Activation of the Hexosamine Pathway Leads to Deterioration of Pancreatic (cid:1) -Cell Function through the Induction of Oxidative Stress*

It is known well that activation of the hexosamine pathway causes insulin resistance, but how this activation influences pancreatic (cid:1) -cell function remains un-clear. In this study, we found that in isolated rat islets adenovirus-mediated overexpression of glutamine:fructose-6-phosphate amidotransferase (GFAT), the first and rate-limiting enzyme of the hexosamine pathway, leads to deterioration of (cid:1) -cell function, which is similar to that found in diabetes. Overexpression of GFAT or treatment with glucosamine results in impaired glucose-stimulated insulin secretion and reduction in the expression levels of several (cid:1) -cell specific genes (insulin, GLUT2, and glucokinase). Additionally, the DNA binding activity of PDX-1, an important transcription factor for these three genes, was markedly reduced. These phenomena were not mimicked by the induction of O -linked glycosylation with an inhibitor of O -GlcNAcase, PUGNAc. It was also found that glucosamine increases hydrogen peroxide levels and that several hexosamine pathway-mediated changes were suppressed by treatment with the antioxidant N -acetyl- L -cysteine. In con- clusion, activation of the hexosamine pathway leads to deterioration of (cid:1) -cell function through the induction of oxidative stress rather than O -linked glycosylation. Thus,

The development of type 2 diabetes is usually associated with a combination of pancreatic ␤-cell dysfunction and insulin resistance. Normal ␤-cells can compensate for insulin resistance by increasing insulin secretion or ␤-cell mass, but insufficient compensation leads to the onset of glucose intolerance. Once hyperglycemia becomes apparent, ␤-cell function deteriorates (1). The adverse effects of chronic hyperglycemia on ␤-cells, called glucose toxicity, have been demonstrated by various in vitro and in vivo studies (2)(3)(4)(5)(6). After chronic exposure to hyperglycemia, insulin gene transcription and glucose-stimu-lated insulin secretion are suppressed. This is often accompanied by a decrease in expression and DNA binding activity of the pancreatic and duodenal homeobox factor-1 (PDX-1) 1 (4, 6 -8). PDX-1 (also known as IDX-1/STF-1/IPF1) (9 -11) is a member of the homeodomain family transcription factors and plays a major role in maintaining normal ␤-cell function by regulating multiple genes, including insulin, GLUT2, and glucokinase (12)(13)(14)(15).
Glucose metabolism through the hexosamine pathway has been implicated in many of the adverse effects of chronic hyperglycemia. In the hexosamine pathway, fructose-6-phosphate is converted to N-acetylglucosamine-6-phosphate by glutamine:fructose-6-phosphate amidotransferase (GFAT). N-Acetylglucosamine-6-phosphate is then converted to N-acetylglucosamine-1,6-phosphate and to UDP-GlcNAc. UDP-Glc-NAc is a substrate for O-linked glycosylation, which is catalyzed by O-GlcNAc transferase. The O-linked glycosylation is reversed by O-GlcNAc ␤-N-acetylglucosaminidase (O-GlcNAcase). The relationship between the hexosamine pathway and subsequent O-GlcNAc modification has been examined extensively (16 -18). Indeed, it has been reported that many proteins including some transcription factors are modified with O-Glc-NAc and that this modification can activate gene transcription (17)(18)(19)(20). Several studies have suggested that the hexosamine pathway is involved in insulin resistance (21)(22)(23)(24)(25). It has been reported that GFAT overexpression in muscle and fat (22) or liver (25) results in insulin resistance. There is also evidence that the O-GlcNAc modification of key transcription factors might contribute to insulin resistance (24). Recently, it has been reported that O-GlcNAc transferase is highly expressed in ␤-cells (26,27), implicating potential involvement of the hexosamine pathway on ␤-cell function. Moreover, it was shown that streptozotocin, a diabetogenic reagent, elevates O-GlcNAc levels in islets, implying that this pathway can contribute to the destruction of ␤-cells (28).
