The Wilms tumor suppressor-1 target gene podocalyxin is transcriptionally repressed by p53.

Wilms tumors are a heterogeneous class of tumors in which Wilms tumor suppressor-1 (WT1) and the p53 tumor suppressor may be variously inactivated by mutation, reduced in expression, or even overexpressed in the wild-type state. The downstream transcriptional targets of WT1 and p53 that are critical for mediating their roles in Wilms tumorigenesis are not well defined. The WiT49 cell line is characteristic of anaplastic Wilms tumors that are refractory to treatment and expresses wild-type WT1 and mutant p53. We have used the small molecule compound CP-31398 (Pfizer) to restore wild-type p53 function to the codon 248 mutant p53 present in WiT49 cells. In these cells, CP-31398 activated transcription of p53-regulated promoters and enhanced UV light-induced apoptosis without altering the overall p53 protein level. These phenotypes were accompanied by restored binding of the p53 protein to promoter sequences in vivo. Gene expression profiling of CP-31398-treated WiT49 cells revealed subsets of putative p53 target genes that were up- or down-regulated. A preferred target of p53-mediated repression in this system is the podocalyxin (PODXL) gene. PODXL is also transcriptionally regulated by WT1 and has roles in cell adhesion and anti-adhesion. Our results show that PODXL is a bona fide target of p53-mediated transcriptional repression while being positively regulated by WT1. We propose that inappropriate expression of PODXL due to changes in WT1 and/or p53 activity may contribute to Wilms tumorigenesis.

Wilms tumors are a heterogeneous class of tumors in which Wilms tumor suppressor-1 (WT1) and the p53 tumor suppressor may be variously inactivated by mutation, reduced in expression, or even overexpressed in the wild-type state. The downstream transcriptional targets of WT1 and p53 that are critical for mediating their roles in Wilms tumorigenesis are not well defined. The WiT49 cell line is characteristic of anaplastic Wilms tumors that are refractory to treatment and expresses wild-type WT1 and mutant p53. We have used the small molecule compound CP-31398 (Pfizer) to restore wildtype p53 function to the codon 248 mutant p53 present in WiT49 cells. In these cells, CP-31398 activated transcription of p53-regulated promoters and enhanced UV lightinduced apoptosis without altering the overall p53 protein level. These phenotypes were accompanied by restored binding of the p53 protein to promoter sequences in vivo. Gene expression profiling of CP-31398treated WiT49 cells revealed subsets of putative p53 target genes that were up-or down-regulated. A preferred target of p53-mediated repression in this system is the podocalyxin (PODXL) gene. PODXL is also transcriptionally regulated by WT1 and has roles in cell adhesion and anti-adhesion. Our results show that PODXL is a bona fide target of p53-mediated transcriptional repression while being positively regulated by WT1. We propose that inappropriate expression of PODXL due to changes in WT1 and/or p53 activity may contribute to Wilms tumorigenesis.
Wilms tumor is a pediatric kidney cancer that affects 1 in 10,000 children. The Wilms tumor suppressor-1 (WT1) 1 gene was identified as a tumor suppressor gene based upon the presence of WT1 mutations in a subset (ϳ10 -15%) of Wilms tumor samples (reviewed in Ref. 1). However, the majority of Wilms tumor cases cannot be explained by genetic alteration of the WT1 locus. There may be disruption of biological pathways either upstream or downstream of WT1, or there may be pathways independent of WT1 leading to this type of neoplasm. Several lines of evidence suggest that p53 also plays an important role in the development and/or progression of Wilms tumors. The p53 and WT1 tumor suppressor genes encode transcription factors that have been shown to interact physically and to modulate each other's function in some experimental situations (2). p53 can alter the transcription regulatory activity of WT1. In addition, WT1 stabilizes the p53 protein, alters its activity as a transcription factor, and prevents p53-induced apoptosis (2,3). Physical and functional interaction has also been observed between the p53 family members p73 and p63 and WT1 (4). Further evidence for biologically relevant interaction between WT1 and p53 includes the fact that the majority of Wilms tumors develop in the presence of wild-type p53 (5). This is in contrast to adult onset human tumors, in which p53 mutations are frequently observed (ϳ50%). The small subset of Wilms tumors that do harbor p53 mutations develop into much more aggressive anaplastic tumors (between 3 and 7% of the total) that are characterized by unfavorable histology, increased metastasis, chemoresistance, and poor prognosis (6 -9). Moreover, Li-Fraumeni patients carrying germ-line p53 mutations are predisposed to development of Wilms tumors (10), and Wilms tumors are one of a small number of cancers that are specifically associated with such mutations (11). To better understand the involvement of both p53 and WT1 in pathways leading to Wilms tumorigenesis, we have used cDNA microarrays to profile gene expression changes dependent upon the activity of these two factors.
To accomplish this, we developed strategies to alter the function of WT1 and p53 in a Wilms tumor cell line, WiT49. This cell line is characteristic of anaplastic Wilms tumors in that it expresses wild-type WT1 and mutant p53. We have previously shown that WT1 activity can be effectively abrogated in WiT49 cells by expression of a naturally occurring dominant-negative mutant version of the protein (DDS5) (12). To restore wild-type p53 function in WiT49 cells, we made use of the small molecule compound CP-31398 developed by Pfizer. CP-31398 was originally shown to stabilize the p53 protein with mutation at codon 173, 175, 249, or 273 in the wild-type conformation and to allow its transcriptional and tumor suppressor functions (13). More recently, it has been shown that CP-31398 can stabilize the wild-type p53 protein as well (14 -16). We found here that CP-31398 restores activity to the codon 248 mutant p53 present in WiT49 cells without altering p53 protein levels. Furthermore, our data provide a direct demonstration that CP-31398 acts, at least in part, by restoring the DNA binding capability of mutant p53 molecules. Together, the use of DDS5 and CP-31398 make it possible to manipulate the activity of two critical tumor suppressors in a setting that physiologically resembles Wilms tumor and to identify resulting changes in downstream target gene expression.
