N-Acetylfarnesylcysteine Is a Novel Class of Peroxisome Proliferator-activated Receptor γ Ligand with Partial and Full Agonist Activity in Vitro and in Vivo*

The thiazolidedione (TZD) class of drugs is clinically approved for the treatment of type 2 diabetes. The therapeutic actions of TZDs are mediated via activation of peroxisome proliferator-activated receptor γ (PPARγ). Despite their widespread use, concern exists regarding the safety of currently used TZDs. This has prompted the development of selective PPARγ modulators (SPPARMs), compounds that promote glucose homeostasis but with reduced side effects due to partial PPARγ agonism. However, this also results in partial agonism with respect to PPARγ target genes promoting glucose homeostasis. Using a gene expression-based screening approach we identified N-acetylfarnesylcysteine (AFC) as both a full and partial agonist depending on the PPARγ target gene (differential SPPARM). AFC activated PPARγ as effectively as rosiglitazone with regard to Adrp, Angptl4, and AdipoQ, but was a partial agonist of aP2, a PPARγ target gene associated with increased adiposity. Induction of adipogenesis by AFC was also attenuated compared with rosiglitazone. Reporter, ligand binding assays, and dynamic modeling demonstrate that AFC binds and activates PPARγ in a unique manner compared with other PPARγ ligands. Importantly, treatment of mice with AFC improved glucose tolerance similar to rosiglitazone, but AFC did not promote weight gain to the same extent. Finally, AFC had effects on adipose tissue remodeling similar to those of rosiglitazone and had enhanced antiinflammatory effects. In conclusion, we describe a new approach for the identification of differential SPPARMs and have identified AFC as a novel class of PPARγ ligand with both full and partial agonist activity in vitro and in vivo.

More than 20 million people in the United States have type 2 diabetes mellitus (T2DM). 2 T2DM is a chronic metabolic dis-ease that results from a variety of conditions, including obesity, insulin resistance, hyperglycemia, and inflammation (1,2). Several different classes of drugs are clinically approved for the treatment of T2DM, including biguinides, sulfonylurea, and thiazolidinediones (TZDs). The TZD class represents one the most common therapies, which include pioglitazone (Actos) and rosiglitazone (Avandia) (3,4). TZDs exert their effects by activating peroxisome proliferator-activated receptor ␥ (PPAR␥), a member of the nuclear receptor superfamily of ligand-activated transcription factors (5)(6)(7)(8) and a master regulator of adipogenesis (9 -12).
Despite their widespread use, concerns exist regarding the safety and efficacy of PPAR␥ agonists (13,14). The first generation TZD, troglitazone, was voluntarily taken off the market due to rare but severe hepatotoxicity (15). Current TZDs, although maintaining a better toxicity profile than troglitazone, are still associated with increased weight gain/adiposity, anemia, increased rate of bone fracture, and liver injury (4,(15)(16)(17). The deleterious effects of PPAR␥ ligands have prompted the development of selective PPAR␥ modulators (SPPARMs). These are ligands that selectively modulate PPAR␥ activity (18,19). Ideally, SPPARMS promote the glucose-lowering effects of PPAR␥ but lack the undesirable effects such as increased adiposity by acting as partial agonists of PPAR␥. To date, identification of new SPPARMs is based on structural modeling, in vitro binding assays, or luciferase-based reporter assays. Although these screens identify compounds that may bind and/or activate PPAR␥, many of the SPPARMS developed do not show differential specificity. Hence, there is partial induction of most PPAR␥ target genes and not only those associated with deleterious effects. Therefore, a screening approach to identify differential SPPARMs, compounds that promote partial and full activity of PPAR␥ depending on target genes, would be beneficial. We describe here a gene expression-based screening approach to identify SPPARMs based on their ability to regulate PPAR␥ target genes differentially. Using this approach, we have identified a novel SPPARM, N-acetylfarnesylcysteine (AFC), that promotes glucose tolerance, has enhanced antiin-flammatory effects and reduced adipogenic activity compared with classic TZDs.

