Oxidative Stress Regulates Vascular Endothelial Growth Factor-A Gene Transcription through Sp1- and Sp3-dependent Activation of Two Proximal GC-rich Promoter Elements*

Enhanced VEGF-A (vascular endothelial growth factor A) gene expression is associated with increased tumor growth and metastatic spread of solid malignancies including gastric cancer. Oxidative stress has been linked to tumor-associated neoangiogenesis; underlying mechanisms, however, remained poorly understood. Therefore, we studied the effect of oxidative stress on VEGF-A gene expression in gastric cancer cells. Oxidative stress generated by H 2 O 2 application potently stim- ulated VEGF-A protein and mRNA levels as determined by enzyme-linked immunosorbent assay and real-time PCR techniques, respectively, and elevated the activity of a transfected ( (cid:1) 2018) VEGF-A promoter reporter gene construct in a time- and dose-dependent manner (4–8-fold). These effects were abolished by the antioxidant N -acetylcysteine, demonstrating DNA-protein complexes were electrophoresed in 6% nondenaturing polyacrylamide gels containing 0.5 (cid:2) TBE (50 m M Tris, 50 m M borate, 2 m M EDTA). Gels were dried and exposed to Eastman Kodak Co. BioMax MR films (Amersham Biosciences) using an intensifying screen. Immunoblotting— For signaling studies, AGS cells were exposed to 1 m M H 2 O 2 , and whole cell lysates were prepared in 20 m M Tris (pH 7.9), 0.42 M NaCl, 1.5 m M MgCl 2 , 0.2 m M EDTA, 10 m M K 2 HPO 4 , 1 m M Na 3 VO 4 , 10 m M NaF, 1.25% Nonidet P-40, and 10% glycerol as de- scribed (38). Following SDS-PAGE and transfer to nylon membranes, Western blot analysis of proteins was performed as described (38) using antibodies specifically recognizing phosphorylated ERK1/2, p38, JNK, or MEK1/2 (New England Biolabs, Beverly, MA) or corresponding non- phosphospecific antibodies to demonstrate equal protein loading.

droxyl radical OH ⅐ , and H 2 O 2 are continuously generated as products of cellular metabolism (1)(2)(3). Increased levels of ROS reflecting oxidative stress have been observed during a number of physiologic and pathologic states, comprising cellular differentiation, reperfusion injury, and ulcerative and inflammatory conditions (2,3). Oxidative stress is also believed to be a major pathogenic factor underlying development and progression of chronic disease states such as artherosclerosis, diabetes, and aging (2)(3)(4). Moreover, ROS participate in malignant transformation as well as metastatic spread of tumor cells (2)(3)(4). Elevated levels of ROS can lead to DNA damage and apoptosis but also act as important signaling molecules influencing various cellular functions including gene expression (2, 6 -9).
The gastric mucosa is permanently exposed to luminal oxidants generated from ingested food, bacteria, and shed mucosal cells, and together with the surface mucous layer, the gastric epithelium represents the first line of defense against luminal oxidative alterations. Increased ROS levels have been linked to peptic lesions of the gastric mucosa triggered by ethanol, nonsteroidal anti-inflammatory drugs, stress, ischemia-reperfusion, and Helicobacter pylori infection (10 -12). In addition to these deleterious effects, ROS are also capable of influencing mucosal repair processes by stimulating epithelial proliferation, production, and release of mucosal growth factors as well as activation of proangiogenic pathways (13). Moreover, several studies have suggested that oxidative stress is also involved in gastric carcinogenesis (14,15). Accordingly, epidemiological studies indicated that application of antioxidants can significantly reduce the risk of gastric cancer (15) and lead to regression of gastric epithelial metaplasia (16), a lesion believed to reflect premalignant states of the gastric mucosa.
