Biochemical Journal

Research article

Characterization of DNA-binding activity in the N-terminal domain of the DNA methyltransferase Dnmt3a

Isao Suetake , Yuichi Mishima , Hironobu Kimura , Young-Ho Lee , Yuji Goto , Hideyuki Takeshima , Takahisa Ikegami , Shoji Tajima

Abstract

The Dnmt3a gene, which encodes de novo-type DNA methyltransferase, encodes two isoforms, full-length Dnmt3a and Dnmt3a2, which lacks the N-terminal 219 amino acid residues. We found that Dnmt3a showed higher DNA-binding and DNA-methylation activities than Dnmt3a2. The N-terminal sequence from residues 1 to 211 was able to bind to DNA, but could not distinguish methylated and unmethylated CpG. Its binding to DNA was inhibited by a major groove binder. Four basic amino acid residues, Lys51, Lys53, Arg177 and Arg179, in the N-terminal region were crucial for the DNA-binding activity. The ectopically expressed N-terminal sequence (residues 1–211) was localized in nuclei, whereas that harbouring mutations at the four basic amino acid residues was also detected in the cytoplasm. The DNA-methylation activity of Dnmt3a with the mutations was suppressed under physiological salt conditions, which is similar that of Dnmt3a2. In addition, ectopically expressed Dnmt3a with mutations, as well as Dnmt3a2, could not be retained efficiently in nuclei on salt extraction. We conclude that the DNA-binding activity of the N-terminal domain contributes to the DNA-methyltransferase activity via anchoring of the whole molecule to DNA under physiological salt conditions.

  • DNA-binding activity
  • DNA methyltransferase
  • DNA methylation
  • subcellular localization

INTRODUCTION

The methylation of cytosine bases is a covalent DNA modification that has a clear function through its association with transcriptional silencing [1]. DNA methylation is essential for proper development [2], genome stability, X-chromosome inactivation, genomic imprinting and silencing of retrotransposons [35]. Dnmt (DNA methyltransferase) catalyses the transfer of a methyl group to the fifth position of the cytosine bases in DNA. Mammalian genomes carry three distinct active Dnmt genes. Two of these three genes, Dnm3a and Dnmt3b, have been reported to show activity towards unmethylated DNA that is comparable with that towards hemi-methylated DNA in vitro [6,7], and are responsible for creating DNA-methylation patterns during embryogenesis and gametogenesis [2,8]. The expression profiles of Dnmt3a and Dnmt3b during embryogenesis and cell differentiation are different. This indicates that the functions of the two methyltransferases are distinct, although they partly complement each others functions [2,9]. Dnmt3b specifically methylates pericentromeric minor satellites, and its dysfunction is reported to be the cause of ICF (immunodeficiency, centromere instability and facial anomalies) syndrome [2,10]. Recently, it was reported that mutation of Dnmt3a was found at a high frequency among patients with acute myeloid leukaemia [11].

The Dnmt3a gene encodes two isoforms, Dnmt3a and Dnmt3a2. Dnmt3a2 is transcribed from an intronic promoter and thus lacks the N-terminal 219 amino acid residues of the full-length Dnmt3a [12]. Dnmt3a is ubiquitously expressed in embryonic mesenchymal cells after day 10.5 [2,9]. On the other hand, Dnmt3a2 is responsible for global DNA methylation, including that of imprinted genes, in male germ cells on embryonic days 16–18 [8,13]. Dnmt3a2 is predominantly expressed in germline stem cells and embryonic stem cells [1315]. Ectopically expressed Dnmt3a has been reported to be located in heterochromatin [12,16] and Dnmt3a2 in euchromatin [12]. When the DNA-methylation activities of recombinant Dnmt3a and Dnmt3a2 towards naked DNA are compared, Dnmt3a shows higher specific activity and its activity is more resistant to moderate salt concentrations than that of Dnmt3a2 [17]. The inhibitory effect of salt on the DNA-methylation activity of Dnmt3a2 is partially reversed in the presence of Dnmt3L [17]. Thus Dnmt3L, a member of the Dnmt3 family lacking DNA-methylation activity, is essential for global DNA methylation in germ cells [1821]. The differences in the enzyme activities between Dnmt3a and Dnmt3a2 led to the idea that these enzymes are under different regulation to determine the site-specific DNA methylation of the genome.

In the present study we focused on characterization of the properties of the N-terminal region that only exists in full-length Dnmt3a. The N-terminal domain showed strong DNA-binding activity. Four basic amino acid residues were crucial for the DNA-binding activity. The salt-resistant DNA-methylation activity of Dnmt3a depends on the DNA-binding activity of the N-terminal domain.

