A novel p53 transcriptional repressor element (p53TRE) and the asymmetrical contribution of two p53 binding sites modulate the response of the placental transforming growth factor-beta promoter to p53.

Previous studies in our laboratory and others identified placental transforming growth factor-beta (PTGF-beta) as an important downstream mediator of DNA damage signaling and a transcriptional target of p53. Here we show that accumulation of PTGF-beta mRNA in response to p53 overexpression is delayed relative to p21(WAF1), whereas the promoters of these genes respond to p53 with similar kinetics. Mutational analyses of two p53 binding sites within the PTGF-beta promoter revealed that site p53-1 (+29 bp) is responsible for as much as 80% of the transcriptional response to p53. This is consistent with electrophoretic mobility shift assays showing that site p53-1 binds p53 with a much higher affinity than site p53-2 (-850 bp). We also describe for the first time a novel 21-bp element (-222 to -242 bp) that acts to down-regulate the PTGF-beta promoter response to p53. Termed the p53 transcriptional repressor element (p53TRE), this sequence was shown to suppress p53 transactivation in a position- and promoter-independent fashion and to associate with a 28-kDa protein expressed in several tumor cell lines. A p53 suppressor element and asymmetric p53 binding sites may participate determining the activation thresholds of p53-responsive promoters in a cell- and context-specific manner.

The p53 tumor suppressor is often referred to as the "guardian of the genome" because of its central role in regulating the cellular response to DNA damage (1). The rapid and substantial increase in p53 protein levels triggered by a genotoxic insult can be attributed primarily to stabilization of the normally labile protein (2). The ability of p53 to suppress transformation is largely dependent on its ability to act as a sequence-specific transcription factor (3). p53 in the nucleus associates with cis-acting DNA sequences composed of two copies of the palindromic motif 5Ј-PuPuPuC(A/T)(A/T)GPy-PyPy-3Ј, separated by a variable linker region of between 0 and 13 bases (4). Studies in lower eukaryotes have led to the estimate that the human genome contains more than 200 genes that are directly responsive to p53 (5). Promoter mapping and microarray studies have already implicated p53 in the regulation of more than 100 genes involved in cellular processes such as DNA repair, angiogenesis, signal transduction, and oxidative stress (6). It is clear that p53 is at the helm of a diverse collection of downstream effectors.
The most intensively studied outcomes of p53 activation are cytostasis and apoptosis (7). A reversible cell cycle arrest response has been largely attributed to p53-dependent up-regulation of the cyclin-dependent kinase inhibitor p21 WAF1 (8,9). p53 has also been implicated in the induction of a premature senescence program in some cell types (10). The mechanisms underlying p53-mediated apoptosis are not as clearly defined, although a multitude of p53 target genes have been implicated in this process (11). However, no single gene appears to be crucial in p53-mediated apoptosis, and it is likely that particular subsets of downstream effectors are activated in a cell-and context-specific fashion, with each contributing in varying degrees to the overall apoptotic response (11). The propensity of p53 to initiate cell cycle arrest and apoptosis varies among different cell types (12,13). Tumor-associated mutations in the DNA binding and hinge domains of p53 have been shown to have different effects on the ability of mutant p53 to activate p53-responsive promoters and can even lead to the specific loss of the apoptotic function of p53 but not its cell cycle arrest activity (14 -17). Oda et al. (18) demonstrated that transactivation of the apoptosis-inducing gene, p53AIP, by wild-type p53 (phosphorylated at Ser-46) occurs prior to the onset of apoptosis but not cell cycle arrest. Together, these observations allude to transcriptional mechanisms that permit the differential regulation of genes downstream of p53 involved in arrest and apoptosis.
