Ryanodine Receptor Type I and Nicotinic Acid Adenine Dinucleotide Phosphate Receptors Mediate Ca2+ Release from Insulin-containing Vesicles in Living Pancreatic β-Cells (MIN6)*

We have demonstrated recently (Mitchell, K. J., Pinton, P., Varadi, A., Tacchetti, C., Ainscow, E. K., Pozzan, T., Rizzuto, R., and Rutter, G. A. (2001)J. Cell Biol. 155, 41–51) that ryanodine receptors (RyR) are present on insulin-containing secretory vesicles. Here we show that pancreatic islets and derived β-cell lines express type I and II, but not type III, RyRs. Purified by subcellular fractionation and membrane immuno-isolation, dense core secretory vesicles were found to possess a similar level of type I RyR immunoreactivity as Golgi/endoplasmic reticulum (ER) membranes but substantially less RyR II than the latter. Monitored in cells expressing appropriately targeted aequorins, dantrolene, an inhibitor of RyR I channels, elevated free Ca2+ concentrations in the secretory vesicle compartment from 40.1 ± 6.7 to 90.4 ± 14.8 μm(n = 4, p < 0.01), while having no effect on ER Ca2+ concentrations. Furthermore, nicotinic acid adenine dinucleotide phosphate (NAADP), a novel Ca2+-mobilizing agent, decreased dense core secretory vesicle but not ER free Ca2+ concentrations in permeabilized MIN6 β-cells, and flash photolysis of caged NAADP released Ca2+ from a thapsigargin-insensitive Ca2+ store in single MIN6 cells. Because dantrolene strongly inhibited glucose-stimulated insulin secretion (from 3.07 ± 0.51-fold stimulation to no significant glucose effect;n = 3, p < 0.01), we conclude that RyR I-mediated Ca2+-induced Ca2+ release from secretory vesicles, possibly potentiated by NAADP, is essential for the activation of insulin secretion.

Increases in cytoplasmic Ca 2ϩ concentration ([Ca 2ϩ ] c ) 1 are important for stimulation of neurosecretion in general (1) and for the activation of insulin secretion from pancreatic islet ␤-cells (1,2). In the latter cell type, increases in [Ca 2ϩ ] c usually occur as a result of either nutrient-induced influx of Ca 2ϩ ions through voltage-gated Ca 2ϩ channels on the plasma membrane (3) or via the release of Ca 2ϩ from intracellular Ca 2ϩ stores (4). The endoplasmic reticulum (ER) (5,6) and Golgi apparatus (7) probably represent the major Ca 2ϩ stores in ␤-cells (8 -10) as in other cell types (11). However, we have recently provided evidence that dense core secretory vesicles also play a role in intracellular Ca 2ϩ signaling in ␤-cells (9,12). Importantly, secretory vesicles occupy a substantial proportion of the intracellular volume of ␤-cells (13,14) and may contain close to half the total cellular Ca 2ϩ . As such, these organelles potentially provide a huge store of mobilizable Ca 2ϩ ions (15).
Previous studies involving measurements of intravesicular free Ca 2ϩ concentration ([Ca 2ϩ ] SV ) and immunoelectron microscopy (9) indicated that ryanodine, but not inositol 1,4,5trisphosphate (11), receptors mediate Ca 2ϩ release from secretory vesicles in ␤-cells (9,10,16). cDNAs encoding three RyR isoforms have so far been identified in mammals. The type I isoform (RyR I) is expressed mainly in skeletal muscle (17), whereas RyR II is abundant in the heart (18). RyR III is present in a variety of tissues and cell types, most notably the brain (19). RyR II has been reported previously to be the most abundantly expressed isoform (at the mRNA level) in wild type (20) and ob/ob mouse islets (20,21), as well as in rat islets (22) and clonal ␤TC3 cells (22). Moreover, the presence of RyR II protein has also been demonstrated in derived INS-1 ␤-cells (23). Lower levels of RyR I and RyR III mRNA have also been detected in ␤TC3 (22) and HIT-T15 cells (24), respectively. However, the physiological role(s) of RyRs in ␤-cells remains unclear, given that RyR II mRNA levels in ob/ob mouse islets are reportedly ϳ1000-fold less than in the heart (21), whereas RyR II protein levels in INS-1 ␤-cells were ϳ10-fold lower than in brain (23).
