Somatostatin Is Only Partly Required for the Glucagonostatic Effect of Glucose but Is Necessary for the Glucagonostatic Effect of K_ATP Channel Blockers

The discovery that hyperglucagonemia (1–3) contributes to diabetes has sparked much interest in understanding the physiology of pancreatic α-cells. However, the mechanisms by which glucose controls glucagon secretion are still poorly understood and are the subject of many hypotheses that are often contradictory (4–8). It has been shown that glucose inhibits (9) or, conversely, stimulates (10,11) glucagon release of isolated α-cells. In the latter situation, the glucagonostatic effect that the sugar exerts on whole islets would be mediated indirectly via one or several paracrine factors released from non–α-cells within the islet. The nature of this factor is unknown. It has been suggested that it could correspond to somatostatin (SST) released by δ-cells (12,13), but it is not unanimously accepted (8,14,15). Moreover, it was reported that in a medium without physiological amino acids, high concentrations of glucose (≥25 mmol/L) fail to inhibit glucagon secretion of islets or even stimulate it (4,8,16,17). The role of K_ATP channels in the control of glucagon release is also highly debated, and variable effects of K_ATP channel blockers have been reported (8,14,18). The reasons for these controversies are unknown and might be linked to the experimental conditions.

In the current study, we evaluated the role of SST in the control of glucagon secretion by K_ATP channel blockers and various glucose concentrations. Two types of preparations were used: isolated islets and the in situ perfused pancreas, which preserves endocrine and paracrine interactions and is the best ex vivo preparation mimicking the physiological situation. Our data suggest that glucose inhibits glucagon release by SST-dependent and SST-independent mechanisms. The contribution of both mechanisms depends on the glucose concentration, with the SST-dependent mechanism being recruited by high concentrations of the sugar. We also present evidence showing that K_ATP channel blockers control glucagon release by two mechanisms: a direct stimulation of α-cells (observed in the absence of SST influence) and an indirect inhibition via SST released from δ-cells. Gene expression analysis shows, as recently suggested (19,20), that α-cells express several SST receptor (SSTR) subtypes, including SSTR2 and SSTR3.

The study was approved by our ethics commission for animal experimentation (project 2014/UCL/MD/016). Several mouse models were used: C57BL/6N mice; Sst^−/− mice (21) and control Sst^+/+ mice with the same genetic background (12); Sst^−/− mice backcrossed for five generations with C57BL/6N mice; two models lacking functional K_ATP channels, i.e., Sur1^−/− (22) and Kir6.2^−/− mice (23); and Sst^−/−/Kir6.2^−/− mice (obtained by crossing Sst^−/− with Kir6.2^−/− mice to create Sst^+/−/Kir6.2^+/− mice and then by crossing Sst^+/−/Kir6.2^+/− mice with each other).

The medium used for all experiments contained (in mmol/L) 124 NaCl, 4.8 KCl, 2.5 CaCl₂, 1.2 MgCl₂, 20 NaHCO₃, 1 mg/mL BSA, and test agents as indicated. It was gassed with O₂:CO₂ (95:5%, pH 7.4). For secretion experiments, it was supplemented with two types of amino acid mixtures: either a 6 mmol/L mixture (MixAA 6) of three amino acids containing (in mmol/L) 2 alanine, 2 glutamine, and 2 arginine or a 2 mmol/L mixture (MixAA 2) containing (in mmol/L) 0.4 alanine, 0.5 glutamine, 0.2 lysine, 0.25 glycine, 0.15 leucine, 0.25 valine, 0.15 threonine, and 0.1 serine. Both MixAAs have been used previously (14,24). The composition of MixAA 2 mimics the physiological concentration of the most abundant amino acids in mouse plasma as measured in our laboratory (Supplementary Table 1). In most cases, we used MixAA 6 for islet perifusion and MixAA 2 for pancreas perfusion. The rationale for using MixAA 6 for perifused islets was to have a rate of glucagon secretion that is high enough to allow detection of the secreted hormone with a reasonable amount of islets (∼200 islets/perifusion chamber). We nevertheless verified that key results obtained with perifused islets and the perfused pancreas were similar with both amino acid mixtures. SST14 and H6056 were from Bachem, pertussis toxin (PTx) was from List Biological Laboratories, CYN154806 and cyclosomatostatin were from Tocris Biosciences, and all other compounds were from Sigma-Aldrich.

