The Small GTPases Ras, Rac, and Cdc42 Transcriptionally Regulate Expression of Human Fibroblast Growth Factor 1*

Four distinct promoters (1A, 1B, 1C, and 1D) of fibroblast growth factor 1 (FGF1), spaced up to 70 kilobase pairs apart, direct the expression of alternatively spliced transcript variants (FGF1.A, -1.B, -1.C, and-1.D) that encode FGF1. These FGF1transcripts can be detected in cultured cells as well as in normal and diseased tissues. These transcripts are differentially regulated in a cell-specific manner. To further delineate the biological function of multiple promoter usage by a single gene, we investigated the transcriptional regulation of these promoters by defined signaling pathways associated with cell proliferation and cell survival. Here we show a specific association of two of the FGF1 promoters, 1C and 1D, with signaling cascades of the Ras superfamily of GTPases. A serum-response element, comprised of the Ets and CArG motifs, present in promoter 1D was shown to be the target of distinct signaling cascades; the Ets motif target of Ras, Rac1, and Cdc42 regulation; and the CArG motif target of de novo protein synthesis-independent cascade. Ras and Rac1 also activated theFGF2 promoter. Further, the transcription factor Ets2 synergistically activated FGF1 gene, but notFGF2, in a Ras- and Rac1-dependent signaling pathway. In support of these conclusions high levels of intracellularFGF1 were detected in cells undergoing cytokinesis. Altogether, our results suggest that FGF1 may play a fundamental role in cell division, spreading, and migration, in addition to cell proliferation.

sively in the kidney (2,3). High FGF1 promoter 1B activity was detected in a glioblastoma cell line U1240MG with promoter construct extending up to nucleotide Ϫ540 from the transcription start site. This activity was attributed to a 23-bp ciselement (Ϫ489 to Ϫ467). A 37-kDa protein, designated p37 brn , was found to associate within this sequence and postulated to positively regulate expression of FGF1.B transcript in the brain (5)(6). In contrast, a basic helix-loop-helix protein, E2-2, negatively regulates the 1B promoter (7). Also, FGF1 promoter 1C construct extending up to Ϫ1614 from the transcription start site displayed activity in PC-3 prostate carcinoma cells and serum inducibility in MDA-MB-231 breast carcinoma cells. The promoter contained cis-elements that included activator protein 1, activator protein 2, and Sp1. These elements showed specific association with nuclear proteins (8). In this study, we have further dissected the signaling cascades involved in FGF1 gene regulation. Here we report molecular mechanisms that explain the differential regulation of these promoters, particularly promoters 1C and 1D, and further provide insight into the biological role of multiple FGF1 promoters.

Southern Blot Analysis
This analysis was performed using standard protocol (9). Tissue and cells expressing variant FGF1 mRNAs were used, i.e. high levels of FGF1.A in human kidney and weak levels in human embryonic kidney 293, high levels of FGF1.B in U-1240 MG and FGF1.C and -1.D in saphenous vein smooth muscle cells and embryo lung fibroblast cells (M426; 2-4). Radiolabeled DNA probes specific for the 1A, 1B, 1C, and 1D promoters were: 2.4-kbp EcoRI-EcoRI containing the 1A promoter, 0.89-kbp HindIII-XhoI containing the 1B promoter, 1.2-kbp BglII-BglII containing the 1C promoter, and 1.2-kbp EcoRI-BglII containing the 1D promoter.
FGF1D Luciferase-The previously described FGF1D minigene (10) was used as DNA template to polymerase chain reaction amplify the FGF1D promoter region Ϫ150 to ϩ40, and including 9 bp of the exon 1 sequence. The primers used were vector primer T3 (5Ј-AAT TAA CCC TCA CTA AAG GG-3Ј) and gene specific primer HBGF D-1 (5Ј-CTT TCA AGA CCA GGT CTT AGC CCA A-3Ј) that overlapped the splice junction of exon Ϫ1D and exon 1 and included 16 nucleotides of exon -1D and 9 nucleotides of exon 1. The amplicon was gel-purified (Mermaid®, Bio 101) and ligated upstream of the luciferase gene in a ligation reaction with pGL2-Basic vector that had been initially digested with HindIII, filled-in with dNTPs and Klenow fragment, and subsequently processed to generate 3Ј-T-overhangs (11). This manipulation facilitated ligation of the polymerase chain reaction product, having 3Ј-A-overhangs, and further recreated the HindIII site. The reporter construct having the 1D fragment in the correct orientation was selected by diagnostic restriction enzyme digestions. The unwanted DNA sequences between the 1D fragment and multicloning site of the vector were removed by further digestion with XhoI and religation. DNA sequencing confirmed the authenticity of this construct. A 1.2-kbp EcoRI-BglII genomic DNA fragment containing the 1D promoter was linearized at EcoRV (present in the vector pBluescript KS(ϩ) upstream of the EcoRI site) and then digested with NcoI to release a 900-bp fragment. This fragment was ligated to the Ϫ150 to ϩ40 1D fragment in pGL2-Basic that had been digested with SmaI (present in the multicloning site, upstream of Ϫ150) and NcoI (present within the Ϫ150 to ϩ40 region of 1D). This generated the Ϫ985 to ϩ40 1D luciferase reporter.
