The Insulin Secretory Granule Is the Major Site of K_ATP Channels of the Endocrine Pancreas

Murine islets were isolated from BALB/c or C57BL/6 mice pancreata by collagenase digestion followed by separation on a Ficoll gradient and purified by hand under a stereo microscope (20). Recombinant Ad.Ins-C-GFP virus was used as described for islet infection (21). Islets were washed using serum-free culture medium and then the medium was nearly completely removed by pipeting, leaving ∼50 μl. The expression levels shown were typical after 1–3 days of infection. Native and infected islets were cultured in RPMI-1640 media containing 7.5 mmol/l glucose and 10% FCS and incubated at 37°C in 5% CO₂. Live cells were used for the experiments reported here, as determined by the LIVE/DEAD Viability/Cytotoxicity Kit (Molecular Probes, Eugene, OR), granule exocytotic capacity, or both.

Green glibenclamide BODIPY FL and red glibenclamide BODIPY TR (Molecular Probes) were each used to achieve a final concentration of 40 nmol/l (22). Islets were washed using Krebs-Ringer buffer (in 7.5 mmol/l glucose) and then labeled by superfusion with the fluorescent glibenclamide within 60 min at 4°C until cellular fluorescence was readily detectable. The islets were then washed twice with Krebs-Ringer buffer and imaged.

Islets were placed into an optical recording chamber (Harvard Apparatus, Holliston, MA) at 37°C. Single-photon confocal microscopy was performed using an Olympus Fluoview 300 Confocal Laser Scanning head with an Olympus IX70 inverted microscope (Olympus, Melville, NY). Excitation of GFP, green glibenclamide BODIPY FL, or green AlexaFluor 488 conjugated to secondary antibody was by the 488 nm Argon laser line and emission detected using sharp cutoff 510IF long-pass and BA530RIF short-pass filters for PMT 1. Excitation of red glibenclamide BODIPY TR or red AlexaFluor 594 conjugated to secondary antibody was by the 543-nm green HeNe laser line and emission detected using a sharp cutoff BA610IF long-pass filter for PMT 2. Coimaging was done by sequential excitation, and simultaneous detection of emission which showed no cross-talk was detectable under these excitation-detection conditions.

Live cell insulin granule fluorescent labeling with the Adlox recombinant, designated simply “Adlox.Ins-C-GFP,” was used as described (21).

Overlap PCR was used to insert in-frame immediately after the coding sequence of mouse SUR1 or mouse K_ir6.2 in pCS2 (23), with an eight-glycine codon linker between coding segments. The simian CMV promoter of pCS2 was used to express SUR1-GFP or K_ir6.2-GFP fusions by incubating 4 μg DNA in 200 μl OptiMem media (Gibco/BRL, Gaithersburg, MD) in one tube and 8 μl Lipofectamine 2000 (Invitrogen, Carlsbad, CA) in 200 μl OptiMem media in a separate tube for 5 min at room temperature. The tubes were then combined, incubated for 20 min at room temperature, and added to islets.

We raised rabbit anti-mouse SUR1 antibodies against the 21 COOH-terminal amino acid residues of mouse SUR1 (KPEKLLSQKDSVFASFVRADK [24]), and they were affinity purified with the peptide. We raised rabbit anti-mouse K_ir6.2 antibodies against either mouse K_ir6.2 residues 377–390 (14 COOH-terminal residues KAKPKFSISPDSL [25]) or K_ir6.2 residues 329–344 (DYSKFGNTIKVPTPLC), and each affinity purified by using the corresponding peptide. The SUR1 and K_ir6.2 antibodies detected bands of the predicted molecular weights by Western analysis, and their Western and immunohistochemical signals were blocked by the cognate peptides used to generate the antibodies. Guinea pig anti-insulin was obtained from Dako (Carpinteria, CA). AlexaFluor 594 goat anti-rabbit IgG(H+L) and AlexaFluor 488 anti-guinea pig IgG(H+L) were from Molecular Probes. Freshly purified islets, naïve or infected by Ad.Ins-C-GFP for 24–48 h, were fixed with 2% paraformaldehyde in PBS for 20 min, washed in PBS three times, and blocked by incubation in 2% BSA in PBS (pH 7.5) overnight at 4°C. The islets were then incubated with primary antibodies (as indicated, each at 5 μg/ml) in blocking buffer overnight at 4°C. The islets were washed three times with PBS, incubated with labeled secondary antibodies in blocking buffer for 2 h at room temperature, and then washed three times with PBS. Confocal microscopy to detect the immunofluorescence was as described above.

