Peroxisome Proliferator-activated Receptor γ-mediated Transcriptional Up-regulation of the Hepatocyte Growth Factor Gene Promoter via a Novel Composite cis-Acting Element*

Hepatocyte growth factor (HGF) is a pleotropic polypeptide that can function as a morphogen, motogen, mitogen, angiogen, carcinogen, and tumor suppressor, depending on the target cell and tissue. Previous studies from our laboratory using transgenic mice have shown that HGF gene expression is tightly regulated at the transcriptional level and that the upstream regulatory elements are crucial for the control of HGF gene transcription. In the present study, we have identified and characterized one of these elements as a peroxisome proliferator-activated receptor γ (PPARγ)-responsive element. This regulatory element was localized at −246 to −233 base pairs upstream from the transcription start site of the HGF gene promoter having the sequence GGGCCAGGTGACCT. Gel mobility shift and supershift assays demonstrated that this cis-acting element strongly binds to the PPARγ isoforms as well as to chicken ovalbumin upstream promoter-transcription factor, a member of the orphan nuclear receptor subfamily. Mutational analysis and gel mobility band shift assays indicated that the binding site is an inverted repeat of the AGGTCA motif with two spacers (inverted repeat 2 configuration) and that the two spacers are important for PPARγ binding. This binding site overlaps with functional binding sites for activating protein-2, nuclear factor 1, and upstream stimulatory factor, and together, they constitute a multifunctional composite binding site through which these different transcription factors exert their regulatory effects on HGF promoter activity. Functional assays revealed that PPARγ, with its ligand, 15-deoxy-prostaglandin J2, strongly stimulates HGF promoter activity. On the other hand, nuclear factor 1, activating protein-2, and chicken ovalbumin upstream promoter-transcription factor transcription factors repress the stimulatory action of PPARγ by competing with PPARγ for their overlapping binding sites. Furthermore, for the first time, our studies demonstrate that the PPARγ ligand, 15-deoxy-prostaglandin J2, induces endogenous HGF mRNA and protein expression in fibroblasts in culture.

Hepatocyte growth factor (HGF) 1 is expressed in the stromal cell compartment of a variety of tissues and acts on neighboring epithelial cells in a paracrine fashion. HGF stimulates cell growth, cell motility, and morphogenesis on its target cells through binding and activating its specific tyrosine kinase cell surface receptor known as Met (1)(2)(3)(4)(5). Gene knockout studies have shown that HGF has an essential role in liver growth during embryonic development (6,7). In adulthood, HGF also plays an important role in supporting the maintenance and renewal of cells in various organs such as liver, lung, and kidney.
HGF gene expression is regulated by extracellular cues such as hormones and cytokines (8 -12). In vivo, the level of HGF mRNA dramatically increases in response to cell loss to facilitate tissue regeneration. For example, HGF mRNA increases in the liver as well as in distal organs such as lung and spleen after loss of hepatic tissue induced by partial hepatectomy or hepatotoxin treatment. Similarly, after injury to the lung or kidney (by toxins or by surgical procedures such as unilateral pneumonectomy or nephrectomy, respectively), HGF gene expression is induced in these organs to promote their regeneration (13,14).
Several studies have shown that HGF gene expression is controlled tightly at the transcriptional level. In vivo analyses of HGF gene promoter regulation in transgenic mice (15) as well as in in vitro transient transfection studies have shown that important regulatory element(s) exist in the proximal promoter region of the HGF gene (16 -20). In recent studies, we have partially characterized this region and revealed that it harbors a composite element located at position Ϫ260 to Ϫ230 bp from the transcriptional start site. We showed that this composite site binds to members of the nuclear factor 1 (NF1), activating protein-2 (AP2), and upstream stimulatory factor (USF) families to regulate HGF gene transcription. Functional studies show that NF1 and AP2 suppress the activity of the HGF gene promoter whereas USF has an activating function (21,22).
In the present studies, we discovered that this region (Ϫ260 to Ϫ230) also harbors a functional PPAR␥-responsive element at Ϫ246 to Ϫ233 bp from the ϩ1 site of the HGF gene promoter having the sequence GGGCCAGGTGACCT. This is apparently a novel PPAR␥ site, which has an inverted RGGTCA motif with two spacer (IR2), rather than the well described DR1 (direct RGGTCA motif with one spacer) (23,24). Gel mobility shift and supershift assays using nuclear extracts from various sources such as whole liver, isolated non-parenchymal liver cells, fibroblast cell lines, and in vitro translated PPAR␥1 and PPAR␥2 protein demonstrated that this element binds to the nuclear receptor PPAR␥ family. Additionally, we found that COUP-TF, an orphan nuclear receptor, also binds to this nuclear hormone binding site and negates the stimulatory effects of PPAR␥. Functional studies revealed that PPAR␥, with its ligand, 15deoxy-PGJ2, strongly stimulates HGF promoter activity. On the other hand, NF1 and AP2 transcription factors repress the stimulatory function of PPAR␥ by competing with PPAR␥ for their individual overlapping binding sites present within this composite element. Moreover, for the first time, our studies demonstrated that the PPAR␥ ligand, 15-deoxy-PGJ2, induces the expression of the endogenous HGF mRNA and protein in cultured fibroblasts.

