Interaction between the Msh2 and Msh6 Nucleotide-binding Sites in the Saccharomyces cerevisiae Msh2-Msh6 Complex*

Indirect evidence has suggested that the Msh2-Msh6 mispair-binding complex undergoes conformational changes upon binding of ATP and mispairs, resulting in the formation of Msh2-Msh6 sliding clamps and licensing the formation of Msh2-Msh6-Mlh1-Pms1 ternary complexes. Here, we have studied eight mutant Msh2-Msh6 complexes with defective responses to nucleotide binding and/or mispair binding and used them to study the conformational changes required for sliding clamp formation and ternary complex assembly. ATP binding to the Msh6 nucleotide-binding site results in a conformational change that allows binding of ATP to the Msh2 nucleotide-binding site, although ATP binding to the two nucleotide-binding sites appears to be uncoupled in some mutant complexes. The formation of Msh2-Msh6-Mlh1-Pms1 ternary complexes requires ATP binding to only the Msh6 nucleotide-binding site, whereas the formation of Msh2-Msh6 sliding clamps requires ATP binding to both the Msh2 and Msh6 nucleotide-binding sites. In addition, the properties of the different mutant complexes suggest that distinct conformational states mediated by communication between the Msh2 and Msh6 nucleotide-binding sites are required for the formation of ternary complexes and sliding clamps.

Structures of truncated forms of bacterial MutS (2,6) and human Msh2-Msh6 (24) in complex with DNA mispairs have been determined by x-ray crystallography. MutS homodimers bind mispaired DNA and form asymmetric rings around the DNA in which only one subunit contacts the mispaired base (2,6,7,25,26). This basic structure is conserved in Msh2-Msh6, with Msh6 being the mispair-contacting subunit (24,(27)(28)(29). Consistent with known structures, challenging the mispairbound forms of MutS, Msh2-Msh6, and Msh2-Msh3 with ATP converts them to a sliding clamp form that slides freely along the DNA but is trapped on an end-blocked DNA (15,16,21,30,31). In contrast, the base pair-bound form undergoes direct dissociation from the DNA when challenged with ATP (21,32). Taken together, these observations suggest that mispair binding may license an ATP binding-dependent conformational change that results in conversion to the sliding form. Furthermore, only the mispair-bound forms of MutS and Msh2-Msh6 form ATP-dependent ternary complexes with MutL and Mlh1-Pms1, respectively, which facilitate subsequent steps in MMR (15-17, 21, 23, 33-35). Attempts to understand the ATPbound conformations of these complexes by soaking ATP and ATP analogs into the MutS-mispair crystals have thus far failed to induce conformational changes expected from the biochemical characterization of these complexes, possibly because of restraints placed on the proteins by the crystal lattice (2, 6, 24 -26). Thus, the nature of the ATP binding-induced conformational changes that link mispair recognition with conversion to the sliding form and/or the form that is competent for ternary complex formation is presently unknown.
The interactions between the bacterial and eukaryotic mispair-binding proteins and ATP are probably best understood for the Msh2-Msh6 complex (32). In the absence of DNA, the Msh6 nucleotide-binding site has high affinity for ATP and low affinity for ADP, whereas the Msh2 nucleotide-binding site has lower affinity for ATP and higher affinity for ADP. ATP binding at the Msh6 nucleotide-binding site results in reduced affinity for ADP in the Msh2 nucleotide-binding site. In the absence of mispair binding, the ATP in the Msh6 nucleotide-binding site is rapidly hydrolyzed, and the resulting ADP dissociates. Thus, in solution, Msh2-Msh6 binds and hydrolyzes ATP, resulting in a form that contains ADP bound in the Msh2 nucleotide-binding site (32). This ADP-bound form of Msh2-Msh6 can bind both base pairs and mispairs (21,29,30,32,36). Interestingly, prebinding of ATP under non-hydrolyzing conditions (absence of Mg 2ϩ ) or the presence of the non-hydrolyzable analog ATP␥S prevents binding of Msh2-Msh6 to any DNA (21,30,32,36). Once Msh2-Msh6 is bound to base-paired DNA, ATP binding causes Msh2-Msh6 to directly dissociate from DNA (21). In contrast, when Msh2-Msh6 is bound to a mispair, ATP hydrolysis at the Msh6 site is inhibited, which then favors binding of ATP at the Msh2 site (32,37,38). A similar dual ATPbound state can be achieved with ATP␥S (21,32). Binding of ATP or ATP␥S converts mispair-bound Msh2-Msh6 into the sliding clamp form (21,30,32,34). A mispair-bound, ATP-bound form of the mispair-binding complexes is also competent for assembling Mlh1-Pms1 ternary complexes (17,21,32), although it has not been definitively established whether the sliding conformation and the Mlh1-Pms1-binding conformation are the same.
