O-GlcNAc Transferase Is Critical for Transducin-like Enhancer of Split (TLE)-mediated Repression of Canonical Wnt Signaling*

Background: The molecular mechanisms by which Groucho/TLEs repress canonical Wnt signaling are incompletely understood. Results: OGT interacts with TLEs and facilitates TLE-mediated transcriptional repression. Removal of O-GlcNAc is essential for gene activation from Wnt-responsive promoters. Conclusion: OGT plays a vital role in TLE-mediated repression of Wnt signaling. Significance: O-GlcNAc modification has a profound influence on diverse signaling pathways. The Drosophila Groucho protein and its mammalian orthologues the transducin-like enhancers of split (TLEs) are critical transcriptional corepressors that repress Wnt and other signaling pathways. Although it is known that Groucho/TLEs are recruited to target genes by pathway-specific transcription factors, molecular events after the corepressor recruitment are largely unclear. We report that association of TLEs with O-GlcNAc transferase, an enzyme that catalyzes posttranslational modification of proteins by O-linked N-acetylglucosamine, is essential for TLE-mediated transcriptional repression. Removal of O-GlcNAc from Wnt-responsive gene promoters is critical for gene activation from Wnt-responsive promoters. Thus, these studies identify a molecular mechanism by which Groucho/TLEs repress gene transcription and provide a model whereby O-GlcNAc may control distinct intracellular signaling pathways.

cellular organisms, does. By placing O-GlcNAc modification on the transcription apparatus, OGT acts in concert with chromatin modification to inhibit transcriptional activators, such as Sp1 (18) and RNA polymerase II (20,21), to maintain transcriptional repression. Nevertheless, specificity for OGT in regulating distinct intracellular pathways at the transcriptional level has not yet been defined. Because OGT contains the TPR structure that is reminiscent of Cyc8, we were prompted to explore the possibility that OGT is a partner for Groucho/TLEs in regulating expression of a distinct subset of developmental genes.
Herein, we provide evidence for pathway-specific regulation of the TCF/LEF locus by O-GlcNAc. We show that OGT is targeted to TCF/LEF sites on endogenous Wnt-responsive gene promoters by its interaction with Groucho/TLE transcriptional corepressors. In addition, these promoters and their downstream genes are specifically regulated by O-GlcNAc modification. Together, these data suggest that O-GlcNAc can function in regulating distinct intracellular signaling pathways. Moreover, these data serve as a model for how O-GlcNAc may specifically regulate other intracellular signaling pathways through targeting of OGT to distinct sites of transcriptional repression via its interaction with certain corepressor molecules.

MATERIALS AND METHODS
Plasmids-Vectors for bacterial expression of the full-length and deletion mutants of OGT in fusion with glutathione S-transferase (GST) were described previously (17). Catalytically dead OGT (D925N) was generated with the QuikChange site-directed mutagenesis kit (Stratagene). Full-length or fragments of human TLE1 and human TLE2 were produced by PCR and subcloned into pcDNA3 mammalian expression plasmid in fusion with the Gal4 DNA-binding domain. pcDNA3.1-His-TLE1 and -TLE2 were kindly provided by G. Stein.
GST Pull-down Assays-All GST fusion proteins were expressed in Escherichia coli and were purified and immobilized by batch affinity chromatography on glutathione-Sepharose 4B (Amersham Biosciences) as described previously (17). [ 35 S]Methionine-labeled proteins were synthesized in vitro with a coupled transcription-translation system, TNT (Promega). 35 S-Labeled proteins were incubated with equal amounts of immobilized GST fusion proteins in binding buffer (50 mM Tris (pH 7.5), 10% glycerol, 100 mM NaCl, 0.1% Nonidet P-40, 1 mM EDTA, 1 mM DTT, 1 mM PMSF, and protease inhibitors) for 2 h at 4°C. Beads were washed five times with the binding buffer. Bound proteins were eluted with 1ϫ SDS-PAGE buffer, separated by SDS-PAGE, and visualized by autoradiography.
