Glycosylation Mediates Up-regulation of a Potent Antiangiogenic and Proatherogenic Protein, Thrombospondin-1, by Glucose in Vascular Smooth Muscle Cells*

Accelerated development of atherosclerotic lesions remains the most frequent and dangerous complication of diabetes, accounting for 80% of deaths among diabetics. However, our understanding of the pathways mediating glucose-induced gene expression in vascular cells remains controversial and incomplete. We have identified an intracellular metabolic pathway activated by high glucose in human aortic smooth muscle cells that mediates up-regulation of thrombospondin-1 (TSP-1). TSP-1 is a potent antiangiogenic and proatherogenic protein that may represent an important link between diabetes and vascular complications. Using different glucose analogs and metabolites sharing distinct, limited metabolic steps with glucose, we demonstrated that activation of TSP-1 transcription is mediated by the hexosamine pathway of glucose catabolism, possibly resulting in modulation of the activity of nuclear proteins activity through their glycosylation. Specific inhibitors of glutamine: fructose 6-phosphate amidotransferase (GFAT), an enzyme controlling the hexosamine pathway, as well as direct inhibitors of protein glycosylation efficiently inhibited TSP-1 transcription and the activity of a TSP-1 promoter-reporter construct stimulated by high glucose. Overexpression of recombinant GFAT resulted in increased TSP-1 levels. Pharmacological inhibition of GFAT or protein glycosylation inhibited increased proliferation of human aortic smooth muscle cells caused by glucose. We have demonstrated that the hexosamine metabolic pathway mediates up-regulation of TSP-1 by high glucose. Our results suggest that the hexosamine pathway and intracellular glycosylation may control important steps in initiation and development of atherosclerotic lesions.

Thrombospondins are matricellular proteins that regulate cell-cell and cell-matrix interactions (1,2). Recent genetic asso-ciation studies link the thrombospondin (TSP) 2 protein family to the development of atherosclerotic lesions (3)(4)(5)(6)(7)(8)(9)(10). TSP-1 was found in early atherosclerotic lesions (11), in injured vascular walls (12,13), and in cardiac allografts where its expression correlated with the degree of vasculopathy (14). The genetic disruption of TSP-1 reduced the atherosclerotic lesion area in the mouse model of atherosclerosis and suggested an important role for TSP-1 in the evolution of plaque and its composition (15). In both in vivo and in vitro studies TSP-1 induced proliferation of vascular smooth muscle cells (SMC) (16,17), and both TSP-1 and TSP-2 inhibited growth of endothelial cells (18 -21); both effects are considered proatherogenic.
Previous studies have documented increased TSP-1 levels in the plasma and kidneys of diabetic patients and diabetic animal models (22)(23)(24)(25). In mesangial cells, the level of TSP-1 was upregulated by glucose by a transcriptional mechanism (26 -28). We have recently reported increased levels of TSP-1 in the blood vessels of diabetic animals (29). Moreover, TSP-1 was up-regulated by high glucose in vitro in major cell types from large blood vessels. These observations suggest that TSP-1 represents an important link between diabetes, hyperglycemia, and accelerated atherogenesis.
A large number of clinical studies and trials have conclusively identified hyperglycemia as an independent risk factor for development of both micro-and macrovascular complications (30 -33). Recently, the Epidemiology of Diabetes Intervention and Complications study reported that, as compared with conventional therapy, intensive glycemic control reduced the risk of the most serious cardiovascular events such as heart attacks, stroke, and death by nearly 60%. These findings clearly underscore the importance of hyperglycemia as a critical player in the development of pathogenic complications associated with diabetes. Glucose regulates the expression of a number of vascular * This work was supported by National Institutes of Health Grants R01 genes, and many of them have been linked to the development of atherosclerosis and abnormal angiogenesis (34). However, the molecular mechanisms activated by glucose that lead to changes in the gene expression profile in vascular cells remain controversial and incomplete.
In this study we have undertaken an unbiased approach to identify specific molecular mechanisms that mediate the upregulation of TSP-1 expression in response to high glucose in cultured HASMC. Using different glucose analogs and metabolites sharing distinct, limited metabolic steps with glucose, we found that the hexosamine pathway of glucose breakdown controls the expression of TSP-1 in HASMC. Our results demonstrate that the transcriptional regulation of TSP-1 is mediated by protein glycosylation, which is increased as a result of activation of the hexosamine pathway.
mRNA Stability Assay-Control cells or glucose-stimulated cells were treated with 5 g/ml actinomycin D and then lysed at different time points from 20 min to 2 h after initiation of actinomycin D treatment. mRNA levels at each time point were quantified by densitometric scans of Northern blots.
