S 26948: a New Specific Peroxisome Proliferator–Activated Receptor γ Modulator With Potent Antidiabetes and Antiatherogenic Effects

The compound S 26948, Servier, a racemate of dimethyl-2-{4-[2-(6-benzoyl-2-oxo-1,3-benzothiazol-3(2H)yl)ethoxy]benzyl}malonate (37), was synthetized at the Pharmaceutical Chemistry Institute, EA 1043, Faculty of Pharmacy of Lille (Prof. D. Lesieur). The reference compound (rosiglitazone) was obtained from the same source. Both compounds were solved in 10 mmol/l DMSO.

Plasmids pGal4-hPPARγ, pGal4-hPPARα, pGal4-hPPARδ, and pG5-TK-pGL3 were constructed as previously described (38).

Cos-7 cells were seeded in 60-mm dishes at a density of 5.5 × 10⁵ cells/dish in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FCS and incubated at 37°C for 24 h before transfection. Cells were transfected in OptiMEM without FCS for 3 h at 37°C, using polyethylenimine, with reporter and expression plasmids, as stated in the figure legends. The plasmid pBluescript (Stratagene, La Jolla, CA) was used as carrier DNA to set the final amount of DNA to 5.5 μg/dish. The pCMV-β-galactosidase expression plasmid was cotransfected as a control for transfection efficiency. Transfection was stopped by the addition of DMEM supplemented with 10% FCS, and cells were then incubated at 37°C. After 16 h, cells were trypsinized and seeded in 96-well plates at the density of 2 × 10⁴ cells/well and incubated for 6 h in 10% FCS–containing DMEM. Cells were then incubated for 24 h in DMEM containing 0.2% FCS and increasing concentrations of the compound tested or vehicle (DMSO). At the end of the experiment, cells were washed once with ice-cold PBS, and the luciferase and β-galactosidase assays were performed as previously described (39). Half-maximal effective concentration (EC₅₀) was estimated using Prism software (GraphPad). All transfection experiments were performed at least three times.

Binding assays, using a human full-length PPARγ construct expressed in bacteria, were performed in 96-well plates. Binding buffer consisted of 10 mmol/l Tris/HCl, pH 8.2, 50 mmol/l KCl, and 1 mmol/l dithiothreitol. Membrane preparations (5 μg/ml) were incubated for 180 min at 4°C in the presence of [³H]rosiglitazone (BRL49653; Amersham Biosciences) (10 nmol/l) and the tested compounds. Nonspecific binding was defined using an excess of unlabeled rosiglitazone (10 μmol/l). Incubation was terminated by the addition of ice-cold 50 mmol/l Tris/HCl buffer, pH 7.4, followed by rapid filtration under reduced pressure through Whatman GF/C filter plates presoaked with ice-cold buffer, followed by three successive washes with the same buffer. Radioactivity was measured in a Top-count apparatus (Packard).

K_i values were calculated according to the equation K_i = IC₅₀/{1 + ([L]/K_d)}, where IC₅₀ is the concentration of test compound required to inhibit 50% of the specific binding of the radioligand, [L] is the concentration of the radioligand used, and K_d is the dissociation constant for the radioligand at the receptor (40). Each experiment was performed twice, and points were in triplicate.

3T3-F442A preadipocytes were cultured in DMEM supplemented with 10% FCS and 1% antibiotics (streptomycin, penicillin, and amphotericyn). Cells were grown to confluence and treated with the same volume of solvent (10 mmol/l DMSO, control cells), rosiglitazone (0.01–10 μmol/l), or S 26948 (0.01–10 μmol/l) every 2 days. Adipogenesis was determined by Oil red O staining or triglyceride quantification using the Triglycerides Enzymatic PAP150 kit from BioMérieux (Marcy-l’Etoile, France), and the expression of adipocyte-specific mRNA markers was measured. Protein content was determined using the DC Protein Assay system from Bio-Rad Laboratories (Hercules, CA).

The protocol used has been published elsewhere with minor modifications (41,42). In a typical binding reaction, 5–10 pmol of in vitro translated ³⁵S-PPARγ (TnT system; Promega) was incubated with DMSO or ligand (10 μmol/l) in 20 mmol/l Tris-HCl (pH 7.5), 150 mmol/l NaCl, 10% glycerol, and 0.1% Triton X-100. Both isoforms (PPARγ₂ and -γ₁) were obtained by coupled transcription/translation, since the cDNA coding for the -γ₂ isoform contains all the sequences necessary for -γ₁ translation. The presence of both isoforms is very likely to result from the alternative use of start codon on the transcribed PPARγ₂ cDNA.

