Role of MutS ATPase Activity in MutS,L-dependent Block of in Vitro Strand Transfer*

In addition to mismatch recognition,Escherichia coli MutS has an associated ATPase activity that is fundamental to repair. Hence, we have characterized two MutS mutant gene products to define the role of ATP hydrolysis in homeologous recombination. These mutants, denoted MutS501 and MutS506, have single point mutations within the Walker A motif, and rate constants for ATP hydrolysis are down 60–100-fold as compared with wild type. Both MutS501 and MutS506 retain mismatch binding and, unlike wild type, fail to relinquish this specificity in the presence of ATP, adenosine 5′-O-(thiotriphosphate), and adenosine 5′-(β,γ-imino)triphosphate. Both MutS501 and MutS506 blocked the level of strand transfer between M13 and fd DNAs. The level of inhibition varied between the mutants and corresponded with the relative affinities to a G/T mispair. Neither MutS501 nor MutS506, however, would afford complete block of full-length heteroduplex in the presence of MutL. DNase I footprinting data are consistent with these results, as the region of protection by MutS501 and MutS506 was unchanged in the presence of ATP and MutL. Taken together, these studies suggest that 1) MutS impedes RecA-mediated homeologous exchange as a distinct mismatch-provoked event and 2) the role of MutL is coupled to MutS-dependent ATP hydrolysis. These observations are in good agreement with the present model for E. coli methyl-directed mismatch repair.

In addition to its role in replication fidelity, mismatch repair contributes to genome stability by controlling the level of recombination between closely related sequences (1)(2)(3)(4)(5)(6)(7)(8)(9). In Escherichia coli, MutS and MutL act to block recombinant yield when phenotypic selection arises in the vicinity of heteroallelic markers (10). Recent work from Zahrt and Maloy (11) provided more evidence implicating mismatch repair in recombination. Specifically, they showed that the efficiency of chromosomal gene transfer between Salmonella typhimurium and the closely related Salmonella typhi is low due to mismatch repair. Zahrt and Maloy (11) attributed this barrier of genetic exchange to DNA divergence and the ability of MMR to correct this event in favor of the recipient strand. Abdulkarim and Hughes (12) also showed that the mutHLSU genes increased the fidelity of exchange 1000-fold between the TufA and TufB translation factors, whose sequences share 99% sequence identity. Mismatch repair in this case ensures that the integrity of the genome is preserved by allowing these genes to evolve in concert.
The first biochemical evidence to implicate mismatch repair in homologous recombination was provided by the in vitro RecA-catalyzed three-strand transfer reaction. E. coli MutS was shown to inhibit exchange between M13 and fd DNAs whose sequence heterology is ϳ3% at the nucleotide level (13). Since MutS was without effect on M13-M13 or fd-fd exchange, we attribute the block to the occurrence of mismatched base pairs in newly formed heteroduplex.
MutL enhanced the MutS-dependent block of M13-fd exchange. Indeed, these proteins completely abolished full-length heteroduplex between these DNAs and is reminiscent of the proposed role of these activities in MMR (14 -19). MutL is believed to add to the MutS mismatch complex in an ATP-dependent fashion. Evidence for this stems from work by Grilley et al. (20), who showed the region of DNase I protection by MutS and ATP is extended by MutL. In the experiments described here, the role of ATP hydrolysis by MutS is examined on RecA-strand transfer. Previous work by Wu and Marinus (21) described dominant negative mutants of MutS from MNNG treatment and showed a large fraction of the isolates were missense mutations within the nucleotide binding domain (21). It is here that we have characterized two of these mutants, MutS501 and MutS506, and show that 1) mismatch binding is retained, 2) ATP hydrolysis is reduced, and 3) mismatch correction with MutS Ϫ extracts is not restored.
These mutants also blocked RecA-catalyzed strand transfer between M13 and fd DNAs. That MutS501, and to a lesser extent MutS506, inhibited M13-fd and not M13-M13 exchange, suggests that mismatch recognition precludes RecA-dependent branch migration in regions of nonhomology. Moreover, that both mutants failed to completely block strand transfer in the presence of MutL suggests that the barrier to homeologous recombination is dependent upon ATP hydrolysis.

EXPERIMENTAL PROCEDURES
E. coli proteins RecA, SSB, 1 and MutL were purified as described previously (13). Restriction endonucleases SnaBI and MspI and S1 nuclease were purchased from U.S. Biochemical Corp./Amersham Pharmacia Biotech. DNase I from bovine pancreas was from Worthington.
