Interaction of ATP Sensor, cAMP Sensor, Ca2+ Sensor, and Voltage-dependent Ca2+ Channel in Insulin Granule Exocytosis*

ATP, cAMP, and Ca2+ are the major signals in the regulation of insulin granule exocytosis in pancreatic β cells. The sensors and regulators of these signals have been characterized individually. The ATP-sensitive K+ channel, acting as the ATP sensor, couples cell metabolism to membrane potential. cAMP-GEFII, acting as a cAMP sensor, mediates cAMP-dependent, protein kinase A-independent exocytosis, which requires interaction with both Piccolo as a Ca2+ sensor and Rim2 as a Rab3 effector. l-type voltage-dependent Ca2+ channels (VDCCs) regulate Ca2+ influx. In the present study, we demonstrate interactions of these molecules. Sulfonylurea receptor 1, a subunit of ATP-sensitive K+ channels, interacts specifically with cAMP-GEFII through nucleotide-binding fold 1, and the interaction is decreased by a high concentration of cAMP. Localization of cAMP-GEFII overlaps with that of Rim2 in plasma membrane of insulin-secreting MIN6 cells. Localization of Rab3 co-incides with that of Rim2. Rim2 mutant lacking the Rab3 binding region, when overexpressed in MIN6 cells, is localized exclusively in cytoplasm, and impairs cAMP-dependent exocytosis in MIN6 cells. In addition, Rim2 and Piccolo bind directly to the α11.2-subunit of VDCC. These results indicate that ATP sensor, cAMP sensor, Ca2+ sensor, and VDCC interact with each other, which further suggests that ATP, cAMP, and Ca2+ signals in insulin granule exocytosis are integrated in a specialized domain of pancreatic β cells to facilitate stimulus-secretion coupling.

The pancreatic ␤ cell is a typical endocrine cell, in which exocytosis of insulin-containing vesicles is regulated by various intracellular signals. ATP, cAMP, and Ca 2ϩ are the major intracellular signals in the regulation of insulin secretion (19). Physiologically, the glucose concentration is the most important determinant of insulin secretion. Glucose stimulation increases the ATP concentration, which closes the K ATP channels, depolarizing the ␤ cell membrane and opening L-type VDCCs, allowing Ca 2ϩ influx and triggering exocytosis of insulin granules. The pancreatic ␤ cell K ATP channel comprises two subunits, pore-forming Kir6.2, a member of the inward rectifier K ϩ channel family, and the regulatory subunit SUR1, a receptor of the sulfonylureas widely used in the treatment of type 2 diabetes mellitus (20). Studies of Kir6.2 and SUR1knockout mice demonstrate that the K ATP channel, acting as an ATP sensor, acts as a key molecule by coupling cell metabolism to membrane potential in glucose-induced insulin secretion (21,22). On the other hand, incretins such as glucagon-like peptide-1 and glucose-dependent insulinotropic polypeptide strongly potentiate glucose-induced insulin secretion by cAMP signaling (23,24). We recently found that cAMP-GEFII/Epac2 (thereafter, cAMP-GEFII) (25)(26)(27), acting as a cAMP sensor, mediates cAMP-dependent, protein kinase A-independent insulin secretion, and that this requires interaction with both Rim2 and Piccolo (27)(28)(29). We also found that Piccolo forms homodimers or heterodimers with Rim2 in a Ca 2ϩ -dependent manner and that Piccolo rather than Rim2 may function as a Ca 2ϩ sensor (29). In addition, L-type VDCCs are known to regulate Ca 2ϩ influx into pancreatic ␤ cells. Although the sensors and regulators of ATP, cAMP, and Ca 2ϩ signals have been characterized individually, the interactions of these molecules have not been demonstrated directly.
In the present study, we have investigated interactions of the K ATP channel, cAMP-GEFII, Rim2, Piccolo, and L-type VDCC. We found that these proteins interact with each other, suggesting the integration of ATP, cAMP, and Ca 2ϩ signals in a specialized domain of pancreatic ␤ cells to facilitate stimulussecretion coupling.
Subcellular Fractionation-MIN6 cells lysed with hypotonic buffer containing 10 mM HEPES, pH 7.8, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride were used. The lysate was incubated on ice for 30 min and centrifuged at 800 ϫ g for 5 min to remove nuclear fractions. The supernatant was centrifuged at 200,000 ϫ g for 1 h and collected as cytosolic fractions. The pellet (membrane fraction) was suspended with 10 mM HEPES, pH 7.4, 1.5 mM MgCl 2 , 1 mM EDTA, 150 mM NaCl, 10% glycerol, and 0.5 mM dithiothreitol. The mixture was sonicated for 30 s twice on ice and centrifuged at 200,000 ϫ g for 1 h. The supernatant was collected as a membrane extract fraction. Each fraction was subjected to immunoblot analysis using anti-cAMP-GEFII antibody and anti-Rim2 antibody raised against Rim2 peptide (amino acid residues 348 -365).
