Down-regulation of PROS1 Gene Expression by 17β-Estradiol via Estrogen Receptor α (ERα)-Sp1 Interaction Recruiting Receptor-interacting Protein 140 and the Corepressor-HDAC3 Complex*

Pregnant women show a low level of protein S (PS) in plasma, which is known to be a risk for deep venous thrombosis. 17β-Estradiol (E2), an estrogen that increases in concentration in the late stages of pregnancy, regulates the expression of various genes via the estrogen receptor (ER). Here, we investigated the molecular mechanisms behind the reduction in PS levels caused by E2 in HepG2-ERα cells, which stably express ERα, and also the genomic ER signaling pathway, which modulates the ligand-dependent repression of the PSα gene (PROS1). We observed that E2 repressed the production of mRNA and antigen of PS. A luciferase reporter assay revealed that E2 down-regulated PROS1 promoter activity and that this E2-dependent repression disappeared upon the deletion or mutation of two adjacent GC-rich motifs in the promoter. An electrophoretic mobility shift assay and DNA pulldown assay revealed that the GC-rich motifs were associated with Sp1, Sp3, and ERα. In a chromatin immunoprecipitation assay, we found ERα-Sp protein-promoter interaction involved in the E2-dependent repression of PROS1 transcription. Furthermore, we demonstrated that E2 treatment recruited RIP140 and the NCoR-SMRT-HDAC3 complex to the PROS1 promoter, which hypoacetylated chromatin. Taken together, this suggested that E2 might repress PROS1 transcription depending upon ERα-Sp1 recruiting transcriptional repressors in HepG2-ERα cells and, consequently, that high levels of E2 leading to reduced levels of plasma PS would be a risk for deep venous thrombosis in pregnant women.

Protein S (PS) 2 is a vitamin K-dependent plasma protein that functions as a nonenzymatic cofactor for activated protein C in the down-regulation of the blood coagulation cascade via pro-teolytic inactivation of coagulant factors Va and VIIIa (1). PS has been also shown to display activated protein C-independent anticoagulant activity in purified systems as well as in plasma (2,3).
Over the past 2 decades, low levels of plasma PS have become a well established risk factor for the development of deep venous thrombosis (4 -6). Hereditary PS deficiency has been shown to be an autosomal dominant trait, and many causative genetic mutations have been described in the PS␣ gene (PROS1) (7). However, PS deficiency can also occur throughout life under acquired conditions such as oral anticoagulant use and liver disease (8). Furthermore, acquired PS deficiency has been reported in individuals with high levels of estrogen during pregnancy and in those taking oral contraception (9 -11).
The major source of circulating plasma PS is the hepatocyte (12), but PS is also produced constitutively at low levels by a variety of other cell types throughout the body (13)(14)(15)(16)(17). PS circulates in human plasma at a concentration of 0.35 M in a free form (40%) and a C4b-binding protein-bound form (60%) (18,19). Two copies of the PS gene located on chromosome 3, the active PS␣ gene (PROS1) and the inactive PS␤ pseudogene (PROS2), share 96% homology in their coding sequences (20 -22). The promoter and first exon are absent from the PROS2 gene, and the promoter region of PROS1 has been poorly investigated in contrast to the coding regions. It has been reported that transcription from the PROS1 promoter is directed from multiple start sites and that the PROS1 5Ј-flanking region lacks the characteristic "CAAT" and "TATA" boxes (23).
Estrogens are important regulators of mammalian growth and metabolism, accomplishing these functions by controlling the expression of specific genes via estrogen receptors (24). The estrogen receptor (ER) is a member of the steroid/nuclear receptor superfamily of transcription factors and is required for the mediation of 17␤-estradiol (E 2 )-induced responses in multiple tissues and organs (25). The classical ER mechanism of action involves ligand-induced formation of an ER homodimer that interacts with estrogen-responsive elements (EREs) in target gene promoters and recruits cofactors necessary for transactivation (25). There is increasing evidence that the formation of a classical genomic ER-ERE complex is only one of several genomic and non-genomic pathways of estrogen actions (26 -28). Genomic ER associates with other transcription factors such as the activating protein-1 (AP-1) complex, nuclear factor B (NFB), and specificity proteins (Sp) to modulate ligand-dependent gene expression (26,27,29). In this study, we investigated the molecular mechanisms of the reduction in PS production caused by E 2 as well as the genomic ER signaling pathway that modulates ligand-dependent PROS1 gene repression in HepG2-ER␣ cells stably expressing human ER␣.

