The ubiquitin–proteasome system is important to maintain pancreatic β-cell function. Inhibition of the proteasome significantly reduced glucose-induced insulin secretion. Key regulators of the stimulus/secretion cascade seem to be affected by protein misfolding if the proteasome is down-regulated as recently reported in humans with Type 2 diabetes. It remains unknown, however, whether the glucose sensor enzyme glucokinase is involved in this process. A direct interaction between glucokinase and ubiquitin could be shown in vivo by FRET, suggesting regulation of glucokinase by the proteasome. After proteasome inhibition glucokinase activity was significantly reduced in MIN6 cells, whereas the protein content was increased, indicating protein misfolding. Enhancing the availability of chaperones by cyclohexamide could induce refolding and restored glucokinase activity. Glucokinase aggregation due to proteasome blocking with MG132, bortezomib, epoxomicin or lactacystin could be detected in MIN6 cells, primary β-cells and hepatocytes using fluorescence-based assays. Glucokinase aggresome formation proceeded microtubule-assisted and was avoided by cyclohexamide. Thus the results of the present study provide support for glucokinase misfolding and aggregation in case of a diminished capacity of the ubiquitin–proteasome system in pancreatic β-cells. In the Type 2 diabetic situation this could contribute to reduced glucose-induced insulin secretion.
- glucose-induced insulin secretion
- pancreatic β-cell
- protein misfolding
- ubiquitin–proteasome system
The cellular protein abundance and turnover rate is controlled by the translation rate and the UPS (ubiquitin–proteasome system). Under physiological conditions, the proteasome degrades ubiquitinated proteins once their specific lifetime has been reached . In cases of oxidative and ER (endoplasmic reticulum) stress, the UPS eliminates misfolded proteins [1,2]. In pancreatic β-cells the balance between protein synthesis, folding and degradation is important to maintain proper glucose-induced insulin secretion . Incubation of human, mouse and rat pancreatic islets, as well as MIN6 β-cells, with a proteasome inhibitor resulted in a significantly reduced glucose-induced insulin secretion, whereas the insulin content remained unchanged [3–5]. Thus, in addition to insulin biosynthesis , regulation of proteins involved in the pathway of insulin secretion such as the ATP-dependent potassium and voltage-dependent calcium channels by the UPS has been discussed [2,5,6]. However, the effect of the UPS on other key regulators of the insulin secretion machinery is far less understood, but of growing interest .
Recently, it has been reported that in pancreatic islets of humans with Type 2 diabetes, the activity of the UPS is reduced compared with healthy individuals [7,8]. Using qPCR (quantitative real-time PCR) it was demonstrated that genes of the 26S proteasome were down-regulated . This is in agreement with another recently published study reporting accumulation of ubiquitinated proteins in β-cells of humans with Type 2 diabetes . Furthermore, low expression and activity of the ubiquitin C-terminal hydrolase L1 has been shown, indicating reduced access of ubiquitinated proteins to the proteasome in humans with Type 2 diabetes . Thus both studies underline that alterations in the UPS can contribute to the pathogenesis of Type 2 diabetes.
The glucose-phosphorylating enzyme glucokinase catalyses not only the first step of glycolysis in β-cells, but has the important role of the glucose sensor [9–13]. Knockdown of glucokinase resulted in a loss of glucose-induced insulin secretion . In pancreatic β-cells glucokinase is mainly regulated at the post-translational level [15–22]. The enzyme is activated by glucose and the bifunctional enzyme phosphofructo-2-kinase/fructose-2,6-bisphosphatase [16,18,19,22]. In vitro it has been described that recombinant glucokinase is ubiquitinated, suggesting quality control of glucokinase by the proteasome . Using a multiplexing strategy the stability of ~8000 human proteins has been investigated. This cell-based approach revealed for the long-living protein GAPDH (glyceraldehyde-3-phosphate dehydrogenase; gene ID 2597) a lifetime index of 6.24. For glucokinase (gene ID 2645) a significantly lower lifetime index of 4.34 has been observed, indicating medium half-life and regulation by the proteasome . Thus the aim of the present study was to investigate the impact of the UPS on glucokinase activity in pancreatic β-cells.
