Induction by glucose of genes coding for glycolytic enzymes in a pancreatic beta-cell line (INS-1).

Chronic elevation in glucose has pleiotropic effects on the pancreatic β-cell including a high rate of insulin secretion at low glucose, β-cell hypertrophy, and hyperplasia. These actions of glucose are expected to be associated with the modulation of the expression of a number of glucose-regulated genes that need to be identified. To further investigate the molecular mechanisms implicated in these adaptation processes to hyperglycemia, we have studied the regulation of genes encoding key glycolytic enzymes in the glucose-responsive β-cell line INS-1. Glucose (from 5 to 25 mM) induced phosphofructokinase-1 (PFK-1) isoform C, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (4-fold), and L-pyruvate kinase (L-PK) (7-fold) mRNAs. In contrast the expression level of the glucokinase (Gk) and 6-phosphofructo-2-kinase transcripts remained unchanged. Following a 3-day exposure to elevated glucose, a similar induction was observed at the protein level for PFK-1 (isoforms C, M, and L), GAPDH, and L-PK, whereas M-PK expression only increased slightly. The study of the mechanism of GAPDH induction indicated that glucose increased the transcriptional rate of the GAPDH gene but that both transcriptional and post transcriptional effects contributed to GAPDH mRNA accumulation. 2-Deoxyglucose did not mimic the inductive effect of glucose, suggesting that increased glucose metabolism is involved in GAPDH gene induction. These changes in glycolytic enzyme expression were associated with a 2-3-fold increase in insulin secretion at low (2-5 mM) glucose. The metabolic activity of the cells was also elevated, as indicated by the reduction of the artificial electron acceptor 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium. A marked deposition of glycogen, which was readily mobilized upon lowering of the ambient glucose, and increased DNA replication were also observed in cells exposed to elevated glucose. The results suggest that a coordinated induction of key glycolytic enzymes as well as massive glycogen deposition are implicated in the adaptation process of the β-cell to hyperglycemia to allow for chronically elevated glucose metabolism, which, in this particular fuel-sensitive cell, is linked to metabolic coupling factor production and cell activation.

The adaptation of the pancreatic ␤-cell to chronic hyperglycemia is characterized by an augmented secretory function, hypertrophy, and hyperplasia (1,2). These pleiotropic actions promoted by elevated glucose are expected to be associated with the induction of a number of glucose-regulated genes that must be identified. Elevated circulating glucose has gained recognition over the last few years as a factor contributing to ␤-cell dysfunction and the subsequent development of type 2 diabetes (2)(3)(4). However, the link between sustained hyperglycemia and long-term alterations in ␤-cell function is not well understood. In animal models, prolonged hyperglycemia causes both an impaired glucose-induced insulin secretion and the development of peripheral insulin resistance (5,6). These alterations can be reversed in part by lowering glucose to normal circulating levels, as occurs following treatment with the glycosuric agent phlorizin (7). These observations support the notion that high circulating glucose concentrations contribute to the development of pathologies associated with diabetes. Other factors in addition to glucose, such as elevated circulating free fatty acids, may also participate in the ␤-cell secretory defect (4 -6, 8).
Previous work has shown that an important component of the ␤-cell adaptation process to hyperglycemia is an increase in glucokinase activity (9,10), whereas the sugar does not modify the expression of the Gk 1 gene (8 -11). Nonetheless, the action of elevated glucose on the ␤-cell likely involves many additional proteins aside from Gk. Noteworthy are the results obtained in hexokinase (Hk) and Gk gene transfer experiments in the ␤-cell showing that other steps in glucose metabolism become ratelimiting after only modest increases in glucose-phosphorylating activity (8,12,13). This suggests that in vivo the activities of enzymes downstream of GK may not be as far in excess as previously thought from experiments using cell homogenates and particular assay conditions that do not match the cell situation.
One major step in linking hyperglycemia to long term phenotypic changes in pancreatic ␤-cells is the recognition of glucose as a main modulator of gene expression (5). Indeed, the sugar activates insulin gene transcription and pro-insulin mRNA translation (14 -17). Glucose also induces the accumulation of Glut-2 (18), L-pyruvate kinase (L-PK) (11), and acetyl-CoA carboxylase (ACC) (19) mRNAs in the ␤-cell. The 5Ј-CACGTG-3Ј motif, which has been found in the promotor region of a number of glucose-responsive genes, has been identified as the element conferring carbohydrate responsiveness to the L-PK and S-14 genes in hepatocytes (20,21). This motif is similar to the consensus sequence that binds the MLTF (major late transcription factor), a member of the c-myc family of transcription factors (20). Overexpression of c-myc in transgenic mice increases the hepatic expression of the L-PK gene (22). It remains to be determined which particular transcription factor(s) of the c-myc/MLTF family mediate(s) glucose induction of the L-PK and S-14 genes. On the other hand, Sp1 binding motifs found in the second promoter of the ACC gene, and not the MLTF/c-myc binding motif, may be responsible for glucose inducibility of the ACC gene in mouse preadipocytes (23). Thus, there appear to exist at least two transcription factors and distinct mechanisms by which glucose modulates metabolic enzyme gene expression in higher eukaryotic cells.