EXPERIMENTAL PROCEDURES
Cell Culture and Reagents-Human hepatoma cell line HepG2 and human breast cancer cell line MCF7 were purchased from American Type Culture Collection (ATCC, Manassas, VA). The cells were cultured in Dulbecco's modified Eagle's medium (DMEM; Wako, Tokyo) supplemented with 5 or 10% fetal bovine serum (FBS; JRH Biosciences, Lenexa, KS) and 100ϫ antibiotic-antimycotic mixed stock solution (Nacalai Tesque, Kyoto, Japan). Human normal hepatocytes (hNhep) were purchased from Lonza (Walkersville, MD) and cultured in collagen I-coated dishes using the HCM TM BulletKit according to the manufacturer's protocols. For estrogen assays, cells were cultured in phenol red-free DMEM (Invitrogen) supplemented with 10% charcoal-stripped FBS (CS-FBS) for 3 days. Next day, the cells were cultured in phenol red-free DMEM supplemented with 1% CS-FBS and treated with ethanol (vehicle) or 100 nM E 2 for 48 h. E 2 and trichostatin A (TSA) were purchased from Sigma-Aldrich, and ICI 182,780 was from Tocris Bioscience (Ellisville, MO).
pER␣ was transfected by a calcium precipitation method as described previously (32), and stable transfectants were established from HepG2 cells through selection with G418. All clones were checked by Western blot analysis as described below, and the subclone showing the highest level of ER␣ (HepG2-ER␣) was used for further study.
Enzyme-linked Immunosorbent Assay for Measurement of PS-HepG2-ER␣ cells were treated with E 2 or vehicle (ethanol only) for 48 h, and the culture medium was harvested. A rabbit IgG against human PS (Dako, Carpinteria, CA) was biotinylated with an ECL TM protein biotinylation module (Amersham Biosciences). An unlabeled anti-PS polyclonal antibody (2.4 ng/100 l) in bicarbonate buffer (15 mM Na 2 CO 3 , 35 mM NaHCO 3 , and 3 mM NaN 3 ) was coated onto each well of a microtiter plate (Nunc, Roskilde, Denmark). After three washes with 150 l of Tris-buffered saline (TBS; 50 mM Tris, pH 7.4, 150 mM NaCl), the wells were blocked with 1% bovine serum albumin in TBS and then incubated with the plasma standards or culture medium samples. After three more washes with TBS, the biotinylated anti-PS antibody (0.1 g/100 l) was added followed by diluted (1:1000) streptavidin-horseradish peroxidase conjugate (Amersham Biosciences). After incubation, the substrate buffer (0.65 mg/ml o-phenylenediamine (Wako) and 0.06% H 2 O 2 in 0.1 M citrate, 0.2 M sodium phosphate buffer, pH 5.0) was added to each well. After further incubation at room temperature for 20 min, the peroxidase reaction was stopped by the addition of 50 l of 2 M H 2 SO 4 , and absorbance was measured at 490 nm.
Quantitative RT-PCR for Measurement of PS mRNA-Total RNA of the cells was extracted using an RNeasy mini kit (Qiagen, GmbH, Germany). The first-strand cDNA was prepared with 5 g of total RNA using the SuperScript III first strand system (Invitrogen). Quantitative RT-PCR was performed with a Power SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA), and ABI PRISM 7000 sequence detection systems (Applied Biosystems) were used for measurement. Relative PS mRNA was calculated as the respective PS mRNA/GAPDH (glyceraldehyde-3-phosphate dehydrogenase) mRNA as described previously (33).
After a 6-h transfection, the cells were washed and treated for 48 h with fresh phenol red-free DMEM supplemented with 1% CS-FBS containing 100 nM E 2 , 100 nM E 2 /1 mM ICI 182,780, 1 mM ICI 182,780 only dissolved in ethanol, or ethanol alone as a vehicle control. The cells were harvested, and subsequently luciferase activity was determined with a luciferase assay system (Promega) according to the manufacturer's directions. Luciferase activity was normalized to the activity of co-transfected ␤-galactosidase as an internal control for transfection efficiency.
Transient Transfection of siRNA-HepG2-ER␣ cells were cultured in phenol red-free DMEM with 10% CS-FBS and transfected with siRNA against Sp1 or Sp3 (Ambion, Austin, TX) or nonspecific siRNA using Oligofectamine reagent (Invitrogen) according to the manufacturer's directions.
