Cannabinoids Inhibit Insulin Receptor Signaling in Pancreatic β-Cells

Sources and dilutions of primary antibodies used in immunoblotting, immunoprecipitation, and immunostaining are listed in Supplementary Table 1. AEA, 2-AG, AEA-d8, 2-AG-d5, WIN55,212-2, arachidonyl-2-chloroethylamide (ACEA), AM251, and AM630 were obtained from Cayman Chemical (Ann Arbor, MI). GFP-HA-tagged CB1R was from K. Mackie (Indiana University). The human IR and Gα_i3 cDNA were amplified by RT-PCR from a human pancreas RNA (Stratagene, La Jolla, CA), with oligo-dT (18 bp) for the reverse transcription. The IR cDNA was incorporated into a 3×Flag vector and the Gα_i3 cDNA was incorporated into an mVenus-C1 vector. IR mutant (IR-3YA), whose Tyr1158/1162/1163 residues were substituted to Ala, was generated from wild-type IR (IR-WT) using a QuikChange II XL site-directed mutagenesis kit (Stratagene).

CB1R^−/− mice and their WT littermates were developed and backcrossed to a C57Bl/6 J background, as previously described (16). The study used male 2- to 3-month-old CB1R^+/+ and CB1R^−/− mice. AM251 (10 mg/kg) was administrated by daily intraperitoneal injection to 1-month-old normal C57Bl/6 J mice for 4 weeks. FAAH^−/− mice were obtained from Dr. Benjamin Cravatt (Scripps Research Institute), and 1-month-old db/db mice were from The Jackson Laboratory (Bar Harbor, ME).

DMSO, AM251 (10 mg/kg), and AM630 (10 mg/kg) were injected daily for 4 weeks. Then, the pancreata were dissected and blood was collected in the ad lib state of eating. Blood glucose concentration was measured using an Elite glucometer (Bayer Healthcare, Tarrytown NY), and plasma insulin was measured using a rat/mouse insulin ELISA Kit (Crystal Chem, Downers Grove, IL). All animal care and experimental procedures followed National Institutes of Health (NIH) guidelines and were approved by the National Institute on Aging Animal Care and Use Committee.

Pancreata were fixed in 4% paraformaldehyde, immersed in 20% sucrose before freezing, and sectioned (7-μm thickness). Human and mouse paraffin-embedded pancreatic sections were immunostained as before (17). After antigen unmasking, the slides were incubated with primary antibodies (Supplementary Table 1), followed by secondary antibodies (Invitrogen, Carlsbad, CA) along with TO-PRO-3 (Invitrogen), in some cases, for nuclear staining.

Slides were viewed using a LSM-710 confocal microscope (Carl Zeiss MicroImaging, München-Hallbergmoos, Germany). Multiple sections from 3 to 5 mice per group, separated by at least 200 μm from each section, were assessed for signal intensity and the number of nuclear p27- or proliferating cell nuclear antigen (PCNA)-positive (PCNA⁺) β-cells with LSM Image Browser software (Carl Zeiss). To determine β-cell area and mass, the pancreata were trimmed of all nonpancreatic tissue and processed for immunostaining. The cross-sectional areas of pancreata and β-cells were determined from multiple sections (n = 5 mice per group), separated by at least 200 μm from each section, using LSM Image Browser software. The relative cross-sectional area of β-cells was determined by quantification of the cross-sectional area covered by insulin-positive cells divided by the cross-sectional area of total pancreas tissue. The β-cell mass per pancreas was estimated as the product of the relative cross-sectional area of β-cells per total tissue and the weight of the pancreas.

Human islets were provided by the Islet Cell Resource Center. Mouse islets were isolated using collagenase digestion, as we previously described (18). Human or mouse islets in Dulbecco’s modified Eagle’s medium (DMEM) containing 4 mmol/L glucose and 1% BSA were pelleted and incubated with DMEM containing 4, 10, or 15 mmol/L glucose, with or without KCl (30 mmol/L) for 10 min. Lipid extraction and tandem mass spectrometry analysis were done next, and details are described in the Supplementary Data.

