Peroxisome proliferator-activated receptor alpha activators improve insulin sensitivity and reduce adiposity.

Fibrates and glitazones are two classes of drugs currently used in the treatment of dyslipidemia and insulin resistance (IR), respectively. Whereas glitazones are insulin sensitizers acting via activation of the peroxisome proliferator-activated receptor (PPAR) gamma subtype, fibrates exert their lipid-lowering activity via PPARalpha. To determine whether PPARalpha activators also improve insulin sensitivity, we measured the capacity of three PPARalpha-selective agonists, fenofibrate, ciprofibrate, and the new compound GW9578, in two rodent models of high fat diet-induced (C57BL/6 mice) or genetic (obese Zucker rats) IR. At doses yielding serum concentrations shown to activate selectively PPARalpha, these compounds markedly lowered hyperinsulinemia and, when present, hyperglycemia in both animal models. This effect relied on the improvement of insulin action on glucose utilization, as indicated by a lower insulin peak in response to intraperitoneal glucose in ciprofibrate-treated IR obese Zucker rats. In addition, fenofibrate treatment prevented high fat diet-induced increase of body weight and adipose tissue mass without influencing caloric intake. The specificity for PPARalpha activation in vivo was demonstrated by marked alterations in the expression of PPARalpha target genes, whereas PPARgamma target gene mRNA levels did not change in treated animals. These results indicate that compounds with a selective PPARalpha activation profile reduce insulin resistance without having adverse effects on body weight and adipose tissue mass in animal models of IR.

MS, 1 which develops as a result of IR (1), is characterized by glucose intolerance, hyperinsulinemia, dyslipidemia, and hypertension. These metabolic abnormalities are frequently associated with visceral obesity (2). The clustering of multiple cardiovascular risk factors in MS results in increased risk for atherosclerotic vascular disease, the major cause of mortality and morbidity in type 2 diabetic patients (3). Pharmacological treatment of MS should therefore aim at ameliorating IR and reducing cardiovascular risk factors.
The PPAR nuclear receptors are important regulators of glucose and lipid homeostasis, which are activated by two classes of drugs: fibrates and glitazones (4). Glitazones are PPAR␥ activators, currently used for the treatment of IR and type 2 diabetes (5). These compounds increase fatty acid (FA) uptake in adipose tissue (6), due to PPAR␥-mediated induction of lipoprotein lipase (LPL) and FA transporter proteins (7,8). These actions are considered major determinants of the effects of glitazones on glucose homeostasis, since they favor the diversion of FA from muscles resulting in a relief of inhibition of peripheral glucose utilization (6). In addition, glitazones may also act by ameliorating TNF␣-induced insulin resistance (9 -11). However, PPAR␥ activation also enhances adipose differentiation and fat storage (12). Moreover, glitazones increase food intake (13,14), at least in part through the repression of leptin gene expression in adipose tissue (14,15). This action is likely to be mediated via PPAR␥ since heterozygous PPAR␥deficient mice display increased adipose tissue leptin expression accompanied with lowered food intake (16). Accordingly, increased body weight gain (13,17) and adipose tissue mass (14) have been reported in rodents upon glitazone treatment, a feature that might also occur in humans (18).
Results from the Helsinki Heart Study demonstrated that fibrates significantly reduce the incidence of cardiovascular disease in patients with type 2 diabetes (19). Fibrates are hypolipidemic drugs that are very efficient in lowering elevated triglyceride concentrations consistently observed in these patients (20). The action of fibrates on lipid metabolism is mediated principally by activation of PPAR␣ leading to altered expression of genes involved in lipid and lipoprotein metabolism in liver (21). Since fibrate treatment results in increased hepatic oxidation of fatty acids and reduced synthesis and secretion of triglycerides (20), as well as decreased plasma concentrations of cytokines, such as TNF␣ (22,23), we hypothesized that selective PPAR␣ activators might also improve glucose homeostasis. To test this hypothesis, we assessed therefore the influence of selective PPAR␣ activators on glucose homeostasis and body weight control in animal models of IR.

