Inhibition of Wilms tumor 1 transactivation by bone marrow zinc finger 2, a novel transcriptional repressor.

The Wilms tumor suppressor gene, wt1, encodes a zinc finger transcription factor that has been implicated in the regulation of a number of genes. Protein-protein interactions are known to modulate the transcription regulatory functions of Wilms tumor (WT1) and have also implicated WT1 in splicing. In this report, we identify a novel WT1-interacting protein, bone marrow zinc finger 2 (BMZF2), by affinity chromatography utilizing immobilized WT1 protein. BMZF2 is a potential transcription factor with 18 zinc fingers. The BMZF2 mRNA is mainly expressed in fetal tissues, and the protein is predominantly nuclear. Co-immunoprecipitation experiments are consistent with an in vivo association between WT1 and BMZF2. Glutathione S-transferase pulldown assays and far Western blots revealed that zinc fingers VI-X (amino acids 231-370) are required for interaction with the zinc finger region of WT1. Functionally, BMZF2 inhibits transcriptional activation by WT1. Moreover, a chimeric protein generated by fusion of BMZF2 to the GAL4 DNA-binding domain significantly decreases promoter activity of a reporter containing GAL4 DNA-binding sites, suggesting the presence of an active repressor domain within BMZF2. Our results suggest that BMZF2 interferes with the transactivation potential of WT1.

characterized and is mutated in 10 -15% of sporadic WTs (2). Germ line wt1 lesions in humans are associated with predisposition to WTs and aberrant differentiation of the urogenital system (3).
The wt1 gene encodes a transcription factor with a prolineglutamine-rich amino terminus and four carboxyl-terminal zinc fingers of the Krü ppel C 2 -H 2 class. The mRNA contains two alternative sites of translation initiation (4,5), two alternatively spliced exons (6,7), and undergoes RNA editing (8), thus potentially encoding 24 different protein isoforms with predicted molecular masses of 36 -65 kDa. The function of the alternative translation initiation events, the RNA editing modification, and the first alternative splicing event (exon V) have not been well defined, although exon V can repress transcription when fused to a heterologous DNA-binding domain (9). Alternative splicing of exon IX inserts or removes three amino acids (ϮKTS) (referred to as WT1(ϩKTS) or WT1(ϪKTS)) between zinc fingers III and IV and changes the DNA binding specificity of WT1 (10). The WT1(ϪKTS) isoforms can bind to two DNA motifs as follows: (i) a GC-rich motif, 5Ј GXGXGGGXG3Ј, related to the EGR-1-binding site (10); and (ii) a (5ЈTCC3Ј) n -containing sequence (11). Recently NMR relaxation studies (12) have indicated that the KTS insertion increases the flexibility of the linker between fingers III and IV and abrogates binding of the fourth zinc finger to its cognate site in the DNA major groove. A number of genes involved in growth regulation and cellular differentiation contain WT1binding sites within their promoters, and their expression can be modulated by WT1 in transfection assays (reviewed in Refs. [13][14][15]. The wt1 gene product has been shown to mediate both transcriptional repression and activation (reviewed in Refs. [13][14][15]. Whether WT1 behaves as an activator or repressor appears to depend on promoter architecture surrounding the WT1-binding sites as well as on the presence of auxiliary transacting factors. Accordingly, it is well accepted that WT1 activity can be controlled by protein-protein interactions. A number of proteins are known to associate with WT1 and these include p53, p73, p63, SF-1, Par-4, Ciao 1, UBC9, Hsp70, U2AF65, CBP/p300, WTAP, and WT1 itself (reviewed in Refs. [13][14][15]. The interaction of some of these proteins with WT1 is associated with modification of WT1 transcriptional properties as well as effects on the properties of the interacting partner (see "Discussion"). An additional role for WT1 in splicing is also postulated based on the subnuclear localization of WT1(ϩKTS) isoforms and the interaction of these isoforms with splicing factors (16 -18).
Utilizing affinity chromatography of nuclear extracts passed over immobilized WT1 protein, followed by mass spectrometric analysis of a specifically retained protein, we have identified a novel WT1-interacting protein, named bone marrow zinc finger 2 (BMZF2). BMZF2 was first identified from a screen of zinc finger proteins expressed in the hematopoietic system (19). The BMZF2 protein has 18 tandem zinc fingers and a Krü ppelrelated amino-terminal domain. Aside from RT-PCR analysis of BMZF2 transcripts demonstrating the presence of BMZF2 transcripts in a large number of tissues (19), no further characterization of this gene has been reported. Here we demonstrate that WT1 physically interacts with BMZF2 and that this interaction inhibits WT1-mediated transcriptional activation. Additionally, fusion of BMZF2 to the GAL4 DNA-binding domain produced a chimera capable of repressing transcription of a reporter gene containing GAL4-binding sites, indicating that BMZF2 is a novel transcriptional repressor. These results suggest that BMZF2 interferes with the transactivation properties of WT1.

