Functional interactions between Sp1 or Sp3 and the helicase-like transcription factor mediate basal expression from the human plasminogen activator inhibitor-1 gene.

Basal expression of the human plasminogen activator inhibitor-1 (PAI-1) is mediated by a promoter element named B box that binds the helicase-like transcription factor (HLTF), homologous to SNF/SWI proteins. Electrophoretic mobility shift assays performed on a set of B box point mutants demonstrated two HLTF sites flanking and partially overlapping with a GT box binding Sp1 and Sp3. Mutations affecting either the Sp1/Sp3 or the two HLTF sites inhibited by 6- and 2.5-fold, respectively, transient expression in HeLa cells of a reporter gene fused to the PAI-1 promoter. In Sp1/Sp3-devoid insect cells, co-expression of PAI-1-lacZ with Sp1 or Sp3 led to a 14-26-fold induction while HLTF had no effect. Simultaneous presence of Sp1 or Sp3 and the short HLTF form (initiating at Met-123) provided an additional 2-3-fold synergistic activation suppressed by mutations that prevented HLTF binding. Moreover, a DNA-independent interaction between HLTFMet123 and Sp1/Sp3 was demonstrated by co-immunoprecipitation from HeLa cell extracts and glutathione S-transferase pull-down experiments. The interaction domains were mapped to the carboxyl-terminal region of each protein; deletion of the last 85 amino acids of HLTFMet123 abolished the synergy with Sp1. This is the first demonstration of a functional interaction between proteins of the Sp1 and SNF/SWI families.

Basal expression of the human plasminogen activator inhibitor-1 (PAI-1) is mediated by a promoter element named B box that binds the helicase-like transcription factor (HLTF), homologous to SNF/SWI proteins. Electrophoretic mobility shift assays performed on a set of B box point mutants demonstrated two HLTF sites flanking and partially overlapping with a GT box binding Sp1 and Sp3. Mutations affecting either the Sp1/Sp3 or the two HLTF sites inhibited by 6-and 2.5-fold, respectively, transient expression in HeLa cells of a reporter gene fused to the PAI-1 promoter. In Sp1/Sp3-devoid insect cells, co-expression of PAI-1-lacZ with Sp1 or Sp3 led to a 14 -26-fold induction while HLTF had no effect. Simultaneous presence of Sp1 or Sp3 and the short HLTF form (initiating at Met-123) provided an additional 2-3-fold synergistic activation suppressed by mutations that prevented HLTF binding. Moreover, a DNA-independent interaction between HLTFMet123 and Sp1/Sp3 was demonstrated by co-immunoprecipitation from HeLa cell extracts and glutathione S-transferase pull-down experiments. The interaction domains were mapped to the carboxyl-terminal region of each protein; deletion of the last 85 amino acids of HLTFMet123 abolished the synergy with Sp1. This is the first demonstration of a functional interaction between proteins of the Sp1 and SNF/ SWI families.
Plasminogen activator inhibitor-1 (PAI-1), 1 a member of the serine protease inhibitor (serpine) family, plays a key role in the regulation of fibrinolysis by binding to and rapidly inactivating both tissue-type and urokinase-type plasminogen activators (reviewed in Ref. 1). Increased plasma levels of PAI-1 have been shown to be associated with venous thrombosis and to predispose to arterial thrombosis (reviewed in Ref. 2). Transgenic mice deficient for PAI-1 show increased endotoxin-induced venous thrombosis and enhanced neointima formation upon vessel wall injury, whereas mice overexpressing PAI-1 suffer from spontaneous venous occlusions (3,4). Components of the fibrinolytic system are also involved in extracellular matrix degradation required for invasion and metastasis of neoplastic cells. PAI-1 is thought to play two independent roles in such processes by protecting the tumor stroma from urokinase-type plasminogen activator-mediated degradation and by favoring detachment of vitronectin-bound cells (5,6). The poor prognosis of high PAI-1 levels in many cancer patients might also stem from its recently demonstrated role in tumor vascularization (7).