Mitochondrial oxidative events with oscillatory changes in the concentrations of ATP and ADP, in conjunction with accelerated anaplerotic input into the citric acid cycle and a rise in malonyl-CoA, are thought to be essential factors implicated in ␤-cell metabolic signaling (5,6,8). Increased glycolytic flux itself may also provide a signal mediating KATP channel closure (24) and a Ca 2ϩ rise independently of glucose-derived pyruvate metabolism in the mitochondria (6,8,25). There is experimental evidence indicating that accelerated NADH production by glyceraldehyde-3-phosphate dehydrogenase (GAPDH), linked to an effective glycerol 3-phosphate shuttle transferring cytosolic reducing equivalents to the mitochondria (26,27), mediates this action of glucose (25). Therefore, metabolic enzymes in the glycolytic, anaplerotic, and lipogenic pathways can be considered components of the transducing machinery, which links nutrient metabolism to ␤-cell activation. Since increased glucose metabolism in this particular fuel-sensitive cell is linked to signal transduction, it is thus attractive to hypothesize that sustained exposure to high glucose concentrations may alter the expression of genes coding for metabolic enzymes involved in glucose metabolism leading to changes in insulin secretion and ␤-cell growth. With respect to ␤-cell proliferation, it is noteworthy that glucose increases mitogen-activated protein kinase activity in INS-1 cells (28).
In order to test this hypothesis and identify the glucosemodulated genes in these pathways, we attempted to study the expression of genes coding for various glycolytic, anaplerotic, and lipogenic enzymes in INS-1 cells (29). In this paper, we show that three genes encoding key glycolytic enzymes are induced by glucose in ␤(INS-1) cells. These enzymes are phosphofructokinase-1 (PFK-1), which participates in the control of the glycolytic flux and metabolic oscillations (30); GAPDH, which provides cytosolic NADH which may act as a coupling factor (25); and L-PK, which catalyzes the formation of glycolytic-derived ATP. Changes in the expression level of these enzymes are associated with exaggerated glucose metabolism and insulin release at low concentrations of the sugar and an increase in ␤-cell proliferation.

EXPERIMENTAL PROCEDURES
Cell Culture and Incubation Conditions-INS-1 cells were seeded in 21-cm 2 Petri dishes (1.4 ϫ 10 6 cells/dish) and grown as described previously (29). When cells reached 80% confluence after approximately 7 days, they were washed twice with PBS (phosphate-buffered saline) at 37°C and preincubated for 48 h in culture medium containing 5 mM glucose. Cells were then washed with PBS and incubated in culture medium at various glucose concentrations for the indicated times.
Protein Analysis and Enzymatic Activities Measurements-Samples of 25 g of INS cell protein extracts, obtained in the presence of 50 mM Tris (pH 7.5), 5 mM EDTA, 5 mM EGTA, 0.5 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 10 g/ml aprotinin, 5 g/ml pepstatin, 5 g/ml antipain, and 10 mM mercaptoethanol, were resolved on 10% SDS-polyacrylamide gels and electrotransferred to nitrocellulose membranes (Bio-Rad). The membranes were incubated with a rabbit anti-rat antibody raised against the C-, M-, and L-PFK-1 isoforms (30). GAPDH was detected with an anti-rabbit monoclonal antibody (kindly provided by Dr. E. Knecht, Instituto de Investigaciones Biomédicas, Valencia, Spain). The L-PK and M-PK isoforms were detected with specific antibodies (kindly provided by Dr. J. Blair, University of West Virginia). The enzymatic activities of low K m Hk and Gk were measured in INS cells extracts as described previously (34).
In Vitro Transcription Assay-Nuclei isolation and nuclear run-on transcription assays were performed according to Ref. 35. Briefly, nascent transcripts were elongated in vitro in the presence of [ 32 P]UTP and 2.1 mg/ml heparin. The [ 32 P]RNAs were subjected to mild alkaline hydrolysis (30 min, 50°C, 50 mM Na 2 CO 3 ) and hybridized to 4 g/dot of the following DNA constructions immobilized on nitrocellulose membranes: a 0.74-kb EcoRI-BamHI fragment (positions 1040 -1780) of mouse rRNA 18 S cDNA subcloned in pUC830 and a 1.289-kb PstI-PstI fragment (positions 1-1289) of rat GAPDH cDNA subcloned in pBSKS.
