Nicotinic Acid Adenine Dinucleotide Phosphate (NAADP) and Endolysosomal Two-pore Channels Modulate Membrane Excitability and Stimulus-Secretion Coupling in Mouse Pancreatic β Cells*

Background: TPCs are regulated by NAADP and other factors. Results: NAADP-induced Ca2+ release from acidic stores evokes depolarizing currents in pancreatic β cells. Inhibition of NAADP signaling or TPC knock out attenuates Ca2+ signaling and insulin secretion. Conclusion: NAADP-evoked Ca2+ release enhances β cell excitability and insulin secretion in response to glucose or sulfonylureas. Significance: NAADP signaling pathways offer novel therapeutic targets for diabetes treatment.

and inwardly rectifying Kir6.2 potassium channel subunits (3), and inactivating or activating mutations in the K ATP channel (where the K ATP channel is the ATP-dependent potassium channel) subunits lead to congenital hyperinsulinemia (4) or neonatal diabetes (5), respectively. However, K ATP channel closure alone is not sufficient to depolarize the membrane to threshold, and activation of an additional depolarizing current has also been postulated (6,7). The existence of an additional glucose-regulated membrane current in ␤ cells is suggested by the finding that mice lacking functional K ATP channels (8), like Sur1 or Kir6.2 knock-out mice), are not hypoglycemic, and islets from adult knock-out mice are still capable of responding to glucose with electrical activity, [Ca 2ϩ ] i oscillations, and insulin secretion (9 -11). The identity and regulation of this membrane conductance remain an enigma.
In contrast to the Ca 2ϩ influx across the plasma membrane that plays a critical role in effecting insulin granule exocytosis, Ca 2ϩ release from intracellular stores has been thought to play a modulatory rather than a triggering role in stimulus-secretion coupling in the pancreatic ␤ cell. [Ca 2ϩ ] i oscillations in response to glucose are modulated by the uptake and release of Ca 2ϩ from endoplasmic reticulum (ER) Ca 2ϩ stores (12) and also from acidic Ca 2ϩ storage organelles (13). In addition, several incretins, such as glucagon-like peptide 1 and acetylcholine, are thought to enhance insulin secretion by mechanisms that are, in part, dependent on Ca 2ϩ release from intracellular stores via intracellular messengers such as cAMP and inositol trisphosphate (IP 3 ) (14,15). However, recent studies have suggested that the newly discovered Ca 2ϩ -mobilizing messenger NAADP might play an important role in ␤ cell Ca 2ϩ signaling (16 -24).
NAADP, the most potent of the Ca 2ϩ -mobilizing messengers described, has been shown to mediate local Ca 2ϩ -signaling events by releasing Ca 2ϩ from acidic, endolysosomal Ca 2ϩ stores in several vertebrate and invertebrate cells (25)(26)(27), and appears to be a critical trigger for many Ca 2ϩ -signaling events (26 -28). The most prominent target Ca 2ϩ release channels for NAADP have recently been identified as the two members of the endolysosomal two-pore channel family, TPC1 and TPC2 (29 -37). Some studies report a lack of NAADP sensitivity in isolated lysosomes (23,38), which may reflect technical issues, but also may be due in part to loss of NAADP binding to an accessory protein (39 -42) forming part of a multiprotein signaling complex in endolysosomal membranes (27,(43)(44)(45). NAADP-induced Ca 2ϩ release in MIN6 cells can be disrupted by the lysomotropic agent glycyl-L-phenylalanine-␤-naphthylamide (GPN) or bafilomycin, which disrupts acidic store Ca 2ϩ storage implicating lysosomally related organelles as the principal target for NAADP in these cells (19,20,23). In the pancreatic ␤ cell line MIN6, and primary mouse ␤ cells, glucose increases NAADP synthesis and hence intracellular levels (18,20,22), consistent with its role as an intracellular messenger. NAADP introduced into mouse pancreatic ␤ cells via a patch pipette was found to evoke a series of oscillatory plasma membrane currents, which were blocked by the NAADP antagonist Ned-19 (21) and were abolished in pancreatic ␤ cells prepared from Tpcn2 Ϫ/Ϫ mice (29). Furthermore, increasing concentrations of Ned-19 abolished glucose-evoked Ca 2ϩ spiking in mouse pancreatic ␤ cells, suggesting an important role for NAADP in stimulus-response coupling in these cells (21). This finding is consistent with our earlier study showing that prior desensitization of NAADP-sensitive Ca 2ϩ release mechanisms block subsequent glucose-evoked Ca 2ϩ signals in MIN6 cells (18).
Glucose (18) and glucagon like-peptide 1 (18,20) have both been reported to increase ␤ cell NAADP levels, effects that may be partially dependent on the ADP-ribosyl cyclase, CD38 (20,22). At present, ADP-ribosyl cyclases, including CD38, are the only characterized enzymes that have been demonstrated to catalyze the synthesis of NAADP, using NADP and nicotinic acid as substrates by a base-exchange mechanism (46,47). It has been suggested that glucose stimulation increases the internalization of CD38 involving cytoskeletal changes (22) with NAADP synthetic sites associated with acidic organelles (20). Furthermore, glucose-evoked Ca 2ϩ signals and insulin secretion are impaired in mouse Cd38 Ϫ/Ϫ pancreatic ␤ cells, and Cd38 Ϫ/Ϫ mice show glucose intolerance (48), and human CD38 autoantibodies and CD38 mutations have been shown to be associated with type-2 diabetes (49,50). Recently, extracellular NAADP was found to be transported into mouse pancreatic ␤ cells where it evoked Ca 2ϩ release from acidic stores (24). Remarkably, intraperitoneal injections of NAADP were found to restore glucose-evoked insulin secretion in the db/db mouse model of type-2 diabetes and to ameliorate blood glucose regulation (24).

