Transcription Factor hDREF Is a Novel SUMO E3 Ligase of Mi2α*

The human transcription factor DNA replication-related element-binding factor (hDREF) is essential for the transcription of a number of housekeeping genes. The mechanisms underlying constitutively active transcription by hDREF were unclear. Here, we provide evidence that hDREF possesses small ubiquitin-like modifier (SUMO) ligase activity and can specifically SUMOylate Mi2α, an ATP-dependent DNA helicase in the nucleosome remodeling and deacetylation complex. Moreover, immunofluorescent staining and biochemical analyses showed that coexpression of hDREF and SUMO-1 resulted in dissociation of Mi2α from chromatin, whereas a SUMOylation-defective Mi2α mutant remained tightly bound to chromatin. Chromatin immunoprecipitation and quantitative RT-PCR analysis demonstrated that Mi2α expression diminished transcription of the ribosomal protein genes, which are positively regulated by hDREF. In contrast, coexpression of hDREF and SUMO-1 suppressed the transcriptional repression by Mi2α. These data indicate that hDREF might incite transcriptional activation by SUMOylating Mi2α, resulting in the dissociation of Mi2α from the gene loci. We propose a novel mechanism for maintaining constitutively active states of a number of hDREF target genes through SUMOylation.

tion, several studies have suggested a novel function for dDREF in the establishment or regulation of transcriptional insulators found in several hundred regions of the Drosophila genome (6,7).
We previously identified hDREF as the human homolog of dDREF and determined its DNA-binding motif (hDRE; TGTCG(C/T)GA(C/T)A) (8). The hDRE sequence is similar to that of Drosophila DRE and perfectly matches the M8 motif, one of the most conserved motifs in the promoters of human genes, as determined by systematic comparative human genomics (9). In addition, hDREF was recently identified as one of the major M8-binding proteins by employing a SILAC-based quantitative proteomics approach (10). Interestingly, genes containing M8 motifs exhibited increased expression in actively proliferating cells. Accordingly, we previously demonstrated that hDREF positively regulates the expression of genes involved in cell proliferation, including histone H1 and plural ribosomal protein (RP) genes (8,11). Moreover, knockdown of hDREF resulted in impairments in cell proliferation and G 1 /S transition, further indicating that hDREF is a functional homolog of dDREF. Despite the importance of these functions (11), the mechanisms underlying the constitutively active transcription of genes involved in cell proliferation and the proteins that interact with DREF are currently unclear.
SUMOylation involves the covalent conjugation of an ϳ100amino acid (aa) small ubiquitin-related modifier (SUMO) to lysine residues in the consensus T⌿KXE (where ⌿ is any hydrophobic residue; X is any aa residue) aa sequence on target proteins (12). Protein modification by SUMO conjugation has emerged as an important modification sufficient to alter the biochemical features or activities of proteins. A number of transcription factors are regulated by SUMO modification. SUMOdependent transcriptional stimulation has been reported for GATA4, PAX6, and the glucocorticoid receptor (13)(14)(15). However, SUMO modification more frequently results in transcriptional repression, as is the case for c-Jun, C/EBP family members, Sp3, IB␣, KAP-1, PPAR␥, and a number of other transcription factors (16 -19).
SUMOylation is catalyzed by an enzymatic cascade consisting of three enzymes (12). After large SUMO precursor proteins are converted to a mature form by cleavage at the C-terminal glycine residue by SUMO protease, SUMO is attached to the heterodimeric E1 enzyme Aos1/Uba2. The activated SUMO is then transferred from the E1 enzyme to Ubc9, an E2-conjugating enzyme capable of forming a thioester intermediate between diglycine residues at the C terminus of mature SUMO proteins and the active cysteine residue of Ubc9. Ubc9 has been demonstrated to be sufficient for SUMO conjugation to sub-strate proteins in vitro. E3 ligases have been reported to act as adaptors between Ubc9 and SUMOylation substrate proteins; thus, the existence of an E3 ligase enhances the recognition of substrate proteins and promotes the SUMOylation conjugation reaction. Several E3 ligase enzymes have been identified, including protein inhibitor of activated STAT (PIAS) family members (20), the nuclear pore complex component RanBP2 (Nup358) (21), histone deacetylase 4 (HDAC4) (22), and the Polycomb group (PcG) protein Pc2 (23).
The nucleosome remodeling and deacetylation (NuRD) complex was first identified in 1998 (24 -26) and contains the histone deacetylases HDAC1/2, the ATP-dependent nucleosome remodelers Mi2␣/␤ (also known as chromodomain helicase DNA-binding proteins 3/4), RBBP4/7 (retinoblastomabinding proteins 4/7), metastasis-associated factors (MTAs), and methyl-CpG-binding proteins (MBD2/3). Because the NuRD complex contains the dual enzymatic activities of HDAC1/2 and Mi2␣/␤, this complex has been proposed to mediate transcriptional repression by regulating chromosome structure (27,28). However, recent studies have indicated that the subunit composition can vary, and thus, NuRD function may be altered by interactions between its individual complex components and other interacting molecules, including transcription factors, histone with epigenetic modifications, and methylated DNA (29 -31). Despite the accumulating knowledge on the molecules that recruit NuRD to the local chromatin region, the molecular mechanism underlying the dissociation of NuRD from chromatin remains unknown.
