The tumor suppressor protein p53 requires a cofactor to activate transcriptionally the human BAX promoter.

An important regulator of the proapoptotic BAX is the tumor suppressor protein p53. Unlike the p21 gene, in which p53-dependent transcriptional activation is mediated by a response element containing two consensus p53 half-sites, it previously was reported that activation of the BAX element by p53 requires additional sequences. Here, it is demonstrated that the minimal BAX response element capable of mediating p53-dependent transcriptional activation consists of two p53 half-sites plus an adjacent 6 base pairs (5'-GGGCGT-3'). This GC-rich region constitutes a "GC box" capable both of binding members of the Sp family of transcription factors, including Sp1 in vitro, and of conferring Sp1-dependent transcriptional activation on a minimal promoter in cells. Mutations within this GC box abrogated the ability of p53 to activate transcription without affecting the affinity of p53 for its binding site, demonstrating that these 6 bases are required for p53-dependent activation. In addition, a positive correlation was observed between the ability of p53 to activate transcription in cells and the ability of Sp1 to bind this response element in vitro. Mutations that inhibited Sp1 binding also blocked the ability of p53 to activate transcription through this element. Together, these results suggest a model in which p53 requires the cooperation of Sp1 or a Sp1-like factor to mediate transcriptional activation of the human BAX promoter.

The BCL-2 family of proteins are key mediators of the apoptotic response. One member of this family is the proapoptotic BAX. Preceding apoptosis, cytosolic BAX translocates to the mitochondria and homodimerizes. Homodimeric BAX then is thought to cause the release of cytochrome c (1-3) which subsequently functions as a coactivator of Apaf-1 in the cleavage of pro-caspase-9, initiating programmed cell death (4). BAX exists in equilibrium with two of its homologs, BCL-2 and BCL-X L . Unlike BAX, these two homologs exert antiapoptotic effects by heterodimerizing with BAX in the mitochondria, blocking its ability to release cytochrome c (5,6). Thus, an important determinant of the apoptotic response of a cell is the balance between the levels of BAX and BCL-2/BCL-X L . In this regard, regulation of the level of expression of BAX protein is key.
An important regulator of BAX gene expression is the tumor suppressor protein p53 (7,8). The p53 protein has been implicated in several growth-related pathways, including apoptosis and cell cycle arrest (9,10). The ability of p53 to function as a sequence-specific DNA-binding protein appears to be central to its role as a tumor suppressor (11,12). At its amino terminus, the protein contains a potent transcriptional activation domain (13) that is linked to a central core domain that mediates sequence-specific DNA binding (14 -16). Both of these domains have been shown to be important for p53-mediated growth suppression (17).
A DNA consensus sequence through which p53 binds and activates transcription has been identified. This sequence consists of two palindromic decamers of 5Ј-RRRCWWGYYY-3Ј (where R is a purine; Y is a pyrimidine; and W is an adenine or thymine) separated by 0 -13 base pairs, forming four repeats of the pentamer 5Ј-RRRCW-3Ј alternating between the top and bottom strands of the DNA duplex (18,19). Through sequences similar to this consensus, p53 has been shown to activate the transcription of many genes, including BAX, p21, mdm2, gadd45, IGF-BP3, and cyclin G (8, 20 -26). When compared with alternate p53 targets, studies demonstrate that the BAX gene is differentially regulated by wild-type p53 in a cell typespecific manner (7,27,28). In the mouse, p53-dependent regulation of BAX expression following ionizing radiation is seen in the prostate, thymus, spleen, small intestine, and lung, as well as sympathetic, Purkinje, and olfactory cortical neurons. In the kidney, heart, liver, and brain, however, no p53-dependent regulation of BAX is observed (7,27). Furthermore, the myeloid leukemia ML-1, Burkitt's lymphoma WMN and AG876, and lymphoblastoid NL2 and FWL cell lines induce BAX following ionizing radiation, whereas the fibroblast AG1522 and WI38, colorectal carcinoma RKO, and osteosarcoma U2-OS cell lines fail to do so (28). In addition, several tumor-derived p53 mutants have been identified that are capable of activating transcription through the promoter of the p21 gene but not through the BAX promoter (29 -32). This correlates with an inability of these mutants to trigger apoptosis (29,31,32), suggesting that a failure in the ability of p53 to transactivate the BAX gene may play an important role in tumor formation and progression. Supporting this, Yin et al. (33) demonstrated that BAX is an obligatory downstream effector for the p53-mediated apoptosis that attenuates choroid plexus tumor growth in the TgT121 mouse model. Thus, a complete understanding of the transcriptional regulation of the BAX promoter by p53 may yield important information relevant to our understanding of tumorigenesis.
