Recombinant Human DNA (Cytosine-5) Methyltransferase

Initial velocity determinations were conducted with human DNA (cytosine-5) methyltransferase (DNMT1) on unmethylated and hemimethylated DNA templates in order to assess the mechanism of the reaction. Initial velocity data with DNA andS-adenosylmethionine (AdoMet) as variable substrates and product inhibition studies with methylated DNA andS-adenosylhomocysteine (AdoHcy) were obtained and evaluated as double-reciprocal plots. These relationships were linear for plasmid DNA, exon-1 from the imprinted small nuclear ribonucleoprotein-associated polypeptide N, (CGG·CCG)12, (m5CGG·CCG)12, and (CGG·CCG)73 but were not linear for (CGG·Cm5CG)12. Inhibition by AdoHcy was apparently competitive versus AdoMet and uncompetitive/noncompetitive versus DNA at ≤20 μm AdoMet. Addition of the product (methylated DNA) to unmethylated plasmid DNA increased V max(app)resulting in mixed stimulation and inhibition. Velocity equations indicated a two-step mechanism as follows: first, activation of DNMT1 by methylated DNA that bound to an allosteric site, and second, the addition of AdoMet and DNA to the catalytic site. The preference of DNMT1 for hemimethylated DNA may be the result of positive cooperativity of AdoMet binding mediated by allosteric activation by the methylated CG steps. We propose that this activation plays a rolein vivo in the regulation of maintenance methylation.

Initial velocity determinations were conducted with human DNA (cytosine-5) methyltransferase (DNMT1) on unmethylated and hemimethylated DNA templates in order to assess the mechanism of the reaction. Initial velocity data with DNA and S-adenosylmethionine (AdoMet) as variable substrates and product inhibition studies with methylated DNA and S-adenosylhomocysteine (AdoHcy) were obtained and evaluated as double-reciprocal plots. These relationships were linear for plasmid DNA, exon-1 from the imprinted small nuclear ribonucleoprotein-associated polypeptide N, (CGG⅐CCG) 12 , (m 5 CGG⅐CCG) 12 , and (CGG⅐CCG) 73 but were not linear for (CGG⅐Cm 5 CG) 12 . Inhibition by AdoHcy was apparently competitive versus AdoMet and uncompetitive/noncompetitive versus DNA at <20 M AdoMet. Addition of the product (methylated DNA) to unmethylated plasmid DNA increased V max(app) resulting in mixed stimulation and inhibition. Velocity equations indicated a two-step mechanism as follows: first, activation of DNMT1 by methylated DNA that bound to an allosteric site, and second, the addition of AdoMet and DNA to the catalytic site. The preference of DNMT1 for hemimethylated DNA may be the result of positive cooperativity of AdoMet binding mediated by allosteric activation by the methylated CG steps. We propose that this activation plays a role in vivo in the regulation of maintenance methylation.
The genome of most organisms contains modified nucleotides including N 6 -methyladenine, N 4 -methylcytosine, and C 5 -methylcytosine (m 5 C) 1 (1)(2)(3). However, both the biological significance and the types of DNA methylation differ greatly between prokaryotes and eukaryotes. In prokaryotes, most modified bases participate in restriction-modification, a defense mechanism that protects the host from heterologous phage infection (4). In addition, N 6 -methyladenine plays a role in the initiation of DNA replication (5) and in post-replicative methyl-directed mismatch repair (6).
In higher eukaryotes, DNA methylation is confined to m 5 C and is implicated in the regulation of development, genomic imprinting (7,8), X chromosome inactivation, gene expression (9), and retrotransposon inactivation (10 -12). In mammals, the patterns of methylation are inherited from both parental genomes but are erased and reconstructed (de novo methylation) in somatic cells following implantation (13,14). Such patterns are then copied and maintained by hemimethylation of the daughter strands during semiconservative DNA synthesis in S phase (maintenance methylation) (15,16). However, the mechanisms by which both de novo and maintenance methylation occur remain to be elucidated.
Several DNA methyltransferases have been isolated from human and mouse (17)(18)(19)(20), but it is not clear whether the de novo and maintenance methylations are carried out by separate proteins in vivo or whether both activities are shared by one or more enzymes (21)(22)(23)(24)(25)(26). Nevertheless, a role for in vivo methylation has been established for isoforms of the human DNMT1 gene (27,28) whose product, DNMT1, methylates C at a CG dinucleotide step, both in single-stranded and doublestranded, unmethylated or hemimethylated, templates (29 -34). Hemimethylated templates are the most effective for the reaction, and methylation rates increase in the neighborhood of pre-existing m 5 C residues (35)(36)(37)(38). However, little is known about the mechanisms responsible for this preference.
