Mismatch Recognition and DNA-dependent Stimulation of the ATPase Activity of hMutSα Is Abolished by a Single Mutation in the hMSH6 Subunit*

The most abundant mismatch binding factor in human cells, hMutSα, is a heterodimer of hMSH2 and hMSH6, two homologues of the bacterial MutS protein. The C-terminal portions of all MutS homologues contain an ATP binding motif and are highly conserved throughout evolution. Although the N termini are generally divergent, they too contain short conserved sequence elements. A phenylalanine → alanine substitution within one such motif, GXFY(X)5DA, has been shown to abolish the mismatch binding activity of the MutS protein ofThermus aquaticus (Malkov, V. A., Biswas, I., Camerini-Otero, R. D., and Hsieh, P. (1997) J. Biol. Chem. 272, 23811–23817). We introduced an identical mutation into one or both subunits of hMutSα. The Phe → Ala substitution in hMSH2 had no effect on the biological activity of the heterodimer. In contrast, the in vitro mismatch binding and mismatch repair functions of hMutSα were severely attenuated when the hMSH6 subunit was mutated. Moreover, this variant heterodimer also displayed a general DNA binding defect. Correspondingly, its ATPase activity could not be stimulated by either heteroduplex or homoduplex DNA. Thus the N-terminal portion of hMSH6 appears to impart on hMutSα not only the specificity for recognition and binding of mismatched substrates but also the ability to bind to homoduplex DNA.

The highly conserved C termini of MutS homologue proteins contain the ATP binding sites as well as helix-turn-helix motifs (20 -22). As nucleotide binding alters the conformation of the DNA-bound factors, such that they dissociate from the mismatch (22)(23)(24), and as helix-turn-helix motifs are frequently involved in DNA binding, it had been assumed that the mismatch recognition capability of these proteins resides in the C terminus. However, Malkov et al. (25) have shown that a short, highly conserved N-terminal motif with the consensus sequence GXFY(X) 5 DA (see Fig. 1) plays a crucial role in mismatch binding. They demonstrated that the MutS protein of Thermus aquaticus could be cross-linked to the heteroduplex substrate via the phenylalanine residue within this motif and that substitution of this residue with alanine effectively abolished mismatch binding. The overall structure of the protein, its ability to dimerize, and its ATPase activity were not impaired by this mutation (25). As both subunits of the human mismatch binding factor hMutS␣ possess the above consensus sequence (Fig. 1), it might be expected that both polypeptides contact DNA. However, our earlier experiments showed that only hMSH6 could be cross-linked to mismatch-containing DNA (20,26,27). These latter results invoke the asymmetric nature of the heterodimer and imply that the two subunits play distinct roles during mismatch recognition. Therefore, we introduced the Phe 3 Ala mutations into hMSH2 and hMSH6 and tested the effects of these changes on mismatch binding and mismatch repair. We showed that the F432A mutation in hMSH6 abolished not only mismatch recognition but also the binding to DNA in general, whereas the substitution of Phe-42 with alanine in hMSH2 was without effect. Our results are in general agreement with those of Alani and co-workers (28) who described a similar study with the MutS␣ heterodimer of S. cerevisiae. However, although the ATPase activity of the human factor was insensitive to the presence of both homo-and heteroduplex DNA, the S. cerevisiae MSH2-msh6 F337A heterodimer could be stimulated by a mismatch-containing oligonucleotide substrate at low salt concentrations. The possible reasons underlying this difference are discussed.

Site-directed Mutagenesis and Production of the Recombinant Baculovirus Vectors-
The conserved phenylalanine residues within hMSH2 and hMSH6 were changed to alanine by PCR-based site-directed mutagenesis as described previously (20,29).
Both plasmids were sequenced to ensure the integrity of the cDNA inserts and to confirm the presence of the mutations. The recombinant baculoviruses were produced using the Bac-to-Bac Baculovirus Expression System (Life Technologies, Inc.) according to the manufacturer's recommendations.
