Mismatch-, MutS-, MutL-, and Helicase II-dependent Unwinding from the Single-strand Break of an Incised Heteroduplex*

Escherichia coli MutS, MutL, and DNA helicase II are sufficient to initiate mismatch-dependent unwinding of an incised heteroduplex (Yamaguchi, M., Dao, V., and Modrich, P. (1998) J. Biol. Chem., 273, 9197–9201). We have studied unwinding of 6.4-kilobase circular G-T heteroduplexes that contain a single-strand incision, 808 base pairs 5′ to the mismatch or 1023 base pairs 3′ to the mispair as viewed along the shorter path between the two DNA sites. Unwinding of both substrates in the presence of MutS, MutL, DNA helicase II, and single-stranded DNA binding protein was mismatch-dependent and initiated at the single-strand break. Although unwinding occurred in both directions from the strand break, it was biased toward the shorter path linking the strand break and the mispair. MutS and MutL are thus sufficient to coordinate mismatch recognition to the orientation-dependent activation of helicase II unwinding at a single-strand break located a kilobase from the mispair.

The strand specificity necessary for correction of DNA biosynthetic errors by the Escherichia coli mismatch repair system is provided by the transient absence of adenine modification of d(GATC) sequences within newly synthesized DNA (1). Repair is initiated by binding of a MutS homodimer to a mismatch followed by addition of MutL to this complex (2)(3)(4). Assembly of this ternary complex activates a MutH-associated endonuclease that cleaves the unmethylated strand at a hemimethylated d(GATC) sequence within newly replicated DNA (5). The single-strand break introduced by MutH, which may occur either 3Ј or 5Ј to the mismatch on the unmethylated strand, directs the excision of that portion of the unmodified strand spanning the d(GATC) sequence and the mispair (6,7). Excision requires MutS, MutL, DNA helicase II (also called MutU), and depending on the strand break to mismatch orientation, a 3Ј 3 5Ј or 5Ј 3 3Ј single-strand exonuclease (6,8).
The accompanying manuscript (9) demonstrates that MutS and MutL greatly enhance the activity of DNA helicase II on incised heteroduplex DNA. In this paper we have used KMnO 4 to determine the site of initiation of mismatch-dependent helix unwinding in DNA substrates containing a site-and strandspecific, single-strand break. Permanganate preferentially attacks single-stranded DNA where it oxidizes the 5,6 double bond of thymine and methylcytosine and reacts to a lesser degree with other bases (10,11). This single-strand selective reagent has been used previously to detect helix opening associated with promoter melting by bacterial and eukaryotic RNA polymerases (12,13). Using this approach we show that MutS-, MutL-, and helicase II-dependent unwinding of an incised heteroduplex initiates at the strand break, with the direction of unwinding being biased toward the shorter path between the strand break and the mismatch in a circular heteroduplex.

EXPERIMENTAL PROCEDURES
Proteins and DNA-E. coli MutS (14), MutL (3), and MutH (15) were purified as described previously. DNA helicase II was isolated from an overproducing strain according to Runyon et al. (16). Single-strand DNA-binding protein (SSB) 1 and T4 polynucleotide kinase were purchased from Amersham Pharmacia Biotech, and restriction endonucleases were purchased from New England Biolabs.
Circular 6440-base pair (bp) G-T heteroduplex and G⅐C homoduplex DNAs containing a strand-and site-specific, single-strand break were prepared using f1MR phage DNAs (2,6,17). The structure of these molecules is illustrated in Fig. 1. DNAs with a single-strand break in the complementary strand at the HincII site are referred to as 5Јheteroduplexes since the nick is 5Ј to the mismatch as viewed along the shorter path (808 base pairs) in the circular DNA. A second configuration, referred to as a 3Ј-substrate, was prepared by MutH incision of the viral strand at the single GATC site (1023 bp from the mismatch, shorter path) in hemimethylated DNA (6,15). Corresponding control homoduplexes containing a G⅐C base pair instead of a mismatch at position 5632 were constructed in a similar manner. DNA size markers were prepared by cleavage of f1MR3 (2) replicative form DNA with appropriate restriction endonucleases.