After 60 min incubation at 20°C (200 μl final volume), 40 μl of a 50% Sepharose-glutathione GST-coactivator slurry was added to the mix and agitated slowly on a rotating wheel for 90 min at 20°C. Unbound material was removed by three successive washes of the Sepharose beads by 10 vol of 1× PBS–0.1% Triton X100. Resin-bound receptors were then resolved by 10% SDS-PAGE and detected and quantified by autoradiography on a PhosphorImager (Molecular Dynamics). All assays were performed at least in triplicate with distinct coactivator extracts.

The assay was based on interaction mating using the Clontech Matchmaker Two Hybrid System 3. Briefly, full-length PPARγ was expressed as a Gal4-DBD fusion protein (pGBT9 vector; Clontech) and transformed into the AH109a yeast strain; the coactivator or corepressor receptor–interacting domain was inserted into the pGAD24 vector to generate a fusion protein with the Gal4 activation domain (DRIP205: amino acid 459–803, PPARγ coactivator [PGC]1: 190–403, SRC1: 459–888, CBP: 8–93, GRIP1: 548–878, and SMRT: 373-1191). After mating, the diploid cells were subjected to two rounds of replication and selected for interacting clones. A fluorogenic substrate was used to quantify the interaction at various ligand concentrations (10 different concentrations, ranging from 50 nmol/l to 50 μmol/l). Results were expressed in fluorescence units as the difference between “no cofactor” control experiments and the signal obtained in cofactor PPARγ studies. Areas under the curve were calculated and plotted as bar graphs. All experiments were carried out at Phenex, Germany.

The care and use of mice were in accordance with the European Community Council Directive 86/609/EEC and approved by the Comité d’Ethique et d’Expérimentation Animale of the University Paul Sabatier (Toulouse, France) and Institut Pasteur (Lille, France). Obese (ob/ob) male C57BL/6 mice were kept under standard conditions of housing (12-h light/dark cycles), feeding (diet: UAR, Villemoissson sur Orge, France; food and water ad libitum), and environment temperature (21°C). Mice were randomly divided into three groups (vehicle control: DMSO 10%/solutol HS 15 [BASF] 15% and sterile water 75% [vol/vol/vol], 10 mg/kg rosiglitazone, and 30 mg/kg S 26948; n = 5–6 animals per group) and injected intraperitonealy once daily (5 ml/kg). Body weight and food intake were measured daily. After 13 days of treatment, mice were killed by decapitation after CO₂ anesthesia and blood samples collected for immediate assessment of glucose levels. Plasma was frozen for posterior metabolite quantification.

Homozygous human apolipoprotein (apo)E2 knockin mice (E2-KI) mice were used in the atherosclerotic lesion study. These mice express human apoE2 instead of mouse apoE (43). Male mice (n = 9–10 per group) were fed a Western diet containing (wt/wt) 0.2% cholesterol and 21% fat (UAR, Epinay sur Orge, France) supplemented without (control group) or with rosiglitazone (10 mg/kg) or S 26948 (30 mg/kg) for 9 weeks. The mice had ad libitum access to water. Blood was collected after a 6-h fasting period by sinus retroorbital puncture under isoflurane-induced anesthesia for biochemical analysis.

Blood glucose levels were measured using Glucotrend and Accu-Check active systems (Roche, Mannheim, Germany). Plasma nonesterified fatty acid (NEFA) was determined using an acyl-CoA oxidase–based colorimetric kit (NEFA-C; Wako Pure Chemicals, Neuss, Germany). Plasma triglycerides were measured using enzymatic colorimetric methods (Triglycerides Enzymatic PAP150; BioMérieux, Marcy-l’Etoile, France). Serum insulin levels were determined by an immunoassay enzymatic system (Mercodia Ultrasensitive Mouse Insulin ELISA; Mercodia, Uppsala, Sweden). Plasma levels of total cholesterol were measured using commercially available kits (Boehringer). HDL cholesterol concentrations were determined after precipitation of apoB-containing lipoproteins with phosphotungstic acid per milligram (Boehringer). Non-HDL cholesterol values were calculated by substracting HDL cholesterol from total cholesterol.