We have previously cloned and characterized a novel transcription factor involved in basal expression of the human PAI-1 gene in HeLa cells (17). The protein, named helicase-like transcription factor (HLTF) has a specific DNA-binding domain, a RING finger domain, and the seven conserved DNA helicase domains; it is homologous to proteins of the SNF/SWI family that play a role in chromatin remodeling and facilitate trans-factor interaction with nucleosomes (reviewed in Refs. 18 -21). Two HLTF proteins differing in translation initiation site were observed, only the smaller of which, HLTFMet123, is transcriptionally active. The same protein was independently isolated by other groups because it interacted with other DNA targets (the human immunodeficiency virus long terminal repeat and the simian virus enhancer, the myosin light chain locus enhancer, the rabbit uteroglobin promoter, a tumor necrosis factor response element), and shown to display DNA-dependent ATPase activity (22)(23)(24)(25). HLTF activates the PAI-1 promoter via specific interaction with the B box that was initially identified as a phorbol ester-responsive element (13), but was later shown only to be involved in basal expression (17). The B box is highly similar to the GT box (also called CACCC motif), which has been shown to bind Sp1 as well as other recently identified members of the expanding Sp family of transcription factors (26 -28). Sp1 is ubiquitously expressed and is essential for early embryonic development (29); it can activate transcription of a large number of regulated and constitutively expressed genes, whether the promoter comprises a TATAA box or not (30). Sp1, Sp3, and Sp4 contain a similar DNA-binding domain, with three zinc fingers and two glutamine-and serine/threonine-rich trans-activation domains (28,31). Sp3 contains an additional inhibitory domain, leading to either activation or repression depending on the promoter and cellular context (31)(32)(33)(34)(35)(36)(37).
In the present work, we investigated the interplay between Sp1/Sp3 and HLTF in basal expression from the proximal PAI-1 promoter.

EXPERIMENTAL PROCEDURES
Materials-The pGL-3 Basic vectors, luciferin, and reporter-lysis buffer were from Promega (Leiden, The Netherlands); the Galacto-Light kit was from Tropix (Bedford, MA). The Protein A-Sepharose poly(dI-dC/dI-dC), radioactive precursors, and the ECL Western blot detection kit were from Amersham Pharmacia Biotech (Merelbeke, Belgium). Culture media, fetal calf serum, media additions, and Lipo-fectAMINE transfection reagent were from Life Technologies, Inc. (Ghent, Belgium). The Qiagen plasmid extraction kits were from Westburg (Leusden, The Netherlands). Oligonucleotides were from Eurogentec (Seraing, Belgium). Monoclonal antibodies against Sp1 and Sp3 were from Santa Cruz Biotechnology (Santa Cruz, CA); the polyclonal antisera against Sp1 or Sp3 have been described elsewhere (31), as has the antiserum against murine plasminogen (38). The fusion protein between glutathione S-transferase and the HLTF DNA-binding domain (GST-HLTF-DBD), previously named GST-6D3, and mAb2F6, a monoclonal antibody directed against 6D3, have been described (17). The HeLa and Drosophila SL2 cells were provided by Dr. C. Backendorf (University of Leiden, Leiden, The Netherlands) and Dr. W. Wahli (University of Lausanne, Lausanne, Switzerland), respectively. The pPacUbxSp1 and pPacUbxSp3 expression vectors and GST-Sp3 mutants have been described elsewhere (34), the GST-Sp1 mutants were kindly provided by Dr. E. Wintersberger (University of Vienna, Vienna, Austria), and the pPacSp1N619(⌬D) and the ⌬44-LacZ vectors were provided by Dr. Tjian (University of California, Berkeley, CA) and Dr. Spear (University of Kentucky, Lexington, KY), respectively.
Electrophoretic Mobility Shift Assay (EMSA)-Nuclear extracts from HeLa cells were prepared as described (17). Double-stranded oligonucleotides (sequence of top strands given in Fig. 2A) were end-labeled with [␥-32 P]ATP and T4 polynucleotide kinase. EMSA was performed with 6 g of HeLa cell nuclear extract that were incubated with 20,000 cpm of 32 P-labeled probe in a 20-l volume containing 15 mM Hepes (pH 7.9), 100 mM (or other concentrations as indicated) NaCl, 10% glycerol, 1 mM EDTA, and 1 g of poly(dI-dC/dI-dC). After 15 min at room temperature, electrophoresis was performed at 4°C on a 4% polyacrylamide gel in 45 mM Tris borate, pH 8.5, 1 mM EDTA, which was prerun for 30 min. When needed, 1 l of antibody (corresponding to 100 ng of proteins) was added to the binding assay mixtures and incubated for 30 min on ice before addition of labeled probes.
