The Organization, Promoter Analysis, and Expression of the Human PPARγ Gene*

PPARγ is a member of the PPAR subfamily of nuclear receptors. In this work, the structure of the human PPARγ cDNA and gene was determined, and its promoters and tissue-specific expression were functionally characterized. Similar to the mouse, two PPAR isoforms, PPARγ1 and PPARγ2, were detected in man. The relative expression of human PPARγ was studied by a newly developed and sensitive reverse transcriptase-competitive polymerase chain reaction method, which allowed us to distinguish between PPARγ1 and γ2 mRNA. In all tissues analyzed, PPARγ2 was much less abundant than PPARγ1. Adipose tissue and large intestine have the highest levels of PPARγ mRNA; kidney, liver, and small intestine have intermediate levels; whereas PPARγ is barely detectable in muscle. This high level expression of PPARγ in colon warrants further study in view of the well established role of fatty acid and arachidonic acid derivatives in colonic disease. Similarly as mouse PPARγs, the human PPARγs are activated by thiazolidinediones and prostaglandin J and bind with high affinity to a PPRE. The human PPARγ gene has nine exons and extends over more than 100 kilobases of genomic DNA. Alternate transcription start sites and alternate splicing generate the PPARγ1 and PPARγ2 mRNAs, which differ at their 5′-ends. PPARγ1 is encoded by eight exons, and PPARγ2 is encoded by seven exons. The 5′-untranslated sequence of PPARγ1 is comprised of exons A1 and A2, whereas that of PPARγ2 plus the additional PPARγ2-specific N-terminal amino acids are encoded by exon B, located between exons A2 and A1. The remaining six exons, termed 1 to 6, are common to the PPARγ1 and γ2. Knowledge of the gene structure will allow screening for PPARγ mutations in humans with metabolic disorders, whereas knowledge of its expression pattern and factors regulating its expression could be of major importance in understanding its biology.

central role in lipid homeostasis and the maintenance of energy balance in vertebrates. These cells store energy in the form of triglycerides during periods of nutritional affluence and release it in the form of free fatty acids at times of nutritional deprivation. Excess of white adipose tissue leads to obesity (1)(2)(3), whereas its absence is associated with lipodystrophic syndromes (4). In contrast to the development of brown adipose tissue, which mainly takes place before birth, the development of white adipose tissue is the result of a continuous differentiation/development process throughout life (2,5). During development, cells that are pluripotent become increasingly restricted to specific differentiation pathways. Adipocyte differentiation results from coordinate changes in the expression of several proteins, which are mostly involved in lipid storage and metabolism, that give rise to the characteristic adipocyte phenotype. The changes in expression of these specialized proteins are mainly the result of alterations in the transcription rates of their genes.
Several transcription factors including the nuclear receptor PPAR␥ (6,7), the family of CCAATT enhancer binding proteins (C/EBP) 1 (8 -13) and the basic helix-loop-helix leucine zipper transcription factor ADD1/SREBP1 (14,15) orchestrate the adipocyte differentiation process (for reviews, see Refs. 1, 3, 16 -18). In contrast to the wide tissue distribution of the various C/EBPs, PPAR␥ has been shown to have an adipose-restricted pattern of expression in mouse. The currently favored hypothesis is that C/EBP␤ and ␦ induce the expression of PPAR␥ (11), which then triggers the adipogenic program. Terminal differentiation then requires the concerted action of both PPAR␥, C/EBP␣, and ADD-1/SREBP1 (7,15). Several arguments support the important role of PPAR␥ in adipocyte differentiation. First, overexpression of PPAR␥ by itself can induce adipocyte conversion of fibroblasts (6). In addition, PPAR␥ together with C/EBP␣ can induce transdifferentiation of myoblasts into adipocytes (19). Second, the description of functional PPREs in the regulatory sequences of several of the genes that are induced during adipocyte differentiation, such as the genes coding for adipocyte fatty acid binding protein, aP2 (6), phosphoenolpyruvate carboxykinase (PEPCK) (20), acyl-CoA synthetase (ACS) (21,22), and lipoprotein lipase (LPL) (23), is consistent with the crucial role attributed to PPAR␥ in adipocyte differentiation. Finally, PPAR activators, such as fibrates (24,25) and fatty acids (7, 26 -28), or synthetic PPAR␥ ligands, such as the thiazolidinediones (7,28,29), induce adipocyte differentiation. In this context, it is interesting to note that prostanoids, which are potent inducers of adipose differentiation programs (30 -32), may be one of the natural ligands of PPAR␥. In addition to PPAR␥, PPAR␣, but not PPAR␦, has been shown to have some, albeit weaker, adipogenic activity (33).
