A non-AUG translational initiation event generates novel WT1 isoforms.

The Wilms' tumor (WT) suppressor gene, WT1, is mutated in a small set of WTs and is essential for proper development of the urogenital system. The gene has three sites of transcriptional initiation and produces mRNA transcripts containing 5′-untranslated regions of more than 350 nucleotides. The mRNA, through two alternative splicing events, is predicted to direct the synthesis of four protein isoforms with molecular masses of 47-49 kDa. In this report, we identify and characterize novel WT1 protein isoforms having predicted molecular masses of 54-56 kDa. Mutational analysis of the murine wt1 mRNA demonstrates that the novel isoforms are the result of translation initiation at a CUG codon 204 bases upstream of and in frame with the initiator AUG. We show that these isoforms are present in both normal murine tissue and in WTs. Like WT1, the larger isoforms localize to the cell nucleus and are capable of mediating transcriptional repression. Our results indicate that regulation of WT1 gene expression is more complex than previously suspected and have important implications for normal and abnormal urogenital system development.

The Wilms' tumor (WT) suppressor gene, WT1, is mutated in a small set of WTs and is essential for proper development of the urogenital system. The gene has three sites of transcriptional initiation and produces mRNA transcripts containing 5-untranslated regions of more than 350 nucleotides. The mRNA, through two alternative splicing events, is predicted to direct the synthesis of four protein isoforms with molecular masses of 47-49 kDa. In this report, we identify and characterize novel WT1 protein isoforms having predicted molecular masses of 54 -56 kDa. Mutational analysis of the murine wt1 mRNA demonstrates that the novel isoforms are the result of translation initiation at a CUG codon 204 bases upstream of and in frame with the initiator AUG. We show that these isoforms are present in both normal murine tissue and in WTs. Like WT1, the larger isoforms localize to the cell nucleus and are capable of mediating transcriptional repression. Our results indicate that regulation of WT1 gene expression is more complex than previously suspected and have important implications for normal and abnormal urogenital system development.
Wilms' tumor (WT) 1 is a embryonal malignancy of the kidney that afflicts one out of every 10,000 children (Matsunaga, 1981). Approximately 7-15% of sporadic WTs contain detectable mutations in the tumor suppressor gene WT1 (Coppes et al., 1993;Little et al., 1992;Varanasi et al., 1994). The WT1 gene encodes a transcription factor belonging to the early growth response family of Cys 2 -His 2 zinc finger proteins. The pre-mRNA is alternatively spliced at two exons to produce four WT1 isoforms . The first alternatively spliced exon inserts or removes 17 residues upstream of the four zinc fingers and is capable of mediating transcriptional repression when fused to a DNA binding domain (Wang et al., 1993a). The second alternatively spliced exon inserts or removes 3 amino acids, KTS, between zinc fingers III and IV and alters the DNA binding specificity of the protein, as well as its subnuclear localization Larrson et al., 1995;Rauscher et al., 1990). The ϪKTS WT1 isoforms recognize a GC-rich motif (5Ј-GCGGGGGCG-3Ј), as well as a (TCC) n repeat (Rauscher et al., 1990;Wang et al., 1993b), and can affect expression of a number of genes harboring these motifs in their regulatory regions (for a review see ). These genes include insulin-like growth factor II (Drummond et al., 1992), insulin-like growth factor 1 receptor (Werner et al., 1993), platelet-derived growth factor A-chain (Gashler et al., 1992;Wang et al., 1992), colony-stimulating factor-1 gene (Harrington et al., 1993), transforming growth factor-␤1 (Dey et al., 1994), the retinoic acid receptor-␣ (Goodyer et al., 1995), and the wt1 gene itself . WT1 can mediate both transcriptional repression and activation, depending on the architecture of the promoter under study (Madden et al., 1991;Drummond et al., 1992Drummond et al., , 1994Wang et al., 1992). Activation and repression are mediated by distinct domains within the WT1 protein (Wang et al., 1993a).
In addition to its role as a tumor suppressor gene, WT1 plays an essential role in the normal development of the urogenital system. The expression pattern of WT1 is not ubiquitous, being mainly restricted to components of the urogenital system: the gonads, developing glomeruli, and the uterus (Pelletier et al., 1991a). Many children with germline WT1 mutations suffer from malformations of the urogenital system, ranging in severity from minor genital anomalies to streak gonads and renal nephropathy (for a review see Bruening and Pelletier (1994)). Consistent with a role for WT1 in the development of the urogenital system is the observation that this system fails to differentiate in wt1-null mice (Kreidberg et al., 1993).
