Distinct MutS DNA-binding Modes That Are Differentially Modulated by ATP Binding and Hydrolysis*

The role of MutS ATPase in mismatch repair is con-troversial. To clarify further the function of this activity, we have examined adenine nucleotide effects on interactions of Escherichia coli MutS with homoduplex and heteroduplex DNAs. In contrast to previous results with human MutS (cid:1) , we find that a physical block at one end of a linear heteroduplex is sufficient to support stable MutS complex formation in the presence of ATP (cid:1) Mg 2 (cid:2) . Surface plasmon resonance analysis at low ionic strength indicates that the lifetime of MutS complexes with heteroduplex DNA depends on the nature of the nucleotide present when MutS binds. Whereas complexes prepared in the absence of nucleotide or in the presence of ADP undergo rapid dissociation upon challenge with ATP (cid:1) Mg 2 (cid:2) , complexes produced in the presence of ATP (cid:1) Mg 2 (cid:2) , adenosine 5 (cid:2) -( (cid:3) , (cid:4) -imino)triphosphate (AMPPNP) (cid:1) Mg 2 (cid:2) , or ATP (no Mg 2 (cid:2) ) are resistant to dissociation upon ATP challenge. AMPPNP (cid:1) Mg 2 (cid:2) and ATP (no Mg 2 (cid:2) ) reduce MutS affinity for heteroduplex but have little effect on homoduplex affinity, resulting in abolition of specificity for mispaired DNA at physiological (EM Science, Gibbstown, NJ) developed in 0.3 M KPO 4 , pH 7.0. Dried plates were quantitated on a Molecular Dynamics PhosphorImager after overnight exposure. Initial steady-state rates of ATP hydrolysis were determined by least squares analysis of the linear portion of the progress curve. Estimation of MutS (cid:1) DNA Dissociation Constants— Values for the dissociation constant ( K d ) for MutS DNA complexes were determined by surface plasmon resonance spectroscopy by titration of chip-bound DNA with increasing concentrations of MutS. Maximum binding values ob- tained in this manner were plotted as a function of protein concentration and data fit to a square hyperbola, from K d value was concentrations or nucleotide, MutS binding to 41-bp homoduplex DNA was not saturable as monitored surface plasmon resonance spectroscopy, not saturable concentrations . cases values were estimated purposes comparison with constants ob- under conditions where saturable behavior was observed. this end, binding isotherms were formulated by fitting binding data to hyperbola constrained saturation value determined DNA under conditions high affinity binding (in or presence ADP).

The role of MutS ATPase in mismatch repair is controversial. To clarify further the function of this activity, we have examined adenine nucleotide effects on interactions of Escherichia coli MutS with homoduplex and heteroduplex DNAs. In contrast to previous results with human MutS␣, we find that a physical block at one end of a linear heteroduplex is sufficient to support stable MutS complex formation in the presence of ATP⅐Mg 2؉ . Surface plasmon resonance analysis at low ionic strength indicates that the lifetime of MutS complexes with heteroduplex DNA depends on the nature of the nucleotide present when MutS binds. Whereas complexes prepared in the absence of nucleotide or in the presence of ADP undergo rapid dissociation upon challenge with ATP⅐Mg 2؉ , complexes produced in the presence of ATP⅐Mg 2؉ , adenosine 5-(␤,␥-imino)triphosphate (AMPPNP)⅐Mg 2؉ , or ATP (no Mg 2؉ ) are resistant to dissociation upon ATP challenge. AMPPNP⅐Mg 2؉ and ATP (no Mg 2؉ ) reduce MutS affinity for heteroduplex but have little effect on homoduplex affinity, resulting in abolition of specificity for mispaired DNA at physiological salt concentrations. Conversely, the highest mismatch specificity is observed in the absence of nucleotide or in the presence of ADP. ADP has only a limited effect on heteroduplex affinity but reduces MutS affinity for homoduplex DNA.
DNA biosynthetic errors that escape the proofreading function of DNA polymerase are corrected by mismatch repair. Although mismatch repair was initially characterized in bacteria (1)(2)(3)(4), analogous systems have been identified in eukaryotes, and defects in the mammalian pathway have been implicated in tumor development (5)(6)(7)(8).
Rectification of DNA biosynthetic errors by these systems relies on secondary signals within the helix that serve to distinguish newly synthesized DNA from the template strand. In Escherichia coli these signals are based on patterns of adenine methylation of d(GATC) sequences, which are hemimethylated in newly replicated DNA (2). Repair is initiated via mismatch recognition by MutS, with assembly of a MutL⅐MutS⅐heteroduplex complex serving to activate the MutH d(GATC) endonuclease in an ATP-dependent reaction (9 -11). Activated MutH cleaves the unmethylated strand of a hemimodified d(GATC) sequence that may be located to either side of the mispair. Assembly of MutL⅐MutS⅐heteroduplex is also sufficient to activate unwinding by DNA helicase II, which enters the helix at the MutH-produced strand break with an orientation bias such that it tracks back toward the mismatch (12). This orientation-dependent activation of the excision system at the strand break implies signaling between the two DNA sites along the helix contour. The fact that MutS and MutL are sufficient for this effect suggests that one or both of these activities function in signal transmission between the two DNA sites.
