Inhibition of Vascular Endothelial Growth Factor Expression in HEC1A Endometrial Cancer Cells through Interactions of Estrogen Receptor α and Sp3 Proteins*

Treatment of HEC1A endometrial cancer cells with 10 nm 17β-estradiol (E2) resulted in decreased vascular endothelial growth factor (VEGF) mRNA expression, and a similar response was observed using a construct, pVEGF1, containing a VEGF gene promoter insert from −2018 to +50. In HEC1A cells transiently transfected with pVEGF1 and a series of deletion plasmids, it was shown that E2-dependent down-regulation was dependent on wild-type estrogen receptor α (ERα) and reversed by the anti-estrogen ICI 182,780, and this response was not affected by progestins. Deletion analysis of the VEGF gene promoter identified an overlapping G/GC-rich site between −66 to −47 that was required for decreased transactivation by E2. Protein-DNA binding studies using electrophoretic mobility shift and DNA footprinting assays showed that both Sp1 and Sp3 proteins bound this region of the VEGF promoter. Coimmunoprecipitation and pull-down assays demonstrated that Sp3 and ERα proteins physically interact, and the interacting domains of both proteins are different from those previously observed for interactions between Sp1 and ERα proteins. Using a dominant negative form of Sp3 and transcriptional activation assays in Schneider SL-2 insect cells, it was confirmed that ERα-Sp3 interactions define a pathway for E2-mediated inhibition of gene expression, and this represents a new mechanism for decreased gene expression by E2.

Angiogenesis is an important physiological process associated with neovascularization and growth and metastasis of many different tumors (1)(2)(3)(4)(5)(6)(7). The complex process of angiogenesis involves tissue-specific interplay between multiple growth factors, cytokines, kinases, and other growth regulatory proteins including vascular endothelial growth factor (VEGF) 1 or vascular permeability factors. VEGF is thought to play a key role in blood vessel formation; disruption of one VEGF allele in transgenic mice leads to embryonic death associated with impaired blood vessel formation, suggesting an important role for VEGF in vascularization (8,9). Angiogenesis is an important process for the growth of solid tumors, and anti-angiogenic drugs are rapidly being developed for treatment of this disease (10 -13). Not surprisingly, VEGF expression is a prognostic factor for many tumor types, and this protein plays a key role in tumor angiogenesis (14 -19).
VEGF and related angiogenic factors are expressed in mammary tumors, and increased levels of VEGF are a negative prognostic factor for survival of women with breast cancer (20 -27). Hormonal regulation of VEGF expression has been reported in the rodent uterus and mammary and in human breast cancer cells (28 -34). Hyder et al. (28,29) show that 17␤-estradiol (E2), tamoxifen, and structurally related triphenylethylene-derived anti-estrogens induced VEGF in the rat uterus, and E2 also induced VEGF mRNA and protein levels in carcinogen-induced rat mammary tumors (30). Regulation of VEGF in breast/endometrial cancer cell lines by steroid hormones has given variable results. One study showed that progestins, but not estrogens, androgens, or glucocorticoids, induced secretion of VEGF protein in T47D human breast cancer cells, whereas the progesterone-induced response was not observed in other breast (MCF-7, ZR-75, and MDA-MB-231) and endometrial (Ishikawa) cancer cell lines (32). Another study reported that E2 and tamoxifen induced VEGF mRNA levels, and VEGF protein was also increased by the hormone in breast cancer cells (33). HEC1A endometrial cancer cells express ER␣ protein, and both E2 and tamoxifen induce cell proliferation and reporter gene activity in cells transfected with E2-responsive constructs containing promoter inserts from the complement C3, creatine kinase B, and cathepsin D genes (35). Therefore, we have used HEC1A cells for investigating the molecular mechanisms of estrogen-dependent regulation of VEGF gene expression. The results showed that E2 decreased VEGF mRNA levels in HEC1A cells, and deletion analysis of the VEGF gene promoter identified a minimal sequence (Ϫ66 to Ϫ47) that was required for hormone-dependent decreased expression. Subsequent mutational analysis and results of gel mobility shift and DNA-footprinting assays indicated that hormonal effects were associated with ER␣/Sp3 interactions at G/GC-rich sites. Transfection studies in Drosophila SL2 Schneider cells in culture also showed that ER␣/Sp3 mediated decreased transactivation of a VEGF gene promoter-derived construct, and both ER␣ and Sp3 proteins physically interact, as determined in coimmunoprecipitation and pull-down assays. Thus, ligand-activated ER␣/Sp3 interaction with G/GC-rich el-ements represents a novel pathway for down-regulation of VEGF, and these interactions may be important for hormonedependent regulation of other genes given the appropriate cellular context.
hER␣ expression plasmid was kindly provided by Dr. Ming-Jer Tsai (Baylor College of Medicine, Houston, TX). hER␣ deletion constructs HE11C, HE15C, and HE19C were originally obtained from Dr. Pierre Chambon (Institut de Genetique et de Biologie Moleculaire et Cellulaire, Illkirch, France) and inserted into vectors pCDNA3 and pCDNA3.1/His C (InVitrogen, Carlsbad, CA) in this laboratory for in vitro translation and expression in mammalian cells. Eukaryotic expression plasmids for human Sp1 and Sp3 proteins were made by excising the Sp1 or Sp3 cDNAs from pPacSp1 (generously supplied by Dr. Robert Tjian, University of California, Berkeley, CA) or pPacUSp3 (kindly provided by Dr. Guntram Suske, Institute fur Molekularbiologie und Turmorforschung, Marburg, Germany) using XhoI and NotI restriction enzymes, respectively. Sp1 and Sp3 cDNAs were then ligated into appropriately digested, calf intestinal phosphatase-treated pCDNA3.1 expression plasmid with oligonucleotide modifications to produce Kozak sequences with in-frame start codons for each protein. pPac-hER␣ was produced by removal of hER␣ cDNA from pCDNA3 or pCDNA3.1 by EcoRI digest and ligation into a modified Drosophila expression plasmid pPacUbx multiple cloning site. Before insertion into pPacUbx, oligonucleotide linkers were added to hER␣ cDNA to ensure proper frame and expression as described (38). Human expression plasmid for progesterone receptor-B form was also a generous gift of Dr. Pierre Chambon. Rat c-Fos expression plasmid was kindly provided by Dr. Tom Curran (Roche Institute of Molecular Biology, Nutley, NJ) and was digested with HindIII and BamHI to release c-Fos cDNA; the insert was then cloned into pCDNA3.1 expression vector. Egr-1 expression plasmid was made by digesting pJDM1196 (a kind gift of Dr. Jeffrey Milbrandt, Washington University School of Medicine, St. Louis, MO) with EcoRI and XbaI to release a 1.8-kilobase pair rat Egr-1 cDNA, which was then ligated into pCDNA3 empty vector digested with the same restriction enzymes. pCDNA3-Egr-1 was used as the template to produce Egr-1 protein in a rabbit reticulocyte-coupled transcription/ translation reaction (Promega). Plasmid pCMV-DNSp3 contains the DNA binding domain and dominant negative form of Sp3 in pCDNA3.1/His C and was generously provided by Drs. Yoshihiro Sowa and Toshiyuki Sakai (Kyoto Prefectural University of Medicine, Kyoto, Japan).
