Hypoglycemic Action of Thiazolidinediones/Peroxisome Proliferator–Activated Receptor γ by Inhibition of the c-Jun NH₂-Terminal Kinase Pathway

For adipocyte differentiation, 2 days after confluency, 3T3-L1 preadipocytes were exposed for 48 h to Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FCS, 500 μmol/l isobutylmethylxanthine, 25 μmol/l dexamethasone, and 5 μg/ml insulin followed by a 48-h incubation in DMEM supplemented with 10% FCS and 5 μg/ml insulin. Thereafter, cells were maintained in DMEM supplemented with 10% FCS. Fat accumulation was visualized by staining with Oil red O. Unless indicated, 5 days after differentiation, cells were incubated for 16 h with DMEM supplemented with 0.5% FCS and rosiglitazone (10 μmol/l), troglitazone (10 μmol/l), 15d-PGJ₂ (3 μmol/l), F-l-Leu (100 μmol/l), bisphenol A diglycidyl ether (BADGE) (10 μmol/l), or vehicle and stimulated with TNF-α (10 ng/ml) for 20 min before harvesting. INS-1E cells (28) were treated with rosiglitazone for 24 h or SP600125 (10 μmol/l) for 20 min and stimulated with IL-1β (10 ng/ml). JNK activity and cell viability were determined after 20 min and 24 h, respectively. Viable and apoptotic cells were differentially labeled using the LIFE/DEAD Viability/Cytotoxicity kit (Molecular Probes). A minimum of 1,000 cells per condition was counted for each experiment.

3T3-L1 preadipocytes seeded in 100-mm tissue culture plates were transiently transfected by lipofection (Invitrogen) and using 3 μg each of the plasmids pCEFL-KZ-HA-JNK and pSV-Sport-PPARγ2 or pSV-Sport. After transfection, cells were treated as described above.

NXE cells were transfected with pSUPERretro (Oligoengine) and pSUPERretro-PPARγ-RNAi (5′GAAGATGACAGACCTCAGG-3′) by standard calcium phosphate precipitation method. Two days after transfection, tissue culture medium was filtered, supplemented with 4 μg/ml polybrene, and used to infect 3T3-L1 preadipocytes. Infected cells were selected with puromycin (2 μg/ml).

JNK and hemagglutinin-JNK were immunoprecipitated using the antibodies sc-474 (Santa Cruz Biotechnology) and 12CA5 (BabCO), respectively, and the JNK activity was determined as described previously (29).

IRS-1, JNK, PPARγ, hemagglutinin-tag, and tyrosine phosphorylation were detected using the antibodies sc-559, sc-474, sc-7273 (Santa Cruz Biotechnology), 12CA5 (BabCO), and 4G10 (Upstate Biotechnology), respectively. Immunoblots were performed using the enhanced chemiluminescence detection system.

RT-PCR analysis was performed using 1 μg total RNA extracted from 3T3-L1 cells with Trizol reagent (Life Technologies) and the following primer pairs: 5′-AGAAGTCACACTCTGACAGG-3′/5′-CAATCGGATGGTTCTTCGGA-3′, 5′-ACTGCCTATGAGCACTTCAC-3′/CAATCGGATGGTTCTTCGGA-3′, and 5′-ACCACAGTCCATGCCATCAC-3′/5′TCCACCACCCTGTTGCTGTA-3′ for PPARγ1, PPARγ2, and glyceraldehyde-3-phosphate dehydrogenase, respectively.

All experimental protocols were approved by the Animal Care Research Committee of the University of Barcelona. Male C57BL/6J obese (ob/ob) mice and lean littermates were obtained from Elevage Janvier (Le Genest-Saint-Isle, France). At 14 weeks of age, rosiglitazone (1 mg/kg) or vehicle (PBS) was orally gavaged once a day for 10 consecutive days. Thereafter, mice were anesthetized with isofluorane, and liver, skeletal muscle (hind quarter), and adipose tissue (epididymal fat pads) were excised and frozen in liquid nitrogen.

