Endocannabinoids and cannabinoid type 1 (CB₁) receptors are important players in the regulation of energy homeostasis, having the ability to fine-tune the activity of metabolically relevant tissues, including the hypothalamus, the adipose tissue and the liver (Bensaid et al., 2003; Bermudez-Silva et al., 2012; Bermúdez-Silva et al., 2008; Cota et al., 2003; Lipina et al., 2010; Osei-Hyiaman et al., 2005). Rimonabant, the first-in-class CB₁ antagonist/inverse agonist was marketed to treat complicated obesity, although central side effects led to its withdrawal later on (reviewed in Bermudez-Silva et al., 2010; Jones, 2008). Rimonabant counteracted the positive energy balance characterizing obese patients by decreasing food intake while inducing energy expenditure and lipolysis, and modulated other peripheral metabolic processes (reviewed in Bermudez-Silva et al., 2012). Among the key metabolic processes modulated by rimonabant, glucose homeostasis is an important one (reviewed in Nogueiras et al., 2008), with two outstanding physiological mechanisms underlying it: insulin sensitivity and insulin release. The latter is controlled by the endocrine pancreas. Nearly all the endocannabinoid system (ECS) components are expressed in the endocrine islets. In particular, CB₁ receptors are expressed in both alpha (α) and beta (β) cells and their activation is coupled to insulin release, whereas their blockade seems to decrease insulin secretion (reviewed in Li et al., 2011; Doyle, 2011). Interestingly, and underlining the physiological relevance of pancreatic ECS, intra-islet endocannabinoid levels are known to increase after a glucose challenge (Bermúdez-Silva et al., 2008; Kim et al., 2011). Moreover, rimonabant decreases insulin hypersecretion in isolated islets from diabetic rats (Getty-Kaushik et al., 2009), and its chronic administration improves pancreatic function and islet morphology in diabetic rats (Duvivier et al., 2009). Finally, recent evidence points to an important role of CB₁ receptors in the regulation of expansion of the β-cell mass (Kim et al., 2011, 2012).

The mammalian or mechanistic target of rapamycin complex 1 (mTORC1) is a rapamycin-sensitive multiprotein complex comprising the serine-threonine kinase mTOR and the proteins raptor, PRAS40 and mLST8 (Laplante and Sabatini, 2012). It functions as a nutrient sensor able to integrate signals from growth factors and hormones through the phosphoinositide-3-kinase–protein kinase B/Akt (PI3K-PKB/Akt) pathway. mTORC1 regulates protein synthesis and different aspects of cell growth and metabolism mainly via p70S6 kinase 1 (p70S6K1) and the eukaryotic translation initiation factor 4E-binding protein 1 (4E-BP1) (Laplante and Sabatini, 2012).

Like the ECS, the mTORC1 pathway has also been involved in β-cell physiology (reviewed in Leibowitz et al., 2008). Transient mTORC1 activation leads to increased β-cell/islet size, β-cell mass expansion and insulin production (Blandino-Rosano et al., 2012; Mori and Guan, 2012; Van de Velde et al., 2011; Velazquez-Garcia et al., 2011), whereas chronic mTORC1 activation has rendered conflicting results depending on the nature of the specific activation and/or the experimental model used. mTORC1 overactivation induced by deletion of tuberous sclerosis complex 1 or 2 (TSC1/2) favours β-cell mass expansion and glucose tolerance (Blandino-Rosano et al., 2012; Rachdi et al., 2008), in sharp contrast with other studies in which stimulation of mTORC1 in β cells induced by glucose or IGF-1 led to inhibition of the IRS2/Akt pathway and β-cell apoptosis (Briaud et al., 2005). Conversely, mice deficient for p70S6K1 (S6K1^−/−) are characterized by the lack of the p70S6K1-dependent negative feedback on IRS1/2, leading to enhanced insulin sensitivity and normal glucose levels in diet-induced obesity (Um et al., 2004). Strikingly, mTORC1 inhibition by rapamycin worsens hyperglycaemia and the metabolic state in type 2 diabetes (Fraenkel et al., 2008; Pereira et al., 2012) while reducing islet engraftment and impairing β-cell function (Bell et al., 2003; Bussiere et al., 2006; Zhang et al., 2006). This apparent paradox seems to have been resolved very recently by the demonstration that the deleterious effects of rapamycin might be mediated through mTOR complex 2 (mTORC2) (Barlow et al., 2012; Lamming et al., 2012).

