Contribution of Human Mlh1 and Pms2 ATPase Activities to DNA Mismatch Repair*

MutL (cid:1) , a heterodimer composed of Mlh1 and Pms2, is the major MutL activity in mammalian DNA mismatch repair. Highly conserved motifs in the N termini of both subunits predict that the protein is an ATPase. To study the significance of these motifs to mismatch repair, we have expressed in insect cells wild type human MutL (cid:1) and forms altered in conserved glutamic acid residues, predicted to catalyze ATP hydrolysis of Mlh1, Pms2, or both. Using an in vitro assay, we showed that MutL (cid:1) proteins altered in either glutamic acid residue were each partially defective in mismatch repair, whereas the double mutant showed no detectable mismatch repair. Neither strand specificity nor directionality of repair was affected in the single mutant proteins. Limited proteolysis studies of MutL (cid:1) demonstrated that both Mlh1 and Pms2 N-terminal domains undergo ATP-induced conformational changes, but the extent of the conformational change for Mlh1 was more apparent than for Pms2. Furthermore, Mlh1 was protected at lower ATP concentrations than Pms2, suggesting Mlh1 binds ATP with higher affinity. These findings imply that ATP hydrolysis is required for MutL (cid:1) activity in mismatch repair and that this activity is associated with differential conformational changes in Mlh1 and Pms2. Mismatch (MMR) replication

Mismatch repair (MMR) 1 helps to protect the genome from replication errors caused by DNA polymerases. Its pivotal role as caretaker of genome stability is exemplified by HNPCC, a hereditary predisposition to colon, rectal, and other cancers associated with germ line mutations in MMR (1).
In Escherichia coli, the major players in the pathway are known to the extent that a full in vitro mismatch repair was reconstituted from purified components (2). A MutS homodimer recognizes and binds a mismatch and recruits a MutL homodimer to form an ATP-dependent ternary complex with DNA (3). MutL recruits and activates MutH (4), an endonuclease that introduces a specific nick on hemimethylated GATC sites, thus providing the strand discrimination signal required to ensure repair of the nascent strand. MutL also activates UvrD (5), a helicase that unwinds the duplex DNA from the nick, providing a single strand DNA substrate for an exonuclease. The repair pathway is bidirectional in that repair can initiate from GATC sites located either 5Ј or 3Ј to the mismatch (6). As a result, of the four exonucleases that have been found to participate in the pathway, two have 5Ј to 3Ј polarity, and the others have 3Јto 5Ј polarity (7). The single strand gap formed by the exonuclease is filled in by DNA polymerase III, and the nick is sealed by DNA ligase.
The pathway in eukaryotes is less understood than in E. coli (reviewed in Refs. 8 and 9). Instead of one MutS, there are three MutS-like proteins that are involved in mutation avoidance, Msh2, Msh3, and Msh6. Msh2 and Msh6 associate to form MutS␣, the major MutS activity in MMR (reviewed in Ref. 9). Illustrating a similar level of complexity, there are four MutL homologs in mammals: Mlh1, Pms2 (closest to Pms1 in yeast), Pms1, and Mlh3. Mlh1 and Pms2 (or Mlh1 and Pms1 in yeast) associate to form MutL␣, the major MutL activity in the pathway (9 -11). Subsequent to mismatch binding by MutS␣, a ternary complex involving MutL␣ forms at the mismatch (12).
The downstream events of MMR and the identity of the proteins involved are even less characterized. No homologue has been identified for MutH, and the mode of strand discrimination is not clear. ExoI, a 5Ј to 3Ј exonuclease, is thought to participate in MMR based on physical and genetic interactions with Mlh1 and Msh2 in yeast and human cells (13)(14)(15)(16). Proliferating cell nuclear antigen, the sliding clamp of DNA polymerase ␦, has been implicated to act in MMR prior to and during the DNA synthesis step (17)(18)(19)(20)(21)(22)(23). The role of proliferating cell nuclear antigen in early MMR may provide a clue to the mode of strand discrimination (24).
MutL has been shown to be a member of the GHL superfamily of ATPases, which includes the homodimeric E. coli gyrase B and eukaryotic Hsp90 (reviewed in Ref. 25). The ATPase motif that resides in the N-terminal domain is distinct in organization and catalytic mechanism from "Walker motif " ATPases, including MutS proteins. Upon ATP binding, a conformational change is associated with dimerization of the N termini of the two protomers (26 -28). ATP hydrolysis, which is dependent on this dimerization (28,29), is catalyzed by a glutamic acid residue, which in gyrase B was shown to act as a general base that activates a water molecule for nucleophilic attack of the ␥-phosphate of ATP (30). Mutation of this glutamic acid in gyrase B affected ATP hydrolysis but not binding (30), whereas the analogous mutation in MutL while eliminating hydrolysis activity also had a small effect on binding (28). MutL is a weak ATPase (31), in support of a model in which MutL acts to recruit and activate downstream effector proteins following mismatch binding by MutS (32).
The MutL-like ATPase domain is highly conserved, suggesting that eukaryotic MutL homologues also possess ATPase activity. Indeed, the N-terminal fragment of human Pms2 has been shown recently to have an ATPase activity catalyzed by the conserved glutamic acid (33). Studies in bacteria and yeast indicate the importance of the ATPase domain in MMR (34,35).
In this report, we present results demonstrating the importance of the ATPase domains of human MutL␣ during MMR. Furthermore, the results suggest an asymmetry within human MutL␣ as reflected by differential conformational changes in response to adenine nucleotides.

EXPERIMENTAL PROCEDURES
Preparation of MP-1 Cell Line-Mouse embryonic fibroblasts deficient for both Mlh1 and Pms2 (MP-1) were derived by crossing doubly heterozygous animals in a C57BL/6 genetic background. Embryos were harvested from timed pregnancies, homogenized, and placed in culture as described (36). Immortalized clones arising spontaneously after senescence were pooled and expanded. Genotyping to identify the Mlh1Ϫ/Ϫ; Pms2Ϫ/Ϫ cells was performed by using a PCR assay (37,38) with high molecular weight DNA harvested from surplus tissue and subsequently was confirmed in the isolated cell line.
