α,β-Unsaturated Ketone Is a Core Moiety of Natural Ligands for Covalent Binding to Peroxisome Proliferator-activated Receptor γ*

Peroxisome proliferator-activated receptor γ (PPARγ) functions in various biological processes, including macrophage and adipocyte differentiation. Several natural lipid metabolites have been shown to activate PPARγ. Here, we report that some PPARγ ligands, including 15-deoxy-Δ12,14-prostaglandin J2, covalently bind to a cysteine residue in the PPARγ ligand binding pocket through a Michael addition reaction by an α,β-unsaturated ketone. Using rhodamine-maleimide as well as mass spectroscopy, we showed that the binding of these ligands is covalent and irreversible. Consistently, mutation at the cysteine residue abolished abilities of these ligands to activate PPARγ, but not of BRL49653, a non-covalent synthetic agonist, indicating that covalent binding of the α,β-unsaturated ketone in the natural ligands was required for their transcriptional activities. Screening of lipid metabolites containing the α,β-unsaturated ketone revealed that several other oxidized metabolites of hydroxyeicosatetraenoic acid, hydroxyeicosadecaenoic acid, and prostaglandins can also function as novel covalent ligands for PPARγ. We propose that PPARγ senses oxidation of fatty acids by recognizing such an α,β-unsaturated ketone as a common moiety.

main (LBD) with or without synthetic ligands (9 -14) have contributed to better understanding of the binding selectivity and activation mechanism of the synthetic PPAR␥ ligands. In the PPAR␥ ligand binding pocket, a large hydrophobic region makes contact with ligands, and a hydrogen bond network between PPAR␥ and ligands stabilizes helix 12, which may promote binding of coactivator proteins (2). However, there is no structural report for the binding mode of the natural PPAR␥ ligands. In the case of PPAR␦, its crystal structure complexed with a fatty acid has been reported (15). The ligand, eicosapentaenoic acid, was attached to its binding pocket in two distinct conformations in the crystal. Meanwhile, it has been reported that a natural PPAR␥ ligand, 15-deoxy-⌬ 12,14 -prostaglandin J 2 (15d-PGJ 2 ), has the potential to bind covalently to some proteins, such as H-Ras (16), NF-B (17,18), IB kinase (19), and AP-1 (20) via a cysteine residue. In addition, it is reported that the synthetic PPAR␥ ligands GW9662, T0070907, and L-764406 (21)(22)(23) have been shown to react with the cysteine residue of the PPAR␥ LBD. These lines of evidence led to us investigate the binding mode of 15d-PGJ 2 to the PPAR␥ LBD.

EXPERIMENTAL PROCEDURES
Chemicals-PPAR␥ ligands and eicosanoids were purchased from Cayman Chemical Co. or Alexis Corporation (Lausen, Switzerland). Rhodamine-maleimide was purchased from Molecular Probes, Inc. Reagents for animal cell culture were obtained from Invitrogen. All other chemicals were purchased from Sigma or WAKO (Osaka, Japan).
Plasmids-For bacterial expression of the PPAR␥ LBD, a PPAR␥ fragment containing amino acid residues 195-477 was amplified by PCR using oligonucleotides (5Ј-GA CAT ATG GCG GAG ATC TCC AGT and 5Ј-G CGG ATC CTA GTA CAA GTC CTT G; underlines indicate introduced NdeI and BamHI sites, respectively) and subcloned into * This work was supported by a research grant endorsed by the New Energy and Industrial Technology Development Organization. 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.
NdeI/BamHI-digested pET28a (Invitrogen). After plasmid DNA was recovered from a colony, the sequence was confirmed.
