Methyl-CpG Binding Domain Proteins and Their Involvement in the Regulation of the MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 Gene Promoters

Promoter hypermethylation is responsible for the restricted expression of the tumor-associated MAGE antigens. In order to elucidate the mechanism underlying methylation-dependent repression, we examined the involvement of methyl-CpG binding proteins, MBD1, MBD2a, and MeCP2, in silencing of MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 genes. Electrophoretic mobility shift assays displayed binding of MBD1 to the methylated and unmethylated MAGE-A promoters. Using chromatin immunoprecipitation assays, in vivo binding of MBD1 and MeCP2 to the promoters could be observed in MCF-7 and T47D cells. Transient transfection assays of MCF-7 cells were done with the transcriptional repression domains (TRD) of MBD1, MBD2a, and MeCP2, and MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 promoters. Whereas the TRD of MBD1 and MeCP2 repressed the MAGE-A promoters, the TRD of MBD2 had no inhibiting effect on the promoter activity. Furthermore, cotransfections of Mbd1-deficient mouse fibroblasts and MCF-7 cells with MBD2a, MeCP2, and the MBD1 splice variants, 1v1 and 1v3, showed that strong methylation-dependent repression of the MAGE-A promoters could not be further down-regulated by these proteins. However, the two MBD1 splice variants, 1v1 and 1v3, were able to repress the basal activity of unmethylated MAGE-A promoters. Additional cotransfection experiments with both isoforms of MBD1 and the transcription factor Ets-1 showed that Ets-1 could not abrogate the MBD1-mediated suppression. In contrast with the repressive effect mediated by MBD1, MBD2a was found to up-regulate the basal activity of the promoters. In conclusion, these data show, for the first time, the involvement of methyl-CpG binding domain proteins in the regulation of the MAGE-A genes. (Mol Cancer Res 2007;5(7):749–59)


Introduction
The MAGE-A gene family is located on chromosome Xq28 (1) and is comprised of 12 members (MAGE-A1 -12). They encode tumor-associated antigens which are recognized by CTLs in conjunction with MHC class I molecules of various haplotypes on the tumor cell surface (2). Their expression is restricted to different histologic tumors types including melanoma, lung, mamma, bladder, and stomach carcinoma (3)(4)(5)(6)(7)(8). In normal adult tissues, they are silent, except for testicular germ cells (spermatogonia and primary spermatocytes) and placenta (9). Hypermethylation of CpG-rich MAGE-A promoters causes gene silencing and prevents Ets transcription factors from accessing their binding sites (1,10,11), which are responsible for the high transcriptional activation of the MAGE-A1 gene (12).
Alterations in DNA methylation occur during the pathogenesis of human tumors, and a global DNA hypomethylation of the genome, including the tumor-associated MAGE-A genes, has been observed in various carcinomas. Recent clinical studies showed that immunotherapy against MAGE-A antigens caused a significant tumor regression in patients with various tumor entities, pointing out that vaccination with a MAGE-A peptide could be a promising target for the treatment of patients with tumors (13)(14)(15)(16). Nevertheless, heterogeneous intratumor expression of the MAGE-A genes may hamper the effectiveness of immunotherapy and urgently requires a detailed examination of the regulation of these genes in tumors. The mechanism leading to partial DNA hypomethylation of the genes during carcinogenesis remains poorly understood. In our previous article, we addressed this issue and found that besides DNA methylation, histone deacetylation, which leads to a compact and transcriptional inactive chromatin structure (17), is also involved in the partial repression of MAGE-A genes in tumor cells and may impede their activation (18). Whether methyl-CpG binding proteins, which recruit histone deacetylases to methylated DNA, contribute to the silencing of MAGE-A genes has not yet been clarified.