Cell-based Transcription Assays-HepG2 cells were transiently cotransfected with expression constructs for TLE1 or TLE2 (full-length and fragments) and OGT (wild-type or catalytically dead) plasmids either individually or in combination, together with the G5-Luc reporter construct as described previously (17). Transient transfection was performed using electroporation or Lipofectamine 2000 (Invitrogen) following the manufacturer's instructions. For luciferase assays, transfection efficiencies were normalized using a cotransfected ␤-galactosidase plasmid. ␤-Catenin/TCF-induced transcriptional activity was determined by using a ␤-catenin/TCF promoter-luciferase reporter construct, pTOPFLASH (22). Briefly, a dual luciferase reporter assay was performed whereby HEK293 cells were transfected with the experimental TCF promoter/luciferase reporter gene (TOPFLASH). A mutated TCF-luciferase reporter construct (pFOPFLASH) served as a negative control for TOPFLASH activity. A control reporter pRL-TK Renilla luciferase (Promega) was co-transfected in each sample to serve as an internal control for transfection efficiency. To promote Wnt signaling, Wnt1 conditioned medium was added 24 h after transfection, and luciferase activities were assayed 48 h after transfection. Experiments were performed in triplicate and repeated three times. Statistical analysis was performed using analysis of variance. A value of p Ͻ 0.05 was considered statistically significant.
Chromatin Immunoprecipitation Assay-ChIP assays were performed using previously described oligonucleotides and adhering generally to methods described previously (16). The oligonucleotide sequences used are listed in Table 1.
RNA Interference-pSUPER-based vectors were constructed that contain DNA templates for the synthesis of siRNAs and transfected using Lipofectamine 2000. The sequences used are listed in Table 2.
Real-time RT-PCR-RNA was extracted using TRIzol reagent (Sigma) and purified with an RNase Easy Kit (Qiagen). cDNAs were synthesized from 0.2 g of DNase-treated total RNA using the cDNA Archive Kit (Applied Biosystems). Relative RNA levels were quantified by real-time RT-PCR technology with an ABI PRISM 7700 detection system and SYBR Green reagent (Applied Biosystems). PCRs contained 1ϫ SYBR Green Master Mix, 66 nM primers, and cDNA equivalent to 10 ng of total RNA in a 15-l volume. Target mRNA levels were normalized against 36B4 or GAPDH mRNA level (as indicated) from the same total RNA sample. The primers used are listed in Table 3.

RESULTS
TLEs Physically Interact with OGT-As a first step to assess a possible interaction between TLEs and OGT, we coexpressed hexahistidine-tagged TLE1 or TLE2 with hemagglutinin (HA)tagged OGT in COS-7 cells. The result showed that TLE1 or TLE2 is coimmunoprecipitated with OGT, demonstrating that TLEs can physically associate with OGT (Fig. 1A). Direct protein-protein interactions between TLEs and OGT were confirmed by a glutathione S-transferase (GST) pull-down assay, in which TLE1 and TLE2 synthesized in vitro were able to bind to the purified GST-OGT fusion protein (Fig. 1B). In addition to multiple tandem TPRs at the N terminus, OGT contains two conserved domains (CDI and CDII) at the C terminus that contribute to its catalytic activity (23,24). Deletion analysis reveals that a region spanning the first six TPR motifs and the CD I FIGURE 1. Physical interaction between OGT and TLEs. A, coimmunoprecipitation analysis. COS-7 cells were transfected with HA-tagged OGT or His-tagged TLE1 or -2 expression vectors and immunoprecipitated with ␣-HA antibody, followed by immunoblotting with ␣-HA and ␣-His antibodies. B-E, mapping interaction domains in OGT and TLEs. TLE proteins and their deletion mutants were synthesized in reticulocyte lysate and were incubated with bacterially produced full-length OGT and deletion mutants. F, schematic representation of primary structures of TLE and OGT and the identified interactions. CcN, a region containing phosphorylation sites for Cdc2 and casein kinase 2; SP, serine/proline-rich domain; WD, WD40-repeat domains in tandem.

