In the present investigation, we studied the modulating effects of several tea catechins and bioflavonoids on DNA methylation catalyzed by prokaryotic SssI DNA methyltransferase (DNMT) and human DNMT1. We found that each of the tea polyphenols [catechin, epicatechin, and (–)-epigallocatechin-3-O-gallate (EGCG)] and bioflavonoids (quercetin, fisetin, and myricetin) inhibited SssI DNMT- and DNMT1-mediated DNA methylation in a concentration-dependent manner. The IC50 values for catechin, epicatechin, and various flavonoids ranged from 1.0 to 8.4 μM, but EGCG was a more potent inhibitor, with IC50 values ranging from 0.21 to 0.47 μM. When epicatechin was used as a model inhibitor, kinetic analyses showed that this catechol-containing dietary polyphenol inhibited enzymatic DNA methylation in vitro largely by increasing the formation of S-adenosyl-l-homocysteine (a potent noncompetitive inhibitor of DNMTs) during the catechol-O-methyltransferase-mediated O-methylation of this dietary catechol. In comparison, the strong inhibitory effect of EGCG on DNMT-mediated DNA methylation was independent of its own methylation and was largely due to its direct inhibition of the DNMTs. This inhibition is strongly enhanced by Mg2+. Computational modeling studies showed that the gallic acid moiety of EGCG plays a crucial role in its high-affinity, direct inhibitory interaction with the catalytic site of the human DNMT1, and its binding with the enzyme is stabilized by Mg2+. The modeling data on the precise molecular mode of EGCG's inhibitory interaction with human DNMT1 agrees perfectly with our experimental finding.
DNA methylation at the C5-position of cytosine within the CpG dinucleotides represents an important mechanism for epigenetic control of gene expression and the maintenance of genome integrity. Although DNA hypermethylation is associated with inactivation of genes, global genomic hypomethylation is often associated with the induction of chromosomal instability (Chen et al., 1998). A number of recent studies have suggested that bioactive food components, including both essential and nonessential nutrients, can modify DNA methylation patterns in complex ways. For instance, consumption of a methyl-deficient diet led to hypomethylation of specific CpG sites within several oncogenes (such as c-myc, c-fos, and c-H-ras), resulting in elevated levels of mRNA for these genes (Wainfan and Poirier, 1992). A recent study showed that tea catechins are effective inhibitors of human DNA methyltransferase (DNMT)-mediated DNA methylation in vitro and in cultured cancer cells (Fang et al., 2003).
DNA methylation is catalyzed by specific DNMTs. Multiple DNMTs are known to be present in humans, animals, and microorganisms, and they have varying degrees of specificity toward unmethylated and hemimethylated DNA substrates (Bestor and Ingram, 1983). S-Adenosyl-l-methionine (SAM) is the methyl donor in the DNMT-mediated DNA methylation, as in many other enzymatic methylation reactions [such as the catechol-O-methyltransferase (COMT)-mediated O-methylation of various catechol substrates], resulting in the formation of S-adenosyl-l-homocysteine (SAH) after donating its methyl group to the DNA substrate.
It is known that various catechol-containing dietary polyphenols are excellent substrates for the COMT-mediated O-methylation (Zhu and Liehr, 1996; Zhu et al., 1996, 2000, 2001). The COMT-mediated methylation of large amounts of dietary catechols would not only reduce the cellular pools of SAM but also form an equimolar amount of demethylated SAM, a feedback inhibitor of various SAM-dependent methylation reactions. Hence, it is hypothesized that the facile metabolic O-methylation of various catechol-containing dietary polyphenols may be an important cumulative factor that affects the rate of DNA methylation (Fig. 1A), and this possibility was investigated in the present study. In particular, we studied the mechanism of inhibition of enzymatic DNA methylation by catechol-containing dietary polyphenols from two common classes: tea catechins (catechin, epicatechin, and EGCG) and bioflavonoids (quercetin, fisetin, and myricetin) (structures shown in Fig. 1B). The recombinant prokaryotic SssI DNMT and human DNMT1 (a dominant form of the DNMTs in humans) were used in the present study as model enzymes for two reasons. First, the SssI DNMT is functionally similar to the human DNMT3A and DNMT3B and methylates both unmethylated and hemimethylated DNA substrates with almost equal efficiency (Okano et al., 1998), whereas the human DNMT1 preferentially methylates hemimethylated DNA over unmethylated DNA. Second, both of the enzymes are commercially available as selectively expressed proteins, which provide a stable source of enzymes needed for our study.
