ATP Increases the Affinity between MutS ATPase Domains

MutS is the key protein of the Escherichia coli DNA mismatch repair system. It recognizes mispaired and unpaired bases and has intrinsic ATPase activity. ATP binding after mismatch recognition by MutS serves as a switch that enables MutL binding and the subsequent initiation of mismatch repair. However, the mechanism of this switch is poorly understood. We have investigated the effects of ATP binding on the MutS structure. Crystallographic studies of ATP-soaked crystals of MutS show a trapped intermediate, with ATP in the nucleotide-binding site. Local rearrangements of several residues around the nucleotide-binding site suggest a movement of the two ATPase domains of the MutS dimer toward each other. Analytical ultracentrifugation experiments confirm such a rearrangement, showing increased affinity between the ATPase domains upon ATP binding and decreased affinity in the presence of ADP. Mutations of specific residues in the nucleotide-binding domain reduce the dimer affinity of the ATPase domains. In addition, ATP-induced release of DNA is strongly reduced in these mutants, suggesting that the two activities are coupled. Hence, it seems plausible that modulation of the affinity between ATPase domains is the driving force for conformational changes in the MutS dimer. These changes are driven by distinct amino acids in the nucleotide-binding site and form the basis for long-range interactions between the ATPase domains and DNA-binding domains and subsequent binding of MutL and initiation of mismatch repair.

match binding holds the MutS dimer in a more closed and robust conformation.
More data are available on the nucleotide-induced conformational changes in MutS. Limited proteolysis of bacterial, yeast, and human MutS homologs has shown that the degradation pattern of MutS is altered in the absence or presence of ADP or ATP (4,13,18). Electron micrographs of human MutS␣ (hMutS␣) show a clear difference between the different nucleotide-bound states (4). In the presence of ADP, hMutS␣ shows an open clamp conformation. In the presence of ATP␥S, the clamp is closed, forming a more disc-like shape. Small angle x-ray scattering data confirm the opening and closing of the MutS clamp (19). In the presence of AMPPNP, the Thermus thermophilus MutS dimer shows the smallest radius of gyration, whereas in the presence of ADP the dimer is in the most extended conformation, with the apoprotein showing an intermediate state.
To elucidate how ATP binding causes the observed conformational changes in MutS, we soaked pre-existing crystals of MutS in complex with a mismatched DNA (17) in ATP solution. Previous attempts by W. Yang and co-workers (5) to obtain a crystal structure of Taq MutS binding ATP did not succeed, because ATP was hydrolyzed to ADP in the crystal. More recently, the structure of Taq MutS in complex with ADP-beryllium fluoride, an ATP analogue, was published (20). Both structures show small rearrangements throughout the molecule when compared with the nucleotide free structure, yet none of these explains the large conformational changes observed in the biochemical experiments.
Here we present the structure of E. coli MutS in complex with ATP. The structure does not show large conformational changes and presents, most likely, a trapped intermediate of a pre-hydrolysis state. However, despite the absence of major movements in this structure, small alterations in and around the nucleotide-binding site suggest that, upon ATP binding, the two ATPase domains of the MutS dimer move closer together. Using analytical ultracentrifugation we demonstrate that ATP binding indeed enhances the affinity of the ATPase domains; conversely, ADP binding decreases the affinity between the two domains. Finally, we discuss how these local changes in the ATPase domains of the MutS dimer are transmitted through other regions in the MutS-DNA complex, causing MutS to release the mismatch and/or enabling it to bind MutL.

EXPERIMENTAL PROCEDURES
Reagents and Sample Preparation-ATP, ADP, AMPPNP, and ATP␥S were obtained from Fluka. All other chemicals were purchased from Merck. All mutants were derived from plasmid pMQ372 (kindly provided by M. Marinus) and pM800 for ⌬C800 MutS (17), a C-terminal deletion construct that lacks the last, nonconserved 53 amino acids. Single site mutations were introduced by a method adapted from the Exsite kit (Stratagene). All MutS proteins and DNA substrates were prepared as described previously (20). Reactions were performed in 150 mM NaCl, 20 mM HEPES, pH 7.5, and 10 mM MgCl 2 .