DNMT1 has a bipartite structure with the C-terminal 570 amino acids containing the catalytic domain. This region shares sequence homologies with all prokaryotic type II cytosine-5 methyltransferases (39,40), including a PC dipeptide motif that is part of the catalytic center in the crystal structures of M.HhaI (41) and M.HaeIII (42), and the binding site for S-adenosylmethionine (AdoMet), the methyl donor for methyltransferases (43,44). The remaining ϳ1000 N-terminal amino acids, which are not present in the prokaryotic enzymes, contain a nuclear localization signal, a replication foci targeting sequence (15), and are important in the discrimination between unmethylated and hemimethylated substrates (33).
Herein, we report kinetic analyses with the human fulllength, recombinant, DNMT1 on a variety of DNA substrates with the aim of learning about the mechanism of the methyl transfer reaction and the role of DNA in regulating the enzyme activity. The results confirm the preference of DNMT1 for pre-methylated DNA; however, the kinetics reveal a complex behavior with DNA substrates that are bound more tightly. The reaction follows a sequential mechanism whereby both substrates (DNA and AdoMet) must bind to the enzyme before any product (methylated DNA and AdoHcy) is released and is consistent with a two-step process. First, DNMT1 binds DNA at an allosteric site (probably in the N-terminal domain) and activates the catalytic center, and second, AdoMet and the DNA (which may either be the same molecule bound to the regulatory site or a new DNA molecule) occupy the catalytic site. Allosteric binding of pre-methylated CG is proposed to increase the accessibility of AdoMet to the catalytic center, which then results in an acceleration of the reaction rate.

EXPERIMENTAL PROCEDURES
Enzyme Assay for Initial Velocities in the Absence of Products-Recombinant DNMT1 was expressed and purified as described (45). Unmethylated or hemimethylated DNA was incubated with 40 nM DNMT1 and S-(5Ј-adenosyl)-L-[methyl-3 H]methionine (AdoMet) for 30 min at 37°C in a total volume of 25 l in buffer A (100 mM Tris⅐HCl, pH 7.8, at 25°C, 1 mM Na 2 EDTA, 1 mM dithiothreitol, 7 g/ml phenylmethylsulfonyl fluoride, 5% glycerol, and 100 g/ml bovine serum albumin). Fig. 1 shows that reactions were linear under these conditions. On the other hand, enzyme concentrations greater than 50 nM (inset to Fig. 1B 12 and (m 5 CGG⅐CCG) 12 , 0.1-1.0 M unmethylated CG and 1.02-10.0 M AdoMet; for (CGG⅐CCG) 73 , 0.12-1.0 M CG and 2.11-20.0 M AdoMet. For d(I-C⅐I-C) ϳ7000 , only 1 nM DNMT1 was used since the reactions were much faster. In this case, the concentration range of the substrates was 0.1-1.0 M for CI and 1.02-20 M for AdoMet. At the end of the reaction, samples were quenched in a dry ice/ethanol bath, spotted on DE81 ion exchange chromatography filters (Whatman), placed on a 12-well manifold apparatus (Millipore), washed, dried, and counted (45). The efficiency of scintillation counting was calculated for every round of washing from the counts/min obtained in two filters that were loaded with an internal control (IC filter). This control consisted of 25 l of plasmid DNA extensively pre-methylated with AdoMet and bacterial SssI methylase (New England Biolabs); unincorporated AdoMet was removed by Sephadex G-50 fine (Amersham Pharmacia Biotech) column chromatography. The eluate was reconstituted in buffer A to a concentration of ϳ6000 cpm/25 l (IC solution). The nanomolar [ 3 H]CH 3 (N) present in the DNA at the end of the reaction was calculated as n ϭ (C Ϫ B)/RF, where C was the counts/min of a sample, B the blank (cpm in the absence of DNMT1), R the ratio of IC filter/IC solution, and F the cpm/nM free AdoMet. The initial velocity (v) was obtained as 1/v ϭ ET/N, where E was the concentration of DNMT1 and T the reaction time. The lower levels of incorporations (which were associated with the data points at the lowest concentrations of unmethylated (CGG⅐CCG) n ) were approximately twice their blank values. Data points were collected in duplicate, graphed on Lineweaver-Burk double-reciprocal plots, and fit to weighted (weight was v 2 ) linear regressions. These plots contain two independent variables, namely the reciprocal of the apparent V max (V max(app) ), given by the intercepts, and the ratio of the Michaelis constant (K m ) (plus the dissociation constant) for the variable substrate to V max(app) , given by the slopes. By re-plotting these intercepts and slopes derived from various combinations of substrates, new linear plots were obtained, which yield the true V max , K m , and the dissociation constant for the substrates. Non-linear fits were not weighted. When appropriate, the coordinates of the intersection for the family of reciprocal plots were calculated from the kinetic constants obtained from the replots of slope and intercept and applied as a constraint during curve fitting.
