The THAP-Zinc Finger Protein THAP1 Associates with Coactivator HCF-1 and O-GlcNAc Transferase

THAP1 is a sequence-specific DNA binding factor that regulates cell proliferation through modulation of target genes such as the cell cycle-specific gene RRM1. Mutations in the THAP1 DNA binding domain, an atypical zinc finger (THAP-zf), have recently been found to cause DYT6 dystonia, a neurological disease characterized by twisting movements and abnormal postures. In this study, we report that THAP1 shares sequence characteristics, in vivo expression patterns and protein partners with THAP3, another THAP-zf protein. Proteomic analyses identified HCF-1, a potent transcriptional coactivator and cell cycle regulator, and O-GlcNAc transferase (OGT), the enzyme that catalyzes the addition of O-GlcNAc, as major cellular partners of THAP3. THAP3 interacts with HCF-1 through a consensus HCF-1-binding motif (HBM), a motif that is also present in THAP1. Accordingly, THAP1 was found to bind HCF-1 in vitro and to associate with HCF-1 and OGT in vivo. THAP1 and THAP3 belong to a large family of HCF-1 binding factors since seven other members of the human THAP-zf protein family were identified, which harbor evolutionary conserved HBMs and bind to HCF-1. Chromatin immunoprecipitation (ChIP) assays and RNA interference experiments showed that endogenous THAP1 mediates the recruitment of HCF-1 to the RRM1 promoter during endothelial cell proliferation and that HCF-1 is essential for transcriptional activation of RRM1. Together, our findings suggest HCF-1 is an important cofactor for THAP1. Interestingly, our results also provide an unexpected link between DYT6 and DYT3 (X-linked dystonia-parkinsonism) dystonias because the gene encoding the THAP1/DYT6 protein partner OGT maps within the DYT3 critical region on Xq13.1.

THAP1 is a sequence-specific DNA binding factor that regulates cell proliferation through modulation of target genes such as the cell cycle-specific gene RRM1. Mutations in the THAP1 DNA binding domain, an atypical zinc finger (THAP-zf), have recently been found to cause DYT6 dystonia, a neurological disease characterized by twisting movements and abnormal postures. In this study, we report that THAP1 shares sequence characteristics, in vivo expression patterns and protein partners with THAP3, another THAP-zf protein. Proteomic analyses identified HCF-1, a potent transcriptional coactivator and cell cycle regulator, and O-GlcNAc transferase (OGT), the enzyme that catalyzes the addition of O-GlcNAc, as major cellular partners of THAP3. THAP3 interacts with HCF-1 through a consensus HCF-1-binding motif (HBM), a motif that is also present in THAP1. Accordingly, THAP1 was found to bind HCF-1 in vitro and to associate with HCF-1 and OGT in vivo. THAP1 and THAP3 belong to a large family of HCF-1 binding factors since seven other members of the human THAP-zf protein family were identified, which harbor evolutionary conserved HBMs and bind to HCF-1. Chromatin immunoprecipitation (ChIP) assays and RNA interference experiments showed that endogenous THAP1 mediates the recruitment of HCF-1 to the RRM1 promoter during endothelial cell proliferation and that HCF-1 is essential for transcriptional activation of RRM1. Together, our findings suggest HCF-1 is an important cofactor for THAP1. Interestingly, our results also provide an unexpected link between DYT6 and DYT3 (X-linked dystonia-parkinsonism) dystonias because the gene encoding the THAP1/DYT6 protein partner OGT maps within the DYT3 critical region on Xq13.1.