Experimental Procedures
Preparation of Islets of Langerhans and Islet ␤ Cell Clusters-Islets of Langerhans were aseptically isolated by collagenase digestion of the pancreases of 8 -10-week-old male mice of the following strains: CD1, Tpcn2 ϩ/ϩ and Tpcn2 Ϫ/Ϫ (29), Tpcn1 ϩ/ϩ and Tpcn1 Ϫ/Ϫ (52), with Tpcn mice in a B6;129 background. All mice were killed by cervical dislocation and ageand sex-matched (and for the latter two, background strainmatched). Except for the hormone release measurements (for which intact islets were used), clusters of islet ␤ cells and single ␤ cells were prepared by dispersing islets in a Ca 2ϩ -free medium and cultured on circular coverslips for 1-4 days in RPMI 1640 culture medium (GIBCO, Paisley, UK) containing 10% heat-inactivated fetal calf serum, 100 IU/ml penicillin, 100 g/ml streptomycin, and 10 mM glucose.
[Ca 2ϩ ] i Measurements-Cultured clusters of islet cells were loaded with 1 M Fura PE3-AM or Fura 2-AM (Teflabs, Austin, TX) for 60 min at 37°C in a bicarbonate-buffered solution containing 10 mM glucose. The coverslip was then used as the bottom of a temperature-controlled perifusion chamber (Bioscience Tools, San Diego) mounted on the stage of an inverted microscope. The flow rate was 1.5 ml/min, and the temperature within the chamber was 37°C. [Ca 2ϩ ] i was measured at dualwavelength (340 and 380 nm) excitation spectrofluorimetry, using a CCD camera (Photon Technologies International, Princeton, NJ) to capture the emitted fluorescence at 510 nm. When [Ca 2ϩ ] i was simultaneously measured in a voltageclamped single cell, the patch pipette contained 100 M Fura 2 pentapotassium salt, and the emitted fluorescence was captured at 510 nm using a photomultiplier (Photon Technologies International, Princeton, NJ).
Measurement of Flavine Adenine Dinucleotide (FAD) Fluorescence-Cultured clusters of ␤ cells were preincubated for 60 min at 37°C in a control medium containing 3 mM glucose and then transferred to the stage of an LSM 510 confocal microscope. After a further 10 min of perfusion by 3 mM glucose, the recording was started. The oxidized form of the FAD was excited at 488 nm. Emitted fluorescence was collected with a 505-nm long-pass filter.
Electrophysiology-All patch clamp measurements were carried out using a multiclamp 700B patch clamp amplifier and the software pClamp 9 (Axon Instruments, Foster City, CA). When using the perforated whole-cell mode of the patch clamp technique, the electrical contact was established by adding the poreforming antibiotic, amphotericin B, to the pipette solution. Amphotericin (stock solution of 60 mg/ml in DMSO) was used at a final concentration of 300 g/ml. The tip of the pipette was filled with antibiotic-free solution, and the pipette was then back-filled with amphotericin-containing solution. The voltage clamp was considered satisfactory when the access resistance was Ͻ30 megohms and stable. In the standard whole-cell configuration, the access resistance was Ͻ15 megohms. Patch pipettes were pulled from borosilicate glass capillaries (World Precision Instruments, Hertfordshire, UK); they had resistances of 3-5 megohms when filled with intracellular solution.
All experiments were carried out on single ␤ cells. Two criteria were used to identify ␤ cells. The capacitance of mouse ␣-, ␦-, and ␤ cells has been reported to be 4.4, 5, and 7.4 picofarads, respectively. Therefore, only large cells with a capacitance of Ͼ5 picofarads were chosen for this study. The average capacitance was 7.6 Ϯ 0.2 picofarads. After verification of the capacitance, a depolarizing protocol was applied to identify the properties of the voltage-dependent Na ϩ current, which is known to be largely inactivated at resting potential in ␤ cells but not in ␣ and ␦ cells. Thus, cells in which a large Na ϩ current could be activated by a small depolarizing pulse from a holding potential of Ϫ70 mV were discarded. By contrast, cells that displayed a Na ϩ current only after a hyperpolarizing pulse to Ϫ140 mV were considered to be ␤ cells and were used for the experiments.
The whole-cell K ATP channel current (I K(ATP) ) was monitored by 100-ms duration pulses of Ϯ20 mV from a holding potential of Ϫ70 mV. Whole-cell Ca 2ϩ currents were recorded by depolarizing the plasma membrane with a 100-ms pulse from Ϫ80 to 10 mV.
Solutions-The medium used for the isolation of islets and for all experiments was a bicarbonate-buffered solution containing (in mM) the following: 120 NaCl, 4.8 KCl, 2.5 CaCl 2 , 1.2 MgCl 2 , and 24 NaHCO 3 . It was gassed with O 2 /CO 2 (94:6) to maintain pH 7.4 at 37°C. Except for the electrophysiological experiments, it was supplemented with 1 mg/ml BSA (fraction V, Roche Applied Science, Mannheim, Germany). When the concentration of KCl was increased, the concentration of NaCl was correspondingly decreased to keep the osmolarity of the medium unchanged.
Insulin Secretion-Islets were isolated from mice and cultured in RPMI 1640 medium overnight before insulin secretion was assessed. Insulin secretion was measured during 1-h static incubations in Krebs-Ringer Buffer (KRB) containing (in mM) the following: 18.5 NaCl, 2.54 CaCl 2 , 1.19 KH 2 PO 4 , 4.74 KCl, 25 NaHCO 3 , 1.19 MgSO 4 , and 10 HEPES (pH 7.4). Samples of the supernatant were assayed for insulin using a mouse insulin ELISA kit (Mercodia, Sweden). Where Ned-19 was used, islets were preincubated for 5 min with the drug prior to the addition of secretagogues.