In this study, we found that the transcription factor hDREF exhibits SUMO E3 ligase activity and specifically conjugates SUMO-1 to Lys-1971 of Mi2␣. Moreover, we demonstrated that hDREF probably activates transcription by facilitating the dissociation of Mi2␣ from the transcribed region of hDREF target genes via SUMOylation.
The oligonucleotides for the FLAG epitope were as follows: 5Ј-tcagCATGGACTACAAGGACCACGATGA-CAAAC-3Ј and 5Ј-tcgaGCTTGTCATCGTCGTCCTTGTA-GTCCATG-3Ј. 5Ј-Overhang ends, which are compatible with staggered ends created by SalI or XhoI digestion, are indicated by lowercase letters.
The primers used in the 5Ј-RACE reaction for amplification of the 5Ј-end of Mi2␣ cDNA were as follows: 5Ј-CTTCCCGG-GCTAGCATGAAGGCGGCAGACACTGTG-3Ј and 5Ј-ATG-TGGTAGGAGGAGATGCACGCGTCACAG-3Ј. Recognition sites for NheI and MluI are underlined.
Yeast Two-hybrid Screening-Yeast two-hybrid screens with pretransformed human fetal brain Matchmaker cDNA library (Clontech) were performed using the full-length hDREF cDNA as bait as described previously (36).
DNA Transfection-Plasmid DNA was transfected into cells by the calcium phosphate method as described previously (19). In the case of 293FT cells, DNA transfections were performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions.
In Vitro Transcription/Translation-In vitro transcription and translation reactions were carried out in 50 l of reaction mixture using the TNT-coupled reticulocyte lysate system (Promega) in the presence of [ 35 S]methionine according to the manufacturer's instructions. The sizes and amounts of the products were analyzed by SDS-PAGE and autoradiography. Signals were quantified by densitometry using ImageJ software.
In Vitro SUMOylation Assay-Recombinant GST-SUMO-1(GG) lacking the two tandem glycine residues at the C terminus of mature SUMO-1, GST-Ubc9, GST-Uba2, GST-Aos1, and GST-hDREF were expressed in BL21 Escherichia coli (37) and purified using glutathione-Sepharose 4B (GE Healthcare) according to the manufacturer's instructions. Recombinant His-Mi2␣ containing the C-terminal 384 aa (residues 1617-2000) was expressed in BL21 (DE3) cells and purified using nickel-nitrilotriacetic acid-agarose (Qiagen) according to the manufacturer's instructions. In vitro SUMOylation reactions were performed as follows. In vitro synthesized full-length 35 S-Mi2␣ (10 l of in vitro transcription/translation reaction) or 0.1 g of His-Mi2␣(1617-2000) was mixed with 2 g of GST-SUMO-1(GG), 1 g of GST-Ubc9, 0.5 g of GST-Uba2, 0.5 g of GST-Aos1 in 50 mM Tris-HCl (pH 8.0), 5 mM MgCl 2 , 5 mM ATP, and 1 mM dithiothreitol at 37°C for 30 min. Reactions were terminated by heating the samples in Laemmli's sample buffer at 85°C for 3 min. The samples were resolved by SDS-PAGE, and signals were detected by autoradiography or immunoblotting analysis using anti-His antibody.
Immunoprecipitation and Immunoblotting-Harvested cells were washed with PBS and lysed in lysis buffer containing 20 mM HEPES (pH 7.4), 150 mM NaCl, 1% Triton X-100, 1 mM dithiothreitol, 0.1 mM EDTA, 20 mM N-ethylmaleimide, 1 mM Na 3 VO 4 , 5 mM NaF, 1 mM sodium pyrophosphate, and protease inhibitor mixture (Roche Applied Science) for 5 min on ice prior to sonication at 4°C and centrifugation at 15,000 ϫ g for 20 min at 4°C. The supernatant was transferred to a new tube, and appropriate antibodies were added at a final concentration of 0.5 g/ml. After incubation for 3 h at 4°C, a mixture of protein G-and protein A-Dynabeads (Invitrogen) (1:1, v/v) was added to the supernatant and rotated for 1 h at 4°C. The Dynabeads were washed with lysis buffer supplemented with 300 mM KCl and boiled in Laemmli's sample buffer for 5 min. After brief centrifugation, the supernatants were subjected to SDS-PAGE, transferred to Immobilon-P membranes (Millipore), and incubated with specific antibodies as indicated. Membranes were developed using chemiluminescence (Thermo Scientific), and signals were quantified by densitometry using ImageJ software.
Immunofluorescence and Fluorescence Microscopy-Cells grown on glass coverslips were washed with PBS, fixed with 2% paraformaldehyde (PFA) in PBS for 20 min, permeabilized in PBS containing 0.3% Triton X-100 for 15 min, and then blocked with 2% BSA in PBS. For digitonin permeabilization, cells on glass coverslips were permeabilized in PBS containing 20 g/ml digitonin for 7 min at room temperature, fixed with 2% PFA in PBS, and then blocked with 2% BSA in PBS. The coverslips were incubated with the appropriate primary and secondary antibodies in PBS containing 0.1% BSA and 0.1% Triton X-100 for 1 h at 25°C. Coverslips were mounted onto a glass slide spotted with ProLong Gold or SlowFade Gold Antifade Reagents (Invitrogen), and fluorescence was then visualized by confocal laser scanning microscopy (LSM 510, Zeiss).
