(2002) 21, 7901 – 7911 ª 2002 Nature Publishing Group All rights reserved 0950 – 9232/02 $25.00 www.nature.com/onc

Groups of target involved in specific p53 downstream effects cluster into different classes of DNA binding sites

Hua Qian1,4, Ting Wang1, Louie Naumovski2, Charles D Lopez3 and Rainer K Brachmann*,1,5

1Division of Oncology, Department of Medicine, Washington University School of Medicine, 660 S Euclid Avenue, Box 8069, St Louis, Missouri, MO 63110, USA; 2Division of Hematology, Oncology and Stem Cell Transplantation, Stanford University, 269 Campus Drive, CCSR Room 1215, Stanford, California, CA 94305, USA; 3Division of Hematology and Medical Oncology, Department of Medicine, Oregon Health Sciences University, 3181 SW Sam Jackson Park Road, Baird Hall Room 3030, Portland, Oregon, OR 97201, USA

The tumor suppressor protein p53, once activated, can and/or apoptosis (Prives and Hall, 1999; Vogelstein et cause either cell cycle arrest or apoptosis through al., 2000). Direct target genes of p53 have one or more transactivation of target genes with p53 DNA binding p53 DNA binding site(s) (DBS) very similar to the p53 sites (DBS). To investigate the role of p53 DBS in the consensus DBS that consists of two copies of the 10 regulation of this profound, yet poorly understood base pair (bp) motif 5’-PuPuPuC(A/T)(T/A)GPyPyPy- decision of life versus death, we systematically studied 3’ separated by 0 – 13 bp (el-Deiry et al., 1992; Funk et all known and potential p53 DBS. We analysed the DBS al., 1992). However, only one p53 DBS, Type IV separated from surrounding regions in yeast Collagenase, is a perfect match for the p53 consensus and mammalian assays with and without DNA damage. DBS. There are over 60 genes with known or potential p53 efficiently utilized the DBS of MDM2 and of genes p53 DBS (see Tables 1 and 2). The biological activities connected to cell cycle arrest, DNA repair and the death of many are consistent with established p53 effects. For receptor pathway of apoptosis. However, p53 was unable example, the p21 and 14-3-3s genes are involved in to utilize two-thirds of the isolated DBS, a subset that p53-mediated G1 and G2 cell cycle arrest, respectively. included almost all DBS of apoptosis-related genes. The GADD45, p53R2 and PCNA genes have roles in Neither ASPP2, a p53-interacting protein reported to DNA repair, and the growing list of apoptosis-related specifically stimulate p53 transcriptional activity on genes includes Bax, Noxa, p53AIP1, PERP and PIG3. apoptosis-related promoters, nor DNA damage resulted Genes that are part of the death receptor pathway of in p53 utilization of isolated DBS of apoptosis-related apoptosis are also represented, such as Fas, KILLER/ genes. Thus, a major regulation of p53 activity occurs at DR5 and Pidd (Bates and Vousden, 1999; Hengartner, the level of p53 DBS themselves by posing additional 2000; Ko and Prives, 1996; Levine, 1997; Prives and requirements for the successful utilization of apoptosis- Hall, 1999; Rich et al., 2000; Sionov and Haupt, 1999; related DBS. Vogelstein et al., 2000; Vousden, 2000). In addition, Oncogene (2002) 21, 7901 – 7911. doi:10.1038/sj.onc. p53 induces MDM2, a target that creates a 1205974 negative feedback loop with p53 to maintain low p53 protein levels in unstressed cells (Haupt et al., 1997; Keywords: apoptosis; cell cycle; DNA binding site; Kubbutat et al., 1997; Prives, 1998). The role of many DNA repair; p53 other known or potential target genes in the p53 context remains poorly defined. For example, p53 may drive the transcription of growth-promoting genes, such as c-fos,c-met, Epidermal Growth Factor Receptor Introduction and Transforming Growth Factor-a (Elkeles et al., 1999; Ludes-Meyers et al., 1996; Seol et al., 1999; Shin et al., p53 exerts its tumor suppressor function, in part, as a 1995). transcription factor that induces effector genes once It is also still unclear why some cells respond to p53 activated by upstream signals, such as DNA damage. activation with cell cycle arrest, while others undergo Activated p53 results in DNA repair, cell cycle arrest apoptosis. Cell type, amount of DNA damage, the presence of survival factors and the inappropriate activity of appear to be important, but what *Correspondence: RK Brachmann; E-mail: [email protected] exact underlying mechanisms govern the profound Current addresses: 4Department of Surgery, Xin Hua Hospital, decision of survival versus death after p53 activation Shanghai Second Medical University, 1665 Kong Jiang Road, remains unknown (Bates and Vousden, 1999; Sionov 5 Shanghai, 200092, PR China; Division of Hematology/Oncology, and Haupt, 1999; Vousden, 2000). Several models have Department of Medicine, University of California at Irvine, Medical Sciences I, C250, Irvine, California, CA 92697, USA been put forth. In the ‘p53 dumb’ model, activated p53 Received 22 April 2002; revised 12 August 2002; accepted 13 always leads to the same downstream signals, including August 2002 upregulated cell cycle arrest and apoptotis genes, and p53 and p53 DNA binding sites H Qian et al 7902 Table 1 Sequences of known and potential p53 DNA binding sites

