Deoxyribonucleic Acid Methyl Transferases 3a and 3b Associate with the Nuclear Orphan Receptor COUP-TFI during Gene Activation

Abstract

Transcriptional activation of silent genes can require the erasure of epigenetic marks such as DNA methylation at CpGs (cytosine-guanine dinucleotide). Active demethylation events have been observed, and associated processes are repeatedly suspected to involve DNA glycosylases such as mCpG binding domain protein 4, thymine DNA glycosylase (TDG), Demeter, and repressor of silencing 1. A complete characterization of the molecular mechanisms occurring in metazoan is nonetheless awaited. Here, we report that activation of the endogenous vitronectin gene in P19 cells by the nuclear receptor chicken ovalbumin upstream promoter-transcription factor I (COUP-TFI) is observed in parallel with the recruitment of TDG and p68 RNA helicase, two components of a putative demethylation complex. Interestingly, when activated, the vitronectin gene was loaded with DNA methyltransferases 3a and 3b (Dnmt3a/b), and a strand-biased decrease in CpG methylation was detected. Dnmt3a was further found to associate with COUP-TFI and TDG in vivo, and cotransfection experiments demonstrated that Dnmt3a/b can enhance COUP-TFI-mediated activation of a methylated reporter gene. These results suggest that Dnmt3a/b could cooperate with the orphan receptor COUP-TFI to regulate transcription of the vitronectin gene.

DNA METHYLATION ON cytosine is an epigenetic mark found in vertebrates, plants, fungi, and, to a lesser extent, insects. Methylation, occurring mainly at CpGs (cytosine-guanine dinucleotide), has profound effects on gene activity. Indeed, it has been recognized that hypomethylated DNA is found in active genes, whereas hypermethylated genes are silent (1, 2). Furthermore, methylated CpGs recruit specific proteins that in turn recruit histone modifiers involved in chromatin condensation (1). The patterns of DNA methylation are established and maintained by DNA methyl transferases (Dnmt) that can act either on hemimethylated CpGs generated throughout the replication of DNA (Dnmt1 or Dnmt3a and 3b), or de novo on unmethylated DNA (Dnmt3a and 3b). Maintenance methylation occurs during the S phase, with this coupling allowed by a physical interaction between Dnmt and the replication machinery. Conversely, mechanisms regulating de novo methylation are poorly understood. However, as described for the oncogenic promyelocytic leukemia-retinoic acid receptor fusion protein (3), targeting could be achieved through direct interaction of the methyltransferases with sequence-specific DNA-binding proteins (4, 5).

It has long been thought that DNA methylation is a stable mark that cannot be enzymatically removed, with multiple rounds of cell division without Dnmt-mediated remethylation required to completely erase this epigenetic mark from DNA. However, in the past recent years, numerous examples of active demethylation events that would not rely on cell division were described (6–9). Active demethylation requires enzymatic machineries that are able either to remove the methyl group branched on cytosines, or to replace the methylated nucleotides by unmethylated ones (10). Although 5-methylcytosine DNA glycosylases have been shown in plants (11–14), the data available from metazoan species are highly controversial and no demethylase has been yet formally identified. Nonetheless, a recent study demonstrated that active demethylation in mammalian cells involves DNA strand breaks suggesting that DNA glycosylases and repair systems are used (15). Among the candidate proteins, the T:G mismatch DNA glycosylase (TDG) copurifies with the p68 RNA helicase (p68) as well as small RNAs, within a complex isolated from chicken embryos, which shows a 5-methylcytosine DNA glycosylase activity (16–19). This enzyme is involved in DNA repair processed through the base excision repair (BER) mechanism. In BER, TDG first recognizes U:G or T:G mismatches generated by spontaneous deamination of cytosines and methylcytosines, respectively. T or U bases are then flipped out of the DNA molecule and cleaved by TDG, generating abasic sites [apurinic/apyrimidic (AP)] that are then repaired by the sequential recruitment of an AP endonuclease, DNA polymerase β, and a DNA ligase (20). From the work of Zhu et al. (19), it is postulated that TDG can flip out 5MeC and cut the modified base. Nonetheless, 5MeC still establishes normal pairing with G and is not likely to distort DNA (21, 22). This could explain the very low efficiency of the reaction when compared with T or U removal from mismatches (19).

TDG has been shown to interact with a number of transcription factors, and especially with nuclear receptors (23–28). The physical association of retinoic acid receptor or estrogen receptors (ER) with TDG results in higher transcriptional activation of reporter genes; and, at least for ER, such transcriptional coactivation does not require a functional glycosylase catalytic domain (23). On the other hand, Jost et al. (25) reported that the interaction of ER with TDG increases the 5-methylcytosine DNA glycosylase activity of the enzyme. These studies indicate that TDG is a multifunctional protein that could be involved in both DNA repair and transcriptional regulation. Correspondingly, the fact that the histone acetyl transferase CBP can interact with and acetylate TDG is likely to reflect a molecular switch regulating the functional commitment of the glycosylase. Acetylation of TDG inhibits its association with the AP endonuclease, suggesting that it could decrease TDG repair activity (26). Another molecular switch could be triggered by the sumoylation of TDG because this posttranslational modification dramatically reduces TDG binding to DNA and abrogates G:T mismatch processing (29). However, the processes regulating TDG sumoylation are yet unknown and it is still unclear whether sumoylated TDG retains its transcriptional activities.

Here, we investigated the recruitment of various factors on the vitronectin gene during activation by the nuclear orphan receptor chicken ovalbumin upstream promoter-transcription factor I (COUP-TFI) in P19 embryonal carcinoma cells (30). We found that, besides the increased level of TDG association upon transcription activation, COUP-TFI binding to the promoter enhanced the recruitment of Dnmt3a, Dnmt3b, and p68. These changes in protein loading of the vitronectin gene occur while the promoter becomes undermethylated and the gene is actively transcribed. Our data suggest that Dnmt3a/b, p68 and TDG may be involved in the dynamic process of transcription activation.