Several models have been proposed to account for interaction of the two DNA sites involved in the methyl-directed reaction. The recent demonstration that a mismatch within one oligonucleotide duplex can activate MutS-and MutL-dependent MutH cleavage of a d(GATC) sequence on a second oligonucleotide duplex has led to the suggestion that the two sites are brought into proximity by DNA bending with the presence of MutS at the mispair signaling activation of downstream activities at the strand signal via a MutL interface (13). However, the efficiency MutH activation reported in these trans-activation experiments is several hundred times lower than that observed for d(GATC) cleavage activated in cis by a mismatch located 1,000 base pairs distant (11). Furthermore, whereas a DNA bending mechanism might account for activation of MutH cleavage at a d(GATC) sequence, it does not explain the MutS-, MutL-, and orientation-dependent activation of the excision system at the ensuing strand break.
Two alternate models, which can account for orientation-dependent signaling between the two DNA sites, invoke ATP-dependent movement of MutS and its eukaryotic homologs along the helix contour (14 -16). The presence of ATP reduces the steady-state affinity of bacterial MutS and eukaryotic MutS␣ for a mismatch (10,(17)(18)(19)(20)(21), and ATP challenge of preformed MutS⅐DNA or hMutS␣⅐DNA complexes can lead to release of the protein from a mismatch (14 -16). Since ATP␥S also results in release of a MutS homolog from a mispair (14,16,21), this effect may be due at least in part to binding of the triphosphate, although ATP hydrolysis by the heteroduplex-bound protein has been invoked by several laboratories (14,15,22). Electron microscopic analysis of the kinetics of evolution of MutS⅐heteroduplex complexes has indicated that the presence of ATP results in the nucleotide-dependent formation of DNA loops, which grow with time and are stabilized by bound MutS at the base (14). However, loop formation was not supported by ATP␥S, 1 and ongoing ATP-dependent DNA loop growth was blocked upon addition of excess AMPPNP. Since the mismatch was found within the DNA loop with most heteroduplexes, these effects were attributed to ATP-dependent tracking of MutS from the mispair in a bidirectional fashion.
Analysis of ATP effects on the interaction of human MutS␣ with linear heteroduplexes that have biotin-avidin end-blocks or with circular DNAs has indicated that the MSH2⅐hMSH6 heterodimer also leaves a mispair in the presence of ATP by movement along the helix (15,16,22). This has led to the suggestion that in the presence of ATP, MutS homologs form a sliding clamp about the helix (15,16). However, two distinct mechanisms for movement of this clamp have been proposed. One model invokes ATP binding and hydrolysis by DNA-bound hMutS␣ to modulate protein conformers leading to directional movement of a sliding clamp along the helix (15). The other posits mismatch recognition by the hMutS␣⅐ADP complex, with the mispair acting as a nucleotide exchange factor that promotes formation of the hMutS␣⅐ATP complex that is capable of free diffusion along the helix (16). In this mechanism ATP hydrolysis occurs after release from DNA and serves to regenerate the mismatch-binding form of the protein. These models differ with respect to an ADP requirement for mismatch recognition, whether movement on DNA is directional and whether ATP hydrolysis is supported by the DNA-bound form of the protein.
As noted above, the substantial evidence supporting ATP-dependent movement of MutS homologs along the helix has been interpreted in terms of a mechanism for signaling between the mismatch and the strand signal that directs repair. However, it is important to note that other potential functional roles for movement of the protein along the helix have not been excluded, for example, involvement in the kinetic path by which mispairs are located.
A distinct function for ATP binding and hydrolysis by MutS homologs has been suggested in the context of the DNA bending model for interaction of the mismatch and strand signal sites that was described above. This proposal stipulates that once mismatch recognition occurs, MutS remains bound to the mispair during the course of the repair reaction. In this scheme, ATP binding and hydrolysis by DNA-bound MutS functions in a kinetic proofreading mechanism that is used to verify mismatch recognition prior to initiation of repair (13). ATP binding by DNA-bound MutS is envisioned to reduce the affinity of the protein for both mismatches and homoduplex sites. The ATP-induced reduction in affinity is postulated to be greater for the MutS⅐homoduplex complex than for the MutS⅐mismatch complex, resulting in preferential dissociation of MutS from correctly paired sequences. This paper addresses the effects of adenine nucleotides on the interaction of MutS with heteroduplex and homoduplex DNAs. We show that as in the case of human MutS␣, ATPpromoted dissociation of a MutS⅐heteroduplex complex requires free DNA ends. Interestingly, a single avidin-biotin endblock is sufficient to prevent dissociation, implying that this protein does not form a simple sliding clamp in the presence of ATP that is capable of free diffusion along a heteroduplex. By using surface plasmon resonance spectroscopy, we also show that MutS is capable of specific mismatch recognition in the presence of ADP, ATP⅐Mg 2ϩ , or in the absence of nucleotide, although mismatch specificity is not detectable at physiological ionic strength in the presence of AMPPNP⅐Mg 2ϩ or in the presence of ATP in the absence of a divalent cation. These observations indicate that adenine nucleotides differentially modulate mismatch and homoduplex interactions and indicate that the protein has two distinct DNA-binding modes.

