Concurrent Replication and Methylation at Mammalian Origins of Replication

Abstract

Observations made with Escherichia coli have suggested that a lag between replication and methylation regulates initiation of replication. To address the question of whether a similar mechanism operates in mammalian cells, we have determined the temporal relationship between initiation of replication and methylation in mammalian cells both at a comprehensive level and at specific sites. First, newly synthesized DNA containing origins of replication was isolated from primate-transformed and primary cell lines (HeLa cells, primary human fibroblasts, African green monkey kidney fibroblasts [CV-1], and primary African green monkey kidney cells) by the nascent-strand extrusion method followed by sucrose gradient sedimentation. By a modified nearest-neighbor analysis, the levels of cytosine methylation residing in all four possible dinucleotide sequences of both nascent and genomic DNAs were determined. The levels of cytosine methylation observed in the nascent and genomic DNAs were equivalent, suggesting that DNA replication and methylation are concomitant events. Okazaki fragments were also demonstrated to be methylated, suggesting that the rapid kinetics of methylation is a feature of both the leading and the lagging strands of nascent DNA. However, in contrast to previous observations, neither nascent nor genomic DNA contained detectable levels of methylated cytosines at dinucleotide contexts other than CpG (i.e., CpA, CpC, and CpT are not methylated). The nearest-neighbor analysis also shows that cancer cell lines are hypermethylated in both nascent and genomic DNAs relative to the primary cell lines. The extent of methylation in nascent and genomic DNAs at specific sites was determined as well by bisulfite mapping of CpG sites at the lamin B2, c-myc, and β-globin origins of replication. The methylation patterns of genomic and nascent clones are the same, confirming the hypothesis that methylation occurs concurrently with replication. Interestingly, the c-myc origin was found to be unmethylated in all clones tested. These results show that, like genes, different origins of replication exhibit different patterns of methylation. In summary, our results demonstrate tight coordination of DNA methylation and replication, which is consistent with recent observations showing that DNA methyltransferase is associated with proliferating cell nuclear antigen in the replication fork.

DNA methylation at cytosine residues at the CpG dinucleotide sequence is now recognized as an important mechanism of epigenetic regulation of genomic function (25–27, 36). Although methylated cytosines, in contrast to other forms of epigenetic control, are part of the covalent structure of the genome, they are inherited by a postreplicative enzymatic transfer of methyl groups from S-adenosyl methionine, which is catalyzed by DNA methyltransferase (MeTase) (1). Unresolved questions involve how the replication of epigenetic and genetic information is coordinated and whether DNA methylation plays a regulatory role in mammalian DNA replication. An interesting biological example of a role for DNA methylation in regulating DNA replication occurs at the origin of replication of Escherichia coli, oriC (5, 6, 29). Methylation of oriC by the Dam MeTase lags 8 min behind its replication, maintaining it at a hemimethylated state throughout replication. The origin is sequestered by the bacterial plasma membrane (6), making it inaccessible to the limiting levels of Dam MeTase available in the cell (31). This hemimethylated state inhibits reinitiation from the origin before a full round of replication is completed. The challenge of preventing reinitiation from multiple origins of replication in eukaryotic cells is far greater than the one facing E. coli. It is possible that eukaryotic cells have developed a similar function for DNA methylation (30). An additional control mechanism in E. coli that is dependent on a lag between replication and methylation is methyl-directed mismatch repair, whereby strand discrimination is based on the difference in methylation between the nascent and parental strands (21).

A recent report has identified cell cycle-dependent densely methylated islands at two chromosomal origins of replication: ori-β, which is located ∼17 kb downstream of the dihydrofolate reductase (DHFR) locus in Chinese hamster ovary cells; and ori-RPS14, which is located at the 5′ region of the ribosomal protein 14 (RPS14) locus (35). These densely methylated islands were reported to contain cytosine residues which were fully methylated at all four possible dinucleotide contexts (CpA, CpC, CpG, and CpT). Methylation of cytosines located in dinucleotide sequences other than CpG has been reported before (39), but the specific concentration of these methylated sequences in chromosomal origins of replication has raised the interesting possibility that they might be involved in regulating specific functions during replication, such as attachment to the nuclear matrix, licensing of specific origins, and inhibiting reactivation of origins (35). However, a more recent report has demonstrated the presence of a high-density cluster of cell cycle-independent, methylated CpG dinucleotides on the 5′ region of both ori-β and ori-RPS14 (28), but methylation of C in other dinucleotide sequences was not observed. It was postulated that methylated CpG clusters mark specific origins for replication through changes in chromatin structure (28). Other origins of replication have not been examined. It is not clear, however, whether heavy methylation is a basic structural characteristic of any origin, as has been suggested elsewhere (28), or whether different origins are differentially methylated, which is consistent with a regulatory role for DNA methylation in origin function. The methylation pattern of one origin might reflect the idiosyncrasy of this origin rather than a general rule.

