Different Residues Mediate Recognition of 1-O-Oleyllysophosphatidic Acid and Rosiglitazone in the Ligand Binding Domain of Peroxisome Proliferator-activated Receptor γ*

Here we showed that a naturally occurring ether analog of lysophosphatidic acid, 1-O-octadecenyl-2-hydroxy-sn-glycero-3-phosphate (AGP), is a high affinity partial agonist of the peroxisome proliferator-activated receptor γ (PPARγ). Binding studies using the PPARγ ligand binding domain showed that [32P]AGP and [3H]rosiglitazone (Rosi) both specifically bind to PPARγ and compete with each other. [32P]AGP bound PPARγ with an affinity (Kd(app) 60 nm) similar to that of Rosi. However, AGP displaced ∼40% of bound [3H]Rosi even when applied at a 2000-fold excess. Activation of PPARγ reporter gene expression by AGP and Rosi showed similar potency, yet AGP-mediated activation was ∼40% that of Rosi. A complex between AGP and PPARγ was generated using molecular modeling based on a PPARγ crystal structure. AGP-interacting residues were compared with Rosi-interacting residues identified within the Rosi-PPARγ co-crystal complex. These comparisons showed that the two ligands occupy partially overlapping positions but make different hydrogen bonding and ion pairing interactions. Site-specific mutants of PPARγ were prepared to examine individual ligand binding. H323A and H449A mutants showed reduced binding of Rosi but maintained binding of AGP. In contrast, the R288A showed reduced AGP binding but maintained Rosi binding. Finally, alanine replacement of Tyr-473 abolished binding and activation by Rosi and AGP. These observations indicate that the endogenous lipid mediator AGP is a high affinity ligand of PPARγ but that it binds via interactions distinct from those involved in Rosi binding. These distinct interactions are likely responsible for the partial PPARγ agonism of AGP.

nanomolar equilibrium binding constants. These synthetic agents are clinically used to improve insulin sensitivity and fat metabolism in type 2 diabetics (23)(24)(25)(26). However, these classical full PPAR␥ agonists elicit a variety of side effects, including weight gain, edema, increased fat mass, and tumor formation in rodents (27). To reduce unwanted side effects associated with TZD therapy, new partial agonists have been developed (28).
PPAR␥ was originally identified as an orphan nuclear receptor and an adipocyte transcription factor (29,30). Activation of PPAR␥ by TZDs leads to altered metabolism in adipose tissue, skeletal muscle, and liver that collectively results in insulin sensitization, lowering of blood glucose, and elevation of triglyceride clearance (31). Clinical data indicate that PPAR␥ agonists protect type 2 diabetics from atherosclerosis (32). On the other hand, there is considerable evidence to support a deleterious role for oxidized phospholipids and fatty acids as important signaling molecules in the context of the atherosclerotic lesion (33). The LDL particle acquires a number of important biological activities as a result of oxidative modification. Mild oxidation of LDL produces prothrombotic and proatherogenic mox-LDL (34 -36). Recent studies suggest that lysophosphatidic acid (LPA) GPCR antagonists abolish platelet aggregation elicited by mox-LDL, indicating that an LPA-like lipid plays an essential role in the thrombogenic effects of mox-LDL (35). Oxidation of LDL creates a host of PPAR␥ agonists (37), including AGP (38). AGP is also formed from alkyl dihydroxyacetone phosphate through enzymatic reduction (39). In the brain, AGP is present at 0.44 nmol/g, ϳ15% that of acyl-LPA levels (40). Here we provide evidence that AGP is a ligand of PPAR␥ with potency similar to that of the TZD Rosi but with only 40% efficacy. Computational and mutational analysis of the AGP-PPAR␥ complex indicates differential interaction with key residues in the ligand binding and activation domains that explain the partial activation elicited by AGP. We further provide evidence that LPA GPCRs are not required for AGP-induced activation of PPAR␥. However, individual LPA GPCRs differentially regulate the transcriptional activity of PPAR␥ 1 .