Similar to many other genes involved in growth regulation, the wt1 mRNA transcript contains an AUG-initiated open reading frame (ORF) preceded by a long, GC-rich, 5Ј-untranslated region (UTR) (Nagpal et al., 1992). There are three sites of transcription initiation within the murine wt1 promoter, producing mRNA species with 5Ј-UTRs of 375, 700, or 720 nucleotides. In the course of characterizing the protein isoforms produced from these transcripts, we noted the presence of isoforms having molecular masses greater than expected. In this report, we demonstrate that these isoforms arise from translation initiation at a CUG codon upstream of and in frame with the wt1 initiator AUG. These novel isoforms are present in normal and malignant tissue and are capable of repressing transcription.
Although AUG codons are essentially exclusively used as initiation codons for eukaryotic mRNAs, there are rare examples of cellular mRNAs where other codons (GUG, ACG, and CUG) are also used for this purpose. These include protooncogenes (MYC, INT-2, PIM-1, and LYL-1) (Acland et al., 1990;Hann et al., 1988;Saris et al., 1991;Mellentin et al., 1989), as well as the basic fibroblast growth factor gene (Prats et al., 1989), retinoic acid receptor ␤4 (Nagpal et al., 1992), krox-24 (also a member of the early growth response family) (Lemaire et al., 1990), and the ltk receptor (Bernards and de la Monte, 1990). The nature of the signals that dictate the use of a CUG codon as an initiation codon within the 5Ј-UTR are not well understood, but immediate downstream sequences can influence the efficiency with which CUG codons are selected (Boeck and Kolakofsky, 1994;Grü nert and Jackson, 1994). Our results show that WT1 gene expression is more complex than previously suspected and that the activity of these novel WT1 isoforms needs to be considered in biological assays involving WT1.

EXPERIMENTAL PROCEDURES
Cell Lines-TM3 and COS-1 cells were obtained from the American Type Culture Collection (ATCC). TM3 cells were maintained in a 1:1 mixture of Dulbecco's modified Eagle's medium and Ham's F-12 medium supplemented with 5% horse serum and 2.5% fetal calf serum. COS-1 cells were maintained in Dulbecco's modified Eagle's medium (high glucose) supplemented with 10% fetal calf serum.
Northern Blots-Total RNA was isolated from mouse testis and TM3 cells by the LiCl 2 /urea procedure (Auffray and Rougeon, 1980). 10 g of RNA was electrophoresed into a 1.2% agarose/37% formaldehyde gel and blotted to nitrocellulose (Schleicher & Schuell). Filters were probed with a 32 P-labeled murine wt1 cDNA as described previously (Pelletier et al., 1991a).
Immunoprecipitations and Western Blots-Crude nuclear extracts were prepared by resuspending washed cells into an equal packed cell volume of 10 mM Hepes, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 0.1% Nonidet P-40, 0.5 mM dithiothreitol, 0.5 g/ml aprotinin, 0.2 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride. Following a 10-min incubation on ice, the suspension was passed ten times through a 23 G needle. The nuclei were pelleted and resuspended in three times the packed cell volume of RIPA buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 0.1% SDS, 0.5% sodium deoxycholate). For Western blots, 30 g of the nuclear extract was fractionated on a 10% SDS-PAGE gel. Following transfer to Immobilon-P (Millipore), the membranes were probed with the indicated antibodies, and proteins were visualized by chemiluminescence using an ECL kit (Amersham Corp.).
For immunoprecipitations, cells were grown to 80% confluency, washed in PBS, and preincubated in methionine-free medium for 1 h. Cellular protein was labeled with 200 Ci/ml of [ 35 S]methionine for 12 h. Following a wash with PBS, the cells were lysed in RIPA buffer. For each immunoprecipitation, approximately 10 9 cpm of lysate was precleared with protein A-Sepharose (Pharmacia Biotech Inc.) and preimmune serum for 1 h at 4°C. Incubation with specific antibody and protein A-Sepharose was then performed for 4 h at 4°C, after which time the immunoprecipitate was washed three times with RIPA buffer and once with PBS. Elution was performed in SDS-PAGE sample buffer. The eluted proteins were fractionated on a 10% SDS-PAGE gel, treated with EN 3 HANCE (New England Nuclear), and detected by exposure to X-Omat film (Kodak).