Much attention has focused on post-translational modifications within the amino-and carboxyl-terminal domains of p53 and their impact on the ability of p53 to transactivate target genes (19). Phosphorylation events within the amino-terminal activation domain have been shown to regulate p53 binding to associated proteins such as Mdm2 and components of the transcription initiation complex (20). More recently, up-regulation of p53 target genes in vivo has been shown to be contingent on acetylation of the carboxyl-terminal domain of p53 (21)(22)(23). Espinosa and Emerson (22) showed that acetylation of p53 can significantly enhance associations with transcriptional coactivators p300 and TRAAP necessary for acetylation of nucleosomes in the promoters of p53 target genes. Consistent with these findings, binding and deacetylation of p53 by the NADdependent deacetylase SIR2␣ were shown to repress p53-mediated transcription and the cellular response to DNA damage (24,25). Several other p53-interacting proteins have been shown to enhance (e.g. BRCA1 (26), BML (27)) or attenuate (e.g. ATF3 (28), S100B (29)) p53-mediated transcription. Among these, Samuels-Lev et al. (30) demonstrated that ASPP proteins could specifically enhance the transcription of p53 target genes involved in apoptosis, whereas Stros et al. (31) found that high mobility group (B1/B2) proteins down-regulate the apoptosis-inducing Bax gene in a cell-specific manner. Because none of these factors have been shown to associate directly with DNA, it is likely that they exert their effects by enhancing or interfering with the ability of p53 to bind DNA or to interact with other important trans-acting factors (e.g. Sp1 (32)) associated with p53-responsive genes. Although it is clear that promoter-specific sequence elements play an essential role in the response of individual genes to p53 activation, the nature of these regulatory elements and the mechanisms used to modulate the response of individual gene promoters to p53 are poorly understood.
Previous studies in our laboratory identified the Placental Transforming Growth Factor-␤ (PTGF-␤) 1 gene as responding to both p53-dependent and p53-independent DNA damagesignaling events (33). Overexpression of PTGF-␤ alone was sufficient to suppress growth and induce apoptosis of MDA-MB-468, MCF-7, and various other breast cancer cell lines, but not untransformed cell lines. We demonstrated that a 1,015base pair region encompassing the 5Ј-end of the PTGF-␤ gene contains a functional p53 binding site (p53-1) within the 5Јuntranslated region. Coincident with our study, Tan et al. (34) also identified the PTGF-␤ gene as p53-responsive and described a second p53 binding site (p53-2) almost 900 bp upstream of site p53-1. Here we compare the kinetics of PTGF-␤ and p21 WAF1 promoter activation by p53, examine more closely the correlation between PTGF-␤ promoter activity and p53 binding to sites p53-1 and p53-2, and search for cis-acting sequence elements that participate in the PTGF-␤ promoter response to p53.

EXPERIMENTAL PROCEDURES
Cell Culture-Conditions for growth of MDA-MB-468 and MCF-7 human breast cancer cells and HeLa human cervical carcinoma cells have been described previously (35). Procedures for the growth and maintenance of adenoviral vectors have also been described previously (33).
Reporter Constructs and Luciferase Assays-The p21-lux luciferase reporter construct was graciously provided by Dr. S. Benchimol (Department of Medical Biophysics, University of Toronto). p21-lux contains ϳ2.4 kb of the p21 WAF1 proximal promoter in the pGL3-Basic luciferase cloning vector (Promega). A 1,015-bp fragment containing 967 bp upstream of the transcriptional start site and 49 bp within the 5Ј-untranslated region of the human PTGF-␤ gene was PCR amplified from human fibroblast genomic DNA using the forward primer SP-WTF1 (5Ј-GTA CCG AGC TCT AGA ACT CTT GAC GT-3Ј) and reverse primer SPWTR1 (5Ј-AGA TCT CGA GTT CTT GCC CGG GCA T-3Ј). This fragment was cloned upstream of a luciferase gene in the pGL2-Basic reporter vector (Promega) and is denoted as pWT.
Variants of pWT having mutations in p53 binding sites were generated using the following primers (see Fig. 2B). The pMT1 forward primer was SPWTF1, and the reverse primer was SPMT53R (5Ј-GGA GTT CGA GAT CGA TCT GGG TCG AAT GGC AAT ACC-3Ј).
