Introduction

In addition to serving as an energy carrier and intracellular signal, ATP has an important role as an extracellular signal and neurotransmitter. After its release from cells by exocytosis or via non-vesicular pathways, ATP activates two types of purinergic P2 receptors in the plasma membrane. P2X receptors are ligand-gated non-selective cation channels, while P2Y receptors are G-protein coupled. In humans, the P2X and P2Y families comprise 7 and 11 isoforms, respectively [1, 2].

ATP is present at millimolar concentrations in insulin granules [3, 4] and is released from beta cells upon glucose stimulation [5–7]. There is evidence for the expression of both P2X and P2Y receptors in rat and mouse beta cells, suggesting that ATP acts as an autocrine signal in islets, although it is debatable whether purinergic signalling stimulates or inhibits insulin secretion [8, 9]. Overall, the few studies that have been conducted in human islets suggest a stimulatory role for ATP [10–12]. However, controversy exists regarding the receptor subtypes and signal transduction pathways involved. While one study proposed that ATP acts principally via P2X3 receptors, membrane depolarisation and increasing the intracellular Ca2+ concentration ([Ca2+]i) [11], a more recent study suggested a prominent role for P2Y1 receptors and activation of protein kinase C [12]. Involvement of P2X7 has also been proposed [13]. An ATP-evoked, P2X-mediated membrane current in human beta cells has been suggested [14], although the effect of ATP on glucose-induced electrical activity has not been investigated.

We sought to characterise the effects of extracellular ATP on membrane currents and membrane potential in human beta cells. We found that the effects of ATP were mimicked by the P2Y agonist ADP and demonstrated that autocrine activation of P2Y1 receptors plays a significant role in the regulation of electrical activity, [Ca2+]i and insulin secretion in human beta cells. Autocrine signalling via P2Y1 represents a novel link between Ca2+ influx and Ca2+ release from intracellular stores.

Methods

Materials

MRS-2279, MRS-2365, MRS-2500, α,β-methyleneadenosine-5-triphosphate (α,β-meATP), 2′,3′-O-(2,4,6-trinitrophenyl)-ATP (TNP-ATP), heparin, thapsigargin and bafilomycin A1 were from R&D Systems (Minneapolis, MN, USA). Fura-2AM and Fura-2 Na+-salt were from Life Technologies (Burlington, ON, Canada). Nucleotides and other chemicals were obtained from Sigma-Aldrich (Oakville, ON, Canada).

Islet isolation, culture and transfection

Human islets were from the Clinical Islet Laboratory at the University of Alberta or the Alberta Diabetes Institute IsletCore [15, 16]. The study was approved by the local Human Research Ethics Board. Islets were dispersed in Ca2+-free buffer and then plated onto plastic or glass-bottom Petri-dishes (In Vitro Scientific, Sunnyville, CA, USA) and incubated in RPMI-1640 medium containing 7.5 mmol/l glucose for at least 24 h before experiments. For measuring ATP release, cells were infected with an adenovirus encoding a P2X2–green fluorescent protein fusion protein (AdP2X2-GFP) for 24–48 h [5]. All experiments, except for the assessment of insulin secretion, were carried out using dispersed beta cells.

Immunohistochemistry

Paraffin-embedded tissue sections were heated in 10 mmol/l Na+-citrate (pH 6) for 10 min. Sections were blocked using 20% goat serum and incubated with anti-P2Y1 (1:50 dilution; P6487; Sigma-Aldrich) and anti-insulin antibodies for 1 h, followed by fluorescently labelled secondary antibodies. Images were captured using a Zeiss Apotome inverted microscope (Carl Zeiss Canada, Toronto, ON, Canada). Identification of beta cells by immunocytochemistry after patch-clamp and Ca2+ imaging was as described previously [17].

Ca2+ imaging

Cells were pre-incubated with Fura-2AM (1 μmol/l) for 15 min. Glass-bottom Petri dishes were mounted onto an inverted microscope (Zeiss Axioobserver, Carl Zeiss Canada Ltd.) equipped with an ICCD-camera and a rapid-switching light source (Oligochrome; Till Photonics, Grafelfing, Germany). Fluorophore, excited at 340 and 380 nm (intensity ratio 10:4) and emission detected at 510 nm, was imaged at 0.5 Hz using Life Acquisition software (Till Photonics). Beta cells were identified by immunostaining and fluorescence ratios were calculated using ImageJ (v1.46r; http://imagej.nih.gov/ij/).

