Deficient in DNA methylation 1 (DDM1) defines a novel family of chromatin-remodeling factors.

Deficient in DNA Methylation 1 (DDM1) protein is required to maintain the DNA methylation status of Arabidopsis thaliana. DDM1 is a member of the broad SWI2/SNF2 protein family. Because of its phylogenetic position, DDM1 has been speculated to act as a chromatin-remodeling factor. Here we used a purified recombinant DDM1 protein to investigate whether it can remodel chromatin in vitro. We show that DDM1 is an ATPase stimulated by both naked and nucleosomal DNA. DDM1 binds to the nucleosome and promotes chromatin remodeling in an ATP-dependent manner. Specifically, it induces nucleosome repositioning on a short DNA fragment. The enzymatic activity of DDM1 is not affected by DNA methylation. The relevance of these findings to the in vivo role of DDM1 is discussed.

Deficient in DNA Methylation 1 (DDM1) protein is required to maintain the DNA methylation status of Arabidopsis thaliana. DDM1 is a member of the broad SWI2/ SNF2 protein family. Because of its phylogenetic position, DDM1 has been speculated to act as a chromatin-remodeling factor. Here we used a purified recombinant DDM1 protein to investigate whether it can remodel chromatin in vitro. We show that DDM1 is an ATPase stimulated by both naked and nucleosomal DNA. DDM1 binds to the nucleosome and promotes chromatin remodeling in an ATP-dependent manner. Specifically, it induces nucleosome repositioning on a short DNA fragment. The enzymatic activity of DDM1 is not affected by DNA methylation. The relevance of these findings to the in vivo role of DDM1 is discussed.
The compaction of eukaryotic genomes into chromatin structures has profound implications for nuclear processes such as replication (1,2), transcription (3,4), DNA repair (5), and recombination (6,7). Evolution has created a diverse repertoire of regulatory mechanisms affecting the dynamics of chromatin structure. One broad group of mechanisms involves a battery of the enzymes that covalently modify histones and DNA (8 -10). Specific regulatory factors recognize the modification status of chromatin fibers and render the structure active or inactive (8).
To ensure sufficient structural flexibility of the chromatin, the cell employs a specialized class of multiprotein complexes, which utilize the energy of ATP hydrolysis to change chromatin folding in the poorly understood process of "chromatin remodeling" (11,12). Central to this activity are SWI2/SNF2-type ATPases that form the catalytic core of these remodeling complexes. The SWI2/SNF2 protein family falls within the large superfamily of DEXD/H-ATPases (13). The distinctive signature of the SWI2/SNF2 family is the SNF2_N domain, a variant of the typical DEXD/H domain that contains a well conserved C-terminal extension of ϳ100 amino acids. 1 Phylogenetic analysis divides the SWI2/SNF2 family into a number of subfamilies (13), which appear to be present in all major taxa. Outside the SNF2_N and HelicC domains, which form the catalytic module, these subfamilies are not well conserved. Members of three closely related subfamilies (ISWI, Mi2/CHD, and SWI2/ SNF2) have been shown to serve as the catalytic subunits of chromatin-remodeling complexes (15)(16)(17)(18). Recently, remodeling activity has also been demonstrated for a complex built around the more distantly related INO80 ATPase (19). All catalytic subunits of the chromatin-remodeling complexes that have been analyzed can act outside the complex context, although they require associated proteins to achieve their full activity (16 -18, 20). CSB/ERCC6 defines a fifth group of remodelers, although it has been analyzed only as the isolated, recombinant protein (21). During recent years it has become clear that energy-dependent remodelers act in concert with histone modifying enzymes (22)(23)(24)(25)(26)(27).
So far only very few family members have been analyzed in biochemical assays. At least one of them, MOT1, acts as a transcriptional repressor that regulates TATA binding protein binding to promoters (28 -30). Therefore, it is likely that only a subset of SWI2/SNF2-like proteins serve as true chromatinremodeling factors.
