E-cadherin is a WT1 target gene.

The WT1 tumor suppressor gene encodes a transcription factor that can activate and repress gene expression. Transcriptional targets relevant for the growth suppression functions of WT1 are poorly understood. We found that mesenchymal NIH 3T3 fibroblasts stably expressing WT1 exhibit growth suppression and features of epithelial differentiation including up-regulation of E-cadherin mRNA. Acute expression of WT1 in NIH 3T3 fibroblasts after retroviral infection induced murine E-cadherin expression. In transient transfection experiments, the human and murine E-cadherin promoters were activated by co-expression of WT1. E-cadherin promoter activity was increased in cells overexpressing WT1 and was blocked by a dominant negative form of WT1. WT1 activated the murine E-cadherin promoter through a conserved GC-rich sequence similar to an EGR-1 binding site as well as through a CAAT box sequence. WT1 produced in vitro or derived from nuclear extracts bound to the WT1-response element within the murine E-cadherin promoter, but not the CAAT box. E-cadherin, a gene important in epithelial differentiation and neoplastic transformation, represents a downstream target gene that links the roles of the WT1 in differentiation and growth control.

The WT1 tumor suppressor gene is associated with Wilms' tumor, which can be considered a disease of disordered development, since the tumors exhibit the three elements of primordial kidney development including blastemal, stromal, and epithelial elements (1). The WT1 gene is predominantly expressed in the developing kidney, but WT1 expression is also detected in the fetal gonad, spleen, mesothelium, breast, and other tissues (reviewed in Ref. 2). During early nephrogenesis, WT1 expression is first detected in the metanephric mesenchyme. At the onset of epithelial differentiation, WT1 expression increases as the mesenchyme surrounding the ureteric bud condenses to form primary vesicles. These cells further differentiate into comma-and S-shaped bodies, and subsequently mature into glomeruli and proximal tubules. This phenomenon and the finding that mice null for WT1 fail to develop kidneys (3) suggest an essential role for WT1 in controlling the mesenchymal-epithelial transition of renal development. WT1 differs from other tumor suppressor genes such as Rb and p53 in that Rb and p53 are widely expressed in a variety of developing and adult tissues, while WT1 expression is restricted. The restricted pattern of WT1 expression suggests that its growthsuppressive properties may be linked to its role in organogenesis.
WT1 gene encodes a Cys-His zinc finger transcription factor, which functions as both a transcriptional repressor and activator (2, 4 -6). The DNA binding activity of WT1 was demonstrated on several different sequences including 5Ј-GCGGGG-GCG-3Ј, which is also recognized by EGR-1, EGR-2, and EGR-3 (4,(7)(8)(9)(10). Alternative splicing of the WT1 gene yields four major isoforms of the WT1 protein. Isoforms C and D include a 3-amino acid (KTS) segment between zinc fingers three and four of the protein, which alters the DNA binding specificity of the protein, while isoforms B and D include a 17-amino acid segment involved in transcriptional regulation. WT1(A) includes neither additional segment (2).
The tumor suppressor activity of WT1 was demonstrated by stable transfection of wild-type WT1 into a Wilms' tumor cell line harboring a mutant WT1 (11), as well as in heterologous osteosarcoma, CV1, and fibroblast cells (12)(13)(14)(15). WT1 can regulate expression of a number of growth and differentiationrelated genes in co-transfection assays including insulin-like growth factor II (16,17), insulin-like growth factor I receptor (18), platelet-derived growth factor A-chain (19,20), Pax-2 (21), syndecan-1 (22), Dax-1 (23), and amphiregulin (24) (reviewed in Refs. 2 and 25). Epidermal growth factor receptor downregulation by WT1 may contribute to growth suppression by inducing apoptosis (12), while syndecan activation (22) and amphiregulin (24) secretion may be important for renal differentiation. A problem in the interpretation of many of these studies has been the lack of a model for the action of WT1 in inducing both differentiation and growth control. We found that expression of WT1 in mesenchymal fibroblasts both suppresses cell growth (13) and induces features of epithelial differentiation (26). These WT1-expressing NIH 3T3 cells exhibit desmosome-like structures and increased expression of several epithelial marker genes such as E-cadherin, type IV collagen, and perlecan (heparin sulfate proteoglycan) along with reduced expression of ␣ 8 integrin and vimentin (26). Furthermore, we recently identified, by representational difference analysis, genes up-regulated in WT1-expressing cell lines which are expressed in the condensing metanephric mesenchyme of the developing mouse kidney. 1 This lends support to our hypothesis that expression of WT1 in this heterologous system mimics some of the effects of WT1 in kidney differentiation. Similarly, work by Haber's group identified amphiregulin as a gene induced by WT1 in an osteosarcoma cell line but also expressed in the condensing, developing human kidney (24). In this study we have focused on E-cadherin as a WT1 target gene, due to its close association with epithelial differentiation. E-cadherin is a calcium-dependent cell adhesion protein localized at the adherens junction that mediates cell-cell interactions. Expression of E-cadherin serves as a marker for epithelial differentiation (27,28). The intracellular portion of E-cadherin interacts with submembrane cytoskeletal proteins including the ␣and ␤-catenins (27)(28)(29)(30)(31). Loss of the E-cadherin expression is frequently seen in human tumors (28) and has been associated with loss of epithelial cell morphology and tumor invasion (32)(33)(34). The expression of E-cadherin and other cell adhesion proteins also play a key role in normal development during the mesenchymal-epithelial transition (35,36). E-cadherin appears in the induced metanephric mesenchyme once condensation into epithelia begins (37,38), the same time at which WT1 expression rises in the developing kidney (39,40).