EXPERIMENTAL PROCEDURES
Proteins and DNAs-MutS was isolated by a modification of the procedure described previously (9). Frozen cell paste (50 g) was resuspended in 20 mM KPO 4 , pH 7.4, 1 mM EDTA, 10 mM 2-mercaptoethanol (2 ml/g of cell paste), and cells were disrupted by sonication. The extract was clarified by centrifugation (15,000 ϫ g, 45 min), and the resultant supernatant was treated with 0.25 volume of 25% (w/v) streptomycin sulfate. After stirring on ice for 30 min, the precipitate was removed by centrifugation (15,000 ϫ g, 30 min), and the supernatant was treated with ammonium sulfate (0.18 g/ml) added over a period of 30 min. After stirring on ice for 45 min, the precipitate was collected by centrifugation (15, (v/v) glycerol. Protein was stored at Ϫ20°C. MutS concentrations were determined from the absorbance at 280 nm using an extinction coefficient of 69420 M Ϫ1 cm Ϫ1 calculated from the primary amino acid sequence (23).
Oligodeoxyribonucleotides were purchased from Oligos Etc. (Wilsonville, OR) and, when indicated, were radiolabeled at the 5Ј-end with T4 polynucleotide kinase and [␥-32 P]ATP (3,000 Ci/mmol, PerkinElmer Life Sciences) to a specific activity of 1 ϫ 10 6 cpm/pmol. A 41-base pair (bp) G-T heteroduplex that contained a terminal biotin at one or both ends of the duplex was prepared by combining 160 nM 5Ј-32 P-labeled d(CGCCGAATTGCTAGCAAGCTGTCGAGTCTAAAAATTCGGCT)-3Јbiotin with 320 nM of unlabeled d(AGCCGAATTTTTAGACTCGATAG-CTTGCTAGCAATTCGGCG) that was synthesized with or without a 3Ј-terminal biotin. DNA samples (100 l) were annealed in 10 mM Tris-HCl, pH 8.0, 1 mM EDTA, and 150 mM NaCl by heating at 99°C in a PerkinElmer Life Sciences Gene Amp 9600 thermocycler for 2 min and cooling to 25°C over a period 90 min. Otherwise identical A⅐T homoduplex substrates were made by annealing 160 nM 5Ј-32 P-labeled d(CGCCGAATTGCTAGCAAGCTATCGAGTCTAAAAATTCGGCT)-3Јbiotin with 320 nM of the latter oligonucleotide above with and without a 3Ј-biotin. Substrates without 3Ј-biotin tags were prepared in a similar fashion.
DNAs used for surface plasmon resonance spectroscopy were prepared by combining 1 M of 5Ј-biotin-d(CGCCGAATTGCTAGCAAGCT-GTCGAGTCTAAAAATTCGGCT) or 5Ј-biotin-d(CGCCGAATTGCTAG-CAAGCTATCGAGTCTAAAAATTCGGCT) with 2 M 5Ј-d(AGCCGAA-TTTTTAGACTCGATAGCTTGCTAGCAATTCGGCG) for 41-bp G-T heteroduplex and A⅐T homoduplex, respectively. Oligonucleotides (100 l) were annealed as described above and diluted 10-fold prior to conjugation to the streptavidin sensor chip as described below.
Two hundred-bp G-T heteroduplex and A⅐T homoduplex DNAs were prepared by the polymerase chain reaction method described previously (15). Briefly, homoduplex and heteroduplex molecules were prepared by hybridizing single strands isolated by denaturing high pressure liquid chromatography from duplex fragments produced by the amplification of base pairs 5531-5732 of bacteriophages f1MR1 and f1MR3 (24). A biotinylated f1MR1 fragment was produced using 5Ј-biotin-d(TACGCG-CAGCGTGACCGCTA) as a forward primer and d(AAGTTTTTT-GGGGTCGAGGT) as a non-biotinylated reverse primer. A non-biotinylated f1MR3 fragment was produced using the same primers lacking biotin. The 5Ј-biotinylated viral strand from f1MR1 (0.3 M) was combined in a 100-l volume with 0.6 M of the complementary sequence prepared from f1MR1 or f1MR3 to form a homoduplex or heteroduplex, respectively. Duplexes were annealed as described above.
Gel Shift Analysis-DNA-binding reactions (20 l) contained 20 mM Tris-HCl, pH 7.6, 1 mM dithiothreitol, 50 g/ml bovine serum albumin, 5 mM MgCl 2 , 50 mM KCl. 41-bp 32 P-homoduplex or 32 P-heteroduplex, 25 g/ml of a BstEII digest of bacteriophage DNA (New England Biolabs) as nonspecific competitor, streptavidin, and ATP were present as indicated. Solutions were preincubated at 25 or 37°C as noted for 10 min, the reaction initiated by addition of MutS, and incubation continued for 10 min. Reactions were terminated by the addition of 2 l of 50% (v/v) glycerol, 0.05% xylene cyanol, 0.05% bromphenol blue, and 20 mM EDTA, placed on ice, and loaded onto 4 or 5% native polyacrylamide gels (acrylamide/bisacrylamide, 37.5:1) in 6.7 mM Tris acetate, pH 7.5, and 1 mM EDTA. Gels were electrophoresed at room temperature at 11.4 V/cm in this buffer. 32 P-Labeled complexes were visualized by autoradiography after drying and quantitated using a Molecular Dynamics PhosphorImager. Surface Plasmon Resonance Spectroscopy-Surface plasmon resonance measurements were performed on a BIAcore 2000. A streptavidin sensor chip was derivatized with 41-bp homoduplex (305 response units) or G-T heteroduplex (306 response units) DNAs described above in which one strand was derivatized with a 5Ј-biotin. Additional sensor chips were derivatized with 299 response units and 298 response units of a 5Ј-biotin-tagged 200-bp homoduplex or G-T heteroduplex, respectively. Solutions (100 l) of MutS in 20 mM Tris-HCl, pH 7.6, 1 mM dithiothreitol, 0.005% surfactant P-20, 5 mM MgCl 2 , containing KCl and nucleotide concentrations as indicated, were flowed across the SA chip at a rate of 20 l/min. After MutS association, the chip was washed at 20 l/min with 20 l of 20 mM Tris, pH 7.6, 1 mM dithiothreitol, 0.005% surfactant P-20, and KCl as indicated, followed by 60 l of the same buffer/KCl solution containing 5 mM MgCl 2 and 1 mM ATP. Measurements were performed at 25°C, and samples were maintained at 4°C prior to injection. The SA chip was regenerated by a 20-l injection of 0.5% sodium dodecyl sulfate.
