Functional characterization of Escherichia coli DNA adenine methyltransferase, a novel target for antibiotics.

We have characterized Escherichia coli DNA adenine methyltransferase, a critical regulator of bacterial virulence. Steady-state kinetics, product inhibition, and isotope exchange studies are consistent with a kinetic mechanism in which the cofactor S-adenosylmethionine binds first, followed by sequence-specific DNA binding and catalysis. The enzyme has a fast methyl transfer step followed by slower product release steps, and we directly demonstrate the competence of the enzyme cofactor complex. Methylation of adjacent GATC sites is distributive with DNA derived from a genetic element that controls the transcription of the adjacent genes. This indicates that the first methylation event is followed by enzyme release. The affinity of the enzyme for both DNA and S-adenosylmethionine was determined. Our studies provide a basis for further structural and functional analysis of this important enzyme and for the identification of inhibitors for potential therapeutic applications.

Bacterial DNA methyltransferases generate N 4 -methylcytosine, C 5 -methylcytosine, and N 6 -methyladenosine in an S-adenosylmethionine-dependent reaction (1). Bacterial DNA methylation plays critical roles, including DNA repair, phage protection, gene regulation, and DNA replication, in diverse biological pathways. The majority of DNA methyltransferases form one-half of a restriction-modification system that protects the host bacteria against bacteriophage infection. Together with cognate restriction endonucleases, which generally cleave a short palindromic sequence, these restriction-modification systems provide the foundation for many recombinant DNA manipulations; the endonucleases and methyltransferases have provided many structural and mechanistic insights into the process of sequence-specific DNA recognition and modification.
Not all DNA methyltransferases have an endonuclease partner or at least one which is known. Thus, DNA adenine methyltransferase (DAM, 1 methylates the adenine in GATC) in ␥-proteobacteria (2,3), and the cell cycle-regulated methyltransferase (CcrM, methylates the adenine in GANTC) in ␣-proteobacteria (3,4) are involved in post-replicative mismatch repair, DNA replication timing, cell cycle regulation, and the control of gene expression. DAM and CcrM have been identified as new targets for antibiotic development (5) because some pathogenic bacteria are either avirulent or not viable when the corresponding genes are removed. DNA adenine methylation regulates the pili formation genes in Escherichia coli and Salmonella, providing one of the first and clearest examples of epigenetic gene regulation (2). This DNA-mediated gene regulation involves differentially methylated GATC sites, which represent a small minority of the ϳ5,000 -20,000 GATC sites found in a typical bacterial genome.
E. coli DAM is a functional monomer of 278 amino acids (6). Our present understanding of how this enzyme functions is based largely on a small number of reports (6 -10). Herman and Modrich (6) first characterized the enzyme with plasmid DNA, providing various kinetic parameters and evidence in support of a non-processive path during the methylation of multisite substrates. The apparent K m values (AdoMet and DNA), and the apparent turnover number (19 min Ϫ1 ) are similar to those determined for other DNA methyltransferases. The finding that the enzyme dissociates from DNA after an initial methylation cycle, at least with DNA in which the next site is 2455 bp away, has several biological implications (6). For example, the methylation state of two GATC sites in the Pap regulon in part determines the phase variation of pili formation (2). We demonstrated that EcoRI DNA methyltransferase (M.EcoRI) is not processive with substrates in which adjacent sites are only 50 bp apart. We also identified the interesting trend that methyltransferases not part of a known restriction-modification system act processively, whereas those that are part of a restriction-modification system act distributively (11,12). The demonstration that CcrM is a processive enzyme certainly fits into this pattern (13). Based on the results of Hermann and Modrich (6), DAM appears not to follow this trend. Subsequently, Bergerat et al. (8), provided evidence that the ability of DAM to methylate a particular GATC site depends on DNA sequences involving 2-3 bp flanking the recognition site. This qualitative analysis provided intriguing insights about how DAM might differentially methylate a subset of the thousands of GATC sites in a bacterial genome. The number and sequence context of such differentially methylated GATC sites remains poorly characterized (8). In contrast to the earlier work, Bergerat et al. (8) showed clear evidence for processive catalysis when adjacent sites were as close as 18 bp apart and in large substrates with longer inter-site distances. Recently, Urig et al. (7) also showed evidence for processive catalysis with DNA of various lengths.