We used the PixCell II Workstation (Arcturus Engineering, Carlsbad, CA) to perform laser capture microdissection and image acquisition of β-cells in pancreas sections, as we previously described (17). Total RNAs were isolated using the PicoPure RNA Isolation Kit (Arcturus), according to the manufacturer’s instructions. After reverse transcription, the resulting materials were used for quantitative (q)RT-PCR amplification using gene-specific primer pairs and SYBR Green PCR master mix (Applied Biosystems, Carlsbad, CA).

β-IRWT and β-IRKO cells were established from mice (2,19,20). β-IRWT, β-IRKO, MIN6, β-TC6, and α-TC1 cells were maintained in DMEM with 10% FBS (Invitrogen). CHO-K1 and CHO-IR (CHO-K1 cells stably transfected with IR) cells were maintained in DMEM/F-12 with 10% FBS. Human neuroblastoma BE(2)-M17 cells were maintained in OPTI-MEM with 10% FBS. Transfections of small interfering (si)RNAs (Santa Cruz Biotechnology, Santa Cruz, CA) for Gα_i3 and CB1R and the expression vectors for CB1R, Gα_i3, and IR were done using lipofectamine RNAiMAX or 2000 (Invitrogen). Scramble siRNA (Silencer Negative Control #1; Ambion, Austin, TX) or empty vector was transfected as the negative control. For exogenous insulin treatment, cells starved overnight in low-glucose DMEM containing 0.1% FBS were incubated for 2 h in glucose-free DMEM with 0.1% BSA, followed by pretreatment with CB1R agonists for 15 min and insulin stimulation for 10 min, with or without CB1R agonists. For cell proliferation studies, cells were plated into 96-well plates with complete medium. After treatment with CB1R antagonist or agonists, or both, the proliferation rate was determined after 48 or 72 h using the CellTiter 96 Aqueous One Solution Cell Proliferation Assay (Promega, Madison, WI), according to the manufacturer’s instructions.

Cell lysates extracted using radioimmunoprecipitation assay buffer (RIPA) containing protease and phosphatase inhibitor cocktails were incubated with the appropriate antibody and subsequently incubated with protein A/G beads. Beads were washed three times with RIPA buffer and underwent Western blot analysis with the primary antibodies and with an HRP-conjugated secondary antibody. Blots were visualized by enhanced chemiluminescence (GE Health, Berne, Switzerland). For Gα_i activation assay, we used the Gα_i Activation Assay Kit (NewEast Biosciences, Malvern, PA) as previously described (21). Cells treated with or without ACEA were lysed with ice-cold 1×Assay/Lysis buffer and underwent immunoprecipitation with anti-Gα_i–guanosine triphosphate (GTP) antibody and Western blot analysis with anti-Gα_i3 antibody. For positive control, β-IRWT cell lysate, not treated with ACEA, was incubated with guanosine 5'-O-[γ-thio] triphosphate (GTP-γS) for 90 min before immunoprecipitation.

Quantitative data are presented as the mean ± SEM. Differences between mean values were compared statistically by Student t test. Comparisons were performed by using GraphPad Prism (GraphPad Software, La Jolla, CA). A P value of < 0.05 was considered statistically significant.

CB1Rs are present in β- and α-cells in both mouse (Fig. 1A) and human (Fig. 1B) islets. To confirm specificity of staining, we used pancreata from CB1R^−/− mice (16) (Fig. 1A) and the corresponding blocking peptides (Fig. 1A). For added proof of our findings, we microdissected β-cells from islets of CB1R^+/+ and CB1R^−/− mice by laser-capture microscopy, and using qRT-PCR of the captured cells, we confirmed CB1R expression (Fig. 1C). Western blot analysis showed that CB1Rs are expressed in mouse insulinoma (β-TC6 and MIN6) and glucagonoma (α-TC1) cells (Supplementary Fig. 1A). CB2Rs are absent in β- and α-cells, but cellular CB2R staining is evident in islets (Fig. 1D). Of general interest, and as previously reported (10), the CB2Rs are present in somatostatin-containing cells in islets (Fig. 1E).

