Induction of Vascular Endothelial Growth Factor by Tumor Necrosis Factor α in Human Glioma Cells POSSIBLE ROLES OF SP-1

Abstract The expression of vascular endothelial growth factor (VEGF) has been implicated in brain tumor angiogenesis, and the promoter region for the VEGF gene contains several SP-1 and AP-1 (c-Fos and c-Jun) binding motifs. Among eight human glioma cell lines, cellular mRNA levels of transcription factors SP-1 and AP-1 (c-Fos and c-Jun) were found to be closely correlated with those of VEGF. VEGF expression appears to be highly susceptible to hypoxia or exogenous cytokines and growth factors. Of various cytokines and growth factors, basic fibroblast growth factor (bFGF), tumor necrosis factor α (TNF-α), and interleukin 1 most potently enhanced VEGF mRNA levels of a glioma cell line, U251. Incubation of the glioma cells with bFGF or TNF-α increased both VEGF and SP-1 mRNA at 30 min and c-Fos mRNA at 1-3 h, over 5-fold. Nuclear run-on assays showed an apparent increase of the transcription of the VEGF gene as well as the SP-1 gene by bFGF or TNF-α. Gel mobility shift assays demonstrated that only SP-1 binding activity was increased 1 h after exposure to bFGF or TNF-α, and also that AP-1, but not SP-1, activity was significantly activated by hypoxia. Mithramycin, an inhibitor of SP-1, at 1-10 nM inhibited activation of the VEGF gene by bFGF or TNF-α but not that by hypoxia. Western blot analysis also demonstrated an increase in cellular amounts of VEGF by TNF-α and a decrease by co-administration with mithramycin. The promoter activity of the VEGF gene, which contains five SP-1 binding sites and one AP-1 binding site but not hypoxia regulatory elements, was enhanced by bFGF or TNF-α but not by hypoxia. The chloramphenicol acetyltransferase assay with VEGF promoter deletion constructs demonstrated that four clusterized SP-1 binding sites in the proximal promoter were essential for the basal transcription and the TNF-α-dependent activation. These data indicated that the expression of the VEGF gene enhanced by bFGF or TNF-α appeared to be mediated in part through the transcription factor SP-1, suggesting a different mechanism from that for hypoxia-induced activation of the VEGF gene.

Tumor angiogenesis supports tumor enlargement, and a transition from limited to rapid tumor growth accompanies neovascularization (1)(2)(3). Malignantly transformed cells produce various growth factors or cytokines, which should balance tumor angiogenesis through paracrine control. Basic fibroblast growth factor (bFGF), 1 EGF, TGF-␣, hepatocyte growth factor, VEGF, PDGF, TNF-␣, TGF-␤, and IL-8 are supposed to be involved in tumor angiogenesis (3)(4)(5)(6). An angiogenic factor, VEGF, induces neovascularization in the chick chorioallantoic membrane and corneal bioassays in vivo, and it also acts as a potent and specific mitogen for vascular endothelial cells in vitro (7)(8)(9)(10)(11)(12). VEGF induces signaling through interaction with its receptor, Flt-1 (13)(14)(15)(16). VEGF expression is up-regulated in central nervous system tumors compared with the normal brain (17). VEGF mRNA is localized in tumor cells within human brain tumor tissue specimens by means of in situ hybridization (18,19). VEGF is expressed in human gliomas (12,19), and tumor growth is inhibited by anti-VEGF antibody (20,21). Glioma cells implanted in nude mice with endothelial cells that harbor mutant Flk-1, a VEGF receptor, fail to grow (22). In human gliomas, VEGF mRNA levels, but not TGF-␣, TGF-␤, and bFGF mRNA levels, are closely correlated with tumor angiogenesis (12,23). VEGF and bFGF induce an in vitro angiogenic response in bovine microvascular endothelial cells that is far greater than additive and that occurs with greater rapidity than the response to either cytokine alone (24). VEGF thus appears to be important for glioma growth and its neovascularization. VEGF expression is enhanced in human glioma cells (18) and retina in a mouse retinopathy model (25) under hypoxic conditions. Analysis of the VEGF promoter region reveals several potential binding sites for the transcription factors SP-1, AP-1, and AP-2 (26). Hypoxic stress induces the activation of AP-1 (27), NFB (28), and hypoxia regulatory elements (29) in various cell types. By contrast, VEGF gene expression is also enhanced by EGF, PDGF, TGF-␤, phorbol esters, and cAMP analogues (26, 30 -34). However, little is known about how such growth factor-induced activation of the VEGF gene is regulated.
