Evidence That the DNA Mismatch Repair System Removes 1-Nucleotide Okazaki Fragment Flaps

Background: The DNA mismatch repair (MMR) system protects humans from cancer. Results: Combining an MMR system defect ( msh2 (cid:2) ) with rad27 (cid:2) causes a strong synergistic increase in the rate of 1-bp insertions, and a reconstituted MMR system removes 1-nt flaps. Conclusion: The MMR system removes 1-nt Okazaki fragment flaps. Significance: A new function of the MMR system was identified. The DNA mismatch repair (MMR) system plays a major role in promoting genome stability and suppressing carcinogenesis. is involved in Okazaki fragment maturation. We found that in the yeast Saccharomyces cerevisiae, the MMR system and the flap endonuclease Rad27 act in overlapping pathways that protect the nuclear genome from 1-bp insertions. In addition, we determined that purified yeast and human MutS (cid:2) proteins recognize 1-nucleotide DNA and RNA flaps. In reconstituted human systems, MutS (cid:2) , proliferating cell nuclear antigen, and replication factor C activate MutL (cid:2) endonuclease to remove the flaps. ATPase and endonuclease mutants of MutL (cid:2) are defective in the flap removal. These results suggest that the MMR system contributes to the removal of 1-nucleotide Okazaki fragment flaps. the of FEN1 the

Genome stability is essential for maintaining life and preventing numerous genetic disorders. The MMR 2 system promotes genome stability by correcting replicative DNA polymerase errors, removing mismatches formed during homologous recombination, impeding homologous recombination, and participating in DNA damage response (1)(2)(3)(4)(5). Genetic or epigenetic inactivation of the MMR system strongly predisposes humans to several types of cancers (6). MMR has been extensively studied in Escherichia coli and eukaryotes (7,8).
MutL␣ (MLH1-PMS2 heterodimer in humans and MLH1-PMS1 heterodimer in yeast), MutS␣ (MSH2-MSH6 heterodimer), MutS␤ (MSH2-MSH3 heterodimer), EXO1, PCNA, and RFC are the key eukaryotic MMR factors (9 -23). Eukaryotic MMR occurs both on the leading and lagging strands, but mismatches on the lagging strands are corrected more efficiently than those on the leading strands (24). The first step in eukaryotic MMR is recognition of the mismatch by the MutS homolog MutS␣ or MutS␤ (11,12,15,19). After mismatch recognition, MutS␣ or MutS␤ and loaded PCNA activate MutL␣ to incise the discontinuous daughter strand in the vicinity of the mismatch (25)(26)(27)(28). The endonuclease activity of MutL␣ depends on the integrity of its ATPase sites and the DQHA(X) 2 E(X) 4 E motif (25,26). A strand break generated by MutL␣ 5Ј to the mismatch serves as the entry site for MutS␣activated exonuclease 1 to degrade a mismatch-containing segment of the daughter strand in a 5Ј 3 3Ј-excision reaction (21,25). The resulting gap is repaired by DNA polymerase ␦ holoenzyme (29). The loss of exonuclease 1 causes only a modest MMR defect in yeast and mice (18,22). Consistent with these observations, a reconstituted system lacking exonuclease 1 is proficient in MMR (30). The reconstituted system bypasses the requirement for exonuclease 1 in the mismatch removal by relying on the strand-displacement activity of DNA polymerase ␦ holoenzyme.
In addition to mismatches, several other aberrant structures with significant mutagenic potential are formed during DNA replication. Among them are Okazaki fragment flaps (31,32). Okazaki fragment maturation is a process that removes the flaps and joins the trimmed ends together producing continuous strands (33,34). Genetic evidence indicates that defective removal of Okazaki fragment flaps causes genome instability (31,32). In eukaryotes, Rad27/FEN1 endonuclease, Dna2 helicase/nuclease, and the 3Ј 3 5Ј-exonuclease activity of DNA polymerase ␦ remove Okazaki fragment flaps (32)(33)(34)(35)(36). PCNA interacts with Rad27, and this interaction strongly stimulates the flap endonuclease activity of Rad27 (37). An important question is whether there are additional proteins that contribute to the removal of Okazaki fragment flaps.
The MMR system corrects DNA polymerase errors on newly replicated DNA (38 -41). It has been unknown whether the MMR system plays a direct role in DNA replication. In this report, we describe genetic and biochemical experiments that indicate that the MMR system removes 1-nt Okazaki fragment flaps.

Experimental Procedures
Yeast Strains and Genetic Methods-Saccharomyces cerevisiae wild-type haploid strains used in this study were as follows:  (44). The wild-type diploid strain FKY1037 was prepared by crossing the E134 and 1B-D770 strains. Gene replacements were generated by transforming yeast haploid or diploid cells with disruption cassettes in the presence of lithium acetate/PEG4000/DMSO. The PMS1 gene located in its natural chromosomal location was mutated to the pms1-E707K allele using the "dellitto perfetto" technique (45). Spontaneous mutation rates were measured, and mutation spectra were determined as described previously (42).
Oligonucleotides-Oligonucleotides used in this work were synthesized by IDT (Coralville, IA). The sequences of the indicated oligonucleotides are shown in Table 1.