To identify transcripts modulated by WT1 and/or p53 activity, we prepared a cDNA microarray containing probes for ϳ1446 putative p53 target genes identified in the literature and in preliminary microarray experiments. We postulated that this group of genes would be likely to include many of the genes of interest in terms of regulation of cell growth and tumorigenesis. Hybridization of fluorescently labeled RNA from WiT49 cells grown in the presence or absence of CP-31398 allowed us to identify genes that are regulated by p53 in this system. A small subset of putative p53 target genes was robustly and reproducibly affected by CP-31398 treatment. A notable target for p53-mediated repression in this system is podocalyxin (PODXL), a cell-surface molecule involved in cell adhesion that is also transcriptionally regulated by WT1 (17,18). In this study, we show that PODXL expression is regulated by both p53 and WT1 and suggest that PODXL may be an important downstream mediator of p53 and WT1 activity in Wilms tumorigenesis.

EXPERIMENTAL PROCEDURES
Cell Culture, Drug Treatments, and Transfection-The WiT49 cell line was grown in Dulbecco's modified Eagle's medium/nutrient mixture F-12 (1:1) supplemented with 10% fetal calf serum, 100 units/ml penicillin, and 100 g/ml streptomycin. CP-31398 was provided as a gift from Pfizer. The compound was dissolved in 100% Me 2 SO and stored in the dark at 4°C. HCT116-p53 ϩ/ϩ and HCT116-p53 Ϫ/Ϫ cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 100 units/ml penicillin, and 100 g/ml streptomycin and treated with doxorubicin at 200 ng/ml. The stably transfected N1 and D5 cell lines were derived from WiT49 cells as described (12) and maintained in WiT49 medium containing 0.5 mg/ml G418 (Invitrogen). All transfections into WiT49 and HCT116 cells were done using Lipo-fectAMINE Plus (Invitrogen). The p53-dependent reporter construct pConALuc (19) and PODXL promoter-reporter constructs (18) were cotransfected with pRLTK (Promega). The indicated amount of CP-31398, an equivalent volume of Me 2 SO solvent, or doxorubicin (200 ng/ml) was added 24 h post-transfection, and cells were harvested 19 h later. Firefly and Renilla luciferase activities were determined using the Dual Luciferase activity kit (Promega). Normalization of firefly luciferase activity to either Renilla luciferase activity or total protein concentration (Bradford assay using Bio-Rad reagent) gave similar results.
Semiquantitative and Quantitative Reverse Transcription (RT)-PCRs-Total RNA was prepared using TRIzol (Invitrogen) followed by DNase treatment (DNA-free, Ambion Inc.). RNA was quantitated by spectrometry, and equal amounts were reverse-transcribed to firststrand cDNA using Superscript II (Invitrogen) and random primers. Conventional PCR was performed for detection of p21 and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNAs using a limited number of cycles (n ϭ 27). 2 For absolute quantitation, primers and fluorescently labeled beacon probes for PODXL and GAPDH were designed and manufactured by Gorilla Genomics. Assays were performed on an ABI 7700 thermal cycler exactly as recommended by Gorilla Genomics, except that reaction volumes were reduced to 15 l. Duplicate reactions were performed for each sample, and the average threshold cycle (C t ) at which product began to accumulate was used to determine the starting mRNA target copy number: copy number ϭ e (CtϪb)/m , where m and b are the slope and y intercept, respectively, of a standard curve run in parallel. The standard curve was derived from reactions containing known input amounts of a double-stranded DNA target synthesized by Gorilla Genomics.
TUNEL Staining-WiT49 cells were exposed to various combinations of UV light (10 s) and/or CP-31398 (15 g/ml). Cells were grown for 19 h after treatment and fixed with 1% paraformaldehyde in phosphatebuffered saline followed by 70% EtOH for terminal deoxynucleotidyltransferase-mediated detection of broken DNA ends using the APO-BRDU kit (BD Biosciences). The percentage of apoptotic cells detected by FACScan analysis for each condition is plotted. The results shown are representative of three independent experiments.
Microarray Hybridization-Our custom-designed /p53 target" cDNA microarray contains probes for ϳ1446 putative p53 target genes that were identified through literature searches or in preliminary unpublished profiling experiments. Probe sequences were amplified from clones purchased from Research Genetics as part of a set of 40,000 sequence-verified human clones. The quality of each probe was analyzed by gel electrophoresis. An SDDC-2 DNA Microarrayer robot (Virtek Vision Inc.) was used to spot purified probes in 1.5ϫ SSC onto poly-L-lysine-coated glass slides. Following printing, the slides were UV light-cross-linked (120 mJ); chemically blocked for 10 min at room temperature in 170 mM succinic anhydride and 70 mM sodium borate in 1-methyl-2-pyrrolidinone; denatured for 2 min in 95°distilled H 2 O; and spun dry. The final array contained Ͼ5100 spots, including each probe in duplicate with multiple probes for some genes. In addition to putative p53 target genes, a large number of rRNA and other housekeeping genes were included for quality control and normalization purposes. 80 -100 g of total RNA prepared from cell lines using TRIzol was directly labeled by cDNA synthesis incorporating Cy3-or Cy5-linked dUTP as described (20). Preparation and labeling of RNA from primary Wilms tumor and normal kidney tissue samples involved one round of RNA amplification followed by an indirect aminoallyl labeling procedure. 3 Competitive hybridization of two different RNA populations labeled with fluorescent dyes having different excitation wavelengths (Cy3, 532 nm; and Cy5, 632 nm) was as described (20). For identification of p53 targets in WiT49 cells, RNA from CP-31398-treated cells was hybridized against RNA from Me 2 SO-treated cells. For analysis of gene expression in primary tissues, each patient sample was hybridized against a common control consisting of pooled normal fetal kidney. Hybridization signals were detected and quantitated using an integrated microscanner and the GenePix analysis software package from Axon Instruments, Inc. Comparative analysis of gene expression levels was performed using GeneSpring Version 4.2 software from Silicon Genetics.