In the diabetic state potentially damaging oxidative stress is provoked in islets as well as in other cells (29 -31). Because the expression levels of antioxidant enzymes such as superoxide dismutase, catalase, and glutathione peroxidase are known to be very low in islets compared with other tissues (32), it may be that ␤-cells are particularly susceptible to oxidative stress. Several studies have provided evidence that oxidative stress caused by diabetes mediates the toxic effect of hyperglycemia on ␤-cells (33)(34)(35)(36)(37), suppressing insulin gene transcription and glucose-stimulated insulin secretion and even producing apoptosis in ␤-cells. PDX-1 expression and DNA binding activity are also suppressed by oxidative stress provoked by the diabetes state (34 -36). Furthermore, some toxic effects of hyperglycemia on ␤-cells in rodent models are reduced by antioxidant treatment (35,36).
In this study, we show that activation of the hexosamine pathway in ␤-cells leads to suppression of PDX-1 DNA binding activity, of several ␤-cell specific genes, and of glucose-stimulated insulin secretion. Also, we show that these phenomena are produced by the induction of oxidative stress rather than increased O-linked glycosylation. These results suggest that activation of the hexosamine pathway may contribute to the ␤-cell dysfunction of diabetes by provoking oxidative stress.

MATERIALS AND METHODS
Isolation and Culture of Rat Pancreatic Islets-Islets were isolated from the pancreas of 200 -250-g male Harlan Sprague-Dawley rats (Taconic Farms, Germantown, NY) by collagenase digestion with a method described previously (38). Briefly, the common bile duct was cannulated and injected with 6 ml of cold M199 medium containing 1.5 mg/ml collagenase (Roche Molecular Biochemicals). The islets were then separated with Histopaque 1077 (Sigma) density gradient. The washed islets were picked individually under a dissecting microscope and cultured overnight in RPMI 1640 medium (11 mM glucose, supplemented with 10% fetal calf serum, 100 units/ml penicillin, and 0.1 mg/ml streptomycin sulfate) in a humidified atmosphere of 5% CO 2 at 37°C. All animal procedures were approved by the Animal Care Committee of the Joslin Diabetes Center. Groups of about 500 islets were cultured for 3 days in 6-cm bacteriologic Petri dishes in 3 ml of RPMI containing either low (5 mM) or high (20 mM) glucose concentrations, prior to the studies described below. For some studies, the islets were exposed for 24 h to 5 mM glucosamine (Sigma) and 20 or 50 M of the O-GlcNAcase inhibitor PUGNAc (39) in RPMI with 11 mM glucose.
Preparation of Recombinant Adenovirus Containing the cDNA Encoding GFAT-A recombinant adenovirus containing the cDNA encoding GFAT was prepared using the AdEasy system (40). In brief, the GFAT encoding region (kindly provided by Dr. Donald A. McClain) was cloned into the HindIII-EcoRV site of a shuttle vector pAdTrack-CMV. One microgram of the resultant plasmid was linearized with PmeI and cotransformed into electrocompetent Escherichia coli BJ5183 cells with 0.1 g of adenoviral backbone plasmid, pAdEasy-1, by homologous recombination with electroporation (2, 500V, 200 Ohms). Then the resultant plasmid was retransformed into E. coli XL-Gold Ultracompetent cells (Strategene, La Jolla, CA). The plasmid was linearized with PacI and then transfected using LipofectAMINE (Life Technologies, Inc.) into the adenovirus packaging 293 cells, which were maintained in Dulbecco's modified Eagle's medium. Ten days after transfection, cell lysate was obtained from the 293 cells. The cell lysate was added to 293 cells again, and when most of the cells were detached, cell lysate was obtained again (this process was repeated three times). Control adenovirus expressing green fluorescent protein (Ad-GFP) was prepared in the same manner. To determine viral titers, confluent 293 cells were infected with a 1:10,000 dilution of the final lysate containing Ad-GFAT (with GFP). After 18 h of incubation, the effective titer was determined by the following formula: 10 7 ϫ the average number of GFP-positive cells/field (ϫ100 magnification) (plaque-forming units/ml). This number was considered to be proportional to the number of infective particles in the original lysates. Isolated rat islets (ϳ500 islets) were infected with Ad-GFAT or Ad-GFP, using a 1-h exposure to 30 l of the adenovirus (1 ϫ 10 8 plaque-forming units/ml). One hour after infection the islets were cultured in 3 ml of RPMI medium (5 or 20 mM glucose) in 6-cm bacteriologic Petri dishes.
