Peroxisome Proliferator-activated Receptor γ Suppresses Proximal α1(I) Collagen Promoter via Inhibition of p300-facilitated NF-I Binding to DNA in Hepatic Stellate Cells*

Depletion of peroxisome proliferator-activated receptor γ (PPARγ) represents one of the key molecular changes that underlie transdifferentiation (activation) of hepatic stellate cells in the genesis of liver fibrosis (Miyahara, T., Schrum, L., Rippe, R., Xiong, S., Yee, H. F., Jr., Motomura, K., Anania, F. A., Willson, T. M., and Tsukamoto, H. (2000) J. Biol. Chem. 275, 35715-35722; Hazra, S., Xiong, S., Wang, J., Rippe, R. A., Krishna, V., Chatterjee, K., and Tsukamoto, H. (2004) J. Biol. Chem. 279, 11392-11401). In support of this notion, ectopic expression of PPARγ suppresses hepatic stellate cells activation markers, most notably expression of α1(I) procollagen. However, the mechanisms underlying this antifibrotic effect are largely unknown. The present study utilized deletion-reporter gene constructs of proximal 2.2-kb α1(I) procollagen promoter to demonstrate that a region proximal to -133 bp is where PPARγ exerts its inhibitory effect. Within this region, two DNase footprints with Sp1 and reverse CCAAT box sites exist. NF-I, but not CCAAT DNA-binding factor/NF-Y, binds to the proximal CCAAT box in hepatic stellate cells. A mutation of this site almost completely abrogates the promoter activity. NF-I mildly but independently stimulates the promoter activity and synergistically promotes Sp1-induced activity. PPARγ inhibits NF-I binding to the most proximal footprint (-97/-85 bp) and inhibits its transactivity. The former effect is mediated by the ability of PPARγ to inhibit p300-facilitated NF-I binding to DNA as demonstrated by chromatin immunoprecipitation assay.

Cirrhosis, the advanced stage of liver fibrosis, is the 12th leading cause of medial mortality in 2002 with 27,257 annual deaths according to a report by the Center for Disease Control. This mortality is even higher among those with productive ages between 45 and 54, ranking it as the fourth leading cause of death, highlighting medical and socioeconomic significance of the disease (1). Currently, there is no medical treatment for the disease other than liver transplantation. Therefore, the understanding of cellular and molecular mechanisms of liver fibrogenesis is of primary importance for the development of new treatments. The effector cell type for liver fibrosis is the hepatic stellate cell (HSC). 2 HSCs are liver mesenchymal cells that are believed to function as pericytes for the liver microcirculatory system called sinusoids. They are located in the perisinusoidal space, an anatomical area nestled between the nonluminal surface of the sinusoidal endothelial cell and the microvilli surface of the hepatocyte (2). HSCs constitute 7-10% of the liver cell population and it stores 85% of the body's total vitamin A content (3,4). They also produce and maintain the normal matrix milieu (basement membrane components) of the perisinusoidal space. In addition, HSC provides direct and indirect homeostatic control over hepatocytes through communication via gap junctions (5) and the release of soluble factors such as hepatocyte growth factor (6), epimorphin (7), and pleiotrophin (8). However, these cells are also responsible for a severalfold increase in the production of extracellular matrix components in the genesis of liver fibrosis (4).
Upon fibrogenic stimulation, quiescent HSCs transdifferentiate to myofibroblastic cells to produce excessive extracellular matrix. This cellular transition process is characterized by the loss of vitamin A storage; cellular proliferation and migration; acquisition of a myofibroblastic phenotype, such as expression of ␣ smooth muscle actin, induction of fibrogenic extracellular matrix genes (collagen type I and III), expression of autocrine cytokines, such as platelet-derived growth factor (9), transforming growth factor ␣, transforming growth factor ␤ (10 -13), and their receptors; and expression of adhesion molecules (intercellular adhesion molecules); and chemokines (MCP-1, CINC) (14). In search of the molecular basis of this unique HSC transdifferentiation phenomenon, we and others recently disclosed that this process is accompanied by reduced levels of a member of the nuclear hormone receptor family, peroxisome proliferator-activated receptor ␥ (PPAR␥) (15,16). Further, our subsequent studies demonstrated that this molecular "defect" in transdifferentiated HSC is part of the loss of adipogenic transcriptional program required for the maintenance of HSC quiescence (17). In fact, this loss of the adipogenic program and the transdifferentiation process are coordinately reversed by ectopic expression of PPAR␥ (18) or sterol regulatory element-binding protein 1c (17), another key adipogenic transcription factor.