Dystonias are neurological diseases characterized by involuntary muscle contractions which result in twisting, repetitive movements, and abnormal postures (1,2). At least six primary forms, where dystonia is the only neurologic feature, have been described (1,2), but the disease gene has been discovered for only two of these, DYT1 and DYT6 (3,4). Early onset DYT1 generalized dystonia is caused by mutations in the gene encoding TorsinA (TOR1A), a member of the AAAϩ family of ATPases, which may function as a chaperone in the nuclear envelope and endoplasmic reticulum (3). Recently, mutations in THAP1 have been identified as a cause of mixed-onset DYT6 primary torsion dystonia (4 -6). THAP1 is the prototype of a previously uncharacterized family of cellular factors (Ͼ 100 distinct members in the animal kingdom), defined by the presence at their amino terminus of the THAPzinc finger (THAP-zf), 3 an atypical zinc-dependent sequencespecific DNA binding domain (7)(8)(9). THAP1 recognizes a consensus DNA target sequence of 11 nucleotides (THABS for THAP1 binding sequence) considerably larger than the 3-4 nucleotide motif typically recognized by classical C2H2 zinc fingers (8). Although THAP1 biological roles are not completely understood, data supporting an important function in cell proliferation and cell cycle pathways have been provided (10). THAP1 was found to be involved in the regulation of endothelial cell proliferation and G1/S cell cycle progression, and RRM1, a pRb/E2F cell cycle target gene involved in S-phase DNA synthesis, was identified as the first direct target gene of THAP1 (10). In addition to cell proliferation, THAP1 may also play roles in cell survival and/or apoptosis because it has been shown to interact with Par-4, a proapoptotic factor linked to prostate cancer and neurodegenerative diseases, including Parkinson disease (11).
Several distinct mutations in THAP1 were found in DYT6 dystonia patients and most of them mapped to the DNA binding THAP-zf, suggesting the mutations may cause disease by disrupting the DNA binding activity of THAP1 (4,5). Indeed, one of the THAP1 missense mutant proteins (F81L) was functionally analyzed and found to exhibit strongly reduced DNA binding affinity (4). Together, these findings supported the possibility that transcriptional dysregulation, because of mutations in THAP1, might contribute to the phenotype of DYT6 dystonia. Interestingly, transcriptional dysregulation because of reduced expression of TAF-1 (RNA polymerase II TATA-boxbinding protein-associated factor 1) has previously been proposed to cause another form of dystonia, DYT3 or X-linked dystonia-parkinsonism (12,13).
In this study, we report that THAP1 shares sequence characteristics, in vivo expression profiles and cellular partners with THAP3, a previously uncharacterized member of the THAP-zf protein family. Using functional proteomics, we show that HCF-1, a potent transcriptional coactivator and cell cycle regulator (14 -19), and OGT, the enzyme that catalyzes the addition of O-GlcNAc (20), are major cellular partners of THAP3. THAP1 also associates with HCF-1 and OGT in vivo and mediates recruitment of HCF-1 to the RRM1 promoter during endothelial cell proliferation. Our results provide an unexpected link between THAP1/DYT6 and OGT, the product of a gene (OGT) mapped within the DYT3 critical interval on Xq13.1 (12,21), suggesting an unexpected link between DYT6 and DYT3 dystonias.

EXPERIMENTAL PROCEDURES
Cell Culture and RNA Interference-Hela cells, Hela Tet-Off cells (Clontech), and Hela HLR cells (Stratagene) were grown in Dulbecco's modified Eagle's medium. Knockdown of THAP1 and HCF-1 expression in primary human endothelial cells (HUVECs) was performed using ON-TARGET plus SMARTpool and individual siRNA duplexes (Dharmacon, Lafayette, CO) as previously described (10).
Plasmid Constructions-The fulllength coding region of human THAP3 was cloned in-frame with an 82-bp linker encoding Flag/HA tags in pTRE-Tight expression vector (Clontech). The resulting vector was co-transfected with pRep-Hygro vector into Hela Tet-Off cells and stable transformants were obtained after 4 weeks of selection in 200 g/ml hygromycin. Human THAP1 ORF was cloned into expression vector pFlag-CMV5a (Sigma) to generate pFlag-THAP1 expression vector. GST, GST-THAP1 1-213 , GST-THAP3 162-239 fusion proteins were produced using pGEX-2T prokaryotic expression vector (Amersham Biosciences), and purified as previously described (11). GST-THAP3 H178A , -THAP3 Y180A , and -THAP3 HBM (DHSY/AAAA) mutants were obtained from wild-type GST-THAP3 162-239 using PCR. Gal4-THAP3 and Gal4-THAP3 HBM mutant expression vectors were generated by inserting the corresponding full-length THAP3 fragments, generated by PCR, into pCMVGT vector downstream of the Gal4-DB (amino acids 1-147). The THAP3 HBM mutant was also cloned into pTRE-Tight expression vector. For two-hybrid assays, the open-reading frames of human THAP-zf proteins were amplified by PCR from full-length clones (Open Biosystems) and inserted into the pGADT7 expression vector (Clontech).