Human gliomas also produce high levels of bFGF (35)(36)(37), and the expression of bFGF mRNA is increased in many human glioma cell lines (38). TGF-␣ and -␤ are often produced at high levels in human glioma and other various human tumors (38 -41). These studies suggest that many growth factors, such as bFGF, EGF, TGF-␣, PDGF, and TGF-␤, in addition to VEGF may modulate human glioma-induced angiogenesis. Our recent study indicates that overexpression of a proto-oncogene, c-fos, mediates neovascularization in the rat brain (42), suggesting a novel involvement of a proto-oncogene or transcription factor in the angiogenic process. One can expect that these various growth factors may change the expression of VEGF, resulting in stimulation of the glioma-dependent neovascularization. In this study, we investigated which growth factor or cytokine was responsible for the enhanced expression of the VEGF gene and also which transcription factor was involved in the enhanced expression of the VEGF gene in response to growth factors and cytokines in human glioma cells.
Cell Culture-Eight glioma cell lines were derived from patients with glioma or astrocytoma. Their properties were as described previously (12,38,45). These cell lines were cultured in DMEM containing 10% FBS, 100 units/ml penicillin, and 60 g/ml kanamycin as reported (38,45).
Hypoxic Conditions-Cells were incubated in Anaerocult A mini (Merck, Darmstadt, Germany) to produce hypoxic conditions. The Anaerocult A mini contains components that chemically bind oxygen quickly and completely, creating an oxygen-free milieu and a CO 2 atmosphere. The reaction zone of Anaerotest strips were moistened with water and placed onto the lids of Petri dishes. Anaerocult A mini was placed into an incubation bag and moistened with 8.0 ml of water. The dishes were immediately placed in the incubation bag, which was sealed with Anaeroclips and incubated at 37°C.
Stability of VEGF mRNA-To determine the stability of VEGF mRNA, U251 cells were incubated in DMEM containing 1% FBS for 4 h, incubated with or without 10 ng/ml bFGF or 100 units/ml TNF-␣ for 1 h. Thereafter, the cells were incubated with 1 g/ml actinomycin D, and total RNAs were extracted at 0, 1, 3, 6, and 12 h later (48). Total RNAs were analyzed by hybridization with a 32 P-labeled cDNA probe.
Nuclear Run-on Transcriptional Assay-One dish (Nunc, Roskilde, Denmark) of quiescent cells or those stimulated with bFGF or TNF-␣ for 1 h was used for nuclei preparation for each assay. Nuclei were isolated from cells as described previously (46,49). The 32 P-labeled RNA was isolated as described (46). pUC19, GAPDH, ␤-actin, VEGF, SP-1, and c-Jun cDNAs (2 g) were digested with restriction enzymes, and then the fragment cDNAs were suspended in 0.2 N NaOH and 6 ϫ SSC, boiled for 10 min, then chilled on ice for 10 min, after which an equal volume of 2 M Tris-HCl, pH 7.4, was added. The DNA was blotted by using a Life Technologies, Inc., slot blot apparatus onto nitrocellulose membranes, which were baked and prehybridized in Hybrisol I (Oncor, Gaithersburg, MD). About 2 ϫ 10 6 cpm/ml were used for the hybridizations, which proceeded in Hybrisol at 42°C for 72 h (46,49).
Western Blot Analysis of VEGF-The whole cell lysates were prepared as described previously (49,51). Samples containing 50 g of protein from the whole cell lysates were mixed with 4 ϫ sample buffer (250 mM Tris-HCI, pH 6.8, containing 20% ␤-mercaptoethanol and 9% SDS) and boiled for 5 min. Samples were then subjected to SDSpolyacrylamide gel electrophoresis (15% gel for VEGF). Proteins were then electrophoretically transferred to a clear blot membrane P (Atto Co., Ltd, Tokyo, Japan). Detection was performed with the enhanced chemiluminescence Western blotting method (ECL, Amersham) (49,51).
Cloning and Sequencing of the VEGF Promoter Region-Synthetic oligonucleotides (20-mer) were prepared on the basis of the published DNA sequence of the VEGF promoter region (26). The primer sequences of the single strand oligonucleotides for PCR were: Ϫ624, 5Ј-AAGC-CCATTCCCTCTTTAGC-3Ј; and ϩ430, 5Ј-GGCAAAGTGAGTGACCT-GCT-3Ј. PCR proceeded using genomic DNA (100 ng) isolated from human placenta in 5 l containing a 1 M primer pair using a DNA thermal cycler (Astec Co., Fukuoka, Japan). An initial 5 min of denaturation at 95°C was followed by 2 min of primer annealing at 55°C and 2 min of polymerization at 72°C. The PCR product was gel purified and subcloned into pMOSBlue. Several independent clones were sequenced by the Taq cycle sequencing method. The reporter plasmids for expression in our cell lines were obtained as follows. One unit of Klenow fragment was added to the NdeI-BamHI VEGF promoter fragment and incubated at room temperature for 20 min to ensure blunt end molecules. Then the promoter fragment was cloned into the HindIII site of pSV00CAT (Nippongene Co., Tokyo, Japan) via HindIII linkers and designated pVEGF-CAT1. VEGF promoter deletions from Ϫ624 to ϩ430 were constructed. The promoter fragment was obtained from a subclone, and from this fragments from Ϫ129, Ϫ94, or Ϫ49 to ϩ430 were isolated. The resulting restriction fragments were made blunt ended using the Klenow fragment of DNA polymerase I and cloned into pSV00CAT using HindIII linkers and designated pVEGF-CAT2-4.