Here is presented a detailed analysis of the p53 response element located in the promoter of the human BAX gene. The minimal BAX response element capable of mediating p53-dependent transcriptional activation is found to consist of two p53 half-sites plus an adjacent 6 base pairs (5Ј-GGGCGT-3Ј) that * This work was supported by NCI Grant CA69161 from the National Institutes of Health and the Breast Cancer Program of the United States Army Medical Research and Materiel Command Grants DAMD-17-97-1-7336, DAMD-17-97-1-7337, and DAMD-17-99-1-9308). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed: Cancer Center, Box demonstrate sequence-specific binding to the transcription factor Sp1. Mutational analysis of this "GC box" shows it to be required for p53-dependent activation, and a positive correlation between the ability of p53 to activate transcription in cells and the ability of Sp1 to bind this response element in vitro is observed. These results are consistent with a model in which p53 requires the cooperation of Sp1 or a Sp1-like factor to mediate transcriptional activation of the human BAX gene. This presents the intriguing possibility that regulation of this cofactor may represent a novel basis for the cell type-specific control of the proapoptotic BAX by wild-type p53. Transfections-Saos-2 cells were transfected using LipofectAMINE Plus Reagent (Life Technologies, Inc.). 2 ϫ 10 5 cells were seeded into 35-mm plates. Cells were transfected 24 h later according to the manufacturer's instructions. Cellular lysates were prepared 24 h post-transfection, and total protein concentration was determined by protein assay (Bio-Rad), and luciferase assays were quantitated using a commercially available kit (Promega) and a TD-20e Luminometer (Turner). Drosophila SL2 cells were transfected using Cellfectin (Life Technologies, Inc.). 60-mm dishes were seeded with 2 ϫ 10 6 cells in Schneider's Drosophila media containing 10% heat-inactivated fetal bovine serum but no penicillin or streptomycin. The DNA to be transfected was added to 500 l of serum-free media containing 8 l of Cellfectin reagent, mixed gently, and incubated at room temperature for 20 min. This mixture then was added directly to the cells. 48 h post-transfection cells were lysed by sonication (6 ϫ 20 s pulse). Total protein and luciferase activity was determined as above.
HeLa Cell Nuclear Extraction-Unless otherwise stated, all procedures were conducted at 4°C. HeLa S3 cells were obtained as a packed cell pellet from the National Cell Culture Center (Minneapolis, MN). Cell pellets were resuspended in 5 volumes of Buffer A (10 mM HEPES, pH 7.6, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM DTT) and incubated on ice for 10 min. Cells then were centrifuged at 500 ϫ g for 12 min. The supernatant was removed, and the pellet was resuspended in two packed cell volumes of Buffer A. Cells were homogenized 10 times in a Dounce homogenizer with pestle A (tight). The resulting solution was centrifuged at 430 ϫ g for 10 min to pellet the nuclei. The supernatant was decanted, and the pellet was recentrifuged at 24,000 ϫ g for 20 min. The supernatant again was removed. The pellet was resuspended in 3 ml of Buffer C (20 mM HEPES, pH 7.6, 25% glycerol, 420 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.5 mM DTT) per 10 9 cells. The solution was homogenized 10 times with pestle B (loose). The resulting solution was transferred to a beaker and stirred for 30 min on ice. The solution then was centrifuged at 24,000 ϫ g for 30 min. The resulting nuclear extract was dialyzed against Buffer D (20 mM HEPES, pH 7.6, 20% glycerol, 0.1 M KCl, 0.2 mM EDTA, 1.5 mM MgCl 2 , 0.5 mM DTT) for 5 h. The extract was clarified by centrifugation at 24,000 ϫ g for 20 min. Nuclear extracts were aliquoted, frozen in a dry ice/ethanol bath, and stored at Ϫ70°C.
Electrophoretic Mobility Shift Assay-Production of baculovirus-infected Sf9 cell extracts and purification of recombinant human p53 protein were done as described previously (34). Purified p53 protein, extract from Sf9 cells expressing recombinant human Sp1 protein, or HeLa cell nuclear extract was incubated with 3 ng of radiolabeled double-stranded oligonucleotide and antibody (Sp1 PEP-2X, p300 N-15X, and CBP 451X, Santa Cruz Biotechnology), where appropriate, in a total volume of 30 l of DNA binding buffer (20 mM HEPES, pH 7.5, 83 mM NaCl, 0.1 mM EDTA, 12% glycerol, 2 mM MgCl 2 , 2 mM spermidine, 0.7 mM DTT, and 17 g/ml poly(dI-dC)) for 20 min at room temperature. Samples were loaded on a native 4% acrylamide gel in 0.5ϫ TBE and electrophoresed at 4°C at 225 V for 2 h. The gel was dried and exposed to Kodak XAR film using an intensifying screen at Ϫ70°C. Phosphorimaging and densitometry data were collected with a Personal Molecular Imager FX and a GS-710 Calibrated Imaging Densitometer (Bio-Rad), and analyzed with Quantity One software (Bio-Rad).

RESULTS
All Three Potential p53 Half-sites Are Required for the p53dependent Transcriptional Activation of the Human BAX Promoter-Previously it was demonstrated that in isolation the p53 response element from the human BAX promoter required sequences from three adjacent half-sites to confer p53-dependent transcriptional activation on a minimal promoter (37). To confirm the requirement of all three half-sites in the context of the BAX promoter, luciferase reporter plasmids with various deletions in the BAX promoter, both in and around the p53 response element, were cotransfected with either pCMV or a wild-type p53 expression vector into the p53-negative osteosarcoma Saos-2 cell line (Fig. 1). The previously characterized p53 response element of the BAX promoter is contained within the sequence from Ϫ113 to Ϫ83 from the start site of transcription.
There was no significant difference between the p53-dependent transactivation of either a reporter construct lacking sequences 5Ј to the p53 response element (pBAX Ϫ127/ϩ51) or the fulllength promoter construct (pBAX Ϫ315/ϩ51) (Fig. 1A). Deletion of a larger fragment, including the p53 response element (pBAX Ϫ76/ϩ51), produced a reporter construct that was unresponsive to wild-type p53 (Fig. 1A). Furthermore, targeted deletion of the promoter region containing the p53 response element (pBAX⌬Ϫ126/Ϫ77) also produced a reporter plasmid that was unresponsive to wild-type p53 (Fig. 1A). These results show that Ϫ113 to Ϫ83 is the only region, within the 366-base pair promoter fragment investigated, that affects the ability of p53 to activate transcription.