Receptors for nicotinic acid adenine dinucleotide phosphate (NAADP), a novel intracellular Ca 2ϩ -mobilizing agent (25), may represent an alternative pathway for Ca 2ϩ efflux from dense core secretory vesicles (26). Although other studies (27)(28)(29)(30) have demonstrated NAADP-induced Ca 2ϩ release in a variety of mammalian cells and cell lines, few data are currently available regarding the role of NAADP in the ␤-cell. Although functional NAADP-sensitive Ca 2ϩ stores were recently revealed in human ␤-cells (31), NAADP-induced Ca 2ϩ release was not observed in dispersed ␤-cells from either normal or ob/ob mouse islets (32). However, in the latter report, neither the RyR agonists caffeine and ryanodine nor cyclic ADP-ribose (cADPr) induced Ca 2ϩ release.
In the present study, we show that islets and MIN6 ␤-cells express two RyR isoforms, RyR I and RyR II, that display distinct subcellular localizations. Thus, whereas type I RyRs are present at approximately equal density in a vesicle/mitochondrial fraction, and in microsomes, RyR II was considerably more abundant on ER membranes. Surprisingly, dantrolene, a selective inhibitor of RyR I, increased steady-state free [Ca 2ϩ ] in secretory vesicles but not in the ER, suggesting the presence on vesicles of a further activator or channel capable of amplifying the effects of RyRs on Ca 2ϩ release. We provide evidence that receptors for NAADP may serve this role, and we thus demonstrate that secretory vesicles, but not the ER, are an NAADP-responsive Ca 2ϩ store in ␤-cells.
Detection of mRNA for Ryanodine Receptors in ␤-Cells and Islets-Total RNA was extracted from cell lines or rat tissue using TRI Reagent TM (Sigma) according to the manufacturer's instructions and reverse-transcribed using Moloney murine leukemia virus-reverse transcriptase (Promega). PCR amplification was performed with primers designed to amplify an isoform-specific region of each of the three RyR subtypes (38) as follows: RyR I (forward, 5Ј-GAAGGTTCTGGAC-AAACACGGG-3Ј; reverse, 5Ј-TCGCTCTTGTTGTAGAATTTGCGG-3Ј); RyR II (forward, 5Ј-GAATCAGTGAGTTACTGGGCATGG-3Ј; reverse, 5Ј-CTGGTCTCTGAGTTCTCCAAAAGC-3Ј); and RyR III (forward, 5Ј-CCTTCGCTATCAACTTCATCCTGC-3Ј; reverse, 5Ј-TCTTCTACTGGG-CTAAAGTCAAGG-3Ј). The PCR mix consisted of 5 l of 10ϫ  55.5°C for 45 s, 72°C for 1 min for 32 cycles and then 72°C for 10 min. Negative controls were performed by omission of the reverse transcription step or by exclusion of the template from the PCR. PCR products were separated by migration on a 2% (w/v) agarose gel. Products were excised and purified using a QIAQuick TM gel extraction kit (Qiagen) and subjected to restriction digest analysis and automated sequencing.
Semi-quantitative PCR-Total RNA from rat islets, skeletal muscle, or heart was reverse-transcribed, and semi-quantitative PCR was performed in the following mix: 5 l of 10ϫ Buffer 3 (Roche Molecular Biochemicals), 6 l of 25 mM MgCl 2 , 1 l of 10 mM dNTPs, 0.25 l of Taq DNA polymerase (Roche Molecular Biochemicals), 5 l of reverse transcriptase-PCR cDNA, 0.4 M RyR I or RyR II primers, 0.1 M ␤-actin primers (39), and distilled H 2 O at a final volume of 50 l. Semiquantitative PCR was then performed as follows: RyR I primers, 94°C for 2 min and then 94°C for 45 s, 55°C for 45 s, 72°C for 1 min, for 25 cycles and then 72°C for 10 min; RyR II primers, 94°C for 2 min and then 94°C for 45 s, 58°C for 45 s, 72°C for 1 min, for 20 cycles and then 72°C for 10 min. Amplification in the linear phase was confirmed in each case by trials with 12-30 cycles (data not shown).