Plasmid Transfection-U251 cells were transfected using Lipofectin (Life Technologies), and then a mixture of pVEGF-CAT1 (10 g) and pRSV-neo (0.5 g) was added. After 8 h, the medium containing DNA and Lipofectin was replaced with fresh medium. After overnight incubation with medium, the cells were incubated in selection medium containing 0.6 mg/ml Geneticin (Life Technologies). Growing colonies were cloned, expanded, and tested for CAT activity.
CAT Assays-The stably transfected cells were exponentially grown in DMEM containing 10% FBS, followed by DMEM containing 1% FBS for 4 h. Thereafter cells were treated with 10 ng/ml bFGF, 100 units/ml TNF-␣, or hypoxia for 24 h. A transient CAT assay was performed after transfection by a calcium co-precipitation method (49,51). The transfected cells were incubated for 6 h with fresh DMEM containing 10% FBS, followed by further incubation for 4 h in DMEM containing 1% FBS. The cells were then treated with 100 units/ml TNF-␣ for 12 h. The cell monolayers were washed with sterile isotonic saline and harvested in 40 mM Tris-HCl, pH 7.5, 1 mM EDTA, and 150 mM NaCl. The cell pellet was resuspend and sonicated in 0.25 M Tris-HCl, pH 7.9, and then assayed for protein content by the method of Bradford (52). Each CAT assay was performed using identical amounts of protein as described (53). The CAT activities of the transient transfection assay were normalized to ␤-galactosidase activity.

Determination of Cellular Basal Levels of VEGF, SP-1, and
AP-1 mRNA-We first determined the cellular levels of VEGF mRNA in eight human glioma cell lines, which were independently isolated from patients with gliomas (38,45). Northern blots demonstrated that various levels of VEGF mRNA were expressed in the eight cell lines (Fig. 1). IN301 and IN500 cells expressed the most among the eight cell lines. There was no apparent expression of VEGF mRNA in IN157 and U343 cells. VEGF mRNA levels in U251 and T98G cells were about 50%, whereas NHG1 and NHG2 cells expressed about 25% of that in IN500 cells. The VEGF gene promoter region contains a single major transcription start and several potential binding sites for the transcription factors SP-1 and AP-1 (c-Fos and c-Jun) (26). We thus determined whether the expression of SP-1 or AP-1 was correlated with VEGF mRNA levels in several human glioma cell lines. Eight glioma cell lines expressed various mRNA levels of SP-1, c-Fos, and c-Jun, the levels of which appeared to be relatively higher in NHG1, NHG2, IN301, IN500, U251, and T98G cells, which also had relatively higher VEGF mRNA levels ( Fig. 1). The mRNA levels of VEGF, SP-1, c-Fos, and c-Jun in each cell line were measured when normalized against the GAPDH mRNA levels in each glioma cell line. Cellular mRNA levels of SP-1, c-Fos, and c-Jun appeared to be correlated with those of VEGF in the eight glioma cell lines.
We also examined whether the cellular mRNA levels of other growth factors, TGF-␣, bFGF, and TGF-␤, were correlated with SP-1 or AP-1 mRNA levels in the eight glioma cell lines. Consistent with our previous study (38), these eight glioma cell lines had various mRNA levels of bFGF, TGF-␣, and TGF-␤. However, we did not find any significant correlation between the mRNA levels of those growth factors and those of the two transcription factors. 2 Although the cellular mRNA levels of other transcription factors, AP-2 and NF-1, the binding sites of which are also located on the VEGF promoter region (26), were not tested, SP-1 or AP-1 might be involved in the expression of the VEGF gene in growing human glioma cells.
