DNA Mismatch Repair in Eukaryotes and Bacteria

DNA mismatch repair (MMR) corrects mismatched base pairs mainly caused by DNA replication errors. The fundamental mechanisms and proteins involved in the early reactions of MMR are highly conserved in almost all organisms ranging from bacteria to human. The significance of this repair system is also indicated by the fact that defects in MMR cause human hereditary nonpolyposis colon cancers as well as sporadic tumors. To date, 2 types of MMRs are known: the human type and Escherichia coli type. The basic features of the former system are expected to be universal among the vast majority of organisms including most bacteria. Here, I review the molecular mechanisms of eukaryotic and bacterial MMR, emphasizing on the similarities between them.


Introduction
DNA mismatch repair (MMR) is a highly conserved DNA repair system ( Table 1) that greatly contributes to maintain genome stability through the correction of mismatched base pairs. The major source of mismatched base pairs is replication error, although it can arise also from other biological processes [1]. In Escherichia coli, MMR increases the accuracy of DNA replication by 20-400-fold [2]. Mutations and epigenetic silencing in MMR genes have been implicated in up to 90% of human hereditary nonpolyposis colon cancers [3][4][5][6][7][8], indicating the significance of this repair system. Postreplicative MMR is performed by the long-patch MMR mechanism in which multiple proteins are involved and a relatively long tract of the oligonucleotide is excised during the repair reaction [9,10]. In contrast, particular kinds of mismatched base pairs are repaired through very short-patch MMR in which a short oligonucleotide tract is excised to remove the lesion [11][12][13]. In this paper, I refer to long-patch MMR as MMR.
Currently, 2 types of MMR mechanisms have been elucidated: one is expected to be employed by eukaryotes and the majority of bacteria, and the other is specific to E. coli and closely related bacteria. As shown in Figures 1(a) and 1(b), MMR in eukaryotes and most bacteria directs the repair to the error-containing strand of the mismatched duplex by recognizing the strand discontinuities. On the other hand, as shown in Figure 1(c), E. coli MMR reads the absence of methylation as a strand discrimination signal. In both MMR systems, strand discrimination is conducted by nicking endonucleases. MutL homologues from eukaryotes and most bacteria incise the discontinuous strand to introduce the entry or termination point for the excision reaction. In E. coli, MutH nicks the unmethylated strand of the duplex to generate the entry point of excision.
Here, I first review the overviews of MMR systems in E. coli, eukaryotes, and the majority of bacteria. Second, I refer to the molecular features of MutS and MutL homologues, the key enzymes in MMR. Third, molecular mechanisms for strand-discrimination and excision reaction are discussed. Finally, I introduce the cellular functions of MMR other than postreplication mismatch correction.

Overview of Methyl-Directed MMR in E. coli
The E. coli MMR system has been well characterized and reconstituted using recombinant proteins (Figure 1(c)) [14].
In this system, a mismatched base is recognized by a MutS homodimer. A MutL homodimer interacts with the MutS-DNA complex, and then a MutH restriction endonuclease is activated by MutL. The MMR system needs to discriminate the newly-synthesized/error-containing strand; however, a mismatched base itself contains no such signal. The E. coli  The involvement of bacterial clamp and clamp-loader in the strand incision reaction has not yet been confirmed. * 2 It is demonstrated that the endonuclease motif in MLH3 is responsible for in vivo MMR [83]; however, the endonuclease activity of MutLγ has not yet been confirmed biochemically. * 3 In yeast and human, EXO1 has the 5 -flap endonuclease activity in addition to 5 -3 exonuclease activity.  MMR system utilizes the absence of methylation at the restriction site to direct the repair to the newly synthesized strand. MutH nicks the unmethylated strand at the hemimethylated GATC site to introduce an entry point for the excision reaction [15][16][17]. The error-containing strand is removed by helicases [18] and exonucleases [19][20][21], and a new strand is synthesized by DNA polymerase III and ligase. The absence of methylation serves as a signal for the discrimination of the error-containing strand, and hence, E. coli MMR is called methyl-directed MMR. Although homologues of E. coli MutS and MutL exist in almost all organisms, no homologue of E. coli MutH has been identified in the majority of organisms including eukaryotes and most bacteria (Table 1) [22].

