A defect in the p53 response pathway induced by de novo purine synthesis inhibition.

p53 is believed to sense cellular ribonucleotide depletion in the absence of DNA strand breaks and to respond by imposition of a p21-dependent G1 cell cycle arrest. We now report that the p53-dependent G1 checkpoint is blocked in human carcinoma cell lines after inhibition of de novo purine synthesis by folate analogs inhibitory to glycinamide ribonucleotide formyltransferase (GART). p53 accumulated in HCT116, MCF7, or A549 carcinoma cells upon GART inhibition, but, surprisingly, transcription of several p53 targets, including p21cip1/waf1, was impaired. The mechanism of this defect was examined. The p53 accumulating in these cells was nuclear but was not phosphorylated at serines 6, 15, and 20, nor was it acetylated at lysines 373 or 382. The DDATHF-stabilized p53 bound to the p21 promoter in vitro and in vivo but did not activate histone acetylation over the p53 binding sites in the p21 promoter that is an integral part of the transcriptional response mediated by the DNA damage pathway. We concluded that the robust initial response of the p53 pathway to GART inhibitors is not transcriptionally propagated to target genes due to a defect in p53 post-translational modifications and a failure to open chromatin structure despite promoter binding of this unmodified p53.

p53 is believed to sense cellular ribonucleotide depletion in the absence of DNA strand breaks and to respond by imposition of a p21-dependent G 1 cell cycle arrest. We now report that the p53-dependent G 1 checkpoint is blocked in human carcinoma cell lines after inhibition of de novo purine synthesis by folate analogs inhibitory to glycinamide ribonucleotide formyltransferase (GART). p53 accumulated in HCT116, MCF7, or A549 carcinoma cells upon GART inhibition, but, surprisingly, transcription of several p53 targets, including p21 cip1/waf1 , was impaired. The mechanism of this defect was examined. The p53 accumulating in these cells was nuclear but was not phosphorylated at serines 6, 15, and 20, nor was it acetylated at lysines 373 or 382. The DDATHF-stabilized p53 bound to the p21 promoter in vitro and in vivo but did not activate histone acetylation over the p53 binding sites in the p21 promoter that is an integral part of the transcriptional response mediated by the DNA damage pathway. We concluded that the robust initial response of the p53 pathway to GART inhibitors is not transcriptionally propagated to target genes due to a defect in p53 post-translational modifications and a failure to open chromatin structure despite promoter binding of this unmodified p53.
The mammalian tumor suppressor protein p53 triggers cytoprotective growth arrests prior to S phase and mitosis in response to genotoxic stresses, induced by ultraviolet (1) or ionizing radiation (2) or treatment with any of several DNAmodifying agents (3). When cellular DNA damage is detected, p53 turnover is prevented, and accumulating p53 activates the transcription of a series of downstream targets, including p21 cip1/waf1 (4), a key inhibitor of cyclin D-cdk 4/cdk 6, cyclin E-cdk 2, and cyclin B-cdk1 kinases (5,6), with subsequent growth arrest in G 1 and/or G 2 phase of the cell cycle.
It is currently thought that either ribonucleotide depletion per se and/or a slowing of the progression of replication forks can also be detected independently of direct DNA damage, resulting in signaling to p53 and initiation of a p53-dependent cell cycle arrest (7,8). Some of the strongest support for this concept comes from the effects of n-phosphonacetyl-L-aspartate (PALA), 1 an inhibitor of an early step of de novo pyrimidine synthesis which causes pyrimidine ribonucleotide depletion but, surprisingly, not DNA strand breaks (8). PALA is an unusual case among the cancer chemotherapeutic agents, for even those drugs which inhibit enzyme targets directly and exclusively usually result in downstream DNA strand breaks.
Another rare cytotoxic compound that does not induce DNA strand breaks is the antimetabolite 5,10-dideazatetrahydrofolate (DDATHF) (9), the prototypical inhibitor of the first folatedependent enzyme in de novo purine synthesis, glycinamide ribonucleotide formyltransferase (GART) (10 -12). DDATHF and its analogs cause a rapid decline in ATP and GTP pools (11,13,14) and, concomitantly, potent inhibition of cell growth (11,(15)(16)(17)(18). Second and third generation GART inhibitors have been developed, and have been or currently are in early phase clinical trials as anticancer agents (19 -25). GART inhibitors are unique in that they are inactive against the classical enzymatic targets for antifolates, i.e. dihydrofolate reductase and thymidylate synthase, but cause a potent inhibition of purine synthesis and consequent cytotoxicity (26) without detectable DNA damage (16,27). Others have reported (16) that GART inhibitors are cytotoxic to carcinoma cells that do not have p53 function but only cytostatic to carcinoma cells that retain wild-type p53 function; this suggestion would be quite compatible with the role of p53 as a direct sensor of nucleotide pools. However, we have found GART inhibitor-induced cytotoxicity to be unaffected by p53 status (28), and, moreover, despite a substantial accumulation of p53 following DDATHF, there was no G 1 arrest, as shown below. This paradox begged explanation.
In this study, we report that human tumor cells respond to de novo purine synthesis blockade by GART inhibitors with p53 stabilization, but that the downstream transcriptional activation events causative of a G 1 arrest are defective. The mechanism of inactivation of the p53 response involves interference with the posttranslational phosphorylation and acetylation of p53 and with histone acetylation at the promoters of p53sensitive downstream targets, even though nuclear translocation and sequence-specific binding of this unmodified p53 are unaffected in vivo. This inactivation of the p53 pathway ex-* This work was supported in part by Grant CA-27605 from the DHHS, National Institutes of Health. 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  plains why p53 wild-type and null function carcinoma cells are equisensitive to the cytotoxicity of the GART inhibitors.
Cell Culture-Human colon carcinoma HCT116 p53ϩ/ϩ and p53Ϫ/Ϫ cells were generously provided by Dr. Bert Vogelstein (Johns Hopkins University), MCF7 (human breast carcinoma) and A549 (human lung carcinoma) cell lines were from the American Type Culture Collection (ATCC). All cell lines were grown in RPMI 1640 (Invitrogen/Life Technologies) supplemented with 10% dialyzed fetal calf serum (dFCS) in a humidified atmosphere of 5% CO 2 at 37°C. Cells were determined free of mycoplasma using a ribosomal RNA PCR kit from ATCC. Irradiation was performed at a dose rate of 4.47 Gy/min using a Mark 137 I-Cesium source.
