Regulation of the Human Cyclin-dependent Kinase Inhibitor p18 INK4c by the Transcription Factors E2F1 and Sp1*

The p18 INK4c cyclin-dependent kinase inhibitor is an important regulator of cell cycle progression and cellular differentiation. We and others found that overexpressed E2F proteins up-regulate p18 expression. To better understand this phenomenon, we performed a functional analysis of the human p18 promoter. Deletion studies revealed that the E2F-responsive elements of the promoter are located within 131 bp upstream of the transcription start site. This region contains putative Sp1- and E2F-binding sites. Mutational inactivation of these elements revealed that the Sp1 sites were important for the basal activity of the promoter but could also mediate the effects of E2F1 on the p18 promoter. Moreover, we found that E2F1 and Sp1 can synergistically enhance the activity of the proximal p18 promoter. Gel shift analyses using p18 promoter-derived probes led to the identification of several multiprotein complexes that were found to contain different combinations of E2F proteins and/or Sp1. Recombinant E2F1 was also capable of binding to the E2F-binding sites. Chromatin immunoprecipitation experiments demonstrated that E2F1 and E2F4 associate with the p18 promoter in unperturbed cells. Based on these findings, we conclude that E2F proteins and Sp1 play an important role in the control of p18 expression.

Progression through the mammalian cell cycle is governed by cyclins, cyclin-dependent kinases, and regulators thereof, whose expression and activity are tightly coordinated through a series of well ordered, but not completely understood, relationships. One event of prime importance during the cell cycle is the passage from the G 1 to the S phase, during which a complex intracellular signalization, involving a transient rise in the levels of G 1 phase cyclin proteins and a concurrent increase in the activity of the associated kinases, leads to the activation of the E2F transcription factor, an important regulator of cell cycle-dependent gene expression (reviewed in Refs. [1][2][3][4][5]. The E2F transcription factor family (comprising E2F1 to E2F6, DP1, and DP2) is known to modulate the expression of a number of proteins implicated in cell division, accounting for its major role in regulating the cell cycle. By associating with their DP dimerization partner, some E2F transcription factors can modulate the expression of proteins involved in DNA synthesis (thymidylate synthase (6), thymidine kinase (7), and DNA polymerase ␣ (8)), DNA repair (uracil-DNA glycosylase (9,10)), and cell cycle control (c-Myc (11), c-Myb (12), and cyclin E (13-15)), among others. Although it is known that E2F1 is a powerful promoter of mitosis (16,17), knock-out experiments in mice have demonstrated a higher incidence of tumor development in mice that lack the E2F1 gene (18). This observation suggests that E2F1-mediated signaling may be part of an antiproliferative pathway.
One of the cell cycle regulatory genes whose expression is induced by E2F is p18 INK4c , a member of the INK4 sub-family of cyclin-dependent kinase inhibitors. p18 shares sequence homology with p16 INK4a , p15 INK4b , and p19 INK4d and acts primarily on CDK4 1 and CDK6 kinases uncomplexed with their cyclin partners (19 -21). DeGregori and co-workers (22) have shown that some members of the E2F family of transcription factors were able, when overexpressed by adenoviral infection, to induce expression of the p18 gene. The fact that E2F induces p18 expression is surprising considering that the p18 protein is involved in cell cycle arrest in a variety of physiological processes such as differentiation of adipocytes (23), B-lymphocytes (24), granulocytes (25), osteoblasts (26), neuroblasts (27), and myoblasts (28,29). p18 is also induced by IL-6 in B-lymphocytes (30) and by progestins in breast cancer cells (31). On the other hand, p18 expression is down-regulated by agents that stimulate cell proliferation such as genistein in breast cancer cells (32), phorbol esters in HL-60 cells (25), and during HTLV infection of T-cells (33).
Mutations that disrupt the ability of cyclin-dependent kinase inhibitors to bind their target CDK have been discovered in a variety of cancers (34,35). Our group has shown previously (36,37) that BT-20 breast cancer cells, as well as three breast tumor biopsies of 35 samples, carry the p18-A72P mutation that decreases the ability of p18 to bind CDK6 and inhibit colony formation in transient transfection assays. Mice whose p18 alleles have been knocked out display a variety of aberrant phenotypes including lymphoproliferative disorders, organomegaly, and pituitary gland hyperplasia. Double knock-out mice for p18 and other members of the cyclin-dependent kinase inhibitor family display more varied and pronounced phenotypes (38 -40), indicating that this gene is important for the proliferative control of different cell lineages. Despite the increasing knowledge concerning the role of p18 in diverse physiological processes, very little is known regarding the precise molecular mechanisms that regulate the expression of this gene.
The human p18 INK4c gene consists of three exons, with the two last exons containing the entire p18 protein coding sequence. In most cell lines, this gene is transcribed in two predominant mRNA species of 2.1 and 1.0 kb. However, several cDNAs of intermediate length have also been cloned. By Northern blot analysis, we have determined that the longest form is the most abundant form in cancer cell lines of various origins including breast cancer (MCF7, MDA-MB-231, and BT-20), erythroleukemia (K562), and cervix carcinoma (HeLa). The 2.1-kb transcript is also the one the most influenced by E2F expression. The p18 promoter, defined in this study as the genomic sequences located upstream of the transcription start site of the long form cDNA (GenBank TM accession number AF041248 (37)), does not contain a TATA box, and its principal feature, reminiscent of other TATA-less promoters, is a region of high G ϩ C content immediately upstream of the transcription initiation region. There are no evident initiator elements (41,42) and no downstream promoter elements (DPE (43)(44)(45)), which is consistent with the presence of more than one transcription initiation site (37,42).
An understanding of the fundamental mechanisms whereby E2F regulates p18 INK4c expression would undoubtedly clarify the roles of both E2F and p18 in cell cycle regulation and might eventually lead to a better comprehension of the paradoxical effects of E2F on cell proliferation. In an attempt to achieve this objective, we have isolated p18 INK4c regulatory sequences and identified those that mediate activation by E2F. We report here that both E2F and Sp1 response elements mediate E2F-induced activation of p18 INK4c gene transcription. Furthermore, we have found that some members of the E2F transcription factor family can bind these elements in vitro and induce the transcription of the p18 gene. Moreover, we found that E2F1 and E2F4 associate with the regulatory region of the p18 gene in vivo.
MCF7 cells were grown in DMEM-F-12 supplemented with 5% FCS, 2 mM glutamine, 1 ϫ 10 Ϫ9 M E2, 100 units of penicillin/ml, and 50 g of streptomycin/ml. HeLa S3 cells were grown in 75-cm 2 flasks as a monolayer culture in high glucose DMEM supplemented with 10% v/v fetal calf serum, 100 units of penicillin/ml, and 50 g of streptomycin/ml. Suspensions of HeLa S3 cells ("spinners") were obtained by changing the medium to S-MEM supplemented with 10% v/v fetal calf serum, 2 mM glutamine, 100 units of penicillin/ml, and 50 g of streptomycin/ml. The cells were split in three 175-cm 2 flasks and cultured for 5 days after which enough detached cells were present in the flasks to inoculate a 100-ml spinner bottle. Cell density was kept between 0.2 and 1.25 ϫ 10 6 cells/ml. SL2 cells were grown as a loosely attached monolayer in Schneider Drosophila medium supplemented with 10% FCS and antibiotics at 25°C. Adenoviral Infections-The recombinant adenoviruses (AdCMV-FLAG-E2F1, AdCMV-FLAG-E2F2, and AdCMV-GFP) were obtained by insertion of the human E2F1, E2F2, or Aequoria victoria green fluorescent protein (GFP) coding sequences into the pAdCMV-FLAG vector, followed by homologous recombination in Escherichia coli (46). AdCMV-GFP has been described previously (47). The FLAG epitope (MAYKD-DDKL) was appended to the N terminus of human E2F1 and E2F2 to allow for easy detection of transgene expression by Western blot (cloning details are available upon request). Viral stocks were produced as described previously (48), and viral titers were determined by a plaque assay in 293 cells and defined as plaque-forming units/ml. Cells were then infected by adding virus stocks directly to the culture medium at an input multiplicity ranging between 100 and 300 viral particles per cell. The infected cells were harvested 18 h later, and total RNA and proteins were extracted for Northern blots and immunodetection.