ATPase Assays-ATP hydrolysis by MutS was measured in 20-l reactions at 37°C in 20 mM Tris-HCl, pH 7.6, 50 mM KCl, 5 mM MgCl 2 , 1 mM dithiothreitol, 50 g/ml bovine serum albumin, 200 M [␣-32 P]ATP (2500 Ci/mmol), 0.5 mg/ml streptavidin, and DNA substrates as indicated. These components, except for ATP, were assembled on ice and incubated for 10 min at room temperature to allow the conjugation of streptavidin to biotin-tagged DNAs. After supplementation with MutS (300 nM monomer equivalent) and incubation at 37°C for 5 min, reactions were initiated by addition of the [␣-32 P]ATP. Rate determinations were based on removal of 2-l samples, which were quenched in 50 l of 0.5 M EDTA, pH 8.0. ATP hydrolysis was determined by chromatography of 1 l of quenched sample on PEI cellulose plates (EM Science, Gibbstown, NJ) developed in 0.3 M KPO 4 , pH 7.0. Dried plates were quantitated on a Molecular Dynamics PhosphorImager after overnight exposure. Initial steady-state rates of ATP hydrolysis were determined by least squares analysis of the linear portion of the progress curve.
Estimation of MutS⅐DNA Dissociation Constants-Values for the dissociation constant (K d ) for MutS DNA complexes were determined by surface plasmon resonance spectroscopy by titration of chip-bound DNA with increasing concentrations of MutS. Maximum binding values obtained in this manner were plotted as a function of the protein concentration and data fit to a square hyperbola, from which the K d value was extracted.
At KCl concentrations above 100 mM in the presence or absence of nucleotide, MutS binding to 41-bp homoduplex DNA was not saturable as monitored by surface plasmon resonance spectroscopy, even at the highest concentration tested (800 nM). Binding to heteroduplex also was not saturable at these salt concentrations in the presence of ATP, AMPPNP, or ATP in the absence of Mg 2ϩ . In these cases K d values were estimated for purposes of comparison with dissociation constants obtained under conditions where saturable behavior was observed. To this end, binding isotherms were formulated by fitting binding data to a hyperbola that was constrained to the saturation value determined with heteroduplex DNA under conditions of high affinity binding (in the absence of nucleotide or in the presence of ADP).

One or Two Streptavidin End-blocks Interfere with ATPpromoted Release of MutS from Linear Heteroduplex DNA-
Terminal streptavidin-biotin complexes have been used to study the mechanism of nucleotide-promoted release of human MutS␣ from linear heteroduplex DNA (15,16,22). We have utilized this method to explore further the nature of bacterial MutS movement along the helix in the presence of ATP. These experiments used a 41-base pair G-T heteroduplex with a cen-trally located mispair or an otherwise identical A⅐T homoduplex DNA (see "Experimental Procedures") possessing either two free termini or one or two biotin-streptavidin terminal end-blocks. The ability of MutS to form stable DNA complexes with these DNAs in the absence or presence of ATP was tested by gel shift analysis.
Due to the presence of unlabeled homoduplex competitor in all reactions, no binding was observed with the A⅐T homoduplex control that contained one or two streptavidin-biotin endblocks (Fig. 1, upper panel, lanes 10 -13), indicating that MutS does not interact nonspecifically with the streptavidin component of such substrates. As observed previously (14), the presence of ATP and Mg 2ϩ reduced the yield of MutS complexes with the streptavidin-free G-T heteroduplex by about 90% (compare lanes 1-3). However, the yield of specific complexes with the G-T DNA containing a bi-terminal streptavidin endblock was unaffected by the presence of the nucleotide (lanes 7-9), a finding similar to those obtained with human MutS␣ (15,16). Furthermore, a single streptavidin block was largely sufficient to stabilize specific complexes in the presence of ATP (lanes 4 -6). The latter finding was independent of placement of the streptavidin complex at the 3Ј-end of one strand or the other, or 3Ј or 5Ј derivatization of a particular duplex terminus (not shown), ruling out strand polarity or position effects with respect to placement of the single end-block. Quantitation of the results shown in Fig. 1 (upper panel) demonstrated that the yield of specific MutS⅐heteroduplex complexes in the presence of ATP, as compared with that observed in the absence of nucleotide, was 10% for the heteroduplex with free ends, 100% for the substrate with the bi-terminal block, and 80% for G-T DNA with a streptavidin-biotin complex at only one end. It is noteworthy that the complexes observed with end-blocked heteroduplexes were obtained in reactions that were initiated by addition of MutS to solutions that contained both DNA and nucleotide ("Experimental Procedures"), i.e. the formation of such complexes did not require pre-binding of MutS to DNA in the absence of nucleotide.