Measurement of C-peptide Secretion-MIN6 cells were cultured in Dulbecco's modified Eagle's medium containing 25 mM glucose, 10% (v/v) fetal bovine serum under a humidified condition of 95% air and 5% CO 2 . MIN6 cells were transfected with human preproinsulin expression vector plus pCMV-luciferase, pCMV-Myc-Rim2⌬N, or pCMV-Myc-Rim2⌬C 2 . As a control, luciferase was used. Three days after transfection, the C-peptide secretory response to 1 mM of 8-bromo-cAMP (8bromoadenosine 3Ј, 5Ј cyclic monophosphate) (Sigma) in the presence of glucose (16.7 mM) for 60 min was evaluated by human C-peptide released into medium. Human C-peptide was measured by human Cpeptide RIA kit (Linco Research Inc., St. Charles, MO).
Immunocytochemistry-MIN6 cells were transfected with pCMV-Myc-full-length Rim2 and Rim2⌬N. The cultured cells were fixed with 3.7% formaldehyde in 0.1 M phosphate buffer, pH 7.4, for 30 min at room temperature and thoroughly rinsed with 0.1 M phosphate-buffered saline. After the samples were pretreated with 0.2% Triton X-100 and 10% normal donkey serum, they were incubated with rabbit anti-Myc antibody and mouse anti-Rab3 antibody, followed by Cy3-labeled donkey anti-rabbit IgG (Jackson ImmunoResearch Laboratories) and fluorescein isothiocyanate-labeled donkey anti-mouse IgG (Jackson Immu-noResearch Laboratories), respectively. The sections were mounted with coverslips using PourmaFlour (Immunon, Pittsburgh, PA) and observed by confocal laser scanning microscopy (LSM410, Carl Zeiss, Tokyo).
Binding of cAMP-GEFII and SUR1 Is Regulated by cAMP-Because cAMP induces conformational change of the protein kinase A regulatory subunit and cAMP-GEFII upon binding (32,33), we investigated whether cAMP binding affects the interaction of SUR1 and cAMP-GEFII. MBP-SUR1-NBF-1 was incubated with lysate of MIN6 cells in 1 or 100 M 8-bromo-cAMP to determine whether endogenous cAMP-GEFII in MIN6 cells interacts with SUR1-NBF-1 and if such an interaction is regulated by cAMP. Interaction between cAMP-GEFII and SUR1 is decreased by a high concentration (100 M) of 8-bromo-cAMP ( Fig. 2A). We also examined whether cAMP regulates the interaction of cAMP-GEFII with Rim2 or Piccolo. MIN6 cells treated with 1 mM 8-bromo-cAMP were subjected to in vitro binding assay with GST-Rim2 (538 -863) (27) or GST-Piccolo (4505-4758) (29) immobilized on glutathione resins. cAMP-GEFII formed a complex with both Rim2 (538 -863) and Piccolo (4505-4758) both in the presence and in the absence of 1 mM 8-bromo-cAMP (Fig. 2B). These results show that endogenous cAMP-GEFII binds to SUR1 and suggests that cAMP inhibits interaction of SUR1 and cAMP-GEFII and does not affect the interaction of cAMP-GEFII and Rim2 or Piccolo.
Protein Domains Responsible for Binding Rim2 and cAMP-GEFII-Both Rim1 and Rim2 have an amino (N)-terminal zinc finger domain, a PDZ domain, and two C 2 domains. It has been suggested that Rim1 functions as a scaffolding protein, which interacts physically through these domains with Rab3, Munc13-1, Ca 2ϩ channels, synaptotagmin I, and ␣-liprin (11, 16 -19). Assuming that Rim2 also functions as a scaffolding protein in determining which region of Rim2 interacts with cAMP-GEFII, we constructed the various Rim2 deletion mutants shown in Fig. 3A and performed a GST pull-down assay. GST-fused proteins were incubated with lysates of MIN6 cells. Fig. 3B shows the interaction of endogenous cAMP-GEFII in MIN6 cells and Rim2 mutants. Although the Rim2 mutant (198 -649) lacking a PDZ domain did not bind to cAMP-GEFII, the mutant having a PDZ domain (Rim2PDZ) bound to cAMP-GEFII, indicating that the PDZ domain contains the cAMP-GEFII binding region. The Rim2 mutant, in which the well conserved amino acids (Arg 682 , Leu 688 , and Gly 689 ) critical for binding to PDZ-interacting protein (34) are replaced with alanines, did not bind to endogenous cAMP-GEFII in MIN6 cells (Fig. 3C). This further confirms that the PDZ domain of Rim2 is responsible for binding to cAMP-GEFII, using its mutants shown in Fig. 3D. All cAMP-GEFII mutants lacking the carboxyl (C)-terminal region bound to Rim2PDZ, but the cAMP-GEFII mutant lacking the N-terminal region (residues 291-1011) did not (Fig. 3E). These results show that cAMP-GEFII, through its N-terminal region, binds to the PDZ domain of Rim2.