Northern Blots-Northern blot analyses were done using 10 g of total RNA isolated from cells with TRI-Reagent and solubilized in Forma-Zol (Molecular Research Center Inc., Cincinnati, OH). All probes were radiolabeled using the random-priming method (49). Glyceraldehyde-3-phosphate dehydrogenase and p18 mRNA were detected using probes comprising the full-length coding sequences of the respective human cDNAs.
Whole Cell Extracts-Cells grown as monolayer cultures were trypsinized, washed twice with PBS, and pelleted in Eppendorf tubes. For suspension cultures, the cells were transferred into 500-ml centrifuge bottles, spun at 1200 ϫ g for 5 min, and washed once with PBS. The cell pellet was then transferred to a centrifuge tube and washed once more with PBS. After centrifugation and removal of the supernatant, cells were lysed according to a method described previously (50) using extraction buffer (buffer A) containing 20 mM Hepes, pH 7.9, 25% glycerol, 0.4 M KCl, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, 1 mM Na 3 VO 4 (Sigma), 50 mM NaF (Sigma), 10 mM ␤-glycerophosphate (Sigma), Complete protease inhibitor mixture, 0.7 g/ml pepstatin, and 1 mM Pefabloc (all three from Roche Molecular Biochemicals). Cells were resuspended and incubated on ice for 20 min before being stored at Ϫ80°C for at least 1 day. Prior to use, lysates were thawed, vortexed vigorously, and centrifuged. Protein concentrations were determined using the Bio-Rad DC protein assay.
Plasmid Constructs-The GST-Sp1-8xHis protein encoded by the pGEX-Sp1-8xHis plasmid contains the sequence encoding the 696 Cterminal amino acids of human Sp1 protein (provided in pCMV-Sp1 by Dr. Tjian, Berkeley, CA) followed by eight histidine residues at the C terminus, cloned in the pGEX-6P-1 plasmid (Amersham Biosciences AB). The pGEX-E2F1-8xHis and pGEX-DP1-8xHis plasmids that encode full-length human E2F1 and DP1 proteins fused to the C terminus of the GST protein and bearing an eight-histidine residue tag at the C terminus were constructed in a similar manner. The pcDNA3-E2F1-HA plasmid was constructed by cloning the entire coding sequence of the human E2F1 gene (provided in pCMV-E2F1 by K. Haelin), in the expression plasmid pcDNA3-HA, in-frame with sequences encoding a C-terminal HA tag. The pcDNA3-FLAG-DP1 plasmid was constructed by inserting the human DP1-coding sequence, preceded by DNA encoding the FLAG epitope, in the pcDNA3 expression plasmid. The expression plasmids for Sp1, E2F1, and DP1 used in D. melanogaster cell transfection experiments were constructed using the pAc5/V5-His plasmid (Invitrogen). Reporter plasmid constructs comprising different regions of the p18 promoter were generated by PCR amplification using gene-specific primers and by cloning the amplicons in the KpnI and NheI sites upstream of the luciferase gene in the pGL3-Basic reporter plasmid (Promega, Madison, WI). The E2F-BS construct containing five copies of an E2F consensus response element was obtained by annealing the oligonucleotides 5Ј-gatcatttaagtttcgcgccctttctcaa-3Ј (upper strand) and 5Ј-gatcttgagaaagggcgcgaaacttaaat-3Ј (lower strand). This double-stranded oligo was then kinased and ligated to form concatamers. Pentamers were purified on an agarose gel and cloned into the BglII site of the E1B-TATA-Luc reporter plasmid. The sequence of the insert was confirmed by sequencing both strands.
Production of Recombinant Proteins-The recombinant GST-Sp1-8xHis protein was produced according to Kadonaga et al. (51) with the following modifications. E. coli BL21 CodonPlus RIL cells (Stratagene, La Jolla, CA) were transformed with the pGEX-Sp1-His plasmid. A culture was prepared by inoculating 20 ml of LB medium (containing 100 g/ml ampicillin, 34 g/ml chloramphenicol, and 2% w/v glucose) with a single colony. The culture was grown overnight at 37°C and then diluted 500-fold in the same medium and allowed to grow until A 600 ϭ 0.5. Isopropyl-1-thio-␤-D-galactopyranoside (Amersham Biosciences) was then added to a final concentration of 0.4 mM. Three hours later, the cells were harvested by centrifugation, washed once in cold PBS, aliquoted in the proper number of tubes, and centrifuged again to obtain pellets originating from 100 ml of broth. The supernatants were thoroughly decanted, and the pellets were stored at Ϫ80°C until use. Unless stated otherwise, all purification steps were carried out at 4°C or in an ice water bath. To purify the overexpressed protein, one pellet was resuspended in 10 ml of ice-cold PBS containing 2% Triton X-100 and vortexed vigorously. This suspension was then subjected to two freeze-thaw cycles, sonicated twice (1 min each), and then slowly agitated on a rocking platform for 30 min. The lysate was centrifuged at maximum speed in a microcentrifuge. The supernatant was decanted, and the pellet containing the GST-Sp1-8xHis protein was resuspended in 10 ml of solubilization buffer (50 mM sodium phosphate buffer, pH 6.8, 6 M urea, 1 mM DTT, and 0.5% Triton X-100). The suspension was sonicated and agitated as described above and dialyzed for 4 h in order to re-nature the fusion protein. The dialysis buffer was composed of PBS supplemented with 2 M urea, 10 M ZnSO 4 , 1 mM DTT, 10% v/v glycerol, and 0.5% Triton X-100. The dialysis buffer was then replaced with a similar buffer without urea, and the dialysis was prolonged for an additional 16 h. The dialyzed lysate was centrifuged to remove insoluble matter and allowed to bind to glutathione-Sepharose beads (Amersham Biosciences) at 4°C overnight. The beads were washed three times in PreScission reaction buffer (50 mM Tris, pH 7.0, 150 mM NaCl, 1 mM DTT, 0.5% Triton X-100) and incubated in 1 ml of this buffer supplemented with 20 units of PreScission protease (Amersham Biosciences) for 24 h at 4°C. The beads were centrifuged at 10,000 ϫ g for 5 min, and the supernatant was conserved. Glycerol was added to the cleaved protein solution to a final concentration of 50% v/v, and the solution was stored in small aliquots at Ϫ80°C. At this step, the recombinant protein appeared to have a purity of at least 75% as judged by Coomassie Blue staining of the protein solution separated by SDS-PAGE (data not shown). The GST-E2F1-8xHis and GST-DP1-8xHis proteins were purified by differential solubilization. The induction and purification procedure for the GST-E2F1-8xHis and GST-DP1-8xHis proteins was the same as for the Sp1 protein, with the following modifications. After solubilization of the cell lysates with 2% Triton X-100 and sonication, the solution was centrifuged, and the pellet was resolubilized in STE buffer supplemented with DTT and Sarkosyl (150 mM NaCl, 10 mM Tris, pH 7.5, 1 mM EDTA, 5 mM DTT, 0.75% N-lauroylsarcosine sodium salt (Sigma)). The solution was vortexed, sonicated briefly, agitated at 4°C for 30 min, and centrifuged at 10,000 ϫ g for 10 min. Glycerol was added to the solubilized protein solutions to a final concentration of 50% v/v, and the solution was stored in small aliquots at Ϫ80°C. At this step, the GST-E2F1-8xHis and GST-DP1-8xHis recombinant proteins appeared to have a purity of at least 75% as judged by Coomassie Blue staining of the protein solution separated by SDS-PAGE (data not shown). Contrary to Sp1-8xHis, the E2F1-8xHis and DP1-8xHis proteins were not separated from GST by protease digestion. The presence of Sarkosyl did not alter the DNA-binding properties of the proteins, because these solutions were highly diluted in the gel shift reaction mixtures (400-fold dilution).
