CB1 Cannabinoid Receptors Couple to Focal Adhesion Kinase to Control Insulin Release*

Background: Endocannabinoids can affect pancreatic β cell physiology. Results: Anandamide and 2-arachidonoylglycerol binding to CB1 receptors induces focal adhesion kinase phosphorylation, which is a prerequisite of insulin release. Conclusion: Focal adhesion kinase activation downstream from CB1 receptors couples cytoskeletal reorganization to insulin release. Significance: Identifies the molecular blueprint of 2-arachidonoylglycerol signaling in the endocrine pancreas, and outlines a kinase activation cascade linking endocannabinoid signals to insulin release. Endocannabinoid signaling has been implicated in modulating insulin release from β cells of the endocrine pancreas. β Cells express CB1 cannabinoid receptors (CB1Rs), and the enzymatic machinery regulating anandamide and 2-arachidonoylglycerol bioavailability. However, the molecular cascade coupling agonist-induced cannabinoid receptor activation to insulin release remains unknown. By combining molecular pharmacology and genetic tools in INS-1E cells and in vivo, we show that CB1R activation by endocannabinoids (anandamide and 2-arachidonoylglycerol) or synthetic agonists acutely or after prolonged exposure induces insulin hypersecretion. In doing so, CB1Rs recruit Akt/PKB and extracellular signal-regulated kinases 1/2 to phosphorylate focal adhesion kinase (FAK). FAK activation induces the formation of focal adhesion plaques, multimolecular platforms for second-phase insulin release. Inhibition of endocannabinoid synthesis or FAK activity precluded insulin release. We conclude that FAK downstream from CB1Rs mediates endocannabinoid-induced insulin release by allowing cytoskeletal reorganization that is required for the exocytosis of secretory vesicles. These findings suggest a mechanistic link between increased circulating and tissue endocannabinoid levels and hyperinsulinemia in type 2 diabetes.

The "storage-limited model" of insulin release (22) recognizes two pools of insulin granules that undergo temporally distinct waves of exocytosis. Its first, the Ca 2ϩ transient-dependent phase involves the rapid fusion of the readily releasable pool of granules at the plasma membrane. Its second phase requires the replenishment of released vesicles through the trafficking of reserve insulin granules to positions adjacent to the cell membrane. This replenishment, however, is impeded by the dense web of filamentous actin (F-actin) present in the subplasmalemmal compartment under non-stimulated conditions. Therefore, stimulus-dependent cytoskeletal remodeling is needed to eliminate the F-actin barrier. However, instead of merely disrupting or degrading the pre-existing F-actin barrier (23), actin polymerization in stress fibers is stabilized at focal adhesion (FA) plaques containing e.g. paxilin, talin, and vinculin and serving to route insulin granules toward the plasma membrane (24). This mechanism rests on focal adhesion kinase (FAK) whose activity regulates FA remodeling and turnover (25,26), and tension signaling (27), facilitating glucose-stimulated insulin secretion (28). This concept is consistent with biphasic insulin release because second phase transport of insulin granules requires F-actin as a motive force (29,30).
It is appealing to link cannabinoid receptor activation to the second phase of insulin secretion because CB 1 Rs can activate FAK (31). Moreover, both CB 1 R and GPR55 in neurons or malignant cells (32)(33)(34) can influence Rho family GTPases, which by interacting with FAK (35), furnish a cellular microenvironment facilitating secretory vesicle exocytosis. However, the contribution of FAK signaling downstream from cannabinoid receptors to regulated insulin secretion from pancreatic ␤ cells remains unknown. Here, we hypothesized that agonist activation of cannabinoid receptors can orchestrate a signaling cascade via Akt (or alternatively via extracellular signal-regulated kinases (ERK1/2)) and FAK to trigger F-actin polymerization and FA plaque formation to facilitate second phase insulin release. We have also explored the ligand and receptor specificity and temporal dynamics of eCB-induced insulin release by combining molecular pharmacology and mouse genetics in INS-1E cells and in primary mouse pancreatic islets. Our data suggest that eCBs coordinately activate ERK1/2/Akt and FAK downstream from CB 1 R to render the cytoskeleton permissive for insulin secretion. (36) were cultured at 37°C in RPMI 1640 medium containing glucose (11 mM), HEPES (10 mM), heat-inactivated fetal bovine serum (FBS; 5%), sodium pyruvate (1 mM), ␤-mercaptoethanol (50 M), pen-icillin (50 g/ml), and streptomycin (100 g/ml; all from Sigma). Cells were routinely subcultured in 24-well plates up to passage 120 and allowed to reach ϳ80% confluence. For immunocytochemistry, INS-1E cells were plated on 12-mm coverslips coated with D-polyornithine (0.001%). The effect of the following drugs on insulin release, alone or in combination, was assessed: ACEA (100 nM), AEA (100 nM-10 M), AM 251 (100 nM), capsazepine ( (37). Insulin release was tested after either 30 min or 24 h of drug exposure. Cytoskeletal reorganization was tested after 30 min of stimulation unless otherwise stated.