One of the most pivotal antifibrotic effects of PPAR␥ is its ability to inhibit type I collagen expression at the level of the transcription (15,16). Type I collagen makes up 40 -50% of the total collagen proteins in the normal liver and is increased to 60 -70% in the cirrhotic liver (3). Type I collagen is a heterotrimeric protein composed of two ␣1(I) and one ␣2(I) collagen polypeptides encoded by two different genes that are coordinately up-regulated in liver fibrogenesis (19). Treatment of activated HSC with a ligand for PPAR␥ (15,16) or transduction of these cells with a PPAR␥ plasmid (15,16) represses basal ␣1(I) procollagen promoter activity that is largely dependent on a proximal 2.2-kb 5Ј-flanking region (19,20). However, the mechanism by which PPAR␥ inhibits type I collagen promoter activity is currently unknown. The present study investigated where in the proximal 2.2 kb ␣1(I) procollagen promoter PPAR␥ renders its inhibition and how it achieves this effect. Our results demonstrate that the 5Ј-flanking ␣1(I) procollagen promoter proximal to Ϫ133 bp is where PPAR␥ renders its inhibitory effect. This inhibition is mediated by the ability of PPAR␥ to suppress NF-I binding and transactivity via inhibition of p300-facilitated NF-I binding.

MATERIALS AND METHODS
HSC Isolation and Cell Culture-HSCs were isolated from normal male Wistar rats as previously described (21). Briefly, nonparenchymal cells were isolated via sequential digestion with Pronase and type IV collagenase, followed by differential low speed centrifugation. A pure fraction of HSC was isolated by arabinogalactan gradient ultracentrifugation and collecting the cells at the interface between the medium and a 1.035 density gradient. Cell purity was determined using phase-contrast microscopy and UV-excited fluorescence microscopy. Cell viability was determined by trypan blue exclusion. Cells were cultured on a 100-mm dish in low glucose Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 100 mg/ml streptomycin, 10,000 units/ml penicillin, and 25 g/ml amphotericin B. Cells were maintained in the culture medium for 7 days, at which time adenoviral vectors were added. Spontaneously immortalized, activated HSCs were established from a rat with cholestatic liver fibrosis and termed as biliary fibrosis-derived stellate cells (BSCs) (22). These cells have the phenotype similar to activated HSC. BSCs were cultured in 10% low glucose medium and primarily used in transient transfection experiments.
Adenoviral Vector Infection-Full-length PPAR␥1 cDNA was cloned from pCMX-PPAR␥1 into the transfer vector, subsequently allowing homologous recombination with the pAdEasy-1 adenoviral plasmid containing a green fluorescent protein (GFP) reporter (Stratagene, La Jolla, CA) as previously described (18). A control vector containing cytomegalovirus-driven GFP reporter gene was also constructed. On the seventh day of primary rat HSC culture, a vector was applied to HSCs at a multiplicity of infection of 100 to infect and transduce PPAR␥ or GFP. The following day, the medium was changed, and the cells were cultured for an additional four days to observe the effects on HSC as described (18).
Plasmids and Transient Transfection-pCMX and pCMX-PPAR␥ were gifts from Ron Evans (The Salk Institute, La Jolla, CA). pSG5 and p300 were gifts from Michael Stallcup (University of Southern California). Mouse collagen promoter deletion constructs were as follows: pCol2-lucif, pCol3-lucif, pCol6-lucif, pCol7-lucif, and pUC-Cat (Ϫ220, Ϫ133, Ϫ120, and Ϫ92 bp/ϩ115 bp) were used as previously described (19,20). pPac, pPac-NF-I and pPac-Sp1 were used to assess the effects of PPAR␥ on NF-I-or Sp1-driven promoter activity. Renilla pRL-TK was purchased from Promega. Luciferase promoter deletion constructs were created via restriction enzyme digestion (XbaI and XhoI) of the pUC-Cat constructs and insertion into the pGL3-luciferase (Promega) backbone. Briefly, BSC or NIH3T3 cells were seeded in 6-or 24-well plates and incubated overnight. A collagen promoter construct, an expression plasmid (pCMX or PPAR␥), and F2 reagent (Targeting System, San Diego, CA) were mixed and added to serum-free, high glucose Dulbecco's modified Eagle's medium and incubated for 25 min at 37°C and then placed onto the cells. Two hours later, the cells were supplemented with 10% high glucose Dulbecco's modified Eagle's medium. The following day, the medium was changed, and the cells were incubated for an additional 8 h. The cell lysates were collected using 5ϫ passive lysis buffer (Promega) and dual luciferase assay (Promega) was performed using a luminometer (E&G Berthold). Mutations of three nucleotides (TGG to CAA) in the most proximal reverse CCAAT box was created by site-directed mutagenesis according to the QuikChange TM protocol (Stratagene). Primers were designed to introduce 3-nucleotide mutations into the wild type CCAAT of the FP-1 region of the luciferase promoter deletion construct (Ϫ133 bp/ϩ115 bp). The DNA sequence of each construct was verified using an ABI Prism 377 sequencer (PerkinElmer Life Sciences). The following primers were used: FP-1-mutant NF-I (forward), 5Ј-gggccaggcagttctgatCA-Actgggggccgggctgctggctc-3Ј; FP-1-mutant NF-I (reverse), 5Ј-gagccagc-agcccggcccccagTTGatcagaactgcctggccc-3Ј.
