The present study was conducted in adult male mongrel dogs, 1–3 years old. We included a subset of animals herein from a previous study [19] to report more extensive data related to β-cell function in vivo and plasma levels of endocannabinoids. Animals were housed in kennels at the vivarium of the Keck School of Medicine, University of Southern California (Los Angeles, CA). The protocol for this study was submitted to, and approved by, the ethics committee of the University of Southern California, and all procedures followed the regulations from the Institutional Animal Care and Use Committee.

At arrival, animals started a standard diet consisting of 825 g of dry chow (a mixture of Laboratory High Density Canine Diet and Prolab Canine 2000, Richmond, IN) for 2–3 weeks. Diet was switched to a weight-maintaining control diet for 3 weeks consisting of 825 g of dry chow and one canned food (Hill’s Pet Nutrition, Topeka, KS). Total daily food presented contained 3,582 kcal (28.1% from proteins, 31.3% from fat, 40.6% from carbohydrates). After body weight stabilization, animals were fed a hypercaloric high-fat diet (HFD) for 22 weeks. HFD consisted of a control diet enriched with lard/bacon grease (6 g/kg of baseline body weight). Total daily calorie content of HFD presented consisted of 5,527 kcal (53.0% from fat). This HFD regimen has been extensively validated in our laboratory to effectively and reproducibly induce insulin resistance in canines [18–21]. Food was presented from 09:00–12:00 h. Water was provided ad libitum. Since experiments in vitro were conducted at the end of the study, our findings in vitro were compared with those from animals fed a control diet for 4–6 weeks (n = 7). The calorie content of the control diet was estimated based on the personal experience of our veterinary staff on mongrel dogs.

A group of 9 animals were fed a HFD. Body weight, abdominal fat content, β-cell function in vivo, whole-body insulin sensitivity, and biochemical analyses (fasting glucose, insulin, C-peptide, glucagon, plasma non-esterified fatty acids, anandamide, 2-AG, oleoyl ethanolamide, linoleoyl ethanolamide, and palmitoyl ethanolamide), were determined before and after HFD. Further metabolic assessments have been published elsewhere [19, 20].

Fat accumulation in the abdominal region was quantified by MRI using a 1.5-T Gemsow Scanner (General Electric). MRI scan included eleven 1-cm thick horizontal slices. Fat area (visceral and subcutaneous adipose tissue) of each slice was estimated based on pixel intensity [19]. Abdominal fat volume was calculated from the product of the fat area and the thickness of the slice. Total abdominal fat volume included the sum of the fat volume obtained from the 11 slices.

Whole-body insulin sensitivity was estimated by the euglycemic hyperinsulinemic clamp, as previously described [20]. Basal samples were taken at t = -30, -20, -10, and -1 min. At t = 0 min, a somatostatin infusion was started and continued for the duration of the experiment. Porcine insulin was infused into a peripheral vein to induce hyperinsulinemia. Glucose was clamped at basal concentration by a variable infusion of glucose. Blood samples were collected every 10 min from t = -30 to 60 min, every 15 min from t = 60 to 120 min, and then every 10 min from t = 120 to 180 min. Whole-body insulin sensitivity (SI_CLAMP) was calculated from the following equation: SI_CLAMP = ΔGINF/(ΔI X G), where ΔGINF is the difference in glucose infusion rate during the steady state period (t = 150–180 min) from basal, ΔI is the difference in plasma insulin at steady state from basal, and G is the steady-state plasma glucose concentration.

Assessment of β-cell function in vivo was performed using the graded-hyperglycemic clamp [18, 22]. Glucose infusion (50%) was injected peripherally at variable rates using a syringe pump (Razel Scientific Instruments, Stamford, CT) in order to maintain blood glucose constantly at three sequential concentrations: 5.6 (t = 0–59 min), 8.3 (t = 60–149 min) and 11.1 mmol/L (t = 150–240 min). Blood samples were collected every 10 min throughout the experiment, starting 20 min prior to the commencement of glucose infusion. Glucose infusion rates were adjusted periodically based on plasma glucose readings using a blood glucose analyzer (YSI 2700, Yellow Springs Instruments, Yellow Springs, OH). β-Cell function was measured as the slope of the relation between insulin (pmol/L) and glucose (mmol/L) during the steady-state at each glucose clamp period (t = 40−60 min: 5.6 mmol/L, t = 130−150 min: 8.3 mmol/L, and t = 210−240 min: 11.1 mmol/L).

Islets were obtained from control-diet and HFD dogs (after completing their corresponding diet periods, as described before), immediately before euthanasia, under general inhalant anesthesia. All animals were 2–3 years old at the time of the pancreas procurement. Islet isolation, islet viability stain, and assessment of β-cell function in vitro were performed as previously described [23].

