Glucose and Pharmacological Modulators of ATP-Sensitive K⁺ Channels Control [Ca²⁺]_c by Different Mechanisms in Isolated Mouse α-Cells

Most experiments were performed with our mouse models expressing EYFP specifically in α- or β-cells and referred to as GYY and RIPYY mice, respectively (19). NMRI mice were used as controls. The study was approved by our Commission d'Ethique d'Experimentation Animale.

Islets were obtained by collagenase digestion of the pancreas, and single cells were prepared by dispersion in a Ca²⁺-free medium. Islet cells were cultured for 1–4 days on coverslips in RPMI 1640 containing 7 mmol/l glucose.

The extracellular solution contained 120 mmol/l NaCl, 4.8 mmol/l KCl, 1.5 mmol/l CaCl₂, 1.2 mmol/l MgCl₂, 24 mmol/l NaHCO₃, and 1 mg/ml BSA (pH 7.4). It was gassed with O₂:CO₂ (94:6%). The 2.5-mmol/l amino acid mixture used in some experiments contained 0.5 mmol/l alanine, 0.5 mmol/l leucine, 0.75 mmol/l glutamine, and 0.75 mmol/l lysine. For I_KATP and membrane potential recordings, the extracellular medium was devoid of BSA and supplemented with 5 mmol/l HEPES. Pipette solution contained 70 mmol/l K₂SO₄, 10 mmol/l NaCl, 10 mmol/l KCl, 3.7 mmol/l MgCl₂, and 5 mmol/l HEPES (pH 7.1).

To identify β-cells from GYY mice, islet cells were infected with the AdRIPBgliDsRed adenovirus ensuring a β-cell specific expression of DsRed (a description is available in an online appendix at http://dx.doi.org/10.2337/db07-1298.).

Cells expressing EYFP (excitation, 490 nm; emission, 535 nm) or DsRed (excitation, 540 nm; emission, 610 nm) were first selected. [Ca²⁺]_c (fura-PE3 or fura-2) and NAD(P)H fluorescences were monitored at 37°C as described previously (19). It was verified that EYFP fluorescence did not contaminate [Ca²⁺]_c and NAD(P)H signals. I_KATP was recorded at 31–32°C in the perforated mode by applying 100-ms-duration pulses of ±20 mV from a holding potential of −80 mV as reported previously (20). Membrane potential measurements were performed at 33°C in the perforated mode in current-clamp. Glucagon secretion from batches of 200 islets of GYY mice was monitored in perifusion experiments as described previously (19).

Data are shown as representative traces or means ± SE of results obtained with the indicated number of cells or batches of 200 islets (Fig. 7D only) from at least three different cultures. The statistical significance of differences between means was assessed by paired Student's t test.

Because EYFP and fura-PE3 excitation spectra do not overlap, [Ca²⁺]_c could be easily monitored in EYFP-expressing α-cells. In the presence of 0.5 mmol/l glucose, [Ca²⁺]_c oscillated in 31% (175 of 555) of α-cells (Fig. 1B) and was stable at basal levels in the others (Fig. 1A). Non-oscillating and oscillating α-cells were equally responsive to arginine and adrenaline, indicating that they are both physiologically normal (not shown). Increasing the glucose concentration from 0.5 (G0.5) to 15 mmol/l (G15) did not affect [Ca²⁺]_c in non-oscillating α-cells (Fig. 1A; n = 21) and slightly decreased it in oscillating α-cells (Fig. 1B). In the latter group, average [Ca²⁺]_c integrated over the last 17 min of perifusion with G15 was 28% lower than average [Ca²⁺]_c in G0.5 (n = 35, P < 0.05).