Crystal Structure Determination-Crystals of ⌬C800 MutS were grown in the presence of 100 M ADP under conditions described earlier (17). Then, increasing volumes (1 ϩ 2 ϩ 4 ϩ 8 l) of an ATP containing solution (1 mM ATP, 30% (w/v) polyethylene glycol 6000, 15% (v/v) glycerol, 300 mM NaCl, 20 mM HEPES, pH 7.5, 10 mM MgCl 2 ), were added to crystal-containing drops (ϳ2 l) with 5-min intervals. Crystals were frozen and stored in liquid nitrogen until data collection. X-ray diffraction data were collected at 100 K on beamline ID14-EH2 at the European Synchrotron Radiation Facility in Grenoble. Data were processed using the HKL suite (21). The model of MutS-GT-ADP (Protein Data Bank accession code 1E3M) was used for rigid body refinement in Refmac5 (22), manual model building was performed in O (23), final refinement was performed with refmac5 using the TLS option (21 domains), and ARP/wARP (24) was used to build the solvent atoms. Details of the crystallographic analysis are given in Table I.
ATPase Assay-ATPase activity of MutS was measured in an ATPregenerating system by coupling ATP hydrolysis to oxidation of NADH (25). The final reaction mixture (100 l) contained 5 M MutS and 0 -50 M ATP. Oxidation of NADH was followed for 5 min in a Beckman spectrophotometer at 340 nm. For determination of the effect of homoand heteroduplex DNA on MutS ATPase activity and its dependence on salt concentration, 5 mM ATP containing 0.01 MBq/l [␣-32 P]ATP was incubated with 5 M MutS and a 30-base pair DNA oligomer (20 M) containing a single GT mismatch (17) in 20 mM HEPES, pH 7.5, and NaCl concentrations ranging from 0 to 400 mM. Reactions were quenched with 25 volumes of 0.5 M EDTA, and 1 l of this mixture was spotted on a polyethyleneimine cellulose-TLC plate (Merck, Darmstadt, Germany) and developed in 1 M orthophosphoric acid, pH 3.8. A strong stimulation of the MutS ATPase activity was observed upon inclusion of DNA in the assay, notably with heteroduplex DNA around physiological salt concentrations (up to 6-fold stimulation at 150 mM NaCl, data not shown). Our results are very similar to the ATPase activities recently reported for E. coli MutS (26) as well as those reported for eukaryotic MutS homologs (3,27,28). Subsequent experiments addressing ATPase stimulation of mutant MutS proteins by heteroduplex DNA were performed at 150 mM NaCl.
Filter Binding Studies-Binding of ATP␥S was determined using filter binding studies as described previously (10). Reaction mixtures (20 l) contained 5 M MutS and 0 -50 M [ 35 S]ATP␥S. Affinities reported in Table II represent total binding and are therefore a summation of nucleotide binding in both sites of the MutS dimer. ATP-induced release of DNA was measured as described previously (10), with final reaction mixtures containing 1 M MutS and 1 M DNA with or without 150 M ATP␥S.
Analytical Ultracentrifugation-Ultracentrifugation data were obtained using an XL-A analytical ultracentrifuge (Beckman Coulter). Equilibrium sedimentation of full-length and ⌬C800 MutS was performed in 50-l volumes in sample cells fitted with a 3-mm centerpiece. Sample cells contained 7 M protein in the absence or presence of nucleotides (50 M) in 250 mM NaCl, 20 mM HEPES, pH 7.5, 10 mM MgCl 2 . The reference cell contained the identical solution without MutS. Data were collected at 280 nm, 20°C, 10,400 rounds/min with a delay of 24 h to reach equilibrium. Theoretical molecular weights, partial specific volume (v bar ), and molar extinction coefficient at 280 nm were calculated using the program Sednterp (29). All data were analyzed using the program Ultraspin (www.mrc-cpe.cam.ac.uk).

