Preferential de novo methylation of cytosine residues in non-CpG sequences by a domains rearranged DNA methyltransferase from tobacco plants.

In plant DNA, cytosines in symmetric CpG and CpNpG (N is A, T, or C) are thought to be methylated by DNA methyltransferases, MET1 and CMT3, respectively. Cytosines in asymmetric CpNpN are also methylated, and genetic analysis has suggested the responsible enzyme to be domains rearranged methyltransferase (DRM). We cloned a tobacco cDNA, encoding a novel protein consisting of 608 amino acids, that resembled DRMs found in maize and Arabidopsis and designated this as NtDRM1. The protein could be shown to be localized exclusively in the nucleus and exhibit methylation activity toward unmethylated synthetic as well as native DNA samples upon expression in Sf9 insect cells. It also methylated hemimethylated DNA, but the activity was lower than that for unmethylated substrates. Methylation mapping of a 962-bp DNA, treated with NtDRM1 in vitro, directly demonstrated methylation of approximately 70% of the cytosines in methylatable CpNpN and CpNpG sequences but only 10% in CpG. Further analyses indicated that the enzyme apparently non-selectively methylates any cytosines except in CpG, regardless of the adjacent nucleotide at both 5' and 3' ends. Transcripts of NtDRM1 ubiquitously accumulated in all tissues and during the cell cycle in tobacco cultured BY2 cells. These results indicate that NtDRM1 is a de novo cytosine methyltransferase, which actively excludes CpG substrate.

Methylation of cytosine residues in DNA is enzymatically catalyzed by DNA methyltransferases, which transfer a methyl group from S-adenosyl-L-methionine (AdoMet) to the 5-position. In mammals, two types of DNA methyltransferase, which differ in DNA substrate preference, have been reported. Those belonging to the Dnmt1 group prefer cytosines in hemimethylated CpG sites, i.e. CpG with m 5 C in only one strand (1), and are considered to be associated with the DNA replication complex in vivo (6) functioning in maintenance of methylation patterns. Those belonging to the Dnmt3 group are reported to methylate cytosines in unmethylated CpG (7) and have been suggested to establish the methylation pattern during embryonic development (7).
In plants, genes encoding three types of DNA methyltransferases have been reported so far. For example, in Arabidopsis, MET1, chromomethylase (CMT), and domains rearranged methyltransfease (DRM) are distinct (8). MET1 is homologous to mammalian Dnmt1 (9), and its suppression results in a drastic reduction of global methylation in transgenic Arabidopsis (10,11). A similar reduction of global methylation and altered phenotypes was also observed in transgenic tobacco plants expressing an antisense tobacco MET1 (NtMET1) gene (12). Thus, MET1 has been suggested to function in maintenance of global genomic methylation in plants (10 -12). The other two types are unique to plants; putative proteins belonging to the CMT group have a chromodomain in catalytic motifs and are reported to be responsible for maintenance of cytosine methylation at CpNpG sites in, for example, retrotransposons (where N is A, T, or C) (13,14). Amino acid sequence analysis indicated that the DRM type has catalytic motifs, thus resembling mammalian de novo enzymes such as Dnmt3 (15), although they differ in possessing a characteristic rearrangement in catalytic motifs, between I-V and VI-X. A recent genetic analysis with mutant lines suggested that Arabidopsis DRMs might be responsible for the methylation of cytosines in CpNpG and asymmetric sequences of transgenes (16). Asymmetric cytosine methylation was also seen in epigenetically silenced loci, with DRMs further suggested to function in epigenetic gene silencing (15,16). However, no studies on biochemical properties of these proteins have been reported so far. In this article, we describe isolation of tobacco DRM. The enzyme expressed in insect cells could be shown to preferentially methylate cytosine residues in CpNpN and also CpNpG, providing concrete evidence for the predicted function of DRMs. BY2 cells was performed as described (12) with modification. After culture in a modified LS medium containing 5 g ml Ϫ1 aphidicolin (Wako, Tokyo) for 24 h, cells were collected and washed with 3% sucrose solution, transferred to fresh medium. and cultured further. They were then harvested by centrifugation and stained with 1% orcein to allow determination of the mitotic index.
