Peptide Mapping of the Murine DNA Methyltransferase Reveals a Major Phosphorylation Site and the Start of Translation*

The murine DNA methyltransferase catalyzes the transfer of methyl groups from S-adenosylmethionine to cytosines within d(CpG) dinucleotides. The enzyme is necessary for normal embryonic development and is implicated in a number of important processes, including the control of gene expression and cancer. Metabolic labeling and high pressure liquid chromatography-electrospray ionization-mass spectrometry (HPLC-ESI-MS) were performed on DNA methyltransferase purified from murine erythroleukemia cells. Serine 514 was identified as a major phosphorylation site that lies in a domain required for targeting of the enzyme to the replication foci. These results present a potential mechanism for the regulation of DNA methylation. HPLC-ESI-MS peptide mapping data demonstrated that the purified murine DNA methyltransferase protein contains the N-terminal regions predicted by the recently revised 5′ gene sequences (Yoder, J. A., Yen, R.-W. C., Vertino, P. M., Bestor, T. H., and Baylin, S. B. (1996) J. Biol. Chem. 271, 31092–31097). The evidence suggests a start of translation at the first predicted methionine, with no alternate translational start sites. Our peptide mapping results provide a more detailed structural characterization of the DNA methyltransferase that will facilitate future structure/function studies.

The DNA methyltransferase enzyme interacts with p23, a known component of the progesterone receptor complex (10), and it is likely that control of genomic DNA methylation is coupled to receptor-mediated signaling systems in some, as yet unknown, way. A detailed knowledge of the factors involved in the regulation of the methyltransferase enzyme will be necessary to know how genomic methylation patterns are established.
The murine DNA methyltransferase is an approximately 180 -190-kDa enzyme that transfers a methyl group from Sadenosylmethionine (AdoMet) onto the C-5 position of cytosine within the d(CpG) dinucleotides of double-stranded DNA. Targeted disruption of the gene for this enzyme is lethal to embryos at the middle stages of gestation (1). A number of functional domains of the enzyme have been identified and are presented in Fig. 1. The catalytic domain lies in the C-terminal third of the protein and contains regions of homology with the prokaryotic DNA methyltransferases, including a conserved AdoMet binding site and catalytic center. The N-terminal twothirds of the enzyme is separated from the catalytic domain by a flexible hinge region consisting of glycine-lysine repeats (11) and contains sequences that bind zinc and DNA independently (11,12). A minor groove DNA-binding motif (SPKK) shown to undergo phosphorylation-dependent attenuation in other proteins (13,14) is located in the non-catalytic region of the protein. The DNA methyltransferase localizes to the replication foci during S phase of the cell cycle (15), and a 200-amino acid segment of the protein is necessary for this cell cycledependent localization.
Various sizes of the murine DNA methyltransferase enzyme (20,21,35), it's cDNA (16,17), and it's mRNA (18,40) have been reported. Based on SDS-PAGE analysis the protein is estimated to be from 170 to 190 kDa. Post-translational processing (20,21) as well as alternate transcriptional start sites have been suggested to account for this range in size (18,40). However, when purified in the presence of protease inhibitors, the DNA methyltransferase from murine erythroleukemia (MEL) cells is detected as a single protein band on SDS-PAGE (22), casting doubt upon the presence of multiple enzyme forms. The original cDNA predicted a protein of 1573 amino acids, corresponding to a mass of 175 kDa (16), but was later revised to 1502 amino acids (169 kDa) (Genbank accession X14805). Further revisions to the 5Ј regions of the human and murine DNA methyltransferase genes now predict a protein of 182 kDa (17). These data implicate a new translational start site resulting in an enzyme 118 amino acids longer than previously determined. Furthermore, these data suggest that a previously identified DNA methyltransferase promoter (19) would lie in an intron (17), upstream of the originally predicted start of translation in MEL cells.