RESULTS

The Endogenous Vitronectin Gene Promoter Becomes Hypomethylated When Activated by COUP-TFI

To investigate the mechanisms involved in COUP-TFI-mediated transcriptional activation of the vitronectin gene, we generated P19 cell lines expressing COUP-TFI tagged with the hemagglutinin (HA) epitope. Stable transfectants were selected for G418 resistance and for their ability to express detectable levels of HA::COUP-TFI fusion protein. The transcription of the transgene is under the control of the cytomegalovirus (CMV) promoter that can be super-activated when cells are treated with 8-bromo-cAMP (8-Br-cAMP). This provides the opportunity to control the levels of HA::COUP-TFI in P19 cells (Fig. 1A), and thus to avoid the retinoic acid treatment which is used to induce endogenous COUP-TFI gene expression (31) to a similar level (30). When cells were exposed to 8-Br-cAMP, an induction of the transgene was detected after 4 h, followed by the induction of the vitronectin gene after 10 h, whereas PO expression remained constant (Fig. 1B). Neither COUP-TFI nor vitronectin endogenous genes were induced in the control cell lines when treated with 8-Br-cAMP (data not shown). To determine the potential variations in the level of CpG methylation during COUP-TFI-induced activation, we ran bisulfite sequencing of the proximal promoter (32). This method allows the precise determination of the methylation status of all CpGs present in a given genomic DNA fragment. We performed these analysis on genomic DNA extracted from control or 8-Br-cAMP-treated cells expressing COUP-TFI for 48 h. As a control, similar treatments were performed on pCR3.1 P19 cells. The results of these experiments, summarized in Fig. 1C, showed: 1) that the vitronectin gene promoter is hypermethylated in undifferentiated P19 cells, correlating with its transcriptionally silent state; 2) that 8-Br-cAMP on its own does not modify CpG methylation levels; and 3) that COUP-TFI expression is correlated to a decrease in CpG methylation levels (Fig. 1C). Furthermore, when comparing the strand-specific methylation patterns, it was apparent that hypomethylation occurred on the transcribed (bottom) strand, with the coding (top) strand remaining largely unaffected. When comparing +/− 8-Br-cAMP situations in a two-sided t test, methylation varied significantly only in the case of the bottom strand in HA::COUP-TFI expressing cells (Fig. 1C). Even if both sequenced strands did not originate from the same initial DNA molecules, these data suggest a bias in strand methylation. In summary, the vitronectin gene proximal promoter is hypermethylated when silent, and CpGs become hypomethylated when transcription occurs.

Fig. 1.

COUP-TFI Induces Hypomethylation of the Vitronectin Gene Promoter DNA A, Western blot analysis of transgene expression with HA antibody. High levels of COUP-TFI protein were detected in HA::COUP-TFI expressing P19 cells treated with 1 mm 8-Br-cAMP for 24 h, whereas low but detectable levels were observed in the absence of 8-Br-cAMP. B, Semiquantitative RT-PCR analysis of COUP-TFI and vitronectin expression in stable cell lines treated with 1 mm 8-Br-cAMP. PO is a ribosomal phosphoprotein gene used as a control. C, DNA was prepared from stable P19 cells treated with 1 mm 8-Br-cAMP for 48 h, and the DNA methylation status of individual CpG sites was determined by sequencing sodium bisulfite-modified DNA. The positions of each CpG, as well as COUP-TFI binding sites, are indicated on the scheme by black dots and gray rectangles, respectively. For each condition, strand-specific PCR products were subcloned, and 10–11 clones were subjected to nucleotide sequence analysis. For each clone, the methylation status is depicted by black and white squares indicating methylated and unmethylated CpGs respectively. Squares above the scheme show methylation status of the top strand (5′–3′), whereas bottom strand (3′–5′) methylation is depicted below the scheme. To determine the statistical significance of the methylation changes observed, the total methylation at all CpGs was calculated for each clone. The distributions obtained for the upper and lower strand in absence and presence of 8-Br-cAMP were compared using a two-sided t test. P values are indicated in parentheses.

COUP-TFI Expression Increases TDG, p68, and Dnmt3a/b Mobilization onto the Vitronectin Gene Promoter in Vivo

TDG DNA glycosylase has been suggested to be part of a demethylation complex together with the p68 RNA helicase (p68), and is known to interact with nuclear receptors (16–19, 23). Because we observed a loss of DNA methylation during activation of the vitronectin gene, a COUP-TFI-mediated loading of the TDG onto the vitronectin promoter was tested in vivo by chromatin immunoprecipitation (ChIP) assays using chromatin extracts of P19 cells expressing COUP-TFI or not, after a 24 h 8-Br-cAMP treatment. Oligonucleotides were designed to amplify the proximal promoter −252/−53 fragment of the vitronectin gene and the proximal promoter −279/−84 region of the PO control gene (Fig. 2A). First, we showed that the activation of the vitronectin gene by COUP-TFI correlates with the binding of the orphan receptor to the promoter in vivo (Fig. 2, B and C). Additionally, this binding correlates with the detection of an activated RNA Pol II ChIP signal (phospho-polymerase, P-RNA Pol II), as well as with the loss of the dimethylated K9 histone H3 signal and the release of Heterochromatin Protein 1α from the promoter, two marks of heterochromatin (Fig. 2, B and C). Finally, although already present in the absence of COUP-TFI, the amount of TDG recruited to the vitronectin promoter was clearly enhanced when COUP-TFI was expressed (Fig. 2, B and C). These data suggest the existence of a functional relationship between the nuclear receptor and the glycosylase. Conversely, the active PO gene, which transcription is nor differentiation dependent nor COUP-TFI dependent, did not show variations in the recruitment of these various proteins upon COUP-TFI expression (Fig. 2, B and C). Because we had shown that CpG methylation decreased during the activation of the vitronectin gene, we also examined the presence of Dnmt1, Dnmt2, Dnmt3a, and Dnmt3b on the vitronectin promoter by ChIP assays, and found that Dnmt3b was present on the silent gene (Fig. 2, D and E). Importantly, the recruitment of Dnmt3a and Dnmt3b was enhanced when the promoter was active (Fig. 2, D and E). Through sequential ChIP (Re-ChIP) experiments, we then analyzed the simultaneous presence of COUP-TFI, TDG, p68, the activated RNA Pol II, Dnmt3a, and Dnmt3b on the vitronectin promoter. TDG was not found associated with COUP-TFI in Re-ChIP assays, suggesting that these two proteins are not simultaneously present in DNA-bound complexes (Fig. 3). Nonetheless, Re-ChIP experiments demonstrated that p68 was recruited to the activated vitronectin promoter and found in both COUP-TFI- and TDG-bound chromatin fractions (Fig. 3A). Interestingly, p68 was not found associated with the vitronectin promoter in the TDG-bound fraction of chromatin when the gene was not transcribed, suggesting that p68 association with TDG-bound chromatin could be dependent on transcription, and that COUP-TFI could mediate the assembly of a p68/TDG complex on the active gene. Re-ChIP experiments further suggested that Dnmt3a and Dnmt3b could be associated to both COUP-TFI and TDG on transcriptionally active chromatin, and that TDG and Dnmt3 are found on inactive chromatin as well.

Fig. 2.