The Ϫ580 to ϩ40 1D luciferase reporter construct was generated by digesting the 1.2-kbp EcoRI-BglII genomic DNA fragment containing the Ϫ1D promoter region with HpaI and NcoI (Ϫ545 to Ϫ82) to release a 460-bp fragment. The gel-purified fragment (Geneclean II, Bio 101) was ligated to the Ϫ150 to ϩ40 1D fragment in pGL2-Basic digested with NheI (present in the multicloning site, upstream of Ϫ150), bluntended with dNTPs and Klenow fragment, and subsequently digested with NcoI (present with the Ϫ150 to ϩ40 region). This construct was propagated in STBL2™ cells (Life Technologies, Inc.), as propagation in the routinely used Escherichia coli strain DH5␣ (Life Technologies, Inc.) cells resulted in deletion in the plasmid DNA. Utilization of STBL2™, suitable for unstable DNA sequences (12), eliminated the problem of deletion, and the correct, intact DNA was recovered.
The Ϫ545 to ϩ40 construct was linearized at SacI (present in the multicloning region, upstream of HpaI, Ϫ545) and subsequently digested with BstXI (Ϫ339) or PstI (Ϫ277), treated with T4 DNA polymerase and dNTPs to generate blunt ends, gel-purified (Geneclean II, Bio 101), and ligated. Again, STBL2™ cells were used for transformation to obtain the correct constructs.
Mutagenesis of Promoter 1D ETS and CArG Sites-The 1D wild type ETS consensus sequence 5Ј-GGA TG-3Ј (Ϫ172 to Ϫ167) and CArG consensus sequence 5Ј-CCA AAT AAG G-3Ј (Ϫ164 to Ϫ155) were mutated using the approach described by Deng and Nickoloff (13; CLON-TECH). The primers used were: ETS-MUT 5Ј-TAG GGT TGG Gtt ccG TGT CCA AAT AAG GCT TGC TCG-3Ј, CArG-MUT 5Ј-GGT TGG GGG ATG TGT aaA Acg Aat tCT TGC TCG AGG 3-Ј and selection primer 5Ј-GCA GCC ACT aGT AAC AGG ATT-3Ј. The lowercase letters denote the mutated nucleotides with the underlined sequence denoting an EcoRI site for CArG-MUT and an SpeI site for the selection primer. The single nucleotide change from guanosine to adenosine, converts an AlwNI restriction enzyme site, present once in the wild type pGL2-Basic vector sequence, to a SpeI site. This manipulation allowed enrichment and selection of the mutant clones by AlwNI restriction digestion.
pCH110 -The ␤-galactosidase-expressing vector, used generally in transient transfections as an internal control, comprises the E. coli lacZ gene driven by the simian virus 40 early promoter (15).
FGF2 Luciferase-pF2.0CAT (Ϫ1800 to ϩ179; 16) was digested with SalI, filled in with dNTPs and Klenow fragment, and then digested with XhoI. The FGF2 promoter fragment was ligated into pGL2-Basic digested with SmaI and XhoI.

Transient Transfections
All plasmid DNA for transfections were prepared using Qiagen plasmid purification kit (Qiagen, Inc.). NIH/3T3 cells (2 ϫ 10 5 ) were seeded on 60-mm plates and allowed to grow for at least 24 h before transfection. Transfections were performed using 30 l of liposomal transfer reagent DOTAP (Roche Molecular Biochemicals) and DNA in a total volume of 200 l (20 mM HEPES, 150 mM NaCl, pH 7.4). This mixture was allowed to incubate for 30 min at room temperature to facilitate stable complex formation between DNA and DOTAP. This was followed by transfer to the cells in 1.5 ml of culture medium containing 10% CS. After incubation at 37°C for at least 20 h, the medium was removed and replaced with fresh medium containing 0.5% serum. Cells were subsequently treated as follows.