The region from Ϫ113 to Ϫ83 contains three potential p53 half-sites (represented in Fig. 1B as the light gray, white, and dark gray boxes). The role of each of these half-sites in the p53-dependent activation of the BAX promoter was examined. Removal of the first half-site from Ϫ113 to Ϫ104 (pBAX⌬Ϫ113/ Ϫ104) significantly reduced the ability of p53 to activate transcription through this promoter (Fig. 1B, compare 63-fold with pBAX Ϫ315/ϩ51 to 7-fold with pBAX⌬Ϫ113/Ϫ104), whereas removal of the second (pBAX⌬Ϫ103/Ϫ93) or the third half-site (pBAX⌬Ϫ92/Ϫ83) completely abolished the ability of p53 to activate transcriptionally the promoter (Fig. 1B). Consistent with the above results, removal of the first and second halfsites in combination (pBAX⌬Ϫ113/Ϫ93) also abolished the ability of p53 to activate transcriptionally the promoter (Fig. 1B).
FIG. 1. All three potential p53 halfsites are required for the p53dependent transcriptional activation of the human BAX promoter. A and B, Saos-2 cells were transfected as described under "Experimental Procedures" with 1 g of the indicated pBAX reporter constructs in the presence of either 10 ng of pCMV (white bars) or 10 ng of pCMV-p53 wt (gray bars). 24 h post-transfection cells were lysed and assayed for total protein and luciferase activity as described under "Experimental Procedures." The indicated values are the average of three independent experiments each performed in duplicate. Error bars correspond to one S.D. The numbers above each bar indicate the fold activation for each reporter construct observed with pCMV-p53 wt as compared with pCMV. A, the previously identified p53 response element is indicated by the dark gray box at Ϫ113 to Ϫ83. B, the three potential p53 half-sites are represented by the light gray (Ϫ113 to Ϫ104), white (Ϫ102 to Ϫ93), and dark gray (Ϫ92 to Ϫ83) boxes.
These results demonstrate that, as was observed with the isolated response element (37), p53 requires sequences from all three potential half-sites to mediate transcriptional activation of the BAX promoter.
The First Two Potential p53 Half-sites Constitute a Bona Fide p53 Response Element-Each of the three potential p53 half-sites located in the BAX promoter from Ϫ113 to Ϫ83 closely resembles the consensus sequence of 5Ј-RRRCWW-GYYY-3Ј (represented in Fig. 2 by the light gray, white, and dark gray boxes). The first, located at Ϫ113 to Ϫ104, deviates from the consensus at 2 bases (Ϫ113 and Ϫ104). The second half-site matches the consensus sequence at all 20 base pairs and is located at Ϫ102 to Ϫ93. The third half-site is located at Ϫ92 to Ϫ83 and deviates from the consensus at three bases (Ϫ84, Ϫ85, and Ϫ88)(see Fig. 2). These three half-sites can combine in different ways to produce a total of three possible p53 complete binding sites (half-sites 1 and 2, 2 and 3, and 1 and 3). Previous studies demonstrated that in electrophoretic mobility shift assays (EMSA), double-stranded oligonucleotides representing both Ϫ113 to Ϫ93 (half-sites 1 and 2) and Ϫ102 to Ϫ83 (half-sites 2 and 3) are capable of binding p53 in a sequence-specific manner with similar affinities (37). When cloned upstream of the adenovirus E1b minimal promoter in the pTATA luciferase reporter plasmid, however, the combination of the first and second half-sites (Ϫ113 to Ϫ93) is unable to mediate p53-dependent transcriptional activation (37). To examine further the ability of p53 to interact with this sequence in cells, the Ϫ113 to Ϫ93 sequence was multimerized (as three copies) and cloned into the pTATA luciferase reporter plasmid. This reporter plasmid was cotransfected with either pCMV or a wild-type p53 expression vector in the Saos-2 cell line (Fig. 2). These three copies of this p53-binding site were capable of mediating a significant degree of activation in response to p53 (Fig. 2, compare 4-fold with pTATA-113/Ϫ93 to 142-fold with pTATA(Ϫ113/Ϫ93) 3 ), demonstrating that the sequence from Ϫ113 to Ϫ93 is indeed a bona fide p53 response element capable of both binding p53 in a sequence-specific manner in vitro and mediating p53-dependent transcriptional activation in cells. Confirming previous results, p53 was able to activate transcription through the second and third half-sites (Ϫ102 to Ϫ83), but this activation was significantly reduced as compared with that mediated by all three half-sites combined (Fig. 2, compare 44-fold with pTATA-102/Ϫ83 and 153-fold with pTATA-113/Ϫ83). To test the ability of half-sites one and three to mediate p53-dependent transcriptional activation, a synthetic oligonucleotide corresponding to Ϫ113 to Ϫ83 of the BAX promoter, with Ϫ102 to Ϫ93 scrambled to remove any contribution of the second half-site, was cloned into the pTATA reporter plasmid. This construct failed to be activated by p53 (Fig. 2, pTATA-113/Ϫ83(sc Ϫ102/Ϫ93)). The third half-site in isolation (Ϫ92 to Ϫ83) also failed to mediate p53-dependent transcriptional activation (Fig. 2, pTATA-92/Ϫ83).
