A Far Upstream Cis-element Is Required for Wilms’ Tumor-1 (WT1) Gene Expression in Renal Cell Culture*

To identify novel cis-regulatory elements responsible for the tissue-restricted expression pattern of the Wilms’ tumor-1 (WT1) gene, we mapped a total of 11 DNase I-hypersensitive sites in the 5′-flanking region and first intron of the human gene, six of which were specific for WT1expressing cell lines. A 1.4-kilobase (kb) fragment from the mousewt1 5′-flanking region contained cross-hybridizing sequence with significant homology to a region of DNase I hypersensitivity in the human WT1 gene which bound to nuclear matrix in human fetal kidney 293 cells. None of the DNase I-hypersensitive sites/matrix attachment regions, either alone or in combination, were sufficient for tissue-specific WT1 expression in transient and stably transfected cell lines. However, stable transfection of an approximately 620-kb yeast artificial chromosome (YAC) that carried the entire mouse wt1 locus into 293 cells resulted inwt1 (trans)gene expression at a level of approximately 30% of the endogenous human gene. Deletion of the 1.4-kb cross-hybridizing mouse fragment, located approximately 15 kb upstream of the transcription start site, caused complete loss of wt1 gene expression in the YAC-transfected 293 cells. In summary, we have identified a far upstream element that contains a region of DNase I hypersensitivity and that binds to nuclear matrix. This element includes phylogenetically conserved sequence and is required, although not sufficient, for mouse wt1 gene expression in human fetal kidney cells in culture.

The Wilms' tumor-1 (WT1) 1 gene plays a critical role in genitourinary development and in the pathogenesis of Wilms' tumors. Wilms' tumor (nephroblastoma) is a childhood malignancy of the kidney that affects 1 in 10,000 children. Wilms' tumors arise when the metanephric mesenchyme fails to differentiate into developing glomeruli and tubules. Histomorpho-logically the tumors often have a triphasic appearance consisting of blastemal, stromal, and epithelial elements (1). These tissue components mimic, although incompletely, normal kidney development suggesting that Wilms' tumors result from an abnormal differentiation program of embryonic kidney cells. Most cases of Wilms' tumors are sporadic and unilateral. Occasionally, tumors develop in both kidneys, in a familial pattern, and are associated with a more complex malformation syndrome. A single WT1 allele loss resulting from a deletion on chromosome 11p13 is found in patients with WAGR syndrome which includes Wilms' tumor, aniridia, genitourinary malformations, and mental retardation (2,3). The Denys-Drash syndrome is caused by a single allele point mutation in the WT1 gene and is characterized by Wilms' tumors, male genital ambiguity, and progressive nephropathy (4 -6). WT1 gene defects are found in up to 20% of Wilms' tumors suggesting that the WT1 gene encodes a tumor suppressor.
The WT1 gene product is a zinc finger protein of the Cys 2 -His 2 type with significant homology to the DNA binding domain of the EGR family of zinc finger transcription factors (7)(8)(9). Alternative splicing of exon 5, which encodes 17 amino acids, and insertion of three additional amino acids (KTS) between zinc fingers three and four results in four different isoforms of WT1 proteins (10,11). These four WT1 isoforms, conserved among mammals both in their structure and relative abundance, differ in their DNA binding affinities and specificities (12,13). RNA editing of WT1 transcripts may contribute to even more heterogeneous WT1 proteins (14). In most contexts, the WT1 protein binds to GC-and TC-rich consensus sequences and acts as a transcriptional repressor (12,15,16). Putative WT1 target genes include the genes for insulin-like growth factor-2 (17), platelet-derived growth factor A-chain (18,19), epidermal growth factor receptor (20), insulinlike growth factor-1 receptor (21), PAX2 (22), and WT1 itself (23,24), among others.
Major sites of WT1 gene expression in the body are the genitourinary tract and the mesothelial cells of heart, lung, and abdomen (25)(26)(27). WT1 is tightly regulated during kidney development. Low levels of WT1 mRNA are detectable in the undifferentiated mesenchyme of the metanephric kidney (27)(28)(29). WT1 message dramatically increases upon induction of mesenchymal cells by the ureteric bud and persists in the renal vesicle, comma-, and S-shaped bodies, where it is restricted to the podocytes of the developing glomerulus (25). WT1 is an essential gene as its homozygous disruption in mice caused agenesis of the kidneys probably resulting from apoptosis of the metanephric blastema (30). In addition, the wt1 knock-out mice also showed hypoplasia of the heart and lungs likely due to defects of the mesothelium (30). The characteristic expression pattern and the results obtained from homozygous germ line disruption studies suggest a critical role for WT1 in mesenchymal-epithelial differentiation.
Little is known about the molecular mechanisms of WT1 gene regulation. The human WT1 gene contains a GC-rich, TATA-and CCAAT-less promoter (31,32) which has several potential consensus sequences for EGR/WT1 (33), Pax-8 (34), and other transcription factors. Recent studies from our laboratory indicate that a substantial part of WT1 gene expression is regulated at the level of transcription (35). We have recently shown that Sp1, which binds to a 9-base pair CTC repeat in the WT1 proximal promoter, accounts for Ϸ80% of WT1 transcription in cultured cells stably transfected with reporter constructs (35). A plasmid containing Ϫ1900 to ϩ200 bp of the human WT1 promoter region linked to a lacZ reporter gene, however, did not mimic the characteristic expression pattern of the endogenous wt1 gene in transgenic mice. 2 These findings suggested that additional cis-elements might be required for the tissue-restricted WT1 expression. A 350-base pair 3Ј-enhancer which contains two GATA motifs (33,36) and a transcriptional silencer in the third intron of the WT1 gene (37) have recently been identified. Both elements were active in cells of the erythroid lineage but failed to direct expression of WT1 in cells of renal origin. Our efforts have therefore focused on the identification of novel cis-regulatory elements responsible for WT1 gene expression in renal cells.

DNase I-hypersensitive Site Mapping
Probe Preparation-A set of unique probes were obtained for the entire WT1 locus. 3 Briefly, cosmids L109, L156, and L159 (gift of Dr. Daniel Haber, described in Ref. 7) were digested with a variety of restriction enzymes, separated by electrophoresis through agarose, stained with ethidium bromide, photographed, and subjected to Southern blotting. Sheared human placental DNA was labeled with 32 P by random priming using a PrimeIt kit (Stratagene, La Jolla, CA) and used as a probe. Following washing under conditions of high stringency, the membranes were exposed to film at Ϫ80°C. Those restriction fragments, identified by ethidium bromide staining but failing to hybridize to sheared human placental genomic DNA, were assumed to be unique WT1 probes devoid of repetitive sequences and were chosen as probes for DNase I-hypersensitive site mapping.
Nuclei Preparation-Nuclei were prepared according to the method of Shapiro et al. (38) for the following cell lines obtained from ATCC: K562, CEM, HFK 293, HeLa, and TK10.
DNase I-hypersensitive Site Mapping-Nuclei were stored as 0.5-ml aliquots of 5 OD 260 nm at Ϫ80°C. Individual aliquots were thawed and subjected to DNase I digestion followed by deproteinization and isolation of DNA. The amount of DNase I chosen for each digest was determined empirically for each cell type. A series of DNase I digests were performed on each cell type to obtain a gradual digestion of the genomic DNA by DNase I. Following the recovery of the DNase I-digested DNA, it was subjected to genomic Southern blotting using the non-repetitive probes identified above. By scanning the lanes from low to high concentration of DNase I, the appearance of sub-bands on the autoradiogram signified the identification of DNase I-hypersensitive sites (Fig. 1A).