Plasmid Constructions-The 336-bp fragment (coordinates Ϫ318 to ϩ18 bp) of the human PAI-1 promoter was obtained by PCR amplification with appropriate primers and cloning into the BglII and HindIII sites of the pGL-3-Basic plasmid yielding PAI-1-luc. For transfection experiment in Drosophila SL2 cells, the test DNA fragments were subcloned into the BamHI and HindIII sites of the ⌬44-LacZ vector (39) replacing the original alkaline phosphatase promoter to yield PAI-1-lacZ. This vector was chosen since, in contrast to PGL3-Basic or pBLCAT3, it presented no induction upon Sp1 or Sp3 transient coexpression (data not shown). Mutagenesis of the PAI-1 promoter B box was done by four-way PCR as described elsewhere (17).
The minimal (Ϫ34 to ϩ33 bp) adenovirus major late promoter was cloned as a BglII/HindIII fragment upstream from the luc gene into the pGL3 vector. Two copies of the B box and its mutants ( Fig. 2A) synthesized with KpnI and BamHI linkers were then inserted into the KpnI and BglII sites to provide the (Bbox) 2 -AdMLp-luc series.
To build Drosophila expression vectors for HLTF, HLTFMet1 and HLTFMet123 cDNAs (17) were amplified by PCR with primers introducing BglII and SalI sites on the 5Ј and 3Ј ends, respectively, and were inserted into the BamHI and XhoI sites of the pPacUbx vector (40).
The expression vector for the GST-Sp3 deletion mutant containing amino acids 495-653, i.e. the zinc finger and D domains (GST-Sp3ZD), was made by inserting into the BamHI and EcoRI sites of pGEX-1T a Sp3 cDNA fragment amplified by PCR with the primers 5Ј-GG-GAAAAAGCAACACATTTG-3Ј and 5Ј-TTACTCCATTGTCTCATT-3Ј carrying a 5Ј BamHI or EcoRI site, respectively. In order to compare easily the GST-Sp1 and -Sp3 deletion mutants from different sources, we used a nomenclature based on the domains retained, i.e. GST-Sp1, GST-Sp1ABC, GST-Sp1A, and GST-Sp1ZD for the mutants described by Ref. 41, originally numbered aa 1-788, aa 1-621, aa 1-293, and aa 622-788, respectively.
All the constructs were confirmed by nucleotide sequence determination, amplified, and purified with Qiagen plasmid extraction kits.
Cell Cultures and Transfections-HeLa cells were grown in Dulbecco's modified Eagle's medium/F-12 medium supplemented with 1 mM glutamine, 100 IU/ml penicillin, 100 g/ml streptomycin, and 10% heat-inactivated fetal calf serum. For transfection, a total of 2.5 ϫ 10 5 HeLa cells were seeded per well of Falcon six-well plates, and grown overnight before addition of 2 g of reporter plasmid, 10 ng of CMV␤ internal control, and 8 l of LipofectAMINE in 1 ml of Opti-MEM. After 5 h, the cells were washed with phosphate-buffered saline and supplemented with growth medium. Cells were harvested for luciferase and ␤-galactosidase assays 16 h after transfection (see below). Luciferase activities were normalized to ␤-galactosidase activities.
Drosophila melanogaster SL2 cells were grown in M3 medium supplemented with 10% heat-inactivated fetal calf serum and 1 mM glutamine at 28°C. One day prior to transfection, cells were plated into 3.5-cm dishes at a density of 2.5 ϫ 10 6 /dish. Cells were transfected by the calcium phosphate method as described (42). Each dish received 4 g of reporter plasmid and different amounts of pPacUbx expression vectors (see figure legends) brought to a total of 1 g with control vector. After DNA addition, the plates were left undisturbed until the time of harvest, 24 h later.
