Role of de Novo DNA Methyltransferases and Methyl CpG-binding Proteins in Gene Silencing in a Rat Hepatoma*

The expression of metallothionein-I (MT-I), a known antioxidant, was suppressed in a transplanted rat hepatoma because of promoter methylation and was induced by heavy metals only after demethylation by 5-azacytidine (5-AzaC). Treatment of the tumor-bearing rats with 5-AzaC resulted in significant regression of the hepatoma. When the inhibitor-treated tumor was allowed to grow in a new host, MT-I promoter was remethylated, which suggested de novo methylation. The activities of both de novo (3-fold) and maintenance DNA methyltransferases (DNMT) (5-fold) were higher in the hepatoma than in the host liver. The mRNA levels of the de novomethyltransferases DNMT3a and DNMT3b were 3- and 6-fold higher, respectively, in the tumor implicating transcriptional up-regulation of these two genes in this tissue. Immunohistochemical analysis showed exclusive localization of DNMT3a in the nuclei of both the liver and hepatoma, whereas DNMT3b was detected in the nuclei as well as the cytoplasm. Immunoblot assay showed that the levels of DNMT1, DNMT3a, and DNMT3b proteins in the hepatoma were 5-, 10-, and 4-fold higher, respectively, than in the liver. The mRNA level of the major methyl CpG-binding protein (MeCP2) was 8-fold higher in the tumor compared with the liver. Immunohistochemical studies showed that MeCP2 is localized exclusively in the nuclei of both tissues. A chromatin immunoprecipitation assay demonstrated that MeCP2 was associated with the MT-I promoter in the hepatoma implicating its involvement in repressing the methylated promoter. Analysis of the DNA isolated from the liver and hepatoma by RLGS-M (restriction landmark genomic scanning with methylation-sensitive enzyme) (NotI) showed that many genes in addition to MT-I were methylated in the hepatoma. These data demonstrate suppression of theMT-I gene and probably other genes in a solid tumor by promoter methylation and have provided potential molecular mechanisms for the altered methylation profile of the genes in this tumor.

In the early studies, DNA methylation was considered a mechanism to control precisely the expression of a small por-tion of the entire genome amid the overwhelming noise of repetitive elements, noncoding DNA, introns, and potentially active transposable elements. Later on, its importance in the development of mammals became apparent (1,2). In the last decade, the silencing of tumor suppressor genes by promoter methylation in tumorigenesis emerged as an important alternative to gene silencing by classical mutation. There are several known tumor suppressor genes that are silenced because of promoter methylation in different tumors, e.g. Rb, p16 INK4a , BRCA-1, VHL, E-cadherin, and MLH1 (3). Besides the well established tumor suppressor genes, the list of hypermethylated genes in tumors that are not documented as tumor suppressors is growing in number. The list of these important genes includes p15 INK4a (4), ER (5), GSTPi (6), TIMP3 (7), SOC (8), DAPK1 (9), and p73 (10). These genes are silenced in different tumors ranging from acute myeloid lymphoma to breast cancer. Metallothionein (MT) is one of the recently identified genes that is silenced in a rat hepatoma (11), human hepatocellular carcinoma and metastatic adenocarcinoma (12), colorectal carcinoma (13), and human and murine hepatocellular tumor (14). The MT gene can also be silenced in cells that over-express the p80 subunit of the Ku protein, an autoantigen (15), which is due, at least in part, to promoter methylation (16). We established the molecular mechanism for the downregulation of the MT-I gene by demonstrating methylation of the CpG islands in its promoter (11,17).
Metallothioneins (MTs) 1 are a family of small cysteine-rich proteins that consists of four different isoforms, namely MT-I, MT-II, MT-III, and MT-IV (18). Of these isoforms, MT-I and MT-II are ubiquitously expressed in all tissues from lower eukaryotes to vertebrates. The expression of MT-III is restricted to glutaminergic neurons of the brain (19), and that of MT-IV is confined to squamous epithelial cells of the skin, tongue, and intestinal lining (20). MT-I and -II are considered important mediators of cellular detoxification mechanism, scavenging heavy metals such as Cd 2ϩ , Hg 2ϩ , and Cu 1ϩ as well as free radicals. These are all strong inducers of these genes (21,22). MT gene expression is also up-regulated in animals under stress (23), in response to viral infection (24) and in Cu,Zn-superoxide dismutase null mice (25). By virtue of their heavy metal binding capacity, MTs participate in maintaining cellular homeostasis of biologically essential metals such as Zn 2ϩ and Cu 1ϩ . Many cellular proteins such as DNA and RNA polymerases along with several transcription factors (e.g. Sp1, TFIIA, and MTF1) require zinc for their optimum activity (26). MTs can act both as zinc donor and acceptor and thus function in controlling the activities of these regulatory proteins (27). The silencing of the MT gene in several human and rodent tumors implicates that the absence of metallothionein protein may facilitate tumor growth.
Methylation of the MT-I promoter in the hepatoma but not in the liver of the same tumor-bearing animal prompted us to compare the expression, abundance, and importance of the DNA methylation machinery in the two tissues. The methylation-mediated silencing of genes involves an intricate interplay of many factors that methylate and assemble a repressor complex on the gene promoter, resulting in heterochromatin formation (28). The key players involved in reactions ranging from initiation to maintenance of the silenced state of a gene include DNA methyltransferases (DNMT), methyl CpG-binding proteins (MBDs), and the co-repressors that assist the MBDs in maintaining a closed chromatin structure.