Materials and Methods
Chemicals. Catechin, epicatechin, EGCG, quercetin, fisetin, myricetin, SAM, SAH, and porcine liver COMT (affinity column-purified, 1396 units/mg of protein) were purchased from the Sigma Chemical Co. (St. Louis, MO). The synthetic double-stranded poly(dG-dC)·poly(dG-dC) and poly(dI-dC)·poly(dI-dC) were obtained from GE Healthcare (Little Chalfont, Buckinghamshire, UK). The average lengths of the poly(dG-dC)·poly(dG-dC) and poly(dI-dC)·poly(dI-dC) were 920 and 5469 base pairs, respectively. The prokaryotic SssI DNMT and the human DNMT1 were obtained from New England Biolabs (Beverly, MA). According to the supplier, the SssI DNMT was isolated from a strain of Escherichia coli transfected with the SssI DNMT gene from the Spiroplasma sp. strain MQ1 (Renbaum et al., 1990), and it was stored at –20°C (at 200 units/50 μl) in a buffer containing 10 mM Tris-HCl, pH 7.4, 0.1 mM EDTA, 1 mM dithiothreitol, 200 μg/ml bovine serum albumin, and 50% glycerol. DNMT1 was stored in a buffer containing 50 mM Tris-HCl, pH 7.5, 1 mM EDTA, 1 mM dithiothreitol, 7 μg/ml phenylmethylsulfonyl fluoride, 50% glycerol, and the protease inhibitor cocktail. The cocktail contained 4-(2-aminoethyl)benzenesulfonyl fluoride, pepstatin A, E64, bestatin, leupeptin, and aprotinin. [methyl-3H]SAM (specific activity, 11.2–13.5 Ci/mmol) was purchased from PerkinElmer Life and Analytical Sciences (Boston, MA). The ScintiVerse BD scintillation cocktail was obtained from Fisher Scientific (Pittsburgh, PA).
Assay of Enzymatic DNA Methylation in Vitro. The following reaction mixtures were freshly prepared on ice immediately before carrying out the in vitro DNA methylation reaction. The poly(dG-dC)·poly(dG-dC) and poly(dI-dC)·poly(dI-dC) substrates were diluted in the TE buffer (containing 10 mM Tris-HCl and 1 mM EDTA, pH 7.6) to desired stock concentrations, usually at 250 μM CpG methylation sites. Note that 1 mol of the double-stranded dG-dC:dG-dC or dI-dC:dI-dC dinucleotides was considered to contain 2 mol of the CpG methylation sites. The stock solutions of SssI DNMT and DNMT1 were further diluted in water, usually to a concentration of 1 unit/5 μl. The methyl donor SAM (containing ∼0.5 μCi of [methyl-3H]SAM) was dissolved in 0.8 mM HEPES at desired stock concentrations depending on the experiments. As recommended by the supplier of the DNMTs, the reaction buffer for SssI DNMT consisted of 50 mM NaCl, 10 mM Tris-HCl, 10 mM MgCl2, 1.0 mM dithiothreitol, pH 7.9, and the reaction buffer for DNMT1 consisted of 50 mM Tris-HCl, 1 mM EDTA, 1 mM dithiothreitol, and 5% glycerol, pH 7.8.
The above freshly prepared solutions were then added in a sequential order to a 1.5-ml microcentrifuge tube on ice: 5 μl of the reaction buffer, 5 μl of the synthetic DNA substrate, 5 μl of the enzyme, 5 μl of water, and 5 μl of SAM (mixed with [methyl-3H]SAM). It is noteworthy that in some of the experiments designed to study the effects of dietary chemicals in the presence of COMT, the 5-μl volume of water was replaced with 2.5 μl of COMT (at 2 units/2.5 μl for SssI DNMT and 4 units/2.5 μl for DNMT1) and 2.5 μl of a dietary chemical (added separately to the reaction mixture). As such, the final reaction mixture usually contained 0.125 to 10 μM CpG methylation sites, 1.25 to 20 μM SAM (containing ∼0.5 μCi of [methyl-3H]SAM), and 0.25 to 4 units of the enzyme in a final volume of 25 μl. Considering that the Mg2+ ion is a necessary cofactor for the activity of COMT, MgCl2 (at a final concentration of 2 mM) was added into the DNMT1-mediated reaction tube because the corresponding reaction buffer provided by the supplier did not contain MgCl2. The incubations were carried out at 37°C for varying lengths of time, and the reactions were arrested by immediately placing the tubes on ice, followed by addition of 145 μl of ice-cold 0.9% sodium chloride and 100 μl of salmon testes DNA (at 1 mg/ml in the TE buffer). To precipitate the DNA, 30 μl of 3 M sodium acetate and 900 μl of ice-cold ethanol (95%) were added, and the samples were placed in a –80°C freezer for 2 h. After centrifugation at ∼10,000 g for 10 min, 100 μl of the heat-deactivated flour (suspended at 40 mg/ml in double-distilled water) was added to each tube, followed by centrifugation at ∼10,000 g for 5 min. This step was designed to firmly secure the precipitated DNA pellet (including the methylated poly[dG-dC]·poly[dG-dC] or poly[dI-dC]·poly[dI-dC]) at the bottom of the microcentrifuge tubes, which would prevent the pellet from being partially washed away during the following washing steps. It is noteworthy that our comparison of various measurements showed that the addition of the flour made the intra-assay as well as interassay variations much smaller compared with the parallel assays without flour. The pellets were then gently washed three times with 70% ethanol, and each wash was followed by centrifugation for 3 min at 10,000g. The pellets (containing DNA and flour) were then resuspended in 30 μl of 100% ethanol and 1 ml of TE buffer, pH 7.6. Each vial was sonicated for ∼10 min and vortexed thoroughly to assure adequate resuspension of the pellet, and then the content was transferred to a scintillation vial (containing 4 ml of ScintiVerse BD) for measurement of 3H-radioactivity with a liquid scintillation counter (Tri-CARB 2900 TR; PerkinElmer Life and Analytical Sciences). Corresponding blank values obtained from incubations in the absence of the DNA substrate were also determined in each individual assay and subtracted. Because the same amount (1 unit) of the SssI DNMT and DNMT1 was used in most of the assays (unless otherwise indicated), the rate of DNA methylation, mediated by SssI DNMT or human DNMT1, is usually expressed as picomoles of methylated sites formed per minute (abbreviated to pmol/min). The kinetic parameters (KM and VMAX values) were calculated manually according to the Eadie-Hofstee plots (V versus V/[S]).