The reaction mixture was resolved on a 6% nondenaturing polyacrylamide gel (Bio-Rad) in 0.5ϫ TBE (45 mM Tris, 45 mM boric acid, 1 mM EDTA). The gel was dried and subjected to phosphorimaging for detection of shifted bands. For a supershift analysis, polyclonal antibodies against NF-I, Sp1, or Sp3 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) were added and incubated on ice for an additional 30 min. For competition analysis, a 200-fold molar excess of a cold probe was added to the reaction mixture just prior to the addition of ␣-32 P-labeled probe.
Chromatin Immunoprecipitation (ChIP) Assay-The ChIP assay was performed using the ChIP assay kit according to the manufacturer's protocol (Upstate Biotechnology, Inc., Lake Placid, NY). In brief, ChIP assay was performed on HSC cultured on plastic for 7 days or NIH3T3 cells without or with transfection with NF-I, PPAR␥, and/or p300 expression plasmids or respective empty vectors. After a 48-h incubation, ϳ4 ϫ10 6 cells/ChIP assay, were cross-linked with 1% formaldehyde at 37°C for 10 min and rinsed twice with ice-cold phosphatebuffered saline. The cells were harvested by brief centrifugation and lysed in SDS-lysis buffer (50 mmol/liter Tris-HCl, pH 8.1, 10 mmol/liter EDTA, 1% SDS, protease inhibitors). The lysates were sonicated on ice with two pulses at 15 s each to achieve chromatin fragments ranging between 200 and 1000 bp in size followed by centrifugation at 15,000 rpm for 10 min at 4°C. Supernatants were collected and diluted 10-fold in a ChIP dilution buffer (a 20-l aliquot was removed to serve as an input sample) followed by preimmunoprecipitation clearing with 80 l of a mixture of salmon sperm DNA/Protein A at 4°C with rotation for 30 min. Immunoprecipitation was carried out with 1 g of antibodies (anti-NF-I, anti-p300, and anti-CCAAT DNA-binding factor (CBF)/ NF-Y antibodies from Santa Cruz Biotechnology) at 4°C overnight with rotation. After immunoprecipitation, 60 l of a mixture of salmon sperm DNA and Protein A was added and incubated at 4°C with rotation for 30 min and followed by brief centrifugation. The precipitates were washed twice with low salt buffer, once with high salt buffer, and once with LiCl buffer. Then the precipitates were washed again with the TE buffer. The immune complexes were extracted twice with 250 l of elution buffer. The extracted complexes and the input were heated at 65°C for 4 h after the addition of 20 l of 5 mol/liter NaCl to reverse cross-link. Following proteinase K treatment, DNA was extracted by phenol/chloroform solution and precipitated with 20 g of glycogen.

