Tissue-Specific Expression of Estrogen Receptor 1 Is Regulated by DNA Methylation in a T-DMR

Tissue samples

Specimens of endometrium, mammary gland, placenta, and skin were obtained from 35 Japanese women, and breast cancer tissue was obtained from 17 Japanese women. Endometrium was obtained from patients who underwent surgery for uterine leiomyoma. Mammary gland tissue and breast cancer tissue were obtained from patients who underwent mastectomy for breast cancer. Placenta tissue was obtained from patients with normal delivery. Skin tissue was obtained from operation scar with a small piece of normal skin tissue at repeated cesarean section. Specimens were dissected immediately after the removal, immersed in liquid nitrogen and stored at −80°C until DNA/RNA extraction as previously reported (13, 14).

Cell culture

ER-α-positive breast cancer cell line (MCF7) was obtained from the Department of Yamaguchi University Center for Gene Research and was cultured in DMEM (Sigma) supplemented with 10% fetal calf serum (FCS) (Biological Industries Ltd), 0.1mM nonessential amino acids (GIBCO BRL), 1mM sodium pyruvate (GIBCO), 100-U/mL penicillin, and 100-mg/mL streptomycin at 37°C under 5% CO₂ in a humidified atmosphere. ER-α-negative breast cancer cell line (MDA-MB-231; ATCC) was cultured in L-15 medium (ATCC) supplemented with 10% FCS, 100-U/mL penicillin, and 100-mg/mL streptomycin under the same conditions. Endometrial stromal cells (ESCs) were isolated from the human proliferative phase endometrium as previously reported (34) and cultured in DMEM (Sigma) supplemented with 10% FCS, 4mM glutamine, 50-μg/mL streptomycin, and 50-U/mL penicillin.

Sodium bisulfite sequencing

Bisulfite reactions were performed using a EpiTect Bisulfite kit (QIAGEN) according to the conditions as follows: 95°C for 5 minutes, 65°C for 85 minutes, 95°C for 5 minutes, and 65°C for 175 minutes as previously reported (15, 35). The bisulfite-converted DNA was amplified by PCR using 6 primer pairs for region I to VI shown in Supplemental Table 1 under the thermocycling conditions (95°C for 10 min, and 40 cycles of 94°C for 30 se, 60°C for 30 s, and 72°C for 1 min followed by 10 min of final extension at 72°C). The resulting products were cloned into pGEM-T easy vector (Promega). After sequencing reaction using a BigDye Terminator V3.1 kit (Applied Biosystems), sequencing was performed with a 3130xl Genetic Analyzer (Applied Biosystems) as previously reported (11, 13, 14). QUMA (http://quma.cdb.riken.jp/) was used to analyze the bisulfite sequencing data (36). The percentage of the methylated CpG sites in a total of the examined CpG sites was calculated and defined as follows: 5%>: unmethylation, 5%–30%: hypomethylation, 31%–70%: moderate methylation, and 71%<: hypermethylation.

RT-PCR and quantitative real-time RT-PCR (qRT-PCR)

Total RNA was isolated from tissues using Isogen (Wako Pure Chemical Industries Ltd), and 1-μg total RNA was reverse-transcribed using a Quantitect Reverse Transcription kit (QIAGEN) according to the manufacturer's protocol as previously reported (13). To distinguish the transcribed products derived from each upstream exon (see Figure 3 below), we synthesized 3 primer pairs as shown in Figure 3 below and Supplemental Table 2 for ESR1. We also made the primer pair inside of Exon1 to amplify any types of transcription variants (Supplemental Table 2). A primer pair for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as an internal control (Supplemental Table 2). PCR was performed under the thermocycling conditions (initial cycle of 94°C for 10 min, then 22–35 cycles of 94°C for 30 s, 55°C–60°C for 30 s and 72°C for 1 min followed by final extension for 10 min at 72°C). The resulting products were subjected to agarose gel electrophoresis. Real-time qRT-PCR was performed using SYBR Premix Ex Taq (TAKARA) and a LightCycler (Roche Applied Science) in ESR1 variants and DNA methyltransferases (DNMTs), including DNMT1, DNMT3A, and DNMT3B, using the primer pairs shown in Supplemental Table 2. All samples were run in duplicate. Melting curves of the products were obtained after cycling by a stepwise increase of temperature form 55°C to 95°C.

