Rimonabant Ameliorates Insulin Resistance via both Adiponectin-dependent and Adiponectin-independent Pathways*

Rimonabant has been shown to not only decrease the food intake and body weight but also to increase serum adiponectin levels. This increase of the serum adiponectin levels has been hypothesized to be related to the rimonabant-induced amelioration of insulin resistance linked to obesity, although experimental evidence to support this hypothesis is lacking. To test this hypothesis experimentally, we generated adiponectin knock-out (adipo(-/-))ob/ob mice. After 21 days of 30 mg/kg rimonabant, the body weight and food intake decreased to similar degrees in the ob/ob and adipo(-/-)ob/ob mice. Significant improvement of insulin resistance was observed in the ob/ob mice following rimonabant treatment, associated with significant up-regulation of the plasma adiponectin levels, in particular, of high molecular weight adiponectin. Amelioration of insulin resistance in the ob/ob mice was attributed to the decrease of glucose production and activation of AMP-activated protein kinase (AMPK) in the liver induced by rimonabant but not to increased glucose uptake by the skeletal muscle. Interestingly, the rimonabant-treated adipo(-/-)ob/ob mice also exhibited significant amelioration of insulin resistance, although the degree of improvement was significantly lower as compared with that in the ob/ob mice. The effects of rimonabant on the liver metabolism, namely decrease of glucose production and activation of AMPK, were also less pronounced in the adipo(-/-)ob/ob mice. Thus, it was concluded that rimonabant ameliorates insulin resistance via both adiponectin-dependent and adiponectin-independent pathways.


Rimonabant has been shown to not only decrease the food intake and body weight but also to increase serum adiponectin levels. This increase of the serum adiponectin levels has been hypothesized to be related to the rimonabant-induced amelioration of insulin resistance linked to obesity, although experimental evidence to support this hypothesis is lacking.
To test this hypothesis experimentally, we generated adiponectin knock-out (adipo(؊/؊))ob/ob mice. After 21 days of 30 mg/kg rimonabant, the body weight and food intake decreased to similar degrees in the ob/ob and adipo(؊/؊)ob/ob mice. Significant improvement of insulin resistance was observed in the ob/ob mice following rimonabant treatment, associated with significant up-regulation of the plasma adiponectin levels, in particular, of high molecular weight adiponectin. Amelioration of insulin resistance in the ob/ob mice was attributed to the decrease of glucose production and activation of AMP-activated protein kinase (AMPK) in the liver induced by rimonabant but not to increased glucose uptake by the skeletal muscle. Interestingly, the rimonabanttreated adipo(؊/؊)ob/ob mice also exhibited significant amelioration of insulin resistance, although the degree of improvement was significantly lower as compared with that in the ob/ob mice. The effects of rimonabant on the liver metabolism, namely decrease of glucose production and activation of AMPK, were also less pronounced in the adipo(؊/؊)ob/ob mice. Thus, it was concluded that rimonabant amelio-rates insulin resistance via both adiponectin-dependent and adiponectin-independent pathways.
The prevalence of obesity has increased dramatically in recent years (1,2). It is commonly associated with type 2 diabetes, coronary artery disease, and hypertension, and the coexistence of these diseases in subjects has been termed the metabolic syndrome (3). There is a demand for effective and safe antiobesity agents that can produce and maintain weight loss and improve the metabolic syndrome.
The newly discovered endocannabinoid system, consisting of the CB-1 (cannabinoid type-1) receptor and endogenous lipid-derived ligands, contributes to the physiological regulation of energy balance, food intake, and lipid and glucose metabolism, through both central orexigenic effects and peripheral metabolic effects (4 -11). The endocannabinoid system is overactivated in genetic animal models of obesity (4,6), and the selective CB-1 blocker, rimonabant, produces weight loss and ameliorates metabolic abnormalities in obese animals (12,13). Patients with obesity and hyperglycemia associated with type 2 diabetes exhibit higher concentrations of endocannabinoids in the visceral fat and serum, respectively, than the corresponding controls (14). Rimonabant has been shown to produce substantial weight loss and reduction of waist circumference and also improve insulin resistance and the profile of several metabolic and cardiovascular risk factors in diabetic as well as nondiabetic obese patients (15)(16)(17)(18).