To date, five methyl-CpG binding proteins have been identified: MBD1, MBD2, MBD3, MBD4, and MeCP2. All of these proteins are involved in the transcriptional repression of methylated DNA (19,20). The methyl-CpG binding domain protein 2 (MeCP2) can interact with both the corepressor SinA 3 and histone deacetylases (21)(22)(23). It can also repress the transcription of distinct promoters independently of histone deacetylases, probably through direct interaction with basal transcription factors (24). MeCP2-deficient mice show neuronal defects, which likely lead to their short survival time of 2 to 3 months (25). The other members of the family (MBD1, MBD2, and MBD3) also associate with histone deacetylases. MBD1 is alternatively spliced to produce five protein isoforms (PCM1, MBD1v1, MBD1v2, MBD1v3, and MBD1v4) which differ in the number of cysteine-rich (CXXC) domains and the carboxyl-terminal sequence. All five MBD1 isoforms repress transcription from methylated promoters and the variants with three CXXC domains additionally repress unmethylated promoters (26,27). Two isoforms of MBD2 are known: MBD2a and MBD2b. The shorter form, MBD2b, starting at the second methionine therefore lacks the NH 2 -terminal sequence of MBD2a (28). Recent reports described MBD2a as both an activator and a repressor of transcription (29)(30)(31). Mbd1-and Mbd2-deficient mice are viable and fertile. However, the Mbd1 À/À mouse shows mild defects in neurogenesis (32), 3 and the Mbd2 À/À mouse in nurturing their offspring (33).
Based on our findings that the association of DNA methylation and histone deacetylation may be responsible for inactivation of the MAGE-A genes (18), we have continued to examine whether methyl-CpG binding proteins participate in silencing these genes. Our promoter studies and transient transfection assays show that both splice variants MBD1v1 and 1v3 were able to repress the unmethylated MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 promoters, and the transactivator Ets-1 could not abrogate the MBD1-mediated gene repression. Conversely, MBD2a could enhance the promoter activity of these genes.

Results
In vitro Binding of MBD1 to the 5 ¶-Regions of MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 Genes In our previous studies, we reported the association of DNA methylation and histone deacetylation in silencing of MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 genes (18). Because methyl-CpG binding proteins can recruit histone deacetylases to methylated DNA, we investigated whether these proteins are able to bind to hypermethylated MAGE-A promoters. For these analyses, we applied two techniques: the in vitro electrophoretic mobility shift assay (EMSA) and in vivo chromatin immunoprecipitation (ChIP) assay.
EMSA was done using the regions extending from À77 to À36 of MAGE-A1, À158 to À83 of MAGE-A2, À160 to À102 of MAGE-A3, and À109 to À63 of MAGE-A12 genes (Fig. 1A). We chose these sequences because they contained Ets binding motifs which have been shown to be important for the MAGE-A1 promoter activity (10). There were four, six, two, and five CpG dinucleotides in the MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 promoter fragments, respectively ( Fig. 1A). In vitro DNA methylation of the fragments was carried out by DNA methylases SssI and HpaII, which methylate each cytosine residue of the CpG dinucleotides and the second cytosine of the sequence CCGG, respectively. For EMSA, we used nuclear extracts derived from three breast cancer cells (MCF-7, T47D, and MDA-MB-231) as well as the epithelioid cervical carcinoma cell line HeLa. Nuclear extracts from HeLa cells were used because methyl-CpG binding proteins were recently found to be components of HeLa cells and served as control for the binding analyses (30,34). When the unmethylated, HpaIIand SssI-methylated probes were incubated with various nuclear extracts, prominent complexes were formed as detected by EMSA. Figure 1B  Binding was competed by the addition of an excess of cognate, unlabeled oligonucleotides which were either unmethylated or methylated indicating that the complexes were specific ( Fig. 1B and C, lanes 5-10). Furthermore, the complexes competed by SssI-methylated oligonucleotides had a weaker intensity ( Fig. 1B and C, lanes 8-10) than those competed by unmethylated oligonucleotides (Fig. 1B and C, lanes 5-7), suggesting that the binding affinity of nuclear extracts was higher to methylated than unmethylated DNA.
To characterize the nature of the complexes, we did supershift experiments using antibodies that specifically react with different methyl-CpG binding domain proteins. When added to the EMSA reaction, an antibody specific for MBD1 produced a supershift (SS, Fig. 2), whereas antibodies specific for MBD2a and MeCP2 had no effect on the binding reaction (data not shown). Our findings show that MBD1 had a binding affinity to both methylated and unmethylated MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 promoters (Fig. 2). Figure 2A shows such a supershift caused by MBD1 using increasing amounts of nuclear extract from MCF-7 (lanes 4 and 5), MDA-MB-231 (lanes 8 and 9), and T47D (lanes 12 and 13) breast cancer cells and the unmethylated MAGE-A1 probe. As shown in Fig. 2B and C, an antibody specific for MBD1 generated supershifted complexes containing nuclear extracts derived from MCF-7 (B) or T47D cells (C) and methylated or unmethylated MAGE-A2, MAGE-A3, and MAGE-A12 probes (lanes 5-7).