Insulin secretion

Fifteen size-matched islets (in triplicates) were pre-incubated in 0.5 ml KRB buffer containing 1 mmol/l glucose and 0.1% BSA for 1 h, followed by a 1 h test incubation in KRB with the indicated glucose concentrations and test substances. The supernatant fraction was removed and the insulin concentration was determined using the MSD human insulin kit (Meso-Scale Discovery, Rockville, MD, USA).

Electrophysiology

Patch-clamp was performed using an EPC-10 amplifier and Patchmaster software (Heka Electronics, Lambrecht, Germany). Patch-pipettes were pulled from borosilicate glass (resistance 3–8 MΩ; Sutter Instruments, Novato, CA, USA). Solutions for whole-cell and perforated-patch recording are detailed in the electronic supplementary materials (ESM) Methods. Cells were continuously superfused (~1 ml/min) with extracellular solution at ~32°C. Rapid application of ATP was performed using a Fast-Step system (Warner Instruments, Hamden, CT, USA). Beta cells were identified by immunostaining or based on cell size (12.5 ± 0.3 pF; n = 189) [17].

PCR analysis

Expression of P2Y receptors (P2RY1–14) was analysed by RT-PCR in RNA purified from isolated human islets, using a previously described protocol [18]. Primer sequences are detailed in ESM Methods.

Data analysis

Data are presented as means ± SEM. The n values represent the number of cells, unless indicated otherwise. Statistical significance was evaluated using Student’s t test, or by multiple-comparison ANOVA and Bonferroni post test when comparing multiple groups.

Results

Membrane currents evoked by purinergic receptor agonists

Using the whole-cell configuration in cells held at −70 mV in 6 mmol/l glucose, extracellular ATP application evoked a detectable inward current in 25% of human beta cells (15 out of 60). The maximal current amplitude was 24 pA (Fig. 1a). In responding cells, the ATP-activated current averaged 6.7 ± 1.7 pA and was reduced 65 ± 17% by the P2 receptor blocker suramin (100 μmol/l; p < 0.05, n = 5).

Fig. 1
figure 1

Effect of purinergic agonists on resting membrane currents. (a) Whole-cell membrane current evoked by extracellular application of 0.1 mmol/l ATP. (b) Average amplitudes (normalised to cell size) of inward currents evoked by ATP (10 μmol/l, n = 15 cells), ADP (2 μmol/l, n = 40 cells), MRS-2365 (0.1 μmol/l, n = 5 cells, two donors) and α,β-meATP (10 μmol/l, n = 4 cells). (c) Membrane currents evoked by voltage ramps under control conditions (black trace) and after application of 2 μmol/l ADP (grey trace). (d) Average ADP-activated inward current measured at −70 mV (n = 21). (e) Membrane current evoked by 2 μmol/l ADP in the absence and presence of 0.2 mmol/l tolbutamide. (f, g) Control (black traces) and ADP (2 μmol/l) evoked membrane currents (grey traces) during voltage ramps under control conditions (f) and after replacement of extracellular Na+ with NMDG+ (g). (h) The ADP-activated currents were quantified at −70 mV (n = 5, two donors). Data is from three to seven donors unless indicated otherwise. *p < 0.05 compared with ATP or Na+

Full size image

In perforated-patch whole-cell recordings, ATP-evoked an inward current in all beta cells (1.2–8.6 pA) and averaged 3.4 ± 0.7 pA (Fig. 1b). Similar responses were obtained with the P2Y(1/12/13) agonist ADP (4.2 ± 0.6 pA; 0.4–11.1 pA) and the P2Y1 agonist MRS-2365 (2.4 ± 0.3 pA), but not the P2X(1/3) agonist α,β-meATP (Fig. 1b). The ATP- or ADP-evoked current was inhibited 78 ± 13% by the P2Y1 antagonist MRS-2279 (2–3 μmol/l, p < 0.05, n = 4). The ADP-evoked current was inward during voltage ramps from −110 to −60 mV (Fig. 1c, d) and was not inhibited by tolbutamide (4.2 ± 1.1 pA, n = 4; Fig. 1e). The rapid upstroke in Fig. 1e following application of tolbutamide and ADP is likely to be an artefact. Instead, the current was attenuated (by 90 ± 25%; p < 0.05, n = 5) when Na+ was replaced by the membrane-impermeable cation N-methyl-d-glucamine (NMDG+) (Fig. 1f–h).

