Absence of an unusual "densely methylated island" at the hamster dhfr ori-beta.

An unusual "densely methylated island" (DMI), in which all cytosine residues are methylated on both strands for 127-516 base pairs, has been reported at mammalian origins of DNA replication. This report had far-reaching implications in understanding of DNA methylation and DNA replication. For example, since this DMI appeared in about 90% of proliferating cells, but not in stationary cells, it may regulate origin activation. In an effort to confirm and extend these observations, the DMI at the well characterized ori-beta locus 17 kilobases downstream of the dhfr gene in chromosomes of Chinese hamster ovary cells was checked for methylated cytosines in genomic DNA. The methylation status of this region was examined in randomly proliferating and stationary cells and in cell populations enriched in the G1, S, or G2 + M phases of their cell division cycle. DNA was subjected to 1) cleavage by methylation-sensitive restriction endonucleases, 2) hydrazine modification of cytosines followed by piperidine cleavage, and 3) permanganate modification of 5-methylcytosines (mC) followed by piperidine cleavage. The permanganate reaction is a novel method for direct detection of mC residues that complements the more commonly used hydrazine method. These methods were capable of detecting mC in 2% of the cells. At the region of the proposed DMI, only one mC at a CpG site was detected. However, the ori-beta DMI was not detected in any of these cell populations using any of these methods.

C in 2% of the cells. At the region of the proposed DMI, only one mC at a CpG site was detected. However, the ori-␤ DMI was not detected in any of these cell populations using any of these methods.
DNA methylation at CpG dinucleotides in mammalian cells has been implicated as an important component of such pivotal processes as transcription (1), imprinting (2), recombination (3), carcinogenesis (4), development (5), and replication timing (6). CpG methylation is carried out by a methyltransferase that is associated with replication foci in the nucleus (7). In general, this enzyme methylates specifically hemimethylated CpG dinucleotides, although de novo methylation of unmethylated CpG dinucleotides can occur but at a low rate (8). In addition, this enzyme occasionally methylates cytosines in sequences other than CpG (9,10). However, given the specificity of the mammalian methyltransferase, this non-target methylation should be of little significance in vivo because the pattern of non-CpG methylation will not be maintained during the subsequent round of DNA replication.
Recently, an unusual form of DNA methylation has been implicated as an important component of mammalian origins of DNA replication. A densely methylated island (DMI) 1 has been reported at three different mammalian origins of DNA replication (11,12). These DMIs were found only in proliferating cell populations and consisted of a 127-, 258-, or 516-bp stretch of DNA in which all cytosines were methylated, regardless of their sequence context (12). The discovery of DMIs has several important implications. First, this unprecedented methylation pattern suggested that a second (or modified) methylation enzyme exists in mammalian cells. Interestingly, limited proteolysis of the mammalian DNA methyltransferase stimulates its ability to methylate unmethylated CpG dinucleotides de novo (13,14). Second, the rapid loss of DMIs when cells stop proliferating implies an active demethylation process. In fact, repair enzymes might be involved in active demethylation (15,16). Third, the characteristics of DMIs strongly suggest a role in initiation of DNA replication. For example, since the 16 known mammalian origins (17) lack a common extensive sequence homology, a DMI may mark initiation sites for DNA replication. Alternatively, DMIs might provide a means to recognize origins that have already replicated. Remethylation at the 11 dam methylation sites at Escherichia coli's ori-C is delayed about 10 min, during which time the hemimethylated ori-C is bound to the outer bacterial membrane in a form that cannot reinitiate replication (18,19). The DMI might play an analogous role in mammalian cells where nuclear structure is required for site-specific initiation of DNA replication (20,21). In fact, site specificity is established during the middle of G 1 phase in each cell cycle (22), about the time when DNA methyltransferase activity appears and then increases during S phase (23)(24)(25). Therefore, it was imperative to confirm the existence of a DMI at a well characterized mammalian replication origin and to determine whether or not a DMI appears at a specific time during the cell division cycle.
The DMIs described above were identified using the bisulfite method to detect 5-methylcytosine ( m C). Bisulfite catalyzes the hydrolytic deamination of cytosines, but not m C, to uracils (26 -28). However, this method suffers from the fact that conversion of cytosines to uracils requires that DNA be in a single strand state (29,30). Thus, failure to properly denature the genomic DNA or partial renaturation during the bisulfite treatment can lead to persistence of cytosines that will be interpreted erroneously as the presence of m C (31). Therefore, we investigated cytosine methylation by three alternative and independent methods. The first method used methylation-sensitive restriction endonucleases to identify specific sites that are methylated and, therefore, resistant to cleavage. Cleavage was monitored by direct visualization of the DNA products using Southern blotting and hybridization. Alternatively, uncleaved genomic DNA was quantitated by PCR amplification. Southern blotting and hybridization imposes the fewest DNA manipulations and provides a direct estimate of the fraction of cells containing a DMI, whereas PCR monitoring allows the detection of very small DMI fractions. However, digestion by restriction endonucleases detects only a subset of potential methylation sites. The second method is able to detect m C at any nucleotide position by modifying DNA with hydrazine and then cleaving the modified cytosines with piperidine. Although hydrazine, like bisulfite, reacts with cytosines, but not with m C (32), hydrazine modification is independent of the strandedness of the DNA. However, because m C is identified by the absence of a band in the cytosine sequence of genomic DNA that is present in the cytosine sequence of cloned DNA (i.e. "negative display" of m C), background cleavage events or closely spaced bands may lead to ambiguity. This problem was overcome by a third method. Permanganate at slightly acidic pH can modify thymines and m C but not cytosines (33). Piperidine then cleaves DNA at these sites ("positive display" of m C). Permanganate reactivity requires that the DNA be single stranded, but in contrast to bisulfite, the strandedness of the DNA can be monitored by the concomitant reaction of thymines, which also takes place only in single-stranded DNA. To our knowledge, the permanganate/piperidine method for detection of m C has never before been applied to complex genomes. Therefore, conditions for the permanganate reaction were optimized for this purpose, and the method is presented here as a valuable tool for investigation of DNA methylation in complex genomes.