Glucose Tolerance Tests-Male Tpcn1 Ϫ/Ϫ (Tpcn1 tm1Dgen ) (52) and Tpcn2 Ϫ/Ϫ (Tpcn2 Gt(YHD437)Byg ) (29) mice and strainmatched wild types aged 66 -76 days were fasted overnight and then given 2 g/kg intraperitoneal glucose (in the form of an autoclaved 20% glucose solution). Blood samples taken from the tail vein at 0, 15, 30, 60, and 120 min were analyzed with an Accu-Chek Compact Plus glucose monitor. Mean values for each time were compared using the Student's t test, with p Ͻ 0.05 taken as significant.
Insulin Secretion from Whole Pancreata-Pancreatic perfusions were performed within 15 min of cervical dislocation in 80 -100-day-old mice essentially as described elsewhere (53). At the end of the perfusion, the pancreas was dissected and transferred in acid/ethanol (ethanol/H 2 O/HCl, 52:17:1). All samples were then stored at Ϫ20°C. Only experiments with an output rate greater than 200 ml/min were assayed for insulin. The hormone assay was done using a commercially available RIA kit (Millipore, Watford, UK).
Electron Microscopy-The islet preparations were fixed in 2% paraformaldehyde and 2% glutaraldehyde in cacodylate buffer in the presence of calcium chloride, washed, post-fixed in 1% osmium tetroxide, and contrast-enhanced by staining en bloc with uranyl acetate. Sections were further contrasted with Reynold's lead.

Results
Characterization of NAADP-AM-evoked Ca 2ϩ Release-We first investigated the effects of NAADP on intracellular Ca 2ϩ concentrations ([Ca 2ϩ ] i ) in primary mouse pancreatic ␤ cells using the membrane-permeant NAADP analogue NAADP-AM (51). Given that the concentration-response curve is bellshaped in mammalian cells (58), different NAADP-AM concentrations were tested to optimize the response (Fig. 1A) (58). 10 nM NAADP-AM gave only a small response, whereas 10 M NAADP-AM gave no response at all. An intermediate concentration of 60 nM was found to give the most consistent and largest Ca 2ϩ response and was used for subsequent studies. In 8/10 clusters of ␤ cells, extracellular application of NAADP-AM (60 nM), in the presence of low glucose (3 mM), evoked delayed [Ca 2ϩ ] i increases (Fig. 1A). The peak was reached Ͼ15 min after application of NAADP-AM (Fig. 1F). To assess the role of the ER in the action of NAADP-AM (59), we treated the cells with the SERCA pump inhibitor thapsigargin in the absence of extracellular Ca 2ϩ to remove functional ER Ca 2ϩ stores. In keeping with earlier observations (13), the NAADP-AM-evoked [Ca 2ϩ ] i transients were larger and occurred more rapidly after thapsigargin treatment ( Fig. 1, B, E, and F) (13). Previous studies have implicated acidic Ca 2ϩ stores as the principal target organelles for NAADP in pancreatic ␤ cells and other cells (19,60). Accordingly, bafilomycin (which inhibits Ca 2ϩ uptake into acidic stores dependent on V-H ϩ -ATPase activity (60)) blocked Ca 2ϩ release in response to NAADP-AM treatment ( Fig. 1, C and E). In addition, the membrane-permeant NAADP antagonist Ned-19 (100 M) (Fig. 2F) (21) also completely abolished the Ca 2ϩ transient evoked by NAADP-AM (Fig. 1, D and E). Collectively, these data suggest that NAADP targets acidic Ca 2ϩ stores rather than the ER in mouse pancreatic ␤ cells.
Modulation of Glucose-evoked Ca 2ϩ Spiking by Ned-19 and Vacuolar Proton Pump Inhibitors-We next examined whether NAADP signaling plays a role in glucose-mediated responses in primary ␤ cells as suggested previously in MIN6 cells (18). Stimulation of mouse pancreatic ␤ cells by 10 mM glucose resulted in [Ca 2ϩ ] i oscillations that were superimposed upon a sustained plateau ( Fig. 2A). Acute application of 60 nM NAADP-AM first enhanced Ca 2ϩ spiking from 5.8 Ϯ 0.7 to 11.6 Ϯ 1. 2B). This finding is consistent with the bell-shaped concentration-response curve to NAADP in mammalian cells (18,51,61); the initial stimulatory effect was mediated by low concentrations, and the subsequent inhibition reflected the build-up of higher self-desensitizing concentrations of NAADP (16,18). The antagonist (18, 51, 61) Ned-19 was also found to inhibit the glucose-induced [Ca 2ϩ ] i rise, and it abolished the Ca 2ϩ oscillations, without affecting the initial [Ca 2ϩ ] i decrease due to ATPenhanced Ca 2ϩ uptake by the ER (Fig. 2C) (62). When Ned-19 was applied after commencement of glucose-evoked Ca 2ϩ responses, the glucose-induced [Ca 2ϩ ] i plateau was abolished (Fig. 2D). The structurally related analogue Ned-20, which is not an NAADP antagonist ( Fig. 2G) (21), was without effect (Fig. 2E). Because NAADP mobilized Ca 2ϩ from acidic organelles ( Fig.  1C) (60), we examined whether selective pharmacological interference of Ca 2ϩ storage by acidic organelles with agents that affect Ca 2ϩ uptake into these stores (60)  as observed previously in the MIN6 ␤ cell line (19). When bafilomycin was applied acutely, it reduced glucose-induced [Ca 2ϩ ] i oscillations (Fig. 2I). Similar results were obtained with concanamycin, another V-type-H ϩ -ATPase blocker (Fig. 2K). The effect of bafilomycin on the frequency of glucose-evoked [Ca 2ϩ ] i transients is summarized in Fig. 2J. These observations are consistent with the hypothesis that NAADP-sensitive acidic Ca 2ϩ stores play a key role in sustaining glucose-induced [Ca 2ϩ ] i oscillations in mouse pancreatic ␤ cells (63).