Antibodies-The anti-WT1 polyclonal antibody C-19 was purchased from Santa Cruz and was generated against a peptide (LVRHHN-MHQRNMTKLQLAL) at the carboxyl terminus of the WT1 protein. The anti-WT1 monoclonal antibodies 13B5 and 8A7 were raised against peptides designed to mimic WT1 isoforms containing or lacking the first alternatively spliced exon. Their characterization has been previously discussed (Mundlos et al., 1993). The rabbit anti-Wt1 polyclonal antibody 605 was raised against a GST-Wt1 fusion protein containing exons 2-9 of the murine wt1 gene. The rabbit polyclonal antibody S3, which recognizes only the CUG-initiated WT1 isoforms, was raised against the multiple antigenic peptide (Posnett et al., 1988) PASTCVPEPAS-QHTLR, located within the unique amino-terminal region produced by the CUG-initiated WT1 isoforms, and present in both mouse and human sequences (Fig. 2). (WT1 isoforms containing or lacking the first alternative splice site will be referred to as ϩ or Ϫ 17 aa in the text.) Plasmid Constructs-Restriction enzymes and other DNA-modifying enzymes were purchased from New England Biolabs. CMV-wt1 and pSP65 constructs lacking the 5Ј-UTR have been previously described (Goodyer et al., 1995;Pelletier et al., 1991b). To construct SP65-wt1 plasmids containing various 5Ј-UTRs, clones isolated from a mouse cDNA library that terminated at various regions within the 5Ј-UTR were used. The 5Ј ends of these clones were excised from pKSIIϩ with EcoRI (which cleaves at the 5Ј end of the cDNA) and NcoI (which cleaves at position 541 downstream of the initiator AUG) and subcloned into the same sites of SP65-wt1. To generate point mutations within the 5Ј-UTR, site-directed mutagenesis was performed using the polymerase chain reaction and overlapping oligonucleotides harboring the desired change (Pelletier et al., 1991b).
In Vitro Transcriptions and Translations-RNA polymerases were purchased from Promega. For in vitro transcriptions, plasmids were linearized with the indicated enzymes, and transcriptions were performed as described previously (Pelletier et al., 1991b). RNA transcripts were quantitated by monitoring incorporation of [ 3 H]CTP (New Eng-land Nuclear). In vitro translations were performed in rabbit reticulocyte lysate using [ 35 S]methionine as directed by the manufacturer (Promega).

RESULTS
The wt1 Gene Encodes Novel 54 -56 kDa Isoforms-To gain better insight into the regulation of wt1 gene expression, we screened a number of mouse cell lines for expression of wt1 mRNA. Of the lines tested, one of these, TM3 (derived from The 47-49-kDa Wt1 isoforms are indicated by a dot and are not well resolved on this gel. The ϳ57-kDa related protein is denoted by a ϫ. Extracts were prepared from untransfected COS-1 cells (lanes 1 and 4), COS-1 cells transfected with CMV/wt1(Ϫ/Ϫ) (lanes 2 and 5), or TM3 cells (lanes 3 and 6). Samples were immunoprecipitated with the antibody 605 (lanes 1-3) or with preimmune serum (lanes 4 -6). Following electrophoresis, the gel was treated with EN 3 Hance (New England Nuclear), dried, and exposed to X-Omat (Kodak) film at Ϫ70°C for 2 days. C, Western blot analysis of nuclear extracts prepared from COS-1 and TM3 cells. Blots were probed with monoclonal antibodies 8A7 (lanes 1-3) and 13B5 (lanes 4 -6), which recognize WT1 isoforms lacking or containing the first alternatively spliced exon, respectively. Extracts were prepared from untransfected COS-1 cells (lanes 1 and 4), COS-1 cells transfected with CMV/wt1(Ϫ/Ϫ) (lane 2), CMV/wt1(ϩ/Ϫ) (lane 5), or TM3 cells (lanes 3 and 6). The dot indicates the position of migration of the 47-49-kDa Wt1 isoforms; whereas the ϫ delineates the position of the ϳ57-kDa isoforms. Blots were exposed to X-Omat film (Kodak) for 2 min. D, expression of Wt1 * in murine ovaries. Extracts were prepared from untransfected COS-1 cells (lane 1), COS-1 cells transfected with CMV/wt1(Ϫ/Ϫ) (lane 2), COS-1 cells transfected with CMV/wt1(ϩ/ϩ) (lane 3), and adult murine ovaries (lane 4). The arrows indicate the positions of migration of the Wt1 isoforms. The blot was probed with the antibody C-19 and exposed to X-Omat (Kodak) film for 5 min. mouse testis), revealed significant levels of wt1 mRNA by Northern blot analysis (Fig. 1A, lane 2). In addition to the expected 3.1-kilobase wt1 mRNA, a second mRNA species of ϳ2.5 kilobases is present in testis (Fig. 1A, lane 1) and in TM3 cells (lane 2). The nature of the 2.5-kilobase cross-hybridizing mRNA species is not well defined but has been previously noted to be testis-specific and may arise from the use of an alternative polyadenylation site within the wt1 3Ј-UTR (Pelletier et al., 1991a).
To detect Wt1 protein from TM3 cells, we made use of the polyclonal anti-Wt1 antibody, 605. This antibody, directed against amino acids 123-299, is capable of immunoprecipitating [ 35 S]methionine-labeled Wt1 from COS-1 cells transfected with a CMV-wt1 expression vector (denoted by a dot in Fig. 1B, lane 2). This polypeptide is not recognized by preimmune serum (Fig. 1B, lane 5), nor is it immunoprecipitated from untransfected COS-1 cells (Fig. 1B, lane 1). In immunoprecipitates from TM3 cells with the 605 antibody, we noted the presence of Wt1 protein (lane 3, denoted by a dot) as well as a polypeptide of ϳ57 kDa (indicated by a ϫ in Fig. 1B, lane 3). Like Wt1, this polypeptide is not visible in immunoprecipitates of TM3 cells with preimmune serum (Fig. 1B, lane 6).