RNA Extraction and cDNA Synthesis-Total RNA was extracted from islets using Trizol (Life Technologies, Inc.). After quantification by spectrophotometry, 500 ng of RNA was heated at 85°C for 3 min and then reverse-transcribed into cDNA in a 25-l solution containing 160 M dNTP, 50 ng of random hexamers, 10 mM dithiothreitol, and 200 units of Superscript II RNase H Ϫ reverse transcriptase (Life Technologies, Inc.). Incubations were performed for 10 min at 25°C, 60 min at 42°C, and 10 min at 95°C. The final cDNA reaction products were then diluted with H 2 O to a concentration corresponding to 20 ng of starting RNA (20-ng RNA equivalents) per 3 l.
Semiquantitative Radioactive Multiplex PCR-Primer design was optimized for multiplex PCR with Eugene TM version 2.2 (Daniben Sys-tems, Cincinnati, OH). Polymerization reactions were performed with a Perkin-Elmer 9700 Thermocycler using a 50-l reaction volume containing 3 l of cDNA (20-ng RNA equivalents), 160 M cold dNTPs, 2.5 Ci of [␣-32 P]dCTP (3000 Ci/mmol), 10 pmol of appropriate oligonucleotide primers, 1.5 mM MgCl 2 , and 5 units of AmpliTaq Gold DNA polymerase (Perkin-Elmer, Norwalk, CT). The oligonucleotide primers and cycle number used for multiplex PCR were as follows: To ensure the validity of the measurement of mRNA levels by semiquantitative-radioactive multiplex PCR, control experiments were performed using normal rat islet cDNA to show that the amount of each amplimer obtained in a multiplex PCR was independent of the presence of the other primers, excluding the possibility of strong interferences between primers as we reported previously (6). In brief, the number of cycles were adjusted to be in the exponential phase of the amplification of each product, and we verified that the amount of each PCR product in a multiplex reaction increases linearly with the amount of starting cDNA (5-40-ng RNA equivalents), ensuring that changes in the ratio of PCR product to control gene product truly reflect a change in mRNA abundance of that gene relative to the control gene.
Western Blotting-Whole cell extracts were obtained from uninfected islets and islets (ϳ500 islets) infected with Ad-GFAT or Ad-GFP. After treatment with lysis buffer (20 mM Tris-HCl, pH 7.5, 150 M NaCl, 1% Nonidet P-40), supernatants were collected. Ten micrograms of cell extracts were fractionated by 10% SDS-polyacrylamide gel electrophoresis, and transferred to polyvinylidene difluoride membranes (Immun-Blot TM PVDF membrane; Bio-Rad) using transfer buffer containing 20% methanol, 25 mM Tris base, and 192 mM glycine (300mA, 2 h). After blocking at room temperature for 1 h in 50 mM Tris-HCl, 150 mM NaCl, 0.1% Tween 20 with 5% nonfat dry milk, the membranes were incubated at 4°C overnight in TBS buffer (50 mM Tris-HCl, 150 mM NaCl) containing a 1:1,000 dilution of monoclonal mouse anti-Olinked glucosamine antibody (RL-2 antibody; Affinity Bioreagents, Golden, CO), and washed three times in TBS buffer with 0.1% Tween 20 (TBS-T). The membranes were then incubated for 1 h at room temperature in TBS containing 1:1,000 dilution of goat anti-mouse IgG antibody (Bio-Rad) coupled to horseradish peroxidase, followed by three 10-min washings with TBS-T. Immunoreactive bands were made visible by incubation with lumiGLO (Cell Signaling, Beverly, MA) and exposed to x-ray film (Kodak, New Haven, CT).
Evaluation of Glucose-stimulated Insulin Secretion-Glucose-stimulated insulin secretion was determined by static incubation. Isolated rat islets (50 islets) were preincubated for 30 min with 2 ml of HEPESbalanced Krebs-Ringer bicarbonate buffer and then incubated for 60 min in the same buffer supplemented with 0.5% bovine serum albumin and either 2.8 or 16.7 mM glucose. The insulin secreted into the medium was determined using a radioimmunoassay kit (Linco Research, St. Charles, MO) using rat insulin as the standard.