Growth Factors and Cytokines Stimulate VEGF mRNA Expression in Glioma Cells-VEGF gene expression is enhanced by EGF, PDGF, and TGF-␤ (30 -32) or by hypoxic conditions (18) in various cultured cells. We examined whether other various growth factors and cytokines could stimulate expression of the VEGF gene in glioma cell. We first examined effect of TGF-␣, bFGF, PDGF, and VEGF at 1, 10 and 100 ng/ml, IL-1 at 1 and 10 ng/ml, and TNF-␣ at 10, 100, and 1000 units/ml on the VEGF gene expression, and we found that the high end of the stimulation range was 10 ng/ml for VEGF, bFGF, TGF-␣, and PDGF, 1 ng/ml for IL-1, and 100 units/ml for TNF-␣. Fig.  2 shows cellular levels of VEGF mRNA in the human glioma cell line U251 when exposed to TGF-␣, bFGF, PDGF, and VEGF at 10 ng/ml, IL-1 at 1 ng/ml, and TNF-␣ at 100 units/ml. Northern blots showed that the major 3.7-kilobase transcripts of the VEGF gene were expressed. A 3-h incubation with bFGF, TNF-␣, and IL-1 enhanced VEGF mRNA levels about 3-fold, whereas PDGF or TGF-␣ increased them about 2-fold over those of controls in the absence of any factor (Fig. 2). VEGF itself did not enhance the VEGF mRNA level in the glioma cells. Treatment of the other glioma cell line, IN301, with the above growth factors or cytokines also demonstrated the most efficient enhancement of VEGF mRNA by bFGF, TNF-␣, and IL-1. 2 TNF-␣ or bFGF Coordinately Stimulates the Expression of Both VEGF and SP-1 Genes-We examined the effect of bFGF or TNF-␣ on the expression of VEGF mRNA in U251 cells. VEGF mRNA expression was induced about 3-fold or more after 30 min in the presence of 10 ng/ml bFGF or 100 units/ml TNF-␣ (Figs. 3 and 5, A and B). SP-1 expression was also induced 5-7-fold after 30 min in the presence of bFGF or TNF-␣. The cellular SP-1 and VEGF mRNA levels were decreased to the initial level after 3 and 6 h, respectively. In contrast, c-Fos and c-Jun gene expression increased at 3 h after exposure to bFGF or TNF-␣. The expression of the VEGF and SP-1 genes seemed to be induced with similar time kinetics in the presence of bFGF or TNF-␣ (Fig. 5, A and B). Goldberg and Schneider (27) have reported that the expression of the VEGF mRNA is induced by hypoxia through AP-1mediated transcription. We also examined the expression of VEGF, SP-1, c-Fos, and c-Jun genes under hypoxic conditions. As shown in Figs. 4 and 5C, the VEGF mRNA expression was increased about 3-5-fold by hypoxia within 3-12 h under hypoxic conditions. The c-Jun mRNA level was increased at 3 h, and that of c-Fos mRNA was induced at 1 h under hypoxic conditions. In contrast, the SP-1 mRNA levels seemed to be maintained constant. The expression of the c-Fos gene preceded that of the VEGF gene, but the enhancement of VEGF and c-Jun gene expression was virtually concomitant during hypoxia in U251 cells (Fig. 5C). AP-1 rather than SP-1 might be associated with increased expression of the VEGF gene in response to hypoxia.
VEGF mRNA Stability in the Presence of bFGF or TNF-␣-Finkenzeller et al. (54) have previously reported that up-regulation of VEGF gene by hypoxic stress is due to stabilization of VEGF mRNA in mouse NIH3T3 cells. The enhanced expression of VEGF mRNA in glioma cells incubated with bFGF or TNF-␣ might be due to its stabilization. We examined the turnover 2 M. Ryuto and M. Kuwano, unpublished observation. Exponentially growing cells were incubated in DMEM containing 1% serum for 4 h, then with TGF-␣ (10 ng/ml), bFGF (10 ng/ml), VEGF (10 ng/ml), PDGF (10 ng/ml), TNF-␣ (100 units/ml), and IL-1 (1 ng/ml) for 3 h. Cellular RNAs were hybridized with 32 P-labeled VEGF and GAPDH cDNA probes. Relative changes of VEGF mRNA levels were normalized by GAPDH mRNA level when that of untreated cells was 100%.
rates of VEGF mRNA in U251 cells incubated with or without bFGF or TNF-␣. VEGF mRNA in both untreated and treated glioma cells was degraded, with similar half-lives of about 3 h in the presence of actinomycin D (Fig. 6, A and B). Exogenous bFGF or TNF-␣ did not appear to alter the stability of VEGF mRNA, suggesting the involvement of other mechanisms besides mRNA stability in the enhanced expression of the VEGF gene by bFGF or TNF-␣.
Nuclear Run-on Assay of VEGF and SP1 Genes in the Presence of bFGF or TNF-␣-To determine whether the enhanced expression of VEGF mRNA is due to increased transcription of the VEGF gene, we performed nuclear run-on assays (Fig. 7). In comparison with the untreated control, transcription of the VEGF gene as well as SP-1 was apparently increased by bFGF or TNF-␣ in U251 cells. Transcription of both VEGF and SP-1 genes was observed after longer exposure. In contrast, c-Jun, ␤-actin, and GAPDH gene transcription was not increased by bFGF or TNF-␣. The bFGF-or TNF-␣-induced enhancement of the VEGF mRNA levels appeared to be due to transcriptional control of the VEGF gene.