Reporter assay for the promoter activity and T-DMR-methylated reporter assay

First, we investigated whether the promoter region has transcriptional activities in the absence of the T-DMR1. To make a construct which only has the promoter region of ESR1, the promoter region was inserted into pGL3 basic vector (Promega), and the construct was designated as pGL3-Promoter. The construct with both the promoter region and the T-DMR was also made and designated as pGL3-Promoter-TDMR. Uterine smooth muscle cells (Lonza) were cultured on a 24-well plate (8 × 10⁴ cells/well) for 24 hours and then transfected with 450 ng of the reporter constructs linearized by Aor5HI (Takara) and 30 ng of pRL-TK vector (Promega) as a normalization control using Lipofectamine2000 (Invitrogen). Twenty-four hours after the transfection, the luciferase activities were measured using a Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer instruction. The assays were performed in duplicate and repeated 3 times.

Secondly, we examined the effect of DNA methylation at the T-DMR1 on the luciferase activity in the presence of the promoter region (corresponding to AB-promoter). For this purpose, T-DMR-methylated reporter assays were performed as previously reported (see Figure 2 below and Supplemental Figure 1) (15). Briefly, a 5′-flanking region (−651 to +37) of the ESR1 gene and the T-DMR1 region from −1244 to −638 were amplified and designated as ESR1-promoter and ESR1-Tdmr, respectively. Both ESR1-promoter and ESR1-Tdmr fragments were inserted together into pGL3 basic vector (Promega). The construct was amplified using the dam⁻ and dcm⁻ Escherichia coli strain SCS110 (Stratagene). The Tdmr fragment (−1244 to −638), excised from the construct, was mock treated or treated with SssI methylase (New England Biolabs). The resultant unmethylated and methylated Tdmr fragments were religated with the linealized vector fragment to create 2 types of reporter constructs: pGL3-TDMR-M (methylated) and pGL3-TDMR_U (unmethylated). The transfection with the constructs was done in uterine smooth muscle cells and the luciferase activities were measured as described above.

Western blotting

Western blotting was performed as previously reported (37). Briefly, whole-cell lysates were prepared using loading buffer reagents (Santa Cruz Biotechnology, Inc) without trypsin treatment. Equal amounts of total protein were electrophoresed on 10% sodium dodecyl sulfate-polyacrylamide gel. After SDS-PAGE was completed, proteins were transferred to the polyvinylidene difluoride membrane (ATTO) with semidry-type blotting system. The transferred membranes were stained by immunochemical technique consisting of the next procedure: after blocking the membrane with blocking solution (5% skimmed milk with 0.1% Tween-20 dissolved in Tris-buffered saline; pH 7.5), the blotted membranes were incubated with antibodies for DNMT1, DNMT3A, Early growth response protein 1 (EGR1) (Cell Signaling Technology), and β-tubulin (Sigma), which were diluted in blocking solution. Then, these membranes were incubated with the peroxidase-conjugated second antibody diluted in blocking solution. Finally, an enhanced chemiluminescence-Western blotting detection system (Amersham) was applied according to the manufacturer's protocol, after which the membranes were exposed to hyperfilm-enhanced chemiluminescence (Amersham). To reuse the blot, the membranes were stripped in Restore Western stripping buffer (Pierce Biotechnology).