Hydrogen Peroxide Assay-Hydrogen peroxide levels were determined using a PeroxiDetect TM kit (Sigma). In brief, samples (100 l) containing different concentrations of glucosamine were incubated for 30 min with 1 ml of aqueous peroxide color reagent (aqueous solution containing 100 mM sorbitol and 125 M xylenol orange) and 10 l of ferrous ammonium sulfate reagent (25 mM ferrous ammonium sulfate in 2.5 M sulfuric acid), and the hydrogen peroxide level was measured by the absorbance at 560 nm. Hydrogen peroxide levels are expressed relative to sample treated with 10 M H 2 O2, with the value being arbitrarily set at 100.
Gel Mobility Shift Assay-Nuclear extracts were obtained from uninfected islets and islets (ϳ500 islets) infected with Ad-GFAT and Ad-GFP. The cells were treated with 1 ml of hypotonic buffer (20 mM HEPES, pH 7.9, 20 mM NaF, 1 mM EDTA, 1 mM EGTA, 1 mM Na 4 P 2 O 7 , 1 mM Na 3 VO 4 , 1 mM dithiothreitol), and high salt buffer (420 mM NaCl and 20% glycerol in above buffer) was added to the pellet, followed by 1 h of incubation at 4°C. The supernatants were used as nuclear extracts. Two micrograms of nuclear extract were incubated with 2 g of poly(dI-dC) at 4°C in a 20 l of reaction buffer (10 mM HEPES, pH 7.8, 0.1 mM EDTA, 75 mM KCl, 2.5 mM MgCl 2 , 1 mM dithiothreitol, and 3% Ficoll). The binding reaction was initiated by adding 32 P-labeled double-stranded oligonucleotide probe. A doublestranded oligonucleotide containing the rat insulin gene PDX-1 binding region and surrounding sequences (ACG TCC TCT TAA GAC TCT AAT TAC CCT ACG T) was used as a binding probe. In some of the reaction mixtures, anti-PDX-1 antiserum (13) was added 1 h before addition of the DNA probes. After the binding reaction, the samples were analyzed by separation on 6% polyacrylamide gel (150 V, 1 h) in 1ϫ TBE buffer (45 mM Tris-HCl, 45 mM boric acid, 1 mM EDTA), followed by autoradiography.
Statistical Analysis-All results are presented as the means Ϯ S.E. Statistical analysis was performed using the unpaired Student's t test.

Activation of the Hexosamine Pathway Leads to Deterioration
of ␤-Cell Function-To examine the possible involvement of the hexosamine pathway on pancreatic ␤-cell function, primary rat islets were exposed to GFAT-expressing adenovirus (Ad-GFAT) or control adenovirus (Ad-GFP) expressing GFP only. Fig. 1A shows representative islets 3 days after infection with Ad-GFAT. Islet cells except for some in the center of the islets were GFAT(GFP)-positive. Total RNA was isolated 3 days after infection, and RT-PCR was performed. As shown in Fig. 1B, infection with Ad-GFAT led to an increase in GFAT expression in the islets as compared with uninfected islets or islets infected with a control adenovirus, Ad-GFP. Also, the extent of O-linked glycosylation was increased after exposure to Ad-GFAT; the immunoreactivity determined with RL-2 antibody identified increases of at least three proteins (ϳ70, 80, and 110 kDa) for O-linked N-acetylglucosamine (Fig. 1B).
To examine the effects of hexosamine pathway activation on the glucose-stimulated insulin secretion, static incubation was performed using the islets exposed to Ad-GFAT or after treatment with glucosamine. As shown in Fig. 1C, glucosamine suppressed glucose-stimulated insulin secretion. Moreover, overexpression of GFAT decreased glucose-stimulated insulin secretion at 3 day of culture in 20 mM glucose but not in 5 mM glucose. In contrast, insulin secretion was not decreased after exposure to Ad-GFP (Fig. 1C).