Matrix Attachment Region Assays
Nuclear matrix was prepared according to the method of Bode and Maass (39). Nuclei from approximately 5 ϫ 10 7 cells were prepared as follows. Cells were washed with 2 ϫ 50 ml of isolation buffer (3.75 mM Tris-HCl, pH 7.4, 0.05 mM spermine, 0.125 mM spermidine, 0.5 mM K-EDTA, 20 mM KCl, and 1% thiodiglycol, freshly adjusted to 0.2 mM with phenylmethylsulfonyl fluoride) and scraped off into 15 ml of isolation buffer containing 0.1% digitonin. Nuclei were released with 15 strokes in a tightly fitting Dounce homogenizer. Following centrifugation (900 ϫ g, 5 min at 4°C) they were washed twice in the same medium. Nuclei were resuspended in 100 l of nuclear stabilization buffer (5 mM Tris-HCL, pH 7.4, 0.05 mM spermine, 0.125 mM spermi-dine, 20 mM KCl, 1% thiodiglycol, 0.1% digitonin, 0.2 mM phenylmethylsulfonyl fluoride and 1% aprotinin) and incubated at 42°C for 30 min. Nuclear halos were obtained by the addition of 2 ml of LIS dilution buffer (20 mM HEPES-NaOH, pH 7.4, 0.1 M lithium acetate, 1 mM K-EDTA, and 0.1% digitonin) and homogenizing with two strokes with a loosely fitting pestle, followed by the addition of 2 ml of 50 mM lithium 3,5-diiodosalicylate (LIS) to extract non-matrix proteins. Halos were washed following centrifugation (2400 ϫ g, 5 min, 4°C) in 4 ϫ 50 ml of sterile digestion buffer (20 mM Tris-HCl, pH 7.4, 0.05 mM spermine, 0.125 mM spermidine, 20 mM KCl, 70 mM NaCl, 10 mM MgCl 2 , 0.2 mM phenylmethylsulfonyl fluoride). Nuclear matrix was obtained by digestion with 200 units each of EcoRI, HindIII, and BamHI for 4 h at 37°C. Matrix was pelleted and resuspended in 700 l of sterile digestion buffer containing 0.5 mg/ml sonicated Escherichia coli DNA, which was needed as a nonspecific competitor in the matrix binding assay.
Double-stranded DNA fragments to be tested for nuclear matrix binding were end-labeled with 32 P, and 30,000 cpm were added to each aliquot and incubated overnight at 37°C with gentle mixing. Following incubation, matrix binding reactions were pelleted at 12,000 ϫ g for 15 min at 4°C. Supernatants were removed to a new tube. Pellets were washed with 3 ϫ 1 ml of digestion buffer and resuspended in 150 l of digestion buffer. Supernatants and pellets were adjusted to 0.1% SDS, 10 mM EDTA, followed by addition of proteinase K to 0.1 mg/ml and incubated overnight at 50°C. DNA was precipitated by the addition of equal volume of 2-propanol, washed with 70% ethanol, air-dried, and resuspended in 100 l of TE. Ten l each of pellet and supernatant fractions were electrophoresed through 1% agarose, dried onto Whatman 3MM paper, and autoradiographed. Fragments associating with the nuclear matrix appeared in the pellet fraction, and those not binding to the matrix remained in the supernatant.
As a negative control, pBluescript II (Stratagene, La Jolla, CA) was linearized with EcoRI and labeled with Klenow. The positive control, SAR 800 , was obtained by EcoRI/HindIII digestion of pCL (supplied by Dr. J. Bode). SAR 800 is an 800-bp matrix binding fragment from human interferon-␤ gene (39). Another positive control used was MAR-H (40). MAR-H was obtained as a 991-bp XbaI fragment from pHEN18-3.

P1 Clones
PCR screening of a mouse P1 library (Genome Systems, St. Louis, MO) was done with two sets of mouse wt1 gene-specific primers. The "upstream" primer pair was located from Ϫ568 to Ϫ547 bp (forward primer) and from Ϫ356 to Ϫ375 (reverse primer) relative to the major transcription start site. A 195-bp sequence from exon 10 of the mouse wt1 gene was amplified with the "downstream" primers. Eight P1 clones (1405-1412) were obtained which were maintained in the Cre recombinase negative bacterial strain NS 3516. Three clones (1405, Ϫ06, Ϫ07) bound the upstream primers only, and four clones (1409, Ϫ10, Ϫ11, Ϫ12) bound the downstream primers. Clone 1408 was found to bind both primer pairs (Fig. 5).
Growth of P1 Clones and DNA Isolation-P1 clones were grown in LB medium containing 25 g/ml kanamycin. To increase copy number, the plasmids were induced for 5 h with 0.5 mM isopropylthio-␤-D-galactosidase (Life Technologies, Inc.). The bacteria were harvested for DNA isolation and resuspended in 1 ml of GTE solution (50 mM glucose, 10 mM EDTA, pH 8.0, 25 mM Tris-HCl, pH 8.0). Lysozyme was added to a final concentration of 1.5 mg/ml, and the cells were incubated at room temperature for 5 min. Alkaline lysis was performed by adding 2 ml of 0.2 M NaOH, 1% SDS and incubating 5 min on ice. 1.5 ml of a 3 M potassium acetate solution was added followed by a 10-min incubation at 4°C. The samples were centrifuged at 10,000 ϫ g for 10 min, and the supernatants were carefully removed and incubated (30 min at 37°C) with DNase-free RNase A at a final concentration of 50 g/ml. After phenol/chloroform (1:1) extraction, the DNA was precipitated with an equal volume of isopropyl alcohol. The DNA pellet was washed with 70% ethanol and resuspended in an appropriate volume of TE buffer. DNA yield was approximately 500 ng/ml culture.
Cross-hybridization of the Mouse wt1 Gene with Human Probes-Ten g of genomic DNA from the mouse wt1 gene in the P1 vector was digested to completion with various restriction enzymes. The digested DNA was separated on a 0.8% agarose gel and transferred to positively charged nylon membrane (Boehringer Mannheim) with 10 ϫ SSC as transfer buffer. After UV cross-linking to the membranes and a 4-h pre-hybridization, the DNA was probed overnight at 65°C in 20 ml of a solution containing 0.25 M sodium phosphate buffer, pH 7, 7% SDS, 1% bovine serum albumin, 1 mM EDTA, and 25 g/ml salmon testis DNA. Membranes were washed at 37°C for 1 h with 0.1 ϫ SSC, 0.1% SDS and exposed to x-ray film with intensifying screens at Ϫ80°C. After the films were developed, the membranes were washed sequentially to increase stringency (55°C and then 65°C) and reexposed each time. Random primed probes (PrimeIt, Stratagene, La Jolla, CA) were generated from human regions 56R13-5 and 56R13-12 (Fig. 1B). DNA bands that were detectable by autoradiography after a high stringency wash (65°C) were subcloned into pBluescript II KS ϩ for further analysis.

YAC Clones
A Ϸ620-kb YAC clone (clone address, YAC-90-1A) containing the mouse wt1 gene in the pYAC4 vector (Fig. 5) was obtained by PCR screening of a mouse super pool YAC library (Genome Systems, St. Louis, MO). A 213-base pair sequence (Ϫ568 to Ϫ356 bp relative to the major transcription start site) from the mouse wt1 proximal promoter was amplified with the following PCR primers: 5Ј-CAATTTCACCTT-GAATCTCAAC-3Ј (forward primer) and 5Ј-TGTTAATCAGAAGGGT-GGGG-3Ј (reverse primer). The YAC was designated Y 620 mWT1 and was maintained in Saccharomyces cerevisiae strain AB1380.
