Targeted TET oxidase activity through methyl‐CpG‐binding domain extensively suppresses cancer cell proliferation

Abstract DNA methyltransferase (DNMT) inhibitors are epigenetic drugs used to treat myelodysplastic syndrome. They not only induce DNA demethylation but also have significant cytostatic and cytotoxic effects; however, the relationships between these characteristics have not been established yet due to the lack of a method to induce only DNA demethylation. Herein, we show that a fusion protein comprised of the methyl‐CpG‐binding domain (MBD) and the catalytic domain of Ten‐eleven translocation protein 1 (TET1‐CD) globally demethylates and upregulates a number of methylated genes. These upregulated genes frequently contained CpG islands (CGIs) within ± 1000 bp of the transcription start site (TSS). Interestingly, 65% of the genes upregulated fivefold or more by MBD‐TET1‐CDwt were also reactivated after treatment with a DNMT inhibitor, 5‐azacytidine (Aza‐CR), suggesting that gene reactivation by both methods primarily shares the same mechanism, DNA demethylation. In order to examine whether DNA demethylation affects the growth of cancer cells, we have established a tetracycline inducible system that can regulate the expression of MBD‐TET1‐CDwt in a prostate cancer cell line, LNCaP. The induction of MBD‐TET1‐CDwt demethylated and upregulated glutathione S‐transferase pi 1 (GSTP1), one of the hypermethylated genes in prostate cancer. In accordance with the reactivation of methylated genes, induction of MBD‐TET1‐CDwt extensively suppressed the growth of LNCaP cells through G1/S arrest. These results clearly indicate that TET oxidase activity recruited at methyl‐CpG sites through MBD induces reactivation of hypermethylated genes by DNA demethylation and allows us to analyze the effect of only global DNA demethylation in a wide variety of cancer cells.


Introduction
Two DNA methyltransferase (DNMT) inhibitors, 5-aza-2'-deoxycytidine (decitabine) and its analog 5-azacytidine (azacitidine), are currently used as epigenetic therapeutic drugs for patients with a subset of cancers such as myelodysplastic syndromes (MDS) and acute myelogenous leukemias (AML) [1,2]. When one of these base analogs is incorporated into DNA, DNMT becomes covalently linked to the modified base; [3] this triggers the proteasomal degradation of DNMTs, [4] leading to global demethylation and gene reactivation. These DNMT inhibitors induce significant cytostatic and cytotoxic effects. These characteristics have also been linked to the formation of covalent adducts between incorporated azacytosine bases and DNMTs, which cause stalled replication forks and pronounced changes in the cell cycle [5]. Therefore, it remains elusive to what extent DNA demethylation contributes to these characteristics. In order to assess the effects of global DNA demethylation, one of the most commonly used approaches is DNMT knockout studies. However, because DNMTs have the potential to repress transcription independently of their methylating activities [6,7], this approach may not suffice. Therefore, we sought

ORIGINAL RESEARCH
Targeted TET oxidase activity through methyl-CpG-binding domain extensively suppresses cancer cell proliferation to use the catalytic activity of the Ten-eleven translocation (TET) family proteins in order to demethylate and upregulate various methylated genes in the genome.
TET proteins are Fe(II)-and αketoglutarate (α-KG)dependent dioxygenases that successively convert 5-methylcytosine (5-mC) into 5-hydroxymethylcytosine (5-hmC), 5-formylcytosine (5-fC), and 5-carboxylcytosine (5-caC). These oxidized methylcytosines are believed to be intermediates in the process of 5-mC demethylation [8]. Previous studies have reported that overexpression of the TET1 catalytic domain (TET1-CD) in human embryonic kidney cell line 293T (HEK293T) can induce DNA demethylation and gene reactivation in exogenous nonreplicable DNA reporters and even in endogenous genomic loci with limited levels [9][10][11]. Interestingly, overexpression of full length TET1 (TET1-FL) did not induce global DNA demethylation in HEK293T cells because of the preferential binding of TET1-FL to unmethylated CpG islands (CGIs) through its CXXC domain [11]. These results indicate that a system targeting TET1-CD at hypermethylated promoter regions is necessary to demethylate and upregulate a number of methylated genes. We previously used the methyl-CpGbinding domain (MBD) [12] and developed a novel method that can reactivate silenced genes by promoter hypermethylation; this method is called "methyl-CpG targeted transcriptional activation," MeTA for short [13][14][15][16]. Reactivation is achieved by the introduction of a plasmid which carries a fusion gene comprised of MBD and the transcriptional activation domain (TAD). In this study, we constructed a fusion gene between TET1-CD and MBD to ask (1) whether this fusion gene can demethylate and upregulate a wide variety of methylated genes, and (2) whether it can suppress the growth of prostate cancer cells.