Preparation of Extracts-Cytoplasmic extracts were prepared as described (39). Cells were grown in up to 40 P-150 plates; trypsinized; washed once with PBS, once with isotonic buffer (20 mM HEPES, pH 7.9, 250 mM sucrose, 5 mM KCl, 1.5 mM MgCl 2 , 1 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl fluoride), and then with hypotonic buffer (20 mM HEPES, pH 7.9, 5 mM KCl, 1.5 mM MgCl 2 , 1 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl fluoride); resuspended in hypotonic buffer at a density of 7 ϫ 10 7 cells/ml; and allowed to swell on ice for 15 min. The swollen cells were lysed in a Dounce homogenizer using pestle B and then incubated on ice for 30 min and centrifuged at 2000 ϫ g for 10 min at 4°C. The supernatant was recentrifuged at 12,000 ϫ g for 10 min at 4°C. The supernatant from this step, which is the cytoplasmic extract, was aliquoted and frozen in liquid nitrogen.
Preparation of DNA Substrates-All substrates were prepared from phage DNA kindly provided by Dr. Paul Modrich. A 5Ј nick G-T mismatch DNA substrate was prepared from f1MR1 phage ssDNA and f1MR3 phage dsDNA linearized with Sau96I (40). The nick is located 125 bp 5Ј to the mismatch on the complementary strand. Repair of the nicked (complementary) strand renders the substrate sensitive to HindIII. Repair of the intact (viral) strand renders the substrate sensitive to XhoI. A 5Ј nick one-base loop substrate was prepared from f1MR22 phage ssDNA and f1MR21 phage dsDNA (41) linearized with Sau96I. The nick is located 125 bp 5Ј to the mismatch on the complementary strand. Repair of the nicked strand renders the substrate sensitive to XcmI. Repair of the intact (viral) strand renders the substrate sensitive to BglI. In this substrate, the mismatch is an unmatched A in the complementary strand.
To prepare a 3Ј nick one-base IDL substrate, a 150-bp PCR fragment site was amplified using pBluescript as a template and the primers 5Ј-GTGACTCTAGAGAATTCATGGTCATAGCTGTTTCC-3Ј and 5Ј-CTGACTCTAGAGAGTTAGCTCACTCATTAGG-3Ј.
The fragment that contained an EcoRI site provided by the upstream primer (underlined) was digested with XbaI and cloned into a unique NheI site in both f1mR21 and f1MR22, generating f1MR21RI and f1MR22RI, respectively. The NheI site is located 10 bp from the mismatch and on the other side of it relative to Sau96I. f1MR21RI dsDNA was linearized with EcoRI and then mixed with f1MR22RI phage ssDNA. In this substrate, the nick is located 130 bp 3Ј to the mismatch on the complementary strand. Sensitivities to restriction enzymes are the same as the 5Ј nick one-base IDL substrate. All substrates were prepared as follows: following linearization of the phage dsDNA, the reaction was desalted using Microcon 30 (Amicon). 5 g of dsDNA was mixed with 7.5 g of the appropriate ssDNA at a volume of 100 l, incubated at 75°C for 15 min, and put on ice for 5 min, followed by the addition of 20ϫ SSC to 2ϫ SSC final concentration and incubation for an additional 10 min on ice. Up to six reactions were loaded on a 1% low melt agarose gel and run in TBE for 36 h at 20 V. The band containing the heteroduplex was cut out, chopped to small pieces, washed with ␤-agarase buffer (10 mM Bis-Tris-HCl, pH 6.5, 1 mM EDTA) twice by incubating on ice for 30 min, and melted at 70°C. After cooling to 40°C, ␤-agarase (New England Biolabs) was added (5 units/200 l) and incubated for 2 h at 40°C. The mixture was extracted sequentially with phenol, phenol/chloroform/isoamyl alcohol mixture, and chloroform and ethanol-precipitated. The DNA pellet was resuspended in Tris-EDTA buffer.
MMR Assays-A typical assay (20 l) contained 100 g of cytoplasmic extract, 100 ng of DNA substrate in 20 mM Tris, pH 7.5, 5 mM MgCl 2 , 0.1 mM dNTPs, 4 mM ATP, 50 g/ml bovine serum albumin, 1 mM glutathione, 50 mM KCl. Reactions were incubated at 37°C, quenched by the addition of 60 l of stop solution (1.2% SDS, 25 mM EDTA), heat-inactivated at 65°C for 10 min, incubated with 1 l of proteinase K (10 mg/ml) at 37°C for 15 min, and extracted sequentially with phenol, phenol/chloroform, and chloroform, and the DNA was ethanol-precipitated. The DNA pellet was resuspended in restriction mix (15 l) containing 5 units of ClaI and either 5 units of Hind III (for G-T substrate) or 5 units of XcmI (for one-base IDL substrates), incubated at 37°C for 60 min, and resolved on 1% agarose gels for 16 h at 40 V in Tris acetate-EDTA buffer. Ethidium bromide-stained bands were visualized using Gel Doc 1000 (BioRad). The gels were quantified using Quantity One software (BioRad). Repair efficiency of 5Ј nick DNA substrates was calculated by dividing the sum of the 3.1-and 3.3-kb bands by the total intensity of the bands in the lane (6.4 kb ϩ 3.3 kb ϩ 3.1 kb). Repair efficiency of 3Ј nick DNA substrates was calculated by dividing the sum of the 3.2-and 3.3-kb bands by the total intensity per lane.