ChIP Assays-ChIP assays were performed using 293FT cells transfected with expression plasmids based on methods described previously (11). Signals were detected by autoradiography and quantified by analyzing the image using ImageJ software. For quantitative analysis, we determined the number of PCR cycles exhibiting exponential amplification for each primer set using 2-fold serial dilutions of input sample.
qRT-PCR-Total RNA was prepared from HeLa cells with an RNeasy kit (Qiagen). One microgram of RNA was subjected to cDNA synthesis with oligo(dT) 18 primers and Superscript II reverse transcriptase (Invitrogen). Quantitative PCRs were carried out in a mixture containing [␣-32 P]dCTP (PerkinElmer Life Sciences) and AmpliTaq Gold (Thermo Fisher Scientific), and PCR products were separated on 6% acrylamide gels as described previously (11). Signals were detected by autoradiography and quantified by analyzing the image using ImageJ. For quantitative analysis, we determined the number of PCR cycles exhibiting exponential amplification of products for each primer set and cDNA template.

hDREF Is a SUMO E3 Ligase of Mi2␣
Isolation of Nuclei and Fractionation of Chromatin-bound and -unbound Proteins-Cells were harvested using cell scrapers, washed twice with ice-cold PBS, and lysed by Dounce homogenization in hypotonic buffer (10 mM HEPES, pH 8.0, 5 mM KCl, 1 mM MgCl 2 , 1 mM EDTA) supplemented with 1 mM N-ethylmaleimide and protease inhibitor mixture. Nuclei were pelleted by centrifugation at 16,000 ϫ g for 10 min and then suspended in buffer containing 15 mM Tris-HCl (pH 8.0), 1 mM EDTA, 0.4 M NaCl, 10% sucrose, 1 mM DTT, 1 mM N-ethylmaleimide, and protease inhibitor mixture. After incubation on ice for 30 min, chromatin-unbound (soluble) proteins were obtained as the supernatant after high speed centrifugation at 100,000 ϫ g for 1 h. The pellet was resuspended in Laemmli's buffer and subjected to sonication at 4°C prior to use as the chromatin-bound fraction. Signals were detected by autoradiography and quantified by analyzing the image using ImageJ software.
Statistical Analysis-Data are presented as means, and error bars correspond to S.E. unless otherwise indicated. Statistical significance was determined using Student's t test. p Ͻ 0.05 was considered statistically significant.

hDREF Interacts with Factors Involved in the SUMOylation Pathway and Is SUMOylated Both in Vitro and in Vivo-
To define the molecular mechanisms by which hDREF activates transcription, we screened for hDREF-interacting proteins using the yeast two-hybrid system with a human brain cDNA library and full-length hDREF as bait. Several proteins related to the SUMO pathway, such as SUMO-1, the E2 enzyme Ubc9, and the E3 enzymes PIAS1 and Pc2 were identified as putative interaction partners in addition to Mi2␣ (Table 1) (36). This raised the possibility that hDREF might be modified by SUMO or have a novel function in the SUMOylation pathway. Therefore, we first investigated whether endogenous hDREF itself was a target for SUMOylation. HeLa cell extracts were subjected to immunoprecipitation with an anti-SUMO-1 antibody followed by immunoblotting analysis. Anti-hDREF antibody detected a single slower migrating band with an apparent molecular mass of 160 kDa, suggestive of SUMO-1 modification of endogenous hDREF (Fig. 1A). Next, the preference of SUMO isoforms was investigated using HeLa cells transfected with HA-tagged hDREF and cyan fluorescent protein (CFP)tagged SUMO-1, -2, or -3 plasmids. Western blots probed with anti-HA antibody revealed a major 180-kDa and minor higher molecular mass species in addition to the 78-kDa hDREF without SUMOylation (Fig. 1B). To confirm that the slower migrating bands were SUMO-conjugated hDREF, the conjugationdefective mutant SUMO-1-G97A (38) was coexpressed with HA-hDREF, and extracts were subjected to immunoprecipitation with anti-HA antibody, followed by immunoblotting using anti-HA and anti-GFP antibodies (Fig. 1C). Both antibodies detected a major 180 kDa band and minor higher migrating bands in samples containing WT SUMO-1, the latter of which were absent in lysates from G97A-expressing cells, indicating that these slower migrating bands corresponded to SUMOylated hDREF. Moreover, the apparent molecular mass of the slow migrating band was equivalent to that of two hDREF molecules and one SUMO protein. The combined results from the in vitro SUMOylation experiments with a series of hDREF deletion mutants (Fig. 1D) and the apparent molecular mass of endogenous hDREF conjugated with SUMO-1 suggested that an hDREF homodimer may be conjugated with a single molecule of SUMO. However, SUMOylated proteins often migrate more slowly on SDS-PAGE than expected by their estimated molecular mass; thus, more studies are required to confirm this notion.