Expression of COUP-TFI Favors TDG and Dnmt3 Loading on the Vitronectin Promoter in Vivo ChIP experiments performed using chromatin prepared from P19 stable cell lines expressing or not HA::COUP-TFI and treated for 24 h with 1 mm 8-Br-cAMP. A, Schematic representation of the PO and vitronectin gene promoters indicating COUP-TFI binding sites (NR1 and NR2, black boxes) as well as the positions of the oligonucleotides used for PCR and QPCR (open arrows). B and D, The recruitment of COUP-TFI (αCOUP-TFI and αHA), phosphorylated RNA Pol II (αP-RNA Pol II), dimethyl histone H3 (αdiMeK9 H3), Heterochromatin Protein 1 α (αHP1α), TDG (αTDG), Dnmt1 (αDnmt1), Dnmt2 (αDnmt2), Dnmt3a (αDnmt3a), and Dnmt3b (αDnmt3b) to the vitronectin and control PO gene promoters was monitored through immunoprecipitation with the indicated antibodies after cross-linking of proteins to DNA, and subsequent amplification of the vitronectin gene −252/−53 fragment and PO gene −279/−84 fragment by PCR. C and E, Quantitative analysis of the ChIP signal through real-time PCR. Results are expressed as a percentage of the signal obtained with the input chromatin DNA.

Fig. 3.

p68 and Dnmts Associate with TDG or COUP-TFI as Separated Complexes onto Activated Vitronectin Gene Promoter ChIP experiments were performed using chromatin prepared from P19 stable cell lines expressing or not HA::COUP-TFI and treated for 24 h with 1 mm 8-Br-cAMP. A second immunoprecipitation (Re-ChIP) was then run on HA or TDG immunoprecipitates with the indicated antibodies before amplification of the vitronectin and PO gene promoters. αFlag antibody was used as a negative control. Note that COUP-TFI and TDG are not simultaneously found on transcriptionally active chromatin. Pol II, Polymerase II.

TDG, p68, and Dnmt3a Interact with COUP-TFI

Because TDG has been shown to interact with all nuclear receptors tested to date, we examined whether it could also interact with COUP-TFI. Even if Re-ChIP data showed the exclusive engagement of COUP-TFI and TDG at the vitronectin promoter, such an interaction could be involved in the observed increased occupancy of the vitronectin promoter by TDG upon COUP-TFI expression. This hypothesis was first stressed by glutathione S-transferase (GST) pull-down experiments that demonstrated that full-length in vitro-translated COUP-TFI interacts with a TDG produced as a GST fusion protein (Fig. 4A). This interaction was also observed in vivo. Indeed, a Flag::TDG fusion protein coimmunoprecipitated with HA::COUP-TFI when coexpressed in COS-7 cells (Fig. 4B). To delineate the domains of COUP-TFI interacting with TDG, several in vitro-translated truncated COUP-TFI proteins were produced and incubated with recombinant GST::TDG (Fig. 4A). Reverse experiments using in vitro-translated TDG and GST::COUP-TFI fusion proteins were also run (Fig. 4C). In accordance with the data shown Fig. 4C, in vitro-translated COUP-TFI 153–423 and bacterially expressed TDG did not interact in pull-down assays (data not shown). These assays determined that the DNA binding domain of COUP-TFI, formed by residues 84–153, interacts with TDG (Fig. 4D).

Fig. 4.

COUP-TFI Interacts with TDG, p68, and Dnmt3a A, Autoradiograph of in vitro-translated ³⁵S-labeled COUP-TFI 1–423 (full-length protein), HA::COUP-TFI 84–423, HA::COUP-TFI 57–317 and HA::COUP-TFI 57–153 pulled down by GST (negative control) or GST::TDG fusion proteins. Input lane corresponds to 20% of the ³⁵S-labeled protein used in the experiments. Note that two bands are generated for HA::COUP-TFI 84–423, HA::COUP-TFI 57–317 and HA::COUP-TFI 57–153 due to internal translation initiation downstream of the HA sequence. B, COUP-TFI directly interacts with TDG in vivo. COS-7 cells were cotransfected with pCDNA3 HA/COUP-TFI, pXJ141-Flag/TDG and empty vectors as indicated. Cell extracts were subjected to immunoprecipitation (IP) with αFlag antibody. The resulting precipitates were analyzed by Western blot using αHA and αFlag antibodies. Five percent of the input (I) used for immunoprecipitation was directly subjected to immunoblotting. C, Autoradiograph of in vitro-translated full-length TDG pulled down by GST (negative control) or GST fusion proteins that include residues 57–423 or 153–423 of COUP-TFI. The bound, in vitro-translated, TDG is indicated by an arrow on the left. Input lane corresponds to 20% of the ³⁵S-labeled in vitro-translated TDG used. D, Schematic representation of the different constructs used to determine the domains of COUP-TFI binding to TDG in GST pull-down assays. The results of the consecutive in vitro interaction assays are summarized on the right with + indicating interaction, and − indicating no interaction. Autoradiograph of in vitro-translated full-length p68 (E) or full-length SRC-1 (F) pulled down by GST or GST::COUP-TFI (aa 57–423) recombinant proteins. Input lanes correspond to 50% or 100% of ³⁵S-labeled in vitro-translated p68 or SRC-1 proteins used in the experiments, respectively. Pull-downs used GST fusions previously in vitro phosphorylated by recombinant PKC or MAPK (Erk2), as indicated. G, Endogenous Dnmt3a from P19 cells coimmunoprecipitates with COUP-TFI. pCR3.1 and HA::COUP-TFI expressing cells were lysed and HA::COUP-TFI was immunoprecipitated (IP). Recovered proteins were subjected to Western blot analysis with αDnmt3a or αHA. H, Endogenous Dnmt3a from P19 cells coimmunoprecipitates with TDG. P19 cells were transfected with expression vectors for wt or mutant protein versions of HA-tagged TDG. Cells were then lysed and HA::TDGs were immunoprecipitated (IP). Recovered proteins were subjected to Western blot analysis with αDnmt3a or αHA.

We also assayed the occurrence of an interaction between COUP-TFI and p68 in vitro by GST pull-down assays. These experiments evidenced the interaction between these two proteins (Fig. 4E). Moreover, when recombinant GST::COUP-TFI protein was phosphorylated in vitro by recombinant Erk2, we observed a marked increase in its ability to interact with p68. In contrast, the in vitro phosphorylation of GST::COUP-TFI by protein kinase C (PKC) had no effect (Fig. 4E). As a control, both kinases did not influence the level of association between COUP-TFI and SRC-1 (Fig. 4F). Interestingly, we have previously demonstrated that MAPKs positively regulate COUP-TFI activity (33). Such a regulation could thus be linked to the enhanced recruitment of p68 by phosphorylated COUP-TFI, as shown in the case of the ERα/p68 interaction (34). When HA-tagged COUP-TFI, TDG, TDG N140A (glycosylase defective; Ref. 35), and TDG K330A (sumoylation defective; Ref. 29) were used to fish out endogenous Dnmt3a from P19 cells by coimmuprecipitation assays, we found that Dnmt3a could interact with the orphan receptor as well as with the wild-type (wt) and mutated glycosylase (Fig. 4, G and H). Furthermore, Flag-Dnmt3b coimmunoprecipitated with COUP-TFI and TDG (data not shown). Together with our Re-ChIP experiments, these interaction data support the hypothesis that COUP-TFI/p68, COUP-TFI/Dnmt3a-3b, TDG/p68 and TDG/Dnmt3a-3b, could be assembled as complexes on chromatin.