EXPERIMENTAL PROCEDURES
Cell Culture and Chemicals-3T3-L1 preadipocytes were obtained from ATCC. PPAR␥ knock-out mouse embryonic fibroblasts (MEF PPAR␥KO ) were obtained from Dr. Evan Rosen (Beth Israel Deaconess Medical Center, Boston MA) (11). MEF PPAR␥KO were transformed with retrovirus-expressing empty vector or PPAR␥ and selected using puromycin as described previously (20). Cells were then treated with compounds as indicated, RNA was isolated, and RT-PCR was performed. All chemicals including the LOPAC library were obtained from Sigma or Alexis Biopharmaceuticals. Tnf-␣ was purchased from R&D Systems. Cell culture media and supplements came from Cellgro (Mediatech), and FBS was from HyClone.
Drug Screening-3T3-L1 cells were seeded in a 96-well plate and grown overnight. The following day cells were treated with a 20 M concentration of each compound for 24 h and RNA isolated using 96-well Turbo RNA isolation kit (Qiagen) according to the manufacturer's directions. RT-PCR was performed using SYBR Green reagent (Applied BioSystems).
Adipocyte Differentiation-3T3-L1 cells were subjected to a differentiation protocol as described previously in the absence or presence of AFC or rosiglitazone (20). The level differentiation was by monitoring lipid accumulation as detected by Oil Red staining.
Western Blot Analysis-Western blotting for phosphorylated and nonphosphorylated PPAR␥ was performed as described previously by Hauser et al. (21). In brief, protein lysates from differentiated 3T3-L1 cells (described above) were separated on a 10% polyacrylamide gel containing 8 M urea and transferred to nitrocellulose. Immunoblotting for PPAR␥ was performed using a rabbit polyclonal antibody (H100; Santa Cruz Biotechnology). Phosphorylation of the CDK5 site on PPAR␥ at serine 273 was performed as previously described (22). Briefly, 3T3-L1 fibroblasts expressing PPAR␥ were pretreated with rosiglitazone, MRL24, or AFC with the indicated doses and then incubated with TNF-␣. Cells were homogenized in lysis buffer, and phosphorylation of PPAR␥ at Ser-273 was analyzed with phospho-specific antibody against PPAR␥ Ser-273. Total PPAR␥ was analyzed with anti-PPAR␥ antibody (Santa Cruz Biotechnology). Western blotting for serine 112 (or 82 of PPAR␥ 1 )phosphorylated and nonphosphorylated PPAR␥ was performed as described previously by Hauser et al. (21). Immunoblotting for PPAR␥ was performed using a rabbit polyclonal antibody (H100).
Ligand Binding Assay-Histidine-tagged PPAR␥-LBD fusion construct (a gift from Dr. John Schwabe, Medical Research Council, UK) was expressed in E. coli as described previously (24). In brief, protein expression was initiated by the addition of isopropyl 1-thio-␤-D-galactopyranoside and grown for an additional 3 h. Cells were harvested, and pellets were isolated and lysed. The lysate was sonicated, and supernatant was added to nickel-nitrilotriacetic acid slurry (Qiagen) for purification according the manufacturer's directions. The presence of Histagged PPAR␥-LBD protein was confirmed by SDS-PAGE. For the ligand binding assay, 100 g of protein was incubated with 10 nM [ 3 H]rosiglitazone (PerkinElmer Life Sciences) in the presence of increasing concentrations of unlabeled AFC for 3 h at 4°C in ligand binding buffer (23,25). After incubation the beads were washed to remove the unbound ligand and binding quantitated using liquid scintillation counting. We also performed a ligand binding assay using labeled AFC. Unlabeled rosiglitazone was used at various concentrations in the presence of 35 S-labeled AFC (Signum Biosciences). The binding isotherms were analyzed for both ligands, and K d values were determined using the one-site binding competition model as implemented in the SigmaPlot (Systat Software, San Jose, CA).
Animal Experiments-All animal work was conducted in accordance with the Animal Care Facility at the University of Maryland, Baltimore. Male C57BL/6J (4 weeks old) mice were obtained from Jackson Laboratory and fed a high fat diet (60 kcal% fat, D12492, Research diet) to develop diet-induced obesity (DIO) and hyperglycemia. The blood glucose levels were checked before the start of drug treatment. Animals were administered dimethyl sulfoxide, rosiglitazone, and AFC by intraperitoneal injection daily. After 2 weeks, fasting glucose and glucose tolerance tests were performed. In brief, mice were fasted overnight and tail vein blood used to measure blood glu-cose levels. For the glucose tolerance test, blood glucose was measured as indicated after intraperitoneal glucose administration (1 g/kg body weight). White adipose tissue was harvested at the terminal point of the experiment for histology or frozen in liquid nitrogen for further analysis. For histological analysis paraformaldehyde-fixed paraffin-embedded sections of white adipose tissue were stained with hematoxylin & eosin to observe changes in adipose tissue morphology. Counting of adipocytes was performed blinded by a pathologist (Dr. Twaddel).