Transient Transfection Assays-Appropriate VEGF-luciferase reporter plasmid, hER␣ expression plasmid, and pCDNA3.1-His-LacZ (InVitrogen) expression plasmid (for normalization of transfection efficiency) were transiently cotransfected into HEC1-A cells using the calcium phosphate-DNA coprecipitation method or LipofectAMINE (Life Technologies, Inc.). pCDNA3 or pCDNA3.1 empty vectors were also transfected to maintain DNA mass balance among different transfection groups. An estrogen-responsive pC3-Luc construct, containing the mouse complement-3 (C3) gene promoter insert, was kindly provided by Dr. Donald P. McDonnell (Duke University Medical School, Durham, NC) and used as a positive control in most experiments to confirm hormone responsiveness of the transfected cells. After 12-24 h, the transfection mixture was aspirated from cells, and the cells, which were then treated for 24 -72 h with fresh serum-free medium containing 10 nM E2 dissolved in Me 2 SO, served as a solvent control. One M concentrations of ICI 182,780 or tamoxifen were used in experiments with anti-estrogens. Drosophila Schneider SL-2 cells were cultured as described above, seeded in 6-well culture dishes, and transfected with pVEGF6, pPacSp1, or pPacUSp3 and hER␣ expression plasmid; SL-2 cells were then treated with Me 2 SO or 10 nM E2. After treatment for 24 -72 h, all cells were harvested by manual scraping in 1ϫ lysis buffer (Promega). An aliquot of soluble protein was obtained by freezing cells in liquid nitrogen (30 s), vortexing (30 s), and centrifuging at 12,000 ϫ g for (1 min). Lysates were assayed for luciferase activity using luciferase assay reagent (Promega); ␤-galactosidase activity was measured using Tropix Galacto-Light Plus assay system (Tropix, Bedford, MA) according to the manufacturer's conditions in a Lumicount micro-well plate reader (Packard Instrument Co.). Relative luciferase activity was normalized to relative ␤-galactosidase units for each transfection experiment.
Cells were cultured in serum-free medium for 3 days before treatment with 10 nM E2 and also in serum-free media supplemented with bovine serum albumin. Total RNA was extracted using RNAzol B (Tel-Test, Friendswood, TX), following manufacturer's protocol. RNA was quantitated by measuring the 260/280-nm absorption ratio, and concentrations of each sample were adjusted to 100 -200 ng/l RNA for use in RT-PCR. RNA were reverse-transcribed at 42°C for 25 min using oligo-d(T) primer, followed by PCR amplification of RT product using 2 mM MgCl 2 , 1 M each gene-specific primer, 1 mM dNTPs, and 2.5 units AmpliTaq DNA polymerase (Perkin-Elmer). Primer sets for both VEGF and ␤-actin were added to each mixture, and the gene products were coamplified (22 cycles). Ater amplification in a PTC-200 thermal cycler (MJ Research, Watertown, MA), 25 l of each sample was loaded on a 5% polyacrylamide gel. Electrophoresis was performed at 110 V in 1ϫ TBE (0.09 M Tris-base, 0.09 M boric acid, 2 mM EDTA, pH 8.3) for 3-4 h, the gel was dried, and individual VEGF bands were quantitated and normalized relative to ␤-actin using an Instant Imager (Packard) and Storm PhosphorImager (Molecular Dynamics, Sunnyvale, CA). Band intensity values were obtained by autoradiography on X-Omat AR film (Eastman Kodak Co.) and scanning densitometry with a Sharp JX-330 scanner (Sharp Electronics, Mahwah, NJ), followed by analysis with a software package, Zero-D, (Scanalytics, Sunnyvale, CA).
Preparation of HEC1A Nuclear Extracts-Cells were cultured in medium without phenol red, supplemented with 2.5% fetal bovine serum, and pretreated with dextran-coated charcoal to remove growth factors and endogenous estrogens. Treatments were added to cells for 1-24 h prior to harvesting by trypsinization or manual scraping. Resuspension of cells in HED (25 mM HEPES, 1.5 mM EDTA, 1 mM dithiothreitol, pH 7.6) to allow swelling was followed by cellular homogenization in HEGD (HED with 10% glycerol), using a drill/pestle apparatus or Type-B Dounce homogenizer (Kontes Glass Co., Vineland, NJ) on ice. Cell viability was checked by trypan-blue staining, and homogenization was stopped when approximately 80 -90% of cells were broken. The cellular homogenate was centrifuged at 800 ϫ g/10 min/4°C to give a nuclear pellet that was incubated in high salt HEGDKϩ buffer (HEGD containing 0.5 M KCl) on ice for 1 h. Nuclei in high salt buffer were centrifuged at 12,000 ϫ g for 30 min at 4°C, and protein concentration of the nuclear extract (supernatant) proteins was determined (39); the nuclear extracts were then stored in small aliquots at Ϫ80°C for use in gel shift assays.
