Identification, characterization, and tissue distribution of human peroxisome proliferator-activated receptor (PPAR) isoforms PPARgamma2 versus PPARgamma1 and activation with retinoid X receptor agonists and antagonists.

We describe the cloning, characterization, and tissue distribution of the two human peroxisome proliferator activated receptor isoforms hPPARγ2 and hPPARγ1. In cotransfection assays the two isoforms were activated to approximately the same extent by known PPARγ activators. Human PPARγ binds to DNA as a heterodimer with the retinoid X receptor (RXR). This heterodimer was activated by both RXR agonists and antagonists and the addition of PPARγ ligands with retinoids resulted in greater than additive activation. Such heterodimer-selective modulators may have a role in the treatment of PPARγ/RXR-modulated diseases like diabetes. Northern blot analysis indicated the presence of PPARγ in skeletal muscle, and a sensitive RNase protection assay confirmed the presence of only PPARγ1 in muscle that was not solely due to fat contamination. However, both PPARγ1 and PPARγ2 RNA were detected in fat, and the ratio of PPARγ1 to PPARγ2 RNA varied in different individuals. The presence of tissue-specific distribution of isoforms and the variable ratio of PPARγ1 to PPARγ2 raised the possibility that isoform expression may be modulated in disease states like non-insulin-dependent diabetes mellitus. Interestingly, a third protected band was detected with fat RNA indicating the possible existence of a third human PPARγ isoform.

Peroxisome proliferator-activated receptors (PPARs) 1 are members of the intracellular receptor superfamily. They play a role in lipid metabolism and metabolic diseases. There are three PPAR subtypes with distinct tissue distribution in Xenopus, mice, and humans: PPAR␣, PPAR␤ (also called NUC1 or PPAR␦), and PPAR␥ (1)(2)(3)(4)(5)(6)(7)(8)(9)(10). PPAR␥ expression is observed in adipose tissue in rodents. Its expression is induced early in differentiation of 3T3-L1 preadipocytes into adipocytes, and its overexpression in fibroblasts induces them to differentiate into adipocytes (11). Two isoforms of mPPAR␥ resulting from different promoters and alternate splicing have been identified (7,12,13). A human isoform, hPPAR␥1, has been cloned from a human hematopoietic cell line and placenta (14,15), and another from human fat (16) has been reported. Although a preliminary report on the distribution of PPAR␥ in human tissues has been published (16), the distribution of PPAR␥1 versus PPAR␥2 has not been reported.
Thiazolidinediones are high affinity ligands and potent activators for PPAR␥. They decrease insulin resistance in insulinresponsive tissues including skeletal muscle (the primary site of insulin-stimulated glucose uptake) in patients with noninsulin-dependent diabetes mellitus (17). It is assumed that PPAR␥ is the therapeutic target for these compounds; yet the presence of PPAR␥ has not been conclusively demonstrated in human muscle. The identification of human PPAR␥ isoforms and their tissue distribution will help in understanding their role in metabolic diseases like non-insulin-dependent diabetes mellitus and obesity.
We undertook to clone and characterize the tissue distribution of human PPAR␥1 and PPAR␥2 and compare it with that of human PPAR␣ and PPAR␤. We compared the ability of PPAR␥ agonists to activate the two isoforms. A PPAR␥ antagonist would be a useful tool to dissect PPAR␥ action and may also block adipocyte differentiation. Such a ligand that competitively antagonizes PPAR␥ activity has not been reported. An alternative approach would be to block PPAR␥/RXR activation with an antagonist of RXR. Surprisingly, an RXR antagonist activated the PPAR␥/RXR heterodimer as did an RXR agonist. Greater than additive activation was seen with PPAR␥ and RXR ligands. EXPERIMENTAL PROCEDURES 5,8,11,Eicosatetraenoic acid and 2-bromopalmitate were purchased from Sigma, and 15-deoxy-⌬ 12,14 -prostaglandin J2 was obtained from Cayman Chemicals. BRL 49653, LG100268, and LG100754 were synthesized at Ligand Pharmaceuticals Inc.