In this report, we provide evidence that the WT1 protein can directly trans-activate E-cadherin gene expression. Both transient and stable expression of WT1 induced the endogenous E-cadherin gene. WT1 protein bound to a proximal GC-rich sequence in the E-cadherin gene promoter. This cis-acting element alone was sufficient for trans-activation of the E-cadherin promoter by WT1 in co-transfection experiments. A dominant negative mutant of WT1, WTAR inhibited the E-cadherin promoter in WT1 expressing NIH 3T3 cells. E-cadherin represents a downstream target gene that may link the functions of WT1 in epithelial differentiation and tumor suppression.

EXPERIMENTAL PROCEDURES
Cell Lines-The NIH 3T3 cells stably expressing WT1 (WR16 and WR35) and the Ras-transformed NIH 3T3 cells stably expressing WT1 (VW9) that were used in this study have been described previously (13,26). Ecotropic PHOENIX cells (gift of W. Pear, University of Pennsylvania, Philadelphia, PA) used to generate retroviral stocks were grown in Dulbecco's modified Eagle's medium with 10% fetal bovine serum.
Plasmid Construction-Expression vectors for wild-type WT1 isoforms (A, B, C, D) and the WT1 (AR) dominant negative mutant were described previously (13,(41)(42)(43). The promoter regions of the murine E-cadherin gene were amplified from murine genomic DNA by PCR 2 using various forward primers containing a BglII adaptor sequence to generate deletions or point mutations. Mutated bases are indicated in bold type. A reverse primer containing a HindIII adaptor sequence at ϩ113 of the murine E-cadherin promoter sequence was used (Sequences 1-17). The human E-cadherin promoter luciferase construct HECad (Ϫ368/ϩ125) was described previously (44). For the construction of HECad (Ϫ217/ϩ125) and HECad (DEL), E-cadherin promoter sequences were amplified by PCR and subcloned into SacI and HindIII of pGL2-Basic (Promega, Madison, WI). A bicistronic retroviral vector for WT1 was constructed by insertion of a Sau3A fragment of WT1(A) into the BglII site of the MIGR1 retroviral vector (gift of W. Pear), which contains the murine stem cell leukemia virus long terminal repeat.
A single reverse primer (R) containing a BglII adaptor was paired with each forward primer (F).
The polymerase chain reaction was performed at a denaturing temperature of 94°C for 30 s, annealing temperature of 60°C for 30 s, and extension for 1 min at 72°C for 40 cycles. After amplification, the PCR fragments were digested with BglII and HindIII and cloned into the pGL3-Basic luciferase reporter construct (Promega), and subjected to automated DNA sequencing.
Transfection Assay-For cotransfection assays involving the murine and human E-cadherin luciferase reporter and WT1 isoforms (A, B, C, D), NIH 3T3 cells and 293T cells were grown in Dulbecco's modified Eagle's medium containing 10% fetal calf serum and penicillin/streptomycin. At 18 h prior to transfection, NIH 3T3 cells were plated at a density of 3 ϫ 10 4 in 24-well dishes. Optimal quantities of effector (RSV-WT1(A, B, C, D), WTAR) and reporter (E-cadherin-Luc) used in transfection assays were empirically determined to be 0.5 g of effector (RSV-WT1(A, B, C, D) or RSV alone), 0.1 g of reporter, and 0.01 g of a thymidine kinase Renilla reporter gene, which serves as an internal control for transfection efficiency. WT1 modestly activated the thymidine kinase Renilla reporter, leading to an underestimate of activation by WT1 but did not alter the pattern of activation noted for E-cadherin promoter deletion constructs. The plasmids were transfected using the SuperFect transfection reagent (Qiagen, Valencia, CA) according to the manufacturer's directions. The NIH 3T3 cells were harvested 60 h after transfection and assayed for luciferase activity using the Dual-Reporter assay system (Promega). All results are shown as the average Ϯ standard deviation of three to six independent experiments. For preparation of nuclear extracts, 293T cells were plated at the density of 5 ϫ 10 6 cells/10-cm dish 18 h prior to transfection and 5 g of RSV-WT1(A) was used for transfection. The 293T cells were harvested 48 h after transfection, and nuclear extract was prepared as described below.
Generation of Retrovirus-A bicistronic vector harboring the green fluorescent protein and WT1(A) cDNAs was transiently transfected using Superfect (Qiagen) into PHOENIX cells. At 48 h after transfection, expression of the vector was monitored by phase contrast fluorescence microscopy (Nikon, Tokyo, Japan) and the supernatant of the packaging cells was collected, filtered, and layered in the presence of 5 g/ml Polybrene (Sigma) on a fresh culture of 60% confluent NIH 3T3 cells. At 24, 48, and 72 h after infection, the cells were monitored for green fluorescence and lysed for the preparation of total RNA using Trizol (Life Technologies, Inc.).