Glucose Metabolism, Glycogen Determination, and Insulin Secretion-INS cells seeded in 24-well plates were used for these studies. Cells were preincubated for 48 h in culture medium at 5 mM glucose and incubated at 5 and 25 mM glucose for 3 days. Culture medium was removed and cells were washed twice with PBS and preincubated for 30 min at 37°C in Krebs-Ringer bicarbonate medium (KRB) containing 10 mM Hepes (pH 7.4), 0.07% bovine serum albumin, and 2.5 mM glucose. Cells were then washed twice with PBS and incubated for 30 min in KRB-Hepes medium containing 0.07% bovine serum albumin, 0.5 mg/ml MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide), and different glucose concentrations. Glucose metabolism was followed by the MTT reduction test, which closely reflects the rate of glucose oxidation, as described previously (36). Incubation media were collected to determine insulin release (37). The cellular glycogen content was determined according to a published procedure (38).
Cell Proliferation and DNA Content-INS-1 cells were seeded (0.8 ϫ 10 5 cells/well) in 24-well plates. When cells reached 50% confluence after approximately 4 days, they were washed twice with PBS and preincubated for 48 h in culture medium containing 5 mM glucose. Cells were washed again with PBS and incubated for 5 days in culture medium containing different glucose concentrations. Cells were then detached by trypsinization and resuspended in 0.8 ml/well RPMI. An aliquot of the cell suspension was diluted in 2 M NaCl, 0.05 M sodium phosphate buffer (pH 7.4), and DNA was measured as described (39). Cells were counted in a Coulter Counter.

RESULTS
Glucose Augments the Expression Level of the mRNA for Several Glycolytic Enzymes in ␤(INS-1) Cells-We elected to study the expression of various transcripts encoding key glycolytic enzymes, which are known to play essential regulatory role in ␤-cell signaling or glycolysis. These include the transporter Glut-2, whose ␤-cell content is decreased in diabetes (40); Gk, which acts as a glucose sensor (9); PFK-1, which is an important determinant of glycolytic flux and oscillations (30); 6-phosphofructo-2-kinase/fructose 2,6-bisphosphatase (PFK-2), which synthesizes fructose 2,6-bisphosphate, a regulator of glycolytic flux (41); GAPDH, which forms NADH which may directly or indirectly (via ATP production) act as a coupling factor (25); and pyruvate kinase, a highly regulated enzyme that converts cytosolic ADP to ATP.
Glut-2 mRNA increased by approximately 3-fold in INS-cells incubated for 24 h at 25 mM glucose in comparison to those cultured at 5 mM glucose (Fig. 1). This confirms findings made by others in pancreatic islets and various ␤-cell lines (18,(42)(43)(44). The Glut-2 probe detected only one abundant transcript of 2.8 kb in INS cells, in contrast with purified ␤-cells, which expressed two mRNA species of 3.9 and 2.8 kb (45). The biological significance of the larger transcript is unknown.
Using a poly(A ϩ ) mRNA fraction, it was observed that the "platelet" PFK-1-C transcript of 3.2 kb was induced by elevated glucose but was not detectable at basal (5 mM) glucose ( Fig. 1). The "liver" type PFK-1-L and "muscle" type PFK-1-M transcripts were undetectable under our experimental conditions using the Northern blot technique. However, the corresponding proteins were detected using specific antibodies (see below). The GAPDH probe recognized one 1.3-kb transcript in INS cells, which was consistently induced (about 4 -5-fold) by incubating INS cells at high glucose (Fig. 1). The L-PK probe has been shown to recognize three transcripts of 3.1, 2.2, and 2.0 kb in cultured hepatocytes and INS-1 cells (11,46). The low molecular weight species are indistinguishable in our electrophoresis conditions. Fig. 1 shows that the different L-PK transcripts are markedly induced by high glucose, confirming the results of a previous study (11). The effect of glucose on the expression level of these inducible transcripts was dose-dependent (data not shown).
As previously reported (11), the Gk mRNA level did not vary in response to glucose. The expression of the transcript of glucose-6-phosphate dehydrogenase also remained similar at low and high glucose (data not shown). PFK-2 mRNA content barely varied in cells cultured at high glucose. The slight apparent decrease of the PFK-2 signal at high glucose is due to a difference in sample loading of the gel in this experiment (data not shown). The cDNA probe used to detect the PFK-2 tran-script recognizes both the L-and M-isoform mRNA species. It is most likely that only the M-isoform is present in INS cells as well as in pancreatic islets. Indeed, reverse transcriptase-polymerase chain reaction experiments revealed that the mRNA of the L-isoform is expressed neither in INS cells nor in islets, while the mRNA for the M-isoform is present in both INS cells and islet tissue. 2 The 18 S ribosomal mRNA content, which was used as a control for gel loading, remained unaffected by glucose ( Fig. 1).