Islets were prepared for electron microscopy as described (21). Briefly, islets were cryofixed, cut into ultrathin (70–100 nm) sections with a Reichert Ultracut U ultramicrotome/FC4S cryo-attachment, and then lifted on a 2.3 mol/l sucrose droplet on Formvar-coated copper grids. Sections were washed with PBS and PBG buffer (PBS, 0.5% BSA, and 0.15% glycine), blocked with 5% normal goat serum PBG (30 min), labeled with rabbit anti-K_ir6.2 “COOH-terminus” at 1:100 in PBG for 1 h, washed in PBG, labeled with goat anti-rabbit gold-conjugated secondary antibodies (each at a dilution of 1:25 for 1 h), washed in PBG and PBS, fixed in 2.5% glutaraldehyde/PBS, washed in PBS, and then washed in double-distilled H₂O. Sections were poststained in 2% neutral uranyl acetate (7 min), washed in double-distilled H₂O, stained in 4% uranyl acetate (2 min), and then embedded in 1.25% methylcellulose. Labeling was observed on a JEOL JEM 1210 electron microscope (Tokyo) at 80 kV.

While covisualizing the arrival of secretagogue with release of insulin granules, we were struck by the observation that the sulfonylurea secretagogue glibenclamide, at low concentrations, localized strongly to the secretory granule itself. We found that either green fluorescent glibenclamide BODIPY FL or red fluorescent glibenclamide BODIPY TR at 40 nmol/l applied to freshly isolated mouse islet cells localized to punctate cytoplasmic structures with spatial and dynamic properties reminiscent of insulin secretory granules (Fig. 1). At up to 10-fold higher concentration, the fluorescent BODIPY dyes by themselves did not localize to subcellular structures, indicating that the glibenclamide moiety directed the fluorescent moieties to the puncta. Furthermore, application of unlabeled 4,000 nmol/l glibenclamide largely prevented labeling of the granules by 40 nmol/l fluorescent glibenclamides. The results suggest high-affinity sulfonylurea receptors localized to the insulin secretory granule, consistent with previous findings (26,27).

We demonstrated that the major receptor site for the glibenclamides was localized to insulin secretory granules by covisualizing red glibenclamide together with the green fluorescent “Ins-C-GFP” reporter of insulin granules (21) expressed in the same islets. Application of 40 nmol/l red glibenclamide onto islets expressing Ins-C-GFP resulted in colocalization of the red and green fluorescent reporters at the same cytoplasmic punctate structures (Fig. 2). As before, the red glibenclamide bound to subcellular puncta. Green Ins-C-GFP fluorescence expressed in some of the β-cells of the same islet colocalized with red glibenclamide patterns. We studied this more closely at the single-cell level (Fig. 3). Insulin secretory granules identified by green Ins-C-GFP overlapped extensively with the red fluorescent glibenclamide. Obvious colocalization observed as yellow fluorescence was found by optically sectioning throughout the β-cell. Furthermore, the fluorescent glibenclamide resulted in high-intensity signals at the granule relative to other cell structures, suggesting that a substantial number of high-affinity sulfonylurea receptors are natively expressed at the insulin secretory granule. Little or no labeling of the β-cell membrane, or any other membranes than those of the secretory granules, was detected. The green glibenclamide had been previously shown to bind SUR of bovine monocytes and muscle cells with a subnanomolar apparent dissociation constant (K_D) ∼40 nmol/l (22). SUR1 is likely the only receptor of the endocrine pancreas competent to appreciably bind 40-nmol/l fluorescently labeled glibenclamides. The results suggest that the high-affinity sulfonylurea receptor at the insulin granule is SUR1 (previously shown to be a subunit of the pancreatic K_ATP channel), which also resides on the β-cell plasma membrane (1–7,12,13).

