Recruitment of Epac2A to Insulin Granule Docking Sites Regulates Priming for Exocytosis

Control of protein localization is an important mechanism in signal transduction and regulation of cell function. The cAMP-activated guanine nucleotide exchange factors Epac1 and Epac2 mediate a variety of cAMP-regulated cellular processes, including neurotransmitter and hormone release, intracellular Ca²⁺ handling, cell adhesion, junction formation, cell proliferation, and gene transcription (1–7). They were recently found to undergo activity-dependent redistribution from the cytoplasm to the plasma membrane in different types of cells (8–10). The two proteins differ somewhat in tissue distribution and domain structure (1,7). Epac1 is relatively widely expressed, whereas Epac2 shows highest expression in the central nervous system, pancreas, and adrenals (1,11,12). Alternative promoter usage and splicing give rise to three isoforms of Epac2 (11,13). Both Epac1 and the Epac2 isoforms are multidomain proteins with an N-terminal regulatory region and a COOH-terminal catalytic region (14). The regulatory region consists of a low-affinity cyclic nucleotide-binding (CNB) domain (CNB-A; only present in the longest Epac2 splice version Epac2A [15]), a disheveled, Egl-10, and pleckstrin (DEP) domain, and a high-affinity CNB domain (CNB-B). The DEP domain is critical for the membrane binding of Epac1 (16), but its role in Epac2 is unclear, and the shortest isoform Epac2C lacks DEP domain. The catalytic region contains a Ras exchange motif, a Ras association (RA) domain, and a CDC25 homology domain. The latter domain catalyzes GTP/GDP exchange on the Rap family of small GTPases (1,12,17). In the absence of cAMP, Epac is inactive and maintained in a compact structure with the regulatory region blocking the catalytic domain (14). cAMP binding to the CNB domain(s) results in a conformational change that releases the autoinhibition and exposes the catalytic region (14,18,19). Only the high-affinity domain, CNB-B (K_d of 1.2 μmol/L [15]), seems important for catalytic regulation of Epac2 (18). The function of the CNB-A domain, with much lower affinity for cAMP (K_d 87 μmol/L [15]), is less clear, but it has been suggested to be involved in the subcellular targeting of Epac2 (11).

Epac2 has been found to positively regulate insulin secretion from pancreatic β-cells (4,20–22). We recently discovered that Epac2 activation results in translocation of the protein from the cytoplasm to the plasma membrane in insulin-secreting β-cells (10) in a process controlled by cAMP and Ca²⁺, which in β-cells are key signals for exocytosis of insulin secretory granules (23). Glucose, the main physiological stimulator of insulin release, acts by increasing the concentrations of Ca²⁺ and cAMP in the β-cell cytoplasm, and the gut hormone glucagon-like peptide 1 (GLP-1) amplifies glucose-stimulated insulin secretion by stimulating cAMP production. Both glucose and GLP-1 were consequently found to induce translocation of Epac2 to the plasma membrane (10), but the detailed molecular targeting mechanisms and functional consequences are not clear. The cAMP-induced translocation of Epac2 requires intact high-affinity CNB-B and RA domains (10). However, the roles of the low-affinity CNB-A domain and the adjacent DEP domain remain unknown. In the current study, we have extended the analysis of the molecular mechanism underlying Epac2 membrane translocation and use a single-cell, real-time imaging approach to define the role of the different domains in the regulatory region of the protein. Furthermore, as Epac2 stimulates insulin secretion from β-cells, we investigated how Epac2 binding to the plasma membrane is related to insulin secretory granules and whether the translocation influences exocytosis. We found that Epac2A promotes granule priming and fusion by clustering at the exocytosis sites beneath the plasma membrane. This translocation involves two distinct localization mechanisms, one mediating membrane binding (CNB-B and RA domain) and another (CNB-A) directing Epac2A to components of the vesicle docking and fusion machinery.

3-Isobutyl-1-methylxanthine (IBMX), cAMP, GLP-1 (7–36 amide), forskolin, and poly-L-lysine were from Sigma-Aldrich (Stockholm, Sweden). Plasmids and adenovirus encoding mouse Epac2A fused to the fluorescent proteins EGFP or mCherry (GFP-Epac2A-wt and mCherry-Epac2A-wt) have been described (10), and a plasmid encoding yellow fluorescent protein–tagged Epac1 was a gift from Drs. J. Bos and H. Rehmann, University of Utrecht (Utrecht, the Netherlands). Point mutations were generated by site-directed mutagenesis, and domain deletions were created by PCR. Plasmids and adenovirus encoding fluorescence-tagged neuropeptide Y (NPY) to visualize secretory granules (24) and a cAMP-dependent protein kinase–based cAMP biosensor (25) have previously been reported.

