Mlh1-Mlh3, a Meiotic Crossover and DNA Mismatch Repair Factor, Is a Msh2-Msh3-stimulated Endonuclease*

Background: Meiotic crossing over requires resolution of Holliday junctions through actions of the DNA mismatch repair factor Mlh1-Mlh3. Results: Mlh1-Mlh3 is a metal-dependent, Msh2-Msh3-stimulated endonuclease. Conclusion: Our observations support a direct role for Mlh1-Mlh3 endonuclease activity in recombination and repair. Significance: An enzymatic activity is identified for a key recombination and repair factor. Crossing over between homologous chromosomes is initiated in meiotic prophase in most sexually reproducing organisms by the appearance of programmed double strand breaks throughout the genome. In Saccharomyces cerevisiae the double-strand breaks are resected to form three prime single-strand tails that primarily invade complementary sequences in unbroken homologs. These invasion intermediates are converted into double Holliday junctions and then resolved into crossovers that facilitate homolog segregation during Meiosis I. Work in yeast suggests that Msh4-Msh5 stabilizes invasion intermediates and double Holliday junctions, which are resolved into crossovers in steps requiring Sgs1 helicase, Exo1, and a putative endonuclease activity encoded by the DNA mismatch repair factor Mlh1-Mlh3. We purified Mlh1-Mlh3 and showed that it is a metal-dependent and Msh2-Msh3-stimulated endonuclease that makes single-strand breaks in supercoiled DNA. These observations support a direct role for an Mlh1-Mlh3 endonuclease activity in resolving recombination intermediates and in DNA mismatch repair.

DNA mismatch repair (MMR) 3 is a conserved mechanism that acts during DNA replication to remove DNA polymerase misincorporation errors such as base-base and insertion/deletion mismatches. In the well characterized Escherichia coli MMR system, MutS recognizes and binds mismatches, recruiting MutL. This interaction relays mismatch recognition by MutS to the MutH endonuclease, resulting in nicking at nearby GATC sites of the unmethylated, newly replicated strand. MutH nicking provides an entry point for UvrD helicase and single-stranded exonucleases that then excise the mismatch, resulting in a gap that is filled in by replicative DNA polymerases (1).
Multiple MutS and MutL homologs (MSH and MLH, respectively) have been identified in eukaryotes. In Saccharomyces cerevisiae, the MSH complexes Msh2-Msh6 and Msh2-Msh3 bind to DNA mismatches and interact primarily with the MLH complex Mlh1-Pms1 to activate mismatch excision and DNA re-synthesis steps (1). Msh2-Msh6 primarily acts to repair basebase and small insertion/deletion mismatches, whereas Msh2-Msh3 primarily acts to repair larger (up to 17 nucleotides) insertion/deletion loop mismatches (1). A minor MMR pathway has been identified that involves recognition of insertion/ deletion mismatches by Msh2-Msh3 followed by interaction with the Mlh1-Mlh3 complex (2,3). Interestingly, subsets of MSH and MLH proteins act in a distinct, interference-dependent, meiotic crossover pathway (see below).
MutH-type endonucleases have not been identified in eukaryotes; however, a subset of MLH complexes displays DNA nicking activity (4). For example, human Mlh1-Pms2 (hMutL␣) and bakers' yeast Mlh1-Pms1 display latent endonuclease activities that are essential for mismatch repair (5,6). Studies on human MMR by Modrich and co-workers (5,6) have suggested that hMutL␣ nicking activity, which is directed to an existing nick, provides access to exonucleases (e.g. Exo1) that act in mismatch excision and may also facilitate strand discrimination. Mutations in a DQHAX 2 EX 4 E metal binding motif present in both human Pms2 and yeast Pms1 disrupt both MLH endonuclease activity and MMR (5)(6)(7). Recently, two groups reported the crystal structures of the endonuclease domain of MLH proteins (8,9). For Bacillus subtilis MutL, the endonuclease domain consisted of dimerization and regulatory subdomains connected by a helical lever spanning the conserved endonuclease motif. Such an organization was proposed to serve as a regulatory module to prevent MutL from promiscuously nicking DNA. The authors also suggest that interactions involving MutL and other protein factors, such as MutS and processivity clamp, trigger conformational changes in this motif that modulate MutL endonuclease activity (8).