Determination of the Methylation Status ofRARβ Gene in Cultured Cancer Cells. The two human breast cancer cell lines (MCF-7 and MDA-MB-231) used in this study were obtained from the American Type Culture Collection (Manassas, VA). The culture of these cells was described in detail in our recent study (Liu and Zhu, 2004). The MCF-7 and MDA-MB-231 cells were treated with 0, 0.2, 1, 5, 25, or 50 μM EGCG for 3 or 6 days, respectively, depending on the rate of cell growth. Cells were then harvested for DNA extraction by using the DNeasy tissue kit (QIAGEN, Valencia, CA). The extracted DNA (1 μg) was modified by EZ DNA methylation kit (Zymo Research, Orange, CA) under conditions specified by the supplier. All unmethylated cytosines were converted to uracils, whereas 5-methylcytosines remained unaltered. The methylation status of the RARβ genes was determined by using a nested two-stage PCR approach (Palmisano et al., 2000). Sodium bisulfite-modified DNAs were first subjected to the stage I PCR to amplify the 425-base pair fragments of the RARβ gene (Fig. 2). The stage I PCR products include a portion of the CpG-rich promoter region of the gene. The primers (from Invitrogen, Frederick, MD) for the stage I PCR recognized the bisulfite-modified template but did not discriminate between methylated and unmethylated alleles. The stage I PCR products were diluted 50 to 1000 times, and 2.5 μl of the diluted PCR products was subjected to a stage II PCR with primers specifically designed for the methylated or unmethylated template. The sequences of the primers, annealing temperatures, and product sizes used in the stage I and II PCR amplifications of the RARβ gene are summarized in Fig. 2. HotStarTaq DNA polymerase (QIAGEN) in a 25-μl reaction volume was used in all PCRs. The conditions for the stage I PCR were as follows: 95°C for 15 min, denaturing at 95°C for 45 s, annealing at 53°C for 1 min, extension at 72°C for 1 min for 40 cycles, followed by a final extension at 72°C for 10 min. Conditions for the stage II PCR included 95°C for 15 min; denaturing at 95°C for 30 s, annealing at 62°C at 30 s, extension at 72°C for 30 s for 18 to 30 cycles; and a final 10-min extension step at 72°C. The number of cycles and the PCR conditions were optimized so that the amplified signal was still on the linear portion of the amplification curve. Controls (without DNA) were included for each set of PCRs. Amplified PCR products were subjected to electrophoresis using 2% agarose gels, stained with ethidium bromide, directly visualized with ultraviolet transillumination, and photographed. The integrated optical density of each band was quantified by densitometry. The relative levels of methylated and unmethylated PCR products were normalized against the PCR products obtained by using another set of primers specifically designed to amplify the regions that served as internal standard for the stage I PCR products (depicted in Fig. 2). The primers for the internal standard would not discriminate between methylated and unmethylated alleles. All PCR amplifications were carried out using a PTC-100 thermal cycler (MJ Research, Waltham, MA).
Computational Modeling of the Binding of EGCG to Human DNMT1. Energy minimizations, molecular dynamics simulations, and molecular docking experiments were performed on a Silicon Graphics Origin-350 workstation (SGI, Mountain View, CA) using the consistent valence force field (Dauber-Osguthorpe et al., 1998) implemented in the Insight II program (Accelrys Inc., San Diego, CA). The cell multipole method (Ding et al., 1992), with a distance-dependent dielectric constant (ϵ = ϵ0r, with ϵ0 = 4.0), was used for summation of nonbonding interactions. For energy minimizations, the steepest descent method was employed first to a 100 kcal/mol Å energy gradient and followed by the Polak and Ribiere conjugate gradients method (Press et al., 1993) until the final convergence criterion reached the 0.01 kcal/mol × Å energy gradient.
Homology Model of Human DNMT1. The primary sequences of human DNMT1 (protein GI number 4503351) and HhaI methylase (protein GI number 127455) were obtained from the NCBI database. Based on a recent suggestion that class I DNMTs share similar catalytic core regions (Schubert et al., 2003), a homology model of the human DNMT1 was thus constructed by using HhaI methylase (Protein Data Bank code 5MHT) as a template after multiple sequence alignments with ClustalW (Higgins et al., 1996). SAM was included in this model because this cofactor is required for maintaining the structural fold and the normal catalytic function of the enzyme (Martin and McMillan, 2002). The variable loop regions were initially constructed using the random loop search method implemented in Insight II. The whole enzyme structure was then subjected to a combination of constrained molecular dynamics and simulated annealing (starting at a temperature of 1500 K and then cooling to a temperature of 300 K) procedures (Fiser et al., 2000), during which the conserved β-sheet and α-helical structures that form the catalytic crevice and bind the DNA double helix were held in a fixed configuration, whereas the carboxylate moiety of SAM was constrained to Cys1147 and Gly1149 of the enzyme with hydrogen bonds. A set of 25 conformations collected from this procedure were then subjected to a sequence of energy minimizations: 1) minimization with the DNA double helix fixed and the hydrogen bonds constrained; 2) minimization with the DNA double helix freed; and finally, 3) minimization without any constraint. Among the lowest energy conformations selected, the final structure was validated using the Profiles-3D program (Bowie et al., 1991) and was used as template for the docking of EGCG molecule.