MATERIALS AND METHODS
Plasmids-0.7 mouse HGF-CAT promoter construct (Ϫ699 to ϩ29 bp) was described in an earlier work from our laboratory (16). 0.7 HGF-CAT/NF1-M, in which the NF1 binding site was mutated, was created by a sequential PCR mutagenesis method (21 Preparation of Nuclear Extracts-Mouse NIH3T3 and 3T3-L1 fibroblasts were originally obtained from the American Type Culture Collection (Manassas, VA) and cultured in the conditions suggested by the American Type Culture Collection. Cells were grown to ϳ90% confluence, washed twice with cold phosphate-buffered saline, and scraped with a rubber policeman in the same buffer. Nuclear protein extracts were prepared as described previously (20). Rat liver non-parenchymal cells were prepared from collagenase-perfused rat liver followed by centrifugation at 50 ϫ g to remove hepatocytes and were provided by Dr. G. K. Michalopoulos in our department. They were used to prepare nuclear protein extract.
A rat liver stellate cell line (Ito cell line) was kindly provided by Dr. M. Rojkind from the Albert Einstein College of Medicine and cultured as described previously (25,26). For preparation of mouse liver nuclear protein extracts, livers were removed from mice and homogenized in buffer A containing protease inhibitors (20). The nuclei were collected, and the nuclear proteins were extracted by the same method as described above.
DNA Transfection and CAT Assay-Mouse 3T3-L1 fibroblasts were cultured in 6-well plates for 24 h and then transfected with various mouse HGF promoter-CAT chimerical plasmids using the DNA calcium phosphate method according to the instructions of the CellPhect transfection kit (Amersham Pharmacia Biotech) as described previously (19). The ␤-galactosidase reference plasmid pCH110 (Amersham Pharmacia Biotech) was used as an internal control for monitoring the transfection efficiency and normalizing the data accordingly. The amount of plasmid used per well in transfection was as follows: 5 g of one chimerical CAT construct and 1 g of pCH110. After coprecipitation of the cells with DNA-calcium phosphate for 16 h in serum-containing medium, the cells were washed twice with serum-free medium and kept in serum-containing medium for an additional 24 h before harvesting for determination of CAT activity. CAT activity was determined as described previously (19,20). Transfections were performed at least three separate times with two independent preparations of purified plasmid DNA.
For cotransfection with COUP-TFI, AP2, NF1/X, RXR␣, and PPAR␥1 in 3T3-L1 cells, the amount of plasmid used per well in transfection was as follows: 5 g of one chimerical CAT construct, 1 g of pCH110, and different amounts of expression plasmids. The total amount of DNA per well was equilibrated by addition of corresponding empty (no insert) expression plasmid.
Gel Retardation Assays-The double-stranded oligonucleotides used in gel mobility shift assays were labeled with [␣-32 P]dCTP by end labeling with the Klenow fragment of DNA polymerase. The labeled oligonucleotide probes were then gel-purified and used in gel mobility shift assays as described previously (21). Two g of poly(dI-dC) (Amersham Pharmacia Biotech) were used as the nonspecific competitor in 10 l of reaction mixture. When antibodies for supershift were used, they (1 l) were incubated with nuclear extracts at room temperature for 20 min before performing the DNA binding shift assays. The binding reactions were carried out at room temperature for another 20 min before loading onto 5% nondenaturing polyacrylamide (19:1, acrylamide/bisacrylamide) gels. The amount of the nuclear protein extract used in each reaction was about 4 g, and that of the labeled probe was between 0.2 and 0.4 ng. For competition experiments, a 100-fold molar excess of unlabeled oligonucleotides was included in reaction mixtures. Gels were run in 0.5ϫ TBE buffer (0.045 M Tris borate, 0.001 M EDTA) at a constant voltage of 210 V, dried, and autoradiographed with intensifying screens. RNA Isolation and Analysis-Total RNA was isolated by using RNAzol B solution (Cinna/Biotecx, Friendswood, TX) according to the manufacturer's instructions. The RNA concentration was determined by measuring the absorbance at 260 nm. For quantitative RT-PCR, 1 g of total RNA was reverse-transcribed by using AMV reverse transcriptase (Roche Molecular Biochemicals) in reverse transcriptase reaction and amplified for 25 cycles (the number of cycles were optimized to ensure that quantitative assessment could be performed) with Taq DNA polymerase (Roche Molecular Biochemicals) per the manufacturer's instructions using primers specific to HGF (sense, 5Ј-ATCAGACACCAC-ACCGGCACAAAT-3Ј; antisense, 5Ј-GAAATAGGGCAATAATCCCAA-GGAA-3Ј) or ␤-actin (CLONTECH Laboratories). ␤-actin served as an RNA integrity and normalization control.
