Involvement of a Novel Zinc Finger Protein, MIZF, in Transcriptional Repression by Interacting with a Methyl-CpG-binding Protein, MBD2*

MBD2, a methyl-CpG-binding protein, is a component of the MeCP1 histone deacetylase (HDAC) complex and plays a critical role in DNA methylation-mediated transcriptional repression. To understand the molecular basis of the methylation-associated repression, we attempted to identify MBD2-interacting proteins by a yeast two-hybrid system. Using MBD2 as bait, we isolated a novel zinc finger protein, referred to as MIZF. A direct interaction between MBD2 and MIZF was confirmed by in vitro binding assays and immunoprecipitation experiments. Four of seven zinc fingers present in the C-terminal region of MIZF are required for binding with MBD2. The MIZF mRNA is expressed in all human tissues and cell lines examined. The subcellular localization of MIZF is distinct from that of MBD2, although both proteins co-localize in some areas of the nuclei; MIZF localizes diffusely in the nucleoplasmic region, whereas MBD2 preferentially localizes in major satellites. A reporter assay demonstrated that MIZF significantly abrogates transcriptional activities. This repression is attenuated by an HDAC inhibitor, trichostatin A, and is completely dependent on the interaction with MBD2. These results suggest that MIZF is abundantly present in cells and functions as a negative regulator of transcription by binding to MBD2 and recruiting HDAC-containing complexes.

MBD2, a methyl-CpG-binding protein, is a component of the MeCP1 histone deacetylase (HDAC) complex and plays a critical role in DNA methylation-mediated transcriptional repression. To understand the molecular basis of the methylation-associated repression, we attempted to identify MBD2-interacting proteins by a yeast two-hybrid system. Using MBD2 as bait, we isolated a novel zinc finger protein, referred to as MIZF. A direct interaction between MBD2 and MIZF was confirmed by in vitro binding assays and immunoprecipitation experiments. Four of seven zinc fingers present in the C-terminal region of MIZF are required for binding with MBD2. The MIZF mRNA is expressed in all human tissues and cell lines examined. The subcellular localization of MIZF is distinct from that of MBD2, although both proteins co-localize in some areas of the nuclei; MIZF localizes diffusely in the nucleoplasmic region, whereas MBD2 preferentially localizes in major satellites. A reporter assay demonstrated that MIZF significantly abrogates transcriptional activities. This repression is attenuated by an HDAC inhibitor, trichostatin A, and is completely dependent on the interaction with MBD2. These results suggest that MIZF is abundantly present in cells and functions as a negative regulator of transcription by binding to MBD2 and recruiting HDACcontaining complexes.
Methylation of CpG dinucleotides is the major epigenetic modification in higher eukaryotic genomes. This reaction is mediated by DNA methyltransferases (1,2), and the biological importance of the CpG methylation is directly demonstrated by the fact that mice lacking the gene of a DNA methyltransferase exhibit a defect in embryogenesis at midgestation (3). CpG methylation also plays important roles in a wide range of biological steps, including tissue-specific gene expression, Xchromosome inactivation, and genomic imprinting (4 -8). In addition, abnormalities in CpG methylation have been linked to altered gene expression in certain genetic diseases, tumorigenesis, and senescence; aberrant methylation of a tumor suppressor gene, p16 ink4A , is closely related to tumor growth characteristics (9 -11). The state of gene methylation is associated with transcriptional repression (12), and the biological consequence of CpG methylation is mediated by a family of methyl-CpG-binding proteins (MeCPs), 1 which contain a common methyl-CpG binding domain (MBD) (13)(14)(15). Although the molecular basis for MBD-dependent repression is still unclear, so far five MBD-containing proteins have been reported in mammals: MeCP2, MBD1, MBD2, MBD3, and MBD4. MeCP2 is the first characterized protein that can bind to methylated CpG pairs (16 -18). MBD1 and MBD2 bind specifically to a symmetric methyl-CpG and function as transcriptional repressors (19,20), whereas MBD3 is a subunit of the nucleosome remodeling histone deacetylase (NuRD) complex that includes Mi-2, HDAC1/2, RbAp46/48, and metastasis-associated protein 2 (21,22). On the other hand, MBD4 (also known as MED1) binds preferentially to methyl-CpG⅐TpG mismatches and removes thymine from a mismatch methyl-CpG site, suggesting that MBD4 may be involved in DNA mismatch repair and the maintenance of genome stability (23).