Adiponectin is an adipokine that is specifically and abundantly expressed in the adipose tissue and released into the circulation, which directly sensitizes the body to insulin (19,20). Administration of recombinant adiponectin to rodents increases the glucose uptake and fat oxidation in muscle, reduces hepatic glucose production, and improves whole body insulin sensitivity (21)(22)(23). Adiponectin-deficient (adipo(Ϫ/Ϫ)) mice exhibit insulin resistance and glucose intolerance (24,25). Previous stud-ies have shown that adiponectin stimulates fatty acid oxidation in the skeletal muscle and inhibits glucose production in the liver by activating AMP-activated protein kinase (AMPK) 2 (26 -29). We also reported that pioglitazone may induce amelioration of insulin resistance and diabetes via an adiponectin-dependent mechanism in the liver and an adiponectin-independent mechanism in the skeletal muscle (30).
Rimonabant has been shown to increase the plasma adiponectin levels in animal models of obesity and diabetes as well as in both diabetic and nondiabetic subjects (15,31,32). The results of the RIO-Lipids study provided evidence of a weight loss-independent effect of rimonabant on the plasma adiponectin levels (15). Furthermore, the metabolic improvements induced by rimonabant could be attributed, at least in part, to a moderate but significant increase in the plasma circulating adiponectin levels (15). However, whether the rimonabant-induced increase in the plasma levels of adiponectin might be causally involved in the effects of rimonabant, in particular its insulin-sensitizing effects, has not been addressed experimentally.
To address this issue, in the present study, we used adipo(Ϫ/Ϫ)ob/ob mice (30) to investigate whether rimonabant might be capable of ameliorating insulin resistance in the absence of adiponectin. We found that rimonabant significantly decreased the body weight and food intake to similar degrees in the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice. Furthermore, we found significant amelioration of the insulin resistance in the ob/ob mice, in association with significant up-regulation of the serum adiponectin levels after 21 days of treatment with rimonabant at 30 mg/kg, body weight. The amelioration of insulin resistance in the ob/ob mice was attributed to the decrease of glucose production and activation of AMPK in the liver but not the increased glucose uptake by the skeletal muscle, induced by the drug. Interestingly, insulin resistance was also significantly, although only partially, improved in the adipo(Ϫ/Ϫ)ob/ob mice. Thus, the results suggest that rimonabant ameliorates insulin resistance via both adiponectin-dependent and adiponectin-independent pathways.

EXPERIMENTAL PROCEDURES
Animals and Genotyping-The mice were housed under a 12-h light/dark cycle and fed standard chow, CE-2 (CLEA Japan Inc., Tokyo, Japan). The composition of the chow was as follows: 25.6% (w/w) protein, 3.8% fiber, 6.9% ash, 50.5% carbohydrates, 4% fat, and 9.2% water. Ob/ob and adipo(Ϫ/Ϫ)ob/ob mice were generated by intercrossing adipo(ϩ/Ϫ)ob/ϩ mice. All the mice were maintained on a C57Bl/6 background (30). All of the experiments in this study were conducted on 16 -20week-old male littermates. The animal care and experimental procedures were approved by the Animal Care Committee of the University of Tokyo.
Rimonabant Treatment Study-Rimonabant (SR141716) or vehicle (0.1% Tween 80 in saline) was administered to ob/ob and adipo(Ϫ/Ϫ)ob/ob mice at a dose of 30 mg/kg body weight by oral gavage, once daily for 21 consecutive days. Rimonabant was kindly provided by Sanofi-Aventis (Montpellier, France). We measured the body weights and food intake of the mice once daily for 21 consecutive days.
Hyperinsulinemic-Euglycemic Clamp Study-Clamp studies were carried out as described previously (30) with slight modifications. In brief, 2 days before the study, an infusion catheter was inserted into the right jugular vein under general anesthesia induced by sodium pentobarbital. Studies were performed on the mice under conscious and unstressed conditions after 8 h of fasting. A primed continuous infusion of insulin (Humulin R; Lilly) was administered (25.0 milliunits/kg/min), and the blood glucose concentration, monitored every 5 min, was maintained at 100 -130 mg/dl by administration of glucose (5 g of glucose/10 ml enriched to ϳ20% with [6,6-2 H 2 ]glucose (Sigma)) for 120 min. Blood was sampled via tail tip bleeds at 90, 105, and 120 min for determination of the rate of glucose disappearance (R d ). R d was calculated according to nonsteady-state equations (30), and endogenous glucose production was calculated as the difference between the R d and the exogenous glucose infusion rate (30).
