Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Methylated DNA Binding Domain Protein 2 (MBD2) Coordinately Silences Gene Expression through Activation of the MicroRNA hsa-mir-496 Promoter in Breast Cancer Cell Line

  • Sebastian Alvarado,

    Affiliation Department of Pharmacology and Therapeutics, McGill University, Montreal, Quebec, Canada

  • Joanne Wyglinski,

    Affiliation Department of Pharmacology and Therapeutics, McGill University, Montreal, Quebec, Canada

  • Matthew Suderman,

    Affiliations Department of Pharmacology and Therapeutics, McGill University, Montreal, Quebec, Canada, McGill Centre for Bioinformatics, McGill University, Montreal, Quebec, Canada

  • Stephen A. Andrews,

    Affiliation Department of Pharmacology and Therapeutics, McGill University, Montreal, Quebec, Canada

  • Moshe Szyf

    Affiliations Department of Pharmacology and Therapeutics, McGill University, Montreal, Quebec, Canada, Sackler Program for Epigenetics and Developmental Psychobiology, McGill University, Montreal, Quebec, Canada

Methylated DNA Binding Domain Protein 2 (MBD2) Coordinately Silences Gene Expression through Activation of the MicroRNA hsa-mir-496 Promoter in Breast Cancer Cell Line

  • Sebastian Alvarado, 
  • Joanne Wyglinski, 
  • Matthew Suderman, 
  • Stephen A. Andrews, 
  • Moshe Szyf


Methylated DNA binding protein 2 (MBD2) binds methylated promoters and suppresses transcription in cis through recruitment of a chromatin modification repressor complex. We show here a new mechanism of action for MBD2: suppression of gene expression indirectly through activation of microRNA hsa-mir-496. Overexpression of MBD2 in breast epithelial cell line MCF-10A results in induced expression and demethylation of hsa-mir-496 while depletion of MBD2 in a human breast cancer cell lines MCF-7 and MDA-MB231 results in suppression of hsa-mir-496. Activation of hsa-mir-496 by MBD2 is associated with silencing of several of its target genes while depletion of MBD2 leads to induction of hsa-mir-496 target genes. Depletion of hsa-mir-496 by locked nucleic acid (LNA) antisense oligonucleotide leads to activation of these target genes in MBD2 overexpressing cells supporting that hsa-mir-496 is mediating in part the effects of MBD2 on gene expression. We demonstrate that MBD2 binds the promoter of hsa-mir-496 in MCF-10A, MCF-7 and MDA-MB-231 cells and that it activates an in vitro methylated hsa-mir-496 promoter driving a CG-less luciferase reporter in a transient transfection assay. The activation of hsa-mir-496 is associated with reduced methylation of the promoter. Taken together these results describe a novel cascade for gene regulation by DNA methylation whereby activation of a methylated microRNA by MBD2 that is associated with loss of methylation triggers repression of downstream targets.


Covalent modification of cytosine in CpG dinucleotides in 5′regulatory regions of genes by methylation has been shown to regulate gene function in cis by suppressing transcription of the juxtaposed gene [1]. However, genome wide analyses of RNA transcription have repeatedly shown only partial correlation between methylation states of promoters and transcription. Most commonly seen are unmethylated promoters that are nevertheless inactive [2]. Remarkably, even pharmacological DNA demethylation does not result uniformly in gene induction but a fraction of the transcriptome is suppressed [3], [4]. This suggests that DNA methylation states are not limited to cis-suppression but might be controlling downstream cascades of gene regulatory events.

An accepted hypothesis is that the DNA methylation signal is read by methylated DNA binding proteins that recruit repressive complexes and suppress gene expression in cis [5], [6] . Methylated DNA binding protein 2 (MBD2) binds methylated DNA and has been shown to partner with the NuRD complex a multisubunit containing histone deacetylase activities (HDAC) [7], thus promoting gene silencing through inactivation of chromatin configuration [8]. There is a wide body of research that has established the role of MBD2 in methylation dependent cis-gene suppression. However, it was also proposed that MBD2 could act as an activator of gene expression either through recruitment of histone acetyl transferases (HAT) and other transcriptional activators such as TACC3 [9] the HTLV-1 TAX1 activator [10] or through promoting DNA demethylation [11], [12] although the DNA demethylation biochemical activity of MBD2 has been contested by several studies [7], [8].

We have previously shown that MBD2 is required for activation and maintaining a demethylated state of prometastatic genes in breast cancer [13] prostate cancer [14] and liver cancer [15] and that inhibition of MBD2 in colorectal and lung cancer cell lines reverses tumor growth as explants in vivo [16] [17] and overexpression of MBD2 leads to activation of Methylated TypeII Hexokinase Gene in hepatocytes [18]. A conserved sequence required for demethylation of cytokines in mature Th2 cells CNS-1 is bound to MBD2 when these cells undergo demethylation [19]. mbd2−/− mice exhibit hypermethylation of certain tumor suppressor genes which is partial in mbd2−/+ mice [20]. In this paper, we examined the hypothesis that MBD2 could cause coordinate gene repression through activation of repressive pathways of gene regulation.

Highly-networked candidate repressors in the cell are microRNAs. microRNAs are small 17–22 nt non coding RNAs expressed in several eukaryotic organisms that regulate the stability and processing of target mRNA through direct binding to 3′UTRs [21], [22]. The average microRNA has the potential to bind up to 100 different targets in the cell positioning them as nodal global regulators of disease where transcriptional programs involving multiple genes change drastically [23]. MicroRNAs have been shown to play a role in cell proliferation, differentiation, apoptosis and development causing small reductions in the level of hundreds of target mRNAs for each microRNA [21], [22]. We tested, therefore, the possibility that MBD2 could affect gene repression of network of genes through activation of a methylated microRNA. Although microRNAs have been demonstrated by several studies to be regulated by DNA methylation [24][27] chromatin modification [28] and transcriptional regulators [29], the possibility that methylated DNA binding proteins could control expression of a group of genes through changing the methylation state of a microRNA has not been considered. We show here that MBD2 could cause gene repression of a group of genes through activating a microRNA that silences these genes.

Materials and Methods

Plasmid Promoter Constructs and in vitro methylation of mir-496

A PCR amplified fragment (Using primers in Table 1) containing the mir-496 promoter 5′ regulatory region (−315→+161 relative to the TSS) was cloned into PCR2.1 and sub-cloned using HindIII and BamHI restriction sites into the CpG-free pCpGl luciferase reporter [30] in sense and antisense directions. Methylation of promoter constructs was carried out with 2 rounds of methylation with the CG specific SssI Methyltransferase and the methyl donor S-adenosylmethionine as recommended by the manufacturer's guidelines (NEB, Cat#. M0226L).