Our results indicate that glucose causes an induction of several genes coding for key glycolytic enzymes in INS-1 cells. The effect of glucose is selective, because some genes are induced, while others are not under the same experimental conditions.
The time dependence of the accumulation of the Glut-2, GAPDH and L-PK transcript is depicted in Fig. 2. The lag time of both Glut-2 and L-PK mRNA induction was approximately 2 h, and a maximal effect occurred at about 6 h of culture at high glucose. In contrast, the onset of GAPDH induction was much slower. Increased expression of GAPDH mRNA in response to elevated glucose required at least 6 h, and a maximal effect was observed at 24 h. We previously reported that ACC mRNA induction by high glucose in INS cells requires about 3 h (19). The reason for which GAPDH mRNA induction displays a relatively long lag time is not known. It possibly reflects a distinct mechanism of gene induction by glucose (see below).
Glucose Increases the Expression Level of GAPDH, L-PK, and Three PFK-1 Isoforms-Experiments were carried out to determine whether the accumulation of the inducible mRNAs is associated with a similar increased expression of the corresponding glycolytic enzymes. Since many glycolytic enzymes such as GAPDH have a long half-life (2-3 days), the expression of the glycolytic enzymes was measured after 3-4 days of incubation at either low (5 mM) or high (25 mM) glucose.
To assess the expression of PFK-1, a polyclonal antibody which recognizes the three PFK-1 isoforms was used. Fig. 3 shows that the C-, M-, and L-isoforms of PFK-1 were all induced by high glucose. The presence of the C-PFK-1 (86.5 kDa) and M-PFK-1 (82.5 kDa) isoforms has been reported in rat islets and INS cells (30). Both PFK-1 subtypes were induced in parallel and by approximately the same extent when INS cells were incubated at the elevated glucose concentration (Fig. 3). The L-PFK-1 isoform is poorly expressed in pancreatic islets (30)  The figure shows the mean values (n ϭ 2 for Glut-2, n ϭ 6 for GAPDH, and n ϭ 2 for L-PK) of densitometric quantifications of the autoradiograms for the different transcripts normalized to those of the 18 S rRNA signal, which did not vary under our experimental conditions. (Fig. 3). Interestingly, the L-isoform (76.7 kDa) of PFK-1 was expressed following a 3-day exposure of INS cells to high glucose. The cellular GAPDH content was only marginally increased after 48 h of cell exposure to 25 mM glucose (data not shown). We therefore studied its expression at 120 h. It is apparent that the amount of the GAPDH protein rose to an extent similar to that for its corresponding transcript in response to high glucose (Fig. 3).
A thorough study of L-PK transcriptional activation by glucose in INS cells has previously been carried out (11). However, the expression of the L-PK enzymes in response to glucose was not investigated. We therefore studied the expression of both the L-and M-isoforms of the enzyme in INS cells incubated for 72 or 120 h at 5 or 25 mM glucose. Immunoblot analysis of L-PK using a polyclonal antibody revealed the presence of two proteins of 55-60 kDa. Both were induced by elevated glucose (Fig. 4A). This result was confirmed using a monoclonal anti-L-PK antibody (Fig. 4B). The specificity of the monoclonal antibody was assessed by showing that it detected a major specific band in liver, whereas it detected no L-PK protein in smooth muscle where this isoform is not expressed (Fig. 4B). The two protein bands detected by the L-PK antibodies might be different phosphorylated forms of the enzyme, spliced variants, or translation products from different transcripts. The results in Fig. 4 also indicate that glucose increases the expression level of the L-isoform of PK. Thus, M-PK, which is also expressed in INS cells and is the most abundant isoform in islets tissue (47), was only slightly induced by the sugar (Fig.  4B). It should be pointed out that the single immunoreactive protein band detected by the M-PK antibody is not due to a cross-reactivity of the antibody with L-PK for the following two reasons. First, the M-PK antibody recognized a single abundant protein band in muscle and revealed two weak signals in liver tissue. Second, although M-PK has a molecular weight indistinguishable from the higher molecular weight form of L-PK, its marked variation of expression at high versus low glucose revealed with both L-PK antibodies was not detected with the M-PK antibody. The results in Fig. 3 are furthermore in agreement with the absence of 5Ј-CACGTG-3Ј or 5Ј-CACGGG-3Ј carbohydrate response elements in the promoter of the rat M-PK gene as compared to the L-PK gene.
In accord with results obtained with rat islets (48) Hence, the glucose induction of the PFK-1, GAPDH, and L-PK transcripts is associated with similar changes at the protein level, which possibly contribute to the late phenotypic alterations in INS cells exposed to an elevated glucose concentration.
