Glucose- and Hormone-Induced cAMP Oscillations in α- and β-Cells Within Intact Pancreatic Islets

Adrenaline, glucagon, GLP-1, 2′5′-dideoxyadenosine (DDA), EGTA, poly-l-lysine, 3-isobutylmethylxanthine (IBMX), somatostatin, and HEPES were from Sigma. Catechol estrogen (CE) [1,3,5(10)-estratrien-2,3,17-β-triol] for AC inhibition and the inactive control compound 1,3,5(10)-estratrien-3,17-β-triol were from Makaira Limited (London, U.K.). KH7 was a gift from Drs. J. Buck and L. Levin, Weill Medical College of Cornell University, New York, NY. The acetoxymethyl esters of the Ca²⁺ indicators Fura-PE3 and Fura Red were obtained from TEFLabs (Austin, TX) and Invitrogen, respectively. Adenoviruses expressing a fluorescent cAMP biosensor have previously been described (12).

Islets of Langerhans were isolated from C57BL6J mice as previously described (22). The Uppsala ethics committee approved all animal experimental procedures. When the pancreas was excised, the lower duodenal part was omitted to reduce contribution of islets with cells producing pancreatic polypeptide (37). After isolation, the islets were cultured for 1–4 days in RPMI-1640 medium containing 5.5 mmol/L glucose, 10% fetal calf serum, 100 μg/mL penicillin, and 100 μg/mL streptomycin at 37°C in an atmosphere of 5% CO₂ in humidified air. Some experiments were performed on single cells prepared by shaking the freshly isolated islets in a Ca²⁺-deficient medium followed by attachment to coverslips during 2–5 days of culture. If not otherwise stated, data were obtained with cells from at least three independent islet isolations. Where indicated, experiments were performed on insulin-secreting MIN6 β-cells cultured as previously described (12). For [cAMP]_pm measurements, the islets or cells were infected with cAMP biosensor-encoding adenoviruses using 10⁵ fluorescence forming units (FFU)/islet or 60 FFU/cell as previously described (12). It was verified that the virus infection did not affect the islet [Ca²⁺]_i response to glucose stimulation.

The islets or cells were transferred to a buffer containing (in mmol/L) 125 NaCl, 4.8 KCl, 1.3 CaCl₂, 1.2 MgCl₂, 3 glucose, and 25 HEPES (pH adjusted to 7.40 with NaOH) and incubated for 30 min at 37°C. In some experiments, this medium was supplemented with 20 mmol/L NaHCO₃ after equimolar reduction of NaCl. For Ca²⁺ recordings, the cells were preincubated in the presence of 1 μmol/L of the acetoxymethyl ester of Fura-PE3 and islets with 10 μmol/L of the acetoxymethyl ester of Fura Red, respectively. Immediately after preincubation, the islets were applied onto a poly-l-lysine–coated 25-mm coverslip and allowed to attach. [Ca²⁺]_i imaging with Fura-PE3 was performed at 37°C with an epifluorescence microscope setup as previously described (5). Measurements of the cytoplasmic Ca²⁺ concentration beneath the plasma membrane ([Ca²⁺]_pm) were made with Fura Red in a total internal reflection fluorescence (TIRF) setup consisting of an Eclipse Ti microscope with a 60× 1.45 NA-objective (Nikon), using 488 nm excitation with fluorescence detection at >630 nm. Using the same TIRF system, [cAMP]_pm was measured in intact islets or cells expressing the cAMP translocation biosensor as previously described (12). Image pairs were acquired every 5 s using a back-illuminated EMCCD camera (DU-887; Andor Technology, Belfast, Northern Ireland) controlled by MetaFluor software (Molecular Devices, Downingtown, PA). The cAMP concentration was expressed as the background-corrected cyan fluorescent protein–to–yellow fluorescent protein (CFP-to-YFP) ratio. The basal ratio was normalized to unity for compensation for the variability in expression levels.

Individual cells in the experimental chamber were identified by immunostaining for insulin or glucagon. Following the fluorescence recordings, the cells were rinsed with 25, 50, and 75% ethanol and eventually fixed in 95% ethanol for 5 min. After sequential rinsing with 3% H₂O₂ and Tris buffer (0.05 mol/L, pH 7.4), serum-free protein block (DAKO) was added to reduce background staining. Polyclonal rabbit anti-glucagon or guinea pig anti-insulin (1:100; DAKO) was added after 10 min and allowed to incubate for 30 min. After rinsing with Tris buffer, the MACH 3 rabbit probe alkaline phosphatase polymer kit (Biocare Medical, Concord, CA) was used for visualization according to the manufacturer’s instructions. The reaction was stopped by addition of Tris buffer. After rinsing with distilled water, cell nuclei were stained with hematoxylin (Histolab) for 0.5–2 min.