After 18–24 h culture, islets sized ~150–200 μm were handpicked and put on 24-well culture plates. Static incubation experiments consisted of 1-h equilibrium period with 3 mmol/L glucose, followed by 1-h additional period with either 3 mmol/L glucose (basal) or 15 mmol/L glucose (GSIS), in the presence or absence of exogenous anandamide and/or cannabinoid receptor antagonists. β-Cell function in vitro was estimated as the rate of insulin secreted (pmol•L⁻¹/islet/h) during the second hour of static incubation (basal insulin concentration during the equilibrium period was subtracted for calculation) and also expressed as the stimulation index: ratio of the insulin secreted at 15 mmol/L glucose (second hour of static incubation) to the basal insulin secreted at 3 mmol/L glucose (equilibrium period) [23].

To explore possible paracrine effects of anandamide in vitro, insulin, glucagon, and somatostatin secretion were measured in same islets during short incubation period (1 h, 100 islets per well) and prolonged static incubation (14 h, 15 islets per well). Since basal hormone secretion did not yield detectable hormone concentrations of glucagon and somatostatin, for this purpose, secretion of islet hormones were expressed as the total amount secreted during the 1-h equilibrium period and the second period (either 1 h or 14 h). Insulin secretion was expressed as pmol•L⁻¹/islet; glucagon and somatostatin release were expressed as ng•L⁻¹/islet.

Islet perifusion experiments were performed as previously described [23], with little modification. Islets were plated onto coverslips coated with BD cell-tak (BD Biosciences, San Jose, CA). After a 60-min equilibrium period (t = 0), islets were stimulated with 15 mmol/L glucose and 10 nmol/L anandamide for 36 min. The effluent was collected at 1-min intervals from t = −3 min until t = 10 min, followed by 2-min sampling until the end of the experiment (t = 36 min). β-Cell function was expressed as the net rate of insulin secreted (pmol•L⁻¹/islet) in response to 15 mmol/L glucose relative to the lowest insulin concentration in the first phase (t = 0 to 8 min). Also, fold increase over average baseline was calculated. Second phase was defined as the period t = 9 to 36 min.

Anandamide was purchased from Sigma-Aldrich (St Louis, MO). Iodoresiniferatoxin and AM630 were purchased from Tocris Bioscience (Minneapolis, MN). Rimonabant was kindly provided by Sanofi-aventis. All cannabinoid drugs were dissolved in dimethyl sulfoxide (0.1% final concentration).

Plasma non-esterified fatty acids were measured using a colorimetric assay [18]. Insulin from plasma samples and in vitro experiments were determined by ELISA [23]. Plasma C-peptide (canine kit, Millipore, St. Charles, MO) was determined by radioimmunoassay in duplicate. In vitro glucagon (canine kit, Millipore) and somatostatin were determined by radioimmunoassay in single samples. Initial experiments were assessed using the somatostatin kit 13-RB306 (American Laboratory Products Company, Windham, NH), then switched to RK-060-14 (Phoenix Pharmaceuticals Inc., Burlingame, CA) because kit 13-RB306 was discontinued. Anandamide, 2-AG, oleoyl ethanolamide, linoleoyl ethanolamide, and palmitoyl ethanolamide concentrations were determined in plasma samples using methanol and acetonitrile for extraction, and liquid chromatography/tandem mass spectrometry (LC/MS/MS system) for analysis, as described in detail elsewhere [24].

Gene expression of CB1R, CB2R, and the transient receptor potential vanilloid 1 (TRPV1) was assessed in intact fresh canine islets. Approximately 200 islets per animal were suspended in TRI Reagent and stored at ‒80°C. RNA was extracted from frozen islets using the Tri-Reagent Kit (Molecular Research Center Inc., Cincinnati, OH). Total RNA concentration was quantified by spectrophotometry, absorbance measured at 260 nm. The 260/280 nm absorption ratio of all preparations ranged between 1.8 and 2.0. RNA integrity was assessed by gel electrophoresis using agarose/ethidium bromide gel. First-strand cDNA was synthesized according to the manufacturer’s protocol, from 1 μg of total RNA using Superscript II (Invitrogen, Carlsbad, CA). Real-time PCR was performed on a Light-Cycler 2.0 instrument (Roche Applied science, Indianapolis, IN). The cDNA was amplified using ‘Universal probe system’ in a glass capillary in a final volume of 10 μL reaction mix containing 2.5 μL, 100-fold diluted cDNA, 2 μL LightCycler Taq-Man Master Mix buffer (Roche Applied Science, Indianapolis, IN), 1 μmol/L specific forward-reverse primers (CB1R forward: 5’-CCTGGTTCTGATCCTTGTGG-3’; CB1R reverse: 5’-ACCATAATCGCAAGCAGAGG-3’; CB2R forward: 5’-TACTTGCCCCTTATGGGATG-3’; CB2R reverse: 5’-ATCAGGGGGAAAAGCTCAG-3’; TRPV1 forward: 5’-GGGAACCAGGGAAAAGTTCT-3’; TRPV1 reverse: 5’-GAACTGTGAGGGCATCAAGC-3’) and 0.5 μL of universal probes (for CB1R, CB2R, and TRPV1 we used #3, #56, and #13, respectively). All primers and universal probes were designed in Roche Applied Science website (http://www.roche-applied-science.com). Light cycler was programmed as follows: 1) Pre-incubation, 10 min at 95°C; 2) denaturation, 50 cycles, 10 s at 95°C; 3) annealing, 40 s at 60°C, followed by extension, 1 s at 72°C. After 45–55 cycles they were cooled at 42°C for 30 s. The quantification of 18S rRNA was used for sample normalization using SYBER Green I kit. PCR was performed according to the manufacturer’s protocol (Roche Applied Science). The specificity of amplification was determined by melting curve analysis.