The kinetics of methylation after replication of mammalian DNA has been previously examined (10, 40), but the results appear to be conflicting. One report has suggested a lag between replication and methylation of as much as 6 h in some parts of the chromosome (40), while others have suggested a much shorter lag of approximately 1 min (10). The identification of fork targeting sequences in the DNA MeTase directing it to sites of replication (17) is consistent with the model that replication and methylation can occur concurrently. Another recent finding (8) showed that the DNA methyltransferase is able to associate with the proliferating cell nuclear antigen (PCNA) and to compete with p21 for its binding site, further supporting the hypothesis that methylation patterns are inherited as replication proceeds. However, the targeting of DNA MeTase to the replication fork and its ability to associate with PCNA do not exclude the possibility that origin sequences are protected from methylation during replication, as they are in E. coli.

To elucidate the role of DNA methylation in replication and to understand how the pattern of methylation is inherited, one has first to determine the kinetics of methylation of newly synthesized DNA containing origins of DNA replication and its dinucleotide sequence specificity. To address this issue, we isolated newly synthesized DNA containing origins of replication from primary and transformed cell lines by extrusion of nascent DNA using a previously established protocol (42), and we used it as a substrate for both nearest-neighbor analyses and bisulfite mapping. This enabled us to look at the rate of methylation in the growing replication fork within a few hundred base pairs from the points of initiation. Since nonsynchronized growing cells were used, these origin-enriched DNA samples include an accurate representation of all active origins in these cells. Using these approaches, we directly measured the methylation status of active origins of replication and Okazaki fragments at the dinucleotide level in primary and transformed cell lines. This study establishes the temporal relationship between initiation of replication and propagation of genetic and epigenetic information encoded by the genome.

MATERIALS AND METHODS

Cell culture.

HeLa and CV-1 cells were purchased from the American Type Culture Collection; primary African green monkey kidney cells and human normal skin fibroblasts were purchased from BioWhittaker. All cells were maintained as monolayers and were grown to approximately 30% confluence in alpha modified Eagle medium supplemented with 10% fetal bovine serum (Flow Lab., McLean, Va.).

Extrusion of nascent DNA.

To isolate sequences of DNA located in close proximity to points of initiation of replication, nascent strands were extruded by branch migration by using previously described methods (12, 34, 42), with slight modifications. The harvested (mid-log-phase) cells were washed three times in 10 ml of ice-cold phosphate-buffered saline and lysed in 4 ml of Hirt lysis buffer (11) with gentle shaking. The lysates were decanted, 0.1 mg of proteinase K per ml was added, and the samples were incubated at 37°C overnight. Following phenol-chloroform extraction and ethanol precipitation, the DNA was dissolved in TE buffer (10 mM Tris, 1 mM EDTA [pH 8]). The nascent DNA strands were extruded by incubation at 50°C for 16 to 18 h. To isolate sequences located at different distances from replication initiation points, the nascent DNA was size fractionated on a 5 to 30% sucrose gradient (0.2 M NaCl, 10 mM TE [pH 8], 0.02% sodium azide) by centrifugation at 24,000 rpm in an SW27 rotor at 9°C for 16 to 18 h and then precipitated in ethanol, dissolved in TE, and analyzed by electrophoresis on a 1% agarose gel (see Fig. 1A). For both the bisulfite mapping and the competitive PCR amplification reactions, the nascent DNA was further size selected (0.4 to 1.2 kb) by gel electrophoresis purification with a Sephaglas BandPrep kit (Pharmacia Biotech).

FIG. 1.

