The Receptor for Advanced Glycation End Products Is Induced by the Glycation Products Themselves and Tumor Necrosis Factor-α through Nuclear Factor-κB, and by 17β-Estradiol through Sp-1 in Human Vascular Endothelial Cells*

The binding of advanced glycation end products (AGE) to the receptor for AGE (RAGE) is known to deteriorate various cell functions and is implicated in the pathogenesis of diabetic vascular complications. Here we show that AGE, tumor necrosis factor-α (TNF-α), and 17β-estradiol (E2) up-regulated RAGE mRNA and protein levels in human microvascular endothelial cells and ECV304 cells, with the mRNA stability being essentially invariant. Transient transfection experiments with human RAGE promoter-luciferase chimeras revealed that the region from nucleotide number −751 to −629 and the region from −239 to −89 in the RAGE 5′-flanking sequence exhibited the AGE/TNF-α and E2 responsiveness, respectively. Site-directed mutation of an nuclear factor-κB (NF-κB) site at −671 or of Sp-1 sites at −189 and −172 residing in those regions resulted in an abrogation of the AGE/TNF-α- or E2-mediated transcriptional activation. Electrophoretic mobility shift assays revealed that ECV304 cell nuclear extracts contained factors which retarded the NF-κB and Sp-1 elements, and that the DNA-protein complexes were supershifted by anti-p65/p50 NF-κB and anti-Sp-1/estrogen receptor α antibodies, respectively. These results suggest that AGE, TNF-α, and E2 can activate the RAGE gene through NF-κB and Sp-1, causing enhanced AGE-RAGE interactions, which would lead to an exacerbation of diabetic microvasculopathy.

accumulate in various tissues at an extremely accelerated rate under a diabetic state, and are implicated in the development of diabetic vascular complications, e.g. retinopathy and nephropathy (1). We have shown previously that AGE exert their effects on endothelial cells and pericytes, the constituents of microvessels, through interactions with a cell-surface receptor for AGE (RAGE); AGE stimulate the growth of microvascular endothelial cells through an induction of vascular endothelial growth factor (VEGF) leading to angiogenesis on the one hand (2), and inhibit prostacyclin production and stimulated plasminogen activator inhibitor-1 synthesis by the endothelial cells on the other (3). AGE exhibit a growth inhibitory action on pericytes (4), which would lead to pericyte loss, the earliest histological hallmark in diabetic retinopathy (5).
RAGE belongs to the immunoglobulin superfamily of cell surface molecules (6,7). It is expressed in multiple tissues (8) and interacts with various ligands including AGE (9,10). The engagement of RAGE by AGE has been reported to induce cellular oxidant stress, activating the transcription factor nuclear factor-B (NF-B) (11,12), resulting in the perturbation of a variety of vascular homeostatic functions (9,10). AGE-RAGE interaction thus has been thought to play a central role in the development of diabetic vasculopathy. To determine how the RAGE gene is regulated under a diabetic state is, therefore, important for clarifying the pathogenesis of diabetic complications as well as for understanding the physiological roles of RAGE.
It has been reported that AGE-rich blood vessels show enhanced RAGE immunoreactivity (13); this implies the possibility that AGE themselves may up-regulate the RAGE expression. Among cytokines, tumor necrosis factor-␣ (TNF-␣) is thought to be involved in the development of diabetes (14). Evidence has accumulated that serum TNF-␣ levels are increased in non-insulin-dependent diabetes mellitus (15)(16)(17) and that TNF-␣ can activate the NF-B pathway (18 -20). Recently, NF-B has been reported to play a role in the basal and lipopolysaccharide-induced expression of the RAGE gene (21). It has also been reported that diabetic vasculopathy is often aggravated during pregnancy, probably due to the increased level of serum estrogen (22)(23)(24). From these observations, the possibility that TNF-␣ and estrogen worsen the diabetic complications through the induction of RAGE gene expression should also be considered.
In the present study, we thus examined the effects of AGE, TNF-␣, and 17␤-estradiol (E 2 ) on RAGE gene expression and found that the three agents are capable of up-regulating the RAGE mRNA and protein levels in human microvascular endothelial cells. On the other hand, non-glycated BSA, other cytokines, and an anti-estrogen did not affect the RAGE mRNA levels. mRNA stability and promoter assays demonstrated that the induction was at the transcriptional level, and that AGE and TNF-␣ induced the RAGE gene through an activation of NF-B while E 2 induced the gene through Sp-1.