Chromatin immunoprecipitation (ChIP) assay

ChIP assays were performed using the SimpleChIP Enzymatic Chromatin IP kit (Cell Signaling Technology) as previously reported (11). Cells were cross-linked by addition of formaldehyde into the medium, and cross-linking was terminated by addition of glycine (0.125M, final concentration). After the nuclear extraction, Micrococcal Nulease was added to digest DNA. Two percent of the supernatant was kept as input controls. The chromatin was incubated with the next antibodies at 4°C overnight: a transcription factor, EGR1 (Cell Signaling Technology) and H3K27me3 (generous gift from Dr Kimura, Tokyo Institute of Technology) (38). Normal rabbit IgG (Cell Signaling Technology and Invitrogen, respectively) was used as a negative control. Immune complexes were collected, and cross-linkings of the immunoprecipitated chromatin complexes (IP) and input controls (INPUT; 2% of the total soluble chromatin) were reversed by heating the samples at 65°C overnight. The DNA fragments were purified and subjected as a template for PCR amplification. The specific primers used for ChIP-PCR analysis to amplify the regions in the T-DMR1 and T-DMR2 (Supplemental Table 3). The relative levels of each target sequence were analyzed by real-time PCR. The cycle threshold (Ct) of IP sample was used for quantification and the value were normalized for their corresponding INPUT value using the next formula (IP/INPUT = 2 ^{CtINPUTDNA − CtIPDNA}) as we reported previously (34).

Lipid-mediated transfection of small interfering RNA (siRNA) duplexes

Knockdowns of EGR1 were performed using siRNA method as reported previously (34). EGR1 ON-TARGET plus SMART pool and ON-TARGET plus Non-Targeting pool siRNA were purchased from Dharmacon. ESCs were plated in the medium lacking antibiotics at approximately 3 × 10⁵ cells in 25-cm² tissue culture flasks and incubated for 2 days. At 30%–50% confluence, siRNA duplexes (20nM) and RNAi MAX (2.5 μL/well; Invitrogen) diluted in Optimem (Invitrogen) were transfected to ESCs at 60% confluence. After the medium was changed 4 hours later, cells were incubated for 72 hours.

Statistical analyses

A paired t test was performed for the comparison between the 2 groups in Figures 3 and 5 below using SPSS for Windows version 11 (SPSS, Inc). Differences were considered significant at P < .05.

ESR1 mRNA expression and DNA methylation statuses of ESR1 promoter

ESR1 mRNA expression was examined in the endometrium, mammary gland, placenta, and skin. ESR1 mRNA expression was high in the endometrium and mammary gland and low or negligible in the placenta and skin (Figure 1A), indicating that ESR1 expression is tissue-specific. We then tested the DNA methylation status of the promoter region (−566 to +229 bp) and the distal region (−1188 to −790 bp) from TSS by sodium bisulfite sequencing in these tissues (Figure 1B). In all of the tissues, the promoter region from −566 to +229 bp was unmethylated, whereas in the distal region from −1188 to −790 bp, the DNA methylation status was variable among the tissues (Figure 1B). Endometrium and mammary gland showed unmethylation and hypomethylation statuses, whereas placenta and skin showed moderate methylation and hypermethylation statuses, respectively. An analysis of additional samples confirmed that the DNA methylation status in the distal region varied among these tissues (Figure 1C). The DNA methylation status in the distal region was strongly associated with ESR1 expression, which suggests that the distal region is a T-DMR. We thus designated it as T-DMR1.

T-DMR-methylated reporter assay

As shown in Figure 2A, the pGL3-Promoter+T-DMR construct had significantly higher reporter activities (P < .05) compared with the empty vector, whereas the promoter alone did not induce the reporter activity, indicating T-DMR1 is essential for the enough reporter activity.

Because the results of Figures 1 and 2A indicated that T-DMR1 is essential for the transcriptional activity of ESR1 and the DNA methylation status of T-DMR1 is associated with ESR1 expression, we examined whether DNA methylation of T-DMR1 actually has an inhibitory effect on ESR1 expression. For this purpose, T-DMR-methylated reporter assay was performed to demonstrate that the DNA methylation of T-DMR1 alone, whereas keeping the promoter region hypomethylated has a suppressive effect on reporter activities. A DNA fragment corresponding to the T-DMR1 was methylated in vitro and inserted into a reporter construct with an unmethylated promoter region (pGL3-TDMR-M construct) (Figure 2B). Unmethylated T-DMR1 was used as a control (pGL3-TDMR-U construct). The unmethylated and methylated statuses of T-DMR1 in the pGL3-TDMR-U and pGL3-TDMR-M constructs were confirmed by sodium bisulfite sequencing (Figure 2C). The pGL3-TDMR-U construct (control) had 5-fold higher reporter activities compared with the empty vector, whereas the pGL3-TDMR-M construct reporter activity was significantly reduced by 47.5% when compared to the pGL3-TDMR-U construct (P < .05) (Figure 2D). These results indicate that DNA methylation of T-DMR1 suppresses ESR1 expression. Therefore, the distal region (−1188 to −790 bp) was determined as the T-DMR that regulates ESR1 gene expression.

