RFTS‐dependent negative regulation of Dnmt1 by nucleosome structure and histone tails

DNA methylation in promoter regions represses gene expression and is copied over mitotic divisions by Dnmt1. Dnmt1 activity is regulated by its replication foci targeting sequence (RFTS) domain which masks the catalytic pocket. It has been shown that Dnmt1 activity on unmethylated DNA is inhibited in nucleosome cores. In the present study, we aimed to assess the effect of nuclesome formation on maintenance methylation at single CpG resolution. We show that Dnmt1 fully methylates naked linker DNA in dinucleosomes, whereas maintenance methylation was repressed at all CpG sites in nucleosome core particles. Deletion of RFTS partly released obstruction of Dnmt1 activity in core particles. Histone H3 tail peptides inhibited Dnmt1 in an RFTS‐dependent manner and repression was modulated by acetylation or methylation. We propose a novel function of RFTS to regulate Dnmt1 activity in nucleosomes.


Introduction
The 5th position of genomic Cytosine is frequently methylated. In gene promoters, methylation represses transcription, whereas there is a positive correlation with expression when methylation is located in gene bodies [1]. DNA methylation is crucial for normal development [2]. Once DNA methylation patterns are established, they are faithfully maintained through cell divisions. Maintenance DNA methylation is essential not only for differentiated cells but also for epigenetic reprogramming from primed to na€ ıve pluripotent cells [3]. When maintenance methylation is impaired, aberrantly differentiated or apoptotic cells are observed [4,5]. Despite numerous analyses of methylation patterns, the regulatory mechanism of DNA methylation is still largely unknown.
Dnmt1 is the enzyme primarily responsible for the accurate perpetuation of DNA methylation patterns following cell division. It can target unmethylated DNA but has a preference for hemimethylated substrate [6,7]. Dnmt1 À/À knockout in mice is lethal at early embryonic stage [8]. Dnmt1 is a multidomain protein [9,10], containing an independent replication foci targeting sequence (RFTS) domain. RFTS masks the catalytic pocket, and this intramolecular interaction increases the activation energy of Dnmt1 [9]. Dnmt1 interacts directly and/or indirectly with histone tail regions. In egg or cell lysate, the RFTS domain binds to ubiquitin, ubiquitylated histone H3, and unmodified H3 [11][12][13]. However, the physiological meaning of the interaction between RFTS and histone H3 is unclear.
The direct effect of nucleosomal structure on Dnmt1 activity in vitro has been reported in several studies, however, the described effects are controversial. The maximal activity of Dnmt1 toward mononucleosomes reconstituted with unmethylated DNA and histones prepared from HeLa cells was unchanged compared to naked DNA but K m values on substrates were increased by nucleosome structure [14]. When mononucleosomes were reconstituted with recombinant histones and unmethylated DNA, Dnmt1 activity toward nucleosomes was slightly reduced compared to naked DNA [15]. In another study, mononucleosomes including some hemimethylated CpG sites were used. DNMT1 showed an intrinsic ability to methylate CpG sites packaged into nucleosomes but the activity was DNA sequence dependent [16]. Analysis of mononucleosomes with unmethylated DNA by bisulfite sequencing showed that Dnmt1 de novo methylation was inhibited in nucleosome cores but not on linker DNA. Dnmt1 activity toward hemimethylated nucleosomal arrays was also reported to be lower than that toward naked DNA [17].
Dnmt1 methylates hemimethylated DNA in a processive manner, whereas unmethylated substrate is modified distributively [7]. In the present study, we therefore specifically address the effect of nucleosome structure on Dnmt1 maintenance methylation activity by using reconstituted dinucleosomes in which all CpG sites are hemimethylated. Using bisulfite sequencing, we mapped Dnmt1 activity at single CpG resolution. As previously shown for de novo methylation [17], core regions in the nucleosomes were refractory to maintenance methylation while linker DNA was not. On naked DNA, histone tail peptides repressed Dnmt1 activity at low micromolar concentration. This repression was lost upon removal of the RFTS domain. Acetylation of histone H3 peptide abolished interaction with RFTS and Dnmt1 inhibition. Our results suggest that histone presence regulates Dnmt1 maintenance methylation activity in two ways. Firstly, nucleosomes obstruct Dnmt1 access to DNA and this steric hindrance is partially released by RFTS removal. Secondly, the interaction with unmodified histone tails can inhibit Dnmt1 activity directly by stabilizing RFTS-dependent autoinhibition.