Flow Cytometry-G 0 /G 1 synchrony was achieved by holding HCT116 p53ϩ/ϩ and p53Ϫ/Ϫ cells at confluence for 48 h, then releasing from postconfluent synchrony by replating to a lower density. For dual parameter flow cytometry experiments, cells were released from G 0 /G 1 synchrony into media containing 10% dFCS, 65 M BrdU (to continuously label cells actively synthesizing DNA) and either PBS, 10 M DDATHF, or 400 M PALA. Forty-eight hours later, cells were harvested by trypsinization (0.1% trypsin/0.04% EDTA), fixed in cold 70% ethanol, and stored at 4°C until flow cytometry. Cells were washed in 0.5% BSA in PBS, and DNA was denatured with 2 N HCl for 20 min at room temperature. The cell suspensions were neutralized by the addition of 0.1 M sodium borate, pH 8.5, and again washed. Cell pellets were incubated for 1 h with mouse monoclonal anti-BrdU FITC-conjugated antibody (BD PharMingen), diluted 1:6 in 0.5% BSA/0.5% Tween 20/ PBS, washed, and resuspended in 50 g/ml propidium iodide (Sigma) and 7 Kunitz units/ml ribonuclease B (Sigma). For single parameter flow cytometry studies, p53ϩ/ϩ cells were released from synchrony into media containing either PBS or 10 M DDATHF. Cells were fixed and stained in 50 g/ml propidium iodide, 0.1% Triton X-100, and 7 Kunitz units/ml RNase B in 3.2 mM sodium citrate buffer. Flow cytometric analysis was performed on an EPICS XL-MCL flow cytometer (Beckman Coulter).
Western Blotting-Cells were lysed in buffer containing 62.5 mM Tris, pH 6.8, 5% glycerol, 2% SDS, 5% 2-mercapothethanol, 50 mM NaF, 0.05 mM Na 3 VO 4 , and 1ϫ protease inhibitor complete mixture (Roche Applied Science). Protein concentrations were determined using a Bio-Rad assay, against a standard of BSA. Total cellular proteins (20 g) were resolved by electrophoresis on 7.5-8.5% SDS-polyacrylamide gels, and were transferred to an Immobilon-P polyvinylidene fluoride membrane (Millipore) or nitrocellulose (Bio-Rad). Primary antibodies were p53 monoclonal Ab-6 (Oncogene) at 1:100, p21 monoclonal Ab-1 (Oncogene) at 1:100, Gadd45 polyclonal (Zymed Laboratories Inc.) at 1:125, Bax polyclonal (Santa Cruz Biotechnology) at 1:2000, and monoclonal antibody active against ␣, ␤, and ␥ isoforms of actin (Oncogene) at 1:4000. Primary monoclonal antibody to Mdm 2 (1:300) was a generous gift from Dr. Swati Deb (Virginia Commonwealth University). Membranes were incubated with secondary antibodies conjugated to horseradish peroxidase (Pierce), and complexes detected by enhanced chemiluminescence using the SuperSignal West Pico Chemiluminescent Substrate Kit (Pierce). Quantification was performed on a Molecular Dynamics SI Densitometer using ImageQuant software. To analyze p53 post-translational modifications, total p53 levels were first quantitated by densitometry of Western blots, and a second experiment was subsequently performed in which loaded lysate or immunoprecipitate volumes were varied to allow equivalent amounts of p53 to be analyzed in each lane. For the phosphorylation-specific antibodies, 100 g or greater lysate (digested with 40 g/ml DNase I for 10 min at room temperature) were loaded, and polyclonal antibodies against p53 peptides with phosphorylated residues were serine 6 (AnaSpec) at 1:500, serine 15 (Cell Signaling) at 1:1000, and serine 20 (Santa Cruz Biotechnology) at 1:200. For the acetylation-specific antibody against lysines 373/382 (Upstate, 1:500 dilution), p53 was first concentrated by immunoprecipitation with the Ab-6 antibody to p53, and then subjected to Western blotting. For analysis of p53 protein stabilization, HCT116 p53ϩ/ϩ cells were either irradiated at 15 Gy or treated with 10 M DDATHF and, after 2.5 h (post-irradiation) or 8 h (DDATHF), cycloheximide was added to a final concentration of 20 g/ml. Cells were collected at various times thereafter, lysed, and subjected to SDS-PAGE and Western blotting as described above.
Northern Blotting-Total RNA was isolated using Trizol (Invitrogen/ Life Technologies) and poly (A) ϩ RNA was selected using an Oligotex mRNA Kit (Qiagen). Poly(A) ϩ RNA was denatured by glyoxal/Me 2 SO treatment at 55°C and 2 g of poly(A) ϩ RNA were separated on a 1% agarose gel. RNA was transferred to ICN Biotrans membrane and probed with random-primed p21 cDNA (ATCC) or mdm2 cDNA (a gift from Dr. Swati Deb, Virginia Commonwealth University) in Gilbert's solution (0.5 M Na 3 PO 4 , pH 7.0, 1 mM EDTA, 1% BSA, 7% SDS] for 16 h at 65°C. The membrane was washed to a stringency of 2ϫ SSC and 0.1% SDS at 55°C. The blot was stripped and reprobed with random primed glyceraldehyde phosphate dehydrogenase (GAPDH) cDNA as described above.
Ribonuclease Protection Assay-Ribonuclease protection assay (RPA) reactions containing 5.8 ϫ 10 5 cpm probe synthesized from the hStress-1 Ribo-Quant Multiprobe Template Set (BD PharMingen), and 20 g of total RNA were hybridized in 80% deionized formamide, 1 mM EDTA, pH 8.0, 400 mM NaCl, and 40 mM PIPES pH 6.7 overnight at 56°C. Digestion with RNase A mixture (100 g/ml ribonuclease A, 10 mM TrisCl, pH 7.4, 1 mM EDTA, 200 mM NaCl, 100 mM LiCl) continued for 30 min at 30°C, and was terminated with the addition of 100 g proteinase K and 0.5% SDS. Protected fragments were ethanol-precipitated, fractionated on a 5% polyacrylamide-urea gel, the gel was dried and exposed to film, and fragments were quantitated by Phosphor-Imager (Molecular Dynamics) analysis.
Immunofluorescence-Following release from G 0 /G 1 synchrony, HCT116 cells were seeded at densities between 8 ϫ 10 4 and 4.5 ϫ 10 5 cells/well on 18 mm glass coverslips in RPMI 1640 ϩ 10% dFCS with or without 10 M DDATHF. Cells were washed, fixed in 100% MeOH at Ϫ20°C for 3 min, rinsed, and stored in PBS at 4°C until samples from all time points were collected. Nonspecific binding was blocked with 5% BSA in Tris-buffered saline containing Tween 20 and sodium azide (TBS-TA), pH 7.5 (0.14 M NaCl, 2.7 mM KCl, 25 mM Tris base, 0.05% Tween 20, 0.05% sodium azide). All incubations were carried out at 37°C for 30 min. Cells were incubated in mouse monoclonal anti-p53 Ab-6 (Oncogene) at a 1:20 dilution in 2% BSA in TBS-TA, and labeled with goat anti-mouse Texas Red antibody (Molecular Probes) at a 1:200 dilution in 2% BSA in TBS-TA. Cell staining was imaged using an Olympus Fastscan 2000 12-bit digital camera on an Olympus IX70 inverted fluorescent microscope (Olympus America) and analyzed using Adobe Photoshop software.
