Transcriptional activation of fibroblast growth factor 1.B promoter is mediated through an 18-base pair cis-acting element.

Four different transcripts encoding fibroblast growth factor 1 (FGF-1, also known as aFGF) have been previously identified in our laboratory. Among them, FGF-1.B is the major transcript expressed specifically in the neuronal cells in brain tissue. Using the transient transfection experiment in a glioblastoma cell line, U1240MG, that expresses 1.B, we previously identified two regulatory regions (RR1 and RR2) in the brain-specific promoter, FGF-1.B. In the present study, we showed that the minimal region required for the DNA-protein interaction in RR2 resides in an 18-base pair (−484 to −467) sequence, by using DNase I footprinting and methylation interference studies and electrophoretic mobility shift assays. This minimal cis-acting element was found to be sufficient in enhancing the reporter activity driven by the heterologous herpes simplex virus thymidine kinase promoter in the 1.B-positive U1240MG cell line. This enhancing effect, however, was not detected in a glioblastoma cell line, U1242MG, which is negative for 1.B expression. By electrophoretic mobility shift assays, we also identified a specific DNA-protein complex, namely complex I, which is specific for 1.B-positive cell lines and human brain tissue. By in situ UV cross-linking experiment, we further showed that complex I contains two major DNA-binding proteins of apparent molecular masses of 37 and 98 kDa. Our results suggest that the formation of complex I, resulting from the heterodimerization of a 37-kDa protein (1.B-specific) and a 98-kDa protein (ubiquitous) may likely be a prerequisite for the enhanced expression of 1.B transcript in neuronal cells.

The FGF 1 family consists of 14 structurally related polypeptide growth factors (1,2). FGF-1, a prototype member, was originally identified as a mitogen for endothelial cells (3) and subsequently for a variety of mesoderm-and neuroectodermderived cells (4,5). This growth factor is involved in a variety of physiological and pathological function such as tissue growth, wound healing, neovascularization, mesoderm development, and angiogenesis (6 -10). FGF-1 confers tumorigenicity when introduced via expression vectors into normal cells (11)(12)(13). Unlike FGF-2, FGF-1 is found primarily in brain and retina, although low levels of its mRNA have been demonstrated in other tissues (14 -16). FGF-1 may function in establishment of retinal fate and ganglion cell differentiation from the invaginated optic vesicle (17). However, the role of FGF-1 in brain is not yet well understood. By in situ hybridization and immunohistochemical analysis, it has been shown that the expression of this growth factor in brain is exclusively in neural cells but not in glial cells (18,19). It was also shown that the expression of FGF-1 mRNA in neurons is correlated with specific developmental events (20). FGF-1 was reported to coexist with tyrosine hydroxylase, a key enzyme for catecholamine synthesis, in neuronal cells (21) and is involved in the expression of this gene (22). FGF-1 mRNA (23) and the protein (24) levels were also reported to elevate in lesioned rat brain. Interestingly, in gliomas, a major human intracranial tumor (25), FGF-1 mRNA expression was reported to increase significantly (7,26). However, the regulatory mechanism of FGF-1 gene expression is only beginning to be understood.
The human FGF-1 gene spans over 120 kilobase pairs containing three protein coding exons and at least four upstream untranslated exons, namely 1A, 1B, 1C, and 1D (27)(28)(29)(30)(31). Splicing of each of these untranslated exons to the first protein coding exon generates four different mRNA transcripts designated as 1.A, 1.B, 1.C, and 1.D. These multiple transcripts generated by alternate promotor usage and splicing are distributed in a tissue-specific manner. The major transcript in human brain, 1.B, is different from that in human kidney or prostate (31,32). By RNase protection assay, it was shown that 1.B transcripts are highly expressed in glioblastoma tissues, as well as in some of the glioblastoma cell lines (26,32). By in situ hybridization with cRNA probe specific for the murine FGF-1.B transcript, our laboratory has recently shown that 1.B in mouse brain are restricted largely to sensory and motor nuclei in the brainstem, and to the ventral spinal cord and cerebellum (33).
We have previously identified a 41-bp DNA sequence (Ϫ507 to Ϫ467) in the brain-specific FGF-1.B promoter, designated as regulatory region 2 (RR2), through luciferase reporter assays (34). RR2 binds nuclear factors through a 23-bp (Ϫ489 to Ϫ467) region and this binding is linked to the enhanced functional activity of the promoter (34). In the present communication, we have further characterized the functional importance of this region. We have narrowed down the minimal cis-acting sequence to an 18-bp region (Ϫ484 to Ϫ467) which can form more than one DNA-protein complex including the 1.B-specific complex, complex I. This 18-bp sequence is sufficient to enhance the luciferase reporter gene expression driven by a heterologous thymidine kinase (tk) promoter in a cell line-specific manner. UV cross-linking analysis of this 1.B-specific complex re-veals that it contains two major DNA-binding proteins of 37 and 98 kDa.

MATERIALS AND METHODS
Cells-U1240MG, U1242MG, U343MG, and U251MG glioblastoma cells (human glioblastoma cell lines originally established from surgical specimens of malignant gliomas) and the normal fetal glial cell line (CHII) were grown in minimal essential media containing 10% calf serum and antibiotics (Life Technologies, Inc.) as described (26,34).