PPAR␥ Inhibits ␣1(I) Procollagen mRNA Expression and Promoter
Activity-PPAR␥ is depleted in activated HSC, whereas activation markers including the ␣1(I) collagen gene are induced (15,16). Ectopic expression of PPAR␥ in culture-activated HSC by an adenoviral vector reduces the level of ␣1(I) procollagen mRNA by half as compared with HSC transduced with a control GFP vector as determined by real time PCR (Fig. 1A). Transient transfection experiment using the HSC cell line (BSC cells) reveals that PPAR␥ expression also decreases by 50% the activity of a proximal 2.2-kb ␣1(I) procollagen promoter known to encompass the highest basal activity in fibroblasts (19,20) (Fig. 1B). These results confirm the previous finding (15,16,18) and further support that the inhibitory effect of PPAR␥ is at the level of the proximal promoter.  . PPAR␥ suppresses the collagen promoter within the ؊220 bp proximal promoter region. A, a schematic diagram of four deletion constructs within the proximal 2.2 kb ␣1(I) collagen promoter. B, relative promoter activities of the collagen promoter deletion constructs as compared with the highest activity achieved by the Ϫ220/ϩ115 bp promoter. *, p Ͻ 0.05 as compared with the activity of pCOL3 (n ϭ 4). C, PPAR␥ equally inhibits each deletion construct by 50%, suggesting that the primary site of the inhibitory effect of PPAR␥ is located within the Ϫ220 bp collagen promoter region. *, p Ͻ 0.05 as compared with pCMXtransfected cells (n ϭ 4).

A Site of the Inhibitory Effect of PPAR␥ Is Located within the Most
Proximal Ϫ220 bp ␣1(I) Collagen Promoter Region-In order to assess the region within the 2.2-kb promoter that is subjected to PPAR␥-mediated inhibition, we performed transient transfection experiments using four deletion constructs of the promoter (Fig. 2A). The relative activity of each of the deletion constructs was tested first by transfection experiments in BSCs (Fig. 2B). Our results reveal that two repressor elements: one between Ϫ2.2 and Ϫ1.8 kb and another between Ϫ1.1 and Ϫ220 bp. An enhancer element is also found between Ϫ1.8 and Ϫ1.1 kb. Of all four deletion constructs, the region proximal to Ϫ220 bp is shown to have the highest promoter activity in consistent with the previous findings (19,20). Co-transfection of each deletion construct with a PPAR␥ or empty vector reveals that PPAR␥ inhibits all of the deletion constructs by ϳ50% (Fig. 2C). Since the Ϫ220 bp region has the highest activity and PPAR␥ expression equally reduces the activity of each promoter construct, we concluded that PPAR␥ primarily exerts its inhibitory effect within the Ϫ220 bp proximal region and ϩ115 bp of the first exon. Our review of the Ϫ220 bp proximal promoter sequence fails to reveal a PPRE. Thus, PPAR␥ must mediate the effect not via direct repression but via its interaction with other trans-acting factor(s) and/or cis-regulatory element(s).
Inhibition of PPAR␥ Is Confined to DNase Footprint (FP)-1 and -2 Regions within the Ϫ133 bp ␣1(I) Collagen Promoter-To further define the site of PPAR␥-mediated inhibition, we designed and created an additional set of deletion constructs within the most proximal Ϫ220 bp region. This region is known to contain four protected footprints as determined by DNase footprinting analysis of activated HSC (24) (Fig.  3A), and newly created deletions are designed to test these FP regions: Ϫ133/ϩ115 (FP-3 and FP-4 deleted but intact FP-2 and -1); Ϫ120/ϩ115 (distal region of FP-2 deleted but intact FP-1); and Ϫ92/ϩ115 bp (only proximal region of FP-1). Transient transfection experiments reveal that the two distal footprints (FP-3 and FP-4) contribute minimally to basal collagen promoter activity (a statistically significant change in the promoter activity is not attained by this deletion: pCOLϪ133) (Fig. 3B). However, an additional deletion, including the distal half of FP-2 reduces the basal promoter activity to 50%, and a further deletion within FP-1 reduces the promoter activity by another 50%. Overexpression of PPAR␥ expression vector results in a 50% inhibition on the pCOLϪ133 promoter (Fig. 3C). This effect is attenuated when FP-2 is disrupted (pCOL-120). PPAR␥ also retains a modest inhibitory effect on the promoter activity rendered only by the most proximal portion of the FP-1 (pCOL-92). However, the absolute magnitude of the activity inhibited by PPAR␥ accounts only for 25% of the inhibition seen with pCOLϪ220 or pCOLϪ133. These results suggest that both FP-1 and FP-2 contain the sites via which PPAR␥ renders a major inhibitory effect on the promoter.