To obtain the C285A mutant, pCMX-GAL4-hPPAR␥1 was amplified by Pyrobest DNA polymerase (TAKARA Bio. Inc.) using oligonucleotides (5Ј-ATC TTT CAG GGC GCT CAG TTT CGC TCC and 5Ј-GGA  GCG AAA CTG AGC GCC CTG AAA GAT; underlines indicate mutated  codon). The template DNA was removed by digestion with DpnI, and the amplified DNA was directly transformed into Escherichia coli DH5␣. After plasmid DNA was recovered from a colony, the sequence was confirmed. The Y475F mutant was obtained by the same procedure using oligonucleotides (5Ј-CTG CAG GAG ATC TTC AAG GAC TTG TAC and 5Ј-GTA CAA GTC CTT GAA GAT CTC CTG CAG; underlines indicate mutated codon).
Protein Purification-The human PPAR␥ LBD was expressed in BL21 (DE3) and purified by nickel affinity chromatography as described previously (14). After removal of a polyhistidine tag by thrombin digestion, the PPAR␥ LBD was further purified by gel filtration using Superdex200 (Amersham Biosciences). We omitted the reducing agents from all the buffers.
TOF Mass Spectrometry Analysis-After the PPAR␥ LBD protein was incubated with the ligands, the protein solution was mixed with 33% acetonitolyl saturated with sinopinic acid at a 1:1 ratio. Then, 2 l of the sample was spotted onto the grid and dried completely under vacuum. A MALDI-TOF mass spectrometry spectrum was obtained with a Voyager Elite (PerSeptive Biosystems, Framingham, MA).
Rhodamine-Maleimide Assay-The purified PPAR␥ LBD (0.1 M) was mixed with various ligands at the indicated concentrations in a buffer containing 20 mM Tris-HCl, pH 7.4, 150 mM NaCl. The mixture was incubated at room temperature for 30 min. After SDS was added to the reaction mixture to a final concentration of 0.5%, rhodamine-maleimide and tris-(2-carboxyethyl)phosphine hydrochloride were added to the mixture to 0.2 and 1 mM final concentrations, respectively. After incubation at room temperature for 30 min, the samples were treated with SDS-PAGE sample buffer containing ␤-mercaptoethanol and were separated by SDS-PAGE. Fluorescent signals derived from rhodamine were visualized by a FM-BIO II (Hitachi).
Calculation of Binding Constant-We calculated the k on values of irreversible binding of ligands as follows.
[NR] ϩ [ligand] 3 [NR-ligand] (binding constant ϭ k on ) Here, NR and NR-ligand indicate a nuclear receptor and a ligandconjugated nuclear receptor, respectively. The reaction speed at certain time (t) is shown as Equation 1.
If a is not equal to b, Equation 1 can be converted to Equation 2.
If a is equal to b, Equation 1 can be converted to Equation 3.
We obtained the k on value after nonlinear least square fitting of the data using Equation 2 and Equation 3 by the Newton method. We calculated the K d value of competitors as follows. [ The reaction speed of the covalent binding at certain time (t) is shown as Equation 5.
We obtained the K a value after nonlinear square fitting of the data with the values generated by a differential equation modified from Equations 4 and 5. K d was obtained as 1/K a . Measurement of UV Spectra-Samples were mixed at indicated con-centrations and incubated for 20 min at room temperature. UV spectra between 250 and 400 nm were recorded using a Beckman DU640 spectrophotometer. For graphical presentation, UV spectrum of PPAR␥ alone was subtracted from that of the ligand-PPAR␥ mixture. Cell Culture and Luciferase Assay-Cell culture, transient transfection, and luciferase assay were described previously (24). In this study, we measured the fluorescence of enhanced yellow fluorescent protein as an internal control of transfection efficiency, using a Fusion ␣ (PerkinElmer, Inc.).