Western Blot Analysis-Tissues were excised and homogenized in ice-cold buffer A (25 mM Tris-HCl (pH 7.4), 10 mM sodium orthovanadate, 10 mM sodium pyrophosphate, 100 mM sodium fluoride, 10 mM EDTA, 10 mM EGTA, and 1 mM phenylmethylsulfonyl fluoride). The sample buffer for analysis under reducing conditions was composed of 3% SDS, 50 mM Tris-HCl (pH 6.8), 5% 2-mercaptoethanol, and 10% glycerol. Samples were mixed with 5ϫ sample buffer, heated at 95°C for 5 min for heat denaturation, separated on polyacrylamide gels, and then transferred to a Hybond-P polyvinylidene difluoride transfer membrane (Amersham Biosciences). Bands were detected with ECL detection reagents (Amersham Biosciences). To examine the Akt and AMPK phosphorylation and protein levels, lysates of liver and muscle were analyzed using anti-phospho-Akt (Cell Signaling Technology, Inc., Beverly, MA), anti-Akt (Cell Signaling Technology, Inc.) antibody, anti-phospho-AMPK (Cell Signaling Technology, Inc., Beverly, MA), and anti-AMPK (Cell Signaling Technology, Inc.) antibodies. For the analysis under nonreducing conditions, 2-mercaptoethanol was excluded from the sample buffer described above. To examine the isoforms of adiponectin, the serum samples were diluted 20-fold. Anti-mouse adiponectin antiserum was obtained by immunizing rabbits with the globular domain of mouse recombinant adiponectin produced in Escherichia coli (21).
Tissue Sampling for Insulin Signaling Pathway Study-Mice were anesthetized after 16 h of starvation, and 0.05 unit of human insulin (Humulin R; Lilly) was injected into the inferior vena cava. After 5 min, the liver was removed, and the specimens were used for protein extraction as described above.
Plasma Adiponectin and Lipid Measurements-The mice were deprived of access to food for 16 h before the measurements. The plasma adiponectin levels were determined with a mouse adiponectin enzyme-linked immunosorbent assay kit (Otsuka Pharmaceutical Co., Ltd., Tokyo, Japan). Serum triglyceride and free fatty acids (Wako Pure Chemical Industries Ltd., Osaka, Japan) were assayed by enzymatic methods.
Measurement of Adipocyte Size-Epididymal white adipose tissue and subcutaneous fat were routinely processed for paraffin embedding, and 4-m sections were cut and mounted on silanized slides. The adipose tissue sections were stained with hematoxylin and eosin, and the total adipocyte area was manually traced and analyzed using the Win ROOF software (Mitani Co. Ltd., Chiba, Japan). The white adipocyte area was measured in 200 or more cells/mouse in each group, in accordance with a previously described method (30), with slight modifications.
Oil Red O Staining and Quantification-Lipid accumulation was assessed by Oil Red O staining in 6-m frozen sections of the liver fixed in phosphate-buffered 4% paraformaldehyde, according to a previously described method (33) with slight modification. In brief, the livers were washed once for 1 min with H 2 O. After an additional wash for 1 min with 60% isopropyl alcohol, the livers were stained for 10 min at 37°C with freshly diluted Oil Red O solution (6 parts of Oil Red O stock solution and 4 parts of H 2 O; the Oil Red O stock solution contained 0.5% Oil Red O in isopropyl alcohol). After one wash for 2 min with 60% isopropyl alcohol and one wash for 1 min with H 2 O, the livers were stained for 5 min with hematoxylin. The stain was then washed off with running water, and the silanized slides were stained. Oil Red O staining was quantified on digital images. Color images were acquired with a Nikon digital camera and analyzed using the Image J software. The percentage of the area of Oil Red O staining was measured from 9 -10 different sections/mouse in each experimental group. Values were expressed as percentage of area.
