Chromatin composition alterations and the critical role of MeCP2 for epigenetic silencing of progesterone receptor-B gene in endometrial cancers

Abstract

Objective

To understand the epigenetic mechanism underlying the PR-B gene silencing in endometrial cancer (EC) cells, we compared the chromatin composition between transcriptionally active and silenced PR-B genes in EC cell lines and cancer tissues.

Methods

Chromatin Immunoprecipitation (ChIP) assay was performed to measure MBD occupancy and histone acetylation/methylation in transcriptionally active and silenced PR-B genes. PR-B-positive/-negative, as well as epigenetic inhibitor-treated/-untreated EC cells were used as study models. Real-time polymerase chain reaction (PCR) and Western blot analysis were applied to measure the mRNA and protein levels of PR-B, MBD, and histones.

Results

A close association among PR-B methylation, MBD binding and PR-B gene silencing was observed. Treatment with epigenetic inhibitors led to dynamic changes in the PR-B chromatin composition and gene expression. Increased H3/H4 acetylation and H3-K4 methylation, and decreased H3-K9 methylation were found to be associated with re-activation of silenced PR-B genes. MeCP2 knockdown resulted in a decreased MeCP2 binding to PR-B genes and an increased PR-B expression. ChIP analysis of MeCP2 binding to PR-B genes in the PR-B-positive/-negative EC samples confirmed the significant role of MeCP2 in PR-B silencing.

Conclusion

PR-B gene expression is regulated by a concerted action of epigenetic factors including DNA methylation, MBD binding, and histone modifications. MeCP2 occupancy of PR-B genes plays a critical role in PR-B gene silencing. These findings enriched our knowledge of the epigenetic regulation of PR-B expression in EC, and suggested that the epigenetic re-activation of PR-B could be explored as a potential strategy to sensitize the PR-B-negative endometrial cancers to progestational therapy.

Introduction

One well-recognized effect of progesterone in the human uterus is the protection of the endometrium against the hyperplastic and tumorigenic activities caused by excessive levels of estrogens [1, 2]. Progestins are routinely prescribed together with estrogen in hormone replacement therapy to prevent endometrial cancers (EC) [3]. Progestational treatment represents a common hormonal therapy approach for patients with breast and gynecologic cancers. Complete reversal of endometrial hyperplasia and EC using high doses of progestins has been reported [4–6]. Progestin regimens were found to be beneficial for advanced or recurrent endometrial cancers in some, but not all clinical trials [7–9]. Improved response rates and relatively mild side effects were observed with a combined regimen consisting of progesterone and tamoxifen [10, 11] or progesterone and an aromatase inhibitor in treating patients with advanced endometrial cancers [12].

The anti-cancer effects of progestins in endometrial cells are primarily mediated by the progesterone receptor (PR), a member of the nuclear receptor family with ligand-dependent transcription activities. When bound to progesterones, the carboxyl terminus of the PR undergoes conformational changes that promote PR dimerization and binding to the target DNA [13]. The activated PR interacts with a variety of co-activators/co-repressors and regulates the expression of diverse downstream genes [14]. Progesterone inhibits EC progression via the suppression of cell cycle progression and induction of apoptosis. Progesterone decreases the expression of the estrogen receptor [15], upregulates the cyclin-dependent kinase inhibitor P27 Kip1 [16], and inhibits EC cell proliferation. In addition, prolonged progestin treatment increases cellular levels of P21Cip/waf1, a key cell-cycle regulator, through activation of P53 [17]. In vivo studies indicated that induction of apoptosis is an early event during treatment of endometrial hyperplasia with progestins [18]. Unfortunately, PR expression is frequently lost in advanced EC, which is considered a most common cause for resistance to hormonal therapy [19, 20]. While the majority (72 %) of PR-positive tumors respond to progesterone treatment, the response rate is much lower (10–20 %) for PR-negative tumors [21].

Endometrial glandular cells express two isoforms of progesterone receptors, PR-A and PR-B. These isoforms are generated through the use of alternative promoters with differing transcription start sites. Two lines of evidence provided support for the crucial role of PR-B in the pathogenesis of EC. First, PR-B accounts for most of the inhibitory effects of progestins on cancer cell growth [22]. Second, clinical studies in endometrial cancers have shown a drastic decrease in mRNA levels of PR isoform B, but not A [23, 24]. The loss of PR-B expression has been found to be associated with cancer invasion and recurrence [25, 26]. The PR-B gene contains a typical CpG island spanning nucleotides −80 to +500 relative to the transcription start site. Aberrant DNA hypermethylation within this region is considered to be the cause of the PR-B-negative phenotype in EC cell lines, as well as in patient samples [27]. Interestingly, we observed that DNA methyltransferases (DNMT) 1 and 3B, the enzymes responsible for genomic DNA methylation, were overexpressed three- to fourfold in endometrioid cancers compared to normal tissues [28]. These findings suggested that a simultaneous upregulation of maintenance (DNMT1) and de novo (DNMT3B) methyltransferases may take a part in the aberrant DNA hypermethylation of target genes such as PR-B. The pivotal role of DNA methylation in PR-B gene inactivation is also supported by the results of in vitro studies. We and others have shown that treatment of PR-B-negative EC cell lines with DNMT inhibitors resulted in PR-B gene demethylation and a concomitant PR-B gene re-activation [27, 29]. While these observations underscored the significance of the DNA methylation-mediated pathway in PR-B silencing, the underlying mechanism(s) remain to be investigated. It is not clear, for example, what alterations in local chromatin components may be involved in the cancer-related PR-B gene inactivation.

Accumulated data suggest that changes in chromatin composition contribute to transcriptional silencing [30–32]. Methyl-CpG binding domain proteins (MBDs) specifically bind to methylated DNA and recruit histone modification enzymes such as histone deacetylases (HDAC), methyltransferases (Mets) and/or histone demethylase [33–37], leading to histone modification alterations and gene silencing. In this study, we characterized the MBD binding and histone modifications in the chromatins associated with the methylated and unmethylated PR-B genes. Moreover, the relationship between chromatin composition and PR-B silencing was determined in EC cells that were pre-treated with DNMT and HDAC inhibitors as well as PR-B-negative and PR-B-positive EC tissues. These experiments have led to the identification of specific factors and mechanisms that are critical for PR-B epigenetic silencing.