Using the methods described above, cytosine methylation was investigated at the DMI locus at ori-␤, an origin of bidirectional replication located 17 kb downstream of the dhfr gene in Chinese hamster ovary (CHO) cells (17). Methylation was examined in randomly proliferating and stationary CHO K1 cells containing two copies of the dhfr gene region per diploid genome and in CHO C400 cells containing ϳ1000 tandem copies of this region. CHO K1 cells were also isolated in the G 1 , S or G 2 ϩ M phases of their proliferation cycle. In no case was the DMI detected, although the sensitivity of the methods allowed detection of m C in 2% of the total DNA.

EXPERIMENTAL PROCEDURES
Cell Culture and Cell Harvesting-CHO were propagated after 1:20 dilution of a confluent starting culture in Dulbecco's modified Eagle's medium (DMEM), supplemented with 0.4 mM L-glutamine and 5 (CHO K1) or 10% (CHO C400) fetal calf serum, at 37°C in 5% CO 2 . Proliferating cells were harvested when they reached ϳ50% confluency (about day 3). To obtain stationary cells (G 0 ), cells at about 30% confluency were washed with PBS and cultured for another 5 days in DMEM ϩ 0.4 mM L-glutamine ϩ 0.5% fetal calf serum (12). Further culturing of cells at low serum, to increase the portion of G 0 , resulted in slow cell death without change of the FACS profile (data not shown). For harvesting, cells were washed with PBS and trypsinized. Trypsin digestion was stopped by adding DMEM ϩ 10% fetal calf serum. The cells were collected by centrifugation at 1000 ϫ g for 10 min at 4°C and resuspended in PBS on ice. The profile of the cell cycle was checked before and after harvesting by FACS analysis (see below). Cells were used for elutriation or immediately for DNA isolation.
Elutriation-Centrifugal elutriation was carried out on 1.2 ϫ 10 8 proliferating growing CHO K1 cells in 5 ml of PBS using a Beckman rotor (JE-5.0) and centrifuge (model J-6M/E) at 2,500 rpm. The flow rate was gradually increased until cells reached the upper region in the chamber. 7-8 fractions were taken at increasing flow rates. The cells were kept on ice, and an aliquot (ϳ1 ϫ 10 5 cells) of each fraction was used for FACS analysis. DNA was prepared from the remaining cells.
FACS Analysis-Prior to harvesting, adherent cells in one 75-cm 2 flask were rinsed once with DMEM lacking serum and stained as described (34) with some modifications. Briefly, 5 ml of sterile cold staining solution (4 mM sodium citrate, pH 7.0, 5 g/ml propidium iodide, 0.5% Triton X-100, and 10 g/ml freshly added RNase A) were added to the cells and then placed at 4°C for 45 min in the dark. Residual adhering cells were dislodged by pipeting. After harvesting or elutriation, the cells were rinsed in PBS, centrifuged at 1,000 ϫ g for 5 min at 4°C, resuspended in the above staining solution, and held at 4°C for 45 min in the dark. The cell suspensions (ϳ1 ϫ 10 6 cells/ml) were analyzed by flow cytometry (Epics XL/MCL 4 Color Coulter). Between 5,000 and 20,000 fluorescent events were counted. The FACS profile of cells treated by this method was stable for at least 2 weeks.
Genomic DNA Extraction, Enzymatic Digestions, and Quantitative Southern-Cells at (ϳ5 ϫ 10 6 per ml in PBS) were mixed with an equal volume of lysis buffer (1% sarcosyl, 75 mM Tris-Cl, pH 8.0, 25 mM EDTA), and 100 g/ml proteinase K was added to the suspension. Cells were lysed by incubation at 50°C overnight. The solution was extracted twice with TE-equilibrated phenol and once with chloroform:isoamyl alcohol (24:1). RNase A was added to 100 g/ml, incubated for 3 h at 37°C, and dialyzed extensively against TE buffer for 2 days. Genomic DNA was concentrated under vacuum to 0.1 of the previous volume and dialyzed once more against 0.1 ϫ TE buffer. After this step, DNA had a final concentration of 0.1 g/l.
