The Antidiabetic Agent LG100754 Sensitizes Cells to Low Concentrations of Peroxisome Proliferator-activated Receptor γ Ligands*

Insulin resistance and non-insulin-dependent diabetes mellitus are major causes of morbidity and mortality in industrialized nations. Despite the alarming rise in the prevalence of this disorder, the initial molecular events that promote insulin resistance remain unclear. The data presented here demonstrate that LG100754, an antidiabetic RXR ligand, defines a novel type of nuclear receptor agonist. Surprisingly, LG100754 has minimal intrinsic transcriptional activity, instead it enhances the potency of proliferator-activated receptor (PPAR) γ-retinoid X receptor heterodimers for PPARγ ligands. The ability of LG100754 to both increase PPARγ sensitivity and relieve insulin resistance implies that a deficiency in endogenous PPARγ ligands may represent an early step in the development of insulin resistance.

Insulin resistance and non-insulin-dependent diabetes mellitus (NIDDM) 1 have reached epidemic status in industrialized societies (1,2). Over 125 million people worldwide suffer from NIDDM, and these individuals face a dramatically increased risk for developing atherosclerotic heart disease, stroke, renal disease, blindness, and limb amputations. It is thus alarming that the number of NIDDM cases have increased 5-fold in the past decade, a trend that is predicted to continue. Equally worrisome is that NIDDM, initially defined as a disease of adult onset, is now appearing in adolescents.
Insulin responsiveness can be modulated by a number of processes including transcriptional cascades controlled by nuclear hormone receptors. Nuclear receptors comprise a superfamily of transcription factors that directly regulate gene expression in response to low molecular weight ligands. Upon binding these ligands, receptors undergo a conformational change that promotes an exchange of coregulatory proteins and ultimately a change in the rate of transcription of specific target genes (3). Compounds that bind to and activate the PPAR␥ subunit of the PPAR␥-RXR nuclear receptor heterodimer (4 -6) alter transcription of genes involved in glucose and lipid homeostasis. Included among these target genes are lipid transporters (CD36, aquaporin), key metabolic enzymes (lipoprotein lipase, phosphoenolpyruvate carboxykinase, uncoupling protein-1), adipocyte-enriched signaling molecules (leptin, resistin, ACRP30, FIAF/PGAR), lipid-modulated nuclear receptors (LXR␣), and an intermediate in the insulin signaling pathway (c-Cbl-associating protein) (2,5,(7)(8)(9)(10)(11)(12)(13)(14)(15). A variety of cellular, molecular, and pharmacologic studies have shown that PPAR␥ activation results in increased adipogenesis, redistribution of fatty acids and triglycerides into fat, and ultimately improved insulin sensitivity (2,6,16). Indeed, PPAR␥-specific ligands such as rosiglitazone are currently used for the clinical treatment of NIDDM (17).
PPAR␥ functions as part of a heterodimeric complex with a nuclear receptor known as RXR (4 -6). RXR serves as a common heterodimeric partner for several nuclear receptors and is modulated by a class of ligands known as rexinoids. Since PPAR␥ functions as an obligate heterodimer with RXR, there has been an interest in developing RXR-specific rexinoids as potential treatments for NIDDM. A particularly interesting compound is LG754, which primarily activates PPAR-RXR heterodimers and retains potent antidiabetic properties (18). We now show that LG754 defines a novel rexinoid agonist that paradoxically has little intrinsic transcriptional activity. Instead, LG754 functions by enhancing the affinity of PPAR␥ for its ligands.
LG754 thus defines a new class of receptor agonist that can be described as a ligand sensitizer. The fact that this PPAR␥ sensitizer relieves insulin resistance suggests that a relative deficiency in endogenous PPAR␥ ligands may play a primary role in the development of insulin resistance. This notion accounts for several critical paradoxes in our understanding of NIDDM.