Triplicate vials, each containing 10⁴ MIN6 β-cells and Krebs-Ringer bicarbonate–HEPES buffer (100 μL) supplemented with 3.3 mmol/L (94 GBq/mol) or 11.1 mmol/L (28 GBq/mol) D-[U-¹⁴C]glucose, were incubated with or without test compounds for 90 min at 37°C under an atmosphere of 95/5% O₂/CO₂ during slow shaking. Metabolism was then stopped by addition of 17 μmol/L antimycin A (Sigma-Aldrich) in ethanol. The generated ¹⁴CO₂ was trapped in 250 μL hyamine 10× (Perkin-Elmer) during incubation at 37°C for 2 h as previously described (38). Radioactivity was measured in a liquid scintillation counter after addition of 5 mL Ultima Gold scintillation fluid (Perkin-Elmer).

Image analysis was made using the MetaFluor or ImageJ (W.S. Rasband, National Institutes of Health, http://rsb.info.nih.gov/ij) software. [cAMP]_pm and [Ca²⁺]_i response magnitudes were calculated from time-average data obtained before and during the stimulation period. Data are presented as means ± SEM. Statistical comparisons were assessed with Student t test.

Figure 1A illustrates TIRF measurements on peripheral mouse islet cells expressing the cAMP biosensor, and Fig. 1B shows TIRF images of the fluorescence-labeled subunits of the cAMP indicator. The CFP-to-YFP ratio was usually stable in peripheral islet cells exposed to 3 mmol/L glucose (Fig. 1C). Cell size and adrenaline responses have previously been used to discriminate between α- and β-cells because only the relatively small α-cells expressing α₁- and β-adrenoceptors (24) respond with intracellular mobilization of Ca²⁺ (39). Because the β-adrenoceptors mediate cAMP elevation in α-cells and the larger β-cells express cAMP-lowering α₂-adrenoceptors (40), we tested whether the cAMP response to adrenaline could be used for cell identification. This was indeed the case, and all of 25 cells with small TIRF footprints that responded to adrenaline with [cAMP]_pm elevation stained positively for glucagon. In contrast, all of 30 cells with large footprint and negative adrenaline response were insulin positive. Figure 1C and D shows that an immune-identified α-cell responded to adrenaline with elevation of [cAMP]_pm, whereas a β-cell (immunostaining not shown) did not respond under these conditions with basal [cAMP]_pm levels. The [cAMP]_pm-lowering effect of adrenaline on β-cells became apparent under conditions where the levels of cAMP were elevated (Fig. 2A and D).

When mouse islets were exposed to glucagon in the presence of 3 mmol/L glucose, there was a prompt increase in [cAMP]_pm with sustained elevation or oscillations in β-cells (36 of 41 [Fig. 2A]) and α-cells (5 of 8 [Fig. 2C]). In β-cells, the integrated response was half-maximal and maximal at 80 and 1,000 nmol/L glucagon, respectively (Fig. 2B). Somatostatin suppressed the glucagon-induced [cAMP]_pm elevation in both β-cells (Fig. 2D) and α-cells (data not shown). Similar responses were seen with the α₂-adrenoceptor agonist clonidine, which lowered [cAMP]_pm in all of 11 β-cells (Fig. 2E) and 6 of 8 α-cells (Fig. 2F), indicating that both cell types express α₂-receptors. However, adrenaline still caused elevation of [cAMP]_pm in the α-cells, consistent with domination of β-adrenoceptors in this cell type (40). At concentrations of 0.5–10 nmol/L, GLP-1 induced [cAMP]_pm oscillations in β-cells (Fig. 3A and D). The oscillatory pattern varied between different cells with frequencies ranging from 0.18 to 2.1 min⁻¹ and amplitudes from 0.06 to 1.5 normalized ratio units. The amplitude and duration of individual oscillations tended to increase with GLP-1 concentration, as illustrated in Fig. 3A. High concentrations of GLP-1 (>50 nmol/L) induced prompt and sustained [cAMP]_pm elevation in β-cells (19 of 23) that rapidly reversed upon omission of the peptide (Fig. 3B). Figure 3C shows the concentration dependence of the integrated [cAMP]_pm response to GLP-1, with half-maximal and maximal effects at 30 and 300 nmol/L, respectively. The [cAMP]_pm oscillations induced by GLP-1 were strikingly well synchronized among β-cells in the islet even between those lacking direct contact (Fig. 3D). Only occasional α-cells responded to GLP-1 with elevation of [cAMP]_pm (5 of 48), and Fig. 3E shows a graded response to 5–50 nmol/L GLP-1 in one of these α-cells identified with adrenaline. To investigate whether the lack of response in many α-cells was because cAMP did not increase above the detection threshold, we exposed the cells to 5 μmol/L IBMX, which is at the border to give a detectable increase of [cAMP]_pm. Of 21 adrenaline-identified α-cells, 10 responded to 5 μmol/L IBMX. Only two of those plus two of the IBMX-unresponsive α-cells reacted to 100 nmol/L GLP-1, and the [cAMP]_pm increase was of a magnitude similar to that induced by adrenaline (Fig. 3F).