To further evaluate the possible involvement of the hexosamine pathway on ␤-cell function, we studied the effects of overexpression of GFAT and glucosamine treatment on ␤-cell specific gene expression. Three days after the infection, total mRNA was isolated and used for multiplex RT-PCR followed by densitometry analyses. Insulin mRNA was decreased by GFAT overexpression in the presence of a high glucose concentration (20 mM), whereas no decrease was observed in the Ad-GFPinfected islets ( Fig. 2A). GLUT2 and glucokinase mRNA levels were also decreased, whereas glucagon, Kir 6.2, and SUR mRNA levels were unchanged. Similarly, treatment with glucosamine also caused a dose-dependent decrease in the expression of several ␤-cell-specific genes (insulin, GLUT2, and glucokinase), whereas glucagon gene expression was unaffected (data not shown). To understand the molecular basis of the suppression of transcription in these three genes, we looked for changes in the DNA binding activity of PDX-1. When primary rat islets were exposed to Ad-GFAT at a high glucose concentration (20 mM), PDX-1 DNA binding was markedly decreased (Fig. 2B). The specificity of the band in Fig. 2B for PDX-1 was confirmed by its ablation by PDX-1 antibody but not by preimmune serum.

Induction of O-Linked Glycosylation Does Not Suppress ␤-Cell Function-O-Linked
glycosylation is thought to be induced by activation of the hexosamine pathway. Overexpression of GFAT was found to increase O-linked glycosylation in islets (Fig. 1B). To examine the possible involvement of Olinked glycosylation in hexosamine pathway-mediated ␤-cell dysfunction, we enhanced O-linked glycosylation using an O-GlcNAcase inhibitor PUGNAc, as shown in Fig. 3A. Despite these changes, insulin, GLUT2, and glucokinase mRNA levels were not affected (Fig. 3B). Moreover, PDX-1 DNA binding was not affected by PUGNAc treatment (Fig. 3C). Thus, O-linked glycosylation alone does not seem to duplicate the suppressive effects of GFAT overexpression.
Oxidative Stress Is Involved in Hexosamine Pathway-mediated ␤-Cell Dysfunction-Oxidative stress is known to be provoked by high glucose concentrations (29 -31) and is suggested to play a role in the ␤-cell dysfunction of diabetes (32)(33)(34)(35)(36). We found that glucosamine itself can provoke oxidative stress.  1. Overexpression of GFAT using adenovirus in rat pancreatic islets and its effect on glucose-stimulated insulin secretion. A, isolated rat islets (ϳ500 islets) were infected with a recombinant adenovirus, Ad-GFAT or Ad-GFP (1 ϫ 10 8 plaque-forming units/ ml) and cultured for 3 days in RPMI medium (11 mM glucose). As seen by GFP in this light micrograph of whole floating cells, many cells are infected at 3 days after infection with Ad-GFAT, which also contains GFP. B, islets were cultured for 3 days in RPMI medium after infection with Ad-GFAT or Ad-GFP. Total RNA was isolated for RT-PCR, and whole cell extracts were used for Western blot analysis with proteins identified with RL-2 antibody for O-linked glycosylation. C, after 3 days in culture at low (5 mM) and high (20 mM) glucose after control and test adenovirus infection, static glucose-stimulated insulin secretion was examined over 60 min with batches of 50 islets, comparing 2.8 with 16.7 mM glucose. Glucose-stimulated insulin secretion was also examined in uninfected islets after 24 h of exposure to 5 mM glucosamine in RPMI medium. The data are expressed as the means Ϯ S.E. in bar graphs (n ϭ 3).**, p Ͻ 0.01. Actually glucosamine can increase the level of H 2 O 2 in a dosedependent manner, with levels being reduced by the addition of NAC (Fig. 4). Thus, to evaluate the possible involvement of oxidative stress in hexosamine pathway-mediated ␤-cell dysfunction, we sought to determine whether oxidative stress is involved in the change of ␤-cell function induced by GFAT overexpression. First, we examined the effects of antioxidant on hexosamine pathway-mediated suppression of glucose-stimulated insulin secretion. As shown in Fig. 5A, glucose-stimulated insulin secretion was partially restored by the addition of the antioxidant NAC. In addition, antioxidant treatment partially or totally restored the expression levels of insulin, GLUT2, and glucokinase mRNA (Fig. 5B). As shown in Fig. 5C, PDX-1 DNA binding was restored by NAC treatment. These results suggest that oxidative stress is likely to be involved in hexosamine pathway-mediated ␤-cell dysfunction. DISCUSSION Chronic hyperglycemia is accompanied by a decline in glucose-stimulated insulin secretion and insulin biosynthesis, a phenomenon known as glucose toxicity (1)(2)(3)(4)(5)(6)(7)(8). To understand the mechanisms behind these changes, we examined the hypothesis that activation of the hexosamine pathway leads to deterioration of ␤-cell function by inducing oxidative stress. In an earlier effort to show a deleterious effect of this pathway on FIG. 2. Effect of GFAT overexpression on ␤-cell-associated gene expression and PDX-1 DNA binding activity. A, isolated rat islets (ϳ500 islets) were infected with recombinant adenovirus, Ad-GFAT, or Ad-GFP at 1 ϫ 10 8 plaque-forming units/ml and cultured for 3 days in the presence of 5 or 20 mM glucose. Total RNA was isolated from the islets and used for RT-PCR. The relative mRNA levels of each gene (insulin, glucagon, GLUT2, glucokinase, Kir 6.2, and SUR) are expressed as the means Ϯ S.E. in bar graphs with values of the Ad-GFP-infected islets at 5 mM glucose being arbitrarily set at 100 (n ϭ 3). *, p Ͻ 0.05; **, p Ͻ 0.01. B, three days after the infection with Ad-GFP or Ad-GFAT in the presence of 5 or 20 mM glucose, nuclear extracts were obtained from the islets, and gel shift assays were performed. A double-stranded oligonucleotide containing the rat insulin gene PDX-1 binding region was used as a binding probe. In the fifth and sixth lanes, anti-PDX-1 antiserum and preimmune serum were employed. Similar results were obtained in three independent experiments.