To better understand the physiological role of PPAR␥ in human physiology, it is crucial that we gain insight into the regulation of PPAR␥ gene expression in man. Therefore, we cloned the human PPAR␥ cDNAs, determined the structure of the human PPAR␥ gene, and studied the expression of the PPAR␥ mRNAs and the regulation of their promoter. Both PPAR␥1 and 2 are produced in human tissues but PPAR␥2 appears to be the minor isoform in man. In addition to adipose tissue, which contains high levels of PPAR␥, we demonstrate high level expression of human PPAR␥ in the colon. The structure of the gene encoding the mouse and human PPAR␥s is highly conserved. Furthermore our results demonstrate that 3 and 1 kb of DNA upstream of the transcription start sites of PPAR␥1 and ␥2, respectively, are sufficient to control basal and tissue-specific PPAR␥ gene expression.

Materials and Oligonucleotides
The oligonucleotides used for various experiments in this manuscript are listed in Table I.

Isolation of the Human PPAR␥ cDNA and Gene, Restriction
Mapping, Determination of Intron/Exon Boundaries, and DNA Sequencing A human adipose tissue gt11 library was screened with a random primed 32 P-labeled 200 bp fragment, covering the DNA-binding domain of the mouse PPAR␥ cDNA. After hybridization, filters were washed in 2 ϫ SSC, 0.1% SDS for 10 min at 20°C and twice for 30 min in 1 ϫ SSC, 0.1% SDS at 50°C and subsequently exposed to x-ray film (X-OMAT-AR, Kodak). Of several positive clones, one clone 407 was characterized in detail. The insert of this clone, starting Ϯ90 bp upstream of the ATG start codon and extending downstream into the 3Ј-untranslated region (UTR) sequence, was subcloned in the EcoRI site of pBluescript SK Ϫ to generate clone 407.2. Sequence analysis of 407.2 confirmed it as being the human homologue of the mouse PPAR␥2 cDNA. While this work was in progress, other groups also reported the isolation of human PPAR␥2 cDNA clones (34,35).
To isolate genomic P1-derived artificial chromosome (PAC) clones containing the entire human PPAR␥ gene, the primer pair LF-3 and LF-14 was used to amplify an 86-bp probe with human genomic DNA as template. This fragment was then used to screen a PAC human genomic library from human foreskin fibroblasts. Three positive clones, P-8854, P-8855, and P-8856, were isolated. Restriction digestion and Southern blotting were performed according to classical protocols as described by Sambrook et al. (36). Sequencing reactions were performed, according to the manufacturer instructions, using the T7 sequencing kit (Pharmacia Biotech Inc.).

Determination of the Transcription Initiation Site: Primer Extension and 5Ј-Rapid Amplification of cDNA Ends (5Ј-RACE)
Primer Extension-The oligonucleotide LF-35 was 32 P-labeled with T4-polynucleotide kinase (Amersham Life Science, Inc) to a specific activity of 10 7 dpm/50 ng and purified by gel electrophoresis. For primer extension, 10 5 dpm of oligonucleotide was added in a final volume of 100 l to 50 g of adipose tissue total RNA isolated from different patients. Primer extension analysis was performed following standard protocols utilizing a mixture of 1.25 units of avian mycloblastosis virus reverse transcriptase (Life Technologies, Inc.) and 100 units of Moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc.). A sequencing reaction and molecular mass standards were used to map the 5Ј-end of the extension products.
5Ј-RACE-The Marathon cDNA amplification kit (CLONTECH) was used to obtain a library of adaptor-ligated double-stranded cDNA from human adipose tissue. 1 g of poly(A) ϩ RNA was used as a template for the first strand synthesis, with the 52-mer CDS primer and 100 units of the MMLV reverse transcriptase in a total volume of 10 l. Synthesis was carried out at 42°C for 1 h. Next, the second strand was synthetized at 16°C for 90 min in a total volume of 80 l containing the enzyme mixture (RNase H, Escherichia coli DNA polymerase I, and E. coli DNA ligase), the second strand buffer, the dNTP mix, and the first strand reaction. cDNA ends were then made blunt by adding to the reaction 10 units of T4 DNA polymerase and incubating at 16°C for 45 min. The double-stranded cDNA was phenol/chloroform extracted, ethanol precipitated, and resuspended in 10 l of water. Half of this volume was used to ligate the adaptor to the cDNA ends (adaptor sequence CTAATACGACTCACTATAGGGCTCGAGCGGCCGCCCGG-GCAGGT) in a total volume of 10 l using 1 unit of T4 DNA ligase. The ligation reaction was incubated 16 h at 16°C. The resulting cDNA library was diluted to a final concentration of 0.1 mg/ml.
