Programming the Transcriptional State of Replicating Methylated DNA*

CpG methylation is maintained in daughter chromatids by the action of DNA methyltransferase at the replication fork. An opportunity exists for transcription factors at replication forks to bind their cognate sequences and thereby prevent remethylation by DNA methyltransferase. To test this hypothesis, we injected a linearized, methylated, and partially single-stranded reporter plasmid into the nuclei of Xenopus oocytes and followed changes in the transcriptional activity after DNA replication. We find that dependent on Gal4-VP16, the action of DNA methyltransferase, and replication-coupled chromatin assembly DNA replication provides a window of time in which regulatory factors can activate or repress gene activity. Demethylation in the promoter region near the GAL4 binding sites of the newly synthesized DNA did not occur even though the Gal4 binding sites were occupied and transcription was activated. We conclude that “passive” demethylation at the replication fork is not simply dependent on the presence of DNA binding transcriptional activators.

CpG methylation is causal for transcriptional silencing, imprinting, and X inactivation (1)(2)(3)(4). CpG island hypermethylation at promoter elements of tumor suppressor genes correlates with their repression (5)(6)(7)(8). The regulatory mechanisms by which DNA methylation interferes with gene activity have begun to be unraveled (2). Large repressor complexes containing methyl-CpG-binding proteins and histone deacetylases have been characterized (9 -11), directly linking CpG methylation to transcriptional repression. To achieve the precise epigenetic control of gene expression, the mechanisms of maintenance methylation, de novo methylation, and demethylation must be strictly regulated.
Several mechanisms have been proposed to keep a promoter in a CpG island methylation-free state (reviewed in Ref. 12). Active demethylation occurs during early mouse development (13). However, the existence of proteins exerting demethylase activity has been controversial and remains obscure (Ref. 14 and references therein). Alternatively, DNA replication might be involved in controlling the methylation state of promoters (15). Replicative processes might provide a window of opportu-nity for transcription factors to bind to hemimethylated DNA and thereby prevent remethylation and subsequent gene repression. Transcription factor binding might interfere with the remethylating activity of DNA methyltransferases at the replication fork (16,17).
Transcription factor binding during the replication of DNA leads to subsequent demethylation concomitant with DNA replication (18). In Xenopus eggs it has been shown that replication, together with factor binding but not ongoing transcription, results in partial locus-specific demethylation (15). Alternatively, nonmethylated CpG islands may result from a transcriptionally active gene, the sequences of which also function as an origin of DNA replication at a very early developmental stage (19).
All enzymatic activities and co-factors necessary for efficient DNA replication are present in the Xenopus oocyte. Although oocytes do not initiate DNA replication on double-stranded DNA, they can convert single-stranded DNA into doublestranded DNA and support the concomitant assembly of chromatin. Previous studies using single-stranded unmethylated DNA showed the relief of nucleosomal repression by transcriptional activators and that the assembly of transcription complexes and replication-coupled chromatin assembly are highly competitive processes (20,21). Here we examine whether transcription factors compete with the DNA methylation maintenance machinery on CpG-methylated templates directly at the replication fork and whether this competition can determine the transcriptional fate of a reporter gene. Our approach is first to in vitro methylate linearized template DNA and subsequently to use the enzyme exonuclease III to generate unidirectional deletions. The modified DNA is then microinjected into frog oocytes. Using this partially single-stranded methylated template, we find that the stable binding of transcriptional activators or repressors during the synthesis of the second DNA strand programs the gene to remain either active or repressed. This programming is, under our experimental conditions, independent of complete demethylation near the binding site. Hence, factor binding during the time window of second strand synthesis might tag the sites for complete demethylation once replication is completed.

EXPERIMENTAL PROCEDURES
Plasmids-The construct pGEMH2BLuc used in this study is derived from the vector PGEM-3Zf(ϩ) (Promega). The plasmid was cut with EcoRI and HindIII and ligated with an EcoRI/HindIII fragment from the plasmid pE4 containing five Gal4 binding sites and the adenovirus E4 promoter (kindly provided by Mike Carey, University of California, Los Angeles). The minimal adenovirus E4 promoter was replaced with the TaqI/MluI fragment of the histone H2B promoter. The luciferase gene sequence ranging from ϩ1 to ϩ200 was polymerase chain reaction-amplified using specific primers containing BamHI and SmaI restriction sites and was subsequently cloned into the SmaI/ BamHI linearized vector. Expression plasmids encoding fusion proteins of MeCP2 and Gal4-VP16 used in this study are described in Ref. 9.