Analysis of mRNA Expression-Cytoplasmic RNAs were isolated 2 The abbreviations used are: PCR, polymerase chain reaction; GFP, green fluorescent protein; RSV, Rous sarcoma virus; GST, glutathione S-transferase; EMSA, electrophoretic mobility shift assay; DTT, dithiothreitol; PBS, phosphate-buffered saline. 5Ј-TAT AAG CTT CGC CGA GCA AAC ACT GAG CT-3Ј SEQUENCES 1-17 from cell lines as described previously (13,26). Cytoplasmic RNAs from each cell line (20 g) were subjected to Northern blot analysis for E-cadherin expression (45) and quantified by PhosphorImager analysis using ImageQuant software (Molecular Dynamics, Sunnyvale, CA). The E-cadherin probe was a gift from S. Mazur (Mount Sinai School of Medicine, New York, NY). E-cadherin expression in retrovirus-infected NIH3T3 cells was detected by reverse transcriptase-PCR. The reverse transcriptase reaction was performed with 1 g of total RNA using Superscript II, (Life Technologies, Inc.). The cDNA was then subjected to 35 rounds of PCR using E-cadherin-specific primers (E-cadherin forward PCR primer, 5Ј-TCA ACG ATC CTG ACC AGC AGT TCG-3Ј; reverse transcriptase primer, 5Ј-GGT GAA CCA TCA TCT GTG GCG ATG-3Ј; reverse nested PCR primer, 5Ј-ATC TTC TGA TCC ATG ACC GTG TCC G-3Ј). Polymerase chain reaction was performed at a denaturing temperature of 94°C for 30 s, annealing temperature of 60°C for 30 s, and extension for 1 min at 72°C for 35 cycles. An internal control for RNA integrity was simultaneously performed using mouse ␤-actinspecific primers (forward, GCC CAG AGC AAG AGA GGT AT; reverse, GGC CAT CTC TTG CTC GAA GT). The PCR products were visualized by agarose gel electrophoresis and ethidium bromide staining.
Electrophoretic Mobility Shift Assay (EMSA)-Double-stranded oligonucleotide probes were created by annealing complementary singlestranded oligonucleotides (Life Technologies, Inc.). The oligonucleotide sequences used were as follows. The putative WT1 binding site is underlined, and mutated position(s) are indicated in bold type.
The HindIII 5Ј overhanging ends of the duplex oligonucleotides were labeled with [␣-32 P]dCTP and the Klenow fragment of DNA polymerase I and gel-purified. In vitro coupled transcription/translations were performed with pSP64-WT1(A) (41) or an insertless pSP64 control vector using rabbit reticulocyte lysate (Promega). Nuclear extract from WT1(A)-expressing cell line VW9 (13), WT1(A)-transfected 293T, and a NIH 3T3 control cell line were extracted using a two-step method as follows. 5-10 ϫ 10 6 cells were grown to 80% confluence and resuspended in 10 mM HEPES (pH 7.9), 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM DTT. The cells were allowed to swell on ice for 10 min, vortexed, and briefly pelleted in a microcentrifuge at 4°C. The pellet was resuspended in 20 mM HEPES (pH 7.9), 25% glycerol, 420 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, and 0.5 mM DTT and incubated on ice for 20 min. The nuclear extract was then centrifuged for 2 min in a microcentrifuge at 4°C, and the supernatant was collected. Protein concentrations were determined using the Bio-Rad DC kit. To compare the binding activity of the alternative ϩKTS and ϪKTS forms of WT1 to the E-cadherin promoter WT1-response element, sequences encoding the zinc-finger domains of WT(A/ϪKTS), WT(C/ϩKTS), and WT(AR) (negative binding control) were cloned into the pGEX2TK GST-expression vector (Amersham Pharmacia Biotech), expressed in Escherichia coli strain BL21 and purified on glutathione beads (1).
DNA binding reactions using in vitro transcribed/translated WT1 were performed with 2.5 l of lysate programmed with pSP64-WT1(A) or the pSP64 control as described previously (46), except that the total volume of the reaction was 20 l and the amount of probe used was 0.2 ng (approximately 10 6 cpm/ng). Binding reactions utilizing nuclear extracts were performed by preincubating 3 g of extract in 20 mM HEPES (pH 7.5), 12% glycerol, 1 mg/ml bovine serum albumin, and 0.5 mM DTT with 2 g of d(I-C) and 1 g of DNase-free RNase A for 15 min at room temperature. Next, 0.2 ng (approximately 10 6 cpm/ng) of radiolabeled probe was added and the mixture was incubated for an additional 20 min at room temperature. Binding reactions utilizing GST-WT1 protein were performed by preincubating 200 ng of protein in 50 mM HEPES (pH 7.4), 50 mM KCl, 50 M ZnCl 2 , 5 mM MgCl 2 , 1 mM DTT, and 20% glycerol, with 1 g of d(I-C) for 10 min at room temperature. After this, 0.2 ng (approximately 10 6 cpm/ng) of radiolabeled probe was added and the mixture was incubated for 30 min at 4°C. The DNA-protein complexes were separated through a 5% non-denaturing polyacrylamide gel for 1.5 h at 300 V at room temperature. For Western blot analysis following the mobility shift assay, the assay was performed with an unlabeled oligonucleotide probe. The polyacrylamide gel was transferred to an Immobilon-P nylon membrane (Millipore, Bedford, MA), and immunoblotting analysis was performed as described below. The relative amount of DNA bound by WT1 protein was quantified utilizing a PhosphorImager and ImageQuant software (Molecular Dynamics, Sunnyvale, CA).