Islets were isolated with collagenase and, except for one series of experiments performed on freshly isolated islets, were cultured overnight in RPMI 1640 medium containing 7 mmol/L glucose and 10% heat-inactivated FCS. For incubation experiments, batches of seven islets were incubated at 37°C in 1 mL of medium containing various test substances. For perifusion experiments, batches of 120–400 islets were perifused at 37°C, at a flow rate of 0.5 mL/min, with various test solutions.

Mice were anesthetized with ketamine (100 mg/kg) and xylazine (16 mg/kg) injected intraperitoneally. Control experiments showed that these anesthetics did not alter glucagon secretion in our experimental conditions (Supplementary Fig. 1). The pancreas was perfused in situ at 37°C at a flow rate of 1 mL/min in a single-pass circuit through both the celiac trunk and the superior mesenteric artery. Therefore, a catheter was inserted in the abdominal aorta above the inferior mesenteric artery, and the venous effluent was collected by another catheter inserted in the portal vein. To minimize leakage of the perfusate by routes other than the portal vein, ligatures were performed at the level of renal arteries and the abdominal aorta above the celiac trunk. Moreover, the entire intestine was resected except for ∼2 cm of duodenum, which was ligatured at its distal part. A ligature was also performed at the level of the stomach. To avoid coagulation, the pancreas was first perfused with 2 mL heparinized (50 IU/mL) PBS. After an initial equilibration period of 30 min with the same solution as that used immediately thereafter, the effluent was collected every 4 min.

Insulin (homemade assay), glucagon (Merck Millipore, Burlington, MA), and SST (Euro-Diagnostica, Malmö, Sweden) were measured by radioimmunoassay. We checked that all drugs did not interfere with the assays.

GYY and RIPYY mice expressing enhanced yellow fluorescent protein under the control of the glucagon promoter or the rat insulin promoter, respectively, were used to obtain pure populations of α- and β-cells (25). A BD FACSAria III Cell Sorter was used to sort the cells (excitation: 488 nm, emission: 530 nm, 100 μm nozzle, 20 psi, 30 kHz). The purity of the sorting, checked on a fluorescence microscope, was >98%. Gene expression was analyzed in purified α- and β-cells (26). Total RNA was extracted using TriPure (Roche). RNA (60 ng) was reverse transcribed into cDNA according to the protocol of Roche Applied Science for High Fidelity Synthesis Kit, except for RevertAid Reverse Transcriptase and Ribolock RNase Inhibitor (Thermo Scientific). Real-time PCR using SYBR Green was performed with an iCycler iQ Real-Time PCR Detection System (Bio-Rad, Hercules, CA). The primers used were as follows (5′→3′ forward and reverse sequences): SSTR1, 5′-CTACTgTCTgACTgTgCT-3′ and 5′-ATgggCAAgATAACCAgTAAT-3′; SSTR2, 5′-TCTTCCgTgTCTgTggC-3′ and 5′-gggATTTgTCCTgCTTACT-3′; SSTR3, 5′-CTggCgAACAgCCTTTAT-3′ and 5′-ggTgCCTgTACCCACTgA-3′; SSTR4, 5′-TgTgCTATTATTCAAACTggCT-3′ and 5′-ggTgTCAACTTCAggATTgT-3′; SSTR5, 5′-CTATgTggTgTTgCggT-3′ and 5′-ggCACAAgAAggAgCCAAA-3′; glucagon, 5′-ATgAACACCAAgAggAACCgg-3′ and 5′-CTTCTgggAAgTCTCgCCTT-3′; insulin, 5′-TCTTCTACACACCCATgCCC-3′ and 5′-ggTgCAgCACTgATCCAC-3′; and GAPDH, 5′-ACCCAgAAgACTgTggATgg-3′ and 5′-ACACATTgggggTAggAACA-3′. The annealing temperature was set to 62°C for all primers. GAPDH was used as internal control for RT-PCR efficiency and subsequent normalization. It is an excellent housekeeping gene because its cycle threshold (Ct), using the same amount of total cDNA, was similar for α- and β-cells, i.e., 26.7 ± 0.5 (n = 3) and 27.3 ± 0.5 (n = 3), respectively. Expression of all genes was reported to that of GAPDH giving a Ct difference (ΔCt) for GAPDH minus the test gene. The results are expressed as 2^−ΔCt.

