A Potent Antidiabetic Thiazolidinedione with Unique Peroxisome Proliferator-activated Receptor g -activating Properties*

Thiazolidinediones (TZDs) constitute an exciting new class of antidiabetic compounds, which function as activating ligands for peroxisome proliferator-activated receptor g (PPAR g ). Until now, there has been an excel-lent correlation between in vivo hypoglycemic potency and in vitro binding and activation of PPAR g by TZDs. We have characterized MCC-555, a novel thiazolidinedione ligand for PPAR g with unique functional properties. The antidiabetic potency of this compound is greater than that of other TZDs, including BRL49653, yet its binding affinity for PPAR g is less than 1 ⁄ 10 that of BRL49653. The effect of MCC-555 binding on PPAR g transcriptional activity is highly context-specific such that it can function as a full agonist, partial agonist, or antagonist depending on the cell type or DNA binding site. These transcriptional properties are partly ex-plained by unique partial agonism of coactivator recruitment to PPAR g . The properties of MCC-555 are mechanistically distinct from those of the estrogen receptor partial agonist and antagonist tamoxifen because the N terminus of PPAR g is not required for activation by MCC-555, and MCC-555 does not stimulate corepressor recruitment to PPAR g . The context selectivity of MCC-555 may contribute to its enhanced hypoglycemic

Thiazolidinediones (TZDs) constitute an exciting new class of antidiabetic compounds, which function as activating ligands for peroxisome proliferator-activated receptor ␥ (PPAR␥). Until now, there has been an excellent correlation between in vivo hypoglycemic potency and in vitro binding and activation of PPAR␥ by TZDs. We have characterized MCC-555, a novel thiazolidinedione ligand for PPAR␥ with unique functional properties. The antidiabetic potency of this compound is greater than that of other TZDs, including BRL49653, yet its binding affinity for PPAR␥ is less than 1 ⁄10 that of BRL49653. The effect of MCC-555 binding on PPAR␥ transcriptional activity is highly context-specific such that it can function as a full agonist, partial agonist, or antagonist depending on the cell type or DNA binding site. These transcriptional properties are partly explained by unique partial agonism of coactivator recruitment to PPAR␥. The properties of MCC-555 are mechanistically distinct from those of the estrogen receptor partial agonist and antagonist tamoxifen because the N terminus of PPAR␥ is not required for activation by MCC-555, and MCC-555 does not stimulate corepressor recruitment to PPAR␥. The context selectivity of MCC-555 may contribute to its enhanced hypoglycemic potency in vivo despite reduced affinity for PPAR␥ relative to other TZDs.
Nuclear hormone receptors (NHRs) 1 constitute a class of transcription factors with activity that is regulated by natural or synthetic lipophilic ligands (1). A number of NHRs are involved in developmental and/or metabolic processes, and modulation of NHR activity is an effective strategy in the treatment of a variety of cancers, such as breast cancer (2), prostate cancer (3), and acute promyleocytic leukemia (4), as well as metabolic diseases including thyroid disease (5) and diabetes. Non-insulin-dependent diabetes mellitus is a major cause of morbidity and mortality in industrialized nations and is characterized by a post-insulin receptor defect that has been difficult to target therapeutically until the recent discovery that thiazolidinediones (TZDs) enhance the actions of insulin at a level distal to the insulin receptor (6).
The mechanism of TZD action is not completely understood, but a number of lines of evidence point to their function as ligands for a member of the NHR superfamily called peroxisome proliferator-activated receptor ␥ (PPAR␥), the natural ligand of which may be derived from or related to prostaglandin J2 (7-9). One of the most potent TZDs, BRL49653, binds to PPAR␥ with an affinity in the nanomolar range (10), and the rank order of TZD potency for in vivo plasma glucose lowering correlates well with their rank order potency for PPAR␥ activation (11,12). Nevertheless, a number of questions remain with regard to the mechanism of TZD potentiation of insulin action. The main problem is that PPAR␥ is primarily expressed in adipose tissue (13,14), whereas muscle is ordinarily the main site of insulin-dependent glucose disposal in mammals. This apparent paradox has not yet been satisfactorily resolved.