EXPERIMENTAL PROCEDURES
Materials and General Methods-Restriction endonucleases, calf intestinal alkaline phosphatase, the Klenow fragment of DNA polymerase I, T4 DNA ligase, and T4 DNA polymerase were purchased from New England Biolabs. The luciferase assay kit was purchased from Promega. [␥- 32  Preparation of plasmid DNA, restriction enzyme digestion, agarose gel electrophoresis of DNA, DNA ligation, and bacterial transformations were carried out using standard methods (20). Clones of DNA PCR amplification products were always sequenced by the chain termination method using double-stranded DNA templates to ensure the absence of mutations.
Cell Culture, Transfections, and CAT and Luciferase Assays-293 and COS-7 cell lines were maintained in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal calf serum (Invitrogen), penicillin, and streptomycin. For transient transfections, cells were plated at a density of 2-5 ϫ 10 5 cells per 100-mm diameter dish 24 h prior to transfection. The cells were transfected by the calcium phosphate precipitation method (20). Individual DNA precipitates were adjusted to contain equal amounts of total DNA by the addition of the empty expression vector, pcDNA3. All transfections and subsequent CAT and luciferase assays were performed at least in duplicate. Cells were washed and refed 16 h post-transfection and harvested ϳ48 h later. Cells were scraped from the dishes following a phosphate-buffered saline (PBS) wash, centrifuged, and resuspended in 150 l of 250 mM Tris (pH 8.0). They were then subjected to three rounds of freezethaw; an aliquot was taken for measurement of ␤-galactosidase activity, and the remainder of the extract was heated to 65°C for 10 min and then assayed for CAT activity (21). Following thin layer chromatography, regions containing acetylated [ 14 C]chloramphenicol, as well as unacetylated [ 14 C]chloramphenicol, were quantitated by direct analysis on a PhosphorImager (Fujix BAS 2000). Luciferase activity was determined using the Promega luciferase assay kit. All CAT and luciferase activity values were normalized to ␤-galactosidase values, which served as internal controls in the transfections.
Preparation of HeLa Nuclear Extracts-Fifty (150-mm) plates of HeLa S3 cells were grown for the preparation of nuclear extracts. Cells were maintained in Dulbecco's modified Eagle's medium with 10% fetal calf serum, penicillin (100 units/ml), streptomycin (100 g/ml), and L-glutamine (4 mM) (Invitrogen) and grown at 37°C in a humidified 5% CO 2 incubator. The cells were harvested by scraping in cold PBS and collected by centrifugation. After washing in PBS, the cell pellet was resuspended in 5 ml of Buffer A (10 mM Hepes (pH 7.9), 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM DTT) supplemented with 1 mM benzamidine HCl, 1 mM phenylmethylsulfonyl fluoride, 1 mM leupeptin, and 1 mM antipain. The cells were lysed using a type B Dounce, and lysis was verified by staining with trypan blue. The lysate was centrifuged at 4°C for 15 min at 1500 ϫ g, and the supernatant was removed. The pellet was resuspended in 5.5 ml of 75 mM NaCl Affinity Chromatography (AC) Buffer (20 mM Hepes (pH 7.5), 10% glycerol, 1 mM DTT, 1 mM EDTA) with protease inhibitors and sonicated for 60 s (2 bursts of 30 s) at 4°C. The lysate was then centrifuged at 4°C for 30 min at 20,800 ϫ g, and the supernatant was collected.
Isolation of BMZF2 by Affinity Chromatography-GST and GST-WT1ZF (WT1 zinc fingers I-IV fused in-frame to GST) recombinant protein were coupled to Affi-Gel 10 at a protein/resin concentration of 0, 0.5, and 2.0 mg/ml. Coupling reactions were performed in a final volume of 200 l of AC buffer by incubating at 4°C overnight and rotating the samples end-over-end. The affinity matrix was washed in AC buffer containing 75 mM NaCl and 80 mM ethanolamine at room temperature for 1 h, followed by an additional 1 h wash in 75 mM NaCl-AC buffer containing 1 mg/ml purified bovine serum albumin. The matrix was then washed with AC buffer containing 1 M NaCl at room temperature for 10 min and then equilibrated and stored in AC buffer containing 75 mM NaCl. Coupling efficiencies were ϳ80%. Ten column volumes (1 ml) of HeLa nuclear extract were applied to a small column containing 100 l of affinity matrix. The resin was washed 3 times with AC buffer containing 75 mM NaCl. Elutions were performed sequentially with 2 times 2 column volumes (2 ϫ 200 l) of each of the following: (i) 75 mM NaCl-AC buffer with 1% Triton X-100; (ii) 300 mM NaCl-AC buffer; (iii) 1 M NaCl-AC buffer; and (iv) 1% SDS-AC buffer.