In this study, we have used Saccharomyces cerevisiae to analyze eight MMR-deficient Msh2-Msh6 complexes for defects in ATP binding, sliding clamp formation, and Msh2-Msh6-Mlh1-Pms1 ternary complex assembly and for ATP binding-induced conformational changes. The results of this study have allowed us to directly detect conformational changes induced by ATP binding to the Msh6 nucleotide-binding site and to relate the ATP-binding properties of the Msh2 and Msh6 nucleotidebinding sites to conformational changes and the ability of Msh2-Msh6 to form sliding clamps and ternary complexes with Mlh1-Pms1.
Partial Proteolysis-To determine whether ATP binding affects the conformation of Msh2-Msh6, 1 g of wild-type or mutant Msh2-Msh6 was incubated with 0 -500 ng of trypsin with or without 100 M ATP␥S in reaction buffer (25 mM Tris, pH 8.0, 110 mM NaCl, 4 mM MgCl 2 , 0.01% IGEPAL, 2 mM dithiothreitol, 2% glycerol) in a final volume of 10 l for 1 h at room temperature. Reactions were stopped by the addition of phenylmethylsulfonyl fluoride to a final concentration of 10 mM. The proteolysis products were separated by SDS-PAGE on a 4 -15% gel (Bio-Rad), and the resulting gel was silver-stained. Western blotting using rabbit polyclonal antibodies to S. cerevisiae Msh2 or Msh6, which were raised in this laboratory, was performed to detect and differentiate Msh2 and Msh6.
UV Cross-linking Experiments-Nucleotide binding was measured by UV cross-linking exactly as described previously (32,44). ATP-binding reactions were performed in the absence of Mg 2ϩ , and ADP-binding reactions were performed in the presence of Mg 2ϩ .
Surface Plasmon Resonance Analysis-Experiments analyzing the interaction between the Msh2-Msh6 complex and mispaired DNA were performed with a Biacore T100 instrument (GE Healthcare) essentially as described previously (21,44,45) using an IAsys instrument and adapted for use with a Biacore T100 instrument. DNA substrates 236 bp in length with biotin conjugated to one end and a centrally located GT mispair were constructed as described (21,41,46). Approximately 20 ng of DNA (140 response units) was conjugated to a flow cell of a streptavidin-coated Biacore SA chip. In the unblocked sliding experiments, 50 nM wild-type or mutant Msh2-Msh6 and 250 M ATP were flowed over the chip at 20 l/min for 120 s in reaction buffer. In the end-blocked sliding experiments, 30 nM LacI was flowed over the chip at 20 l/min for 120 s in reaction buffer before and during Msh2-Msh6 binding. Dissociation of Msh2-Msh6 from the DNA end was induced by the addition of 1 mM isopropyl ␤-D-thiogalactopyranoside to the sample. The DNA-coated surface was regenerated with a 20-l pulse of 3 M NaCl. Data were analyzed with BIAevaluation 3.1 software. Reference subtraction was made from an unmodified flow cell. All experiments were performed at 25°C.