Analysis of O 2 Consumption-Oxygen consumption was measured every 3 min for 24 h in the fasting mice using an O 2 /CO 2 metabolism measurement device (model MK-5000; Muromachikikai, Tokyo, Japan). After rimonabant treatment for 21 days, each mouse was placed in a sealed chamber (560-ml volume) with an air flow rate of 500 ml/min at room temperature. The amount of oxygen consumed was converted to ml/min by multiplying it with the flow rate.
RNA Preparation and Taqman PCR-Total RNA was extracted from various tissues in vivo with TRIzol reagent (Invitrogen), in accordance with the manufacturer's instructions. After treatment with RQ1 RNase-free DNase (Promega, Madison, WI) to remove genomic DNA, cDNA was synthesized with MultiScribe reverse transcriptase (Applied Biosystems, Foster City, CA). Total RNA was prepared from 3T3L1 cells in vitro with an RNeasy Mini Kit (Qiagen Co., Düsseldorf, Germany), in accordance with the manufacturer's instructions. mRNA levels were quantitatively analyzed by fluorescencebased reverse transcriptase-PCR. The reverse transcription mixture was amplified with specific primers using an ABI Prism 7000 sequence detector equipped with a thermocycler. The primers used for MCP-1 (monocyte chemoattractant protein-1), resistin, phosphoenolpyruvate carboxykinase (PEPCK), carnitine palmitoyltransferase-1A, the hepatic isoform of carnitine palmitoyltransferase-1, protein phosphatase 2C, and cyclophilin were purchased from Applied Biosystems (Foster City, CA). The relative expression levels were compared by normalization to the expression levels of cyclophilin.
Cell Culture and Differentiation of 3T3L1 Adipocytes and Rimonabant Treatment-3T3L1 preadipocytes were cultured in Dulbecco's modified Eagle's medium containing 25 mM glucose and 10% fetal bovine serum at 37°C. Confluent cultures were induced to differentiate into adipocytes by incubation in Dulbecco's modified Eagle's medium containing 25 mM glucose, 10% fetal bovine serum, 0.25 units/ml insulin, 0.25 M dexamethasone, and 0.5 mM isobutyl-1-methylxanthine. After 2 days, the medium was changed to Dulbecco's modified Eagle's medium containing 25 mM glucose, 10% fetal bovine serum, and 0.025 units/ml insulin. All studies were performed on adipocytes 10 days after the initiation of differentiation (Day 0). Rimonabant treatment (100 nM and 1 M) was started on Day 0, and DMSO was used as the vehicle. Prior to the start of the experiments, the differentiated adipocytes were serum-starved in Dulbecco's modified Eagle's medium containing 25 mM glucose for 16 h at 37°C.

Absence of Adiponectin Had No Effect on Rimonabant-induced Suppression of Body Weight and Daily
Food Intake-The body weight gain was similar between the untreated ob/ob and adipo(Ϫ/Ϫ)ob/ob mice (Fig. 1A), as reported previously (30). The food intake was also comparable between the untreated ob/ob and adipo(Ϫ/Ϫ)ob/ob mice (Fig. 1B). Rimonabant significantly decreased the body weight and food intake to similar degrees in the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice (Fig. 1, A and  B). After 21 days of rimonabant treatment, both the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice weighed 10% less than the corresponding untreated mice (Fig. 1A). Moreover, rimonabant treatment significantly decreased the white adipose tissue (WAT) mass in both subcutaneous and visceral (epididymal, mesenteric, and retroperitoneal) fat to similar degrees in the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice (Fig. 1C). To determine whether the presence of adiponectin is required for the reduction of the average adipocyte size induced by rimonabant treatment, we histologically analyzed the epididymal fat pad and subcutaneous WAT after fixation and quantitation of the adipocyte size. The distribution of the adipocyte size in the rimonabant-treated ob/ob and adipo(Ϫ/Ϫ)ob/ob mice was similarly narrowed to that in the untreated ob/ob and adipo(Ϫ/ Ϫ)ob/ob mice (Fig. 1, D and E), and rimonabant treatment significantly reduced the average adipocyte size in the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice to a similar degree (Fig. 1F). These findings indicate that the absence of adiponectin had no effect on either the rimonabant-induced decrease of the body weight or the food intake of the mice and that rimonabant treatment can induce a reduction of adipocyte size in the absence of adiponectin or leptin or both.