MATERIALS AND METHODS
Plasmids and Reagents-The PPAR␥ luciferase reporter construct PPRE ϫ 3 TK-Luc contains the herpesvirus thymidine kinase promoter (Ϫ105/ϩ51) linked to three copies of the rat acyl-CoA oxidase PPRE (5Ј-AGGGGACCAGGACAAAGGTCACGTTCGGGA-3Ј). The GAL4 reporter was as described previously (19). A cytomegalovirus expression vector with a T7 promoter was used to express the following proteins in cells and/or in vitro: PPAR␥ (mouse PPAR␥1, GenBank TM accession number U10374), RXR (human RXR␣, GenBank TM accession number X52773), Gal-PBP (human PBP, GenBank TM accession number AF283812, Val 574 -Ser 649 ), VP-PPAR␥ (mouse PPAR␥ ligand binding domain, GenBank TM accession number U10374, Cys 163 -Tyr 475 ). For two-hybrid studies, an RXR ligand binding domain expression vector was used that contains the SV40 TAg nuclear localization signal (AP-KKKRKVG) fused upstream of the RXR ligand binding domain (human RXR␣, GenBank TM accession number X52773, Glu 203 -Thr 462 ). Gal4 fusions contained the indicated fragments fused to the C-terminal end of the yeast Gal4 DNA binding domain; VP16 fusions contained the 78 amino acid herpesvirus VP16 transactivation domain. For bacterial expression, p160 coactivator proteins were expressed as fusion proteins containing GST upstream of the 3 receptor interaction domains of SRC-1 (human SRC-1, GenBank TM accession number U59302, Asp 617 -Asp 769 ), ACTR (human ACTR, GenBank TM accession number AF036892, Gly 615 -Gln 768 ) or GRIP1 (mouse GRIP1, GenBank TM accession number U39060, Arg 625 -Lys 765 ). The PBP fusion contained the two * 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.

RESULTS
Cesario et al. (18) have shown that the RXR ligand LG754 is a selective activator of PPAR-RXR heterodimers. As expected for a PPAR␥-RXR agonist, LG754 was also shown to promote adipogenesis in 3T3-L1 cells and to relieve insulin resistance in db/db mice. This prompted us to directly compare the extent of PPAR␥ activation by LG754 and known PPAR␥ ligands. To determine the relative efficacy of LG754, we transiently transfected RXR-expressing CV-1 cells with PPAR␥ reporter and expression vectors and compared the activation of LG754 with that of rosiglitazone, a standard PPAR␥ agonist (20) (Fig. 1A). Rosiglitazone (1 M) produced the expected strong activation (28-fold) of PPAR␥, but the RXR ligand LG754 (1 M) yielded only a 5-fold activation. These findings were unexpected in light of the fact that LG754 effectively mimics the biological effects of PPAR␥ ligands in vivo.
Nuclear hormone receptors activate transcription by recruiting transcriptional coactivator proteins. Using a mammalian two-hybrid assay, Cesario et al. (18) demonstrated that LG754 specifically recruits the coactivator PBP (also known as TRAP220 and DRIP205) to the PPAR␥-RXR heterodimer. Since LG754 had less intrinsic transcriptional activity than PPAR␥ ligands (Fig. 1A), we compared the relative ability of LG754 and rosiglitazone to recruit PBP using the same two-hybrid assay described above. In this system, reporter expression is activated if the herpesvirus VP16 transactivation domain becomes tethered to the promoter via a PPAR-coactivator interaction. Using this assay, we found that rosiglitazone produced a strong 10-fold increase in the recruitment of PPAR␥ to PBP, whereas LG754 had only a 2-3-fold effect on this interaction (Fig. 1B). The relative differences in coactivator recruitment by these two ligands closely paralleled the weak effect of LG754 on activation of the PPAR␥-RXR heterodimer (Fig. 1A).
To further explore the effect of LG754 on coactivator recruitment, we utilized an electrophoretic mobility shift assay. Unlike the two-hybrid assay, this approach examines the effect of ligands on DNA-bound PPAR␥-RXR heterodimers, i.e. native receptor complexes. Thus, PPAR␥, RXR, and affinity-purified GST-coactivator fusion proteins were incubated with a 32 Plabeled response element and separated by electrophoresis through a nondenaturing gel. In this experiment, we compared LG754 to LG268, another RXR-specific rexinoid with antidiabetic activity (22). LG268 effectively recruited the p160 family of coactivators (SRC1, ACTR, GRIP) but had no effect on the GST control (Fig. 2). A quantitatively smaller, but highly reproducible, shift was also seen with PBP (Fig. 2). This demonstrates that RXR ligands can recruit coactivators to the DNAbound PPAR␥-RXR heterodimer. In contrast, LG754 failed to recruit any of these proteins (Fig. 2), confirming that LG754 does not effectively recruit coactivators to the PPAR␥-RXR complex. Thus, in contrast to previously described nuclear receptor agonists, LG754 is unique in that does not possess strong transcriptional activating properties.
