Recognition of native DNA methylation by the PvuII restriction ...

© 2000 Oxford University Press

Nucleic Acids Research, 2000, Vol. 28, No. 16 3143–3150

Recognition of native DNA methylation by the PvuII restriction endonuclease Manda R. Rice and Robert M. Blumenthal* Department of Microbiology and Immunology, Medical College of Ohio, 3055 Arlington Avenue, Toledo, OH 43614-5806, USA Received March 27, 2000; Revised and Accepted June 22, 2000

ABSTRACT Recognizing the methylation status of specific DNA sequences is central to the function of many systems in eukaryotes and prokaryotes. Restriction–modification systems have to distinguish between ‘self’ and ‘non-self’ DNA and depend on the inability of restriction endonucleases to cleave their DNA substrates when the DNA is appropriately methylated. These endonucleases thus provide a model system for studying the recognition of DNA methylation by proteins. We have characterized the interaction of R·PvuII with DNA containing the physiologically relevant N4-methylcytosine modification. R·PvuII binds N4mC-modified DNA and cleaves it very slowly. Methylated strands in hemimethylated duplexes were cleaved at a higher rate than in fully methylated duplexes, in parallel with a higher binding affinity for hemimethylated DNA. The co-crystal structures of R·PvuII–DNA, together with a mutagenesis study, have implicated specific amino acids in recognition of the methylatable base; one of these is His84. We report that replacing His84 with Ala reduced the rate of cleavage of unmodified DNA but, in contrast, slightly increased the cleavage of N4mC-modified DNA. INTRODUCTION The interaction of DNA-binding proteins and methylated DNA is important to a wide variety of cellular processes. These include regulation of gene expression in both prokaryotes and eukaryotes (1,2), controlling the replication of bacterial chromosomes (3) and discrimination between parental and daughter strands of DNA during mismatch repair (4,5). Restriction–modification systems provide bacteria with a form of immunity against bacteriophage infection (6); the role of the modification methyltransferase in these systems is to protect the cell DNA from the restriction endonuclease. In type II restriction–modification systems the restriction endonuclease and DNA methyltransferase are separate, independently active proteins (7). Several restriction endonucleases have been structurally characterized and important insights have been gained regarding the mechanisms of DNA sequence

recognition and phosphodiester cleavage (8–13), however, important questions remain. One central feature of restriction endonucleases that is not well understood is how DNA methylation prevents cleavage. If this is not due entirely to effects on binding, then there must be signal propagation through the endonuclease; the methyl groups (whether in N4-methylC, 5-methylC or N6-methylA) are positioned near the center of the major groove while the cleavage reaction occurs at the sugar–phosphate backbone roughly 10 Å off to either side. The PvuII restriction modification system has been cloned and sequenced and both the DNA methyltransferase and the restriction endonuclease proteins have been studied (14–17). The DNA methyltransferase (M·PvuII) catalyzes the transfer of a methyl group from the donor S-adenosyl-L-methionine to the N4 position of the internal cytosine in the recognition sequence (CAGN4mCTG) (15,18). This modification inhibits DNA cleavage by the restriction endonuclease (R·PvuII). The PvuII restriction–modification system is unique in that it is the only one for which both the restriction endonuclease and the DNA methyltransferase have been crystallographically characterized (19–22). R·PvuII has been characterized structurally as an apoenzyme, in complex with unmodified DNA and in co-crystals with DNA containing 5-iodocytosine in place of the methylatable cytosine (20,23,24). Each monomer of the homodimeric protein is composed of three regions: a catalytic region, a subunit interface region and a DNA recognition region. The subunit interface and DNA recognition regions are structurally distinct from those of other characterized restriction endonucleases, while the catalytic region resembles those of other restriction endonucleases (25). R·PvuII approaches DNA from the minor groove side and wraps around the double helix, recognizing the base pairs primarily via the major groove, though an important minor groove recognition contact is made by D34 (26). A histidine triplet (H83–H85) located in the DNA recognition region is positioned to play a major role in the recognition of methylated bases: in the R·PvuII–DNA co-crystal, the Cβ of H84 is just 4.3 Å from N4 of the methylatable cytosine, and when this cytosine is replaced by 5-iodocytosine (a structural analog of 5-methylcytosine) His84 adopts an altered conformation (20). The role of specific amino acids in binding and catalysis was also studied by site-directed mutagenesis (27). R·PvuII mutants were created in which His83, His84 and His85 were individually replaced by alanine. These mutants were characterized in crude extracts for their ability to cleave

*To whom correspondence should be addressed. Tel: +1 419 383 5422; Fax: +1 419 383 3002; Email: [email protected]