We have used sensitive mass spectroscopic techniques to map the peptides in the purified MEL DNA methyltransferase, studied post-translational modifications, and examined the Nterminal regions of the protein. Our data identify a major phosphorylation site on the DNA methyltransferase from murine erythroleukemia cells. The site lies in a region of the protein that is required for targeting of the methyltransferase to the replication foci during S phase of the cell cycle (15). Our results suggest that the murine DNA methyltransferase contains the newly reported N terminus, supporting a single start of translation at this new site.

MATERIALS AND METHODS
Electrophoresis, Blotting, and DNA Methyltransferase Preparation-SDS-polyacrylamide gel electrophoresis (SDS-PAGE) was performed using 7.5% polyacrylamide, using broad range molecular weight protein standards (New England Biolabs). Transfer of proteins from polyacrylamide gels to nitrocellulose (MSI Inc.) or PVDF membranes (Millipore, Inc.) was performed using an Integrated Separation Systems semi-dry electroblotting apparatus set at 200 mA for 50 min.
Purification of DNA methyltransferase from MEL cells was performed as described previously (22). Protein concentrations were determined with a Bio-Rad protein assay kit using bovine serum albumin and myosin as standards and validated by comparison with these standards on Coomassie-stained SDS-PAGE gels.
Glycosylation Analysis-Purified homogeneous MEL DNA methyltransferase was subjected to SDS-PAGE and transferred to nitrocellulose membranes. N-Glycosylated and O-glycosylated polypeptides were detected by using a Glycotrack (Oxford Glycosystems) carbohydrate detection kit. Briefly, the putative glycoprotein immobilized to the nitrocellulose membranes was oxidized with a periodate solution and subsequently biotinylated with biotin hydrazide. The biotinylated glycopolypeptides were detected with a streptavidin-alkaline phosphatase conjugate. Ovalbumin (about 5% carbohydrate content) and biotinylated size markers were used as positive controls, and non-biotinylated size markers were used as negative controls.
Preparation of MEL Nuclear Extracts-Cell pellets were homogenized in 0.2-2.0 ml of nuclear extraction buffer (10% sucrose, 0.3% Triton X-100, 20 mM Tris, pH 7.4, 3 mM MgCl 2 , supplemented with 1.0 mM of the phosphatase inhibitors, sodium molybdate, sodium fluoride, sodium vanadate, sodium orthophosphate, and 10 M of the protease inhibitors, E-64, phenylmethylsulfonyl fluoride, leupeptin, L-1-tosylamido-2-phenylethyl chloromethyl ketone, N ␣ -p-tosyl-L-lysine chloromethyl ketone) by five passages through a 20-gauge needle. The extract was centrifuged at 4000 ϫ g for 15 min. The pellet was suspended in nuclear extraction buffer containing 250 mM NaCl and homogenized by 12 presses in a small Dounce homogenizer. The suspension was centrifuged at 50,000 ϫ g for 60 min and the supernatant processed as described below.
Immunoprecipitation-Cell-free extracts were added to 0.8 -4.0 ml of immunoprecipitation buffer (50 mM Tris, pH 8.0, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS with protease inhibitors, and phosphatase inhibitors as described in the previous section), and the final NaCl concentration was brought to 150 mM.
Immunoprecipitations were carried out by the addition of 5-30 l of pATH 52 antiserum directed against a Trp E-methyltransferase Nterminal fusion encoding amino acids 256 -754 (11) or non-immune rabbit antiserum to the cell-free extracts, incubated for 1 h on ice, then added to a 50-l pellet volume of protein A-Sepharose CL 4B beads (Pharmacia Biotech Inc.), and incubated with frequent agitation for 1 h on ice. The suspension was pelleted by centrifugation at 2000 ϫ g for 5 min and washed three times at room temperature in 150 mM NaCl, 20 mM Tris, pH 8.0, 0.1% Tween 20. The final pellet was dissolved in 20 l of SDS-PAGE sample buffer supplemented with 2 mM dithiothreitol, heated at 70°C for 10 min, and subjected to SDS-PAGE for analysis.