In vivo Binding of MBD1 and MeCP2 to the 5 ¶-Regions of MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 Genes As EMSA revealed in vitro binding, specifically of MBD1 to methylated and unmethylated MAGE-A promoters, we did ChIP assays to monitor in vivo binding of methyl-CpG binding proteins to the MAGE-A promoters in MCF-7 and T47D cells. Following cross-linking of nuclear proteins to DNA with formaldehyde and immunoprecipitation of the protein-DNA complexes with the antibodies specific for MBD1, MBD2, or MeCP2, the immunoprecipitated DNA fragments were amplified by PCR using gene-specific primers and visualized by gel electrophoresis (Fig. 3A). Compared with DNA precipitated with IgG, DNA precipitated with antibodies against MBD1 and MeCP2 contained significant increases of MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 promoter DNA (Fig. 3A). Densitometric evaluation showed that amplification of the MBD1-and MeCP2-immunoprecipitated DNA resulted in a 6-and 9-fold, 2-and 5-fold, 4-and 3-fold, and 3-and 8-fold increase using primers specific for MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 genes, respectively, in comparison with the PCR products of the IgG-immunoprecipitated DNA. MeCP2 had a higher affinity with the MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 promoters than MBD1; possibly caused by the higher precipitation efficiency of the MeCP2specific antibody (Fig. 3A).
To verify the data of the standard PCR, we additionally carried out a quantitative real-time PCR with primers specific for the MAGE-A1 sequence and the RPLP0 (ribosomal protein, large protein) housekeeping gene. The primer pairs were chosen due to the stringent conditions of the real-time PCR. A low amount of immunoprecipitated RPLP0 DNA could be observed in the bar chart of Fig. 3B based on the characteristics of a housekeeping gene, which is constitutively expressed, whereas its promoter region is unmethylated. In MCF-7 cells which do not express MAGE-A1 (18), the yield of MAGE-A1 DNA immunoprecipitated with antibodies against MBD1 and MeCP2 was 2-and 2.5-fold higher, respectively, than the yield of DNA immunoprecipitated with the antibody against MBD2 or IgG, which served as a negative control (Fig. 3B). When DNA originated from T47D cells, there was no difference in the yield of MAGE-A1 DNA precipitated with antibodies against MBD1, MBD2, and MeCP2, in comparison to the DNA precipitated with IgG ( Fig. 3B), indicating that these cells expressed MAGE-A1 (18). Taken together, these findings show that MBD1, as well as MeCP2, interact in vivo with all four MAGE-A promoter sequences in MCF-7 cells. To functionally test the effect of the methyl-CpG binding proteins on MAGE-A promoters, we constructed a series of reporter plasmids in which we inserted sequences derived from the MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 promoters immediately downstream of the Gal4 DNA-binding sites. The expression plasmids contained the transcriptional repression domains (TRD) of MBD1, MBD2, and MeCP2 linked to the heterologous Gal4 DNA-binding domain (28). Both reporter and expression plasmids were transiently cotransfected with a reference plasmid into MCF-7 cells; cell lysates were then assayed for luciferase activity 48 h posttransfection. Immunoblot analysis using antibodies specific for MBD1, MBD2, and MeCP2 documented efficient expression of the fusion proteins in these cells (data not shown). Targeting of the TRD of MBD1 to Gal4 DNA-binding sites upstream of the MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 promoters produced a 7-, 25-, 25-, and 8-fold decrease in promoterdriven luciferase activity, respectively, when compared with the basal activity of Gal4-MAGE-A constructs (Fig. 4). Although the Gal4-linked TRD of MeCP2 had a weaker effect, it was able to repress the MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 transcription by 2-, 4-, 5-and 2-fold, respectively (Fig. 4). These data show that the TRDs of MBD1, and to a lesser extent, MeCP2, but not that of MBD2, are able to inhibit each of the tested MAGE-A promoter activities.