The 5Ј-end of PPAR␥1 was PCR-amplified using 5 l of the library as a template with the oligonucleotides AP-1 (binding to the adaptor) and LF-45 (binding antisense to the 5Ј-end of the PPAR␥1). After an initial denaturing step at 95°C for 3 min, 25 cycles were done at the following conditions: 10 s at 95°C, 20 s at 60°C, and 30 s at 72°C. The resulting PCR product was reamplified for 30 additional cycles at the same conditions using the nested oligonucleotides AP2 (nested to AP1) and LF-2 (nested to LF-45). The PCR product was analyzed on a 2% agarose gel, treated with Pfu polymerase (Stratagene) and cloned into the EcoRV site of pBluescript SK ϩ . A total of 20 white colonies were grown and sequenced from both ends using the oligonucleotides T3 and T7 (Dye Terminator Cycle sequencing kit, Applied Biosystems).
For the determination of the 5Ј-end of PPAR␥2, the same procedure was followed except that the oligonucleotide LF-14 (specific for the PPAR␥2 5Ј-UTR) was used in the first round PCR, and the oligonucleotide LF-35 (nested to LF-14) was used in the second round PCR with the same cycling conditions.

Tissue Biopsies and Cell Culture
Omental adipose tissue, small and large intestine, kidney, muscle, and liver biopsies were obtained from non-obese adult subjects undergoing elective surgery or endoscopy. All subjects had fasted overnight before surgery (between 8.00 p.m. and 10 a.m.) and received intravenous saline infusion. They had given informed consent, and the project was approved by the ethics committee of the University of Lille. All tissue was immediately frozen in liquid nitrogen until RNA preparation.
Standard cell culture conditions were used to maintain 3T3-L1 (obtained from ATCC), CV-1 (a kind gift from Dr. R. Evans, Salk Institute, La Jolla, CA), and Hep G2 cells (ATCC). BRL-49,653, supplied by Ligand Pharmaceuticals, San Diego, CA (in DMSO) and fatty acids (in ethanol) were added to the medium at the concentrations and times indicated. Control cells received vehicle only. Fatty acids were com-

mRNA Analysis by RT-Competitive PCR Assay
RNA preparation of total cellular RNA was performed as described previously (37). The absolute mRNA concentration of the differentially spliced PPAR␥ variants was measured by reverse transcription reaction followed by competitive polymerase chain reaction (RT-competitive PCR) in the presence of known amounts of competitor DNA yielding amplicons of different size allowing the separation and the quantification of the PCR products. The competitor was constructed by deletion of a 74-bp fragment (nucleotides ϩ433 to ϩ507 by HindIII digestion) of PPAR␥1 cloned into pBluescript KS ϩ , yielding pBSCompPPAR␥. Working solution of the competitor was prepared by in vitro transcription followed by serial dilution in 10 mM Tris-HCl (pH 8.3), 1 mM EDTA buffer. For RT-competitive PCR, the antisense primer hybridized to the 3Ј-end of exon 3 (␥AS:5Ј-GCATTATGAGCATCCCCAC-3Ј, nt ϩ600 to ϩ620) and the sense primer to exon 1 (␥S:5Ј-TCTCTCCGTAATGGAA-GACC-3Ј, nt ϩ146 to ϩ165) or to the B exon (␥2S:5Ј-GCGATTCCT-TCACTGATAC-3Ј, nt ϩ41 to ϩ59). Therefore, the same competitor served to measure either total PPAR␥ mRNAs (␥1 ϩ ␥2; with primers ␥AS and ␥S) or, specifically, PPAR␥2 mRNA (with primers ␥AS and ␥2S). The ␥AS/␥S primer pair gave PCR products of 474 and 400 bp for the PPAR␥ mRNAs and competitor, respectively. The primer pair ␥AS/ ␥2S gave 580 bp for PPAR␥2 mRNA and 506 bp for the competitor. For analysis of the PCR products, the sense primers ␥S and ␥2S were 5Ј-end labeled with the fluorescent dye Cy-5 (Eurogentec, Belgium).