NAADP Modulation of Plasma Membrane Currents and Membrane Potential-We have previously shown that intracellular application of NAADP in ␤ cells evokes oscillatory currents (21,29). To examine the impact of NAADP-evoked Ca 2ϩ release on the excitability of pancreatic ␤ cells in more detail, we infused NAADP (10 -100 nM) through a patch pipette in the standard whole-cell configuration while simultaneously measuring the membrane currents. In ␤ cells clamped at Ϫ70 mV, intracellular application of 100 nM NAADP evoked intermittent inward currents of varying amplitudes (Fig. 3A). These currents contrast with the periodic outward hyperpolarizing K ϩ currents evoked by IP 3 -evoked Ca 2ϩ release in ␤ cells (64), and they underscore the differential actions of these two Ca 2ϩmobilizing messengers. The inward currents were absent in control cells where no NAADP was added to the intracellular solution (Fig. 3B), as well as when highly desensitizing concentrations (Ͼ1 M) of NAADP were used (Fig. 3C), consistent with the bell-shaped concentration-response curve due to selfdesensitization of the NAADP receptor (18,61). Moreover, these currents were also prevented by addition of the Ca 2ϩ chelator BAPTA to the pipette solution, demonstrating that they are likely dependent on Ca 2ϩ release ( Fig. 3D and see Fig.  7A). When N-methyl-D-glucamine replaced extracellular K ϩ and Na ϩ , no inward currents could be seen in response to 100 nM NAADP (Fig. 3E). Taken together with the BAPTA studies, these data are consistent with the inward currents being due to opening of Ca 2ϩ -activated cation channels. The NAADP antagonist Ned-19 (100 M) also abolished the NAADP-evoked currents, as shown previously (Fig. 3F) (21). NAADP-evoked Ca 2ϩ release has recently been reported to activate TRPM4 channels in HeLa cells (65), and because this Ca 2ϩ -activated nonselective cation channel has been proposed to control insulin secretion (66) along with TRPM5 (67-69), we tested the effect of 9-phenanthrol, a TRPM4 blocker (70), on the NAADPevoked currents in ␤ cells. We found that the currents were inhibited by this drug (Fig. 3G), suggesting a possible involvement of TRPM4 channels.
Importantly, like NAADP, high glucose (15 mM) concentrations also evoked spontaneous inward currents in cells voltageclamped at Ϫ70 mV (Fig. 3H). These glucose-induced inward currents were also blocked by Ned-19 (100 M) (Fig. 3I). The frequencies and amplitudes of the above NAADP-and glucoseevoked currents from Fig. 3, A-I, are summarized in Fig. 3, J and K, respectively. Taken together, these data raise the exciting possibility that NAADP-mediated Ca 2ϩ release from acidic stores may modulate the glucose-mediated membrane currents that in turn initiate ␤ cell electrical activity and insulin secretion.
We next examined the effect of intracellular NAADP on the pancreatic ␤ cell plasma membrane potential. A nonstimulatory level of glucose (3 mM) alone (Fig. 4A) had no effect on resting membrane potential (around Ϫ70 mV). In the presence of 3 mM glucose, NAADP-AM (60 nM) evoked low amplitude voltage oscillations that did reach the threshold for action potential firing (Fig. 4B). At 10 mM glucose, glucose-induced electrical activity consisting of 30 -40-mV action potentials was observed. Under these conditions, the blockade of NAADP action with Ned-19 (100 M) resulted in membrane hyperpolarization and suppression of electrical activity (Fig. 4C). If NAADP-evoked Ca 2ϩ release is involved in the regulation of the membrane potential by glucose, pharmacological manipulation of ion fluxes across the endomembranes involved would be predicted to impact on glucose-mediated changes in membrane excitability. It is therefore of interest that bafilomycin reversibly suppressed glucose-evoked action potentials (Fig.  4D).

Lack of Effects of Ned-19 and Vacuolar Proton Pump Inhibitors on Plasma Membrane Currents and Cell Metabolism and
Validation of Compound Used-Although the profound effects of Ned-19 and bafilomycin on glucose-mediated Ca 2ϩ signaling and electrical changes above were ascribed to antagonism of NAADP and abrogation of acidic organelle Ca 2ϩ storage, it was important to rule out other targets that could potentially account for the effects of these two agents on glucose action.
Ned-19 exerted no significant effects upon glucose metabolism, as indicated by the persistence of glucose-induced decrease in mitochondrial FAD fluorescence (Fig. 5A). The inhibitor was also without effect on K ATP channel activity; in the presence of 10 mM glucose, the whole-cell conductance averaged 0.21 Ϯ 0.04 and 0.20 Ϯ 0.06 nanosiemens in the absence or presence of Ned-19 (Fig. 5B). By contrast, the combination of the K ATP channel activator diazoxide and the mitochondrial inhibitor sodium azide (1 mM) resulted in a large increase in K ATP channel activity. Similarly, there was no inhibitory effect of Ned-19 on the voltage-gated Ca 2ϩ currents; the peak current during depolarization from Ϫ70 to 0 mV averaged 110 Ϯ 1 pA (n ϭ 10) and 114 Ϯ 1 pA (n ϭ 10) in the absence and presence of Ned-19, respectively (Fig. 5C). Thus, the suppression of electrical activity cannot simply be attributed to activation of K ATP channels or inhibition of VDCCs. In addition, Ned-19 was without effect on Ca 2ϩ release induced by the stimulation of muscarinic receptors with acetylcholine (100 M) (Fig. 5D), which leads to the opening of IP 3 receptors and discharge of ER stores (71). These data are therefore consistent with a high degree of selectivity of Ned-19 as an antagonist of NAADP and demonstrate that its effects are consistent with a major role for NAADP-induced Ca 2ϩ release in glucose-induced Ca 2ϩ signaling.