RESULTS
AFC Regulates PPAR␥ Activity-We used a small library of pharmacologically active compounds (LOPAC, Sigma) to screen for molecules with differential PPAR␥ activating abilities. The advantage of this library is that it contains characterized pharmacologically active compounds for which structure and mechanisms have been annotated (see Sigma-Aldrich Web site). Additional compounds that alter signal transduction pathways known to regulate PPAR␥ were also included (27)(28)(29). We performed real-time PCR analysis of well characterized PPAR␥ target genes Angptl4, Adrp, and aP2 as readouts of PPAR␥ activation (30 -33). Although the specific targets associated with effects of PPAR␥ are not fully understood, Angptl4 and Adrp are associated with antidiabetic effects of PPAR␥ (32,34). In contrast, the fatty acid-binding protein aP2 plays an important role in adiposity associated with PPAR␥ activation (31). We compared changes in gene expression to rosiglitazone with a focus on similar induction of Angptl4 and Adrp and partial induction of aP2. A number of compounds induced Angptl4 and Adrp but not aP2. Although aP2 was induced by cAMP and glucocorticoid-related compounds in the library (which are known to promote adipogenesis) (35), they did not induce Ang-ptl4 or Adrp as effectively as rosiglitazone (data not shown). However, one compound, AFC, appeared to have differential selectivity toward activation of PPAR␥ (Fig. 1A). Angptl4 and Adrp were induced to a similar level by rosiglitazone and AFC (supplemental Fig. S1, A and B). In contrast, aP2 was induced more robustly by rosiglitazone than AFC (supplemental Fig.  S1C).
Next, we confirmed the results of the screen and performed a dose response for AFC and rosiglitazone to evaluate the apparent selective partial agonism by AFC. There was a dose-dependent increase in PPAR␥ target genes Angplt4 and Adrp by both AFC and rosiglitazone (Fig. 1, B and C). In contrast, at all the doses that had a similar effect on Adrp and Angplt4, rosiglitazone was much more effective at inducing aP2 than AFC (Fig.  1D). For example, at the highest dose of each AFC and rosiglitazone (150 M and 250 nM, respectively), Angptl4 and Adrp were induced similarly. However, rosiglitazone treatment led to a 7-fold greater induction of aP2 than AFC, demonstrating a partial agonism of AFC to activate PPAR␥ with respect to aP2. Recent studies have also shown that the insulin-sensitizing effects of PPAR␥ are also mediated in part by induction of the adipokine adiponectin (AdipoQ) (30). AdipoQ was induced to a similar level by rosiglitazone and AFC (supplemental Fig. S1D). This further supports that AFC is a selective partial ligand for aP2 but is a full agonist on other PPAR␥ target genes.
The ability of AFC to induce several PPAR␥ target genes suggests that AFC functions by activating PPAR␥. To confirm this, we treated PPAR␥ knock-out MEFs (MEF PPAR␥KO ) and MEF PPAR␥KO ectopically expressing PPAR␥ (MEF PPAR␥KO -PPAR␥) with AFC. AFC treatment did not induce Adrp or Adi-poQ expression in MEF PPAR␥KO (Fig. 2, A and B). In the absence of PPAR␥, Angptl4 and aP2 were not detected irrespective of treatment (Fig. 2, C and D). However, in the MEF PPAR␥KO -PPAR␥ cells, AFC treatment led to an induction of all four genes (Fig. 2, A-D), demonstrating that AFC is inducing these genes in a PPAR␥-dependent mechanism.
Next, we subjected 3T3-L1 preadipocytes to a differentiation protocol using a hormonal mixture in the presence and absence of rosiglitazone or AFC. Compared with rosiglitazone, AFC was less effective at promoting adipogenesis (Fig. 2E). We then examined the expression of PPAR␥ target genes from the differentiated adipocytes. Although Adrp and Angptl4 were equally induced by rosiglitazone and AFC (supplemental Fig.  S2, A and B); unlike rosiglitazone, AFC did not induce aP2 expression (supplemental Fig. S2C). This further demonstrates that AFC differentially activates PPAR␥.