Materials and methods

Cell lines and reagents

The human endometrial cancer cell lines AN3, KLE, RL-95, HEC-1A, and HEC-1B were purchased from American Type Culture Collection (ATCC, Rockville, MD). These cells were grown in DMEM/F12 medium. The well-differentiated human endometrioid adenocarcinoma Ishikawa cell line was generously provided by Dr. Masato Nishida (Kasumigaura National Hospital, Japan) [38]. Ishikawa cells are maintained in MEMα medium. All the media are supplemented with 10 % fetal bovine serum (BioWhitaker, Walkersville, MD), 100 μg/ml streptomycin, 100 units/ml penicillin, and 2 mM l-glutamine. Cells were grown to 20 % confluence in 10-cm dishes and treated with different concentrations of aza-deoxycytidine (ADC, Sigma, St. Louis, MO) or trichostatin A (TSA, Sigma, St. Louis, MO) as indicated in the figure legends.

Antibodies against MeCP2, histone H3 and H4, acetyl-histone H3, acetyl-histone H4, dimethyl-histone H3 at Lysine-4 (Met H3-K4), and dimethyl-histone H3 at Lysine-9 (Met H3-K9) were obtained from Upstate Biotechnology, Inc (Lake Placid, NY). Rabbit antibody for β-actin, and rabbit and goat antibodies for MBD1, MBD2, MBD3, and MBD4 were purchased from Santa Cruz Biotechnology (Santa Cruz, CA).

Collection of tissue samples

PR-B-positive and -negative endometrial cancer tissue samples (four cases in each group) were collected from patients treated at the Mayo Clinic. The PR-B expression status of tissues was diagnosed by a pathologist based on the pathological review of the immunostaining results. Fresh tissue samples were snap frozen and stored at −80 °C. This study was approved by the Institutional Review Board of the Mayo Foundation. In accordance with the Minnesota Statute for the Use of Medical Information in Research, only patients who consented to the use of their cancer tissues and medical records were included in the study.

Real-time PCR

Total RNA was isolated using TRIzol reagent (Invitrogen, Carlsbad, CA). cDNA was synthesized with 1 μg RNA using a SuperScript™ kit (Invitrogen, Carlsbad, CA). The 20 μl products of reverse transcription were diluted to 100 μl, and 2 μl was used for each real-time polymerase chain reaction (PCR) assay. PCR was performed in a total volume of 25 μl containing 140 ng forward and backward primers, respectively, and 12.5 μl SYBR green Master Mix (Stratagene, Cedar Creek, TX). The mRNA levels of PR-B and GAPDH (glyceraldehyde-3-phosphate dehydrogenase) genes were measured using the following primers: PRB-F: 5′-ACT GAG CTG AAG GCA AAG GGT; PRB-R: 5′-GTC CTG TCC CTG GCA GGG C; MeCP2-F: 5′-CAG GCA AAG CAG AGA CAT CA-3′; MeCP2-R: 5′-GCT TAA GCT TCC GTG TCC AG-3′; GAPDH-F: 5′-GAA GGT GAA GGT CGG AGT C-3′; GAPDH-R: 5′-GAA GAT GGT GAT GGG ATT TC-3′. PCR conditions were: initial denaturing at 95 °C for 5 min, followed by 40 cycles of denaturing at 95 °C for 15 s, annealing at 56 °C for 30 s, and extension at 72 °C for 30 s. The specificity of the real-time PCR was verified by a clear, single DNA band with the predicted size of final PCR products resolved in the agarose gel electrophoresis.

siRNA knockdown of MeCP2

siRNA oligonucleotides of MeCP2 (5′-GCU CUA AAG UGG AGU UGA UUU -3′) were purchased from Dharmacon Technology, Inc (Chicago, IL). KLE cells were seeded on six-well plates at 60 % confluence 24 h before the transfection. Cells were transfected with siRNA oligo at a final concentration of 100 nM using the DharmaFECT 1 transfection reagents (Dharmacon, Chicago, IL). Total RNA was isolated from the cells and reverse transcription and real-time PCR were performed to measure MeCP2 and PR-B mRNA levels. For ChIP Assay, cells were seeded on 100-mm dishes and transfected with the same reagents.

Methylation-specific PCR

Genomic DNA was isolated from cell cultures using DNAzol reagent (Molecular Research Center, Cincinnati, OH) following the manufacturer’s instructions. The EZ DNA methylation kit (Zymo Research, Orange, CA) was used for sodium bisulfite conversion of genomic DNA. Methylation-specific PCR was performed with the primers specific for either methylated or unmethylated PR-B CpG islands as previously published [27]. 10 μl of PCR products were resolved in agarose gel electrophoresis and the DNA bands were visualized by ethidium bromide staining.