MATERIALS AND METHODS
MG132 (Z-Leu-Leu-Leu-al) and the ProteoStat® Aggresome Detection kit were from EnzoLife Sciences. CHX (cyclohexamide), camptothecin and nocodazole were from Sigma–Aldrich. Epoxomicin was from Merck. Bortezomib was from New England Biolabs. All primers, including random hexamer primers, and chemicals for TaqMan assays were obtained from Life Technologies. The RevertAid™ H Minus M-MuLV reverse transcriptase was from Fermentas. The GoTaq® Taq polymerase was purchased from Promega, and dNTPs were from Genecraft.
The cDNA of human β-cell glucokinase was subcloned in-frame (ApaI and BamHI restriction sites) in the pDendra2-C vector (Evrogen). Generation of pECFP–glucokinase has been described previously . ECFP was replaced by mCherry (Clontech) using AgeI and BspEI to generate mCherry–glucokinase. Tubulin from the pEYFP–Tub vector (Clontech) was subcloned into the Dendra2-C vector using XhoI and BamHI to generate Dendra2–α-tubulin. pEGFP–C1-Ub (addgene plasmid 11928) and pEGFP–C1-UbKO.G76V (addgene plasmid 11932) were generated and deposited by Dantuma et al. . EGFP was replaced by EYFP using AgeI and BsrGI restriction sites to produce EYFP–ubiquitin and EYFP–ubiquitinK0,G76V. The pcDNA3.3-d2eGFP vector (addgene plasmid 26821) was deposited by Rossi and co-workers [27,28].
Cell culture, isolation of primary cells and treatment
MIN6 and COS cells were grown in DMEM (Dulbecco's modified Eagle's medium; Biochrom) supplemented with 25 mmol/l glucose, 10% (v/v) FBS, 10 units/ml penicillin, 10 μg/ml streptomycin and 2 mmol/l glutamine in a humidified atmosphere at 37°C and 5% CO2. MIN6 cells were transfected with the vector DNA by the use of jetPEI (Qbiogene). Stable clones were selected through resistance against G418 (1200 μg/ml) and characterized further for Dendra2–glucokinase by fluorescence microscopy. Cells were seeded and grown 48 h before experiments. NMRI and C57/BL6 mice used in the present study were housed at the central animal care facility of the faculties, and islet and hepatocyte isolation were approved by the state's Animal Care Committee. Pancreatic islets were isolated from 11-week-old female NMRI mice by collagenase digestion in bicarbonate-buffered Krebs–Ringer solution. β-Cells were obtained by dispersion in calcium-free Krebs–Ringer solution and kept in RPMI 1640 medium supplemented with 5 mmol/l glucose for 24 h. Primary hepatocytes were isolated from C57/BL6 mice . Isolated cells were suspended in Williams’ medium E supplemented with 10 mmol/l glucose, 5% (v/v) FBS, 1×10−4 mmol/l dexamethasone and 1×10−5 mmol/l insulin. Hepatocytes were seeded at a density of 4×104 cells on glass coverslips and incubated for 24 h in a humidified atmosphere at 37°C and 5% CO2. MG132 (10 μmol/l), epoxomicin (1 μmol/l), CHX (10 μg/ml) and bortezomib (10, 50 and 100 nmol/l) were added to the culture medium for 3 or 12 h, and camptothecin (2.5, 5 and 10 μmol/l) for 36 h.
Glucokinase enzyme activity
MIN6 cells were seeded in 10-cm dishes at a density of 3×106 cells and treated as indicated. Glucokinase enzyme activity was measured in an enzyme-coupled photometric assay . Cells were homogenized in PBS (pH 7.4) and the protein concentration was quantified using a Bio-Rad protein assay. Enzyme activity was measured spectrophotometrically at 1 mmol/l glucose and was subtracted from the value obtained at 10 mmol/l glucose to compensate for cellular hexokinase activity.