Transcriptional Activation by Glucose of Genes Encoding Glycolytic Enzymes-An increased transcriptional rate of the Glut-2 (18) and L-PK (11) genes has been documented in INS cells challenged with high glucose. We sought to determine whether a transcriptional effect is also implicated in the increased expression of GAPDH. Run-on transcriptional assays were initially carried out following a 2-or 4-h incubation period at either 5 or 25 mM glucose. In three separate experiments, a similar transcriptional activity of the GAPDH gene at the two glucose concentrations was measured (data not shown). However, when the assays were carried out at 8 or 24 h with elevated glucose, a 1.7-fold increased transcriptional rate was observed in response to high glucose (Fig. 5). Thus, the enhanced transcriptional activity of the GAPDH gene is delayed in comparison to the Glut-2 (18) and L-PK (11) genes and accounts for about half of the accumulation of the GAPDH mRNA or protein in response to elevated glucose. These results suggest that additional mechanism(s) are implicated in the glucose regulation of GAPDH expression. These may include altered mRNA stability or differential processing of the transcript.
We addressed the first possibility by stimulating INS cells at 25 mM glucose for 24 h and incubating them afterward for 1-8 h in the presence of high (25 mM) or low (5 mM) glucose in the presence of the transcription inhibitor actinomycin D. The results indicated that the GAPDH transcript has a long half-life, since its content was reduced by only 20% following an 8-h incubation period in the presence of the inhibitor. Longer times were not tested because the drug affected cell viability beyond 8 h. The results obtained showed no apparent difference in the GAPDH mRNA decay at low versus high glucose (data not shown). Nonetheless, these results do not entirely discount a possible action of glucose in modulating GAPDH mRNA stability. Thus, due to the limitation imposed by the cell toxicity of actinomycin D, longer incubation periods could not be tested to rigorously assess this issue. It is noteworthy that the transcriptional induction of the Glut-2 gene also does not fully account for the glucose increased expression of Glut-2 mRNA (18). Whether glucose modulates the stability of transcripts encoding glycolytic enzymes remains to be determined.
Mechanism of Glucose Regulation of GAPDH Gene Expression-Previous work documented that the non-metabolizable glucose analog 2-deoxyglucose (2-DOG) induces the L-PK (11) and ACC (19) genes in INS cells. This suggested that the metabolism of the sugar beyond the Gk step is not required for the induction of some genes encoding metabolic enzymes (11,19,20). To gain insight into the mechanism of GAPDH induction, we tested several glucose analogs as well as nutrient stimuli and pharmacological agents, which activate various signaling pathways and insulin secretion (Fig. 6). Mannose, the epimer in position 2 of glucose, which is well metabolized in the ␤-cell (6), increased the expression level of GAPDH mRNA. The effect of glucose was suppressed by the Gk inhibitor mannoheptulose and was not mimicked by 2-DOG, which is phosphorylated but is not metabolized beyond this step (49). 6-Deoxyglucose and 3-O-methylglucose, two analogs that enter mammalian cells but cannot be phosphorylated, were ineffective. High concentrations of pyruvate and glutamine plus leucine, which efficiently promote insulin secretion (6, 50), did not alter the expression level of the GAPDH transcript. Finally, elevated K ϩ , phorbol 12-myristate 13-acetate, and forskolin, which are good secretagogues and, respectively, activate the Ca 2ϩ , protein kinase C, and cAMP signaling systems (51), barely modified the GAPDH mRNA content (Fig. 6). 10 Ϫ7 M insulin (autocrine control) and 30 mM sucrose (osmolarity control) had no effect on the expression level of Glut-2 and GAPDH mRNAs (data not shown).
The results demonstrate that glucose needs to be metabolized beyond the glucokinase step to induce the GAPDH gene and that the Ca 2ϩ , cAMP, and protein kinase C transduction systems are not implicated in this process.

MTT Reduction, Glycogen Deposition, Glucose-induced Insulin Secretion, and DNA Synthesis in INS Cells Exposed during
Culture to High Glucose Concentrations-To examine whether the glucose-induced changes in the expression level of key glycolytic enzymes in ␤(INS-1) cells were associated with similar phenotypic changes already observed in pancreatic islets (1,2,8,9,45,51,52), which might in part result from these modifications, we measured several parameters of ␤-cell activation.