Limited Proteolysis-A typical reaction (20 l) contained 200 ng of MutL␣, up to 10 mM ATP, 30 mM Tris, pH 7.5, 150 mM NaCl, 10 mM MgCl 2 , 1 mM dithiothreitol. The protein was preincubated with ATP at 30°C for 5-15 min, and then 20 -50 ng trypsin (Promega) were added and incubation continued for 5-20 min. The reaction was terminated by adding 4 l of 6ϫ SDS-PAGE loading buffer and immediately boiled at 100°C for 8 min. Two equal volumes from a given reaction were each separated by 8% SDS-PAGE, transferred to polyvinylidene difluoride membranes (Immobilon-P; Millipore Corp.), and blotted with rabbit polyclonal antibodies to human Pms2 or Mlh1 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) at a 1:500 dilution followed by horseradish peroxidase-conjugated goat anti-rabbit IgG (Oncogene Research Products) at a 1:2000 dilution. The blots were developed using the Renaissance chemiluminescence kit (PerkinElmer Life Sciences) and exposed to x-ray film. The films were scanned using ScanMaker5 (Microtek) and quantified using Multi-Analyst software (Bio-Rad). The antibodies used were as follows: sc-581, anti-human Mlh1 N terminus; sc-582, antihuman Mlh1 C terminus; sc-617, anti-human Pms2 N terminus; sc-618, anti-human Pms2 C terminus.
Site-directed Mutagenesis-The glutamic acid to alanine substitutions have been made using the QuikChange site-directed mutagenesis kit (Stratagene). Two 36-mer mutagenic oligonucleotides containing the desired mutation (underlined, from GAG to GCC) in their middle (5Ј-GCACTGCGGTAAAGGCCTTAGTAGAAAACAGTCTGG-3Ј and 5Ј-CC-AGACTGTTTTCTACTAAGGCCTTTACCGCAGTGC-3Ј) were annealed to the same sequence on opposite strands of a plasmid containing the human Pms2 cDNA. Similarly, two 38-mer mutagenic oligonucleotides (5Ј-CGGCCAGCTAATGCTATCAAAGCCATGATTGAGAACTG-3Ј and 5Ј-CAGTTCTCAATCATGGCTTTGATAGCATTAGCTGGCCG-3Ј) were annealed to another plasmid containing the human Mlh1 cDNA.
PCR was carried out with Pfu DNA polymerase using 16 cycles of 95°C for 30 s, 55°C for 1 min, and 68°C for 12 min each. The products, which are nicked circular strands, were incubated with DpnI to digest the methylated, nonmutated parental DNA template and were used to transform DH10B E. coli cells. Plasmids were retrieved from transformants and sequenced to confirm the presence of the mutation. Fragments spanning the mutation were cut out and replaced the corresponding wild type fragments in Mlh1 or Pms2 cDNAs.
Expression and Purification of Human MutLa-The Bac-to-Bac baculovirus (Invitrogen) expression system was used to express MutL␣ in Spodoptera frugiperda (Sf9) cells infected with recombinant baculovirus.
Human Pms2 cDNA was cloned into pFastBac DUAL in two steps. A 760-bp fragment from the 5Ј-end was PCR-amplified using an upstream primer (5Ј-ATCAGCTGGGATCCATGCATCACCATCACCATCACGAG-CGAGCTGAGAGCTCGAG) encoding six consecutive histidines in front of the second amino acid and 5Ј-GGGGCAGCTGAACAAAAGG as downstream primer. The amplified fragment was digested with BamHI and PvuII and subcloned into pFastBac DUAL between BbsI and PvuII under the p10 promoter. This plasmid was cleaved with PvuII and ligated to the 3Ј fragment of Pms2. Mlh1 cDNA was subcloned into the resulting plasmid between XbaI and EcoRI in the other multicloning site under the polyhedrin promoter. Recombinant pFastBac DUAL plasmids were used to transform DH10Bac E. coli. cells (Invitrogen) that contain baculovirus shuttle vector (bacmid). Transformants in which the expression cassettes containing the cloned cDNAs from the pFastBac DUAL plasmids were transferred by transposition into the bacmid were isolated. Recombinant bacmid preparations from these transformants were used to transfect Sf9 cells. 150 -300 ml of logarithmic phase Sf9 cells grown in Sf-900IISFM medium (Invitrogen) supplemented with 10% fetal calf serum (Hyclone) were infected with baculovirus stocks at a multiplicity of infection of 3 for 48 h. The cells were lysed with five packed cell volumes of Sf9 lysis buffer (50 mM Tris, pH 7.9, 100 mM KCl, 1% Nonidet P-40, 5 mM 2-mercaptoethanol, 5 mM phenylmethylsulfonyl fluoride, one EDTA free protease inhibitor mixture tablet (Roche Molecular Biochemicals)). The supernatant, supplemented with 20 mM imidazole, was loaded on an Ni 2ϩ -nitrilotriacetic acid column (Qiagen) equilibrated with buffer H (25 mM Tris, pH 7.9, 0.5 M NaCl, 10% glycerol, 5 mM 2-mercaptoethanol) plus 20 mM imidazole. The column was washed with 10 volumes of buffer H plus 20 mM imidazole, and the bound protein was eluted with buffer H plus 500 mM imidazole. The eluate from the nickel column was diluted to 100 mM NaCl and loaded on a 1-ml Resource Q column (Amersham Biosciences), washed with buffer T (50 mM Tris, pH 7.8, 10% glycerol, 0.01% Nonidet P-40) plus 100 mM NaCl, and eluted with a 0.1-1 M NaCl gradient in buffer T. Fractions containing MutL␣ that eluted at ϳ0.3 M NaCl were pooled and concentrated using a Centricon 30 concentrator (Amicon), and the buffer was exchanged to MutL storage buffer (20 mM Tris, pH 7.5, 100 mM NaCl, 10% glycerol, 1 mM dithiothreitol, 0.1 mM EDTA).

Human MutL␣ Complements Extract from MMR-deficient Mouse Embryonic Fibroblasts (MEFs)-Human
MutL␣, purified after expression in insect cells (Fig. 1A), was tested for complementation of extracts from a MEF cell line (MP-1) genetically deficient in both Mlh1 and Pms2. The validity of this "human/mouse" approach is supported by reports showing that a human Mlh1 cDNA can complement the "MMR" phenotypes of Mlh1-deficient MEFs (42) and that human chromosome 2 that contains the MSH2 allele complements mouse Msh2 Ϫ/Ϫ cells (43). These results indicate the high degree of conservation of the mammalian MMR system.