A human heart 5Ј-stretch cDNA library (Clontech) was screened with a mouse PPAR␥ (7) probe at low stringency (35% formamide, 5 ϫ SSC, 0.1% SDS, 100 g/ml fish sperm DNA at 37°C). Several positive clones were isolated and sequenced. Comparison with the mPPAR␥ sequence indicated that one clone encoded the N terminus and another the C terminus of hPPAR␥, and their sequences overlapped by 485 base pairs. The complete hPPAR␥-coding region was reconstructed by a triple ligation using pBKCMV (Stratagene) digested with EcoRI and KpnI and utilizing the unique ScaI site in the coding region. This plasmid was then digested with NcoI, blunt-ended with Klenow enzyme, and redigested with KpnI. The liberated fragment was subcloned into pBKCMV at the XbaI site (blunted with Klenow enzyme) and the KpnI site to give pCMVhPPAR␥1.
A third positive clone was isolated and sequenced. This sequence * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) U79012.
Northern Blots-Human multiple tissue Northern blots were purchased from Clontech. Hybridization was done according to the manufacturer's protocol. The probe for hPPAR␣ has been described (6). pC-MVhPPAR␤ (pCMVhNUC1) (18) was digested with EcoRI and the 500-base pair fragment was isolated. pCMVhPPAR␥1 was digested with ScaI and KpnI and the 1-kilobase pair fragment isolated. This probe will recognize hPPAR␥1 and hPPAR␥2 RNA. All probes were labeled by random priming.
RNase Protection Assays-Four human white fat and two skeletal muscle samples were obtained from the National Disease Research Interchange (NDRI, Philadelphia) or the University of California (San Diego) tissue bank. Total RNA was isolated using standard techniques. A sample of human skeletal muscle RNA was purchased from Clontech.
A partial cDNA containing nucleotides 1-252 of hPPAR␥2 ( Fig. 1) was subcloned into the pCRII vector (Invitrogen). This was linearized with CelII and labeled antisense riboprobe made with the T7 RNA polymerase and Maxiscript in vitro transcription kit (Ambion). The adipocyte protein 2 (aP2, a kind gift from Dr. Bruce Spiegelman) cDNA was liberated with BamHI and subcloned into pGEM 3Zf(Ϫ) (Promega). The DNA was linearized with BclI and riboprobe made using SP6 RNA polymerase. RNase protection assay was done with an Ambion direct protect lysate RNase protection assay kit. Band intensities were quantitated by a PhosphorImager (Molecular Dynamics).
Co-transfection Assays-Transfections in CV-1 cells were performed as described (6,23). The reporter plasmid pPPREA3-tk-Luc containing three copies of the PPRE identified in the acyl CoA oxidase (AOX) gene has been described (24). The ␤-galactosidase expression plasmid pCH110 was used to normalize difference in transfection efficiencies. The normalized response is the luciferase activity of the extract divided by the ␤-galactosidase activity of the same. Compounds were dissolved in Me 2 SO (vehicle). Each data point is the mean of triplicate transfections, and the error bars represent the standard error of the mean. Each experiment was repeated at least two times. A representative experiment is shown in each case.

RESULTS
A human heart cDNA library was screened with a probe corresponding to the mouse PPAR␥ (7). Three overlapping clones were identified, purified, and sequenced. The nucleotide sequence is shown in Fig. 1. The longest open reading frame starting from the nucleotide at position 91 coded for a polypeptide of 505 amino acids. There was an in-frame stop codon upstream of this methionine suggesting the translation initiation occurred from this codon. The second and third methionine codons were at positions 29 and 31 in the amino acid sequence. The first and third methionine codons in hPPAR␥2 were in a context appropriate for translation initiation, i.e. the Kozak sequence (25), and were conserved between mice and man. The second methionine codon was unique to human PPAR␥2, and the corresponding amino acid in mice was isoleucine.