Next, we investigated the output of COUP-TFI/TDG interaction in terms of DNA recognition. Indeed, the fact that TDG interacts with the orphan receptor through its DNA binding domain suggested that such a complex would not be able to bind DNA. We first ran EMSA with probes containing one COUP-TFI binding site from the vitronectin promoter, directly adjacent to a CpG (position −164, relative to the transcription start site) found to be heavily methylated in the absence of COUP-TFI and methylated to a much lower extent following the expression of COUP-TFI (see Fig. 1). Three probes were used: one with a hemimethylated CpG (5MeC), one with a T:G mismatch in the CpG (with the T in the upper strand), and one with no modification (Control). These experiments used recombinant full-length TDG and COUP-TFI [amino acids (aa) 57–423] purified to homogeneity. The assays showed that COUP-TFI binds equally well to the three probes, whereas TDG was better retained on the mismatch-containing probe (Fig. 5). The high stability of TDG on a mismatch probe has already been described and is explained by the fact that TDG cleaves the mismatch oligonucleotide in conditions used for EMSA and establishes a very strong interaction with the abasic site produced by the reaction (35). When both proteins were coincubated with the probes, no super-retarded complex appeared, suggesting that TDG may not interact with COUP-TFI when bound to DNA. This was further suggested by the exclusion of TDG from the mismatch probe observed when increasing the amounts of COUP-TFI (Fig. 5). In accordance with Re-ChIP data, these experiments suggest that TDG and COUP-TFI independently bind to DNA, with their respective binding likely exclusive. These data also indicate that a preformed COUP-TFI/TDG complex may be disrupted by DNA binding of one or both partners. Then it could be hypothesized that such complexes may allow the glycosylase to load and scan COUP-TFI targeted promoters.

Fig. 5.

COUP-TFI Binding to DNA Decreases TDG/DNA Association The interactions between COUP-TFI, TDG, and DNA were analyzed by EMSAs using ³²P-labeled probes corresponding to a 36-bp fragment of the vitronectin promoter containing one binding site for COUP-TFI directly adjacent to a CpG dinucleotide (see Materials and Methods). The Mismatch oligonucleotide contains a T:G mismatch, the 5MeC oligonucleotide is hemimethylated on the CpG, and the control oligonucleotide contains neither methylated CpG nor mismatch. Probes were incubated with 1.25, 2.5, 5, or 10 ng of purified recombinant COUP-TFI and 500 ng of purified recombinant TDG. Protein-DNA complexes and free oligonucleotides were separated on a 6% polyacrylamide gel. The asterisk and the arrow indicate the positions of the probe bound to COUP-TFI and TDG, respectively.

Dnmt3 Enhance COUP-TFI Activation of a Methylated Reporter Gene

To examine the functional significance of these various complexes, we next run glycosylase assays as well as cotransfection experiments. In glycosylase assays, TDG, either alone or combined with p68, was unable to cleave DNA at MeCs, suggesting the requirement for additional factors (data not shown). These could be either RNA cofactors (16), or proteins that would deaminate 5MeC in the context of CpGs, providing T:G mismatches to TDG for repair. Importantly, Dnmts being able to catalyze deamination of CpGs (36–39), their recruitment to methylated promoters during transcriptional activation could participate to active demethylation processes. This hypothesis was further substantiated in another experimental system, in which Dnmt3a exhibited cytosine and 5MeC deamination ability (Métivier, R., R. Gallais, C. Tiffoche, C. Le Péron, R. Jurkowska, R. P. Carmouche, D. Ibberson, P. Barath, F. Demay, G. Reid, V. Benes, A. Jeltsch, F. Gannon, and G. Salbert, manuscript in preparation).

Cotransfection experiments were run in P19 cells with in vitro methylated pVitro-Luc, as well as expression vectors for COUP-TFI, and various Dnmts. We first show that, whereas Flag-Dnmt1 and Myc-Dnmt3b had no effect on transcription of the mock-methylated and methylated reporter genes, Myc-Dnmt3b could specifically enhance COUP-TFI-mediated transcriptional activation of the methylated reporter gene (Fig. 6A). We then compared the ability of Flag-Dnmt3a and 3b, and Myc-Dnmt3b to regulate transcription via COUP-TFI and found that all tagged Dnmts could enhance transcription but that Flag-Dnmt3b and Myc-Dnmt3b were more potent than Flag-Dnmt3a at high dosage (Fig. 6B). Surprisingly, when combined with TDG, Dnmt3b did not coactivate transcription of the methylated reporter gene in the presence of COUP-TFI (Fig. 6C). Interestingly, the glycosylase-defective N140A mutant TDG allowed Dnmt3b coactivation of COUP-TFI, albeit to a lower level than when no glycosylase was added. Conversely, the sumoylation-defective K330 mutant was still able to fully repress COUP-TFI/Dnmt3b-mediated transcription. Finally, in these conditions, nor Dnmt3b nor wt or mutated TDG could modify COUP-TFI-mediated transactivation of the mock-methylated reporter gene (data not shown). Collectively, these transfection experiments suggest that, when targeted by COUP-TFI, Dnmt3 can enhance the transcription of a methylated reporter gene; and that TDG can counteract this effect, either through titration of the methyltransferases, or through a repair activity requiring the glycosylase function, or both.

Fig. 6.

Dnmt3a Stimulates COUP-TFI-Mediated Activation of a Methylated Reporter Gene A and B, P19 cells were transfected with a mock-treated or in vitro-methylated reporter gene containing the proximal promoter of the mouse vitronectin gene coupled to the luciferase coding sequence (note that methylation of the reporter gene resulted in a lower basal activity). Expression vectors for COUP-TFI and Dnmts were cotransfected as indicated (M- and F- mean Myc and Flag, respectively). Amounts of transfected Dnmt expression vectors were 10, 20, or 40 ng/well. C, Similar experiments were run in the additional presence of wt or mutant TDG expression vectors. The N140A mutant is glycosylase defective, whereas the K330A mutant cannot be sumoylated. Graphs show the mean ± sem (n = 6–9 depending on experiments) of data obtained in two to three separate experiments. Asterisks indicate statistical differences between COUP-TFI (dark gray bars) and COUP-TFI + Dnmt3b (light gray bars) conditions, in the presence or absence of wt and mutant TDG (ANOVA Fisher’s test, P < 0.05). Note that in B and C the data are shown as fold induction, whereas in A data are shown as relative luciferase units (RLU).

