Human Adipocyte Fatty Acid-Binding Protein (aP2) Gene Promoter-Driven Reporter Assay Discriminates Nonlipogenic Peroxisome Proliferator-Activated Receptor γ Ligands

Metabolic syndrome is characterized by the clustering of at least three risk factors among hypertension, certain types of dyslipidemia, impaired glucose tolerance and type II diabetes, and obesity. These metabolic abnormalities lead to atherosclerosis and related complications (Haffner and Taegtmeyer, 2003). The control of lipid and carbohydrate metabolism, including physiologic and pharmacological treatments, represents a valid rationale to reduce cardiovascular diseases in patients with metabolic syndrome (Beckman et al., 2003; Wilson and Grundy, 2003).

PPARs are a subclass of the nuclear receptor superfamily acting as ligand-dependent transcription factors (Kersten et al., 2000). Three subtypes were identified; the α isoform is the primary subtype expressed in liver, but also in heart and kidney, and acts as a major regulator of metabolism of fats, catabolism of fatty acids, and synthesis and catabolism of lipoproteins (Barbier et al., 2002). PPARα is also involved in cholesterol efflux from peripheral tissues and the high-density lipoprotein reverse cholesterol-transport pathway (Chinetti et al., 2001).

The PPARγ isoform is essentially expressed in adipocytes, where it stimulates lipoprotein lipase and promotes fatty acid uptake and storage in mature adipocytes. These effects, along with increased insulin sensitivity, decrease circulating free fatty acids and triglycerides (Reginato and Lazar, 1999). Like PPARα, PPARγ also influences lipid metabolism in macrophages from the arterial wall by up-regulating the expression of ATP-binding cassette transporter A1, which results in an increased cholesterol efflux (Chawla et al., 2001). The functions of the ubiquitous PPARδ isoform remain quite mysterious so far, despite 1) a role in epidermal maturation and skin wound healing (Wahli, 2002), 2) properties related to lipid metabolism, as well as 3) a putative role in fat metabolism recently reported (Oliver et al., 2001; Wang et al., 2003).

Thiazolidinediones (TZDs) are efficacious insulin-sensitizing agents used in the treatment of type II diabetes and work through the binding and activation of PPARγ. However, both in pharmacological and clinical use, TZDs increase adiposity and body weight, and elicit clinical side effects including edema, plasma volume expansion, and hemodilution (Aleo et al., 2003; Schöfl and Lübben, 2003; Yang et al., 2003). These drawbacks stimulate the search for PPARγ modulators with modified pharmacological profiles that regulate fatty acids and glucose metabolism with a reduced increase in adiposity, since this undesirable effect may favor the development of obesity, a risk factor often associated with insulin resistance (Oberfield et al., 1999). The adipocyte-specific PPARγ2 isoform contains additionally 30 amino acids at the amino terminus DNA-binding domain as compared with the PPARγ1 isoform, and is characterized by a greater constitutive transcription activation function than PPARγ1 (Werman et al., 1997). It has been hypothesized that PPARγ2 may regulate adipogenesis specifically, whereas in the absence of ligand, PPARγ1 does not (Ren et al., 2002); nevertheless, the selective modulation of PPARγ1 versus PPARγ2 has yet to be established and may represent a complex pharmacological challenge.

The identification of partial agonists for PPARγ represents an attractive pharmacological rationale which needs further validation to fulfill the aforementioned criteria. Regarding nuclear receptors, partial agonism can be achieved by either 1) modulating cofactor interactions (i.e., activator recruitment, repressor releasing, stoichiometry), 2) reducing the heterodimerization with retinoid X-receptor, or 3) controlling the interaction between the PPAR DNA-binding domain and response elements in the promoter region of target genes. In the present study, this latter hypothesis was addressed, and the partial efficacy for PPARγ modulators monitored in reporter-gene and lipogenesis assays was related to lower neutral lipid accumulation in vitro. This paper also describes the cloning and functional characterization of the full-length promoter of the human homolog for aP2, namely, fatty acid-binding protein-4 (FABP4), which has not been described so far. Its in vitro regulation may represent either an interesting complement or an alternative approach to adipocyte differentiation assays.