RNA Isolation and Gene Expression Analysis-Total RNA was isolated from INS-1E cells, rat spleen, cortex, and cerebellum (as tissue-specific positive controls) (38,40,41) using the RNeasy Mini Kit (Qiagen) followed by DNase digestion, and verifying RNA integrity on 2% agarose gels (500 ng of RNA). cDNA was prepared by reverse transcription with random primers using the high-capacity cDNA Reverse Transcription Kit (Applied Biosystems) and PCR amplified (30 cycles) by ratspecific primer pairs (Table 1). PCR products were resolved on 2% agarose gels and imaged on a ChemiDoc XRS ϩ system (Bio-Rad).
INS-1E cells, pancreas, and brain sections were inspected on a Zeiss LSM700 laser-scanning microscope equipped with a ϫ40 water immersion objective using ϫ1.0 -2.5 optical zoom and differential interference contrast when required (Fig. 2). Comparison of null and wild-type tissues was performed when keeping threshold and intensity settings identical. Images were acquired in the ZEN2010 software package. Cytoskeletal remodeling, defined as the CB 1 R-dependent formation of stress fibers and FA plaques, was quantified at ϫ40 primary magnification in n ϭ 30 -80 INS-1E cells/condition. Data from triplicate experiments were statistically analyzed (see below). Multi-panel images were assembled in CorelDraw X5 (Corel Corp.).

Measurement of Monoacylglycerol Lipase-like Activity by 4-Nitrophenylacetate-INS-1E
cells were treated with increasing concentrations of JZL 184 or vehicle for 24 h. Then, cells were washed, scraped, and sonicated in 50 mM sodium phosphate buffer (pH 8.0) containing 0.3 M sucrose (ϳ40 s) at 4°C. Samples were then centrifuged at 100,000 ϫ g at 4°C for 50 min. Pellets containing the cell membrane fraction were collected, re-suspended in 10 mM Tris-HCl (pH 7.4), and stored at Ϫ80°C until use. MGL-like activity was assayed in a 96-well plate format (microtiter plates, 200 l total volume) as described (47). Samples (4 g of protein/well in 150 l of Tris-HCl (pH 7.4) also containing 0.1% fatty acid-free BSA (Sigma)) were aliquoted and mixed with JZL 184 (200 or 500 nM) or vehicle, the latter used as controls. 4-Nitrophenylacetate (Sigma), as substrate, was dissolved in 250 mM Tris-HCl (pH 7.4) and added to each well. MGL-like activity was determined by measuring the rate of 4-nitrophenylacetate hydrolysis at 10-min intervals in a colorimetric assay on an Infinite M1000 PRO microplate reader (Tecan) tuned at 405 nm at 37°C. Samples containing buffer only were used as blanks.
Liquid Chromatography-Atmospheric Pressure Chemical Ionization-Mass Spectrometry-INS-1E cells were treated with JZL 184 (200 nM) for 30 min or 24 h (Fig. 2F). Cell-free supernatants were separated and flash-frozen in liquid N 2 . Cells were scraped in 2 ml of ice-cold methanol and kept at Ϫ20°C until processing for liquid chromatography-atmospheric pressure chemical ionization-mass spectrometry. Extraction, purification, and quantification of 2-AG and AEA followed published protocols (5,48). After lipid extraction and pre-purification on silica gel columns, endocannabinoid levels from 2-4 independent samples/condition were analyzed by isotope dilution using liquid chromatography-atmospheric pressure chemical ionization-mass spectrometry (Shimadzu LCMS-2020).