The formation of Msh2-Msh6-Mlh1-Pms1 ternary complexes on mispaired DNA was also analyzed with a Biacore T100 instrument essentially as described previously (21,44,45). The minus Mlh1-Pms1 control was performed similarly to the end-blocked sliding experiment described above;

Characterization of Mutations Affecting the Msh2 and Msh6
ATP-binding Regions-A previous study of three dominantnegative S. cerevisiae msh6 mutations, msh6-S1036P, msh6-G1067D, and msh6-G1142D, revealed that the resulting mutant Msh2-Msh6 complexes were proficient for binding mispaired bases in DNA but were not converted to the sliding clamp form upon ATP binding. This suggested either that they occluded mispairs from detection by functional Msh2-Msh6 and Msh2-Msh3 complexes or that they were unable to form appropriate ternary complexes with MutL homolog complexes (29,42,44). These mutations resulted in amino acid substitutions near the Msh2 and Msh6 nucleotide-binding sites but were not predicted to alter the structure of the nucleotide-binding sites. Here, we have extended our prior studies and have included five additional mutations in our analysis (Fig. 1). One of the new mutations is a dominant msh6 mutation, msh6-R1024C (47). The four other mutations are msh2 mutations, msh2-R730C, msh2-S742P, msh2-T773D, and msh2-G855D, which cause amino acid substitutions in Msh2 equivalent to those caused by the dominant msh6 mutations. These mutations cause amino acid substitutions near the Msh6 ATP-binding site (msh2-S742P, msh2-T773D, and msh6-G1142D), near the Msh2 ATPbinding site (msh6-S1036P, msh6-G1067D, and msh2-G855D), or at positions between the two ATP-binding sites (msh6-R1024C and msh2-R730C) (Fig. 1).
The mutation rates of strains harboring each of the eight mutant genes on low copy number plasmids were assessed qualitatively using patch tests. Plasmids containing the mutant msh2 alleles did not complement the high frameshift mutation rate of an msh2⌬ strain (supplemental Fig. S1), but they had no effect on the frameshift mutation rate of a wild-type strain (supplemental Fig. S2). These msh2 mutations therefore result in a strong loss of function but do not cause a dominant phenotype like that caused by the equivalent msh6 mutations (29).
Partial Proteolysis Reveals That Msh6 Undergoes Conformational Changes upon Binding ATP␥S-To monitor nucleotide-induced conformational changes in the Msh2-Msh6 complex, we incubated wildtype Msh2-Msh6 with increasing concentrations of trypsin in the presence or absence of ATP␥S in reactions containing Mg 2ϩ , and the digestion of Msh2-Msh6 was monitored by SDS-PAGE (Fig. 2) as well as by Western blotting with Msh2and Msh6-specific antibodies (supplemental Fig. S3). In the absence of ATP␥S, Msh6 was extensively degraded at low trypsin concentrations, whereas Msh2 was degraded only at higher trypsin concentrations (Ͼ100 ng/reaction). Msh6 degradation correlated with the appearance of three species of ϳ90, 80, and 75 kDa that were derived from Msh6 as indicated by Western blotting. As the trypsin concentration increased (Ն50 ng/reaction), the 90-and 80-kDa species were degraded, and the 75-kDa species remained until much higher trypsin concentrations (Ͼ150 ng/reaction). We have shown previously that the 90-kDa species corresponds to an N-terminal Msh6 truncation product that is missing the first 441 residues (43). Msh2 degradation correlated with the appearance of lower molecular weight species, many of which were detected with the anti-Msh2 antibody (supplemental Fig. S3).
The addition of 100 M ATP␥S in the presence of Mg 2ϩ resulted in protection of Msh2-Msh6 from digestion by trypsin in three ways (Fig. 2). First, at low trypsin concentrations, only the 90-kDa species was present, and at higher trypsin concentrations, both the 90-and 80-kDa Msh6 species were present and remained highly resistant to digestion with trypsin. Second, the 75-kDa Msh6 species, which is derived from the 90-and 80-kDa species, 3 was not formed. Finally, Msh2 remained intact at intermediate trypsin concentrations, and at higher trypsin concentrations, it was cleaved into fragments in the size range of 40 -50 kDa that remained highly resistant to degradation and were not seen at all in the absence of ATP␥S. Identical digestion patterns were observed with ATP␥S in the absence of Mg 2ϩ or when ATP␥S was replaced with ATP in the absence of Mg 2ϩ (data not shown).