AFC Is a PPAR␥ Ligand-AFC has a structure similar to several lipid-derived molecules that are known to be PPAR␥ ligands (36,37). The ability of AFC to interact with and activate PPAR␥ as a ligand was tested using a chimeric reporter plasmid containing a Gal4-DNA binding domain fused to the LBD of PPAR␥. We also used a PPAR␥-LBD with a glutamine to proline mutation at amino acid 286 as a negative control. This mutation renders PPAR␥ unable to bind ligands and activate gene expression (23). AFC induced the activity of the Gal4-PPAR␥ fusion construct ϳ 3-fold (Fig. 3A). In contrast, neither rosiglitazone nor AFC was able to activate the Gal4-PPAR␥-LBD QP mutant construct ( Fig. 3A and supplemental Fig. S3A), demonstrating a direct effect of AFC on the LBD of PPAR␥. Although AFC activated the Gal4-PPAR␥-LBD less than rosiglitazone (supplemental Fig. S3A), these data strongly suggest that AFC is interacting with the LBD of PPAR␥.
Next, we performed ligand binding assays to confirm that AFC interacts directly with PPAR␥. As shown in Fig. 3B, AFC was able to displace rosiglitazone from the LBD of PPAR␥ with a K d of 11 Ϯ 3 M. We also determined whether unlabeled rosiglitazone could displace radiolabeled 35 S-labeled AFC from PPAR␥ (supplemental Fig. S3B). Increasing concentrations of rosiglitazone effectively displaced AFC from PPAR␥ with a K d for rosiglitazone of 20 nM. This is in line with the reported K d for rosiglitazone (23,38,39). Next, we performed molecular dynamics simulation of AFC bound to PPAR␥. The modeling predicts that the carboxylate moiety of AFC molecule makes hydrogen bonds to Tyr-473 and His-449. These two residues interact similarly with a majority of the known PPAR␥ ligands. In addition, the model predicts that the acetyl group of AFC forms a novel interaction with the backbone nitrogen of His-466 (Fig. 3C). This is a unique interaction that has not been observed for other PPAR␥ ligands.
AFC is most well known for its ability to inhibit Ras/Rho-GTPase signaling by competing for farnesyl binding domains and inhibiting carboxymethylation of GTPases because it mimics the C-terminal farnesylcysteine group of Ras-GTPase (40,41). Because GTPase signaling inhibits PPAR␥ activity we wanted to determine whether AFC was activating PPAR␥ by virtue of its ability to inhibit GTPase signaling. We treated 3T3-L1 cells with different compounds known to inhibit GTPase processing. Farnesyl transferase and geranylgeranyl transferase inhibitors work upstream of isoprenylcysteine carboxyl methyltransferase by inhibiting the farnesylation or geranylgeranylation of C-terminal CAAX-containing proteins (41). Neither farnesyl transferase inhibitors nor geranylgeranyl transferase inhibitors induced Angptl4 or aP2 in the 3T3-L1 cells (supplemental Fig. S4A). We also used S-adenosylhomocysteine, an inhibitor of isoprenylcysteine carboxyl methyltransferase that is structurally distinct from AFC. We did not observe a change in the expression of PPAR␥ target genes (supplemental Fig. S4A). Additionally, two compounds in the LOPAC library mevastatin and pravastatin, known to inhibit isoprenylation of GTPases, did not alter PPAR␥ target gene expression (data not shown). Therefore, it does not appear that inhibition of GTPase processing by AFC is mediating its effects on PPAR␥.
We then wanted to determine whether AFC alters other signal transduction pathways regulating PPAR␥ function. Phosphorylation of PPAR␥ in its N terminus by several MAPKs reduces the activity of PPAR␥ (21,42,43). Previous studies show that AFC decreases MAPK signaling (40,41). Therefore, we wanted to determine whether the effect of AFC on PPAR␥ was a result of decreasing the phosphorylation of PPAR␥. Fully differentiated 3T3-L1 cells were treated with rosiglitazone, or AFC and the phosphorylation status of PPAR␥ was determined as described under "Experimental Procedures." Phosphorylation of PPAR␥ in differentiation adipocytes was not altered following treatment with AFC (supplemental Fig. S4B, upper  band). Therefore, it is unlikely that AFC is increasing PPAR␥ activity by reducing the phosphorylation of PPAR␥. Interestingly, AFC decreased nonphosphorylated PPAR␥, in a fashion similar to rosiglitazone. The reduction in unphosphorylated PPAR␥ protein by PPAR␥ ligands is a general property of PPAR␥ ligands (21). This further supports our findings showing that AFC is a PPAR␥ ligand.