Electrophoretic Mobility Shift Assay (EMSA)-Nuclear extracts were prepared from HepG2-ER␣ cells using NE-PER nuclear and cytoplasmic extraction reagents (Pierce) and stored in aliquots at Ϫ80°C until further use. The protein concentration of the nuclear extracts was measured with the Bio-Rad protein assay kit (Bio-Rad). DNA probes containing the PROS1 promoter fragment (from Ϫ176 to Ϫ147) were synthesized, biotinylated, and annealed. EMSA was performed according to a method described previously (32). Briefly, nuclear extract (5 g) and a biotin-labeled double-stranded DNA probe (600 fmol), with or without an unlabeled competitor, were treated with a LightShift TM chemiluminescent EMSA kit (Pierce) according to the manufacturer's instructions. In supershift experiments, the nuclear extract was incubated on ice for 10 min with the biotin-labeled double-stranded DNA probe after which an anti-Sp1, anti-Sp3, or anti-ER␣ antibody was added, and the incubation was continued for another 20 min. Samples were loaded on a 6% nondenaturing polyacrylamide gel in 0.5ϫ TBE buffer (0.089 M Tris borate, pH 8.0, 0.089 M boric acid, and 10 mM EDTA) and electrophoresed for 3.5 h at 100 V. Biotin-labeled DNA probes were transferred to Hybond TM -Nϩ membranes (Amersham Biosciences) and then integrated with streptoavidin-horseradish peroxidase conjugate.
DNA Pulldown Assay-The DNA pulldown assay (DNA affinity precipitation; DNAP assay) was carried out with biotinlabeled DNA probes as described previously (34). The nuclear extracts (100 g) were prepared from HepG2-ER␣ cells and incubated with biotin-labeled DNA probes (100 pmol) and 15 g of polydI-dC in DNAP buffer (20 mM HEPES-KOH, pH 7.9, 80 mM KCl, 1 mM MgCl 2 , 0.2 mM EDTA, 0.5 mM dithiothreitol, 10% glycerol, and 0.1% Triton X-100) on ice for 45 min. Subsequently, 500 g of Dynabeads M-280 streptavidin (Invitrogen) was added and incubated further at 4°C for 1 h. The beads were washed three times with DNAP buffer, and the bound proteins were eluted in SDS sample buffer, separated by 10% SDS-PAGE, and characterized by Western blot analysis with the respective specific antibodies.
In the ChIP re-IP assay, the precleared chromatin supernatants were immunoprecipitated with the first antibody, anti-ER␣ or anti-RIP140, at 4°C for overnight. The protein-antibody complexes were incubated with protein G PLUSagarose at 4°C for 2-4 h, eluted by incubation with 10 mM dithiothreitol at 37°C for 30 min, and diluted 1:50 in dilution buffer. After centrifugation, the supernatants were divided in aliquots and reimmunoprecipitated with their respective second antibodies individually. The second immunocomplexes were extracted from the beads followed by PCR amplifications of a 149-bp region of the PROS1 promoter containing target GC-rich motifs from bound DNA as described above.
StatisticalAnalysis-Dataarepresented as the mean Ϯ S.D. and are representative of at least three independent experiments. Significant differences between experimental groups in the quantitative RT-PCR, enzyme-linked immunosorbent assay, and luciferase assay were analyzed using Student's t test. Differences were considered to be significant when p was less than 0.05.

Down-regulation of PROS1 Expression by E 2 in HepG2-ER␣ Cells and Human Normal
Heptocytes-Because ER␣ was undetectable in the original HepG2 cells by Western blotting (Fig. 1A), we established a HepG2-derived cell line stably expressing human ER␣ (HepG2-ER␣) as described under "Experimental Procedures." We confirmed that the HepG2-ER␣ cells expressed large amounts of ER␣ protein equivalent to breast cancer-derived MCF7 cells as determined by Western blot analysis (Fig. 1A). The HepG2-ER␣ cells treated with E 2 showed significantly decreased levels of PS mRNA, but the original HepG2 cells and HepG2-Mock cells did not (Fig. 1B). Also, E 2 treatment down-regulated PS antigen significantly in the HepG2-ER␣ cells (Fig. 1C). In addition, we also demonstrated that E 2 treatment down-regulated PS mRNA by 60% in hNHep, which expressed a high level of ER␣ protein (Fig. 1, D and E).