Preparation of AGE-BSA-AGE-BSA was prepared by incubating BSA with 0.5 M glucose at 37°C for 6 weeks under sterile conditions as described previously (2). After unincorporated sugars were removed by dialysis against phosphate-buffered saline (PBS), glucose-modified higher molecular weight materials were used as AGE-BSA. Control non-glycated BSA was incubated under the same conditions except for the absence of glucose. The concentration of AGE-BSA and control BSA were determined by the method of Bradford (26).
Measurement of RAGE mRNA by Quantitative Reverse Transcription-PCR (RT-PCR) Method-Poly(A) ϩ RNAs were isolated from subconfluent cultures of HMVEC or ECV304 cells incubated under various conditions, using a Quick Prep Micro mRNA purification kit (Amersham Pharmacia Biotech, Buckinghamshire, United Kingdom), and analyzed by RT-PCR with a GeneAmp RNA PCR kit (Perkin-Elmer, Foster City, CA) as described previously (27). Oligodeoxyribonucleotide primers and probes for human RAGE, VEGF, and ␤-actin mRNA were the same as described previously (4,28). The amounts of poly(A) ϩ RNA templates and cycle numbers for amplification were chosen in quantitative ranges where reactions proceeded linearly as described previously (28,29); 30 ng of templates and 30 cycles were chosen for amplifying human RAGE mRNA, 30 ng and 40 cycles for human VEGF mRNA, and 30 ng and 20 cycles for ␤-actin mRNA. The fragments amplified with PCR were sequence-verified on both strands by the chain termination method (30). Five-l aliquots of each RT-PCR reaction mixture were electrophoresed on a 2% agarose gel and transferred to a Hybond-N ϩ nylon membrane (Amersham Pharmacia Biotech), and the membrane was hybridized with the respective probes, which had been 32 P-end-labeled with polynucleotide kinase (27). Signal intensities of hybridized bands were measured by a BAS 1000 BioImage analyzer (Fuji Photo Film Co. Ltd., Hamamatsu, Japan).
Analysis of RAGE mRNA Stability-HMVEC or ECV304 cells were treated with TNF-␣, E 2 , or AGE-BSA for 4 h, and further cultured in the presence of 10 g/ml of actinomycin D (Sigma) for various time periods. Total RNAs were isolated from the cultures with Isogen (Nippon Gene, Toyama, Japan) according to the method described by Chomczynski and Sacchi (31), and analyzed by the quantitative RT-PCR method described above. The amounts of total RNA templates and cycle numbers for amplification were chosen in quantitative ranges; 300 ng of templates and 35 cycles were chosen for amplifying human RAGE mRNA, and 300 ng and 20 cycles for ␤-actin mRNA.
Western Blot Analysis-Subconfluent cultures of HMVEC or ECV304 cells were incubated with TNF-␣, E 2 , or AGE-BSA for 24 h. After the incubation, cells were washed with cold PBS, scraped off in cold PBS, and pelleted by centrifugation at 300 ϫ g at 4°C for 5 min. The cells were lysed immediately by sonication in SDS-polyacrylamide gel electrophoresis (PAGE) sample buffer (62.5 mM Tris-HCl (pH 6.8), 2% SDS, 5% 2-mercaptoethanol, 10% glycerol, and 0.002% bromphenol blue) and boiled at 95°C for 5 min. Twelve g of the cell lysates were resolved by SDS-PAGE (12.5%), and then transferred onto a nitrocellulose membrane (Amersham Pharmacia Biotech). The membrane was treated with the anti-RAGE polyclonal antibodies, and the immunoreactive bands were visualized with an ECL detection system (Amarsham Pharmacia Biotech) as described previously (32). Signal intensities of the resultant bands were determined by densitometry using BIO-PROFIL 1-D (version 5.08) software (Vilber Lourmat Biotechnology, Marne La Vallée, France).
Construction of the RAGE Promoter-Luciferase Chimeras-The chimeric genes for transfection experiments were constructed by ligating the 5Ј-flanking regions of differing lengths of the human RAGE gene upstream of the luciferase gene in a pGL3-basic vector (Promega Corp., Madison, WI) (see Fig. 5A). The genomic DNA fragments of the human RAGE gene were amplified by PCR using a cosmid named KS71 (33) as a template, which was kindly provided by Professor Toshimichi Ikemura (National Institute of Genetics, Shizuoka, Japan). The PCR primers employed in the amplification reactions are shown in Table I. Furthermore, we constructed two additional chimeric genes. The DNA fragments containing exons 1-11 plus introns 1-10 or the 3Ј-flanking region of the human RAGE gene were amplified by PCR using the

5Ј-GAACGCGT TACCTCGGAGGGAGTTTCTG
Ϫ1528 to Ϫ1508 pGL-4 Forward primers pGL-el-11 5Ј-GAGTCGAC AAGAGGTGGAACAAGAGAAG ϩ6318 to ϩ6298 cosmid KS71 and specific primers shown in Table I, and were ligated to downstream of the luciferase gene in the pGL3-Basic vector that had carried the longest fragment of the 5Ј-flanking region (see Fig. 5A). All the fragments obtained were sequence-verified.
