Central Endocannabinoid Signaling Regulates Hepatic Glucose Production and Systemic Lipolysis

Animal protocols were approved by the Mount Sinai School of Medicine Institutional Animal Care and Use Committee. Male Sprague-Dawley (SD) rats (8 weeks old) (Charles River Breeding Laboratories, Wilmington, MA) were fed a standard rodent diet (Rodent Diet 5001; LabDiet, St. Louis, MO) and housed in a temperature- and light-controlled (21°C, 12-h light-dark cycle) facility. Rats were stereotaxically fitted with 22-gauge intracerebroventricular guide cannulae (PlasticsOne, Roanoke, VA) targeting the third ventricle, 2.5 mm posterior from bregma, on the midline 9 mm below the cortical surface, 2 weeks before the clamp study. One week later carotid arterial and jugular venous catheters were implanted for blood sampling and infusion, respectively. Rats recovered for at least 4 days and returned to within 10% of their presurgical body weight before being studied. After recovering from surgery, rats participating in the rimonabant study were given highly palatable 10% lard diet (HFD) (Modified Laboratory w/ 10% Lard 57IR, LabDiet) for 3 days before the clamp while food intake was monitored.

The rat clamp studies have been described previously (10) and are depicted schematically in Fig. 1A. In brief, intracerebroventricular infusion cannulae were inserted and a 5 μL/h intracerebroventricular infusion was initiated (t = −120 min) and maintained throughout the study. Treatments dissolved in artificial cerebrospinal fluid (aCSF) (Harvard Apparatus, Holliston, MA) contained 0.25 μg/μL WIN55,212-2 (WIN) (Sigma-Aldrich, St. Louis, MO; soluble with 5% DMSO, 0.5% Tween-80), 36.5 ng/h arachidonoyl-2'-chloroethylamide (ACEA) (with 0.036% DMSO), 0.2 μg/μL rimonabant (with 0.35% DMSO, 0.02% Tween-80) (both Cayman Chemical, Ann Arbor, MI), or the appropriate vehicle composed of aCSF, DMSO, and Tween 80 as needed for treatment solubility.

At t = 0 min, we started a tracer equilibration period consisting of primed-continuous infusions of [³H-3]glucose (radiochemical concentration >97%; PerkinElmer, Waltham, MA) and [²H-5]glycerol (98 atom percent excess from Isotec [Miamisburg, OH]) with boluses of 20 μCi and 13.3 μmol followed by continuous infusions of 0.5 μCi/min and 0.334 μmol/min, respectively. At t = 90 min, blood samples were taken every 10 min to determine basal hGP and R_a glycerol. At t = 120 min, an insulin clamp began with primed-continuous infusions of human insulin (Humulin R, Lilly; Indianapolis, IN) and somatostatin (Bachem Biosciences; Torrance, CA) with boluses of 53.2 mU · kg⁻¹ and 53.2 μg · kg⁻¹ followed by infusions at 3 mU · kg⁻¹ · min⁻¹ and 3 μg · kg⁻¹ · min⁻¹, respectively, while continuing the tracer infusions at the aforementioned rates. Euglycemia was maintained by measuring blood glucose and infusing 25% glucose as necessary. Blood samples were obtained every 10 min for the last hour of the experiment to calculate hGP, R_d of glucose, and R_a glycerol under clamped conditions.

At the end of the clamp, rats were anesthetized and killed and tissue samples were freeze-clamped in liquid nitrogen and stored at −80°C. Blood was collected in EDTA tubes. Blood glucose was measured by the GM9-D glucose analyzer (Analox Instruments, London, U.K.). Plasma insulin was analyzed by a rat insulin ELISA from Mercodia (Uppsala, Sweden). Plasma nonesterified fatty acid (NEFA) and trigylcerides were measured by kits from Wako Diagnostics (Richmond, VA) and Sigma-Aldrich.

Plasma samples were deproteinated using barium hydroxide and zinc sulfate, centrifuged, and the supernatant was dried overnight to eliminate tritiated water. Glucose was redissolved in water and counted using Ultima Gold in a MicroBeta TriLux (both PerkinElmer) liquid scintillation counter. Under preclamp steady-state conditions, hGP equals the glucose turnover rate, which was determined from the ratio of the [³H-3]glucose tracer infusion rate and the specific activity of plasma glucose. During the clamp, hGP was calculated by subtracting the exogenous glucose infusion rate (GIR) from the glucose turnover rate, which in a steady state equals the R_d.