Cell extracts from transfected cells were prepared in reporter/lysis buffer. The protein concentration was determined with the Bradford assay, and equal amounts of cell extracts were used for luciferase assay with luciferin or for ␤-galactosidase activity with Galacton chemiluminescent substrate according to the manufacturer's instructions. Transfections were done in triplicate and repeated three to five times with different preparations of the same plasmid.
Protein/Protein Interactions- 35 S-Labeled HLTF proteins were obtained by in vitro transcription/translation in the TNT reticulocyte lysate of pGEM4Z constructs into which the cDNA was under control of the T7 promoter. The GST-Sp1, GST-Sp3, and mutant proteins were produced in Escherichia coli from pGEX vectors and purified by adsorption onto glutathione-Sepharose according to the manufacturer. The GST pull-down experiments were performed as described (41). Briefly, 20 l of a 50:50 slurry of the glutathione-Sepharose-bound GST fusion protein (1 g) were incubated with either 10 l of reticulocyte lysate translation mix containing 35 S-labeled HLTF protein (wild type or mutant) or 200 g of HeLa nuclear extract in 200 l of binding buffer at 4°C for 1 h. The binding buffer was 20 mM HEPES, pH 7.9, 1 mM MgCl 2 , 40 mM KCl, 0.1 mM EDTA, 0.1% Nonidet P-40, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 100 g/ml ethidium bromide. The beads were washed four times with binding buffer and then boiled in 2ϫ SDS sample buffer, followed by analysis on 8% SDS-PAGE. Coomassie Blue staining demonstrated the presence of similar amounts of GST fusion proteins. The HLTF proteins were visualized either by autoradiography ( 35 S label) or Western blotting using 10 g/ml mAb2F6 and the ECL detection system.
For co-immunoprecipitation, 200 g of HeLa nuclear extract were incubated with 10 g of polyclonal antibody directed against either 6D3 or murine plasminogen (negative control) in 500 l of NET buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.1% Nonidet P-40, 1 mM EDTA, 0.25% gelatin) as described (43) at 4°C for 2 h. 20 l of protein A-Sepharose were added, and the mixture was incubated for another 1 h. Beads were collected by centrifugation and washed two times with 1 ml of NET buffer and then once with 10 mM Tris-HCl, pH 7.5, 0.1% Nonidet P-40. Bound proteins were analyzed on 10% SDS-PAGE and revealed by Western blotting with polyclonal antibodies against either HLTF, Sp1, or Sp3 and the ECL detection system.
DNA Analysis-The DNA dideoxynucleotide sequencing reaction was performed with the AutoRead sequencing kit in presence of either fluorescence-labeled or unlabeled primers combined with fluorescencelabeled dATP. Samples were processed on the Automated Laser Fluorescent DNA Sequencer system from Amersham Pharmacia Biotech. Generated sequences were analyzed with software (44) from the University of Wisconsin Genetics Computer Group provided by the Belgian EMBL Node (Brussels, Belgium).
Statistical Analysis-Student's t test for paired values was used to evaluate luciferase and ␤-galactosidase transient expression.

HLTF, Sp1, and Sp3
Bind to the B Box of PAI-1 Promoter-In a previous study, EMSAs performed with a 32 P-labeled B box oligonucleotide (coordinates Ϫ86 to Ϫ60 bp in the human PAI-1 promoter; Ref. 13) and HeLa nuclear extracts had shown a major shifted band corresponding to several protein/DNA complexes (17). In order to identify the proteins involved in these complexes, EMSA conditions have now been modified allowing resolution of four distinct complexes (C1, C2, and the C3a/C3b doublet in Fig. 1). Protein/DNA interactions were strongly affected by the salt concentration of the binding buffer. At 150 mM NaCl, a strong C3a/C3b doublet and a weak C1 complex were observed (lane 1) while decreasing the salt concentration led to progressive decline of the doublet intensity and simultaneous appearance of a strong C1 and a weak C2 complex (lanes 2-5). Addition of antibodies directed against Sp3 to the EMSA reaction caused a supershift of complex C2 (Fig. 1B, lane 3). A partial supershift of complex C1 was caused by Sp1 antibodies (lane 4), and further addition of Sp3 antibodies suppressed most of the remaining complex (lane 2). Formation of the C3a complex was inhibited by addition of mAb2F6 directed against HLTF (lane 6) but not by an unspecific antibody (lane 5). None of the antibodies used suppressed complex C3b. In aggregate, these data indicate that HLTF (complex C3a), Sp1 (complex C1), and Sp3 (complexes C1 and C2) can bind to the B box. . The B box of the PAI-1 promoter contains a tandem repeat of sequences (5Ј-AGTGGGTG-3Ј and 5Ј-GGCTGGAA-3Ј) that present one or two mismatches (bold) with this target. These repeats flank the GT box (5Ј-GGGTG-3Ј) that partially overlaps with the upstream putative HLTF site ( Fig. 2A, upper line).