We next analyzed the effects of bafilomycin on the same functional parameters. The effects of the latter inhibitor were independent of alterations in glucose-evoked changes in mitochondrial membrane potential or cell metabolism (Fig. 5, E-G), VDCCs (Fig. 5H), or modulation of K ATP channels (Fig. 5I). Moreover, in the absence of extracellular Ca 2ϩ , bafilomycin increased [Ca 2ϩ ] i after thapsigargin treatment (Fig. 5J), and vice versa (Fig. 5K), confirming that bafilomycin-sensitive Ca 2ϩ stores are distinct from the ER. The application of the lysomotropic agent GPN, a lysosomotropic agent that abrogates Ca 2ϩ storage by lysosomes (19,25), exerted effects that resembled those of bafilomycin (Fig. 5L), again indicating that a lysosomerelated organelle is the likely source of Ca 2ϩ release.
TPC2 Expression and Subcellular Localization in Endocrine Pancreas-Turning to the molecular targets for NAADP, we have recently identified TPC2, encoded by the Tpcn2 gene, as a critical mediator of the NAADP response (29), and we have shown that it is required to couple stimuli to Ca 2ϩ release from acidic stores (72). This channel is localized on acidic stores, but not at the ER or plasma membrane, and in particular it co-localizes with organelles of the endolysosomal system, most prominently the lysosomes (29). RT-PCR analysis of mouse islets indicates that Tpcn2 is expressed in mouse islets (Fig. 6A). Affirming the localization of TPC2 to acidic stores in ␤ cells, immunolocalization of the endogenous TPC2 in primary human ␤ cells revealed substantial overlap with immunoreactivity of LAMP1, a major lysosomal marker (Fig. 6B). There was substantially less overlap with immunoreactivity for insulin or EEA1, used as markers for insulin granules (73) and endosomes (29), respectively. In a complementary approach, TPC2-mCherry was expressed in MIN6 cells (Fig. 6C). A similar lysosomal localization of TPC2-mCherry transfected into this murine ␤ cell line was seen, but some co-localization with insulin granules could also be seen.

Glucose and Tolbutamide-evoked Ca 2ϩ Responses Are
Reduced in Isolated Pancreatic ␤ Cells from Tpcn2 Ϫ/Ϫ Mice-We next examined the effects of infusing NAADP (100 nM) into mouse pancreatic ␤ cells from age-, sex-, and backgroundmatched wild-type and Tpcn2 Ϫ/Ϫ mice via the patch pipette during standard whole-cell recordings while simultaneously measuring intracellular Ca 2ϩ concentrations (by Fura 2) and inward membrane currents. In wild-type ␤ cells, NAADP (100 nM) evoked [Ca 2ϩ ] i transients, some of which associated with transient inward currents (Fig. 7A). However, both NAADP-evoked Ca 2ϩ transients and currents were absent in cells prepared from Tpcn2 Ϫ/Ϫ mice (Fig. 7B), mirroring the effect of Ned-19 on the wild-type cells (Fig. 3F). Detailed morphological comparison of ␤ cells from Tpcn2 Ϫ/Ϫ and wild-type mice by electron microscopy indicated no substantial differences in morphology, organelle number, or distribution, including those of insulin granules (Fig. 7C), making it unlikely that changes in these parameters were the underlying cause for alterations in NAADP responses observed in Tpcn2 Ϫ/Ϫ null ␤ cells.
Next, we examined the Ca 2ϩ responses to glucose of ␤ cells prepared from wild-type and Tpcn2 Ϫ/Ϫ mice (Fig. 8, A and B). In Tpcn2 Ϫ/Ϫ ␤ cells, glucose-evoked Ca 2ϩ transients were either abolished, reduced in amplitude, or delayed (Fig. 8, B and C) compared with the robust responses observed in wild-type cells (Fig. 8, A and C). The average [Ca 2ϩ ] i rises evoked by high glucose were substantially reduced (but not abolished) in all Tpcn2 Ϫ/Ϫ ␤ cells studied (Fig. 8C), although activation of VDCCs by membrane depolarization by K ϩ (45 mM) in the presence of the K ATP channel opener, diazoxide (100 M), still evoked a large [Ca 2ϩ ] i response (Fig. 8B). Similar results were obtained from Tpcn1 Ϫ/Ϫ ␤ cells (Fig. 8, D-F). These studies indicate that the lysosomal TPC2 and endosomal TPC1 channels play a significant role in the generation of glucose-induced Ca 2ϩ signals in pancreatic ␤ cells, likely through their modulation of ␤ cell electrical activity.
It has been noted that a minimum concentration of glucose "fuel" is required for threshold concentrations of the oral hypoglycemic agent and K ATP inhibitor tolbutamide to mimic the electrical effects of raised glucose levels in mouse ␤ cells (74). In agreement, we also found that in the absence of glucose (0 mM), tolbutamide (25 M) treatment failed to increase [Ca 2ϩ ] i , in contrast to the effect in the presence of 3 mM glucose. However, pretreatment with NAADP-AM (60 nM) partially reconstitutes the Ca 2ϩ signal with tolbutamide in the absence of glucose (Fig.  8, G and H). Furthermore, we found that in wild-type mouse ␤ cells exposed to 3 mM glucose, tolbutamide (25 M) evoked a rise in [Ca 2ϩ ] i (Fig. 8I), whereas cells from Tpcn2 Ϫ/Ϫ mice failed to respond to tolbutamide.