JNK1-deficient (JNK1^−/−) mice and wild-type littermates were obtained from parental JNK1 heterozygous mice backcrossed for five generations onto the C57BL/6J strain. For each genotype, 4-week-old male mice were divided randomly into three groups. One group was fed standard chow and the other two received a high-fat, high-carbohydrate diet (diets F4031 and F3282; BioServ). At 17 weeks of age, one of the high-fat, high-carbohydrate diet groups from each genotype was subjected to rosiglitazone treatment in the same conditions, as described above, while the rest were treated with vehicle alone. For the glucose tolerance test (GTT), on day 9 of treatment, mice were fasted for 6 h before oral administration of glucose (2 mg/g body wt), and glucose was determined in blood collected from the tail with an automatic glucometer (Elite; Bayer). The trapezoid rule was used to determine the area under curve (AUC) for plasma glucose concentrations from 1 to 60 min. Plasma insulin and triglycerides were determined by ELISA (Mercodia) and enzymatically (Accuntrend GCT meter; Roche), respectively. Tissue samples were obtained on day 17 of treatment, as described above.

Eight-week-old male ob/ob, ob/ob JNK1^−/−, and lean mice were treated with rosiglitazone (3 mg/kg) or vehicle once a day for 4 consecutive days. Thereafter, epididymal fat pads were dissected, minced in Krebs-Ringer solution supplemented with 2 mmol/l sodium pyruvate and 3% BSA, and digested with 1.5 mg/ml collagenase. Adipocytes were filtered, washed three times in the same buffer, and placed in plastic vials in a final volume of 400 μl. In triplicates, cells were treated with vehicle, rosiglitazone, insulin, or rosiglitazone plus insulin for 10 min at 37°C before 2-deoxy-d-[³H]glucose (2-DG) was added at a final concentration of 0.1 mmol/l (0.4 μCi). After 10 min, 100 μl of 100 μmol/l cytochalasin B was added, and adipocytes were separated by centrifugation in microtubes containing phthalic acid dinonyl ester (density 0.98 g/ml). Incorporation of labeled 2-DG was measured by liquid scintillation.

Data were analyzed with an unpaired Student's test. Values are presented as means ± SE; *P < 0.05; **P < 0.01.

To gain insight into the molecular mechanism behind the insulin-sensitizing action of TZDs, we examined the effect of rosiglitazone on TNF-α–induced JNK activity in 3T3-L1 adipocytes. Treatment of cells with rosiglitazone inhibited TNF-α–induced JNK activity to 52 ± 11.8% (Fig. 1A). This action was consistent with the capacity of rosiglitazone to restore the insulin-induced tyrosine phosphorylation of IRS-1 regardless the presence of TNF-α (Fig. 1B).

Because the antidiabetic actions of TZDs are driven, to a great extent, by PPARγ (12), we performed a series of assays to assess whether this nuclear receptor mediates the action of rosiglitazone on the JNK pathway. Dose-response analyses performed in 3T3-L1 adipocytes showed that effective rosiglitazone doses to inhibit JNK activity correlated with those previously reported to bind to and activate PPARγ (30) (Fig. 2A). Next, we tested several PPARγ agonist ligands, including the TZD troglitazone and other chemically distinct compounds, such as 15d-PGJ₂ and F-l-Leu, for their capacity to inhibit TNF-α–induced activation of JNK in 3T3-L1 adipocytes. JNK activity induced by this cytokine was reduced to 74 ± 10, 69 ± 4, and 70 ± 11% by treatment with troglitazone, 15d-PGJ₂, and F-l-Leu, respectively (Fig. 2B). Like rosiglitazone, none of these PPARγ ligands modified basal JNK activity or the overall amount of this kinase, as assessed by immunoblot analysis (Fig. 1A; data not shown). Moreover, the downregulation of JNK activation induced by rosiglitazone in 3T3-L1 adipocytes was efficiently prevented by the PPARγ antagonist ligand BADGE (31) (Fig. 2C). We also analyzed whether the overexpression of PPARγ enhanced JNK inhibition by rosiglitazone in 3T3-L1 preadipocytes, which express only low levels of PPARγ1 (Fig. 3A). Overexpression of PPARγ2 in these cells improved the inhibitory action of the drug on the JNK pathway (JNK activation was decreased 32 ± 9.4 and 63 ± 3.8% in pSV and pSV-PPARγ-transfected cells, respectively) (Fig. 3B). The PPARγ1 expression found in 3T3-L1 preadipocytes may account for the inhibition of JNK activity in pSV-transfected cells. Accordingly, BADGE also blocked the inhibitory action of rosiglitazone on the TNF-α–induced activity of JNK in these cells (Fig. 3C). Finally, PPARγ expression in 3T3-L1 cells was downregulated by RNA interference (RNAi) through retroviral infection with pSUPER-PPARγ RNAi. Effectiveness of PPARγ RNAi was assessed by the decrease in the levels of PPARγ mRNA (Fig. 3D) and protein (Fig. 3E) and by the inhibition of 3T3-L1 cell differentiation to adipocytes as determined by Oil red O staining (Fig. 3F). In contrast to the cells infected with the control vector, cells in which PPARγ was downregulated by RNAi concomitantly failed to inhibit TNF-α–induced activation of JNK in response to rosiglitazone (Fig. 3G). A similar result was obtained in the infected 3T3-L1 preadipocytes (Supplemental Fig. 1, which is detailed in the online appendix [available at http://dx.doi.org/10.2337/db06–1293]), indicating that PPARγ1 also mediates the inhibitory action of rosiglitazone on the JNK pathway in these undifferentiated cells.