Production and Purification of the Recombinant Proteins-Co-infection of Spodoptera frugiperda 9 (Sf9) cells with different pairs of recombinant baculoviruses gave rise to the four possible heterodimers: hMSH2/hMSH6, hMSH2FA/hMSH6, hMSH2/hMSH6FA, and hMSH2-FA/hMSH6FA (hMutS␣, hMutS␣2FA, hMutS␣6FA, and hMutS␣2,6FA, respectively). After 72 h of incubation, the cells were harvested and the total protein extracts were prepared as described (26). The proteins were purified by fast protein liquid chromatography (FPLC) essentially as described (24) except that the heparin-Sepharose fractions containing hMutS␣ were first loaded on a Resource-S FPLC column. Even though the desired protein was in the flow-through, the latter matrix bound many contaminating polypeptides. The Resource-S flow-through was then loaded onto a Resource-Q column (all FPLC columns were from Amersham Pharmacia Biotech). The protein preparations were judged by 7.5% SDS-PAGE to be Ͼ95% pure (Fig. 2). Protein concentrations were estimated in triplicate by the Bradford assay as directed by the manufacturer (Bio-Rad) using bovine serum albumin as a standard.
Electrophoretic Mobility Shift Assays (EMSAs)-All EMSAs, including the determinations of the dissociation constants, were carried out as described previously (24); 100 ng of purified protein were incubated with 40 fmol of radioactively labeled DNA substrate (G/C, G/T, ϩ2) for 20 min at room temperature. The oligonucleotide duplexes were obtained by annealing the 32  To test the sensitivity of the DNA-protein complexes to ATP, the nucleotide was added to the reactions to a final concentration of 1 mM after 10 min of incubation time. Following electrophoresis on 6% non-denaturing polyacrylamide gels, the results were visualized on Biomax TM MR films (Kodak) and quantified using the ImageQuant v1.2 software (Molecular Dynamics). The determination of the dissociation constants was performed as described (20).
ATPase Assays-The ATPase assays were carried out at 37°C in a 20-l mixture containing 20 mM Tris-HCl (pH 7.5), 5 mM MgCl 2 , 120 mM KCl, 1 mM dithiothreitol, 0.5 M (2.6 g, 10 pmol) purified recombinant protein, 4 M ATP, and 33 nM [␥-32 P]ATP. When the assays were carried out in the presence of DNA, the substrates (2 M homo-or heteroduplex DNA G/T) were added at the beginning of the incubation period, and the protein concentration was reduced to 0.1 M (0.52 g, 2 pmol) so as to make the stimulation effect more apparent. At selected time points, 2-l aliquots were removed, the reaction was stopped by the addition of 5 l of 80% formamide dye, and 2 l of this mixture were loaded on a 20% sequencing gel as described previously (27). The results were visualized on Biomax TM MR films (Kodak) and quantified using the ImageQuant v1.2 software (Molecular Dynamics). The hMutS␣2,6KR variant, which is defective in ATP binding and which was used in these assays as a negative control, was prepared as described previously (20). All assays were carried out in triplicate.
In Vitro Mismatch Repair Assays-M13mp2 heteroduplexes containing either a single G/T mispair (G/T) or a loop of two extrahelical nucleotides (ϩ2) were incubated with the cytoplasmic cell extracts as described (20). The DNA was then purified, electroporated into a mutS strain of E. coli, and plated along with the ␣-complementation strain CSH50, isopropyl-1-thio-␤-D-galactopyranoside, and 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside (X-gal). If no repair occurred, mixed plaques were observed containing both blue and colorless progeny. Reduction in the percentage of mixed plaques and a concomitant increase in single-color plaques was indicative of repair. The recombinant hMutS␣ variants (0.18 g) were used to complement the cytoplasmic extracts (50 g) of the mismatch repair-deficient cell lines HCT15 (hMSH6 Ϫ/Ϫ ) and LoVo (hMSH2 Ϫ/Ϫ ). The repair reactions were allowed to proceed at 37°C for 18 min. Repair efficiency (%) ϭ 100 ϫ (1 Ϫ (% hMSH6 Point Mutation Abolishes Mismatch Recognition mixed plaques in extract-treated sample)/(% mixed plaques in extractuntreated sample)).