Sp1 Binds with Sequence Specificity to and Activates Transcription through the p53 Response Element from the Human BAX Promoter-We previously reported the identification of a nuclear factor, termed Binder of BAX 1 (BoB1), that interacts with sequence specificity with the same region of the human BAX promoter that is required for p53-dependent transcriptional activation (37). These previous studies demonstrated that this factor binds to sequences within the region of Ϫ102 to Ϫ83. Analysis of this region using a MatInspector search of the TRANSFAC data base (38,39) showed that it contains a sequence that potentially could bind the transcription factor Sp1. To test this, a synthetic oligonucleotide corresponding to Ϫ102 to Ϫ83 of the BAX promoter was used as a radiolabeled probe in an EMSA with HeLa cell nuclear extract (Fig. 3). As reported previously for Saos-2 (37), HeLa cell nuclear extract contains a factor that demonstrated marked sequence specificity for the labeled BAX probe. This factor was successfully competed by increasing amounts of unlabeled probe (Fig. 3, lanes 7-9) as well as by increasing amounts of oligonucleotide corresponding to the DNA-binding consensus sequence of Sp1 (Fig. 3, lanes [13][14][15]. This binding was specific, as an oligonucleotide corresponding to the 5Ј p53 response element from the human p21 promoter failed to compete for binding (Fig. 3, lanes 10 -12). In addition, this factor was successfully bound by an anti-Sp1 antibody, as demonstrated by a "supershifted" complex ( Fig. 3,  lanes 2 and 3), whereas a control anti-p300 antibody failed to bind the factor (Fig. 3, lanes 4 and 5). Together, these data demon-FIG. 2. The first two potential p53 half-sites constitute a bona fide p53 response element. Saos-2 cells were transfected as described under "Experimental Procedures" with 1 g of the indicated pTATA reporter constructs in the presence of either 10 ng of pCMV (white bars) or 10 ng of pCMV-p53 wt (gray bars). 24 h post-transfection cells were lysed and assayed for total protein and luciferase activity as described under "Experimental Procedures." The indicated values are the average of three independent experiments each performed in duplicate. Error bars correspond to 1 S.D. The numbers above each bar indicate the fold activation for each reporter construct observed with pCMV-p53 wt as compared with pCMV. The sequence of the BAX promoter from Ϫ113 to Ϫ83 is given at the top of the figure. Potential p53 quartersites are indicated by the solid bars above and below the sequence. Bases that deviate from the p53 DNA-binding consensus sequence are indicated by asterisks. The three potential half-sites are indicated by the brackets labeled 1-3, respectively, and are represented graphically as the light gray, white, and dark gray boxes, respectively. The vertical arrow above the BAX sequence indicates the 1-base pair insert between the first and second half-sites.
p53-dependent Transactivation of bax strate that Sp1 can bind a portion of the p53 response element from the human BAX promoter in a sequence-specific manner.
To delineate further the sequences important for Sp1 binding, oligonucleotides were synthesized that replaced portions of the BAX sequence with corresponding sequence from the p21 5Ј p53 response element. The sequence from Ϫ102 to Ϫ83 in the BAX promoter contains two p53 half-sites (Ϫ102 to Ϫ93 and Ϫ92 to Ϫ83), and the p21 5Ј element also consists of two p53 half-sites. Hybrid oligonucleotides were synthesized in which the first of the two half-sites in the BAX element was combined with the second half-site of the p21 5Ј element and vice versa. The oligonucleotide corresponding to Ϫ102 to Ϫ83 of the BAX promoter again was used as a radiolabeled probe with HeLa nuclear extract in an EMSA (Fig. 4). Competitions, using unlabeled probe as well as the oligonucleotides corresponding to the p21 5Ј element and the two hybrid elements, were conducted. Sp1 bound the radiolabeled probe (Fig. 4, lane 1) and was recognized by an anti-Sp1 antibody (Fig. 4, lane 2) but not by a control anti-CBP antibody (Fig. 4, lane 3). Both unlabeled probe and the Sp1 DNA-binding consensus site oligonucleotide effectively competed for Sp1 binding (Fig. 4, lanes 4 -5 and  12-13, respectively), whereas the p21 5Ј element did not (Fig. 4,  lanes 10 -11). Consistent with the notion that Sp1 binds DNA through GC box regions, the hybrid oligonucleotide in which the first half-site is derived from the p21 sequence and the second half-site from the BAX sequence (Ϫ92 to Ϫ83, 5Ј-GGGCGTGGGC-3Ј) effectively competed for Sp1 binding (Fig.  4, lanes 8 -9), whereas the other hybrid oligonucleotide that replaces this GC-rich region with sequence from the p21 5Ј element demonstrated a significantly reduced affinity for Sp1 binding (Fig. 4, lanes 6 -7). These data indicate that Sp1 binds to sequence within Ϫ92 to Ϫ83 of the BAX promoter.
To determine whether or not Sp1 can interact with this element in cells, a pTATA luciferase reporter plasmid containing Ϫ113 to Ϫ77 of the human BAX promoter was cotransfected with increasing amounts of an Sp1 expression vector into the Sp1-deficient Drosophila SL2 cell line (Fig. 5). Expression of Sp1 successfully activated transcription of this reporter and yet failed to activate transcription of a control plasmid containing the 5Ј p53 response element of the p21 promoter (Fig. 5).
Consistent with the in vitro EMSA results, this confirms that Sp1 is capable of activating transcription through the p53 response element of the human BAX promoter.