Crystal Structure of the ATP-bound MutS-G:T Complex-
Previously (17) we solved the crystal structure of a C-terminal deletion construct of MutS, termed ⌬C800, in complex with substrate DNA. This construct lacks a nonconserved region at the C terminus but retains most functions of the full-length protein, being able to bind to mismatched DNA, hydrolyze ATP, and complement a MutS deficient strain in vivo, albeit with slightly reduced levels (17). In the ⌬C800 structure an ADP molecule (⌬C800-GT-ADP) is present in only the monomer that directly binds the mismatched base (monomer A) (17). To understand the role of ATP binding and its effects on the conformation of MutS, we soaked ATP into pre-grown crystals of the ⌬C800-GT-ADP complex. Soaking of the crystals in the nucleotide solution did not lead to any macroscopic changes of the crystals such as cracking, and no large changes in unit cell dimension were observed when compared with the ⌬C800-GT-ADP (data not shown). Soaking experiments with ATP, ATPvanadate, ATP-aluminum fluoride, and AMPPNP resulted in essentially identical structures; we have described only the ATP structure here. The ⌬C800-GT-ATP structure, on the whole, does not show major changes from the previously determined structure with ADP (⌬C800-GT-ADP). Contrary to what can be expected from the biochemical data, no rearrangement of domains or release of the DNA is visible (Fig. 1A). The ATP molecule is bound in the nucleotide-binding site in a conformation characteristic for P-loop hydrolases (30,31), yet hydrolysis of ATP did not take place here. This suggests that the ⌬C800-GT-ATP structure represents a trapped state. Apparently, the complex is caught within the crystal contacts and cannot carry out the movements required to complete the active site or cause conformational changes as observed in biochemical experiments. The only changes observed are restricted to the ATPase domains where several residues have been affected by the newly bound ATP molecules ( Fig. 1, D-F).
In monomer A we find clear density for ATP-Mg 2ϩ (Fig. 1, B and C). To accommodate the newly bound ATP, asparagine A616 (Asn A616 ), located at the tip of the P-loop, has rotated away from its original position to make place for the ␥-phosphate. Simultaneously, histidine A728 (His A728 ) has rotated toward Asn A616 . In this new position, Asn A616 and His A728 form a hydrogen bond and stabilize each other's orientation (Fig.  1E). The rotation of Asn A616 exposes the ␥-phosphate of the ATP to a serine in the opposing monomer B. This serine (Ser B668 ) is conserved in the MutS homologs and is located in a loop (residues 659 -670) that is equivalent to the signature motif of the ABC membrane transporters. Previously, no electron density was observed for this loop in the structure of ⌬C800-GT-ADP, even though the remainder of the complex showed well defined electron density. In the new structure with ATP, Ser B668 shows clear density, indicating that its position is stabilized by the presence of ATP, whereas the loop on which it resides is more ordered and shows partially interpretable density.
In monomer B (not contacting the mismatched base), which has no nucleotide in the previous structure with ADP, clear electron density is present that fits the base and sugar of the nucleotide. However, the tail of the nucleotide containing the three phosphates is flexible and can be placed in two different conformations, both of which are different from the position of the ATP in monomer A. Also, there is no density present that could fit a magnesium ion. In addition, the movements of Asn B616 and His B728 , as observed in monomer A, are not seen here. In fact, the side chain of Asn B616 and the adjacent backbone are disordered and show no electron density. Finally, the serine from the opposing monomer (Ser A668 ) and the loop on which it resides show weaker electron density than observed in the other nucleotide-binding site.
In conclusion, the two ATP-binding sites both bind ATP (Fig.  1, B and C) but show differences in conformation (Fig. 1, E and F), in agreement with the previously described asymmetry of the ATPase domains (10,(32)(33)(34)(35). An ADP molecule is found in the nucleotide-binding site of the ⌬C800-GT-ADP structure monomer A, which contacts the mismatch directly. The other monomer (B) does not make any contacts to the mismatch and has an empty nucleotide-binding site. In the MutS-GT-ATP structure, this asymmetry is retained with monomer A binding ATP "specifically," whereas monomer B binds ATP in an apparent "nonspecific" manner.