To extend these results and determine whether the ϳ57-kDa polypeptide species was related to Wt1, we made use of two anti-WT1 monoclonal antibodies (Mundlos et al., 1993). Antibody 8A7 specifically recognizes WT1 isoforms lacking the 17 amino acids introduced by alternative splicing of exon 5 (Ϫ17 aa), whereas 13B5 specifically recognizes WT1 isoforms containing this alternatively spliced exon (ϩ17 aa) (Mundlos et al., 1993). On Western blots of COS-1 cells transfected with CMV/ wt1(Ϫ/Ϫ), 8A7 recognizes the Wt1(Ϫ/Ϫ) protein isoform as expected (denoted by a dot, Fig. 1C, lane 2). This polypeptide is not present in untransfected COS-1 cells (Fig. 1C, lane 1). In TM3 cells, a ϳ57-kDa polypeptide is also recognized by this antibody (indicated by a ϫ in Fig. 1C, lane 3). Antibody 13B5 recognizes the Wt1(ϩ/Ϫ) isoform produced in COS-1 cells transfected with CMV/wt1(ϩ/Ϫ) (indicated by a dot in Fig. 1C, lane 5). In TM3 cells, a ϳ57-kDa polypeptide is also detected with this antibody (indicated by a ϫ in Fig. 1C, lane 6). Upon prolonged exposure of these blots, the 47-49-kDa wt1 isoforms can also be detected in extracts of TM3 cells (data not shown). These results strongly suggest the existence of novel Wt1 isoforms containing or lacking the Wt1 exon V and having molecular masses ϳ10 kDa higher than Wt1. In this manuscript, we will refer to these larger isoforms as Wt1 * to distinguish them from the 47-49-kDa Wt1 isoforms.
To determine whether Wt1 * could be detected in normal tissues, nuclear extracts were prepared from murine ovaries FIG. 2. Detection of Wt1 * in sporadic WTs. A, schematic diagram of the genomic organization of the human WT1 gene. The nucleotide sequence and predicted polypeptide sequence of both the human and murine genes upstream of the AUG codon is indicated. The white boxes represent alternatively spliced exons, whereas hatched boxes represent the zinc fingers. Each exon is numbered with a roman numeral. The relative position of two CUG codons present upstream from and in frame with the initiator AUG codon in the human and murine 5Ј-UTRs is shown. The putative CUG initiation codons are underlined, whereas the initiator AUG codon is underlined and in bold. The peptide sequence used to generated the polyclonal serum S3 is boxed. Amino acid homology of the predicted amino-terminal extension from the human and murine cDNAs is presented. Black circles signify identity, whereas open circles signify conservative changes. The methionine encoded by the initiator AUG is denoted in bold. hWT1, human WT1 cDNA sequence; mWT1, murine wt1 cDNA sequence; hwt1, human WT1 polypeptide sequence; mwt1, murine Wt1 polypeptide sequence. B, analysis of WT1 * in WTs. A Western blot of nuclear extracts from murine TM3 cells (lane 1) or whole cell extracts from WTs (lanes 2-8) were probed with anti-WT1 * serum, S3. The WT1 * isoforms are indicated by two arrows. Molecular mass markers (in kDa) are indicated to the right of the panel. and probed with the anti-WT1 antibody, C-19, directed against the WT1 carboxyl terminus. Extracts prepared from COS-1 cells transfected with CMV/wt1(Ϫ/Ϫ) (Fig. 1D, lane 2) or CMV/ wt1(ϩ/ϩ) (Fig. 1D, lane 3) produced the appropriate wt1 isoforms and acted as standards on this blot. No Wt1 protein was present in untransfected COS-1 cells (Fig. 1D, lane 1). Western blot analysis of nuclear extracts from murine ovaries revealed the presence of the 47-49-kDa Wt1 isoforms, as well as the Wt1 * isoforms (Fig. 1D, lane 4). A similar result was obtained with nuclear extracts from murine testis (data not shown). These results indicate that Wt1 * is present in normal murine tissue.