Western Blot Analysis-Proteins from whole cell lysates were separated by 10% SDS-PAGE and transferred to a polyvinylidene difluoride membrane (Bio-Rad). The membrane was blocked with 5% (w/v) nonfat dried milk in 20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 0.05% Tween 20 (TBST) for 1 h and incubated with primary antibody in TBST with 5% (w/v) nonfat dried milk for 12 h. Secondary immunoreactive bands were visualized using the SuperSignal chemiluminescent substrate kit (Pierce).
Ligand Binding Assays-Hexahistidine (His 6 ) tag PPAR␥ fusion proteins or His 6 and thrombin recognition site containing empty vector controls were expressed in BL-21(DE3) cells. Transformed BL-21 cells were induced using 0.3 mM isopropyl 1-␤-D-galactopyranoside (Fisher) for 12 h at 25°C and collected by centrifugation. Recombinant LBD-PPAR␥ was extracted with lysis buffer (50 mM HEPES-KOH, pH 6.8, 200 mM NaCl, 5 mM DTT, 1 mM phenylmethanesulfonyl fluoride, 0.5% Triton X-100, and 15% glycerol) and centrifugation at 12,000 ϫ g for 20 min. 1 ml of the supernatant was incubated with 50 l of TALON metal affinity resin (BD Biosciences) at 4°C for 1 h in lysis buffer. The resin was washed five times with wash buffer (50 mM HEPES-KOH, pH 6.8, 200 mM NaCl, 5 mM DTT, 15% glycerol, and 5 mM imidazole) and eluted with 150 mM imidazole in wash buffer. The PPAR␥-LBD protein was quantified using the Bradford protein assay (Pierce) and Coomassie Blue staining. For saturation binding assay, 1 g of His 6 -PPAR␥-LBD fusion protein was incubated at 18°C for 1 h in binding buffer (50 mM HEPES, pH 6.8, 100 mM NaCl, 5 mM EDTA, 5 mM DTT) and [ 32 P]AGP. The radioligand-LBD fusion protein was precipitated using 3.3% (w/v) ␥-globulin (Sigma) and 36% (w/v) polyethylene glycol 8000 (Fisher), collected on a DEAE-81 filter disks (Whatman), and quantified by liquid scintillation counting. Nonspecific binding was determined in the presence of nonradioactive AGP (10 M). For competition binding assays, 1 g of His 6 -PPAR␥-LBD protein was incubated at 18°C for 1 h in 200 l of binding buffer in the presence of 5 nM [ 3 H]Rosi or 5 nM [ 32 P]AGP with or without cold compounds. [ 32 P]AGP-bound His 6 -PPAR␥ was precipitated using ␥-globulin and polyethylene glycol 8000, collected on a DEAE-81 filter disk. [ 3 H]Rosi-bound protein was collected on a filter disks. Radioligand bound to His 6 -PPAR␥ was washed three times with wash buffer (50 mM HEPES, pH 6.8, 100 mM NaCl, 5 mM DTT), and bound radioactivity was quantified in a Beckman scintillation counter LS-6500.
Photoaffinity Labeling of PPAR␥-LBD with [ 32 P]Benzophenone-AGP 18:1-The synthesis and radioactive labeling of 1-alkylbenzophenone-2-hydroxy-sn-glycero-3-phosphate-[ 32 P] (AGP-BP) was done as described previously (23). Purified PPAR␥-LBD (0.25 g/l) was combined with AGP-BP at a 200 nM final concentration with and without 10 M AGP or Rosi. The samples were irradiated using a UV cross-linker (Spectroline) on ice for up to 15 min. Samples were subsequently diluted with SDS sample buffer and separated by 10% SDS-PAGE. Gels were dried and radioactive bands detected by autoradiography using BioMax MS x-ray film (Eastman Kodak Co.) and a BioMax Transcreen-HE intensifying screen (Kodak).