Model Building-The 15d-PGJ 2 -PPAR␥ covalently bound complex was built using the crystal structure of the PPAR␥-BRL49653-coactivator peptide complex (Protein Data Bank accession code 1FM6) (12) as the template. The force field and programs used for the potential-energy calculations of the PPAR␥ and the coactivator peptide were described in a previous report (25). One of the conformations of 15d-PGJ 2 , which was randomly generated with a molecular dynamics simulation at 400 K in vacuo, was introduced into the ligand binding pocket of PPAR␥ by visual manipulation. Here, the carboxyl group and the cyclopentenone ring in 15d-PGJ 2 were placed in the vicinity of the BRL49653 head group and the sulfur atom in Cys-285, respectively. All of the Asp, Glu, Arg, Lys, and His residues and the carboxyl group in 15d-PGJ 2 were treated as the charged forms. Force field parameters for 15d-PGJ 2 covalently bound to Cys-285 were taken from those of the analogous fragments in the AMBER general force field (www.amber.scripps.edu/) for the organic molecule data base. Its partial charges were obtained from charge fitting with the RESP program (26) (www.amber. scripps.edu/Questions/resp.html) after geometry optimization with the Gaussian 98 program (27) using the Hartree-Fock/6 -31G* basis set. All of the energy minimization and molecular dynamics calculations were carried out in vacuo. The potential energy of the complex was minimized by using the conjugated gradient method via the following steps. First, the geometry of 15d-PGJ 2 -conjugated Cys-285 in PPAR␥ was allowed to be flexible while the position of the other part was restrained with a harmonic potential, and then all of the complex was allowed to be flexible without restraint. After the energy minimization procedures, a 100-ps heating from 0 to 300 K and a 500-ps equilibration were done. Subsequently, a 500-ps sampling simulation was executed. Here, all of the main-chain atoms of PPAR␥ and the coactivator peptide were restrained by applying a harmonic potential to the minimized conformation, and the time step of the molecular dynamics simulation was 1 fs. Finally, the closest structure to the average of the sampled ones was chosen as the complex model.
Reverse Transcription-PCR-THP-1 cells were maintained in RPMI1640 with 10% fetal calf serum at 37°C. 2 ϫ 10 6 cells were seeded in 6-cm dishes and treated with ligands for 10 h. Cells were collected by centrifugation, and total RNA was purified from cells by TRIzol reagent (Invitrogen) according to the manufacturer's instructions. Quality of isolated RNA was determined by electrophoresis. Single-stranded cDNA was synthesized from 1 g of total RNA using reverse transcriptase (Promega) with an oligo(dT) 15 primer. PCR primers were designed according to the uniSTS data base; detail is available upon request. PCR products were separated by PAGE and stained with SYBR Gold (Molecular Probes). Images of the gels were obtained by the FM-BIO II (Hitachi).

RESULTS
Model Building of the 15d-PGJ 2 -PPAR␥ Complex-As the first step toward understanding the binding mechanism of the natural PPAR␥ ligands, we built a model structure of the 15d-PGJ 2 -PPAR␥ complex according to the superimposed structures of the PPAR␥ LBD complexed with several synthetic ligands (Protein Data Bank accession codes 1KNU, 1FM6, 1I7I, 4PRG). In the superimposed structures, the polar part of the ligands tended to localize near helix 12 of PPAR␥ and the non-polar part near helix 3 (Fig. 1A). To fit 15d-PGJ 2 to the binding site of BRL49653, we considered four possible orientations of 15d-PGJ 2 in the PPAR␥ ligand binding pocket (Fig. 1B). In the models, the cyclopentenone ring (Fig. 1B, a and  b) or the carboxyl group (Fig. 1B, c and d) of 15d-PGJ 2 could be assigned as a polar moiety. In the latter case, we noticed that a cysteine residue (Cys-285) of PPAR␥ located at close proximity to the electrophilic carbon(s) of an ␣,␤-unsaturated ketone in 15d-PGJ 2 (Fig. 1B, c and d). The PPAR␥ ligand binding pocket was large enough to accommodate 15d-PGJ 2 (Fig. 1C).