Cleavage of genomic DNA with AluI and MboII was performed according to the instructions of the manufacturer. An unrelated ϳ100-bp DNA fragment containing an AluI site, radioactively labeled at one 5Ј-end, was added to the AluI digests of genomic DNA. Completeness of AluI cleavage was monitored by autoradiography after an aliquot of the AluI digestion was subjected to gel electrophoresis. Only the genomic DNA from reactions in which the labeled fragment was completely cleaved was used for Southern blotting (below). A second control for completion of both enzymatic digestions was performed by monitoring the average length of DNA after gel electrophoresis and ethidium bromide staining. In order not to obscure the subsequent Southern blot analysis, the radioactive phosphate was removed from the end-labeled fragment by treatment with calf intestine phosphatase. Digested genomic DNA was phenol-extracted, ethanol-precipitated, and dissolved in TE at 1 g/l.
For Southern blot analysis, equal amounts of AluI-or MboII-cleaved genomic DNA of logarithmically growing or stationary CHO cells were run in parallel with radioactively end-labeled DNA molecular weight markers on a 2.5% BIOZYM "small DNA" agarose gel in TBE buffer. Ten micrograms of CHO K1 DNA and 1 g of CHO C400 DNA were used. After electrophoresis, the DNA was denatured by soaking the gel in 0.2 M NaOH, 5 mM EDTA for 40 min. The gel was neutralized with 0.5 M sodium phosphate buffer, pH 6.5, and equilibrated in 10 mM sodium phosphate buffer, pH 6.5. The DNA was electrophoretically transferred onto a nylon Hybond-Nϩ membrane (Amersham Corp.) in a Bio-Rad Mini-PROTEAN II blotting chamber in 10 mM sodium phosphate buffer, pH 6.5, at 250 mA and 30 V overnight. Complete transfer was verified by comparing the autoradiographs of the labeled DNA marker in the gel and on the membrane prior to hybridization. The transferred DNA was fixed by placing the membrane on Whatman paper soaked in 0.4 M NaOH for 20 min and baking it for 2 h at 80°C. This treatment ensures complete retention of DNA on the membrane during hybridization, as verified by autoradiography of the labeled DNA marker prior to and after hybridization. Genomic DNA was hybridized with end-labeled primers 6C and 6E in Church buffer (32) supplemented with 250 g/ml yeast RNA at 50°C overnight and washed as described (32). Hybridization signals were visualized by autoradiography.
PCR Detection of Cleavage Pattern of Methylation-sensitive Restriction Enzymes at Genomic DNA from Synchronized Cells-Proliferating and serum-arrested CHO K1 cells were grown as described above. Cells grown to early-, mid-, or late-G 1 were harvested at 40, 120, or 195 min, respectively, after re-plating cells obtained by mitotic selection (21). Progression into S phase, which occurred at approximately 4 h after replating, was monitored by incorporation of [ 3 H]thymidine into parallel cultures (21). At the indicated time, cells were washed free of medium using cold (4°C) PBS, released by trypsinization, and counted using a hemacytometer. The cells were collected by centrifugation and placed on ice. After the pellet was rinsed in cold hypotonic buffer (20 mM HEPES, pH 7.6, 5 mM KCl, 1.5 mM MgCl 2 ), the cells were swelled at 5 ϫ 10 6 cells/ml in hypotonic buffer on ice for 5 min. The cells were again collected by centrifugation and resuspended in hypotonic buffer to give a total concentration of 2.5 ϫ 10 4 cells per l. The cell suspension was snap-frozen in liquid nitrogen and stored at Ϫ80°C.
Nuclei were isolated from the frozen cell suspension by addition of 1 ml of cold hypotonic buffer (at this point, all the cells are lysed, leaving nuclei and cellular debris), followed by centrifugation for 3 min at high speed in an Eppendorf centrifuge. The pellet was resuspended to 1 ϫ 10 5 nuclei/l in DNA isolation buffer (20 mM Tris-HCl, pH 7.5, 1 mM EDTA, 0.5% SDS), supplemented with 0.5 mg/ml RNase A, and incubated at 37°C for 3 h. The reaction was then supplemented with 1 mg/ml proteinase K and continued for 3 h. DNA was purified by organic extraction and precipitated with ethanol. After centrifugation, the pellet was air-dried and resuspended to Ͼ600 g/ml in 10 mM Tris-HCl, pH 7.5. DNA fragments averaged 20 to 25 kb in length.