The primers for pMT2 were generated in a two-step PCR using fragments generated from forward primer SPWTF1 with SPINMTRB (5Ј-TCG AGA TCG ATC TGG GTC GAA TGG CA-3Ј; fragment A) and forward primer SPINMTFA (5Ј-TGC CAT TCG ACC CAG ATC GAT CTC GA-3Ј) with reverse primer SPMT53R (fragment B) in a final PCR including fragments A and B with forward primer SPWTF1 and reverse primer SPWTR1.
pDLPTR (see Fig. 6) was generated using a four-step cloning procedure as follows. The forward primers SPWTF1 and DLPTRR (5Ј-GCT GTC GCG GAC ATT GTT ACT ATG TG-3Ј) were used in a PCR to generate fragment C, and forward primers DLPTRF (5Ј-GCT GAT ATC ATA TCG AGG AAG AGG A-3Ј) and SPWTR1 were used to generate fragment D. Both fragments were cloned separately into pCR2.1-TOPO cloning vector (Invitrogen). Fragment D was cloned into the EcoRV/ HindIII restriction sites downstream of fragment C, and the resulting SacI/XhoI fragment containing C and D was then cloned into pGL3-Basic.
pDCR was generated in a fashion similar to that for pDLPTR; however, primer PCRDCF (5Ј-GCC GAT ATC CTC ATA TCG AGG AAG AG-3Ј) was used instead of DLPTRF, and DLPTRR was replaced by PCRDCR (5Ј-GGA GTC TTT TTG GAG G-3Ј).
pWT-TRE2 was generated by cloning a 171-bp PCR fragment corresponding to the region between Ϫ81 and Ϫ251 bp using forward primer OLI716F (5Ј-CTC GAG CTC CAA AAA GAC TCC CAG-3Ј) and reverse primer OLI716R (5Ј-GGT ACC TGC CTG CAG AGC AAA CAC-3Ј) into the KpnI site in pWT. p21-TRE2 was generated in a similar fashion. All reporter constructs were sequenced using vector-and plasmid-specific primers on a LI-COR model 400 automated sequencer (ACGT Corp., Toronto, Ontario). Luciferase reporter constructs were cotransfected into MDA-MB-468 cells with a constitutively expressed ␤-galactosidase reporter plasmid (CMV-␤-galactosidase) using a calcium phosphate precipitation method (37). Luciferase was quantitated using the Dual-Light kit (PE Biosystems/Tropix) using a Berthold Lumat LB9507 luminometer and normalized to ␤-galactosidase activity measured in the same fashion.
Electrophoretic Mobility Gel Shift Assay (EMSA) and UV-Cross-linking Assay-Complementary, single-stranded DNA oligonucleotides were combined at 2.5 M and end-labeled with [␥-32 P]ATP. Excess nucleotides were removed by gel filtration using Sephadex-G25 fine spin columns (Roche). Oligonucleotides were resuspended in water at ϳ350 nM, boiled, and left to cool to room temperature overnight. Oligonucleotides (forward sequence only is indicated): p53-1, forward primer P53BS1 (5Ј-CAC CAG CCA TGC CCG GGC AAG AAC TCA-3Ј); MTp53-1, forward primer P53MTFGS (5Ј-CAC AGC TCG ACC CGG GTC GAA ACT CA-3Ј); p53-2, forward primer SPINWTFA (5Ј-TGC CAT CTT GCC CAG ACT TGT CTC GA-3Ј); MTp53-2, forward primer SPIN-MTFA (described above). Sequences included in oligonucleotides R A , R B , R C , R D , R E , R F , and R are depicted in Fig. 6A. Oligonucleotide R MT corresponds to the sequence in forward primer GS724.751F (5Ј-AGA CTC CAA GGG CGA ATT CTG CAG ATA TCC TCA-3Ј). Conditions for DNA binding reactions and EMSAs have been described previously (38). For UV cross-linking assays, incubation reactions containing oligonucleotides and nuclear protein extracts were exposed to 8,000 J of UV radiation in a Stratalinker (Stratagene) and fractionated on a 4 -20% Tris-glycine SDS-polyacrylamide gel for 2 h at 100 V. All gels were dried and exposed to x-ray film overnight.

RESULTS
The PTGF-␤ Promoter Exhibits an Immediate-Early Response to p53 Activation-Microarray studies in colon and lung carcinoma cell lines overexpressing wild-type p53 (6,39) have suggested that PTGF-␤ is induced with kinetics that are characteristic of an immediate-early p53 target gene. Northern blot analyses were employed to compare more definitively the steady-state levels of PTGF-␤ and p21 WAF1 transcripts in MDA-MB-468 cells at various times after infection with a recombinant adenovirus expressing wild-type p53 (Adp53) at 100 pfu/cell. As shown in Fig. 1A, p21 WAF1 transcripts were detected by 12 h and peaked by 24 h postinfection with Adp53. PTGF-␤ mRNA was not detectable at 12 h but also peaked by 24 h postinfection.