We next tested the effect of glucose on α-cell [Ca²⁺]_c in the presence of a 2.5-mmol/l mixture of amino acids that potentiate glucagon secretion (21). This mild stimulatory condition increased the proportion of α-cells displaying [Ca²⁺]_c oscillations in G0.5 to 70% (180 of 255), which made it easier to study the inhibitory effect of high glucose. Increasing the glucose concentration from 0.5 to 15 mmol/l decreased [Ca²⁺]_c (integrated over the last 27 min in G15) to a similar extent (by 24%; n = 54, P < 0.05) as without amino acids (Fig. 1C). We also monitored [Ca²⁺]_c in NMRI α-cells identified by their response to adrenaline applied at the end of the experiment (17,22). In the presence of G0.5, [Ca²⁺]_c oscillated in 20% (16 of 82) of α-cells in the absence of amino acids and in 67% (67 of 100) of α-cells in the presence of 2.5-mmol/l amino acid mixture. The effect of G15 on [Ca²⁺]_c in NMRI α-cells was similar to that observed in GYY α-cells (Fig. 1D), with a 32% drop in average [Ca²⁺]_c (integrated over the last 27 min in G15) compared with initial [Ca²⁺]_c in low glucose (n = 46, P < 0.01).

[Ca²⁺]_c oscillations had very heterogeneous patterns, being irregular or mixed, composed of oscillations of various amplitude and frequency (Fig. 2A and beginning of Fig. 3B and E). The pattern was not obviously affected by glucose or amino acids. The oscillations resulted from concomitant membrane potential oscillations with bursts of spikes (Fig. 2B). The electrical spiking involved Ca²⁺ channels because it was abolished in a Ca²⁺-free medium (Fig. 2C).

We next tested modulators of K_ATP channels on α-cell [Ca²⁺]_c. When [Ca²⁺]_c was low and stable in G0.5, addition of 10 μmol/l tolbutamide increased [Ca²⁺]_c (Fig. 3A). Subsequent addition of 100 μmol/l diazoxide reversed the effects of tolbutamide (Fig. 3A). When [Ca²⁺]_c was oscillating in low glucose, 100 μmol/l diazoxide decreased [Ca²⁺]_c to basal levels, and this effect was reversed by 100 μmol/l tolbutamide (Fig. 3B). Similar results were observed in the presence of 2.5-mmol/l amino acid mixture (19). Low diazoxide concentrations (1–3 μmol/l) failed to affect oscillating or basal [Ca²⁺]_c (Fig. 3C). Diazoxide was also ineffective at 10 μmol/l in non-oscillating cells perifused with G15 (n = 8, not shown). Figure 3D shows that 100 μmol/l diazoxide prevented 10 mmol/l arginine from increasing [Ca²⁺]_c in α-cells. Overall, these data indicate that [Ca²⁺]_c oscillations in the presence of glucose alone and Ca²⁺ influx induced by arginine occur only when most K_ATP channels are closed.

We next tested whether α-cell [Ca²⁺]_c was affected by perturbation of cell metabolism. In the presence of 2.5-mmol/l amino acid mixture, the mitochondrial poison azide reversibly abolished spontaneous [Ca²⁺]_c oscillations occurring in G0.5 and lowered [Ca²⁺]_c to basal levels. Subsequent closure of K_ATP channels with 500 μmol/l tolbutamide reversed this inhibition (Fig. 3E). A similar effect of azide was observed in the absence of amino acids and occurred in glucose-stimulated β-cells (not shown). The comparable response of α- and β-cells to azide suggests that the membrane potential is influenced by metabolism in both cell types. We therefore compared their I_KATP.

EYFP-expressing cells from GYY mice had a capacitance of 4.5 ± 0.19 pF/cell (n = 42) versus 6.77 ± 0.37 pF/cell (n = 18) for those from RIPYY mice. These results agree with previous reports showing that α-cells are smaller than β-cells (14,23,24).

I_KATP was measured in the perforated mode of the patch-clamp technique in α- and β-cells from GYY mice and in β-cells from RIPYY mice. For some experiments, islet cells from GYY mice were infected with the recombinant adenovirus AdRIPBgliDsRed 2 days before the experiments to permit easy identification of β-cells (red) and α-cells (yellow) before recordings. Insulin immunodetection showed that DsRed was exclusively targeted to β-cells (supplementary Fig. 1A–C, available in the online appendix) but that only ∼45% β-cells were fluorescent for DsRed (43 of 94). Importantly, DsRed and EYFP were consistently expressed in distinct cell types (supplementary Fig. 1D).