Enzyme Assay for Initial Velocities in the Presence of Products-Addition of known concentrations of the products to an enzymatic reaction (product inhibition) is a powerful strategy for deciphering an enzyme mechanism, i.e. whether the substrates bind (and the products dissociate) in an ordered or random fashion. Methylation reactions on supercoiled pRW3602 were performed in the presence of added S-(5Јadenosyl)-L-homocysteine (AdoHcy) or methylated DNA as product inhibitors. Assays were performed as described but, in addition, they contained fixed amounts of either of the inhibitors. In the experiments with AdoHcy, the concentration range of the reactants was 4.0 -25.0 M CG, 2.0 -40.0 M AdoMet, and 2.5-25.0 M AdoHcy. Two methylated DNAs were used as products. The first was a 40-bp duplex oligonucleotide [( Me CG) 20 ]) containing canonical Watson-Crick pairs but with 5-methylcytosines substituting for all cytosines. The second was a 36-bp duplex oligonucleotide with the sequence CGG(F 5 CGG) 11 ⅐(Cm 5 CG) 12 , named ( F/Me CG) 12 , where F 5 C designates 5-fluorocytosine. The rationale for using this fluorinated oligonucleotide was that, contrary to [( Me CG) 20 ], ( F/Me CG) 12  Chemicals and DNA-AdoMet at the specific activities of 15.0 Ci/ mmol (1 Ci ϭ 37 GBq), 500 mCi/mmol, and 60 -85 Ci/mmol was obtained from Amersham Pharmacia Biotech. AdoHcy (Sigma) was prepared by dissolving 100 mg into 5 ml of 1 N HCl to give a 52 mM solution. Dilutions were made fresh in distilled water.
Plasmid pRW3602 was a derivative of pUC9 that contained a 40-bp insert (TTAAGCAGCAGTATCCTCTTGGGGGCGCCTTCCCCACACT) from the human ␥-globin promoters in its HincII site (46). The plasmid was 2705 bp in length with a total of 338 CG steps (on average, 1 every 8 bp). (CGG⅐CCG) 73 was an XbaI-BamHI restriction fragment from plasmid pRW3691 and contained a total of 156 CG steps (including 10 from the triplet repeat flanking sequences) (47). d(I-C⅐I-C) ϳ7000 (48) was from Amersham Pharmacia Biotech. (CGG⅐CCG) 12 , (CGG⅐Cm 5 CG) 12 , (m 5 CGG⅐CCG) 12  give a stock solution of 500 M CG steps. Oligonucleotide concentrations were obtained by optical absorption on a Beckman DU 640 spectrophotometer using the extinction coefficients calculated from the known coefficients of the component nucleotides. Since this synthetic 40-mer contains only m 5 C at all CG steps and contains no CNG sequences, it does serve as a substrate for DNMT1 and was studied as a product inhibitor. A list of all the substrates and inhibitors used and their relevant properties is given in Table I. Fig. 2 shows the basic scheme for the reaction of DNMT1 with DNA and AdoMet as obtained from the initial velocity data and product inhibition studies. Knowledge of the reaction sequence is essential for the derivation of all the kinetic constants, since the meaning of slopes and intercepts of the double-reciprocal plots (as well as their replots) is based on the velocity equation derived from the mechanism itself. Fig.  2A shows that the first step involves the binding of DNA (D) to a regulatory site in DNMT1 (E), distinct from the catalytic site, to give a DNA-DNMT1 initial complex. The second step consists in the binding of a CG⅐CG (or CG⅐m 5 CG or m 5 CG⅐CG) and AdoMet (Am) to the catalytic site to give the ternary complex competent for catalysis.