Transfection of Cultured Cells and Reporter Gene Assay-Determination of PPAR␥ activation in CV-1 and B103 cells transfected with the acyl-coenzyme A oxidase-luciferase (PPRE-ACox-Luc) reporter gene construct was performed as reported previously (38). RH7777 cells were transduced with 10 multiplicities of infection of ad5-PPRE-ACox-Luc recombinant virus. Two days post-transduction, Rosi and AGP 18:1 dissolved in Opti-MEM ⌱ (Invitrogen), 1% fetal bovine serum, and 1% Me 2 SO were applied for 20 h. Luciferase and ␤-galactosidase activities were measured with the Steady-Glo luciferase assay system (Promega) and the Galacto-Light Plus TM system (Applied Biosystems, Foster City, CA), respectively. Samples were run in quadruplicate, and the means Ϯ S.E. were calculated. Data are representative of at least two independent transfections. Student's t test was used for null hypothesis testing, and p Ͻ 0.05 was considered significant.
Ligand Docking-The PPAR␥ protein structure was derived from Protein Data Bank entry 1FM9 (41). The agonist farglitazar (GI262570) and crystallographic water molecules were removed. Hydrogen atoms of protein amino acids and conserved water molecules were added and optimized to a root mean square gradient of 0.01 kcal/mol⅐Å using the MMFF94 (41) force field. Nonpolar hydrogen atoms were removed prior to docking. The docking box volume was 30 ϫ 30 ϫ 30 Å, and the center was near the sulfur atom of residue Cys-285. The default software docking parameters were used except the following changes: maximum number of energy evaluations equaled 9 ϫ 10 9 ; maximum number of generations equaled 4 ϫ 10 4 ; number of iterations of Solis and Wets local search equaled 3,000. AGP 18:1 was built in the Ϫ1 ionization state and minimized to a root mean square gradient of 0.01 kcal/mol⅐Å using the MMFF94 force field (41) in the MOE software program (42). Docking studies were performed using the Autodock 3.0 software (43,44).

AGP 18:1 Directly Interacts with the Ligand Binding
Pocket of PPAR␥-To obtain direct proof for the interaction of AGP with the ligand binding domain of PPAR␥, a benzophenone derivative of AGP (AGP-BP) was synthesized ( Fig. 1) (45). Cold AGP-BP elicited a dosedependent activation of the PPRE-ACox-Luc reporter gene in CV-1 cells, similar to that of Rosi ( Fig. 2A). The PPAR␥-LBD was incubated with 200 nM [ 32 P]AGP-BP in the presence or absence of 10 M cold AGP or Rosi on ice and illuminated under ultraviolet light for up to 15 min. Modified PPAR␥-LBD was isolated from free [ 32 P]AGP-BP by SDS-PAGE, and covalently bound radioactivity was monitored by autoradiography. PPAR␥-LBD showed a time-dependent increase in labeling by [ 32 P]AGP-BP, which was decreased by a 50-fold excess of either AGP or Rosi (Fig. 2B). This result suggests that AGP-BP interacts with a binding site within the PPAR␥-LBD utilized by both AGP and Rosi. AGP 18:1 and 16:0 have been shown to activate LPA GPCR (46,47). To establish that none of the four known LPA GPCRs is required for activation of PPAR␥, we sought cell lines that lack endogenous expression of known cell surface and intracellular LPA receptors. The rat hepatoma McA-RH7777 and the rat neuroblastoma B103 have been widely used to study LPA receptors because these cells do not respond to LPA when subjected to standard assays of GPCR (48,49). When tested using genespecific primers with reverse transcription-PCR, RH7777 showed a lack of transcripts encoding each of the four known LPA GPCRs. B103 cells, even though not responsive to LPA, showed the expression of the LPA 4 transcript ( Fig. 2C and supplemental Fig. S1). RH7777 