Covalent Binding of 15d-PGJ 2 to PPAR␥ LBD-The model building described above raised the possibility that 15d-PGJ 2 is bound to PPAR␥ covalently. To test this possibility, we performed a mass spectroscopy experiment. Generally, macromol-ecules such as proteins are denatured during MALDI-TOF mass spectroscopy analyses, and so small molecules, such as ligands, are detached from the proteins. Actually, with or without the synthetic ligand BRL49653, the PPAR␥ LBD showed the same molecular weight as that calculated from the amino acid composition (MW ϭ 34, 534) ( Fig. 2A). In contrast, in the presence of 15d-PGJ 2 , the PPAR␥ LBD exhibited a higher molecular weight, which was close to the molecular weight of the PPAR␥ LBD plus 15d-PGJ 2 (calculated MW ϭ 34, 850) ( Fig.  2A). This result indicated that 15d-PGJ 2 was bound to PPAR␥ even after the PPAR␥ was denatured and suggested that 15d-PGJ 2 covalently bound to PPAR␥.
To confirm that 15d-PGJ 2 modifies the cysteine residue in the PPAR␥ LBD, we took advantage of rhodamine-maleimide to label free cysteine residues in proteins. Fortunately, there is only one cysteine residue in the PPAR␥ LBD. In these experiments, we denatured the protein by adding 0.5% SDS after ligand-protein complex formation, to facilitate rhodamine-maleimide access to the cysteine residue in the PPAR␥ LBD (Fig.  2B). The amount of the rhodamine-maleimide-labeled PPAR␥ LBD decreased by incubating with 15d-PGJ 2 in a concentrationdependent manner (Fig. 2C), indicating that 15d-PGJ 2 covalently bound to the cysteine residue of PPAR␥. Based on the irreversible binding reaction model, we calculated k on value of 15d-PGJ 2 as 47,165.3 M Ϫ1 min Ϫ1 . We next added the known PPAR␥ ligands to the reaction mixture simultaneously with 15d-PGJ 2 (Fig. 2D). Excess amount of the competitors can occupy the ligand binding pocket, and so 15d-PGJ 2 cannot reach to the cysteine residue. Then addition of 0.5% SDS removes the non-covalently bound competitors, with the result that the cysteine residue in the denatured PPAR␥ LBD can be labeled by rhodamine-maleimide. This result indicates that the non-covalent ligands and 15d-PGJ 2 share their binding site. Using the k on value of 15d-PGJ 2 , the K d values of BRL49653, MCC-555, and 13(S)-HODE were calculated to be 129 nM, 577 nM, and 199 M, respectively.
Rhodamine-maleimide assay revealed that 15d-PGJ 2 specifically modified the cysteine residue of the PPAR␥ LBD. Concomitantly, 15d-PGJ 2 itself must be modified by binding to the PPAR␥ LBD. Then, we measured the UV spectrum of 15d-PGJ 2 in the absence or presence of the PPAR␥ LBD (Fig. 3A). In the absence of the PPAR␥ LBD, 15d-PGJ 2 showed peak absorbance at 320 nm ( Fig. 3A, a, thin line). After incubating with the PPAR␥ LBD, the resultant peak absorbance of 15d-PGJ 2 shifted to 295 nm (Fig. 3A, a, thick line). This blue shift of the peak absorbance was also observed even when 15d-PGJ 2 was incubated in a free cysteine solution (Fig. 3A, b). However, the resultant peak absorbance was 310 nm, which was different from the PPAR␥-induced peak absorbance but rather similar to that of a 15d-PGJ 2 analogue, CAY10410, which lacks the cyclopentenone ring (Fig. 3A, c, thick line). The results suggested that the cysteine residue of the PPAR␥ LBD targeted an electrophilic carbon within 15d-PGJ 2 that is different from that of free cysteine targets. Considering the similarity of the peak absorbance, the free cysteine may target C9 of 15d-PGJ 2 (Fig.  3B), whereas the cysteine residue in the PPAR␥ LBD may target C13 of 15d-PGJ 2 (Fig. 3B). The amplitude of the UV spectrum of CAY10410 was reduced by binding to PPAR␥ (Fig.  3A, c), whose peak absorbance was probably shifted to a shorter wavelength than that we had measured. The result indicated that the cysteine residue in the PPAR␥ LBD may also target C13 of CAY10410 (Fig. 3B) as well as 15d-PGJ 2 .