We first compared I_KATP in α- and β-cells from GYY mice. During perifusion with G15, I_KATP was small in α-cells (39.5 ± 9.5 pS/pF, n = 5) and β-cells (21.5 ± 3.5 pS/pF, n = 5) (Fig. 4A and B). The maximal density of I_KATP was estimated by perifusing the cells with 250 μmol/l diazoxide and 1 mmol/l sodium azide. I_KATP was 30% smaller (but not statistically different) in α- (490 ± 72 pS/pF) than β-cells (695 ± 130 pS/pF). Closure of K_ATP channels by 250 μmol/l tolbutamide in the presence of G15 decreased I_KATP to slightly lower, but not significantly different, values (29.7 ± 7 pS/pF in α-cells and 15.2 ± 1.5 pS/pF in β-cells) than those measured in G15 alone (Fig. 4A and B). This indicates that in the sole presence of G15, most, although not all, K_ATP channels are closed in α-cells and in β-cells (25). To determine whether infection had affected the current density, we performed similar experiments in noninfected α-cells from GYY mice. No effect of infection was observed: I_KATP was 34.2 ± 4.5 pS/pF (n = 6) in G15, 432 ± 92 pS/pF in the presence of diazoxide and azide, and 23 ± 3 pS/pF after addition of 250 μmol/l tolbutamide (not shown). Overall, these results highlight the similarities of I_KATP between α- and β-cells.

We next investigated whether glucose influences α-cell I_KATP. To validate our measurements, I_KATP was first recorded in β-cells from RIPYY mice. As expected, the large I_KATP in G0.5 was reversibly inhibited from 242 ± 72 to 27.5 ± 8 pS/pF by G15 (n = 5; Fig. 4C) and from 211 ± 106 to 30 ± 15 pS/pF by 250 μmol/l tolbutamide (n = 4, not shown). By contrast, glucose did not affect I_KATP in α-cells (38.5 ± 8.7 and 39.2 ± 9.5 pS/pF (n = 5) in G0.5 and G15, respectively; Fig. 4D). However, α-cell I_KATP was reduced by 30% by 250 μmol/l tolbutamide (from 33.2 ± 8.7 to 21.0 ± 1.7 pS/pF, n = 12, not statistically different, not shown) and increased by diazoxide (from 38.5 ± 8.7 to 141 ± 51 pS/pF, n = 5; Fig. 4D). In the presence of G15, 2 mmol/l azide reversibly increased I_KATP from 33 ± 8.2 to 138 ± 39 pS/pF (Fig. 4E; n = 5, P < 0.05) indicating that K_ATP channels in α-cells can be controlled by changes in cell metabolism.

NAD(P)H fluorescence can be used to monitor nutrient-induced changes in β-cell metabolism. We therefore compared the influence of glucose on NAD(P)H fluorescence in isolated α-cells and EYFP-negative islet cells (most of them presumably being β-cells). Figure 5A shows that increasing the glucose concentration from 0.5 to 15 mmol/l induced a reversible increase of NAD(P)H fluorescence in EYFP-negative cells (β-cells) (n = 14) but had no effect on α-cell NAD(P)H fluorescence (n = 30). Azide, which blocks the electron transport chain and inhibits NAD(P)H oxidation, evoked a small increase of NAD(P)H fluorescence in α- and β-cells. Similar results were observed in the presence of 2.5-mmol/l amino acid mixture (not shown).

The same experiment was performed in NMRI islet cells. After each NAD(P)H measurement, cells were loaded with fura-2 on the stage of the microscope, and the [Ca²⁺]_c response to adrenaline was monitored to distinguish α-cells from non–α-cells. Figure 5B shows that G15 reversibly raised NAD(P)H fluorescence in adrenaline-nonresponsive cells (most of them presumably being β-cells, n = 18), whereas it barely affected NAD(P)H fluorescence in adrenaline-responsive cells (presumably α-cells, n = 33). These results suggest that oxidative metabolism, hence ATP synthesis, is not significantly accelerated by high glucose in α-cells and may explain the lack of effect of glucose on I_KATP in α-cells.