The results are presented as mean traces (±SE) of experiments with islets or pancreas obtained from at least three different preparations of mice. Statistical significance of differences was evaluated by paired and unpaired Student t tests or one-way ANOVA followed by Dunnett or Tukey test.

To evaluate the role of SST in the control of glucagon secretion, islets from C57BL/6 mice were pretreated or not with PTx (200 ng/mL, 18 h). By ADP ribosylating the α subunit of G_i/o proteins, PTx locks them in a GDP-bound inactive state and is expected to block the effect of SST (27). We first tested whether PTx pretreatment effectively prevented the glucagonostatic effect of SST. These experiments were performed on incubated islets in the presence of a 6 mmol/L mixture of three amino acids to stimulate glucagon secretion and better disclose any inhibitory effect. As expected, SST inhibited glucagon release of control islets (Fig. 1A). By contrast, pretreatment with PTx prevented the glucagonostatic effect of SST, demonstrating that PTx was effective. It also increased glucagon release. Since glucagon content was similar in control and PTx-treated islets (14), this suggests that SST exerts a tonic inhibition on glucagon release of control islets. We next tested the possible involvement of SST in the effect of various glucose concentrations (0–30 mmol/L [G0-G30]) (Fig. 1B). In control islets, glucose dose-dependently inhibited glucagon secretion. The glucagonostatic effect was close to maximal at G7, with no major additional inhibition at higher glucose concentrations. By contrast, the dose-response curve was different in PTx-treated islets. The relative amplitude of the inhibition was smaller. Moreover, the inhibition was maximal at G7 but progressively waned as glucose further increased above G7. At G15 and G30, glucose even failed to inhibit glucagon release.

To verify the difference in glucagon secretion between G7 and G30 in PTx-treated islets, we performed dynamic perifusion experiments. To be closer to the physiological situation, these experiments were performed in the presence of a 2 mmol/L mixture of eight amino acids present at physiological concentrations. In control islets, switching between G7 and G30 very slightly inhibited glucagon release (Fig. 1C and E). The rate of glucagon release was much higher in PTx-pretreated islets than in control islets, and it was even larger in response to G30. Similar effects were produced by G20 (Fig. 1D and F). Overall, these results suggest that, in control islets, glucose exerts an inhibitory effect that is partly independent of SST (not abrogated by PTx) at low concentrations of the sugar and that is due to SST (abrogated by PTx) released from δ-cells at glucose concentrations >7 mmol/L. Since PTx might block G_i/o-coupled receptors other than those activated by SST, we verified more directly the involvement of SST by performing experiments with islets from Sst^+/+ and Sst^−/− mice.

As for the experiments with PTx, these experiments were performed in the presence of a 6 or 2 mmol/L mixture of amino acids (Fig. 2). In Sst^+/+ islets, switching between G7 and G30 did not affect glucagon release whereas G1 stimulated glucagon secretion (Fig. 2A–D). In Sst^−/− islets, glucagon secretion was higher than in Sst^+/+ islets. This cannot be explained by a higher glucagon content since the content was, on the contrary, slightly lower (∼20%) in Sst^−/− than in Sst^+/+ islets (14). These results again demonstrate the tonic inhibitory effect of SST on glucagon release. Switching from G7 to G30 stimulated glucagon release of Sst^−/− islets to a similar extent as G1 (Fig. 2A–D). This stimulation was not due to an osmotic effect because it was not reproduced by the addition of 23 mmol/L sucrose to a medium containing G7 (Fig. 2C and D). G20 also stimulated glucagon secretion from Sst^−/− islets but not from Sst^+/+ islets (Fig. 2E and F). These results confirm that SST is partly involved in the glucagonostatic effect of glucose, at least at high concentrations of the sugar. We next evaluated which SSTR subtypes are involved in this glucagonostatic effect.

Gene expression of SSTRs was investigated in α- and β-cells purified by FACS from GYY and RIPYY mice, respectively. Control experiments showed that our cell preparations were highly enriched in α- or β-cells. Indeed, glucagon gene expression was 1,500 times higher in fluorescent cells (enhanced yellow fluorescent protein positive) from GYY mice than from RIPYY mice (Fig. 3A). On the other hand, insulin gene expression was 20 times higher in fluorescent cells from RIPYY mice than from GYY mice. Among the five SSTR genes, α-cells mainly express SSTR2 and SSTR3 (Fig. 3B). SSTR2 is the major isoform and is 17 times more expressed in α- than β-cells. SSTR3 is equally expressed in α- and β-cells. The other isoforms are poorly or not significantly expressed in α-cells.