When the glucose concentration was elevated from 3 to 11, 20, or 30 mmol/L, most islet β-cells responded with a pronounced rise of [cAMP]_pm followed by a stable plateau (16 of 41) or slow oscillations with frequencies in the 0.05–0.4/min range and amplitudes of 0.05–0.6 ratio units (21 of 41) (Fig. 4A and B). The remaining cells showed a single [cAMP]_pm peak followed by return to the baseline (not shown). As previously described in isolated β-cells (12), the glucose-induced elevation of [cAMP]_pm was preceded by a small reduction in many cells. This phenomenon is illustrated in Fig. 4C. Figure 4A–D also shows that adrenaline abolished the glucose-induced elevation of [cAMP]_pm. Like the GLP-1–triggered oscillations of [cAMP]_pm, those induced by glucose were often synchronized among β-cells within an islet (Fig. 4D). The glucose and hormone effects were additive, and the [cAMP]_pm elevation induced by a combination of 20 mmol/L glucose and 10 nmol/L glucagon or 100 nmol/L GLP-1 was larger than the effect of either stimulus alone (Fig. 4E and F).

The amplitudes of the [cAMP]_pm oscillations were significantly suppressed, but the oscillations often persisted after removal of Ca²⁺ and addition of 2 mmol/L EGTA to prevent changes in [Ca²⁺]_i (4 of 6 cells) (Fig. 5A). The [Ca²⁺]_i response of intact islets differs from that of isolated cells in displaying a pattern of rapid (2–4/min) regular oscillations that depends not only on Ca²⁺ influx but also on release from intracellular stores (2). Corresponding rapid oscillations of [cAMP]_pm were never observed, even after elevating the extracellular Ca²⁺ concentration from 1.3 to 2.6 mmol/L (data not shown), which typically promotes the appearance of the fast Ca²⁺ pattern. Simultaneous measurements of [cAMP]_pm and [Ca²⁺]_pm in islet β-cells loaded with the fluorescent indicator Fura Red demonstrated that [cAMP]_pm and [Ca²⁺]_pm elevations were synchronized (Fig. 5B), but the temporal relationship of the initial glucose-induced [cAMP]_pm and [Ca²⁺]_pm elevations varied with [cAMP]_pm, sometimes increasing before (Fig. 5C) and sometimes after (Fig. 5B) [Ca²⁺]_pm. Together, these findings indicate that Ca²⁺ amplifies the [cAMP]_pm response but that [Ca²⁺]_i oscillations are not essential for glucose-induced oscillations of [cAMP]_pm.

In both β-cells within islets (n = 4) (Fig. 6A) and MIN6 β-cells (n = 37) (Fig. 6B), the glucose-induced [cAMP]_pm oscillations were completely inhibited by 30 μmol/L KH7, which has been described as a specific sAC inhibitor (41). Similar inhibition was obtained with 100 μmol/L of the tmAC inhibitor DDA in clonal (Fig. 6B) and primary (n = 5) (Fig. 6C) β-cells. Also, the [cAMP]_pm elevation stimulated by 50 nmol/L GLP-1 was inhibited by both DDA and KH7 (n = 12) (Fig. 6D). However, whereas DDA was without effect on glucose-induced [Ca²⁺]_i oscillations, KH7 completely abolished this response (n = 8) (Fig. 6E), indicating that the compound might interfere with upstream effects in glucose stimulus-secretion coupling. Indeed, KH7 (30 μmol/L) markedly suppressed glucose oxidation in MIN6 cells. This effect could not be attributed to lowering of cAMP because the effect was not reversed by addition of 10 mmol/L of the membrane-permeable analog 8-Br-cAMP. In contrast, a slight tendency of DDA to inhibit glucose oxidation was completely reversed by 8-Br-cAMP (Fig. 6F).