Electrophoretic Mobility Shift Assays-Probes encompassing different regions of the promoter were synthesized by annealing complementary synthetic oligonucleotides. The upper strand oligonucleotide of each pair bore an additional "GG" dinucleotide at its 5Ј terminus to allow radiolabeling of the probe using the Klenow fragment of DNA polymerase and [␣-32 P]dCTP. Unincorporated radioactive nucleotides were removed by gel filtration on Bio-Spin 30 columns (Bio-Rad). The binding conditions that gave satisfying results for both E2F and Sp1 binding contained 5 g of total cellular proteins, 500 ng of sheared salmon sperm DNA, 25 mM Hepes, pH 7.6, 25 mM KCl, 1 mM EGTA, 2 mM MgCl 2 , 0.05% Nonidet P-40, and 10% glycerol, in a final assay volume of 20 l. The probe was added to the binding reaction mixture, and binding was allowed to occur for 25 min at room temperature. 40,000 cpm of probe (generally representing 5-25 fmol of doublestranded oligonucleotide) were used in each EMSA reaction. For competition experiments, 1 pmol of unlabeled oligonucleotide was added prior to the addition of the labeled probe. When EMSA experiments were performed using bacterially produced or IVT proteins, 50 ng of the Sp1-8xHis protein, 50 ng of GST-DP1-8xHis protein, 5 ng of GST-E2F1-8xHis protein, or 2 l of each in vitro translated (IVT) protein were used, and 100 ng of sheared salmon sperm DNA and 40 g of bovine serum albumin (Sigma) were included in the binding mixtures. IVT reactions were performed as specified by the manufacturer (Promega) in 50-l reactions programmed with 0.5 g of T7 promoter-driven expression plasmid. An IVT reaction programmed with the empty pcDNA3 plasmid was used as a negative control. For antibody supershift experiments, 1 g of antibody was added at room temperature 10 min before addition of the probe. The antibodies used are as follows: anti-Sp1 Pep2 (Santa Cruz Biotechnology, catalog number SC-59), anti-Sp3 (Santa Cruz Biotechnology, catalog number SC-644), anti-E2F1 (NeoMarkers, catalog number MS-878), anti-E2F3 (Geneka, Montréal, Canada, catalog number 32010020, clone PG-30), anti-E2F4 (NeoMarkers, catalog number MS-1057), and anti-DP1 (Neomarkers, catalog number MS-1056). The binding reactions were loaded on 4% non-denaturing polyacrylamide gels using 1ϫ TGE (25 mM Tris, pH 8.5, 190 mM glycine, 1 mM EDTA) as the buffering system. The gels were run at 25 mA per gel at 4°C for ϳ2.5 h. After drying, the gels were autoradiographed on Hyperfilm (Amersham Biosciences) for 24 h with intensifying screens at Ϫ80°C.
Chromatin Immunoprecipitation Assays-Detection of promoterbound E2F1 and E2F4 proteins was assessed by chromatin immunoprecipitation assays, essentially as described previously (52), except that the chromatin purification step on CsCl gradient was omitted. The chromatin was sonicated in order to obtain fragments of approximately 600 bp in length. E2F1-and E2F4-containing complexes were immunoprecipitated using anti-E2F1 (Santa Cruz Biotechnology, catalog number SC-193) and anti-E2F4 (Santa Cruz Biotechnology, catalog number SC-1082) antibodies. An antibody against the FLAG epitope (Santa Cruz Biotechnology, catalog number SC-807) was used as a negative control. To detect the p18 gene in protein-DNA complexes, a 170-bp fragment of exon I was amplified by PCR using oligonucleotides 5Ј-ctctgccgagcctccttaaaact-3Ј (nucleotides ϩ1 to ϩ23 of exon I) and 5Ј-ttttcgctgaaacaattgctgct-3Ј (nucleotides ϩ170 to ϩ148 of exon I). The primers used to detect the actin gene promoter region are the same as used by others (52). The two genes were detected by 34 cycles of PCR, with an annealing temperature of 58°C for the p18 gene and 60°C for the actin gene.
Transient Transfections and Promoter Activity Assays-MCF7 cells were seeded the day before transfection in 24-well plates at a density of 1.2 ϫ 10 5 cells per well. The DNA (500 ng of reporter plasmid, 10 ng of pCMV-RL, 2 ng of pCMV-LacZ, and a total of 100 ng of expression vector for E2F1 and/or DP1 and/or empty expression plasmid) was diluted in 150 mM NaCl to a final volume of 3 l. ExGen 500 (MBI Fermentas, Burlington, Canada) was diluted 8-fold in 150 mM NaCl, and 20 l of this solution was added to the DNA. The DNA/ExGen mixtures were incubated for 20 min at room temperature. The cells were rinsed with serum-free medium, and 250 l of this medium was added to the cells. The DNA/ExGen mixture was added to the cells and incubated for 3 h at 37°C, after which the DNA/ExGen solution was removed by aspiration, and fresh serum-containing medium was added to the cells. Cells were harvested 24 h later by replacing the growth medium with 150 l of Passive Lysis Buffer (PLB, Promega), and incubating the cells in PLB on a rocking platform for 15 min at room temperature. For each transfected well, 20 l of lysate were transferred to a 96-well plate, and firefly and Renilla luciferase activities were measured on an automated computer-assisted luminometer (Berthold, Germany), using the dual-luciferase assay kit (Promega), according to the manufacturer's instructions. To correct for well-to-well variability in transfection efficiency, the firefly luciferase activity values were divided by those of the Renilla luciferase activity, which are assumed to reflect transfection efficiency. This quotient is referred to as the normalized firefly luciferase activity.
Schneider-2 (SL2) Drosophila cells were seeded at 5 ϫ 10 5 cells per well in 24-well plates. The next day, the cells were transfected using the FuGENE 6 reagent (Roche Molecular Biochemicals). The DNA (120 ng of reporter plasmid, 20 ng of pRL-null, 5 ng of pAc5/V5-His/lacZ, and a total of 80 ng of expression vector for E2F1, DP1, or Sp1) was diluted in a final volume of 2 l in TE buffer. FuGENE 6 reagent was diluted 33-fold in Opti-MEM (Invitrogen), incubated 10 min at room temperature, and 18 l of this dilution was added to the DNA solution. This mixture was further incubated for 20 min and was then added directly to the wells containing 250 l of fresh culture medium. The DNA was left in contact with the cells until harvest, 48 h later. The procedures for cell harvest and determination of reporter gene activity were as described for MCF7 cells.

E2F Up-regulates p18 mRNA and Protein Levels-Previous
reports (22,53) showed that the p18 mRNA could be induced by adenoviral overexpression of E2F1 in rat and murine embryonic fibroblasts. As a preliminary to a detailed analysis of the p18 promoter, we decided to repeat these experiments and extended them to determine whether increased p18 mRNA levels lead to increased p18 protein levels.