DISCUSSION

In the present report, we have demonstrated the occurrence of physical and/or functional interactions between the orphan nuclear receptor COUP-TFI, the glycosylase TDG, the p68 RNA helicase, and the DNA methyltransferases Dnmt3a-3b. Importantly, ChIP and Re-ChIP data support the presence of a TDG/Dnmt3a-3b complex on untranscribed chromatin (in P19/pCR3.1 cells). Such a complex may participate in heterochromatin maintenance, being involved in the repair of DNA mismatches generated by the spontaneous deamination of cytosines and methylcytosines in CpGs, and their subsequent (re-)methylation (Fig. 7). Interestingly, TDG was able to suppress the coactivation function of Dnmt3b on COUP-TFI in cotransfection experiments. This antagonistic behavior does not support a coordinated role of TDG, Dnmt3b and COUP-TFI in the transcriptional activation of the vitronectin gene but rather suggests that TDG/Dnmt3b could be required for repair and methylation of transcription-induced mismatches. Indeed, this is supported by our cotransfection data showing that the glycosylase-defective TDG was less effective than wtTDG in suppressing coactivation by Dnmt3b. Accordingly, the COUP-TFI/Dnmt3 coactivation process would involve the generation of mismatches, in a Dnmt3-dependent manner. Such a hypothesis is compatible with the known ability of bacterial Dnmts to catalyze the deamination of cytosines (39).

Fig. 7.

Schematic Representation of the Hypothetic Inactive vs. Active Chromatin States of the Vitronectin Gene Proximal Promoter The inactive vitronectin promoter harbors heterochromatin epigenetic marks like diMeK9 of histone H3 and diMeCpGs. This condensed chromatin is bound by HP1 as well as TDG probably complexed to Dnmt3b and Dnmt3a. This complex may participate in the maintenance of CpG methylation state when spontaneous deamination of methylcytosines occurs. In the active configuration, the vitronectin promoter is loaded with RNA Pol II and transcription regulatory complexes such as COUP-TFI/p68/Dnmt3 or TDG/p68/Dnmt3. These two complexes are not simultaneously present on chromatin. PIC, Preinitiation complex.

Although they are also recruited on COUP-TFI-activated vitronectin gene promoter, these factors are included within exclusive TDG/Dnmt3a-3b and COUP-TFI/Dnmt3a-3b complexes that may also contain p68, as indicated by our Re-ChIP assays. Given that our Re-ChIP experiments used unsynchronized cells stochastically transcribing the vitronectin gene to different extents, the recruitment of the TDG/Dnmt3a-3b complex may concern only a subpopulation of COUP-TFI/HA-expressing cells treated with 8-Br-cAMP. This subpopulation would not, at the time of the experiment, express the vitronectin gene which promoter would therefore include the TDG/Dnmt3a-3b complex. Accordingly, the TDG/Dnmt3b complex may favor heterochromatin formation, with the COUP-TFI/Dnmt3b complex conversely favoring euchromatin formation. However, our Re-ChIP experiments suggest that TDG could be complexed to Dnmt3a on active chromatin, whereas TDG/Dnmt3b association with chromatin does not seem to depend on transcription. This may add further complexity to the model presented within Fig. 7 because this observation may imply that Dnmt3a/TDG or Dnmt3b/TDG could coexist in different complexes having divergent outputs. Additionally, the exclusive recruitment of the TDG/Dnmt3a-3b and COUP-TFI/Dnmt3a-3b complexes may suggest that COUP-TFI could serve as a platform for the assembly or recruitment of transcription regulatory complexes that would be released once formed. Transcription of the vitronectin gene may thus be associated with the occurrence of transient and exclusive states leading to timely separated populations of chromatin-bound complexes in single cells. As described in another model of transcription initiation, a transcriptional ratchet may control the occurrence of this sequence of events (40). In the case depicted in this report, different functions of the TDG/Dnmt3 complexes could be regulated depending on protein partners and/or posttranslational modifications that would occur during transcription. Importantly, the presence of the p68 RNA helicase in the TDG/Dnmt3 complex when associated with active chromatin would be an important step toward the acquisition of new functional properties of the complex. Finally, whatever the mechanism involved is, which would require further attention, a true antagonistic behavior of the above-described complexes would nevertheless largely depend on their local respective concentrations at the vitronectin gene locus.

Correlated with the recruitment of these complexes, CpGs from the vitronectin promoter become hypomethylated. Because the above-mentioned proteins have been suggested to directly or indirectly regulate CpG methylation, their association with the vitronectin promoter is likely to regulate its methylation status. TDG is supposed to be part of an active demethylation complex recruited through a direct interaction with ERα (25) and suggested to be required for the hormonal control of the transcription of a methylated template (28). Thus, TDG, through its glycosylase activity, could be involved in an active demethylation process of COUP-TFI target genes such as vitronectin. Nonetheless, we provide evidence that TDG is unable to initiate 5MeCpG demethylation through a BER-type mechanism, at least in vitro. This directly contrasts with the view that TDG can cleave 5MeC within CpGs and suggests that the initiation of demethylation is likely to require additional components of a putative demethylation complex. Such additional components could be specific 5MeC deaminases. Indeed deamination of cytosine or 5MeC leads to U:G or T:G mismatches respectively, both being corrected by TDG. Being expressed in embryonic stem cells and able to deaminate 5MeC and C, it has been proposed that cytidine deaminases such as activation-induced deaminase and apolipoprotein B editing complex 1, could play a significant role in DNA demethylation by generating mismatches (41, 42). Although such a scenario is plausible, it would, however, provide an unspecific mechanism leading to the deamination and repair of both methylated and unmethylated cytosines, including non-CpG cytosines. More specific and costless mechanisms targeting methylcytosine in CpGs are then likely to have evolved, but the enzymes involved in these processes have not yet been characterized.

An appealing hypothesis is that the same systems that regulate DNA methylation at CpGs would also be involved in active demethylation. In support of that hypothesis, bacterial DNA methyltransferases are known to actively deaminate cytosine and 5MeC in experimental conditions where the methyl donor concentration is low (39, 43–45). Even if such a reaction is not favored in vitro in the case of a purified mouse Dnmt1 (46), it could however be triggered by the interaction of Dnmts with specific factors regulating transcription, and would allow the selective deamination of cytosines in the context of CpGs. Such deaminated cytosines would then be repaired through regular BER reactions. Our data support this hypothesis because 1) ChIP assays revealed that Dnmt3a and Dnmt3b are associated with the activated vitronectin gene; 2) TDG and Dnmt3a-3b are found together on chromatin, in support of a coordinated role for Dnmt3 and TDG in the control of DNA (de)methylation; and 3) Dnmt3a is able to catalyze deamination of cytosine and 5MeC in vitro (Métivier, R., R. Gallais, C. Tiffoche, C. Le Péron, R. Jurkowska, R. P. Carmouche, D. Ibberson, P. Barath, F. Demay, G. Reid, V. Benes, A. Jeltsch, F. Gannon, and G. Salbert, manuscript in preparation). Furthermore, Li et al. (47) recently demonstrated that TDG can inhibit Dnmt3a methylation activity. Then, it can be proposed that, depending on regulatory mechanisms such as posttranslational modifications or through protein/protein interactions, Dnmt3 could alternate between active and inactive states, with the active state leading to cytosine methylation and the inactive state to cytosine deamination. This would provide sufficient flexibility for the occurrence of a dynamic CpG methylation profile. This hypothesis will need further attention to formally demonstrate the involvement of Dnmt3a in 5MeC deamination in vivo. Such a demonstration will, however, be hampered by the lack of appropriate tools, and will certainly require innovative experimental approaches as well as specific drugs targeting Dnmts without being incorporated into DNA.

