Absence of cannabinoid 1 receptor in beta cells protects against high-fat/high-sugar diet-induced beta cell dysfunction and inflammation in murine islets

All materials and mouse strains used in this study are detailed in ESM Tables 1–7.

Animal care and procedures were approved by the National Institute on Aging Animal Care and Use Committee. Mice were housed in groups of four using 12 h dark/light cycles, provided with water and fed ad libitum. Cnr1^flox/flox mice were generated as described in the ESM Methods. Cnr1^flox/flox mice were mated with MIP-Cre/ERT mice (University of Chicago, Chicago, IL, USA) and injected daily for 5 days with i.p. tamoxifen to obtain β-CB1R^−/− mice (Cnr1^flox/flox:MIP-Cre/ERT⁺), β-CB1R^+/+ control littermates (Cnr1^flox/flox:MIP-Cre/ERT⁻) and MIP-Cre/ERT mice (Cnr1^wt/wt:MIP-Cre/ERT⁺). CB1KO mice (NIH, Bethesda, MD, USA) were bred as described in the ESM Methods. Male mice (n = 6–7 mice/group) were aged to 25 weeks and body weights and metabolic variables were analysed. Body composition was analysed using NMR. Pancreases were fixed and processed for immunohistochemistry (anti-insulin [1:100; Dako], anti-glucagon [1:500; Sigma-Aldrich], anti-BrdU [1:100; Accurate Chemicals]). Islet size was quantified using Pancreas++ [25]. Hormones were quantified using ELISA. Methanol–chloroform-extracted ECs from plasma were analysed using LC-MS/MS, as described in the ESM Methods.

Male mice (6–8 weeks old; n = 6–7/group) were randomised to a standard diet (16.7% kJ fat and 12.4% kJ sugar wt/wt) or high-fat/high-sugar diet (HFHS; 49.2% kJ fat and 32.2% kJ sugar wt/wt) for 15 weeks. Body weight and food consumption were measured weekly. At the end of the study, GTTs and metabolic measurements were performed. Metabolic measurements were measured using a Columbus Instruments Comprehensive Lab Monitoring System (CLAMS; Columbus Instruments), as described in ESM. Tissues (pancreas, liver and white adipose tissue) were dissected, weighed and flash frozen or fixed for immunohistochemistry (anti-CD3, anti-CD68, anti-thioredoxin-interacting protein [TXNIP] and anti-phospho [p][S536]-p65 [1:100; Abcam], anti-ceramide [1:10; Enzo Life Sciences]). Immunoprecipitated insulin receptor (IR) or IRS-2 from liver protein extracts (anti-IRS2 [Cell Signaling], anti-IR [Santa Cruz Biotechnology]; 1 μg) were subjected to Tris-glycine PAGE, immunoblotted with anti-IRS2 or anti-p-Tyr (1:1000; Santa Cruz Biotechnology), visualised by enhanced chemiluminescence and quantified using ImageJ.

Freshly isolated islets were cultured in non-stimulatory conditions (2 mmol/l glucose) to quantify basal cAMP (n = 100 islets/mouse) and resting insulin secretion (n = 40 islets/mouse). Islets were cultured with exendin-4 (Ex-4) and 3-isobutyl-1-methylxanthine (IBMX; 25 μmol/l; Sigma-Aldrich), and intra-islet cAMP levels and insulin secretion were quantified. Islets (n = 100–120 islets/mouse) were infused with glucose (7.5 mmol/l) in a perifusion system and insulin secretion was quantified. Freshly isolated islets (n = 30 islets/group) were used to measure the oxygen consumption rate (OCR), extracellular acidification rate (ECAR) using an XFe24 Seahorse Analyzer (Agilent) and intra-islet levels of reactive oxygen species (ROS) as described in ESM. Islets (n = 30–50 islets/group) were cultured in the presence or absence of 500 μmol/l palmitate and 16.5 mmol/l glucose for 24 h. Islet viability, cAMP levels, mRNA expression (Il1b, Nlrp3, Tnfa [also known as Tnf], Ldha), cytokine secretion (IL-1β, TNF-α), protein phosphorylation (p-Erk1/2, p-SAPK/JNK, p-p38, p-p53, p-Stat1, p-HSP27) and cleavage (caspase 3, PARP) were measured. Islets (n = 15 islets/group) were incubated with a mixture of cytokines (10 ng/ml IL-1β, 50 ng/ml TNF-α and 50 ng/ml IFN-γ) or vehicle for 18 h and cytotoxicity was assayed. There were at least three replicates in all experiments.