Lipoprotein cholesterol and triglycerides distribution profiles were obtained by gel filtration chromatography using a superose 6 HR 10/30 column (Amersham Pharmacia Biotech) on pooled plasma samples from each group.

Total RNA from cultured cells or frozen tissues was extracted using the Extract-all solution (Eurobio). Complementary DNA was synthesized with the Super ScriptTM reverse transcriptase and poly dT primers (Invitrogen). The real-time PCR measurement of individual cDNAs was performed using SYBR Green dye to measure duplex DNA formation with the Gene Amp 5700 (Applied Biosystem) and normalized to the amount of cyclophilin RNA. The following primers were used: adiponectin forward TCCTGGAGAGAAGGGAGAGAAAG, reverse CCCTTCAGCTCCTGTCATTCC; aP2 forward GATGCCTTTGTGGGAACCTG, reverse GCCATGCCTGCCACTTTC; CCAAT/enhancer-binding protein α forward GAGCCGAGATAAAGCCAAACA, reverse GCGCAGGCGGTCATTG; Cyclophilin forward GATGAGAACTTCATCCTAAAGCATACA, reverse TCAGTCTTGGCAGTGCAGATAAA; LPL forward GTGGCCGAGAGCGAGAAC, reverse CCACCTCCGTGTAAATCAAGAAG; PEPCK forward GCCACAGCTGCTCGCAGAA, reverse GAAGGGTCGCATGGCAAA; PPARγ₂ forward AGTGTGAATTACAGCAAATCTCTGTTTT, reverse GCACCATGCTCTGGGTCAA; and tumor necrosis factor (TNF)α forward GCAGTCTGTGTCTGCTGGGAT, reverse CGCAGAACGGGATGAAGC.

Adipose tissue specimens were fixed in 95% ethanol overnight for histological analysis and embedded in paraffin. Sections were cut at a thickness of 5 μm and stained with hemalun/eosin. Inguinal white adipose tissue (iWAT) sections were incubated for 1 h at room temperature with 0.5 μg/ml uncoupled protein (UCP)11-A. Second antibody coupled to alkaline phosphatase (1:200) was visualized using BCIP/NBT. Endogenous alkaline phosphatase activity was inhibited by levamisole. Control experiments were performed using purified rabbit IgG. Adipocyte size and distribution were determined by the vertical sections method (44,45).

Palmitate oxidation rates were measured in liver homogenates as previously described (46). Small fragments of liver were homogenized in Set buffer (250 mmol/l sucrose, 2 mmol/l EDTA, and 10 mmol/l Tris, pH 7.4), and 75 μl were incubated in 300 μl of oxidation buffer (27 mmol/l KCl, 10 mmol/l KH₂PO₄, 5 mmol/l MgCl₂, 1 mmol/l EDTA, 25 mmol/l sucrose, 75 mmol/l Tris, 5 mmol/l ATP, 1 mmol/l NAD⁺, 25 μmol/l cytochrome c, 0.1 mmol/l coenzyme A, 0.5 mmol/l l-carnitine, and 0.5 mmol/l l-malate, pH7.4) in glass vials. After a 5-min preincubation at 37°C in a shaking-water bath, the reaction was started by adding 100 μl of 600 μmol/l [1-¹⁴C]palmitic acid (Amersham Pharmacia Biotech, Orsay, France) bound to fatty acid–free BSA in a 5:1 molar ratio. Incubation was carried out for 30 min and stopped by adding 200 μl of 3M perchloric acid. The ¹⁴CO₂ produced was trapped in 300 μl ethanolamine/ethylene glycol (1:2 vol/vol) and measured using a liquid scintillation counter. After 90 min at 4°C, the acid incubation mixture was centrifuged (5 min at 10,000g), and 500 μl of supernatant containing the ¹⁴C-perchloric acid acid-soluble products was assayed for radioactivity by liquid scintillation. Palmitate oxidation rates were calculated from the sum of ¹⁴CO₂ and ¹⁴C-percholic acid-soluble products and expressed in nanomoles of palmitate per minute per gram tissue weight or picomoles of palmitate per minute per milligram of protein.