EMSA-VEGF promoter-derived oligonucleotides were synthesized and annealed, and 5 pmol were 5Ј-end-labeled using T4 kinase (Roche Molecular Biochemicals) and [␥-32 P]ATP (NEN Life Science Products). A 20 -30-l EMSA reaction mixture contained 75-150 mM buffer (a mixture of HEGD and HEGDKϩ to achieve the proper salt concentration), 1-10 g of nuclear protein, 500 -1000 ng of poly(dI-dC) or poly(dA-dT) (Roche Molecular Biochemicals) with or without unlabeled competitor oligonucleotide, and 10 fmol of labeled probe; the mixture was incubated for 15 min at room temperature or on ice. Commercially available antibodies to Sp1, Sp3, and hER␣ proteins were sometimes added to the incubation mixtures immediately after the addition of the oligonucleotide probe and incubated for 15-60 min further. Protein-DNA complexes were resolved by 5% PAGE at 110 -200 V at room temperature or 4°C for 1.5-4 h. Specific antibody-protein interactions were observed as supershifted or immuno-depleted complexes.
The assay for determining Egr-1 DNA binding used a similar buffering system supplemented with 40 M ZnCl 2 . Specific nuclear extract-VEGF promoter complexes formed were separated from free probe by electrophoresis at room temperature or at 4°C on a 3.5-5% polyacrylamide gel in 1ϫ TBE, 100 -180 V, for 2-4 h. For EMSA of pure proteins or in vitro transcribed/translated proteins, 3-5 l of rabbit reticulocyte lysate containing the protein of interest was incubated in similar conditions as used for nuclear extract. All reactions were balanced for equal amounts of rabbit reticulocyte lysate that "mock"-transcribed/translated empty pCDNA3 or pCDNA3.1 vector. To verify specific components of DNA-protein complexes formed, a commercially available antibody against Egr-1 was added to some reaction mixtures immediately after labeled probe, incubated on ice for 15-60 min, and analyzed as described above.
SssI Methylase Footprinting-The protocol for SssI methylase footprinting was used as described (40,41), with some modifications. Briefly, a plasmid containing the human VEGF 5Ј-untranslated region was restricted at a single site to linearize the vector. The VEGF gene promoter, at a concentration of 10 ng/l, was incubated with varying concentrations of nuclear extracts from Me 2 SO-or E2-treated cells or in vitro translated Sp1 and Sp3 proteins in a reaction buffer containing 5% glycerol, 17.65 mM MgCl 2 , 0.18 mM S-adenosylmethionine, and 5 mM dithiothreitol. The reaction was incubated on ice for 5 min, then at room temperature for 20 min to allow protein-DNA interactions. Diluted SssI methylase (New England Biolabs Inc., Beverly, MA) was added to each reaction, and all components were incubated at 30°C for 15 min to allow for the enzyme to methylate CpG islands. Samples were then heat-inactivated at 75°C for 15 min, and a deamination buffer was added to each tube. Fresh deamination buffer consisted of 0.9 M NaOH, 25 mM EDTA, and 200 g/ml sonicated fish sperm DNA. After the addition of saturated aqueous sodium metabisulfite, the samples were boiled for 5 min, and the deaminated VEGF promoter product was subjected to PCR using modified gene-specific primers. PCR products were then separated by agarose gel electrophoresis, extracted from the gel using a commercially available kit (Qiagen Inc., Santa Clarita, CA), and sequenced using either one of the VEGF gene-specific primers mentioned above and Sequitherm Thermostable DNA polymerase (Epicentre Technologies, Madison, WI). Dideoxy-GTP was used for chain termination, and the products were resolved on a denaturing 6% poly-acrylamide-urea gel. After gel drying and exposure to a phosphor storage screen, sequence data were visualized on a Storm scanner (Molecular Dynamics) for determination of specific footprint patterns.
hER␣/Sp3 Immunoprecipitation-For immunoprecipitation of hER␣ by polyclonal antibodies against Sp3, in vitro transcribed/translated proteins were obtained using the rabbit reticulocyte lysate system (Promega), and hER␣ was labeled with [ 35 S]methionine (Amersham Pharmacia Biotech). Equal volumes (10 l) of lysate containing [ 35 S]Met-hER␣ and cold Sp3 protein were incubated for 1 h at room temperature. To this mixture, Sp3 polyclonal antibodies were added, and the samples were subsequently incubated for 2 h on ice. Protein G-agarose beads (Santa Cruz) were added then added, and the samples were incubated at 4°C for 3-5 h. Immunoprecipitation of protein G-bound proteins was conducted according to the manufacturer's recommendations. Proteins were separated by SDS-PAGE (10% gel) and visualized by Storm Phos-phorImager (Molecular Dynamics).
GST-Sp3/hER␣ Pull-down Assay-Plasmid constructs containing the glutathione S-transferase (GST) cDNA linked to human Sp3 cDNA were also generously provided by Dr. Guntram Suske and included pGEX-Sp3 II (or Sp3-full), pGEX-Sp3-B, and pGEX-Sp3-BC. Empty vector pGEX-2TK was produced from pGEX-Sp3-BC by removing the Sp3 cDNA fragment with an XbaI and XhoI digest and religating the vector. Subsequently, pGEX-Sp3-ZD was generated by PCR amplification of the DNA binding domain of Sp3 using primers previously described (42). The Sp3-ZD product was digested with BamHI and EcoRI and ligated into the correspondingly cut pGEX-2TK vector. BL 21 Escherichia coli competent cells (ATCC) were transformed with the appropriate pGEX plasmid and induced by 1 mM isopropyl-1-thio-␤-D-galactopyranoside (Promega) treatment, and bacterial extracts of GST-Sp3 fusion proteins were isolated according to the manufacturer's protocol (Pierce). Cell lysates were resolved by 10% SDS-PAGE and visualized by Coomassie Blue staining to verify protein expression ratios for each Sp3 deletion. GST-Sp3 fusion proteins were purified from bacterial cell lysates by incubation with Sepharose 4B beads conjugated to glutathione (Amersham Pharmacia Biotech). Subsequently, beads and attached Sp3 protein deletions were incubated at 4°C for 2-12 h with in vitro translated [ 35 S]methionine-labeled hER␣ proteins (WT and variants described above) in hER binding buffer (250 mM NaCl, 0.1% Nonidet P-40, 50 mM HEPES, pH 7.5, 5.0 mM EDTA, 0.5 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 10 g/l aprotinin), as described (43). Beads were then washed with the same buffer 4ϫ, and specific protein interactions were eluted by boiling in 1ϫ Laemmli's solution. Eluted proteins were resolved on a 10% SDS-PAGE gel, and the gel was dried and visualized by exposure to a phospho-storage screen scanned on a Storm PhosphorImager or by autoradiography.
