T:G Mismatch-specific Thymine-DNA Glycosylase Potentiates Transcription of Estrogen-regulated Genes through Direct Interaction with Estrogen Receptor α*

Nuclear receptors (NR) classically regulate gene expression by stimulating transcription upon binding to their cognate ligands. It is now well established that NR-mediated transcriptional activation requires the recruitment of coregulator complexes, which facilitate recruitment of the basal transcription machinery through direct interactions with the basal transcription machinery and/or through chromatin remodeling. However, a number of recently described NR coactivators have been implicated in cross-talk with other nuclear processes including RNA splicing and DNA repair. T:G mismatch-specific thymine DNA glycosylase (TDG) is required for base excision repair of deaminated methylcytosine. Here we show that TDG is a coactivator for estrogen receptor α (ERα). We demonstrate that TDG interacts with ERα in vitro and in vivo and suggest a separate role for TDG to its established role in DNA repair. We show that this involves helix 12 of ERα. The region of interaction in TDG is mapped to a putative α-helical motif containing a motif distinct from but similar to the LXXLL motif that mediates interaction with NR. Together with recent reports linking TFIIH in regulating NR function, our findings provide new data to further support an important link between DNA repair proteins and nuclear receptor function.

Nuclear receptors (NR) 1 regulate gene expression upon binding to small lipophilic molecules. The NR superfamily encompasses high affinity receptors for the steroid hormones, vitamin D3, thyroid hormone, and retinoic acid as well as so-called "orphan receptors" that bind with low affinity to dietary lipids such as fatty acids, oxysterols, bile acids, and xenobiotics and a large number of receptors with no known ligand (1). Two related estrogen receptors (ER␣) are responsible for the major estrogenic responses in mammals, being required for male and female reproductive function, but also for bone maintenance in the cardiovascular system and in regulating certain brain functions (2,3).
NR share a common modular structure with a core DNA binding domain and a C-terminal ligand binding domain (LBD). The LBD functions as a ligand-dependent transcription activation domain (AF2) as well as providing a surface for interaction with other NR. Transcription activation is mediated by synergism between AF2 and a second transactivation domain (AF1) located N-terminal to the DNA binding domain (4,5). Whereas AF2 activity requires ligand binding, AF1 activity is subject to regulation by phosphorylation (6 -8).
Transcription regulation by NR occurs through direct interaction with the transcription machinery and/or by recruitment of coregulator proteins that facilitate interaction with the transcription machinery. Moreover, chromatin-remodeling complexes are now known to play an important role in gene transcription (9 -12). Several coactivator proteins associate with ligand-bound NR and include the p160 proteins, SRC1, TIF2/ GRIP1, and pCIP/ACTR/AIB1/RAC3/TRAM-1 (13). The thyroid hormone receptor-associated protein complex (14), which is very similar to the vitamin D3 receptor-interacting protein complex (15), interacts with liganded NR and is required for thyroid hormone and vitamin D3 receptor as well as ER␣ activities in vitro (16). CBP and the highly related p300, which copurify with RNA polymerase II (17) and associate with p160 coactivators, can directly interact with NR (18 -20). Additionally, a number of NR-interacting proteins, including PGC-1, TIF1, SUG1, and RIP140, mediate transcriptional regulation by NR (21)(22)(23)(24).
Coactivators associate with the LBD of NR via small ␣-helical motifs with a sequence conforming to the consensus LXXLL (25)(26)(27). Although the LXXLL sequence is essential for association with the LBD, sequences outside this motif are required for specific interaction because different LXXLL motifs are preferentially bound by different NR (28 -30). In the unliganded state, several NR bind to their responsive genes to repress gene expression, at least in part, through the recruitment of the corepressors N-CoR and SMRT (31)(32)(33)(34). NR antagonists also recruit N-CoR and SMRT to repress gene expression, for example, tamoxifen in the case of ER␣ (35,36) and GW6471 in the case of peroxisome proliferator-activated receptor ␣ (PPAR␣) (37). N-CoR/SMRT recruitment is mediated by an ␣-helical motif conforming to the consensus LXXX(I/H)-XXX(I/L) where I is isoleucine and H is histidine (38 -40), a similar but more extended sequence than the coactivator LXXLL motif. The highly conserved LBD is a "wedge"-shaped structure generally comprised of 12 ␣-helices containing a ligand binding pocket (41). Ligand binding displaces helix 12, revealing a charge clamp with which LXXLL coactivator peptides associate (28,42,43). In the case of antagonist-bound PPAR␣, the extended ␣-helical corepressor motif further displaces helix 12, preventing it from taking up its active position (37).