Transfection and Analysis of Promoter Activity-U1240MG cells were plated in 60-mm Falcon tissue culture dishes (Becton Dickinson Labware, Lincoln Park, NJ) to achieve 60 -80% confluence by day 2. On day 2, cells were transfected with 10 g of plasmid DNA using the cationic lipid transfection reagent, DOTAP (Boehringer Mannheim). Plasmid DNA was prepared using the Qiagen plasmid kit (Qiagen Inc., Chatsworth, CA). In each experiment, all constructs were transfected in duplicate. The results show the mean and standard error (S.E.) of at least two experiments. The transfection protocols were the same as described previously (34). The luciferase activity was normalized as described and was expressed as total light units per milligram of protein (34). The Bio-Rad DC protein assay system was used to determine the amount of protein in the cell lysates.
Construction of the FGF-1.B Insertion Clones-A pair of complementary oligonucleotides containing nucleotides Ϫ492 to Ϫ467 of FGF-1.B promoter were synthesized. The sequence of the sense oligonucleotide is 5Ј-GATCTACGACCTGCTGTTTCCCTGGCAACTCG-3Ј with the underlined portion showing the BglII site, and the antisense 3Ј-ATGCTG-GACGACAAAGGGACCGTTGAGCCTAG-5Ј with the underlined portion showing the BamHI site. The two oligonucleotides were annealed and ligated into the BamHI site of either pBluescriptII KS(ϩ) or pUC19. A clone with four repeating units in the same orientation was selected, double-digested with SpeI and BamHI, resulting in the 104-bp fragment. The 104-bp tetramer fragment was blunt-ended using Klenow enzyme and subcloned into blunt-ended XhoI site of Del-2. Del-2 contains a deletion of nucleotides Ϫ490 to Ϫ467 in the FGF-1.B Ϫ540 promoter reporter construct as described previously (34). The Ϫ540 construct contains nucleotides Ϫ540 to ϩ31 of FGF-1.B promoter sequence in the pGL2-Basic luciferase vector. The resulting tetramer inserted clones were designated as 4U(ϩ) with the tetramer in the sense orientation and 4U(Ϫ) in antisense orientation. A clone containing one repeat unit in the direct orientation cloned into the BamHI site of pUC19 was designated as pFGF-1.B Ϫ492/Ϫ467. The reporter plasmid containing the full-length tk promoter was constructed by inserting nucleotides Ϫ200 to ϩ50 of ptk-LUC (provided by Dr. Ghyselinck;Ref. 35) into the SmaI-SacI sites of pGL2-Basic. This plasmid was designated as tk(Ϫ200). Similarly, a minimal tk promoter fragment (from Ϫ80 to ϩ50) was cloned into pGL2-Basic and was designated as tk. The heterologous promoter-reporter plasmids 26bp/tk and 18bp/tk contain the 26-bp (Ϫ492 to Ϫ467) and 18-bp (Ϫ484 to Ϫ467) of FGF-1.B promoter in the tk, respectively. Four pairs of oligonucleotides with SmaI and KpnI sites representing the 18-bp wild type (WT) sequence (Ϫ484 to Ϫ467) in either orientation (ϩ or Ϫ) and two mutant (mut) sequences (mut 456 and mut 56, Table I) were synthesized, annealed and subsequently cloned upstream of tk plasmid at SmaI-KpnI site. The resulting constructs are designated as 18bp(ϩ)/tk, 18bp(Ϫ)/tk, mut 456/tk, and mut 56/tk. The copy number and orientation of each insert were verified by sequencing.
Preparation of Nuclear Extracts (NE)-NE was prepared from the above cell lines as described before (34) with slight modification. The hypotonically swollen cells were homogenized with a pestle (20 strokes) in a Dounce homogenizer and then treated with Nonidet P-40 (Nonidet P-40) to a final concentration of 0.5%. The nuclei were pelleted by centrifugation for 15 min at 3,300 ϫ g, and the cytoplasmic fraction was removed. The nuclei were resuspended with one packed nuclei volume of low salt buffer, buffer B (20 mM Hepes (pH 7.9), 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.5 mM DTT, 20% glycerol, 0.5 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, and 0.9 unit/ml aprotinin) plus 0.02 M KCl. To this suspension one packed nuclei volume of high salt buffer, buffer B plus 1 M KCl, was added dropwise with slow mixing and rocked at 4°C for 30 min. The extract was centrifuged for 30 min at 4°C at 25,000 ϫ g. The supernatant was dialyzed against 50 volumes of dialysis buffer, buffer B plus 0.1 M KCl for 1.5 h at 4°C. The NE was stored in aliquots at Ϫ80°C following centrifugation for 20 min at 4°C at 25,000 ϫ g. Protein concentration, measured by the Bio-Rad assay using bovine serum albumin (BSA) as standard, showed 4 -6 mg of protein/ml of the NE. NE from human brain tissue and placenta was prepared as described by Gorski et al. (36).