Cell Viability and Proliferation Assays-To assess cell viability, INS-1E cells were stimulated with AEA (10 M) or vehicle for 24 h. Cell Titer Blue Reagent (Promega), diluted 1:10 in the culture medium, was added during the last 3 h of the incubation period. Media were then transferred onto 96-well plates and their ϭ 570/590 nm absorbance ratio measured on an Infinite M1000 PRO microplate reader (Tecan). The absorbance of the   CB 1 R Ϫ/Ϫ and wild-type pancreatic islets were plated in 24-well plates (10 islets/well) and incubated in KRBH for 1 h. Next, islets were incubated in KRBH supplemented with either 2.75 or 16.5 mM glucose to which ACEA (100 nM) or vehicle had been added at 37°C for 30 min.
KRBH buffers from basal conditions and after stimulation were recovered to determine secreted insulin levels. The amount of insulin in the incubation buffers was measured by a rat/mouse Insulin ELISA Kit (Millipore) as per the manufacturer's recommendations. Insulin concentrations were normalized to the total protein content of INS-1E samples or to the number of stimulated islets.
Protein Measurement-Protein concentrations were determined by the Bradford method (Bio-Rad) with optical density of the total protein content measured at 595 nm. Protein concentrations were calculated against BSA standards.
Statistics-Phosphoprotein levels were normalized to those of the respective non-phosphorylated ("total") proteins. Experiments were performed in triplicate unless stated otherwise. Data were expressed as mean Ϯ S.D. p Ͻ 0.05 was considered statistically significant. Data were analyzed using one-way ANOVA followed by pairwise comparisons when appropriate. Student's t test (independent group design) was used to statistically analyze data on immunofluorescence intensities. Correlation coefficients were calculated by Pearson's test.

Molecular Organization of Endocannabinoid Signaling Networks in INS-1E Cells-For rat ␤ cell-derived INS-1E cells to
serve as adequate tools to study the relationship of eCB signals and insulin release, they must recapitulate signaling arrangements seen in vivo (4). Particularly, because the complexity of eCB signals in whole islets (50) may be vastly different from that seen in functionally segregated cell types. Therefore, we first profiled the presence of relevant mRNAs by PCR, and validated our results against reference tissues (38,41,46).
In INS-1E cells, both CB 1 R and DAGL␣ were abundant at the plasmalemma, and also filled most of the cytoplasm and processes ( Fig. 1, C and C 1 ). MGL immunoreactivity appeared punctate in both the perisomatic compartment and processes ( Fig. 1C 2 ). Given our primary focus on 2-AG signaling, we also verified the specificity of CB 1 R, DAGL␣, and MGL immunosignals in pancreas and/or nervous tissues from respective null mice ( Fig. 1, D-G 1 ).
We suggest that AEA signaling might also operate in INS-1E cells because both mRNA (Fig. 1H) and protein (Fig. 1I) of the enzymes NAPE-PLD and FAAH, partaking in AEA biosynthesis and degradation, respectively (9, 52), were detected.

CB 1 R Activation Promotes DAGL␣ Expression in INS-1E
Cells-Presence of the molecular constituents required for both 2-AG and AEA signaling prompted us to test whether agonist stimulation of cannabinoid receptors modulates the "2-AG signaling cassette" (that is, metabolic enzymes and receptors). In testing cell-or tissue-specific eCB signal specificity we focused on 2-AG because the enzymes required for its metabolism have been more unequivocally identified (10,53). We exploited direct and indirect stimulus protocols: first, in pharmacological protocols we applied exogenous AEA for 30 min or 24 h, whose (ϳ12 h) half-life is compatible with pro-longed stimulation in vitro (54). Second, we inhibited MGL activity by JZL 184 (55) ). B, Western analysis of antibodies used to reveal minimal requirements of 2-AG signaling. C-C 2 , CB 1 R, MGL, and DAGL␣ localization in INS-1E cells combining fluorescence and differential interference contrast (DIC) microscopy. D-G 1 , histochemical controls demonstrating the specificity of rabbit anti-CB 1 R (D and D 1 ), guinea pig anti-CB 1 R (E and E 1 ), anti-MGL (F and F 1 ), and guinea pig anti-DAGL␣ (G and G 1 ) in the pancreas (D and D 1 ), hippocampus (E and F 1 ), or cerebellum (G and G 1 ) of adult wild-type and respective null mice (38,42,43). The lack of residual immunolabeling in knock-out tissues unequivocally supports the specificity of the antibodies used. Hoechst 33,342 was used as nuclear counterstain. mRNA (H) and protein (I) detection for NAPE-PLD (N-PLD) and fatty acid amide hydrolase (FAAH), involved in anandamide metabolism (4,9), in INS-1E cells. Arrows in B and I point to immunoreactive protein bands at the calculated molecular weight of each target. Abbreviations: mol, molecular layer of the cerebellum; PC, Purkinje cell layer; pyr, pyramidal layer of the CA1 hippocampal subfield. Scale bars ϭ 100 m (D 1 ), 15 m (C-C 2 ), and 10 m (E 1 and F 1 ).