Recent studies have also implicated DNA repair proteins in transcription regulation by NR. The basal transcription factor TFIIH, required for nucleotide excision repair, also regulates the activity of a number of transcription factors including retinoic acid receptors ␣ and ␥, ER␣, and the androgen receptor (AR) (44 -47). BRCA1, which has been implicated in double strand break repair, represses ER␣ and stimulates AR activity (48 -51). Components of base excision repair are also implicated in regulating gene expression. A/P endonuclease Ref-1 activates c-Jun and p53 (52,53). TDG was identified as a c-Jun-interacting protein by yeast two-hybrid screening (54) and found to repress the activity of the homeodomain-containing thyroid transcription factor-1 (55). In the present study, we have investigated the role for TDG as a coactivator for ER␣. We show that TDG interacts with many NR, representative of the major groupings within the superfamily, and that TDG associates with ER␣ in a ligand-dependent manner. This interaction is mediated by a motif in TDG, which is similar to the NR coactivator motif required for coactivator recruitment by NR. Moreover, we demonstrate that TDG potentiates transcription activation by ER␣ and show that it is recruited to the promoters of estrogen-regulated genes in vivo.

MATERIALS AND METHODS
Plasmids-The expression vector pSG5 and constructs expressing human ER␣ and deletion mutants expressing N-or C-terminal portions of ER␣ (⌬AF2 and ⌬AF1, respectively) have been described as having HA-tagged TDG and TDG-N140A (56 -59). All of the other TDG mutants were obtained by site-directed mutagenesis. pGEX-AF2 and helix 12 mutants were kind gifts of M. Parker (60). ERE-2-TATA-CAT was kindly provided by B. Katzenellenbogen.
GST Pull-down Assays-In vitro transcription/translations were performed using TNT rabbit reticulocyte lysates (Promega) in the presence of [ 35 S]methionine. GST proteins were induced, and lysates were prepared as described previously (24). For pull-down assays, GST fusion proteins were purified by affinity chromatography on glutathione-agarose beads and retained as a 50% slurry in buffer C (20 mM Hepes, pH 7.6, 100 mM KCl, 1 mM EDTA, 1 mM dithiothreitol, 20% glycerol, protease inhibitors). 100 l of glutathione-agarose beads loaded with 10 g of GST fusion proteins were then used directly in binding assays with 10 l of radiolabeled polypeptides from in vitro translation by the addition of 890 l of low salt buffer (50 mM Hepes, pH 7.6, 250 mM NaCl, 0.5% Nonidet P-40, 5 mM EDTA, 0.1% bovine serum albumin, 0.5 mM dithiothreitol, 0.005% SDS, protease inhibitors). Following 1-h incubation at room temperature, the beads were washed twice with low salt buffer and twice with high salt buffer (low salt buffer containing 1 M NaCl). Samples were boiled for 10 min in 80 l of Laemmli buffer and fractionated by SDS-PAGE. Gels were dried and autoradiographed.