Luciferase Assay

For luciferase assays; HEK293 cells were plated at a density of 1×105 cells per well in 6 well plates. MBD2b and mtMBD2b (with a deletion of the methylated DNA binding domain) were subcloned into the pEF6 (Invitrogen) expression vector from the previously described pcDNA3.1-His-MBD2 [11]. To generate the mutant MBD deleted plasmid, pcDNA3.1-His-MBD2 was digested (KpnI) to remove the MBD domain (nucleotide 601–812) and blunted by Klenow (Roche). Subsequently, the pcDNA3.1 plasmid was digested (NotI) and the released fragment was ligated into pEF6 (Invitrogen). Co-transfections for luciferase assays were performed with CaCl2 precipitation as described previously [31] with a total of 2 ug of plasmid transfected using a fixed 200 ng of luciferase promoter and a ratio of 1∶3 of expression construct (MBD2/mtMBD2 to empty backbone, pef6). Constructs were transfected into HEK293 cells, harvested at 72 hours and assayed with the Luciferase Assay System as recommended by the manufacturer's guidelines (Promega, Cat#. E1483). Luciferase activity per condition was normalized to total protein concentration.

Cell culture and transfections

Human non-invasive breast cancer cells MCF-7, invasive breast cancer cells MDA-MB-231 and nontransformed immortalized breast epithelial cells MCF-10A were purchased from American Type Culture Collection (ATCC, Manassas, Virginia). MCF-7 cells were cultured in minimum Eagle's medium with 10 µg/ml of insulin (Invitrogen, Carlsbad, California). MDA-MB-231 cells were cultured in Dulbecco's modified Eagle's medium (Invitrogen). All media were supplemented with 10% fetal bovine serum, 2 mM glutamine, 100 U/ml penicillin and 100 µg/ml streptomycin. MCF-10A cells were cultured in Mammary Epithelial Basal Media (MEGM®; Lonza/Clonetics Corporation, CC-3150) supplemented with the BulletKit® provided by the manufacturer. The GA-1000 (gentamycin–amphotericin B mix) from the BulletKit® was not added to the media as recommended by ATCC. Transient transfections were carried out using lipofectin (Invitrogen, Cat. #18292-037) as described previously [32] with minor modifications, siRNA was transfected at a final concentration of 70 nM. SiRNAs were selected from one of the siGENOME SMARTpool® sequences provided by Dharmacon, (Lafayette, Colorado). This concentration was optimized in preliminary experiments using transient knockdowns at several concentrations from 20 to 200 nM. Scrambled control was used as control (target sequence ordered: scrambled control 5′-GCCUUGGCAGCCUAGGCGA-3′, and siMBD2 5′-UUACUAGGCAUCAUCUUUCUU-3′). Both Anti-mir-496 (Exiqon, Cat.# 410275-00 5′-AGATTGGCCATGTAATACTC-3′) locked nucleic acids (LNA) and negative control LNA (Exiqon, Cat. # 199004-00 5′GTGTAACACGTCTATACGCCCA-3′) were transfected with lipofectin to a final concentration of 75 nM as described above. Lentiviral infection of MCF-10A was performed as described previously [33]. Briefly, 3×106 HEK293 cells were cultured 24 hours before transfected with 5 ug of pMD2G-VSVG, 5 ug of pCMV-R8.91 and 5 ug of MBD2-lenti plasmid (Open Biosystems) using Fugene (Roche). Cells were washed with PBS and incubated 48 hours with DMEM (Invitrogen). Supernatant was collected and filtered with 0.45 um disk filter and used to infect MCF-10A (plated at 500,000 cells 24 hours prior to infection).

Expression arrays and validation

RNA was extracted using Trizol according to manufacturer's instructions (Invitrogen, 15596-026). For transcriptome analysis, 1 µg of RNA from MCF-10A cells infected with an MBD2 lentivirus or control empty virus and were subjected to microarray expression analysis using Affymetrix Human Genome U133_Plus 2.0 (Array hybridization was performed at the Génome Québec Innovation Centre, Montréal, Canada). Biological replicates were normalized using the RMA method. Differentially expressed genes were chosen to be those with greater than 1.5-fold increase or less than 0.5-fold decrease in each sample as compared to untreated control. For expression array validation of the microarray results a total of 3 ug of RNA was used for RT-PCR cDNA conversion with AMV reverse transcriptase (Roche, Life Technologies) and random hexamers. Primers were designed to extend across exon boundaries to rule out genomic DNA contamination (Autoprime) and expression was normalized relative to GAPDH expression (primers in Table 1). Quantitative PCR (qPCR) was carried out using the SYBR Green method (Roche, Life Technologies) on LightCycler 480. MicroRNA cDNA conversion was carried out using the Miscript PCR System (Qiagen,Cat #218160). Expression of the mature mir-496 was measured using miScript primer assays (Qiagen, MS00007707) that spanned the entire mature miRNA and calculated its levels relative to RNU1A1 noncoding housekeeping RNA control . This control was suggested by the manufacturer to be stable across different conditions. qPCR was carried out using miScript SYBR Green PCR kit (Qiagen, Cat. # 218073). All data was Analyzed using the Absolute Relative Quantification LightCycler 480 software.

MBD2 Chromatin Immunoprecipitation

MCF-10A infected with MBD2 lentivirus and control were enriched for MBD2-bound DNA through immunoprecipitation as described previously. Briefly, 3 million cells from each cell line was fixed with 1% formaldehyde for 15 minutes at 37°C in the presence of protease inhibitor (Complete, Roche). Fixed cells were then lysed and subjected to sonication. Each sample was pre-cleared with protein G agarose and divided into three sub-samples: Input (100 µl), Bound (900 µl)-to be incubated with 50 ug anti-MBD2 sheep polyclonal antibody at 1 mg/ml (Upstate-Millipore, 07-198), and control-to be incubated with sheep IgG non-specific antibody (negative control, Santa-Cruz Biotechnology, sc-2717) overnight at 4°C. The next day, the unbound fraction was removed, and DNA bound to the beads was subjected to multiple salt washes. Washes were carried out with low salt wash (0.1% SDS, 1% Triton-X, 2 mM EDTA, 20 mM Tris, 150 mM NaCl), high salt(Same as low salt only 500 mM NaCl) , LiCl wash (0.25 M LiCl, 1% NP-40, 1% deoxycholate, 1 mM EDTA, 10 mM Tris pH 8) followed by six TE washes. The bound fractions were then eluted and the antibodies were degraded by protease K treatment. ChIP DNA was used as a template for QPCR.