cells abundantly expressed the PPAR␥ transcript, although we were unable to detect PPAR␥ transcript in B103 cells in experiments using up to 40 cycles of PCR amplification. Western blot analysis using a PPAR␥-specific antibody confirmed the presence of this transcription factor in RH7777 cells (Fig. 2D). In contrast, we could not detect PPAR␥ in B103 cells or CV-1 cells. We also confirmed that Rosi or AGP can activate endogenous PPAR␥ expressed in LPA 1/2/3/4 -negative RH7777 cells (Fig. 2E). We discounted the role of LPA 4 because in B103 cells that lack PPAR␥ and LPA 1/2/3 but express LPA 4 , AGP did not activate PPAR␥ unless it was introduced heterologously. When B103 cells were transfected with empty pcDNA3.1 vector, we could not detect transcriptional activity of the PPRE-ACox-Luc reporter gene after exposure to either Rosi, AGP, or AGP ϩ Rosi (Fig. 2F). Heterologous expression of PPAR␥ introduced responsiveness to both ligands. Ligand-induced activation of the reporter gene was abolished by the specific PPAR␥ inhibitor GW9662 (Fig. 2G). These results together suggest that AGP specifically interacts with PPAR␥ and can activate the PPRE-ACox-Luc reporter gene independently of the expression of known LPA GPCRS. Henceforth, we used B103 cells to further characterize the effect of AGP on PPAR␥.
AGP 18:1 Is a High Affinity Ligand for PPAR␥-To quantitatively characterize the binding properties of AGP with PPAR␥-LBD, [ 32 P]AGP 18:1 was synthesized. Equilibrium binding studies showed that the binding of [ 32 P]AGP was concentration-dependent and saturable with an apparent K d of ϳ60 nM (Fig. 3A). For comparison, we also determined binding of [ 3 H]Rosi to the same batch of purified His 6 -PPAR␥-LBD. In agreement with previous reports (50)   Normalized luciferase activity (mean Ϯ S.E., n ϭ 5) was measured. **, p Ͻ 0.01, significant differences over vehicle control. G, PPRE-ACox-Luc expression is blocked by the PPAR␥-specific antagonist GW9662 (500 nM). Transfection and assay conditions were the same as in F. **, p Ͻ 0.01, significant differences over vehicle control. FEBRUARY 10, 2006 • VOLUME 281 • NUMBER 6 concentration-dependent, saturable binding to PPAR␥-LBD with an apparent K d(app) of ϳ40 nM (data not shown). Thus, these two ligands show similar high affinity binding to PPAR␥. Next, we determined whether AGP bound to PPAR␥ as effectively as Rosi by directly comparing the concentration dependence of [ 32 P]AGP binding to [ 3 H]Rosi binding at nanomolar concentrations, a range representative of physiological ligand concentrations. We found that specific [ 32 P]AGP binding was slightly lower than [ 3 H]Rosi binding at all concentrations, which is consistent with its lower apparent K d (Fig. 3B).

Differential Ligand Recognition by PPAR␥
[ 32 P]AGP was displaced by cold AGP in a concentration-dependent manner. Over 80% of bound [ 32 P]AGP was displaced with a 1,000-fold excess of cold AGP (Fig. 3C). In contrast, only ϳ55% of bound [ 32 P]AGP could be displaced by Rosi even with a 10,000-fold excess of the cold competitor (Fig. 3C). When the labeled ligands were reversed (Fig. 3D), ϳ95% of [ 3 H]Rosi was displaced using a 2,000-fold excess of cold Rosi, whereas only ϳ40% of [ 3 H]Rosi was displaced by the same 2,000-fold excess of cold AGP. The displacement curve was biphasic with as little as 1 nM cold AGP able to displace ϳ45% of bound [ 3 H]Rosi (Fig. 3D), whereas levels of cold AGP Ͼ1 M only brought about an additional ϳ10% increase in displacement. These results suggest that both ligands bind tightly to PPAR␥ and that their binding sites are overlapping but not identical.