We have studied a novel TZD called MCC-555 that is more potent than BRL49653 in vivo yet has lower affinity for PPAR␥. Moreover, MCC-555 binding activates transcription by the PPAR␥ ligand-binding domain (LBD) differently in different contexts, in part because of novel effects on coactivator recruitment. This raises the possibility that only a subset of the functions of activated PPAR␥ contributes to insulin sensitivity, and that therapeutic strategies for non-insulin-dependent diabetes mellitus that target these functions may lead to compounds with increased potency and decreased toxicity.

MATERIALS AND METHODS
Hypoglycemic Potencies of TZDs-Male KK-A y mice obtained from CLEA Japan (Tokyo, Japan) were housed in individual cages in an animal room and given laboratory chow (MF, Oriental Yeast Co. Ltd.) and sterile water ad libitum. They were used for the experiment at the age of 11 weeks. MCC-555, pioglitazone, and BRL49653 were synthesized in crystalline forms, and purity was confirmed using standard criteria. Each test drug was suspended in 0.5% sodium carboxymethyl cellulose solution and administered orally for 4 days at a dose volume of 10 ml/kg of body weight. The control mice were given the sodium carboxymethyl cellulose solution in the same manner. Blood was collected in heparinized hematocrit capillary tubes under fed state on the day after the last administration. The capillary tubes were centrifuged, and plasma samples were separated. Plasma glucose levels were determined by a glucose oxidase method with a commercial kit (Glu Neo Shino-Test, Shino-Test Co.). The significance of differences between the control and each treatment were assessed by analysis of variance followed by Dunnett's two-tailed test. The dose that produces a 25% decrease in plasma glucose level from the control (ED 25 ) was calculated by linear regression from the concentration-related efficacy curve. The relative hypoglycemic activities were estimated by the parallel line assay.
Northern Analysis-8-week old, male KK-A y mice were used. For * 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. § Present address: Dept. of Cell Biology, Harvard Medical School, Boston, MA 02115.
RNA determinations, MCC-555 and BRL49653 (30 mg/kg, suspended in 0.5% sodium carboxymethyl cellulose solution) were administered orally daily for 4 days. On day 5, mice were killed, and epididymal fat pads were removed and quickly frozen in liquid nitrogen. mRNA was prepared as described previously by homogenizing epididymal fat from KK-A y mice with a Polytron homogenizer in 7 ml of 4 M guanidine thiocyanate, 20 mM sodium citrate (pH 7.0), 17 mM N-lauroyl sarcosine, and 100 mM ␤-mercaptoethenol. After centrifugation (10,000 rpm for 10 min at 12°C) the supernatant was sheared through a 20-gauge syringe three times. Total RNA was isolated from homogenates, and Northern blots were performed as described previously (15). The cDNA probes for aP2, PPAR␥, and leptin were labeled with 32 P by using random hexamers.
Coactivator Recruitment Assay-The coactivator-dependent receptor ligand assay was performed as initially described by Krey et al. (18) with modifications described below. Briefly, GST fusion proteins were purified with glutathione-Sepharose beads (Amersham Pharmacia Biotech). GST fusion proteins bound to beads (100 l of a 50% slurry) in GST binding buffer (50 mM KCl, 20 mM HEPES, pH 7.9, 2 mM EDTA, 0.1% Nonidet P-40, 10% glycerol, 0.5% nonfat dry milk, 5 mM dithiothreitol) was mixed with 5 l of radiolabeled in vitro-translated SRC-1 and indicated ligand concentration. Ligand or vehicle control made up Ͻ2% of the total volume in all experiments. Fusion proteins, SRC-1, and ligand were incubated at 4°C for 1 h with gentle rocking. The beads were then vigorously washed with 1 ml of GST binding buffer. Bound proteins were eluted by boiling in 20 l of 2 ϫ SDS loading buffer and analyzed by SDS-polyacrylamide gel electrophoresis. The GST fusion proteins were stained with Coomassie Blue for every experiment to ensure equal loading, and bound proteins were visualized by autoradiography and quantitated using a phosphorimager. Corepressor interaction experiments were performed under identical conditions using GST and GST-silencing mediator of retinoid and thyroid receptor (SMRT) receptor interaction domain (amino acids 982-1485) as described previously (33), except that 10 nM thyroid hormone or 50 M BRL49653 and MCC-555 were added as indicated.