ATP Binding, Hydrolysis, and DNA Release-To further analyze the function of the three residues that were affected by the presence of ATP in the crystal structure, we mutated each to alanine, both in full-length MutS and in ⌬C800. N616A and H728A are located within or close to the nucleotide-binding site, respectively, and S668A is located further away from its own nucleotide-binding site but contributes to the opposing site in trans (see Fig. 1) (5). The two mutations, N616A and S668A, both show an ϳ4-fold reduction in ATPase activity, whereas the H728A mutation renders the protein virtually inactive (Table II). However, the effects of the three mutations on ATP binding are somewhat different; N616A and H728A both show a reduced affinity for ATP␥S, whereas S668A shows an affinity slightly stronger than wild-type MutS. The effects of these mutations on ⌬C800 are similar to the results obtained with the full-length product, confirming that the C-terminal deletion does not affect the conformation of the ATPase domains.
Similar to the ATPase activities, the binding to mismatched DNA is affected in the three mutants. N616A and S668A show a somewhat reduced DNA binding and no longer release it in the presence of ATP␥S as readily as wild-type protein. H728A has an even weaker affinity for DNA, which improves slightly in the presence of ATP␥S. Finally, although the ATPase activity of wild-type MutS in the presence of mismatched DNA is stimulated 6-fold, this stimulation is reduced for N616A and S668A and completely lost in H728A. Hence, this suggests that the residues that were affected by the binding of ATP in the MutS crystal structure play a role not only in nucleotide binding and hydrolysis but also in inducing the conformational changes that signal between ATPase domains and DNA-binding domains.
Oligomerization of MutS-The ATPase domain of MutS belongs to the family of ABC ATPases. For several members of this family it has been shown that the ATPase domains dimerize upon binding of ATP (36 -40). In the crystal structure of MutS, two interfaces of the MutS dimer are present. The major interface is found between the two ATPase domains (2922 Å 2 ). A second, much smaller interface (451 Å 2 ) is mediated through the DNA. Therefore, in the absence of DNA, the changes in the dimerization of the ⌬C800 MutS will mainly reflect changes in the affinity of the ATPase domains. Whereas the full-length MutS protein exists in equilibrium between dimers and tetramer, ⌬C800 MutS shows a monomer-dimer equilibrium. For this reason, we have used ⌬C800 MutS to analyze the effect of nucleotide on the dimerization of the ATPase domains.
We have used analytical ultracentrifugation to examine the oligomeric state of ⌬C800 and full-length MutS. The data were fitted either to a monomer-dimer or dimer-tetramer equilibrium using fixed calculated molecular masses for ⌬C800 (residues 1-800, 89.5 kDa) and full-length MutS (853 residues, 95.2 kDa). In addition, ADP, ATP, or the nonhydrolyzable AMPPNP was added to test its effect on the oligomeric state of the protein. Full-length MutS can be fitted to a dimer-tetramer equilibrium (K d ϭ 2.2 M) without any indication of monomers or higher order oligomers (Table III). The oligomerization of full-length MutS is not affected by AMPPNP because the equilibrium is identical in the absence or presence of the nucleotide.
The deletion construct ⌬C800 on the other hand, is strongly affected by the presence of nucleotides (Fig. 2). In the absence of added nucleotide (ADP bound to monomer A only), ⌬C800 MutS is found to be in a monomer-dimer equilibrium (K d ϭ 13 M). The addition of AMPPNP enhances the affinity of the ⌬C800 monomers, decreasing the K d for dimerization 26-fold (K d ϭ 0.5 M). The addition of ADP or ATP (which is hydrolyzed during the course of the experiment), on the other hand, strongly drives the equilibrium toward the monomeric species (K d Ͼ 1000 M) (data not shown). These results indicate that ATP binding tightens the MutS dimer, whereas binding of ADP to both nucleotide-binding pockets leads to loosening of the intertwined ATPase domains, resulting in monomerization. Ultimately, this resulted in precipitation of the protein.