nificantly decreased MGL availability after 2 and 24 h (p Ͻ 0.05; Fig. 2, A and D), as measured by quantitative immunocytochemistry. Simultaneous detection of CB 1 Rs revealed the removal of this receptor from the cell surface and accumulation in the perinuclear cytoplasm (Fig. 2D 1 ). This coincided with decreased CB 1 R content (Fig. 2B) suggesting intracellular degradation (56). In contrast, the DAGL␣ protein, which is present at greater levels than DAGL␤ in the endocrine pancreas (57), significantly increased (Fig. 2, C and D 2 ).
Next, we inhibited MGL (by JZL 184, 200 nM) to increase 2-AG availability (55), and to test whether the above molecular rearrangements were selective to AEA. First, we confirmed that JZL 184 rapidly inhibited MGL-like enzyme activity (Fig. 2E), defined as the significantly reduced hydrolysis of 4-nitrophenylacetate in JZL 184-exposed INS-1E cells. Subsequently, we measured 2-AG and AEA concentrations in supernatants and cell pellets of INS-1E cells exposed to JZL 184 for 30 min or 24 h. We found a substantial JZL184-induced increase in extracellular 2-AG content that persisted up to 24 h ("supernatant,"  2G). CB 1 R levels remained reduced after 1 and 24 h of JZL 184 treatment (Fig. 2H). Likewise, the increased DAGL␣ level upon JZL 184 exposure recapitulated AEA effects (Fig. 2I). These data suggest that both "direct" and "indirect" CB 1 R agonism might drive 2-AG synthesis to maintain the efficacy of eCB signaling even if CB 1 Rs are desensitized or degraded. CB 1 R activity can induce apoptosis in pancreatic ␤ cells (57). Therefore, we tested whether AEA affected INS-1E cell survival. AEA did not decrease INS-1E cell viability (Fig. 2J)  184 ϩ O-2050), fold) and ACEA, a selective CB 1 R agonist, failed to induce INS-1E cell proliferation (Fig. 2K).
Agonist-induced CB 1 R Activation Drives Insulin Release from INS-1E Cells-Endocannabinoid signals have been suggested to regulate insulin secretion (5,6,59). However, the precise molecular mechanism of eCB action, including receptor identity and signal transduction sequence, on insulin secretion remains uncertain. AEA rapidly increased basal insulin secretion  Fig. 3A). The effect of AEA (10 M) on basal insulin secretion under glucose-free conditions was time-dependent, reaching quasi-saturation by 30 min (Fig. 3B). Therefore, we used 30-min stimuli throughout. AEA is a pleiotropic ligand, being an agonist at CB 1 R, CB 2 R, and TRPV1 and other receptors and channels (60). Therefore, we first dissected receptor requirements of AEA-induced insulin release. O-2050, a silent CB 1 R antagonist (F (1,12) Fig. 3E). Although being statistically ineffective on its own (p Ͼ 0.2), AM 251, a CB 1 R inverse agonist, also decreased AEA-induced insulin hypersecretion under both basal and glucose stimulatory conditions (F (1,12) ϭ 13.73, p Ͻ 0.001; Fig. 3F), corroborating earlier findings with SR141716 (7). These data, together with the inability of JWH133 (a CB 2 R agonist) to affect insulin release (Fig. 3D), suggest that CB 1 Rs drives the AEAinduced insulin secretion from INS-1E cells and augments the immediacy of insulin release upon stimulation with high glucose.