Because SSTR2 is highly expressed in α-cells, we used CYN154806, a SSTR2 antagonist, to test its role in the control of glucagon secretion. The addition of 300 nmol/L CYN154806 to a medium containing G7 strongly stimulated glucagon secretion from islets of C57BL/6 mice, indicating that glucagon release was already inhibited by the tonic activation of SSTR2 (Fig. 4A). However, CYN154806 did not prevent G1 from stimulating glucagon release. We next tested the role of SSTR2 in the control of glucagon secretion by G30. In the absence of CYN154806, switching from G7 to G30 very slightly inhibited glucagon release and stimulated SST secretion, whereas G1 stimulated glucagon release and inhibited SST secretion (Fig. 4B and C). In the continuous presence of CYN154806, glucagon secretion was much higher, and G30 still inhibited glucagon release. Because we found that α-cells also express SSTR3, we tested the effect of CYN154806 in combination with H6056 (PRL2915, 1 µmol/L), an antagonist that has been reported to effectively block both SSTR2 and SSTR3 (28). The combination of both antagonists removed the inhibitory effect of G30 (Fig. 4B). This was not due to a reduced SST release since the antagonists did not prevent the stimulatory effect of G30 on SST secretion (Fig. 4C). The fact that the antagonists did not unmask a stimulatory effect of G30 is, at first sight, at variance with the glucagonotropic effect of G30 observed in islets devoid of the paracrine influence of SST (Figs. 1 and 2) unless the antagonists did not fully block the action of SST. To verify the effectiveness of the SSTR antagonists, we tested whether they could block the glucagonostatic effect of SST. These experiments were performed with islets of Sst^−/− mice because their glucagon secretion is not tonically inhibited by endogenous SST, as is the case in control islets, and, hence, the effect of SST is more easily detected. SST14 (1 nmol/L) applied to a medium containing G2 strongly inhibited glucagon release (Fig. 4D and E). This inhibition was attenuated but not abolished by CYN154806 or the combination of CYN154806 + H6056 (Fig. 4D and E). Cyclosomatostatin (10 µmol/L), a nonselective SSTR antagonist, also failed to block the action of SST (not shown). It is unclear if these results reflect a poor efficacy of the antagonists as reported in some systems (29). The impossibility to fully block the action of SST with antagonists prevented us from further investigating the involvement of SSTR subtypes in the control of glucagon secretion of control islets.

Another objective of the study was to evaluate the role of SST in the control of glucagon secretion by K_ATP channel blockers. Therefore, islets from Sst^+/+ and Sst^−/− mice were perifused with a 6 mmol/L mixture of amino acids, and the effects of glucose and two K_ATP channel blockers, tolbutamide and gliclazide, were tested. In Sst^+/+ islets, switching between G1 and G7 reversibly inhibited glucagon secretion, whereas tolbutamide and gliclazide were ineffective (Fig. 5A and B). In Sst^−/− islets, switching between G1 and G7 inhibited glucagon release. By contrast, tolbutamide and gliclazide exerted a glucagonotropic effect that was obvious upon removal of the drugs (Fig. 5A and B) (see also shorter protocol in Supplementary Fig. 2). Similar results were obtained in freshly isolated islets (Supplementary Fig. 3). To better detect the glucagonotropic effect of the sulfonylureas, additional experiments were performed in a nonstimulating medium that lacked amino acids and contained G7. Low concentrations of the drugs already very efficiently stimulated glucagon release (Fig. 5C and D). Similar glucagonotropic effects of low concentrations of tolbutamide were also observed in C57BL/6 islets pretreated with PTx (Fig. 5E) or perifused with the SSTR2 and SSTR3 antagonists CYN154806 and H6056 (Fig. 5F). Insulin secretion measurements revealed similar responses of Sst^+/+ and Sst^−/− islets to low concentrations of the sulfonylureas, demonstrating that removal of the SST paracrine influence did not affect the sensitivity to the K_ATP channel blockers (Supplementary Fig. 4).