The shown sequences are the sequences used in our study. For genes with several p53 DNA binding sites (DBS), we tried to use the terminology of the original papers as much as possible. With a few exceptions, we used the actual p53 DBS plus 5 bp up- and downstream (if the information was available). The p53 consensus DBS (Nat Genet 1, 450) is shown on top (R=A or G; W=A or T; Y=C or T). Our alignment tries as much as possible to take previous alignments into consideration. Bold blue nucleotides in capitals conform to the p53 consensus DBS. Regular blue nucleotides in capitals are additional nucleotides that conform to motifs of the p53 consensus DBS. Bold green nucleotides represent a motif (TGCCT) that was proposed to be important for p53 binding (Science 252, 1708). Both c-fos and MDM2 have two complete p53 DBS, and we have aligned those to each other. References for the p53 DBS sequences can be found in Table 2.

these signals are then modulated by other pathways. In genes (Vousden, 2000). ASPP1 and ASPP2 have been the ‘p53 smart’ model, increased p53 protein levels, characterized as p53 coactivators that specifically specific post-translational p53 modifications and/or the stimulate the apoptotic function of p53. They appear recruitment of specific coactivators determine the to do so by enhancing the DNA binding and expression levels of cell cycle arrest and apoptotis transcriptional activity of p53 at apoptosis-related

Oncogene p53 and p53 DNA binding sites H Qian et al 7903 Table 2 for 67 p53 yeast assay

The phenotypes at 308C and 378C were the same, except where indicated. ‘-’: no growth, ‘+’: minimal growth, ‘++’: moderate growth, ‘+++’: growth similar to p53 consensus DBS. All DBS are of human origin, except where indicated (amouse, brat). The ‘m2’ column indicates the mismatches with the p53 consensus DBS. Bold, shadowed numbers indicate that interspersed sequences between the two half-sites of p53 DBS are present. MDM2 and c-fos have two adjacent p53 DBS and the mismatches of each DBS are shown. Underlined, italic numbers point out the 5 p53 DBS that do not conform to our proposed rules for p53 DBS. References: 1Molecular Cell 1,3;2EMBO J 17, 5015; 3Nat Genet 14, 482; 4EMBO J 13, 4816; 5Oncogene 13, 1965; 6Cell 75, 817; 7Genes Dev 12, 2102; 8Proc Natl Acad Sci U S A 97, 109; 9J Biol Chem 275, 20127; 10J Biol Chem 269, 6383; 11Cell 71, 587; 12Nature 404, 42; 13Proc Natl Acad Sci U S A 93, 895; 14J Biol Chem 275, 3867; 15Oncogene 19, 1735; 16Nat Genet 26, 122; 17Cell 80, 293; 18Oncogene 16, 2177; 19Mol Cell Biol 20, 233; 20Nature 377, 646; 21Mol Cell Biol 20, 5602; 22Science 288, 1053; 23Cell 102, 849; 24Genes Dev 14, 704; 25Nature 389, 300; 26Oncogene 15, 2145; 27Hum Mol Genet 7, 1039; 28Genes Dev 7, 1126; 29Nucleic Acids Res 23, 2584; 30J Biol Chem 273, 15387; 31Oncogene 18, 127; 32Science 252, 1708; 33FEBS Lett 445, 265; 34Mol Cell Biol 17, 6330; 35J Biol Chem 275, 15557; 36Mol Cell Biol 19, 2594; 37J Biol Chem 274, 3565; 38Anticancer Res 16, 105; 39Oncogene 19, 1843; 40J Biol Chem 274, 37679; 41Blood 90, 1373; 42Mol Cell Biol 16, 6009; 43Cancer Res 60, 3722; 44J Biol Chem 274, 12061; 45Nucleic Acids Res 25, 983; 46Nat Med 1, 570; 47Proc Natl Acad Sci U S A 95, 11307; 48Cell Growth Differ 10, 295; 49J Biol Chem 275, 6051; 50Genes Dev 6, 1143; 51FEBS Lett 436, 139; 52Oncogene 16, 2479; 53Cancer Res 57, 3281; 54Nucleic Acids Res 23, 3710; 55J Biol Chem 271, 12414; 56Oncogene 16, 1299; 57Oncogene 18, 7740; 58Mol Cell Biol 15, 4694; 59Gene 193,5;60J Biol Chem 270, 19312; 61Carcinogenesis 17, 2559. p53 DBS for the following genes were published too late for our study: Apaf-1 (Nat Cell Biol 3, 552), APC (J Biol Chem 276, 18193), c-Ha-Ras (Oncogene 19, 5831), Caspase-1 (J Biol Chem 276, 10585), hMSH2 (J Biol Chem 276, 27363), p53DINP1 (Mol Cell 8, 85), PTEN (Mol Cell 8, 317) and PUMA (Mol Cell 7, 673; Mol Cell 7, 683) promoters (Lane, 2001; Samuels-Lev et al., 2001). regulatory mechanisms by providing more or less Thus, the activity of ASPP proteins would fit into the perfect binding sites for p53. Differences between p53 ‘p53 smart’ model (Vousden, 2000). In addition, DBS have been previously noted. For example, p53 genetic approaches in mice and chromatin immuno- mutants were found to have differential DNA precipitation (ChIP) analysis have recently shown that binding properties and transcriptional activities when p53 requires its family members p73 and p63 to be tested with several p53 DBS (Campomenosi et al., present in order to bind to the promoters of and 2001; Di Como and Prives, 1998; Friedlander et al., transactivate apoptosis genes (Flores et al., 2002). 1996). Several studies suggested that wild-type p53 If a cell’s response to activated p53 is indeed may adopt more than one conformation to utilize determined at the level of promoters, then p53 DBS certain DBS (Thornborrow and Manfredi, 1999; themselves may be important components of the 2001). Lastly, Szak et al. (2001) have recently