Expression of WT1 * in WTs-We felt it unlikely that WT1 * was being generated from an alternatively spliced transcript because the structure of this mRNA has been extensively analyzed by S1 nuclease (Pelletier et al., 1991a) and RNase protection . We considered the possibility that an alternative translation initiation event was responsible for producing the novel WT1 polypeptides because examination of the murine and human nucleotide sequences upstream of the initiator AUG revealed the presence of two CUG codons ϳ200 nucleotides upstream from and in frame with the initiator AUG codon (Fig. 2A). These CUG codons are conserved between human and mouse cDNAs, although their positions relative to the initiator AUG differ slightly. Because CUG codons are used as translation initiation codons by some eukaryotic mRNAs (see Introduction), we investigated the possibility that WT1 * was generated by alternative translational initiation. We generated a polyclonal antibody against a peptide representing amino acids 9 -24 of the predicted amino-terminal extension (boxed in Fig. 2A). This antibody, called S3, was used to screen a series of WTs for expression of WT1 * . Two Wt1 * isoforms (ϩ/Ϫ 17 aa) are detected by S3 in murine TM3 cells (Fig. 2B, lane 1). Two of the tumors examined also showed expression of WT1 * (Fig. 2B, lanes 2 and 8). In our small sample cohort, there was a direct correlation between the presence of WT1 * and the presence of WT1 protein (data not shown). We hypothesize that WT1 * arises from alternative translation initiation on the WT1 mRNA template and contains a unique amino-terminal extension not present in WT1.
WT1 * Initiates at a CUG Codon Upstream from and in Frame with the Initiator AUG Codon-To directly demonstrate that the amino-terminal extension of WT1 * is due to alternative translational initiation, a series of deletion mutants was generated from the murine and human cDNAS (Fig. 3A). Two of the constructs {pKS/mwt1[ϩ182] and pSP/hWT1[ϩ200]} contain shortened 5Ј-UTRs that lack the above described CUG codons, whereas others {pKS/mwt1[ϩ475] and pKS/ hWT1[ϩ378]} had 5Ј-UTRs that included these CUG codons (Fig. 3A). Following in vitro transcription and translation, [ 35 S]methionine-labeled protein products were analyzed by SDS-PAGE. In the absence of exogenously added mRNA, no protein product is detectable (Fig. 3B, lane 1). The truncation mutants, pKS/mwt1[ϩ182] and pSP/hWT1[ϩ200], produced only one polypeptide species having a molecular mass of ϳ50 kDa (Fig. 3B, lanes 2 and 4, indicated by an arrowhead). These results are consistent with translational initiation occurring at the predicted AUG codon (underlined and bold in Fig. 2A). Translation of mRNA produced from pKS/mwt1[ϩ475] and pKS/hWT1[ϩ378] produced two polypeptide species, one having the expected molecular mass for an initiation event occurring at the initiator AUG, and one having a molecular mass identical to that of WT1 * (Fig. 3B, lanes 3 and 5). Similar results were obtained upon translation of these mRNAs in a wheat germ extract, indicating that the above results are not specific to reticulocyte lysate (data not shown). These results are consistent with our hypothesis that alternative translational initiation is responsible for generating the Wt1 * protein isoforms and demonstrate that both the human and murine mRNAs are competent for producing this protein species.
To define the codon responsible for directing translation initiation of Wt1 * , a series of deletion and site-directed mutants were generated within the murine 5Ј-UTR (Fig. 4A). In addition to truncating various portions of the 5Ј-UTR, stop codons were introduced that flanked the CUG codons identified above ( Fig. 2A), and mutations affecting the nature of each CUG codon were generated. Deletions within the first 507 nucleotides from the 5Ј-UTR of the wt1 mRNA did not abolish production of Wt1 * (Fig. 4B, lanes 2-5), whereas a deletion removing all but 182 nucleotides of 5Ј-UTR generated a transcript unable to produce Wt1 * (Fig. 4B, lane 11). Deletion of the Wt1 initiator AUG codon abolished production of Wt1, instead generating a polypeptide having a molecular mass of ϳ37 kDa (Fig. 4B, lane 12, indicated by a small arrow). The molecular mass of this polypeptide is similar to that expected from internal initiation of ribosomes at a downstream, in frame AUG codon at position 378 (relative to the A of the initiator AUG codon). Introduction of a UGA codon 229 nucleotides upstream from and in frame with the AUG initiator codon did not affect FIG. 3. In vitro synthesis of Wt1 * . A, diagram of in vitro wt1 expression vectors. The black boxes represent the coding regions, whereas the stippled and hatched boxes represent the murine and human 5Ј-UTRs, respectively. The dotted and checkered boxes represent the T7 and SP6 RNA polymerase promoters. The small black rectangle above the murine 5Ј-UTR represents a small ORF present in this portion of the cDNA. The two black dots represent the upstream CUG codons highlighted in Fig. 2A. B, in vitro translation products obtained from RNA generated by 5Ј-UTR deletion mutants. RNA was generated from constructs shown in A and translated in rabbit reticulocyte lysate at a mRNA concentration of 16 g/ml. RNA prepared from the following vectors was used as input . 3 l of the translation reaction were fractionated on a 10% SDS-PAGE gel. Following electrophoresis, the gel was treated with EN 3 Hance, dried, and exposed to Fuji x-ray film at Ϫ70°C for 12 h. production of Wt1 * (Fig. 4B, lane 6), indicating that the Wt1 * initiation codon must lie downstream of this site. On the other hand, placing a UGA codon at position 183, in frame, and upstream of the initiator codon generated an mRNA no longer capable of synthesizing Wt1 * (Fig. 4B, lane 7). These data suggest that the signals responsible for Wt1 * translation initiation lie between nucleotides ϩ183 and ϩ229. Site-directed mutagenesis was used to abolish the individual CUG codons at position 204 and 192 (see Fig. 2A). Converting CUG 204 to CUC 204 drastically affected expression of Wt1 * while having little effect on production of Wt1 (Fig. 4B, lane 8). Expression of Wt1 * was reduced ϳ10-fold but not completely abolished. Mutagenesis of CUG 192 to CUC 192 had no effect on production of Wt1 * (Fig. 4B, lane 9), and mutagenesis of both CUG codons The smaller arrow indicates the position of migration of the truncated Wt1 polypeptide predicted to arise from translation initiation at the downstream AUG codon. Following electrophoresis in a 10% SDS/polyacrylamide gel, the gel was treated with EN 3 Hance, dried, and exposed to X-Omat (Kodak) film at Ϫ70°C with an intensifying screen overnight.