hDREF Is a SUMO E3 Ligase-From the previous analyses, we noticed that hDREF expression enhanced the signals of SUMO-conjugated proteins (Fig. 1B). Moreover, anti-GFP antibody detected numerous SUMOylated proteins, which coimmunoprecipitated with HA-hDREF (Fig. 1C). As such, we hypothesized that hDREF is a novel transcription factor possessing SUMO ligase activity. E3 ligases are known to function as adaptors between the E2-SUMO thioester and substrate protein to facilitate SUMOylation. Thus, we examined whether hDREF physically associates with SUMO and Ubc9 (E2) by GST pull-down assay. As shown in Fig. 2A, in vitro-translated hDREF bound GST-SUMO-1 and GST-Ubc9 but was unable to bind the E1 proteins GST-Aos1 and Uba2. The physical association of hDREF with both SUMO-1 and Ubc9 E2 enzyme is compatible with our hypothesis that hDREF may be a SUMO ligase. To investigate the function of hDREF in the SUMOylation further, we sought to determine the subcellular localization of hDREF, SUMO isoforms, and Ubc9. Notably, hDREF colocalized with SUMO isoforms and Ubc9 in discrete nuclear foci (Fig. 2B). The distribution and size of SUMO-hDREF foci resembled those of promyelocytic leukemia (PML) (39) or PcG nuclear bodies (40). Both specialized nuclear bodies are formed by non-covalent interactions between the SUMOylated proteins and SUMO interaction motif (SIM)-containing proteins and are considered to regulate SUMOylation in nuclei (41). Thus, to examine the relationships between hDREF-SUMO foci and PML or PcG nuclear bodies, cells expressing HA-hDREF and Myc-tagged PIAS1 or Pc2, both of which are known to be SUMO E3 ligases, were immunostained with anti-HA and anti-Myc antibodies (Fig. 2B). Most of hDREF foci overlapped with PIAS1 foci, suggesting that hDREF might exist in the PML nuclear bodies. On the other hand, hDREF and Pc2 appeared to localize to mutually exclusive areas. Because a number of studies have demonstrated a crucial role of PML nuclear bodies in regulating the SUMOylation of nuclear proteins and sequestrating SUMOylated partners through its SIM sequence (39), the observed association of hDREF with PML bodies is consistent with our hypothesis that hDREF may be a SUMO ligase.
To validate this, we performed in vivo depletion assays with shRNA against hDREF (Fig. 2C) (11). To this end, HeLa cells were transduced with lentivirus expressing hDREF shRNA and then transfected with CFP-SUMO isoforms. The relative amounts of hDREF and SUMOylated proteins were then visualized on immunoblots with anti-hDREF antibody and anti-GFP antibody, respectively. As expected, hDREF knockdown markedly reduced the SUMOylation of a number of proteins, suggesting that hDREF may play a role in promoting endogenous SUMO modification. Finally, we prepared various hDREF mutants to investigate their effect on autoSUMOylation and SUMOylation of endogenous proteins other than hDREF. As shown in Fig. 2D, hDREF with mutations in the BED finger (C47A, C50A, C47A/C50A, C61A, and C71A) yielded insufficient autoSUMOylation, whereas others (LLVL/AAAA, ⌬⌴360, ⌬L401, ⌬ATC, and W590A/W591A) could not be conjugated with SUMO-1. Interestingly, expression of these SUMOylation-defective hDREF mutants markedly weakened the SUMOylation of endogenous proteins, confirming hDREF to be a SUMO ligase.
hDREF Stimulates SUMO-1 Conjugation to Mi2␣-To confirm that hDREF possesses SUMO E3 ligase activity, we needed to show that hDREF can stimulate the SUMOylation of target proteins other than itself. We hypothesized that Mi2␣ might be hDREF Is a SUMO E3 Ligase of Mi2␣ MAY 27, 2016 • VOLUME 291 • NUMBER 22 a substrate of hDREF for the following reasons: 1) our twohybrid screen revealed an interaction between hDREF and a 400-aa region at the Mi2␣ C terminus; 2) SUMO plot analysis showed that human Mi2␣ possesses five putative consensus sequences for SUMO conjugation in its C terminus (42,43); and 3) the STRING database annotated a physical interaction between Mi2␣ and SUMO-1 and Ubc9. To test this hypothesis, HA-tagged Mi2␣ was transiently expressed with or without FLAG-hDREF and CFP-SUMO-1 in HeLa cells, and its SUMOylation status was examined by Western blotting analysis (Fig. 3A). Anti-HA antibody exhibited a strong slower migrating HA-Mi2␣ species with an apparent molecular mass of 290 kDa only when HA-Mi2␣ was co-expressed with FLAG-hDREF and CFP-SUMO-1, indicating that hDREF effectively stimulates Mi2␣ SUMOylation in vivo. Next, preference of SUMO isoforms was determined. As shown in Fig. 3B, HA-Mi2␣ polypeptide was effectively conjugated with SUMO-1 but not with SUMO-2 and SUMO-3. Finally, we determined the SUMO acceptor lysine(s) on the Mi2␣ polypeptide. The Mi2␣ C terminus contains five putative SUMOylation consensus motifs (Fig. 3C, left). Therefore, we analyzed a set of Lys 3 Arg mutants for their ability to act as SUMO acceptor sites (44). Significantly, the slower migrating band was nearly undetectable only in the K1971R lysate (Fig. 3C, lane 7), suggesting that Lys-1971 is the SUMO acceptor site of Mi2␣. It should be noted that the amount of the wild-type Mi2␣ protein often tended to be significantly increased by hDREF expression (Fig. 3C, lanes 1  and 2), raising the possibility that Mi2␣ may be stabilized by hDREF expression.
