Proc. Natl. Acad. Sci. USA Vol. 80, pp. 4639-4643, August 1983 Biochemistry

Methyl-directed repair of DNA base-pair mismatches in vitro (mutagenesis/gene conversion/DNA methylation) A.-LIEN Lu, SUSANNA CLARK, AND PAUL MODRICH Department of Biochemistry, Duke University Medical Center, Durham, North Carolina 27710 Communicated by Robert L. Hill, April 18, 1983

ABSTRACT An assay has been developed that permits anal- system requires not only detection of base-pair mismatches but ysis of DNA mismatch repair in cell-free extracts of Escherichia a mechanism for discrimination of parental and newly synthe- coli The method relies on repair of heteroduplex molecules of fl sized strands as well. These authors suggested that the transient R229 DNA, which contain a base-pair mismatch within the single undermethylation of the newly synthesized strand might pro- EcoRI site of the molecule. As observed with mismatch hetero- vide the bias for such discrimination. Indeed, several lines of duplexes of A DNA [Pukila, P. J., Peterson, J., Herman, G., evidence indicate that dam methylation of d(G-A-T-C) se- Modrich, P. & Meselson, M. (1983) Genetics, in press], in vivo mis- quences functions in this respect. Thus, deficiency or over- match correction of fl heteroduplexes is directed by the state of production of this DNA methylase results in a mutator phe- dam methylation of d(G-A-T-C) sequences within the DNA du- notype (13, 14). In addition, genetic analysis has suggested that plex. Thus, the heteroduplex dam methylase participates in a pathway involving mutH, mutL, 5'-G-A-A-T-T-C and mutS function (15, 16). However, the most compelling evidence has been provided 3'-T-T-T-A-A-G by the transfection experiments of Meselson and colleagues (refs. is repaired in vivo to an EcoRI-sensitive form if the strand bearing 12 and 17; M. Rykowski and M. Meselson, personal commu- the wild-type EcoRI sequence carries the dam modification and nication), which employed A heteroduplexes in defined states the other does not. Such molecules are also subject to mismatch of dam methylation. Hemi-methylated heteroduplexes were repair by E. coli extracts. The in vitro activity is also dependent subject to mismatch repair almost exclusively on the unmeth- on ATP, the state of dam methylation of mismatch heterodu- ylated strand to yield the genotype of the methylated strand, plexes, and products of mutH, mutL, mutS, and uvrE loci. How- and repair was found to be substantially decreased if both strands ever, crude fractions deficient in these gene products do comple- of the heteroduplex were fully modified at d(G-A-T-C) se- ment in the cell-free system, thus providing assays for their isolation. quences. The in vitro reaction is accompanied by repair synthesis on the In this paper we describe a substrate that permits in vitro unmethylated DNA strand. analysis of mismatch repair. The reaction in crude cell fractions is dependent on the state of DNA methylation and requires Base-pair mismatches within the DNA double helix may arise ATP as well as the products of mutH, mutL, mutS, and uvrE. in several ways. Spontaneous deamination of cytosine or ad- enine leads to G-U or I-T mispairs, respectively. Such lesions MATERIALS AND METHODS are thought to be repaired via action of DNA glycosylases (1, 2). Mismatched base pairs may also occur during homologous Bacterial and Phage Strains. E. coli K38 (Hfr C; ref. 18) was genetic recombination if allelic differences are included within from R. E. Webster of this department; RS5033 (Hfr H, metBI the heteroduplex region formed by breakage and rejoining of rell strlO0 azi7 lacMS286 thi dam4 480dIIlacBKl), from E. B. parental molecules. The phenomenon of gene conversion has Konrad (19); andJC4583.