Oncogene p53 and p53 DNA binding sites H Qian et al 7904 demonstrated by gel shift assays that the PIG3 DBS (as originally reported by Polyak et al., 1997) is a low- affinity p53 DBS, as compared to the p21,5’ site, and MDM2 DBS. However, the question whether p53 DBS play a broader role in determining a cell’s response to activated p53 has never been stringently and compre- hensively addressed. Therefore, we studied isolated p53 DBS of all known and putative p53 downstream target genes systematically in yeast and mammalian assays with and without DNA damage. Strikingly, two thirds of p53 DBS were unable to direct p53-mediated transcription, and almost all DBS from apoptosis- related genes were included in this set. To the contrary, the DBS of MDM2 and of genes connected to cell cycle arrest, DNA repair and the death receptor pathway of apoptosis were efficiently used by p53. Based on our findings, we propose rules for p53 DBS that successfully predict the behavior of 62 out of 67 DBS in our assays. The results suggest that p53 DBS themselves are an important part of the regulatory mechanisms that guide a cell on the path to either cell cycle arrest or apoptosis.

Results

To determine the ability of p53 to utilize isolated DBS in yeast, we generated URA3 reporter gene plasmids for 67 known and potential p53 DBS representing 57 known and potential p53 downstream target genes (see Tables 1 and 2). The reporter plasmids were identical except for the DBS itself; 5 bp of flanking region were added, if the information was available (see Figure 1a, Tables 1 and 2). p53 transcriptional activity was evaluated through the Ura- of yeast strains at 30 and 378C. Unexpectedly, only 23 of 67 DBS (34%) were utilized by p53 in the yeast assay (Table 2) representing 22 of 57 p53 target genes (39%). Strikingly, all p53 target genes related to cell cycle arrest and DNA repair (Ko and Prives, 1996; Levine, 1997; Vogelstein et al., 2000) had at least one DBS that was functional in yeast (Figure 1b,c; for simplicity, hereafter referred to as cell-cycle-like DBS). The only exceptions were BTG/PC3/Tob and RB. All known p53-dependent genes that are part of the death receptor pathway of apoptosis (Bates and Vousden, 1999; Figure 1 Systematic analysis of p53 DNA binding sites (DBS) in Hengartner, 2000; Rich et al., 2000; Sionov and Haupt, yeast reporter gene assays. (a) Two reporter gene systems were 1999; Vousden, 2000) also had DBS that were used in this study. For yeast reporter constructs, 67 p53 DBS were cloned into the SPO13 promoter upstream of the URA3 re- functional in yeast (Figure 1d; Fas (APO-1/CD95), porter gene. Promoter fragments containing the DBS were then KILLER/DR5 and Pidd). In contrast, all other DBS of moved into a mammalian luciferase reporter construct. PTA de- genes that are part of apoptotic pathways (apoptosis- notes the TATA box of the herpes simplex virus thymidine like DBS) did not function in yeast, except Noxa and promoter. (b,c) All cell-cycle-arrest- and DNA-repair-related p53AIP1 (Figure 1d) (Bates and Vousden, 1999; genes had at least one DBS that was utilized by p53 in the yeast assay (called cell-cycle-like DBS for simplicity), resulting in a Hengartner, 2000; Rich et al., 2000; Sionov and Haupt, Ura+ phenotype, except BTG/PC3/Tob and RB. Three indepen- 1999; Vousden, 2000). Several other genes with varying dent transformants are shown per DBS. (d) DBS of genes that biological activities also had DBS that were utilized by are part of the death receptor pathway also conferred a Ura+ p53 in yeast, including MDM2, the p53 target gene phenotype, while the remainder of DBS of genes connected to apoptosis (apoptosis-like DBS) were not utilized by p53. Results that is part of a negative feedback loop with p53 (see for the MDM2 DBS are shown in (b) through (d) for comparison, Figure 1b – d, and Table 2) (Haupt et al., 1997; since this DBS conferred the highest p53 transcriptional activity Kubbutat et al., 1997; Prives, 1998). in mammalian assays