Mice were killed by cervical dislocation; the heart of each animal was fixed with 4% phosphate-buffered paraformaldehyde (pH 7.0), and serial 10-μm-thick sections were cut between the valves and the aortic arch for quantitative analysis of lipid deposition by Oil red O. Images were captured by use of a JVC 3-charge-coupled device video camera. Sections were analyzed using the computer-assisted Quips Image analysis system (Leica Mikroskopic und System, Wetzlar, Germany).

Data are expressed as means ± SE. Prism Software (GraphPad, San Diego, CA) was used for all statistical analysis. A one-way ANOVA followed by Tukey's multiple comparison test was used to assess the statistical difference between groups. When appropriate, an unpaired t test was performed.

To examine whether S 26948 activated PPARγ, we used a transient transfection reporter assay in COS-7 cells. S 26948 was unable to significantly activate RXRα (Fig. 1A), human PPARα (Fig. 1B), and PPARβ/δ (Fig. 1C) at concentrations up to 10 μmol/l. By contrast, it strongly activated human PPARγ in a dose-dependent manner using chimeric protein construct (Fig. 1D) or using full-length construct cotransfected with RXRα (Fig. 1E). The EC₅₀ for S 26948 was not significantly different from rosiglitazone (Fig. 1D and E). Furthermore, in a binding assay using [³H]rosiglitazone and hPPAR-LBD (45), S 26948 bound PPARγ with the same affinity as rosiglitazone (Fig. 1F); K_i values were not significantly different.

3T3-F442A cells were grown to differentiate in the absence or presence of various concentrations of S 26948 or rosiglitazone. As an index of differentiation, the triglyceride content per milligram protein of the cells was assessed (Fig. 2A). As expected, rosiglitazone significantly increased the triglyceride content in a dose-dependent manner after 8 days of treatment compared with control values (P < 0.001). By contrast, S 26948 did not increase the triglyceride content of the cells at any concentration tested. Microscopic observation of the cell culture showed that at the end of the differentiation (day 8), cells treated with 100 nmol/l S 26948 had morphologic features similar to those of control cells. Lipid droplets in S 26948–treated cells were smaller and less abundant than those in rosiglitazone-treated cells and were indeed very similar to control cells (Fig. 2B).

The rate of triglyceride accumulation was the same in control and S 26948–treated cells in a time-course analysis (Fig. 2C). By contrast, cells treated with 100 nmol/l rosiglitazone accumulated significantly more triglycerides than control or S 26948 cells as soon as 4 days after confluence. On the other hand, S 26948 slightly induced mRNA expression of adipocyte marker genes aP2, LPL, PEPCK, and PPARγ2 but to a lesser extent than rosiglitazone (Fig. 2D).

The ligand-binding domain of PPARγ interacts with cofactors such as DRIP205, CBP, PGC-1α, GRIP1, and SRC-1 (47,48). GST pull-down experiments were first carried out to assess the affinity of PPARγ for GST fusion proteins with the indicated cofactors (Fig. 3A and B). Compared with control conditions, both S 26948 and rosiglitazone at 10 μmol/l had no significant effect on the interaction between the PPARγ-binding domain and SRC-1 (Fig. 3A). By contrast, whereas rosiglitazone stimulated the formation of PPARγ/DRIP205 and PPARγ/PGC-1α complexes, S 26948 was unable to elicit the formation of such complexes. However, S 26948 and rosiglitazone recruited GRIP1 with a seemingly equal efficiency (Fig. 3A and B). Similar results were obtained with 1 μmol/l of each ligand (data not shown).

A mating yeast two-hybrid assay was then used to assess the affinity of these coregulators for PPARγ. End point results obtained at the highest concentrations used (50 μmol/l in this assay) fully support our results shown in panel A and B (data not shown). Results were thus expressed as the area under the curve, obtained from mating interaction assays carried out at ligand concentrations ranging from 50 nmol/l to 50 μmol/l (Fig. 3C). There is therefore a reflection of the affinity of PPARγ for the various coregulators used here. These affinity assays revealed again that S 26948 promotes a weaker interaction of PPARγ with DRIP205, PGCα, and SRC1—the latter being undetectable in GST pull-down assays. A similar behavior was detected for CBP, whose interaction with PPARγ was barely above threshold in GST pull-down assays. In this assay, rosiglitazone was also more efficient at displacing SMRT than S 26948. Finally, GRIP1/TIF2 exhibited an equal affinity for rosiglitazone- or S 26948–bound PPARγ. Taken collectively, these two approaches clearly suggested that S 26948 promoted a PPARγ conformation distinct from that elicited by rosiglitazone.