(endAl ga144 thi-l thyA48 thyR27 lc6l), been attributed to repair of such mismatches (3). DNA repli- from A. J. Clark (University of-California, Berkeley). B. Glick- cation errors may also contribute to generation of mispaired man (National Institutes of Environmental Health Sciences) bases. In the case of the chromosome of Escherichia coli, the provided strains KMBL 3752 (endA101 thyA306 lysA65 argA103 error rate has been estimated to be 108 to 10-11 per base pair bio-87 metE72 pheA97 purA aroB cysC), KMBL 3773 (as KMBL replicated (4, 5), substantially lower than the in vitro error rate 3752 but mutHl01), KMBL 3774 (as KMBL 3752 but mutL101), of DNA replication systems derived from this organism (6). KMBL 3775 (as KMBL 3752 but mutS101), and KMBL 3789 (as Direct evidence for existence of a mechanism for correction KMBL 3752 but uvrE502) (15). of mismatched base pairs has been provided in the E. coli sys- Bacteriophage fl R229 containing an EcoRI site in the in- tem by transfection with A (7-9), 4X174 (10), and T7 hetero- tragenic region (20) was provided by R. E. Webster. Deriva- duplexes (11) marked genetically on the two DNA strands. Such tives containing mutations within the EcoRI sequence were experiments have demonstrated that incorrect base pairs can be constructed by minor modification of the procedure of Shortle eliminated from heteroduplexes prior to replication and, fur- and Nathans (21) and structure was determined by DNA se- thermore, have implicated the products of E. coli mutH, mutL, quence analysis according to Maxam and Gilbert (22). mutS, and uvrE loci in this process (9, 11, 12). Because strains DNA Preparations. Bacteriophage fl replicative form (RF) defective with respect to these loci exhibit a mutator phenotype I preparations were isolated (23) from chloramphenicol-treated (5), it seems likely that this set of genes directs a system in- infected cells (24). Single-stranded viral DNA was isolated from volved in postreplication repair of DNA biosynthetic errors. As purified virions (25, 26). DNAs devoid of dam methylation at pointed out by Wagner and Meselson (8), function of such a d(G-A-T-C) sequences were prepared by using strain RS5033 as host. DNAs modified at this sequence were prepared by us- The publication costs of this article were defrayed in part by page charge ing K38 or K38 harboring pGG503, a plasmid which leads to 10- payment. This article must therefore be hereby marked "advertise- ment" in accordance with 18 U. S.C. §1734 solely to indicate this fact. Abbreviations: RF, replicative form; kb, kilobase(s).

4639 Downloaded by guest on September 28, 2021 X%.rxo Biochemistry: Lu et al. Proc. Natl. Acad, Sci. USA 80 (1983) to 40-fold overproduction of the dam methylase (27). Table 1. Methyl-directed mismatch repair of fl R229 in vivo DNA heteroduplexes were prepared by mixing fl duplexes Transfectants (100 of RF III, linearized with HincII) with a 10-fold molar C/V ,ug Mixed Total excess of viral strands, followed by alkaline denaturation and methylation EcoRIP EcoRT/ annealing as described (28). After isolation by hydroxylapatite +/- 48 1 2 51 chromatography (29), double-stranded DNA was dialyzed against -/+ 0 17 0 17 0.01 M Tris-HCI, pH 8.0/1 mM EDTA, and then was subjected -/- 22 18 1 41 to closure with E. coli DNA ligase (30) in the presence of ethid- F1 heteroduplexes containing a G-T mismatch within the EcoRI site ium bromide (96 mmol of dye per mol of nucleotide). Cova- were constructed in several states of methylation. The state of dam lently closed DNA circles were then isolated by equilibrium methylation ofcomplementary (C) and viral (V) strands is indicated by centrifugation in CsCl/ethidium bromide as described (31). + and -, respectively. The strands are indicated as Hemi-methylated heteroduplexes prepared by using meth- C 5'-G-A-A-T-T-C ylated RF and unmethylated viral strands were resistant to cleavage by Mbo I, indicating that all d(G-A-T-C) sites were in V 3'-T-T-T-A-A-G. the hemi-methylated state. In contrast, hemi-methylated mol- In the case ofviral strand modification, incompletely hemi-methylated ecules prepared from unmethylated RF and methylated viral molecules were eliminated by Mbo I cleavage. Heteroduplexes were in- strands were subject to some cleavage by this enzyme. Because troduced into K38by transfection (35), and cells were immediately