(A) Fractionation of nascent DNA from human normal skin fibroblasts. DNA was prepared and extruded by branch migration as described in Materials and Methods from human normal skin fibroblasts and size fractionated on a sucrose gradient. The fractions were electrophoresed on a 1% agarose gel and ethidium bromide stained. Okazaki fragments (fractions 1 to 4) as well as higher-molecular-weight nascent DNA (fractions 8 to 10) were used as substrates for nearest-neighbor analysis Lane 1 (M), a 100-bp ladder marker; lanes 2 to 14, nascent DNA fractions with increasing molecular weights. (B) Fractions 1 to 4 containing mostly Okazaki fragments. DNA from fractions 1 to 4 was 5′ labeled with [γ-³²P]ATP with PNK and subjected to alkaline treatment with 0.4 M NaOH. The alkaline-treated and untreated samples were electrophoresed through a 5% polyacrylamide alkaline gel. An autoradiogram of the dried gel is shown. The lability of the 5′ label in NaOH suggests that most of the DNAs in these fractions bear RNA nucleotides as expected from Okazaki fragments.

Competitive PCR analysis.

To verify that our method isolates highly enriched origin DNA, we determined whether our nascent DNA fractions contained non-origin-derived DNA by the highly sensitive competitive PCR analysis as previously described (34). Both nascent (≈20 to 260 ng) and genomic HeLa DNAs (≈100 ng) were used as a template for PCR amplification reactions. Two primer sets from human c-myc (GenBank accession no. J00120) were used. The first set (5′-TGCCGTGGAATAACACAAAA-3′ [sense, starting at bp 761], 5′-CTTTCCAGGTCCTCTTTCCC-3′ [antisense, starting at 1134], and 5′-TAACACAAAAGATCATTTCAGGGAGCAAAC-3′ [primer used to design competitor at the c-myc ori]) amplifies the region of the c-myc origin of replication (34, 37), and the second set (5′-GGTTCTAAGATGCTTCCTGG-3′ [sense, starting at 7848], 5′-ATGGGTCCAGATTGCTGCTT-3′ [antisense, starting at 8299], and 5′-TGCT TCCTGGGAGAAGGTGAGAGGTAGGCA-3′ [primer used to design competitor downstream of the c-myc ori]) amplifies a region located 6,711 bp downstream from the first set of primers. PCR conditions were as follows: 94°C for 1 min, 55°C for 1 min, and 72°C for 1 min (30 cycles).

Nearest-neighbor analysis.

Nearest-neighbor analysis was performed as previously described (26, 33). A 1-μg amount of nascent DNA from each of the samples was incubated at 37°C for 15 min with 0.1 U of DNase I. Then, 1 μl of either α-³²P-labeled (10 mCi/ml; Amersham) dATP, dCTP, dGTP, or dTTP was added, together with 1 U of Kornberg DNA polymerase, and the mixture was incubated for 15 min at 30°C. A 30-μl volume of water was added to the reaction mixture, and the unincorporated nucleotides were removed by spinning through a Microspin S-300 HR column (Pharmacia Biotech, Inc.). The labeled DNA was digested with 70 μg of micrococcal nuclease (Pharmacia) in the manufacturer’s recommended buffer for 10 h at 37°C. The samples were loaded on thin-layer chromatography (TLC) phosphocellulose plates (13255 cellulose; Eastman-Kodak), and the 3′-mononucleotides were separated in the first dimension (isobutyric acid-H₂O-NH₄OH [66:33:1]) and in the second dimension [(NH₄)₂SO₄-isopropanol-Na acetate [80:2:18]). Two labeled controls were used to indicate the relative migrations of [³²P]methyl-dCMP and [³²P]dCMP. Fully methylated methyl-dC-dG or nonmethylated dC-dG double-stranded oligomers were labeled with [α-³²P]dGTP and digested to 3′-methyl-dCMP or 3′-dCMP as previously described (33). The chromatograms were exposed to Fuji phosphoimaging plates and scanned in a BAS 2000 PhosphorImager, and percentages of corresponding cytosines and 5-methylcytosines were calculated after respective quantifications. In general, the standard deviation of the assay was in the range of 1 to 3%.

Okazaki fragment identification.

Fractions 1 to 4 from the sucrose gradient contained DNA with the size of the Okazaki fragment (see Fig. 1A). The 5′ ends of the Okazaki DNA fraction (100 to 250 bp), which contained RNA primers, were labeled with polynucleotide kinase (PNK) for 1 h with 5 μl of [γ-³²P]ATP (10 mCi/ml; Amersham). The sample was separated into two; the first half was treated with 0.4 M NaOH, and the other half was left untreated. The two halves were fractionated by 5% polyacrylamide alkaline gel electrophoresis and were analyzed by autoradiography (Fig. 1B).