Methylated DNA Immunoprecipitation (MeDIP) and Array Hybridization

MCF-10A infected with MBD2 lentivirus and control were enriched for methylated DNA through immunoprecipitation as described previously by Cedar's group [34] For array hybridization, labeled input and bound DNA samples were hybridized to custom designed 244K promoter tiling array (Agilent Technologies) that contained probes covering all transcription start sites at intervals from 800 bp upstream to 200 bp downstream of all genes described in Ensembl (version 44) and within 250 bp of approximately 400 microRNAs from miRBase, all at 100 bp-spacing. The array covered 36,957 transcription start sites corresponding to 18,468 genes. All the steps of hybridization, washing, and scanning were done following the Agilent protocol for ChIP-on-chip analysis. Immunoprecipitation validation was calculated by measuring qPCR of bound over input using primers in Table 1 on Roche LightCycler 480. All microarray data are MIAME compliant and the raw data have been deposited in Gene Expression Omnibus (GEO) at NCBI (, accession numbers: GSE47857 (Methylation profiles of MCF-10A cells infected with MBD2 lentivirus) and GSE47873 (Gene expression profiles MCF-10A cells overexpressing MBD2)

Bisulfite mapping

Bisulfite mapping of DNA was carried out as described previously [35]on a total of 3 ug of DNA. In MCF-10A, MCF-7 and MDA-231, The genomic mir-496 locus was amplified with a nested PCR using primers in Table 1 (n = 50 clones). PCR products were then cloned into PCR2.1 using a TOPO TA cloning kit (Invitrogen, Cat. # K4500-02) and sequenced with a Beckman Coulter CEQ sequencer. Bisulfite mapping of the transiently transfected hsa-mir-496-pCpGl Luciferase expression plasmid was carried out using pyrosequencing with a first round of outer PCR primers located within the pCpGl construct (to differentiate between the exogenous transfected promoter and the endogenous promoter) and a second round of inner PCR within the cloned mir-496 sequence with biotinylated primers in Table 1. Pyrosequencing was carried out directly on the PCR product using a Biotage Q24 Pyrosequencer according to the manufacturer's guidelines. Data was analyzed with Pyromark Q24 Software.

Statistical Analysis

Statistical analysis for bisulfite mapping and qPCR data was performed using unpaired t-tests. Each value represents the mean ± S.E.M. of three independent experiments. The results were considered statistically significant when *P<0.05, **P<.001, ***P<.0001.


Over expression of MBD2 triggers demethylation and activation of hsa-mir-496

MCF-10A, MCF-7 and MDA-MB-231 mammary cell lines were used in our study since they express varying levels of endogenous MBD2 (Fig. 1A Fig. 1C) ranging from low (MCF-10A) to high, (MDA-MB-231). Deregulation of MBD2 expression was previously observed in tumors. For example Stefanska et al reported that MBD2 expression is highly elevated in hepatocellular carcinoma biopsies relative to control normal liver tissue. Similarly Billard et al. reported that in invasive breast ductal carcinoma the level of MBD2 expression is significantly associated with tumor size [36]. We then chose the cell line with the lowest level of MBD2 (MCF-10A) (Fig. 1A) as a model system to determine the impact of increased MBD2 expression level on the methylome by overexpressing ectopic MBD2 in these cells (Fig. 1C). This cell line was then subject to methylated DNA immunoprecipitation and array hybridization (and mRNA expression array analysis). From our meDIP array we identified a panel of demethylated microRNAs in response to MBD2 overexpression (Table 2). We focused our study in this paper on hsa-mir-496 which showed robust demethylation by ectopic expression of MBD2 (Fig. 1 B,D). It is clear however that levels of MBD2 per se are not exclusively determining the steady state levels of hsa-mir-496 since MCF-7 cells express lower levels of hsa-mir-496 (Fig. 1D) than MCF-10A cells despite their higher levels of endogenous MBD2 (Fig. 1C). Nevertheless, endogenous MBD2 is required for expression of hsa-mir-496 cells in both MCF-7 and MDA-MB-231 cells since depletion of MBD2 results in concomitant reduction of hsa-mir-496 expression (Fig. 1D).

Figure 1. Depletion of MBD2 in mammary breast cancer cell lines leads to induction of hsa-mir-496 expression.

(A) qPCR of endogenous MBD2 mRNA in MCF-10A, MCF-7 and MDA-231. (B) A track showing the position of demethylated probes (descending grey bars) in the hsa-mir-496 promoter region as determined by MeDIP enrichment for methylated DNA and hybridization to a genome wide promoter array. (C) qPCR of MBD2 mRNA levels in MBD2 transfected (black) MCF-10A and controls (empty), and siRNA-MBD2 treated MCF-7 and MDA-231 cells (empty boxes) and controls (black boxes). Bottom panel is a Western blot analysis with an anti MBD2 antibody (D) qPCR quantification of hsa-mir-496 in MBD2 transfected (black) MCF-10A and controls (empty), and siRNA-MBD2 treated MCF-7 and MDA-231 cells (empty) and controls (black).

Table 2. microRNAs that become demethylated upon overexpression of MBD2 in MCF-10A as determined by meDIP array.

We then validated the demethylation of hsa-mir-496 in response to expression of MBD2 as predicted by the DNA methylation array (Fig. 1B). The microRNA hsa-mir-496 5′ region upstream to the transcription start site (TSS) (Fig. 2A) is highly methylated in control MCF-10A and MCF-7 cells and is hypomethylated in MDA-MB-231 which express higher levels of hsa-mir-496 (Fig. 2B) as determined by DNA methylation mapping analysis using bisulfite converted DNA. We show that expression of ectopic MBD2 in MCF-10A cells results in almost complete demethylation of the hsa-mir-496 promoter (results of bisulfite analysis in Fig. 2D).

Figure 2. MBD2 overexpression in MCF-10A cells induces hsa-mir-496 expression and demethylation through binding to the TSS.

(A) Physical map of the 5′ region of hsa-mir-496. The balloons represent CG dinucleotide sequences. The transcription start site TSS is indicated. The position of primers used to amplify in qChIP analysis the are indicated. (B) Results of bisulfite mapping of the 5′ hsa-mir-496 in MCF-10A (empty), MCF-7 (grey) and MDA-MB-231 cells (dark) (C) qPCR ChIP of MBD2 in MCF-10A, MCF-7 and MDA-MB-231 with primers as outlined in panel A. MCF-10A [C], MCF-10A expressing ectopic MBD2 [+M], MCF-7 Control [C] and MBD2 depleted MCF-7 cells [-M], Control [C] and MBD2 depleted [-M] MDA-MB-231 cells with primers as outlined in panel A (D) Results of bisulfite mapping of the 5′ hsa-mir-496 in MCF-10A (empty) [C] transfected with MBD2 (black). (E) Results of bisulfite mapping of the 5′ hsa-mir-496 in MCF-7 cells (empty) or MBD2 depleted MCF-7 cells (Dark). (F) Results of bisulfite mapping of the 5′ hsa-mir-496 in MDA-MB-231 cells (empty) or MBD2 depleted MDA-MB-231 cells (Dark).

To determine whether MBD2 directly interacts with hsa-mir-496 promoter in the human breast cancer cell line examined, we performed a chromatin immunoprecipitation with an MBD2 antibody as previously described and amplification with primers covering the transcriptional start site (TSS) [15]. Our assay demonstrates binding of MBD2 within the (TSS) (−11–+246) of hsa-mir-496 that correlates with MBD2 expression in the different cell lines (Fig. 2C). MBD2 binding to the hypomethylated hsa-mir-496 promoter in MDA-MB-231 (Fig. 2C) is higher than its binding to the methylated promoter in MCF-10A and MCF-7 cells, consistent with a role for MBD2 in interacting with an active and demethylated hsa-mir-496 (Fig. 2C). Ectopic expression of MBD2 results in dramatically increased binding of MBD2 to the region proximal to the TSS (Fig. 2C) and loss of DNA methylation of the same region (specifically CpGs at −17,−15, and −6 are demethylated in MCF-10A cells transfected with MBD2) (Fig. 2D).

