L-764406 Is a Partial Agonist of Human Peroxisome Proliferator-activated Receptor γ

Insulin-sensitizing thiazolidinedione (TZD) compounds are high affinity ligands for a member of the nuclear receptor family, peroxisome proliferator-activated receptor (PPAR) γ. A scintillation proximity assay for measurement of3H-radiolabeled TZD binding to human PPARγ under homogeneous conditions was developed. Using this approach, a novel non-TZD compound (L-764406) was shown to be a potent (apparent binding IC50 of 70 nm) PPARγ ligand. Preincubation of PPARγ with L-764406 prevented binding of the [3H]TZD, suggesting a covalent interaction with the receptor; in addition, structurally related analogues ofL-764406, which would be predicted not to interact with PPARγ in a covalent fashion, did not displace [3H]TZD binding to PPARγ. Covalent binding of L-764406 was proven by an observed molecular weight shift of a tryptic PPARγ ligand binding domain (LBD) peptide by mass spectrometric analysis. A specific cysteine residue (Cys313 in helix 3 of hPPARγ2) was identified as the attachment site for this compound. In protease protection experiments, the liganded receptor adopted a typical agonist conformation. L-764406 exhibited partial agonist activity in cells expressing a chimeric receptor containing the PPARγ LBD and a cognate reporter gene and also induced the expression of the adipocyte-specific gene aP2 in 3T3-L1 cells. In contrast,L-764406 did not exhibit activity in cells transfected with chimeric receptors containing PPARα or PPARδ LBDs. The partial agonist properties of L-764406 were also evident in a co-activator association assay, indicating that the increased transcription in cells was co-activator mediated. Thus,L-764406 is a novel non-TZD ligand for PPARγ and is also the first known partial agonist for this receptor. The results suggest a critical functional role for Cys313, and helix 3, in contributing to ligand binding and subsequent agonist-induced conformational changes.

gand-activated transcription factors (1,2). They are currently divided into three subtypes: PPAR␣, PPAR␦ (also known as hNUC1 or PPAR␤), and PPAR␥, with each being encoded by a distinct gene. The presence of at least two promoters in the 5Ј-flanking region of PPAR␥ results in the production of two isoforms, ␥1 and ␥2 (3), where ␥2 possesses an additional 30 residues at its amino terminus (4). Like other members of this superfamily, the PPARs exert their effects by regulating gene transcription and all three members bind to DR1 response elements (reviewed in Ref. 1).
Although the three subtypes have been grouped together based on sequence homology, it is clear that they have distinct functions. PPAR␣ is expressed predominantly in the liver and is involved in peroxisome proliferation and regulation of fatty acid catabolism (5). PPAR␥ plays a pivotal role in fat cell differentiation and lipid storage (6), while PPAR␦ is expressed in most cell types but its role remains unclear (7). Consistent with their distinct physiological roles, each receptor has been shown to bind a discrete set of ligands, although, in general, these receptors seem to be regulated by fatty acids and eicosanoids (8 -10). Thus, for PPAR␣, selected fatty acids serve as potential natural ligands; in addition, the fibrate class of hypolipidemic drugs and a group of structurally diverse peroxisome proliferators can also function as PPAR␣ ligands and/or activators. The synthetic thiazolidinedione insulin sensitizers (11) and the prostaglandin derivative 15-deoxy-⌬ 12,14 -prostaglandin J 2 bind and activate PPAR␥ (12,13), while several polyunsaturated fatty acids and eicosanoids can serve to activate PPAR␦ (7,9).
These receptors play a central role in lipid homeostasis where imbalances can lead to cardiovascular disease, obesity, and diabetes. Indeed, medications targeted to PPARs have been demonstrated to be effective treatments for hyperlipidemia (fibrates, Ref. 14), and insulin resistance (thiazolidinediones, Ref. 15). Thus, there has been considerable interest in developing new and specific ligands for these receptors (16,17). It is, therefore, clear that further characterization of the spectrum of natural or synthetic molecules, which can function as PPAR ligands and agonists is an important undertaking. Moreover, a determination of the precise residues in the PPAR ligand binding domains (LBDs), which make contact with agonist ligands would greatly facilitate the subsequent discovery of new and therapeutically useful modulators of these receptors.