To evaluate the K_ATP channel and SST dependence of the effects of glucose on glucagon secretion, we compared the effect of G1 and G7 on glucagon release from three types of islets (Sst^−/−, Kir6.2^−/−, and Sst^−/−/Kir6.2^−/− islets) (Fig. 5G). As expected, the K_ATP channel opener diazoxide inhibited glucagon release of Sst^−/− islets but was ineffective in islets without K_ATP channels (Kir6.2^−/− and Sst^−/−/Kir6.2^−/−). The rate of glucagon release was the lowest in Kir6.2^−/− islets but the highest in Sst^−/−/Kir6.2^−/− islets. The difference in secretion rates cannot be explained by a difference in glucagon content, which was similar in both types of islets (576 ± 76 pg/Kir6.2^−/− islet, n = 5 vs. 544 ± 65 pg/Sst^−/−/Kir6.2^−/− islet, n = 4). This result suggests that SST is responsible for the very low glucagon secretion rate in Kir6.2^−/− islets. On the other hand, the higher rate of secretion in Sst^−/−/Kir6.2^−/− than in Sst^−/− islets suggests that ablation of K_ATP channels stimulates glucagon release if SST is not expressed. Switching from G7 to G1 reversibly stimulated glucagon release of the three islet types (P < 0.05), demonstrating that the inhibitory effect of G7 is independent of SST and K_ATP channels. We next tested the effect of G30. In Sst^−/− islets with K_ATP channels, switching from G7 to G30 strongly stimulated glucagon release (P < 0.05) to a similar extent as gliclazide applied thereafter (Fig. 5H). By contrast, both G30 and gliclazide failed to affect glucagon secretion from Sst^−/−/Kir6.2^−/− islets. This result suggests that the glucagonotropic effect of G30 in the absence of the paracrine influence of SST results from a closure of α-cell K_ATP channels.

The above experiments were performed on perifused isolated islets in which paracrine and endocrine interactions might be different from the in vivo situation. To be as close as possible to the physiological situation, additional experiments were performed on the in situ perfused mouse pancreas, which preserved endocrine and paracrine interactions. These experiments aimed at validating key observations obtained with isolated islets.

We first tested the K_ATP channel dependence of the glucagonostatic effect of G7. In pancreas from C57BL/6 mice perifused with a medium containing a 6 mmol/L mixture of amino acids, increasing the glucose concentration from 1 to 7 mmol/L or the addition of tolbutamide reversibly inhibited glucagon secretion (Fig. 6A and B). In pancreas from mice without K_ATP channels (Sur1^−/− or Kir6.2^−/− mice), the rate of glucagon secretion in G1 was lower than in control mice (Fig. 6C–H). This was not due to differences in pancreatic glucagon content (Supplementary Fig. 5). Switching from G1 to G7 inhibited glucagon secretion, whereas both tolbutamide and diazoxide were ineffective (Fig. 6C–H).

The role of SST was studied by comparing secretion from the pancreas of Sst^+/+ and Sst^−/−mice. The first series of experiments was performed in the presence of a 6 mmol/L mixture of amino acids. Glucagon secretion from Sst^+/+ pancreas was similarly inhibited by a rise of the glucose concentration from 1 to 7 mmol/L and by tolbutamide (Fig. 7A). By contrast, glucagon secretion from Sst^−/− pancreas was inhibited by G7 but stimulated by tolbutamide. Similar results were obtained in the presence of a 2 mmol/L mixture of amino acids (Fig. 7B) and with islets from Sst^−/− mice backcrossed for five generations with C57 mice (Fig. 7C). Like tolbutamide, gliclazide applied to G1 exerted a glucagonostatic effect in Sst^+/+ mice and a glucagonotropic effect in Sst^−/− mice (Fig. 7D).

Our islet perifusion data suggested that removal of SST influence strongly stimulates glucagon release in islets without K_ATP channels (Fig. 5G). We therefore compared glucagon secretion of the perfused pancreas of Kir6.2^−/− and Sst^−/−/Kir6.2^−/− mice (Fig. 7E). Surprisingly, glucagon secretion was similar in the pancreas of both types of mice (with a similar glucagon content) (Supplementary Fig. 5), suggesting that a compensatory inhibitory mechanism is present in the perfused pancreas but not in perifused islets. However, application of CYN154806 + H6056 strongly stimulated glucagon release from Kir6.2^−/− pancreas but was ineffective in Sst^−/−/Kir6.2^−/− pancreas. This demonstrates the strong glucagonostatic effect of SST in Kir6.2^−/− pancreas. Switching from G7 to G1 reversibly stimulated glucagon release in all conditions and in both types of pancreas (P < 0.05). We next investigated whether the glucagonostatic effects of glucose and K_ATP channel blockers were affected by the SSTR antagonists. Acute application of CYN154806 + H6056 reversed the inhibitory effect of tolbutamide (Supplementary Fig. 6). The continuous presence of the antagonists strongly stimulated glucagon release (P < 0.01) (Fig. 7F). The antagonists did not prevent the glucagonostatic effect of G7 but attenuated the glucagonostatic effect of tolbutamide (percentage of inhibition was 31% in the presence of CYN154806 + H6056 and 32% in the presence of CYN154806 vs. 80% in the absence of the SSTR antagonists). However, as for the experiments with perifused islets, none of the SSTR antagonists blocked the effect of SST applied at the end of the perfusion, suggesting that the failure of tolbutamide to stimulate glucagon release in the presence of the SSTR antagonists results from their inability to fully block SST action (Fig. 7F).