Sequence Analysis-Multiple sequence alignments of the THAP-zf proteins and their HBMs were generated with ClustalW according to the Blosum matrix and colored with Boxshade. Prediction of coiled-coil domains was performed with COILS and PAIRCOIL programs.
Immunoaffinity Purification and Mass Spectrometry Analysis-Nuclear extracts from induced Hela-t-THAP3 cells were prepared as previously described (22) with minor modifica- tions. Nuclei were incubated for 30 min in buffer B: 20 mM Tris pH 7.4; 400 mM NaCl, 5 mM MgCl 2 , 10 mM ␤-mercaptoethanol, 0,5% Nonidet, 1 mM phenylmethylsulfonyl fluoride, and Complete protease inhibitor mixture (Roche). The salt concentration was adjusted to 150 mM NaCl, and nuclear extracts were loaded onto a 4-ml 10 -40% glycerol gradient in 150 mM NaCl buffer B, centrifuged at 50,000 rpm (200,000 ϫ g) for 4 h. Fractions corresponding to a peak determined after immunoblotting were pooled and purified by immunoprecipitation with anti-FlagM2 antibody-conjugated agarose beads (Sigma), which were washed with 150 mM NaCl-containing IP buffer and eluted with a solution at 500 g/ml of Flag peptide. As a control, a mock purification was performed from uninduced HeLa cells. Samples from induced and non-induced HeLa cells were separated by SDS-PAGE, and submitted for proteomic analysis (see supplemental information).
Reporter Gene Assays-Hela HLR cells, which contain a Firefly luciferase chromatin-integrated reporter gene under the control of five GAL4 binding sites, were co-transfected with 50 ng of pCMVGT-Gal4-THAP3 or -Gal4-THAP3 HBM constructs and 50 ng of Renilla luciferase construct (pRL-CMV, Promega), with or without 1 g of HCF-1-V5 expression vector (15) using JetPEI reagent (2 l/well, Polyplus Transfection). After 48h, Firefly and Renilla luciferase activities were assayed with the Dual luciferase assay system (Promega). Firefly luciferase activities were normalized to Renilla luciferase to control for transfection efficiency. Chromatin Immunoprecipitations (ChIPs), qPCR Assays, and Northern Blot-ChIP-qPCR assays were performed as previously described (10) using 5 g of antibodies specific for THAP1 (10) or HCF-1 (15). Sequences of the primers are available in the supplemental information. QuantiTect Primer Assays (Qiagen) were used for analysis of HCF-1, RRM1, and GAPDH expression by qPCR. For Northern blot analysis, blots of poly(A) ϩ RNA from adult mouse and human tissues were hybridized according to the manufacturer's instructions (Clontech), using random-primed human THAP3 and THAP1 cDNA probes.

RESULTS
THAP3 Shares Similarities with THAP1-Human THAPzf proteins are not known to share homologies outside the THAP-zf. However, we observed that, among the 12 human THAPzf proteins (7), THAP3, shares sequence similarities with THAP1 not only within the THAP-zf (7) but also within the carboxyl-terminal domain, which contains in both proteins, a conserved bipartite nuclear localization sequence (NLS), and a coiled-coil domain predicted by COILS and PAIRCOIL programs (Fig. 1A). Northern blot analyses revealed that THAP3 and THAP1 also share striking similarities in their expression profile both in mouse and human tissues (Fig. 1B). Together, these findings indicated THAP3 and THAP1 exhibit similarities not only in their primary structure and sequence but also in their expression pattern in vivo.
