Glucose-Regulated Gene Expression Maintaining the Glucose-Responsive State of β-Cells Free

In general, the acute minute-to-minute effects of glucose regulation of β-cells primarily depend on the allosteric or covalent modification of existing effector or regulator proteins. Examples of this type of regulation are the closure of ATP-sensitive potassium channels (K⁺_ATP channels) by a rise in the cellular [ATP]/[ADP] ratio (14) and GTP-induced activation of the small G-protein Rab-3a, which controls vesicle budding and exocytosis (15). On the other hand, it is widely assumed that the subacute (hours) and chronic (days) effects of glucose require, in addition to the acute regulatory effects, changes in the expression level of the very genes that encode proteins involved in the acute regulation of β-cell function. In line with the above examples, we have recently shown that 24-h exposure of rat β-cells to 10 mmol/l glucose increases mRNA encoding both subunits of the K⁺_ATP channels (SUR1 and Kir6.2) (16), whereas at the same time, mRNA encoding the Rab-geranylgeranyl transferase is reduced compared with cells cultured in 3 mmol/l glucose (D.F., V. Berger, M. Kruhoeffer, T. Orntoft, D.P., F.S., unpublished data). The number and identity of genes undergoing glucose regulation in β-cells is not yet known, but knowledge in this field is expanding rapidly (16–18; D.F., V. Berger, M. Kruhoeffer, T. Orntoft, D.P., F.S., unpublished data).

In pioneering studies, Collins et al. (19,20) analyzed via two-dimensional gel electrophoresis ³⁵S-methionine-labeled proteins extracted from whole islets that were exposed in vivo or in vitro to either low or high glucose. Using this technique, ∼2,000 different islet protein spots were identified. From this pool of islet proteins, ∼1.5% was reported to be regulated by glucose in the physiological concentration range (20). Assuming that the total number of proteins that are expressed in β-cells may be as high as 10,000, the number of glucose-regulated proteins may exceed a few hundred. The need to identify such large numbers of genes involved has shifted research efforts from the “classical” candidate gene approach to representational difference analysis (differential display [18]) or microarrays allowing gene expression analysis (16,17; D.F., V. Berger, M. Kruhoeffer, T. Orntoft, D.P., F.S., unpublished data). Although the candidate gene approach has given useful insights with respect to regulation of insulin (21,22), IAPP (23,24), PC1/3 (25), GLUT2 (26), GAD65 (27), and IA-2 (28) genes, it will not allow full comprehension of the scale and details of a glucose-regulated gene network. The prediction that glucose probably regulates hundreds of β-cell genes is borne out of a recent large-scale mRNA expression analysis. The possibility of undertaking such screening depended on high-density microarrays, allowing the measurement of thousands of different transcripts in one particular sample. Using the glucose-responsive murine β-cell line MIN6, Webb et al. (17) identified 75 transcripts from which abundance was regulated by altering the glucose concentration in the culture medium from 5.5 to 25 mmol/l. Following a similar approach in primary cultures of fluorescence-activated cell sorter (FACS)-purified rat β-cells, our laboratory recently identified more than 150 glucose-regulated transcripts for which abundance depended on the glucose concentration (3 vs. 10 mmol/l) during the 24-h tissue culture (16; D.F., V. Berger, M. Kruhoeffer, T. Orntoft, D.P., F.S., unpublished data). Messenger RNAs that were more abundant in G10 cells than in G3 cells (n = 76) and transcripts with a higher abundance in cells cultured at 3 mmol/l glucose (n = 103) could be classified into several functional groups (Fig. 1). The glucose-upregulated transcripts encode a number of proteins from which expression was previously reported to be glucose dependent in the candidate gene approach. Examples are both preproinsulins (21), the prohormone convertase PC1/3 (25), the glucose transporter GLUT2 (26), and the major β-cell autoantigens GAD65 (27) and IA-2 (28). The analysis of MIN6 (17) and the FACS-purified β-cell genes (16) was performed with Affymetrix mouse/rat A arrays that cover only part of all genes in these murine genomes. Consequently, it can be predicted that the number of glucose-regulated transcripts in β-cells may be a lot higher than currently reported.