Since these experiments were performed under conditions where MutS is catalyzing ATP hydrolysis, the enhanced yields of specific MutS⅐heteroduplex complexes observed with endblocked DNAs is indicative of increased steady-state levels of these protein⅐DNA complexes. This effect is due at least in part to an increased lifetime of such complexes as judged by substrate challenge experiments. As shown in Fig. 1 (lower panel), MutS complexes with G-T heteroduplex lacking a terminal end-block dissociated rapidly and completely upon addition of a 600-fold excess of unlabeled G-T heteroduplex. By contrast, complexes with DNA containing a bi-terminal end-block did not dissociate significantly over a 30-min period under these conditions. MutS complexes with heteroduplex DNA blocked at one terminus behaved in an intermediate fashion. About half of the complexes dissociated upon challenge with unlabeled heteroduplex, but the remainder was stable over a 30-min period.
The results obtained with MutS and the heteroduplex containing a bi-terminal block are similar to those previously obtained with human MutS␣, where stabilization of specific complexes in the presence of ATP was also observed (15,16). However, results with the bacterial protein differ from those obtained with hMutS␣ for substrates with a single end-block. In contrast to the substantial stabilization afforded specific complexes of the bacterial protein in the presence of ATP, a single end-block is not sufficient to stabilize hMutS␣⅐heteroduplex complexes in the presence of this nucleotide (15,16). Our findings suggest that as in the case of hMutS␣, DNA termini are required for MutS dissociation in the absence of other repair activities, but they also imply that a single free terminus does not support efficient release of the bacterial protein.
Surface Plasmon Resonance Spectroscopy of MutS-DNA Complexes and the Effect of Adenine Nucleotides on Association and Dissociation-The entrapment of MutS on end-blocked heteroduplexes suggested that MutS might retain the ability to recognize a mismatch in the presence of ATP and Mg 2ϩ . We further examined the effects of adenine nucleotides on the dynamics of formation and dissociation of MutS⅐heteroduplex complexes using the real time technique of surface plasmon resonance spectroscopy (SPRS). Since analysis of protein-DNA interactions by SPRS is based on linkage of DNA to a streptavidin-derivatized sensor chip via a single terminal biotin (see "Experimental Procedures"), this experimental system is similar to that described above using a heteroduplex with a single end-block.
Representative sensorgrams performed in the presence of 50 mM KCl display three elements (Fig. 2) as follows: a binding phase during protein flow across the chip, a dissociation phase upon buffer wash, and second dissociation phase upon challenge with ATP⅐Mg 2ϩ . MutS readily associated with chipbound 41-base pair G-T heteroduplex in the presence of ADP⅐Mg 2ϩ (A), ATP⅐Mg 2ϩ (B), AMPPNP⅐Mg 2ϩ (C), or in the presence of ATP but in the absence of MgCl 2 (D). In each case a major component of the binding observed was mismatch-dependent, as demonstrated by parallel analyses with an otherwise identical homoduplex control (the effects of adenine nucleotides on heteroduplex and homoduplex affinities are considered in detail below). The demonstration of stable, specific complexes in the presence of ATP⅐Mg 2ϩ (B) confirms the results obtained from the gel shift experiments of Fig. 1. Stable, specific MutS⅐heteroduplex complexes were also formed in the presence of nonhydrolyzable AMPPNP or in the presence of ATP but in the absence of divalent cation (C and D), suggesting that formation of these complexes does not require ATP hydrolysis. Furthermore, the resistance of MutS⅐DNA complexes to dissociation upon subsequent ATP⅐Mg 2ϩ challenge was dependent upon the nature of the nucleotide present during the binding phase. Complexes prepared with heteroduplex or homoduplex DNA in the presence of ADP dissociated rapidly and completely upon subsequent challenge with ATP⅐Mg 2ϩ (A). Similar results were obtained if MutS was allowed to bind in the absence of nucleotide (not shown). However, MutS⅐DNA complexes prepared in the presence of ATP⅐Mg 2ϩ , AMPPNP⅐Mg 2ϩ , or ATP (no Mg 2ϩ ) were largely refractory to dissociation upon ATP challenge (B-D).

MutS Affinities for Homoduplex and Heteroduplex DNA Are Differentially Modulated by Adenine Nucleotides and Ionic
Strength-Dissociation constants (K d ) were estimated by SPRS by evaluating the extent of MutS⅐DNA complex formation as a function of MutS concentration in the absence or presence of adenine nucleotides (see "Experimental Procedures" and Fig.  3). In order to address the generality of the relative binding affinities obtained, these analyses were performed at several KCl concentrations, and the results are summarized in Table I and Fig. 4.