Pancreatic lymphocytes were isolated after pancreas perfusion and cells were stained using PE anti-mouse CD8a and FITC anti-mouse CD69 (Biolegend), as detailed in the ESM Methods. Populations were determined by using BD FACSCanto II and BD FACSDiva software (BD Biosciences).

Microarray experiments and analysis were performed as previously described [26].

Age and sex-matched littermate mice were randomly assigned to vehicle, S961, standard diet or HFHS.

Data are presented as means ± SEM. No statistical method was used to predetermine the sample size. Differences between mean values for variables within individual experiments were compared statistically using Student’s t test or ANOVA, as appropriate. Comparisons were performed using GraphPad Prism version 6.0. p ≤ 0.05 was considered statistically significant.

Conditional β-CB1R^−/− mice were obtained by mating Cnr1^flox/flox with MIP-Cre/ERT mice (ESM Fig. 1). Six-week old β-CB1R^−/− and β-CB1R^+/+ mice were injected with tamoxifen for 5 days and monitored until 25 weeks of age (Fig. 1a–d). No difference in body weight (Fig. 1b) or body composition (ESM Fig. 2a–d) was observed between β-CB1R^+/+ and β-CB1R^−/− mice. Fasting plasma insulin levels were significantly increased (153 ± 23%) by 10 weeks of age in β-CB1R^−/− mice compared with baseline and were significantly higher than in β-CB1R^+/+ (Fig. 1c) and MIP-Cre/ERT mice (ESM Fig. 2e), which was reflected in lower fasting blood glucose (FBG) levels (Fig. 1d, ESM Fig. 2f). β-CB1R^−/− mice were more glucose tolerant than β-CB1R^+/+ mice in an IPGTT (Fig. 1e). After oral glucose-lipid challenge, β-CB1R^−/− mice had higher circulating insulin levels than β-CB1R^+/+ mice (Fig. 1f, ESM Fig. 2g). At 10 weeks, beta cell proliferation was increased 5.6-fold in β-CB1R^−/− mice (Fig. 1g–i), and the islet area was significantly greater (1.9-fold) than in β-CB1R^+/+ (Fig. 1j) and MIP-Cre/ERT (ESM Fig. 2h, i) mice, with no change in islet number (ESM Fig. 2j). Beta cell proliferation is regulated by insulin, cytokines and prolactin, which stimulate Igf1 expression in beta cells [27]. The trophic changes observed were independent of intra-islet IGF-1, since Igf1 expression was decreased in β-CB1R^−/− compared with β-CB1R^+/+ islets (Fig. 1k).