First-strand cDNA synthesis was performed from total RNA (0.1 g) in the presence of the antisense primer ␥AS (15 pmol) and of thermostable reverse transcriptase (2.5 units; Tth DNA polymerase, Promega) as described (38). After the reaction, half of the RT volume was added to the PCR mix (90 l) containing the primer pair ␥AS/␥S for the assay of PPAR␥ total mRNA, whereas the other half was added to a PCR mix (10 mM Tris-HCl, pH 8.3, 100 mM KCl, 0.75 M EGTA, 5% glycerol, 0.2 mM dNTP, 5 units of Taq polymerase) containing the primer pair ␥AS/␥2S for the assay of PPAR␥2 mRNA. Four aliquots (20 l) of the mixture were then transferred to microtubes containing a different, but known, amount of competitor. After 120 s at 95°C, the samples were subjected to 40 PCR cycles (40 s at 95°C, 50 s at 55°C, and 50 s at 72°C). The fluorescent-labeled PCR products were analyzed by 4% denaturing polyacrylamide gel electrophoresis using an automated laser fluorescence DNA sequencer (ALFexpress, Pharmacia, Uppsala, Sweden), and integration of the area under the curve using the Fragment manager software (Pharmacia) was performed as described (38).
To validate this technique, human PPAR␥2 mRNA was synthesized by in vitro transcription from the expression vector pSG5hPPAR␥ (Riboprobe system, Promega) and quantified by competitive PCR over a wide range of concentrations (0.25-25 attomole (amol) added in the RT reaction). Standard curves obtained when assaying PPAR␥ total mRNA or PPAR␥2 mRNA are shown in Fig. 2C. The linearity (r ϭ 0.99) and the slopes of the standard curves (0.98 and 1.11) indicated that the RTcompetitive PCR is quantitative and that all the mRNA molecules are copied into cDNA during the RT step. The lower limit of the assay was about 0.05 amol of mRNA in the RT reaction, and the interassay variation of the RT-competitive PCR was 7% with six separated determinations of the same amount of PPAR␥ mRNA.

Analysis of Promoter Activity
To test the activity of the human PPAR␥ promoters several reporter constructs were made. A 1-kb fragment of PAC clone 8856 was isolated by PCR using the oligonucleotides LF-35 (binding antisense in the PPAR␥2 5Ј-UTR) and the oligonucleotide LF-58 (binding sense at position -1000 of the PPAR␥2), was sequenced, and was inserted into EcoRV site of pBluescript (Stratagene, La Jolla, CA). After digestion of plasmid pBS␥2p1000 with SmaI and KpnI, the insert was cloned into the reporter vector pGL3 (Promega), creating the expression vector pGL3␥2p1000. To isolate the PPAR␥1 promoter, an 8-kb EcoRI fragment, which hybridized with the oligonucleotide LF-2 (corresponding to  35 S-labeled proteins. B, validation of a PPAR␥ antibody. mPPAR␣, mPPAR␥, haPPAR␥, and hPPAR␥ were synthesised in rabbit reticulocyte lysate and electrophoresed on a 10% SDS-PAGE gel. The proteins were transferred to nitrocellulose and developed with a rabbit anti-PPAR␥ antibody. As indicated, the antibody specifically recognized PPAR␥ but did not recognize PPAR␣. C, Western blot of three different human adipose tissue samples. Two different protein extracts of the same adipose tissue sample were run side-by-side. WAT, white adipose tissue. as well as the different-sized amplicons obtained are indicated. B, typical analysis of the fluorescence-labeled PCR products on an automated fluorescence DNA sequencer using a denaturing 4% polyacrylamide gel electrophoresis. C, validation of the RT-competitive PCR assay and standard curves obtained when assaying PPAR␥ total mRNA or PPAR␥2 mRNA. The linearity (r ϭ 0.99) and the slopes of the standard curves (0.98 and 1.11) indicated that the RT-competitive PCR is really quantitative and that all the mRNA molecules are copied into cDNA during the RT step.
the 5Ј-UTR of ␥1), was cloned into pBluescript. Partial mapping and sequencing of this clone revealed the presence of a 3-kb fragment upstream of the transcription initiation site. To test for promoter activity, a SacI/XhoI digestion of this clone containing the 3-kb promoter was inserted in the same sites of pGL3, resulting in the final vector pGL3␥1p3000. The pSG5-haPPAR␥ (39) and pMSV-C/EBP␣ (10) expression vectors were described elsewhere. Transfections were carried out in 60-mm plates using standard calcium phosphate precipitation techniques (for 3T3-L1, CV-1, and COS cells) (22). Luciferase and ␤-galactosidase assays were carried out exactly as described previously (22).