Luciferase Reporter Assay-We next examined PROS1 promoter activity by conducting a luciferase reporter assay and observed that E 2 decreased the luciferase activity of pPROS1/ Ϫ4229 in HepG2-ER␣ cells ( Fig. 2A). In a series of 5Ј-truncated constructs obtained by restriction enzyme digestion and selfligation of pPROS1/Ϫ4229, we also observed E 2 -dependent repression of the luciferase activity, although the basal activity was gradually reduced (Fig. 2A). Meanwhile, a computer search for putative nuclear factor binding sites between Ϫ338 and Ϫ236 revealed AP-1 and GC-rich sites at Ϫ281 and Ϫ244 that might play a role in the basal expression of PROS1 in HepG2-ER␣ cells, respectively.
In a luciferase assay of further truncated forms, pPROS1/Ϫ175 showed E 2 -dependent repression, but pPROS1/Ϫ137 did not, indicating that the Ϫ175 to Ϫ137 region of the PROS1 promoter, containing two adjacent GC-rich motifs at Ϫ172 to Ϫ163 and Ϫ162 to Ϫ153, was critical for E 2 -induced downregulation (Fig. 2B). To clarify the importance of those GC-rich motifs, we transfected a series of constructs containing mutations of a single GC-rich site (pPROS1/ Ϫ175Mut1 or pPROS1/Ϫ175Mut2) or of both sites (pPROS1/ Ϫ175Mut3). We observed that the luciferase reporter activity of the HepG2-ER␣ cells transfected with pPROS1/Ϫ175Mut2 was reduced by E 2 treatment, but that of the cells transfected with pPROS1/ Ϫ175Mut1 or pPROS1/Ϫ175Mut3 was not (Fig. 2C).
To further investigate the importance of the two GC-rich motifs in a full-length promoter, we prepared a luciferase reporter vector with two mutated GC-rich motifs (pPROS1/Ϫ4229Mut) derived from pPROS1/Ϫ4229. In comparison with pPROS1/Ϫ4229, pPROS1/ Ϫ4229Mut showed a reduced luciferase activity and loss of its E 2 -dependent repression in HepG2-ER␣ cells (Fig. 2D). In the original HepG2 cells, which lack ER␣ expression, pPROS1/Ϫ4229 also did not show apparent E 2 -dependent repression of the luciferase activity.
Requirement of ER␣ for E 2 -dependent Down-regulation of PROS1 in HepG2 Cells-We also examined the ER␣ requirement for E 2 -induced inhibition of PROS1 promoter activity in HepG2 cells transfected with pPROS1/Ϫ175. The HepG2-ER␣ cells stably expressing human ER␣ showed E 2 -induced repression of luciferase activity, whereas the HepG2 mock-transfected cells with no ER␣ expression did not (Fig. 3A). Consistent with these observations, ICI 182,780 (a

. Transient expression of PROS1 promoter-reporter gene constructs in HepG2-ER␣ cells with or without E 2 treatment. Luciferase activity of the HepG2-ER␣ cells transiently transfected with pPROS1/
Ϫ4229 or its 5Ј-deleted constructs (A and B) and pPROS1/Ϫ175 or its mutant (C), with or without E 2 treatment, was measured. WT (pPROS1/Ϫ4229) or the full-length promoter construct with mutation (pPROS1/Ϫ4229Mut) was transiently transfected into HepG2-ER␣ cells or original HepG2 cells treated with or without E 2 , and luciferase activities were measured (D). The luciferase activity levels of the respective constructs were expressed as relative values to that of the pGL3 Basic-derived empty vector without E 2 treatment. A, results for the constructs pPROS1/Ϫ4229 to pPROS1/Ϫ236. B, results for the constructs pPROS1/Ϫ236 to pPROS1/Ϫ137 with putative transcription factor binding sites. pure antagonist of ER) reversed the effects of E 2 on luciferase activity in HepG2-ER␣ cells transfected with pPROS1/Ϫ175, whereas ICI 182,780 alone had no effect (Fig. 3B).