MATERIALS AND METHODS

Recombinant proteins, expression and purification

Truncated and mutated cDNAs of Dnmt3a, i. e. Dnmt3a-(1–211) with the mutations K51A and K53A (N-mut), R177A and R179A (C-mut), and K51A, K53A, R177A and R179A (D-mut), were PCR-amplified with recombinant Taq DNA polymerase (Toyobo) with the primers harbouring the mutations, and then subcloned into pGEX6P1 (GE Healthcare) for expression in Escherichia coli. The sequence encoding the MBD (methylated DNA-binding domain) (residues 1–75) of MBD1 [MBD1-(1–75)] was provided by Dr Mitsuyoshi Nakao (Institute of Molecular Embryology and Genetics, Kumamoto University, Kumamoto, Japan), and was subcloned into pET30a with a GST (glutathione transferase) tag added between the histidine tag and MBD1-(1–75) in-frame. The expression plasmid encoding the N-terminal DNA-binding sequence (residues 119–197) of Dnmt1 [Dnmt1-(119–197)] has been described previously [22]. The DNA sequences of all of the plasmids constructed in the present study were confirmed using the dideoxy method.

Since the molecular masses of the full-length Dnmt3a, its mutants and Dnmt3a2 are large, it was difficult to highly purify the recombinant proteins utilizing E. coli as a host [7]. For this, all of the recombinant full-length Dnmt3as were expressed in Sf9 cells with a histidine tag at their N-termini, and purified as described previously [23]. As for the truncated forms of Dnmt3a, including the mutants, the proteins were successfully expressed in E. coli and then purified. The expression of the GST-tagged N-terminal region of Dnmt3a was induced in BL21-CodonPlus (DE3)-R1L cells at 18 °C with 0.5 mM IPTG (isopropyl β-D-thiogalactopyranoside). Harvested cells were suspended in 20 mM Tris/HCl (pH 7.6) containing 0.04% Triton X-100, 0.1% PI (protease inhibitor cocktail, Nacalai Tesque) and 1 mM MgCl2. The cell suspension was sonicated at 40 W for 5 s, and then 125 units of benzonase (Sigma) were added, followed by incubation at room temperature (22 °C) for 90 min. After incubation, NaCl and Triton X-100 were added to the solution to final concentrations of 0.25 M and 1% (w/v) respectively. The mixture was centrifuged at 10000 g for 10 min, and the supernatant fraction was loaded on to a glutathione–Sepharose column (GE Healthcare), and eluted with a solubilization buffer supplemented with 10 mM GSH.

The N-terminal region of Dnmt1 was purified as described by Suetake et al. [22]. Histidine- and GST-tagged MBD1-(1–75) was expressed in BL21-CodonPlus (DE3)-R1L cells at 23 °C with 0.5 mM IPTG. The protein-expressing cells were suspended in a buffer comprising 50 mM Tris/HCl (pH 7.5), 0.5 M NaCl, 0.1% Triton-X100, 0.1% PI and 1 mM DTT (dithiothreitol). The suspension was sonicated at 40 W for 5 s and then centrifuged at 10000 g for 5 min. The resulting supernatant was loaded on to a Ni2+-chelating column pre-equilibrated with a buffer comprising 20 mM Pipes (pH 6.2), 10% (w/v) glycerol, 0.35 M NaCl, 0.01% Nikkol (Nikkoh Chemicals), 15 mM 2-mercaptoethanol, 0.001% PI and 20 mM imidazole, washed with the same buffer, and then eluted with the washing buffer supplemented with 0.4 M imidazole. The eluate was loaded on to a glutathione–Sepharose column equilibrated with a buffer comprising 50 mM Tris/HCl (pH 8.0), 1 M NaCl and 1 mM DTT, and then eluted with 10 mM GSH. The eluate was dialysed against 20 mM Tris/HCl (pH 8.0), 0.1 M NaCl and 1 mM DTT. Protein concentrations were determined using a BCA (bicinchoninic acid) assay kit (Pierce).

Purified recombinant proteins were electrophoresed using SDS/PAGE (10% or 12% gels), and the protein bands were stained with CBB (Coomassie Brilliant Blue R-250).

DNA-binding assay

DNA fragments, i.e. MMTV145, 601.2×2 (two tandem repeats of 601.2) and 5S(RR), were prepared as described previously [24,25]. Fully methylated, hemimethylated and unmethylated oligonucleotides of 42 bp [22,26] and 26AT [22] were also used.

For gel-shift assays, the indicated amounts of protein and DNA were mixed in 20 mM Tris/HCl (pH 7.6) containing 5 mM EDTA, 2.7 M glycerol, 0.2 mM DTT and 0.2 mM PMSF. The reaction mixtures were incubated at 37 °C for 30 min, and then electrophoresed in a 0.7% agarose gel or 5% acrylamide gel in a 1×TBE buffer (45 mM Tris/borate and 1 mM EDTA) [24]. When the effects of DNA-binding chemicals on the protein binding were examined, DNA and the indicated amounts of chemicals were pre-mixed and incubated on ice for 10 min. DNA bands were visualized by staining with GelGreen (Biotium) and determined in a FLA-5000 image analyser (Fuji Photo Film). The bound and free MMTV145 DNA bands stained with GelGreen were quantified using Imagegause software (Fuji Photo Film).