Data were not normally distributed, as determined by the Shapiro-Wilk's W test. Data were expressed as means±S.E.M., unless otherwise indicated. Spearman correlation was used to determine bivariate association between in vivo metabolic variables. Wilcoxon matched pairs test was used to evaluate differences within groups and the Mann-Whitney U test was used for comparison between groups. Friedman test was used to compare the effects of multiple drugs (groups) on β-cell function in islet batches from same animals. Friedman test was followed by Wilcoxon test if P < 0.05. A mathematical model approach (mixed-effects linear regression [25]) was used to determine the possible interaction between insulin, glucagon, and somatostatin secretion in response to anandamide during static incubation, while accounting for the number of replicates, glucose concentration, and type of diet. Differences were statistically significant if P<0.05. All analyses were performed using Statistica (StatSoft Inc., Tulsa, OK, USA) and Stata/SE 10.0 for Windows (StataCorp LP, College Station, TX).

HFD for 22 weeks induced a significant increase in body weight by 13.4±2.5% (P<0.01), and total abdominal fat content by 81.3±21.6% (P<0.01) (Table 1 and Fig 1A). Likewise, HFD induced a marked reduction in insulin sensitivity by 31.3±12.1% (P<0.05). In addition, there was a decrease in fasting plasma 2-arachidonoyl glycerol but not anandamide, oleoyl ethanolamide, linoleoyl ethanolamide, or palmitoyl ethanolamide (Table 1). Overall fasting insulin and C-peptide did not significantly increase after prolonged fat diet given the variable response in the animals studied. However, additional analysis in a subset of animals showing consistent increase in fasting insulin (n = 5, from 55.0±10.4 to 78.9±18.3 pmol/L, P<0.05) revealed no increase in plasma anandamide levels (P = 0.893). Likewise, animals with increased fasting C-peptide (n = 5, from 138.0±14.8 to 196.7±29.5 pmol/L, P<0.05) showed no changes in anandamide (P = 0.50). Moreover, in animals with increased insulin slope during hyperglycemic clamp (n = 5, from 75.0±19.8 to 110.3±19.7 pmol•L^-1/mmol•L^-1, P<0.05), anandamide showed rather a tendency to decrease (from 122.8±22.5 to 78.6±11.6 pmol/L, P = 0.080).

This study also found no correlation between body weight or abdominal fat adiposity and plasma anandamide or 2-AG. Likewise, we found no correlation between plasma anandamide and insulin or between plasma anandamide and C-peptide (Fig 1B). However, we found an inverse correlation between plasma 2-AG and insulin (P<0.05) and between plasma 2-AG and C-peptide (P<0.05) after 22 weeks of HFD (Fig 1C). Collectively, these findings do not support a direct association between fasting plasma anandamide and β-cell function in vivo.

Islets were isolated from control-diet dogs (n = 7, body weight: 28.8±1.0 kg) and from dogs fed a HFD for 19–22 weeks (21.5±0.5 weeks) (n = 6, body weight: 34.5±1.3 kg). In control dogs, islets incubated with 10 μmol/L anandamide had a higher rate of basal insulin secretion (10.6±5.7 pmol•L⁻¹/islet/h) as compared with no treatment (3.0±1.0 pmol•L⁻¹/islet/h; P<0.05) (Fig 2A). In HFD dogs, islets incubated with 10 μmol/L anandamide also had a higher rate of basal insulin secretion (6.9±1.3 pmol•L⁻¹/islet/h) as compared with no treatment (0.6±0.2 pmol•L⁻¹/islet/h; P<0.05). Despite the difference in body weight between groups (P<0.05), the insulinotropic effect of anandamide was not different (P = 0.886).

