Generation of Nicotinic Acid Adenine Dinucleotide Phosphate and Cyclic ADP-Ribose by Glucagon-Like Peptide-1 Evokes Ca²⁺ Signal That Is Essential for Insulin Secretion in Mouse Pancreatic Islets

Anti-mouse CD38 monoclonal antibody was obtained from eBioscience (San Diego, CA,). Dulbecco’s modified Eagle’s medium containing low glucose and antibiotics was from Life Technologies (Grand Island, NY). GLP-1 (7-36) amide was purchased from American Peptide Company (Vista, CA). A ¹²⁵I-insulin radioimmunoassay kit was from Linco Research (Charles, MO). Biolog Life Science (Bremen, Germany) provided 8-pCPT-2′-O-Me-cAMP and N₆-benzoyl-cAMP. Recombinant nicotinic acid mononucleotide adenylyltransferase was a gift from Dr. Se Won Suh (Department of Chemistry, Seoul National University, Seoul, Korea) (24). All other reagents were obtained from Sigma.

Cd38^−/− mice with genetic background ICR were inbred in the animal facility of Chonbuk National University Medical School. The generation and characterization of Cd38^−/− mice have been described previously (14). All experimental animals used were under a protocol approved by the institutional animal care and use committee of the Chonbuk National University Medical School. Standard guidelines for laboratory animal care were followed (25).

Pancreatic islets were isolated from Cd38^+/+ and Cd38^−/− mice weighing 25–30 g using a collagenase method (14,26), with the exception of changing Krebs-Ringer buffer to Krebs-Ringer bicarbonate buffer (KRBB) (in millimoles per liter: 2 CaCl₂, 2.8 glucose, 145 NaCl, 1.19 KCl, 2.54 MgCl₂, 1.19 KH₂PO₄, 5 NaHCO₃, and 20 HEPES, pH 7.3). Briefly, mouse pancreata were distended by infusion of KRBB containing 0.15 mg/ml type V collagenase through the bile duct. Islets were isolated, washed with KRBB, and stabilized by culturing in a humidified incubator (95% air, 5% CO₂) overnight at 37°C in low glucose (5 mmol/l), Dulbecco’s modified Eagle’s medium supplemented with 10% (vol/vol) fetal bovine serum, 100 units/ml penicillin G, and 100 μg/ml streptomycin (culture media).

Complete methods, including a description of the procedures for measurement of [Ca²⁺]_i, measurement of intracellular cyclic ADPR and NAADP concentration, separation of organelles, reconstitution study, analysis of CD38 in organelles, and measurement of insulin, are provided in the online appendix (available at http://dx.doi.org/10.2337/db07-0443).

Data represent means ± SEM of at least three separate experiments. Statistical analysis was performed using Student’s t test. A value of P < 0.05 was considered significant.

Dependency of glucose concentrations on GLP-1–mediated regulation of [Ca²⁺]_i in β-cells was first examined. At 2.8 mmol/l glucose, GLP-1 was not able to elevate [Ca²⁺]_i (Fig. 1A). In contrast, the addition of 12 mmol/l glucose increased [Ca²⁺]_i and [Ca²⁺]_i slowly decreased, but not to the basal level, and sustained (Fig. 1B). At the sustained Ca²⁺ levels, application of GLP-1 induced a rapid and large rise of [Ca²⁺]_i that was also sustained (Fig. 1B). It should be noted here that our preliminary studies showed that at 2.8 mmol/l glucose, none of the molecules used in this study generated Ca²⁺ signals (data not shown).

To examine whether the GLP-1–induced Ca²⁺ signal is mediated by cAMP, we utilized forskolin, an activator of adenylyl cyclase. Treatment of β-cells with forskolin produced an increase in the Ca²⁺ signal similar to that observed with GLP-1 (Fig. 1C). Next, we asked whether PKA and Epac mediate the GLP-1–induced Ca²⁺ signal. As shown in Fig. 1D, activation of cAMP-dependent PKA or Epac also induced a rapid and sustained Ca²⁺ signal and the GLP-1–induced Ca²⁺ signaling or insulin secretion was only partially blocked by PKA inhibitors (10 μmol/l H89 or 100 μmol/l Rp-cAMP; online appendix Fig. 1). These results indicate that these two cAMP-sensitive molecules play a role in GLP-1–induced Ca²⁺ signal.