Western blotting
Western blotting was performed according to the previous protocols [19]. Membranes were probed, using mouse monoclonal anti-FLAG M2 antibody at 1:1000 dilution and mouse monoclonal anti-βactin (Sigma) antibody at 1:3000 dilution. Detection was carried out using peroxidase-conjugated antimouse IgG (GE Healthcare, Buckinghamshire, UK) antibody. Signals were visualized by reaction with ECL Detection Reagent (GE Healthcare) and digitally processed using LAS 4000 (Fujifilm, Tokyo, Japan).

Sodium bisulfite sequencing
Bisulfite treatment of the relevant genomic DNA was carried out with EpiTect Bisulfite Kit (Qiagen, Valencia, CA, USA). PCR was performed to amplify the sequences around the promoter regions of TRH and MAL. PCR products were subcloned into pKRX vector by TA cloning [22] and sequenced on an ABI sequencer with dye terminators (Applied Biosystems). Primers used for bisulfite sequencing are described previously [15] and in Table S2.

Quantitation of 5-mC and 5-hmC
EpiMark 5-hmC and 5-mC Analysis Kit (New England Biolabs, Ipswich, MA, USA) was used to quantitate the amounts of C, 5-mC, and 5-hmC in the MAL promoter region. Genomic DNA was treated with T4 βglucosyltransferase (T4-BGT) and digested with restriction endonuclease HpaII or MspI. These endonucleases recognize the same CCGG sequence, but the former cannot cut the DNA containing a methylated CpG sequence, whereas the latter can cut irrespective of methylation status. Then quantitative genomic PCR using SYBR Green was performed for quantifying the methylation status of the CpG site in CCGG sequences. Nucleotide sequences of this PCR are as follows:

Microarray
Microarray analyses were performed according to methods described previously [15]. Cy-3 labeled cRNAs from HEK293T stable cell lines were hybridized to Agilent Human Gene Expression 4x44K v2 Microarray Kit. The cut-off value for fivefold upregulation was employed for selection of the genes. High-throughput microarray data are available in the Gene Expression Omnibus database (http://www.ncbi.nlom.nih.gov/geo) under the accession number: GSE74744.

CGI analysis
We performed two types of searches; (1) between 2-kb upstream from the transcription start site (TSS) to the end of the gene, and (2) between 1-kb upstream from the TSS to 1-kb downstream. The former was termed as CGI, and the latter as CGI ± 1000-bp. The CpG islands searcher (http://cpgislands.usc.edu/) was used for Gardiner-Garden and Frommer's criteria, %GC > 50, length > 200bp, and observed CpG/expected CpG > 0.6 [23] and for Takai and Jones' criteria, %GC ≥ 55, length ≥ 500-bp, and observed CpG/expected CpG ≥ 0.65 [24].

Cell proliferation assay
LNCaP-derived 3 × 10 3 cells were seeded in 24-well dishes at day -1. Cells were treated with or without 1 μg/mL tetracycline, and cell viability was assessed at days 0, 2, 4, 6, and 8 by alamarBlue assay (Invitrogen) as described previously [18]. Experiments were performed in quadruplicate and repeated three times.