Isolation of NtDRM1 cDNA-Initially, a DNA fragment encoding NtDRM1 was obtained by differential display of a cDNA population derived from wounded tobacco leaves (18). A full-length 2.4-kb cDNA of NtDRM1 was isolated by colony hybridization of a tobacco (N. tabacum) cDNA library using the NtDRM1 fragment as a probe and sequenced for both strands using a Big Dye terminator sequencing kit (Applied Biosystems, Foster City, CA). Motif prediction was performed using the on-line data bases PROSITE (tw.expasy.org/prosite/) and ISREC motif scan (hits.isb-sib.ch/cgi-bin/PFSCAN). Phylogenetic analysis was accomplished with ClustalW (clustalw.genome.ad.jp/).
Transformation of BY2 Cells and Histochemical Analysis-The Nt-DRM1 coding region (1824 bp) was fused in-frame via an engineered NcoI site to the N terminus of the green fluorescent protein (GFP) open reading frame. The construct was then cloned into the SmaI and EcoRI sites of pBI121 (19) and transformed into Agrobacterium tumefaciens strain EHA105 (20) by the heat-shock method. Following a 2-day cocultivation of the individually transformed A. tumefaciens with 5 ml of 4-day-old cultured tobacco BY-2 cells in the dark, transformed calli were selected using 100 g/ml kanamycin and 250 g/ml calbenicyrin over a 4-week period (21). DNA staining was performed with an aliquot of MS medium containing 1 mg/ml 4Ј,6-diamidino-2-phenylindole (DAPI). GFP fluorescence was observed using an AX70 fluorescence microscope equipped with UV-and B-excitation filters (Olympus, Tokyo) and standard fluorescein isothiocyanate filters. Digital images were captured with a cooled charge-coupled device camera (CoolSNAP-HQ, Photometrics, Tucson, AZ).
Expression and Purification of a GST::NtDRM1 Fusion Protein-Full-length NtDRM1 was fused to the C terminus of a glutathione S-transferase (GST) and expressed in insect cells using the GATEWAY cloning system (Invitrogen) according to the manufacturer's instructions. Amplification with a set of primers (forward, 5Ј-AAAAAGCAGG-CTTTATGGACAACAATCTTTCTGGAGAAGAC-3Ј; reverse, 5Ј-AGAA-AGCTGGGTACTAATGTTTATGTCTGGACATTATGGACTT-3Ј) was achieved, and the resulting fragment was subjected to a second PCR with another set of primers (forward, 5-GGGGACAAGTTTGTACAAA-AAAGCAGGC-3; reverse, 5Ј-GGGGACCACTTTGTACAAGAAAGCTG-GGT-3Ј). The final fragment was cloned into the pDEST20 vector plasmid and transformed into Escherichia coli together with bacumid (genome of a baculovirus) to allow in vivo recombination, and resulting single plaques were picked up and further propagated. Sf9 cells (6 ϫ 10 9 ), maintained in Grace's insect medium (Invitrogen) supplemented with 10% fetal bovine serum (Invitrogen) and 500 g/ml gentamycin, were infected with the recombinant baculovirus (500 l) and incubated at 27°C for 4 days. Cells from one dish were suspended in 1 ml of lysis buffer (20 mM Tris-HCl (pH 7.5), 5 mM EDTA, 1% Nonidet P-40, and 25% (v/v) glycerol, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 100 g/ml aprotinin) and sonicated twice for 10 s. GSTfused NtDRM1 with 77 kDa was purified through a glutathione-Sepharose column according to the manufacturer's instructions (Amersham Biosciences). Protein concentrations were estimated by the Bradford method, and immunoblot analysis was performed as described (22).