Oligonucleotides (Table I), which were purchased from Oligos Etc., were 5Ј-32 P-end-labeled using [␥-32 P]ATP (3000 Ci/mmol, New England Nuclear) and T4 polynucleotide kinase according to the recommendations of the manufacturer. Labeling was terminated by addition of EDTA to 10 mM and heating at 65°C for 10 min. Unincorporated label was removed by passing the solution through Sephadex G-25 (Amersham Pharmacia Biotech) equilibrated with 10 mM Tris/HCl (pH 7.6), 1 mM EDTA, and 100 mM NaCl. Labeled oligonucleotide was ethanol precipitated and resuspended in 10 mM Tris/HCl (pH 7.6), 1 mM EDTA.
Chemical Quench Analysis-Chemical quench experiments utilized a KinTek apparatus (KinTek Instruments). Unless specified otherwise, a solution (40 l) containing 0.8 g (0.19 pmol) heteroduplex DNA, 1.4 g MutL (10 pmol as dimer), 2.8 g MutS (15 pmol as dimer), 0.8 g of DNA helicase II (10 pmol as monomer) in 50 mM Hepes/KOH (pH 8.0), 20 mM KCl, 6 mM MgCl 2 , 50 g/ml bovine serum albumin, 1 mM dithiothreitol was mixed with 40 l of 2 mM ATP containing 4 g of SSB in the same buffer. Mixing syringes were maintained at 37°C and reaction times varied between 50 ms and 5 s. Reactions were quenched by injection of 40 l of freshly prepared 30 mM KMnO 4 in H 2 O, and samples were collected in tubes on ice. Approximately 2 s after collection, permanganate oxidation was terminated by addition of 10 l of 1 M dithiothreitol. Samples were supplemented with 2 l of 0.5 M EDTA and 10 l of 10 mM Tris/HCl (pH 7.6), 1 mM EDTA, and passed through a spin-column containing S-300 (Amersham Pharmacia Biotech) equilibrated with 10 mM Tris/HCl (pH 7.6), 1 mM EDTA, 0.3 M NaCl. The column flow-through was extracted with phenol, precipitated with eth-* This work was supported in part by Grant GM23719 from the National Institute of General Medical Sciences. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  anol, and DNA was resuspended in 10 mM Tris/HCl (pH 7.6), 1 mM EDTA. After hydrolysis with restriction endonucleases as indicated, DNA digests (50 l) were supplemented with 5.6 l of piperidine and incubated at 90°C for 30 min. Piperidine-treated samples were evaporated to dryness, the residue was dissolved in 30 l of H 2 O, and the solution taken to dryness. After repetition of the H 2 O wash, the DNA pellet was dissolved in 6 l of 85% formamide, 20 mM EDTA, 0.05% bromphenol blue, and 0.05% xylene cyanol FF, and subjected to electrophoresis through 6% polyacrylamide gels containing 8 M urea. DNA fragments were transferred to nylon membranes (ICN Biotran TM ), and sites of permanganate oxidation mapped by indirect end-labeling with 5Ј-32 P oligonucleotides as indicated (18). DNA species were visualized by autoradiography and quantitated using a Molecular Dynamics PhosphorImager.

RESULTS
The excision step of methyl-directed mismatch repair requires MutS, MutL, DNA helicase II, an appropriate exonuclease, and a heteroduplex containing a single-strand break (6 -8). The accompanying article demonstrates that MutL greatly stimulates helicase II on conventional helicase substrates and that MutS and MutL activate the unwinding activity of helicase II on an incised DNA containing a mismatch (9). We show here that MutS-, MutL-, and helicase II-dependent unwinding of heteroduplex DNA initiates at the strand break.