Four different isoforms of DNMTs, encoded by different genes, have been identified and characterized. DNMT1, DNMT3a, and DNMT3b are enzymatically active (2,29). DNMT1 is also called maintenance methylase, as it preferentially methylates hemimethylated substrates, whereas DNMT3a and -3b are characterized as de novo methylases for their ability to methylate unmethylated DNA. The expression of DNMT1 and -3b appears to be cell cycle-dependent, whereas DNMT3a can be expressed independently of cell cycle (30). Recent studies have shown that DNMTs not only methylate CpG islands but can also form repressor complex and facilitate the formation of heterochromatin. For example, DNMT1 is known to associate with the co-repressors HDAC2 and DMAP1, which in turn bind the transcriptional repressor TSG101 and form a repressor complex at the DNA replication foci (31). Likewise, DNMT3a can bind and form complexes with HDAC1 and RP58, a transcriptional repressor associated with the heterochromatin (32). Both DNMT3a and DNMT3b can repress transcription in a methylation-independent manner and colocalize with the heterochromatin protein HP-1␣ at pericentromeric heterochromatin throughout the cell cycle (33). This diverse function of DNMTs makes it difficult to assign and understand the role of an individual protein or a complex in silencing a gene.
The silencing of methylated promoters requires binding of MBDs to symmetrically methylated CpGs (28,34). Five such proteins, which share a common methyl-binding domain, have been identified so far (35). Among these proteins, MBD1, MBD2, and MeCP2 repress the methylated promoters through recruitment of co-repressor complexes (36,37). Although mammalian MBD3, the DNA binding component of the chromatin remodeling complex NuRD, cannot bind to methylated CpG dinucleotides on its own, it is recruited to the methylated promoters through its interaction with MBD2 (38,39). These proteins are widely expressed, except in embryonic stem cells where DNA methylation is dispensable (40). By virtue of their interaction with co-repressors like HDAC, Sin3A, and Mi-2, these MBDs can form corepressor complex on a methylated CpG island, resulting in chromatin compaction and gene silencing. MBD4 encodes a uracil DNA glycosylase that repairs CpG: TpG mismatches (41).
Previous study in our laboratory showed suppression of the MT-I gene in a solid rat hepatoma due to promoter methylation (11). The C-5 of cytosine in all of the 21 CpG dinucleotides within the MT-I promoter spanning the region from Ϫ225 to ϩ1 bp are methylated (16). None of the CpGs were methylated in the liver from the tumor-bearing rats, which is consistent with the robust MT-I induction in this tissue in response to heavy metals and free radicals (42). Genomic footprinting showed that the cis-acting elements within the promoter were inaccessible to the transcription factors as a result of methylation in these regions (16). Interestingly, the level of most transcription factors that include MTF-1, a key factor involved in the specific expression of the MT-I and MT-II genes (43), was either the same or higher in the tumor relative to the host liver. These data were similar to those generated earlier in our study, where the MT-I promoter was shown to be silenced in mouse lymphosarcoma cells due to promoter methylation (17). The present study was undertaken to explore further the DNA methylation machinery in the hepatoma and liver and to determine a potential relationship between the robust growth of the hepatoma in the host and the expression and activities of maintenance and de novo DNA methyltransferases and methyl C-binding proteins.

Maintenance of Morris Hepatoma in Rat and 5-AzaC Treatment-
Morris hepatoma 3924A was grown in ACI rats by transplanting a 0.5 ϫ 2-3-mm slice of the solid tumor into their hind legs as described previously (16). For 5-AzaC treatment, rats were injected intraperitoneally with the drug (5 mg/kg body weight) dissolved in physiological saline or with saline alone (control) after 2 weeks of tumor transplantation, when the tumor growth was visible. The animals were sacrificed when the control tumor grew to 15-20 g in size (4 -6 weeks). In 5-AzaC-treated rats the tumor growth was significantly reduced. Thin slices of both 5-AzaC-treated and control tumor were retransplanted in new donor rats, and tumor growth was followed until it attained its original size.
Bisulfite Genomic Sequencing-Preparation of genomic DNA, treatment with sodium bisulfite, and amplification of the MT-I promoter was done according to the protocol optimized in our laboratory (44). The amplified DNA was digested with ApoI and Tsp509I to check complete conversion of unmethylated cytosines to uracils and sequenced with [ 33 P]ddNTP using the Thermosequenase radiolabeled terminator cycle sequencing kit (United States Biochemical).
Isolation of RNA and Northern Blot Analysis-Total RNA was isolated from the liver and hepatoma by the guanidinium thiocyanate-acid phenol method (45). Poly(A) ϩ RNA was isolated from the total RNA using the poly(A)-Tract mRNA isolation kit (Promega). Five micrograms of poly(A) ϩ RNA was fractionated by formaldehyde-agarose (1.2%) gel electrophoresis and transferred to a nylon membrane. The membrane was then hybridized with 32 P-labeled mouse MBD2 or COX-1 cDNA fragments as probe in rapid hybridization buffer (Amersham Biosciences) following the manufacturer's protocol.
RT-PCR-For RT-PCR both total RNA and poly(A) ϩ selected RNA were used. Reverse transcription was carried out with random hexamers (PerkinElmer Life Sciences) and M-MLV reverse transcriptase from 1 g of total RNA or 25 ng of poly(A) ϩ RNA following the protocol provided in the GeneAmp RNA PCR kit (PerkinElmer Life Sciences). One-fifth of the RT reaction was subsequently PCR-amplified for each of the genes of interest with dNTP (Roche Molecular Biochemicals) and Isolation of the Nuclei from the Liver and Hepatoma and Preparation of Nuclear Extract-The liver and the hepatoma nuclei were isolated by homogenization in a high density sucrose buffer following the protocol of Gorski et al. (46) and Rose et al. (47), respectively. For the DNA methyltransferase assay, the isolated nuclei were extracted with buffer containing 0.35 M KCl following the protocol of Wadzinski et al. (48). For Western blot analysis of methyl CpG-binding proteins, nuclei were resuspended in radioimmune precipitation buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1.0% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS) and sonicated (five 5-s pulses) at 4°C. After removal of the debris by centrifugation at 12,000 ϫ g for 15 min at 4°C, the nuclear proteins in the supernatant were separated by SDS-PAGE. The protein concentration in the nuclear extracts was measured with Bio-Rad reagent according to Bradford's method using bovine serum albumin as standard.