3144 Nucleic Acids Research, 2000, Vol. 28, No. 16 bacteriophage λ DNA and for their ability to bind DNA fragments in gel mobility shift assays. Two of these mutants, H84A and H85A, are analyzed in this study in order to better understand the interaction of R·PvuII and methylated DNA. Recently N4-methylcytosine (N4mC) phosphoramidite became commercially available, allowing the production of synthetic DNAs in which full methylation of the recognition site could be assured and supporting in vitro studies. We thus analyzed the interaction of R·PvuII with its physiologically relevant methylation. We report here that the H84A substitution alters the effects of native DNA methylation on R·PvuII activity. These results are consistent with the structural evidence, suggesting that His84 plays a key role in detecting DNA methylation. MATERIALS AND METHODS

The modified oligonucleotides contained N4mC in place of the internal cytosine in the single PvuII site. The upper and lower strands of each duplex DNA were radiolabeled separately using T4 polynucleotide kinase (Life Technologies) and [γ-32P]ATP (6000 Ci/mmol; Amersham Pharmacia Biotech, Piscataway, NJ). Unincorporated nucleotides were removed using QiaQuick columns (Qiagen, Valencia, CA). The concentration of the single-stranded radiolabeled DNA was determined from the A260 nm and equimolar amounts of each strand were annealed by heating to 90°C and slowly cooling. The specific radioactivity was determined as previously described (28). To fill-in the 5'-overhang, radiolabeled duplexes were treated with the Klenow fragment of DNA polymerase (New England Biolabs) in the presence of all four nucleoside triphosphates (33 µM each). The efficiency of this reaction was assessed by electrophoresis on a sequencing gel.

Plasmid vector and strains

DNA cleavage assays

All plasmids used in the experiments are derivatives of the vector pMal (New England Biolabs, Beverly, MA). When expressed as a fusion to maltose binding protein (MBP), proteins were purified on an amylose affinity column. Escherichia coli strains used included K802 (WA802) and K802 (λCAC1.3A), both obtained from Dr William Jack at New Engand Biolabs. The λ prophage in the latter strain carries pvuIIM, protecting the DNA so that active endonuclease genes can be introduced.

To assay wild-type R·PvuII, radiolabeled oligonucleotides were used in NEB buffer 2 (50 mM NaCl, 10 mM Tris–HCl, 10 mM MgCl2, 1 mM dithiothreitol, pH 7.9), using commercial preparations of R·PvuII (50 U/µl; New England Biolabs). The R·PvuII was diluted immediately before use in NEB diluent A (50 mM KCl, 10 mM Tris–HCl, 0.1 mM EDTA, 1 mM dithiothreitol, 200 µg/ml bovine serum albumin, 50% v/v glycerol, pH 7.4). All reactions contained 10 nM duplex radiolabeled DNA in a total volume of 50 µl and a final glycerol concentration of 5% (v/v). Assays using unmodified DNA included 20 pM (0.15 U) R·PvuII dimer, while reactions with methylated substrates contained 170 pM (1.25 U) R·PvuII dimer. Reaction components were combined and preincubated at 37°C for 2 min and reactions were initiated by adding R·PvuII. Samples were withdrawn, added to Circumvent Stop/ Loading Dye (deionized formamide containing 0.3% xylene cyanole FF, 0.3% bromophenol blue, 0.37% EDTA, pH 7.0; New England Biolabs) and placed in a dry ice/ethanol bath. Samples were resolved on a 20% sequencing gel cast in glycerol-tolerant buffer (31) and the dried gel was analyzed with a PhosphorImager as described above. The activities of R·PvuII–MBP fusion proteins were determined using two substrates: λ DNA and a synthetic oligonucleotide (described above). Oligonucleotide-based assays were carried out as described above. Lambda assays contained 1 µg λ DNA (Gibco BRL) and NEB buffer 2 in a total volume of 50 µl. After 1 h at 37°C, Circumvent Stop/Loading Dye was added and the samples were resolved on a 1% agarose gel in Tris–acetate buffer. The effects of crude extract on the activity of the enzymes was tested with a centrifuged sonicate of the R·PvuII-nonproducing E.coli strain JM107MA2 (28). This extract was added to an equal volume of R·PvuII H84A, incubated at room temperature for 2 h and assayed on λ DNA as described above. The effect of cleaving MBP from the R·PvuII fusion was tested by treating the H84A variant with Genenase I (New England Biolabs), monitoring the efficiency of cleavage by SDS– PAGE. As another control, purified MBP (New England Biolabs) was added to a reaction containing commercial wildtype R·PvuII. Competition cleavage assays were performed to estimate the relative affinities of R·PvuII for unmodified and N4mC-modified