For filter-binding assays, the 5′ end of the 5S(RR) DNA was labelled with [γ-32P]ATP (185 GBq/mmol; GE Healthcare) and T4 polynucleotide kinase. The labelled DNA was separated from unincorporated nucleotides by chromatography on a Sephadex G-50 column (GE Healthcare). Labelled DNA (39 nM) and purified Dnmt3a or Dnmt3a2 (40 nM) in 25 μl of buffer identical with that used for the gel-shift assays were incubated at 37 °C for 1 h. After incubation, the reaction mixtures were spotted on to a nitrocellulose filter (Pall) that had been pre-washed with 20 mM Tris/HCl (pH 7.4) containing 20 mM NaCl, and then washed with the identical buffer supplemented with 20% (w/v) glycerol. The dried filter was exposed to an imaging plate (Fuji Photo Film), and then the radioactivity adsorbed on to the filter was determined with a BAS2000 instrument (Fuji Photo Film).

Determination of DNA-methylation activity

DNA-methylation activity was determined as described previously [23]. Briefly, 25 μl of a reaction mixture comprising 5 mM EDTA, 2.7 M glycerol, 0.2 mM DTT, 0.2 mM PMSF, 20 mM Tris/HCl (pH 7.6), 100 ng of poly(dG-dC), 133 pmol of [3H]SAM (S-adenosylmethionine; GE Healthcare) and the indicated amount of Dnmt3a, Dnmt3a2 or Dnmt3a-D-mut was incubated at the indicated temperature for 1 h. At the end of the incubation, the reaction mixture was added to 1.5 mM non-radioactive SAM. The incorporated radioactivity was determined.

Cell culture and transfection

HeLa cells were cultured in MEM (minimal essential medium, Nissui) supplemented with 3% (v/v) FBS (fetal bovine serum), 100 units/ml penicillin and 100 μg/ml streptomycin. HEK-293T [HEK (human embryonic kidney)-293 cells expressing the large T-antigen of SV40 (simian virus 40)] cells were maintained in DMEM (Dulbecco's modified Eagle's medium, Nissui) supplemented with 10% (v/v) FBS, 100 units/ml penicillin and 100 μg/ml streptomycin. Sf9 cells were maintained at 27 °C in Grace's medium (Invitrogen) supplemented with 0.322% Bacto TC lactalbumin hydrolysate (Difco), 0.322% Bacto TC yeastolate (Difco) 10% (v/v) FBS, 100 units/ml penicillin and 100 μg/ml streptomycin.

Full-length Dnmt3a, Dnmt3a with D-mut (Dnmt3a-D-mut), Dnmt3a2, Dnmt3a-(1–211) and Dnmt3a-(1–211) with D-mut [Dnmt3a-(1–211)-D-mut] were subcloned into pcDNA3 (Invitrogen) with a FLAG tag at their N-termini in-frame. Transient transfection experiments involving HEK-293T cells were performed with Lipofectamine™ according to the manufacturer's protocol (Invitrogen). To obtain clones stably expressing the N-terminal region of Dnmt3a in HeLa cells, expression plasmids were transfected by electroporation at 1200 V and 25 μF with a Gene Pulser (Bio-Rad). The transfected cells were selected in culture medium supplemented with 0.4 mg/ml G418, and several independent clones were isolated.

Immunohistochemistry

Isolated HeLa cells stably expressing FLAG–Dnmt3a-(1–211) or FLAG–Dnmt3a-(1–211)-D-mut were fixed with 4% (v/v) formaldehyde, and then permeabilized with 0.1% Triton X-100 in PBS. The ectopically expressed proteins were immunodetected with a specific monoclonal antibody against the FLAG tag (M2, Sigma) and Alexa Fluor® 488-conjugated anti-mouse IgG antibodies as the secondary antibodies (Molecular Probes). The cells were observed under a laser confocal microscope (LSM 510, Zeiss) with an objective lens and ×63 Plan-Apochromat (Zeiss). The images were analysed with Pascal software (Zeiss).