Role of NAADP and TPC2 in Glucose-induced Insulin Secretion-Having implicated a key role for NAADP, TPC2, and lysosomal Ca 2ϩ stores in Ca 2ϩ signaling and electrical activity, we finally examined the effect of disrupting NAADP signalinguponinsulinsecretionitself.Insulinsecretionwasmeasured from isolated whole islets in response to glucose or 45 mM K ϩ . Prior treatment of islets with Ned-19 (100 M) for 5 min substantially inhibited insulin secretion induced by 15 mM glucose but not by 45 mM K ϩ (Fig. 9A). The absence of an effect of Ned-19 on insulin secretion induced by high K ϩ makes it unlikely that there is a relevant off-target effect of the inhibitor upon the exocytotic machinery.
Finally, we performed intraperitoneal blood glucose tests in fasted TPC knock-out mice in comparison with the corre- sponding strain-matched wild-type animals to examine the effect of perturbing TPC expression upon glucose homeostasis in the whole animal. This revealed that in Tpcn1 Ϫ/Ϫ mice (Fig.  9E) the blood glucose levels peaked at significantly higher levels than WT, and the time course revealed an impaired glucose tolerance capacity. In contrast, Tpcn2 Ϫ/Ϫ mice (Fig. 9F) were less affected.

Discussion
This study highlights the importance of NAADP-sensitive acidic stores and the newly identified endolysosomal channels TPC1 and TPC2 in Ca 2ϩ signaling during stimulus-secretion coupling in mouse pancreatic ␤ cells. Since its discovery as a potent Ca 2ϩ -mobilizing agent in sea urchin egg homogenates (75), NAADP has been widely demonstrated to evoke Ca 2ϩ signals in an extensive range of mammalian cells, including those of both the endocrine and exocrine pancreas (76). NAADP is an alternative product of multifunctional ADP-ribosyl cyclase enzymes, which is also responsible for the synthesis of the ryanodine receptor-regulating messenger cADPR (46). Building on early studies suggesting that cADPR is an important regulator of Ca 2ϩ signaling during secretion-coupling in pancreatic ␤ cells (77), we now reported that NAADP also mobilizes Ca 2ϩ in pancreatic ␤ cells (16 -18).
In contrast to the other two principal mobilizing messengers IP 3 and cADPR, the major target organelles for NAADP in sea urchin eggs are acidic stores rather than the ER. Pharmacological approaches and cell fractionation studies revealed that NAADP releases Ca 2ϩ from a separate organelle to the ER (78,79), identified as acidic lysosomally related organelles (25). This principle was later extended to mammalian cells and NAADP release from acidic stores and has now been established in a large number cell types (60,80). In many cells, signaling domains at lysosome-ER junctions have been observed (81). Thus, NAADP-evoked Ca 2ϩ release from acidic stores may trigger further Ca 2ϩ release from larger ER Ca 2ϩ stores through the mediation of IP 3 receptors and ryanodine receptors (59). Here, we have found that Ca 2ϩ signals evoked by the mem-  AUGUST 28, 2015 • VOLUME 290 • NUMBER 35

JOURNAL OF BIOLOGICAL CHEMISTRY 21385
brane-permeant NAADP analogue NAADP-AM are from intracellular stores because they persist in the absence of extracellular Ca 2ϩ and that the NAADP antagonist Ned-19 blocks this effect (Fig. 1).
Comparison of the effects of drugs that effect Ca 2ϩ uptake and storage in different organelles supports a role for acidic stores rather than the ER as the target of NAADP. Bafilomycin selectively inhibits vacuolar H ϩ pumps that acidify acidic stores, and it has been shown that Ca 2ϩ uptake into acidic organelles is pH-dependent and probably mediated by Ca 2ϩ /H ϩ exchange (60). Bafilomycin treatment was thus found to abolish NAADP-AM-evoked Ca 2ϩ release (Fig. 1C). In contrast, thapsigargin (a SERCA pump inhibitor that blocks Ca 2ϩ uptake into the ER) was found to enhance NAADP-AM-induced Ca 2ϩ release. This suggests that NAADP-evoked Ca 2ϩ release in the ␤ cell does not trigger further Ca 2ϩ release through ER mechanisms. Rather the predominant role of the ER here is to act to buffer Ca 2ϩ rather than as a source for release, and the functional removal of the ER decreases Ca 2ϩ buffering, allowing Ca 2ϩ release from acidic stores to increase further in the cytoplasm. The role of the ER to buffer Ca 2ϩ during signaling has also been noted for glucose-evoked Ca 2ϩ signals where glucose first decreases cytoplasmic Ca 2ϩ due to increased ATP generation and stimulation of SERCA pumps (12,82,83). Previous studies also support acidic stores as targets for NAADP. Ca 2ϩ indicators targeted to acidic granules or ER in MIN6 cells showed that NAADP releases Ca 2ϩ from acidic organelles but not the ER (17). Bafilomycin and the lysosomotropic agent GPN abolishes Ca 2ϩ release by photolysis of caged NAADP in MIN6 cells, but it does not affect IP 3 -evoked Ca 2ϩ release (19). In primary mouse ␤ cells, NAADP-evoked Ca 2ϩ release was inhibited by GPN, which was shown to lyse acidic stores selectively (13,23). The delay in Ca 2ϩ responses seen with NAADP-AM (Figs. 1 and 4) may also be determined partly by the time for hydrolysis of ester groups by intracellular endogenous esterases, which varies between cells (51), but the delay is probably largely due to initial buffering of Ca 2ϩ by the ER because it was decreased by thapsigargin treatment. In addition, by analogy with the situation in the ER (83), uptake of Ca 2ϩ into acidic stores might be enhanced by glucose-stimulated ATP production, and because luminal Ca 2ϩ sensitizes TPCs to low NAADP concentrations, this could promote Ca 2ϩ release from these stores (84,85), an effect that would be enhanced by removing competing ER stores.