R·PvuII–MBP fusions Plasmids containing pvuIIR, pvuIIRH84A and pvuIIRH85A (27) were provided by Dr Paul Riggs (New England Biolabs). The pvuIIRH85A clone was provided as a pMal construct. We subcloned the wild-type and H84A mutant genes into the pMal vector with BglII and XbaI, ligating the resulting fragments into the corresponding sites in a pMal construct containing pvuIIR interrupted with a stuffer fragment (provided by Dr Riggs). The ligations were used to transform E.coli strain K802 (λCAC1.3A) (28) and plated onto LB agar containing carbenicillin (120 µg/ml) and kanamycin (10 µg/ml). Plasmid minipreps were screened for loss of the stuffer fragment and potential clones were screened by transforming a strain that lacks pvuIIM; plasmids producing active endonuclease yield very few transformants in the absence of a protecting methylase. Sequences were confirmed by the dideoxy chain termination method (29) (Sequitherm Excel Kit; Epicentre Technologies, Madison, WI). Expression and purification of the pMal fusion clones from 1 l cultures was carried out according to the manufacturer’s recommendations (New England Biolabs). Column fractions were collected and analyzed by A280 and SDS–PAGE electrophoresis. The concentration of pooled fractions was determined using a Bradford assay (30) (Life Technologies, Rockville, MD). The yield was 5–15 mg protein/l culture and the resulting protein gave a single band on analysis by SDS– PAGE. Synthetic DNA substrates PAGE-purified oligonucleotides were purchased from New England Biolabs and were designed to yield duplexes with one blunt end and the other having a five base 5'-extension (Fig. 1).

Nucleic Acids Research, 2000, Vol. 28, No. 16 3145

Figure 1. Synthetic oligonucleotide substrates contained either (A) a single PvuII site (CAGCTG), (B) a PvuII site in which the central cytosine was substituted with N4-methylcytosine (CAGMTG) or (C) a reversed PvuII site for use as a control oligonucleotide (GTCGAC).

oligonucleotides. The rate of cleavage of 10 nM radiolabeled Pvu40R (unmethylated) was measured as described above, in the presence of varying amounts of competitor DNAs. Competitor and radiolabeled DNAs were mixed prior to addition of R·PvuII. RESULTS DNA containing N4-methylcytosine on both strands is cleaved very slowly The protective methyltransferase of the PvuII restriction– modification system converts CAGCTG sequences to CAGN4mCTG (15,32). In the absence of this methylation, significant expression of pvuIIR kills the cell while even tiny levels of expression rapidly lead to the accumulation of pvuIIR mutants (33). In simple assays R·PvuII does not appear to cleave the methylated DNA; however, this has not been carefully examined and cleavage of N4mC-hemimethylated DNA, a normal product of DNA replication, has not been studied at all. We carried out cleavage assays with an excess of R·PvuII for up to 4 h using the substrates shown in Figure 1. These substrates contained N4mC on neither, one or both strands. Each strand’s labeled cleavage products had distinct lengths and were resolved on sequencing gels and quantitated using a PhosphorImager. As a control, the substrates were digested with AluI restriction endonuclease (R·AluI; not shown), which recognizes AGCT, the central four bases of the PvuII site. R·AluI does not cleave N4mC-modified DNA (18), although the activity of R·AluI on N4mC-hemimethylated DNA has not been investigated previously. The control reactions using R·AluI showed no cleavage, even with hemimethylated substrates. In addition to revealing the N4-hemimethylation sensitivity of R·AluI, this result indicates that the methylated oligonucleotide substrates did not contain a detectable amount of contaminating unmodified DNA. Surprisingly, the DNA containing N4mC on both strands was cleaved by R·PvuII, though as one would expect this cleavage was extremely slow. Figure 2 shows the progress of the reaction using completely unmethylated (Fig. 2A and B) or fully methylated (Fig. 2C and D) DNA substrates. Note the time scale difference as well as the fact that the unmodified DNA was digested with a substantially lower amount of R·PvuII. As discussed below, the digestion of DNA with N4mC on both strands may represent nicking (cleavage of only one strand per