Transcriptional Repression of Unmethylated MAGE-A Genes by the Splice Variants of MBD1, and Activation by MBD2a
Our recent studies on various cell lines indicated that promoter hypermethylation and histone deacetylation are associated with silencing of MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 genes (18). Furthermore, our present findings suggest that MBD1 and MeCP2 may be able to bind to and repress MAGE-A promoters (Figs. [1][2][3][4]. To more precisely define the role of the methyl-CpG binding proteins in MAGE-A gene repression, we transiently cotransfected fulllength MBD1, MBD2a, and MeCP2 together with methylated and unmethylated MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 reporter constructs into MCF-7 cells, and assessed their influence on transcription in response to methylation by SssI and HpaII. We expanded our transfection assay with embryonic fibroblast cells from a Mbd1-knockout mouse 3 to exclude interference from endogenous MBD1. The Mbd1-knockout mouse was analyzed by PCR using genomic DNA from tail biopsies and a primer set gene-specific for the wild-type and knockout allele, as illustrated in the schematic view of Fig. 5A. The primers binding at À695 and À266 bp relative to the start site of MBD1 (ATG) produced a PCR product of 430 bp in size when genomic DNA from wild-type mice were amplified (Fig. 5B, wt lane). No PCR product of 430 bp was generated when DNA originated from homozygous Mbd1-knockout mice (Fig. 5B, Mbd1 À/À lane).
The primer pair at À695 bp and +1383 bp amplified a PCR product of 2078 bp in size when DNA was derived from wildtype mice (data not shown), and a 520-bp-long product when deleted DNA from Mbd1-knockout mice was used (Fig. 5B, Mbd1 À/À lane). Consequently, amplification of DNA from heterozygous mice formed three PCR products at 430, 520, and 2078 bp (Fig. 5B, Mbd1 +/À lane).
The genetic characterization of embryonic mouse fibroblasts from the Mbd1-knockout mouse was carried out by amplification of their genomic DNA using the same primer set. Whereas amplification with the primers specific for the wild-type allele did not lead to a PCR product of 430 bp in size (Fig. 5C, wt lane), amplification with the primers specific for the Mbd1 sequence with the deleted exon resulted in a 520-bp-long PCR product (Fig. 5C, Mbd1 À/À lane). For cotransfection of MCF-7 cells and Mbd1-deficient embryonic fibroblast cells, we used two isoforms and a mutant form of MBD1. Whereas MBD1v1 contains two CXXC domains, MBD1v3 harbors three CXXC domains. The mutant form of MBD1 is functionally inactivated by a 126-bp deletion in the methyl-CpG binding domain and a mutation in the TRD (amino acids 537-556), and served as a negative control. Evaluation of promoter-driven luciferase activities showed that DNA methylation of the reporter plasmids by SssI caused a nearly complete loss of MAGE-A promoter activity in mouse Mbd1 À/À fibroblasts (Fig. 5D). Similar results were obtained when transfection assays were done with SssIand HpaII-methylated constructs in MCF-7 cells, as described in our recent article (18). Due to the strong methylation-dependent repression, cotransfected methyl-CpG binding proteins had no further inhibitory effect on methylated MAGE-A constructs in Mbd1-deficient mouse cells (Fig. 5D) and MCF-7 cells (data not shown). In contrast, the transcriptional activity of the unmethylated MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 promoters in Mbd1-deficient mouse cells was efficiently reduced 3.5-, 2.5-, 3-, and 8.5-fold by isoform MBD1v1 and 1.5-, 3-, 2-, and 1.5-fold by isoform MBD1v3, respectively (Fig. 5D). In MCF-7 cells, MBD1v1 repressed the basal activity of unmethylated MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 promoters by 8-, 6-, 3-, and 17-fold, respectively, whereas MBD1v3 did not affect the promoters (Fig. 6). MBD1v1 with its three CXXC domains usually affected the repression more strongly than MBD1v3 with its two CXXC domains. The mutant version of MBD1 failed to repress the reporter plasmids indicating the specificity of repression by MBD1v1 and MBD1v3 (Figs. 5D and 6). Contrary to the repressive effect by the MBD1 isoforms, MBD2a was able to stimulate f1.5-, 3-, 10-, and 1.5-fold the MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 promoters, respectively, in MCF-7 cells (Fig. 7). A similar stimulation by MBD2a was observed in Mbd1-deficient mouse cells (data not shown). Finally, MeCP2 had no effect on the unmethylated promoters (data not shown).