To delineate further the differences between binding of [ 3 H]Rosi and [ 32 P]AGP, we examined a panel of related lipids as cold competitors. All cold lipids were applied at 2.5 M, corresponding to a 500-fold excess over the radiolabeled ligand (Fig. 4, A and B). In this competition binding assay, only linoleic acid blocked Ͼ95% of  place labeled Rosi at the concentration tested. We also tested these lipids for their ability to block [ 32 P]AGP binding (Fig. 4B). In this case, AGP 16:0, AGP 18:1, LPA 18:1, cPA 18:1, LPA 20:0, and PC displaced more than 50% of the radioligand. Most surprisingly, 15-deoxy-PGJ 2 , oleic acid, and linoleic acid, compounds widely regarded as endogenous PPAR␥ agonists, were unable to displace [ 32 P]AGP 18:1 at the concentration applied. The differences in the displacement of the two radioligands further support the notion that they interact differently with PPAR␥.
AGP 18:1 Is a Partial Agonist of PPAR␥-B103 cells that lack PPAR␥ and LPA 1/2/3 were transfected with the PPRE-ACox-Luc reporter con-struct and treated with increasing concentrations of AGP, Rosi, and S1P. We found the threshold of AGP-induced reporter gene activation to be ϳ10 nM (ϳ1.5-fold; Fig. 4C). A further increase in AGP concentration elicited only modest incremental increases in transcriptional activation so that maximal expression was achieved at 10 M (ϳ1.9fold). Rosi also induced a concentration-dependent increase in transcriptional activity with a threshold concentration ϳ10 nM. However, the magnitude of this response was greater than that of AGP reaching a maximum of ϳ3.4-fold activation in this cell line. The related lipid mediator S1P failed to activate (Fig. 4C) the reporter gene despite that B103 cells express the S1P 1 GPCR (data not shown). Because as little as 10 nM AGP or Rosi was effective in eliciting a modest, yet significant increase in transcriptional activation of the reporter gene, we compared the transcriptional activity elicited by several putative, endogenous activators of PPAR␥ at this concentration. Tokumura et al. (51) established previously that 10 -14% of AGP remains intact in the cytoplasm within 30 min after extracellular application, indicating that a significant portion of AGP is imported into the cell. B103 cells transfected with the reporter gene construct were treated with 10 nM AGP 18:1, oleic acid, Az-PC, 15-deoxy-PGJ 2 , LPA 18:1, or Rosi. Among these putative endogenous PPAR␥ ligands, only AGP and Rosi induced significant increases in transcriptional activity (1.5-and 1.7-fold, respectively; Fig. 4D). These results, taken together with the binding data, are consistent with the notion that AGP functions as a partial agonist of PPAR␥ that is capable of activating this transcription factor at low nanomolar concentrations. Furthermore, AGP is more potent in activating the PPRE-ACox-Luc reporter gene construct than several putative endogenous PPAR␥ agonists. AGP activation is surpassed only by the synthetic full agonist, Rosi.