Northern Blots-Total RNA was extracted using TriZol reagent (Invitrogen). Isolated RNA was stored in diethyl pyrocarbonate-treated water at Ϫ80°C. The purity and concentration were determined by measuring the optical density at 260 and 280 nm before use. The optical density ratio at 260/280 ranged from 1.7 to 2.0. For Northern blot analysis, 10 g of SMC RNA was electrophoresed in 16.7% agarose-formaldehyde gels, transferred to nylon membranes (PerkinElmer Life Sciences), and hybridized to appropriate 32 P-labeled cDNA probes using a Random prime labeling kit (Amersham Biosciences). The cDNA probe was a 600-bp fragment of TSP-1 cDNA corresponding to the N-terminal part of TSP-1 protein that has no homology with other thrombospondins. Membranes were prehybridized in ExpressHyb Hybridization solution (BD Biosciences) for 1 h at 68°C. The heat-denatured cDNA probe was added, and hybridization was performed at 68°C for 2 h. The membranes were then washed and exposed to Kodak BioMax MR film at Ϫ80°C. Signal intensity was quantified using a Digital Science Imaging System (Version 2.0.1, Eastman Kodak Co.).
Nuclear Run-on Assay-Nuclear run-on transcriptional assays were performed as described by Greenberg and Ziff (35). After incubation of HASMC with glucose, medium was aspi-rated, and lysis buffer (10 mM Tris-HCL, pH 7.5, 10 mM NaCl, 3 mM MgCl 2 , 0.5% Nonidet P-40, protease inhibitors) was added to the cells. All the buffers were prepared in diethyl pyrocarbonate-treated water. After shaking at 4°C for 10 min, cells were scraped and spun for 5 min at 1000 ϫ g at 4°C, and the pellet was washed with lysis buffer. Nuclei were resuspended in storage buffer (100 l/plate) (50 mM Tris pH 8.3, 40% glycerol, 5 mM MgCl 2 , 0.1 mM EDTA) and stored at Ϫ80°C until used. For the in vitro transcription reactions, 20 million nuclei were resuspended in the reaction buffer (40 mM Tris, pH 8.5, 150 mM NH 4 Cl, 7.5 mM MgCl 2 , 0.62 mM ATP, 0.31 mM GTP, 0.31 mM CTP, 80 units of RNase inhibitor, 125 Ci of [ 32 P]UTP) and incubated for 20 min at 27°C. 3 l of 10 mM UTP was added, and the incubation was continued for 10 min. This step was followed by the addition of 2 l of RNase-free DNase I (10 units/ml), and the incubation was further continued for 10 min at 27°C. Then, 75 l of 20% SDS, 15 l of 0.5 M EDTA, 5 l of 1 M Tris-HCl, pH 7.4, and 55 l of 10 mg/ml proteinase K were added, and the resulting mixture was incubated for 2 h at 42°C. After phenol extraction and ethanol precipitation, RNA was dissolved in 50 l of diethyl pyrocarbonate-treated water. 5 g of TSP-1 or ␤-actin cDNA and 0.25 g of 28 S cDNA were dissolved in a denaturing solution (25 l of 10 M NaOH, 4 M NaCl/ml), spotted on nylon membranes (PerkinElmer), UVcross-linked, and dried in air. Membranes were prehybridized for 1 h at 42°C. The heat-denatured RNA probe and RNase inhibitor were added, and hybridization was performed at 42°C overnight. After two 90-min washes in 30 mM sodium citrate, pH 7, 300 mM NaCl, 0.1% SDS at 42°C, membranes were exposed to a Kodak BioMax MR film.
Inhibition of Specific Metabolic Pathways-HASMC were incubated with 30 mM D-glucose for 24 h in the presence or absence of different glycosylation inhibitors (40 M 6-diazo-5oxonorleucine (DON), 40 M azaserine, 1 mM benzyl 2-deoxy-␣-D-galactopyranoside, BG). In the case of DON, cells were preincubated for 6 -18 h prior to stimulation. To demonstrate that the effect of DON on TSP-1 expression was due to the inhibition of GFAT and the absence of the downstream metabolites of the hexosamine pathway, we used glucosamine to overcome DON inhibition. In these experiments cells were preincubated with 40 M DON for 6 h followed by stimulation with 2 mM glucosamine for 24 h.
Proliferation of Cultured Smooth Muscle Cells-HASMC were plated in 24-well clusters (Costar) (5000 cells/well) in 10% FBS, DMEM/F-12. 24 h later, the medium was changed to low glucose (5 mM) DMEM, and treatments were initiated as follows. After 4 h of pretreatment with inhibitors as indicated in the figure legends, 30 mM D-glucose, L-glucose, mannose, fructose, galactose, 2 mM glucosamine, 100 M PUGNAc, and 0.8 g/ml TSP-1 (Sigma) were added. TSP-1 was added both at the time of stimulation and again the next day to imitate continuous production of TSP-1 by sugar-stimulated cells. In some of these experiments, as shown in Fig. 8A, cells were treated with 2 g/ml of anti-TSP-1 antibody (Clone 6.1, LabVision) or a control antibody against unrelated protein added at the time of stimulation and once again the next day. The control antibody did not affect HASMC proliferation (not shown). After 4 days of proliferation, plates were washed with phosphate-buffered saline, and the amount of cell DNA/well was measured using a CyQuant Cell proliferation assay kit (Molecular Probes). The time point was chosen because when the cells were plated at the density of 5000 cells/well (lower limit of the linear zone of detection in CyQuant assay), the difference in proliferation between glucose-stimulated and non-stimulated cells was greatest at 4 days. In another series of experiments, siRNA for TSP-1 and control siRNA (Ambion) were delivered to HASMC by nucleofection (Amaxa) according to the manufacturer's protocol.