Results
Repression of the methylation activity of Dnmt1 in the nucleosome core region At single base resolution, analysis of Dnmt1-catalyzed methylation has so far been performed on nucleosomes prepared with unmethylated DNA [17]. To elucidate the effect of nucleosomal structure on mouse Dnmt1 activity on its intrinsic target, we prepared high-purity hemimethylated DNA which can be analyzed by sodium bisulfite sequencing (Fig. 1A). More than 95% of CpG sites on the upper strand in purified hemimethylated DNA were methylated, while less than 5% of CpG sites on the lower strand were sporadically methylated (Fig. 1B). The N-terminal 290 amino acids of Dnmt1 contain a sequence which binds to PCNA Schematically illustrated procedure of hemimethylated DNA preparation (left panel). Methylated cytosine in CpG sequences is indicated by red circles. DNA for each step was separated in 5% native PAGE, and fluorescently detected (right panel). PCR product, with a length of 570 bp, (lane 1) was methylated with M. SssI, and then treated with kexo. After kexo treatment, two bands (lane 2), which are derived from double-(570 bp DNA) and single-stranded DNA (arrowhead), were detected. The contaminating double-stranded DNA was removed with an acrylamide disk gel (lane 3), and then the lower strand was filled in using T4 polymerase (lane 4). The DNA was digested with BamHI and HindIII (lane 5), and the product was purified with an acrylamide disk gel (lane 6). The samples were run on the same gel, but intervening lanes were removed. The splice site is indicated by a black vertical line in the figure. DNA sequences of the primers are shown in Table S3 during replication. Dnmt1 291 lacking this domain is functionally similar to full-length Dnmt1 in vivo [18] and in terms of enzymatic activity on naked DNA [6]. did not observe specific sites exempt from methylation ( Fig. 1C), indicating that all CpG sites are recognized by Dnmt1 291 in equal manner.
We then reconstituted dinucleosomes from hemimethylated DNA and recombinant histones. In the nucleosomes, contamination by free DNA was negligible ( Fig. 2A), nucleosome core particles were positioned as expected (Fig. 2B), and the size of DNA wrapped on the core particles was similar to unmethylated template (Fig. 2C). The prepared nucleosomes were then used in a methylation reaction with Dnmt1 291 (Fig. 2D). To monitor the integrity of the nucleosomes during the reaction, bacterial methyltransferase M. AluI was added [19]. In three independent experiments, CpG sites in both core particle regions core 1 (CpG number 6-18) and core 2 (CpG 22-34) were scarcely methylated, while more than 90% of CpG sites in all linker DNA regions (linkers 1, 2, and 3) were methylated ( Fig. 2E and Fig. S2).  Full-length Dnmt1 (Dnmt1 FL), as well as Dnmt1 291, methylated more than 90% of CpG sites in the naked hemimethylated DNA (compare Fig. 3A and 1C). The prepared nucleosomes were then methylated with Dnmt1 FL (Fig. 3B). CpG sites in both core 1 and core 2 were scarcely methylated, while more than 90% of CpG sites in all linker DNA regions were methylated (Figs 3C and S3). The methylation pattern introduced by Dnmt1 FL was comparable to that by Dnmt1 291 (compare Figs 3C and 2E). In addition, no significant difference between the percentage of DNA methylation in each CpG catalyzed by Dnmt1 FL and Dnmt1 291 was observed (Fig. 3D). These suggest that the N-terminal region (amino acids 1-290) does not contribute to Dnmt1 inhibition by nucleosome structure.
Increase in Dnmt1 access to nucleosome entry/ exit sites by removing RFTS In Xenopus egg lysate, the N-terminal region (amino acids 180-470) lacking the PCNA binding domain but including the RFTS region of Xenopus Dnmt1 has been reported to bind to unmodified histone H3 [13]. We also detected binding of GST-mouse Dnmt1 to the N-terminal region of histone H3 (amino acids  which is limited to the RFTS domain (amino acids 291-602) (Fig. 4A). Crystal structure suggests that RFTS interacts with the catalytic pocket to block access to substrate DNA (Fig. 4B), thereby increasing activation energy [9]. We examined whether the truncation of RFTS changes the inhibitory effect of nucleosomes, by preparing Dnmt1 with a deletion of amino acids 1-601 (Dnmt1 602) (Fig. 4C). As for Dnmt1 291, we observed that the methylation activity of Dnmt1 602 was repressed by nucleosome structure (Figs 4D, S4, and 5A). The percentage of methylated CpGs in core regions methylated with Dnmt1 602 was slightly higher compared to Dnmt1 291 (compare  22 within the M1 region which are 24 and 22 bp away from the core, respectively. In contrast, toward naked DNA regions (A1, linker2, and A3) in nucleosomes, Dnmt1 291 and 602 showed similar DNA methylation activity (Fig. 5D). Thus, RFTS obstructs Dnmt1 activity toward CpG sites within the entry/exit region of the nucleosome core which are subject to spontaneous unwrapping described as 'breathing' [20,21]. In addition, the methylation pattern in the two consecutive nucleosomes was similar (Fig. 5B), indicating that methylation activity at entry/exit sites was not affected by the length of the adjacent linker DNA.