RT-PCR target
Primer sequence region in OGT independently bind to TLE1 and TLE2 (Fig. 1, B and C). As for the TLEs, both the conserved N-terminal glutamine-rich domain (Q domain) and the C-terminal WDrepeat domain interact with OGT (Fig. 1D). Interestingly, the Q domain of TLEs is able to bind to either TPR1-6 or CDI of OGT, whereas the WD domain mainly contacts OGT CDI (Fig.  1E). Together, the results suggest that there are multiple interfaces between an OGT molecule and a TLE molecule (Fig. 1F).
OGT Facilitates Transcriptional Repression of TLEs-Previous studies suggest that Groucho/TLEs mediate repression in part by recruiting HDACs (10,25). Surprisingly, when TLE1 was tethered to a luciferase reporter plasmid by fusion to the Gal4 DBD, it retained potent repression activity irrespective of the presence of the HDAC inhibitor, TSA ( Fig. 2A). In contrast, TSA was able to elevate Gal4 DBD-induced transcription from the reporter. These results suggest that HDACs are not sufficient to mediate TLE repression. It is known that both the Q and WD domains of the TLEs possess repression activity (26). These two domains physically interacted with OGT (Fig. 1, D and E), supporting the idea that their repression activity is dependent on OGT. Indeed, our results show that overexpression of OGT largely enhances transcriptional repression by either the Q or WD domain of TLE1 and -2, suggesting that both domains mediate repression by recruiting OGT (Fig. 2B). We next evaluated whether OGT could cooperate with TLE to inhibit transcription. The results indicate that although a moderate dose of exogenous OGT alone did not inhibit the reporter transcription by itself, OGT was capable of enhancing transcriptional repression by Gal4-TLE1 (Fig. 2C). OGT requires its endogenous catalytic activity to transfer O-GlcNAc moieties to proteins. We therefore used a catalytically dead OGT mutant lacking the ability to augment TLE1 repression by O-GlcNAc modification. Cotransfection of this inactive OGT mutant with TLE relieved TLE-mediated repression of the Gal4 reporter, suggesting that the enzymatic activity of OGT is necessary for TLE-mediated transcriptional repression (Fig. 2D).
TLEs Associate with OGT in the Absence of Wnt Signaling-To evaluate the in vivo role for OGT in TLE-mediated repres- sion, we analyzed the association of OGT with proteins known to form a complex with TCF/LEF during transcriptional inhibition and activation using Wnt1-conditioned media. OGT was immunoprecipitated with TLE1 or -2 in the absence of Wnt, and this association was disrupted in the presence of Wnt (Fig.  3, A and B). The reciprocal immunoprecipitation experiments further demonstrate that OGT is a component of the TLE corepressor complex on TCF/LEF in the absence of Wnt (Fig. 3, C  and D) but not in the presence of Wnt when activated ␤-catenin (␤-catenin*) associated with TCF/LEF (Fig. 3D). These data suggest that OGT interacts with TLE and TCF/LEF in a Wntresponsive manner. To assess whether OGT associates with endogenous Wnt-responsive gene promoters, we used a ChIP assay to identify proteins present at consensus TCF/LEF-binding sites on the promoters of cyclin D1 and c-MYC genes (27,28). In the absence of Wnt, OGT, TLE1, and TLE2 were associated with the LEF locus on cyclin D1 and c-MYC promoters, but Wnt activation displaces these repressor proteins from these promoters (Fig. 3E). Using an antibody against the activated form of ␤-catenin (␤-catenin*), we found that ␤-catenin* replaced OGT/TLE on TCF/LEF in response to Wnt activation (Fig. 3E). As a control, we found that OGT, TLE1, and TLE2 were constitutively bound to the Notch-responsive HES-1 locus within the COL2A1 gene in the absence and presence of Wnt (Fig. 3E). These results indicate that OGT is an endogenous component of the TLE-LEF complex that represses Wnt signal.