DNA Methyltransferase Assay-DNA methyltransferase activity was monitored by incorporation of the 3 H-labeled methyl group of AdoMet into recipient DNA substrate. For the initial assay, 2 g of poly(dI-dC)poly(dI-dC) (Sigma) was used as the substrate. For methylation of native DNA, samples from pGEX propagated in E. coli and Sf9 cells were used. For sequence specificity assays, the synthetic oligonucleotides 5Ј-ACGATCGTACGATCGTACGATCGT-3Ј (for CpG), 5Ј-ACTGC-AGTACTGCAGTACTGCAGT-3Ј (for CpNpG, where N is A or T), and 5Ј-AGCATGCTAGCATGCTAGCATGCT-3Ј (for CpNpN) were prepared. All these sequences are palindromic, which form duplexes. For hemimethylation analysis, synthetic 28-mer oligonucleotides containing five m 5 Cs and its complementary strand without m 5 C were independently prepared and annealed to form duplexes. The substrates for CpG were 5Ј-ATTCGATCGAATCGTATACGTACGTATT-3Ј and 3Ј-TAAGCTAGC-TTAGCATATGCATGCATAA-5Ј (m 5 C is underlined), and those for CpNpG were 5Ј-ATTCAGTCAGATCTGATCAGTACTGATT-3Ј and 3Ј-TAAGTCAGTCTAGACTAGTCATGACTAA-5Ј. The 100-l reaction mixture contained 20 mM MOPS-NaOH (pH 7.0), 5 mM EDTA, 200 g/ml bovine serum albumin, 25% (v/v) glycerol, 1 mM dithiothreitol, 100 g/ml RNase A. 2 M AdoMet (methyl-3 H, specific activity 307.1 GBq/mmol) (PerkinElmer Life Sciences) and appropriate amounts of substrate DNA and enzyme preparation were incubated at 37°C for the indicated period. To the reaction mixture was added 1 mg/ml of proteinase K (Invitrogen) in 500 l of proteinase K-SDS buffer, containing 1% SDS, 2 mM EDTA, 125 mM NaCl, and 0.5 mg/ml salmon sperm DNA. After further incubation at 50°C for 1 h, substrate DNA was extracted by phenol/chloroform treatment, precipitated with ethanol, and spotted onto DEAE paper, which was washed with 0.5 M sodium phosphate, dried, and assessed for radioactivity. Under these experimental conditions, the background level of 3 H radioactivity was ϳ50 cpm per assay, indicating the quenching to be suppressed to a minimum. Kinetic parameters were calculated from the Michaelis-Menten equation with the Anemona program (23).
Quantitative Analysis of m 5 C-Identification of m 5 C by HPLC was performed using substrate DNA isolated from Sf9 cells. A 200-l reaction mixture, containing 100 g of DNA, 2 mM AdoMet (Sigma), and the appropriate amounts of NtDRM1 in the methylation buffer as described above, was incubated at 37°C for 16 h and then further with 110 ng of RNase A (Nacalai Tesque, Kyoto, Japan) at 37°C for 2 h. DNA was extracted with phenol/chloroform, precipitated with ethanol, and digested with 2 units of nuclease P1 (Sigma) in 100-l buffer containing 3 mM sodium acetate (pH 5.4) and 0.5 mM ZnSO 4 at 37°C for 20 h. The resultant nucleotides were dephosphorylated with 20 units of calf intestine alkaline phosphatase (Takara, Otsu, Japan) at 37°C for 2 h. Samples were then fractionated by ultrafiltration (Ultrafree-MC PL-10 microcentrifuge tubes, Millipore, Bedford, MA), and the permeate was injected into a Supelcosil LC-18-S column (Supelco, Bellafonte, PA). Separation was performed with a 2.5-20% methanol gradient in the presence of 50 mM KH 2 PO 4 (pH 4.3).