EXPERIMENTAL PROCEDURES
Animals-A first series of experiments was performed on male C57BL/6 mice, 8 weeks of age at the start of the experiment, which were randomly assigned to three different diets for 14 weeks. The mice received a low fat diet (UAR AO4), a high fat diet containing coconut oil (29% w/w) as described (24), or the same high fat diet supplemented with fenofibrate (0.05% w/w). A second series of experiments was performed with male Zucker rats of different ages, either bred at the U465 INSERM animal facility from pairs originally provided by the Harriet G. Bird Laboratory (Stow, MA) or obtained from Iffa-Credo (L'Arbresle, France). In the first experiment, 5-week-old obese fa/fa Zucker rats (n ϭ 6 per group; U465 INSERM breeding facility) were fed a standard diet with or without ciprofibrate (0.005% w/w) for 15 days and subsequently subjected to an intravenous glucose tolerance test (IVGTT). In the second experiment, 8 lean Fa/? and 14 obese fa/fa 21-week-old Zucker rats (Iffa-Credo breeding facility) were randomized in two groups per genotype, based on base-line body weight and serum triglyceride and glucose concentrations. Rats of each group were given a standard rat diet with or without ciprofibrate (0.005% w/w) for 21 days. In the third experiment, 12 obese fa/fa 20-week-old Zucker rats (U465 INSERM breeding facility) were randomized in two groups/genotype based on base-line body weight and treated for 9 days once daily by oral gavage with GW9578 (5 mg/kg/day) (25) or vehicle.
The food intake of the animals was carefully monitored by weighing special gridded metal food containers at regular intervals. In all experiments, body weights were monitored throughout the treatment period. Except when glucose tolerance tests were performed, food was removed at 8 a.m. and blood samples collected 4 h later at the end of the treatment. Animals were euthanized and tissues collected and weighed. Serum or plasma was isolated and stored at Ϫ20°C until further analysis.
Intravenous Glucose Tolerance Test-Animals were anesthetized at 2:00 p.m. after a 5-h fast by an intraperitoneal injection of sodium pentobarbital (50 mg/kg). Rats were injected with glucose (0.55 g/kg) in the saphenous vein and blood samples were collected from the tail vein in heparinized tubes at 0, 5, 10, 15, 20, and 30 min after the glucose load. Samples were kept on ice, and plasma was isolated and stored at Ϫ20°C until analysis.
Serum Assays-Glucose concentrations were measured using enzymatic methods, insulin (CIS Bio) and leptin (Linco) by radioimmunoassay.

RESULTS
To determine whether selective PPAR␣ activators influence insulin and glucose homeostasis, experiments were performed in animal models of IR using ciprofibrate and fenofibrate, which are the most PPAR␣-selective fibrates currently available in clinics, as well as with the novel PPAR␣ subtypeselective agonist GW9578 (25). Results from in vitro transactivation assays demonstrated that ciprofibrate (data not shown) and fenofibrate (25) activate PPAR␣ with EC 50 values of 20 and 30 M, respectively, whereas PPAR␥ is only marginally activated by any of these compounds (EC 50 value for PPAR␥ of 300 and Ͼ300 M for fenofibrate and ciprofibrate respectively). Neither ciprofibrate nor fenofibrate activates PPAR␦ (Ref. 25 and data not shown). The third compound, GW9578, was chosen on the basis of its high activity and specificity for PPAR␣, with EC 50 values for murine PPAR␣ of 5 nM (as opposed to 1.5 M for PPAR␥ and 2.6 M for PPAR␦) (25).
The influence of fenofibrate on glucose homeostasis was first analyzed in C57BL/6 mice, which develop obesity and IR when fed a high fat diet (24). Feeding C57BL/6 mice with the high fat diet resulted in hyperinsulinemia and mild hyperglycemia, which were both corrected by low dose fenofibrate treatment (Fig. 1, A and B). By contrast, treatment of chow-fed C57BL/6 mice (n ϭ 10/group) during 10 weeks with fenofibrate (0.1% w/w) incorporated in the diet did not influence serum glucose (untreated: 1.89 Ϯ 0.33 versus treated: 2.03 Ϯ 0.48 g/liter) or insulin (untreated: 17.6 Ϯ 6.0 versus treated: 16.3 Ϯ 7.8 microunits/ml) concentrations, indicating that the effects of fenofibrate on glucose homeostasis occur only in IR fat-fed mice.