Covalent Binding of Natural Ligands to PPAR␥-PPAR␥ is reportedly activated by several natural and synthetic ligands (5)(6)(7)(8). For example, Nagy et al. (7) showed that oxidized fatty acids, e.g. 9-and 13-hydroxyoctadecadienoic acid (HODE) and 9-and 13-oxo-octadecadienoic acid (oxoODE) derived from oxidized low density lipoprotein (oxLDL), activated PPAR␥ and mediated the cellular effect of oxLDL in macrophages. Then we assessed whether such known PPAR␥ ligands possess the ability to bind covalently to PPAR␥. We showed that the molecules (e.g. 15d-PGJ 2 , CAY10410, 9-oxoODE, 13-oxoODE, and T0070907) covalently bound to the cysteine residue in the PPAR␥ LBD in the rhodamine-maleimide assay (Fig. 4A). CAY10410 also showed weaker blocking activity than 15d-PGJ 2 . The natural ligands and CAY10410, which covalently bound to the PPAR␥ LBD, share the common chemical structure of an ␣,␤-unsaturated ketone (Fig. 4B, boxes). Although 9(S)-HODE and 13(S)-HODE have structures quite similar to those of 9-oxoODE and 13-oxoODE, respectively, they lack an ␣,␤-unsaturated ketone and did not covalently bind to the PPAR␥ LBD, indicating that the ␣,␤-unsaturated ketone was a crucial structure for the covalent binding. It is reported that synthetic ligands GW9662, T0070907, and L-764406 covalently bind to the cysteine residue in the PPAR␥ LBD (21)(22)(23). In this assay, we were able to confirm the cysteine modification by T0070907 (Fig. 4A), which has been shown to react with the cysteine residue of the PPAR␥ LBD through a nucleophilic aromatic substitution of the chlorine in the structure (21). Because the cysteine residue of the PPAR␥ LBD targets the ␤-carbon in the ␣,␤-unsaturated ketone, as shown in Fig. 3, the different mode of action was evident between the 15d-PGJ 2represented natural ligands and these synthetic ligands.
Although the binding mode by 15d-PGJ 2 , 9-oxoODE and 13-oxoODE was achieved by the same chemical reaction, they showed different IC 50 to block the rhodamine-maleimide to label the free cysteine residue in the PPAR␥ LBD (Fig. 4C). We calculated the k on values of 15d-PGJ 2 , 13-oxoODE, and 9-oxoODE as 47,165.3, 8,073.7, and 1,535.3 M Ϫ1 min Ϫ1 , respectively. Blue shift of the peak wavelength of 15d-PGJ 2 (10 M) was also induced by incubating with 2 mM cysteine (b). The resultant peak (thick spectrum) in cysteine-reacted 15d-PGJ 2 corresponded to that of a 15d-PGJ 2 analogue, CAY10410 (c). UV spectra of CAY10410 (10 M) were also altered by incubating with 58 M PPAR␥. The peak wavelengths of 15d-PGJ 2 and CAY10410 are indicated as vertical dashed lines. B, explanation of UV spectral rearrangement by the cysteine residue in PPAR␥ and the free cysteine. The cysteine residue of the PPAR␥ LBD reacts with C13 in 15d-PGJ 2 and CAY10410, whereas a free cysteine (Cys) reacts with C9 in 15d-PGJ 2 .

FIG. 2. Covalent binding of 15d-PGJ 2 to the PPAR␥ LBD.