Cycloheximide Induction Studies-Cells were incubated for 34 h followed by a medium change. The fresh medium comprised of either 0.5% serum only, 10% serum only, 0.5% serum plus cycloheximide (10 g/ml tissue culture grade in ethanol; Sigma), or 10% serum plus cycloheximide. After 12 h of incubation the medium was removed, cells were washed twice with fresh medium containing 0.5% serum only, followed by an additional incubation of 2 h in medium with 0.5% serum (cycloheximide-washout). The cells were washed twice with phosphatebuffered saline and subsequently harvested using lysis reagent (Promega). The reporter activity was determined using a luminometer (Lumat LB 9501, EG &G Berthold, Berthold Systems Inc., Pittsburgh, PA). The ␤-galactosidase activity of the internal control was measured using the same instrument using a chemiluminescent assay (Galacto-Light™, Tropix, Bedford, MA). Transfections were performed in duplicate or in triplicate, and the data were expressed as the ratio of reporter activity versus ␤-galactosidase activity. One g of pCH110 and 4 g of reporter were used for all experiments. Sera from various companies, including Life Technologies, Inc., BioWhittaker, and Hyclone, were tested. HyClone Cosmic Calf™ serum gave the best results and was used in all subsequent experiments.
GTP/GDP-binding Protein Regulation Studies-Cells were incubated in 0.5% serum for a total of 48 h with a medium change once at 24 h. Cells were harvested, and luciferase activity and ␤-galactosidase activity were determined. For the Rho protein studies, 1 g of pCH110, 4 g of reporter, and 1 or 3 g of EFRhoA-V14, Rac1-V12, and Cdc42Hs-V12 plasmids each, were co-transfected, and specific dose effects were noted. The truncated promoter constructs Ets WT/CArG MUT (Ϫ179 to ϩ40), and Ets MUT/CArG WT (Ϫ183 to ϩ40) showed a similar trend compared with the SRE mutations in the context of the (Ϫ985 to ϩ40) construct. The expression plasmid EFplink was used to make up the total amount of DNA to 8 g.
Ras responsiveness was initially determined by looking at the difference in reporter activities without and with effector Ras (expression plasmid Ϫ/ϩ Ras gene, pHO6, and pHO6T1; 17). A positive control pRDO53 showed greater than 40-fold Ras responsiveness, whereas the FGF1C and -1D promoters each showed ϳ14-fold Ras responsiveness. The FGF1A and -1B promoters showed a nonspecific response of 3.5 and 3.1, respectively, which was comparable to the promoter-less vector, pGL2-Basic. Subsequently, the Ras responsiveness was re-examined by titrating the Ras expression plasmid (pDCRRasV12; 18), with similar results. Expression plasmids (0.125 or 0.5 g) with cDNAs encoding RasV12, RasV12C40, or RasV12S35 were used along with 1 g of HSV-␤-galactosidase expression plasmid and 4 g of reporter plasmid. The expression plasmid pDCR was used to make up the total amount of DNA to 6 g.
Mek WT, Mek MUT, and Rac1 MUT Studies-Cells were incubated in 0.5% serum for a total of 48 h with a medium change once at 24 h. Ras responsiveness (pDCRRasV12, 0.5 g) of luciferase reporters (4 g) was examined in the presence of wild type Mek1 (Mek WT, 0.5 g; 19) or constitutively active dominant negative Mek1 (Mek MUT, 0.5 g; 19). HSV-␤-galactosidase expression plasmid (1 g) was used in all experiments. Also, the effect of dominant inhibitory Rac1 (Rac1 N17, 0.5 g; 20) was tested in the presence of Ras. The Mek MUT gave a specific dose effect (inhibitory) when tested with 0.125 and 0.5 g of expression plasmid, whereas Rac1 N17 showed no such effect. The expression plasmids for Ras (pDCR), Mek1 (pcDNA3), and Rac1 N17 (pEXV MYC) served as controls and were also used to adjust the total amount of DNA to 6 g. Cells were harvested, and luciferase activity and ␤-galactosidase activity were determined.