To determine whether endogenous MBD2 plays a role in hsa-mir-496 DNA methylation we measured the effects of depletion of MBD2 mRNA in two breast cancer cell lines MCF-7 and MDA-MB-231 that express significant levels of MBD2 (Fig. 1C). Hsa-mir-496 is highly methylated in MCF-10A cells and we reasoned that depletion of MBD2 in these cells would have very little further impact on DNA methylation. SiRNA knock down achieved significant MBD2 depletion of 80% in MCF-7 and 60% in MDA-MB-231) (Fig. 1C for QPCR) and depletion of MBD2 binding to the TSS of hsa-mir-496 in both cell lines (Fig. 2C). Following MBD2 knockdown in MDA-231 cells methylation at the hsa-mir-496 promoter is significantly increased at −40 while remaining sites are relatively hypomethylated following depletion of MBD2 (Fig. 2F). In MCF-7 cells, which express lower levels of MBD2 than MDA-MB-231 cells the hsa-mir-496 promoter is heavily methylated except at site −103 which is completely hypomethylated. This site is completely methylated in response to MBD2 depletion (Fig. 2E) while sites at −40 and +16, which are fully methylated in these cells, are partially demethylated following MBD2 depletion (Fig. 2E). In summary, endogenous MBD2 depletion results in alteration of the DNA methylation state of hsa-mir-496 with increased methylation of specific sites as well as demethylation of other sites. Interestingly, different sites are impacted in response to MBD2 depletion in different breast cancer cell lines. However, in both cases MBD2 depletion resulted in concomitant partial reduction in hsa-mir-496 expression (Fig. 1D; 1.76-fold and 3-fold decrease in MCF-7 cells and MDA-MB-231 respectively). It is unclear whether these partial changes in DNA methylation are involved in the inhibition of MBD2 expression or whether MBD2 has an impact on hsa-mir-496 expression that is independent of DNA methylation.

MBD2 activates methylated hsa-mir-496 promoter activity in a luciferase reporter transient transfection assay

Our data derived from forced expression of MBD2 in MCF10A cells suggested that MBD2 expression could activate and lead to demethylation of hsa-mir-496 (Fig. 1 and 2). To directly test the hypothesis that MBD2 could bind and activate the hsa-mir-496 promoter and alter its state of methylation we used a transient transfection reporter assay. The hsa-mir-496 region overlapping the predicted TSS, whose state of methylation was examined by bisulfite mapping in Fig. 2B–F, was cloned into CpG-free pCpGl –luciferase reporter (hsa-mir-496-pCpGl) (Fig. 3A). This construct has CG DNA methylation target sequences only in the hsa-mir-496 promoter region and the assay is therefore not confounded by vector DNA methylation and measures directly the effects of DNA methylation and MBD2 on the transcriptional activity of this region as well its state of methylation. We first demonstrate that this region directs transcriptional activity as indicated by the fact that the region cloned in the sense orientation directed 45-fold higher luciferase activity than in the antisense direction (Fig. 3B). In vitro methylation of all CGs in this region with Sss1 DNA methyltransferase which recapitulates the situation in MCF-10A cells reduced luciferase activity 59-fold relative to the unmethylated control supporting the conclusion that methylation of CG sites in the hsa-mir-496 promoter region silenced its activity (Fig. 3B). The transcriptional activity of methylated hsa-mir-496-pCpGl increases when the methylated reporter is co-transfected with MBD2 expression vector as compared with an empty backbone (pEF6) (1.8 fold) and does not increase with an MBD2 methylated DNA binding domain (MBD) deletion mutant (mtMBD2) (Fig. 3C). A ChIP assay with an anti-MBD2 antibody demonstrated that MBD2 directly interacts with the ectopic hsa-mir-496-promoter region (our primer set selectively amplifies ectopic hsa-mir-496 (see description in materials and methods) but not to a distal region in the vector, which served as a negative control (Fig. 3D).

Figure 3. DNA methylation silences hsa-mir-496 and ectopic MBD2 induces methylated hsa-mir-496 by transient transfection luciferase assay.

(A) Physical map of the hsa-mir-496- pCpGl Luciferase reporter. The position of CG dinucleotide sequences are indicated as balloons which are all located in the hsa-mir-496 5′ region. Arrows indicate position of primers used for Q-Chip. The position of primer used for pyrosequencing is indicated by a horizontal arrows under the scheme. (B) Relative luciferase activity in HEK 293 cells transiently transfected with hsa-mir-496 promoter cloned into pCpGl in Sense [S], Antisense [AS] and in vitro methylated sense hsa-mir-496- pCpGl [mS]. (C) Relative luciferase activity in HEK293 cells co-transfected with methylated hsa-mir-496- pCpGl and empty pEF6 vector [Control], MBD2 expression vector [MBD2] or MBD2 mutant without the MBD domain [mtMBD2]. (D) Ectopic MBD2 binding to methylated hsa-mir-496 region in the hsa-mir-496- pCpGl plasmid in transiently transfected HEK 293 cells as determined by QPCR of a ChIP assay with antiMBD2 antibody. The position of primers used for amplification is indicated in (A). (E) Bisulfite pyrosequencing of methylated hsa-mir-496-pCpGl (grey), MBD2 immunoprecipitation and bisulfite sequencing of hsa-mir-496 -pCpGl following transient-co-transfection experiment of methylated hsa-mir-496 with either pEF6 plasmids (empty) or pEF-MBD2 expression vector (black) in HEK-293 cells.

We then tested whether the binding of MBD2 to ectopic hsa-mir-496 promoter results in a change in its state of methylation. Any demethylation detected must be active since the plasmid is transiently transfected and does not contain an origin of replication. We captured the ectopic hsa-mir-496 promoter DNA molecules that were interacting with MBD2 (in both empty vector transfection where the transfected hsa-mir-496 was interacting with endogenous MBD2 and ectopic MBD2 transfectants which had an excess of ectopically transfected MBD2) by ChIP using anti MBD2 antibody and treated the captured DNA with sodium bisulfite. The bisulfite converted ectopic hsa-mir-496 promoter was amplified with specific primers (that don't amplify the endogenous gene, see Table 1) and was subjected to pyrosequencing. The results presented in Fig. 3E show that sites − 44,−40,−17,−15,−6 and +16 in methylated-hsa-mir-496-pCpGl were demethylated in endogenous MBD2 bound DNA (Fig. 3E, empty and dark box) when compared to naked methylated DNA (grey box). Ectopic expression of MBD2 increased demethylation at −108, −98, −40, −17,−15, and −6 and +16 suggesting that increasing levels of MBD2 over endogenous levels enhanced the extent of demethylation of transiently transfected hsa-mir-496 promoter bound to MBD2 (Fig. 3E dark box). Since our assay measures demethylation in molecules that are physically bound to the MBD2 protein, these results show that binding of MBD2 to the hsa-mir-496 promoter is associated with site specific demethylation.