Measurement of insulin secretion
MIN6 cells were seeded in six-well plates at a density of 3×105 cells and treated as indicated. Finally, cells were incubated for 1 h in bicarbonate-buffered Krebs–Ringer solution without glucose, supplemented with 0.1% BSA and thereafter stimulated for 1 h with 3 or 25 mmol/l glucose. Thereafter, 1 ml of the incubation buffer from each well was carefully harvested and gently centrifuged to remove detached cells (720 g for 5 min at 22°C). In these supernatants the secreted insulin was determined by RIA against a rat insulin standard. The protein concentration was quantified using the Bradford protein assay.
MTT cell viability and caspase 3 activity assays
MIN6 cells were seeded in 96-well plates at a density of 20000 cells and treated as indicated. Cell viability was then determined using a microplate-based MTT assay . MIN6 cells were seeded at a density of 50000 cells per well and caspase 3-positive cells were detected using the NucView™ 488 Caspase-3 assay kit (Biotium). The total number of cells was determined by nuclear staining with 1 μmol/l Hoechst 33342 for 15 min and corresponding fluorescence images were taken automatically using a HC 387/11-433(50)-517(40)-613(60) filter set (AHF Analysentechnik) and quantified with a scanR/IX81 microscope system (Olympus) as described previously .
Western blot analyses
MIN6 cells and hepatocytes were seeded in 6-cm dishes at a density of 2.5×105 cells and treated as indicated. Finally, cells were homogenized by sonication in lysis buffer containing 5 mmol/l Tris (pH 7.5), 5 mmol/l EDTA, 1 mmol/l PMSF, 1% Triton X-100 and protease inhibitors and insoluble material was pelleted by centrifugation (7000 g for 10 min at 4°C). Total cellular protein was fractioned by reducing SDS/PAGE (10% gel) and electroblotted on to PVDF membranes. Non-specific-binding sites of the membranes were blocked with Odyssey Blocking Buffer (Li-Cor Biosciences) for 30 min at room temperature (22 °C). Blots were incubated with glucokinase antibody [Santa Cruz Biotechnology; catalogue number sc-7908 (diluted 1:200) or catalogue number sc-1980 (diluted 1:500)] for β-cell and liver glucokinase respectively and GAPDH [Santa Cruz Biotechnology; catalogue number sc-137179 (diluted 1:2000)] at 4°C overnight, followed by incubation with the appropriate IRDye secondary antibodies (Li-Cor Biosciences) for 30 min at room temperature. Specific protein bands were visualized in the Li-Cor Infrared Imaging System (Li-Cor Biosciences). Quantification of specific protein bands was performed using Odyssey application software (Li-Cor Biosciences).
Total RNA was isolated from MIN6 or MIN6 Dendra-GK cells using the Qiagen RNeasy kit and glucokinase (GCK) gene expression was measured with TaqMan assays (human GCK, Hs01564555_m1; mouse Gck, Mm00439129_m1). The reactions were performed using the ViiA 7 real-time PCR system (Life Technologies) [31,32]. The housekeeping genes Actb (β-actin), G6pdx (glucose-6-phosphate dehydrogenase X-linked), Gapdh and Ppia (peptidylprolyl isomerase A) were used for normalization. Data analysis was performed with qBasePLUS (Biogazelle) [31,32].
MIN6 cells and hepatocytes at a density of 80000 cells, and primary β-cells from 20 islets were seeded on to glass coverslips and treated as indicated. Thereafter cells were washed twice with PBS and fixed overnight with 4% paraformaldehyde. Cells were permeabilized for 30 min with 0.5% Triton X-100 and 3 mM EDTA in ProteoStat® aggresome assay buffer (pH 8) on ice. Non-specific-binding sites were blocked with 1% BSA and 1% Triton X-100 in PBS for 20 min at room temperature. Cells were immunostained as described previously  with a goat anti-glucokinase antibody [Santa Cruz Biotechnology; catalogue number sc-1979 (diluted 1:100)] and the appropriate Cy2 (carbocyanine)-labelled secondary antibody (diluted 1:200, Dianova) for 1 h at room temperature. Thereafter cells were stained with the aggresome detection reagent containing nuclear staining (Hoechst 33342) and Proteostat® red dye. Cells were mounted with Mowiol/DABCO anti-photobleaching mounting media (Sigma). Finally, samples were analysed with an Olympus Fluoview1000 confocal microscope system. For sequential scanning the multi-line argon laser (488 nm) to excite Cy2, the diode laser (405 nm) to excite Hoechst 33342, and the DPSS (diode-pumped solid state) laser (559 nm) to excite the ProteoStat® dye and an UPLSAPO 60×1.35 NA (numerical aperture) oil-immersion objective were used. Hoechst 33342 was detected at a 425–475 nm wavelength range together with Cy2 at 500–545 nm. Light emitted by ProteoStat® dye was detected at 575–675 nm. Image processing was done using FV10-ASW software (Olympus).