The Phe 3 Ala Mutations in hMSH2 and hMSH6 Do Not
Affect the Stability of hMutS␣-The conserved phenylalanine residues 42 and 432 in hMSH2 and hMSH6, respectively, were substituted with alanine by site-directed mutagenesis of the respective cDNAs. Baculovirus vectors expressing the mutated and wild type proteins were constructed and used to infect Sf9 cells. By co-infecting with all the possible combinations, four heterodimers were obtained: wild type hMutS␣, hMutS␣2FA, hMutS␣6FA, and hMutS␣2,6FA. Examination of the total protein extracts of the infected Sf9 cells by SDS-PAGE showed that all polypeptides were expressed with similar efficiency (Fig. 2, lanes Sf9 TE). During subsequent purification by FPLC, all four hMutS␣ variants behaved identically and were obtained in similar yields; this suggests that the mutations did not bring about any significant alterations in the structural properties of the polypeptides. As shown in Fig. 2 (lanes purified hMutS␣), all the mutated polypeptides migrated through SDS-PAGE gels with velocities identical to the respective wild type proteins, and the hMSH2/hMSH6 stoichiometry of the four hMutS␣ variants was also similar to the wild type heterodimer. Because the stability of hMSH6 depends on its interaction with hMSH2 (14,30), these results indicate that the mutant polypeptides were not significantly altered structurally and were able to interact with their cognate partners to form stable heterodimeric factors.
Mutation of Phenylalanine 432 of hMSH6 to Alanine Abolishes Mismatch Recognition-To test whether the Phe 3 Ala mutations in hMSH2 and hMSH6 affected the specific binding of the purified recombinant hMutS␣ variants to base-base mismatches and small IDLs, EMSAs were carried out using radioactively labeled, double-stranded 34-mer homo-or heteroduplex oligonucleotides. As shown in Fig. 3A, wild type hMutS␣ and the hMutS␣2FA variant formed specific protein-DNA complexes with substrates carrying either a G/T mismatch or a loop of two extrahelical thymidines. In contrast, the hMutS␣6FA and hMutS␣2,6FA factors failed to form stable complexes with these substrates under identical conditions. Similar results were obtained when a substrate carrying a 1-nucleotide loop was used (data not shown). The binding affinities of the wild type hMutS␣ and the hMutS␣2FA variant for the G/T-containing substrate were comparable, with both apparent dissociation constants being ϳ4 nM (data not shown, for details see "Experimental Procedures"). This value agrees well with that reported previously (20). The data showed that substitution of alanine for phenylalanine in the putative mismatch binding motif of hMSH6 abolished the mismatch recognition ability of hMutS␣, although an identical mutation in hMSH2 was without effect in the mismatch binding assays. This finding helps explain the results of our earlier experiments where only hMSH6 could be cross-linked to mismatched substrates (20,26,27). The fact that identical substitutions in the MSH2 and MSH6 subunits of MutS␣ of S. cerevisiae were described to have similar phenotypes (28) further underscores the conservation of function of eukaryotic MutS homologues (see also "Discussion").
The F432A Mutation in hMSH6 Abolishes the Binding of hMutS␣6FA and hMutS␣2,6FA to Homoduplex DNA-Wild type human and yeast MutS␣ is also able to bind homoduplex DNA albeit about one order of magnitude less efficiently than heteroduplex substrates (9,20,28,31). As the Phe 3 Ala mutation in hMSH6 abolished mismatch recognition, we were interested to see if this mutation affected also the nonspecific binding to homoduplex DNA. To this end, we modified the conditions of our standard EMSA so as to augment binding to the latter substrate. By reducing the amount of nonspecific competitor DNA 10-fold, we were able to detect the formation of protein-DNA complexes between the 34-mer homoduplex G/C and hMutS␣ or hMutS␣2FA. In contrast, no such complexes were noticeable when hMutS␣6FA or hMutS␣2,6FA were incubated with this substrate (Fig. 3B). This indicates that the F432A mutation in hMSH6 affects the binding of hMutS␣ not only to mismatch-containing substrates but also to DNA in general.