Materials and Methods

Plasmids. Chimeric receptors containing the yeast GAL4 DNA-binding domain fused to either human PPARα or PPARγ ligand-binding domain were generated. A pFA-CMV plasmid (Stratagene, La Jolla, CA) containing a yeast GAL4 DNA-binding domain (BD) downstream of a multiple cloning site was used as backbone vector. Human PPARα (GenBank ID: S74349) and PPARγ (GenBank ID: U63415) ligand-binding domain (position 619 to 1530 for PPARα, position 607 to 1518 for PPARγ) was PCR-amplified, and a BamHI restriction site was inserted at the 5′ end of the ligand-binding domain sequence and in frame with the GAL4 BD reading frame to generate chimeric receptor genes. Each construct was controlled by automated DNA sequencing to validate the GAL4 BD-PPAR ligand-binding domain chimeric gene nucleotide sequence. The corresponding reporter plasmid for these GAL4 chimeric receptors (pFR-Luc; Stratagene) contained five upstream activation sequences of the yeast GAL4 gene promoter upstream of a canonical TATA box and adjacent to a luciferase reporter gene (pFR-Luc).

Cloning of the human full-length PPARα and PPARγ2 plasmids and the construction of the corresponding reporter plasmid containing three copies of the consensus PPAR response element (PPRE3-HSV-luc) were reported previously (Wurch et al., 2002). The human PPARγ1 subtype (GenBank ID: L 40904) was PCR-amplified and cloned according to the same strategy.

Construction of the Human Fatty Acid-Binding Protein-4 Gene Promoter Reporter Plasmid. The promoter-reporter plasmid, containing the luciferase coding sequence under the transcriptional control of a 5400-bp-long fragment having both the enhancer and core promoter elements of the human FABP4 gene promoter, was obtained according to the following strategy. A sequence coding for human FABP4 (GenBank accession number J02874) was identified on bacterial artificial chromosome clone RP11-157I4 (accession number AC018616) of the Whitehead Institute Center for Genome Research (Cambridge, MA); this bacterial artificial chromosome encompassed about 166,000 bp of human chromosome 8. The start codon of the FABP4 was located at position 24,025 of AC018616, and the 5′ end of its mRNA was mapped to 23,978. By analogy to the mouse aP2 promoter (Graves et al., 1992), a 5400-bp-long genomic sequence upstream of the putative transcription start point was amplified by PCR on human genomic DNA and sense (5′-CATTCAGAAAGGAACTTTGTTTCAAATAAAAGGAGAG-3′) and reverse (5′-ATTATT CTTCAAGGAGAGAAGGAAGCTGCA-3′) primers and a long-range polymerase mixture (Expand Long Template PCR system; Roche Applied Science, Indianapolis, IN). PCR products were cloned into pCR4.1 and fully sequenced on an automated DNA sequencer (ABI310 Genetic Analyzer; Applied Biosystems, Foster City, CA). Clones from independent amplifications were compared to rule out PCR errors. The 5403-bp-long promoter was further subcloned into a promoterless pGL3Basic vector (Promega, Charbonnieres, France) upstream of a luciferase reporter gene.

Reporter-Gene Assays. COS-7 cells [American Type Culture Collection (ATCC), Manassas, VA; CRL-1651] were seeded (12–15 × 10³ cells/well in 96-well plates) in high-glucose (4.5 g/l) Dulbecco's modified Eagle's medium containing 50 μg/ml gentamycin, 2.5 μg/ml fungizone, and supplemented with 10% fetal calf serum. After 24 h in a humidified incubator (37°C, 5% CO₂ in air), they were up to 60 to 80% confluence before transfection in serum-free and antibiotic-free medium by LipofectAMINE Plus reagent (Invitrogen, Cergy, France) according to the instructions of the manufacturer. Cells were incubated for 4 h with transfection mixtures containing 9 ng of PPAR (either wild-type or chimeric) and 37 ng of the corresponding reporter plasmid, and then washed before the addition for 24 h of fresh medium containing delipidated charcoal-stripped fetal calf serum and antibiotics. Then, cells were treated with either compound or vehicle [0.1% dimethyl sulfoxide (DMSO)] for an additional 24 h, and luciferase activity was measured as previously described (Wurch et al., 2002).

Cells treated with vehicle corresponded to the basal transcription level of luciferase gene, which was subtracted from the ligand responses. Experimental points were best-fitted using the Sigma Plot v.4.0.1 software (SPSS Inc., Chicago, IL; four-parameter logistic equation), and EC₅₀ values were deduced as the concentration of ligand that yielded 50% of its own maximal response. In case a plateau phase could not be reached due to insolubility or toxicity of the compound at high concentrations, an apparent EC₅₀ value was calculated as the concentration of ligand yielding 50% of the maximal response at the highest investigated concentration.