Electrophoretic Mobility Shift Assay-The procedures for nuclear extract preparation and EMSA were adapted from Siliciano et al. (29). For each condition, ϳ5 ϫ 10 6 HCT116 p53ϩ/ϩ cells were washed twice with ice-cold PBS, scraped, pelleted, and lysed on ice for 3-5 min in buffer containing 20 mM HEPES pH 7.6, 25% glycerol, 1.5 mM MgCl 2 , 10 mM NaCl, 0.2 mM EDTA, 2 mM dithiothreitol, 0.1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml pepstatin A, 20 g/ml leupeptin, 10 g/ml aprotinin, 1 mM Na 3 VO 4 , and 1 mM NaF. Nuclei were pelleted, extracted for 45 min at 4°C with end-over-end rocking in 1.5ϫ volumes of the above lysis buffer adjusted to 400 mM NaCl, and cleared by centrifugation. 25 g of nuclear extract or 25 ng of recombinant p53 protein (kindly supplied by Dr. Sumitra Deb, Virginia Commonwealth University) were incubated for 20 min at room temperature in binding buffer containing 100 mM Tris-HCl, pH 7.6, 5% glycerol, 1 mM EDTA, 2 mM MgCl 2 , 2 mM dithiothreitol, 75 mM NaCl, 25 g BSA, and 1 g sonicated salmon sperm DNA. 32 P-end-labeled p21 oligonucleotide (2 ng) containing the 5Ј p53 consensus binding site at the p21 promoter (5Ј-ATC AAT TCT CGA GGA ACA TGT CCC AAC ATG TTG CTC GAG GAT-3Ј) was added, and incubation continued for another 20 min. For supershift analysis, nuclear extract or recombinant p53 was preincubated with 0.5 g of Ab-6 p53 antibody (Oncogene) for 15 min prior to the addition of binding buffer. For competition studies, 20 or 40-fold molar excess cold wild-type p21 or cold mutant p21 (5Ј-ATC AAT TCT CGA GGA AAC GTT CCC AAA CGT TTG CTC GAG GAT-3Ј) oligonucleotide was added simultaneously with the wild-type 32 P-labeled p21 probe. Samples were loaded onto a pre-run 4% non-denaturing polyacrylamide gel and DNA: protein complexes were resolved by electrophoresis in 0.5ϫ Tris borate/EDTA buffer (45 mM Tris, 45 mM boric acid, 1 mM EDTA, pH 8.0). Gels were dried and visualized by autoradiography.
Semi-quantitative and Real Time PCR-Semi-quantitative PCR was performed with 10 -20 ng of input DNA and 0.5 or 1 l of immunoprecipitated DNA in reactions containing 1 mM MgCl 2 , 0.2 mM dNTPs, and 0.6 M primers. For the upstream p53 consensus binding site on the p21 promoter, the primers were 5Ј-CTT TCC ACC TTT CAC CAT TCC-3Ј (sense), 5Ј-AAG GAC AAA ATA GCC ACC AGC-3Ј (antisense), and the product size was 234 bp. Amplification conditions were: 94°C 30 s, 53°C 30 s, 72°C 30 s. For the downstream p53 binding site on the p21 promoter, the primers first described by Barlev et al. (31) were 5Ј-CCA GCC CTT TGG ATG GTT T-3Ј (sense) and 5Ј-GCC TCC TTT CTG TGC CTG A-3Ј (antisense), and the product size was 420 bp. Reactions contained 1 mM MgCl 2 , 0.1 mM dNTPs, and 0.6 M primers, and amplification conditions were: 94°C 1 min, 58.5°C 1 min, 72°C 1 min. PCR cycle number was varied from 25-30 cycles to keep the reaction within the linear range. Quantitative real time PCR was performed using the Quantitect SYBR Green PCR Master Mix (Qiagen). The primers flanking the 5Ј p53 binding site were those described above, whereas the primers flanking the 3Ј p53 binding site were 5Ј-GAG GTC AGC TGC GTT AGA GG-3Ј (sense), and 5Ј-TGC AGA GGA TGG ATT GTT CA-3Ј (antisense), and these primers generated a 154 bp product. Real-time PCR analysis was performed with three-step amplification plus melt curve, and the amplification conditions were: 94°C 15 s, T m 30 s, 72°C 30 s, with T m values of 53°C and 58.5°C for the upstream and downstream p53 binding sites, respectively. ChIP real-time PCR experiments were performed twice, each with triplicate samples.

RESULTS
Inactivation of the p53-dependent G 1 Checkpoint by DDATHF-Normal embryonic skin fibroblasts have been re-ported to arrest in G 1 and G 2 after induction of a nucleotide deficiency with the de novo pyrimidine synthesis inhibitor PALA, whereas fibroblasts in which the p53 pathway had been inactivated by adenoviral E6 expression advance into and accumulate in S phase (8). We found some striking differences when we compared the effects of p53 on cell cycle traverse of carcinoma cells treated with the de novo purine synthesis inhibitor DDATHF and the de novo pyrimidine inhibitor PALA. HCT116 cells with wild-type (ϩ/ϩ) p53 function and isogenic HCT116 cells bearing two null alleles of p53 were synchronized in G 0 /G 1 , then released into medium containing BrdU and either 10 M DDATHF 2 or 400 M PALA. Cell cycle traverse was then monitored via dual parameter flow cytometry over the next 48 h (Fig. 1). The post-confluent G 0 /G 1 synchronization technique induced equivalent levels of synchronization for HCT116 cells that were either p53ϩ/ϩ or Ϫ/Ϫ (Fig. 1A), and both cell types entered the cell cycle with equivalent kinetics (not shown), reestablishing the distribution of cells in all phases of the cycle by 48 h (Fig. 1B). As expected, after 48 h of treatment with DDATHF, 77-80% of HCT116 cells bearing p53 null alleles had bypassed the G 1 checkpoint and accumulated in mid-S phase, apparently due to depletion of purine nucleotide pools (Fig. 1C). Surprisingly, however, HCT116 cells with the p53ϩ/ϩ genotype treated with DDATHF also entered S phase with the same time course. Similarly to published studies on PALA-treated human fibroblasts (8), HCT116 cells with wild-type p53 initiated and maintained G 1 and G 2 arrests after PALA treatment, while HCT116 cells without functional p53 markedly accumulated in S phase (Fig. 1D). Hence, p53 function allowed detection of a nucleotide deficiency induced by de novo pyrimidine synthesis inhibition, but not by purine synthesis inhibition. Interestingly, exit from the G 0 /G 1 synchrony and transit into early S in DDATHF-treated p53Ϫ/Ϫ cells was slowed compared with that seen with PALA-treated cells, an effect not seen in our previous studies with unsynchronized cells (28). Time course experiments in synchronized (Fig. 1) and 2 Previous dose-response curves indicated that this concentration of DDATHF resulted in maximal growth inhibition and cell kill (12,26,28). unsynchronized cells (28) treated with DDATHF demonstrated an equally facile exit from G 2 of HCT116 cells with p53 function or without functional p53 alleles. Thus, the p53 signaling pathway detected the PALA-induced changes in pyrimidine ribonucleotide pools, but the p53-dependent G 1 and G 2 checkpoints appeared to be inactivated following a treatment with DDATHF previously shown (11) to decrease ATP and GTP pools.