SWI2/SNF2-like proteins participate in various nuclear activities including transcriptional control (24,31), DNA repair (Ref. 13 and references therein), chromosome segregation (32), and chromosome folding (32,33). The members of two SWI2/ SNF2-type subfamilies, ATRX and DDM1, are involved in the control of DNA methylation status (34,35). In the plant Arabidopsis thaliana, the ddm1 genetic background results in a 70% decrease in the DNA methylation level (36). In consequence, this leads to deregulation of gene expression and the reactivation of silent transposons (37)(38)(39)(40)(41)(42). DDM1 belongs to a small protein family conserved in plants (A. thaliana DDM1, NCBI accession number AAD28303; Zea mays, NCBI AAL73042), fungi (Saccharomyces cerevisiae Yfr038wp, NCBI NP_116696; Aspergillus fumigatus, NCBI CAD28443), and mammals (Mus musculus lymphocyte-specific helicase, NCBI NP_032260 and Homo sapiens proliferation-associated SNF2like gene, NCBI AAF82262). No genes encoding DDM1-like proteins have been found in the two fully sequenced invertebrate genomes. Evolutionary analysis places the DDM1 subfamily close to ISWI and SNF2/SWI2 groups (43). 2 It is presently unclear how DDM1 acts to maintain DNA methylation status, although it has been speculated that it could remodel chromatin structure to facilitate the access of DNA methylases to the substrate (35). However, the remodeling activity of DDM1 has not been analyzed, and this hypothesis is based solely on the phylogenetic position of the DDM1 subfamily.
To address the above issue we have expressed recombinant DDM1 and characterized its biochemical activities. We find that DDM1 is an ATPase stimulated by both naked and nucleosomal DNA. It binds to nucleosomes and promotes nucleosome repositioning in an ATP-dependent manner. Our data indicate that DDM1 defines a novel class of chromatin-remod-eling factors. Remarkably, considering its apparent function in the nucleus, we find that DDM1 activity is not affected by DNA methylation status.

EXPERIMENTAL PROCEDURES
Production of Recombinant DDM1 Protein-The DDM1 cDNA was amplified by PCR and cloned into the pFAST BAC 1 vector (Invitrogen) cleaved with BamHI and SalI sites. The 3Ј primer contained a DNA sequence encoding a hexahistidine tag. The PCR product encoded a DDM1 fusion protein carrying a C-terminal His tag. The sequence of this fragment was verified by DNA sequencing. After transfection and amplification, the recombinant virus was used to infect SF21 insect cells. Three days after infection, the cells were collected, washed with phosphate-buffered saline, and resuspended in extraction buffer (20 mM HEPES-KOH, pH 7.8, 300 mM KCl, 5 mM imidazole,10% glycerol, 0.1% Nonidet P-40). After homogenization cellular debris were removed by centrifugation, and the extract was applied to a nickel-agarose column. The resin was washed extensively with homogenization buffer containing 25 mM imidazole. Bound proteins were then eluted with the buffer containing 500 mM imidazole. Further purification and imidazole removal were achieved by gel filtration chromatography. Purity was estimated to be about 90% by SDS-PAGE. The protein identity was confirmed by mass fingerprinting using a QTof (Micromass) spectrometer. The preparation was tested for contaminating nuclease activity. To exclude any effect of contaminants, a "mock" preparation was made using uninfected cells. This preparation was tested in all experiments and did not show any activity in either the enzymatic or band-shift assays.
DNA Fragments and Chromatin Reconstitution-For the initial ATPase assays, pBR322 plasmid DNA was used. Chromatin was reconstituted on the pBR322 DNA by the salt dialysis method using purified chicken erythrocyte histones (44). Assembly quality was verified by micrococcal nuclease digestion. For nucleosome mobility assay, mononucleosomes were assembled on a 248-bp fragment of mouse rDNA (45). The DNA was labeled with 32 P by PCR. After gel purification, the PCR products were converted into nucleosomes by the addition of core histones and salt dialysis. Different ratios of DNA to histones were used, and reconstitution products were analyzed by electromobility gel-shift assay to reveal the optimal conditions for nucleosome assembly. For nucleosome mobility assays, nucleosomes positioned on specific DNA sequences were resolved by electrophoresis on a 4.5% polyacrylamide gel and purified (45). For assays involving DNA methylation, the 210-bp ClaI fragment of the A. thaliana FWA promoter was used (positions 1287-1496 of NCBI accession number AF178688). DNA was methylated with SssI DNA methylase (New England Biolabs), and methylation was confirmed by digestion with Fnu4HI. Labeling of DNA and the optimization of nucleosome assembly conditions were performed as described above.
ATPase Assay-The reactions were performed in buffer consisting of 20 mM Tris-HCl, pH 8.0, 50 mM NaCl, 1.5 mM MgCl 2 , 1 mM ␤-mercaptoethanol, and 100 g/ml bovine serum albumin. Ten fmol of recombinant DDM1 were incubated with 100 ng of free or nucleosomal DNA in the presence of 66 M ATP and a trace amount of [␥-32 P]ATP. After a 60-min incubation at 25°C, 1-l samples were spotted onto polyethyleneimine-cellulose plates, and the reaction products were separated from nonhydrolyzed ATP by thin layer chromatography. For kinetic analysis, the samples were incubated with ATP at concentrations ranging from 10 to 500 M for 30 min. Spots were quantified with a Storm PhosphorImager using ImageQuant software (Molecular Dynamics).