Crossing over in meiotic prophase in most sexually reproducing organisms creates physical connections between chro-mosome homologs that are critical for their proper segregation in Meiosis I (10). In bakers' yeast, meiotic recombination is initiated by the formation of about 140 -170 double-strand breaks located throughout the genome (11)(12)(13)(14)(15). Approximately 40% of these double-strand breaks are converted into crossovers between homologous chromosomes, and the remainder are repaired as noncrossovers or by using a sister chromatid as a template (16). These double-strand breaks are resected to form 3Ј single-strand tails that primarily invade the unbroken homolog and undergo pairing with complementary sequences. Stabilized invasion intermediates (single-end invasion intermediates) are converted into double Holliday junctions (dHJs) and ultimately resolved into crossovers (17,18). Work in bakers' yeast has suggested that the majority of crossovers are formed in an interference-dependent pathway where dHJs, stabilized by Msh4-Msh5 (19 -21), are resolved through the actions of the Sgs1 helicase, the Exo1 XPG nuclease (in steps not requiring its exonuclease activity), and a putative Mlh1-Mlh3 endonuclease activity (22)(23)(24)(25)(26)(27).
Little is known at the biochemical level about how Sgs1, Exo1, and Mlh1-Mlh3 act on crossover intermediates and how Msh4-Msh5 might coordinate Mlh1-Mlh3. One model gaining increasing attention is that Msh4-Msh5 restricts a putative Mlh1-Mlh3 endonuclease activity to recombination intermediates that are resolved to crossover products, and Sgs1 acts as a pro-crossover factor creating a specific dHJ structure that can be readily cleaved by an endonuclease activity intrinsic to Mlh1-Mlh3 (25). These ideas are supported by the genetic and physical studies cited above as well as the following. 1) Msh4-Msh5 can bind to both single-end invasion and Holliday junction substrates in vitro (20). 2) Mlh3 contains a DQHAX 2 EX 4 E metal binding motif (Fig. 1A). 3) Cell biological assays show that Mlh1-Mlh3 acts downstream of Msh4-Msh5 in meiosis (28 -30). 4) Crossing over in bakers' yeast meiosis is reduced by ϳ2-fold in mlh3⌬ mutants and to a much greater extent (7-16fold) in mms4⌬ mlh3⌬ mutants defective in both interference-dependent and -independent crossover pathways (21,27,(31)(32)(33).
A DQHAX 2 EX 4 E metal binding motif found in MLH endonucleases was identified in the beginning of the C-terminal domain of Mlh3 (Fig. 1A). Nishant et al. (26) showed that a mutation in this motif, mlh3-D523N, created a dominant negative allele and conferred a null phenotype for Mlh3 functions in both MMR and meiotic crossing over. They also suggested that Mlh1-Mlh3 possesses an endonuclease activity required for its meiotic crossing over and MMR functions. We present the purification of bakers' yeast Mlh1-Mlh3 and demonstrate that it is a metal-dependent endonuclease that nicks super-coiled DNA. Mlh1-Mlh3 also facilitates binding of Msh2-Msh3 to insertion/deletion mismatches in vitro, and its endonuclease activity on supercoiled DNA is stimulated by Msh2-Msh3. In conjunction with previous genetic data, our work supports a role for the Mlh1-Mlh3 endonuclease in MMR and in resolving recombination intermediates.
FLAG Tag Insertion in MLH1-A DNA fragment containing the MLH1 open reading frame and a FLAG tag inserted after amino acid 448 in Mlh1 (39) was cloned between the XhoI and SphI sites of pFastBacDual (Invitrogen; downstream of the p10 promoter) to create pEAE348.
HIS 10 Tag Insertion in MLH3(S288C)-A His 10 tag was inserted at the N terminus of Mlh3. The resulting His 10 -MLH3 construct was inserted into the SalI and NotI sites of pFast-BacDual (Invitrogen, downstream of the pPH promoter) to create pEAE352.