Binding Assay-Ligand binding assay was performed essentially as described (8). Briefly, GST-PPAR␥ fusion protein isolated from Escherichia coli DH5␣ cells was pelleted and lysed by sonication in Tris-EDTA-NaCl buffer (10 mM Tris-HCl (pH 8.0), 1 mM EDTA, 100 mM NaCl). For competition binding assays, bacterial extracts (ϳ300 g of protein) were incubated at 4°C for 2 h with 200 nM [ 3 H]BRL49653 (specific activity, 18 Ci/mmol) in the absence or presence of unlabeled TZD in buffer containing 10 mM Tris (pH 8.0), 50 mM KCl, and 10 mM dithiothreitol. Bound was separated from free radioactivity by elution through 1-ml Sephadex G-25 desalting columns (Amersham Pharmacia Biotech). Bound radioactivity eluted from the column void volume was quantitated using a liquid scintillation counter.
Transient Transfection Transcription Assay-293T cells and JEG-3 cells were maintained in high glucose Dulbecco's modified Eagle's medium containing 10% fetal bovine serum and then switched to high glucose Dulbecco's modified Eagle's medium with 10% charcoalstripped fetal calf serum just before transfection. All transfections were normalized for total DNA and promoter concentration. All transfection experiments except the PPAR␥ ⌬N experiment (see Fig. 7A) were performed in 60-mm-diameter dishes using the calcium phosphate precipitation method with 1 g of receptor expression vectors, 1 g of luciferase reporter, and 0.5 g of ␤-galactosidase (␤-gal) expression vector. The calcium phosphate transfection experiments shown in Fig. 7A were performed using 24-well plates, with 20 ng of receptor expression vector, 50 ng of luciferase reporter, and 50 ng of ␤-gal vector. BRL49653, MCC-555, or vehicle control (dimethyl sulfoxide) were added 16 h after transfections at concentrations described in legends. The following day, cells transfected were lysed in Triton X-100, and cell lysates were subjected to luciferase and ␤-gal assays. Results were normalized to ␤-gal activity, and fold activation was calculated.

MCC-555 Is a Potent Antidiabetic Thiazolidinedione-
The structure of MCC-555 is shown and compared with that of BRL49653 and other TZDs in Fig. 1A. Fig. 1B shows that MCC-555 was more effective than BRL49653 or other TZDs in lowering plasma glucose in the mouse KK-A y model of obesity and non-insulin-dependent diabetes mellitus. As indicated in the Fig. 1, the ED 25 for MCC-555 was nearly 3-fold lower than that of BRL49653, the most potent of other TZDs tested (20).
MCC-555 Is an Activating Ligand for PPAR␥-Antidiabetic TZDs have been shown to function as activating ligands for PPAR␥, and the abilities of TZDs to activate PPAR␥ and to reduce plasma glucose are well correlated (11,12). Using the well characterized PPAR binding site from the acyl-coenzyme A oxidase gene upstream of a luciferase reporter, MCC-555 activated PPAR␥-dependent transcription in a transient transfection assay, with maximal activation approximately equal to that of BRL49653 (Fig. 2). However, despite its greater antihyperglycemic potency, MCC-555 was less potent than BRL49653 in this activity, with an EC 50 nearly 10-fold greater (Fig. 2). Using a direct binding displacement assay, we confirmed that MCC-555 binds directly to PPAR␥ (Fig. 3). However, the ability of MCC-555 to bind to PPAR␥ was more than a log order less than that of BRL49653; the EC 50 of MCC-555 was 8 M, whereas that of BRL49653 was 200 nM, similar to that reported by others (9,10). This agreed well with the order of potency in PPAR␥ activation assays but was reversed relative to the in vivo antidiabetic potencies of the TZDs.