To find further evidence that hDREF acts as an endogenous SUMO ligase for Mi2␣, the effect of hDREF depletion by expressing shRNA against hDREF on HA-Mi2␣ SUMOylation hDREF Is a SUMO E3 Ligase of Mi2␣ was examined. As shown in Fig. 3D, the signal of the slower migrating SUMOylated Mi2␣ was significantly diminished by knocking down hDREF expression, and cotransfection of plasmid expressing shRNA-resistant mRNA for hDREF again accumulated SUMOylated Mi2␣. Furthermore, expression of SUMOylation-defective hDREF mutants resulted in a marked decrease of SUMOylated Mi2␣ (Fig. 3E). These results established that hDREF stimulates Mi2␣ SUMOylation in vivo.
Binding of hDREF to Mi2a Depends on the Presence of SUMO-1-To determine the trigger for Mi2␣ SUMOylation by hDREF, we first examined whether their physical interaction was modulated by SUMOylation. Interactions between hDREF and Mi2␣ were investigated by immunoprecipitation using 293FT cells transfected with either WT or K1971R HA-Mi2␣ in various combinations with FLAG-hDREF and CFP-SUMO-1. Interestingly, both WT and K1971R Mi2␣ immunoprecipitated a considerable amount of hDREF with or without SUMOylation in the presence of SUMO-1 expression (Fig. 5A, lanes 4 and  8); however, a small amount of hDREF was pulled down in the absence of SUMO-1 expression (Fig. 5A, lanes 2 and 6). Immunoprecipitation using anti-FLAG antibody similarly showed an
Mi2␣ is a component of the NuRD chromatin-remodeling complex (46,47); therefore, we tested whether Mi2␣ SUMOylation alters NuRD complex formation. To this end, we transfected FLAG-HDAC1 and HA-Mi2␣ plasmids with or without HA-hDREF and CFP-SUMO-1 plasmids into 293FT cells and examined the association of Mi2␣ with NuRD complex by immu-noprecipitation with anti-FLAG antibody. As shown in Fig. 5C, WT and K1971R Mi2␣ were similarly co-immunoprecipitated with HA-HDAC2 and other members of the NuRD complex, including RbAp48, MTA2, and MBD3, regardless of hDREF and SUMO-1 expression. These data indicate that SUMOylation of Mi2␣ does not affect NuRD complex formation.
hDREF Dissociates Mi2␣ from Chromatin in a SUMO-1-dependent Manner-To assess the role of Mi2␣ SUMOylation, we investigated the effect of hDREF/SUMO-1 coexpression on the subcellular localization of Mi2␣. FLAG-hDREF and HA-WT or HA-K1971R Mi2␣ with or without Myc-SUMO-1 were expressed in HeLa cells, and signals for hDREF and Mi2␣ were immunofluorescently detected (Fig. 6A, left). Notably, without SUMO-1 expression, hDREF and HA-WT or -K1971R Mi2␣ were distributed in the nucleoplasm, and significant portions of hDREF and Mi2␣ were colocalized on diffused foci. However, when Myc-SUMO-1 was coexpressed, hDREF signals were observed as clear foci and colocalized with both WT and K1971R Mi2␣, corroborating that the association between hDREF and Mi2␣ may depend on the presence of SUMO-1. Next, we examined hDREF and Mi2␣ signals after treatment with digitonin, a detergent that effectively releases the nuclear contents (Fig. 6A, right). Under these conditions, hDREF was detected as strong foci, and the number of foci increased by SUMO-1 expression. Staining of cells simultaneously expressing FLAG-hDREF, HA-Mi2␣, and Myc-SUMO-1 using anti-PML, anti-FLAG, and anti-Myc antibodies showed colocalization of hDREF with endogenous PML and Myc-SUMO-1 proteins in these foci, indicating that they corresponded to the PML bodies and that hDREF and Mi2␣ within the PML nuclear bodies are resistant to digitonin treatment (Fig. 6B). The distribution pattern of WT Mi2␣ without SUMO-1 expression in digitonin-treated cells was similar to that in non-treated cells. Surprisingly, digitonin treatment of SUMO-1-expressing cells led to the disappearance of the diffuse WT Mi2␣ signals from the nucleoplasm, whereas the Mi2␣ signals within the PML nuclear bodies remained with hDREF. In contrast, K1971R Mi2␣ in the nucleoplasm as well as within the PML nuclear bodies was not diminished by digitonin treatment regardless of the presence or absence of SUMO-1. These results suggested that the solubility of Mi2␣ from nucleoplasm may be increased by SUMO modification. Therefore, we speculated that SUMOylation at Lys-1971 of Mi2␣ may weaken the binding affinity of Mi2␣ to global chromatin. To confirm this possibility, we carried out biochemical experiments to determine whether the chromatin-binding activities of Mi2␣ and hDREF were altered by SUMO modification (Fig. 6, C and D). To this end, we prepared soluble and chromatin-bound protein fractions from the cell lysates of 293FT cells expressing WT or K1971R Mi2␣ and hDREF with or without SUMO-1. Both fractions were subjected to Western analysis using anti-HA and anti-FLAG antibodies. Interestingly, WT Mi2␣ was enriched in the soluble fraction (Fig. 6C, lanes 5 and 6), and the amount of chromatin-bound Mi2␣ was significantly reduced by coexpression of SUMO-1 (Fig. 6C, lane 10). Notably, SUMOylated Mi2␣ was detected only in the soluble fraction (Fig. 6C, lane 6). In contrast, a larger portion of K1971R Mi2␣ was detected in the chromatin-bound fraction regardless of SUMO-1 expression (Fig. 6C, lanes 11 and 12). In summary, the biochemical data were consistent with those of immunofluorescence analysis; thus, we concluded that Mi2␣ SUMOylation may cause Mi2␣ dissociation from global chromatin.