Chromatin Immunoprecipitation-Chromatin immunoprecipitation (ChIP) was carried out as described (21). Briefly, DNA-binding proteins were cross-linked to DNA in their native setting by addition of formaldehyde to the medium of growing cells. The cross-linked chromatin was sonicated to an average length of ϳ600 bp and subjected to immunoprecipitation with anti-p53, anti-WT1, and control antibodies. For Fig.  1D, chromatin was prepared from WiT49 cells that were grown in the presence of 15 g/ml CP-31398 or an equivalent volume of Me 2 SO solvent for 16 h. Chromatin was quantitated by spectrometry, and 100 g of chromatin was used in each ChIP reaction. Chromatin was immunoprecipitated with 1 g of rabbit polyclonal antibody (either antihemagglutinin antibody as a nonspecific control (sc-805) or anti-p53 antibody (sc-6243; both from Santa Cruz Biotechnology)). ''Total input'' chromatin samples consisted of 20% of the supernatant recovered from immunoprecipitation reactions containing no antibody. For Fig. 6B, each ChIP reaction contained input chromatin equivalent to ϳ10 7 ge-2 Primer sequences are available upon request. 3 W. Li and B. R. G. Williams, submitted for publication.
nomes from untreated WiT49 cells. 2 g of rabbit polyclonal antibody was used for anti-WT1 (equal mixture of the N terminal (sc-846) and the C terminus (sc-192)), anti-p38 (sc-535), and anti-MEK3 (sc-960; all from Santa Cruz Biotechnology) immunoprecipitation. 2 g of monoclonal antibody DO1 followed by sheep anti-mouse Ig was used for anti-p53 immunoprecipitation. Two sequential rounds of immunoprecipitation were performed in this experiment. For both experiments, antibodybound protein-DNA complexes were collected using protein A-Agarose (16-157C, Upstate Biotechnology, Inc.) and washed extensively. The immunoprecipitated material was purified and assayed by PCR for the enrichment of particular DNA sequences. 10% of the recovered ChIP product was used as the template for PCRs (30 -35 cycles) with p21 or PODXL promoter primers. 2 Products were visualized on ethidium bromide-stained agarose gels.

Restoration of p53 Activity in Wilms Tumor
Cells-The small molecule compound CP-31398 restores wild-type activity to the codon 248 mutant p53 expressed in WiT49 cells as illustrated by both activation of a p53-dependent luciferase reporter construct ( Fig. 1A) and endogenous p21 expression (Fig. 1B). Transcription was activated in a dose-dependent manner in both experiments and presumably resulted from restoration of the proper conformation of the p53 DNA-binding domain rather than from an increase in total p53 levels. This mode of action for the compound was suggested by Foster et al. (13) when it was originally discovered. More recently, several groups of investigators have suggested that CP-31398 restores p53 activity, at least in part, by increasing p53 protein levels (14 -16).
Notably, we observed no change in the total amount of p53 present in WiT49 cells following exposure to CP-31398 ( Fig. 1C).
To directly address the mechanism by which CP-31398 restores p53 transcriptional function in WiT49 cells, we performed ChIP assays with anti-p53 antibodies on chromatin isolated from control or CP-31398-treated cells. As shown in Fig. 1D, p53 occupancy of a defined p53-binding site within the p21 promoter (22) was observed only in WiT49 cells that were treated with CP-31398. This confirms that the codon 248 mutant p53 present in WiT49 cells is defective in DNA binding. Furthermore, the observed induction of p21 transcription by CP-31398 was accompanied by restoration of p53 binding to the p21 promoter in vivo. Thus, these results provide direct in vivo confirmation of the mode of action originally proposed for CP-31398 (13).
In addition to activating p53-dependent transcription, CP-31398 treatment resulted in sensitization of WiT49 cells to UV light-induced apoptosis. Western blot analysis of PARP cleavage indicated that the ability of WiT49 cells to undergo apoptosis in response to UV irradiation was enhanced by coincident treatment with CP-31398, presumably through its restoration of p53 function ( Fig. 2A). In this experiment, staurosporine treatment provided a positive control for PARP cleavage. Staurosporine interferes with both protein kinase C and members of the cyclin/cyclin-dependent kinase machinery to FIG. 1. CP-31398 treatment of WiT49 cells restores wild-type DNA binding and transcriptional regulation by mutant p53 without changing p53 protein levels. A, subconfluent WiT49 cells were transiently transfected with a control Renilla luciferase reporter construct (pRLTK) and a p53-dependent firefly luciferase construct (pConALuc). 24 h later, the indicated concentration of CP-31398 (x axis) was added. Cells were grown for an additional 19 h before being lysed and tested for luciferase activity. Firefly luciferase activity that has been normalized for transfection efficiency using the corresponding Renilla values is shown on the y axis. B, cDNA prepared from WiT49 cells that were treated with the indicated amounts of CP-31398 for 19 h was used as the template for PCR amplification of p21 WAF1 (upper panel) and GAPDH (lower panel) transcripts. The last two lanes contained distilled H 2 O (dH20) and total genomic DNA instead of cDNA template, respectively. The relative amount of p21 signal (normalized to GAPDH) is indicated below each lane for the first four lanes. Signal quantitation was performed using ImageQuant software on scanned images of ethidium bromide-stained gels. C, equal masses of whole cell extracts were separated by SDS-PAGE and analyzed by Western blotting. The membrane was probed with antibody against p53 (upper panel), stripped, and then probed with antibody against ␤-actin (lower panel). Extracts included those from untreated WiT49 cells (lanes 1 and 2) and those exposed to Me 2 SO (DSMO; lanes 3 and 4) and 15 g/ml (lanes 5 and 6) and abrogate the G 2 checkpoint and to induce apoptosis in a p53independent manner (23). Thus, treatment of mutant p53expressing WiT49 cells with staurosporine alone was sufficient to induce apoptosis as indicated by detachment of a fraction of the cells from the plate and PARP cleavage in both adherent and detached cells. In contrast, untreated WiT49 cells, as well as those treated with either UV irradiation or CP-31398 alone, did not contain significant amounts of cleaved PARP protein. However, increased PARP cleavage was apparent in cells that received DNA damage stress via UV light in the presence of CP-31398. TUNEL staining of broken DNA ends followed by FACScan analysis also showed that CP-31398 treatment promoted UV light-induced apoptosis in WiT49 cells in a synergistic manner (Fig. 2B). These results provide further indication that CP-31398 effectively restores wild-type p53 function in the WiT49 cell line. In addition, we observed that CP-31398 treatment led to activation of the AKT kinase (protein kinase B) as indicated by Ser 473 phosphorylation (data not shown). Activation of AKT is indicative of survival pathways being induced (24). This may explain the lack of significant cell death in WiT49 cells treated with CP-31398 alone under the conditions used for these experiments.
