A Mutation in the MSH6 Subunit of the Saccharomyces cerevisiae MSH2-MSH6 Complex Disrupts Mismatch Recognition*

In yeast, MSH2 interacts with MSH6 to repair base pair mismatches and single nucleotide insertion/deletion mismatches and with MSH3 to recognize small loop insertion/deletion mismatches. We identified a msh6 mutation (msh6-F337A) that when overexpressed in wild type strains conferred a defect in both MSH2-MSH6- and MSH2-MSH3-dependent mismatch repair pathways. Genetic analysis suggested that this phenotype was due to msh6-F337A sequestering MSH2 and preventing it from interacting with MSH3 and MSH6. In UV cross-linking, filter binding, and gel retardation assays, the MSH2-msh6-F337A complex displayed a mismatch recognition defect. These observations, in conjunction with ATPase and dissociation rate analysis, suggested that MSH2-msh6-F337A formed an unproductive complex that was unable to stably bind to mismatch DNA.

DNA mismatches can arise through DNA replication errors, physical damage, and heteroduplex formation during genetic recombination. If left unrepaired, these mismatches become fixed in the genome as mutations. The best understood mismatch repair system is the Escherichia coli mutHLS long-patch repair pathway (1). A model for the initiation of mismatch repair by MutH, MutL, and MutS immediately after passage of a DNA replication fork has been developed based upon in vivo studies and an in vitro mismatch repair system reconstituted from purified components (1)(2)(3)(4)(5)(6). In this model, a mismatch is first recognized and bound by a dimer of MutS that displays an intrinsic ATPase activity. In a reaction that requires ATP, a dimer of MutL binds to MutS and then activates the methylation sensitive endonuclease MutH. Activation of MutH results in cleavage of the unmethylated DNA strand at hemimethylated d(GATC) sites that are transiently present after replication fork passage, providing an entry point for excision and replication proteins to remove the mismatch and repair the resulting DNA gap using the parental DNA strand as a template.
The ability of MutS to both recognize base pair mismatches and trigger downstream events that can be located several kilobases away from a mismatch site suggests that it can bind to DNA in at least two different modes. The first mode allows mismatch recognition, and the second mode allows MutS protein to use the energy of ATP hydrolysis to translocate along DNA with MutL so that it can activate MutH at GATC sites (1,7,8). Support for the presence of multiple MutS DNA binding modes was initially obtained from DNA binding assays showing that MutS protein can specifically recognize base pair mismatches in the absence of ATP and that the addition of ATP resulted in the loss of mismatch binding specificity (8). An elegant electron microscopy analysis performed by the Griffith laboratory (7) and the Modrich laboratory (9) provides further evidence for this idea. They showed that MutS can form ATPdependent loop structures on DNA substrates that contain a mismatch site. The size of loop was dependent on incubation time and the presence of a mismatch site. They hypothesized that MutS can bind to a mismatch substrate in an ATP independent step. After recognition, a second binding mode is activated through an ATP-dependent conformational change in MutS resulting in the loss of its mismatch recognition functions and a shift into a mode that allows it to bidirectionally translocate along DNA away from a mismatch site. This activity results in the formation of loop structures that are thought to serve as topological intermediates for strand excision and allow MutS to scan along DNA until it identifies and activates MutH bound at GATC sites.
At present, little is known about which domains in MutS are important for mismatch recognition. Recently, a DNA crosslinking analysis performed by Malkov et al. (10) revealed that the phenylalanine 39 residue in Thermus aquaticus MutS was critical for photocross-linking of MutS to a mismatch substrate. They also showed that a mutant derivative of MutS, mutS-F39A, displayed a reduced affinity for mismatch substrate. Although this study identified a domain that is important for mismatch recognition, it did not address whether this domain functions in general DNA binding and/or in mismatch binding and whether it functions during one or more DNA binding modes.
Whereas studies in bacteria have greatly increased our understanding of mismatch repair, analogous repair systems in eukaryotes appear to be more complex (1,11,12). Several homologs of mutS and mutL have been identified in eukaryotic organisms. A feature of the MutS homologs is they all contain a highly conserved ATP binding domain. In the yeast Saccharomyces cerevisiae, six mutS homologs and four mutL homologs have been identified, and the gene products of the MSH2, MSH3, MSH6, MLH1, and PMS1 genes have been identified as components of nuclear mismatch repair. Several studies have suggested that during mismatch recognition, MSH2 acts a scaffold for mismatch binding, whereas MSH6 and MSH3 act as specificity factors (13)(14)(15)(16)(17). This hypothesis was based on genetic studies showing that msh2 mutants are defective in the repair of both base pair and loop insertion/deletion mismatches, whereas msh3 mutants are principally defective in * 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 U.S.C. Section 1734 solely to indicate this fact.
‡ These authors contributed equally to this work. Supported by National Institutes of Health predoctoral training grants.
§ the repair of 2-4-nucleotide loop mismatches and msh6 mutants are principally defective in the repair of base-base and single nucleotide insertion/deletion mismatches (14,16,18). Interestingly, MSH3 and MSH6 functions have been shown to overlap, as msh3 and msh6 single mutants do not display as strong a defect in mismatch repair in their respective mismatch repair assays as do msh2 mutants. msh3 msh6 double mutants, however, display a mutator phenotype that is identical to that observed in msh2 mutants (14,16). Subsequent biochemical studies have shown that MSH2 can interact with MSH3 and MSH6 to form distinct complexes that bind subsets of mismatches predicted by the genetic studies and that an MLH1-PMS1 heterodimer can interact with these complexes when they are bound to a mismatch site (13, 17, 19 -22).
In this study, we used genetic and biochemical assays to investigate the mismatch recognition properties of MSH2-MSH6. Based on the pioneering studies performed by Malkov et al. (10), we created site-specific mutations in putative DNA binding domains in the MSH2 and MSH6 subunits and examined the effect of these mutations in UV cross-linking and DNA binding studies. Our analysis, described below, is consistent with MSH2 acting primarily as a scaffold for interactions with MSH6 and MSH3 that confer specificity to mismatch recognition. Interestingly, a mutation in MSH6 (msh6-F337A) was identified that conferred a dominant negative phenotype that was consistent with msh6-F337A sequestering MSH2 from functioning in mismatch recognition. This analysis also suggested that subunit interactions, mismatch recognition functions, and ATPase activities are coordinated to allow for the recognition and repair of DNA mismatches.