The Ability of Sp1 to Bind the p53 Response Element of the BAX Promoter in Vitro Correlates with the Ability of p53 to Activate Transcription through This Element in Cells-To explore the significance of the Sp1-binding site to the ability of p53 to activate transcription through the BAX promoter, nucleotide substitutions were identified that differentially affected the ability of p53 to activate transcription through its response element in the BAX promoter (Ϫ113 to Ϫ83). Two mutated forms of the p53 response element from the BAX promoter, in which the indicated guanine bases were replaced with adenines (Fig. 6A, GGϪ92/Ϫ91AA and GGϪ85/Ϫ84AA), were cloned into the pTATA luciferase reporter plasmid. In cotransfection assays with a wild-type p53 expression vector in FIG. 3. Sp1 binds with sequence specificity to the p53 response element from the human BAX promoter. An electrophoretic mobility shift assay was performed using an oligonucleotide corresponding to the Ϫ102/Ϫ83 sequence from the human BAX promoter as radiolabeled probe. 8 g of HeLa cell nuclear extract was incubated with 3 ng of the probe alone (lanes 1 and 6) 10 and 11), BAX/p21 5Ј hybrid oligonucleotide (lanes 6 and 7) or p21 5Ј/BAX hybrid oligonucleotide (lanes 8 and 9), or a 10-or 20-fold molar excess of the unlabeled Sp1 consensus oligonucleotide (lanes 12 and 13). The arrows indicate the positions of the Sp1-DNA and the supershifted antibody-Sp1-DNA complexes. B, the sequences of the human BAX promoter from Ϫ102 to Ϫ83 from the start site of transcription (gray boxes) and the human p21 promoter from Ϫ2281 to Ϫ2262 from the start site of transcription (white boxes; corresponding to the p21 5Ј oligonucleotide) are shown. Each sequence is divided into two with the first half indicated as A and the second half indicated as B. Oligonucleotides in A are represented graphically according to this color and letter scheme. For example, the BAX/p21 5Ј hybrid oligonucleotide that corresponds to the first half of the BAX sequence followed by the second half of the p21 sequence is indicated by a gray box labeled A followed by a white box labeled B.

p53-dependent Transactivation of bax
the Saos-2 cell line, substitution of bases Ϫ92 and Ϫ91 completely abolished the ability of p53 to activate transcription through this element (Fig. 6A, compare Ϫ113/Ϫ83 to GGϪ92/ Ϫ91AA), whereas substitution of bases Ϫ85 and Ϫ84 did not (Fig. 6A). As observed in Figs. 1 and 2, removal of the third potential half-site (Ϫ92 to Ϫ83) inhibited the ability of p53 to mediate transcriptional activation through this element (Fig.  6A, compare Ϫ113/Ϫ83 and Ϫ113/Ϫ93), demonstrating the requirement for this Sp1-binding sequence in the p53-dependent transcriptional activation of this element.
Both of these mutant sequences were assayed for their ability to bind purified p53 in an EMSA. An oligonucleotide corresponding to Ϫ113 to Ϫ77 of the BAX promoter was used as a radiolabeled probe with purified p53 in an EMSA (Fig. 6B). Competitions were performed with increasing amounts of an oligonucleotide corresponding to Ϫ113 to Ϫ83 of the BAX promoter and the two mutant oligonucleotides. When compared with the wild-type oligonucleotide, both mutant oligonucleotides displayed a slightly decreased affinity for p53 (Fig. 6B,  compare lanes 2-4 with lanes 5-7 and 8 -10; Fig. 6C). Compared with one another, however, both mutant oligonucleotides demonstrated a comparable affinity for p53 (Fig. 6, A and B), suggesting that the differences in p53-dependent transcriptional activation observed in Fig. 6A are not due to differences in the affinity of p53 for the two sequences. In contrast, the abilities of the two mutant sequences to bind Sp1 differed (Fig.  6, D and E). An oligonucleotide corresponding to Ϫ113 to Ϫ77 of the BAX promoter was used as radiolabeled probe with extract from Sf9 cells expressing recombinant human Sp1 protein in an EMSA (Fig. 6D). Sp1 bound the probe and was recognized by an anti-Sp1 antibody (Fig. 6D, lanes 1-2). Sp1 binding was successfully competed by unlabeled BAX Ϫ113/ Ϫ83 oligonucleotide as well as by the GGϪ85/Ϫ84AA mutated oligonucleotide (Fig. 6D, lanes 3-4 and 9 -11, respectively; Fig.  6E). The GGϪ92/Ϫ91AA mutant, however, demonstrated a significant decrease in affinity for Sp1 (Fig. 6D, compare lanes  3-5 and 9 -11 to lanes 6 -8; Fig. 6E).
The results with the GGϪ85/Ϫ84AA mutant presented in Fig. 6 suggest that not all of the bases contained within the third potential half-site of the p53 response element are required for p53-dependent transcriptional activation. To iden-tify the minimal sequence elements required to mediate p53dependent transactivation, a series of oligonucleotides was synthesized in which each of the 10 bases of the third potential half-site (Ϫ92 to Ϫ83) was individually replaced. These mutant oligonucleotides then were cloned into the pTATA luciferase reporter plasmid and tested for their responsiveness to p53 in a cotransfection assay in the Saos-2 cell line (Fig. 7). Consistent with the results in Fig. 6A, substitution of the bases at either Ϫ85 or Ϫ84 did not inhibit the ability of p53 to activate transcription through this element (Fig. 7, GϪ85T and GϪ84T). Furthermore, substitution of Ϫ86 and Ϫ83 also failed to affect significantly the ability of p53 to activate transcription (Fig. 7, compare Ϫ113/Ϫ83 to GϪ86T and CϪ83A). Substitution of the base at Ϫ87, however, significantly reduced the ability of p53 to activate transcription through this element (Fig. 7, compare Ϫ113/Ϫ83 to TϪ87G). Together, these results suggest that the minimal response element consists of sequence from Ϫ113 to Ϫ87, with Ϫ86 to Ϫ83 being dispensable for p53-dependent transactivation.