Assay of de Novo and Maintenance DNA Methyltransferase Activities-Both de novo and maintenance DNA methyltransferase activities were measured as described (49). Briefly, nuclear extract (5-60 g) was incubated in assay buffer (10 mM Hepes buffer, pH 7.9, 3 mM MgCl 2 , 100 mM KCl, and 5% glycerol) with 500 ng of double-stranded oligonucleotide (unmethylated or hemimethylated) and 3 Ci of S-adenosyl-L-[methyl-3 H]methionine (SAM, Amersham Biosciences) in a total volume of 100 l for 2 h at 37°C. The reaction was terminated by adding 300 l of stop solution (1.0% SDS, 2.0 mM EDTA, 3.0% 4-amino salicylate, 5.0% butanol, 0.25 mg/ml calf thymus DNA, 1.0 mg/ml proteinase K) and incubating at 37°C for 45 min. The reaction mix was then spotted on a Whatman GF/C filter, dried, and washed with 5% trichloroacetic acid followed by 70% ethanol. The dried filters were then counted using a Beckman liquid scintillation counter. All of the reactions were done in triplicate. The oligonucleotide sequences were as follows: top, 5Ј-GGG-GGCCAAGCGCGCGCCTGGCGCCCGGGCCGGCTCAAGCGCGCGC-CTGGCGCCCGGATC-3Ј; bottom, 5Ј-GATCCGGGCGCCAGGCGCGC-GCTTGAGCCGGCCCGGGCGCCAGGCGCGCGCTTGG-3Ј. For hemimethylated substrate, the bottom strand oligonucleotide was synthesized with 5-methyl cytosine at all cytosines followed by guanines and annealed to the same unmethylated top strand.
Antibody Production against Recombinant Proteins-To generate antibodies specific for DNMT3a and DNMT3b, we expressed the fragments of the mouse proteins lacking the N-terminal DNA binding domain as a histidine-tagged protein in Escherichia coli. Similarly, rat MeCP2 and mouse MBD2 proteins with deleted N-terminal MBDs were expressed as His-tagged proteins. DNMT3a, DNMT3b (generous gifts from Dr. En Li), and MeCP2 (a gift from Dr. Adrian Bird) partial cDNAs were amplified from the full-length cDNAs with gene-specific primers and cloned into BamHI and KpnI sites of pQE-30 vector to express as histidine-tagged proteins (Qiagen). Mouse MBD2 partial cDNA was amplified from a lung cDNA library (Stratagene) using gene-specific primers and subcloned into the KpnI and PstI sites of the same vector. The identity of the amplified products was confirmed by sequencing. Bacteria (M15) were treated with isopropyl-1-thio-␤-D-galactopyranoside (1 mM) for induction of the proteins, and the recombinant proteins were then purified through Ni-NTA resins (Qiagen) following the manufacturer's protocol. The purified proteins were separated on SDS-PAGE and stained with 0.25 M KCl to visualize the polypeptides, which were then excised out of the gel. The resulting product was then emulsified with Freund's complete adjuvant and injected into a New Zealand male rabbit to raise antibodies. The following sets of primers were used to amplify the cDNA fragments (the production of antibodies was monitored by Western blot analysis): DNMT3a-F (5Ј-CTACGCGGATC-CGCGGAGCATGCCCTCCAGCGGCCC-3Ј) and DNMT3a-R (5Ј-CTAC-GGGGTACCCCGACTCAGGCTCATCGTCGGCTGC-3Ј); DNMT 3b-F (5Ј-CTACGCGGATCCGCGATGAAGGGAGACAGCAGACATC-3Ј) and DNMT3b-R (5Ј-CTACGGGGTACCCCGATACTCTGTGCTGTCTC-CATC-3Ј); MeCP2-F (5Ј-TACGCGGATCCCCTTCCAGGAGAGAACAG-AAACCA-3Ј) and MeCP2-R (5Ј-TACGGGGTACCTGCAATCCGCTCTA-TGTAAAGTCA-3Ј); MBD2-F (5Ј-GGGGTACCGCTGTTGACCTTAGCA-GTTTTGAC-3Ј) and MBD2-R (5Ј-AACTGCAGTTACGCCTCATCTCCA-CTGTCCAT-3Ј).
Western Blot Analysis-Nuclear extract (200 -300 g) was separated on 7.5% SDS-PAGE, transferred to nitrocellulose membrane, blocked in 5% milk in TBST (Tris-buffered saline at pH 7.5 containing 0.1% Tween 20). For the DNMT3a antibody, a dilution of 1:6000 in TBST (0.1% Tween) was used. For the DNMT3b antibody, a dilution of 1:1000 in TBST (0.5% Tween) was used. For detection of MeCP2 and MBD2, a 1:5000 dilution in TBST (0.1% Tween) of the polyclonal antisera against the recombinant proteins was used. Horseradish peroxidase-conjugated anti-rabbit IgG (Amersham Biosciences) at a dilution of 1:5000 in TBST (0.1% Tween) was used as the secondary antibody. An ECL detection kit (Amersham Biosciences) was used to detect the signal in the Western blots.
Immunohistochemistry Studies-Tissue samples were harvested from the hepatoma and liver of the tumor-bearing rats. The samples were frozen immediately on dry ice and cut on a cryostat to 15-m thick sections. The sections were fixed in 4% formaldehyde for 15 min and then washed in PBS for 5 min. The sections were then treated in 0.3% Triton in PBS for 15 min. After another wash in PBS, the nonspecific binding sites were blocked by incubating the sections in 5% milk at room temperature. The endogenous peroxidase activity was quenched by incubating the sections in 0.3% H 2 O 2 in methanol for 30 min. The sections were then incubated with the primary antibodies (1:2000 dilu-tion in PBS) at 4°C overnight. After washing the section three times in PBS for 5 min, they were incubated with biotinylated anti-rabbit antibody (1:200) for 1 h at room temperature and visualized by the traditional horseradish peroxidase method. The images were acquired with a digital camera mounted on an Olympus microscope.