A panel of B box oligonucleotide mutants affecting G nucleotides either in the putative HLTF binding sites or in the GT box ( Fig. 2A) were synthesized. They were evaluated in a competition assay for their ability to suppress, at a 50-fold molar excess, complexes formed by the radiolabeled wild type B box in EMSA performed as above (Fig. 1B). Mutant 1 bearing mutations at bp Ϫ72, Ϫ73, Ϫ75, and Ϫ76 had previously been shown to have strongly reduced affinity for HLTF (17); consistently, it could suppress neither HLTF-nor Sp1/Sp3-containing complexes (Fig. 2B). Mutant 3 with a single mutation at bp Ϫ76 could not interact with Sp1/Sp3 either but was able to bind HLTF since it suppressed its complex (C3a). Mutants 4, 5 (data not shown), 6, 9, and 10 could still bind Sp1/Sp3 since they prevented formation of their complexes (C1 and C2) with the wild type B box but had reduced affinity for HLTF, only weakly affecting its complex. Mutant 8 affecting bp Ϫ70 and Ϫ71 bound Sp1/Sp3 less effectively since it only suppressed partially the complexes these proteins form with the wild type B box.
In aggregate, these data indicate that the sequences AGTGG and GGCTG are binding motifs for HLTF and that both have to be mutated simultaneously (Mut10) to suppress HLTF interaction; these mutations do not affect Sp1/Sp3 binding to the overlapping GT box, while a point mutation could be found that affects only Sp1/Sp3 binding (Mut3).
Transcriptional Activity of the B Box Mutants-Two copies of either the B box oligonucleotide or its mutants were fused to a luciferase reporter gene with a minimal adenovirus major late promoter (TATA-luc) and the resulting (Bbox) 2 -TATA-luc reporter vectors were introduced into HeLa cells. Transient luciferase expression was not affected by mutations in either of the two HLTF binding sites (Fig. 3A, Mut6 and Mut9); however, the combined mutations (Mut10) led to a 2-fold reduced luc activity as compared with the wild type B box. Mutations in the Sp1/Sp3 binding site (Mut1 or Mut3) reduced transient luciferase activity to the level of the control TATA-luc construct.
The mutations suppressing only either Sp1/Sp3 (Mut3) or HLTF (Mut10) binding were introduced into the 318-bp PAI-1 promoter fused to the luc gene and the constructs evaluated by transient expression in HeLa cells. The single G Ϫ76 mutation in the Sp1/Sp3 site (Fig. 3B, PAI-1 Mut3-Luc) led to a 6-fold reduced luciferase activity as compared with the wild type promoter (PAI-1-Luc) while the combined mutations affecting both HLTF sites (PAI-1 Mut10-Luc) resulted in a 2.5-fold inhibition. These data demonstrate a positive correlation between binding of HLTF and Sp1/Sp3 to these sites and transcriptional activity.