Ets Studies-Cells were incubated in 0.5% serum for a total of 48 h with medium change once at 24 h. Luciferase reporter constructs (4 g each) were co-transfected with 0.5 g of expression plasmid for Ets2 or PEA3 transcription factors (21)(22) or control expression vector (pCGN) in the presence of Ras (0.125 g), Rac1 (1 g), and Cdc42 (1 g), along with 1 g of HSV-␤-galactosidase expression plasmid. The total amount of DNA was adjusted to 6 or 7 g with empty expression plasmid. The amounts of activated GTPases indicated above showed submaximal promoter responses, which was further increased specifically for FGF1 promoters 1C and 1D by Ets2. The effect of Ets transcription factors on luciferase reporters was also tested in the absence of activated GTPases (4-g reporter plasmid, 1-g HSV-␤-galactosidase expression plasmid, expression plasmid for Ets2 or PEA3 (0 g or 0.5 g), and pCGN (1.0 g or 0.5 g)). Cells were harvested, and luciferase activity and ␤-galactosidase activity were determined. For all transfection experiments, at least two independent experiments in duplicate or triplicate were performed.

Synchronization and Immunohistochemistry of NIH/3T3 Cells
NIH/3T3 cells were synchronized using serum starvation as described previously (23). Briefly, cells were seeded on 8-chamber slides at a density of 5 ϫ 10 3 /chamber, and allowed to grow for four days to confluence. On day 5, confluent cells were starved with Dulbecco's modified Eagle's medium containing 0.5% CS for two days. CS was added to a final concentration of 20% on day 7. At indicated intervals after serum stimulation (0, 2, 6, 16, and 24 h), cells were fixed with ice-cold acetone for 2 min and prepared for immunohistochemistry. The slides were treated with 3% H 2 O 2 to block endogenous peroxidase followed by phosphate-buffered saline rinses (3 ϫ 5 min). The cells were then incubated in blocking sera for 30 min at room temperature and then incubated in anti-FGF1 (R&D, 1:50 dilution) antibody followed by phosphate-buffered saline rinses (3 ϫ 5 min). After incubation with secondary antibody at room temperature for 45 min and phosphatebuffered saline rinses (3 ϫ 5 min), the cells were reacted with the ABC reagent for 45 min at room temperature. The color reaction was performed using AEC chromagen (Vector Labs), and the cells were counterstained with hematoxylin.
For each time point, the numbers of FGF1 positive and negative cells during the cytokinesis phase of mitosis were scored by counting five separate fields at 400ϫ magnification. Each field contains ϳ400 -500 cells. The percentages of FGF1 positive cells as well as the total number of mitotic cells at each time point were thus determined.

FGF1 Promoter 1D Is Uniquely Superinduced by Serum/
Cycloheximide-To rule out the possibility of gene reorganization as a potential mechanism for cell-specific expression and regulation, the critical regulatory regions were examined by Southern blot analysis using probes specific for each of the four FGF1 promoters (Fig. 1). The analysis was performed on genomic DNA from tissues and cells expressing variant FGF1 transcripts. No rearrangement was noted in any of the genomic DNA analyzed (Fig. 1). A polymorphism was noted in the promoter 1D region in DNA isolated from smooth muscle cells (Fig. 1). These data suggested that deletion or rearrangement of regulatory regions, as seen in immunoglobulin gene rearrangement, was not the mechanism of regulation. These data further suggested that the differential regulation of FGF1 mRNAs was probably because of cell-specific transcriptional regulation. To test this hypothesis, we pursued our previous observation of FGF1.D mRNA superinduction by serum in the presence of the protein synthesis inhibitor cycloheximide noted in smooth muscle cells and fibroblasts (4). This superinduction was reminiscent of a phenomenon previously reported for c-fos and ␥-actin and was mediated via a cis element identified as the CArG box, also known as the core element of the SRE (24). The functionality of promoter 1D was measured using luciferase reporter activity in transient transfection assays. In comparison to FGF1 promoters 1A, 1B, and 1C, promoter 1D was uniquely superinduced and behaved similarly to the endogenous gene (Fig. 2, CS/cycloheximide versus Starved and CS). Furthermore, heat inactivation of serum at 56°C for 30 min and repeated freezing-thawing of serum stocks dramatically affected the extent of this inducibility (data not shown). The data showed that (i) cycloheximide alone was not the inducing agent, and (ii) cycloheximide inducibility may depend on serum component(s) that are heat labile. Altogether, these observations supported the notion of a transcriptional enhancement of promoter 1D activity with serum in the presence of cycloheximide via a signaling pathway activated by serum component(s) and independent of de novo protein synthesis. Because CS was unable to superinduce promoter 1D, negative factors likely were involved in promoter 1D expression. That these factors were labile was demonstrated by the induction of 1D by cycloheximide alone. Thus, the combination of CS and cycloheximide elicits the phenomenon of superinduction.