Amino acid sequence comparison indicated 97% identity overall between hPPAR␥2 and mPPAR␥2. The DNA binding domains were 83% conserved between hPPAR␥2 and hPPAR␣ or hPPAR␤. Further, three amino acids were present between the two cysteines in the D-box (amino acids 177-179), a characteristic feature of all PPARs known to date. Based on these observations we believe this human isoform is hPPAR␥2.
To determine if there are multiple translation start sites for hPPAR␥2, as in mPPAR␥2, coupled in vitro transcription/ translation reactions were performed in the presence of [ 35 S]methionine and pCMVhPPAR␥2 as template. Two bands were observed by PAGE (Fig. 2). The upper band (57 kDa) corresponded to translation initiation from the methionine at position 1. The lower band (53 kDa) probably corresponded to translation initiation from the methionine at position 31. We cannot rigorously discount translation initiation from the methionine at position 29. However, since this methionine was not within a good Kozak sequence and was also absent in mP-PAR␥2, we think it is unlikely. Indeed, in vitro transcription/ translation of pCMXmPPAR␥1 (10) and pCMVhPPAR␥1 also gives rise to bands that comigrate with the lower band observed with pCMVPPAR␥2. Hence, in analogy with mPPAR␥1, we called the smaller polypeptide hPPAR␥1.
Since PPAR␥2 binds to PPREs as a heterodimer with RXR, we next determined the transcriptional response of the PPAR␥2/RXR heterodimer to an RXR ligand. LG100268 (28) is a highly selective RXR ligand (K d ϳ3 nM). Both BRL 49653 and LG100268 transcriptionally activated the PPAR␥2/RXR heterodimer (Fig. 5A), and the transcriptional response observed with both ligands was greater than that observed individually. RXR agonists activated a reporter containing the hydratase (bifunctional enzyme) PPRE. They also induced expression of the hydratase gene in vivo, and increased induction is seen with a combination of RXR and PPAR agonists.  1, 4, and 7), acyl-coenzyme A oxidase (lanes 2, 5, and 8) and apoA-1 (lanes 3, 6, and 9) genes were used as probes. Recently LG100754, another high affinity RXR ligand (K d ϳ12 nM), has been described as an RXR/RXR homodimer antagonist on a CRBPII-tk-Luc reporter (29). To determine if the response of PPAR␥2/RXR to BRL 49653 will be antagonized by LG100754 binding to RXR, a cotransfection assay with PPAR␥2/RXR was performed (Fig. 5A). Surprisingly, LG100754, like LG100268, is an agonist of hPPAR␥2/RXR, and activation by the combination of BRL 49653 and LG100754 is greater than the individual compounds. It is also an agonist of hPPAR␥1/RXR (data not shown) and hPPAR␣/RXR (30).
Since LG100754 is a high affinity RXR ligand and also activates the hPPAR␥/RXR heterodimer, we determined whether LG100754 also activates RXR homodimers using the same reporter used for the PPAR assays (pPPREA3-tk-Luc). Since the consensus PPRE and RXR response element are of the DR-1 type (24), it would be interesting to compare the effect of RXR modulators on the two response elements.
LG100268 strongly activated the RXR/RXR homodimer on PPREA3-tk-Luc (Fig. 5B). LG100754 was a very weak activator of the RXR homodimer. Interestingly, it antagonized the activation of RXR/RXR by LG100268 (Fig. 5B). Hence, LG100754 acted as a PPAR␥/RXR heterodimer agonist but as an RXR homodimer antagonist on the same response element, a PPRE. This was the first demonstration of an RXR ligand having such dimerselective effects on the same reporter. This dimer-selective activity is probably not due to LG100754 binding with high affinity to PPAR␥ since LG100754 displaces labeled BRL 49653 from PPAR␥ only at very high concentrations in a DNA-dependent ligand binding assay using PPAR␥/RXR heterodimers (40) (data not shown).
We next determined the tissue distribution of human PPAR␥ RNA (using a probe common to PPAR␥1 and PPAR␥2) and compared it with that of hPPAR␣ and hPPAR␤ by Northern blotting (Fig. 6A). Human PPAR␣ was found predominantly in skeletal muscle, liver, heart, and kidney, a distribution similar to that reported for mPPAR␣. PPAR␤ RNA was more ubiquitously expressed with maximal expression in placenta and skeletal muscle. One band approximately 2 kilobases in length was observed with the PPAR␥ probe. Human PPAR␥ is expressed in the insulin-responsive tissues (skeletal muscle, heart, and liver) and is consistent with the distribution in mice (31).