DNA methylation statuses around upstream exons and mRNA expression of ESR1 variants in normal tissues

ESR1 has several upstream exons (upstream Exon-A to upstream Exon-E1), in which transcription starts from the corresponding TSSs (Figure 3A) (21). Three upstream exons, Exon-A, Exon-B, and Exon-C, are often used in the tissues with high ESR1 expression (21). The transcription products from upstream Exon-A, Exon-B, and Exon-C are spliced to generate mature ESR1 mRNAs, called variant1, variant2, and variant3, respectively (Figure 3B). ESR1 mRNA expression of variant1, variant2, and variant3 was analyzed by RT-PCR and qRT-PCR. All 3 variants were highly expressed in the endometrium and mammary gland, whereas none of the variants were expressed in the placenta and skin (Figure 3, C and D), indicating that the transcription from each upstream exon is tissue-specific.

Next, we analyzed DNA methylation status around upstream Exon-A, Exon-B, and Exon-C (−2953 to +229 bp). CpG sites located from −566 to +229 bp were unmethylated in all the tissues, which was similar to the result shown in Figure 1 (Figure 3E). We designated this region as AB-promoter. CpG sites of T-DMR1 (−1188 to −790 bp) were unmethylated in the endometrium, hypermethylated in the mammary gland, moderately methylated in the placenta and hypermethylated in the skin, which was also similar to the result shown in Figure 1 (Figure 3E). CpG sites in the region from −2099 to −1876 bp, corresponding to upstream Exon-C, were unmethylated or hypomethylated in all the tissues (Figure 3E). We designated this region as C-promoter. The region from −2953 to −2302 bp was unmethylated in the endometrium, hypomethylated in the mammary gland, and moderately methylated in the placenta and skin (Figure 3E), indicating that the region from −2953 to −2302 bp is another T-DMR that regulates variant3 expression. We designated this T-DMR as T-DMR2 (Figure 3E). From these findings, it is suggested that each upstream exon has its own T-DMR.

DNA methylation statuses around upstream exons and mRNA expression of ESR1 variants in breast cancer

To investigate whether DNA methylation of the T-DMRs is involved in the regulation of transcription of the ESR1 mRNA variants in breast cancer, we examined the mRNA expression levels of variant1, variant2, and variant3 in 17 breast cancer tissues by real-time qRT-PCR and DNA methylation statuses around upstream Exon-A, Exon-B, and Exon-C by sodium bisulfite sequencing. Seventeen tissue samples were classified into 3 groups (Figure 4A); all 3 variants were expressed (group X), none of the variants were expressed (group Y), and variant1 and variant2 were expressed (group Z). In this study, mRNA expression was defined as positive when the ratio of mRNA of each variant to GADPH is more than 0.3 according to the expression level of the mammary gland shown in Figure 3D. We also examined the mRNA expression of the 3 variants in MCF7 and MDA-MB-231 cells, which are known as an ER-α-positive and an ER-α-negative breast cancer cell line, respectively. As shown in Figure 4B, in MCF7, variant1 and variant2, but not variant3, were expressed whereas any of the variants were not expressed in MDA-MB-231. Then, we examined DNA methylation statuses of AB-promoter, C-promoter, T-DMR1, and T-DMR2 in each group and both cell lines (Figure 4C). In case BK-3 of group X expressing all 3 variants, the AB-promoter and C-promoter were unmethylated and hypomethylated, and T-DMR1 and T-DMR2 were hypomethylated and moderately methylated, respectively, similar to what was observed in the DNA methylation profiles of the endometrium and mammary gland. In group Y expressing none of the variants, T-DMR1 and T-DMR2 were hypermethylated as was observed in the placenta and skin, but interestingly, the methylation statuses of the C-promoter varied among the cases such as moderate methylation in cases BK-8, BK-10, BK-12, and BK-13, and unmethylation in case BK-11. In group Z, the DNA methylation pattern of case BK-17 was similar to that of case BK-3 of group X, but interestingly, case BK-17 expressed only variant1 and variant2. In addition, in MCF7, although both T-DMR1 and T-DMR2 were moderately methylated, MCF7 expressed variant1 and variant2 but not variant3. These findings indicate that transcriptional regulation of ESR1 expression is abnormal in some breast cancers. In other words, some cases of breast cancer may have other regulation mechanisms in ESR1 expression than DNA methylation at T-DMRs. Furthermore, in MDA-MB-231 expressing none of the variants, even B-promoter and C-promoter in addition to T-DMR1 and T-DMR2 were moderately methylated and hypermethylated, respectively. It is also suggested that DNA methylation aberrantly occurs in breast cancer.