Exonuclease III Digestion-Templates for second strand synthesis in * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  vitro and in vivo were generated using exonuclease III (exo III) 1 (Promega), which produces unidirectional deletions starting at the 3Ј end. To protect one end from digestion, the plasmid pGEMH2Bluc was cut with KpnI and SmaI. Deletions were produced starting from the SmaI end and heading into the regulatory sites of the H2B promoter and upstream Gal4 sites. 10 g of the linear fragment were incubated with 500 units of exonuclease III at 37°C for 10 min in a buffer containing 66 mM Tris/Cl, pH 8.0, and 0.66 mM MgCl 2 . To stop the reaction, samples were heat-inactivated by incubation at 65°C for 20 min. The lengths of the deletions were determined by the treatment of aliquots from the reactions with mung bean nuclease to cleave single-stranded DNA overhangs and by comparing them with the size of the wild-type unmodified template. After purification of the modified DNA by standard methods, the fragments were either microinjected into oocytes or subjected to second strand synthesis in a cell-free egg extract derived from Xenopus laevis.
In Vitro Methylation of DNA-All CpG sites in the plasmid PGEMH2BLuc were methylated in vitro using SssI methylase following the instructions of the manufacturer (New England Biolabs). After phenol-chloroform extraction, the DNA was precipitated. The efficiency and amount of methylation were checked by restriction analysis with the methylation-sensitive restriction endonuclease HpaII.
Preparation of X. laevis Low Spin Supernatant Extract and Second Strand Synthesis-Female frogs were primed with 200 IU of human choriogonadotropin, and eggs were dejellied and crushed in a buffer containing 50 mM Tris/Cl, pH 7.5, 50 mM KCl, 2.5 mM MgCl 2 , and 2 mM 2-mercaptoethanol as described (21). 5 l of egg extract (50 g of protein), 500 M each dCTP, dTTP, and dGTP, and 0.5 Ci of [ 32 P]dATP were mixed with 500 ng of exonuclease III-digested template, and the reaction was allowed to proceed for 2 h at room temperature. The buffer system used contained 10 nM ␤-glycerophosphate, 3 mM creatine phosphate, 5 mM MgCl 2 , 3 mM ATP, and 10 ng creatine kinase. The reactions were stopped with proteinase K and incubation for 1 h at 37°C. DNA fragments were further purified using the Rapid PCR Purification system according to the manufacturer's instructions (Life Technologies, Inc.).
Microinjection into X. laevis Oocytes and Primer Extension-Oocytes preparation and DNA and RNA microinjections were performed as described (22). In vitro transcribed mRNA (2-5 ng) encoding the Gal4hybrid protein was injected into the cytoplasm, and 6 -8 h later 1 ng of DNA template was injected into each oocyte nucleus. Total RNA of the injected samples was precipitated, and the pellets were washed with 70% ethanol and dissolved in 20 l of diethyl pyrocarbonate-treated water. The transcription was monitored by primer extension as described (22). An oligonucleotide complementary to the luciferase gene from ϩ126 to ϩ154 was used for primer extension analyses (5Ј-A-TGTTCACCTCGATATGTGCATCTGTAAA-3Ј).
DNase I-hypersensitive Site Analysis-Oocytes (20/sample) were injected with mRNA and template DNA as described above and homogenized in DNase I digestion buffer (10 mM HEPES, pH 8.0, 50 mM KCl, 5 mM MgCl 2 , 1 mM dithiothreitol, 0.1% Nonidet P-40, and 5% glycerol). The oocyte suspension was divided into four fractions and subsequently digested with 10 -100 units of DNase I (Life Technologies, Inc.) at room temperature for 1 min and was stopped with 1 volume of 1% SDS/20 mM EDTA. The samples were then treated with 3 l of DNase I-free RNase (Roche Molecular Biochemicals) for 30 min at 37°C, and proteinase K was added for 2 h at 55°C. Proteins were extracted with 1 volume of phenol-chloroform, and the clear supernatants were precipitated with isopropanol and washed with 70% ethanol. The DNA pellets were dissolved in 20 l of TE buffer (50 mM Tris/HCl, 1 mM EDTA, pH 7.6) and subjected to restriction digestion with 20 units of KpnI/sample. Digested DNA samples were analyzed by electrophoresis using 2% agarose gels, blotted onto Hybond-N membranes, and hybridized with a random primed probe comprising the KpnI/HindIII fragment of the subcloned Gal4-H2B-Luciferase insert.