Rimonabant Increased the Energy Expenditure and Decreased the Serum Triglyceride and Free Fatty Acid Levels to a Similar Degree in the ob/ob and adipo(Ϫ/Ϫ)ob/ob Mice-In addition to suppressing food intake, rimonabant has been demonstrated to increase the energy expenditure (10,34), and the increase in energy expenditure has also been shown in CB-1 receptor knock-out mice (35). Since the involvement of adiponectin in this action of rimonabant remains unclear, we investigated the effects of rimonabant on energy expenditure. First, we measured the rectal temperature in the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice. The temperature was essentially the same ( Fig. 2A), and rimonabant treatment significantly increased the rectal temperature of the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice to a similar degree ( Fig. 2A). Second, we investigated the oxygen consumption after 21-day treatment with rimonabant and found that in the dark phase of the daily light cycle, rimonabant increased the energy expenditure to a similar degree in both the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice (Fig. 2B). This effect of rimonabant on the energy expenditure in the ob/ob mice did not require the presence of adiponectin. We next investigated the effects of rimonabant treatment on the serum lipid levels. In addition to reducing the body weight, rimonabant has been demonstrated to reduce the serum triglyceride (TG) (15)(16)(17)(18)32) and free fatty acid (FFA) levels (32). However, the involvement of adiponectin in this action of rimonabant remains unclear. Both the serum TG and FFA levels were indistinguishable between the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice (Fig. 2, C  and D), and rimonabant treatment significantly decreased the levels of both to similar degrees in the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice (Fig. 2,  C and D). This effect of rimonabant on the serum lipids in the ob/ob mice did not require the presence of adiponectin. MCP-1 and resistin have been shown to be important mediators of insulin resistance linked to obesity (36 -39). We analyzed the expression of MCP-1 and resistin in the epididymal WAT. The expressions of both MCP-1 and resistin were indistinguishable between the untreated and rimonabant-treated mice of either genotype (Fig. 2, E and F)  able in either the untreated or rimonabant-treated adipo(Ϫ/Ϫ)ob/ob mice (Fig. 3A). High molecular weight (HMW) adiponectin is known to be the most active, and its serum levels have been reported to be decreased in obese individuals and murine models, which is associated with a decrease of the hepatic and muscle AMPK activity and fatty acid combustion and, thereby, exacerbation of insulin resistance (19,20). Therefore, we analyzed the plasma levels of this isoform of adiponectin by Western blotting. Rimonabant treatment significantly increased the serum levels of HMW adiponectin in the ob/ob mice (Fig. 3B). On the other hand, the plasma levels of middle molecular weight and low molecular weight adiponectin were slightly, but not significantly, increased in the rimo-nabant-treated ob/ob mice (Fig.  3B). In regard to the adipo(Ϫ/Ϫ)ob/ob mice, plasma adiponectin was not detectable in either the untreated or rimonabant-treated mice (Fig. 3B). Rimonabant has been reported to increase adiponectin expression and secretion in 3T3F442A adipocyte (6,40). We next investigated the direct effect of rimonabant on adiponectin secretion using the murine adipocyte cell line 3T3L1 and confirmed that treatment with 100 nM and 1 M rimonabant actually increased the expression and secretion into the medium of adiponectin (Fig. 3, C  and D).