Vesicular insulin exocytosis is ubiquitously dependent on intracellular Ca 2ϩ signals (22). Therefore, we tested the reliance of CB 1 R-dependent insulin secretion from INS-1E cells on extracellular (bath-applied) Ca 2ϩ . Both basal and glucose-stimulated insulin secretion were inhibited in Ca 2ϩ -free extracellular medium. In contrast, the ability of AEA to significantly augment glucose-induced insulin release remained unaffected (Fig.  4). Because cytoplasmic Ca 2ϩ underpins insulin secretion (61), our data suggest that intracellular Ca 2ϩ mobilization is a candidate mechanism to couple CB 1 R activation to insulin release (5, 59).   NOVEMBER 8, 2013 • VOLUME 288 • NUMBER 45

JOURNAL OF BIOLOGICAL CHEMISTRY 32691
Prolonged Endocannabinoid Exposure Maintains the Readiness of CB 1 R-dependent Insulin Release-Receptor recycling and desensitization, dampening of signal transduction efficacy or metabolic checkpoints of ligand availability can modify the impact of prolonged endocannabinoid exposure on insulin release. Therefore, we first tested whether AEA exposure for 24 h followed by 30 min stimulation with low or high glucose in the presence of AEA could increase insulin release. We showed that AEA significantly augmented both basal (1.41 Ϯ 0.12-fold of control; p Ͻ 0.05) and high glucose-induced insulin release (2.4 Ϯ 0.25-fold of control; p Ͻ 0.05; Fig. 5A) even if present for 24 h prior to probing insulin release. Second, we addressed whether modifying endogenous 2-monoacylglycerol availability by MGL inhibition affected insulin secretion. JZL 184 treatment for 24 h significantly increased both basal and glucosestimulated insulin release (F (1,8) Fig. 5B). Besides inhibiting MGL, JZL 184 dose dependently can affect "off targets." Therefore, we used OMDM 188, a DAGL inhibitor (37), to test 2-AG involvement. OMDM 188 application for 24 h alone did not affect basal insulin release (Fig. 5C). In contrast, OMDM 188 alone (24 h) occluded insulin secretion brought about by high glucose (overall effect: F (1,8) ϭ 57.77, p Ͻ 0.001), suggesting that DAGL activity is required for insulin release. Moreover, OMDM 188 also counteracted the effect of JZL 184 on glucose-induced insulin release (F (1,8) ϭ 13.06, p ϭ 0.002; Fig. 5C), with combined treatment decreasing insulin secretion to the level observed in low glucose. These data suggest that high glucose increases 2-AG synthesis in ␤ cells to tonically drive insulin release, which agrees with a previous finding in rat RIN-m5F cells (5).

2-AG Signaling Networks in and CB 1 R-dependent Insulin
Release from Pancreatic Islets-We affirmed the physiological relevance of the above findings by histochemical profiling of the distribution and cell type-specificity of DAGL, MGL, and ABHD6 in relationship to CB 1 R localization and cell type diversity in the adult mouse pancreas. First, we confirmed (4) that mouse insulin ϩ (␤) cells expressed CB 1 Rs, which often appeared membrane-bound (Fig. 6A). In contrast, glucagon ϩ (␣) cells remained mostly unlabeled (Fig. 6, A 1 and A 2 ). Both insulin ϩ and glucagon ϩ cells expressed DAGL␣ (Fig. 6, B and B 2 ). Unexpectedly, MGL concentrated in glucagon ϩ cells with only low to moderate MGL immunoreactivity seen in glucagon Ϫ cells, including ␤ cells (Fig. 6C). Instead, ABHD6 appeared as a major 2-AG-degrading enzyme in glucagon Ϫ cells (Fig. 6D) including ␤ cells (Fig. 6D 1 ) and pancreatic polypeptide ϩ F cells (62) (Fig. 6D 2 ). In contrast, somatostatin ϩ ␦ cells (62) appeared as ABHD6 Ϫ (Fig. 6D 3 ). MGL is highly upregulated in malignant cells (63). Therefore, immortalized INS-1E cells might favor MGL expression. Taken together, these data suggest that both autocrine and paracrine 2-AG signaling can occur in ␤ cells via CB 1 Rs. In fact, genetic disruption of CB 1 R expression prevented ACEA-stimulated insulin secretion from native pancreatic islets (p Ͻ 0.05; Fig. 6E).