Other Materials-Proteinase K, phosphocreatine kinase, and phosphocreatine were purchased from Sigma. Oligonucleotides were from Oligos Etc. Replicative form and single-stranded M13 and fd bacteriophage DNAs were prepared as described (13). The replicative forms of M13 and fd DNAs were linearized with SnaBI restriction endonuclease. Linear DNA (50 pmol of 5Ј-ends) was 5Ј-end-labeled with [␥-32 P]ATP using T4 polynucleotide kinase and purified as described previously (13).
DNA Substrates-Oligodeoxynucleotides were purchased from Oligos Etc. Purified oligomers were annealed and isolated as described previously (22). Covalently closed circular G/T heteroduplex used in the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  MMR assays was prepared as described previously (23)(24)(25)(26). Repair was scored by determining the fraction of molecules rendered sensitive to XhoI cleavage due to T 3 C correction on the unmethylated strand.
Determining Kinetic Parameters K m and k cat for MutS-MutS ATPase activity was carried out as described by Hughes and Jiricny (28). Reactions (24 l) were performed in 50 mM HEPES-KOH, containing 0.75 Ci of [␥-32 P]ATP and 2, 4, 8, 18, 25, 37.5, 50, or 75 M ATP. MutS (0.33 M; MutS501 and MutS506, 2 M) was added to initiate the reaction at 37°C, and aliquots (2 l) were quenched (4 l; 8.3 mM EDTA, pH 8.0; 2.0 l formamide) and resolved for 0.5 h at 25-mA constant current as described in the figure legends. Rates for ATP hydrolysis were determined by quantitating the amount of labeled P i liberated as described above. Kinetic determinations were accomplished by two methods from three independent experiments (29).
Effects of MutS501, MutS506, and MutL on Strand Exchange-Reactions were carried out as described previously, except MutS (75 nM) and MutL (75 nM) were added just prior to SSB and ATP (13). Control reactions were supplemented with MutS and MutL diluent buffer.
For reactions with 32 P-end-labeled DNA (ϳ50 ng; 35,000 cpm) concentration of linear duplex was adjusted with cold homoduplex DNA.
Labeled duplex (75 pM; 0.5 l) was premixed with unlabeled duplex (10.4 M; 1.2 l) and then incubated with the RecA-single-stranded DNA reaction as described.

Purification of Hexameric Histidine Recombinant MutS-
Both MutS mutants (501 and 506) and wild-type were overexpressed and purified using a N-terminal hexameric histidine tag. Induction with 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside resulted in high levels of expression, since ϳ30% of the total E. coli lysate ( Fig. 1.) was MutS. However, a vast majority (ϳ80%) of the recombinant protein precipitated as inclusion bodies (30). This was corrected by reducing the isopropyl-1thio-␤-D-galactopyranoside to 0.05-0.1 mM and lowering the growth temperature to 25°C. Recombinant His 6 -tagged proteins were purified to greater than 95% purity using nickel affinity chromatography (Fig. 1). The yield from 4-liter preparations was 40 -50 mg of protein at 2.5-3.0 mg/ml.
Purification of wild-type MutS (31) was done in parallel to examine the effects of the N-terminal tag on mismatch repair. In complementing MutS Ϫ extracts, His 6 -tagged MutS was as proficient in repair as wild-type biochemical properties (mismatch binding/repair) for wild-type and His 6 -tagged MutS. This is in good agreement with studies from previous laboratories (32,33) illustrating no detectable perturbations due to the N-terminal tag (34).
Gel Retardation Assay/NheI Endonuclease Assays-Both MutS501 and MutS506 were examined for mismatch recogni- Using the QIAexpressionist System (Qiagen), MutS501 and MutS506 were cloned into the plasmid pQE10 to generate recombinant His 6tagged proteins as described under "Experimental Procedures." Lanes 1, 4, and 7, uninduced lysates from wild-type MutS, MutS501, and MutS506, respectively. Lanes 2, 5, and 8, lysates after induction with isopropyl-1-thio-␤-D-galactopyranoside at 25°C. Lanes 3, 6, and 9, nickel affinity purification of His 6 -tagged recombinant proteins (2.5 g). All purified samples had an apparent molecular mass of 97 kDa, similar to nontagged MutS. tion by the ability to retard a 51-mer G/T-containing oligonucleotide. Titration of His 6 -tagged MutS revealed a K D for a G/T mispair at 40 nM ( Fig. 2a) (15). MutS501 mismatch was marginally affected with a K D of 55 nM. Mismatch recognition by MutS506, however, was down almost 3-fold with a K D of 125 nM. The gel retardation pattern was identical in nature to both tagged and nontagged wild-type MutS (data not shown). Discrimination for G/T heteroduplex over homoduplex DNA was 10-fold for both wild-type and MutS501 (Fig. 2b).