FGF1 Promoter 1D Superinducibility Is Mediated by a CArG Motif-To identify the promoter 1D cycloheximide inducible element, reporter constructs with various 5Ј-truncations were tested in transient transfections. These experiments narrowed down the element to nucleotides Ϫ277 and Ϫ150, as reflected by a decrease in promoter inducibility by 43% (Fig. 3A). Sequence analysis of promoter 1D revealed a putative SRE comprising a CArG element between Ϫ164 and Ϫ155, along with an adjacent Ets element between Ϫ172 and Ϫ167. Thus, the SRE was a good candidate for site-directed mutagenesis. Mutagenesis of the SRE was based on previously described contact points of Ets serum response factors with the c-fos SRE (25). Indeed, mutation of the CArG element, but not the adjacent Ets  promoters (1A (Ϫ826 to ϩ77), 1B (Ϫ831 to ϩ31), 1C (Ϫ786 to ϩ88), and 1D (Ϫ985 to ϩ40)) and HSV-TK basal promoter (TK, Ϫ80 to ϩ50) driving the luciferase reporter gene were individually tested using transient transfections. The promoterless pGL2-Basic (pGL2) vector was also used as a control for background activity. The average of triplicate sets with standard error is shown. site, significantly attenuated (up to 74%) the serum/cycloheximide effect giving activity comparable to the Ϫ150 promoter construct (Fig. 3). Thus, a signaling pathway independent of de novo protein synthesis was activated by serum component(s) and was linked to FGF promoter 1D via the CArG box component of a serum response element. Consequently, cycloheximide alone leads to increased promoter activity (see Fig. 2, 1D, cycloheximide versus Starved). The CArG element plays an important role in transcription in several muscle-specific promoters (25). To test the function of FGF1 promoter 1D CArG element in muscle-specific expression, we tested the activity of FGF1 promoter 1D in proliferating myoblasts. The results indicated that promoter 1D was indeed active in this cell type. Promoter 1D mutants showed that the SRE played a significant role in promoter activity, with the CArG element displaying a more significant role than the adjacent Ets site; compared with wild type, there was a 79% drop in promoter activity with CArG MUT/Ets WT versus a 34% drop for Ets MUT/CArG WT. Altogether, these data suggested that the SRE played a crucial role in regulating promoter 1D activity.
Signaling by Small GTPases Is Linked to the Induction of FGF1 Promoters 1C and 1D-To further elucidate the signaling circuit for promoter 1D regulation, the involvement of Rho family GTPases was examined. The rationale for this study was based on a report that showed the c-fos SRE was the nuclear target of regulation by Rho family GTPases Rac1, RhoA, and Cdc42 (14). These activated small GTPases, along with Ras, were tested in transient co-transfections. The results showed that Rac1 and Cdc42 were involved in the regulation of two FGF1 promoters; Rac1 and Cdc42 activated 1D expression, whereas Rac1 activated 1C (Fig. 4A). Rac1 also activated the FGF2 promoter (Fig. 4C). The c-fos SRE (five tandem copies) was used as a positive control and showed response to Rac1, Cdc42, and a modest response to RhoA (Fig. 4C). This response to RhoA was not observed for any FGF1 promoter nor for the FGF2 promoter (Fig. 4C). Interestingly, the c-fos SRE in the context of its own promoter showed only a weak response to activated GTPases (Fig. 4D). The result underscores the robustness of the FGF1 and FGF2 promoter responses. Further, Ras was involved in regulation of both promoters 1D and 1C, as well as the FGF2 promoter (Fig. 5A). We also examined the bladder carcinoma cell line J82 overexpressing oncogenic Ras (stable transfections). The results showed a 3-4-fold increase in steady state levels of FGF1.D mRNA compared with control; another bladder carcinoma cell line EJ also expressed detectable levels of 1D mRNA as determined by RNase protection assays (data not shown). High levels of FGF1.C mRNA were not detected in these cells, which is consistent with the kinetics FIG. 3. Identification of the promoter 1D cycloheximide inducible cis element in NIH/3T3 cells. Transfections were performed as described under "Experimental Procedures," with 10% serum, in the absence or presence of cycloheximide (Ϫ/ϩ). A, the wild type 5Ј-truncated FGF1D promoter constructs (Ϫ985 to ϩ40, Ϫ545 to ϩ40, Ϫ339 to ϩ40, Ϫ277 to ϩ40, and Ϫ150 to ϩ40) were tested. The results allowed localization of the cycloheximide inducible element to nucleotides Ϫ277 and Ϫ150. B, site-directed mutagenesis of CArG box between Ϫ164 and Ϫ155 (CArG MUT) and not the flanking Ets site (between Ϫ172 and Ϫ167, Ets-MUT) leads to a loss of cycloheximide inducibility. These mutations were introduced in the Ϫ985 to ϩ40 reporter construct. The average of duplicate sets with standard error is shown. The relative reporter activity is shown with the activity of (WT) Ϫ985 to ϩ40 construct taken as 100%. The pGL2-Basic vector was used as a control for background activity.