To investigate whether this induction profile could be recapitulated at the transcriptional level, the PTGF-␤ and p21 WAF1 promoters were cloned into luciferase reporter vectors to generate pWT and p21-lux, respectively, and their response to p53 was examined. The PTGF-␤ promoter in pWT contains two functional p53 binding sites: site p53-1 at ϩ29 bp within the 5Ј-untranslated region and site p53-2 at Ϫ850 bp (33); p21-lux contains 2.4 kb of the p21 WAF1 promoter (40). MDA-MB-468 cells were transiently transfected with either the p21-lux or pWT reporter, and luciferase assays were performed at 6, 12, 18, and 24 h postinfection with Adp53. As shown in Fig. 1B, induction levels of the two promoters were virtually identical at 12 h. By 18 h the p21 WAF1 promoter displayed near maximal 15-fold induction levels and remained at this level at the 24 h time point. Induction of the PTGF-␤ promoter proceeded at a slower rate, reaching 6-fold by 18 h and 13-fold by 24 h postinfection. No induction was observed in cells infected with a control adenovirus expressing the ␤-galactosidase gene (data not shown). Thus, although p53-mediated induction of the PTGF-␤ promoter is somewhat delayed relative to the p21 WAF1 promoter at 18 h, induction levels are comparable by 24 h, consistent with the levels of endogenous mRNA transcripts observed at this time point. The absence of a significant difference in promoter induction levels at 12 h suggests that the delay in endogenous PTGF-␤ transcript accumulation at this time point is not caused by differences in the transcriptional response of these promoters to p53.
Binding Sites p53-1 and p53-2 Make Distinct Contributions to the p53 Response of the PTGF-␤ Promoter-To examine the contributions of p53 binding sites p53-1 and p53-2 to PTGF-␤ promoter responsiveness, we examined the activity of promoter constructs bearing mutations in each site. As depicted in Fig.  2A, mutant p53 binding sites were created by nucleotide substitutions (lowercase) within the core (shaded) of each 10-bp consensus repeat motif. Mutations of this nature have been shown to be sufficient to abolish p53 binding (4). Luciferase assays were performed 24 h postinfection of MDA-MB-468 cells with Adp53. As shown in Fig. 2B, mutation of site p53-1 resulted in a large decline in p53 induction: from 24-fold in pWT to 4-fold in pMT1. Mutation of p53-2 had a less dramatic effect, reducing induction from 24-fold in pWT to 16-fold (pMT2) after Adp53 infection. Mutation of both sites (pMT1/2) completely abolished p53 responsiveness. This was confirmed in an analysis of the p53 response of constructs with a mutated p53-1 and progressively longer deletions from the 5Ј-end of the PTGF-␤ promoter (Fig. 2C). These results are consistent with the notion that site p53-1 makes a larger contribution to the p53 response than p53-2, and that the combined contribution of these sites to PTGF-␤ promoter induction is additive.

p53-mediated Transactivation of the PTGF-␤ Promoter
for p53 binding to a consensus probe (lanes 18 -20) (38). These results suggest that the affinity of site p53-1 for p53 is much higher than site p53-2 under our EMSA binding conditions. Previous work by Tan et al. (34) suggested that p53 binding to each of these sites was dependent on the presence of monoclonal antibody Ab421 in EMSA binding reactions. As shown in Fig. 3B, inclusion of Ab421 in EMSA binding reactions produced a p53-specific band with the p53-2 probe (lane 9) which could be effectively competed by a 10-and 100-fold excess of p53-1 (lanes 10 and 11) but not by a mutant form of p53-2 ( lanes  12 and 13). This EMSA complex was also supershifted by an antibody specific for the amino terminus of p53 (lane 14). These results confirmed that site p53-2 is also a specific target for wild-type p53 binding.