PPAR␥ Inhibits NF-I Binding to FP-1 but Has No Effects on Sp1
Binding to FP-2-Since FP-1 and FP-2 are shown to be the most likely sites of the inhibitory action of PPAR␥, we next investigated what trans-acting factors bind to these regions in activated HSC. NF-I and Sp1 are shown to comprise protein components that bind to FP-1 in NIH3T3 fibroblasts (19) and activated HSC (20). We used electrophoretic mobility shift assays (EMSA) to characterize these DNA-protein interactions of FP-1 in activated HSC. Using nuclear extracts from culture-activated HSC and a radiolabeled oligonucleotide probe containing the FP-1 sequence, one major DNA-protein complex and another faint, higher molecular weight band are detected. To identify proteins bound to FP-1, supershift assays were performed using antibodies against NF-I, Sp1, and Sp3. Although a supershifted band is not detected, NF-I antibody clearly diminishes the major band and Sp1 antibody abrogates only the upper, faint band. Sp3 antibody fails to affect either band (Fig. 4A). The addition of a cold consensus NF-I binding sequence in excess, completely eliminates the protein binding to the FP-1 probe, as does the cold FP-1 probe (Fig. 4C). The addition of a cold Sp1 consensus element in excess eliminates the upper protein-DNA complex without affecting the lower major band (Fig. 4C). These results demonstrate that the lower major band of DNA-protein complex contains NF-I, and the upper band contains Sp1 and maybe NF-I.
We also performed EMSA to determine proteins bound to the FP-2 DNA. Using nuclear extracts from activated HSC and a FP-2 probe, two DNA-protein bands are identified. Supershift assays reveal that Sp1  -1 and FP-2 sites. A, a schematic diagram of deletion constructs within the Ϫ220 bp ␣1(I) collagen promoter in reference to the known four DNase footprints. B, relative activities of the collagen promoter deletion constructs transfected in BSCs. *, p Ͻ 0.05 as compared with pCOLϪ220 (n ϭ 5). C, co-transfection experiments using deletion constructs and a PPAR␥ expression vector reveals that PPAR␥ inhibits the Ϫ220 and Ϫ133 bp promoters by 45-50%, but this inhibition is attenuated when FP-2 is disrupted in pCOL-120. PPAR␥ still suppresses pCOL-92 containing the proximal portion of FP-1. *, p Ͻ 0.05 as compared with pCMX-transfected cells (n ϭ 5-7). antibody diminishes the upper band, whereas Sp3 antibody completely abrogates the lower band (Fig. 4B). NF-I antibody has no effects. The addition of a cold consensus Sp1 binding element in excess completely eliminates the upper band, whereas an excess amount of unlabeled FP-2 decreases both upper and lower bands (Fig. 4D). Thus, these results demonstrate that the upper DNA-protein complex contains Sp1, a 105/ 95-kDa protein, and the lower band contains Sp3, an 80-kDa protein, in activated HSC.
Next, we tested whether expression of PPAR␥ affects the binding of the nuclear proteins to FP-1 and FP-2. We performed EMSA using  DECEMBER 9, 2005 • VOLUME 280 • NUMBER 49 nuclear extracts from activated HSC infected with either Ad.PPAR␥ or Ad.GFP. Our results show that PPAR␥ does not affect protein binding to FP-2 (Fig. 4E, middle panel) but reduces the intensity of complexes formed with the FP-1 that contain NF-I (second lane of the left panel) (Fig. 4E). A PPAR␥-specific DNA binding sequence, ARE-7, was used to confirm increased binding of PPAR␥ by the nuclear extracts of the Ad.PPAR␥-infected cells (Fig. 4E, right panel). CBF/NF-Y binds to a CCAAT box in the proximal murine ␣2(I) procollagen promoter in NIH3T3 cells (25)(26)(27). Thus, the binding of this factor to a reverse CCAAT site in FP-1 of ␣1(I) collagen promoter is possible. To test this, we first used anti-CBF antibody for EMSA with the FP-1 probe and HSC nuclear extracts. As shown in the fourth lane of the left panel of Fig.  4E, this antibody does not affect the DNA-protein complex formation, whereas anti-NF-I antibody clearly decreases it (last lane). Using the same antibody, we also performed a ChIP assay. For this, we used both NIH3T3 cells and HSC spontaneously activated in culture on plastic. CBF binding to FP-1 was evident in NIH3T3 cells but not in HSC (Fig.  4F). These results demonstrate that CBF does not bind to the reverse CCAT site in FP-1 in HSC.