Media, Reagents, and Chemicals-E. coli strains were grown in LB broth or on LB agar that was supplemented with 100 g/ml ampicillin when required (29). Yeast strains were grown in either YPD or minimal selective media (30). 2% glucose, 2% sucrose, 3% glycerol, 2% lactate, and 2% galactose were included as carbon sources as indicated. When required, canavanine (Sigma) was included in minimal selective media lacking arginine at 60 mg/liter. 5-FOA (U. S. Biological, San Antonio, TX) plates were prepared as described (30). Polyclonal antibodies raised against MSH2 and MSH6 were obtained as described previously (15).
Nucleic Acid Techniques-Site-directed mutagenesis of MSH2 and MSH6 to create the msh2-Y42A, msh6-F337A, msh6-G335D, and msh6D343A alleles was performed using overlapping polymerase chain reaction mutagenesis (31). Oligonucleotide synthesis and doublestranded DNA sequencing of the entire subcloned fragments used to make the msh2 and msh6 mutant alleles were performed at the Cornell Biotechnology Analytical-Synthesis Facility (Ithaca, NY). All restriction endonucleases, T4 polynucleotide kinase, T4 DNA ligase, T4 DNA polymerase, and Vent polymerase were from New England Biolabs (Beverly, MA) and used according to the manufacturers' specifications. Plasmid DNA was isolated by alkaline lysis, and all DNA manipulations were performed as described previously (32).
Biochemical Techniques-Overexpression and purification of MSH2-MSH6 and msh2-Y42A-MSH6 was performed as described previously (13,15). MSH2-msh6-F337A and MSH2-msh6-F337A/G987D were purified using the same procedure with the exception that these purifications required loading and washing the ssDNA cellulose column at 100 mM NaCl Buffer A (25 mM Tris, pH 7.5, 1 mM EDTA, 10 mM ␤-mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride) instead of the 200 mM salt conditions used to load and wash the MSH2-MSH6 complex. In cell extracts obtained from overexpression strains, the levels of soluble wild type and mutant MSH2-MSH6 proteins were indistinguishable and were found to be equivalent to 3% of total protein. This was determined in SDS-PAGE by directly comparing the intensity of MSH2 and MSH6 bands in crude extracts to known amounts of purified complex. Through Western blot analysis, we determined that the number of MSH2 monomers per wild type mid-log cell was approximately 400; a 700-fold increase in this level was observed in strains that overexpressed MSH2 using the GAL10 promoter. 2 The purity of protein preparations was monitored by SDS-PAGE (8% gels) (33) (Fig. 3A). The approximate molecular weight of the MSH2-MSH6 and MSH2-msh6-F337A was determined by measuring the time required to elute polypeptides from a Superose 6HR gel filtration column as described previously (13). This analysis was kindly performed by William Enslow at the Cornell Biotechnology Analytical-Synthesis Facility.
UV Cross-linking-In the UV cross-linking studies, 120 nM (1.5 g) MSH2-MSH6 or MSH2-msh6-F337A was incubated in 50-l reactions for 15 min at 30°C with 20 nM 32 P-labeled duplex DNA substrate in buffer containing 25 mM Tris, pH 7.5, 0.01 mM EDTA, 0.1 mM dithiothreitol, and 2.0 mM MgCl 2 . The immunoprecipitation reactions presented in Fig. 1B and the cross-linking reactions presented in Fig. 4A were performed in the absence of MgCl 2 . Samples were then transferred to polystyrene tubes and irradiated in a Rayonet Photochemical reactor, model RPR100 (Branford, CT) equipped with 350 nm bulbs for 0 -60 min in the absence of visible light. After UV irradiation, samples were immediately boiled for 3 min in SDS-PAGE sample buffer and loaded onto an 8% SDS-PAGE gel for analysis by gel electrophoresis. Following 1 The abbreviations used are: 5-FOA, 5-fluoroorotic acid; hMSH, human MSH; PAGE, polyacrylamide gel electrophoresis; 5-IdUrd, iododeoxyuridine; ssDNA, single-stranded DNA. 2 T. Sokolsky and E. Alani, unpublished data. electrophoresis, the SDS-PAGE gel was stained with Coomassie Blue, dried, and visualized using a phosphorimager screen. Immunoprecipitations of denatured cross-linked samples were carried out using methods described by Iaccarino et al. (17). 10 l of cross-linked protein-DNA samples were boiled for 5 min in SDS sample buffer (New England Biolabs) and then added at a 10% final concentration to reactions containing anti-MSH2, anti-MSH6, or no antibody suspended in 0.1 M NaCl Buffer A. After a 45-min incubation at 4°C, 30 l of protein A-Sepharose were added to each reaction followed by another 45-min incubation. Protein A beads were spun down at 2500 rpm for 45 s. Supernatant was removed from the beads, which were then washed with 200 l of 0.1 M NaCl Buffer A. Beads were spun again, supernatant removed, and the beads were resuspended in SDS sample buffer, boiled for 5 min, and loaded onto SDS-PAGE gels. All gels were visualized using a phosphorimager screen (Molecular Dynamics), and bands were quantified using Mac IQ Imagequant software. DNA Binding, Dissociation, and ATPase Assays-DNA filter binding assays were performed as described previously (25). The standard buffer for the DNA binding assay contained 25 mM Tris, pH 7.5, 0.1 mM dithiothreitol, 0.01 mM EDTA, and 40 g/ml bovine serum albumin. When indicated, ATP was added to a final concentration of 1.5-2.0 mM. Binding was performed at 30°C for 15 min in a 60-l reaction. This time point was chosen because maximal binding for MSH2-MSH6 to oligonucleotide substrate was observed at 15 min, and no change in binding was observed in incubations that ranged from 15 to 90 min. Each reaction contained 16.7 nM 32 P-labeled ϩ1 substrate, 0 -333 nM unlabeled ϩ1 or homoduplex competitor, and 0.15-1.5 g (10 -100 nM) of MSH2-MSH6, msh2-Y42A-MSH6, or MSH2-msh6-F337A. Following incubation, samples were analyzed by filter binding to KOH-treated nitrocellulose filters (Schleicher and Schuell, Keene, NH) (34) as described by Chi and Kolodner (35). Dissociation of wild type and mutant MSH2-MSH6-DNA complexes was examined in the standard DNA binding buffer conditions as described in the text.