Transfection Experiments and Luciferase Assay-ECV304 cells or HMVEC (2 ϫ 10 5 cells) were plated into the wells of six-well tissue culture plates (Becton Dickinson Labware) 1 day before transfection. For ECV304 cells, the DNA/cationic lipid mixture for transfection was composed of test plasmid (2 g), pRL-SV40 vector (0.5 g) (Promega) that served as an internal control to normalize luciferase activities, TransFast TM transfection reagent (7.5 l) (Promega), and 1 ml of the experimental medium. Cells were exposed to the DNA/cationic lipid mixture for 2 h, then received 5 ml of the FBS-containing medium and were further incubated for 36 h at 37°C. For transfection of HMVEC, test plasmid (2 g) and pRL-CMV vector (1 g) (Promega) were added to a tube containing 100 l of Opti-MEM I (Life Technologies, Inc.). Lipofectin reagent (6 l) (Life Technologies, Inc.) was added to another tube containing 100 l of Opti-MEM I. The plasmid DNA and lipofectin reagent were then mixed together. After incubation at room temperature for 30 min, DNA-liposome complex was diluted with 800 l of Opti-MEM I. Cells were exposed to the mixture for 8 h, then received 2 ml of the FBS-containing medium and were further incubated for 36 h at 37°C. After the incubation, the cells were treated with TNF-␣, E 2 , or AGE-BSA for 8 h in the experimental medium. Luciferase activities were measured using a Dual-Luciferase reporter assay system (Promega) according to the manufacturer's protocol with a luminometer (Fluoroskan Ascent FL version 2.2.4, Labsystems, Helsinki, Finland).
Preparation of Nuclear Extracts from Cultured Cells-Nuclear extracts were prepared essentially as described by Schreiber et al. (34). Briefly, ECV304 cells (2 ϫ 10 6 ) were plated onto 75-cm 2 tissue culture flasks (Becton Dickinson Labware) in the complete medium and left for 24 h at 37°C. The cells were further incubated in the experimental medium at 37°C for 12 h and then treated with TNF-␣, E 2 , or AGE-BSA for 4 h. After the treatment, the cells were washed twice with ice-cold PBS, scraped off in PBS (10 ml), and pelleted by centrifugation at 3,000 rpm in a Beckman GH3.7 roter at 4°C for 5 min. The pelleted cells were resuspended in 0.4 ml of ice-cold buffer A (10 mM HEPES (pH 7.9), 10 mM KCl, 0.1 mM EDTA, 1 mM dithiothreitol, and 0.5 mM phenylmethylsulfonyl fluoride) and placed on ice for 15 min. After the addition of 25 l of 10% Nonidet P-40, the suspension was vortexed for 15 s and centrifuged at 13,000 rpm in a Hitachi T15S roter at 4°C for 30 min. The resultant nuclear pellets were washed with buffer A and resuspended in 0.1 ml of a solution containing 20 mM HEPES (pH 7.9), 400 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride by constant agitation for 15 min at 4°C. The nuclear lysates were centrifugated at 13,000 rpm at 4°C for 10 min, and the supernatants were collected as nuclear extracts. The protein concentrations of the nuclear extracts were determined by the method of Bradford (26).
Electrophoretic Mobility Shift Assay and Supershift Assay-The wild type and mutant double-stranded oligodeoxyribonucleotides encompassing the NF-B site (nucleotide numbers Ϫ671 to Ϫ663) or the two Sp-1 sites (Ϫ189 to Ϫ181 and Ϫ172 to Ϫ166) were prepared. Their sequences were 5Ј-AGAGTGGGGAACCCCTCCCA-3Ј and 5Ј-AGAGT-GGGGTTCCCCTCCCA-3Ј (Ϫ677 to Ϫ658) as the wild and mutated NF-B elements, respectively, and 5Ј-CCCAGAGGCTGGTAGTAC-CCAGGGGTGGGGTGA-3Ј and 5Ј-CCCAGAAACTGGTAGTACCCAG-GAATGGGGTGA-3Ј (Ϫ193 to Ϫ161) as the wild and mutated Sp-1 elements, respectively (mutated sites are indicated by underlines). Twenty-five fmol of wild-type oligodeoxyribonucleotides, which had been 32 P-end-labeled with polynucleotide kinase (27), were incubated with 5 g of nuclear extracts at room temperature for 30 min. Samples were then loaded onto 6% polyacrylamide gels and run in 0.2ϫ Tris borate/EDTA electrophoresis buffer at 10 V/cm for 2-3 h. The gels were dried and autoradiographed at Ϫ80°C overnight. For competition assay, nuclear extracts were first incubated with a 50-fold excess of unlabeled wild-type or mutant oligodeoxyribonucleotides at room temperature for 15 min, and then incubated with the labeled wild-type probe for 30 min under the same conditions as described above. For supershift assays, antibodies to NF-B p65, NF-B p50, Sp-1, estrogen receptor (ER) ␣, or ER␤ (Santa Cruz Biotechnology Inc., Santa Cruz, CA) were added to the nuclear extracts and incubated at 4°C for 12 h. The antibody-treated nuclear extracts were subsequently incubated with the labeled oligodeoxyribonucleotides for 30 min under the same conditions as described above.