Genes silenced in response to MBD2 overexpression are targets of hsa-mir-496

We tested the hypothesis that MBD2 would repress specific genes through activation of microRNA. in silico scanning using mirANDA identified a set of predicted targets of hsa-mir-496. We then tested whether some of these genes were silenced in MCF-10A cells that overexpressed MBD2 (Table S2)by examining an Affymetrix gene expression array of control and MBD2 overexpressing MCF-10A cells.In a preliminary scan we examined by qPCR 20 mRNAs. We then tested whether these 20 target mRNAs would also be affected by changes in MBD2 in other cell lines. Three of these 20 genes that are in silico targets of hsa-mir-496 that were found to be silenced in response to MBD2 overexpression were also regulated by MBD2 in other cell lines: Cathespin H (CTSH), POU domain class 2 transcription factor 3(POU2F3) and prostaglandin-endoperoxide synthase 1(PTGS1). CTSH, POU2F3 and PTGS1 were all down regulated in MCF-10A cells in response to MBD2 overexpression (3.4-fold, 10.7-fold and 2.13-fold respectively Fig. 4 A–C). Upon endogenous MBD2 depletion CTSH and POU2F3 were upregulated in MCF-7 (1.61-fold and 1.56-fold, respectively) and MDA-MB-231 cells (1.93-fold, and 1.91-fold respectively) (Fig. 4A–C). This supports the conclusion that endogenous MBD2 is indeed involved in silencing of these genes.

Figure 4. Repressed targets of MBD2 in MBD2 overexpressing cells are putative targets of hsa-mir-496.

(A) CTSH expression in MBD2 overexpressing MCF-10A [+M] and siMBD2 depleted MCF-7 [−M ] and MDA-231 cell lines and controls [C]. (B) POU2F3 expression in MBD2 overexpressing MCF-10A [+M] and in response to transient depletion of MBD2 [−M] in MCF-7 and MDA-231 and controls [C]. (C) PTGS1 expression in MBD2 overexpressing MCF-10A [+M] and in response to transient depletion of MBD2 [−M] in MCF-7 and MDA-231 and controls [C]. (D) hsa-mir-496 expression as determined by QPCR analysis in LNA treated MCF-10A, MCF-7 and MDA-MB-231 cells. (E) CTSH expression in response to transient knockdown of hsa-mir-496 in MCF-10A overexpressing MBD2 , MCF-7 and MDA-231 and controls . (F) POU2F3 expression in a transient knockdown of hsa-mir-496 in MCF-10A overexpressing MBD2 , MCF-7 and MDA-231 and controls. (G) PTGS1 expression in a transient knockdown of hsa-mir-496 in MCF-10A overexpressing MBD2 , MCF-7 and MDA-231 and controls.

We then determined whether the effect of up or down regulation of MBD2 on expression of these genes was mediated by hsa-mir-496. We depleted hsa-mir-496 with a locked nucleotide antisense oligonucleotide targeting hsa-mir-496 (using a scrambled LNA as a control) in MCF-10A that overexpress ectopic MBD2 as well as MCF-7 and MDA-MB-231 cells which express endogenous MBD2 (Fig. 4D). Following hsa-mir-496 knockdown, CTSH, POU2F3 were induced in all three cell lines as expected if these genes are downregulated by hsa-mir-496 (Fig. 4C,E) while PTGS1 was induced in MCF-10A cells overexpressing MBD2 and MDA-MB-231 cells expressing high level of MBD2 but not in MCF-7 cells (Fig. 4G). These data are consistent with the hypothesis that CTSH and POU2F3 suppression by MBD2 is mediated by hsa-mir-496 while PTGS1 regulation by MBD2 seems to involve other factors in MCF-7 cells.

Down regulated mRNAs with MBD2 overexpression identify networks of putative targets of hsa-mir-496

Expression analysis of mRNA in MCF-10A stably overexpressing MBD2 identified 5129 genes that were significantly (p<0.005) repressed (<0.9 ratio fold change) in comparison with control MCF-10A cells (Table S1). Cross-referencing this list with a computed list of putative hsa-mir-496 targets (using miRANDA) identified a dataset of 141 (Table S2) genes repressed by MBD2 that are putative targets of hsa-mir-496. We used Ingenuity Pathways Analysis (IPA) suite to delineate the gene networks that the genes in the list fall into. IPA identified networks with a unique role in cell movement, antigen presentation, cell cycle and cell death (Fig. 5A). Associated network functions identified down regulated mRNAs integral to pathways that promote migration (p =  6.79E-5 - 9.1E-5) and haptotaxis (p = 8.62E-5 - 2.94E-2) (Fig. 5B). Interestingly, down regulation of MBD2 was previously shown to reverse invasiveness and metastasis in breast cancer [37] and prostate cancer cell lines [14].

Figure 5. Ingenuity pathway analysis of putative targets of the MBD2-hsa-mir-496 pathway in MCF-10A cells overexpressing MBD2.

(A) A list of genes repressed by MBD2 overexpression in MCF-10A cells was compared to a computed list of hsa-mir-496 targets (miRANDA) and subjected to Ingenuity pathway analysis. (B) Associated network functions identified a network with a role in cell migration and haptotaxis. Down regulated mRNA and putative hsa-mir-496 targets are highlighted in bold and light blue outline. Data were analyzed through the use of IPA (Ingenuity® Systems,


The cis silencing of gene expression by DNA methylation is very well documented. Methylated DNA binding proteins including MBD2 bind methylated CGs and recruit chromatin repressor complexes that silence in cis gene expression [7]. The data presented here delineates a potential novel pathway of gene regulation by MBD2 that amplifies a DNA methylation signal by affecting downstream genes through mechanisms that don't necessarily require cis-DNA methylation in the affected genes.

We provide here several lines of evidence that in combination provide strong support for this hypothesis. We show first that ectopic MBD2 expression in untransformed epithelial cells results in upregulation of a microRNA hsa-mir-496 (Fig. 1D). Second, expression of hsa-mir-496 is partially dependent on endogenous MBD2 in two breast cancer cell lines; partial depletion of MBD2 results in reduction in hsa-mir-496 expression (Fig. 1D). Third, overexpression of MBD2 in MCF-10A cells triggers demethylation of the promoter of hsa-mir-496 (Fig. 2D). Fourth, MBD2 interacts with chromatin at the hsa-mir-496 promoter as determined by a ChIP assay, which is consistent with the hypothesis that MBD2 activates hsa-mir-496 directly in cis rather than the alternative hypothesis that MBD2 activation of hsa-mir-496 is mediated through MBD2 repressive activity on a putative trans acting repressor gene (Fig. 2B). Fifth, a luciferase reporter assay demonstrates that the hsa-mir-496 promoter region per se is silenced by DNA methylation and that it is activated with ectopic expression of MBD2 (Fig. 3 B and C). Sixth, MBD2 targets and binds an ectopic hsa-mir-496 promoter (Fig. 3D). This provides further evidence for direct action of MBD2 on the hsa-mir-496 promoter as an MBD2 antibody pulls down the hsa-mir-496 DNA region and not other regions on the vector. Seventh, the physically MBD2 bound ectopic hsa-mir-496 promoter molecules are partially demethylated delineating a tight relationship between MBD2 binding and DNA demethylation of the hsa-mir-496 promoter (Fig. 3E). Taken together these results are consistent with the hypothesis that MBD2 could regulate hsa-mir-496 promoter activity and its DNA methylation state.