The Phe 3 Ala Mutations in hMSH2 and hMSH6 Do Not Affect the ATPase Activities of the Variant Heterodimers-In previous studies, the specific DNA-protein complex formed by the mismatched substrate and hMutS␣ could be shown to be sensitive to ATP (20,22,23). The binding of the nucleotide to the so-called Walker type A motifs at the C termini of the MutS homologues was shown to induce distinct conformational changes, which lead to the dissociation of the heterodimer from the substrate (22,24). To ensure that the F42A mutation had not affected the ability of hMutS␣2FA to dissociate from mismatch-containing DNA substrates in the presence of the nucleotide, we added ATP to the EMSA mixtures after a 10-min incubation. As shown in Fig. 4, both factors, the wild type hMutS␣ and the hMutS␣2FA mutant, dissociated from the

hMSH6 Point Mutation Abolishes Mismatch Recognition
substrates to a similar extent upon nucleotide addition. Therefore, we conclude that the constitution of the hMutS␣2FA-DNA complex, as well as its sensitivity to ATP, are unaffected by the F42A substitution in hMSH2.
Because the hMutS␣6FA and hMutS␣2,6FA variants failed to bind mismatch-containing DNA, we could not examine their ATP binding and/or hydrolysis properties by EMSA. To test the unlikely possibility that the F432A mutation in hMSH6 altered the DNA binding properties of this variant through a structural change in the ATP binding domain, we carried out ATPase assays. As shown in Fig. 5, the kinetics of ATP hydrolysis by hMutS␣ and hMutS␣6FA were comparable and clearly distinguishable from those of the hMutS␣2,6KR variant defective in ATP binding (20). This suggests that the Phe 3 Ala mutation in hMSH6 affected only mismatch and DNA binding but not the ATPase activity of the hMutS␣6FA heterodimer.
The ATPase Activity of the hMutS␣6FA and hMutS␣2,6FA Variants Is Insensitive to DNA-Though weak, the ATPase activity of MutS␣ is essential for its function in MMR as shown by the phenotype of hMutS␣ variants in which ATP binding was diminished through Lys 3 Arg mutations in the Walker type A ATP binding motifs (20,21). Previous studies showed that the ATPase activity of MutS homologues could be stimulated by homoduplex and heteroduplex DNA albeit to different extents (28,33,34). We wanted to test whether this stimulatory effect was absent in hMutS␣ variants bearing mutations that affect heteroduplex and homoduplex recognition.
The purified heterodimers were incubated with [␥-32 P]ATP either in the presence or in the absence of DNA. In these experiments, five times lower protein concentrations (0.1 M as opposed to 0.5 M used in the experiments described in the paragraph above) were used to make the stimulation effect more apparent. As shown in Fig. 6, the ATPase activities of all four hMutS␣ variants were comparable in the absence of DNA. The Phe 3 Ala substitutions thus did not affect the DNAindependent ATPase activity. Addition of homoduplex DNA to the reactions resulted in a 3-fold stimulation of ATP hydrolysis by hMutS␣ or hMutS␣2FA, whereas heteroduplex DNA stimulated the ATP hydrolysis by these factors approximately 7-fold (Fig. 6, A and B, respectively). In contrast, the ATPase activity of hMutS␣6FA or hMutS␣2,6FA was unaffected by both homoduplex and heteroduplex DNA (Fig. 6, C and D,  respectively). These results differ somewhat from those of Alani and co-workers (28) who reported that the ATPase activity of the yeast heterodimer MSH2-msh6 F337A (equivalent of hMutS␣6FA) was stimulated by heteroduplex DNA. This effect is hard to explain as the yeast heterodimer, similarly to its human counterpart studied here, demonstrated no mismatch binding activity. However, as the stimulation of the yeast factor was seen only at low salt concentrations, this effect may be of questionable relevance (see also "Discussion"). The human factor differs from its yeast homologue in that no stimulation of the hMutS␣6FA ATPase was observed under a range of different salt concentrations. 