ATP␥S titration experiments were performed to determine whether the observed changes in protease sensitivity resulted  from filling of only the high affinity ATP-binding site in Msh6 or from filling of both of the ATP-binding sites in the Msh2-Msh6 complex. At 0.5 M ATP␥S and 10 ng of trypsin, we observed substantial protection of Msh2 as well as protection of the 90-and 80-kDa Msh6 species and reduced formation of the 75-kDa Msh6 species. Maximal protection was seen in the presence of 1 M ATP␥S (Fig. 3), which correlates with filling of only the high affinity Msh6 ATP-binding site (32). Importantly, the level of protection did not increase further at higher ATP␥S concentrations (Fig. 3), suggesting that filling of the Msh2 site with ATP and consequentially any conformational changes required to convert Msh2-Msh6 to a sliding clamp are not revealed by this proteolysis assay. Partial proteolysis was also performed in the presence of Mg 2ϩ with mutant complexes in which the Msh6 ATPase site (Msh2-Msh6-G1142D, Msh2-S742P-Msh6, and Msh2-T773D-Msh6) or an intersite position (Msh2-R730C-Msh6) was affected (Fig. 2). The first three complexes were selected for analysis because assessment of ATP binding by UV cross-link-ing (see below) suggested that they were defective in ATP binding to the Msh6 nucleotide-binding site, and the latter complex was selected as a control. The digestion patterns of these complexes were the same as those of wild-type Msh2-Msh6. These mutants therefore undergo the same ATP␥S-induced conformational changes in the presence of 100 M ATP␥S and Mg 2ϩ as wildtype Msh2-Msh6; we also confirmed that 2 M ATP␥S induced protease protection of Msh2-Msh6-G1142D and Msh2-T773D-Msh6 (data not shown). However, this result seems inconsistent with the observation that the Msh2-Msh6-G1142D and Msh2-T773D-Msh6 complexes were defective in binding ATP in the Msh6 nucleotide-binding site when ATP binding was assessed by UV cross-linking (see below) ( Table 1). The proteolysis and cross-linking experiments differed in that the former was performed in the presence of Mg 2ϩ , whereas the latter was performed in the absence of Mg 2ϩ to prevent ATP hydrolysis. Therefore, to determine the role that Mg 2ϩ plays in the ATP␥S-induced conformational change in Msh2-Msh6, the partial proteolysis reactions were repeated in the absence of Mg 2ϩ (Fig. 4). The digestion pattern of the wild-type Msh2-Msh6 complex was not affected by the absence of Mg 2ϩ either with or without ATP␥S or by the absence of Mg 2ϩ when ATP was substituted for ATP␥S (data not shown). In contrast, the digestion patterns of the Msh2-Msh6-G1142D and Msh2-T773D-Msh6 complexes were different from those of wild-type Msh2-Msh6 in the absenceofMg 2ϩ .Here,thepresenceofMg 2ϩ resultedinATP␥Sdependent protection of the 90-and 80-kDa species, whereas in the absence of Mg 2ϩ the 90-and 80-kDa species were not protected by ATP␥S, and the 75-kDa species was formed (Fig. 4). In addition, the digestion pattern of the Msh2-Msh6-G1142D complex was the same when ATP was substituted for ATP␥S in the absence of Mg 2ϩ (data not shown). Thus, the Msh6 proteins in these mutant complexes require Mg 2ϩ to undergo the ATP␥S-induced conformational change, suggesting that the Msh6-G1142D and Msh2-T773D amino acid substitutions destabilize ATP binding, potentially through electrostatics, and therefore require Mg 2ϩ to stabilize binding of ATP to the Msh6 site. The Msh2-Msh6-S742P complex appeared to be partially protected from digestion by trypsin in the presence of ATP␥S and absence of Mg 2ϩ ( Fig. 4), consistent with its relatively small defect in ATP binding in the absence of Mg 2ϩ (Table 1).
Mutant Msh2-Msh6 Complexes Have Different ATP-binding Defects-To investigate the effects of the Msh2 and Msh6 amino acid substitutions on nucleotide binding at the high affinity Msh6 and low affinity Msh2 ATP-binding sites (32), we used UV cross-linking to determine the affinities of the eight different mutant Msh2-Msh6 protein complexes for ATP under non-hydrolyzing conditions (absence of Mg 2ϩ ) and for ADP in the presence of Mg 2ϩ (Table 1) as described previously (32,44). The intersite mutant complexes Msh2-R730C-Msh6 and Msh2-Msh6-R1024C bound ATP at the Msh2 and Msh6 sites with affinities that were similar to those of wild-type Msh2-Msh6. The amino acid substitutions affecting the Msh2 site (Msh2-G855D, Msh6-S1036P, and Msh6-G1067D) strongly reduced binding of ATP to the Msh2 site, but they did not affect binding of ATP to the Msh6 site. All five of these mutant complexes displayed wild-type affinities of binding ADP to the Msh2 nucleotide-binding site (Table 1).