O-GlcNAc Is Required for Gene Repression by TLEs in the Wnt Pathway-The TOPFLASH system is commonly used to assess responsiveness at the TCF/LEF locus (29). This synthetic system compares expression of luciferase in response to TCF/LEFbinding sites (TOPFLASH) with that of mutant TCF/LEF-binding sites (FOPLASH). To evaluate the role of endogenous OGT in this system, we used several short hairpin RNAs combined with Wnt1-conditioned media, which was shown to activate expression of downstream genes (Fig. 4, A and B). Using shRNA to knock down levels of OGT (shOGT) caused a decrease in the O-GlcNAc level (Fig. 4A). Knockdown of OGT resulted in a dose-dependent derepression of the TCF/LEF locus on the TOPFLASH reporter (Fig. 4C). Additionally, luciferase expression on the FOPFLASH reporter was also increased to a certain extent, which is probably related to global derepression by reduction of O-GlcNAc levels (Fig. 4C). In the presence of Wnt, we observed a less striking difference between the TOPFLASH and FOPFLASH reporters (Fig. 4D). These results suggest that OGT represses transcription via TCF/LEF-dependent as well as -independent pathways.  O-GlcNAc moieties are removed from proteins enzymatically by O-GlcNAcase (OGA) (30). We examined whether OGA is essential for activation of the TCF/LEF locus. Knockdown of OGA by shOGA causes an increase in intracellular O-GlcNAc levels (Fig. 4A). OGA knockdown inhibited activation of the TCF/LEF locus by Wnt (Fig. 4E). Hence, OGA is required for Wnt activation and suppressed expression of Wntresponsive genes at both the translational (Fig. 4F) and transcriptional levels (Fig. 4, G-L). Thus, OGA is a distinct requirement for canonical Wnt signaling.
Amino Enhancer of Split (AES) Antagonizes TLE Repression by Competition with OGT-The AES protein is homologous to the N-terminal Q and GP domains of TLEs, and it can partially antagonize TLE-mediated repression in a dominant negative fashion (31,32). We employed AES as a tool to evaluate the structure-function relationship for the OGT-TLE interaction. To this end, we determined whether AES function as a derepressor lies in its ability to interfere with the TLE-OGT inter-action via the Q-domain of AES. The results show that the addition of AES inhibits TLE1 binding to OGT in a pull-down assay (Fig. 5A). In doing so, AES might bind directly to OGT through its Q domains and prevent this enzyme from associating with TLEs. Alternatively, AES might bind to TLEs and compete off OGT. In an attempt to distinguish these two possibilities, we show that, although AES contains the highly conserved Q domain, it fails to bind to OGT (Fig. 5B). In contrast, it directly interacts with TLE1 (Fig. 5C), which can be explained by the intrinsic affinity between the Q domains (33). Hence, the dominant negative function of AES is probably due to its competition with OGT for binding to the Q domains of TLEs.
Genetic and biochemical studies suggest that AES does not abrogate TLE repressive function entirely but rather fine tunes its function. Consistent with this notion, overexpression of AES only partially diminishes repression activity of the full-length TLE1 fused to Gal4 DBD in a luciferase reporter assay (Fig. 5D). Further analysis shows that AES specifically relieves repression imposed by the interaction between the Q domain of TLE1 and OGT, whereas it has no effect on repression by the WD domain via OGT (Fig. 5D). These results lead to a mechanistic interpretation for AES function as a moderate or partial "derepressor" (Fig. 5D).

DISCUSSION
OGT is a ubiquitous transcriptional regulator that plays important roles in gene repression (34 -36). This work provides evidence that O-GlcNAc modification can specifically regulate the canonical Wnt locus. We show that the enzyme OGT can directly interact with Groucho/TLE transcription factors. OGT associates with a LEF/TLE repression complex in vivo (Fig. 3). This repression complex is disrupted upon Wnt stimulation (Fig. 4). We provide a framework for the role of O-GlcNAc in regulating the canonical Wnt locus (Fig. 6).