Design of WT and mut Oligonucleotides-The WT oligonucleotide representing the sequences Ϫ489 to Ϫ467 or Ϫ484 to Ϫ467 of FGF-1.B promoter and the mutants are described in Table I. To localize the minimal region of DNA-protein interaction by electrophoretic mobility shift assays (EMSA), we also synthesized the following oligonucleotides, which upon annealing and end-filling represent the sequences Ϫ484 to Ϫ467, Ϫ484 to Ϫ473, and Ϫ480 to Ϫ467. The sequences of the pairs of oligonucleotides are, respectively, as follows 5Ј-GATCTCTGTTTCCCT-GGCAACTCG and 5Ј-GATCCGAGTTGCCAGGGAAACAGA, 5Ј-GAT-CTTCCCTGGCAACTCG and 5Ј-GATCCGAGTTGCCAGGGAA, and 5Ј-GATCTGTTTCCCTG and 5Ј-GATCCAGGGAAACA.
DNA Probes-The pGL2-Basic reporter clones containing Ϫ540 to ϩ31 or Ϫ594 to ϩ31 of the FGF-1.B promoter were digested with BbsI. The BbsI fragment containing Ϫ540 to Ϫ357 or Ϫ594 to Ϫ357 was labeled with [␣-32 P]dCTP by end-filling following gel purification and digested with SmaI. Finally, the SmaI (Ϫ540 or Ϫ594)-BbsI (Ϫ357) fragment was gel-purified. The labeled HindIII-EcoRI fragment containing Ϫ492 to Ϫ467 sequences was obtained by digesting the plasmid pFGF-1.B Ϫ492/Ϫ467 with HindIII and EcoRI, followed by labeling with Klenow enzyme in the presence of [␣-32 P]dATP. The oligonucleotide probes were labeled with [␥-32 P]ATP using polynucleotide kinase following annealing of the complementary strand and purified by passing through spin columns (Sephadex G25).
EMSA-The binding reaction contained the binding buffer (25 mM Hepes, pH 7.9, 1 mM MgCl 2 , 0.5 mM DTT, 40 mM KCl, and 5% glycerol) with radiolabeled DNA probe (30,000 cpm), 2 g of poly(dI-dC), 5 g of nuclear protein, and cold competitor when desired in a total volume of 20 l. Each reaction was initiated by addition of protein and incubated for 30 min at room temperature, and then 12 l was analyzed in a 4% polyacrylamide gel in 0.25 ϫ TBE buffer (22.25 mM Tris borate, 0.5 mM EDTA, pH 8.0). Following electrophoresis, the gel was dried under vacuum and autoradiographed.
Sodium Deoxycholate (DOC) Treatment of U1240MG NE-To exclude the possibility that formation of additional complexes was due to the interactions occurring between the DNA-binding protein and non-DNA-binding proteins, the NE was treated with DOC as described by Chellappan et al. (37). U1240MG NE (50 g) was incubated in ice in a total volume of 10 l with 0.7% DOC (final concentration) for 20 min, and then Nonidet P-40 was added to a final concentration of 1%. One l of this reaction mix was assayed in EMSA in a total volume of 20 l. An equivalent amount of untreated NE was also tested for comparison of the binding activity in the presence of 0.05% Nonidet P-40.
DNase I Footprinting Assay-An aliquot containing 0.3 ng (30,000 cpm) of Ϫ540 to Ϫ357 or Ϫ594 to Ϫ357 SmaI-BbsI fragment of FGF-1.B promoter, labeled at the 3Ј-end at the BbsI site (sense strand), was incubated with increasing amount of U1240MG nuclear extract (12-60 g of protein) or 20 g of acetylated BSA in a total volume of 30 l containing 25 mM Hepes (pH 7.9), 1 mM MgCl 2 , 0.5 mM DTT, 40 -60 mM KCl, 5% glycerol, and 4 g of poly(dI-dC) (Pharmacia). After 30 min of incubation, an equal volume (30 l) of a solution containing 8 mM MgCl 2 and 4 mM CaCl 2 was added, followed by the addition of 10 -25 ng (depending on the amount of nuclear extract used) of freshly diluted DNase I (Boehringer Mannheim). The digestions were carried out for 60 s. The reaction was stopped by adding 140 l of stop buffer (TE plus 0.1 M NaCl, 0.2% SDS), 200 l of phenol, and 200 l of CHCl 3 . The aqueous layer was extracted further with 200 of l CHCl 3 and ethanolprecipitated in the presence of 5 g of tRNA. The precipitates were denatured, and equal counts were analyzed in a 6% acrylamide sequencing gel.
Methylation Interference Assay-The HindIII-EcoRI fragment from plasmid pFGF-1.B Ϫ492/Ϫ467 was 3Ј-end-labeled at EcoRI site (sense strand), partially methylated with dimethyl sulfate, and used as a probe in binding reactions containing U1240MG NE. DNA-protein complexes were separated from the free DNA by EMSA, eluted from the gel and cleaved with piperidine, and then analyzed in an 8% sequencing gel as described (34).