DNA Methylation Mapping-Substrate DNA was prepared by PCRamplifying a 962-bp fragment of pGEX-4T-1 (positions 4521-513, containing the position 1) then methylated in vitro in a reaction mixture containing 25 g of DNA, 2 mM AdoMet (Sigma), and 93.75 g of NtDRM1 in methylation buffer, as described above, at 37°C for 16 h. After phenol extraction and precipitation with ethanol, DNA was subjected to bisulfite modification (24). The method was developed to identify m 5 C in arbitrary sequences, based on the resistance of m 5 C to bisulfite, which changes C into U. After bisulfite treatment, the modified DNA can be amplified by PCR, cloned, and directly sequenced, the C and m 5 C in the initial sequence being replaced with T and C, respectively. Experimentally, modified DNA was subjected to first PCR using ExTaq TM enzyme (Takara) and specific forward, 5Ј-GTTGTGTAGTTT-GAATGGTGAATGG-3Ј, and reverse, 5Ј-CAACCACCCAACATATTAT-ACTTATCAAC-3Ј primers. Both were designed for predicted sequences after modification, in which C was assumed to be converted into T in the former and G into A in the latter. The resulting product was subjected to subsequent second PCR with the same forward primer and another reverse primer, 5Ј-CACAAAACCCTTAATTTTCCAATAACC-3Ј (G was replaced with A). Amplified 594-bp DNA samples were ligated to the pGEM-T easy vector TM (Promega, Madison, WI) and cloned in DH5␣ (Stratagene, La Jolla, CA). Sequences were determined with an ABI PRISM BigDye TM Terminator DNA sequencing kit and a 3100 Genetic Analyzer automated sequencer (Applied Biosystems). Fifteen clones were sequenced, and the average methylation at each site was calculated.
DNA and RNA Isolation and Gel Blot Hybridization-Genomic DNA was extracted by the cetyl-trimethylammonium bromide method (25). Total RNA was isolated from indicated tissues or from BY2 cells by the acid guanidinium/phenol/chloroform method (26), and hybridization analyses were performed as described (27) with probes synthesized with a pair of primers specific for each gene.

RESULTS
Identification of NtDRM1-During screening for genes whose transcripts accumulate in the early stage after wounding of tobacco leaves by the modified differential display, a particular fragment of 1.6 kb was identified. Although the clone was found later not necessarily to be specific to the wound response by Northern hybridization, it was further characterized because homology searches indicated resemblance to genes for DNA methyltransferases. Subsequent screening of a tobacco cDNA library yielded a full length of 2,540-bp cDNA, encoding a protein of 608 amino acids. Sequence analysis showed it to contain at least six highly conserved motifs found in DNA methyltransferases, but their order was unusual. In contrast to the majority of eukaryotic DNA methyltransferases, having conserved motifs that are arranged in the order of I-IV-VIII-X in the C terminus, the present putative protein was found to possess motifs in the rearranged order of VI-VIII-IX-X-I-IV (Fig. 1A). A homology search indicated high homology to DRMs from Arabidopsis (DRM1 and DRM2) and maize (Zmet3) (15) (Fig. 1B). Consequently, we concluded that the isolated gene encodes a DRM, and we designated it as NtDRM1 (N. tabacum domains rearranged methyltransferase 1). In the Nterminal region, two domains (amino acid positions 61-97 and 166 -204) were found that resembled the ubiquitin association (UBA) domains of human p62 and yeast RAD23 (Fig. 1A). UBA domains, considered to function in protein-protein interactions, have also been identified in DRMs of Arabidopsis and maize (15). A nuclear localization signal is present at amino acid positions 234 -237 (Fig. 1A). A phylogenetic tree generated with the conserved catalytic motifs I-IV indicated that DRMs are more closely related to the de novo methyltransferases from mammals (Dnmt3a and Dnmt3b) than to other plant DNA methyltransferases (Fig. 1B). A similar tree was also obtained when motifs VI-X were aligned (data not shown).
Subcellular Localization-To identify the cellular localiza-tion, a reporter gene encoding GFP was selected, and a plasmid expressing the NtDRM1-GFP fusion protein under the control of the cauliflower mosaic virus (CaMV) 35S promoter was constructed. As the control, a plasmid expressing only GFP was used. Each plasmid was introduced into onion epidermis cells by biolistic bombardment, and interference contrast images for whole cell structures (Fig. 2, A and C) and GFP fluorescence for localization of GFP proteins (Fig. 2, B and D) were examined. Although CaMV 35S::GFP control construct showed GFP signals in both cytoplasm and nucleus (Fig. 2B), with CaMV 35S::NtDRM1-GFP they were predominantly in the nucleus (Fig. 2D). To confirm the nuclear localization of NtDRM1 in planta, tobacco BY-2 cells were stably transformed with the same plasmids, and GFP fluorescence was detected in living cells (Fig. 2, E-H). As in onion cells, fluorescence of NtDRM1-GFP was observed only in the nucleus (Fig. 2, F and H). This pattern was identical with that of DAPI, which stains the nucleus (Fig. 2G). A nuclear localization of NtDRM1 was ubiquitously observed in transformed cells (Fig. 2, E and F). It is evident from these results that NtDRM1 is exclusively localized in nuclei.