duplex) but nevertheless indicates cleavage activity on fully methylated DNA. Less surprisingly, the two hemimethylated DNAs were also cleaved by R·PvuII (Fig. 3). In both cases the cleavage rate was ∼1000-fold lower than that of the unmodified DNA, though substantially higher than that of the fully modified duplex. It was important to determine if only the unmodified strand was being cut. As shown in the lower halves of Figure 3A and C, both strands are cleaved by R·PvuII. Interestingly, when only the lower strand is methylated the upper strand is preferentially cleaved; while when only the upper strand is methylated the two strands are cleaved at comparable rates. The upper and lower strands of the substrates differ in length by 5 nt and that difference also might contribute to the distinct cleavage rates of the two strands in hemimethylated substrates. To test this, the radiolabeled duplex oligonucleotides were treated with the Klenow fragment of DNA polymerase I to make both strands equal in length (while not changing the length disparity between the two 5'-labeled cleavage products). Filling in the 5'-overhang had significant effects on the cleavage of only the native (unmethylated) DNA, with the filled DNA being cleaved at a higher initial rate and the progress curves showing a more hyperbolic character (Fig. 2A and B). Filling in the 5'-overhang also improved cleavage of the fully methylated substrate, though comparison is difficult because cleavage of the unfilled DNA was barely detectable (Fig. 2C). Methylation status of DNA affects R·PvuII binding affinity Methylation of the DNA substrate could affect R·PvuII binding, catalysis or both. Binding studies carried out under non-catalytic conditions can be difficult to interpret, because the presence of divalent cations can profoundly affect binding specificity (34,35). Accordingly, we used competition cleavage assays to study the effects of N4mC on R·PvuII binding. Radiolabeled unmodified oligonucleotide duplex was used as the substrate (Fig. 1A) and increasing amounts of unlabeled oligonucleotide competitors were added. DNAs tested as competitors included the doubly N4mC-modified duplex, the two hemimethylated duplexes and a non-site oligonucleotide (Fig. 1C). Unlabeled unmodified oligo was used as a control. Figure 4 shows representative progress curves from the competition assays; the results with the doubly methylated competitor are shown. The reaction rates from plots such as Figure 4 were compared to the rate in the absence of competitor to calculate percent inhibition, which was then plotted against the concentration of competitor. The concentrations of various competitors needed to inhibit cleavage by 50% (IC50, which should approximate the Kd) are shown in Table 1. R·PvuII recognizes N4mC-modified CAGCTG sites, as they yield an IC50 2- (fully methylated) to 4-fold (hemimethylated) lower than non-site DNA. The IC50 values for the hemimethylated DNAs were about half that of the fully modified DNA, consistent with differences in their relative rates of cleavage. Analysis of H84A and H85A mutants in vivo and in vitro To explore the role of H84 and H85 in recognition of the methylatable cytosine, we employed two His→Ala mutants generated by others (27). The H84A mutant was reported to bind DNA in gel mobility shift assays and in an in vivo binding assay and to have cleavage activity in crude extracts ∼100-fold

3146 Nucleic Acids Research, 2000, Vol. 28, No. 16

Figure 2. Cleavage of unmethylated or fully methylated oligonucleotide duplexes by wild-type R·PvuII. 5'-End-labeled duplex oligonucleotides (0.5 pmol, 10 nM) were digested with R·PvuII (0.15 U, ∼1 fmol, 20 pM, or 0.02 U for the unmodified substrate) and electrophoresed through a 20% acrylamide sequencing gel. The upper and lower strands (whether or not filled-in with Klenow polymerase) yield 22 and 18 nt labeled cleavage products, respectively (see Fig. 1). (A) Surface plot (generated by NIH Image) of the digestion time course for unmethylated DNA; peak height is proportional to phosphorimager signal strength. (B) Fraction of unmethylated DNA that had been cleaved. Filled symbols represent data from Klenow-filled duplexes, while open symbols represent data from unfilled duplexes. Squares represent upper strands and circles represent lower strands. (C) Surface plot of the digestion time course for DNA having N4mC on both strands. Only the region containing the cleavage products is shown. For comparison, cleaved unmethylated DNA was run in adjacent lanes. (D) Fraction of fully methylated DNA that had been cleaved. Only the Klenow-filled substrate was cleaved sufficiently to allow reliable quantitation and even then only as the summed signal from both strands. The dotted line is from linear regression of the data, not including the earliest point.

lower than wild-type enzyme, but the activity was still sufficient to greatly reduce transformation efficiency of the plasmid carrying pvuIIRH84A in the absence of the PvuII methylase. The H85A mutant was also found to bind DNA specifically, but had such low activity that it could be expressed in the absence of protective methylase (27). The in vivo activity of MBP fusions of H84A and H85A was tested by transforming two isogenic strains, one carrying pvuIIM on a lysogenic λ phage (which should allow transformation with an active endonuclease gene) and the other without the PvuII methylase gene. As a control, we used a MBP fusion to wild-type pvuIIR. Transformation with the H84A fusion clone was extremely inefficient in the absence of a protecting methylase, showing the same pattern as wild-type enzyme (Table 2) and confirming that the MBP fusion is

active. The H85A fusion clone, as expected, transformed both pvuIIM+ and pvuIIM– strains with equal efficiency. Overexpression of each fusion protein by treatment with IPTG was toxic to the strains regardless of methylase status (not shown). We carried out several control experiments to facilitate comparison between our studies of the H84A variant of R·PvuII and those of Nastri et al. (27). Crude extract, from a restrictionless strain of E.coli, had no effect on cleavage rates nor did MBP, whether present in fused or free form (not shown). Accordingly we used MBP-fused R·PvuII in the remaining experiments. We next examined the cleavage patterns generated by the fusion proteins, using DNA from bacteriophage λ. The wildtype and H84A enzymes showed similar levels of activity


Figure 3. Cleavage of hemimethylated oligonucleotide duplexes by wild-type R·PvuII. 5'-End-labeled unfilled duplex oligonucleotides (0.5 pmol, 10 nM) were digested with R·PvuII (0.15 U, ∼1 fmol, 20 pM) and electrophoresed through a 20% acrylamide sequencing gel. The upper and lower strands yield 22 and 18 nt labeled cleavage products, respectively (see Fig. 1). (A) Surface plot (generated by NIH Image) of the digestion time course for DNA containing N4mC in the lower strand only; peak height is proportional to phosphorimager signal strength. (B) Fraction of hemimethylated DNA (lower strand methylated) that had been cleaved. (C) Surface plot of the digestion time course for DNA having N4mC on the upper strand only. (D) Fraction of hemimethylated DNA (upper strand methylated) that had been cleaved.