DNA (500 g/ml) was digested with XbaI according to the manufacturer's directions to reduce viscosity. XbaI released a 4.5-kb fragment that contained ori-␤ and the proposed DMI near its center. XbaI-digested genomic DNA (2 g) was digested for 4 h at the appropriate temperature with 10 units of AccI, AluI, Alw26I or MslI using buffers provided by the manufacturer (New England Biolabs, Boehringer Mannheim). Spermidine (1 mM) was included in AccI digestions. These conditions represent a 15-25-fold "over-digestion" with respect to incubation time and units of enzyme. Restriction digestion of PCR-generated fully cytosine-methylated fragments was performed using 8 units of enzyme, 85 ng of DNA, and 6 h incubation time (50 -350-fold over-digestion). PCR reactions were carried out in 20 l using 100 ng of template DNA, 10 pmol of primer T (AGA CAA ATG TCA GCA TGA AGG CAG G), 10 pmol of primer L (TCT CCA GCA CCA TGT CTG CCT GTG G), 0.5 units Takara ExTaq polymerase (Oncor-Appligene), and buffer supplied by the manufacturer. Thirty cycles consisted of 45 s each of 95, 63, and 72°C. The size of the expected product is 418 bp.
Fragment Methylation in Vitro and Cleavage with Enzymatic and Chemical Probes-A DNA fragment of ϳ100 bp encompassing an MspI/ HpaII site was radioactively labeled at one 5Ј-end by standard methods. CpG sites of an aliquot were methylated by the bacterial SssI methylase (New England Biolabs). Methylated and unmethylated fragments were cleaved with MspI (New England Biolabs) or HpaII (Boehringer Mannheim) according to the manufacturer's instructions. Hydrazine/piperidine treatment of methylated and unmethylated fragments was performed as described (35). KMnO 4 /piperidine treatment was as described (33) with the modification that the pH of the sodium acetate buffer was lowered to 4.1. The products of all cleavage reactions were analyzed on a sequencing gel.
Analyses of DNA Methylation in a Synthetic PCR DNA Fragment by Chemical Probes-A ϳ790-bp fragment of the DNA region around the proposed DMI was synthesized by PCR on a cloned DNA template by using primer 1D (Table I) and a second primer 5Ј-ATA ATA AAA AAA CTA GTT TTG AGT CAT TTT ATG G-3Ј (positions 3475-3442) and either dCTP or methyl-dCTP at a final nucleoside triphosphate concentration of 200 M. The PCR product was purified from a low melting point agarose gel by isotachophoresis (36). Hydrazine/piperidine or KMnO 4 /piperidine treatment of fully methylated and fully unmethylated fragments was performed as described above. Guanine cleavage was as described (35). Cleavage products were visualized by LM-PCR.
Analyses of DNA Methylation in Genomic DNA by Chemical Probes-Conditions for base-specific modification of genomic DNA by hydrazine in vitro, followed by piperidine-catalyzed cleavage, have been described previously (32,37). Genomic DNA was cut with PstI to reduce its viscosity before treatment with hydrazine/piperidine. Ligation-mediated polymerase chain reaction (LM-PCR) was carried out as described previously (38) and modified (39,40). Conditions for base-specific modification of genomic DNA by permanganate in vitro followed by piperidine-catalyzed cleavage have been described previously (33) and were modified as follows. High molecular weight DNA was cut with PstI to reduce its viscosity, phenol-extracted, and ethanol-precipitated. A 50-l sample containing 50 g of genomic DNA in TE buffer was heated to 90°C for 2 min and immediately placed on ice. 200 l of 30 mM NaOAc, pH 4.1, was added at room temperature (22°C), before adding 50 l of freshly prepared 6 mM KMnO 4 . The sample was incubated for 3 min at room temperature and terminated by adding 20 l of ␤-mercaptoethanol. DNA was precipitated in 0.3 M NaOAc, pH 5.0, and 2.5 volumes of  Absence of DMI at Replication Origins ethanol, dissolved in TE buffer, and precipitated again. The sample was dissolved in 100 l of 10% piperidine, incubated for 30 min at 90°C, and then quenched at 0°C. Piperidine was removed by vacuum centrifugation overnight, and the DNA was analyzed by LM-PCR (cf. above). In an effort to confirm these results, the methylation state of the putative ori-␤ DMI was investigated using a method that required the fewest manipulations of genomic DNA. DNA was digested with a restriction endonuclease, AluI, that recognizes and cleaves AGCT, a non-CpG site, but is inhibited when the cytosine in either strand is methylated (41). DNA products were then analyzed by Southern blotting and hybridization. AluI has three recognition sites within the DMI region (Fig.  2C). A methylation-insensitive restriction endonuclease, MboII, was used to determine the general accessibility of DNA to restriction enzymes (41). MboII has two recognition sites with the sequence GAAGA⅐TCTTC in the DMI (Fig. 2C). Inhibition of AluI cleavage due to methylation of the recognition sequence and insensitivity of MboII was confirmed by digestion of a PCR fragment spanning the DMI region, which was synthesized in the presence of either dCTP or 5 m dCTP ("Materials and Methods"; data not shown). DNA from CHO K1 cells containing a single copy of the dhfr locus per haploid genome and from CHO C400 cells, which have amplified this locus about 500 times (42), were examined. DNA was extracted from exponentially growing and from serum-deprived, stationary cells, assumed to be in G 0 phase. Cell cycle was examined by FACS analysis (Fig. 2A). Proliferating cells displayed on average a G 1 :S:(G 2 ϩ M) ratio of about 2.5:1.7:1, respectively. After 5 days of serum deprivation, the peak reflecting the G 0 phase was about 80 -85% of the total population.