Extraction of ectopically expressed Dnmt3a

HEK-293T cells transiently expressing Dnmt3a, Dnmt3a-D-mut, Dnmt3a2, Dnmt3a-(1–211) or Dnmt3a-(1–211)-D-mut were washed with Dulbecco's PBS, and then suspended in 20 mM Tris/HCl (pH 7.4) containing 1 M sucrose, 0.3% Triton X-100, 3 mM MgCl2, 15 mM 2-mercaptoethanol and 0.001% PI. The suspensions were centrifuged at 2400 g at 4 °C for 10 min. The precipitated nuclear fractions were resuspended in the above buffer supplemented with 0.3 M NaCl, and then incubated on ice for 30 min. After incubation, the mixtures were centrifuged at 10000 rev./min in a Beckman TLA 120.2 rotor at 4 °C for 10 min. An aliquot taken at each step was subjected to SDS/PAGE (10% gel), and Dnmt3a was immunodetected with specific antibodies [9].

RESULTS

It has been shown that the Dnmt3a gene is responsible for the DNA methylation of imprinted genes and repetitive sequences in germ cells [8]. It has also been shown that it is not the gene product of full-length Dnmt3a, but an isoform Dnmt3a2, which lacks the N-terminal 219 amino acid residues, that is responsible for the creation of the methylation patterns of the genome in male germ cells [13,17]. On the other hand, since the full-length Dnmt3a is ubiquitously expressed in somatic cells, its activity is expected to be strictly regulated so as not to allow aberrant methylation in the genome [2,9]. We estimated that the N-terminal region of Dnmt3a, which is the only difference between full-length Dnmt3a and Dnmt3a2, may contribute to the regulation of Dnmt3a-methylation activity or its localization to chromatin. In the present study, we focused on determining the properties of the N-terminal region of Dnmt3a.

DNA binding of full-length Dnmt3a and Dnmt3a2

Dnmt3a2, an isoform of Dnmt3a, lacks the 219 amino acid residues at the N-terminus of the full-length Dnmt3a, and thus Dnmt3a2 possesses an identical catalytic domain with that of Dnmt3a (Figure 1A). Nevertheless, Dnmt3a2 apparently shows a lower specific activity than Dnmt3a for DNA methylation [17]. To characterize further the catalytic activities of Dnmt3a and Dnmt3a2, histidine-tagged Dnmt3a and Dnmt3a2 were expressed in Sf9 cells and purified (Supplementary Figure S1 at http://www.BiochemJ.org/bj/437/bj4370141add.htm), and then the temperature-dependent DNA-methylation activities of the isoforms were determined and fitted to the Arrhenius equation (Figure 1B). The similar slopes of the plots for the two enzymes indicate that the activation energies of the DNA-methylation activities of Dnmt3a and Dnmt3a2 are almost identical (Table 1).

Figure 1 DNA-methylation and DNA-binding activities of Dnmt3a and Dnmt3a2

(A) Schematic illustration of the Dnmt3a and Dnmt3a2 structures. PHD, plant homeodomain. (B) DNA-methylation activities of Dnmt3a and Dnmt3a2 expressed in Sf9 cells determined at 293, 303 and 310 K. The methylation reaction was performed with 40 nM Dnmt3a or Dnmt3a2. Three measurements were made and are expressed as the mean±S.E.M. The DNA-methylation activities determined (d.p.m.) and the incubation temperature (1/T) were plotted and fitted to a linear regression curve obtained with the Arrhenius equation to calculate the Arrhenius energy activation values for the methyl group transfer activities. (C) Gel-shift assay of Dnmt3a and Dnmt3a2. The DNA fragment of 16 nM MMTV145 or 601.2×2 was incubated with no (−), 1, 30 or 200 nM Dnmt3a or Dnmt3a2 in the buffer described in the Materials and methods section, with 20 mM NaCl. Closed and open arrowheads indicate free and shifted DNA bands respectively. (D) Determination of the amounts of the shifted DNA bands. The DNA concentration-dependent-binding activities of Dnmt3a (●) and Dnmt3a2 (○) are shown. The values were taken from (C) and two other independent gel-shift assays (means±S.D.). (E) DNA-binding activities of Dnmt3a and Dnmt3a2 on filter-binding assays. The results obtained in two independent experiments are shown (exp. 1 and exp. 2). The vertical axis indicates radioactivity normalized to that bound to Dnmt3a2.

View this table:
Table 1 Activation energies of the DNA-methylation activities of Dnmt3a and Dnmt3a2

The activation energies of Dnmt3a and Dnmt3a2 were calculated from the slopes of the Arrhenius plot shown in Figure 1(B).

Recently, Jeong et al. [27] reported that the N-terminal region of human DNMT3B, which is positioned on the N-terminal side of the PWWP domain, is responsible for salt-resistant nucleosome binding. Similar to the N-terminal region of Dnmt3b, the N-terminal sequence of Dnmt3a is rich in basic amino acid residues (Figure 2A). When Dnmt3a was incubated with DNA, the N-terminal region of approximately 36 kDa containing a histidine tag and the extra sequence was protected from chymotrypsin attack (Supplementary Figure S2 at http://www.BiochemJ.org/bj/437/bj4370141add.htm). The N-terminal region appeared to form a complex with DNA and to form a domain that is resistant to protease digestion.