One-quarter of each fraction was analyzed on 12.5% SDS-polyacrylamide gels. Gels were prepared for silver staining by fixing overnight in 50% methanol, 10% acetic acid followed by a 10-min rinse in 20% ethanol and a 10-min rinse in water. Gels were then reduced with sodium thiosulfate (0.2 g/liter) for 1 min, rinsed twice with water for 20 s, and incubated in silver nitrate (2.0 g/liter) for 30 min. Gels were washed once with developing solution (sodium carbonate (30 g/liter), formaldehyde (1.4 ml of 37% solution/liter), sodium thiosulfate (10 mg/liter)) for 30 s and incubated in the developing solution until the desired intensity was reached. The reaction was stopped by exchanging the developing solution with 1% acetic acid for a minimum of 20 min. Specifically eluted bands were excised from the gel. The tryptic digestions of the protein samples were performed by Borealis Biosciences Inc., and the molecular mass of the tryptic fragments was determined with a Perspective Biosystems Voyager Elite MALDI-TOF (Toronto, Canada). The protein was identified by matching the observed proteolytic masses obtained in the MALDI-TOF spectra with the hypothetical tryptic peptide masses derived from the NCBI non-redundant translated GenBank TM data base.
For the analysis of the interaction between endogenous WT1 and BMZF2, K562 cells were lysed in lysis buffer (20 mM Tris-HCl (pH 7.4), 10 mM KCl, 10 mM MgCl 2 , 2 mM EDTA, 10% glycerol, 1% Triton X-100, 2.5 mM ␤-glycerol phosphate, 1 mM NaF, 1 mM DTT, 1 g/ml of aprotinin, 1 g/ml of leupeptin, 1 g/ml of Pefabloc, and 1 g/ml of pepstatin A) for 10 min on ice. Lysates were sonicated twice for 15 s and incubated with 420 mM NaCl in lysis buffer for 1 h on ice. Extracts were incubated with the indicated antisera or antibodies, and immune complexes were collected with protein A-Sepharose beads at 4°C for 1 h. The beads were washed 6 times with lysis buffer, and the proteins were eluted with 1ϫ SDS loading buffer. Immunoprecipitates were separated by SDS-PAGE and detected by Western blotting using anti-WT1 antibody (F-6, Santa Cruz Biotechnology) or anti-BMZF2 antibody.
In Vitro Transcription and Translation-In vitro transcriptions of pKSII/BMZF2-(1-622) and deletion constructs were performed using T3 RNA polymerase on templates that had been linearized with XbaI. In vitro translations were performed in the presence of [ 35 S]methionine in rabbit reticulocyte lysates, essentially as described by the manufacturer (Promega). For in vitro synthesis of p53, an SP65-based plasmid containing the p53 gene (linearized with HindIII) was used in transcription reactions with SP6 RNA polymerase. In vitro synthesized translation products were electrophoresed on 10% SDS-polyacrylamide gels, treated with EN 3 HANCE, dried, and exposed to X-Omat film (Eastman Kodak Co.).
GST Pulldown Assays-GST and GST-WT1ZF (containing WT1 zinc fingers I-IV fused to GST) recombinant proteins were purified as described previously (23). Briefly, bacterial extracts producing GST or GST-WT1 recombinant proteins were incubated with a 50% slurry of glutathione-agarose beads (Amersham Biosciences) in TNE-150 buffer (50 mM Tris-HCl (pH 8.0), 1% Nonidet P-40, 2 mM EDTA, 150 mM NaCl) with rocking. GST proteins bound to the beads were then collected by brief centrifugation (12,000 ϫ g) and washed four times with TNE-150 buffer. The captured protein was then used in GST pulldown assays. An aliquot of the captured protein was analyzed by SDS-PAGE and the yield estimated by Coomassie Blue staining. [ 35 S]Methionine-labeled proteins synthesized from in vitro translation reactions were incubated with immobilized GST or GST-WT1 proteins in binding buffer (50 mM Tris-HCl (pH 7.5), 12.5 mM MgCl 2 , 10% glycerol, 1% Nonidet P-40, 150 mM NaCl, 1 mM DTT, 200 g/ml bovine serum albumin, 200 g/ml ethidium bromide) for 1 h at 4°C. The beads were collected by centrifugation and washed 4 times with 1 ml of binding buffer. Bound proteins were eluted from the beads by boiling in 1ϫ SDS loading buffer (62.5 mM Tris-HCl (pH 6.9), 10% glycerol, 2% SDS, 5% ␤-mercaptoethanol) and were separated by electrophoresis.
Nucleotide Sequence Accession Number-The nucleotide sequence of a full-length human BMZF2 cDNA has been deposited in the Gen-Bank TM data base under accession number AY148489.