FIG. 4. Rho family GTPases regulate FGF1 gene expression in NIH/3T3 cells.
A, activated Rho proteins, Rac1-V12 and Cdc42-V12, are involved in promoter 1D regulation. FGF1 promoter regions tested were: 1A (Ϫ826 to ϩ77), 1B (Ϫ831 to ϩ31), ⌬1C (Ϫ786 to ϩ88), 1C (Ϫ1601 to ϩ88), or 1D (Ϫ985 to ϩ40). Data using 3 g of effector plasmids are shown. B, mutation of the SRE Ets site of the 1D promoter region (Ϫ985 to ϩ40) impairs Rac1 and Cdc42 regulation. Here the relative reporter activity (in percent) of the WT sequence is compared with the activity of the reporter with the CArG mutation (CArG-MUT) or the Ets mutation (Ets MUT) in response to 3 g of effector plasmid. The average of duplicate or triplicate sets with standard error is shown. C, Rho protein gene regulation of FGF2, c-fos SRE (five copies in tandem array), and (D) 0.75-kbp c-fos promoter containing a single copy of the SRE. The pGL2-Basic vector was used as a control for background activity.
of expression of this transcript, i.e. acute versus transient (4). These results supported the involvement of Ras in FGF gene expression. Further analysis revealed that the nuclear target of promoter 1D regulation by Rac1, Cdc42, and Ras was the Ets site, which was indispensable for this signaling (Fig. 4B and  5B). Notably, compared with the WT, CArG MUT response to Cdc42 and Ras was diminished by 33 and 27%, respectively.
Ras signaling includes at least two pathways, one regulating gene expression in the nucleus (MAP kinase pathway, which involves Mek1) and the other modulating actin cytoskeleton organization (cell morphology pathway, which involves Rac1; 18). Both pathways are necessary for the mitogenic effect of Ras. Dominant inhibitory Mek1 (Mek MUT/Ser 222 to Ala (S222A) (19)) partially blocked Ras induction of FGF1 promoters 1C and 1D, as well as FGF2 (Fig. 5C; 50% inhibition for 1C, 40% inhibition for 1D, and 31% inhibition for FGF2). However, dominant inhibitory Rac1 (Rac1 N17; 20) was unable to suppress this induction to any extent, and in fact the data showed a slight increase in promoter 1C and FGF2 activity by 28 and 15%, respectively (Fig. 5C). We also tested two Ras effector pathway mutants; RasV12C40 is deficient in MAP kinase pathway signaling but efficient in inducing membrane ruffling, whereas RasV12S35 is defective in inducing membrane ruffling but effective in inducing the MAP kinase pathway signaling (18). Both Ras mutants failed to induce FGF1C, -1D, FGF2, and c-fos SRE promoters. Unexpectedly, co-transfections with both RasV12C40 and RasV12S35 failed to rescue gene expression (data not shown). These data suggested that in NIH/3T3 cells, unlike in rat embryo fibroblast cells (18), both amino acid residues 35 and 40 of Ras are required in cis for proper effector interaction(s). Collectively (i) the Ets site-binding protein is a target of regulation and (ii) the Ets and CArG-binding protein cooperation does not appear to be important. This regulation is unlike the previously proposed model of a ternary complex formation at the SRE that requires Ets factor binding along with the serum response factor for promoter activation by the Ras-Raf-MAP kinase pathway. It is also in sharp contrast to the c-fos regulation by Rac1 and Cdc42 GTPases that require only the CArG/serum response factor interaction and not the adjacent Ets site (14).