Growth of the YAC Clones-Yeast cells were grown at 30°C in liquid medium containing (per liter) 1.7 g of yeast nitrogen base (Difco), 5 g of ammonium sulfate, 20 g of dextrose (Sigma), 100 mg of adenine hemisulfate, 0.72 g of complete supplement mixture lacking L-tryptophan and uracil (Bio 101, La Jolla, CA), pH 5.8. The cell density was determined spectrophotometrically at a wavelength of 600 nm. Solid medium differed from the liquid broth as it contained adenine hemisulfate at 40 mg/liter and 20 g/liter of bacto agar (Difco).
Preparation of Yeast Genomic DNA in Agarose Plugs-A 50-ml saturated yeast culture (Ϸ1 ϫ 10 8 cells/ml) was pelleted at 1000 ϫ g for 5 min and washed once in 50 mM EDTA. The cells were resuspended in 1 ml of SEM (1 M sorbitol, 20 mM EDTA, 14 mM 2-mercaptoethanol) supplemented with 100 g/ml zymolyase 100T (ICN Biomedicals, Costa Mesa, CA). One ml of a 2% low melting temperature agarose (SeaPlaque GTG, FMC, Rockland, ME) was added, mixed, and dispensed into the slots of the plug mold (Bio-Rad) taking care to avoid air bubbles. After cooling for 20 min at 4°C, the agarose plugs were removed and incubated in 30 ml of SEM with zymolyase 100T (100 g/ml) at 37°C for 2 h. The solution was decanted and replaced with 20 ml of yeast lysis solution containing 1% (w/v) lithium dodecyl sulfate, 100 mM EDTA, 10 mM Tris-HCl, pH 8. The agarose plugs were incubated overnight at 37°C, washed once in TE buffer, and stored at 4°C in 500 mM EDTA solution.
Pulsed-field Gel Electrophoresis-The agarose plugs (up to 5 plugs) were equilibrated in 50 ml of 0.5 ϫ TBE buffer at 4°C for 3 ϫ 30 min, loaded into the slots of a 1% agarose gel (SeaPlaque GTG), and sealed with the same 1% agarose. The following conditions were applied for efficient separation of yeast chromosomes by pulsed-field gel electrophoresis on a CHEF DRIII apparatus (Bio-Rad): 0.5 ϫ TBE running buffer (14°C), switching time 60 -120 s, 6 V/cm, 120°field angle, 24 h running time. After visualizing the DNA with 0.5 g/ml ethidium bromide the YAC could be easily identified as an additional chromosome of approximately 620 kb.
Mapping of the YAC Clones-Restriction mapping of the YAC clones was done by partial digestion of agarose-embedded yeast DNA. Increasing concentrations (0.1-20 units) of several restriction enzymes including MluI, SalI, and PvuI were used for partial digestion of genomic DNA. The digested DNA was separated on a pulsed-field gel, blotted onto nylon membrane (Boehringer Mannheim), and hybridized with specific probes for both YAC vector arms. The centric vector arm probe was made from a 2.7-kb BamHI/PvuII fragment, and the acentric vector probe was obtained from the 1.7-kb BamHI/PvuII fragment of pBR322 (41). The relative fragment lengths from both ends were determined to construct a restriction map of the YAC clone. As wt1 gene-specific probes we used a 0.5-kb BglII/EcoRI fragment from exon 10 and a 0.5-kb NheI/EagI fragment from the mouse wt1 proximal promoter.
Fluorescent in Situ Hybridization Analysis-Fluorescent in situ hybridization analysis was performed by Genome Systems (St. Louis, MO) according to their standard procedure.
"Retrofitting" the YAC with a Dominant Selectable Marker-A neomycin resistance gene was targeted into the URA3 site of the acentric YAC vector arm using homologous recombination in yeast. The retrofitting plasmid contained a neomycin resistance gene and a lacZ gene both driven by the pgk promoter, as well as a LYS2 marker gene for positive selection of yeast transformants (42). YAC containing yeast cells were transformed by the lithium acetate method (43). Yeast cultures were grown in 100 ml of dropout medium (SC Ura Ϫ / Trp Ϫ ) to a density of approximately 10 7 cells/ml. After pelleting (5 min at 1000 ϫ g) and washing in sterile H 2 O the cells were incubated at 30°C for 30 min in 15 ml of a solution containing 100 mM lithium acetate, pH 7.5, 10 mM Tris-HCl, pH 7.5, and 1 mM EDTA. A 200-l aliquot of the yeast suspension was carefully mixed with 5 l (3 g) of the DNA to be transformed and 15 l (150 g) of denatured salmon sperm DNA as a carrier (CLONTECH, Palo Alto, CA). 1.2 ml of polyethylene glycol solution (8 volumes of 50% PEG 4000, 1 volume of 10 ϫ TE buffer, pH 7.5, 1 volume 1 M lithium acetate, pH 7.5) was added, and the samples were shaken for 30 min at 30°C. Heat shock was performed at 42°C for 15 min. The cells were collected and resuspended in 300 l of TE buffer. A 100-l aliquot of the suspension was plated on a 10-cm Petri dish and grown for 3 days on an SC Trp Ϫ /Lys Ϫ solid dropout medium. The clones were replica-plated on SC Trp Ϫ /Lys Ϫ /Ura Ϫ medium, and colonies that grew on the SC Trp Ϫ / Lys Ϫ plates but did not replicate on SC Trp Ϫ /Lys Ϫ /Ura Ϫ plates were picked and grown to saturation in liquid medium.
The transformants were tested for integration of the neomycin resistance gene by PCR amplification of a 722-bp fragment spanning the URA3/pgk-promoter junction. 17 out of 22 clones examined gave the expected PCR product. Pulsed-field gel electrophoresis was performed with genomic DNA obtained from these clones. As compared with the original YAC, the retrofitted clones ran at a slightly higher (ϩ10 kb) molecular weight due to integration of the ␤geoLys2 cassette into the acentric vector arm. The PCR results were confirmed by Southern transfer and hybridization with a neomycin resistance gene-specific probe. The retrofitted YAC was designated Y 620 mWT1neo R .
Transfection of Human Fetal Kidney 293 Cells with the YACs-Human fetal kidney (HFK) 293 cells (ATCC CRL 1573) were transfected by lipofection as described (42). Yeast DNA was prepared at high density in agarose plugs (Ϸ4 ϫ 10 9 yeast cells/ml), and the chromosomes were separated by pulsed-field gel electrophoresis according to the following protocol: 0.8% low-melting agarose gel (Sea Plaque GTG), 0.5 ϫ TBE buffer, 30 -60 s switching time, 6 V/cm, 110°field angle, 24 h running time. The YAC band was excised from the gel and dialyzed overnight at 4°C in 20 mM Tris-HCl, 1 mM EDTA, 100 M spermine (Sigma), pH 7.6. On the day of the transfection, the agarose slice was divided into four 1.5-ml pieces and transferred to a 15-ml polystyrene tube. Poly-L-lysine (Sigma) was added to each slice at a final concentration of 4 g/ml. The gel slices were melted at 65°C and incubated with 10 units of ␤-agarase (New England Biolabs, Beverly, MA) each at 40°C for 90 min. After the ␤-agarase treatment, 50 l (100 g) of LipofectAMINE reagent (Life Technologies, Inc.) was added to the DNA, gently mixed, and incubated at room temperature for 30 min. The lipid-DNA complex was supplemented with 500 l of 10 ϫ Dulbecco's modified Eagle's medium in a final volume of 5 ml using Opti-MEM (Life Technologies, Inc.) and applied to a 10-cm dish with 90% confluent HFK 293 cells. The cells were incubated with the transfection complex for 8 h at 37°C. Transfection was stopped by washing the cells twice with Opti-MEM and changing to fresh Dulbecco's modified Eagle's medium containing 10% fetal calf serum. The cells were grown for 48 h and then split into selective medium containing 300 g/ml G418 (Life Technologies, Inc.). G418 selection was performed for at least 2 weeks before single colonies of approximately 5 mm diameter were transferred into 96-well plates using cloning cylinders. Cells were expanded in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and 150 g/ml G418. To test for chromosomal integration of the YAC, the cell lines were grown in the absence of G418 for approximately 30 cell cycles. Those lines that continued to grow after readdition of G418 (300 g/ml) were found to have at least one copy of the YAC stably integrated.