Gel Mobility Shift Assays-Gel mobility shift assays that used the oligonucleotide-based substrates (Fig. 2) were carried out as described below. The oligonucleotide-based substrates were produced using oligonucleotides 1-8 ( Table 1). Each of the substrates contained oligonucleotide 1, which was labeled with 32 P at the 5Ј end with T4 polynucleotide kinase. In addition, the homoduplex, 1-nt insertion, dynamic 1-nt DNA flap, static 1-nt 3Ј DNA flap, static 1-nt 5Ј DNA flap, and nicked substrates contained oligonucleotides 2, 3, 4 and 5, 6 and 7, 4 and 8, and 4 and 7, respectively. To make the DNA substrates, the indicated oligonucleotides were mixed and annealed. The annealing was carried out in a buffer containing 20 mM HEPES-NaOH, pH 7.4, and 100 mM KCl at 40°C for 4 h, followed by incubation of the mixtures at 20°C for 30 min. After annealing, the resulting duplex DNAs were separated on native 6% polyacrylamide gels and then purified from the gels. The gel-purified DNAs were used as substrates in the DNA-binding reactions. The DNA-binding reactions were carried out in 20-l mixtures each containing 20 mM HEPES-NaOH, pH 7.4, 5 mM MgCl 2 , 140 mM KCl, 0.2 mg/ml BSA, 2 mM DTT, 20 nM of a competitor 40-bp DNA, 2 nM of the indicated 32 P-labeled DNA substrate, and purified yeast or human MutS␣. Yeast MutS␣ concentration in the mixtures varied in the range of 5-1600 nM (the actual concentrations used were 5, 10, 20, 40, 100, 200, 400, 550, 800, 1200, and 1600 nM). Human MutS␣ concentration in the mixtures was in the range of 5-800 nM (the actual concentrations used were 5,10,20,40,100,200,400,550, and 800 nM). The competitor 40-bp DNA was prepared by annealing two complementary phosphorylated 40-mer oligonucleotides 9 and 10. Reaction mixtures containing yeast MutS␣ were incubated for 10 min at 30°C, and reaction mixtures containing human MutS␣ were incubated for 5 min at 37°C. The reaction products were immediately subjected to electrophoresis on 6% polyacrylamide gels in the 0.5ϫ Tris borate/EDTA running buffer at 4°C. The gels were dried, and 32 P-labeled DNAs were visualized with a Typhoon phosphorimager (GE Healthcare). Each experiment was repeated at least twice. After quantification of the images with ImageQuant software (GE Healthcare), the apparent K d values were determined using GraphPad Prism 6 software. The data were fit into the equation of nonlinear regression curve with Hill slope (Y ϭ B max ⅐X h /(K d ϩ X h )). In this equation, Y is the concentration of MutS␣-DNA complexes; B max is the maximum concentration of MutS␣-DNA complexes; X is the concentration of MutS␣; K d is the apparent dissociation constant, and h is the Hill coefficient.
Gel mobility shift assays that used 2-kb circular DNA substrates ( Fig. 3) were performed as detailed below. The substrates were prepared using the pSYAH1A plasmid DNA containing a 36-nt gap (47). The gap was generated according to a described protocol (47). The no-flap, G-T, 1-nt DNA flap, and 1-nt RNA flap substrates were prepared by annealing the gapped pSYAH1A DNA with oligonucleotides 11, 12, 13, and 14, respectively. The G-T and no-flap substrates each contain two ligatable nicks that are 36 nt apart. Cleavage with restriction endonucleases HindIII and HpyCH4III was utilized to determine what fraction of each of the substrates contains the annealed oligonucleotide. These restriction endonucleases do not cleave DNA within a gap due to the destruction of their sites by the gap. Based on this approach, we determined that ϳ95% of each of the circular substrates contained the annealed oligonucleotide.
To determine apparent K d values for binding of human MutS␣ to the circular DNAs, the reactions were carried out in 20-l mixtures each containing 20 mM HEPES-NaOH, pH 7.4, 120 mM KCl, 5 mM MgCl 2 , 0.2 mM ATP, 0.2 mg/ml BSA, 2 mM DTT, 1.9 nM (50 ng) of the indicated circular 2-kb DNA, 50 nM of the competitor 40-bp DNA, and human MutS␣ (5,10,20,40,100,200, 400, 550, or 800 nM). After a 5-min incubation at 37°C, each reaction mixture was mixed with 3 l of loading buffer (1ϫ TAE, 40% glycerol, and 0.02% bromphenol blue), and the reaction products were immediately subjected to electrophoresis on 1.2% agarose gels in 1ϫ TAE at 4°C, followed by ethidium bromide staining of the gels. The separated DNAs were transferred onto nylon membranes and hybridized with 32 P-labeled oligonucleotide 15. The labeled DNAs were visual- The sequences of oligonucleotides described in this report  were gel-purified by IDT (Coralville, IA).