We used mismatch-containing plasmid substrates as previously described (44) with the modification of using cytoplasmic extracts instead of nuclear. As shown in Fig. 2, A (lane 1) and B (lane 1), MP-1 cytoplasmic extract was inefficient in the repair of a 5Ј G-T and a 5Ј one-base IDL (insertion deletion loop) heteroduplexes, respectively. The addition of recombinant human MutL␣ restored repair activity (Fig. 2, A (lane 3) and B (lane 2)). These repair levels were comparable with those of MMR-proficient MEF and HeLa cell cytoplasmic extracts (data not shown) and were in the range reported for HeLa cell nuclear extracts (10,44). This repair activity meets other criteria characteristic of in vitro MMR including sensitivity to aphidicolin (data not shown) and being limited to the nicked strand ( Fig. 2A, compare lanes 3 and 4).
Conserved Glutamic Acid Residues in the N Termini of Mlh1 and Pms2 Are Each Required for in Vitro MMR-To test the role of the putative ATPase domain of human MutL␣ in MMR, we mutated a conserved glutamic acid residue in the N terminus of Mlh1 and Pms2 (Fig. 1B). We chose to mutate this residue because it was clearly implicated in the catalysis of ATP hydrolysis in human Pms2 (33) as well as E. coli MutL (28) and gyrase B (30) and has been shown to be essential for in vivo MMR in E. coli (45) and yeast (35).
We have purified human MutL␣ expressed in insect cells, in  5-10). The DNA from each reaction was purified and was split to two reactions. One half was digested with ClaI plus HindIII to probe repair of the complementary (nicked) strand of the DNA substrate (C). The other half was digested with ClaI plus XhoI to probe repair of the viral (continuous) strand (V). The arrows indicate unrepaired DNA (6.4-kb band) and repaired DNA (3.3-and 3.1-kb bands). The intermediate band (3.2-kb) is a cleavage product of trace amounts of linear homoduplex DNA "reagent" used in the preparation of the heteroduplex substrate, found in some heteroduplex DNA preparations. In the substrate illustration, Sau96I* denotes the cleavage site of the phage dsDNA used to prepare the heteroduplex. This is also the location of the 5Ј-nick in the heteroduplex. B, a similar in vitro repair assay was carried out with a 5Ј nicked one-base IDL heteroduplex, with the DNA digested with ClaI plus XcmI to probe repair of the complementary (nicked) strand. which the glutamic acid residue was substituted to alanine in either Mlh1 at position 34 (MutL␣ mE34A), Pms2 at position 41 (MutL␣ pE41A), or both (MutL␣ mE34A/pE41A) (Fig. 1A). Studies showed that heterodimer formation is dependent upon interaction of the C termini of Mlh1 and Pms2 (46). Therefore, not surprisingly, these mutations did not affect heterodimer formation as evidenced by similar ratios of the two protomers in both the wild type and mutant dimers (Fig. 1A). Moreover, the expression levels and protein yields were similar to wild type MutL␣, suggesting that neither mutation affected overall structure or stability.
A previous report suggested that "5Ј repair" might be dependent primarily on PMS2, whereas "3Ј repair" is dependent primarily on MLH1 (47). To test whether the conserved glutamic acid might be involved in determining "directionality" of repair, we repeated the assay using a similar substrate in which the nick was located at a similar distance but 3Ј relative to the one-base IDL (Fig. 3B). As shown, similar to the 5Ј nicked one-base IDL substrate, the two single mutants each showed similar and partial repair efficiency. Whereas repair of the two substrates by wild type MutL␣ was saturated early in the reaction, repair efficiencies of the single mutants continued to increase.
A similar trend of repair was observed for 5Ј nicked G-T substrate ( Fig. 2A). For both of the single mutant MutL␣ proteins, residual repair of the G-T mismatch was limited to the nicked strand ( Fig. 2A).
ATP-dependent Conformational Changes in Wild Type MutL␣ Studied by Limited Proteolysis-We next asked whether the putative ATPase activity of MutL␣ drives conformational changes similar to MutL, and asked what is the contribution of the individual protomers. Limited proteolysis analysis is often used to detect conformational changes in proteins (48). Previous analysis of yeast MutL␣ has suggested ATP-dependent conformational change in Mlh1 (35). ATP-dependent conformational changes have been reported recently for an N-terminal fragment of yeast Pms1 (49) but not in the context of the full-length MutL␣.
Here, we performed limited proteolysis followed by Western blotting, using specific antibodies to the N termini of Mlh1 or Pms2, to study the effect of ATP on possible conformational changes in each protomer of human MutL␣. Because the epitope for the Mlh1 and Pms2 antibodies is located at the N terminus of both protomers (residues 1-20), all of the detected fragments in Fig. 4, A and B, should contain the intact N termini of Mlh1 and Pms2, respectively, but have different C termini.
Using V8 protease, Mlh1 was digested from the full-length size of 80 kDa to fragments of 49, 40, and 30 kDa and smaller sizes (Fig. 4A). In the presence of 5 mM ATP, the 40-and 30-kDa N-terminal fragments were protected, compared with continued degradation without ATP. Pms2 was cleaved by V8 protease from the full-length size of 100 kDa to a 47-kDa N-terminal fragment and subsequently to a 40-kDa N-terminal fragment (Fig. 4B). ATP (5 mM) retarded the degradation slightly, as shown by the presence of an 18-kDa fragment seen only without ATP. However, there was no detectable protection of the 47-and 40-kDa fragments by ATP.
Trypsin, which has different substrate specificity than V8, digested MutL␣ more rapidly, and ATP-induced protection of both Mlh1 and Pms2 was more apparent (Fig. 5). The protec- tion of Mlh1 N-terminal fragments by ATP was substantial; in the absence of ATP, no N-terminal fragments of Mlh1 were detected (Fig. 5A). However, incubation with 10 mM ATP protected a 30-kDa fragment, similar in size to the fragment protected from V8 protease by ATP. The C terminus of Mlh1 was stable under the conditions used and was not affected by ATP (Fig. 5B).