Chromatin immunoprecipitation (ChIP) assays

ChIP was performed using the ChIP Assay Kit (Upstate Biotechnology, Inc., Lake Placid, NY) with some modifications on recommended protocols. Briefly, one 10-cm dish of cell culture containing approximately 5 × 10⁶ cells, or 50–100 μg of EC tissues, were used for each ChIP assay. Protein-DNA cross-linking was carried out by exposure to 1 % formaldehyde at 37 °C for 10 min. The medium was removed and cells were washed three times with ice-cold phosphate buffered saline (PBS). Then, 1.5 ml PBS was added and the cells were scraped off the culture dishes, and transferred into conical tubes. Cells were collected by centrifugation at 2,000 rpm for 4 min at 4 °C. The PBS was removed and 200 μl of SDS lysis buffer (1 % SDS, 10 mM EDTA, 50 mM Tris–HCl, pH 8.1) supplemented with protease inhibitors (1 mM PMSF, 1 μg/ml aprotinin, and 1 μg/ml pepstatin A) was added to re-suspend the cell pellets. Sonication was performed on ice using a sonicator (Sonic Dismembrator, Model 500, Fisher Scientific) pre-set for 10-s pulses with 10-s intervals. Four repeated sonication cycles were applied to achieve chromatin fragmentation of 200–1,000 bp. Samples were centrifuged at 13,000 rpm for 10 min and the supernatants were transferred to new tubes. The samples were diluted tenfold with ChIP dilution buffer (0.01 % SDS, 1 % Triton X-100, 2 mM EDTA, 16.7 mM Tris–HCl, pH 8.1, 150 mM NaCl). A 20-μl aliquot of the sample was removed to serve as the input control. To reduce the non-specific background, the DNA–protein complexes were pre-cleared by incubation with 75 μl of Protein A agarose beads (50 % slurry containing salmon sperm DNA). The pre-absorption was carried out at 4 °C with constant rotation for 2 h. Specific antibodies against MeCP2, MBD1, MBD2, MBD3, MBD4, total histones, acetylated histones, and methylated histones were used for immunoprecipitation. Antibody binding was accomplished by incubation at 4 °C overnight with constant rotation. For negative controls, a non-specific antibody was used instead of specific antibodies. To collect immune complexes, 60 μl of Protein A agarose-salmon sperm DNA (50 % slurry) was added to each tube and incubation continued for 2 h at 4 °C. Agarose beads were recovered by gentle centrifugation at 2,000 rpm for 2 min. The beads were washed sequentially with 1 ml buffer for 5 min in the following order: two times with low-salt buffer (0.1 % SDS, 1 % Triton X-100, 2 mM EDTA, 20 mM Tris–HCl, pH 8.1, 150 mM NaCl), two times with high-salt buffer (0.1 % SDS, 1 % Triton X-100, 2 mM EDTA, 20 mM Tris–HCl, pH 8.1, 500 mM NaCl), once with LiCl buffer (0.25 mM LiCl, 0.5 % deoxycholic acid, 1 mM EDTA, 10 mM Tris, pH 8.1), and once with 1X TE buffer. After washing, 500 μl fresh 1 % SDS and 0.1 M NaHCO₃ was added to elute the immune complexes. Formaldehyde cross-links were reversed by adding 20 μl 5 M NaCl to 500 μl eluates and heating at 65 °C for 4 h. DNA fragments were recovered by ethanol precipitation after proteinase K digestion and phenol/chloroform extraction.

Following immunoprecipitation, PCR was performed with PR-B gene-specific primers: 5′-TCA GAA TAA CGG GTG GAA ATG-3′ and 5′-TCT AAC AAC GCC TCC TCC TC-3′. The 108 bp amplicon containing four CpG sites represents the CpG islands from the PR-B promoter region. PCR conditions were: 94 °C for 5 min for initial denaturation, followed by 30 cycles of denaturation at 94 °C for 45 s, annealing at 56 °C for 45 s, and extension at 72 °C for 1 min. The final PCR products were analyzed with 2 % agarose gels electrophoresis and ethidium bromide staining. The ChIP results were documented by Polaroid photograph with UV excitation.

Western blot analysis

Cell cultures were rinsed three times with cold PBS and harvested by scraping in lysis buffer (20 mM Hepes, pH 7.2, 25 % glycerol, 0.4 M NaCl, 1.5 mM MgCl₂, 0.2 mM EDTA, 0.5 mM DTT, and 0.5 mM phenylmethylsulfonyl fluoride). The lysis buffer was supplemented with 1X protease inhibitor cocktail (Sigma, St. Louis, MO). Cellular proteins were quantified and resolved by SDS polyacrylamide gel electrophoresis and transferred to ImmunoBlot™ PDVF membranes (Bio-Rad Laboratories, Hercules, CA) as previously described [39]. Expression levels of MeCP2, MBD1, total H3 and H4, and acetylated H3 and H4 were determined using specific antibodies following the manufacturer’s instructions. Chemiluminescence detection was performed with the ECLplus™ Western Blotting Detection System (Amersham Corp, Arlington Heights, IL). The blots were re-probed with β-actin antibody and the results provided controls for protein loading.

Data analysis

Relative PR-B mRNA levels of endometrial cancer cell lines were calculated based on the real-time PCR data. The threshold cycle number (CT) for PR-B was normalized against the GAPDH internal reference gene by the formula: ΔCT = CT_PR-B − CT_GAPDH. The difference between PR-B and GAPDH was further converted to a relative fold (F = 2^ΔCT). The PR-B mRNA level of the Ishikawa cell was arbitrarily set at 100, and PR-B mRNA levels in AN3, KLE, RL-95, HEC-1A and HEC-1B cells were expressed as a percentage of the level in Ishikawa cells. The results of the ChIP experiments were documented with an HP Q3190A scanner and analyzed by densitometry using the NIH Image program. All data groups from the real-time PCR and ChIP experiments were analyzed by a multivariate analysis of variance (ANOVA) to determine if there was a significant difference among the groups. For data groups that satisfied the initial ANOVA criterion, individual comparisons were performed with the use of post-hoc Bonferroni t tests with the assumption of two-tail distribution and two samples with equal variance. The statistical significance (p < 0.05) was marked by asterisks in the figures. Average values and standard errors were calculated and graphically presented for quantitative data from real-time PCR and ChIP experiments.

Results

Establishment of study models and optimal experimental conditions

We measured PR-B mRNA levels in six cell lines using real-time PCR to identify those with high and low PR-B expression. As shown in Fig. 1, Ishikawa, a well-differentiated, non-tumorigenic endometrial cancer cell line, expressed the highest levels of PR-B mRNA. KLE and HEC-1B, the two least-differentiated cell lines, expressed little or no PR-B mRNA. These three cell lines with the divergent PR-B expression levels were chosen for further comparative investigation. Treatment with either the DNMT inhibitor 5-aza-2′-deoxycytidine (ADC) or the HDAC inhibitor Trichostatin A (TSA) did not affect PR-B expression in the PR-B-positive Ishikawa cells, but dramatically induced PR-B mRNA expression in PR-B-negative KLE and HEC-1B cells (Fig. 2a), suggesting the involvement of epigenetic control mechanisms. We performed methylation-specific PCR (MS-PCR) to verify the methylation status of PR-B gene in these cell lines. As shown in Fig. 2b, KLE and HEC-1B cells (PR-B-negative) contained mostly methylated PR-B, while Ishikawa cells (PR-B-positive) carried mainly the unmethylated PR-B gene, a result largely consistent with their PR-B expression patterns. It should be pointed out that although for the simplicity of description, we marked PR-B methylation status as ‘unmethylated’ in Ishikawa cells and ‘methylated’ in KLE cells and HEC-1B cells, the MS-PCR only assesses the methylation status of two CpG sites, and the possible divergence in DNA methylation in other CpG sites could not be ruled out. For example, a relatively low level of PR-B gene methylation in Ishikawa cells, or partial methylation of the PR-B gene in KLE and HEC-1B cells may exist.