Identification of p53-regulated Target Genes in Wilms Tumor Cells-To identify genes that are regulated at the transcriptional level by p53 in WiT49 cells, we generated gene expres-sion profiles for cells treated with CP-31398 in comparison with cells treated with Me 2 SO, the solvent for the compound. Total RNA was prepared from WiT49 cells grown for 19 h in the presence of 15 g/ml CP-31398. Drug-treated and control RNAs were labeled and competitively hybridized against the p53 target microarray in five independent experiments. The vast majority of the ϳ1446 unique genes in the array did not show any change in expression level upon drug treatment when the five experiments were averaged (Fig. 3). This indicates that CP-31398 does not alter gene expression in a nonspecific manner. However, subsets of genes were significantly up-or downregulated (Tables I and II, respectively). Genes were classified as ''p53-regulated'' in this system if they showed an average change in expression between control and CP-31398-treated cells of Ͼ2.5-fold over the five experiments. As illustrated in Tables I and II, most of the p53-regulated genes we identified showed changes of 2.5-fold or more in the majority (if not all) of the experiments. The incorporation of data from five independent experiments makes this a stringent selection. Thus, there may be some important p53-regulated target genes missing from our lists. However, those that are listed are done so with a high degree of confidence.
Interestingly, the genes that were most significantly up-or down-regulated by CP-31398 treatment of WiT49 cells are not the most well known ''classical'' p53 target genes and, in many cases, have not been previously reported. Moreover, CP-31398induced gene expression changes in the WiT49 cell line do not correlate strongly with those reported for two colon cancer cell lines (14). As has been shown for physiologically activated p53, it is likely that CP-31398-restored p53 displays cell-type specificity in the spectrum of genes that it transactivates or represses.
The WT1 Target Gene Podocalyxin Is Negatively Regulated by p53-Of the genes that were significantly and reproducibly repressed by CP-31398 treatment in WiT49 cells, PODXL was particularly interesting to us since it has previously been shown to be a transcriptional target of WT1 (18,25). To confirm the expression data generated by microarray hybridization, we performed quantitative RT-PCR using a fluorescent beacon probe (Fig. 4). This assay proved to be linear over a large range of target concentrations and confirmed that expression of the endogenous PODXL gene was reduced by ϳ80% following restoration of p53 activity by CP-31398.
To further verify that p53 regulates PODXL expression, we utilized luciferase reporter constructs containing portions of the sequence upstream of the endogenous PODXL gene (18). PODXL promoter-luciferase construct A contains 1.65 kb of sequence upstream of the transcriptional start site of PODXL inserted into the promoterless luciferase reporter construct pGL3-basic. PODXL promoter-luciferase construct F contains a 310-bp portion of the promoter that includes two nested WT1binding sites. Both of these constructs have been shown to be transcriptionally activated by isoforms of WT1 lacking three amino acids in their DNA-binding domain (ϪKTS), but not by isoforms containing the alternatively spliced sequence (ϩKTS) (18). We transiently transfected WiT49 cells with PODXL promoter-luciferase constructs A and F or with the pGL3-basic plasmid. 3 h post-transfection, the cells were treated with CP-31398 (15 g/ml) or Me 2 SO solvent for 19 h. Promoter activity was then quantitated by luciferase assay (Fig. 5A). As expected, there was no significant change in expression of the promoterless construct pGL3-basic upon CP-31398 treatment. The minimum WT1-responsive portion of the PODXL promoter (construct F) was also insensitive to CP-31398 treatment. In contrast, the activity of the intact 1.65-kb PODXL promoter (construct A) was repressed by CP-31398. This CP-31398-me- diated repression mirrors what was seen for the endogenous PODXL gene in the microarray and quantitative RT-PCR assays described above, although the degree of repression was somewhat less striking (ϳ60% as compared with Ͼ80%). The difference in response to CP-31398 between constructs A and F indicates that p53 interacts with a portion of the PODXL promoter sequence that is distinct from the WT1-responsive sites identified by Wang et al. (18).
Several reports have indicated that CP-31398 is likely to have p53-independent, as well as p53-dependent, effects on gene expression (14 -16, 26). To confirm that repression of PODXL transcription following CP-31398 treatment is mediated by p53 and is not a nonspecific side effect of the drug, we analyzed the activity of PODXL promoter-luciferase construct A in isogenic human colon carcinoma cell lines that are either wild-type (HCT116-p53 ϩ/ϩ ) or null (HCT116-p53 Ϫ/Ϫ ) for p53. The two cell types were transiently transfected with PODXL promoter-luciferase construct A or the pGL3-basic reporter construct in the presence or absence of doxorubicin, a DNAdamaging agent known to activate p53. As shown in Fig. 5B, in HCT116-p53 Ϫ/Ϫ cells, PODXL promoter-luciferase construct A behaved no differently from pGL3-basic upon treatment with doxorubicin. In contrast, upon doxorubicin treatment of HCT116-p53 ϩ/ϩ cells, the PODXL promoter was specifically repressed by ϳ60% in comparison with pGL3-basic. These data show that the PODXL promoter is negatively regulated by p53 in a system that does not rely upon use of CP-31398.