Wild type and mutant MSH2-MSH6-DNA complexes were also examined in gel retardation assays. Binding was performed at 30°C for 15 min in 10-l reactions containing 360 nM (0.9 g) MSH2-MSH6 or MSH2-msh6-F337A, 10 nM 32 P-labeled ϩ1 substrate, 25 mM Tris, pH 7.5, and 0.1 mM dithiothreitol. Unlabeled ϩ1 and homoduplex substrate were added as competitor as indicated in Fig. 3. After incubation, an equal volume of buffer containing 50 mM Tris, pH 7.5, 3 mM EDTA, and 50% glycerol was added to the binding reaction, and the samples were loaded onto a 4% polyacrylamide-bis (29:1) gel. The gel was run at room temperature in 1ϫ Tris-Borate EDTA buffer (31) at 200 V, and the extent of binding was determined after the gel was dried using a phosphorimager (Molecular Dynamics) and analyzed with Mac IQ Imagequant software.
The affinity of MSH2-MSH6, msh2-Y42A-MSH6, and MSH2-msh6-F337A for ssDNA cellulose was measured using a minicolumn protein retention assay. Approximately 30 g of wild type and mutant complex were applied to a 100-l bed volume ssDNA cellulose column in Buffer A containing 0.1 M NaCl. Each of these columns was eluted with successive 10 column volumes of 1ϫ buffer containing 0.1, 0.25, 0.30, and 1.0 M NaCl buffer. The amount of protein eluted with each wash was determined using Bradford assays (36).

UV Cross-linking Indicates That Both MSH2 and MSH6 Are
Cross-linked to a Duplex Oligonucleotide at the Site of a Mismatch-Previous studies have led to the proposal that the MSH6 subunit of the S. cerevisiae MSH2-MSH6 complex, which is principally involved in repairing base-base and single insertion/deletion mismatches, acts as a mismatch recognition specificity factor for MSH2 (13)(14)(15). A similar proposal for the human mismatch repair proteins, based on UV cross-linking analysis of hMSH2-hMSH6 to mismatch substrates, was made by Iaccarino et al. (17), in which it was found that only the hMSH6 subunit could be cross-linked to a mismatch substrate. Recently, Malkov et al. (10) demonstrated that mismatch DNA substrates containing a iododeoxyuridine residue (5-IdUrd) at the site of the mismatch could be specifically cross-linked to T. aquaticus MutS at phenylalanine 39, which is located in a region that is highly conserved among MutS homologs. Malkov et al. (10) also showed that a substitution of alanine at this position (mutS-F39A) abolished binding of MutS to mismatch substrates. We performed a similar analysis to identify residues involved in yeast MSH2-MSH6 mismatch recognition. 38-mer homoduplex (AT-I) and single nucleotide loop substrates (⌬1-I) were synthesized that contained single 5-IdUrd substitutions at the indicated thymine residues (see under "Experimental Procedures"). In the ⌬1-I substrate, the 5-IdUrd substitution constitutes an extrahelical loop of one nucleotide (see under "Experimental Procedures"). In filter binding competition assays (see below), MSH2-MSH6 displayed a 10-fold binding specificity for the ⌬1-I substrate compared with the AT-I substrate: this specificity was similar to that observed with ϩ1 and homoduplex substrates tested previously (see Fig.  3A and data not shown).
UV-dependent cross-linking of mismatch and homoduplex substrates to the MSH2-MSH6 complex was observed by incubating 32 P-labeled ⌬1-I or AT-I substrate with MSH2-MSH6 under standard binding conditions ( Fig. 1; see under "Experimental Procedures"). After incubation, samples were irradiated with UV light for 0 -60 min, resolved by SDS-PAGE, and then exposed to film to detect radioactively labeled protein-DNA complexes. As shown in Fig. 1A, in binding reactions involving either the ⌬1-I or the AT-I substrates, a major band migrating at ϳ180 kDa (Band B) was observed that was dependent on MSH2-MSH6 and UV exposure and migrated in SDS-PAGE slightly above the position of MSH6 (142 kDa). The size of the labeled complex was consistent with a ⌬1-I or AT-I substrate (ϳ30 kDa) cross-linked to MSH6 (142 kDa). At UV exposures 15 min and longer, a band migrating at greater than 205 kDa (Band A) was observed, as well as a faint band migrating at ϳ116 kDa (Band C). At 30 min of UV exposure, a cross-linking efficiency of 17% (as measured by densitometry) was observed in the ⌬1-I reactions; this cross-linking efficiency was 7-fold higher than that found in reactions involving the AT-I substrate.
Immunoprecipitation analysis was performed on denatured cross-linked species to determine whether Bands A-C contained MSH2 or MSH6. As shown in Fig. 1B, Band B was recognized and immunoprecipitated only by MSH6 antibody (lane 2), and Band C was immunoprecipitated only by MSH2 antibody (lane 3). Band A was recognized by both MSH2-and MSH6-specific antibodies, suggesting that this species contained cross-links that physically joined all three components of the MSH2-MSH6-⌬1-I complex. It is important to note that in addition to Bands A-C, lower intensity bands were observed in the cross-linking reactions shown in Fig. 1A; based on immunoprecipitation analysis and the mobility of these bands relative to Bands A-C in SDS-PAGE, we hypothesize that they represent complexes containing photocleavage products of the 5-IdUrd DNA substrates.