Statistical Analysis-Paired t tests and one-way analysis of variance (ANOVA) with Tukey's range tests were used to test for significant differences between groups. All experiments were carried out at least three times.

TNF-␣, E 2 , and AGE-BSA Increased the RAGE mRNA and
Protein Levels in HMVEC and ECV304 Cells-To examine the effects of TNF-␣, E 2 , and AGE-BSA on the RAGE mRNA level in HMVEC, poly(A) ϩ RNAs were isolated from cells that had been exposed to various concentrations of these agents for 4 h, and analyzed by the quantitative RT-PCR method. As shown in Fig. 1A, TNF-␣, E 2 , and AGE-BSA increased the RAGE mRNA levels in dose-dependent manners. The extents of induction and the peak concentrations were about 3-fold at 100 ng/ml TNF-␣, 10 nM E 2 , and 50 g/ml AGE-BSA, respectively, when normalized by ␤-actin mRNA-derived signals used as an internal control. Next, we examined the time course of the RAGE mRNA induction. HMVEC were treated for various time periods with TNF-␣, E 2, and AGE-BSA at their most effective doses. As shown in Fig. 1B, the mRNA levels began to increase at 2 h and reached a maximum at ϳ8 h after the addition of either of the three agents. On the other hand, exposure of HMVEC to TGF-␤1 (10 ng/ml), IFN-␥ (165 ng/ml), and nonglycated BSA (50 g/ml) for 4 h did not affect the RAGE mRNA levels (Fig. 1C). 4-OH tamoxifen (10 nM), an anti-estrogen, abolished the E 2 -induced RAGE mRNA induction (Fig. 1C).
We next determined the RAGE mRNA levels in ECV304 cells, an immortalized cell line derived from human umbilical vein endothelial cells (25). This cell line exhibited the same responsiveness to AGE-BSA with respect to VEGF induction ( Fig. 2A) as did HMVEC (2), the most effective dosage of AGE-BSA being 50 g/ml in both cultures. The extent of induction was about 4-fold. When exposed to TNF-␣ (100 ng/ml), E 2 (10 nM), or AGE-BSA (50 g/ml) for 4 h, the RAGE mRNA levels in the ECV304 cells were also increased about 2-fold compared with those in the control unexposed cells (Fig. 2B).
We next examined whether the increase in RAGE mRNA was actually followed by an increase in RAGE proteins in HMVEC and ECV304 cells. The cells were treated with TNF-␣, E 2 , or AGE-BSA for 24 h, and subjected to Western blot analysis with anti-RAGE polyclonal antibodies. As shown in Fig. 3, one major immunoreactive band was marked at 46 kDa in either HMVEC (Fig. 3A) or ECV304 cells (Fig. 3B), being consistent with our previous report (3), and its intensity was increased by the treatment with the three agents. The RAGE protein levels in HMVEC and ECV304 cells treated with TNF-␣, E 2 , or AGE-BSA were about 2-fold higher than the basal levels. The results indicated that ECV304 cells retained the ability to respond to those agents as did primary cultured endothelial cells, and subsequently we mainly used this cell line to examine the mechanisms of RAGE gene induction.
Effects of TNF-␣, E 2 , and AGE-BSA on RAGE mRNA Stability in HMVEC and ECV304 Cells-We next determined the RAGE mRNA stability in HMVEC and ECV304 cells exposed or not exposed to the three agents to determine which step of gene expression accounted for the increase in RAGE mRNA levels. The cells were incubated in the presence or absence of TNF-␣, E 2 , or AGE-BSA for 4 h, then incubated with actinomycin D for various time periods, and underwent quantitative RT-PCR analyses. As shown in Fig. 4 (A and B), the half-lives of RAGE mRNA in TNF-␣-, E 2 -, and AGE-BSA-treated or untreated HMVEC and ECV304 cells were calculated from the RAGE and ␤-actin mRNA-derived signals to be between 2.1 and 2.8 h, and there was no statistically significant difference among them. The results suggested that the TNF-␣-, E 2 -, and AGE-BSA-induced increase in RAGE mRNA was achieved at the transcriptional level.