The phenotype of differentiated β-cells is remarkable not only for its production, storage, and secretion of insulin, but also for its unique metabolic organization that may be key to the exquisite glucose responsiveness of the cells. Several aspects of glucose metabolism in the β-cells have been previously reviewed (29–32). Studies on FACS-purified rat β-cells have contributed to four particular aspects of metabolic organization in these cells: a low hexokinase-over-glucokinase ratio (33), little anaerobic glycolysis (34), a low [ATP]/[ADP] ratio at basal (5 mmol/l) glucose (14), and high rates of anaplerosis/cataplerosis (16,34). As mentioned above, the insulin secretory phenotype of rat β-cells to glucose can be maintained chronically in tissue culture if a glucose concentration of 10 mmol/l is present (7). Both lowering (6 mmol/l) and augmenting (20 mmol/l) the glucose concentration during the culture period resulted in a marked loss of the capacity of the cells to synthesize, store, and secrete insulin upon further acute glucose stimulation (7). β-Cells can also be dedifferentiated by chronic hyperglycemia in vivo (10), strongly indicating that this phenomenon is not an artifact of the in vitro serum-free culture system. Similar effects can be obtained by exposing β-cells for at least 2 h to either low (3 mmol/l) glucose (16) (Fig. 2) or high glucose in the presence of the protein synthesis inhibitor cycloheximide (35) (Fig. 2). From these results, it seems likely that the observed changes in the β phenotype (shift from a glucose-responsive to a glucose-unresponsive state) are linked to an altered presence of at least one critical (regulatory or effector) protein for which function is required for insulin secretion. To explore this possibility, we performed two-dimensional gel electrophoresis of ³⁵S-methionine-labeled proteins synthesized over 4 h at 10 mmol/l glucose in FACS-purified rat β-cells after 10 days of culture in 6, 10, or 20 mmol/l glucose (Fig. 3). In these conditions, the amount of total protein synthesized per β-cell increased from 7.6 ± 2.5 cpm/β-cell at 6 mmol/l glucose preculture to 13.9 ± 4.0 cpm/β-cell at 10 mmol/l glucose preculture and remained the same (12.7 ± 6.0 cpm/β-cell) after culture in 20 mmol/l glucose (mean ± SE; n = 3). Laser densitometric scanning of spot intensities could distinguish between several different patterns of glucose regulation. A first pattern, concentration-dependent upregulation by glucose (6–20 mmol/l), was obtained for the proinsulin 2 protein (marked “a” in Fig. 3). Intensity of the two combined proinsulin 2 spots (localized in two-dimensional gels of immunoprecipitated extracts; data not shown) was increased sixfold between 6 and 10 mmol/l glucose and another 2.7-fold between 10 and 20 mmol/l glucose culture. The second pattern exhibited an optimum glucose concentration (10 mmol/l) for translation and lower ³⁵S-incorporation in cells cultured at 6 or 20 mmol/l glucose. This pattern is exemplified by four 82-kDa proteins (labeled “b” in Fig. 3) from which synthesis was between 2.5- and 4-fold higher at 10 mmol/l glucose culture than in the other two culture conditions. This expression pattern is of potential interest because it coincides with the apparent optimal glucose level to maintain the differentiated state of primary β-cells both in tissue culture (7–9) and in vivo (10,11). A third pattern (illustrated by “c” in Fig. 3) shows suppression of translation when the culture glucose concentration increases. The experimental setup did not allow further mass spectrometric analysis. Nevertheless, these data underline the point (7) that the protein biosynthetic activity of β-cells depends on the glucose concentration to which the cells are chronically exposed. The shift in phenotype (7) seems to be the result of a change in the type of protein that is synthesized, which in turn depends on the mRNA profile of the cells.