CE is an alternative inhibitor of sAC (42), and this compound was without effect on glucose-induced [Ca²⁺]_i oscillations (Fig. 6E). In 60% of the MIN6 β-cells, CE caused temporary interruption of the [cAMP]_pm oscillations, but after a few minutes, there were [cAMP]_pm oscillations indistinguishable from those under control conditions (Fig. 6G). The interruption probably was not mediated by changes in cAMP, given that it was observed also with the inactive control compound 1,3,5(10)-estratrien-3,17-β-triol (data not shown). Similar results were obtained when the experiments were performed in medium containing 20 mmol/L NaHCO₃, which promotes sAC activity (Fig. 6H).

In the presence of 3–7 mmol/L glucose, [cAMP]_pm was low and stable in α-cells identified by the adrenaline-induced [cAMP]_pm elevation. However, like in β-cells, 11–30 mmol/L glucose triggered [cAMP]_pm oscillations in many α-cells (3 of 11 at 11 mmol/L, 13 of 28 at 20 mmol/L, and 8 of 17 at 30 mmol/L glucose) (Fig. 7A). In some α-cells, the nadirs and peaks of the oscillations reflected inhibition and stimulation, respectively, in relation to the baseline (Fig. 7B). In the few islets that permitted analysis of the phase relationship of [cAMP]_pm oscillations between different α-cells (n = 1) and between α-cells and β-cells (n = 4), there was no apparent synchronization of the oscillations (Fig. 7C and D).

The current study uses a TIRF microscopy approach to investigate cAMP signaling in intact pancreatic islets. The method images the plasma membrane and submembrane cytoplasm of the different types of cells in the islet mantle in direct contact with the coverslip. Although innervation and normal microcirculation is lacking, the preparation allows analysis of cAMP signaling in cells able to interact within the micro-organ. Unfortunately, immunostaining could not be used to identify particular cells within an islet after the TIRF recording because the islets already detach from the coverslip during attempted fixation. Free islet cells attach more firmly, and it became apparent that immune-identified α- and β-cells displayed distinct [cAMP]_pm responses to adrenaline, reflecting differences in adrenoceptor expression (40), and these responses were used for identification of the two cell types in the islet. Islets were isolated only from the parts of the pancreas where PP-cells are scarce (37), and it seems unlikely that δ-cells responding to adrenergic receptor activation with lowering of [cAMP]_pm incorrectly have been taken for β-cells. The δ-cells are much less abundant and smaller than the β-cells. Small cells and cells with small footprints were not taken as β-cells irrespective of response.

As expected, glucagon and GLP-1 evoked cAMP signaling in the islets. However, whereas GLP-1 triggered [cAMP]_pm oscillations in most β-cells, only occasional α-cells responded to the hormone with elevation of [cAMP]_pm. It has been controversial whether GLP-1 receptors are present in α-cells. Some studies of rat α-cells have failed to demonstrate GLP-1 receptor expression (43,44) while others have shown that GLP-1 increases the cAMP content (35) and exocytosis (35,45). The present data from mouse α-cells is in line with observations that GLP-1 receptors are expressed only in a small subpopulation of rat (44) and mouse (36) α-cells. It is well accepted that GLP-1 is an inhibitor of glucagon secretion, but the underlying signaling is unclear. It was recently suggested that very small cAMP elevations inhibit glucagon secretion by a protein kinase A (PKA)-dependent mechanism (36). However, from the present data, GLP-1 inhibition of glucagon secretion seems to involve mechanisms other than changes of [cAMP]_pm.

Glucagon elicited [cAMP]_pm oscillations in most α- and β-cells. Given that cAMP is a positive modulator of glucagon secretion (23,24,34,35), our observations support a positive autocrine feedback effect of the hormone. The reason for the absence of [cAMP]_pm oscillations in low glucose when glucagon secretion is stimulated is currently unknown. In islet β-cells, the cAMP responses to glucagon and GLP-1 were significantly enhanced by high glucose, as expected from earlier biochemical studies (17). Glucose alone also induced pronounced changes in [cAMP]_pm, reminiscent of those of [Ca²⁺]_i, with an initial lowering followed by a prominent rise. This rise may reflect both direct effects of glucose on the β-cell and an early peak of glucagon secretion within the islet (32,33). The period of the subsequent oscillations was usually several minutes, as in isolated cells (12), and faster oscillations typical for islet recordings of membrane potential and [Ca²⁺]_i were not observed. Because Ca²⁺ is known to increase islet cAMP via Ca²⁺/calmodulin-sensitive ACs (17), it seems possible that the changes of [cAMP]_pm are driven by those of [Ca²⁺]_i (46). However, like in isolated β-cells (12), [cAMP]_pm oscillations in intact islets often persisted when glucose-stimulated Ca²⁺ entry was prevented, and the cAMP signal sometimes preceded that of Ca²⁺ during the initial glucose-induced response. Comparable findings were reported using transgenic mouse islets expressing another cAMP sensor, with glucose-induced stable cAMP elevation preceding the increase of [Ca²⁺]_i and persisting in the absence of extracellular Ca²⁺ (13). There are mutual interactions between cAMP and Ca²⁺, and both signals depend on intracellular ATP (12). ATP concentrations are in turn affected by Ca²⁺, and interestingly, oscillations in β-cell metabolism (47) show a Ca²⁺ dependence reminiscent of that of the presently observed cAMP oscillations. In light of these interactions, it is not surprising that there are slight variations in timing between the Ca²⁺ and cAMP signals.