Identification of the cis-Elements Responsive to TNF-␣, E 2 , and AGE-BSA in the RAGE Promoter-To confirm whether TNF-␣, E 2 , and AGE-BSA did induce the RAGE gene transcription and, if so, to delimit the regions involved in such transcriptional activations, a series of chimeric 5Ј-deletion promoterluciferase reporter constructs were prepared. Schematic representations of the constructs are shown in Fig. 5A. ECV304 cells were transiently transfected with the constructs, and the effects of TNF-␣, E 2 , and AGE-BSA on the luciferase activity in the transfected cells were determined. pGL-1 carried the longest 5Ј-flanking region of the RAGE gene (1689 base pairs upstream of the transcription start site; Ref. 14), and when the pGL-1-transfected cells were exposed to TNF-␣ (Fig. 5B), E 2 (Fig. 5C), or AGE-BSA (Fig. 5D), the promoter activities (closed columns) increased significantly (1.5-to ϳ 2-fold) compared with those in unexposed cells (open columns). The same con- were then isolated and analyzed by RT-PCR as described under "Experimental Procedures." B, HMVEC were treated with TNF-␣ (100 ng/ml), E 2 (10 nM), or AGE-BSA (50 g/ml) for the indicated time periods, and assayed for RAGE mRNA levels. C, HMVEC were treated with TNF-␣ (100 ng/ml ϭ 5 ϫ 10 3 units/ml), TGF-␤1 (10 ng/ml), IFN-␥ (165 ng/ml ϭ 5 ϫ 10 3 units/ml), E 2 (10 nM), 4-OH tamoxifen (10 nM), E 2 plus 4-OH tamoxifen, AGE-BSA (50 g/ml), or non-glycated BSA (50 g/ml) for 4 h, and assayed for RAGE mRNA levels. The TGF-␤1 and IFN-␥ concentrations employed were those that caused near maximal biological effects (57,58). Graphs indicate the quantification of the RAGE mRNA levels. Intensities of the RAGE mRNA signals were normalized with those of ␤-actin mRNA-derived signals, and are related to the value of the control without additives. Columns and bars indicate means and standard deviations, respectively, in three independent experiments. *, p Ͻ 0.05; **, p Ͻ 0.01 (versus control). centration of non-glycated BSA did not induce the luciferase activity in the transfected cells (data not shown). Deletion of the 5Ј-flanking region of the RAGE gene to Ϫ751 (pGL-2 to -5) did not affect the TNF-␣-and AGE-induced luciferase expression, but deletion to Ϫ629 (pGL-6) abolished the induction (Fig.  5, B and D). In contrast, the construct with a deletion to Ϫ239 (pGL-7) still retained the E 2 -induced luciferase expression but deletion to Ϫ89 (pGL-8) abolished the induction (Fig. 5C). The Ϫ751 to Ϫ629 region contained an NF-B site (Ϫ671 to Ϫ663), and the Ϫ239 to Ϫ89 region contained two Sp-1 sites (Ϫ189 to Ϫ181 and Ϫ172 to Ϫ166). The results thus suggested that the NF-B site and the Sp-1 sites might be required for the TNF-␣/AGEand E 2 -induced activation of the RAGE gene promoter, respectively.

FIG. 2. AGE-induction of VEGF mRNA and effects of TNF-␣, E 2 , and AGE-BSA on RAGE mRNA levels in ECV304 cells. A,
We also tested additional constructs that carried genomic fragments containing exons 1-11 and introns 1-10 or the 3Јflanking region (pGL-e1-11 and pGL-3Јfl, respectively) in addition to the pGL-1 5Ј-promoter region. As shown in Fig. 5E, when the pGL-e1-11-or pGL-3Јfl-tranfected cells were stimulated by TNF-␣ (closed columns), E 2 (lattice columns), or AGE-BSA (dotted columns), the extents of luciferase induction were almost indistinguishable from those in the pGL-1-transfected cells. The results indicated that there were neither stimulatory nor silencing elements in the exon/intron or 3Ј-flanking region of the RAGE gene that could affect its responsiveness to TNF-␣, E 2 , or AGE-BSA.
