CB1 and GLP-1 Receptors Cross Talk Provides New Therapies for Obesity

Few, and often subeffective, drugs are available to address the increasing prevalence of obesity and type 2 diabetes (1). Glucagon-like peptide 1 receptor (GLP-1R) agonists improve glycemic control (2) but show suboptimal efficacy against adiposity and insulin resistance (2–4). The first generation of brain-penetrant cannabinoid receptor type 1 (CB1R) antagonists displayed relevant body weight (BW)-lowering action but also caused neuropsychiatric adverse effects (5,6). Recently developed peripherally restricted CB1R inhibitors overcome these safety concerns, while potently lowering BW in animal models of obesity (7,8). These compounds also improve systemic and tissue-specific metabolic complications, including insulin and leptin resistance, dyslipidemia, nonalcoholic fatty liver disease, inflammation, β-cell loss, and diabetic nephropathy (6–10). Notably, GLP-1 secretion is modulated by changes in CB1R signaling (11), while the insulinotropic action of GLP-1R agonism is enhanced in mice lacking CB1R (12), suggesting the existence of reciprocal functional interactions between GLP-1R and CB1R.

Here we demonstrate that a cross talk between GLP-1R and peripheral CB1R signaling modulates food intake and BW. Coadministration of the peripheral CB1R blocker JD-5037 with GLP-1R agonists to diet-induced obese (DIO) mice has greater efficacy to lower BW than monotherapies, with superior and greater beneficial effects on insulin resistance, dyslipidemia, and hepatic steatosis. This combinatorial strategy may be used to correct obesity and its comorbidities and warrants further clinical investigation.

The experiments were in compliance with European Union Directives (2010/63/EU), the University of Bordeaux Ethical Committee (DIR1354), and the University of Cincinnati Animal Care Committee. Male mice were housed at 22 ± 2°C (12 h dark/light cycle). CB1-knockout (CB1-KO) mice (13) were fed a regular chow diet (SAFE A03; SAFE diets). Eight-week-old male C57BL/6J mice (Janvier Laboratories, Le Genest-Saint-Isles, France) were fed a chow diet (SAFE A03) or a high-fat diet (HFD, D12492 or D12331; Research Diets) for 6 months before the study. GLP-1R-knockout (GLP-1R-KO) mice (14) were fed a HFD (D12331; Research Diets) for 16 weeks before the study. JD-5037 (#530481; MedKoo Biosciences) and rimonabant (9000484; Cayman Chemical) were given i.p., and IUB48 (14) or semaglutide (synthetized at Novo Nordisk Research Center of Indianapolis, IN) were given subcutaneously. JD-5037 was prepared in 0.9% NaCl with Tween 80 (1%) and DMSO (4%). Semaglutide was prepared in 50 mmol/L phosphate buffer (pH 7.4) with 0.05% Tween 80. Mice were randomly assigned to pharmacological treatment groups, and the experimenters were not blinded to the intervention groups. Animals were single housed or group housed, as specified in the figure legends.

Energy expenditure was calculated using an energy balance technique (total energy expenditure balance [TEEbal]) (15).

Plasma insulin and leptin was determined by ELISA kit (Mercodia 10-1132-01 and 10-1113-01, and Millipore EZML-82K, respectively) and plasma and hepatic triglycerides by Thermo Scientific TR22421. For hepatic triglycerides extraction, livers (200 mg) were incubated overnight at 55°C in 30% ethanolic KOH. After 50% ethanol was added, they were centrifuged (13,000 rpm, 5 min), and 100 µL of MgCl₂ (1 mol/L) was added to 100 µL of supernatant. Samples were centrifuged again (13,000 rpm, 5 min) before quantification. Unknown triolein equivalents were interpolated using logistic regression. Triglycerides content was calculated in mg/g tissue by triolein equivalents ∗ 2 ∗ 0.012/(tissue weight in grams).

Body composition analysis was performed with an EchoMRI-900 (Echo Medical Systems).

Glucose tolerance, insulin tolerance, and glucose-induced insulin release were performed as previously described (14) using fasting durations and glucose/insulin concentrations detailed in the figure legends. The glucose disappearance rate (KITT) was calculated as the slope of the decreasing line of blood glucose levels over 30 min from insulin administration, as previously described (16).

RNA tissue extraction and quantitative (q)PCR were performed as previously reported (17). Primers are provided in Supplementary Table 1.