Ets-1 Did Not Affect MBD1-Mediated Suppression and MBD2-Mediated Activation
The present data suggests that promoter hypermethylation is sufficient for silencing of MAGE-A genes. Furthermore, it shows that MBD1v1 and MBD1v3 are able to bind to the unmethylated MAGE-A promoters and down-regulate their basal activity. The MAGE-A1 promoter region contains two motifs for Ets transcription factors. Promoter demethylation and binding of the Ets factors to their binding sites have been shown to drive 90% of the MAGE-A1 transcription (10). In contrast with the investigated MAGE-A1 promoter, only little is known of the MAGE-A2, MAGE-A3, and MAGE-A12 promoters. However, it is assumed that they are regulated in a similar manner. By DNA sequence analysis and alignments, we identified Ets-1 consensus motifs at positions À103/À90, À143/À134, and À154/À145 of MAGE-A2, at positions À113/À104 and À151/À141 of MAGE-A3, and at position À100/À90 of the MAGE-A12 promoter. To examine the involvement of Ets-1 in MAGE-A2, MAGE-A3, and MAGE-A12 transcription, we transiently transfected MCF-7 cells and embryonic mouse fibroblasts with an expression plasmid encoding Ets-1. Ets-1 could stimulate the promoter activity of MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 in Mbd1-deficient fibroblasts and MCF-7 cells 2-, 4-, 2-, and 2-fold (Fig. 5D) and 2-, 4-, 6-, and 3-fold (Fig. 7), respectively. We conclude that Ets is not only a transactivator for MAGE-A1 but for the other tested promoters as well.
By additional cotransfection experiments, we examined the influence of Ets-1 on MBD1-mediated repression and MBD2amediated activation. As shown in Figs. 5D and 7, Ets-1 could neither alleviate the suppression (Fig. 5D) nor affect the activation (Fig. 7) of the promoters by MBD1 and MBD2a, respectively.

Discussion
The current study shows for the first time the different regulations for unmethylated MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 promoters by methyl-CpG binding proteins. Thus far, these epigenetic factors have been found to be mainly associated with methylated promoters of tumor suppressor genes (32)(33)(34)(35)(36).
Our transfection assays of MCF-7 cells and embryonic mouse Mbd1 À/À fibroblasts show that promoter hypermethylation seems to be sufficient for silencing of the MAGE-A genes. However, considering that mice deficient of the different methyl-CpG binding proteins and the resulting phenotypes (25,32,33), it could be possible that other methyl-CpG binding proteins, like MeCP2, could compensate for the loss of MBD1 in the repression of methylated MAGE-A promoters. Moreover, our binding analyses of EMSA show that MBD1 bound to methylated as well as to unmethylated MAGE-A promoters, and sustains the transfection data that binding of MBD1 to unmethylated DNA may lead to repression of the promoters. The ability of MBD1 to repress unmethylated promoters in vitro is supported by a recently published study showing methylation-independent repression (35). Repression of unmethylated genes depends on the third CXXC domain, and repression of methylated genes requires the methyl-CpG binding domain (26,36). The isoform MBD1v1 contains the additional third CXXC domain and could therefore repress promoters regardless of their methylation status. However, MBD1v3 lacks the third CXXC domain. Although the repression by MBD1v3 was weaker than the suppression by MBD1v1 and only observed in Mbd1-deficient mouse fibroblasts, our findings indicate that the two other CXXC domains may also contribute to the repression of unmethylated promoters, however, with a weaker affinity. The repressive effect on the examined MAGE-A promoters by MBD1v3 could also be ascribed to the fact that endogenous DNA methylases 3a and 3b in the transfected cells might methylate de novo the introduced reporter plasmids containing each of the MAGE-A promoters (37). Furthermore, the specific repression of MAGE-A promoters by MBD1v1 and MBD1v3 was confirmed by the use of mutated MBD1 in our overexpression studies. MBD1mut, which lacks the methyl-CpG binding domain and harbors a nonfunctional TRD, congruously showed no effect on MAGE-A gene expression.
The MAGE-A promoters contain putative Ets motifs and the transcription factor Ets has been shown to be responsible for the high transcriptional activation of MAGE-A1 (10). Our results show that transfection of expression plasmid encoding Ets-1 resulted in the activation of all tested MAGE-A promoters. It is noteworthy that cotransfection of MBD1v1 or MBD1v3 together with Ets-1 did not lead to an abrogation of MBD1mediated repression. Concerning the potential role of MBD1 as a repressor of unmethylated MAGE-A promoters and the mechanisms underlying the transcriptional repression, it is of considerable interest to investigate the relation between MBD1 and Ets-1. It is unclear whether, in the cotransfection assay using both MBD1 and Ets-1, the MBD1-mediated repression depends on basal or Ets-1 -mediated activity and whether MBD1 is able to prevent the binding of Ets-1 to its motif.