Here, we used a radiolabeled thiazolidinedione and recombi-nant human PPAR␥ to develop a scintillation proximity assay (SPA) in order to characterize ligands that bind this receptor. Using this approach, a novel non-TZD compound, L-764406, was shown to be a high affinity PPAR␥ ligand. Several approaches were used to demonstrate that this compound acts as a specific agonist for PPAR␥. Importantly, L-764406 was also shown to bind covalently to PPAR␥. This phenomenon was exploited in order to determine that the compound bound directly to Cys 313 in helix 3 of the LBD of human PPAR␥2. The identification of Cys 313 as the attachment site for L-764406 defines an important role for this residue and for helix 3 in ligand binding and activation of PPAR␥.

EXPERIMENTAL PROCEDURES
Preparation of Recombinant PPAR␥-The 1.5-kilobase pair PPAR␥ insert was released from the pCRII cloning vector by digestion with the restriction enzyme SmaI (4) and ligated into the SmaI site of the bacterial expression vector pGEX-4T-2 (Amersham Pharmacia Biotech) to produce the vector pGEX-hPPAR␥2 containing the full-length human PPAR␥2 cDNA fused to glutathione S-transferase. Escherichia coli BL-21 cells were transformed with pGEX-hPPAR␥2 plasmid DNA. Cells were cultured and induced with isopropyl-␤-D-thiogalactopyranoside as described by the supplier. The cells were pelleted by centrifugation for 20 min at 2,000 rpm and the pellet was resuspended in 30 ml of phosphate-buffered saline containing 0.25 mM phenylmethylsulfonyl fluoride/1,000 ml of culture medium. Two passes through a French press were used to disrupt the cells, and cellular debris was removed by centrifugation at 10,000 rpm, 4°C. Recombinant GST-hPPAR␥ was isolated batchwise using glutathione-Sepharose as described by the supplier (Amersham Pharmacia Biotech). Typically, 2 mg of GST-hP-PAR␥ at approximately 50% purity was obtained per liter of bacterial culture.
SPA Binding Assay-The binding assay was developed for use with microtiter plates (Dynex Technologies, catalog number 011-010-7905) using a total volume of 100 l of assay buffer: 10 mM Tris-Cl, pH 7.2, 1 mM EDTA, 10% (w/v) glycerol, 10 mM sodium molybdate, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 2 g/ml benzamidine, and 0.1% dry milk powder. Each bottle of protein A-yttrium silicate SPA beads (Amersham Pharmacia Biotech catalog number RPN143) was suspended in 25 ml of assay buffer but omitting the dry milk powder and adding sodium azide to a final concentration of 0.01%. [ 3 H 2 ]5-[4-[2-(5-metyl-2-phenyl-4-oxazolyl)-2-hydroxyethoxy]benzyl]-2,4-thiazolidinedione ([ 3 H]TZD), 21 Ci/mmol, was dissolved in ethanol and used at a final concentration of 10 nM. The recombinant GST-hPPAR␥ preparations were usually used at a dilution of 800 ϫ producing a final concentration of approximately 5 nM. Goat anti-GST antibodies were obtained from Amersham Pharmacia Biotech (catalog number 27-4577-01) and used at a 400-fold final dilution. The GST-hPPAR␥, goat anti-GST antibodies, and [ 3 H]TZD were diluted in assay buffer and combined in a total volume of 70 l in the microtiter plate. Five l of test compound was added so that the final concentration of Me 2 SO did not exceed 2%. Following the addition of 25 l of protein A-yttrium silicate SPA beads to each well, the plate was incubated at 15°C for 24 h with shaking. Radioactivity was quantified in a Packard Topcount scintillation counter.