To test the possible antidiabetes effects of S 26948, 8- to 12-week-old male ob/ob mice were treated with 30 mg/kg S 26948, 10 mg/kg rosiglitazone, or vehicle (control mice) by daily intraperitoneal administration for 13 days. These doses were chosen based on their efficiency in reducing blood glucose levels to the same levels in preliminary pharmacokinetic studies (data not shown). As shown in Table 1, treatment with S 26948 resulted in decreased blood glucose levels (52%) and a concomitant strong reduction of plasma insulin levels (95%). This suggests that S 26948 increased insulin sensitivity. Furthermore, S 26948 treatment was capable of normalizing serum triglyceride levels (46% reduction) and lowering serum NEFA levels (55%). All these effects paralleled those of rosiglitazone treatment (Table 1).

One of the undesirable effects of TZD treatment in type 2 diabetic subjects is body-weight gain (31,33,34). To determine the effects of S 26948 treatment, body weight was measured daily and body-weight gain calculated as the difference in body weight between each day and day 0. Figure 4A shows the body-weight gain at the end of the treatment, as well as food intake and efficiency. Whereas rosiglitazone significantly increased body weight compared with weight of controls, S 26948 had an opposite effect. Thus, at the end of the treatment, the body-weight gain was 5.52 ± 0.42 g, 3.63 ± 0.31 g, and 0.87 ± 0.82 g for rosiglitazone-, control, and S 26948–treated mice, respectively. Indeed, no difference in food intake was measured, resulting in a profoundly decreased food efficiency in the S 26948–treated mice compared either with control or rosiglitazone-treated animals.

S 26948 treatment did not increase white adipose tissue (WAT) weight compared either with rosiglitazone or control mice. As expected, rosiglitazone treatment increased epidydimal (15.5 and 21.9% vs. control and S 26948–treated animals, respectively), inguinal (20 and 40.2%, respectively), and total WAT weight (as the sum of iWAT and epididimal WAT weight depots; 17.5 and 36.6%, respectively) (Fig. 4B). Furthermore, S 26948 treatment decreases mRNA expression of the adipogenic transcription factors PPARγ and CCAAT/enhancer-binding protein α and the adipocyte marker gene LPL in iWAT (Fig. 4C). The expression of the adipokines adiponectin and TNFα was also different in iWAT of S 26948–treated mice (Fig. 4C). Thus, whereas rosiglitazone induced increase in both adiponectin and TNFα mRNA expression, S 26948 augmented the expression of adiponectin mRNA but not that of TNFα. Histological examination of the tissue also revealed changes in the morphology of the cells (Fig. 5A). Thus, both rosiglitazone- and S 26948–treated mice presented an increased number of small adipocytes in WAT deposits, most of them being positive for UCP1 (Fig. 5A). Nevertheless, only adipocytes from S 26948–treated animals were smaller when measured as adipocyte area (Fig. 5B) or as adipocyte size distribution (Fig. 5C). Interscapular brown adipose tissue weight was significantly increased in rosiglitazone-treated animals (110% increase vs. control mice and 45% vs. S 26948–treated animals) (Fig. 4B), with bigger cells and increased lipid droplets compared with control or S 26948–treated mice (Fig. 5D).

Knowing the lipid profile, body and adipose tissue weight profiles, and adipocyte characteristics of the S 26948–treated mice, we sought to know whether this compound could exert its effects in other tissue. Skeletal muscle (gastrocnemius muscle) weight was similarly reduced in rosiglitazone- and S 26948–treated mice, but liver weight was specifically decreased in S 26948–treated animals (Fig. 6A) compared with both untreated and rosiglitazone-treated animals. Histological examination of the livers revealed that S 26948 was able to notably reduce lipid droplets in hepatocytes (Fig. 6C) by increasing the lipid oxidation capacity of the tissue (Fig. 6C and D).