Activation of DNA-binding Protein, SP-1, or AP-1 by bFGF and TNF-␣ or by Hypoxia-We demonstrated that TNF-␣ induced the rapid activation of SP-1 and NFB in human microvascular endothelial cells (46,50). In contrast, the expression of the proto-oncogenes c-jun and c-fos is up-regulated after exposure to hypoxia in the human hepatoma cell line Hep3B (27). We used a gel mobility shift assay to examine whether TNF-␣, bFGF, or hypoxia could modulate the activation of SP-1 or AP-1. U251 cells under hypoxia for 1 and 3 h induced AP-1 activation but did not when incubated for 1 h with TNF-␣ or bFGF (Fig. 8A). The binding of AP-1 to its consensus DNA recognition sequence increased 2-3-fold within 3 h after exposure to hypoxia. In contrast, SP-1 activation was induced after incubation for 1 or 3 h with TNF-␣ or bFGF, respectively (Fig.  8B). TNF-␣ and bFGF induced 3-4-and 2-fold increases in the SP-1 binding activity, respectively, suggesting a greater activation of SP-1 by TNF-␣ than by bFGF. Hypoxia stress appeared not to enhance the SP-1 binding to its consensus sequence. To asses the DNA-protein complex, competition experiments with unlabeled, double stranded oligonucleotides were performed. As can be seen in Fig. 8C, the upper band was efficiently diminished when competed by preincubation with a 50-fold excess of SP-1 consensus oligonucleotide. Other oligonucleotides such as mutated SP-1 and AP-1 did not compete the binding.
Effect of Mithramycin on TNF-␣or bFGF-induced Activation of the VEGF Gene-Mithramycin binds to the GC box and prevents sequential SP-1 binding, resulting in specific inhibition of expression of several genes (55)(56)(57). We examined whether mithramycin could inhibit bFGF-or TNF-␣-induced activation of the VEGF gene. Treatment with mithramycin at 1 and 10 nM inhibited enhanced expression of the VEGF gene by bFGF or TNF-␣ (Fig. 9A). Mithramycin at 10 nM decreased the VEGF mRNA level stimulated by bFGF or TNF-␣ to that in the  (Fig. 3) and under hypoxic stress (Fig. 4) were quantified from Northern blotting data when normalized by cellular levels of GAPDH mRNA (Fig. 3) and rRNA (Fig. 4). Radioactivity of the corresponding area on the filter was determined by BAS2000 bioimage analyzer. The signal intensity at time 0 in Figs. 3 and 4 was set at 1.0. Ⅺ, VEGF; f, SP-1; E, c-Fos; q, c-Jun. absence of both factors. However, we could not observe any apparent decrease of VEGF mRNA levels stimulated under hypoxic stress by mithramycin at 1 and 10 nM (Fig. 9B). We observed that VEGF gene activation by the hypoxic stress was inhibited in the presence of 100 nM mithramycin (data not shown).
Four VEGF species, VEGF 121 , VEGF 165 , VEGF 189 , and VEGF 206 , are generated in various cell types by differential splicing of the VEGF mRNA (26). Western blot analysis with a specific antibody against anti-human VEGF demonstrated production of VEGF 165 (Fig. 10, band a) and VEGF 121 (Fig. 10,  band b) in the U251 glioma cells, and Fig. 10, band c, showed three different size molecules of 12-15 kDa, possibly proteolytic products of VEGF. The cellular level of VEGF 165 was increased 2-and 3.5-fold higher, respectively, at 6 and 12 h after treat-ment with TNF-␣ than at time 0. Mithramycin inhibited the TNF-␣-induced up-regulation of VEGF molecules, and the drug at 10 nM almost completely inhibited the up-regulation.
Induction of VEGF-CAT Activity by TNF-␣ or bFGF-Based on the results obtained from nuclear run-on assays, we expected that TNF-␣ or bFGF could stimulate the transcriptional activation of the VEGF gene in U251 glioma cells. We cloned the promoter region of the VEGF gene between Ϫ624 and ϩ430. This promoter region contains five SP-1 binding sites and one AP-1 binding site but not hypoxia regulatory elements, which have been reported (29). We constructed the VEGF promoter-CAT and transfected it stably into U251 glioma cells. We then analyzed the promoter activity in the stable U251 glioma cells. The CAT activity was increased about 3-4-fold higher by TNF-␣ or bFGF (Fig. 11). By contrast, the CAT activity was not changed under hypoxic condition. We examined the effect of TNF-␣ at 10, 100, and 1000 units/ml on the VEGF promoter activity, and we observed 4-fold higher maximum stimulation at 100 units/ml TNF-␣. To map the regions that confer both basal and TNF-␣-induced activities on the VEGF promoters, we constructed a series of promoter deletion mutants carrying FIG. 6. Stability of VEGF mRNA in the absence and presence of bFGF or TNF-␣. Glioma cells were incubated in DMEM containing 1% serum for 4 h, then with or without bFGF or TNF-␣ for 1 h. Cells were further exposed to 1 g/ml actinomycin D (Act D), and RNAs prepared at the indicated periods were hybridized with 32 P-labeled VEGF and GAPDH cDNA probes. A, Northern blot patterns. B, decay rates of VEGF mRNA were determined when mRNA levels at the indicated periods were normalized by the mRNA level at time 0 in the absence or presence of bFGF or TNF-␣. Ⅺ, control; q, bFGF; E, TNF-␣.