Divergent Signaling Pathways Converge on the SRE of FGF1 Promoter 1D-We examined the involvement of Mek in the fetal bovine serum/cycloheximide superinduction of promoter 1D by using the dominant inhibitory Mek1 mutant (Mek MUT/ Ser 222 to Ala (S222A)) and wild type constructs (Mek WT). The results showed that Mek MUT had no inhibitory effect on the superinduction with the Mek WT displaying up to 17% inhibition, when compared with vector control (taken as 100%). A pitfall in this experiment may be the use of cycloheximide itself, i.e., cycloheximide may confound these results by interfering with the expression of MUT plasmids. To address this possibility, we repeated the experiments with a specific Mek 1 blocker PD98059 (26). As a control, 5X SRE-driven luciferase reporter was also used. The Ets motif in this c-fos SRE is the target of regulation by the MAP kinases (27). The results showed that PD98059 inhibited the superinduction weakly (8.7 and 4%, compared with control with vehicle, taken as 100%, when tested at two concentrations of 2 and 10 M, respectively). The IC 50 of PD98059 for Mek1 activity is 2 M. Also, the 5X SRE control showed a similar trend, with 6.6 and 26.6% inhibition (compared with control with vehicle, taken as 100%). This effect was consistent with the accepted role of the CArG motif and not the Ets motif, in participating in the superinduction (24,25). We also examined the effect of SB202190, a specific inhibitor of the p38␣ and p38␤ stress-activated protein kinase (28). When used at 4 M (14-fold above IC 50 of 280 nM), this inhibitor showed no effect on the superinduction. Altogether, these results demonstrate that the Ets and CArG motifs of promoter 1D are targets of divergent signaling pathways.
The Transcription Factor Ets2 Synergizes Ras and Rac1 Responsiveness of FGF1 Promoters-The involvement of the Ets family of transcription factors was directly examined, specifically Ets2 and PEA3. The Ras-Raf-MAP kinase-dependent selective phosphorylation of a conserved threonine in the pointed domains of Ets1 and Ets2 enhances their trans-activation potential (21,22). When transfected singly, Ets2 (0.5 g) provided a 2.7-and 1.7-fold induction for the FGF1 promoter 1C and 1D, respectively (data not shown), whereas RasV12 (0.5 g) provided a 8.3-and 4-fold induction for promoters 1C and 1D, respectively (Fig. 5C). To demonstrate the possible synergistic effect of Ras and Ets2, 0.5 g of Ets2 was co-transfected with a suboptimal amount of Ras (0.125 g). Ets2 demonstrated another 3.5-and 2-fold induction for promoters 1C and 1D, respectively, over cells transfected with Ras alone (Fig. 6A). These results suggested a synergistic effect of Ras and Ets. A similar effect was noted for Rac1 and Ets2 (Fig. 6, B and C). Mutation of the Ets site in promoter 1D attenuated this effect (Fig. 6C). This effect was, however, not observed for FGF2 with Ets2 alone (data not shown) or with Ras plus Ets2 and Rac1 plus Ets2 (Fig. 6, A and B). In contrast, PEA3, another member of the Ets family of transcription factors, showed no synergistic effect with Ras, Rac1, or Cdc42 (Fig. 6, A-D). Collectively, these results underscore the biological significance of Ets protein(s) in FGF1 gene regulation.
FGF1 Is Expressed at High Levels during Cytokinesis-Finally, we examined the profile of endogenous FGF1 expression in NIH/3T3 cells. FGF1 expression was monitored in cells synchronized by serum starvation and then stimulated to enter cell cycle through serum stimulation. FGF1 was not detected in serum-starved cells (0 h), nor at 2 h of serum addition (Fig. 7,   A and B). However, at 16 h after serum addition, FGF1 was detected specifically during the cytokinesis phase of mitosis (Fig. 7C), with as high as 98.3% cells expressing FGF1 (Table  I). This trend continued up to 24 h after serum addition (Fig.  7D), although the percentage of FGF1 positive cells decreased slightly (90.5%).