Immunoblotting-Cell lines were grown to 80% confluence in 10-cm plates, scraped into phosphate-buffered saline (PBS), and centrifuged at 1000 ϫ g. The cell pellets were resuspended in 200 l of PBS and solubilized in 200 l of 2ϫ Laemmli buffer (60 mM Tris, pH 6.8, 2% SDS, 10% glycerol) for 1 h at 4°C, The protein samples were then incubated at 100°C for 10 min and centrifuged in a microcentrifuge at 4°C for 15 min, and the resulting supernatant was collected. Protein concentrations were determined using the Bio-Rad DC kit. ␤-Mercaptoethanol was added to the samples to a final concentration of 0.5 M, and then heated at 85°C for 10 min prior to loading on a gel. Protein samples (40 g) were separated by electrophoresis through a 12% SDS-polyacrylamide gel and transferred to an Immobilon-P (Millipore, Bedford, MA) nylon membrane. The filters were blocked with 5% dry milk in PBS for 2 h, incubated with a rabbit polyclonal anti-WT1 antibody (C19, Santa Cruz Inc.) at 200 ng/ml, or with a rat monoclonal anti-mouse E-cadherin antibody (ECCD-2, Zymed Laboratories Inc., San Francisco, CA) at 1 g/ml in 0.5% dry milk Tris-buffered saline with 0.05% Tween 20 for 1 h, rinsed in Tris-buffered saline with Tween 20, and probed with a horseradish peroxidase conjugated anti-rabbit (Roche Molecular Biochemicals) (1:7500 dilution) or anti-rat (Santa Cruz Biotechnology) (1:1000 dilution) secondary antibody for 1 h. After rinsing in Trisbuffered saline with Tween 20, the immunoblots were developed by WT E-cad: Activates the E-cadherin Gene chemiluminescence (ECL, Amersham Pharmacia Biotech).

Induction of E-cadherin Expression in NIH 3T3 Cells by Stable Transfection and Retroviral Infection-In previous studies, we created a number of normal (WR16 and WR35) and
Ras-transformed (VW9) NIH 3T3 cell lines stably expressing WT1 isoform A (Ϫ17 amino acids, ϪKTS). In these cells, a number of epithelial markers were consistently up-regulated, including the E-cadherin gene ( Fig. 1A and Ref. 26). To examine if the endogenous E-cadherin gene is directly and immediately regulated by WT1(A), NIH 3T3 cells were infected with a bicistronic retrovirus carrying WT1(A) and GFP (Fig. 1B). Ecadherin mRNA was induced in NIH 3T3 cells as early as 24 h after infection with a retrovirus harboring WT1(A) and continued to increase at 48 and 72 h after infection. No induction of E-cadherin was noted in a control culture infected with a virus containing only GFP. These results suggested that E-cadherin induction was a direct result of WT1 expression rather than an effect of clonal selection.
Transcription of the Mouse and Human E-cadherin Promoter Is Activated by WT1-In order to determine if WT1 directly regulated the E-cadherin gene, we tested the ability of WT1 to activate both the mouse and human E-cadherin promoters. Previous studies indicated that the proximal promoter region of murine E-cadherin (Ϫ178 to ϩ113) is required for epithelial expression (47-49) (Fig. 2). Within the proximal promoter there are three regions that contribute to epithelial expression ( Fig. 2A) (50,51): 1) a palindromic element (Pal/Ap2) located at (Ϫ98 to Ϫ78), which stimulates transcription in epithelial cell and represses E-cadherin expression in mesenchymal cells; 2) a CAAT box (Ϫ65 to Ϫ61); and 3) two G-C rich regions I (Ϫ52 to Ϫ44) and II (Ϫ41 to Ϫ32), which confer basic epithelial promoter activity. A WT1/EGR-1 binding site (CCGGGGGCG) was found at position Ϫ51 to Ϫ43 of the murine E-cadherin promoter ( Fig. 2A), suggesting that WT1 could directly regulate E-cadherin. Sixteen different murine E-cadherin promoter fragments were tested for their response to the WT1 protein (Fig. 2B). Initially, the full-length proximal (Ϫ178) reporter construct was co-transfected into NIH 3T3 cells with vectors expressing WT1 from the Rous sarcoma virus long terminal repeat (Fig. 3). All four major isotypes of WT1 (A, B, C, D) strongly activated the E-cadherin promoter. Particularly striking activation was noted in response to WT1(A) and WT1(B), the isoforms of WT1 that lack the KTS sequence between zinc fingers 3 and 4, while isoforms C and D (ϩKTS) stimulated the promoter to a lesser extent. Since our previous studies were performed using the WT1(A) protein, we continued analysis with this isoform.
Four deletion constructs of the E-cadherin promoter were utilized to test the importance of the three regions in E-cadherin promoter that contribute to epithelial expression. The E-cad(Ϫ178/ϩ113) construct contains wild-type proximal promoter elements which, as described previously, are sufficientspecific for epithelial expression (47). The E-cad(Ϫ74/ϩ113) and E-cad(Ϫ67/ϩ113) constructs are deleted for the palindromic element (E-pal), but retain a CAAT box and a GC-rich potential WT1 binding site consensus sequence. The E-cad(Ϫ61/ϩ113) construct is deleted for the E-Pal and CAAT box, but retains the GC sequence. The E-cad(Ϫ61/ϩ113[GCm]) construct harbors a point mutation in the GC/WT1 element. The (Ϫ30) construct is a complete deletion of the three Ecadherin epithelial responsive elements (Fig. 2).