MIN6 β-cells (26) of passages 21–30 were cultured and transfected as described (10). Human islets were obtained from three normoglycemic (two male and one female; aged 67–73 years) and four cadaveric organ donors with type 2 diabetes (two male and two female; aged 55–59 years) via the Nordic Network for Clinical Islet Transplantation in Uppsala with approval from the Uppsala human ethics committee. The isolated islets were cultured up to 7 days as detailed in Tian et al. (27). For some experiments, the islets were dispersed into single cells and small clusters using cell dissociation buffer (Thermo Fisher Scientific) containing 10% trypsin. The cell suspension was added onto poly-L-lysine–coated coverslips and cultured overnight. The islets or cells were infected with adenovirus by 3- to 4-h incubation with 105 fluorescence-forming units per islet, followed by overnight culture before use.

Most Epac translocation experiments were performed in experimental buffer containing 138 mmol NaCl, 4.8 mmol KCl, 1.2 mmol MgCl₂, 1.3 mmol CaCl₂, 3 mmol d-glucose, 25 mmol HEPES (pH 7.40), and 0.1% BSA. Coverslips with cells or islets were preincubated in this medium during 30–60 min, mounted in an open 50-µL chamber, and superfused with buffer at a rate of 0.2 mL/min. Where indicated, MIN6 cells were permeabilized with α-toxin as previously described (10). All of these experiments were performed at 37°C.

In granule localization experiments, the cells were maintained at 32°C in a solution containing 138 mmol NaCl, 5.6 mmol KCl, 1.2 mmol MgCl₂, 2.6 mmol CaCl₂, 3 mmol d-glucose, and 5 mmol HEPES (pH 7.4 with NaOH). In Figs. 7A and 8, exocytosis was evoked in medium containing 10 mmol/L d-glucose, 2 μmol/L forskolin, and 200 μmol/L diazoxide by computer-timed local application of high K⁺ (75 mmol/L KCl equimolarly replacing NaCl) through a pressurized glass electrode. In Fig. 7B, exocytosis was stimulated by elevation of d-glucose to 20 mmol/L.

The subcellular localization of the fluorescent Epac constructs was assessed using spinning-disk (Andor Technology, Belfast, U.K.) (10) and point-scanning (Zeiss 780; Carl Zeiss, Oberkochen, Germany) confocal microscopes with 60×, numerical aperture 1.4 objectives (Nikon, Tokyo, Japan). Epac translocation, granule localization, and exocytosis were recorded with total internal reflection fluorescence (TIRF) microscopy. Most Epac2 experiments and the cAMP recordings were performed with an Eclipse Ti microscope (Nikon) with a TIRF illuminator and a 60×, numerical aperture 1.45 objective. Diode-pumped solid-state lasers (Cobolt, Stockholm, Sweden) provided excitation light at 491 (for EGFP) and 561 nm (mCherry). Fluorescence at 530-nm/50-nm half-bandwidth (EGFP) and 620-nm long-pass (mCherry) was detected with a back-illuminated EMCCD camera (iXON DU897; Andor Technology) controlled by MetaFluor software (Molecular Devices, Downingtown, PA). Images were acquired every 5 s. Granule localization and exocytosis experiments were made using a custom-built TIRF microscope based on an AxioObserver Z1 with a 100×/1.45 objective (Carl Zeiss) as described in Gandasi and Barg (28).

Epac translocation was analyzed as changes in TIRF intensity over time in regions of interest corresponding to single cells. The signals were normalized to the prestimulatory level after background subtraction. All experiments were repeated on at least two different cell preparations. The data are expressed as means ± SE or as scatter plots with highlighted median values. Two-tailed Student t test or Wilcoxon signed-rank test was used to assess statistical differences.

Colocalization of Epac with secretory granules was quantified essentially as described (29). In brief, an algorithm implemented in MetaMorph read the average fluorescence in a central circle (c) of three-pixel (0.5 µm) diameter, a surrounding annulus (a) with a five-pixel outer diameter, and a cell-free background area (bg). The specific on-granule fluorescence was defined as ∆F = c − a. To obtain off-granule fluorescence, S, the annulus value was background-corrected (S = a − bg). ΔF/S is a measure of the degree of binding to the granule location (28). Positive values indicate enrichment and negative values exclusion from the granule site.