Plasmids-Chimeric receptors containing the yeast GAL4 DNA binding domain fused to either human PPAR␣, PPAR␦, or PPAR␥ were created by insertion of a BamHI/HindIII fragment from pFC DNA binding domain (Stratagene) encoding the GAL4 DNA binding domain into the same sites within the mammalian expression vector pcDNA3.1(ϩ) (Invitrogen) to generate the vector pcDNA3.1-GAL4. The locations of the LBDs for each PPAR receptor were determined by sequence alignment programs from the Wisconsin Sequence Analysis Package (18). The fragments were generated by polymerase chain reaction using appropriate primers, which provided polymerase chain reaction products flanked by BamHI and NotI sites at their 5Ј and 3Ј ends, respectively. These fragments were digested with BamHI and NotI and ligated into the vector pcDNA3.1-GAL4, which had been digested with the same enzymes. The fragments begin at amino acid 167 for PPAR␣ (GenBank TM accession number L02932), 139 for PPAR␦ (GenBank TM accession number L07592), and 203 for PPAR␥2 (Gen-Bank TM accession number U63415), and extend to the COOH terminus for each receptor. To confirm accuracy, the DNA sequence for each construction was determined. The reporter plasmid for these GAL4 chimeric receptors (pUAS(5ϫ)-tk-luc) contains five repeats of the GAL4 response element (UAS) upstream of a minimal thymidine kinase pro-moter that is adjacent to the luciferase gene. The control vector, pCMV-lacZ, contains the CMV promoter adjacent to the galactosidase Z gene.
Cell Culture and PPAR Transactivation-COS-1 cells were seeded at 12 ϫ 10 3 cells/well in 96-well cell culture plates in high glucose Dulbecco's modified Eagle's medium containing 10% charcoal-stripped fetal calf serum (Gemini Bio-Products), nonessential amino acids, 100 units/ml penicillin G, and 100 mg/ml streptomycin sulfate at 37°C in a humidified atmosphere of 10% CO 2 . After 24 h, transfections were performed with LipofectAMINE (Life Technologies, Inc.) according to the instructions of the manufacturer. Briefly, transfection mixes for each well contained 0.48 l of LipofectAMINE, 0.00075 g of pcDNA3-GAL4/PPAR expression vector, 0.045 g of pUAS(5ϫ)-tk-luc reporter vector, and 0.0002 g of pCMV-lacZ as an internal control for transfection efficiency. Cells were incubated in the transfection mixture for 5 h at 37°C in an atmosphere of 10% CO 2. The cells were then incubated for ϳ48 h in fresh high glucose Dulbecco's modified Eagle's medium containing 5% charcoal-stripped fetal calf serum, nonessential amino acids, 100 units/ml penicillin G, and 100 mg/ml streptomycin sulfate Ϯ increasing concentrations of test compound. Since the compounds were solubilized in Me 2 SO, control cells were incubated with equivalent concentrations of Me 2 SO; final Me 2 SO concentrations were ϭ 0.1%, a concentration that was shown not to affect transactivation activity. Cell lysates were produced using Reporter Lysis Buffer (Promega, Madison, WI) according to the manufacturer's instructions. Luciferase activity in cell extracts was determined using Luciferase Assay Buffer (Promega) in an ML3000 luminometer (Dynatech Laboratories). ␤-Galactosidase activity was determined using ␤-D-galactopyranoside (Calbiochem) as described previously (19).
Measurement of aP2 mRNA-Confluent 3T3-L1 cells were incubated in Dulbecco's modified Eagle's medium with 10% fetal calf serum, 100 units/ml penicillin G, 100 g/ml streptomycin sulfate, 1 M dexamethasone, 150 nM insulin, in the absence or presence of increasing concentrations of test compound for 4 days at 37°C in 5% CO 2 (with one medium change). Total RNA was prepared from cells using the Biotecx Ultraspec TM RNA isolation kit, and RNA concentration was estimated from absorbency at 260 nM. RNA (20 g) was denatured in formamide/ formaldehyde and slot blotted onto Hybond TM -N membrane. Prehybridization was performed at 42°C for 1-3 h in 50% formamide and Thomas solution A containing 25 mM sodium phosphate, pH 7.4, 0.9 M sodium chloride, 50 mM sodium citrate, 0.1% each of gelatin, Ficoll, and polyvinylpyrollidone, 0.5% SDS, and 100 g/ml denatured salmon sperm DNA. Hybridization was carried out at the same temperature for 20 h in the same solution with a 32 P-labeled aP2 cDNA probe (2 ϫ 10 6 cpm/ml). After washing the membranes under appropriately stringent conditions, the hybridization signals were analyzed with a PhosphorImager (Molecular Dynamics).