Our interest in DAM from E. coli and other bacteria is to understand how DNA adenine methylation regulates gene ex-pression and to develop therapeutics based on intervening in this process, either directly against DAM or against other components of the regulatory machinery. Here we present a further characterization of E. coli DAM. The enzyme's order of substrate and product binding and dissociation have been identified; the rate-limiting step during catalysis is product dissociation; and the distributive catalysis of the enzyme on a regulatory genetic element containing two GATC sites is shown. We also report on the stability constants of several enzyme complexes involving cofactors, cofactor analogs, and DNA. Our results provide the basis for a better understanding of this critical bacterial protein.

MATERIALS AND METHODS
SyberGold, calf thymus DNA, and ScintiVerse scintillation fluid were purchased from Molecular Probes, Roche Applied Science, and Fisher, respectively. Vistra Green, [methyl-3 H]AdoMet (average specific activity: 75 Ci/mmol), and [␥-32 P]ATP were from Amersham Biosciences. DE81 filters were from Whatman. S-Adenosylmethionine and sinefungin were from Sigma. T4 polynucleotide kinase was from New England Biolabs.
Determination of Protein Concentration-DAM concentrations were determined based on a molecular mass of 32 kDa and an extinction coefficient of ⑀ ϭ 1.15 ml/mg cm (6).
Cofactor Purification-S-Adenosylmethionine was further purified as described previously (14). All cofactor dilutions were in 0.1 N HCl.
Oligonucleotide Synthesis and Purification-Mechanistic analysis required the use of a single site DNA substrate. The hairpin substrate and unmethylated and hemimethylated 20-mer substrates are shown below. The target base and the recognition sequence are bolded and underlined; the complementary strand, if it does not contain an adenine that can undergo methylation, is simply underlined; and the hairpin turn is italic and underlined.
Substrate oligonucleotides were synthesized by Midland and purified by high pressure liquid chromatography using a Dynamax PureDNA column (Rainin Instrument Co.) according to the manufacturer's specifications. Oligonucleotides were stored in 10 mM Tris, pH 8.0, 1 mM EDTA. Concentrations were determined using calculated extinction coefficients. For gel mobility shift assays, DNA substrates were radiolabeled using [␥-32 P]ATP (Amersham Biosciences) and T4 polynucleotide kinase (New England Biolabs). Unincorporated label was removed with Bio-Gel P-6 spin columns (Bio-Rad).
Preparation and Purification of DAM-We purified DAM from two expression constructs. DAM, overexpressed in E. coli strain XL-2Blue (Stratagene) harboring plasmid pDAL572 (provided by Bruce A. Braaten, University of California, Santa Barbara, CA) was grown in LB media (10 g of bacto-tryptone, 5 g of bacto-yeast, and 10 g of NaCl/liter) supplemented with 25 g/ml kanamycin and 12.5 g/ml tetracycline at 37°C until an A 600 of 0.60 was reached. The cells were induced with 1 mM isopropyl 1-thio-␤-D-galactopyranoside and 0.05% L-arabinose, grown for 4 h at 37°C, centrifuged immediately, and stored at Ϫ20°C. Cells were resuspended in 120 ml of P-11 buffer (50 mM potassium phosphate buffer, pH 7.4, 10 mM ␤-mercaptoethanol, 10 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 10% glycerol), protease inhibitor mixture tablets (Roche Applied Science), and 0.8 M NaCl and then lysed by French press. The lysate was centrifuged at 14,000 rpm for 45 min in a Sorvall SA 600 rotor. The supernatant was loaded onto a 350-ml Whatman P-11 column and eluted with a salt gradient of 0.2-0.8 M NaCl. Fractions containing DAM were pooled, dialyzed in B-S buffer (20 mM potassium phosphate buffer, pH 7.0, 10 mM ␤-mercaptoethanol, 10 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 10% glycerol), and loaded onto a 30-ml blue Sepharose column (Pharmacia Corp.). DAM was eluted with a salt gradient of 0 -1.5 M NaCl, and the appropriate fractions were pooled, flash frozen, and stored at Ϫ70°C.