Velocity Equations, Initial Forward Velocity in the Absence of Products-
Based on the initial velocity patterns that were obtained, the reaction is treated at steady state (the turnover rate of the enzyme is not limited by how fast the chemical reaction takes place) except for the interconversion between free DNMT1 and DNA-DNMT1, which must occur at equilibrium within the steady state. Addition of the substrates is shown to be ordered, with AdoMet preceding the DNA; however, this sequence of addition is not proven by the present data, as discussed later. Indeed, the reaction may be ordered with DNA binding before AdoMet or random, where either substrate can bind first. The velocity equations for an ordered or random sequential bireactant system (such as this) are identical, and both V max and the Michaelis constant for the substrates (K m CG and K m AdoMet ) can be derived, even though the mechanism is unknown. However, the dissociation constant K ia for the first substrate that adds to the reaction cannot be assigned unless it is identified. Our derivation of the following velocity Equation 1 was based on the formulations by King-Altman and Cha, which are described in detail by Segel (49). The method requires that the section of the reaction at equilibrium be grouped into a single corner of the scheme, as illustrated in Fig. 2B. The rate k 3 [AdoMet] is then corrected for f 1 , the ratio of DNA-DNMT1 to DNA-DNMT1 plus free DNMT1. In the following equations, [CG] is given the same meaning as [D] (DNA) in Fig. 2.
The velocity Equation 1, in the absence of products, is as follows. v FIG. 2. Scheme of an ordered Bi Bi mechanism with allosteric activation for the methyltransferase reaction by DNMT1. A, free DNMT1 (E) binds the activator DNA (D) at a regulatory site and subsequently AdoMet (Am) and a second molecule of DNA to the catalytic site. The sequence of addition of AdoMet and the DNA to the catalytic site is arbitrary. The kinetic data do not distinguish whether AdoMet, or the DNA, or both are the first to bind to this site. B, E, and DE are grouped together since their interconversion is at equilibrium relative to the other enzyme species. This assumption is made in order to develop a velocity equation that yields linear responses; f 1 is DE/(E ϩ DE); M and Ah signify the products of the reaction, methylated DNA and AdoHcy, respectively.  12 Chemically synthesized oligonucleotide (36 bp) containing the triplet repeat sequence whose expansion in the 5Ј-UTR a of the FMR1 gene (Xq27.3) causes fragile-X syndrome (m 5 CGG ⅐ CCG) 12 Chemically synthesized oligonucleotide (36 bp) where the upper strand of the triplet repeat contains 5-methylcytosine at all CG steps (DNA hemimethylated on the upper strand) (CGG ⅐ Cm 5 CG) 12 Chemically synthesized oligonucleotide (36 bp) where the lower strand of the triplet repeat contains 5-methylcytosine at all CG steps (DNA hemimethylated on the lower strand) (CGG ⅐ CCG) 73 Restriction fragment of ϳ256 bp from pRW3691 (47) containing 146 CG steps from 73 consecutive copies of CGG ⅐ CCG and an additional 8 CG steps from the flanking DNA sequences poly d(I-C) ⅐ poly d(I-C) Synthetic polymer routinely used to assay for DNA methyltransferase activity DNA Inhibitors ( Me CG) 20 Chemically synthesized 40-bp duplex oligonucleotide ("Experimental Procedures") containing 5methylcytosine at all 20 CG steps but containing no CNG steps ( F/Me CG) 12 Chemically synthesized 36-bp duplex oligonucleotide containing the FMR1-associated triplet repeat (CGG ⅐ CCG) substituted with 5-fluorocytosines on the upper strand and 5-methylcytosines on the lower strand at the CG steps a UTR, untranslated region.