Interestingly, the HCT116-p53 ϩ/ϩ and HCT116-p53 Ϫ/Ϫ cells responded very differently to treatment with CP-31398. Whereas the drug showed no signs of toxicity in HCT116-p53 Ϫ/Ϫ cells under standard conditions (15 g/ml for 19 h), HCT116-p53 ϩ/ϩ cells were killed by drug treatment (data not shown). Although this toxicity prevented us from analyzing gene expression in the two HCT116 cell lines treated with CP-31398, it indicates that the biological activity of CP-31398 requires p53. Our finding of p53-dependent cell death induced by CP-31398 in HCT116-p53 ϩ/ϩ cells corroborates previously reported data (14,16).
Transcriptional Activation of PODXL by WT1-Wang et al. (18) reported that PODXL is transcriptionally activated by isoforms of WT1 lacking three amino acids (ϪKTS) between zinc fingers 3 and 4 and defined a site within the PODXL promoter sequence that is primarily responsible for the effect. Through analysis of stably transfected derivatives of the WiT49 cell line, we have provided further support for WT1-mediated transactivation of the PODXL gene. We have shown previously that expression of the naturally occurring dominant-negative mutant DDS form of WT1 is an effective way to modulate the activity of WT1 in vivo (12). Here, we used quantitative RT-PCR with fluorescent beacon probes to assess PODXL expres- sion in a control transfectant clone (N1) and in a DDS-expressing clone (D5). Like the parental WiT49 cells, N1 cells expressed the four major isoforms of wild-type WT1 (with or without exon 5 and with or without KTS) at a ratio of ϳ1:1:1:1, and it is expected that DDS WT1 inhibits the activity of all four isoforms. As shown in Fig. 6A, we found that basal (Me 2 SOtreated) PODXL expression was reduced in D5 cells (which express DDS WT1) compared with control N1 cells (which express only wild-type WT1). Although these results illustrate a correlation between endogenous PODXL expression and WT1 activity, it is relatively weak since the difference in PODXL levels between the D5 and N1 clones was only ϳ2-fold. There are several possible explanations for this. First, it is possible that DDS WT1 does not fully abrogate all activities of endogenous wild-type WT1. Second, WT1 may be only one of a number of regulators involved in PODXL expression in vivo. Finally, it is also possible that WT1 is not a very strong inducer of PODXL expression and that the effect of blocking WT1 function is therefore not that striking. This possibility is supported by data showing that expression of the full-length PODXL promoter-reporter construct is activated only 3-fold by ectopically expressed WT1 (18). Nevertheless, our demonstration that inhibition of endogenous WT1 in a Wilms tumor cell line is accompanied by reduced levels of PODXL mRNA indicates that PODXL is likely a physiologically relevant WT1-activated gene.
Treatment of the D5 and N1 stable transfectants of WiT49 cells with CP-31398 showed that both responded similarly to the parental cell line in that PODXL expression was suppressed (Fig. 6A). As expected because of the lower level of basal PODXL expression in WT1-negative D5 cells compared with WT1-positive N1 cells, levels following CP-31398 treat-TABLE I Gene induction following CP-31398 treatment 28 genes were found to be consistently up-regulated in WiT49 cells treated with CP-31398. Five independent cDNA array hybridizations were performed with total RNA extracted from cells treated with 15 g/ml CP-31398 or an equal volume of Me 2 SO solvent for 19 h. Listed below are the minimum, maximum, and average ratios of CP-31398/Me 2 SO hybridization intensities for those genes with an average ratio of 2.5 or greater over the five experiments. The number of independent measurements in which the ratio was Ͼ2.5 out of the total number of analyzable measurements is indicated in the last column. Measurements were not included if the spot was flagged due to poor quality. More than five measurements were available if a clone was spotted more than once on the array or if multiple different clones for a gene were spotted. EST, expressed sequence tags; TNF, tumor necrosis factor.

TABLE II
Gene repression following CP-31398 treatment Six genes were consistently down-regulated in WiT49 cells treated with CP-31398. Data are from the five replicate experiments described in the legend for Table I. Genes were considered down-regulated if the average normalized ratio was 0.4 or less (2.5-fold reduction in gene expression with drug treatment). ment were also somewhat lower in D5 cells than in N1 cells. However, the extent of PODXL repression following CP-31398 treatment was similar in the two cell lines (84.7% for N1 and 88.6% for D5). This result suggests that p53-mediated suppression of PODXL expression can occur in the presence or absence of WT1 and that it can override gene activation by WT1 in some experimental situations.
To determine whether WT1 binds directly to the PODXL promoter in vivo, we utilized ChIP. ChIP carried out using formaldehyde-cross-linked chromatin from untreated WiT49 cells showed that the PODXL promoter sequence was specifically enriched by immunoprecipitation with anti-WT1 antibody compared with immunoprecipitation with irrelevant antibodies against p38 and MEK3, two cytosolic signaling molecules, or with mock immunoprecipitation with no antibody (Fig. 6B). This enrichment was confirmed using semiquantitative RT-PCR. DNA recovered from ChIP reactions was amplified using the same PCR primers specific for the PODXL promoter and SYBR Green Master Mix (PerkinElmer Life Sciences) on an ABI 7700 thermal cycler. The C t at which the PODXL promoter product was detected in the anti-WT1 ChIP reaction was between 4.49 and 5.15 cycles lower than the C t for various control antibody ChIP reactions (data not shown). These significantly lower C t values indicate enrichment of PODXL promoter DNA by anti-WT1 ChIP. Together, these results indicate that the WT1 protein is physically associated with the endogenous FIG. 4. Quantitative RT-PCR confirms that PODXL mRNA levels decrease upon CP-31398 treatment of WiT49 cells. A, shown is a standard curve for Gorilla Genomics beacon quantitative RT-PCR assay for PODXL. PCRs contained between 50 and 5 ϫ 10 6 copies of a synthetic double-stranded DNA target corresponding to the PODXL target sequence. The average threshold cycle at which PCR product accumulated from duplicate reactions is plotted against input target copy number. The equation for the resulting line is used to calculate the number of PODXL mRNA copies present in unknown samples. The R 2 value for the standard curve indicates the efficiency of the PCR amplification. A similar standard curve was generated for the GAPDH beacon assay (data not shown). B, cDNA from WiT49 cells treated with various doses of CP-31398 for 19 h was analyzed using Gorilla Genomics beacon quantitative RT-PCR assays for PODXL and GAPDH. The level of PODXL expression relative to untreated cells is shown, with PODXL expression normalized to GAPDH to control for input cDNA amount.