Although our data shows that interaction of ectopic MBD2 with the hsa-mir-496 promoter results in demethylation, our data does not directly demonstrate that MBD2 is demethylating hsa-mir-496 promoters in cells nor does it claim that MBD2 is involved in its demethylation. Our data is also consistent with the alternative possibility that MBD2 interaction with this promoter results indirectly in recruitment of other mechanisms that were recently proposed to cause DNA demethylation such as complexes including TET enzymes and base excision repair activity (BER) [38] or hydroxymethylation mediated [39] or direct demethylation by DNMTs [40]. Although we have no evidence for these indirect mechanism in the case of hsa-mir-496 promoter, further experiments are required to test this hypothesis that are beyond the scope of this paper.

We also show in two breast cancer cell lines that endogenous MBD2 is required for expression of hsa-mir-496 since depletion of MBD2 results in reduction of hsa-mir-496 expression (Fig. 1D). However, not surprisingly the expression of hsa-mir-496 in these cells is not determined exclusively by the levels of MBD2. MCF10A cells express higher levels of hsa-mir-496 than MCF-7, which have higher steady-state levels of MBD2 (Fig. 1 A, C, D). Genes are known to be regulated by networks of factors that differ from cell type to cell type, nevertheless in both cell lines expression of MBD2 is required for expression.

In contrast to over expression of MBD2 in MCF-10A cells which results in a dramatic hypomethylation, depletion of MBD2 in MCF-7 and MDA-MB-231 cells results in very limited hypermethylation in spite of a significant decrease in expression of hsa-mir-496. This is consistent with the conclusion that MBD2 is required for hsa-mir-496 activity independently of the state of methylation. However, it is also possible that our assay wasn't sensitive enough to detect partial changes in DNA methylation in response to partial depletion of MBD2. The fact that MBD2 depletion results in both increase and decrease in DNA methylation of different sites suggests that MBD2 has a complex effect on the state of methylation.

A single microRNA has several hundred putative targets in the transcriptome. This allows for amplification and coordination of gene regulation events in the cell. Thus, activation of one microRNA could result in suppression of several RNAs. It has been previously shown that DNA methylation regulates expression of several microRNAs and that hypermethylation of microRNA silences them particularly in cancer leading to gene activation [41][43]. However, a microRNA dependent pathway of suppression of gene expression by activation of a microRNA by a methylated DNA binding protein has not been described before. The presence of such a pathway can explain the paradoxical observation that treating cells with DNA methylation inhibitors could result in not only in gene induction but gene silencing as well.

Although some reports have indicated a role of hsa-mir-496 in alcohol exposure [44], and aging [45] its expression has not yet been functionally linked to any repressed targets. By comparing the list of repressed genes in response to MBD2 in MCF-10A cells from an expression array and in silico predicted hsa-mir-496 targets we derived possible gene targets for MBD2-hsa-mir-496 pathway. We focused on the genes CTSH, POU2F3 and PTGS1 since they were validated to be repressed by increased MBD2 levels in MBD2 transfected MCF-10A cells. Furthermore CTSH and POU2F3 were then shown to be induced by depletion of MBD2 in MDA-MB-231 cells and MCF-7 cells (Fig. 4A,B,C). Hsa-mir-496 is required for MBD2 repression of these genes since depletion of hsa-mir-496 (Fig. 4D) in MBD2 overexpressing MCF-10A cells results in relief of repression (Fig. 4E, F,G). Hsa-mir-496 depletion in MCF-7 and MDA-MB-231 results in induction of CTSH and POU2F3 (Fig. 4E,F). PTGS1 is induced in MDA-MB-231 cells by hsa-mir-496 depletion (Fig. 4G) but not in MCF-7 cells suggesting that other factors regulate this gene in MCF-7 cells (Fig. 4G). Conversely, although PTGS1 is responsive to hsa-mir-496 depletion in MDA-MB-231 cells (Fig. 4G) it is not affected by MBD2 depletion in these cells (Fig. 4C). This could be explained by insensitivity of PTGS1 to the extent of reduction of hsa-mir-496 that is brought about by partial MBD2 depletion in MDA-MB-231 cells using siMBD2 treatment (Fig. 1D).

MBD2 was previously implicated in cancer growth and metastasis in several types of cancers [13][17] including breast cancer [13]. Part of this effect is mediated through activation of prometastatic genes that is associated with their demethylation. Although targets of MBD2 have been identified to play functional roles in cancer, few have identified targets that exert direct higher-order regulation such as a microRNAs or transcription factors. microRNAs have been extensively characterized to be hypermethylated in different cancers [41][43]. The results described here offer a different mechanism by which MBD2 could affect gene expression in cancer and other physiological states: coordinated repression of genes through activation of microRNA.

While here we present a case for the activation of a microRNA through binding of a methylated DNA binding protein that involves promoter demethylation, it most probably involves several other mechanisms including recruitment of transcriptional machineries [46], [47]and histone complexes [48]. Further experiments are required to unravel these additional events regulating hsa-mir-496 expression in different cell types.

The genes that were validated here as targets of hsa-mir-496 are known to be involved in different aspects of cancer progression. CTSH (Cathespin H) is a cysteine protease whose activity is often upregulated during cancer metastasis [49]. It is surprising therefore that this gene is down regulated by MBD2 and hsa-mir-496 in highly invasive breast cancer cells MDA-MB-231. It is clear however that what defines the metastatic state is not a select list of genes but a complex network and the overall output of the network. On the other hand downregulation of CTSH has been observed during osteolysis in highly metastatic breast cancers which is consistent with a role in promoting metastasis [50]. POU2F3 is a transcription factor that has been largely silenced in cervical cancer [51] and has been highlighted as a tumor suppressor gating the transformation of primary cell lines to metastatic melanomas [52]. Its silencing by overexpression of MBD2 through hsa-mir-496 is consistent with a role in cancer. These data suggest a different mechanism for suppression of tumor suppressors in cancer than the known mechanism of suppression by cis DNA methylation; long-range suppression through demethylation of regulatory microRNA. PTGS1 is involved in prostaglandin synthesis and it is deregulated in pancreatic cancer [53].

A limitation of our studies is that we only used mir-496 antagonists in the current study. Follow up studies should focus on other assays to strengthen this relationship (3′UTR assays, exogenous mir-496, etc.).