Previous studies have shown E. coli MutS leaves the mismatch in the presence of ATP (31,35). This nucleotide-driven process was tested for wild-type, His 6 -tagged wild-type, MutS501, and MutS506 in the presence of 2 mM ATP. Like nontagged wild-type, His 6 -tagged MutS failed to bind the mismatch site in the presence of ATP (Fig. 2B). In contrast, MutS501 and MutS506 retained mismatch binding. Both mutants essentially bound a G/T mismatch with the same affinity as that when ATP was not present. Nonspecific binding (G:C homoduplex) by these proteins was slightly affected by ATP. Further studies with nonhydrolyzable ATP analogs ATP␥S and AMP-PNP revealed similar results.
In a related experiment, we tested binding specificity of these mutants by examining the relative accessibility of a restriction site adjacent to the mismatch (31). As detailed in Fig.  3a. G/T-containing duplex DNA was protected against digestion by NheI at increasing concentrations of His 6 -tagged MutS and abolished by 2 mM ATP. G:C homoduplex was sensitive to cleavage by NheI at the highest level of MutS, demonstrating the mismatch-specific nature of protection.
MutS501 protected against endonucleolytic attack in a sim-ilar fashion as wild-type. MutS506 was less efficient in its ability to protect against cleavage. As illustrated in Fig. 3b, a 2-fold increase in sensitivity to NheI digestion was seen for MutS506 as compared with MutS501. Even at substoichiometric amounts of protein, both MutS501 and MutS506 remained bound to the mismatch site in the presence of ATP. These results further reinforce the gel retardation data in illustrating  the mismatch specificity of these mutants in the presence of ATP.
Kinetics of ATP Binding/Hydrolysis-ATPase activity was examined for MutS501 and MutS506 to further characterize the dominant negative phenotype (21). Kinetic parameters, K m and k cat were determined for wild type, MutS501, and MutS506 to verify their role in MMR. As shown in Fig. 4, E. coli MutS hydrolyzes ATP to ADP ϩ P i and thus corroborates the early findings of Grilley et al. (20). From three independent experiments, k cat and K m values were determined, and the data were analyzed by both Lineweaver-Burk and Eadie-Hofstee double reciprocal plots. As a comparison, values for MutS from S. typhimurium are presented (29). In terms of k cat , E. coli wildtype MutS is approximately 30 times faster. This rate of hydrolysis does not seem to be dependent upon better binding, since the K m values are virtually identical (Table I). Moreover, it is unlikely that this difference in k cat is due to an ATPase contaminant in the E. coli preparations, because both MutS501 and MutS506 possessed different rates of ATP hydrolysis.
Based upon the mutations (29) defining MutS506 and MutS501, we observed a dramatic drop in ATPase activity (Fig.  5) as compared with wild type. A turnover number of 7.4 min Ϫ1 for wild type was down 60 -100-fold for MutS501 and MutS506 with values of 0.11 and 0.08, respectively (Table I). K m values were slightly affected with a 2-3-fold decrease in binding as compared with wild-type. It is important to note that the ATPase activities presented here were measured in the absence of DNA. We observed no difference in ATP hydrolysis in the presence of duplex DNA (data not shown), but effects of a mismatch were not tested.
Mismatch Correction-Mismatch repair was examined to compare the repair efficiency of MutS501 and MutS506 with His 6tagged wild-type by complementing mutS-(MutS201::Tn5) extracts. Maximum repair of closed circular G/T heteroduplex by His 6 -tagged MutS occurred at 0.2 pmol of protein (Table II). This corresponded well with wild-type MutS as optimum repair occurred between 0.2 and 0.8 pmol of MutS. Unlike wild type, His 6tagged MutS started to inhibit repair above 0.3 pmol. 2 Neither MutS501 nor MutS506 was able to complement MutS Ϫ extracts (Table II). Increasing the protein concentration an additional 5-fold resulted only in minimal repair (Ͻ7 fmol). These proteins also failed to block repair of wild-type MutS when added in a 1:1 ratio at 0.4 pmol of total protein (data not shown).