RESULTS
Cloning of the human PPAR␥ cDNA-A cDNA probe containing a 200-bp (KpnI-BglII) fragment encoding the DNA binding domain of the mouse PPAR␥ (44) was used to screen a human adipose tissue cDNA library. Several independent human PPAR␥ cDNA clones, representing both the PPAR␥1 and PPAR␥2 subtypes, were isolated and sequenced (Fig. 1A). The human PPAR␥ protein shows a 99% similarity and a 95% identity on the amino acid level with mouse PPAR␥ (Fig. 1A). Interestingly, the initiation codon for human PPAR␥1 is different from the mouse PPAR␥1 (Fig. 1B). Therefore, human PPAR␥1 is 2 amino acid residues longer than its mouse homologue.
Expression of PPAR␥ mRNA and Protein-To analyze the expression pattern of the two PPAR␥ isoforms, we developed a sensitive RT-competitive PCR assay in which relative amounts of PPAR␥1 and ␥2 mRNA could be measured from minute quantities of RNA (0.1 g total RNA). This method relies on the co-amplification in the same tube of known amounts of competitor DNA ( Fig. 2A) with PPAR␥ cDNA, obtained after reverse transcription from total tissue RNA. The competitor and the target use the same fluorescently labeled PCR primers but yield amplicons with a different size (Fig. 2, A and B), allowing their separation and quantification on an automated sequencing gel at the end of the reaction (Fig. 2C). All tissue preparations were carefully dissected, and the RNA was shown to be free of contamination with adipose tissue as evidenced by the absence of human leptin mRNA by RT-competitive PCR assay (38) (data not shown). PPAR␥1 mRNA was the predominant PPAR␥ isoform in all human tissues analyzed (Fig. 3). PPAR␥2 was detected in both liver and adipose tissue where it accounted for 15% of all PPAR␥ mRNA. Interestingly, in addition to the high level of expression of PPAR␥ mRNA expected in adipose tissue, we found a very high level of PPAR␥1 in large intestine. In contrast to adipose tissue, large intestine contained no PPAR␥2 mRNA. Kidney, liver, and small intestine contained intermediate levels of PPAR␥ mRNA, whereas PPAR␥ mRNA was barely detectable in skeletal muscle (Fig. 3).
Next, the expression of the human PPAR␥ protein was analyzed in human adipose tissue. A PPAR␥ specific antibody, raised against a peptide corresponding to amino acids 20 -104 of mPPAR␥, was used. This antibody is highly specific for PPAR␥ and does not cross-react with PPAR␣ and ␦ in Western blot experiments (Fig. 4, A and B). Using this antibody in a Western blot of protein extracts from human adipose tissue, we detected a band (potentially representing a doublet) with an approximate molecular mass of 60 kDa, consistent with the predicted mass of PPAR␥1 and 2 and with the protein product generated by in vitro transcription/translation in the presence of [ 35 S]methionine (Fig. 4, B and C).
PPAR␥2 Binds and Transactivates through a PPRE-To an- alyze whether PPAR␥ could bind to a PPRE, classically composed of direct repeats spaced by one intervening nucleotide (DR-1), EMSA was performed using in vitro transcribed/translated PPAR␥2 protein. An oligonucleotide containing a high affinity PPRE, previously identified in the apoA-II promoter J site, was used in EMSA (29). This oligonucleotide was capable of binding both human and hamster PPAR␥/mRXR␣ heterodimers in EMSA (Fig. 5, lanes 5 and 6). Homodimers of either hPPAR␥ or mRXR␣, however, were incapable of binding to this oligonucleotide. When increasing concentrations of unlabeled apoA-II J site were added as competitor, binding of the hPPAR␥/mRXR␣ heterodimer to the labeled PPRE was almost completely inhibited (Fig. 5, lanes 7-9). In addition, oligonucleotides corresponding to the PPRE elements of the ACO or LPL genes competed, albeit less efficiently (Fig. 5, compare  lanes 7 with 10 and 13).
We next verified that the human PPAR␥2 cDNA was capable of activating gene transcription through a PPRE. Therefore, 3T3-L1 preadipocytes were cotransfected with the PPAR␥2 expression vector pSG5hPPAR␥2 and a PPRE-driven luciferase reporter gene. The luciferase gene was under the control of a multimerized ACO-PPRE site and the TK promoter (Fig. 6). hPPAR␥2 was capable of activating this PPRE-based reporter 2-fold, an effect which was substantially enhanced when hPPAR␥2 was cotransfected together with RXR␣. Upon the addition of the PPAR␥ ligand BRL-14653, luciferase expression was increased 6-fold when the transfection was done with hPPAR␥2 alone or at least 10-fold when the cells were cotransfected with both hPPAR␥2 and mRXR␣. Similar results were obtained when prostaglandin J2 was used as a PPAR␥ ligand (data not shown).