The above-described experiments showed that glucose slightly decreased [Ca²⁺]_c without affecting I_KATP. Our conclusion is supported by experiments performed after maximal closure of K_ATP channels with 500 μmol/l tolbutamide. Increasing the glucose concentration from 0.5 to 15 mmol/l under these conditions again induced a 42% drop of average [Ca²⁺]_c (Fig. 6A; n = 26, P < 0.01). To determine whether this [Ca²⁺]_c decrease results from a direct inhibition of VDCCs, glucose was tested in α-cells depolarized with 30 mmol/l K⁺. Under these conditions, [Ca²⁺]_c was steadily elevated because of the forced opening of VDCCs, and glucose was ineffective (Fig. 6B; n = 19).

We tested the effect of three candidate paracrine factors released by β-cells, GABA, zinc, and insulin, on α-cell [Ca²⁺]_c in a medium containing G0.5 and 2.5-mmol/l amino acid mixture. GABA (100 μmol/l) and zinc (3 and 30 μmol/l) did not affect [Ca²⁺]_c except for an initial, small, transient increase by zinc (Fig. 7A and B). Addition of 100 nmol/l insulin decreased [Ca²⁺]_c in only 5 of 21 cells, leaving [Ca²⁺]_c unchanged in the others. On average, insulin slightly decreased [Ca²⁺]_c from 234 ± 10 to 212 ± 12 nmol/l (Fig. 7C; n = 21). However, it failed to affect glucagon secretion and did not prevent the strong (70%) inhibitory effect of glucose on glucagon secretion (Fig. 7D).

We finally evaluated the importance of Ca²⁺ influx through L-type VDCCs in α-cells. As illustrated by Fig. 8, 1 μmol/l nimodipine inhibited the [Ca²⁺]_c increase occurring during the spontaneous [Ca²⁺]_c oscillations in G0.5 (by 74%) or induced by tolbutamide or arginine (by 85% for both agents). By contrast, as expected (22), the drug did not prevent the [Ca²⁺]_c elevation elicited by 10 μmol/l adrenaline.

In this study, we used our GYY mouse model expressing EYFP in α-cells (19) to study the mechanisms by which glucose controls α-cell [Ca²⁺]_c. The validation of our GYY mouse is extended here by the similarity of key results in GYY α-cells and NMRI α-cells identified by their response to adrenaline (22).

In the presence of a low glucose concentration, [Ca²⁺]_c was found to oscillate in only ∼30% of α-cells; similar proportions were reported by others (3,17). As expected, the percentage of oscillating α-cells increased to ∼70% in the presence of a 2.5-mmol/l mixture of amino acids known to stimulate glucagon release (21). These oscillations resulted from rhythmic Ca²⁺-dependent spiking activity.

Whether the inhibition of glucagon secretion by glucose results from direct or indirect effects remains disputed (3,5,9,16,18,26–28). Our observation that high glucose induced a small [Ca²⁺]_c decrease in isolated α-cells supports a direct effect. Several mechanisms have been suggested to explain direct effects of glucose on α-cells. One hypothesis implicates K_ATP channels. With only one exception (29), previous studies agree that α-cells possess K_ATP channels (9,11,15,16,23,30,31). However, their possible role in stimulus-secretion coupling remains obscure. Thus, glucagon secretion has been shown to be stimulated (9,11), unaffected (32), or inhibited (15,33,34) on closure of K_ATP channels by sulfonylureas. Species differences can only partly account for these contradictions (9,14,30), because results are also controversial within the same species. For instance, tolbutamide was reported to decrease (33), not to affect (29), or to increase (17) [Ca²⁺]_c in mouse α-cells. In the present study, we demonstrated that α-cells display several essential features of β-cells. I_KATP is inhibited by tolbutamide and increased by diazoxide. Its maximal amplitude is only 30% smaller than in β-cells. It is controlled by cell metabolism because mitochondrial poisoning of the cells with azide reversibly increased it. As in β-cells, K_ATP channels control the membrane potential and [Ca²⁺]_c. Thus, closing K_ATP channels with tolbutamide triggered a [Ca²⁺]_c rise. In contrast, opening K_ATP channels with diazoxide or by decreasing cell metabolism with azide decreases [Ca²⁺]_c, very likely as a result of plasma membrane hyperpolarization and arrest of Ca²⁺ influx through VDCCs. All of these results indicate that there is a coupling between K_ATP channel activity and Ca²⁺ influx through VDCCs in isolated mouse α-cells as in β-cells.