A, mass spectroscopy analysis of PPAR␥ LBD with or without ligands. After incubating the PPAR␥ LBD with either Me 2 SO, BRL49653, or 15d-PGJ 2 , the molecular weight of the ligand-PPAR␥ complex was investigated by MALDI-TOF mass spectroscopy. B, schematic representation of the method used to analyze the covalent binding of ligands to the cysteine residue in the PPAR␥ LBD. After incubating the PPAR␥ LBD with ligands, proteins were denatured by adding SDS. Free cysteine residues were then detected by rhodamine-maleimide. C, concentrationdependent modification of the cysteine residue in the PPAR␥ LBD by 15d-PGJ 2 . D, concentration-dependent competition of 15d-PGJ 2 binding to PPAR␥ with other ligands. Various concentrations of competitor ligands were simultaneously added to the PPAR␥ LBD with 5 M 15d-PGJ 2 , and the cysteine modification by 15d-PGJ 2 was investigated by rhodamine-maleimide.

Requirement of the Covalent Binding for PPAR␥ Activation-
We evaluated the relationship between the binding modes of the ligands and their activities in a cell-based transcription assay (Fig. 5A). BRL49653 was the most potent activator of PPAR␥ among the ligands we tested (Fig. 5A, left  panel). CAY10410 showed transcriptional activity similar to that of 15d-PGJ 2 (Fig. 5A, middle panel). On the other hand, when we compared the activities of structurally related molecules, e.g. 13(S)-HODE versus 13-oxoODE and 9(S)-HODE versus 9-oxoODE (Fig. 5A, right panel), the ligands that covalently bound to PPAR␥ always showed significantly higher activities than the analogous ligands lacking an ␣,␤-unsaturated ketone. This result is consistent with a previous report that oxo me-tabolites showed greater activities and affinities than HODE did (7). These results suggested that the covalent binding to the receptor helped the ligands to activate PPAR␥.
We made three PPAR␥ mutants with the cysteine exchanged to alanine (C285A), valine (C285V), or serine (C285S). All of the mutants were still activated by BRL49653 at levels equivalent to that of the wild type PPAR␥, whereas they totally lost responsiveness to 15d-PGJ 2 and 13-oxoODE (Fig. 5B). Thus, these putative natural ligands indeed functioned as ligands, via novel covalent binding to the cysteine residue in the ligand binding pocket, to activate transcription.
Next, we analyzed the effect of an antagonist on activation by both synthetic and natural ligands for PPAR␥. Two hours after agonist addition, the PPAR␥-specific antagonist GW9662 was applied to the cells expressing GAL4-PPAR␥ protein and a luciferase gene as a reporter (Fig. 5C). Both 15d-PGJ 2 -and BRL49653-dependent activities of PPAR␥ decreased with the addition of GW9662. However, inhibition of 15d-PGJ 2 -dependent activity of PPAR␥ by GW9662 was more moderate than that of BRL49653-dependent activity. Rapid inhibition of BRL49653-dependent activity might be due to the displacement of the ligand by the antagonist. On the other hand, resistance of 15d-PGJ 2 against the antagonist indicated that 15d-PGJ 2 binding to PPAR␥ in the cells was irreversible. We considered that slower and more moderate inhibition of 15d-PGJ 2 -dependent activity was not achieved by the displacement of the ligand but by the competitive binding between 15d-PGJ 2 and GW9662 to newly synthesized PPAR␥.