Cell Transfection and Luciferase Assay-HASMC were plated in 24-well clusters (Costar) (0.3 ϫ 10 6 cells/well) in 10% FBS, DMEM/F-12 media. The next day cells were transiently transfected with a Ϫ2033/ϩ66 pTHBS-1 promoter-luciferase reporter construct, a control pGL3 vector, or with a Ϫ121/ϩ66 pTHBS-1 promoter-luciferase construct, which was found to be unresponsive to glucose but still had high constitutive activity comparable with the full promoter reporter construct. The transfection procedure was carried out using Lipofectin reagent (Invitrogen) following the manufacturer's protocol. 6 h post-transfection, the plasmid DNA-containing medium was changed to low glucose (5 mM) 10% FBS, DMEM, and the transfected cells were treated with 30 mM glucose, fructose, mannose, or galactose, as appropriate. For inhibitor studies, cells were incubated with 30 mM glucose in the presence or absence of different glycosylation inhibitors, as indicated in the figures. After 42 h of incubation, cell extracts were assayed for luciferase activity using a luciferase assay kit (Promega). Protein concentrations in cell lysates were analyzed using the BCA protein assay reagent (Pierce), and the activity of luciferase was normalized to total protein concentrations in lysates.
Overexpression of Glutamine; Fructose 6-Phosphate Amidotransferase-HASMC were trypsinized, washed, and plated in 6-well clusters (Costar) (0.8 ϫ 10 6 cells per well) in 10% FBS, DMEM/F-12 media. Confluent cells were transiently transfected with 2 g GFAT-pcDNA3.1 or with pcDNA3.1 as a control in Opti-MEM-1 reduced serum medium using Lipofectin reagent (Invitrogen) according to the manufacturer's protocol. 5 h post-transfection, the plasmid DNA-containing medium was replaced by 10% FBS, DMEM/F-12 media. Total RNA was extracted from the transfected cells 48 h post-transfection and subjected to Northern blotting, as previously described. To detect GFAT protein expression, whole cell lysates were prepared from the transfected cells using a hypotonic cell lysis buffer containing Nonidet P-40. Cell pellets were resuspended in protein sample buffer and boiled for 10 min, and samples were resolved in 8% SDS-PAGE gel. Proteins were visualized by staining with Coomassie Blue to detect overexpression of the 80-kDa protein.
Western Blot-Lysates of cells transfected with GFAT or controls were analyzed by SDS gel electrophoresis, and Western blotting was performed using human GFAT antibody (36) kindly provided by Dr. E. D. Schleicher, University of Tubingen, Germany, anti-O-GlcNAc, an antibody recognizing cytoplasmic and intranuclear O-linked glycoproteins (Affinity Bioreagents, Inc.), and ␤-actin antibodies. Lysates of cells treated with different sugars, glucosamine, STZ, PUGNAc, and DON, as indicated in the figure legend, were resolved in 12% SDS-PAGE gel, and Western blotting was performed using anti-O-GlcNAc. The membrane was also stained with Ponceau S to determine equal protein loading.
Detection of Intracellular Glycosylation by Immunofluorescence-Immunofluorescence was done using anti-O-GlcNAc antibody as described by Chen et al. (37). HASMC were grown on coverslips in 6-well clusters (Costar; 0.75 ϫ 10 6 cells per well) in 10% FBS, DMEM/F-12 medium. Confluent cells were placed in 0.2% FBS, low glucose (5 mM) DMEM for 24 h and then incubated with 30 mM D-glucose, mannose, and galactose for varying periods of time. At the appropriate end points, the media were aspirated, and the cells were fixed for 20 min at 25°C in a solution containing 4% paraformaldehyde and 0.2% Triton-X. The cells were blocked in 5% donkey serum for 60 min at 25°C. Subsequently, cells were incubated for 60 min at 25°C with a 1:100 dilution of anti-O-GlcNAc (clone RL2) mouse monoclonal antibody against O-linked N-acetylglycosylated proteins in 5% donkey serum followed by incubation with a 1:50 dilution of rhodamine red-X-conjugated donkey anti-mouse immunoglobulin G (Jackson Immunoresearch Inc.) for 60 min at 25°C. Quantitation of immunofluorescence was performed with Image-Pro Plus 4.5.1 (Media Cybernetics).