Much evidence has accumulated to indicate a close association between methylation-dependent repression and histone deacetylation. Among these MBD proteins, MeCP2 functions as a transcriptional repressor in concert with Sin3A, histone deacetylase 1 (HDAC1), and HDAC2. MBD2 is also likely to function as a molecular link between methyl-CpG and HDACs, because MBD2 is a component of the MeCP1 HDAC complex that includes RbAp46/48, Sin3A, SAP30, and SAP18 (24), and MBD2 can also interact with the NuRD complex (20). To understand the precise mechanisms for the recruitment of the two HDAC complexes by MBD2 and for MBD2-dependent transcriptional repression, further identification and characterization of novel proteins associated with MBD2 is required. In the present study, we performed a yeast two-hybrid screening using mouse MBD2 as bait to search for MBD2-interacting molecules and found a novel zinc finger protein. This MBD2-binding zinc finger (MIZF) protein represses transcription by associating with MBD2 and a histone deacetylase complex. The results suggest that MIZF, in concert with MBD2, recruits HDAC complexes, which in turn results in transcriptional repression.
3-aminotriazole as described (25). The colonies were tested for ␤-galactosidase activity, and positive clones were analyzed for the insert sequences. To isolate the full-length cDNA, 5Ј-rapid amplification of cDNA ends combined with nested polymerase chain reaction was performed using a human fetal brain cDNA library (Marathon-Ready cDNA, CLONTECH) as a template.
Plasmid Construction-For the expression of a green fluorescent protein (GFP) fused to MBD2, the MBD2b cDNA was subcloned into the appropriate sites of pEGFP-C1 (CLONTECH) to obtain pGFP-MBD2. Epitope-tagged derivatives of MBD2 and MIZF, containing amino-terminal FLAG and carboxyl-terminal Myc tags, respectively, were generated using a pcDNA3 expression vector (Invitrogen, San Diego, CA). For the preparation of glutathione S-transferase (GST) fusion recombinant proteins, MBD2b and MIZF cDNA were subcloned into pGEX-4T-1 (Amersham Pharmacia Biotech) (GST-MBD2 and GST-MIZF). Deletion derivatives of MBD2b were generated by inserting the cDNA fragments into pGEX-4T-1 using EcoRI and HincII for GST-N1-52, EcoRI and NcoI for GST-N1-154, EcoRI and HindIII for GST-N1-212, HincII and SalI for GST-C53-262, NcoI and SalI for GST-C154 -262, and HindIII and SalI for GST-C212-262. To transcribe and translate MBD2 and MIZF in vitro, the MBD2b and MIZF cDNAs were subcloned into the appropriate sites of pGBKT7 (CLONTECH). Deletion constructs of MIZF were generated by inserting the cDNA fragments into pGBKT7 using PstI and XbaI for MIZF-N1-429, EcoRV and BamHI for MIZF-C125-517, and XbaI and BamHI for MIZF-C429 -517. For MIZF-N1-201 and MIZF-C201-517, MIZF cDNA was mutagenized by polymerase chain reaction to create a SphI site at codon 201 using the primer 5Ј-GTGGTAGCATGCCCCACCTGTG. The polymerase chain reaction products were digested and then subcloned into the corresponding sites of pGBKT7. Plasmids expressing fusion proteins of the GAL4-DBD and MBD2, MIZF, the N-terminal half of MIZF (N1-201), or the C-terminal half of MIZF (C201-517) were constructed using the pM1 expression vector described elsewhere (26). A firefly luciferase reporter plasmid, pGL3-G5pol␤, was constructed as follows. A 496-base pair fragment of the human DNA polymerase ␤ promoter (Ϫ422 to ϩ73) (27) was amplified with specific primers 5Ј-ATCTCTAGAGAAAGTTTTGACAGT-GTGACG and 5Ј-CAAGCTTGAAGGAGGTACCAGGACTTGGAG, digested with XbaI and HindIII, and then subcloned into the corresponding sites of pGEM-3Z (Promega, Madison, WI) to generate pGEM-pol␤. The KpnI-NheI fragment containing five GAL4-binding sites excised from pG5luc (Promega) was inserted into the KpnI and XbaI sites of pGEM-pol␤, and the insert was excised with SacI and HindIII and subcloned into the corresponding sites of the pGL3-enhancer (Promega).