HA Tag Insertion in MLH3 (S288C)-Overlap-extension PCR (40) was used to insert an HA tag (underlined) flanked by alanine residues (AAAYPYDVPDYAAAA) immediately after aspartic acid 466 of Mlh3 in pEAE352 to create pEAE358. A His 10 -MLH3-HA fragment was then excised from pEAE358 and introduced into pEAE348 to form pEAE360, which con- GCCGATATTGACAAGACGGCAAAGATGTCCTAGCAATCCATTGGT GATCACTGGTAGCGC tains the MLH1-FLAG gene downstream of the p10 promoter and His 10 -MLH3-HA downstream of the pPH promoter. mlh3-D523N (S288c)-The mlh3-D523N mutation was introduced into pEAE358 by Quick Change (Stratagene, La Jolla, CA). The resulting vector, pEAE367, contained His 10 -MLH3-D523N-HA downstream of the pPH promoter. The His 10 -mlh3-D523N-HA fragment was then excised from pEAE367 and introduced into pEAE348 to form pEAE369, which contains the MLH1-FLAG gene downstream of the p10 promoter and His 10 -mlh3-D523N-HA downstream of the pPH promoter. All PCR-amplified fragments were verified by DNA sequencing (Cornell BioResource Center).
Purification of Mlh1-Mlh3 from Baculovirus-infected Sf9 Cells-pFastBacDual constructs pEAE360 (Mlh1-Mlh3) and pEAE369 (Mlh1-mlh3-D523N) were used in the Bac-to-Bac baculovirus expression system (Invitrogen) according to the manufacturer's instructions. Mlh1-FLAG and His 10 -Mlh3-HA or His 10 -mlh3-D523N-HA were co-expressed by infecting Sf9 insect cells (Invitrogen) with amplified baculovirus at a multiplicity of infection of 5 in SF-900 III SFM medium (Invitrogen). For a typical purification, 300 ml of cells were harvested 60 h post-infection by centrifugation at 4000 rpm for 10 min. Cells were washed with PBS, pelleted, and flash-frozen in liquid N 2 .
Purified Mlh1-Mlh3 was resolved by Coomassie-stained SDS-PAGE. Individual bands were excised and analyzed by the Cornell University Proteomics Facility using a Thermo LTQ Orbitrap Velos Mass Spectrometer (Fig. 1C). Protein concentrations were determined by the Bradford assay (47) using BSA standards.
For assays on oligonucleotides, 1 nM end-labeled homoduplex, ϩ8, or HJ substrates was combined with 2 mM MgCl 2 , 1 mM ATP, and 150 nM Mlh1-Mlh3 in a buffer containing 20 mM HEPES, pH 7.5, 40 g/ml BSA, 20 mM KCl, and 1% glycerol on ice in a final volume of 15 l. Reactions were incubated and terminated as above. Formamide and bromphenol blue were added to final concentrations of 48% and 0.5 mg/ml, respectively. Samples were boiled for 5 min and resolved by 12% PAGE containing 8 M urea and 1ϫ Tris borate-EDTA, run at 300 V for 2 h. Gels were dried on 3MM Whatman paper and visualized by phosphorimaging.
Immunodepletion of Mlh1-Mlh3 Endonuclease Activity-Immunodepletion of Mlh1(FLAG)-Mlh3(HA) with anti-FLAG mouse IgG (Sigma) or control (anti-FliC (flagellin)) mouse IgG (BioLegend) was performed as described previously (5) with minor modifications. Briefly, protein A-agarose beads (Roche Applied Science) were washed extensively with 25 mM HEPES-KOH, pH 7.5, 0.15 M KCl, and 0.02% Nonidet P-40 (equilibration buffer). The beads were incubated with anti-FLAG or control IgG (2.5 g/l of packed beads) for 2 h at 4°C and washed with 100 volumes of equilibration buffer. Mlh1(FLAG)-Mlh3(HA) (0.4 g/l of beads) was incubated at 4°C for 1.5 h with immobilized antibodies in equilibration buffer containing 0.2 M KCl. Supernatants were analyzed by SDS-PAGE and for endonuclease activity on supercoiled DNA.
DNA filter binding assays were performed as described (49) with the following modifications. Briefly, 40-l reactions containing 20 mM Tris, pH 7.5, 0.01 mM EDTA, 2 mM MgCl 2 , 40 g/ml BSA, 0.1 mM DTT, and the indicated amounts of 32 Pend-labeled DNA and Mlh1-Mlh3 were incubated for 8 min at 30°C. A 32-l portion of each reaction was filtered through KOH-treated nitrocellulose filters (50) using a Hoefer FH225V filtration unit. Filters were analyzed by scintillation counting.