Micrococcal Nuclease Treatment of Oocytes-Injected oocytes (20/ sample) were homogenized in MNase buffer (DNase I buffer ϩ 3 mM CaCl 2 ). The oocyte extract was divided into four fractions and subsequently digested with increasing amounts (1-10 units) of MNase (Worthington) at room temperature for 15 min. Reactions were stopped with 1 volume of a buffer containing 1% SDS/20 mM EDTA. DNA was prepared according to standard methods and analyzed by electrophoresis, blotted, and hybridized as described above.

Establishing a System Suitable for Analysis of the Transcription of Replicating
Methylated DNA-To examine whether transacting factor binding interferes with the process of DNA remethylation during the time course of replicative processes, we used a single-stranded methylated H2B minimal promoter fused to five upstream Gal4 binding sites to assay for second strand synthesis and transcription (Fig. 1A). Because it is presently not possible to methylate cytosines efficiently within CpG sequences of single-stranded DNA using a bacterially derived methylase, we methylated double-stranded DNA and then generated heteroduplex DNA using exonuclease III, which creates unidirectional deletions starting from the 3Ј end of the template. After linearization of the plasmid DNA with SmaI and KpnI and subsequent treatment with exonuclease III, we were able to generate deletions ranging from 400 to 3000 base pairs (Fig. 1B). Using a low spin supernatant derived from X. laevis eggs (containing the functional replication machinery) we were able to demonstrate incorporation of radioactive nucleotides into newly synthesized DNA (Fig. 2, A and B). Cleavage with HpaII and comparison of methylated and non-FIG. 1. Constructs used in the present study and the strategy for generating single-stranded DNA. A, plasmid DNA used in this study was generated as described under "Experimental Procedures." 5ϫ Gal4 binding sites were cloned in front of a complete histone H2B promoter and a partial luciferase sequence. After linearization with KpnI and SmaI, one strand was deleted unidirectionally with exo III. B, the length of deleted DNA under our experimental conditions varies between 1 and 2.5 kilobase pairs (kB) as determined by mung bean nuclease treatment and subsequent religation of samples. Lanes 2-6 represent time points for exo III digestion for 3 (lanes 2 and 3), 5 (lanes 4 and 5), and 10 min (lane 6). Lane 1 represents no digestion. bp, base pairs. methylated templates shows that there is no major change in the methylation status of all HpaII sites in the H2B promoter after DNA synthesis (Fig. 2, C and D, lanes 4-6). Aphidicolin sensitivity indicates that DNA polymerase ␣ is responsible for the polymerizing activity (Fig. 2, C and D, lanes 10-12). We conclude that in vitro methylated and further modified templates are subjected to second strand synthesis carried out by DNA polymerase ␣-primase.
The Efficient Transcription of Linearized Methylated Templates Depends on a Transcriptional Activator-As described previously, DNA topology plays a significant role in transcription and template linearization and has been reported to exert a dramatic decrease in transcription efficiency (23,24). The exact mechanism of this effect is not known and may be explained by a decreased affinity of transacting factors for their cognate sequences. Transcription of microinjected linearized DNA into frog oocytes is restored by co-injection of strong activators like Gal4-VP16 (Fig. 3A, lanes 8, 9, 12, 13, 16, and  17). In contrast, when we use a supercoiled circular template, the stimulation of transcription caused by the activator is less pronounced (Fig. 3A, compare lanes 2 and 3 with lanes 4 and 5). Template methylation abolishes transcription completely in the absence of the activator (Fig. 3A, lanes 6 and 7) (25). Gal4-VP16 is capable of binding to methylated DNA when present before chromatin assembly rendering the microinjected templates transcriptionally active (Fig. 3A, lanes 8, 9, 12, 13, 16, and 17). In addition, microinjected linear template DNA is sufficiently stable in the oocyte to allow efficient packaging into nucleosomal arrays during an incubation of 12 h after injection (Fig. 3B).

