Introduction
Glucose-dependent insulinotropic polypeptide (GIP) is a hormone from the proximal small intestine that is secreted in response to food ingestion [
1]. Although it acts as an incretin hormone to promote insulin secretion, mice lacking GIP receptors show only a mild impairment of postprandial glucose homeostasis [
2], but appear to be protected from the obesogenic effects of either a high-fat diet or the hyperphagic leptin deficient
ob/
ob background [
3]. Similar amelioration of diet-induced obesity, with correspondingly improved insulin sensitivity, was observed in other mouse models with impaired GIP signalling due to treatment with a GIP-receptor antagonist (Pro
3-GIP) [
4] or ablation of GIP-secreting K cells [
5]. These results have been used to argue that inhibiting GIP signalling may have beneficial effects in obese individuals [
6]. A similar outcome could potentially be achieved by reducing GIP secretion.
GIP release is triggered by the arrival of nutrients in the upper small intestine, where it is produced by enteroendocrine K cells located predominantly in the duodenal epithelium [
1]. These are an open-type endocrine cell, with apical surfaces facing into the gut lumen, capable of directly sampling the luminal contents [
7,
8]. They are stimulated by a variety of nutrients, including monosaccharides and fats [
1,
9]. Glucose sensing involves uptake by the sodium-dependent glucose cotransporter 1 (SGLT1), a sodium-coupled transporter, as exemplified by the loss of glucose-triggered GIP secretion in SGLT1-deficient mice [
10]. The electrogenic SGLT1 signal, however, operates synergistically with cytosolic cAMP levels [
11], providing opportunities for G-protein-coupled receptor pathways to influence GIP release. The expression of the G
αs-coupled receptor
Gpr119 in K cells, for example, may contribute to fat-dependent GIP secretion via the locally generated ligands 2-mono-oleoyl glycerol and oleoylethanolamide, produced by luminal triacylglycerol digestion and local tissue metabolism, respectively [
12,
13]. Relatively little is known about inhibitory signals that reduce cAMP levels in K cells, mediated by G
αi-coupled receptors.
Transgenic mice producing the yellow fluorescent protein (YFP) Venus under the control of the GIP promoter have provided a new tool for identifying and purifying K cells for transcriptomic and functional analysis [
11]. Here we report that the expression of
Sstr5 and the cannabinoid receptor type 1 (
Cnr1, CB1), examples of predominantly G
αi-coupled receptors, is enriched in K cells and that these receptors can be targeted to manipulate GIP secretion.
Discussion
We demonstrate here that GIP release can be inhibited through the recruitment of Gαi-coupled receptors. Most importantly, CB1 receptor stimulation suppresses GIP but not GLP-1 secretion. Expression of cannabinoid receptors has, to our knowledge, not previously been reported in enteroendocrine cells, yet our expression analysis identified the CB1 receptor message (Cnr1) as one of the highest expressed G-protein-coupled receptor mRNAs in FACS-purified K cells. Consistent with this observation, GIP secretion was inhibited by the CB1 receptor agonist mAEA in vitro, an effect partly reversed by co-application of the inverse agonist AM251.
As expected, we observed a robust inhibition of both GIP and GLP-1 secretion in vitro by somatostatin. A suppressive effect of somatostatin on incretin secretion has long been known. Consistent with the predominant G
αi coupling of SSTRs, somatostatin suppressed cytosolic cAMP responses in the GLUTag cell line, at least in part via SSTR5. A role for SSTR5 in primary murine K and L cells was also evident from secretion studies in primary intestinal cultures. Additional roles for SSTR2 and SSTR3 in both cell types are suggested by the mRNA expression data and the incomplete suppression of the inhibitory effect of somatostatin by the SSTR5 antagonist. This is in agreement with a previous report that GLP-1 secretion from fetal rat intestinal cultures was suppressed by specific agonists of SSTRs in the rank order SSTR5>SSTR 2>SSTR 3 [
22].
Although we detected mRNA expression of
Cnr1 in small intestinal L cells, we observed no inhibition of GLP-1 secretion from primary cultures by mAEA. The CB1-dependent inhibition of GIP but not GLP-1 release did not reflect interexperimental variability as the difference was still observed when supernatant fractions from the same wells were analysed for both incretins. A lack of effect of mAEA on GLP-1 secretion was also observed in colonic cultures, consistent with the absence of
Cnr1 expression in purified colonic L cells. We thus conclude that the observed
Cnr1 mRNA expression in small intestinal L cells either does not translate into a sufficient density of functional receptors or reflects only a minority of L cells. In this context, it is interesting to note that there is an established overlap in the small intestinal K and L cell populations, with up to 20% of L cells also expressing GIP [
23‐
25].