Oligonucleotide
Oligonucleotide sequence ized with a Typhoon phosphorimager. The data were quantified and analyzed as described above. DNA Incision Reactions-Circular DNAs were used as substrates in the incision reactions (Figs. 4 -8). Each of the substrates was prepared by annealing of an appropriate 5Ј-phosphorylated or 5Ј-32 P-labeled oligonucleotide to the gapped pSYAH1A DNA in a mixture containing the oligonucleotide and gapped DNA in a 1:1 molar ratio. The diagnostic cleavage with HindIII and HpyCH4III outlined above showed that 92-96% of each of the substrates contained the annealed oligonucleotide. The 5Ј-32 P label was introduced into the oligonucleotides by T4 polynucleotide kinase. The incision reactions were performed in 25-40-l mixtures each containing 20 mM HEPES-NaOH, pH 7.4, 120 mM KCl, 5 mM MgCl 2 , 3 mM ATP, 0.2 mg/ml BSA, 2 mM DTT, 1.5 nM (60 fmol) of the indicated DNA substrate, and the indicated human proteins. When MutS␣, MutL␣, PCNA, RFC, RPA, CAF-1, MutL␣-E705K, MutL␣-D699N, and MutL␣-EA were present in the reaction mixtures, their concentrations were 40,16,24,4,40,24,16,16, and 16 nM, respectively. Some DNA incision reactions (Figs. 6 -8) occurred in the presence of histone H3-H4 heterodimer (22, 44, or 88 nM). The DNA incision reactions were incubated at 37°C for 10 -30 min as indicated. Unless noted otherwise, the reactions were stopped and analyzed as described below. At the specified times, 8-or 11-l aliquots of the reactions were mixed with 20 l of a gel-loading buffer containing 90% formamide and 20 mM EDTA. DNA products of the stopped reactions were separated on 15% polyacrylamide gels containing 6 M urea. The gels were dried, and the 32 P-labeled DNA species were visualized by phosphorimaging. The data were quantified using ImageQuant software (GE Healthcare).

MMR System and Rad27 Flap Endonuclease Have Overlapping Functions Involved in the Maintenance of Genome
Stability-We began this work to investigate whether the MMR system contributes to the removal of Okazaki fragment flaps. The Rad27/FEN1 endonuclease is the key enzyme that removes short flaps during Okazaki fragment maturation (33,34). Accordingly, S. cerevisiae strains lacking Rad27 are genetically unstable (31,32,48,49). Previous research has demonstrated that both the MMR system and Rad27 are necessary for the suppression of mutations in the ϩ1 frameshift reporter his7-2 (32, 38 -40, 44). ϩ1 frameshifts that occur in a 51-bp his7-2 sequence containing an A 7 run revert the phenotype of the cells to His ϩ (42,44). To study whether there is a functional overlap between the MMR system and the Rad27 flap endonuclease, we determined the his7-2 mutation rates in the haploid and diploid yeast strains shown in Tables 2 and 3. The his7-2 mutation rate in the haploid double mutant msh2⌬ rad27⌬ (6700 ϫ 10 Ϫ8 ) was 33 times higher than the sum of the his7-2 mutation rates in the haploid single mutants msh2⌬ and rad27⌬ (i.e. combining msh2⌬ with rad27⌬ resulted in a 33-fold synergistic increase in the his7-2 mutation rate) ( Table 2). Likewise, the his7-2 mutation rate for the diploid double mutant msh2⌬/msh2⌬ rad27⌬/ rad27⌬ (13,000 ϫ 10 Ϫ8 ) was increased 36 times relative to the sum of the his7-2 mutation rates for the diploid single mutants msh2⌬/msh2⌬ RAD27/RAD27 and MSH2/MSH2 rad27⌬/ rad27⌬ (Table 3). These findings indicate that there is a functional overlap between the MMR system and Rad27 in haploid and diploid yeast S. cerevisiae. lys2::InsE-A 8 is a yeast ϩ1 frameshift reporter (43). ϩ1 frameshifts that are formed within a 71-bp lys2::InsE-A 8 region, including an A 8 run, produce Lys ϩ cells (43). To ascertain that the above findings (Tables 2 and 3) were not reporter-specific, we measured the lys2::InsE-A 8 mutation rates in the msh2⌬, rad27⌬, and msh2⌬ rad27⌬ mutants ( Table 4). Analysis of the data demonstrated that the lys2::InsE-A 8 mutation rate in the msh2⌬ rad27⌬ double mutant (21,000 ϫ 10 Ϫ8 ) was 24 times higher than the sum of the lys2::InsE-A 8 mutation rates in the msh2⌬ and rad27⌬ single mutants. Thus, the use of the lys2::InsE-A 8 mutation assay provided additional evidence that a genetic stabilization function of the MMR system overlaps with a genetic stabilization function of the Rad27 flap endonuclease. Collectively, these genetic experiments suggest that an MMR system-dependent mechanism and a different mechanism dependent on the Rad27 flap endonuclease repair the same or related types of pre-mutagenic intermediates which, if left unrepaired, give rise to ϩ1 frameshifts.
Next, we used DNA sequencing to identify ϩ1 frameshifts that reverted his7-2 in the msh2⌬, rad27⌬, and msh2⌬ rad27⌬ mutants ( Table 2). The results revealed that all of the his7-2 reversions in the msh2⌬ and msh2⌬ rad27⌬ spectra and a majority of the reversions in the rad27⌬ spectrum were 1-bp insertions, each of which extended the A 7 run into an A 8 run ( Table 2). In addition, we found that combining msh2⌬ with rad27⌬ led to a 40-fold synergistic increase in the rate of 1-bp insertions ( Table 2). This finding implies that one or several related types of pre-mutagenic intermediates producing 1-bp insertions are repaired by both an MMR system-dependent mechanism and a Rad27-dependent mechanism.

Impact of deletion of MSH2 and RAD27 on rates of his7-2 mutations
The mutant strains are isogenic to E134 (wild type) and were obtained by dissection of tetrads of MSH2/msh2⌬ RAD27/rad27⌬ diploids. 95% confidence intervals are in parentheses, and the relative rates of 1-bp insertions are in braces.