Molecular Docking of EGCG. After the DNA double helical structure was removed from the structural model of DNMT1, an energy-minimized EGCG structure generated by the Builder module in thr Insight II program was manually positioned within the catalytic core region between SAM and the catalytic site, where the target cytosine base is positioned in a flipped-out position. In this study, the binding pocket was defined to include all the residues within 16 Å of the EGCG molecule, and then a combination of Monte Carlo and molecular dynamics simulation procedures were performed to explore the binding conformations of EGCG using the Affinity module of the Insight II program. During this procedure, EGCG and the side chains of the amino acid residues within the defined binding site were allowed to rotate freely. Among 100 conformations obtained from the Monte Carlo sampling, the most likely binding conformation was determined, which would have the lowest energy among the minimized structures and would have maximal hydrogen bonding interaction.
Binding Energy Analysis. The binding energy calculations were performed using the semiempirical PM3 method (Stewart, 1989) for a model system comprising EGCG, SAM, and a group of key catalytic site residues (Lys1482, Cys1225, Glu1265, Asn1266, Val1267, Arg1311, and Asn1577) within 3 Å reach of the docked EGCG molecule. During the calculation, atoms in the backbone-heavy amino acids were in fixed positions. Furthermore, to explore the impact of Mg2+ on the inhibitory potency of EGCG, the binding energy (ΔEbind) was calculated after Mg2+ was appropriately introduced to form a plausible Mg2+-coordinated EGCG complex in the catalytic site. It was assumed that Mg2+ would be in a monodentate state, for this coordination mode is likely to be predominant, based on previous X-ray structural studies of several other Mg2+-containing proteins (Dudev and Lim, 2003). Therefore, water molecules were also included in our proposed computational model to satisfy the preferred octahedral coordination geometry of Mg2+ (Dudev and Lim, 2003).
Inhibition of DNMT-Mediated DNA Methylation by Tea Polyphenols and Bioflavonoids
DNA Methylation Conditions. Before we studied the effects of various dietary chemicals on DNA methylation, we first optimized the reaction conditions for the in vitro DNA methylation by determining the effects of different incubation times and the effects of different concentrations of the enzyme (prokaryotic SssI DNMT or human DNMT1), the methyl donor SAM, and the synthetic DNA substrates. Based on these measurements, a common reaction condition was devised for both SssI DNMT- and DNMT1-mediated DNA methylation reactions, which included an incubation time of 30 min, an enzyme concentration of 1 unit/25 μl, a SAM concentration of 5 μM, and substrate concentrations from 0.1 to 5 μM (data not shown). It is noteworthy that in some of the experiments designed to determine the modulating effects of the catechol-containing dietary chemicals on the rate of DNA methylation in vitro, COMT was also added to the reaction mixture, in addition to the dietary chemical. The concentration of porcine liver COMT used in the present study was 2 units/25 μl for SssI DNMT and 4 units/25 μl for DNMT1. Our measurements showed that COMT at these concentrations provided physiologically relevant rates of methylation of representative catechol substrates tested in this study (such as catechol estrogens and tea catechins) (data not shown).
Inhibition by Tea Catechins. Catechin and epicatechin inhibited the SssI DNMT-mediated DNA methylation in a concentration-dependent manner (Fig. 3, A and B). The IC50 values of catechin and epicatechin for inhibition of DNA methylation were 6.5 and 8.1 μM, respectively, and the maximal inhibition at the highest concentration (20 μM) of these two inhibitors was ∼60% in the presence of COMT. In comparison, EGCG exerted a more potent and efficacious inhibition of the SssI DNMT-mediated DNA methylation (Fig. 3C). The apparent IC50 value of EGCG was ∼0.21 μM, which was ∼30 times more potent than catechin and ∼40 times more potent than epicatechin, and a nearly complete inhibition was seen at 5 μM EGCG. The strong inhibition of DNA methylation by EGCG was repeated more than three times, and consistent results were obtained. It should be noted that when COMT was not added to the reaction system, catechin or epicatechin alone at 20 μM also exerted varying degrees of direct inhibition of SssI DNMT-mediated DNA methylation (Fig. 3, A and B). In contrast, EGCG had a strong direct inhibition of DNA methylation (Fig. 3C). Although the potency of EGCG's direct inhibition was lower than its overall inhibition when COMT was present, the efficacy of its direct inhibition was comparable with its overall inhibition (Fig. 3C). Taken together, these results showed that whereas the metabolic O-methylation of catechin and epicatechin markedly increased their overall activity for inhibition of SssI DNMT-mediated DNA methylation, the methylation of EGCG did not seem to significantly increase its inhibitory efficacy on DNA methylation.
When human DNMT1 was tested as the enzyme, the inhibitory effect of catechin, epicatechin, and EGCG on enzymatic DNA methylation was similar to their inhibition of SssI DNMT-mediated DNA methylation (Fig. 3, D–F). The apparent IC50 values of catechin and epicatechin for inhibition of DNMT1-mediated DNA methylation were 4.6 and 8.4 μM, respectively, and their maximal inhibition at the highest 20 μM inhibitor concentration was ∼80%. In comparison, EGCG (IC50 = 0.47 μM) exerted a 10- to 20-fold more potent inhibition of the human DNMT1 than did catechin and epicatechin. It is noteworthy that in the absence of COMT, catechin or epicatechin alone at 20 μM also exerted certain degrees of direct inhibition of the DNMT1-mediated DNA methylation (Figs. 2E and 3D). In comparison, EGCG had a stronger direct inhibition of DNA methylation, which had similar potency and efficacy as in the presence of COMT (Fig. 3F).
