Glucokinase Is an Integral Component of the Insulin Granules in Glucose-Responsive Insulin Secretory Cells and Does Not Translocate During Glucose Stimulation

MIN6 cells (p20–29) were cultured in Dulbecco’s modified Eagle’s medium containing 25 mmol/l glucose, 50 μmol/l β-mercaptoethanol, and 15% (vol/vol) fetal bovine serum (FBS) (23). Using a process similar to that for MIN6 cells, βTC-tet cells were cultured in medium, but the serum supplement was 2.5% FBS and 15% horse serum. Once confluent, tetracycline (1 μg/ml) was added to induce growth arrest, and the cells were cultured for 16–20 days (24). INS 1E cells were cultured in RPMI-1640 media containing 50 μmol/l β-mercaptoethanol, 1 mmol/l sodium pyruvate, and 5% FBS (25). For determination of glucose phosphorylation, the cells were cultured in 24-well plates and incubated in medium containing [2-³H]glucose and either 5 or 25 mmol/l glucose as described previously (20).

MIN6 and INS 1E cells were cultured in 24-well plates and washed with Krebs-Ringer buffer (26). They were preincubated for 30 min at 37°C in Krebs Ringer buffer containing 5 mmol/l glucose, followed by a 2-h incubation at the indicated glucose concentration for determination of insulin secretion (26). Insulin was determined in the medium using a rat insulin enzyme-linked immunosorbent assay (Mercodia, Uppsala, Sweden).

Free and bound glucokinase activity and immunoreactivity in MIN6 cells were determined after incubation in medium with the glucose concentrations indicated. Digitonin permeabilization was carried out as previously described (20). For cell homogenization, the monolayers were scraped in 150 mmol/l KCl, 3 mmol/l Hepes, and 1 mmol/l dithiothreitol (pH 7.2) and passed six times through 27-gauge and then 23-gauge needles. Half of the homogenate was centrifuged at 13,000g (15 min) to generate supernatant and pellet fractions.

MIN6 organelles were fractionated on a Percoll gradient based on a method described by Hutton and Peshivaria (27). Briefly, cells from six flasks (75 cm²) were scraped in buffer containing 275 mmol/l sucrose, 10 mmol/l MES, 1 mmol/l EGTA, 0.2 mmol/l benzamidine, 0.2 mol/l PMSF, 10 μg/ml leupeptin, 10 μg/ml antipain, and 10 μg/ml pepstatin and passed six times through 27-gauge and then 23-gauge needles. The homogenate was precleared at 1,700g and the supernatant layered on 27% (wt/vol) Percoll in extraction buffer and centrifuged at 35,000g (50 min). The top 10 ml were discarded, and 1-ml fractions were collected and washed twice with extraction buffer at 13,000g (15 min) to remove the Percoll. Pellets were suspended in extraction buffer for determination of enzyme activity, immunoreactivity, and protein.

For hypotonic lysis, fractions containing the insulin granules were pooled, sedimented at 13,000g, and lysed in 5 mmol/l sodium carbonate (28) for 20 min. The lysate was layered on a sucrose gradient containing 250 μl 90% sucrose, 500 μl 35% sucrose/2.5 mmol/l sodium carbonate, and 500 μl 5% sucrose/2.5 mmol/l sodium carbonate and centrifuged at 140,000g (16 h), and 1 ml fractions were removed for immunoblotting. For detergent lysis, the granules were suspended in HSE buffer (20 mmol/l Hepes, 150 mmol/l NaCl, and 5 mmol/l EDTA, pH 7.4) without or with 1% (wt/vol) Triton-X100 (29) for 30 min. The lysate was centrifuged at 13,000g, and protein in the supernatant and pellet was determined by immunoblotting.

Glucokinase and low-K_m hexokinase activity were determined as previously described (30) except for assays on the insulin granules, where a radiochemical assay was used (31). Glutamate dehydrogenase (32) and aryl sulfatase (33) were determined as previously described. Protein was determined using Biorad reagent, and enzyme activities were expressed as milliunits per milligrams of cell protein, where 1 mU is the amount converting 1 nmol of substrate per minute.