THAP1 and THAP3 Associate with HCF-1 and OGT-To identify THAP3-interacting partners in vivo, nuclear extracts were prepared from Hela-t-THAP3 cells, conditionally expressing a carboxyl-terminal Flag/HA-tagged THAP3, and fractionated in a glycerol gradient. A predominant THAP3-containing complex of ϳ0.6 MDa (supplemental Fig. S1) was identified, immunoprecipitated from the corresponding fractions with anti-Flag antibody, and analyzed by nanoLC-MS/MS mass spectrometry (Fig. 2A). The purification was reproducible, as several THAP3-interacting proteins were repeatedly identified in three independent affinity purification/ proteomic analyses (Table 1). Strikingly, many of the THAP3interacting proteins corresponded to differentially processed forms of HCF-1. Numerous peptides were identified within both N-and C-terminal parts of HCF-1, including many peptides bearing O-glycosylated residues within the HCF-1 basic domain (supplemental Fig. S2). Interestingly, the human OGT enzyme (20) and protein chaperones with demonstrated O-GlcNAc lectin activity, such as the HSP70-like FIGURE 3. THAP1 and THAP3 belong to a large family of HCF-1-binding THAP-zf proteins. A, multiple alignment of the HBMs found in human THAP-zf proteins. A consensus is shown. B, primary structure of the nine human THAP-zf proteins containing consensus HBMs or HBM-like motifs. The asterisk indicates the position of the HBM, which is always located 4 -9 amino acids upstream the predicted coiled-coil domain. C, twohybrid assay. A yeast strain expressing the GAL4DB-HCF-1 kelch domain was transformed with constructs expressing the indicated human THAP-zf proteins or control protein CREB3. Transformants were patched onto SD-Leu/Trp (LT) and SD-Leu/Trp/His (LTH) plates. D, ␤-galactosidase assays. The ␤-galactosidase activity of each strain is expressed as relative light units per second (RLU/s). The result shown for each strain is the average of two independent colonies after subtraction of the background level obtained from the GAL4DB-HCF-1 kelch strain transformed with pGADT7 vector alone. proteins Hsc71 and GRP-78 (23), were also reproducibly identified in the THAP3-associated protein complexes ( Fig. 2A and Table 1). The association of HCF-1 and OGT with THAP3 was further confirmed by co-immunoprecipitation assays and immunoblot analyses of nuclear extracts fractionated in glycerol gradients (supplemental Fig. S1). Together, these proteomic analyses indicated that O-GlcNAcylated-HCF-1, O-GlcNAc transferase OGT, and O-GlcNAc lectins constitute the core components of the THAP3-associated protein complexes in human Hela cells. Numerous viral and cellular factors bind the HCF-1 aminoterminal kelch domain, via a tetrapeptide motif ((D/E)HXY), designated the HCF-1-binding motif (HBM) (19,24,25). A consensus HBM (DHSY) is also present in THAP3 (Fig. 1A), and GST pulldown assays revealed that wild-type THAP3 binds to HCF-1 in vitro. Mutations of the THAP3-HBM strongly reduced (single amino acid mutants) or completely abrogated (quadruple mutant) the THAP3-HCF-1 interaction in vitro (Fig. 2B). Moreover, co-immunoprecipitation assays in human cells revealed that THAP3, but not the THAP3 HBM mutant, is able to precipitate HCF-1 and OGT (Fig. 2C). The association of THAP3 HBM with OGT was also significantly reduced, suggesting THAP3 associates with OGT in vivo mainly through its interaction with HCF-1. We also observed that THAP1 contains a consensus HBM (DHNY) (Fig. 1A) and interacts with HCF-1 in two-hybrid (see below) and GST pulldown assays in vitro (Fig. 2D). In addition, THAP1 associates with HCF-1 and OGT in co-immunoprecipitation assays in vivo (Fig. 2E). These later findings indicated THAP3 shares its two major cellular partners with THAP1.