Localization of a Cis-acting Sequence Element That Downregulates the PTGF-␤ Promoter Response to p53-To investigate whether other regions within the PTGF-␤ promoter are involved in p53 induction, a series of 5Ј-end deletion constructs were generated and tested for responses to infection of MDA-MB-468 cells with a recombinant adenovirus expressing wildtype p53. Interestingly, removal of 284 bp between Ϫ389 and Ϫ105 bp resulted in a 3-fold increase in p53 induction (compare pWT⌬389 with pWT⌬105), pointing to the existence of a negative regulatory element within this region (Fig. 4A). Functional analysis of constructs containing additional deletions through this region (Fig. 4B) localized this negative regulatory element, designated the "p53 Transcriptional Repressor Element" (p53TRE), to a 35-bp region between Ϫ251 and Ϫ216 bp within the PTGF-␤ promoter.
EMSA Complex Formation Mediated by the p53TRE-To determine whether one or more trans-acting factors specifically associate with the p53TRE, a 40-bp double-stranded DNA oligonucleotide (denoted R in Fig. 5) containing the putative repressor region between Ϫ251 and Ϫ212 bp was radiolabeled and used in EMSA binding reactions containing nuclear extracts from MDA-MB-468 cells. As shown in Fig. 5, two bandshifts of equal intensity, denoted R1 and R2 for the upper and lower bands, respectively, were observed (lane 2). Both bands could be competed with a 10 -100-fold molar excess of unlabeled oligonucleotide R (lanes 3-5), although oligonucleotide concentrations needed to compete R1 binding were typically 10-fold greater than for R2. Furthermore, a 10 -100-fold molar excess of an unrelated competitor (MTp53-2) was seen to compete R1 but not R2 effectively (lanes 6 -8). These results suggest that EMSA complex R1 is nonspecific and that complex R2 formation involves high affinity trans-acting factor binding to specific sequence elements within the 35-bp p53TRE. Identical EMSA binding patterns were observed using extracts prepared 24 h postinfection of MDA-MB-468 cells with Adp53 (lanes 9 -15), indicating that p53 has no direct effect on the binding or expression of factors associated with the p53TRE.
To delimit sequence elements involved in band R2 formation further, a series of EMSA competition experiments were performed using oligonucleotides containing different portions of the p53TRE. The results, summarized in Fig. 6A, demonstrate that of the six overlapping oligonucleotides tested, only oligonucleotide R A , containing sequences from Ϫ228 to Ϫ238 bp, did not show some level of competition for band R2 formation (Fig.  6B, lanes 11-14). Oligonucleotides R B and R D , which share a 21-bp overlap region between Ϫ222 and Ϫ242 bp, were the most effective competitors (Fig. 6B, lanes 15-18 and 23-26), followed by R C (lanes 19 -22), R E (lanes [27][28][29][30], and R F (lanes [31][32][33][34]. Oligonucleotides R A and R B define a 6-bp overlap region between Ϫ222 and Ϫ228 bp which appears to be necessary for R2 complex formation. Oligonucleotides R C and R E demonstrate that sequences at the 5Ј-end (Ϫ232 to Ϫ242 bp) of the 21-bp region defined by oligonucleotide R B also contribute to R2 complex formation. Reciprocal experiments using oligonucleotide R B as the probe (Fig. 6B, lanes 35-44) resulted in the formation of two complexes that appear to be identical to those formed with oligonucleotide R. Complex R2 was competed efficiently by unlabeled oligonucleotide R D (lanes 37-40) but not by oligonucleotide R MT , in which sequences within oligonucleotide R D are scrambled (lanes 41-44). Together, these results define a novel 21-bp p53TRE (Ϫ222 to Ϫ242 bp) that mediates the formation of a specific DNA-protein complex involved in modulating transactivation of the PTGF-␤ gene by p53.
The p53TRE Functions in a Position-independent Manner-To investigate the effect that disruption of the p53TRE has in the context of the full-length PTGF-␤ promoter, promoter constructs with mutations spanning the p53TRE (pDLPTR) or within the p53TRE (pDCR) were generated (Fig.  7). Mutant p53TRE sequences in pDCR are identical to oligonucleotide R MT , which was unable to compete efficiently for EMSA complex R2 (Fig. 6). Promoter responses were compared with that of the intact promoter (pWT) in MDA-MB-468 cells 24 h after treatment with Adp53. In each case, mutations involving the p53TRE region resulted in a significant increase in promoter induction by p53 relative to the wild-type promoter (Fig. 7).  10 -12) or mutant site p53-2 (MTp53-2; lanes 6 -8, 13-15). Relative positions of EMSA complexes R1 (nonspecific) and R2 (specific) are indicated.