Interestingly, fenofibrate treatment prevented the high fat diet-induced increase in body weight (Fig. 2) and adipose tissue mass ( Fig. 1, C and D). Identical results were obtained in a separate experiment with C57BL/6 mice submitted to the same nutritional protocol (Fig. 3). In these animals, serum leptin concentrations were measured and found to be positively correlated with body weight and epididymal adipose tissue weight, a relationship that was not influenced by fenofibrate treatment (Fig.  3). These observations indicate that fenofibrate does not exert a specific regulatory effect on leptin production. Interestingly, despite lower leptin concentrations, food intake was not increased in the mice fed the fenofibrate-enriched diet (13.8 Ϯ 1.0 kcal/day/ animal; n ϭ 14) versus high fat diet alone (13.8 Ϯ 0.9 kcal/day/ animal; n ϭ 14). These data further show that the effects of fenofibrate on glucose homeostasis, body weight, and adipose tissue mass are not driven by a reduction in caloric intake. To evaluate which PPAR subtype was activated in vivo, Northern blot analysis of PPAR␣-and PPAR␥-specific target gene expression was performed. To this aim, the mRNA levels of the gene coding for the fatty acid transporter CD36/FAT, which has been implicated in the development of the IR syndrome (27), were measured based on the observation that its expression is regulated in a tissue-selective manner by activators of PPAR␣ in liver and PPAR␥ in adipose tissue (28). In the C57BL/6 mice treated with fenofibrate, a marked increase in CD36/FAT mRNA levels were observed in the liver, whereas mRNA levels of CD36/FAT remained unchanged in epididymal adipose tissue (Fig. 4). These observations suggest that fenofibrate treatment resulted in selective activation of PPAR␣ in liver, but not of PPAR␥ in adipose tissue of these mice.
To test whether PPAR␣ activators might also improve glucose homeostasis in another model of IR, we chose the obese Zucker rat, which bears a mutation in the leptin receptor gene resulting in early onset obesity and marked hyperinsulinemia (29). Depending on the genetic background, these rats may also develop hyperglycemia later in life. Ciprofibrate treatment of 5-week-old obese Zucker rats lowered body weight gain and epididymal adipose tissue mass and reduced plasma insulin concentrations by almost 50% (Table I). Furthermore, the plasma insulin response to glucose during IVGTT was markedly decreased (Fig. 5), demonstrating a clear-cut improvement of insulin action on glucose utilization. Serum glucose concentrations and IVGTT glucose curves were normal and comparable between treated and untreated obese rats (Table I and Fig. 5).
At the dose administered to the obese Zucker rats, peak serum concentrations of ciprofibrate of 91 Ϯ 3 M were reached indicating, based on the EC 50 values for PPAR␣ (20 M) and PPAR␥ (Ͼ 300 M) activation, that ciprofibrate likely activates selectively PPAR␣ in the obese rats. This was further demonstrated by a marked increase in CD36/FAT mRNA levels observed in the liver, with no change in both epididymal and perirenal adipose tissue (Fig. 7A). Furthermore, mRNA levels of LPL and leptin, two other genes regulated by PPAR␥, but not by PPAR␣ activators in adipose tissue of rats (7,14,15), were similar in epididymal adipose tissue of treated and control animals (LPL: untreated, n ϭ 7:100 Ϯ 16% versus treated, n ϭ 7:109 Ϯ 11%; lep: untreated, n ϭ 7:100 Ϯ 18% versus treated, n ϭ 7:126 Ϯ 8%). By contrast, hepatic mRNA levels of the PPAR␣ target genes (21, 28) apoA-I (untreated, n ϭ 7:100 Ϯ 37% versus treated, n ϭ 7:35 Ϯ 12%) and apoC-III (untreated, n ϭ 7:100 Ϯ 12% versus treated, n ϭ 7:35 Ϯ 4%) were significantly affected by ciprofibrate treatment. In addition, the ac-   5. Ciprofibrate decreases the plasma insulin response to glucose in obese Zucker rats. 5-Week-old obese Zucker rats (n ϭ 5/group) were treated for 15 days with ciprofibrate (0.005% w/w) and subsequently submitted to an IVGTT. Results are represented as the mean Ϯ S.E. AUC, area under the curve. Statistically significant differences between groups (t test, p Ͻ 0.01) are indicated by an asterisk. A, insulin curve; B, glucose curve. tivity and mRNA levels of CPT-I and CPT-II as well as ␤-oxidation rates in the liver were significantly enhanced by ciprofibrate treatment (data not shown). These data indicate that ciprofibrate treatment results in selective PPAR␣, but not PPAR␥ activation in these obese Zucker rats.