Comparison of adenine nucleotide effects as a function of KCl concentration (Table I and Fig. 4) indicates that MutS affinities for homoduplex and G-T heteroduplex DNAs are differentially modulated by the nature of the adenine nucleotide present. Whereas AMPPNP⅐Mg 2ϩ and ATP (no Mg 2ϩ ) have only a modest effect on homoduplex affinity as KCl concentration is increased (Table I and Fig. 4, upper panel), the presence of ADP results in a substantial affinity reduction that is particularly evident at salt concentrations in the physiological range. A similar KCl-dependent reduction in MutS homoduplex affinity  4 -6), both (lanes 7-9), or neither (lanes 1-3) duplex terminus. After preincubation for 10 min at 37°C to permit streptavidin-biotin conjugation, reactions were initiated by addition of MutS to 35 nM (monomer equivalents). After a 10-min additional incubation, reactions were terminated and subjected to gel shift analysis as described under "Experimental Procedures." Lower panel, gel shift reactions ("Experimental Procedures") contained 250 M ATP, 0.5 mg/ml streptavidin, and 1.6 nM 41-bp 32 P-labeled G-T heteroduplex with a 3Ј-terminal biotin at one (f), both (q), or neither (E) duplex terminus. Streptavidin was allowed to conjugate with DNA terminal biotin for 10 min at 25°C, and reactions were then initiated by addition of MutS to 100 nM (monomer equivalents). After an additional 10-min incubation at 25°C, unlabeled 41-bp G-T heteroduplex lacking a terminal biotin tag was added as competitor (1 M final concentration, time ϭ 0 on plot). Incubation was continued at 25°C, and reactions were sampled and subjected to gel shift analysis as indicated.
was observed in the presence of ATP⅐Mg 2ϩ where ATP hydrolysis is ongoing. Conversely, the highest heteroduplex affinities were observed in the absence of nucleotide or in the presence of ADP⅐Mg 2ϩ , and these affinities are relatively insensitive to increased KCl concentration (Table I and Fig. 4, lower panel). By contrast, heteroduplex affinity is reduced dramatically, particularly at salt concentrations in the physiological range, in the presence of AMPPNP⅐Mg 2ϩ , ATP (no Mg 2ϩ ), or under hydrolytic conditions in the presence of ATP⅐Mg 2ϩ .
The observation that MutS affinities for both homoduplex and heteroduplex are reduced to a similar degree in the presence of ATP⅐Mg 2ϩ may seem surprising, given the differential effects of ADP⅐Mg 2ϩ and AMPPNP⅐Mg 2ϩ on affinities for the two types of DNA. However, these findings are consistent with results of pre-steady-state studies, which have shown that the rate-limiting step for ATP turnover by MutS depends on the nature of the DNA cofactor present (25). In the presence of a homoduplex DNA, the rate-limiting step occurs subsequent to hydrolysis, implying significant binding site occupancy by ADP, but in the presence of heteroduplex DNA, this step occurs at or prior to covalent chemistry implying occupancy by ATP.
These results suggest that MutS has at least two distinct DNA-binding modes and that these can be modulated by the phosphorylation state(s) of bound adenine nucleotides. The differential effects of the di-and triphosphate on homoduplex   Fig. 4 shows that this effect is largely due to a dramatic reduction in heteroduplex affinity, consistent with previous work (14,16,21) demonstrating that nonhydrolyzable ATP analogs can promote release of a MutS homolog from a mispair. ADP Increases Mismatch Specificity Primarily by Reducing Homoduplex Interaction-The SPRS experiments above, which utilized 41-base pair DNA substrates, demonstrated a large enhancement of MutS heteroduplex specificity with increased KCl concentration provided that the experiments were done in the absence of nucleotide or in the presence of ADP. Further ADP-dependent specificity enhancement, beyond that observed in the absence of nucleotide, was demonstrable with larger DNAs capable of accommodating multiple MutS molecules. SPRS analysis of MutS binding at 150 mM KCl to a 200-bp G-T heteroduplex with a centrally located mismatch provided a clear example of two MutS-binding modes to a single DNA substrate (Fig. 5). Two distinct phases of MutS binding were resolved under these conditions: a high affinity phase with an apparent K d of 9 nM and a low affinity phase that did not saturate even at the highest MutS concentration tested (800 nM). Since binding to an otherwise identical A⅐T homoduplex indicated that the low affinity phase is due to nonspecific interactions (Fig. 5), the high affinity binding phase therefore corresponds to MutS⅐mismatch complex formation. The inclusion of ADP significantly reduced the nonspecific interaction of MutS with perfectly paired sequences as judged by reduced binding to control A⅐T homoduplex DNA, as well as by reduction of the low affinity binding phase to the G-T heteroduplex. By contrast, the nucleotide had little effect on mismatch recognition as the K d values for the specific phase of the two binding curves were similar (9.1 Ϯ 1.5 and 10 Ϯ 1.9 nM without and with ADP, respectively). ADP therefore increases MutS specificity by reducing the affinity of the protein for perfectly paired DNA sequences.