Although EMSA and ChIP assays show that MBD2a lacked binding activity with any of the examined MAGE-A promoters, transfected MBD2a could stimulate the luciferase activity of unmethylated reporter plasmids containing MAGE-A promoter fragments. The stimulatory effect of MBD2a on unmethylated promoters was recently reported for cyclic AMP -responsive genes (29). The absence of binding activity of MBD2a to these B. Characterization of the Mbd1-knockout mouse was done by PCR using genomic DNA from tail biopsies of homozygous (Mbd1 À/À ), heterozygous (Mbd1 +/À ) and wild-type mice and gene-specific primers in A. C. Genetic characterization of embryonic mouse fibroblasts from the Mbd1-deficient mouse was done by PCR with genomic DNA from these cells and the same primer set in A. D. Embryonic fibroblasts were transfected with unmethylated (unmeth. ) and methylated (meth. ) reporter plasmids containing fragments of MAGE-A1, MAGE-A2, MAGE-A3, or MAGE-A12 promoters and expression plasmids encoding MBD1mut, MBD1v1, MBD1v3, or the transcription factor Ets-1. The signals derived from the reference plasmid pCMV LacZ were used to normalize the variability in transfection efficiency. The activities are shown relative to the activity of the reporter construct containing the MAGE-A1 sequence which was arbitrarily set to 100%.

Methyl-CpG Binding Proteins and MAGE Genes
Mol Cancer Res 2007;5(7). July 2007 genes and their activation by MBD2a were explained by its interaction with the RNA helicase A and by a hypothetical model illustrating MBD2a as a factor in the transcriptional coactivator complex which is associated with the RNA polymerase II (29). Whether MBD2a forms a link with one of the components of the basal transcription machinery or even interacts with Ets-1 will require further investigation. Moreover, we cannot completely exclude that MBD2a may promote transcription mediated by other transcription factors, like Sp1 which is also involved in MAGE-A1 promoter activation (10).
Our ChIP assays show that both MBD1 and MeCP2 were able to bind to each of the four examined MAGE-A promoters in MCF-7 cells. As we previously reported, MAGE-A promoters are heavily methylated in this cell line (18). The cells have been shown to possess a high maintenance of DNA methyltransferase activity (38). In contrast with our ChIP results, the in vitro data obtained by EMSA show the binding activity of MBD1 to the MAGE-A promoter fragments, but no binding of MeCP2. The reason for this discrepancy seems to be the use of short DNA fragments extending not more than 80 bp 5 ¶ from the start site of the promoters in the EMSA assays. For in vitro studies, longer DNA sequences were amplified in the PCR reaction following the ChIP assays. These findings suggest that MBD1 does not bind to the same site of the promoter as MeCP2 does. Apparently, MBD1 and MeCP2 have different binding preferences to the MAGE-A promoters.
In addition, a putative repressive effect on these promoters was caused by MeCP2 when its TRD was linked with the heterologous Gal4-binding domain. However, the use of fulllength MeCP2 affected neither methylated nor unmethylated MAGE-A promoters. The discrepancy of our results could allude to the availability of endogenous MeCP2 in MCF-7 cells and Mbd1-deficient mouse fibroblasts. Besides its affinity to hypermethylated DNA (36,39,40), the binding of MeCP2 to unmethylated promoters was recently described for the metallothionein I gene, and was explained by its TRD which harbors a DNA binding motif in addition to the NH 2 -terminal methyl-CpG binding domain (35).
Despite the global decrease in DNA methylation, hypermethylated regions leading to transcriptional silencing of tumor suppressor and DNA repair genes have also been detected in malignant cells (41). Loriot et al. proposed a model that could explain how hypomethylation in tumors is maintained despite methylation activities. They reported that during the neoplastic process, cancerous cells undergo a transient phase of epigenetic instability leading to demethylation of the MAGE-A1 gene. In cells containing factors that exert a sufficient level of transcriptional activation to protect the region against remethylation, the promoter of MAGE-A1 will be maintained unmethylated (37). Because MBD1 and MeCP2 can interact with the MAGE-A promoters, it could be interesting to investigate whether they are involved in this process to prevent remethylation by their interaction with this promoter.
In conclusion, our results refer to substantial aspects of the MAGE-A gene repression and reveal why promoter demethylation must not result in the activation of the MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 genes. It is noteworthy that we could show for the first time that binding of MBD1 to the unmethylated MAGE-A promoters lead to gene repression which could not be abrogated by the transcription factor Ets-1. Studies are under way to investigate the interplay between MBD1 and Ets-1 at unmethylated MAGE-A promoters.