Immunoprecipitations and Immunoblotting-COS-1 cells were plated in 9-cm dishes in Dulbecco's modified Eagle's medium without phenol red supplemented with 5% dextran-coated charcoal-stripped fetal calf serum 16 -24 h prior to transfection with 4 g of expression plasmids using LipofectAMINE 2000 (Invitrogen). After an additional 48 h, the cells were treated with 10 nM E2 or vehicle (ethanol) for 2 h. Whole cell lysates were prepared in radioimmune precipitation buffer (150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholic acid, 0.1% SDS, 50 mM Tris-HCl, pH 7.5) supplemented with 1 mM phenylmethylsulfonyl fluoride and complete protease inhibitors (Roche Applied Science). The lysates (2 mg) were incubated with protein G-Sepharose (Sigma) for 1 h at 4°C followed by the addition of 2 g of HA antibody (Santa Cruz Biotechnology) for 30 min at 4°C. Immunoprecipitation using a monoclonal ␣-tubulin antibody (Sigma) was used as a control. The immunoprecipitates were washed in radioimmune precipitation buffer and resuspended in Laemmli buffer containing 5% v/v dithiothreitol, and immunoblotting was carried out using anti-HA peroxidase (Roche Applied Science) and anti-ER␣ (NovoCastra).
Chromatin Immunoprecipitation-Chromatin immunoprecipitation (ChIP) assays were performed following published methodology (62). E2 was added to a final concentration of 10 nM for 30 min prior to cell fixation and harvesting. Immunoprecipitations were performed using previously described antibodies against ER␣ (63), SRC1 (64), and TDG (59). Acetylated H3 monoclonal antibody was purchased from Upstate Biotechnology. PCR for the pS2 gene was performed as described previously (62). PCR for the progesterone receptor gene was performed using primers with the sequences 5Ј-AAAGGGGAGTCCAGTCGT-CATG-3Ј and 5Ј-TGCTGGTCCTGCGTCTTTTC-3Ј and span sequences in the human progesterone receptor (PR) gene that contains sequences identical to the ERE named E3 in the rat PR gene (65). PCR for the PSA gene was performed using oligonucleotides in the human PSA gene. These primers have been described previously as primers G/H (66).

RESULTS
T:G Mismatch-specific Thymine DNA Glycosylase Stimulates ER␣ Activity in a Ligand-dependent Manner-In COS-1 cells, TDG stimulated ER␣ activity ϳ3-fold in the presence of E2 and OHT but not in the absence of ligand or in the presence of ICI (Fig. 1B). TDG cotransfection with ER␣ lacking AF1 (⌬AF1) or AF2 (⌬AF2) (Fig. 1A) showed that deletion of either activation function prevented coactivation by TDG (Fig. 1C). Mutation of key residues in helix 12 inhibits recruitment of coactivators (60). Substitution of the hydrophobic residues at amino acid positions 539 and/or 540 abolished ER␣ activity and prevented coactivation by TDG (Fig. 1D). Substitution of the charged residues at position 542 or 538/542/545 significantly reduced ER␣ activity and reduced coactivation by TDG in a proportional manner. Substitution of tyrosine 537 by serine has been shown to result in ligand-independent ER␣ activity and SRC1 recruitment (69,70). By contrast, TDG only stimulated its activity in the presence of E2. These data are indicative of a requirement for helix 12 sequences for ER␣ interaction with TDG but suggest mechanistic differences in the recruitment of coactivators such as SRC1 and TDG by ER␣. Moreover, a mutation that abolishes the DNA glycosylase activity of TDG (N140A) (58) stimulated ER␣ activity in a similar manner to wild-type TDG, indicating that coactivation by TDG does not require the DNA glycosylase activity associated with TDG (see Fig. 3).
TDG Interacts with ER␣ in a Ligand-dependent Manner in Vitro-In GST pull-down assays, ER␣ and ER␣-⌬AF1 associated with TDG in a ligand-regulated manner. No interaction was evident for ER␣-⌬AF2 ( Fig. 2A). TDG interacted with AF2 in the presence of E2 (Fig. 2B). Some association was observed with OHT but not in the absence of ligand or in the presence of ICI. Helix 12 plays a crucial role in the ligand-dependent recruitment of coactivators by the LBD and mutations in helix 12 inhibit or prevent coactivator binding (60). In the case of mouse ER␣, substitution of the aspartic acid at positions 542 and 549 by asparagines together with the replacement of glutamic acid at position 546 by glutamine prevents SRC1 binding to AF2. TDG did not bind to this mutant (Fig. 2C). Substitution of the leucine 542 and 544 or the methionine 547 and leucine 548 by alanine prevented TDG interaction as has also been shown for SRC1. Interaction was unaffected by substitution of tyrosine 541 by phenylalanine as described for SRC1 (70). Taken together, these data indicate that TDG interacts with ER␣ in a ligand-dependent manner and that this interaction requires the integrity of AF2. Note also that substitution of glutamic acid at position 546 by alanine, which has little adverse effect on SRC1 binding to AF2 in vitro, did not bind TDG, which is suggestive of difference as well as similarities in the way that TDG and SRC1 interact with AF2.