Chromatin Immunoprecipitation (ChIP) Assay-Intact nuclei were isolated from the liver and tumor without disturbing the internal macromolecular interaction, as described above, and then washed with PBS at 1000 ϫ g for 10 min. The pellet was resuspended in PBS (10 7 nuclei/ml) and treated with 1.0% formaldehyde at room temperature to cross-link the DNA-binding proteins to cognate cis-acting elements. The nuclei were harvested after 30 min, washed with PBS, and solubilized in lysis buffer (50 mM Tris-HCl, pH 8.1, 1% SDS, 10 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin) at a density of 10 7 nuclei/ml. The lysate was sonicated to shear chromosomal DNA to a size of ϳ500 -750 bp. The insoluble material was removed by centrifugation in a microcentrifuge at maximum speed for 15 min at 4°C. The soluble supernatant (100 l) was then diluted 10-fold with ChIP dilution buffer (16.7 mM Tris-HCl, pH 8.1, 1.1% Triton X-100, 1.2 mM EDTA, 167 mM NaCl), and chromatin immunoprecipitation was performed following the published protocol (50) with some modifications. The diluted chromatin was precleared by incubating it for 3 h with 5 l of preimmune serum-coupled protein A-agarose beads saturated with bovine serum albumin (1 mg/ml) and salmon sperm DNA (0.4 mg/ml). The mixture was spun at 2000 rpm for 2 min in a microcentrifuge at 4°C, and the supernatant was allowed to bind either to 2 l of preimmune serum or to antiserum raised against recombinant MeCP2 overnight at 4°C. Immune complexes were pulled down by protein A-agarose beads and washed extensively with different buffers as described (50). DNA precipitated by MeCP2-antibody complex as eluted, de-cross-linked, purified DNA was dissolved in 20 l of Tris-EDTA, and 1 l was used for semiquantitative PCR with gene-specific primers. DNA purified from the chromatin (input DNA) was used as the control for PCR. The primers used for amplification of rat MT-I promoter are 5Ј-TGTTCCACAC-GTCACACGG-3Ј and 5Ј-GCGGACAGTCTGCTCTCTTTATAG-3Ј. In the semiquantitative PCR, one of the primers (2 pmol/reaction) was labeled with [␥-32 P]ATP catalyzed by polynucleotide kinase. Immunoprecipitated DNA (1 l) as well as input DNA (1/100th of the total) was subjected to hot-start PCR under the following conditions: 94°C, 4 min; 25 cycles of 94°C for 30 s, 55°C for 30 s, 72°C for 30 s, and one cycle at 72°C for 5 min. The reaction products were separated on a polyacrylamide gel (5% acrylamide) with Tris borate-EDTA as running buffer. The dried gel was subjected to autoradiography and quantitated by phosphorimaging analysis.
RLGS-M Analysis-Genomic DNA was isolated from the liver and hepatoma of tumor-bearing rats, which was then subjected to RLGS-M analysis following the protocol of Okazaki et al. (51). Briefly, isolated DNA was first digested with NotI and a commonly occurring restriction enzyme, EcoRV, end-labeled with [␣-32 P]dGTP and [␣-32 P]dCTP using Klenow, and separated on a gel. A third digestion with HinfI, performed in-gel to fragment the DNA further, was run in a second dimension. The two-dimensional gel was then exposed to an x-ray film and analyzed.

MT-I Promoter Is Remethylated following Retransplantation of the 5-Azacytidine-treated Tumors into New Donor Rats
We have previously shown an example of silencing of the MT-I gene due to promoter methylation in a rapidly growing rat hepatoma (Morris hepatoma 3924A) (see "Materials and Methods" for the tumor transplantation). The CpG dinucleotides in the rat MT-I promoter are depicted in Fig. 1A. Treatment of the tumor-bearing rats 3 weeks after tumor transplantation (when tumor growth was visible) with 5-azacytidine, a potent inhibitor of DNA methyltransferases (DNMTs), resulted in demethylation of the MT-I promoter and MT induction by heavy metals (16). Significant regression of the tumor was observed upon 5-AzaC treatment compared with the untreated control. We extended this study to see the consequence of retransplanting the regressed tumor into new donor rats. Our particular interest was to study the state of MT-I promoter methylation upon retransplantation and growing the tumor in absence of the inhibitor 5-AzaC.
It has been reported previously that de novo methylation of a gene promoter can be the result of random methylation of an entire CpG island and maintenance thereafter, as was observed with many tumor suppressor genes (52). The other proposed mechanism is that the promoter methylation can be signaled through a specific chromatin structure or DNA sequence such as short or long interspersed nuclear elements (SINEs and LINEs), and thereafter the methylation spreads from the specific methylation hot spots (53). If methylation of the MT-I promoter is a random process, it is unlikely that remethylation of the promoter will occur upon growing the tumor in 5-AzaC-free medium. On the other hand, if there are any methylation hot spots on the MT-I promoter, remethylation should occur immediately after removal of 5-AzaC. To analyze the promoter remethylation event, a thin slice (ϳ0.5 ϫ 2 mm) of the regressed tumor was transplanted to a new donor and allowed to grow in the absence of 5-AzaC. As a control, an untreated tumor slice of the same size was transplanted to another donor rat. Both tumors grew back to comparable sizes, but the time required for the inhibitor-treated tumor to grow was twice (12 versus 6 weeks) that of the untreated control.
To determine the extent of MT-I promoter remethylation, genomic DNA was isolated from the hepatoma before (ϪAzaC) and after 5-AzaC (ϩAzaC) treatment and following retransplantation of the AzaC into new donors (ϩ/ϪAzaC). The isolated DNAs were treated with sodium bisulfite reagent (16), and the MT-I immediate promoter (Ϫ304 to ϩ148 bp) was PCR-amplified (see Fig. 1A) with upper strand-specific primers. After confirming the completion of bisulfite conversion by restriction enzyme digestion, the MT-I promoter was sequenced. Consistent with our previous results (16), exposure of the animals bearing the hepatoma to 5-AzaC resulted in demethylation of MT-I promoter (Fig. 1B, compare panel c with  b), as shown by conversion of demethylated C to T by bisulfite treatment. Remethylation of the MT-I promoter at all CpG sites in the tumor occurred following retransplantation (ϩ/ ϪAzaC) and growth in the absence of 5-AzaC (Fig. 1B, panel d).