Fig. 1.

PR-B mRNA expression in well-differentiated (Ishikawa) and poorly-differentiated (AN3, KLE, RL-95, HEC-1A and HEC-1B) endometrial cancer cell lines. Real-time PCR was performed as described in “Materials and methods”. The levels of PR-B mRNA were standardized by those of the GAPDH gene as internal controls. The PR-B mRNA level of the Ishikawa cell was arbitrarily set at 100. Relative PR-B mRNA levels for the remaining cell lines were expressed as a percentage of those in Ishikawa cells. Averages and standard errors were calculated from three repeated experiments. Note that KLE and HEC-1B cells express 2–3 orders lower levels of PR-B mRNA than Ishikawa cells

Fig. 2.

Induction of PR-B expression in KLE and HEC-1B cells by epigenetic inhibitor. a PCR products representing PR-B and GAPDH mRNA were resolved in agarose gel electrophoresis. The single-band pattern indicated specific amplification by PCR. PCR assays performed with mock reverse transcription without reverse transcriptase were included as controls for genomic DNA contamination. Following 25 cycles of PCR, no DNA band was observed in KLE and HEC-1B. Following treatment with 2.5 μM ADC for 48 h or 7 nM TSA for 24 h, PR-B mRNA expression was detected. b Results of MS-PCR. Methylated or unmethylated PR-B genes were found in KLE/HEC-1B or Ishikawa cells, respectively. Note that while TSA treatment did not affect the DNA methylation, ADC treatment induced the demethylation of PR-B genes in KLE and HEC-1B cells. “M”, PCR with primers targeting the methylated PR-B gene. “U”, PCR using primers targeting unmethylated PR-B gene. c–e Dose- and time-dependent responses. HEC-1B cells were treated with increasing concentrations (panels c, e) and time (panels d, f) of ADC (panels c, d) or TSA (panels e, f) as indicated. PR-B mRNA levels were measured by real-time PCR and standardized against those of GAPDH. PR-B levels from drug-treated cells were compared to those of the untreated control cells. Statistical significance (* p < 0.05) are marked with asterisks on the top the standard error bars. 2.5 μM of ADC and treatment for 48 h, or 7 nM of TSA for treatment of 24 h, were identified as optimal treatment conditions

MS-PCR results indicated that ADC treatment can effectively reverse the PR-B DNA methylation status in KLE and HEC-1B cells. In contrast, TSA treatment did not affect PR-B gene methylation (Fig. 2b). To establish the optimal conditions of drug treatment for PR-B re-activation, we treated PR-B-negative HEC-1B cells with varied concentrations and durations of ADC and TSA, and quantified the PR-B mRNA levels using real-time PCR. Figure 2c and d demonstrated that treatment with 2.5 μM of ADC for 48 h led to the highest induction of PR-B mRNA levels. Similarly, Fig. 2e and f showed that 7 nM for 24 h represented the optimal treatment conditions for TAS. These conditions were used in the subsequent chromatin immunoprecipitation experiments.

Differential MBD occupancy at methylated and unmethylated PR-B genes

We have investigated the possible involvement of MBDs in the control of the PR-B gene by comparing their occupancy at the PR-B gene in PR-B-positive and -negative cells, as well as before and after drug treatment. The binding of MeCP2, MBD1, MBD2, MBD3, and MBD4 to the PR-B gene was examined using ChIP assays as described under “Materials and methods”. The left panels of Fig. 3 show one representative set of ChIP results from KLE (a), HEC-1B (b), and Ishikawa cells (c). The right panels of Fig. 3 document the average MBD binding levels based on densitometry analyses of results from three repeated ChIP experiments in the three cell lines. The results can be summarized as following: (1) substantial MeCP2 binding was detected in the methylated PR-B gene of KLE and HEC-1B cells. In contrast, relatively little or no MeCP2 binding to the unmethylated PR-B gene was detected in Ishikawa cells. In addition, treatment with ADC resulted in reduced binding of MeCP2 in KLE and HEC-1B cells, but not in Ishikawa cells. (2) MBD1 and MBD2 binding was detected in both methylated (KLE and HEC-1B cells) and unmethylated (Ishikawa cells) PR-B genes. Treatment with ADC reduced MBD1 occupancy within the methylated, but not unmethylated, PR-B gene. The same treatment reduced MBD2 binding in KLE, but not in HEC-1B or Ishikawa cells. (3) No significant MBD3 or MBD4 binding was detected in any of the cell lines examined. (4) TSA treatment did not change MBD binding patterns in any of the cell lines examined.

Fig. 3.

Representative results of the ChIP assay to determine the MBD occupancy of the PR-B gene in KLE and HEC-1B (PR-B-negative, PR-B gene methylated) and Ishikawa (PR-B-positive, PR-B gene unmethylated) cells. Cells were treated with ADC (2.5 μM, 48 h) or TSA (7 nM, 24 h) or DMSO solvent as control. Left panels are representative results of the ChIP experiments using specific antibodies raised against MeCP2, MBD1, MBD2, MBD3, and MBD4. The PCR products representing the PR-B gene were resolved in 2 % agarose gels and visualized by ethidium bromide staining. Positive (input) and negative (using non-specific antibody, N.S.) controls were included. Right panels show densitometry analyses of ChIP results in the three cell lines. Averages and standard errors were calculated from three repeated experiments. Statistical significance (* p < 0.05) between treated and untreated cells were indicated by asterisks

Thus, different MBD family members exhibited diverse and dynamic profiles of interaction with the PR-B gene. Specifically, these results demonstrated a close association between MeCP2 occupancy of the methylated PR-B gene and PR-B expression inactivation. The weak MBD3 and MBD4 occupancy at the PR-B gene suggests that these two factors are unlikely to play a major role in the DNA methylation-mediated PR-B regulation. One interesting result is the considerable levels of MBD1 and MBD2 binding to the unmethylated PR-B gene in Ishikawa cells. As discussed further below, this data pointed to the involvement of these factors in a regulation mechanism(s) most likely unrelated to DNA methylation.