Figure 2 DNA-binding activity of Dnmt3a-(1–211)

(A) Amino acid sequence of the N-terminal domain (residues 1–211) of Dnmt3a [Dnmt3a-(1–211)]. Basic amino acid residues, lysine and arginine, are indicated in bold, and the 1–57 sequence is underlined. The mutated Lys51, Lys53, Arg177 and Arg179 to produce Dnmt3a-N-mut (K51A and K53A), Dnmt3a-C-mut (R177A and R178A) and Dnmt3a-D-mut (K51A, K53A, R177A and R179A) are shown in italics with double-underlines. (B) SDS/PAGE of the purified recombinant GST–Dnmt3a-(1–211). The purity of the protein (1 μg per lane) was more than 95%, as judged by densitometry. Molecular mass markers (kDa) are indicated. (C) GST–Dnmt3a-(1–211) (0, 160 and 800 nM) was incubated with MMTV145 (42 nM) or 601.2×2 (16 nM), and then gel-shift assays were performed as described in Figure 1(C). (D) Double-stranded oligonucleotides of 42 bp containing 12 CpG pairs with unmethylated (UN), hemimethylated (HM) and full methylated (FM) DNA were incubated with (+) or without (−) 640 nM GST–Dnmt3a-(1–211) or 80 nM full-length Dnmt3a, and then gel-shift assays were performed as described in Figure 1(C). Closed and open arrowheads in (C and D) indicate free and shifted DNA bands respectively.

On the basis of this finding, we expected that the N-terminal sequence, which is missing in Dnmt3a2, forms a domain that possesses DNA-binding activity. Thus we performed gel-shift assays with the two different sequence and size DNA fragments: MMTV145, 145 bp in length with 57.2% G+C content, and 601.2×2, 360 bp in length with 56.9% G+C content. Dnmt3a showed strong DNA-binding activities towards these DNA fragments, the activities being much stronger than that of Dnmt3a2 (Figure 1C). The shifted and free bands for MMTV145 DNA in Figure 1(C) and two other bands independently determined by gel-shift assay were semi-quantitatively determined in an image analyser and then plotted (Figure 1D). The DNA-binding activity of Dnmt3a, which was estimated to be less than 100 nM, was significantly higher than that of Dnmt3a2 (Figure 1D). When an alternative assay method involving filter binding for determination of the DNA-binding activity at a fixed concentration of another DNA fragment of 5S(RR) (length of 155 bp with 62.6% G+C content) was used, again the DNA-binding activity of Dnmt3a was higher than that of Dnmt3a2 (Figure 1E). These results indicate that the N-terminal domain of Dnmt3a is the major determinant of its DNA-binding activity with no preference for the sequences.

DNA-binding activity of the N-terminal domain of Dnmt3a

To determine whether or not residues 1–219 of the N-terminus by itself exhibit DNA-binding activity, we tried to purify this region. However, since the 1–219 sequence fused to GST could not be solubilized (results not shown), we expressed and purified the N-terminal sequence (residues 1–211) of Dnmt3a [Dnmt3a-(1–211)] instead (Figure 2B). Dnmt3a-(1–211) gave a single peak, with a 99.9% recovery, corresponding to an estimated molecular mass of 64.6±11.9 kDa (mean±S.D.) on dynamic light-scattering measurement and approximately 60 kDa on gel-filtration chromatography (Supplementary Figure S3 at http://www.BiochemJ.org/bj/437/bj4370141add.htm). The observed apparent masses were the mass of a dimer.

Recombinant Dnmt3a-(1–211) showed significant DNA-binding activity towards DNA fragments (Figure 2C), supporting the notion that the strong DNA-binding activity of Dnmt3a is mainly due to the N-terminal domain of Dnmt3a. The binding affinity of Dnmt3a-(1–211) was estimated to be less than 100 nM (Supplementary Figure S4 at http://www.BiochemJ.org/bj/437/bj4370141add.htm). As Jeong et al. [27] reported that Dnmt3a preferentially associates with methylated genomic targets in vivo, we examined whether or not Dnmt3a-(1–211) selectively binds to methylated DNA. As shown in Figure 2(D), neither full-length Dnmt3a nor Dnmt3a-(1–211) selectively recognized the methylated DNA. Thus something other than the direct binding to DNA through the N-terminal domain of Dnmt3a must be responsible for the recognition of the methylated DNA in vivo.

