocrinology
2012
;
153
:
3603
12
.
25.
Hashida
Y
,
Nemoto
Y
,
Imajoh
M
,
Murakami
M
,
Taniguchi
A
,
Komatsu
N
, et al
Promoter methylation of the bone morphogenetic protein 6 gene in multiple myeloma
.
Oncol Rep
2012
;
27
:
825
30
.
26.
Kimura
K
,
Toyooka
S
,
Tsukuda
K
,
Yamamoto
H
,
Suehisa
H
,
Soh
J
, et al
The aberrant promoter methylation of BMP3b and BMP6 in malignant pleural mesotheliomas
.
Oncol Rep
2008
;
20
:
1265
8
.
27.
Wen
XZ
,
Akiyama
Y
,
Baylin
SB
,
Yuasa
Y
. 
Frequent epigenetic silencing of the bone morphogenetic protein 2 gene through methylation in gastric carcinomas
.
Oncogene
2006
;
25
:
2666
73
.
28.
Yang
ZJ
,
Liu
FX
,
Yang
YS
,
Yang
X
,
Zhu
GX
. 
Expression of bone-morphogenetic protein 2 and tumor necrosis factor alpha correlates with bone metastases in bladder urothelial carcinoma
.
Ann Diagn Pathol
2013
;
17
:
51
3
.
29.
Guo
X
,
Xiong
L
,
Zou
L
,
Zhao
J
. 
Upregulation of bone morphogenetic protein 4 is associated with poor prognosis in patients with hepatocellular carcinoma
.
Pathol Oncol Res
2012
;
18
:
635
40
.
30.
Xu
G
,
Tang
S
,
Yang
J
,
Chen
K
,
Kang
J
,
Zhao
G
, et al
BMP7 expression in esophageal squamous cell carcinoma and its potential role in modulating metastasis
.
Dig Dis Sci
2013
;
58
:
1871
9
31.
Munz
M
,
Kieu
C
,
Mack
B
,
Schmitt
B
,
Zeidler
R
,
Gires
O
. 
The carcinoma-associated antigen EpCAM upregulates c-myc and induces cell proliferation
.
Oncogene
2004
;
23
:
5748
58
.
32.
van der Gun
BT
,
Melchers
LJ
,
Ruiters
MH
,
de Leij
LF
,
McLaughlin
PM
,
Rots
MG
. 
EpCAM in carcinogenesis: the good, the bad or the ugly
.
Carcinogenesis
2010
;
31
:
1913
21
.
33.
De La Garza
EM
,
Binkley
PA
,
Ganapathy
M
,
Krishnegowda
NK
,
Tekmal
RR
,
Schenken
RS
, et al
Raf-1, a potential therapeutic target, mediates early steps in endometriosis lesion development by endometrial epithelial and stromal cells
.
Endocrinology
2012
;
153
:
3911
21
.
34.
Hsu
PY
,
Hsu
HK
,
Singer
GA
,
Yan
PS
,
Rodriguez
BA
,
Liu
JC
, et al
Estrogen-mediated epigenetic repression of large chromosomal regions through DNA looping
.
Genome Res
2010
;
20
:
733
44
.
35.
Ehrich
M
,
Turner
J
,
Gibbs
P
,
Lipton
L
,
Giovanneti
M
,
Cantor
C
, et al
Cytosine methylation profiling of cancer cell lines
.
Proc Natl Acad Sci U S A
2008
;
105
:
4844
9
.
36.
Robinson
MD
,
Stirzaker
C
,
Statham
AL
,
Coolen
MW
,
Song
JZ
,
Nair
SS
, et al
Evaluation of affinity-based genome-wide DNA methylation data: effects of CpG density, amplification bias, and copy number variation
.
Genome Res
2010
;
20
:
1719
29
.
37.
Serre
D
,
Lee
BH
,
Ting
AH
. 
MBD-isolated Genome Sequencing provides a high-throughput and comprehensive survey of DNA methylation in the human genome
.
Nucleic Acids Res
2010
;
38
:
391
9
.