Phosphoamino Acid Analysis-SDS-PAGE gels of the immunoprecipitates were transferred onto Immobilon PVDF membranes as described previously (23). The membrane was air-dried and exposed to Fuji RX film at Ϫ70°C for 24 h. The [ 32 P]orthophosphate-labeled DNA methyltransferase was excised and extracted from the filter, acid-hydrolyzed as described previously (24), and then subjected to thin layer cellulose chromatography as described (25) using 0.5 M NH 4 OH/isobutyric acid (5:3, v/v) as a solvent. Negative controls were performed by the excision of nonspecifically labeled bands from a lane on the PVDF membrane that was immunoprecipitated with non-immune rabbit serum. Phosphoserine, phosphothreonine, and phosphotyrosine standards (500 ng, Sigma) were detected by ninhydrin spray (Sigma), and the experimental lanes were air-dried and detected by autoradiographic exposure to Fuji RX film at Ϫ70°C for 36 h.
Proteolytic Digestion of 32 P-Labeled Murine DNA Methyltransferase-In situ digestion and extraction of the DNA methyltransferase in SDS-PAGE gels was performed according to the method of Williams (26). Briefly, DNA methyltransferase-containing bands (5-10 g) were excised from Coomassie-stained SDS-PAGE gels of immunoprecipitates from 32 P-labeled MEL cells. The bands were washed by shaking for 30 min in ice-cold acetone, dried in a speed-vac, and then re-hydrated in 100 l of 100 mM NH 4 HCO 3 , pH 7.8, for 4 h with gentle agitation. The buffer was changed once, and the bands were cut into three equal pieces followed by the addition of either trypsin (Sigma), L-1-tosylamido-2phenylethyl chloromethyl ketone-treated chymotrypsin (Boehringer Mannheim), or V8 protease (Boehringer Mannheim) (1 g) and incubated for 16 h at 37°C. The supernatant was collected, and the remaining gel slices were extracted by shaking with an additional 100 l of 100 mM NH 4 HCO 3 for 4 h. The supernatant was saved, and a further extraction was performed with 2 M urea for 12 h. The collected supernatants were pooled and dried on a speed-vac. Samples were dissolved in 50 l of HPLC grade water for subsequent chromatography. The recovery efficiency of the proteolytic fragment peptides from the gel was monitored by Cerenkov counting and ranged between 10% for V8 digests and 85% for tryptic digests.
Proteolytic Digestion of Murine DNA Methyltransferase Purified from MEL Cells-Murine DNA methyltransferase was purified to homogeneity at a concentration of 1.7 M (300 g/ml) from MEL cells as described previously (21). This preparation (500 l) was precipitated with 2 volumes of ice-cold acetone and centrifuged for 30 min at 20,000 ϫ g, and the pellet was washed with 70% ice-cold acetone. The pellet was resuspended in 10 l of 8 M urea with repeated (three times) mixing and heating at 37°C. The solution was brought to a final concentration of 17 M protein, 2 M urea, and 100 mM NH 4 HCO 3 in a 50-l volume, pH 7.8, by the addition of HPLC grade water and a stock solution of 1 M NH 4 HCO 3 . Calcium chloride was added to a concentration of 5 mM in the tryptic digests.
L-1-Tosylamido-2-phenylethyl chloromethyl ketone-treated chymotrypsin (Boehringer Mannheim) (7 g), trypsin (Sigma) (5 g), or V8 protease (Boehringer Mannheim) were added from 10 ϫ stock solutions, and the digest was incubated at 37°C for 12 h, sonicated for 5 min, and then supplemented with an additional 5 g of protease. The digestion was allowed to continue for an additional 12 h prior to HPLC/ESI-MS analysis.