Focal Adhesion Kinase Activity Links Endocannabinoid Signaling to Insulin Release-AEA (15 min) did not impair actin polymerization and/or stability per se, as indicated by the unaltered G/F-actin ratio in AEA-exposed INS-1E (Fig. 9A). Thus, and considering the formation of vinculin ϩ FA plaques upon CB 1 R activation, we hypothesized that FAK, a tyrosine kinase, might be poised to mediate the interaction between integrins and actin cytoskeleton in response to CB 1 R activation (31). FAK, also activated by growth factors (27) and glucose (26), participates in the multimolecular complex making up focal contact sites (FA plaques) to transduce extracellular signals to i.e. initiate the exocytosis of insulin granules (26). Therefore, FA plaque formation upon CB 1 R stimulation (Fig. 7D) is well suited to link cytoskeletal remodeling to insulin release. FAK phosphorylation (Ͼ10-fold increase versus control) occurred within 30 min of AEA stimulation (Fig. 9B), was O-2050 sensitive, and mimicked the effect of high glucose (Fig. 9B). This FAK phosphorylation predominantly occurred at the tips of processes protruding from INS-1E cells, which provided adhesion points as judged by them being vinculin ϩ (Fig. 9, C and C 1 Ј). The AEA-induced concentration of vinculin ϩ /phospho-FAK ϩ FA plaques served to anchor the stress fibers (Fig. 9C 1 Ј). The over-FIGURE 6. CB 1 R, DAGL␣, MGL, and ABHD6 distribution in the endocrine pancreas of mouse. A and A 2 , CB 1 R immunoreactivity was seen in insulin ϩ ␤ cells, but not glucagon ϩ ␣ cells, in pancreatic islets. Arrows point to membrane-associated CB 1 Rs. B and B 2 , both insulin ϩ (␤) and glucagon ϩ (␣) cells harbored DAGL␣ immunoreactivity. Arrows in B 1 , inset, denote DAGL in glucagon ϩ ␣ cells. C and D, generally, MGL and ABHD6 immunoreactivity was moderate in pancreatic islets of adult mice. C, MGL preferentially localized to glucagon ϩ ␣ cells (arrows). Inset depicts the co-existence of cytoplasmic MGL and insulin immunoeactivities (arrow). D, ABHD6 was found in both glucagon ϩ ␣ cells (arrows) and glucagon Ϫ cells (open arrowheads). Of the glucagon Ϫ cells, insulin ϩ ␤ cells (D 1 and D 1 Ј) and pancreatic polypeptide ϩ F-cells (D 2 ) but not somatostatin ϩ ␦ cells (D 3 ) harbored ABHD6 immunosignal. E, pancreatic islets isolated from CB 1 R Ϫ/Ϫ and wild-type mice were challenged with ACEA (30 min) to release insulin. Note that ACEA failed to augment glucose-induced insulin release in CB 1 R Ϫ/Ϫ islets. Hoechst 33,342 was used as nuclear counterstain. n.s., non-significant. Scale bars ϭ 60 m (A 2 and B 2 ), 18 m (C, D, and D 3 ), 9 m (C/inset, D 1 Ј, and D 2 ). Data in E were normalized to control in low glucose, and expressed as mean Ϯ S.D. fold-values. *, p Ͻ 0.05 versus control (pairwise comparisons after one-way ANOVA are shown). NOVEMBER 8, 2013 • VOLUME 288 • NUMBER 45

JOURNAL OF BIOLOGICAL CHEMISTRY 32693
all phospho-FAK immunoreactivity of AEA-treated INS-1E cells markedly increased (Fig. 9C 1 ), recapitulating the results of our Western analysis on total FAK levels upon AEA exposure (Fig. 9B). O-2050 reduced phospho-FAK immunoreactivity, the density of vinculin ϩ FA plaques and their association with stress fibers (Fig. 9, C 2 and C 2 Ј).

DISCUSSION
Our report identifies CB 1 R-mediated signaling events facilitating insulin release from INS-1E cells, immortalized rat ␤ cells, and native mouse ␤ cells. These findings highlight FAK activation as a key step in eCB-induced insulin release. Moreover, and although INS-1E and native pancreatic ␤ cells reportedly express a variety of (endo-) cannabinoid-sensing receptors (3,16,59), eCB-stimulated insulin release appears to rely, in a large part, on CB 1 Rs. Impaired energy balance is considered to be the main cause for obesity and type 2 diabetes (68,69). For energy balance to be kept within a physiologically neutral range, the tight regulatory interplay of major centers, including the nervous system, gastrointestinal tract, pancreas, adipose tissue, liver, and skeletal muscles has evolved to continuously monitor nutrient and energy status at the periphery. Integration of vital peripheral information requires long-range communication among various organ systems. eCB signaling at CB 1 Rs has recently emerged as a particularly efficacious signaling network to control central aspects of food intake and energy homeostasis (68). Similarly, eCB signaling in insulin-sensitive peripheral tissues impacts glucose utilization and lipid homeostasis (70,71). Despite mounting interest (3,4,6,59,68), the molecular link between eCB signaling and insulin release from the endocrine pancreas remain poorly understood.