In our analysis of 24-h glucose-regulated transcripts in FACS-purified rat β-cells (16), we observed that most of the 76 transcripts for which abundance was higher after culture in 10 mmol/l glucose encode proteins that are related to specialized β-cell function (Fig. 1). Examples are enzymes of glucose, lipid, and amino acid metabolism (23 transcripts) as well as proteins involved in regulated insulin synthesis and secretion (12 transcripts upregulated), signaling proteins (13 upregulated in 10 mmol/l glucose), and channels/transporters (8 upregulated). On the contrary, most of the 103 transcripts for which abundance was higher after culture in 3 mmol/l glucose encode proteins involved in general cellular maintenance, repair, protein synthesis/degradation, transcription, and RNA splicing. To further study the functional role of these glucose-regulated transcripts in rat β-cells, we recently concentrated on two gene clusters that appear involved in the triggering and amplifying pathway of glucose-induced insulin release (16). For the triggering pathway, we observed higher expression of GLUT2, suppression of inactivating kinase of the pyruvate dehydrogenase complex, and upregulation of several subunits of the respiratory chain (16). These changes correlated to higher rates of glucose oxidation. Moreover, transcripts encoding both subunits (SUR1 and Kir6.2) of the K⁺_ATP channels and the calcium sensor calmodulin were more abundant in cells cultured at 10 mmol/l glucose. In the same study, we reported that expression of multiple genes involved in cataplerosis, the export of mitochondrial metabolites into the cytosol for the production of lipid intermediates, was higher in cells cultured at 10 mmol/l glucose than at 3 mmol/l glucose (16). The identified transcripts encode the first common step, ATP-citrate lyase, several enzymes of fatty acid synthesis and desaturation, and cholesterol or isoprene biosynthesis. These changes correlated well with enhanced abundance of the transcription factor ADD1 (SREBP1c) (36) as well as with enhanced rates of incorporation of ¹⁴C-labeled carbon from glucose into hydrophobic molecules that could be extracted from the cells. The functional importance of the ATP-citrate lyase reaction for maintaining stimulated insulin release was documented using two pharmacological inhibitors, radicicol and (−)-hydroxycitrate, which abrogated part of the acute secretory capacity of the cells (16). These data support the idea (37) that cataplerosis is important to maintain the functional state of the β-cells.

In contrast to the well-defined triggering pathway of insulin release (1,38), the molecular targets and precise signaling molecules involved in cataplerosis remain poorly understood. They can vary from increased metabolic flux through various NADPH-generating cycles such as the citrate/pyruvate and citrate/malate cycles. It was previously estimated (34,39–41) that pyruvate carboxylation accounts for a high percentage of total pyruvate metabolism in β-cells. What fraction of carboxylated pyruvate serves to feed citrate/pyruvate and citrate/malate cycles may become clear by following a novel nuclear magnetic resonance (NMR) spectroscopic approach (41). Interestingly, β-cells cultured in 10 mmol/l glucose simultaneously upregulate expression of ATP-citrate lyase and downregulate expression of isocitrate dehydrogenase (16), an enzyme that would draw citrate substrate away from the two above-mentioned cycles. Alternatively, the cataplerotic flux through the ATP-citrate lyase reaction can serve primarily to generate building blocks for membrane biogenesis, which most likely will be enhanced under conditions of increased insulin biosynthesis, processing, and exocytosis. Balancing the relative contribution of fatty acid unsaturation and cholesterol production would affect significant functional membrane variables such as lipid raft density, membrane fluidity, or protein lipidation (42–44). The produced long-chain acyl-CoAs may also play a role in signaling as direct metabolic precursors of diacylglycerol, an activator of protein kinase C, or eicosanoids. Finally, it is conceivable that amplification effects are generated at the level of an integrated transcriptional response because glucose metabolism not only induces the synthesis of transcription factors that promote lipogenic gene expression (16), it generates (still unknown) ligands for various nuclear receptors (for example, the HNF family).

Relatively little is known about the molecular mechanisms that are involved in glucose regulation of β-cell gene expression. Over the past decades, insight into control of mammalian insulin gene expression has markedly progressed, both at the level of gene transcription (45) and at post-transcriptional steps (46); therefore, this information can be taken as a paradigm of glucose-regulated genes in β-cells. Unraveling the daunting complexity of the regulatory network of metabolites, signaling molecules, kinases/phosphatases, and transcription factors (Fig. 4) seems to be a formidable task for one gene and almost impossible for a whole network of interacting genes. At the transcriptional level, glucose metabolism can alter gene expression directly by (in)activation of proteins that are already bound to the preinitiation complex on the gene promoter. This may be achieved via kinases/phosphatases that sense the abundance of metabolic flux in the β-cells or the overall energy status of the cell. Some of the (in)activation pathways may depend on glucose-induced influx of calcium in the cells (which triggers activation of calcium-dependent kinases and phosphatases) or, even more indirectly, on glucose activation of insulin release and insulin interactions with insulin receptors on β-cells (Fig. 4). Changes in the concentration of the transcription factor protein in the nucleus can be obtained via translocation of existing molecules or glucose-induced changes in the balance between synthesis and degradation of transcription factors. A leverage effect can be obtained for transcription factors encoded by glucose-responsive genes because of the overall stimulatory effect of glucose on the translation rate of produced mRNA (3,4).