FIG. 5. The isolated PODXL promoter is repressed by p53 in both WiT49 and HCT116 cells.
A, WiT49 cells were transiently transfected with luciferase reporter constructs lacking a promoter (pGL3-basic; bars 1 and 2) or containing the full-length 1.65-kb PODXL proximal promoter (PODXL promoter-luciferase construct A; bars 3 and 4) or a 310-bp 5Ј-and 3Ј-deleted fragment of the PODXL promoter containing a WT1responsive sequence (PODXL promoter-luciferase construct F; bars 5 and 6). CP-31398 (15 g/ml) or Me 2 SO (DMSO) solvent was added 3 h post-transfection. Luciferase activity was normalized to total protein concentration. B, HCT116-p53 ϩ/ϩ and HCT116-p53 Ϫ/Ϫ cells were transiently transfected with luciferase reporter constructs lacking a promoter (pGL3-basic) or containing the PODXL promoter (PODXL promoter-luciferase construct A). 3 h post-transfection, cells were treated with doxorubicin (doxo; 200 ng/ml) for an additional 18 h to activate p53. Luciferase activity for each condition was normalized to the total amount of protein in the lysate. The graph shows the ratio of PODXL promoter-luciferase construct A activity in the presence of doxorubicin divided by its activity in the absence of doxorubicin, relative to the same ratio for pGL3-basic. PODXL promoter-luciferase construct A was specifically repressed only in HCT116-p53 ϩ/ϩ cells that had been treated with doxorubicin.
PODXL promoter in actively growing WiT49 cells.
We did not detect p53 binding to PODXL promoter sequences in the anti-p53 ChIP assay shown in Fig. 6B (lane 5). This result is expected since the chromatin was prepared from untreated WiT49 cells, which express mutant p53 that is incapable of binding to DNA. However, we also performed ChIP assays on chromatin from WiT49 cells treated with CP-31398 to restore the DNA binding capacity of p53. As illustrated by enhanced occupancy of the p53 recognition element in the p21 promoter (Fig. 1D), this drug treatment did indeed restore DNA binding activity to the codon 248 mutant p53 present in WiT49 cells. However, even following CP-31398 treatment, we were unable to detect p53 association with the PODXL promoter in WiT49 cells by ChIP (data not shown). ChIP was performed with several different monoclonal antibodies and a polyclonal antibody against p53 with chromatin from both CP-31398-treated WiT49 cells and doxorubicin-treated HCT116-p53 ϩ/ϩ cells. In all cases, binding of p53 to the p21 promoter was readily detected, whereas binding to the PODXL promoter was not. One possible explanation for this might be that the PCR readout of the ChIP experiments was not actually assaying the relevant p53-binding sites. The PCR primers used in the ChIP experiments were designed against the promoter region of PODXL (GenBank TM /EBI accession number AF395890, bp Ϫ1374 to Ϫ970) and therefore would be unlikely to detect immunoprecipitated p53-binding sites located either farther upstream or downstream. Use of an algorithm to identify likely p53-binding sites (27) revealed a number of highly significant matches scattered throughout the PODXL gene. The 10 most significant binding sites found within the PODXL genomic sequence (including 10 kb upstream of the transcriptional start site) had scores between 83.43 and 98.22. One of the predicted p53-binding sites is located in the PODXL pro-moter (bp Ϫ1843) just upstream of the most likely WT1-binding site at bp Ϫ1227 (18). The remaining sites are all at least 10 kb farther downstream within the introns and exons of the gene. Since the chromatin used in our ChIP experiments is sonicated to an average length of ϳ600 bp, the possible enrichment of these downstream p53 sites in ChIP products cannot be assayed using PCR primers designed against the PODXL promoter sequence. Moreover, it has been suggested in the literature that p53-mediated repression of transcription might occur through p53 interaction with non-classical binding sites or with other components of the transcriptional or chromatin remodeling machinery (28 -30). Thus, further study will be required to identify whether there are sequences within the PODXL gene that are responsible for p53-mediated repression of PODXL expression.
Expression of PODXL in Primary Wilms Tumors-Given our findings of p53-and WT1-mediated regulation of PODXL expression in cell lines, we were interested in the status of PODXL expression in primary Wilms tumors. Expression of PODXL in primary Wilms tumor and normal adjacent kidney samples was analyzed by cDNA microarray hybridization. Each patient's labeled cDNA was competitively hybridized against a pooled control normal fetal kidney cDNA sample. In general, PODXL expression was significantly reduced in Wilms tumors relative to normal fetal kidney (Fig. 7A). The median and average ratio of PODXL expression in 64 Wilms tumors relative to those in normal fetal kidney were 0.26 and 0.29, respectively. Adjacent normal kidney tissue was available for analysis from six of the Wilms tumor patients; and in four of these cases, expression of PODXL was much higher in the normal kidney tissue than in the tumor tissue (Fig. 7B).  6 and 7). The positive control for recovery of amplifiable DNA from the immunoprecipitation reaction was the supernatant from the no-antibody reaction (Total input; lane 8). The negative and positive controls for PCR amplification were no template (distilled H 2 O (dH20); lane 9) and total genomic DNA (lane 10). MW std, molecular weight standards.