Digestion of Pms2 yielded N-terminal fragments similar in size to those seen with V8 protease (Fig. 5C). Without ATP, the Pms2 N-terminal domain was more resistant to trypsin than the Mlh1 N-terminal domain (compare lanes 2-4 in Fig. 5C with lanes 2-4 in Fig. 5A). The presence of ATP stabilized Pms2 further (Fig. 5C, compare lanes 6 -9 with lanes 2-5). Nevertheless, ATP provided more protection to Mlh1 than to Pms2, as apparent by comparing the relative stability of the fragments (compare lanes 6 -9 in Fig. 5, A and C, respectively, and also Fig. 5E). The Pms2 C-terminal domain was resistant to digestion and was not affected by ATP, similar to Mlh1 (Fig. 5D).

ATP-dependent Conformational Changes in Mlh1 and Pms2
Show Differential Sensitivities to ATP Concentration-In the above experiment, ATP was used at a concentration of 10 mM. Under these conditions, both Pms2 and Mlh1 are expected to be saturated with ATP, based on the reported K m of E. coli MutL (28) and human Pms2 N-terminal fragment (33). To study whether the extent of protection correlates with the concentration of ATP, we repeated the assay but with various ATP concentrations (Fig. 6A). We also used a higher trypsin/protein ratio (1:4 compared with 1:10 in the previous experiment) in order to facilitate quantitation of Pms2 protected fragments.
Similar to the previous conditions, we could not detect the N-terminal fragments of Mlh1 in the absence of ATP (Fig. 6A,  lanes 1-3). In contrast to the previous conditions, the N-terminal fragments of Pms2 were not detected without ATP, reflecting the higher trypsin concentration in this assay. The degradation rate of Mlh1 and Pms2 was more rapid than in the previous assay (compare Fig. 6C with Fig. 5E), again consistent with higher trypsin concentrations in this assay. Here, again, Mlh1 was more resistant than Pms2 (Fig. 6C).
Mlh1 and Pms2 N termini were protected by ATP in a concentration-dependent manner ( Fig. 6A; quantitation in Fig.  6B). We calculated K1 ⁄2 values for Mlh1 and Pms2. These are the concentrations of ATP that give the half-maximal effect on proteolysis and are expected to reflect dissociation constants or the affinities for the nucleotide. Assuming that at 10 mM ATP Mlh1 or Pms2 are saturated and therefore maximally protected, K1 ⁄2 for the Mlh1 30-kDa fragment occurred at Ͻ0.1 mM ATP, whereas for the Pms2 40-kDa fragment K1 ⁄2 occurred between 0.1 and 0.5 mM ATP (Fig. 6B).
Effect of Glutamic Acid Substitutions on ATP-induced Conformational Changes-We next studied the effects of altering the conserved glutamic acid residues on the trypsin digestion patterns of the N-terminal domains of human MutL␣ using essentially the same assay as described above. Similar to wild type MutL␣, the three altered forms were protected by ATP in a concentration-dependent manner (Fig. 6, D, G, and J).
The limited proteolysis results for Mlh1 and Pms2 N termini in MutL␣ mE34A (Fig. 6D) were similar to that of wild type MutL␣, both in terms of the K1 ⁄2 values (Fig. 6E) and stability of the N-terminal fragments (Fig. 6F). The digestion pattern of the Mlh1 N terminus in MutL␣ pE41A (Fig. 6G) was similar to Mlh1 in wild type MutL␣. In contrast, the Pms2 N terminus cleavage pattern in MutL␣ pE41A was different than in wild type MutL␣ in that a higher ATP concentration was required for protection. Specifically, the K1 ⁄2 value for Pms2 occurred at 0.5-10 mM ATP (Fig. 6H) compared with 0.1-0.5 mM for wild type MutL␣ (Fig. 6B) and MutL␣ mE34A (Fig. 6E). One explanation for this difference is that the Pms2 E41A mutation may reduce ATP binding in addition to abolishing ATP hydrolysis. This explanation would be consistent with the finding that the equivalent mutation in MutL, E29A, reduced ATP binding (28).
The proteolytic pattern of the double mutant MutL␣ mE34A/ pE41A was different from that of the other three forms of MutL␣ (Fig. 6J) in that the Mlh1 and Pms2 N termini were more protected by ATP, as evident by the slower degradation rate (Fig. 6L, compare with Fig. 6, C, F, and I). The Mlh1 30-kDa fragment was essentially stable in the double mutant. Interestingly, the K1 ⁄2 value of Mlh1 N-terminal domain in the double mutant protein was higher (0.1-0.5 mM) than that of the same protomer in the single mutant protein, MutL␣ mE34A (Ͻ0.1 mM) (Fig. 6, compare K with E). Thus, Mlh1 carrying the E34A substitution showed different apparent affinities to ATP, depending on Pms2 status. In contrast, Pms2 carrying the E41A substitution showed an "inherent" ATP binding defect with a K1 ⁄2 in the range of 0.5-10 mM ATP, similar to the same Pms2 protomer in the single mutant protein, MutL␣ pE41A. Therefore, the E34A substitution by itself may not affect ATP binding to Mlh1.