RFTS-dependent repression of Dnmt1 activity by histone N-terminal region
Because the histone H3 N terminus is protruding at 3 bp distant from nucleosome entry/exit sites, we next addressed the possibility that histone tails directly inhibit the activity of Dnmt1 [22]. The activity of Dnmt1 291 toward 42-bp oligonucleotides containing 12 hemimethylated CpG sites (42 bp/12hmCpG) (Table S2) was repressed by histone tail peptides. Histone H3 (1-30) repressed the activity of Dnmt1 291 with a half maximal inhibitory concentration (IC50) of approximately 2 lM (Fig. 6A and Table S1) while it did not affect the bacterial DNA methyltransferase M. SssI (Fig. 6B). The activity of Dnmt1 FL was also reduced by H3 (1-30), but the effect was slightly weaker compared to Dnmt1 291 (Fig. 6C), suggesting that the N-terminal region (1-290 aa) of Dnmt1 did not contribute to repression.
Substrate binding to Dnmt1 can be monitored by using an oligonucleotide in which the target cytosine in a CpG site is substituted with fluorocytosine (F-oligo). This substrate is covalently cross-linked upon binding the catalytic center and the cross-linked product can be monitored by slower migration of Dnmt1 in SDS/PAGE [23,24]. The formation of the reaction intermediate was reduced by adding H3 (1-30) to Dnmt1 291 (Fig. 6D). As the RFTS domain of Dnmt1 (291-602) bound to the histone H3 tail and RFTS removal derepressed methylation in the marginal region of nucleosomes, we next examined the effect of H3 (1-30) on the activity of Dnmt1 602. Dnmt1 602 activity was not suppressed by H3 (1-30), and rather slightly stimulated at low concentration (Fig. 6E). Substrate binding to Dnmt1 602 as monitored by using F-oligo was not repressed by H3 (1-30) (Fig. 6D). These results indicate that the inhibitory effect of histone tails on Dnmt1 activity is not due to capture of negatively charged DNA substrates by basic peptides but rather caused by interaction with RFTS.
Dnmt1 processively methylates hemimethylated DNA, whereas unmethylated substrate is modified distributively [7]. We next tested whether the observed repression by H3 tail was due to specific inhibition of the processive reaction. We therefore used 42-bp oligonucleotides as substrate in which only one CpG site was hemimethylated (42 bp/1hmCpG) (Table S2). Toward 42 bp/1hmCpG, the activity of Dnmt1 291 was also repressed (Fig. 7D), suggesting that processive and nonprocessive methylation were equally affected by H3 .
To further narrow down histone tail regions required for inhibition, we prepared truncated H3 peptides. H3 (10-30) did not inhibit Dnmt1 291 activity, while H3 (1-20) showed a biphasic effect (Fig. 8A). Dnmt1 291 activity was slightly stimulated at less than 10 lM H3 (1-20), but the stimulatory effect was lost at higher concentration. H3  or  repressed Dnmt1 291 activity, but the effect was milder than H3 . Loss of inhibition with truncated peptides suggests that a large binding interface is needed for Dnmt1 repression by histone H3.
A peptide (H3 rev) in which the amino acid sequence of H3 was inverted (Table S1) inhibited Dnmt1 activity with similar potency (Fig. 8B), suggesting that not the amino acid sequence but the total charge of the peptide is essential for the repression. We further examined whether the repressive effect is dependent on amino acid sequence. H4 , with a pI similar to H3 (10-30) (Table S1), inhibited the activity of Dnmt1 291 (Fig. 8C), however, the repressive effect was weaker than H3 (1-30) (compare to   (Table S1), did not inhibit the activity of Dnmt1 291 (Fig. 8D). In summary, histone tails can modulate Dnmt1 activity in a histone moleculedependent manner, and their inhibitory effect is influenced by post-translational modifications.