As shown in Fig. 4A, the full-length construct (Ϫ178), was activated about 15-fold by WT1(A) in NIH 3T3 cells relative to a control transfection with the insertless RSV expression vector. The (Ϫ74) and (Ϫ67) constructs, which were deleted for the E-pal element, were still strongly activated by WT1(A) (25-and 15-fold, respectively), which showed that E-pal segment is not important for WT1-mediated activation. Deletion of the CAAT box and GC elements (Ϫ30) significantly abrogated the ability of WT1(A) to activate reporter gene activity. Therefore, the minimal set of cis-acting sequences responsive to WT1(A) included the GC-I box (Ϫ61). Consistent with this view, when the GC-I sequence was mutated from CCGGGGGCG to CCGGGT-GCG, the ability of the (Ϫ61) reporter construct to be activated by WT1 was abolished in a manner similar to deletion of the GC-I sequence (i.e. the Ϫ30 construct) (Fig. 4A). It must be noted that the deletion mutations of the E-cadherin promoter, in addition to eliminating the ability of WT1 to activate tran-scription, severely impaired the basal activity of the promoter. This may be due to elimination of GC-rich sequences, which can bind the widely expressed Sp1 and CAAT-binding transcription factors. Potentially WT1 may need to cooperate with such basal elements and transcription factors in order to activate gene transcription.
Constructs Ϫ74 and Ϫ67 had the strongest response to WT1, which suggested that both the CAAT box and the GC-I region were essential for full activation. Therefore, we next performed a fine mutational analysis of the region between Ϫ67 and Ϫ44 to pinpoint the WT1-responsive element(s). For convenience, we divided the region into three subregions (C is the CAAT box; I is the intermediate region between CAAT box and GC-I; and GC is the GC-I site) (Fig. 2B). Whereas the wild-type Our findings did not rule out the possibility that intermediate sequences, including nucleotides "GCGG" located between the CAAT box and GC-I site were able to substitute for the loss of the GC-I site and that the observed response of the CAAT element to WT1 was really due to a second cryptic WT1-response element. To test this hypothesis, the GCGG sequence was drastically altered to TTTT and co-transfection analysis with WT1 was performed (Fig. 4C). When the CAAT box and GC-I sites were intact, mutation of the intermediate sequence (C-Im-GC) alone did not affect the ability of the construct to be induced by WT1. Mutation of the intermediate sequence in combination with either a mutant CAAT box (Cm-Im-GC) or mutant GC-I site (C-Im-GCm, C-Im-GCmm) (Fig. 4C) did not drastically affect the reporter activity (Fig. 4, compare B and  C). When the CAAT, intermediate, and GC-I sequences were simultaneously mutated (Cm-Im-GCmm in Fig. 4C), response to WT1 was lost. Collectively, our findings indicate that WT1 mediates activation of the murine E-cadherin promoter through the GC-I element and the CAAT-box (see Table I). The GC element was sufficient to mediate significant induction of the promoter, but the combination of the two sites led to an additive effect on transcription in the presence of WT1.
The human and murine E-cadherin promoters are very similar (47,49), and the GC element and CAAT box are conserved.

FIG. 3. Transcription of the murine E-cadherin promoter by the different isoforms of WT1.
A murine E-cadherin promoter/luciferase construct (0.1 g) was transiently co-transfected into NIH3T3 cells along with an expression vector for each of the four major WT1 spliced isoforms (WTA, WTB, WTC, WTD) (0.01, 0.1, 0.5, and 1 g) and a Renilla luciferase internal control vector (0.01 g). At 60 h after transfection, a dual luciferase assay was performed. Normalized luciferase expression Ϯ S.D. of triplicate experiments is plotted. White bar, 0.01 g; pale gray bar, 0.1 g; dark gray bar, 0.5 g; black bar, 1 g.
The human E-cadherin promoter responded to WT1 in a pattern qualitatively similar to the murine promoter (Fig. 5). Expression from a reporter containing sequences from Ϫ368 to ϩ125 of human promoter HE-cad (Ϫ368/ϩ125) was activated about 3-fold by WT1(A) relative to RSV control transfection, while in this set of experiments the murine promoter was activated 5-fold.
HE-cad (Ϫ217/ϩ125), which still contains the essential elements of the human E-cadherin promoter, was activated to a similar extent as the full proximal promoter. HE-cad(DEL), a construct in which all three critical elements of the promoter (Pal/Ap2, CAAT, and GC-I) were deleted, was not responsive to WT1. The CAAT box region and GC-I are well conserved between human and mouse, but the Pal/Ap2 region is not, which suggests the importance of CAAT and GC-rich region in Ecadherin gene regulation ( Fig. 2A) and the potential evolutionary conservation of a WT1/E-cadherin regulatory circuit.
Disruption of E-cadherin Expression by a Dominant Negative Form of WT1-Dominant negative form of WT1 mutants were previously identified both in tumor specimens as well as in patients with the Denys Drash syndrome (2). We showed that one particular mutant, WT1(AR) (43), which is deleted for exon 9 sequences encoding the third zinc finger motif of WT1, inhibited the ability of wild-type WT1 to activate transcription in a cotransfection assay (41). Therefore, we determined if WTAR could disrupt the activation of E-cadherin in cells stably and transiently expressing wild-type WT1. In a WT1-expressing NIH 3T3 cell line (WR16), E-cadherin promoter activity was approximately 5 times higher than in control NIH 3T3 cells that do not express WT1 (Fig. 6A). However, when WR16 cells were transiently transfected with the E-cadherin promoter along with a vector expressing the WT1(AR) mutant allele, E-cadherin promoter activity decreased by over 80% to a level similar to that obtained in control NIH 3T3 cells (Fig. 6B). When WTAR was transiently cotransfected with WT1(A) in the NIH 3T3 cells, WTAR blocked the WT1A-mediated activation of the E-cadherin promoter by more than 50% (Fig. 6C).