Western Immunoblot-HEC1A cells in serum-free medium were treated with Me 2 SO or 10 nM E2 at various times and harvested in lysis buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, 10% (v/v) glycerol, 1% Triton X-100, 1.5 mM MgCl 2 , 1 mM EGTA, 10 g/ml aprotinin, 50 mM phenylmethylsulfonyl fluoride, 50 mM sodium orthovanadate), placed on a rocker at 4°C to extract soluble protein, and centrifuged at 14,000 ϫ g for 5 min at 4°C. Protein was quantitated (37), and a 50-g aliquot of protein diluted with loading buffer was boiled and loaded on a 7.5% SDS-polyacrylamide gel. Samples were electrophoresed at 150 -180 V for 3-4 h, and the separated proteins were transferred (in a buffer containing 48 mM Tris-HCl, 29 mM glycine, and 0.025% SDS) to PVDF membrane (Bio-Rad). Specific proteins were detected by incubation with polyclonal primary antibodies Sp1-PEP2 and Sp3-D20 (both 1:1000 dilution) against Sp1 and Sp3 proteins, respectively, followed by blotting with horseradish peroxidase-conjugated anti-rabbit secondary antibody (1:5000 dilution). Blots were then exposed to chemiluminescent substrate (ECL) (NEN Life Science Products) and placed on Kodak X-Omat AR autoradiography film. Band intensities were evaluated by scanning laser densitometry (Sharp Electronics Corp.) and background subtraction using Zero-D Scanalytics software (Scanalytics Corp.).
Statistical Analysis-Experiments were repeated two or more times, and representative results are expressed as the mean Ϯ S.E. for at least three replicates for each treatment group. Data were subjected to analysis of variance, and statistical differences between treatment groups were demonstrated by Scheffe's test. Treatments were considered statistically significantly different from controls if p Ͻ 0.05.

RESULTS
E2-dependent Down-regulation of VEGF mRNA and Reporter Gene Activity in HEC1A Endometrial Cancer Cells-Semiquantitative RT-PCR analysis of VEGF mRNA levels showed that treatment of HEC1A endometrial cells with 10 nM E2 caused a decrease 4 h after treatment, and there was a significant (Ͼ50%) decrease in mRNA levels after 24 h (Fig.  1A). E2 also decreased luciferase activity in HEC1A cells transiently transfected with pVEGF1, pVEGF2, and pVEGF3 constructs containing the Ϫ2018 to ϩ50, Ϫ789 to ϩ50, and Ϫ267 to ϩ50 regions of the VEGF gene promoter linked to a bacterial luciferase reporter gene (Fig. 1B). Further deletion analysis with pVEGF4 (Ϫ131 to ϩ379), pVEGF5 (Ϫ131 to ϩ54), and pVEGF6 (Ϫ66 to ϩ54) showed that E2 decreased luciferase activity 40 -60% in HEC1A cells transfected with all three constructs; however, a significant decrease in basal activity was also observed with pVEGF6 (data not shown), indicating that the Ϫ131 to Ϫ66 region of the VEGF gene promoter was important for basal activity in HEC1A cells.
The specificity of the ER␣-mediated down-regulation response was further investigated using pVEGF4 and pVEGF5, which contained VEGF 5Ј-regulatory sequences up to Ϫ131 and pVEGF6 containing only the Ϫ66 to ϩ54 region of the promoter. pVEGF4, pVEGF5, and pVEGF6 gave similar responses in transient transfection studies showing that ER␣, but not progesterone receptor (PR), mediates the hormone-dependent decrease in luciferase activity ( Fig. 2A). Moreover, the decreased response in HEC1A cells transfected with pVEGF4 or pVEGF5 caused by E2 is reversed after cotreatment with 1 M concentration of the anti-estrogen ICI 182,780 (Fig. 2B). In parallel studies with pVEGF6, the results were similar to those obtained for the longer constructs (pVEGF4 and pVEGF5); wild-type ER␣, but not PR, mediated a hormone-dependent decrease in activity, and the anti-estrogen ICI 182,780 reversed the estrogenic response (Fig. 2C). The requirement for activation function 1 (AF1) or AF2 domains of ER␣ were also investigated in HEC1A cells cotransfected with pVEGF6 and variants expressing AF2 (HE19), AF1 (HE15), or both AF1 and AF2 but not the DNA binding domain (HE11). The results showed that only wild-type ER␣ mediated hormone-dependent downregulation of luciferase activity in HEC1A cells transfected with pVEGF6. The downstream Ϫ66 to ϩ54 region of the VEGF gene promoter contains two overlapping G/GC-rich regions (Ϫ66/Ϫ47) and a third downstream GC-rich site; results of deletion (pVEGF7) and mutational (pVEGF6m) analysis indicate the former site was required for the hormone-induced response (Fig. 2D). Moreover, the Ϫ66/Ϫ47 site alone (pVEGF8) was sufficient for significant E2-dependent decrease in luciferase activity, but the magnitude of this response was lower than observed for constructs containing the endogenous proximal promoter region. Mutation of the overlapping G/GCrich site (pVEGF8m) resulted in loss of hormone responsiveness, and similar results were obtained using constructs containing a wild-type consensus GC-rich Sp1 binding site (pSp1) and a mutant Sp1 binding site (pSp1m). In all of these experiments, E2 responsiveness of the cells was monitored (data not shown) by showing that 10 nM E2 induced a 5-20-fold increase of luciferase activity in HEC1A cells transfected with pC3-luc, a construct containing the mouse complement C3 gene promoter insert (35).