Inhibition by Bioflavonoids. Several common catechol-containing bioflavonoids (quercetin, fisetin, and myricetin) were also tested for their activity in inhibiting DNA methylation. Each of the tested bioflavonoids inhibited SssI DNMT-mediated DNA methylation in a concentration-dependent manner (Fig. 4A). The IC50 values of quercetin, fisetin, and myricetin for inhibition of SssI DNMT were 1.6, 1.0, and 0.7 μM, respectively (Fig. 4A). Myricetin was slightly more potent and efficacious than quercetin and fisetin for inhibition of DNA methylation. When COMT was not added to the test system, quercetin or fisetin at the highest concentration tested (20 μM) had only a very weak direct inhibition of SssI DNMT-mediated DNA methylation. In comparison, myricetin at the same concentration had a stronger direct inhibition (∼60%) of DNA methylation. This stronger direct inhibition by myricetin probably also determines its higher overall inhibitory efficacy on DNA methylation.
The overall inhibitory effect of quercetin, fisetin, and myricetin on human DNMT1-mediated DNA methylation was similar to their inhibition of SssI DNMT-mediated DNA methylation (Fig. 4B). Although the direct inhibition by quercetin or fisetin (at 20 μM) in the absence of COMT was rather weak, myricetin exerted quite a strong direct inhibition. In the presence of COMT, the apparent IC50 values of quercetin, fisetin, and myricetin for inhibition of the DNMT1-mediated DNA methylation were 1.6, 3.5, and 1.2 μM, respectively.
Inhibition of the Promoter Region DNA Methylation in Cultured Cells by EGCG and Catechin. A recent study showed that treatment of cultured cells with EGCG affected the methylation status of certain genes in human esophageal squamous cell carcinoma cell lines KYSE 510 and KYSE 150 (Fang et al., 2003). In the present study, we compared the effects of EGCG with catechin on the methylation status of RARβ gene in two human breast cancer cell lines (MCF-7 and MDA-MB-231). It is known that RARβ gene promoter region in MCF-7 and MDA-MB-231 cells was hypermethylated (Sirchia et al., 2000). From a morphological standpoint, the MCF-7 and MDA-MB-231 cells treated with EGCG or catechin at concentrations up to 50 μM showed little sign of cytotoxicity. A representative data set for MCF-7 cells is shown in Fig. 5. Based on the measurement of the amount of total isolated DNA from each treatment group of the cultured cancer cells, the rate of cell growth was only slightly inhibited (<15%) by EGCG or catechin (at 50 μM) in MDA-MB-231 cells, and it was basically not affected in MCF-7 cells by these two tea polyphenols (data not shown).
The methylation-specific and unmethylation-specific bands of RARβ gene were both detected in these two human breast cancer cell lines when EGCG was not present. It is noteworthy that the relative density of the methylation- and unmethylation-specific bands of the genes as shown in Fig. 6 did not represent the actual ratios between the methylation and unmethylation bands, because the ratios might have been altered after two stages of nested PCR amplification. After treatment of MCF-7 cells with EGCG at 0, 0.2, 1, 5, 25, and 50 μM for 6 days, the unmethylation-specific bands for the promoter region of the RARβ gene were found to be increased in a dose-dependent manner, whereas the methylation-specific bands were decreased dose dependently (Fig. 6). In the MDA-MB-231 cells, the change of the methylation status of the RARβ promoter region by EGCG treatment was very similar to that observed in MCF-7 cells (Fig. 6). In similar ways, the unmethylation-specific bands for the promoter region of p16 gene were increased in cultured T-47D human breast cancer cells after treatment with 5 or 20 μM EGCG, whereas the methylation-specific bands were decreased (data not shown). For comparison, we also determined the effects of catechin on the methylation status of the RARβ gene in MCF-7 and MDA-MB-231 cells. The overall change of the methylation status in these cells after treatment with catechin was similar to the treatment with EGCG, but EGCG had a more potent and somewhat stronger inhibitory effect on DNA methylation than did catechin (Fig. 6).
Kinetic Mechanisms for the Inhibition of DNA Methylation by Tea Polyphenols
Epicatechin. Using epicatechin as a representative dietary catechol, we determined the kinetic features of its inhibition of DNMT-mediated DNA methylation. We found that at a relatively low inhibitor concentration (2 μM), epicatechin decreased the VMAX values of the SssI DNMT-mediated DNA methylation, but it also slightly decreased the apparent KM values. At a higher inhibitor concentration (10 μM), the VMAX values were further decreased in a concentration-dependent manner, but the KM values were basically not further changed (Fig. 7A). This suggests a predominantly noncompetitive mechanism of inhibition when higher inhibitor concentrations were present. A decrease in the VMAX values was also observed when the human DNMT1-mediated DNA methylation was assayed in the presence of epicatechin (Fig. 7B).