Oncogene p53 and p53 DNA binding sites H Qian et al 7905 We next determined the ability of transfected p53 to utilize the DBS for cell cycle arrest, DNA repair and apoptosis genes in the p53-negative H1299 cell line. The DBS were cloned into a mammalian luciferase reporter plasmid using the same strategy to ensure that all DBS were tested under identical conditions (Figure 1a). We found the pattern of DBS utilization by p53 to be essentially the same in H1299 cells as in yeast (Figure 2a,b). The two exceptions were the PTGF-b, SBS01 and the Bax DBS (negative in yeast, but activity above background in H1299 cells). Of note, the MDM2 DBS resulted in 13-fold more activity than the p21,5’ site, DBS (used as a reference of 100% throughout the figures). We investigated whether higher p53 protein levels could lead to utilization of apoptosis-like DBS in H1299 cells, but the results were essentially unchanged as compared to Figure 2a,b (Figure 2c). We considered the possibility that transiently over- expressed p53 may be sufficient to utilize cell-cycle-, but not apoptosis-like DBS. We therefore decided to evaluate the utilization of DBS by endogenous wild- type p53 in the absence and presence of DNA damage. We chose A549 cells with endogenous wild-type p53 because they show markedly increased p53 protein levels and an up to fivefold increase in p53 transcrip- tional activity after doxorubicin-induced DNA damage (Wang et al., 2001). In A549 cells without DNA damage, the pattern for cell-cycle-like DBS was similar to H1299 cells. Additionally, all DBS that were positive in yeast led to a 2 – 5-fold increase in p53 transcrip- tional activity after DNA damage. Consistent with the H1299 experiments, MDM2 was the DBS most efficiently utilized by p53 (Figure 3a). In contrast, apoptosis-like DBS, except Noxa and p53AIP1, were unable to direct p53-mediated transcription, with or without DNA damage (Figure 3b). The Bax DBS was also negative, suggesting cell-dependent differences in the utilization of the Bax DBS or that p53 over- expression in H1299 cells caused the observed moderate amount of transcriptional activity (compare Figures 2b to 3b). These results suggested that DNA-damage-induced Figure 2 Systematic analysis of p53 DBS in mammalian reporter post-translational modifications of p53 were unable to gene assays using H1299 cells. Results that were essentially super- compensate for the underutilized apoptosis-like DBS imposable on the p53 yeast assay were obtained in mammalian re- (Figure 6I). In theory, A549 cells may lack certain porter gene assays in p53-negative H1299 cells transiently transfected with 50 ng of the p53 expression plasmid pCMV- pathways upstream of p53, or they may be devoid of p53. (a) Cell-cycle-like DBS conferred p53 transcriptional activity an essential p53 cofactor that needs to interact with similar to the reference DBS p21,5’ site (always shown as 100%). post-translationally modified p53 for the efficient The DBS of MDM2, a gene that is important for a negative feed- utilization of apoptosis-like DBS (Figure 6II). To find back loop with p53, led to markedly enhanced luciferase activity. a condition in which p53 utilizes apoptosis-like DBS, (b) Apoptosis-like DBS were not utilized by p53, while DBS of death-receptor-related genes were. The control for (a) and (b)is we evaluated representative DBS in HepG2, MCF7, the SPO13 promoter without a p53 DBS; all DBS constructs were U2OS (all with endogenous wild-type p53), SW480 also tested in the absence of p53 with results similar to the shown (mutated endogenous p53 plus transfected wild-type control. (c) Increased levels of p53 did not lead to increased uti- p53) and H1299 (p53-negative plus transfected wild- lization of apoptosis-related DBS by p53. Controls are the specific DBS constructs without p53. The expression plasmid pC53-SN3 type p53) cells with and without DNA damaging was used because of higher p53 expression levels than pCMV- treatments known to induce apoptosis in these cell p53. Immunoblotting was performed with anti-p53 antibody lines. None of the tested conditions resulted in the DO-1 after the cell lysates of the shown reporter gene assay utilization of apoptosis-like DBS by p53 (Figure 3c – f, had been adjusted to control for Renilla luciferase activity. The and data not shown). This may have been due to the reduced transcriptional read-out for the p21,5’ site, DBS with 500 ng as compared to 50 ng p53 expression plasmid is likely fact that all tested cell lines have genetic alterations due to increased p53-mediated apoptosis that the assay system that result in the inability of p53 to utilize apoptosis- does not completely control for

Oncogene p53 and p53 DNA binding sites H Qian et al 7906

Figure 3 Systematic analysis of p53 DBS in A549 cells with and without DNA damage and evaluation of apoptosis-like DBS un- der various assay conditions. Experiments were performed as in Figure 2a – c, except that many of the reporter gene assays relied on endogenous p53 in the absence or presence of DNA damage. (a) In A549 cells, cell-cycle-like DBS were again positive (14-3-3s, BDS-2; PTGF-b; GADD45 and PCNA to a lesser extent than in H1299 cells), and all positive DBS showed significantly enhanced p53 transcriptional activity after exposure of the cells to doxorubicin at 0.6 mg/ml for 16 h. The MDM2 DBS conferred markedly enhanced activity compared to the others. (b) Apoptosis-like DBS were again negative and showed no stimulation of p53 transcrip- tional activity after DNA damage. As in yeast and H1299 cells, exceptions were Noxa and p53AIP1, as well as genes of the death receptor pathway. The Bax DBS (approximately 30% activity in p53-negative H1299 cells with transiently transfected, overexpressed p53) was negative in A549 cells, suggesting that the cell context is important for this DBS or that large amounts of p53 can lead to some utilization of the Bax DBS. (c – f) Apoptosis-like DBS were also negative in several other cell lines tested with and without DNA damage (Doxorubicin or g-irradiation). Similar to H1299 cells, the Bax DBS led to transcriptional activity above background in SW480 cells (with mutated p53) that were transfected with p53