Recently phosphorylation of Ser-273 of PPAR␥ by CDK5 was shown to alter the ability of PPAR␥ to activate a subset of PPAR␥ target genes (22). Treatment of 3T3-L1 cells with rosiglitazone or MRL24 reduced the TNF-␣-induced phosphorylation of Ser-273 of PPAR␥ as reported previously (22). In contrast, AFC did not alter the phosphorylation of Ser-273 (supplemental Fig. S4C). Therefore, the ability of AFC to regulate PPAR␥ differentially does not appear to be mediated via altering phosphorylation of Ser-273 of PPAR␥.
AFC Improves Glucose Homeostasis in Diabetic Mice-Next, we wanted to determine whether AFC could restore glucose homeostasis in a rodent model of diabetes. Experiments were also performed with rosiglitazone for comparison. After 20 weeks on a high fat diet, fasting glucose levels of the mice were increased, indicating the presence of T2DM (Ͼ140 mg/dl). Mice were then treated with AFC (7.5 mg/kg per day) or rosiglitazone (10 mg/kg per day) for 4 weeks at which point fasting glucose in control mice was still Ͼ140 mg/dl. However, fasting glucose was reduced significantly in the rosiglitazone and AFCtreated mice (Fig. 4A). Mice were then subjected to a glucose tolerance test. Rosiglitazone-and AFC-treated mice had similar efficacy in their ability to reduce blood glucose compared with control mice (Fig. 4B). Body weight of control and rosiglita-zone-treated mice increased ϳ12 and 14%, respectively (Table 1). In contrast, the weight of AFC-treated mice only increased 5.8%. The weight of AFC-treated mice was not significantly different from control mice. However, AFC-treated mice gained significantly less weight than the rosiglitazone-treated mice (p Ͻ 0.05). The reduced weight gain/resistance to weight gain did not appear due to food intake, which appeared similar (data not shown). There was a trend toward a decrease in total epididymal white adipose tissue (WAT) weight in the AFC-treated mice compared with control or rosiglitazone-treated mice, but this was not statistically significant (Table 1). We also examined the expression of liver enzymes, aspartate aminotransferase and alanine transaminase, to determine whether the decreased weight gain of AFCtreated mice was due to toxicity. Expression of these markers was unaltered in livers from AFC-treated mice compared with mice in the control group (data not shown). These studies demonstrate that AFC is as effective as rosiglitazone at reducing plasma glucose levels without the weight gain typically associated with TZD treatment.
Histological analysis of epididymal WAT of rosiglitazoneand AFC-treated mice showed that adipocytes of treated mice were smaller than the adipocytes of the control-treated mice (Fig. 4C). This is noteworthy because previous studies have shown that smaller adipocytes are associated with glucose tolerance (44). In addition, a decrease in adipocyte size is reported to be in part responsible for the effects of PPAR␥ ligands (10,44,45). We also examined the effect of AFC on the expression of PPAR␥ target genes in WAT. Similar to what we observed in vitro, treatment of mice with AFC led to an increase in Adrp (Fig. 4D) and AdipoQ (Fig. 4E) to an extent similar to that observed with rosiglitazone. In contrast, the induction of aP2 by rosiglitazone was greater than AFC (Fig. 4F). Therefore AFC promotes glucose homeostasis without the adipogenic effects associated with classic TZDs.