In addition to the role of IL-1β in type 1 diabetes, its glucose-induced production in β-cells contributes to β-cell glucotoxicity in the pathogenesis of type 2 diabetes (32). Because JNK plays a role in the death of pancreatic β-cells induced by IL-1β (9), we analyzed the effect of rosiglitazone on IL-1β–induced JNK activity and cell death in the pancreatic β-cell line INS-1E. Consistent with the PPARγ expression in these cells, rosiglitazone efficiently inhibited JNK activity triggered by IL-1β (Fig. 4A). Moreover, this inhibitory action correlated with the capacity of rosiglitazone to prevent IL-1β–induced apoptosis to a similar extent as the JNK inhibitor SP600125 (Fig. 4B). Similar results were obtained in another pancreatic β-cell line, β-TC3 (Supplemental Fig. 2).

To study the action of rosiglitazone on the JNK pathway in vivo, we took advantage of the exacerbated JNK activity found in peripheral tissues in the obese/hyperglycemic ob/ob mouse (6). Remarkably, rosiglitazone treatment significantly inhibited JNK activity in liver and adipose tissues in the ob/ob mice, whereas no significant effect was detected in any of these tissues in the lean littermates (Fig. 5A). In addition, extracts from liver were analyzed for IRS-1 tyrosine phosphorylation and total IRS-1 content. In agreement with a previous report (33), the amount of hepatic IRS-1 was decreased in the ob/ob mice. Although treatment with rosiglitazone did not restore the overall amount of IRS-1 in these animals and thus it remained lower than their lean littermates, it significantly increased IRS-1 tyrosine phosphorylation (Fig. 5B).

To analyze the effects of rosiglitazone on insulin-induced glucose uptake, primary adipocytes were isolated from 8-week-old lean, ob/ob, and ob/ob JNK1^−/− mice. In contrast to lean mice, adipocytes from untreated ob/ob and ob/ob JNK1^−/− mice showed a reduced glucose uptake in response to insulin, which is consistent with their obesity-induced insulin resistance condition, respectively (Fig. 6). In agreement with the capacity of rosiglitazone to improve insulin sensitivity, treatment of the ob/ob mice with this drug clearly enhanced insulin-induced glucose uptake in primary adipocytes, whereas no effect was observed in those from lean mice. Notably, rosiglitazone failed to enhance glucose uptake in response to insulin in adipocytes from ob/ob JNK1^−/− mice (Fig. 6).

Rosiglitazone treatment was also effective in lowering JNK activity in peripheral tissues in high-fat diet–induced obese mice (Fig. 7A). JNK1^−/− mice on a high-fat, high-carbohydrate diet are partially protected from adiposity (Fig. 7; Supplemental Fig. 3A); nonetheless, they still develop insulin resistance, although milder than their wild-type littermates, as shown by GTT (Fig. 7B and C) and plasma insulin levels (Fig. 7D). In contrast to wild-type animals, treatment of JNK1^−/− mice with rosiglitazone neither improved their glucose tolerance (Fig. 7B and C) nor reduced their hyperinsulinemia (Fig. 7D). Regarding lipid metabolism, in response to a high-fat, high-carbohydrate diet, plasma FFAs did not increase in the JNK1^−/− in comparison with the wild-type mice (Supplemental Fig. 3B), whereas plasma triglycerides were found equally elevated in both types of mice. Remarkably, although the hypoglycemic action of rosiglitazone was abolished in the JNK1^−/− mice, another one of its beneficial actions, such as lowering plasma triglyceride level, was observed only in those mice lacking JNK1 (Fig. 7E).