In Vitro Transcription/Translation-A transcription and translation-coupled reticulocyte lysate system was used to prepare in vitro translated COUP-TFI, mouse PPAR␥1, PPAR␥2, PPAR␣, and mRXR␣ proteins from their cognate expression vectors driven by the T7 promoter (27). Transcription/translation reactions were carried out in a 50-l reaction volume as recommended by the supplier (Promega). The authenticity of translation products was always confirmed in parallel experiments by including radioactive [ 35 S]methionine and radioactive [ 35 S]cysteine in the translation reaction followed by SDS-polyacrylamide gel electrophoresis and autoradiography. Translation reactions using empty vectors were always included as negative controls in these analyses, including the gel shift assays, to exclude potential false signals. Translation products were stored at Ϫ80°C. For each gel mobility shift assay, about 1 l was used from each translated reaction.

Identification of a Binding Site for PPAR␥ Transcription
Factor in the HGF Gene Promoter-Our recent studies of the HGF promoter revealed that the proximal promoter region at position Ϫ260 to Ϫ230 is a multifunctional composite site to which different transcription factors such as NF1, USF, and AP2 can bind and regulate HGF gene transcription (21,22). During those studies, we found that distinct factor(s) also specifically bind to this element that are not related to the NF1, USF, or AP2 families (21). Within this 30-bp region (Ϫ260 to Ϫ230) of the promoter, we identified RGGTCA hexanucleotide motifs, which are potential binding sites for the nuclear hormone receptor superfamily (23,28). To further characterize the composite binding site and its cognate transcription factors, we radiolabeled the HGF promoter composite element (Ϫ260 to Ϫ230) with 32 P and used this oligonucleotide as a probe to perform gel mobility band shift assays. In these experiments, nuclear protein extracts from 3T3-L1 and NIH3T3 fibroblast cell lines were used. As shown in Fig. 1, this DNA element forms several specific complexes, of which the major complex(es) (shown by a large arrow head) are efficiently abrogated by an excess amount of binding site for NF1 (compare lane 2 with lane 4 and lane 7 with lane 9, respectively). The remaining two complexes, which we have labeled in the figure as C1 and C2, formed by nuclear protein extract from NIH3T3 cells, are abolished by an excess amount of USF and nuclear hormone binding sites (RXRE), respectively ( Fig. 1, lanes 9 -11). We have previously shown by super shift assays that the major com-plex(es) C contain NF1 isoform(s) and that the C1 complex is composed of USF1 and USF2 isoforms (21). As shown in Fig. 1, 3T3-L1 fibroblasts do not have detectable C1 and only form the C2 complex, which is totally abrogated by RXRE (DR1, direct AGGTCA repeat with one spacing) (Fig. 1, lane 6). Thus, these results extend our previous observation and demonstrate that the complex C2 may contain nuclear hormone receptor(s), because it can be competed by a binding site having an AGGTCA motif (please see Ref. 21). To better define the binding site for complex C2, several mutant versions of the promoter element were synthesized (see Fig. 2E for nucleotide sequences) and used as competitors in electrophoretic mobility band shift assays using NF1-depleted 3T3-L1 nuclear protein extracts. As depicted in Fig. 2A, deletion of up to 11 base pairs from the 5Ј region did not have a notable effect on the binding activity (because these oligonucleotides still effectively competed with the radiolabeled wild-type probe; Fig. 2A, lanes 3 and 4, oligonucleotides designated S1 and S2). However, when six nucleotides from the 3Ј region were truncated, the resulting oligonucleotide totally lost its competitive ability ( Fig. 2A, lane 5). These results revealed that the binding site for C2 should be between Ϫ249 and Ϫ230. Close examination of the promoter element (Ϫ249 to Ϫ230) indicated that it harbors a potential RGGTCA inverted repetitive sequence separated by two nucleotides (IR2) (Fig. 2B, shown by arrows). The AGGTCA site is known as the perfect nuclear hormone binding half-site, although in almost every case, it is a variation of this sequence known as imperfect motif (28). Indeed, comparison of the nucleotide sequence of this 31-base pair element among the mouse, rat, and human HGF promoters revealed perfect conservation of this site (Fig. 2B). To define the binding site for the C2 complex in more detail, several mutated oligonucleotides were synthesized and used in gel shift competition assays, as above, using fibroblast nuclear extracts (Fig. 2, C and D).
