|Home | About | Journals | Submit | Contact Us | Français|
Obesity and its metabolic consequences are a major public health concern worldwide. Obesity is associated with overactivity of the endocannabinoid system, which is involved in the regulation of appetite, lipogenesis, and insulin resistance. Cannabinoid-1 receptor (CB1R) antagonists reduce body weight and improve cardiometabolic abnormalities in experimental and human obesity, but their therapeutic potential is limited by neuropsychiatric side effects. Here we have demonstrated that a CB1R neutral antagonist largely restricted to the periphery does not affect behavioral responses mediated by CB1R in the brains of mice with genetic or diet-induced obesity, but it does cause weight-independent improvements in glucose homeostasis, fatty liver, and plasma lipid profile. These effects were due to blockade of CB1R in peripheral tissues, including the liver, as verified through the use of CB1R-deficient mice with or without transgenic expression of CB1R in the liver. These results suggest that targeting peripheral CB1R has therapeutic potential for alleviating cardiometabolic risk in obese patients.
Endocannabinoids are endogenous lipid mediators that interact with the same G protein–coupled receptors — CB1R and CB2R — that recognize plant-derived cannabinoids, and they regulate a broad range of physiological functions. CB1Rs are expressed at very high levels in the brain but are also present at much lower yet functionally relevant concentrations in various peripheral tissues, whereas the expression of CB2Rs is largely limited to cells of the immune and hematopoietic systems. Activation of CB1R results in increased appetite, insulin resistance, and increased hepatic lipogenesis, which suggests the involvement of the endocannabinoid/CB1R system in obesity and its metabolic consequences (1). Indeed, obesity and its metabolic complications are characterized by an overactive endocannabinoid system (2–5), and chronic treatment with CB1R antagonists leads to weight loss and improved cardiometabolic risk profile in obese rodents (6, 7) and humans (8–11). However, concern over neuropsychiatric side effects, including anxiety, depression, and suicidal ideation (12), prevented approval of the first-in-class CB1R antagonist rimonabant in the United States and led to its withdrawal from the European market as well as the withdrawal of related compounds from preclinical development (13). Although the exact role of the endocannabinoid system in the control of mood and anxiety-like behaviors is not clear, CB1R in the prefrontal cortex, amygdala, and the mesolimbic dopaminergic reward pathway have been linked to the control of these behaviors (14). On the other hand, CB1Rs are also present in peripheral tissues including the liver (15–17), skeletal muscle (18, 19), endocrine pancreas (20, 21), and fat (16, 22, 23), where their activation contributes to obesity-related metabolic and hormonal abnormalities (24–26). This suggests that selective targeting of peripheral CB1R may result in an improved hormonal-metabolic profile in obesity without the untoward behavioral effects observed following treatment with brain-penetrant CB1R antagonists. To this end, we have developed a potent, selective, orally bioavailable CB1R antagonist with limited brain penetrance, and tested its behavioral and metabolic activity profile in two experimental models of obesity as well as in transgenic mice with CB1R expression limited to the liver. The results indicate that selective targeting of peripheral CB1R has value in the management of cardiometabolic risk in obesity.
First-generation CB1R antagonists, such as rimonabant, are highly lipid soluble and readily penetrate the blood-brain barrier. In order to reduce brain penetrance, we introduced several modifications into the structure of rimonabant. The analog AM6545 [5-(4-[4-cyanobut-1-ynyl]phenyl)-1-(2,4-dichlorophenyl)-4-methyl-N-(1,1-dioxo-thiomorpholino)-1H-pyrazole-3-carboxamide] (Figure (Figure1A)1A) is less lipid soluble than rimonabant (estimated partition coefficient [log P], 3.3 vs. 6.4 for rimonabant) but retains high affinity and selectivity for CB1R. In radioligand displacement assays, AM6545 has a KI of 3.3 nM for CB1R, which is similar to that of rimonabant, and greater than 100-fold CB1/CB2 selectivity (Figure (Figure1B).1B). Unlike rimonabant, AM6545 does not reduce GTPγS binding in mouse brain membranes and is therefore a neutral antagonist (Figure (Figure1B).1B). Importantly, AM6545 displays markedly reduced brain penetrance, as reflected by a brain/plasma concentration ratio of 0.03 following acute parenteral or oral administration and removal of intravascular fluids, as compared with a ratio of 0.8 for rimonabant (Figure (Figure1C).1C). Brain and plasma levels were also measured 12 hours after dose administration, following chronic (28 days) treatment. The brain/plasma ratio of AM6545 remained low (i.e., 0.07), and the brain concentration of rimonabant continued to be much higher (approximately 14-fold) than that of AM6545 (Figure (Figure1D). 1D).
In both cell-based and in vitro biochemical assays, AM6545 interacts at low micromolar concentrations with the ABC transporter P-glycoprotein (P-gp; Abcb1), which is localized on the luminal surface of capillary endothelial cells in the brain forming the blood-brain barrier (27) (Supplemental Figure 1; supplemental material available online with this article; doi: 10.1172/JCI42551DS1). This suggests that transporter-mediated extrusion of AM6545 from the brain may contribute to its low apparent brain penetrance. This was further verified in P-gp–deficient (Mdr1a/b–/–) mice and in mice deficient in both P-gp and Bcrp/Abcg2, along with their wild-type controls. As shown in Figure Figure1E,1E, brain penetrance of AM6545 increased dramatically in both knockout strains, resulting in brain/plasma ratios of 0.64 and 0.69, respectively. This indicates that the low brain penetrance of AM6545 in wild-type mice is due predominantly to active extrusion by P-gp. Furthermore, when tested in Caco-2 intestinal epithelial cell monolayers (28), AM6545 displays high permeability, with no evidence for active efflux (efflux ratio of 1.1). This suggests good oral bioavailability, also supported by similar plasma concentrations following i.p. and oral administration (Figure (Figure1C). 1C).