Bisulfite mapping.

Bisulfite mapping was performed as described previously, with small modifications (7). A 3.6 M solution of sodium bisulfite (ACS grade, pH 5; Sigma) was prepared fresh each time, and a 20 mM stock of solution of hydroquinone was prepared and stored at −20°C. A 5-μg amount of DNA (digested with EcoRI) was incubated for 15 min at 37°C with 60 μl of 0.3 N NaOH. Following this incubation, 431 μl of a 3.6 M Nabisulfite–1 mM hydroquinone solution was added. A 100-μl volume of mineral oil was added to overlay the solution, and the tube was heated at 55°C for 12 h. The bisulfite reaction was recovered from beneath the mineral oil and desalted by using the Promega Wizard Prep (following the manufacturer’s protocol). A 6-μl volume of 3 N NaOH was added to the desalted solution, and the tube was incubated for 15 min at 37°C. Following ethanol precipitation (in the presence of 0.3 M NH₄OAc), the DNA was resuspended in 100 μl in double-distilled H₂O. Approximately 50 ng of DNA was used in each of the PCR amplifications. PCR products were used as templates for subsequent PCRs utilizing nested primers. The PCR products of the second reaction were then subcloned with the Invitrogen TA cloning kit (we followed the manufacturer’s protocol), and the clones were sequenced with the T7 sequencing kit (we followed the manufacturer’s protocol [procedure C]). The primers used for the c-myc origin (GenBank accession no. J00120) were MYC(IN)1 (5′-CCTTTCCCAAATCCTCTTTCC-3′ [positions 1135 to 1116]), MYC(in)2 (5′-GTGAGGGATTAAGGATGAGA-3′ [positions 721 to 730]), MYC(OUT)1 (5′-AACCATTAACTCTTTCCTCC-3′ [positions 1178 to 1159]), and MYC(out)2 (5′-TTAAAATGTTTTTGGGTGAGG-3′ [positions 706 to 726]). The primers used for the human lamin B2 origin (GenBank accession no. M94363) were HL.B2.IN.A (5′-AAAAAAAAACCCTAACTTAACC-3′ [positions 4372 to 4349]), HL.B2.OUT.A (5′-AAAAACTACAACTCCCACAC-3′ [positions 4502 to 4483]), HL.B2.OUT.S (5′-TTTTTAAGAAGATGTATGTTTAG-3′ [positions 3871 to 3893]), and HL.B2.IN.S (5′-TTAATGATTTGTAATATATATTTTAT-3′ [positions 3852 to 3876]). The primers used for the human β-globin origin of replication (GenBank accession no. HUMHH 73008) were HBG.IN1 (5′-TTTTTTGGGGATTTGTTTATTTTT-3′ [positions 62449 to 62473]), HBG.OUT1 (5′-TTAGGTTGTTGGTGGTTTATTT-3′ [positions 62404 to 62426]), HBG.IN2 (5′-AAAATATTTCCTTTTATTATACACA-3′ [positions 62910 to 62885]), and HBG.OUT2 (5′-TCCAAATAATAATATACTAAACAAA-3′ [positions 62974 to 62950]).

RESULTS

Methylation of cytosines in CpG dinucleotides at origins of replication occurs concurrently with their replication.

How fast are origins of replication methylated relative to initiation of replication? To answer this question while avoiding biases arising from particular methylation patterns associated with specific origins, such as, for example, ori-β and ori-RPS14, we chose first to examine a population of origins represented in origin-enriched DNA (12, 34, 42). Nascent DNAs, which can be extruded from bulk DNA due to the unique physical properties of the replication bubble (34, 42), were prepared from human and monkey cell lines in order to establish whether methylation at origins of replication represents a species-specific property. Logarithmically growing CV-1 and HeLa cells were used in order to obtain a representative sample of the population of functional origins of replication in these cells. Extrusion of nascent DNA (12, 34, 42) was followed by size fractionation of the DNA in a neutral sucrose gradient. Recent reports have demonstrated that additional steps, such as labeling the nascent DNA with bromodeoxyuridine (BrdU) followed by anti-BrdU immunoprecipitation, are unnecessary (15, 34). Fractions 8 to 10, comprising newly synthesized DNA of 500 to 1,000 bp, were used in the nearest-neighbor analyses (Fig. 1A).