RESULTS
Initial velocity experiments enable the evaluation of kinetic constants and provide insights into the mechanism of a reaction. A comprehension of the mechanism of this key enzyme, human DNMT1, is critical for understanding its role in developmental processes and in the etiology of fragile X syndrome and other diseases (7)(8)(9)(10)50). A variety of DNA templates (Table I) were methylated to less than 5-8% of the total CG steps by purified, recombinant DNMT1 with the aim of assessing the effect of sequences flanking the substrate CG on the kinetic constants, the role of negative supercoiling, and the mechanism of the reaction. Experimental data were analyzed by graphing the extent of methylation as a function of concentration of DNA or AdoMet, as variable substrates, on doublereciprocal Lineweaver-Burk plots. This report is the second of a series of three papers describing the purification and characterization of DNMT1 (45) and focuses on the mechanism of the methyltransferase reaction. The third paper 2 will describe the effect of DNA topology on the reaction rates at CG sites in random as well as CGG⅐CCG repeat tracts and compare the kinetic properties of DNMT1 with the bacterial M.SssI.
Linear Velocity Responses-For bireactant enzymes (such as DNMT1), double-reciprocal plots generally give linear responses where 1/v is graphed as a function of 1/substrate. For most of the DNAs used, which included supercoiled pRW3602 as purified from Escherichia coli (Ϫ ϭ 0.045), relaxed circular (Ϫ ϭ 0), or linear, as well as the SNRPN oligonucleotide (unmethylated or hemimethylated), (CGG⅐CCG) 12 , (m 5 CGG⅐CCG) 12 , and (CGG⅐CCG) 73 , the velocity curves were linear with respect to the variable substrate, whether this was the DNA or AdoMet. Fig. 3 shows the data with supercoiled pRW3602. Fig. 3A shows the concentration of [ 3 H]CH 3 groups incorporated when the DNA was the variable substrate (on the x axis) and AdoMet the fixed substrate. Conversely, Fig. 3B shows the results with AdoMet as the variable substrate and DNA as the fixed substrate. For all of the DNA templates listed above, the velocity patterns were as in Fig. 3, i.e. they converged to the left of the y axis and above or below the x axis for both substrates. Fig. 4 shows the double-reciprocal plots for the triplet repeat sequences (CGG⅐CCG) 12 (Fig. 4, A and B), (m 5 CGG⅐CCG) 12 (Fig. 4, C and D), and (CGG⅐CCG) 73 (Fig. 4,  E and F). These patterns contrast with those obtained with M.HhaI (51) and M.SssI 2 methylases, where the families of lines converge on the y axis when AdoMet is the variable substrate. This result shows that AdoMet and DNA do not bind by an ordered and rapid equilibrium mechanism to DNMT1, as in the case with M.HhaI and M.SssI. Fig. 5 shows the replots of the slope and y axis intercept (1/V max(app) ) for each of the lines in Fig. 3 that were used to derive the kinetic constants. As shown in Equations 2 and 3, four replots are possible that give the following constants: (a) 1/V max on the y axis intercept (Fig. 5A) Fig. 3B as a function of 1/CG. Therefore, these four replots yield the maximum velocity and the Michaelis constants for AdoMet (at DNA ϭ ϱ) and DNA (at AdoMet ϭ ϱ). As pointed out previously, the dissociation constant K ia cannot be assigned to the DNA or AdoMet because the method does not distinguish the order of addition. The experimental values for the constants are reported in the companion papers 2 (45).
Curved Velocity Responses-Unexpectedly, two template DNAs (d(I-C⅐I-C) ϳ7000 and (CGG⅐Cm 5 CG) 12 ) gave non-linear initial velocity curves. The results for d(I-C⅐I-C) ϳ7000 are described in the accompanying paper (45), whereas the doublereciprocal plots for (CGG⅐Cm 5 CG) 12 are reported in Fig. 6. Fig.  6A shows that the methylation rate was linear when DNA was the variable substrate. The data also indicate that, contrary to Figs. 3 and 4, velocities were maximal at 1.74 M AdoMet, such that further increases (up to 10.0 M) did not result in a decrease in slope or intercept. Fig. 6B shows that plots were not linear when AdoMet was the variable substrate. Velocities were unchanged, and plots were parallel to the x axis when AdoMet concentrations rose above ϳ2 M. The responses were still dependent on DNA concentration, since 1/V max(app) (y axis intercepts) decreased with increasing CG content. A replot of 1/V max(app) versus 1/CG was linear (Fig. 7A), whereas both intercept and slope replots from the data at fixed AdoMet were curved (Fig. 7B).