DISCUSSION
In this study we have identified two of the four FGF1 promoters (1C and 1D) to be specifically under the regulation by the Ras superfamily GTPases Ras, Rac1, and Cdc42. These findings combined with our previous reports showing that the brain-specific FGF1 promoter 1B is under the control of other transcription factors, E2-2 and p37 brn (6, 7), provide definitive evidence for a single gene to be regulated by distinct signaling mechanisms through different promoters. This type of regulation sets a new paradigm for gene expression. The molecular basis of the FGF1 gene expression, which we uncovered in this report, further emphasizes the necessity in fine tuning the pleiotropic effects exerted by FGF1.
Altogether, the results presented in this study provide a unique perspective into overlapping yet discrete regulatory pathways of the two closely related growth factors, FGF1 and FGF2. These results link signaling pathways previously shown to be associated with cell proliferation (18) and cytoskeletal organization (29,30) to growth factor gene activation and identify a cis element in FGF1 promoter 1D that can be defined as an SRE. This SRE is comprised of Ets and CArG sites. These sites are targets of regulation by divergent signaling pathways; Ets site target of Ras, Rac1, and Cdc42 regulation; and the  7. FGF1 expression during cytokinesis phase of mitosis. NIH/3T3 cells were synchronized as described under "Experimental Procedures." The cells at indicated time points following serum stimulation were reacted to antibodies to FGF1 and counterstained with hematoxylin. The quantification of FGF1 positive cells is shown in Table I. CArG site target of de novo protein synthesis-independent cascade. This cascade does not involve Mek1 nor p38 MAP kinase. This cascade may have particular biological relevance in protection of tissue via acute ischemic preconditioning (31). Thus, temporal expression of FGF1 upon brief periods of hypoxia may provide a protective effect from detrimental effects of prolonged hypoxia, consistent with the role of FGF1 as a survival factor.
Our results also demonstrate regulation of the FGF1 gene by the Ets family of transcription factors. Indeed the FGF1 promoter 1D constructs with Ets mutations have low level basal activities. As shown in Fig. 4B, the percentage of induction for the Ets mutant by Rac1 and Cdc42 is 43 and 114%, respectively. These figures are not high in comparison to those for the wild type promoter, 262 and 488%, and for the CArG mutant, 248 and 303%, respectively. These results clearly indicate that the Ets element plays a significant role in Rac1 and Cdc42 induction. This observation is also reflected in Fig. 6C, demonstrating synergy between Ets and GTPases. In contrast, Ras manifests a less dramatic effect than Rac1 and Cdc42 (Fig. 5B, also Fig. 6A). These observations are in agreement with previous reports describing phosphorylation of Thr 72 of Ets2 by Ras, which is necessary for increasing the trans-activation potential of this transcription factor (21). That this effect was not because of overexpression in our transient assays was demonstrated by selective activation by Ets2 and not PEA3. Together, these results directly link FGF1 function to the established role of Ets in development and angiogenesis (32).
Ras superfamily GTPases regulate growth, differentiation, as well as link cell surface receptors to organization of the actin cytoskeleton. These GTPases regulate actin dynamics and regulate fundamental processes such as cell movement, cell cycle progression, cytokinesis, as well as gene expression in the nucleus (29,30). Rac1 and Cdc42 are wound-activated and are linked to actin-polymerized structures lamellipodia and filopodia formation, respectively. Rac1 is essential for producing leading edge protrusions, necessary for forward movement in the wound; Cdc42 is important for maintaining the polarized phenotype of migrating cells, whereas Ras is involved in stress fiber formation and focal adhesion turnover and also is required for cell movement. Altogether, it is the coordinated role of these small GTPases that leads to cell movement. Combined with the data presented in this study, FGF1 is likely involved in this fundamental process.
Not all GTPases regulate FGF1, and this observation further eliminates nonspecific effects because of overexpression. For example, RhoA showed no effect on FGF1 nor FGF2 regulation. The transient transfection data were supported by endogenous FGF1 expression in dividing cells (Fig. 7). These observations suggest that intracrine/autocrine signals involved in cell division require FGF1. Because cell migration, cell cycle progres-sion, and cytokinesis involve actin reorganization and GTPases, we suggest that FGF1 is indeed necessary in these fundamental processes.