PAI-1 Promoter Activation by Sp1 and Sp3
-In order to further study the interplay between Sp1/Sp3 and HLTF on the B box, Drosophila SL2 cells, which lack endogenous Sp1-like activity, were used for transient expression experiments. The 318-bp PAI-1 promoter fused to the lacZ reporter gene was activated in a dose-dependent manner by co-expression with either Sp1 or Sp3, reaching maximal 26.0 Ϯ 0.3-and 14.3 Ϯ 0.2-fold stimulation in the presence of 500 ng of pPacUbxSp1 or pPacUbxSp3 expression vector, respectively. This effect was mediated by the GT box, since only minor stimulations (1.3 Ϯ 0.1-fold and 1.4 Ϯ 0.1-fold with 100 ng of Sp1 or Sp3 expression vector, respectively) were observed with the PAI-1-Mut3-LacZ into which the Sp1/Sp3 binding site was mutated, as compared with the activations reached with the same amount (1 g) of wild type PAI-1-LacZ (6.2 Ϯ 0.2-fold and 5.9 Ϯ 0.2-fold with 100 ng of Sp1 or Sp3 expression vector, respectively).
In conclusion, Sp1 and Sp3 can stimulate the PAI-1 promoter provided their binding site is intact in the B box.
Transcriptional Synergy between HLTF and Sp1 or Sp3-The use of alternative translation initiation starts in the HLTF mRNAs was previously shown to yield, both in vitro and in vivo, two protein variants (HLTFMet1 and HLTFMet123), of which only the shorter activated basal expression of the PAI-1 promoter in HeLa cells (17). Introduction of expression vectors for either HLTFMet1 or HLTFMet123 in SL2 cells did not affect transient expression of PAI-1-LacZ (Fig. 4, A and B, lanes  3 and 4). Similar expression of HLTFMet1 and HLTFMet123 in these experiments was demonstrated by Western blotting, using whole cell extracts from transfected or control SL2 cells (data not shown). Co-expression of HLTFMet123 enhanced by an additional 2-fold the Sp1-mediated activation of PAI-1-LacZ (Fig. 4A, lane 6); similarly, an additional 3-fold enhancement was observed upon co-expression of HLTFMet123 with Sp3 (Fig. 4B, lane 6). HLTFMet1 had no such effect (Fig. 4, A and B,  lanes 5). These synergistic activations were dependent on HLTF binding since they were not observed when the reporter vector carried the mutations simultaneously affecting the two HLTF sites (PAI-1 Mut10-LacZ; Fig. 4, A and B, lanes 12).
In conclusion, the transcriptionally active form of HLTF can synergize with either Sp1 or Sp3 to stimulate the PAI-1 promoter, and this activation requires the presence of both HLTFand Sp1/Sp3-binding sites into the B box.
Physical Interaction in Vitro between HLTF and Sp1/ Sp3-We then investigated whether direct HLTF-Sp1 or HLTF-Sp3 interactions could be involved in the observed transcriptional synergies. GST pull-down experiments were performed in the presence of ethidium bromide, which eliminates any possible interference by, or dependence on, DNA in the protein-protein interaction (62). 35 S-Labeled HLTFMet123 was synthesized by in vitro translation in a reticulocyte lysate and incubated with GST-Sp1 immobilized onto glutathione-Sepharose. The bound material was analyzed by SDS-PAGE and autoradiography, showing that approximately 10% of the input HLTFMet123 was retained by GST-Sp1 (Fig. 5B, lane 1 versus  lane 10) while nothing could be detected in the control adsorbed on GST alone (lane 9). In order to map the Sp1 domain involved in the interaction, identical amounts of deletion mutants (41) were evaluated in the same assay. Fig. 5A presents schematically the domains of Sp1 and Sp3: A, B, C, and D refer to the activation domains (40) and Z to the three zinc fingers constituting the DNA-binding domain. HLTF was only retained by the ZD mutant containing the carboxyl-terminal part of the Sp1 protein (lane 4 versus lanes 2, 3, and 5). Some Sp3 mutants were evaluated in the same way, similarly demonstrating strong binding for the ZD domains only (lane 6 versus lanes 7  and 8).
To map the HLTF domain required for interaction with Sp1 and Sp3, a series of HLTF cDNA deletion mutants was prepared (Fig. 6A) 2, 5, 6, 8, 11, and 12).  2-6). The material adsorbed on glutathione beads was analyzed by SDS-PAGE, followed by autoradiography. The molecular size markers are indicated, and HLTF is shown by an arrow. into which the same amount of each protein was incubated with GST-Sp1ZD (Fig. 6C) or GST-Sp3ZD (Fig. 6D) immobilized on glutathione-Sepharose. Only deletion of 85 amino acids from the carboxyl terminus (mutant ⌬H VI) greatly reduced the ability of HLTF to interact with Sp1 (Fig. 6C, lane 7) or Sp3 (Fig. 6D, lane 7).