In Vitro Binding Assay-GST fusion proteins were expressed in Escherichia coli BL21 (DE3) and purified using glutathione-Sepharose beads (Amersham Pharmacia Biotech). [ 35 S]Methionine-labeled fulllength MIZF, various deletion mutants of MIZF, and MBD2 were synthesized using the indicated pGBKT7 plasmids as templates in a TNT T7 quick coupled transcription-translation system (Promega). The labeled, in vitro-translated proteins were incubated with various GST fusion proteins immobilized on glutathione-Sepharose beads for 4 h at 4°C in buffer containing 25 mM Tris-HCl (pH 7.2), 150 mM NaCl, 0.2% Nonidet P-40, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 1 g/ml leupeptin. After washing with the same buffer, the bound proteins were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE).
Subcellular Localization-COS-7 cells growing on glass coverslips were transfected with pGFP-MBD2 or pMIZF-Myc using Effectene Transfection Reagent (Qiagen, Hilden, Germany) as described elsewhere (28). The cells were fixed with 3.7% formaldehyde in phosphatebuffered saline and permeabilized with 0.2% Triton X-100 in phosphate-buffered saline for 5 min. For observation of Myc-tagged MIZF, cells were incubated with anti-Myc antibody (9E10) and stained with fluorescein isothiocyanate-conjugated goat anti-mouse antibody (BIO-SOURCE, Camalliro, CA). The cell preparation was observed using a confocal laser-scanning microscope.
Immunoprecipitation and Western Blot Analysis-Transfected 293 Tissues including brain, heart, skeletal muscle (Sk. muscle), colon, thymus, spleen, kidney, liver, small intestine, placenta, lung, and peripheral blood leukocyte (PBL) were analyzed by hybridization with a digoxigenin-labeled MIZF cDNA probe (upper panel). The same filter was rehybridized with a digoxigenin-labeled human glyceraldehyde-3phosphate dehydrogenase cDNA probe to quantify RNA loading (lower panel). Similarly, human cell lines were analyzed by hybridization with a MIZF cDNA probe (upper panel) or a glyceraldehyde-3-phosphate dehydrogenase (GAPDH) cDNA probe (lower panel). Size markers and the corresponding mRNA are indicated on the left and right sides, respectively. C, subcellular localization of MIZF. COS-7 cells transiently transfected with pMIZF-Myc (a and b) or pGFP-MBD2 (c and d) were stained for MIZF with anti-Myc antibody. Each pair of panels shows a fluorescence image (a and c) and the corresponding difference interference contrast image (b and d). The bar represents 20 m. cells were lysed in buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.1% Nonidet P-40, 1 mM dithiothreitol, 1 mM EDTA, 0.1 mM phenylmethylsulfonyl fluoride, and 1 g/ml leupeptin) and centrifuged. The lysates were immunoprecipitated with the indicated antibodies and protein G-Sepharose (Zymed Laboratories Inc., San Francisco, CA) for 4 h at 4°C, and the resultant immunoprecipitates were subjected to SDS-PAGE. Proteins were transferred to a filter and incubated with either anti-Myc, anti-HDAC1 (Santa Cruz Biotechnology), anti-FLAG (Sigma), or anti-GFP antibody (MBL, Nagoya, Japan), followed by a horseradish peroxidase-conjugated anti-mouse (Promega) or anti-rabbit IgG polyclonal antibody (Dako, Kyoto, Japan). The positive bands were visualized using an enhanced chemiluminescence system (Amersham Pharmacia Biotech).
Northern Blot Analysis-RNA was isolated from several human cell lines by the guanidine isothiocyanate method using Isogen reagent (Wako, Osaka, Japan). Total RNA (20 g) was resolved in 1% agarose gels and transferred onto a positively charged nylon membrane (Amersham Pharmacia Biotech). Hybridization and detection were performed using a digoxigenin-labeled MIZF cDNA probe (nucleotides 745-1704) (Roche Diagnostics, Tokyo, Japan). The tissue expression pattern of MIZF mRNA was also examined with a human multiple tissue Northern blot (CLONTECH). All blots were stripped and reprobed with a cDNA fragment encoding the constitutively expressed human glyceraldehyde-3-phosphate dehydrogenase as a control.