Lactate Dehydrogenase Assay-The cytotoxicity of all sugars and inhibitors at the concentrations used was detected by the measurement of lactate dehydrogenase activity released from the cytosol of damaged cells into the supernatant, according to the manufacturer's protocol (Roche Applied Science). Briefly, HASMC were grown in 6-well clusters in full growth medium. 24 h before an experiment, cells were placed in low glucose (5 mM) DMEM, 0.2% FBS and incubated with different glucose analogs in the presence or absence of glycosylation inhibitors, as previously described. At the appropriate end point, 100 l of supernatant was placed in a 96-well microtiter plate followed by the addition of 100 l of the reaction mixture provided in the assay kit. The resulting mixture was incubated at room temperature for 30 min, and the absorbance of the samples was measured at 490 nm using a microplate reader. To determine the maximum lactate dehydrogenase release, the supernatant collected from Triton X-100-treated cells was included as a positive control.

XTT (Sodium 3,3Ј-{1-[Phenylamino)carbonyl]-3,4-tetrazoli-um}-bis(4-methoxy-6-nitro) Benzene Sulfonic Acid Hydrate)
Cell Viability Assay-The toxicity of glycosylation inhibitors at concentrations used in the present study was also assessed by the XTT viability assay based on the measurement of the activity of mitochondrial enzymes, according to the manufacturer's protocol (Biotium, Inc.). Briefly, HASMC were plated in 96-well tissue culture plates. Confluent cells were placed in 100 l of low glucose (5 mM) DMEM, 0.2% FBS for 24 h before stimulation with sugars. Cells were then treated with different glucose analogs as well as with glycosylation and GFAT inhibitors, as previously described. At the appropriate end point, activated XTT solution was added to each well, incubation was continued for 2-5 h, and the absorbance of the samples was measured at 490 nm using a microplate reader.
Statistical Analyses-All values are presented as the means Ϯ S.E., and significant differences between means were evaluated

Regulation of TSP-1 Expression by Glucose in Vascular SMC
by the paired Student's t test. Probability values of Յ0.05 were considered to be statistically significant.

Identification of the Hexosamine Pathway as a Candidate Mediator of TSP-1 Up-regulation by High Glucose in HASMC-
Most of the effects of glucose depend on its intracellular breakdown (38,39). However, some effects result from a change in osmolarity or cell membrane properties and do not require glucose uptake by cells (40). To find out whether intracellular glucose metabolism is required for TSP-1 expression as well as to identify the specific steps in intracellular glucose breakdown controlling TSP-1 expression, we used different glucose analogs sharing distinct metabolic steps with glucose as follows; 1) glucose analogs that are transported into the cell and share limited metabolic steps with glucose (fructose, mannose, and galactose), 2) cell-permeable glucose analogs that cannot be metabolized (2-deoxyglucose, which cannot be metabolized beyond 2-deoxyglucose 6-phosphate), 3) a downstream metabolite of the glycolytic pathway (dihydroxyacetone), and 4) a cellimpermeable glucose analog to provide osmolarity control (biologically inactive L-glucose). 30 mM L-glucose did not affect TSP-1 mRNA levels, excluding osmolarity change as a cause of TSP-1 up-regulation (Fig. 1). 30 mM 2-deoxyglucose and dihydroxyacetone failed to induce TSP-1 expression, suggesting that a limited breakdown of glucose is required and sufficient for up-regulation of TSP-1 mRNA levels. Other sugars (30 mM fructose, mannose, and galactose) increased TSP-1 mRNA expression up to 4-fold as compared with control cells (Fig. 1).
This effect is similar to that of 30 mM D-glucose, suggesting that these four sugars share the metabolic pathway controlling TSP-1 expression.
Production of metabolites for protein glycosylation is the common pathway shared by all of these sugars (Fig. 2). In glucose metabolism, formation of metabolites of glycosylation is mediated by an activation of the hexosamine biosynthetic pathway. The hexosamine pathway starts with fructose 6-phosphate, and our results with 2-deoxyglucose and dihydroxyacetone also limited the relevant steps of glycolysis to this intermediate, suggesting the hexosamine pathway as a candidate for a metabolic pathway mediating TSP-1 up-regulation by high levels of glucose.