Structural Basis for the Differential Interaction of AGP and Rosi with PPAR␥-Because crystal structures of ligand-free and Rosi-bound PPAR␥ structures are available, we used molecular modeling techniques to computationally dock AGP within the PPAR␥ structure (Protein Data Bank entry 1FM9). The energetically most favorable model placed AGP in a position that partially overlapped with the position occupied by Rosi in its co-crystal structure with PPAR␥ (Fig. 5, A and B). Histidine residues 323 and 449 form critical interactions with the thiozolidine ring of Rosi that have been experimentally validated to play a role in TZD binding and activation (52). Tyrosine 473 in the AF-2 helix plays an essential role in Rosi-mediated activation of PPAR␥. Mutation of this residue renders PPAR␥ nonactivable by TZD (53). The predicted position of AGP within the PPAR␥ ligand binding pocket shows significant overlap with that of the Rosi-occupied volume (Fig. 5C), particu-larly relative to residues His-323, His-449, and Tyr-473. However, AGP hydrocarbons make no hydrogen bonding interactions with the two histidines like Rosi does. In contrast, the phosphate head group of AGP is predicted to make a salt bridge with Arg-288, a residue that is not engaged by Rosi (Fig. 5C). Furthermore, Ser-289 makes van der Waals contact with the central portion of the AGP hydrophobic tail. The terminus of the AGP hydrophobic tail interacts with residues that are not engaged by Rosi. We hypothesized that Arg-288 is likely to play a differentiating role in the interactions between PPAR␥ and AGP or Rosi. In contrast, we predicted that Tyr-473 is in the vicinity of the double bond and is likely to be important for interactions with both ligands. To test these hypotheses, we generated alanine or phenylalanine replacement mutants of these residues in the PPAR␥-LBD and tested their binding to and activation by Rosi and AGP. In agreement with our predictions, H323A, H449A, H323A/H449A, and Y473F completely lost their ability to bind Rosi (Fig. 6A). In contrast, H323A maintained ϳ80% and H449A maintained normal AGP binding (Fig. 6A). The R288A mutant showed reduced AGP, but normal Rosi binding and the Y473A mutant were unable to bind either ligand.
To examine the impact of these mutations on PPAR activation, we examined their transcriptional activity using the PPRE-ACox-Luc reporter gene system. B103 cells were transfected with individual PPAR␥ mutants, the reporter gene construct, and SV40-␤-galactosidase for normalization. The transcriptional activity of wild type PPAR␥ increased 2.5-fold after exposure to 10 M Rosi. In contrast, the H323A, H449A, Y473F, and H323/449A mutants all showed significantly reduced transcriptional activity in response to Rosi (Fig. 6B). The R288A and Y473F mutants showed reduced transcriptional activity in response to 10 M AGP, whereas activation of the H323A, H449A, and H323A/ H449A mutants was indistinguishable from that of the wild type (Fig.  6C). These results highlight distinct interactions between AGP or Rosi with select residues within the PPAR␥-LBD.

DISCUSSION
Recent studies have provided evidence that select forms of lysophosphatidic acid applied extracellularly activate the lipid-regulated transcription factor PPAR␥. LPA, the cognate ligand for four mammalian GPCR, has been viewed primarily as an extracellular mediator (34,38,54,55). AGP has been detected in several vertebrate biological fluids and tissues, including rat brain, human ascitic fluid, saliva, and hen egg (40, 56 -58). In the brain, the steady-state concentration of AGP was 0.44 nmol/g (40), whereas in ascites fluids from ovarian cancer patients AGP was 3.7 nmol/g relative to 0.4 nmol/g in controls (56). Human saliva contains 0.104 nmol/ml AGP (57), and hen egg contains 1.09 nmol/g of egg yolk and 0.02 nmol/g of egg white (58). It is noteworthy that yolk is enriched in AGP and that this concentration is sufficient to activate PPAR␥. Nuclei have the capability to synthesize PA from LPA and AGP and also break it down (59 -63). The significance of the nuclear lipid metabolic pathways is presently unclear because of the paucity of data.