Neoangiogenesis is a general pathophysiological mechanism critically involved in healing of inflammatory and ulcerative epithelial lesions as well as tumor growth and metastasis (17). Among the proangiogenic factors identified so far, VEGF-A (vascular endothelial growth factor A) represents one of the most potent stimuli of neoangiogenesis (17,18). In the stomach, enhanced VEGF-A gene expression has been identified to critically contribute to peptic ulcer healing (13, 19 -21). In addi-tion, gastric adenocarcinomas were found to frequently display high levels of VEGF-A expression accompanied by increased intratumoral microvessel density (22,23), whereas injection of VEGF-A-specific neutralizing antibodies were capable of potently inhibiting the growth of gastric cancer xenotransplants in rodents (24). Despite the fact that these studies clearly established an important role of VEGF-A-dependent angiogenesis in mucosal regeneration, peptic ulcer healing, and gastric cancer, the pathways controlling VEGF-A gene expression in these settings have not yet been defined.
Previous studies revealed that VEGF-A gene expression can be influenced by extracellular growth factors, cytokines, and genetic alterations or hypoxia (17,25). Moreover, current studies indicated that angiogenic processes may be influenced by cellular redox processes (26 -30); potential mechanisms linking oxidative stress to proangiogenic factors like VEGF-A, however, remained unclear (29). Transcriptional activation of the VEGF-A promoter represents a core mechanism through which expression of the VEGF-A gene can be regulated (17,18,25). Several cis-acting promoter elements including a hypoxia-responsive site at Ϫ975, proximal GC-rich elements, and Egr-1 and AP2 recognition motifs were found to participate in VEGF-A gene regulation (31)(32)(33)(34)(35). Depending on the cellular context and/or the stimulus investigated, these recognition motifs and/or their respective binding factors variably contribute to VEGF-A expression control (31)(32)(33)(34)(35). Moreover, VEGF-A gene expression has been shown to be regulated through several signaling cascades comprising MAP kinase ERK-, JNK-, and p38-dependent cascades as well as NF-B pathways (17,18,25,32). In contrast, signaling pathways mediating the effects of oxidative stress on the VEGF-A gene await clarification.
Here we demonstrate that oxidative stress generated by the model oxidant H 2 O 2 at submillimolar concentrations potently stimulates release and production of VEGF-A in gastric cancer cells and provides evidence that transcriptional activation of the VEGF-A promoter represents the underlying core mechanism. Moreover, we identify the zinc finger transcription factors Sp1 and Sp3 as molecular mediators of the VEGF-A redox response and show that Sp1/Sp3-activated GC-boxes located in the proximal VEGF-A promoter differentially participate in oxidative stress-dependent VEGF-A gene regulation. Furthermore, we show that MAPK ERK-related and to a lesser extent JNK-related signaling pathways are crucial for transmission of oxidative stress effects on the human VEGF-A gene in gastric cancer cells.
TaqMan Quantitative Real Time Reverse Transcription-PCR Analysis-Primers were designed using Primer Express software (PerkinElmer Life Sciences). Expression levels of human VEGF-A and housekeeping gene ␤-actin were determined using the following primer pairs: forward VEGF-A (5Ј-cttgccttgctgctctacc-3Ј) and reverse VEGF-A (5Ј-cacacaggatggcttgaag-3Ј); forward ␤-actin (5Ј-tcagcaagcaggagtatgacga-3Ј) and reverse ␤-actin (5Ј-cgcaactaagtcatagtccgcc-3Ј). Specificity of the products was demonstrated for each fragment by melting curve analysis, gel electrophoresis, and sequencing. The SYBR Green I assay and the ABI Prism 7700 sequence detection system (Applied Biosystems) were used for detecting real-time quantitative PCR products from 0.25-2.5 ng of reverse-transcribed cDNA samples. SYBR Green I dye intercalation into the minor groove of double-stranded DNA reaches an emission maximum at 530 nm. PCRs for each sample were done in triplicate for both target gene and ␤-actin control. Quantitation of mRNA expression was carried out by relating the PCR threshold cycle obtained from tissue samples to a cDNA standard curve. In the case of ␤-actin, the PCR threshold cycle was related to the number of AGS cells from which ␤-actin was extracted. The normalized amount of VEGF-A expression was obtained by dividing the averaged sample value by the averaged ␤-actin value of AGS cells and given in copies of VEGF-A per AGS cell.