We next tested whether the reduced brain penetrance of AM6545 is associated with an attenuation of behavioral effects. Of the cannabinoid-induced behavioral tetrad (29), analgesia is known to have an important peripheral component (30). We therefore compared the effects of AM6545 and rimonabant in antagonizing cannabinoid-induced catalepsy, hypomotility, and hypothermia. Cannabinoid-induced catalepsy, quantified by the bar assay, is mediated exclusively by central CB1R (31). The marked increase in immobility induced in mice by the cannabinoid agonist HU-210 (30 μg/kg, i.p.) was completely blocked by rimonabant (10 mg/kg, i.p.) but was unaffected by a similar dose of AM6545 (Figure (Figure2A).2A). Similarly, rimonabant completely prevented the hypomotility induced by either i.p. or intracerebroventricular (i.c.v.) administration of the cannabinoid WIN55,212-2, whereas AM6545 at the same dose was without effect (Figure (Figure2B).2B). The hypothermic response to WIN55,212-2 was maximal following an i.c.v. dose of 30 μg or an i.p. dose of 10 mg/kg. The response to the centrally administered agonist peaked at 30 minutes and returned to baseline within 2 hours, and it was completely blocked by pretreatment with 10 mg/kg rimonabant i.p. but was unaffected by the same dose of AM6545 (Figure (Figure2C).2C). The hypothermia was more profound after peripheral as compared with central administration of WIN55,212-2 and lasted up to 8 hours. This effect was again completely blocked by rimonabant, 10 mg/kg, i.p., and was partially inhibited by AM6545, with the residual response being similar in magnitude and duration to the control response to centrally administered WIN55,212-2 (Figure (Figure2C).2C). These findings suggest that both central and peripheral sites contribute to the hypothermic response to WIN55,212-2 and that AM6545 does not affect hypothermia mediated via central CB1R.
Rimonabant, but not AM6545 (each at 10 mg/kg, i.p.), also induced a marked and long-lasting increase in ambulatory activity in wild-type but not CB1R–/– mice (Figure (Figure2D).2D). Interestingly, when tested in the P-gp–deficient Mdr1a/b–/– and Mdr1a/b–/–abcg2–/– mice, 10 mg/kg AM6545 caused a robust increase in ambulatory activity similar to rimonabant, reflecting its increased brain penetrance in these mice (Supplemental Figure 2). Similarly, rimonabant but not AM6545 (each at 10 mg/kg, i.p.) induced a robust anxiogenic response in the elevated plus maze (EPM) paradigm (Figure (Figure2E).2E). Thus, AM6545 is devoid of the centrally mediated behavioral effects that are predictive of the neuropsychiatric side effects of brain-penetrant CB1R antagonists. In contrast, 10 mg/kg AM6545 completely blocked the anandamide-induced inhibition of upper gastrointestinal motility, an effect mediated by prejunctional CB1R in peripheral neurons innervating the gut (32) (Supplemental Figure 3). Because the dose of anandamide used (10 mg/kg, i.p.) was reported to be maximally effective in this paradigm (33), CB1R occupancy by AM6545 can be assumed to be near maximal at the 10 mg/kg dose used in most of our experiments.
The metabolic profile of AM6545 and rimonabant was next examined in mice with diet-induced obesity (DIO). Male C57BL/6J mice fed a high-fat diet (HFD) for 14 weeks became obese and were then started on daily i.p. injections of vehicle, rimonabant, or AM6545 (both at 10 mg/kg/d) for an additional 28 days. Age- and sex-matched mice on standard chow served as controls. The overweight and increased adiposity of mice on HFD were significantly reduced by both antagonists (Figure (Figure3,3, A and B). HFD-induced fatty liver, as reflected in elevated hepatic triglyceride (TG) content, and hepatocellular damage, as indicated by increased plasma ALT levels, were completely reversed by either rimonabant or AM6545 (Figure (Figure3,3, C and D). Similar effects were observed after oral administration of AM6545 at 10 or 30 mg/kg/d for 14 days (Supplemental Figure 4), indicating good oral bioavailability.
The effect of AM6545 on substrate utilization was analyzed by indirect calorimetry. In mice kept on standard chow, a bolus i.p. dose of 10 mg/kg AM6545 at the start of the light period caused a marked and sustained reduction in respiratory quotient (RQ), resulting from an increase in fat oxidation and decrease in carbohydrate oxidation (CHO) (Figure (Figure3E),3E), which were similar to the effects of rimonabant (Figure (Figure3F).3F). Only transient effects were seen in CB1R–/– mice (Figure (Figure3E),3E), indicating that the sustained changes were due to CB1R blockade. Similar effects of AM6545 were seen in mice on HFD, although they were smaller than in mice on standard diet, possibly due to the diet-induced decrease in basal levels of RQ and increase in fat oxidation (Figure (Figure3E).3E). Nevertheless, chronic (28 days) treatment of mice on HFD resulted in increased total energy expenditure in the case of both antagonists (Supplemental Figure 5). Thus, the modest decrease in body weight by AM6545 was probably due to increased lipid oxidation.