A Block between p53 Accumulation and p21 Induction Following DDATHF Treatment-We sought to understand why the p53 response pathway does not initiate a G 1 checkpoint response following DDATHF treatment. In response to GART inhibition by DDATHF in HCT116 cells, p53 accumulated in a time-dependent manner, peaking after 12 h at levels 7-fold higher than in untreated controls ( Fig. 2A). Despite the robust accumulation of p53 seen in DDATHF-treated p53ϩ/ϩ HCT116 cells, p21 levels failed to increase ( Fig. 2A). To compare this with the DNA damage response pathway, when HCT116 cells were treated with VP16, p53 levels rose 8.5-fold. The p53-dependent p21 induction is intact in this cell line (Fig. 2B), for both VP16 and ionizing radiation induced p53 and p21 to similar levels in the p53ϩ/ϩ cells (Fig. 2B). These genotoxic stresses did not induce p21 in the p53Ϫ/Ϫ cells (data not shown).
When a source of preformed purines, such as hypoxanthine or inosine, was added to cell cultures treated with DDATHF, the purine salvage pathway prevented the decline in ATP and GTP pools caused by GART inhibition and rescued the growth inhibitory effects of DDATHF on tumor cells (11). Likewise, addition of 50 M inosine to HCT116 p53ϩ/ϩ cultures prevented the increase in p53 protein levels induced by DDATHF (Fig. 2C). Hence, the effect of DDATHF on p53 levels can be linked to purine nucleotide pool decreases caused by GART inhibition.
To determine whether the unexpected response of HCT116 cells to DDATHF was general for dividing carcinoma cells or was cell-context specific, a human breast carcinoma cell line, MCF7, and a human lung carcinoma cell line, A549, were also treated with DDATHF. These cell lines express wild-type p53 and respond to ionizing irradiation with a strong G 1 block (32); yet, these cell lines also progressed into S phase upon treatment with DDATHF (data not shown). Both MCF7 and A549 cells accumulated substantial levels of p53 following DDATHF treatment but failed to induce p21 as a result (Fig. 2D). We concluded that the block between p53 accumulation and induction of p21 was a general phenomenon. Impairment of Transcriptional Activation or Repression by the p53 Accumulating in GART Inhibitor-treated Cells-We questioned whether the p53 accumulating after DDATHF treatment in these cells was active as a transcriptional activator by examining the steady-state mRNA levels for p21 and several other transcriptional targets of p53. In the HCT116 p53ϩ/ϩ cells, there was a barely detectable increase of p21 mRNA following treatment with DDATHF (Fig. 3A), measurable only by RPA (Fig. 3B and Table I). Likewise, mRNA levels of mdm2, whose transcription is normally regulated by p53 (33), remained constant after GART inhibition when normalized by gapdh levels (Fig. 3A). Average values of mdm2/gapdh in two experiments were 122 and 108% of untreated controls after 12 and 24 h of DDATHF exposure, respectively. In contrast, there was a marked induction of both p21 and mdm2 (9-fold and 6-fold, respectively) in HCT116 cells treated with VP16 as a control (Fig. 3A). Interestingly, a reproducible, but transient increase in Mdm2 protein was observed after 6 and 12 h of GART inhibition, apparently due to post-transcriptional mechanisms that were dependent on p53 (compare Fig. 3, C with A). The role of this increased Mdm2 in limiting the transcriptional activation of the accumulating p53 is unclear (see "Discussion").
RPAs were used to analyze the effect of DDATHF on the transcription of several p53-regulated stress-inducible genes ( Fig. 3B and Table I). The impairment of DDATHF-stabilized p53 to regulate transcription was reflected in several other downstream genes, in addition to p21 and mdm2. The proapoptotic p53-regulated gene bax was induced 3.4-fold upon VP16 treatment, but following GART inhibition, neither bax nor the anti-apoptotic gene mcl1 were induced (Fig. 3B); this agreed with a constant level of Bax protein expression in DDATHF-treated cells (Fig. 3C). However, there was an increase in Gadd45 protein found after both VP16 and DDATHF exposure, but Western blots suggest this increase is not a p53-dependent change, as HCT116 p53Ϫ/Ϫ cells also demonstrated an elevation in Gadd45 protein (Fig. 3C). Quite interestingly, the anti-apoptotic gene bclx, whose promoter has been shown to be repressed by p53 (34) was induced up to 3.6-fold after 24 h of purine synthesis inhibition. These data collectively suggested that the p53 accumulating after GART inhibition is ineffective at activating (p21, mdm2, bax) or repressing (bclx) downstream p53 target genes. p53ϩ/ϩ and p53Ϫ/Ϫ cells exposed to 10 M DDATHF for the indicated time. Cell lysate protein (20 g) was subjected to SDS-PAGE and immunoblotting with antibodies against p53 or p21. The first lane of each panel was loaded with protein from HCT116 p53ϩ/ϩ cells exposed to 10 M etoposide (VP16) for 24 h. Western blots were stripped and reprobed with an actin antibody. B, HCT116 p53ϩ/ϩ cells were treated with either 10 M VP16 for 24 h, or exposed to 12 or 15 Gy irradiation and harvested after 3.5 h, and Western blotting was performed. C, Western blot analysis of p53 protein levels in HCT116 p53ϩ/ϩ cells upon treatment with 10 M DDATHF in the presence or absence of the purine salvage metabolite inosine (50 M). The first two lanes were loaded with lysates from p53ϩ/ϩ cells treated with 10 M VP16, or untreated. D, cell lysates of human breast carcinoma cells (MCF7) and human lung carcinoma cells (A549) were prepared after treatment with 10 M DDATHF for the indicated periods, and Western blot analysis was conducted. Actin levels were equal in D.