Also as previously reported (11), EC-synthetic enzymes, N-acyl-phosphatidyl ethanolamine phospholipase D (NAPE-PLD; Fig. 2A) and diacylglycerol lipase-α (DAGLα) (Fig. 2B) are present mainly in β-cells, with little if any expression in α-cells in both human and mouse islets. The EC-degrading enzyme, fatty acid amide hydrolase (FAAH), is present mainly in β-cells (Fig. 2C), and monoacyl glycerol lipase (MAGL) is present in both β- and α-cells (Fig. 2D) in human and mouse islets. In agreement with published literature (9,10), increasing glucose concentrations increased 2-AG and AEA levels in human (Fig. 2E) and mouse (Fig. 2F) islets, and membrane depolarization with KCl also increased 2-AG and AEA levels in human islets (Fig. 2E). Consistently, glucose and KCl also increased 2-AG levels in insulin-secreting immortalized β-cells (β-IRWT) established from mice (2,19,20) (Supplementary Fig. 1B). Thus, these results favor the presence of an entire self-contained ECS in mouse and human islets.

The CB1R^−/− mice are known to have a lean phenotype and are resistant to weight gain, even when fed a high-fat diet, and to have improved glucose tolerance and insulin sensitivity compared with WT mice (22,23). Consistently, treatment with CB1R ligands leads to glucose intolerance and insulin resistance in rodents (23–25) and conversely, peripheral, but not central, CB1R antagonism caused weight-independent improvement in glucose tolerance, insulin sensitivity, fatty liver, and plasma lipid profile even in mice with genetic or diet-induced obesity (23,25,26). These effects are independent of food intake or central CB1R activation (6,23,26).

Their pancreata have not been previously studied. We found that the amount of phosphorylated IR at Y1162/1163 (p-IR), IRS1/2 at Y612 (p-IRS1/2), and AKT at S473 (p-AKT) were all higher in the islets of CB1R^−/− mice compared with age-matched CB1R^+/+ mice (Fig. 3A). Because IR signaling is a key regulator of β-cell proliferation (1–4,27–29), we investigated β-cell area and islet size in CB1R^−/− mice. In keeping with enhanced IR signaling in isolated islets of CB1R^−/− mice, β-cell area (Fig. 3B) and islet size (Fig. 3C), but not pancreas wet weight of CB1R^+/+ (0.216 g ± 0.02 g) versus CB1R^−/− (0.196 ± 0.02 g, n = 6 per genotype, P = 0.48), were increased, and the distribution of islet size (Fig. 3D) was shifted toward bigger islets in CB1R^−/− compared with CB1R^+/+ mice. PCNA, a marker of cell proliferation, was more readily apparent in nuclei of β-cells in CB1R^−/− mice than in CB1R^+/+ mice (Fig. 3E). Ki-67 staining, another marker of cell proliferation, generated similar results (data not shown). Concordant with these observations, CB1R blockade by AM251 also led to increases in β-cell area (Fig. 3F) and PCNA⁺ β-cells (Fig. 3G), compared with DMSO-treated animals.