We finally compared the effect of high glucose on glucagon secretion of Sst^+/+ and Sst^−/−mice. Switching from G7 to G30 inhibited glucagon release from Sst^+/+ mice (P < 0.05 in MixAA 2 and P < 0.01 in MixAA 6). However, it stimulated that of Sst^−/− mice (P < 0.01) (Fig. 7G and H).

In the current study, we used pharmacological tools and different transgenic mouse strains to investigate the role of SST in the control of glucagon secretion by various glucose concentrations and K_ATP channel blockers. The experiments were performed on isolated islets or the in situ perfused pancreas.

SST inhibits the secretion of many hormones. Pancreatic δ-cells mainly release SST14, whereas intestinal δ-cells mainly release SST28 (30,31). Although SST released from pancreatic δ-cells contributes to <5% of circulating SST (32,33), it is SST locally released by δ-cells that is more important for the control of glucagon secretion (34). α-Cells are more susceptible than β-cells to the local paracrine influence of SST (34,35) and show signs of preferential contacts with δ-cells (36).

The possibility that SST exerts a tonic glucagonostatic effect is not obvious. Indeed, experiments of anterograde perfusion of the pancreas with anti-SST antibodies argue against this hypothesis (37). Moreover, our experiments on the perfused pancreas showed no major difference in absolute rates of glucagon secretion between Sst^+/+ and Sst^−/− mice. However, experiments with inhibitors of SST secretion (13) or antibodies (38,39) support a tonic glucagonostatic effect of SST. This is also strongly supported by the present results and other results (12,14) showing that the rate of glucagon secretion of isolated islets is much higher in islets pretreated with PTx (to block SST action) or in islets from Sst^−/− mice than in control islets, or upon application of SSTR antagonists.

It has been suggested that direct closure of K_ATP channels of α-cells inhibits glucagon release (24,40). The proposed mechanism involves inactivation of low-threshold voltage-dependent channels due to the depolarization of the plasma membrane induced by K_ATP channel closure. Our data support the opposite notion, i.e., that closure of K_ATP channels of α-cells stimulates glucagon release. Indeed, in the absence of the paracrine influence of SST (Sst^−/− mice), K_ATP channel blockers exerted a strong glucagonotropic effect that was more apparent in the absence of amino acids, i.e., when glucagon secretion was not stimulated. Moreover, glucagon secretion was much higher in Sst^−/−/Kir6.2^−/− islets than in Sst^−/− islets. Since we (41) and others (15) observed that K_ATP channel blockers increase [Ca²⁺]_c in α-cells, we think that these agents stimulate exocytosis in α-cells by the same mechanisms as in β-cells.

It has been suggested that in β-cells, high concentrations of tolbutamide (such as those used in several of our experiments) stimulate secretion by an effect that is additional to that resulting from the closure of K_ATP channels and that involves Epac2 (RAPGEF4), a cAMP-dependent effector controlling exocytosis (42). Our experiments indicate that this Epac2-dependent effect is not responsible for the glucagonotropic effect of tolbutamide observed in Sst^−/− islets. Indeed, a low concentration of tolbutamide (10 µmol/L), which barely affects Epac2 (42), stimulated glucagon secretion almost as efficiently as 100 µmol/L of the drug (Fig. 5C–F). Moreover, low concentrations of gliclazide (1–10 µmol/L), a K_ATP channel blocker that does not activate Epac2 (42), potently stimulated glucagon release (Fig. 5D). These results suggest that the glucagonotropic effects of tolbutamide and gliclazide in the absence of the paracrine influence of SST result from their action on K_ATP channels and not on Epac2.