Mutations in the AGGTCA half-site in the 3Ј region abolished the binding capability (mutant oligonucleotides named M4 and M5 in which AGGTCA is mutated to AGGGAC and CTTTCA, respectively) (Fig. 2C, lanes 6 and 7). Conservative mutations in the 5Ј-half-site RGGTCA motif (GGGCCA to AAGCCA or GGGAAA and also mutations in the two spacers (GG to AT) (oligonucleotides named M1, M2, and M3, respectively)) did not dramatically affect binding, because these oligonucleotides still competed with the labeled probe for binding (Fig. 2C, lanes  3-5). However, mutation in the 5Ј-half-site that totally changes the hexanucleotide from GGGCCA to TTTGGC completely abolished its binding activity (Fig. 2D, lane 8). To define the role of the two spacers in the IR2 configuration, we generated additional mutant oligonucleotides as indicated in Fig. 2E and used them as competitors. As shown in Fig. 2D, deleting of one or both spacer nucleotides (GG) dramatically reduced the binding activity (Fig. 2D, lanes 5 and 6). On the other hand, mutating the two putative spacers from GG to TT did not affect the binding activity (Fig. 2D, lane 7). These results implied that the RGGTCA half-sites and the two-spacer configuration are important for binding of complex C2 to this region. As mentioned above, it is well known that the RGGTCA is the binding halfsite for some members of the nuclear receptor superfamily (23,28). Therefore, several antibodies against the members of the nuclear receptor superfamily were used to identify complex C2 by supershift assays using fibroblast nuclear extract; these antibodies were against retinoic acid receptor, OR-1, thyroid hormone receptor, COUP-TF, vitamin D receptor, and nerve growth factor-induced B. None of these antibodies reacted with the C2 complex (data not shown). Further analysis using additional antibodies against other nuclear receptors revealed that anti-PPAR␥ antibody reacted with and totally supershifted complex C2 formed by the fibroblast nuclear extracts (Fig. 3, lanes 2 and 6). Neither anti-RXR␣ antibody nor an unrelated antibody (anti-Sp1; see Fig. 3, lanes 4 and 8) reacted with complex C2. Taken together, the data clearly show that PPAR␥ binds to the HGF promoter element located at Ϫ246 to Ϫ233 bp from the transcription start site.
Regulatory Function of PPAR␥ and Its Ligand 15-Deoxy-PGJ2 on the HGF Gene Promoter Activity-To demonstrate whether the PPAR␥ binding site and its binding transcription factor PPAR␥ have regulatory function on the HGF gene promoter, we performed cotransfection experiments with HGF-CAT promoter constructs and PPAR␥1 expression vector. Because the PPAR␥ binding site overlaps with those of NF1 and USF and because we have shown that NF1 strongly suppresses HGF gene promoter activity (21), we decided to eliminate the interference of NF1 with the PPAR␥ function. To do this, two different HGF-CAT constructs were generated and used: the wild-type 0.7 HGF-CAT construct and its mutated version, the 0.7 HGF-CAT/NF1-M construct, in which the NF1 site is mutated but the PPAR␥ binding site remains intact. Another HGF-CAT construct, 0.1 HGF-CAT, which does not have the NF1 and PPAR␥ binding sites, was also used as a negative control. We selected the 3T3-L1 cell line because the nuclear protein extract from this cell line did not form a USF complex (C1) (see Figs. 1 and 3) with this element that may interfere with the PPAR␥ function. As shown in Fig. 4, 15-deoxy-PGJ2, the PPAR␥ ligand, significantly induced the promoter activities of both the 0.7 HGF-CAT and the 0.7 HGF-CAT/NF1-M constructs but not the 0.1 HGF-CAT construct (Fig. 4). Overexpression of PPAR␥1 and treatment with its ligand, 15-deoxy-PGJ2, had an even stronger stimulatory effect on both the 0.7 HGF-CAT and the 0.7 HGF-CAT/NF1-M constructs (Fig. 4). It is believed that RXR␣ is the functional partner of PPAR␥ through heterodimer formation (28). Even though we did not Ϫ260 to Ϫ230 bp in the HGF gene promoter were labeled with 32 P and incubated with nuclear protein extract (NPE) from NIH3T3 and 3T3-L1 fibroblasts. The binding reactions were performed at room temperature for 20 min in the presence of the nonspecific competitor poly(dI-dC). Consensus NF1, USF, and RXRE (DR1), as well as unlabeled selfdouble-stranded oligonucleotides, were used as competitors. The complexes labeled C and indicated by a large arrow contain NF1 isoforms; the complex labeled C1 contains USF1 and USF2 (21); and the C2 complex contains a putative nuclear receptor; F denotes free 32 P-labeled oligonucleotide probe.