Preparation of Cell Lysates and Membrane Fractions-MIN6 cells or homogenized rat skeletal muscle were extracted into radioimmunoprecipitation assay (RIPA) buffer, consisting of phosphate-buffered saline supplemented with 1% (v/v) Nonidet P-40, 0.5% (w/v) sodium deoxycholate, and 0.1% (w/v) SDS. Secretory vesicle protein was obtained by immunoadsorption of phogrin.EGFP-containing vesicles (40). In brief, MIN6 cells infected with phogrin.EGFP were homogenized, and the nuclear fraction was removed by centrifugation. Pre-cleared homogenate was then incubated with anti-GFP antibody (Roche Molecular Biochemicals) bound to protein A-Sepharose. Immunoadsorbed vesicles were washed then incubated at 99°C for 10 min in SDS-PAGE loading buffer to dissociate vesicle proteins.
Ryanodine Receptor Type I Antiserum Production and Purification-Rabbit polyclonal antiserum was raised to a RyR I-specific sequence of 15 amino acids (residues 830 -845; RREGPRGPHLVGPSRC) that is 100% conserved in all known mammalian RyRI sequences and absent from the mammalian RyR II and III sequences. Rabbits (New Zealand White) were immunized with the keyhole limpet hemocyanin-conjugated peptide as described previously (41), and antibody specificity was confirmed by enzyme-linked immunosorbent assay and immunoblot analysis with brain, skeletal, and cardiac muscle microsomes, prepared as described previously (41). For immunoblot analysis, microsomes were separated on a 5% (v/v) polyacrylamide gel (30 g of protein/lane), and proteins were electrophoretically transferred to polyvinylidene difluoride membrane before probing with antibody at a dilution of 1:1000. Affinity-purified antibody (anti-RyR I; number 2142) was prepared by acid elution following incubation of the crude RyR I antisera either with skeletal muscle RyR protein immobilized on polyvinylidene difluoride membrane strips or on protein A-agarose columns (Sigma).
Intact cells were perifused with KRB plus additions as stated at 2 ml⅐min Ϫ1 in a thermostatted chamber (37°C) in close proximity to a photomultiplier tube (ThornEMI) (44). Where indicated, cells were permeabilized with 20 M digitonin for 1 min at 37°C and subsequently perifused in intracellular buffer (IB: 140 mM KCl, 10 mM NaCl, 1 mM KH 2 PO 4 , 5.5 mM glucose, 2 mM MgSO 4 , 1 mM ATP, 2 mM sodium succinate, 20 mM Hepes, pH 7.05). Additions to this buffer were as Microinjection, Flash Photolysis, and Ca 2ϩ Imaging-MIN6 cells were seeded onto 24 mm poly-L-lysine-coated coverslips and microinjected, as previously described (45)(46)(47), with Oregon Green 488 1,2bis(2-aminophenoxy)ethane-N,N,NЈ,NЈ-tetraacetic acid-1 dextran (Molecular Probes, Eugene, OR; 2.5 mg⅐ml Ϫ1 ) in Tris-HCl buffer, pH 8.0, in the absence (control) or presence of 2 M caged NAADP (48) (kindly provided by Dr. Luigia Santella, University of Naples). The concentration of caged NAADP in the microinjection pipette was 2 M, resulting in a final intracellular concentration of 50 -150 nM, given an injection volume of 2.5-7.5% of the total cell volume (49). 2-4 h after microinjection, cells were imaged on a Leica TCS-SP2 laser-scanning confocal illumination system attached to a Leica DM IRBE inverted epifluorescence microscope, using a 63ϫ PL Apo 1.4 numerical aperture oil immersion objective. Fluorescence was excited at 488 nm (argon laser), and fluorescence was detected at wavelengths longer than 515 nm using a long pass cut-off filter at this wavelength. A second laser (Coherent TM ) provided light for 1 s at wavelengths of 351 and 364 nm (via the objective lens) and was used for photolysis of caged NAADP at selected regions of interest within the cell (corresponding to ϳ1% of the surface of the confocal slice). Images were acquired at 5-s intervals before and after the UV pulse. Where indicated, ER/Golgi Ca 2ϩ stores were depleted of Ca 2ϩ by washing the cells in Ca 2ϩ -free KRB and incubation in Ca 2ϩ -free KRB supplemented with 1 M thapsigargin, 10 M cyclopiazonic acid, and 1 mM EGTA for 10 min. Cells were then maintained in Ca 2ϩ -free KRB during confocal imaging and photorelease of NAADP.