The nascent DNA fraction containing 100 to 250 bp should contain mostly Okazaki fragments. In order to examine this supposition, we alkali treated the corresponding DNA fractions (100 to 250 bp) in order to degrade the 5′ RNA primers from the Okazaki fragments. The results from Okazaki identification (Fig. 1B) show that nearly all of the labeled 100- to 250-bp fraction is sensitive to alkali treatment, suggesting that >90% of this fraction is indeed composed of Okazaki DNA.

We verified by competitive PCR that our isolation protocol results in fractions that are highly enriched for nascent DNA located within ∼500 bp from the points of initiation of DNA replication and demonstrated that the enriched nascent DNA does not contain non-origin-related genomic DNA. As shown in Fig. 2, sequences bearing the point of initiation of the c-myc origin of replication (34, 37) are amplified from the nascent DNA fractions (Fig. 2A), whereas a sequence located approximately 7 kb downstream of the c-myc origin is not amplified from the same fractions (Fig. 2B). Competitive PCR amplifications were performed to exclude nonspecific inhibition of the amplification reaction as an explanation for the absence of the amplified product and to normalize the reaction product to a known standard as described in Materials and Methods. The fact that no genomic DNA was detected by competitive PCR in our nascent fractions indicates that our nascent DNA was not contaminated with broken fragments of genomic DNA.

FIG. 2.

Nascent DNA is highly enriched for sequences corresponding to sites of initiation of replication. (A) HeLa cell nascent DNA (0 to 13 μl as indicated [≈20 ng/μl]) and 200 molecules of competitor were used as a template for competitive PCR reactions with primers targeted to the c-myc origin of replication. (B) HeLa cell nascent DNA (0 to 13 μl as indicated [≈20 ng/μl]) and 200 molecules of competitor were used as a template for competitive PCRs, but with primers and competitor targeted to a region ∼7 kb downstream of the c-myc origin.

The state of cytosine methylation in CpG dinucleotides in the different DNA fractions was determined as described in Materials and Methods (Fig. 3A). The results (Fig. 3B) demonstrate that newly synthesized DNA located approximately 250 to 500 bp on either side of origins of replication is nearly fully methylated (79% for HeLa cells and 72% for CV-1 cells [Fig. 3B, gray bars]) compared to the state of methylation of genomic DNA (80% for both HeLa and CV-1 cells [Fig. 3B, black bars]). These data suggest that unlike in E. coli, methylation of origin sequences in mammals is initiated before the synthesis of ∼250 bp is completed. This rapid rate of methylation of vertebrate DNA differs from previous published values (10, 40) and might be due to the fact that previous studies did not specifically look at origin-enriched DNA.

FIG. 3.

HeLa and CV-1 cell nascent DNA is rapidly methylated. (A) Nascent DNA fractions 8 to 10 (500 bp to 1 kb) and genomic DNA prepared from HeLa and CV-1 cells as well as a poly(methyl-dCdG) control (which served as a control for migration of 3′-methyl-dCMP) were used as substrates for nearest-neighbor analysis with [α-³²P]dGTP as described in Materials and Methods. Two different reactions, each loaded twice (four lanes per sample), were performed for each cell type. The labeled DNA was digested to 3′ mononucleotides, which were then separated by TLC. The positions of migration of cold mononucleotide standards are indicated. (B) Quantification of a triplicate assay similar to the one presented in panel A. The percentages of methylated cytosines relative to total cytosines were determined by PhosphorImager quantification of the signals obtained for 5′-methyl-dCMP and dCMP. The results are averages of three independent determinations ± standard deviations. Filled boxes, nascent DNA; shaded boxes, genomic DNA; 5 mC, 5′-methylcytosine.

Origins of replication are rapidly methylated irrespective of their states of transformation.

Previous reports have suggested that CpG-rich sequences are hypermethylated in cancer cells (2, 20, 23) and that this hypermethylation reflects an increase in the activity of the DNA methylation machinery (9, 13). It has also been previously suggested that changes in the kinetics of DNA methylation of origins of replication might be involved in cellular transformation (30).