This is further indicated by the decreased gene expression of the lipogenic enzymes fatty acid synthase (FAS) and stearoyl coenzyme-A desaturase 1 (Scd1) in both subcutaneous and visceral adipose tissue, whereas the expression of the lipolytic enzyme hormone-sensitive lipase (HSL) was increased in subcutaneous and decreased in visceral fat (Figure (Figure4,4, A and C), similar to the reported effects of rimonabant (16). Decreased lipogenic and increased lipolytic gene expression was also evident in the liver of DIO mice, where AM6545 treatment resulted in decreased Scd1, FAS, and acetyl coenzyme A carboxylase-1 (Acaca) and increased carnitine palmitoyl transferase–1 (CPT1) expression (Figure (Figure4E).4E). In agreement with the reduced lipogenic gene expression, AM6545 blocked cannabinoid-induced de novo lipogenesis in vitro in isolated mouse hepatocytes (Supplemental Figure 6). AM6545 treatment also reduced CB1R expression in both adipose tissue and liver (Figure (Figure4,4, A, C, and E), which is again similar to recent findings with rimonabant (16).
The greater efficacy of rimonabant over AM6545 in reducing body weight is probably related to its additional ability to reduce total caloric intake (Supplemental Figure 7A). A smaller and more transient reduction in food intake by AM6545 did not affect total caloric intake over the treatment period (Supplemental Figure 7B). Nevertheless, to test whether the transient reduction in food intake may contribute to metabolic changes induced by short-term CB1 antagonist treatment, the effects of treating DIO mice with 10 mg/kg/d AM6545 for 7 days were compared with the effects of pair feeding. As illustrated in Figure Figure5,5, short-term AM6545 treatment caused significant reductions in caloric intake, body weight, blood glucose and insulin levels, serum ALT, and hepatic TG content; normalized the impaired glucose tolerance; and improved insulin sensitivity relative to that in free-feeding, vehicle-treated mice. In the pair-fed group, food intake was held at the same reduced level as in AM6545-treated mice, yet all other parameters were similar to those in vehicle-treated controls. Thus, the transient reduction of caloric intake by AM6545 does not contribute to its metabolic effects.
In addition to reducing hepatic lipogenesis and increasing fatty acid oxidation, AM6545 treatment also resulted in a marked increase in the hepatic production and release of TG-rich VLDL, as measured in tyloxapol-treated DIO mice (Figure (Figure6).6). All of these effects could contribute to the reduction of hepatic steatosis.
A key component of the obesity-related metabolic syndrome is insulin resistance, which manifested in the simultaneous presence of hyperglycemia and hyperinsulinemia (Figure (Figure7,7, A and B). DIO mice also displayed hyperleptinemia (Figure (Figure7C),7C), an indication of leptin resistance, whereas the plasma level of adiponectin, an adipokine that promotes fatty acid oxidation, was reduced (Figure (Figure7D).7D). All of these changes were attenuated or completely reversed by CB1R blockade, rimonabant being somewhat more efficacious than AM6545. DIO mice also displayed reduced glucose tolerance (Figure (Figure7E)7E) and moderately decreased sensitivity to the hypoglycemic effect of insulin (Figure (Figure7F).7F). The reduced glucose tolerance, reflected by the increase in AUC from 327 ± 24 to 437 ± 14 mg/dl/h (P < 0.01) according to the i.p. glucose tolerance test, was reversed to control levels by chronic treatment with either rimonabant (AUC, 279 ± 35 mg/dl/h) or AM6545 (358 ± 14 mg/dl/h). Insulin resistance, as assessed from the insulin-induced change in the normalized blood glucose values, was increased by the HFD, and this increase was reversed by rimonabant but not by AM6545 treatment of DIO mice. When insulin resistance was assessed by the homeostasis model assessment–insulin resistance (HOMA-IR) method and the insulin sensitivity index, both antagonists significantly improved insulin sensitivity, but the effect of rimonabant was stronger (Supplemental Figure 8).
In order to test whether peripheral CB1R blockade affects metabolic parameters in more than one model of obesity, we used genetically leptin-deficient ob/ob mice that develop extreme obesity associated with diabetes, fatty liver, and dyslipidemias. Treatment end points were determined at 7 as well as at 28 days to gain insight into the time course of their development. Daily treatment of adult male ob/ob mice for 7 days with 10 mg/kg of either AM6545 or rimonabant had no significant effect on body weight and adiposity index (Figure (Figure8A),8A), yet both antagonists significantly improved glucose tolerance and insulin resistance (Figure (Figure8B)8B) and reduced the elevated levels of plasma glucose and insulin, with AM6545 being somewhat more effective than rimonabant (Figure (Figure8C).8C). AM6545, but not rimonabant, also markedly lowered liver TG content and reduced hepatocellular damage, as reflected by the lowering of plasma levels of ALT (Figure (Figure8D),8D), whereas after 28 days of treatment, both antagonists were effective in reducing hepatic TG content (Supplemental Figure 9C). AM6545, but not rimonabant, increased HDL cholesterol, and both antagonists reduced LDL cholesterol levels. Plasma TG markedly increased after 7 days of treatment with AM6545, probably as a result of the rapid and robust clearance of liver TGs (Figure (Figure8E).8E). Indeed, AM6545 treatment caused an increase in VLDL-TG production in ob/ob mice similar to that observed in DIO mice (Figure (Figure6). 6).
In ob/ob mice treated for 28 days, rimonabant, but not AM6545, caused a modest but significant reduction in body weight and adiposity, and both compounds improved dyslipidemia by reducing plasma TGs and LDL cholesterol levels and increasing plasma HDL cholesterol (Supplemental Figures 9 and 10).