Concentration Dependence of PPAR␥ Activation-Concentration dependence of PPAR␥ activation by oxidized prostaglandins revealed that 15-keto-PGF 1 and 15-keto-PGE 1 showed higher EC 50 and maximal activation than 15d-PGJ 2 did (Fig.  8A). The order of EC 50 of oxidized prostaglandins was 15d-PGJ 2 , 15-keto-PGE 1 , and 15-keto-PGF 1 . BRL49653 showed both low EC 50 and high maximal activation. This phenomenon indicated that EC 50 and maximal activation were independently regulated by each ligand. These oxidized prostaglandins also blocked the maleimide labeling of the PPAR␥ LBD in a concentration-dependent manner (Fig. 8B), indicating that FIG. 4. Naturally occurring ligands covalently bind to the PPAR␥ LBD. A, modification of the cysteine residue in the PPAR␥ LBD by ligand binding. B, natural ligands that can bind covalently to PPAR␥ have a common chemical structure, an ␣,␤-unsaturated ketone (indicated as a box). An electrophilic carbon in the ␣,␤-unsaturated ketone is indicated by an arrow. C, dose-dependent modification of PPAR␥ by ligands. The PPAR␥ LBD was incubated with various concentrations of 15d-PGJ 2 13-oxoODE or 9-oxoODE, and the covalent binding was detected by using rhodamine-maleimide. Quantifications of the band intensities are plotted. they bound to the cysteine residue in the PPAR␥ covalently. The order of the affinities was 15d-PGJ 2 , 15-keto-PGE 1 , and 15-keto-PGF 1 , which was correlated with that of EC 50 .
Structural Models of the Natural Ligand-PPAR␥ Complexes-To investigate whether covalent binding of 15d-PGJ 2 to the PPAR␥ LBD is structurally achievable, we built a structural model of the 15d-PGJ 2 -PPAR␥ complex, based upon the crystal structure of the PPAR␥/retinoid X receptor ␣ heterodimer complexed with BRL49653 and a coactivator peptide (Protein Data Bank accession code 1FM6) (12). The obtained structure was superimposed onto the PPAR␥-BRL49653 complex, suggesting some similarities and differences between the BRL49653-and 15d-PGJ 2 -PPAR␥ complexes (Fig. 8C). The hydrophobic tail of 15d-PGJ 2 , which BRL49653 lacked, extended toward the tip of helix 11 (L452) and was stabilized through hydrophobic interactions. This binding mode was different from that observed in the eicosapentaenoic acid-PPAR␦ complex, where the hydrophobic tail of eicosapentaenoic acid extended toward helix 2 or helix 5, rather than helix 11 (15).
To analyze what causes the difference in the maximal activation, we also built model structures of the PPAR␥ LBD covalently bound by 15-keto-PGE 1 and 15-keto-PGF 1 . Because these oxidized prostaglandins have very similar structures but very different chemical and physical natures, we took advantage of molecular dynamics simulation to get the model structures instead of homology modeling. Carboxyl groups of these ligands made a hydrogen bond with Tyr-473 (Fig. 8D), which resides in helix 12 and is a key ligand-recognizing residue (12)(13)(14). Hydrophobic clusters assisting the ligand binding were also conserved ( Fig. 8C and data not shown). In contrast, we observed two differences in hydrogen bonds between the main-chain carbonyl group of Leu-340 and a hydroxyl group of the ligands (Fig. 8D, a) and between the sulfur atom of the cysteine residue and another hydroxyl group (Fig. 8D, b). The former was observed only in the 15-keto-PGF 1 -PPAR␥ complex, and the latter was observed both in the 15-keto-PGF 1 -and 15-keto-PGE 1 -PPAR␥ complexes, but not in the 15d-PGJ 2 -PPAR␥ complex. These structural differences may produce more stable binding of 15-keto-PGF 1 or 15-keto-PGE 1 than 15d-PGJ 2 and exert the receptor to have higher maximal activity, namely, efficacy.