Expression of the Mouse wt1 Gene in Stably Transfected HFK 293 Cells-RT-PCR and RNase protection assays were used to measure WT1 gene expression in HFK 293 cells stably transfected with the YACs. Total RNA was isolated from the G418-resistant colonies with Trizol reagent (Life Technologies, Inc.). After a 15-min incubation with DNase I (Life Technologies, Inc.) 2 g of total RNA was reverse-transcribed using oligo(dT) as a primer (Superscript II, Life Technologies, Inc.). PCR amplification of the cDNA was performed with the following primer pairs: mouse wt1 gene-specific primers from the untranslated region of exon 10 (forward primer: 5Ј-TTCAAAGGACACGACTGTG-GATC-3Ј; reverse primer: 5Ј-CAGCCAGACCTCTGAAATTCTGTAC-3Ј) and an exon 10 primer pair specific for the human WT1 gene (forward primer: 5Ј-TCTAACATTCCCGAGGTCAGCC-3Ј; reverse primer: 5Ј-AT-TCCCCTCCATTTGTGCAAG-3Ј). As an internal control co-amplification of human ␤-actin transcripts was performed. Twenty nine PCR cycles were done on a 480 Thermal Cycler (Perkin Elmer) using 60-s denaturation at 94°C, 120-s primer annealing at 58°C, and 120-s primer extension at 72°C. The PCR products were analyzed on a 1.5% agarose gel.
RNase protection assay (HybSpeed RPA kit, Ambion, Austin, TX) was performed with 30 g of total RNA isolated from stably transfected HFK 293 cell lines using Trizol reagent (Life Technologies, Inc.). Antisense riboprobes with a specific activity of 5 ϫ 10 8 cpm/g were obtained by in vitro transcription with T7 RNA polymerase (MaxiScript T7 kit, Ambion, Austin, TX) from linearized template DNA. The template for the mouse wt1 riboprobe consisted of a 429-bp PCR product from the untranslated region of exon 10 (1918 -2347 bp of the published mouse wt1 cDNA sequence) subcloned into the pCR2.1 vector (Invitrogen, San Diego). The human WT1 riboprobe was transcribed from a 299-bp PCR product of exon 10 (1867-2166 bp of the published cDNA sequence) in pCR2.1. RNase protection assays were done according to the protocol supplied by the manufacturer. The protected RNA fragments of 299 nucleotides (human WT1) and 211 nucleotides (mouse wt1) were separated on a 8% denaturing polyacrylamide gel. The gel was dried on a gel drier (Bio-Rad) and autoradiographed. The relative intensities of the hybridization signals were determined by densitometry scanning of the autoradiographs (Molecular Dynamics).
Introducing Deletions into the YAC-Deletions were targeted into the YAC (Y 620 mWT1neo R ) by homologous recombination in yeast. The 1.4-kb 5Ј-EcoRI fragment that cross-hybridized at high stringency to a DNase I-hypersensitive site and matrix attachment region in the human WT1 gene (56R13-5) was deleted from the YAC by a transplacement technique (44). A 5-kb XbaI fragment encompassing the 1.4-kb EcoRI piece was first subcloned from the mouse P1 clone 1407 (Genome Systems, St. Louis, MO) into the yeast integration plasmid pRS 406 (Stratagene, La Jolla, CA). The 1.4-kb EcoRI fragment was cut out from the insert that was blunt-end religated thus leaving at least 1.5 kb of flanking sequence on either side of the 1.4-kb deletion. Lithium acetate transformation of the YAC containing yeast cells was performed with 3 g of linearized (SacI) and gel-purified plasmid as described above. "Pop-in" transformants were selected on SC Trp Ϫ /Lys Ϫ /Ura Ϫ plates, picked, and grown to saturation in liquid medium. Genomic DNA was prepared from these cultures and screened by PCR for integration of the fragmentation plasmid. PCR was done with a primer pair which flanked the 1.4-kb EcoRI deletion. Transformants that had incorporated the deletion vector were grown for "pop-out" in SC Trp Ϫ /Lys Ϫ liquid medium and plated onto SC Trp Ϫ /Lys Ϫ plates supplemented with 1 mg/ml fluoroorotic acid (Sigma). Colonies growing on these plates were replica-plated onto SC Trp Ϫ /Lys Ϫ and SC Trp Ϫ /Lys Ϫ /Ura Ϫ plates, respectively. Transformants that grew on SC Trp Ϫ /Lys Ϫ plates but that did not grow on SC Trp Ϫ /Lys Ϫ /Ura Ϫ plates were further analyzed. Genomic DNA was prepared from these colonies, run on a pulsed-field gel, and transferred to nylon membranes (Boehringer Mannheim) using 10 ϫ SSC as a transfer buffer. After UV cross-linking and a 30-min pre-hybridization at 65°C, the DNA was probed with the 1.4-kb EcoRI fragment. Membranes were washed for 40 min at 65°C with 0.1 ϫ SSC, 0.1% SDS and exposed on film at Ϫ80°C with intensifying screens for 24 h.
Absence of major fragmentations/deletions was suggested by the appearance of specific PCR products.
Transient Transfection of HFK 293 and HeLa Cells-Transient transfections were done by lipofection as described above. DNA from the same preparation was used to transfect simultaneously HFK 293 and HeLa cells. 48 h after transfection the cells were harvested and total RNA prepared. Human and mouse WT1 gene expressions were determined in transiently transfected cell lines by RT-PCR (34 cycles) using the above primers. As a positive control for gene expression from the transfected YAC, we amplified transcripts of the neomycin resistance gene using the following primers: 5Ј-AACAAGATGGATTGCACG-CAG-3Ј (forward primer) and 5Ј-TGCAGTTCATTCAGGGCACC-3Ј (reverse primer).

RESULTS
The long term goal of our experiments is to identify cisregulatory elements residing in the WT1 locus that faithfully reproduce the tissue and developmental pattern of WT1 gene expression. We began by utilizing DNase I-hypersensitive site mapping as an unbiased approach to uncover such elements. This technique allows for the identification of regions of "naked" chromatin, existing within the highly ordered nucleosome structure. These regions have been shown to bind tissue-specific transcription factors (45,46) and contain locus control regions (47,48) and nuclear matrix attachment sites (49,50). It is thought that a gene has to be accessible for binding by numerous protein factors to become transcriptionally active. DNase I-hypersensitive site mapping allows one to quickly scan a gene segment for regions of open chromatin, thus identifying potential functional sites.
As a prerequisite to mapping the entire human WT1 locus for DNase I-hypersensitive sites, a series of probes unique to the WT1 gene and devoid of repetitive sequences had to be identified. Three cosmid clones, L156, L109 and L159 (7), which spanned the entire human WT1 locus of approximately 85 kilobases (kb) of DNA were obtained from Dr. Daniel Haber. To identify DNA fragments to use as probes for DNase I-hypersensitive site mapping, a simple method was devised. 3 Three regions were identified in this manner suitable for DNase I-hypersensitive site mapping of the first intron and the 5Јflanking region of the WT1 gene (Fig. 1B).