MutS-, MutL-, Helicase II-, and Mismatch-dependent Unwinding Initiates at the Strand Break-To determine the site of initiation of heteroduplex unwinding, we have used permanganate, which preferentially oxidizes thymidylate residues in sin-gle-stranded DNA (10,11). Since thymidylate oxidation renders the phosphodiester bond subject to hydrolysis by strong base, the location of oxidized residues can be determined. The circular 6.4-kilobase DNAs used in this work contained a G-T mismatch (or an G⅐C base pair) and a site-and strand-specific nick located 808 bp 5Ј or 1023 bp 3Ј to the mismatch as viewed along the short path linking the two DNA sites in the circular molecules ( Fig. 1). For convenience, the two substrate orientations are referred to as 5Ј-or 3Ј-heteroduplexes, respectively.
A solution of G-T heteroduplex (or G⅐C homoduplex), MutL, MutS, and helicase II was mixed in a chemical quench device with a solution of ATP and SSB, and after brief incubation reactions were subjected to a 2 s oxidation with KMnO 4 (see "Experimental Procedures"). Sites of KMnO 4 hydrolysis were mapped relative to an appropriate restriction site by indirect end-labeling (18) after piperidine hydrolysis and electrophoresis through denaturing polyacrylamide gels. As shown in Fig. 2  (lanes 1-3), initiation of unwinding of the 5Ј-G-T heteroduplex . DNAs with a nick in the C strand are designated as 5Ј-heteroduplexes, since the incision lies 5Ј to the mismatch along the shorter path that joins the two sites in the circular molecule. DNAs with a nick on the V strand are dubbed 3Ј-heteroduplexes for a similar reason. Sites of cleavage by restriction endonucleases used in the experiments are also indicated with coordinates shown corresponding to the nucleotide 5Ј to the phosphodiester attacked. Shaded regions correspond to oligonucleotides (Table I) used for indirect end-labeling.

FIG. 2.
Helix unwinding initiates at the nick with 5-heteroduplex DNA. 5Ј-G-T heteroduplex or 5Ј-G⅐C homoduplex DNA containing a nick in the C strand at the HincII site ( Fig. 1) was incubated as indicated with MutS, MutL, DNA helicase II, SSB, and ATP followed by a 2 s quench with 10 mM KMnO 4 (see "Experimental Procedures"). After cleavage with DraI and piperidine treatment to cleave strands at the sites of permanganate oxidation, DNA samples were subjected to electrophoresis through denaturing polyacrylamide gels and electrotransferred to a nylon membrane. Fragments of interest were visualized by indirect end-labeling (7) using 32 P-oligonucleotides V6287 and C6289 as probes. As summarized in Table I, these probes hybridize to individual strands of the heteroduplex near the DraI site. Upper panel, permanganate reactive sites on the incised C strand; lower panel, permanganate reactive sites on the continuous V strand. Lanes: 1, complete system, 50 msec reaction; 2, complete, 1 s reaction; 3, complete, 5 s reaction; 4, helicase II omitted, 1 s reaction; 5, MutS omitted, 1 s reaction; 6, complete system, 1 s reaction but no KMnO 4 quench; 7, all proteins omitted, 1 s reaction; 8, complete system, 1 s reaction but G⅐C homoduplex substituted for heteroduplex; 9, marker for location of the strand break. The strong band that runs with the marker for the strand break in the lower panel is the result of permanganate oxidation of the closed circular V strand near the nick. It was not produced in the absence of the oxidizing agent (lane 6).
at the single-strand break was evident as judged by conversion of either the incised (upper panel) or the continuous (lower panel) strand of the molecule to a permanganate-sensitive form. Conversion of that region of the molecule to permanganate-sensitivity in the vicinity of the nick was rapid with a maximal unwinding rate achieved in 5 s or less. An otherwise identical G⅐C homoduplex did not support the reaction (lane 8), and increased permanganate reactivity was not observed in the absence of helicase II or MutS (lanes 4 and 5) or in the absence of MutL (not shown). Consequently, unwinding observed at the strand break is dependent on MutS, MutL, and helicase II and on the recognition of a mismatch 808-bp distant. Analysis of permanganate sensitivity of the continuous strand of the heteroduplex (Fig. 2, lower panel) revealed that mismatch-provoked unwinding of 5Ј-heteroduplex occurred to either side of the nick. This observation will be considered further below. As shown in Fig. 3, virtually identical results were obtained with a 3Ј-G-T heteroduplex in which the mismatch and strand break were separated by 1023 bp. However, in contrast to results obtained with the 5Ј-heteroduplex described above, the degree of unwinding of the 3Ј-substrate increased significantly between 1 and 5 s, perhaps due to the increased distance between the two DNA sites in the latter molecule.