Isolation of BMZF2 as a Novel WT1-interacting Protein-
Genetic screens and chemical cross-linkers have identified a small number of proteins that interact with the WT1 zinc finger domain (see "Discussion"). We looked to employ a biochemical approach to validate the interaction of previously described WT1-interacting proteins, as well as identify potentially new protein(s) that could interact with the WT1 zinc finger domain. To this end, HeLa nuclear extracts were incubated with Affi-Gel 10 resin that had been coupled to GST or GST-WT1ZF fusion proteins. Elutions obtained with 300 mM NaCl, 1 M NaCl, and 1% SDS were analyzed by SDS-PAGE and silver staining. Two proteins, an ϳ40-kDa (denoted by an asterisk) and ϳ30-kDa (denoted by an arrow) species, were visible in the 1 M NaCl elution (Fig. 1A). The 40-kDa species is present in elutions from both the GST and GST-WT1ZF affinity matrices (Fig. 1A, lanes 2, 3, and 5-7) and was not observed if HeLa A, affinity chromatography of HeLa nuclear extracts. HeLa cell nuclear extracts were incubated with 0, 0.5, and 2.0 mg/ml Affi-Gel 10 resin coupled to GST or GST-WT1ZF. Following elution with 1 M NaCl, 40 l (25%) of each eluent was analyzed on a 12.5% SDS-polyacrylamide gel and visualized by silver staining. The position of migration of an ϳ40and an ϳ30-kDa protein species is indicated by an asterisk and an arrowhead, respectively. The protein to resin cross-linking ratio, as well as whether or not nuclear extract had been applied to the affinity matrix, is indicated above the panel. The positions of migration of molecular mass markers (New England Biolabs) is indicated to the left. B, nucleotide and amino acid sequence of BMZF2 cDNA. The nucleotide and amino acid sequences of BMZF2 are shown with numbering of nucleotides and amino acids displayed to the right. The matched BMZF2 peptides identified by MALDI-TOF are in rectangles. The zinc fingers are shown in boldface type. The amino acid sequence differences between our sequence and that in GenBank TM , accession numbers NM 005774 and AF067164, are highlighted by an underline. nuclear extract were omitted from the GST or GST-WT1ZF columns (Fig. 1A, lanes 1 and 4). These results suggest that the 40-kDa protein species is not specifically retained by the WT1ZF domain, and characterization of this species was not further pursued. The ϳ30-kDa protein species was observed only in elutions from the GST-WT1ZF matrix (compare lanes 6 and 7 to lane 3), was not observed in elutions from columns that had no GST-WT1ZF coupled to them but had been exposed to nuclear extract (lanes 2 and 5), and was not observed in elutions where the GST-WT1ZF column had not been exposed to nuclear extract (lane 4). As well, this protein species was not observed in the 300 mM NaCl elutions from the GST-WT1ZF column (data not shown). These results indicate that the ϳ30-kDa protein species was specifically retained by and eluted from the GST-WT1ZF affinity matrix.
Mass spectrometry of 15 peptides obtained from the excised protein from the SDS gel identified two proteins, 9 peptides matched to BMZF2 and will be the focus of the current study. The experimentally determined masses of the 9 peptides matched bone marrow zinc finger 2 (denoted by gray boxes in . In order to pursue functional studies with BMZF2, we used RT-PCR to obtain a full coding version of the gene from HeLa cell mRNA (Fig. 1B). BMZF contains a Krü ppel-related amino-terminal domain, named the Krü ppel-related novel box (KRNB), and 18 Krü ppel-like zinc fingers (19). These features suggest that BMZF2 is a nucleic acid-binding protein with potential transcriptional activity.
Sequence analysis of our clones, as well as PCR products, revealed several discrepancies between our sequence and the BMZF2 sequence deposited in GenBank TM (GenBank TM accession numbers NM005774 and AF067164). The reported sequence indicates the presence of 2 adenosine residues at nucleotides 153 and 154, 2 adenosine residues at nucleotides 192 and 193 (on our sequence nucleotide 191), and 2 thymidine residues at nucleotides 200 and 201 (on our sequence nucleotide 198), whereas we find a single adenosine and thymidine residue at these corresponding positions. This has the net effect of altering the reading frame of a portion of the amino-terminal domain of BMZF2 (denoted by an underline in Fig. 1B) and reducing the size of the predicted BMZF2 protein product by 1 amino acid. These differences may reflect errors in the reported sequence of BMZF2 or alternative splicing events. Additionally, there is an adenosine residue at position 946 (corresponds to a thymidine at our position 943), a cytosine at position 949 (corresponds to an adenosine at our position 946), a thymidine at position 983 (corresponds to an adenosine residue at our position 980), and a cytosine at position 1309 (corresponds to a thymidine at position 1306). The amino acids altered by these differences are underlined in the sequence presented in Fig. 1B and may reflect sequencing errors in the original sequence or polymorphisms. We find these same sequence differences in RT-PCR products obtained from RNA isolated from non-transformed cells, indicating that they are not specific to transformed HeLa cells (data not shown).