Transfection Studies-Transient transfections of cultured AGS cells were carried out using the calcium phosphate precipitation technique (DNA transfection kit; 5 Prime 3 3 Prime, Inc., Boulder, CO) as previously described (36 -38). Briefly, AGS cells were transfected with 0.5 g of reporter gene plasmid per well unless otherwise indicated. To correct for transfection efficiency, 50 ng/well of Renilla luciferase construct pRL-TK (Promega) were cotransfected. After transfection, cells were maintained in serum-free UltraCulture (BioWhittaker) for 24 h. Unless otherwise indicated, cells were stimulated for 6 h with 500 M hydrogen peroxide or 10 Ϫ8 M phorbol 12-myristate 13-acetate (PMA; Biomol, Plymouth Meeting, PA). To block p38 MAP kinase, incubations were performed in the presence of 25 M SB 202190 (Calbiochem). For transfection of SL-2 cells, 10 6 cells/well were transfected with 3 g of VEGF-A(Ϫ88/Ϫ50)-Luc along with 1 g of expression constructs (Sp1, Sp3, Sp1-DBD, Sp3-DBD) or empty vectors. Luciferase activities were detected in a monolight Luminometer (EG Berthold, Bad Wildbach, Germany) using the "Dual Luciferase Reporter Assay" (Promega). Incubations were performed in triplicates, and results were normalized for transfection efficiency and calculated as means Ϯ S.E. Values were expressed as arbitrary light units or -fold increases in luciferase activity compared with controls. Statistical significances were calculated using Student's t test (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001).

Oxidative Stress Stimulates Production and Release of VEGF-A Protein and Increases VEGF-A mRNA Levels in AGS
Cells-To determine the effects of oxidative stress on VEGF-A gene expression, AGS cells were exposed to H 2 O 2 and analyzed for VEGF-A protein and mRNA levels. H 2 O 2 time-and dosedependently stimulated production and release of VEGF-A protein, having a maximum (ϳ3-fold elevation) at 12 h (Fig. 1, A and B). These responses were redox-specific, as shown by their complete reversal through application of the antioxidant Nacetylcysteine. To investigate potential cell lysis due to H 2 O 2 treatment, lactate dehydrogenase activity in cell culture supernatants was analyzed in parallel ( Fig. 1C) and found to be unaltered (Fig. 1C). Similar to the changes observed on the protein level, H 2 O 2 treatment increased VEGF-A mRNA levels (maximum at 8 -12 h), suggesting that enhanced transcription represents the mechanism underlying oxidative stress responsiveness of the VEGF-A gene (Fig. 1D). Oxidative Stress Regulates the VEGF-A Promoter through a Proximal 39-bp Promoter Element-To identify promoter regions mediating oxidative stress responsiveness of the human VEGF-A gene, functional 5Ј deletion analysis was performed. These studies revealed that loss of the region spanning Ϫ2018 to Ϫ85 had no substantial influence on VEGF-A promoter activity. In contrast, removal of an additional 33 nucleotides abolished basal and H 2 O 2 -stimulated VEGF-A promoter activity (Fig. 3A), suggesting that Ϫ85/Ϫ52 comprises essential regulatory elements. To investigate the enhancer properties of this promoter region in detail, it was transferred to a heterologous promoter system (pT81-Luc) and functionally analyzed (Fig. 3B). The presence of the Ϫ88/Ϫ50 element conferred elevation of basal transcriptional activity (10 -15-fold) as well as H 2 O 2 responsiveness to the per se redox-insensitive pT81-Luc vector (Fig. 3B). Redox responses of VEGF-A Ϫ88/Ϫ50 were abolished by the antioxidant N-acteylcysteine (Fig. 3B). Analysis of the Ϫ88/Ϫ50 promoter region in EMSA studies showed that under basal and H 2 O 2 -stimulated conditions, two major   3C, lanes 3 and 4). Only infrequently we observed a third complex (complex III), which was not influenced by any oligonucleotides used in EMSA studies (data not shown) and therefore must be regarded as nonspecific. H 2 O 2 treatment maximally enhanced complex I and II after 10 min, whereas prolonged exposure decreased complex formation (lanes [5][6][7][8]. Under identical conditions, H 2 O 2 did not increase the binding of CREB, USF-1, and USF-2 transcription factors to a minimal cox-2 (cyclooxygenase 2) promoter fragment, 2 demonstrating that oxidative stress-triggered protein/DNA interaction at the VEGF-A promoter is not a random nonspecific process.