Thus, based on (1) the important role of both the ECS and the mTORC1 pathway in whole-body energy homeostasis, (2) their localization and involvement in key physiological processes within the islets of Langerhans, including insulin secretion and β-cell mass expansion, and (3) previous findings linking both signalling systems in other tissues (Puighermanal et al., 2009; Senin et al., 2013), we hypothesized the existence of a functional link between the ECS and the mTORC1 pathway in the regulation of glucose-stimulated insulin secretion (GSIS) in the islets of Langerhans. Here, we demonstrate that CB₁ receptor antagonism decreased GSIS in isolated islets of Langerhans from chow-fed lean mice, with this effect requiring a functional mTORC1 pathway. Furthermore, CB₁ receptor antagonism promoted glucose intolerance in vivo, an effect that was counteracted by acute pharmacological blockade of the mTORC1 pathway.

In this manuscript, we present evidence that cannabinoid CB₁ receptors and the mTORC1 pathway interact in the islets of Langerhans to modulate GSIS. Furthermore, systemic pharmacological blockade of these two metabolic signalling systems involved in energy homeostasis and nutrient sensing reveals a functional interaction in the regulation of glucose homeostasis.

Pharmacological modulation of CB₁ receptors in the islets modulates GSIS (for recent reviews see Li et al., 2011; Doyle, 2011) and β-cell mass (Duvivier et al., 2009; Getty-Kaushik et al., 2009; Kim et al., 2011, 2012). Thus, it looks likely that both GSIS and β-cell mass expansion are cellular processes intrinsically modulated by endocannabinoids. However, there are conflicting data in the literature regarding which cell types express (or not) CB₁ within the endocrine pancreas and in which way CB₁ activation alters GSIS. Adding more controversy to this issue, rimonabant has been shown to act not only through CB₁ receptors but also through GPR55 (Kapur et al., 2009; Lauckner et al., 2008; Yin et al., 2009), and GPR55 is expressed in the endocrine pancreas, with its activation increasing GSIS (Romero-Zerbo et al., 2011; McKillop et al., 2013). Our results on C57BL/6, CB₁^+/+ and CB₁^−/− islets show that rimonabant promotes both CB₁-dependent and CB₁-independent decreases of GSIS, with doses ranging from 0.1 to 1 µM requiring action on CB₁ receptors. Indeed, these doses decreased plasma insulin levels and induced glucose intolerance in C57BL/6 and CB₁^+/+, but not in CB₁^−/− littermates.

In previous studies, we have investigated glucose tolerance in chow-fed rats using another CB₁ antagonist, AM251, and this drug was found to increase glucose tolerance at low doses (0.01-0.2 mg/kg), whereas it had no effect at higher doses (1 mg/kg) (Bermudez-Silva et al., 2007; Bermúdez-Siva et al., 2006). A couple of reasons could account for the different effects observed with AM251. First, several investigations have shown that AM251 is a potent agonist of GPR55 (Kapur et al., 2009), and GPR55 is expressed in rat islets, as previously mentioned (Romero-Zerbo et al., 2011), which suggests that some of the previously published findings using AM251 could be partly due to GPR55 activation. Second, it has been reported that there is some degree of species-specific expression for CB₁, CB₂ and GPR55 (Li et al., 2011; Romero-Zerbo et al., 2011), with CB₁ and GPR55 being more intensely expressed in rat than in mouse β cells. Of note, in the present study, we validated our findings in CB₁^−/− mice, showing that rimonabant-induced glucose intolerance is mediated through CB₁ receptors. Indeed, lower plasma insulin levels were detected in mice injected with 3 mg/kg rimonabant when compared to vehicle-injected mice, suggesting that, at least in part, glucose intolerance was due to decreased insulin secretion, probably as a consequence of the inhibitory effects of rimonabant on islet function. Taken together, these results indicate that CB₁ receptors in the islets of chow-fed lean mice are important in the modulation of whole-body glucose homeostasis.