hDREF Stimulates Transcription of the hDREF Target Genes by Dissociating Mi2␣ from Their Gene Body via SUMOylation-Finally, we examined the effect of Mi2␣ SUMOylation on the transcriptional activity of hDREF target genes. We first investigated transcription of the RPS6 gene containing an hDRE 51 bp upstream of the transcription start site (TSS). Sole expression of hDREF, SUMO-1, and WT Mi2␣ resulted in a 110% increase, a 27% increase, and a 70% decrease in RPS6 mRNA, respectively (Fig. 7A). Interestingly, sole expression of K1971R Mi2␣ also reduced RPS6 mRNA to the same extent as WT, suggesting that SUMOylation-defective Mi2␣ may retain transcriptional repressive activity. Next, we investigated whether Mi2␣ can repress RPS6 transcription upon stimulation by hDREF expression (Fig. 7B). Without SUMO-1 expression, WT or K1971R Mi2␣ did not apparently reduce RPS6 transcription, suggesting that Mi2␣ could not substantially repress RPS6 hDREF-stimulated transcription. Additionally, coexpression of SUMO-1 with WT Mi2␣ significantly increased RPS6 gene transcription in the presence of hDREF. Surprisingly, co-expression of SUMO-1 with K1971R Mi2␣ did not enhance RPS6 transcription in the cells expressing hDREF. These results indicate that hDREF may suppress the transcriptional repressive activity of Mi2␣ by SUMOylation at Lys-1971. Supportive evidence was also obtained by qRT-PCR analysis of the hDREF target genes RPL10 and RPL12 (Fig. 7C).
We analyzed the binding status of hDREF and Mi2␣, phosphorylated RNA polymerase II (RNAPII) recruitment, and histone modifications in the proximal and distal regions of the RPS6 gene promoter by ChIP analysis (Fig. 7D). Enriched hDREF signals at the promoter showed that hDREF was evidently bound to the DRE 51 bp upstream of the TSS regardless of the simultaneous expression of Mi2␣ or SUMO-1. Interestingly, a significant amount of RNAPII accumulated at the promoter region dependent on hDREF occupancy, suggesting that hDREF stimulates recruitment of RNAPII at the promoter. Surprisingly, Mi2␣ was also extremely concentrated at the promoter region when hDREF and SUMO-1 were coexpressed (Fig. 7E, top). The accumulation of Mi2␣ was consistent with the immunoprecipitation data indicating that affinity between Mi2␣ and hDREF was enhanced by SUMO-1 expression. Interestingly, there was an overall inverse correlation between the binding status of Mi2␣ and RNAPII at a 2-kb region downstream from the TSS (corresponding to the fourth intron) (Fig.  7E, bottom). Mi2␣ expressed solely or together with hDREF or SUMO-1 occupied the fourth intron, but Mi2␣ signal disappeared when expressed simultaneously with hDREF and SUMO-1. RNAPII signal in the fourth intron, presumably reflecting active transcriptional elongation, was decreased when Mi2␣ occupied the region but was markedly increased when Mi2␣ was dissociated from the region. Furthermore, the pattern of histone H3 K4 trimethylation, a marker of active transcription, was mostly similar to that of RNAPII. Moreover, trimethylated histone H3 Lys-9, a marker of inactive transcription, exhibited trimethylation patterns inverse to those of H3 Lys-4. Together, the binding status of Mi2␣ and RNAPII seems to be mutually exclusive; thus, the absence of Mi2␣ and the presence of RNAPII in the gene body may be required for active RPS6 transcription. Finally, we performed ChIP with cells expressing K1971R Mi2␣ (Fig. 7F). Expectedly, SUMOylationdefective Mi2␣ bound in the fourth intron, even with hDREF and SUMO-1 coexpression. Under this condition, RNAPII was hardly detected in that location. Collectively, these results strongly suggested that hDREF may stimulate RPS6 expression by Mi2␣ SUMOylation that causes the release Mi2␣ from the gene body of RPS6. At 48 h after transfection, total RNA was extracted, and qRT-PCR was performed for RPS6 and GAPDH. The intensity of individual bands was quantified and expressed relative to the control (no transfection). Amounts of RPS6 mRNA were normalized to those of GAPDH as an internal control. Data are the mean