PPAR␥ Inhibition of Collagen Promoter
NF-I Is Most Important for Proximal ␣1(I) Procollagen Promoter Activity-The above data suggest that NF-I is a likely target of PPAR␥. To better understand the functional importance of NF-I in the context of the promoter of interest, 3 nucleotides in the reverse CCAAT site of FP-1 that is shown to be critical for NF-I binding (19) were mutated (Ϫ98/Ϫ96; TGG 3 CAA), and the activity of the mutated pCOLϪ133 promoter was assessed in BSCs. As shown in Fig. 5A, this mutation almost completely abrogates the promoter activity. Because of very low promoter activity of the mutated promoter, the effect of PPAR␥ could not be assessed. Nonetheless, this result demonstrates the utmost importance of NF-I binding to the CCAAT box in FP-1 for proximal ␣1(I)pro-collagen promoter activity in HSC that appears to be a target of PPAR␥.
PPAR␥ Inhibits NF-I but Not Sp1-mediated Stimulation of the Ϫ133 bp Collagen Promoter-To further test the effects of PPAR␥ on NF-Imediated proximal promoter activity, transient transfection experiments were performed using the Ϫ133 promoter-reporter plasmid (pCOLϪ133) and an expression vector for NF-I, Sp1, or both in the absence or presence of PPAR␥ expression. We first examined the regulatory effects of NF-I and Sp1 on the promoter. For this purpose, we used NIH3T3 fibroblasts that have been previously used to characterize NF-I and Sp1 regulation of this specific promoter (19). A transient transfection experiment revealed that NF-I overexpression modestly but dose-dependently transactivated the pCOLϪ133 promoter. A mild 1.8-fold increase in promoter activity is achieved with 300 ng of NF-I plasmid, the maximum amount used for this experiment (Fig. 5B). On the other hand, Sp1 overexpression markedly increases the promoter activity as much as 350-fold with 400 ng of Sp1 (Fig. 5C). We also tested the combined effect of NF-I and Sp1 on the collagen promoter by cotransfecting with 100 ng each of both plasmids. This co-expression results in a synergistic induction of the promoter activity in NIH3T3 (Fig. 5D), and the same effect is also confirmed in BSCs (Fig. 5E). We then tested the effects of PPAR␥ on NF-I-or Sp1-mediated activation of the collagen promoter as well as the effect of PPAR␥ on the synergistic activation of this promoter by both factors. Our results show that PPAR␥ significantly inhibits NF-I-mediated promoter activity in NIH3T3 and BSCs (Fig. 5, D and E). In contrast and to our surprise, PPAR␥ does not inhibit Sp1 transactivity toward this promoter in either cell type (Fig. 5, D and E). In addition, PPAR␥ significantly reduces NF-I/Sp1 synergistic induction of the collagen promoter by in NIH3T3 and BSCs (Fig. 5, D and E). These results confirm NF-I as a primary target for the inhibitory effect of PPAR␥ on the proximal ␣1(I) collagen promoter.
p300 Rescues the Inhibitory Effect of PPAR␥ on NF-I-mediated Collagen Promoter Activity-Our results demonstrate that PPAR␥ inhibits basal ␣1(I) procollagen promoter activity at the site upstream of Ϫ133 bp comprising overlapping NF-I and Sp1 sites. They also demonstrate that the promoter activity stimulated by NF-I but not Sp1 is inhibited by PPAR␥, and this effect is probably due to reduced NF-I binding to FP-1. Since NF-I and PPAR␥ share common co-activators, such as p300, and a competition for p300 by PPAR␥ with other transcription factors is possible, we examined whether the inhibitory effect of PPAR␥ on NF-I involves p300. To address this question, we overexpressed p300 to determine its effects on PPAR␥-mediated inhibition of NF-I-dependent promoter activity. Our results show that 330 ng of the p300 expression plasmid partially rescues NF-I from the inhibitory effect of PPAR␥ on each of the three collagen promoter deletion constructs tested, pCOLϪ133/ϩ115bp, pCOLϪ120/ϩ115bp, and pCOL-92/ϩ115bp, suggesting the involvement of p300 within FP-1 (Fig. 6A). In addition, p300 dose-dependently relieves the suppression by PPAR␥ of basal pCOLϪ133 promoter activity (Fig. 6B, second set of bar graphs). More importantly, p300 dose-dependently rescues NF-I from the suppressive effect of PPAR␥ on the promoter (Fig. 6B, third set of bar graphs). Thus, these results suggest that PPAR␥ inhibits NF-I-mediated collagen promoter activity at least in part via competition between these two transcription factors for p300.