6.1-fold higher in patient 174, and 5.7-fold higher in patient 129. Together, these data suggest that reduced PODXL expression is a characteristic of Wilms tumors.
By simply looking at mRNA expression levels, there was no evidence of a correlation between PODXL and WT1 in primary Wilms tumors (Fig. 7A). Our tumor sample set had a large range of WT1 expression levels (ratios of 0.07-3.03), whereas PODXL expression was much less dynamic (ratios of 0.147-0.875), and the variation that was seen does not seem to be associated with WT1 mRNA levels in any way. Unfortunately, we do not have information regarding WT1 protein levels, isoform ratios, or mutations for this tumor set. A subset of Wilms tumors contains mutations in the WT1 gene that abrogate (or alter) its function, and it is also possible that WT1 mRNA levels do not reflect protein levels. Both of these situations would mask a correlation between WT1 and PODXL. In a similar analysis, we did not see a correlation between PODXL and p53 mRNA levels (data not shown); but as with WT1, it will be important to assess p53 protein levels and DNA sequence to determine whether p53 status plays a role in PODXL expression in primary Wilms tumors. An additional obstacle to meaningful interpretation of expression data from primary tumors is the complex cellular composition of primary tissue samples. Thus, it is possible that differences in gene expression might simply reflect differences in cell types populating various samples rather than a true difference in transcript level per cell.
Microarray data for PODXL expression were also analyzed, focusing on eight anaplastic Wilms tumors and 40 non-anaplastic favorable histology tumors. PODXL expression was lower in both groups of tumors relative to normal fetal kidney. However, there was a statistically significant difference (1.48-fold) in PODXL expression between anaplastic and non-anaplastic tumors (ratio of 0.334 versus 0.226; p ϭ 0.056). Given that p53 mutations in Wilms tumor are nearly always associated with anaplasia, this finding fits with the hypothesis that p53 may normally function to suppress PODXL expression and that loss of this function via mutation may play a role in development or progression of anaplasia.

DISCUSSION
Activity of CP-31398 -The small molecule drug CP-31398 was originally identified by virtue of its ability to stabilize the DNA-binding domain of mutant p53 molecules in the wild-type conformation (13). This study showed that CP-31398 treatment led to activation of p53-regulated transcription and curbed the growth of tumors with p53 mutations in mice; however, the absolute requirement for p53 for the observed effects was not addressed. The promise of CP-31398 as a therapeutic means for restoring p53 activity has been complicated by more recent studies that question the specificity of the compound. Rippin et al. (26) were unable to detect in vitro interaction between CP-31398 and recombinant core domain p53 in a variety of biophysical assays. They also suggested that induction of some downstream target genes in response to CP-31398 might not require p53. However, this conclusion was based upon analysis of just two target genes (one of which did show p53 dependence) in a tetracycline-regulated system that was not shown to tightly control p53 expression. Nevertheless, other reports have also indicated that some effects of CP-31398 are, in part, p53-independent, whereas others appear to require p53 (14 -16, 26). For example, a subset of known p53 target genes was induced at least to some extent upon CP-31398 treatment in HCT116-p53 Ϫ/Ϫ colon carcinoma cells as well as in HCT116-p53 ϩ/ϩ cells (14,16). Notably, p53-independent activation of many classical p53 target genes has been observed in other systems as well, including UV irradiation (31,32). The mechanisms responsible for these parallel, yet overlapping pathways of gene regulation remain unclear. p53-independent ef- fects of CP-31398 were suggested by Rippin et al. (26) to result from intercalation of the compound into DNA and alteration of p53 binding to DNA. Other work suggests that, following exposure to CP-31398, p53 may have altered DNA binding specificity due to a drug-induced conformation that is not entirely equivalent to wild-type different post-translational modifications, altered association with other proteins, or persistent interaction with MDM2 without subsequent ubiquitination (15). The ultimate CP-31398-driven phenotype of cell cycle arrest, apoptosis, or enhanced chemosensitivity following exposure to chemotherapeutic agents does, however, appear to be p53-dependent (14 -16). This suggests that the critical downstream effectors of p53 function are induced by CP-31398stabilized p53 even if the entire spectrum of possible p53 targets is not.
Using a stringent selection, we have identified a small number of putative p53 target genes that are up-or down-regulated upon CP-31398-mediated restoration of p53 function in a mutant p53-expressing Wilms tumor cell line. Given that only a subset of the genes in our cDNA microarray were affected by CP-31398 treatment, it seems unlikely that the drug simply alters transcription in a nonspecific manner by DNA intercalation as suggested by Rippin et al. (26). In addition, our observation of both gene induction and repression argues against a nonspecific mechanism. Finally, as shown in Fig. 5A, CP-31398-mediated repression of a luciferase reporter construct required the presence of specific PODXL promoter sequences. A 1.65-kb portion of the promoter was repressed by CP-31398, whereas several other portions contained therein (ranging from 310 bp to 1.195 kb) were not ( Fig. 5A and data not shown). The failure of many of the putative p53 target genes in the array to respond to CP-31398-stabilized p53 may be due to a number of factors, including 1) another factor required for gene expression is not present in WiT49 cells; 2) CP-31398 does not restore all p53 functions under the conditions used; 3) some genes are not actually regulated by p53 in vivo, at least in this cell type; 4) an additional stress signal is required to activate restored p53; or 5) some probes in the cDNA microarray may not be functional. Together with the findings of other investigators, our study suggests that the gene regulatory function of CP-31398-restored p53 is not identical to that of physiologically activated wild-type p53. However, gene expression is not altered in a nonspecific manner and does correlate with p53-dependent tumor-suppressive phenotypes such as sensitization to apoptosis. Thus, genes showing expression changes following CP-31398 treatment may be important downstream effectors of p53.