However, we found that I_KATP in α-cells was not affected by glucose, remaining similarly low in the presence of 0.5 and 15 mmol/l glucose. This contrasts markedly with β-cells in which I_KATP is much larger in low than in high glucose. These results are in keeping with our NAD(P)H measurements, indicating that stimulation by high glucose did not significantly increase α-cell metabolism while producing its expected acceleration of β-cell metabolism. The reasons for the lack of effect of glucose on α-cell metabolism are unclear. Unlike β-cells, α-cells do not express GLUT2 but the higher affinity GLUT1 transporter (35), and both cell types possess glucokinase (36). Our results are consistent with previous studies on α-cell metabolism showing that glucose does not affect NAD(P)H (37) and flavin adenine dinucleotide fluorescence (38,39) or the ATP-to-ADP ratio (40). They are at variance with other data on α-cells reporting an hyperpolarization of the mitochondrial membrane potential (41), a small increase in ATP concentration (3,10), and an inhibition of I_KATP by glucose (9). The reasons for these discrepancies are unclear and could be due to differences in experimental procedures (including selection of cells) and/or species. The observation that α-cell I_KATP was small in the presence of 0.5 mmol/l glucose suggests that the ATP-to-ADP ratio is high in low glucose. A previous report from our laboratory has shown that the ATP-to-ADP ratio is much higher in α- than β-cells maintained in a low-glucose concentration (40). It is also possible that K_ATP channels are more sensitive to ATP inhibition in α- than β-cells (23).

It is unclear why, at low glucose, [Ca²⁺]_c oscillates in some α-cells and remains at basal levels in others. The similar responsiveness of the two groups of α-cells to adrenaline (this study) and arginine (19) rules out a trivial explanation of malfunctioning. Oscillating and silent cells could correspond to two populations of α-cells either equipped with different sets of VDCCs (42) or maintaining small differences in input resistance (percentage of closed K_ATP channels). The high input resistance measured at 0.5 mmol/l glucose (>5 GΩ in our whole-cell recordings) can explain why [Ca²⁺]_c is elevated in one-third of the α-cells in low glucose. This is consistent with the observation that tolbutamide, which only slightly increased input resistance, abruptly raised [Ca²⁺]_c in all non-oscillating α-cells. That [Ca²⁺]_c oscillates in some α-cells and remains basal in others is reminiscent of the situation found in isolated mouse β-cells perifused with 7–8 mmol/l glucose, a threshold concentration for β-cells. The high input resistance of α-cells also explains why arginine increases [Ca²⁺]_c in α- but not in β-cells at a low glucose concentration. Thus, decreasing the α-cell input resistance with diazoxide prevented the effect of arginine.

It has been suggested that closure of most α-cell K_ATP channels might depolarize the plasma membrane to such an extent that voltage-dependent channels participating in action potential generation inactivate (15,16,33). Supporting this model, low diazoxide concentrations were reported to reverse glucose inhibition of glucagon secretion by slightly reactivating K_ATP channels and relieving the inactivation of voltage-dependent channels (16). Two of our observations argue against this proposal. Tolbutamide similarly increased [Ca²⁺]_c at high and low concentrations, and low diazoxide concentrations did not increase [Ca²⁺]_c, even in silent cells in G15.

Our findings that glucose did not significantly affect I_KATP and slightly decreased [Ca²⁺]_c in the absence or presence of a high concentration of tolbutamide, which was expected to maximally close K_ATP channels, suggest that glucose decreases [Ca²⁺]_c independently from an action on K_ATP channels. Its lack of effect during depolarization with KCl also indicates that glucose does not inhibit VDCCs. The small inhibitory effect of glucose on [Ca²⁺]_c likely results from a change in membrane potential, the mechanisms of which remain to be identified.