AFC Reduces Inflammation in WAT-Numerous studies have shown that induction of macrophage-derived inflammatory cytokine such as Tnf-␣, Il6, and serum amyloid proteins 2 and 3 (Saa2 and Saa3) are key mediators of insulin resistance (46 -49). Reducing the production of proinflammatory cytokines is one of the mechanisms mediating the antidiabetic effects of PPAR␥. In addition, AFC has been shown to inhibit inflammation (50). However, examination of WAT from AFCtreated mice showed increased infiltration of macrophages. This would imply that AFC is promoting inflammation in the WAT, despite improved glucose homeostasis. Therefore we investigated whether AFC could alter cytokine production associated with diabetes. 3T3-L1 preadipocytes were treated with Tnf-␣ in the presence or absence of AFC or rosiglitazone, and the gene expression of proinflammatory cytokines was measured. Saa2 expression was reduced by both rosiglitazone and AFC (Fig. 5A). Interestingly, whereas AFC reduced Saa3 mRNA levels ϳ50%, rosiglitazone did not alter its expression (Fig. 5B). In addition, AFC reduced Il6 expression almost 80%, whereas rosiglitazone reduced it by ϳ20% (Fig. 5C). Next, we wanted to confirm the antiinflammatory effects of AFC in vivo. The expression of Il6, Tnf-␣, Saa2, and Saa3 was reduced in the WAT of AFC-treated mice (Fig. 5, D-G). Surprisingly, we did not observe a significant effect of rosiglitazone on cytokine gene expression in WAT (supplemental Fig. S5). These data demonstrate that AFC represses proinflammatory cytokines associated with insulin resistance to a greater degree than rosiglitazone despite the increase in macrophage infiltration.
The reduction in inflammatory cytokines in WAT from AFC-treated mice, yet increased macrophage infiltration appears paradoxical. Adipose tissue macrophages are typically   (52). Insulin resistance is associated with an increase in M1/M2 macrophages, whereas insulin sensitivity is associated with an increase in M2, which is thought to promote tissue repair and glucose homeostasis. This is supported by data showing that recruitment of M2 macrophages to fat is reduced in diabetic mice (53). Furthermore, loss of PPAR␥ results in reduced M2 macrophage activation (54). Therefore, we investigated the effect of AFC on macrophage activation. AFC induced the mRNA expression of arginase I (ArgI) and Clec7a, two markers of M2 macrophage activation (Fig. 5, H and I) (54 -57). In contrast, AFC decreased the expression of Itgax (which encodes CD11c), a marker of M1 macrophages that is elevated in the WAT of diabetic mice and humans (Fig. 5J) (56,58,59). These data suggest that the antiinflammatory effects of AFC may be partially mediated by alternative macrophage activation.

DISCUSSION
TZDs are a class of PPAR␥ agonists used in the treatment of T2DM. However, PPAR␥ agonists are associated with numerous side effects, which have prompted the development of SPPARMs. Most of these molecules do not show differential activation of PPAR␥ target genes associated with beneficial versus deleterious effects. Therefore, to identify differential SPPARMs, we developed a gene expression-based approach based on a method described previously (60) This approach enables the identification of both full and partial agonist effects on PPAR␥ depending on the genes of interest. Using this approach we have identified AFC as a novel class of PPAR␥ ligand.
AFC was initially described as a synthetic compound that inhibits GTPase processing by blocking farnesyl binding to target proteins and/or carboxymethylation of Ras/Rho-GTPases. However, the ability of AFC to activate PPAR␥ is independent of its ability to inhibit carboxymethylation. A unique aspect of AFC function is its specificity with regard to different PPAR␥ target genes. AFC acts as both a partial agonist for certain PPAR␥ target genes and a full agonist for others, both in vitro and in vivo. Although the target genes mediating both antidiabetic and deleterious effects of PPAR␥ ligands are not well understood, AFC induced PPAR␥ target genes associated with insulin sensitivity such as Angptl4, Adrp and AdipoQ to a similar extent as rosiglitazone. In contrast, induction of aP2, a PPAR␥ target gene promoting adiposity, was attenuated compared with rosiglitazone. Although we do not know the mechanisms of differential specificity by AFC, other SPPARMs have been shown to alter the interaction of PPAR␥ with transcriptional coactivators. Therefore, it is likely that AFC also alters the ability of PPAR␥ to interact with its coactivators and hence promote differential recruitment of PPAR␥ to the promoters of its target genes. The unique manner in which AFC interacts with PPAR␥ supports the idea of novel changes in structure that may alter PPAR␥ function. Subsequent studies should elucidate these mechanisms and the manner in which PPAR␥ target genes are differentially induced by AFC.