Although presence of a G-T mismatch is known to enhance helix dynamics in the vicinity of the mispair (19, 20), the mismatched thymidylate in the G-T heteroduplex did not de-tectably react with permanganate under the mild oxidation conditions used. 2 Furthermore, under conditions where singlestrand character was rapidly generated at the strand break in the G-T heteroduplex in the presence of MutS, MutL, helicase II, and SSB, permanganate oxidation products were not detected in the vicinity of the mismatch after 5 s incubation. This is illustrated in Fig. 4 for the 5Ј-G-T heteroduplex, and identical results were obtained with the 3Ј-substrate (not shown). These results imply that the single-strand character that develops at the strand break in an incised heteroduplex is not the consequence of an unwinding event that initiates at the mispair and propagates to the mismatch. Consequently, we have concluded that unwinding by activated helicase II initiates at the strand break. The extent of unwinding observed with 5Јand 3Ј-heteroduplexes was about 50 -100 nucleotides.
Bias in the Direction of the DNA Helicase II Unwinding-As mentioned above and shown in the lower panels of Figs. 2 and 3, analysis of permanganate sensitivity of the continuous heteroduplex strand demonstrated that unwinding occurs in both directions from the strand break. However, analysis of mismatch-provoked methyl-directed excision tracts produced in extracts and in a purified system has demonstrated that excision is largely restricted to the shorter path between the strand signal and the mismatch in circular heteroduplexes similar to those used here (7). The unwinding reactions described above utilized only a subset of the proteins required for methyldirected mismatch repair, and helicase II is known to load FIG. 3. Helix unwinding initiates at the nick with 3-heteroduplex DNA. Substrates were 3Ј-G-T heteroduplex or 3Ј-G⅐C homoduplex DNA containing a nick in the V strand at the MboI site (Fig. 1). Reactions were performed as described under "Experimental Procedures" and in the legend to Fig. 2. DNAs were cleaved with HincII prior to electrophoresis, and oligonucleotides C005 and V005 were used for indirect end-labeling. Upper panel, permanganate reactive sites on the V strand; lower panel, permanganate reactive sites on the C strand. Lanes: 1, complete system, 50 msec reaction; 2, complete system, 1 s reaction; 3, complete system, 5 s reaction; 4, helicase II omitted, 1 s reaction; 5, MutS omitted, 1 s reaction; 6, complete system, 1 s reaction but no KMnO 4 quench; 7, all proteins omitted, 1 s reaction; 8, complete system, 1 s reaction but G⅐C homoduplex substituted for heteroduplex; 9, marker for location of nick. As noted in the legend to Fig. 2, the strong band in the lower panel results from permanganate oxidation of the covalently continuous strand near the site of the strand break in the open strand.

FIG. 4. Helix opening does not occur near the mismatch early in the repair reaction.
Reactions with the 5Ј-G-T heteroduplex containing a nick in the C strand at the HincII site were performed as described under "Experimental Procedures" and the legend to Fig. 2. DNAs were cleaved with HgiAI prior to electrophoresis, and indirect end-labeling with oligonucleotides V5470 or C5470 was used to visualize permanganate reactive sites. Upper panel, permanganate reactive sites on the C strand; lower panel, permanganate reactive sites on the V strand. Lanes: 1, complete system, 50 msec reaction; 2, complete system, 1 s reaction; 3, complete system, 5 s reaction; 4, marker for location of nick; 5, marker for location of mismatch.
preferentially onto single-strand regions within otherwise duplex DNA (21). Consequently, it was possible that a directional unwinding preference was masked to some degree by secondary events in which the helicase loaded onto single-stranded DNA produced by orientation-dependent unwinding from the strand break. This possibility was assessed in several ways with potential unwinding preference estimated by summing integrated band intensities of oxidation products produced on the continuous heteroduplex strand to either side of the nick.