Since LG754 has only weak intrinsic transcriptional activity, its ability to mimic PPAR␥ ligands in vivo implies that this compound is functioning by an alternative mechanism. Previous studies have demonstrated that individual subunits of nuclear receptor heterodimers can have dramatic allosteric effects on its partner receptor. For example, the insect ecdysone receptor does not bind its ligand with high affinity, instead ligand binding requires association of the ecdysone receptor with its heterodimeric partner, ultraspiracle (USP) (23). The existence of such allosteric interactions among receptor heterodimers prompted us to ask whether LG754 can increase the affinity of the PPAR␥-RXR complex for PPAR␥ ligands. Cell-based transfection assays were used to examine the effect of LG754 on rosiglitazone-mediated activation of PPAR␥. CV-1 cells were transfected with PPAR␥ reporter and expression vectors and treated with suboptimal amounts of rosiglitazone (60 nM) in the presence of increasing amounts of LG754 (Fig. 3A). As expected, rosiglitazone (60 nM) activated PPAR␥, but notably this activation was further enhanced by LG754. The optimal effect of LG754 was seen at a concentration of 1 M (Fig. 3A), which is similar to the optimal doses required for adipogenesis in 3T3-L1 cells (18). We next examined the effect of LG754 on the PPAR␥ dose-response curve. As expected from Fig. 1A, LG754 had minimal activity by itself. However, it shifted the rosiglitazone dose response curve leading to a 3-fold increase in the apparent potency of rosiglitazone (Fig. 3B). LG754 had a similar effect on the  2. LG754 fails to effectively recruit coactivators to DNAbound PPAR␥-RXR heterodimers. Electrophoretic mobility shift complexes using a 32 P-labeled PPAR response element (acyl-CoA oxidase) in vitro translated PPAR␥-RXR and affinity-purified GST (control) or GST-coactivator fusion proteins. LG268 (100 nM) and LG754 (1 M) were added as indicated. The position of the PPAR␥-RXR () and PPAR␥-RXR-coactivator (‹) complexes are shown. potency of other PPAR␥ ligands including 15-deoxy-⌬ 12,14prostaglandin J 2 (20) (Fig. 3C). These findings suggest that LG754 increases the affinity of PPAR␥ for its ligands.
To assess the effect of LG754 on the ligand binding activity of PPAR␥, an assay was required that can efficiently measure ligand binding to PPAR␥-RXR heterodimers. Standard solution-based binding assays are useful for measuring binding to PPAR␥ monomers. While such assays can be performed in the presence of RXR, it can be difficult to ensure that all PPAR␥ molecules in the reaction are complexed with RXR. Therefore, solution-based assays may measure binding to a mixed population of PPAR␥ monomers and PPAR␥-RXR heterodimers. To overcome this limitation, ligand binding was measured in a modified electrophoretic mobility shift assay where heterodimer-containing complexes can be separated and identified within the gel. Complexes were analyzed as in Fig. 2, but instead of using radiolabeled DNA, the complexes were visualized with a 125 I-labeled PPAR␥ ligand (21). Note that LG754 significantly increased the amount of 125 I-labeled PPAR␥ ligand in the complex (Fig. 3D) without affecting the total amount of complex formed (Fig. 2 and data not shown). These data suggest that LG754 enhances the affinity of PPAR␥ for its ligands. DISCUSSION A large body evidence of biochemical, structural, and genetic data have firmly demonstrated that nuclear receptors activate transcription by recruiting transcriptional coactivator proteins (3). This has led to the commonly accepted paradigm that nuclear receptor agonists function by inducing a conformation change that favors a more stable receptor-coactivator complex. The data presented here indicate that LG754 defines a new class of nuclear receptor agonist that has minimal coactivator recruitment activity and therefore minimal inherent transcriptional activity. Instead, this compound activates transcription by allosterically enhancing the ligand binding activity of its partner receptor, PPAR␥. We refer to this agonist class as a "sensitizer." LG754 therefore represents the first example of a nuclear receptor-sensitizing agent.