(differing ∼8-fold); however, the pattern of digestion was subtly different for the two enzymes (not shown). The final fragment patterns were indistinguishable, so the differences appear to reflect different relative preferences for the 14 PvuII sites present in λ DNA. Although the relatively low activity of H85A makes analysis of site preference difficult, the pattern of digestion observed was apparently the same as that generated by wild-type enzyme (not shown). We also analyzed the digestion of oligonucleotides by these fusion proteins, using the unmodified oligonucleotide (Fig. 1A). Figure 5A shows the progress curves for the wild-type and H84A fusion proteins. The rate of cleavage by H84A– MBP was at least 5-fold lower than that by the WT–MBP protein and H85A–MBP exhibited a further 1000-fold reduction in activity (not shown). As noted above, the H85A

mutation so severely impairs activity that the clone can be maintained in the absence of a protecting methylase, and its extremely low activity precluded further characterization. Effect of H84A on the cleavage of methylated DNA We studied the abilities of the H84A and wild-type enzymes to recognize and cleave oligonucleotide substrates containing N4mC in both strands. Cleavage products were analyzed as described above. While both enzymes cleaved methylated DNA at a substantially lower rate than unmodified DNA, the H84A mutant cleaved N4mC-modified DNA ∼50% faster than did the wild-type enzyme (Fig. 5B and Table 3). This difference, while small, is striking in comparison to the much higher activity of wild-type than H84A enzyme on unmodified DNA.


Table 1. Competition cleavage assays with R·PvuII IC50a (nM)

Competitor DNA C/Cb

10

N4mC/C

152

C/N4mC

183

N4mC/N4mC

354

Non-site duplex

658

aConcentrations of competitor DNA required to achieve 50% inhibition of the initial reaction rate (IC50) were determined from plotting the initial rates versus the concentration of competitor DNA, using data such as that shown in Figure 4. bAn aliquot of 0.5 pmol (10 nM) radiolabeled substrate DNA was used in all cases, together with 0.15 U (∼1 fmol or 20 pM) R·PvuII. Target and competitor DNAs were premixed and reactions initiated with R·PvuII.

Table 2. In vivo activity of various R·PvuII alleles pvuIIR plasmida

Host

0.001 µg

0.01 µg

pMal R·PvuII (WT)

M–

0

0

M+

2739

6943

M–

0

0

M+

427

M–

7

M+

270

pMal H84A

pMal H85A

Concentration 0.1 µg

1 µg

1

0

tntc

tntc

0

1

5478

tntc

tntc

159

1911

tntc

3886

tntc

tntc

Figure 4. Competition cleavage assay to measure the relative affinity of R·PvuII for various modified substrates under catalytic conditions. Unmodified radiolabeled DNA was cleaved by R·PvuII in the presence of increasing concentrations of unlabeled competitor DNA. Data shown is from competition with N4mC/N4mC substrates and is representative of all data sets. The lines shown result from linear regression and r values were 0.98 and above. The following concentrations of competitor DNA were used: no competitor (open circle); 20 nM (closed circle); 200 nM (square); 1000 nM (diamond); 2000 nM (triangle). These correspond to the fold molar excesses over labeled substrate that are indicated in the figure.

aVarious microgram amounts of three R·PvuII plasmids were used to transform strains carrying pvuIIM (M+) or lacking pvuIIM (M–). The more active the endonuclease allele, the fewer transformants will result in the absence of pvuIIM. Numbers indicate the number of carbenicillin-resistant transformants. tntc, too numerous to count.

Table 3. H84A versus wild-type R·PvuII initial cleavage ratesa Form of R·PvuII

Unmodifiedb (C/C)

Methylatedb (N4mC/N4mC)

Wild-type

≥4.3 × 10–1 (1.0)

6.4 × 10–5 (0.00015)

H84A

4.3 ×

10–2

(0.1)

1.0 × 10–4 (0.00023)

aInitial

rates of cleavage were extrapolated from the fits to the data shown in Figure 5. Units are pmol DNA cleaved per s. Normalized values are shown in parentheses. bThe duplexes shown in Figure 1; unmodified in Figure 1A and fully methylated in Figure 1B.