The effect of K_ATP channel blockers was different in the presence of a paracrine influence of SST. Indeed, in control islets, application of tolbutamide or gliclazide had no effect or inhibited glucagon release. Since K_ATP channel blockers stimulate SST release (14) and since SST potently inhibits glucagon secretion, we suggest that SST released upon K_ATP channel closure counteracts the direct stimulatory effect of K_ATP channel blockers on glucagon release. Hence, the net effect of K_ATP channel blockers on glucagon secretion results from a balance between a direct stimulation of α-cells and an indirect inhibition via SST (see model in Fig. 8A). It depends on the rate of glucagon release, which is affected by the glucose concentration. When the rate is already high, the inhibition is more apparent. This is why K_ATP channel closure/ablation tends to inhibit glucagon release in G1. By contrast, when the rate is already low, the stimulation is apparent. This is why K_ATP channel blockers stimulate glucagon release in G7 (14). Consequently, SSTR antagonists, by partially attenuating the effect of SST, will attenuate the glucagonostatic effect of sulfonylureas at G1 (Fig. 7F) and exacerbate their glucagonotropic effect at G7 (Fig. 5F).

It is unclear whether SST is involved or not in the control of glucagon secretion by glucose. Some studies support (12,13) and others do not support such a role (8,14,15). We investigated the involvement of SST in the glucagonostatic effect of glucose using two different experimental models: PTx treatment and genetic invalidation of SST (Sst^−/− mice). Both models yielded convergent results. Increasing the glucose concentration from 0 to 7 mmol/L clearly inhibited glucagon release independently of SST. However, there is also an SST-dependent control of glucagon secretion that contributes gradually more as the glucose concentration increases (see model in Fig. 8B). Indeed, our incubation experiments showed that glucagon secretion of control islets was maximally inhibited at ∼7 mmol/L glucose and that the amplitude of this inhibition remained similar at higher concentrations of the sugar. By contrast, glucagon secretion of PTx-treated islets was maximally inhibited at ∼7 mmol/L glucose, but the amplitude of this inhibition decreased at higher glucose concentrations. This was confirmed by our dynamic perifusion/perfusion experiments that showed that a rise of the glucose concentration from 7 to 20 or 30 mmol/L slightly attenuated glucagon release of control islets but stimulated that of islets/pancreas devoid of the paracrine influence of SST. The attenuation of glucagon release was associated with a stimulation of SST secretion.

The fact that in the absence of the paracrine influence of SST glucose inhibited (G1→G7) or stimulated (G7→G20/G30) glucagon release depending on its concentration suggests that it might activate two distinct mechanisms. The glucagonostatic effect (G7) is likely independent of K_ATP channel closure since G7 inhibited glucagon secretion of the pancreas of Kir6.2^−/− and Sur1^−/− mice. It is also, at least partly, independent of SST since G7 inhibited glucagon release from Sst^−/−/Kir6.2^−/− islets. On the other hand, it might be speculated that the glucagonotropic effect of high glucose concentrations results from a closure of K_ATP channels since genetic ablation of K_ATP channels removed the stimulatory effect of G30 (Fig. 5H). The stronger glucagonotropic effect of G30 with 6 mmol/L than with 2 mmol/L amino acids (Fig. 7H vs. Fig. 7G) might reflect a stronger closure of K_ATP channels in the presence of the highest amino acid concentration.

Several differences were observed between the different types of preparations used in this study. First, in Sst^+/+ mice, tolbutamide and G30 inhibited glucagon secretion from the perfused pancreas, whereas they barely affected glucagon release of isolated islets. This difference cannot be explained by cell culture alone because tolbutamide only tended to inhibit glucagon release of freshly perifused islets. A possible explanation is that the influence of δ-cells is stronger when the solutions are applied via blood vessels (perfused pancreas) rather than around the islets (perifused islets). Application of the solutions via blood vessels may indeed preserve local paracrine interactions between two adjacent cells that are distant from blood vessels. By contrast, the constant flow around perifused islets might easily wash away paracrine signals and reduce paracrine interactions, particularly for cells such as α- and δ-cells that are localized at the islet periphery. Other unknown reasons might also be involved. Second, the rate of glucagon secretion was much higher in Sst^−/− than in Sst^+/+ isolated islets, whereas it was similar in the perfused pancreas of both types of mice. This difference cannot be explained by cell culture because the rate of secretion was also higher in islets freshly isolated from Sst^−/− versus Sst^+/+ mice. It may, however, result from a compensatory inhibition in the perfused pancreas that is lost upon islet isolation. It should be emphasized that these differences between preparations do not at all invalidate our conclusion that in the absence of SST, K_ATP channel blockers and high glucose stimulate glucagon release. We also confirmed our results in Sst^−/− mice backcrossed for five generations on the C57BL/6N background, demonstrating that the observed effects are not strain specific (Fig. 7C).