Hexosamine Pathway and Protein Glycosylation Mediate the Increase in TSP-1 mRNA Levels-To confirm that the hexosamine pathway and protein glycosylation mediate the up-regulation of TSP-1, we used specific inhibitors of GFAT, the ratelimiting enzyme of the hexosamine pathway, and direct glycosylation inhibitors that interfered with the transfer of a sugar moiety to a protein. Preincubation of HASMC with GFAT inhibitors, DON (40 M, 18 h), and azaserine (40 M, added at the time of glucose stimulation) inhibited the increase in glucose-induced TSP-1 mRNA levels by up to 9-fold and up to 12.9-fold, respectively (Fig. 3, A and B), as quantified by densitometric scans of Northern blots. Similarly, when HASMC were preincubated with 30 mM glucose in the presence of the direct glycosylation inhibitor BG (1 mM, added at the time of stimulation with glucose), there was no increase in TSP-1 mRNA expression. As shown in Fig. 3C, TSP-1 mRNA expression in response to 30 mM glucose was inhibited by 3-fold. The inhibitors also down-regulated the basal level of TSP-1, suggesting that the physiological concentration of glucose maintains constitutive TSP-1 expression. However, this down-regulation was not statistically significant, as shown in Fig. 3 (lower panel, representative of 3-4 independent experiments). The toxicity of the inhibitors was assessed by the release of lactate dehydrogenase and by the activity of mitochondrial enzymes, as described under "Experimental Procedures." Neither the viability nor the metabolic activity was affected by any of the inhibitors in the indicated concentrations for the times used in our experiments (data not shown).   FEBRUARY 23, 2007 • VOLUME 282 • NUMBER 8

JOURNAL OF BIOLOGICAL CHEMISTRY 5707
To confirm that the activation of the hexosamine pathway results in increased expression of TSP-1, we incubated HASMC for 24 h with 2 mM glucosamine, a downstream metabolite of the hexosamine pathway. In addition, cells were treated for 24 h with STZ (5 mM) and PUGNAc (100 M). Both compounds are known to increase O-linked protein glycosylation by effectively inhibiting ␤-N-acetylglucosaminidase, an enzyme responsible for cleavage of O-GlcNAc residues from intracellular proteins. As shown in Fig.  3D, glucosamine, STZ, and PUGNAc significantly increased TSP-1 mRNA expression by 3.8-, 5.2-, and 5.2-fold, respectively, as compared with control cells. To show that the effect of GFAT inhibitors on TSP-1 expression is the result of specific inhibition of the hexosamine pathway and that the addition of downstream metabolites can overcome the effect of GFAT inhibitors, glucosa-mine was used to stimulate cells pretreated with DON. The inhibitory effect of DON (40 M) on TSP-1 mRNA expression was completely reversed when cells were treated with glucosamine (Fig. 3E).
Overexpression of GFAT Results in Increased TSP-1 mRNA Levels-After the hexosamine pathway and protein glycosylation were identified as mediators of up-regulation of TSP-1 levels by glucose in HASMC in our experiments with inhibitors, we overexpressed recombinant GFAT in HASMC to confirm its effect on TSP-1 mRNA levels. Human rGFAT cDNA was transiently overexpressed in HASMC as described under "Experimental Procedures" and detected by Coomassie Blue staining of cell lysates, resolved by SDS-PAGE. A band with the expected size of 80 kDa was detected only in the lysates of GFAT-transfected cells (Fig. 4A,  upper panel). This was further confirmed by Western blotting using an antibody specific for human GFAT (Fig. 4A, lower panel). To assure that the achieved level of GFAT overexpression was sufficient to increase O-linked glycosylation of proteins, GFAT-transfected and control cell lysates were subjected to Western blotting using the anti-O-GlcNAc antibody that specifically recognizes O-linked glycoproteins. As shown in Fig. 4B (upper panel), we observed increased O-linked glycosylation of some proteins in the GFAT-transfected cells, as indicated by the arrows. We then evaluated whether the mRNA level of TSP-1 was altered by overexpression of GFAT. As shown in Fig. 4C, we observed that after 48 h of cell culture, TSP-1 mRNA expression was increased in cells overexpressing GFAT as compared with HASMC transfected with a control vector. These results clearly demonstrate that overexpression of the rate-limiting enzyme GFAT, which provides key metabolic precursors for protein glycosylation, regulates TSP-1 synthesis in HASMC.
Transcriptional Regulation of TSP-1 mRNA Levels-The effect of high glucose on TSP-1 levels is direct and is not mediated by secreted factors; thus, when media collected from glucose-stimulated cells were added to unstimulated HASMC, they did not cause any detectable increase in TSP-1 levels (data not shown). We, therefore, proceeded to identify the molecular mechanism of up-regulation of TSP-1 mRNA by high glucose in HASMC and to confirm that the hexosamine pathway affects this mechanism directly.
Although TSP-1 mRNA levels increased dramatically after stimulation with glucose (Fig. 5A, upper panel), TSP-1 mRNA stability did not change (Fig. 5A, lower panel). The half-life of TSP-1 mRNA in HASMC determined from these experiments was estimated to be between 60 and 70 min. These data demonstrate that the increased expression of TSP-1 mRNA is not due to an increase in mRNA stability and suggest a transcriptional mechanism of up-regulation.