Genetic and pharmacologic blockade of CB1R in mice with diet-induced obesity results in improved glucose tolerance and insulin sensitivity (23,25,26,30). Because genetic CB1R ablation resulted in larger islets and enhanced β-cell proliferation, despite improved insulin sensitivity, we investigated if CB1R modulation would be beneficial to β-cells in a mouse model of type 2 diabetes. We injected DMSO, AM251, or AM630 (a CB2R antagonist) daily into 4-week-old db/db mice for 4 weeks (Fig. 4A). Consistent with previous reports (23,25,26,30), AM251-treated mice had lower body weight (Fig. 4B) and blood glucose (Fig. 4C) and plasma insulin (Fig. 4D) levels than DMSO-treated mice, whereas AM630 had no obvious effects (Fig. 4B–D). Pancreatic sections showed increased intraislet insulin content and total β-cell mass in AM251-treated mice compared with DMSO-treated mice (Fig. 4E–G), which was most likely due to enhanced β-cell proliferation (Fig. 4H).

Given that AKT regulates p27^Kip1 (p27) activity, an inhibitor of cell cycle progression, by affecting its abundance and subcellular localization (31,32), and that accumulation of p27 in the nuclei of β-cells contributes to deficient β-cell mass and proliferation during the development of type 2 diabetes in Irs2^−/− and db/db mice (28), we examined the expression and subcellular localization of p27. Immunostaining of pancreatic sections from AM251-treated mice showed a significant decrease in the total amount and nuclear localization of p27 in β-cells, with most of the protein being localized in cytoplasm, compared with DMSO-treated mice (Fig. 4I). To evaluate whether the increased β-cell proliferation seen by CB1R antagonism associates with any changes in IR signaling, we examined phosphorylation levels of IR. The amount of p-IR was significantly increased in AM251-treated mice compared with DMSO-treated mice (Fig. 4G).

Because pharmacologic and genetic blockade of CB1Rs led to enhanced IR signaling and β-cell proliferation, we next investigated the potential role of IRs as a/the mediator of CB1R-controlled β-cell proliferation. We used β-cells established from control (β-IRWT) and β-cell–specific IR knockout (β-IRKO) mice (2,19,20) in which expression levels of CB1Rs are similar (Fig. 5A). The selective synthetic CB1R agonist ACEA slowed proliferation of β-IRWT cells that was prevented by AM251 (Fig. 5B).

As previously reported (19), β-IRKO cells already had significantly reduced proliferation rates compared with β-IRWT cells (Fig. 5C). The proliferation rate of β-IRWT was increased by AM251 (Fig. 5D) and decreased by ACEA (Fig. 5E), whereas both compounds had no or a lesser effect on β-IRKO cells (Fig. 5D and E). Consistently, ACEA dose-dependently decreased levels of p-IRS1/2, p-AKT, and p-FoxO1 (Fig. 5F) and increased p27 expression (Fig. 5A) in β-IRWT cells, but not in β-IRKO cells. CB1R blockade by AM251 in normal mice also resulted in a significant decrease in both the amount and nuclear localization of p27 (Fig. 5G) compared with DMSO-treated animals. These results suggest that CB1R signaling functions as a negative regulator of the proliferative effects of endogenously secreted insulin from β-cells.

To test the direct effect of CB1R activation on IR activity, isolated mouse islets (Fig. 6A) and β-IRWT cells (Fig. 6B) were pretreated with AEA or 2-AG and ACEA, respectively, before addition of exogenous insulin. All agonists markedly prevented exogenous insulin-stimulated IR autophosphorylation and downstream signaling. This was also true in BE(2)-M17 human neuroblastoma cells (data not shown), which express CB1Rs (Supplementary Fig. 1A). In addition, ACEA diminished exogenous insulin-stimulated IR autophosphorylation and levels of p-IRS1/2, p-PDK1, p-AKT, and p-FoxO1 in MIN6 cells (data not shown). The inhibitory effects of CB1R agonists on IR autophosphorylation and p-AKT were not observed in CHO cells (CHO cells lack CB1Rs, see Supplementary Fig. 1A; Fig. 6C), but ACEA treatment in CHO-IR (CHO cells stably transfected with IR) cells transfected with the expression vector containing GFP-HA–tagged CB1R led to reduced IR autophosphorylation (Fig. 6D).