DNMTs expression

ESR1 expression is reported to be activated in response to 5-aza-2′-deoxycytidine, an inhibitor of DNMTs, in ER-negative breast cancer cells (28), suggesting that DNMTs are involved in the regulation of DNA methylation of the ESR1 promoter region. Single nucleotide polymorphisms of DNMT1 are associated with the risk of triple-negative (ER-negative, progesterone receptor-negative, and Her2/neu-negative) breast cancers (39). These reports suggest that DNMTs are associated with the DNA methylation in ESR1 gene. Therefore, to investigate whether the expression levels of DNMTs are involved in the regulation of DNA methylation levels of the T-DMRs, we examined the mRNA and protein levels of DNMTs in the normal tissues and breast cancer tissues. Endometrium, mammary gland, placenta, skin, and breast cancer tissues expressed the mRNA of DNMT1 and DNMT3A, but the expression level of DNMT3B was low compared with DNMT1 and DNMT3A (Supplemental Figure 2A). There was no association between the mRNA or protein levels of DNMTs and the DNA methylation levels at the T-DMRs in normal tissues and breast cancer tissues (Supplemental Figure 2).

Potential transcription factors and histone modifications at T-DMRs

We searched the transcription factor that binds to the T-DMRs analyzing a transcription factor motif using The Multiple Em for Motif Eliciation (MEME) suite (http://meme-suite.org) (40), and EGR1 was extracted as a potential transcription factor. EGR1 has the consensus DNA sequence (GCGTGGGCG) in both T-DMR1 and T-DMR2 with extremely low P value and q value (0.00000561399 and 0.00343731, respectively) (Figure 5A). EGR1 belongs to the EGR family of C2H2-type zinc-finger proteins (41). EGR1 is a nuclear protein and functions as a transcriptional regulator (41). Therefore, to show that EGR1 actually regulates ESR1 expression, the effect of EGR1 knockdown on ESR1 expression was examined by transfecting EGR1 siRNA into human ESC, which express ESR1. As shown in Figure 5B, EGR1 was clearly suppressed by EGR1 siRNA transfection, and mRNA and protein expressions of ESR1 were significantly suppressed by EGR1 knockdown.

Because DNA hypomethylation within the promoter region of the gene facilitates the EGR1 binding to its consensus motif (42), we next investigated using a ChIP assay whether the binding of EGR1 to the T-DMRs is affected by DNA methylation at the T-DMRs. For this purpose, ESC isolated from the human endometrium expressing all 3 ESR1 variants and MDA-MB-231 expressing none of the ESR1 variants were used. As shown in Figure 5C, the EGR1 bindings to the T-DMR1 and T-DMR2 were significantly higher in ESC than in MDA-MB-231, suggesting that EGR1 binds to T-DMR1 and T-DMR2 in the human endometrium in which T-DMR1 and T-DMR2 are unmethylated, but the EGR1 bindings are inhibited in MDA-MB-231 in which T-DMR1 and T-DMR2 are hypermethylated.

DNA methylation is associated with a repressive histone modification, H3K27me3, which is involved in gene silencing (9, 10, 43). Therefore, we investigated the H3K27me3 status in T-DMR1 and T-DMR2 in ESC, in which T-DMRs are hypomethylated, and MDA-MB-231, in which T-DMRs are hypermethylated. The H3K27me3 levels in both T-DMR1 and T-DMR2 were significantly lower in ESC than in MDA-MB-231 (Supplemental Figure 3).