Despite the pharmacological significance of TZDs, little is known about the molecular mechanism(s) behind the insulin-sensitizing action of these antidiabetic agents. These drugs ameliorate insulin resistance by acting on the early steps of the insulin signal transduction cascade, restoring insulin receptor and IRS-1 tyrosine phosphorylation levels in response to insulin in cell culture (13,14) and in vivo (this study; 15,16). This action is consistent with the inhibitory role of TZDs on the JNK pathway observed in our study given that IRS-1 is a JNK target and IRS-1 phosphorylation by JNK leads to an inhibition of insulin signaling (5). In addition, in genetically and diet-induced mouse models for obesity and diabetes, JNK inhibitors function, similarly as TZDs, as insulin-sensitizing agents, decreasing nonfasting blood glucose levels, improving glucose tolerance, decreasing fasting and nonfasting plasma insulin concentrations, and improving insulin tolerance test performance (7,8). Moreover, the failure of rosiglitazone to improve glucose uptake in response to insulin in primary adipocytes from ob/ob JNK1^−/− mice, and GTT and hyperinsulinemia in the high-fat diet–induced obese JNK1^−/− mice strongly supports our hypothesis that inhibition of JNK pathway is a relevant mechanism to accomplish the hypoglycemic action of TZDs. Nonetheless, it should be mentioned that not all of the beneficial actions of rosiglitazone are abolished in the JNK1^−/− mice; an effect of this drug on the high-fat diet–induced increase in plasma triglycerides is only observed in the JNK1-deficient mice, suggesting an inhibitory role of JNK on PPARγ activity (see below). In this regard, it has been suggested that TZDs regulate triglyceride partitioning among various tissues independently of peripheral insulin resistance (34).

In 3T3-L1 cells, we have shown that TZDs block JNK activation by exogenously added TNF-α, indicating that the action of these drugs is exerted directly on the pathway and is not caused by a decreased production of an extracellular signaling molecule. Nonetheless, TZDs inhibit TNF-α expression in the adipose tissue of rodents, probably because of their inhibitory effects on activator protein-1 and nuclear factor-κB activities (35,36). Therefore, it cannot be excluded that the inhibition of the JNK activity by TZDs in vivo is a result of two processes, a direct targeting on and a decreased production of an extracellular activator(s) of the JNK pathway. These two processes may be interdependent because, as demonstrated for other nuclear receptors, inhibition of the JNK pathway downregulates AP-1 activity (37,38), a transcriptional regulator involved in the expression of pro-inflammatory cytokine genes. In fact, rosiglitazone also inhibits lipopolysaccharide-induced activation of JNK in primary macrophages (Supplemental Fig. 4). Therefore, inhibition of the JNK pathway may also be relevant to the anti-inflammatory action of TZDs, a property that may also contribute to ameliorate insulin resistance (20). In line with this, in 3T3-L1 adipocytes and in vivo, rosiglitazone also inhibited TNF-α–induced activation and activity, respectively, of the pro-inflammatory p38 mitogen-activated protein kinase (MAPK) pathway, whereas it had no effect on extracellular-regulated protein kinase activity (Supplemental Fig. 5). Nonetheless, although rosiglitazone inhibitory action on the p38MAPK pathway may be relevant to other of its biological properties such as adipogenesis (39), the inverse correlation between obesity and p38MAPK activity found in adipose tissue (39), liver (Supplemental Fig. 5C), and skeletal muscle (data not shown) seems to preclude the direct involvement of this protein kinase in insulin resistance and, hence, a prominent role of its inhibition by TZDs in conducting the hypoglycemic action of these drugs.

In addition to enhancing insulin sensitivity, TZDs also improve pancreatic β-cell function in type 2 diabetic patients and murine models for type 1 and type 2 diabetes (16–19). Independent studies indicate that JNK activity mediates pancreatic β-cell death in response to pro-inflammatory cytokines, such as IL-1β (9), and that JNK inhibitors exert a protective action in these circumstances (8–10). Moreover, in the obese/hyperglycemic db/db mice, JNK inhibitors increase insulin gene expression (8) and delay pancreatic failure (7). In the present study, we report that the TZD rosiglitazone blocks JNK activation in response to IL-1β in two insulin-secreting cell lines and that this action correlates with the capacity of this drug to prevent the IL-1β–triggered death of these cells. On the basis of these preliminary observations in stable cell lines, we propose that TZDs may exert a direct protective effect on pancreatic β-cells in vivo. In addition to other indirect beneficial effects (such as lowering insulin secretory demands and pro-inflammatory cytokine levels, which eventually may also be related to the inhibitory action on the JNK pathway), this may contribute to delaying pancreatic failure, which marks the establishment of a diabetic condition.