Characterization of the Transcription Initiation Site of the Human PPAR␥ Gene-To unambiguously identify the 5Ј-end of the cDNA, several approaches were undertaken. First, primer extension experiments were performed, utilizing different human adipose tissue RNA samples, and results were independently confirmed by using 5Ј-RACE. Several primer extension products were seen for the PPAR␥1 mRNA using primer LF2  A (panel A) and B (panel B) exons of the human PPAR␥ gene. Transcription initiation sites as determined by primer extension (long arrows) and 5Ј-RACE (asterisks) are indicated. C, primer extension. Total human adipose tissue was used in primer extension. The major extension products are indicated by arrows. Size standards indicated on the right consist of a sequencing reaction. The sequence corresponding to the 5Ј-UTR is shown in panel A. Exons are denoted by gray or black rectangles and introns by a solid line. Restriction site for BamHI, is indicated by a B. The location of the ATG start-codon is indicated. The asterisk indicates the different ATG used in mPPAR␥1 (see Fig. 1). (Fig. 7, A and C). The relative positions of the transcription initiation sites as determined by the 5Ј-RACE were in agreement with the results for primer extension.
One major extension product of 62 bp was observed consistently with the primer LF-35 for PPAR␥2. A second extension product of 96 bp was found using the same primer (Fig. 7, B and  C). The results of 5Ј-RACE were consistent with the primer extension (Fig. 7B). The transcription initiation sites identified correlated well with the transcription initiation sites observed for the mouse PPAR␥2 mRNA (45). A striking feature of the human PPAR␥2 5Ј-UTR is its high degree of sequence conservation with the mouse 5Ј-UTR (see Fig. 9). It awaits further study to determine the exact implications of this conservation.
Structural Organization of the Human PPAR␥ Gene-To clone the human PPAR␥ gene and to determine its promoter sequence, we screened a PAC human genomic library derived from human foreskin fibroblasts. Three positive clones (P-8854, P-8855, and P-8856), each spanning Ͼ100 kb of genomic sequence, were isolated. All three clones were next shown to hybridize with the oligos LF-14 (corresponding to exon B) and LF-36 (exon 6), which indicates that they span most of the PPAR␥ coding region. More importantly, clone P-8856 also hybridized to oligo LF-2 and, hence, contains the transcription initiation site for PPAR␥1 and 2. This clone was further characterized by Southern blotting and partial sequence analysis, which allowed the construction of a physical map of the human PPAR␥ locus (Fig. 8). The human PPAR␥ gene spans more than 100 kb. The PPAR␥1 and PPAR␥2 mRNAs are encoded by 8 and 7 exons, respectively. The 5Ј-untranslated region of the PPAR␥1 mRNA is encoded by two exons, which we, in analogy to the nomenclature used for the mouse gene, named exon A1 and A2. The coding region of PPAR␥1 is contained in the next six exons (exons 1 to 6). Exons 1 to 6 also encode the majority of PPAR␥2 mRNA. The additional 28 amino acids of PPAR␥2 as well as the 5Ј-UTR are encoded by the B exon, which is located between exons A2 and A1.
The length of the introns was determined by long-range PCR (CLONTECH Tth polymerase mix) using the oligonucleotide pairs LF-3/LF-18, LF-20/LF-21, LF-22/LF-23, LF-24/LF-25, LF-26/LF-27, and LF-28/LF-29 and the PAC clone P-8856 as a template. The intron-exon boundaries were sequenced using genomic DNA as template. The 5Ј donor and 3Ј acceptor splice sites were found to be conforming to the consensus splice donor and acceptor sequences (Table II). The DNA binding domain of the receptors is encoded by exons 2 and 3, each encoding a separate zinc finger. The entire ligand binding domain is encoded by exons 5 and 6, which are separated by 16.3 kb of intron sequence.