In the present study we examined the interactions between alkyl analogs of LPA and PPAR␥. We showed previously that AGP but not acyl-LPA accumulates in oxidatively modified LDL (35). Likewise, AGP was found to be the most potent activator of PPAR␥-mediated transcription compared with the acyl-LPA species (38). Furthermore, Tokumura et al. (51) have established that 10 -14% of extracellularly applied AGP can be recovered intact, from the cytoplasm, 30 min after application. Recent evidence indicates that PPAR␥ is present in the cytoplasm (64) where it is potentially available for direct interaction with AGP. Therefore, we first examined the quantitative aspects of AGP interaction with the LBD of PPAR␥ to establish whether physiologic concentrations are capable of eliciting activation. We utilized a photoaffinity analog of AGP to examine specific binding to purified PPAR␥-LBD. These experiments revealed time-dependent, specific labeling of PPAR␥-LBD with AGP-BP that was competitively reduced by excess cold AGP or Rosi. By using radioligand binding assays, we determined that AGP 18:1 has high affinity for PPAR␥ with an apparent K d of 60 nM. This value is comparable with PPAR␥ binding of the TZD compound, Rosi, reported in the literature (40 -325 nM) (18,50,65). However, despite the high binding affinity of AGP, competition with Rosi indicated that the two ligands cannot mutually and fully displace each other. A 500-fold excess of various putative endogenous PPAR␥ ligands competed with AGP or Rosi (Fig. 4, A and B) with varying success, revealing substantial differences in efficacies. Long chain (ՆC18) LPA species, oleate, linoleate, and 15-deoxy-PGJ 2 effectively competed with Rosi. In contrast, the latter three lipids did not displace [ 32 P]AGP. Most interestingly, the endogenous phospholipid ligand Az-PC (21) and the LPA analogs were effective in competing with AGP for binding. The Rosi displacement curve showed two distinct parts. In the picomolar to low nanomolar range, AGP was a more effective competitor than Rosi; however, in the micromolar range AGP failed to displace more than ϳ55% of bound TZD. The distinct pattern of displacement by endogenous PPAR␥ ligands combined with the biphasic displacement curve suggests that AGP and Rosi interact differentially with the PPAR␥-LBD. To examine this hypothesis, we applied a computational modeling-guided mutagenesis strategy, which utilized an agonist-bound PPAR␥ crystal structure as the docking target. Computational docking of AGP and LPA analogs with the PPAR␥ binding pocket indicated only partial common ligandoccupied volume and predicted several overlapping and distinct interactions with key ligand binding pocket residues (Fig. 5). Residues His-323, His-449, and the AF2 activation loop residue Tyr-473 were predicted to be in close proximity to both ligands. These residues are essential for TZD binding and activation (52,66). In contrast the AGP head group was shown to make a salt bridge with Arg-288. Alanine replacement mutations confirmed predictions of the model in that AGP binding was selectively diminished in the R288A mutant without any effect on Rosi binding. This finding is consistent with the demonstrated impairment of binding of natural, but not synthetic, ligands to the R288H mutant identified in sporadic colon cancers (53). Alanine replacement of Tyr-473, an essential residue for TZD binding to PPAR␥ (67), abolished binding to both Rosi and AGP, suggesting a common mechanism of activation. Altogether, these findings provide strong evidence in support of the hypothesis that AGP is a high affinity endogenous ligand of PPAR␥. Nonetheless, AGP interacts with a different set of residues in the large ligand binding pocket of PPAR␥.
Our next goal was to assess how the high affinity binding of AGP translates to activation of PPAR␥. To address this question, we tested several cell lines to identify a PPAR␥ and LPA GPCR null background, where using a receptor add-back approach could also be pursued to examine the role of LPA GPCR modulation of PPAR␥ activation. We confirmed that RH7777 hepatoma cells, which lack the four known LPA GPCRs (48) but endogenously express PPAR␥, were activated by both Rosi and AGP (Fig. 2E). This observation strongly supports the hypothesis that LPA GPCRs are not necessary for the activation of PPAR␥ by AGP. We found that the B103 cell line expressed traces of LPA 4 transcript but was devoid of other LPA GPCRS and endogenous PPAR␥ transcripts, hence providing a unique tool to study the effect of LPA GPCRS on PPAR␥ activation. Wild type B103 cells showed no AGP-or Rosi-elicited transcriptional activation of the reporter gene. Heterologous expression of PPAR␥ introduced transcriptional activation to both ligands that was abolished by the specific and irreversible PPAR␥ antagonist GW9662. These results provide strong evidence that AGP can specifically activate transcription of a PPAR␥ target gene in a mammalian cell line. This extends the previous observation by McIntyre et al. (34) obtained in yeast cells that lack all nuclear hormone receptors and  FEBRUARY 10, 2006 • VOLUME 281 • NUMBER 6 contain only a single GPCR-linked receptor for mating factor. They (34) found that LPA directly acts on PPAR␥, rather than stimulating an unknown ligand through an LPA GPCR.