To gain a better understanding of the kinetic mechanism(s) for the observed inhibition of DNA methylation by catechol-containing dietary polyphenols, we also determined the effects of SAH on the SssI DNMT- and human DNMT1-mediated DNA methylation. Representative data for the regulation of human DNMT1-mediated DNA methylation by SAH are shown in Fig. 8. We found that SAH was a very potent inhibitor of the DNMT1-mediated DNA methylation in vitro, and the estimated IC50 value for SAH was ∼0.34 μM (Fig. 8A). It is noteworthy that this inhibition by SAH is independent of the DNA substrate concentrations used (data not shown). Enzyme kinetic analyses showed that when a fixed concentration of SAM was present, increasing the concentrations of SAH resulted in continuous decreases of the VMAX values of the DNMT1-mediated DNA methylation, but the KM values were essentially not altered (Fig. 8B), clearly indicating that SAH was a noncompetitive inhibitor with respect to the formation of the methylated DNA products. This finding is in agreement with earlier studies showing that SAH is essentially a noncompetitive inhibitor of the DNMTs (James et al., 2003).
Taking together all the data we gathered, we came to the conclusion that there are two mechanistic components involved in the inhibition of DNA methylation by some of the catechol-containing dietary polyphenols (such as catechin and epicatechin): one is the direct inhibition of the DNMTs (independent of COMT-mediated methylation), and the other is the indirect inhibition of the enzyme through increased formation of SAH during COMT-mediated methylation of the dietary chemicals.
EGCG. Because our earlier studies showed that EGCG was O-methylated by COMT to a much lesser degree than catechin and epicatechin, its strong inhibition activity toward DNA methylation as observed in the present study mostly did not result from its metabolic O-methylation and the subsequent formation of SAH. In agreement with this suggestion, we found that EGCG displayed a similar efficacy for inhibition of DNA methylation in the absence of COMT, whereas this was not the case with catechin and epicatechin (Fig. 3). During our search for the possible mechanism(s) for the potent inhibition of DNA methylation by EGCG, we found that although SssI DNMT-mediated DNA methylation was essentially not affected by MgCl2 at up to 10 mM (the highest concentration tested), the human DNMT1-mediated DNA methylation was inhibited by the presence of MgCl2 in a concentration-dependent manner (Fig. 9, A and C). Most importantly, the inhibition potency (IC50 value) of EGCG for human DNMT1-mediated DNA methylation was markedly enhanced by MgCl2 (Fig. 9D). In comparison, the inhibition of SssI DNMT-mediated DNA methylation by EGCG was only slightly enhanced by MgCl2 (Fig. 9B). Because the presence of equimolar concentrations of Cl– (from NaCl or KCl) showed no appreciable effects as observed with MgCl2 (data not shown), it was believed that the inhibitory effect of MgCl2 was attributable to the Mg2+ ion.
In the presence of Mg2+, EGCG decreased the VMAX value of DNMT1-mediated DNA methylation, but it basically did not affect its KM value, thus suggesting that EGCG predominantly functions as a noncompetitive inhibitor of DNA methylation. One representative data set is shown in Fig. 10. Given all the data we gathered, it became quite clear that the strong inhibitory effect of EGCG on human DNMT1-mediated DNA methylation was independent of the COMT-mediated methylation of EGCG, but it is mainly attributable to its strong, direct inhibition of this enzyme, and the inhibition is greatly enhanced by Mg2+.
Computational Analysis of the Interaction of EGCG with Human DNMT1. To gain insights into the molecular mechanism of DNMT inhibition by EGCG, computational modeling studies were also conducted. A comparison of the X-ray structures of the class I DNMTs showed that they shared highly similar structures, especially in the catalytic core regions, which consist of several conserved α-helical and β-sheet structures (Schubert et al., 2003). From analyzing the multiple sequence alignments of several known DNMTs, it became apparent that the human DNMT1 contained unique, long variable regions, which made it more difficult to predict its conformations solely based on computational modeling analysis. Guided by the Profiles-3D scores for determining the best conformations for the variable regions, a homology model for human DNMT1 was constructed by using the lowest energy conformations. Figure 11A shows the mode that was developed in the present study for EGCG binding in the catalytic core region of the human DNMT1. In this binding mode, the key binding site residues on human DNMT1 included those that are within the 3-Å reach of the bound EGCG molecule (i.e., Pro1224, Cys1225, Phe1228, Glu1265, Asn1266, Val1267, Arg1309, Arg1310, Arg1311, Val478, Gly1481, Lys1482, and Asn1577). Four of these residues are predicted to form H-bonds with EGCG, which include Glu1265 (Oϵ with the D-ring OH), Arg1310 (Namide with the B-ring OH), Arg1311 (Nϵ with the B-ring OH and NNH2 with the D-ring OH), and Lys1482 (Namide with the A-ring OH).