In agreement with the presence of the HBM, several members of the human THAP-zf protein family were identified in large scale two-hybrid screens with the HCF-1 amino-terminal kelch domain (supplemental Fig. S4). Additional two-hybrid assays (Fig. 3, C and D) revealed that the HCF-1 kelch domain binds to human THAP-zf proteins that contain an HBM (THAP1 to THAP7 and THAP11) or HBM-like motif (THAP4) but not to those which do not possess HBMs (THAP8 and THAP10). Based on our bioinformatics analyses and two-hybrid results, we concluded that THAP-zf proteins represent a large family of HCF-1 binding factors.
THAP3 Recruits HCF-1 to a Chromatin-integrated Promoter for Transcriptional Activation-We asked if interaction of THAP3 with HCF-1 might play a role in transcriptional regulation. Because THAP3 target genes have not yet been identified, we used a chromatin-integrated Gal4-dependent luciferase reporter construct (Hela-HLR cell line). Wild-type THAP3 or THAP3 HBM mutant was expressed as Gal4 DNA binding domain-fusion proteins in Hela HLR cells, with or without V5-epitope-tagged HCF-1 (Fig. 4A). ChIP assays revealed that both Gal4-THAP3 and Gal4-THAP3 HBM were significantly enriched at the Gal4 binding sites in the Gal4 reporter gene (Fig. 4B). In contrast, ChIP assays with anti-V5 antibodies showed that the V5-epitope-tagged HCF-1 was recruited to the Gal4 binding sites when co-expressed with Gal4-THAP3 but not when co-expressed with Gal4-THAP3 HBM (Fig. 4C). Moreover, recruitment of HCF-1 by Gal4-THAP3 resulted in enhanced transcriptional activity of the Gal4-luciferase reporter gene (Fig. 4D), whereas no effect was found with the Gal4-THAP3 HBM mutant. We conclude that THAP3, through its HBM, is able to recruit HCF-1 to a chromatin-integrated promoter for transcriptional activation.
Endogenous THAP1 Recruits HCF-1 to the RRM1 Promoter-We next asked if HCF-1 might play a role in transcriptional regulation by endogenous THAP1. Using primary human endothelial cells (HUVECs), we have recently shown that THAP1 associates in vivo with consensus THAP1 binding sites found in the RRM1 promoter (10). Because HCF-1 has previously been found to be required for multiple stages of cell cycle progression (18), we determined whether binding of HCF-1 to Results are shown as means and SDs of three separate data points and are representative of at least two independent experiments. D, recruitment of HCF-1 by Gal4-THAP3 results in enhanced transcriptional activity of the Gal4luciferase reporter, whereas no effect was found with the Gal4-THAP3 HBM mutant. Normalized luciferase activities are shown as means and SDs of three independent transfection experiments.

THAP1/DYT6 Associates with HCF-1 and OGT
THAP1 may play a role in transcriptional regulation of RRM1. The RRM1 promoter (Fig. 5A) contains binding sites for both THAP1 and E2F transcription factors. ChIP assays were performed in HUVECs after knockdown of endogenous THAP1 expression with a pool of specific siRNAs (Fig. 5B). ChIP experiments performed in the presence of control siRNAs revealed that THAP1 and HCF-1 are recruited to the promoter region containing the THAP1 binding sites (Ϫ178) but not to the 5Ј-untranslated region containing the E2F site (ϩ84) nor to a site close to the STOP codon (ϩ2630) (Fig.  5C). In cells treated with THAP1 siRNAs, there was a significant reduction of THAP1 association. Interestingly, the reduction in THAP1 binding correlated with a parallel reduction in HCF-1 association with the RRM1 promoter ( Fig. 5C and supplemental Fig. S5). In contrast, the association of HCF-1 with the p107 promoter, a promoter bound by E2Fs (19) but not by THAP1, was increased after knockdown of THAP1 (Fig. 5D). These experiments performed in primary human cells indicated that THAP1, rather than E2Fs, is important for HCF-1 recruitment to the RRM1 promoter during endothelial cell proliferation.