FIG. 3. Effect of O-linked glycosylation on ␤-cell function.
A, rat islets were treated for 24 h with glucosamine (5 mM) or PUGNAc (20 M) in RPMI medium (11 mM glucose), whole cell extracts were obtained, and Western blotting was performed with RL-2 antibody for O-linked glycosylation. B, three days after treatment with glucosamine (5 mM) or PUGNAc (20 and 50 M), total RNA was isolated from the islets and used for RT-PCR. The relative expression of each gene (insulin, GLUT2, and glucokinase) is expressed as the mean Ϯ S.E. in bar graphs, with values of uninfected islets being arbitrarily set at 100 (n ϭ 3). *, p Ͻ 0.05; **, p Ͻ 0.01. C, three days after treatment with glucosamine (5 mM) or PUGNAc (20 and 50 M), nuclear extract was obtained from the islets, and gel shift assay was performed. The binding probe for the rat insulin gene PDX-1 binding region was the same as in Fig. 2B. Similar results were obtained in three independent experiments.
␤-cells, Tang et al. (41) overexpressed GFAT in the ␤-cells of transgenic mice and found hyperinsulinemia and insulin resistance but no hyperglycemia except for mild hyperglycemia in male mice. We suspect that the failure to find more impairment of ␤-cell function may be due to substrate limitation; that is, hyperglycemia may be a requirement for the pathway to become important even when overexpressed.
In the present study, we combined GFAT overexpression with hyperglycemia. Glucose-stimulated insulin secretion and the mRNA levels of several ␤-cell specific gene (insulin, GLUT2, and glucokinase) were suppressed by GFAT overexpression or glucosamine treatment. It was surprising that insulin secretion did not seem to be differentially influenced by 3 days of exposure to 5 versus 20 mM glucose for 3 days. In the experiments with Ad-GFP of Fig. 1, the basal secretion in the islets exposed to high glucose was elevated, as might have been expected; these changes, however, were not statistically different. Both GFAT overexpression under high glucose and glucosamine treatment were accompanied by a decrease of PDX-1 DNA binding activity (Figs. 2 and 3). Moreover, these hexosamine pathway-mediated changes were not mimicked by induction of O-linked glycosylation (Fig. 3), but they were suppressed by antioxidant treatment (Fig. 5). PDX-1, a homeodomaincontaining transcription factor, is expressed in pancreas and duodenum and contributes to the activation of insulin, GLUT2, and glucokinase genes (9 -14). We showed a significant decrease of PDX-1 DNA binding activity after GFAT overexpression, which may explain at least part of the suppression of these ␤-cell specific genes. It has been reported that after chronic exposure to a high glucose concentration, PDX-1 expression or activity is reduced in association with a reduction of insulin gene transcription (4 -8, 35, 36). Thus, we speculate that the activation of the hexosamine pathway might be involved in the decrease of PDX-1 activity found under hyperglycemic conditions. Consistent with this hypothesis are reports that heterozygous mutations of PDX-1 in humans are associated with diabetes (42) and that heterozygous PDX-1 knockout mice display impaired glucose tolerance (43).