UV Cross-linking Experiment-Analysis of the EMSA complexes was done by in situ UV cross-linking according to the method of Wu et al. (38) with a slight modification. The 32 P-labeled photoaffinity probe used in this experiment was made by annealing two complementary strands (5Ј-GATCTGTTTCCCTG-3Ј and 5Ј-GATCCGAGTTGCCAGGGAAAC-AGA-3Ј) followed by end-filling with dGTP, 5-bromo-dUTP, [␣-32 P] dATP, and [␣-32 P]dCTP by Klenow reaction. The gel shift binding reaction, carried out using the labeled photoaffinity probe (representing Ϫ484 to Ϫ467), was scaled up in the absence or presence of a specific cold competitor. The reaction mixture was run in 1% low melting agarose gel in 0.5 ϫ TAE buffer (20 mM Tris acetate, 1 mM EDTA, pH 8.0) at 4°C. The gel was covered with plastic wrap and exposed to UV light (254 nm) for 20 min at 4°C. The DNA-protein complexes I, II, and III were localized following autoradiography, and the gel slices containing each of these complexes were melted. The labeled DNA cross-linked with the specific DNA-binding protein was concentrated on glassfog beads (Mermaid Kits), eluted by boiling with SDS-gel loading dye. The proteins thus labeled were separated in SDS-polyacrylamide gel electrophoresis and visualized following autoradiography of the dried gel.

Functional Rescue of FGF-1.B Del-2 Luciferase Reporter Construct with Four Tandem Repeats of Ϫ492 to Ϫ467 in Sense Orientation
To strengthen the functional link between the DNA-protein interaction and the FGF-1.B promoter activity, we inserted four tandem repeats of the DNA sequences Ϫ492 to Ϫ467 in the sense (4U(ϩ)) or antisense (4U(Ϫ)) orientation into the Del-2 construct (Ϫ540 to ϩ31 with the deletion of nucleotides Ϫ490 to Ϫ467). These constructs were transfected into U1240MG cells and assayed for the luciferase reporter activity. Fig. 1 showed that the insertion of the four repeats in the sense orientation significantly increased the reporter activity (4-fold) in comparison to that with Del-2 construct. The 4U(ϩ) construct had a 20% increase in luciferase activity in comparison to the wild type construct (Ϫ540 to ϩ31). In contrast, the insertion of the four repeats in the antisense orientation had no effect in comparison to that observed with Del-2 construct. These results suggest that the enhancer activity of the nucleotide sequences Ϫ492 to Ϫ467, unlike the conventional enhancer, is orientation-dependent.

Minimal Sequence Requirement for the DNA-Protein Interaction
To determine the minimal sequence in RR2 required for the specific DNA-protein interaction, we performed (i) DNase I footprinting assay, (ii) methylation interference assay, and (iii) EMSA with overlapping sequences between Ϫ492 and Ϫ467.
DNase I Footprinting Assay-To characterize the minimal cis-acting element for the specific DNA-protein interaction, we used end-labeled (sense strand labeled at its 3Ј-end) DNA from Ϫ540 to Ϫ357 ( Fig. 2A) or Ϫ594 to Ϫ357 (Fig. 2B) in DNase I footprinting assay with increasing amount of U1240MG NE. The results showed a single footprint in the same position (Ϫ485 to Ϫ467) when either probe was used and indicated that the minimal sequence required for the DNA-protein interaction resides in the sequence from Ϫ485 to Ϫ467. Methylation of Guanine Nucleotide at Ϫ482, Ϫ474, and Ϫ473 in the Sense Strand Strongly Affects the Specific DNA-Protein Complex Formation-To identify the precise location of the nucleotides in the sense strand contacting with the nuclear factors, methylation interference analysis using a probe labeled at the 3Ј-end of the sense strand was performed. The previous study could not detect the contact sites in the sense strand (34); this may be due to the full methylation of the genomic fragment (Ϫ507 to Ϫ467) having only 10 guanine nucleotides in the same strand. This problem was overcome by using the DNA fragment (Ϫ492 to Ϫ467) flanked by vector sequences at both ends. It is evident from the results (Fig. 3) that the methylation of guanine nucleotides in the sense strand at positions Ϫ482, Ϫ474, and Ϫ473 strongly affects the specific DNA-protein interaction.
The Minimal cis-Acting Element Required for the Specific

DNA-Protein Interaction Is from
Ϫ484 to Ϫ467-The DNase I footprinting assay (Fig. 2) showed that the site for the DNAprotein interaction lies in the sequences from Ϫ485 to Ϫ467. We used oligonucleotides of different sizes between Ϫ492 and Ϫ467 having overlapping sequence either as cold competitors (Fig. 4A) or as probes (Fig. 4B) in EMSA to determine the minimal sequence requirement for the DNA-protein complex formation with U1240MG NE. In EMSA, we detected three distinct complexes (I, II, and III) with the above NE (Fig. 4A). It is evident from this figure that the sequence Ϫ484 to Ϫ467 is equally effective to compete these complexes as the wild type oligonucleotide (Ϫ489 to Ϫ467) does. The oligonucleotide Ϫ480 to Ϫ467 only partially competes for these complexes, even at 100 molar excess, while Ϫ484 to Ϫ473 does not compete at all. Likewise, when these oligonucleotides were used as probes (Fig. 4B), only Ϫ484 to Ϫ467 as well as Ϫ492 to Ϫ467 could specifically bind to the nuclear factors. The oligonucleotide Ϫ480 to Ϫ467 showed very poor binding, and no specific binding was detected with the Ϫ484 to Ϫ473 probe. These three experiments together determine the minimal cis-acting element for the specific DNA-protein interaction lies from Ϫ484 to Ϫ467.