To investigate the effects of S 26948 on dyslipidemia and atherosclerosis, homozygous human apoE2 knockin (E2-KI) mice were used. These mice display mixed dyslipidemia and develop atheroslerotic plaques (43). Male E2-KI mice were fed a Western diet and treated with or without rosiglitazone or S 26948 for 9 weeks. Compared with those in control mice, S 26948 reduces plasma total cholesterol (P < 0.001) and triglyceride (P < 0.05) concentrations with a decrease of non-HDL cholesterol (P < 0.001) levels (Table 2). Rosiglitazone had a tendency to reduce (NS) total cholesterol and non-HDL cholesterol (NS) levels (Table 2). Triglyceride levels did not change in the rosiglitazone-treated mice. HDL cholesterol levels were similar between the groups. Analysis of cholesterol and triglyceride distribution profiles confirmed the lipid results: S 26948 reduced cholesterol and triglyceride contents in the LDLs (including VLDL, intermediate-density lipoprotein [IDL], and LDL) (Fig. 6A and B). Rosiglitazone only slightly reduced cholesterol in the IDL+LDL fractions (Fig. 7A and B).

The effects on atherosclerosis were assessed in these mice at the end of treatment by measuring Oil red O–stained surfaces at the aortic sinus. S 26948 significantly reduced atherosclerotic lesion surfaces by 46% compared with those in control mice (0.033 ± 0.014 vs. 0.061 ± 0.015 mm², P < 0.01) (Fig. 8A and B). In contrast, rosiglitazone had no effect on atherosclerotic lesion size (0.062 ± 0.022 mm²) (Fig. 8A and B).

The results of the current study establish that S 26948, a non-TZD compound, is a new SPPARM that, by contrast to most of the molecules developed so far, does not promote adipogenesis in vitro and in vivo. Furthermore, S 26948 has in vivo antidiabetes and antiatherogenic activity, improves insulin action, and corrects dyslipidemia in rodent models.

Reporter gene assays established that S 26948 is a full PPARγ-specific agonist, at least as potent as rosiglitazone, having similar EC₅₀ at the nanomole per liter order (Fig. 1D) and an affinity of S 26948 for its receptor as high as that of rosiglitazone (Fig. 1F). However, the finding that it has no effect on adipogenesis—in addition to the significant effect in reducing atherosclerosis—showed functional differences with rosiglitazone and other PPARγ agonists (21–27). This establishes S 26948 as a new specific PPARγ modulator that is able to drive normal adipocyte differentiation—in contrast to rosiglitazone, which exacerbates this process. Thus, 3T3-F442A cells differentiated with rosiglitazone largely increased triglyceride content and PPARγ adipocyte target gene expression such as LPL and aP2, whereas S 26948 did not influence these parameters (Fig. 2) in sharp contrast with other PPARγ ligands (22,23).

The mechanisms by which nuclear receptors regulate transcription of target genes are not fully understood. However, evidence indicates that ligand binding results in a conformational change that allows the dissociation of corepressors (such as NcoR) and SMRT, as well as the association of coactivators (such as SRC1, DRIP205, CBP/p300, or PGC1α, among others) (48). Interestingly, it is clear from our data that rosiglitazone and S 26948 did not induce a similar pattern of cofactor recruitment by PPARγ (Fig. 2). Whereas rosiglitazone induced the recruitment of DRIP205 and PGC1α, S 26948 was unable to trigger these interactions, providing evidence of structural differences between rosiglitazone and S 26948 liganded PPARγ, which may explain in part the different biological properties of these drugs in vitro and in vivo (see below). Since DRIP205 (TRAP220) has been described as a coactivator of PPARγ required for adipogenesis (49), the decreased adipogenic activity of S 26948 might be explained in part by the lower efficiency of S 26948 than rosiglitazone to recruit this coactivator.

Recently it has been proposed that moderate activation of PPARγ would improve metabolic features without increasing adipocyte differentiation (50). Moderated activation of PPARγ could be achieved by two mechanisms: 1) by partial agonism or antagonism of the receptor or 2) by differential coactivator recruitment. S 26948 exerts its activity as a full agonist of PPARγ, as demonstrated by the binding experiences, but with a specific coactivator recruitment that decreases its adipogenic capacity compared with rosiglitazone and thus acts as a specific modulator of PPARγ activity.