PPAR␥ Inhibits NF-I Binding to FP-1 in Vivo via Its Suppressive
Effects on p300-stimulated NF-I-DNA Binding-We next tested whether PPAR␥ inhibits NF-I binding to FP-1 in vivo by the ChIP assay. First, using antibodies against NF-I and primers flanking FP-1, we amplified the FP-1 region that is bound with NF-I in NIH3T3 cells transduced 1) with or without PPAR␥, 2) with NF-I with or without PPAR␥. Under the basal condition (no NF-I overexpression), the amplified band for FP-1 in agarose gel is too weak to ascertain the effects of PPAR␥. Thus, real time PCR using SyBer Green was performed to quantify the amplified FP-1 DNA. Using this technique, the binding of NF-I to FP-1 is shown to be reduced by 40% by PPAR␥ (Fig. 7A). ChIP assay also reveals NF-I binding to FP-1 under NF-I transduction is clearly inhibited by PPAR␥ (Fig.  7B, lanes 1 and 2). These results confirm the conclusion from our in vitro data obtained by EMSA (Fig. 4E). Next, we examined the binding of p300 to FP-1 by immunoprecipitating DNA with anti-p300 antibodies. As shown in the Fig. 7B, lanes 3 and 4, the p300 binding is coordinately suppressed by PPAR␥. To determine whether p300 binding to FP-1 via NF-I is reduced, we performed a two-step ChIP assay. For this, we first immunoprecipitated DNA with anti-NF-I antibodies, followed by PCR and then reimmunoprecipitation with anti-p300 antibodies after a brief treatment with dithiothreitol for the second PCR. As shown in the 5th and 6th lanes of Fig. 7B, PPAR␥ reduces p300 association with FP-1 via NF-I. As a positive control for PPAR␥ transduction, we amplified the PPRE site for the aP2 gene after p300 immunoprecipitation. As shown in the last two lanes of Fig. 7B, transfection with the PPAR␥ vector expectedly increases the binding of p300 to the PPRE. These results may suggest that NF-I binding to FP-1 is reduced by PPAR␥, and p300 recruit-ment to NF-I is consequently decreased. Then why does p300 overexpression rescue the PPAR␥ inhibitory effect on NF-I-driven promoter activity (Fig. 6)? Since the effect of PPAR␥ is mediated by suppressed NF-I binding, p300 must promote NF-I binding. This notion was tested by examining the effects of p300 on NF-I binding to FP-1 and on PPAR␥-mediated inhibition of NF-I binding using the ChIP assay. Indeed, p300 promotes the binding of NF-I to FP-1 (Fig. 7C, lane 3 versus lane 1). Further, PPAR␥-mediated inhibition of NF-I binding is prevented by p300 overexpression (Fig. 7C, lane 2 versus lane 4). These results suggest that PPAR␥ suppresses NF-I binding to FP-1 via its inhibition of p300-facilitated NF-I binding to its cognate binding site.

DISCUSSION
The present study identified a region proximal to Ϫ133 in the proximal 5Ј-flanking ␣1(I) procollagen promoter as the site of PPAR␥-mediated inhibition. This region encompasses two known DNase footprints designated as FP-1 and FP-2 as determined using nuclear extracts from fibroblasts (19,20) and activated HSC (24). Both footprints have reverse CCAAT binding sites and two identical 12-bp GC repeats (19). The CBF is shown to bind to the most proximal CCATT site of murine ␣2(I) procollagen promoter in NIH3T3 cells (25)(26)(27) and of human ␣1(I) procollagen promoter in skin fibroblasts (28). However, the primary proteins that are shown to bind to FP-2 are Sp1 and Sp3, and NF-I is shown to bind to the Ϫ100/Ϫ96 bp CCAAT box in FP-1 in both NIH3T3 cells and HSC (19,24,29). In fact, our EMSA and ChIP assays fail to demonstrate the binding of CBF to FP-1 in HSC despite the fact that CBF binding to the same site is evident in NIH3T3 cells (Fig. 4, E and F) and skin fibroblasts (28). The literature suggests PPAR␥ causes negative cross-coupling with Sp1 as in the case for thromboxane receptor gene promoter activity in vascular smooth muscle cells (30). However, our transient transfection analysis and EMSA reveal inhibition of neither Sp1-induced pCOLϪ133␣1(I) promoter activity nor Sp1 binding to FP-2 by PPAR␥. Instead, both NF-I-stimulated promoter activity and NF-I-mediated synergistic induction of Sp1 transactivity are suppressed by PPAR␥, demonstrating NF-I as a primary molecular target of PPAR␥. To our knowledge, this is the first demonstration of negative regulation of NF-I by PPAR␥. Our mutation experiment clearly demonstrates the utmost importance of the Ϫ100/Ϫ96 CCAAT site in FP-1 to the proximal promoter activity (Fig. 5A). Because of the very low activity of the mutated promoter, the effect of PPAR␥ could not be examined. However, it seems apparent from our data that PPAR␥ targets NF-I interaction with this CCAAT site. As we look into the underlying mechanisms, PPAR␥ is shown to inhibit NF-I binding to FP-1 as demonstrated by EMSA and ChIP assay. More importantly, this effect is mediated by suppression of p300-facilitated NF-I binding to its binding element as revealed by a reversal by p300 of the inhibitory effects of PPAR␥ on both the promoter activity (Fig. 6) and NF-I binding to FP-1 (Fig. 7C).