Specific histone modification patterns in methylated and unmethylated PR-B genes

We next investigated the significance of histone modification in PR-B regulation. H3 and H4 acetylation, as well as H3 methylation at the lysine 9 (H3-K9) and lysine 4 (H3-K4) positions in the PR-B gene were examined in the ChIP experiments using specific antibodies. The left panels of Fig. 4 show representative results from repeated ChIP experiments. Average densities of the PCR products representing the chromatin composition of the PR-B gene were calculated for each experimental group (right panels, Fig. 4). These results are summarized as following: (1) While total H3 and H4 binding levels were comparable among the three cell lines, relatively lower acetylation levels at both H3 and H4 were found in methylated (KLE and HEC-1B cells) compared to unmethylated (Ishikawa cells) PR-B genes. (2) Treatment with either ADC or TSA led to significant increases in histone acetylation on the methylated PR-B gene. These alterations were all statistically significant, except for H4 acetylation in the HEC-1B cells following ADC treatment. (3) H3-K9 methylation levels appeared to be higher in the methylated PR-B gene compared to the unmethylated PR-B gene, while H3-K4 methylation displayed an opposite trend. (4) Treatment of KLE cells with ADC and TSA resulted in demethylation on H3-K9, but increased H3-K4 methylation. Similar results were found in HEC-1B cells, with the exception that the increase of H3-K4 methylation induced by TSA treatment did not reach a statistical significance. (5) ADC and TSA treatment did not significantly affect histone acetylation or methylation in the unmethylated PR-B gene (Ishikawa cells), a result correspondent to that for MBDs binding. Thus, covalent histone modifications appear to undergo dramatic changes following treatment with TSA and ADC. Overall, these results indicate a general association of higher levels of histone acetylation and H3-K4 methylation, and lower levels of H3-K9 methylation, with the transcriptional activation of the methylated and silenced PR-B gene.

Fig. 4.

ChIP assays on histone modifications in PR-B genes. KLE, HEC-1B, and Ishikawa cells were treated with ADC and TSA, as described in Fig. 3. Left panels are representative results of ChIP experiments performed with antibodies against total H3 and H4, acetylated H3 and H4, and methylated H3 at K4 and K9 positions. The PCR products representing PR-B genes were resolved and visualized in 2 % agarose gels. Positive (input) and negative (using non-specific antibody, N.S.) controls were included. Right panels show densitometry analyses on histone modifications of PR-B genes in KLE, HEC-1B, and Ishikawa cells. The results were calculated from three repeated experiments. Statistical significance (* p < 0.05) between drug-treated and -untreated cells were indicated by asterisks

Expression levels of MBDs and histones

We questioned whether the alterations in PR-B chromatin composition observed above could be caused by global differences in MBDs or histone protein expression levels among these EC cell lines. To answer this question, we performed a Western blot analysis to measure cellular levels of relevant MBDs and histones, including MeCP2, MBD1, MBD2, and deacetylated and acetylated H3 and H4. As shown in Fig. 5 (upper panels), all the EC cell lines express comparable levels of MeCP2 (75 kDa), MBD1 (80 kDa), and MBD2 (40 kDa) despite their dramatic differences in PR-B expression levels. More importantly, the expression of these factors was not significantly affected by ADC or TSA treatment. These results suggest that the varied MBD binding detected by the ChIP assay is most likely due to local, PR-B gene-specific alterations rather than these factors’ global changes on protein expression. Similarly, comparable levels of total H3 and H4 were found in the three cell lines, and the H3 and H4 levels were not affected by either ADC or TSA treatment. Treatment with TSA, but not ADC, increased the global acetylation levels in H3 and H4 in these cell lines (Fig. 5, lower panels). These results suggested a concomitant increase in histone acetylation at the PR-B gene, as observed in ChIP analysis (Fig. 4), with the increased global histone acetylation levels detected by Western blot analysis.

Fig. 5.

Western blot analysis of global MBD and histone levels. KLE, HEC-1B and Ishikawa cells were treated under the conditions described in Fig. 3. Protein extracts were isolated and quantified as described under “Materials and methods”. Cellular levels of MBDs and total or acetylated histones were detected using specific antibodies. The results of β-actin (43 KDa) provide controls for protein loading

Confirmation on the inhibitory role of MeCP2 in PR-B epigenetic regulation

Based on the above results from the ChIP assays suggesting a strong MeCP2 involvement in the PR-B silencing, we performed MeCP2 knockdown experiments to verify the potential role of MeCP2 in PR-B epigenetic regulation. First, we confirmed that compared to the KLE cells transfected with control siRNA (Ctl siRNA), KLE cells transfected with MeCP2-specific siRNA (MeCP2 siRNA) expressed dramatically decreased MeCP2 mRNA levels at 48 h postransfection (Fig. 6a). ChIP analysis showed that following MeCP2 knockdown, MeCP2 binding to the methylated PR-B gene was lower compared to the control group (Fig. 6b). Moreover, measurement of the PR-B mRNA level indicated that the reduced MeCP2 occupancy of the PR-B gene was accompanied by a concomitant increase in PR-B mRNA levels, indicating re-activation of the silenced PR-B gene (Fig. 6c). These results strongly suggested that the binding of MeCP2 to the PR-B gene is critical for the DNA methylation-mediated PR-B silencing in a PR-B-negative EC cell line.

Fig. 6.