DNA-binding inhibitors such as minor groove binders [DAPI (4′,6-diamidino-2-phenylindole) and netropin] [28] did not inhibit DNA binding of Dnmt3a-(1–211) towards 26-mer 26AT DNA, but a major groove binder (Methyl Green) [29] did inhibit the DNA binding (Figure 3, top panel). Under identical conditions, the binding of the MBD domain of MBD1-(1–75) (which binds to the major groove containing methylated DNA) towards methylated 42-mer DNA was inhibited by Methyl Green, and the binding of Dnmt1-(119–197) (which binds to the minor groove of AT-rich DNA) towards 26-mer AT-rich DNA AT26 was inhibited by DAPI and netropin [22,30] (Figure 3, middle and bottom panels).

Figure 3 Effects of DNA-binding inhibitors on the DNA binding of Dnmt3a-(1–211)

No (−), 10 and 50 μM minor groove binders (DAPI and netropsin) or a major groove binder (Methyl Green) were incubated with 3 nM DNA and 80 nM GST–Dnmt3a-(1–211) (top panel), 160 nM His–GST–MBD1-(1–75) (middle panel) or GST–Dnmt1-(119–197) (bottom panel). The DNA used for the binding reactions for GST–Dnmt3a-(1–211), His–GST–MBD1-(1–75) and GST–Dnmt1-(119–197) were 26AT, 42 bp containing 12-CpG-methylated pairs and 26AT respectively. After the reaction, the mixtures containing GST–Dnmt3a-(1–211) and GST–Dnmt1-(119–197) were electrophoresed on a 0.8% agarose gel, and that of His–GST–MBD1-(1–75) in a 5% (w/v) acrylamide gel. The shifted DNA bands were visualized by staining with GelGreen as described in Figure 1(C). Closed and open arrowheads indicate free and shifted DNA bands.

Amino acid residues in Dnmt3a-(1–211) responsible for DNA binding

To identify the sequence in Dnmt3a-(1–211) responsible for DNA binding, Dnmt3a-(1–211) was dissected into the 1–161, 1–146, 1–57, 58–211, 58–161 and 162–211 sequences, and the resultant constructs were expressed as GST-fusion proteins (Figure 4A). These recombinant proteins were examined for DNA-binding activity using gel-shift assays. Among the constructs, the 1–57 construct bound to DNA with almost the same level as Dnmt3a-(1–211) (Figure 4B). The binding activities of the 58–211 and 162–211 constructs were significantly reduced. The 58–161 construct did not bind to DNA at all. On the other hand, the 1–161 and 1–146 constructs, although containing the 1–57 sequence, did not show as high binding activity levels as that of 1–57. This suggests that the DNA-binding activity of Dnmt3a-(1–211) may be negatively regulated by the 58–146 sequence. Removal of the large GST tag at the N-terminus from the 1–211 and 1–57 constructs did not have any effect on the binding affinity to DNA (Supplementary Figure S5 at http://www.BiochemJ.org/bj/437/bj4370141add.htm). In conclusion, at least two sequences, 1–57 and 162–211, mainly contribute to the binding of Dnmt3a-(1–211) to DNA.

Figure 4 The amino acid sequence responsible for the DNA-binding activity of Dnmt3a-(1–211)

(A) Truncated constructs of Dnmt3a-(1–211) were expressed as GST-fusion proteins and purified. Molecular mass markers (kDa) are indicated. (B) Gel-shift assay of the recombinant proteins (2 μg, 1.6–2.8 μM) with 16 nM 601.2×2. The positions of the free DNA (closed arrowheads) and shifted (open arrowheads) bands are indicated. (C) Wild-type (WT), Dnmt3a-N-mut (N-mut), Dnmt3a-C-mut (C-mut) and Dnmt3a-D-mut (D-mut) were purified. CBB staining of 1 μg each of the recombinant proteins WT, N-mut, C-mut and D-mut after SDS/PAGE is shown. (D) Gel-shift assays of WT, N-mut, C-mut and D-mut (1.6 μM each) with 16 nM 601.2×2 as described in Figure 1(C). Closed and open arrowheads indicate free and shifted DNA bands respectively.

To identify crucial amino acid residues for DNA binding, we introduced site-directed mutations into Dnmt3a-(1–211). Since the N-terminal domain of human DNMT3B was reported to be responsible for the binding to nucleosomes in vivo [27], we searched for homologous sequences in Dnmt3a-(1–211) and the N-terminal region of Dnmt3b, paying particular attention to basic amino acid residues, which are often responsible for DNA binding. We found two candidate sites at 50–54 in the 1–57 construct and 176–181 in the 161–211 construct. Three mutants of Dnmt3a-(1–211) with K51A and K53A (N-mut), R177A and R179A (C-mut), and K51A, K53A, R177A and R179A (D-mut) respectively were expressed as GST-fusion proteins and purified (Figure 4C). The DNA-binding activity was drastically reduced for N-mut, but less so for C-mut (Figure 4D). These results are consistent with the finding that the 1–57 construct bound to DNA more efficiently than the 161–211 construct, indicating that Lys51 and Lys53 mainly contribute to DNA binding. The mutant with four amino acid residues mutated [Dnmt3a-(1–211)-D-mut] completely lost binding activity to DNA (Figure 4D). Removal of GST from the N-terminus of the mutant again did not have any effect on the DNA-binding activity (Supplementary Figure S5).