Overview of Eukaryotic MMR
In eukaryotes, strand discontinuity serves as a signal that directs MMR to the discontinuous strand of a mismatched duplex (Figure 1(a)). In newly synthesized strands, discontinuities can exist as 3 -ends or termini of Okazaki fragments. These termini of the DNA strand may function as strand discrimination signals in vivo. For the biochemical characterization of eukaryotic MMR, nicked plasmid DNAs have been used as a model substrate containing a strand discontinuity, on which several reviews elaborated [10,23,24]. For this assay system, the shorter path from a nick to the mismatch is removed by the excision reaction, indicating that 5 -and 3directed MMRs are distinct ( Figure 2) [25][26][27]. Intriguingly, the 5 -3 exonuclease activity of exonuclease 1 (EXO1) is required for both 5 -and 3 -directed strand removals [28,29]. The reason for the apparently contradictory requirement of 5 -3 exonuclease activity for the 3 -discontinuity-directed excision reaction had been unknown because it was thought that strand discontinuity was also the entry point for excision reaction. This issue was resolved by the discovery that the human MutL homologue MutLα (MLH1-PMS2 heterodimer) and yeast MutLα (MLH1-PMS1 heterodimer) harbor latent endonuclease activity that nicks the discontinuous strand of the mismatched duplex in a MutSα-, PCNA-, RFC-, and ATP-dependent manner ( Figure 3) [10,30,31]. In eukaryotic 5 -and 3 -directed MMR, the 3 -and 5sides of a mismatch are incised, respectively, by MutLα, and the 5 -3 exonuclease activity of EXO1 removes the DNA segment spanning the mismatch, that is, in eukaryotic 3directed MMR, the preexisting strand discontinuity does not serve as an entry point for the excision reaction, and the entry point is introduced by the endonuclease activity of MutLα ( Figure 3(b)). Intriguingly, MutLα is essential for 3 -nick-directed MMR but it is dispensable for 5 -nick-directed MMR, although MutLα possesses an enhancing effect on 5 -nickdirected MMR [32,33]. This result indicates that there is an alternative pathway for 5 -nick-directed MMR. In vitro experiments suggest that MutLα-independent 5 -nickdirected human MMR requires at least 3 proteins, MutSα, EXO1, and replication protein A (RPA) [32].

Overview of MMR in mutH -Less Bacteria
The DQHA(X) 2 E(X) 4 motif in the C-terminal domain of the PMS2 subunit of human MutLα comprises a metal-binding site that is essential for the endonuclease activity of MutLα and MMR activity of nuclear extract [30]. Except for E. coli and closely related bacteria, most mutH-lacking bacteria possess MutL homologues that contain this metal-binding motif [30,34]. Therefore, the molecular mechanism established on the basis of the results obtained from eukaryotic MMR systems is expected to be universal for organisms lacking mutH. In agreement with this prediction, several studies demonstrated that MutL homologues from mutHless bacteria, for example, Thermus thermophilus, Aquifex aeolicus, and Neisseria gonorrhoeae, possess endonuclease activity that is dependent on the DQHA(X) 2 E(X) 4 motif [34][35][36]. Furthermore, in T. thermophilus, it is clarified that the endonuclease activity of MutL is essential for in vivo DNA repair activity [34]. Thus, the molecular mechanism of MMR in mutH-less bacteria is expected to resemble that of eukaryotic one (Figure 1(b)).