Because the tmACs expressed in islets have in vitro K_m values for ATP in the submillimolar range (48), it is not clear how they would be regulated by the approximately 10-fold higher concentrations of ATP in the cytoplasm. We now considered the possibility that glucose stimulates cAMP formation via sAC, which is regulated by bicarbonate, Ca²⁺, and millimolar concentrations of ATP (18,19,49) and which was recently proposed to be present in insulin-secreting cells (20). However, the small molecule inhibitor KH7 that is often used to define sAC activity (20,41) showed deleterious effects on β-cell metabolism, limiting its usefulness for studies of sAC in intact cells. The less selective sAC inhibitor CE failed to provide evidence that sAC mediates glucose-induced cAMP formation in the subplasma membrane space of insulin-secreting cells. Future studies will have to clarify which particular isoforms of tmACs sense changes in β-cell metabolism and how the additive effects of glucose and hormones on [cAMP]_pm are mediated. Interestingly, although glucose is an inhibitor of glucagon secretion and cAMP stimulates exocytosis, glucose also triggered cAMP oscillations in α-cells. However, glucose-inhibited glucagon secretion is pulsatile, and the reduced time-average secretion in both rat (32) and human (33) islets is composed of alternating periods of inhibition and stimulation. The present findings are consistent with cAMP being involved in the stimulatory component, perhaps also explaining how very high glucose concentrations can paradoxically increase glucagon release (31).

Pulsatile hormone secretion from the intact islet obviously requires synchronization of the secretory activity among the cells within the islet. Functional coupling was indeed observed in GLP-1–stimulated β-cells, but the underlying mechanism is unknown. However, gap junction–mediated diffusional exchange and electrical coupling and the stimulatory action of Ca²⁺ on cAMP formation are factors favoring cAMP signal propagation and synchronization between cells within functional islet subdomains. The synchronization was less striking during glucose stimulation, possibly because paracrine influences become perturbed by insufficient exchange of superfusion medium around the islet cells adhering to the coverslip. Not only do glucose and glucagon independently evoke [cAMP]_pm oscillations in α- and β-cells, but somatostatin, whose secretion is stimulated by glucose and glucagon (50), suppresses cAMP levels in these cells. In summary, we found that the cAMP signaling properties of isolated β-cells are maintained when they are located within the intact pancreatic islet. Moreover, we discovered that glucose and glucagon induce [cAMP]_pm oscillations in α-cells. Clarification of the temporal relationship of cAMP signaling in α- and β-cells will help to explain the generation of antisynchronous pulses of insulin and glucagon underlying the 20- to 100-fold changes in insulin-to-glucagon ratio (33) that determine glucose homeostasis in the liver.

This work was supported by a European Foundation for the Study of Diabetes/MSD grant as well as by grants from the Family Ernfors Foundation, the Magnus Bergvall’s Foundation, the Swedish Diabetes Association, and the Swedish Research Council (67X-14643, 67P-21262, and 12X-6240). This work was also supported by a grant from the Novo Nordisk Foundation.

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

G.T. performed all of the experiments, analyzed data, and reviewed and edited the manuscript. S.S. contributed to glucose oxidation experiments and discussion of results and reviewed and edited the manuscript. E.G. contributed to discussion of results and reviewed and edited the manuscript. A.T. contributed to discussion and wrote the manuscript.

The authors thank Drs. Jochen Buck and Lonnie Levin (Department of Pharmacology, Weill Medical College, Cornell University, Ithaca, New York) for the gift of KH7 and Jun-Ichi Miyazaki (Division of Stem Cell Regulation Research, Osaka University Medical School, Osaka) for MIN6 β-cells. The authors are indebted to Ing-Marie Mörsare, Ing-Britt Hallgren, and Heléne Dansk (Department of Medical Cell Biology, Uppsala University, Uppsala, Sweden) for skillful technical assistance.