The ability of AM6545 treatment to reduce body weight and adiposity in DIO mice but not in leptin-deficient ob/ob mice suggests that AM6545 treatment may sensitize adipose tissue to the lipolytic effects of endogenous leptin. Indeed, the AM6545-induced changes in lipogenic gene expression were completely absent in both subcutaneous and visceral adipose tissue of similarly treated ob/ob mice (Figure (Figure4,4, B and D). In contrast to adipose tissue, lipogenic gene expression in the liver was similarly reduced by AM6545 in DIO and ob/ob mice (Figure (Figure4,4, E and F). AM6545 also reversed the HFD-induced loss of the appetite-suppressing effect of leptin (Supplemental Figure 11).
Selective deletion of CB1R in hepatocytes has been shown to protect mice from HFD-induced glucose intolerance and insulin resistance and to attenuate the associated steatosis, which implicates endocannabinoids acting via hepatic CB1R in these processes (24). We generated mice transgenically expressing CB1R in hepatocytes on a global CB1R-knockout and C57BL/6J background (htgCB1R–/– mice) by injecting a DNA construct containing the mouse albumin promoter and the CB1R coding region into homozygous CB1R–/– oocytes from superovulated donors. CB1R–/– mice heterozygous for the transgene and their CB1R–/– littermates were used to test whether activation of hepatic CB1R is sufficient to induce diet-induced glucose intolerance/insulin resistance and hepatic steatosis and whether their blockade may contribute to the improved glucose tolerance following short-term treatment with a peripheral CB1R antagonist. htgCB1R–/– mice were viable and displayed no gross defects or obvious functional abnormalities. Similar to their global CB1R–/– littermates, they did not express CB1R in extrahepatic tissues, including the brain, but had both CB1R mRNA and protein present in hepatocytes at levels somewhat higher than those in wild-type C57BL/6J mice (Figure (Figure9A).9A). Exposure of htgCB1R–/– mice to HFD resulted in a much smaller, though significant, elevation of hepatic TG levels relative to those of wild-type mice on HFD, whereas no such change was observed in their CB1R–/– littermates (Figure (Figure9B).9B). In contrast, the robust HFD-induced glucose intolerance and reduced insulin sensitivity in htgCB1R–/– mice (Figure (Figure9C)9C) were similar to those reported earlier in wild-type mice, as opposed to CB1R–/– mice, which remain insulin sensitive (34). HFD also resulted in the activation of the ER stress response in the liver as indicated by increased phosphorylation of eukaryotic translation initiation factor 2α (eIF2α) (Figure (Figure9D),9D), a process that has been linked to insulin resistance (35). Furthermore, daily treatment of htgCB1R–/– mice on HFD for 7 days with 10 mg/kg AM6545, but not with vehicle, reversed all the above diet-induced changes (Figure (Figure9). 9).
Here we have demonstrated that treatment of obese mice with a peripherally restricted CB1R antagonist improves glycemic control and dyslipidemia and reverses hepatic steatosis, which strongly supports the contribution of peripheral CB1R activation to these hormonal/metabolic abnormalities. In the absence of any change in body weight in ob/ob mice treated with AM6545, the observed improvements in metabolic parameters were clearly weight independent. In mice with DIO, AM6545 did reduce body weight, but it was less efficacious than the brain-penetrant CB1R antagonist rimonabant. This is in agreement with the widely held notion that weight reduction by global CB1R blockade is the combined result of a centrally mediated decrease in caloric intake and a peripherally mediated increase in energy expenditure (11, 36). Indeed, rimonabant, but not AM6545, reduced total caloric intake during the 4-week treatment period, most likely through blockade of CB1R in the brain (37), and selective blockade of peripheral CB1R by AM6545 in DIO mice was associated with increased total energy expenditure due to increased fat burning (Figure (Figure3E3E and Supplemental Figure 5). These metabolic effects of AM6545 were fully evident as early as after 7 days of treatment and were not replicated by pair feeding (Figure (Figure5),5), which rules out the role of the transient reduction in food intake by AM6545 as their cause.
A likely target of AM6545 is CB1R in adipocytes, blockade of which results in increased adiponectin secretion (23). AM6545 treatment reversed the diet-induced suppression of plasma adiponectin (Figure (Figure7D),7D), which probably contributed to the observed increase in fat oxidation. The rise in adiponectin could also contribute to the alleviation of hepatic insulin resistance (38). Blocking adipocyte CB1R may also directly inhibit lipogenesis by decreasing the gene expression of the lipogenic enzymes Scd1 and FAS in adipose tissue of AM6545-treated DIO mice (Figure (Figure4,4, A and C). Interestingly, AM6545 increased HSL expression in subcutaneous fat but decreased it in visceral fat, similar to the reported effects of rimonabant in DIO mice (16). Although we found that the reduction in adipose tissue mass by AM6545 was similar in the two types of fat tissue (data not shown), the differential regulation of HSL expression may reflect qualitative differences in the effects of endocannabinoids on subcutaneous and visceral adipose tissue.
A possible alternative target of AM6545 in adipose tissue is presynaptic CB1R on sympathetic nerve terminals, stimulation of which will inhibit norepinephrine release (39). Adipocytes produce and secrete endocannabinoids (40) that could tonically activate such receptors. Rimonabant has been shown to increase norepinephrine turnover in peripheral tissues of wild-type mice but not mice that lack CB1R in forebrain and sympathetic neurons (41), which makes the above mechanism plausible. The role of such presynaptic CB1R is compatible with recent findings that hypothalamic leptin suppresses anandamide in white adipose tissue, and when this suppression of endocannabinoid tone is prevented by systemic CB1R activation or by sympathetic denervation of fat pads, leptin fails to suppress adipose tissue lipogenesis (42). Regardless of the specific cellular target of CB1R antagonists in adipose tissue, the above-described effects of AM6545 reinforce recent indirect evidence for a partially peripheral mechanism for the reduction of body weight and adiposity by brain-penetrant CB1R antagonists (11, 26, 43, 44).