The crystal structures of C-terminal dimerization domain-deleted bacterial MutS [37,38] and human MutSα [43] were solved in complex with mismatched DNA ( Figure 4). The structures revealed that the mismatchrecognition mechanisms of bacterial MutS and eukaryotic MutSα fundamentally resemble each other. The bacterial MutS homodimer/DNA complex shows a functionally asymmetric protein dimer in which only 1 of the 2 subunits recognizes the mismatched or unpaired base. Thus, although bacterial MutS binds to double-stranded DNA as a homodimer, its functionality is heterodimeric.
Bacterial MutS and human MutSα recognize the heteroduplex by stacking the mismatched base with a phenylalanine residue (Phe36 and Phe432 in E. coli MutS and human MSH6, resp.) that is intercalated from the minor groove side [38] (Figure 4). This phenylalanine residue is perfectly conserved among bacterial MutS homologues and eukaryotic MSH6 [37], and the alteration of this residue to an alanine results in a drastic decrease in the mismatchrecognition ability of Thermus aquaticus MutS [49]. residue is also conserved among bacterial MutS and eukaryotic MSH6 [37], and the mutation of this glutamate into an alanine or a glutamine abolishes in vivo MMR activity [50][51][52]. Elimination of this hydrogen bond by replacing the thymidine with 2,4-difluorotoluene, which lacks the N3, resulted in the decrease in mismatch selectivities of E. coli MutS and yeast MutSα [50,52].
The excellent crystallographic analyses of E. coli MutS complexes with various kinds of mismatched heteroduplexes remarkably enhanced our understanding of the mismatch recognition mechanism at an atomic resolution [53]. MutS recognizes a wide range of mismatches in a common manner. Heteroduplexes bound by MutS homologues are sharply kinked with the minor groove wide-opened ( Figure 5(d)), which allows the mismatched base to be in close contact with the phenylalanine and the glutamate residues of the mismatch-recognition domain. The carbonyl oxygen (OE2) of Glu38 in E. coli MutS forms a hydrogen bond with N3 of mismatched pyrimidine or N7 of mismatched purine base [53]. It is expected that N7 of mismatched purine or OE2 of the glutamate is protonated. As previously discussed, this might be the reason why the mutation of Glu38 to Gln in E. coli MutS eliminates MMR activity though Gln also can form hydrogen bonds [53]. The acidity of the glutamate but not glutamine might be suitable for the required pK a shift. Interestingly, E38A mutant of E. coli MutS exhibited the enhanced DNA-binding activity to a perfectly matched homoduplex DNA [50]. Schofield   that Glu38 would take part in the promotion of DNA kinking by destabilizing the unkinked DNA-MutS complex through the electrostatic repulsion of the Glu side-chain with phosphate backbone [50]. The lack of a crystal structure for MutS homologues with a homoduplex has prevented us from further understanding the mechanism; however, atomic force microscopy is a powerful tool that can be used to investigate the events that occur during the initial substrate binding by MutS homologues. Observations of the E. coli MutS-DNA complex using atomic force microscopy revealed that the DNA is bent at the perfectly matched site while it is bent and unbent at the mismatched site [54]. Taken together with previous biochemical and crystallographic data, it is proposed that MutS nonspecifically binds to DNA and bends it in order to search for a mismatch, when it encounters a mismatch it kinks the DNA at the mismatched site and finally forms an ultimate recognition complex in which the DNA is unbent [54]. In order to explain why an unbent conformation is more stable at a mismatch than at a homoduplex site, Wang et al. speculated the possibility that a mismatched base is flipped out upon mismatch recognition by MutS [54]. Base flipping is one of the major base-recognition mechanisms observed among base-processing enzymes such as DNA glycosylases or methyltransferases [55]. Although further experimental evidence should be required, this idea may be attractive especially when we remember the recent report that a nonenzymatic protein, alkylguanine alkyltransferaselike protein, also flips out the methylated base upon substrate recognition [56,57].