Sequences in TDG Required for Interaction with ER␣ Are Similar to the NR Coactivator/Corepressor Motifs-The sequences in TDG required for interaction with ER␣ were investigated using deletion mutants of TDG in a yeast two-hybrid system. We employed a yeast strain in which the genomically encoded URA3 gene had been functionally replaced by an estrogen-responsive URA3 gene (ERE-URA3). In the absence of uracil, there is no growth unless ER␣ is introduced and E2 at a concentration of 1 nM or greater is present (data not shown). In the presence of 0.1 nM E2, there was no growth if uracil was omitted. Transformation with SRC1 fused to the activation domain of VP16, however, enabled growth in the presence of 0.1 nM E2, indicative of a ligand-dependent interaction between ER␣ and SRC1 (data not shown). VP16 fused to amino acids 32-421 of murine TDG also supported growth in the presence of ER␣ and 0.1 nM E2, consistent with an interaction between ER␣ and TDG. Deletions from the C terminus were suggestive of a requirement for sequences between amino acids 307 and 346. Deletions from the N terminus demonstrated that sequences 47-122 are important for TDG interaction with ER␣ (Fig. 3A).
Examination of the TDG sequence highlighted an evolutionarily conserved region between amino acids 115 and 146 of human TDG with a similarity to the NR-interacting motifs found in NR coactivators and corepressors (Fig. 3B). GST pulldown assays demonstrated that substitution of the leucine 132, valine 135, and isoleucine 136 by alanine prevented interaction with AF2, whereas other mutations in this region did not significantly affect the interaction (Fig. 3C). The requirement for the region of TDG around amino acids 132-136 for its interaction with AF2 was further explored by performing competition experiments employing peptides corresponding to amino acids 115-146 of human TDG. A peptide corresponding to wild-type TDG sequences or a peptide in which the leucines corresponding to amino acids 119 and 120 of TDG were substituted competed for interaction of TDG with ER␣ AF2 (Fig. 3D, lanes 4 -7 and 8 -11). A peptide in which the leucine at residue 132 was substituted by alanine did not block the interaction of TDG with GST-AF2 (lanes 12-15). A peptide in which the valine at residue 135 and isoleucine at residue 136 were substituted by alanine also did not compete (lanes 16 -19). Similarly, a peptide in which the residues at positions 132, 135, and 136 were substituted by alanine did not block TDG binding to GST-AF2 (lanes 20 -23) nor was competition observed using a peptide corresponding to amino acids 330 -346 of TDG (P6) (lanes 24 -28).
In transient transfection experiments, TDG-132A and TDG-135A/136A failed to stimulate ER␣ activity, whereas TDG-119A/120A, TDG-124A, and TDG-127A/128A stimulated ER␣ activity (Fig. 3E). Immunoblotting of the lysates showed that expression levels of the TDG mutants and of ER␣ were similar in each case. The TDG-N140A mutant stimulated ER␣ activity as well as wild-type TDG, suggesting that the DNA glycosylase activity is not required for coactivation by TDG.
Finally, in a mammalian two-hybrid assay, amino acids 116 -146 of TDG stimulated the activity of the Gal4 DNA binding domain fused to AF2 (Fig. 3F). Substitution of valine 135 and isoleucine 136 prevented the interaction. Taken together, these data indicate that a motif similar to LXXLL motif is required and is sufficient for interaction with the LBD/AF2 of ER␣.
TDG Interacts with ER␣ in an Estrogen-dependent Manner in vivo-Co-transfection of HA-tagged TDG and ER␣ followed by immunoprecipitation with antibodies to the HA tag resulted in the coprecipitation of ER␣ in the presence but not in the absence of E2, indicating that TDG interacts with ER␣ in a ligand-dependent manner in vivo (Fig. 4A, compare lanes 9  and 18).