had the same effect as altering only CUG 204 (Fig. 4B, lane 10). These results strongly indicate that translational initiation at CUG 204 is responsible for production of Wt1 * . We interpret the residual production of Wt1 * observed with wt1[ϩ475]-(CTC@204) and wt1 [ϩ475](CTC@204/192) to indicate that the CUG to CUC mutation is leaky and that nucleotide sequences flanking the CUG codon are influencing translation initiation at this site (Boeck and Kolakofsky, 1994;Grü nert and Jackson, 1994).
It is clear from the translation results obtained with wt1[ϩ475](TGA@229) and wt1[ϩ475](TGA@183) (Fig. 4B) that the Wt1 * translation codon must lie between nucleotides ϩ183 and ϩ229. To rule out the possibility that the initiator codon for Wt1 * was upstream or downstream of CUG 204 and that the effects on translation initiation of mutating CUG 204 were not the result of altering a motif that indirectly influenced the efficiency of initiation, four additional constructs were generated (Fig. 5A)  wt1[ϩ475](⌬229 -205) generates a 24-base pair in-frame deletion immediately upstream of the CUG 204 codon and will abolish production of Wt1 * only if the initiation codon lies upstream of CUG 204 . Following in vitro transcription/translation of these plasmids and their respective mRNAs, the [ 35 S]methioninelabeled protein products were analyzed by SDS-PAGE. Translation of wt1[ϩ475] produced both Wt1 * and Wt1 as expected (Fig. 5B, lane 2). Only Wt1 was produced following translation of RNA synthesized from wt1[ϩ202], consistent with the notion that the Wt1 * initiation codon does not lie downstream of CUG 204 . Mutation of CUG 204 to an AUG codon increased the production of Wt1 * (Fig. 5B, lane 4), consistent with the assignment of CUG 204 as the initiation codon. Conversion of CUG 204 to a UGA codon abolished production of Wt1 * , whereas production of Wt1 was still observed (Fig. 5B, lane 5). Removal of nucleotides 229 -205 affected the overall translational efficiency of the mRNA produced (Fig. 5B, compare intensity of products in lane 7 with those in lane 6); however, the production of both Wt1 * and Wt1 was still observed (Fig. 5B, lane 7), indicating that the Wt1 * initiator codon does not lie upstream of CUG 204 . These results indicate that CUG 204 is the Wt1 * initiation codon.
Wt1 * Is Present in the Nucleus-To determine the subcellular localization of Wt1 * , an HA tag was placed in frame with and directly upstream from the CUG 204 initiation codon (Fig. 6A). It was necessary to place the HA tag at the amino terminus because translation initiation at CUG 204 is leaky and allows scanning ribosomes to initiate at the downstream, in frame ATG codon (See Figs. 3, 4, and 5), resulting in the production of both Wt1 * and Wt1. The introduction of an HA tag immediately upstream of CUG 204 was advantageous because the anti-HA monoclonal antibody, 12CA5, would specifically detect expression of only the 57-kDa isoform (Fig. 6B, lane 2). The construct CMV/wt1[ϩ42] (Fig. 6A) produced only the 47-kDa isoform (Fig. 6B, lane 1) and contained an HA epitope at the carboxyl terminus. Immunofluorescence of COS-1 cells transfected with either CMV/wt1[ϩ42] or CMV/wt1[HA/wt1 * ] revealed that both isoforms localize to the nucleus (Fig. 6, C and D). Wt1 * Mediates Transcriptional Repression-In most transient transfection assays, WT1 functions as a repressor of tran-scription (see Introduction). To determine whether the novel amino-terminal domain of Wt1 * altered the transcriptional properties of Wt1, 3 wt1/GAL4 expression vectors were generated (Fig. 7A). The Wt1 zinc fingers were replaced with the DNA binding domain of GAL4 (Ma and Ptashne, 1987), enabling the scoring of Wt1 activity on a promoter containing GAL4 binding sites. In such a context, Wt1 produced from CMV/wt1[ϩ182/GAL4] is able to repress transcription of the cat gene under TK promoter control ϳ10-fold, similar to levels obtained with CMV/wt1[ϩ475/GAL4], which produces Wt1 * and Wt1 (Fig. 7B). Western blot analysis indicated that Wt1 and Wt1 * were produced from CMV/wt1[ϩ475/GAL4] to approximately equal levels. 2 Transient transfections with CMV/ wt1[HA/wt1 * /GAL4], which only produces Wt1 * (data not shown), represses CAT production ϳ 5-fold (Fig. 7B). These results demonstrate that in the system we have analyzed, the amino-terminal extension of Wt1 * does not significantly change the transcriptional properties of Wt1.