p53-mediated Transactivation of the PTGF-␤ Promoter
To determine whether the p53TRE can function in a position-independent manner, a 342-bp fragment containing two copies of the 171-bp region between Ϫ81 and Ϫ251 bp was cloned upstream of the 966-bp PTGF-␤ promoter. As shown in Fig. 7, the addition of p53TRE elements upstream of site p53-2 (pWT-TRE2) reduced p53 responsiveness to 68% of the parental pWT construct. Similarly, the addition of p53TREs to the 5Ј-end of the p21 WAF1 promoter (p21-TRE2) reduced p53 activation to 58% of the parental p21 WAF1 promoter construct. These observations suggest that p53TRE sequences can modulate p53-dependent transcriptional activation in a positionand promoter-independent manner.
The analysis of an 8.8-kb region upstream of the PTGF-␤ gene transcriptional start site (GenBank accession no. AF305420) revealed the existence of candidate p53TRE sequences at Ϫ8,637, Ϫ8,194, Ϫ7,436, Ϫ5,253, and ϩ2,320 bp. Interestingly, such sequences were situated adjacent to putative p53 binding sites identified at Ϫ8,594, Ϫ8,318, Ϫ7,562, Ϫ4,888, and Ϫ3,957, suggesting a possible functional association. Analysis of other p53-responsive genes for the presence of sequences homologous to the p53TRE identified p53TRE elements within the 14-3-3, PIG-3, and proliferating cell nuclear antigen genes, and single sites within the p21 WAF1 and GML promoters ( Table I). Alignment of these 10 candidate elements with the p53TRE at Ϫ242 bp of the PTGF-␤ promoter identified a putative consensus binding region containing a highly con-served 11-bp core sequence (CCCAGCCTGGA). No matches to the 21-bp p53TRE were obtained in a search of the vertebrate matrix group of the TRANSFAC 5.0 database (transfac.gbf.de/ TRANSFAC/) using MatInspector Professional software (genomatix.gsf.de).
The p53TRE Associates with a 28-kDa Trans-acting Factor Expressed in Several Human Tumor Cell Lines-To examine whether trans-acting factors specific for the p53TRE are expressed in other cell types, nuclear extracts prepared from MCF-7 and HeLa cells were used in EMSA binding reactions containing the oligonucleotide R probe. As shown in Fig. 8A, complex R2 formation was observed in all EMSA binding reactions (lanes 2-7), suggesting that transcription factors specific for the p53TRE are widely expressed. UV cross-linking the EMSA binding reactions containing oligonucleotide R and different concentrations of nuclear extracts from MDA-MB-468 cells, followed by fractionation of UV-cross-linked binding reactions on SDS-polyacrylamide gels, identified three distinct protein-DNA complexes estimated to be 180, 90, and 28 kDa in size (Fig. 8B, lanes 2-4). Because bands at 180 and 90 kDa were also observed in parallel experiments containing oligonucleotide p53-1 (lane 6), these were considered to be nonspecific. Only the 28 kDa band was specific for binding reactions mediated by the p53TRE oligonucleotide, implicating this p53TREbinding protein (p53TRE-BP) in the attenuation of PTGF-␤ promoter transactivation by p53.