We first determined whether the kinetics of DNA methylation of origins of replication is different in transformed human (HeLa) cells from that in primary normal human skin fibroblasts. Then, we compared the cytosine methylation states of CpG dinucleotides in monkey-transformed (CV-1) cells and African green monkey primary kidney cells. The states of methylation of sequences located ∼500 bp from the point of initiation of replication were compared with the average state of methylation of genomic DNA. The results (Fig. 4A) show that nascent DNA from primary human cells is 68% methylated (gray bar) at CpG sequences, compared to 72% of genomic DNA (black bar). African green monkey primary kidney cells have 52% of their CpGs methylated at the nascent DNA (gray bar), compared to 53% in genomic DNA (black bar). The small difference observed in primary human cells between cytosine methylation of CpG sequences in nascent DNA (68%) and genomic DNA (72%) is close to the standard error of our assay. However, this small difference might alternatively suggest that few CpG sites remain nonmethylated in primary nascent DNA. These results imply that similarly to transformed cells (HeLa and CV-1 cells), primary human and monkey cells initiate methylation immediately after replication. Methylation of cytosine residues residing in the dinucleotide CpG sequences is initiated before ∼500 bp are synthesized following initiation of replication. The fact that the level of methylation of origin DNA is similar to the level of total genomic DNA indicates that, on average, origins are not more methylated than the rest of the genome, as has been previously proposed (35). Whereas the kinetics of methylation after replication seem to be similar in transformed and primary cells, the overall level of methylation of the primary cells in the present study is lower. This result is consistent with the general hyperactivation of DNA MeTase activity observed in cancer cells (3). Alternatively, in the case of HeLa cells versus primary human fibroblasts, the lower level of methylation might reflect the specific cell type of the untransformed cells.

FIG. 4.

(A) Nascent DNA is rapidly methylated in transformed and untransformed cells. Nascent DNA (shaded boxes) and genomic DNA (filled boxes) prepared from CV-1, African green monkey (AGM), HeLa, and human skin fibroblast cells were subjected to nearest-neighbor analysis of DNA methylation at CpG dinucleotides as described in Materials and Methods. The results presented are averages of three determinations ± standard deviations. (B) Methylation of Okazaki fragments. Okazaki fragment containing DNA was prepared as described in Materials and Methods (fractions 1 to 4) and subjected to nearest-neighbor analysis of DNA methylation at CpG dinucleotides. The results are averages of three independent determinations ± standard deviations.

Okazaki fragments are methylated.

To determine whether rapid methylation is characteristic of sequences near or at the point of initiation of replication or whether all nascent DNA is methylated at similar rates, we determined the state of methylation of Okazaki fragments. Okazaki fragments are replication intermediates synthesized on the lagging strand of DNA both near and distal to the point of initiation of DNA replication. To determine that the Okazaki fractions isolated by our procedure are indeed Okazaki fragments rather than sheared genomic DNA, we took advantage of the fact that Okazaki fragments contain RNA primers at their 5′ ends which are sensitive to NaOH treatment (Fig. 1B), whereas genomic DNA is not. Fractions 1 to 4 (Fig. 1A and B), containing DNA with the size of the Okazaki fragment (100 to 250 bp), were collected following sucrose gradient fractionation, and the percentage of CpG methylation was determined as described in Materials and Methods. The results (Fig. 4B) show that Okazaki fragments are partially methylated, suggesting that methylation of DNA is initiated before Okazaki fragments are ligated to form-longer DNAs. Since the majority of Okazaki fragments are derived from the growing points of replication forks and not from origins of replication, this suggests that rapid kinetics of methylation is a feature of all nascent DNA. One interesting observation is that different cell lines show different levels of methylation of Okazaki fragments. Human cells (HeLa cells and human skin fibroblasts) exhibit lower percentages of CpG methylation (30 and 35%, respectively) than African green monkey (CV-1 and primary kidney) cells (65 and 47%, respectively) (Fig. 4B). These differences do not correlate with the state of transformation of the cells. One possible explanation for the partial methylation of Okazaki fragments compared to origin-derived DNA might be a differential association of DNA MeTase with the leading- and lagging-strand replication machinery.

CpG is the only methylated dinucleotide sequence in origins of replication.