ATPase Assays-ATPase activity was determined using the Norit A absorption method (51)

RESULTS
Purification of Yeast Mlh1-Mlh3 from Sf9 Insect Cells-Bakers' yeast Mlh1-Mlh3 was purified from Sf9 insect cells infected with pFastBacDual baculovirus constructs expressing both MLH1 and MLH3 ("Experimental Procedures"). In these constructs a FLAG tag (DYKDDDDK) was inserted after amino acid 448 in Mlh1 and His 10 , and HA (YPYDVPDYA) tags were inserted in Mlh3 at the N terminus and after amino acid 466, respectively (Fig. 1B). The MLH1-FLAG and HIS 10 -MLH3-HA constructs conferred wild-type functions for DNA mismatch repair in the lys2A 14 reversion assay (39, 43; Table 2). Extracts prepared from Sf9 cells overexpressing Mlh1-FLAG and His 10 -Mlh3-HA were applied to a nickel-nitrilotriacetic acid-agarose column. Peak fractions containing Mlh1-Mlh3 were purified further using heparin and Q-Sepharose chromatography ("Experimental Procedures"). The identity of Mlh1-Mlh3, which purified as a heterodimer with an apparent 1:1 stoichiometry, was confirmed by mass spectrometry (Fig. 1, B and C, and "Experimental Procedures"). Mlh1-mlh3-D523N, containing a mutation in the putative endonuclease domain of Mlh3 shown previously to confer a null phenotype for both DNA MMR and meiotic functions (26), was purified using the same procedure (Fig. 1B). It is important to note that although Mlh1-Mlh3 remained as an intact heterodimer in both column chromatography (Fig. 1) and immunoprecipitation analyses, the subunits separated upon elution from a Superdex 200 HR 10/30 (GE Healthcare Life Sciences) gel-filtration column using buffer containing 50 mM NaH 2 PO 4 , pH 7.5, 150 mM NaCl, and 0.01% Tween 20 and variations thereof. We do not have a good explanation for this observation; however, one possibility is that Mlh1-Mlh3 interactions become unstable in specific buffer conditions. Consistent with this idea is a recent structural analysis of bakers' yeast  FEBRUARY 28, 2014 • VOLUME 289 • NUMBER 9

JOURNAL OF BIOLOGICAL CHEMISTRY 5667
Mlh1-Pms1, suggesting that yeast Mlh3 lacks 11 amino acids present in the C terminus of Pms1 that comprises part of the Mlh1-Pms1 interaction domain (9).

Mlh1-Mlh3 Is a DNA Endonuclease That Nicks Supercoiled DNA-Previously
Hall et al. (52) showed using filter binding and electron microscopy assays that the yeast Mlh1-Pms1 complex binds long duplex DNA molecules with positive cooperativity. As shown in an agarose gel shift assay ( Fig. 2A), Mlh1-Mlh3 shifted the mobility of duplex plasmid DNA and then formed protein-DNA complexes that became trapped in the well within a narrow concentration range (150 -200 nM). These observations encouraged us to test if Mlh1-Mlh3 can cleave plasmid substrates. We tested Mlh1-Mlh3 for endonuclease activity using the supercoiled plasmid pUC19 in reaction conditions similar to those used for B. subtilis MutL ("Experimental Procedures"). As shown in Fig. 2B, supercoiled pUC19 incubated with Mlh1-Mlh3 showed concentration-dependent conversion to the nicked circular form. This nicking activity was not seen in reactions containing Mlh1-mlh3-D523N. The endonuclease activity observed in purified Mlh1-FLAG-Mlh3-HA preparations could be immunodepleted using anti-FLAG but not using a nonspecific control antibody (Fig. 2C). Together these studies provide strong evidence that the Mlh1-Mlh3 heterodimer contains an intrinsic endonuclease activity.
Mlh1-Mlh3 Endonuclease Activity Is Metal-dependent but Not Stimulated by ATP or RFC and PCNA-As shown in Fig. 2B and Fig. 3A, Mlh1-Mlh3 displays a metal-dependent endonuclease activity. We tested different metals in a variety of concentrations and found that optimum stimulatory effects were typically seen at 1 mM concentrations of metal ion. Endonuclease stimulation was observed in the presence of Mn 2ϩ and Co 2ϩ and to lesser extents, Mg 2ϩ and Ca 2ϩ . These data suggest a broader metal specificity for Mlh1-Mlh3 endonuclease compared with Mlh1-Pms1, which shows a strict dependence on manganese (5,6).