Data are mean ± SD and were analyzed using Prism 8 (GraphPad). ANOVA, followed by the appropriate post hoc test, was used as specified in the figure legends.

All data and resources are available from the corresponding authors upon request.

To explore possible functional interactions between CB1R and GLP-1R, we analyzed short-term changes in food intake and glycemic control in chow-fed CB1-KO or wild-type (WT) littermates acutely treated with multiple doses of a GLP-1R agonist (IUB48) (14) after 18 h of fasting, which activates the endocannabinoid system (18). At 10 nmol/kg, IUB48 suppressed food intake in both WT and CB1-KO mice, with a greater effect in the KO (Fig. 1A). A lower dose (0.5 nmol/kg) of IUB48 still reduced food intake in CB1-KO mice but not in WT controls (Fig. 1B). No genotype-dependent changes in systemic glucose tolerance were observed after testing several doses of IUB48 (Fig. 1C and Supplementary Fig. 1A). Accordingly, no changes in glucose-induced insulin release were observed after treatment with a low (0.1 nmol/kg) IUB48 dose (Supplementary Fig. 1B).

Next, we probed the effects of the coadministration of different doses of IUB48 and the CB1R antagonist rimonabant on fasting-induced food intake in chow-fed BL6 mice. Coadministration of appetite-suppressant doses of IUB48 and rimonabant reduced food intake to a greater extent relative to monotherapies (Fig. 1D and E). A greater hypophagic effect was also observed after cotreatment with an effective dose of IUB48 and a noneffective dose of rimonabant (Fig. 1F). Similarly, the combination of noneffective doses for both molecules reduced food intake in both refed (Fig. 1G) and free-fed (Supplementary Fig. 1D) animals. Thus, functional comodulation of CB1R and GLP-1R synergistically reduced food intake but did not impact acute changes in glycemic control.

Because peripheral CB1R antagonists have promising antiobesity and antidiabetes action (8), we tested the antiobesity effects of the combined chronic administration of IUB48 and the peripherally restricted CB1R inhibitor JD-5037 (7) in DIO mice. To uncover possible synergistic effects, we used a dose of JD-5037 (1 mg/kg) with subthreshold weight-lowering and hypophagic action (Supplementary Fig. 2A) and a dose of IUB48 (100 nmol/kg) with clear effects on insulin secretion but minimal efficacy against BW loss (14). As expected, 2 weeks of IUB48 monotherapy did not alter BW, food intake, or fat mass, whereas JD-5037 significantly reduced these parameters (7) (Fig. 2A–C and Supplementary Fig. 2B). The cotreatment caused even greater reductions than monotherapies (Fig. 2A–C), without affecting lean mass (Fig. 2C), and also had greater hypophagic action than the monotherapies at day 1 and during week 2 (Fig. 2B), in agreement with our previous findings (Fig. 1).

JD-5037 reduced plasma leptin levels (Fig. 2D) and leptin mRNA expression in brown and white adipose tissues (Fig. 2E). The cotreatment caused similar or even greater reduction in these markers (Fig. 2D and E). No changes in systemic insulin action were found among treatments at day 8, as assessed by the analysis of the percentage blood glucose changes (Supplementary Fig. 2C) and by the KITT (Supplementary Fig. 2D). However, after 14 days, the cotreatment, but not monotherapies, significantly improved systemic insulin sensitivity (Fig. 2F and G). JD-5037–mediated effects on hepatic insulin resistance involve reduced hepatic content of ceramide via the lowered expression of the ceramide-producing enzyme SPTLC3 (9). Likewise, JD-5037 and JD-5037 + IUB48, but not IUB48 alone, reduced hepatic SPTLC3 mRNA expression after 2 weeks (Fig. 2H). This was paralleled by the increased hepatic expression of the insulin sensitivity marker IRS1 (Fig. 2I) and by a marked reduction in hepatic and systemic triglyceride levels in mice receiving the cotreatment (Fig. 2J–L).

The antiobesity action of the cotreatment was due to both food intake-dependent and -independent effects, since mice treated with IUB48 + JD-5037 had higher estimated energy expenditure relative to the vehicle (Fig. 2M). Consequently, animals pair-fed to the intake of mice receiving IUB48 + JD-5037 (Supplementary Fig. 2E) had less pronounced reductions in BW and fat mass than the cotreatment group (Supplementary Fig. 2F and G).