The msh6-F337A Mutation Conferred a Mismatch Repair Defect-We constructed alanine substitutions at residue 42 in MSH2 (msh2-Y42A) and residue 337 in MSH6 (msh6-F337A) based on the fact that these residues can be aligned with the T. aquaticus MutS Phe-39 residue ( Fig. 2A) (10). The msh2-Y42A and msh6-F337A alleles were cloned into GAL10, 2 overexpression plasmids; galactose induction studies indicated that each mutant was overexpressed to levels similar to that observed for the corresponding wild type proteins ( Fig. 2B and data not shown). The effect of these mutations was then assessed for defects in mismatch repair.
The msh2-Y42A and msh6-F337A alleles were tested to de-termine whether they can complement the mutator phenotype exhibited by msh2⌬ and msh6⌬ strains, respectively (14). In an assay that assesses mutation rate (principally base-base and single nucleotide insertion/deletion mutations) through the acquisition of resistance to the arginine analog canavanine, msh2⌬ strains exhibit a 13-20-fold higher mutation rate than either wild type strains or msh2⌬ strains that overexpress MSH2 (14,38,39). The msh2-Y42A allele fully complemented the msh2⌬ mutator phenotype when its protein product was expressed in both GAL10 promoter inducing (galactose) and repressing (glucose) conditions (  (15) analyzed a mutation in the MSH6 ATP binding domain (msh6-G987D) that severely inhibited the ATPase activity of the MSH2-MSH6 complex. When overexpressed in wild type strains, the msh6-G987D allele conferred a weak dominant negative mutator phenotype that was similar to the phenotype found in msh6⌬ strains, as it was limited to defects in the MSH2-MSH6 repair pathway (Table I) (15). In contrast, when the msh6-F337A allele was similarly overexpressed in wild type strains, a mutator phenotype was observed that was similar to that observed in msh2⌬ strains. In the canavanine assay, wild type strains overexpressing msh6-F337A displayed a 21-fold higher mutation rate than strains lacking this plasmid and a 5-fold higher rate than msh6⌬ strains or wild type strains that overexpressed msh6-G987D (Table I). In DNA slippage assays (26) that detect 2-4 insertion/deletion mismatches formed in an in-frame (TG) 16 T cassette placed within the URA3 gene, msh2⌬ strains exhibited an approximately 17-fold higher rate of slippage events, as measured by resistance to 5-fluoroorotic acid, than wild type. msh6⌬ strains or a wild type strain overexpressing MSH6 displayed DNA slippage frequencies that were indistinguishable from wild type. However, strains overexpressing msh6-F337A displayed a DNA slippage rate that was similar to that observed in msh2⌬ strains (Table I). It is important to note that the overexpression of all msh6 single and double mutant protein combinations in this study was similar to that observed for overexpression of wild type MSH6 (data not shown).
MSH2 Overexpression Suppresses the Dominant Negative Phenotype Conferred by Overexpressing msh6-F337A-As described above, the msh6-F337A mutation conferred a dominant negative phenotype that was similar to the mutator phenotype observed in msh2⌬ strains. We hypothesized that this resulted from msh6-F337A sequestering MSH2 in an inactive complex that was prevented from interacting with either MSH6 or MSH3 subunits. We tested this idea by performing canavanine mutator assays on wild type strains overexpressing the msh6-F337A, MSH2, and MSH6 subunits individually or in combination. As shown in Table I, co-overexpression of MSH2 or MSH6 with msh6-F337A fully suppressed the dominant negative phenotype observed in strains that only overexpressed msh6-F337A, consistent with the idea that MSH2 was sequestered from interacting with MSH3 or MSH6. This experiment, however, did not rule out the possibility that excess MSH2 suppressed the dominant negative phenotype conferred by the  To test this, mutator assays were performed in a msh6⌬ strain that overexpressed both MSH2 and msh6-F337A. As shown in Table I, this strain displayed a mutation rate that was similar to that observed in msh6⌬ strains, indicating that overexpression of MSH2 cannot suppress the msh6-F337A mismatch repair defect to achieve wild type mismatch repair function.
Analysis of msh2-Y42A-MSH6 and MSH2-msh6-F337A Complexes-MSH2-msh6-F337A and msh2-Y42A-MSH6 complexes were purified as described in Fig. 2B and under "Experimental Procedures." The complexes each contained a subunit stoichiometry that was similar to the wild type complex, suggesting that the site-specific mutations did not compromise interactions between subunits or disrupt the overall structure of the mutant proteins. Approximately the same yield of purified protein per unit of weight of induced cells was observed for each of the wild type and mutant heterodimers, and coimmunoprecipitation and gel filtration analysis indicated that the mutant complexes behaved similarly to wild type in these assays and formed stable heterodimers (data not shown) (13,15). In ATPase assays performed in the absence of DNA, the ATPase activities of MSH2-msh6-F337A and MSH2-MSH6 were identical (see Fig. 5) (15) (K m and V max were identical; data not shown). The msh2-Y42A-MSH6 complex was indistin-guishable from wild type in all genetic and biochemical assays described in this report (data not shown).
We obtained evidence that the MSH2-msh6-F337A complex was defective in DNA binding during the purification of the complex. Under conditions where the wild type complex was retained on a ssDNA cellulose column (200 mM NaCl), the MSH2-msh6-F337A complex was not retained. The purification procedure for the MSH2-msh6-F337A complex was then modified by loading MSH2-msh6-F337A fractions onto the ssDNA column at 100 mM NaCl (see under "Experimental Procedures"). To better quantify the binding of MSH2-MSH6 and MSH2-msh6-F337A complexes to DNA containing an insertion mismatch or homoduplex DNA, filter binding assays were performed with ϩ1 and homoduplex DNA substrates ( Fig. 2C; Refs. 13 and 33; see under "Experimental Procedures"). The 37-mer homoduplex substrate is identical to the ϩ1 substrate with the exception that the ϩ1 substrate contains an adenine insertion after base pair 15. In this assay, increasing amounts of protein were added to a constant amount (16.7 nM) of 32 Plabeled ϩ1 or homoduplex DNA. As seen in Fig. 2C, MSH2-MSH6 binding to ϩ1 substrate was maximal at 3.6 pmol (60 nM) of protein with nearly 50% of total counts bound. In contrast, less than 10% of total counts were bound in a binding reaction containing 60 nM MSH2-msh6-F337A complex and either homoduplex or ϩ1 substrate. In addition to a general DNA binding defect, the MSH2-msh6-F337A complex appeared to display a defect in discriminating between ϩ1 and homoduplex substrates. At 0.6 pmol (10 nM) of complex, 33 and 17% of ϩ1 and homoduplex substrate, respectively, were bound to MSH2-MSH6 complex. In contrast, 8.5 and 7.5% of ϩ1 and homoduplex substrate, respectively, were bound by 3.6 pmol (60 nM) of MSH2-msh6-F337A complex (Fig. 2C). It is important to note that DNA binding in all of the reactions was inhibited at high protein concentrations because the NaCl present in the protein preparations inhibited DNA binding when present at final concentrations greater than 50 mM; this is illustrated by the fact that the percentage of binding for all of the reactions reached a plateau at protein levels greater than 3.6 pmol and started to decrease at levels greater than 7.2 pmol ( Fig. 2C and data not shown).