Absolute mutation rate (؋10 ؊8 )
Relative rate a Each of the complex mutations consisted of an insertion and four or more other genetic alterations, all located within an ϳ20-bp DNA segment. b The wild-type data are from a previous report (42).

1-bp insertions in the
The MMR system contains two mismatch recognition complexes, MutS␣ and MutS␤. As shown in Table 5, the his7-2 mutation rate in the msh3⌬ msh6⌬ mutant was indistinguishable from that in the msh2⌬ mutant but 23 times higher than the sum of those in the msh3⌬ and msh6⌬ mutants. This result indicates that the partially overlapping activities of MutS␣ and MutS␤ (19, 50 -52) are engaged in the suppression of ϩ1 frameshifts in his7-2. To study whether an MMR system-dependent function overlapping with a Rad27 function involves MutS␣ and/or MutS␤, we determined the his7-2 mutation rates for the msh2⌬, rad27⌬, msh2⌬ rad27⌬, msh3⌬ msh6⌬ rad27⌬, msh3⌬ rad27⌬, and msh6⌬ rad27⌬ mutants (Table 5). We found that the his7-2 mutation rate for the msh3⌬ msh6⌬ rad27⌬ mutant did not differ from the his7-2 mutation rate for the msh2⌬ rad27⌬ mutant, but it was ϳ12 or ϳ70 times higher than the rate for the msh6⌬ rad27⌬ or msh3⌬ rad27⌬ mutants, respectively. These data indicate that both MutS␣ and MutS␤ participate in an MMR system-dependent function that overlaps with a Rad27 function. We also found that the his7-2 mutation rate in msh6⌬ rad27⌬ exceeded that in msh3⌬ rad27⌬ by 6-fold (Table 5). This result is consistent with the view that compared with MutS␤, MutS␣ plays a more important role in an MMR system-dependent function that overlaps with a Rad27 function.
The results described above were obtained using the his7-2 and lys2::InsE-A 8 reversion assays that only allow scoring of ϩ1 frameshifts. Unlike the his7-2 and lys2::InsE-A 8 reversion assays, the CAN1 forward mutation assay allows scoring of many different types of genetic alterations, including 1-bp insertions, base substitutions, and 1-bp deletions. The CAN1 forward mutation assay takes advantage of the fact that mutational inactivation of the CAN1 gene encoding arginine permease makes the yeast cell resistant to canavanine, a structural analog of arginine. In this assay, Can r cells are selected on a synthetic medium that lacks arginine and contains canavanine. To determine the can1 mutation spectrum in an msh2⌬ rad27⌬ strain, we performed a series of experiments summarized in Fig.  1. We started this series of experiments by measuring the CAN1 mutation rates in two sets of msh2⌬, rad27⌬, and msh2⌬ rad27⌬ strains (Fig. 1A). One set of the strains was prepared on the wild-type strain E134 background and the other on the wild-type strain BY4742 background. We chose to measure CAN1 mutation rates in two sets of yeast strains to exclude the possibility that the data are strain-specific. The results demonstrated that the relative CAN1 mutation rate in either msh2⌬ rad27⌬ mutant was ϳ2 times higher than the sum of the relative CAN1 mutation rates in the isogenic single mutants (i.e. the relative CAN1 mutation rates in the isogenic msh2⌬ and rad27⌬ mutants are in a weak synergistic relationship) (Fig.  1A). Similar results were obtained in two earlier studies (48,49). We next determined the can1 mutation spectra in the wildtype, msh2⌬, rad27⌬, and msh2⌬ rad27⌬ mutants (Fig. 1B). The rates of base substitutions and 1-nt deletions in the msh2⌬ rad27⌬ mutant did not differ significantly from those in the msh2⌬ mutant. However, the rate of 1-nt insertions in the msh2⌬ rad27⌬ mutant was 12 times higher than sum of those   in the msh2⌬ and rad27⌬ mutants. This information supports the view that one or several related types of pre-mutagenic intermediates causing 1-nt insertions are removed by both an MMR system-dependent mechanism and a Rad27-dependent mechanism.
Duplications are formed at a high rate in rad27⌬ mutants (31,32). These duplications have been suggested to be the products of unprocessed Okazaki fragment flaps (31,49). Strikingly, 6 -14-bp duplications were produced at a rate of 630 ϫ 10 Ϫ8 in CAN1 in the msh2⌬ rad27⌬ strain, but they were absent in the can1 spectra of the rad27⌬ and msh2⌬ mutants (Fig. 1B). These data suggest that one or several related types of pre-mutagenic intermediates triggering 6 -14-bp duplications are removed by both an MMR-dependent mechanism and a Rad27-dependent mechanism.
The Dna2 helicase/nuclease is an essential enzyme that participates in the removal of flaps during Okazaki fragment maturation (33)(34)(35)(36). Yeast strains carrying a dna2 allele, dna2-1, are temperature-sensitive (55,56) and show a weak defect in the maintenance of dinucleotide repeats (56). We established that the his7-2 mutation rate in the dna2-1 strain was increased 10-fold relative to that in the wild-type strain ( Table 6). Sequencing of 10 independent HIS7 revertants produced in the dna2-1 background showed that nine mutants contained an identical mutation, which was an A insertion in the his7-2 A 7 run, and one mutant had a deletion of two As in the same run. We then studied the effect of combining dna2-1 with msh2⌬ on the his7-2 mutation rate (Table 6). We found that the his7-2 mutation rate in the dna2-1 msh2⌬ double mutant was two times higher than the sum of those in the single mutants. This observation is consistent with the idea that one or several related types of pre-mutagenic intermediates causing ϩ1 frameshifts are repaired by both an MMR system-dependent mechanism and a Dna2-dependent mechanism.