Analyses of MIN6 Dendra2–glucokinase cells and FRET experiments were performed with a cellR/Olympus IX81 inverted microscope system (Olympus) equipped with a Cellcubator (Olympus) using 60% humidity, 37°C and 5% CO2. Glass-bottomed dishes were fixed on the microscope stage and images were taken with an UPLSAPO 60×1.35 NA oil-immersion objective (Olympus). ET 470/40 and 556/20 filter sets (AHF) were used to excite green and red Dendra2 respectively. Dendra2 green and red emission were detected using a 510/30-630/100 dual-band filter (AHF). An S360/40 filter (AHF) was used for photoconversion of Dendra2. Image processing was done using xcellence software (Olympus). The FRET set-up has been described previously . Sensitized emission-based FRET efficiency (FRETN) was calculated from the ECFP emission with excitation at 436 nm, EYFP emission with excitation at 436 nm and EYFP emission with excitation at 500 nm, on the basis of the calculation of Vanderklish et al. .
Data are expressed as means±S.E.M. Statistical analyses were performed by ANOVA followed by Bonferroni's test for multiple comparisons or by Student's t test using the Prism analyses program (Graphpad).
Effect of MG132 and CHX on glucose-induced insulin secretion and cell viability
MIN6 β-cells were responsive to a glucose stimulus, whereas treatment with the reversible proteasome inhibitor MG132, the translation inhibitor CHX or a combination of both for 12 h led to a loss of insulin secretion at basal (Figure 1A) and stimulating (Figure 1B) glucose concentrations. Treatment with bortezomib, a proteasome inhibitor specific for the β5 proteasome subunit, significantly reduced both basal and glucose-induced insulin secretion in a concentration-dependent manner. Cell viability determined using the MTT test was reduced only by 10 and 30% for bortezomib (50 nmol/l) and MG132 and/or CHX respectively (Figure 1C), and the number of caspase 3-positive MIN6 cells was only increased 2-fold (Figure 1D). In contrast, camptothecin, with a comparable effect on insulin secretion (Figures 1A and 1B), decreased cell viability by 50% (Figure 1C) and resulted in a 7-fold higher number of caspase 3-positive cells (Figure 1D). Therefore loss of cell viability could not be the major reason for the loss of insulin secretion observed by proteasome inhibition.
Effect of MG132, epoxomicin, lactacystin and CHX on glucokinase enzyme activity and protein content
Treatment of MIN6 cells with MG132 significantly decreased the activity of the glucose sensor enzyme glucokinase by approximately 50% (Figure 2A). Although CHX had no effect on glucokinase activity compared with the control, co-treatment with MG132 could counteract the reduction caused by MG132. In agreement with previous studies , glucokinase-overexpressing MIN6 Dendra2–glucokinase cells showed a higher glucokinase activity compared with MIN6 cells alone (results not shown). Treatment with MG132 and/or CHX resulted in a significantly reduced activity compared with the control (Figure 2B). To test whether the reduction in glucokinase activity was due to a reduced protein content, Western blot analyses were performed. Quantification of the typical line at ~53 kDa showed that MG132 treatment of MIN6 cells resulted in a significant increase in the glucokinase protein content (Figure 2C). Treatment of MIN6 cells for 12 h with more specific proteasome inhibitors, namely epoxomicin (1 μmol/l) or lactacystin (10 μmol/l) also resulted in a higher glucokinase protein content (results not shown). Glucokinase protein expression was reduced after 12 h treatment with the translation inhibitor CHX (Figure 2C). Simultaneous inhibition of the proteasome and translation resulted in a protein expression comparable with the control situation. To test whether long-living proteins are also influenced by treatment with MG132 and/or CHX, the immunoreactivity of GAPDH was quantified (Figure 2D). In contrast with the effect on glucokinase, a 12 h treatment with the inhibitors alone or in combination had no significant effect on GAPDH expression in MIN6 cells. qPCR analyses revealed a down-regulation in endogenous GCK gene expression after treatment with MG132 and/or CHX after 3 and 12 h by ~40% and 60% respectively (Supplementary Figure S1A at http://www.biochemj.org/bj/456/bj4560173add.htm). MIN6 Dendra2–glucokinase cells showed a comparable response to MG132 and CHX with respect to Gck expression (Supplementary Figure S1B), whereas human GCK mRNA under the control of the CMV (cytomegalovirus) promoter was solely significantly influenced by 12 h treatment with CHX (Supplementary Figure S1C), indicating that inhibition of the translation has only an effect on transgene expression.