C-18 Reverse Phase Chromatography/Electrospray Mass Spectrometry-A typical sample contained 20 pmol of 32 P-labeled DNA methyltransferase digested with protease in the gel, containing 300 -1700 dpm, mixed with 200 pmol of purified DNA methyltransferase digested in the solution as described in the previous section. All chromatography was performed using a Michrom Ultrafast HPLC system (Michrom Bioresources, Sunnyvale, CA), equipped with a micro-flow cell UV absorbance detector set at 215-nm wavelength. A 1.0-mm inner diameter ϫ 15.0 cm long C 18 , 300-m reverse phase HPLC column (Michrom Bioresources, Sunnyvale, CA) was used to separate peptides derived from proteolytic digests of the DNA methyltransferase. Chromatography solvents were 0.1% trifluoroacetic acid, 1% acetonitrile (solvent A)  (11) with homology to prokaryotic DNA methyltransferases. The region starting at the N terminus and going to the solid box is a putative regulatory domain (11). The hatched box is the replication foci targeting domain (15). The dotted box is predicted by the newly identified 5Ј sequences of the gene (17). Binding sites for zinc (Zn) (12,13), and DNA (12) are indicated. The position of serine 514, identified as a major phosphorylation site, is indicated. and 0.1% trifluoroacetic acid, 95% acetonitrile (solvent B). After loading the column with the proteolyzed DNA methyltransferase, a 5-10-min isocratic wash of 0% B was followed by a gradient of 0 to 95% solvent B over 30 min to 100 min at a flow rate of 50 l/min.
For analyses that required Cerenkov counting, the outlet of the UV absorbance detector flow cell was connected to a Pharmacia fraction collector with 30 cm of PEEK tubing (0.005-inch diameter, Upchurch Scientific Inc., Oak Harbor, WA). Fractions were collected at 1-min intervals and were counted for 2 min in a Beckman scintillation counter.
For analyses that required electrospray mass spectrometry, the outlet of the UV absorbance detector flow cell was connected to the electrospray probe with 20 cm of PEEK tubing (0.005-inch diameter). Mass spectrometry was performed using a Fisons VG Platform II quadrupole mass spectrometer, equipped with a pneumatically assisted ESI source. The mass spectrometer was scanned repetitively over a mass to charge ratio (m/z) range of either 300 -1500 or 400 -1500, at a scan time of 2 s/scan, a 25-35 V orifice potential, and a cone temperature of 70°C. Data were collected in centroid mode.
MassLynx (Micromass Inc.) software provided with the mass spectrometer permitted mass spectra to be displayed for any observed peak of ion detection events, allowed background subtraction, as well as extraction of a defined input m/z from the data set. A Biolynx peptide analysis algorithm (Micromass Inc.) was used to predict the peptide products of proteolytic digests of the murine DNA methyltransferase (Genbank accession number X14805) and to search ESI-MS data for these products. Manual searching of data sets was also used.

Post-translational Modification: Serine Phosphorylation and
Absence of Glycosylation-The glycosylation state of the purified MEL DNA methyltransferase was analyzed using a Glycotrack carbohydrate detection kit. Our results showed that glycosylation of the homogeneous enzyme was not detectable. We were able to detect glycosyl moieties in 35 ng of ovalbumin which is 5% glycosylated by weight, whereas glycosylation was not detected in 4 g of either MEL-DNA methyltransferase or recombinant DNA methyltransferase (data not shown).
Post-translational phosphorylation was detected by 32 P labeling of the DNA methyltransferase in MEL cells. Immunoprecipitations with anti-methyltransferase antibodies from [ 32 P]orthophosphate-treated MEL cells resulted in the recovery of 32 P-labeled DNA methyltransferase as detected by x-ray film exposed to SDS-PAGE gels of the immunoprecipitates ( Fig. 2A).