MeCP2 knockdown experiments in the PR-B-negative KLE cells. a Validation of the siRNA-mediated MeCP2 knockdown. Following siRNA transfection, MeCP2 mRNA levels were measured with real-time PCR. Compared to the control siRNA (Ctl siRNA), transfection with MeCP2-specific siRNA (MeCP2 siRNA) resulted in a significant reduction of MeCP2 mRNA levels (* p < 0.01). b ChIP analysis on MeCP2 binding to the PR-B gene. While the input control gave a strong signal, no signal was detected in the blank PCR reaction without a DNA template (Neg Ctl, negative control). A weak signal was found in the IgG control (IgG). Knockdown of MeCP2 resulted in a reduction of MeCP2 binding to the PR-B gene compared to the cells transfected with control siRNA. c PR-B mRNA levels were examined in the KLE cells by real-time PCR. A significantly increased PR-B mRNA level was observed following MeCP2 knockdown (* p < 0.05)

Determination of MeCP2 occupancy of the PR-B genes in PR-B-negative and -positive EC tissues

To further verify the MeCP2 involvement in PR-B silencing in EC cells, we examined the PR-B-negative and PR-B-positive EC tissues diagnosed by the immunostaining with PR antibodies (data not shown). We first measured the PR-B mRNA levels in these samples to confirm their PR-B expression status (Fig. 7a). The real-time PCR results indicated that the four PR-B-positive EC tissues indeed contained much higher levels of PR-B mRNA than the PR-B-negative EC tissues. The methylation status of the PR-B gene was determined by MS-PCR. As shown in Fig. 7b, the four PR-B-positive EC tissues displayed a largely ummethylated pattern (top panel). In the four PR-B-negative EC tissues, both the primers designed for the unmethylated (U) and methylated (M) PR-B genes detected some signals, but the signals for methylated PR-B gene was stronger than those for unmethylated PR-B gene, suggesting a partial, instead of complete, methylation of the PR-B gene in these tissues. In the PR-B-negative (HEC-1B) cell line, which was included as a control, the primers for methylated, but not the primers for unmethylated, PR-B genes, detected strong signals, validating the accuracy of the MS-PCR. Densitometry analyses of the data demonstrated a more than six-fold higher level of methylation index (a ratio of DNA bands representing methylated PR-B verses that represent unmethylated PR-B) in the PR-B-negative cancer tissues than in PR-B-positive tissues (Fig. 7c, p < 0.01).

Fig. 7.

Differential MeCP2 binding to PR-B genes in PR-B-negative and PR-B-positive EC tissues. a Measurement of PR-B mRNA levels with real-time PCR confirmed that all the four PR-B-negative EC tissues expressed significantly lower PR-B mRNA levels than the four PR-B-positive tissues, confirming the diagnosis of the PR-B expression status by immunostaining. b Results of methylation-specific PCR. In PR-B-positive EC tissues, PCR with primers for unmethylated PR-B- genes (U) generated much stronger signals than primers targeting the methylated PR-B (M). In contrast, in PR-B-negative EC tissues, while both sets of primers produced some products, primers targeting methylated PR-B generated stronger signals than primers for unmethylated PR-B. DNA from the PR-B-negative HEC-1B cells was included as a control. c Quantitative analyses using MS PCR results indicated that on the average, the PR-B-negative tissues have a significantly higher methylation index than the PR-B negative tissues (** p < 0.01). d ChIP analysis on MeCP2 binding to PR-B genes. As negative and positive controls, PCR were performed without DNA templates or with genomic DNA isolated from EC tissue sample 33A (33A gDNA). For each EC tissue, ChIP experiments with input genomic DNA and nonspecific antibodies (anti-IgG) were included as internal controls. e Quantitative analyses of the ChIP results indicated that on average, MeCP2 binding to PR-B genes is significantly stronger in PR-B-negative than in PR-B-positive tissues (** p < 0.01)

During ChIP analyses of EC tissues, we purposely performed the experiment on PR-B-positive and -negative tissues side by side with an alternate order of the two types of tissues to achieve a more accurate comparison, thus avoiding possible bias caused by variations in ChIP procedures as well as gel electrophoresis. As shown in Fig. 7d, PCR detected a stronger MeCP2 binding to PR-B promoters in PR-B-negative cancer tissues than in PR-B-positive tissues. Densitometry analyses of the results showed a four-fold difference in the average MeCP2 binding between the two groups (Fig. 7e, p < 0.01). These results confirmed that MeCP2 binding to the PR-B gene is intimately associated with PR-B gene methylation as well as the decreased PR-B expression in native EC tissues, offering an in vivo pathological relevance for our findings in EC cell lines on the critical role of MeCP2 for PR-B gene silencing.

Discussion

DNA methylation and covalent histone modifications represent the two critical sets of epigenetic events leading to gene silencing. Two potential models concerning the functional interactions between DNA methylation and histone modification have been proposed [40–42]. The first model suggests that histone modifications play a dominant role over DNA methylation during epigenetic regulation. One observation in support of this model is the requirement of Dim-5, a histone methyltransferase, for DNA methylation in fungus [43]. Another line of evidence comes from the DNA demethylation following administration of HDAC inhibitors [44, 45]. In Neurospora crassa, TSA treatment caused a selective DNA demethylation and activation of specific genes [46]. Moreover, we previously observed that HDAC inhibitor treatment of EC cells caused destabilization of DNMT3B mRNA, providing one mechanism by which TSA could induce DNA demethylation [39].

The second model emphasizes a central role of DNA methylation for epigenetic silencing [40]. This model is supported by the observation that DNA demethylation induced by either chemicals or the knockdown of DNMT often resulted in alterations of histone modifications in mammalian cells [47–49]. Mechanistic studies have shown that when bound to methylated DNA, MBDs are capable of forming protein complexes with enzymes catalyzing histone acetylation and/or methylation [50, 51]. According to this model, MBDs actively recruit histone modification enzymes to methylated genomic domains, leading to chromatin remodeling, chromatin conformation changes, and ultimately, transcriptional silencing. Our data showing that ADC-induced PR-B-gene demethylation resulted in dramatic changes in histone modification as well as increased PR-B expression appear to support the DNA methylation-directed gene silencing.