Aggregation of endogenous glucokinase after proteasome inhibition
Glucokinase aggregation after inhibition of the proteasome was further analysed by fluorescence microscopy. MIN6 cells were treated with MG132 and/or CHX for 12 h and thereafter stained for glucokinase and the ProteoStat® dye specific for aggresomes . In MIN6 cells, glucokinase was homogenously distributed in the cytoplasm (Figure 3A). Inhibition of degradation by MG132 resulted in small aggregates which co-localize with the aggresome detection marker (Figure 3B). The aggregates were distributed in the cytoplasm with a somewhat higher density in the perinuclear region. Treatment with CHX had nearly no effect on glucokinase distribution (Figure 3C). Simultaneous inhibition of translation and degradation reduced cytoplasmic glucokinase aggregation (Figure 3D). Likewise, treatment with the β5 proteasome-subunit-specific inhibitor bortezomib evoked formation of glucokinase aggregates in the cell cytoplasm which co-localized with the aggresome detection marker. For quantification (Figure 3H), regions of interest were determined in the cytoplasm. Proteasome inhibition increased the range of pixel intensity significantly representing aggregation of glucokinase, whereas co-treatment with CHX could significantly diminish this aggregation. Aggregation of endogenous glucokinase could be confirmed in primary mouse β-cells (Figure 3F). Treatment for 12 h with MG132 resulted in glucokinase aggregates in the cytoplasm, which co-localized with the aggresome detection marker (Figure 3G).
Effect of MG132 and CHX on glucokinase in hepatocytes
Glucokinase expression in hepatocytes is significantly higher compared with pancreatic β-cells . After incubation at high glucose, glucokinase showed a homogenous distribution in mouse hepatocytes (Figure 4A). Inhibition of the proteasome evoked a strong aggregation of glucokinase in the cytoplasm with significant co-localization with the ProteoStat® dye (Figure 4B). Simultaneous inhibition of the translation could counteract this aggregation (Figure 4C). Proteasome inhibition resulted in an increase in glucokinase in relation to GAPDH expression in hepatocytes (Figure 4D), which could be partly diminished by co-treatment with CHX.
Interaction of glucokinase with ubiquitin
In vitro ubiquitination of glucokinase has been reported previously using a rabbit reticulocyte lysate system . In addition to wild-type conjugation-efficient EYFP–ubiquitin, a mutant conjugation-deficient EYFP–ubiquitinK0,G76V lacking all internal lysine residues and the C-terminal glycine residue  was used to investigate ubiquitination of glucokinase in vivo. EYFP–ubiquitin (Figure 5A) and EYFP–ubiquitinK0,G76V (Figure 5B) were visible in the cytoplasm and the nucleus in MIN6 cells. EYFP–ubiquitin fluorescence was highest in the nucleus. In the cytoplasm a punctate pattern was observed, partly co-localized with glucokinase. In contrast, the conjugation-deficient mutant showed a homogeneous expression pattern throughout the cytoplasm and the nucleus. For FRET analyses, ECFP–glucokinase and EYFP as a control, EYFP–ubiquitin wild-type or mutant were co-overexpressed in COS cells. In ECFP–glucokinase and EYFP–ubiquitin co-overexpressing cells FRETN was 4.5-fold higher than in the negative control (Figure 5C). In contrast, no interaction was detected with the EYFP–ubiquitinK0,G76V mutant (Figure 5C).