DNase I-hypersensitive Sites-A total of 11 hypersensitive sites were identified, some of which were tissue-restricted in their appearance ( Fig. 1 and Table I). Most notably, the promoter region of those cell lines that expressed WT1 mRNA (HFK 293, CEM, and K562) yielded bands corresponding to DNase I-hypersensitive regions, whereas those cell lines negative for WT1 expression (HeLa, TK10) failed to show DNase I hypersensitivity ( Fig. 1 and Table I). HeLa cell nuclei exhibited five DNase I-hypersensitive sites in the far upstream region (HS I to HS V in Fig. 1). Although negative for WT1 mRNA by Northern blotting, HeLa cells showed low levels of WT1 expression after 40 cycles of RT-PCR (data not shown). For comparison, no DNase I-hypersensitive sites were detectable in TK10 cells that did not show WT1 transcripts even after 40 PCR cycles (data not shown). We also identified two regions of DNase I super hypersensitivity in the far upstream region (HS I and HS III in Fig. 1) most easily detected in nuclei from CEM cells. These sites were evident with the minimal amount of DNase I added, and surprisingly, in some experiments, they were strongly present even in untreated nuclei, presumably due to activation of endogenous DNases by the addition of reaction buffer.
The 56R13-5 region also contained a DNase I-hypersensitive site (HS II in Fig. 1) and may overlap with the region of DNase I super hypersensitivity (HS III in Fig. 1). Due to the resolution of the technique of DNase I-hypersensitive site mapping and the size of the WT1 locus, individual hypersensitive sites were "placed" plus or minus approximately 500 bp. Another interesting feature of this analysis was the observation that 56R13-12 bound to nuclear matrix prepared from 293 cells but not to matrix obtained from HeLa cells, which do not express WT1 at a significant level (Fig. 2, see below). DNase I-hypersensitive site VI (HS VI) was located within 56R13-12 (Fig. 1) and was present in those cells lines expressing a moderate to minimal amount of WT1 message (HFK 293 and K562) but was absent from both the non-expressing cell lines (HeLa, TK10) and CEM cells which exhibit robust levels of WT1 message (Table I). This region may be a target for binding of a negative regulatory molecule after the fragment is made accessible and transcriptionally competent. The absence of HS VI from HeLa and TK10 cells may reflect the fact that both cell lines are negative for WT1 expression, and therefore the chromatin at this region is in a closed configuration.
We next examined the possible role of these DNase I-hypersensitive sites in WT1 transcriptional regulation. A series of constructs were made to test whether these sites conferred the tissue-restricted expression pattern of the WT1 gene to a heterologous promoter assayed in tissue culture. Serial deletions of a WT1-reporter construct containing up to 24 kb of 5Ј-flanking sequence had significant effects on gene expression suggesting enhancer function for at least some of the DNase I-hypersensitive sites (35). None of the elements, however, tested both alone and in combination, proved to be specific for renal FIG. 1. A, DNase I-hypersensitive sites mapping to the first intron (HS X) and to the upstream region (HS I to HS IX) in the WT1 gene. Nuclei from HFK 293, K562, HeLa, and CEM cells were treated with increasing amounts of DNase I from left to right, digested with EcoRI, and then probed with non-repetitive sequence from the human WT1 gene. DNase I-hypersensitive sites I, II, III, and VI were identified using the 1.4-kb EcoRI fragment 56R13-5 (B) as a probe. DNase I-hypersensitive sites VII to X were obtained by hybridizing with the 3.4-kb SacI fragment 56S34 (B). Arrows indicate DNase I-hypersensitive sites present in WT1 expressing HFK 293, K562, and/or CEM cells but not in non-expressing HeLa cells. B, distribution of DNase I-hypersensitive sites (HS, downward arrows) in the first intron (HS X and HS XI) and 5Ј-flanking sequence (HS I to HS IX) of the human WT1 gene. Black boxes indicate non-repetitive DNA sequence unique to the human WT1 gene (see "Experimental Procedures"). The restriction map is adapted with modifications from Tadokoro et al. (69).

TABLE I Summary of DNase I-hypersensitive site data in various WT1 expressing (ϩ) and non-expressing (Ϫ) cell lines
ϫ/Ϫ refers to the presence/absence of DNase I-hypersensitive sites in the human WT1 gene. No DNase I-hypersensitive sites were detectable in the first intron and 5Ј-flanking region of the WT1 gene in non-expressing renal TK10 cells. HeLa cells while being negative in the first intron and proximal promoter showed 5 DNase I-hypersensitive sites (HS I to HS V) in the far upstream region of the WT1 gene. The significance of these DNase I-hypersensitive sites has yet to be determined but may reflect the fact that low levels of WT1 transcripts were detectable in HeLa cells after 40 cells. 4 In addition to transcriptional control, DNase I-hypersensitive sites may mediate other functions such as insulators (51,52), locus control regions (47,48), or nuclear matrix attachment sites (49,50). We therefore asked whether these elements could mediate attachment of DNA in the WT1 locus to nuclear matrix. To this end a matrix attachment assay was performed using as probes regions surrounding the DNase I-hypersensitive sites identified above. Matrix Attachment Regions-Nearly the entire WT1 locus was scanned for its ability to bind nuclear matrix. Nuclear matrix was prepared from 293 and HeLa cells and probed with end-labeled WT1 genomic fragments. Reactions were separated into pellet and supernatant fractions and run on analytical agarose gels. Several sites proved to bind to the nuclear matrix in vitro as evidenced by the probe partitioning into the matrix pellet (Fig. 2). Two fragments were used as positive controls, SAR 800 (39) and MAR-H (40), obtained from the interferon-␤ and immunoglobulin heavy chain, respectively. Nearly all of the radioactivity remained in the pellet (Fig. 2). Using pBluescript as a negative control, most of the radioactivity was found in the supernatant and was not associated with the pellet (Fig.  2). Two WT1 fragments, designated 56R13-5 and 56R13-12, bound to the matrix (Fig. 2). Data on 56R13-12 will be described elsewhere. 56R13-5 was digested with XhoI that cuts the fragment into two pieces. The site of attachment to the nuclear matrix could be localized to a smaller subregion. When sequenced, this region proved to be AT-rich, with AT content over approximately 300 bp being as high as 70% (Fig. 4). AT richness is consistent with published nuclear matrix attachment sites (39,49).
P1 Clones-To test for sequence conservation in the human and mouse WT1 gene, we used a non-repetitive 1.4-kb EcoRI fragment (56R13-5) that was located approximately 15 kb upstream of the transcription start site in the human WT1 gene as a probe. We chose this fragment because it contained a matrix attachment region that was particularly sensitive to DNase I. P1 clone 1407 was digested to completion with several restriction enzymes, and a Southern blot of the digest was hybridized at increasing stringency with 56R13-5 as a probe. As shown in Fig. 3, a single band of comparable intensity was seen in each lane after a high stringency wash (0.1 ϫ SSC, 0.1% SDS, 65°C) and a 5-day exposure on film. Notably, the human WT1 probe hybridized to the more upstream P1 clones (P 1405 to P 1408) only but not to the downstream clones (P 1409 to P 1411, see "Experimental Procedures"). The Ϸ6-kb SacI fragment from P1 clone 1407 detected with the human probe 56R13-5 ( Fig. 3) was subcloned into pBuescriptIIKS ϩ for further analysis. This plasmid was digested with a variety of restriction enzymes and reprobed at high stringency with 56R13-5. The smallest band obtained was a 1.4-kb EcoRI fragment which mapped to a position approximately 15 kb upstream of the transcription start site in the mouse gene. This fragment was subcloned into pBluescript, sequenced, and compared with the full-length sequence of the human 56R13-5. The sequence of the human fragment is shown in Fig. 4. The human and mouse WT1 genes both contained an AT-rich (Ϸ70%) region at the same position. Moreover, there were 20 -30 base pair regions of exquisite (Ͼ80%) sequence conservation in addition to more extended stretches (Ͼ60 bp) of lower (Ͻ40%) homology. A Blast search of this 1.4-kb upstream region indicated no significant homology to any published sequence.