A histidine-tagged DAM (hDAM) was generated by amplifying the dam gene with the Expand TM high fidelity PCR system (Roche Applied Science) using 0.1 g of plasmid pTP166 input DNA (15). The purified PCR product was digested with NcoI and XhoI and cloned into PET-28a (Novagen). Three clones were sequenced, and one was found to have the correct nucleotide sequence. This plasmid was then transferred to the BL21(DE3)pLysS expression strain and selected on LB kanamycin/ chloramphenicol plates. Plating and growth were done in LB with kanamycin/chloramphenicol selection. The enzyme was overexpressed as an N-terminal His 6 -tagged protein. A 12-liter culture was grown to an A 595 of 0.6 and induced with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside for 3 h before harvesting. The cell pellet was stored at Ϫ80°C.
As needed, the cell pellet was thawed and resuspended in buffer A (50 mM potassium phosphate, pH 8.0, 0.5 M NaCl, 10% glycerol, 10 mM ␤-mercaptoethanol, and 20 mM imidazole) containing protease inhibitors (Roche Applied Science). The cells were lysed (1 mg/ml lysozyme, stirring for 30 min at 4°C), sonicated, and centrifuged. The cell lysate supernatant was loaded onto 1.5-2-ml nickel-nitrilotriacetic acid Superflow columns (Qiagen Inc.) pre-equilibrated on fast protein liquid chromatography with buffer A. An imidazole gradient in buffer A (20 -300 mM imidazole) eluted the protein, and the pooled fractions were dialyzed overnight in buffer C (50 mM potassium phosphate, pH 7.0, 50 mM NaCl, 10 mM EDTA, 10 mM ␤-mercaptoethanol, and 5% glycerol) at 4°C. This was loaded onto an SP-Sepharose FF FPLC column (AP Biotech), and the protein was collected across a salt gradient of 50 -600 mM NaCl. Pooled fractions were dialyzed overnight in DAM storage buffer (20 mM potassium phosphate, pH 7.5, 200 mM NaCl, 20% glycerol, 2 mM DTT, and 0.2 mM EDTA), flash frozen in small aliquots, and stored at Ϫ80°C. An 8-liter growth yielded 5 mg of 90% pure hDAM. Circular Dichroism Analysis of Protein Structure-Circular dichroism spectra were obtained using purified DAM and hDAM in 50 mM sodium phosphate, pH 7.0, at room temperature. Data were collected on an Aviv 202 circular dichroism spectrophotometer using a 500-l quartz fluorescence cuvette with a 0.2-cm slit width (Starna). Data were collected between 190 and 265 nm, and spectral data were fit to curves using KaleidaGraph (Synergy Software). Secondary predictions of both DAM and hDAM were done using multiple CD spectra and the program CDSSTR from DichroWeb (16).
Methyltransferase Assays-Filter binding assays monitored the incorporation of tritium-labeled methyl groups into DNA (14). Reactions contained 100 mM Tris, pH 8.0, 10 mM EDTA, 10 mM DTT, and 0.4 mg/ml BSA. Substrate concentrations were noted in each experiment. The methyltransferase was diluted in protein dilution buffer (20 mM KH 2 PO 4 , 200 mM NaCl, 0.2 mM EDTA, 0.2 mg/ml BSA, 2 mM DTT, and 10% glycerol). After incubation at 37°C, reactions were stopped by transferring aliquots to 2.5-cm Whatman DE81 circular filter papers. Filters were washed three times with 50 mM KH 2 PO 4 , once with 80% ethanol, once with 95% ethanol, and once with diethyl ether, for 10 min each time. Filters were air-dried, and tritium content was determined in 4 ml of Liquiscint using a Beckman model LS 1701 scintillation counter or ScintiVerse scintillation fluid using a Beckman model LS 6500.