The results provide strong evidence for high-affinity sulfonylurea receptors localized to insulin secretory granules but do not directly identify the receptor as SUR1. The high intensity of the fluorescent glibenclamide signals at the insulin granule suggests fluorescent tags directly on SUR1 should be detectable. To test this, we determined whether SUR1 tagged with GFP localizes green fluorescence to insulin granules. We used the CMV promoter to express a COOH-terminal GFP fusion to the mouse SUR1 (“SUR1-GFP”) subunit of the mouse pancreatic K_ATP channel (23). We determined whether islet cells expressing green SUR1-GFP puncta could be colabeled with the red glibenclamide. Figure 4 demonstrates the colabeling at granule-like structures observed. Small and large size classes of punctate structures were labeled by SUR1-GFP and red glibenclamide. The small granules exhibited more uniform diameters averaging ∼400 nm and are bona fide insulin secretory granules, consistent with previous measurements (21). Colocalization of green SUR1-GFP and red fluorescent glibenclamide at secretory granules was evident by the small yellow puncta in merged images. Little or no fluorescence was detectable at the cell perimeter, consistent with a dramatically lower density of K_ATP channel protein at the plasma membrane (7). The large punctate structures strictly correlated with long-term culturing, which was done to achieve substantial expression of the fluorescent K_ATP channel subunit-GFP reporters. Large puncta were found in naive islets, as well as islets expressing transgenes, and were detectable by Ins-C-GFP or anti-insulin, indicating a close relationship to insulin granules. The large puncta might plausibly be lysosomal intermediates containing insulin-dense cores (28) or nascent insulin granules that failed to mature (29). Further study of the large puncta was obviated by their absence in fresh islets where the anti-insulin and anti-K_ATP channel subunit antibodies always identified more characteristic, small-sized insulin secretory granules (see below). Previously, we had demonstrated that the GFP used in the construction of SUR1-GFP, expressed alone, resulted in uniform green fluorescence in the cytoplasm, which accumulates in the nucleus (21). Clearly, SUR1 protein sequences directed the GFP fluorescent moiety not to these compartments but to insulin secretory granules. The SUR1-GFP results, in addition to the fluorescent glibenclamide results, suggest that native SUR1 might be highly trafficked to insulin granules.

To directly test for native SUR1 localized to insulin secretory granules, we took an immunohistochemical approach. Antibodies against a COOH-terminal peptide of mouse SUR1 were used with anti-insulin antibodies to probe freshly purified mouse islets. Figure 5 shows extensive colocalization of the anti-SUR1 and anti-insulin antibodies at secretory granules. Colocalization is evident by extensive yellow patterns of fluorescence in the merged image (n = 34 islets in seven experiments). Note that in these islets freshly prepared from the animal, only the small uniformly sized puncta that best approximate insulin granules are observed. Preincubation with the peptide that generated the anti-SUR1 antibodies, or omission of the SUR1 antibodies, resulted in no detectable fluorescence. The results suggest native SUR1 are highly localized to insulin secretory granules, as compared with the β-cell membrane. Because the SUR1 subunit has thus far always associated with the pore-forming subunit K_ir6.2 in the K_ATP channel (6,8–10), the results might be further supported by testing for a granule site for K_ir6.2.

We tested for targeting to granules of expressed K_ir6.2-GFP fusions of the mouse pancreatic K_ATP channel (23). Figure 6 shows that simultaneous expression of green fluorescent K_ir6.2-GFP and labeling by 40 nmol/l red glibenclamide results in colabeling indicated by yellow puncta in islet cells. Extensive colabeling was observed throughout optical sections in all cells (n = 47 islets in five experiments). The average fluorescence diameter was 0.435 ± 0.231 μm (n = 283 granules in seven cells of five islets in three experiments). A minority of granules showed one but not the other label, consistent with high K_ir6.2-GFP expression and low diffusion of red glibenclamide, or low K_ir6.2-GFP expression and complete diffusion. No plasma membrane fluorescence was detectable. As mentioned, exogenously expressed GFP alone is not targeted to the insulin granule (21). Localization to secretory granule of pore-forming K_ir6.2-GFP fusions, but not GFP alone, suggests that native K_ATP channel subunits traffick to insulin secretory granules.

We tested for native K_ir6.2 localized to the granule by labeling with either of two distinct antibodies specific to the COOH-terminal cytoplasmic domain of mouse K_ir6.2 and colabeling by anti-insulin. Figure 7 shows that native K_ir6.2 is localized to the insulin secretory granule. An antibody to the C-terminus of K_ir6.2 directed red fluorescent secondary antibody to insulin granules, identified by the anti-insulin labeled by green fluorescent secondary antibody (n = 24 islets in five experiments). Native K_ir6.2 in the β-cell membrane was not detectable under these conditions, in which granule K_ir6.2 protein labeling was strong. A second antibody directed against another peptide of the COOH-terminal cytoplasmic domain of K_ir6.2 showed highly similar results (data not shown). The K_ir6.2 peptides used to generate the antibodies prevented all labeling by their respective antibodies, as did omission of the primary antibodies. Given that K_ir6.2 is an integral membrane protein, antibodies against K_ir6.2 might label specifically at the membrane perimeter of the secretory granule.