FIG. 7.
Nuclear run-on analysis after exposure to bFGF or TNF-␣. Nuclei were prepared from cells incubated with or without bFGF or TNF-␣ for 1 h. Transcription in the isolated nuclei was analyzed by hybridization of 32 P-labeled transcript to 5 g of pUC19, GAPDH, ␤-actin, VEGF, SP-1, and c-Jun cDNA fragments immobilized on individual nitrocellulose membranes. The data are from after exposure to the imaging plate of BAS2000 for 12 h. Apparent expression of VEGF and SP-1 genes, but not pUC19, in the absence of bFGF or TNF-␣ (None) was observed when exposed for 72 h.

FIG. 8. Effect of bFGF, TNF-␣, and hypoxic stress on nuclear factor binding to AP-1 (A) and SP-1 (B and C) consensus fragments.
Cells under serum starvation (1% serum) were incubated for 1 or 3 h with 10 ng/ml bFGF or 100 units/ml TNF-␣. Under hypoxic conditions, exponentially growing cells were cultured in the presence of 10% serum for the indicated periods. Nuclear extracts prepared from the cells treated under various conditions were incubated with 32 P-labeled oligonucleotides for AP-1 or SP-1 and resolved by gel electrophoresis. A and B, a 1000-fold excess of unlabeled oligonucleotides was added for competition. C, nuclear extract was preincubated with or without a 50-fold excess of various consensus oligonucleotides before adding labeled probe. Arrowhead, specifically retarded DNA-protein complex of AP-1 or SP-1.
variable 5Ј-ends from nucleotides Ϫ624 to Ϫ49 and a common 3Ј-end at nucleotide ϩ430 (Fig. 12A). Removal of the regions located between nucleotides Ϫ129 and Ϫ94 (pVEGF-CAT3) dramatically decreased the promoter activity (Fig. 12, B and C). TNF-␣ at 100 units/ml induced 2-3-fold more promoter activity driven by the VEGF promoter carrying a deletion until nucleotide Ϫ129, but it could not enhance when the regions between Ϫ129 and Ϫ94 were deleted (Fig. 12, B and C). DISCUSSION Several potential binding sites for the transcription factors AP-1, AP-2, and SP-1 are localized in the VEGF gene promoter (26). Among eight glioma cell lines, glioma cell lines with higher VEGF mRNA levels had higher levels of SP-1, c-Fos, and c-Jun mRNA, whereas those with lower VEGF mRNA levels had lower levels of SP-1, c-Fos, and c-Jun mRNA (Fig. 1). Steady state levels of VEGF mRNA in growing glioma cells thus might be closely correlated with those of the transcription factors SP-1 and AP-1, suggesting the involvement of both transcription factors in basal level expression. On the other hand, high levels of VEGF mRNA are expressed in small anaplastic glioma cells, which line up in close proximity to necrotic areas, suggesting that VEGF mRNA expression can be induced in vivo under hypoxic conditions (18). Yao et al. (58) have reported that the hypoxic response of NADPH-oxidoreductasediaphorase expression in human colon cancer cells was mediated through AP-1, initially by an increase of c-Jun, then followed by a further increase of c-Fos. In our present study, hypoxic stress induced a rapid increase of c-Fos mRNA, followed by the up-regulation of both c-Jun and VEGF genes 3 h later in glioma cells, but SP-1 gene expression was not affected by the hypoxic conditions (Fig. 5). A relevant study by Goldberg and Schneider (27) has demonstrated up-regulation of c-Jun and VEGF genes in hepatoma cells under hypoxic stress conditions. Gel shift assays demonstrated that AP-1 binding to the phorbol ester (12-O-tetradecanoylphorbol-13-acetate) responsive element was apparently enhanced 1-3 h after exposure to hypoxia, but that SP-1 was not evidently activated (Fig. 8, A   FIG. 9. Effect of mithramycin on cellular increase in VEGF mRNA levels by bFGF or TNF-␣ (A) and hypoxia (B). Cells were incubated in DMEM containing 1% serum with bFGF or TNF-␣ for 1 h in the absence or presence of 1 and 10 nM mithramycin. Under hypoxic condition, cells were incubated in DMEM containing 10% serum for 6 h with or without mithramycin. Then cellular levels of VEGF mRNA were determined by Northern blotting. The amount of mRNA in each lane is presented after ethidium bromide staining. Radioactivity of the corresponding area on the filter was examined by BAS2000 bioimage analyzer. Signal intensity at time 0 in the absence of any treatment was normalized as 100%.

FIG. 10. Effect of mithramycin on cellular levels of VEGF in TNF-␣-treated cells.