The Coupled Equilibria between Nucleotide and DNA Binding Can Be Used to Estimate Nucleotide K d Values-The experiments described above demonstrate that heteroduplex affinity is reduced substantially in the presence of AMPPNP⅐Mg 2ϩ or ATP (no Mg 2ϩ ). The simplest interpretation of these observations is that binding of the triphosphate is responsible for the reduction in heteroduplex affinity. However, since the adenine nucleotide concentration in these experiments was 1 mM and since the AMPPNP and ATP preparations used were contaminated at the several percent level by ADP, it could be argued that this effect not only requires binding of triphosphate but is also dependent on binding of the contaminating diphosphate as well. To address this question we have exploited the observation that adenine nucleotides have clear effects on the MutS binding maxima observed in SPRS experiments at the steady-state phase of the sensorgram trace (Fig. 2). The nature of this coupled equilibrium permits measurement of the apparent binding constant for the triphosphate with respect to its role in reducing heteroduplex affinity. This approach was used to determine the binding affinity of MutS for AMPPNP⅐Mg 2ϩ (Fig. 6) where sensorgram traces were obtained at a constant protein concentration using a 41-bp G-T heteroduplex. Increasing concentrations of AMPPNP resulted in a diminution of MutS binding to the DNA chip, an effect that was saturable and hyperbolic (Fig. 6, inset), yielding an apparent K d for AMPPNP of 0.31 Ϯ 0.02 M. This high affinity for AMPPNP implies that triphosphate binding alone is sufficient to reduce heteroduplex affinity because the concentration of contaminating ADP in this experiment would be in the low nM range, well below the dissociation constant of the MutS⅐ADP complex, which is in the M range. 2 End-blocked DNA Substrates Stimulate the ATPase Activity of MutS-As discussed above, the several models for MutS action differentially invoke ATP hydrolysis by the DNA-bound and free forms of the protein. To address the possibility that MutS⅐DNA complexes may support ATP hydrolysis, we examined the ATPase activity in the presence of DNA substrates with zero, one, or two streptavidin-biotin end-blocks (Fig. 7). If ATPase activity were dependent on the ability of MutS to dissociate from the DNA, MutS complexes trapped on endblocked substrates would be expected to have reduced ATPase activity relative to that of the free protein.
ATPase assay conditions were similar to those of the gel shift experiments of Fig. 1 and included streptavidin. As shown in Fig. 7, titration of MutS with each type of DNA substrate increased the rate of ATP hydrolysis in a manner similar to that observed previously (25). However, the degree of stimulation was reduced ϳ2-fold for the heteroduplex with a bi-terminal end-block as compared with DNAs that lacked or contained only one biotin-streptavidin terminal complex. In no case did the presence of terminal blocks inhibit ATPase activity. The stimulation of MutS ATPase activity by DNA with a bi-termi-nal end-block cannot be attributed to contaminating, underivatized DNA cofactor because the ATPase rate saturates and becomes independent of the concentration of the DNA cofactor. Furthermore, the gel shift studies shown in Fig. 1 indicate that virtually all of the biotinylated DNA substrate is conjugated with streptavidin under these conditions. Significantly, data from the gel shift studies and SPRS demonstrate that MutS remains associated with the DNA when the end(s) of the DNA are blocked. Together, these data strongly suggest that MutS can hydrolyze ATP while bound to DNA.

DISCUSSION
As discussed above, two classes of models have been proposed with respect to the role of the MutS ATPase in mismatch repair. One class invokes signaling between the mismatch and the strand signal via ATP-dependent movement of a MutS clamp-like structure along the helix, and two models of this type have been described. These differ in that the translocation model invokes directional tracking driven by ATP hydrolysis that occurs while MutS is DNA-bound (14,15), whereas the molecular switch model postulates free diffusion of the MutS⅐ATP complex along the helix, with ATP hydrolysis occurring subsequent to dissociation from DNA (16).
The second class of model argues that MutS remains bound to the mismatch during the course of repair and that interaction of the two DNA sites occurs by a DNA bending mechanism that brings the distal DNA sites into proximity (13). This model posits that ATP binding and hydrolysis by MutS functions in a kinetic proofreading capacity to enhance mismatch specificity of the protein.
As in the case of hMutS␣ (15,16,22), the nature of the interaction of bacterial MutS with end-blocked heteroduplexes is consistent with formation of clamp-like structure that can move along the helix in the presence of ATP. However, results with the bacterial protein differ from those obtained with the eukaryotic protein in the case of heteroduplex DNA with a single end-block. In contrast to hMutS␣ where bi-terminal end-blocks are required to prevent ATP-promoted release of the protein from a heteroduplex (15,16,22), a single end-block is largely sufficient to prevent release in the case of the bacterial protein (Fig. 1). Although this sort of behavior can be viewed as consistent with a mechanism that supports directional movement, it is seemingly incompatible with a mechanism that invokes free diffusion of a protein clamp along the helix. The sliding clamp model also postulates that ATP hydrolysis occurs upon DNA release. By using oligonucleotide heteroduplexes with sterically blocked ends as an ATPase cofactor, we have observed stimulation of ATPase activity, not the diminution predicted by this model. The gel shift experiments (Fig. 1) and SPRS studies (Fig. 2) performed with substrates identical to those used for ATPase assay indicate that MutS remains DNA bound while undergoing multiple ATP hydrolytic turnovers. Together, these observations seem more consistent with an active role for ATP hydrolysis in the function of DNA-bound MutS as opposed to the passive ATP-binding role envisioned in the molecular switch model.