Alternatively, rather than achieving cytosine deamination, the recruitment of Dnmt3a to the active vitronectin promoter may serve to maintain an asymmetric methylation profile that could be a mark of transiently active chromatin. Indeed, our data suggest a strand-biased methylation profile of the active vitronectin promoter. This could reflect either an asymmetric demethylation or an asymmetric re-methylation of DNA. Strand-biased methylation has been described in a few cases such as centromeric DNA in Arabidopsis (48), the active Dc8 gene in carrot (49), and the embryonic ρ-globin gene in chicken (50). In the latter, the unmethylated active ρ-globin gene becomes fully methylated during erythroid cell differentiation, with the coding strand being targeted first by methyltransferases. Similarly, we suggest here that the vitronectin promoter is the target of a strand-specific methylation mechanism during transcriptional activation. Interestingly, activity of Dnmt3a exhibits sequence specificity and strand asymmetry (51). However, examination of the vitronectin gene promoter sequence did not allow to predict that one particular strand would be methylated by Dnmt3a, suggesting that sequence-independent requirements may exist. It would be interesting to determine how stable is the asymmetric DNA methylation of the vitronectin promoter, and to examine the methylation status of other differentially regulated genes, to understand whether strand-specific methylation could define a class of active genes or is rather a transient mark of remodeling chromatin.

MATERIALS AND METHODS

Plasmids

PCR3.1 was from Invitrogen (Carlsbad, CA). pSG5 from Stratagene (La Jolla, CA), pSG5-hTDGwt and hTDG N140A and K330A mutants were from P. Schär (University of Basel, Basel, Switzerland). The pSG5-mTDG (aa 1–421) vector was a gift from Prof. P. Chambon (Institut de Génétique et de Biologie Moléculaire et Cellulaire, Strasbourg, France). An EcoRI-BamHI fragment of the pSG5-mTDG vector was subcloned in pBluescript II KS (pBS). Then the EcoRI-NotI fragment of pBS-mTDG was cloned in pGEX-6P1 (Amersham Biosciences, Arlington Heights, IL) and the BamHI-BamHI fragment of the resulting vector was subcloned in pXJ141-Flag (52). pVitro-Luc, pCDNA3 hCOUP-TFI and pGEX expression vectors encoding GST fused to hCOUP-TFI 57–423, and hCOUP-TFI 153–423 have already been described (33). pCDNA3-based expression vector for HA::COUP-TFI 57–153, HA::COUP-TFI 57–317, HA::COUP-TFI 84–423, and pCR3.1-HA-COUP-TFI 1–423 have been described (53). SRC-1 and p68 expression vectors were gift from B. W. O’Malley (Baylor College of Medicine, Houston, TX) and S. Kato (University of Tokyo, Tokyo, Japan) respectively. The Dnmt1 (54) and Myc-Dnmt3b (55) vectors were obtained from S. Baylin (The Johns Hopkins University, Baltimore, MD) and E. Li (Novartis, Cambridge, MA), whereas Flag-Dnmt3a, and Flag-Dnmt3b (56) were gift from H. Leonhardt (Ludwig Maximilians University Munich, Munich, Germany).

Antibodies

Antibodies were purchased from Santa Cruz [HA-probe (Y11), αCOUP-TFI (N-19), αDnmt1 (H-300), αDnmt2 (H-271), αDnmt3a (H-295), αDnmt3b (H-230)], Sigma (St. Louis, MO) (anti-Flag M2 monoclonal antibody, anti-Flag M2 monoclonal antibody conjugated to agarose beads), Upstate (Lake Placid, NY) [anti-dimethyl-histone H3 (Lys9), αphospho-RNA polymerase II (CTD4H8, lot no. 20463)], Euromedex (Souffelmeyershaim, France) [αHP-1α (2HP-1H5)]. The anti-TDG was prepared as described below and αp68 is a gift from Dr. F. Fuller-Pace (University of Dundee, Dundee, Scotland, UK).

Cell Culture and Transfections

P19 embryonal carcinoma cells and COS-7 cells were grown as monolayers in DMEM (Invitrogen) supplemented with 5% fetal calf serum (Invitrogen). For transient transfection assays, P19 cells were plated in 24-well plates and were transfected using FuGENE 6 transfection reagent (Roche, Indianapolis, IN). Typically, 150 ng of luciferase-based reporter plasmid containing the proximal promoter of the mouse vitronectin gene (pVitro-Luc), 100 ng of β-galactosidase-based control plasmid, and 10–40 ng of CMV or simian virus 40 promoter driven expression vectors were used. The amount of CMV and simian virus 40 promoters in each transfection mix was kept constant with empty pCR3.1 and pSG5 vectors. Cells were assayed for luciferase and β-galactosidase activity 36 h after transfection. In vitro methylation of plasmid DNA, used 10 μg of pVitro-Luc plasmid incubated with SssI methyltransferase (New England Biolabs, Boston, MA) in a total volume of 40 μl at 37 C for 3 h with 160 μm S-adenosylmethionine. The procedure for mock methylation was identical but no enzyme was included. For stable transfection, 5.10⁵ P19 cells were plated in 100-mm dishes and transfected with 10 μg of expression vector. After 24 h, cells were grown in the presence of 400 μg/ml of G418 (Sigma). After 10 d of selection, clones of resistant cells were isolated and propagated in the presence of G418.

Semiquantitative RT-PCR Analysis

Total RNAs from P19 cells were purified using the Trizol reagent (Life Technologies, Inc.). Two micrograms of RNA were used as template for Moloney murine leukemia virus reverse transcriptase (Invitrogen) and Pd(N)6 random hexamers (Amersham Biosciences). For PCR, 1/20 of the reverse transcription reaction was used with 1 μm of the following primers (Proligo, Evry, France): AAGCACTACGGCCAATTCAC (COUP-TFI up), AGCTCGCAGATGTTCTCGAT (COUP-TFI down) [28 cycles]; CAGCTCTGGAGAAACTGCTG (PO up), GTGTACTCAGTCTCCACAGA (PO down) [27 cycles]; TACTATCAGAGCTGCTG-39 (Vitronectin up), AGTTGATGCGAGTGAAG (Vitronectin down) [33 cycles].