In contrast, each of the three mutations affecting residues near the Msh6 ATP-binding site caused distinct defects. The Msh2-T773D substitution strongly reduced binding to the Msh6 ATP-binding site but did not affect binding to the Msh2 ATP-binding site. The Msh6-G1142D substitution caused a strong reduction in binding of ATP to both the Msh6 and Msh2 nucleotide-binding sites. Finally, the Msh2-S742P substitution caused only a small reduction in binding of ATP to the Msh6 nucleotide-binding site but caused a strong reduction in binding of ATP to the Msh2 nucleotide-binding site. It should be noted that the proteolysis experiments discussed above indi-cate that the ATP-binding defects at the Msh6 nucleotidebinding sites of the Msh2-Msh6-G1142D and Msh2-T773D-Msh6 complexes likely result from a Mg 2ϩ dependence of ATP binding by these mutants. Unfortunately, ATP binding cannot be assessed by UV cross-linking in the presence of Mg 2ϩ because the ATP is hydrolyzed too rapidly. The results with the msh2-T773D mutation indicate that ATP binding to the high affinity Msh6 ATP-binding site is not required for ATP binding to the lower affinity Msh2 site. However, the results with the msh6-G1142D and msh2-S742P mutations suggest that conformational changes linking the Msh6 and Msh2 ATP-binding sites can affect ATP binding by the Msh2 site. All three of these mutant complexes had wild-type affinities for binding of ADP to the Msh2 nucleotide-binding site (Table 1).
Mutant Msh2-Msh6 Complexes Are Defective in Sliding Clamp Formation-To monitor sliding clamp formation, we modified an established method to evaluate binding of Msh2-Msh6 to a DNA duplex containing a central mispaired base with or without a LacI protein-mediated end block for use with a Biacore T100 instrument (21,44,45). Briefly, wild-type Msh2-Msh6 was flowed over mismatched DNA that was immobilized at one end to a streptavidin-coated chip in the presence of ATP.

Msh2-S742P-Msh6
Msh6 a These data are the average of two independent experiments. b Affinities reported as Ͼ200 M were difficult to measure due to low signal-tobackground ratios.
After hydrolyzing the ATP to ADP, Msh2-Msh6 binds the mismatch (21,32,36). After exchange of ADP for ATP, wild-type Msh2-Msh6 forms a sliding clamp that slides off the DNA end (21,30,36,44), which gives rise to a moderate level of steadystate binding (Fig. 5). The lac operator site is present at the free end of the immobilized DNA. When LacI is added, it binds the operator site and acts as a block to prevent dissociation of Msh2-Msh6 from the DNA end, giving rise to a higher level of steady-state Msh2-Msh6 binding than in the absence of LacI (Fig. 5) (21). Upon the addition of isopropyl ␤-D-thiogalactopyranoside, LacI is released from the operator site, and the accumulated Msh2-Msh6 slides off of the free DNA end, resulting in reduced steady-state Msh2-Msh6 binding comparable with that seen in the absence of LacI (Fig. 5) (21). Like wild-type Msh2-Msh6, complexes containing each of the four Msh2 mutations or the Msh6-R1024C mutation all showed higher binding to the blocked mispaired DNA substrate compared with the blocked base-paired DNA substrate in the presence of ATP, indicating that these complexes retained specificity for mispaired DNA (supplemental Fig. S4). However, all of the mutant complexes displayed lower absolute levels of binding to both end-blocked and unblocked mispaired DNA substrates relative to wild-type Msh2-Msh6 in the presence of ATP ( Fig. 5 and supplemental Fig. S4). The mutant complexes also responded differently than the wild-type complex to the LacI end block (Fig. 5). The addition of LacI did not increase the steady-state levels of binding of these mutant complexes relative to those seen for binding to the unblocked substrate, and the addition of isopropyl ␤-D-thiogalactopyranoside to release the LacI end block did not cause significant dissociation of the mutant Msh2-Msh6 complexes from end-blocked mispaired substrates (Fig. 5). These results are consistent with the mutant complexes failing to undergo the ATP-induced conformational change to the sliding clamp form, as seen for the previously studied dominant mutant complexes containing the Msh6-S1036P, Msh6-G1067D, and Msh6-G1142D substitutions (42,44).