Cell Cultures
The human cell lines HeLa derived from an epithelioid cervical carcinoma and the three breast cancer cell lines, MCF-7, MDA-MB-231, and T47D, as well as the primary cell  culture of embryonic fibroblast cells from a Mbd1-deficient mouse were maintained in DMEM (Invitrogen), supplemented with 10% FCS, and 2 mmol/L of L-glutamine (Invitrogen).

Preparation of Primary Mouse Embryonic Fibroblasts from Mbd1-Knockout Mice
For the construction of the Mbd1-deficient mouse, the second exon of Mbd1 containing the first ATG of the translation start site was deleted. A second ATG is located in the third exon in another reading frame, and is directly followed by a stop codon. By introducing the deletion of the second exon, no functional Mbd1 protein can be translated. The generated Mbd1-knockout mice are, however, viable and fertile. 3 The embryonic fibroblasts were isolated from these mice at embryonic day 13. Embryos were dissected, and head and organs were removed. The embryonic tissues were washed with PBS and mechanically digested following incubation for 2 h on ice with trypsin/EDTA (Invitrogen). The genotypes of the Mbd1-knockout mouse and primary embryonic fibroblasts were verified by PCR using the following forward primer and two reverse primers specific for MBD1: 5 ¶-CCT CGA GAC GGG AGA AAT AGT TAG A-3 ¶ (À695 to À671), 5 ¶-CCC GGT GTT CTA GCT GCA TTT ATG G-3 ¶ (À290 to À266), and 5 ¶-GAC TAA GAG CCA ATG GCT TTA GAG G-3 ¶ (+1359 to +1383). The primer position was relative to the first ATG in exon 2.

Preparation of Nuclear Extracts and EMSA
Nuclear extracts were isolated from three breast cancer cell lines (MCF-7, MDA-MB-231, and T47D) and HeLa cells derived from an epithelioid cervical carcinoma with 300 AL of a buffer containing 10 mmol/L of HEPES (pH 7.9), 10 mmol/L of KCl, 0.1 mmol/L of EDTA, 0.1 mmol/L of EGTA, 1 mmol/L of DTT, and 0.5 mmol/L of phenylmethylsulfonyl fluoride. The suspension was incubated for 15 min on ice and 20 AL of 10% Nonidet P40 was added and subsequently centrifuged for 60 s at 10,000 Â g. The nuclear pellet obtained was resuspended in 100 AL of a buffer containing 20 mmol/L of HEPES (pH 7.9), 25% glycerol, 0.4 mmol/L of NaCl, 1 mmol/L of EDTA, 1 mmol/L of EGTA, 1 mmol/L of DTT, and 0.5 mmol/L of phenylmethylsulfonyl fluoride, and incubated for 20 min at 4jC. After centrifugation for 5 min, 10,000 Â g at 4jC, the supernatant, corresponding to the nuclear extract, was collected.

ChIP Assays
MCF-7 and T47D cells at 80% cell confluency were fixed with 20 mL of 1% formaldehyde in minimal medium for 10 min at room temperature. The DNA/protein cross-linking reaction was stopped by adding a glycine stop-fix solution. The cells were washed with ice-cold PBS, scraped and pelleted by centrifugation for 10 min, 720 Â g at 4jC. Cells were lysed with a hypotonic lysis buffer, and the nuclei were pelleted by centrifugation for 10 min, 2,400 Â g at 4jC. The nuclei pellet was sheared in 1 mL of shearing buffer by sonication at 25% power for 4 min on ice (Sonicator UP50H; Dr. Hielscher GmbH) to chromatin fragment lengths of 200 to 1,000 bp. The stop-fix solutions, the hypotonic lysis buffer, and the shearing buffer were obtained from the ChIP-IT kit (Active Motif). The extract was precleared with protein G beads and an aliquot of the precleared supernatant was used as a positive control (gDNA). Aliquots (170 AL) of the supernatant were immunoprecipitated overnight using 4 Ag of the antibodies specific for IgG (Active Motif), MBD1 (Abcam), MBD2a (Abcam), or MeCP2 (Abcam) at 4jC. The DNA/protein/antibody complexes were incubated with the blocked protein G beads for 2 h at 4jC. After washing the beads, the immunoprecipitated DNA was eluted from the beads with 100 AL elution buffer containing 1% SDS and 50 mmol/L of NaHCO 3 for 15 min and protein-DNA crosslinks were reversed with 200 mmol/L of NaCl by incubation at 65jC for 4 h. Digestion of the proteins was done with 0.1 mmol/L of EDTA, 20 mmol/L of Tris-HCl (pH 6.5) and 2 AL of Proteinase K solution (Active Motif) for 2 h at 42jC. The DNA purified by mini-columns (Active Motif) was amplified by standard PCR or quantitative real-time PCR.