Effect of proteasome inhibition on mCherry–glucokinase in comparison with d2EGFP in MIN6 cells
The EGFP has a long half-life and high protein stability, and showed homogenous distribution in mammalian cells after transfection [27,28]. Peptide sequences which are rich in proline, glutamic acid, serine and threonine residues, also called PEST sequences, are known to serve as a signal for protein degradation. Subcloning of a PEST sequence to the C-terminus of EGFP resulted in d2EGFP, an EGFP variant with a short half-life [27,28]. MIN6 cells were transiently transfected with mCherry–glucokinase, a red fluorescence glucokinase fusion protein and d2EGFP and analysed over 9 h (Figure 6A). The horizontal intensity profile demonstrated distribution of both fluorescence proteins at 0, 3, 6 and 9 h (Figure 6B). MG132 induced aggregation of mCherry–glucokinase (Figure 6C), as observed for endogenous glucokinase in MIN6 cells (Figure 3). Aggregated mCherry–glucokinase near the cell nucleus accumulated over time, reflected by the strong increase in fluorescence at the aggregation point in the profile (Figure 6D). In contrast, despite the short half-life, d2EGFP showed a homogenous distribution without aggregation (Figure 6C). The intensity profile was comparable at 0, 3, 6 and 9 h (Figure 6D). Treatment of MIN6 cells with epoxomicin (Supplementary Figures S2A and S2B at http://www.biochemj.org/bj/456/bj4560173add.htm) and bortezomib (Supplementary Figures S2C and S2D) resulted in the same effect.
Aggregation of Dendra2–glucokinase in MIN6 cells
MIN6 cells, transiently transfected (Figures 7A–7C) or stably overexpressing a Dendra2 fusion protein with glucokinase (Figures 7D–7L) were incubated with MG132 alone or in combination with CHX. No differences were detected in the distribution of Dendra2–glucokinase compared with endogenous glucokinase (Figure 3), mCherry–glucokinase (Figure 6), and between transiently transfected or stably overexpressing cells (Figures 7A and 7D). The proteasome inhibitors MG132 (Figures 7B and 7E), lactacystin (Figure 7G), epoxomicin (Figures 7H–7L) and bortezomib (Supplementary Figures S3B and S3C at http://www.biochemj.org/bj/456/bj4560173add.htm) evoked aggregation of Dendra2–glucokinase in the cytoplasm with a higher localization around the nucleus, whereas co-treatment with CHX could counteract this aggregation (Figures 7C and 7F). In the 3D slice view the compactness of Dendra2–glucokinase aggregation is clearly visible (Figure 7I). Smaller aggregates arranged around a dense core, resulting in a growing globular aggregate. The homogeneous distribution of Dendra2–glucokinase (Figure 7J) changed to small (Figure 7K) and globular (Figure 7L) aggregates 6 and 8 h after treatment with epoxomicin respectively. The fluorescent protein Dendra2 was homogeneously distributed in MIN6 cells (Supplementary Figure S3D) and did not aggregate after 12 h treatment with MG132 (Supplementary Figure S3E) or bortezomib (Supplementary Figure S3F). Susceptibility of the tubulin network to proteasome inhibitors, because of hyperacetylation of α-tubulin, has been reported previously . Dendra2–α-tubulin showed a comparable filamentous structure in MIN6 cells (Supplementary Figure 3G) as the endogenous protein . After 12 h treatment with MG132 or bortezomib tubulin bundles and focal aggresomes were detectable (Supplementary Figures S3H and S3I).