Recognition of 1-nt DNA Flaps by MutS␣-We considered two models to explain the observation that combining msh2⌬ with rad27⌬ leads to the strong synergistic increases in the rates of spontaneous 1-bp insertions ( Table 2 and Fig. 1B). In the first model, DNA polymerase ␣ errors are corrected not only by MMR (57) but also by a Rad27-dependent mechanism, and DNA polymerase ␣ errors that escape both MMR and the Rad27-dependent mechanism produce mutations, including 1-bp insertions. However, this model is not supported by the observation that the deletion of RAD27 in the msh2⌬ strain does not significantly increase the rate of base substitutions (Fig. 1B), which are the most common products of DNA polymerase ␣ errors (57, 58). Thus, it is unlikely that a considerable fraction of 1-bp insertions formed in msh2⌬ rad27⌬ mutants  CAN1 mutation rates and can1 mutation spectra in the wild-type, msh2⌬, rad27⌬, and msh2⌬ rad27⌬ strains. A, CAN1 mutation rates. Each of the mutants was made in the two different wild-type backgrounds: E134 and BY4742. The numbers above the bars are the relative mutation rates. B, can1 mutation spectra in the wild-type strain E134 and its mutant derivatives. The relative mutation rates are in parentheses. a , all 1-bp insertions were formed in mononucleotide runs that were ՆN 2 . originate from DNA polymerase ␣ errors. The second model is based on the knowledge that the key function of the 5Ј flap endonuclease Rad27 is the removal of short Okazaki fragment flaps (33,34). In this model, 1-nt Okazaki fragment flaps are removed by both a Rad27-dependent mechanism and an MMR system-dependent mechanism, and the unprocessed flaps are converted by misalignment and ligation into 1-bp insertions. Thus, this model suggests that the majority of 1-bp insertions produced in msh2⌬ rad27⌬ mutants are formed from 1-nt Okazaki fragment flaps. Because Okazaki fragment flaps do not cause base substitutions, the second model is consistent with our genetic data ( Table 2 and Fig. 1B).
The second model postulates that the MMR system removes 1-nt Okazaki fragment flaps. To determine whether there is evidence for this, we carried out the biochemical experiments described below. We first examined whether yeast MutS␣ recognizes 1-nt DNA flaps present on the 32 P-labeled oligonucleotide-based substrates (Fig. 2). The data revealed that yeast MutS␣ bound the substrate containing the dynamic 1-nt flap with an apparent K d of 38 Ϯ 2 nM (Fig. 2, A and B). The control experiments indicated that yeast MutS␣ bound the 1-nt insertion-containing DNA, nicked DNA, and homoduplex DNA with apparent K d values of 25 Ϯ 1, 180 Ϯ 10, and 200 Ϯ 8 nM, respectively (Fig. 2, A and B). Therefore, these experiments demonstrate that yeast MutS␣ recognizes the dynamic 1-nt flap nearly as efficiently as the 1-nt insertion. We then investigated whether yeast MutS␣ recognizes static 1-nt 3Ј and 5Ј flaps. The experiments showed that yeast MutS␣ bound the static 1-nt 3Ј and 5Ј flaps with apparent K d values of 60 Ϯ 2 and 55 Ϯ 3 nM, respectively. Thus, yeast MutS␣ recognizes the static 1-nt 3Ј and 5Ј flaps with the same affinity. Surprisingly, yeast MutS␣ detected the static 1-nt 3Ј and 5Ј flaps somewhat less efficiently than the dynamic 1-nt flap (Fig. 2, A and B). Because a dAMP residue forms the flap in the dynamic substrate and a dCMP residue produces the flaps in the static substrates, it is possible that yeast MutS␣ recognizes a flapped dCMP residue less efficiently than a flapped dAMP residue.
We also studied whether human MutS␣ recognizes the dynamic 1-nt flap (Fig. 2C). Our experiments indicated that human MutS␣ bound the dynamic 1-nt flap with an apparent K d of 30 Ϯ 1 nM. An apparent K d value for binding of human MutS␣ to the 1-nt insertion is 30 Ϯ 6 nM. These K d values are 7-12 times lower than those for binding of human MutS␣ to the nicked and homoduplex DNAs (Fig. 2C). Thus, human MutS␣ efficiently recognizes the dynamic 1-nt flap. Collectively, these findings support the view that the ability to recognize 1-nt DNA flaps is conserved in eukaryotic MutS␣ proteins.
We also analyzed whether human MutS␣ recognizes a dynamic 1-nt flap present on a circular 2-kb DNA (Fig. 3). Each of the substrates contained a 1-nt DNA flap, a 1-nt RNA flap, no flap, or a G-T mispair (Fig. 3A). The results revealed that MutS␣ bound the 1-nt DNA and RNA flap-containing DNAs with K d values of 119 Ϯ 3 and 115 Ϯ 10 nM, respectively (Fig. 3B). These K d values are half that of 254 Ϯ 35 nM for the binding of MutS␣ to the control no-flap DNA. Thus, MutS␣ detects that the cir-cular DNA carries a 1-nt flap, which may be a deoxyribonucleotide or ribonucleotide residue.