Five SSTRs (SSTR1–5) encoded by different genes have been described (43). While SSTR1–4 display equal or slightly higher affinity for SST14 than for SST28, SSTR5 preferentially binds SST28 (43). Experiments using SSTR2-selective agonists and antagonists as well as Sstr2^−/− mice have suggested that SSTR2 is the main mediator of SST-induced inhibition of glucagon release (44,45). Moreover, it was reported that SSTR2 is expressed only in α-cells (19,46). This has sometimes led to the belief, erroneously, that SSTR2 is the only SSTR subtype in α-cells and that blockade of SSTR2 fully alleviates the glucagonostatic effect of SST. However, some reports suggest that α-cells also express other SSTRs (31,47,48). Two recent studies of gene expression analysis on purified mouse α- and β-cells showed that SSTR2 is α-cell specific, that SSTR3 is equally expressed in α- and β-cells, and that SSTR1, SSSTR4, and SSTR5 are almost absent in both cell types (19,20). Our gene expression analysis of FACS-sorted α- and β-cells fully confirms these data. These results also suggest that, contrary to previous reports (31,43), SSTR5 is poorly expressed in β-cells.

Our results clearly show that the tonic glucagonostatic effect of SST involves SSTR2 and SSTR3 because antagonists of these receptors strongly stimulated glucagon release of C57BL/6 islets but had no effect on glucagon secretion of Sst^−/− islets. However, we could not get a clear picture of the contribution of both receptors in the glucagonostatic effect of high glucose concentrations or K_ATP channel blockers because the SSTR antagonists failed to fully suppress the glucagonostatic effect of exogenous SST. We cannot exclude that SST also acts on other types of receptors than SSTR2 and SSTR3.

We suggest that the inhibition of glucagon secretion by glucose is mainly independent of SST for low glucose concentrations but starts to involve SST for high concentrations of the sugar (Fig. 8). We also provide evidence that the closure of K_ATP channels controls glucagon secretion by two mechanisms: a direct stimulation of α-cells (independent of Epac2 activation and observed in the absence of SST influence) and an indirect inhibition via SST released from δ-cells. The inhibitory paracrine effect of SST involves several SSTRs, including SSTR2 and SSTR3. Since δ-cells might be electrically coupled to β-cells (49), it is possible that β-cells control glucagon release via SST, although they might also control glucagon secretion independently of δ-cells. Any alteration of the paracrine influence of SST, as reported in diabetes (31,50), might indirectly affect glucagon release.

Acknowledgments. The authors thank Fabien Knockaert (Pôle d’Endocrinologie, Diabète et Nutrition, Université Catholique de Louvain, Brussels, Belgium) for technical help with hormonal assays and Joseph Bryan (Pacific Northwest Diabetes Research Institute, Seattle, WA) for the supply of Sur1^−/− mice.

Funding. B.-K.L. holds a fellowship from the FRIA/FNRS (Brussels, Belgium). J.-C.J. and P.Gi. are Research Directors of the Fonds de la Recherche Scientifique-FNRS (Brussels, Belgium). This study was funded by Actions de Recherche Concertées (13/18-051 from the Communauté Française de Belgique), the Fonds de la Recherche Scientifique Médicale (PDR grant T.0124.15), European Foundation for the Study of Diabetes (EFSD)/Boehringer Ingelheim, and the Société Francophone du Diabète (Paris, France) (grants 2015 and 2017).

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

Author Contributions. B.-K.L., H.C., A.G.-R., P.C., P.Ga., and N.A. designed and performed the experiments and contributed to the discussion. C.B., J.-C.J., and S.S. provided material and advice for experiments and reviewed and edited the manuscript. V.S. provided material. P.Gi. conceived the study, designed the experiments, and wrote the manuscript. All authors approved the manuscript. P.Gi. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 51st Annual Meeting of the European Association for the Study of Diabetes, Stockholm, Sweden, 14–18 September 2015.