To see whether TDG is recruited to the promoters of endogenous estrogen-responsive genes following estrogen treatment, ChIP assays were performed. The ER␣-positive MCF-7 cells were grown in the absence of estrogen prior to the addition of 10 nM E2 for 30 min. The presence of endogenous TDG and ER␣ on the estrogen-responsive regions of the PR and pS2 genes was determined by semiquantitative PCR using specific pairs of primers spanning the estrogen-responsive regions in the FIG. 3. Interaction of TDG with ER␣ is mediated by a motif similar to LXXLL. A, VP16 fused to regions of mouse TDG was coexpressed with ER␣ in the presence of 1 nM E2, and the yeast were scored for growth on medium lacking uracil. B, amino acid sequences of TDG from different species are aligned with the NR-interacting regions of N-CoR and SMRT. Also shown is the consensus sequence for the corepressor and coactivator binding motifs for NR (37). Positions of TDG residues 132/135/136 are shown by dots. C, pull-down assays were done following incubation of TDG or mutants with GST-AF2 with 100 nM E2. D, GST pull-down assays using 35 S-labeled TDG in the presence of 1 M (lanes 4, 8, 12, 16 promoters of these genes. Treatment with E2 induced a significant increase in the occupancy of the pS2 and PR gene promoters by ER␣ and SRC1 (Fig. 4B, lanes 5-6 and 9-10), in agreement with previous reports (62,71). The addition of E2 also stimulated histone H3 acetylation (lanes 11 and 12), whereas estrogen-responsive gene promoters were not immunoprecipitated using an irrelevant (FLAG) antibody (lanes 3 and 4). TDG was recruited following E2 treatment as seen for ER␣ and SRC1 (lanes 7 and 8), indicating that TDG is recruited to estrogen-responsive gene promoters by ER␣ in a ligand-dependent manner.
Other Nuclear Receptors Interact with TDG but Show Differing Ligand and Sequence Requirements for their Interactions-We have shown that TDG interacts with ER␣ in a liganddependent manner. GST pull-down assays showed that the AR, glucocorticoid receptor, and PR associate with TDG in the presence of their cognate ligand but not in its absence and do not interact with TDG(AAA) (Fig. 5). By contrast, all-transretinoic acid ␣, PPAR␥, thyroid hormone ␣, and vitamin D3 were pulled down in the presence or absence of ligand by TDG and by TDG(AAA). These preliminary data indicate that TDG can interact with many nuclear receptors but suggest that the mechanisms of their interactions are likely to be different. DISCUSSION TDG is a T:G and U:G mismatch-specific DNA repair enzyme that has previously been shown to stimulate the activity of retinoid receptors (54,68,72). We have shown that TDG stimulates the activity of ER␣ in a ligand-dependent manner. This effect was mediated by the LBD/AF2 and required the integrity of helix 12 sequences. TDG also interacted with ER␣ in GST pull-down assays. Mapping of the region(s) of TDG required for the interaction in yeast demonstrated a requirement for amino acids 47-346, whereas previous studies have demonstrated that sequences 122-346 are required for TDG interaction with retinoid receptors (68). An analysis of these sequences highlighted a putative ␣-helical motif between amino acids 115 and 146, similar to the LXXLL coactivator motif and the corepressor motif (consensus sequence, LXX(I/H)IXXX(L/I)), required for interaction with NR LBD/AF2. Mutation of leucine 132, valine 135, and isoleucine 136 prevented TDG interaction with ER␣ in vitro, indicating that the interaction is mediated by the sequence motif LDIVI, similar to the LXXLL coactivator motif.