Limited Proteolysis of Wild Type MutL␣ with Nonhydrolyzable ATP Analogs-The protection from proteolysis of mutant MutL␣ proteins predicted to be deficient in ATP hydrolysis suggests that ATP binding but not hydrolysis is necessary for this effect. To test this possibility further, we studied the ef- In the Western blot shown, wild type human MutL␣ was preincubated with or without 10 mM ATP for 15 min, followed by the addition of trypsin (Promega). Samples (20 ml) were taken just prior to or 5, 10, 15 and 20 min after the addition of the protease, immediately boiled for 8 min in the presence of SDS-PAGE sample buffer, separated by SDS-PAGE, blotted on Immobilon-P membranes, and probed with antibodies as indicated below. The blots were scanned using ScanMaker 5 (Microtek), and the scans were quantified using Multi Analyst software (Bio-Rad). A, the membrane was probed with an antibody against the N terminus of Mlh1. B, the membrane from A was stripped and reprobed with an antibody against the C terminus of Mlh1. C, the membrane was probed with an antibody against the N terminus of Pms2. D, the membrane from C was stripped and reprobed with an antibody against the C terminus of Pms2. E, stability of the N-terminal fragments detected in A and C. The intensities of the "ϩATP" 30-kDa Mlh1 bands in A (lanes 6 -9) and the combined intensities of the Pms2 40-and 47-kDa bands in C (lanes 2-6 and 6 -9) were quantified as described above. In each case, the band intensities at 5 min were normalized to 100% relative to the 10-, 15-, and 20-min time points. Lane 1 in panels A-D shows full-length Mlh1 or Pms2. fects of a nonhydrolyzable ATP analog on limited proteolysis patterns of wild type MutL␣. As shown in Fig. 7, AMP-PNP protected N-terminal fragments of both Mlh1 (Fig. 7C) and Pms2 (Fig. 7A) from proteolysis. These fragments were of similar size to those protected by ATP (compare lanes 6 -8 with lanes 10 -12 in Fig. 7, A and C). Interestingly, AMP-PNP had a differential protection effect on MutL␣. It protected Pms2 better than ATP (Fig. 7B), whereas ATP protected Mlh1 better than AMP-PNP, when these nucleotides were at 10 mM (Fig.  7D). Thus, the protection of wild type MutL␣ by AMP-PNP is not similar to the protection of the double mutant MutL␣ by ATP (Fig. 6), although a priori the former should imitate the hydrolysis mutant. The reason for this may be explained by different affinities of Pms2 and Mlh1 for AMP-PNP. As shown in Fig. 7E, Mlh1 and Pms2 N termini were protected by AMP-PNP in a concentration-dependent manner. However, calculation of the K1 ⁄2 values (Fig. 7G) suggests that Mlh1 binds AMP-PNP very poorly (K1 ⁄2 of 0.5-10 mM for AMP-PNP compared with 0.1 mM or less for ATP), whereas Pms2 binding is less affected (Fig. 7F). Using the same assay, Mlh1 showed a similar "binding" defect for ATP␥S, whereas Pms2 had no defect compared with ATP (data not shown). The results with nonhydrolyzable ATP analogs suggest that ATP binding is sufficient for protection of MutL␣ from proteolysis.
Limited Proteolysis of Wild Type and the Double Mutant MutL␣ with ADP-To study further whether ATP binding is sufficient to drive a conformational change in MutL␣, we performed limited trypsin proteolysis of wild type MutL␣ in the presence of ADP (Fig. 8A). ADP addition resulted in some protection of both Mlh1 and Pms2 (compare lanes 5 and 9 with lane 1). However, the protection of Mlh1 by ADP was negligible compared with ATP (Fig. 8C, ATP/ADP intensity ratio Ͼ10 at 10 mM) and was apparent only at 10 mM. In contrast, the protection of Pms2 by ADP was similar to that of ATP (Fig. 8C, ATP/ADP intensity ratio 0.9 at 10 mM) and was apparent at lower nucleotide concentrations. Because wild type MutL␣ is expected to hydrolyze ATP to ADP, the differential protective effect of Pms2 by ADP might result from ADP generated during hydrolysis. To explore this possibility, we tested the double mutant MutL␣ mE34A/pE41A, which is predicted not to hydrolyze ATP. The analysis shows (Fig. 8B) that both Pms2 and Mlh1 in the double mutant are protected by ADP, as was shown for the wild type protein (Fig. 8B, compare lanes 9 and 12 with lane 1). However, in contrast to the wild type MutL␣, ATP protected Pms2 in the double mutant to a higher extent than ADP (Fig. 8C, ATP/ADP intensity ratio 2.8 at 0.5 mM and 4.6 at 10 mM). Similar to the wild type protein, ATP protected Mlh1 in the double mutant Ͼ10-fold more than ADP. These results suggest that Mlh1 and Pms2 differ in their ADP binding properties and/or ADP-induced conformational changes and that the major nucleotide that drives these conformational changes is ATP. DISCUSSION In this report, we have studied the role of the N-terminal ATPase domains of human Mlh1 and Pms2, which comprise MutL␣, in MMR. We compared wild type MutL␣ with forms in FIG. 6. Limited trypsin proteolysis analysis of wild type and mutant MutL␣ proteins as a function of ATP concentration. Left column, wild type MutL␣ (A) or mutant MutL␣ proteins (D, G, and J) were preincubated for 5 min with varying ATP concentrations for 5 min prior to trypsin addition. Subsequently, samples (20 l) were taken at 5, 10, and 20 min, immediately boiled for 8 min in the presence of SDS-PAGE sample buffer, and analyzed by SDS-PAGE followed by Western blotting with antibody against the N terminus of either Mlh1 or Pms2. Only protected bands are shown (40-kDa Pms2 fragment and 30-kDa Mlh1 fragment). The blots were quantitated as described for Fig. 5. The graphs on the right quantitate the corresponding blot on the left column (e.g. B and C represent the blot shown on A). The schematic diagram on the left represents the MutL␣ proteins in each row: Mlh1 (M), Pms2 (P), and glutamic acid substitution (*). For each Western blot shown, there are two readouts: the extent of conformational change as reflected by the stability of the protected band at a given ATP concentration (right column, Stability) and affinity forATPasreflectedbytheATPconcentrationdependent protection (middle column, ATP titration). Middle column, proteolysis pattern as a function of ATP concentration. Quantitation of band intensities at 5 min is shown . For each blot (A, D, G, and J) in the left column, band intensity in lane 10 (10 mM ATP) was taken as 100% relative to reactions with lower ATP concentrations (lanes 1, 4, and 7). Stippled bars, Pms2; black bars, Mlh1. Right column, stability of the Mlh1 and Pms2 N termini in the presence of 10 mM ATP. For each blot shown in the left column, the band intensity at 5 min (lane 10) was taken as 100% relative to 10and 20-min time points (lanes 11 and 12, respectively). Stippled bars, Pms2; black bars, Mlh1. which Mlh1 and/or Pms2 were altered at conserved glutamic acid residues, predicted to be crucial for ATP hydrolysis (28). Using an in vitro complementation assay with recombinant forms of MutL␣, we have shown that these conserved residues were important for MMR activity. We have used these full-length biologically active proteins in limited proteolysis assays and showed that in the presence of ATP both Mlh1 and Pms2 were more resistant to protease digestion, suggesting that ATP binding induces conformational changes. We also reported differential nucleotide binding and conformational changes between Mlh1 and Pms2, suggesting an asymmetry within MutL␣.