Assessment of Receptor Conformation by Partial Protease Digestion-
The protease digestion assays were performed by the method of Allan et al. (20), with previously described modifications (21). The pSG5-hP-PAR␥ plasmid was used to synthesize 35 S-radiolabeled PPAR␥, in a coupled transcription/translation system according to the protocol of the manufacturer (Promega). The transcription/translation reactions were subsequently aliquoted into 22.5-l volumes, and 2.5 l of phosphate-buffered saline Ϯ compound were added. These mixtures were incubated for 20 min at 25°C, separated into 4.5-l aliquots, and 0.5 l of distilled H 2 O or distilled H 2 O-solubilized trypsin were added. The protease digestions were allowed to proceed for 10 min at 25°C, then terminated by the addition of 95 l of denaturing gel loading buffer and boiling for 5 min. The products of the digestion were separated by electrophoresis through a 1.5-mm 4 -20% polyacrylamide-sodium dodecyl sulfate gel. After electrophoresis, the gels were fixed in 10% acetic acid (v/v):40% methanol (v/v) for 30 min, treated in EN 3 HANCE for a further 30 min and dried under vacuum for 2 h at 80°C. Autoradiography was then performed to visualize the radiolabeled digestion products.
Mass Spectrometry-Liquid chromatography-electrospray ionization-mass spectrometry (LC-ESI-MS) analysis was performed using a Finnigan TSQ7000, as described previously (22). Mass measurement of receptor-ligand complex was performed as follows. Samples were loaded on a C4 reverse phase column (1 ϫ 100 mm) at a flow rate of 100 l/min with 0.075% aqueous trifluoroacetic acid and eluted with a gradient of 2-60% acetonitrile over 40 min. The effluent was fed directly to the ESI interface of the mass spectrometer. Ions were detected throughout the entire LC gradient over a m/z (mass to charge ratio) range of 500 -2,000. Receptor-ligand complex was digested with trypsin (sequence grade, Boehringer Mannheim), and the peptide fragments were analyzed by LC-ESI-MS and LC-ESI-tandem mass spectrometry (LC-ESI-MS/MS) on a Finnigan LCQ.

Construction of C313A PPAR␥ Mutation-Cys 313 of hPPAR␥ is located within a 72-base pair restriction fragment bracketed by unique
MscI and BsmI sites. The C313A point mutation was made by synthesizing complimentary oligonucleotides containing the appropriate codon and anticodon for Ala at position 313 and flanked with MscI and BsmI sites. The complimenary oligonucleotides were hybridized at room temperature in restriction enzyme buffer and digested with the enzymes MscI and BsmI. The digested oligonucleotides were used to replace the same fragment in the vector pGEX-hPPAR␥2. The identity of the mutation was confirmed by DNA sequencing. Mutant PPAR␥ receptor protein was expressed in bacteria and purified as described above. Binding activity was determined using the SPA assay described above.
Co-activator Association Assays-A homogeneous time-resolved fluorescence assay (HTRF) was used to examine the interaction of PPAR␥ and the mutant receptors with the co-activator CBP (CREB-binding protein). A complete description of this assay has been published elsewhere (23); briefly, 198 l of reaction mixture (100 mM HEPES, 125 mM KF. 0.125% (w/v) CHAPS, 0.05% dry milk, 1 nM GST-PPAR␥ LBD or 5 nM GST-PPAR␥, 2 nM anti-GST-(Eu)K, 10 nM biotin-CBP 1-453 , 20 nM SA/XL665) were added to each well, followed by addition of 2 l of test compound or vehicle (Me 2 SO) in appropriate wells. Plates were mixed by hand and covered with TopSeal. The reaction was incubated overnight at 4°C, followed by measurement of fluorescence reading on a Discovery instrument (Packard). Data were expressed as the ratio, multiplied by a factor of 10 4 , of the emission intensity at 665 nm to that at 620 nm.