Identification of Proteins Interacting with the Ϫ66/Ϫ47 Region of the VEGF Gene Promoter-The G-rich Ϫ66 to Ϫ47 region of the VEGF gene promoter is sufficient for down-regulation by ER␣ and incubation of nuclear extracts from HEC1A cells with 32 P-(Ϫ66/Ϫ47), and analysis by gel electrophoretic mobility shift assays gave a high molecular weight broad band (Fig. 3A, lane 2) and a less intense lower band. Intensities of the higher molecular weight band were decreased after competition with wild-type (but not mutant Ϫ66/Ϫ47) and consensus GC-rich oligonucleotides (lanes 3-6). Moreover, after coincubation with antibodies to Sp3 or Sp1 proteins or nonspecific IgG (lanes 8 -10), Sp1 antibody supershifted the major band (Sp1-SS), whereas a smaller component of the major band and the lower molecular weight complex were immunodepleted by Sp3 antibody. Nonspecific IgG had no effect. In a separate experiment using a different Sp3 antibody, a supershifted complex was observed (data not shown). This same region of the VEGF promoter did not bind recombinant human ER␣ (data not shown), and competition with unlabeled estrogen-responsive element did not decrease intensities of the retarded bands formed with 32  RT-PCR analysis of VEGF mRNA levels in HEC1A cells. Cells were seed in 2.5% charcoal-stripped serum medium for 1 day and cultured under serum-free conditions for 3 days and treated with Me 2 SO (DMSO) (vehicle) or 10 nM E2 for the indicated times, and mRNA levels were determined as described under "Materials and Methods." E2 significantly (*p Ͻ 0.05) decreased steady state VEGF mRNA levels after 24 h. Results are expressed as mean Ϯ S.E. B, effects of E2 treatment on VEGF gene promoter-luciferase reporter activity in HEC1A cells. Cells were transiently transfected with 500 ng of VEGF gene promoterluciferase reporter construct (pVEGF1-6), 500 ng of hER␣ expression plasmid, and 500 ng of ␤-galactosidase expression vector and then treated for 24 -72 h with Me 2 SO or 10 nM E2. Luciferase activity was determined as described under "Materials and Methods" and normalized to ␤-galactosidase activity; results are presented as percent Me 2 SO for each transfection group. In most experiments, cells were transfected in parallel with a highly E2-inducible promoter-reporter construct pC3-Luc to verify E2 responsiveness. Results are expressed as mean Ϯ S.E., and significant (p Ͻ 0.05) inhibition (*) was observed with all constructs. 49). Therefore, we investigated the role of Egr-1 in mediating the down-regulation of VEGF since a putative Egr-1 binding site was identified at Ϫ60 to Ϫ51 in the VEGF gene promoter, although direct binding by Egr-1 protein was not demonstrated (36). Results summarized in Fig. 3B compare binding of nuclear extracts from E2-treated HEC1A cells to 32 P-(Ϫ66/Ϫ47) and [ 32 P]Egr-1 oligonucleotides. Protein binding to the Ϫ66 to Ϫ47 region gave bands (lanes 1-6) that were not competitively

FIG. 2. Deletional analysis and hormone or anti-estrogen responsiveness of pVEGF constructs in HEC1A cells.
A, comparative effects of E2 and progesterone. Cells were transiently transfected with 500 ng of reporter (pVEGF4 -6), 500 ng of hER␣ or human PR-B form, and 500 ng of ␤-galactosidase expression plasmid, treated with Me 2 SO (DMSO), 10 nM E2, or 10 nM progesterone (P) for 24 -72 h, and luciferase activity was determined as described under "Materials and Methods." E2 significantly (*p Ͻ 0.05) decreased pVEGF4 -6 luciferase activity, whereas progesterone had no effect on activity. B, effect of the anti-estrogen ICI 182,780. Cells were transiently transfected as described above and treated with Me 2 SO, 10 nM E2, 10 nM E2 ϩ 1 M ICI 182,780 (I), or 1 M ICI 182,780 alone for 24 -72 h. The pure anti-estrogen (ICI 182,780 (I)) significantly (*p Ͻ 0.05) reversed the effects of E2 on luciferase activity but did not affect control activity. C, comparative effect of wild-type and variant hER␣. Cells were transiently transfected with 500 ng of pVEGF6, 500 ng of WT, or variant (HE11C, HE15C, and HE19C) hER␣ as described above and treated with Me 2 SO (vehicle), 10 nM E2, 1 M ICI 182,780 (I), or E2 ϩ ICI 182,780 for 24 -72 h. Decreased luciferase activity was only observed with WT hER␣, and this activity was reversed by cotreatment with ICI 182,780 (*p Ͻ 0.05). D, mutational analysis of pVEGF6. HEC1A cells were transiently transfected with pVEGF6, pVEGF6m (mutation in both GC-rich sites), pVEGF7 (GC-rich sequence deleted), pVEGF8 (GC-element alone inserted upstream of the TATA box), or pVEGF8 m (GC-rich site mutated and placed upstream of the TATA box), as described above, and treated with Me 2 SO or 10 nM E2 for 24 -72 h. The GC-rich Ϫ66/Ϫ47 VEGF promoter sequence was necessary and sufficient for mediating E2-decreased (*p Ͻ 0.05) luciferase activity. As a control, consensus Sp1 and mutant Sp1 oligonucleotides were synthesized, inserted in TATA-pGL2, and assayed for response to E2 treatment. Down-regulation of luciferase activity through a consensus Sp1 site was significant (*p Ͻ 0.05) and similar to results obtained with pVEGF8. Oligonucleotide sequences are included, with lowercase letters representing 5Ј-KpnI and 3Ј-NheI restriction site overhangs; bold-faced letters denote mutated G 3 T sequences. decreased with consensus Egr-1 oligonucleotide or supershifted with Egr-1 antibodies. In contrast, the