This result indicates that when this particular DNA was bound to DNMT1, the velocity of the reaction was already maximal at a very low (ϳ2 M) AdoMet concentration, contrary to the expectation that maximum velocity requires infinite amounts of AdoMet. The plots in Figs. 6 and 7 still enable the calculation of 1/V max , 1/K m CG , and 1/K m AdoMet ; however, K ia cannot be derived. The most significant conclusion that can be drawn from these results is that the kinetic behavior of DNMT1 may be dramatically altered by both sequence of the DNA template and its pre-methylation status. Obviously, the scheme in Fig. 2 and its velocity equations are not adequate to describe the results with (CGG⅐Cm 5 CG) 12 which implies that DNMT1 is capable of complex kinetics. Overall, this combination of linear plus non-linear responses indicates a steady-state mechanism, where the DNA can act simultaneously both as a substrate and as an activator for the reaction. The activation is proposed to occur through binding of a DNA molecule at a site distinct from the catalytic center, i.e. an allosteric, or regulatory, site. The precise sequence of the chemical steps that lead to such nonlinear responses, however, is unknown. Nevertheless, it seems clear that the role of the DNA bound to the allosteric site is to increase the affinity of the DNA-DNMT1 complex for AdoMet since low levels of AdoMet are sufficient to maximally drive the reaction.
It is noteworthy that the complex enzymatic behavior is associated with a DNA sequence, the (CGG⅐CCG) n triplet repeat, whose expansion in the chromosomal FRAX locus leads to aberrant methylation and to disease in humans.
In summary, these studies indicate that the methylation reaction by DNMT1 may follow complex mechanisms, and both the sequence composition as well as the methylation status of the DNA substrate contribute to this complexity.
Product Inhibition with AdoHcy-Product inhibition studies of bireactant enzymes provide a means to distinguish random from ordered sequential Bi Bi mechanisms. Ordered systems give competitive patterns with the first substrate that binds to the enzyme versus the last product that leaves the enzyme (the plots converge on the y axis) and non-competitive patterns with the other combinations (the lines converge to the left of the y axis). On the contrary, random mechanisms give competitive inhibition between like substrates and products (with similar chemical structures) and non-competitive patterns between unlike reactants. In our case, determining whether the reaction is ordered or random would enable the assignment of the dissociation constant K ia to the DNA, to AdoMet, or to both in the case of a random system.
In In double-reciprocal plots where AdoMet was the variable substrate, AdoHcy the changing-fixed inhibitor (like substrate and product), and CG the fixed co-substrate (4.0 to 25.0 M), velocities increased, as expected, up to 20.0 M AdoMet; however, higher concentrations caused strong inhibition by AdoHcy, a result that was not anticipated. In fact, the expectation was that AdoMet would progressively overcome the inhibition by AdoHcy, linearly increasing the reaction rates as its concentration rose. The slopes obtained from the 1/v versus 1/CG plots (which were linear) at various AdoHcy concentrations were graphed as a function of AdoMet. Increasing AdoMet up to 50 M progressively reduced the slopes (higher velocities) to plateau levels in the absence of AdoHcy, as expected. AdoHcy increased all of the slopes as a result of inhibition, and increasing AdoMet progressively reduced such an inhibition. However, at concentrations of AdoMet Ͼ20 M, there was a new, strong non-competitive inhibition by AdoHcy that was then reduced by higher levels of AdoMet. This pattern found at greater than 20.0 M AdoMet is anomalous; the reason for this behavior is uncertain. When these unusual data at Ͼ20.0 M AdoMet were excluded, the pattern of inhibition versus AdoHcy (like substrate and product) was competitive (Fig. 8A).
To test whether there was evidence for more than one binding site for AdoMet in DNMT1, a Dixon plot (49) was constructed. Slopes 1/CG were replotted as a function of 1/AdoMet for AdoMet Յ20.0 M, at each fixed value of AdoHcy. The replots were linear, and their slopes (Slope CG/AdoMet ) were finally graphed as a function of AdoHcy (Fig. 8B). The result was a linear, rather than a parabolic, curve indicating that only one AdoMet-binding site per DNMT1 molecule was detected. The x axis intercept gives the K i for AdoHcy, which is ϳ14 M.