In Figs. 4–6, the ∆F/S value for each cell represents an average from a time-series of 50 consecutive frames. In Fig. 7, ΔF/S was calculated for every frame, and only cells with intensity level (cell − background) >200 (16-bit digitization) were included. Note that ∆F is given as per-pixel average for the entire 0.5-µm circle, and ∆F/S values are therefore seemingly small. Assuming a cluster size of 50 nm, ΔF/S = 0.1 corresponds to at least 20-fold enrichment in the cluster beneath a granule. Single-granule exocytosis events were detected manually as a sudden (less than three frames) drop in fluorescence.

To determine the mechanisms underlying Epac2A binding to the plasma membrane, fluorescent protein–tagged deletion mutants lacking either the low-affinity cAMP-binding CNB-A domain (mCherry-Epac2A-ΔCNB-A) or the DEP domain (mCherry-Epac2A-ΔDEP) were expressed together with wild-type Epac2A (GFP-Epac2A-wt) in MIN6 cells (Fig. 1A). As previously reported (10), cAMP elevation by the phosphodiesterase inhibitor IBMX (100 μmol/L) induced translocation of wild-type Epac2A, recorded in TIRF microscopy as an increase of fluorescence in the cell footprint area (Fig. 1B). Quantification of the IBMX responses showed that the translocation of the ΔCNB-A mutant was more pronounced than wild-type and that of ΔDEP impaired (Fig. 1C and D). Control experiments showed nearly identical translocation of coexpressed GFP- and mCherry-tagged Epac2A-wt, indicating that translocation responses can be quantitatively compared between different Epac2 constructs (Fig. 1E and F). The reduced translocation of the ΔDEP mutant may be explained by increased prestimulatory membrane association, as evaluated by confocal imaging of wild-type and mutant Epac2A (Fig. 1G). Both constructs showed a fairly homogenous cytoplasmic distribution. The relative membrane/cytoplasm intensity for ΔDEP was 0.89 ± 0.04 vs. 0.73 ± 0.04 for coexpressed wild-type Epac2A (P < 0.01; n = 14 cells and 4 experiments) and 0.72 ± 0.03 for ΔCNB-A vs. 0.72 ± 0.05 for wild-type (n = 12 cells and 4 experiments) in unstimulated cells.

The cAMP dependence of Epac2A translocation was analyzed in permeabilized MIN6 cells. Superfusion of increasing cAMP concentrations resulted in dose-dependent membrane translocation of Epac2A-wt with half-maximal translocation at 21 μmol/L and maximal effect at ∼100 μmol/L (Fig. 2A–C). The ΔCNB-A mutant showed more pronounced translocation than wild-type at 1–10 μmol/L cAMP, but the maximal translocation amplitude was smaller and obtained already at 30 μmol/L cAMP (Fig. 2D and F). The DEP-deficient mutant showed very modest translocation, reaching maximum at 10 μmol/L cAMP (Fig. 2E and F). The averaged data in Fig. 2F show that the cAMP dependence of the translocation was left-shifted in the mutants with half-maximal effect obtained at 5.6 ± 1.2 and 1.7 ± 1.9 μmol/L for ΔCNB-A and ΔDEP, respectively.

We next investigated how a physiological stimulus of cAMP formation in β-cells, the insulinotropic hormone GLP-1, influenced Epac2A localization. At 1 nmol/L, GLP-1 evoked a prompt rise of the cAMP concentration, which was associated with translocation of Epac2A to the plasma membrane (Fig. 3A). GLP-1– and cAMP-induced Epac2A translocation was not restricted to the cell line, but also observed in human pancreatic islet cells (Fig. 3B). The molecular determinants of the translocation were further characterized in MIN6 cells. Like IBMX, GLP-1 stimulated more pronounced translocation of the ΔCNB-A mutant than of Epac2-wt (Fig. 3C). The translocation response was amplified also by a point mutation that prevents cAMP binding to the CNB-A domain (G114E [2]) (Fig. 3D). In contrast, mCherry-Epac2A-ΔDEP showed poor translocation in response to GLP-1 (Fig. 3E). Consistent with previous observations (10), a point mutation preventing cAMP binding of the CNB-B domain (G422D [2]), or one in the RA domain preventing binding to Ras-GTP (K684E [30]), strongly suppressed GLP-1–induced Epac2A translocation (Fig. 3F–H).