The Putative ori-␤ DMI Was Not Detected by a Methylationsensitive Restriction Endonuclease, Southern Blot Detection-
Complete digestion of genomic DNA with AluI was monitored by the concomitant cleavage of a short, radioactively labeled fragment containing a unique AluI site, which was included in the cleavage reaction. Genomic DNA on the membrane was hybridized with two radioactive primers located at adjacent sites at the center of the DMI. As shown in Fig. 2B, lanes 4 and 7, cleavage of genomic DNA of both logarithmically growing CHO K1 and CHO C400 cells yields a fragment of 221 bp (arrows), which is expected when AluI is not inhibited by methylation (compare Fig. 2C). The same fragment with the same intensity was visible when DNA from stationary cells was analyzed (Fig. 2B, compare lanes 3 and 4 and lanes 7 and 8). All four possible AluI fragments (Fig. 2C) appeared when the membrane was hybridized with a synthetic PCR fragment spanning the entire DMI region (data not shown). In no case did we obtain a fragment of 740 bp, which should appear if all three AluI sites within the DMI were resistant to cleavage due to methylation (see Figs. 2B and 3C, broken lines). Cleavage of all DNAs with MboII yielded the expected fragment of 411 bp, indicating unimpaired accessibility of the DNA at this region (Fig. 2B, lanes 1, 2, 9 and 10, arrows; Fig. 2C). These results reveal that at least some non-CpG sites of the proposed DMI region of most proliferating or stationary CHO cells must be unmethylated.
The Putative ori-␤ DMI Was Not Detected by Four Methylation-sensitive Restriction Endonucleases, PCR Detection-The fraction of genomic DNA that resists cleavage to AluI or other methylation-sensitive restriction endonucleases can be quantitated by PCR using flanking primers, since only those molecules that resist cleavage allow synthesis of a full-length PCR product. Using this assay Tasheva and Roufa (12) concluded   FIG. 1. Location of the proposed DMI with respect to the dhfr ori-␤. A portion of the dihydrofolate reductase (dhfr) locus in CHO cells is shown. The origin of bidirectional replication (OBR) was mapped to the indicated position by several methods (17). The boundaries of the DMI slightly differ from those given by Tasheva and Roufa (12) since the ori-␤ sequence has recently been corrected (Genbank accession number X94372).

FIG. 2. Southern blot analysis of proliferating and stationary CHO cells.
A, FACS analysis of proliferating (log) and stationary (stat) CHO K1 and CHO C400 cells. Abscissa, fluorescence of propidium iodide-stained nuclei in arbitrary units. Ordinate, count events. B, autoradiographs of genomic DNA from proliferating and stationary CHO K1 and CHO C400 cells, cleaved with AluI or MboII, and hybridized to two genomic primers (see text). Lanes 5 and 6 contain 5Ј-endlabeled DNA length markers (X174, cleaved with HaeIII). Arrows indicate hybridizing AluI and MboII genomic fragments. The expected position of a 740-bp AluI genomic fragment is indicated by broken lines. C, map of the putative DMI region (gray box) indicating relevant AluI and MboII cleavage sites and fragment lengths, expected by hybridization of the two end-labeled genomic primers 6C and 6E (arrows within the DMI). that ϳ86% of proliferating CHO cells contained a DMI at ori-␤ and ori-RPS14 (12). To determine whether or not a DMI could be detected at ori-␤ using the same strategy employed by Tasheva and Roufa (12), the methylation-sensitive restriction endonucleases (41) AluI (AGCT), AccI (GTMKAC), Alw26I (GGATC), or MslI (CAYN 4 RTG) were incubated individually with DNA from proliferating CHO cells isolated in our laboratory or with DNA samples provided by Tasheva and Roufa. All four of these enzymes cut within the putative DMI at ori-␤, and AluI and MslI cut within the putative DMI at ori-RPS14, another locus shown to contain an OBR (43). DNA products were amplified using PCR under conditions designed to give a linear response between the amount of template included in the reaction and the amount of product obtained (data not shown). Following digestion with these enzymes, no amplification product was detected at ori-␤ (Fig. 3, lanes 1-4). On the other hand, when genomic DNA was digested with XbaI, a restriction endonuclease that cuts outside the putative DMI, the putative DMI region was amplified (Fig. 3, lane 5). Equivalent results were obtained with genomic DNA from quiescent CHO cells and from CHO cells synchronized in their early, mid, or late G 1 phases (data not shown). Control digestions using a PCR DNA fragment that had been generated in the presence of 5-methyl-dCTP demonstrated that the enzymes used would not cut methylated DNA even when an excess of incubation time and enzyme is used (see "Materials and Methods"; data not shown). To provide a quantitative assay for the detection of the highest portion of the cells possessing a DMI, the PCR conditions were adjusted to detect residual uncleaved molecules. It was estimated that a DMI, as defined by Tasheva and Roufa (12), cannot exist in more than 2% of the cells. Taken together, these results confirmed the absence of the putative DMI at ori-␤. Similar results were obtained at ori-RPS14.