DISCUSSION R·PvuII cleaves CAGCTG sites that contain N4-methylcytosine Restriction–modification systems can provide protection from bacteriophage infection, but the potentially lethal restriction endonuclease must be strictly controlled to prevent damage to the cell’s own DNA. There are many mechanisms known to control restriction activity, including transcriptional and posttranscriptional regulation (36–38). Methylation of cellular

Figure 5. Comparison of the cleavage of unmodified (A) and N4-methylcytosine-modified oligonucleotides (B) by wild-type R·PvuII (open circles) and H84A (closed circles). The MBP fusion proteins were used to digest radiolabeled oligonucleotides. The proportion of DNA cleaved was calculated from phosphorimager analysis of acrylamide sequencing gels. The data are fitted to an inverse exponential function; r values were >0.98.


DNA that blocks cleavage by a restriction endonuclease is a hallmark of restriction–modification systems. When R·EcoRI is expressed in the absence of protective methylation, E.coli survival depends on DNA ligase (39). This ligase is not efficient, however, at repairing blunt end doublestrand cleavages such as those catalyzed by R·PvuII; while the ligase could repair nicks, nicks may lead to collapse of replication forks that can only be repaired by recombination (40). Earlier studies from this laboratory have shown that very low level expression of wild-type R·PvuII is tolerated by E.coli in the absence of protective methylation, even in a recA background (17). Nevertheless, the pvuIIR genes in such cells rapidly accumulate mutations. It is therefore surprising that R·PvuII cleavage of N4mCmethylated DNA, while very slow, is detectable (under the conditions used). Control reactions with R·AluI showed that the observed cleavage was not due to the presence of contaminating unmodified DNA. It is important to note that differences between in vivo and in vitro conditions might affect the in vivo consequences of a small amount of cleavage. These differences include the relative concentrations of endonuclease and DNA, the huge excess of competitor (non-site) chromosomal DNA, ionic conditions, DNA-binding proteins such as HU and the presence of various repair enzymes. The cleavage of properly methylated DNA may illustrate the principle that selection yields adequacy, not necessarily perfection. Effects of N4-methylcytosine on R·PvuII Competition assays were performed in solution under catalytic conditions in order to study the relative binding affinities of R·PvuII for unmodified and N4mC-methylated oligonucleotides. The affinity of R·PvuII for fully N4mC-methylated DNA was 35-fold lower than for unmodified DNA, though still ~2-fold greater than the affinity for non-specific DNA (Table 1). Nevertheless, these results indicate that the great majority of the reduction in R·PvuII cleavage caused by N4mC is due to reduced catalysis or some other post-binding step. Our results are similar in several respects to those from the EcoRI restriction–modification system. R·EcoRI recognizes the sequence GAATTC and methylation of the second adenine blocks cleavage. The presence of N6-methyladenine on one strand of the DNA had a modest effect on binding, but a profound effect on the overall cleavage rate, reducing it by ~1800-fold (41). Cleavage of both strands was reduced, but cleavage of the methylated strand in the hemimethylated duplex was much greater than that of either methylated strand in the doubly modified DNA. Similarly, the EcoK methyltransferase prefers hemimethylated to unmethylated substrates and has no activity on fully methylated sites, yet this enzyme has similar binding affinities for all three types of sites (42). R·PvuII is also known to be sensitive to flanking sequences, as illustrated by its poor cleavage of CAGCTGATC if the underlined A is methylated (43). H84A has altered cleavage rates and patterns on unmodified DNAs We next studied the role of specific amino acids implicated by structural analyses in binding and cleavage of substrate DNA. The R·PvuII–DNA co-crystal with unmodified DNA suggests that addition of a methyl group to the N4 position of the internal cytosine interferes with a contact made between Cβ of