To analyze the distribution of Epac2 in relation to the localization of near-membrane insulin granules, the cells were cotransfected with NPY-EGFP as granule marker and mCherry-labeled Epac2 versions and imaged with TIRF microscopy. In unstimulated cells exposed to 3 mmol/L glucose, mCherry-Epac2A-wt distributed diffusely throughout the footprint area of the cell with slight preference for granules (∆F/S = 0.010 ± 0.003 vs. 0 for random locations; P < 0.001) (Fig. 4). After elevation of cAMP with 2 μmol/L of the adenylyl cyclase activator forskolin, ∼40% of the cells showed apparent clustering of Epac2A at the sites of docked granules. The extent of clustering was related to expression level and most pronounced in cells with weak mCherry-Epac2A fluorescence (data not shown). Average ∆F/S for all cells increased sixfold in the presence of forskolin (P < 0.001) (Fig. 4).

Forskolin failed to induce clustering of Epac2A with point mutations in the high-affinity cAMP-binding CNB-B domain (G422D) and the RA domain (K684E) (Fig. 4). Granule-associated clusters were also absent with ∆CNB-A and the mutant with cAMP binding–deficient CNB-A domain (G114E). In contrast, the ΔDEP mutant clustered to the same extent as the forskolin-treated wild-type (P < 0.001 vs. wild-type control), irrespective of forskolin (Fig. 4). These findings could not be explained by systematic differences in expression of the various constructs (data not shown). Instead, the data demonstrate that the DEP domain is not required for the targeting of Epac2A to granule sites and that the CNB-A domain is dispensable for membrane recruitment, but important for the specific interaction with the granule docking sites.

Epac2A clustering was also induced by GLP-1 (10 nmol/L) in both MIN6 cells and human islet cells (Fig. 5A). This clustering was reduced in cells from donors with type 2 diabetes (Fig. 5A). Although not statistically significant, the granule clustering (ΔF/S) tended to correlate inversely with HbA_1c for all human donors (R² = 0.86) (Fig. 5B). Consistent with a role for the CNB-A domain for localization of Epac2A to granule sites, expression of the isolated CNB-A domain (including the preceding N-terminal 42 amino acid residues) demonstrated significant colocalization with vesicles (ΔF/S more than fivefold over random locations; P = 0.0024) (Fig. 6A). GLP-1 stimulation tended to reduce this clustering, possibly because of competition from endogenous Epac2A. We also tested if glucose induced Epac2A clustering. Indeed, cells exposed to 20 mmol/L glucose showed significantly higher ΔF/S than those maintained at 3 mmol/L (Fig. 6B).

To further investigate the mechanisms underlying Epac2A binding to granule sites, we investigated the closely related Epac1, which lacks an N-terminal CNB domain and therefore resembles the ∆CNB-A mutant of Epac2A. Yellow fluorescent protein–tagged Epac1 translocated to the plasma membrane in response to cAMP elevation with IBMX and forskolin (Fig. 6C). Comparison of coexpressed Epac1 and Epac2A in permeabilized cells showed lower maximal translocation amplitude for Epac1 (Fig. 6D and E). Moreover, the dose response was left-shifted with half-maximal translocation at 7.7 μmol/L cAMP for Epac1 vs. 17 μmol/L for Epac2A and maximal translocation at ∼30 and 100 μmol/L, respectively (Fig. 6E). Analyzing the protein distribution in the plasma membrane, Epac1 showed only a weak, albeit significant (P < 0.005), tendency to cluster near granule docking sites in response to forskolin (Fig. 6F and G).

To test how Epac2A association with granule docking sites at the plasma membrane affects exocytosis, the cells were stimulated by depolarization with 75 mmol/L K⁺ in the presence of 10 mmol/L glucose, 200 μmol/L diazoxide, and 2 μmol/L forskolin. Exocytotic events, seen as rapid disappearance of NPY-GFP, were observed immediately after onset of the stimulation (Fig. 7A). The analysis was restricted to cells that showed clustered Epac2A fluorescence. All granules undergoing exocytosis were selected and matched with other granules in the same cell that failed to respond to the stimulation. The former group corresponds to primed granules, whereas the latter are merely docked (28). Fluorescence from granules undergoing exocytosis decayed to zero, whereas it remained constant in time-matched controls (Fig. 7A). The granule-associated Epac2A signal, quantified as ∆F/S, was significantly higher in exocytosing granules compared with the failures (P < 0.001) (Fig. 7A). We also analyzed glucose-induced exocytosis events. Granules undergoing exocytosis in cells stimulated with 20 mmol/L glucose showed less Epac2A clustering than with forskolin treatment, but significantly more than granules failing to fuse (P < 0.001) (Fig. 7B). The data indicate that recruitment of Epac2A to granules correlates with their priming.