same nuclear extracts from HEC1A cells formed a more mobile specifically bound band [ 32 P]Egr-1 (lanes 7-12) that was competitively decreased by wild-type (but not mutant) Egr-1 oligonucleotide and supershifted by Egr-1 antibodies. These results indicate that, whereas Egr-1 is expressed in HEC1A cells, this protein does not bind 32 P-(Ϫ66/Ϫ47). We further investigated potential binding of Egr-1 to the VEGF gene promoter by comparing interactions of in vitro translated protein with a consensus [ 32 P]Egr-1 oligonucleotide and with 32 P-(Ϫ66/Ϫ47) (derived from the VEGF gene promoter) (Fig. 3C). In vitro translated Egr-1 bound consensus [ 32 P]Egr-1 to give a retarded band (lanes 9 -14), and intensity of this band was decreased after competition with unlabeled consensus wild-type Egr-1 but not mutant Egr-1 oligonucleotides (lanes 11 and 12). Moreover, Egr-1 antibodies supershifted the Egr-1-DNA complex (lane 13). In contrast, in vitro translated Egr-1 did not bind 32 P-(Ϫ66/ Ϫ47) to form a retarded band (lanes 2-7), suggesting that Egr-1 did not play a significant role in down-regulating VEGF through direct interactions with the Ϫ66/Ϫ47 region of the promoter.  6 and 7), decreased intensity of the Sp1-DNA retarded band. In addition, Sp1 protein bound a consensus wildtype oligonucleotide to give a retarded band complex (lane 8) with mobility similar to the Sp1-32 P-(Ϫ66/Ϫ47) band. Incubation of reticulocyte lysate alone with 32 P-(Ϫ66/Ϫ47) gave two nonspecific bands (Fig. 4B, lane 1), and incubation with in vitro translated Sp3 protein primarily gave a single higher molecular weight retarded band (lanes 2-5) that was decreased in intensity after incubation with 100-fold excess of unlabeled Ϫ66/Ϫ47 oligonucleotide (lane 6) but not mutant Ϫ66/Ϫ47m oligonucleotide (lane 7). Thus, the G/GC-rich Ϫ66/Ϫ47 region of the VEGF gene promoter binds Sp1 and Sp3 proteins but not Egr-1. In some studies, variability of Sp1 and Sp3 ratios can account for treatment-dependent effects on transactivation (50 -52), and therefore, the effects of 10 nM E2 on immunoreactive Sp1 and Sp3 proteins were determined in HEC1A cells (Fig. 4C). The results show that levels of Sp1 protein do not significantly change over a period of 24 h, whereas a 20 -30% increase of immunoreactive Sp3 protein was observed within FIG. 3. Binding of nuclear extracts and Egr-1 to 32 P-(؊66/؊47) (EMSA analysis). A, nuclear extracts from E2-treated HEC1A cells bind 32 P-(Ϫ66/Ϫ47) and consensus [ 32 P]Sp1 oligonucleotides. Binding of nuclear extracts to oligonucleotides and antibody supershifts were determined by EMSA, as described under "Materials and Methods." Nuclear protein extracts formed specific Sp1 and Sp3 complexes with 32 P-(Ϫ66/Ϫ47) (lanes 1-8) and [ 32 P]Sp1 (lanes 9 -15), and these bands were supershifted with antibodies to Sp1 (lanes 5 and 12) and Sp3 (lanes 7 and 14) proteins. Normal rabbit and goat IgG did not supershift any bands (lanes 6 and 8; lanes 13 and 15). B, binding of nuclear extracts from E2-treated HEC1A cells with 32 P-(Ϫ66/Ϫ47) or consensus [ 32 P]Egr. This assay was carried out as described in A; Egr-1 protein does not bind 32 P-(Ϫ66/Ϫ47) (lanes 1-6) but forms a specifically bound complex with [ 32 P]Egr (lanes 7-12) that is supershifted (S.S.) with Egr-1 antibody (lane 11). C, binding of in vitro translated rat Egr-1 protein and 32 P-(Ϫ66/Ϫ47) or [ 32 P]Egr. In vitro Egr-1 does not bind 32 P-(Ϫ66/Ϫ47) (lanes 1-7) but binds [ 32 P]Egr (lanes 8 -14) and forms an immuno-reactive complex with consensus [ 32 P]Egr (lane 13).
3 h after treatment with E2. This shows that there is an overall significant increase in Sp3 protein (at 3 and 24 h); however, the increased levels are modest.
Transactivation of pVEGF6 by ER␣/Sp1 and ER␣/Sp3 in Schneider SL2 Cells and Effect of Dominant Negative Sp3 in HEC1A Cells-Previous studies in this laboratory have demonstrated functional interactions of Sp1 and ER␣ at GC-rich sites in several gene promoters where ER␣/Sp1 is associated with hormone-dependent up-regulation of gene expression in human breast cancer cell lines (38,(53)(54)(55)(56)(57)(58)(59)(60)(61). Therefore, we investigated ER␣ (ϩE2) interactions with both Sp1 and Sp3 proteins in Drosophila Schneider SL-2 cells that do not constitutively express ER␣, Sp1, or Sp3 proteins. In SL-2 cells transiently transfected with pVEGF6 and expression plasmids for Sp1 and ER␣, the results show that ER␣ enhances Sp1-dependent activation of pVEGF6, and a Ͼ3-fold increase in luciferase activity was observed using 500 ng of ER␣ expression plasmid (Fig. 5A). Similar results were observed using a G/GC-rich construct from the bcl-2 gene promoter (38). Enhanced activity was lower using 1000 ng of ER␣, whereas ER␣ expression plasmid alone (2000 ng) did not enhance transactivation compared with controls (no plasmids added). Using a similar approach in SL-2 cells cotransfected with Sp3 expression plasmid (100 ng) and pVEGF6 (1 g), cotransfection with increasing amounts of ER␣ expression plasmid (200 -1000 ng) resulted in a Ͼ50% decrease in luciferase activity. These data are consistent with the observed hormone-dependent ER␣/Sp3-mediated down-regulation of VEGF gene expression in HEC1A cells, and in this same cell line, overexpression of dominant negative Sp3 expression plasmid partially reversed the hormone-dependent response (Fig. 5C).