In aggregate, these data demonstrate direct protein/protein interactions between HLTF and Sp1 or Sp3 that are mediated by the carboxyl-terminal region of each partner.
Physical Interaction in Vivo between HLTF and Sp1/Sp3-HLTF synthesized in vivo could also specifically interact with Sp1 and Sp3, as shown when glutathione-Sepharose-bound GST-Sp1ZD or GST-Sp3ZD were incubated with HeLa nuclear extracts and the adsorbed material analyzed by SDS-PAGE and Western blotting with the mAb2F6 monoclonal antibody against HLTF (Fig. 7A, lanes 1 and 3 versus the GST control in  lane 2). This antibody only recognizes HLTFMet1. However, both forms of HLTF (HLTFMet1 and HLTFMet123) were visualized with a polyclonal antibody against HLTF in a similar experiment (data not shown).
To determine whether HLTF and Sp1/Sp3 could form a complex in vivo, HeLa nuclear extracts were immunoprecipitated with antibodies against HLTF, and the immune complexes were analyzed by SDS-PAGE and Western blotting demonstrating the presence of Sp1 or Sp3 (Fig. 7, B and C, lanes 1) as well as HLTF (data not shown). The additional proteins observed were also present when the control antibodies against murine plasminogen were used, which did not immunoprecipi-tate Sp1 nor Sp3 (Fig. 7, A and B, lanes 2).
The Carboxyl-terminal Domain of HLTF Is Required for Transcriptional Synergy with Sp1 or Sp3-In order to investigate a transcriptional role for the HLTF domain involved in the protein/protein interactions with Sp1/Sp3, transient activation by Sp1 or Sp3 of the PAI-1-lacZ reporter vector in insect SL2 cells was challenged by co-expression of an HLTFMet123 mutant with deletion of the carboxyl terminus (⌬H VI; Fig. 8); no synergy could be observed in either case, as compared with the control conditions, where the presence of full-size HLTFMet123 yielded a 2-3-fold activation of Sp1-or Sp3-mediated enhancement.

DISCUSSION
In the present study, we demonstrate that the cis-element (B box) that mediates basal expression of the PAI-1 promoter in HeLa cells comprises a GT box interacting with Sp1 or Sp3 flanked by two HLTF-binding sites. This element mediates transcriptional activation by Sp1 and Sp3, and additional synergy with HLTF, which requires DNA binding and protein/ protein interaction mediated by the carboxyl-terminal region of either protein. Although the PAI-1 promoter was shown to harbor another Sp1-binding site (bp Ϫ45 to Ϫ40; Ref. 14 and 16) besides the one studied here (B box, bp Ϫ82 to Ϫ65), our data indicate that basal expression in HeLa cells was mostly mediated by the B box site (Fig. 3B).
Protein Binding to the B Box-Binding sites for Sp1/Sp3 and HLTF were found on a very short DNA region, raising the question of whether these large proteins could be present simultaneously on the B box. The desoxyribonuclease I footprinting technique cannot provide an answer to this question since the region protected in vitro by purified Sp1 extends from bp Ϫ83 to Ϫ66 in the PAI-1 promoter (14), thus including the two HLTF sites. Classical EMSA conditions use excess oligonucleotide and only allow observation of B box complexes involving individual proteins: Sp1, Sp3, or HLTF (Fig. 1). When a much reduced amount of B box was saturated with the recombinant HLTF DNA-binding domain (GST-HLTF-DBD) in EMSA, a supershift was observed upon addition of either Sp1 or Sp3, indicating that the two proteins could bind simultaneously to the oligonucleotide (data not shown). Although the full-size HLTF (116 kDa) is larger than GST-HLTF-DBD (48 kDa), indirect evidence that it binds in the presence of Sp1 is provided by the observation that the Sp1/or Sp3/HLTF synergy is abolished by mutation of the two HLTF binding sites (Fig. 4,  lanes 12).