The Minimal 18-bp cis-Acting element (Ϫ484 to Ϫ467) Is Sufficient to Enhance the Activity of the Heterologous tk Promoter-To address the functional consequence of the DNA-protein interaction with the minimal 18-bp (Ϫ484 to Ϫ467) region and the cell line specificity of such interaction, we tested the activity of the minimal tk promoter in absence or presence of the 18-bp or the 26-bp (Ϫ492 to Ϫ467) DNA sequence of FGF-1.B in U1240MG or U1242MG cell lines. The results showed that the insertion of the 18-bp sequence upstream of the tk promoter enhanced the reporter activity by 6-fold relative to tk (Fig. 5A, left panel). It is also clear that the 18-bp DNA sequence is sufficient to enhance the activity of the heterologous tk promoter, since the 26-bp construct has no further effect on the enhancement of the promoter activity. This enhancement is cell-specific, since neither 18-bp nor 26-bp DNA sequence enhanced the tk promoter activity in U1242MG cells (Fig. 5A, right panel). It is important to note that a comparable level of reporter activity (10-fold) was scored by both U1240MG and U1242MG cells when tk(Ϫ200) was used in transient transfection assay (Fig. 5A). We also assayed the reporter activity of tk promoter in the absence or presence of the 18-bp oligonucleotide in either orientation (Fig. 5B). The results showed that the 18-bp sequence, when placed upstream of the tk promoter in the sense orientation, activated it by 6.5-fold. When placed in the antisense orientation, the activity was reduced to 1.7-fold relative to tk. This result is consistent with the data shown in Fig. 1, demonstrating that the regulatory effect of the 18-bp sequence is orientation-dependent. The insertion of mutant oligonucleotide, mut 456 and mut 56, upstream of the tk (mut 456/tk or mut 56/tk) resulted in a 71% and 57% reduction of the promoter activity respectively relative to 18-bp(ϩ)/tk (Fig. 5B).

Correlation of DNA-Protein Complex I in EMSA with the Expression of 1.B Transcript
Since the activation of the tk promoter through the 18-bp or 26-bp DNA sequence was demonstrated to be cell-specific, we were interested to know whether there is any cell-specific DNA-protein interaction with the same DNA sequence and, if so, whether that correlates with the expression of 1.B. We assayed the NE prepared from five different cell lines (U1240MG, U1242MG, U251MG, U343MG, and CHII) of glial origin for the specific binding of the nuclear protein(s) with the DNA probe having the sequence Ϫ492 to Ϫ467 (Fig. 6). By RNase protection, it has been shown that U1240MG and U251MG are positive for 1.B transcript, whereas the remaining three cell lines are negative (26,32). Fig. 6A showed that three distinct complexes (I, II, and III) were formed with U1240MG NE (lane 1) and these complexes were competed by a specific cold competitor (Ϫ489 to Ϫ467) (lane 2). Notably, the fastest mobility complex, complex I, is exclusively present in the 1.Bpositive cell lines (U1240MG and U251MG) but not in the 1.B-negative cell lines (CHII, U343MG, and U1242MG) (Fig. 6,  A and B). To understand whether there is any tissue-specific expression of similar factors, we also performed EMSA with the NE prepared from human brain and placenta (Fig. 6B). Interestingly, three specific DNA-protein complexes including complex I were also formed with the NE from the human brain (Fig. 6B, lane 5), but the nuclear proteins in the human placenta failed to form any of these complexes (lane 7). These results are consistent with the previous RNase protection studies, which showed the human brain tissue is positive for 1.B (26,32) whereas placenta is negative for FGF-1 mRNA (5).

The DNA-Protein Complexes I, II, and III Share a Common DNA Binding Site
Since the methylation of guanine (G) residues at Ϫ482, Ϫ474, and Ϫ473 in the sense strand affected the formation of three complexes (I, II, and III) (Fig. 3), we were interested to know whether these complexes result from the binding of nuclear factors with the DNA sequence Ϫ492 to Ϫ467 at distinctly different site or an overlapping site. We tested each of these complexes in methylation interference assay and looked for the methylation of particular G residues that affect individual complex. Fig. 7 showed that each of these complexes preferably bound to the labeled and methylated DNA, which had no methylation of the G residues at Ϫ474 and Ϫ473 in the sense strand. In addition, methylation of the G residue at Ϫ482 was also shown to interfere with complex II formation and to some degree with complex I formation. These results indicate that these three complexes share a common binding site including the nucleotides Ϫ474 and Ϫ473. We also made several mutant oligonucleotides of the sequence Ϫ484 to Ϫ467 or Ϫ489 to Ϫ467 considering the G residues important for DNA-protein interaction in the sense strand (Fig. 3) and antisense strand (34) and used them in EMSA (Fig. 8). The sequences of these mutants are described in Table I. We used these mutant oligonucleotides as well as the WT oligonucleotide in EMSA as cold competitors (Fig. 8, A and C) or as probes (Fig. 8B) to compare their abilities to affect the formation of these three complexes I, II, and III. Fig. 8A showed that the WT oligonucleotide as well as mut 1, mut 3, or mut 13 could compete the three complexes I, II, and III (lane 1) to the same extent. The competition was drastically reduced by mutants that contain mutation at nucleotide Ϫ478 including mut 2, mut 12, mut 23, or mut 123. Lane 10 showed the inability of an irrelevant (IR) oligonucleotide to compete these complexes. A comparison of the abilities of these oligonucleotides (WT and mutants) to form specific DNA-protein complexes (Fig. 8B) reveals that the binding ability of the single point mutant mut 1 or mut 3 is virtually the same as that of the WT. However, the complex formation is significantly reduced for the single point mutant mut 2. These results showed that a single mutation at Ϫ478 (mut 2) affects the formation of each of these three complexes to the same proportion. Similarly, mutations at Ϫ482, Ϫ474, and Ϫ473 (mut 456) or at Ϫ474 and Ϫ473 (mut 56) resulted in the loss of ability to compete these complexes (Fig. 8C, Table I). When used as probes, both mut 456 and mut 56 failed to form these specific complexes (Table I). These results suggest that all these complexes may share a DNA binding site including the sequence from Ϫ478 to Ϫ473. The nearby G residues in the sense strand at Ϫ482 and antisense strand at Ϫ484 and Ϫ472 may be involved in the stabilization of these complexes.