K d AdoMet Determination-Native protein fluorescence-based K d AdoMet determination was done on a PerkinElmer Life Sciences LS50B as described previously (17). K d DNA Determination-DNA substrates were radiolabeled using [␥-32 P]ATP and T4 polynucleotide kinase. For K d DNA determination in the presence of sinefungin, binding assays containing 5 nM 32 P-labeled DNA and DAM from 10 -2000 nM in MRB (20 mM Tris, pH 8.0, 10 mM EDTA, 0.20 mg/ml BSA, and 10 mM DTT) and 50 M sinefungin were incubated at 37°C for 10 min (18). DAM was diluted into protein dilution buffer. Samples were loaded onto pre-run 10% non-denaturing polyacrylamide gels. Gels were run at 300 V for 2 h at 4°C. Gels were dried, exposed to image plates, and analyzed on a STORM 840 densitometer (Amersham Biosciences). Densitometry was performed using ImageQuant software (Amersham Biosciences). Dissociation constants were derived from data fit to rectangular hyperbolic equations using KaleidaGraph.
Processivity Analysis-Processivity was determined with a 326-bp PCR fragment with GATC sites at positions 122 and 225, which in-cludes the Pap operon (20), using a processivity assay as described previously (11,12). DAM (10 nM  Circular Dichroism Results-Circular dichroism spectra of DAM and hDAM were used in conjunction with CDSSTR (16) to make secondary structure predictions. Similar values were predicted for both enzymes. The predicted DAM percentages were 49% ␣-helix, 25% ␤-sheet, 7% turn, and 17% unordered, whereas the predicted hDAM percentages were 54% ␣-helix, 24% ␤-sheet, 5% turn, and 15% unordered. DNA methyltransferases utilize a broad range of secondary structures, and our calculations fall within these values.
Kinetic Mechanism-The kinetic mechanism was analyzed by double-reciprocal analysis using hDAM methyltransferase (14). The 1/velocity versus 1/DNA concentration plot shows intersecting lines in the second quadrant, whereas the 1/velocity versus 1/AdoMet concentration plot shows lines intersecting on the y-axis (Fig. 1, A and B). A ping-pong mechanism is unlikely because the patterns in Fig. 1, A and B are intersecting, hence the ternary complex methyltransferase-DNA-AdoMet is indicated. Product inhibition studies with S-adenosylhomocysteine showed uncompetitive inhibition with DNA (Fig. 1E) and competitive inhibition with AdoMet (Fig. 1F). The simple competitive plot for AdoMet and AdoHcy at a constant DNA concentration rules out a random steady-state mechanism. The uncompetitive inhibition observed with AdoHcy and DNA at a constant AdoMet concentration also argues against random mechanisms; this pattern suggests that a binary methyltransferase-DNA complex can bind AdoHcy. Product inhibition studies with methylated DNA showed uncompetitive inhibition with DNA (Fig. 1C) and mixed type inhibition with AdoMet (Fig. 1D). These data are consistent with a steadystate ordered Bi Bi mechanism where the cofactor S-adenosylmethionine binds first. Additional evidence against a random rapid equilibrium mechanism and a DNA first-ordered mechanism is provided by the uncompetitive inhibition pattern observed with methylated DNA and DNA (Fig. 1C) because a competitive plot is expected in these cases. Although consistent with both an ordered AdoMet first mechanism (rapid equilibrium binding) and random steady-state mechanisms, such mechanisms are inconsistent with much of our other data. The reconciliation of the double-reciprocal analysis and product inhibition data forms the basis of the kinetic mechanism assignment. The k cat (0.93 Ϯ 0.06 min Ϫ1 ), K m DNA (17.4 Ϯ 3.0 nM), and K m AdoMet (5.6 Ϯ 1.1 M) were obtained using linear replots and SigmaPlot-calculated values. For all product inhibition analyses, data were fit to standard competitive, non-competitive, mixed type, and uncompetitive inhibition equations using Sig-maPlot (21). The fit with the highest R 2 value was selected as the best fit.