From the initial binding energy analyses of the EGCG binding conformations using the consistent valence force field in the Affinity docking program, we predicted that the presence of Mg2+ would alter EGCG's binding energy and, accordingly, its inhibitory activity. A more detailed calculation of the binding energy was then performed using a model system built in the present study that was composed of EGCG and a group of key catalytic site residues from the DNMT1. The PM3-derived binding energy values for the formation of complexes between EGCG and DNMT1 in the absence and presence of Mg2+ are summarized in Table 1. The ΔEbind value for the formation of EGCG-DNMT1 complex in the absence of Mg2+ was calculated to be –5.7 kcal/mol (Table 1 and Fig. 11A), indicating that the binding of EGCG to DNMT1 would be energetically favored even in the absence of Mg2+. This prediction is certainly consistent with our experimental observation that EGCG inhibited DNMTs in the absence of Mg2+. However, in the presence of Mg2+ and its coordinated water molecules, the ΔEbind value for the formation of DNMT1-EGCG complex was only at –8.1 kcal/mol when three coordinated water molecules were also present (Table 1 and Fig. 11B). The difference in the binding energy values in the presence and absence of Mg2+ clearly suggests that Mg2+ coordination in the catalytic site of DNMT1 would markedly enhance the binding force between EGCG and the enzyme (an increase of as much as ∼250-fold in the computed binding force) by stabilizing their binding interactions. This information is in accord with our experimental observations described in the present study which showed that the presence of Mg2+ at 2 mM markedly increased the apparent inhibition potency of EGCG by ∼10-fold (refer to Fig. 9D).
In the present study, we examined several representative catechol-containing dietary polyphenols (tea catechins and bioflavonoids) for their activity in inhibiting DNA methylation catalyzed by SssI DNMT and human DNMT1. Our data showed that all these dietary polyphenols are capable of inhibiting DNMT-mediated DNA methylation in vitro, but with varying potencies and efficacies. It is noteworthy that among all the dietary polyphenols tested in this study, EGCG was found to be a more potent and efficacious inhibitor of enzymatic DNA methylation in vitro. The IC50 values for its inhibition of SssI DNMT- and DNMT1-mediated DNA methylation were 0.21 and 0.47 μM, respectively (Fig. 3). Additional experiments using cultured MCF-7 and MDA-MB-231 breast cancer cells showed that treatment with nontoxic concentrations of EGCG or catechin partially inhibited the methylation status of the promoter regions of RARβ gene. It is noteworthy that although EGCG was more potent and efficacious than catechin, the differences in their potency and efficacy did not seem to be as pronounced as the results from the in vitro DNMT inhibition experiments. As explained below, this may be partly attributable to the fact that catechin is a far better substrate for COMT than EGCG (Zhu et al., 2000, 2001), which may contribute to an increased intracellular pool of SAH. In addition, EGCG usually has a relatively lower intracellular bioavailability than catechin because the former generally has a faster metabolic breakdown.
Our earlier studies have shown that many of the catechol-containing dietary polyphenols are excellent substrates for the O-methylation by human or rodent COMT (Zhu and Liehr, 1996; Zhu et al., 1996, 2000, 2001). From a mechanistic standpoint, we hypothesized that these dietary catechols, regardless of whether they can also exert a direct inhibition of the DNMTs, may cause an indirect inhibition of the DNMT-mediated DNA methylation through increased formation of SAH (a potent noncompetitive inhibitor of the DNMTs) during their methylation metabolism by COMT. This mechanistic explanation (illustrated in Fig. 1) is supported by our data showing that copresence of a complete COMT methylation system strongly enhanced the inhibition potency and efficacy of those dietary polyphenols that are suitable substrates for COMT.
It is interesting that the results of our present study also revealed that the strong, direct inhibition of DNA methylation by EGCG is independent of COMT but instead is dependent on the presence of Mg2+. This was particularly the case with respect to human DNMT1. It is noteworthy that a recent study reported that the apparent KI value for inhibition of the human DNMT-mediated DNA methylation by EGCG in vitro was ∼7 μM, with IC50 of ∼20 μM (Fang et al., 2003). This value was obtained by using a nuclear fraction (rich in human DNMT1) under reaction conditions in which Mg2+ was not present. Our data showed that the presence of physiologically relevant concentrations of Mg2+ (such as 2 mM) drastically enhanced the inhibitory potency of EGCG in human DNMT1-mediated DNA methylation reaction in vitro by decreasing its IC50 value from the previously reported 20 μM to ∼0.5 μM (a ∼40-fold increase in potency).
To shed light on the mechanism of DNMT1 inhibition at the molecular level, we also conducted computational modeling analyses to probe the interactions between EGCG and human DNMT1. Consistent with the recently proposed binding site model of human DNMT1 for EGCG (Fang et al., 2003), we identified Glu1265 as a key catalytic site residue that interacts with EGCG by forming H-bonds. However, our binding mode predicted that the p-OH group of EGCG's D-ring (the gallic acid moiety) would form an H-bond with Glu1265, whereas the model by Fang et al. (2003) suggested that the same amino acid residue would form an H-bond with one of the m-OH groups of the D-ring. We believe that the p-OH oxygen on the D-ring would be a better donor for the formation of H-bonding with Glu1265 than the corresponding m-OH oxygen because of its stronger acidity.
In addition, the data from our modeling studies suggest that Mg2+ is probably coordinated to Glu1265 at the catalytic core site. This conclusion was derived, in part, from a detailed comparison of the active-site conformations of the human DNMT1 with the COMT protein using multiple modeling tools, and we suggested that Glu1265 on human DNMT1 is probably equivalent to one of the two conserved acidic residues in COMT that are known to coordinate Mg2+ (Vidgren et al., 1994). Based on our proposed mode of interactions between EGCG and DNMT1 (Fig. 11), detailed binding energy calculations revealed that the formation of the EGCG-DNMT1 complex in the presence of Mg2+ would be much preferred over the complex without Mg2+ (ΔΔEbind = –2.4 kcal/mol; Table 1), and it was estimated that EGCG's inhibitory potency would be markedly enhanced in the presence of Mg2+ compared with that in the absence of Mg2+. This estimate, albeit indirectly derived from computations based on our predicted mode of interactions, is perfectly in agreement with our experimental observation described in the present study.