Other models propose that glucose-induced inhibition of glucagon secretion is indirect and mediated by β-cell–derived paracrine factors: GABA, insulin, or zinc (27). This hypothesis is also contested. GABA has been reported to inhibit glucagon secretion by activating GABA_A receptor channels in α-cells (6,8,43), but some studies (18,41,44), including the present one, failed to detect an effect of GABA in α-cells. However, it is important to bear in mind that GABA_A receptors quickly desensitize and that a prolonged application of the neurotransmitter might not reproduce the in vivo situation. Other experiments tested whether GABA_A receptor antagonists prevent the inhibition of glucagon secretion by glucose, and again, conclusions in favor (6,43) or against (7,16) the hypothesis were reached. Species differences have been put forward to explain these conflicting results (6,16,27).

Because of the inverse regulation of glucagon and insulin secretion by glucose, insulin is an appealing candidate to mediate indirect inhibition of glucagon secretion. Several reports suggest that insulin can at least partly mediate the effect of glucose in α-cells (3–5,9,45), but this hypothesis is refuted by others (18,21,46). In the present study, we found a very small inhibitory effect of insulin on [Ca²⁺]_c that did not affect insulin secretion. Zinc, which helps insulin storage in secretory granules, is coreleased with insulin and has been suggested to mediate the indirect effect of glucose on α-cells (10,46), possibly by opening α-cell K_ATP channels (11). However, this hypothesis is at variance with the observations that chelation of zinc does not reverse glucose-induced inhibition of glucagon release in mouse islets (16) and that zinc does not decrease α-cell [Ca²⁺]_c (this study) or even accelerates [Ca²⁺]_c oscillations (3,18). Again, species differences might explain these contradictory results.

Although we acknowledge that isolated and cultured α-cells, as studied here, may behave differently from α-cells within intact islets, the fact that inhibition of glucagon secretion by glucose largely occurs over a concentration range that is below the threshold for stimulation of insulin secretion (3,9,18,47,48) is difficult to reconcile with the proposed intervention of β-cell–derived paracrine factors.

The nature of the VDCCs present in α-cells is another controversial issue. It has been reported that mouse α-cells in freshly isolated islets possess at least three types of VDCCs, T-, N-, and L-type channels (14,16,49), and that in the absence of cAMP production, N-type channels are more important than L-type in the control of Ca²⁺ influx and exocytosis (49,50). However, others have been unable to identify N-type Ca²⁺ channels in mouse α-cells (24) and have suggested that Ca²⁺ influx mainly occurs through L-type channels (17,24,34). Here, using nimodipine, a selective L-type Ca²⁺ channel blocker, we showed that Ca²⁺ influx stimulated by arginine and tolbutamide or occurring in the sole presence of glucose takes place mainly through L-type channels and, to a lesser extent, through non–L-type channels.

Our data show that, as in β-cells, K_ATP channels can transduce changes in cell metabolism into changes in membrane potential and Ca²⁺ influx through VDCCs in mouse α-cells. We also show that, in contrast to β-cells, isolated α-cells are poorly responsive to glucose that slightly lowers [Ca²⁺]_c without significantly affecting cell metabolism or K_ATP channel activity. Hence, glucose and K_ATP channel modulators exert distinct effects on α-cell [Ca²⁺]_c. The lowering of [Ca²⁺]_c resulting from a direct action of glucose on α-cells is modest and probably insufficient to account for the robust inhibition of glucagon secretion produced by glucose in whole islets (19). It is therefore likely that glucose-induced inhibition of glucagon secretion in the intact islets results from a combination of both effects of the sugar and indirect effects by islet factors.

N.Q. and M.C.B. have received a research fellowship from the Fonds pour la formation à la Recherche dans l'Industrie et dans l'Agriculture, Brussels. P.G. is Research Director of the Fonds National de la Recherche Scientifique, Brussels. This work was supported by the Fonds de la Recherche Scientifique Médicale (Brussels) (grant 3.4552.04), by the General Direction of Scientific Research of the French Community of Belgium (grant ARC 05/10-328), by the Interuniversity Poles of Attraction Programme (PAI 6/40) from the Belgian Science Policy, and by Juvenile Diabetes Research Fund Project Grant 2007-685.

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

We thank M. Stevens and J. Carpent for technical assistance, Dr. P.L. Herrera for the gift of GluCre and RIPCre mice, and Dr. T. Nguyen for skillful help in the construction of AdRIPBgliDsRed.