CBP/p300 are promiscuous co-activators that contribute to transcriptional activation by many transcription factors. They serve to recruit components of the general transcriptional machinery such as TFIID, TFIIB, and RNA polymerase. Its binding to activation domains of transcription factors brings histone acetyltransferases close to specific nucleosomes in target gene promoters (for a review, see Ref. 31). They also possess intrinsic acetyltransferase activity toward not only histones but also transcription factors. As exemplified by the regulation of p53 by p300, p300-mediated acetylation of transcription factors increases their binding to DNA (32). Our results also demonstrate a p300-mediated increase in NF-I binding to FP-1 of the proximal ␣1(I) procollagen promoter (Fig. 7C), and this mechanism is shown to be a target of the inhibitory action of PPAR␥. p300 also interacts with multiple transcription factors to facilitate a synergism for transcriptional activation. In our study, synergism between NF-I and Sp1 is demonstrated for pCOLϪ133 activity, although NF-I alone has a modest stimulatory effect (Fig. 5B). Due to the overlapping binding sites for these two transcription factors in FP-1 and FP-2 and their proximity, this synergism is most likely facilitated by p300. If that were the case, PPAR␥ would readily reduce the synergistic activation of the collagen promoter by its inhibitory effect on p300. As demonstrated by the deletion analysis of the Ϫ220/ϩ115 bp promoter, the intact FP-1 and FP-2 are required for the maximal PPAR␥ effect (Fig. 3C). Since the Sp1 and NF-I sites in FP-2 are disrupted in pCOL-120, the extent of inhibition is clearly attenuated (Figs. 3C and 5A). However, our EMSA data show the binding of Sp1 and Sp3 but not NF-I to FP-2 (Fig. 4B). Thus, these results suggest that the synergistic interaction of NF-I bound to FP-1 (Fig. 4A) and Sp1 bound to FP-2 is the target of PPAR␥, and this is disrupted in the deletion construct pCOL-120. This specific aspect will need to be addressed further by our future study.
Although our study identifies NF-I as a target for PPAR␥-mediated inhibition of the proximal ␣1(I)collagen promoter, our deletion analysis demonstrates that PPAR␥ still possesses a small inhibitory effect on pCOL-92 that has the proximal NF-I site disrupted (Figs. 3C and 5A). The absolute magnitude of this inhibition accounts only for 25% of the inhibition observed with pCOLϪ220 or pCOLϪ133. Nevertheless, this result still suggests an NF-I-independent mechanism for PPAR␥-mediated inhibition. Indeed, the most proximal region upstream of Ϫ92 is known to have the binding sites for other proteins. For instance, YY-1, also known as NF-E1, binds to the element consisting of (C/t/ a)CATN(T/a)(T/g/c) located at regions Ϫ40 to Ϫ34 bp and Ϫ35 to Ϫ29 bp of the ␣1(I) collagen promoter and up-regulates the transcription via stabilizing the interaction between TBP and TFIID and other components of the transcription machinery (33). YB-1, a member of the cold shock domain protein superfamily, binds to a specific DNA sequence (CTGATTGG) at Ϫ83 to Ϫ59 bp and inhibits the promoter activity due to its ability to separate DNA strands and to prevent the binding of other transcription factors (29). Thus, PPAR␥ may affect the binding or transcriptional regulation by these factors to render the observed inhibition of pCOL-92 activity. Further, PPAR␥ may also interact with other factors around the transcription start site, including the RFX family, that serve as repressors for both collagen type I genes (34).
In summary, the present study demonstrates that PPAR␥ reduces NF-I-mediated ␣1(I) collagen promoter activity via its ability to inhibit p300-facilitated binding of NF-I to DNA. The identification of the molecular target for PPAR␥ should aid in understanding the molecular basis of the antifibrotic effects mediated by PPAR␥.