As shown in Figs. 5 and 6 (upper panels), directional unwinding from the strand break toward the mismatch via the shorter path in the circular heteroduplex could be demonstrated with both 3Ј-and 5Ј-heteroduplexes, but the magnitude of preference for the shorter path decreased monotonically with increasing helicase II concentration. A decrease in preferential unwinding along the shorter path with both substrates also occurred as reaction time increased (Figs. 5 and 6, lower panels), and whereas heteroduplex unwinding from the strand break did not require SSB, presence of the protein conferred a modest increase in directional unwinding (not shown). MutS and MutL are therefore not only sufficient to activate helicase II unwinding from the strand break of an incised heteroduplex, but they also confer directionality on this process. DISCUSSION A single hemimodified d(GATC) sequence, which may reside to either side of the mismatch, is sufficient to provide strand specificity to heteroduplex repair by the E. coli methyl-directed pathway (8,22), with mismatch-provoked incision of the unmethylated strand of the d(GATC) site providing a strand break that directs removal of that portion of the new strand spanning the nick and the mispair (5-7). Analysis of reaction intermediates has suggested that excision initiates at the strand break by a mechanism in which helicase II displacement renders the incised strand sensitive to an appropriate 3Ј 3 5Ј or 5Ј 3 3Ј single-strand exonuclease, depending on location of the nick 3Ј or 5Ј to the mispair (7). The experiments described here are compatible with this mechanism and demonstrate that MutS and MutL are sufficient to coordinate mismatch recognition to activation of helicase II unwinding at a singlestrand break that can be located 800 -1,000 bp from the mismatch.
As noted previously (7,8), the bidirectional nature of the methyl-directed system requires loading of the appropriate hydrolytic activity at the incised d(GATC) sequence to ensure that excision proceeds toward the mispair. The finding that helicase II activation by MutS and MutL results in a significant degree of orientation-dependent unwinding on a nicked heteroduplex suggests that the latter proteins are sufficient to evaluate placement of the strand break 3Ј or 5Ј to the mismatch. Despite their separation distance, interaction of the mismatch and the strand break during the course of this reaction is fast FIG. 5. Helicase II unwinding is biased toward the shorter path between the nick and mismatch with a 3-heteroduplex. Complete reactions utilized a 3Ј-G-T heteroduplex containing a nick in the V strand at the MboI site (Fig. 1) and were performed by a modification of the procedure described under "Experimental Procedures" and in the legend to Fig. 2. Helicase II, which was omitted from the DNA syringe, was present in the syringe containing SSB and ATP, and incubation was for 1 s at 37°C prior to permanganate quench. Unwinding bias to either side of the strand break was estimated by summing radiolabel present in oxidized species produced to either side of the strand break on the continuous C strand (e.g. the species evident in the lower panels of Figs. 2 and 3 that migrate below or above the position of the strand break). Error bars are Ϯ 1 S.D. Upper panel, helicase II was varied as indicated. Reaction time was 1 s. Lower panel, SSB and helicase II were present at 8 and 0.6 g, respectively, and reaction time was varied as indicated.

FIG. 6. Helicase II unwinding is biased toward the shorter path between the nick and mismatch with a 5-heteroduplex.