In Vivo and in Vitro Interaction between BMZF2 and WT1-There was a clear discrepancy in masses between the ϳ30-kDa protein species identified by affinity chromatography and the predicted mass for BMZF2 (72 kDa). To resolve this, we performed co-immunoprecipitation experiments from extracts prepared from human K562 erythroleukemia cells that express both WT1 (26) and BMZF2 (19) (Fig. 2). Immunoprecipitation with either non-immune rabbit serum (Fig. 2A, lane 1) or a rabbit polyclonal anti-BMZF2 antibody (Fig. 2A, lane 2) was performed on extracts prepared from K562 cells. Following fractionation of the immunoprecipitates by SDS-PAGE, Western blotting analysis was performed using an anti-WT1 antibody ( Fig. 2A). The presence of an immunoreactive protein species of ϳ50 kDa was detected only when anti-BMZF2 was used as the immunoprecipitating antibody (compare lane 2 to 1) and is consistent with WT1 isoforms generated from the first ATG initiation codon (4). When anti-BMZF2 antibodies were used to probe total cell extracts from K562 cells (Fig. 2B, lane  3), 6 immunoreactive protein species were visible, ranging in size from ϳ25 to 72 kDa. These protein species are either cross-reacting with our antibody preparation or represent BMZF2 isoforms that contain a portion of the BMZF2 amino terminus (because our polyclonal was raised against the first 80 amino acids of the protein). Immunoprecipitations utilizing either an ␣-HA antibody (as negative control) (Fig. 2B, lane 1) or a monoclonal ␣-WT1 antibody (Fig. 2B, lane 2), followed by probing for the presence of BMZF2 revealed that four of these species were retained by WT1 (compare lane 2 to 1). We note that one of these species is ϳ30 kDa in molecular mass and likely corresponds to the species that we initially identified by affinity chromatography (Fig. 1A).
To confirm these results, co-immunoprecipitation experiments were conducted with extracts prepared from 293 cells transiently transfected with expression vectors driving synthesis of WT1 and HA-tagged BMZF2, Par-4, and SIM-2. After transfections, cells were lysed, immunoprecipitated with a polyclonal anti-WT1 antibody, and subjected to Western blot analysis utilizing an anti-HA antibody. In this experiment, Par-4, a protein known to interaction with WT1 (27), acts as our positive control, whereas SIM-2, a member of the PAS (Per-Arnt-Sim) family of transcription factors and not known to interact with WT1, acts as our negative control. As shown in Fig. 2C, Par-4 is pulled down from cells by WT1 (compare lane 4 to 1), whereas SIM2 is not (compare lane 3 to 1). BMZF2 was coimmunoprecipitated with WT1 (compare lane 5 to 1). Coimmunoprecipitation of BMZF2 was not due to cross-reactivity of BMZF2 with the anti-WT1 antibodies, because BMZF2 was not present in immunoprecipitation reactions performed with this antibody on extracts that lacked WT1 (lane 2). Taken together, these results indicate that WT1 and BMZF2 interact in vivo.
We performed in vitro pulldown assays to determine the protein region(s) required for BMZF2 and WT1 interaction. Luciferase, BMZF2, and p53 were produced in in vitro translation systems and tested for their ability to bind to GST-WT1ZF immobilized on glutathione resin. As expected, luciferase did not bind to immobilized GST or GST-WT1ZF (Fig. 3A,  compare lanes 4 and 7 to lane 1). p53, a factor known to interact with WT1, was specifically retained by immobilized GST-WT1ZF but not by the GST affinity column (compare lane 9 to 6). Similarly, BMZF2 was retained by GST-WT1ZF but not by GST (compare lane 8 to 5). A set of BMZF2 deletion mutants was generated and used to map the region responsible for the interaction with WT1ZF (Fig. 3B). WT1ZF bound to BMZF2-(81-622) (lacking the first 80 amino-terminal amino acids) but not to BMZF2-(1-80), indicating that the amino-terminal 80 amino acids are not responsible for the binding to WT1ZF (Fig.  3B). WT1ZF also bind to deletion mutants BMZF2-(81-370) and BMZF2-(231-370) but not to BMZF2-(371-622) or BMZF2-(81-230), indicating that amino acids 231-370 (zinc fingers 6 -10) contain the WT1-binding site (Fig. 3B).