To determine whether the transcription of the endogenous TSP-1 gene is increased in response to acute stimulation with 30 mM glucose, we performed an in vitro nuclear run-on transcriptional assay using nuclei isolated from glucose-stimulated or control HASMC. As shown in Fig. 5B, treatment of HASMC with 30 mM glucose increased the in vitro synthesis of endogenous TSP-1 mRNA by 3-fold, as compared with control cells treated with 5 mM glucose. Under these conditions, the transcription of control genes, ␤-actin, and 28 S ribosomal RNA did not change appreciably.
To further confirm that glucose activates the transcription of the TSP-1 gene (THBS1), we transiently transfected cultured HASMC with a human THBS1 promoter-luciferase reporter gene construct and compared its activation in glucose-stimulated and control HASMC. We used Ϫ2033/ϩ66 pTHBS1, a control vector, pGL3, and a shorter pTHBS1 construct that is non-responsive to glucose (Ϫ121/ϩ66 pTHBS1) but still retains a high level of basal activity. After glucose stimulation, the activity of the pTHBS1 Ϫ2033/ϩ66-luciferase reporter construct was increased by 11-fold (Fig. 5C). Taken together, these results show that glucose activates the TSP-1 gene at the level of transcription. The effect of glucose analogs and GFAT/ glycosylation inhibitors on THSB1 promoter activation can, therefore, be assessed directly.
Effect of Glucose Analogs on THBS1 Promoter Activity-We have compared the effect of glucose analogs that up-regulated endogenous TSP-1 mRNA levels on the activity of a luciferase reporter gene under the control of the THBS1 promoter (Fig.  6A). Similar to the effect of glucose (Fig. 5C), incubation of cells transfected with pTHBS1-2033/ϩ66-luciferase with 30 mM fructose, mannose, or galactose resulted in activation of the   (n ϭ 4). B, 30 mM glucose increased in vitro synthesis of endogenous TSP-1 mRNA. Nuclei isolated from control or glucose-stimulated HASMC were subjected to nuclear run-on transcriptional assays. 28 S rRNA and ␤-actin were used as internal controls. Upper panel, two representative TSP-1 Northern blots; lower panel, densitometric quantification of three independent experiments. C, glucose stimulation increased the activity of pTHBS1 Ϫ2033/ϩ66 luciferase reporter construct as compared with controls. HASMC were transiently transfected with Ϫ2033/ϩ66 pTHBS1 and Ϫ121/ϩ66 pTHBS1 driving the expression of a luciferase gene reporter or control pGL3 vector. After a 6-h recovery, the media were aspirated, and cells were placed in 5 or 30 mM glucose. Luciferase activity was measured 42 h later. Results are the means Ϯ S.E. of 3-4 independent experiments. FEBRUARY 23, 2007 • VOLUME 282 • NUMBER 8 JOURNAL OF BIOLOGICAL CHEMISTRY 5709 THBS1 promoter. Fructose, mannose, and galactose increased the activity of luciferase expressed from the construct pTHBS1-2033/ϩ66 (2.0-, 4.0-, and 3.5-fold, in each case), whereas the activity of the pTHBS1 Ϫ121/ϩ66 construct and the activity of a control vector remained unchanged.

Regulation of TSP-1 Expression by Glucose in Vascular SMC
Effect of the Inhibitors of GFAT and Glycosylation Inhibitors on THBS1 Promoter Activity-The activity of the THBS1 promoter-luciferase reporter construct in response to 30 mM glucose was efficiently inhibited by DON (40 M) and azaserine (40 M) (Fig. 6B). Thus, specific GFAT inhibitors block the transcriptional activation of the THBS1 promoter in response to high glucose. The effect of these inhibitors was similar to their effects on endogenous TSP-1 mRNA levels induced by high glucose.
Increased Nuclear Glycosylation Induced by Glucose and Glucose Analogs-The function of a number of transcription factors and other nuclear and cytosolic proteins can be modi-fied by glycosylation (41,42). The direct effect of GFAT and glycosylation inhibitors on the activity of the THBS1 promoter suggests that transcription factor(s) responsible for this activation or co-activator(s) may be modified by glycosylation. Therefore, to examine the level of nuclear glycosylation induced by glucose and the glucose analogs that increase the expression of TSP-1, we utilized an immunofluorescence technique using the anti-O-GlcNAc antibody, which specifically recognizes all O-linked N-acetylglucosamine moieties on proteins. HASMC treated with glucose, mannose, or galactose (30 mM) for 24 h displayed a dramatic increase in immunofluorescence in their nuclei as compared with control cells (Fig. 7A). Quantification of the fluorescence intensity demonstrated a statistically significant increase (up to 4.5-fold) in nuclear glycosylation by these sugars. This increase in nuclear glycosylation was detected as early as after 1 h of incubation with each of the glucose analogs (Fig. 7B) and persisted for at least 48 h (not shown). Furthermore, Western blotting of HASMC lysates using anti-O-GlcNAc antibody revealed increased O-glycosylation after treatment with different sugars. As shown in Fig. 7C, glucose, fructose, mannose, and galactose all increased protein glycosylation as compared with control cells. Glucosamine, which bypasses the rate-limiting enzyme of the hexosamine pathway GFAT, and PUGNAc, a potent in vitro inhibitor of ␤-N-acetylglucosaminidase, increased O-glycosylation of intracellular proteins. Incubation of cells with the different sugars in the presence of the GFAT inhibitor, DON, revealed a marked decrease in protein glycosylation as compared with cells treated with sugars alone. The inhibitory effect of DON was completely reversed by glucosamine, confirming the specificity of DON effect.