To confirm that the bases from Ϫ86 to Ϫ83 are not required for p53-dependent transcriptional activation, two additional mutant oligonucleotides were synthesized. The first mutant was generated by replacing all 10 nucleotides from Ϫ92 to Ϫ83 (Fig. 8A, scϪ92/Ϫ83). The 4 bases from Ϫ86 to Ϫ83 were substituted as indicated to generate the second mutant oligonucleotide (Fig. 8A, scϪ86/Ϫ83). Each oligonucleotide was cloned into the pTATA vector and tested for its responsiveness to p53 in a cotransfection assay (Fig. 8A). As observed with the reporter plasmid in which the sequence from Ϫ92 to Ϫ83 is removed entirely (pTATAϪ113/Ϫ93), the first mutant, in which all 10 bases of the third potential half-site (Ϫ92 to Ϫ83) are replaced, showed little to no response to p53 (Fig. 8A, compare pTATAϪ113/Ϫ93 to pTATAϪ113/Ϫ83 and pTATAscϪ92/Ϫ83). In contrast, the second mutant, in which only the last 4 bases of the element (Ϫ86 to Ϫ83) are replaced, was efficiently activated by p53 (Fig. 8A, compare 312-fold with pTATAϪ113/Ϫ83 to 323-fold with pTATAscϪ86/Ϫ83). This result demonstrates that the minimal p53 response element in the BAX promoter consists of sequence from Ϫ113 to Ϫ87. In an EMSA both mutants displayed a decreased affinity for p53 as compared with the wild-type sequence (Fig. 8B, compare lanes  2-4 to lanes 8 -10 and 11-13; Fig. 8C). When compared with each other, there was no significant difference in the affinity of p53 for the two mutant sequences (Fig. 8B, compare lanes 8 -10  to 11-13; Fig. 8C). This suggests that the differences in transcriptional activation observed in Fig. 8A cannot be explained by differences in p53 affinities. Furthermore, the oligonucleotide corresponding to Ϫ113 to Ϫ93 displayed a similar affinity for p53 as the two mutant oligonucleotides (Fig. 8B, compare  lanes 5-7 to lanes 8 -10 and 11-13; Fig. 8C) consistent with the idea that, in the case of the two mutants, p53 is interacting with the first and the second half-sites only. The sc Ϫ86/Ϫ83 mutant oligonucleotide efficiently competed for Sp1 binding in an EMSA (Fig. 8D, compare lanes 2-4 to lanes 8 -10; Fig. 8E), whereas the ability of the sc Ϫ92/Ϫ83 mutant to bind Sp1 was significantly reduced compared with the wild-type sequence (Fig. 8D, compare lanes 2-4 to lanes 5-7; Fig. 8E), further strengthening the correlation between Sp1 binding in vitro and p53 activation in cells. DISCUSSION The data presented in this report demonstrate that the minimum p53 response element in the BAX promoter consists of the sequence from Ϫ113 to Ϫ87 from the start site of transcription. This sequence contains a p53-binding site (Ϫ113 to Ϫ93) that can function as a bona fide response element as demonstrated by its ability when multimerized to confer p53-depend- p53-dependent Transactivation of bax ent transcriptional activation on a minimal promoter (Fig. 2). Immediately adjacent to this p53-binding site are 6 base pairs that are GC-rich in nature (Ϫ92 to Ϫ87: 5Ј-GGGCGT-3Ј). These 6 bases are required for p53-dependent transcriptional activation as deletion or mutation of this region in the context of either the promoter or the isolated response element completely abrogates the ability of p53 to activate transcription through this sequence (Figs. 1B, 2, 6A, and 8A). The addition of these bases to the Ϫ113/Ϫ93 sequence appears to have little effect on the affinity of p53 for this sequence (Fig. 8, B and C), consistent with a model in which these 6 bases function to recruit a co-activator as opposed to simply enhancing p53 binding. Furthermore, these 6 base pairs mediate sequence-specific binding to the Sp1 transcription factor (Figs. 3, 4, 6D, and 8D), and a positive correlation is seen between the ability of Sp1 to bind this element in vitro and the ability of p53 to mediate transcriptional activation through its response element in cells (Figs. 6 and 8). In addition, the results with electrophoretic mobility shift assays with the GGϪ92/Ϫ91AA mutant oligonucleotide (Fig. 6B) are not consistent with the published p53 DNA-binding consensus sequence of (RRRCWWGYYY) 2 (18,19). This consensus allows for a purine in the first three posi-FIG. 6. A mutant element that fails to bind Sp1 in vitro also fails to confer p53-dependent transcriptional activation in cells. A, Saos-2 cells were transfected as described under "Experimental Procedures" with 1 g of the indicated pTATA reporter constructs in the presence of either 10 ng of pCMV (white bars) or 10 ng of pCMV-p53 wt (gray bars). 24 h post-transfection cells were lysed and assayed for total protein and luciferase activity as described under "Experimental Procedures." The indicated values are the average of five independent experiments each performed in duplicate. Error bars correspond to 1 S.D. The GC-rich region that binds Sp1 is shown by the boxed sequence. Bases in the wild-type sequence that were mutated are shown in gray with the corresponding mutations indicated above. B, an electrophoretic mobility shift assay was performed using an oligonucleotide corresponding to the Ϫ113/Ϫ77 sequence from the human BAX promoter as radiolabeled probe. 50 ng of purified p53 was incubated with 3 ng of the probe alone (lane 1) or in the presence of a 500-, 1000-, or 1500-fold molar excess of either the unlabeled BAX Ϫ113/Ϫ83 oligonucleotide (lanes 2-4), the GGϪ92/Ϫ91AA oligonucleotide (lanes 5-7), or the GGϪ85/Ϫ84AA oligonucleotide (lanes 8 -10). The arrow indicates the position of the p53-DNA complex. The vertical bar between lanes 4 and 5 represents the removal of irrelevant lanes from the gel. C, bands were quantitated by densitometry. D, an electrophoretic mobility shift assay was performed using an oligonucleotide corresponding to the Ϫ113/Ϫ77 sequence from the human BAX promoter as radiolabeled probe. Extract from Sf9 cells expressing human recombinant p53-dependent Transactivation of bax tions of each half-site. The GGϪ92/Ϫ91AA mutant contains a conservative substitution of purines (adenines) for purines (guanines) and, as such, does not represent a substantive change in terms of the p53 DNA-binding consensus sequence. This substitution, however, did produce a significant decrease in the ability of p53 to bind to this oligonucleotide in vitro (Fig.  6B), suggesting that, in these limited circumstances, the p53 DNA-binding sequence involves greater specificity than implied by the consensus.