Assay of Insulin Secretion-MIN6 cells were incubated in full growth medium (see above) containing 3 mM glucose for 16 h and then incubated in KRB supplemented with 3 mM glucose for 30 min at 37°C. The medium was removed and retained, and cells were then stimulated with KRB supplemented with 30 mM glucose for a further 30 min at 37°C. Released and total insulin were measured by radioimmunoassay (34).
Statistics-Free [Ca 2ϩ ] was calculated using METLIG software (50). Data represent the means Ϯ S.E. of at least three separate experiments. Statistical analysis was performed using the paired Student's t test.

Detection of Ryanodine Receptor mRNAs in ␤-Cell Lines and
Primary Islets-Using isoform-specific primers, PCR products corresponding to RyR I and RyR II cDNA were readily amplified from MIN6-, INS1-, or rat islet-derived cDNAs (data not shown). By contrast, RyR III cDNA was not amplified from these sources using the chosen primer pair (see "Experimental Procedures"), although RyR III from brain cDNA was amplified, as expected. The identity of each of the generated PCR products was confirmed by both restriction analysis and by automatic sequencing, which revealed 100% identity with the corresponding mouse (GenBank TM accession numbers X83932 and AF295105 for RyR I and RyR II, respectively) and rat (GenBank TM accession numbers AF130879 and U95157) cDNAs. Semi-quantitative PCR revealed ϳ5-fold lower RyR I mRNA levels in rat islet-derived than in skeletal muscle-derived cDNA, and RyR II mRNA levels were ϳ8-fold lower in islet than in heart cDNA (data not shown).
Expression of Ryanodine Receptor Protein in MIN6 Cells-In order to confirm the presence, and identify the intracellular localization, of RyRs in MIN6 ␤-cells, subcellular fractionation and immunoblotting (Western) was performed. Probing of crude MIN6 cell fractions with a subtype-specific RyR antibody, which recognizes type I (and III) isoforms, indicated similar levels of RyR immunoreactivity in both mitochondrial/ dense core secretory vesicle and microsomal fractions (Fig. 1A,  upper panel). By contrast, RyR II immunoreactivity was much more abundant on the latter fraction, with only weak staining for RyR II in the crude vesicle/mitochondria fraction (Fig. 1A,  lower panel). To demonstrate that RyR I immunoreactivity was present on secretory vesicles in the crude secretory vesicle/ mitochondrial fraction examined above, immunoblotting was also performed with immunopurified dense core secretory vesicles (40). Immunoreactivity toward both anti-RyR I/III, and to a selective anti-RyR I antibody (see "Experimental Procedures"), was clearly evident in these membranes (Fig. 1B, upper panels), whereas reactivity to RyR III (Fig. 1B, lower panel) was undetectable.
Effect of Ryanodine Receptor Inhibition on Secretory Vesicle and ER Ca 2ϩ Concentrations-The above fractionation studies indicated that the relative abundance of type I/type II RyRs was higher in secretory vesicles than the ER but provided no information on the relative abundance of the two isoforms on either membrane, considered alone. To determine the relative importance of RyR I and RyR II on each organelle, we therefore used a functional approach in living MIN6 cells. Recombinant aequorins, targeted specifically to either organelle by the addition of appropriate presequences (9), were utilized to monitor free Ca 2ϩ concentrations in each compartment. Concentrations of ryanodine (10 M) expected to lead to closure of all RyR isoforms (51) substantially raised the steady-state concentrations of free Ca 2ϩ in both the secretory vesicle matrix ([Ca 2ϩ ] SV ) and the ER lumen ([Ca 2ϩ ] ER ) (Fig. 2, A and B). In contrast, the skeletal muscle relaxant dantrolene (52) significantly increased [Ca 2ϩ ] SV (from 40.1 Ϯ 6.7 to 90.4 Ϯ 14.8 M; n ϭ 4, p Ͻ 0.01; Fig. 2C), whereas neither [Ca 2ϩ ] ER (Fig. 2D) nor [Ca 2ϩ ] c (Fig. 2E) was affected by this agent. These findings are in agreement with previous studies where dantrolene was shown to bind directly to and inhibit pig and rabbit skeletal muscle (types I and III), but not cardiac (type II), RyRs (52-55), suggesting that dantrolene inhibits the type I, but not the type II, RyR in MIN6 ␤-cells.