MSH2-msh6-F337A Is Defective in Both Mismatch Binding and in General DNA Binding-To further explore the mismatch recognition properties of wild type and mutant complexes, the DNA binding activity of MSH2-MSH6 and MSH2-msh6-F337A were analyzed in filter binding and gel retardation competition assays. In the filter binding assays (Fig. 3, A and B) 20 nM (0.3 g) MSH2-MSH6 or MSH2-msh6-F337A complex was incubated with 16.7 nM 32 P-labeled ϩ1 substrate in the presence and absence of various concentrations of unlabeled ϩ1 or homoduplex competitor. In these reactions the molar ratio of protein to DNA substrate was approximately 1:1 and was chosen based on the binding titrations shown in Fig. 2C. Mismatch binding between the two competitors was measured by determining the maximal horizontal separation between the binding curves (13,33). As shown for the MSH2-MSH6 complex in Fig. 3A, the ϩ1 substrate is an approximately 8-fold more effective competitor than homoduplex substrate; this result is consistent with that found previously (13,15). In competitions involving the MSH2-msh6-F337A complex (Fig. 3B), the ϩ1 and homoduplex competitions were similar and suggested a dramatic decrease in mismatch binding specificity compared with the MSH2-MSH6 complex. A loss in mismatch binding specificity was also observed in MSH2-MSH6 binding experiments performed in the presence of 2.0 mM ATP; ATP addition was previously shown to eliminate mismatch recognition (13) (data not shown).
In gel retardation assays, 360 nM (0.9 g) MSH2-MSH6 or MSH2-msh6-F337A complex was incubated with 10 nM 32 Plabeled ϩ1 substrate in the presence and absence of various concentrations of unlabeled ϩ1 or homoduplex competitor substrates. As shown in Fig. 3, D and E, in the absence of unlabeled competitor, MSH2-MSH6 completely retarded the mobility of the 32 P-labeled ϩ1 substrate; in competition reactions, the ϩ1 unlabeled substrate was an approximately 20-fold more effective competitor than homoduplex substrate. In reactions involving MSH2-msh6-F337A, a discrete gel shift was not observed; instead, a diffuse gel shift pattern was observed that spanned from the position of unbound 32 P-labeled ϩ1 substrate to the position of the MSH2-MSH6-ϩ1 complex (Fig. 3D).
ATP Enhances UV Cross-linking and Binding of MSH2-msh6-F337A to DNA-UV DNA cross-linking studies were performed to examine in greater detail the mismatch binding properties of the wild type and MSH2-msh6-F337A complexes. These experiments were carried out as described under "Experimental Procedures." 24 nM (0.3 g) MSH2-MSH6 or MSH2-msh6-F337A was incubated with 20 nM 32 P-labeled ⌬1-I DNA at 30°C for 15 min, UV-irradiated for 15 min, electrophoresed on 8% SDS-PAGE gels, and then quantified by using a phosphorimager (see under "Experimental Procedures"). As shown in Fig. 4A, the MSH2-msh6-F337A complex displayed a 6-fold lower level of cross-linking to the ⌬1-I substrate compared with MSH2-MSH6 (Fig. 4A, lanes 1 and 2). When 1.5 mM ATP was included in the binding reaction prior to UV cross-linking, a 2.7-fold reduction in the level of MSH2-MSH6 cross-linking to the ⌬1-I substrate was observed (lane 4). However, in the MSH2-msh6-F337A reactions, ATP enhanced cross-linking of the complex to the ⌬1-I substrate by 2.5-fold (Fig. 4A, lanes 1  and 3).