Samples containing 20 μg protein were resolved by SDS-PAGE on either 10% (glucokinase, phogrin, glyceraldehyde-3-phosphate dehydrogenase [GAPDH], cathepsin D, and PFK-2 [6-phosphofructo-2-kinase-2/fructose-2,6-bisphosphatase]) or 15% (insulin) polyacrylamide gels. After electrotransfer of protein to nitrocellulose membranes, these were incubated for 90 min with primary antibody: glucokinase (H-88; Santa Cruz Biotech, Santa Cruz, CA), insulin (Chemicon, Temecula, CA), GAPDH (Hytest; Biotrend, Köln, Germany), cathepsin D (Santa Cruz Biotech, Santa Cruz, CA), phogrin (gift from Dr. John Hutton) (34), and PFK-2 (chicken polyclonal IgY against recombinant fructose 2,6-bisphosphatase domain) (35). After washing, membranes were incubated for 1 h with the appropriate peroxidase-conjugated anti-IgG (Jackson Immunoresearch Laboratories). Immunoreactive bands were visualized using an enhanced chemiluminescence kit (Amersham Biosciences) and determined using the Alpha Imager Program (Alpha Immunotech).

Cells cultured on coverslips coated with 1% gelatin were fixed and immunostained as previously described (20). Confocal images were visualized and captured using a Leica TCS-SP2-UV microscope with a 63×, NA 1.3 objective.

Colocalization between glucokinase and insulin granules has been reported previously in intact islets (18) and MIN6 cells (20) as well as in glucose-starved βTC cells (21). The latter study reported glucose-induced translocation of granule-associated glucokinase to the cytoplasm. In the present study, colocalization of glucokinase with insulin granules was observed in MIN6 cells incubated at either 5 or 25 mmol/l glucose, with no apparent difference between low and high glucose (Fig. 1). Similarly, colocalization between glucokinase and insulin granules was observed in INS1E and βTC-tet cells incubated at both 5 or 25 mmol/l glucose (data not shown). Glucokinase catalytic activity was detectable spectrometrically in both the supernatant and pellet fractions of MIN6 cells but only in the supernatant of INS1E and βTC-tet cells (Table 1), possibly due to the low granule content or high pellet NADPH oxidase activity, respectively. However, glucokinase immunoreactivity was detectable in the pellet fraction of all three cell lines (data not shown). Glucose-responsive insulin secretion in MIN6 and INS1E was in agreement with previously reported values (26,36). Although the ratio of glucose phosphorylation at high versus low glucose was similar for all three cell lines (Table 1), the activity of low-K_m hexokinase was lowest in INS1E and highest in βTC-tet. MIN6 rather than INS1E was used for isolation of granules (see below) because of their high granule content.

Because immunofluorescence staining is not sufficiently sensitive to detect adaptive changes in glucokinase distribution in β-cells, we used digitonin permeabilization and cell homogenization to test for changes in distribution of glucokinase activity and immunoreactivity. MIN6 cells are conventionally cultured in medium containing 25 mmol/l glucose (23). In these conditions, the digitonin method released 45% of catalytic activity and 60% of immunoreactive glucokinase, while homogenization of the cells released 70% of immunoreactive and catalytically active glucokinase (Fig. 2A–D, t = 0). This indicates that the bound fraction, as determined by the digitonin method, includes enzymes that are either weakly bound to organelles or associated with subcellular structures that are disrupted by homogenization.

Incubation of the cells in glucose-free medium for 1 or 3 h caused a decrease in glucokinase activity and immunoreactivity associated with the bound fraction by the digitonin technique (Fig. 2A and C). Qualitatively similar but smaller changes were observed by cell homogenization (Fig. 2B and D). Incubation of the cells for 2 h in glucose-free medium followed by 1 h with either 5 or 25 mmol/l glucose was associated with an increase in bound glucokinase activity and a decrease in free activity, indicating reversal of the changes caused by the glucose-free medium (Fig. 2E and F). These experiments on MIN6 and others on INS1E (data not shown) did not provide evidence for dissociation of bound glucokinase by glucose.