Oxidative Stress Potently Transactivates the VEGF-A Gene
Oxidative Stress Stimulates Binding of Sp1 and Sp3 to the VEGF-A Promoter-To further characterize the nuclear factors binding to the VEGF-A(Ϫ88/Ϫ50) element, EMSA competition studies were carried out (Fig. 4). The selection of consensus oligonucleotides used in these studies (Sp1, AP2, and Egr-1) was based on the presence of these sequences within the Ϫ88/ Ϫ50 sequence. We found that complex I and II, which were found in unstimulated (Fig. 4A, lanes 1-7) as well as stimulated samples (lanes 8 -14) were completely competed out by an excess of Sp1 consensus sequence (lanes 2 and 9). In contrast, AP2 consensus sequence was clearly less effective (lanes 6 and 13), whereas Egr-1 consensus oligonucleotides had no effect (lanes 4 and 11). To confirm these results, EMSA supershift experiments using specific antibodies recognizing Sp1, Sp3, Sp4, AP2, or Egr-1 were performed. These studies identified complex I as containing Sp1 protein (Fig. 4, B (lanes 3 and 8) and C (lanes 2, 5, 6, and 8)), whereas complex II was shown to consist of Sp3 (Fig. 4C, lanes 3, 5, 7, and 8). In contrast, zinc finger protein Sp4 as well as AP2 and Egr-1 were not detected. In control EMSAs, the specificity of the AP2 antibody was confirmed using HeLa cell nuclear extracts together with a labeled AP2 consensus site as a probe (data not shown). Functionality of the Egr-1 antibody was confirmed in supershifts employing AGS nuclear extracts together with a proximal fragment of the mouse chromogranin A gene promoter as radiolabeled probe (38).  (41). Overexpression of either Sp1 or Sp3 potently stimulated the VEGF-A(Ϫ88/Ϫ50) element (7-8-fold), demonstrating that the VEGF-A promoter is highly reactive to both transcription factors (Fig. 5). Cotransfection of both factors did not yield in higher effects compared with the expression of the individual factors. Sp1 and Sp3 mutants lacking their DNA-binding domain exhibited virtually no transactivating effect on the VEGF-A(-88/-50) element (Fig.  5), confirming the specificity of results obtained with wild type expression constructs.
Oxidative Stress Stimulates Sp1 but Not Sp3 Transactivating Capacity-To investigate the influence of oxidative stress on the transactivating capacity of Sp1 and Sp3, we cotransfected AGS cells with Gal4-Sp1 or Gal4-Sp3 expression constructs and the 5ϫGal4-Luc reporter plasmid (Fig. 6). Whereas transfection of 5ϫGal4-Luc or Sp1/Sp3 transactivator plasmids alone produced reporter gene activity close to background levels, cotransfection of either Gal4-Sp1 or Gal4-Sp3 along with the 5ϫGal4-Luc reporter plasmid significantly increased basal promoter activity (ϳ50-fold). Moreover, oxidative stress clearly increased Sp1 transactivating capacity (ϳ3-fold) but did not significantly influence Sp3. These data confirm the regulatory influence of oxidative stress on Sp1 and suggest potential differences in the regulatory mechanisms controlling the activity of Sp1 and Sp3 in response to cellular oxidative stress in gastric cancer cells.