The differential effectiveness of AM6545 in reducing body weight and adiposity in DIO mice but not in ob/ob mice may be related to leptin. DIO is known to trigger a primarily peripheral form of leptin resistance (45, 46), which is reflected in a compensatory increase in plasma leptin levels (6, 15). In turn, the observed reversal of hyperleptinemia by CB1R antagonists reflects improved leptin sensitivity, analogous to the enhanced leptin sensitivity of CB1R–/– mice (47). In agreement with the above reports, we found that the acute anorexigenic effect of leptin is reduced in mice on HFD, and this reduction was reversed following daily treatment with AM6545 for 9 days (Supplemental Figure 11). Leptin is known to promote fatty acid oxidation independently of its effects on food intake through activation of AMP kinase in muscle and adipose tissue (48). This parallels the ability of CB1R antagonists to increase the activity of the mitochondrial fatty acid oxidative pathway in liver (24) and increased mitochondrial biogenesis in white adipose tissue (49). Thus, sensitization to the metabolic actions of endogenous leptin may contribute to the increase in fat oxidation and secondary weight loss following AM6545 treatment in DIO mice but not in leptin-deficient ob/ob mice. Direct evidence for such a mechanism is provided by the present findings that, in contrast to its effects in DIO mice, AM6545 treatment failed to decrease lipogenic and increase lipolytic gene expression in adipose tissue of ob/ob mice (Figure (Figure4,4, B and D).
Apart from showing a small increase in hepatic TGs, htgCB1R–/– mice were largely resistant to HFD-induced steatosis (Figure (Figure9B).9B). Together with earlier findings that indicate partial protection of liver-specific CB1R–/– mice from HFD-induced steatosis (24), this points to a dominant role for extrahepatic CB1R in the development of diet-induced fatty liver. Adipose tissue is a likely candidate, as it may provide free fatty acids for hepatic TG synthesis (16). However, the essentially complete and rapid reversal of hepatic steatosis by AM6545, which indicates a dominant role for peripheral CB1R, may not be due to a simple reversal of the processes that led to its development. Hepatic CB1Rs are likely involved, as indicated by the AM6545-induced marked increase in TG-rich VLDL secretion, which paralleled the rapid and near complete elimination of excess hepatic TGs. Preferential targeting of hepatic CB1R was further facilitated by the accumulation of AM6545 in the liver (Figure (Figure1C).1C). In a recent study, treatment of mice with a novel inhibitor of monoglyceride lipase caused an acute increase in hepatic levels of the endocannabinoid 2-arachidonoylglycerol (2-AG) accompanied by a CB1R-mediated increase in plasma TGs. This was due to reduced TG clearance with no change in TG secretion by either CB1R agonist or antagonist (25), which is different from the present findings. It is possible, however, that the link between CB1R and TG secretion requires long-term activation of the hepatic endocannabinoid system, as induced by HFD. The latter, unlike the acute model, is associated with steatosis, upregulation of hepatic CB1R (15, 16), and elevated hepatic anandamide rather than 2-AG levels (15).
At the 10 mg/kg dose of AM6545 used here, CB1R occupancy is near maximal, given that the plasma concentration of the drug (approximately 3 μM, Figure Figure1C)1C) exceeded its KI by 3 orders of magnitude. A further indication that blockade of peripheral CB1R is near complete at this dose is that AM6545 completely blocked inhibition of gut motility by a dose of anandamide (10 mg/kg, Supplemental Figure 3) previously shown to be maximally effective in this paradigm (33). Furthermore, no additional reduction of hepatic TGs and tissue damage markers was observed when the dose of AM6545 was increased from 10 to 30 mg/kg (Supplemental Figure 4). The observed absence of behavioral effects is therefore noteworthy, as it suggests that the therapeutic efficacy of AM6545 could be maximized without the risk of neuropsychiatric side effects. In recent clinical trials with brain-penetrant CB1R antagonists, concern about neuropsychiatric side effects limited the daily dose used to levels that resulted in only partial (approximately 30%) occupancy of CB1R (8–11, 50), which may have limited their therapeutic efficacy.
Both rimonabant and AM6545 were effective in attenuating obesity-related glucose intolerance and insulin resistance and normalizing the elevated blood glucose and insulin levels. Although in DIO mice, rimonabant was more efficacious than AM6545 in this regard, in ob/ob mice AM6545 was as effective as rimonabant, suggesting that the improved glycemic control is primarily due to blockade of peripheral CB1R. A similar conclusion is supported by the finding that insulin sensitivity is improved in DIO rats by chronic systemic, but not by i.c.v., administration of rimonabant, even though the two types of treatment had similar effects on food intake (26). CB1R on hepatocytes is one of the targets of AM6545. Hepatic CB1Rs play an important role in insulin resistance, as mice with either global or hepatocyte-specific knockout of CB1R remain insulin sensitive on HFD (24). The present findings using a rescue model in which CB1Rs are expressed only in hepatocytes indicate that hepatic CB1Rs are not only necessary, but sufficient to account for the diet-induced glucose intolerance and insulin resistance. When kept on HFD, these transgenic mice develop glucose intolerance and moderate insulin resistance, whereas their global CB1R-knockout littermates are resistant to these diet-induced changes. Furthermore, AM6545 treatment normalizes glycemic control in the transgenic mice, which directly implicates CB1R on hepatocytes as a peripheral target for AM6545. Indeed, our recent experiments using euglycemic-hyperinsulinemic clamp indicate that hepatic CB1Rs play a dominant role in HFD-induced hepatic insulin resistance by increasing hepatic glucose production and inhibiting the antigluconeogenic effect of insulin (J. Liu et al., unpublished observations). The possible additional contribution of CB1R in skeletal muscle, adipose tissue, and pancreatic β-cells to the effects of AM6545 on glycemic control remains to be explored.