Rimonabant Improved Hepatic Insulin Resistance in both the ob/ob and adipo(Ϫ/Ϫ)ob/ob Mice, although the Effect Was Significantly Less Pronounced in the adipo(Ϫ/Ϫ)ob/ob Mice-We carried out hyperinsulinemic-euglycemic clamp studies in the ob/ob and adipo(Ϫ/Ϫ)-
ob/ob mice to investigate the effect of rimonabant on the insulin resistance in the liver and skeletal muscle. Without rimonabant treatment, the glucose infusion rates were comparable in the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice (Fig. 4A). After 21 days of rimonabant treatment, the glucose infusion rates were significantly increased in both the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice (Fig.  4A); however, the increase was significantly less pronounced in the adipo(Ϫ/Ϫ)ob/ob mice. Rimonabant treatment also produced a significant decrease of the endogenous glucose production in both the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice, but the effect was significantly less pronounced in the adipo(Ϫ/Ϫ)ob/ob mice (Fig. 4B). The rates of R d were indistinguishable between the untreated ob/ob and adipo(Ϫ/Ϫ)ob/ob mice, and rimonabant treatment had no effect on this parameter in either genotype (Fig. 4C). We next studied the effects on insulin signaling and the downstream reactions in the liver (Fig. 4, D and E). Insulin-stimulated Akt phosphorylation was significantly increased in rimonabanttreated ob/ob mice as compared with that in the untreated ob/ob mice (Fig. 4D), whereas insulin-stimulated Akt phosphorylation only tended to be increased in the rimonabant-treated adipo(Ϫ/Ϫ)ob/ob mice as compared with that in the corresponding untreated mice. The PEPCK expression levels in the liver were comparable in the untreated ob/ob and adipo(Ϫ/Ϫ)ob/ob mice (Fig. 4F). Rimonabant treatment significantly decreased the expression of PEPCK in both the ob/ob and adipo(Ϫ/Ϫ)mice, but the effect was significantly less pronounced in the adipo(Ϫ/Ϫ)ob/ob mice (Fig. 4F). These find-ings indicate that rimonabant ameliorates hepatic but not muscle insulin resistance in mice with an ob/ob background, in both an adiponectin-dependent and adiponectin-independent manner.
Rimonabant Increased the Hepatic AMPK Activities and CPT-1 (Carnitine Palmitoyltransferase-1) Expression Levels in both ob/ob Mice and adipo(Ϫ/Ϫ)ob/ob Mice, but Its Effect was Significantly Less Pronounced in the adipo(Ϫ/Ϫ)ob/ob Mice-We carried out analysis of the liver metabolic activity after the clamp studies to investigate the effect of rimonabant on amelioration of insulin resistance. The AMPK activities were comparable in the untreated ob/ob and adipo(Ϫ/Ϫ)ob/ob mice (Fig. 5A). Rimonabant treatment for 21 days increased the AMPK activities in both the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice, but its effect was significantly less pronounced in the adipo(Ϫ/Ϫ)ob/ob mice (Fig. 5A). The expression levels of CPT-1, the rate-limiting enzyme in fatty acid ␤-oxidation, were also comparable in the untreated ob/ob and adipo(Ϫ/Ϫ)ob/ob mice (Fig. 5B). Rimonabant treatment increased the CPT-1 expression in both ob/ob and adipo(Ϫ/Ϫ)ob/ob mice, but its effect was significantly less pronounced in the adipo(Ϫ/Ϫ)ob/ob mice (Fig. 5B). The expression levels of protein phosphatase 2C were indistinguishable between the untreated ob/ob and adipo(Ϫ/Ϫ)ob/ob mice, and rimonabant treatment had no effect on the protein phosphatase 2C expression in either genotype (Fig. 5C). As reported previously (26,41), fatty acid oxidation is positively regulated by AMPK in the liver; therefore, we next carried out analysis of the hepatic TG content by Oil Red O staining. The percentage of areas of Oil Red O staining in the liver were comparable in the untreated ob/ob and adipo(Ϫ/Ϫ)ob/ob mice (Fig. 5D). Rimonabant treatment significantly decreased the hepatic TG content in both the ob/ob and adipo(Ϫ/Ϫ)mice, but its effect was significantly less pronounced in the adipo(Ϫ/Ϫ)ob/ob mice  (Fig. 5D). We also investigated the AMPK activities in the muscle after the clamp studies. The AMPK activities in the muscle were indistinguishable between the untreated ob/ob and adipo(Ϫ/Ϫ)ob/ob mice, and rimonabant treatment had no effect on the muscle AMPK activity in either genotype (Fig. 5E). These findings indicate that rimonabant activates hepatic but not muscle AMPK in mice with an ob/ob background in both an adiponectin-dependent and adiponectin-independent manner.