plated this problem was less severe if phage were propagated on the for infective centers. Individual plaques were picked into 1 ml of R broth overproducer of the dam methylase (above), this host was em- (32), which was then supplemented with 0.2 ml ofan overnight culture ployed for preparation of methylated viral strands. In this case, of K38. After incubation at 370C for 3-4 hr, a 0.1-ml sample was re- = 69% of the heteroduplexes were resistant to Mbo I, 21% were moved and used to infect a logarithmic culture of K38 (1.5 ml, A590 1.5 in Rbroth). Twenty minutes later chloramphenicol was added to 30 cleaved once, and 10% were cleaved more than once. If the tsg/ml, and after an additional 60 min cells were collected by centrif- unmethylated sites are distributed randomly with respect to ugation. RF DNAwas preparedby amini-lysate procedure (36) and tested the four d(G-A-T-C) sites of fl viral strands, a binomial distri- for cleavage byEcoRI endonuclease. DNA scored as sensitive to the en- bution with a probability of methylation of 0.9 for each d(G-A- zyme was cleaved to an extent of >95%, whereas resistant DNA was T-C) site would predict 66%, 29%, and 5% for these three classes. subject to <5% cleavage. Intermediate values were scored as mixed. Although perhaps not entirely random, it seems reasonable to conclude that methylation of viral single strands is about 90% ner et al. (33) except that thawed cell suspensions were sup- complete at each d(G-A-T-C) sequence, provided that phage plemented with 1.2 mM dithiothreitol/0. 15 M KCI/0.23 mg of are propagated on the methylase overproducer. When indi- lysozyme per ml, and heat shock at 370C was for a time suf- cated, partially hemi-methylated molecules were eliminated from ficient to yield a final suspension temperature of 20'C. After heteroduplex preparations by Mbo I cleavage and rebanding in centrifugation, the supernatant was treated with solid (NH4)2SO4 CsCI/ethidium bromide. (0.42 g/ml), and the precipitate was collected by centrifuga- Fully methylated heteroduplexes were prepared by in vitro tion, resuspended in 0.3 ml of 0.025 M Hepes, pH 7.6/0.1 mM methylation of hemi-methylated heteroduplexes by using ho- EDTA/2 mM dithiothreitol/100 mM KCI, and dialyzed against mogeneous dam methylase (27) and S-[methyl-3H]adenosyl-L- this buffer until the conductivity achieved a value equivalent methionine (80 Ci/mmol; 1 Ci = 3.7 x 1010 Bq; New England to that of 0.2 M KCI (about 90 min). This material (fraction I, Nuclear). 40-70 mg/ml) was frozen in small samples at -70°C. Preparation of Cell Extracts. Bacteria were grown in 1-liter Mismatch Repair Assay. The assay scores conversion of fl cultures of LB medium (32) supplemented with 0.1% glucose R229 heteroduplexes, containing a base-pair mismatch within to an A590 of 1.0-1.2, collected by centrifugation at 15°C, sus- the EcoRI site, to a form which is subject to cleavage by EcoRI pended in 2 ml of 0.05 M Tris HCI, pH 7.6/10% sucrose, and endonuclease (Fig. 1). Reactions (final volume, 10 ,l) con- frozen in dry ice/ethanol. Cells were lysed according to Wick- tained 0.02 M Tris-HCI, pH 7.6/5 mM MgCl2/50 ,ug of bovine serum albumin per ml/1 mM glutathione/0.2 mM spermi- H/incI dam dine/1.5 mM ATP/0.5 mM NAD/100 ,M (each) the four deoxyribonucleoside-5'-triphosphates/10 ,ug of DNA per ml and fraction I (added in 2-5 ,l), as indicated. After incubation for 1 hr at 37°C, 30 Al of 25 mM EDTA was added, and samples were extracted twice with phenol and twice with diethyl ether. DNA, collected by ethanol precipitation, was dissolved in 16 ,ul of 0.05 M Tris-HCI, pH 7.6/0.1 M NaCl/5 mM MgCl2/1 mM dithiothreitol/50 ,g of bovine serum albumin per ml and was fI R229 W hydrolyzed with EcoRI (31) and BamHI (Bethesda Research dam Laboratories) endonucleases. Digestion products were sepa- \\ ~~~~~2000W rated by electrophoresis on a 1% agarose gel in 0.036 M Tris-HCl/ 0.03 M NaH2PO4/1 mM EDTA at 5.3 V/cm for 4 hr. Ethid- \A~g4000 X ium-stained gels were photographed (Kodak Tri-X film) and DNA dam/Bam HI content of bands