Determination of K d AdoMet -E. coli DAM has two tryptophans at positions 10 and 236, which provide a basis for probing ligand-induced conformational changes (6). AdoMet was added at 10 concentrations from 1 to 200 M in MRB, and the decrease in native protein fluorescence was plotted as a function of AdoMet concentration. Triplicate determination was used to calculate the thermodynamic dissociation constant and associated standard error, 18.7 Ϯ 1.81 (data not shown). The K d   (Table I), indicating that AdoMet is bound with comparable affinity in the binary complex and in the presence of DNA.

AdoMet
Determination of K d DNA -The affinity of the enzyme (DAM) for an unmethylated duplex (20-mer) was determined by gel mobility shift assay (Fig. 2) using 50 M sinefungin. The dissociation constant (K d DNA ), 119 Ϯ 7.3 nM (Table I), was determined from the scanned gel and resultant binding isotherm. Although similar to K m DNA (Table I), direct comparison is compromised by the use of sinefungin for the K d DNA and AdoMet for K m DNA determinations. Burst Experiments-Burst experiments provide insight about which steps limit catalytic turnover, the amount of active enzyme, and an apparent k cat determination (19,22). Under high enzyme and substrate (DNA and AdoMet) conditions, DAM manifests a burst of product formation, followed by a slower rate (Fig. 3A). Because the filter binding assay measures formation of methylated DNA, which includes product that is still associated with the enzyme, the simplest interpretation of these results is that methylation is faster than subsequent product release steps. The pre-steady-state phase (before 10 s) is followed by the steady-state phase, and the slope of the latter provides a measure of the apparent k cat . The slope of the higher enzyme condition (squares, 150 nM) is correspondingly greater than the lower enzyme condition (circles, 60 nM). Extrapolation of both plots to zero time provides an estimate of the active enzyme, which within experimental error is Ͼ95% active. Increasing the AdoMet concentration from the 30 M used in these experiments does not result in any changes in the product formation profiles (data not shown).
Isotope Partition-Our steady-state kinetic analysis suggests that the binary enzyme-AdoMet complex is catalytically competent. This can be directly tested through the use of an isotope-partitioning experiment (14). As in the burst experiment (Fig. 3A), preincubation of radiolabeled AdoMet and enzyme followed by the addition of DNA and more radiolabeled AdoMet in a chase results in the rapid formation of product. Adding a 25-fold excess of unlabeled AdoMet in the chase along with DNA results in a 3-fold decrease in the vertical axis intercept. A non-zero vertical intercept under these conditions requires the initial binary enzyme-AdoMet complex to be catalytically competent because a 25-fold reduction in AdoMetspecific activity results in a near zero level of detected product when the labeled and unlabeled AdoMet are premixed (data not shown). The 3-fold decrease in vertical intercept shows that upon DNA addition, the binary complex dissociates to a greater extent than proceeding to form the ternary complex and catalysis (14).
Processivity Analysis-The data in Fig. 4 show that DAM acted distributively on the DNA fragment derived from the Pap regulon, which contains two GATC sites separated by 103 bp (11,12,20). The gel shows the digestions (DpnII) which generated the singly methylated fragments (S1 and S2) and unmethylated fragments (U1 and U2). The percent conversion, which relates the singly (S1 and S2) and doubly (D) methylated band to the sum of all bands, is plotted. The calculated percent conversions are shown in the lower graph, and lines are shown for a hypothetical completely distributive enzyme (Fig. 4) (11,12). Similar results were obtained when NaCl was varied up to 100 mM, and AdoMet was used at 3 M (data not shown).

DISCUSSION
Although several bacterial DNA methyltransferases are well characterized both structurally and functionally (1), few have the potential medical importance of DAM and CcrM, shown to be essential for virulence and viability in several human pathogens (2, 4). Our aim was to provide a functional description of DAM to form the basis of identifying potent inhibitors. 2 We purified the recombinant enzyme from two expression vectors in an effort to obtain large quantities for both structural and inhibitor screening studies. The two forms, with (hDAM) and without (DAM) an N-terminal histidine tag, showed similar kinetic parameters (see "Results").