Next, we investigated how Epac2A mutants with different granule localization influenced insulin secretion by quantifying the number of exocytosis events after stimulation with high K⁺. Figure 8A shows that forskolin amplified secretion several-fold in cells expressing mCherry-Epac2A-wt. The response was significantly (P < 0.001) attenuated in cells expressing the translocation-incapable G422D mutant, the ΔCNB-A mutant, unable to accumulate at granules, and the ΔDEP mutant, which constitutively localizes to the granule sites (Fig. 8B). These findings show that Epac2A localization to the docking sites is necessary, but not sufficient, to promote exocytosis.

The Epac proteins are widely expressed cAMP sensors, which recently have been found to translocate from the cytoplasm to the plasma membrane upon activation (8–10). In the current study, we investigated the mechanisms and functional relevance of this spatiotemporal regulation in β-cells and found that Epac2A clusters at secretory vesicle docking sites. This distribution pattern is achieved by a dual targeting mechanism: one that directs the protein to the plasma membrane (the CNB-B and RA domains) and one mediating the specific interaction with the exocytosis machinery (CNB-A domain). We also showed that the localization process regulates granule priming and exocytosis. These findings provide novel insights into the role of cAMP in the regulation of insulin secretion.

Following cAMP binding, Epac2 changes conformation from a closed, autoinhibited state with the N-terminal regulatory region sterically hindering substrate binding at the catalytic site to an open configuration, in which the autoinhibition is relieved (18). This activation results in translocation of the protein from the cytoplasm to the plasma membrane in a cAMP- and Ca²⁺-dependent manner (10). Previous studies have demonstrated that Epac2 binds to the plasma membrane by interacting with Ras-GTP via the RA domain within its catalytic region (30,31). Accordingly, a point mutation (K684E) in this domain markedly suppresses the activity-dependent Epac2 translocation (Fig. 3) (10). cAMP binding to the second (more COOH-terminal), high-affinity CNB-B domain was also essential to enable membrane interaction (Fig. 3) (10). The role of the first (most N-terminal), low-affinity CNB-A domain is less clear, but it has been suggested that it is essential for membrane targeting (11).

In the current study, we investigated the role of the CNB-A and DEP domains using fluorescence-labeled Epac2A mutants and an assay that allows direct comparison of wild-type and mutant protein translocation dynamics in single cells. Experiments with cAMP addition to permeabilized cells showed a left-shifted dose-response relationship for Epac2A translocation after deletion of the CNB-A domain. The deletion potentially increases the accessibility of cAMP to the high-affinity CNB-B domain, because the two domains in the crystal structure occlude each other (14). A similarly improved Epac2A translocation at low to moderate cAMP levels was also observed when cAMP binding to the CNB-A domain was prevented by a single amino acid substitution, probably as a result of increased flexibility of the interface between the two CNB domains (32). The low-affinity domain seemed to contribute to membrane binding only at very high (>30 μmol/L) cAMP concentrations, but it is questionable whether this effect has any physiological relevance, because the translocation response to a high concentration of GLP-1 in intact cells was not reduced by the CNB-A deletion.

DEP domains are typical membrane association domains (33), and in Epac1, the DEP domain is important for cAMP-induced translocation by binding to phosphatidic acid (16). In Epac2, cAMP binding induces significant conformational changes in the DEP domain (32), but its role for Epac2 localization has been unclear. The present data showed that translocation of the DEP deletion mutant was markedly suppressed. Because the assay measures fluorescence intensity changes in relation to the baseline, this effect likely reflects increased Epac2A-ΔDEP association with the membrane under basal conditions rather than a critical role of the DEP domain for membrane binding. This conclusion is supported by the observation of increased membrane localization in confocal images (Fig. 1) and that the mutant protein was present at granule sites irrespective of cAMP (Fig. 4).

The incretin hormone GLP-1 amplifies postprandial glucose-triggered insulin secretion mainly by increasing the β-cell level of cAMP (34), although other mechanisms may also contribute (35). GLP-1 was now found to trigger Epac2A translocation to granule docking sites in both MIN6 cells and primary human islet cells. This effect paralleled the stimulatory action of GLP-1 on cAMP production and was reduced in cells from donors with type 2 diabetes. The latter finding is interesting in light of our recent demonstration that defective cAMP signaling underlies impaired islet function under lipotoxic conditions (27). The domain dependence of the translocation was similar to that for pharmacological increase of intracellular cAMP. These observations indicate that Epac2A translocation to the plasma membrane is required for the physiologically important stimulatory effect of GLP-1 on insulin secretion.