quantitated by microdensitometer scan. RESULTS FIG. 1. Restriction map offl R229. The DNA of this phage, the se- Methyl-Directed Mismatch Repair in Vivo. To determine quence ofwhich is available (34), contains one EcoRI site (position 5,616) whether repair of mismatches within duplex fl DNA is methyl- and four d(G-A-T-C) sites recognized by dam methylase (positions 216, directed, as has been previously observed in the case of A mol- 1,382, 1,714, and 2,221; the last of these is also a BamHI site). Mis- ecules (12, 17), heteroduplexes of fl R229 containing a G-T matched heteroduplexes used in this work were constructed by using in R229 and a derivative containing a C-to-T transition at position 5,621. mismatch within the EcoRI site (Fig. 1) were constructed Downloaded by guest on September 28, 2021 Biochemistry: Lu et aL Proc. Natl. Acad. Sci. USA 80 (1983) 4641

1 2 34 5 6 7 8 9 Table 2. Requirements for in vitro mismatch repair - EcoRI sites Condition repaired, fmol

Il- Complete 11 Fully methylated DNA 1.8 Without ATP 0.6 Without NAD 13 3.3 kb- 3.1 kb- Without deoxyribonucleoside- 5'-triphosphates 6.3 Results shown were determined by densitometric analysis of pho- 11 12 13 14 15 16 17 18 19 20 tographs like those shown in Fig. 2 and represent the average of sev- 10 eral experiments. Extent ofrepair was calculated on the basis of peak areas of 3.1- and 3.3-kb restriction products generated upon cleavage ofrepairedmolecules withEcoRIandBamHIendonucleases. Complete repair would correspond to 24 fimol. Mismatch Repair in Vitro. The EcoRI resistance of fl R229 III- _ heteroduplexes containing a mismatch within the EcoRI site provides a simple means to test for mismatch repair activity in 3.3 k b- 3.1 kb- an in vitro system. As shown in Fig. 2 (lanes 1-4), hemi-meth- ylated heteroduplexes containing a G-T mismatch within the EcoRI recognition sequence were subject to mismatch correc- tion by crude E. coli fractions, as judged by generation of the FIG. 2. Mismatch repair in vitro. DNAs were incubated as indi- cated with cell extracts or restriction endonucleases, or both, and then appropriate restriction products upon subsequent cleavage with were fractionated on 1% agarose gels, which were stained with ethid- EcoRI and BamHI (Fig. 1). Several lines of evidence indicate

ium bromide. I, 1, and m designate covalently closed, nicked, and lin- that this reaction is related to the mechanism of mismatch re- earformsoffl DNA. Digestionoffl R229duplex circles withEcoRI and pair as it occurs in the cell. First, repair of heteroduplexes BamHI endonucleases yields 3.1- and 3.3-kilobase (kb) products (Fig. methylated on both DNA strands was only 16% of that ob- 1). (Upper) Lane 1, untreated hemi-methylated fl heteroduplex sub- served with hemi-methylated molecules (Fig. 2, lane 6, and strate containing a G-T mismatch (see Table 1); lane 2, as lane 1, but hydrolyzedwithEcoRI andBamHl endonucleases; lane 3, as lane 1, but Table 2). This is consistent with the finding that the efficiency incubated with fraction I (135 lg ofprotein) derived from KMBL 3752 of in vivo mismatch repair of A heteroduplexes is greatly de- (mut+); lane 4, as lane 3, but hydrolyzed with EcoRI and BamHI en- creased if both DNA strands are fully modified by the dam en- donucleases; lane 5, as lane 4, but the hemi-methylated substrate con- zyme (12). More convincing, however, is the observation that tained a wild-typeEcoRl sequence in both strands; lane 6, as lane 4, but in vitro mismatch correction occurs poorly, if at all, in extracts mismatch heteroduplex was methylated on both DNA strands; and lanes prepared from isogeneic strains defective in mutH, mutL, mutS, 7-9, samples run on these last three lanes were prepared as for lane 4, except incubation with fraction I was performed in the absence of add- or uvrE function (Fig. 2, lanes 10-14; see Table 4). These genes ed ATP, NAD, or deoxyribonucleoside-5'-triphosphates, respectively. have been previously implicated in the process of mismatch (Lower) Samples were as for lane 4, except