Mechanism of Increase in p53 Levels Following DDATHF Treatment-The p53 elevation following DDATHF treatment was not due to enhanced p53 transcription (Fig. 3B). However, upon treatment with DDATHF followed by addition of the protein synthesis inhibitor cycloheximide, p53 protein turned over with a half-life greater than 2 h, roughly equivalent to the stability of p53 accumulating in cells after ionizing radiation (Fig. 4A). p53 in untreated cells decayed with a half-time of  3. The transcriptional activating and repressing functions of p53 are impaired after inhibition of de novo purine synthesis by DDATHF. A, Northern blot analysis of p21 and mdm2 induction following treatment of HCT116 p53ϩ/ϩ cells with 10 M DDATHF. Total RNA was isolated after the indicated times of DDATHF exposure and 2 g poly(A)ϩ RNA were subjected to Northern blotting. The sample in the first lane was 2 g of mRNA from HCT116 p53ϩ/ϩ cells treated with 10 M VP16 for 24 h. The blot was stripped and reprobed for gapdh. B, RPA of p53-regulated genes following exposure of p53ϩ/ϩ cells to DDATHF. Twenty micrograms of total RNA were hybridized to a probe synthesized from a stress-inducible multiprobe template. Shown in the first lane are the migration positions of the unhybridized probe and, in the last lane, yeast tRNA hybridized with probe as a test of specificity of hybridization. For details, see "Experimental Procedures." C, Western blot analysis of p53 downstream transcriptional targets Mdm2, Gadd45, and Bax was performed on lysates of HCT116 p53ϩ/ϩ and p53Ϫ/Ϫ cells exposed to 10 M DDATHF for the indicated times. Stripping and reprobing of the blots with actin was used to assess protein loading. about 10 min (Fig. 4A), as expected from previous studies. Hence, de novo purine synthesis inhibition leads to p53 protein stabilization.
Normal turnover of the p53 protein is modulated by the ubiquitin/proteasomal pathway whereby the activity of a ubiquitin ligase, such as Mdm2, results in progressive addition of several residues of ubiquitin, a 9 kDa protein. Because ubiquitination and subsequent proteasomal degradation of the tagged p53 requires ATP, we reasoned that the limited purine pools might be affecting turnover of p53 in DDATHF-treated cells. The marking of p53 for degradation by ubiquitination was studied in HCT116 p53ϩ/ϩ cells treated with VP16 or DDATHF. As can be seen in Fig. 2A, the levels of p53 protein rose within 6 h and peaked after 12 h of GART inhibition in HCT116 cells; p53 was immunoprecipitated over this time course and p53 ubiquitination was analyzed by Western blotting. Ubiquitinated p53 conjugates were detected after treatment with VP16, as evidenced by a ladder of species ranging from 53 to 90 kDa, that were immunoreactive with a p53 antibody (Fig. 4B). This increase was expected and appears to reflect the ubiquitin ligase activity of Mdm2, based upon previous literature (35). However, when p53ϩ/ϩ HCT116 cells were exposed to DDATHF, p53 ubiquitination was not detectable over this time period, despite the increased steady state levels of p53 and Mdm2 protein. Thus, inhibition of de novo purine synthesis results in the inability to properly ubiquitinate p53, resulting in stabilization of this protein against proteasomal degradation. Subcellular Localization of p53 Following DDATHF Treatment-The ability of p53 to translocate to the nucleus is a prerequisite for transcriptional activation; yet this active process might well be compromised by diminished levels of purine nucleotides. To investigate this possibility, G 0 /G 1 synchronized HCT116 p53ϩ/ϩ cells were released from synchrony into media containing either PBS or 10 M DDATHF, and cells were analyzed by flow cytometry, by Western blotting for p53 and p21 expression, and by immunofluorescence to determine the subcellular localization of p53 (Fig. 5). There was a clear increase in p53 nuclear staining concomitant with the accumulation of cells in S phase and with p53 expression levels, which peaked after 24 h of DDATHF exposure and remained elevated after 48 h (Fig. 5C). This elevation in p53 nuclear staining was similar to that seen in p53ϩ/ϩ cells treated with VP16 for 24 h (Fig. 5D). Nuclear p53 staining was not seen in control p53Ϫ/Ϫ cells, even at high magnification (Fig. 5E). We concluded that, as carcinoma cells reach the G 1 /S border in the presence of DDATHF, p53 becomes stabilized, and there is no impediment to the cytoplasmic-nuclear translocation step in the activation of p53.
Disruption of p53 Signaling to p21 Is Unique for GART Inhibitors Among the Various Classes of Antifolates-Other folate analogs inhibitory to purine synthesis were studied as well as agents inhibitory to other steps in folate metabolism. Both the second generation GART inhibitors LY309887 and AG2034 and the third generation GART inhibitor AG2037 had effects identical to DDATHF, causing robust p53 accumulation, but no p21 induction (Fig. 6A). In contrast, treatment of HCT116 cells with the dihydrofolate reductase inhibitor methotrexate resulted in p53 accumulation and marked p21 transcription, as did the antifolate thymidylate synthase inhibitor D1694 (tomudex, raltitrexed) (Fig. 6B). The induction of p21 was clear within 6 h of methotrexate treatment, and was quite striking after 24 h. Hence, the observed defect in the p53 response pathway is a general and distinctive characteristic of any GART inhibitor, but this response is not shared by other antifolates. This specificity is very interesting, given that methotrexate and all of the other known inhibitors of mammalian dihydrofolate reductase are thought to result in an indirect inhibition of de novo purine synthesis as well as of thymidylate synthesis.
Post-translational Modifications of the p53 That Accumulates After GART Inhibition-Several serine residues in the N-terminal domain of p53 are known to be phosphorylated in response to genotoxic insults such as ionizing radiation and ultraviolet light, and these events are thought to contribute to the stabilization and dissociation of p53 from human Mdm2 (36) and the nuclear translocation of the accumulating p53 (37). Similarly, acetylation of lysines in the C terminus of p53 has been implicated in the binding of nuclear p53 to motifs in p53-sensitive promoters (38,39). In distinct contrast, upon DDATHF treatment, phosphorylation could not be detected using an antibody specific for p53 phosphorylated at serine 15 (compare Fig. 7A with Fig. 2A). Serine 15 was strongly phosphorylated in response to VP16, methotrexate, and D1694, with a time course similar to that of p21 induction (Fig. 7A). When equivalent amounts of p53 were loaded onto a gel from each lysate to allow assessment of the ratio of phosphate groups per p53 molecule, the level of serine 15 phosphorylation of DDATHF-stabilized p53 was very low, in contrast to the strong phosphorylation levels in response to VP16 or ionizing radiation (Fig. 7B). Similarly, serines 6 and 20 of p53 remained hypophosphorylated in response to DDATHF (Fig. 7B).