Luciferase Assay-Human 293 cells plated on 24-well plates were transfected with 50 ng of pGL3-G5pol␤, 5 ng of pRL-TK (Promega), and the indicated expression plasmids. Total amounts of expression plas-mids were normalized by pcDNA3. After incubation for 24 h, transfected cells were split into two dishes and further incubated in the presence or absence of 100 ng/ml trichostatin A (TSA) (Wako) for 10 h. Cell lysates were prepared, and dual luciferase assays were carried out using the dual luciferase reporter assay system (Promega).
Histone Deacetylase Assay-Histone deacetylase activity was measured using mouse core histones radiolabeled with [ 3 H]acetate in vivo as a substrate (29). Aliquots (2 l) of eluates from the immunoprecipitates were incubated with 10 g of [ 3 H]histones in 100 l of assay buffer (20 mM Tris-HCl, pH 8.0, 75 mM NaCl, 1 mM dithiothreitol) for 30 min at 37°C. The reaction was stopped by the addition of 20 l of 12 N HCl. The released [ 3 H]acetic acid was extracted with 1 ml of ethyl acetate and measured by scintillation counting of the solvent layer.

RESULTS
Identification of MIZF-When a total of 5 ϫ 10 6 clones were tested in a yeast two-hybrid screening system with MBD2 cDNA, ϳ300 cDNA clones were obtained as first candidates exhibiting His(ϩ) and Ade(ϩ) properties. Most of these were, however, eliminated by a second screening for their ability to activate the ␤-galactosidase gene. One clone showed strong ␤-galactosidase activity and was chosen for further analysis. To obtain the full-length sequence of the clone, 5Ј-rapid amplification of cDNA ends combined with polymerase chain reaction was performed using a human fetal cDNA library. Finally, we obtained a cDNA comprising 2285 base pairs including an open reading frame encoding a polypeptide of 517 amino acids with a predicted molecular mass of 59.7 kDa. BLAST search revealed that the cDNA is identical to human cDNA MGC:4317 (accession number BC001945). This cDNA encodes a novel protein with homology to zinc finger proteins. Therefore, we name this protein MIZF (MBD2-interacting zinc finger protein). As shown in Fig. 1, the MIZF protein contains seven zinc finger domains similar to the C2H2 zinc finger motif (30) and a stretch of negatively charged amino acids residues from position 49 to 63.
Tissue and Subcellular Expression of MIZF-Expression of MIZF was examined by Northern blot analysis with RNA samples from various human tissues and cell lines. The expression level of MIZF was detectable as a predominant single signal in all tissues and cell lines examined (Fig. 2, A and B). The apparent size of the MIZF transcript is ϳ2.3 kilobase, consistent with the size of the cDNA clone. Among tissues examined, the highest level of MIZF was detected in the brain, heart, skeletal muscle, and kidney, with moderate levels of the transcript seen in the colon, thymus, spleen, liver, small intestine, placenta, and lung ( Fig. 2A). The MIZF transcript was detected in all cell lines examined at relatively constant levels (Fig. 2B).
We examined the subcellular localization of MIZF and compared it with that of MBD2. When COS-7 cells were transiently transfected with pMIZF-Myc, MIZF localized exclusively to the cell nucleus except the nucleolus (Fig. 2C, a). The nuclear localization of MIZF proteins was confirmed by difference interference contrast image (Fig. 2C, b). Consistent with the previous observations on MBD2, GFP-MBD2 displays a diffuse nucleoplasmic staining pattern with prominent nuclear dots that are known to be highly methylated regions of the genome (Fig. 2C, c and d) (15). These results indicate that MIZF is a nuclear protein with subcellular localization distinct from that of MBD2.
Association of MIZF with MBD2-The association of the MIZF protein with MBD2 was confirmed by in vitro binding assays. The MIZF protein was produced by an in vitro translation system and tested for binding to GST-MBD2, and vice versa. As shown in Fig. 3A, MIZF bound to GST-MBD2 but not to GST alone. Similarly, the in vitro translated MBD2 protein bound to GST-MIZF but not to GST alone (Fig. 3B). This association was further confirmed in 293 cells subjected to cotransfection with pMIZF-Myc and pGFP-MBD2. When GFP-MBD2 was immunoprecipitated with anti-GFP antibody, MIZF-Myc (68 kDa) was present in the resultant immunoprecipitates (Fig. 3C). Similarly, GFP-MBD2 was detected in immunoprecipitates with anti-Myc antibody. These results demonstrate that MIZF associates specifically with MBD2 in vitro, as well as inside the cell.