These in vitro results are consistent with the in vivo data on body weight in obese mice. It is well known that most PPARγ ligands are able to normalize glycemia and insulinemia in spite of increasing body weight (22,24). This has first been observed with TZDs in animal studies (31,32,51) and then confirmed in mildly obese humans treated with such compounds (33,34). In humans, the severity of weight gain is proportional to the level of glycemic control achieved and is thus often related to the dose used. In the present study, we show that rosiglitazone, a full PPARγ agonist, has a net effect on body-weight gain. Indeed, a clear increase in adipose tissue mass was observed upon rosiglitazone treatment, whereas S 26948 had an opposite effect. This is in accordance with the in vitro results.

Compared with rosiglitazone, S 26948 was as potent to lower glucose, NEFA, triglyceride, and insulin levels in ob/ob mice—a model of type 2 diabetes (Table 1). The exact mechanism by which PPARγ agonists exert their insulin-sensitizing effect is still not well understood. This could be related to the ability to decrease serum free fatty acids (FFAs), since increased FFA levels have been shown to induce insulin resistance (52). Lowered FFA level, observed on TZD treatment, has been proposed to be the consequence (in adipose tissue) of both an increased storage and/or a decreased release. These explanations might hold true for rosiglitazone but cannot occur with regard to S 26948, which did not promote body-weight gain. An alternative explanation for the beneficial effect of S 26948 on glucose homeostasis might be its distinctive regulatory properties in other organ(s), such as liver. In fact, both molecules increased lipid oxidation in liver, but S 26948 appears to have a greater effect than rosiglitazone (higher palmitic oxidation and lower lipid infiltration) (Fig. 6). This amelioration on hepatic lipid catabolism may explain part of the glucose homeostasis improvement and lipid-lowering effect of S 26948. In addition to this increment of hepatic lipid oxidation, also noted is the enhanced expression of UCP1 in adipose tissue of treated animals (Fig. 5A), which may also be involved in increasing energy expenditure in this tissue, although no obvious differences could be observed between rosiglitazone and S 26948. Selective biological effect(s) of S 26948, when compared with other PPARγ agonists, might stem from its inability to promote the recruitment of two coactivators, DRIP205 and PGC-1, to PPARγ in our conditions. As both have been shown to have a crucial role in adipocyte differentiation, this may provide a molecular explanation for the nonadipogenic properties of S 26948. By extension, the effect of S 26948 on glucose homeostasis may be related to its specific properties on coactivator recruitment, consistent with the concept of SPPARM (12,13). In this regard, the differential regulation of adiponectin and TNFα of both compounds (Fig. 4C) may also be the result of their special coactivator recruitment and indeed contribute to the glucose- and lipid-lowering effects of S 26948. Thus, S 26948 restores the balance between adipose adiponectin and TNFα gene expression in our in vivo model, which is in clear contrast to rosiglitazone.

Different studies have documented that PPAR ligands inhibit the development of atherosclerosis in different mouse models (56). In E2-KI mice, PPARγ agonists like rosiglitazone or pioglitazone do not reduce the development of atherosclerosis (57). In the present study, we confirmed that in this model rosiglitazone has no significant effect on plasma lipoprotein (Table 2 and Fig. 6) or atherosclerosis (Fig. 7). By contrast, S 26948 reduced atherosclerosis. This effect was associated with a decrease of cholesterol and triglyceride content in the LDL particles (VLDL, IDL, and LDL). This suggests that the reduction of atherosclerosis with S 26948 treatment is in part due to a decrease of atherogenic lipoproteins.

In summary, we identified a new compound (S 26948) that exerts antidiabetes effects similar to most of the other PPARγ ligands. However, when compared with these ligands, S 26948 is highly specific and selective for PPARγ and does not promote adipogenesis and body-weight gain. Furthermore, it shows a strong efficiency in correcting dyslipidemia and atherosclerosis. Although the exact mechanism beyond these effects remains to be determined, S 26948 or related compounds might represent a new class of therapeutic molecules for the treatment of type 2 diabetes and atherosclerosis.

We thank B. Spiegelman (Dana-Farber Institute, Boston, MA), B.W. O’Malley (Baylor College of Medecine, Houston, TX), C. Rachez (INSERM U459, Lille, France), and L.P. Freedman (Merck, West Point, PA) for the gift of plasmids.

We fully acknowledge M. Coevoet, J. Vanhoutte, M. Nibbelink, Y. Jeanson, and M. Boisson for excellent technical assistance, as well as the assistance of the staff of the animal quarter of the IFR31, in particular Y. Barreira and J.M. Lerme.