The acetylation of lysines 373 and 382 of p53 was assessed by immunoprecipitating with a pantropic p53 antibody, followed FIG. 4. DDATHF-induced stabilization of p53. A, Western blot displaying p53 protein levels following inhibition of protein synthesis. HCT116 p53ϩ/ϩ cells were treated with 10 M DDATHF for 8 h, or were exposed to 15 Gy ionizing radiation and incubated for 2.5 h, at which time 20 g/ml cycloheximide was added. Lysates were prepared at the indicated times and Western blotting was performed with the p53 antibody Ab-6. B, in vivo ubiquitination assay showing laddering of ubiquitinated p53 conjugates. HCT116 p53ϩ/ϩ or p53Ϫ/Ϫ cells were either untreated or exposed to 10 M DDATHF for 6 or 12 h, or to 10 M VP16 for 24 h. p53 was immunoprecipitated from whole cell lysates and ubiquitination was assessed by Western blotting with a p53 antibody, Ab-6. The heavy (H) and light (L) chains of the immunoprecipitating antibody are indicated.
by Western blotting. In order to compare the level of acetylation present on p53 before and after drug treatment, p53 levels in immunoprecipitates were first quantitated, and then equivalent amounts of immunoprecipitated p53 were loaded onto a second gel. The gel was subjected to Western blotting with an antibody against acetylated lysines 373 and 382. Acetylation of these residues was not stimulated after DDATHF or VP16 treatment, and, in fact, the number of acetyl groups per p53 molecule dropped substantially below untreated levels by 12 and 24 h of GART inhibition as p53 accumulated, implying that the p53 being synthesized after DDATHF exposure was not being acetylated (Fig. 7C). Thus, although p53 stabilization occurs upon DDATHF treatment, both N-terminal phosphorylation and C-terminal acetylation were severely compromised on the accumulating p53 protein.
In Vitro Binding of DDATHF-stabilized p53 to a p53 Consensus Sequence in the Human p21 Promoter-One of the reported consequences of p53 N-terminal phosphorylation, particularly at serine 15, is enhanced recruitment, by p53, of the histone acetyltransferase CBP/p300, and enhanced acetylation of p53 (40,41). CBP/p300 is capable of transferring acetyl groups to both the p53 C terminus and to the chromatin itself, thereby enabling transcription. Therefore, to determine whether the largely unmodified p53 that accumulates in the nucleus following GART inhibition was able to bind to the p21 promoter, electrophoretic mobility shift assays (EMSA) were employed. Nuclear extracts from p53ϩ/ϩ cells treated with 10 M VP16 exhibit binding to an oligonucleotide containing the upstream p53 binding site of the p21 promoter previously studied by EMSAs (29) (Fig. 8A). Surprisingly, after only 6 h of purine synthesis inhibition by DDATHF, nuclear extracts from p53 wild-type cells were capable of p21 probe binding, and binding continued for 24 h. Furthermore, the specificity of binding to the p21 oligonucleotide was demonstrated upon the addition of excess cold wild-type and mutant oligonucleotides to the binding reactions (Fig. 8B). Twenty-fold excess cold wildtype oligonucleotide interfered with binding of the 32 P-labeled probe, while up to 40-fold excess of a related oligonucleotide with the p53 binding site mutated failed to compete out the wild-type probe (Fig. 8B). The specificity of this binding for nuclear extracts from either VP16 or DDATHF-treated carcinoma cells was confirmed by supershift analysis with the p53 antibody Ab-6 ( Fig. 8B). Hence, the largely unmodified species of p53 that accumulated in DDATHF-treated cells was capable of binding to a p53 binding site of the p21 promoter despite the absence of acetyl groups on the C-terminal domain, in agreement with the conclusions of recent in vitro studies on this promoter (42).
Binding of p53 to the p21 Promoter in Intact Cells Treated with DDATHF-Chromatin immunoprecipitation (ChIP) experiments were performed to study the events of transcriptional activation of the p21 promoter in DDATHF-treated cells FIG. 5. Synchronous entrance of HCT116 cells into S phase correlates with an elevation in p53 protein and an increase in p53 nuclear staining. p53ϩ/ϩ cells were released from a G 0 /G 1 synchrony into media containing either PBS or DDATHF, and cells were collected and analyzed via Western blotting, flow cytometry, or immunofluorescence. A, cell lysates were harvested at the indicated times after release from synchrony, and p53 or p21 protein expression was determined by immunoblotting. In the first lane are lysates from asynchronous p53ϩ/ϩ cells treated with VP16, which demonstrate a positive response to p53 and p21. B, cells were collected and stained with propidium iodide, and DNA content was assessed by flow cytometry. C, cells were fixed and stained with p53 antibody (Ab-6) and imaged at 200ϫ magnification. D, HCT116 p53ϩ/ϩ cells were treated with VP16 for 24 h, fixed, and stained with the p53 antibody Ab-6. E, untreated HCT116 p53Ϫ/Ϫ cells were fixed and stained with p53 antibody, as shown in the left panel, and imaged at 300ϫ magnification. The right panel is a phase contrast image.
in vivo. Two functional p53 binding sites have been identified on the human p21 promoter (4,43). In our experiments, genomic DNA was sheared to fragments predominantly on the distribution range of 200 -1000 bp to allow distinction of p53 binding to the upstream and downstream p53 binding sites located at approximately Ϫ2300 and Ϫ1300 bp, respectively (Fig. 9A). When chromatin from VP16-treated p53ϩ/ϩ cells was immunoprecipitated with a p53 antibody, a strong 234 bp product could be amplified from this immunoprecipitate using PCR primers flanking the 5Ј p53 binding site of the p21 promoter (Fig. 9, A and B), and a second DNA fragment centered on the 3Ј p53 binding site was, likewise, present in the immunoprecipitate (Fig. 9B). The same 5Ј and 3Ј PCR products of similar intensity to those seen in cells bearing DNA damage by VP16 were also seen in the chromatin extracted from DDATHF-treated p53ϩ/ϩ cells, indicating that the p53 accumulating upon GART inhibition is, indeed, capable of binding to the p21 promoter at both the upstream and downstream p53 binding sites in intact cells (Fig. 8B). Virtually identical results were obtained for VP16 (Fig. 9B) or IR-treated cells (data not shown). When real-time PCR was used to quantitate p53 bound to these immunoprecipitated DNAs, the increase of p53 resident at each binding site in the p21 promoter induced by DDATHF treatment was as extensive as that seen with VP16treated cells, with p53 binding at least 4-fold higher than untreated controls (Fig. 9B). Hence, the defect in the p53 pathway responsible for a failure to transcriptionally activate p21 was downstream of nuclear accumulation of p53 and binding to the previously defined p53 sites in the p21 promoter.