Proximal GC-boxes Differentially Contribute to VEGF-A Promoter Regulation-To elucidate the functional importance of individual GC-boxes located within the Ϫ88/Ϫ50 region, a series of VEGF-A(Ϫ88/Ϫ50) mutants was characterized in transfection assays and EMSA studies (Fig. 7). Compared with the wild type VEGF-A(Ϫ88/Ϫ50) element, mutation of GC-box I (Mut 1) did not affect oxidative stress-triggered activation of this sequence (Fig. 7A). Functional loss of GC-box II or GC-box III, however, significantly inhibited basal and stimulated VEGF-A promoter activity (Mut 4, Mut 6), suggesting that oxidative stress responsiveness of the VEGF-A(Ϫ88/Ϫ50) region requires the functional integrity of these two elements. In line with this view, simultaneous inactivation of GC-box II and III in the presence of an intact GC-box I (Mut 5), reduced basal and stimulated promoter activity to background levels, a result that was also observed after mutation of all three GC-boxes (Mut 3). Furthermore, when GC-box I was mutated together with either GC-box II (Mut 2) or GC-box III (Mut 7), promoter activity was decreased to a similar degree as observed after individual mutagenesis of GC-boxes II or III (ϳ50%), further confirming the lack of functional relevance of GC-box I in VEGF-A(Ϫ88/Ϫ50) regulation. EMSA studies using mutated VEGF-A(Ϫ88/Ϫ50) oligonucleotides either as competitors (Fig.  7B) or as radiolabeled probes (Fig. 7C, data only shown for mutants Mut 1-3) revealed that mutation of individual GCboxes I, II, or III had little or no effect on factor binding, supporting the concept that the loss of single GC-boxes can be compensated regarding Sp1/Sp3 binding by those GC-boxes remaining intact (Fig. 7, B (lanes 2, 5, and 7) and C (lane 2)). Examination of double mutations VEGF-A(Ϫ88/Ϫ50) Mut 2, Mut 5, and Mut 7 showed reduction of Sp factor binding to variable degrees (Fig. 7, B and C, Mut 2), whereas alteration of all three GC-boxes abolished the ability of VEGF-A(Ϫ88/Ϫ50) to bind nuclear proteins (Fig. 7, B and C, Mut 3). Interestingly, mutation of GC-box I together with either GC-box II (Mut 2) or GC-box III (Mut 7) showed only little effect on factor binding ( Fig. 7B and data not shown), which correlates with the retained function of these mutants in transfection assays (Fig.  7A). In contrast, double mutation of GC-boxes II and III clearly reduced Sp1/Sp3 binding (Fig. 7, B (lane 5) and C (lane 3)), being in accordance with the reduced functional activity of this mutant. Together, these data strongly suggest that GC-boxes II and III are indispensable for basal and oxidative stress-triggered activity of the VEGF-A(Ϫ88/Ϫ50) element, whereas GCbox I, which also possesses the capacity to bind Sp1/Sp3 transcription factors, has obviously no functional importance in this context.

Analysis of Signaling Cascades Mediating Oxidative Stress Responsiveness of the VEGF-A Promoter-
To explore the signaling pathways activated by oxidative stress, the influence of H 2 O 2 on the phosphorylation status of key kinases related to redox-triggered signaling was investigated. Exposure of AGS cells to oxidative stress led to a rapid increase in ERK1/2, JNK, and p38 kinases phosphorylated (Fig. 8A), reaching maximal effects after 5-10 min and sustained hyperphosphorylation throughout the entire experimental period of 120 min. Similarly, the upstream ERK1/2 kinase MEK1 was hyperphospho-rylated in response to oxidative stress, showing a similar timeresponse relationship. After identification of MEK1/2/ERK1/2, JNK, and p38 as signaling targets of oxidative stress, we examined the functional role of these kinases for VEGF-A promoter regulation in transfection assays employing DN kinase mutants (Fig. 8B). Application of DN ERK1 and/or DN ERK2 decreased H 2 O 2 responsiveness of the VEGF-A promoter by 40 -50% (Fig. 8B), whereas the PMA response was almost abolished. Similarly, functional impairment of Raf-1 or Ras, both of which have been located upstream of the MEK1/2 3 ERK1/2 signaling module, potently inhibited H 2 O 2 -and PMA-dependent VEGF-A promoter activation (ϳ60 -70% reduction). In con- trast, application of a p38-specific kinase inhibitor (SB 202190) had no effect. Interruption of JNK-related signaling by transfection of a dominant-negative MKK4 mutant resulted in modest but reproducible inhibition of H 2 O 2 -dependent VEGF-A transactivation (ϳ30% inhibition), however, without reaching statistical significance. PMA responsiveness of the VEGF-A promoter was not