MutS501 and MutS506 Inhibit M13-fd RecA-catalyzed Strand Exchange-To better define the mechanism behind the MutS,L block of RecA-mediated strand transfer, we examined the role of MutS ATPase activity in this process. Functional His 6 -tagged MutS was therefore tested against the ATPase-defective mutants in the ability to block M13-fd exchange (13). As illustrated in Fig. 5., His 6 -tagged wild type limits full-length heteroduplex formation at increasing MutS concentrations. This concentration-dependent block is in good agreement with previous studies on nontagged MutS and again illustrates no unusual perturbations due to the N-terminal histidines (34). Maximal inhibition was observed at a MutS to mismatch ratio of 1:2.5. MutS501 shows a similar ability to inhibit strand transfer as wild-type MutS. The extent of inhibition by MutS506, however, was down approximately 2-3-fold.
Effects of MutS,L on Homologous Exchange-A concentration of MutS to mismatch of 1:4 yielded a 2-fold decrease in full-length heteroduplex. Consistent with previous studies (13), the rate of strand transfer between M13 and M13 was about 3 times faster than M13-fd (Fig. 6, a and b), thus illustrating that homology search by RecA is sensitive to non-Watson-Crick base pairing. M13 duplex was taken up by the RecA filamented single-stranded circular M13 within the first 10 min of incubation. The population of branched intermediates for M13-M13 exchange appeared more dispersed, indicating no barrier to branch migration. MutS failed to change this distribution in the homologous reaction (Fig. 6a), an observation consistent   with the formation of normal canonical base pairing. Conversely, branched intermediates between M13 duplex and fd single-strands accumulated as a defined species that appeared to stabilize when MutS was present (Fig. 6b). His 6 -tagged MutS and MutL revealed a similar kinetics profile as that observed for nontagged wild type (13). These combined repair activities completely blocked formation of fulllength heteroduplex DNA between M13 and fd (Fig. 7). This barrier to RecA-catalyzed strand transfer was MutS-dependent, since MutL alone had no effect on either M13-M13 or M13-fd exchange. As illustrated in Fig. 6b, it appears that MutL participates in this block by destabilizing branched intermediates that have accumulated in the presence of MutS (13).
Effects of MutL on RecA-catalyzed Strand Transfer in the Presence of Mutants MutS501 and MutS506 -In examining the role of mismatch repair proteins in homeologous recombination, we tested the effects of MutS501 and MutS506 on both M13-M13 and M13-fd exchange in the presence of MutL. As illustrated in Fig. 8 and 9a, we observed that MutS501 inhibits M13-fd exchange similar to that observed for wild type. In contrast, the level of inhibition by MutS506 was down 2-fold (Fig. 9b). This inhibition of M13-fd exchange was dependent upon heteroduplex formation as both mutants contribute little to the block of M13-M13 strand transfer (data not shown).
MutS501 did not completely block full-length heteroduplex in the presence of MutL during M13-fd exchange. The extent of product formation was the same whether MutL was present or not ( Figs. 8 and 9a). It appears from these results that the effect of MutL requires MutS and ATP hydrolysis. In addition, it suggests the block by MutL is mismatch-dependent. A similar loss of MutL function was observed with MutS506. The amount of heteroduplex DNA, however, was 2-3 times higher for MutS506 than MutS501 (Figs. 8 and 9b).
Footprinting of Wild-type MutS, MutS501, and MutS506 at a Mismatch-We further characterized the DNA binding properties of MutS501 and MutS506 using DNase I footprinting. Grilley et al. (20) showed that MutL will extend the MutS footprint at a G/T mispair in the presence of ATP. MutS501 protein retains its mismatch binding specificity and protects a region of ϳ20 base pairs spanning the G/T mismatch (Fig. 10a). This footprint is identical to that observed for wild type. Consistent with previous studies with gel retardation, MutS501 remained at the mismatch site in the presence of ATP. His 6tagged wild-type MutS, however, does not protect against DNase I digestion in the presence of ATP (Fig. 10b). This observation is different from Grilley et al. (20) with nontagged MutS. We do, however, point out that in the studies of Grilley et al. they did observe a decrease in protection due to ATP. Evidence here and elsewhere suggests MutS leaves the mismatch in the presence of ATP (31). MutS506 bound specifically at the G/T mismatch and produced a different protection pattern than MutS501 or wild-type (Fig. 10a). A region of approximately 8 -10 bp 5Ј to the mismatch is more sensitive to DNA cleavage with MutS506 than MutS501 or wild-type (Figs. 8 and 10a).