For standard PCR, the following primers were used: MAGE- was amplified in 35 cycles. After agarose gel electrophoresis, the PCR products were quantified densitometrically using the software NIH Image 6.2f (NIH).
Quantitative real-time PCR was done on a realplex 4 system (Mastercycler ep gradient S; Eppendorf AG). PCR reactions contained 8 AL of purified immunoprecipitated DNA, 5 pmol of each primer set, 10 AL of DNA QuantiTect SYBR Green PCR kit (Qiagen), and water to a final volume of 20 AL. The immunoprecipitated DNA was amplified during 45 cycles of 95jC for 30 s, 58jC for 30 s, and 72jC for 15 s. The following primers were used: MAGE-A1, 5 ¶-GTT CCC GCC AGG AAA CAT C-3 ¶ and 5 ¶-GAA CTC TAC GCC GTC CCT CAG-3 ¶; RPLP0 (housekeeping gene), 5 ¶-TTA GTT TGC TGA GCT CGC CAG-3 ¶ and 5 ¶-CTC TGA GCT GCT GCC ACC TG-3 ¶. After amplification, melting curve analysis was used to determine the specificity of the PCR products. For quantification, a serial dilution of genomic DNA was used in each run. Each sample was thermocycled in triplicate and all experiments were repeated at least thrice.

RT-PCR
For cloning of the different MBD1 isoforms and MBD2a, total RNA was prepared from HeLa cells using the RNeasy RNA-Isolation Kit (Qiagen) and were done according to the manufacturer's description. cDNA synthesis was carried out using the SuperScript First-Strand System and priming with the oligonucleotides dT (Invitrogen). PCR amplification of cDNA was done with primers specific for MBD1: 5 ¶-ATG GCT GAG GAC TGG CTG GAC TGC CCG-3 ¶ and 5 ¶-TCT AGT CTG TAG AAG CCT CCA GTC TAC TGC-3 ¶ and for MBD1mut: 5 ¶-GAC AGG ATG CGA AGC AAA GTT GAG CTG-3 ¶ and 5 ¶-TCT AGT CTG TAG AAG CCT CCA GTC TAC TGC-3 ¶, and for MBD2a: 5 ¶-ATG CGC GCG CAC CCG GGG-3 ¶ and 5 ¶-TTA GGC TTC ATC TCC ACT-3 ¶. The reaction was done in a final volume of 20 AL containing PCR Buffer (Qiagen), 200 Amol/L of each deoxynucleotide triphosphate (Roche Applied Science), 0.5 Amol/L of each primer, and 2.5 units of Pfu turbo hot start polymerase (Stratagene). Template DNA was amplified in 35 cycles.

In vitro Methylation of Plasmid DNA
Twenty micrograms of the reporter plasmid constructs containing MAGE-A1, MAGE-A2, MAGE-A3, and MAGE-A12 promoter fragments were methylated by SssI or HpaII methylase (New England Biolabs) for 4 h. Efficient methylation of plasmid DNA was confirmed by its resistance to digestion with the methylation-sensitive restriction enzyme HpaII (New England Biolabs). A control digestion was done using the isoschizomer MspI (New England Biolabs).

Transfection and Luciferase Reporter Assay
The MCF-7 cell line was transiently transfected with 0.5 Ag of reporter plasmids and expression plasmids using the FuGENE 6 Reagent (Roche Applied Science). The mouse embryonic fibroblasts were transfected with 0.4 Ag of the same plasmids, however, the FuGENE HD Reagent was used. For efficiency control of transfection assays, 0.2 Ag of a CMV-hgalactosidase vector was additionally transfected as an internal control. Following 48 h of incubation, the transfected cells were lysed using the Luciferase Reporter Gene Assay Lysis Buffer (Roche Applied Science). Promoter-driven luciferase activity was measured by a 20/20 n Luminometer (Turner Biosystems) and normalized by h-galactosidase activity (Galacto-Light; Applied Biosystems). Each transfection experiment was carried out in triplicate wells and repeated at least thrice.