fraction I was derived from repair in vivo (9, 11, 12). isogeneic wild-type and mutator strains. Lane 10, mute; lane 11, Requirements for the cell-free reaction are shown in Fig. 2 mutfi0l; lane 12, mu1tL101; lane 13, nutS101; lane 14, uvrE502; lane and Table 2. As mentioned mismatch correction was 15, mutH101 and mutL101; lane 16, mutHi101 and mutS101; lane 17, above, mutH101 and uvrE502; lane 18, mutL101 and mutS101; lane 19, mutL101 greatly decreased if both DNA strands were methylated at d(G- and uvrE502; lane 20, mutSlO1 and uvrE502. In those experiments em- A-T-C) sites. The reaction is also highly dependent on ATP, with ploying a single fraction I, reactions contained about 250 jig ofprotein, mismatch correction in the absence of this nucleotide being only whereas those employing fraction I from two strains contained about about 5% (Fig. 2, lane 7) of that in its presence. However, 125 jug from each. omission of the ligase cofactor NAD from the crude system was without significant effect on the reaction (Fig. 2, lane 8). Al- several states of dam methylation and introduced into strain though in vitro mismatch repair occurred in the absence of ex- K38 by transfection (35). Individual plaques were isolated and ogenous deoxyribonucleoside-5'-triphosphates, the extent of used to prepared RF DNA, which was then tested for sensi- reaction was decreased (Fig. 2, lane 9). As will be shown below, tivity to cleavage by EcoRI endonuclease (Table 1). As can be because hemi-methylated mismatch heteroduplexes are sub- seen, sensitivity to the endonuclease was governed by the state ject to strand-specific repair DNA synthesis, mismatch repair of dam methylation of transfecting heteroduplexes. In the case of hemi-methylated heteroduplexes, the majority of progeny Table 3. Methyl dependence and mismatch dependence of repair phage contained an EcoRI-sensitive site if the heteroduplex was DNA synthesis methylated on the strand containing the canonical EcoRI se- C/V C/V Relative dTMP quence. However, hemi-methylated heteroduplexes modified mismatch methylation incorporation on the strand containing the mutant EcoRI site yielded progeny G-T phage lacking the EcoRI sequence. In contrast, unmethylated +/- 1.0 G-T 0.27 ± 0.06 mismatch heteroduplexes yielded either EcoRI-sensitive or +/+ None (G-C) +/- 0.33 ± 0.11 -resistant progeny in a ratio of about 1.2:1. These findings are in accord with results previously obtained by using bacterio- Reactions (20 pm) were performed as described in Materials and phage A heteroduplexes (refs. 12 and 17; M. Rykowski and M. Methods, except that [a-32P]dTTP (5 x 104 cpm/pmol, New England Meselson, personal communication) and, in particular, are con- Nuclear) was present at 50 MLM, fl heteroduplexes, at 5 pg/ml, and fraction I (prepared from JC4583), at 2.6 mg/ml. After incubation at sistent with existence of an E. coli mismatch repair system, the 370C for 30 min, perchloric acid-insoluble label was determined. One- strand specificity of which can be directed by the state of meth- hundred percent incorporation corresponds to 9.8 pmol of dTMP. Re- ylation of d(G-A-T-C) sequences. sults represent mean ± SD. Downloaded by guest on September 28, 2021 4642 Biochemistry: Lu et al. Proc. Natl. Acad. Sci. USA 80 (1983)

+ .- c. -.+ _ I E A B C A 6 C A

C

0

._ mutH Fraction Lugl FIG. 3. Methyl-directed repair DNA synthesis on mismatch het- 0 eroduplexes. Reactions were performed as in the legend to Table 3 by using the G-T mismatch heteroduplex (Table 1) in the two possible states FE ofhemi-methylation. After phenol extraction, dialysis, and concentra- 10 tion by ethanol precipitation, 32P-labeled DNAs were hydrolyzed with Bsp I endonuclease and cleavage products were separated by electro- phoresis on a 6% polyacrylamide gel (22). The three largest cleavage products (A, B, and C shown in Fig. 1) were eluted and subjected to elec- trophoresis on a strand separation gel (22), with bands visualized by autoradiography. A 5'-end labeled Bsp I hydrolysate of fl R229 was 5- included on gels to serve as marker. Origin ofseparated strands offrag- ments A, B, and C was determined in separate experiments by hy- bridization to circular viral strands. The state of methylation of viral (V) and complementary (C) strands is indicated by + and -. 