Adipocyte Differentiation Assays. Murine 3T3-L1 cells (ATCC, CCL 92.1 passage 5–12) were seeded at 8 to 8.5 × 10³ cells/well of 96-well plates and were grown to confluence at 37°C in 5% CO₂ in high-glucose Dulbecco's modified Eagle's medium containing pyruvate, 10% fetal calf serum, gentamycin (50 μg/ml), and fungizone (2.5 μg/ml). Two days after reaching confluence, cells were incubated in the same medium containing a so-called differentiation cocktail [1 μM dexamethasone + 5 nM insulin + 0.5 mM isobutylmethylxanthine] along with various concentrations of PPAR modulators or their vehicle (0.1% DMSO).

For measurement of murine aP2 (maP2) mRNA, 3T3-L1 cells were maintained in the presence of the differentiation cocktail and drug treatment for 3 days. Then, they were washed with phosphate-buffered saline without Ca²⁺ and Mg²⁺, dried, and frozen at –80°C. Total RNA was isolated from murine 3T3-L1 cells using an RNeasy Mini Kit according to manufacturer specifications (QIAGEN, Valencia, CA). Contaminating DNA in the RNA preparation was removed by DNase I treatment on-column at room temperature for 15 min. For reverse transcription, 500 ng of total RNA was used to generate cDNA in a total volume of 20 μl. iScript (Bio-Rad, Hercules, CA) was used for first-strand cDNA synthesis during 40 min at 42°C. Reaction was stopped by a 5-min step at 85°C. Real-time PCR was carried out on the iCycler iQ Real Time PCR Detection System (Bio-Rad) using gene-specific primers and iQ SYBR green Supermix (Bio-Rad). The sequences of the primers are as follows: maP2 (NM_024406) forward 5′-GGGCGTGGAATTCGATGAAATCA-3′, maP2 (NM_024406) reverse 5′-CCCGCCATCTAGGGTTATGAT-3′, m36B4 (NM_007475) forward 5′-GGACCCGAGAAGACCTCCTT-3′, and m36B4 (NM_007475) reverse 5′-AATGGTGCCTCTGGAGATTTTCG-3′. The PCRs were performed in a final volume of 25 μl, as follows: 5 min at 95°C to activate “hot start” enzyme and 40 cycles at 94°C for 30 s, 60°C for 30 s, and 72°C for 30 s, followed by a melt curve from 65°C to 95°C (0.5°C every 10 s), to determine specific temperature melting of each amplicon. m36B4 was used as internal reference for normalization of maP2 relative quantifications.

To examine lipid accumulation during adipogenesis, 3T3-L1 cells were maintained in the presence of the differentiation cocktail and drug treatment for 7 days without medium exchange. Then, cells were washed gently with phosphate-buffered saline in the presence of Ca²⁺ and Mg²⁺, before the addition of 5 μl of AdipoRed reagent (Cambrex Bio Science Walkersville, Inc., Walkersville, MD) per well. AdipoRed is a solution of Nile red whose maximum emission of the fluorescent signal (λ_ex = 485 nm, λ_em = 572 nm) depends on the hydrophobic environment, especially when the stain is partitioned in triglyceride droplets.

Statistical Analysis. All data were expressed as mean ± S.E.M. Statistical significance (for results in Tables 2 and 3) was determined with Dunnett's multiple comparison procedure applied after a one-way analysis of variance conducted on data (P value of less than 0.05 was considered significant).