To assess the effect of E2F on p18 expression, adenovirus vectors expressing either the green fluorescent protein (GFP), human E2F1, or human E2F2 were introduced into MDA-MB-468 human breast adenocarcinoma cells and into a human cell line of finite life span, namely WI-38 normal lung fibroblasts. Infected cells were harvested 18 h post-infection, and p18 mRNA and protein levels were determined by Northern blot and Western blot, respectively ( Fig. 1).
In MDA-MB-468 cells, E2F1 overexpression caused a 19-fold increase in p18 mRNA levels and a 13-fold increase in p18 protein levels (Fig. 1, A and B, upper panels). In the same cell line, E2F2 overexpression caused a 16-fold increase in p18 mRNA levels and a 7-fold increase in p18 protein levels. In WI-38 cells, E2F1 caused 9-fold increases in p18 mRNA and protein levels, whereas E2F2 caused 5-fold increases in p18 mRNA and protein levels. As controls we assessed the mRNA levels of the housekeeping gene glyceraldehyde-phosphate dehydrogenase, as well as the protein levels of TRADD, a protein involved in apoptosis. Neither were altered by overexpression of E2F1 or E2F2 (Fig. 1, A and B, lower panels). These experiments confirmed that E2F1 or E2F2 overexpression leads to increased p18 mRNA levels and showed that p18 protein levels are also up-regulated.
Putative Regulatory Elements of the p18 Promoter-One important observation from the previous experiment is that the predominant p18 transcript expressed in both cell lines is the long (2.1 kb) mRNA form and that this transcript is highly induced by both E2F proteins. Subsequent studies to determine whether the increased p18 expression observed in E2F-overexpressing cells is a direct effect of E2F therefore focused on regulatory elements of the p18 gene that lie upstream of exon I. The first 150 bp of the promoter contain no TATA box ( Fig.  2A) but have a very high G ϩ C content, which is a hallmark of several TATA-less promoters (42). This proximal portion of the p18 promoter constitutes the 3Ј end of a CpG island that extends from Ϫ640 to ϩ35 relative to the transcription start site, as determined by the CpGPlot software (54). The 131 nucleotides that precede the transcription initiation site contain two elements, located at Ϫ121 and Ϫ23 relative to the start site, that loosely fit the consensus DNA sequence of E2Fbinding sites (TTTSSCGC, where S is G or C (55)). The G ϩ C-rich region also contains numerous putative Sp1-binding sites (GC boxes (56)) that are located at Ϫ121 and Ϫ98 and in a cluster at Ϫ56, Ϫ49, Ϫ42, Ϫ38, and Ϫ32. Note that the element at Ϫ121 fits the consensus sequence of both transcription factors. The preliminary genomic sequence of the murine p18 gene (Arachne assembly contig_50802) reveals that the putative regulatory elements contained in the human promoter are conserved in the murine promoter (Fig. 2B). This high degree of conservation suggests that these elements are likely to play an important role in the regulation of p18 expression.
E2F1 and Sp1 Transactivate the p18 INK4c Gene Promoter-To determine whether the putative E2F-and Sp1-binding elements of the p18 promoter can mediate the constitutive or E2F-induced activity of the promoter, we tested the function of these elements in transient expression assays.
We constructed a reporter plasmid consisting of 1600 bp (construct A) of p18 promoter sequence fused to the luciferase reporter gene and transfected it, as well as a series of 5Јdeletion mutants, in MCF7 human breast cancer cells (Fig. 3). We found that promoter construct B, which contains 131 bp of 5Ј-regulatory DNA, is as active as construct A. On this basis, we concluded that the region from Ϫ1600 to Ϫ132 is dispensable for basal p18 promoter activity. Moreover, we found that the region from Ϫ131 to Ϫ32 contains elements that are important for basal transcription because its deletion (construct P) almost completely abrogates luciferase activity. This region includes the putative Sp1 DNA-binding sites identified in Fig.  2. Mutation of the five Sp1-binding sites clustered between Ϫ56 and Ϫ28 (construct E) abolished all basal transcriptional activity. In the absence of a functional Sp1 cluster, additional mutations (constructs G and I-K) did not further reduce promoter activity. Mutations of the other putative regulatory elements of the promoter, either alone or in combination, were considerably less deleterious than mutation of the Sp1 cluster (see constructs B-D, F, and H). In agreement with these results, deletion of nucleotides Ϫ131 to Ϫ91 (construct N) did not have a pronounced effect on the basal activity of the promoter, but mutation of the Sp1 cluster (construct O) disrupted the activity of the Ϫ90 promoter.
To determine which putative regulatory elements mediate inducibility by E2F, similar experiments were conducted in MCF7 cells using expression vectors for E2F1 and DP1 (Fig. 4). DP1 was included in these experiments to avoid that the lim- ited pool of endogenous DP1 be a limiting factor for E2F activity. As noted previously, DP1 alone had no effect on luciferase activity (data not shown (57)). E2F1 had no effect on the control reporter plasmid (construct M) but caused a 38-fold increase in the activity of p18 promoter construct A, which contains 1600 bp of promoter sequence. E2F1 expression caused a 53-fold increase in the activity of promoter construct B, which contains the most proximal 131 bp of promoter sequence, demonstrating that the p18 promoter region spanning nucleotides Ϫ1600 to Ϫ132 is dispensable for induction by E2F. Mutation of the Ϫ121 Sp1/E2F or Ϫ96 Sp1 sites alone (constructs C and D) had negligible effects on p18 promoter activation by E2F1. On the other hand, mutation of the Sp1 cluster or the Ϫ23 E2F site alone (constructs E and F) or in combination with mutations of the Ϫ121 and Ϫ98 elements (constructs G and H) caused more significant (50 -80%) decreases in E2F-induced p18 promoter activity. Activation of the p18 promoter by E2F was reduced to negligible levels (i.e. less that 10% compared with construct B) only when both the Sp1 cluster and the Ϫ23 E2F site were inactivated (constructs I-K). These results show that optimal activation of the p18 promoter by E2F requires multiple regulatory elements.
E2F1 and Sp1 Cooperate to Induce Transcription from the p18 INK4c Promoter-To confirm that the proximal promoter is directly implicated in Sp1-mediated transactivation, we tested the same promoter constructs in SL2 D. melanogaster cells that are devoid of Sp1 activity (58). The promoter constructs were co-transfected with an Sp1 expression plasmid or a control plasmid (Table II; see under "Materials and Methods" for details concerning data treatment and statistical analysis). Sp1 caused a 38-fold increase in the activity of promoter construct B in SL2 cells. This effect is specific to the p18 promoter, because an artificial reporter construct (E2F-BS) consisting of five E2F DNA-binding sites cloned upstream of the E1B-TATAbox and the luciferase gene is not transactivated by Sp1 (Table  II, compare constructs B and L). Disruption of the Sp1-binding sites located at Ϫ121 (construct C) and at Ϫ96 (construct D) diminished the luciferase activity induced by Sp1 by 10 and 4%, respectively. However, it is clear that the major element conferring Sp1 responsiveness to the promoter is the cluster of Sp1-binding sites located between Ϫ56 to Ϫ28 because mutation of this cluster decreased Sp1-induced luciferase activity by 83% (compare constructs B and E). Even mutations that disrupted all but the Sp1 cluster were less deleterious (construct H). These results strongly suggest that the cluster of Sp1 sites, which is contained within Ϫ56 to Ϫ32, is responsible for basal p18 promoter activity.