These sequences lie within the glycosylase domain of TDG, raising the possibility that mutations in this region fail to interact with ER␣ because they disrupt the glycosylase domain. However, peptides corresponding to this region competed TDG interaction with AF2, whereas peptides containing alanine at positions 132 or 135 and 136 did not compete, indicative of the importance of these residues for interaction between TDG and ER␣. Although we cannot completely rule out the possibility that disruption of the glycosylase activity may be important in preventing interaction, we note that while mutation of leucine 132 or valine 135 and isoleucine 136 prevented stimulation of ER␣ activity, TDG-N140A, which lacks glycosylase activity, stimulated ER␣ activity as well as wild-type TDG. Thus, the DNA glycosylase activity is not required for stimulation of ER␣ activity by TDG.

FIG. 4. TDG interacts with ER␣ in vivo and is recruited to the promoters of E2-regulated genes in an E2-dependent manner. A,
COS-1 cell extracts were immunoprecipitated with an HA antibody following transfection with ER␣ and HA-TDG. E2 (10 nM) was added 2 h prior to harvesting. Immunoprecipitates were immunoblotted using ER␣ and TDG antibodies. Immunoprecipitations for ␣-tubulin served as a control. The polypeptide observed with this antibody is likely to represent immunoglobulins. B, MCF-7 cells were stimulated with 100 nM E2 for 30 min prior to fixation and harvesting. Immunoprecipitations were performed using antibodies to the FLAG epitope, ER␣, TDG, SRC1, or acetylated histone H3. Immunoprecipitated DNA was amplified by PCR (21-25 cycles) using primers for the pS2 and PR genes directed against regions of the genes as shown schematically. Positions of the EREs and the size of the PCR products are shown. GST pull-down assays confirmed previous findings, which showed that retinoid receptors all-trans-retinoic acid ␣ and 9-cis-retinoic acid ␣ interact with TDG in a ligand-independent manner (68). PPAR␥, thyroid hormone ␣, and vitamin D3 also interacted with TDG in a ligand-independent manner, whereas the steroid receptors AR, glucocorticoid receptor, and PR showed ligand-dependent association. Furthermore, mutation of amino acids 132, 135, and 136 in TDG prevented interaction with the steroid receptors but did not inhibit non-steroid receptor binding, suggestive of intrinsic differences in the way that the steroid and non-steroid receptors interact with TDG.
In addition to the demonstration that TDG stimulates ER␣ activity in reporter gene assays, further evidence for ER␣/TDG interaction was provided by coimmunoprecipitation experiments, which demonstrated that TDG interacts with ER␣ in a ligand-dependent manner in vivo. Moreover, the ChIP assay showed that TDG is recruited to the promoters of estrogenregulated genes in an estrogen-dependent manner in the ER␣positive MCF-7 breast cancer cell line. These data are indicative of an interaction between endogenous ER␣ and TDG at the promoters of estrogen-regulated genes.
The fact that TDG-N140A is able to stimulate ER␣ activity as well as wild-type TDG suggests that coactivation by TDG does not require DNA glycosylase activity. Indeed, a recent study has shown that TDG associates with CBP and stimulates its transcriptional activity, and it has been postulated that the recruitment of CBP by TDG may stimulate histone acetylation to facilitate base excision repair (73). Additionally, however, TDG may facilitate and/or stabilize CBP recruitment to transcription factors. We have also found that TDG can also interact with SRC1, lending further support to the conclusion that TDG can recruit coactivators to DNA to stimulate transcription 2 and suggesting that TDG is present in complexes with CBP/p300 and the p160 coactivators.
Much evidence is accumulating to link DNA repair proteins to the processes of gene regulation, primarily by modulating the activities of transcription factors. Regulation of transcription factor activities by DNA repair proteins may be due at least in part to the interaction of the latter with transcriptional coactivators as evidenced by the interaction of TDG and BRCA1 with CBP. In addition, transcription factors such as ER␣ may be capable of sensing DNA lesions on or around their binding sites in gene promoters and recruit DNA repair enzymes such as TDG. Transcription factor and DNA repair protein recruitment of chromatin-remodeling complexes would facilitate repair of such lesions. A recent study showing that all-trans-retinoic acid ␣ can interact with and stimulate DNA repair by TFIIH (74) indicates that transcription factors do indeed play a role in DNA repair. Whether ER␣ and other nuclear receptors stimulate DNA repair by TDG remains to be determined.