In filter binding assays, we compared the effect of 1.5 mM ATP on the binding of ϩ1 and homoduplex DNA by the MSH2-TABLE I Average rates of spontaneous mutations and DNA slippage events in msh2⌬, msh6⌬, and wild type strains bearing the msh6-Y42A, msh6-F337A, and msh6-G987D alleles on GAL10, 2 plasmids msh2⌬ (EAY 281), msh6⌬ (EAY338), and wild type (FY23) strains were transformed individually or in pairs with pEAE51 (pGAL10-MSH6), pEAE20 or pEAE86 (pGAL10-MSH2), pEAE91 (pGAL10-msh2-Y42A), pEAE84 (pGAL10-msh6-G987D), and pEAE88 (pGAL10-msh6-F337A) and tested for forward mutations to can r . These strains were also transformed with pEAA69 ((TG) 16 T-URA3, ARS, CEN) and, where indicated, tested in the (TG) 16 T tract alteration assay described under "Experimental Procedures." All assays presented were performed using media containing 2% galactose, 2% sucrose as a carbon source. None refers to pRS control plasmids (26) that lacked MSH2 or MSH6 sequences. The median frequency was determined from 11 independent colonies for each experiment, and the mutation rate per cell generation was determined as described by Lea and Coulson (28). The average rate of two to six independent experiments for each strain is presented. Note that overexpressed MSH6 confers a weak dominant negative phenotype in the canavanine resistance assay and only partially complements the msh6⌬ mutator phenotype (15). This partial complementation and weak dominance phenotype was not observed in strains that expressed the GAL10-MSH6 plasmid at lower levels (15 MSH6 and MSH2-msh6-F337A complexes (Fig. 4B). In reactions containing 16.7 nM ϩ1 substrate and 20 nM (0.3 g) MSH2-MSH6, 34% of total 32 P-labeled ϩ1 substrate was bound; in identical reactions containing MSH2-msh6-F337A, 7% total binding was observed. The addition of ATP to the MSH2-MSH6 ϩ1 binding reaction resulted in a significant reduction in binding from 34 to 25% (Fig. 4B). In reactions involving MSH2-msh6-F337A, the addition of ATP resulted in an increase in ϩ1 substrate binding from 7 to 14%. In binding reactions involving homoduplex DNA that were performed under the same conditions, the addition of ATP to either the MSH2-MSH6 or MSH2-msh6-F337A reactions resulted in both cases in an increase in homoduplex binding (Fig. 4B). These experiments were performed at several different ATP concentrations with qualitatively similar results (data not shown). Taken together, the filter binding results are consistent with those obtained in the UV-cross-linking experiments. Mismatch Substrates Specifically Stimulate the ATPase Activity of MSH2-msh6-F337A While Inhibiting the ATPase Activity of MSH2-MSH6 -Previous analysis of the yeast MSH2-MSH6 complex indicated that under low salt conditions (Ͻ50 mM NaCl), the ATPase activity of MSH2-MSH6 was reduced in the presence of homoduplex DNA and reduced even further in the presence of ϩ1 substrate (15,39) (Fig. 5). Genetic analysis of the msh6-F337A mutation and UV cross-linking analysis of the MSH2-msh6-F337A complex encouraged us to examine the effect of DNA substrate on the ATPase activity of the MSH2-msh6-F337A complex. As described earlier, in the absence of DNA, the ATPase activity of MSH2-msh6-F337A was identical to that observed for the MSH2-MSH6 complex (Fig. 5B). However, unlike the MSH2-MSH6 complex, ϩ1 substrate addition (167 nM) to the MSH2-msh6-F337A ATPase reaction resulted in a nearly 4-fold increase in its ATPase activity (Fig. 5). The addition at 167 nM single-stranded oligonucleotide or a mismatch substrate (ϩ2) that is not recognized by the MSH2-MSH6 complex (13) did not affect the ATPase activity of the MSH2-msh6-F337A complex, suggesting that the enhanced ATPase activity was not the result of a contaminating activity such as that conferred by a helicase (data not shown).
The stimulation of the MSH2-msh6-F337A ATPase activity by the ϩ1 substrate was further investigated in experiments performed at a constant concentration of ATP (33 M) and varying amounts of ϩ1 or homoduplex substrate. As shown in Fig. 5C, at all concentrations tested, the ϩ1 substrate displayed a stronger inhibition of MSH2-MSH6 ATPase activity than the homoduplex substrate, and increasing concentrations of either DNA substrate resulted in a gradual reduction in ATPase activity (Fig. 5). In contrast, increasing concentrations of homoduplex DNA had no effect on the ATPase activity of the MSH2-msh6-F337A complex; however, the ϩ1 substrate displayed a concentration-dependent effect on ATPase activity. At low ϩ1 substrate concentrations, the ϩ1 substrate stimulated the ATPase activity of the MSH2-msh6-F337A complex with maximal stimulation observed at 167 nM ϩ1 substrate. At ϩ1 substrate concentrations greater than 167 nM, the ATPase activity progressively decreased; at 660 nM ϩ1 substrate, the ATPase activity of the MSH2-msh6-F337A complex was inhibited to a level below that observed with homoduplex substrate. Studamire et al. (15) showed that the ATPase activity of complexes defective in the MSH2 ATPase was still modulated by mismatch and homoduplex substrates; this modulation, however, was not observed in complexes deficient in the MSH6 ATPase activity. To test whether the enhancement of the MSH2-msh6-F337A ATPase activity by ϩ1 substrate was due to the ATPase activity of the msh6-F337A subunit, we purified a complex containing both the msh6-F337A mutation and a msh6 mutation (msh6-G987D) that was shown previously to severely inhibit the MSH6 ATPase activity (15). MSH2-msh6-F337A/G987D complex was purified as was described for MSH2-msh6-F337A. A similar yield, purity, and subunit ratio were obtained for purified MSH2-msh6-F337A/G987D compared with MSH2-MSH6 and MSH2-msh6-F337A (data not shown). We also found that the ratio of MSH2 to MSH6 in immunoprecipitations involving MSH2-specific antibody and crude extracts was indistinguishable for wild type and MSH2-msh6-F337A/G987D complexes (13) (data not shown). In DNA binding assays, the MSH2-msh6-F337A/G987D complex displayed a defect in mismatch binding that was indistinguishable from that observed for the MSH2-msh6-F337A complex (Fig. 3C).
The ATPase activity of the MSH2-msh6-F337A/G987D complex was tested in the presence and absence of DNA substrate (Fig. 6). In the absence of DNA substrate, the ATPase activity of the MSH2-msh6-F337A/G987D complex was less than half of that observed for the MSH2-msh6-F337A complex but was similar to that observed for the MSH2-msh6-G987D complex characterized previously (15). The ATPase activity of the MSH2-msh6-F337A/G987D complex was unaffected by the presence of either homoduplex or ϩ1 substrate; this result is consistent with the MSH6 ATPase playing an important role in the enhancement of the MSH2-msh6-F337A ATPase by ϩ1 substrate and also supports the idea that the MSH6 ATPase plays a critical role in signaling mismatch recognition (15).
Dissociation of the MSH2-MSH6 and MSH2-msh6-F337A-Oligonucleotide Complexes-The finding that MSH2-msh6-F337A was defective in mismatch binding but was capable of detecting mismatch substrates as demonstrated by mismatchdependent changes in its ATPase activity suggested that it was incapable of forming stable complexes with mismatch substrates. We tested this idea by performing the dissociation experiments presented in Fig. 7. In these studies, 5.7 nM (0.3 g) MSH2-MSH6 and MSH2-msh6-F337A were preincubated with either 1.4 nM 32 P-labeled homoduplex or ϩ1 substrate for 15 min, after which a 300-fold excess of unlabeled competitor was added. The stability of the original protein-DNA complexes was then measured as a function of time in filter binding assays.