Live-cell imaging of MIN6 Dendra2–glucokinase cells after treatment with MG132
Using the special feature of Dendra2, irreversible photoconversion , the effect of the proteasome inhibitor MG132 on folded (red Dendra2) and newly synthesized Dendra2–glucokinase (green Dendra2) in living cells can be analysed over time. MIN6 Dendra2–glucokinase cells were treated with MG132 and fluorescence images of living cells were taken every hour after initial irreversible photoconversion of green into red Dendra2 (Figure 8A). After 5 h of treatment small aggregates were visible in the cytoplasm with a somewhat higher concentration near the nucleus. After 9 h the small aggregates accumulated in the perinuclear region forming a dense globular aggregate. After 24 h a high amount of aggregated Dendra2–glucokinase was detectable around the nucleus. Nocodazole, an inhibitor of microtubule polymerization, prevented accumulation of the aggregates near the nucleus and after 12 h small aggregates were distributed in the whole cytoplasm (Figure 8B).
In pancreatic β-cells the rate of transcription and translation of the glucose sensor glucokinase appears to be very stable [38,39]. Regulation of glucokinase enzyme activity occurs mainly at the post-translational level  by interaction with the bifunctional enzyme phosphofructo-2-kinase/fructose-2,6-bisphosphatase [16,18,19,22], cytoplasmic compartmentation via binding to insulin granules mediated through S-nitrosylation [15,20], to mitochondria , and association with tubulin filaments . However, another aspect of post-translational regulation, namely glucokinase protein stability, remains an unexploited field in vivo. Unstructured regions and amino acids which are correlated with protein instability were mostly postulated from in vitro experiments. However, a global protein stability profiling, which has been performed in mammalian cells indicates the need for cellular approaches to investigate protein half-life regulation . In the present study we could demonstrate ubiquitination of glucokinase in vivo by FRET analyses, as previously demonstrated in vitro .
Thermal instability, structural changes and self-association of glucokinase mutations causing maturity-onset diabetes of the young have been shown [40–45]. We investigated aggregation of the wild-type glucokinase protein using the computer algorithm TANGO . This program predicted regions between Val181 and Leu185, and between Val302 and Asp311 regions with high β-sheet aggregation propensities. Additionally taking into account that the glucokinase protein has a high flexibility and can be present in at least five different conformations by undergoing extensive conformational changes [10,47,48], it becomes obvious that glucokinase has high susceptibility to protein misfolding and, thus, requires a sufficient control by the UPS.
Two decades ago the half-life of glucokinase in rat hepatocytes has been estimated to be 12.7 h . In the same study the half-life of lactate dehydrogenase was 17 h, and the authors concluded that non-autophagic mechanisms are involved in cytoplasmic enzyme turnover . Interestingly, the recently performed cellular proteome-scale protein-turnover analyses using the proteasome inhibitor MG132 suggested that both proteins are regulated by the proteasome and provided, in agreement with the results obtained by Kopitz et al. , a higher lifetime index for lactate dehydrogenase (gene ID 3945) than for glucokinase (4.80 compared with 4.34) . Treatment of MIN6 cells with the proteasome inhibitor MG132 for 3 h resulted in an increase in the glucokinase protein by 50% . Extending the incubation time to 12 h, which corresponds to the calculated half-life of the enzyme, we observed a comparable increase in glucokinase protein. As a control, GAPDH immunoreactivity was taken. The content of this long-living protein (lifetime index 6.24) remained unchanged. Glucokinase gene expression decreased, most probably as an adaptive process, after incubation with MG132, but notably did not increase and, thus, cannot account for causing the higher glucokinase protein content. However, despite the high protein content we observed a 50% decrease in glucokinase activity, indicating formation of inactive misfolded glucokinase in the absence of proteasome activity. In fact, incubation of MIN6 cells with MG132 and in addition with CHX, a compound inhibiting translation and increasing the availability of chaperones, restored glucokinase activity to a certain extent.