FIG. 2.
Characterization of the binding motif(s) for the nuclear receptor in the HGF promoter regulatory region. A, truncated (shortened from 5Ј or 3Ј ends as indicated in E) oligonucleotides S1, S2, and S3 corresponding to the promoter element were synthesized and used as competitors in gel mobility competition assays using NF1-depleted nuclear protein extract from 3T3-L1 fibroblasts and the wild-type HGF promoter element (Ϫ260 to Ϫ230) as a probe. B, analysis of the potential RGGTCA motifs within the regulatory region (from Ϫ260 to Ϫ230 bp) of the mouse HGF gene promoter. RGGTCA inverted repeats separated by two nucleotides (IR2) were identified as indicated by arrows. For comparison, the nucleotide sequences of the rat and human HGF promoter element are also presented. The functional binding sites for NF1, USF, AP2, and an E box (i.e. USF binding site) are also shown. C, mutated oligonucleotides M1-M5 (for sequence information, please see E) were synthesized and used as competitors as described under A. D, the S2 oligonucleotide was labeled and used as a probe with NIH3T3 nuclear extracts to further define the IR2 site. The nuclear extract used in this gel shift assay was not depleted of NF1. The DR1 site (direct AGGTCA motif with one spacing) was also used as competitor. E, the nucleotide sequences of the wild-type and mutant versions of the promoter element used to define the PPAR␥ binding site are shown. The mutated nucleotides are underlined. C2 denotes complex C2. Self-competitor (wild-type unlabeled probe) was used as a positive control. see an obvious supershift complex by anti-RXR␣ antibody using fibroblast nuclear extracts (see Fig. 3), it is possible that the nuclear protein extracts from NIH3T3 and 3T3-L1 cells do not contain a sufficient amount of RXR␣ binding activity. In fact, we show that the HGF PPAR␥ element binds to both PPAR␥ and RXR␣ using liver nuclear protein extract, which is presumably rich in RXR␣ (see below). To find out whether RXR␣ can regulate the HGF gene promoter through binding to this IR2 element as a partner of PPAR␥, RXR␣ expression vector and its ligand, 9-cis-RA, were used in cotransfection CAT assay experiments using a fibroblast cell line. As shown in Fig. 4, 9-cis-RA itself had a modest effect, but a combination of 9-cis-RA plus its receptor, RXR␣, strongly stimulated HGF promoter activity, similar to 15-deoxy-PGJ2 with its receptor, PPAR␥. The strongest induction of the promoter was achieved by cotransfection of PPAR␥ and RXR␣ expression vectors in the presence of 15-deoxy-PGJ2 (the PPAR␥ ligand) and 9-cis-RA (the RXR␣ ligand) (Fig. 4). Because PPAR␥/RXR␣ could induce the promoter activity of the 0.7 HGF-CAT/NF1-M construct, these results indicate that PPAR␥/RXR␣ may not only indirectly (via competing with NF1 binding), but also directly, stimulate the HGF promoter through binding to a functional RGGTCA motif in the HGF proximal promoter.