It should be emphasized that while DNA methylation may serve as an initiating event, changes in histone modification codes following ADC treatment are still required for transducing the DNA methylation signals to transcriptional suppression. In fact, TSA treatment is able to activate the PR-B gene (Fig. 2a) without significant disturbance of DNA methylation patterns (Fig. 2b). Such an “override” or “bypass” of DNA demethylation by HDAC modification has been previously observed in other genes and cell types [52, 53], suggesting that DNA demethylation is not a constant prerequisite for gene activation, at least under the pharmacological conditions. At this time, it is not clear if this bypass mechanism bears any physiological significance. Further studies investigating the cause-effect relationship between DNA methylation and histone modification are required to define the hierarchical order of these epigenetic events in the epigenetic regulation of the PR-B gene.

H3-K9 methylation and H3-K4 demethylation are considered a hallmark of transcriptionally inactive genomic domains found in heterochromatin, imprinted genes, and silenced genes in cancer cells [54–57]. Our observations on the H3 methylation patterns in active and silenced PR-B genes are consistent with this concept. The functional significance of the opposite changes in H3-K4 and H3-K9 methylation in PR-B gene silencing is supported by the results from drug treatment experiments (Fig. 4). When PR-B-negative cells were treated with ADC, re-activation of PR-B gene expression (Fig. 2a) was accompanied by the changes in H3-K4 and H3-K9 methylation (Fig. 4). Interestingly, the effects of TSA treatment were not limited to changes in histone acetylation, rather, H3-K4 and K9 methylation in the PR-B gene was altered as well (Fig. 4). This result appears to be contradictory to a previous study showing that TSA alone had little effect on H3-K9 methylation in methylated P16, MLH1, and O6-methylguanine-DNA methyltransferase genes in a colorectal cancer cell line [58]. It is noteworthy that only a moderate increase in histone acetylation and minimal effects on gene expression were observed after treatment with TSA in that study. The discrepancy may be explained by divergent cell sensitivity to TSA treatment and/or different drug treatment conditions applied in the two studies.

Our observations are reminiscent of the findings from two other laboratories. Satoshi Fujii et al. found that the epigenetically silenced ARHI gene in breast cancer cell lines was efficiently reactivated by TSA treatment. This result was accompanied by an increase in H3 acetylation and a significant reduction in H3 K9 methylation [59]. Similarly, Jiwen Li et al. reported that transcriptional repression by an unliganded thyroid hormone receptor is associated with a substantial increase in H3-K9 methylation, as well as a decrease in H3-K4 methylation. Moreover, TSA treatment resulted in not only histone hyperacetylation, but also an increase in H3-K4 methylation and a simultaneous reduction in H3-K9 methylation [60]. These observations point to the existence of a cross-talk between histone hyperacetylation and H3-K9 demethylation during TSA-induced re-activation of a methylated and transcriptionally silenced PR-B gene in at least some cancer cells.

The mechanism by which an HDAC inhibitor like TSA could cause histone methylation modifications is unclear. One possible mechanism may involve chromatin conformation changes. H3 hyperacetylation may trigger re-configuration of the local chromatin structure; this “in cis effect” may affect methylation modification by allowing or preventing access of methylation modification to the key lysine residues. Alternatively, H3 acetylation itself, or a related change in another chromatin component, could mediate the active recruitment/release of histone methyltransferases/demethylases. Histone demethylase KDM7 and/or LSD1 with substrate specificity for methylated histones [35, 37, 61] may be candidate enzymes carrying out the active H3-K9 demethylation. In this context, the related mechanisms may explain the ADC effects on histone methylation changes. ADC treatment-caused DNA methylation pattern alterations will readily lead to differential MBD binding, MBDs-mediated recruitment of histone acetylation modification enzymes. Alterations in histone acetylation status may affect the methylation status by the “in cis effect”. Direct recruitment of histone methylation enzymes by MBD may provide an alternative mechanism.

Previous studies have documented divergent promoter binding profiles for different MBD family members [62, 63]. The ability of these factors to form extensive protein–protein interactions adds another dimension to the complexity of their interactions with genomic DNA. MBDs have been identified as a component of the transcription repression complexes Sin3, Mi-2/NuRD, and SWI2/SNF2 that also contain chromatin remodeling enzymes such as ATPase and histone deacetylase [51, 64]. In addition, tissue-specific expression of these factors [62, 65] may contribute to their differential binding to a given promoter. In this study, we observed markedly diverged PR-B gene binding patterns by individual MBDs. MBD3 and MBD4 displayed little binding to methylated CpG islands in PR-B gene promoter, suggesting that these MBDs may be less likely involved in DNA methylation-mediated PR-B regulation. Since MBD3 and MBD4 are expressed in all EC cell lines examined, their low occupancy of methylated PR-B gene in PR-negative cell lines may be caused by differential posttranslational modification and/or constraints in their access to a PR-B promoter, e.g., a restricted cellular localization. It is noteworthy that since the MS-PCR only examines the methylation status of a limited number of CpG sites, and ChIP assays in this study only cover a relatively small region containing 108 nucleotides and four CpG sites, we could not exclude the possibility that the non-MeCP2 MBDs, including MBD3 and 4, may nevertheless contribute to PR-B regulation by binding to other methylated region(s) of PR-B genes. On the other side, the possible divergence in DNA methylation levels along the PR-B gene, e.g., a lower level of PR-B methylation in the two PR-B-negative cell lines, could contribute to the relatively low reactivation of PR-B in response to MeCP2 knockdown (Fig. 6c).

MeCP2 is a multifunctional, high-affinity, methylated-DNA binding protein implicated in gene silencing, activation, chromatin architecture, and regulation of RNA splicing [66]. MeCP2 mutation has been linked to the classic Rett syndrome, variant Rett syndrome, and mild learning disabilities [67]. Several studies have demonstrated that MeCP2 expression in mammalian cells is tightly controlled, and even a small change in MeCP2 levels may lead to functional alterations in the cell. Alessandro Brero et al. have shown that pericentric heterochromatin aggregates during myogenic differentiation. This clustering leads to the formation of large chromocenters and correlates with increased levels of MeCP2 and pericentric DNA methylation. Ectopic overexpression of MeCP2 mimicked this effect, causing a dose-dependent clustering of chromocenters in the absence of differentiation [68]. In another study, Ann Collins et al. [69] examined the role of MeCP2 in a mouse model and found that the moderate overexpression of MeCP2 caused a progressive neurological disorder. In the current study, Western blot results demonstrate that MeCP2 expression is remarkably stable (Fig. 5), which may suggest a critical role(s) of this protein for cancer cell functions. Indeed, it has been reported that MeCP2 is directly involved in cell apoptosis [70]. The requirement for a stringent control of MeCP2 levels seemed to be reflected by the cell response to forced MeCP2 knockdown (Fig. 6) in this study. Reduction of MeCP2 levels by siRNA caused a decreased MeCP2 binding to the methylated PR-B gene in KLE cells, and as a result, the re-activation of PR-B gene transcription.