Product Inhibition with Methylated and Fluorinated DNA-The second part of the inhibition studies consisted of the use of the other product, fully methylated DNA. In conjunction with the data with AdoHcy, fully methylated DNA was expected to give non-competitive inhibition patterns versus both substrates for an ordered steady-state mechanism or competitive inhibition versus DNA for a random mechanism. A concentration range of 2.5 to 60.0 M of m 5 CG steps containing 5-methylcytosine in a 40-bp duplex synthetic oligonucleotide ( Me CG) 20 was added to 4.00 to 25.0 M CG from supercoiled pRW3602 (like substrate and product) and 6.67 or 30.0 M AdoMet. Fig. 9A shows the replot of intercepts from double-reciprocal plots where pRW3602 was the variable substrate and ( Me CG) 20 the changing-fixed inhibitor, with AdoMet fixed at 6.67 M. Intercepts were expected not to change for a competitive system or increase with increasing m 5 CG for a non-competitive mechanism. On the contrary, whereas some scatter was observed, 1/V max(app) decreased as more inhibitor was added, indicating enzyme activation rather than inhibition. Slope effects were more complex but not substantial. At 30.0 M AdoMet, velocities with 2.5, 5.0, and 10.0 M ( Me CG) 20 were indistinguishable from those in the absence of inhibitor, whereas 20, 30, and 40 M ( Me CG) 20 caused inhibition. These latter three concentrations gave parabolic curves, whereas, in all cases, intercepts were unchanged (Fig. 9B). Overall, these data indicate that fully methylated DNA acted in two ways as follows: first, it inhibited the reaction by competing with unmethylated DNA for the catalytic center, and second, it accelerated the turnover number of the enzyme by binding to an allosteric site.
To verify this dual role of methylated DNA further, experiments were carried out with an oligonucleotide (CGG(F 5 CGG) 11 ⅐(Cm 5 CG) 12 ) that contained m 5 CG on one strand and F 5 CG (5-fluorocytosine) steps on the complementary strand [( F/Me CG) 12 ]. ( F/Me CG) 12 has two characteristics as follows: on the one hand, it acts as a dead-end inhibitor since F 5 CG binds to DNMT1 in the presence of AdoMet and traps the enzyme into a stable DNMT1-AdoMet-DNA complex that does not proceed through catalysis (52)(53)(54). As a result, this causes inhibition. On the other hand, the DNA sequence and methylation status of ( F/Me CG) 12 is identical to that of (CGG⅐Cm 5 CG) 12 , the substrate that produced the complex enzymatic patterns in Figs. 6 and 7. Thus, it was of interest to determine if ( F/Me CG) 12 acted as an inhibitor, an activator, or both. Eleven nM to 40.0 M of modified CG steps from ( F/Me CG) 12 was added to 4.00 to 25.0 M CG from supercoiled pRW3602 and 6.67 or 30.0 M AdoMet. Fig. 10, A and B, shows the intercept and slope replots from double-reciprocal plots where pRW3602 was the variable substrate and ( F/Me CG) 12 the fixed inhibitor (like substrate and inhibitor). The Lineweaver-Burk plots were linear. ( F/Me CG) 12 caused a decrease in both intercepts and slopes at 6.67 M AdoMet (filled circles) and had no effect at 30.0 M AdoMet (open circles). Interestingly, the decreases in intercepts occurred at nanomolar concentrations of the added oligonucleotide, suggesting that ( F/Me CG) 12 was far more potent as an activator than as an inhibitor.
To verify this conclusion, control reactions were performed in FIG. 7. Replots of slopes and intercepts for initial velocities with (CGG⅐Cm 5 CG) 12 . A, replot of y axis intercepts from the 1/v versus 1/AdoMet data shown in Fig. 6B. B, replot of y axis intercepts and slopes from the 1/v versus 1/CG shown in Fig. 6A. the presence of ( F/Me CG) 12 alone since DNMT1-AdoMet-DNA complexes, which are retained on DE81 filters during the sample processing, would lead to erroneously high values of labeled methyl groups for pRW3602. These controls confirmed that the oligonucleotide acted as an activator rather than an inhibitor.