With the present results we expand the knowledge about the downstream mechanisms being recruited by CB₁ receptors in the islets to modulate GSIS. In fact, both genetic and pharmacological blockade of mTORC1 abolish CB₁-mediated rimonabant action on GSIS. Interestingly, a dose of rapamycin lacking effect on GSIS is sufficient to block rimonabant action, whereas a higher dose decreases GSIS. This evidence suggests that rapamycin acts simultaneously on different targets to modulate GSIS. Whereas a low dose inhibits the rimonabant-induced effect on GSIS, thus unmasking the dependence of rimonabant action on an intact mTORC1 pathway, higher doses of rapamycin decrease GSIS. Rapamycin has been found to inhibit mTORC2 (Barlow et al., 2012; Lamming et al., 2012) and to cause insulin resistance (Lamming et al., 2012). However, doses reported to induce insulin resistance were higher than the ones used in our study and they were chronically administered. By contrast, we demonstrated that rapamycin administered acutely at 0.1 mg/kg, although decreasing insulin levels, increased insulin sensitivity. Although we have not measured rapamycin plasma levels in our animals, according to previous findings (Guba et al., 2005), 0.1 mg/kg rapamycin should be leading to around 20-30 nM of the compound in the plasma, a dose that we have found to decrease GSIS through mTORC1. Thus, the effect of rapamycin on GSIS is dependent on mTORC1, although further studies are needed in order to unequivocally establish that action on mTORC2 or other molecular pathways is not required.

A CB₁-acting dose of rimonabant increased phosphorylation of both p70S6K1 and rpS6, downstream targets of the mTORC1 pathway, and decreased GSIS. However, the mechanisms linking S6K1/rpS6 phosphorylation and GSIS are unclear at present. mTORC1 is an important player in islet physiology, especially in the control of β-cell mass (Pende et al., 2000), which has also been shown to be involved in the pathophysiology of type 2 diabetes (Leibowitz et al., 2008). Although β-cell mass investigation was beyond the scope of our work, it is noteworthy that Kim and colleagues have shown that CB₁ blockade in islets is coupled to increased insulin receptor activation through impaired Gαi/o-coupled inhibition, Akt phosphorylation and increased β-cell mass (Kim et al., 2011, 2012). Given that Akt is a well-known activator of mTORC1, the rimonabant-induced increase in mTORC1 activation might be mediated by the Akt pathway. Likewise, given the central role of mTORC1 in modulating β-cell mass (Rachdi et al., 2008; Leibowitz et al., 2008), our results suggest that the previously reported CB₁-dependent β-cell mass expansion (Kim et al., 2011) could involve mTORC1. Rimonabant has been described to decrease plasma insulin levels and insulin hyper-secretion from isolated islets of diabetic Zucker rats (Duvivier et al., 2009; Getty-Kaushik et al., 2009). Furthermore, histo-morphological analysis of pancreas from rimonabant-treated Zucker rats has suggested a protective role of CB₁ blockade on islet integrity (Duvivier et al., 2009). Consequently, beyond the important role of islet CB₁ receptors in physiology, rimonabant treatment in diabetes improves islet function by mechanisms that are still not well understood. Our findings suggest that mTORC1 activation could be one of these mechanisms, although additional experiments on diabetic animals need to be carried out in the future.