Discussion
Dnmt1 activity toward nucleosomes reconstituted with unmethylated or hemimethylated DNA has been reported [14][15][16][17]. Schrader et al. examined Dnmt1 activity toward nucleosomes including hemimethylated DNA, and reported an overall repressive effect, however, not resolved by nucleosome regions [17]. As Dnmt1 switches from distributive to processive methylation when presented with hemimethylated substrate, we analyzed Dnmt1 maintenance methylation activity at single CpG resolution on dinucleosomes containing hemimethylated DNA [7]. We found that Dnmt1 maintenance methylation activity was blocked at CpGs in nucleosome cores and the repression was partially released by removal of the RFTS domain. The RFTS guides Dnmt1 to replication foci in S-phase [31] and increases activation energy by intramolecular binding to the catalytic domain via hydrogen bonds [9,23]. The interaction between RFTS and the catalytic domain of Dnmt1 is released by interaction with the SRA domain of Uhrf1 [23], and Uhrf1 is crucial for DNA methylation [30]. Although several roles of RFTS in Dnmt1 function have been reported, its function in a nucleosomal context remains to be elucidated. In the present study, we observed that RFTS obstructs Dnmt1 access to nucleosome entry/exit sites. We also found that RFTS is directly bound by the N terminus of histone H3 and that this interaction inhibits Dnmt1 catalytic activity, potentially by stabilizing RFTS-dependent autoinhibition. Thus, RFTS regulates Dnmt1 activity not only by intramolecular interaction [9,32] but it also mediates Dnmt1 regulation by histones. Furthermore, as free, modified histones not incorporated in chromatin are observed in the nucleus [33], the repression of Dnmt1 by histone tails could not be limited to nucleosome structures.
In contrast to Dnmt1, de novo DNA methyltransferase Dnmt3a, as well as the Dnmt3a/Dnmt3L complex methylate linker DNA preferentially in a nucleosomal context [34,35]. The histone H3 tail stimulates the enzymatic activity of Dnmt3a in vitro and stimulation was abrogated by H3K4 methylation [35][36][37]. We showed that maintenance methylation by Dnmt1 is inhibited by histone H3, H3 methylated at lysine 9 and H4, and that the inhibitory effect of histone H3 is released by acetylation. Thus, de novo methyltransferase Dnmt3a and maintenance DNA methyltransferase Dnmt1 are oppositely regulated by unmodified histone H3.
We observed that H3 peptide with inverted amino acid sequence inhibited Dnmt1 with similar potency as the native H3 tail. Therefore, the polykationic nature of the peptide might be more important for inhibition than the peptide structure, possibly affecting protein or DNA conformation. In conclusion, this study shows a role of RFTS in regulating Dnmt1 function in a nucleosomal context and in mediating inhibition of Dnmt1 by histone tails. Future investigation should address the interaction between histone modifications and Dnmt1 to elucidate the regulation of DNA maintenance methylation.