The mismatch-recognizing property of MutS homologues is closely related to their ATP-binding/hydrolyzing activity [58][59][60][61][62][63]. MutS homologues contain a Walker's ATPbinding motif in each subunit that is formed through the association of the subunits. It has been shown that E. coli MutS exchanges ADP to ATP upon mismatch binding then undergoes a conformational change to form a sliding clamp [61,63]. The ATP-binding-dependent formation of the sliding clamp was also confirmed in eukaryotic MutS homologues [63][64][65][66]. Several studies revealed that the ATPase activity of the E. coli MutS homodimer is also heterodimeric [67,68]. One of the 2 nucleotide-binding sites exhibits a high affinity to ADP, and the other shows a high affinity to ATP [60,69,70], therefore, we should discriminate between the different adenine nucleotide-binding forms of MutS homologues in MMR. Recently, a computational study using normal-mode analysis was applied to assess the conformational dynamics of E. coli MutS and human MutSα [71]. Normal-mode analysis is one of the simulation methodologies which can deal with the dynamics of large molecules [72]. The analyses suggested the existence of the distinct conformational states which are expected to reflect the functional cycles including DNA scanning, mismatch recognition, repair initiation, and sliding along DNA after mismatch recognition [71]. These computational studies can provide start points for further experiments. It is thought that the sliding clamp of MutS leaves the mismatch and diffuses along the DNA to activate the downstream reactions, which is called the "moving" or "cis" model [23,74]. The "moving" model is supported by the result that blockages between the mismatch and the nick prevent in vitro E. coli MMR [75]. In contrast, another mechanism, which is called the "stationary" or "trans" model, is also proposed to account for the activation of downstream reactions [23,74]. In the "stationary" model, MutS remains bound at the mismatch after recognition, and the ATPase activity of MutS is required for verification of mismatch recognition. The "stationary" model is supported by the observation that MutS-mismatch complex can activate MutH on separate homoduplex molecule [67].
The dynamism and transiency of the ternary complex of MutS (MutSα)-MutL (MutLα)-mismatch had prevented us from characterizing their physical interactions. Hydrogen/deuterium exchange mass spectrometry (DXMS) is suitable to study the interactions in large and transient complexes. DXMS analysis of the formation of E. coli MutS-MutL-mismatch complex revealed that a relatively small region in domain II (Figure 4) of MutS serves as an interface for binding MutL [76]. On the basis of the structural analyses of the MutS and MutL N-terminal domains, Wei Yang and coworkers also predicted this region to be a MutL-interacting site [37]. Although the sequence similarity of this region is limited, a structurally conserved region in the MSH2 subunit of Saccharomyces cerevisiae MutSα is also essential for interaction with MutLα [76].
E. coli MutS proteins exist not only in a dimeric form but also in a tetrameric form at high concentrations [41,77]. The C-terminal domain (the last 53 amino acids in E. coli MutS) is responsible for the tetramerization of full-length MutS. Although the crystallographic analyses were achieved by using C-terminal truncations, it has been pointed out that the deletion of the C-terminal domain causes defects in mismatch recognition, MutH stimulation, and in vivo MMR activity [77][78][79]. On the basis of the crystallographic analysis of the C-terminal domain (Figure 6), tetramer-disrupting mutants of E. coli MutS were prepared [44]. Examination of the effects of those mutations revealed that dimerization but not tetramerization of the MutS C-terminal domain is essential for in vivo mismatch repair [44].

Molecular Functions of MutL Homologues
Bacterial MutL homologues exist as homodimers [47] while eukaryotic MutL homologues form the heterodimers MutLα, MutLβ, and MutLγ [80][81][82][83]. It is suggested that a large portion of eukaryotic MMR is performed by MutLα. In contrast,  [45,85]. The N-terminal ATPase domain contains a single ATPbinding motif per subunit. MutL homologues belong to the GHKL ATPase superfamily that includes homologues of DNA gyrase, Hsp90, and histidine kinase in addition to MutL homologues [46,86]. The GHKL superfamily proteins undergo large conformational changes upon ATP binding and/or hydrolysis. X-ray crystallographic analysis clarified the ATP-binding-dependent conformational change of the N-terminal domain of E. coli MutL (Figure 8(a)) [46]. In addition, study using atomic force microscopy revealed that full-length S. cerevisiae MutLα can exist in several ATPbinding species with specific conformations in a solution [73]. The conformational change of the full-length MutL homologue seems to involve the interaction between the N-terminal and C-terminal domains (Figure 8(c)). ATPdependent conformational changes are also implicated for bacterial MutL endonucleases [34,87]. Such ATPase-cycledependent conformational changes in MutL homologue should be necessary to perform the MMR reaction [88][89][90].