38.
Irizarry
RA
,
Ladd-Acosta
C
,
Wen
B
,
Wu
Z
,
Montano
C
,
Onyango
P
, et al
The human colon cancer methylome shows similar hypo- and hypermethylation at conserved tissue-specific CpG island shores
.
Nat Genet
2009
;
41
:
178
86
.
39.
Peng
DF
,
Kanai
Y
,
Sawada
M
,
Ushijima
S
,
Hiraoka
N
,
Kosuge
T
, et al
Increased DNA methyltransferase 1 (DNMT1) protein expression in precancerous conditions and ductal carcinomas of the pancreas
.
Cancer Sci
2005
;
96
:
403
8
.
40.
Huang
SS
,
Fraenkel
E
. 
Integrating proteomic, transcriptional, and interactome data reveals hidden components of signaling and regulatory networks
.
Sci Signal
2009
;
2
:
ra40
.
41.
Keshava Prasad
TS
,
Goel
R
,
Kandasamy
K
,
Keerthikumar
S
,
Kumar
S
,
Mathivanan
S
, et al
Human protein reference database—2009 update
.
Nucleic Acids Res
2009
;
37
:
D767
72
.
42.
Pasquale
EB
. 
Eph-ephrin bidirectional signaling in physiology and disease
.
Cell
2008
;
133
:
38
52
.
43.
Thiery
JP
,
Sleeman
JP
. 
Complex networks orchestrate epithelial–mesenchymal transitions
.
Nat Rev Mol Cell Biol
2006
;
7
:
131
42
.
44.
Bailey
JM
,
Singh
PK
,
Hollingsworth
MA
. 
Cancer metastasis facilitated by developmental pathways: sonic hedgehog, Notch, and bone morphogenic proteins
.
J Cell Biochem
2007
;
102
:
829
39
.
45.
Berx
G
,
van Roy
F
. 
Involvement of members of the cadherin superfamily in cancer
.
Cold Spring Harb Perspect Biol
2009
;
1
:
a003129
.
46.
Kandoth
C
,
Schultz
N
,
Cherniack
AD
,
Akbani
R
,
Liu
Y
,
Shen
H
, et al
Integrated genomic characterization of endometrial carcinoma
.
Nature
2013
;
497
:
67
73
.
47.
Larsen
AB
,
Pedersen
MW
,
Stockhausen
MT
,
Grandal
MV
,
van Deurs
B
,
Poulsen
HS
. 
Activation of the EGFR gene target EphA2 inhibits epidermal growth factor–induced cancer cell motility
.
Mol Cancer Res
2007
;
5
:
283
93
.
48.
Scambia
G
,
Benedetti Panici
P
,
Ferrandina
G
,
Battaglia
F
,
Distefano
M
,
D'Andrea
G
, et al
Significance of epidermal growth factor receptor expression in primary human endometrial cancer
.
Int J Cancer
1994
;
56
:
26
30
.
49.
Pandiyan
K
,
You
JS
,
Yang
X
,
Dai
C
,
Zhou
XJ
,
Baylin
SB
, et al
Functional DNA demethylation is accompanied by chromatin accessibility
.
Nucleic Acids Res
2013
;
41
:
3973
85
.
50.
McGarvey
KM
,
Fahrner
JA
,
Greene
E
,
Martens
J
,
Jenuwein
T
,
Baylin
SB
. 
Silenced tumor suppressor genes reactivated by DNA demethylation do not return to a fully euchromatic chromatin state
.
Cancer Res
2006
;
66
:
3541
9
.
51.
Nieto
MA
. 
The ins and outs of the epithelial to mesenchymal transition in health and disease
.
Annu Rev Cell Dev Biol
2011
;
27
:
347
76
.
52.
Thiery
JP
,
Acloque
H
,
Huang
RY
,
Nieto
MA
. 
Epithelial–mesenchymal transitions in development and disease
.
Cell
2009
;
139
:
871
90
.
53.