The DNA methyltransferase band was excised from the gel along with the nonspecific gel slice from a non-immune rabbit antiserum. In 10-mCi labeling experiments using 2 ϫ 10 8 to 5 ϫ 10 8 MEL cells, the DNA methyltransferase band repeatedly contained 9000 -11,400 dpm by Cerenkov counting, whereas negative control gel slices, produced by a non-immune antiserum, resulted in 680 dpm, indicating that the 32 P label was specifically associated with the Coomassie-detectable DNA methyltransferase and not with any other portion of the gel. The phosphorylation state of the DNA methyltransferase was not affected by serum starvation for 30 h prior to labeling, serum starvation, and replenishment or by treatment of MEL cells with the phosphatase inhibitor, okadaic acid (0.1 M), for 24 h prior to labeling.
The major sites of protein phosphorylation in vertebrate cells are on serine, threonine, and tyrosine residues (27). To determine which type of amino acid was phosphorylated, the 32 Plabeled DNA methyltransferase was acid-hydrolyzed into individual amino acids, and the amino acids were separated by thin layer chromatography. The presence of [ 32 P]phosphate was detected by x-ray film. Fig. 3 shows that the primary sites of phosphorylation reside on serine or threonine, because the majority of label co-migrated with the phosphoserine and phosphothreonine standards.
Peptide Mapping and Identification of Serine 514 as a Major Phosphorylation Site-HPLC was used to separate peptides derived from proteolytic digests of the DNA methyltransferase. Cerenkov counting of the HPLC column fractions was used to detect 32 P-labeled peptides, and electrospray ionization mass spectrometry was used to measure the masses of these peptides. These studies required the use of highly pure DNA methyltransferase (Fig. 2C), since other proteins in the preparation might result in the detection of spurious masses. 32 P-Labeled DNA methyltransferase was obtained from immunoprecipitated nuclear extracts of metabolically labeled MEL cells, which was excised from SDS-PAGE gels (Fig. 2B). The 32 Plabeled DNA methyltransferase was separated from the IgG subunits of the antibody by excision of the methyltransferase from the SDS-PAGE gel (see "Materials and Methods"). Because only a small amount (3-10 pmol) of the DNA methyltransferase-derived peptides were extracted for any single digest, the samples were supplemented with digested, nonlabeled (Fig. 2C) DNA methyltransferase purified from MEL cells (50 -200 pmol). In these experiments, 60 -95% of the total 32 P label (200 -1200 dpm) eluted in a small retention window (3 min), and the remaining 5-40% of loaded counts were distributed throughout the HPLC gradient. Over 60% of the predicted masses were detected in each digest, resulting in the assignment of over 80% of the DNA methyltransferase sequence, spanning the entire protein. Less than 8% of the detected m/z's were not predicted by the cDNA.
Two criteria were used to identify phosphorylated peptides as follows: (i) the peak of [ 32 P] in the HPLC chromatogram, and (ii) an 80-atomic mass unit increase from the predicted mass of peptides generated by sequence-specific proteases. Our data identified serine 514 as a major phosphorylation site. Digests with trypsin, chymotrypsin, and Staphylococcus aureus V8 proteases demonstrated masses that differed by 80 atomic mass units from the predicted peptides containing serine 514. A summary of the mass spectrometry data showing the structure of the phosphorylated peptide, and its predicted cleavage sites by S. aureus V8, trypsin, and chymotrypsin is shown in Fig. 7. In tryptic and chymotryptic digests the non-phosphorylated peptide was detected at a retarded retention time consistent with its increased hydrophobicity.
The V8, chymotryptic, and tryptic phosphopeptides containing serine 514 were co-retained with the 32 P label on C 18 columns. Figs. 4 -6 present the data from peptide mapping experiments using C 18 chromatography to separate digests of the 32 P-labeled DNA methyltransferase. The tryptic digest (Fig. 4A) produced a peak of 32 P label in the pre-gradient volume, indicating that the peptide had very low affinity for the stationary phase. The phosphopeptide and peak of 32 P label eluted with many unassigned large molecule contaminants in the pre-gradient volume. This fraction was collected and purified on the same C 18 column coupled to the mass spectrometer. This additional step served to separate the phosphopeptide from these contaminants. Fig. 4 shows the co-retention of m/z 702 with the 32 P label, corresponding to the singly charged phosphopeptide, IYISpK (T84 ϩ P i ), containing serine 514.