SMC Proliferation Induced by High Glucose Is Due to Activation of the Hexosamine Pathway and Increased Expression of TSP-1-Previous studies have suggested that proliferation of vascular SMC is stimulated by TSP-1 (17). Our finding of dramatically elevated levels of TSP-1 in diabetic vessels in basal conditions and after intravascular injury (29) suggests its contribution to accelerated restenosis and atherosclerosis in diabetes (43,44). We evaluated the effect of glucose analogs that, similar to glucose, increased TSP-1 levels on proliferation of cultured HASMC. All analogs increased SMC proliferation, similar to the effect of glucose (Fig. 8A). Consistent with the inhibition of glucose-induced TSP-1 expression by the inhibitors of GFAT or glycosylation inhibitors, we observed that pharmacological inhibition of GFAT or protein glycosylation efficiently inhibited increased proliferation of SMC caused by high levels of glucose. The GFAT inhibitors DON and azaserine (40 M) as well as an inhibitor of direct protein glycosylation, BG (1 mM), significantly decreased SMC proliferation induced by 30 mM glucose (44.2% inhibition with DON, 20.1% with azaserine, 66.3% with BG) (Fig. 8A), as measured on the fourth day after stimulation of HASMC with high glucose. Consistent with our hypothesis that proliferation of HASMC in response to high glucose is caused by the increased levels of TSP-1, treatment of cells with anti-TSP-1 antibody significantly blocked proliferation of HASMC in response to glucose, glucosamine, and PUGNAc, whereas the incubation of cells with TSP-1 resulted in a significant increase in cell proliferation (Fig. 8A). In addition, TSP-1-targeted siRNA inhibited proliferation of cultured HASMC (Fig. 8B). Taken together, these experiments demonstrate that all sugars and activators of the hexosamine pathway and protein glycosylation that induce the expression of TSP-1 also induce HASMC proliferation. On the other hand, inhibition of the hexosamine pathway, direct inhibition of glycosylation, or direct inhibition of TSP-1 block the proliferation of HASMC.

DISCUSSION
As a protein with a number of proatherogenic properties, TSP-1 may represent an important link between diabetes and the accelerated development of atherosclerosis. We have previously reported a dramatically increased expression of TSP-1 in diabetic large vessels and an up-regulation of TSP-1 levels by high glucose in cultured cells from large blood vessels (29). However, the molecular mechanism by which the expression of TSP-1 is regulated in large blood vessels has not been addressed.
Using glucose analogs that share distinct metabolic steps with glucose, we now demonstrate that the steps of glycolysis upstream of glucose 6-phosphate formation are insufficient for induction of THBS1. Therefore, glucose has to be metabolized to increase THBS1 expression. However, only limited glucose metabolism is required, because the glycolytic steps downstream of fructose 1,6-bisphosphate are not required for induction of THBS1. Both 2-deoxyglucose, a cell-permeable glucose analog that cannot be metabolized beyond 2-deoxyglucose 6-phosphate, and dihydroxyacetone, a downstream metabolite of glycolysis, failed to induce THBS1 expression. These data suggest that the hexosamine pathway, which starts with fructose 6-phosphate and is dramatically activated in diabetes (45)(46)(47), is a mediator of THBS1 expression (see Fig. 2). Activation of the hexosamine pathway leads to the formation of intermediates used for protein glycosylation. We used fructose, mannose, and galactose sugars that share the glycosylation pathway with glucose as a means to assess the effect of glycosylation on TSP-1 expression. Glycosylation as a mediator of the increase in TSP-1 levels was further suggested by the fact that fructose, mannose, and galactose dramatically increased the expression of THBS1 in a manner similar to the effect of glucose; all these sugars can be metabolized to produce precursors for glycosylation, although their metabolic pathways are different from one another. Using inhibitors of GFAT, an enzyme controlling the hexosamine pathway, a direct inhibitor of glycosylation, overexpression of GFAT, and downstream metabolites of the hexosamine pathway, we have confirmed that the hexosamine pathway of intracellular glucose metabolism controls up-regulation of TSP-1 mRNA levels in response to high glucose levels. The hexosamine pathway has been widely implicated in the development of pathologic com- plications associated with diabetes (48,49). Previous studies have shown that flux through the hexosamine pathway plays an important role in glucose-induced increases in transforming growth factor ␤1 and plasminogen activator inhibitor type 1 synthesis in mesangial cells (36,50). In addition, increased flux through the hexosamine pathway has also been shown to be responsible for NF-B-dependent promoter activation in mesangial cells (51). Furthermore, the hexosamine pathway has been linked to the development of macrovascular complications associated with diabetes (48).