GST Pull-Down Assays

Radiolabeled proteins were synthesized in vitro using the T7 RNA polymerase rabbit reticulocyte-coupled transcription/translation kit (Promega Corp., Madison, WI). GST fusion proteins were produced in Escherichia coli BL21-CodonPlus RP (Stratagene). After 12 h of induction with 0.5 mm isopropyl β-d-thiogalactoside, cells were harvested by centrifugation, resuspended in PBS/1% Triton (TPBSI) containing a protease inhibitor cocktail (Roche) and then lysed by sonication. Lysates were clarified by centrifugation and incubated overnight with a 50% suspension of glutathione-agarose beads (Sigma) in PBS. Washed fusion proteins bound to the beads were resuspended in TPBSI and incubated for 2 h with 5 μl of [³⁵S]methionine-labeled proteins in a total volume of 200 μl. Beads were washed ten times in 500 μl of TPBSI supplemented with 100 mm NaCl. Washed beads were denatured at 95 C for 10 min, and the supernatant was resolved on SDS-PAGE. For in vitro phosphorylation of the GST or GST::COUP-TFI fusion proteins, 25 ng of recombinant Erk2 (Sigma) or PKC purified from rat brain (Sigma) were incubated in the presence of 100 to 200 ng of GST fusion proteins bound to glutathione-agarose beads in 50 μl of MAPK phosphorylation buffer [50 mm Tris-HCl (pH 8), 0.5 mm EDTA, 25 mm MgCl₂, 1 mm dithiothreitol (DTT), 250 μm ATP, 2% glycerol] or 50 μl of PKC phosphorylation buffer [20 mm HEPES (pH 7.4) 10 mm MgCl₂, 1.7 mm CaCl₂, 600 μg/ml phosphatidylserine (Sigma); 1 mm DTT, and 250 μm ATP] for 30 min at 30 C, in the presence of protease inhibitors (Roche) and phosphatase inhibitors (1 μm okadaic acid, 200 μm Na₃VO₄; Sigma). Effective phosphorylation was verified by running parallel reactions with 50 μm cold ATP and 5 μCi of [γ-³²P]ATP (MP Biomedicals, Irvine, CA). Beads were then extensively washed before adding radiolabeled proteins for pull-down assays.

Expression and Purification of hCOUP-TFI and mTDG

All purification procedures were performed using ÁktaExplorer 10 and chromatographic material supplied by Amersham Biosciences. COUP-TFI (aa 57–423) was expressed as a GST fusion protein from pGEX-6P1 vector in E. coli strain BL21-CodonPlus-RP (Stratagene). After induction by 0.1 mm isopropyl β-d-thiogalactoside for 4 h at 25 C, bacteria were harvested by centrifugation, washed in water, and the pellets were resuspended in lysis buffer [50 mm Tris-Cl (pH 7.5), 200 mm NaCl, 5 mm MgCl₂, 5 mm DTT, 5 mm 3-[3-(cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS), 5% glycerol] containing deoxyribonuclease I (10 U/ml) and a mix of protease inhibitors (Complete-EDTA; Roche). Bacteria were disrupted by French press and the lysate was clarified by centrifugation (30 min; 25,000 × g). Recombinant COUP-TFI was purified on Heparine Sepharose using PBS with 700 mm NaCl supplement for elution. Eluate was applied to GSTrap 5-ml column and after extensive wash by PBS, diluted PreScission Protease (10 U/ml in PBS supplemented with 2 mm CHAPS and 1 mm DTT) was injected to the column. After removing the GST tag by overnight on-column cleavage at 4 C, target protein was eluted and applied on Resource S 1-ml column. COUP-TFI was eluted by salt gradient with peak at PBS with 1 mm CHAPS containing 100 mm NaCl. Bacterial expression of mTDG (aa 1–421) and preparation of the lysate were performed using the same protocol as for COUP-TFI excluding CHAPS. Clarified lysate was applied to Heparine Sepharose FF 26/10 column and recombinant protein was eluted with PBS supplemented with 700 mm NaCl. GST::TDG fusion protein was further retained on GSTprep 16/10 column and after extensive washings by PBS, diluted Prescission Protease [10 U/ml in 20 mm Tris-Cl (pH 8.8), 50 mm NaCl, 1 mm EDTA, 1 mm DTT] was injected to the column. After the removal of the GST tag by overnight on-column cleavage at 4 C, the protein sample was then eluted and applied on Resource Q 6ml column and resolved by the salt gradient (100–200 mm NaCl) in 20 mm Tris-Cl (pH 8.5). Fractions containing TDG were concentrated and separated by gel filtration (Superdex 200 HR 16/60; GE Healthcare, Orsay, France) in 20 mm Tris-Cl (pH 8.0), 150 mm NaCl, 1 mm EDTA, 1 mm DTT. The purified protein was used for immunization of rabbits. The collected polyclonal antibody was affinity purified before use in ChIP assays.

EMSAs

The following oligonucleotides NR2 Up 5′-ATCCCTCCTGCCCGTGACCCTATGACTTTGGTCTTG-3′, MeC Up 5′-ATCCCTCCTGCC^5MeCGTGACCCTATGACTTTGGTCTTG-3′, and Mismatch Up 5′-ATCCCTCCTGCCTGTGACCCTATGACTTTGGTCTTG-3′, were 5′-labeled with T4 polynucleotide kinase and [γ-³²P]ATP before annealing with the complementary strand NR2 Down 5′-CAAGACCAAAGTCATAGGGTCACGGGCAGGAGGGAT-3′. Purified proteins (1.25, 2.5, 5, or 10 ng COUP-TFI, 250 or 500 ng TDG) were preincubated 15 min at room temperature with 0.5 mg/ml BSA and 5 mm dithiothreitol. After addition of 0.05 mg/ml poly(deoxyinosine-deoxycytosine) and binding buffer [10 mm HEPES (pH 7.8), 50 mm KCl, 1 mm EDTA, 5 mm MgCl₂, 10% glycerol], samples were incubated for 15 min. The samples were then incubated with the ³²P-labeled DNA duplexes (20,000 cpm) for 20 min. Protein-DNA complexes were separated on a native 5% polyacrylamide gel in 0.5× TBE [1× TBE: 89 mm Tris, 89 mm borate, 2 mm EDTA (pH 8)] at 4 C.