YAC Clones-PCR screening of a mouse super pool YAC library (Genome Systems, St. Louis, MO) provided a clone of approximately 620 kb in size. This YAC was designated Y 620 mWT1. A rough restriction mapping of the YAC was done with probes from the proximal promoter and exon 10 of the mouse wt1 gene and with probes for both vector arms. The YAC clone contained the entire wt1 structural gene (Ϸ100 kb) on a SacII fragment that was flanked by 240 and 300 kb of 5Ј-and 3Ј-sequence, respectively. The 5Ј-to 3Ј-orientation of the wt1 gene was toward the centric vector arm (Fig. 5). Fluorescent in FIG. 2. Identification of two MARs contained within 56R13-5 and 56R13-12 (Fig. 1) in HFK 293 cells. Note that the Bg 4.1-kb fragment that was obtained from the far upstream region of the human WT1 gene (see Fig. 4) is essentially negative in this assay. The positive controls are the matrix attachment regions (MAR-H) from the immunoglobulin heavy chain enhancer and the scaffold attachment region (SAR 800 ) from the interferon-␤ gene. The Bluescript plasmid (pBS) served as a negative control. P indicates the pellet fractions, i.e. DNA that has attached to the pellet, and S refers to the supernatant. The size of the fragments is not of relevance but depends on the enzyme used for the preparation of the nuclear matrix and probes (see "Experimental Procedures").

FIG. 3. Cross-hybridization at high stringency between human
and mouse WT1 genomic DNA sequence. The mouse P1 clone 1407 (Fig. 5) containing wt1 upstream sequence was digested to completion with several restriction enzymes and hybridized by the Southern technique with 56R13-5 (Fig. 1) as a human probe. The membrane was washed at 65°C in 0.1 ϫ SSC, 0.1% SDS and exposed to x-ray film with intensifying screens for 5 days. A single band of similar intensity was detectable in each lane. The Ϸ6-kb SacI fragment indicated by asterisks was subcloned into pBluescript and re-hybridized with 56R13-5. As a result a 1.4-kb EcoRI fragment was obtained which was compared with the full-length sequence of human 56R13-5.
situ hybridization analysis (Genome Systems, St. Louis, MO) was used to determine the chromosomal localization of the YAC in the mouse genome. For this purpose YAC DNA was hybridized to metaphase chromosomes from mouse embryonic fibroblasts. A total of 80 metaphase cells were analyzed with all of them exhibiting specific labeling of a large sized chromosome. No evidence for any additional chromosomal location hybridizing with the YAC DNA was obtained. On the basis of 4,6diamidino-2-phenylindole staining and co-hybridization with a chromosome-specific probe, the YAC clone was found to map to a position on chromosome 2 that was 52% of the distance from the heterochromatic-euchromatic boundary to the telomere. This area corresponded to band 2E1 where the endogenous mouse wt1 gene is located (data not shown).
For positive selection in mammalian cells a neomycin resistance gene was targeted into the URA3 site of the acentric YAC vector arm utilizing homologous recombination in yeast. The "retrofitted" YAC (Y 620 mWT1neo R ) was transfected into a human fetal kidney cell line (HFK 293) by a lipofection method. These cells were known to express endogenous WT1. After 3 weeks of G418 selection (300 g/ml) single colonies of Ϸ5 mm diameter were cloned and expanded. A total of 16 stable HFK 293 cell lines could be established as a result of two transfection experiments using YAC DNA from different preparations each time.
Mouse wt1 gene expression in stably transfected 293 cells was measured by RT-PCR and RNase protection assay. Using specific PCR primer pairs from the untranslated region of exon 10, 12 out of 16 stable 293 cell lines were found to express the mouse wt1 gene in addition to the endogenous human gene (data not shown). To quantitate the WT1 mRNA in the transfected cells, we used RNase protection assay with antisense riboprobes for exon 10 of the human and mouse WT1 gene. As shown in Fig. 6, three representative cell lines had similar amounts of the endogenous human WT1 mRNA. The level of mouse wt1 expression was found to be comparable in the transfected lines and amounted to approximately 30% of the level of expression of the endogenous human WT1 gene (Fig. 6).
The copy number of the transfected YACs was determined by Southern hybridization of genomic DNA with a 1.7-kb BamHI/ PvuII fragment from pBR322 as a probe for the acentric YAC vector arm. It turned out that most if not all examined cell lines had a single copy of the YAC incorporated in their genome (data not shown). Therefore, the theoretical maximum level of mouse wt1 expression would be 50% compared with the endog-enous human WT1 mRNA.
To determine whether wt1 expression of the YAC was cell type-specific, we also transfected HeLa cells that are negative for WT1 mRNA by Northern blotting. In three independent transfection experiments, we could not establish a single stable cell line possibly due to the very low level expression of the neomycin resistance gene in HeLa cells (Fig. 7B). As an alternative approach, we performed transient transfections simultaneously of HFK 293 and HeLa cells. Using YAC DNA from the same preparation for the transfection of both cell lines, HFK 293 but not HeLa cells were found by RT-PCR (34 cycles) to express the mouse wt1 gene (Fig. 7A). Low amounts of neomycin resistance gene transcripts could be detected in HeLa cells after 34 PCR cycles indicating transcriptional activity of the transfected Y 620 mWT1neo R (Fig. 7B). Although the YAC could not be tested in the chromosomal context of a non-WT1 expressing cell line, the results obtained with transiently transfected 293 and HeLa cells strongly suggest a cell type specific regulation of the mouse wt1 (trans)gene.