2 The F432A Mutation in hMSH6 Abolishes Mismatch Repair in Vitro-In an attempt to test the functionality of the four hMutS␣ variants, we used the recombinant proteins in in vitro MMR assays. To this end, we used extracts of MMR-deficient lines that lack hMutS␣: HCT15 that carries truncating mutations in both alleles of hMSH6 and LoVo, carrying a homozygous deletion in hMSH2. As anticipated, purified recombinant hMutS␣ or hMutS␣2FA fully complemented the MMR deficiency of the LoVo extracts on circular heteroduplex substrates with either a single G/T mismatch or a 2-nucleotide IDL. Similarly, the G/T repair efficiency was restored in HCT15 extracts by these recombinant proteins; the ϩ2 substrate was not tested in HCT15 as these extracts contain a full complement of hMutS␤ and are thus proficient in loop repair. In contrast, addition of hMutS␣6FA or hMutS␣2,6FA had no significant effect on the MMR efficiency of these extracts (Fig. 7). DISCUSSION Although the eukaryotic mismatch binding factor MutS␣ is a heterodimer of two highly conserved polypeptides, it binds to mismatch-containing substrates in an asymmetric fashion. Our experiments with whole cell extracts (26) and later with the purified factor (27) have shown that only a single polypeptide of about 160-kDa molecular mass could be cross-linked to the DNA. We were subsequently able to show by immunoprecipitation of the cross-linked protein-DNA complexes that the polypeptide covalently bound to the mismatch-containing substrate was hMSH6 (20). When Malkov et al. (25) showed that the MutS protein of T. aquaticus was cross-linked to DNA via a phenylalanine residue in a highly conserved N-terminal motif (Fig. 1), we decided to examine this region in hMutS␣. As our [␥-32 P]ATP was incubated with the purified hMutS␣ variants at 37°C. Aliquots were taken after the indicated times, and the reactions were stopped and loaded onto a 20% denaturing polyacrylamide gel (see "Experimental Procedures"). The amounts of liberated inorganic phosphate ( 32 P i ) were compared with the total amount of radioactivity (sum of [␥-32 P]ATP and 32 P i ) in the respective assay by quantification of the intensity of the corresponding bands on an autoradiograph. Open circle, control reactions carried out without protein.
The experiment was carried out in triplicate. Error bars represent the standard deviations from the mean. present data show, mutation of this conserved phenylalanine to alanine in hMSH6 appears to have severely attenuated binding of the heterodimer to mismatch-containing substrates (Fig. 3A) as well as to homoduplex DNA (Fig. 3B). The fact that an identical mutation in hMSH2 was entirely without effect in our assays suggests that this latter polypeptide does not participate in mismatch recognition. It is interesting to note that the Phe 3 Ala mutation in hMSH6 did not affect the basal level of ATPase activity of the heterodimer, but that it abolished the ability of DNA, both homo-and heteroduplex, to stimulate ATP hydrolysis (Fig. 6, C and D). These results differ somewhat from those reported by Alani and co-workers (28) for the S. cerevisiae factor who reported that, at low salt concentrations, the ATPase activity of the yeast MSH2-msh6, F337A variant could be stimulated by heteroduplex DNA even though its binding to heteroduplex DNA was affected as severely as in the case of the human factor. As noted above, the ATPase activity of the human hMutS␣ variant hMSH2/hMSH6F432A could not be stimulated by DNA substrates, not even at low salt concentrations (data not shown). The partial proteolytic pattern of the F39A mutant of T. aquaticus MutS (Fig. 1) in the presence of ATP was recently described to be unaltered by the presence of heteroduplex DNA (35), which further strengthens the suggestion that this mutation abolishes mismatch binding. The anomalous behavior of the yeast mismatch binding factor cannot be easily explained. However, the latter factor does appear to differ somewhat from the human hMutS␣. In cross-linking experiments, both MSH2 and MSH6 could be seen to form a covalent complex with the heteroduplex DNA (28), whereas no such reaction was observed with the human protein (20). The yeast factor appears to bind mismatches containing 8-oxoguanine and act in their correction (36), whereas no recognition of the 8-oxoguanine/C or 8-oxoguanine/A-containing substrates was observed with the human heterodimer. 3 In addition, the mismatch binding motif of MSH6 contains a second phenylalanine residue at position 338 (Fig. 1), which might conceivably compensate, at least to some extent, for the loss of Phe-337 in heteroduplex recognition in the presence of ATP and low salt; the bacterial and the human polypeptides have a tyrosine at this site (Fig. 1). The finding that the mutation of only a single site in MutS could abolish mismatch recognition (25) was exciting because it helped identify the domain of the protein required for DNA binding. The fact that mutating this site in only the MSH6 subunit of the respective heterodimers MutS␣ (28) or hMutS␣ (present study) was sufficient to bring about the loss of binding to heteroduplex substrates served to accentuate the role of the latter polypeptide in mismatch recognition. However, neither experiment helped us to understand the role of this conserved site in binding of hMutS␣ to DNA. This 3 G. Crouse and J. Jiricny, unpublished data.