Materials. All molecular biology reagents were obtained from either Invitrogen, BD Biosciences Clontech (Palo Alto, CA), Roche Applied Science, Stratagene, Promega, or Applied Biosystems. Cells were obtained from ATCC. Dulbecco's modified Eagle's medium, phosphate-buffered saline, gentamycin, fungizone, and LipofectAMINE Plus reagent were purchased from Invitrogen; delipidated and charcoal-stripped fetal calf serum was obtained from Hyclone Laboratories (Erembodegem, Belgium). Luminescence was measured by using a luciferase assay system from Promega. Triglyceride accumulation was quantified by using AdipoRed assay reagent (Cambrex Bio Science Walkersville, Inc.). Rosiglitazone, 2,2-dichloro-12-(4-chlorophenyl)dodecanoic acid (BM-17.0744), 2-(4-[2-(3-[2,4-difluorophenyl]-1-heptylureido)ethyl]phenoxy)-2-methyl-butyric acid (GW-2331), 4-(4-((2S,5S)-5-(2-(bis(phenylmethyl)amino)-2-oxoethyl)-2-heptyl-4-oxo-3-thiazolidinyl)butyl)-benzoic acid (GW-0072), 5-(2,4-dioxothiazolidin-5-ylmethyl)-2-methoxy-N-(4-trifluoromethylbenzyl)-benzamide (KRP-297), and [4-[3-(4-acetyl-3-hydroxy-2-propylphenoxy)-propoxyl]phenoxy]-acetic acid (L-165041) were synthesized by the Medicinal Chemistry Division of Pierre Fabre Research Center. Pioglitazone was purified from the medicine Actos (Takeda Pharmaceuticals, Japan). Fenofibrate, indomethacin, and diclofenac sodium were obtained from Sigma (St. Quentin, France). All compounds were dissolved in DMSO, and successive dilutions were made to obtain a final maximal DMSO concentration of 0.1%.

Results

Homologous Cloning of the Human Fatty Acid-Binding Protein-4 Gene Promoter. The promoter of the mouse aP2 gene has been identified as containing several cis- and trans-acting elements (Graves et al., 1992) as well as PPAR response elements (Elbrecht et al., 1996). The human ortholog was cloned by homologous cloning: a gene encoding human acidic FABP4 (or aP2) has been retrieved from the public databases (GenBank accession number J02874; Baxa et al., 1989), and it contained, besides the entire coding sequence, 63 bp of the 5′ untranslated region. The entire sequence J02874 was used in a BLAST search against all publicly available human sequences. A high BLAST score (398 bits, E-value: 1.0 10^–108) was obtained for entry AC018616, a bacterial artificial chromosome of human chromosome 8 containing clone RP11-157I4 (Whitehead Institute Center for Genome Research). This location is in accordance with the chromosomal mapping of human FABP4 to position 8q21 (Prinsen et al., 1997). Four exons of haP2 were localized within the AC018616 sequence (Table 1). The exon/intron junctions corresponded to those previously identified for the mouse aP2 gene (Phillips et al., 1986). By analogy to the 5400-bp-long aP2 promoter identified in mice (Graves et al., 1992), an equivalent human sequence was PCR-amplified from human genomic DNA, and its promoter activity was evaluated in a luciferase reporter-gene assay. Four nucleotide modifications were observed in the promoter sequence presented herein (deposited to the European Molecular Biology Laboratory database as AJ627200) as compared with the AC018616 entry (Fig. 1): an AA doublet (position 3808 in AJ627200) within a poly-A stretch of 13 adenosine residues and a CA doublet (position 5030 in AJ627200) located within a CA-dinucleotide stretch were missing.

Characterization of PPAR Modulators in GAL4 Reporter-Gene Assays. To compare the efficacy of PPAR agonists, a selection of ligands or their vehicle (0.1% DMSO) was incubated with COS-7 cells coexpressing chimeric receptors formed by the GAL4 DNA-binding domain and a human PPAR ligand-binding domain along with a GAL4-responsive reporter plasmid. Rosiglitazone transactivated PPARγ with an 8-fold higher potency than pioglitazone, and both behaved as rather selective PPARγ full agonists despite a significant transactivation of GAL4:PPARα by pioglitazone at the highest concentration tested (100 μM). The well known fenofibric acid and the novel compound BM-17.7044 induced a selective transactivation mediated by GAL4:PPARα with apparent potencies in the high micromolar range for both agents and a slightly lower efficacy for BM-17.0744 relative to fenofibric acid (Table 2). Although the non-subtype-selective PPARαγ ligands GW-2331 and KRP-297 elicited a maximal PPARα transactivation comparable with that of fenofibric acid, they displayed partial agonist properties in GAL4:PPARγ reporter-gene assays regarding both TZDs. The EC₅₀ values for the compound L-165041 were 211 ± 66 (obtained in a GAL4: PPARδ-specific set of experiments) and 8507 ± 2836 nM for GAL4:PPARδ and GAL4:PPARγ chimeric proteins, respectively; it displayed a partial efficacy (65 ± 11%) for GAL4: PPARγ and remained inactive for GAL4:PPARα (Table 2). The last three compounds tested, namely, GW-0072, indomethacin, and diclofenac, behaved as a subtype-selective PPARγ agonist with a partial efficacy between 35 and 50% relative to pioglitazone. The evaluation of the nonsteroidal anti-inflammatory drugs (NSAIDs) indomethacin and diclofenac on GAL4:PPAR reporter-gene assays was based on their previously reported low affinity for the γ isoform (Lehmann et al., 1997; Adamson et al., 2002).