We also show that genetic manipulation of OGT and OGA is sufficient to modulate a TCF/LEF reporter activity (TOPFLASH) (Fig. 4, C-E). Activation of the canonical Wnt locus is linked to expression of target genes, including c-MYC, ID2, and cyclin D1, whose expression levels are enhanced by lowering O-GlcNAc levels (Fig. 4, G-L). However, although O-GlcNAc modification appears to repress the canonical Wnt locus and prevent expression of Wnt-responsive genes, removal of O-GlcNAc moieties from proteins is not by itself indicative of activation of the canonical Wnt locus (Fig. 4) (i.e. shRNA  against OGT causes a similar activation of mutant TCF reporter versus wild-type TCF-reporter in the absence of Wnt). In contrast, for TCF/LEF activation, the removal of O-GlcNAc modification from proteins seems to be essential along with the presence of ␤-catenin for the Wnt conditioned media to achieve full activation. It should be pointed out that the studies by Olivier-Van Stichelen et al. (37,38) show that OGT enhances the transcriptional capability of ␤-catenin through stabilizing the protein in the cytosol. The contrary effect of OGT on ␤-catenin activity may reflect the spatiotemporal difference in OGT regulation in distinct cellular contexts.
AES has been hypothesized to antagnonize repression at the canonical Wnt locus (31,32). Herein, we use AES to demonstrate critical features of the structure-function relationship that exists between OGT and TLE. AES contains high homology with the Q domain of full-length TLE. We observed that multiple interaction motifs may exist between OGT and TLE ( Fig. 1) but that the Q domain of TLE, in particular, was more important for repression via OGT ( Fig. 2A). Using AES, we showed that AES may antagonize the ability of OGT to associate with TLE, specifically by relieving repression through the Q domain (Fig. 5D). This supports the notion that a physical interaction between OGT and TLE is functionally important in mediating repression by TLE.
Groucho/TLE-mediated transcriptional repression may involve multiple mechanisms, such as expression of partner repressors, competition with coactivators, and posttranslational modifications (13). The experiments presented here support a model in which OGT is a component of the TLE-LEF repression complex in the nucleus. OGT associates directly with the TLEs, thereby specifically repressing canonical Wnt signals. There are multiple interaction surfaces between the TLE and OGT proteins. Thus, the TLEs are able to effectively recruit OGT and form a versatile corepressor pair. It has been shown that mSin3A interacts with HCF-1 to enhance transcriptional repression (39). We and others demonstrate that OGT physically interacts with mSin3A and HCF-1 (17,40,41). The association of OGT with the general corepressor mSin3A (17) suggests that the covalent modification of the transcriptional apparatus by O-GlcNAc represents a general mechanism for gene repression in diverse signaling pathways. However, some degree of specificity must be required to maintain transcriptional control. Based on the TPR domain of OGT (42), it has been hypothesized that specificity for O-GlcNAc modification is derived from the association of OGT with specific repressor molecules. The current work supports this notion. Although TLEs may also recruit mSin3A, HDACs, and thereby OGT, the direct association of the TLEs with OGT may provide more specificity to the repression complex. Thus, our evidence supports a distinct role for O-GlcNAc modification regulating the canonical Wnt locus.
Our proposed model (Fig. 6) suggests that during transcriptional repression, OGT is targeted to the canonical Wnt locus by its interaction with the TLEs, where it may modify critical elements of the transcriptional apparatus, such as Sp1 and RNA polymerase II (17,21). AES may fine tune the system by antagonizing the interaction of OGT with TLEs. For activation of the TCF/LEF locus, stabilization of ␤-catenin and removal of O-GlcNAc residues by OGA are both necessary. Considering the wide array of proteins modified by O-GlcNAc, including transcriptional coactivators (43) and the proteasome (44), identification of a mammalian phenotype resultant from specific O-GlcNAc regulation of the canonical Wnt pathway may not be possible. Nevertheless, these data indicate that OGT activity may play a role in a variety of developmental processes regulated through the TCF/LEF locus.