Hugo
HJ
,
Kokkinos
MI
,
Blick
T
,
Ackland
ML
,
Thompson
EW
,
Newgreen
DF
. 
Defining the E-cadherin repressor interactome in epithelial–mesenchymal transition: the PMC42 model as a case study
.
Cells Tissues Organs
2011
;
193
:
23
40
.
54.
Beltran
M
,
Puig
I
,
Pena
C
,
Garcia
JM
,
Alvarez
AB
,
Pena
R
, et al
A natural antisense transcript regulates Zeb2/Sip1 gene expression during Snail1-induced epithelial–mesenchymal transition
.
Genes Dev
2008
;
22
:
756
69
.
55.
Denzel
S
,
Maetzel
D
,
Mack
B
,
Eggert
C
,
Barr
G
,
Gires
O
. 
Initial activation of EpCAM cleavage via cell-to-cell contact
.
BMC Cancer
2009
;
9
:
402
.
56.
Bandyopadhyay
A
,
Yadav
PS
,
Prashar
P
. 
BMP signaling in development and diseases: a pharmacological perspective
.
Biochem Pharmacol
2013
;
85
:
857
64
.
57.
Chen
A
,
Wang
D
,
Liu
X
,
He
S
,
Yu
Z
,
Wang
J
. 
Inhibitory effect of BMP-2 on the proliferation of breast cancer cells
.
Mol Med Rep
2012
;
6
:
615
20
.
58.
Le Page
C
,
Puiffe
ML
,
Meunier
L
,
Zietarska
M
,
de Ladurantaye
M
,
Tonin
PN
, et al
BMP-2 signaling in ovarian cancer and its association with poor prognosis
.
J Ovarian Res
2009
;
2
:
4
.
59.
Kang
MH
,
Kim
JS
,
Seo
JE
,
Oh
SC
,
Yoo
YA
. 
BMP2 accelerates the motility and invasiveness of gastric cancer cells via activation of the phosphatidylinositol 3-kinase (PI3K)/Akt pathway
.
Exp Cell Res
2010
;
316
:
24
37
.
60.
Lee
YC
,
Cheng
CJ
,
Bilen
MA
,
Lu
JF
,
Satcher
RL
,
Yu-Lee
LY
, et al
BMP4 promotes prostate tumor growth in bone through osteogenesis
.
Cancer Res
2011
;
71
:
5194
203
.
61.
Karreth
F
,
Tuveson
DA
. 
Twist induces an epithelial–mesenchymal transition to facilitate tumor metastasis
.
Cancer Biol Ther
2004
;
3
:
1058
9
.
62.
Al Moustafa
AE
,
Achkhar
A
,
Yasmeen
A
. 
EGF-receptor signaling and epithelial–mesenchymal transition in human carcinomas
.
Front Biosci
2012
;
4
:
671
84
.
63.
Livasy
CA
,
Reading
FC
,
Moore
DT
,
Boggess
JF
,
Lininger
RA
. 
EGFR expression and HER2/neu overexpression/amplification in endometrial carcinosarcoma
.
Gynecol Oncol
2006
;
100
:
101
6
.
64.
Niikura
H
,
Sasano
H
,
Matsunaga
G
,
Watanabe
K
,
Ito
K
,
Sato
S
, et al
Prognostic value of epidermal growth factor receptor expression in endometrioid endometrial carcinoma
.
Hum Pathol
1995
;
26
:
892
6
.
65.
Maetzel
D
,
Denzel
S
,
Mack
B
,
Canis
M
,
Went
P
,
Benk
M
, et al
Nuclear signalling by tumour-associated antigen EpCAM
.
Nat Cell Biol
2009
;
11
:
162
71
.
66.
Bachmann
IM
,
Halvorsen
OJ
,
Collett
K
,
Stefansson
IM
,
Straume
O
,
Haukaas
SA
, et al
EZH2 expression is associated with high proliferation rate and aggressive tumor subgroups in cutaneous melanoma and cancers of the endometrium, prostate, and breast
.
J Clin Oncol
2006
;
24
:
268
73
.

Supplementary data