Similar experiments with S. aureus V8 and chymotrypsin are presented in Figs. 5 and 6. In these experiments, 32 Plabeled digests were divided into two equal aliquots and separated by HPLC. The first aliquot was subjected to HPLC, and the fractions were collected and Cerenkov counted. The second aliquot was subjected to HPLC and mass spectrometry. In the  (1 min, 50 l) were collected, and Cerenkov counts were measured for each fraction (q, 100% ϭ 220 dpm). The y axes measure the relative intensity of the signal. The x axes measure time in minutes. chymotryptic digest (Fig. 5), an m/z of 1016 was co-retained with the 32 P label, corresponding to the phosphorylated chymotryptic peptide ISpKIVVEF, containing serine 514. In the V8 digest (Fig. 6), an m/z of 1272 was co-retained with the 32 P label, corresponding to the phosphorylated V8 peptide, KIYIS-pKIVVE, containing serine 514. Thus, the unambiguous assignment of serine 514 as a major phosphorylation site was possible because no isomassive peptides were predicted by the cDNA, masses of peptides containing serine 514 were detected in digests with three independent proteases, and these peptides were co-retained with the 32 P label on C 18 reverse phase HPLC.
The presence of minor peaks of 32 P label in the S. aureus V8 and tryptic digests (Figs. 5 and 6), as well as the detection of some threonine phosphorylation on thin layer chromatograms (Fig. 3), suggests that other sites of phosphorylation are present. Our mass spectrometry studies indicated that these peptides were probably below the limit of detection of the mass spectrometer (approximately 10 pmol for peptide standards) or were not suitably ionized, because no masses differing from the predicted peptides by 80 atomic mass units were detected at these retention times. In addition to the phosphopeptide containing serine 514 (Figs. 4 -7), 4 out of 300 detected masses were consistent with an 80-atomic mass unit shift in the mass of threonine-containing peptides from the predicted cDNA; however, these were not detected in more than one of the three independent protease digestions and did not co-migrate with the 32 P label. It is possible that these sites were (i) sulfated, which would also result in an 80-atomic mass unit signature, (ii) the products of nonspecific cleavage, (iii) non-DNA methyltransferase-derived contaminants, or (iv) phosphorylated sites that did not turnover during the 12-h cell labeling period.
The DNA methyltransferase contains a number of putative recognition sites for kinases implicated in the modulation of DNA binding proteins, which were detected in the non-phosphorylated form. An SPKK minor groove binding motif (14) resides in the N-terminal of the murine DNA methyltransferase and is proximal to a zinc-binding domain (11,12). A potential p34 cdc2 -cyclin B protein kinase site (SPPK) (27) is found in the middle of the non-catalytic domain. These peptides were detected at retention times where only minimal amount of 32 P label was measured, without any additional mass of 80, and were therefore unphosphorylated. Unlike the major phosphorylation site that we have identified, these sequences are not conserved between murine, chicken, human, sea urchin, and frog DNA methyltransferases (16,28,29,36,41).
Identification of the N-terminal Region of the DNA Methyltransferase-Reports on the size of the DNA methyltransferase have been variable with respect to the migration on SDS-PAGE gels, the length of its messenger RNA (18,40), and the 5Ј ends of the gene (17,19). Recently, a revision of the 5Ј regions of the methyltransferase gene was reported, suggesting new 5Ј exons and a new start of translation (17). The formerly predicted start of translation would lie 357 base pairs downstream and code for an internal methionine. Our HPLC-ESI-MS results are consistent with the start of translation residing at the newly identified start codon (Table I, Fig. 8). Peptide A1-2 includes the start methionine and peptide T19 spans the first internal methionine, suggesting that translation of the DNA methyltransferase begins prior to this previously suggested start site. The detection of peptides corresponding to the newly determined 5Ј region suggests that the DNA methyltransferase from MEL cells is not significantly proteolyzed.