Nucleosome Mobility Assay-The nucleosome mobility assay was performed essentially as described by Laengst et al. (45). Gel-purified nucleosomes (60 fmol) were incubated with recombinant DDM1 (1-50 fmol) in the ATPase buffer containing 1 mM ATP for 1 h at 25°C in a final volume of 10 l. The reaction was stopped by the addition of 500 ng of competing plasmid DNA and further incubation for 10 min. Nucleosomes were then separated by native gel electrophoresis on a 4.5% polyacrylamide gel in 1ϫ Tris acetate EDTA buffer. Gel-resolved bands were visualized with the PhosphorImager.

DDM1
Interacts with Nucleosomal DNA-Sequence analysis of proteins of the DDM1 family did not reveal any known domain that might be implicated in DNA or nucleosome recognition. Therefore, it was necessary to determine whether DDM1 itself can bind to its potential substrates, naked DNA, or nucleosomes. To this end purified, recombinant DDM1 was used in band-shift assays with short fragments of DNA and with in vitro assembled mononucleosomes.
To produce sufficient quantities of DDM1 protein, a recombinant baculovirus expressing A. thaliana DDM1 was constructed and used to infect SF21 cells. The expressed DDM1 protein was isolated and found to be ϳ90% pure (Fig. 1A). A 248-bp fragment of mouse rDNA was amplified by PCR and used as free DNA or in a form of mononucleosomes assembled with purified chicken core histones. A band-shift assay with DDM1 and free DNA demonstrated the formation of nonspecific DNA-protein complexes (Fig. 1B). Increasing the protein concentration resulted in a smearing pattern, indicating the binding at the multiple sites. In contrast, the interaction of DDM1 with the mononucleosomes resulted in the formation of one distinct shifted band (Fig. 1C), indicating the production of one bound species. Despite these differences in the nature of the complexes formed, band-shift assay showed that DDM1 binds to both targets with similar affinity.
ATPase Activity of DDM1 Is Stimulated by Both Free and Nucleosomal DNA-Although SWI2/SNF2-type chromatin-remodeling ATPases occur in vivo in multi-protein complexes, all of the enzymes analyzed thus far display ATPase activity in the absence of other subunits. To examine the activity of purified DDM1 protein, the ATPase assays were performed in the presence of core histones, plasmid DNA, or plasmid-assembled chromatin ( Fig. 2A). Buffer was used as reference. To exclude any effect of potential contaminants, the mock preparation from uninfected cells was assayed in parallel reactions. To optimize the assay, DDM1 activity was analyzed under a broad spectrum of different conditions. The optimal reaction conditions were a buffer of pH 8.0 containing 50 mM NaCl, 100 g/ml bovine serum albumin, 1 mM ␤-mercaptoethanol, and 1 mM MgCl 2 and a temperature of 25°C. In such a reaction ATPase activity of DDM1 was clearly stimulated by free DNA. Very little additional stimulation was seen in the presence of an equivalent amount of DNA assembled into chromatin. The addition of core histones had no effect on the DDM1 activity. These findings are consistent with the results of band-shift assays and imply that interaction with both DNA and nucleosomes stimulates the ATPase activity to similar levels.
To understand how DNA and nucleosomes stimulate the ATPase of DDM1, a kinetic analysis of the ATP hydrolysis reaction was performed. A time course of the ATP hydrolysis showed that the reaction is linear for 90 min (data not shown). To ensure linearity of the reaction, the assays were stopped after 30 min. The velocities of the ATP hydrolysis by recombinant DDM1 in the absence or presence of DNA or nucleosomes were determined by titrating with ATP (Fig. 2B). The maximal velocity (V max ) and the Michaelis-Menten constant (K m ) were then determined by curve fitting and double-reciprocal plot (Fig. 2C). Both V max and K m of ATP hydrolysis by recombinant DDM1 were higher than the corresponding values determined for the isolated Mi2␣-remodeling ATPase (17), although they were of the same order of magnitude (Table I). The analysis revealed that DNA and nucleosomes increased the V max 2.9and 3.1-fold, respectively, and the K m only 1.2-fold. Because V max measures the turnover number, our data indicate that both cofactors stimulate the ATPase of DDM1 mainly by increasing its turnover number while changing the affinity of DDM1 to ATP only slightly.