Endogenous HCF-1 Regulates RRM1 mRNA Levels-We have previously shown that silencing of endogenous THAP1 in primary HUVECs led to a significant decrease in RRM1 mRNA levels (10). Therefore, we asked whether HCF-1, a potent transcriptional coactivator (24), may also play a role in the regula-tion of RRM1 expression. Knockdown of endogenous HCF-1 expression with a pool of siRNAs optimized to reduce off-target effects (ON-TARGET-plus SMARTpool siRNAs) resulted in down-regulation of RRM1 mRNA levels (Fig.  6A), indicating HCF-1 is essential for the activation of RRM1 during cell cycle progression. This result was confirmed using 4 individual siRNAs, which similarly reduced HCF-1 levels in HUVECs, as shown by Western blot (Fig. 6B) and qPCR (Fig. 6C) analyses. Importantly, the 4 HCF-1 siRNAs reduced RRM1 mRNA levels (ϳ50% inhibition), but had no significant effects on mRNA levels of actin, GAPDH, and THAP1. Together, these data provide strong evidence that endogenous HCF-1 modulates RRM1 mRNA levels in primary human endothelial cells.
Finally, we performed knockdown experiments to address the potential role of OGT in RRM1 regulation. As shown in supplemental Fig. S6, reduction of OGT levels did not affect RRM1 expression in HUVECs. However, this does not preclude a role for OGT in RRM1 regulation in other cell types and for the THAP1/HCF-1/OGT complex in the regulation of other promoters.

DISCUSSION
In this study, we report that dystonia protein THAP1/DYT6 exhibits striking similarities with THAP3, another member of the THAP-zf protein family, in terms of primary structure, in vivo expression pattern and cellular partners. Similarly to THAP3, THAP1 associates with HCF-1, a cell cycle factor and potent transcriptional coactivator, and OGT, the enzyme that mediates O-GlcNAcylation of proteins. THAP3 binding to HCF-1 is mediated by a consensus HBM, a motif that is also found in THAP1. In addition to THAP1 and THAP3, seven other members of the human THAP-zf protein family, contain evolutionary conserved HBMs and possess HCF-1 binding properties, indicating THAP-zf proteins represent a large family of HCF-1-binding factors.
THAP1/HCF-1 Interaction and Transcriptional Regulation during the Cell Cycle-We found that THAP3 and THAP1 are both able to recruit HCF-1 to chromatin-integrated promoters for transcriptional activation. HCF-1 was recruited to the THAP1 binding sites, rather than the E2F site, in the promoter of cell cycle-specific gene RRM1, and endogenous THAP1 was found to be essential for HCF-1 recruitment during endothelial cell proliferation. This is an important result because, although recruitment of HCF-1 on cell cycle-specific promoters has been FIGURE 5. Endogenous THAP1 recruits endogenous HCF-1 to the promoter of cell cycle-specific gene RRM1 during cell proliferation in vivo. A, schematic representation of the human RRM1 promoter. The binding sites for THAP1 (THABS) and E2Fs are indicated. The position of the DNA fragments analyzed in ChIP-qPCR assays are shown. B, knockdown of endogenous THAP1 in primary HUVECs. THAP1, HCF-1, and Tub␣ (loading control) expression levels were analyzed by Western blot. Down-regulation of THAP1 does not modify HCF-1 levels. C, knockdown of THAP1 inhibits recruitment of THAP1 and HCF1 to the RRM1 promoter in vivo. ChIP-qPCR assays with anti-THAP1 or anti-HCF1 antibodies were performed using proliferating primary HUVECs treated with control or THAP1 siRNAs. Immunoprecipitated DNA was quantified in triplicate by qPCR. Fold enrichment was calculated by dividing the amount of the RRM1 promoter Ϫ178 and ϩ84 DNA fragments precipitated to the amount of RRM1 ϩ 2530 DNA fragment precipitated (negative control). Results are representative of at least two independent experiments. D, knockdown of THAP1 results in increased recruitment of HCF1 to the p107 promoter in vivo. ChIP-qPCR assays with anti-THAP1 or anti-HCF1 antibodies were performed as described in C.