RESULTS
Characterization of the SPA Assay-The major advantage of the scintillation proximity assay over other approaches for measuring ligand binding is that it is a single step, homogeneous assay format, so there is no need to separate bound isotope from free. This technical innovation was achieved through the development of beads impregnated with scintillant. Once the receptor is attached to the bead, and ligand is bound, they are sufficiently close to allow the ␤ emission from the tritium to be absorbed by the scintillant which will then shift this energy to produce light (Fig. 1). ␤ emissions from unbound tritiated li-gand will dissipate in the buffer. The background in the assay, as determined by the number of counts/min obtained in the presence of 100-fold excess unlabeled TZD, was less than 50 cpm. At a concentration of 10 nM [ 3 H]TZD, the total counts/min in the assay was approximately 1,000, providing a 20-fold window of specific binding activity. Counting efficiency for this assay is difficult to determine, since it would be necessary to know the efficiency of all coupling reactions to the bead, but it is estimated to be approximately 50% by the manufacturer.
We used increasing concentrations of [ 3 H]TZD in the SPA assay to produce a saturation curve (Fig. 2A). These data were regraphed in the form of a Scatchard plot (24), which demonstrates a single population of binding sites with a K d of 11 nM (Fig. 2B). From the number of binding sites obtained from the Scatchard plot, and since the GST-hPPAR␥ preparation was approximately 50% pure, as determined by SDS-gel electrophore- sis, we estimate that Ͼ20% of the protein was active in binding.
Known thiazolidinedione PPAR␥ ligands, including TZD, BRL49653, and CS-045 (troglitazone), were titrated in the SPA assay. IC 50 values were determined to be 13, 314, and approximately 1,700 nM, respectively (data not shown). These IC 50 values cover 2 orders of magnitude and agree well with previously published binding activities determined using dextran/ gelatin-coated charcoal to separate bound from free ligand (4). Furthermore, the rank order of these compounds is reflected in both transactivation assays and in their in vivo glucose lowering activity (4).
L-764406 Is a Novel PPAR␥ Ligand-Using the SPA assay approach, L-764406 was shown to be a potent PPAR␥ ligand. Thus, titration of this compound revealed an apparent IC 50 of 70 nM (results not shown). Compared with known synthetic thiazolidinedione ligands for PPAR␥, the molecular structure and relatively small size of L-764406 (Table I) suggested that this compound might interact covalently with the receptor. This hypothesis was supported by additional results demonstrating that a structurally related deschloro compound, shown in Table I, did not bind to PPAR␥. In particular, the O-methyl derivative, L-273422, would occupy a similar volume to L-764406, thereby indicating that the bulk provided by the chlorine residue at this position is not a critical determinant for binding.
[ 3 H]TZD Cannot Displace L-764406 in the SPA Assay-The SPA assay was used to determine whether a TZD could displace L-764406 from PPAR␥. The experiment was designed to also ensure that PPAR␥ binding activity could still be detected after 48 h. Thus, recombinant receptor was incubated with 10 nM [ 3 H]TZD and submaximal amounts of either L-764406 or unlabeled TZD. After 24 h, the amount of [ 3 H]TZD was increased to 50 nM. Under these conditions, the amount of specific binding should increase with time as the system progresses toward equilibrium ( Fig. 2A), but only if binding sites are still available. As seen in Fig. 3A, this only occurs with the samples preincubated with either 15 or 20 nM TZD. Those samples preincubated with either 150 or 300 nM L-764406 did not exhibit an increase in specific counts, suggesting that there were no free binding sites available to be occupied by the increased amount of [ 3 H]TZD. This would be expected if L-764406 was a covalent ligand, since binding equilibrium would not be achieved. These results were confirmed by preincubation of PPAR␥ with L-764406 at a maximally effective concentration of 4 M (Fig. 3B). Under these conditions, no [ 3 H]TZD binding activity could be detected after 24 h, while a second sample incubated in parallel with vehicle alone retained binding activity.