MCC-555 Is a Partial Agonist and/or Antagonist of Chimeric Forms of PPAR␥-One of the cardinal features of members of the NHR superfamily is the ability of LBD to function independently when fused to a heterologous polypeptide (21). This feature has been exploited for many NHRs, including PPAR␥. Indeed, the binding of ligand to PPAR␥ observed by others and in the present work involves the fusion of the LBD to glutathione S-transferase (8 -10). Moreover, much of the work on liganddependent activation of PPAR␥ is derived from studies fusing the PPAR␥ LBD to a heterologous DNA-binding domain (DBD), most frequently that of Gal4. We therefore decided to compare the properties of MCC-555 and BRL49653 in the context of the Gal4-PPAR␥ fusion protein. Fig. 4A shows that both MCC-555 and BRL49653 activated transcription by Gal4-PPAR␥, and the EC 50 value for the compounds again were consistent with the reduced binding affinity of MCC-555. However, unlike the activation of wild-type PPAR␥ by MCC-555, the maximal activation achieved by MCC-555 was consistently ϳ2 fold less than that caused by saturating concentrations of BRL49653. Thus, in this context, MCC-555 functioned as a partial agonist of PPAR␥ activation. This predicted that at high concentration MCC-555 would displace BRL49653 and function as an inhibitor of BRL49653-stimulated activation of Gal4-PPAR␥. Indeed, Fig. 4B shows that MCC-555 antagonized BRL49653-dependent activation when present at a very high relative concentration. This effect was not attributable to destabilization of the Gal4-PPAR␥ chimera, because Western analysis with Gal4 antibody confirmed equal concentrations of the protein in the presence and absence of MCC-555 (data not shown).
The behavior of MCC-555 as a partial agonist of Gal4-PPAR␥ contrasted with its full activation of wild-type PPAR␥. To determine whether this was in some way an artifact of the chimeric receptor, we fused the PPAR␥ LBD to an alternate DBD from the tetracycline operon. This chimeric receptor was previously used to analyze prostaglandin activators of PPAR␥ (7). Fig. 5A shows that tet-PPAR␥-dependent transcription was markedly stimulated by BRL49653. In contrast, however, MCC-555 was not at all effective until near millimolar concentrations were provided (Fig. 5B). Moreover, Fig. 5C shows that in the presence of activating concentrations of BRL49653, MCC-555 now functioned essentially as a pure antagonist. This effect of MCC-555 was attributable to competitive binding to PPAR␥ rather than interference with expression of the chimeric PPAR␥, because the tet-PPAR␥ remained able to interact with retinoid X receptor as demonstrated by mammalian twohybrid assay (Fig. 5D).

MCC-555 Functions as a Partial Agonist in Recruitment of Coactivators to PPAR␥-
The paradigm of ligand-dependent activation by NHRs suggests that one or more coactivators are recruited to the LBD in the presence of ligand (22)(23)(24). Fig. 6A shows that a saturating concentration of BRL49653 indeed caused recruitment of two coactivators, cAMP response element-binding protein (CBP) (25) and SRC-1 (26), to the PPAR␥ LBD. The results with a saturating concentration of MCC-555 are shown for comparison. Note, however, that the magnitude of coactivator binding, especially SRC-1, in Fig. 6A is less in the presence of saturating concentration of MCC-555 than of BRL49653. This is seen better in the dose-response study for ligand-induced recruitment of SRC-1 to GST-PPAR␥ shown in Fig. 6B. As predicted from the reduced affinity of MCC-555 for the PPAR␥, the EC 50 for maximal SRC-1 binding was approximately one log greater for MCC-555 than for BRL49653. However, even at saturating concentrations of MCC-555, the maximum amount of SRC-1 recruited to the same amount of GST-PPAR␥ is remarkably less than that recruited by saturating concentrations of BRL49653. Phosphorimager quantitation of three similar experiments indicates that maximal binding of SRC-1 to GST-PPAR␥ was ϳ5-fold greater in the presence of BRL49653 than MCC-555 (Fig. 6C). We propose this to be the novel mechanism whereby MCC-555 functions as a variable partial agonist and antagonist of PPAR␥ LBD function.