To determine the role of the NF-B site at Ϫ671 to Ϫ663 in the TNF-␣ and AGE activation of the RAGE promoter, site-directed mutagenesis was performed at that site (pGL-5 NF-B2m) (Fig. 6A). When luciferase activities were assayed in cells transfected with the mutant, the inducibility by both TNF-␣ and AGE-BSA was found to be totally abolished (Fig.  6B). Similarly, we also determined the role of the two Sp-1 sites at Ϫ189 to Ϫ181 and Ϫ172 to Ϫ166 in the E 2 -dependent transcriptional activation. We altered each of the two Sp-1 sites (pGL-7 Im and pGL-7 IIm) or both sites (pGL-7 ImIIm) (Fig.  6A), and the E 2 effect on luciferase activities in the mutant transfectants was tested (Fig. 6C). The E 2 inducibility of

FIG. 4. RAGE mRNA stability. HMVEC (A) and ECV304 cells (B)
were treated with TNF-␣ (100 ng/ml), E 2 (10 nM), or AGE-BSA (50 g/ml) for 4 h, and further cultured in the presence of 10 g/ml actinomycin D for the indicated time periods. RAGE mRNA levels were determined by RT-PCR with total RNAs extracted from the cells. Control, cultures without the additives but with actinomycin D. Graphs depict quantitative representations of the RAGE mRNA levels in HM-VEC (A) and ECV304 cells (B). Data were normalized by ␤-actin mRNA-derived signals. Essentially the same results were obtained in three independent experiments. luciferase activities was decreased in pGL-7 Im-and pGL-7 IIm-transfected cells and totally abolished in pGL-7 ImIImtransfected cells. The results indicated that the NF-B binding site and the two Sp-1 binding sites were essential for the TNF-␣/AGE-and E 2 -induced activation of the RAGE gene, respectively.
TNF-␣/AGE-BSA and E 2 Induced the Nuclear Protein-DNA Complex Formation on the NF-B and Sp-1 Sites-To further characterize the roles of the NF-B and Sp-1 sites in the regulation of RAGE gene expression, we conducted electrophoretic mobility shift assay using synthetic oligodeoxyribonucleotides corresponding to the putative NF-B binding site and to the two Sp-1 binding sites, and nuclear extracts (5 g) from ECV304 cells. As shown in Fig. 7 A and B, lane 1). In E 2 -treated cells, the binding activities to the Sp-1 sites also increased compared with the basal conditions (Fig. 7C). In each case, the specificity of the gelretarded band was demonstrated by competition with a 50-fold excess of unlabeled wild-type or mutant oligodeoxyribonucleotides, namely preincubation with wild-type probes resulted in decreased binding to the NF-B site (Fig. 7, A and B, lane 3)
Next, we examined by supershift assays which members of the NF-B family were responsible for the stimulation of the RAGE gene expression by TNF-␣ and AGE-BSA. When the nuclear extracts from the cells exposed to TNF-␣ (Fig. 7D) or AGE-BSA (Fig. 7E) were incubated with either anti-p65 or anti-p50 antibody, a more slowly migrating band (Fig. 7, D and E, arrow) newly appeared with a concomitant decrease in the original complex (Fig. 7, D and E, closed arrowhead). We also examined which factors were involved in the E 2 induction of the RAGE gene. When the nuclear extracts from the cells treated with E 2 (Fig. 7F) were incubated with either anti-Sp-1 or anti-ER␣ antibody, the original band DNA-protein complex was supershifted upward (Fig. 7F, lanes 2 and 3), whereas anti-ER␤ antibody gave no effect (Fig. 7F, lane 4). The results indicated that the DNA-binding complex for the NF-B site was composed mainly of p50 and p65 members of the NF-B family, and that the DNA-binding complex for the Sp-1 sites was composed mainly of Sp-1 and ER␣. pGL-5 NF-B2m, control versus TNF-␣, AGE-BSA, not significant. C, pGL-7 and pGL-7 Im; control versus E 2 , p Ͻ 0.01; pGL-7 IIm and pGL-7 ImIIm; control versus E 2 , not significant.

FIG. 7. Electrophoretic mobility shift (A-C) and supershift (D-F) assays of the NF-B and Sp-1 binding sites in the RAGE promoter region.