DISCUSSION
p53 is a sequence-specific transcription factor that regulates the expression of multiple genes; many are involved in the induction of cell cycle arrest and apoptosis in response to DNA damage. p53 functions in a context-and cell-specific fashion (42). Such functional flexibility alludes to the existence of intricate mechanisms that regulate the p53-mediated transcriptional response of individual target genes. Previous studies in our laboratory (33) and others (34) have implicated PTGF-␤ as an important downstream target of p53-dependent DNA damage signaling. Here we demonstrate that the PTGF-␤ promoter responds to p53 with kinetics comparable with the p21 WAF1 promoter and that this response is mediated by two p53 binding sites: a high affinity site (p53-1) at ϩ29 bp that is responsible for as much as 80% of the PTGF-␤ promoter response, and a low affinity site (p53-2) at Ϫ850 bp. Interestingly, site p53-1 was shown to bind p53 in the absence of Ab421, whereas latent p53 binding was not observed with site p53-2. Deletion studies also identified a novel cis-acting sequence element between Ϫ222 and Ϫ242 bp which further modulates the PTGF-␤ response to p53. Termed the p53-transcriptional repressor element, these sequences were shown to suppress p53-mediated transactivation in a position-and promoter-independent manner and to associate specifically with a 28-kDa trans-acting factor expressed in several human tumor cell lines. These results evoke a model for PTGF-␤ gene regulation in which the high affinity site (p53-1) provides for a rapid response to small changes in cellular p53 levels, the low affinity site (p53-2) allows for further induction in the face of persistent, high levels of p53, and the p53TRE provides an independent means of modulating the PTGF-␤ promoter response to p53 in a cell-and context-specific manner.
PTGF-␤ promoter induction by p53 proceeds more slowly than the p21 WAF1 promoter between 12 and 18 h postinfection with Adp53, but no significant differences were observed at 12 h, and both promoters reached similar levels of induction (13-15-fold) by 24 h. This is in contrast to our Northern blot results showing that PTGF-␤ transcript accumulation is delayed by as much as 12 h relative to p21 WAF1 . Recently, Baek et al. (43) reported that the dietary phenolic compound, resveratrol, and the DNA-damaging agent, etoposide, could induce the accumulation of PTGF-␤ mRNA and protein in HCT-116 cells in a p53-dependent manner. Consistent with our observations, p53 was seen to accumulate as early as 3 h post-treatment, whereas increased levels of PTGF-␤ mRNA and protein remained undetectable until at least 20 h later. Based on our comparison of promoter activities, the apparent delay in PTGF-␤ mRNA accumulation relative to p21 WAF1 at 12 h does not appear to involve differences in the transcriptional response of these genes to p53. Rather, this delay is likely a reflection of significantly lower levels of basal PTGF-␤ gene expression in these cells, with the result that PTGF-␤ gene transcripts require more time to accumulate to detectable levels after p53 induction. PTGF-␤ levels may cross the threshold of detection only in the face of sustained high level expression of p53. In support of this view we have observed that the PTGF-␤ promoter is much less active than the p21 WAF1 promoter in the absence of p53. It is also unlikely that p53mediated changes in factor binding at the p53TRE contribute to this delay because p53 overexpression was seen to have no effect on p53TRE binding at the 24 h time point corresponding to the peak of PTGF-␤ promoter induction. Rather, levels of factor binding to the p53TRE may contribute to the maintenance of low basal levels of PTGF-␤ gene expression and may have an important influence on the rate and maximal level of transcriptional induction of this gene by p53.
Of the two known p53 binding sites within the PTGF-␤ gene, site p53-1 has a much higher affinity for p53 but displays lower homology to the p53 consensus (17/20-bp match) than site p53-2 (18/20-bp match). In fact, our analysis of the sequence of p53 binding sites from other p53 target genes such as p21 WAF1 (40), 14-3-3 (44), and cyclin G 1 (45) indicated that the degree of adherence to the consensus sequence did not accurately predict the contribution of each binding site to p53 responsiveness. Clearly, individual nucleotides, or the arrangement of specific subsets of nucleotides, must have an important influence on the affinity of these binding sites for p53. In support of this idea, Kim et al. (46) found that p53 binding sites having internal symmetry shift more readily toward a cruciform or stem-loop structure that acts as a more favorable binding substrate for p53. The existence of six potential intrastrand bp  7. Functional analysis of PTGF-␤ promoter constructs containing mutations or insertions of the p53TRE. MDA-MB-468 cells were transfected with luciferase reporter constructs containing the PTGF-␤ promoter (pWT; ϩ49 to Ϫ966 bp) mutated for p53TRE sequences between Ϫ219 and Ϫ258 bp (pDLPTR), between Ϫ222 and Ϫ236 bp (pDCR), or with the addition of two 171-bp p53TRE fragments to the extreme 5Ј-end (pWT-TRE2). Also included was a p21 WAF1 promoter construct containing two 171-bp p53TRE elements at the extreme 5Ј-end (p21-TRE2). Luciferase activities were measured 24 h post-treatment with Adp53 (100 pfu/cell). Results are expressed as the percent mean fold induction Ϯ S.E. relative to the intact PTGF-␤ promoter (pWT) for pDLPTR, pDCR, and pWT-TRE2; activity of the p21-TRE2 is expressed relative to the intact p21 WAF1 promoter (p21-lux; not shown). Relative positions of p53 binding sites (1 and 2 in pWT; A, B, and C in p21 WAF1 ) and the p53TRE (R) in each promoter construct are depicted schematically.