The finding that the double mutant MutL␣ lacked detectable in vitro MMR activity suggests that ATP hydrolysis by human MutL␣ is essential for its function in MMR and is in agreement with genetic and biochemical studies of MutL proteins in E. coli and S. cerevisiae (28,35). The data in Fig. 3 suggest that whereas repair by wild type MutL␣ was saturated early in the MMR reaction, repair by the single mutants continued to increase, suggesting that total repair might approach wild type values with longer incubation times. This suggests that ATP hydrolysis by one protomer may be sufficient for continuing the cycling of MutL␣, albeit at a slower rate. In vivo, the time window allowing repair may be limiting, therefore rendering a more severe phenotype to the single mutants. Indeed, Mlh1 null mouse cells expressing human Mlh1E34A show a mutator phenotype indistinguishable from their uncomplemented parent cells. 2 Based on previous in vitro studies, mismatch repair in mam-malian cells has bidirectional capability to initiate repair from nicks located either 5Ј or 3Ј to the mispair (50). As proposed for E. coli MutL, the ATPase activity of MutL␣ may coordinate the downstream steps in MMR. A priori, the directionality of repair may depend on inherent asymmetry within MutL␣. Using substrates containing either 5Ј or 3Ј nicks with the single mutants, our results suggest that the ATPase activities of Mlh1 and Pms2 are both important for 5Ј and 3Ј in vitro MMR. Thus, they are in apparent contrast to previous studies suggesting that Pms2 was more important for 5Ј to 3Ј repair and Mlh1 was more important for 3Ј to 5Ј repair (47). Whether other functions of MutL␣ will show additional effects on the directionality of repair remains to be seen. We also observed that the single mutants partially repaired two mispairs, G-T and one-base IDL. Taken together, our findings suggest that the lack of ATPase activity in MutL␣ results in a general defect in MMR. The protection of Mlh1 and Pms2 from proteases in the presence of ATP suggests that the N-terminal domains of both protomers undergo conformational changes. In theory, conformational changes may also render a protein more sensitive to proteases. The finding that each protomer became more resistant is consistent with structural data on MutL, according to which disordered residues in five loops become ordered and form secondary structures upon ATP binding (28). Secondary structures are likely to be digested slower than disordered regions. ATP-induced conformational changes were demonstrated with E. coli MutL (28), yeast Mlh1 (35), and other members of the GHL superfamily such as gyrase B (26) and yeast Hsp90 (27). Based on our findings, we propose a more  5, and 9) or 5, 10, and 20 min following trypsin addition (1:4 trypsin/ protein ratio) and were processed as described for Fig. 6. A, limited proteolysis of Pms2 N terminus. B, stability of the 40-kDa Pms2 fragment in A. The band intensity at lane 6 was taken as 100% relative to lanes 7-12. C, limited proteolysis of Mlh1 N terminus. D, stability of the 30-kDa Mlh1 fragment in C. The band intensity at lane 6 was taken as 100% relative to lanes 7-12. E, wild type MutL␣ was preincubated without nucleotide or with various concentrations of AMP-PNP for 10 min. Reactions started by the addition of trypsin. The samples were processed as described in the legend of Fig. 6. The digestion of Pms2 N terminus (top) or Mlh1 N terminus (bottom) is shown as a function of AMP-PNP concentration. F, relative protection of the N terminus of Pms2, at 5 min, as a function of AMP-PNP or ATP concentrations. For the top blot in E, band intensity in lane 10 (10 mM AMP-PNP) was taken as 100% relative to reactions with lower AMP-PNP concentrations (lanes 1, 4, and 7). The data for Pms2 protection by ATP were taken from Fig. 6B. G, relative protection of the N terminus of Mlh1, at 5 min, as a function of AMP-PNP or ATP concentrations. For the bottom blot in E, band intensity in lane 10 (10 mM AMP-PNP) was taken as 100% relative to reactions with lower AMP-PNP concentrations (lanes 1, 4, and  7). The data for Mlh1 protection by ATP was taken form Fig. 6B. extensive ATP-induced conformational change in Mlh1 compared with Pms2. A less extensive conformational change in human Pms2 is consistent with recent studies on the structure of the N terminus of human Pms2 (33), in which three of the above mentioned five loops are ordered in the monomeric Pms2 N-terminal fragment in the absence of ATP and do not change conformation further with ATP. Although the structure of the N terminus of human Mlh1 has not been reported, we predict, based on our limited proteolysis studies, that more residues become ordered in Mlh1 upon ATP binding than in Pms2.
Another differential effect detected in the limited proteolysis studies was that Mlh1 was protected at lower ATP concentrations than Pms2, suggesting a higher affinity of Mlh1 for ATP, a finding that is consistent with a recent study of N-terminal fragments of yeast MutL␣ (49).
The finding that the glutamic acid substitutions did not prevent the ATP-induced conformational changes of the altered MutL␣ proteins suggests that the conformational change depends primarily on ATP binding, as was proposed for E. coli MutL (28). Also consistent with this idea are our findings that nonhydrolyzable ATP analogs, AMP-PNP and ATP␥S, protected Mlh1 and Pms2 from proteolysis.
Interestingly, ADP conferred similar protection from proteolysis to wild type Pms2 as ATP, but ATP protected Pms2 in the double mutant better than ADP (Fig. 8). In contrast, ATP protected Mlh1 far better than ADP in both wild type and the mutant protein. One possible explanation for this differential protection effect of Pms2 versus Mlh1 is that ADP formed during hydrolysis stays bound to Pms2 while dissociating faster from Mlh1. This is based on the observation that wild type Mlh1 required 10 mM ADP to show protection, whereas Pms2 was protected at lower ADP concentrations (Fig. 8A). Furthermore, Pms2 may hydrolyze ATP faster than Mlh1, as has been suggested recently for yeast Pms1 (49). Taken together, the nucleotide binding site of Pms2 may be occupied with ADP for longer periods of time than Mlh1, thus preventing rebinding of ATP. Although ADP can provide protection from proteolysis, the conformational change it elicits in the protein is smaller compared with ATP, as was suggested for bacterial MutL (28). In contrast to the wild type protein, the double mutant does not hydrolyze ATP; hence, ADP is not present to block ATP binding, and the full extent of conformational change in Pms2 is detected.