Glucokinase immunoreactivity associated with the bound fraction after either digitonin permeabilization or cell homogenization (Fig. 2) may represent glucokinase binding to more than one type of organelle. To investigate this further, cell homogenates were fractionated on a Percoll gradient to separate the insulin granules from other organelles, and the fractions were immunoblotted for glucokinase, insulin, and phogrin (a marker for the insulin granule membrane) (34). Marker enzymes for mitochondria (glutamate dehydrogenase) and lysosomes (arylsuphatase) were also determined. There were two main peaks of glucokinase immunoreactivity that coincided with the major peak of insulin immunoreactivity (fractions 2–7 from the bottom of the gradient) corresponding to the dense insulin granules and the mitochondrial peak (fractions 17–22) based on colocalization with glutamate dehydrogenase (Fig. 3). Glucokinase activity was detectable radiochemically in the insulin granule fractions (118 ± 33 pmol · min^–1 · mg protein^–1, n = 6). There was negligible activity of glutamate dehydrogenase in the insulin granule fractions, but there was immunoreactivity to insulin and phogrin in the mitochondria-rich fractions. The ratio of insulin to phogrin immunoreactivity was higher in the dense insulin granule fractions compared with the low-density mitochondria-rich fractions, suggesting that the latter most likely contain either immature granules or remnant vesicles that have released their insulin. The ratio of glucokinase to phogrin immunoreactivity was fourfold higher in the low-density fractions (Table 2). Fractionation of INS-1E and βTC-tet organelles confirmed the association of glucokinase immunoreactivity with insulin granules in these cell lines (data not shown).

To test for adaptive changes of glucokinase association with organelles, MIN6 was incubated in either glucose-free medium or medium containing 25 mmol/l glucose for 1 h before isolation of the organelles. Immunoreactive glucokinase associated with both the granules and the low-density fractions was not released by high compared with low glucose (Fig. 3, Table 2).

Percoll fractions were immunoblotted for PFK-2, a glucokinase-binding protein (35), and for GAPDH, which has been implicated in granule movement and membrane fusion (37). Both the insulin granules and low-density organelles showed immunoreactivity to PFK-2 and GAPDH (Table 2). When compared with recovery of phogrin, recovery of GAPDH was similar in the insulin granules and low-density fractions, whereas the recovery of PFK-2 relative to that of phogrin was twofold higher in the latter fractions, similar to the higher glucokinase-to-phogrin ratio (Table 2).

To determine whether the glucokinase immunoreactivity recovered with the insulin granules is associated with the core or the periphery of the granule, insulin granules were lysed hypotonically and the lysate fractionated on a sucrose gradient. The fractions were immunoblotted for insulin, phogrin, glucokinase, GAPDH, PFK-2, and the lysosomal marker cathepsin-D, since the lysosomal peak overlaps with the insulin granule peak (Fig. 4). There was negligible overlap in distribution of phogrin and insulin immunoreactivity, confirming total release of the insulin following hypotonic lysis. Glucokinase was recovered with phogrin and not with the insulin, suggesting that it is not associated with the core components. Likewise, GAPDH and PFK-2 cofractionated with glucokinase and phogrin, whereas cathepsin D did not colocalize with these enzymes or with phogrin, confirming that immunoreactivity to these proteins could not be accounted for by lysosomal contamination.

Granule-associated glucokinase characteristically appears as a double band within the 50-kDa region by immunoblotting. Following sonication and/or treatment of the granules with detergent and subsequent centrifugation, the upper glucokinase band is released more easily into the supernatant, while the lower band remains associated with the pellet fraction (Fig. 5). This is suggestive of different molecular forms of glucokinase that may differ in their association with the insulin granules.

In this study, we confirmed colocalization of glucokinase with insulin granules by fluorescence microscopy in MIN6 cells cultured at either low or high glucose in agreement with previous studies using cell microscopy techniques (18–21). Since techniques based on fluorescence microscopy do not allow quantitative determination of the proportion of total cellular glucokinase associated with a particular organelle type and also do not distinguish between enzyme that is an integral component of the organelle as opposed to weakly bound to the periphery, we used subcellular fractionation to determine the association of glucokinase with different organelles

We show in this study that ∼30–50% of the glucokinase is associated with the particulate fraction after homogenization of MIN6 cells. After fractionation of the organelles, the glucokinase is distributed primarily into two fractions, one of which corresponds to the dense insulin granules, while the other cofractionates with mitochondria and low-density phogrin-containing organelles with a low insulin content. Although the insulin granule peak overlaps, in part, with the lysosomes, we could exclude the possibility that granule-associated glucokinase included immunoreactive protein entrapped in lysosomes because after hypotonic lysis and fractionation on a sucrose gradient, glucokinase did not colocalize with a lysosomal marker. The glucokinase immunoreactivity associated with the low-density vesicles could represent remnant granules after insulin secretion with partial loss of membrane components, as suggested by the increase in the glucokinase-to-phogrin ratio or the association of glucokinase with mitochondria.