Our mutagenesis analysis delineated a minimal GT box from bp Ϫ77 to Ϫ73, a location partly overlapping with the 5Ј HLTF site but at a 5-6-bp distance from the 3Ј HLTF site (Fig. 2A). This spacing suggests that HLTF bound to the 3Ј site would be on the opposite side of the DNA double helix as compared with bound Sp1, an arrangement reducing steric hindrance. More space could even probably be available since the DNA might be locally unwound (with 11.2 bp/turn instead of 10) by the Sp1 zinc fingers as shown in vitro (46).
The two HLTF-binding sites of the B box are located 10 bp apart, i.e. on the same side of the DNA double helix, probably precluding simultaneous binding of such large molecules. HLTF might have equivalent affinities for its two binding sites since no significant difference in binding competition of GST-HLTF-DBD with oligonucleotide mutants affecting either one site or the other was observed neither in EMSA (Fig. 2, mutants 6, 8, and 9) nor in a more sensitive method allowing quantitation of DNA/protein interaction at equilibrium (enzyme-linked DNA/protein interaction assay, ELDIA; Ref. 47) (data not shown). Evaluation of the different mutants by transient expression experiments in HeLa cells demonstrated no significant activity difference upon disruption of any single HLTF site as compared with the wild type B box (Fig. 3A). In aggregate, these data suggest that binding of a single HLTF molecule in the B box is sufficient to mediate the synergy with Sp1 or Sp3.
In addition, in several published EMSAs where Sp1/Sp3 binding to a GT box was investigated, additional uncharacterized faster migrating complexes were found (e.g. Refs. 48 -52), which our data suggest might result from HLTF binding. Because the HLTF proteins are found in most adult tissues (17,23,24), synergies similar to the ones described here might occur on several other promoters and might have remained unnoticed till now.
Role of Protein/Protein Interactions-Protein/protein interactions between HLTF and Sp1/Sp3 might further stabilize the complex formed by these factors and DNA. The Sp1 domain involved in the interaction with HLTF was delineated in the present study to the ZD region. Transient co-expression experiments performed with Sp1 lacking domain D demonstrated a synergy with HLTF similar to that observed with full-size Sp1, suggesting that the zinc finger domain was involved (data not shown). Further studies will require point mutagenesis of Sp1 since deletion of the zinc finger domain abrogates its DNA binding required for activation of the B box. Several cellular transcription factors such as YY1 (53), the p65 subunit of NF-B (54), and GATA-1 (55) have been shown to functionally interact with the zinc finger domain of Sp1. Similarly, the carboxyl-terminal region of Sp1 (Z and D domains) was found to mediate functional interaction with the cell cycle-regulated E2F transcription factor (41,56); a more refined mapping will be required to determine if the same domain is involved in HLTF and E2F interaction. In these examples for E2F/Sp1 interaction, the mouse thymidine kinase promoter carrying the two protein binding sites required the interaction of both factors with DNA (41) as observed here for Sp1/Sp3 and HLTF, while hamster dihydrofolate reductase promoter mutants mediated the E2F/Sp1 synergy with a single site for either factor, the missing factor being brought by protein/protein interaction (56).
The carboxyl terminus of HLTF that was shown in this study to mediate the interaction with Sp1 comprises helicase domain VI; a single amino acid change in the homologous domain of the yeast SNF2/SWI2 protein produces a dominant negative mutant unable to exert ATP-dependent nucleosome disruption (57). Although, like other SNF/SWI proteins, HLTF has no demonstrated DNA helicase activity, it was shown to be a DNA-dependent ATPase (23)(24)(25). The homologous SNF/SWI proteins require such an activity to facilitate trans-factor binding to nucleosomal DNA (58,59); they are associated with the transcription initiation complex and facilitate chromatin remodeling at the promoter (60). Likewise, B box-bound HLTF might provide local chromatin opening in the proximal PAI-1 promoter and facilitate transcriptional activation by Sp1/Sp3 that might involve the recently described Sp1 cofactor complex (61).
The present study was performed in transient expression and demonstrated a 1.5-2.5-fold increase by HLTF of Sp1/Sp3 activation; use of stably transformed cells into which the integrated reporter constructs present a more structured chromatin would probably demonstrate a stronger synergy.