In summary, these product inhibition studies support the finding obtained with (CGG⅐Cm 5 CG) 12 that methylated DNA is an activator of the methyl transfer reaction. On the other hand, since this behavior complicated the inhibition patterns, it could not be used in conjunction with the AdoHcy experiments to distinguish whether the reaction proceeds through a random or an ordered mechanism. Therefore, the dissociation constant K ia cannot be assigned to either the DNA or AdoMet. DISCUSSION Mammalian DNMT1 was reported to methylate hemimethylated DNA to a greater extent (2-5-fold) than unmethylated DNA (32)(33)(34)38), an observation confirmed here with a new recombinant enzyme that displays an ϳ30-fold higher activity than previously obtained (34). It was also observed that methylation rates increase when duplex, or single-stranded, DNA contains randomly pre-methylated m 5 CG⅐CG, m 5 CG⅐m 5 CG, or m 5 CG steps (a phenomenon known as methylation spreading) and that such a stimulation extends in trans to unmethylated molecules (Refs. 36 and 37 and references therein and Refs. 55 and 56). Furthermore, allosteric transitions in DNMT1-methylated DNA complexes have been proposed, based on non-Michaelis-Menten patterns of methylation with hemimethyl-ated templates (57).
Our studies indicate that the presence of m 5 CG (either in hemimethylated of fully methylated templates) stimulates the reaction. The effect is likely mediated by DNA binding to an allosteric site, since neither ( Me CG) 20 nor the ( F/Me CG) 12 duplexes served as alternative substrates in control experiments.
This effect could be achieved in two ways as follows: (a) m 5 CG "exposes" the enzyme active site, which is otherwise less accessible and/or, (b) m 5 CG binding modifies the active site conformation, improving its fit for AdoMet and/or the DNA. The data with (CGG⅐Cm 5 CG) 12 , which showed saturation at 2 M AdoMet, suggest an active role for m 5 CG in shaping the AdoMet binding pocket. Furthermore, the results reveal a role for the DNA primary sequence in modulating substrate binding specificity and, ultimately, the catalytic rates (45,58).
The turnover number for the enzyme varied considerably among the DNA templates, from ϳ1 to 50 h Ϫ1 , but was lower than for d(I-C⅐I-C) ϳ7000 , 184 h Ϫ1 (45). These results seem to exclude the methyl transfer as the rate-limiting step in the reaction. On the other hand, it was found that methylation rates increase with negative supercoiling 2 suggesting that, instead, substrate binding and/or product release limit the turnover rate (59).
Overall, the initial velocity data are consistent with a steadystate sequential Bi Bi mechanism (either ordered or random). The linear double-reciprocal plots are adequately described by the reaction scheme in Fig. 2 and velocity equations (Equations 1-3) developed according to steady-state assumptions that require the addition of both substrates before any product is released. However, the scheme is not sufficient to explain the curved responses obtained with (CGG⅐Cm 5 CG) 12  C) ϳ7000 , suggesting that the reaction may take alternative pathways.
It was reported that DNMT1 is processive, based on the observations that methylation rates increase with the length of the DNA, that NaCl inhibits the reaction in a concentrationdependent manner (57,60), and that DNA-protein associations are rather stable (61,62). The linear velocity patterns obtained here with the various DNA templates, including the closely spaced CG steps in (CGG⅐CCG) n , were accounted for by a reaction scheme whereby the enzyme dissociates from the DNA after each reaction cycle. Thus, additional terms associated with processivity were not necessary in the final velocity equation. The higher V max observed for longer, rather than shorter, polymers such as (CGG⅐CCG) 73 versus (CGG⅐CCG) 12 (45) 2 does not require a processive mechanism. The one-dimensional limited diffusion (63)(64)(65)(66) and/or intersegment transfer (67)(68)(69)(70) processes, characteristic of DNA-binding enzymes, account for this result. The theory underlying such mechanisms states that macromolecular collision in solution is not elastic, so that when a protein collides with a DNA molecule, it stays along the contour length of the DNA through repetitive microcollisions (which entail dissociations and re-associations), rather than drifting away. Consequently, the time spent by a protein on one DNA molecule of length n would be greater than the time spent by the same protein on combined m DNA molecules of length n/m. Therefore, we cannot conclude that DNMT1 acted processively.
These kinetic determinations have implications for the in vivo reactions of maintenance methylation. For example, the pre-methylated parental strand during DNA replication may act as an allosteric activator to direct reactions on the unmethylated daughter strand. Also, the ability of the enzyme to bind AdoMet in equilibrium may be exploited to accelerate reaction rates. It is possible that turnover rates are higher in vivo, where DNMT1 may associate with multienzyme complexes, such as the replication or the mismatch repair systems, that afford high processivity rates that can compensate for the limitations imposed in vitro with DNMT1 alone. In summary, the human DNMT1 is a complex enzyme from the standpoint of protein structure, the ability of substrates to conduct regulatory functions, and kinetic behavior on different types of substrates.