For the MSH2-MSH6-homoduplex complex, a rapid decay was observed immediately after unlabeled homoduplex substrate was added. As shown in Fig. 7A, approximately 70% of the MSH2-MSH6-homoduplex complexes dissociated within 1 min after unlabeled homoduplex was added. For the MSH2-MSH6-ϩ1 complex, the addition of unlabeled ϩ1 substrate resulted in a biphasic decay, indicating the presence of two types of complexes. One complex was unstable, with a t1 ⁄2 of ϳ 5 min, and the other was stable, with a t1 ⁄2 of ϳ 80 min (Fig. 7A). These results indicated that ϳ40% of the MSH2-MSH6-ϩ1 complexes formed stable complexes. In experiments involving MSH2-msh6-F337A, similar dissociation profiles were observed for the MSH2-msh6-F337A-homoduplex and MSH2-msh6-F337A-ϩ1 complexes, and the dissociation of these complexes was more rapid than was observed for the MSH2-MSH6homoduplex complex (Fig. 7B). The inclusion of 2.0 mM ATP during the MSH2-MSH6-ϩ1 preincubation phase resulted in a dissociation profile that was similar to that found for the MSH2-MSH6-homoduplex preincubated in the presence or absence of ATP (Fig. 7C). The addition of ATP during the formation of MSH2-msh6-F337A-homoduplex and MSH2-msh6-F337A-ϩ1 complexes had no effect on the dissociation of these complexes (data not shown).
The finding that MSH2-msh6-F337A complexed with ϩ1 or homoduplex substrate displayed a dissociation profile that was even more rapid than was observed for MSH2-MSH6 provides an explanation for why we were able to detect MSH2-msh6-F337A-ϩ1 complexes in filter binding but were unable to observe discrete complexes in the gel shift assay. We hypothesize that the MSH2-msh6-F337A-ϩ1 complexes were detected in filter binding because these unstable complexes have a greater chance to be recovered on nitrocellulose filters, as the filter binding step was rapid and was performed in standard DNA binding buffer (see under "Experimental Procedures"). We believe that these unstable complexes were not observed in the gel shift assay because the gel electrophoresis step provides an extended time period to allow for dissociation of the complex. This argument can also explain why a greater mismatch binding specificity for MSH2-MSH6 was observed in the gel shift assay compared with the filter binding assay. An important caveat in our hypothesis is that unlike gel retardation analysis, in which specific protein-DNA complexes can be identified, it is difficult to determine in filter binding whether the complexes that were formed resulted from specific protein-DNA interactions or resulted from nonspecific protein-DNA aggregates.

The msh6-F337A Mutation Confers a Mismatch Recognition
Defect-We analyzed mutations in MSH2 (msh2-Y42A) and MSH6 (msh6-F337A) that are analogous to a mutation in T. aquaticus MutS that disrupts binding to mismatch DNA (10). Genetic analysis indicated that the msh6-F337A but not the msh2-Y42A mutation conferred a mismatch repair defect. In UV cross-linking and DNA binding assays that involved titration analysis and substrate competitions, MSH2-msh6-F337A displayed defects in both general and mismatch-specific DNA binding (Figs. 2 and 3).
Additional support for the idea that the msh6-F337A mutation conferred a mismatch binding defect came from experiments that examined the effect of ATP on the mismatch binding activity of wild type and MSH2-msh6-F337A complexes. In one set of experiments that involved measuring the ATPase activity of complexes in the presence of DNA substrate, we observed that in contrast to wild type complex, the ATPase activity of MSH2-msh6-F337A was specifically stimulated by ϩ1 substrate (Fig. 5). However, as the concentration of ϩ1 substrate was raised above 167 nM, the ATPase activity of the MSH2-msh6-F337A complex decreased and eventually reached a level that was below that observed in the absence of ϩ1 substrate. Homoduplex DNA did not affect ATPase activity, suggesting that msh6-F337A confers specific defects in mismatch recognition that can be overcome in the presence of a sufficiently high concentration of ϩ1 substrate, leading to a wild type response (decrease) in ATPase activity.
In a second set of experiments, the binding of MSH2-msh6-F337A to single nucleotide insertion/deletion (ϩ1, ⌬1-I) and homoduplex substrates was measured in UV cross-linking and in filter binding assays performed in the presence and absence of ATP (Fig. 4). Previous studies indicated that the specific binding of MutS homolog complexes to mismatch substrates in vitro was abolished by ATP (8,13,19,41,42). This loss of mismatch binding specificity in the presence of ATP most likely reflects a switch from a recognition mode to a translocation mode (7,9). In contrast, the binding of MSH2-msh6-F337A complex to both homoduplex and single nucleotide insertion/ deletion substrates was improved in the presence of ATP, and approached one-half of the level observed for wild type MSH2-MSH6 (Fig. 4). Based on the two binding mode model for MutS described in the Introduction, we propose that the mutant complex is defective in mismatch recognition but can still function in a second mode that allows it to bind and/or translocate along DNA in an ATP-dependent fashion. It is important to note that in the presence of ATP, the MSH2-msh6-F337A complex did not reach the level of binding to homoduplex DNA that was observed for MSH2-MSH6. This finding can be reconciled in the context of a two binding mode model if the initial loading of the MSH2-MSH6 complex onto DNA is accomplished primarily through the first binding mode, the disruption of which would be expected to reduce both general DNA binding and mismatch recognition. At present it is unclear whether the ATP-dependent binding of the MSH2-msh6-F337A complex is mediated by residues in the conserved domain presented in Fig. 2A or is located elsewhere.