Effect of purinergic agonists on the membrane potential

At 6 mmol/l glucose, isolated beta cells exhibited varying degrees of action potential firing (Fig. 2). Application of ATP (10 μmol/l) depolarised cells (from −53.7 ± 2.2 to −39.3 ± 2.9 mV, Fig. 2a, e) and stimulated or augmented action potential firing. This effect was mimicked by ADP, which depolarised human beta cells from −53.6 ± 2.1 to −38.6 ± 2 mV (Fig. 2b, e). UTP, an agonist at P2Y(2/6/8) receptors, had no effect (data not shown). The P2Y1 agonist MRS-2365 also potently depolarised the cells (from −53.3 ± 4 to −44.2 ± 2.9 mV, Fig. 2c, e), while the P2X(1/3) agonist α,β-meATP depolarised the membrane potential only slightly (from −59.7 ± 4.8 to −57.2 ± 4.1 mV, Fig. 2c, e). The effect of ATP was prevented in four out of five cells by the P2Y1 antagonist MRS-2279 (Fig. 2d). ATP did not directly modulate voltage-gated Ca2+ or K+ currents (ESM Fig. 1).

Fig. 2
figure 2

Effect of purinergic agonists on the membrane potential. Membrane potential recordings from human beta cells by perforated patch. (a) ATP (10 μmol/l) was applied as indicated by the bar. (b) ADP (1 μmol/l) was added as indicated. (c) Effect of α,β-meATP (10 μmol/l) and the agonist MRS-2365 (0.1 μmol/l) in the same cell. (d) ATP (10 μmol/l) was applied in the absence or presence of the antagonist MRS-2279 (1 μmol/l) as indicated. (e) Average depolarisation evoked by ATP (10 μmol/l, n = 16), ADP (1–2 μmol/l, n = 11), MRS-2365 (0.1 μmol/l, n = 11) and α,β-meATP (10 μmol/l, n = 7). Data are from four to six donors in each experiment. **p < 0.01 compared with ATP

Full size image

Effect of purinergic agonists on [Ca2+]i

At 6 mmol/l glucose, ATP induced a biphasic increase in [Ca2+]i consisting of an initial rapid spike followed by a plateau (Fig. 3a). In the absence of extracellular Ca2+, the [Ca2+]i spike was largely unchanged, whereas the plateau was reduced (Fig. 3b). Pretreatment of cells with thapsigargin removed the [Ca2+]i spike without affecting the plateau when extracellular Ca2+ was present, and completely suppressed the [Ca2+]i increase under Ca2+-free conditions (Fig. 3c, d, k). In contrast bafilomycin A1, which depletes acidic Ca2+ stores, had little effect on the [Ca2+]i signal (Fig. 3e, k). The effect of ATP on [Ca2+]i was mimicked by ADP and MRS-2365 (Fig. 3f, h, k). UTP and α,β-meATP evoked only small responses (Fig. 3g, i, k). Beta cells from a donor with type 2 diabetes were observed to have an ATP-sensitive Ca2+ response that appeared smaller than the response of healthy beta cells, but did not attain statistical significance and lacked the initial rapid spike (Fig. 3j).