Previous studies reporting colocalization of glucokinase with insulin granules did not address whether glucokinase is an integral component of the insulin granule or is loosely bound to the granule periphery. In this study, hypotonic lysis of insulin granules followed by sucrose gradient fractionation enabled separation of insulin from the granule membrane marker phogrin. Because immunoreactive glucokinase distributed with phogrin and not insulin, it can be inferred that glucokinase is associated with the outer structure rather than the granule core. Like phogrin, glucokinase remained associated with the granules after detergent extraction, suggesting that it is an integral component of the granules. The double band of granule-associated glucokinase, as shown by immunoblotting, suggests that there may be different molecular forms of glucokinase associated with the granules, possibly representing either cross-linking of disulphide bonds (38) or nitrosylation of cysteine residues (22). The faster migrating form was more resistant to detergent extraction.

Two other proteins that colocalized with glucokinase and phogrin are GAPDH and PFK-2. The latter protein catalyzes the formation and degradation of fructose 2,6-bisphosphate and is a binding partner for glucokinase (35). Since a higher fraction of PFK-2 than glucokinase was recovered with both the insulin granules and the low-density organelles, it raises a question as to whether PFK2 acts as the binding protein or molecular scaffold for glucokinase or whether it has a role in stabilizing glucokinase, as was recently shown in insulin secretory cells (39).

We tested for association of GAPDH with the insulin granules because a previous study on hamster insulinoma tumor insulinoma lysates showed that a cytoplasmic protein with immunoreactivity to GAPDH catalyzed fusion of insulin granules with plasma membrane vesicles in vitro (37). In this study, we demonstrated the association of GAPDH with insulin granules. This raises a question as to whether the GAPDH isoform that catalyzes fusion is also present in the granules in addition to the cytoplasm and is causing granule fusion with the membrane. The association of glucokinase with phogrin and GAPDH therefore suggests a role for glucokinase in glucose-induced granule movement or membrane fusion.

Two possible roles can be envisaged for granule-associated glucokinase. It either represents a store of “kinetically inactive” enzyme that is released to the cytoplasm during glucose-induced insulin secretion, or it has a direct kinetic function when associated with the granule that determines glucose-mediated changes in granule maturation, movement, or membrane fusion during insulin secretion. The attraction of the former hypothesis is its analogy to the nuclear compartmentation of glucokinase in the hepatocyte. Rizzo and colleagues (21,22) reported that glucokinase translocates from the granules to the cytoplasm in response to glucose or insulin based on a decline in glucokinase immunoreactivity that is associated with the digitonin bound fraction after the addition of glucose and also based on differences in the rate of granule fluorescence recovery after photobleaching (FRAP) in cells expressing a glucokinase-–yellow fluorescent protein fusion protein. Using similar digitonin permeablization or cell homogenization, we found no evidence for either glucose-induced translocation (present study) or insulin-induced translocation (L.A. and C.A., unpublished observations). To enable direct analysis of the effect of glucose on granule-associated glucokinase, we used Percoll gradient fractionation to separate the granules from other organelles and found no evidence for adaptive changes in the glucokinase content of the granules. A tentative explanation for the apparently conflicting results between the present study and the previous study (21) is that in the present study we determined a glucokinase fraction that is “integral” to the granule and is retained after density gradient fractionation, whereas the FRAP method used previously measures a diffusible glucokinase fraction in the vicinity of the granules. Differences in rates of diffusion may reflect changes in macromolecular crowding around the granule in response to glucose or insulin.

In summary, we provide direct evidence for the association of glucokinase with insulin granules based on fractionation of organelles and show that following hypotonic lysis, glucokinase cofractionates with phogrin, GAPDH, and PFK2 but not insulin. We suggest that glucokinase is an integral component of the granule, that it does not translocate in response to glucose, and that it may be involved in glucose-mediated changes in granule movement or membrane fusion.

Funding of a PhD studentship to C.A. by the Development Trust at the University of Newcastle is gratefully acknowledged. We also thank the Wellcome Trust, Diabetes U.K. (grant BDA RD01/0002364), and the Medical Research Council (grant G0100348) for equipment grant support.

We thank Drs. Jun-ichi Miyazaki, Claus Wollheim, and Shimon Efrat for providing MIN6, INS 1E, and βTC-tet cells, respectively, and Dr. John Hutton for the kind gift of the phogrin antibody.