Pharmacological Characterization of the Human aP2 Promoter. COS-7 cells were transiently cotransfected with mammalian expression vectors containing either full-length human PPARα, PPARγ1, or PPARγ2, along with the human aP2-Luc plasmid. All selected compounds that displayed substantial PPARγ-activating properties in the GAL4 reporter-gene assay elicited a concentration-dependent increase in luciferase expression/activity under the control of the aP2 promoter, whatever PPARγ1 (Fig. 2A) or PPARγ2 (Fig. 2B) coexpressed. In accordance with data obtained in GAL4 reporter-gene assays, GW-0072 and indomethacin also displayed a low efficacy in this human aP2 gene promoter-driven reporter assay, whereas diclofenac remained ineffective. The PPARα ligands fenofibric acid and BM-17.0744 remained inefficient not only in the full-length PPARγ-aP2 reporter-gene assays, which was in accordance with their PPAR subtype-binding selectivity, but were also inactive in the PPARα-aP2 reporter-gene assay, thus demonstrating the specificity of the interaction between the PPARγ DNA-binding domains and this functional promoter aP2 (Fig. 2; Table 3). In contrast to previous results describing the activation of a common “consensus” PPRE by all full-length PPAR subtypes (Wurch et al., 2002), the present data suggest that the 5400-bp fragment upstream of the putative transcription start of the human FABP4 mRNA can be used in a reportergene assay selective for both PPARγ1 and PPARγ2 but not PPARα. Moreover, besides the TZDs rosiglitazone, pioglitazone, and KRP-297, which induced a maximal receptor activation, the maximal efficacy (E_max) achieved by the nonsubtype-selective agent GW-2331 relative to rosiglitazone was about 85% and 99%, respectively, for PPARγ1 and PPARγ2. L-165041 also elicited a concentration-dependent increase in luciferase activity with a maximal activity comparable to that of rosiglitazone, but with a lower potency (Table 3). No significant difference between PPARγ1 and PPARγ2 was detected with L-165041.

To further define the putative relationship between the activation profiles of the selected compounds on chimeric versus full-length PPARγ proteins, aP2-dependent reporter-gene assays were performed in COS-7 cells that were treated for 24 h simultaneously with a constant and maximally effective concentration of rosiglitazone (100 nM) and increasing concentrations of L-165041 up to 30 μM. This agent reduced rosiglitazone-stimulated PPARγ transcriptional activity in a concentration-dependent manner, and to a larger extent in PPARγ2 cells (41 ± 4% reduction in PPARγ1; 48 ± 9% reduction in PPARγ2); nevertheless the efficacy between rosiglitazone and L-165041 was almost comparable during this head-to-head comparison. Hence, the well known PPARγ partial agonist GW-0072 was investigated (PPARγ1, Fig. 3A; or PPARγ2, Fig. 3C), and it elicited a concentration-dependent inhibition of rosiglitazone-mediated luciferase expression in this aP2 promoter controlled cell-based assay (Fig. 3, B and D for, respectively, PPARγ1 and PPARγ2). Beyond PPARγ reporter-gene assays, so-called “physiological experiments” on adipocyte differentiation were carried out to determine whether a relationship between these in vitro models may exist, at least for this PPAR modulator selection.

Effects of PPARs on 3T3-L1 Adipocytes. Confluent murine 3T3-L1 preadipocytes differentiated into adipocytes upon treatment with dexamethasone, insulin, and isobutylmethylxanthine. Adipogenesis was stimulated in a concentration-dependent manner by rosiglitazone, pioglitazone, GW-2331, and L-165041, with regard to the up-regulation of maP2 mRNA after a 3-day treatment period; the threshold concentrations for these compounds were 0.1 and 1 μM, respectively, for TZD and non-TZD agents. By contrast, the selective PPARα activator BM-17.0744 remained ineffective, as did indomethacin and diclofenac. A full range of apparent intrinsic activities was observed with the highest concentration of ligands tested, which displayed some PPARγ agonist properties (Fig. 4A). The level of maP2 mRNA appeared to be related to the development of the adipocyte phenotype, and it was used as a rather early and sensitive marker of adipogenesis in the present study.