WT1 Binds to the GC-I Element of the E-cadherin Promoter-As shown above, deletion analysis of the murine E-cadherin promoter indicated that a GC-rich, EGR-1-like element and the CAAT box were required for WT1-mediated activation. To demonstrate a direct interaction between WT1 protein and these promoter sequences, a electrophoretic mobility shift assay (EMSA) was performed. An oligonucleotide corresponding Activates the E-cadherin Gene to nucleotides Ϫ62 to Ϫ24 of the E-cadherin promoter, which contains the GC-I/WT1 binding site, was incubated with in vitro transcribed and translated WT1 and subjected to electro-phoresis. A DNA-protein complex was formed when the oligonucleotide was incubated with WT1-programmed rabbit reticulocyte lysate (Fig. 7A, lane 2, lower arrow), but not with unprogrammed lysate (lane 1). This complex was specifically retarded by a WT1 antibody (Fig. 7A, lane 4, upper arrow) but was unaffected by a control antibody, G4 (lane 6). No WT1/DNA complex was formed on a mutant oligonucleotide representing the same GCm sequence that disabled the response of the murine E-cadherin promoter to co-expressed WT1 (data not shown). Unlabeled GC oligonucleotides efficiently competed for WT1 binding to the wild-type probe, while the GCm sequence could not effectively compete for binding even when present in the reaction at a 2000-fold molar excess (Fig. 7B).
Since all four isoforms of WT1 could transactivate the Ecadherin promoter (Fig. 3), we tested whether both the ϩ and Ϫ KTS forms of WT1 could bind to the GC-I/WT1-response element. GST fusion proteins containing the four WT1(A/ϪKTS) zinc fingers, WT1(C/ϩKTS) zinc fingers, and three zinc fingers of the dominant negative/DNA binding-deficient WT(AR) mutant were allowed to bind an oligonucleotide containing the GC and CAAT elements of the E-cadherin promoter (Fig. 8). As expected (43), GST-WT1(AR) did not bind to DNA. GST-WT1(A/ϪKTS) bound relatively strongly to the promoter sequence, while WT1(C/ϩKTS) yielded 1/10 the amount of DNAprotein complex. This is consistent with transfection data, which showed that WT1(C) and WT1(D) have a reduced but definite capacity to activate the E-cadherin promoter (Fig. 3).
To map the WT1 binding element within the E-cadherin promoter, oligonucleotides containing both the CAAT box and the GC-I/WT1 binding site (Sequences 18 -27) as well as mutant versions of these sites were utilized in an EMSA with in vitro translated WT1(A) (Fig. 9). Since the CAAT box and the GC-I site recognize many widely expressed transcription factors, it was necessary to add a WT1 antibody to the reaction to more clearly identify the WT1/DNA complex (Fig. 9, A-D, lane  1, top arrow). A WT1 antibody supershift complex was observed only when oligonucleotides containing the wild-type GC-I/WT1 binding site were used as probes (Fig. 9, A (lanes 5 and 9), and B (lane 5)). Mutations in the CAAT box or the intermediate sequence still allowed the WT1 protein to bind to GC-I site (Fig.  9, A (lane 9) and B (lane 5)). An oligonucleotide containing the CAAT box alone did not have any WT1 binding activity (Fig.  9B, lanes 7-10). Oligonucleotides with a wild-type CAAT box and mutant GC-I site (Fig. 9, C and D, lanes 3-6) did not bind to WT1 even though the CAAT box could mediate transactivation in response to WT1. To show that WT1 produced in tran- FIG. 5. Activation of the human Ecadherin promoter by WT1. NIH 3T3 cells were transfected with 0.5 g of a RSV control expression vector (white bar) or 0.5 g of RSV-WT1(A) (black bar) along with 0.1 g of the indicated human Ecadherin promoter construct. The HECad (Del) is devoid of all three major promoter elements (Pal/Ap2, CAAT, and GC-I) and is similar to Ϫ30 murine E-cadherin promoter construct. A full-length murine Ecadherin promoter construct (MECad Ϫ178) was used as a positive control. A Renilla luciferase expression vector (0.01 g) was used to normalize for transfection efficiencies. The average normalized luciferase activity Ϯ S.D. of triplicate experiments is presented.

FIG. 6. A dominant negative WT1 mutant inhibits the E-cadherin promoter.