DOC Treatment Has Little Effect on the Number of the Complexes Formed with U1240MG NE
Since complexes I, II, and III share a common DNA binding site, we addressed the question whether these three complexes result from protein-protein association of a common DNA-binding protein with non-DNA-binding proteins. For example, the DNA-binding protein E2F forms DNA-protein complexes of different mobilities in EMSA by associating with other proteins that are not directly associated with the DNA. In such a case, these complexes can be dissociated to a single complex in the same assay by prior treatment with DOC (37). We treated the NE with DOC to dissociate the protein-protein association of similar nature (if any) and assayed in EMSA to detect the number of specific DNA-protein complexes. Our results showed that DOC treatment did not abolish any of the three complexes, although their intensities were reduced in general (Fig. 9, lane  4). This observation implies that each of these three complexes results from the binding of DNA directly and that there are more than one specific DNA-binding protein recognizing the sequence Ϫ484 to Ϫ467.

FIG. 6. Complex I is unique for cell lines expressing FGF-1.B transcript.
Panel A, approximately 5 g of NE from different cell lines was assayed in EMSA in absence (Ϫ) or presence (ϩ) of 50-fold molar excess of specific cold competitor (Ϫ489 to Ϫ467) using the oligonucleotide A probe as described in Fig. 4B. Panel B, approximately 5 g of NE from different cell lines or 2-3 g of NE from the brain or placenta was assayed in absence (Ϫ) or presence (ϩ) of 50-fold molar excess of cold specific competitor (Ϫ489 to Ϫ467) using the same probe and condition as described in Fig. 4A. I, II, and III represent the three specific complexes formed with U1240MG NE.

FIG. 7. Methylation interference showing that complex I, II
, and III share a common DNA binding site around the nucleotides ؊474 and ؊473. The methylated HindIII-EcoRI fragment containing nucleotides Ϫ492 to Ϫ467 and labeled in the sense strand was incubated with U1240MG NE. Individual complex (I, II, or III) and free DNA, localized following electrophoresis through polyacrylamide gel, were eluted, treated with piperidine, and analyzed on a 8% sequencing gel. Lane G, partial chemical degradation specific for G residues; lane F, free probe DNA; lanes I, II, and III, bound DNA probe from complexes I, II, and III, respectively. The positions of guanine nucleotides in the promoter important for the complex formation (Ϫ482, Ϫ474, and Ϫ473) and positions Ϫ485 and Ϫ490 are indicated.

Complex I Contains Two DNA-binding Proteins of 37 and 98 kDa
We used U1240MG NE and a photo affinity probe (as described under "Materials and Methods") to analyze the DNAprotein complexes I, II and III by in situ UV cross-linking (Fig.  10). The bromo-dUTP-labeled photoaffinity probe also formed the same three specific complexes I, II, and III (data not shown) as formed by the regular probe Ϫ492 to Ϫ467 (Fig. 6). We analyzed each of these complexes in denaturing polyacrylamide gels. Fig. 10 showed that complex I contains two DNA-binding proteins of apparent molecular masses of 37 and 98 kDa. Complex II contains a single protein of 98 kDa, and complex III has two proteins of 98 and 145 kDa. Based on the protein profile ( Fig. 10) and the mobility of each complex in native gels (Fig. 6), it appears that complex I results from the heterodimerization of p37 and p98. Complex II may likely be a homodimer of p98 and complex III a heterodimer of p98 and p145 (Fig. 11). DISCUSSION The study of the tissue-or cell-specific distribution of FGF-1 transcripts has led to the identification of four different transcripts having the same protein coding exons but different 5Ј-untranslated exons (27)(28)(29)(30)(31). FGF-1.B is the predominant transcript in brain, gliomas, and some glioblastoma cell lines (e.g. U1240MG and U251MG) (26,32). Using U1240MG cell line, we have recently identified two regulatory regions, RR1 FIG. 8. EMSA showing that three specific complexes share a common DNA binding site. Panel A, U1240MG NE (5 g) was assayed for the binding activity using the same probe and condition as described in Fig. 4A in the absence (Ϫ) or presence of cold WT oligonucleotide (B in Fig.  4A) or different mutant oligonucleotides (as described in Table I) or an oligonucleotide (IR) of irrelevant sequence. Three distinct specific complexes are denoted as I, II, and III. Panel B, the WT oligonucleotide and the seven mutant oligonucleotides (described in Table I) were end-labeled and equal counts (30,000 cpm) of each oligonucleotide was incubated with 5 g of U1240MG NE in the absence (Ϫ) or presence (ϩ) of 50-fold molar excess of cold WT oligonucleotide. Panel C, experiments were performed the same way as described in panel A. Ϫ Ϫ a In the sense strand (shown in the table), the cytosine nucleotide at position Ϫ484, Ϫ478, or Ϫ472 (underlined) was replaced by adenine nucleotide (2)(3)(4)(5)(6)(7)(8) and the guanine nucleotide at position Ϫ482, Ϫ474, or Ϫ473 was replaced by thymidine nucleotide (10 and 11). The lowercase letters in the oligonucleotides (9 -11) represent the sequences used for the generation of restriction sites. b Degree of competition or specific binding was scored by visual assessment of the band shift using the oligonucleotides (1)(2)(3)(4)(5)(6)(7)(8) in EMSA (Fig. 8,  A and B) or using the oligonucleotides (9 -11) (Fig. 8, C).