We observed synergistic interactions involving metal cofactors. As shown in Fig. 2D, the addition of Zn 2ϩ , Cd 2ϩ , and Ca 2ϩ to reactions containing Mn 2ϩ resulted in a significant stimulation of Mlh1-Mlh3 endonuclease activity. However, Mg 2ϩ did not further stimulate the Mn 2ϩ -dependent nicking activity of Mlh1-Mlh3. These results are consistent with studies showing that the MLH proteins contain conserved motifs surrounding the nuclease motif that are rich in cysteine and histidine resi-dues and are predicted to chelate Zn 2ϩ but not Mg 2ϩ ions (e.g. Refs. 9 and 53). They also appear consistent with work reported by Pillon et al. (8) for BsMutL; they showed that Zn 2ϩ and Co 2ϩ stimulated BsMutL endonuclease incubated with Mn 2ϩ to a greater extent than Mg 2ϩ . Furthermore, crystal structures of the endonuclease domain of BsMutL show that Zn 2ϩ binding induces a significant conformational change to the domain. Based on these observations Pillon et al. (8) hypothesized that BsMutL contains a regulatory Zn 2ϩ -binding site adjacent to the nuclease active site that when occupied establishes the active conformation of the C-terminal domain, thereby enhancing its nicking activity. Interestingly, the C-terminal domain of human Mlh1-Pms2 appears to bind Zn 2ϩ tighter than the C-terminal domain of BsMutL; Zn 2ϩ is retained during the purification of the C-terminal domain of Mlh1-Pms2, whereas it is lost for BsMutL (8,53). Conceivably, Mlh1-Mlh3 may also have a tight affinity for Zn 2ϩ , thereby relaxing the strict requirement for Mn 2ϩ as the metal ion required for endonuclease activity.
ATP stimulates the endonuclease activity of human and yeast MutL␣ and B. subtilis MutL (5,6,8). In contrast, ATP inhibits the endonuclease activity of full-length Neisseria gonorrhoeae MutL (54). ATP had also been shown to inhibit the endonuclease activity of Thermus thermophilus and Aquifex aeolicus MutL (55), although inhibition was later linked to a concentration effect (56). The addition of ATP (1 to 500 M) in the presence of Mg 2ϩ and Mn 2ϩ (Mg 2ϩ is included to facilitate binding of ATP to Mlh1-Mlh3 and to prevent ATP from titrating Mn 2ϩ away from the endonuclease reaction) did not stimulate the endonuclease activity of Mlh1-Mlh3 (Fig. 3, A and B). These experiments were performed in the presence of Mn 2ϩ or Mg 2ϩ under conditions where endonuclease activity in the absence of ATP ranged from 10 to 50%, and ionic strength of the reaction ranged from 10 to 100 mM KCl (25 mM KCl and 15 mM NaCl are present in the experiments in Fig. 3A). These conditions were tested because for human MutL␣ ATP-stimulated endonuclease activity only in low ionic strength (5). Consistent with a lack of an ATP effect, Mlh1-Mlh3 displayed weak ATPase activity (ϳ0.06 min Ϫ1 , a few -fold lower than that seen previously for E. coli MutL and yeast Mlh1-Pms1; Refs. 57 and 58) that was not stimulated by the presence of a 10-fold molar excess of 69-bp homoduplex DNA (Fig. 3c).
Kadyrov et al. (5,6) showed that MutL␣ endonuclease activity on supercoiled DNA is stimulated by the replication factors RFC (clamp loader) and PCNA (clamp) in steps requiring ATP. One explanation for this observation is that PCNA loaded by RFC onto negatively supercoiled DNA stimulates MutL␣ endonuclease through protein-protein interactions (6,59,60). We were unable to detect stimulation of Mlh1-Mlh3 endonuclease activity by yeast PCNA and RFC (Fig. 3D), but RFC and PCNA stimulated human MutL␣ as seen previously. 4 Msh2-Msh3 Stimulates Mlh1-Mlh3 Endonuclease Activity-Gel shift assays showed that Msh2-Msh3 binds with higher affinity to insertion/deletion loop mismatches (50 nM K d ) compared with homoduplex DNA (220 nM K d ; Refs. 37 and 61). Previously Habraken et al. (62) showed that the binding of 4 J. Jiricny, personal communication.