The ability of MCC-555 binding to PPAR␥ to recruit both CBP and SRC-1 contrasts with the mechanism of action of a widely studied estrogen receptor (ER) partial agonist, tamoxifen. Tamoxifen is unable to induce the conformational change required for the ER LBD to bind coactivator (23,27). Indeed, tamoxifen binds competitively with estrogen to the ER LBD but fails to activate through the activation function 2 (AF2) activation domains. Transcriptional activation by tamoxifen requires the N-terminal A/B domain of the ER that is not present in Gal4 fusions (28). In contrast, MCC-555 activation of full length PPAR␥ on PPREs does not require the N-terminal A/B domain of PPAR␥. Fig. 7A shows that although the ED 50 of MCC-555 was higher than for BRL49653, consistent with its reduced affinity for PPAR␥, saturating concentrations of the two ligands were equally effective in activating PPAR␥ ⌬N (lacking amino acids 1-118).
Another property that contributes to the ability of tamoxifen to function as an antagonist for the ER is stabilization of corepressor binding to the ER LBD (29 -32). We have previously shown that PPAR␥ interacts in solution with nuclear receptor corepressor and SMRT (33), and SMRT has been suggested to be involved in PPAR␥ function (30) . Fig. 7, B and C, shows that PPAR␥ interacted with the receptor interaction domain of SMRT to a considerably lesser extent than thyroid hormone receptor and that, unlike triiodothyronine binding to thyroid hormone receptor, BRL49653 binding did not lead to significant displacement of SMRT from PPAR␥. Moreover, MCC-555 did not enhance the binding of SMRT to PPAR␥. Similar results were obtained with nuclear receptor co-repressor (N-CoR) (data not shown). Thus there are three major differences between the properties of MCC-555 and tamoxifen on PPAR␥ and ER, respectively: (i) MCC-555 recruits coactivators to PPAR␥, albeit to a lesser degree than BRL49653; (ii) transcriptional activation by MCC-555 does not require the A/B domain of PPAR␥; and (iii) MCC-555 does not enhance PPAR␥ binding to corepressors.
Potency of MCC-555 in Adipocyte Differentiation-We considered the possibility that reporter assays might not reflect the true potency of MCC-555. Therefore we assessed the ability of MCC-555 to induce adipocyte differentiation of 3T3-L1 cells, A mammalian two-hybrid experiment was performed with tet-luciferase reporter, using cotransfected retinoid X receptor-VP16 as described previously (45). 293T cells were transfected with 1 g of reporter plasmid tet-TKluciferase and pSG5-tet-PPAR␥ LBD and then treated with indicated TZD or appropriate vehicle control (dimethyl sulfoxide). a well established property of TZDs (34) that is mediated by PPAR␥ (19,35). MCC-555 induced adipocyte differentiation as measured by induction of the adipocyte-specific, PPAR␥-responsive aP2 gene (Fig. 8) as well as by morphologic criteria (data not shown). The maximal induction of aP2 was similar for MCC-555 and BRL49653. However, the EC 50 for MCC-555 was ϳ10-fold greater for than for BRL49653, again consistent with the relative abilities of these compounds to bind and activate PPAR␥ but inconsistent with their abilities to reduce blood glucose levels in vivo.
Effects of MCC-555 on PPAR␥-regulated Gene Expression In Vivo-Although MCC-555 was less potent in adipocyte differentiation in vitro, we considered that its increased in vivo potency might still involve PPAR␥ but relate to a favorable pharmacokinetic profile relative to that of the multiple other TZDs for which in vitro and in vivo effects are better correlated. We therefore examined epididymal adipose tissue of KK-A y mice for the expression of PPAR␥-responsive genes. We chose aP2 as a gene that is up-regulated by PPAR␥ ligands (34,35) and leptin as a gene that is down-regulated by PPAR␥ ligands (36 -39). We also studied the expression of PPAR␥ itself, which is only minimally responsive to TZDs in adipocytes where it is already expressed at high levels (36). Fig. 9 shows that aP2 was induced from its already high levels by treatment with BRL49653 as well as by MCC-555. Interestingly, the effect of BRL49653 was greater after 1 day of treatment, whereas after 4 days aP2 expression was greatest in mice treated with MCC-555. Similar conclusions regarding the in vivo potency of the TZDs pertained to the down-regulation of leptin gene expression, although this was more modest. PPAR␥ expression was not detectably different in adipose from TZD-treated mice in this experiment. These results suggest that MCC-555 is at least as potent as BRL49653 at regulating PPAR␥-responsive genes in vivo.