Binding of the Nuclear Hormone Element with RXR␣, PPAR␥, and COUP-TF from Liver Tissue and Related Cells-It is well known that the HGF gene is expressed in mesenchymal, but not epithelial, cells of a given tissue. In the liver, HGF is expressed in the non-parenchymal mesenchymal cells (especially stellate cells, also known as the Ito cells or lipocytes because they store lipids and vitamin A). We tested whether the HGF promoter element could bind to PPAR␥ and possibly other transcription factors present in non-parenchymal liver cells. Nuclear protein extracts from whole liver and cells isolated from liver, such as freshly isolated non-parenchymal cells, as well as an Ito cell line (25,26), were prepared. We then subjected these extracts to gel mobility shift assay using the HGF promoter composite site as a probe. The probe bound to PPAR␥ present in all three different nuclear protein extracts, as shown by supershift assays (Fig. 5, lanes 2 and 7, and data not shown for whole liver). A surprising finding in these experiments was that this element could also strongly bind to COUP-TF, a member of the orphan nuclear receptor subfamily, which also binds to the RGGTCA motif (see Fig. 5, lanes 3 and 8) (28). The fact that anti-COUP-TF totally supershifted the C2 complex whereas anti-PPAR␥ partially did so (Fig. 5, compare  lanes 2 and 3) suggests that PPAR␥ and COUP-TF may form heterodimers on the HGF promoter element. Of course, it is known that COUP-TF also forms homodimers; thus, it is likely that the C2 complex in the nuclear extract of liver cells is mainly composed of COUP-TF homodimer as well as COUP-TF and PPAR␥ heterodimers. On the other hand, the C2 complex produced by fibroblast nuclear extracts lacks any COUP-TF, because it did not react with anti-COUP-TF (data not shown; also see Fig. 3). Anti-RXR␣ antibody was also used in these experiments, because RXR␣ is believed to be the partner of PPAR␥, as mentioned above. We did detect weak RXR␣ binding Wild-type HGF-CAT reporter construct (0.7 HGF-CAT) and the mutated HGF-CAT reporter plasmid (0.7 HGF-CAT/NF1-M), which harbors a mutation in the NF1 binding site, were used to determine the function of PPAR␥ on the activity of the HGF promoter. 3T3-L1 cells were transiently transfected with the HGF-CAT constructs described above. For cotransfection, 5 g of RXR␣ and PPAR␥1 expression vectors per each reaction were added. RXR␣ ligand, 9-cis-RA, and PPAR␥ ligand, 15-deoxy PGJ2, were also added to activate their cognate receptors. The 0.1 HGF-CAT construct was used as a negative control. The data are presented as relative CAT activity (percent conversion) and are from three independent experiments performed in duplicate. Error bars represent the standard deviation.
FIG. 5. The IR2 element in the HGF promoter composite site binds to the nuclear receptors PPAR␥ and COUP-TF present in the nuclear protein extracts from liver cells. Supershift assays were carried out using various antibodies as indicated and nuclear protein extracts from an Ito cell line derived from rat liver and from freshly isolated non-parenchymal rat liver cells (NPC). In this experiment, the HGF promoter composite element was used as a probe. Unlabeled probe (cold self) was used as a competitor to confirm specificity. S denotes the supershifted complex.
to this element with the nuclear protein extracts from whole liver tissue (the shifted band was very faint and required a very long exposure of the film) (data not shown). Taken together, the results imply that the PPAR␥ binding site in the HGF gene promoter also binds to COUP-TF and RXR␣, depending on the cell types/tissues analyzed (i.e. fibroblasts versus liver cells).
To confirm that COUP-TF truly binds to the HGF proximal promoter composite element, in vitro translated COUP-TF transcription factor was prepared and used in gel mobility shift assays. Indeed, in vitro translated COUP-TF could specifically and avidly bind to this region, as shown in Fig. 6, lane 1. Unlabeled excess probe totally competed for this complex (Fig.   6, lane 2). Moreover, anti-COUP-TF antibody reacted with and shifted this complex (Fig. 6, lane 4). On the other hand, anti-PPAR␥ antibody did not react with this complex (Fig. 6, lane 3). These results prove that COUP-TF binds specifically to this region. To define the binding motif(s) for COUP-TF in this region, gel shift competition assays were performed using the truncated oligonucleotides S1-S3 and the mutated oligonucleotides M1-M5 (see Fig. 2E for sequences). S1 oligonucleotide retained all of its binding activity. S2 oligonucleotide retained some of its binding ability, but the S3 oligonucleotide totally lost its binding ability for COUP-TF (data not shown). Similarly, the competitive ability of M1, M2, and M3 mutated oligonucleotides reduced only modestly. However, the binding ability for COUP-TF was totally abrogated when the AGGTCA motif in the 3Ј end of the element was mutated (oligonucleotides M4 and M5; data not shown). Combining all of the above results, we conclude that COUP-TF also binds to the composite site in this regulatory region of the HGF gene promoter (see FIG. 6. Characterization of the COUP-TF binding motif in the HGF promoter composite element. In vitro translated COUP-TFI protein was used in these experiments. Supershift assays were performed with antibody against COUP-TF and the wild-type HGF promoter element as a probe. Self-competitor was used to show the specific binding of COUP-TFI with this element. Antibody against PPAR␥ was used as a negative control. The COUP-TF1 complex is indicated by an arrow. S denotes the supershift complex. Negative controls for the in vitro translation product (empty vector control subjected to the in vitro translation reaction) and gel shift assays were also carried out to also ensure specificity (data not shown). The left lane is the molecular weight marker X174/HaeIII. B, MRC-5 human fibroblasts were cultured in 6-well plates and treated with or without (control) 15-deoxy PGJ2 in duplicates. Culture medium was collected at the indicated time points and then subjected to a sensitive sandwich enzyme-linked immunosorbent assay using the R&D HGF enzyme-linked immunosorbent assay kit. The amounts of HGF were determined in the culture medium from the standard curve using pure HGF as a standard as recommended by the supplier. Asterisks indicate statistically significant differences between the control and treated samples (p Ͻ 0.05). Fig. 9). It should be added that the specific binding of the PPAR␥ isoforms (PPAR␥1 and PPAR␥2) to the HGF promoter element was also confirmed in a series of gel mobility band shift and super shift assays using in vitro translated mouse PPAR␥1, PPAR␥2, and RXR␣ (data not shown).