To further analyze the structural alterations affecting interactions between the ATPase domains, and therefore the dimerization of ⌬C800 MutS, we have used the three mutants N616A, S668A, and H728A in our analytical ultracentrifugation experiments. N616A displays a strong reduction in the affinity between the ATPase domains, and the addition of AMPPNP stimulates dimerization only 2.5-fold. The S668A mutant appears to be less affected in dimerization, whereas the H728A mutant starts to dissociate into monomers upon addition of AMPPNP rather than dimerizing. Because these effects do not correlate with the ATP binding affinity (Tables II and  III), this behavior cannot be attributed to decreased affinity for ATP alone. Another possible cause for the decreased dimerization might be that the mutations prevent interactions at the tighter dimer interface; this, however, does not seem very likely to us. All three mutations result in a less bulky side chain (alanine) than the original side chain (asparagine, serine, arginine) and therefore would not lead to steric clashes in a tighter dimer interface. Alternatively, it could be that the residues are involved in stabilization of a tighter dimer by forming bonds with residues across the dimer interface. This option also does not appear likely to us. The dimer interface is very large (ϳ3000 Å 2 ), involving ϳ40 residues from each monomer, and hence a single mutation is not likely to have a great effect. This is exemplified by arginine 697, which is located at the dimer interface and makes contacts with the backbone of the other monomer. Yet mutation of this residue to an alanine does not change the dimerization properties, as observed with analytical ultracentrifugation (data not shown), although it shows reduced ATP binding and hydrolysis (10). Hence, these results support the hypothesis that these three residues are not only involved in nucleotide binding and hydrolysis but also play an important role in inducing/signaling conformational changes throughout the MutS complex.

DISCUSSION
Structure of ATP-bound MutS-To obtain insights into ATPinduced conformational changes, we solved the crystal structure of the MutS-DNA complex binding ATP, MutS-GT-ATP. However, despite the binding of the nucleotide, no large structural changes were observed. In addition, ATP is not hydrolyzed, indicating that the active site is not completed. Most likely, the protein is trapped by crystal contacts with neighboring molecules and thus prevented from moving into the activated state. In a similar experiment with crystals of Taq MutS, it was found that the nucleotide was hydrolyzed within the crystals (5). More recently, the structure of Taq MutS bound to ADP-beryllium fluoride, a non-hydrolyzable ATP analogue, was published (41). In this structure one of the two disordered loops becomes more defined upon binding of the nucleotide, similar to what we observed after ATP binding in our structure. An arginine residue that was indicated as transmitter residue in Taq  The non-equivalent binding of ATP in the two ATPase domains of the MutS dimer is consistent with the described asymmetry (10,(32)(33)(34)(35). In the ⌬C800-GT-ATP crystal structure, monomer A binds the nucleotide in a manner typical for P-loop NTPases (30,31), whereas monomer B binds the nucleotide in a less specific manner. A possible explanation is that monomer A retains the high affinity site, whereas monomer B retains the low affinity site. In the structure the nucleotide in monomer A appears to be ready for hydrolysis, waiting only for the active site to be completed, whereas the ATP in monomer B is not in a position suitable for hydrolysis, with the phosphate tail disordered and no magnesium ion present (which is essential for catalysis). It is plausible that the conformational changes needed to complete the active site in monomer A and subsequently to hydrolyze ATP also cause a switch of the asymmetry. Thus, after hydrolysis of ATP residing in monomer A, the previous low affinity site (B) is expected to change into a high affinity site, primed for hydrolysis, whereas the previous high affinity site (A) becomes the low affinity site, enabling it to release the product. This mechanism is expected to result in the observed alternation of ATP hydrolysis by the two nucleotide-binding sites, as described (10). However, the precise order of events requires additional study. Bjornson et al. (34) have shown that in the absence of DNA, when one monomer binds ADP with high affinity, the second monomer is simultaneously able to bind ATP with high affinity, implying two separate high affinity binding sites for di-and trinucleotides, respectively. In contrast, in our ⌬C800-GT-ADP structure, monomer A has the high affinity site for ADP, and it is the same site that has high affinity for the adenosine trinucleotide in the ⌬C800-GT-ATP structure. We conclude therefore that mismatched DNA modulates the high affinity states for the different nucleotides in a particular site.