A, the E-cadherin promoter is stimulated in WT1expressing NIH 3T3 cells. The E-cadherin luciferase construct (0.1 g) was transfected into parental NIH 3T3 cells or WT1-expressing WR16 cells along with 0.01 g of the Renilla luciferase control expression vector to normalize transfection efficiency. B, the E-cadherin promoter (0.1 g) was transfected into WR16 cells along with an insertless RSV expression vector or RSV-WT1(AR) (0.5 g) encoding a dominant negative form of WT1. C, the E-cadherin promoter (0.1 g) was co-transfected into NIH 3T3 cells with RSV-WTA (0.5 g) in the presence or absence of RSV-WT1(AR) (0.5 g). In all experiments, normalized luciferase activity was determined 60 h after transfection. The average normalized luciferase activity Ϯ S.D. of triplicate experiments is presented. sient transfections could bind the GC element, we attempted mobility shift assays utilizing extracts from transfected cells but a large number of background complexes were noted, potentially representing proteins such as SP1 and Ap2 that can also bind to GC-rich elements. To determine if WT1 was among these complexes, we performed an EMSA with extract from 293T cells transfected with RSV-WT1(A). followed by transfer of the gel for immunoblot analysis (Fig. 10). A WT1-DNA complex signal was detected only in 293T cell nuclear extracts transfected with WT1 (lanes 3 and 4) and not in NIH 3T3 cell nuclear extract (Fig. 10, lanes 1 and 2) or mock-transfected 293T cell extract (Fig. 10, lanes 5 and 6). A summary of DNA binding activity and transcriptional response of the mutant promoter constructs is presented in Table I. DISCUSSION Although it had been shown that WT1 protein could activate or repress the expression of a number of promoters in co-transfection experiments (2), then endogenous forms of many of these putative target genes were later shown not to be regulated by WT1 in biological systems (25,52). Much of this discrepancy is due to the mode by which WT1 target genes were identified, namely by searching the promoters of promising candidate target genes for sequences matching the EGR1/WT1 binding site. Given the relatively GC-rich nature of many promoters, this approach identified a number of putative target gene. True WT1 target genes should have a specific binding site for the protein, respond to WT1 expression in transient assays, and be activated or repressed by both transient and stable co-transfection. However, in addition, the expression of the endogenous gene, not just a promoter construct should be affected by expression of WT1. Furthermore, a WT1 target gene should be expressed in the developing and adult organism in a temporal and spatial pattern that is concordant with WT1 expression. The transcriptional activity of WT1 required for growth control has been controversial. Recently, we demonstrated that missense mutants of WT1 defective for transcriptional activation, but competent for transcriptional repression, failed to suppress cell growth (42). This indicates that genes whose transcription is activated by WT1 may be critical for the biological activity of this tumor suppressor.
We have attempted to identify WT1 target genes using a model system in which engineered expression of WT1 converts the NIH 3T3 fibroblast cell line into cells with many features of epithelia (26). This system was used to survey the expression of many epithelial markers without prior knowledge of the nature of their promoter sequences. E-cadherin was among those genes up-regulated in WT1-expressing cells. Here we report that the E-cadherin gene meets criteria for a WT1 target gene potentially relevant for the roles of WT1 in both inhibiting cell growth and inducing differentiation. Though up-regulation of E-cadherin in the original cell lines could possibly be attributed to clonal selection, we found that E-cadherin mRNA expression was induced within 24 h of infection of NIH 3T3 cells with a retrovirus expressing WT1, consistent with a direct effect of WT1.
Deletions and point mutational analysis of the murine and human E-cadherin promoters and mobility shift assays indicated that an EGR-1-like cis-acting sequence was necessary and sufficient for WT1-mediated transactivation. We then demonstrated that the GC element and the CAAT box were required for full and robust WT1-mediated transactivation of E-cadherin in mesenchymal NIH 3T3 fibroblasts (Fig. 4). Furthermore, a dominant negative mutant of WT1 WT1(AR) inhibited E-cadherin promoter activity in cells expressing wild-  1-6), or a mutant oligonucleotide (lanes 7-12) containing a G to T transversion was incubated with control reticulocyte lysates (C) or in vitro translated WT1 (W). Antibodies used for supershift assay were anti-WT1 (C19) (lanes 3 and 4) and an anti-Gal4 (G4) antibody as a negative control (lanes 5 and 6). B, competition EMSA showing specificity of WT1 binding. In vitro translated WT1 was allowed to bind to the GC element in the absence of competitor or in the presence of the indicated -fold excess of wild-type or mutant GC sequence from the E-cadherin promoter. type WT1 (Fig. 6). Finally, we found that WT1 protein derived from nuclear extracts could bind to the WT1-response element in the E-cadherin promoter (Fig. 10).
It must be noted that detection of WT1 DNA binding activity in nuclear extracts has been reported in only a few publications (53)(54)(55). Curiously, WT1 also appeared to mediate activation of the E-cadherin promoter through a CAAT box, because mutation of this element, but not sequences between the CAAT box and the GC-I element, significantly impaired the ability of WT1 to activate the E-cadherin promoter. Furthermore, a construct with a mutation in the GC box, but with an intact CAAT box, still supported transcriptional activation by WT1 (summarized in Table I). This is consistent with prior information that WT1 can affect transcription through protein-protein interactions. WT1 was shown to interact with p53 and modulate the transcription of reporter genes containing only p53 binding sites (56). WT1 also affects the transcription of the HSP70 promoter, presumably by sequestration of HSP70 from the heat shock transcription factor, thus leading to the activation of the heat shock transcription factor (57). The CAAT box can bind to the CTF/NF1 factors as well as the heteromeric CP1/CP2 factor. Whether WT1 directly binds to these factors to augment their transcriptional effect or sequesters negative regulators that interact directly or indirectly with the E-cadherin CAAT box is currently under investigation. However, the finding that a CAAT box may act as a WT1-responsive element suggests a novel mechanism of WT1 action and a broader potential range of target genes.