These data suggest that MT-I promoter methylation is not a stochastic event. Northern blot analysis showed that the suppression of MT-I expression and promoter remethylation were concurrent (Fig. 2). No significant MT-I transcript was detected in the tumor when the animal was treated with zinc (Fig. 2, lane 5), whereas treatment with 5-AzaC resulted in MT-I gene transcription, which increased 12-fold following treatment with zinc (lane 6). The MT-I mRNA was barely detectable even after zinc treatment when the 5-AzaC-treated tumor was retransplanted and allowed to grow in the absence of the nucleotide analog (Fig. 2, lanes 9 and 10). MT-I was, therefore, not inducible by zinc in the retransplanted tumor. On the other hand, a low level of constitutive but a very high level of zinc-induced MT-I expression was observed in the liver from the control rats, (lanes 1 and 2) and from retransplanted rats (lanes 7 and 8).

Exploration of the Potential Molecular Mechanisms of Altered Promoter Methylation
The Expression and Activities of Both de Novo and Maintenance DNA Methyltransferases Are Higher in the Rat Hepatoma as Compared with the Liver-To get an insight into the potential mechanism of altered promoter methylation in the hepatoma, we determined the expression and/or activity of DNMT isoforms and MBDs in the liver and hepatoma. The levels of mRNA expression, the amount of DNMT proteins, and both de novo and maintenance methyltransferase activities were compared between the liver and hepatoma of the same tumor-bearing rats. Northern blot analysis of poly(A) ϩ RNA isolated from both the liver and hepatoma using gene-specific probes showed a considerably higher expression level of all the three DNMTs (DNMT1, DNMT3a, and DNMT3b) in the hepatoma compared with the liver (Fig. 3A). Phosphorimaging analysis revealed increase in the DNMT1 (3-fold), DNMT3a (2-3fold), and DNMT3b (6-fold) in the hepatoma compared with the liver. Three different sizes of DNMT3a mRNAs were detected in the hepatoma, whereas only two were detected in the liver. We do not know whether alternative splicing generates all of these forms. Although other groups (29) have shown expression of different spliced variants of human DNMT3b in both normal and tumor tissues (colon, pancreas, lung, spleen, placenta) by RT-PCR, we found predominant expression of only one species of DNMT3b in both liver and hepatoma by Northern blot analysis. It is likely that alternative forms are similar in size and could not be analyzed by Northern blotting. Western blot analysis of nuclear extract from hepatoma and liver revealed that DNMT1 (ϳ190 kDa) protein was expressed at a significantly higher level (5-fold) in the hepatoma compared with the liver (Fig. 3B). Similarly, the levels of DNMT3a and DNMT3b were also significantly higher (10-and 4-fold, respectively) in the hepatoma compared with the liver, which correlated with the Northern blot data. Antibodies against DNMT3a detected an ϳ130-kDa polypeptide in the hepatoma and an ϳ120-kDa polypeptide in the liver (Fig. 3C), with the difference in size likely due to post-translational modification. Two polypeptides of apparent molecular sizes of 103 and 98 kDa were specifically detected with DNMT3b antibodies in both tissues. The levels of mRNA as well as proteins of all three isoforms were significantly higher in the hepatoma relative to the liver.
To gain further insight into the potential role of DNMTs in hypermethylation of certain promoters in tumor tissues, we performed immunohistochemical analysis of DNMT3a and DNMT3b in the liver and hepatoma. We analyzed 15-m sections of both the liver and hepatoma that were probed with the respective immune sera with preimmune sera as the control. DNMT3a antisera detected the protein only in the nuclei of the hepatoma and the liver, whereas detection with the preimmune sera was negligible (Fig. 4A). These data indicate that DNMT3a is localized only in the nuclei and that the antibody raised against DNMT3a is very specific. A comparison of the liver and hepatoma with respect to DNMT3a expression showed extensive staining in the tumor compared with the liver. This may also in part be due to the larger nuclei of the hepatoma, in which the ratio of nuclei to cell volume is significantly higher than that in the liver. Analysis of DNMT3b expression showed (Fig. 4B) that DNMT3b is also localized predominantly in the nucleus. It was, however, also detected in the cytoplasm of both tumor cells and the hepatocytes. DNMT3b antiserum was also very specific, as staining by the corresponding preimmune serum was negligible.
Next we determined the de novo and maintenance DNA methyltransferase activities in the liver and hepatoma nuclear extracts using unmethylated and hemimethylated synthetic oligonucleotides as substrate and S-adenosyl-L-[methyl-3 H]methionine as methyl group donor, respectively (Fig. 4C). The de novo activity was 3-fold higher and maintenance methyltransferases activity was 5-fold higher in the hepatoma compared with the liver. Because 5-AzaC is a known inhibitor of the DNMTs, it is very likely that both of the enzyme activities would be reduced significantly in the inhibitor-treated hepatoma. This observation suggests that a clear relationship exists between the growth of the hepatoma and the increased level of DNMT, both maintenance and de novo, in this tissue.

The Expression of Different Methyl CpG-binding Proteins in the Hepatoma Is Different from That in the Liver-The methyl
CpG-binding proteins are required for the assembly of repressor complex on a methylated promoter in order to silence its expression. We, therefore, compared the levels of expression of different MBDs in the hepatoma and liver of the tumor-bearing rats. We were particularly interested in identifying the MBD that is associated with the MT-I promoter and plays a key role in silencing the gene in the tumor. This could provide a model for studying the mechanism of methylation-mediated silencing of a promoter. To determine the mRNA levels of different MBDs, we performed RT-PCR with gene-specific primers. To   FIG. 3. The expression levels of all three of the DNMTs are higher in the hepatoma than in the liver. A, Poly(A) ϩ RNA (2.5 g) isolated from both the liver and the hepatoma was subjected to Northern blot analysis to compare the level of DNMT1, DNMT3a, and DNMT3b mRNAs in the two tissues. COX1 cDNA was used to reprobe the membrane, and the corresponding mRNA served as control for RNA integrity and equalization. B, Western blot analysis of DNMT1; C, Western blot analysis of DNMT3a and DNMT3b. 100 g of nuclear extract from hepatomabearing rats were subjected to SDS-PAGE and blotted with rabbit polyclonal antibodies against DNMT1, DNMT3a, or DNMT3b. LNE, liver nuclear extract; HNE, hepatoma nuclear extracts. Western blot with Ku-70 antibody served as protein loading and transfer control. eliminate the possible amplification from the contaminating DNA, we included RNA samples without reverse transcription. The MBD2 message was analyzed by Northern blot analysis with poly(A) ϩ RNA isolated from the two tissues. We used RT-PCR for other MBDs, as their expression was too low to be detected by Northern blotting. Of the five MBDs identified thus far, the expression levels of MBD1, MBD2, and MBD4 were comparable between the liver and the hepatoma (Fig. 5, A-C). The expression of MeCP2 was 8-fold (Fig. 5D) and that of MBD3 was 2-fold higher (Fig. 5E) in the hepatoma than that in the liver.