Cells were treated for 0, 6, and 12 h with or without TNF-␣ in the absence or presence of mithramycin, and cellular lysates were fractionated by 15% SDS-polyacrylamide gel electrophoresis. VEGF was subsequently detected by immunoblotting with a polyclonal anti-human VEGF antibody. Band a, VEGF 165 (23 kDa); band b, VEGF 121 (18 kDa); band c, products of processed VEGF (12-15 kDa). The nonimmune antibody did not recognize any immunoreactive bands (data not shown). The signal intensity of band a in the absence of TNF-␣ and mithramycin was normalized as 100%.

FIG. 11. A structure of the VEGF promoter region (A) and induction of VEGF-CAT activity by TNF-␣ , bFGF, and hypoxia (B and C).
Promoter analysis of the promoter regions of the human VEGF gene was performed. A, nucleotides are numbered from the transcription start site. Ⅺ, SP-1 binding site; E, AP-1 binding site; f, hypoxia regulatory element. B and C, U251 stable cell lines were treated with TNF-␣, bFGF, or hypoxia and assayed 24 h thereafter. U251 cells were incubated in DMEM containing 1% serum. The relative fold increase was determined when normalized by endogenous CAT activity in the absence of any treatment. and B). The hypoxia-induced activation of the VEGF gene was not inhibited by mithramycin, an SP-1-prone inhibitor (Fig.  9B). Finkenzeller et al. (54) have reported that enhanced expression of the VEGF gene under hypoxic conditions is independent of AP-1 but dependent on the stability of VEGF mRNA in mouse NIH3T3 cells. Minchenko et al. (29) have reported the existence of a new element, hypoxia regulatory element, in the 5Ј-and 3Ј-region of the VEGF gene, but we did not examine this element in our present study. Hypoxia also causes the activation of another transcription factor, NFB, in Jurkat T cells (28), but such a nuclear factor motif does not exist in the VEGF promoter. The activation of AP-1 rather than that of SP-1 appeared to mediate in part the hypoxia-induced expression of the VEGF gene in glioma cells. However, other elements or mechanisms are also expected to be involved in the hypoxiainduced expression of the VEGF gene in human glioma cells.
VEGF expression is enhanced by several growth factors, such as EGF (30), PDGF (31), and TGF-␤ (32), in various cell types, but the underlying mechanism is not known. VEGF mRNA expression is stimulated by phorbol esters and cAMP analogues, indicating that both protein kinase C-and protein kinase A-mediated pathways are involved in the regulation of VEGF gene expression (33). In this study, the cellular VEGF mRNA level was most potently enhanced in response to bFGF, TNF-␣, and IL-1 and moderately enhanced in response to TGF-␣ and PDGF (Fig. 2). We previously demonstrated that TNF-␣ specifically destabilizes ␤-actin mRNA in human microvascular endothelial cells (48) but that TNF-␣ appears not to alter the longevity of the VEGF mRNA half-life (Fig. 6, A and  B). Nuclear run-on assays demonstrated the activation of VEGF and SP-1 genes by TNF-␣ or bFGF (Fig. 7), suggesting that the enhanced expression of the VEGF gene might be due to transcriptional activation. VEGF and SP-1 gene expression was enhanced during shorter periods within 3 h in response to bFGF or TNF-␣. Although AP-1 mediates cellular responses to various growth-modulating factors and phorbol ester (59 -61), cellular levels of c-Fos and c-Jun were enhanced at 3 h after exposure to TNF-␣ or bFGF (Fig. 5). Gel shift assays demonstrated apparent activation of SP-1, but not AP-1, during a short exposure (1 h) to TNF-␣ or bFGF (Fig. 8, A and B). Treatment with mithramycin, an inhibitor of SP-1 interaction with its consensus sequence (55)(56)(57), at 1 and 10 nM inhibited the TNF-␣-or bFGF-induced expression of the VEGF gene (Figs. 9, A and B, and 10). The rapid increase of VEGF by TNF-␣ or bFGF might thus be mainly mediated through the activation of SP-1 in the glioma cells.
In the 5Ј-flanking region of the VEGF gene, there are six SP-1 binding sites besides those for AP-1, AP-2, and NF-1 (26). TNF-␣ often induces the expression of many immunologically relevant and other irrelevant genes through the transcription factor NFB (62). Consistent with this report, exposing human microvascular endothelial cells to TNF-␣ activates not only NFB but also SP-1 (46,50). The expression of the LDL receptor is dramatically increased in response to TNF-␣, but the promoter for the LDL receptor gene has no NFB motif, suggesting that TNF-␣-dependent LDL receptor gene expression is closely coupled with an SP-1 element on the LDL receptor promoter (46). The expression of VEGF as well as the LDL receptor gene might be mediated in part through a transcription factor, SP-1, in response to exogenous stimulation by TNF-␣, and possibly by bFGF. On the other hand, our recent study has demonstrated that H 2 O 2 induces production of an angiogenic factor, IL-8, through NFB activation, resulting in tubular morphogenesis by human microvascular endothelial cells (63). Stoltz et al. (64) have reported that an arachidonic acid metabolite, 12(R)-hydroxyeicosatrienoic acid, stimulates tubular morphogenesis by colonary microvessel endothelial cells through NFB activation. The activation of NFB thus appears to involve angiogenesis in other model systems.