ChIPs

pCR3.1 and COUP-TFI/HA P19 Cells (10⁶) were plated in 100-mm dishes. Twenty-four hours later, cells were treated with 1 mm 8-Bc-cAMP (Sigma) for the next 24 h. Chromatin was cross-linked using 1% formaldehyde during 10 min at room temperature, and the cells were collected after two washings with PBS in 1-ml cell collection buffer [100 mm Tris-HCl (pH 9.4), 100 mm DTT]. ChIP were then performed with minor modifications of the procedure described by Métivier et al. (40). Precipitates were serially washed with 300 μl of Washing Buffer I [2 mm EDTA, 20 mm Tris-HCl (pH 8.1), 0.1% sodium dodecyl sulfate (SDS), 1% Triton X-100, 150 mm NaCl], Washing Buffer II [2 mm EDTA, 20 mm Tris-HCl (pH 8.1), 0.1% SDS, 1% Triton X-100, 500 mm NaCl], Washing Buffer III [1 mm EDTA, 10 mm Tris-HCl (pH 8.1), 1% NP-40, 1% Deoxycholate, 250 mm LiCl] and then twice with 1 mm EDTA, 10 mm Tris-HCl (pH 8.1). Precipitated chromatin complexes were removed from the beads through three successive 10-min incubations of the beads with 50 μl of 1% SDS, 0.1 m NaHCO₃. Cross-linking was reversed by an overnight incubation at 65 C. DNA was purified with QIAquick columns (QIAGEN, Hilden, Germany). Re-ChIP assays were run as described (40). The primers used to amplify the vitronectin and PO promoters were agggagattaacgggactgg (pVitro Up2), attgccctcctccaaacac (pVitro Dw2), CGCAGAAAGTTGTTTTGCTG (pPO Up), and CTGGTTCCATCGACTGTCCT (pPO Dw).

Immunoprecipitation and Western Blot Analysis

The 10⁶ COS-7 cells or 5.10⁵ P19 cells were plated in 100-mm dishes, and were transfected 24 h later with 10 μg of expression vectors. Cells were collected 36 h after transfection. When pCR3.1 and COUP-TFI/HA P19 cells were used, 10⁶ were plated and treated 24 h later with 1 mm 8-Br-cAMP for the next 24 h. Whole-cell lysates of transiently transfected COS-7 cells were prepared in lysis buffer [20 mm Tris (pH 7.9), 120 mm MgCl₂, 0.2% Nonidet P-40, 5 mm EDTA, 10% glycerol] containing a protease inhibitor cocktail (Roche). The soluble fraction was immunoprecipitated with anti-Flag M2 monoclonal antibody conjugated to agarose beads (Sigma) at 4 C for 4 h in 500 μl of lysis buffer. Immunocomplexes were washed seven times in lysis buffer and analyzed by Western blotting. P19 cells extracts were prepared in lysis buffer with 0.5% Nonidet P-40 and 0.5% Empigen BB (Sigma) and sonicated by 10 pulses at 50% duty and power 3 (Sonifer Cell Disruptor, Branson, Danbury, CT). After centrifugation, the supernatants were diluted with 1 Volume IP buffer used for ChIP assay and subjected to immunoprecipitation with 1 μg antibody overnight after a 3 h preclearing at 4 C with 50 μl of a 50% protein A-Sepharose beads slurry (Amersham Biosciences). Complexes were recovered after 3 h incubation at 4 C with 50 μl of protein A Sepharose. Precipitates were serially washed once with Washing buffer I and twice with Washing Buffer II (described above) and analyzed by Western blotting. Antibodies were diluted 1:1000 in 5% of fat-free milk in PBS, 0.1% Tween 20.

Bisulfite Sequencing

Genomic DNA (40 μg) was denatured by NaOH (0.5 m final) in a volume of 100 μl, for 15 min at 37 C. Ninety microliters of 10 mm hydroquinone and 1.4 ml of freshly prepared sodium bisulphite solution [7.6 g Na metabisulphite in 20 ml of 6.2 m Urea, 120 mm NaOH (pH 5)] were added to the denatured DNA. The sample was gently mixed and incubated at 50 C for 16 h. Bisulfite-modified DNA was then purified using Geneclean III kit (Q-Biogene, France), according to the manufacturer’s instructions. Modification was completed by NaOH (0.3 m final) treatment for 15 min at room temperature. The solution was then neutralized by addition of NH₄OAc to 3 m final and the DNA was ethanol precipitated. DNA was resuspended in 5 mm Tris (pH 8.1) and used immediately for strand-specific PCR amplification, cloning and sequencing. The transcribed strand was amplified (39 cycles) with the following primers: AAAATACTTACCAACTCTCC (up) and GAATTTTGTTGTTTTGTTTGAAAA (down). The nontranscribed strand was amplified (39 cycles) with the following primers: GAGAGAATAAGGAGAGGTTAG (up) and CTAACCCAACCCCATTACCC (down2), after 39 cycles with the oligonucleotide CTTCTAAAATCTCACTATCCTACC (down 1).

Acknowledgments

We are grateful to Prof. P. Chambon (Institut de Génétique et de Biologie Moléculaire et Cellulaire, Strasbourg, France), S. Baylin (The Johns Hopkins University, Baltimore, MD), H. Leonhardt (Ludwig Maximilians University Munich, Munich, Germany), E. Li (Novartis, Cambridge, MA), B. W. O’Malley (Baylor College of Medicine, Houston, TX), and S. Kato (University of Tokyo, Tokyo, Japan), for the generous gift of plasmids; and to Dr. F. Fuller-Pace (University of Dundee, Dundee, Scotland, UK) for the kind gift of the p68 antibody. We thank P. Schär (University of Basel, Basel, Switzerland) for giving plasmids and for critical reading of the manuscript. We also thank C. Martin (University of Rennes), A. Barbelivien (University of Rennes), F. Percevault (University of Rennes), C. Ralliere (Institut National de la Recherche Agronomique), and C. Tascon (Centre National de la Recherche Scientifique) for technical assistance.

This work was supported by funds from the Ministère de l’Education Nationale de l’Enseignement Supérieur et de la Recherche (MENESR) and from the Centre National de la Recherche Scientifique. R.G. was supported by a grant from the MENESR and a fellowship from the Ligue Nationale Contre le Cancer.

Disclosure Statement: The authors have nothing to disclose.

Abbreviations

 
• aa

  Amino acids;

 
• AP

  apurinic/apyrimidic;

 
• BER

  base excision repair;

 
• 8-Br-cAMP

  8-bromo-cAMP;

 
• ChIP

  Chromatin immunoprecipitation;

 
• CMV

  cytomegalovirus;

 
• CpG

  cytosine-guanine dinucleotide;

 
• Dnmt

  DNA methyl transferases;

 
• DTT

  dithiothreitol;

 
• ER

  estrogen receptor;

 
• GST

  glutathione S-transferase;

 
• HA

  hemagglutinin;

 
• PKC

  protein kinase C;

 
• Re-ChIP

  sequential ChIP;

 
• SDS

  sodium dodecyl sulfate;

 
• TDG

  thymine DNA glycosylase;

 
• TPBSI

  PBS/1% Triton;

 
• wt

  wild type.