A Model for MSH2-MSH6 Binding to Base Pair Mismatches-The observation that MSH2-MSH6 forms unstable and stable complexes with DNA provides an explanation for how the modest selectivity of the complex for mismatches observed in filter binding assays could result in efficient repair. Based on the off rate studies presented in Fig. 7, we propose that MSH2-MSH6 initially binds to DNA and forms an unstable and nonspecific complex that rapidly dissociates, allowing a new cycle of binding to occur. A proportion of the complexes that form are stable, and the formation of these stable complexes requires the presence of mispaired DNA in the substrate. The stability of the MSH2-MSH6-ϩ1 complex in dissociation experiments suggests that MSH2-MSH6 undergoes a change in binding conformation when it is bound to a mismatch substrate; this results in the stable binding of MSH2-MSH6 to a mismatch and provides a recognition signal for interactions with repair proteins that are involved in subsequent steps. Such a mechanism could amplify the selectivity of the complex for mispaired bases (24,33). A similar binding mode has been demonstrated for the binding of MSH1, MSH2, and hMSH2-hMSH6 to mismatch DNA (9, 24, 33). We propose that because the MSH2-msh6-F337A complex is unable to form a stable complex on mismatch DNA, it fails to provide a mismatch recognition signal for interactions with mismatch repair proteins and instead undergoes futile cycles of mismatch binding and release that involves ATP hydrolysis by the MSH6 subunit.
The above model for mismatch binding does not address the genetic data presented in Table I that showed that strains overexpressing msh6-F337A displayed a strong dominant negative phenotype in both the mutator and DNA slippage assays. Overexpression of MSH6, on the other hand, did not dramatically affect mismatch repair in either assay. As shown in the biochemical assays presented in this paper, the MSH2-msh6-F337A complex was unable to stably bind to mismatch DNA substrates. A hypothesis that can account for these genetic and biochemical observations is that in wild type cells, MSH2 is able to switch between MSH3 and MSH6 subunits when present in a mismatch binding competent complex. In such a scenario, the presence of a high level of one MSH partner would not prevent interactions with a partner present at lower levels. Based on this idea, we suggest that the inability of MSH2-msh6-F337A to stably bind to DNA prevents MSH2 within the mutant complex from switching between MSH3 and MSH6 subunits and thus withdraws MSH2 from pools available for repair. Experiments to test this idea are in progress.
Does MSH2 Contribute to Mismatch Binding?-The observation that the MSH2-msh6-F337A complex is defective in mismatch binding whereas its ATPase activity responds to the presence of a mismatch suggests that additional residues in the MSH2-MSH6 complex are involved in mismatch recognition; this idea is also supported by the finding that the mutant complex showed residual cross-linking to the ϩ1 substrate (Fig.  4). Does the entire mismatch recognition activity of the MSH2-MSH6 complex reside in the MSH6 subunit, or is MSH2 also involved? Previous analysis of MSH2 and MSH6 purified individually indicated that neither polypeptide demonstrated the mismatch binding specificity displayed by the complex (13,25). In cross-linking analysis, we observed a Band C at lower intensity that was shown by immunoprecipitation analysis to contain MSH2 (Figs. 1, 4). The intensity of this band could be increased in binding reactions that omitted magnesium chloride or in reactions that involved incubating MSH2-MSH6 complex with DNA substrates containing 5-IdUrd substitutions at different distances away from the mismatch site. 3 At present, these data cannot distinguish whether MSH2 is playing a role in general DNA binding or in recognition. In an initial attempt to investigate the putative mismatch binding domain shown in Fig. 2A, we made two substitutions in MSH6 (msh6-G335D and msh6-D343A) that map to residues that are in the vicinity of Phe-337 and are highly conserved in MutS homologs. Unfortunately, these substitutions resulted in the destabilization of the mutant msh6 proteins and prevented their purification (data not shown). We are currently analyzing the contribution of the MSH2 and MSH6 subunits in mismatch recognition by examining a set of msh2 and msh6 mutants that display genetic phenotypes consistent with a defect in mismatch binding.
The MSH6 ATPase Acts as a Signaling Factor in Mismatch Recognition-Previously, Studamire et al. (15) individually mutagenized the ATP binding domains of MSH2 and MSH6 and tested the mutant complexes for ATPase activity in the presence and absence of mismatch substrate. Their results supported a role for MSH6 as a mismatch recognition signaling factor, as complexes containing only a functional MSH6 ATPase could be modulated by mismatch binding in a manner qualitatively similar to that observed for the MSH2-MSH6 complex (Fig. 5). The ATPase activity of complexes that contained only a functional MSH2 ATPase was unchanged by the 3 T. Quach, J. Bowers, and E. Alani, unpublished data. ). In C, MSH2-MSH6 was preincubated with DNA substrate in the presence of 2.0 mM ATP. After preincubation, a 300-fold excess of the respective unlabeled substrate was added, and at various time points a 30-l aliquot was removed and the amount of 32 P-labeled substrate that remained bound was measured by filter binding. The percentage of binding is normalized to a control reaction in which unlabeled competitor was not added. The percentages of binding in the absence of competitor were as follows: MSH2-MSH6, 43% to ϩ1 and 26% to homoduplex; MSH2-msh6-F337A, 8.9% to ϩ1 and 9.3% to homoduplex; and MSH2-MSH6 in the presence of 2.0 mM ATP, 27% to ϩ1 and 30% to homoduplex. Background levels of binding were observed for both ϩ1 and homoduplex reactions when unlabeled competitor was included in the preincubation reaction. presence of DNA substrate (15). The data presented in this paper provide further support for this idea, as biochemical analysis showed that the enhanced ATPase activity of the MSH2-msh6-F337A complex in the presence of a mismatch was eliminated if a second mutation was introduced into the MSH6 subunit that inactivated the MSH6 ATPase (MSH2-msh6-F337A/G987D). We hypothesize that the MSH6-dependent mismatch signaling activity plays an important role in recruiting downstream mismatch repair factors, such as the MLH1-PMS1 complex (15). Recent data from the Prakash laboratory (22) are consistent with this idea, as they observed that ATP was required for the assembly of a ternary complex consisting of a DNA mismatch and the yeast MSH2-MSH6 and MLH1-PMS1 complexes.