His84 and the cytosine N4 (19). N4C methylation was suggested to keep Cβ farther from cytosine N4, thus interfering with the hydrogen bonding that occurs between Nδ1 of His84 and O6 of the paired guanine. This suggestion was strengthened by analysis of co-crystals of R·PvuII and DNA containing 5-iodocytosine. Even though His84 is 6.0 Å from the cytosine C5, compared to 4.3 Å from N4, the 5-iodo substitution still caused a significant conformational shift in His84 (20). The van der Waals radius of iodine (2.15 Å) is close to that of CH3 (2.0 Å) (44). The effects reported here of H84A substitution on discrimination between unmethylated and methylated DNA support the predictions made from the protein–DNA co-crystal structures. Our results, consistent with those of Nastri et al. (27), indicate that R·PvuII H84A has sufficient cleavage activity to cause lethal autorestriction in the absence of a protective methylase. Our in vitro studies indicate that the H84A alteration reduced the rate of cleavage of unmodified oligonucleotide substrates of at least 10-fold relative to the wild-type enzyme (Fig. 5 and Table 3). Nastri et al. saw a 100-fold reduction in the in vitro cleavage activity in crude extracts (27). Other observed effects of the H84A substitution included a subtle change in the relative site preference for λ DNA PvuII sites. Since the normal final fragment pattern was eventually generated, there was no cleavage of sites other than the cognate CAGCTG sites. Thus a side chain involved in recognizing the central base pairs of the PvuII site appears to affect preferences for the flanking DNA sequence. H84A cleaves N4C-methylated oligonucleotides at a higher rate than does wild-type R·PvuII Our finding that the H84A substitution reduces the rate of cleavage of unmodified DNA is consistent with His84 playing a role in DNA recognition, but does not address the recognition of methylation status. In contrast to its reduced cleavage of unmodified DNA, R·PvuII H84A cleaves N4C-methylated oligonucleotides slightly faster than does wild-type R·PvuII (Fig. 5 and Table 3). The structural data suggest that a steric clash between the H84 side chain in the wild-type enzyme and the methyl group at the N4 position of cytosine contributes to the blockage of cleavage. The smaller size of the alanine side chain may allow a better fit of this part of the enzyme to methylated DNA, with a correspondingly closer approach of the catalytic side chains to the scissile bond. Also, hydrophobic interactions between the alanine and the N4C methyl group may contribute to binding and or cleavage. SUMMARY We have studied the interaction of R·PvuII with DNA substrates containing N4mC under catalytic conditions, in vivo and in vitro. R·PvuII binds to and cleaves PvuII sites containing N4mC. While this cleavage occurs at a very low but measurable rate, the binding is reduced by full methylation by a much smaller factor. The interaction of R·PvuII and N4mC resembles the interaction of R·EcoRI and N6-methyladenine (41,45) in that methylation affects binding much less than it affects cleavage and in that the methylated strand in a hemimethylated duplex is cleaved at a substantially higher rate than either strand in a fully methylated duplex. Replacing His84 of R·PvuII with Ala reduces the cleavage rate on unmodified DNA but slightly enhances cleavage of N4C-methylated DNA,


consistent with the predicted interaction between His84 and the methylatable cytosine. While H84A has only lost a fraction of the discrimination shown by wild-type R·PvuII against fully methylated sites, this may reflect the complex and relatively inflexible interface between restriction endonucleases and their DNA substrates (46). ACKNOWLEDGEMENTS The authors would like to thank Drs Paul Riggs and Horacio Nastri (New England Biolabs) for providing several pvuIIR clones, including the original H84A mutant, and Drs Ashok S. Bhagwat and Joan C. Dunbar (Wayne State University), J. David Dignam (Medical College of Ohio) and Xiaodong Cheng (Emory University) for critically reviewing the manuscript. We also thank Dr Viktoras Butkus (Fermentas Inc.) for supplying a phosphoramidite of N4-methylcytosine that was used in pilot studies. This work is supported by National Science Foundation grant no. MCB-9631137. REFERENCES 1. Zyskind,J.W. and Smith,D.W. (1992) Cell, 69, 5–8. 2. Plumbridge,J. (1987) Biochimie, 69, 439–443. 3. Reisenauer,A., Kahng,L.S., McCollum,S. and Shapiro,L. (1999) J. Bacteriol., 181, 5135–5139. 4. Mol,C.D., Parikh,S.S., Putnam,C.D., Lo,T.P. and Tainer,J.A. (1999) Annu. Rev. Biophys. Biomol. Struct., 28, 101–128. 5. Kolodner,R.D. and Marsischky,G.T. (1999) Curr. Opin. Genet. Dev., 9, 89–96. 6. Bickle,T.A. (1987) In Neidhardt,F.C. (ed.), DNA Restriction and Modification Systems. Escherichia coli and Salmonella typhimurium, 1. American Society for Microbiology, Washington, DC. 7. Pingoud,A. and Jeltsch,A. (1997) Eur. J. Biochem., 246, 1–22. 8. McClarin,J.A., Frederick,C.A., Wang,B.C., Greene,P., Boyer,H.W., Grable,J. and Rosenberg,J.M. (1986) Science, 234, 1526–1541. 9. Kim,Y., Grable,J.C., Love,R., Greene,P.J. and Rosenberg,J.M. (1990) Science, 249, 1307–1309. 10. Winkler,F.K., Banner,D.W., Oefner,C., Tsernoglou,D., Brown,R.S., Heathman,S.P., Bryan,R.K., Martin,P.D., Petratos,K. and Wilson,K.S. (1993) EMBO J., 12, 1781–1795. 11. Newman,M., Lunnen,K., Wilson,G., Greci,J., Schildkraut,I. and Phillips,S.E.V. (1998) EMBO J., 17, 5466–5476. 12. Newman,M., Strzelecka,T., Dorner,L.F., Schildkraut,I. and Aggarwal,A.K. (1994) Nature, 368, 660–664. 13. Huai,Q., Colandene,J.D., Chen,Y., Luo,F., Zhao,Y., Topal,M.D. and Ke,H. (2000) EMBO J., 19, 3110–3118. 14. Blumenthal,R.M. (1987) In Chirikjian,J.G. (ed.), The PvuII Restriction– Modification System: Cloning, Characterization and use in Revealing in E. coli Barrier to Certain Methylases or Methylated DNAs. Gene Amplification and Analysis, 5. Elsevier North Holland, New York, NY. 15. Blumenthal,R.M., Gregory,S.A. and Cooperider,J.S. (1985) J. Bacteriol., 164, 501–509.