Modulation of PPAR␥ Function on the HGF Gene Promoter by COUP-TF, NF1, and AP2-Because our data showed that COUP-TF binds to the same binding site as that for PPAR␥, it is reasonable to suppose that COUP-TF homodimers could modify the stimulatory function of PPAR␥/RXR␣ by competing with one another for their common binding site. Additionally, published data demonstrate that COUP-TF can form heterodimers with RXR␣ in solution in the absence of their binding site. In this manner, COUP-TF sequesters RXR␣, preventing PPAR␥/RXR␣ heterodimer formation and the stimulatory function of PPAR␥/RXR␣. Furthermore, the binding sites for COUP-TF and PPAR␥/RXR␣ overlap with the binding sites for the suppressive transcription factors NF1 and AP2. Therefore, the functionality of this unique composite element to regulate HGF gene expression in a given tissue or cell may be dependent on the specific combination, and/or the various concentrations, of transcription factors. To show the individual functions of these transcription factors and how they interact to exert their effects, we used the 0.3 HGF-CAT promoter construct, which has a basal promoter and the proximal region of the HGF promoter having the composite site containing the NF1, AP2, and IR2 binding motifs. We also used the 0.1 HGF-CAT construct as a negative control, because it only contains the basal promoter. As shown in Fig. 7, COUP-TF dose-dependently repressed the stimulatory function of PPAR␥/RXR␣ on the HGF gene promoter in the 0.3 HGF-CAT construct. It did not have significant repressive function on the 0.1 HGF-CAT construct (Fig. 7). In similar experiments using the expression vectors for the NF1/X isoform and AP2 co-transfected with HGF-CAT constructs we found that they also inhibited the stimulatory function of PPAR␥/RXR␣ on the HGF gene promoter in the 0.3 HGF-CAT construct but not in the 0.1 HGF-CAT construct (data not shown). In these experiments, empty vector controls did not have any effect on the activity of either HGF-CAT construct (data not shown). These experiments indicate that NF1, AP2, and COUP-TFI could individually down-regulate the stimulatory function of PPAR␥/ RXR␣ on HGF gene expression and that they may do this through competing with PPAR␥/RXR␣ for the overlapping composite binding site in the HGF proximal promoter region.

Induction of Endogenous HGF mRNA and Protein Expression by PPAR␥ Ligand, 15-Deoxy-PGJ2-
To determine whether 15deoxy PGJ2 can induce endogenous HGF gene expression at the mRNA and protein levels, we treated fibroblasts, which naturally express PPAR␥, with 15-deoxy PGJ2 and determined the level of HGF mRNA by semi-quantitative RT-PCR and the level of HGF protein by a very sensitive sandwich enzyme-linked immunosorbent assay. The results demonstrated that HGF expression is significantly up-regulated after treatment with 15-deoxy-PGJ2 (Fig. 8, A and B, respectively). DISCUSSION In this study, we demonstrated that the nuclear receptors PPAR␥ and COUP-TF bind to a composite regulatory element in the HGF upstream promoter region located at Ϫ246 to Ϫ233 bp from the transcription start site. This binding site also overlaps/harbors functional binding sites for NF1, AP2, and the USF family of transcription factors (21,22), and together, they constitute a multifunctional composite site to which these various transcription factors can bind and regulate the activity of the HGF promoter. Cotransfection assays showed that PPAR␥1, with its partner, RXR␣, in the presence of their corresponding ligands, 15-deoxy PGJ2 and 9-cis-RA, respectively, strongly stimulated HGF promoter activity. NF1, AP2, and COUP-TFI suppressed the stimulatory function of PPAR␥ by competing for their individual binding sites. Moreover, we demonstrated that 15-deoxy-PGJ2, the ligand of PPAR␥, induces the expression of the endogenous HGF gene at the mRNA and protein levels in cultured fibroblasts.