Expression of BMZF2 mRNA in Human Tissues and Subcellular Localization of BMZF2-To determine the expression pattern of BMZF2 mRNA, we analyzed adult and fetal human mRNAs from a variety of tissues by Northern blotting. To avoid cross-hybridization to other zinc finger transcripts, the blots were probed with the amino-terminal domain of BMZF2 (nonzinc finger domain). We could not detect any expression of BMZF2 in adult tissues (data not shown). We could only detect BMZF2 expression in fetal tissues, where we observed three major transcripts. Two of these transcripts, of ϳ5.0 and ϳ4.0 kb, were detected in fetal brain, lung, liver, and kidney (Fig.  4A). A shorter transcript of ϳ3.4 kb was detected in fetal lung tissue (Fig. 4A). We have not determined the structure of the three different transcripts, but these may arise from alternative transcription initiation, splicing, or differential use of polyadenylation sites. This raises the possibility that several protein isoforms may exist for BMZF2. As a control for the amount of RNA present in each lane, we reprobed the Northern blot with a human ␤-actin probe (Fig. 4A).
The subcellular localization of BMZF2 was also examined. CMV/BMZF2-(1-622) and CMV/WT1(Ϫ/Ϫ) were transfected into 293 cells. Cells were lysed, and cytosolic and nuclear protein fractions were prepared and subjected to SDS-PAGE and Western blot analysis. As shown in Fig. 4B, BMZF2 and WT1 were present in the nuclear fraction (Fig. 4B, upper and  middle panel). As a further control for the quality of the fractionation procedure, a cytoplasmic protein, Grb2 (28), was detected only in the cytoplasmic fraction (Fig. 4B, lower panel). Immunofluorescence analysis of transfected H1299 cells with CMV/BMZF2-(1-622) revealed the presence of BMZF2 in the nucleus (data not shown). Taken together, these results identify BMZF2 as a nuclear protein.
Inhibition of WT1-mediated Activation by BMZF2-To investigate the functional consequences of the WT1-BMZF2 interaction, we examined whether introduction of BMZF2 would affect transcriptional activation by WT1. WT1(ϪKTS) isoforms have been shown previously (25) to activate a reporter construct containing a WT1-binding site within the human vitamin D receptor (VDR) promoter. A series of reporter and expression constructs were utilized to analyze the effect of BMZF2 on the functional properties of WT1 (Fig. 5A). When Ϫ960phVDR/ Luc was transfected with the empty expression vector pcDNA3, very little luciferase activity was observed (Fig. 5B,  lane 2). Transfection with CMV/WT1(ϪKTS) resulted in a 5.2-fold activation of the VDR promoter (Fig. 5B, lane 3), similar to results reported previously (25). Transfection of CMV/BMZF2-(1-622) with Ϫ960phVDR/Luc did not significantly affect the levels of luciferase produced from Ϫ960ph-VDR/Luc (lane 4) indicating that under these conditions BMZF2 does not affect expression from the VDR promoter. Transfection of increasing amounts of CMV/BMZF2-(1-622) resulted in a dose-dependent decrease in WT1-mediated transcriptional activation (Fig. 5B, compare lanes 5-8). Western blotting of nuclear extracts from the transfected cells demonstrated that increasing amounts of CMV/BMZF2-(1-622) were synthesized in response to increasing amounts of transfected plasmid (Fig. 5C). These results indicate that BMZF2 can inhibit WT1-mediated activation.
To demonstrate that BMZF2 binding to WT1 was required to obtain the observed effects on WT1-mediated transcriptional activation, we constructed an BMZF2 deletion mutant that lacks the WT1-binding domain (amino acids 231 to 370), called CMV/BMZF2-(⌬231-370). As with CMV/BMZF2-(1-622), transfection of CMV/BMZF2-(⌬231-370) with Ϫ960phVDR/Luc had little effect on VDR promoter activity (Fig. 5E . Taken together, these results suggest that BMZF2 specifically inhibits WT1-mediated transcriptional activation through its physical association with WT1. BMZF2 Is a Transcriptional Repressor-To assess the transcriptional properties of BMZF2, BMZF2 was fused to the Gal4 DNA-binding domain, generating CMV/GAL4/BMZF2. Two reporter plasmids were used in these experiments, pTECAT/ 5XGAL4, which contains five GAL4-binding sites upstream of the TK minimal promoter, and pTECAT, which lacks GAL4binding sites and serves as a negative control (Fig. 6A). An expression vector driving the synthesis of only the GAL4 DNAbinding domain, CMV/GAL4, had no effect on the levels of CAT produced from either pTECAT or pTECAT/5XGAL4 when transfected into COS-7 cells (Fig. 6B, lanes 2-5 and 11-14). On the representation of the constructs. [ 35 S]Methionine-labeled BMZF2 and deletion mutants produced by in vitro translation were incubated with immobilized GST-WT1ZF. Following washing, the bound proteins were eluted with SDS loading buffer, and proteins were analyzed by 10% SDS-PAGE. Gels