To evaluate whether ADPR cyclase/CD38 is involved in the GLP-1–induced Ca²⁺ signal and to identify a Ca²⁺ store, we first examined ER Ca²⁺ store. Thapsigargin, an ER Ca²⁺ ATPase inhibitor, completely blocked only the late phase of Ca²⁺ signals, while maintaining an initial sharp Ca²⁺ rise (Fig. 1E). Pretreatment of the cells with high concentrations of ryanodine, which is known to inhibit ryanodine receptor or a cyclic ADPR antagonistic analog 8-Br–cyclic ADPR, resulted in a significant inhibition of only the GLP-1–induced sustained Ca²⁺ signal (Figs. 1F and G, respectively). Xestospongin C, an IP₃ receptor inhibitor, did not show any effects on the GLP-1–induced Ca²⁺ signal (Fig. 1H). However, acetylcholine-induced Ca²⁺ signaling, which was proved to be due to IP₃ production (27), was blocked by the reagent (online appendix Fig. 2). The effects of various inhibitors on the initial and sustained Ca²⁺ increases are summarized in Fig. 1I. Together, these data suggest that GLP-1 induces a rapid and sustained Ca²⁺ signal by releasing Ca²⁺ from ER and non-ER Ca²⁺ stores in β-cells through generation of cAMP and that ADPR cyclase/CD38 is involved in the GLP-1–mediated sustained rise of [Ca²⁺]_i via generation of cyclic ADPR.

Considering the possibility that additional Ca²⁺-mobilizing messengers targeting intracellular Ca²⁺ stores beside ER could be involved in GLP-1–induced Ca²⁺ signaling, we thought NAADP would be an interesting candidate since it is one of the most potent Ca²⁺-releasing messengers found thus far (28,29). Therefore, we initially tested whether NAADP can be transported into β-cells when applied extracellularly because NAADP is known to be transported in certain cell types (30). Figure 2A shows that NAADP was transported in a time-dependent manner and that the NAADP transport was Ca²⁺ and glucose dependent and partially blocked by dipyridamole (Fig. 2B). Interestingly, cyclic ADPR externally applied to β-cells was also similarly transported, except in a Ca²⁺-independent manner (online appendix Fig. 3). To rule out any possibility of the effect on purinoreceptors by extracellularly added NAADP, we also tested the effects of equimolar concentration of NAD and NADP under the same condition. However, we could not find any significant signals with even high concentration of the nucleotides (online appendix Fig. 4). Treatment of β-cells with exogenous NAADP elicited Ca²⁺ signals in a bell-shaped concentration response, and the Ca²⁺ signal peaked at a 50 nmol/l concentration of NAADP (Fig. 2C). However, the NAADP-induced Ca²⁺ signal was not observed in the presence of 2.8 mmol/l glucose or in the Ca²⁺-free condition (online appendix Fig. 5), where NAADP was not transported into the cells (Fig. 2B). As shown in Fig. 2D, NAADP (50 nmol/l) induced a rapid and sustained [Ca²⁺]_i increase in a manner similar to that observed with GLP-1 or forskolin (see Fig. 1C). An important property of NAADP signaling in mammalian cells is a self-desensitization mechanism induced by its high concentrations (31). Similarly, treatment of β-cells with high concentrations of NAADP (10 μmol/l) blocked the low NAADP concentration (50 nmol/l)–mediated Ca²⁺ signals (Fig. 2E). Analyses of characteristics of NAADP-mediated Ca²⁺ signals revealed that NAADP-mediated Ca²⁺ signals were very similar to those observed with GLP-1. Thus, pretreatment of β-cells with thapsigargin blocked NAADP-mediated late-phase Ca²⁺ signal but not the initial spiky Ca²⁺ rise (Fig. 2F). Ryanodine and 8-Br–cyclic ADPR also inhibited the NAADP-mediated sustained Ca²⁺ signal (Figs. 2G and H). The effects of various inhibitors on NAADP-induced Ca²⁺ increases are summarized in Fig. 2I.