An interesting example of interacting kinases in glucose-stimulated gene expression is found in the L-type pyruvate kinase gene, which is nutrient responsive in β-cells (47) and in liver (48). The molecular mechanism that is involved has been studied in most detail in liver tissue (48–50). Transcription is stimulated by carbohydrate response element binding protein (48) via mechanisms involving both cAMP-mediated phosphorylation (49) and inactivation of AMP-activated protein kinase (50). It was suggested that glucose-dependent transcription of the L-type pyruvate kinase gene in MIN6 cells also depends on inactivation of AMP-activated protein kinase because the henomenon can be suppressed by the AMP analog 5-aminoimidazole-4-carboxamide (AICA)-riboside and mimicked by intracellular injection of antibodies directed against the AMP-activated protein kinase α-2 or β-2 chains (51).

An example of glucose regulation of gene expression via signaling pathways involving elevated intracellular calcium has recently been documented for the insulin gene (52). Synergism between glucose (nutrient stimulation) and cAMP (glucagon-like peptide 1 stimulation) at the level of insulin gene expression was reported to integrate at the level of nuclear factor of activated T-cells (NFAT). Studies with the calcineurin inhibitor FK506 and the calcium chelator BAPTA further suggest that the transcriptional response requires activation of calcineurin, a Ca²⁺-calmodulin-dependent protein phosphatase. A long signaling loop can be proposed in which high extracellular glucose elevates intracellular calcium, stimulates exocytosis, raises extracellular insulin, and causes insulin activation of receptors on β-cells (Fig. 4). It has been suggested (6) that such a signaling loop contributes in a rapid manner to the glucose-induced pool of preproinsulin mRNA in the β-cell. The fast-time kinetics of this process seem to some extent contradictory with earlier observations (53). This potentially interesting novel observation suggests that insulin action on β-cells contributes to the glucose-induced transcriptional response to glucose. This finding could explain why mice with β-cell-specific disruption of insulin receptor gene expression have defects in glucose-induced insulin release and diabetes (54).

Glucose not only alters transcription of many genes but also accelerates the translation of existing transcripts. This was first demonstrated for proinsulin mRNA (4,5,53). It is possible that glucose effects on overall rates of translation require direct intervention of metabolite(s), nutrient-derived signals or second messengers, or kinases/phosphatases at the level of the ribosome. The nature of these signals is yet unknown. The preferential translation of preproinsulin mRNA over that of other β-cell mRNAs (3) was proposed to depend on glucose-dependent relief of the signal peptide recognition particle (SRP)-mediated arrest of elongation (5). Alternatively, it may be linked to interactions with (still poorly characterized) proteins on the 3′ and 5′ untranslated regions of the transcript (55). Translational effects of glucose on β-cells could also interact with the long signaling loop that is outlined in Fig. 4, which requires the activation of insulin receptors, phosphatidylinositol-3 kinases, and protein kinase B. This pathway has been shown in other mammalian cell types to require the mammalian target of rapamycin (mTOR), which activates the initiation factor eIF-4E by phosphorylation of its binding protein 4E-BP1 (56). In line with this idea, rat islets exposed to glucose exhibit rapid phosphorylation of 4E-BP1, which could be mimicked by adding exogenous insulin and could be suppressed by rapamycin (57).

In summary, a new area of β-cell research is rapidly unfolding. The behavior of large groups of genes under different conditions of β-cell stimulation can be studied in vitro using relatively low amounts of RNA. So far, this approach has been used for the study of glucose effects on the mRNA profile of rodent β-cells. When applied to human β-cells from nondiabetic and diabetic pancreases, a novel strategy for drug discovery can be planned by identifying innovating targets for pharmacological intervention in β-cells.

This study was supported by grant 9.0130.99 from the Flemish Fund for Scientific Research (FWO Vlaanderen), the Ministerie van de Vlaamse Gemeenschap, Departement Onderwijs (Geconcerteerde Onderzoeksactie 1807), and the Belgian Program on Interuniversity Poles of Attraction, initiated by the Belgian State (IUAP-PAI).

The authors thank E. Quartier and V. Berger for valuable technical support.