Preparation of hemimethylated DNA
Hemimethylated DNA was prepared basically as described by Hermann et al. [38] but including a purification step to reduce contamination by fully methylated DNA in the product. The 601.2 sequence (Fig. S1) was amplified with primer A and B, and then 2 mg of DNA was methylated with 0.2 mg of M. SssI (NEB). Methylated DNA (2 mg) was treated with 3 lg of k exo at 37°C for 2 h to remove the lower strand, and then purified. The purification was performed using 6% acrylamide disk gel in Mini Prep Cell (BioRad, Tokyo, Japan) with 1 watt. The size of eluted DNA was monitored with 5% native PAGE in 19 TBE buffer. The fractions containing only single-stranded DNA were collected, annealed with primer C by incubation at 95°C for 2 min, and cooled down to 37°C with a 15 min linear gradient. The annealed DNA (3 lg) was mixed with 1.3 lg of T4 polymerase in a buffer containing 7 mM MgCl 2 , 15 mM (NH 4 ) 2 SO 4 , 0.5 mM DTT, 0.1 mM EDTA, 0.5 mM dNTP, and 50 mM Tris-HCl, pH8.5, and incubated 37°C for 5 min to fill in the lower strand. The filled in DNA was digested with HindIII and BamHI and the product was purified with 9% acrylamide disk gel in Mini Prep Cell. The preparation steps were summarized in Fig. 1A.

Reconstitution of nucleosomes
Dinucleosomes were prepared and purified as described by Mishima et al. [41]. Briefly, histone octamers and DNA were mixed in buffer containing 2 M KCl, and salt concentration was gradually reduced by dialysis. After reconstitution, contaminating free DNA was removed by ultracentrifugation with a glycerol gradient. The nucleosome concentration was determined via absorbance at 260 nm.

DNA methylation assay
Radioisotope-dependent assay of steady state methylation was performed as described by Takeshita et al. [9]. Briefly, annealed DNA (100 nM) was methylated with Dnmt1 at 37°C for 1 h. Radioactivity incorporated in DNA was measured by liquid scintillation counter, as described by Takeshita [9]. For sodium bisulfite (BS) sequence analysis, 1.4 nM dinucleosomes or naked DNA were methylated with 30 nM of the indicated Dnmt1 versions, under conditions as described above for the methylation assay. To monitor the nucleosome deposition, 0.1 UnitÁlL À1 of bacterial DNA methyltransferase M. AluI (NEB) was added in the reaction mixture. An aliquot of the reaction mixture was treated with Epitect bisulfite kit (Qiagen, Tokyo, Japan). The bisulfite-treated DNA was amplified by PCR with primers (SBS upper sd Fwd and SBS upper sd Rev, or SBS lower sd Fwd and SBS lower sd Rev) ( Table S2). The amplified DNA was subcloned into pBlueScriptII and its sequence was determined by ABI 3100 DNA sequencer. CpG methylation was analyzed using QUMA [42].

F-oligo binding to Dnmt1
The two complementary oligonucleotides (20 bp/1CpG-F sense M and 20 bp/1CpG-antisense in Table S1) were annealed, and the annealed oligonucleotides (200 nM) were methylated with full-length or truncated Dnmt1 (80 nM), under the same conditions as for measuring methylation activity. After incubation at 37°C for 1 h, the reaction mixtures were separated by 7% SDS/PAGE. Proteins were stained with Lumitein (Biotium, Inc, CA, USA), and bands were detected with a fluoro-imager, FLA9500, as described by Mishima et al. [43]. The percentage of shifted Dnmt1 over input Dnmt1 was calculated.

Supporting information
Additional Supporting Information may be found online in the supporting information tab for this article: Table S1. Amino acids sequence of H3 and H4 peptides. Table S2. Oligonucleotides used for analyzing Dnmt1 activity. Table S3. Oligonucleotide for PCR and fill-in reaction. Fig. S1. Sequence of hemimethylated DNA. Fig. S2. Reproducibility of Dnmt1 291 methylation activity. Fig. S3. Nucleosome structure-dependent DNA methylation activity of Dnmt1 FL.