Tissue-specific Determinants of the Human PPAR␥ Promoter-We next subcloned the region 5Ј to the transcription initiation sites of PPAR␥1 and ␥2 and sequenced the proximal promoters (Figs. 7 and 9). No canonical TATA box was found in the PPAR␥1 promoter region close to the transcription initiation site (Fig. 7). The sequence immediately upstream of the transcription initation site is extremely GC-rich, including several consensus Sp1 binding sites. Also a CCAAC box was found in the proximal promoter. Whether any of these factors are important for the regulation of PPAR␥1 gene expression awaits further study. The PPAR␥2 promoter contains a TATA-like element at position Ϫ68, relative to the transcription initiation site. Furthermore, sequence analysis identified a potential CAAT-like consensus C/EBP protein binding site at Ϫ56 (CCAATT) and a perfect AP-1 site at ϩ10 (TGACTCA) (Figs. 7 and 9).
In experiments to evaluate the tissue specificity of these promoters, DNA fragment extending from about Ϫ3 kb to ϩ110 bp and Ϫ1 kb to ϩ122 bp relative to the transcription initiation sites of PPAR␥1 and PPAR␥2 were inserted into the pGL3basic luciferase vector (Promega) to generate the constructs pGL3-␥1p3000 and pGL3-␥2p1000 (Fig. 10). These vectors were then transfected into mouse 3T3-L1 and Hep G2 cells. Transfection efficiency of the various cell lines was monitored by evaluation of the activity of control vectors. Relative to the promoterless parent vector, the human PPAR␥1 promoter fragment stimulated luciferase expression up to 3.5-fold in 3T3-L1 cells, maintained under non-differentiating conditions. In Hep G2 cells, luciferase expression was 9-fold higher with the pGL3-␥1p3000 vector relative to the pGL3-basic vector (Fig. 10). Similar results were obtained with COS cells (data not shown). The expression of the pGL3-␥2p1000 construct containing the PPAR␥2 promoter was not different from the pGL3-basic promoterless vector in Hep G2 cells. In undifferentiated 3T3-L1 cells, the PPAR␥2 promoter induced luciferase expression 2-fold relative to the promoterless control. DISCUSSION Two important findings recently underlined the importance of the PPAR␥ transcription factor. First, PPAR␥ has been iden-     tified as one of the key factors controlling adipocyte differentiation and function in rodent systems (6,7). Second, the recent identification of prostaglandin J2 derivatives and antidiabetic thiazolidinediones as natural and synthetic PPAR␥ ligands, respectively (28, 29, 46 -48). Thiazolidinediones are a new group of anti-diabetic drugs which improve insulin-resistance (for review, see Refs. 49 and 50). The identification of thiazolidinediones as PPAR␥ ligands together with the central role that adipose tissue plays in the pathogenesis of important metabolic disorders, such as obesity and non-insulin-dependent diabetes mellitus (NIDDM), have generated a major interest to determine the role of this PPAR subtype in normal and abnormal adipocyte function in humans.
The PPAR␥ gene spans about 100 kb and is composed of 9 exons, which give rise to PPAR␥1 and PPAR␥2 mRNAs by differential promoter usage and differential splicing. The gene structure as well as the sequence of the encoded protein are well conserved between human and mice (45) (99% similarity and 95% identity). Relative to the mouse, hamster, and Xenopus PPAR␥ (6,39,51), the human protein contains two additional amino acids. This is in agreement with the previous reports on the human PPAR␥ cDNA (34,35,52). The availability of the structure of the human PPAR␥ gene and protein will now allow for genetic studies, evaluating its role in disorders such as insulin resistance, NIDDM, and diseases characterized by altered adipose tissue function such as obesity or lipodystrophic syndromes.
To determine tissue-specific patterns of expression of the human PPAR␥ gene, we developed an RT-competitive PCR assay. Unlike results of previous reports, which used commercially available blots or single RNA samples (34,35), we used multiple independent samples to base our conclusions on. As was observed in rodents (6, 7), we found PPAR␥ to be strongly expressed in adipose tissue. In addition to adipose tissue, the large intestine had high levels of PPAR␥ expression. Several other tissues, such as liver, kidney, and small intestine contained lower but nevertheless considerable levels of PPAR␥ RNA. Skeletal muscle, in contrast, contained only trace amounts of PPAR␥ mRNA.