Cell cycle analysis
Cells were grown in the presence of 1 μg/mL tetracycline (replaced every other day) and were harvested at days 2, 4, 6, and 8. The cell cycle distribution of samples was determined by flow cytometric analysis (FACSCanto II; BD Biosciences) as described previously [25].

Results
The TET1-CD fused to MBD extensively upregulates epigenetically silenced genes In previous studies, overexpression of TET1-CD in HEK293T induced DNA demethylation and gene reactivation with limited induction levels [11]. In order to maximize the effect of TET1-CD, we invented a system that should recruit TET1-CD at hypermethylated promoters in a positive manner by using MBD. We first constructed fusion genes that contain the wild-type (wt) TET1 catalytic domain (TET1-CDwt) as well as the catalytically inactive TET1-CD mutant (TET1-CDmut) with or without MBD and transiently transfected each DNA construct into HEK293T cells. Immunofluorescence analyses using anti-5-hmC antibody clearly indicated that TET1-CDwt, but not TET1-CDmut, is functional and produces 5-hmC in cell nuclei (Fig. S1). Then we asked whether these fusion constructs could upregulate a number of methylated genes. Because insufficient transfection efficiency will mask the precise effects of TET1-CD, we made two independent stable cell lines expressing each fusion construct, as shown in Figure 1A, and analyzed the expression of five methylated genes, including SRY (sex determining region Y)-box 17 (SOX17), mutL homolog 1 (MLH1), myelin and lymphocyte protein (MAL), metallothionein 1M (MT1M), and thyrotropin-releasing hormone (TRH) [15]. These five genes carrying CGIs at their promoter regions were randomly selected from our previous studies on hypermethylated genes that were confirmed by genomic bisulfite sequencing in HEK293T cells [13,15]. It is notable that only stable cell lines expressing MBD-TET1-CDwt extensively reactivated all five methylated genes; other constructs did not upregulate any methylated genes (Fig. 1B). These results suggest that the accumulation of TET1 catalytic activity at hypermethylated promoters through MBD facilitates the extensive upregulation of endogenous methylated genes.

MBD-TET1-CDwt demethylates hypermethylated promoters
We hypothesized that MBD-TET1-CDwt should upregulate a number of genes by demethylating their hypermethylated promoters. TET1 catalytic activity likely oxidizes 5-mC with the help of MBD; subsequently, several mechanisms including passive DNA demethylation through replication [26] and active DNA demethylation through base excision repair (BER) [27] converts these oxidized methylcytosines into cytosines. In order to confirm DNA demethylation, we analyzed CpG sites in promoters of TRH and MAL by genomic bisulfite sequencing. Results are schematically shown in Figure 2. MBD-TET1-CDwt-expressing HEK293T showed frequent demethylation, but the other two vectortransfected stable cell lines along while parental HEK293T remained highly methylated. To further analyze the cause(s) of demethylation, we stoichiometrically examined 5-mC and 5-hmC in the MAL promoter, using the EpiMark 5-hmC and 5-mC Analysis Kit (Fig. 2C). When 5-hmC occurs in the context of CCGG, this modification converts a cleavable MspI site to a noncleavable one. Real-time PCR analysis clearly indicates that production of 5-hmC with extensive amounts of cytosine was observed only in the MAL promoter of MBD-TET1-CDwt #10.
In addition to hypermethylated promoters, distal, and proximal enhancer regions are critical to regulate gene expression and can be targeted by MBD-TET1-CDwt. Long interspersed nucleotide elements (LINEs) were also analyzed as an indicator for global DNA demethylation and we found that DNA demethylation only increased in MBD-TET1-CDwt #10 (data not shown).