Fig. 3
figure 3

Effect of purinergic agonists on [Ca2+]i. (a–j) Averaged [Ca2+]i responses shown to the same scales (AU, arbitrary units) at 6 mmol/l glucose ([Ca2+]i spikes of individual beta cells are not apparent in these averaged traces). ATP (10 μmol/l) was applied under the following conditions: (a) under control conditions (n = 68); (b) in the absence of extracellular Ca2+ (n = 63); (c) to cells pretreated with thapsigargin (10 μmol/l, 10 min, n = 27); (d) to cells pretreated with thapsigargin in the absence of extracellular free Ca2+ (n = 21, two donors) and (e) to cells pretreated with bafilomycin A1 (0.1–2 μmol/l, 10 min) in the absence of extracellular Ca2+ (n = 27). (f) ADP (2 μmol/l) was applied (n = 18). (g) Effect of UTP (10 μmol/l, n = 13). (h) Effect of MRS-2365 (0.1 μmol/l, n = 9). (i) α,β-meATP (10 μmol/l) was added (n = 20). (j) ATP (10 μmol/l) was added to beta cells from a donor with type 2 diabetes (n = 7). (k) Bar graphs showing average AUC (baseline-subtracted) during agonist application (Baf, bafilomycin A1; Thaps, thapsigargin). Data are from three to seven donors unless stated otherwise. ***p < 0.001 compared with ATP

Full size image

To identify the intracellular Ca2+ channel underlying ATP-induced Ca2+ release, [Ca2+]i was measured in cells voltage-clamped at −70 mV and infused with the InsP3 receptor blocker heparin. The [Ca2+]i increase evoked by extracellular ATP was completely suppressed by heparin (Fig. 4a, b), suggesting that the ATP-evoked [Ca2+]i spike results from InsP3-dependent endoplasmic reticulum (ER) Ca2+ release while the plateau is due to Ca2+ influx through plasma membrane channels.

Fig. 4
figure 4

Effect of ATP on Ca2+ release and exocytosis. (a) [Ca2+]i was monitored in voltage-clamped cells, without (control, black trace) or with addition of 0.2 mg/ml heparin (grey trace). ATP was added as indicated. The [Ca2+]i signal evoked by a 500 ms depolarisation from −70 to 0 mV (arrow) was used as a control. (b) Average integrated [Ca2+]i responses in experiments as described in (a) (n = 7 and 4). (c) Representative traces showing capacitance responses before (black trace) and after addition of ATP (10 μmol/l, grey trace) in the same cell. (d) Average changes in exocytotic responses (ΔCm) after mock application (n = 5) or application of ATP (10 μmol/l, n = 15) or ADP (1 μmol/l, n = 9), normalised to control values in the same cells. (e) Capacitance response evoked by application of ATP (100 μmol/l) in a cell clamped at −70 mV. (f) Average exocytotic responses (ΔCm) evoked by ATP under control conditions (n = 18) and in cells pretreated with thapsigargin (Thaps; 10 μmol/l, 10 min; n = 6), calculated as the change in average Cm from the 5 s immediately before and immediately after ATP application. Data are from three to four donors. *p < 0.05 compared with control

Full size image

Effect of ATP on exocytosis

Exocytosis was elicited by voltage-clamp depolarisations from −70 to 0 mV, which triggers Ca2+ influx through voltage-gated Ca2+ channels. Exocytosis was potentiated 64% and 32% by extracellular application of ATP and ADP, respectively (Fig. 4c, d). We next monitored the membrane capacitance in cells clamped at −70 mV to prevent opening of voltage-gated Ca2+ channels, and in the presence of 5 mmol/l glucose. ATP application alone was sufficient to evoke a clear exocytotic response under these conditions (Fig. 4e). This response was strongly inhibited in cells pretreated with thapsigargin (Fig. 4f).

Expression of P2Y receptors in human beta cells

Expression of P2Y receptor isoforms in human islets was analysed by RT-PCR. Transcripts were identified for P2Y1, P2Y2, P2Y11 and P2Y14 (Fig. 5a). The expression of P2Y1 in beta cells was confirmed by co-immunostaining of human pancreatic tissue sections with anti-P2Y1 and anti-insulin (Fig. 5b).