We observed that all three ligands required the cysteine residue (Fig. 8E, gray bars). According to the structural model described above, we made an additional PPAR␥ mutant (Y473F) that showed no response to any ligands (Fig. 8E), supporting our model structure. Regulation of PPAR␥ Target Genes by Oxidized Prostaglandins-Finally, we determined whether oxidized prostaglandins activate PPAR␥ target genes in THP-1 cells, which endogenously express PPAR␥ protein. Adipophilin (ADFP) was upregulated by PPAR␥ ligands (Fig. 9A, left panel), and the upregulation was blocked by simultaneous application of PPAR␥ antagonist T0070907 (Fig. 9A, right panel), suggesting that ADFP expression was regulated through PPAR␥ (28,29). Among several PPAR␥-regulated genes that we have identified by microarray, 2 we successfully observed that oxidized prostag-landins 15-ketoPGE 1 and 15-ketoPGF 1 as well as 15d-PGJ 2 and 13-oxoODE activated PPAR␥-regulated genes, ADFP, interleukin-1 receptor type I, cyclin-dependent kinase inhibitor 1A, and aldehyde oxidase 1 (Fig. 9B). These results suggested that oxidized prostaglandins acted as PPAR␥ ligands and regulated the expression of PPAR␥ target genes in vivo. DISCUSSION By this study, the currently proposed natural PPAR␥ ligands were shown for the first time to share a common structure, the ␣,␤-unsaturated ketone, that enables them to bind to PPAR␥ covalently and to exhibit their activities. Although we do not neces-2 T. Shiraki and H. Jingami, unpublished data. sarily exclude the possibility that as yet unrevealed high affinity ligands with novel moieties still exist, our studies provide several advances toward answering the enigmas in PPAR␥ functions.
First, one can easily imagine that the covalent binding of the ligands to the receptor not only acts as a rigid switch but also enables the receptors to accumulate as active forms, allowing such ligands with low concentrations to function in vivo. Unknown factor(s) may facilitate this covalent binding.
Second, our studies provide clear evidence that the structural moieties within a single molecule are used differently in PPAR␥-dependent and -independent functions. In the case of 15d-PGJ 2 , the electrophilic carbon (C9) within a cyclopentenone ring has been considered to react with the cysteine residue in NF-B and other proteins (17,30). On the other hand, we consider that the carbon at position 13 reacts with the sulfur atom of the cysteine residue in PPAR␥, according to results from UV absorbance measurements and the activity of CAY10410 lacking the cyclopentenone ring (Figs. 2A, 3A, and 5A). Furthermore, the carboxyl group of 15d-PGJ 2 is required for the formation of a hydrogen bond with Tyr-473 in helix 12 of PPAR␥. The 15d-PGJ 2 biotinylated at the carboxyl group reportedly failed to pull down PPAR␥ from cells, whereas it still bound to AP-1 and other proteins (20,30). Thus, the covalent binding of 15d-PGJ 2 to the PPAR␥ LBD is different from that in PPAR␥-independent actions.
Third, our finding that several oxidized eicosanoids act as PPAR␥ ligands fits well with the expected PPAR␥ functions: PPAR␥ may function under oxidative conditions and/or inflammation. For example, glucose consumption is related to mitochondrial respiration, which generates free radical. In atherosclerosis, macrophages play important roles in which oxidized low density lipoprotein accumulates as a major source of oxidized lipid (7). Under inflammatory conditions, macrophages express not only cyclooxygease-2 and 12/15-lipoxygenase, to produce prostaglandins and leukotrienes, respectively (31,32), but also nitric-oxide synthase to generate nitric oxide, another potent oxidative agent (33). In addition, expressions of prostaglandin transporter and 15-hydroxyprostaglandin dehydrogenase, both of which are involved in prostaglandin metabolism to form 15-keto-protstaglandins (34,35), were modulated during lipopolysaccharide-induced fever (36). Recently, it has been shown that oxidative stress stimulates inflammatory cells to produce 5-oxoETE from 5(S)-hydroxy-6,8,11,14-eicosatetrae- noic acid (37). These reports suggest that the production of oxidized eicosanoid metabolites is physiologically regulated under certain conditions, such as inflammation. Thus, PPAR␥ may sense such oxidized fatty acids as signaling molecules and function as an inflammation modulator.
On the other hand, the physiological relevance of the lipid metabolites, especially 15d-PGJ 2 , has long been debated (38). Further studies will be needed to determine whether they really act as endogenous ligands on PPAR␥.