To determine the region of MIZF responsible for binding to MBD2, several deletion mutants were examined (Fig. 4A). As shown in Fig. 4B, C125-517 and C201-517 could bind to MBD2 (lanes 10 -15). The removal of an additional 228 amino acids (C429 -517) resulted in a complete loss of the binding activity to MBD2 (lanes 16 -18), indicating that the N-terminal ϳ200 amino acids are dispensable for the interaction with MBD2. In fact, a mutant N1-201 failed to bind to MBD2 (lanes 4 -6).
Deletion of the C terminus (N1-429) did not affect the binding ability, suggesting that the C-terminal 89 amino acids are also dispensable for the interaction with MBD2 (lanes 7-9). These results indicate that the region comprising amino acids 201-429 of the MIZF protein, containing four zinc finger domains (ZF4 -ZF7) (Fig. 1B), is required for its association with MBD2.
A series of MBD2 deletion mutants was also examined to map the region responsible for the interaction with MIZF (Fig.  5A). MIZF could bind to the MBD2 mutant C53-262, which lacks the 52 amino acids at the N terminus (Fig. 5B, lane 7), but not to mutants N1-52, C154 -262, and C212-262, indicating that the N-terminal 52 amino acids are not responsible for the binding to MIZF (lanes 4, 8, and 9). In addition, MIZF could bind to deletion mutants N1-154 and N1-212, indicating that the C-terminal 108 amino acids are dispensable for the interaction with MIZF ( lanes 5 and 6). Thus, the region from amino acid 53 to 154 of MBD2 is necessary for the association with MIZF.
Involvement of MIZF in Transcriptional Repression-The effect of MIZF on transcription was examined using a reporter assay with a DNA polymerase ␤ promoter and GAL4-binding sites (Fig. 6A). When the full-length MIZF cDNA fused with GAL4-DBD was transfected into cells, transcription of the reporter gene was significantly inhibited by MIZF in a dose-dependent manner (Fig. 6B). Similar results were obtained using an MIZF mutant, C201-517, which possesses binding activity to MBD2. These results suggest that MIZF functions as a negative regulator for transcription and that MIZF-mediated transcriptional repression is dependent on the binding to MBD2, and, in turn, the association with an HDAC complex. In contrast, another MIZF mutant, N1-201, which does not bind to MBD2, did not repress the transcription (Fig. 6B). In addition, the results shown in Fig. 6B demonstrate that transcriptional repression by MIZF can be substantially relieved by TSA, an inhibitor of histone deacetylases. To examine whether MIZF interacts with endogenous HDAC1, an immunoprecipitation experiment followed by Western blotting analysis was performed on lysates from 293 cells transiently expressing MIZF-Myc. As expected, HDAC1 was detected in the immunoprecipitates containing MIZF. MIZF-Myc was specifically coimmunoprecipitated with HDAC1 by an anti-HDAC1 antibody but not by normal rabbit serum (Fig. 6C).
Enhancement of the MBD2-mediated Repression by MIZF-To confirm whether MIZF can modulate MBD2-mediated repression, a luciferase reporter assay was performed in the presence or absence of MIZF. MIZF enhanced the MBD2mediated repression in a dose-dependent manner (Fig. 7A). Basal transcription of GAL4-DBD alone was not affected by MIZF. To investigate the involvement of HDAC in this effect of MIZF, cells expressing FLAG-MBD2 and MIZF-Myc were analyzed by an immunoprecipitation assay. As shown in Fig. 7, B and C, the expression of MIZF produced significant increases in the levels of both the HDAC1 protein and HDAC activity, an effect that was TSA-sensitive. These results suggest that MIZF may enhance MBD2-mediated repression by recruiting HDAC activity to the complex. DISCUSSION A close relationship between methylation-dependent repression and histone deacetylation has been demonstrated. The transcriptional effects of CpG methylation are mediated by MBD-containing proteins. Among MBD proteins, MeCP2 is known to bind to methylated CpG pairs and functions as a transcriptional repressor by recruiting Sin3A and HDACs (16 -18). MBD2, another methyl-CpG binding protein, has also been shown to be a component of the MeCP1-HDAC complex that includes RbAp46/48 and Sin3A (24). In addition, MBD2 interacts with the NuRD complex (21). These findings suggest that MBD2 is likely to function as a molecular link between methyl-CpG and HDACs. To understand the mechanisms underlying the MBD2-dependent transcriptional repression and recruitment of HDACs, we identify MBD2-interacting molecules using a yeast two-hybrid system. When a human fetal brain cDNA library was screened using mouse MBD2 as bait, a novel zinc finger protein, MIZF, was specifically isolated. This MIZF pro-tein binds to MBD2 and represses the promoter-driven transcription activity. Like the MBD2 protein, the MIZF protein is expressed in a wide variety of tissues and cells. These results suggest that MIZF is expressed abundantly in cells and functions as a negative regulator for transcription by binding to MBD2 and to HDAC systems.