In order to evaluate the combined effect of E2F1 and Sp1 on p18 promoter activity, E2F1 and DP1 were co-transfected in combination with Sp1 along with our panel of reporter con- structs in SL2 cells (Table II). E2F1 and DP1 co-expression caused a 102-fold induction in promoter activity of construct B in SL2 cells. Co-expression of E2F1, DP1, and Sp1 leads to a 2343-fold increase in promoter activity. The effect of E2F1, DP1, and Sp1 together is far more than additive and leads us to suspect a functional cooperation between the two transcription factors. This possible cooperative effect of E2F1, DP1, and Sp1 is specific to the p18 promoter because the co-expression of Sp1 with E2F1 and DP1 reduced the activity of an artificial reporter construct containing five tandemly repeated E2F-binding sites (construct L).
We also examined the effect of E2F1 and Sp1 on the activity of p18 promoter constructs containing mutations in one or three of the four response elements (constructs C-J). Although the effect of combined expression of E2F1, DP1, and Sp1 on each of these constructs was mathematically more than additive, the cooperativity between E2F1 and Sp1 relied principally on an intact cluster of Sp1 sites and the Ϫ23 E2F-binding site (Table II, compare constructs C-F).
E2F and Sp1 Interact with Elements of the p18 Promoter-To verify if the putative transcription factor-binding sites identified in the promoter sequence are able to recruit E2F or Sp1 proteins, we performed EMSA experiments using two sets of overlapping probes that span the principal putative regulatory elements contained within the 131 bp upstream of the transcription initiation site (Fig. 2C). Probes A and B correspond to nucleotides Ϫ130 to Ϫ101 and Ϫ110 to Ϫ81, respectively, whereas probes C and D correspond to nucleotides Ϫ70 to Ϫ21 and Ϫ30 to Ϫ1, respectively. The name, position on the promoter, and sequence of the oligonucleotides used in the following EMSA experiments are presented in Table I and Figs. 5 and 6.
First, in order to determine whether Sp1 can bind the putative Sp1-binding sites, we identified in the p18 promoter, probes containing these assumed DNA-binding sites (A-C) were incubated with recombinant bacterially expressed human Sp1 (Fig. 5). The results show that Sp1 can bind to probes A-C (Fig. 5, left, middle, and right panels, respectively), whereas probe D is unable to recruit Sp1 (data not shown). The specificity of the interaction between Sp1 and probes A-C was confirmed by competition experiments in which an excess of cold double-stranded oligonucleotide containing a wild-type consensus Sp1 DNA-binding site (Fig. 5, lanes 2, 7, and 14) or a mutated version thereof (lanes 3, 8, and 15) was included in the binding reaction. Similar results were obtained using the unlabeled oligonucleotide probe or mutated variants thereof as competitors. In addition to the Sp1 consensus at Ϫ98, probe B contains a sequence at Ϫ110 (GGGCGCGGGA) with limited homology to an Sp1-binding element. We evaluated the ability of Sp1 to bind to both elements by performing the competition experiments using three different mutated oligonucleotides (mutants M1, M2, and M1 ϩ 2). The M1 oligonucleotide is mutated at the imperfect site at Ϫ110, whereas the M2 oligonucleotide is mutated at the consensus Sp1-binding site at Ϫ98, and the M1 ϩ 2 oligonucleotide is mutated at both sites. The M1 oligonucleotide was as potent as the wild-type unlabeled probe in displacing the protein-DNA complex, indicating that the imperfect site at Ϫ110 is not recognized by Sp1. Thus, the mutated probes contained mutations at critical positions of the putative Sp1-binding sites, thereby confirming that the Sp1-DNA complexes that we detected are truly dependent on Sp1 DNA binding activity, that these probes contain bona fide binding sites for Sp1, and that Sp1 binding on probe B depends solely on sequences beginning at Ϫ98.
The same type of experiment was conducted using IVT human E2F1 and DP1 proteins (Fig. 6) and probes A and D to determine whether the p18 promoter contains authentic E2Fbinding sites. Although E2F1 alone is able to specifically bind DNA, we included the DP1 protein in the binding assays because DP1 enhances the binding affinity of E2F (59). As shown in Fig. 6, E2F1 and DP1 formed complexes with probes A and D. In both cases, these complexes were competed by an unlabeled double-stranded oligonucleotide containing a wild-type consensus E2F DNA-binding site, but not by an oligonucleotide containing a mutated E2F-binding site. Similarly, unlabeled wild-type probes A and D displaced E2F and DP1 from the corresponding labeled probes, whereas mutated variants of these probes did not. The E2F site in probe D contains two repeats of the sequence TCCCGC. In order to evaluate the contribution of each of these repeats, we inserted mutations in the 5Ј-most (M1), the 3Ј-most (M2), or both (M1 ϩ 2) repeats. It appears that mutation of both repeats is necessary to abolish completely the competition, suggesting that both elements participate in the recruitment of E2F to the probe.
Together these results show that the p18 promoter contains Sp1 DNA-binding sites at positions Ϫ121, Ϫ98, and in the region from Ϫ56 to Ϫ28, whereas E2F-binding sites are present at positions Ϫ121 and Ϫ23. Interestingly, mutation of nucleotides Ϫ120 to Ϫ117 on probe A abolishes the binding of both transcription factors. This is not surprising considering the sequence of the E2F1/Sp1-binding site in probe A, ggcgggaa, which loosely fits the consensus DNA-binding sequence for both transcription factors. On the other hand, the sequence recognized by E2F1-DP1 on probe D, ttcccgctcccgc, fits the E2F consensus, but it does not fit the Sp1 consensus.
E2F1 Binds the Ϫ70 to Ϫ21 Region of the p18 Promoter Independently of Sp1-Because the experiments presented in Fig. 4 and Table II showed that E2F1 can activate the proximal p18 promoter via the cluster of Sp1 DNA-binding sites located between Ϫ56 and Ϫ28, we sought to determine whether the E2F1 protein could directly bind this DNA region in vitro by performing an EMSA experiment using recombinant transcription factors. We incubated radiolabeled probe C or probe D with recombinant bacterially expressed Sp1, GST-E2F1, GST-DP1, or GST proteins in different combinations (Fig. 7). The GST and GST-DP1 proteins alone did not bind to either probe. As presented in Fig. 5, the Sp1 protein could specifically bind to probe C (Fig. 7, lanes 1-7) but not to probe D (Fig. 7, lane 24), whereas GST-E2F1 in combination with GST-DP1 could bind to both probes (Fig. 7, lanes 10 -16 and 27-29). The binding of these factors to the probes is specific, as assessed by oligonucleotide competitions using an Sp1-binding site oligo, an E2Fbinding site oligo, or the unlabeled oligonucleotide probe and mutated variants thereof. Interestingly, the binding of Sp1 to probe C was unaffected by the E2F-binding site oligo (Fig. 7,  lanes 2 and 3), and the binding of GST-E2F1 and GST-DP1 to this probe was unaffected by the Sp1-binding site oligo (Fig. 7,  lanes 13 and 14), indicating that the GST-E2F1/GST-DP1 dimer can discriminate between different GC-rich templates. The strength and specificity of binding of GST-E2F1/GST-DP1 to probe C was unaffected by the presence of Sp1 and vice versa (Fig. 7, lanes 17-23).
Taken together, the results of this last experiment support those presented in Fig. 4 and Table II showing that E2F1 can transactivate from the p18 INK4c promoter by binding to the GC-rich region located between Ϫ56 and Ϫ28.

Cellular E2F and Sp1
Interact with the Ϫ121 Promoter Element-By having determined in vitro that the p18 promoter contains E2F and Sp1 DNA response elements, we proceeded to verify if these transcription factor-binding sites were recognized by cellular proteins. To do so, we incubated probes A-D with total cellular extracts of HeLa S3 cells.