Purification and Enzymatic Analysis-NtDRM1 cDNA was fused in-frame to the GST gene and expressed using the baculovirus-mediated insect cell (Sf9) expression system, which lacks endogenous DNA methyltransferase activity. After three cycles of propagation of infected Sf9 cells, crude proteins were recovered and further purified through a glutathione-Sepharose column. SDS-PAGE indicated the product to be almost homogeneous (Fig. 3A), and this was confirmed to be the recombinant protein by immunoblot staining with anti-GST antibodies (Fig. 3A). By using the crude protein preparation, methylation activity was preliminarily assayed by measuring 3 H-labeled methyl group transfer from AdoMet into the synthetic oligonucleotide, poly(dI-dC)-poly (dI-dC). As positive and negative controls, mammalian Dnmt3a, a de novo DNA methyltransferase, and bacterial ␤-glucuronidase (GUS), respectively, were similarly expressed and subjected to the same assay. The amounts of enzyme protein in the crude samples were estimated by densitometric quantitation of immunostaining after gel electrophoresis, and activity was normalized. Methyl group incorporation was clear in the reaction mixture containing GST-NtDRM1 or GST-Dnmt3a proteins, whereas it was undetectable with the GST-GUS preparation (Fig. 3B). The activity of NtDRM1 was 10-fold higher than that of Dnmt3a, suggesting an identity as a de novo DNA methyltransferase. The optimal pH for the reaction was found to be 7.0 (data not shown). To obtain kinetic information, the activity was further assayed using purified GST-NtDRM1 and varying concentrations of poly(dG-dC)-poly(dG-dC) and also poly(dI-dC)-poly(dI-dC), because the former frequently forms multistranded structures and even nonspecific aggregation, whereas the latter does not (Fig. 3C). Methylation activity increased in a substrate-dependent manner and poly(dI-dC) proved to be a better substrate for the enzyme than poly(dG-dC), probably reflecting the structural features. The apparent K m values for poly(dI-dC) and poly(dG-dC) were 2.58 Ϯ 0.03 and 6.36 Ϯ 0.53 M (in methylatable cytosine mononucleotides), respectively, and the V max values were calculated to be 718.9 Ϯ 10.5 and 312.0 Ϯ 58.9 fmol/min/nmol protein, respectively (Fig. 3C). The ratio of methyl incorporation to the substrate cytosines in poly(dI-dC) was estimated to be around 1/10 5 under the experimental conditions with low concentrations of enzyme and AdoMet and a reaction time of only 30 min. The results indicate de novo cytosine methylation by NtDRM1.
Sequence Specificity-In order to determine the sequence specificity for NtDRM1-mediated methylation, three synthetic oligonucleotides, with different sequences but the same base composition, were prepared. Each palindromic 24-mer contained 6 sites for CpG, CpNpG, or CpNpN, respectively, where N is A or T. The highest methylation was seen with the CpNpN substrate, 3 H incorporation increasing almost linearly with incubation period up to 1.5 h, whereas methyl transfer efficiency was lower with CpNpG than CpNpN and almost null with the CpG substrate (data not shown). The effect of substrate concentration on enzymatic activity was then examined with a fixed concentration of the enzyme and varying concentrations of DNA (Fig. 4A). The best substrate was CpNpN, with a substrate-dependent hyperbolic increase in the reaction velocity (Fig. 4A). The apparent K m and V max values were 45.9 Ϯ 3.04 nM and 224.4 Ϯ 4.12 fmol/min/nmol protein (based on the methylatable cytosine mononucleotides), respectively. The molar ratio of methyl group incorporated to the substrate was ϳ1/10 3 . The transfer rate of the methyl group into CpNpG was 2/3 of that into CpNpN and almost zero into CpG (Fig. 4A). The apparent K m and V max values were 78.3 Ϯ 5.26 nM and 192.2 Ϯ 4.62 fmol/min/nmol protein (based on the methylatable cytosine mononucleotides), respectively. These kinetic experiments suggested NtDRM1 to preferentially methylate cytosines in non-CpG sequences.