R_a glycerol (in μmol · kg⁻¹ · min⁻¹) was calculated by the equation:

where ENR_inf is the fractional isotopic enrichment of the infused glycerol in atom percent excess and ENR_pl is the enrichment in the plasma sample. R is the rate of isotope infusion in μmol · kg⁻¹ · min⁻¹ (36). The ²H-labeling of plasma glycerol was determined as follows: 20 μl of plasma was deproteinated with 200 μl methanol by centrifugation. The fluid fraction was dried and reacted with 50 μl bis(trimethylsilyl)trifluoroacetamide plus 10% trimethylchlorosilane for 20 min at 75°C. Isotope enrichment was determined by gas chromatography–mass spectrometry. Ions of 205–208 mass-to-charge ratios were monitored.

Western blots were performed as in Ref. 10. In brief, proteins were extracted from tissue, separated on 4–12% NuPAGE gels (Invitrogen, Carlsbad, CA), blotted onto Immobilon-FL PVDF (Millipore, Billerica, MA) membranes, blocked, and incubated with primary antibodies. Antibodies used were: phospho-Akt473, GSK3-α/β, phospho-hormone–sensitive lipase (HSL) (Ser563 and Ser660), adipose triglyceride lipase (ATGL), PKA Substrate (all Cell Signaling Technology, Beverly, MA), β-actin, total HSL, and GAPDH (all Abcam, Cambridge, MA). Perilipin and CGI-58 were gifts from Drs. Andrew Greenberg (37) and Dawn Brasaemle (38), respectively. Phosphoperilipin was assessed by the PKA Substrate motif antibody. Blots were washed, incubated with Dylight conjugated secondary antibodies (Thermo Fisher Scientific, Waltham, MA) against rabbit and mouse IgG, and then washed before being scanned with the LI-COR Odyssey (LI-COR, Lincoln, NE) and quantified with Odyssey 3.0 software on the basis of direct fluorescence measurement.

Values are presented as means ± SE. Comparisons between groups were made using unpaired, two-tailed Student t tests. Differences were considered statistically significant at P < 0.05.

To assess the effects of central CB₁ activation on systemic insulin action, the authors performed hyperinsulinemic (3 mU ⋅ kg⁻¹ ⋅ min⁻¹), euglycemic clamp studies in SD rats while infusing the CB₁ agonist WIN (protocol in Fig. 1A). This protocol consisted of a 4-h baseline period followed by a 2-h hyperinsulinemic clamp. Glucose levels were controlled, and insulin concentrations were equal between groups before and during the clamp (Fig. 1B and Supplementary Fig. 1A). We used tracer dilution methodology to assess glucose and lipid fluxes simultaneously through the utilization of [³H-3]glucose and [²H-5]glycerol, respectively.

During baseline, the intracerebroventricular infusion of WIN tended to elevate hGP compared with vehicle, although this did not reach statistical significance. During the clamp, whereas systemic hyperinsulinemia lowered hGP in the vehicle infused rats, intracerebroventricular administration of WIN significantly blunted this effect (Fig. 1C). Although vehicle-infused rats suppressed hGP by 85% during the clamp, rats infused with WIN only suppressed hGP by 52% (Fig. 1D), indicating impaired hepatic insulin action. Of note, the rate of glucose utilization (R_d) in peripheral tissues, such as muscle and fat, tended to be higher in the WIN-infused animals (Fig. 1E), which could account for the fact that we did not find differences in the GIR (Supplementary Fig. 1B). These data demonstrate that the activation of the EC system in the brain is sufficient to induce hepatic insulin resistance.

Because WIN can activate both CB₁ and CB₂ (39), we wanted to test whether it is the activation of CB₁ rather than CB₂ or a drug-specific effect of WIN that impairs hepatic insulin action. Therefore, we infused the highly specific CB₁ agonist ACEA intracerebroventricularly while performing hyperinsulinemic, euglycemic clamp studies (Fig. 1A). Blood glucose was controlled throughout the clamp, and hyperinsulinemia was achieved in all rats (Fig. 1B and Supplementary Fig. 1C). As with intracerebroventricular WIN, intracerebroventricular ACEA infusion tended to elevate hGP at baseline and resulted in a significantly higher hGP during the hyperinsulinemic clamp (Fig. 1F). Therefore, intracerebroventricular ACEA infusion significantly impaired the insulin-induced suppression of hGP (66% with vehicle vs. 52% with ACEA) (Fig. 1G). It is noteworthy that the R_d in the intracerebroventricular ACEA-infused group tended to be higher, as seen with intracerebroventricular WIN infusion (Fig. 1H), which could explain why the GIR was not different despite the markedly blunted suppression of hGP (Supplementary Fig. 1D). These data confirm our above finding that central EC system activation induces hepatic insulin resistance, further highlighting the importance of brain CB₁ in regulating hepatic insulin action.