16. Athanasiadis,A., Gregoriu,M., Thanos,D., Kokkinidis,M. and Papamatheakis,J. (1990) Nucleic Acids Res., 18, 6434. 17. Tao,T. and Blumenthal,R.M. (1992) J. Bacteriol., 174, 3395–3398. 18. Butkus,V., Klimasauskas,S., Petrauskiene,L., Maneliene,Z., Lebionka,A. and Janulaitis,A.A. (1987) Biochim. Biophys. Acta, 909, 201–207. 19. Cheng,X., Balendiran,K., Schildkraut,I. and Anderson,J.E. (1994) EMBO J., 13, 3927–3935. 20. Horton,J.R., Bonventre,J. and Cheng,X. (1998) Biol. Chem., 379, 451– 458. 21. O’Gara,M., Adams,G.M., Gong,W., Kobayashi,R., Blumenthal,R.M. and Cheng,X. (1997) Eur. J. Biochem., 247, 1009–1018. 22. Gong,W., O’Gara,M., Blumenthal,R.M. and Cheng,X. (1997) Nucleic Acids Res., 25, 2702–2715. 23. Balendiran,K., Bonventre,J., Knott,R., Jack,W., Benner,J., Schildkraut,I. and Anderson,J.E. (1994) Proteins, 19, 77–79. 24. Cheng,X., Balendiran,K., Schildkraut,I. and Anderson,J.E. (1995) Gene, 157, 139–140. 25. Kovall,R.A. and Matthews,B.W. (1999) Curr. Opin. Chem. Biol., 3, 578– 583. 26. Horton,J.R., Nastri,H.G., Riggs,P.D. and Cheng,X. (1998) J. Mol. Biol., 284, 1491–1504. 27. Nastri,H.G., Evans,P.D., Walker,I.H. and Riggs,P.D. (1997) J. Biol. Chem., 272, 25761–25767. 28. Sambrook,J., Frisch,E.F. and Maniatis,T. (1989) Molecular Cloning: A Laboratory Manual, 2nd Edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. 29. Sanger,F., Nicklen,S. and Coulson,A.R. (1992) Biotechnology, 24, 104– 108. 30. Bradford,M.M. (1976) Anal. Biochem., 72, 248–254. 31. Pisa-Williamson,D. and Fuller,C.W. (1992) Comments 19. US Biochemical, Cleveland, OH. 32. Butkus,V., Petrauskiene,L., Maneliene,Z., Klimasauskas,S., Laucys,V. and Janulaitis,A.A. (1987) Nucleic Acids Res., 15, 7091–7102. 33. Tao,T. (1992) Doctoral dissertation. Department of Microbiology and Immunology, Medical College of Ohio, Toledo, OH. 34. Rice,M.R., Koons,M.D. and Blumenthal,R.M. (1999) Nucleic Acids Res., 27, 1032–1038. 35. Erskine,S.G. and Halford,S.E. (1998) J. Mol. Biol., 275, 759–772. 36. Sohail,A., Ives,C.L. and Brooks,J.E. (1995) Gene, 157, 227–228. 37. Adams,G.M. and Blumenthal,R.M. (1995) Gene, 157, 193–199. 38. Tao,T., Bourne,J.C. and Blumenthal,R.M. (1991) J. Bacteriol., 173, 1367–1375. 39. Heitman,J., Zinder,N.D. and Model,P. (1989) Proc. Natl Acad. Sci. USA, 86, 2281–2285. 40. Heitman,J., Ivanenko,T. and Kiss,A. (1999) Mol. Microbiol., 33, 1141– 1151. 41. Jen-Jacobson,L., Engler,L.E., Lesser,D.R., Kurpiewski,M.R., Yee,C. and McVerry,B. (1996) EMBO J., 15, 2870–2882. 42. Powell,L.M., Dryden,D.T.F., Willcock,D.F., Pain,R.H. and Murray,N.E. (1993) J. Mol. Biol., 234, 60–71. 43. Chen,D., Liu,Q., Chen,X., Zhao,X. and Chen,Y. (1991) Nucleic Acids Res., 19, 5703–5705. 44. Pauling,L. (1970) Chem. Ber., 6, 468–472. 45. Engler,L.E., Welch,K.K. and Jen-Jacobson,L. (1997) J. Mol. Biol., 269, 82–101. 46. Lukacs,C.M., Kucera,R., Schildkraut,I. and Aggarwal,A.K. (2000) Nature Struct. Biol., 7, 134–140.