The C-terminal domain of MutL homologues is the endonuclease domain in mutH-less organisms [30,31]. The inactivation of the metal-binding motif in the C-terminal domain of Homo sapiens PMS2, S. cerevisiae PMS1, T. thermophilus MutL, A. aeolicus MutL, and N. gonorrhoeae MutL diminishes their endonuclease activity [34][35][36]. Eukaryotic MutLα and bacterial MutL show apparently nonspecific endonuclease activity against lesionless DNA, indicating that MMR requires the sequence-or structure-nonspecific endonuclease activity to introduce excision entry point wherever it is needed [30,34]. The regulatory mechanism of this apparently nonspecific endonuclease activity has been argued [10,91].
The endonuclease activity of MutL homologues is affected by the binding of ATP to their N-terminal domain. Isolated T. thermophilus MutL tightly binds ATP in the absence of the MutS-mismatch complex and the ATPbinding form of MutL has decreased endonuclease activity against perfectly matched substrates in vitro [34]. The in vitro endonuclease activities of A. aeolicus and N. gonorrhoeae MutL are also suppressed by the addition of ATP [34,35]. In addition, T. thermophilus MutL is stably associated with a MutS-mismatch complex in the presence of ATP [34]. Since it has been known that the ATPase activity of MutL is activated by its interaction with MutS, formation of MutS-mismatch-MutL complex is expected to stimulate the endonuclease activity of MutL by canceling the ATPdependent suppression (Figure 9). In other words, ATP may be utilized to suppress the apparently nonspecific endonuclease activity of MutL until it is required. However, it remains to be determined whether the regulatory mechanism is used by other MutL homologues, including eukaryotic MutLα, because the endonuclease activity of MutLα is enhanced by the addition of ATP instead of being suppressed [30,31]. Interestingly, it is clarified that ATP stimulates the endonuclease activity of a relatively high concentration of A. aeolicus MutL in the absence of a MutS-mismatch complex [36], suggesting that the effect of ATP on the MutL endonuclease activity depends on the concentration of MutL. This finding also strongly indicates that the ATPase activity of MutL is required for its endonuclease activity, that is, ATP is utilized not only to suppress the nonspecific endonuclease activity of MutL but also to actively enhance its activity.
The C-terminal endonuclease domain of MutL homologues contains the highly conserved motif, CPHGRP. Mutations in this motif result in the deficiency of in vivo and in vitro MMR activities [34,85]. Bioinformatic analysis indicated that this motif takes part in the formation of the metalbinding motif and it resembles a metal-dependent transcriptional regulator [85]. Using biochemical procedures, the Cterminal domain of the human PMS2 subunit of MutLα was demonstrated to bind a zinc ion [85]. Although the detailed functions of this zinc ion remain unknown, the involvement of the CPHGRP motif in the ATP-dependent conformational change in T. thermophilus MutL is suggested [34].
As mentioned above, recent biochemical studies on the endonuclease activity of MutL homologues have been achieved by using homologues from thermophilic bacteria, such as T. thermophilus, A. aeolicus, and Thermotoga maritima [34,36,87,92]. Proteins from these bacteria are extremely stable and suitable for physicochemical examinations including crystallographic analysis. In addition, a variety of gene manipulating procedures are established in T. thermophilus. T. thermophilus may be one of the ideal model organisms for the study of nick-directed MMR.