Oncogene p53 and p53 DNA binding sites H Qian et al 7907 like DBS. More likely, apoptosis-like DBS alone may not be sufficient to allow essential p53 cofactors to aid p53 (see Figures 4 and 5). Removal of the carboxy-terminal 30 amino acids (residues 364 – 393) of p53 leads to its activation under some conditions (Hupp et al., 1992; Ko and Prives, 1996). With the p21,5’ site, DBS, p53D364 – 393 showed higher activity than wild-type p53. However, improved utilization of PERP, 7218, as well as DBS for Cathepsin D, ei24/PIG8, IGF-BP3, MCG10 and PIG3, was not observed (Figure 4a, and data not shown). We reasoned that cis elements adjacent to p53 DBS that allow cofactors to colocalize and interact with p53 (Figure 6III) might be required for the utilization of apoptosis-like DBS. In yeast, no differences were observed between the isolated DBS and reporter constructs for the p21,5’ site, Bax and PIG3 DBS that contained 100 bp up- and downstream of the p53 DBS itself (Figure 4b). In A549 cells, the p21,5’ site, DBS with addition of promoter fragments showed enhanced p53 transcriptional activity with or without DNA damage (Figure 4c). In contrast, adding 200 bp of native promoter sequence around the Bax and PIG3 DBS did not result in an enhanced usage of these DBS by p53 (Figure 4c). This suggested that cis elements in close proximity to a p53 DBS enhance p53 activity in the case of a cell-cycle- (p21,5’ site), but not apoptosis- like DBS (Bax and PIG3). We then compared full- length promoters for p21, Bax and PIG3 to p53 transcriptional activity with the isolated p21,5’ site, DBS. The presence of additional promoter elements resulted in p53 transcriptional activity that was at least equivalent to the p21,5’ site, DBS for all tested promoters (Figure 4d). ASPP1 and ASPP2 have been characterized as p53 coactivators that specifically enhance p53 transcrip- tional activity at apoptosis-related promoters. The behavior of both ASPP proteins appeared to be very similar (Samuels-Lev et al., 2001). We therefore evaluated whether ASPP2, expressed in A549 cells with endogenous wild-type p53, facilitates p53 tran- scriptional activity with apoptosis-like DBS lacking surrounding promoter sequences. Although ASPP2 enhanced p53 transcriptional activity with cell-cycle- like DBS, it was unable to promote the utilization of apoptosis-like DBS by p53 (Figure 5a). Providing 100 bp of additional cis elements up- and downstream of the p21,5’ site, Bax and PIG3 DBS did not change Figure 4 Search for conditions that allow p53 to utilize apopto- sis-like DBS. (a) Removal of the last 30 amino acids of p53, a po- these findings (Figure 5b). tential negative autoregulatory domain (p53D364 – 393), led to a Mouse knock-out models have recently shown that doubling in luciferase activity with the p21,5’ site, DBS, a rather the p53 family members p73 and p63 are required for small increase, considering the significantly higher protein levels p53-mediated transactivation of apoptosis genes. ChIP for p53D364 – 393. However, it did not lead to p53 activity with the apoptosis-like DBS PERP, 7218, as well as several others assays further established that p63 proteins are present (data not shown). (b) The addition of 100 bp up- and downstream at the promoters of apoptosis-related target genes in of the p53 DBS did not change the yeast phenotypes for p21,5’ order to facilitate p53-mediated transactivation (Flores site, Bax and PIG3.(c) These results were confirmed in mamma- et al., 2002). We tested two of the numerous p73 and lian reporter gene assays in A549 cells. The addition of p21 pro- p63 family members for their ability to assist p53 in moter elements enhanced p53 transcriptional activity, but also abolished the increase in p53 activity after DNA damage that utilizing isolated apoptosis-like DBS. In H1299 cells was seen with the isolated DBS. (d) The evaluation of whole pro- with transiently transfected p53, the Bax DBS results moters for p21, Bax and PIG3 led to p53 transcriptional activity in some p53 transcriptional activity (see Figures 2b and that was at least as high as for the isolated p21,5’ site, DBS

Oncogene p53 and p53 DNA binding sites H Qian et al 7908

Figure 6 ‘Rules of engagement’ for p53 and its DBS. Inactive p53 is unable to utilize DBS, with the likely exception of the MDM2 DBS that showed markedly enhanced activity in our mammalian assays. This fits nicely with the role of the MDM2 gene in a negative feedback loop that maintains relatively inactive p53 at low protein levels in unstressed cells. The MDM2 DBS ac- tually contains two p53 DBS; this in itself, however, does not ex- plain our findings, since the negative c-fos DBS does so as well (see Table 1). Partially or fully activated p53 is sufficient to utilize cell-cycle-like DBS (I). This is illustrated in our mammalian re- porter gene assays where p53 shows transcriptional activity with cell-cycle-like DBS which can be further enhanced by DNA da- mage. Our data is also consistent with the requirement for addi- tional cofactors that bind to post-translationally modified p53 in the context of a DBS alone in order to augment p53 activity (II). In contrast, activated p53 is unable to utilize apoptosis-like DBS alone (I and II). Additional cis elements surrounding apoptosis- like DBS (100 bp up- and downstream) do not change this situa- tion (III), suggesting that the ‘rules of engagement’ for p53 and apoptosis-like DBS are much more complex than for p53 and cell-cycle-like DBS. ASPP proteins have been characterized as p53 coactivators that specifically enhance p53 transcriptional ac- tivity at apoptosis-related promoters. Our results suggest that this coactivator role must require complex regulatory mechanisms and rely on additional cofactors and cis elements. The p73 and p63 fa- milies of p53-related proteins have recently been shown to be re- quired for p53 to activate apoptosis-related target genes. Our data for p73a and p73b suggest that negative apoptosis-like DBS are Figure 5 Evaluation of the p53-interacting protein ASPP2 and not sufficient to allow for p73-mediated enhancement of p53 tran- p73 family members with p53 and isolated p53 DBS. (a) ASPP2 scriptional activity. However, a broader evaluation of apoptosis- enhanced p53 transcriptional activity with isolated cell-cycle-like like DBS and the at least 17 p73 and p63 family members is DBS, in the absence and presence of DNA damage, but was un- needed able to aid p53 in utilizing apoptosis-related DBS. (b) These re- sults remained unchanged when p53 DBS were evaluated with additional 100 bp up- and downstream of the DBS. (c,d) p73a and/or p73b were able to increase p53 transcriptional activity in 3f). In this setting, p73a and p73b were able to enhance the case of the Bax DBS that shows approximately 30% activity p53 transcriptional activity to at least the level of the of the p21,5’ site, DBS in H1299 cells with overexpressed p53. However, the two p73 family members were unable to increase p21,5’ site, DBS (Figure 5c,d). g-irradiation in this p53 transcriptional activity with the IGF-BP3, Box A and the assay system did not lead to enhanced p53 transcrip- PIG3 DBS tional activity (see Figure 3f) suggesting that essential