Under in vitro conditions, the inverse CB1 agonist AM251 did not enhance basal or IBMX-triggered GIP secretion in the absence of an exogenous CB1 receptor agonist. This indicates that there is no significant constitutive activity of CB1 receptors in K cells capable of overcoming a rise in cAMP in the presence of phosphodiesterase inhibition. Interestingly, however, the SSTR5 antagonist also failed to enhance GIP secretion from small intestinal cultures in the absence of added somatostatin, despite the evident somatostatin tone in the same cultures as demonstrated by the doubling of GLP-1 release under the same conditions. Indeed, the measured somatostatin concentration of approximately 50 pmol/l in supernatant fractions from small intestinal cultures is close to the reported affinity of rodent SSTR5 (
K
d 50–300 pmol/l [
26]). Endogenously released somatostatin was also reported to block GIP release in canine intestinal cultures enriched by elutriation for endocrine cells [
21]. As the SSTR5 antagonist was capable of partially reversing the inhibition of GIP secretion by exogenously applied somatostatin, it is possible that the observed GIP secretion in the presence of IBMX represents a maximal stimulation of K cells under our culture conditions.
The expression of
Cnr1 and in vitro effects of CB1 ligands on GIP release were reflected by our in vivo observations in rats, which showed that GIP but not GLP-1 responses to an OGTT were reduced by mAEA pretreatment. As endocannabinoids are known to inhibit gastric emptying [
27], a trivial explanation for the observed delayed appearance of GIP in the plasma could have been a slowed transit of glucose into the duodenum. This seems an unlikely explanation for our findings, however, as we observed no significant differences in the time courses of glucose, acetaminophen and GLP-1 appearance in the plasma between the two groups. It thus appears likely that direct CB1-dependent inhibition of K cells at least in part underlies the reduction in the GIP response caused by mAEA in vivo. In a similar experiment to determine the in vivo effect of AM251, GIP levels were found to increase in the fasting state, but it was not possible to assess endogenous CB1 tone after a glucose challenge because AM251 unexpectedly slowed gastric emptying. This is likely to have reduced glucose delivery to the duodenal K cells, accounting for the observed lower plasma GIP and insulin levels, and correspondingly higher plasma glucose concentration at 30 min.
In the stomach and small intestine of rodents, arachidonylethanolamide and, in some but not all studies, the related endocannabinoid 2-arachidonyl-glycerol have been reported to increase in concentration upon fasting, with levels returning to baseline within an hour of refeeding [
28‐
30]. Intestinally generated endocannabinoids are thought to exert orexigenic effects by signalling through vagal afferent neurons to satiety centres in the central nervous system (CNS) [
27,
30]. The stimulatory effect of systemic cannabinoids on feeding is, however, believed to depend mostly on CB1 receptor activation in the CNS [
31]. Rimonabant, an inverse agonist at the CB1 receptor, was used clinically for the treatment of obesity due to its anorectic effects, but was withdrawn from the market because of psychological side effects, including depression and increased suicidal thoughts [
32].
Although most effects of rimonabant on feeding are mirrored in a mouse with CNS-restricted CB1 receptor knockout [
33], there is still considerable interest in developing drugs targeting peripheral cannabinoid receptors for the treatment of obesity [
27,
31]. Our results suggest that an additional effect of antagonising peripheral CB1 receptors may be to increase GIP secretion. Although this may beneficially enhance the incretin effect, GIP is thought to promote storage of energy resources in peripheral tissues, possibly due to direct effects on adipocytes [
34,
35]. This would potentially oppose the beneficial effects of CB1 receptor antagonism on body weight, although it should be noted that the effect of GIP on nutrient storage is controversial, with some laboratories reporting lipolytic actions on adipocyte models in vitro [
36]. Indeed, in humans, GIP had only minor effects on triacylglycerol clearance that became evident only under conditions of a hyperglycaemic hyperinsulinaemic clamp [
37].
The multiple roles of the endocannabinoid system, both peripherally and centrally, have made CB1 a difficult receptor to target successfully for the therapy of diabetes and obesity [
32]. Our results suggest that there is a small but significant CB1-dependent inhibition of K cells during fasting that might serve to increase the sensitivity to feeding-related signals, as the postprandial reduction in endocannabinoid levels would activate K cells in concert with nutrient stimuli arriving in the duodenal lumen. Regulation of GIP secretion by the endocannabinoid system could be envisaged as part of the transition from a preprandial to a postprandial metabolic state. Although the rate of delivery of glucose to the duodenum appears to provide a dominant degree of control over K cell secretion, as evident from the in vivo experiments with AM251, CB1 activation might provide a useful tool to dissociate GIP and GLP-1 secretion.
Acknowledgements
We thank K. Burling and the MRC-CORD for performing plasma assays for GLP-1, insulin and acetaminophen and G. Yeo (MRC-CORD) for assistance with microarray analysis. B. Moss provided assistance with RT-PCRs during a work experience placement. D. Drucker (Samuel Lunenfeld Research Institute, Toronto, Canada) and M. Lohse (University of Wuerzburg, Germany) kindly provided the GLUTag cell line and Epac2-camps sensor, respectively.