To determine whether the impaired hepatic insulin action seen with intracerebroventricular WIN and ACEA were because of deficits in hepatic insulin signaling, we performed Western blot analyses of livers harvested at the end of the hyperinsulinemic clamps (Fig. 2A and B). Phosphorylation and expression of AKT and glycogen synthase kinase 3 (GSK3), major downstream targets of insulin signaling, were unchanged with either WIN or ACEA infusions. These findings suggest that central CB₁ activation impairs hGP independent of hepatic insulin signaling, consistent with our observations that both central insulin and leptin (40) acutely suppress hGP without altering hepatic insulin signaling.

Because AMP-activated protein kinase (AMPK) is believed to play an important role in regulating hGP (41), we also assessed AMPK phosphorylation. We found that ACEA infusion moderately decreased AMPK phosphorylation (Fig. 2B), and although we were not able to detect this difference with intracerebroventricular WIN infusion, it may in part contribute to the hepatic insulin resistance induced by central EC system activation.

Increased lipolytic flux from WAT to the liver could provide a mechanism for the observed hepatic insulin resistance since glycerol released from WAT can act as a substrate for gluconeogenesis in the liver (42). Because we have demonstrated recently that brain insulin regulates systemic lipolysis (10) and studies have shown that lipolytic flux can regulate hGP (5,6), we investigated whether central EC system activation might induce peripheral lipolysis, consistent with impaired adipose tissue insulin action. To this end, we assessed R_a of glycerol during hyperinsulinemic, euglycemic clamp studies by infusing [²H-5]glycerol. The euglycemic clamp protocol used here enables us to study the effects of central EC system activation on lipolytic flux while controlling circulating insulin and glucose, two major regulators of lipolysis in adipocytes.

During baseline, intracerebroventricular WIN-infused rats tended to have elevated R_a glycerol (Fig. 3A), although this did not reach statistical significance. However, during the clamp, hyperinsulinemia suppressed lipolysis in vehicle-infused rats, but not in intracerebroventricular WIN-infused rats (Fig. 3A). Indeed, intracerebroventricular vehicle-infused rats suppressed R_a glycerol by 47% during moderate hyperinsulinemia, whereas intracerebroventricular WIN-infused rats only suppressed R_a glycerol by 7% (Fig. 3B). These data show that, in addition to the blunted suppression of hGP with systemic hyperinsulinemia, intracerebroventricular WIN infusion impairs the ability of insulin to suppress lipolysis. The inability of the WIN-infused animals to restrain lipolysis could contribute to the impaired suppression of hGP, because it has been shown that lipolytic flux can regulate hGP (5,6). This is highlighted by an inverse correlation between R_a glycerol and hGP suppression (Fig. 3C), and although correlation does not prove causation, it raises the possibility that the impaired suppression of lipolysis is involved in the mechanism regulating hGP.

To investigate the molecular mechanisms responsible for the elevated lipolytic flux in WIN-infused animals, we performed Western blot analyses on WAT from clamped rats. We assessed the activation states and protein expression of several key lipolytic enzymes, including comparative gene identification-58 (CGI-58), HSL, ATGL, and perilipin (43). Perilipin and CGI-58 are lipid droplet-associated proteins. Upon phosphorylation, perilipin facilitates the access of lipolytic enzymes, such as ATGL and HSL, to the lipid droplet and is believed to release CGI-58, which in turn activates ATGL. ATGL hydrolyzes triacylglycerols to diacylglycerols, which are then further hydrolyzed to monoacylglycerols by HSL (44). We found that intracerebroventricular WIN infusion increased perilipin phosphorylation and CGI-58 expression (Fig. 3D), suggesting that ATGL is activated. We also saw consistent trends for intracerebroventricular WIN infusion to increase activating serine phosphorylation sites on HSL. Overall, this increased activation of several key lipolytic enzymes could contribute to the elevated lipolytic flux seen with intracerebroventricular WIN infusion.

We next wanted to understand the role of brain CB₁ in a pathophysiological state of insulin resistance. We induced insulin resistance by exposing SD rats to HFD for 3 days and allowing them to overeat (Supplementary Fig. 2A). This model is characterized by hepatic insulin resistance, whereas peripheral glucose utilization is not yet reduced (45,46). Thus this is a model of early insulin resistance lacking secondary confounding factors, including impaired glucose utilization in muscle and WAT and a chronic proinflammatory state, that are associated with long-standing insulin resistance and diabetes. One of the key defects in this model is impaired hypothalamic insulin action, which is a failure of MBH-infused insulin to suppress hGP (13).