A previous report has indicated that two Chinese hamster origins of replication, ori-β and ori-RPS14, bear a high concentration of methylated cytosines that do not reside in the consensus CpG dinucleotide sequence (35), but this observation has not been confirmed. Using the nearest-neighbor assay, which allows determination of the state of cytosine methylation at each of the four possible CpX dinucleotide sequences, we addressed the question of whether origins of replication, in general, have cytosines methylated in dinucleotide sequences other than CpG. Since this assay could detect fewer than 1 methylated cytosine in 100 cytosines (26, 33), a cluster of methylated cytosines per origin would be easily detected. Nascent and genomic DNAs prepared from HeLa cells were labeled with either α-³²P-labeled dCTP, dATP, dGTP, or dTTP, digested to 3′ mononucleotides, and separated by TLC in one or two dimensions. The results show the absence of methylation of CpC or CpT (Fig. 5A) and CpA (Fig. 5B). The control experiment shows where the 5′-methylcytosine migrates in a two-dimensional TLC relative to the unmethylated cytosine (Fig. 5C). These results support the conclusion that methylation of cytosines at the CpG dinucleotide sequence is the main modification of DNA located at origins of replication and is probably carried out by the same enzymatic machinery responsible for methylation of the rest of the genome.

FIG. 5.

CpG is the only dinucleotide sequence methylated in nascent DNA. (A) HeLa cell nascent DNA prepared as described in Materials and Methods (500 to 1,000 bp, fractions 8 to 10) was subjected to nearest-neighbor analysis for CpC (labeled with [α-³²P]dCTP), CpT (labeled with [α-³²P]dTTP), and CpG (labeled with [α-³²P]dGTP) methylation. (B) To study methylation at dCpA sequences, the 3′ mononucleotides were separated in two dimensions as described in Materials and Methods. Two dimensions were used, since dADP (a degradation product of the labeled dATP) comigrates with 5′-methyl-dCMP. Open circle, migration of 5′-methylcytosine. (C) Two-dimensional analysis of dG neighbors showing 80% methylation at CpG dinucleotide sequences. Open circle, migration of the unmethylated cytosine. 5′ mC, 5′-methylcytosine; con., control methylated CpG oligonucleotide.

Specific CpG sites are rapidly methylated.

To determine if specific sites were being methylated concurrently with replication, we decided to perform bisulfite mapping of the lamin B2, the c-myc, and the β-globin origins of replication (Fig. 6 and 7) and compare the methylation patterns of genomic DNA with those of nascent DNA. This technique allowed visualization of methylated cytosines at a single base resolution (7). The lamin B2 origin was found to be partially methylated (Fig. 6A and B and 7) in all clones tested (five nascent and five genomic clones were tested). Interestingly, one specific CpG was found methylated in all five genomic clones (Fig. 6A) and in all five nascent clones (Fig. 6B), supporting our hypothesis that replication and methylation occur concomitantly. The c-myc origin was found unmethylated in all clones tested (five nascent and five genomic clones), both genomic (Fig. 6C) and nascent (Fig. 6D). As a control, we also sequenced this region using nonbisulfited DNA (Fig. 6E). Four of five CpG sites included in the β-globin origin of replication (4, 14, 22) were found to be methylated in all nascent and genomic clones tested (Fig. 6F and G). This result suggests that not all active origins have to be associated with a high-density cluster of methylated CpG dinucleotides as previously suggested (28), but rather that origins are differentially methylated. Some origins, such as the DHFR ori-β and ori-RPS14 and human β-globin, are heavily methylated (28); other origins, such as the lamin B2 origin, are partially methylated, and the c-myc origin is not methylated. This is consistent with the observation that the general level of methylation of CpG dinucleotides residing near origins of DNA replication is not different from that for general genomic DNA.

FIG. 6.

Specific methylated CpG sites are rapidly methylated, but not all active origins are methylated. (A) HeLa cell genomic DNA was used as a template for bisulfite mapping of sites within the lamin B2 origin of replication (positions 3910 to 4100). Lollipops, methylated CpGs; thin lines, unmethylated CpGs. (B) HeLa cell nascent DNA was used as a template for the same assay mapping the same positions of the lamin B2 origin. (C) HeLa cell genomic DNA was used as a template for bisulfite mapping of sites within the c-myc origin of replication (positions 850 to 980). (D) HeLa cell nascent DNA was used as a template for the same assay mapping the same positions of the c-myc origin. (E) Non-bisulfite-treated HeLa cell genomic DNA was used as a template for the same assay mapping the same positions of the c-myc origin, as a control for panels C and D. (F) HeLa cell genomic DNA was used as a template for bisulfite mapping of sites within the β-globin origin (positions 62449 to 62935). (G) HeLa cell nascent DNA was used as a template for the same assay mapping the same positions of the β-globin origin.