Here, by using INS-1E cells and pancreatic islets from CB 1 R Ϫ/Ϫ mice, we demonstrated the upstream role of eCB signals to regulated insulin secretion. We show that INS-1E cells, like native ␤ cells, can co-express CB 1 Rs, CB 2 Rs, and TRPV1 to sense and integrate eCB signals. Yet data from direct (AEA, ACEA) and indirect (JZL 184) agonist pharmacology and CB 1 R Ϫ/Ϫ mice cumulatively demonstrate that CB 1 R activation dominates eCB-stimulated insulin secretion in a dose-and time-dependent manner. eCB action is relatively fast, and promotes adaptive and long-lasting reorganization of eCB metabolism with opposing changes in MGL and DAGL␣ levels in ␤ cells. Decreased CB 1 R levels suggest rapid receptor recycling and degradation. The rapid loss of MGL also suggests protein degradation, which reportedly involves CB 1 R-dependent activation of breast cancer-associated protein 1 (BRCA1) (72,73), an E3 ubiquitin ligase. In contrast, increased DAGL␣ expression suggests compensatory eCB synthesis in response to disrupted second messenger signaling upon CB 1 R desensitization. This notion is supported by increased CB 1 R content in INS-1E cells treated with the DAGL inhibitor OMDM 188, which may compensate for decreased 2-AG availability by potentially conferring receptor hypersensitivity.
Both immortalized INS-1E cells and native mouse ␤ cells express the metabolic enzymes and receptors sufficient for autocrine 2-AG signaling. Nevertheless, differences in experimental designs (e.g. ␣ cells can contribute to 2-AG production/ degradation in isolated pancreatic islets) have limited our abil-ity to determine whether autocrine or paracrine eCB signals regulate insulin secretion. They might also explain differences between AEA and JZL 184-stimulated insulin release from INS-1E cells reported here and largely non-CB 1 R-mediated inhibition of insulin secretion reported elsewhere (6,59). Inconsistencies as to the cellular sites of CB 1 R expression in pancreatic islets exist in the literature (4,6), primarily due to the lack of appropriate genetic controls. We find mouse ␤ cells endowed with CB 1 Rs, confirming data reported by Starowicz et al. (4) using immunoperoxidase methods in rat and mouse pancreas. Notably, ␣ cells do not harbor detectable CB 1 R levels in vivo, suggesting that glucagon secretion might either be insensitive to endocannabinoid signals or modulated via non-CB 1 Rmediated mechanisms. To maintain glucose utilization and homeostasis, however, eCB actions on glucagon levels in blood might be indirect because eCB-driven insulin release from ␤ cells could per se serve to prime glucagon secretion. ␣ Cells, lining the outer border of pancreatic islets, appear to express both MGL and ABHD6 at levels exceeding those of ␤ cells. This raises the possibility that ␣ cells can restrict the spatial spread of eCBs, and isolate the endocrine pancreas from surrounding tissues. This hypothesis is supported by pronounced ABHD6-like immunoreactivity in pancreatic polypeptide-containing F cells. Unexpectedly, we detected marked ABHD6-like immunosignals in ␤ cells. Although the robustness of immunoreactivity FIGURE 9. CB 1 R-mediated insulin release requires focal adhesion kinase activity. A, comparative assessment of G-and F-actin concentrations upon AEA challenge. Optical density of G-and F-actin was measured, and expressed as the "G/F-actin ratio." B, treatment of INS-1E cells with AEA (10 M) for 30 min induced FAK kinase phosphorylation (p-FAK). Fold-changes were normalized to the phosphorylation level in control, and relative to actin. A representative Western blot is shown. C-C 2 , immunofluorescence detection of increased FAK phosphorylation in INS-1E cells exposed to AEA (30 min). Open rectangles denote the positions of insets. Hoechst 33,432 nuclear localization signal was color-coded in yellow. CЈ-C 1 Ј, insets demonstrate the accumulation of p-FAK at vinculin ϩ focal adhesion points, which also anchor actin filaments (stress fibers, solid arrowheads) upon AEA application. O-2050 prevented stress fiber assembly (open arrowheads). D-D 3 , FAK inhibitor 14 (FAKi14; 1 M, 10 min), a small molecule inhibitor of FAK (67), prevented AEA or high glucose-induced vinculin ϩ FA plaque (C 2 ), and stress fiber assembly (C 3 ). E, media were collected and insulin concentrations determined by ELISA. FAKi14 abrogated AEA-induced insulin secretion. Scale bars ϭ 10 m (B-B 2 and C) and 3 m (BЈ-B 2 Ј, and C 1 ). Data were expressed as mean Ϯ S.D.; n ϭ 80 cells/group; **, p Ͻ 0.01; *, p Ͻ 0.05 (pairwise comparisons after one-way ANOVA). F and F 1 , correlation analysis revealed a positive relationship between the number of INS1-E cells containing FA plaques (F) or stress fibers (F 1 ) and the amount of insulin secreted (ng/g protein). Note that particularly close correlation was seen in the presence of high glucose. Pearson's correlation co-efficient are shown (dashed lines: linear regression plots). cannot be taken indicative of enzyme activity, our histochemical results suggest that MGL and ABHD6 might cooperate in terminating eCB signaling in the endocrine pancreas, including ␤ cells. Moreover, our molecular identification of the coexistence of the biosynthesis and degradation machinery for 2-AG (and probably also AEA) highlights that both autocrine and (juxta-)paracrine eCB signaling must be considered in specific cellular contexts when assessing the functionality and influences of these signaling networks in the pancreas (4,70). The notion that DAGL inactivation occludes insulin release, particularly when glucose concentrations are high, suggest the unexpectedly broad metabolic significance of these enzymes, perhaps beyond regulating 2-AG signaling at CB 1 Rs.