Physical Interactions of ER␣ and Sp3 Proteins-Previous studies have shown the ER␣ does not supershift an Sp1-DNA complex in gel mobility shift assays but enhances retarded band intensity. This type of interaction has been observed for other proteins in gel mobility shift assays; for example, human T cell leukemia virus type-1 Tax, sterol regulatory elementbinding protein, and cyclin D1 enhanced bZIP (and GATA-4), Sp1, and ER binding to their respective cognate DNA enhancer sequences but did not form a supershifted band even though the binary mixture of proteins physically interacts (43,(62)(63)(64)(65). The results illustrated in Fig. 6A show that recombinant human ER␣ protein enhances the intensity of in vitro translated Sp1 protein-32 P-(Ϫ66/Ϫ47) interaction (lanes 2-4), and the enhanced band is supershifted by Sp1 antibody (lane 5). Similar result were also observed for the in vitro translated Sp3-32 P-(Ϫ66/Ϫ47) retarded band (lanes 6 -8), and at the highest concentration of ER␣, a 160% increase in retarded band intensity was observed (lane 8). The Sp1-DNA retarded band was both supershifted and immuno-depleted using Sp3 antibodies (lane 9), whereas only immuno-depletion was observed with a different Sp3 antibody preparation (see Fig. 3A). Physical interactions between ER␣ and Sp3 were determined in coimmunoprecipitation assays using in vitro translated [ 35 S]Sp3 and ER␣ (Fig. 6B, lanes 2 and 3). Sp3 antibody immunoprecipitates  6 and 7). In contrast, ER␣ primarily interacted with the ZD domain of Sp1, and only minimal interaction with other regions of the protein were observed (65). Interactions of wild-type radiolabeled ER␣ (Fig. 6D, lanes 1-3), HE11 (lanes 4 -6), HE15 (lanes 7-9), and HE19 (lanes 10 -12) with GST-Sp3 demonstrated that Sp3 specifically bound with wild-type ER␣, HE11, and HE15, whereas only nonspecific interactions were observed for HE19, suggesting specific interactions of Sp3 with the C-terminal (AF-1 and DNA binding domain) region of ER␣. In parallel studies, GST-Sp1 interacted with wild-type ER␣, HE11, HE15, and H19 (data not shown).
We have also used a highly sensitive viral SssI CpG methylase (40,41) to footprint the G/GC-rich region of the VEGF gene promoter (Fig. 7). Incubation with increased amounts of nuclear extracts from HEC1A cells treated with 10 nM E2 increase footprinting in the GC-box area, and there was a Ͼ50% increase in the cytosine footprint at Ϫ56 and Ϫ64 (data not shown). In vitro footprinting using expressed recombinant proteins (Fig. 7) showed that Sp1 (lanes 5 and 6) and Sp3 (lanes 7 and 8) proteins alone or in combination with ER␣ (lanes 10 and 11) protected several GC sites within the proximal region of the VEGF promoter, and this was particularly evident at cytosines Ϫ56 and Ϫ61. Surprisingly, coincubation with Sp1 plus Sp3 proteins (lane 9) significantly reversed the footprint observed for the proteins alone, and this may be due to competitive protein-protein interactions resulting in decreased protein-DNA binding. DISCUSSION VEGF plays a pivotal role in angiogenesis and endothelial cell growth as well as the growth and metastasis of various primary tumors (1)(2)(3)(4)(5)(6)(7)(8)(9). Not surprisingly, cell type-specific regulation of VEGF expression has been extensively investigated, and multiple mechanisms and factors can affect VEGF mRNA and/or protein levels. VEGF and related family members are essential components for endothelial cell growth and, in turn, a variety of growth factors, cytokines, and kinases also up-regulate VEGF mRNA and/or protein levels in different cell types (36, 37, 66 -75). Hypoxia also induces VEGF (76 -84), and there are also several cellular conditions or factors that down-regulate expression of this gene product, including the tumor suppressor gene p53 and the von Hippel-Lindau tumor suppressor gene product (85)(86)(87)(88)(89).