In adipose tissue and liver, about 15% of all PPAR␥ mRNA was of the PPAR␥2 type, whereas in the remaining tissues no PPAR␥2 mRNA was detected. These observations have several important implications. First, our data question the relative importance of PPAR␥2. Indeed, our results in humans as well as the data by Xue et al. (53) in rodent adipocytes show consistently lower levels of PPAR␥2 mRNA and protein relative to the PPAR␥1 subtype. These observations are in line with the previous observations that the N-terminal domain of PPAR␥ was dispensable, both regarding transcriptional activation and capacity to induce adipocyte differentiation in vitro (7). However, the N-terminal domain is highly conserved between different species, suggesting it might have an important function in vivo. Second, PPAR␥ expression is much more widespread than previously realized, which implies that PPAR␥ controls gene expression in several tissues in addition to adipose tissue. Especially striking is the high level of PPAR␥ expression in the human large intestine. These reports are consistent with the reported high level expression of PPAR␥ in colonic mucosa in mouse (54). It is interesting to note that fatty acids, potential PPAR activators, have been shown to play an important role in modulating the function of the large intestine. For instance diets enriched in saturated lipids have been shown to predispose to the development of colon cancer (55). Furthermore, it has been shown that diets enriched in -3 fatty acids, powerful PPAR activators, have a beneficial response on inflammatory diseases of the gastrointestinal tract such as colitis ulcerosa and Crohn's disease (56,57). Since the high level expression of PPAR␥ suggest that it might play an important role in normal and abnormal colonic function, further studies aimed at exploring this are definitely needed. Finally, the low levels of PPAR␥ expression in skeletal muscle cells also deserve some reflection. Muscle is responsible for clearance of the majority of glucose in the body and abnormal muscle glucose uptake is one of the prime features of insulin resistance and NIDDM. The low levels of PPAR␥ in muscle argue, therefore, that the beneficial effects of thiazolidinedione antidiabetic agents are not likely to FIG. 10. Tissue-specific activity of the PPAR␥ promoter. A, normalized luciferase activity of the pGL3-␥1p3000 construct containing 3000 bp of regulatory sequence of the human PPAR␥1 gene after transfection in 3T3-L1 and Hep G2 cells. Transfections were performed as described under "Experimental Procedures." Scheme of the reporter constructs pGL3-␥1p3000 used in transfection assays is shown above the graphic. B, normalized luciferase activity of the pGL3-␥2p1000 construct containing 1000 bp of regulatory sequence of the human PPAR␥2 gene after transfection in 3T3-L1 and Hep G2 cells. Transfections were performed as described under "Experimental Procedures." Scheme of the reporter constructs pGL3-␥2p1000 used in transfection assays is shown above the graphic. be due to a direct effect of these agents on PPAR␥ present in the muscle. In fact, even though the liver has considerably higher levels of PPAR␥ relative to muscle, thiazolidinediones do not seem to affect PPAR responsive genes in liver tissue at the concentrations commonly used to lower glucose levels (23). This observation together with the observed tissue distribution of PPAR␥ suggests that the glucose lowering effects of the thiazolidinedione PPAR␥ ligands are primarily a result of their activity on adipose tissue, which then, via a secreted signal, might influence muscle glucose uptake.
To identify the molecular circuitry underlying tissue-specific expression of PPAR, we cloned and performed an initial characterization of the human PPAR␥ promoters. As shown, 3000 bp of the PPAR␥1 and 1000 bp of the PPAR␥2 promoter account for substantial levels of basal promoter activity. Further functional studies are underway to determine elements necessary for tissue-specific and regulated expression of the PPAR␥ gene. In this context, it will be interesting to determine the effects of transcription factors known to induce adipocyte differentiation on PPAR␥ expression in this tissue and to define the hierachical role that PPAR␥ plays in this process. PPAR␥ is not the only transcription factor involved in adipocyte differentiation. In addition to PPAR␥, the basic helix-loop-helix leucine zipper factor ADD-1/SREBP1 and transcription factors of the C/EBP family also play a role in determining adipocyte differentiation. It is interesting to note that, as in the mouse PPAR␥2 promoter (45), a potential consensus C/EBP response element could be identified in the human PPAR␥2 promoter by homology searches. This observation fits well with the previous observation that forced expression of C/EBP␤ could induce PPAR␥ expression and further studies on this subject are underway (11,12).
In conclusion, we report the characterization of the human PPAR␥ gene structure and furthermore define the structure of the PPAR␥1 and ␥2 promoter. In addition, our data show that human PPAR␥ has a similar structure and similar transactivation function as the rodent PPARs. The expression patterns of PPAR␥1 and ␥2 show that in man, PPAR␥1 is the predominant form. Our results furthermore demonstrate that, in addition to adipose tissue, human colon expresses high levels of PPAR␥. It is expected that the gene structure will facilitate our analysis of eventual PPAR␥ mutations in humans, whereas knowledge of expression patterns and sequence elements, as well as factors regulating PPAR␥ gene expression, could be of major importance in understanding PPAR biology.