A number of protein complexes interacted specifically with probe A (Fig. 8A, compare lanes 1, 6, and 7). Competitions using an Sp1 consensus binding site prevented the formation of most of these protein complexes (Fig. 8A, lane 2), indicating that Sp1 is a major component of these complexes. Competitions using an E2F consensus binding site prevented the appearance of the lower molecular weight complexes (Fig. 8A,   FIG. 5. Binding of recombinant Sp1 to p18 promoter elements. Recombinant Sp1-8xHis protein was produced in E. coli and purified as described under "Materials and Methods." The protein was used in electrophoretic mobility shift assays on oligonucleotide probes A-C (left, middle and right panels, respectively). Unlabeled oligonucleotide competitors containing a consensus wild-type (WT) or mutated (M) Sp1 DNA-binding site were included in lanes 2, 3, 7, 8, 14, and 15 to assess binding specificity. Competitions using unlabeled probes A-C or versions mutated at critical sites are shown in lanes 4, 5, 9 -12, 16, and 17. The sequences of wild-type and mutated probes A-C are shown below (dots designate identical nucleotides). lane 4) indicating that E2F proteins are contained in a subset of probe A-associated protein complexes. The mutated forms of each consensus oligonucleotide did not alter the formation of protein-DNA complexes (Fig. 8A, lanes 3 and 5). These experiments confirmed that Sp1 and E2F, or proteins with similar DNA-binding affinities, associate with elements contained between Ϫ130 and Ϫ101 of the human p18 promoter.
To identify which of the probe A-associated complexes contained E2F, we performed antibody supershift/interference assays using antibodies raised against E2F1, E2F3, E2F4, and DP1. These assays were performed in the presence of excess Sp1 consensus oligonucleotide in order to more easily identify E2F-DNA complexes. As shown in Fig. 8B, three readily identifiable protein-DNA complexes (designated A1, A2, and A3) were detected under these binding conditions. These complexes were eliminated by addition of excess E2F consensus oligonucleotide but not by a mutated E2F-binding site (Fig. 8B, lanes  1-3), thereby confirming that they are genuine E2F complexes. Co-incubation with antibodies against DP1 caused the disappearance of the complexes and the appearance of a higher molecular weight complex (Fig. 8B, lane 9), indicating that DP1 is present in complexes A1-A3. Antibodies against E2F1 interfered with complex A1, whereas antibodies against E2F4 interfered with complex A2. Interestingly, none of the anti-E2F antibodies used interfered with complex A3. The fact that we included an Sp1 consensus oligonucleotide in the binding assays did not interfere with our appraisal of E2F-DNA complexes because we obtained the same results when the excess Sp1 consensus oligo was omitted, except that some complexes that were supershifted by anti-E2F/DP antibodies were partially concealed by complexes that contain Sp1 (data not shown).
Conversely, in order to analyze the interaction between cellular Sp1 proteins and probe A, E2F was depleted from the EMSAs using an excess of wild-type E2F consensus oligonucleotide (Fig. 8C). Four specific protein-DNA complexes (denoted A4 to A7) were resolved in this manner. The identities of the DNA-bound proteins were assessed by supershift experiments by adding antibodies specific for Sp1 or Sp3 to the binding mixtures (Fig. 8C, lanes 4 and 5, respectively). Complex A4 contains Sp1, whereas complexes A5 and A7 contain Sp3. These two Sp3-containing complexes are most probably due to two alternatively translated forms of the Sp3 protein (60). We concluded that complex A6 contains a protein different from Sp1, Sp3, or an E2F family member because it is not competed by any oligonucleotide other than the probe itself (compare Fig.  8A, lanes 2 and 4, and Fig. 8B, lane 2). Inclusion of E2F-binding oligonucleotides in the EMSA mixtures did not alter the association of Sp1/Sp3 with p18 promoter elements. We were able to observe supershifts using antibodies against Sp1/Sp3 in the absence of excess E2F oligo, but the disappearance of the original complexes was masked by the presence of E2F complexes with similar gel mobilities (data not shown).
These results confirm those obtained with recombinant proteins, namely that E2F and Sp1 transcription factors can interact with regulatory elements contained in probe A.
Cellular Sp1 Interacts with the Ϫ96 and Ϫ56 to Ϫ32 Regulatory Elements-Incubating probe B with cellular proteins yielded four major protein-DNA complexes, designated B1, B2, B3, and B4 (Fig. 9). Three of these complexes (complexes B1, B2, and B4) contain an Sp1-related protein because their formation was prevented by competition with an Sp1 consensus oligonucleotide but not by a mutant thereof (Fig. 9, lanes 1-3). To confirm that Sp1 and/or Sp3 are present in complexes B1, B2, and B4, supershift experiments were conducted using anti-Sp1 and/or anti-Sp3 antibodies (lanes 6 -8). These showed that complex B1 contains Sp1, whereas complexes B2 and B4 contain the Sp3 protein. These results confirm those obtained using recombinant Sp1. The fourth protein complex (B3) does not contain Sp1 because it was unaffected by competition with a wild-type or mutant Sp1-binding site. Moreover, antibodies against Sp1 and Sp3 did not affect B3. On the other hand, both wild-type probe B and probe B containing a mutated Sp1binding site prevented the formation of B3. The identity of this protein was not pursued further. Cellular proteins produced four major complexes (denoted C1, C2, C3, and C4) when incubated with probe C (Fig. 10). All four complexes are attributable to an Sp1-like DNA binding activity because they were displaced from the probe by an excess of wild-type Sp1 consensus oligo but not by an excess of mutated Sp1 consensus oligo (Fig. 10, lanes 1-3). Probe C contains up to five putative Sp1-binding sites. We have attempted to determine whether only a subset of these is responsible for Sp1 binding. However, under the experimental conditions described here (i.e. EMSA experiments using either recombinant Sp1 or total cellular extracts and a series of mutated probes), it seems that all five putative binding sites have the ability to mediate Sp1 binding to the probe (data not shown). Supershift experiments using anti-Sp1 and/or anti-Sp3 antibodies (lanes 6 -8) demonstrated that complex C1 contains the Sp1 protein, whereas complexes C2, C3, and C4 contain Sp3. The appearance of complexes C1 and C2 as a doublet of bands of slightly different gel mobilities is more evident on a shorter exposure of the film (data not shown).

Cellular E2F Associates with the Proximal E2F-binding
Site-The binding of total cellular proteins on probe D (Fig. 11) yielded four major specific complexes (designated D1, D2, D3, and D4). Competition experiments using an E2F consensus oligo revealed that all four complexes contain an E2F-like DNA TABLE II Transactivation of the p18 promoter by exogenous Sp1, E2F1, and DP1 in SL2 cells Drosophila melanogaster SL2 cells were transfected with reporter constructs consisting of a wild-type or mutated p18 promoter fragment cloned upstream of the firefly luciferase reporter gene, as described in the legend of Fig. 3, and different combinations of pAc5 expression vectors for Sp1, FLAG-E2F1, and DP1. The cells were harvested 24 h after transfection and lysed in Passive Lysis Buffer. Firefly and Renilla luciferase activities were measured in a luminometer, and the normalized luciferase activity was calculated. The results are presented, for each promoter construct, as the fold increase in normalized luciferase activity relative to that obtained with construct B co-transfected with the empty pAc5 expression plasmid (arbitrarily set to 1), and are the means Ϯ S.D. of three experiments conducted in triplicate. 7. Binding of E2F1, DP1, and Sp1 to the Sp1 cluster and proximal E2F site. Recombinant GST, GST-E2F1-8xHis, GST-DP1-8xHis, and Sp1-8xHis were produced in E. coli and purified as described. The proteins were incubated, alone or in various combinations, with radiolabeled probe C (lanes 1-23) or probe D (lanes 24 -29) in an EMSA experiment. Unlabeled oligonucleotides were added as indicated in order to evaluate binding specificity. WT, wild-type; M, mutated.