Identification of the Binding Site for L-764406 -Electrophoretically pure hPPAR␥ LBD (amino acid residues 204 -505, GenBank TM accession number U63416) was used to identify the specific binding site for L-764406 on the receptor. This fragment includes all residues from the hinge domain to the COOH terminus of the receptor as well as 2 residues, glycine and serine, which are encoded at the NH 2 terminus after insertion of a BamHI site. The purified fragment was incubated with L-764406 or vehicle alone and analyzed by high performance liquid chromatography electrospray ionization mass spectrometry (LC-ESI-MS) using a C8 1.0 ϫ 50-mm column. The results indicated that all PPAR␥ protein in the reaction formed a complex with L-764406, since no free PPAR␥ protein was observed in the mass spectrum (Fig. 4A). To identify the binding site of L-764406, a tryptic digest was performed using a sample of PPAR␥ that was preincubated with L-764406. Mass mapping of the digest by LC-ESI-MS revealed a single peptide with a size of 1,282 Da that was 283 Da greater than the predicted tryptic fragment (Table II). This molecular mass difference corresponds exactly to the additional mass provided by the compound (with the concomitant loss of HCl), indicating covalent binding of L-764406 to this peptide fragment of PPAR␥. LC-ESI-MS/MS was performed on this peptide. Analysis of the resultant MS/MS spectrum revealed that L-764406 was linked to the single cysteine residue within this fragment ( Fig. 4B and Fig. 5). This cysteine corresponds to Cys 313 in the sequence of full-length hPPAR␥2 (4) and is located within helix 3 of the receptor's LBD.
Mutation of Cys 313 Prevents Binding of L-764406 to hPPAR␥-Recombinant GST-hPPAR␥ protein with a C313A point mutation was purified as described above and used in the SPA assay. As expected, L-764406 (10 M) did not compete with the [ 3 H]TZD for binding to the mutated receptor (data not shown). Binding activity for TZD was also greatly reduced so that the number of specific cpm was decreased by approximately 75% (data not shown) and with a rightward shift in the IC 50 (Fig. 6). Thus, Cys 313 is necessary for binding of L-764406 and dramatically reduces the binding of TZD as well.
L-764406 Is a Partial PPAR␥ Agonist-Several experiments were performed to determine whether L-764406 could function as a PPAR␥ agonist. It has been shown for several members of the nuclear receptor family that binding of agonist ligands induces a conformational change in the LBD. In the case of estrogen receptor ␣, a major component of this conformational change has been shown to involve folding of the AF-2 domain into the core of the LBD, which results in a more compact structure (25). This conformational change is reflected by the increased resistance of the receptor LBD to partial digestion by proteases. Furthermore, distinct protease digestion patterns can be identified upon binding of agonists versus antagonists (25). A typical agonist-like protease protection pattern (21) was obtained when L-764406 was bound to hPPAR␥ (Fig. 7). The most obvious feature of this pattern is the major 27-kDa core fragment, which could also be seen with the thiazolidinedione ligand BRL49653 (4), indicating that L-764406 induces an agonist-like conformational change in PPAR␥.
When the GAL4/hPPAR␥ Chimeric Receptor and the Reporter Construct, 5ϫUAS-TK-luc, were cotransfected into COS-1 cells, we found that L-764406 served to induce PPAR␥dependent transcription of the luciferase gene (Fig. 8). The compound was specific for the PPAR␥ LBD, since it failed to induce transcription mediated by either GAL4/hPPAR␣ or GAL4/hPPAR␦. The EC 50 for transcriptional activation of GAL4/hPPAR␥ in COS-1 cells was 69 nM, which is in good agreement with the IC 50 of 70 nM obtained with the SPA binding assay. Interestingly, maximal transcriptional activation attained using L-764406 was only 25% of that obtained with thiazolidinediones, indicating that this compound was a partial agonist under these conditions. The above findings were confirmed in experiments designed to measure the ability of endogenous wild type PPAR␥ to regulate a classic PPAR␥ adipocyte-specific target gene, adipose fatty acid-binding protein (aP2). Using murine 3T3-L1 preadipocytes, we assessed the ability of L-764406 to promote adipocyte differentiation as measured by induction of aP2 mRNA expression. As depicted in Fig. 9, treatment of 3T3-L1 preadipocytes with L-764406 resulted in a substantial increase in aP2 mRNA expression with an EC 50 value of between 100 and 1,000 nM. As seen in transfected COS-1 cells, this effect of L-764406 was submaximal, achieving approximately 25% of the highest activity obtained with TZD.