AEA in glucose-free conditions induced insulin release (Fig.  3B). This finding supports that glucose and AEA action are additive on insulin release, corroborating earlier reports (6,59). CB 1 R agonists significantly increased insulin release from INS-1E and ␤ cells, by a mechanism involving intracellular Ca 2ϩ mobilization. These data in conjunction with high glucose-induced AEA and 2-AG synthesis (5) suggest a tight feed forward coupling between eCB production and signaling and glucose availability. eCBs can rapidly mobilize intracellular Ca 2ϩ through inositol 1,4,5-trisphosphate signaling (74,75) or by inositol 1,3,4-phosphate receptors on the endoplasmic reticulum (8). Such rapid Ca 2ϩ transients, as seen in insulinomas cells upon AEA stimulation (8), are compatible with a role in first phase insulin release.
A rise of intracellular Ca 2ϩ being a prerequisite of insulin release is also compatible with FAK and FAK-related kinase (76,77) activation and cytoskeletal rearrangements during the second phase of insulin secretion, particularly because vesicle docking during exocytosis is reliant on the Ca 2ϩ -driven assembly of the SNARE machinery. Exocytosis involves a series of highly coordinated and sequential steps. Among these, FAK interacts with integrins, vinculin, paxillin, talin, and ␣-actinin to link FA plaques to filamentous actin. FAK in neurons is a key effector of CB 1 R-induced cytoarchitectural reorganization (31). FAK is an appealing target because it can suppress Rho GTPase activity, thereby promoting FA turnover (35). Considering that CB 1 Rs can modulate the activity of RhoA (and other small GTPases) (34,76), FAK is poised to orchestrate cystoskeletal remodeling, particularly FA plaque assembly, required for insulin release. CB 1 R activation establishes maximal FAK activity upon cooperative signaling with cell surface tyrosine kinase receptors and the sequential phosphorylation of Tyr-397 and Tyr-576/577 FAK residues (76). Here, we showed FAK phosphorylation at Tyr-397, which is considered a permissive step for the recruitment of Src-family kinases, whose second-phase phosphorylation of an additional five Tyr residues in the activation loop of the kinase, confers maximal FAK activity (76). Because CB 1 R-dependent Tyr-397 phosphorylation persists for Ͼ1 h, requires integrin activation, and exhibits adhesion dependence (76), and is abated by FAKi14 (67), we conclude that FAK activation is an essential transducer and integrator of eCB and integrin signaling in pancreatic ␤ cells.
Two of the clinicopathological implications of our findings, i.e. that (i) DAGL activity is a minimal requirement for insulin secretion and (ii) prolonged exposure to heightened eCB concentrations drives insulin release, outline a link between increased circulating eCB levels (5), elevated tissue 2-AG and AEA content, and up-regulated DAGL␣ in ␤ cells (4) and insulin hypersecretion in obesity (51,84). The mechanism we identified suggests that acute CB 1 R activation may be critical for the adaptation of pancreatic ␤ cells to insulin resistance. Our findings provide a molecular backbone to develop tissue selective CB 1 R antagonists, which, given their insulin re-sensitizing effects (71), can offer the rescue of insulin secretion to maintain adequate to tissue demands.