were treated with EN 3 HANCE, dried, and proteins visualized by autoradiography. The ϩ or Ϫ symbols to the right refer to the ability or inability, respectively, to bind to GST-WT1ZF.  other hand, CMV/GAL4/BMZF2 repressed CAT production from pTECAT/5XGAL4 in a dose-responsive manner (lanes 6 -9) but did not affect expression from pTECAT (lanes [15][16][17][18], indicating that repression by CMV/GAL4/BMZF2 is dependent on the presence of GAL4-binding sites within the reporter vector. Additionally, CMV/BMZF2-(1-622), which lacks a GAL4 DNA-binding domain, did not inhibit CAT expression from pTECAT/5XGAL4 (Fig. 6C, compare lane 4 to 3 and 2), consistent with the need for BMZF2 to bind to the reporter construct to achieve repression of transcription. These results demonstrate that BMZF2 is capable of repressing transcription. DISCUSSION The identification of BMZF2 as a WT1 interacting partner was achieved by affinity chromatography of HeLa extracts on immobilized WT1ZF. Although the BMZF2 transcript should produce a protein of 72 kDa, we identified BMZF2 through mass spectrometry analysis of a ϳ30-kDa protein species ( . The amounts of transfected plasmids are indicated below the lanes. The total transfected DNA concentration was kept constant by the addition of the empty expression vector, pcDNA3, to make up for differences in amounts between transfections. The error bars represent the S.D. of three independent experiments, with each sample transfected in duplicate. Luciferase activity of each transfection was set relative to the activity obtained by transfecting pcDNA3, Ϫ960phVDR/Luc, and pRSV/␤-galactosidase (lane 2; which was set at 1). 6. BMZF2 is a transcriptional repressor. A, schematic representation of the pTECAT and pTECAT/5XGAL4 reporter constructs used in this experiment. The white box denotes the thymidine kinase (TK) promoter, the box with the gray gradient represents the 5 GAL4-binding sites, and the CAT open reading frame is illustrated by a black box. The right-angled arrow indicates the transcription start site. Nucleotide positions demarcate the boundaries of the TK promoter, which are identical in the two reporter constructs. B, COS-7 cells were co-transfected with the indicated amounts of pTECAT/5XGAL4 or pTECAT and increasing amounts of CMV/GAL4 or CMV/GAL4/BMZF2 vector. The amounts of transfected plasmids are indicated below the lanes. The total transfected DNA concentration was kept constant by the addition of the empty expression vector, pcDNA3, to make up for differences in amounts between transfections. The error bars represent the S.E. of three separate experiments, with each sample transfected in duplicate. CAT activity of each transfection was set relative to the activity obtained by transfecting pcDNA3, pTECAT/5XGAL4, and pRSV/␤-galactosidase (lane 1; which was set at 1). C, COS-7 cells were co-transfected with the indicated amounts of pTECAT/5XGAL4 and CMV/GAL4, CMV/GAL4/BMZF2, or CMV/BMZF2-(1-622) expression vector. The amounts of transfected plasmids are indicated below the lanes. The total transfected DNA concentration was kept constant by the addition of the empty expression vector, pcDNA3, to make up for differences in amounts between transfections. To normalize for transfection efficiency, the cells were co-transfected with 1 g of pRSV/␤-gal. At 48 h after transfection, the cells were harvested and assayed for ␤-galactosidase and CAT activity. The average fold activation and S.E. for CAT determinations are indicated below the representative chromatogram and represent the value obtained from two independent experiments. CAT activity of each transfection was set relative to the activity obtained by transfecting pcDNA3, pTECAT/5XGAL4, and pRSV/␤-galactosidase (lane 1; which was set at 1). Ac-Cm, acetylated chloramphenicol; Cm, chloramphenicol; O, origin. 1). Native immunoprecipitations from K562 extracts revealed the association of several anti-BMZF2 cross-reacting protein species with WT1 (Fig. 2). The simplest interpretation of our data is that several BMZF2 isoforms exist and can interact with WT1 (Fig. 2). We do not know the origin of the different BMZF2 isoforms.

FIG.
The coding region of BMZF2 is 1866 nucleotides (Fig. 1B). Although 36 nucleotides upstream of the postulated BMZF2 initiation codon (GenBank TM accession number NM005774) is a TAA stop codon, indicating that the postulated ATG is correct, the deposited cDNA sequence (GenBank TM accession number NM005774; 3 kbp) contains a long (1086 nucleotides) 5Ј-untranslated region and 14 upstream open reading frames. These features are unusual for a eukaryotic mRNA and may indicate sophisticated post-transcriptional regulation of BMZF2 or the presence of an intron at the 5Ј end of the cDNA, in which case the protein sequence presented might be partial.