DISCUSSION
The WT1 gene products are necessary for regulating normal differentiation of the urogenital system. Mutations in the WT1 gene result in malformations of the urogenital system as well as predispose to WTs. The structure of the WT1 gene indicates that it should produce four alternatively spliced mRNAs that direct the synthesis of isoforms having molecular masses of 47-49 kDa. In this report, we demonstrate that WT1 gene expression is more complex than anticipated, with a non-AUG translational initiation event producing additional protein isoforms of 54 -56 kDa (WT1 * ). We have found WT1 * to be expressed in a number of tissues and cells including murine TM3 cells (Fig. 1), murine ovaries (Fig. 1), murine testis (data not shown), immortalized rat granulosa cells (data not shown), and human K562 cells (data not shown). We have been unable to accurately establish the relative ratio of WT1 to WT1 * in these cells and tissues, because the antibodies we are employing recognize the isoforms with different efficiencies (e.g. Fig. 1,  compare C with B), perhaps due to differences in protein con-2 W. Bruening, unpublished data.
FIG. 6. Nuclear localization of Wt1 * . A, schematic representation of expression vectors used in this study. A black box signifies the wt1 ORF, a hatched box represents the 5Ј-UTR, and a stippled box represents the HA epitope tag (Pelletier et al., 1991a), whereas a checkered box signifies the CMV promoter. CUG 204 is represented by a filled circle. B, Western blot analysis of COS-1 cells transfected with wt1 expression vectors. Cell extracts were prepared and processed for immunoblotting with the anti-HA antibody 12CA5 as described previously (Goodyer et al., 1995). C and D, immunofluroescent detection of wt1 and wt1 * . COS-1 cells were electroporated in PBS ϩ 20 mM Hepes, pH 7.5, at 260 V/960 microfarads with 10 g of plasmid DNA. The cells were seeded into chamber slides and allowed to recover for 48 h. After washing with PBS, the cells were fixed for 20 min in 3.7% formaldehyde/PBS followed by 5 min in methanol at Ϫ20°C. Slides were blocked with 10% goat serum/0.3% Triton X-100 and probed with the anti-HA antibody 12CA5. Protein was visualized with an fluorescein isothiocyanate-conjugated goat anti-mouse IgG (Jackson Labs). After washing, the cells were mounted in Immunomount (Fisher) and photographed using T-Max 100 black and white film (Kodak). Cells were transfected with CMV/wt1[ϩ42] (C) or with CMV/wt1[HA/wt1 * ] (D).
formation or isoform-specific post-translational modifications. However, our data directly demonstrates that WT1 * isoforms containing or lacking the first alternatively spliced exon are generated. Using isoform-specific antibodies, we were able to detect both isoforms in TM3 cells (Fig. 1C). In addition, two WT1 * isoforms are clearly detectable on Western blots of whole cell extracts of WTs (Fig. 2B). Although we have not analyzed the presence or the absence of the second alternative splice site within WT1 * , there is no reason not to suspect its presence. Our data, taken together with the documented alternative splicing of the WT1 gene  and RNA editing (Sharma et al., 1994), indicate that this gene may produce as many as 16 different protein isoforms. To date, WT1 functional studies have generally been performed using expression vectors capable of producing only the 47-49-kDa isoforms. Given that the WT1 gene products can oligomerize (Reddy et al., 1995;Moffett et al., 1995) and that the isoform ratio is important for proper urogenital system development (Bruening et al., 1992), functional studies need to take into account the contribution of WT1 * isoforms to the overall biological function of the WT1 gene, including the possible interplay of the various isoforms.