FIG. 6. Stimulation of ATPase activity of the recombinant hMutS␣ variants by homoduplex G/C (squares) and heteroduplex G/T (filled triangles) 34-mer oligonucleotides.
A, wild type hMutS␣; B, hMutS␣2FA; C, hMutS␣6FA; D, hMutS␣2,6FA. The assays were performed as described under "Experimental Procedures." Note that in these experiments five times less protein (0.1 M) was used than in the experiment shown in Fig. 5 to make the stimulation effect more apparent. The 10-min time point of reactions in which the ATP binding-deficient variant hMutS␣2,6KR (open triangles) was incubated with the heteroduplex substrate G/T was used as the negative control. Diamonds designate reactions carried out in the absence of DNA. The experiment was carried out in triplicate. Error bars represent the standard deviations from the mean.

FIG. 7. Complementation of protein extracts from mismatch repair-deficient cell lines with recombinant hMutS␣ variants in
in vitro mismatch repair assays. Circular substrates containing a single mismatch (G/T) or a 2-nucleotide IDL (ϩ2) were incubated with extracts of either HCT15 (hMSH6 Ϫ/Ϫ ) or LoVo (hMSH2 Ϫ/Ϫ ) cells, respectively. After an 18-min incubation at 37°C, the proportion of repaired substrate was determined as described previously (20) and depicted as % repair efficiency. no compl., no recombinant protein added to the extracts; wt, complementation with the wild type hMutS␣; 2FA, complementation with hMutS␣2FA; 6FA, complementation with hMutS␣6FA; 2,6FA, complementation with hMutS␣2,6FA. The figure shows the results of two independent experiments (deviations are shown by error bars). situation has changed with the resolution of the crystal structure of the bacterial MutS proteins from E. coli (37) and T. aquaticus (32). The structure of co-crystals of these proteins with DNA reveals that the phenylalanine residue in question is located in an ␣-helix that contacts the DNA in the vicinity of the structural distortion and that the phenylalanine residue partially intercalates into the DNA at the site of the mismatch. This explains its key role in mismatch recognition. However, the most exciting finding concerns the fact that the two subunits of the homodimeric MutS factor are not equivalent inasmuch as only one subunit contacts the substrate DNA via its N-terminal ␣-helix, whereas the equivalent helical motif of the second subunit is turned away from the DNA. Thus, the mismatch binding factor is a functional heterodimer already in prokaryotes. Extrapolation of these crystallographic data to the situation in eukaryotes suggests that the ␣-helix of MSH6 that contains the conserved phenylalanine residue mediates the mismatch contacts, whereas the equivalent motif of hMSH2 is turned away.
The MutS structures help to elucidate one further point. It appears that although much of the protein is unstructured in the absence of DNA, mismatch recognition induces the formation of a dimeric clamp on the substrate. The clamp formation might also be induced through protein contacts with homoduplex DNA, albeit less efficiently. In both cases, the initial protein-DNA contacts would be mediated by the ␣-helix of MSH6 that contains the conserved motif GXFY(X) 5 DA. A disruption of this region would thus result not only in the loss of mismatch recognition, but also in an inability to form a clamp. This would explain why a Phe 3 Ala mutation within this motif of MutS and MSH6 abolishes both specific and nonspecific DNA binding. The availability of the crystal structures of the bacterial MutS proteins makes this hypothesis testable by site-directed mutagenesis of other residues predicted to be required for protein-DNA complex formation.