The cAMP-induced translocation of Epac2A to the plasma membrane resulted in clustering of the protein in distinct puncta, which colocalized with secretory granules present at the membrane. There were no signs of association between Epac2A and granules in the cytoplasm as evaluated from confocal images (Fig. 1), and because the membrane binding and granule accumulation processes can be dissociated, we suggest that Epac2A first translocates to the plasma membrane and then becomes associated with the granule docking sites where the protein contributes to the priming process required for vesicle exocytosis (Fig. 8C). Although not required for the translocation of Epac2A to the plasma membrane, the CNB-A domain was responsible for this clustering near granules. The granular accumulation did not require cAMP concentrations corresponding to the very high K_d (87 μmol/L) of the CNB-A domain, because it was observed with GLP-1, and this stimulation did not seem to trigger cAMP elevations >30 μmol/L (compare data in Figs. 2 and 3). Moreover, clustering was observed also in response to glucose, which is less efficient than GLP-1 in generating cAMP (36).

The CNB-A domain is only present in the Epac2A splice variant, which is the dominating version in secretory cells, like neurons and β-cells (11). The role of the CNB-A domain for association with the exocytosis machinery is consistent with the localization of the isolated CNB-A domain in clusters at granule docking sites (Fig. 6A). Moreover, Epac1, which lacks the CNB-A domain, did not exhibit such clustering, although the protein readily translocated to the plasma membrane in response to cAMP (Fig. 6). Interestingly, the dose dependence of cAMP-induced translocation for Epac1 was more similar to the ΔCNB-A mutant than to wild-type Epac2A. Further studies are warranted to clarify the differential roles of Epac1 and Epac2 at the plasma membrane.

We suggest that the priming process, by which the vesicles are prepared for exocytosis, is regulated by cAMP-dependent clustering of Epac2A to the granule docking sites. Although the levels of endogenous Epac2A could not be assessed in the living cells, we found that granules that underwent exocytosis in response to glucose or K⁺ depolarization were associated with more fluorescence-labeled Epac2A than granules, which were docked but failed to release. Moreover, expression of Epac2A mutants associated with impaired translocation or granule accumulation reduced the number of exocytosis events. It is not clear how Epac2A promotes priming and exocytosis, but it is probably related to its interaction with RIM2α (37) or other components of the secretory machinery, such as the SUR1 subunit of the K_ATP channel (2), the granular sulphonylurea receptor (21), the active zone protein piccolo (38), or SNAP25 (39).

In summary, we studied the spatiotemporal regulation of Epac2A in β-cells and demonstrate that cAMP-elevating stimuli evokes translocation of the protein from the cytoplasm to the plasma membrane with subsequent clustering at the sites of docked secretory granules, where it regulates priming and exocytosis. Membrane localization is achieved independently of conventional lipid-binding protein modules and instead involves the RA domain that tethers Epac2A to the plasma membrane allowing the CNB-A domain to interact with the exocytosis machinery. Pharmacological stimulation of the processes that localize Epac2A to the vesicles may provide a new strategy for improving insulin secretion in type 2 diabetes.

Funding. This study was supported by grants from the Diabetes Research and Wellness Foundation, the European Foundation for the Study of Diabetes (EFSD-MSD), the Stiftelsen Familjen Ernfors Fond, Novo Nordisk Foundation, the Swedish Diabetes Foundation, and the Swedish Research Council. N.R.G. was supported by the EFSD/Lilly European Diabetes Research Programme and the Swedish Society for Medical Research. Human islets were obtained from the Nordic Network for Clinical Islet Transplantation supported by grants from the JDRF and the strategic grant consortium Excellence of Diabetes Research in Sweden (EXODIAB).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. I.A. performed and analyzed Epac2 membrane-binding experiments. N.R.G. performed and analyzed granule imaging and exocytosis experiments. S.B. designed experiments, analyzed data, and edited the manuscript. A.T. conceived the study, designed experiments, analyzed data, and wrote the manuscript. I.A., N.R.G., S.B., and A.T. edited and approved the final version of the manuscript. A.T. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 51st annual meeting of the European Association for the Study of Diabetes, Stockholm, Sweden, 14–18 September 2015.