The silencing of methylated promoter requires participation of methyl CpG-binding proteins that do not show their sequence specificity in vitro. The number of methylated CpGs required for binding varies among MBDs. For example, MeCP2 can bind to a single symmetrically methylated CpG, whereas MBD2 requires 13 methylated CpGs for optimum binding in vitro. The in vivo targets of different MBDs with very similar DNA binding domains is yet to be explored. To determine the MBD that is associated with methylated MT-I promoter, we used polyclonal antibodies against MeCP2 in a ChIP assay. We selected MeCP2 for its higher expression in the hepatoma and the ability of MeCP2 antibodies to immunoprecipitate the protein from cell extracts. Immunoblot analysis with specific antibodies showed that MeCP2 level was 2-3-fold higher in the hepatoma compared with the liver, whereas that of MBD2 (DNA binding component of the MeCP1 complex) was similar in the two tissues (see Fig. 6A). In vivo DNA-protein complexes were cross-linked with formaldehyde in intact nuclei isolated from both liver and hepatoma. The protein bound chromatin was immunoprecipitated using the antibodies with the corresponding preimmune sera used as the control (Fig. 6B). Western blot analysis of the immunoprecipitated proteins with antibodies against MeCP2 showed specific pull-down of an ϳ84-kDa polypeptide in the liver (Fig. 6B, lane 2). In the hepatoma two polypeptides of ϳ84 and 72 kDa were detected (lane 4), which may be due to the occurrence of different forms of MeCP2 in the hepatoma or to proteolytic cleavage during isolation of nuclei. After precipitation of the chromatin, the DNA-protein complex was de-cross-linked and isolated from the complex. MT-I promoter was then PCR-amplified from the precipitated DNA with primers specific to the MT-I promoter (Fig. 6C). The MT-I promoter could be amplified only from the DNA precipitated with MeCP2 antibodies from the hepatoma but not from the liver (Fig. 6C, lane 5). To show that the antibodies did not pull down DNA nonspecifically from the hepatoma chromatin, we amplified a segment of cytochrome c oxidase I (COX-1) cDNA, which was amplified at a very low but equal level in all of the immunoprecipitated DNA (Fig. 6C,  lanes 1, 2, 4, and 5). These results clearly demonstrated a strong association of MeCP2 with MT-I promoter in the hepatoma but not in the liver. Lack of association of MeCP2 with the MT-I promoter in the liver is consistent with the unmethylated status of the promoter. These data suggest that MeCP2, along with other co-repressors, participate in suppression of the methylated MT-I promoter in the hepatoma.
The strong association of the MT-I promoter with MeCP2 in the hepatoma, along with the RT-PCR and immunoblot data suggesting a higher level of MeCP2 expression in the hepatoma relative to the liver, prompted us to analyze its expression in these two tissues using immunohistochemical studies. Immu- FIG. 4. Immunohistochemical analysis shows that DNMT3a and -3b are localized predominantly in the nuclei. A, rabbit polyclonal antibody against DNMT 3a was used to stain 15-m sections of frozen liver and hepatoma. Preimmune serum from the corresponding rabbit was used as control. B, immunohistochemical analysis of hepa-toma and the liver slices with DNMT3b antiserum and preimmune serum. C, liver nuclear extract (LNE) and hepatoma nuclear extracts (HNE) were used to measure maintenance and de novo DNMT activities. Incorporation of 3 H-methyl group from S-adenosyl-L-methionine to hemimethylated or unmethylated DNA/g protein was evaluated.
nostaining of MeCP2 was localized predominantly in the nuclei of both the hepatoma and liver, where the staining in the hepatoma was significantly more pronounced than in the liver (Fig. 7). The preimmune sera failed to stain either the liver or the hepatoma section, indicating the specificity of the MeCP2 antibody. This series of data further confirms the notion that MeCP2 plays an important role in silencing the methylated MT-I promoter in the rat hepatoma.