The significance of the differential binding of nucleotides and conformational changes for Mlh1 and Pms2 to in vivo MMR is not clear. One possible scenario is that Mlh1, which has higher affinity to ATP, binds the nucleotide first and undergoes conformational change. Next, Pms2 binds a second ATP molecule, changes conformation, and hydrolyzes ATP to ADP. The presence of ADP or the conformational change for Pms2 may elicit ATP hydrolysis by Mlh1, which in turn may facilitate ADP release by Pms2. This putative highly ordered cycling between nucleotide-free and -bound forms may regulate the timing of protein-protein interactions during different stages of mismatch repair. A simple model for the ATPase cycle MutL␣ that shows nucleotide binding and the associated conformational The Nterminal domains that contain the ATP binding sites are in light gray (for Pms2) and dark gray (for Mlh1). A, the N termini do not interact in the absence of bound nucleotide. The circular shapes represent the presumed "unordered" conformation that is sensitive to trypsin. B, ATP binding induces conformational changes in the N-terminal domains of both proteins, depicted schematically as transition to rectangular shapes. ATP causes a conformational change, possibly rendering residues in the N termini "ordered" and resistant to trypsin. C, ATPinduced conformational changes cause dimerization of the N termini. Although the conformational changes depicted in B and the resulting dimerization may occur simultaneously, for illustration purposes these events are separated into discrete steps. The interface between the protomers may impart further resistance to trypsin digestion. D, hydrolysis of ATP. Based on the proteolysis data, Mlh1 appears to bind ADP with less affinity than Pms2. Thus, ADP release from Mlh1 may precede that of Pms2. D to A, ADP release cycles MutL␣ back to the nucleotide-free form.

FIG. 8. Limited trypsin proteolysis analysis of MutL␣ with ADP versus ATP.
A, wild type human MutL␣ was preincubated either without nucleotide or with varying concentrations of ATP or ADP for 5 min, followed by the addition of trypsin (1:4 trypsin/protein ratio). Samples were taken 5 min later and processed as described for Fig. 6. B, MutL␣ mE34A/pE41A was preincubated without nucleotide or with 0.5 and 10 mM ATP or ADP for 5 min, prior to the addition of trypsin. Samples (20 l) were taken at 5, 10, and 20 min and processed as described in the legend of Fig. 6. C, relative intensity of protected bands with ATP compared with ADP. The graph shows that for wild type Pms2, ATP-dependent protection was similar to ADP, whereas in the double mutant, ATP resulted in better protection of Pms2 than ADP. As shown, for both wild type and the double mutant, Mlh1 was protected Ͼ10-fold more by ATP compared with ADP. The data was extracted from the gels in A and B. changes in Mlh1 and Pms2 is shown in Fig. 9.
The double glutamic acid mutant showed increased resistance of both Mlh1 and Pms2 protomers to trypsin relative to the other three MutL␣ proteins. One interpretation is that the ATPinduced conformational changes result in dimerization of the N termini of Mlh1 and Pms2, which in turn increase protection from protease digestion. Protein-protein interaction can protect from proteolysis; for example, heterodimerization of the ␣ and ␥ subunits of a G protein protected the former from tryptic digestion (51). Evidence for N-terminal dimerization has been reported for E. coli MutL (28), yeast MutL␣ (35), and other members of the GHL superfamily, including Hsp90 (27) and bacterial gyrase B (26). We suggest that this dimerization occurs in the wild type and the single mutant forms as well, but the dimer intermediate of these forms is more transient. ATP hydrolysis ultimately results in disruption of the dimer as the ATPase cycle proceeds (see Fig. 9). However, when both protomers cannot hydrolyze ATP due to mutation, the N termini are trapped in the dimer state, as suggested for gyrase B (52). Although not necessary for the ATPase activity of human Pms2 (33), N-terminal dimerization is likely to be required for mediating downstream events in MMR, possibly by generating new interfaces for protein-protein interactions, as was suggested for Hsp90 and E. coli MutL (53,54). In turn, ATP hydrolysis would ensure these interactions are transient by resetting the repair cycle intermediates.
Although the conformational changes detected in the limited proteolysis assay probably reflect the inherent characteristics of Mlh1 and Pms2, each protomer may allosterically or physically affect its partner by virtue of C-terminal or N-terminal interactions, thus adding to the complexity level of the assay. For example, the apparent affinity of Mlh1E34A to ATP seems to be influenced by the ability of Pms2 to cycle between ATP-bound and free forms (Fig. 6). Therefore, a more thorough analysis will require limited proteolysis of both Mlh1 and Pms2, separately. Unfortunately, our attempts to purify full-length human Pms2 in the absence of Mlh1 have not been successful. 3 Genetic studies with S. cerevisiae suggest a functional asymmetry within MutL␣ in that mutations in the conserved glutamic acid residues of MLH1 had a more severe mutator phenotype than the equivalent mutations in PMS1 (35). In our study, we did not however observe a difference in repair activity in vitro between the single mutant MutL␣ forms. The basis for this apparent discrepancy may reflect differences between ATP binding properties of yeast and human MutL␣. For example, the K m for ATP of yeast Pms1 N-terminal fragment was 19-fold higher than that of a similar human Pms2 fragment (33,49), although an N-terminal deletion in the former might account for such a difference. Alternatively, the yeast-human difference may reflect the comparison of in vitro versus in vivo assays. The MMR machinery in vivo may be coupled to DNA replication, possibly by proliferating cell nuclear antigen, as suggested by several groups (17)(18)(19)(20)(21)(22)(23). This coupling might facilitate strand discrimination during MMR, a process to which Mlh1 or Pms2 may contribute in an asymmetric manner. Therefore, the current in vitro MMR assay that operates independent of DNA replication may not fully represent repair as it occurs in vivo. Further studies in vivo and in vitro are required to more fully delineate the contributions of Mlh1 and Pms2 to MMR in mammalian cells.