To study the in vivo relevance of the interaction between the ECS and mTORC1 signalling on glucose homeostasis, we first analyzed glucose tolerance and insulin plasma levels in rapamycin-injected mice. We found no effect on glucose tolerance when rapamycin was used in the range of 0.001-0.1 mg/kg, but the highest dose decreased plasma insulin levels. Then, to study the putative interaction with CB₁-dependent signalling, we combined the intraperitoneal administration of rapamycin with rimonabant. Rapamycin at 0.1 mg/kg counteracted rimonabant action on glucose responses during a GTT. However, both rimonabant and 0.1 mg/kg rapamycin decreased plasma insulin levels, with their combination not altering plasma insulin levels further. This latter result implies that rapamycin prevents rimonabant-induced glucose intolerance by increasing insulin sensitivity, as was shown by an ITT. Thus, these pharmacological in vivo experiments suggest that the pancreatic interaction between CB₁ signalling and the mTORC1 pathway has a whole-body impact on glucose homeostasis (see Fig. 8). However, additional extra-pancreatic effects of rimonabant on glucose homeostasis cannot be ruled out and those of rapamycin are evident. In fact, rapamycin at the dose of 0.1 mg/kg has insulin-sensitising properties in mice, an effect that could be related to increased glucose uptake in insulin target tissues, such as the muscle or the adipose tissue. Administration of rapamycin in vivo has generally led to insulin resistance and glucose intolerance (Fraenkel et al., 2008; Rachdi et al., 2008). In the muscle, rapamycin did not improve either insulin sensitivity nor glucose tolerance (Miller et al., 2008), although higher doses (5 mg/kg) that likely targeted mTORC2 were used. In isolated human subcutaneous and omental adipocytes, ex vivo incubation with low doses of rapamycin (10 nM) impaired insulin signalling and glucose uptake (Pereira et al., 2012), but the extent to which this effect contributes to peripheral insulin sensitivity in humans is unknown. Our findings instead suggest that low doses of rapamycin, by specifically inhibiting mTORC1, could be beneficial in pathological conditions where an excessive insulin secretion and signalling is present, such as in insulin resistance (Fig. 8).

Thus, we propose a molecular model by which CB₁ receptor antagonism in islets leads to decreased GSIS through mTORC1 activation. This in turn would decrease plasma insulin levels without altering insulin sensitivity and hence inducing glucose intolerance (Fig. 8A). On the other hand, mTORC1 blockade would prevent the rimonabant-induced decrease in GSIS, although it decreases GSIS per se, while increasing peripheral insulin sensitivity, thus maintaining glucose homeostasis (Fig. 8B). Our experiments do not allow the determination of whether CB₁ receptors are functionally coupled to mTORC1 in metabolic processes at other peripheral tissues, but this possibility cannot be ruled out and will need to be investigated in future studies. Unravelling the coupling between CB₁ receptors and the mTORC1 pathway also deserves more experimental work, but, based on the published literature, it is tempting to speculate that rimonabant is blocking the inhibitory action of the CB₁-Gi protein on phosphorylation of insulin receptors (or other growth-factor receptors) present in islets and in turn phosphorylation of IRS1/2. In fact, the inhibitory actions of endocannabinoids on insulin receptor phosphorylation in islets have been reported previously (Kim et al., 2011).

Overall, our results support the notion that CB₁ and mTORC1 signalling share common molecular circuits in the endocrine pancreas so as to modulate insulin secretion and that this interaction is relevant at the whole-body level to modulate glucose homeostasis. These observations open new research avenues regarding the potential therapeutic value of strategies exploiting this functional interaction in diabetes. The next steps should include an in-depth exploration of the CB₁-mTORC1 interaction in the modulation of both glucose-stimulated insulin secretion and β-cell mass expansion both in human islets and in animal models of type 2 diabetes.

We thank Dr S. Kozma and Dr G. Thomas (University of Cincinnati and IDIBELL Barcelona) for kindly donating the S6K1^−/− mouse strain. We also thank Melissa Elie, Samantha Clark and Vanesa Espinosa-Jimenez for technical assistance. The authors also gratefully acknowledge all the staff of the animal facilities at the Neurocentre Magendie and at the IBIMA (ECAI de Experimentación Animal), and especially IBIMA animal facility coordinator Dr Ricardo González Carrascosa.