Our finding that CP-31398 enhances the ability of mutant p53-expressing WiT49 cells to undergo apoptosis in response to UV irradiation adds to evidence from other groups suggesting that CP-31398 may be an effective chemosensitizing agent. CP-31398 was shown to enhance cell death following treatment with several DNA-damaging agents or with TRAIL in both wild-type and mutant p53 cell lines (14). The compound also promotes apoptosis in response to UVB irradiation of a wildtype p53 melanoma cell line (33). There was no effect, however, on cell death induced by some chemotherapeutic agents in either wild-type or mutant p53 melanoma cells in another study (34). A further complication is added by the fact that HCT116 human colon carcinoma cells expressing the wild-type p53 protein undergo spontaneous apoptosis when treated with CP-31398 alone with no additional stress signal (see ''Results'') (16). Thus, although it remains unclear how CP-31398 functions in promoting apoptosis, there are certainly scenarios in which it does so in a p53-dependent manner (14,16).
Our data showing that steady-state p53 protein levels are not altered by treatment of WiT49 cells with CP-31398 suggest that, although the drug modifies the conformation of the p53 DNA-binding domain, it does not affect the stability of the protein. This mechanism of p53 activation was proposed by Foster et al. (13); however, more recent reports suggest that CP-31398 promotes p53 activity, at least in part, by increasing p53 protein levels. CP-31398 leads to increased steady-state p53 protein levels in cell lines with wild-type p53 as well as some cell lines (but not others) expressing mutant p53 (14 -16, 33, 34). A mechanism for CP-31398-induced stabilization of wild-type p53 has recently been outlined by a study showing that the drug prevents ubiquitination of p53 without altering its association with MDM2 (15). Given the heterogeneous responses of different cell types to CP-31398, it is apparent that multiple mechanisms (perhaps dependent upon the p53 genotype) may play a role in the ability of CP-31398 to modify p53 function. Nevertheless, our demonstration in WiT49 cells that CP-31398 restored DNA binding by mutant p53 in vivo without altering p53 protein levels (Fig. 1, C and D) clarifies at least one mode of action for this compound.
Identification of PODXL as a Target of p53-mediated Transcriptional Repression-The PODXL gene encodes a mucin-like cell-surface sialoglycoprotein that has been shown to have opposite roles in cell adhesion in different tissues (35,36). For example, in the podocytes of the kidney glomerulus, PODXL repels cells from one another (37), whereas in vascular endothelial cells, it plays a pro-adhesive role as a ligand for Lselectin (38). These differences in function are likely due to differences in the glycosylation state of the large extracellular domain of the protein. The intracellular portion of PODXL is linked to the actin cytoskeleton through interactions with ezrin and likely plays a role in cell morphology (39,40). Full-length PODXL is an integral membrane glycoprotein with a single transmembrane span; however, a soluble form of the protein has also been identified (41). The function and possible ligands of soluble PODXL remain unknown. Expression of PODXL in normal tissue has been reported as limited to glomerular podocytes, vascular endothelial cells, platelets, hematopoietic stem cells, and mesothelial cells (38,42,43) as well as to prostate tissue. 4 The anti-adhesive function of PODXL in the kidney glomerulus is essential for maintaining the slit diaphragms between podocyte foot processes through which the urine is filtered (37). PODXL Ϫ/Ϫ mice die soon after birth due to anuric renal failure (44). Reduced expression of wt1 in a mouse model results in glomerulosclerosis reminiscent of that seen in human patients with WT1 mutations (DDS) or deletions (Wilms, Aniridia, Genitourinary syndrome) (25). This is accompanied by down-regulation of PODXL expression, and additional work showed that WT1 directly activates the PODXL promoter (18). Thus, it is clear that WT1-activated expression of PODXL plays an important role in glomerular differentiation and podocyte function. We propose, based on our finding that PODXL is a target of p53-mediated repression, that PODXL may also be involved in Wilms tumorigenesis. Several recent studies linking PODXL to cancer processes provide support for this proposal. Schopperle et al. (41) identified a tumor antigen expressed on malignant stem cells of testicular tumors as PODXL. Since PODXL is not expressed in normal testicular tissue, this finding suggests that inappropriate expression of PODXL may play a role in the etiology or progression of neoplasia. Furthermore, particular variations in the PODXL gene sequence have been linked with prostate cancer aggressiveness. 4 The sequence variations associated with increased aggressiveness alter the ectodomain of the protein and may affect its conformation or glycosylation state. Finally, the fact that PODXL is positively regulated by the pro-angiogenic product of the ETS proto-oncogene also supports the possibility that PODXL may be involved in tumorigenesis (45).
Given the provocative roles that PODXL plays in cell adhesion and anti-adhesion, we hypothesize that inappropriate expression or function of PODXL might result in changes in cell adhesion and migration that drive tumor progression. Inappropriate function of PODXL might be caused by mutation of the PODXL gene itself, as has been proposed for prostate cancer. 4 Alternatively, mutation of a negative regulator of PODXL expression (such as p53) or aberrant expression of a positive regulator (such as WT1) could lead to changes in PODXL expression and function. In light of the latter possibility, it is interesting that WT1, although originally identified as a tumor suppressor gene, may function as an oncogene as well (reviewed in Ref. 46). Tumor-specific overexpression of wild-type WT1 has been observed in acute leukemia, renal cell carcinoma, breast cancer, malignant mesothelioma, and ovarian epithelial tumors. Furthermore, WT1 is mutated only in a small fraction of Wilms tumors, with wild-type WT1 being overexpressed in the remaining majority. Thus, it is possible that WT1 might function as an oncogene in a subset of Wilms tumors. This function might involve transcriptional activation of PODXL by WT1 in a cell type other than the podocyte. It is interesting that WT1 has been implicated as an oncogene in malignant mesothelioma and leukemia, cancers derived from cells that express PODXL. It is difficult to predict what the consequences of altered transcriptional regulation of PODXL might be since it is apparent that cell type-specific post-translational modifications are critical for the function of this protein. Nevertheless, identification of PODXL as a p53-and WT1regulated gene suggests that alterations in this pathway may be involved in Wilms tumors as well as in other tumor types and provides a rationale for further investigation of PODXL function.