Sp1 and Sp3 Bind to GC-rich Motifs of the PROS1 Promoter in
Vitro-To identify the transcription factors binding to the GC-rich motifs at Ϫ173 and Ϫ162 of the PROS1 promoter in HepG2-ER␣ cells treated with E 2 , EMSAs and DNA pulldown assays were carried out. The oligonucleotides used in the EMSAs and DNA pulldown assays are shown in Fig. 4A. In the EMSAs, three shifted bands (ac) were detected using wild-type probes (GC-WT), two (a and b) of which were not detected using mutated oligonucleotides (GC-Mut) that destroyed both GC-rich motifs (Fig. 4B). Furthermore, co-incubation with a 100-fold excess of nonlabeled GC-WT oligonucleotides reduced the intensity of each of these two bands, but the third shifted band (c) was still present. These results indicate that the first two bands (a and b) were specific shifted bands, and the third (c) was nonspecific. In the supershift experiment, co-incubation with antibody against either Sp1 (a) or Sp3 (b) gave reduced shifted bands and additional supershifted bands (SS). Antibodies against ER␣ and nonspecific IgG, however, did not affect the intensity of the shifted bands. DNA pulldown assays gave results consistent with the EMSAs, indicating that Sp1 and Sp3 were apparently co-purified with the wild-type oligonucleotides but not with the mutated oligonucleotides (Fig. 4C). The Sp proteins were detected in the nuclear extracts regardless of E 2 treatment. In contrast, we detected more ER␣ signals in the nuclear extracts from E 2 -treated cells than in those from untreated cells, probably because of the increase in ER␣ proteins in the nucleus due to E 2 stimulation.
Knockdown Experiments with siRNA for Sp Proteins-To verify the function of Sp1 and Sp3 in the   retarded (a, b, and c) and supershifted (SS) complexes. FP, free probe. C, DNA pulldown assays carried out by incubating biotin-labeled oligonucleotides containing WT or mutated GC-rich motifs with nuclear extract from HepG2-ER␣ cells with or without E 2 treatment. Specifically bound proteins were eluted and subjected to Western blotting using specific antibodies against Sp1, Sp3, and ER␣, respectively. Similar results were obtained in multiple independent experiments. Veh, vehicle.
repression of PROS1 by E 2 , we carried out RNA interference experiments. A Western blot analysis of whole lysate from HepG2-ER␣ cells transfected with nonspecific siRNA (iNS) showed that Sp1 and Sp3 were almost equally expressed compared with levels in untransfected control cells (Fig. 5A). However, in the cells transfected with siRNAs for Sp1 (iSp1) and Sp3 (iSp3), a decreased expression of Sp1 and Sp3 proteins was observed, respectively. We next investigated the effects of Sp1 or Sp3 knockdown by conducting luciferase reporter experiments. After the co-transfection of both siRNA and the luciferase reporter construct (pPROS1/Ϫ175), HepG2-ER␣ cells were treated with E 2 , and luciferase activity was determined (Fig. 5B). Basal luciferase activity was decreased in the cells transfected with iSp1 or iSp3 compared with the cells transfected with iNS. E 2 -dependent repression of luciferase activity occurred in the cells transfected with iNS or iSp3 but not in the cells transfected with iSp1. We also observed consistent results for PS mRNA levels in the cells transfected with siRNAs (Fig. 5C). These results indicate that Sp1 and Sp3 are important for basal PROS1 transcription and that Sp1 also has a crucial role in E 2 -dependent repression.
ChIP and ChIP Re-IP Assays-The interactions of ER␣, Sp proteins, corepressors, and HDACs with the proximal region of the PROS1 promoter were further investigated in ChIP and ChIP re-IP assays. In the ChIP assay, the chromatin supernatants from HepG2-ER␣ cells treated with E 2 or vehicle were immunoprecipitated with specific antibodies against nuclear proteins (ER␣, Sp1, Sp3, NCoR, SMRT, HDAC1, HDAC3, HDAC4, HDAC5, and AcH4) or nonspecific IgG, and the eluted DNA fragments were used as subsequent PCR templates for amplification of the region containing target GC-rich motifs in the PROS1 promoter (Fig. 6A). We also used 5Ј-upstream regions containing non-target GC-rich motifs or no GC-rich motif for control PCR amplifications. As shown in Fig. 6B, ER␣ was recruited to the PROS1 promoter region containing target GC-rich motifs in an E 2 -dependent manner, although little was recruited to the control regions. Sp1 was also detected without E 2 treatment, but its expression was clearly enhanced by E 2 , whereas Sp3 was recruited in an E 2 -independent manner. We further investigated the possible commitment of other corepressors and found that NCoR and SMRT were also recruited to the region containing target GC-rich motifs in an E 2 -dependent manner (Fig. 6B, left).