At the single-cell level we observed direct evidence for regulation of glucokinase by the UPS. Using a novel fluorescence-based assay  it could be demonstrated in MIN6 cells, primary mouse β-cells and hepatocytes that glucokinase is within aggresomes after inhibition of the proteasome and that this aggregation could be, at least in part, avoided by CHX through chaperone-mediated refolding. The same effect could be shown using different classes of proteasome inhibitors, namely the peptide aldehyde MG132, a reversible proteasome inhibitor, the peptide epoxyketone epoxomicin and the naturally occurring lactacystin, both irreversible proteasome inhibitors interacting with β subunits of the proteasome, and bortezomib, which has, with the β5 proteasome subunit, a specific target structure . Fluorescent fusion proteins of glucokinase were used in the present study to investigate the process of protein aggregation in living MIN6 cells over time. We are aware that the glucokinase protein content in MIN6 cells after overexpression was somewhat higher than in primary β-cells and in the range of that of hepatocytes. However, aggregation due to proteasome inhibition resulted from transiently and stably overexpressed glucokinase in MIN6 cells in a manner comparable with that of the endogenous protein.
The use of a fusion protein between glucokinase and the photoconvertible protein Dendra2  in a time series analysis, which allows discrimination between older and newly synthesized protein, argue in favour of microtubule-assisted aggresome formation in β-cells. At the beginning of proteasome inhibition, small glucokinase-containing aggresomes were distributed in the cytoplasm, whereas later on they formed bigger aggregates and were transported to a so-called MTOC (microtubule-organizing centre) [51,52], forming a ‘ribbon-like’ huge aggresome near the nucleus  (Figure 8C). Consistently, blocking both microtubule polymerization and proteasomal degradation prevented formation at the MTOC. Because unfolded polypeptides can attach to such MTOC-located aggresomes this is a dangerous knock-on process for the cell, which has been associated so far with neurodegenerative diseases, cancer and metabolic syndromes . Impaired clearance of misfolded and ubiquitinated proteins due to reduced proteasome activity has been recently reported in patients with Type 2 diabetes [7,8]. In agreement with previous studies [3,4] we demonstrated that insulin secretion is significantly reduced after inhibition of the proteasome. Despite its protective effect against MG132-induced glucokinase aggregation CHX could not restore insulin secretion, because it interrupts translation, and thus, biosynthesis of insulin and other key proteins of β-cells. An effect of MG132 on cell viability was detectable and has been described previously [3,4], but at the concentration of 10 μmol/l used the number of apoptotic cells was negligible compared with that of the cytotoxic quinoline alkaloid camptothecin, which had in the present study a comparable effect on insulin secretion. In addition, bortezomib, at a concentration of 10 nmol/l, caused a significant aggregation of glucokinase and a reduction in glucose-induced insulin secretion by 80%, but had virtually no effect on cell viability. This clearly indicates that beyond autophagy [2,54], the UPS is essential for regulating key factors of the stimulus/secretion cascade in pancreatic β-cells.
In conclusion, we demonstrated that glucokinase is regulated by the UPS to avoid misfolding and reduced activity of the enzyme. In addition, our fluorescence-based single-cell analyses suggest a microtubule-assisted aggresome formation of misfolded proteins in β-cells. This helps to understand the pathogenesis of Type 2 diabetes, where a reduced proteasome activity has been reported [7,8]. Inclusion of glucokinase in aggresomes in β-cells of patients with Type 2 diabetes could impair glucose recognition and, thus contribute, at least in part, to a reduced glucose-induced insulin secretion.
Anke Hofmeister-Brix performed experiments, analysed and interpreted data, and wrote the paper. Sigurd Lenzen discussed the data and revised the paper. Simon Baltrusch designed and supervised the study, performed experiments, analysed and interpreted data, and wrote the paper.
This work was supported by the Innovative Medicines Initiative Joint Undertaking [grant number 155005 (IMIDIA)], resources of which are composed of financial contribution from the European Union's Seventh Framework Programme [grant number FP7/2007-2013] and EFPIA (European Federation of Pharmaceutical Industries and Associations) companies.
We thank Jasmin Kresse and Rica Waterstradt for technical assistance. We also thank Dr Ortwin Naujok for support with real-time PCR analyses.
Abbreviations: CHX, cyclohexamide; Cy2, carbocyanine; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; MTOC, microtubule-organizing centre; NA, numerical aperture; qPCR, quantitative real-time PCR; UPS, ubiquitin–proteasome system
- © The Authors Journal compilation © 2013 Biochemical Society