Effect of NAADP on Secretory Vesicle and ER Ca 2ϩ Concentrations-The above results suggest that although ryanodinesensitive Ca 2ϩ efflux from secretory vesicles is likely to be mediated principally via type I RyRs, this channel subtype apparently plays a minor role, if any, in mediating Ca 2ϩ release from the ER in MIN6 ␤-cells. This result was unexpected given that subcellular fractions enriched with ER membranes apparently contained the same amount or more immunoreactivity to RyR I as a crude secretory vesicle/mitochondria fraction (Fig.  1A, upper panel). One simple explanation of this observation is that the absolute number of type II RyRs on the ER is very much greater than type I receptors, a difference that may not be apparent given the different antibodies and dilutions used to quantitate each of these isoforms (Fig. 1).
However, as an alternative explanation, we next explored the possibility that another Ca 2ϩ release channel may be functional on secretory vesicles, whose presence may stimulate the activity of neighboring RyR I channels. The effects on [Ca 2ϩ ] SV of the recently identified Ca 2ϩ -mobilizing molecule NAADP (25,56) were therefore examined in permeabilized cells at a concentration of this compound previously shown to be optimal in human ␤-cells (31) and other mammalian cell types (26). 100 nM NAADP caused a small but highly significant decrease in [Ca 2ϩ ] SV (Fig. 3A) but was completely without effect on [Ca 2ϩ ] ER (Fig. 3B).
To determine whether the effects of NAADP may be mediated by a receptor identical or similar to that described previously (26) in mammalian and other cell systems, we next explored the pharmacology of the observed NAADP-induced changes in secretory vesicle [Ca 2ϩ ]. Concentrations of ryanodine sufficient to inhibit all RyR isoforms (51), but known to have no effect on NAADP receptor activity (57), failed to alter NAADP-induced [Ca 2ϩ ] SV changes (Fig. 3C). Similarly, NAADP-induced release of secretory vesicle Ca 2ϩ was unaffected by dantrolene (Fig. 3D). By contrast, nimodipine, an inhibitor of L-type Ca 2ϩ channels shown previously (26) to block NAADP receptors, completely blocked NAADP-induced changes in [Ca 2ϩ ] SV (Fig. 3E).

Effect of Photorelease of Caged NAADP on [Ca 2ϩ ] c in Intact MIN6
Cells-To determine whether (i) NAADP may mediate Ca 2ϩ release selectively from dense core vesicles in living cells, and (ii) to explore the impact of this release of cytosolic Ca 2ϩ concentrations, we next micro-injected an inactive precursor of NAADP ("caged NAADP") (48) into single MIN6 ␤-cells, and we monitored the impact of its rapid uncaging by flash photolysis (Fig. 4). Photo-released NAADP provoked an increase in the fluorescence of the co-microinjected Ca 2ϩ reporter, Oregon Green, of 4.3 Ϯ 2.1% with respect to basal fluorescence (n ϭ 7 cells; Fig. 4A), consistent with the mobilization of intracellular Ca 2ϩ . The magnitude of this increase was not significantly affected by depletion of ER/Golgi Ca 2ϩ stores with the sarco-(endo)plasmic reticulum Ca 2ϩ -ATPase inhibitor, thapsigargin (5.6 Ϯ 1.4%; n ϭ 8 cells; Fig. 4B), nor by incubation with ryanodine (3.7 Ϯ 0.9%; n ϭ 7 cells; Fig. 4C).