p53-mediated Transactivation of the PTGF-␤ Promoter within site p53-1 compared with just three in site p53-2, predicts that site p53-1 can more readily assume this stem-loop conformation. This may account for the two major differences observed between these two p53 binding sites: the greater affinity of site p53-1 for p53, and the ability of site p53-1 to bind p53 in the absence of Ab421. Latent p53 binding to site p53-1 contrasts with EMSA results reported by Tan et al. (34) showing that Ab421 was required for p53 binding to both sites. Work by several groups has led to the allosteric model of p53 latency, which suggests that the extreme carboxyl terminus of p53 sterically hinders its own DNA binding domain (47). Modification of p53 through carboxyl-terminal acetylation, truncation, or binding by Ab421 results in a conformational change that alleviates this interference (48 -50). However, recent NMR spectroscopic studies examining the structure of activated and latent p53 (51), along with chromatin immunoprecipitation experiments studying the in vivo binding of p53 to endogenous target genes (52), dispute the idea that p53 exists in a latent state in which it is unable to bind DNA. Indeed, Espinosa and Emerson (22) showed that unmodified p53 could associate with p53 binding sites in EMSAs when these elements were flanked by longer fragments of DNA (ϳ70 bp), suggesting that DNA flanking p53 binding sites may help to stabilize secondary structure formation. Our observation that site p53-1 binds p53 in the absence of Ab421 may be related to our use of a longer (26 bp) oligonucleotide probe as compared with Tan et al. (34) (20 bp). The addition of 6 bp, combined with the high internal symmetry of site p53-1, may have permitted formation of a stem-loop structure under our EMSA conditions. In contrast, the internal symmetry of site p53-2 may not be sufficient to allow formation of this secondary structure.
There are numerous examples of trans-acting factors that bind p53 and inhibit transactivation of target genes (28,29,31,53), but none is known to be a DNA-binding protein. To our knowledge the p53TRE is the first example of a cis-acting sequence element identified on the basis of its involvement in the suppression of p53-mediated promoter activation. Evidence for the existence of homologous DNA elements in p53 target genes such as p21 WAF1 , 14-3-3, PIG-3, GML, and proliferating cell nuclear antigen, along with expression of a p53TRE-binding protein in at least three different cancer cell lines, is consistent with a general role for this regulatory element in modulating the cellular response to p53 activation. Recently, a negative regulatory region within the murine and human Bax promoters was able to mute the activity of an endogenous p53 binding site (54); however, sequence analysis of this region has not revealed any homology to the p53TRE. The mechanism by which factor binding to the p53TRE inhibits the PTGF-␤ promoter response to p53 remains to be determined. However, it is likely that the p53TRE interferes directly with p53 binding, perhaps by tethering p53 away from its binding sites or by destabilizing stem-loop structures in p53 binding sites. Alternatively, the p53TRE-binding protein could block associations between p53 and other essential trans-acting factors (e.g. Sp1) within the PTGF-␤ promoter (21,22). Because many of these cofactors associate with either the amino-or carboxyl-terminal domains of p53, it would be interesting to determine whether p53TRE function is dependent on post-translational modifications of one or both of these domains. In combination with p53 binding sites having different affinities for p53, the p53TRE could function to restrict promoter activation until a given threshold concentration of p53 is achieved within the nucleus. Suppressor elements such as the p53TRE represent a novel mechanism for the regulation of p53 transactivation thresholds at the level of individual gene promoters. Further characterization of the p53TRE-binding protein and its association with p53 and associated trans-acting factors will provide valuable insights into transcriptional mechanisms governing the cellular response to p53.