Previous studies on hormone-dependent regulation of VEGF expression indicate that the responses are cell/tissue-specific (28 -34). Recent studies in this laboratory have demonstrated that the HEC1A endometrial carcinoma cell line expresses ER␣ and is E2-responsive (35); however, analysis of VEGF mRNA levels and activities of constructs containing VEGF gene promoter inserts (Fig. 1) show that E2 down-regulates VEGF. This hormone-mediated response is dependent on ER␣ and can be reversed by the pure anti-estrogen ICI 182,780 (but not tamoxifen; data not shown); however, in cells cotransfected with pVEGF4 or pVEGF5 and PR expression plasmid, treatment with progesterone did not alter reporter gene activity. Thus down-regulation of VEGF by E2 parallels results of previous studies showing that p53 and von Hippel-Lindau factor also decreased VEGF activity and/or VEGF gene promoter-dependent constructs (85)(86)(87)(88)(89). Analysis of the VEGF gene promoter has shown that high basal activity of constructs is dependent on a GC/G-rich sequence from Ϫ88 to Ϫ50 that contains putative binding sites for Sp1-like proteins, Egr-1 and AP-2 (36,37,74,75), and a similar pattern of basal activity was also observed in HEC1A cells (data not shown). Interestingly transforming growth factor-␣ and p42/p44 mitogen-activated protein kinase activate VEGF gene promoter constructs at GC-rich regions between Ϫ88 to Ϫ66 that bind AP-2 and Sp1 proteins, and both transcription factors play a role in this response in hamster CCL39 fibroblast and A431 human epidermoid carcinoma cells (37,74). In contrast, mutation of the AP-2 site did not affect platelet-derived growth factor-induced responses that appeared to be dependent on Sp1 and/or Sp3 interactions with three GC-rich sites between Ϫ85 to Ϫ50 (36). Results of deletion analysis of the VEGF gene promoter in HEC1A cells showed (Figs. 1B and 2D) that comparable ER␣-mediated FIG. 5. Activation of pVEGF6 by Sp1, Sp3, and hER␣ expression in Drosophila Schneider SL-2 cells and effects of dominant negative (DN)-Sp3 on pVEGF6 in HEC1A cells. A, effects of hER␣ cotransfection on Sp1-mediated transactivation of pVEGF6. In SL-2 cells transfected with Sp1 expression plasmid (100 ng), cotransfection with increasing amounts of hER␣ (20 -500 g) along with 10 nM E2 treatment significantly (*p Ͻ 0.05) increased luciferase activity, whereas hER␣ alone was inactive. B, effects of hER␣ cotransfection on Sp3 mediated transactivation of pVEGF6. Cells were transfected with Sp3 expression plasmid (100 ng) and treated with 10 nM E2; hER␣ (500 and 1000 ng) significantly (*p Ͻ 0.05) decreased Sp3-dependent activation of pVEGF6 (a*, plotted from a separate experiment in which a new batch of cells, culture medium, and fetal bovine serum were used). Luciferase activities were corrected for total protein instead of ␤-galactosidase activity. C, effects of dominant negative (DN)-Sp3 on E2decreased pVEGF6 activity in HEC1A cells. Cotransfection of increasing amounts (0.2-3.0 g) of pCMV-DNSp3 (expressing a DNA binding form of Sp3 that competes with endogenous Sp3 proteins but does not transactivate) resulted in a dose-dependent reversal of E2-mediated down-regulation of luciferase reporter activity in cells transiently transfected with pVEGF6 (0.5 g) and hER␣ (0.5 g). Significant (p Ͻ 0.05) reversal of the hormone-decreased response was observed using 0.5, 2.0, and 3.0 g of pCMV-DNSp3. In vitro translated Sp1 and Sp3 proteins were incubated with the VEGF promoter and SssI methylase footprint analysis was carried out as described under "Materials and Methods." Sp1 and Sp3 proteins dosedependently enhanced the footprint of GC-rich elements in the VEGF gene promoter (lanes 5-8); Sp1 and Sp3 footprints were enhanced by pure hER␣ protein (lanes10 -11). hER␣ enhanced footprints of Sp1 and Sp3 proteins at the GC-site critical for E2-mediated down-regulation of VEGF and also at adjacent 5Ј and 3Ј GC-rich sequences.
binding of both Sp1 (lanes 2-5) and Sp3 (lanes 6 -9) proteins was enhanced (2-3-fold) after coincubation with increasing amounts of recombinant hER␣ protein in the presence of 3.3 nM E2 treatment. However, hER␣ protein alone did not bind 32 P-(Ϫ66/Ϫ47). B, coimmunoprecipitation assay. This assay was performed as described under "Materials and Methods," and E2 was not added. In vitro translated [ 35 S]hER␣ protein is specifically precipitated after incubation with in vitro translated Sp3 protein and antibody to Sp3 (␣Sp3) (lane 6). As a negative control, incubation of [ 35 S]hER␣ with in vitro translated rat c-Fos protein and c-Fos antibody did not precipitate [ 35 S]hER␣ (data not shown). C, GST pull-down assay. GST-Sp3 fusion proteins were expressed in BL 21 E. coli and isolated; fusion protein concentrations and sizes were verified by SDS-PAGE and Coomassie Blue staining, and pull-down assays were carried out as described under "Materials and Methods" without the addition of E2. Interaction of Sp3 protein with HE19C was nonspecific, and the radiolabeled band was not higher than background binding observed for GST-Sp3, GSH-conjugated beads, or GST alone (lanes 10 -12). Negative control, rat c-Fos protein (lane 13), does not bind GST-Sp3 fusion protein (lanes 14 -16). S.S., supershift. down-regulation was observed for both pVEGF6 and pVEGF8, but not pVEGF7, suggesting that the Ϫ66 to Ϫ52 region of the promoter was required for the hormone-dependent inhibitory response.
This region of the promoter contains two GC-rich binding sites for Sp-like proteins and an overlapping Egr-1 site at Ϫ60 to Ϫ51. Results of gel mobility shift assays indicated that ER␣ did not bind directly to this region of the promoter and nuclear extracts formed retarded bands that primarily contained Sp1 and Sp3 proteins. Egr-1 competition with Sp1 binding to GCrich sites can result in repressed gene expression (48,49), and previous studies indicate the Egr-1 can be induced by E2 in some hormone-dependent cell lines (44). Our studies show the in vitro expressed Egr-1 did not bind the Ϫ66 to Ϫ47 oligonucleotide, suggesting that Sp1 and/or Sp3 are the major proteins directly interacting with this region of the gene promoter.
Sp1 physically and functionally interacts with several nuclear receptor superfamily proteins including ER␣ (53)(54)(55)(56)(57)(58)(59)(60)(61)65), retinoic acid receptors, chicken ovalbumin upstream promoter transcription factor (COUP-TF), and steroidgenic factor-1, and the progesterone receptor interacts with Sp1 and basal transcription element binding protein (an Sp family member) to modulate gene expression through GC-rich promoter elements (93)(94)(95)(96)(97)(98)(99). This paper describes a novel ER␣/Sp3 pathway for hormone-mediated decreased gene expression through a G/GCrich element in the VEGF gene promoter, and this mechanism may be important for hormone-mediated down-regulation of other genes. Results of preliminary studies with pVEGF constructs in other breast and endometrial cancer cell lines suggest that changes in Sp3/Sp1 protein ratios do not necessarily dictate differential ER␣/Sp3 or ER␣/Sp1 action at GC-rich sites. Therefore, current research is focused on identifying other cellular factors that differentially regulate ER␣ interactions with Sp1 or Sp3 proteins.