MutL␣ Endonuclease-dependent Removal of 1-nt Flaps-Having shown that MutS␣ recognizes the 1-nt DNA and RNA flaps on the circular DNA, we carried out and analyzed the reconstituted reactions to determine whether these flaps activate human MutL␣ endonuclease to incise the discontinuous strand in the presence of human MutS␣, PCNA, RFC, and RPA (Fig. 4). The circular DNAs were used as substrates in these reactions because loaded PCNA, required for the activation of MutL␣ endonuclease (25)(26)(27), slides off of linear DNA. The reactions were performed under conditions that were very similar to those used for the identification of the MutS␣-, PCNA-, RFC-, mismatch-, and ATP-dependent endonuclease activity of human MutL␣ (25). Analysis of the reactions (Fig. 4, A and B) led to the following observations. First, 34 Ϯ 5% of the discontinuous strand of the 1-nt DNA flap-containing substrate was incised by MutL␣, whereas the endonuclease cleaved only 10 Ϯ 2% of the discontinuous strand of the control flap-free substrate. Second, MutL␣ incised 30 Ϯ 1% of the discontinuous strand of the 1-nt RNA flap. Third, an endonuclease-deficient MutL␣ variant, MutL␣-E705K (25), did not incise the discontinuous strands of the tested substrates. Together, these observations indicate that 1-nt flaps activate MutL␣ endonuclease to incise the discontinuous strand in the presence of MutS␣, PCNA, RFC, and RPA.
To determine whether incision of the discontinuous strand by MutL␣ results in the removal of flaps, we performed experiments summarized in Fig. 5. As shown in lane 2 of Fig. 5A, the incubation of MutL␣, MutS␣, PCNA, RFC, and RPA with the 1-nt DNA flap-containing circular substrate led to incision of the 32 P-labeled 37-nt fragment at several sites. The most abundant product of the incision reaction had an apparent length of 5 nt, indicating that the incision occurred at a site that is four nucleotides 3Ј to the flap. The incision products were not formed when MutS␣, MutL␣, RFC, or PCNA was omitted from the reaction mixture, but the omission of RPA did not have a significant effect on the incision (Fig. 5, A, lanes 3, 4, 6, and 7,  and B). These results indicate that MutS␣, MutL␣, RFC, and PCNA are required for the incision, but RPA is not. The time course experiments demonstrated that the incision reaction produced the 5-nt fragment in a time-dependent manner (Fig.  5C). The efficiency of the incision of the site located 4 nt downstream from a 1-nt flap was three times higher than that of the same site on the control flap-free substrate (Fig. 5, A, lanes 2  and 10, and B and C). Thus, the flap dependence of the MutL␣ incision was 3-fold. Changing the incubation temperature from 37 to 25°C decreased the flap dependence of the MutL␣ incision from 3-to 2-fold (data not shown). MutL␣, MutS␣, PCNA, and RFC were also required for the incision of the 1-nt RNA flap-containing substrate (Fig. 5, A, lanes 18 -20, 22, and 23, and  B). Consistent with a previous study (59), the 5-nt incision product containing the 5Ј-ribonucleotide residue migrated in the gel slightly slower than the 5-nt incision product lacking a ribonucleotide residue (Fig. 5A, lanes 2 and 18).
We also studied whether the endonuclease activity of MutL␣ is necessary for the incision of the discontinuous strand at a 1-nt flap (Fig. 5, A and B). The replacement of the wild-type MutL␣ with the endonuclease-deficient MutL␣-E705K led to the disappearance of the incision products indicating that the endonuclease activity of MutL␣ is responsible for the incisions (Fig. 5, A, lanes 5 and 21, and B). Further analysis revealed that the presence of a 1-nt flap did not activate the MutL␣ endonuclease to incise the discontinuous strand immediately upstream from the flap (data not shown). Taken together, these experiments demonstrate that MutS␣, RFC, and PCNA activate MutL␣ endonuclease to incise the discontinuous strand 4 nt downstream from a 1-nt DNA or RNA flap. Because the incision is so close to the flaps, it triggers their dissociation from the substrates.
Newly replicated DNA is rapidly assembled into nucleosomes by a mechanism that depends on the histone H3-H4 chaperone CAF-1 (60 -62). The first step in CAF-1-dependent nucleosome assembly is the deposition of histone H3-H4 tetramers. CAF-1-dependent nucleosome assembly probably impacts many processes that take place on the nascent DNA. Consistent with this idea, CAF-1-dependent nucleosome assembly modulates MMR (46,63). Because the MMR systemdependent flap removal (Fig. 5) is likely to occur during CAF-1-dependent nucleosome assembly, we studied whether histone H3-H4 deposition by CAF-1 affects the flap-removing activity of the MMR system. We determined that CAF-1-dependent histone H3-H4 deposition stimulated the flap-removing activity of the MMR system by 2-fold (Fig. 6, A, lanes 11 and  12, and B) and increased the flap dependence of the incision from 3-to 6-fold (Fig. 6C). The efficiency of the flap removal was not changed when MutS␣ and MutL␣ were added to the reaction mixtures that were incubated with CAF-1, the histone After a 10-min incubation, the reactions were stopped by the addition of NaOH and EDTA to the final concentrations of 40 and 5 mM, respectively. The reaction products were separated on alkaline 1.2% agarose gels, transferred onto nylon membranes, hybridized with 32 P-labeled oligonucleotide 16, and visualized by phosphorimaging. A, representative images showing incision of the discontinuous strands in the presence of MutL␣, MutS␣, PCNA, RFC, and RPA. The diagrams outline the circular DNA substrates. Each diagram also shows the relative position of the hybridization probe (bar with an asterisk). The hybridization probe is complementary to the discontinuous strand. B, summary of incision of the discontinuous strands of the indicated DNA substrates at sites that are 4-nt 3Ј to the flap or control nick. The data were obtained by quantification of images, including those shown in A, and are presented as averages Ϯ 1 S.D., n Ն3.