To test whether increased EC tone in the CNS is responsible for the impaired insulin action, we blocked EC signaling in the brain by infusing rimonabant intracerebroventricularly while elevating circulating insulin levels and controlling glucose with a hyperinsulinemic, euglycemic clamp (scheme in Fig. 4A, insulin and glucose levels in Supplementary Fig. 2B and C). During baseline, hGP was similar between groups, whereas during the clamp the rats infused with rimonabant had significantly lower hGP than HFD vehicle rats (Fig. 4B). Thus although HFD vehicle-treated rats suppressed hGP by only 53%, rats infused with rimonabant suppressed hGP by 79% (Fig. 4C). It is noteworthy that intracerebroventricular rimonabant infusion tended to decrease the R_d (Fig. 4D) and, although not statistically significant, it complements the elevation in R_d seen with CB₁ activation by WIN and ACEA infusions. Intracerebroventricular rimonabant mildly increased the GIR (Supplementary Fig. 2D), although this did not reach statistical significance. These data demonstrate that central CB₁ blockade can restore hepatic insulin action, ascribing a critical role to increased EC tone in the brain in the development of early insulin resistance.

Here we demonstrated that elevated EC tone in the brain is sufficient to induce hepatic insulin resistance. By use of pancreatic clamp studies, we showed that central infusions of the CB₁ and CB₂ agonist WIN and the specific CB₁ agonist ACEA impair the ability of systemic hyperinsulinemia to suppress hGP. Conversely, in a model of early insulin resistance characterized by hepatic insulin resistance, central CB₁ blockade was sufficient to restore hepatic insulin sensitivity. After 3 days of HFD, rats develop hepatic and hypothalamic insulin resistance (13). Because animal models of obesity are accompanied by elevated brain and hypothalamic EC levels (20,21), we hypothesized that elevated central EC tone might impair hypothalamic insulin action. By infusing rimonabant intracerebroventricularly, we were able to restore hepatic insulin action caused by overfeeding, emphasizing the relevance of brain CB₁ in the control of hepatic metabolism.

Our second major finding is that central CB₁ activation increases systemic lipolysis. We have recently shown that both systemic lipolysis and hGP can be regulated through brain insulin signaling (10), and because lipolytic flux can affect hGP (5,6), we investigated lipolysis after central EC system activation. The elevation in lipolysis after intracerebroventricular WIN administration increases the flux of glycerol from WAT to the liver, where glycerol would serve as a gluconeogenic substrate, which then contributes to the increased hGP observed during the clamp studies. Therefore, our data support the notion that brain EC tone impairs hypothalamic insulin action not only at the liver but also in WAT.

We measured lipolysis with R_a glycerol through the tracer dilution technique (36), which provides a good estimate of WAT lipolysis (42) for two reasons. First, WAT has very low glycerokinase activity (approximately 1,000-fold lower than liver [47]), and therefore WAT lipolysis results in glycerol release. Second, the liver releases less than 10% of all systemic glycerol (48). Thus, although WAT glycerol release does not entirely account for the R_a glycerol, WAT is the major contributor. We further complement this estimate of whole body glycerol flux by measuring the activation states of HSL and perilipin as estimates of local lipolytic activity. Both measurements support our conclusion that the central EC system regulates lipolysis in WAT.

It is interesting that both CB₁ agonists tended to increase peripheral glucose utilization, whereas the antagonist rimonabant tended to decrease this parameter during the intracerebroventricular infusions. Although the effect size was small and our data did not reach statistical significance, the trends were consistent in each of the three studies, raising the possibility that central CB₁ signaling increases peripheral glucose utilization. We speculate that this could be mediated through increased sympathetic outflow to muscle as has been described recently (49), because muscle is the major glucose utilizing organ, although other tissues may be involved. Therefore, with central CB₁ activation, the elevated R_d could compensate for the decreased hGP suppression, resulting in an unchanged GIR. Likewise, in the rimonabant-infused rats, enhanced hGP suppression could be tempered by this decreased R_d, resulting in a moderate overall increase in GIR.

Whether our studies translate to human physiology is unclear. Although the relevance of neuronal insulin signaling in regulating hGP has been established in rodents through pharmacological and genetic studies (4), the physiological relevance of this regulatory pathway remains uncertain in higher species. Cherrington’s group recently found that brain insulin does regulate hGP in dogs (50), although the effect magnitude was modest, possibly because of species discrepancies and differences in the experimental design of the clamp studies compared with earlier rodent studies. In humans, insulin regulates neuronal activity (51) and participates in weight regulation (52). However, whether brain insulin participates in the regulation of hGP and/or lipolysis remains unstudied in humans. One of the reasons why brain insulin could be more relevant in rodent physiology than in human is that rodents may exhibit a higher basal neural tone to the liver as has been previously suggested (53).