In summary, AM6545 is a novel, peripherally restricted CB1R antagonist with nanomolar affinity, selectivity for CB1R, and therapeutically relevant efficacy in reducing weight and improving hormonal/metabolic abnormalities in animal models of obesity. The previously published, peripherally restricted compound URB447 has approximately 100 times lower CB1R affinity (IC50, 313 nM) and lacks CB1/CB2 selectivity (51), and another recently published compound with reduced brain penetrance has similarly low CB1R affinity (159 nM) (52). An article describing another peripherally restricted CB1R antagonist with nanomolar potency and a high degree of selectivity for CB1 over CB2 receptors appeared after the present study had been submitted for publication. This antagonist was found to reduce body weight in obese rodents and to inhibit cannabinoid-induced hypothermia only at a high dose, but its metabolic and behavioral profile remains to be established (53).
The absence of neurobehavioral effects of AM6545 similar to those observed with rimonabant suggests that AM6545 will have fewer neuropsychiatric side effects in humans. As a neutral CB1R antagonist, AM6545 is also expected to cause less nausea than inverse CB1R agonists do (54). Therefore, further testing and development of compounds of this class may be warranted for the pharmacotherapy of the metabolic syndrome.
5-(4-[4-Cyanobut-1-ynyl]phenyl)-1-(2,4-dichlorophenyl)-4-methyl-N-(1,1-dioxo-thio-morpholino)-1H-pyrazole-3-carboxamide (AM6545) was synthesized from 1-(2,4-dichlorophenyl)-4-methyl-5-(4-iodophenyl)-N-(1,1-dioxothiomorpholino)-1H-pyrazole-3-carboxamide and pent-4-ynenitrile based on reported procedures (55, 56). 1H NMR (Varian Unity-INOVA spectrometer, 500 MHz, CDCl3): 8.08 (s, 1H), 7.43 (s, 1H), 7.37 (d, J = 7.81 Hz, 2H), 7.28–7.34 (m, 1H), 7.24 (s, 1H), 7.06 (d, J = 7.81 Hz, 2H), 3.56 (d, J = 4.88 Hz, 4H), 3.26 (br. s., 4H), 2.72–2.87 (m, 2H), 2.57–2.71 (m, 2H), 2.37 (s, 3H); 13C NMR (125 MHz, CDCl3): 161.21, 144.23, 143.94, 136.69, 136.25, 133.36, 132.44, 130.97, 130.90, 129.92, 128.74, 128.49, 123.90, 118.92, 118.69, 87.43, 82.95, 53.50, 51.62, 18.12, 17.24, 9.85; HRMS-ESI (Q-TOF, Mass Spectrometry Laboratory, School of Chemical Sciences, University of Illinois, Urbana-Champaign, Illinois, USA) calculated for C26H23Cl2N5O3S (M+ +H): 556.0977: found: 556.0980. Elemental analysis (Robertson Microlit Laboratories) calculated for C26H23Cl2N5O3S: C% 56.12, H% 4.17, N% 12.59, S% 5.76; found: C% 56.2, H% 4.04, N% 12.37, S% 5.59.
Animal protocols were approved by the Institutional Animal Care and Use Committee of the NIAAA, NIH. Male 8- to 10-week old C57BL/6J mice and adult male genetically obese ob/ob mice with their lean controls were obtained from The Jackson Laboratory. Mice were maintained under a 12-hour light/12-hour dark cycle and fed ad libitum. To generate DIO, C57BL/6J mice were fed either HFD (TD97070, Harlan Teklad; 60% of calories from fat, 18.8% from protein, and 21.3% from carbohydrates) or a standard laboratory diet (STD, NIH-31 rodent diet) for 14 or 20 weeks as indicated.
CB1R–/– mice and their wild-type littermates were generated and backcrossed to a C57BL/6J background as described previously (57). For generation of liver CB1R transgenic mice on a global CB1R–/– background (htgCB1R–/– mice), a mouse cDNA containing the entire CB1R reading frame was cloned into the XhoI and NotI sites of pGEMAlb-SVpA. A fragment containing the CB1R coding region, the albumin enhancer, albumin promoter, and SV40 polyadenylation site was then excised from the construct at ApaI and MluI sites for microinjection into fertilized oocytes at the single-cell stage, derived from mating CB1R–/– parents. Transgenic offspring were identified by double-band genotyping using tail DNA and were mated to CB1R–/– mice to generate mice that were heterozygous for the transgene, with their CB1R–/– littermates serving as controls. Mice deficient in Mdr1a/b or in Mdr1a/b plus Bcrp/Abcg2 and their wild-type controls were from Taconic.