Recruitment of the Elements of the Transcriptional Initiation Complex to the p21 Promoter in DDATHF-treated Cells-The residency of the most common universal coactivator of tran-scription, CBP/p300 (44), by p53 bound to the upstream and downstream p53-binding sites was studied by ChIP experiments. Interestingly, this coactivator was recruited to the 5Ј and 3Ј p53 binding sites in HCT116 cells poorly, if at all, by either the fully functional p53 that accumulated in response to the DNA damage or the p53 lacking N-terminal phosphorylation and C-terminal acetylation modifications after DDATHF treatment (Fig. 9C). This result was obtained with each of two antibodies raised against CBP/p300. The recruitment of RNA polymerase II to these regions of the p21 promoter was also analyzed and was found to reflect the binding of p53 (Fig. 9C).   7. Analysis of the post-translational modifications of the p53 that accumulates following GART inhibition indicates lack of serine phosphorylation and lack of lysine acetylation. A, HCT116 p53ϩ/ϩ cells were exposed to either 10 M DDATHF, 10 M VP16, 10 M D1694, or 10 M methotrexate for the indicated times. SDS-PAGE and immunoblotting were performed on 25 g of protein, phosphorylation of p53 was determined by probing with an antibody specific to phosphorylated p53 at serine residue 15, and the membranes were stripped and reprobed for actin. B, cells were treated with 10 M DDATHF (12 h), 10 M VP16 (24 h), or irradiated at 15 Gy and harvested after 3.5 h. A preliminary Western blot was performed to quantitate total p53 levels, and equivalent amounts of p53-containing lysate were subsequently loaded and subjected to immunoblotting with antibodies specific for p53 phosphorylated at serines 6, 15, or 20. C, immunoprecipitation with an antibody to p53 was performed on HCT116 p53ϩ/ϩ cell lysates prior to immunoblotting. p53 protein levels were determined, and equal amounts of p53-containing immunoprecipitate were subjected to Western blotting. Acetylation of C-terminal residues of p53 was examined utilizing an acetylation-specific antibody to lysines 373/382. Blots were stripped and reprobed with an antibody to p53 to confirm p53 loading.
Thus, genomic DNA fragments corresponding to both 5Ј and 3Ј p53 binding sites were enriched in complexes immunoprecipitated by an antibody against RNA polymerase II in cells containing VP16-induced DNA damage or in DDATHF-treated cells to an equivalent extent (2-fold over control cells). Hence, the recruitment of RNA polymerase II by p53 binding was independent of the post-translational modifications of p53 studied.
Acetylation of Chromatin Decorating the p53 Binding Sites of the p21 Promoter-When the acetylation of histone H3 at lysine residues 9 and 14 was examined over the p53 binding sites in the p21 promoter by ChIP, the DNA from VP16-treated cells exhibited strong histone H3 acetylation over the DNA fragments surrounding the 5Ј-and the 3Ј-p53 binding sites, but this was not seen with DDATHF-treated cells (Fig. 9C). Ionizing radiation induced histone H3 acetylation to levels similar to those seen with VP16 treatment (data not shown). Hence, upon GART inhibition, we find p53 nuclear translocation and p21 promoter binding, despite the lack of key post-translational modifications; yet, there remains insufficient chromatin acetylation and remodeling to facilitate transcriptional activation. DISCUSSION We have found a major difference in the response of the p53 pathway to inhibition of de novo purine synthesis compared with the response previously reported for inhibition of de novo pyrimidine synthesis. Upon inhibition of de novo pyrimidine synthesis with PALA, p53 protein accumulates in the absence of measurable DNA damage, an observation that led to the concept that p53 can accumulate as a direct result of nucleotide pool imbalances (8). Our results provide additional support for this concept: following de novo purine synthesis blockade by the GART inhibitors studied, p53 levels also markedly increase in tumor cells (Figs. 2 and 6). However, the responses of the p53 pathway for PALA and DDATHF diverged immediately past the point of p53 accumulation. PALA-initiated p53 accumulation strongly induces p21 (8,45). In contrast, the p53 accumulating after GART inhibition is almost inactive as a transcriptional activator at several downstream p53 target genes, including p21 (Figs. 2 and 3). As a result, PALA-inhibited normal human diploid fibroblasts (8) or human colon carcinoma HCT116 p53ϩ/ϩ cells (Fig. 1) respond to accumulating p53 by the induction of G 1 and G 2 cell cycle blocks; after exposure to DDATHF, p53 wild-type carcinoma cells progress into and accumulate in S phase (Fig. 1).
We believe that GART inhibitor-induced p53 stabilization results, not from DNA strand breaks, but rather, from purine ribonucleotide depletion. Restoration of ATP and GTP pools via the salvage pathway using extracellular inosine prevented elevation of p53 levels (Fig. 2C). Moreover, it is known that DDATHF does not induce detectable DNA strand breaks at early time points when analyzed by filter elution studies (27); only after 3 days of treatment with AG2034 do DNA strand breaks become detectable by the much more sensitive TUNEL assay (16), that is, well after the induction of p53 by the GART inhibitors. Easily detectable DNA strand breaks are induced early in the course of action of antifolates inhibitory to thymidylate synthase (46,47) or dihydrofolate reductase (48,49), as a result of DNA synthesis in the absence of TTP but anomolous levels of dUTP, with subsequent patch excision of dUMP in DNA by DNA-uracil glycosylase. Presumably, stabilization of p53 in these latter cases is a response to DNA damage, although it may have a component that is initiated by depletion of TTP per se.
Although the classic view of p53 stabilization involves stressinduced phosphorylation of p53 which releases Mdm2 binding, this view has been challenged by more recent studies. It has been shown that p53 phosphorylation is not essential for DNA damage-induced stabilization of p53, and that p53 mutants incapable of phosphorylation are still capable of stabilization following UV irradiation or adriamycin (50,51). Our results furnish support for the concept that phosphorylation in the p53 peptide involved in binding Mdm2 is not essential for the accumulation of p53. Another mechanism for p53 accumulation seems operative after GART inhibition: the limited purine pools appear to decrease p53 ubiquitination and proteasomal degradation, allowing a prolonged stability of the protein (Fig.  4). It was initially surprising to note that the level of Mdm2 transiently increases after DDATHF treatment (Fig. 3C), without a measurable increase in mdm2 mRNA (Fig. 3A). However, we note that the decreased p53 ubiquitination after purine nucleotide depletion would most likely also reflect a decreased ubiquitination of Mdm2, explaining the increased stability of this protein. Whatever role this increased Mdm2 plays in limiting the effects of p53 would appear to be minor, based on the facts that nuclear p53 accumulates in DDATHF-treated cells FIG. 8. The p53 accumulating in the nucleus after GART inhibition is capable of binding an oligonucleotide containing the p53 binding element of the p21 promoter. A, EMSA examining the in vitro binding of p53 to its 5Ј consensus element on the p21 promoter. The first three lanes show binding of recombinant p53, while the remaining lanes show binding of p53 extracted from nuclei of DDATHF or VP16-treated HCT116 p53ϩ/ϩ cells exposed to drug as in Fig. 2. The bottom of the film, which had bands indicating free probe and nonspecific binding, is not shown. B, EMSA displaying the specificity of binding to the p21 oligonucleotide. The top gel shows data from VP16-treated nuclei, and the bottom gel, from DDATHF-treated nuclei. Either 20-or 40-fold molar excess of cold wild-type or mutant p21 oligonucleotide were added simultaneously with 32 P labeled wild-type p21 probe. Specific p53 binding is supershifted by the addition of the p53 antibody Ab-6. For details, see "Experimental Procedures." and that the nuclear p53 binds to the p21 promoter equally well as is seen with IR or VP16-treated cells (Figs. 8 and 9).