A-C, 5 g of nuclear extracts from ECV304 cells treated for 4 h with TNF-␣ (100 ng/ml) (A), AGE-BSA (50 g/ml) (B), or E 2 (10 nM) (C) were incubated with 25 fmol of the respective 32 P-endlabeled double-stranded oligodeoxyribonucleotides for 30 min as described under "Experimental Procedures." For the competition assay, a 50-fold excess of unlabeled wild-type or mutant oligodeoxyribonucleotides was added to the extracts and preincubated for 15 min prior to the incubation with the 32 P-end-labeled probe. The formation of the protein/ DNA complexes was analyzed by electrophoresis on a 6% polyacrylamide gel. Closed arrows indicate NF-B/DNA and Sp-1/DNA complexes. Open and closed arrowheads indicate nonspecific binding to the probe and free probe, respectively. Essentially the same results were obtained in four independent experiments. D-F, for the supershift assay, the indicated antibodies were added to the extracts and preincubated at 4°C for 12 h prior to incubation with the 32 P-end-labeled probe. cells, we performed luciferase assays in HMVEC using pGL-5 and its mutants for the TNF-␣/AGE-BSA-induced transcriptional activation, and pGL-7 and its mutants for the E 2 -induced activation. As shown in Fig. 8A, TNF-␣ and AGE-BSA induced luciferase expression ϳ3-fold in pGL-5-transfected cells, the extent of induction being comparable with those of the RAGE mRNA induction in HMVEC (Fig. 1). The inducibility by both TNF-␣ and AGE-BSA was found to be totally abolished in pGL-5 NF-B2m-or pGL-6-transfected cells (Fig. 8A). E 2 induced luciferase expression about 2.6-fold in pGL-7-transfected cells (Fig. 8B). This extent of induction was also comparable with that of the RAGE mRNA induction by E 2 in HMVEC (Fig.  1). The E 2 inducibility of luciferase activities was totally abolished in pGL-7 ImIIm-or pGL-8-transfected cells (Fig. 8B). DISCUSSION We (2)(3)(4)35) and others (36 -39) have shown previously that interactions between AGE and RAGE cause phenotypic changes in microvascular endothelial cells and pericytes that are characteristic of diabetic vasculopathy. In this paper, we have demonstrated for the first time that AGE, the RAGE ligand itself, TNF-␣, and E 2 specifically up-regulate RAGE mRNA and protein levels in human vascular endothelial cells, and that this process is mediated by two distinct nuclear complexes, namely p65/p50 NF-B and Sp-1/ER␣.
It has been reported that AGE-rich vasculature exhibits enhanced RAGE immunoreactivity in the sites of diabetic microvascular injury (13). This finding suggested that AGE them-selves could activate the RAGE gene expression, and that increased RAGE might further transduce AGE signals within microvascular cells. The present study has clearly shown that AGE-BSA did, but non-glycated BSA did not, up-regulate the RAGE gene expression and that this was achieved at the level of transcription, because RAGE mRNA half-lives were unchanged by the treatment with AGE (Fig. 4), and because AGE elicited the expression of the reporter gene linked to the 5Ј promoter region of the RAGE gene (Fig. 5). The experiments with the RAGE promoter-reporter gene constructs demonstrated the presence of an AGE-responsive element in the Ϫ751 to Ϫ629 region of the human RAGE gene, and that an NF-B site residing at Ϫ671 conferred the responsiveness to AGE on the RAGE gene (Figs. 5, 6, and 8). Further, the AGE-dependent formation of the NF-B element-p65/p50 complex was demonstrated (Fig. 7). These results are considered to be consistent with the recent observations by others (11,12) that AGE engagement of RAGE induces cellular oxidant stress, thereby activating the transcription factor NF-B. The fact that AGE themselves induce the RAGE gene appears to be important when considering the mechanisms of development of diabetic vascular complications. Such a positive feedback loop in the diabetic state may exacerbate diabetic vasculopathy, exemplified by retinopathy and nephropathy.
We also demonstrated that TNF-␣ is another inducer of the RAGE gene (Figs. 1-3). TNF-␣ is known to be overexpressed in adipose tissue under obese and diabetic states (15)(16)(17) and to cause insulin resistance, the central and early component of non-insulin-dependent diabetes mellitus (18). TNF-␣ affects not only insulin sensitivity by suppressing insulin-stimulated tyrosine phosphorylation of insulin receptor substrate-1 (40) but also cell survival by NF-B activation (20 -22). We thus propose that an increased TNF-␣ level in non-insulindependent diabetes mellitus patients may worsen diabetic vasculopathy via RAGE gene induction. The TNF-␣-induced stimulation of human RAGE gene was found to be also transcriptional and to be achieved by the same NF-B element and binding complex (Figs. 5-8) as was the AGE stimulation. Recently, Li et al. (14) reported that the RAGE gene was activated by lipopolysaccharide via NF-B sites at Ϫ671 and Ϫ467 in bovine aortic endothelial cells and rat vascular smooth muscle cells, and that both sites were required for full activation by lipopolysaccharide. Our results indicated that only one NF-B site at Ϫ671 was required for RAGE gene induction by AGE-BSA and TNF-␣ in human microvascular endothelial cells (Figs. 5, 6, and 8). The discrepancy may be due to differences in the species or ligand specificity of the gene activation. The results suggest that the factors that have the ability to activate the NF-B pathway can induce RAGE gene expression and have the potential to aggravate the AGE-mediated diabetic complications.