Fig. 5
figure 5

Expression of P2Y receptors in human islets. (a) Expression of P2Y receptor isoforms and the peter pan homologue-P2YR11 transcript (PPAN) were analysed by RT-PCR. The horizontal line indicates the 500 base pair marker. (b) A human pancreatic tissue section was co-immunostained with antibodies against the P2Y1 receptor and insulin. Data are representative of results from two donors

Full size image

Exocytotic release of ATP

Rat and mouse beta cells release ATP by Ca2+-dependent exocytosis of insulin granules [5–7, 19, 20]. To examine whether ATP is secreted from human beta cells we overexpressed P2X2 receptors [6, 7] and stimulated exocytosis by infusion of Ca2+ (2 μmol/l) via the patch pipette. Resultant transient inward currents were blocked by the P2 antagonist suramin (Fig. 6a; n = 4). These events reflect release of ATP from single insulin granules [5, 7]. Beta cells were then incubated in 1 or 10 mmol/l glucose for 1 h with 100 μmol/l diazoxide to prevent KATP-mediated depolarisation. To avoid artefacts resulting from P2X2 receptor overexpression, transient inward currents were compared with the current produced by 300 μmol/l ATP. Upon infusion of Ca2+ (2 μmol/l) the ATP release events were larger (Fig. 6c, p < 0.05) and tended to be more frequent (Fig. 6d) following glucose stimulation.

Fig. 6
figure 6

Exocytotic release of ATP from human beta cells. (a) A beta cell infected with AdP2X2-GFP was clamped at −70 mV and infused with solution containing 2 μmol/l free Ca2+. Suramin (100 μmol/l) was added as indicated (n = 4). (b) Sample trace of ATP transient inward current in human cells in 10 and 1 mmol/l glucose for 1 h (n = 11). (c) The normalised amplitudes of transient inward currents (n = 11). (d) The frequency of transient inward currents (n = 11). Data are from four donors. *p < 0.05 compared with 10 mmol/l glucose. Max, maximum

Full size image

Autocrine activation of P2Y1 receptors potentiates glucose-induced [Ca2+]i signals, electrical activity and insulin secretion

We applied P2 receptor antagonists in the absence of exogenous nucleotides. The P2Y1 antagonists MRS-2279 and MRS-2500, but not the P2X(1/3) blocker TNP-ATP, reduced [Ca2+]i in beta cells stimulated with 6 mmol/l glucose (Fig. 7a). Both MRS-2279 and MRS-2500 reversibly hyperpolarised the beta cells in the absence of exogenous ATP and inhibited glucose-induced electrical activity (Fig. 7b). The P2Y1 receptor antagonists decreased the membrane potential by 3.5 ± 1.4 mV (from −47.4 ± 2.8 to −50.8 ± 2.2 mV, p < 0.05; Fig. 7b). In control experiments, the P2Y1 antagonists had no direct effects on voltage-gated Ca2+, Na+ or K+ currents or on KATP current in human beta cells (ESM Fig. 1). In islets from four donors, MRS-2500 reduced the secretory response to glucose by 35% (Fig. 7c).

Fig. 7
figure 7

Effect of P2Y1 antagonists on electrical activity, [Ca2+]i and insulin secretion. (a) Effect of antagonists MRS-2279 (1 μmol/l, n = 22), MRS-2500 (1 μmol/l, n = 18) and TNP-ATP (1 μmol/l, n = 41) on [Ca2+]i (at 6 mmol/l glucose). (b) Effect of MRS-2279 and MRS-2500 (both at 1 μmol/l) on glucose (6 mmol/l)-induced electrical activity (n = 8). (c) Insulin secretion was measured from isolated human islets at 1 and 8 mmol/l glucose (Gluc) in the absence and presence of MRS-2500 (n = 4 donors). Data are from three to six donors. **p < 0.01 and ***p < 0.001 compared with 1 mmol/l glucose or as indicated

Full size image

P2Y1 receptors couple Ca2+ influx to Ca2+ release from stores

In voltage-clamped cells [Ca2+]i was monitored during a series of short depolarisations mimicking glucose-induced electrical activity. In four of nine cells, the capacitance increase was nearly linear and ceased immediately after the end of the stimulation (Fig. 8a) and the [Ca2+]i signal plateaued after a steep initial rise, returning to baseline with a time constant (τ) of ~2 s. In these cells neither [Ca2+]i signal nor exocytosis and Ca2+ influx were affected by MRS-2279 (Fig. 8a, c–f). In five of nine cells, both exocytotic response and [Ca2+]i signal displayed a secondary acceleration during the second half of the stimulation (Fig. 8b); also, the [Ca2+]i signal returned to baseline with a significantly slower τ (~4 s). In these cells MRS-2279 strongly and reversibly inhibited exocytosis, the [Ca2+]i response and the τ of [Ca2+]i decline to levels observed in cells lacking secondary acceleration (Fig. 8b–f). The Ca2+ influx evoked by the depolarisations was not different between groups of cells (and unaffected by MRS-2279; Fig. 8f).