MBD-TET1-CDwt upregulates a number of genes with CGIs in the promoter
To understand the characteristics of genes upregulated by MBD-TET1-CDwt, we performed gene expression microarray analysis using two distinct, independently isolated transfectants obtained from each vector as follows: empty vector, TET1-CDwt, MBD-TET1-CDwt, and MBD-TET1-CDmut. We first validated the adequacy of this system by the examining genes analyzed in Figure 1B. Signal intensities of all the analyzed genes (SOX17, MLH1, MAL, MT1M and TRH) were significantly higher in transfectants with MBD-TET1-CDwt #3 and #10 than in ones with the empty vector, TET1-CDwt, or MBD-TET1-CDmut. Analysis using normalized intensity values clearly indicated that more genes are upregulated in cells containing MBD-TET1-CDwt in comparison with ones containing empty vector, TET1-CDwt, or MBD-TET1-CDmut (Fig. S2). The number of genes upregulated by fivefold or more in transfectants with TET1-CDwt, MBD-TET1-CDwt, and MBD-TET1-CDmut, when compared with empty vector, were 1, 60, and 6, respectively (Table 1 and S1). We searched the CGIs of each upregulated gene according to two accepted criteria: one by Gardiner-Garden and Frommer [23] and the other by Takai and Jones [24]. Interestingly, the frequency of CGIs within ± 1000-bp of the TSS were significantly higher in genes upregulated by MBD-TET1-CDwt by the former criteria. Because the core region of about 1000-bp centered at the TSS is generally unmethylated and is associated with open chromatin structure and transcriptional activation, these data may indicate that MBD-TET1-CDwt efficiently affects these distinct regions.

Genes upregulated by MBD-TET1-CDwt are frequently overlapped with ones upregulated by Aza-CR
We previously demonstrated that MeTA can identify different hypermethylated genes from those recognized by Aza-CR and discussed the differences in reactivation mechanisms of these two methods. In the present study, we analyzed upregulated genes by three methods; Aza-CR, MeTA, and MBD-TET1-CDwt. Because we used the Whole Human Genome DNA Microarray 4x44K v1 in a previous study and v2 in this study, we excluded the results of newly added probes in v2 to fairly compare the data. The genes upregulated fivefold or more were selected as shown by a Venn diagram (Fig. 3).  Thirteen commonly upregulated genes were identified. It is notable that 33 of 51 (65%) upregulated genes by MBD-TET1-CDwt were also upregulated by Aza-CR. This result may be caused by DNA demethylation as the common reactivation mechanism. To confirm this hypothesis, we analyzed methylation statuses of four cancer/testis (CT) antigen genes, MAGEA2B, MAGEA8, MAGEA9, and XAGE1B that were commonly upregulated by both MBD-TET1-CDwt and Aza-CR. The genomic bisulfite sequencing analyses clearly demonstrated that the promoter regions of all four CT antigen genes were hypermetylated in HEK293T cells but were demethylated in MBD-TET1-CDwt-expressing HEK293T cells. The representative results were shown in Figure S3. On the other hand, no genes upregulated by both MBD-TET1-CDwt and MeTA excluding Aza-CR were found.

Induction of MBD-TET1-CDwt demethylates and upregulates GSTP1 gene in LNCaP cells
To clarify whether DNA demethylation affects the growth of cancer cells, we established a tetracycline inducible system of MBD-TET1-CDwt in a human prostate cancer cell line, LNCaP. As shown in Figure 4A, MBD-TET1-CDwt inducible cell line LNCaP_TR15-MBD-TET1-CDwt #16 extensively induced MBD-TET1-CDwt protein at 2 days after tetracycline addition compared to another cell line, LNCaP_TR9-MBD-TET1-CDwt #17. Accordingly, glutathione S-transferase pi 1 (GSTP1) that is known to be one of the frequently hypermethylated genes in prostate cancer, gradually upregulated its expression from day 2 after tetracycline addition (Fig. 4B). We further performed genomic bisulfite sequencing of the GSTP1 promoter region that covers 37 CpG sites in exon 1 and intron 1 and identified that most of the CpG sites were methylated in the parental cell line LNCaP_TR15 and its derivative LNCaP_TR15-MBD-TET1-CDwt #16 without tetracycline, whereas LNCaP_TR15-MBD-TET1-CDwt #16 cells with tetracycline treatment showed prominent demethylation (Fig. 4C).