of three independent experiments. Error bars, S.E. **, p Ͻ 0.02; n ϭ 3. B, HeLa cells were transfected with FLAG-hDREF, HA-Mi2␣ (WT or K1971R), or CFP-SUMO-1 in the indicated combinations. At 48 h after transfection, total RNA was extracted, and qRT-PCR was performed for RPS6 and GAPDH. Amounts of RPS6 mRNA were normalized by those of GAPDH as an internal control. Data are the average of three independent experiments. Error bars, S.E. *, p Ͻ 0.05; **, p Ͻ 0.02; n ϭ 3. C, HeLa cells were transfected with FLAG-hDREF, HA-Mi2␣ (WT or K1971R), and CFP-SUMO-1 as indicated. At 48 h after transfection, total RNA was extracted, and qRT-PCR was performed for RPS6, RPL10, RPL12, and GAPDH. Amounts of mRNA of RPS6, RPL10, and RPL12 were normalized by those of GAPDH as an internal control. Data are expressed as the average value relative to a no transfection control. n ϭ 3. D, schematic illustration of the RPS6 gene. Arrowhead, a hDREF binding site. Solid bars, the region amplified by PCR in ChIP analysis. E, 293FT cells were transfected with FLAG-hDREF, HA-Mi2␣, and CFP-SUMO-1 as indicated. At 36 h after transfection, cells were cross-linked with 1.1% formaldehyde for 5 min, lysed, and sonicated. After centrifugation, the cleared supernatant containing chromatin was subjected to immunoprecipitation using antibodies (control IgG, anti-FLAG, anti-HA, anti-RNA polymerase II (phospho-CTD), anti-histone H3 Lys-4-trimethyl, and anti-histone H3 Lys-9-trimethyl). PCR was quantitatively performed in the reaction mixture containing 32 P-dCTP using genomic DNA from the input extract and DNA recovered from the immunoprecipitates as indicated. PCR products were separated by electrophoresis on 6% acrylamide gels and detected by autoradiography. The intensity of individual bands was quantified and expressed as a percentage of input. Data are the average of three independent experiments. Error bars, S.E. (*, p Ͻ 0.05; **, p Ͻ 0.02). F, ChIP assays were performed using antibodies as described. Instead of WT Mi2␣, SUMOylation-defective K1971R mutant was expressed in 293FT cells. Data are the average of three independent experiments. Error bars, S.E. (*, p Ͻ 0.05; **, p Ͻ 0.02).

Discussion
To our knowledge, this study is the first to demonstrate that the sequence-specific transcription factor hDREF possesses SUMO ligase activity and stimulates transcription of target genes via direct SUMO ligation to Mi2␣, an ATP-dependent nucleosome-remodeling enzyme in the NuRD complex. We propose a novel mechanism by which the sequence-specific transcription factor cancels Mi2␣-mediated transcriptional repression. Several studies have revealed that the NuRD complex is recruited to chromatin through protein-DNA interactions between MBD2 and methylated DNA and protein-protein interactions between Mi2␣/␤ and histone H3 (27,28) and between RBBP7/4 and histone H4. In addition, considerable evidence indicates that the NuRD can also be recruited to focal chromatin regions by specific interactions with transcription factors or corepressors (48 -50). However, it remains unclear how the repressive NuRD complex dissociates from chromatin when necessary. Thus, our findings provide new insights into SUMOylation-dependent dissociation of the NuRD complex from chromatin.
Immunofluorescence and biochemical analyses clearly showed that hDREF and SUMO-1 coexpression resulted in an increased amount of unbound Mi2␣, whereas the SUMOylation-defective K1971R mutant tightly bound to chromatin (Fig.  6, A and C). This result indicated that Mi2␣ SUMOylation by hDREF is sufficient to dissociate Mi2␣ from chromatin, directly inhibit Mi2␣ binding to chromatin, or prevent the incorporation of Mi2␣ into the NuRD complex. However, immunoprecipitation revealed that Mi2␣ could associate with the NuRD complex with or without SUMO-1 to a similar extent, suggesting that Mi2␣ SUMOylation does not affect NuRD complex formation. Thus, it is more likely that the chromatin binding affinity of Mi2␣ in the NuRD complex (Mi2␣-NuRD) may be weakened as a result of Mi2␣ SUMOylation.