50 100 150 under such conditions presumably reflects contamination of the mutS Fraction lugl crude system by residual deoxyribonucleotides. Furthermore, FIG. 4. Complementation assays for mutH and mutS components. heteroduplexes repaired by some fraction I preparations in the Reactions contained 120 pug offraction Ifrom a mutS101 strain and pro- absence of deoxyribonucleotides yielded anomalous products tein as indicated from a mutH101 strain (Upper). In the converse ex- upon cleavage with EcoRI and BamHI endonucleases. In par- periment (Lower) reactions contained 132 ,ug of a mutH101 fraction I ticular, the smaller 3.1-kb product was under-represented in and a mutS101 fraction I as indicated. such cases by as much as 60-80%. An example of this effect is shown in lane 9 of Fig. 2. However, in the presence of deoxyri- presence of radioactive deoxyribonucleotides. As shown in Ta- bonucleotides the two EcoRI-BamHI products were always ob- ble 3, the extent of repair synthesis was dependent on both the tained in near molar equivalence (0.88 ± 0.08 mol of the 3.1- state of heteroduplex methylation and the presence of a base- kb product per mol of 3.3-kb fragment). Thus, the shorter EcoRI- pair mismatch. In particular, repair synthesis on symmetrically BamHI fragment, which contains all of the d(G-A-T-C) sites of methylated molecules containing a mismatch or on hemi-meth- fl (Fig. 1), was selectively lost, or present in altered form, after ylated molecules lacking a mismatch was only 30% of that ob- incubation with extracts in the absence of deoxyribonucleo- served for hemi-methylated heteroduplexes containing a G-T tides. mismatch within the EcoRI site. Thus, a significant component In Vitro Mismatch Correction Is Accompanied by Methyl- of the repair synthesis requires both the presence of a mis- Directed Repair DNA Synthesis. To assess possible involve- match and d(G-A-T-C) sequences unmethylated on at least one ment of repair DNA synthesis during in vitro mismatch cor- strand. Furthermore, this component of the repair synthesis rection, heteroduplexes were incubated with fraction I in the was also found to be dependent on the presence of functional products of mutH, mutL, and mutS loci (unpublished data). Table 4. In vitro complementation of mutator extracts in The dependence of repair synthesis on state of DNA meth- mismatch repair ylation led us to examine possible strand-specificity of this re- EcoRI sites repaired, action. Heteroduplexes containing a G-T mismatch within the in states of hemi- Source of fraction I fmol/mg of protein EcoRI site were constructed both possible methylation and were incubated with fraction I in the presence Wild-type (KMBL 3752) 42 ± 9 of 32P-labeled deoxyribonucleotides. After isolation, these DNAs 4 mutH101 were subjected to cleavage with Bsp I endonuclease and the mutL101 <4 three restriction products (fragments A, B, and C; Fig. <4 larger mutSlOl 1) were isolated. These fragments, which span most of the mol- uvrE502 <4 ecule, were analyzed by polyacrylamide gel electrophoresis un- der strand separation conditions (22). Repair synthesis on hemi- mutH101 + mutL101 10 molecules occurred on the unmeth- mutH101 + mutS101 45 methylated predominantly Reversal of the re- mutH101 + uvrE502 42 ylated strand (Fig. 3). methylation pattern from one strand to the other. mutL101 + mutS101 16 sulted in switch of synthesis or uvrE Function mutL101 + uvrE502 7 Extracts Deficient in mutH, mutL, mutS, mutS101 + uvrE502 36 Complement in Vitro. Although functions derived from mutH, mutL, mutS, and uvrE are deficient in mismatch correction in Mismatch repair assay was performed and reactions (10 p.l) con- the in vitro system, activity is recovered upon mixing of ex- a tained 120-270 tkg of fraction I in the case of assays utilizing single are