HFD-fed obese mice received vehicle, AM6545, or rimonabant daily for 28 days, either by i.p. injection of 10 mg/kg or gavage of 10 or 30 mg/kg/d, as indicated. Age-matched control mice on STD received vehicle daily. Body weight and food intake were monitored daily. Glucose tolerance and insulin resistance were determined as described below. Other mice were sacrificed by cervical dislocation; the brain, liver, and combined fat pads were removed, weighed, and snap-frozen; and trunk blood was collected for determining endocrine and biochemical parameters. Adiposity index was defined as the combined weight of the epididymal, retroperitoneal, and inguinal fat pads, expressed as a percentage of total body weight.
htgCB1R–/– mice and their CB1R–/– littermates were kept on HFD for 20 weeks, with matched controls kept on STD. Mice were then treated daily for 7 days with vehicle or AM6545, 10 mg/kg, i.p. Twelve hours after the last dose and following an overnight fast, the mice were tested for glucose tolerance and insulin resistance.
Adult male genetically obese ob/ob mice were treated with AM6545 or rimonabant for 7 or 28 days, and hormonal and metabolic parameters were determined as in mice with DIO.
For acute treatment, mice received a single dose (10 mg/kg, i.p. or po) of AM6545 or rimonabant and were sacrificed 1 hour later. Blood was collected, and the mice were perfused with phosphate buffered saline for 1 minute to remove drug from the intravascular space before removal of the brain and liver. For chronic treatment, mice received daily i.p. injections of 10 mg/kg of either antagonist for 28 days and were sacrificed 12 hours following the last dose. Drug levels in tissue homogenates and plasma were determined at Apredica Inc. using LC-MS/MS.
AM6545 binding to CB1R was assessed in competition displacement assays using [3H]CP-55,940 as the radioligand and crude membranes from mouse brain for CB1R or HEK293 cells stably transfected with the mouse CB2R, as described previously (58). KI values were determined from 3 independent experiments performed in triplicate.
Mouse brains were dissected and P2 membranes prepared (59) and resuspended at approximately 6 μg protein/μl in 1 ml assay buffer (50 mM Tris HCl, 9 mM MgCl2, 0.2 mM EGTA, 150 mM NaCl; pH 7.4). Ligand-stimulated [35S]GTPγS binding was assayed as described previously (59). Briefly, membranes (10 μg protein) were incubated in assay buffer containing 100 μM GDP, 0.05 nM [35S]GTPγS, test compounds (HU-210, AM6545, and rimonabant) at 1 nM–1 μM, and 1.4 mg/ml fatty acid–free BSA in siliconized glass tubes. Bound ligand was separated from free ligand by vacuum filtration. Nonspecific binding was determined using 10 μM GTPS. Basal binding was assayed in the absence of the ligand and in the presence of GDP.
Catalepsy was assayed using the bar test (29). Briefly, mice were removed from their home cage, and their forepaws were placed on a horizontal bar, 0.5 cm in diameter, positioned 4 cm above the bench surface. Vehicle-treated mice routinely let go of the bar within 2 seconds. Cataleptic behavior was defined as the time the animals remained motionless holding on to the bar, with an arbitrary cutoff of 30 seconds. The test was performed 15 minutes after drug administration.
Locomotor activity was quantified by the number of disruptions of infrared beams in 2 dimensions in an activity chamber.
Mice were implanted with an intraabdominal thermoprobe, and core body temperature was monitored telemetrically. One hour after the i.p. administration of 10 mg/kg rimonabant or AM6545, the mice received a maximally effective i.p. or i.c.v. dose of the cannabinoid agonist WIN-55,212-2, and temperature was monitored for 8 hours.
Anxiety-related behaviors were assessed using the EPM test as described previously (60), 1 hour after drug or vehicle administration. Time spent in the open or closed arms and number of entries was measured by using the automated HindSight software system.
The amount of food eaten and body weight were recorded daily. For measuring the effect of leptin on food intake, mice on HFD (22 weeks) were treated for 9 days with AM6545 (10 mg/kg, i.p.) or vehicle. A third group of age-matched controls on STD were treated with vehicle for 9 days. After the final dosing, mice were fasted for 24 hours. Half an hour before the beginning of the dark phase, mice received leptin (10 mg/kg, i.p.) or vehicle, and food intake was measured for the indicated period of time.
Mice on HFD for 22 weeks were treated once daily with vehicle or with 10 mg/kg AM6545 i.p. for 1 week. In a third, pair-fed group, animals were fed the same amount of HFD as consumed by the AM6545-treated animals over the preceding 24 hours. An i.p. glucose tolerance test (ipGTT) was performed on mice fasted overnight following the last treatment day, with an i.p. insulin tolerance test (ipITT) done on the next day in the fed state. One day later, blood was collected from infraorbital veins for serum glucose, insulin, and ALT measurements. Mice were sacrificed, and the liver was removed for TG measurements.
Blood was collected at the time the mice were sacrificed. Serum ALT, AST, and TGs were quantified by using IDEXX Vet Test Analyzer 8008 (IDEXX Laboratories). Blood glucose was determined using the Elite glucometer (Bayer). Serum insulin was determined using the Ultra Sensitive Mouse Insulin ELISA kit (Crystal Chem Inc.). Serum leptin and adiponectin were determined by ELISA (ALPCO Diagnostics). HDL and LDL cholesterol were quantified colorimetrically using the EnzyChrom kit (BioAssay Systems).
Liver tissue was extracted as described previously (61) and its TG content determined using the EnzyChrom Triglyceride Assay kit (BioAssay Systems).