Since the initial reports linking NAADP-evoked Ca 2ϩ release to two-pore channels (29 -31), there have been numerous reports of TPCs playing an essential role in mediating NAADPevoked Ca 2ϩ release from acidic stores (58,86). However, recent evidence points to a separate NAADP-binding protein that interacts with TPCs to confer NAADP sensitivity (39). Indeed, the requirement of NAADP-binding proteins may suggest that under certain circumstances these proteins may inter-act with multiple channel types (87). This may explain a recent report of the loss of sensitivity of isolated lysosomes to NAADP (23,38), and the finding that glucose may apparently still evoke Ca 2ϩ signals in ␤ cells from Tpcn1 Ϫ/Ϫ /Tpcn2 Ϫ/Ϫ mice (23), although it should be noted that the concentrations of NAADP-AM used in the report by Wang et al. (23) are more than 3 orders of magnitude greater than those found to be effective here. However, consistent with the data presented here, these authors (23) also reported that NAADP mobilizes Ca 2ϩ from acidic stores in the INS-1 ␤ cell line, an effect blocked by Ned-19, and evoke membrane depolarization and spike generation in this cell line. We show here that NAADP-evoked Ca 2ϩ transients in ␤ cells are abolished in cells prepared from Tpcn1 Ϫ/Ϫ and Tpcn2 Ϫ/Ϫ mice (Fig. 7, A and B). We found that endogenous TPC2 proteins in human ␤ cells co-localized with lysosomal markers (Fig. 6B). Interestingly, TPC2 did not appear to colocalize with insulin granules, which have also been proposed to function as NAADP-sensitive Ca 2ϩ stores in ␤ cells (13,17). In addition to mobilizing Ca 2ϩ from acidic stores, NAADP was also found here to evoke plasma membrane cation currents and to depolarize the plasma membrane. NAADP applied through the patch pipette at low concentrations evoked a series of inward current transients (Fig. 3, A-G). Application of higher NAADP (100 M) gave no response, consistent with the bell-shaped concentration-response curve for NAADP in mammalian systems and paralleling the concentration dependence of NAADP for Ca 2ϩ release in ␤ cells (Fig. 1A) (16,18,20) and other mammalian cells (58). These currents were blocked by Ned-19 and by BAPTA, suggesting that they are Ca 2ϩ -activated. Their abolition by replacing Na ϩ with N-methyl-D-glucamine suggests that the currents are cation currents largely carried by Na ϩ ions. Interestingly, a nonselective cation current has also been reported to be activated by GLP-1 (where GLP-1 is glucagon-like peptide 1), an agonist that has also been reported to elevate NAADP levels (20), in the HIT-T15 ␤ cell line (88). The identity of the channels responsible has not been established, but our results with 9-phenanthrol may tentatively point to some involvement of the TRPM4 channels. NAADPevoked Ca 2ϩ release has recently been shown to activate TRPM4 channels in HeLa cells (65). Moreover, TRPM4 and TRPM5 channels have been proposed to mediate in part a Ca 2ϩ -dependent depolarization of the plasma membrane in the INS-1 ␤ cell line (66 -68, 89) and may play a general role as key components of membrane-based Ca 2ϩ oscillators providing initial cell membrane depolarization for cell activation (90). The NAADP-evoked currents were found to be coincident with small NAADP-evoked Ca 2ϩ transients (Fig. 7A), and neither NAADP-evoked Ca 2ϩ transients nor currents were observed in cells from Tpcn2 Ϫ/Ϫ mice (Fig. 7B). We propose that these currents are due to NAADP-evoked Ca 2ϩ release from endolysosomal stores via TPC2 channels and that this, in turn, via elevation of [Ca 2ϩ ] i , leads to Ca 2ϩ -dependent activation of plasma membrane cation channels, possibly TRPM4 or TRPM5. Activation of these channels would then result in membrane depolarization. The finding that application of NAADP-AM elicited a series of membrane potential spikes (Fig. 4B) is consistent with this sce-nario and in agreement with a report that NAADP causes membrane depolarization in INS1 cells (23).
To investigate the role of NAADP-mediated Ca 2ϩ signaling in glucose-induced electrical activity and [Ca 2ϩ ] i , four different approaches were used to block NAADP signaling. These were as follows: (i) abrogation of Ca 2ϩ storage by acidic stores with vacuolar proton pump inhibitors and GPN; (ii) inhibition of the NAADP receptor by Ned-19; (iii) self-desensitization of the NAADP receptor by NAADP; and (iv) knock-out of Tpcn2 and Tpcn1, genes encoding proposed NAADP target channels. Intriguingly, high glucose was also found to evoke small Ned-19-sensitive currents similar to those evoked by pipette application of NAADP (Fig. 3H). Thus, NAADP signaling may contribute, at least partly, to bringing the membrane potential from rest to the threshold for activation of VDCCs (Fig. 10). As has been recognized for a long time, closure of K ATP channels is not sufficient to explain how glucose depolarizes the pancreatic ␤ cell; a depolarizing membrane current is also required (7). We propose that NAADP/TPC1/2-dependent mobilization of Ca 2ϩ from an acidic intracellular store results in activation of depolarizing cation-conducting plasmalemmal ion channels and that this brings the membrane potential to the threshold for action potential firing. This is consistent with our finding that in the absence of NAADP-evoked Ca 2ϩ signals in cells from Tpcn2 Ϫ/Ϫ mice, the K ATP channel blocker, tolbutamide, at threshold concentrations fails to evoke Ca 2ϩ signals as seen in wild-type cells (Fig. 8I). Indeed, it is remarkable that tolbutamide cannot by itself mimic glucose-induced Ca 2ϩ signals but requires NAADP/TPCs. Indeed, tolbutamide will only evoke Ca 2ϩ signals when the acidic vesicle pathway is co-stimulated either with subthreshold concentrations of NAADP/AM or with a permissive subthreshold glucose (3 mM) concentration (Fig. 8, G and H).