We examined whether β-CB1R^−/− mice would still respond to acute insulin resistance using S961, a competitive antagonist of the IR [28]. Ten-week-old β-CB1R^+/+ and β-CB1R^−/− mice were implanted with miniosmotic pumps containing vehicle or S961. Glucose and insulin levels were followed daily for 6 days (Fig. 2a). By day 6, non-FBG was significantly lower in S961-treated β-CB1R^−/− mice (19.7 ± 1.9 mmol/l) than in β-CB1R^+/+ (28.3 ± 1.3 mmol/l) (Fig. 2b) and MIP-Cre/ERT mice (ESM Fig. 3a). S961 treatment increased circulating insulin levels in β-CB1R^+/+ mice, but even more so in β-CB1R^−/− mice (Fig. 2c). By day 6, S961 had caused a 21-fold increase in beta cell proliferation per islet in β-CB1R^+/+ mice compared with baseline, and to a lesser extent (6.5-fold) in β-CB1R^−/− mice (Fig. 2d, e, ESM Fig. 3b). Increased beta cell proliferation in S961-β-CB1R^+/+ mice resulted in increased beta cell number and larger islets (Fig. 2f, g). β-CB1R^−/− mice already had significantly more beta cells than β-CB1R^+/+ mice; a 6.5-fold increase in proliferation in β-CB1R^−/− mice was sufficient to further increase total beta cell numbers when mice were treated with S961 (Fig. 2f). Alpha cell numbers were comparable between strains; S961 treatment resulted in increased alpha cell numbers only in β-CB1R^+/+ mice (ESM Fig. 3c). S961 treatment did not lead to a further increase in islet size in β-CB1R^−/− mice compared with β-CB1R^+/+ mice (Fig. 2g) because beta cell turnover was close to its maximum potential due to the absence of CB1R, and a further slight increase in proliferation did not significantly increase size. Beta cell apoptosis, as assessed by TUNEL staining (ESM Fig. 3d, e), and immune cell infiltration were not observed in or around islets (ESM Fig. 3f).

β-CB1R^+/+ and β-CB1R^−/− mice, 10–12 weeks old, were fed an HFHS diet (HFHS-mice) for 15 weeks (Fig. 3a) as a chronic stimulus for insulin secretion/production. β-CB1R^−/− mice had similar body weight (35–36 g) to β-CB1R^+/+ mice when fed a standard diet (SD-mice). However, HFHS-β-CB1R^−/− mice gained 21 ± 4% more weight than HFHS-β-CB1R^+/+ and MIP-Cre/ERT mice (Fig. 3b, ESM Fig. 4), with comparable food intake (Fig. 3c). This phenotype is distinct from the global CB1KO mouse phenotype, with the latter weighing less than SD- or HFHS-fed β-CB1R^+/+ and β-CB1R^−/− mice (ESM Fig. 5a), demonstrating that this effect is specific to CB1R ablation in beta cells. Liver and subcutaneous adipose tissue from HFHS-β-CB1R^−/− mice weighed significantly more than that from control littermates (Fig. 3d), while CB1KO mouse liver and subcutaneous adipose tissue weighed less than that from control littermates (ESM Fig. 5b). HFHS-β-CB1R^+/+ mice had a higher respiratory exchange rate (RER) during daylight compared with HFHS-β-CB1R^−/− mice (0.79 ± 0.01 vs 0.74 ± 0.01) (Fig. 3e). RER, the ratio of carbon dioxide production/oxygen consumption, estimates the fuel source since glucose metabolism requires more oxygen than fat [29]. At night-time the fuel source is the consumed food. The lower RER in HFHS-β-CB1R^−/− mice during daylight reflects increased fat utilisation because of differences in fat depots. HFHS-β-CB1R^−/− mice had lower activity during the night (active cycle) than HFHS-β-CB1R^+/+ mice (Fig. 3f). HFHS-β-CB1R^−/− mice became more insulin intolerant, as demonstrated by higher FBG (Fig. 4a) and fasting plasma insulin (Fig. 4b) levels than HFHS-β-CB1R^+/+ mice. Liver IRS-2 protein levels, an indicator of insulin resistance [30], were lower in HFHS-β-CB1R^+/+ compared with SD-β-CB1R^+/+ mice (Fig. 4c, ESM Fig. 5c). SD-β-CB1R^−/− mice had lower liver IRS-2 levels than SD-β-CB1R^+/+ mice, which were unchanged on an HFHS diet (Fig. 4c, ESM Fig. 5c). This suggests that even on a standard diet, insulin resistance occurs in the livers of β-CB1R^−/− mice, probably as an adaptation to mitigate hypoglycaemia. IR phosphorylation was also decreased in HFHS-β-CB1R^−/− mice compared with HFHS-β-CB1R^+/+ mice (ESM Fig. 5d). Blockade of CB1R in the liver alleviates diet-induced insulin resistance by increasing insulin sensitivity [12]. To determine whether acute pharmacological inhibition of peripheral CB1R (i.e. liver and/or muscle) could alleviate insulin resistance in HFHS-β-CB1R^−/− mice, we administered the peripherally restricted CB1R antagonist JD-5037 [19] i.p. 30 min before an ITT. JD-5037 increased insulin sensitivity in HFHS-β-CB1R^+/+ mice (ESM Fig. 5e), and more so in the insulin-intolerant HFHS-β-CB1R^−/− mice (ESM Fig. 5f). During an IPGTT, HFHS-β-CB1R^−/− mice had higher blood glucose than HFHS-β-CB1R^+/+ mice, probably because of increased insulin resistance (Fig. 4d). Interestingly, early insulin secretion was protected in HFHS-β-CB1R^−/− mice (Fig. 4e). Similar results were obtained during an OGTT (Fig. 4f, g). Stimulated GLP-1 and GIP plasma levels were significantly lower in HFHS-β-CB1R^−/− compared with HFHS-β-CB1R^+/+ mice 20 min after glucose-lipid challenge (Fig. 4h, i), which probably indicates a homeostatic negative feedback response to prevent hyperinsulinaemia.