Previous studies have suggested a connection between p53 and Sp1. The two proteins physically interact under certain circumstances (40 -42), and, transcriptionally, p53 and Sp1 have been shown to function in a cooperative manner in some settings and an antagonistic manner in others (41,43,44). In addition to p53, Sp1 has been found to synergize with other transcription factors, including YY1 and SREBP (45)(46)(47). Studies with the Sp family of transcription factors, however, are complicated by the fact that there are at least 16 mammalian members of this family. Due to marked conservation in the DNA-binding domain, many of these family members have similar if not identical in vitro DNA binding characteristics (48,49). Originally, this led to the misclassification of many GC boxes solely as Sp1-binding sites because of the ubiquitous nature of Sp1 and the fact that it was the first family member cloned. Given this, the possibility exists that the true in vivo cofactor required for the p53-dependent transactivation of the BAX promoter is an Sp1-related family member that is ob-scured in in vitro assays by the sheer abundance of Sp1 in nuclear extracts from tissue culture cells. Consistent with this, antibodies used in a supershift EMSA identified other Sp family members as minor components of the Sp1-DNA complex. 2 Furthermore, cotransfection assays in the Sp1-deficient Drosophila SL2 cell line failed to demonstrate cooperation between Sp1 and p53 in transcriptionally activating the p53 response element of the BAX promoter. 2 The Drosophila assays, however, are difficult to interpret as the ability of p53 alone to activate transcription through a control plasmid was significantly impaired in the SL2 cell line. Complicating interpretation of the results in the Drosophila system is the recent identification of a Drosophila p53 homolog (50,51) that may affect the ability of transiently expressed human p53 to function properly in this system.
Regardless of whether the cofactor required for the p53-dependent transactivation of the BAX promoter is Sp1 or a related family member, the requirement of this cooperating protein suggests a model for the observed cell type-and tumor type-specific regulation of the BAX gene by wild-type p53 (Fig.  9). In this model, cells that are permissive to p53-dependent up-regulation of the BAX gene express both p53 and the cofactor, and these proteins function together to activate transcriptionally the gene. In those cells that fail to show p53-dependent 2 E. C. Thornborrow 7. Mutational analysis shows that the BAX promoter sequence from ؊86 to ؊83 is not required for p53-dependent transcriptional activation. Saos-2 cells were transfected as described under "Experimental Procedures" with 1 g of the indicated pTATA reporter constructs in the presence of either 10 ng of pCMV (white bars) or 10 ng of pCMV-p53 wt (gray bars). 24 h posttransfection cells were lysed and assayed for total protein and luciferase activity as described under "Experimental Procedures." The indicated values are the average of three independent experiments each performed in duplicate. Error bars correspond to 1 S.D. The GC-rich region that binds Sp1 is shown by the boxed sequence. Bases in the wild-type sequence that were mutated are shown in gray with the corresponding mutations indicated above.