To directly confirm the inhibitory effects of CB1R on IR autophosphorylation, we transfected Flag-tagged IR-WT and IR mutant (IR-3YA), whose Tyr1158/1162/1163 residues were substituted to Ala (Fig. 6E), into β-IRKO cells and then treated the cells with ACEA for 40 h (Fig. 6F). IR autophosphorylation was detected in IR-WT–transfected β-IRKO cells, presumably due to endogenous insulin secretion, and ACEA reduced levels of p-IR and p-AKT in those cells, but in IR-3YA–transfected β-IRKO cells, p-AKT levels were not altered (Fig. 6F). In sum, these results suggest a functional interaction between CB1R and IR signaling upon IR autophosphorylation.

Activated CB1Rs mediate their intracellular actions through Gα_i proteins (5). Gα_i3 appears to be expressed mainly in β-cells of both human and mouse (Fig. 7A), and its activity was increased by CB1R activation in both β-IRWT and β-IRKO cells (Fig. 7B). ACEA dose-dependently increased the association between Gα_i3 and IR (Fig. 7C), whereas siRNA-mediated attenuation of CB1Rs (Supplementary Fig. 2A) reduced the association (Fig. 7C). Furthermore, exogenous insulin treatment led to a decrease in Gα_i3/IR association, which was impeded by ACEA (Fig. 7D).

We further confirmed an EC-mediated Gα_i3/IR association using IR-WT and IR-3YA–transfected β-IRKO cells with or without ACEA treatment for 15 min. As shown in Fig. 6E, ACEA reduced levels of p-IR and p-AKT in IR-WT–transfected β-IRKO cells, but p-AKT was not reduced in IR-3YA–transfected β-IRKO cells (Fig. 7E), and most interestingly, Gα_i3 association with IR-3YA was actually increased compared with IR-WT (Fig. 7F and G). ACEA increased Gα_i3/IR association (Fig. 7F), which conversely was reduced by AM251 (Fig. 7G). These results suggest that Gα_i mediates the inhibitory effect of CB1R activation on IR autophosphorylation by its association with IR. Furthermore, over-expression of Gα_i3 in β-IRWT cells led to decreased levels of p-IR and p-IRS1/2 (Fig. 8A) and knockdown of CB1Rs (Supplementary Fig. 2A) or Gα_i3 (Supplementary Fig. 2B) by siRNA in β-IRWT cells abolished the ability of ACEA to inhibit insulin-stimulated phosphorylation of IR, IRS1/2, and AKT (Fig. 8B) as well as its ability to increase p27 expression (Fig. 8C). Consistently, knockdown of CB1Rs or Gα_i3 resulted in an increase in β-cell proliferation (Fig. 8D) and a loss of the inhibitory actions of ACEA (Fig. 8E). These results suggest that the CB1R-mediated effects on IR and downstream signaling involve EC-induced activation of Gα_i.

In contrast to adipose tissue, liver, and muscle, the presence and effects of CB1Rs in β-cells have been inconclusive. Here, we demonstrate that β-cells contain all of the components of a self-contained ECS: CB1R, the necessary enzymes for catalyzing EC biosynthesis and degradation, and the capacity to generate ECs in response to glucose stimulation and depolarization, even when isolated from the pancreas. We used several antibodies (33,34), including the L15 antibody to the C-terminus of CB1R that exhibits the expression pattern most consistent with in-situ hybridization (33) as well as qRT-PCR of laser-captured β-cells from islets, to affirm that CB1Rs, but not CB2Rs, are present on β-cells of mice and men. Bermúdez-Silva et al. (14) have very recently written that using the L15 antibody they also found a CB1R signal with insulin costaining; after using a commercially available CB1R antibody, they reported that CB1Rs were mainly expressed in α-cells (10). CB2Rs are absent from α- and β-cells. Most reports, including this study, found that β-cells contain EC synthetic and degrading enzymes (9–11,14).