H3-H4 complex, PCNA, RFC, and RPA for 15 min suggesting that the MMR system efficiently removes 1-nt DNA flaps in the presence of pre-loaded H3-H4 tetramers (data not shown). The omission of CAF-1 significantly decreased both the efficiency and flap dependence of the incision (Fig. 6C). Control experiments revealed that the flap removal occurring in the presence of CAF-1-dependent histone H3-H4 deposition required both MutS␣ and MutL␣ (Fig. 6, A and B). An endonuclease-deficient MutL␣ variant, MutL␣-D699N (25), as well as a MutL␣ ATPase mutant, MutL␣-EA (64), could not substitute for the wild-type MutL␣ in the incision reaction. Thus, these experiments demonstrate that the CAF-1-dependent histone H3-H4 deposition promotes the removal of 1-nt DNA flaps by the activated MutL␣ endonuclease.
We then studied how CAF-1 and the histone H3-H4 complex affect the incision of the discontinuous strand at sites that are distant from the 1-nt flap (Fig. 7). Strikingly, the presence of CAF-1 and the histone H3-H4 complex suppressed the MutL␣ endonuclease-dependent incision of the discontinuous strand at the remote sites (Fig. 7, A, lanes 9 and 14 -16, and B). A similar suppression of the MutL␣ endonuclease-dependent incision of the discontinuous strand was observed in the sixprotein system containing the histone H3-H4 complex but not CAF-1 (Fig. 7, A, lanes 9 -12, and B). These findings imply that both CAF-1-dependent histone H3-H4 deposition onto the DNA and nonspecific binding of the histone H3-H4 complex to the DNA protect the remote sites from the incision by the activated MutL␣ endonuclease.
Next, we performed experiments to study whether the reconstituted MMR system is able to remove flaps in the presence of FEN1 (Fig. 8). The data showed that increasing the FEN1 concen-tration decreased the yield of the product of MutL␣ endonucleasedependent flap removal and increased the yield of the product of FEN1-dependent flap removal (Fig. 8, A-C). In addition, the data indicated that one or several proteins present in the eight-protein system suppressed the flap endonuclease activity of FEN1 (Fig. 8,  A, lanes 3-10, and C). These experiments provide evidence that the MMR system removes flaps in the presence of FEN1 and sug- gest that the flap endonuclease activities of FEN1 and the MMR system compete with each other.

Discussion
High fidelity DNA replication is required for the maintenance of genome integrity and the suppression of human diseases (65). The MMR system improves the fidelity of DNA replication by correcting the errors of DNA polymerization (3,4,7). We have used genetic analysis and reconstituted systems to study whether the MMR system contributes to the removal of Okazaki fragment flaps. The major findings described in this report are: 1) combining rad27⌬ with msh2⌬ produces strong synergistic increases in the rates of 1-bp insertions in his7-2 and CAN1 ( Table 2 and Fig. 1B); 2) combining rad27⌬ with mlh1⌬, pms1⌬, or pms1-E707K causes a 20 -26 times synergistic increase in the rate of ϩ1 frameshifts in his7-2 (Table 5); 3) purified yeast and human MutS␣ proteins recognize 1-nt flaps (Figs. 2 and 3); 4) MutL␣ endonuclease activated by MutS␣, RFC, and PCNA removes 1-nt flaps (Fig. 5); 5) the flap-removing activity of the reconstituted MMR system is stimulated by CAF-1-dependent histone H3-H4 deposition (Fig. 6); and 6) the reconstituted MMR system removes 1-nt flaps in the presence of FEN1 (Fig. 8).
These findings indicate that the eukaryotic MMR system removes a subset of 1-nt Okazaki fragment flaps and support a model illustrated in Fig. 9. This model suggests that MutS␣, FIGURE 7. CAF-1-dependent histone H3-H4 deposition protects the remote sites from incision by MutL␣ endonuclease. Each DNA incision reaction was performed in the mixture containing the indicated human proteins and DNA substrate (1.5 nM). When MutS␣, MutL␣, PCNA, RFC, RPA, and CAF-1 were present in the reaction mixtures, their concentrations were 40, 16, 24, 4, 40, and 24 nM, respectively. After a 30-min incubation, the incision reactions were stopped and analyzed as described in Fig. 4. A, image showing the effects of the different protein combinations on incision of the discontinuous strands of the 1-nt flap-containing and flap-free DNA substrates. The diagrams outline the DNA substrates. Each diagram also shows the relative position of the hybridization probe (bar with an asterisk), which is complementary to the discontinuous strand. B and C, incision of the discontinuous strands of the 1-nt flap-containing and flap-free DNA substrates as a function of concentration of histone H3-H4 heterodimers. The data were obtained by quantification of images, including the one shown in A, and are presented as averages Ϯ 1 S.D., n ϭ 2.