Adenine nucleotides differentially modulate MutS affinities for heteroduplex and homoduplex DNAs in a manner that can dramatically alter the specificity of the protein for a mismatched base pair. Conditions where nucleotide-binding site occupancy is restricted to the triphosphate (i.e. in the presence of AMPPNP⅐Mg 2ϩ or ATP (no Mg 2ϩ )) result in a dramatic reduction in heteroduplex affinity but have only a modest effect on homoduplex affinity. As a consequence, specificity of MutS for a mismatch is abolished, or nearly so, at salt concentrations of 100 mM or above (Fig. 4, Table I). Conversely, restriction of nucleotide-binding site occupancy to ADP⅐Mg 2ϩ has little effect on heteroduplex affinity but confers reduced homoduplex affinity and consequently high mismatch specificity. This effect was particularly evident with a 200-bp heteroduplex DNA capable of binding multiple MutS molecules in nonspecific manner (Fig.  5). It is important to note that a similar loss of heteroduplex specificity in the presence of ATP (no Mg 2ϩ ) has been observed with yeast MutS␣, although reduction of homoduplex affinity by ADP has not been observed in this system (26).
Interestingly, in the presence of ATP⅐Mg 2ϩ where hydrolysis is occurring, apparent homoduplex affinity is similar to that observed in the presence of ADP⅐Mg 2ϩ , whereas heteroduplex affinity is similar to that determined in the presence of the triphosphate under conditions where hydrolysis is blocked (i.e. in the presence of AMPPNP⅐Mg 2ϩ or ATP (no Mg 2ϩ ), Fig. 4). At physiological salt concentration (150 mM KCl), this results in a specificity intermediate between those observed when binding site occupancy is restricted to the diphosphate or the triphosphate. There are several potential explanations for this interesting effect. As noted above, pre-steady-state studies have shown that the rate-limiting step for ATP hydrolysis in the presence of a homoduplex cofactor occurs subsequent to hydrolysis, implying occupancy of at least one binding site by ADP (25). However, in the presence of heteroduplex, the rate-limiting step occurs at or prior to covalent chemistry, indicating significant binding site occupancy by ATP. Thus, in the presence of heteroduplex and ATP⅐Mg 2ϩ , MutS might be expected to behave as the triphosphate-bound form, whereas in the presence of homoduplex and ATP⅐Mg 2ϩ it would behave as the diphosphate form. This is consistent with what we observe, but the effects of ATP⅐Mg 2ϩ on heteroduplex/homoduplex affinities could also be indicative of a MutS conformational state in which one class of nucleotide-binding site is occupied by ATP and a second by ADP.
These nucleotide effects on MutS specificity have implications for the postulated kinetic proofreading role of adenine nucleotides in the mismatch verification model (13). This model suggests that ATP binding by DNA-bound MutS preferentially reduces the affinity of the protein for homoduplex DNA sites, thereby resulting in enhanced mismatch specificity. We do not find this to be the case. With the exception of specificity values obtained at 50 mM KCl, which only varied over a 3-fold range in the absence or presence of various adenine nucleotides (Table  I), heteroduplex specificity was abolished in the presence of AMPPNP⅐Mg 2ϩ or ATP (no Mg 2ϩ ) at KCl concentrations of 100 mM or higher. In fact, at KCl concentrations of 100 mM or above, the highest specificity was observed in the absence of nucleotide or in the presence of ADP⅐Mg 2ϩ . The dramatic and differential effects of adenine nucleotides on heteroduplex and homoduplex affinities imply that MutS is capable of several distinct modes of DNA interaction that are modulated by ATP binding and hydrolysis. This idea is also consistent with the finding that the kinetic stability of MutS⅐heteroduplex complexes upon ATP challenge depends on the nature of the nucleotide present during the DNA binding step (Fig. 2). Although the molecular bases of the nucleotide effects on mismatch and nonspecific affinities are unclear, the structure of the bacterial MutS⅐heteroduplex complex may bear on this point (27,28). The amino-terminal 800 residues of MutS compose five domains (mismatch binding, connector, core-levers, DNA clamp, and ATPase) that dimerize into a ⌰-like structure that possesses two channels (27,28). One channel formed by the connector and DNA clamp is occupied by heteroduplex, and the second channel, which is unoccupied in the complex and of unknown function, could also accommodate a region of helix (27,29,30). A two DNA-binding site mechanism involving a sliding clamp and a static DNA-binding site has been postulated to account for MutS movement in the directional translocation model (15). An alternate possibility is based on the fact that the MutS ATPase is member of the ABC ATPase family which also includes Rad50. Although structural information on the MutS-ATP complex is not available, the Rad50-ATP complex has been solved (31). The structural homology of MutS and Rad50 ATPase domains has led to the suggestion that ATP binding by MutS could lead to repositioning of the mismatch binding domain to yield a simple sliding clamp structure (30) like that proposed by Fishel and colleagues (16).
Unfortunately, little is known concerning the possible states of occupancy of the MutS nucleotide-binding sites. This is an important question since the nucleotide binding center of each MutS monomer subunit can potentially exist in three possible states: empty, occupied by triphosphate, or occupied by diphosphate. Given that MutS exists as dimers and tetramers in solution (9,32,33) and the extensive evidence indicating interaction of nucleotide and DNA binding center(s), the number of potential conformational states available to the protein may be substantial (29). In fact, nucleotide occupancy has been established only in the case of the crystallographic ADP⅐MutS⅐heteroduplex complexes obtained with carboxyl-terminal truncated forms of the Taq and E. coli proteins. These results differ, however; in the Taq complex both nucleotide-binding sites are occupied by ADP (13), whereas in the E. coli structure one site is occupied and one is empty (28). Elucidation of the possible states of nucleotide occupancy in the absence and presence of DNA cofactors should further clarify the role of the MutS ATPase in mismatch repair.