Hydrazine and Permanganate Are Complementary Probes for Methylcytosine-The methylation state of the ori-␤ DMI was further investigated at nucleotide resolution using two chemical probes, hydrazine and permanganate, that discriminate between cytosine and methylcytosine. Hydrazine reacts only with cytosine (32). To ensure specificity of this method, a 5Јend-labeled DNA fragment (unrelated to ori-␤) was methylated at all CpG dinucleotides using bacterial SssI methylase. Complete methylation was confirmed by digestion of an aliquot of the DNA product with MspI and HpaII, two isoschizomers that have opposite sensitivity to DNA methylation (41). Both enzymes cleaved the unmethylated DNA at the unique CCGG site 3Ј to C 17 (Fig. 4, lanes 3 and 4). 2 As expected, methylation at C 18 did not impair cleavage by MspI (Fig. 4, lane 5) but did inhibit cleavage by HpaII (Fig. 4, lane 6).
Both the unmethylated and CpG-methylated fragments were subjected to hydrazine modification in the presence or absence of 1.5 M NaCl with subsequent cleavage by piperidine. All methylated cytosines could be unequivocally identified by their absence in the sequence ladders for cytosines (Fig. 4, compare  lanes 7 and 8) or cytosines ϩ thymines (Fig. 4, compare lanes 9  and 10). In contrast, unmethylated cytosines (non-CpG dinucleotides) were present. Close inspection of methylated versus unmethylated DNA bands revealed that migration of methylated DNA (Fig. 4, lanes 8 and 9 and also lanes 11 and 2 One must take into account that restriction enzyme cleavage (Fig.  3A, lanes 3-5) of a 5Ј-labeled fragment 3Ј to nucleobase B n yields a band that migrates ϳ1.5 bands slower than the corresponding band B n Ϫ 1 of a Maxam and Gilbert reference sequence (e.g. Fig. 3A, lane 7); this last lacks the 3Ј-terminal nucleoside, which accounts for the one-band difference and carries a 3Ј-terminal phosphate, which effects an additional faster migration of 0.5 band-to-band distance. It is not known why HpaII also cleaved the unmethylated DNA 3Ј to T 16 (Fig. 3, lane 4). 13-17 that are discussed later) was slightly more retarded relative to unmethylated DNA (Fig. 4, lanes 7 and 10 and also  lanes 12 and 15-18, see below). This most probably resulted from the increased mass due to the methyl groups.
Permanganate modifies thymines and methylcytosines in single-stranded DNA but not cytosines (33). Modified residues can be cleaved by treatment with piperidine. The specificity of this method was also confirmed using the same DNA fragments as described above for hydrazine. To specifically enhance the methylcytosine reaction, several parameters were optimized, the most important of which was reducing the pH from 4.8 to 4.1. Under these conditions, all methylcytosines were cleaved to the same extent as thymines (Fig. 4, compare lanes 11 or 13  with lane 12). In addition, a single guanine at position G 28 was also cleaved. This was not surprising, because some depurina-tion (creating piperidine-sensitive cleavage sites) was expected due to the slightly acidic pH (33). Most importantly, all cytosines remained unreactive. This method, therefore, complements the hydrazine method and has the same degree of specificity, albeit in a reciprocal manner.
It was also necessary to determine the sensitivity limit of chemical probing of the DNA, i.e. what percentage of methylated cytosines at a particular site can be still detected in a mixed, inhomogeneous DNA molecule population also containing unmethylated cytosines at the same site. This was achieved by mixing increasing amounts of unmethylated to decreasing amounts of methylated fragment, both of which had been probed with hydrazine, keeping total DNA amount constant (Fig. 4, lanes 14 -18). Analysis of these mixtures revealed that at least 25% of methylcytosine is required to give a slight visible decrease in band intensity compared with a homogeneous population with only cytosine at that site (Fig. 4, lane 17  versus 18). In these analyses, greater retardation of T and C bands of methyl-DNA than those of methyl-free DNA in the same lane for the reason explained above is clearly evident.
Hydrazine and Permanganate Detected a Methylcytosine at ori-␤, but Did Not Detect a Densely Methylated Island-Before analyzing genomic DNA using the hydrazine and permanganate methods, we determined whether or not a DMI could be recognized by these methods when the piperidine cleavage products were detected by LM-PCR. For this purpose, PCR fragments encompassing the proposed DMI were synthesized either in the presence of dCTP or 5 m dCTP triphosphate (see "Materials and Methods"). These fragments were treated with hydrazine or potassium permanganate as described above and cleaved with piperidine, and cleavage products were visualized by LM-PCR. The results (Fig. 5A) demonstrated that the hydrazine/LM-PCR methodology was capable of detecting the putative DMI at ori-␤. In the sample that contained exclusively m C, all bands representing methylcytosine were absent (lane 4, arrows), whereas in the sample that contained exclusively C, all cytosines were present (lanes 2 or 3). Analogous results were obtained after treatment of the fragments with potassium permanganate (data not shown).