However, pair-feeding and the cotreatment similarly improved systemic insulin tolerance (Supplementary Fig. 2H and I). Thus, in response to combination therapy, compound-induced hypophagia is implicated in the observed ameliorations in insulin sensitivity, while mechanisms independent from food intake contribute to weight loss.

Our data suggest that loss of CB1R function potentiates the hypophagic effects of GLP-1R agonism. However, the opposite relationship (i.e., a GLP-1R–mediated influence on CB1R signaling) could also be true. To address this second possibility, we tested the pharmacological efficacy of JD-5037 in GLP-1R-KO or WT littermates fed a HFD. After acute administration, JD-5037 did not modify the glycemic response to intraperitoneal or oral glucose loading in either WT or GLP-1R-KO mice (Fig. 3A–C). This is in agreement with the lack of functional interaction observed on acute glucose responses (Fig. 1C). However, the BW-lowering effect of JD-5037 was blunted in GLP-1R-KO animals (Fig. 3D), and this was reflected in a decreased hypophagic action of this molecule in GLP-1R-KO mice relative to WT controls (Fig. 3E). These results were not influenced by the initial BW differences between GLP-1R-KO and WT mice at day 0 of the study, since no correlations were found between these initial BW values and the BW changes induced by the drug (Fig. 3F). Thus, GLP-1R partially mediates the weight-lowering and hypophagic effects of JD-5037, which highlights the existence of a bidirectional cross talk between GLP-1R and CB1R signaling.

The long-acting GLP-1R agonist semaglutide shows greater glycemic benefits and a superior weight-lowering action than cognate molecules (2). To further investigate the therapeutic potential of GLP-1R and CB1R cotargeting, we assessed the efficacy of different doses of JD-5037 combined with one dose of semaglutide having submaximal BW-lowering effects (data not shown) in DIO mice. Semaglutide lowered BW, but it did not improve systemic insulin sensitivity relative to the vehicle, albeit it significantly lowered fasting blood glucose levels (Fig. 4 and Supplementary Fig. 3). Accordingly, the glycemic effects of GLP-1R agonists are not due to direct effects on systemic insulin action and are mainly due to improved insulin secretion and inhibition of glucagon production (3,4). When semaglutide was coadministered with a subthreshold dose of JD-5037 (0.3 mg/kg), which does not affect insulin sensitivity, this led to BW loss and to a significant improvement in insulin sensitivity relative to vehicle treatment, as assessed by the analysis of percentage blood glucose changes (Fig. 4A). Cotreatment of semaglutide with higher doses of JD-5037 (1 mg/kg and 3 mg/kg) caused greater BW loss than monotherapies (Fig. 4B and C). At these higher doses, the cotreatment and JD-5037 monotherapy, but not semaglutide, improved insulin sensitivity, as assessed by both the analysis of percentage blood glucose changes and by KITT analysis (Fig. 4B and C). The cotreatment at the higher doses also lowered basal blood glucose levels relative to vehicle administration (Supplementary Fig. 3B and C). Thus, the combination of JD-5037 with semaglutide causes superior weight loss and greater insulin-sensitizing effects than semaglutide monotherapy.

We report that GLP-1R and peripheral CB1R signaling interact in a bidirectional manner to control energy homeostasis. Blockade of CB1R potentiates GLP-1–mediated effects on food intake and fat mass, whereas GLP-1R signaling is partially required for endocannabinoid-mediated actions on energy balance, because the weight loss induced by peripheral CB1 blockade is blunted in GLP-1R-KO mice. Such a cross talk can be pharmacologically manipulated to achieve potent antiobesity and antidiabetic solutions, and this may have clinical implications. Based on our data, the addition of peripheral CB1R inhibitors to currently available GLP-1R mimetics may allow potent weight loss and larger improvements in insulin resistance, systemic dyslipidemia, and nonalcoholic fatty liver disease, relative to monotherapies, without inducing the neuropsychiatric alterations observed with brain-penetrant CB1R antagonists (6,7). This approach could be also effective against nonalcoholic steatohepatitis, which still lacks adequate medical treatments, a possibility supported by the potent reduction in hepatic lipid accumulation observed herein, by the known ability of JD-5037 to attenuate liver fibrosis in preclinical models (19) and by on-going or recently terminated clinical trials investigating the use of both CB1R inhibitors and semaglutide on this liver dysfunction.