Subcellular Localization of cAMP-GEFII and Rim2-Localization of cAMP-GEFII and Rim2 in MIN6 cells was examined using subcellular fractionation (Fig. 4A). cAMP-GEFII is localized both in cytosolic and membrane fractions, whereas Rim2 is localized almost exclusively in membrane fractions. Rim2 was also fractionated by sucrose gradient centrifugation from MIN6 cells (Fig. 4B). Rim2 is present in plasma membrane fractions (fractions 2 and 3; Na ϩ -K ϩ -ATPase ␣-1 as a marker) and large dense-core granule fractions (fractions 7 and 8; VAMP-2 is a marker). These results show that although the intracellular localizations of cAMP-GEFII and Rim2 differ, they overlap in the membrane fraction.
The N Terminus of Rim2 Is Important for both Subcellular Localization and cAMP-dependent Exocytosis-The N-terminal regions of Rim1 and Rim2 have been shown to bind Rab3 (11, 27). We examined the role of this region in subcellular localization using the N-terminal deletion mutant (Rim2⌬N) (Fig.  5A). Wild-type Rim2 transfected into MIN6 cells is localized with Rab3 in both insulin granules and plasma membrane. In contrast, Rim2⌬N is localized exclusively in the cytoplasm. Localization of Rim2⌬N is quite different from that of wild-type Rim2, indicating that the N terminus of Rim2 is critical in localization with the granules (Fig. 5B). We next examined the effect of Rim2⌬N on cAMP-dependent insulin secretion (Fig.  5C). Insulin secretion was monitored indirectly by measuring C-peptide secretion from MIN6 cells, and transfected with human preproinsulin cDNA as reported previously (27). cAMPdependent C-peptide secretion from Rim2⌬N-transfected MIN6 cells was reduced significantly, compared with that from luciferase-transfected MIN6 cells (control). However, Rim2 mutant lacking C 2 domains (Rim2⌬C 2 ) did not inhibit cAMP-dependent C-peptide secretion. These results show that the interaction of Rim2 and Rab3 is required in cAMP-dependent insulin secretion.
Both Piccolo and Rim2 Interact Directly with L-type VDCC-Because synaptotagmin I, which has a C 2 domain, is known to interact with ␣-subunits of VDCC through their cytoplasmic loops which connect domains II and III, we examined whether Piccolo and Rim2, both of which have C 2 domains, also interact with the ␣-subunit. The C 2 domains of Piccolo bind directly to the cytoplasmic loop of the ␣ 1 1.2-subunit but not the ␣ 1 1.3subunit of L-type VDCC (Fig. 6). Similarly, C 2 domains of Rim2 bind directly to the ␣ 1 1.2-subunit but not to the ␣ 1 1.3-subunit. DISCUSSION Stimulus-secretion coupling is a critical event in pancreatic ␤ cells, where the K ATP channel controls glucose-induced and sulfonylurea-induced insulin release. Pancreatic ␤ cell K ATP channels are composed of Kir6.2 and SUR1 subunits, the poreforming subunit, and a regulatory subunit that primarily confers activation by MgADP, potassium channel openers, and inhibition by sulfonylureas, respectively. SUR1 also is required for the functional expression of the K ATP channel at the cell surface (35). In our search for a molecule that interacted with NBF-1 of SUR1, we identified cAMP-GEFII (also referred to as Epac2) (27). Because the NBF is conserved among ABC proteins, cAMP-GEFII might well bind to other ABC proteins. However, cAMP-GEFII did not bind to SUR1-NBF-2, SUR2A-NBF-1, or MDR1-NBF-1, indicating that cAMP-GEFII binds specifically to NBF-1 of SUR. In addition, cAMP-GEFII had no effect upon K ATP channel activity under various conditions (data not shown). Accordingly, it is possible that SUR1 functions as a scaffolding protein as well as the regulatory subunit of the K ATP channel. cAMP-GEFII has been shown to bind to Rim2 and Piccolo (27)(28)(29), both of which are known as CAZ proteins (8,9,11). Although the PDZ domain of Piccolo was found to bind to cAMP-GEFII (29), the region of Rim2 responsible for binding to cAMP-GEFII is not known. Using various Rim2 deletion mutants, we found that the PDZ domain of Rim2 is necessary for binding. These findings indicate that CAZ proteins having PDZ domains