Our kinetic mechanism results are most consistent with a steady-state ordered mechanism involving an initial enzyme-AdoMet complex (Figs. 1 and 3; see "Results"). A limited number of DNA methyltransferases have been characterized at this level, but this overall kinetic mechanism is similar to that determined for the DNA adenine methyltransferases M.EcoRI (14), M.EcoRV (23), and T4DAM (24). Other DNA adenine methyltransferases appear to have either random or alternative ordered mechanisms (25,26). In contrast, the DNA cytosine methyltransferases M.HhaI (27,28), M.MspI (29), and Dnmt1 (30) either have obligate DNA first mechanisms or have a strong preference for this pathway. A recent study of DAM described pre-steady-state experiments in which the preformed enzyme-AdoMet complex leads to faster rates than the preformed enzyme-DNA complex (7). The authors interpret these results as evidence for a DNA first kinetic mechanism. Although such results have many interpretations, they provide little direct evidence for any order of substrate addition. Another recent study showed that at high concentrations, DAM can bind its cognate DNA in the absence of any cofactor or cofactor analogs, which can be reconciled (31) with our proposed kinetic mechanism because the initial velocity studies describe the kinetic preferences under much lower concentrations. Further evidence for an AdoMet first kinetic preference is provided by our isotope-partitioning results (Fig. 3B) which clearly show that the enzyme-AdoMet complex is competent; this result eliminates any mechanism in which sequence-specific DNA binding first is an obligate step. Interestingly, the binary complex is shown under the conditions of the assay to partition largely back to the free enzyme and AdoMet rather than lead to the ternary complex and catalysis. As described for M.EcoRI (14), an AdoMet first mechanism requires only that the cofactor is bound prior to sequence-specific DNA binding, not to the binding of nonspecific DNA adjacent to the target site. This has implications for the mechanism of overall site location, which has not been clearly identified for DAM but was shown to involve facilitated diffusion for other DNA methyltransferases (11,12).
Many of the structural and functional characteristics of DAM presented here are similar to those described for other DNA methyltransferases. Thus, the observed burst kinetics (Fig. 3A) show that a product release step after methyltransfer limits the catalytic turnover, as has been described for the majority of DNA methyltransferases (1). The affinity for DNA (Fig. 2) and inhibition by both AdoHcy and methylated DNA (Table I) of the enzyme are similar to those described for other DNA adenine methyltransferases. Because there is no high resolution structure for DAM, we used secondary structure predictions based on CD analysis to determine whether the structure of the protein fits that of previously characterized DNA methyltransferases and to confirm that the His-tagged enzyme is folded properly. Our secondary structure predictions for both DAM and hDAM are very similar, are within the range of observed values determined previously for DAM (32), and are from the inspection of the five related adenine methyltransferase crystal structures T4DAM, M.TaqI, M.DpnM, M.RsrI, and M.PvuII.
The proposed kinetic mechanism has implications for the design and interpretation of assays to identify DAM inhibitors. For example, despite the AdoMet first kinetic mechanism, compounds that interfere with AdoMet binding may leave the ability of the enzyme to bind DNA unperturbed. This characteristic could derive from the ability of the enzyme to bind DNA nonspecifically in the absence of its cofactor, which would not necessarily be altered if the AdoMet site were changed. Alternatively, compounds that interfere with DNA binding are likely to cause a significant reduction in the amount of enzyme-DNA complex formation. These two mechanistic outcomes are likely to result in very different cellular phenotypes even if the compounds demonstrate similar inhibition potency. In the former case, the inhibited enzyme will likely be bound nonspecifically to DNA; in the latter case, the inhibited enzyme will likely be free from DNA. Methyltransferase inhibitors that sequester the target enzyme onto the DNA are known to have significant nonspecific cellular toxicity (33).