FIG.
binding activity because they were abolished when the wildtype oligonucleotide was added but were unaffected by the mutated E2F-binding site (Fig. 11, lanes 2 and 3). Supershift/ interference experiments using antibodies raised against dif-ferent members of the E2F family showed that complexes D1, D3, and D4 did not form when the anti-DP1 antibody was added to the binding mixture, and thus must contain the DP1 protein (Fig. 11, lane 9). Complex D3 was supershifted by the anti-E2F4 antibody and thus contains the E2F4 protein ( Fig.   FIG. 8. Binding of cellular proteins to the ؊130 to ؊101 region of the p18 promoter. HeLa S3 high salt cellular protein extracts were incubated with probe A in EMSA experiments (A). Binding specificity was assessed by adding wild-type (WT) or mutated (M) versions of Sp1 consensus oligo, E2F oligo, or probe A (lanes 2-7). Antibody supershift experiments were performed by adding 1 g of antibody against E2F/DP1 (B) or Sp1/Sp3 (C) in the binding reaction before the addition of the labeled probe. NS indicates nonspecific binding or background signal. In order to visualize protein-DNA complexes that were concealed by other more abundant complexes of similar electrophoretic mobility, the EMSA experiments in B were conducted in the presence of excess wild-type Sp1 consensus oligo whereas the experiments in C included excess wild-type E2F consensus oligo. Binding specificity was assessed by adding oligonucleotides containing the corresponding wild-type or mutated binding sites.

11, lane 8).
The E2F1 and E2F3 members are also present in some complexes because the anti-E2F1 and anti-E2F3 antibodies yielded weak but nevertheless detectable supershifts (Fig.  11, lanes 6 and 7, marked with asterisks). As expected from the competition experiment using an E2F consensus oligonucleotide, complex D2 was unaffected by the antibodies used.
Taken together, these results confirm that the p18 INK4c promoter is able to recruit proteins of the Sp1 and E2F families of transcription factors.
In Vivo Association of E2F Proteins with the p18 Gene-To confirm that E2F1 and E2F4 associate with regulatory elements of the p18 promoter in vivo as well as in vitro, we performed chromatin immunoprecipitation experiments in WI-38 cells (Fig. 12). Cells growing in log phase were treated with formaldehyde to form cross-links between E2F and associated promoter regions. Chromatin was then isolated, fragmented by sonication, and subjected to immunoprecipitation by using antibodies directed against either E2F1 or E2F4, two of the E2F family members that associate with p18 promoter probes A and D. The presence of the proximal p18 gene promoter in the immunoprecipitated chromatin was detected by amplifying the 5Ј end of exon I by PCR. A DNA fragment corresponding to the actin gene promoter was amplified as a control. As shown in lanes 6 and 7, a genomic DNA fragment containing p18 exon I co-immunoprecipitated with both E2F1 and E2F4 whereas the actin gene promoter was absent from the immunoprecipitates. Both genes were efficiently detected by PCR when the input chromatin (after fragmentation but before immunoprecipitation) was subjected to PCR amplification (lane 3). Negative controls for the immunoprecipitation consisted of performing the immunoprecipitation with an irrelevant antibody (anti-FLAG) or no antibody at all. As expected, no amplification was detected when water was substituted for chromatin in the PCR (lane 1) or in the immunoprecipitation reaction (lane 2, mock ip).
These results conclusively show that the binding of the E2F1 and E2F4 transcription factors to the p18 gene promoter occurs in normal cells. DISCUSSION Understanding the complex mechanisms that regulate cell cycle progression will contribute to a better comprehension of the control of cell proliferation and differentiation in physiological circumstances, as well as of the pathological disruptions of this growth control pathway that occur in cancer. We have examined the regulation of p18 promoter activity by the cell cycle regulator E2F in order to understand, at the molecular level, how a transcription factor renowned for stimulating cell proliferation can up-regulate a gene whose product serves to prevent passage from G 1 to S phase. Previous studies had shown that E2F up-regulates p18 mRNA levels. We have extended these observations by demonstrating that E2F-induced increases in p18 mRNA levels are accompanied by corresponding increases in p18 protein levels. Moreover, we found that the p18 promoter is a direct transcriptional target of E2F, and we have identified promoter elements that mediate the effect of E2F, and we have identified transcription factors that associate with these elements.
To examine the various aspects of the regulation of p18 promoter activity by E2F, we used different experimental models that we deemed most informative. Adenovirus infections and chromatin immunoprecipitation assays were performed in WI-38 cells because these cells represent a "normal" cell line with an intact Rb-E2F pathway. MCF7 cells were chosen for promoter deletion/mutation analyses because they have relatively low endogenous E2F activity, and they are easily transfected. Likewise, SL2 cells were chosen for experiments requiring overexpression of Sp1 because they lack endogenous Sp1. To identify transcription factors that associate with the E2Fresponsive region of the p18 promoter, we performed electrophoretic mobility shift assays using HeLa cell extracts because these have been extensively used to characterize E2F. The sum of these experiments provides a set of consistent data that paints a clearer picture of the mechanism of p18 promoter regulation by E2F. These data allow us to draw a number of conclusions.
First, the elements of the p18 promoter that are required for induction by E2F reside within the first 131 bp upstream of the transcription initiation site. Transient transfection experiments performed in MCF7 cells showed that the 131-bp promoter was as responsive to E2F as a construct containing 1600 bp of promoter DNA. Although we cannot exclude the possibility that sites upstream of Ϫ1600 could be involved in activation by E2F, we believe that the Ϫ131 promoter contains the most essential E2F-responsive elements because the magnitude of the response of the Ϫ1600 and Ϫ131 promoters to E2F in transient expression experiments (38 -53- The Ϫ131 promoter contains four regulatory elements that could potentially bind transcription factors. These include a combined E2F/Sp1-binding site at Ϫ121, an Sp1-binding site at Ϫ98, a cluster of five Sp1-binding sites between Ϫ56 and Ϫ32, and an E2F-binding site at Ϫ23. The specificity of each response element was confirmed by EMSAs by incubating the corresponding labeled DNA probes with excess amounts of oligonucleotides containing E2F or Sp1 consensus binding sites. In addition, as predicted, the Ϫ121 and Ϫ23 E2F-binding sites could bind to recombinant (synthesized in vitro) E2F1 and DP1, whereas the Ϫ98 Sp1 site and the Sp1 cluster could interact with purified Sp1 protein. The physiological relevance of these interactions was further confirmed by performing EMSA antibody interference (supershift) experiments using HeLa cell extracts. These experiments showed that different members of the E2F family (E2F1, E2F3, E2F4, and DP1) and of the Sp1 family (Sp1 and Sp3) interact with the corresponding response elements. These results were confirmed by EMSA experiments performed with protein extracts prepared from T98G and MCF7 cells (data not shown).
One particularly interesting finding of the EMSA experiments was that E2F1 and DP1 (expressed as GST fusion proteins), as well as Sp1, can interact with the Sp1 cluster of the p18 promoter. We could detect a strong interaction between bacterially produced GST-E2F1 and the Ϫ70 to Ϫ21 region of the p18 gene promoter, and this binding was unaffected by the binding of Sp1 to the probe (i.e. neither mutually exclusive nor cooperative). This observation agrees with data concerning the herpes simplex virus-thymidine kinase gene, and its induction by E2F involves a direct interaction between E2F1 and a GCrich portion of the promoter of this gene (61).