Since transcriptional activation is mediated through the interaction of nuclear receptors with co-activators (2), we used a co-activator association assay (23) to show that L-764406 induced binding of PPAR␥ to CBP. Preincubation of the hPPAR␥ LBD with 4 M L-764406 induced binding of the receptor to CBP, however, as with the transcriptional activation of aP2 gene in 3T3-L1 cells and the luciferase reporter in COS-1 cells, this activation was limited to approximately 25% of that seen with TZD (Fig. 10A). Furthermore, this activation could not be increased by subsequent addition of the more potent TZD, indicating that the L-764406 could not be displaced by TZD and that TZD was unable to activate hPPAR␥ when the binding pocket was occupied by this relatively small covalent agonist. Conversely, an excess of the high affinity TZD ligand can protect against the alkyating effect of L-764406 (Fig. 10A). As was expected from the results of the SPA binding assay (results are not shown), TZD was not able to promote the association of the C313A mutation of hPPAR␥ with CBP (Fig. 10B). DISCUSSION We developed a novel SPA-based binding assay for the nuclear receptor, PPAR␥. This is a simple assay that does not require separation of bound and free radiolabeled ligands. Us-  ing this approach, L-764406 was shown to function as a potent ligand for hPPAR␥. Based upon the chemical structure of this compound, we suspected that it might bind covalently to the PPAR␥ LBD. This concept was also suggested by the observa-tion that three closely related analogue compounds lacking the chlorine found in L-764406 were inactive in the PPAR␥ SPA binding assay. Experimental evidence in favor of this hypothesis includes the fact that preincubation of PPAR␥ with this compound abrogated the ability of a potent TZD ligand to displace L-764406 binding. Moreover, covalent binding of L-764406 to PPAR␥ was definitively proven by mass spectral analysis of PPAR␥ LBD incubated with this compound; a molecular mass increase in one peptide fragment was detected, which corresponded precisely to the molecular weight of L-764406, minus 35 Da for the loss of chlorine.
Although L-764406 completely displaced [ 3 H]TZD at concen- trations greater than 1 M, it was only capable of producing partial agonist activity at these concentrations when its ability to induce PPAR␥-mediated gene transcription was assessed. This was noted using either GAL4/hPPAR␥ chimeric receptors in COS-1 cells or in 3T3-L1 preadipocytes where wild type murine receptors are present and can serve to induce aP2 expression. In both cases, only 25% of the maximal activity obtained with a TZD was observed. This partial agonist profile is unique for a PPAR␥ ligand, since other known ligands including several thiazolidinedione insulin sensitizers (11,21) and 15-deoxy-⌬ 12,14 -prostaglandin J 2 (12, 13) function as full agonists in both transfected cells and pre-adipocytes.
Although the molecular basis for partial agonist activity is not well understood, it could be due to the production of a receptor conformation with a reduced affinity for co-activators.
McDonnell et al. (25) have suggested that the estrogen receptor can assume conformations distinct from those of the normal ligand-bound receptor when bound to ligands with varying agonist activities. Their studies also demonstrate that these activities are dependent on cellular and promoter contexts; thus, it is possible that L-764406 might behave as a full agonist under different experimental conditions. In protease protection experiments, L-764406 produced a trypsin digestion profile that was indistinguishable from the thiazolidinedione agonist BRL49653. Although estrogen receptor antagonists are known to produce a distinct protease protection pattern from that of 17␤-estradiol (20,25), we conclude that possible conformational differences responsible for the partial agonist activity of L-764406 may be too subtle for detection by the protease protection assays performed here.