PPAR␥ is a member of the nuclear receptor superfamily that includes receptors for the steroid, thyroid, and retinoid hormones (23,28). It is known that there are three related but quite distinct PPAR proteins: PPAR␣, PPAR␦, and PPAR␥. Two forms of PPAR␥, ␥1 and ␥2, exist as products of alternative promoter usage (29,30). Like other members of this superfamily, PPAR␥ contains a central DNA binding domain that binds to a cis-acting element in the promoter of its target genes. Most of the PPAR␥ response elements described are composed of a directly repeating core site separated by one nucleotide (RG-GTCA-N-RGGTCA) called DR1 (24,28). Interestingly, the PPAR␥ response element in the HGF promoter region is an inverted repeating core site separated by two nucleotides (GGGCCA-NN-TGACCT) that we have called IR2. Published data indicate that PPAR␥ heterodimerizes with RXR␣ and that the PAR␥/RXR␣ heterodimer binds to its response element and activates its target genes (23,31). Our results indicated that the IR2 site in the HGF gene promoter also binds to RXR␣, though very weakly. Generally, RXR␣ is said to be a silent partner. Nonetheless, there are several examples in which RXR␣ can be an active partner (31). Similarly, both RXR␣ and PPAR␥ in the RXR␣/PPAR␥ heterodimers bound to the IR2 element in HGF gene promoter region are independently responsive and are synergistically activated in the presence of both ligands, 15-deoxy PGJ2 and 9-cis-RA.
PPAR␥ plays an essential role in cell growth and differentiation as well as in oncogenesis. Controversies exist on whether PPAR␥ acts as a tumor promoter or tumor suppressor. For example, ligand activation of PPAR␥ causes most, but not all, colon cancer cell lines to undergo a differentiating response and reverse their malignant phenotype (32). Paradoxically, ligand activation of PPAR␥ in min mice, an animal model for familial adenomatous polyposis, seemed to cause increased polyposis (33,34). It is demonstrated that loss-of-function mutations in PPAR␥ are also associated with human colon cancer (35). It is not clear whether this association between colon cancer and PPAR␥ is involved in HGF expression in colon cancer tissue, because HGF expression is also aberrant in colon cancer tissues. Additionally, HGF can also function paradoxically as a tumor suppressor and tumor promoter in different settings (2,3). Little is known about the target genes activated by PPAR␥; our present findings shed significant light on the matter in the sense that HGF may mediate some of the functions of PPAR␥ in tissue growth and differentiation and in cancer.
Like HGF, PPAR␥ is expressed in various tissues, but predominantly in the mesenchymal cells, namely adipocytes/fibroblasts (36). Interestingly, in the liver, HGF is expressed in a specialized fibroblastic cell type called Ito cells, which are also known as lipocytes (because they store lipids and vitamin A). Like HGF, PPAR␥ is also expressed in the placenta, and its deficiency leads to placental dysfunction and death by E10 (37). A qualitatively similar placental abnormality has been seen in RXR␣ knockout mice, suggesting that these phenotypes may be due to impaired PPAR␥-mediated signaling in the placenta (37). Interestingly, HGF knockout mice revealed that homozygous HGF mutant embryos have severely impaired placentas with markedly reduced numbers of labyrinthine trophoblast cells, and the embryos die before birth by E15 (6,7). Combining these data, it seems possible that HGF may be one of the mediators of the growth regulatory action of PPAR␥/RXR␣. So far, we have identified several functional regulatory elements and their binding transcription factors that act on the HGF gene promoter (Fig. 9). Based on the current findings and previous published results, we believe that transcriptional regulation of the HGF gene is very complex and is dependent on the combination of all of these transcription factors, including COUP-TF, CCAAT/enhancer-binding protein, Sp1, estrogen receptor, NF1, AP2, USF, and PPAR␥ (15)(16)(17)(18)(19)(20)(21)(22). CCAAT/enhancerbinding protein ␤, SP1, estrogen receptor, USF, and PPAR␥ are positive regulators. COUP-TF, AP2, and NF1 are mainly negative regulators. COUP-TFs can compete with both estrogen receptor and PPAR␥ for their respective binding sites and suppress the stimulatory functions of both estrogen receptor and PPAR␥ in the presence of their own ligands (estrogen and 15-deoxy PGJ2, respectively). This result implies that COUP-TFs and the NF1 family play an important role in regulation of HGF gene expression. As mentioned above, PPAR␥ and CCAAT/enhancer-binding protein ␤ are believed to play an important role in adipocyte differentiation (38). Interestingly, we noted that in the 3T3-L1 fibroblasts that differentiate into adipocytes when placed in differentiation medium, a marked increase in PPAR␥ and CCAAT/ enhancer-binding protein ␤ binding to the HGF promoter elements occurred and is depicted in Fig. 9. The increase in the binding activity correlated well with an increase in HGF and HGF receptor (Met) expression in these cells, implying that the HGF/Met signaling system may have a role in adipocyte differentiation. A summary of the positive and negative regulators of the HGF gene promoter that we have identified and characterized thus far is presented in Fig. 9.