Because liganded ER␣ is not known to recruit NCoR-SMRT corepressors directly, we tried ChIP assays for RIP140, which is known to interact with liganded ER and modulate ER-mediated transcription (35), and for LCoR, which is an NR-box-containing factor the same as RIP140 (36). After E 2 treatment, only RIP140 was strongly recruited to the region including target GC-rich motifs but not to the other regions (Fig. 6B, middle).
In addition, we investigated the recruitment of HDACs, which are known to interact with RIP140 (37), and observed class I HDACs; HDAC3 was expressed robustly, and HDAC1 was expressed slightly after E 2 treatment (Fig. 6B, right). Class II HDAC4 and HDAC5 were not affected by E 2 . Meanwhile, acetylated histone H4 was deacetylated in the region around the PROS1 promoter containing target GC-rich motifs following E 2 treatment (Fig. 6B, right).
Furthermore, to confirm the E 2 -dependent complex formation of those repressive proteins on the PROS1 promoter, we tried ChIP re-IP assays. We used anti-ER␣ or anti-RIP140 for primary immunoprecipitation and antibodies against the respective nuclear factors for secondary immunoprecipitation. As expected, we observed similar results in both of the ChIP re-IP assays using anti-ER␣ and anti-RIP140 as the primary antibodies, respectively, which showed that ER␣, RIP140, Sp1, Sp3, NCoR-SMRT, and HDAC3 were present on the same region of the PROS1 promoter containing target GC-rich motifs (Fig. 6C). Additionally, we observed that the blocking of deacetylation by an HDAC inhibitor, TSA, cancelled the E 2 -dependent PROS1 gene repression (Fig. 6D).

DISCUSSION
The action of estrogen in target cells is regulated via estrogen receptors that modulate gene expression either positively or negatively. Recent studies have clarified that E 2 -ER functions as a gene modulator, although its negative effects on gene expression are less well understood than its positive effects. We investigated here the molecular mechanisms by which E 2 -ER negatively regulates the gene expression of the anticoagulant PS.
First, we established a cell line (HepG2-ER␣) stably expressing human ER␣, because ER␣ was undetectable in HepG2 cells, and found that E 2 treatment of HepG2-ER␣ cells significantly down-regulated PS expression. The luciferase assays suggested that the GC-rich motif at Ϫ172 of the PROS1 promoter plays an important role in E 2 -dependent gene repression. In subsequent FIGURE 5. Knockdown of Sp1 or Sp3 by RNA interference and its effects on E 2 -dependent PROS1 repression. A, knockdown of Sp was determined by Western blotting. HepG2-ER␣ cells were transfected with iNS, iSp1, or iSp3, and whole cell lysate was analyzed by Western blotting as described under "Experimental Procedures." The experiments were repeated at least three times, and similar results were obtained. ␤-Actin was used as a loading control (Cont.). B, RNA interference-luciferase reporter analysis of HepG2-ER␣ cells. siRNAs (50 nM) were transfected, and the next day pPROS1/Ϫ175 was transfected with or without E 2 treatment. Luciferase activity was expressed relative to that of the cells without E 2 treatment. C, quantitative RT-PCR analysis after siRNA transfection in HepG2-ER␣ cells. Cells were transfected with respective siRNAs for 4 h and treated with E 2 for 48 h. Values are the mean Ϯ S.D. for at least three independent experiments. *, p Ͻ 0.05 versus vehicle control.
EMSAs and DNA pulldown analyses, we observed that Sp1 and Sp3 bound to the GC-rich motif of PROS1, which was consistent with the findings of de Wolf et al. (38).
We did not observe the binding of ER␣ or interaction of ER␣-Sp1 in the EMSAs; however, DNA pulldown assays showed that ER␣ bound to the PROS1 promoter (Ϫ176/Ϫ147) in an E 2 -dependent manner. This weak ER␣ binding seemed to depend on transfer from the cytoplasm to the nucleus by E 2 treatment. We also tried co-immunoprecipitation analyses but could not detect either the ER␣-Sp1 or the ER␣-Sp3 complex (data not shown). These observations were consistent with the report that a ternary ER␣-Sp-DNA complex was not detected in gel mobility shift assays and that ER␣ enhances the formation of the Sp1-DNA complex and increases its stability (39). We hypothesized that ER␣ would bind to Sp-DNA complex in vivo feebly and that it could be too fragile to be detected as a supershifted band in EMSAs or in an ER␣-Sp complex in co-immunoprecipitation analyses.