Similar results were obtained in a separate series of old obese Zucker rats treated for 9 days with the highly specific PPAR␣ agonist GW9578 (25). In contrast to the ciprofibrate experiment, these obese Zucker rats were still normoglycemic at the age of 21 weeks, but serum insulin concentrations were significantly elevated, indicating a state of insulin resistance. Treatment with GW9578 resulted in markedly reduced serum insulin concentrations, whereas serum glucose levels were not affected (Fig. 6). Neither serum leptin levels nor body weights were changed, probably due to the short time of treatment in this particular experiment (Fig. 6). No effect of the treatment was observed on food consumption (untreated: 25.6 Ϯ 6.7 g/day versus treated:25.0 Ϯ 3.2 g/day), precluding an effect via dietary changes. When mRNA levels of CD36/FAT were measured, a pronounced increase of liver CD36/FAT mRNA levels was observed, whereas both epididymal and perirenal adipose tissue CD36/FAT mRNA levels remained unchanged after GW9578 treatment (Fig. 7). Furthermore, GW9578 treatment did not influence LPL or leptin mRNA levels in both adipose tissue depots (data not shown). Thus, as with ciprofibrate, treatment with GW9578 resulted in efficient PPAR␣ activation, whereas PPAR␥ was not activated in these animals.

DISCUSSION
Previous studies have demonstrated that certain naturally occurring and ␤-substituted FA (such as conjugated linoleic acid and MEDICA 16) (30, 31) and the fibrate bezafibrate (32) improve insulin sensitivity and normalize impaired glucose tolerance in rat models of IR. In humans bezafibrate may also improve glucose homeostasis (33)(34)(35). Although fatty acids are PPAR␣ activators, they also activate PPAR␥ and PPAR␦ equally well (36). Similarly, in contrast to ciprofibrate and fenofibrate, bezafibrate activates PPAR␣, PPAR␥, and PPAR␦ with comparable EC 50 values (PPAR␣: 50; PPAR␥: 60 M; PPAR␦: 20 M; Ref. 25). Therefore, it is impossible to conclude via which PPAR form the effects of these compounds on glucose homeostasis are mediated. In the present study we demonstrate that, in two models of diet-induced and genetic obesitylinked IR, PPAR␣ activators correct elevated serum glucose and insulin concentrations by increasing insulin action on glucose utilization. Most importantly, fibrate PPAR␣ activators also decrease adipose tissue mass, by a mechanism independent of changes in food intake and leptin gene expression. This effect is in sharp contrast to PPAR␥ activators, which increase body weight (13,17,18) and epididymal adipose tissue mass in rodents (14).
Several lines of evidence indicate that the nuclear receptor PPAR␣ mediates the actions of ciprofibrate, fenofibrate and GW9578 on glucose homeostasis. First, based on the EC 50 values of these compounds for the different PPARs and their serum concentrations attained in the obese Zucker rats, ciprofibrate and fenofibrate likely activate PPAR␣ maximally, whereas PPAR␥ activation is negligible. Furthermore, ciprofibrate treatment increased mitochondrial ␤-oxidation and serum concentrations of ketone bodies (data not shown) in obese Zucker rats, which is consistent with PPAR␣ activation. This is in contrast to the effect of glitazone PPAR␥ activators, which decrease serum ketone bodies (37,38) and have either no effect on (39) or may even decrease fatty acid oxidation (6,38). Second, treatment of obese Zucker rats with the highly specific PPAR␣ ligand GW9578 resulted in a decrease of serum insulin concentrations. Finally, whereas glitazones have been shown to markedly influence gene expression in adipose tissue of both normal and obese Zucker rats (7,13,28), GW9578 and ciprofibrate did not influence the expression of any known PPAR␥ target genes, including CD36/FAT, in adipose tissue. In sharp contrast, PPAR␣ target gene expression in liver was significantly altered. These data provide in vivo evidence for selective activation of PPAR␣, but not PPAR␥ in the ciprofibrate-and GW9578-treated Zucker rats as well as in the fenofibratetreated C57BL/6 mice.