To detect the putative DMI at ori-␤ in genomic DNA with nucleotide resolution, DNA was extracted from proliferating and stationary CHO K1 cells, cleaved with restriction enzymes to reduce its viscosity, and then treated with either hydrazine or permanganate as described above. Two examples of these analyses are provided. One region should contain the border of the DMI at position 2806 ( Fig. 5B; Fig. 1). The other region should contain an internal portion of the DMI (Fig. 5C). The products of a hydrazine reaction designed to display cytosines only were obtained using an unmethylated PCR fragment as a substrate (Fig. 5B, lane 1). The products from a hydrazine reaction designed to display both cytosines and thymines using DNA from stationary and proliferating cells are shown in Fig.  5B, lanes 2 and 3, respectively. Reduced cytosine reactivity compared with the reactivity of nearby thymine residues indicates that a cytosine is methylated. For example, the intensity of the cytosine band at position 2849 was much less than that of a nearby thymine (position 2852) but was equivalent to the background in these lanes (Fig. 5B, lanes 2 and 3). This indicates that this cytosine is methylated in most, and probably all, genomic DNA molecules. This cytosine is a constituent of a CpG site and served as an internal control for the specificity and sensitivity of the hydrazine reaction. In contrast, all other cytosines of the displayed genomic region (Fig. 5B, lanes 2 and  3) were as reactive toward hydrazine as nearby thymines in the same lanes. Therefore, given the sensitivity of this method (Fig.  4), at least 75% of the genomic DNA must be unmethylated.
These conclusions were confirmed using the permanganate method on single-stranded DNA of the same genomic region. The cytosine at position 2849 was methylated, because it reacted with permanganate in DNA from stationary or proliferating cells (Fig. 5B, lanes 5 and 6). No other cytosine residue gave a signal above background in this analysis (Fig. 5B, compare lanes 5 and 6 with lane 4), suggesting that no other cytosine in this region was methylated. Since all thymines in this region reacted with permanganate, the non-reactivity of cytosines did not result from the presence of double-stranded DNA. Beside thymines, guanines also gave slightly weaker cleavage signals. These signals may have resulted from limited permanganate oxidation or from acidic depurination of genomic DNA (33). This was more apparent here than in pilot experiments (Fig. 4), probably due to enhanced DNA fragmentation caused by the initial heating at 95°C in LM-PCR.
Analysis of an internal DMI region that was expected to be fully methylated revealed that all cytosines in at least 75% of the DNA from stationary or proliferating CHO K1 cells were unmethylated (Fig. 5C). No cytosine residue was resistant to modification by hydrazine (Fig. 5C, compare lanes 2 and 3 with  lane 1), and all cytosine residues were insensitive to permanganate (Fig. 5C, compare lanes 5 and 6 with lane 4). A total of about 90% of the entire putative DMI region and both DMI borders were analyzed in proliferating CHO K1 cells, and parts of the DMI were analyzed in stationary CHO K1 and CHO C400 cells (Fig. 5 and data not shown). Taken together, the data obtained from hydrazine and permanganate methods for m C detection revealed only one m C residue that is a constituent of the single CpG within the putative ori-␤ DMI. Given the sensitivity of these methods (Fig. 4), the remaining cytosines on either strand of the DMI region were unmethylated in at least 75% of the DNA from either proliferating or stationary cells.
To determine whether the DMI was present in only a small portion of the DNA from an asynchronous proliferating population of cells, DNA was isolated from CHO K1 cells in different phases of their cell division cycle. Proliferating CHO K1 cells were fractionated by centrifugal elutriation and analyzed by FACS (Fig. 6) ϩ M phases (fraction 6) was 2.5-4-fold. Analysis of the DNA from these fractions by either the hydrazine or permanganate methods failed to reveal the presence of a DMI (data not shown). Since the detection limit for m C was 25% (Fig. 3), and assuming that the DMI would be present in only one of these phases, at least 90 -94% of the DNA molecules in the total population lacked a DMI at ori-␤. DISCUSSION The report of an unusual form of DNA methylation closely associated with origins of bidirectional replication exclusively in the chromosomes of actively proliferating mammalian cells constituted an important discovery of what appeared to be a major component of mammalian replication origins (11,12). The experiments described here had two objectives. The first was to confirm the existence of a DMI at the well characterized dhfr ori-␤ locus in proliferating CHO cells, and the second was to determine whether or not this DMI appeared at a specific time during the cell division cycle.
In the original report describing densely methylated islands (DMI), two methods were used to demonstrate the presence of a DMI at the dhfr ori-␤ locus in 86% of proliferating CHO cells (12), 1) conversion of cytosines to uracils by bisulfite treatment of DNA, and 2) sensitivity of genomic DNA to cleavage by AluI.