Anandamide also potentiated GSIS in control-diet and HFD animals (Fig 2B), without compromising islet viability (Fig 2C). In control dogs, islets incubated with anandamide had higher GSIS at 15 mmol/L glucose (30.5±9.7 pmol•L⁻¹/islet/h) as compared with no treatment (8.6±1.5 pmol•L⁻¹/islet/h; P<0.05). Likewise, the stimulation index (fold increase of insulin secretion over basal) was higher with anandamide (16.5±2.6 versus 5.9±0.6; P<0.05). The latter finding was not dependent on basal insulin differences (P = 0.499). In HFD dogs, islets incubated with anandamide also had higher GSIS (37.2±14.1 pmol•L⁻¹/islet/h) as compared with no treatment (9.6±4.7 pmol•L⁻¹/islet/h; P<0.05). The stimulation index was also higher with anandamide (21.1±7.0 versus 5.9±1.5; P<0.05), independent of basal insulin differences (P = 0.345). The insulinotropic effect of anandamide at high glucose concentrations was not different between HFD and control-diet dogs, either when comparing insulin secretion rates (P = 0.775) or the stimulation indexes (P = 1.00). Moreover, we found lower insulin secretion at low glucose concentrations in islets from HFD dogs (P = 0.010) as compared with islets from lean dogs (Fig 2A). However, we found no differences in GSIS (P = 0.568) between groups (Fig 2B).

Potentiation of GSIS by anandamide was substantially diminished by the CB1R antagonist rimonabant at 10 μmol/L (Fig 2B), suggesting a CB1R-mediated process, at least in part. In fact, we confirmed mRNA expression of CB1R in intact pancreatic canine islets (Fig 2D). However, we found no differences in gene expression of CB1R, CB2R or TRPV1 between control-diet and HFD animals (CB1R: P = 1.00; CB2R: P = 0.57; TRPV1: P = 0.71).

In another set of control-diet animals (n = 7, body weight: 26.0±0.5 kg), we explored whether lower concentrations of anandamide, similar to those reported in plasma from moderate obese subjects [10], could stimulate insulin secretion in isolated islets. Islets incubated with 10 nmol/L anandamide did not show significant higher rates of basal insulin secretion (1.3±0.1 pmol•L⁻¹/islet/h) as compared with no treatment (1.1±0.1 pmol•L⁻¹/islet/h; P = 0.131, Friedman test) (Fig 3A). At same concentrations, anandamide-stimulated islets did not show significant higher GSIS (7.5±1.6 pmol•L⁻¹/islet/h) as compared with non-stimulated islets (5.7±1.4 pmol•L⁻¹/islet/h; P = 0.052, Friedman test). However, Fig 3A clearly shows a trend, in all dogs, for an increase in insulin secretion with anandamide, that appears to be decreased by rimonabant and not by the CB2R antagonist AM630 [26] or the TRPV1 antagonist iodoresiniferatoxin [27]. None of these cannabinoid antagonists impaired islet viability (Fig 3B).

Data from islet perifusion experiments showed an insulinotropic effect of low concentrations of anandamide. Although anandamide had no effect on the first phase (P = 0.198), it did increase the second phase of insulin secretion (P<0.01) (Fig 3C). These in vitro findings suggest that anandamide might play a physiological role in the regulation of insulin secretion.

During static incubation, batches of 100 islets from control-diet animals (n = 7, body weight: 26.0±0.5 kg) incubated with 10 μmol/L anandamide for 1 h significantly increased basal insulin and glucagon secretion (Fig 4). Anandamide also potentiated GSIS and stimulated glucagon secretion. Similar results were found when batches of 15 islets from control-diet (n = 7, body weight: 28.8±1.0 kg) and HFD (n = 5, body weight: 34.2±1.6 kg) animals were incubated for 14 h with 10 μmol/L anandamide. However, we found no differences in anandamide-induced glucagon secretion between control and HFD animals at low glucose (5.5±2.1 and 3.6±0.9 ng•L⁻¹/islet; P = 0.705) or high glucose concentrations (5.0±1.7 and 3.9±0.8 ng•L⁻¹/islet; P = 0.705, control and HFD, respectively).

Using a regression model to determine the possible interaction among islet hormones, we found no association between anandamide and insulin secretion (P = 0.643) while controlling for glucose (P<0.001), glucagon (P<0.05), somatostatin (P<0.05), and type of diet (P = 0.059). Conversely, anandamide remained significantly associated with glucagon secretion (P<0.001) while controlling for glucose (P = 0.094), insulin (P = 0.054), somatostatin (P<0.05), and type of diet (P = 0.835). This analysis suggests that anandamide-stimulated insulin secretion is dependent on glucose concentration and paracrine effects, most likely glucagon since somatostatin did not significantly increase (Fig 4). Conversely, anandamide-stimulated glucagon secretion could result from a direct effect of this endocannabinoid on the pancreatic α-cells.