Nevertheless, since many other genes and regulators may play key roles in this process, the accumulation of neutral lipid content was monitored in longer experiments (7 days) using AdipoRed reagent (Fig. 4B). Rosiglitazone and pioglitazone elicited the largest concentration-dependent lipid accumulation. All other compounds except BM-17.0744 were able to induce a concentration-dependent lipid accumulation, albeit with a different efficiency because of less adipogenic potential than TZDs (Fig. 4B). During the course of this assay, the highest concentration of BM-17.0744 (10 μM) did not appear to inhibit differentiation, but lipid accumulation, if any, was marginally increased (Fig. 4B).

Photomicrographs illustrated the rosiglitazone-dependent accumulation of intracellular lipids (Fig. 5, C and D), as monitored by AdipoRed staining. By contrast, BM-17.0744-treated 3T3-L1 cells (Fig. 5B) remained comparable with differentiated control cells (Fig. 5A).

Discussion

This report presents the cloning strategy of the human adipocyte fatty acid-binding protein (or aP2) gene promoter and its functional characterization based on human PPAR reporter-gene pharmacological assays. A key result of this study appears to be the selective activation of this promoter by liganded PPARγ of both PPARγ1 and PPARγ2 isoforms, since PPARα remained unable to mediate any aP2-driven reporter-gene expression upon stimulation by either a selective PPARα agonist (BM-17.0744) or a dual PPARαγ agonist (KRP-297). These data are in accordance with the properties of PPARγ agonists on adipocyte differentiation and the up-regulation of murine aP2 both in vitro and in vivo (Lehmann et al., 1995; Henke et al., 1998). The lack of effect PPARα was not related to technical issues since a positive transactivation signal was obtained with both BM-17.0744 and KRP-297 when COS-7 cells were transiently cotransfected with full-length PPARα and consensus PPRE_3x-HSV-Luc expression vectors (data not shown); positive non-subtype-selective interactions between ligand-bound PPAR (α for instance) were also previously shown in reporter-gene assays conducted with the same expression vectors but in HepG2 recipient cells (Wurch et al., 2002). Hence, besides 1) the tissue distribution of the various PPAR isoforms, 2) the specific physiological programs relative to a tissue, 3) the availability of cofactors required for ligand-dependent transcription factor activity, and 4) the affinity of ligands for PPAR ligand-binding domains, one can speculate that the specific interaction between the PPAR-DNA-binding domain and responsive element(s) is of key relevance for the regulation of certain signaling pathways.

The selection of various PPAR modulators was made according to their reported potency and efficacy on the different human PPAR subtypes. Nevertheless, their respective pharmacological properties were established under our experimental conditions with chimeric GAL4:PPAR reporter-gene assays. A rather wide range for GAL4:PPARγ activation was observed with rosiglitazone and pioglitazone (full agonists), KRP-297, GW-2331, and L-165041 (non-subtype-selective agonists), GW-0072, indomethacin, and diclofenac (partial PPARγ agonists), and BM-17.0744 and fenofibric acid (almost inactive), in accordance with previous results (Lehmann et al., 1997; Oberfield et al., 1999; Willson et al., 2000; Adamson et al., 2002). Although they are artificial, GAL4: PPAR ligand-binding domain-related assays are often used to further characterize the agonist efficacy, but their putative relationship in a comparable model with the wild-type receptors and full promoter activities is sometimes lacking (Berger et al., 2003) or not related to the corresponding physiological function, namely, adipogenesis (Reginato et al., 1998; Camp et al., 2000). Hence the cloning of the human adipocyte fatty acid-binding protein gene promoter was performed by homologous cloning, starting from the coding sequence of the human aP2 gene and retrieving a sequence with a genomic organization similar to that described for the mouse aP2 gene (Phillips et al., 1986). Genomic location on human chromosome 8, the homolog of mouse chromosome 3 containing the aP2 gene (Heuckeroth et al., 1987), a 64-nt-long 5′ untranslated region, and exon/intron junctions similar to those described for the mouse aP2 gene (Phillips et al., 1986), were identified. The functionality of this putative promoter sequence was assayed upon fusion upstream to a luciferase reporter gene.