In an effort to confirm this report, three independent, stringently controlled techniques were applied to investigate the ori-␤ DMI, 1) sensitivity of genomic DNA to cleavage by different methylation-sensitive restriction enzymes, 2) sensitivity of cytosines to cleavage by hydrazine/piperidine treatment, and 3) sensitivity of methylcytosines to cleavage by permanganate/ piperidine treatment. Control experiments demonstrated that the methods for detecting m C in genomic DNA at the nucleotide level would have detected a DMI if present in as little as 25% of the total cell population. In addition to comparing proliferating and nonproliferating CHO and CHO C400 cells, CHO cells were also fractionated by centrifugal elutriation to enrich for populations of G 1 , S, and G 2 ϩ M phase cells. Analysis of these individual cell populations meant that a DMI would have been detected in as few as 6 -10% of the cells examined. The PCR-based detection of DNA molecules resistant to cleavage by methylation-sensitive restriction endonucleases would have detected a DMI if present in as little as 2% of the cells.
Both the hydrazine and permanganate methods identified the single, stable m CpG within the putative ori-␤ DMI. This m CpG was present in Ͼ90% of the proliferating CHO K1 (two copies of this locus) and CHO C400 (ϳ1000 copies of this locus) cells, and the same m CpG also was present in stationary cells. Therefore, these methods had no difficulty in detecting methylated cytosines in genomic DNA. However, none of the three methods used detected a DMI. How might the discrepancy between these results and those reported previously (12) be explained?
The bisulfite method is subject to an artifact that could easily account for the appearance of DMIs. Bisulfite converts cytosines to uracils only when the DNA is single stranded. If the DNA sample is not completely denatured, sequences that failed to react with bisulfite will contain stretches of unreacted cytosines that will falsely indicate cytosine methylation (29,30, see also Ref. 31). If PCR primers are used to amplify the unreacted regions in the belief that they contain methylated cytosines, then even a small contamination of undenatured DNA may yield a PCR product that contains only unconverted cytosines. One way to distinguish between cytosines that did not react with bisulfite because they were methylated from cytosines that did not react because they were present in double-stranded DNA is to apply independent methods for detection of methylated cytosines that are not subject to the same artifact.
One such method used by Tasheva and Roufa (11,12) to determine the fraction of genomic DNA that contained a DMI was to measure the fraction of DNA that could be cleaved by digestion with AluI, a restriction endonuclease that cannot cut when its recognition site is methylated. However, the inability of AluI to cut genomic DNA can result either from methylation of its DNA recognition site or simply from incomplete digestion. In fact, while plasmid DNA or PCR-generated DNA was easily cut by AluI in our hands, we initially encountered considerable difficulty cutting genomic DNA to completion. Incompletely digested genomic DNA yielded a full-length product after PCR. However, if this same genomic DNA was re-digested with AluI, no PCR product was obtained (data not shown). Thus, any method using AluI, or any other methylation-sensitive enzyme as a tool for methylation analysis, would be subject to possible incomplete reactivity. Therefore, a stringent control for the enzyme activity should be included. Such a control was included in our experiments by monitoring the degree of cleavage of an unrelated fragment in the same incubation reaction where cleavage of genomic DNA took place. This control was not provided in previous studies of DMIs (12) and therefore leaves open the possibility that AluI simply failed to cut genomic DNA efficiently in these experiments. Additionally, DNA from stationary cells was not investigated to show that in the absence of the putative DMI, AluI would cut DNA to completion.
To determine the methylation status at every cytosine within the putative DMI, we used hydrazine to modify cytosines, which allows cleavage at these positions by piperidine. Hydrazine is the classical method for sequencing DNA and for identifying methylated cytosines ( m C) in genomic DNA (32,35). Since hydrazine does not modify m C, piperidine will not cleave at these positions, and the corresponding cytosine band will be absent in a sequencing gel. The advantage of this method over bisulfite is that its reactivity and specificity does not depend on the strandedness of the DNA. The hydrazine method detected only one m C within the putative ori-␤ DMI, and this was at a CpG dinucleotide, the classical site for m C. Moreover, it was present in stationary cells as well as proliferating cells.
To confirm this conclusion, we developed the use of permanganate to complement the hydrazine method. At low pH, permanganate modifies m C as well as T and allows cleavage at these positions by piperidine. Since permanganate at low pH does not modify C, the presence of a band at a cytosine position in the sequencing gel indicates that the genomic DNA contains m C instead of C at this position. Permanganate is the only reagent for detection of methylation on the nucleotide level that shows positive reactivity with methylated cytosines. This method, therefore, complements the hydrazine method. Like bisulfite, permanganate reacts only with m C in single-stranded DNA. However, since permanganate also reacts with thymines under these conditions, again only in single-stranded DNA, thymines provide an internal standard to ensure that the sequences under evaluation have reacted with permanganate. The results with the permanganate method were in full accord with the results we obtained using methylation-sensitive restriction endonucleases and the hydrazine method.
Stringent application of the bisulfite method that was employed by Tasheva and Roufa (11,12) confirms the absence of a DMI at ori-␤ and also at the RPS14 origin (31). Thus, replication origins are not associated with the unusual methylation pattern known as a DMI. Unusual, dense non-CpG DNA methylation is most likely an artifact of the bisulfite method (31). Since other important findings of unusual DNA methylation were based on this method (e.g. Ref. 44), we suggest that any such findings be reevaluated either by independent methods as presented here or by a modified bisulfite procedure (31).