Mice were fed HFD for 22 weeks, then treated for 7 days with AM6545 (10 mg/kg, i.p.) or vehicle. Mice were fasted overnight before measurement of VLDL-TG secretion into the systemic circulation as described previously (62). Briefly, Triton WR-1339 (100 mg/ml in 0.9% NaCl; Sigma-Aldrich) was injected i.p. at 500 mg/kg, and blood samples were collected at 0, 1, 2, 3 hours after injection from the infraorbital veins. The rate of a linear increase in VLDL-TG was adjusted to plasma volume as a measure of the total TG secretion rate (TGSR). The plasma volume (PV) was estimated from the linear relationship between body weight and plasma volume using the following the formula: PV = body weight (g) × 0.07. TGSR was calculated from the slope of the TG concentration versus time curve and expressed as mg/kg/h.
Mice fasted overnight were injected with glucose (2 g/kg, i.p.), followed by tail blood collection at 0, 15, 30, 45, 60, 90, and 120 minutes. Blood glucose levels were determined using the Elite glucometer (Bayer). After blood glucose returned to baseline, the animals received an injection of insulin (0.75 U/kg, i.p.; Eli Lilly), and blood glucose levels were determined at the same intervals as above, except for the pair-feeding experiments, for which insulin resistance was assessed on the following day. AUCglucose and AUCinsulin between 0 and 120 minutes were determined using a trapezoidal model (63).
O2 consumption and CO2 production were monitored by indirect calorimetry (Oxymax, Columbus Instruments) with 1 mouse per chamber. Each chamber was equipped with a 2-dimensional (xy) infrared beam system (OPTO-M3) to record locomotor activity. Conditioned fresh air at 21 ± 0.5°C and 55% ± 5% relative humidity was pumped into the chambers at 0.6 l/min. The animals were acclimated to the chambers for 1 day, and VO2 and VCO2 measurements taken every 30 minutes were collected and recorded on a computer over the next 24 hours. During the 12-hour dark/12-hour light phases, mice had free access to food and water. AM6545, rimonabant (each at 10 mg/kg, i.p.), or vehicle were administered at the beginning of the light phase, and the mice were monitored for an additional 12 hours. RQ was calculated as CO2 (VCO2) production/O2 (VO2) consumption, with the values of 1 or 0.7 indicating 100% CHO or 100% fat oxidation (FO), respectively. Total energy expenditure (TEE) was calculated as VO2 • (3.815 + 1.232 • RQ), normalized to (body mass)0.75, and expressed as kcal/h/kg0.75. FO and CHO were calculated (64) as: FO = 1.69 • VO2 – 1.69 • VCO2 and CHO = 4.57 • VCO2 – 3.23 • VO2 and expressed as g/h/kg0.75.
The relative tissue content of CB1R protein was assessed by Western blotting (65), using an antibody against a peptide corresponding to aa 1–14 of the extracellular N-terminus of the CB1R (Cayman Chemical), a region displaying 100% sequence homology between the rodent an human receptor.
CB1R mRNA expression was assessed by RT-PCR and normalized with β-actin. Total RNA was isolated from mouse brain and liver tissue using TRIzol and reverse transcribed using the SuperScript First-Strand Synthesis System (Invitrogen). The resulting single-stranded cDNA (5 μl) was denatured at 95°C for 3 minutes and subjected to 30 cycles of amplification, each consisting of 15 seconds at 94°C, 30 seconds at 50°C, and 1 minute at 72°C, with a 7-minute final extension at 72°C. The following forward and reverse primers were used: CB1 (U22948), 5′-GTACCATCACCACAGACCTCCTC-3′ and 5′-GGATTCAGAATCATGAAGCACTCCA-3′ (300-bp amplicon); and mouse β-actin 5′-AAGGTGACAGCATTGCTTCT-3′ and 5′-GGCTGCCTCAACACCTCA-3′ (189-bp amplicon). PCR products were separated by electrophoresis on a 1% agarose gel. RNA without reverse transcription did not yield any amplicon, indicating the absence of genomic DNA contamination.
Total mRNA from liver, inguinal, and epididymal adipose tissue was extracted using the RNeasy kit (QIAGEN) followed by DNase I treatment (Invitrogen) and reverse-transcribed using the Iscript cDNA kit (Bio-Rad). Real-time PCR was performed using a Bio-Rad iCycler iQ and the QuantiTect Primer Assays (QT00169057, QT00149240, QT00291151, QT01748831, QT01554441, and QT00106820) against mouse HSL, FAS, Scd1, CB1R, Acaca, and CPT1, respectively. The normalization was performed with the QuantiTect Primer Assay against mouse Gapdh (QT01658692).
Rimonabant and CP-55,940 were obtained from the National Institute of Drug Abuse Drug Supply Program. HU-210 and WIN55,212-2 were purchased from Tocris. All compounds were dissolved in DMSO using gentle heating before being diluted with Tween 80 and saline (4% DMSO; 1% Tween 80; 95% saline) for i.p. injections or oral administration by gavage.
Values are expressed as mean ± SEM. Unpaired 2-tailed Student’s t test was used to determine differences between vehicle- and compound-treated groups. Time-dependent variables were analyzed, and results in multiple groups were compared by ANOVA followed by Bonferroni test. Nonlinear regression analysis was used to determine the KI values of ligands in binding assays (GraphPad Prism version 5 for Windows). P values less than 0.05 were considered significant.
This work was supported by the NIH (intramural funds from the NIAAA to G. Kunos and from the NCI to S.V. Ambudkar, and 2R01 DA007215 to A. Makriyannis). We thank Christoph Buettner for helpful discussions.
Conflict of interest: The authors have declared that no conflict of interest exists.
Citation for this article: J Clin Invest. 2010;120(8):2953–2966. doi:10.1172/JCI42551.
See the related Commentary beginning on page 2646.