Data obtained from both chimeric and full-length protein reporter-gene assays suggest that a discrepancy may exist regarding the efficacy of so-called partial agonists for PPARγ (e.g., L-165041 and diclofenac). The former elicited a maximal response comparable with that of rosiglitazone in wild-type/human aP2 reporter-gene assays, although it is partial with the PPARγ ligand-binding domain fused to a heterologous GAL4 DNA-binding domain, as was also reported for another compound, MCC-555 (isaglitazone) (Reginato et al., 1998); the latter behaved as a partial agonist in the chimeric assay, whereas it was silent in the PPARγ full-length/human aP2 assay. These data fit with those reported by Adamson et al. (2002) showing that diclofenac antagonizes the PPARγ full-length reporter-gene assay and suggest that apparent efficacy differs from one pharmacological model to another. As a consequence, since the interdomain relationship in PPAR is likely crucial for its own activation, it appears quite difficult to characterize the so-called “intrinsic efficacy” of a PPAR ligand, which can also vary upon cofactor recruitment. Nevertheless, data related to other PPARγ agonists (TZDs, GW-0072, and indomethacin) suggest a coherence between the GAL4 and wild-type/human aP2 procedure. The relevance of this model (reporter-gene assay controlled by the human aP2 promoter) for characterizing the putative antagonistic properties of a PPARγ partial agonist is strengthened by the results obtained with GW-0072 added simultaneously with rosiglitazone, even if an antagonistic effect down to its partial intrinsic efficacy could not be achieved under our experimental conditions. GW-0072 was previously characterized with chimeric PPAR proteins as a partial agonist/antagonist for rosiglitazone (Oberfield et al., 1999).

Despite slight differences in the efficacy of GW-2331 for instance (85 ± 5% and 99 ± 5% for PPARγ1 and PPARγ2, respectively), one cannot speculate upon the putative selectivity of a ligand toward PPARγ1 relative to PPARγ2, and no clear-cut interpretation can be drawn regarding the adipogenic potential of a compound relying only on the presence of an additional 30 amino acids at the amino terminus DNA-binding domain of PPARγ2, which, moreover, are located outside the ligand-binding domain. Even when PPARγ2 activation was proposed to play a unique role in adipogenesis (Ren et al., 2002), it was further demonstrated that PPARγ1 can also drive the differentiation of fat cells, albeit with a lesser sensitivity, depending on both isoform expression level and ligand concentration (Mueller et al., 2002).

Not only a peculiar distribution of a compound but also a specific interaction profile with cofactors may modulate PPAR activation and, putatively, its adipogenic effect (Berger et al., 2003). The in vitro physiological significance of data from reporter-gene assays was addressed by studying the adipogenic potential of some PPAR modulators in 3T3-L1 adipocytes, a well documented (albeit murine) model. GW-2331 and L-165041 showed a less potent and efficacious up-regulation of maP2 mRNA than did rosiglitazone and pioglitazone after 3 days of treatment, which led to a lower accumulation of intracellular neutral lipids after 7 days. GW-0072 and the NSAIDs indomethacin and diclofenac behaved as low lipogenic compounds, as demonstrated in aP2 mRNA and lipid accumulation experiments. The adipogenic properties of these compounds were only observed at the highest concentration (efficacy) and remained quite marginal at lower concentrations. As was foreseeable, a treatment by the PPARα agonist BM-17.0744 resulted in a low increase in maP2 mRNA as compared with the control situation, which was associated with a very low level of neutral lipids.

Altogether, these data suggest that a positive relationship can exist between these in vitro pharmacological and physiological assays despite some specific compound behaviors; even when they are not totally parallel, reporter-gene assays constitute a more rapid and sensitive procedure than adipocyte differentiation to identify agents with low adipogenic properties. These results on the functionality and specificity of the human aP2 promoter represent an interesting issue for addressing part of the lipogenesis signaling pathway and warrant future investigations, at least to determine the nature of DNA-binding domain/promoter interactions, for instance. Efficacy but also potency in full-length PPAR/aP2 reporter-gene assays constitutes a “key” predictive parameter, and subtle differences between compounds within a common chemical series should be tested to further strengthen the validity of the model. Other target genes for PPARγ, such as resistin (Way et al., 2001), adiponectin (Maeda et al., 2001), or aquaporin (Kishida et al., 2001), could also be evaluated following a comparable strategy to characterize, at least in vitro and with a larger predictability, attractive selective PPAR modulators.