TABLE 2 MLH3 constructs are functional in MMR
EAY2037 bearing ARS CEN plasmids with the indicated MLH3 genotypes were tested in the lys2::insE-A 14 mutator assay for reversion to Lys ϩ ("Experimental Procedures"). Six independent cultures were tested for each genotype, and the mutation rate with the 95% confidence interval (CI) is presented.

Mlh1-Mlh3 Is an Endonuclease
Msh2-Msh3 to loop mismatches was enhanced in the presence of yeast Mlh1-Pms1. The enhancement of Msh2-Msh3 binding by Mlh1-Pms1 did not result in a supershift complex containing MSH and MLH proteins as was seen for interactions involving Msh2-Msh6 and Mlh1-Pms1 (e.g. Refs. 63 and 64). One way to explain this result is that Mlh1-Pms1 transiently interacts with loop mismatches to facilitate stable binding of Msh2-Msh3 (see below). Consistent with this idea, Mlh1-Mlh3 also stimulated the binding of Msh2-Msh3 to loop mismatches at concentrations where a stable MLH-DNA complex was not observed in the gel shift assay (Fig. 4A). Similar to studies presented by Habraken et al. (62), this stimulation did not result in the formation of a supershift complex. The above findings, coupled with previous studies showing that MSH and MLH proteins interact (1,62), encouraged us to test if Msh2-Msh3 could stimulate Mlh1-Mlh3 endonuclease activity on supercoiled DNA. As shown in Fig. 4, yeast Msh2-Msh3 stimulated the endonuclease activity of yeast Mlh1-Mlh3 but not yeast Mlh1-mlh3-D523N. This stimulation was specific to the yeast proteins because yeast Msh2-Msh3 was unable to   FEBRUARY 28, 2014 • VOLUME 289 • NUMBER 9 stimulate the endonuclease activity of human Mlh1-Mlh3 (Fig.  4B). These data suggest that physical interactions between Msh2-Msh3 and Mlh1-Mlh3 can act to coordinate the Mlh1-Mlh3 endonuclease activity during MMR and provide a model for how Msh4-Msh5 could act as a specificity factor for Mlh1-Mlh3 to resolve HJs in meiosis. Fig. 5A, Mlh1-Mlh3 bound to duplex oligonucleotides with a moderate preference for a ϩ8 loop mismatch (ϳ50% of total binding at 50 nM Mlh1-Mlh3) compared with homoduplex substrate (ϳ50% of total binding at 90 nM). We found this result surprising because previous studies showed that MLH proteins such as Mlh1-Pms1 do not have a preference for binding to mismatch DNA (e.g. Ref. 64). These results encouraged us to test if Mlh1-Mlh3 can cleave oligonucleotide substrates. As shown in Fig. 5B, we were unable to detect Mlh1-Mlh3 endonuclease activity on oligonucleotide substrates such as a linear duplex, a ϩ8 loop insertion, or a model Holliday junction substrate (38; "Experimental Procedures"). One possible explanation for the lack of activity on these substrates is that Mlh1-Mlh3 endonuclease activity requires substrates that mimic characteristics of supercoiled DNA. However, as shown in Fig. 5C, Mlh1-Mlh3 made additional nicks on a nicked DNA substrate to form linear DNA molecules at levels similar to those seen when Mlh1-Mlh3 was initially incubated with supercoiled DNA. This result suggests that other factors (e.g. DNA substrate length) are likely to impact Mlh1-Mlh3 endonuclease activity.

DISCUSSION
In this study we showed that the bakers' yeast Mlh1-Mlh3 complex displayed an endonuclease activity. We also provided the first purification protocol for Mlh1-Mlh3, which will be important in mechanistic studies in MMR and meiotic crossover control (outlined below). This activity was dependent on divalent metal but was not stimulated by ATP or RFC ϩ PCNA. Interestingly Mlh1-Mlh3 bound to various oligonucleotide substrates but did not cleave them. Finally, we detected interactions between Mlh1-Mlh3 and Msh2-Msh3 in gel shift assays and showed that Msh2-Msh3 stimulated Mlh1-Mlh3 endonuclease activity.