We show here that the increased synthesis of TSP-1 after acute stimulation with high glucose and other sugars is direct and results from the transcriptional activation of the THBS1 gene. The lack of a change in TSP-1 mRNA stability, increased transcription of the endogenous THBS1 gene, and the activation of its promoter suggest a transcriptional mechanism of up-regulation. The lack of increased production of TSP-1 by conditioned media from glucose-stimulated cells as well as the effect of GFAT and glycosylation inhibitors on the activity of the THBS1 promoter demonstrate that high glucose levels directly activate THBS1 transcription by a mechanism that involves the hexosamine pathway and intracellular glycosylation.
Increased flux through the hexosamine pathway exerts an effect on gene expression by increasing the availability of UDP-N-acetylglucosamine, a critical precursor of O-glycosylation of proteins (36, 45, 48 -51). Based on our findings in the present study, we speculate that glycosylation of a cytosolic signaling protein or a transcription factor mediates the synthesis of TSP-1 in response to high glucose levels.
Previous reports by Wang et al. (26,27) have also described the regulation by glucose of TSP-1 expression. These studies, performed in mesangial cells, showed that upstream stimulatory factors (USFs) 1 and 2 bound to an 18-bp sequence in the TSP1 promoter to stimulate TSP-1 transcription. Downstream steps in this pathway were shown to include protein kinase C and p38 mitogen-activated protein kinase. Our findings do not contradict, but rather complement the studies of Wang et al. (26,27). At the present time we do not know how and whether the hexosamine metabolic pathway that we describe interacts with the protein kinase C pathway established by Wang et al. (26,27). Neither USF1 nor USF2 is glycosylated in HASMC (data not shown). Although USF-2 is not glycosylated, USF-2 levels depend on the activation of the hexosamine pathway (52). Alternatively, USFs may require an unidentified transcriptional co-regulator that may be glycosylated. In addition, cytoplasmic proteins are also subject to modification by glycosylation. One such target is endothelial nitric-oxide synthase that has previously been shown to regulate TSP-1 expression (53,54). Furthermore, the signaling pathways and molecular mechanisms employed by HASMC and mesangial cells to regulate TSP-1 expression could differ in some respects since mesangial cells are specialized pericytes that share some, but not all of the properties and functions of HASMC.
Thus, glucose may up-regulate TSP-1 gene transcription through either a glycosylation-mediated modification of a cytoplasmic protein or directly through glycosylation of a promoter binding nuclear factor. Consistent with our suggestion that glycosylation of a nuclear protein could result in increased transcription of TSP-1 gene, the treatment of HASMC with different sugars resulted in a dramatically increased glycosylation of intracellular cytosolic and nuclear proteins as early as 1 h after the start of the stimulation, and this effect persisted for up to 48 h, in parallel with the time course of increase in TSP-1 levels (29). Evidence suggests that glycosylation of cytosolic or nuclear proteins is complex and may comprise a combination of glycosylation events, involving both O-and N-linked modifications (41,42,55). The hexosamine pathway and subsequent addition of O-linked N-acetylglucosamine are required for THBS1 up-regulation. However, this pathway may function in accord with another glycosylation event(s) induced by mannose or galactose. Indeed, the sugar moieties detected in glycosylated intracellular proteins are complex (56).
Vascular SMC migration and proliferation represent key events in the development of atherosclerotic lesions. SMC from diabetic individuals exhibit increased proliferation, adhesion, and migration (57). The effect of TSP-1 on SMC proliferation has been well documented in multiple reports (17,58). Stimulation of SMC by increased levels of TSP-1 in the diabetic vessel wall may provide an explanation for the enhanced proliferation of SMC. Thus, TSP-1 may function as a link between hyperglycemia and accelerated restenosis and atherosclerosis in diabetics, and the intracellular metabolic pathways up-regulating THBS1 expression may control SMC proliferation. We have shown in the present study that, whereas inhibitors of GFAT and glycosylation inhibitors that prevent up-regulation of THBS1 efficiently block HASMC proliferation induced by high glucose, metabolites of the hexosamine pathway such as glucosamine and agents that increase intracellular glycosylation (e.g. PUGNAc) stimulate HASMC proliferation. In addition, the use of an anti-TSP-1 antibody and TSP-1-targeted siRNA, both of which block SMC proliferation induced by high glucose, directly confirm the role of TSP-1 as a link between hyperglycemia and enhanced proliferation of SMC.
In summary, we have identified the hexosamine pathway as a metabolic pathway that is activated by high glucose in HASMC and mediates increased TSP-1 synthesis through an increase in transcription of the THBS1 gene. This pathway also controls increased proliferation of cultured HASMC in response to high glucose levels.