FIG. 7.

Methylation patterns at different sites. Lollipops, CpG dinucleotides analyzed in this study (filled lollipops, CpG dinucleotides methylated in all clones tested; shaded lollipops, CpG dinucleotides methylated in 50% of the clones tested); thin vertical lines, unmethylated CpG dinucleotides.

DISCUSSION

Whereas DNA methylation is now accepted as an important epigenetic mechanism for regulation of genome function, the mechanism by which replication of the genome and its methylation are coordinated has been unknown. Different models have been proposed for the possible function that DNA methylation plays in replication (5, 6, 28, 35), but there have been no data to support or nullify these hypotheses. The study presented here defines some of the basic rules that govern methylation patterns at the earliest events in replication, that is, methylation of origins of replication. We first showed that methylation occurs immediately after initiation of replication, where the replication fork has advanced less than 500 bp after the point of initiation of replication. Second, tight coordination of initiation of replication and methylation is a characteristic of mammalian cells regardless of their state of transformation. Third, methylation occurs rapidly also at segments of the genome that are located distal to origins of replication, since methylation occurs before Okazaki fragments are ligated to form longer nascent strands. The level of methylation observed in the lagging strand is less than that occurring in the leading strand. This might reflect a difference between the kinetics of methylation at the origin of replication, which is fully methylated, and at sites distal to it. Either these sites may be less tightly coordinated with replication, or the lagging strand might be under a differential action of the MeTase. Methylation was complete before 3,000 bp of DNA were synthesized (data not shown). Assuming that forks move at 3 kb/min, this would allow a window of only 30 s for the methyl-directed mismatch machinery to work on the hemimethylated template. These results are inconsistent with methylation being a factor in strand discrimination during mismatch repair, in contrast to the mechanism proposed for bacterial cells (21). Fourth, origins of replication are methylated only at the CpG dinucleotide sequence, as is the rest of the genome. This is consistent with the hypothesis that methylation of origin DNA occurs by the same enzymatic machinery that methylates the rest of the genome.

The data indicate a tight coordination between replication and methylation. This coordination is probably maintained to ensure that the pattern of methylation is appropriately inherited. Upregulation (41) and downregulation (16, 19, 24) of DNA MeTase activity have been previously shown to alter cellular phenotype. Furthermore, embryonic deficiency in DNA MeTase expression is lethal (18). Several mechanisms are probably involved in ascertaining that these processes are coordinated. The expression of DNA MeTase is regulated at the posttranscriptional level with DNA synthesis (31), and the DNA MeTase bears a replication fork targeting signal (17). The simplest explanation for the data in this study is that the DNA MeTase is part of the DNA replication fork complex. This hypothesis is strengthened by the observation that the DNA MeTase associates with PCNA (8). The fact that only CpG sequences are methylated is consistent with the conclusion that the known CpG-specific DNA MeTase is the only DNA MeTase included in the replication fork complex. Interesting questions include whether DNA MeTase is limited to the replication fork and, if so, how repair patches are methylated. It was previously shown that methylation of repair patches is an inefficient process (38), probably because of the localization of DNA MeTase to replication forks. Alternatively, another population of DNA MeTase that is not targeted to the replication fork might exist.

Does DNA methylation play a role in regulating the activity of the replication fork? The data here suggest that differential methylation of active origins during the cell cycle does not play a role in regulating origin function, as has been previously proposed (30) based on the E. coli model (5, 6). However, recent observations suggest that clusters of methylated CpGs might be necessary to mark functional origins (28). We show here that, as with genes, different origins are differently methylated. Some origins, such as lamin B2, are partially methylated, some, such as c-myc, are not methylated, and some, such as the previously reported ori-β and ori-RPS14 and the human β-globin, are heavily methylated. Future experiments will determine whether differential methylation of origins plays a role in the regulation of origins of replication. Thus, the number and the time of replication of origins in a cell might be regulated by methylation. It has been previously observed that ectopic expression of DNA MeTase can lead to cellular transformation (41), while inhibition of DNA MeTase can reverse transformation (16, 19, 24). The state of methylation of origins might be one mechanism through which hypermethylation plays a role in carcinogenesis (30).

In summary, the data presented here demonstrate that initiation of replication and methylation are tightly coordinated and that different origins exhibit different patterns of methylation. Whether methylation of specific sites plays a role in regulating replication activity remains an open question.