Since the probe used in the Northern blot experiments could not distinguish between PPAR␥1 and PPAR␥2 RNA, we developed an RNase protection assay to distinguish the two isoforms. The majority of insulin-stimulated glucose uptake occurs in skeletal muscle, therefore, we determined the expression of PPAR␥2 versus PPAR␥1 in muscle. Since mP-PAR␥2 expression is restricted to fat (32), and the commercial blot used in the Northern analysis did not have a sample of fat RNA, we included human fat RNA in the study (Fig. 6B). Two bands (78 and 252 nucleotides long) were observed in all adipose tissue samples arising from protection of the probe by PPAR␥1 and PPAR␥2 RNA, respectively, as shown. In contrast to the findings in mice (31), PPAR␥1 was expressed at higher levels in all human fat samples studied. Quantitation of the band intensities indicated that the ratio of PPAR␥1 to -␥2 varied in the human samples (from 2 (lane 5) to 10 (lane 7)).
With RNA from human skeletal muscle, we observed the protected fragment due to hPPAR␥1 but not from PPAR␥2 in all three samples. This was not observed with yeast RNA, which was used as a negative control.
To test whether PPAR␥1 RNA observed in muscle was solely due to fat contaminating the muscle samples, we performed RNase protection assays with the muscle RNA and a mouse aP2 probe and compared that with a sample of fat RNA. aP2 (adipocyte protein 2) gene expression is fat-specific (33). Very little specific protection of the probe was seen with 10 g of muscle RNA (Fig. 6C, lanes 5-7) while an intense band is seen with only 2 g of fat RNA (lane 4). The autoradiogram was deliberately overexposed (see lane 3) to reveal any protected bands in lanes 5-7. The smear observed in lanes 5-7 was probably due to nonspecific hybridization between human RNA and the mouse probe and is also seen with yeast RNA (lane 8). We concluded that PPAR␥ was expressed in human skeletal muscle and PPAR␥1 was the predominant isoform in this tissue. In contrast, both PPAR␥1 and PPAR␥2 were expressed in human fat and at much higher levels compared with muscle.
Interestingly, a third protected fragment (170 nucleotides long) was also observed (denoted by an asterisk) in all four fat samples but not in the muscle samples (Fig. 6B). This could be simply due to RNase digestion in regions of imperfect hybridization. However, the ratio of the intensity of this fragment compared with PPAR␥2 varied in the different fat samples, hence, it is unlikely that this was due to breakdown of the larger protected fragment. We therefore hypothesized a third isoform of PPAR␥ in humans that may arise due to alternate splicing and promoter usage. DISCUSSION We have cloned the cDNA for a second isoform of the human PPAR␥, hPPAR␥2. Sequence comparison with mPPAR␥2 revealed 97% amino acid identity. Human PPAR␥2 bound to PPREs as a heterodimer with RXR and was activated by the PPAR␥ ligands BRL 49653 and 15-deoxy-⌬ 12,14 -prostaglandin J2. Based on these observations we believe that hPPAR␥2 was a genuine member of the PPAR subfamily. The amino acid sequence shown in Fig. 1 was identical to the hPPAR␥2 amino acid sequence predicted from the cDNA isolated from a human adipose library (16). However, our clone contained an additional 90 nucleotides of 5Ј-untranslated sequence including the upstream in-frame translation stop codon.
PPAR␥2, like PPAR␥1, bound to PPREs as a heterodimer with RXRs as do all the PPARs known to date. Human PPAR␥1 and hPPAR␥2 have similar activation profiles in reponse to BRL 49653 and 15-deoxy-⌬ 12,14 -prostaglandin J2. Interestingly, there was only 63% identity in the N-terminal 30 amino acids between human and mouse PPAR␥2, far less than in the rest of the polypeptide (98%). This suggests that the N terminus coded by a different exon (13) has diverged more rapidly than the rest of the protein during evolution. The function of these amino acids is unclear.