For the immunoelectron microscopy experiments, we prepared 70- to 100-nm thin sections of mouse islets. Figure 8 shows that gold particle-conjugated secondary antibodies against anti-K_ir6.2 COOH-terminus antibodies localized around dense cores, decorating the perimeter of secretory granules, while little or none (≤0.02) localized to plasma membrane, mitochondria, nucleus, or other cellular structures (n = 11 islets in three experiments). The gold particle labeling was associated with the membrane of the secretory granule and not the dense core. The results show that native K_ir6.2 subunits are largely localized to the membranes of insulin secretory granules in β-cells.

The main finding of this study is the demonstration by a variety of approaches that the major site for the pancreatic K_ATP channel is the insulin secretory granule. That the fluorescent glibenclamides labeled native sulfonylurea receptors localized to insulin secretory granules was further supported by the observation that red fluorescent glibenclamide colocalized with the insulin fluorescent reporter, Ins-C-GFP. We have shown that Ins-C-GFP labels secretory granules that can be mobilized and exocytosed by insulin secretagogues (21). Our results corroborate previous evidence for a sulfonylurea receptor at the insulin secretory granule. For example, early work by Hellman et al. (30) showed evidence for dramatic uptake of glibenclamide into islets. Another study combined ³H-glibenclamide labeling, autoradiography, and electron microscopy to determine an insulin granule receptor for the sulfonylurea (26).

We extended our results by using a quite different second approach. Expression in islet β-cells of the pancreatic high-affinity sulfonylurea receptor fused to GFP (SUR1-GFP), but not the GFP expressed alone, localized fluorescence to the same intracellular punctate structures labeled by red fluorescent glibenclamide. Anti-SUR1 antibody immunohistochemistry was a third approach that indicated insulin secretory granules as the major site for SUR1. The three results each support the conclusion that native SUR1 is targeted to insulin secretory granules. Previous observations are consistent with our findings. The pancreatic SUR1 protein has an apparent molecular weight of 140 kDa with more highly glycosylated forms occurring (8). Ozanne et al. (27) showed, by density gradient centrifugation and ³H-glibenclamide labeling, that >90% of the glibenclamide mapped to intracellular structures, with the majority associated with insulin secretory granules. They showed evidence for ³H-glibenclamide binding to intracellular receptors with nanomolar affinity and for ³H-glibenclamide cross-linking to 140- and 170-kDa proteins. Additional results implicate a sulfonylurea receptor protein of 65-kDa in zymogen as well as insulin granules (31), which might have also contributed to the granule fluorescence reported here.

A secretory granule SUR1 suggested its K_ir6.2 counterpart in the pancreatic K_ATP channel would also be targeted to secretory granules. We showed that expressed K_ir6.2-GFP fusions localized to insulin granules labeled by red fluorescent glibenclamide. Second, two different anti-K_ir6.2 antibodies indicated insulin secretory granules as the major β-cell site for native K_ir6.2. Third, immunoelectron microscopy with the anti-K_ir6.2 antibodies demonstrated a membrane location of dense-core granules for K_ir6.2. Taken together, our results provide the first evidence specifically identifying SUR1 and K_ir6.2 localized to insulin secretory granules. Given that SUR1 and K_ir6.2 assemble into K_ATP channels in the endoplasmic reticulum (32–34), we propose that pancreatic K_ATP channels are then targeted by unknown mechanisms to insulin secretory granules. Once inserted into the ER membrane, the K_ATP channel, like other proteins, is expected to maintain its membrane topology for all cellular membranes. Thus, the SUR1 COOH-terminus and K_ir6.2 NH₂- and COOH-termini are cytoplasmic at granule as well as plasma membranes.

Nearly all studies on glucose-stimulated insulin secretion signaling cite solely a plasma membrane β-cell K_ATP channel. Early on, Gylfe et al. (35) reviewed numerous published data on islet uptake characteristics of radioactive sulfonylurea compounds. The five sulfonylurea compounds studied, including tolbutamide, showed rapid saturated binding with little uptake, suggesting interactions with the surface of the β-cells. Glibenclamide, however, did not rapidly reach uptake equilibrium, but rather progressively accumulated in substantial quantity within the islets (30). While discussing this exceptional mode of interaction of glibenclamide, Gylfe et al. nevertheless concluded that the insulin-releasing action of these drugs is due to binding with the β-cell membrane, because most of the sulfonylurea compounds were not appreciably internalized yet act as insulin secretagogues. The overall model presented, that cell-surface interaction of sulfonylureas decreases potassium permeability, leading to Ca²⁺ influx and insulin release, seemed reasonable at the time and explains much of what we currently know of first-phase insulin release. Like most studies on the action of sulfonylureas on insulin secretion, these studies fail to address any other cellular domain but the plasma membrane. Yet the experimental results are nearly always interpreted to mean that the isulinotropic action of sulfonylurea compounds is exclusively at the plasma membrane, and not internalized. Our results indicate that this view is incomplete and cellular mechanisms of pancreatic K_ATP channels might be productively studied at the insulin secretory granule membrane, as well as the plasma membrane.