DISCUSSION
The selective CB-1 blocker rimonabant has been reported to produce weight loss and ameliorate insulin resistance and metabolic abnormalities in obese animals (12,13), as also in patients with obesity (15)(16)(17)(18). Rimonabant has also been reported to increase the plasma adiponectin levels in animal models of obesity and diabetes, as also in diabetic or nondiabetic subjects (15,31,32). Adiponectin has been proposed to be a major insulin-sensitizing adipokine (19,20) and is a plausible candidate as the adipokine mediating the rimonabant-induced amelioration of insulin resistance. Therefore, in this study, we used two obesity models, the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice, to investigate whether the rimonabant-induced increase of plasma adiponectin might be causally involved in the insulin-sensitizing effects of the drug.
Rimonabant treatment decreased the body weight, food intake, and weight of the WAT to similar degrees in the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice. Furthermore, it also increased the energy expenditure and decreased the serum TG and FFA to similar degrees in the ob/ob and adipo(Ϫ/Ϫ)ob/ob mice. Thus, the involvement of adiponectin was not requited for rimonabant to exert its effects.
Significant improvement of the insulin resistance was observed in the ob/ob mice following rimonabant treatment, in association with significant up-regulation of the plasma adiponectin levels, in particular of HMW. Amelioration of insulin resistance in the ob/ob mice was considered to be attributable to improvement of the hepatic but not muscle insulin resistance. Interestingly, these improvements induced by rimonabant were significantly less pronounced in the adipo(Ϫ/Ϫ)ob/ob mice, indicating that adiponectin is involved in the rimonabant-mediated amelioration of hepatic insulin resistance. In fact, although a significant decrease of the PEPCK expression levels was observed, the AMPK activity was significantly increased, and the hepatic TG content was decreased in the ob/ob mice; all of these changes were significantly less pronounced in the adipo(Ϫ/Ϫ)ob/ob mice lacking adiponectin. We reported from a previous study that adiponectin, especially HMW adiponectin, stimulates AMPK activation in the liver (26,42). These findings suggest that rimonabant treatment activates AMPK in the liver via increasing the secretion of HMW adiponectin and then decreases the expression of PEPCK to inhibit glucose production and increase CPT-1 expression, thereby stimulating fatty acid oxidation in the liver.
On the other hand, rimonabant treatment also produced significant amelioration of hepatic insulin resistance in the absence of adiponectin. This amelioration was possibly attributable to the reduction of body weight (Fig. 1A) but not to suppression of MCP-1 and resistin expression (Fig. 2, E and F). Alternatively, this amelioration was possibly due to the direct activation of AMPK by rimonabant in the liver. In fact, recent reports have shown that AMPK activity was significantly higher in the liver of hepatocyte-specific CB-1 receptor knock-out mice, although the serum adiponectin levels in these animals remained unchanged (35,43), suggesting that rimonabant treatment directly activates hepatic AMPK, even without the mediation of adiponectin, and decreases the expression of PEPCK to inhibit glucose production in the liver.
In addition, Osei-Hyiaman et al. (35) have reported that CPT-1 activity in the liver was significantly increased when systemic CB-1 receptors were blocked pharmacologically in wildtype mice. Moreover, hepatic CPT-1 activity increased, and hepatic TG content decreased when hepatic CB-1 receptors were blocked genetically (35). These data suggest that CB-1 receptor blockade stimulates CPT-1 activity and increases fatty acid combustion to decrease the TG content in the liver. Consistent with this, rimonabant actually increased CPT-1 expression and decreased the TG content in the livers of ob/ob and adipo(Ϫ/Ϫ)ob/ob mice. However, these effects were markedly attenuated in the adipo(Ϫ/Ϫ)ob/ob mice, suggesting that increased CPT-1 expression and decreased hepatic TG content by rimonabant were also mediated by adiponectin-dependent as well as adiponectin-independent pathways.