Effects of hemimethylation were then analyzed. A synthetic 28-mer oligonucleotide containing either CpG or CpNpG was prepared, in which all cytosines were substituted with m 5 C. A non-methylated complementary strand was also generated and annealed to form a double-stranded substrate. Methyl transfer activity was then assayed with a fixed concentration of the enzyme and varying concentrations of the substrate (Fig. 4B). These kinetic analyses showed the enzyme to be active on the hemimethylated CpNpG at a lower velocity than on the unmethylated CpNpG substrate. By taking account of the fact that the number of available cytosines in the former is only half that in the latter, the K m and V max values were calculated to be 21.  (Fig. 4B). The results suggested that the enzyme does not recognize the hemimethylated state of CpG and CpNpG and . Tobacco BY2 cultured cells were transformed with pNtDRM1-GFP, and stable transformants were observed by interference contrast (E) or by epifluorescence for GFP (F). Single cells were similarly observed for GFP fluorescence (H) and stained with DAPI to identify the position of nuclei (G). therefore confirmed its non-maintenance, de novo properties.
Direct Methylation Mapping-The sequence specificity was directly determined by methylation mapping. Native DNA was intensively methylated in vitro with excess AdoMet and the enzyme, modified by the bisulfite method, amplified with PCR, cloned, and directly sequenced. The conversion efficiency of cytosine (C) into uracil (U) was directly estimated by aligning sequences and found to be nearly complete as all cytosines in the untreated sequence were converted into thymine (T) (Fig. 5A). The methylation frequencies were estimated for 15 clones, as-signed to each cytosine, and expressed as percentages (Fig. 5B). The enzyme showed the highest methylation of cytosines in CpNpN followed by CpNpG and least in CpG. The percentages of methylated sites in the total in 15 clones were ϳ75% for CpNpN and 60% for CpNpG, in sharp contrast with the 10% for CpG (Table I). When nucleotide triplets were classified, and the nearest neighbor was taken into account, the average methylation rates for CpT, CpA, and CpC were 87, 75, and 70%, respectively (Table II). It appears that the third nucleotide, including guanine, does not appreciably influence the methylation activity, although T located at the 3Ј end of CpC and CpT reduced the frequency (Table II). The nucleotide located at the 5Ј end of target cytosines showed no apparent effects on site specificity (Table  III). These observations suggest that NtDRM1 essentially recognizes and methylates all cytosines, although some particular combinations, such as CpG and CpCpT, appear to be less favorable. H]AdoMet, and the indicated amount of synthetic oligonucleotides was incubated for 30 min and processed as described above. Experiments were repeated three times, and mean values were estimated with standard deviations. B, effects of hemimethylated substrate. The hemimethylated duplex sequences of 28-mer oligonucleotides used in this assay were synthesized as described in the text. The number of methylatable cytosines was 5 in one strand, and thus contained 5 and 10 sites in each duplex of hemimethylated and unmethylated substrate, respectively. Substrates were unmethylated CpG (umCpG, open circles), hemimethylated CpG (hmCpG, closed circles), unmethylated CpNpG (umCpNpG, open squares), and hemimethylated CpNpG (hmCpNpG, closed squares). Assays were performed as described above.