In addition to experimentally validating several targets of hsa-mir-496, by cross-referencing of hsa-mir-496 in silico targets with down regulated mRNAs in MBD2 overexpressing MCF-10A cells we derived a list of 141 genes whose repression is potentially downstream to the MBD2-hsa-mir-496 pathway. Ingenuity Pathway analysis of this list revealed highly significant functional gene networks involved in cellular movement, cell cycle, cell death and antigen presentation (Fig. 5A). These are molecular pathways that are potentially involved at different stages of cancer progression. Within this subset we looked directly at the pathways of down regulated mRNAs and putative targets of hsa-mir-496 to identify a possible role in migration and haptotaxis. Future studies need to test the hypothesis that this is a mechanism for a coordinated repression of important gene networks in cancer by DNA methylation regulators such as MBD2.

In summary, our data points to the intricate ways by which DNA methylation and its binding proteins could regulate gene expression. Several genome-wide studies have tried to correlate overall gene expression patterns and cis-DNA methylation states. Invariably, these are not perfect correlations. Although these inconsistencies could easily be explained by DNA methylation independent mechanisms, our data shows that bona fide DNA methylation regulators such as MBD2 could trigger a sequence of gene expression events downstream from the initial cis acting DNA methylation signals (Fig. 6 for model). The data illustrates how a DNA methylation signal in a single region could be amplified and affect multiple downstream targets without necessarily altering their state of methylation. If the targets fall into discrete functional pathways (Fig. 5) this mechanism could coordinate responses to single DNA methylation regulators such as MBD2. Similarly, pharmacological demethylation as well as global hypomethylation in cancer and other diseases must not result in gene induction exclusively as is commonly thought, but could result in gene repression as well through pathways such as the one delineated here. Future computations of the impact of DNA methylation in genome-wide and transcriptome wide studies need to take into account these hierarchies of gene expression-repression that could be triggered by single demethylation events.

Figure 6. MBD2 mediated repression through the direct activation of microRNA a model.

MBD2 represses methylated genes in cis by recruiting chromatin repressor complexes. It can also induce gene expression through demethylation or recruitment of chromatin activation complexes. A new pathway for long-range repression mediated through activation and demethylation of microRNA is supported by the data presented in this paper.

Supporting Information

Table S1.

Repressed transcripts following MBD2 overexpression in MCF-10A.


Table S2.

Repressed transcripts following MBD2 overexpression in MCF-10A that overlap with putative targets of hsa-mir-496.


Author Contributions

Conceived and designed the experiments: SA M. Szyf. Performed the experiments: SA JW. Analyzed the data: SA M. Suderman. Contributed reagents/materials/analysis tools: SAA. Wrote the paper: M. Szyf SA.