RLGS-M Analysis Reveals Methylation of Other Genes besides MT-I Gene in the Rat
Hepatoma-It is conceivable that additional genes could also be silenced in the hepatoma, as aberrant hypermethylation of some gene promoters occurs during tumorigenesis along with global hypomethylation. It is known that de novo methylation of growth regulatory genes, particularly those of some tumor suppressor genes, and their consequent silencing contribute to tumorigenesis. To address this issue we explored the possibility of silencing known tumor suppressor genes like Rb, p53, and p16 by Northern blot analysis. We found that all of them are expressed in the hepatoma to an extent comparable with the host liver (data not shown). We also analyzed the expression level of other growth regulatory genes in this hepatoma using a Gene Array blot. With the exception of cyclin E, which is equally expressed in both liver and hepatoma, all other genes were expressed at much higher level in the hepatoma because of its highly proliferating state (data not shown). To detect additional hypermethylated promoters, we performed RLGS-M analysis with NotI enzyme of DNA isolated from the liver and hepatoma of the same animal and DNA isolated from 5-AzaC-treated hepatoma. This relatively novel procedure has been used successfully to identify the genes that are hypermethylated in a variety of tumors (54 -58). The two-dimensional separation of the end-labeled FIG. 5. The liver and hepatoma contain comparable levels of MBD1, MBD2, and MBD4 mRNAs. A, expression of MBD1 mRNA in the liver (L) and hepatoma (H). poly(A) ϩ RNA prepared from the liver and hepatoma total RNA was used as template for cDNA synthesis. Gene-specific primers (see "Materials and Methods") for MBD1 (lanes 1 and 2) and COX-1 (lanes 3 and 4) were used for PCR amplification from the cDNAs (25 cycles). RNA used directly for PCR without cDNA synthesis were the negative control (lanes 6 and 7). B, Northern blot analysis of MBD2. poly(A) ϩ RNA (2.5 g) from the liver and hepatoma was subjected to Northern blot analysis, and the same blot was probed with a 32 P-labeled MBD2 or COX-1 cDNA fragment. C, expression of MBD4 in liver and hepatoma. RT-PCR analysis was the same as described in A. Lanes 1 and 2, COX-1; DNA fragments creates an array of spots when exposed to x-ray film, and this pattern is highly reproducible from sample to sample. If a NotI site is methylated in the experimental sample, NotI will fail to cleave the DNA at this site, and the corresponding fragment will not be generated, resulting in the loss of the specific spot on the autoradiogram. The RLGS-M profile of DNA was very reproducible, and different samples could be compared to determine the spots lost or gained. Because NotI sites are present predominantly on the gene promoters, partial or complete loss of spots indicates methylation of the promoter, and a gain of spots will be due to hypomethylation of the promoter. A representative RLGS-M pattern of the liver and hepatoma DNA isolated from three separate tumorbearing rats is depicted in Fig. 8. Relative to the host liver, several spots were lost (indicated by red arrows) in the hepatoma, and a few spots were also gained (indicated by blue arrows). Of 1006 spots analyzed in the liver, 181 were missing (total and partial) in the hepatoma, and 172 were regained after 5-AazC treatment. 18% of the NotI-specific spots were lost, and 4.5% of the spots were gained in the hepatoma compared with the liver, whereas 90% of the spots lost in the hepatoma were regained after 5-AzaC treatment due to demethylation of the gene. This study demonstrates that many genes besides the MT-I gene are methylated in the hepatoma, some of which may have a significant role in tumor progression. Although identification and characterization of these genes from a NotI/EcoRV rat DNA library is in progress, it is beyond the scope of this paper. The critical issue here is that MT-I is not the only gene that is methylated in the hepatoma. DISCUSSION In the past decade, silencing of gene expression was considered the most important epigenetic alteration in tumorigenesis. The growing list of hypermethylated genes identified in the human cancer includes Rb, p16 INK4a , APC, BRCA1, E-cadherin, LKB1, MLH1, and VHL (59). There has been considerable interest in deciphering the molecular mechanism(s) of this epigenetic alteration in cancer. In an effort to understand the FIG. 6. MeCP2 associates with the methylated MT-I promoter in the rat hepatoma. A, Western blot analysis of MeCP2 and MBD2 in the liver and hepatoma nuclear extracts (LNE and HNE). Identical amount (100 g of protein) of liver and hepatoma nuclear extracts were separated by SDS-PAGE and subjected to immunoblot analysis with antibodies against MeCP2, MBD2, and the p70 subunit of Ku (N3H10, Neomarker) respectively. B, immunoprecipitation and Western blot analysis with anti-MeCP2 antibodies. Identical amount (100 g protein) of formaldehydecross-linked liver and hepatoma nuclear extracts in radioimmune precipitation buffer were immunoprecipitated with 2 l of preimmune serum (Pre-I) or immune serum (MeCP2) coupled to protein A beads after preclearing with protein A beads alone. Immunoprecipitated proteins were subjected to Western blot analysis with anti-MeCP2 antibodies (1:5000 dilution in TBST). C, specific amplification of MT-I promoter from the hepatoma nuclei after ChIP with anti-MeCP2 antibodies. Formaldehyde-cross-linked chromatin (containing equal amount of DNA) isolated from the liver and hepatoma was immunoprecipitated as described in A, and the co-precipitated DNA was de-cross-linked, purified, and amplified by semiquantitative PCR 32 P-labeled with MT-I and COX-1 gene-specific primers. alterations in the DNA methylation machinery, we selected a rat tumor model system that has been used in our laboratory for many years (60 -62). The advantages of this system are as follows. First, the corresponding parental tissue is the liver of the tumor bearing animals; both control and test samples, therefore, are derived from the same animal. Second, the tumor is a rapidly growing hepatoma that yields as much as 20 -30 g/rat following transplantation of a relatively small tumor slice into each hind leg. The high yield of the tumor is beneficial for obtaining the proteins or enzymes of interest in relatively large quantities that is required for extensive purification and characterization. Third, unlike random methylation of CpG islands that frequently occurs in cell culture, DNA/promoter methylation in the animals is physiologically significant. Further, most of the human tumors are solid tumors.
Previous study in our laboratory has shown that the MT-I gene is silenced in the rat hepatoma by promoter methylation (11). The present study addressed the following issues. First, will the MT-I promoter reactivated by demethylation with 5-AzaC treatment in vivo undergo remethylation and silencing following retransplantation into donor rats not treated with the inhibitor? Second, are there significant alterations in the DNA methylation machinery in the hepatoma relative to the host liver? Third, does the altered methylation profile of the hepatoma relative to the liver require a distinct molecular mechanism(s)?
Retransplantation of the regressed tumor led to significantly slow but comparable growth of the tumor. The promoter was completely methylated after the tumor was allowed to grow in the absence of 5-AzaC. The bisulfite sequencing of the MT-I promoter indicates complete demethylation after treatment with the inhibitor. Consequently, remethylation of the promoter should be directed by de novo methyltransferase activity, as the maintenance DNMT utilizes only hemimethylated DNA as the substrate. This is the first report to demonstrate remethylation of completely demethylated DNA in intact animals and therefore to suggest the physiological relevance of the de novo methylation of the MT-I promoter. This result is consistent with that obtained in a study with T24 cells in culture, where p16 promoter demethylated by 5-AzaC treatment was remethylated to its original level after 21 population doublings (63). During 12 weeks of growth, the 5-AzaC-treated tumor underwent at least 25 cell divisions, facilitating de novo methylation of the demethylated promoters.