Msh2-Msh6-Mlh1-Pms1 Ternary Complex Formation Occurs at ATP Concentrations at Which Only the Msh6 Nucleotide-binding Site Binds ATP-Analysis of the ATP concentration dependence of Msh2-Msh6-Mlh1-Pms1 ternary complex formation is compromised by the fact that, as mispair-bound Msh2-Msh6 binds ATP, it slides off the mispair, allowing additional Msh2-Msh6 to bind the mispair (21). It is thus difficult to distinguish the initial binding of Mlh1-Pms1 to Msh2-Msh6 from increased binding of Msh2-Msh6 upon sliding clamp formation due to the addition of increasing concentrations of ATP. We therefore took advantage of the unique properties of the Msh2-T773D-Msh6 and Msh2-Msh6-G1142D complexes, which form ternary complexes with Mlh1-Pms1 in the presence of 250 M ATP but are defective for conversion of mispairbound Msh2-Msh6 to the sliding form. As established by the conformational changes probed by limited proteolysis, these mutant complexes bind ATP␥S in the presence of Mg 2ϩ in their Msh6 nucleotide-binding sites with a wild type-like affinity (wild-type Msh2-Msh6 saturated at 1 M). To determine the ATP-binding requirements for ternary complex formation, we repeated the above assay with a low concentration (2 M) of ATP. Wild-type Msh2-Msh6, Msh2-T773D-Msh6, and Msh2-Msh6-G1142D appeared to form ternary complexes with Mlh1-Pms1 in the presence of 2 M ATP (supplemental Fig. S5). Msh2-T773D-Msh6 and Msh2-Msh6-G1142D formed similar amounts of ternary complex at 2 and 250 M ATP (compare supplemental Fig. S5 with Fig. 6 and with Fig. 4 of Ref. 44). Wild-type Msh2-Msh6 formed less apparent ternary complex at 2 M ATP than at 250 M ATP, most likely because the higher concentration of ATP allows sliding clamp formation and thus additional Msh2-Msh6 binding (32). Although the precise affinities for ATP binding at the Msh6 nucleotide-binding sites of the Msh2-T773D-Msh6 and Msh2-Msh6-G1142D complexes in the presence of Mg 2ϩ are not known, proteolysis protection assays indicate that these sites are filled at 2 M ATP (data not shown). Because the affinity for ATP binding to the Msh2 nucleotide-binding site of the wild-type Msh2-Msh6 and Msh2-T773D-Msh6 complexes is much higher than 2 M ATP (the precise affinity for binding to the Msh2 nucleotide-binding site of the Msh2-G1142D-Msh6 complex is not known), it appears that only ATP binding to the Msh6 site is required for ternary complex formation.

DISCUSSION
The detailed biochemical analysis of eight MMR-deficient Msh2-Msh6 complexes with various defects in ATP binding, sliding clamp formation, and Msh2-Msh6-Mlh1-Pms1 ternary complex assembly (Table 2) supports several key conclusions. First, binding of ATP to the Msh6 nucleotide-binding site results in conformational changes that are communicated to the Msh2 ATP-binding site; however, binding of ATP to both sites can be uncoupled. Second, conversion of Msh2-Msh6 to the sliding clamp form appears to require ATP binding to both sites as well as communication between the two ATP-binding sites. Third, ternary complex formation appears to require ATP binding only at the Msh6 nucleotide-binding site. Finally, sliding clamp formation and ternary complex formation can be uncoupled.
In previous studies, it was observed that the N terminus of Msh6 in the Msh2-Msh6 complex was highly sensitive to digestion with trypsin (43). Digestion first resulted in the loss of the unstructured N-terminal 251 amino acid residues, followed by digestion up to residue 441, leaving a stable 90-kDa C-terminal core of Msh6 complexed with apparently intact Msh2. Here, we found that, at higher trypsin concentrations, the C-terminal Msh6 core was cleaved at two additional sites, resulting in 80and 75-kDa fragments. We have found that the 75-and 80-kDa species result from cleavage of an N-terminally truncated Msh6 species at two different sites between the Msh6 C terminus and the Msh6 ATP-binding site. 3 ATP and ATP␥S protected the 90-and 80-kDa species and prevented the formation of the 75-kDa fragment at concentrations expected to fill only the high affinity Msh6 nucleotide-binding site. These results support the view that binding of ATP in the Msh6 nucleotidebinding site induces a significant conformational change within the Msh2-Msh6 complex. The nature of this conformational change, suggested by examination of the crystal structure of the related Rad50 ATP-binding cassette ATPase in the ATP-bound form (48), is likely an open-to-closed transition of the Msh2-Msh6 ATPase domains relative to the ADP-and DNA-bound conformational states of the MutS and Msh2-Msh6 crystal structures (2, 6, 24 -26). Modeling this closed state ( Fig. 1 and  supplemental Fig. S6) revealed that the ATP-binding cassette ATPase domain "signature motif" of one subunit made conserved interactions with ATP bound in the other subunit and revealed that the observed trypsin digestion sites were protected.