Differential Ligand Recognition by PPAR␥
The lack of LPA 1/2/3 expression in B103 cells permitted the evaluation of the effect of these receptors on PPAR␥ activation. These experiments revealed a pattern of interaction between AGP-and Rosi-induced transcriptional activity of PPAR␥. LPA 1 overexpression significantly augmented the transcriptional activity of PPAR␥ elicited by the simultaneous application of Rosi and AGP. In contrast, LPA 2 and LPA 3 expression significantly attenuated reporter-gene activity elicited by the simultaneous application of Rosi and AGP (data not shown). The observation that LPA 1 expression synergizes with PPAR␥ activation is consistent with our previous observation that uncoupling of LPA 1 signaling by pertussis toxin partially attenuated the LPA-induced PPAR␥ transcriptional activity in vitro and in vivo (38). These results are in sharp contrast to the report by Simon et al. (68), who showed an attenuation of PPAR␥ activity in adipocytes treated with LPA 18:1. Because of the complexity and many possible points of interaction between the GPCR and PPAR␥ signaling pathways (69), dissection of this mechanism is beyond the scope of the present study.
High affinity binding of AGP to PPAR␥ is supported by the reporter gene activation data, indicating that as little as 10 nM AGP or Rosi were effective in eliciting a modest but significant transcriptional activation of the reporter gene in B103 cells (Fig. 4D). This is in stark contrast to previously identified, putative endogenous PPAR␥ ligands, including Az-PC, oleic acid, 15-deoxy-PGJ 2 , and LPA 18:1, which were not effective at this low concentration. Based on these findings, we propose AGP be considered as an additional endogenous ligand of PPAR␥. The efficacy of AGP does not match that of full agonist TZD analogs Rosi (Fig.  4C), pioglitazone, or troglitazone (data not shown). AGP is a partial agonist compared with the TZD compounds, which are clearly more efficacious than any endogenous ligand known to date. Nevertheless, in terms of potency AGP and Rosi show similar levels of activation within the low nanomolar concentration range. Our mutagenesis studies support an hypothesis that the differential interactions with key residues in the PPAR␥-LBD are responsible for the partial agonist properties of AGP as compared with the full agonist Rosi. The computational model will allow us to design novel AGP analogs that show more favorable interactions with the residues uniquely interacting with TZD full agonists, such as His-323 and -449. We predict that AGP analogs can be designed that will have efficacy comparable with TZD. It is important that we acknowledge a new trend in the PPAR␥ field that puts more emphasis on the use of partial agonists because of reduced side effects (23). By taking this into account, AGP might serve as a useful scaffold for lead optimization studies. Another important consideration for future investigation is based on the well established differences in signaling properties and gene activation patterns between full versus partial PPAR␥ agonists (70). In this context, it is expected that AGP activates a different set of PPAR␥ target genes, because of differences in the recruitment of co-activators and co-repressors as compared with full agonists. Again, experimental testing of this hypothesis is beyond the scope of the present study.
Previously we found that AGP content was elevated 6-fold upon mild oxidation of LDL, with the octadecenyl species showing a 10-fold increase over native LDL (38). Most interestingly, the rank order of AGP species present in mox-LDL was the same as that reported in the lipid core of human atherosclerotic plaques (35). These observations serve as important corollaries to the hypothesis that endogenous ligands of PPAR␥ regulate cellular events within the vessel wall during atherogenesis. In conclusion, we propose that AGP should be considered as endogenous ligand of PPAR␥. The simultaneous activations of PPAR␥ and LPA GPCR are likely to lead to complex cellular responses that will be deciphered in future studies.