Since at least 1979, insulin mediators, also referred to as insulin second messengers, are known to be generated from lipid precursors present on plasma membranes in response to IR activation and consequent downstream phospholipase activation (35). We are now suggesting that ECs, also generated from lipid precursors in β-cells, influence IR activation (Supplementary Fig. 3). This is especially relevant because insulin concentrations would be expected to be at their highest levels surrounding β-cells and therefore the islet ECS potentially evolved to prevent an over-exuberant β-cell IR signaling cascade. Favoring this view are our findings that in various β-cell lines, isolated islets and a mouse model of diabetes CB1R signaling counteracts the effects of insulin on β-cells by preventing IR autophosphorylation and downstream signals. This finding was not unique to pancreatic β-cells because activation of CB1Rs also impeded exogenous insulin-stimulated IR autophosphorylation in non-insulin-secreting cells.

We also found that Gα_i, which is involved in the regulation of insulin secretion (36) and β-cell proliferation (37), mediates the inhibitory effect of CB1R activation on IR activity by its association with IR. CB1R activation increased Gα_i3 activity and Gα_i3/IR association was strengthened by CB1R activation and substitution of Tyr1158/1162/1163 residues of IR with Ala, which, conversely, was weakened by suppression of CB1R activity and by insulin. Furthermore, knockdown of Gα_i3 by siRNA abolished the ability of CB1R to inhibit exogenous insulin-stimulated IR autophosphorylation and β-cell proliferation.

These results suggest a functional and physical crosstalk between CB1R and IR signaling upon IR autophosphorylation in a Gα_i-dependent manner. Given that binding of insulin to the extracellular α-chains of IR causes a change within the quaternary structure of IR that places the phosphorylation sites of one β-chain within reach of the active site of the other β-chain and that results in autophosphorylation at Tyr1158/1162/1163 residues in the activation loop of the β-chains (38,39), we propose that Gα_i3 activated by CB1R associates with unphosphorylated IR at the Tyr1158/1162/1163 residues, preventing a conformational change that secures the activation loop in a catalytically competent configuration upon ligand binding.

Because these receptors are found to be present within caveolae, a cholesterol-rich microdomain that performs a number of signaling functions (40), it is possible that the closeness of the CB1Rs causes them to be involved in modulating IR-mediated signaling. Although the detailed molecular mechanism underlying CB1R and Gα_i3 as regulatory components in the IR signaling pathway awaits further exploration, collectively, our results imply that the alteration in IR activity by CB1Rs is a reflection of a direct inhibition of IR autophosphorylation in a Gα_i-dependent manner and that ECs directly regulate proliferation through activation of CB1Rs expressed in β-cells. Through these actions, CB1Rs are likely to set a threshold level for IR-mediated responses, which depends on the level of expression or activation of CB1R. Alteration in IR activity by CB1Rs may additionally be due to the change in autocrine activation because of altered insulin secretion. CB1R-mediated suppression of insulin secretion in a Ca²⁺-dependent manner has been reported (12,41); however, there are also reports to the contrary (9,10,15,42,43).

We demonstrate the therapeutic advantage of CB1R modulation in a type 2 diabetic condition. Inhibition of CB1R activity in db/db mice led to reduced blood glucose and increased β-cell proliferation, coupled with enhanced IR signaling. There is also evidence that insulin itself reduces glucose-stimulated EC synthesis in β-cells, which would serve as a negative feedback loop to reduce intraislet EC levels (9). This would logically mean that when IR function is reduced, as in type 2 diabetes, such a robust feedback would also be impaired, leading to nonphysiologic EC levels in islets (in addition to in fat and liver) and consequent CB1R-mediated β-cell dysfunction through further impeding IR activity. Blocking CB1Rs would therefore be expected to improve β-cell function in db/db mice, as we found. CB2R antagonism had no such effects.