Additional experiments will be needed to determine whether other transcription factors, particularly the other members of the E2F family, interact with the Ϫ131 p18 promoter. For instance, the sequences of the Ϫ121 and Ϫ23 E2F-binding sites of the p18 promoter are highly similar to the "class I E2F consensus site," 5Ј-TTTCCCGC-3Ј, which is a preferred binding site for the E2F6 protein (62). Although the structure of the E2F6 protein suggests that it could act as a transcriptional repressor or as an inducer of E2F-regulated genes, depending on the cellular context, the function of this protein in cell cycle regulation is not yet well defined. In addition, because there exist different isoforms of some of these proteins which may be differentially regulated or expressed during the cell cycle, it will be important to determine which of these interact with the p18 promoter. For example, two human E2F3 proteins, designated E2F3a and E2F3b, have been described in HL-60 cells (63). The antibody we used to detect E2F3 in supershift experiments detects a predominant polypeptide of ϳ55 kDa (presumably E2F3a) in Western blots of HeLa extracts. Similarly, the SP3 transcript gives rise by internal translation initiation to at least three proteins of different sizes. The shorter forms are truncated in their transactivation domain, a fact that could in part explain why Sp3 can also act as a transcriptional repressor (60). Notwithstanding these limitations, which will be resolved in part when appropriate immunologic reagents are available, the results of the EMSA experiments provide a good indication of the type of proteins that interact with the E2F-responsive region of the human p18 promoter.
The functional relevance of the aforementioned regulatory elements of the p18 promoter was tested in human MCF7 breast cancer cells and D. melanogaster SL2 cells, as well as in other cell lines (data not shown), in transient expression experiments using p18 promoter constructs harboring mutations in one or more of these elements. These experiments showed that the Sp1-binding sites in the Ϫ56 to Ϫ28 cluster play a particularly important role in controlling basal p18 promoter activity. The protein-DNA interaction data suggest that the Sp1 and/or Sp3 transcription factors are responsible for the basal level of p18 gene expression. These factors belong to a large family of zinc finger transcription factors that are not, as was once believed, only implicated in the transcription of housekeeping genes. The ubiquitous Sp1 and Sp3 proteins are also important for the control of the expression of several genes in response to a variety of stimuli such as mitogenic signaling and differentiation. Moreover, they have the ability to interact with different transcriptional modulators, including E2F1 and pRb (64).
In terms of the response to E2F1, we found that either the Ϫ23 E2F site alone or the Sp1 cluster could mediate a robust (ϳ10-fold) stimulation by E2F. On the other hand, both basal promoter activity as well as the magnitude of the response to E2F were severely impaired when both the Sp1 cluster and the Ϫ23 E2F site were mutated in combination with mutations of the Ϫ121 E2F/Sp1 and/or Ϫ98 Sp1 sites. These results indicate that these elements are functionally redundant, at least in terms of their role in mediating activation by E2F1. Interestingly, we observed a strong cooperative effect between E2F1 and Sp1 on p18 promoter activity. The promoter of the dihydrofolate reductase gene, which has been well characterized in hamster, mouse, and human, also contains Sp1 and E2F response elements in the vicinity of its transcription initiation region (42). The basal transcription of this gene relies on the presence of these sites, because the position of the E2F site influences the transcription initiation position, and because the absence of functional Sp1 protein or Sp1-binding sites in the promoter abolishes transcription. Moreover, it was found that the E2F site of the dihydrofolate reductase promoter is involved in promoter silencing in serum-starved conditions, whereas serum responsiveness relies on the Sp1 sites (65). Others have suggested that E2F and Sp1 could act in a synergistic fashion on the promoter of different genes, including dihydrofolate reductase, to recruit the transcriptional apparatus and to enhance transcriptional elongation (55, 66 -68). There is also mounting evidence that the E2F-Sp1 cooperation may form, along with pRb and HDAC1, the core of a "transcriptional switch," whose function is to couple transcriptional regulation with cell cycle phase transition (69,70). These topics are beyond the scope of the work reported here but clearly deserve to be investigated in the future, because they could help to better understand the coupling of transcriptional regulation and cell cycle progression.
Previous studies that identified E2F as an inducer of p18 expression relied solely on overexpression of E2F proteins (22,53,71). We have also used adenoviral vectors and transient expression assays to assess the role of E2F on p18 expression and promoter activity. However, these models are not without flaws because overexpression of E2F may cause major perturbations in a cell and thereby elicit nonspecific responses. By using a more physiological approach, i.e. chromatin immunoprecipitation assays, we showed here that E2F1 and E2F4 can associate with the promoter of the p18 gene in unperturbed WI-38 cells. These results constitute additional and compelling evidence that E2F proteins are involved in the regulation of p18 gene expression.
Of the seven cyclin-dependent kinase inhibitor genes characterized to date, three (p18, p19, and p21) have been shown to be transcriptionally induced by E2F. The induction of p21 expression by E2F is not at odds with the properties of E2F when one considers that p21 can act as an assembly factor for cyclin-CDK complexes (72,73). On the other hand, the observation that E2F induces p18 and p19 is not so easily rationalized. Hirai et al. (20) hypothesized that the maximum levels of p19 INK4d (and of p18 INK4c ) in S phase could serve to inhibit CDK4 and CDK6 activity once the G 1 phase is over. This concept was reiterated by DeGregori et al. (22) who proposed that the E2F-p18 link could constitute a retro-feedback loop that would prevent E2F activity beyond the restriction point. Others have proposed that p18 mRNAs derived from alternate promoters of the p18 gene may not be translated with equal efficiency such that differential expression of said transcripts could influence p18 protein levels. This hypothesis is based on the work of Phelps et al. (29) who observed that a translationally inefficient long form of the p18 mRNA was down-regulated during myogenic differentiation of murine C2C12 myoblasts, whereas the short form of p18 mRNA, which is translated efficiently, was up-regulated during this process. The translational efficiency of the short and long forms of human p18 mRNA has not yet been determined, but the fact that p18 protein levels correlated to p18 mRNA levels in WI-38 cells infected with E2F-expressing adenovirus would argue against such a mechanism in human cells.
The present work provides a solid basis to investigate further the regulation of p18 expression by E2F, particularly during the different phases of the cell cycle. In fact, whereas the data presented herein clearly indicate that E2F1 stimulates p18 promoter activity and up-regulates p18 mRNA levels, the role of the other E2F proteins is less clear. For example, E2F4 associates with the p18 promoter in EMSA and chromatin immunoprecipitation experiments, but transient overexpression of E2F4 in MCF7 cells caused only minor increases in p18 promoter activity (data not shown). Because both EMSA and chromatin immunoprecipitation assays were performed with asynchronously growing cells, it is possible that E2F4, as well as other members of the E2F family, acts in complex with a pocket protein (and possibly other co-repressors) as a repressor of p18 promoter activity in early G 1 phase, as is the case for numerous cell cycle-regulated genes (52 ,75). This observation is in accordance with previously published work (22) and with the fact that E2F4, in contrast to E2F1, localizes preferentially to the cell cytoplasm in proliferating cells and associates with pocket proteins in the nucleus only in G 0 and early G 1 phases (76 -78). In the future it will be possible to determine which E2F proteins associate with the p18 promoter in various phases of the cell cycle by performing chromatin immunoprecipitation using extracts from synchronized cells. Combining this ap-proach with stable or transient expression of wild-type and mutated p18 promoter constructs could provide additional information on the respective role of each response element.