In Vivo Methylation-To examine whether or not NtDRM1 functions in vivo, native DNA from Sf9 was analyzed by HPLC (Fig. 6). Samples from Sf9 cells expressing NtDRM1 were isolated, digested with nuclease P1, dephosphorylated with alkaline phosphatase, and assayed by HPLC. As the control, DNA from wild-type Sf9 cells was used. The positive control was prepared by methylating DNA in vitro by NtDRM1, and the negative control was intact Sf9 DNA. HPLC fractionation of the resulting nucleosides showed that DNA samples prepared from trans-formed cells yielded m 5 C at 1.8% of total cytosines, showing a similar elution profile with the control in vitro methylated DNA containing m 5 C, at 2.7% of total cytosines (Fig. 6, B and C). The low amount of m 5 C from transformed cells may be due to the expression system, in which Sf9 cells are dying when the protein is produced. However, since intact DNA from untransformed cells did not show any m 5 C (Fig. 6D), the results suggest Nt-DRM1 to be active as a de novo cytosine methyltransferase in vivo.  II Methylation frequency in triplets containing m 5 C at 5Ј end The sums of total m 5 C numbers in the total methylatable cytosines for the indicated triplets in 15 clones are presented as numbers observed and as percentages, as described in the legend for Table I

TABLE III Methylation frequency in triplets containing m 5 C at 3Ј end
The sums of total m 5 C numbers in the total methylatable cytosines for the indicated triplets in 15 clones are presented as numbers observed and as percentages, as described in the legend for Table I Genomic Organization and Transcript Accumulation-Genomic DNA of tobacco (N. tabacum) was digested with appropriate restriction enzymes and hybridized with a specific probe prepared from the 3Ј-untranslated region of NtDRM1 (positions 2064 -2493). Under high stringency conditions, the probe showed four distinct signals after digestion with BamHI, EcoRI, and HindIII, respectively (Fig. 7A). Because none of these restriction sites was present in the probe sequence, and since N. tabacum is an amphidiploid, it is conceivable that NtDRM1 forms a multigene family existing probably as pairs in each chromosome set originating from its ancestor lines, Nicotiana sylvestris and Nicotiana tomentosiformis. RNA hybridization analyses indicated Nt-DRM1 transcripts to accumulate ubiquitously in leaves, stems, flowers, and roots (Fig. 7B). Levels were also high in all floral organs except pistils (Fig. 7B). During the cell cycle of synchronously cultured BY2 cells, NtDRM1 transcripts accumulated throughout (Fig. 7C), in marked contrast to those for NtMET1 encoding a maintenance methyltransferase, expressed predominantly in the S-phase (Fig. 7C). DISCUSSION Methylation of DNA is characterized by two distinct features in eukaryotes: maintenance of the preexisting methylation patterns and methylation of previously unmethylated sites, the responsible enzymes being referred to as maintenance and de novo methyltransferases, respectively. In plants, proteins belonging to the former type have been biochemically characterized in several species (8), but examples of the latter have not been isolated so far. The presently identified NtDRM1 catalyzed methylation of both synthetic oligonucleotides and native DNA in vitro. Hemimethylated substrates were also methylated at a low efficiency. When expressed in insect cells, it methylated host DNA in vivo. It was thus concluded that Nt-FIG. 6. Identification of methylation products. DNA samples were digested to mononucleosides, dephosphorylated, and subjected to HPLC analysis. A, elution profile of standard authentic cytosine (open arrowhead) and m 5 C (closed arrowhead). The retention times of cytosine and m 5 C were 4.6 and 8.7 min, respectively. B, elution profile of in vitro methylated DNA from wild-type Sf9 cells. A 100-g DNA sample was methylated with NtDRM1 under the standard condition and then digested to mononucleosides. C, elution profile of native DNA from NtDRM1-transformed Sf9 cells. A 100-g of DNA sample was directly digested to mononucleosides. D, elution profile of native DNA from wild-type Sf9 cells. A 100-g of DNA sample was directly digested to mononucleosides.
FIG. 7. Genomic organization and tissue-specific expression of NtDRM1. A, Southern hybridization analysis. A 10-g aliquot of total DNA was digested with the indicated restriction enzymes and subjected to DNA hybridization with the NtDRM1 probe. B, tissue-and organspecific transcript accumulation. Total RNA was extracted from indicated tissues or flower organs, and a 10-g aliquot was fractionated by agarose gel electrophoresis, transferred to a nylon membrane, and subjected to hybridization with the NtDRM1 probe. Flower organs were ovule (Ov), sepals (Se), pistils (Pi), stamen (St), and petal (Pe). As an internal standard, ribosomal protein cDNA (rrm18) was used (bottom panel). C, cell cycle-dependent expression of NtDRM1 and NtMET1 in synchronized tobacco BY2 cells. Membranes bearing 15 g of total RNA per lane were successively probed with the NtDRM1-and NtMET1specific probes. As the internal standard, actin cDNA (Actin) was used (bottom panel). The relative levels of transcripts were quantified based on densitometric measurement of each signal, normalized to actin signals. The indicated phase of the cell cycle was estimated from the change of mitotic index after aphidicolin release. DRM1 is a de novo methyltransferase, to our knowledge, the first clear example of this family in plants.