Numerous studies (18,32–34) have indicated that NAADP receptors exist in acidic organelles in a variety of cell types, including β-cells. We hypothesized that the initial Ca²⁺ rise, which was not blocked by thapsigargin, was due to its release from acidic organelles. Treatment of β-cells with bafilomycin A1, an inhibitor of vacuolar H⁺-ATPase, transiently increased [Ca²⁺]_i and, following treatment with NAADP, failed to induce an increase of [Ca²⁺]_I; nevertheless, thapsigargin-induced elevation of [Ca²⁺]_i was observed (Fig. 3A). A similar result was also observed by treatment of the cells with glycylphenylalanine 2-naphthylamide (GPN), which selectively disrupts these organelles via osmotic lysis (data not shown). After treatment of β-cells with bafilomycin A1 (Fig. 3B) or GPN (Fig. 3C), the increase of [Ca²⁺]_i mediated by GLP-1 or forskolin could not be found. High concentrations of NAADP prevented the induction of Ca²⁺ signals mediated by GLP-1 or forskolin (Fig. 3D). In contrast, cyclic ADPR–mediated Ca²⁺ increase was not blocked by the treatment of 10 μmol/l NAADP (Fig. 3E). The cyclic ADPR–mediated Ca²⁺ increase was also not affected by the treatment with either bafilomycin A1 or GPN (online appendix Fig. 6). In addition, bafilomycin A1 and GPN blocked generation of Epac- or PKA-induced Ca²⁺ signals but not depolarization-evoked Ca²⁺ influx (Figs. 3F and G). High concentrations of NAADP also completely abolished the Epac- or PKA-induced Ca²⁺ signals (Fig. 3H). Together, these results show that NAADP acts as the initiator and propagator of the GLP-1–induced Ca²⁺ signal. The effects of inhibitors on various stimulator-induced Ca²⁺ increases are summarized in Fig. 3I.

We evaluated the effects of acidic organelles disturbing agents on the insulin secretion induced by 12 mmol/l glucose or GLP-1 signaling molecules. Bafilomycin A1 and concanamycin A, inhibitors of vacuolar H⁺-ATPase, have been reported to inhibit glucose-induced insulin secretion by mouse islets (35). The glucose-induced insulin secretions were not affected by GPN or bafilomycin A1 (Fig. 4A) or by concanamycin A (online appendix Fig. 7). In contrast, however, pretreatment of the cells with GPN or bafilomycin A1 completely blocked GPL-1–, PKA-, or Epac-induced insulin secretion (Fig. 4A). Indeed, treatment with NAADP resulted in enhancement of insulin secretion at high glucose concentrations only, which was abolished by GPN or bafilomycin A1 (Fig. 4B). These results indicate that acidic organelles, which involve NAADP-mediated signaling and probably Ca²⁺ release, play critical roles in GLP-1–induced insulin secretion but not in glucose-induced insulin secretion.

Since the above observations suggest that NAADP and cyclic ADPR differentially regulate Ca²⁺ signals mediated by GLP-1 and that NAADP initiates and propagates the GLP-1–induced Ca²⁺ signal, we examined kinetics of the production of these two Ca²⁺-mobilizing messengers in response to GLP-1. Treatment of islets with GLP-1 generated NAADP first and then cyclic ADPR, with a delay of ∼10 s (Fig. 5A), suggesting that the production of cyclic ADPR depends on the NAADP-induced initial Ca²⁺ signal (Fig. 2D and F). Moreover, treatment of islets with NAADP increased the formation of cyclic ADPR, which was dependent on the presence of extracellular Ca²⁺ (Fig. 5B). We tested whether the cyclic ADPR production would be stimulated by a general Ca²⁺ signal. Ionomycin-induced Ca²⁺ signals did not affect cyclic ADPR production (online appendix Fig. 8) as in lymphokine-activated killer cells (8), indicating that NAADP-induced Ca²⁺ signaling specifically produces cyclic ADPR, as in sea urchin eggs (36). We next evaluated dependency of glucose and Ca²⁺ on the GLP-1–stimulated production of NAADP and cyclic ADPR. As shown in Fig. 5C, high glucose concentrations increased the levels of NAADP and cyclic ADPR compared with low glucose concentrations. Addition of GLP-1 in the presence of high glucose concentrations further increased the production of NAADP and cyclic ADPR by ∼3- and 2.5-fold, respectively. GLP-1–mediated production of NAADP and cyclic ADPR was significantly lower in the presence of 2.8 mmol/l glucose or absence of extracellular Ca²⁺ than that in the presence of high glucose and extracellular Ca²⁺ concentrations. These results indicate that high glucose and extracellular Ca²⁺ concentrations are prerequisite for the GLP-1–induced production of the Ca²⁺-mobilizing second messengers. Bafilomycin A1 completely blocked the GLP-1–induced formation of NAADP and cyclic ADPR (Fig. 5D), suggesting that NAADP is formed in acidic organelles. These results also indicate that the formation of cyclic ADPR may depend on NAADP signaling.