We next examined the effect of the DNA-binding activity of Dnmt3a-(1–211) on the DNA-methylation activity of Dnmt3a. The purified recombinant Dnmt3a, Dnmt3a with D-mut (Dnmt3a-D-mut) and Dnmt3a2 (Supplementary Figure S1) were subjected to DNA-methylation activity determination under different salt concentrations (Figure 5A). The DNA-methylation activity of Dnmt3a-D-mut became sensitive to the salt concentration, which is similar to Dnmt3a2 lacking the N-terminal 1–219 sequence. These results strongly indicate that the DNA-binding activity of the N-terminal domain contributes to the salt-concentration sensitivity of the DNA-methylation activity.

Figure 5 Effect of the DNA-binding activity of the N-terminal domain on DNA-methylation activity and nuclear localization

(A) DNA-methylation activities of His–Dnmt3a, His–Dnmt3a2 and His–Dnmt3a-D-mut. The recombinant His-tagged Dnmt3a (●), Dnmt3a2 (▲) and Dnmt3a-D-mut (○) were determined using 4 pmol of enzyme with the indicated salt (KCl and NaCl) concentrations. The radioactivities (d.p.m.) incorporated into DNA after 1 h of incubation at 37 °C are shown. (B) Immunostaining of ectopically expressed Dnmt3a-(1–211) and Dnmt3a-(1–211)-D-mut. Stable transfectants expressing FLAG-tagged Dnmt3a-(1–211) and Dnmt3a-(1–211)-D-mut in HeLa cells were immunostained with the anti-FLAG antibody, followed by anti-mouse IgG antibodies conjugated with Alexa Fluor® 488, and then observed under a confocal microscope (left-hand panels). Nomarski images are shown in the right-hand panels. (C) Salt-dependent extraction of ectopically expressed various Dnmt3a constructs. The left-hand panel is a flow chart of the extraction procedure. HEK-293T cells, which ectopically expressed FLAG-tagged constructs of Dnmt3a, Dnmt3a2, Dnmt3a-D-mut, Dnmt3a-(1–211) and Dnmt3a-(1–211)-D-mut, were fractionated into postnuclear (S1) and nuclear (P1) fractions. The P1 fractions were extracted further with a buffer containing 0.3 M NaCl, and then fractionated into supernatant (S2) and precipitate (P2) fractions. Equivalent amounts to the starting materials (T) of the fractions were subjected to SDS/PAGE and then immunostained with anti-Dnmt3a antibodies (right-hand panel).

Nuclear localization of Dnmt3a is partly regulated by the N-terminal domain through its DNA-binding activity

It has been reported previously that ectopically expressed Dnmt3a and Dnmt3a2 are localized differently in somatic cells, that is Dnmt3a is localized to DAPI-dense heterochromatin and Dnmt3a2 is localized to euchromatin [12]. Jeong et al. [27] reported that almost all human DNMT3A is strongly anchored to nucleosomes.

To determine whether or not the DNA-binding activity of the N-terminal domain contributes to its nuclear localization in vivo, we isolated cells stably expressing FLAG-tagged Dnmt3a-(1–211) and Dnmt3a-(1–211)-D-mut. Ectopically expressed Dnmt3a-(1–211) was immunostained with an anti-FLAG antibody. Dnmt3a-(1–211) was localized inside the nuclei, whereas Dnmt3a-(1–211)-D-mut was diffusely distributed in both the nuclei and cytoplasm (Figure 5B). The N-terminal domain of Dnmt3a seems to have the ability to remain inside the nuclei owing to its DNA-binding activity, and thus this region may affect the localization of the whole Dnmt3a on chromatin.

Next, the nuclei of cells that ectopically expressed Dnmt3a, Dnmt3a2, Dnmt3a-D-mut, Dnmt3a-(1–211) and Dnmt3a-(1–211)-D-mut were prepared, and then extracted with 0.3 M NaCl. The majority of the Dnmt3a was resistant to extraction of nuclei with 0.3 M NaCl (compare Figure 5C, S2 and P2). On the other hand, almost all Dnmt3a-(1–211)-D-mut was found in the postnuclear fraction with a low salt buffer. As for Dnmt3a-D-mut, Dnmt3a2 and Dnmt3a-(1–211), more than half leaked out during the preparation of nuclei with the low-salt buffer (see S1 and P1), and approximately 23% of the remainder was extracted on the following 0.3 M NaCl treatment (compare S2 and P2). This indicates that the four basic amino acid residues in the N-terminal domain of Dnmt3a contribute to the localization in the nuclei via DNA-binding activity.