Discussion
DNA demethylation has been induced by DNMT inhibitors or disruption of the DNMT gene in a wide variety of cancer cells [28]. These DNMT-dependent methods often show phenotypic changes such as growth suppression and apoptosis in cancer cells. However, DNMTs have DNA methylation-independent transcriptional repression activity [6,7] thus, cellular effects caused by DNMT inhibition might be quite irrelevant to DNA demethylation. In addition, DNMT inhibitors impair DNA synthesis by forming azacytosine base-DNMT adducts and may affect cellular phenotypes [5]. For these reasons, we tried to isolate the effects of DNA demethylation alone in cancer cells by using a new DNMT-independent method; the TET catalytic domain is targeted at hypermethylated promoters through MBD to oxidize 5-mC. These oxidized methylcytosines were successively converted to cytosines. Our results using this DNMT-independent method strongly suggest that the reactivation of hypermethylated genes by DNA demethylation suppresses the growth of LNCaP prostate cancer cells, largely consistent with previous results using DNMT inhibitor kazinol Q [29]. MBD-TET1-CDwt induced growth suppression without apoptosis; kazinol Q induced apoptosis in LNCaP cells. The difference in cell phenotypes induced by these two methods may be explained by differences in the DNA demethylation level or by the presence of other factors, including methylationindependent gene reactivation and inhibition of DNA synthesis.
It is notable that 33 of 51 (65%) genes upregulated by fivefold or more in MBD-TET1-CDwt expressing HEK293T cells were included in Aza-CR-mediated upregulated genes (see Fig. 3). Because both methods are based on DNA hypomethylation, the frequency of commonly upregulated genes tends to be higher. Using the MeTA method, on the other hand, the spectrum of selected genes differs, because this method upregulates hypermethylated genes without DNA demethylation; Aza-CR upregulated genes classified as cancer/testis (CT) antigen such as MAGE (melanoma antigen), GAGE (G antigen), and XAGE (X antigen), whereas MeTA did not upregulate these genes [15]. However, MBD-TET1-CDwt transfectants did upregulate these CT antigen genes. The majority of CT antigen  genes contain short CGIs defined by Gardiner-Garden and Frommer at their promoter regions. These results may suggest that effector molecules such as TET1-CDwt and NFκB-TAD can control the key that determines whether the hypermethylated genes are activated even though MBD always targets CGIs in the CT antigen genes. A future question remains of how the effectors differ in their activation of hypermethylated genes.
Recently, customizable transcription activator-like effector (TALE) proteins and TET1 catalytic domain enabled targeted DNA demethylation; [30] this kind of technology can clarify the biological significance of individual hypermethylation-mediated silenced genes. Our targeted DNA demethylation system has created another system for genome-wide global demethylation using MBD, rather than TALE proteins. This newly developed system, MBD-TET1-CDwt, will discover yet to be identified methylated genes critical for cellular phenotypes in cancer. For instance, the common upregulated gene(s) in several prostate cancer cell lines by MBD-TET1-CD system will provide clues to clarify gene(s) associated with cellular proliferation. In addition, our global DNA demethylation system will help understanding of the role of aberrant CGI hypermethylation in cancer [31-33].

Supporting Information
Additional supporting information may be found in the online version of this article: Figure S1. Schematic diagram of five fusion constructs used in this study and catalytic activity of TET1-CDwt. Figure S2. An upward shift of gene expression in MBD-TET1-CDwt stable cells. Figure S3. Bisulfite sequencing analyses of MAGEA8 and MAGEA9 genes in HEK293T and MBD-TET1-CDwtexpressing HEK293T cells. Figure S4. Cell proliferation assay of LNCaP_TR9-derived cells. Figure S5. Flow cytometry analysis of LNCaP_TR9-MBD-TET1-CDwt #17 cells. Table S1. Properties of genes upregulated by MBD-TET1-CDwt. Table S2. Nucleotide sequences of primers used in bisulfite sequences.