In this study glucosamine was found to increase hydrogen peroxide levels (Fig. 4), indicating that glucosamine in ␤-cells could be a toxic reactive oxygen species. Further support for an important role for oxidative stress is the finding that antioxidant treatment prevented some of the adverse effects caused by activation of the hexosamine pathway (Fig. 5). These results are compatible with previous results showing that oxidative stress suppresses insulin gene transcription (34 -36). Thus, in the diabetic state, activation of the hexosamine pathway may play a role in the development of ␤-cell dysfunction by provoking oxidative stress. Indeed, it has been reported that in the diabetic state, reactive oxygen species are produced in ␤-cells as well as in various tissues. The levels of 8-hydroxy-2Ј-deoxyguanosine and 4-hydroxy-2-nonenal modified proteins, markers for oxidative stress, are induced in ␤-cells of diabetic rats (29). Also, the expression levels of antioxidant enzymes, such as superoxide dismutase, catalase, and glutathione peroxidase, are very low in islets as compared with other tissues and cells (32). Therefore, once the ␤-cells face oxidative stress, they are likely to be vulnerable. This vulnerability adds further weight to the probability that oxidative stress mediates the toxic effect of hyperglycemia. It has been reported recently that hypergly- FIG. 5. Involvement of oxidative stress in hexosamine pathway-mediated ␤-cell dysfunction. A, rat islets were incubated for 3 days at 20 mM glucose after infection with Ad-GFAT or Ad-GFP with or without NAC (5 mM), followed by glucose-stimulated insulin secretion studies. Batches of 50 islets were exposed to 2.8 or 16.7 mM glucose for 60 min. The data are expressed as the means Ϯ S.E. in bar graphs (n ϭ 3). **, p Ͻ 0.01. B, three days after the infection with Ad-GFAT or Ad-GFP with or without NAC (5 mM) in the presence of 20 mM glucose, total RNA was isolated from the islets and used for RT-PCR. The relative mRNA levels of each gene (insulin, GLUT2, and glucokinase) are expressed as the means Ϯ S.E. in bar graphs with values of Ad-GFP-infected islets without NAC being arbitrarily set at 100 (n ϭ 3). *, p Ͻ 0.05; **, p Ͻ 0.01. C, three days after infection of Ad-GFP or Ad-GFAT with or without NAC (5 mM) in the presence of 20 mM glucose, nuclear extracts were obtained from the islets, and gel shift assay was performed. The binding probe for the PDX-1 binding region of the insulin gene was the same as in Fig. 2B. Similar results were obtained in three independent experiments. cemia-induced oxidative stress (mitochondrial superoxide) leads to activation of the hexosamine pathway (44). Such a finding raises the possibility that oxidative stress generated in the mitochondria could amplify production of reactive oxygen species from the hexosamine pathway, thus creating a particularly deleterious situation for ␤-cells. Antioxidant treatment, which can provide some protection for ␤-cells against glucose toxicity and ameliorate glucose tolerance (35,36), may exert its beneficial effects by suppressing oxidative stress originating from both mitochondria and the hexosamine pathway.
It is well known that O-linked glycosylation is induced through activation of the hexosamine pathway and that many proteins are O-glycosylated (16 -20). It has also been suggested that O-linked glycosylation could exert a destructive effect upon ␤-cells; STZ, a diabetogenic reagent, elevates O-GlcNAc levels in islets, and GlcNAc modification is correlated with diabetes, suggesting the adverse effects of O-linked glycosylation on ␤-cell function (28). In our study, O-linked glycosylation was increased by GFAT overexpression (Fig. 1B). As shown in Fig. 3, after treatment with an O-GlcNAcase inhibitor PUGNAc, the extent of O-linked glycosylation was increased, but ␤-cell-specific gene expression levels were not affected, suggesting that the simple induction of O-linked glycosylation is not enough to explain all of the changes by the hexosamine pathway. However, we cannot exclude the possibility that Olinked glycosylation could have some less obvious adverse effects on ␤-cell function.
In conclusion, these studies indicate that activation of the hexosamine pathway causes deterioration of ␤-cell function by inducing oxidative stress, implying that this mechanism could be responsible for at least some of the ␤-cell glucose toxicity found in diabetes.