altered by DN MKK4. To further explore the role of the Raf 3 MEK1/2 3 ERK1/2 pathway in VEGF-A regulation, ectopic expression of these kinases was performed. Forced expression of ERK1 or ERK2 dose-dependently transactivated the VEGF-A promoter, whereas co-expression of both kinases led to additive effects (Fig. 8C). Similar to ERKs, overexpression of MEK1 or Raf-1 potently stimulated VEGF-A promoter activity. DISCUSSION In the present study, we demonstrate oxidative stress-dependent stimulation of VEGF-A production and release in gas-tric cancer cells and provide clear evidence that oxidative stress regulates VEGF-A gene expression through transcriptional mechanisms. Previous studies in epithelial and nonepithelial cell models suggested a link between the cellular redox status and enhanced VEGF-A production and/or secretion (26 -30); underlying molecular determinants, however, have not yet been clarified. To provide a detailed functional analysis of molecular pathways mediating the effects of oxidative stress on the human VEGF-A gene, we initially investigated participating cis-and trans-activating factors employing 5Ј deletion analysis of the VEGF-A promoter, DNA element transfer studies, and systematic core promoter mutagenesis. These studies revealed that a region spanning Ϫ88 to Ϫ50 is indispensable for basal as well as oxidative stress-triggered VEGF-A promoter activity. This GC-rich promoter region, which comprises consensus binding elements for Sp-like zinc finger proteins as well as AP2 and Egr-1 transcription factors, has been demonstrated to be involved in VEGF-A gene regulation in various cell systems including fibroblasts (32,33), keratinocytes (34), and glioma cells (35). Our study revealed that in gastric cells the Ϫ88/Ϫ50 element is bound by the zinc finger transcription factors Sp1 and Sp3 (Fig. 4). Application of oxidative stress increased DNA-protein complex formation at the Ϫ88/Ϫ50 site, strongly suggesting that enhanced binding of Sp1/Sp3 to the VEGF-A promoter represents an important mechanism through which oxidative stress transactivates the VEGF-A gene (Fig. 3). This mechanism is clearly different from previous observations made in platelet-derived growth factor-stimulated human fibroblasts and tumor growth factor-␣-stimulated skin keratinocytes, showing constitutive binding of Sp1 and/or Sp3 to the proximal VEGF-A promoter without detectable changes upon stimulation (33,34). Enhanced binding of Sp1 to the proximal VEGF-A promoter has been observed after activation of the MAPK/ERK signaling pathway in hamster fibroblasts (32). In this cell type, however, Sp1 was found to functionally cooperate with the transcription factor AP2, whereas Sp3 was not involved (32). Therefore, oxidative stress-triggered transactivation of the VEGF-A promoter through enhanced binding of both Sp1 and Sp3 represents a novel molecular mode of VEGF-A transactivation.
Transcription factor Sp1 belongs to the superfamily of Splike zinc finger proteins and has been implicated in the regulation of constitutively expressed "housekeeping genes" as well as genes influencing growth and differentiation (for a review, see Ref. 43). In addition, current studies demonstrated that Sp1 also participates in the regulation of inducible gene expression and that interaction of Sp1 with other transcription factors and/or cofactors such as CREB-binding protein, p300, or CRSP 84 may represent an important transcriptional control mechanism (43). More recently, it became clear that Sp1 can also be regulated through changes of its phosphorylation state (44), and subsequently, different signaling pathways including Rasdependent activation of the MEK1/ERK cascade have been identified to target Sp1 (45)(46)(47). Similar to Sp1, Sp3 represents a zinc finger transcription factor comprising highly conserved DNA-binding domains, but analysis of their functional properties revealed significant differences between these two proteins (42,43). Sp3 activates GC-rich DNA elements with affinities similar to other Sp proteins (42,43). In some systems, however, Sp3 has been described to lack intrinsic activity and accordingly can act as transcriptional repressor of other transcription factors binding to the same element (42,43). As demonstrated by experiments in SL-2 Schneider cells, which represent the appropriate cell model for analysis of Sp factor-dependent effects, Sp1 and Sp3 exert equipotent transactivating properties on the VEGF-A promoter, without any indication