An inadequate expansion of β-cell mass or failure of the existing β-cell mass to compensate for the changing insulin demand are hallmarks of type 2 diabetes, and these prominent features may result from defective IR signaling (1–4,27–29). Therefore, our data should result in the resumption of attention being paid to ECS as a key factor in β-cell physiology and may lead to development of a new therapeutic strategy aiming to preserve better functioning β-cells.

EC levels, not only in the circulating blood but also in the pancreas, are said to be elevated in diabetes and obesity (9,11,44,45), and elevated EC levels are associated with increased DAGLα and decreased FAAH levels in β-cells (11). Thus, it is possible that increased EC tone (due to increased EC synthesis, receptor expression or activity) affects the well-described glucose-unresponsiveness of β-cells and the development of insulin resistance by impeding IR autophosphorylation in insulin-sensitive tissues. Indeed, AEA was recently found to impair insulin-stimulated AKT phosphorylation and decrease glucose uptake in skeletal muscle cells (7), and CB1R antagonism enhanced insulin responsiveness of skeletal muscle (8).

In addition, pharmacologic blockade of CB1R in obese fa/fa Zucker rats decreased blood glucose levels and preserved β-cell mass (46), and eliminating CB1Rs in liver protected against fatty liver and improved glucose tolerance and insulin sensitivity in high-fat diet–fed mice (23); IR function in those mice was not investigated. Peripheral, but not central, blockade of CB1R was recently reported to improve overall insulin sensitivity and glucose homeostasis (26) and a non-brain-penetrant CB1R antagonist improved glucose homeostasis, insulin sensitivity, and fatty liver in a weight-independent manner (25). This is a very important point, because a centrally acting CB1R antagonist, rimonabant, used for treating obesity, was removed from patient use because of potentially life-threatening psychiatric problems (47). Therefore, CB1R antagonists with poor brain penetrance might be useful therapies in type 2 diabetes where they would be expected to lessen insulin resistance in skeletal muscle and liver, ameliorate or prevent fatty liver, and improve β-cell function/proliferation.

This work was supported by the Intramural Research Program of the National Institute on Aging (NIA)/NIH. R.N.K. is supported by NIH R01-DK-67536 and 68721. M.E.D. is supported by JDRF 26-2008-864.

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

W.K. designed and performed experiments, contributed to discussion, and wrote, reviewed, and edited the manuscript. M.E.D. designed and performed some experiments and provided advice and reagents. Z.L. performed some experiments and analyzed data. Q.L. designed and performed some experiments and provided advice and reagents. Y.-K.S., O.D.C., and H.S.K. performed some experiments and analyzed data. S.T. and J.K.N. analyzed data. E.K.L. contributed to the design of some experiments, the interpretation of data, and discussion. R.M. and Y.W. performed some experiments and analyzed data. S.M. and B.M. contributed to the design of some experiments, the interpretation of data, and discussion. R.N.K. provided advice and reagents and contributed to the interpretation of data and discussion. J.M.E. designed and performed experiments, contributed to the interpretation of data, and wrote, reviewed, and edited the manuscript.

The authors thank Dr. K. Mackie (Indiana University) for specific CB1R (L15 antibody, in particular) and DAGLα antibodies, and for the expression vector containing GFP-HA-tagged rat CB1R cDNA, and Dr. Benjamin Cravatt (Scripps Research Institute) for providing FAAH^−/− mice. The authors are grateful to the staff of the Bioanalytical Chemistry Section/NIA (chief, Dr. I. Wainer) for developing the methodology for quantifying ECs in islets, to the staff in the Confocal Microscopy Unit/NIA (chief, Dr. F. Indig), and to the Histology Core/NIA (Dr. M. Xu, in particular). Dr. J. Pickel, NIMH Transgenic Core/NIH, provided the CB1R^−/− mice and the animal facilities of NIA/NIH and carried out the genotyping and husbandry. The authors thank Dr. R. Broch (NIA/NIH) for discussions of technical procedures and manuscript readings. The ongoing support and encouragement of our Scientific Director, Dr. D. Longo, were vital for all stages of this work.