Through the RNA interference analysis, we found dual roles for Sp1 at the PROS1 promoter in HepG2-ER␣ cells. Thus, knockdown of Sp1 or Sp3 resulted in substantially decreased transcriptional activity, indicating that both Sp1 and Sp3 take part in basal PROS1 transcription as reported previously (38,40). Intriguingly, we observed a loss of E 2 -dependent PROS1 repression in the cells transfected with iSp1, suggesting that Sp1 has a crucial role in the down-regulation of PROS1 transcription by E 2 . This highlights the potential dual function of Sp1 for basal activation and E 2 -dependent repression of PROS1 transcription.
Our study also showed that ER␣ might interact with Sp1 and repress PROS1 expression via GC-rich motifs in its promoter region. GCrich regions are known to be involved in ER␣-mediated repression at the p21/WAF1 and cyclin G 2 gene promoters, where interplay with Sp proteins seems to occur (41,42). Moreover, direct ER-Sp1 binding has been well documented in estrogen-stimulated genes (43). Actually, through chromatin immunoprecipitation assays, we showed that ER␣ and Sp1 bound to the responsive regions of the PROS1 promoter simultaneously. Safe and Kim (39) stated previously that ER␣ enhances the formation of an Sp-DNA complex and increases its stability. Taken together, these findings suggested that Sp1 might retain its stability and bind more strongly to the regions of the PROS1 promoter responsible to ER␣.
Two pathways of E 2 -ER signaling, a "classical" and a "nonclassical" pathway, have been reported (25,39). In the classical pathway, ligand-bound homodimeric ERs bind directly to a palindromic ERE or half-ERE in the promoter region of a target gene and modulate gene transcription by recruiting transcriptional coactivators or chromatin remodeling complexes. In the FIGURE 6. ChIP and ChIP re-IP analyses of the E 2 -responsive region or control regions of the PROS1 promoter and the effect of TSA on E 2 -dependent PROS1 repression. A, ChIP analysis of the E 2 -responsive region containing GC-rich motifs in the PROS1 promoter and of the control regions containing no GC-rich motif a or non-target GC-rich motif. Specific primers for detection of respective regions (Target GC-rich, Ϫ236 and Ϫ88; Untarget GC-rich, Ϫ3748 and Ϫ3574; No GC-rich, Ϫ4033 and Ϫ3875) are represented by thick arrows. B, ChIP assays for HepG2-ER␣ cells treated with E 2 or vehicle control performed using various antibodies. Similar results were obtained in multiple independent experiments. C, ChIP re-IP assays for HepG2-ER␣ cells treated with E 2 or vehicle control performed using anti-ER␣ antibody or anti-RIP140 antibody for primary immunoprecipitation (IP). Re-immunoprecipitation was performed with respective antibodies as described. D, HepG2-ER␣ cells were treated with E 2 , TSA, or the two in combination, and total RNA was harvested and analyzed by quantitative RT-PCR. Values are the mean Ϯ S.D. for at least three independent experiments. *, p Ͻ 0.05 versus vehicle control; **, p Ͻ 0.05 versus TSA(Ϫ) ϩ E 2 .
non-classical pathway, ligand-bound ERs do not bind to ERE directly; instead, they interact with other transcription factors such as Sp and AP-1 (39). ER-Sp or ER-AP-1 interaction mediates transcriptional gene regulation, recruiting cofactors or chromatin remodeling complexes to GC-rich motifs or to the AP-1 site in the target gene promoter. In this study, we have demonstrated that ER␣-Sp1-RIP140 interaction regulates PROS1 expression by recruiting the NCoR-SMRT corepressor complex and also HDAC3, which induces histone hypoacetylation and less permissive transcription of the PROS1 gene.
An in vivo chromatin immunoprecipitation analysis of the E 2 -responsive region in the PROS1 promoter further revealed the recruitment of NCoR and SMRT to the PROS1 modulator complex. NCoR and SMRT are now documented corepressors for nuclear receptors such as antagonist-bound estrogen receptors and progesterone receptors (44). It was reported that NCoR forms several different complexes with other transcription factors such as SMRT, Sin3a, and HDACs (44). In this study, we observed the recruitment of NCoR and SMRT corepressors to the PROS1 promoter after E 2 treatment.