CD38 (transmembrane glycoprotein) is known to catalyze the synthesis of NAADP and cyclic ADPR (7,11,12). However, it has not yet been clarified which organelles in the pancreatic β-cells are involved in the production of NAADP and cyclic ADPR. To address the question, we resolved cellular organelles into at least three parts: plasma membranes, lysosomes, and secretory granules (Fig. 6A). NADase activity was detected in all three organelles; however, the lysosomes-containing fraction contained the highest NADase activity among the three fractions (Fig. 6B and C). To determine the presence of enzyme(s) other than CD38 in these compartments, CD38 in each fraction was immunoprecipitated with a CD38-specific antibody, and NADase activity in the precipitates and supernatants was further determined. CD38 activity was found mainly in the lysosome fraction and less in the plasma membrane fraction. However, in the supernatants from all fractions, significant levels of NADase activity were also observed consistently (Figs. 6B and C), suggesting an existence of non-CD38 enzyme(s) in β-cells.

We next performed reconstitution studies to determine the production of NAADP and cyclic ADPR in the organelles isolated. The reconstitution study revealed that cyclic ADPR was produced in both plasma membrane–and secretory granule–containing fractions (Fig. 6D). Interestingly, CD38 was not detected in the fractions of secretory granules. On the other hand, the production of NAADP was observed only in fractions containing lysosomes (Fig. 6D). Moreover, the effect of PKA or Epac activator on the production of NAADP and cyclic ADPR showed specificity for the organelles; both PKA and Epac stimulated the generation of NAADP in the lysosomes only (Fig. 6E). Epac activated the cyclic ADPR production in both plasma membranes and secretory granules, and PKA stimulated the cyclic ADPR formation only in the plasma membranes (Fig. 6F). These data indicate that activation of CD38/ADPR cyclase(s) by GLP-1 depends on their intracellular localization, where assembly of cognate signaling proteins such as PKA and Epac may be different.

To further confirm the above observations that enzymes producing NAADP and/or cyclic ADPR exist in β-cells, β-cells isolated from CD38 knockout mice (Cd38^−/−) and the littermates (Cd38^+/+) were used. In the presence of 12 mmol/l glucose, GLP-1–mediated Ca²⁺ signals in these β-cells were somewhat different. The early Ca²⁺ increases were similar to each other; however, the late phase of Ca²⁺ signals in Cd38^−/− β-cells was gradually decreased while the Ca²⁺ signals in Cd38^+/+ β-cells were sustained (Fig. 7A). When the production of NAADP and cyclic ADPR was determined, no alterations in the levels of NAADP and cyclic ADPR were observed in the presence of 12 mmol/l glucose only. However, GLP-1–stimulated production of NAADP and cyclic ADPR in Cd38^−/− islets was substantially higher than that with glucose alone and significantly reduced compared with that in Cd38^+/+ islets (Fig. 7B). Insulin secretion of Cd38^−/− islets stimulated by glucose or GLP-1 was similar to that of Cd38^+/+ islets (Fig. 7C). These findings clearly demonstrate that ADPR cyclase(s)/NAADP-producing enzyme(s) other than CD38 exists in β-cells.

In this study, we showed that GLP-1 elevates Ca²⁺ via stimulation of NAADP and cyclic ADPR production and that NAADP-induced Ca²⁺ mobilization from acidic stores contributes to the GLP-1–stimulated Ca²⁺ signal. The action of GLP-1 is complex, with multiple kinetic components shaping the Ca²⁺ response. Our results firmly established a role of thapsigargin- and ryanodine-sensitive stores and also provide new data implicating a source of Ca²⁺ in acidic organelles.

Many equivocal results have been reported (9,13,14) on the possible role of CD38 and/or cyclic ADPR in glucose signaling in β-cells. Our results showed that high glucose concentrations alone could elevate cyclic ADPR and NAADP levels (Fig. 5C). Consistent with our observations, Masgrau et al. (17) showed that high glucose levels (20 mmol/l) increase NAADP levels in MIN6 cells, a clonal pancreatic β-cell line. However, the glucose-induced production of NAADP and cyclic ADPR was not altered in Cd38^−/− islets compared with that in Cd38^+/+ islets (Fig. 7B). These findings together with immunoprecipitation (Figs. 6B and C) and reconstitution studies (Fig. 6D) suggest that additional ADPR cyclase(s) and/or NAADP-producing enzyme(s) besides CD38 exist in islets. Recently, an existence of enzymes capable of generating NAADP and/or cyclic ADPR beside CD38 has been reported in various tissues (37).