The greater efficacy observed with GLP-1R and CB1R comodulation likely results from multiple mechanisms that involve both food intake reduction and increased energy dissipation. JD-5037 may amplify the anorectic activity of GLP-1 (or vice versa) via periphery-to-brain signals involving the vagus nerve, which coexpress CB1R and GLP-1R (20) thus contributing to the hypophagic effects of CB1R inhibitors (18). The antiobesity effects of JD-5037 derive from changes in both food intake and energy expenditure (6,7), and these changes are due to restoration of endogenous leptin sensitivity (7,21,22) and lowered leptin production from adipocytes (7,22) via blockade of adipocytes CB1R (6,23). These same processes may be implicated in the antiobesity effects observed herein, since our cotreatment led to profound reductions in both circulating leptin levels and production. JD-5037 ameliorates insulin resistance via inhibition of hepatic CB1R (9) and via reduced accumulation of hepatic ceramide species interfering with insulin action (9,21). This mechanism, which involves an inhibitory effect on the ceramide-producing enzyme SPTLC3 (9), may be engaged by our cotreatment, which lowers hepatic SPTLC3 expression and improves markers of systemic and hepatic insulin action.

Thus, multiple neuronal and nonneuronal peripheral cell types expressing CB1R and GLP-1R may be involved in the antiobesity effects observed. A limitation of our work is that we do not dissect the exact identity of these peripheral sites of actions, which will require follow-up investigations.

Obesity is a heterogeneous disease requiring pharmacological approaches able to target multiple pathways. Unimolecular GLP-1R–based multiagonists affecting multiple metabolic targets have exceptional preclinical efficacy (4,14) and are being investigated in clinical studies. Chemical coupling of these newly developed molecules with a growing number of peripheral CB1R inhibitors (8) may have a transformative impact in obesity and diabetes pharmacotherapy, an exciting perspective given the structural and functional plasticity of CB1R (24) and the emergence of novel generations of “fusion” peripheral CB1R inhibitors (8).

Acknowledgments. The authors thank the animal facility, genotyping, and transcriptome platforms of the INSERM U1215 NeuroCentre Magendie, funded by INSERM and LabEx Bordeaux Region Aquitaine Initiative for Neurosciences (BRAIN) (ANR-10-LABX-43), for animal care, mouse genotyping, and qPCR analyses. In particular, the help of Delphine Gonzales (Neurocentre Magendie, U1215, Bordeaux), Fiona Corailler (Neurocentre Magendie, U1215, Bordeaux), Helizabeth Huc (Neurocentre Magendie, U1215, Bordeaux), and Helene Doat (Neurocentre Magendie, U1215, Bordeaux) is acknowledged. The authors thank Charlotte Svendsen (Novo Nordisk Foundation Center for Basic Metabolic Research, Faculty of Health and Medical Sciences, University of Copenhagen, Denmark) for excellent laboratory support.

Funding. C.C. is supported by research grants from the Lundbeck Foundation (Fellowship R238-2016-2859) and the Novo Nordisk Foundation (grant number NNF17OC0026114). D.C. and C.Q. are supported by INSERM. D.C. is also supported by Nouvelle Aquitaine Region, LabEx Bordeaux Region Aquitaine Initiative for Neurosciences (BRAIN) (ANR-10-LABX-43), and Agence Nationale de la Recherche (ANR-10-EQX-008-1 OPTOPATH, ANR-17-CE14-0007 BABrain, and ANR-18-CE14-0029 Mitobesity). C.Q. is also supported by the Francophone Society of Diabetes (SFD), the French Society of Endocrinology (SFE), and the French Society of Nutrition (SFN). Novo Nordisk Foundation Center for Basic Metabolic Research is an independent Research Center, based at the University of Copenhagen, Denmark, and partially funded by an unconditional donation from the Novo Nordisk Foundation (www.cbmr.ku.dk) (grant number NNF18CC0034900).

Duality of Interest. B.F. and R.H. are current employees of Novo Nordisk. D.C. and G.M. are cofounder and stakeholders of the biotech Aelis Farma. D.P.-T. maintains research collaborations and receives funding from Novo Nordisk, Cohbar Inc., and MBX Biosciences Inc. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. P.Z. performed most of the studies and analyzed data. R.H., S.F., L.B., C.A., S.C., T.L.-L., G.M., C.C., and D.P.-T. performed studies, analyzed data, and gave critical inputs. B.F., D.C., and C.Q. conceptualized and designed the experiments, analyzed data, and wrote the manuscript. C.Q. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.