Fig. 8
figure 8

Contribution of Ca2+ release to the autocrine P2Y1-mediated Ca2+ signal and exocytosis. (a, b) Beta cells were stimulated by depolarisations from −70 to 0 mV (20 ms each, at 4 Hz) and membrane capacitance, Ca2+ currents and [Ca2+]i were monitored simultaneously. Tetrodotoxin (0.1 μmol/l) was included to block voltage-gated Na+ currents. The stimulation was performed under control conditions (black traces, black squares), with 1 μmol/l MRS-2279 (MRS, dark-grey traces, dark-grey circles) and after wash-out of the P2Y1 antagonist (light-grey traces, light-grey triangles) in the same cells. Cells that did not (a) or did (b) show a secondary acceleration of exocytosis and [Ca2+]i increase are shown. (c) Average exocytotic responses under control conditions, with 1 μmol/l MRS-2279 and after wash-out in cells without (No acceleration, n = 4) or with (Acceleration, n = 5) secondary acceleration of exocytosis and [Ca2+]i increase during the stimulation protocol. (d) As for (c), showing the integrated [Ca2+]i increase over baseline (AUC). (e) As for (c), showing the time constant (τ) of the [Ca2+]i signal return to baseline after the depolarisation series. (f) As for (c), showing the integrated Ca2+ current evoked by the depolarisations. Data are from two donors. *p < 0.05 compared with the ‘acceleration’ control

Full size image

Discussion

This study provides evidence that ATP acts as a positive autocrine feedback signal in human beta cells, by amplifying glucose-induced [Ca2+]i responses. Several findings support a central role for P2Y1 in this: (1) The effects of ATP were mimicked by ADP (at five- to tenfold lower concentrations), which selectively activates P2Y(1/12/13) (only P2Y1 was detected in human islets) [2]; (2) the selective P2Y1 agonist MRS-2365 increased electrical activity and [Ca2+]i, while the P2X(1/3) agonist α,β-meATP did not; (3) P2Y1 inhibition blocked the ATP-evoked membrane depolarisation; (4) MRS-2500 reduced insulin secretion to a similar extent as the non-specific P2 antagonist suramin [11]. While in agreement with the findings of a recent study [12], our findings vary from those of another study suggesting a dominant role for P2X3 [11]. However, the latter study employed pyridoxalphosphate-6-azophenyl-2′,5′-disulfonic acid (iso-PPADS) and oxidised ATP at concentrations that also strongly inhibit P2Y1 [21, 22]. The more selective P2X(1/3) blocker TNP-ATP [23] does not affect [Ca2+]i (Fig. 7a). While others have also suggested a role for P2X7 [13], human P2X7 has a very low affinity for ATP (half maximal effective concentration [ED50] 0.78 mmol/l) and is insensitive to ADP and AMP [24], making a role for P2X7 unlikely here.

P2Y1 has a ~20-fold lower affinity for ATP than P2X3 [1, 2]. However, it has been reported that insulin granules contain similar concentrations of ATP and ADP [3], suggesting that both nucleotides play an important role. We demonstrated that human beta cells secrete ATP in response to increased [Ca2+]i, and the magnitude of ATP-release events was increased by glucose. This could result from intragranular ATP accumulation via granule-resident vesicular nucleotide transporter [25]. It should be noted, however, that in intact rodent islets extracellular ATP plays an important role in synchronising the electrical and Ca2+ responses among beta cells within and between islets through the induction of Ca2+ release from InsP3-sensitive stores [26–28], in addition to stimulating exocytosis.