MutL␣, PCNA, and RFC provide the minimal set of activities required for the removal of 1-nt Okazaki fragment flaps by the MMR system. According to this model, the mechanism of the removal of a 1-nt Okazaki fragment flap by the MMR system can be divided into three key steps as follows: recognition of the flap by MutS␣; activation of MutL␣ endonuclease by MutS␣, PCNA, and RFC; and the removal of the flap by the activated MutL␣ endonuclease. Our genetic results also suggest that there is an Msh2-dependent, MutL␣-independent mechanism of removal of 1-nt Okazaki fragment flaps (Table 5). In addition, our genetic results are compatible with another model. In this model, misalignment and ligation converts some 1-nt Okazaki fragment flaps into 1-nt loops, which are then removed by the strand-specific MMR (8,11,19,28,66). However, it has not yet been demonstrated that a replicative DNA ligase is able to convert 1-nt flaps into 1-nt loops in the presence of Rad27/FEN1 and/or the MMR system.
The absolute his7-2 mutation rate in the rad27⌬/rad27⌬ msh2⌬/msh2⌬ diploid ( Table 3) is half that of the previously described strong mutator diploid pol3-01/pol3-01 msh2⌬/ msh2⌬ (40). (The pol3-01 mutation inactivates the proofreading activity of DNA polymerase ␦.) This observation reveals that the MMR system is nearly as important for the removal of ϩ1 frameshift intermediates in rad27⌬ strains as for the repair of ϩ1 frameshift intermediates in pol3-01 strains. Genetic interactions between the MMR system and Rad27 have been investigated in the past (31,48,49), but none of the previous studies utilized a ϩ1 frameshift assay or determined can1 mutation spectrum in a strain that lacks an MMR gene and RAD27. Nevertheless, Johnson et al. (48) reported that the relative CAN1 mutation rate in the msh2⌬ mutant is in a weak synergistic relationship with that in the rad27⌬ mutant. Thus, the results of the measurements of the relative CAN1 mutation rates in the msh2⌬, rad27⌬, and msh2⌬ rad27⌬ mutants obtained in this work (Fig. 1A) and the study of Johnson et al. (48) are consistent with each other.
MutS␣ was initially identified as an MMR factor that detects single base-base mismatches (11,12). Subsequent work established that MutS␣ recognizes 1-12-nt insertion/deletion loops (15,19) and damaged base pairs (67). We have described in this report that MutS␣ recognizes 1-nt DNA/RNA flaps (Figs. 2 and  3). This finding extends the range of potentially mutagenic DNA structures recognized by MutS␣. Our genetic experiments support the idea that MutS␤ plays a role in the MMR system-dependent removal of 1-nt Okazaki fragment flaps (Table 5). Thus, it is possible that MutS␤, like MutS␣, recog- nizes 1-nt DNA/RNA flaps and activates MutL␣ endonuclease to remove them. This would be in line with previous work that identified that MutS␤ specifically binds a variety of DNA recombination structures, including the noncomplementary 5Ј DNA flaps and 3Ј tails (68).
In the crystal structure of MutS␣⅐G-T DNA complex, the Glu-434 residue of the conserved mismatch recognition FXE motif forms a hydrogen bond with the mispaired T, the conserved Phe-432 residue stacks onto the T, DNA is sharply bent at the mismatch, and there are several nonspecific protein-DNA interactions (69). These features are also present in the structures of the prokaryotic MutS-mismatch-containing DNA complexes (70,71). It has been proposed that during mismatch recognition, MutS stacks the conserved Phe on an unpaired nucleotide residue and bends DNA (70). The intrinsic bendability of duplex DNA at a mismatch is thought to strongly contribute to the recognition of the mismatch by MutS. A recent study has shown that the same mechanism of mismatch recognition is employed by MutS␣ (69). We speculate that the MSH6 FXE motif is responsible for the recognition of flaps by MutS␣. If this is the case, the conserved Glu-434 is a strong candidate to interact with a flapped deoxy-or ribonucleotide residue via a hydrogen bond. It has been described that duplex DNA bends at nicks (72). Therefore, DNA bending at a nick that accompanies the flap may facilitate the flap recognition by MutS␣. Because the MMR system is conserved from bacteria to humans (69 -71), it is possible that the MMR system also contributes to the removal of Okazaki fragment flaps in bacteria.
Eukaryotic DNA transactions occur in the nucleosomal environment. The fact that the size of naked nascent DNA strands at a eukaryotic replication fork is only ϳ450 bp (73) is consistent with the view that the newly replicated DNA is rapidly assembled into nucleosomes by the histone chaperone CAF-1 (60). Our analysis demonstrates that the CAF-1-dependent histone H3-H4 deposition increases the efficiency and specificity of the flap removal by MutL␣ and protects the discontinuous strand from MutL␣ incision at the remote sites (Figs. 6 and 7). The mechanism behind these effects is not known. We speculate that the loaded histones H3-H4 tetramers trap the MutL␣containing incision complex at the flap-containing site where it was assembled, and as a result the MutL␣ is not able to incise the discontinuous strand at the remote sites and instead removes the flap.
Previous research demonstrated that during eukaryotic Okazaki fragment maturation, the strand displacement activity of DNA polymerase ␦ (30,33,74) produces flaps that are removed by the Rad27/FEN1 endonuclease (33,34), the 3Ј-5Ј-exonuclease activity of DNA polymerase ␦ (32), and the nuclease/ helicase Dna2 (36). In this report, we have described evidence that the eukaryotic MMR system contributes to the removal of Okazaki fragment flaps.
Author Contributions-F. A. K. and L. Y. K. designed experiments. L. Y. K., B. K. D., and F. A. K. performed experiments and analyzed data. F. A. K. and L. Y. K. wrote the paper.