One of the notable features of NtDRM1 is its high specificity toward cytosines in non-CpG sites. Methylation of plant DNA occurs at cytosine residues located in symmetric CpG and CpNpG, and also in asymmetric CpNpN sequences (8). The methylation frequency for these sequences has been estimated to be over 70, ϳ50, and ϳ8%, respectively, in maize DNA (28). This is a clear contrast to the case with mammalian DNA, which is almost exclusively methylated in symmetric CpG, at a frequency at more than 80% in many tissues (29). The methylation of CpG in mammals and plants is mainly mediated by DNA methyltransferases of maintenance type, which, by recognizing m 5 CpG in the mother strand, methylates opposite CpG in newly replicated daughter strands after semiconservative DNA replication. This essentially results in the maintenance of the same methylation pattern throughout cell division (30). In contrast, the methylation pattern at asymmetric cytosines is usually not maintained after DNA replication. For it to occur, an asymmetric de novo methyltransferase would be required, and no clear evidence had been presented for the existence of such enzymes until the recent reports describing Dnmt3a to methylate not only CpG but also asymmetric cytosines (31), and DIM-2 of Neurospora crassa to be responsible for cytosine methylation at symmetrical and asymmetrical sites (32). In plants, no enzyme has been assigned so far to catalyze cytosine methylation in asymmetric sites. Reverse genetic analyses have suggested that methylation of CpNpG and CpNpN is catalyzed by CMTs and DRMs, respectively, suggesting a distinct role among enzymes (16,33). Under certain circumstances, DRM was also shown to be responsible for CpNpG methylation at some loci, suggesting that DRM and CMT may act in a partially redundant and locus-specific manner to control methylation of CpNpG and CpNpN (34). Our present findings are essentially consistent with this view, and further provide biochemical evidence that, at least in tobacco plants, NtDRM1 is generally responsible for non-CpG methylation.
The molecular basis for preferential methylation of non-CpG sites is worthy of consideration. Results with direct methylation mapping suggest that NtDRM1 does not necessarily favor CpNpN and its CpNpG subgroup but rather evades CpG sequences, resulting in an apparent preference. Such rejection of a particular sequence has never been reported, but a recognition mechanism for m 5 C by methylated DNA binding domain proteins is suggestive of a role of specific amino acids. For example, two guanines in the symmetric m 5 CpG doublet have been shown to interact specifically with Arg (Arg-133 and Arg-111) in MeCP2 protein (35). It is of interest to examine whether some amino acids in NtDRM1 repel CpG in such a way as to hinder mutual interaction. Another possibility is that the rearranged catalytic domain is responsible for non-CpG recognition. We think that this is probable, since only plants exhibit methylation in non-CpG sites and have DRMs. Identification of a cytosine-recognition region in NtDRM1 would substantiate this speculation.
The biological significance of selective methylation at asymmetric cytosines is not completely clear but may be of particular importance for establishment of cell-specific methylation patterns (36). Asymmetric cytosine methylation would provide an opportunity for individual cells to establish independent methylation patterns in response to environmental conditions. An example is de novo cytosine methylation at CpNpN sites during RNA-dependent DNA methylation, which has been inferred to function in cosuppression of plant genes upon introduction of foreign DNA (37,38). It was reported recently that the demethylating enzyme, m 5 C-specific glycosylase, regulates gene expression (39,40). In this case, methylation-mediated suppression is alleviated by demethylation, thereby reversibly controlling gene expression in the cell. A similar cycle of methylation-demethylation of genomic DNA can also be achieved through asymmetric methylation and subsequent passive demethylation through cell division.
It must be mentioned that the putative amino acid sequences of DRMs so far identified from maize, Arabidopsis and soybean indicated the presence of a UBA domain (15), shown to be a common region for protein-protein interactions (41). Because NtDRM1 also possesses a UBA domain, and is constitutively expressed throughout the cell cycle and in all tissues, it is conceivable that it forms complexes with multiple proteins that contribute to chromatin structure and thereby methylates particular regions of DNA in a de novo fashion.