In addition to being a PPAR␥ sensitizer, previous studies have demonstrated that LG754 relieves insulin resistance in vivo (18). These findings have important implications, since the molecular events that result in insulin resistance remain obscure. It is well known that PPAR␥ ligands reverse insulin resistance both in humans and in a variety of animal models. PPAR␥ agonists have the interesting property of lowering glucose in diabetic animals but not in non-diabetic animals (24). This implies that PPAR␥ ligands reverse or replace a deficiency that is unique to the diabetic state. It is intriguing to speculate that insulin resistance arises from a relative deficiency in endogenous PPAR␥ ligands and that PPAR␥ agonists are effective antidiabetic agents, because they correct this deficiency. In principle, this hypothesis could be directly tested by determining the levels of endogenous PPAR␥ ligands in normal and diabetic individuals. Several fatty acid derivatives and prostanoids have been shown to bind to PPAR␥ (4,20); however, these ligands are not specific for PPAR␥, and the precise identity of the endogenous PPAR␥ ligand is unknown (25). Therefore, a direct quantitation of endogenous PPAR␥ ligand levels is not currently possible and an alternative approach is required to test this "PPAR␥ ligand deficiency" model. Since LG754 is a PPAR␥ sensitizer, this compound reverses the biological effects that result from a deficiency in PPAR␥ ligands. The previous demonstration that LG754 relieves insulin in db/db mice (18) provides support for the hypothesis that insulin resistance is secondary to suboptimal levels of the yet-to-be identified PPAR␥ ligand.
A genetic test of the "ligand deficiency" hypothesis might include the development of animals expressing PPAR␥ mutants with diminished ligand binding affinity. These animals would not respond to endogenous PPAR␥ ligands and would be predicted to develop insulin resistance. Such animals have not been described, although PPAR␥-null mice have been established (26,27). However, as PPAR␥ integrates both positive (endogenous ligands) and negative signals (MAP kinase) (28), the effects of chronic PPAR␥ ablation cannot be equated with those resulting from ligand deficiency. While appropriate animal models do not exist, several patients have been described with point mutations in PPAR␥ that result in defects in ligand binding and/or transactivation (29). These individuals provide insights into the pathological consequences associated with a diminished response to endogenous PPAR␥ ligands. Indeed, these patients develop lipodystrophy and severe insulin resistance as would be predicted by the PPAR␥ ligand-deficiency hypothesis (Fig. 3E).
There are a number of critical gaps in our understanding of the NIDDM-PPAR␥ connection. For example, PPAR␥ is required for adipogenesis (27,30,31), and its synthetic agonists increase adipose mass in vivo (24). This is unexpected, since insulin resistance worsens in most patients as fat mass in- creases. This raises a question as to how an adipogenic agent can also act as an antidiabetic agent (32)? Another gap is highlighted by the fact that certain rare forms of NIDDM are paradoxically associated with diminished fat mass (lipodystrophy) (33), and perhaps even more surprising is the observation that PPAR␥ activators can effectively treat both obesity-dependent and lipodystrophic diabetes (6,33).
The PPAR␥ ligand deficiency hypothesis (Fig. 3E) is intriguing as it provides a rationale to close these gaps. Given the dual role of PPAR␥ ligands in enhancing adipose mass and insulin responsiveness, I suggest that a primary defect in the synthesis or accumulation of an endogenous PPAR␥ ligand might be the molecular event that underlies lipodystrophic diabetes. Although the identity of the endogenous PPAR␥ ligand is unknown, it has been suggested that the transcription factor SREBP-1c (ADD1) is required to produce an endogenous ligand in adipocytes (34). The PPAR␥ ligand deficiency hypothesis would predict that treatments which lower SREBP-1c levels should result in lipodystrophy, insulin resistance, and decreased adipogenesis. Indeed, two human immunodeficiency virus protease inhibitors (indinavir and nelfinavir) that promote lipodystrophic diabetes in vivo (35) have been shown to inhibit adipogenesis and to reduce SREBP-1 activity (36,37). These observations provide further support for the PPAR␥ ligand deficiency hypothesis.
In the opposing state of obesity, it is reasonable to imagine that feedback mechanisms are triggered in an attempt to limit further lipid storage. In principle, this could be accomplished by decreasing lipogenic signals such as those represented by endogenous PPAR␥ ligands. Indeed, several groups have shown that SREBP-1c levels are lower in obesity (38 -40), suggesting that the obese state may be associated with a corresponding decrease in endogenous PPAR␥ ligands. While this response may provide short term benefits, a chronic deficiency in PPAR␥ ligands could eventually lead to the development of insulin resistance. In effect, lipodystrophy may represent a primary PPAR␥ ligand deficiency, whereas in obesity-related diabetes the deficiency would be secondary to increasing fat mass. Thus, the ligand deficiency hypothesis accounts for the paradoxical association of NIDDM with both obesity and lipodystrophy. It also explains how PPAR␥ ligands can treat both disorders. Given the rising toll of NIDDM, these findings provide further impetus to identify and quantitate the endogenous PPAR␥ ligand.