Our results with brain infusions of the CB₁ antagonist rimonabant differ from those obtained by Nogueiras et al. (32), where prolonged intracerebroventricular rimonabant treatment in a long-standing model of diet-induced obesity did not improve hepatic insulin sensitivity. It is important to point out some differences in the models used in these studies. We studied an early model of high-fat feeding-induced insulin resistance that is chiefly associated with hepatic insulin resistance (45) and characterized by impaired hypothalamic insulin action (13). Thus this model lacks the secondary sequela of long-term high-fat feeding, such as impaired glucose utilization and chronic inflammation. Conversely, Nogueiras et al. studied a model of prolonged high-fat feeding (32) where these complications are likely present and where the peripherally increased EC tone may become more prominent (18,54). Furthermore, our rimonabant dose was different. Whereas Nogueiras et al. (32) infused 5 μg per day, we infused 6.25 μg over a 6-h period. Our chosen intracerebroventricular dose was approximately 1% of the widely used dose of 10 mg/kg ip, which was also used in the abovementioned studies (32). It is possible that the intracerebroventricular dose chosen by Nogueiras et al. (32) was not sufficient to reveal effects of isolated brain CB₁ antagonism, or that in a model of longer-standing insulin resistance isolated central CB₁ antagonism is not enough to reverse the metabolic phenotype.

The role of neuronal EC signaling in regulating energy homeostasis has been highlighted in a study where CB₁ receptors were knocked out in CAMKIIα expressing neurons, including forebrain and sympathetic ganglions (34). These mice displayed a lean phenotype and were resistant to diet-induced obesity and insulin resistance. Although these mice were essentially unaffected by rimonabant treatment, the study did not answer the question whether the beneficial effects of pharmacological CB₁ antagonism are mediated through central versus peripheral mechanisms. Our study demonstrates that brain EC signaling does play an important role in regulating lipolysis, hGP, and possibly peripheral glucose utilization.

It has been shown that short-term overfeeding leads to hypothalamic insulin resistance (13). We speculate that this could be due, in large part, to the dysregulation of the EC system. The activation of the EC system in the brain blunts the ability of insulin to suppress both hGP and lipolysis, similar to what we found by infusing an insulin antagonist into the MBH (10). Under normal conditions, insulin could generate a signal in the MBH that is mediated through descending neuronal pathways that innervate WAT (12,55), resulting in the suppression of lipolysis. However, during states of elevated EC tone, ECs may act as retrograde inhibitors of synaptic transmission and thereby impair hypothalamic insulin action, potentially even in the absence of impaired intracellular insulin signaling. By infusing rimonabant intracerebroventricularly, we released this blockade on insulin-induced synaptic transmission and restored hepatic and WAT insulin action. These data strongly suggest that central EC tone plays an important role in the development of hypothalamic insulin resistance. Although our study does not negate prior work suggesting that the EC system has important peripheral effects, we do demonstrate that the central EC system acutely participates in the regulation of systemic glucose and lipid homeostasis. Therefore, a peripherally restricted CB₁ antagonist might not be as therapeutically potent as the brain-penetrating rimonabant, although it is likely that the feared psychiatric side effects would be avoided.

This work was supported by NIH grants DK-074873, DK-083568, and DK-082724, and the American Diabetes Association, to C.B. C.B. is the recipient of an Irma T. Hirschl Scholar Award.

No potential conflicts of interest relevant to this article were reported.

J.D.O. wrote the article, researched data, and contributed to discussion. E.Z. researched data and contributed to discussion. B.C. researched data. T.S. researched data and contributed to discussion. C.B. designed and supervised the studies, wrote the article, and contributed to discussion.

The authors thank Judith Agudo-Cantero, Claudia Lindtner, Anna Mechling, Mark Real, Toni Salama, Scott Schoninger, Lilya Semenova, Ashlie Sewdass, and Kai Su (all from Mount Sinai) for excellent technical assistance; Drs. Andrew Greenberg (Tufts University) and Dawn Brasaemle (Rutgers School of Environmental and Biological Sciences) for the Perilipin and CGI-58 antibodies, respectively; and Drs. George Kunos (National Institute on Alcohol Abuse and Alcoholism) and Matthias Tschöp (University of Cincinnati) for critical reading of the article. The authors acknowledge the Case Western University MMPC, which is supported by U24 DK-76169, for the mass spectrometry analyses.