Our data indicate that CUG 204 is the initiation codon responsible for directing synthesis of Wt1 * . A number of mRNAs use non-AUG initiation events to generate amino-terminally extended isoforms (see Introduction). The Wt1 initiator CUG is in a favorable consensus sequence (5Ј-XXAXXCUGG-3Ј, where the important nucleotide residues are underlined) (Kozak, 1987), and these likely play a role in directing translational initiation. The identity of the codon at this position is important in determining the efficiency with which this site is recognized, because mutagenesis of the initiation codon CUG 204 to CUC 204 significantly reduced (ϳ15-fold) the amount of Wt1 * produced (Fig. 4). However, it is clear that additional signals influence initiation at this site since the CUG to CUC mutation reduced but did not completely abolish translation initiation (Fig. 4). Surprisingly, deleting 24 nucleotides immediately upstream of CUG 204 significantly reduced the overall translational efficiency of the template mRNA (Fig. 5B, compare lanes  7 and 6). Sequences downstream of non-AUG initiation codons have been shown to be important for determining efficiency of recognition of the start codons (Boeck and Kolakofsky, 1994;Grü nert and Jackson, 1994). However, in the case of the wt1 template, the deletion of sequences upstream of CUG 204 has a global effect on translation initiation, including at the downstream AUG codon. In the case of fibroblast growth factor-2 mRNA, a ribosome landing pad is postulated to directly recruit 43S preinitiation complexes and guide them to a non-AUG codon (Vagner et al., 1995). An internal ribosome binding site near CUG 204 on the wt1 mRNA template could account for the effects seen on global initiation by wt1[ϩ475](⌬229 -205) (Fig.  5B). Alternatively, the overall stability of the template may be affected by this 24-nucleotide deletion.
The degree of conservation between the amino-terminal extension of the murine and human WT1 * isoforms is not as high as between the WT1 proteins. At the amino acid level, the murine and human WT1 isoforms are 96% identical (Buckler et FIG. 7. Trans-repression by Wt1 and Wt1 * . A, schematic representation of vectors used in this study. See legend to Fig.  6 for details. The cross-hatched box represents the cat gene, whereas the TK promoter is shown by an open box. The two dark rectangles within the TK promoter illustrate GAL4 binding sites. B, CAT assays of transient transfections performed with the expression and reporter vectors in A. NIH 3T3 cells were transfected by calcium phosphate and allowed to recover for 48 h. The cells were lysed by freezing and thawing 3 times in 0.25 M Tris, pH 8.0. Following removal of the cellular debris by centrifugation (12,000 ϫ g for 5 min), CAT assays were performed on the supernatant as described previously (Gorman, 1985). A vector containing the CMV promoter driving expression of the ␤-gal gene was included in all transfections and used to standardize efficiencies. CMV/⌬ is the backbone vector used to generate the expression vectors in A but lacking a gene downstream of the CMV promoter (Goodyer et al., 1995(Goodyer et al., ). al., 1991, but the amino-terminal extensions of the human and murine WT1 * isoforms are only 65% identical. However, the first 35 amino acids of WT1 * are quite conserved (88% identical) and may encode a functional domain. Alternatively, overall charge or conformation may be more important to the function of this domain rather than primary sequence per se.
We have analyzed the genomic region encoding the aminoterminal extension of WT1 * for possible mutations in WTs using polymerase chain reaction single stranded conformational polymorphisms (Varanasi et al., 1994). 2 Examination of this region in over 100 WTs failed to detect any mutations. This is not particularly surprising because the majority of WT1 mutations occur within the DNA binding domain (Bruening et al., 1992;Pelletier et al., 1991b;Varanasi et al., 1994) with very few mutations occurring outside of this region.
Several investigators have noted the presence of a ϳ54 -56-kDa contaminating polypeptide in studies involving immunoprecipitation of WT1. Given our results, these polypeptides likely represent WT1 * isoforms generated by translation initiation at CUG 204 . The presence of such isoforms is apparent in Fig. 3F of Larrson et al. (1995), displaying an immunoprecipitation on nuclear extracts with anti-WT1 antibodies from the mesonephric cell line, M15. A protein with similar electrophoretic properties as WT1 * was also detected by Rackley et al. (1993) on an immunoblot analysis of nuclear extracts prepared from the WT1-expressing kidney cell line, 293 (see Fig. 3a; Rackley et al. (1993)). The functional significance of producing WT1 isoforms with an amino-terminal extension is not obvious. The amino-terminal domain on its own does not have a strong trans effect on transcription because fusion constructs containing this domain fused to the GAL4 DNA binding domain did not significantly affect transcription of an appropriate reporter in transient transfection experiments. 2 This domain may be involved in a number of other activities such as heterodimer formation, subnuclear localization, or regulation of WT1 intrinsic activity. Our description of a non-AUG translation event at the WT1 locus indicates that the overall complexity of WT1 regulation is greater than previously suspected. Biological assays involving all WT1 isoforms should present a better picture of the role of WT1 * in normal and abnormal urogenital system development.