DDM1 Is a Chromatin-remodeling Factor-The interaction with mononucleosomes and the stimulation of ATPase activity by nucleosomal DNA suggested that DDM1 could be a chromatin-remodeling factor. To address this issue directly a well established sliding assay, which measures the movement of a histone octamer along a short DNA fragment, was used (45). The histone octamers were deposited on a 248-bp rDNA fragment. Gel electrophoresis of the assembled products revealed two discrete species (Fig. 3A) that correspond to nucleosomes positioned at the center or at the end of the DNA fragment (45). Positioned nucleosomes were gel-purified and incubated with increasing amounts of DDM1 protein. DDM1 was able to induce, in an ATP-dependent manner, the redistribution of the histone octamer (Fig. 3B). DDM1 was found to move nucleosomes from the end to the center of the DNA fragment with much greater efficiency than in the opposite direction. DDM1 ATPase Activity Is Not Stimulated by DNA Methylation-DDM1 is involved in the maintenance of CpG DNA methylation in vivo (36). Therefore, we wished to see whether DNA methylation would affect the activity of recombinant DDM1 in in vitro assays. To this end DDM1 activity was compared in the ATPase assay performed with nucleosomes assembled on methylated and on non-methylated DNA. We decided to use an ATPase assay since it gives the most quantitative and directly comparable results. A fragment of A. thaliana FWA gene promoter was used because it contains a high number of CpG sequences (18 methylcytosines in a 210-bp-long DNA fragment) and has been shown to undergo demethylation in ddm1 mutant plants (40). The DNA was amplified by PCR and methylated with SssI methylase. The efficiency of DNA methylation was checked by cleavage with the restriction enzyme Fnu4HI, which is sensitive to methylation of its recognition sequence (Fig. 4A). Methylated and non-methylated FWA fragments were then used in ATPase assays as both free and nucleosomal DNA. This experiment showed that there was no influence of DNA methylation on the ATPase activity of DDM1 (Fig. 4B). DISCUSSION The SWI2/SNF2 protein family falls within the DEXD/H superfamily of ATPases. It is defined by the SNF2_N domain, a variant of DEXD/H-ATPase domain that contains an additional conserved region of about 100 amino acids extending C-terminally to the classical DEXD/H domain. 1 Based on phylogenetic analysis the SWI2/SNF2 family can be divided into several subfamilies that share the SNF2_N and HelicC domains but are not well conserved outside these regions (13). 2 Chromatin-remodeling activity has been demonstrated for proteins of the closely related SWI2/SNF2, ISWI, and Mi2 subfamilies (15)(16)(17)18). Recently the proteins of two other more distant subfamilies, INO80 and ERCC6/CSB, have been shown to act in a similar manner (19,21). Here we present experimental evidences of chromatin-remodeling activity by a member of the DDM1 subfamily.
DDM1 Interacts with the Nucleosome-Our results show that DDM1 binds to both free DNA and nucleosomes with similar affinity (Fig. 1). However, the binding to the nucleosomal target resulted in a different pattern of gel-resolved complexes compared with the binding to free DNA. The diffused pattern formed by free DNA-DDM1 complexes indicated nonspecific interactions, whereas binding to the nucleosome resulted in the formation of a single distinct complex. These results indicate that in contrast to free DNA, the nucleosomal structure has a preferred binding site for DDM1. We did not observe any stable interaction of DDM1 with core histones in a pull-down assay (data not shown). These data suggest that although DNA is a major target for DDM1 binding, it is the nucleosomal structure that provides a specific structural framework for complex formation. This phenomenon is reminiscent of substrate binding by ISWI, another member of SWI2/SNF2 family. 3 Enzymatic Activity of DDM1-Purified recombinant DDM1 has weak intrinsic ATPase activity in vitro. ATP hydrolysis was greatly stimulated by naked DNA. A slight additional stimulation by nucleosomal DNA was seen (Fig. 2). The catalytic subunits of all chromatin-remodeling complexes characterized so far display ATPase activity when assayed as isolated polypeptides. However, they differ in their response to DNA and chromatin. SWI2/SNF2-like proteins are fully active in the presence of DNA, and no additional stimulation by chromatin is observed (43). In contrast, other remodelers such as Miz␣ and ISWI require nucleosomal DNA to achieve their full activity (15)(16)(17). This observation suggests that functionally, DDM1 may be more closely related to the SWI2/SNF2 subfamily than to ISWI or CHD/Mi2.
The determination of the kinetic parameters of ATP hydrolysis in the absence or presence of DNA or nucleosomes showed that both DNA and nucleosomes increased primarily the V max of the reaction (2.9 and 3.1 times, respectively) while only marginally affecting the K m (1.2-fold increase in both cases). Because V max measures the turnover number of an enzyme, the above kinetic data indicate that both free DNA and nucleo-somes stimulate the ATPase of DDM1 mainly by increasing its turnover number. Similar dependence was found earlier for Mi2␣-remodeling ATPase (17).