DISCUSSION

DNA-binding activity of the N-terminal domain and its function

The DNA-binding affinity of the N-terminal domain (residues 1–211) of Dnmt3a was quite high, which was estimated to be less than 100 nM (Supplementary Figure S4). Although Dnmt3a possesses other DNA-binding regions such as the catalytic domain and the PWWP domain, of which affinity for DNA are one order of magnitude lower than that of the N-terminal domain [31], the N-terminal domain seems to play a major role in DNA binding. This was shown the DNA-binding affinity of Dnmt3a being less than 100 nM (Figure 1C), which was similar to that of the N-terminal domain, and Dnmt3a2 lacking the N-terminal domain was weaker than that of Dnmt3a (Figures 1C–1E). In addition, the DNA-methylation activities of Dnmt3a-D-mut and Dnmt3a2, both of which lack DNA-binding activity due to the N-terminal domain, were sensitive to physiological salt conditions (Figure 5A). This allows the conclusion that Dnmt3a2 alone cannot methylate the genome in vivo because its activity was completely suppressed under physiological salt concentrations in vitro. The results of the present study lead to the idea that Dnmt3L, which interacts directly with Dnmt3a2 and endows Dnmt3a2 with salt resistance [32], may enhance the DNA binding of Dnmt3a2 to allow DNA methylation to occur in male germ cells [18,19] at the stage of global genome methylation, where Dnmt3a2 is the only active DNA methyltransferase [13,17].

The DNA-binding activity of the N-terminal domain did not show any selectivity to the nucleotide sequence (Figure 2C) or the methylation state (Figure 2D). In addition, the activation energies of Dnmt3a and Dnmt3a2 to the DNA-methylation reaction were similar to each other (Figure 1B and Table 1). Furthermore, the DNA-methylation activity of Dnmt3a2 was not enhanced upon the addition of purified Dnmt3a-(1–211) (results not shown). These results indicate that the N-terminal domain does not interact with the catalytic domain or affect the substrate sequence preference of Dnmt3a. The effect of the N-terminal domain on the ability of DNA methylation by Dnmt3a is likely to increase the accessibility to the substrate DNA.

Effect of the N-terminal domain on the localization of Dnmt3a

In cultured cells, ectopically expressed Dnmt3a-(1–211) did not show prominent localization to the constitutive heterochromatin regions, which are stained in a DAPI-dense manner, and the majority of the Dnmt3a-(1–211)-D-mut expressed was located outside of the nuclei. However, ectopically expressed Dnmt3a-D-mut was located partially in the DAPI-dense regions and completely inside the nuclei (Supplementary Figure S6 at http://www.BiochemJ.org/bj/437/bj4370141add.htm). This clearly indicates that the localization of the whole Dnmt3a molecule is regulated not only by the DNA-binding activity of the N-terminal domain, but also by the PWWP domain, which binds to DNA [31] and histone H3K36me3 [33], and the PHD (plant homeodomain) domain, which recognizes unmethylated Lys4 of histone H3 [34]. Thus the DNA-binding activity of the N-terminal domain contributes partly to the nuclear localization of Dnmt3a in vivo. It is rational to speculate that, because Dnmt3a2 does not contain the N-terminal DNA-binding domain, this domain may contribute not only to the salt-resistance of the DNA-methylation activity of Dnmt3a, but also to determination of the isoform-specific localization on genome DNA [35].

AUTHOR CONTRIBUTION

Isao Suetake and Yuichi Mishima performed most of the experiments and contributed to preparation of the manuscript. Hironobu Kimura performed immunofluorescence microscopy. Hideyuki Takeshima contributed to the determination of DNA-methylation activity. Young-Ho Lee, Yuji Goto and Takehisa Ikegami contributed to analysing the data. Isao Suetake and Shoji Tajima designed experiments, analysed data and prepared the manuscript.

FUNDING

This work was supported by the Japan Society for the Promotion of Science Grants-in-Aid for Scientific Research [B-22370053 (to S.T.) and C-21570136 (to I.S.)].

Acknowledgments

We wish to thank Dr Kohei Takeshita and Dr Atushi Nakagawa of our Institute for measurement of the dynamic light scattering of Dnmt3a-(1–211).

Abbreviations: CBB, Coomassie Brilliant Blue R-250; DAPI, 4′,6-diamidino-2-phenylindole; Dnmt, DNA methyltransferase; DTT, dithiothreitol; FBS, fetal bovine serum; GST, glutathione transferase; HEK-293T, HEK (human embryonic kidney)-293 cells expressing the large T-antigen of SV40 (simian virus 40); IPTG, isopropyl β-D-thiogalactopyranoside; MBD, methylated DNA-binding domain; PI, protease inhibitor cocktail; SAM, S-adenosylmethionine

References

View Abstract