Gal4-VP16 Can Overcome Methylation-dependent Repression of Transcription by Binding during Second Strand Synthesis-
The key question to be addressed in our assays is whether binding of a transcriptional activator to its cognate sequence can overcome transcriptional repression caused by cytosine methylation concomitant with replication-coupled chromatin assembly during the step of DNA synthesis. We microinjected the RNA for Gal4-VP16 6 h prior to the injection of template DNA to enable the activator to be present during the synthesis of the new strand and replication-coupled positioning of the nucleosomes. The presence of Gal4-VP16 renders both the methylated double-stranded control template and the partially single-stranded methylated template transcriptionally active (Fig. 4A, lanes 7 and 9). Transcription depends on the activation domain of VP16 (Fig. 4C, compare lanes 2 and 3 with lanes  6 and 7). We next addressed the effects of the kinetics of chromatin assembly on transcription under these conditions. Double-stranded DNA is slowly packaged into chromatin, whereas single-stranded DNA is rapidly assembled into nucleosomes because of a replication-coupled mechanism (20,21). To test the dependence of chromatinization in our system, RNA encoding Gal4-VP16 and CpG-methylated DNA were injected at different time points (as shown in the flow diagram in Fig.  5A). When present before chromatin assembly, Gal4-VP16 leads to transcriptional activation of both the double-stranded and exo III-treated template (Fig. 5B, lanes 1-4). But when the

FIG. 3. Template linearization leads to loss of transcriptional efficiency in the absence of transcriptional activators like Gal4-VP16.
A, 20 oocytes for each lane were microinjected with 2 and 5 ng of RNA encoding the activator Gal4-VP16 6 h prior to DNA template injection (1 ng). After overnight incubation at 18°C, transcription was assayed via primer extension as described under "Experimental Procedures." H2BLuc denotes the position of the test RNA, and H4 denotes the position of the endogenous H4 RNA as an RNA recovery control. Nonmethylated was compared with methylated DNA, and supercoiled (sc) was compared with linearized (lin) DNA. B, linearized template DNA is effectively packaged into nucleosomal arrays in the oocyte. 12 h after injection half of the oocytes used for the transcription experiments were crushed in micrococcal nuclease buffer and subsequently treated with increasing amounts of micrococcal nuclease (Worthington). The reactions were stopped, and DNA was prepared according to standard procedures and blotted onto Hybond-N membranes after electrophoresis. The membranes were subsequently hybridized against a randomly labeled fragment comprising the inserted DNA used in this study. activator RNA is injected during or after replication-coupled chromatin assembly, the activating potential of Gal4-VP16 is decreased in the case of exo III-treated DNA (Fig. 5B, lanes 7  and 8 and lanes 11 and 12). As shown in a DNase I hypersensitivity assay (Fig. 5D), the accessibility of Gal4-VP16 to its binding sites (and therefore the transactivating potential) decreases once rapid chromatinization after the injection of exo III-treated DNA occurs. DNase I-hypersensitive sites around the H2B promoter are lost when activator RNA is injected 2 h later than the partially single-stranded template (Fig. 5D,  lanes 7 and 8 and lanes 11 and 12).
The Transcriptional State of a Gene Can Be Programmed during Replication of Methylated DNA-To determine whether second strand synthesis of methylated DNA provides a "window of opportunity" to assemble either an active or a repressive transcriptional state, we microinjected RNA encoding the MeCP2 repression domain, fused to a Gal4-DNA binding domain, into oocytes before injection of the nuclease-modified template DNA. MeCP2 has been reported to repress transcription in vivo by recruiting histone deacetylases (9,10). Incubation of the injected oocytes with increasing amounts of trichostatin A, a potent inhibitor of histone deacetylases, relieved the transcriptional inhibition because of endogenous histonedeacetylase repressor complexes recruited to methylated DNA as compared with injection of non-trichostatin A-treated oocytes (Fig. 6, compare lane 2 with lanes 3-6). However, in the presence of the hybrid MeCP2-Gal4, this alleviation of repression by trichostatin A is less pronounced (lanes 8-11). One explanation for this finding is the specific targeting of abundant protein repressor complexes to Gal4 binding sites upstream of the histone H2B promoter. The process of replication or second strand synthesis is therefore a time window for commitment of the transcriptional status toward repression.