Oncogene p53 and p53 DNA binding sites H Qian et al 7909 cofactors for p53 may become limiting under the into our class of apoptosis-like DBS as well. However, conditions of p53 overexpression. For two apoptosis- the lack of conformity to the consensus p53 DBS may like DBS that had been negative under all tested be compensated for by the repeat of the pentanucleo- conditions, an increase in p53 transcriptional activity tide sequence of up to 17 times. The report of this was not observed (IGF-BP3, Box A and PIG3, Figure intriguing new class of p53 DBS appeared too late to 5c,d). be incorporated into the experiments of our studies. We attempted to further define the ‘rules of engagement’ between p53 and its DBS through Discussion activation of p53 by DNA damage. For cell-cycle-like DBS, we demonstrated that no additional promoter We have performed a comprehensive and stringent regions were required for enhanced p53 transcriptional study of known and potential p53 DBS that have been activity after DNA damage. This may simply reflect identified in the promoters, introns or exons of 57 increased p53 protein levels, but is also consistent with known and potential p53 downstream target genes. To p53 being activated by post-translational modifications our surprise, a substantial number of reported p53 (Figure 6I) that, in addition, may lead to the DBS (66%) were not utilized by p53 in both yeast and recruitment of p53 cofactors (Figure 6II). In contrast, mammalian assays. This included DBS for almost all we were unable to identify conditions in which p53 genes that are part of the mitochondrial pathway of could utilize apoptosis-like DBS, demonstrating that apoptosis. Equally remarkable was the fact that almost increased p53 protein levels, post-translational modifi- all genes related to cell cycle arrest, DNA repair or the cations of p53 and/or cofactors that function in close death receptor pathway of apoptosis were represented proximity to a p53 DBS (100 bp) are not sufficient to in the collection of positive DBS. drive transcription of apoptosis-related genes (Figure The question arises how seemingly equivalent DBS, 6I – III). Whole promoters for Bax and PIG3 resulted all conforming to the p53 consensus DBS (El-Deiry et in p53 transcriptional activity that was at least as high al., 1992) except for very few base pair changes, can as for the isolated p21,5’ site, DBS. This strongly behave so differently. A closer look, however, reveals suggests that p53 can utilize apoptosis-like DBS only in striking patterns that lead us to postulate the following the context of additional promoter elements that are rules for p53 DBS. Cell-cycle-like DBS that are utilized likely involved in recruiting essential p53 cofactors. by p53 do not have sequences interspersed between the ASPP1 and ASPP2 have been characterized as p53 two half-sites of the p53 DBS (22 of 24) and have two coactivators that are specific for apoptosis-related or fewer mismatches with the p53 consensus DBS (22 genes and that lead to p53-mediated apoptosis, but of 24). Apoptosis-like DBS that are not utilized by p53 not cell cycle arrest (Lane, 2001; Samuels-Lev et al., have interspersed sequences between the two half-sites 2001). Using A549 cells and reporter plasmids contain- of the p53 DBS (29 of 43) and/or three or more ing isolated DBS or DBS flanked by 100 bp of 5’ and mismatches (32 of 43). All p53 DBS, except for GPX, 3’ promoter sequence, we found that ASPP2 stimulated p53AIP1, PAI-1, Pidd and RGC, conform to these p53 transcriptional activity with cell-cycle-like p53 DBS rules (see Tables 1 and 2). (p21,5’ site, and KILLER/DR5), but not apoptosis-like The negative effect of interspersed sequences on p53- DBS (Bax; ei24/PIG8; IGF-BP3, Box A, and PIG3). mediated transcription has been previously recognized DNA damage further enhanced ASPP2 stimulation of (Tokino et al., 1994). Such additional sequences and p53 transcriptional activity with cell-cycle-like DBS, increasing numbers of mismatches with the p53 but did not lead to p53 utilization of apoptosis-like consensus DBS likely result in low affinity DBS, as DBS. These results are opposite to what the model of has been demonstrated by gel shift assays in a Samuels-Lev et al. (2001) would predict and may be comparison of the p21,5’ site, MDM2 and PIG3 different because our study focused on isolated p53 DBS (Szak et al., 2001). These and our results suggest DBS. a hierarchy of p53 DBS: the MDM2 DBS that is used The findings for p53, ASPP2 and isolated p53 DBS by p53 in a relatively inactive state, the collection of lead us to propose two hypotheses. Since ASPP2 has cell-cycle-like DBS that are utilized by a more activated been characterized as a p53 coactivator that is specific pool of p53 and lastly apoptosis-like DBS that have for apoptosis-related genes (Samuels-Lev et al., 2001), additional requirements for usage by p53. its activity must be very tightly regulated. Cells must It should be noted that recently an additional class have a specific inhibitory mechanism to prevent of p53 DBS has been identified by Contente et al. ASPP2-directed p53 stimulation at cell cycle arrest (2002). Detailed studies of the PIG3 promoter showed promoters (since cell-cycle-like DBS alone with ASPP2 that the previously reported PIG3 DBS for p53 (Polyak are sufficient to result in p53 stimulation). Likewise, et al., 1997; used in our study) was not required for cells must have complex mechanisms to specifically p53-mediated transactivation of luciferase reporter enhance p53 activity at apoptosis-related promoters constructs. Rather, p53 depended on and interacted (since apoptosis-like DBS alone with ASPP2 were with repeats of a pentanucleotide motif that repre- unable to do so). Such regulatory mechanisms most sented a microsatellite sequence. The pentanucleotide likely rely on additional, more distant cis elements that repeats have limited homology with the consensus DBS were not addressed in our study. Since ASPP proteins reported by El-Deiry et al. (1992), and thus would fall do not appear to bind DNA directly, these cis elements