Purified DDM1 showed nucleosome-remodeling activity in vitro. It is able to catalyze the redistribution of histone octamers on short DNA fragments in a process that requires the energy provided by ATP hydrolysis. Similar "nucleosome-sliding" activity has been observed for ISWI (15) and dMi2 (16).
As reported previously (46), ISWI repositioned only nucleosomes located at the center of a 248-bp rDNA fragment but not those located at the end. Interestingly, the same group has reported that the ISWI containing complex ACF and purified dMi2 behave in exactly the opposite way (20,16). Most likely, these observations reflect mechanistic differences between the remodelers. We found that DDM1, like ACF and dMi2, was considerably more efficient at mobilizing nucleosomes occupying the fragment termini than located at the center of the 248-bp rDNA. The mechanistic reasons for this directionality of nucleosome sliding are not clear.
The nucleosome sliding activity of DDM1 appears to be rel-3 G. Langst, personal communication.  atively weak. To move ϳ50% of the nucleosomes positioned at the end of a 248-bp rDNA fragment (ϳ30 fmol) to the center of the DNA molecule, 50 fmol of recombinant DDM1 were required. To achieve comparable efficiency it is sufficient to use about 7 times less of isolated, recombinant ISWI or 150 times less of ACF complex (20). However, it is not clear if different recombinant proteins can be directly compared. This observation relates DDM1 to other remodelers that require the context of the native complex for full activity. As demonstrated for other SWI2/SNF2-type proteins such as hBrm, BRG1 (47), and Mi2␣ (17), the associated proteins are essential to increase the rate of ATP hydrolysis. ATPaseassociated subunits have also been shown to act as coupling factors that allow more efficient energy usage (20). However, we can only speculate on the role of putative DDM1associated proteins, as DDM1 has not yet been shown to form a larger complex.
The low efficiency of DDM1 in remodeling assays may also reflect the intrinsic limitations of our experimental system. To demonstrate the remodeling activity of DDM1, standard chromatin-remodeling assay has been used. However, it is possible that this assay, developed for other remodelers, is not perfectly suited for DDM1. Chromatin-remodeling complexes appear to have evolved to fulfill various nuclear functions and to induce different changes in the structure of chromatin. It is likely that some novel assay system is required to fully characterize the mechanisms of DDM1induced chromatin remodeling.
DDM1 Activity and DNA Methylation Status-DDM1 is involved in the maintenance of DNA methylation (36). Similarly to ddm1, mutations in Lsh, a gene encoding a mouse homolog of DDM1, also cause alterations in the DNA methylation pattern (46). Therefore, we wished to see whether DNA methylation can affect DDM1 activity. It has been suggested that DDM1 may alter the chromatin structure to facilitate methyltransferase access to DNA (35). According to this scenario, fully CpG-methylated DNA would represent the final product of the DDM1-assisted reaction. Therefore, it might be anticipated that methylated DNA would inhibit or at least not activate DDM1. Alternatively, DDM1 could act to compact the chromatin structure of methylated DNA to render these regions of the genome inaccessible to de-methylation and, therefore, help to preserve the methylation pattern. If this were the case, preferential activation of DDM1 by methylated DNA might be predicted. Surprisingly, we did not observe any effect of DNA methylation on the ATPase activity of DDM1 (Fig. 4). Band-shift assays also failed to show any effect of DNA methylation on substrate binding affinity (data not shown). One possible explanation for these findings is that DDM1 requires other associated proteins to sense the DNA methylation status. Alternatively, DNA methylation may have some subtle qualitative influence on the DDM1catalyzed reaction that was not detected in the experimental system used.
Recently the CpXpG DNA methylation catalyzed by the chromomethylase 3 was found to depend on the activity of kryptonite (47), the product of one of 15 genes in Arabidopsis that potentially encode histone H3 Lys-9 methyltransferases (48). However, ddm1 mutations also affect CpG DNA methylation (37), which is catalyzed by another DNA methylase MET1 (49,50). It has been demonstrated that in Neurospora crassa the CpG DNA methylation acts genetically downstream of dim5 (51), which encodes a homolog of Drosophila melanogaster histone H3 Lys-9 methyltransferase Su(var)3-9. Interestingly, Gendrel et al. (52) recently demonstrated the alteration of the chromosomal distribution of Lys-9-methylated histone H3 in ddm1 mutants. Therefore, we speculate that DDM1 could be involved in histone H3 Lys-9 methylation. According to this scenario the effect on DNA methylation, which is seen in ddm1 plants, would be indirect. We are actively investigating this possibility.