An inverse relationship between actively transcribing genes and promoter methylation has also been speculated on by other laboratories. Brandeis et al. (64) have shown that CpG islands in mouse and hamster aprt gene undergo de novo methylation when the Sp1 site is mutated to block its binding to the cognate site. The binding of Sp1 to its cis-acting element prevents promoter methylation and its spreading downstream (65). Bender et al. (63) attributed the preferential methylation of p16 exon 2 CpG islands over the promoter CpG islands to occupancy of the promoter by transcription factors. The basal level of MT-I gene expression is normally very low in all cells and tissues unless challenged with heavy metals or agents that generate free radicals. In the hepatoma, where the gene is in the dormant state, the promoter is not accessible to transacting factors. We have previously shown that MT-I expression in the liver of tumor-bearing animals is significantly higher than that of an age-matched control rat due to stress caused by tumor load (11), as stress in animals is a known inducer of MT expression in the liver (23). It is logical to assume that the low but transcriptionally active state of the promoter coupled with the lesser abundance of DNA methylation machinery in the liver keeps the MT-I promoter free of methylation. Remethylation of the promoter upon removal of the drug suggests that a methylation hot spot(s) might be present on the MT gene promoter, which would facilitate the remethylation process. It is also possible that a minor fraction of cells in the regressed tumor harbors the methylated MT-I gene and that the maintenance methyltransferase methylates the hemimethylated promoter. It is, however, evident that repression of genes by promoter methylation that negatively regulates cell proliferation will have a clear advantage over the cells in which the corresponding gene is demethylated and active.
It is conceivable that the relative lack of MT induction in the tumor could lead to an increased availability of functional zinc for a variety of zinc finger proteins that include transcriptional activators. Zinc is known to play a key role in cell proliferation and growth. Indeed, treatment of the animals bearing the hepatoma with 5-AzaC resulted in significant regression of the FIG. 8. Several genes besides MT-I gene are silenced due to methylation in the hepatoma. Genomic DNA isolated from the livers of hepatoma-bearing rats and from the hepatomas under different conditions of 5-AzaC treatment were subjected to NotI/EcoRV digestion, end-labeled, and run on two-dimensional gels (see "Materials and Methods" for details). The appearance or disappearance of individual spots was compared by analyzing the autoradiograms. A section of the autoradiogram displaying spots lost in hepatoma compared with the liver and subsequent reappearance upon 5-AzaC treatment is shown. Spots present in the liver but lost in the hepatoma are indicated by red arrows, and the spots gained in the hepatoma compared with the liver are indicated by blue arrows.
tumor in parallel with decreased methylation and reactivation of the MT-I promoter. A direct link between MT-I promoter silencing and tumorigenesis is yet to be established and is beyond the scope of the present study. Admittedly, the expression of other genes that control tumor growth may also be suppressed by methylation. Nevertheless, the potential role of zinc in tumor growth and its retention in induced MT could play a key role in regulating tumor growth.
The aberrant de novo methylation in the hepatoma compared with the host liver and remethylation of the MT-I promoter provided us the impetus to study the status of the three catalytically active DNMT genes in these two tissues. As anticipated, the level of expression and activity of all three DNMTs are significantly higher in the hepatoma compared with the liver. This result is consistent with the increased expression of the DNMT1 gene in fos-transformed cells compared with normal fibroblasts and the reversal of cell transformation following inhibition of DNMT1 expression (66). Significant elevation of DNMT1 expression has also been associated with tumorigenesis, particularly in colon cancer (30). Contrary to the relatively modest degree of DNMT1 expression reported in tumors compared with the normal counterpart (6,30), our study has revealed significant elevation (3-fold) of its activity and its expression in the rat hepatoma relative to the liver. It has also been observed that in B cell neoplasm (67) and neuroblastoma (68) the human chromosome 2p23 that houses the Dnmt3a gene is amplified. Overexpression of DNMT3b has been more frequently associated with tumorigenesis (30).
Another important observation is in regard to the modification in the levels of MBD in the tumor, particularly that of MeCP2 and its interaction with the MT-I promoter. Based on the solution structure of the methyl C-binding domain, Ohki et al. (69) and Wakefield et al. (70) have proposed that the interaction between methyl CpG and MBD occurs along the major groove of a standard B-form DNA. Since no sequence specificity has been assigned for binding of these proteins to DNA, it remains uncertain as to which of these proteins will associate with a certain methylated promoter and how this specific process is accomplished. The present study has addressed the first issue by performing a ChIP assay using antibodies against MeCP2. The data clearly show that MeCP2 can associate robustly with the MT-I promoter. Because each MBD associates with a particular set or class of promoters destined for methylation and assembles a repressor complex recruiting a specific set of repressors, the co-repressors associated with MeCP2 will also have a role in the methylation and suppression of the gene. Identification of the co-repressors involved in this process is in progress.
Finally, the methylation of other genes in the hepatoma deserves comment. An important question is whether any other gene(s) besides the MT-I gene is (are) methylated. Interestingly, some of the genes for known tumor suppressor proteins, including Rb, p53, and p16, were not silenced as shown by Northern blot analysis (data not presented). On the other hand, the expression levels of most of the growth promoting genes, e.g. cyclins and cdks, were all increased in the hepatoma, which correlates with its highly proliferating state (data not presented). We used a highly reproducible technique, namely RLGS-M, to identify other genes methylated in the hepatoma. It was evident from this analysis that other genes in addition to MT-I are also methylated, along with some genes that were demethylated in the tumor relative to the liver. Since this analysis preferentially selects the methylated promoters (54), it is logical to assume that the majority of the spots lost represent the methylated genes and are confined to the promoter region. We have not identified these genes and studied their expression levels, which is beyond the scope of the present study. Nevertheless, this study has concluded that the methylation and the resultant suppression are not restricted to the MT-I gene in this tumor.