Pregnancy is known to worsen diabetic retinopathy (22)(23)(24). Schocket et al. (41) showed that a decrease in the retinal volumetric blood flow in diabetic patients during pregnancy might exacerbate retinal ischemia and aggravate retinopathy. Suzuma et al. (42) demonstrated that E 2 at the concentration often observed during pregnancy (ϳ10 nM) stimulates VEGFdependent angiogenesis through the up-regulation of VEGF receptor-2 expression. However, the mechanisms underlying the adverse effects of E 2 on diabetic complications are not yet fully understood. In this study, we demonstrated that E 2 is an alternative inducer of the RAGE gene in human endothelial cells, enhancing its transcription at the concentration of 10 nM (Figs. 1-3). An anti-estrogen, 4-OH tamoxifen, totally abolished the E 2 -induced RAGE mRNA induction in HMVEC, while itself caused no change in the RAGE mRNA level ( control versus E 2 , p Ͻ 0.01; pGL-7 ImIIm; control versus E 2 , not significant; pGL-8; control versus E 2 , not significant. 1C). This was regarded as an indication that E 2 may act on RAGE gene through an estrogen receptor. The RAGE promoter does not contain any classical estrogen-responsive element (43) but contains several GC-rich boxes, which can bind to the transcription factor Sp-1 (14,33,44). Clearly, E 2 utilizes a device that is different from those employed by AGE and TNF-␣ in the induction of RAGE gene. Two Sp-1 binding sites at Ϫ189 and Ϫ172 and an Sp-1/ER complex were involved in the E 2 activation, and full E 2 responsiveness required both cis-acting elements (Figs. [5][6][7][8]. Recent studies from other laboratories have shown that an interaction of Sp-1/ER complex with GC-rich motifs in the promoter region is required for the transcriptional activation of several E 2 -responsive genes (45)(46)(47)(48)(49). The results obtained in this study suggest that the E 2 induction of RAGE may partly underlie the exacerbation of diabetic retinopathy during pregnancy.
There are many genes that are regulated by the combination of NF-B and Sp-1, and in some cases a direct interaction between NF-B protein and Sp-1 protein has been demonstrated (50,51). In the induction of the RAGE gene, however, such a direct interaction between the two factors would seem unlikely to occur because the Sp-1-mediated E 2 responsiveness was still retained in the constructs with the deletion of the more than 500-nucleotide region encompassing the NF-B element (Fig. 5).
Cytokines that did not affect the endothelial cell expression of RAGE gene included TGF-␤1 and IFN-␥ (Fig. 1C). This indicated that Smad (52) or Janus kinase-signal transducers and activators of transcription (53) systems would not be involved in the regulation of RAGE gene.
In summary, we found that AGE and TNF-␣ enhanced RAGE expression through activation of the p65/p50 complex of NF-B, and that E 2 also activated RAGE expression through activation of the Sp-1/ER complex. Although the extent was rather modest, the three agents consistently increased the RAGE mRNA and protein levels in vascular endothelial cells. Chronic exposure to AGE, TNF-␣, and/or E 2 and sustained enhancement of RAGE expression may cause a further accumulation of AGE in the vasculature, resulting in an exacerbation of AGE-RAGE-mediated vascular dysfunctions. Such mechanisms of RAGE gene activation probably have evolved to regulate functions of this multiligand receptor in various biologic processes, such as neurite formation of cortical neurons mediated by amphotericin engagement (54) and proinflammatory responses mediated by a recently discovered endogenous ligand, EN-RAGE (55). Obviously, diabetes abuses the molecular devices for the RAGE gene regulation, resulting in the formation of a vicious circle which may eventually lead to the development and progression of diabetic vasculopathy. The enhanced interaction between AGE and RAGE may further increase VEGF expression in endothelial cells and/or retinal pigment epithelial cells (2,56), resulting in angiogenesis, and also inhibit pericyte growth, leading to pericyte loss (4). Although more studies are needed to better clarify the significance of the RAGE gene activation by AGE, TNF-␣, and E 2 as well as immediate post-RAGE signaling events that lead to NF-B and Sp-1 activations, inhibition of the RAGE gene activation may become a promising target for the prevention and treatment of diabetic vascular complications.