Liganded ER␣, however, is not known to recruit NCoR-SMRT corepressors directly. Therefore, we tried ChIP assays for RIP140, which is known to interact with liganded ER and modulate ER-mediated transcription (35), and for LCoR, which is an NR-box-containing factor the same as RIP140 (36). We demonstrated that RIP140 was recruited to the target GC-rich region of the PROS1 promoter after E 2 stimulation, but LCoR was not. The recruitment of RIP140 was theorized to correlate with the interaction between ER␣ and Sp1 because the effect of E 2 was destroyed by the knockdown of Sp1 in RNA interference analysis. We also found that the recruitment of NCoR-SMRT corepressors depended on ER␣-Sp1 interaction and that the complex contained HDAC3. The NCoR-SMRT-HDAC3 inter-action has been reported by several groups (45)(46)(47)(48), consistent with our analysis of the PROS1 promoter. The ChIP re-IP assays demonstrated that ER␣-Sp1, RIP140, and NCoR-SMRT-HDAC3 were present in the same region of the PROS1 promoter. Taken together with these data, this suggests that NCoR and SMRT were indirectly recruited as components of the HDAC3 complex by RIP140 through ER␣-Sp1 interaction following E 2 stimulation.
Our proposed model for the down-regulation of PROS1 expression by E 2 is depicted in Fig. 7. The repression of PROS1 resembles the situation described for p21/WAF1 or cyclin G 2 (39,49). In the absence of E 2 , basal transcriptional regulation of PROS1 seems to be mediated via multiple GC-rich regions including one at Ϫ172 to Ϫ153, as shown in this study and previously (38). In basal PROS1 transcription, Sp1 and Sp3 are recruited to the promoter, and other transcription factors or coactivators may be involved. Upon E 2 treatment, ER␣ interacted and formed a complex with Sp1 at the GC-rich motifs of the PROS1 promoter, and consequently ER␣ reinforced the stability of ER␣-Sp1-DNA binding. Moreover, ER␣-Sp1 interaction recruited RIP140, and consequently, RIP140 recruited the NCoR-SMRT-HDAC3 complex to the PROS1 promoter. The NCoR-SMRT-HDAC3 complex might induce the hypoacetylation of histones, which could lead to greater stabilization of the nucleosome structure, limiting accessibility of the basal transcription factors and thus down-regulating PROS1 expression.
To date, many genes have been reported to be stimulated by estrogen. Almost half of these estrogen-responsive genes may be repressed by estrogen (50); the mechanisms of estrogen-dependent repression have been elucidated. In this study, we have demonstrated novel mechanisms of E 2 -induced PROS1 repression, which could contribute to the elucidation of the estrogen action.
Meanwhile, our study has revealed that PROS1 is a unique gene regulated both positively and negatively by Sp1 interaction in the same cell line. VEGFR2 is also regulated both positively and negatively by ER␣-Sp protein interaction, but the type of regulation depends on the cell line, MCF7 or ZR-75 (51,52). Interestingly, we reported previously a novel missense mutation at Ϫ168 from the ATG of the PROS1 promoter (g.Ϫ168cϾt), which results in weak promoter activity of PROS1 (53). The mutation is located in the distal GC-rich motif, which could be critical for E 2 -dependent repression, as shown in the current study. In addition, we obtained similar results in the EMSA using the oligo probe containing only distal GC-rich motif with or without mutation (data not shown). These data suggest that the distal GC-rich motif may be more important A, in HepG2-ER␣ cells without E 2 treatment, basal PROS1 expression was regulated by Sp1 and Sp3, probably together with other transcriptional coactivators. B, in HepG2-ER␣ cells with E 2 treatment, PROS1 expression was repressed by ER␣-Sp1 interaction recruiting RIP140 and the NCoR-SMRT-HDAC3 corepressor complex sequentially, which consequently induced histone deacetylation. Ac, acetyl residue. than the proximal GC-rich motif. A patient carrying the g.Ϫ168cϾt mutation showed decreased levels of plasma PS but supposedly did not show E 2 -dependent PROS1 repression (53).
In conclusion, we have demonstrated that PROS1 expression is down-regulated by 17␤-estradiol via ER␣. ER␣ interacts with Sp1 and recruits RIP140. RIP140 associates directly with the HDAC3 complex containing NCoR-SMRT corepressors and induces the deacetylation of histones in the PROS1 promoter. We have also revealed the dual roles of Sp1, which regulates PROS1 expression both positively and negatively. E 2dependent repression of PROS1 results in reduced plasma PS levels, leading to the risk of deep venous thrombosis during pregnancy and oral contraceptive use. Further study will be required to fully characterize the mechanisms of reduction in PS and other possible mechanisms of regulation, such as using other hormones during pregnancy.