Our HPLC-ESI-MS analysis was unable to assign any masses consistent with N-terminal acetylation, formylation, or pyroglutamation. Furthermore, we were unable to find any evidence for clipping of the N terminus. Peptide T2 is isomassive with an acetylated T1 peptide (Table I). Since the nonacetylated N-terminal peptide was detected in the V8 digests, we assigned the m/z 516 to T2.  (31,32), and alteration of subcellular localization (33). Since DNA methylation is an essential component of embryonic development and the mechanisms of regulation of DNA methylation in metazoans are poorly understood, a detailed characterization of the structure of the mammalian DNA methyltransferase is critical. Our   FIG. 6. HPLC-ESI-MS of 32 P-labeled S. aureus V8-digested peptides of the DNA methyltransferase. 104 pmol (4 pmol of 32 P-labeled (320 dpm) and 100 pmol of unlabeled) V8 digest of the murine DNA methyltransferase was loaded onto C-18 column connected to either the ESI-MS (A and B) or fraction collector (C). B, HPLC-ESI-MS, reconstructed ion chromatogram of m/z 1271.9 peptide corresponding to the phosphorylated KIYISp-KIVVE (A72 ϩ P i ). An isomassive peptide was assigned (A95-96). A, background subtracted mass spectrum of the chromatographic peak at 22 min showing a singly charged species corresponding to phosphorylated KIYISpKIVVE (A72 ϩ P i ). Other co-retained DNA methyltransferase-derived V8 peptides are noted (A120, A183, etc.). C, relative absorbance at 214 nm of HPLC eluant of the identical sample from B. Fractions (1 min, 50 l) were collected, and Cerenkov counts were measured for each fraction (q, 100% ϭ 180 dpm). The y axes are scaled according to relative intensity of the signal. The x axes measure time in minutes. data show that the murine DNA methyltransferase is posttranslationally modified. Although the DNA methyltransferase was not glycosylated, phosphorylation was detected ( Fig. 2A). HPLC-ESI-MS experiments did not detect any other significant modifications, despite the detection of 80% of the predicted unmodified peptides derived from proteolytic digests of the enzyme. In vivo labeling demonstrated that the 32 P label appeared predominantly on serine and threonine residues. Further characterization of the modification sites relied on a combination of 32 P label isolation and mass spectrometry analysis of HPLC-isolated peptides. Digestion of the DNA methyltransferase with three different proteases showed that the 32 P label was localized predominantly to a small number of fractions. Masses of 80 atomic mass units higher than expected were co-retained with the 32 P label. The combined results show a predominant phosphorylation of serine 514, with relatively small amounts of phosphorylation at other sites.
The peptide sequence surrounding serine 514 is conserved between human, murine, chicken, sea urchin, and frog DNA methyltransferases (16,28,29,36,41). The human and murine sequence contains IYISKIVVE 2 at this position (16,28), and both chicken and frog sequences contain IYMSKIVVE (36,41). The sea urchin (Paracentrotus lividus) sequence contains IYM-SKILIE at the same position in the protein sequence (29). Although the prediction of the kinase that phosphorylates the DNA methyltransferase is highly speculative, the IYISp-KIVVE peptide has some structural similarities to the phosphorylation site on the oncogene product of c-Myb, which is phosphorylated by casein kinase II on SIYSpSpDDDE. Phosphorylation of transcription factor c-myb at this site inhibits binding to DNA, and deletion of this site results in oncogenic activation (31). Similar casein kinase II sites are found in calmodulin (27) and the progesterone receptor (34), and the DNA methyltransferase interacts with a component of the progesterone receptor complex (10).