shown in 2 contained 120-135 of each frac- tracts. Results of this type of analysis Fig. (lanes extract. Complementation assays jig 4. With the ex- tion I, and specific activity refers to total protein present. Results rep- 15-20) and are quantitated in Table possible resent mean ± SD. ception of the mutL, uvrE combination, complementation was Downloaded by guest on September 28, 2021 Biochemistry: Lu et al. Proc. Natl. Acad. Sci. USA 80 (1983) 4643 observed with all possible pairs of extracts and in many cases pairs. However, in the fl heteroduplexes described here, the resulted in a level of repair comparable to that observed with shortest distance between the mismatch and a d(G-A-T-C) site fractions derived from wild-type cells. Complementation of ex- was 1,008 base pairs. Nevertheless, repair occurred with high tracts derived from the mutL101 strain was consistently low, efficiency, indicating that necessary recognition events can oc- although the basis of this effect is not yet understood. As in the cur over large distances. Elucidation of the mechanism of this case of wild-type fractions, mismatch repair in mixed extracts reaction must await its reconstitution in a purified system. The is also dependent on the state of DNA methylation, being de- in vitro assay described here provides a basis for isolation of the creased by about 80% if heteroduplexes are methylated on both necessary components. strands (not shown). Furthermore, in vitro complementation, We gratefully acknowledge the useful comments of Drs. Matthew based on repair of a mismatch within the EcoRI site of fl R229, Meselson, Patricia Pukkila, and Barry Glickman during the course of can be employed as a quantitative assay for components in- this study. This work was supported by Grant GM 23719 from the Na- volved in this process. Application of this method for assay of tional Institute of General Medical Sciences. P.M. is recipient of Re- components deficient in mutH and mutS extracts is presented search Career Development Award CA 00495 from the National Can- in Fig. 4. cer Institute. 1. Lindahl, T., Ljungquist, S., Siegert, W., Nyberg, B. & Sperens, DISCUSSION B. (1977)J. Biol Chem. 252, 3286-3294. The results reported here are in agreement with the Wagner- 2. Karran, P. & Lindahl, T. (1978)J. Biol Chem. 253, 5877-5879. Meselson hypothesis (8) that E. coli possesses a methyl-directed 3. Holliday, R. A. (1964) Genet. Res. 5, 282-304. postreplication repair system for elimination of replication er- 4. Drake, J. W. (1969) The Molecular Basis of Mutation (Holden-Day, rors within newly synthesized DNA. Mismatch correction by San Francisco). 5. Cox, E. C. (1976) Annu. Rev. Genet. 10, 135-156. this system is directed to the newly synthesized strand by vir- 6. Hibner, U. & Alberts, B. M. (1980) Nature (London) 285, 300- tue of its transient undermethylation at d(G-A-T-C) sequences. 305. We have shown that an artificially constructed heteroduplex of 7. Wildenberg, J. & Meselson, M. (1975) Proc. Natl. Acad. Sci. USA fl R229 DNA containing a G-T mismatch (Table 1) is subject 72, 2202-2206. to methyl-directed mismatch repair upon introduction into E. 8. Wagner, R. & Meselson, M. (1976) Proc. Natl Acad. Sci. USA 73, coli by transfection. These results are essentially identical to 4135-4139. 9. Nevers, P. & Spatz, H. (1975) Mol. Gen. Genet. 139, 233-243. those obtained previously with A DNA by Meselson and co- 10. Baas, P. D. & Jansz, H. S. (1972)J. Mol Biol. 63, 557-568. workers (refs. 12 and 17; M. Rykowski and M. Meselson), al- 11. Bauer, J., Krammer, G. & Knippers, R. (1981) Mol Gen. Genet. though in those experiments the nature of the mismatches was 181, 541-547. not known. These data are also in accord with the work of Kra- 12. Pukkila, P. J., Peterson, J., Herman, G., Modrich, P. & Mesel- mer et al. (37), who demonstrated an influence of d(G-A-T-C) son, M. (1983) Genetics, in press. methylation on marker recovery in directed mutagenesis ex- 13. Marinus, M. G. & Morris, N. R. (1974)J. Mol. Biol. 85, 309-322. 