Using the available crystal structures of MutS and human Msh2-Msh6 (2, 6, 24 -26), the local defects in the various symmetric mutants can be understood. The Msh2-S742P and Msh6-S1036P amino acid substitutions disrupt a key residue of the ATP-binding cassette ATPase domain signature motif that probably makes direct interactions with the ATP ␥-phosphate when ATP is bound in the opposite subunit. The defects in Msh2-S742P-Msh6 and Msh2-Msh6-S1036P are consistent with defects in stabilization of the closed conformation of the ATPase domains. The strong Mg 2ϩ dependence of ATP binding to the Msh6 nucleotide-binding site of the mutant Msh2-T773D-Msh6 and Msh2-Msh6-G1142D complexes is consistent with a simple model of destabilization of ATP binding in the Msh6 site by electrostatic repulsion. Despite this, Mg 2ϩ is unable to mask either the defects caused by the equivalent Msh2-G855D and Msh6-G1067D amino acid substitutions in ternary complex formation or the defects caused by all four amino acid substitutions in the formation of the sliding clamp, suggesting that these amino acid substitutions cause additional defects. Finally, the intersite amino acid substitutions Msh2-R730C and Msh6-R1024C appear to be positioned to play roles in stabilizing the closed conformation.
Closing of the ATPase domains induced by binding of ATP to the Msh6 nucleotide-binding site is also consistent with evidence for communication between the high affinity Msh6 and low affinity Msh2 nucleotide-binding sites. The Msh6-G1142D amino acid substitution results in a defect in binding of ATP at both nucleotide-binding sites in the absence of Mg 2ϩ even though this mutation does not alter an amino acid residue near the Msh2 nucleotide-binding site. The Msh2-S742P amino acid substitution, which disrupts a key residue of the signature motif predicted to interact with the ␥-phosphate of Msh6-bound ATP, also appears to disrupt communication between the two nucleotide-binding sites, as it causes only a slight decrease in the affinity for binding of ATP by Msh6 but causes a substantial defect in ATP binding at the Msh2 nucleotide-binding site. The properties of these two mutations suggest that communication between the Msh6 and Msh2 nucleotide-binding sites is required for filling the Msh2 site with ATP. Importantly, the reciprocal amino acid substitutions at the Msh2 ATPase site (Msh2-G855D and Msh6-S1036P) have no effect on ATP binding by Msh6. This result suggests that these amino acid substitutions are unlikely to disrupt ATP binding at the Msh2 site by preventing ATP binding to the Msh6 site. Instead, they likely prevent a conformational change in the ATPase domains initiated by binding of ATP to the high affinity Msh6 site, and this inhibits ATP binding to the low affinity Msh2 site. In contrast, the Msh2-T773D amino acid substitution, which affects the Msh6 nucleotide-binding site and causes a strong defect in ATP Defect classes are as follows: I, failure of the Msh2 nucleotide-binding site to bind ATP and disruption of conformational changes required for both ternary complex and sliding clamp formation; II, possibly reduced affinity of the Msh2 nucleotide-binding site for ATP and disruption of conformational changes required for sliding clamp formation; and III, disruption of conformational change required for sliding clamp formation.
binding at this site, does not cause a defect in ATP binding at the Msh2 nucleotide-binding site. This finding suggests that communication between the two nucleotide-binding sites can be uncoupled such that nucleotide binding at each site can occur independently.