The agonist activity of L-764406 is apparently specific for PPAR␥, since no activity was noted with GAL4 chimeric receptors containing either PPAR␣ or PPAR␦ LBDs. A multiple sequence alignment of the LBDs for hPPAR␥, hPPAR␣, and hPPAR␦ (Fig. 6) shows that the single cysteine in PPAR␥ (Cys 313 ) is conserved in the other two receptors. Although PPAR␥ has only one cysteine in the LBD, the PPAR␣ LBD contains two additional cysteine residues flanking the conserved cysteine, with one of these additional cysteines also being present in PPAR␦. Thus, despite the availability of potential attachment sites in the LBDs of hPPAR␣ and hPPAR␦, L-764406 does not activate these receptors (Fig. 5).
Cys 313 in the LBD of PPAR␥ is within a predicted helical domain 2 that corresponds to helix 3 as previously diagrammed in the crystal structures of the RAR␥ (26) and ER␣ (27) LBDs. Importantly, residues within helix 3 of the ER␣ LBD have been shown to interact with the A-ring of 17␤-estradiol (28), as well as the ER antagonist raloxifene (27). The specific residues of the ER␣ LBD that interact with 17␤-estradiol include Leu 345 , Thr 345 , and Glu 353 and are not conserved in hPPAR␥. Although a homologous cysteine has not been identified in the estrogen receptors, a cysteine corresponding to Cys 313 is conserved in the LBDs of human RXR␣, -␤, and -␥, but not in RAR␣, -␤, and -␥. In addition to providing contacts with its ligand, all-transretanoic acid, helix 3 from the LBD of RAR␥ forms intramolecular interactions with helix 12 and thus may be involved in the orientation of the AF-2 domain (26). This relationship suggests a pathway whereby ligand interactions at helix 3 are translated into interactions with co-activators through the AF-2 domain. Indeed, we have shown that the partial agonist L-764406 exhibits a diminished interaction with CBP (Fig. 10). Since this interaction could not be augmented by subsequent addition of a more potent agonist, TZD (Fig. 10), it suggests that despite the relatively small size of L-764406, the modified Cys 313 disrupts the ligand binding pocket sufficiently to prevent activation by TZD. We know that Cys 313 is not an absolute requirement for TZD binding since the C313A mutation is still active in this regard, albeit at a dramatically reduced level. Together these experiments suggest that the activity of L-764406 as a partial agonist is mediated by a limited (versus full agonists) interaction with co-activators. However, it should be noted that although L-764406 functions as a partial agonist in the cell-based transactivation assays (Figs. 8 and 9) and in the in vitro co-activator association assay (Fig. 10), it is possible that this compound could function as a full agonist in another context.
Although RARs do not posses a cysteine residue homologous to the one found in the LBD of PPAR␥, RAR␣ does have a serine residue at this position. In fact, the replacement of Ala 225 for Ser 232 is the only difference between the LBDs of RAR␤ and RAR␣ and has been shown to account for their ligand specificity (26). This shows that the residue at this position in helix 3 is involved in the determination of ligand specificity for other nuclear receptors.
Since the completion of our studies, Nolte et al. (29) reported their results where x-ray crystallography was used to determine the PPAR␥ LBD structure. Their findings indicate that Cys 313 is indeed located within helix 3 and that helix 3 forms an important component of the ligand binding pocket when occupied by the TZD, BRL49653. Flanking Cys 313 are two residues, Phe 310 and Gln 314 , which form part of a hydrophobic pocket occupied by the sulfur atom of the TZD ring in BRL49653 (29). Cys 313 is believed to form part of a narrow pocket occupied by the central benzene ring of BRL49653 (29). Our results obtained using a novel covalent ligand provide an independent line of evidence which shows the involvement of helix 3, and in our case, Cys 313 , in forming critical components of the PPAR␥ ligand binding pocket.
In summary, L-764406 was shown to possess agonist activity in cells and produced an LBD protease protection pattern that was similar to that caused by a known TZD; thus, it is likely that the interaction of ligands with helix 3 (and Cys 313 ) is important for the induction of conformational changes which mediate co-activator recruitment and activation of transcription. L-764406 is also unique among known PPAR␥ ligands in that it functions as a partial agonist (in co-activator association, transactivation, and adipogenesis). This finding supports the notion that PPAR␥ ligands, which might exhibit more restricted (tissue-or even gene-specific) effects, and hence different therapeutic or toxicity profiles, await discovery.