Although not the subject of the present study, PPAR␣ ligands may influence body weight and glucose homeostasis FIG. 6. PPAR␣ activators decrease serum insulin concentrations in older obese Zucker rats. Two independent experiments were performed using obese Zucker rats. In both experiments, animals were randomized in two groups. Body weights and serum leptin, insulin, and glucose concentrations were not significantly different between control and treated groups at the beginning of each treatment period (data not shown). In both experiments, the control group consisted of obese Zucker rats treated with vehicle. In the first experiment, animals were treated with ciprofibrate (0.005% w/w in chow for 21 days; n ϭ 7/group), whereas, in the second experiment, treatment was done with GW9578 (5 mg/kg/day by gavage for 9 days; n ϭ 6/group). Body weights and serum insulin, glucose, and leptin concentrations were measured at the end of each treatment period. Results are expressed as the mean Ϯ S.D. Statistically significant differences between treated and control (t test; p Ͻ 0.05; *) groups are indicated.
FIG. 7. PPAR␣ activators increase CD36/FAT mRNA levels in liver, but not in adipose tissue of older obese Zucker rats. Obese Zucker rats were treated as described under Fig. 6. RNA was extracted from liver, epididymal, and perirenal adipose tissue from individual animals (ciprofibrate experiment: n ϭ 7/group; GW9578 experiment: n ϭ 6/group) and CD36/FAT mRNA levels measured as described under "Experimental Procedures." Results, expressed as the mean Ϯ S.D., are normalized to control 36B4 mRNA levels. Statistically significant differences between treated and control groups were determined using t test (*, p Ͻ 0.01). A, ciprofibrate experiment; B, GW9578 experiment. through different mechanisms. At the doses employed in the present study, fibrates do not appear to have major effects on adipose tissue (present study and Refs. 7 and 8). In contrast, fibrates increase hepatic ␤-oxidation in obese Zucker rats (data not shown). This catabolic action on FA would result in an increased FA flux from peripheral tissues, such as skeletal muscle and adipose tissue, to the liver, a decreased FA synthesis, and a lowered delivery of triglycerides to peripheral tissues. As such fibrates might alleviate the FA-mediated inhibition of insulin-stimulated oxidative and non-oxidative glucose disposal in skeletal muscle (40,41), thus ameliorating IR. Furthermore, by lowering plasma triglycerides, fibrates may decrease skeletal muscle triglyceride content, which is significantly related to IR and obesity (42). As an alternative or concomitant mechanism, fibrates might improve insulin action by decreasing production of cytokines, such as interleukin-6 and TNF␣. TNF␣ concentrations are increased in IR obese humans (43) and Zucker rats (44), as well as in rodents fed a high fat diet (45). This cytokine has been implicated in the development of IR by interfering negatively with insulin signaling (44). In support for this hypothesis is the demonstration that PPAR␣ activators inhibit NFB signaling, resulting in lowered production of cytokines by smooth muscle cells and decreased plasma concentrations of cytokines (22,23). Interestingly, such a mechanism has also been suggested to participate, at least in part, in the insulin-sensitizing effect of glitazone PPAR␥ activators (11,46). Finally, it cannot be excluded that PPAR␣ agonists exert direct insulin-sensitizing actions.
In contrast to glitazones, which are high affinity PPAR␥ ligands, the current clinically used fibrates are low affinity PPAR␣ ligands. In conclusion, the results from this study suggest that PPAR␣ ligands with higher affinity, in addition to being useful for the treatment of dyslipidemia, may also be of use to improve insulin sensitivity. Further studies in patients with MS are required to determine whether highly active and selective PPAR␣ agonists also improve glucose homeostasis in man.