The intracellular reduction of the tetrazolium dye MTT allows determination of the metabolic activity of the ␤-cell (36). The assay measures the reduction of the tetrazolium salt into insoluble colored formazan crystals. A very good correlation has been shown between MTT reduction, glycolytic flux, glu-cose oxidation, and insulin secretion in INS cells (26,36). MTT is reduced both in the cytosol and the mitochondria of living cells (26). It provides a suitable index of overall (glycolytic plus mitochondrial) metabolism via reducing equivalent production, simultaneously with secretion measurements (36). INS cells were cultured for 3 days at 5 or 25 mM glucose and then incubated for 30 min in the presence of MTT at various glucose concentrations. In cells preincubated for 3 days at low glucose, a subsequent variation of the sugar from 2.5 to 25 mM elicited a dose-dependent increase in MTT reduction, reaching a maximal value at 15 mM (Fig. 7, upper panel). Cells incubated for 3 days at 25 mM glucose displayed a 3-fold higher MTT reduction at low (2.5-5 mM) glucose and a 50% increased reduction of the dye at maximal (15-20 mM) concentrations of the carbohydrate. The data indicate that the basal metabolic activity of INS cells previously challenged with high glucose is 3 times that of cells exposed to a low physiological concentration of the sugar.
In the same set of experiments, we performed insulin secretion measurements in parallel (Fig. 7, lower panel). The patterns found for insulin secretion were very similar to those observed for MTT reduction. Cells cultured at low glucose subsequently exhibited a dose-dependent increase in secretion of 6-fold from 2.5 mM to 25 mM glucose. However, cells preincubated at 25 mM glucose for 3 days showed an increased basal insulin secretion and an absence of response to glucose between 5 and 25 mM glucose.
Glucose increases the glucose 6-phosphate content of rat islets (8,9) and INS cells (19). This glycolytic intermediate can also enter the glycogen biosynthetic pathway leading to intracellular glycogen accumulation, whose exaggerated deposition has been postulated to be toxic for the ␤-cell (53). Fig. 8 (left  panel) shows the dose dependence of glycogen deposition as a function of the glucose concentration. A Ͼ20-fold increased accumulation of glycogen in cells cultured at high (20 mM) versus low (5 mM) glucose was observed. The marked glycogen deposition could act as a glucose store, which, upon lowering of external glucose, could be metabolized to maintain elevated glucose metabolism and insulin secretion at basal concentrations of the sugar. To test this hypothesis, we measured the glycogen content of INS cells as a function of time following a lowering of glucose from 25 mM (for 3 days) to 5 mM glucose. The results in Fig. 8 (right panel) indicate that 60% of the accumulated glycogen was metabolized in a 6-h time period.
Since glucose acts as a growth factor for islet tissue and stimulates the proliferation of normal ␤-cells (1, 52, 54), we determined whether similar action of the sugar occurs in ␤(INS) cells. Glucose promoted a dose-dependent increase in DNA synthesis. The following values were observed after 5 days exposure to various glucose concentrations: glucose (5 mM), 4.7 Ϯ 0.8 g of DNA/well; glucose (11 mM), 8.7 Ϯ 0.4 g of DNA/well; glucose (25 mM), 14.8 Ϯ 0.8 g of DNA/well (means Ϯ S.E. of four wells). Cell proliferation measurements provided identical results (data not shown) and have been reported previously (54). DISCUSSION To address the question of the molecular nature of the ␤-cell adaptation to hyperglycemia, we used the ␤-cell line INS (29). As shown previously (19) and documented in greater detail in this study, the secretory properties of INS cells incubated for a long period of time (3 days) in the presence of elevated (25 mM) glucose concentrations display two characteristic features: a markedly elevated secretion at low (2-5 mM) glucose and a lack of response to higher concentrations of the sugar. This sensitization to glucose is similar to that observed for rat islets in vivo and in vitro (2,9,55) and human islets in culture (56). In addition, glucose within its physiological range of concentrations promotes the proliferation of ␤(INS) cells as well as normal ␤-cells (1). Finally, INS cells incubated for long periods of time at high glucose accumulate massive amounts of glycogen, like islet tissue (53,57). As far as metabolic gene expression is concerned, like in rat islets (11,18), glucose induces Glut-2 but not Gk mRNA in INS cells. In addition, the C-and M-isoforms of PFK-1 (30), the M-isoform of PFK-2, and the M-isoform of PK (47) are well expressed in both pancreatic islets and INS cells (this study). Thus, INS cells appear to be a good model to address the above mentioned questions and to carry on a detailed study of metabolic enzyme expression in a system where only ␤-cells are studied. Indeed, normal islet tissue contains 40% non-␤-cells, which may be very different as far as metabolism is concerned. For instance, studies in which sorted islet cells were studied have indicated that FAD-linked glycerol phosphate dehydrogenase is elevated in the ␤-cell, whereas lactate dehydrogenase is very poorly expressed, and the same is true for INS cells (26). In contrast, the reverse expression of these enzymes occurs in non-␤-cells (26).