Here it is interesting to point out that myricetin, which has a pyrogallic acid moiety (similar to the gallic acid moiety of EGCG), also showed a stronger inhibition potency and efficacy than quercetin and fisetin in the presence of Mg2+. These experimental data are in agreement with the conclusion based on our modeling studies that the gallic acid moiety of EGCG plays a crucial role in its high-affinity interaction with the catalytic site of human DNMT1. By comparison, catechin, epicatechin, quercetin, and fisetin, which all lack a gallic/pyrogallic acid moiety, cannot form a similarly strong coordination with Mg2+ and Glu1265. Therefore, these four catechol-containing dietary polyphenols would have a weaker direct inhibition of the DNMTs as predicted according to our modeling data, but they could still exert a considerably strong indirect inhibition of the DNMTs through the formation of SAH during their metabolic methylation by COMT.
Hypermethylation of DNA is known to be a key epigenetic mechanism for the silencing of many genes, including those of tumor suppressors, DNA repair enzymes, and receptors (Jones and Takai, 2001; Lichtenstein and Kisseljova, 2001; Esteller, 2002; Jones and Baylin, 2002). Several tumor suppressor and receptor genes have been reported to be hypermethylated and transcriptionally silenced during the development of different types of cancers. It has been suggested that inhibition of DNMTs, along with inhibition of the histone deacetylase, may prevent the hypermethylation and silencing of these key genes, and this inhibition may contribute to the treatment of cancer and/or prevention of carcinogenesis. Considerable initial research efforts have already been made in this direction in recent years (Saunders et al., 1999; Bovenzi and Momparler, 2001; Yoshida et al., 2001; Zhou et al., 2002; Aparicio et al., 2003; Shaker et al., 2003; Shi et al., 2003). For instance, there are data showing a reduced intestinal tumorigenesis because of DNMT1 deficiency and use of 5-aza-2′-deoxycytidine in the MIN mice, which carry a mutated APC gene (Laird et al., 1995; Eads et al., 2002).
Animal studies have shown that oral administration of green tea could inhibit chemical carcinogenesis in some organs, and it is believed that multiple mechanisms are involved (Lu et al., 2001). However, it remains to be determined whether and to what extent this cancer-preventive activity is due to the inhibition of DNMTs by EGCG. Based on a recent report of the very high KI and IC50 values (6.89 and 20 μM, respectively) of EGCG, it was accordingly suggested that these effective concentrations are much higher than those achievable in the internal organs but might only be achievable in the oral cavity (in the saliva), stomach, esophagus, and intestines, where there is direct contact between EGCG and the epithelial cells. However, in light of the new results from our present study, showing lower IC50 values (at submicromolar concentrations) of EGCG for inhibition of human DNMT1-mediated DNA methylation in vitro and in cultured cancer cells, our finding may increase the possibility that tea polyphenols may be able to regulate DNA methylation status even in the internal organs.
Finally, it is of considerable interest to point out that a strong inhibition of DNA methylation by catechol-containing dietary polyphenols may also represent an antibacterial mechanism, especially in those bacteria that express high levels of COMT-like enzymes for the rapid methylation of dietary catechols and also for the formation of SAH. In Chinese traditional medicine, it is generally believed that drinking tea enhances oral hygiene. Modern experimental studies have also suggested that drinking tea would inhibit bacterial growth in the oral cavity. However, it remains to be determined whether this inhibition of bacterial growth is due in part to a strong inhibition of their DNA methylation.
In summary, tea catechins (catechin, epicatechin, and EGCG) and bioflavonoids (quercetin, fisetin, and myricetin) each inhibited SssI DNMT- and DNMT1-mediated DNA methylation. The IC50 values for the in vitro inhibition of human DNMT1 by catechin, epicatechin and various flavonoids ranged from 1.0 to 8.4 μM, but EGCG was a more potent inhibitor, with IC50 values at 0.21 to 0.47 μM. Kinetic analyses suggested that catechol-containing dietary polyphenol could inhibit DNA methylation through two mechanisms: direct inhibition of the DNMTs plus indirect inhibition of the enzymes through increased formation of SAH (a potent noncompetitive inhibitor of DNMTs) during COMT-mediated O-methylation of the catecholic dietary polyphenols. However, the strong inhibitory activity of EGCG for human DNMT1-mediated DNA methylation was largely independent of its own methylation and is essentially due to a Mg2+-dependent direct inhibition of the enzyme.
B.T.Z. is Frank and Josie P. Fletcher Professor of Pharmacology and an American Cancer Society Research Scholar.
Article, publication date, and citation information can be found at http://molpharm.aspetjournals.org.
ABBREVIATIONS: DNMT, DNA methyltransferase; SAM, S-adenosyl-l-methionine; COMT, catechol-O-methyltransferase; SAH, S-adenosyl-l-homocysteine; EGCG, (–)-epigallocatechin-3-O-gallate; TE, Tris-HCl/EDTA.
- Received October 15, 2004.
- Accepted July 21, 2005.
- The American Society for Pharmacology and Experimental Therapeutics