TABLE I Competition and binding strengths of WT and mut oligonucleotides
(Ϫ145 to Ϫ114) and RR2 (Ϫ507 to Ϫ467), in the promoter of FGF-1 gene, which are important for FGF-1.B expression (34). In the same study, we have also shown that the deletion of Ϫ490 to Ϫ467 sequence (Del-2) significantly reduced the reporter activity relative to the wild type promoter (Ϫ540 to ϩ31). In this study, we addressed the question whether the insertion of a tandem repeats of the same sequence into Del-2 in either orientation can functionally rescue the reporter activity above the wild type level. As expected, the insertion of four tandem repeats in the sense orientation (4U(ϩ)) restored the reporter activity above the wild type level. However, the insertion of the same tandem repeats in the antisense orientation (4(Ϫ)) had little effect on the reporter activity relative to Del-2. This result suggests that the sequence Ϫ492 to Ϫ467 positively regulates the FGF-1.B promoter in an orientation-dependent manner. It is noted that other regulatory elements, such as the one for the platelet-derived growth factor B gene, are orientation-specific (38).
The region Ϫ540 to Ϫ467 has been shown previously (34) to be important for the enhanced activity of FGF-1.B promoter. The DNase I footprinting experiment carried out in the present study with two overlapping genomic fragments Ϫ540 to Ϫ357, and Ϫ594 to Ϫ357 revealed that the nuclear protein(s) in U1240MG cells protects a common region Ϫ485 to Ϫ467 in both fragments. The methylation interference study as well as EMSA using overlapping oligonucleotides also established that the minimal region for the DNA-protein interaction resides at Ϫ484 to Ϫ467. It is noted that mutation at Ϫ484, Ϫ478, and Ϫ472 (mut 123), which disrupts the formation of DNA-protein complexes, also reduces the transcriptional activity of the reporter activity by 76% in the context of Ϫ540 promoter of FGF-1.B (34). These results indicate that Ϫ484 to Ϫ467 is an essential element contributing to the promoter activation through the DNA-protein interaction.
To further test the above prediction, we determined the activity of the minimal tk promoter in the absence or presence of the 18-bp (Ϫ484 to Ϫ467) or 26-bp (Ϫ492 to Ϫ467) sequence in U1240MG or U1242MG cells. It was demonstrated that both sequences enhanced the reporter activity driven by the heterologous tk promoter to a similar extent in U1240MG. The data indicated that the 18-bp DNA sequence, which is sufficient for DNA-protein interaction, is also sufficient to enhance activity of the tk promoter. This enhancer activity is cell-specific, since none of these two chimeric promoter-reporter constructs enhanced the tk promoter activity when transfected into U1242MG cells. By contrast, the full-length tk construct (tk(Ϫ200)) activated the reporter activity to the similar extent in both cell lines. The activation of tk promoter in U1240MG cells through the 18-bp element is also site-specific because the mutant oligonucleotides (mut 456 or mut 56), which are unable to form specific DNA-protein complex, failed to enhance the activity of tk promoter. Thus, the 18-bp cis-acting sequence DNA-protein interactions occurred in complexes I, II, and III are diagrammed. The p37 brn protein was detected both in a Southwestern analysis using oligonucleotide Ϫ489 to Ϫ467 as a probe (34) and in complex I (Fig. 10). The p98 protein was detected in all three complexes, while p145 was detected only in complex III. We hypothesize that it is this concerted interaction among the three different proteins and the RR2 that regulates brain-specific transcription of the 1.B. Asterisks indicate the three guanine residues in the antisense strand that directly contact the nuclear factors. Only the sense strand of FGF-1.B promoter is shown. The arrows indicate the major transcription start site for 1.B. The underlined sequence is identical to the corresponding mouse FGF-1.B promoter sequence.
functions not only in the context of the native FGF-1.B promoter (34) but also in the context of the heterologous tk promoter (Fig. 5). The orientation dependence of this 18-bp sequence is, however, unconventional but consistent with the data shown in Fig. 1. Moreover, these results also indicate the presence of cell-specific factor(s) interacting with the 18-bp sequence is important for the enhancement of the promoter activity.