of Sp3 acting FIG. 8. Analysis of signaling cascades mediating oxidative stress responsiveness of the VEGF-A promoter. A, phosphorylation of ERK1/2, p38, JNK, and MEK1/2 in response to H 2 O 2 (500 M) was investigated using phosphospecific antibodies in Western blot analysis (upper panels). As a loading control, blots were stripped and reprobed with corresponding nonphosphospecific antibodies (lower panels). Data shown represent a typical result obtained from a series of three independent experiments. B, AGS cells were transfected with construct VEGF-A(Ϫ88/Ϫ50)-Luc along with various dominant negative kinase mutants or the corresponding empty vectors and exposed to H 2 O 2 (500 M) or PMA (10 nM). Additionally, inhibitor SB 202190 was applied to inhibit MAPK p38-dependent effects. Results represent means Ϯ S.E. of three separate experiments, and asterisks indicate statistically significant differences. *, p Ͻ 0.05; **, p Ͻ 0.01. C, AGS cells transfected with overexpression constructs encoding wild type ERK1/2, MEK1, or Raf-1 along with VEGF-A(Ϫ88/Ϫ50)-Luc were analyzed for luciferase activity. Results were expressed as -fold increase of control transfectants, and data shown represent a typical result obtained from a series of three independent experiments. as an inhibitor of Sp1-dependent effects (Fig. 5). To explore the influence of oxidative stress on the transactivation capacity of Sp1 and Sp3, we employed appropriate Gal4-Sp1-or Gal4-Sp3/ Gal4-luciferase cotransfection systems (48,49). Interestingly, we observed that application of oxidative stress enhanced the transactivating capacity of Sp1 but not of Sp3, suggesting potential differences in the regulatory mechanisms controlling the activity of these transcription factors in response to cellular oxidative stress (Fig. 6). Recent studies demonstrated that the transactivating capacity of Sp3 can be suppressed through SUMO-dependent modification of the transcription factor (49,50). To what extent this mechanism also contributes to regulation of Sp3 in context of VEGF-A gene regulation remains to be explored in future studies.
Previous studies demonstrated that depending on the cellular context and/or the stimulus investigated, the VEGF-A promoter can also be bound and transactivated by non-Sp transcription factors (30,34). In contrast to these studies, in gastric cancer cells, no transcription factors other than Sp1 and Sp3 were detected to bind to the VEGF-A(Ϫ88/Ϫ50) region. Initial competition studies using unlabeled oligonucleotides representing consensus binding sites for candidate transcription factors binding to the Ϫ88/Ϫ50 region suggested that AP2 may be interacting with this element (Fig. 4A, lane 13). Detailed analysis of this phenomenon, however, revealed that in AGS cells the AP2 consensus sequence binds Sp transcription factors (data not shown), and therefore, its effect in EMSA competition studies reflects interaction with Sp1 and Sp3 binding.
After identifying the proximal GC-rich site at Ϫ88/Ϫ50 as the critical promoter element mediating VEGF-A oxidative stress responsiveness, we aimed to define the functional properties of three GC-boxes as putative Sp1/Sp3 binding sites located within the Ϫ88/Ϫ50 region. For this purpose, a systematic mutational analysis of these elements was performed in functional transfection studies. In parallel, the influence of these mutations on transcription factor binding was investigated in EMSAs. We found that the structural integrity of the two most 3Ј-located GC-boxes II and III was indispensable for full basal and oxidative stress-stimulated VEGF-A promoter activity, whereas the 5Ј-located GC-box I has only minor importance for VEGF-A gene regulation in gastric cancer cells (Figs. 7 and 9). To our knowledge, this mapping of structure/ function relationships within the human VEGF-A core promoter employing a combination of EMSA techniques and functional promoter studies represents the first detailed evaluation of Sp/XKLF binding sites for basal and stimulated transcriptional regulation of the human VEGF-A gene. A similar approach had currently been used to characterize the importance of the proximal Sp factor binding sites for basal VEGF-A gene regulation in pancreatic cancer cells (31). This study, however, focused entirely on stimulus-independent VEGF-A transcription and GC-box mutagenesis, was less complete, not clearly allowing the determination of how individual promoter elements contribute to VEGF-A transcription (31).