ATP Binding and Hydrolysis in MutS-The data presented in this work indicate that the two ATPase domains of the MutS dimer need to move closer to each other in order to complete the active site and that the three residues Asn 616 , Ser 668 , and His 728 play an important role in this rearrangement. The conserved serine (Ser 668 ) of the signature loop has been implicated in nucleotide hydrolysis in the active site of the opposing monomer, as it is too far away (Ͼ20 Å) from its own nucleotidebinding site (5). Yet, in the structure of ⌬C800-GT-ATP, Ser 668 is still located ϳ5 Å away from the ␥-phosphate of the ATP in  the opposing monomer. Therefore, to approach the ATP, the two ATPase domains have to move closer together. Such an ATP-dependent movement is conserved in several ABC AT-Pases (36 -40). In the crystal structure of the ABC ATPase RAD50 in complex with AMPPNP, a tighter ATPase dimer is observed, with the conserved serine in close proximity to the nucleotide, where it can act in hydrolysis (36). Hopfner et al. (36) suggest that the serine might act in a similar manner to the "arginine finger" found in a group of molecular switches called small G-proteins. In these proteins, the arginine finger is thought to assist hydrolysis by reducing the negative charge of the ␥-phosphate in the pentavalent transition intermediate (31). In a similar way, the serine, located at the beginning of a helix, can deliver the positive charge of the helix dipole (42) to the ␥-phosphate of the nucleotide in the opposing monomer. Moreover, in the small G-proteins the arginine is provided by another protein (the G-protein-activating protein (GAP)), and hydrolysis is controlled by association and dissociation of the two proteins. For RAD50 (36) and shorter versions of MutS (43) it has been shown that monomers need to dimerize for nucleotide hydrolysis to take place. The data presented here indicate that also in full-length MutS, although it forms a stable dimer, ATP hydrolysis is modulated in a similar way. In the ADPbound state, the two ATPase domains have extensive contacts with each other, but the Ser 668 is at a distance from the active site of the opposing monomer. The binding of ATP brings in a strong negative charge opposite to the serine, which will attract the other monomer and bring the two domains closer together. The rotation of the Asn 616 and His 728 away from their position at the dimer interface makes it possible for Ser 668 to move closer to the newly bound ATP and complete the active site.
Mutation of the Ser 668 to alanine results in a 3-fold decrease in ATPase activity. Similar results were previously obtained for E. coli MutS (5) and yeast MSH6 (27). This relatively mild effect of the S668A mutation on the ATPase activity of MutS could be explained by the fact that the positive charge created by the helix dipole is still present but cannot be delivered close to the ␥-phosphate as efficiently. In a similar way, the N616A mutation could also affect the delivery of the positive charge. The analytical ultracentrifugation data show that the ATPase domains no longer dimerize as in the wild type, implying that the movement of Ser 668 closer to the ␥-phosphate no longer takes place with the same efficiency. The H728A mutation, on the other hand, renders the protein virtually inactive. Like N616A, this mutant is deficient in ATP-induced dimerization; but in addition, His 728 also assists in the rotation of Asn 616 to accommodate the ␥-phosphate of the ATP. As it is no longer stabilized in its "rotated-out" position, Asn 616 will compete for the same space as the ␥-phosphate, thereby interfering with the proper positioning of the ATP. Thus, the mutation of His 728 works in two ways, by counteracting binding of the nucleotide and by preventing subsequent completion of the active site by reducing the dimerization of the ATPase domains upon ATP binding. Hopfner et al. (36) proposed that the equivalent histidine in RAD50 might act in ATP hydrolysis by activating a water molecule for the hydrophilic attack on the ␥-phosphate. This would explain the reduced ATP hydrolysis but does not explain the changes in ATP binding. Furthermore, the importance of His 728 is exemplified by a mutation of the equivalent of His 728 in the human MutS homolog MSH6 (H1248D), which was found in a patient with HNPCC, a hereditary form of colon cancer caused by mismatch repair deficiency (44). In addition, mutations at positions equivalent to Ser 668 and His 728 (Ser 1036 and His 1096 , respectively) caused MMR deficiency in yeast MSH6 (45).