EMSA revealed that complex I is unique when using the NE of U1240MG and U251MG cells; both cell lines are known to express 1.B (26,32). The absence of complex I when using the NE of U1242MG or other 1.B-negative cells supports the view that nuclear factor(s) allowing the formation of complex I may be important for the activation of FGF-1.B promoter. Interestingly, the NE from the human brain tissue also formed three specific DNA-protein complexes including complex I, and this tissue predominantly expresses 1.B (26,32). The human placenta is known to be negative for FGF-1 mRNA (5), and the NE prepared from this tissue failed to show any of these complexes. These data suggest tissue-or cell-specific expression of nuclear factor(s) may be important for the tissue-or cell-specific expression of 1.B.
Among the three complexes (I, II, and III), since complex I appears to be linked with 1.B expression, we attempted to precisely locate the binding site that is crucial for this complex formation. The methylation interference assay using the bound probes eluted from each individual complex revealed that complexes I, II, and III resulted from the contact of nuclear factor(s) with G residues at Ϫ474 and Ϫ473 in the sense strand. In addition, complex II has also been found to contact with the G residue at Ϫ482 in the same strand. EMSA using oligonucleotide with mutation at Ϫ482, Ϫ474 and Ϫ473 (mut 456) or at Ϫ474 and Ϫ473 (mut 56) also shows that the formation of each of these complexes is affected to a similar extent in both cases. Furthermore, EMSA using the oligonucleotide (Ϫ489 to Ϫ467) with mutation at different sites (Ϫ484, Ϫ478, and Ϫ472) also failed to differentiate these three complexes in terms of their specific binding sites and showed that the G residue at Ϫ478 is crucial for the formation of these complexes. These results suggest that these complexes share a common DNA binding site including the sequence Ϫ478 to Ϫ473.
DNA-protein complexes with different mobilities may be formed by the interaction of multiple DNA-binding proteins of different sizes to a common binding site (39) or by the association of DNA-binding protein with other proteins that do not directly bind the DNA (37). Our results showed that more than one DNA-binding protein binds to the sequence Ϫ484 to Ϫ467. In situ UV cross-linking study of these three complexes reveals that complex I contains two DNA-binding proteins of 37 and 98 kDa, complex II contains a single DNA-binding protein of 98 kDa, and complex III contains two DNA-binding proteins of 98 and 145 kDa. Based on the protein profile and the mobility of each complex in native gels, it is most likely that complex I results from heterodimerization of p37 and p98. Complexes II and III may likely be a homodimer of p98 and a heterodimer of p98 and p145, respectively. These results suggest that formation of the activating complex in the FGF-1.B promoter requires both p37 and p98. Remarkably, a DNA-binding protein of 37 kDa has been shown to be present in human brain tissue and U1240MG cells (but not in U1242MG cells) that can specifically bind to the sequence Ϫ489 to Ϫ467 in Southwestern analysis (34). In EMSA, however, we could not detect the specific binding of a 37-kDa protein in U1240MG NE that can alone bind the same DNA sequence. This discrepancy may be due to the variation of the assay conditions used to study the DNA-protein interaction. It is pertinent to mention that a specific faster mobility complex (faster than complex I) was demonstrated in EMSA in the same cell line when 1 g or less than 1 g of poly(dI-dC) was used as nonspecific competitor (data not shown). Moreover, it is significant that a DNA-binding protein of the same size (i.e. 37 kDa) from the same source (i.e. U1240MG cells) and the same binding specificity (34) is present in complex I, which is also unique for 1.B-positive cells or tissues.
Our data, taken together, allow us to propose a testable model (Fig. 11) for the DNA-protein interactions at the RR2 (precisely at positions Ϫ484 to Ϫ467) that may explain the cell line-specific activation of FGF-1.B. Based on the present study, it is likely that a 98-kDa protein binds to a minimal DNA sequence Ϫ484 to Ϫ467 as a homodimer. Two other DNAbinding proteins of 37 and 145 kDa bind to the 98-kDa protein, forming two heterodimers of different mobilities. Based on the finding that complex I is present exclusively in 1.B-positive cells (U1240MG, U251MG, and in human brain), we hypothesize that the expression of the p37 followed by its heterodimerization with p98 resulting in the formation of complex I is a prerequisite for the enhancer activity of RR2. Whether the FGF-1.B promoter commits itself to activation in a given cell may depend on the stoichiometric amounts of p37, p98, and p145. It is likely that the availability of p37 is crucial for the expression of 1.B. Since the identified cis-acting sequence shows no similarity with sequences recognized by any other known transcription factors (GenBank, October 1996 release), it is most likely that p37 represents a novel transcriptional factor, which we tentatively designated as p37 brn . Whether p98⅐p98 homodimer and p98⅐p145 heterodimer compete with p37⅐p98 heterodimer for the common binding site and thereby neutralizing the positive trans-acting effect of the latter remains to be determined. As a corroboration to the model proposed here, we have cloned the mouse FGF-1.B promoter (33) and showed that the sequences most crucial for binding to p37 brn (between nucleotides Ϫ483 and Ϫ473) are identical and positioned in the same context relative to the transcription start sites between the two species (33). Isolation of the cDNA expressing these DNA-binding proteins including p37 brn , either by screening the expression library using the minimal binding sequence or by protein purification, will help us verify the above model and study the regulation of 1.B expression in brain as well as in glioblastoma.