Based on our findings, we propose that there are two distinct pathways by which rimonabant ameliorates insulin resistance, one an adiponectin-dependent pathway and the other an adiponectin-independent pathway (Fig. 6). Rimonabant increases the plasma levels of adiponectin, in particular of HMW adiponectin, which induces AMPK activation and decreases gluconeogenesis in the liver, thereby ameliorating insulin resistance. On the other hand, in a manner independent of adiponectin, rimonabant directly induces AMPK activation and decreases gluconeogenesis in the liver, possibly via the hepatic CB-1 receptor (35,43), which also contributes to ameliorating insulin resistance. In addition, rimonabant decreases food intake and increases energy expenditure, which are related to reduction of body weight. This body weight loss may be also associated with ameliorating insulin resistance via adiponectindependent and adiponectin-independent pathways (Fig. 6). Rimonabant is metabolized in the liver by cytochrome P-450 CYP3A4 and amidohydrolase and excreted into the bile (44,45). The oral bioavailability of rimonabant is low to moderate; this is due to the extensive first pass metabolism of the drug (European Medicines Agency). Therefore, in this study, the concentration in the liver of the orally administered rimonabant might be higher than that in other tissues, such as the muscle, because of the first pass effect of the liver. Although intraperitoneally administered rimonabant was reported in a previous study to significantly increase the glucose uptake in the soleus muscle of ob/ob mice (10), no improvement of the insulin resistance in the muscle was observed in our study. One of the reasons for this difference may be the lower concentration of rimonabant in the muscle due to the first pass effect of the liver.
In the four double-blind trials (RIO-Lipids (15), RIO-Europe (16), RIO-North America (17), and RIO-Diabetes (18)) the most frequent adverse events among individuals treated with rimonabant were nausea, dizziness, diarrhea, and insomnia, each occurring at a 1-9% greater frequency than that in the placebo group. In the RIO-Lipids, RIO-Europe, and RIO-North America, the drug had to be discontinued due to the development of psychiatric disorders (mainly depression) in 6 -7% of rimonabant-treated individuals, an absolute increase of 2-5% over the frequency in the placebo group (44). Substance dependence with rimonabant has not been reported. The absence of the appearance of clinical signs in toxicology studies with a recovery period indicates that rimonabant does not possess the potential to produce withdrawal syndrome (European Medicines Agency).
Many reports have shown the efficacy of cannabinoid agonists in chronic pain (46). In a rodent model of inflammatory pain, anandamide, one of the endogenous cannabinoids, suppressed the development and maintenance of thermal hyperalgesia (47). This analgesic effect was diminished by concurrent administration of the CB-1 antagonist, rimonabant, and anandamide. Although rimonabant alters the sensitivity to pain (47), it does not necessarily induce pain itself. On the contrary, rimonabant has recently been shown to prevent indomethacininduced intestinal injury by decreasing the levels of the proinflammatory cytokine, tumor necrosis factor ␣, in rodents (48), indicating its potential anti-inflammatory activity in acute and chronic diseases. In neurogenic inflammatory pain, including arthritis and neuropathy, many cytokines, especially tumor necrosis factor ␣, play a key role in the generation and maintenance of hyperalgesia (49). On the basis of these findings, Costa (50) indicated that the anti-tumor necrosis factor ␣ effect of rimonabant might contribute to its anti-inflammatory activity and consequently to the relief of pain. However, further investigation and accumulation of further evidence on the effect of rimonabant on pain are needed. At least, in the four clinical trials mentioned above, side effects associated with pain, such as hyperalgesia or hypoalgesia, were not reported. Furthermore, it has been suggested that although females might perceive pain differently from males (51,52), the anti-obesity effects of rimonabant appeared to be similar in males and females (European Medicines Agency).
In conclusion, this study demonstrated for the first time that rimonabant ameliorates insulin resistance via both adiponectindependent and adiponectin-independent pathways.
FIGURE6.Rimonabantamelioratesinsulinresistanceviabothadiponectindependent and adiponectin-independent pathways. There are two distinct pathways by which rimonabant ameliorates insulin resistance, one an adiponectin-dependent pathway and the other an adiponectin-independent pathway. Rimonabant increases the plasma levels of adiponectin, in particular of HMW adiponectin, which induces AMPK activation and decreases gluconeogenesis in the liver, thereby ameliorating insulin resistance. On the other hand, in a manner independent of adiponectin, rimonabant directly induces AMPK activation and decreases gluconeogenesis in the liver, possibly via hepatic CB-1 receptor, which also contributes to ameliorating insulin resistance. In addition, rimonabant decreases food intake and increases energy expenditure, which are related to reduction of body weight. This body weight loss may be also associated with ameliorating insulin resistance via adiponectin-dependent and adiponectin-independent pathways.