The results presented here indicate that ATP concentrations sufficient for binding of ATP to the Msh6 nucleotide-binding site are sufficient for the formation of Msh2-Msh6-Mlh1-Pms1 ternary complexes. It is possible that this ATP requirement for ternary complex formation reflects a requirement for ATP binding to Mlh1-Pms1 rather than the Msh6 nucleotide-binding site; however, ATP binding by MutS and not by MutL is required for the formation of MutS-MutL ternary complexes (15,16), mutations affecting ATP hydrolysis by human Mlh1 and Pms2 (homolog of S. cerevisiae Pms1) do not affect the formation of Msh2-Msh6-Mlh1-Pms2 ternary complexes (35), and mutations that are predicted to prevent ATP binding by Mlh1 and Pms1 individually cause only a small MMR defect (49). Four (Msh6-S1036P, Msh6-G1067D, Msh2-G855D, and Msh2-S742P) of the eight amino acid substitutions studied caused complete defects in ternary complex formation, and each of these had relatively normal ATP binding at the Msh6 site but substantially reduced ATP binding at the Msh2 site (Tables 1 and 2). It seems likely that these mutations cause defects both in ATP binding to the Msh2 nucleotide-binding site and in a conformational change required for ternary complex formation that is driven by ATP binding to the Msh6 nucleotide-binding site. Three of the mutant complexes (Msh2-T773D-Msh6, Msh2-R730C-Msh6, and Msh2-Msh6-G1142D) were proficient in formation of Mlh1-Pms1 ternary complexes, and one (Msh2-Msh6-R1024C) was partially proficient in ternary complex formation. The Msh2-T773D-Msh6 and Msh2-Msh6-G1142D complexes appeared to undergo the same conformational change as wild-type Msh2-Msh6 in response to ATP binding to the Msh6 nucleotide-binding site. Furthermore, the wild-type Msh2-Msh6, Msh2-T773D-Msh6, and Msh2-Msh6-G1142D complexes (Msh2-Msh6-R1024C and Msh2-R730C-Msh6 were not tested) were proficient in ternary complex formation at ATP concentrations predicted to saturate only the Msh6 nucleotide-binding site. It therefore seems likely that these amino acid substitutions do not disrupt (Msh2-T773D, Msh6-G1142D, and Msh2-R730C) or partially disrupt (Msh6-R1024C) the conformational change required for ternary complex formation that is driven by ATP binding to the Msh6 nucleotide-binding site even though they appear to disrupt the conformational changes required for sliding clamp formation. Recently, Msh2 domain II was identified as containing the interaction interface between Msh2-Msh6 and Mlh1-Pms1, and an isolated Msh2 domain II was shown to bind Mlh1-Pms1 independently of nucleotide and mispaired DNA (45). Thus, ATP binding to the Msh6 nucleotide-binding site likely induces the conformational change that exposes the Mlh1-Pms1 interface and mediates ternary complex formation.
Conversion of mispair-bound Msh2-Msh6 complexes to sliding clamps has been suggested to require ATP binding to both the Msh6 and Msh2 nucleotide-binding sites, based on the ATP concentration dependence of sliding clamp formation (21,32). Remarkably, all eight amino acid substitutions studied here caused defects in sliding clamp formation, regardless of whether or not they affected ATP binding to the Msh6 or Msh2 nucleotide-binding site. This group included mutant complexes that were competent for ATP␥S-induced protease protection of Msh6 and ATP-induced ternary complex formation (Msh2-T773D, Msh6-G1142D, and Msh2-R730C). These results suggest that the conformational changes involved in sliding clamp formation are distinguishable from those changes that can be monitored by limited trypsin proteolysis and ternary complex formation, and thus, they indicate that ternary complex formation and sliding clamp formation are separable. One intriguing explanation for the sensitivity of sliding clamp formation relative to other functions of Msh2-Msh6 is that sliding clamp formation likely requires energy to drive dissociation of the Msh6 mispair-binding domain from the bound mispair and as such may be more sensitive to subtle misalignments and/or destabilization of the dual ATP-bound state of Msh2-Msh6 (32) than are other conformational changes affecting ATP binding at the Msh2 nucleotide-binding site or ternary complex formation. Moreover, these data also indicate that formation of an Msh2-Msh6 sliding clamp is critical for MMR, as mutations encoding complexes that have defects only in sliding clamp formation (msh2-T773D, msh6-G1142D, msh2-R730C, and msh6-R1024C) are MMR-defective in vivo.