Using ChIP analysis, we examined the binding of Mi2␣ to the hDREF target gene RPS6, and we observed that Mi2␣ distribution patterns at the promoter region and the gene body differed depending on the combination of hDREF and SUMO-1 expression. When Mi2␣ was expressed alone or with hDREF or SUMO-1, Mi2␣ signal at the promoter region was relatively weak, whereas a more intense signal was detected in the gene body. Inversely, Mi2␣ was enriched at the promoter and completely absent in the gene body upon coexpression of hDREF and SUMO-1. In contrast, RNAPII and histone modification signals for active gene transcription disappeared with Mi2␣ expression and were enhanced by the simultaneous expression of hDREF and SUMO-1. Thus, it is evident that signals for active transcription emerged in gene body concurrent with the absence of Mi2␣. Thus, the absence of Mi2␣ (or Mi2␣-NuRD) in the gene body, rather than at the promoter, may be important to incite transcription elongation, and RNAPII stalled at the promoter-proximal region may be released, dependent upon the simultaneous expression of hDREF and SUMO-1. This led us to question how Mi2␣ is dissociated from the gene body in the presence of hDREF and SUMO-1. We developed a model based upon the following lines of evidence: 1) the interaction between hDREF and Mi2␣ was enhanced in the presence of SUMO-1 (Fig. 5, A and B); 2) Mi2␣ exhibited a similar affinity to both SUMOylated hDREF and non-SUMOylated hDREF (Fig.  5A); 3) SUMOylation-defective K1971R Mi2␣ was coimmunoprecipitated with hDREF much more effectively than WT Mi2␣ (Fig. 5B); 4) even in the presence of hDREF and SUMO-1, K1971R Mi2␣ existed in the gene body, whereas WT Mi2␣ disappeared in such situation (Fig. 7, E and F). Thus, we presumed that Mi2␣ may be locally recruited to the promoter region through a direct protein-protein interaction with SUMOylated hDREF, conjugated with SUMO-1 by hDREF, and subsequently dissociated from chromatin, thereby limiting Mi2␣ occupancy at the distal promoter region. To confirm this model, further studies on binding and dissociation kinetics between hDREF and Mi2␣ with or without SUMO modification using an in vitro reconstitution system should be performed. Moreover, it would be of great interest to elucidate the role of PML nuclear bodies, to which a large part of hDREF and Mi2␣ colocalize dependent on SUMO-1 expression. PML nuclear bodies are proposed to regulate a wide variety of processes through sequestration of PML-interacting proteins via its SUMO-SIM interaction; therefore, we speculate that hDREF may tether unbound Mi2␣ to PML nuclear bodies through its SUMOylation in order to maintain active transcription. To verify this possibility, future studies are necessary to clarify molecular interactions among hDREF, Mi2␣, and PML.
We cannot exclude the possibility that hDREF stimulates transcription by tethering Mi2␣-NuRD to the promoter region and switching Mi2␣-NuRD function from transcription repression to activation through Mi2␣. Several reports support this possibility. For example, the Mi2␣-NuRD complex enhances transcription by acting as a c-Myb coactivator (51). More recently, the IKAROS transcription factor has been shown to stimulate transcription elongation in a hematopoietic cell lineage by concurrently recruiting NuRD complex (49) and the positive transcription elongation factor ␤ (P-TEF␤) (49). Moreover, Shimbo et al. demonstrated that the MBD3-containing NuRD complex occupies CpG-rich promoters marked by histone H3 Lys-4 trimethylation; thus, the MBD3-NuRD complex may play multiple roles in fine-tuning the expression of both active and silent genes (31). These findings suggest that the NuRD complex acts as transcription activator depending on its interacting partners. Further studies to analyze the characteristics of SUMOylated Mi2␣-NuRD are required to fully understand its molecular functions. For this purpose, reconstitution with a SUMOylation mimic mutant of Mi2␣ may be of use.
Liu et al. (52) showed that SUMO-1 occupies the proximal promoter regions of numerous human housekeeping genes. They also demonstrated that ribosomal protein genes, which are most actively and constitutively expressed, are remarkably enriched with SUMO-1 at the chromatin surrounding the TSSs and that depletion of SUMO-1 resulted in down-regulation of these genes. Surprisingly, six ribosomal protein genes in their list (RPL5, RPL7A, RPL10A, RPL17, RPL26, and RPS16) commonly contain hDRE at the Ϫ41 to Ϫ67 positions from their TSSs (11). Thus, it is possible that SUMO-1 enrichment at the proximal promoter regions of these genes may reflect that the SUMOylated hDREF bound to the hDREs exists in the proximal promoter regions and that hDREF may be important for con-stitutively active transcription of a set of housekeeping genes. Indeed, most dDREF-binding motifs localize proximal to the TSSs of Drosophila housekeeping genes, and dDRE motifs are required and sufficient for enhancer function of housekeeping genes (4). Further research on the distribution of Mi2␣, hDREF, SUMO-1, and RNAPII over a wide range of chromatin using ChIP-sequencing and RNA-sequencing analyses will be helpful to reveal whether or not SUMO ligase activity of hDREF is generally involved in active transcription of hDRE-containing housekeeping genes.
Additionally, we speculate that the position of hDRE from the TSS might be critical for hDREF function. The hDRE-binding position of candidate hDREF target genes exhibited a narrow peak at 62 bp upstream of the TSS (9). Considering that RNAPII was definitely recruited to the promoter and that histone modifications reflecting active transcription were concentrated at the promoter-proximal region when hDREF was bound to the hDRE (Fig. 7E) with or without the presence of SUMO-1 and Mi2␣, hDREF may solely promote epigenetic changes at the promoter region to facilitate transcription initiation. Our preliminary result obtained by immunoprecipitation using anti-hDREF antibody showed specific interaction between hDREF and CBP or TFIIF. CBP is a well characterized transcriptional coactivator with histone acetyltransferase activity (53). TFIIF is associated with RNAPII and assists RNA-PII preinitiation complex formation as well as direct elongation initiation complex formation (54,55). Future detailed analysis of the interaction between hDREF and histone acetyltransferase or general transcription factors is needed.
In summary, we provided evidence that hDREF acts as a SUMO ligase and may cancel transcription repression by the Mi2␣-NuRD complex through specific SUMO conjugation on Mi2␣. We speculate that hDREF is kept transcriptionally active in a set of housekeeping genes carrying the hDRE. Further research will be necessary to identify other substrates of hDREF to clarify the possible molecular mechanism for keeping constitutively active states of a number of hDREF target genes.