Molecular modeling suggests one potential mechanism of how PPAR␥ is activated differentially by AFC. AFC is bound by PPAR␥ in a unique manner compared with TZDs and other SPPARMs. The model predicts that similar to classic TZDs, AFC makes crucial contacts with His-449 in helix 11 (H11) and Tyr473 in helix 12 (H12). The ability of AFC to interact with Tyr-473 represents a novel feature for a partial PPAR␥ agonist, which usually does not make this interaction (61)(62)(63)(64)(65). Moreover, unlike other full agonists and partial agonists described to date, AFC interacts with His-466 of PPAR␥. This would result in a substantial shift in the orientation of H12, which is known to be a key element in PPAR␥ activation. The interaction of AFC with Tyr-473 and His-466 likely determines its specificity for full and partial agonism toward different PPAR␥ target genes. These are dynamic modeling studies and not actually AFC/PPAR␥ crystal structures. Therefore, current crystal structure studies are seeking to validate these results, and subsequent studies will determine the role of these residues in mediating the differential effects of AFC on PPAR␥ activation.
Treatment of mice with AFC was as effective at lowering fasting glucose levels in diabetic mice as rosiglitazone. In addition, rosiglitazone and AFC treatment improved glucose tolerance to a similar degree. However, treatment of mice with AFC was not associated with one of the known side effects of TZDs, increased weight gain. Therefore, although AFC lowers glucose levels to a similar degree as rosiglitazone, mice treated with AFC either did not gain as much weight as rosiglitazone-treated mice or were resistant to weight gain from a high fat diet.
Proinflammatory cytokines Saa2, Saa3, Tnf-␣, and Il6 play an important role in insulin resistance (46 -49). PPAR␥ represses the expression of a number of these proinflammatory cytokines (54, 66 -68). Although rosiglitazone reduced cytokine mRNA expression in vitro (albeit less effectively than AFC), it remained unchanged in vivo, despite improved glucose tolerance. Although this was somewhat surprising, conflicting results in mRNA expression for inflammatory genes after TZD treatment have been reported previously. TZDs reduce the levels of IL6, TNF-␣ and SAA in adipose tissue from humans (66,69,70). However, other studies, show that in diabetic mice there was no change in inflammatory gene expression from the WAT of mice (71,72). In addition, it was interesting to note that there appeared to be greater macrophage infiltration into WAT of AFC-treated mice. This prompted us to investigate whether these were type I or type II macrophages. Indeed, we observe a shift from M1 to M2 macrophages in WAT from AFC-treated mice. Although PPAR␥ is known to be required for M2 macrophage activation, the TZD class of ligands does not appear to increase M2 macrophages considerably. This suggests that some of the antiinflammatory effects of AFC compared with rosiglitazone may be due to differences in macrophage activation. It is tempting to speculate that ligand activation of PPAR␥ by AFC promotes M2 macrophages to a greater extent than rosiglitazone, although it is unclear at this time whether it is PPAR␥-dependent. Regardless, this suggests that AFC is either more effective at promoting the antiinflammatory effects of PPAR␥ or that it is functioning in a PPAR␥-independent manner.
Several cell-based and in vitro molecular approaches have been used in the past to identify SPPARMs (26,63,(73)(74)(75). However, these approaches only identify general PPAR␥ ligands and do not take in account differential activity of endogenous PPAR␥ target genes. Our differential gene expressionbased approach enables the identification of activators of PPAR␥ with both full and partial agonist effects on PPAR␥ depending on the genes of interest. Indeed, as additional targets are identified, it will help refine and develop this methodology. This approach can also be applied to other systems where identification of molecules or drugs with differential effects is being investigated. A potential drawback of this approach, as with any cell-based technique, is that it identifies compounds that alter PPAR␥ activity through direct and indirect mechanisms. Hence, this approach does not discover compounds that are only PPAR␥ ligands. We actually view this as an advantage because it identifies not only ligands, but also other molecules that can alter PPAR␥ activity independent of PPAR␥ binding. Indeed, a report several years ago identified harmine as an inducible adipose-selective regulator of PPAR␥ expression that increases PPAR␥ activity independent of ligand binding (26). In addition, recent work from Spiegelman and co-workers highlights the effect of ligands on differential induction of PPAR␥ gene expression distinct from ligand binding (22).
The approach described here to identify SPPARMs is rapid and cost-effective, making it amenable to screening large drug libraries. Although the actual PPAR␥ target genes and mechanisms responsible for some of the side effects of TZDs are not well understood, these studies represent advancement toward identifying SPPARMs with differential activity and mechanisms mediating the effects of PPAR␥ ligands. In conclusion, AFC represents a novel class of PPAR␥ ligand with both full and partial agonist activity. Importantly, AFC promotes glucose homeostasis without the adipogenic action of class TZDs and hence supports further studies into the role of AFC and its derivates as antidiabetic therapies.