The present study shows that long term exposure of INS cells to elevated glucose causes a coordinated induction of several glycolytic enzymes, which may play a key role in the coupling process from nutrient metabolism to insulin release and other biological functions including insulin biosynthesis and ␤-cell replication (1,5,8,14,17). These enzymes are PFK-1, GAPDH, and L-PK. The increase of Glut-2, PFK-1, GAPDH, and L-PK is associated with an exaggerated metabolic activity of INS cells, increased secretion at low glucose, and enhanced cell proliferation. Consistent with these observations, fasting of rats for 72-120 h caused a 20% reduction in the activity of pancreatic islet Gk (58), PFK-1 (58,59), and GAPDH (58), and a decrease in the expression level of L-PK mRNA (11).
The question arises as to how the augmented expression level of each of these particular proteins might relate to the observed phenotypic changes. Concerning PFK-1, the M-isoform of this enzyme plays a key role in glycolytic oscillations (5,6,30), which are thought to be implicated in the oscillatory nature of ␤-cell metabolism (60) and insulin secretion (5,6). The expression level of the M-and C-isoforms are increased by glucose as is the L-isoform, which cannot be detected at low concentrations of the sugar. The different isoforms may form heterotetramers displaying intermediate kinetic properties (61,62). It is attractive to hypothesize that the induction of the L-isoform favors the formation of new tetramers, which can lead to a different pattern of glycolytic oscillations resulting in impaired insulin secretion. It should also be noted that mathematical modeling of ␤-cell glycolysis indicates that the meta- bolic flux control coefficient of Gk and PFK-1 are similar at low (around 4 mM) glucose (9). Thus, an increased expression of PFK-1 may be required for increased ␤-cell growth and for high glycolytic flux and secretion as well as to match the augmentation of Gk activity in both rat islets (48) and INS cells (present study), which occurs with long exposures to elevated glucose (9,55). It should be noted that an increase in the quantity and activity (V max ) of a highly regulated allosteric enzyme like PFK-1 will influence glycolytic flux only when the enzyme is activated as it presumably is during metabolic oscillations, since at basal metabolite levels and ATP/ADP ratios, the enzyme is largely inhibited. Thus, PFK-1 induction might possibly permit larger glycolytic oscillations. Another potential role for some PFK-1 isoforms is suggested by a recent report documenting an association between a glucose and ATP-stimulatable Ca 2ϩ -independent phospholipase A 2 in the ␤-cell with a phosphofructokinase-like protein closely related or identical to M-isoenzyme (63).
The glucose induction of GAPDH is of particular interest in view of a novel hypothesis, proposing that a glycolytic signal which occurs at a step catalyzed by this enzyme (8) is responsible for KATP channel closure by glucose. Dukes and coworkers (25) have provided evidence in support of the notion that NADH, generated at the GAPDH step, fuels mitochondrial ATP production via the glycerol-phosphate shuttle (27). This event might be causally linked to the action of glucose on the KATP channel and Ca 2ϩ influx (8,25). Also consistent with the idea that GAPDH deserves much more attention in ␤-cell signaling than previously suspected is the finding that of all glucose metabolites, glyceraldehyde 3-phosphate is the most potent insulin secretagogue (64). With respect to insulin exocytosis noteworthy is the identification of a fusogenic role of a brain isoform of GAPDH, which catalyzes an extremely rapid fusion of phospholipid vesicles (65). The concept is also emerging that some metabolic enzymes have important cellular functions in addition to their role as enzymes. In this regard, numerous non-glycolytic activities have been attributed to GAPDH, including interaction with the cytoskeleton, nucleic acid binding, and DNA repair (66 -69). It remains to be proven whether any of these alternative functions ascribed to GAPDH, which might be implicated in the late phenotypic changes caused by elevated glucose, are operative in the ␤-cell. A previous report showed that the GAPDH activity of pancreatic islets in culture remains constant after a 24-h exposure to high glucose (70). The results of this study are consistent with this previous observation, since we also observed no change in GAPDH activity after a 24-h period but only after 3 days of incubation at elevated glucose.
Pancreatic islets contain the M-isoenzyme of pyruvate kinase (47) and express the L-type PK gene at a low level (11,71). L-PK/T antigen transgenic mice fed a high carbohydrate diet were shown to frequently develop endocrine pancreatic tumors (72), and reverse transcriptase-polymerase chain reaction experiments indicated that glucose intake increased L-PK mRNA in islet tissue (11). The results of this study are in accordance with these observations and furthermore show that the Misoenzyme is constitutively expressed. By contrast, glucose markedly induce the accumulation of the two protein isoforms L and LЈ encoded by the L-PK gene. Altogether, these observations indicate that L-PK protein expression is induced by elevated glucose in the ␤(INS)-cell. The significance of this inductive process is uncertain. As stated above for PFK-1, increased L-PK expression may be required for the accelerated glycolytic flux and the resulting increased secretion and ␤-cell growth because the expression of the M-isoenzyme is not induced by the sugar. In the liver L-PK is a highly regulated enzyme whose activity is changed by its phosphorylation state (72). The Lisoenzyme is allosterically activated by fructose bisphosphate, whereas the muscle isoenzyme is not (73). Whether L-PK is also regulated similarly in the ␤-cell is not known. L-PK provides glycolytic ATP. Perhaps the enzyme activity might cause local ATP gradients in the cytoplasm, in particular in the vicinity of the plasma membrane to influence more directly the open state probability of KATP channels or the activity of phospholipase A 2 (ATP-stimulatable), which liberates free arachidonic acid, a potent stimulator of insulin secretion (74).