Nucleotide-induced Conformational Changes in MutS-The data presented here show that ATP and ADP binding modulate the dimerization of the ATPase domains in ⌬C800 MutS. Yet how does this affect full-length protein? A dimer-tetramer equilibrium, similar to the 2.2 M described here, has also been observed for Taq MutS (46), and T. thermophilus MutS (47,48); however, Bjornson et al. (49) reported a 50-fold lower dissociation constant for the dimer-tetramer equilibrium. A possible explanation for this discrepancy with our data may lie in the conditions under which the analytical ultracentrifugation experiments were performed (i.e. temperatures were 20 and 4°C, respectively) and/or in differences in protein preparation. A possible role for the tetramer in MutS function has not yet been resolved, but it is clear that the full-length protein can exist as a stable dimer. Thus it is safe to assume that in full-length MutS, nucleotide binding causes a "tightening" and "relaxing" of the ATPase dimers. At the same time, nucleotide binding functions as an effector of DNA binding (7,13,14). Our data suggest that the "breathing" of the ATPase domains is coupled to the modulation of DNA binding by nucleotides. The observation that the three mutants N616A, S668A, and H728A are not only diminished in nucleotide-induced modulation of the ATPase dimer affinity but also display reduced ATP-induced DNA release supports the idea that the two activities are linked. In yeast MSH6 a mutation at the equivalent of Ser 668 (yMSH6 Ser 1036 ) also changed nucleotide-induced release of DNA; in contrast, mutation of a residue equivalent to His 728 (His 1096 ) did not show altered behavior (27). It is therefore possible that this residue is primarily important in the MSH2 ATPase domain, as the two monomers are known to be nonequivalent (33, 35). Because the ATPase domains and DNA-binding domains are separated from one another by ϳ60 Å, long-range communication must take place throughout the MutS dimer. To get an idea of how ATP binding could cause conformational changes in domains located at the other end of the complex, we have used the structure of RAD50, which has been solved in complex with AMPPNP and shows a more compact ATPase dimer (36). When we modeled the MutS-ATP structure on the RAD50-AMPPNP structure, we observed a 5-Å translation of the two domains closer together. In addition, we also found a ϳ25 o rotation of the two monomers toward one another, thus forming a narrower dimer configuration (Fig. 3A). Such a movement is in agreement with the small angle x-ray scattering experiments of Kato et al. (19). In their studies they show that the T. thermophilus MutS dimer becomes more compact in the AMPPNPbound state and more relaxed in the ADP-bound state and adopts an intermediate shape in the nucleotide-free form. In addition, an ATP-induced closing of the clamp is also found in electron microscopy images of human MutS␣, which show a closed clamp in the ATP-bound state and an open clamp in the presence of ADP (4). Different DNA binding assays have shown that ATP binding causes MutS to become a sliding clamp that releases the mismatch and moves along the length of the DNA (3,6,18,26,50). In Fig. 3 we have modeled the ATP-induced sliding clamp in MutS. In the ADP-bound state and the absence of DNA (Fig. 3B), the DNA-binding domains are flexible and the clamp is opened up to accommodate the incoming DNA. When DNA is bound (Fig. 3C), the clamp closes, encircling the DNA. Subsequent ATP binding (Fig. 3D) causes the two monomers to rotate toward one another. However, such a rotation will cause steric clashes, notably in the mismatch binding domains. Therefore, upon ATP binding, the two mismatch binding domains could possibly rotate away, thereby making MutS into a sliding clamp. Hence, ATP binding and subsequent hydrolysis to ADP provide a mechanism for MutS to modify the dimer interactions, which in turn are transmitted to other regions on the dimer causing conformational changes needed for DNA release and/or MutL complex formation.