Oncogene p53 and p53 DNA binding sites H Qian et al 7910 may recruit regulatory proteins that then interact with Materials and methods the amino-terminus of ASPP proteins, as suggested by Samuels-Lev et al. (2001). Our model system will p53 assays in S cerevisiae facilitate the systematic identification and examination For URA3 reporter plasmids, annealed with of these cis elements that are important for regulating suitable linkers representing p53 DBS plus 5 bp up- and ASPP function. downstream (if the information was available, see Table 1) The recent finding that the p53-related families of p73 were cloned into the PstI–EcoRI sites of a modified pMV252 and p63 proteins are necessary for the appropriate (Vidal et al., 1996) that had the – 368 to – 170 EcoRI transactivation of apoptosis genes by p53 (Flores et al., fragment replaced with a PstI–EcoRI linker (Figure 1a). 2002) has raised an intriguing question. What is the exact After sequence-confirmation, the NotI-SPO13-promoter- URA3-gene-XhoI cassettes were cloned into pRS414 (Sikorski interplay of these three proteins? Flores et al. (2002) and Hieter, 1989). For constructs with 100 bp up- and established by ChIP analysis that p63 did engage with the downstream of the DBS, the promoter regions were amplified promoters of established p53-dependent apoptosis genes from human genomic DNA for p21 or reporter constructs for for successful p53-mediated transactivation. Yet, at Bax and PIG3 (Miyashita and Reed, 1995; Polyak et al., present it is not known whether all involved factors bind 1997) with Pfx (Invitrogen, CA, USA) and primers with to the promoters directly at all times, what the specific linkers that allowed in vivo of the promoter elements are and which p73 and/or p63 family promoter fragment into the PstI–EcoRI gap of the URA3 members assist in the response of p53 to DNA damage. reporter plasmid. Plasmids were then rescued from yeast and This has led to the proposal of two models by Urist and sequence-verified. p53 yeast assays were performed as Prives (2002). In the ‘Dynamic Exchange Model’, p53 described (Brachmann et al., 1996; 1998; Vidal et al., 1996). and family members rely on a single ‘p53 Family Response Element’; the family members would reside p53 assays in mammalian cells in a stabilized complex, but only one of them could be For luciferase reporter plasmids, the SacII – EcoRI fragments bound to the response element at any given time. In the of yeast reporter plasmids were cloned into a modified pp53- ‘Dual Site Stabilization Model’, at least two ‘p53 Family TA-Luc (Clontech Laboratories, CA, USA) that had the Response Elements’ are required to allow the simulta- KpnI–BglII fragment upstream of the minimal promoter neous binding of p53 and p73 and/or p63 family replaced with a SacII – EcoRI linker (Figure 1a). For members, thus resulting in the recruitment of other experiments in H1299 cells, the p53 expression plasmids required cofactors. Our results for three apoptosis-like pCMV-p53 (50 – 100 ng; Clontech Laboratories, CA, USA) DBS suggest that both models might be correct. In the and pC53-SN3 (25 – 500 ng; Baker et al., 1990) were used. To obtain a matching set of expression plasmids for p53 and case of the Bax DBS, fairly low transcriptional activity of p53D364 – 393, the natural open reading frame (ORF) of p53 p53 could be enhanced by p73a and/or p73b. In contrast, in pCMV-p53 was replaced by p53 ORFs with silent nucleic p73a and/or p73b were not sufficient to enhance p53 acid substitutions (T Wang and RK Brachmann, unpublished transcriptional activity with either the PIG3 or IGF-BP3, data) encoding for wild-type p53 or p53 lacking codons 364 – Box A DBS. Additional studies are needed to establish 393 (50 ng each per reporter gene experiment). To induce which of the at least eight p73 and nine p63 family DNA damage, A549, HepG2 and SW480 cells were exposed members (Yang et al., 2002) are involved in p53- to doxorubicin (SIGMA, MO, USA) at 0.6 mg/ml for 16 – mediated activation of apoptosis genes. Our system 24 h prior to lysis, and H1299 and MCF7 cells to 6000 rad of lends itself to dissecting the complexity of the family g-irradiation 7 h prior to lysis. Reporter gene assays were members and is an excellent tool to identify the promoter performed as described (Brachmann et al., 1998), except that a Renilla luciferase construct (Promega, WI, USA) was used elements that are sufficient to result in p53 transcrip- to normalize results. For ASPP2-related experiments, a Cep4- tional activity. based plasmid expressing full-length ASPP2 (500 ng) was The comprehensive and systematic approach of our used. G Melino provided mammalian expression plasmids for study required assay systems that were relatively easy HA-tagged p73a and p73b (50 ng per experiment; de to manipulate. Clearly, the regulation of p53 tran- Laurenzi et al., 1998). The luciferase reporter constructs for scriptional activity in normal cells is much more whole promoters were provided by WS El-Deiry for p21 complex, as it involves full-length promoters and (based on El-Deiry et al., 1993), M Oren for Bax (based on occurs in the context of chromatin. However, the Miyashita and Reed, 1995) and B Vogelstein for PIG3 complexity will only be fully understood when every (Polyak et al., 1997). Standard procedures and the DO-1 single regulatory component has been appropriately antibody (Santa Cruz Biotechnology, CA, USA) were used for anti-p53 immunoblotting. identified and characterized. Our classification of p53 DBS is an important step forward in systematically dissecting the mechanisms that determine cellular outcomes after p53 activation. The clear differences between p53 DBS strongly argue that p53 DBS Acknowledgements themselves are important determinants of a cell’s We thank WS El-Deiry, H McLeod, G Melino, M Oren and B Vogelstein for providing reagents and Nataya decision to undergo either cell cycle arrest or Boonmark for constructing the Cep4-ASPP2 plasmid. This apoptosis. Our rules for p53 DBS will help to assign work was supported in part by grants from the James S p53 target genes to different p53 downstream path- McDonnell Foundation and National Institutes of Health ways and identify and characterize the specific grants CA81511 (RK Brachmann) and CA76316 (L requirements for activating these pathways. Naumouski).