Compelling evidence supports the notion that the insulin-sensitizing action of TZDs is mediated to a great extent by PPARγ. In this regard, there is a strong correlation between the binding of TZDs to PPARγ in vitro and the hypoglycemic action of these drugs in vivo (27,28). In addition, non-TZD PPARγ ligands, such as F-l-Leu, also increase insulin sensitivity (25), and activators of the PPARγ heterodimer partner, the retinoid X receptor, also have antidiabetic properties (40). Finally, patients who harbor a dominant-negative PPARγ allele in heterozygosis show severe insulin resistance and diabetes (41). In the present study, we demonstrate that effective concentrations of rosiglitazone in JNK inhibition draw a parallel with those required to bind to and activate PPARγ (30). In line with this, inhibition of JNK activation may be achieved by other TZDs, such as troglitazone, with a comparable potency to rosiglitazone, as well as, by other PPARγ agonist ligands, such as 15d-PGJ₂ and F-l-Leu, and can be blocked by the PPARγ antagonist BADGE. We also show that the action of rosiglitazone on the JNK pathway may be modulated in opposite ways by up- and downregulation of PPARγ protein level. In this regard, the inhibitory effect of this drug on JNK activity is enhanced in 3T3-L1 preadipocytes by overexpression of PPARγ. In contrast, PPARγ RNAi abrogates the action of this drug on the JNK pathway in 3T3-L1 preadipocytes and adipocytes. Taken together, our data indicate that the inhibition of JNK activation by TZDs is mediated by PPARγ.

In the light of the present study, we conclude that TZDs/PPARγ inhibit the JNK pathway in cell cultures and in vivo. From a general point of view, this interaction should be added to the growing list of inhibitory crosstalks between the nuclear receptor superfamily and the MAPK signaling pathways. In this regard, previous studies have shown a negative regulation of PPARγ transcriptional activity by direct JNK phosphorylation (42). Consistently, constitutive activation of the JNK pathway by expression of a MKK7 constitutively-activated mutant leads to the inhibition of PPARγ transcriptional activity (Supplemental Fig. 6A). The concomitant abrogation of rosiglitazone action on the JNK pathway in these conditions (Supplemental Fig. 6B) suggests that a PPARγ transcriptional-dependent mechanism is mediating the inhibition of JNK by TZDs.

In the context of type 2 diabetes and in addition to other actions of TZDs, such as interference with the IKK/NF-κB pathway, inhibition of the JNK pathway may be relevant for the hypoglycemic properties of TZDs/PPARγ because exacerbated JNK activity is involved in insulin resistance and pancreatic β-cell death. In fact, after original studies showing the insulin-sensitizing and pancreatic β-cell protective activity of JNK inhibitors (7,8), current drug discovery efforts are focused on the development of molecules with such inhibitory property for the treatment of diabetes (7,43). In summary, this study contributes to a better understanding of the mechanism that underlies the pharmacological actions of TZDs and provides further support to and rationale for research on JNK and JNK inhibitors as therapeutic target and agents, respectively, for the treatment of diabetes.

J.D.-D. has received a predoctoral fellowship from the Ministerio de Educación y Cultura from Spain. M.M. has received a predoctoral fellowship from the Ministerio de Educación y Cultura from Spain. This study has received grants from the Ministerio de Ciencia y Tecnología (BFU2004-02096/BMC), the Ministerio de Sanidad y Consumo (PI021326), and the Generalitat de Catalunya, Spain.

We thank Drs. P. Maechler, M. Camps, R.A. Flavell, and B.M. Spiegelman for providing the INS-1E and 3T3-L1 cell lines, JNK1^−/− mice, and the PPARγ2 expression vector, respectively; C. Vila and Luc Marti for help with the in vivo and primary adipocyte studies, respectively; and T. Yates for style correction.