DISCUSSION
The treatment of non-insulin-dependent diabetes mellitus has recently been improved by the identification of TZDs as novel agents with the unique ability to potentiate insulin action at a level of signaling that is distal to the insulin receptor. The discovery that TZDs are nanomolar PPAR␥ ligands and that their abilities to activate PPAR␥ correlate with their enhancement of insulin action suggests that these two actions of TZDs are related. Our results are very important in this context. There are a number of current theories of TZD action. PPAR␥ is mainly expressed in adipose tissue, and thus one possibility is that PPAR␥ activation stimulates insulin-mediated glucose uptake to such a large extent in adipose tissue that blood sugar is lowered despite the fact that muscle is ordinarily the main site of insulin-dependent glucose disposal in mammals. Another possibility is that TZD activation of PPAR␥ induces adipocytes to send an insulin-enhancing signal to muscle; possibilities include reduced levels of adipocyte-derived tumor necrosis factor ␣ (40) or free fatty acids (41). It is also possible that the very low amounts of PPAR␥ in muscle are sufficient for TZDs to directly affect insulin action in that tissue (42), or that TZDs have other direct effects on muscle besides activation of PPAR␥. FIG. 8. MCC-555 induces adipogenesis but with lower potency than BRL49653. 3T3-L1 cells were exposed to different concentrations of TZDs as indicated. Northern analysis of 10 g of total RNA probed with aP2 cDNA is shown. Equal RNA loading was confirmed by rRNA staining (not shown).
FIG. 9. Regulation of PPAR-responsive genes from adipose tissue of TZD treated KK-A y mice. Northern analysis of epididymal adipose tissue of KK-A y mice treated with BRL49653 or MCC-555 (30 mg/kg) for either 1 or 4 days as indicated. 3T3-L1 day 0 (D0, preadipocyte) and day 8 (D8, adipocyte) RNA was used as control. 10 g of total RNA was probed with cDNAs for aP2, leptin, and PPAR␥. 28 S rRNA ethidium bromide staining is also shown as evidence of equal loading.
The ability of MCC-555 to function as a potent antidiabetic agent and to possess novel PPAR␥-activating properties has a number of potential implications. Certainly, if PPAR␥ activation is sufficient for the antidiabetic effects of TZDs, the potency of MCC-555 indicates that partial or context-dependent agonism of PPAR␥ is sufficient for these antidiabetic effects. Moreover, the antidiabetic potency of MCC-555 is increased relative to BRL49653, which has higher affinity for PPAR␥. The effects on adipose gene expression in vivo are likely to be related to more favorable pharmacokinetics of MCC-555, leading to stimulation of PPAR␥-responsive genes in adipose tissue. However, MCC-555 is also associated with reduced cardiac and hematopoietic side effects relative to other TZDs (20). Thus, it is possible either that 1) the differential ability of MCC-555 to activate the PPAR␥ LBD depending on the context of the DBD to which it is fused is somehow favorable in terms of antidiabetic effects or 2) a subset of the full range of agonistic effects of other TZDs, such as BRL49653, is somehow detrimental to the therapeutic effects of these full agonists. Of course, it is also possible that MCC-555 has other targets in addition to PPAR␥. There may thus be additional or alternative targets of MCC-555 in fat or muscle that contribute to its unique pharmacological and therapeutic profile in diabetes.
The unique properties of MCC-555 are not completely understood. Unlike the ER partial agonist tamoxifen, the A/B domain of PPAR␥ was not essential for full activation from a PPREcontaining reporter gene. While this work was under review, Blanco et al. (43) reported that p300/CBP-associated factor functions as a nuclear receptor coactivator by binding to the receptor DBD (43). This could explain the ability of MCC-555 to fully activate PPAR␥ transcription from the PPRE, because that process requires the PPAR␥ DBD, which is absent from the Gal4 and tet fusion proteins. By contrast, the ability of MCC-555 to function as a full agonist with full-length PPAR␥ suggests that SRC-1, the recruitment of which by MCC-555 is submaximal, is not rate-limiting in this situation. This is consistent with the recent description of promoter-specific requirements for p300/CBP-associated factor, CBP, SRC-1, and other coactivators (44). We also cannot rule out the possibility of a specific coactivator that plays a critical role in insulin action and is differentially recruited by MCC-555.