We examined the promoter (Ϫ624 to ϩ430) activity of the human VEGF gene using the transfectants after treatment with TNF-␣ or bFGF (Fig. 11). This region contains five SP-1 binding sites and one AP-1 binding site, and our present study demonstrated that this region could respond to TNF-␣ or bFGF but not hypoxia. Ikeda et al. (65) have reported that deletion of Ϫ1180 to Ϫ887 deprives the fragment of its hypoxia responsiveness in mouse glioma C6 cells. Minchenko et al. (29) also reported that the region from Ϫ844 to Ϫ744 contained hypoxia regulatory elements in human cervical cancer HeLa cells. The VEGF promoter in the VEGF-CAT1 construct carries no hypoxia regulatory elements but one AP-1 motif (Fig. 11). Hypoxia stress did not enhance CAT activity driven by the VEGF promoter, suggesting that the AP-1 consensus sequence at this region might not be required for the hypoxia-induced activation of the VEGF gene. TNF-␣ or bFGF appears to stimulate expression of the VEGF gene through SP-1 on its promoter. This CAT assay also indicated that the mechanism of VEGF gene activation by TNF-␣ or bFGF were different from that by hypoxia.
The human VEGF promoter lacks the TATA box in the vicinity of the transcription start sites but is very GC rich (26). Transcription of many TATA-less promoters is often activated  ␣ (B and C). The VEGF promoter-CAT gene fusion plasmids pVEGF-CAT1-4 were transiently co-transfected with pSV2-␤-galactosidase. A, restriction enzymes used to generate the 5Ј-ends of the deletion constructs are shown on the left. Restriction sites are shown in parentheses, numbered relative to ATG ϩ1. The 3Ј-end of each promoter fragment is at ϩ430. Ⅺ, SP-1 binding site; ⅜, AP-1 binding site. B and C, CAT activities were corrected for differences in transfection efficiency among the cell lines as estimated by ␤-galactosidase activity and then normalized to corrected activity of transient pVEGF-CAT-transfected cells. The relative fold increase was determined when normalized by endogenous CAT activity of pVEGF-CAT1-transfected cells in the absence of any treatment. by SP-1, and clusterization of SP-1 binding sites in the proximity of the transcription start site appears to be typical of the TATA-less promoters (66 -70). In the human VEGF promoter, four proximal SP-1 binding sites are clusterized between nucleotides Ϫ129 and ϩ1. Constructs containing the four clustered SP-1 binding sites were sufficient to allow trans-activation of the VEGF promoter by SP-1 in the absence and presence of TNF-␣. pVEGF-CAT3, containing three SP-1 binding sites, abolished its promoter activity in the absence and presence of TNF-␣ (Fig. 12), implying that four SP-1 binding sites are critical for basal promoter activity and also for TNF-␣ responsiveness. The insulin-like growth factor-binding protein-2 gene promoter lacks a TATA box but has three proximal clustered SP-1 binding sites (71). All three of these SP-1 binding sites are required for promoter activity of the insulin-like growth factorbinding protein-2 gene. Consistent with this report (71), our present study also indicated that four clustered SP-1 binding sites are important for both basal transcription and the TNF-␣-induced transcription of the VEGF gene.
Incubating U251 cells with TNF-␣ or IL-1 increased the bFGF mRNA level severalfold. 2 Okamura et al. (72) have previously demonstrated that TNF-␣ enhances bFGF expression in human microvascular endothelial cells. This TNF-␣-induced bFGF further modulates the expression of interleukin 6 and collagenase in the microvascular endothelial cells. Exogenous VEGF induces tubular morphogenesis in bovine aortic endothelial cells, and incubation with both VEGF and bFGF also synergistically enhances tubular morphogenesis (24). Human microvascular endothelial cells efficiently induce tubular morphogenesis in type I collagen gels in the presence of both VEGF and bFGF and when co-cultured with VEGF-producing human glioma cells (12). The enhanced expression of both VEGF and bFGF in tumor cells in response to TNF-␣ or IL-1 might induce a paracrine loop for neovascularization under pathological conditions, including tumors, diabetic retinopathy, and inflammation.
This study demonstrated that gene expression of the potent angiogenic factor VEGF induced by bFGF or TNF-␣ might be mediated through the transcription factor SP-1 rather than another transcription factor, AP-1, in human glioma cells. The VEGF promoter-driven reporter assay also indicated a possible involvement of SP-1 in VEGF gene expression in response to TNF-␣ or bFGF. However, hypoxia stress did not induce activation of SP-1. The regulatory mechanism underlying the TNF-␣and bFGF-induced expression of the VEGF gene appeared to differ from that induced by hypoxia.