Our data support a model whereby DDATHF-induced purine ribonucleotide depletion is sensed, either directly or indirectly by p53, which then is stabilized and accumulates (Fig. 10). However, the limiting purine pools render the kinases responsible for p53 phosphorylation inactive, resulting in a hypophosphorylated and hypoacetylated p53 species. It has been suggested that phosphorylation of p53 at serine 15 is a critical event for p53 transcriptional activation (40), and is the primary signal that is then amplified by phosphorylation of residues 6, 9, and 18 (52). These phosphoserines, in turn, promote first acetylation of p53 itself, and then the recruitment of HATs to target promoters and the activation of localized histone acetylation (52). Despite the lack of post-translational modifications of p53 in DDATHF-treated cells, our ChIP experiments ( Fig. 9) provide direct in vivo evidence that this unmodified p53 is capable of nuclear retention and binding to the p21 promoter at the 5Ј and 3Ј p53 binding elements. This is in agreement with the observation by Espinosa and Emerson (42) that unacetylated p53 can bind to the p21 promoter in in vitro assembled chromatin. RNA polymerase II is also recruited to these regions of the p21 promoter, but remains in an inactive pre-initiation complex (Fig. 9). Recently, it has been demonstrated that acetylation-defective p53 mutants did not recruit the transcriptional coactivators CBP and TRRAP and that, as a result, the acetylation of regional histone H3 and H4 was diminished at p53 targets (31). This would suggest that the lack of C-terminal acetylation in the p53 accumulating in DDATHF-inhibited cells is responsible for the inability to recruit HATS to the p21 promoter in our studies. We have not been able to identify the HAT/FAT involved in acetylation of p53 and histone H3 in our HCT116 experiments, but it did not appear to be a likely candidate, CBP/p300 (Fig. 9). Nevertheless, the proper balance of coactivators with histone acetyltransferase activity or of corepressors with histone deacetylase activity was not attained at the p21 promoter in our studies on DDATHF-treated cells, leaving the nucleosomes decorating the region hypoacetylated FIG. 9. DDATHF-stabilized binds the p21 promoter in vivo, but histone acetylation at the p21 promoter is compromised. A, schematic of the p21 promoter, indicating the 5Ј and 3Ј p53 consensus binding motifs. The pentameric p53 binding elements (43) in the p21 promoter are underlined. The arrows depict relative positions of the PCR primers flanking the two p53 binding sites. B and C, in vivo ChIP analysis of the p21 promoter in HCT116 p53ϩ/ϩ and p53Ϫ/Ϫ cells. HCT116 p53ϩ/ϩ cells were treated with 10 M DDATHF for 12 h, 10 M VP16 for 24 h, or p53ϩ/ϩ and p53Ϫ/Ϫ cells were left untreated. Protein-DNA complexes were cross-linked, and the chromatin was sheared and subjected to immunoprecipitation with antibodies to p53 (B), RNA polymerase II, CBP/ p300, or acetyl-lysine histone H3 (C) as described under "Experimental Procedures." DNA was subjected to semi-quantitative and real-time PCR using primers flanking the 5Ј (black bars) or 3Ј (gray bars) p53 binding elements on the p21 promoter. For semi-quantitative analysis, the ratio of immunoprecipitated DNA to input DNA is indicated beneath each gel. For real-time analysis, fold induction or binding of the indicated proteins to the p21 promoter is shown, along with standard deviation. For details, see "Experimental Procedures. "   FIG. 10. Defects in the p53 response following GART inhibition by DDATHF. DDATHF inhibits the enzyme GART, resulting in inhibition of de novo purine synthesis and a reduction in ATP and GTP. These decreases in ribonucleotide pools may be sensed by p53, causing stabilization and accumulation of a p53 protein that is hypophosphorylated at key N-terminal serine residues, such as serines 6, 15, and 20, and hypoacetylated at C-terminal lysines 373/382. Despite its lack of post-translational modifications, this p53 species is capable of nuclear retention and p21 promoter binding, but there is a lack of chromatin acetylation at the p21 promoter, notably of histone H3, and a failure to stimulate transcription. and the promoter transcriptionally inactive.
The response of the p53 pathway to GART inhibitors appears to represent a "non-genotoxic stress" and, hence, invites comparison with those initiated by hypoxia or hydroxyurea treatment. Recent studies have reported that hypoxic stress or cell cycle arrest with hydroxyurea can also result in the accumulation of a p53 protein without transcriptional activation of downstream genes, but the intermediate steps in the mechanisms involved differed from that observed with GART inhibitors. Hypoxia, a classic "non-genotoxic" stress, leads to stabilization of a transcriptionally latent p53 that is not acetylated at the C terminus but is phosphorylated at residues in the Nterminal peptide (53). Hydroxyurea (HU) was reported to induce the accumulation and nuclear localization of a p53 species incapable of target gene induction despite the modification of both N-terminal phosphorylation sites and C-terminal acetylation sites (7). The mechanism of the inactivity of p53 as a transcriptional activator after HU has not yet been solved, but it is very interesting to note that cells treated with HU failed to mount an effective p53 response to a subsequent challenge with IR (7). To our knowledge, the mechanisms whereby p53 accumulates within the nucleus following treatment with other non-genotoxic stimuli such as hypoxia or PALA remain unclear.
We conclude that there are multiple factors needed for a transcriptionally productive p53 response and that binding of p53 to its cognate sites in the promoters of downstream target genes is necessary but not sufficient for transactivation. Although p53 is activated upon de novo purine synthesis inhibition by DDATHF, insufficient recruitment of chromatin-remodeling complexes prevents transcriptional activation of the p21 gene and allows for S phase entrance of tumor cells. This explains why tumor cells with wild-type or null/mutant p53 alleles are equally sensitive to cell growth inhibition and cell kill by DDATHF or AG2034 (28): p53 wild-type cells become functionally p53 null after treatment with GART inhibitors.