Whenever we observed intense fluorescence localized to secretory granules, little if any cell membrane labeling was detected. The relative density of the pancreatic K_ATP channel in the secretory granule membrane is likely high relative to the β-cell membrane. The observation is consistent with electrophysiological estimates of the plasma membrane K_ATP channel, where hundreds are thought to reside (7). On one hand, the K_ATP channels spread across the β-cell plasma membrane would be largely undetectable by the confocal microscopy conditions used here, but are readily detectable by electrophysiology. On the other hand, the electrophysiology that detects K_ATP channels at the cell membrane is incapable of testing for K_ATP channels at granule membranes. Interestingly, glibenclamide is distinguished from other sulfonylureas, not only by its exceptional ability to be internalized within β-cells (30,35), but also by its superior secretory efficacy (3). The observations call for sulfonylurea drug design in the future to specifically target the intracellular secretory granule sites.

Our results suggest there are two roles for pancreatic K_ATP channels that depend on cellular position. K_ATP channels localized to the β-cell membrane play a well-established role in coupling glucose metabolism to plasma membrane excitability that switches calcium influx on for release. The calcium influx triggers exocytosis of primed insulin granules docked at the β plasma membrane, observed as first-phase release (36). K_ATP channels localized to the insulin granule membrane might also play a role in coupling glucose metabolism in a dose-dependent manner to granule transport and exocytosis, observed as second-phase release (see below). Transgenic studies show that inactivation of the pore-forming K_ir6.2 subunit renders glucose ineffective on insulin secretion from isolated islets, with a very small first-phase possibly remaining (37). Transgenic inactivation of SUR1 also disrupts the response of both phases to glucose, where only a dramatically delayed and blunted release was found that might or might not be mechanistically related to wild-type second-phase release (38). Even with calcium chronically elevated by the SUR1 inactivation, second- as well as first-phase response to elevated glucose was disrupted, consistent with granule K_ATP (gK_ATP) channels mediating an amplifying action of glucose on insulin release.

gK_ATP channels are designed and positioned to confer glucose dose dependency to the second phase of insulin release, where calcium is unlikely rate limiting (14,16,39,40). There is evidence that the ATP/ADP ratio exerts control strength over signal flow, coupling glucose metabolism and insulin exocytosis rates. In particular, increases in ATP/ADP ratios that signal elevated blood glucose levels accelerate steps distal to calcium influx through the plasma membrane, including insulin granule translocation and exocytosis (16,18,19,41). With ATP sites on K_ir6.2 that inhibit activity and ADP sites on SUR1 that antagonize the inhibition, gK_ATP channels are designed as ATP and ADP sensors of glucose metabolism. The increased ATP/ADP ratio could inhibit granule membrane potassium conductance of the gK_ATP channel, leading to changes in granule membrane potential, ion composition, or both. Verdugo and colleagues (42,43) have reported evidence that secretory granules require potassium channels for appropriate release. The gK_ATP channel might not only function as a channel as it does in the plasma membrane. The gK_ATP channel might play additional roles due to different ligand, phospholipid, and protein interactions between the two membrane domains. Thus, via increases in ATP and decreases in ADP that combine with changes in intracellular calcium and phosphoinositides, glucose metabolism could alter gK_ATP channel activities, involving interactions with lipids, motors, and associated proteins that mediate transport, recruitment, and exocytosis of the granule at the β-cell membrane (44–50). Whereas the minority of K_ATP channels in the plasma membrane provide on/off switch regulation for Ca²⁺ influx that mediates first-phase release and is permissive for second-phase release (11,14–18), the majority of K_ATP channels in secretory granule membranes are situated to accelerate, in a graded way with glucose metabolism, granule translocation, priming, and exocytosis rates that otherwise limit second-phase insulin release.

This study was supported by the Department of Cell Biology and Physiology of University of Pittsburgh School of Medicine, the Juvenile Diabetes Research Foundation (grant no. 4-1999-845), and the National Science Foundation (grant no. MCB 9817116).

The authors thank Drs. Meir Aridor, Bob Bridges, Ray Frizzell, Linton Traub, and Massimo Trucco for helpful discussions, and Rita Bottino, Bala Balamurugan, and Yigang Chen for generously providing highly pure mouse islets. This study is dedicated to the memory of Paul F. Drain, who taught us measure and patience.