To help clarify MutL function in recombination, we examined this activity with MutS501 and MutS506 in the presence of ATP. Consistent with studies of Grilley et al. (20) we see that His 6 -tagged wild-type MutS protection from DNase I increases upon the addition of MutL and ATP (Fig. 10b, right). Neither MutS501 nor MutS506 extended the region of protection in the presence of MutL and ATP. Both mutants remained in the vicinity of the G/T mispair. Studies with ATP alone revealed a similar pattern of protection (Fig. 10b, middle). This is in good agreement with the results from the gel retardation assays, where the presence of ATP failed to disrupt mismatch binding. DISCUSSION Here, we present data that define the enzymatic properties of MutS mutants MutS501 and MutS506 in MMR and recom- bination. Gel retardation analysis shows both MutS501 and 506 retain the ability to bind mismatches. Indeed, MutS501 was as efficient as wild-type in binding a G/T mispair. MutS506 binding specificity, however, was reduced almost 3-fold. This difference in mismatch binding is consistent with the in vivo studies of Wu and Marinus (21), who showed repair of a 2-base deletion was more efficient in mutS501 than mutS506 mutant strains. That both mutS501 and mutS506 are dominant over wild type suggests mismatch binding is not the only criterion defining this phenotype. We do believe the results presented here offer a plausible explanation for the relative difference in repair efficiencies by these mutants (21).
The initial goal of these studies was to define the role of MutS,L during RecA-catalyzed strand transfer between closely related DNAs. Previous work has shown that the formation of mismatched base pairs in heteroduplex is modulated by MutS,L (13). In E. coli, MutS initiates repair by binding to the mismatch site. In subsequent steps, incision at a hemimethylated d(GATC) site is dependent upon MutS, MutL, MutH, and ATP hydrolysis. It is known that a persistent and strandspecific nick will direct repair and thus bypass the MutH requirement (15,36,38,39). That MutS,L and ATP hydrolysis are still required for excision reflects an intimate interaction between these proteins in governing a repair event. In this study, we examined how these repair activities interact during strand transfer by examining the role of ATP hydrolysis by MutS in this process. We demonstrate that the extent of exchange between homeologous DNAs is affected by the ability of MutS to bind newly formed mismatches. More precisely, strand exchange between M13 and fd DNAs is hampered by both MutS501 and MutS506. Since both mutants have reduced ATPase activity (60 -100-fold) and failed to leave the mismatch in the presence of ATP, we attribute this block of strand exchange to mismatch binding.
Based on studies from Cox (40,41), RecA promotes the exchange of homologous DNAs at a rate of 350 bp min Ϫ1 . Given the size of M13 phage DNA, RecA should drive the reaction to completion in approximately 18 min. Indeed, we observed 100% conversion of linear duplex to nicked circles between M13 DNAs in 20 min. The rate of branch migration for M13-fd exchange was slower at ϳ105 bp min Ϫ1 . Because of this difference in rates, as manifested by regions of nonhomology, we believe MutS has time to test for the presence of non-Watson-Crick base pairing. Indeed, it would be fair to assume that mismatch binding interferes with normal RecA reassociation in regions where the density of mispairs is energetically unfavorable (13).
The effect by MutL on homologous strand transfer could result from MutS-dependent translocation along the DNA via a-shaped loop structures, a process that would help stabilize MutS,L on duplex DNA after mismatch binding. Recent studies from Allen et al. (31) are consistent with this idea. The rate of loop formation catalyzed by MutS and ATP increased 2-fold in the presence of MutL. Moreover, the ATPase-defective mutants MutS501 and MutS506 failed to show the enhanced block of full-length heteroduplex between M13 and fd DNAs in the presence of MutL. It appears the role of MutL is confined to a step following mismatch recognition and is probably coupled to MutS ATPase activity. DNase I footprinting studies presented here are consistent with this line of reasoning. Unlike wild-type MutS, both mutants failed to expand the region of nuclease protection in the presence of ATP and MutL. However, it is not immediately obvious whether binding or hydrolysis is required for MutL recruitment to the complex. MutS could in principle undergo a conformational change once ATP is bound. Mismatch binding would then be lost, and the hydrolytic step would ensure that MutS remains in this state, thus facilitating DNA translocation over mismatch recognition. It is also likely that MutL somehow contributes to this form of MutS by stabilizing the complex on duplex DNA (28). Taken together, these results suggest MMR exerts a similar strategy in testing for the occurrence of mismatched base pairs during RecA-catalyzed strand transfer.
Certainly, further studies are needed to help define other roles of mismatch repair in recombination. Based on studies in yeast S. cerevisiae two MutS homologues, MSH4 and MSH5 are implicated in meiotic recombination (42)(43)(44). Just recently, the human homologue of yMSH4 was identified (45). This, along with studies illustrating yMSH2-specific binding to Holliday junctions (46), could implicate further roles for mismatch repair in genome stability.