Epithelial oxidative stress responses comprise activation of various signaling cascades including the proliferation-associated MAPK-ERK pathway as well as stress-related MAPK-JNK and -p38 kinase cascades (2,(5)(6)(7). In gastric cancer cells, Ras-dependent activation of the Raf 3 MEK1 3 ERK1/2 kinase module has been shown to be a major pathway through which oxidative stress can influence transcriptional responses (37). Our current study shows that H 2 O 2 treatment of gastric cancer cells at a concentration of 1 mM led to increased phosphorylation of ERK1/2, MEK1, p38, and JNK, indicating activation of associated upstream kinases (Fig. 8A). The finding that JNK is hyperphosphorylated in response to H 2 O 2 differs from our previous observations (37) and may be explained by the fact that H 2 O 2 concentrations required for VEGF-A transactivation were at least 2,5-fold lower than the doses at which stimulation of HDC promoter activity was obtained (2.5-10 mM). In full agreement with our previous work, functional studies revealed that the Raf-1/MEK1/ERK1/2 kinase module represents the major pathway through which oxidative stress transactivates the VEGF-A promoter. This view is supported by the finding that interruption of this cascade at different levels resulted in substantial impairment of H 2 O 2 -triggered VEGF-A promoter activity (70 -80% inhibition; Fig. 8B). Conversely, activation of the signaling cascade upstream of ERKs by overexpression of Raf or MEK, respectively, resulted in robust transcriptional VEGF-A responses, which were also obtained after ERK1 and/or ERK2 overexpression (Fig. 8C). Additionally, application of an inhibitory Ras mutant (N15) substantially impaired the effect of H 2 O 2 on the VEGF-A promoter, confirming our previous observation that oxidative stress activates the Raf-1/MEK1/ERK1/2 kinase cascade in gastric cancer cells through (a) Ras-dependent mechanism(s) (37). These findings are compatible with the concept of the Raf/MEK1/ERK1/2 cascade being a critical signaling in VEGF-A gene regulation as shown in other experimental systems including acidic pH challenge of human glioma cells (51), insulin-like growth factor-I stimulation of NIH3T3 fibroblasts (52), and hamster fibroblast models (30). The finding that H 2 O 2 stimulated JNK phosphorylation, whereas functional interruption of JNK signaling by application of a dominant negative MKK4 mutant reduced VEGF-A promoter activity by Յ30% (Fig. 8B), is well compatible with the MKK4/JNK module acting as additional signaling route mediating VEGF-A oxidant responsiveness in gastric cells (Fig. 9). Similar to these findings, evidence for dual regulation of VEGF-A expression through both MAPK-ERK and -JNK pathways has been observed in primary rat astrocytes FIG. 9. Regulation of VEGF-A transcription by oxidative stress. Oxidative stress potently up-regulates release and production of human VEGF-A in gastric cancer cells, and transcriptional activation of the VEGF-A gene promoter represents the underlying core mechanism. Enhanced binding of Sp1 and Sp3 to two GC-boxes at Ϫ73/Ϫ66 and Ϫ58/Ϫ52 represents the core mechanism of oxidative stress-triggered VEGF-A transactivation. In addition, redox-triggered posttranslational modification of Sp1, but not Sp3, represents an additional mechanism of VEGF-A regulation. Activation of Ras 3 Raf 3 MEK1 3 ERK1/2 signaling and to a lesser extent JNK-related pathways are crucial for transmission of oxidative stress effects on the human VEGF-A gene. and glioblastoma cells exposed to ionizing radiation (53) as well as up-regulation of the GADD45 gene in response to UV radiation (54).
Together, identifying oxidative stress as a potent stimulus of VEGF-A gene expression in gastric cancer cells, our results support the concept that changes in the cellular redox status can directly exert regulatory effects on proangiogenic processes. Furthermore, the detailed delineation of pathways and structural elements mediating oxidative stress responsiveness of the VEGF-A gene can help to develop novel therapeutic approaches to target VEGF-A gene expression in benign and/or malignant (patho)biological settings.