Importance of Secretory Vesicle Ca 2ϩ Release for Glucosestimulated Insulin Secretion-The above studies indicated that release of Ca 2ϩ from secretory vesicles, mediated by either type I RyR or NAADP receptors, may be important for the triggering of insulin secretion by nutrients. To determine whether Ca 2ϩinduced Ca 2ϩ release via type I RyRs or Ca 2ϩ release through NAADP receptors was qualitatively the more important pathway for Ca 2ϩ efflux in living cells, we blocked the former with dantrolene (see above). In accordance with previous findings in islets (58) and INS1E cells (59), the stimulation of insulin release from MIN6 ␤-cells by 30 mM (versus 3 mM) glucose (3.07 Ϯ 0.51-fold; n ϭ 3, p Ͻ 0.01) was completely inhibited in the presence of the drug (Fig. 5A). Similarly, glucose-induced insulin secretion was abolished (from 2.02 Ϯ 0.11-fold stimulation to no significant effect; n ϭ 4, p Ͻ 0.01) in the presence of 100 M ryanodine (Fig. 5B). This latter finding is in contrast to previous reports where ryanodine had little or no effect on insulin secretion in mouse ␤-cells (60) nor in INS1E cells (59), a result attributed to the poor permeation of ryanodine across the cell membrane in these systems. By contrast, both the present (Fig. 2, A and B) and previous studies (61) indicate that intact MIN6 cells may be more permeable to this drug.

Ryanodine Receptor mRNA Expression in Pancreatic
␤-Cells-In this study, we found that mRNAs encoding both RyR I and RyR II are expressed in ␤-cell lines as well as in rat islets at approximately equal levels. These findings are in apparent contrast with previous reports (20 -22) indicating that RyR II mRNA was by far the most abundant RyR isoform, at least in mouse ␤-cells. Thus, in these earlier studies, only faint reverse transcriptase-PCR products were generated with RyR I primers in mouse-derived ␤TC3 cells, whereas the same primers did not detect any RyR I mRNA in rat islets (22). Similarly, ob/ob mouse islets, which are enriched in ␤-cells, were shown to contain ϳ1000-fold less RyR II mRNA than the heart (21). A number of factors may underlie these apparently discrepant results. First, the primers used to detect RyR I in previous reports may have been less efficient in amplifying the ␤-cell RyR I than those used in the present study. Consistent with this view, the primers used by Takasawa et al. (20) scarcely detected RyR I mRNA in the brain, a tissue in which this isoform is abundant (19,62). Second, the presence and density of RyRs in different ␤-cell lines and rodent strains may well differ. Thus, RyR II protein was hardly detectable in RINm5F cells, whereas INS1 cells were shown to express significant amounts (23). Similarly, one group (21) described low levels of RyR II mRNA in ob/ob mice islets, whereas another (20) failed to detect this message in islets from distinct colonies of ob/ob mice. An intriguing possibility is that these differences in RyR II expression in the islet may contribute to the differing severities of diabetes in the different mouse strains.
Ryanodine Receptor Protein Expression and Subcellular Localization in MIN6 ␤-Cells-RyR subtypes are expressed in various combinations in specific tissues and cell types. Thus, RyR II and III are expressed in the heart, RyR I and RyR III in skeletal muscle, and all three subtypes in smooth muscle and brain (62). Although the relevance of multiple isoform expression is not clearly understood, one possibility is that co-expression serves to amplify Ca 2ϩ signals. Indeed, such a mechanism may explain why the presence of both RyR I and RyR II is required for the activation of Ca 2ϩ -induced Ca 2ϩ release upon Ca 2ϩ influx in vascular myocytes (63).
The present study revealed the presence of both RyR I and RyR II in MIN6 ␤-cells and islets at both the mRNA and protein levels (Fig. 1). Moreover, using subcellular fractionation as well as vesicle immunopurification, we demonstrate a distinct subcellular localization for each RyR isoform in this cell type. Thus, type I RyRs were present on both the ER and secretory vesicles, whereas RyR II immunoreactivity was more abundant in the former. Although we also detected a small amount of RyR II immunoreactivity in the crude mitochondrial/ secretory vesicle fraction (Fig. 1A), it should be noted that at least part of this reactivity may result from contamination of this fraction with ER/Golgi fragments and/or cross-reactivity with RyR I, because the anti-type II RyR antibody used also weakly cross-reacts with RyR I.
Supporting the view that Ca 2ϩ -induced Ca 2ϩ release from secretory vesicles is mediated principally via RyR I channels, and from the ER via RyR II, blockade of RyR I receptors in whole cells with dantrolene (55) affected steady-state Ca 2ϩ concentrations only in the former compartment (Fig. 2, C versus D). Together, these data therefore provide both structural and functional evidence that RyR I and RyR II are located on distinct organelles in MIN6 ␤-cells.