can form a complex with cAMP-GEFII. It is generally thought that the PDZ domain binds to consensus sequence E/D/S/T/-X-(where is a hydrophobic amino acid and X is any amino acid) of the C-terminal region of PDZ-interacting proteins (34). However, we find that cAMP-GEFII, through its N-terminal region that does not have such a consensus sequence, binds to the PDZ domain of Rim2, suggesting that cAMP-GEFII binds uniquely to the PDZ domain. As deletion of the C-terminal region with GEF activity toward Rap1 does not affect binding of cAMP-GEFII to either Rim2 or Piccolo, Rap1 is not involved. cAMP-GEFII has been shown to be responsible in cAMP-dependent, protein kinase A-independent exocytosis (27,28). Accordingly, we examined the effects of cAMP upon the binding of SUR1 and cAMP-GEFII and upon the binding of cAMP-GEFII and Rim2 or Piccolo. Interestingly, cAMP-GEFII and SUR1 binding is decreased by a high cAMP concentration. It has been shown also that cAMP binding to cAMP-GEFII induces its conformational change (33). Such an alteration might dissociate cAMP-GEFII from SUR1 and induce the interaction of cAMP-GEFII and Rim2. We found that Ca 2ϩ does not affect the binding of cAMP-GEFII to Rim2 or Piccolo. These results indicate that neither cAMP nor Ca 2ϩ are factors in the binding of cAMP-GEFII and Rim2 or Piccolo. Analyses of the subcellular localizations of cAMP-GEFII and Rim2 show that they overlap at least in the plasma membrane. Piccolo also has been shown to be located in the plasma membrane in neurons (8). Taken together, these findings suggest that cAMP-GEFII interacts with Rim2 and Piccolo in the plasma membrane.
The N-terminal region of Rim1 binds to Rab3, which is involved in Ca 2ϩ -triggered exocytosis (11). We find that deletion of the N-terminal region of Rim2 (Rim2⌬N) results in the loss of binding ability to Rab3 and that, unlike wild-type Rim2, Rim2⌬N is distributed diffusely in the cytoplasm; localization is also quite different from that of Rab3. Overexpression of Rim2⌬N in MIN6 cells shows that cAMP-dependent, Ca 2ϩtriggered secretion is reduced significantly compared with control (wild-type Rim2-transfected MIN6 cells). This reduction probably occurs because Rim2⌬N cannot transduce the cAMP signal to Rab3-mediated exocytosis and/or cannot couple to the exocytotic machinery. We reported recently that Piccolo functions as a Ca 2ϩ sensor in pancreatic ␤ cells (29). The present finding that deletion of the C 2 domain of Rim2 does not affect cAMP-potentiated, Ca 2ϩ -triggered secretion supports this report.
Synaptotagmin I, a Ca 2ϩ sensor in neurons, interacts with the ␣ 1 2.1-subunit of P/Q-type VDCC (36) and ␣ 1 2.2-subunit of N-type VDCC (37). In addition, synaptotagmin I binds to the ␣ 1 1.2-subunit of L-type VDCC in pancreatic ␤ cells (38). Rim1 has also been shown to bind to ␣ 1 1.2-subunit through C 2 domains (17). However, whether or not Piccolo binds to VDCC was unknown. Our data indicate that Rim2 and Piccolo bind directly to ␣ 1 1.2-subunits through their C 2 domains. Together, these findings clarify the link between Ca 2ϩ sensors and L-type VDCCs in exocytosis of insulin granules. The L-type VDCCs are concentrated in the area of the pancreatic ␤ cell that contains the insulin granules, suggesting that VDCCs and the granules are targeted to the same region in the pancreatic ␤ cell and that the molecular organization is critical in regulated exocytosis within a zone of voltage-dependent Ca 2ϩ entry (39).
Based on the present in vitro findings, we propose interactions of the K ATP channel as an ATP sensor, cAMP-GEFII as a cAMP sensor, Piccolo as a Ca 2ϩ sensor, and the L-type VDCC (Fig. 7). The K ATP channel interacts specifically with cAMP-GEFII through NBF-1 of SUR1. The increase in cAMP concentration dissociates cAMP-GEFII from SUR1, inducing dimerization of Rim2 or Piccolo in a Ca 2ϩ -dependent manner. Rab3 links Rim2 to the insulin granules. Further studies are required to determine the spatial and temporal regulation of the interactions of these molecules in vivo.