To study the enhanced beta cell function observed in β-CB1R^−/− mice, we isolated islets for ex vivo analysis (Fig. 5a). β-CB1R^−/− islets were visually bigger than those from β-CB1R^+/+ mice (Fig. 5b), consistent with in situ measurements (Fig. 1j). CB1R is constitutively active under non-stimulated conditions, and thus inverse agonists or genetic ablation should eliminate basal activity [31]. Non-stimulated, resting intracellular cAMP levels (Fig. 5c) and insulin secretion (Fig. 5d) were increased in β-CB1R^−/− compared with β-CB1R^+/+ islets (217 ± 33% and 175 ± 10%, respectively). Ex-4-stimulated AC activity resulted in a 391 ± 9% increase in cAMP levels in β-CB1R^−/− islets compared with 176 ± 4% in β-CB1R^+/+ islets (Fig. 5e). Ex-4-stimulated glucose-mediated insulin secretion was also significantly higher in β-CB1R^−/− compared with control islets (Fig. 5f). Islets were placed in a perifusion system [8] to examine glucose-stimulated insulin secretion (Fig. 5g). β-CB1R^−/− islets had increased insulin secretion, an earlier peak of the first phase and greater AUC for insulin than β-CB1R^+/+ islets (Fig. 5g), all likely resulting from increased AC activity (Fig. 5c). Similar results were found in CB1KO islets (ESM Fig. 6a). To mimic diet-induced beta cell damage, we cultured isolated islets with or without high glucose and palmitate (HGP) for 24 h (Fig. 5h). Loss of viability as seen in HGP-β-CB1R^+/+ islets was not observed in HGP-β-CB1R^−/− islets (Fig. 5i). HGP-β-CB1R^−/− islets also had higher intracellular cAMP levels (required for maximal insulin secretion) than control littermates, reflecting increased beta cell functional capacity (Fig. 5j).

CB1R impacts mitochondrial metabolism in muscle and neurons [32, 33], and ATP generation via glucose metabolism is required for insulin secretion [34]. We analysed mitochondrial metabolism in β-CB1R^−/− islets by measuring the OCR and ECAR (Fig. 6a–c). Islets were pre-incubated for 1 h in 2 mmol/l glucose; the addition of 16.5 mmol/l glucose increased mitochondrial respiration (oligomycin-inhibited ΔOCR) by 185 ± 8% in β-CB1R^+/+ islets, and to a lesser extent (158 ± 9%) in β-CB1R^−/− islets (Fig. 6b, ESM Fig. 6b). The OCR was inhibited by oligomycin in both strains, indicating that the oxygen consumption was due to mitochondrial respiration. The basal ECAR (an indicator of lactate production) was equal in both strains, and the addition of glucose significantly increased the ECAR in β-CB1R^−/− islets compared with β-CB1R^+/+ islets (509 ± 94% vs 311 ± 28%; Fig. 6c). Mitochondria are the primary source of ROS, and a high respiration rate is associated with ROS production [35]. Islets from fasted β-CB1R^−/− mice had significantly lower ROS production compared with β-CB1R^+/+ mice (Fig. 6d). Ldha expression significantly increased only in HGP-β-CB1R^−/− islets (Fig. 6e), in concordance with the ECAR data.