To test whether the 1.4-kb EcoRI mouse fragment that crosshybridized to the human 56R13-5 (Figs. 3 and 4) contained cis-regulatory elements required for wt1 gene expression, we deleted this region from Y 620 mWT1neo R using homologous recombination in yeast. The deletion construct consisted of a pRS406 vector (Stratagene, La Jolla, CA) which had Ϸ1.5 kb of flanking sequence on either side of the targeted 1.4-kb EcoRI piece subcloned into the XbaI site. Yeast transformants were screened by PCR and Southern hybridization of genomic DNA. Using the 1.4-kb EcoRI fragment from the mouse wt1 gene as a probe, one out of 6 clones was found to have the expected deletion (data not shown). A single band of the same size (Ϸ620 kb) as the original (undeleted) Y 620 mWT1neo R was observed when the DNA blot was re-hybridized with a specific probe for the centric vector arm suggesting that the YAC was largely intact and no fragmentation had occurred during the transformation procedure (data not shown). Genomic DNA from this YAC clone, designated Y 620 mWT1neo R ⌬EcoRI, was prepared in agarose plugs and separated on a pulsed-field gel. The YAC band was excised from the gel and transfected into HFK 293 cells. Three transfection experiments using YAC DNA from two different preparations provided a total of 11 stable cells lines. RT-PCR was performed with 2 g of total RNA from stably transfected cells using human and mouse WT1 gene-specific primer pairs. Transcripts of the endogenous human WT1 gene were detectable in all cell lines after 34 PCR cycles (Fig. 8). However, we could not detect significant levels of mouse wt1 mRNA in any of the 11 stable lines that had the 1.4-kb upstream element missing on the YAC (Fig. 8). To confirm this result, we performed another transfection experiment with Y 620 mWT1neo R ⌬EcoRI in which we measured WT1 mRNA in pools of stable 293 cell lines. Again, after 34 cycles of RT-PCR, significant amounts of mouse wt1 mRNA could not be detected, whereas the endogenous human WT1 gene was normally expressed (Fig. 8). To ensure that the transfected YAC DNA was largely intact and no major fragmentation and/or deletion had occurred during the lipofection procedure, we performed repeated PCR on genomic DNA from the transfected cell lines. Specific PCR primer pairs were used to amplify 90 -350 base pair stretches of DNA from both YAC vector arms and from exon 10, exon 6, and a 15-kb upstream region in the mouse wt1 gene. The expected PCR products were obtained with at least 8 out of 11 lines transfected with Y 620 mWT1neo R ⌬EcoRI suggesting that the transfected YAC was largely intact. DISCUSSION Studies on the regulation of WT1 are important for several reason. First, since WT1 is expressed in induced blastema and in podocytes, these studies may provide new insights into some of the earliest transcriptional signals mediating kidney development and into podocyte-specific gene expression. Second, since WT1 has been suggested to play a role in mesenchymalepithelial conversion (25,30), regulators of WT1 might be involved in epithelial cell differentiation as well. Finally, WT1 regulation studies may reveal molecular events leading to the formation of Wilms' tumors. In an effort to study the transcriptional control of WT1, we and others (34 -37) have previously localized several regulatory elements on the gene. None of the cis-elements identified so far, however, have mimicked the tissue-specific expression pattern of WT1. Likewise, transgenic mice harboring a Ϫ1.9 to ϩ0.2 kb WT1-lacZ construct did not express the transgene in a tissue-restricted fashion. 2 We therefore reasoned that additional, possibly far upstream and/or downstream elements, might be important for WT1 gene regulation. In this study we used human fetal kidney 293 cells and various other WT1 expressing and non-expressing cell lines to map DNase I-hypersensitive sites and matrix attachment regions in the WT1 gene locus.
We report the following novel findings. 1) A total of 11 DNase I-hypersensitive sites were identified in the first intron and 5Ј-flanking region of the human WT1 gene. Six of the discov-ered DNase I-hypersensitive sites were found in WT1 expressing cells only suggesting a role in cell type-specific gene regulation. 2) A 1.4-kb fragment of the human WT1 gene, residing approximately 15 kb upstream of the transcription start site, which contained DNase I-hypersensitive sites and also bound to the nuclear matrix in vitro, showed significant sequence conservation with an upstream fragment of similar size from the mouse wt1 gene. 3) Stable transfection of a Ϸ620-kb YAC clone carrying the mouse wt1 locus into human fetal kidney cells resulted in (trans)gene expression at a level comparable to the endogenous WT1 gene. 4) Deletion of this 1.4-kb 5Ј-element from the YAC caused a loss of wt1 gene expression in human fetal kidney cells. We have therefore identified a novel cisregulatory region in the mouse wt1 gene critical for expression in renal cells.
DNase I-hypersensitive sites occur in the chromatin of transcriptionally competent genes probably due to a local relaxation of chromatin organization which allows the binding of protein factors, a subset of which may be tissue-specific transcription factors (53,54). Occasionally DNase I-hypersensitive sites are located within AT-rich matrix attachment regions (MARs). MARs are thought to anchor the DNA to the nuclear matrix, and they may also insulate the DNase I-hypersensitive sites from the influence of outside sequences (49,55). MARs have also been found to be responsible for the tissue-restricted expression of immunoglobulin chains in a transgenic setting (40). Moreover, several transcription factors have been shown to be localized to the nuclear matrix (56,57).
None of the DNase I-hypersensitive sites and/or MARs that we have identified in the first intron and 5Ј-flanking region of the WT1 gene had a tissue-specific effect on the transcriptional activity of transiently and stably transfected reporter constructs both in the context of the natural WT1 minimal promoter and an SV40 promoter. These initial results were not completely unexpected. Although DNase I-hypersensitive regions have been found to correlate very closely with transcription in a number of genes studied so far, the complicated pattern of tissue-specific expression of the WT1 gene may require additional elements that were missing from these smaller constructs. For example, the developmental regulation of the human ␤-globin gene relies on a locus control region at the 5Ј-end of the ␤-globin cluster. The locus control region consists of four subdomains that exhibit exquisite hypersensitivity to DNase I in erythroid but not in non-erythroid cells (47,58,59). An upstream DNase I-hypersensitive site corresponding to a distal enhancer element was also required for the tissue-specific expression of hepatocyte nuclear factor 4 in transgenic FIG. 5. Schematic of the mouse and human WT1 loci and the genomic mouse YAC and P1 clones used. The original YAC clone (Y 620 mWT1) has been retrofitted with a neomycin resistance gene cassette (␤geoLys) allowing positive selection in mammalian cells. Extensive mapping of the retrofitted YAC, designated as Y 620 mWT1neo R , was performed with a variety of restriction enzymes (data not shown). Y 620 mWT1neo R contained the mouse wt1 structural gene on a Ϸ100-kb SacII fragment flanked by 240 and 300 kb of 5Ј-and 3Ј-sequence, respectively. The 5Ј-to 3Ј-orientation of the mouse wt1 gene on the YAC was toward the pYAC4 centric vector arm. Eight genomic DNA clones (P 1405 to P 1412) containing different segments of the wt1 gene were obtained by PCR screening of a mouse P1 library (see "Experimental Procedures"). The diagram is not drawn to scale.
FIG. 6. WT1 expression determined by RNase protection assay in three representative 293 cell clones stably transfected with Y 620 mWT1neo R . The assays were performed with 30 g of total RNA from the stable 293 clones and with 20 g of total RNA from adult mouse kidney and HFK 293 cells as controls. A mix of specific riboprobes from the untranslated region of exon 10 of the human and mouse WT1 gene was used for hybridization (see "Experimental Procedures"). Both riboprobes had roughly the same uridine content of approximately 28% and were labeled to the same specific activity of 5 ϫ 10 8 cpm/g. Note that the level of mouse wt1 expression is approximately the same in different stable clones and amounts to approximately 30% of the endogenous human gene. The relative intensities of the hybridization signals were determined by densitometry (Molecular Dynamics). ntds, nucleotides.
animals (60). Recent findings indicate that the transcriptional core of a 3Ј-enhancer in the mouse immunoglobulin kappa gene becomes DNase I-hypersensitive during early B cell development (61). These and other studies provide strong evidence for a functional role of DNase I-hypersensitive sites in gene transcription. For a number of other genes, cis-regulatory elements located far away from the transcription start site have been found. Far upstream enhancers between Ϫ13.5 and Ϫ19.5 kb, for example, were necessary for high level expression of the mouse pro-␣2(I) collagen gene (62). Recent mutation analysis and transgenic studies indicated a region approximately 150 kb downstream important for PAX 6 transcription (63). It is conceivable that a far upstream and/or downstream sequence is also required for the regulation of the WT1 gene that is located next to PAX 6 on the human chromosome 11p13. We reasoned that these proposed sites were absent from the smaller constructs and would therefore not allow us to identify those elements sufficient for WT1 tissue-specific expression.