Oncogene p53 and p53 DNA binding sites H Qian et al 7911 References

Baker SJ, Markowitz S, Fearon ER, Willson JK and Ludes-Meyers JH, Subler MA, Shivakumar CV, Munoz RM, Vogelstein B. (1990). Science, 249, 912 – 915. Jiang P, Bigger JE, Brown DR, Deb SP and Deb S. (1996). Bates S and Vousden KH. (1999). Cell. Mol. Life Sci., 55, Mol. Cell. Biol., 16, 6009 – 6019. 28 – 37. Miyashita T and Reed JC. (1995). Cell, 80, 293 – 299. Brachmann RK, Vidal M and Boeke JD. (1996). Proc. Natl. Polyak K, Xia Y, Zweier JL, Kinzler KW and Vogelstein B. Acad. Sci. U S A, 93, 4091 – 4095. (1997). Nature, 389, 300 – 305. Brachmann RK, Yu K, Eby Y, Pavletich NP and Boeke JD. Prives C. (1998). Cell, 95, 5–8. (1998). EMBO J., 17, 1847 – 1859. Prives C and Hall PA. (1999). J. Pathol., 187, 112 – 126. Campomenosi P, Monti P, Aprile A, Abbondandolo A, Rich T, Allen RL and Wyllie AH. (2000). Nature, 407, 777 – Frebourg T, Gold B, Crook T, Inga A, Resnick MA, Iggo 783. R and Fronza G. (2001). Oncogene, 20, 3573 – 3579. Samuels-Lev Y, O’Connor DJ, Bergamaschi D, Trigiante G, Contente A, Dittmer A, Koch MC, Roth J and Dobbelstein Hsieh JK, Zhong S, Campargue I, Naumovski L, Crook T M. (2002). Nat. Genet., 30, 315 – 320. and Lu X. (2001). Mol. Cell., 8, 781 – 794. De Laurenzi V, Costanzo A, Barcaroli D, Terrinoni A, Falco Seol DW, Chen Q, Smith ML and Zarnegar R. (1999). J. M, Annicchiarico-Petruzzelli M, Levrero M and Melino Biol. Chem., 274, 3565 – 3572. G. (1998). J. Exp. Med., 188, 1763 – 1768. Shin TH, Paterson AJ and Kudlow JE. (1995). Mol. Cell. Di Como CJ and Prives C. (1998). Oncogene, 16, 2527 – 2539. Biol., 15, 4694 – 4701. el-Deiry WS, Kern SE, Pietenpol JA, Kinzler KW and Sikorski RS and Hieter P. (1989). Genetics, 122, 19 – 27. Vogelstein B. (1992). Nat. Genet., 1, 45 – 49. Sionov RV and Haupt Y. (1999). Oncogene, 18, 6145 – 6157. el-Deiry WS, Tokino T, Velculescu VE, Levy DB, Parsons R, Szak ST, Mays D and Pietenpol JA. (2001). Mol. Cell. Biol., Trent JM, Lin D, Mercer WE, Kinzler KW and Vogelstein 21, 3375 – 3386. B. (1993). Cell, 75, 817 – 825. Thornborrow EC and Manfredi JJ. (1999). J. Biol. Chem., Elkeles A, Juven-Gershon T, Israeli D, Wilder S, Zalcenstein 274, 33747 – 33756. A and Oren M. (1999). Mol. Cell. Biol., 19, 2594 – 2600. Thornborrow EC and Manfredi JJ. (2001). J. Biol/ Chem., FloresER,TsaiKY,CrowleyD,SenguptaS,YangA, 276, 15598 – 15608. McKeon F and Jacks T. (2002). Nature, 416, 560 – 564. Tokino T, Thiagalingam S, el-Deiry WS, Waldman T, Friedlander P, Legros Y, Soussi T and Prives C. (1996). J. Kinzler KW and Vogelstein B. (1994). Hum. Mol. Genet., Biol. Chem., 271, 25468 – 25478. 3, 1537 – 1542. Funk WD, Pak DT, Karas RH, Wright WE and Shay JW. Urist M and Prives C. (2002). Cancer Cell, 1, 311 – 313. (1992). Mol. Cell. Biol., 12, 2866 – 2871. Vidal M, Brachmann RK, Fattaey A, Harlow E and Boeke Haupt Y, Maya R, Kazaz A and Oren M. (1997). Nature, JD. (1996). Proc.Natl.Acad.Sci.USA,93, 10315 – 10320. 387, 296 – 299. Vogelstein B, Lane D and Levine AJ. (2000). Nature, 408, Hengartner MO. (2000). Nature, 407, 770 – 776. 307 – 310. Hupp TR, Meek DW, Midgley CA and Lane DP. (1992). Vousden KH. (2000). Cell, 103, 691 – 694. Cell, 71, 875 – 886. WangT,KobayashiT,TakimotoR,DenesAE,SnyderEL, Ko LJ and Prives C. (1996). Genes Dev., 10, 1054 – 1072. el-Deiry WS and Brachmann RK. (2001). EMBO J., 20, Kubbutat MH, Jones SN and Vousden KH. (1997). Nature, 6404 – 6413. 387, 299 – 303. Yang A, Kaghad M, Caput D and McKeon F. (2002). Trends Lane D. (2001). Nature, 414, 25 – 26. Genet., 18, 90 – 95. Levine AJ. (1997). Cell, 88, 323 – 331.

Oncogene