Obesity-related molecules such as ROS, ceramides, palmitate and elevated circulating glucose activate the NLRP3 (NLR family, pyrin domain containing 3) inflammasome through phosphorylation of NFκB p65, leading to immune cell infiltration [36‐41]. ROS activates NLRP3 by inducing a direct interaction with TXNIP [36‐38]. HFHS-β-CB1R^−/− mice (Fig. 7a) had lower intra-islet levels of TXNIP and ceramide than HFHS-β-CB1R^+/+ mice (Fig. 7b). Intra-islet levels of p-p65 were significantly higher in HFHS-β-CB1R^+/+ compared with HFHS-β-CB1R^−/− mice (Fig. 7b). Infiltrating cells, such as T cells (CD3⁺) and macrophages (CD68⁺), were observed in close proximity to blood vessels, and in and around the islets of HFHS-β-CB1R^+/+ mice (yellow arrows, Fig. 7b, c, ESM Fig. 7a). Flow cytometry analysis showed that the T cells were CD69⁺ (ESM Fig. 7a). Such infiltration was absent in β-CB1R^−/− mice fed an HFHS diet for 15 weeks (Fig. 7a, c, ESM Fig. 7b). Circulating EC levels remained unchanged (ESM Fig. 7c) and intra-islet Cnr2 expression was significantly lower (ESM Fig. 7d) in β-CB1R^−/− mice; there was no compensation by the cannabinoid 2 receptor (CB2R), the main EC receptor in immune cells, or activation of CB1R in other islet cell types that explained the reduced immune cell infiltration. β-CB1R^−/− islets had lower Il1b, Nlrp3 and Tnfa expression than β-CB1R^+/+ islets (Fig. 7d–f). Microarray analysis confirmed a downregulation of CB1R, TXNIP, NFκB, TNF-α and IL-1β pathways in β-CB1R^−/− relative to β-CB1R^+/+ islets (ESM Fig. 7e). Tnfa expression was increased only in HGP-β-CB1R^+/+ but not in HGP-β-CB1R^−/− islets (Fig. 7g, h). Although Il1b expression was increased in HGP-β-CB1R^−/− islets (Fig. 7h), it was significantly less than in HGP-β-CB1R^+/+ mice. Secretion of IL-1β and TNF-α from HGP-β-CB1R^−/− was significantly reduced or undetectable, respectively, compared with HGP-β-CB1R^+/+ islets (Fig. 7i, j). Activation of JNK and p38, as happens with proinflammatory cytokines such as IL-1β and TNF-α, leads to beta cell death [42, 43] and EC-mediated cytokine secretion from macrophages [18]. HGP-β-CB1R^−/− islets had significantly lower levels of p-p38 than β-CB1R^+/+ islets, while p-Erk1/2 or p-SAPK/JNK did not change (Fig. 7k). Downstream signalling of p38 was also reduced (p-p53, p-Stat1 and p-HSP27), together with reduced cleaved-caspase 3 and cleaved-PARP (Fig. 7k), indicating reduced cytokine-induced apoptosis [44]. To investigate if lower MAPK activation was due to lower cytokine secretion or sensitivity, islets were treated ex vivo with mixtures of cytokines (IL-1β, TNF-α and IFN- γ [45]) for 18 h (Fig. 7l). β-CB1R^−/− islets had significantly lower levels of cytotoxicity than β-CB1R^+/+ islets, with levels never reaching above those of vehicle-treated β-CB1R^+/+ islets (Fig. 7m). These data show that β-CB1R^−/− islets have lower basal cytotoxicity and, while responsive to cytokines, lower cytokine secretion lessens the inflammatory cascade protecting the islet.