In this study, we demonstrated a close association between MeCP2 binding and PR-B transcription silencing. Although Western blot analyses indicated comparable concentrations of MeCP2 protein in the PR-B-positive and -negative cell lines, and its expression levels were not significantly altered by drug treatment (Fig. 5), MeCP2 selectively occupied the methylated, but not unmethylated PR-B gene, indicating a methylation-dependent binding pattern. The specific interaction between MeCP2 and methylated PR-B is demonstrated by decreased MeCP2 binding when PR-B undergoes demethylation following ADC treatment (Fig. 3). Follow-up studies indicated that MeCP2 knockdown led to a reduced MeCP2 binding to PR-B and a concomitant increase of PR-B expression (Fig. 6). The significance of MeCP2 binding to methylated PR-B during PR-B silencing is also supported by results from native EC samples. Since progesterone deficiency is a well-recognized etiologic factor for EC, and PR-B silencing may contribute to EC development, it would be of great interest to examine the expression and role(s) of MeCP2 in EC pathogenesis on larger-scale studies using more clinical specimens.

Compared to MeCP2, MBD1 and MBD2 displayed complicated PR-B binding patterns. MBD1 and MBD2 were present at both the methylated and unmethylated PR-B genes (Fig. 3). Thus, interaction with PR-B gene by these two factors with the PR-B gene does not seem to be strictly methylation-dependent. Although MBD1 and MBD2 were originally isolated as methyl CpG binding proteins, their abilities to bind to unmethylated DNA have been observed in subsequent studies [71, 72]. It was proposed that these proteins use distinct mechanisms to interact with target DNA. Mouse MBD1 uses CXX3-3 rather than the MBD domain to interact with the methylated, as well as unmethylated CpG islands [71]. The MBD2 gene produces two splicing isoforms, MBD2a and MBD2b. Interestingly, the full-length MBD2a has been shown to interact with RNA helicase A and promote the activation of unmethylated CREB-responsive genes [71]. Our results showed that both factors responded to ADC treatment (MBD1 in KLE and HEC-1B, MBD2 in KLE), which suggested that they nevertheless could be involved in the regulation of the methylated PR-B gene. Due to the limited amounts of EC tissues available for this study, we couldn’t afford to use the ChIP assay to confirm these observations in clinical samples. On the other hand, their possible interactions with the unmethylated PR-B gene point to an alternative, methylation-unrelated mechanism in Ishikawa cells. Thus, it is fully possible that MBD1 and MBD2 may participate via different regulation pathways targeting the methylated, as well as the unmethylated genes. Indeed, while MeCP2 is knocked down to <10 % of the original level, PR-B expression was only moderately increased (Fig. 6), suggesting the possible involvement of additional MBDs, e.g., MBD1 and MBD2. The nature and significance for MBD1- and MBD2-mediated PR-B regulation, particularly, the implication of their binding to the unmethylated PR-B gene, are intriguing questions that require further investigation.

It was previously observed that ADC treatment could affect H3-K9 acetylation and methylation in unmethylated genes via a mechanism unrelated to its ability to induce DNA demethylation [73]. We did not observe such effects in the current study (Fig. 4). Instead, our data showed that the histone modification pattern in the unmethylated PR-B gene in Ishikawa cells was rather stable. Local histone acetylation at the PR-B gene remained unchanged even when the global histone acetylation level was increased by TSA treatment. This is in sharp contrast to the data from the methylated PR-B gene in KLE and HEC-1B cells. One possible explanation is that the histone acetylation in the active PR-B gene has reached a saturated level and has become insensitive to further signaling for additional acetylation. This opinion is consistent with the PR-B expression data showing that TSA treatment failed to induce substantial changes in chromatin composition (Fig. 4) or further increase of PR-B expression in PR-B-positive Ishikawa cells (Fig. 2a and data not shown). Thus, HDAC and DNMT inhibitors are capable of re-activating the methylation-silenced PR-B gene in cancer cells, but do not interfere with PR-B expression in PR-B-positive cells. This selective action on the silenced PR-B gene may mediate a cancer cell-specific effect in vivo, which may prove to be a benefit when applying epigenetic inhibitors to treat the PR-B-negative endometrial cancers.

Liang et al. [74] reported that H3-K4 methylation and H3-K4 and H3-K14 acetylation are highly concentrated at the 5′-region of transcriptionally active genes, but are greatly decreased in the downstream of the transcription start site. This finding indicates that the histone modification profile may vary along the sequences in a given gene. It is important to point out that the current investigation focuses on the CpG island close to the transcription start site. Further studies on extended regions are required to draw a complete picture of the entire PR-B locus including the upstream region, the transcribed region, and the 3′-prime regions of the PR-B gene. Nevertheless, our investigation has identified a distinct set of epigenetic changes in local chromatin associated with PR-B silencing. The finding that both MBD factors and histone modifications participate in PR-B silencing may explain the previously observed synergism between DNMT and HDAC inhibitors in PR-B re-activation [29]. Given the strong involvement of MeCP2 in PR-B suppression, it is possible that the knockdown of MeCP2 expression or blocking of its binding to the methylated PR-B gene may lead to the re-activation of the PR-B gene and therefore, a possible re-sensitization of PR-B-negative endometrial cancers to progestational therapy. Such epigenetic approaches may be especially effective for treating advanced or recurrent EC cases that are often PR-B–negative, and account for most deaths caused by endometrial cancers.