BMZF2 has 18 contiguous zinc finger motifs homologous to Krü ppel-type C 2 -H 2 zinc fingers. Initial sequence analysis of the amino-terminal domain of BMZF2 by Han et al. (19) suggested the presence of a Krü ppel-related novel box, which they nominated as KRNB. We demonstrate here that the KRNB motif of BMZF2 is also a transcriptional repressor module (Fig.  6). The BMZF2 amino-terminal domain shows homology to other C 2 -H 2 zinc finger proteins, including ZNF 224 (66% identity over 92 amino acids) (GenBank TM accession number AAF04106), ZNF 221 (51% identity over 92 amino acids) (Gen-Bank TM accession number XP_009237), ZNF 222 (47% identity over 92 amino acids) (GenBank TM accession number AAF66075), and ZNF 155 (25% identity over 92 amino acids) (GenBank TM accession number AAF18684). Although little functional analysis of these proteins has been performed, the presence of a KRNB domain within these proteins would suggest that they are also transcriptional repressors.
We confirmed the association of BMZF2 and WT1 by GST pulldown assays, far Western blot assays, and co-immunoprecipitations (Figs. 2 and 3). We have performed the far Western experiments with nuclease-treated lysate (Fig. 3D), suggesting that the interaction between the two proteins is not mediated by nucleic acid. It is interesting that 5 of 18 zinc fingers of BMZF2 are required for interacting with WT1, implying functional differences among the BMZF2 zinc fingers. It is recognized that zinc finger motifs can participate in protein-protein interactions. For example, the Ikaros protein plays a central role in the development of lymphoid cells and is capable of mediating both DNA-binding and protein-protein interactions (30). Ikaros contains up to 6 Krü ppel-like fingers (depending on splicing patterns), with the first four involved in sequencespecific DNA binding and the last two involved in homodimerization and binding to a second zinc finger protein, Aiolos (31). Dimerization of Ikaros and Aiolos modulates their respective ability to bind DNA and activate transcription. Furthermore, the Krü ppel-like zinc fingers of the transcription factors Sp1 and EKLF appear also to be capable of binding both DNA and proteins simultaneously (32).
Several previous reports (33,34) have identified protein partners that interact with WT1 through either the aminoterminal domain or through the zinc finger domain. The WT1 protein can homodimerize through association via its aminoterminal domain. SF-1 (35), UBC9 (36), and Hsp70 (37) have been shown to interact with the amino-terminal domain of WT1. The interaction of WT1 with these factors has been shown to lead to elevated transcription of downstream genes (e.g. Mü llerian inhibitory substance) due to competition of WT1 with Dax-1 for SF-1 and to the promotion of WT1-mediated growth arrest due to its interaction with Hsp70 (37). Proteins that interact with WT1 through the zinc finger domain include p53, p73, p63, Par-4, Ciao 1, CBP/P300, and U2AF65 (reviewed in Refs. 14 and 15). There is a functional consequence of WT1 interacting with BMZF2, inhibition of WT1-mediated transactivation (Fig. 5).
BMZF2 is expressed in fetal brain, lung, liver, and kidney (Fig. 4). We failed to detect any expression by Northern blotting in adult tissues (Fig. 4). Although Han et al. (19) reported that BMZF2 is expressed to low levels in the heart, brain, lung, kidney, testis, bone marrow, liver, spleen, pancreas, stomach, placenta, and in six leukemic cell lines, they had to resort to RT-PCR to detect expression in these tissues. It is tempting to speculate that BMZF2 is a developmentally regulated transcriptional repressor that is not present in many adult tissues or is present to very low levels in these tissue. We observed three transcripts (3.4, 4, and 5 kb) by Northern blotting of RNA isolated from fetal tissue utilizing the unique amino-terminal domain of BMZF2 as a probe and under stringent hybridization and wash conditions. The 3.4-kb transcript was present in lung tissue and absent in brain, liver, and kidney. The other two transcripts were present in all four tissues. We have yet to explain the underlying structure features that are responsible for generating these three different transcripts.
Hematopoiesis is a complex physiological process that requires intricate regulation of gene expression during embryogenesis, fetal life, and adult life. BMZF2 was first identified from an acute promyelocytic leukemia cell line, NB4 and is expressed in a number of leukemia cell lines (19). WT1 is expressed in early bone marrow precursors and rapidly downregulated following differentiation of these cells and leukemiaderived cell lines, suggesting that it may also play a role in early hematopoiesis (38 -41). WT1 is also highly expressed in many human acute leukemias, suggesting that mis-expression of WT1 may also be a contributor to hematopoietic malignancies (42,43). Recently, WT1 has also been shown to induce growth arrest and differentiation in primary hematopoietic progenitors (44). Conversely, the loss of wt1 gene function has also been implicated in the development of malignancies including acute leukemias (45). This correlates with the tumorsuppressive effects of WT1 expression in leukemia cell lines and suggests that WT1 acts as a differentiation-promoting gene during hematopoiesis and that the loss of functional WT1 expression may contribute to leukemogenesis. Understanding the molecular relationship between WT1 and BMZF2 may provide insight into the role of these two proteins in normal hematopoiesis, as well as understanding events that become deregulated during development of hematopoietic malignancies.