Complete reactions, which contained a 5Ј-G-T heteroduplex with a nick in the C strand at the HincII site (Fig. 1), were performed as described under "Experimental Procedures" and in the legend to Fig. 5. SSB was present at 8 g. Upper panel, helicase II was varied as indicated, and reactions were stopped after 350 ms. Lower panel, helicase II was present at 0.6 g, and reaction time was varied as shown.
with maximal initiation of unwinding achieved after 1-5 s under conditions of MutS, MutL, and helicase II excess. Although the molecular events responsible for interaction of the two DNA sites are not fully understood, recent electron microscopy experiments (4) have suggested that MutS translocation along the heteroduplex contour may play a role in this process. Whereas the MutS dimer initially binds to heteroduplex DNA at the mismatch, this complex is converted in the presence of ATP to an ␣-shaped DNA structure that is stabilized by MutS at the base. The mismatch in such complexes is usually found in the DNA loop. This rearrangement has been attributed to a mechanism in which the two subunits of the MutS dimer act as ATP-driven divergent motors that translocate from the mispair in a bidirectional fashion along the helix contour. MutL stimulates this reaction and when present migrates along the helix with MutS. Under the buffer conditions used for the experiments described here, the rate of MutS-catalyzed formation of ␣-shaped DNA loops approaches 10,000 bp per min in the absence of MutL (4), sufficiently fast to account for the interaction of the two sites observed in the experiments described here. Fig. 7 illustrates a mechanism for MutS-and MutL-dependent activation of helicase II unwinding that incorporates these electron microscopy results, as well as the findings presented here and in the accompanying paper (9). Helicase II activation initiates by binding of a MutS dimer to the mismatch (4). MutL adds to the MutS⅐DNA complex in a reaction that requires ATP but apparently not ATP hydrolysis (3). Although MutL exists as a dimer in solution (3), the stoichiometry of MutL addition has not been established. In a reaction that depends on ATP hydrolysis, the subunits of the MutS dimer leave the mismatch, usually in a bidirectional manner, with MutL moving along the helix with MutS. At a stage in the reaction that remains to be determined, helicase II adds to MutS⅐MutL⅐DNA complex, and when a strand break is encountered the activity enters the helix in such a way that unwinding tends to proceed toward the mismatch, irrespective of placement of the nick 3Ј or 5Ј to the mispair on the incised strand. Since MutL greatly stimulates the activity of helicase II and since the two proteins interact physically (9,23), it is likely that MutL directly promotes initiation of unwinding by helicase, perhaps by physically facilitating helix entry of the activity at the strand break.
DNA helicases show a preferred polarity during initiation of unwinding (24). The orientation-dependent unwinding from the nick toward the mismatch described here can be understood in terms of this polarity preference, which for helicase II is 3Ј to 5Ј (25). Thus, one need only invoke loading of the unwinding activity onto the incised strand when the nick is 3Ј to the mismatch or onto the continuous strand when nick is 5Ј to the mispair.
The suggestion that MutL may have an important role in activating the excision step of bacterial mismatch repair may have implications for the eukaryotic reaction. The mammalian pathway has a mispair specificity similar to that of the bacterial reaction and occurs by a similar bidirectional mechanism. Defects in this system have been implicated in both inherited and sporadic cancers, as well as in cellular resistance to certain DNA damaging agents (26 -29). However, in contrast to the MutS and MutL homodimers that are active in the bacterial pathway, human mismatch repair is dependent on MutS␣, a heterodimer of the MutS homologs MSH2 and MSH6, and MutL␣, a heterodimer of the MutL homologs MLH1 and PMS2 (30 -32). Certain MLH1 and PMS2 mutations confer selective directional defects in mismatch repair. Thus, some MLH1 mutations are selectively defective in repair directed by a strand break located 3Ј to the mismatch but are proficient in mismatch correction directed by a 5Ј-strand signal (33). 3 Conversely, a PMS2 mutation has been identified that blocks repair from the 5Ј-side of the mismatch but not from the 3Ј-side (34). One interpretation of these findings is that like bacterial MutL, human MutL␣ functions to activate the mismatch repair excision system, but in the case of the human pathway the two subunits of MutL␣ differentially function to load a 3Ј to 5Ј or 5Ј to 3Ј excision system, depending on the location of the strand break that directs the reaction.