A somewhat surprising result is our demonstration that DAM shows no evidence for processive catalysis in our assay on the DNA segment derived from the regulatory region of the Pap   FIG. 4. DAM is a distributive enzyme with the Pap regulatory DNA element. The intensity of D and S1ϩS2 were divided by the sum of DϩS1ϩS2ϩU1ϩU2 to determine the percent conversion of doubly and singly methylated species for a given time of reaction. D, doubly methylated fragment; S1, singly methylated (GATC I); S2, singly methylated (GATC II); U1, fragment generated from single cleavage (GATC II); U2, fragment generated from single cleavage (GATC I).
operon, which contains two GATC sites separated by 103 bp (Fig. 4) (20). Unlike the DNA substrates used in the prior studies of DAM processivity, this DNA element contains GATC sites with a methylation known to directly influence the expression of an adjacent operon (2). Although prior investigations of DAM processivity have provided conflicting interpretations, the demonstration that the enzyme can act processively (7) is compelling because many factors could interfere with this type of function under in vitro conditions. The prior work by Herman and Modrich (6) on DNA, in which two GATC sites are separated by 2455 bp, showed that the sites are methylated independently. Although not directly addressing processivity, the work by Bergerat et al. (8) using largely plasmid substrates provided indirect evidence that DAM is processive. The most definitive study on DAM processivity was provided by Urig et al. (7) using short synthetic DNA, PCR fragments, and DNA. These authors showed that DAM is extensively processive, with an average of 55 GATC sites being methylated for one enzyme binding event with large DNA. Trivial reconciliations of these results with our demonstrated lack of processivity can be excluded. Our experimental results did not change with variations in salt concentrations or variations in AdoMet concentrations, and a similar insensitivity to these parameters was described by Urig et al. (7). The lack of observed DAM processivity is not likely to be because of rapid dissociation of the protein off the proximal DNA ends, as Urig et al. (7) demonstrated processive character on short substrates as well.
We suggest that the different results might be understood in the context of sequences flanking the GATC recognition sites. Evidence for this was provided in the study on plasmid DNA methylation, in which a distinct ranking of preferred GATC sites was revealed (8). The 22 GATC sites within the pBR322 plasmid have diverse flanking sequences. A small subset is highly preferred by DAM and has the flanking sequences CGC-gatcATG, GAAgatcGGG, GATgatcGGC, CTTgatcCGG, and CATgatcCCC. Although no consensus site was identified, sites with three G/C pairs on the 3Ј side and two A/T pairs and one G/C pair on the other side are preferred. A number of sites were shown to be particularly disfavored in that no or little methylation was detected for the sites. These sites include TTCgat-cACT, GAAgatcCTT, AACgatcAAG, and TTTgatcTTT. The two sites found in the Pap fragment used for our processivity studies (GACgatcTTT and AAAgatcGTT) have features found in several poorly methylated sites (e.g. TTT on the 3Ј side, AAA on the 5Ј side) and lack the features seen in the preferred sites (e.g. 3Ј G/C pairs) (2,20). Thus, it seems plausible that DAM is not processive with particular GATC sites, as demonstrated here for the Pap-associated element, which derives from the particular sequence context of the two sites. A significant determinant of the processivity of an enzyme is the partition between remaining on its DNA after catalysis and leaving the DNA (7,11,12). One manner in which a poorly methylated site could modulate this partitioning is by increasing the time spent on a particular site prior to methylation. Thus, without altering the off-rate kinetics, a poorly methylated site would appear to be less processively modified than another site that is methylated rapidly. Alternatively, the off-rate kinetics from a particular site could be increased, resulting again in a site that shows minimal processivity. DAM has many biological functions, several of which require the enzyme to methylate the majority of genomic GATC sites (2,6,7). For example, although the exact number of the ϳ19,000 GATC sites in the E. coli genome that are unmethylated or partially methylated is unknown, various studies implicate 10 -50 such sites (7,34,35). The remaining sites are faithfully methylated during the normal cell division processes, demanding an efficient and nearly complete methylation process. As proposed by Urig et al. (7), a highly processive DAM would ensure that this large number of sites is efficiently methylated. However, other DNA methyltransferases with comparable site frequencies in the E. coli genome are either not processive, or certainly not as processive as DAM with the substrates used by Urig et al. (7). We propose that a processive DAM on the majority of sites, as described by Urig et al. (7), is compatible with our results when one considers the flanking sequence context, the different biological roles of the majority of genomic GATC sites, and the GATC sites known to be involved in gene regulation, such as those probed in our study within the Pap regulon (2,20).