It must be pointed out that the action of the sugar is specific for some genes coding for glycolytic enzymes, as not all the induced genes are activated to the same extent and with the same kinetics. The data also support the notion that there may exist several mechanisms implicated in the action of glucose on glycolytic enzyme gene expression in the ␤-cell. A transcriptional activation is involved in the increased expression of the Glut-2 (18), GAPDH, and L-PK (11) genes, although additional mechanisms such as alterations in transcript stability likely contribute to the increased Glut-2 (18) and GAPDH mRNA accumulation at high glucose. Moreover, glucose 6-phosphate (19,75) or an intermediate of the pentose pathway (76) may mediate gene activation for L-PK because 2-deoxyglucose is also effective (11). In contrast, GAPDH induction seems to be dependent of the metabolism of glucose because of the inability of 2-deoxyglucose to mimic the action of glucose.
The induction of key glycolytic enzymes is associated with a marked increase in metabolic activity of INS cells. The MTT reduction experiments show an elevated ␤-cell oxidative metabolism at all tested glucose concentrations in cells exposed to elevated glucose for 3 days. A parallel pattern is observed for insulin secretion. The massive deposition of glycogen may contribute to the hypersecretion of insulin at low glucose by providing glucose through glycogenolysis at low external concentrations of the sugar. Thus, since glucose-6-phosphatase is barely expressed in rat islets (9) and INS cells, 3 the vast majority of the glucose mobilized from glycogen should enter the glycolytic pathway. If it can be extended to the in vivo situation, these changes in glycogen metabolism and ␤-cell glycolysis could be related to the hyperinsulinemia developed in the first stages of non-insulin-dependent diabetes mellitus. Increased glycolytic enzyme gene expression and metabolic activity may also be implicated in the stimulation of ␤-cell proliferation. Indeed, glucose is a major growth factor in this particular fuel-sensitive cell both in vivo and in vitro (1,2). Elevated glycolytic flux should lead to the enhanced production of various signaling molecules and activate transduction pathways implicated in the regulation of cell growth such as the protein kinase C, mitogen-activated protein kinase, and Ca 2ϩ signaling systems. Interestingly, glucose activates mitogenactivated protein kinases in INS cells and induces a number of immediate early response genes implicated in cell growth regulation (28).
An important issue is to know whether the action of glucose on the expression of genes encoding metabolic enzymes is the consequence of glucose-directed gene regulation, increased metabolism of the sugar, or its mitogenic action. With respect to the L-PK and ACC genes, a "direct" action of the sugar independent of accelerated glycolytic flux is likely. Thus, the nonmetabolizable analog 2-DOG, which has no mitogenic action, induces the L-PK (11) and ACC (19) genes. Consensus carbohydrate response elements are present in the promoter region of the L-PK gene (20). With respect to ACC, two short sequences that resemble the consensus sequence are also present within its promoter (19). It is unclear at present whether these sequences contribute to the glucose regulation of the ACC gene in the ␤-cell. Concerning Glut-2, requirement of glucose metabolism has been documented in both liver (77) and rat islets (78) to regulate the Glut-2 gene. The induction of the GAPDH gene by glucose is relatively delayed with respect to the other investigated genes, it is not mimicked by 2-DOG, and a consensus response element is not found in its promoter. We therefore favor the view that GAPDH gene induction by glucose is the consequence of accelerated glucose metabolism and possibly early mitogenic events cause by the sugar. Nonetheless, these inductive processes, whatever the diversity of the mechanisms involved, pinpoint candidate genes and proteins that might be crucially involved for the normal adaptation response to elevated glucose (i.e. hypersecretion, hypertrophy, and hyperplasia of the ␤-cell). Possibly a defect in glucose regulation, either direct or indirect, of one of these genes could contribute to the pathogenesis of type 2 diabetes where the adaptation of the ␤-cell to hyperglycemia has failed. Finally, glucose is not the only calorigenic nutrient that may participate in the phenotypic alterations of insulin secreting cells during diabetic pathologies. In addition to glucose, other circulating factors, such as elevated fat, appear to play an instrumental role (5, 6, 8, 79 -81). The mechanism whereby each individual calorigenic nutrient or combination of them participate in the etiology of diabetes remains largely an open question.