It was previously reported that epithelial-specific transcription of E-cadherin required a palindromic sequence as well as two tandem AP-2 transcription factor binding sites within the proximal GC-rich portion of the promoter, which we have demonstrated can also bind to WT1 (47,50,51). Rb and Myc were later shown to activate E-cadherin gene expression via interaction with AP-2, through the E-Pal and GC-rich region in epithelial cells, but not in mesenchymal NIH 3T3 cells (58). Consistent with these data, we found that transfection of Rb did not activate the E-cadherin promoter and co-transfection of WT1 and Rb did not yield synergistic or additive activation on E-cadherin expression in NIH 3T3 cells (data not shown). However, unlike Rb or Myc, WT1 could transactivate the E-cadherin gene in non-epithelial NIH 3T3 cells. This suggests an independent pathway for epithelial differentiation exerted by WT1 and is consistent with the notion that WT1 might upregulate E-cadherin expression in mesenchymal cells during the mesenchymal-epithelial transition. However, our studies have not excluded the possibility of inteaction between WT1 and AP-2 or other proteins bound to the E-Pal site. WT1mediated activation was most efficient when the E-Pal element was deleted (Figs. 3 and 4). This finding suggests the presence of a negative regulator bound to the E-Pal element in mesenchymal cells, consistent with previous findings (47,50,51). Such a protein might act to counter WT1-mediated activation of E-cadherin in certain tissues.
E-cadherin expression is characteristic of epithelia and is critical for epithelial development (27). E-cadherin is required for embryogenesis, as targeted disruption of this gene leads to failure of early murine development due to a defect in embryo compaction (59,60). More specifically, in the kidney, E-cad- Oligonucleotides representing nucleotides Ϫ67 to Ϫ24 of the E-cadherin promoter, which contains both CAAT box and GC-I WT1 binding site and its mutant versions, were incubated with in vitro translated WT1 (W) or unprogrammed lysate (C). Antibodies used for supershift assay were anti-WT1 C-19 antibody (C19), while a anti-Gal4 antibody (G4) was used as a negative control. A, probes C-I-GC and Cm-Im-GC. B, probes Cm-I-GC and CAAT box alone. C, probes C-Im-GCmm and Cm-Im-GCmm. D, probes C-I-GCmm and Cm-I-GCmm.
FIG. 10. EMSA/Western analysis of WT1 protein bound to GC-I/WT1 binding site. An EMSA assay was performed allowing an unlabeled Ϫ62/Ϫ24 oligonucleotide to interact with nuclear extracts derived from NIH 3T3 cells (lanes 1 and 2), mock-transfected 293T cells (lanes 5 and 6), and 293T cells transfected with a WT1 expression plasmid (lanes 3 and 4). The DNA-WT1 complex was separated by non-denaturing gel electrophoresis and transferred to a nylon membrane, followed by immunoblot with an anti-WT1 antibody. herin expression is induced in the condensing metanephric mesenchyme during the mesenchymal-epithelial transition (37,38,61) at the same time that WT1 is up-regulated in this tissue (39,40). This fact, along with the other evidence outlined, makes a compelling case for E-cadherin as a direct WT1 target gene. Direct transactivation of E-cadherin by WT1 demonstrates an evident link between the tumor suppressive activity of WT1 and a gene that plays a major role in differentiation. It confirms our previous report that WT1 plays a direct role in both the induction of the epithelial phenotype and inhibition of tumorigenesis by controlling the expression of a subset of genes encoding molecules that control epithelial characteristics (26). E-cadherin might mediate many facets of growth suppression by WT1. Homotypic interactions mediated by E-cadherin lead to increased cell adhesion and decreased invasiveness. Furthermore, E-cadherin was recently demonstrated to have an anti-proliferative effect of its own, correlated with the up-regulation of the CDK inhibitor, p27 (62). We checked for p27 up-regulation after infection of 3T3 cells with the WT1 retrovirus and found that p27 was expressed at a basal level in 3T3 but was not up-regulated by WT(A) (Data not shown). This may be due to the relatively low level of Ecadherin induced after transient expression and does not preclude the possibility that E-cadherin expression in the developing kidney may affect the expression of p27 or other cyclindependent kinase inhibitors.
The E-cadherin gene is frequently inactivated in human cancer, and its down-regulation is particularly associated with the acquisition of an invasive tumor phenotype (28,34,63). Allelic loss of heterozygosity of E-cadherin has been reported in cervical, prostrate, breast, hepatocellular, and colorectal carcinomas as well as in Wilms' tumor itself (28, 64 -68). Point mutations of the E-cadherin gene were found in gastric, renal, lung, colorectal, and breast cancers (28, 66, 69 -74). Defects in trans-acting pathways regulating E-cadherin expression have been suggested in breast cancer (44), and epigenetic inactivation of E-cadherin gene was also implicated by changes in chromatin structure and the methylation status of CpG sites within of the E-cadherin promoter (50). Recently, E-cadherin expression within Wilms' tumors was found only in a limited set of cells resembling renal tubules (75). These tumors coexpressed cytokeratin and vimentin, underscoring the aberrant differentiation of the tumor. Altered E-cadherin expression as a result of lost or mutated WT1 could play a role in the pathogenesis of Wilms' tumors. Furthermore, E-cadherin complexes with ␤-catenin, a protein involved in transduction of wnt signaling to the nuclear TCF/LEF transcription factors (76). Activating mutations of ␤-catenin were recently identified in Wilms' tumor lending more evidence to the potential importance of disruption of the E-cadherin/catenin axis in this tumor (77).
In summary, E-cadherin is a gene coordinately regulated with WT1 during kidney development, whose expression can be activated in response to stable or transient expression of WT1 and whose promoter can be bound and activated by WT1. Hence, E-cadherin is a likely a bona fide target gene of WT1.