p53-dependent Transactivation of bax BAX expression, one can propose three possible mechanisms to explain the apparent failure of wild-type p53 to activate the BAX gene (Fig. 9). First, the required cofactor may be absent, either due to mutation or due to cell type-specific limitations on its expression. Second, this factor may be inactivated by posttranslational modification. Finally, another factor that cannot cooperate with p53 may compete with the cofactor for binding to its site in the BAX promoter. Data with the Sp family of transcription factors support each of these possibilities. Although several of the Sp family members, like Sp1, are ubiquitously expressed, other members of the family display high degrees of tissue specificity (48,49). Even the ubiquitously expressed family members fluctuate in levels under particular cellular conditions (52)(53)(54)(55). Sp1 mRNA, for example, varies up to 100-fold depending on the cell type and developmental stage of the mouse (56). Consistent with a model of post-translational modification, certain Sp family members, including Sp1 and EKLF, are phosphorylated, glycosylated, and acetylated (57)(58)(59). Finally, given the high level of conservation in the DNAbinding domain of the Sp family of transcription factors, it is not surprising that DNA binding competition can be observed between various members of this family. In certain cases, including Sp1/Sp3, BTEB1/AP-2rep, and BKLF/EKLF, this competition has ramifications on gene expression (60 -62). In each FIG. 8. The minimal element from the BAX promoter that confers p53-dependent transcriptional activation consists of a single p53-binding site and an adjacent Sp1-binding site. A, Saos-2 cells were transfected as described under "Experimental Procedures" with 1 g of the indicated pTATA reporter constructs in the presence of either 10 ng of pCMV (white bars) or 10 ng of pCMV-p53 wt (gray bars). 24 h post-transfection cells were lysed and assayed for total protein and luciferase activity as described under "Experimental Procedures." The indicated values are the average of three independent experiments each performed in duplicate. Error bars correspond to 1 S.D. The numbers above each bar indicate the fold activation for each reporter construct observed with pCMV-p53 wt as compared with pCMV. The GC-rich region that binds Sp1 is shown by the boxed sequence. Bases in the wild-type sequence that were mutated are shown in gray with the corresponding mutations indicated above. B, an electrophoretic mobility shift assay was performed using an oligonucleotide corresponding to the Ϫ113/Ϫ83 sequence from the human BAX promoter as radiolabeled probe. 50 ng of purified p53 was incubated with 3 ng of the probe alone (lane 1) or in the presence of a 500-, 1000-, or 1500-fold molar excess of either the unlabeled BAX Ϫ113/Ϫ83 oligonucleotide (lanes 2-4), the BAX Ϫ113/Ϫ93 oligonucleotide, the BAX sc Ϫ92/Ϫ83 oligonucleotide ( lanes 5-7), or the BAX sc Ϫ86/Ϫ83 oligonucleotide (lanes 8 -10). The arrow indicates the position of the p53-DNA complex. C, bands were quantitated by phosphorimaging. D, an electrophoretic mobility shift assay was performed using an oligonucleotide corresponding to the Ϫ113/Ϫ83 sequence from the human BAX promoter as radiolabeled probe. Extract from Sf9 cells expressing human recombinant Sp1 protein was incubated with 3 ng of the probe alone (lane 1) or in the presence of a 10-, 50-, or 100-fold molar excess of either the unlabeled BAX Ϫ113/Ϫ83 oligonucleotide (lanes 2-4), the BAX sc Ϫ92/Ϫ83 oligonucleotide ( lanes 5-7), or the BAX sc Ϫ86/Ϫ83 oligonucleotide (lanes 8 -10). The arrow indicates the position of the Sp1-DNA complex. E, bands were quantitated by phosphorimaging.
p53-dependent Transactivation of bax case, transcriptional activation by one family member is repressed by the other member by competing for the same DNAbinding site. The data in this report, in combination with the previous studies of the Sp family of transcription factors, support a model in which the regulation of a required cofactor controls cell type-specific p53-dependent expression of the BAX gene.
The ability of the proapoptotic BAX to function as a tumor suppressor protein has been substantiated by several studies. In certain mouse models, BAX has been shown to be an important mediator of p53-dependent apoptosis and a suppressor of oncogenic transformation, with loss of BAX leading to accelerated rates of tumor growth, increased tumor numbers, larger tumor mass, and decreased survival rates (63, 64). A significant correlation between decreased BAX expression and both a corresponding resistance to apoptotic stimuli, as well as a shorter survival period also have been observed in a number of human tumor types, including breast, ovarian, pancreatic, colorectal, and non-Hodgkin's lymphoma (65)(66)(67)(68)(69). In addition, in colon and gastric cancers of the microsatellite mutator phenotype mutational inactivation of the BAX gene has been shown to confer a strong survival advantage during tumor clonal evolution (70). Complimenting these data are observations showing that overexpression of the BAX protein in certain tumor cell lines both sensitizes these cells to chemotherapy-and radiation-induced apoptosis and reduces their ability to form tumors in SCID mice (71)(72)(73). Together, these results strongly support a tumor suppressor role for the BAX protein.
An important regulator of the BAX gene is the tumor suppressor protein p53. Several reports have demonstrated the significance of the p53-BAX pathway in tumor suppression. Both the identification of tumor-derived p53 mutants that selectively fail to activate transcription through the BAX promoter and subsequently fail to induce apoptosis (29 -32) as well as the TgT121 transgenic studies that demonstrate that BAX is an obligatory downstream effector of p53 in the suppression of choroid plexus tumor growth (33) suggest that the ability of p53 to activate transcription through the BAX promoter is important to the tumor suppressor function of p53. Furthermore, the resistance of certain tumor cell lines to radiation therapy is associated with a failure of wild-type p53 to induce BAX expression (28,74), and certain human tumors have been identified that are genetically wild-type for both p53 and BAX and yet fail to express significant levels of BAX protein (75). Thus, a complete understanding of the transcriptional regulation of the BAX gene by the tumor suppressor p53 may provide important information concerning both the molecular origins of cancer as well as the development of tumor resistance to certain cancer treatments. FIG. 9. Model for the cell type-specific regulation of the BAX promoter by the tumor suppressor protein p53. A, in cells that are permissive to p53-dependent transcriptional activation of the BAX gene, p53 and the required cofactor cooperate to mediate activation. In cells that do not support the p53-BAX pathway, three possible mechanisms may explain the apparent failure of wild-type p53 to activate the BAX gene. B, the cofactor may be absent due to mutation or to cell type-specific limitations on its expression. C, the cofactor may be inactivated by post-translational modification as follows: P, phosphorylation; G, glycosylation; or A, acetylation. D, another factor that cannot cooperate with p53 may compete with the required cofactor for binding to its site in the BAX promoter. The p53-binding site (Ϫ113 to Ϫ93) is represented by the black box. The Sp1-binding site (Ϫ93 to Ϫ87) is represented by the white box. p53, the required cofactor, and the inhibitory factor are represented by the gray circle, the dotted oval, and the cross-hatched triangle, respectively.