Genetic studies suggested that Mlh1-Mlh3 intrinsic endonuclease domain is important for resolving dHJs into crossovers, yet the biochemical mechanism remains unknown (25,26). To our knowledge MLH proteins have not been observed to cleave any type of model substrate constructed from oligonucleotides. However, a large number of MLH proteins that contain an endonuclease domain motif display endonuclease activity on supercoiled and nicked plasmids (4 -6), suggesting a DNA binding specificity for MLH proteins that is not seen on small synthetic substrates. Our work shows that Mlh1-Mlh3 can nick both supercoiled and pre-nicked DNA; one way to explain our observations is that Mlh1-Mlh3-mediated endonuclease cleavage requires longer oligonucleotide substrates than those used in our study. MLH proteins have been shown to bind cooperatively to DNA (52); thus longer DNA substrates may be needed to stabilize MLH binding to promote DNA cleavage. Consistent with this idea, we hypothesize that Msh2-Msh3 stimulates Mlh1-Mlh3 endonuclease activity by binding non-specifically to supercoiled DNA or possibly to loops in supercoiled DNA created by the extrusion of inverted repeat sequences (65). In this model Mlh1-Mlh3 is specifically recruited to Msh2-Msh3 A representative assay for five trials is shown. B, nicking (n) activities of yeast and human Mlh1-Mlh3 (120 nM) were measured in the presence of 1 mM Mn 2ϩ and yeast Msh2-Msh3 (120 nM) as indicated. The % of supercoiled substrate (sc) that was cleaved is shown Ϯ S.D. C, endonuclease assays were performed as in B but contained 0 -500 M ATP, 120 nM Mlh1-Mlh3, and 120 nM Msh2-Msh3 when indicated. Reactions were performed in triplicate, samples were resolved on agarose gels, and the percent of DNA that was nicked is plotted. D, endonucleasedeficient yeast Mlh1-Mlh3 is not stimulated by yeast Msh2-Msh3. Nicking activity of yMlh1-Mlh3 and yMlh1-mlh3-D523N (each at 120 nM) was measured in the presence of 1 mM Mn 2ϩ , 0.5 mM ATP, and yMsh2-Msh3 (120 nM) as indicated. The % of supercoiled substrate that was cleaved is shown.
that is stably bound to DNA, thus promoting activation of the Mlh1-Mlh3 endonuclease.
Unlike other MLH endonucleases, the Mlh1-Mlh3 endonuclease was not stimulated by ATP or RFC ϩ PCNA. Based on these results we suggest that Mlh1-Mlh3 endonuclease activity on Holliday junctions requires novel and complex interactions with other pro-crossover factors. In support of this idea, Zakharyevich et al. (25) suggested that Msh4-Msh5 interacts with an Exo1-Mlh1-Mlh3 nuclease ensemble to resolve dHJs and that Sgs1 acts as a pro-crossover factor to interact with Mlh1-Mlh3 and ZMM meiotic factors (ZIP, MSH, MER) to stabilize strand exchange intermediates and to coordinate the formation of dHJs (see also 24). They also proposed that the Sgs1 helicase could act in this role by promoting a specific type of dHJ conformation that can be readily cleaved by an endonuclease. Such interactions between Mlh1-Mlh3 and the above factors may be necessary to facilitate large conformational changes within the Mlh1-Mlh3 complex to allow for cleavage of HJs. In support of this idea, Gueneau et al. (9) purified and crystallized the MutL␣ C-terminal domain. They did not observe an endonuclease activity associated with the C-terminal domain and suggested, as did others (e.g. 8), that conformational rearrangements involving DNA, the N-terminal domain of MutL␣, and the linker regions connecting the N-and C-terminal domains and metal cations are required for this activity.
We showed that Msh2-Msh3 stimulates the endonuclease activity of Mlh1-Mlh3. Interactions between MSH and MLH proteins has been characterized in several organisms and have been shown to involve the ATP binding domain of the MLH factor studied (66, 67), thereby reinforcing the idea that the crosstalk between the N-and C-terminal domains of Mlh1-Mlh3 is important to promote resolution of dHJs into crossovers. We have not explored whether post-translational modifications are required to activate Mlh1-Mlh3 during meiotic prophase. Such modifications were detected by Matos et al. (68). They found that the Mus81-Mms4 endonuclease was phosphorylated and hyper-activated by Cdc5 during meiosis. This activation was critical to generate crossing over required for proper chromosome segregation in Meiosis I.