Gal4-VP16 Expression Does Not Lead to Demethylation during Second Strand Synthesis-To determine whether binding of Gal4-VP16 precluded remethylation of newly synthesized templates, we performed restriction analyses to study the methylation status of microinjected DNA after the microinjection of either double-stranded or exonuclease III-treated singlestranded templates (Fig. 7). It is apparent that the bulk of template treated with exonuclease is size-reduced (Fig. 7A,  compare lanes 1 and 3). In contrast, when injected into oocytes followed by purification and subsequent restriction, the fulllength HindIII fragment appears, indicating the resynthesis of the second DNA strand (Fig. 7B, lanes 4, 10, and 12). The sensitivity of HpaII against restriction cleavage is restored in the case of the DNA that is injected in single-stranded methylated form (compare lanes 6 and 12). This indicates that no complete demethylation of both strands occurs either in the absence (Fig. 7B) or presence (Fig. 7C) of functional Gal4-VP16. On the contrary, no apparent de novo methylation is observed for the injected nonmethylated template (Fig. 7B, lanes 4-6), from which we conclude that demethylation is not induced by transcription factor binding near the HpaII sites within the H2B promoter in the frog oocyte during ongoing synthesis of the second strand under our experimental conditions.  1, 2, 5, 6, 9, and 10 versus lanes 3, 4, 7, 8, 11, and 12). A quantification of the results is shown in C. A.U., arbitrary units. D, DNase I-hypersensitive site analysis for the conditions shown in B. After an incubation of 12 h the oocytes were crushed in DNase I buffer and subsequently treated with increasing amounts of DNase I (Promega). The reactions were stopped, and DNA was purified and digested with KpnI to counteract any possible religations that may have occurred in the oocyte. After electrophoresis DNA fragments were blotted onto a Hybond-N membrane and hybridized against a randomly labeled probe hybridizing from the beginning of the luciferase sequence to the Gal4 binding sites. Arrows denote the sites located in the promoter region of the template, which seem to be hypersensitive against DNase I treatment. H2BLuc denotes the position of the test RNA, and H4 denotes the position of the endogenous H4 RNA as an RNA recovery control.

DISCUSSION
There are two major findings made in this study. First, transcriptional activators or repressors are capable of programming the transcriptional state in the context of methylated CpG dinucleotides within regulatory sites during DNA synthesis. Second, upon binding of transactivators, no demethylation events were observed on newly synthesized CpG-methylated DNA, indicating that the target sites for DNMTs are accessible even in the presence of site-specific DNA-binding proteins.
DNA Methylation, Methyl CpG-dependent Gene Silencing, and the Replication Machinery Are Associated-Components of the DNA methylation machinery such as DNMT1 associate with replication foci (26). DNMT1 forms a complex with proliferating cell nuclear antigen in vivo (27). In addition, the inhibition of DNMT1 diminishes DNA replication, and DNMT1 interacts with histone deacetylases and the transcriptional repressor DMAP1 at replication foci (28,29). It has been proposed that remethylation of newly synthesized CpG islands is directly coupled to the migrating replication fork. After DNA methylation and once chromatin is assembled, methyl-CpGbinding proteins could recruit histone deacetylases (9,10). Methyl-CpG-binding proteins like methyl-CpG-binding protein-2 and -3 have been reported to bind hemimethylated DNA at the replication fork in late S phase (30). DNA methylation at specific cytosine residues might also serve as a signal recognized by the replication machinery, and this may lead to initiation of replication at more specific genomic sites. This mechanism may be used only by a subset of origins because not every origin of replication appears to have methylated CpG sequences (31).