Interestingly, at low concentrations of glucose (2.8 mmol/l), none of the following molecules generate Ca²⁺ signals in β-cells: GLP-1, forskolin, activators of PKA and Epac, and NAADP. These molecules generate Ca²⁺ signals only under the condition of high glucose–induced elevated basal Ca²⁺ level (Fig. 1B). In agreement with our observations, several studies have also reported that GLP-1 has no significant effects on β-cells at low concentrations of glucose (38,39). However, the reasons for the insensitivity of GLP-1 and other downstream signaling molecules at low concentrations of glucose remain completely unknown. Intriguingly, KCl-induced Ca²⁺ signals could not substitute for the effect of high glucose concentrations in triggering GLP-1–induced Ca²⁺ signals (online appendix Fig. 9). Glucose-induced Ca²⁺ signals as well as metabolites of glucose may coordinate the cognate signaling molecules and/or organelles for the effective responses to subsequent stimuli in β-cells.

In the present study, exogenous NAADP could be a useful tool for analyzing NAADP-mediated Ca²⁺ signaling by virtue of the property of NAADP transport into β-cells. The characteristics of NAADP transport were very similar to those observed in RBL-2H3 cells, which were originally observed (30). However, the exact mechanisms involved in the transport of NAADP in pancreatic islets remain to be clarified. Dependencies of the NAADP transport on high glucose and external Ca²⁺ are the important issues to be solved in conjunction with the Ca²⁺ signals in β-cells.

Our data (Fig. 1E) show that thapsigargin greatly reduced the overall Ca²⁺ response, while leaving Ca²⁺ transient. Previously, Holz et al. (40) demonstrated that GLP-1, as well as cAMP (through Epac), mobilized Ca²⁺ from intracellular Ca²⁺ stores, which was abolished by thapsigargin. Our data suggest that Ca²⁺ signaling induced by GLP-1 consists of transient and sustained components and that the former is thapsigargin insensitive and the latter is sensitive to thapsigargin or ryanodine. Furthermore, we could differentiate the effects of bafilomycin or GPN on the transient and sustained Ca²⁺ by GLP-1; bafilomycin or GPN pretreatment had no effects on cyclic ADPR–induced Ca²⁺, which represents sustained Ca²⁺ (online appendix Fig. 6), while completely blocking the NAADP-induced Ca²⁺, which represents transient Ca²⁺ (Fig. 3A).

Mitchell et al. (34) showed that NAADP decreased free Ca²⁺ concentrations of dense-core secretory vesicles in permeabilized MIN6 β-cells. Duman et al. (41) reported also that dense-core secretory granules comprise a GPN-sensitive acidic Ca²⁺ store. Our results also showed a similar result in that 50 μmol/l GPN released Ca²⁺ from secretory granule–containing fractions and lysosome-containing fractions (online appendix Fig. 10). Therefore, further studies are needed to clarify these NAADP-responsive Ca²⁺ stores in β-cells.

Johnson and Misler (16) suggested that NAADP initiates subsequent oscillatory Ca²⁺ signaling in insulin signaling in human β-cells. However, their results are quite different from ours in two aspects: 1) Ca²⁺-free solutions did not affect initial Ca²⁺ rise in the insulin-mediated Ca²⁺ signals, whereas GLP-induced Ca²⁺ signals were completely blocked in the Ca²⁺-free solutions (data not shown), and 2) thapsigargin resulted in a complete abolishment of the insulin-mediated Ca²⁺ signals, while the agent inhibited the late phase of GLP-1–induced Ca²⁺ signals but not the initial Ca²⁺ spike (Fig. 1E). Indeed, stimulation of mouse β-cells with 200 nmol/l insulin under the same conditions used by Johnson and Misler did not affect NAADP levels (online appendix Fig. 11).

In conclusion, our results revealed for the first time that GLP-1 elevates [Ca²⁺]_i via stimulation of NAADP and cyclic ADPR production and that NAADP acts as an initiator for GLP-1–mediated Ca²⁺ signals and effectively increases insulin secretion as much as GLP-1 does. Our results also indicated that the downstream modulators for GLP-1 consist of CD38 and hitherto unidentified ADPR cyclase(s)/NAADP–producing enzyme(s), which produces NAADP and cyclic ADPR spatiotemporally or differentially. Finally, consistent with previous findings (42,43) of GLP-1–evoked interrelationships between cAMP and Ca²⁺, our results indicate that both NAADP and cyclic ADPR may act as key modulators interdependent of [cAMP]_i and [Ca²⁺]_i oscillation.

M.-J.I. and B.-J.K. are the recipients of the BK21 Program of the Ministry of Education of Korea. This work was supported by Korea Research Foundation Grant KRF-2004-005-E00108, a fund of International Collaboration Study from Chonbuk National University (U.-H.K.), and a grant from the National R&D Program for Cancer Control, Ministry of Health and Welfare of Korea (0620220-1) (C.-Y.Y.)

We thank Sang-Hee Yu for the excellent technical support.