The final step in the stimulus-secretion coupling is the exocytosis of insulin-containing granules. In isolated islets, Ned-19 completely blocked glucose-evoked insulin secretion. Ned-19, however, had no effect on secretion evoked by depolarizing islet cells with high extracellular K ϩ , which bypasses electrical activity and depolarizes the membrane potential to ϳϪ10 mV and opens the VDCCs. This finding makes it possible to discard the explanation that Ned-19 inhibits insulin secretion by an off- NAADP-induced Ca 2ϩ release synergizes with the K ATP -dependent pathway to depolarize the plasma membrane and activate VDCCs. The ATP-mediated closure of K ATP channels increases membrane resistance, which together with Ca 2ϩ -dependent depolarizing currents (possibly TRPM4/5 channels), activated by NAADP-induced Ca 2ϩ -release via TPC2 expressed on acidic stores, may depolarize the plasma membrane to threshold for VDCC activation (AS, acidic stores; NSCC, nonselective cation channel; R m , membrane resistance; VDCC, voltage-dependent Ca 2ϩ channels; Glc, glucose). In addition, NAADP-induced Ca 2ϩrelease together with VDCC-mediated Ca 2ϩ influx are both required for exocytosis of insulin granules. target effect on the exocytotic machinery. In a more in vivo setting, glucose-evoked insulin secretion from perfused whole pancreata from Tpcn2 Ϫ/Ϫ and Tpcn1 Ϫ/Ϫ mice was investigated. Insulin secretion stimulated by glucose (20 mM) was substantially reduced compared with that from wild-type animals (Fig. 9B). Thus, we provide evidence that NAADP signaling is an important regulator of stimulus-secretion coupling in pancreatic ␤ cells (Fig. 10).
Surprisingly, Tpcn Ϫ/Ϫ mice are only mildly diabetic as assessed by glucose tolerance tests (Fig. 9), with a significant impairment in Tpcn1 Ϫ/Ϫ mice. However, a recent study has implicated TPC2 as a novel gene for diabetic traits in mice, rats, and humans (91), with a decrease in fasting glucose and insulin levels reported in Tpcn2 Ϫ/Ϫ mice. The effects of knocking out Tpcn genes in mice may result in complex phenotypes, including compensatory mechanisms, with regard to blood glucose and insulin levels because NAADP-mediated Ca 2ϩ release has been implicated in GLUT4 translocation in murine skeletal muscle (92). Furthermore, NAADP signaling has been implicated in the action of peroxisome proliferator-activated receptor ␥ agonists in their insulin-sensitizing actions to ameliorate insulin resistance (93). Future studies with tissue-specific inactivation of these genes will be required to address these questions.
Although NAADP levels in ␤ cells have been reported to be increased in response to elevated glucose, and also in response to the incretin hormone GLP-1 (18,20), the mechanisms are not well understood. However, ADP-ribosyl cyclases have been implicated in NAADP synthesis in ␤ cells (20). Although CD38 is a membrane-bound ecto-enzyme, glucose treatment of ␤ cells induces endocytosis of CD38 that requires cytoskeletal changes (22). Inhibition of CD38 internalization with jasplakinolide, which promotes actin polymerization, blocks glucose-stimulated NAADP levels and impairs glucose-evoked Ca 2ϩ signaling (22). NAADP synthesis is found to be associated with lysosomal membrane fractions (20). We have previously argued that cADPR (and NAADP) synthesis may occur within the acidic organelles (94). The luminal acidic pH of acidic organelles would provide an optimal environment for NAADP synthesis by ADP-ribosyl cyclases (46). We showed that pyridine nucleotides are transported into organelles, and second messenger products are transported into the cytoplasm to their site of action (94).
Antibodies to CD38 (95) and a missense mutation in the CD38 gene (50) have been linked to type 2 diabetes, and this could potentially be accounted for by reduced NAADP synthesis. Remarkably, in a recent study, it was found that intraperitoneal injections of NAADP could restore defective insulin secretion and blood glucose regulation in db/db mice, an animal model of type 2 diabetes (24), presumably via the NAADP transport mechanisms described above.

Conclusions
We propose that NAADP-evoked Ca 2ϩ release from acidic stores via TPC2 or TPC1 channels evokes a small local Ca 2ϩ signal that activates Ca 2ϩ -dependent cation currents in the plasma membrane. The finding that the membrane currents evoked by intracellular Ca 2ϩ mobilization are blocked by 9-phenanthrol implicates TRPM4 channels in this process. Additional direct effects of NAADP-evoked Ca 2ϩ release on exocytosis itself cannot be excluded at this stage, with a possible contribution from exocytotic granules themselves (17,96). The NAADP/TPC/acidic organelle pathway represents a new component of the glucose-evoked trigger to depolarize the plasma membrane upon K ATP channel closure resulting in VDCC-mediated Ca 2ϩ influx and insulin secretion (Fig. 10). This new pathway may offer new targets for novel diabetic therapies.