DNA Demethylation during Replication-Little is known concerning factors that are involved in regulation of the cellular methylation pattern. DNA methylation at the specific dinucleotide sequence "CpG" plays a key regulatory role for the transcriptional activity for some loci and potentially for other nuclear processes like replication, DNA repair, and recombination (31,32,33). In addition, aberrant methylation of CpG islands in the vicinity of active promoters correlates strongly with the escape from normal cell cycle regulation and therefore oncogenesis (34). Under normal physiological conditions, these CpG islands are methylation-free except for sequences located on the inactive X chromosome where CpG methylation is part of the cellular dosage compensation mechanism in females (reviewed in Ref. 35). Progress has been made in elucidating the underlying mechanisms for specification and maintenance of an unmethylated CpG island. Protein-DNA interactions at specific CpG dinucleotides are crucial for defining the unmethylated state. In those studies the oriP region of the Epstein-Barr virus was used to demonstrate a site-specific demethylation once the specific factor Epstein-Barr nuclear antigen-1 was bound followed by subsequent replication (18).
DNA demethylation was shown to be a two-step process requiring transacting factor association and replication of a specific sequence. Other transcription factors were shown to operate in an analogous manner in other systems. It has been demonstrated for X. laevis embryos that a massive demethylation occurs at specific regulatory sites of reporter genes after the mid-blastula transition (15). These observations lead to the hypothesis that the replicative processes concomitant with the binding of transcription factors play a role in keeping a CpG sequence free of methylation. An alternative hypothesis is that active demethylating proteins are involved in establishing the demethylated state. The existence of such demethylases has been controversial (10,11,36). At present, to our knowledge none of the studies described attempted to address the question of how the transcriptional state of a gene is affected once the regulatory site for a transcription factor has been methylated and DNA synthesis of the second strand is ongoing. Two possible scenarios can be envisaged. (i) Binding of the transcriptional regulator and the remethylating activity like DNMT1 would compete for their target sites, depending on the affinity toward the particular site, and the influence of remethylation may dominate over transcription factor binding, resulting in transcriptional repression. (ii) Another possibility is that the transacting factor would dominate over remethylating activities and therefore program the transcriptional state, allowing rapid initiation of transcription.
It is technically challenging to set up an experimental system that allows the investigation of newly synthesized methylated DNA and its effect on gene transcription. Because of the low activity of bacterial CpG methylases on single-stranded DNA, in vitro CpG methylation of single-stranded circular DNA is not efficient. To circumvent this problem we linearized plasmid DNA harboring regulatory sequences and then in vitro methylated the template DNA with the methylase SssI. To generate a partially singlestranded methylated template, we digested the linear DNA with exonuclease III, giving rise to unidirectional deletions of one strand as a function of time. The resulting heteroduplex DNA was used as a template in oocyte microinjection studies. From initial studies using oocyte microinjections, we concluded that linearized DNA is sufficiently stable in the oocyte after a 12-h incubation and is assembled into nucleosomes (see Fig. 3B). However, as reported by others, linearization of supercoiled plasmid DNA leads to a decrease in transcriptional efficiency (23,24). Our studies confirmed these previous findings, but it became clear that binding of a strong transcriptional activator such as Gal4-VP16 is capable of overcom- ing the negative influence of template linearization on transcription. We tested other systems such as the thyroid hormone receptor-␤-promoter and found that transcriptional efficiency is completely lost after linearization (data not shown).
Using single-stranded circular DNA as a template for transcription, earlier studies have demonstrated that transcription complex formation and nucleosome assembly, involving all four core histones, are competitive processes (21). Via its interaction with the replication protein-proliferating cell nuclear antigen, the chromatin assembly factor-1 directs chromatin assembly of newly replicated DNA toward repression (37,38). Furthermore, it has been proposed that heavily methylated DNA-like repetitive sequences and satellite DNA replicate late in S phase and that transcription of these sequences is down-regulated. It is tempting to speculate that in order to render methylated genes transcriptionally active replication processes should provide a time window allowing genes to be programmed by functional transcription complexes. From our studies it became clear that the binding of transcriptional activators or repressors during the formation of the second strand and targeting them to a specific regulatory site are sufficient to direct the transcriptional status of a methylated promoter, perhaps by directly interfering with the chromatin assembly factor-1 chromatin assembly pathway.