Post-translational phosphorylation is a plausible means for regulating some aspect of the DNA methyltransferase's function. Because the phosphorylation site lies in a domain required for protein targeting to the replication foci during S phase of the cell cycle, phosphorylation may affect the subcellular localization of the enzyme. The large N-terminal twothirds of the protein is unnecessary for catalysis and contains distinct DNA and zinc binding sites (Fig. 1) (11, 12). We recently showed that an allosteric site on the enzyme is involved in both substrate inhibition and the binding of inhibitory single-stranded DNA sequences 3 ; phosphorylation might serve to attenuate this allosteric effect. Preliminary studies of the enzyme treated with alkaline phosphatase, -phosphatase, and casein kinase resulted in no significant effect on the catalytic rate in steady-state assays using double-stranded poly(dI-dC) as a substrate.
Characterization of the Translational Start Site-Previous studies have suggested alternate transcriptional, translational, or proteolytic products of the MEL DNA methyltransferase (17)(18)(19)(20)(21)41), and functional roles for different sizes of the enzyme have been proposed (20,21). Reports on the size of the mammalian proteins on SDS-PAGE range from 120 to 190 kDa   (20,21,35). However, our studies have suggested that the DNA methyltransferase is very susceptible to proteolysis and can be purified as a single band on SDS-PAGE gels, migrating between the 158-and 212-kDa size markers (Fig. 2C) (21,37). The original report on the translational product of the MEL DNA methyltransferase cDNA predicting a 1573-amino acid protein with a mass of 175 kDa (16) was later revised to predict a 1502-amino acid protein with a mass of 169 kDa (Genbank accession X14805). Cloning of the human DNA methyltransferase (28) suggested a new open reading frame upstream of the originally predicted start of translation identified in the MEL DNA methyltransferase sequence. Reports of DNA methyltransferase cDNAs from chicken, frog, and sea urchin show considerable divergence with respect to the length of the sequence and the N-terminal regions (29,36,42). Furthermore, different size mRNAs for the DNA methyltransferase in mouse testis have been reported (18,41). A recent, detailed study (17) has characterized new 5Ј exons of the human and murine DNA methyltransferase genes and predicts a murine translational product that is 118 amino acids longer than previously suspected (1620 amino acids, predicted mass of 183 kDa). However, this would mean that a previously identified promoter (19) would now lie in an intron.
Our mass spectrometry data clearly identify the presence of the 118-amino acid N-terminal peptide. The assignment of masses from V8 and tryptic digests which include the N-terminal peptide (A1-2), a peptide spanning the first internal methionine (T19), which was the formerly proposed start of translation, and an array of peptides covering the N-terminal region, demonstrates the existence of the newly reported extra sequence in the MEL-derived DNA methyltransferase (Table I  and Fig. 8). No masses were detected for any N-terminally proteolyzed peptide, and our SDS-PAGE analysis suggests a single, non-proteolyzed species. No evidence for post-translational modification of the N terminus was detected.
Functional comparisons of the MEL-derived (which contains an intact N terminus) and an N-terminal truncated recombinant enzyme show that the two forms of enzyme have very similar steady-state kinetic parameters (37). Thus, our identification of the extended N-terminal region of the purified murine DNA methyltransferase protein allows us to conclude that this region has no effect on the k cat , K m DNA or K m AdoMet of the enzyme. However, differential processing of the DNA methyltransferase in different tissues or stages of development remains a possibility.
Our work will facilitate the future characterization of the functional significance of phosphorylation of the DNA methyltransferase. The preliminary results showing that phosphorylation does not alter the enzymatic activity of the DNA methyltransferase is understandable in light of the fact that the phosphorylated residue is in a domain that is unessential for catalysis. These in vitro experiments need to be extended to other types of functional assays including modulation of inhi-bition by nucleic acids and interactions with other proteins. Our ability to express and purify recombinant DNA methyltransferase should facilitate in vitro studies of site-directed mutants (37,38). Other model systems, such as COS cell expression (39) or transgenic mice (1) might be useful in assessing the functional significance of phosphorylation of serine 514 in vivo.