  1. 1. Razin A, Riggs AD (1980) DNA methylation and gene function. Science 210: 604–610.
  2. 2. Munnes M, Patrone G, Schmitz B, Romeo G, Doerfler W (1998) A 5′-CG-3′-rich region in the promoter of the transcriptionally frequently silenced RET protooncogene lacks methylated cytidine residues. Oncogene 17: 2573–2583.
  3. 3. Weber M, Hellmann I, Stadler MB, Ramos L, Paabo S, et al. (2007) Distribution, silencing potential and evolutionary impact of promoter DNA methylation in the human genome. Nat Genet 39: 457–466.
  4. 4. Broday L, Lee YW, Costa M (1999) 5-azacytidine induces transgene silencing by DNA methylation in Chinese hamster cells. Mol Cell Biol 19: 3198–3204.
  5. 5. Hendrich B, Bird A (1998) Identification and characterization of a family of mammalian methyl-CpG binding proteins. Mol Cell Biol 18: 6538–6547.
  6. 6. Cross SH, Meehan RR, Nan X, Bird A (1997) A component of the transcriptional repressor MeCP1 shares a motif with DNA methyltransferase and HRX proteins. Nat Genet 16: 256–259.
  7. 7. Ng HH, Zhang Y, Hendrich B, Johnson CA, Turner BM, et al. (1999) MBD2 is a transcriptional repressor belonging to the MeCP1 histone deacetylase complex. Nat Genet 23: 58–61.
  8. 8. Zhang Y, Ng HH, Erdjument-Bromage H, Tempst P, Bird A, et al. (1999) Analysis of the NuRD subunits reveals a histone deacetylase core complex and a connection with DNA methylation. Genes Dev 13: 1924–1935.
  9. 9. Angrisano T, Lembo F, Pero R, Natale F, Fusco A, et al. (2006) TACC3 mediates the association of MBD2 with histone acetyltransferases and relieves transcriptional repression of methylated promoters. Nucleic Acids Res 34: 364–372.
  10. 10. Ego T, Tanaka Y, Shimotohno K (2005) Interaction of HTLV-1 Tax and methyl-CpG-binding domain 2 positively regulates the gene expression from the hypermethylated LTR. Oncogene 24: 1914–1923.
  11. 11. Bhattacharya SK, Ramchandani S, Cervoni N, Szyf M (1999) A mammalian protein with specific demethylase activity for mCpG DNA. Nature 397: 579–583.
  12. 12. Detich N, Theberge J, Szyf M (2002) Promoter-specific activation and demethylation by MBD2/demethylase. J Biol Chem 277: 35791–35794.
  13. 13. Pakneshan P, Tetu B, Rabbani SA (2004) Demethylation of urokinase promoter as a prognostic marker in patients with breast carcinoma. Clin Cancer Res 10: 3035–3041.
  14. 14. Shukeir N, Pakneshan P, Chen G, Szyf M, Rabbani SA (2006) Alteration of the methylation status of tumor-promoting genes decreases prostate cancer cell invasiveness and tumorigenesis in vitro and in vivo. Cancer Res 66: 9202–9210.
  15. 15. Stefanska B, Huang J, Bhattacharyya B, Suderman M, Hallett M, et al. (2011) Definition of the landscape of promoter DNA hypomethylation in liver cancer. Cancer Res 71: 5891–5903.
  16. 16. Campbell PM, Bovenzi V, Szyf M (2004) Methylated DNA-binding protein 2 antisense inhibitors suppress tumourigenesis of human cancer cell lines in vitro and in vivo. Carcinogenesis 25: 499–507.
  17. 17. Slack A, Bovenzi V, Bigey P, Ivanov MA, Ramchandani S, et al. (2002) Antisense MBD2 gene therapy inhibits tumorigenesis. J Gene Med 4: 381–389.
  18. 18. Goel A, Mathupala SP, Pedersen PL (2003) Glucose metabolism in cancer. Evidence that demethylation events play a role in activating type II hexokinase gene expression. J Biol Chem 278: 15333–15340.
  19. 19. Aoki K, Sato N, Yamaguchi A, Kaminuma O, Hosozawa T, et al. (2009) Regulation of DNA demethylation during maturation of CD4+ naive T cells by the conserved noncoding sequence 1. J Immunol 182: 7698–7707.
  20. 20. Sansom OJ, Berger J, Bishop SM, Hendrich B, Bird A, et al. (2003) Deficiency of Mbd2 suppresses intestinal tumorigenesis. Nat Genet 34: 145–147.
  21. 21. Bartel DP (2004) MicroRNAs: genomics, biogenesis, mechanism, and function. Cell 116: 281–297.
  22. 22. He L, Hannon GJ (2004) MicroRNAs: small RNAs with a big role in gene regulation. Nat Rev Genet 5: 522–531.
  23. 23. Brennecke J, Stark A, Russell RB, Cohen SM (2005) Principles of microRNA-target recognition. PLoS Biol 3: e85.
  24. 24. Esteller M (2007) Epigenetic gene silencing in cancer: the DNA hypermethylome. Hum Mol Genet 16 Spec No 1: R50–59.
  25. 25. Fabbri M, Garzon R, Cimmino A, Liu Z, Zanesi N, et al. (2007) MicroRNA-29 family reverts aberrant methylation in lung cancer by targeting DNA methyltransferases 3A and 3B. Proc Natl Acad Sci U S A 104: 15805–15810.
  26. 26. Han L, Witmer PD, Casey E, Valle D, Sukumar S (2007) DNA methylation regulates MicroRNA expression. Cancer Biol Ther 6: 1284–1288.
  27. 27. Saito Y, Jones PA (2006) Epigenetic activation of tumor suppressor microRNAs in human cancer cells. Cell Cycle 5: 2220–2222.
  28. 28. Ozsolak F, Poling LL, Wang Z, Liu H, Liu XS, et al. (2008) Chromatin structure analyses identify miRNA promoters. Genes Dev 22: 3172–3183.
  29. 29. O'Donnell KA, Wentzel EA, Zeller KI, Dang CV, Mendell JT (2005) c-Myc-regulated microRNAs modulate E2F1 expression. Nature 435: 839–843.
  30. 30. Klug M, Rehli M (2006) Functional analysis of promoter CpG methylation using a CpG-free luciferase reporter vector. Epigenetics 1: 127–130.
  31. 31. Rouleau J, Tanigawa G, Szyf M (1992) The mouse DNA methyltransferase 5′-region. A unique housekeeping gene promoter. J Biol Chem 267: 7368–7377.
  32. 32. Zucman J, Delattre O, Desmaze C, Azambuja C, Rouleau G, et al. (1992) Rapid isolation of cosmids from defined subregions by differential Alu-PCR hybridization on chromosome 22-specific library. Genomics 13: 395–401.
  33. 33. Wiznerowicz M, Trono D (2003) Conditional suppression of cellular genes: lentivirus vector-mediated drug-inducible RNA interference. J Virol 77: 8957–8961.
  34. 34. Keshet I, Schlesinger Y, Farkash S, Rand E, Hecht M, et al. (2006) Evidence for an instructive mechanism of de novo methylation in cancer cells. Nat Genet 38: 149–153.
  35. 35. Clark SJ, Harrison J, Paul CL, Frommer M (1994) High sensitivity mapping of methylated cytosines. Nucleic Acids Res 22: 2990–2997.
  36. 36. Billard LM, Magdinier F, Lenoir GM, Frappart L, Dante R (2002) MeCP2 and MBD2 expression during normal and pathological growth of the human mammary gland. Oncogene 21: 2704–2712.
  37. 37. Pakneshan P, Szyf M, Farias-Eisner R, Rabbani SA (2004) Reversal of the hypomethylation status of urokinase (uPA) promoter blocks breast cancer growth and metastasis. J Biol Chem 279: 31735–31744.
  38. 38. Mohr F, Dohner K, Buske C, Rawat VP (2011) TET genes: new players in DNA demethylation and important determinants for stemness. Exp Hematol 39: 272–281.
  39. 39. Chen CC, Wang KY, Shen CK (2012) The Mammalian de Novo DNA Methyltransferases DNMT3A and DNMT3B Are Also DNA 5-Hydroxymethylcytosine Dehydroxymethylases. J Biol Chem 287: 33116–33121.
  40. 40. Chen CC, Wang KY, Shen CK (2013) DNA 5-Methylcytosine Demethylation Activities of the Mammalian DNA Methyltransferases. J Biol Chem
  41. 41. Davalos V, Moutinho C, Villanueva A, Boque R, Silva P, et al. (2012) Dynamic epigenetic regulation of the microRNA-200 family mediates epithelial and mesenchymal transitions in human tumorigenesis. Oncogene 31: 2062–2074.
  42. 42. Lehmann U, Hasemeier B, Christgen M, Muller M, Romermann D, et al. (2008) Epigenetic inactivation of microRNA gene hsa-mir-9-1 in human breast cancer. J Pathol 214: 17–24.
  43. 43. Yu F, Jiao Y, Zhu Y, Wang Y, Zhu J, et al. (2012) MicroRNA 34c gene down-regulation via DNA methylation promotes self-renewal and epithelial-mesenchymal transition in breast tumor-initiating cells. J Biol Chem 287: 465–473.
  44. 44. Wang LL, Zhang Z, Li Q, Yang R, Pei X, et al. (2009) Ethanol exposure induces differential microRNA and target gene expression and teratogenic effects which can be suppressed by folic acid supplementation. Hum Reprod 24: 562–579.
  45. 45. Noren Hooten N, Abdelmohsen K, Gorospe M, Ejiogu N, Zonderman AB, et al. (2010) microRNA expression patterns reveal differential expression of target genes with age. PLoS One 5: e10724.
  46. 46. Lee Y, Kim M, Han J, Yeom KH, Lee S, et al. (2004) MicroRNA genes are transcribed by RNA polymerase II. EMBO J 23: 4051–4060.
  47. 47. Saito Y, Saito H (2012) Role of CTCF in the regulation of microRNA expression. Front Genet 3: 186.
  48. 48. Buurman R, Gurlevik E, Schaffer V, Eilers M, Sandbothe M, et al. (2012) Histone deacetylases activate hepatocyte growth factor signaling by repressing microRNA-449 in hepatocellular carcinoma cells. Gastroenterology 143: 811–820 e811–815.
  49. 49. Kirschke H (1997) Lysosomal cysteine peptidases and malignant tumours. Adv Exp Med Biol 421: 253–257.
  50. 50. Ishibashi O, Mori Y, Kurokawa T, Kumegawa M (1999) Breast cancer cells express cathepsins B and L but not cathepsins K or H. Cancer Biochem Biophys 17: 69–78.
  51. 51. Zhang Z, Huettner PC, Nguyen L, Bidder M, Funk MC, et al. (2006) Aberrant promoter methylation and silencing of the POU2F3 gene in cervical cancer. Oncogene 25: 5436–5445.
  52. 52. Riker AI, Enkemann SA, Fodstad O, Liu S, Ren S, et al. (2008) The gene expression profiles of primary and metastatic melanoma yields a transition point of tumor progression and metastasis. BMC Med Genomics 1: 13.
  53. 53. Omura N, Griffith M, Vincent A, Li A, Hong SM, et al. (2010) Cyclooxygenase-deficient pancreatic cancer cells use exogenous sources of prostaglandins. Mol Cancer Res 8: 821–832.