Interestingly, inhibition of type I RyRs, shown here (Fig. 5) and in earlier studies (58) to block glucose-induced insulin secretion, now seems likely to involve largely a blockade of Ca 2ϩ release from secretory vesicles, rather than from the ER (Fig. 2). This result, which accords well with the reported effects of depleting vesicle Ca 2ϩ on insulin release (64), reinforces the view that the release of vesicle Ca 2ϩ plays an important role in triggering or facilitating the exocytotic release of insulin (Fig. 6).
Role of NAADP Receptors in ␤-Cells-An unexpected finding of the present study was that dantrolene, which blocks RyR I channels, affected Ca 2ϩ concentrations in secretory vesicles but not in the ER despite the presence of RyR I receptors on both organelles. One possible explanation for this result is our demonstration that MIN6 ␤-cells possess an NAADP-sensitive intracellular Ca 2ϩ store that appears to coincide, at least in part, with secretory vesicles (Figs. 3 and 4). Importantly, these data are consistent with previous findings that have suggested NAADP releases Ca 2ϩ from a non-ER Ca 2ϩ pool at the frog neuromuscular junction (65) and that the NAADP-sensitive Ca 2ϩ pool in sea urchin eggs is thapsigargin-insensitive (66).
We show here that in pancreatic ␤-cells, NAADP-induced Ca 2ϩ release is insensitive to ryanodine and dantrolene, confirming previous reports (26) that NAADP acts on a channel distinct from the RyR (Figs. 3, C and D, and 4C). Although the receptor for NAADP is as yet unidentified in molecular terms,  6. Potential cross-talk between two different but converging messenger pathways that lead to Ca 2؉ release from dense core secretory vesicles. Exposure of islet ␤-cells or MIN6 cells to nutrients causes an increase in intracellular free [ATP], which in turn may modulate the activity of CD38, increasing the intracellular concentrations of NAADP and cADPr. NAADP-induced Ca 2ϩ release from the secretory vesicles (1) could potentially activate Ca 2ϩ -induced Ca 2ϩ release from the neighboring type I RyRs (2), previously sensitized to Ca 2ϩ by cADPr (3). This would result in a domain of high Ca 2ϩ concentration underneath the plasma membrane, independent of Ca 2ϩ influx through voltage-gated Ca 2ϩ channels, which may contribute to the stimulation of insulin secretion.
in agreement with previous findings in sea urchin eggs (67), brain (28), smooth muscle (68), and heart (57), L-type Ca 2ϩ channel inhibitors were found here to block NAADP-induced Ca 2ϩ release from ␤-cell vesicles (Fig. 3E), suggesting that a common or similar receptor is involved in each of these cellular systems. Interestingly, NAADP-mediated Ca 2ϩ release in living ␤-cells was not significantly affected by blockade of RyRs (Fig. 4, C versus A), a result also consistent with an action of this messenger via a non-RyR channel.
Two Converging Pathways Are Involved in Ca 2ϩ Release from Secretory Vesicles-cADPr, the proposed endogenous ligand for RyRs (69,70), and NAADP are likely to be synthesized in ␤-cells by the same enzyme, an ADP ribosyl cyclase termed CD38 (71). At the cell surface, the catalytic domain of CD38 is likely to be exposed to the extracellular space (72). However, internalized catalytically active CD38 is found in non-clathrincoated vesicles (73). Moreover, endocytotic vesicles have been shown previously to accumulate the precursor for cADPr, ␤-NAD, via an unidentified dinucleotide transport system (74). As well as catalyzing the conversion of ␤-NAD to cADPr, CD38 itself is thought to be involved in pumping the cyclic nucleotide into the cytosol (75), thus providing a potential mechanism for the accumulation of intracellular cADPr and potentially NAADP.
Interestingly, transgenic mice overexpressing CD38 specifically in ␤-cells show enhanced glucose-induced insulin secretion (76), whereas deletion of both alleles of CD38 gives rise to glucose intolerance (77). Because CD38 activity is regulated allosterically by ATP (76), glucose-induced increases in the intracellular free concentration of this nucleotide (47, 78) may lead to increases in the intracellular concentrations of either cADPr, NAADP, or both. Although glucose-dependent increases in cADPr have been reported previously (69), key future goals will be to determine whether glucose is able to increase NAADP levels in ␤-cells and whether such increases are altered in models of type 2 diabetes mellitus.