We demonstrated that the C-terminal region of MIZF (amino acids 201-517) is required for the association with MBD2 (Fig.  4) and for the TSA-sensitive repression (Fig. 6). The N-terminal region of MIZF (amino acids 1-201), which lacks MBD2-binding activity, does not repress transcription (Figs. 4 and 6). These results suggest that MIZF-derived repression is totally dependent on binding to MBD2 and HDACs. Indeed, MIZF was coimmunoprecipitated with endogenous HDAC1 (Fig. 6). It has become clear that several transcriptional repressors, such as YY1, Bcl-6, and Rb family proteins, are able to associate directly with HDAC1 or HDAC2 (31)(32)(33). The minimal structures required for the association with HDAC are thus far identified as the LXCXE-like motif in Rb family proteins, a 30-amino acid glycine-rich region in YY1, and the POZ domain in Bcl-6. However, MIZF does not possess regions homologous to these motifs. In addition, we could not detect the interaction with any known HDACs in a yeast two-hybrid assay when MIZF cDNA was used as bait, 2 indicating that MIZF is unable to associate directly with HDACs. It has been reported that MBD2 binds to MBD3, an integral component of the NuRD complex (34), and also interacts with a distinct HDAC-containing complex, MeCP1 (24). Therefore, it is rather possible that MIZF binds to MBD2, which in turn recruits HDACs to the protein complex. A series of associations of these proteins would then result in transcriptional repression. This is supported at least in part by the finding that the coexpression of MIZF and MBD2 significantly enhances HDAC protein recruitment and activity (Fig. 7).
MIZF is a unique, C2H2-type zinc finger protein involved in methylation-dependent transcriptional repression. The C2H2 zinc finger structure was initially identified in DNA-binding molecules such as the transcription factors TFIIIA and Krü ppel, and at present a number of transcription factors are known to utilize C2H2 zinc fingers as DNA-binding domains (35). Most of these proteins contain multiple fingers, which are required for the recognition of a specific range of DNA sequences. Concerning an analogy with these proteins, it is possible that MIZF is a DNA-binding protein, and our preliminary results reveal that MIZF binds to some DNA sequences (data not shown). Thus, it is probable that MIZF functions as a repressor, in terms of its suppressive effect on transcription and DNA binding activity. As shown in Fig. 2C, we demonstrate here that MIZF and MBD2 are expressed quite abundantly and ubiquitously in cells and tissues, suggesting their important functions in cell. Although these two proteins colocalize in some areas in nuclei, their main localizations are quite different; MIZF localizes diffusely in the nucleoplasmic region, whereas MBD2 preferentially localizes in the major satellite, which is known to be a highly methylated region of the genome. Thus, it is possible that MIZF functions as a repressor in the regulation of the transcription of specific genes. Identification of the DNA sequences that interact with MIZF is necessary to understand this issue.
It has become clear that zinc fingers are also involved in protein-protein interactions. Homodimerization of the multifinger protein Ikaros is mediated by its two C-terminal fingers (36). Likewise, metastasis-associated protein 2, a subunit of the NuRD complex, which contains a C4-type zinc finger, also modulates HDAC activity by interacting with MBD3 (20). MIZF contains seven zinc fingers (ZF1-ZF7), and four fingers in the C-terminal region are required for the interaction with MBD2. Our preliminary study demonstrates that MIZF also associates with MBD3 in a manner similar to its binding with MBD2 (data not shown). MBD3 is a component of the NuRD complex (34). Thus, it is possible that MIZF recruits the NuRD complex through its interaction with MBD3. The analysis of MIZF-interacting proteins may provide a clue to understanding the mechanism of transcriptional repression induced by MIZF.