To incorporate large DNA fragments for transfection and to manipulate these fragments, we resorted to YAC wt1 clones and initially focused on showing that an intact YAC gives wt1 (trans)gene expression in a renal cell line expressing endogenous WT1. Then we deleted a small (1.4 kb) region of DNase I hypersensitivity and exquisite sequence conservation between the human and mouse WT1 gene to ask whether it was necessary for wt1 expression. For these purposes we used an approximately 620-kb YAC carrying the entire mouse wt1 locus. Compared with reporter constructs the use of a YAC clone offered the following advantages. First, YACs allow one to introduce genes into tissue culture cells and animals in a more natural chromosomal context. Second, gene expression from YACs in transgenic animals and transfected cells normally occurs in a copy number-dependent and integration site-independent way (64). Third, YACs are relatively easy to manipulate (introducing deletions, insertions, etc.) by homologous recombination in yeast. We have chosen a mouse YAC (instead of a human), because most WT1 expressing cell lines are derived from human tissue. Transfection of the mouse wt1 gene into human cells allowed us to quantitate "transgene" expression directly by an RNase protection assay with a mouse wt1-specific riboprobe. The use of YACs in gene regulation studies is limited by transfection efficiency into mammalian cells which is significantly lower than for regular plasmid DNA (65). We have therefore taken highly transfectable HFK 293 cells in which we introduced a YAC carrying the mouse wt1 gene by the lipofection method (42). Twelve of the 16 stable cell lines expressed the mouse wt1 gene at a level of at least 30% of the endogenous human gene (Fig. 6). Cell lines in which we could not detect mouse wt1 mRNA had portions of the structural gene missing likely due to DNA shearing during the transfection (data not shown). These results encouraged us to utilize stably transfected human fetal kidney cells as a model to narrow down cis-regulatory elements on the mouse wt1 gene.
To this end we examined whether one of the DNase I-hypersensitive sites, although not sufficient, might still be required for wt1 gene expression in HFK 293 cells. We focused our initial studies on a 1.4-kb EcoRI fragment located approximately 15 kb upstream of the transcription start site in the mouse wt1 gene. The reasons for this were as follows. First, the corresponding sequence in the human gene was found to be hypersensitive to DNase I suggesting a transcriptionally active site (Fig. 1). Second, this human analogue bound to the nuclear matrix in vitro (Fig. 2). Third, a sequence comparison of this region revealed significant homology between the human and mouse gene. In particular, we found 20 -30 base pair stretches FIG. 7. a, WT1 expression in transiently transfected (Y 620 mWT1neo R ) HFK 293 and HeLa cells as evaluated by RT-PCR. The same human and mouse WT1-specific primers were used as for RT-PCR of the stable 293 cell clones (Fig. 6). The RNA samples were incubated with DNase I to avoid co-amplification of genomic DNA. Ϯ RT refers to the presence and absence of reverse transcriptase. An equimolar mix of human and mouse WT1 cDNA plasmids was used as a positive control. After 34 PCR cycles mouse wt1 transcripts were readily detectable in transient HFK 293 but not in HeLa cells that did not express the endogenous WT1 gene. Occasionally, two faint bands were obtained by RT-PCR of transient HeLa cells. These bands were of different size than the RT-PCR products for human and mouse WT1 and did not hybridize with human and mouse WT1 cDNA, respectively, suggesting co-amplification of unspecific (non-WT1) transcripts. b, expression of the neomycin resistance gene from Y 620 mWT1neo R in transiently transfected HeLa cells. RNA samples were treated with DNase I. Ϯ RT indicates the presence/absence of reverse transcriptase. Specific primers for the neomycin resistance gene were used for the PCR reaction (see "Experimental Procedures"). pSV 2 neo was used as a positive control. After 34 PCR cycles a faint signal of Ϸ190 base pairs could be detected which was absent without RT.
FIG. 8. RT-PCR evaluation of WT1 expression in HFK 293 cells stably transfected with Y 620 mWT1neo R ⌬EcoRI. This YAC clone had a 1.4-kb upstream fragment in the mouse wt1 gene deleted which cross-hybridized with the human WT1 sequence 56R13-5 containing a site of DNase I hypersensitivity and a matrix attachment region (Figs. 1 and 2). Notably, none of the 11 clonal isolates, obtained from two independent transfection experiments, had mouse wt1 transcripts detectable after 34 PCR cycles. No significant expression of the mouse wt1 transgene was found also in a pool of stable 293 cells obtained in an additional transfection experiment. Note that the endogenous human WT1 gene was expressed in all examined clones. of Ͼ80% sequence conservation which were roughly co-linear in the WT1 gene of human and mouse (Fig. 4). Most interestingly, deletion of this 1.4-kb upstream fragment from the YAC resulted in complete loss of mouse wt1 gene expression in stably transfected HFK 293 cells (Fig. 8). These results strongly suggested that a cis-regulatory element in the 1.4-kb EcoRI region was required for transcription of the wt1 gene. There was still a minor chance, however, that accidental fragmentation of the YAC in addition to the targeted deletion had occurred and was responsible for the loss of wt1 gene expression. We addressed this issue by showing that sequence-tagged sites along the entire length of the YAC could be amplified by PCR in at least 8 out of 11 stable cell lines (data not shown). These findings suggested that more than 70% of the transfectants had an intact copy of the YAC stably integrated into their genome. For comparison, a similar percentage of cell lines that had been stably transfected with the original (undeleted) YAC expressed the mouse wt1 (trans)gene at a significant level. Moreover, the 1.4 kb-deleted YAC ran on a pulsed-field gel at roughly the same molecular size as the original YAC indicating absence of major deletions or fragmentation (data not shown).
Notably, the 1.4-kb EcoRI fragment that had no homology to any known sequence was not sufficient for the tissue-specific expression of the wt1 gene but was functioning only in the context of a YAC carrying the entire mouse wt1 gene. These findings together with the observation that the human sequence analogue contained a matrix attachment region are consistent with a role for chromatin organization in WT1 gene regulation. A novel class of gene regulatory elements, termed "facilitators," which have been linked to chromatin unfolding has recently been described. Facilitators, although having no effect on reporter gene expression in transient assays, were essential for copy number proportional and integration siteindependent expression in transgenic animals (66). A lack of facilitator function prevented DNase I hypersensitivity and significantly reduced transcriptional activating function of a T cell-specific enhancer in the human adenosine deaminase gene (66). To date, we have no direct experimental evidence to indicate that the identified upstream element meets the criteria for a facilitator. It is also possible that the critical region we have identified in some way interacts with other cis-elements to activate WT1 gene expression. Cooperative interaction between distant regulatory regions has previously been found to be important for gene transcription. For example, a 3-kb upstream DNase I-hypersensitive site in the gene for the murine stem cell antigen CD34 conferred position-independent gene expression in stably transfected hematopoietic progenitor cells only after addition of 3Ј-downstream sequence (67). A model of mutual interaction between distant enhancer/locus control regions and proximal promoter sequence has also been suggested for the mechanism of chromatin opening in the chicken ␤ Aglobin gene (68).
In conclusion, we have identified a novel cis-element in the mouse wt1 gene required for expression in human fetal kidney cells. This regulatory region is located approximately 15-kb upstream of the transcription start site and contains a sequence which is highly conserved between the human and mouse. The human region is also located Ϸ15-kb upstream of the human WT1 transcription start site, shows DNase I hypersensitivity, and contains a site for attachment of nuclear matrix. Future investigations will further narrow down the 1.4-kb region, identify putative transfactors binding to it, explore its mechanistic role, and assay its importance in transgenic mice in tissue-restricted WT1 expression, both in the kidney and outside.