14. Herman, G. E. & Modrich, P. (1981)J. Bacteriol, 145, 644-646. periments involving DNA cloned into M13 vehicles. 15. Glickman, B. W. & Radman, M. (1980) Proc. Natl Acad. Sci. USA We have also found that hemi-methylated fl heteroduplexes 77, 1063-1067. containing the G-T mismatch are subject to repair by E. coli 16. McGraw, B. R. & Marinus, M. G. (1980) Mol. Gen. Genet. 178, extracts in an ATP-dependent reaction. Because the assay em- 309-315. ployed monitors conversion of an EcoRI site containing a mis- 17. Radman, M., Wagner, R. E., Glickman, B. W. & Meselson, M. match to a form sensitive to cleavage by the endonuclease, it (1980) in Progress in Environmental Mutagenesis, ed. Alacevic, M. (Elsevier/North Holland, Amsterdam), pp. 121-130. has been possible to directly examine in vitro repair of this mis- 18. Lyons, L. B. & Zinder, N. D. (1972) Virology 49, 45-60. match with only one of the two possible hemi-methylated het- 19. Konrad, E. B. (1977)J. Bacteriol 130, 167-172. eroduplex configurations. Several lines of evidence indicate that 20. Boeke, J. D. (1981) Mol Gen. Genet. 181, 288-291. the cell-free reaction is analogous to that which has been ob- 21. Shortle, D. & Nathans, D. (1978) Proc. Nati Acad. Sci. USA 75, served in vivo. First, extent of in vitro mismatch correction is 2170-2174. markedly decreased if both DNA strands are methylated at d(G- 22. Maxam, A. & Gilbert, W. (1980) Methods Enzymol. 65, 499-560. the EcoRI endonuclease 23. Hardies, S. C. & Wells, R. D. (1979) Gene 7, 1-14. A-T-C) sites. Second, although mis- 24. Model, P. & Zinder, N. D. (1974)J. Mol Biol 83, 231-251. match repair assay can detect correction in only one of the two 25. Lin, T.-C., Webster, R. E. & Konigsberg, W. (1980)J. Biol. Chem. hemi-methylated configurations, mismatch-dependent repair 255, 10331-10337. DNA synthesis has been examined in both. This synthesis oc- 26. Webster, R. E., Grant, R. A. & Hamilton, L. A. (1981) 1 Mol Biol curs predominantly on the unmethylated DNA strand. Lastly, 152, 357-374. cell-free mismatch correction requires components that are 27. Herman, G. E. & Modrich, P. (1982)J. Biol Chem. 257, 2605-2612. 28. Horiuchi, K. & Zinder, N. D. (1972) Proc. Nati Acad. Sci. USA lacking in strains defective in mutH, mutL, mutS, and uvrE loci 69, 3220-3224. previously implicated in the process of mismatch repair in vivo 29. Britten, R. J., Graham, D. E. & Neufeld, B. R. (1974) Methods (9, 11, 12). Furthermore, preliminary experiments indicate that Enzymol 29E, 363-418. an A-C mismatch within the EcoRI site of fl R229 heterodu- 30. Modrich, P., Anraku, Y. & Lehman, I. R. (1973)J. Biol Chem. 248, plexes is also subject to mismatch repair by this in vitro system 7495-7501. in a reaction that appears analogous to that observed with the 31. Modrich, P. & Zabel, D. (1976) J. BioL Chem. 251, 5866-5874. Thus, the system appears to rec- 32. Miller, J. H. (1972) Experiments in Molecular Genetics (Cold Spring G-T mismatch reported here. Harbor Laboratory, Cold Spring Harbor, NY). ognize both transition mispairs (unpublished data). 33. Wickner, W., Brutlag, D., Schekman, R. & Kornberg, A. (1972) To repair mismatches in a strand-directed fashion, it is clear Proc. Natl Acad. Sci. USA 69, 965-969. that the system has to recognize both a mismatch and a hemi- 34. Beck, E. & Zink, B. (1981) Gene 16, 35-58. methylated dam site. The tetranucleotide sequence d(G-A-T-C) 35. Lederberg, E. M. & Cohen, S. N. (1974) J. BacterioL 119, 1072- would be expected to occur on the average once every 256 base 1074. in a 36. Birnboim, H. C. & Doly, J. (1979) Nucleic Acids Res. 7, 1513-1523. pairs random sequence, and available evidence is consis- 37. Kramer, W, Schughart, K. & Fritz, H.-J. (1982) Nucleic Acids Res. tent with this sort of frequency within the E. coli chromosome 10, 6475-6485. (38). Therefore, the average distance between a mismatch and 38. Szyf, M., Gruenbaum, Y., Urieli-Shoval, S. & Razin, A. (1982) the closest d(G-A-T-C) site is expected to be about 130 base Nucleic Acids Res. 10, 7247-7259. Downloaded by guest on September 28, 2021