The hPPAR␥2/RXR and hPPAR␥1/RXR heterodimers were activated by the RXR modulators LG100268 and LG100754. They increased the transcriptional response seen with the PPAR␥ agonist BRL 49653. This is consistent with our previous studies showing that RXR modulators increase the responsiveness of the PPAR␣/RXR heterodimer (6,24). LG100754 was interesting because it is an agonist of PPAR␥/RXR but an RXR/RXR antagonist. Binding of LG100754 to RXR may lead to distinct conformational changes of the receptor dimer such that PPAR␥/RXR is read as an activator by the transcriptional machinery, but the RXR/RXR homodimer is transcriptionally silent. Such compounds like LG100268, LG100754, and BRL 49653 may therefore modulate distinct but overlapping sets of target genes and might have a role in the treatment of PPAR/ RXR-modulated diseases like diabetes.
The tissue distribution of hPPAR␥ is important for the therapeutic activity of drugs targeting the PPAR␥/RXR heterodimer. Thiazolidinediones act as insulin sensitizers in skeletal muscle and are high affinity PPAR␥ ligands (34). Structure activity relationship indicates a good correlation between in vivo potency and in vitro activity of thiazolidinediones (35), implicating PPAR␥ as the therapeutic target for these compounds. However, earlier data indicated PPAR␥ is expressed at high levels, specifically in adipose tissue in rodents (11,32), and is essentially undetectable in muscle where approximately 80% of insulin-stimulated glucose uptake occurs (36). It was not clear how PPAR␥ expressed almost exclusively in adipose tissue could have the effect of insulin sensitization in skeletal muscle and raised the possibility that insulin sensitization by thiazolidinediones in skeletal muscle was not mediated by PPAR␥-dependent mechanisms.
Our data demonstrated that PPAR␥ was expressed in human skeletal muscle, fat, and heart, tissues where the majority of insulin-stimulated glucose uptake occurs. Further, while both PPAR␥1 and PPAR␥2 are expressed in human fat, the dominant isoform in human muscle is PPAR␥1. These findings are consistent with recently published data in mice (31) and suggest that PPAR␥1 might be the relevant target for thiazolidinediones in human skeletal muscle. The close conservation of sequence, subtype, and tissue distribution of PPAR␥ between mice and humans is consistent with the observation that thiazolidinediones act as insulin sensitizers in both species (17,37). However, we do not yet know the distribution of PPAR␥1 versus PPAR␥2 protein in these tissues.
The ratio of the intensities of the PPAR␥1 and -␥2 isoforms varied in the four individual fat samples, hinting that isoform expression may be modulated. Further, only PPAR␥1 was detected in muscle, not PPAR␥2, pointing to differential expression of PPAR␥ isoforms in tissues. Therefore, an analysis of PPAR␥ isoform distribution in skeletal muscle and fat in normal, obese, and diabetic individuals might yield valuable information and is currently underway.
We used a commercially available Northern blot to determine PPAR␥ distribution. Although we did not observe hybridization to placenta and lung RNA on this blot we note that PPAR␥ expression was observed in these tissues (15). We cannot explain this other than as a variation between individuals. For an accurate determination it was important to assay expression levels in several individuals as was done in our RNase-protection assays.
Thiazolidinediones appear to act as insulin sensitizers in vivo through activation of PPAR␥/RXR. Our data indicate the presence of PPAR␥ in insulin-responsive tissues in humans. One may speculate that thiazolidinediones bind to and activate PPAR␥ altering the expression of key genes in target tissues rendering them more responsive to insulin, increasing glucose uptake, lowering hepatic glucose output, and lowering hyperglycemia. Since RXR modulators are also able to activate the PPAR␥/RXR heterodimer, they could activate a set of thiazolidinedione-responsive genes and may therefore either alone or in combination with thiazolidinediones have utility in the treatment of non-insulin-dependent diabetes mellitus.