Some studies conducted in rodents, particularly mice, have found that ATP inhibits insulin secretion [8] and that insulin secretion in islets from mice lacking P2Y1 is elevated [29]. This was attributed to direct inhibitory effects of ATP on exocytosis [30] or voltage-gated Ca2+ currents [31]. It has been reported that adenosine, acting on P1 receptors, inhibits insulin secretion from INS-1 cells [32], but this was not confirmed in human islets [11]. We show here that ATP stimulates depolarisation-evoked exocytosis without affecting Ca2+ currents in human beta cells. We found no evidence for a negative role of ATP in insulin secretion from human islets, consistent with the potentiation of insulin secretion from human islets following block of extracellular ATP degradation [11, 33].

ATP increased [Ca2+]i in a biphasic manner, with an initial peak reflecting Ca2+ release from stores and a plateau reflecting Ca2+ influx. Similar to rat beta cells [34], Ca2+ was released via heparin-sensitive InsP3 receptors, but was from thapsigargin-sensitive (ER) rather than bafilomycin-sensitive (acidic) compartments. Our findings differ from those of Jacques-Silva et al, who concluded that Ca2+ stores contribute little to the ATP-evoked Ca2+ signal in human beta cells [11], but this may be explained by experimental differences: the Ca2+ response under Ca2+-free conditions is transient and will appear small (as AUC) when compared with prolonged agonist application in the presence of extracellular Ca2+. Although a recent study suggests that autocrine activation of P2Y1 stimulates diacylglycerol production in rodent and human beta cells [12], we were unable to determine a role for phospholipase C as the inhibitor U-73122 (5–10 μmol/l) also suppressed KCl-evoked Ca2+ responses (data not shown).

In mice, P2Y1 receptors depolarise beta cells via inhibition of KATP channels [30]. In contrast, the ADP-evoked membrane current in human beta cells did not reverse at the K+ equilibrium potential and was insensitive to tolbutamide. Instead, the current was abolished by removal or replacement of Na+, indicating a Na+- or non-selective cation conductance similar to P2Y1-activated currents in neurons [35–37]. This effect was insensitive to thapsigargin (arguing against a store-operated channel) and to Gd3+, a blocker of the Na+ leak channel NALCN [38]. While the molecular identity of the P2Y1-activated leak channel remains unclear, candidates include members of the transient receptor potential channel family [39].

We now show that blocking P2Y1 receptors also inhibits electrical activity, [Ca2+]i signalling and insulin secretion in human beta cells in the absence of exogenous ATP (Fig. 7). While P2Y1 blockade inhibits spontaneous [Ca2+]i transients in mouse beta cells [40], this was in the presence of a Ca2+ channel blocker and thus not likely caused by electrical activity and Ca2+ influx. We obtained similar results with two different, selective P2Y1 antagonists [41, 42] that lacked non-specific effects on a number of human beta cell ion channels (ESM Fig. 1).

In cells stimulated with action potential-like depolarisations, P2Y1 blockade reduces [Ca2+]i and exocytosis without affecting Ca2+ currents (Fig. 8). Thus the MRS-2279-sensitive component of the [Ca2+]i increase reflects Ca2+ release from stores triggered by autocrine activation of P2Y1 receptors. This, in combination with activation of diacylglycerol and protein kinase C (DAG/PKC) [12], potentiates exocytosis. The Ca2+ signal required for insulin secretion is largely generated by Ca2+ influx through voltage-gated Ca2+ channels [43, 44]; our results are compatible with this because membrane depolarisation and Ca2+ influx are necessary to evoke ATP release and initiate the feedback loop. The secondary acceleration of depolarisation-evoked exocytosis, which we suggest reflects ER Ca2+ release, is observed in human [17, 45] but not mouse [46] beta cells.

Our data indicate that the contribution of Ca2+ release to the glucose-induced Ca2+ signal may have been underestimated. We present some limited data suggesting that release of Ca2+ from thapsigargin-sensitive stores during ATP stimulation is absent in a donor with type 2 diabetes; this could contribute to impaired secretion. There is evidence that ER stress is involved in the pathogenesis of type 2 diabetes [47] and is associated with reduced SERCA2b [48, 49], the main ER Ca2+ pump in beta cells [50]. The resulting lowering of ER Ca2+ levels may not only promote apoptosis but also impair beta cell stimulus–secretion coupling.