Strand Discrimination in Nick-Directed MMR
Accumulating evidence indicates that a pre-existing strand break can serve as a signal to discriminate the Domain I Domain II Domain III Domain IV Figure 11: Crystal structure of T. thermophilus RecJ (PDB ID: 2ZXR). Full-length T. thermophilus RecJ is comprised of domains I-IV and forms a ring-like structure. The catalytic active site is located in the cavity between domains I and II as indicated by an arrow. Domain III shows a structural similarity to the oligonucleotide/oligosaccharide-binding fold that is often found in single-stranded DNA-binding proteins. The ring-like structure and oligonucleotide/oligosaccharide-binding fold will ensure the high processivity and strict specificity for single-stranded DNA.
error-containing strand in eukaryotic MMR [27,30]. Since newly synthesized strands always contain strand break as 3 -ends or 5 -termini of Okazaki fragments, these ends can be utilized as strand discrimination signals in vivo. This is consistent with the observation that MutSα-dependent yeast MMR corrects mismatches more efficiently in the lagging strand than in the leading strand [93]. As mentioned above, MutLα is responsible for the strand-discrimination by nicking the discontinuous strand of the mismatched duplex (Figures 1 and 3). Interestingly, it is demonstrated that MutLα incises the discontinuous strand at a distal site from the pre-existing strand break. How does MutLα discriminate the discontinuous strand at a site distant from the strand-discrimination signal? One possible explanation is that MutSα (or MutSβ) and MutLα are loaded onto the substrate DNA through their interactions with the replication machinery such as proliferating cell nuclear antigen (PCNA), replication factor C (RFC), and/or DNA polymerase δ to render the newly synthesized strand in the catalytic site of the MutLα endonuclease domain [23,94,95]. It is known that MSH3 and MSH6 contain the PCNA-interacting motif QX 2 (L/I)X 2 FF [96]. In fact, several studies have demonstrated the associations of MutSα and MutSβ with PCNA [95,97,98]. In addition, PCNA and RFC are necessary not only for repair synthesis [99] but also for the mismatch-provoked incision and excision reactions [30,100]. Inhibitors of PCNA abolish 3 -nickdirected excision and 40-50% of 5 -nick-directed excision [99][100][101]. The excision reaction of MutLα-independent 5 -nick-directed MMR may be performed independently of PCNA. In mutH-less bacteria, the C-terminal region of MutS contains the putative β-clamp-binding motif QLSFF [102]. The deletion of this region abolishes the in vitro interaction of MutS with the β-clamp and in vivo MMR activity [103]; furthermore, this interaction is necessary for the in vivo localization of MutS and MutL in response to mismatches [103]. These interactions may also be responsible for the strand discrimination in bacterial nick-directed MMR.

Downstream Events in Nick-Directed MMR
The excision reaction of in vitro eukaryotic MMR is performed by the 5 -3 single-stranded DNA-specific exonuclease EXO1. To date, EXO1 is a unique exonuclease that is involved in eukaryotic MMR [30,104]. In addition, no reports have identified the MMR-related eukaryotic DNA helicase. The exonuclease activity of EXO1 is enhanced by its direct interaction with MutSα [99]. As mentioned above, MutSα forms a sliding clamp and diffuses along the DNA after mismatch recognition. The purpose of the diffusion of MutSα from the mismatch may be to activate EXO1 at the 5terminus. In mutH-less bacteria, A. aeolicus MutL enhances the DNA helicase activity of UvrD [105] whose amino acid sequence is ubiquitous among bacteria. Furthermore, a genetic study implied the simultaneous involvements of the 5 -3 exonuclease RecJ and a 3 -5 exonuclease ExoI in T. thermophilus MMR [106]. In mutH-less bacteria, the errorcontaining DNA segment generated by the endonuclease activity of MutL may be removed bi-directionally by the cooperative function of multiple exonucleases and helicases ( Figure 10). The termination of the EXO1-dependent excision reaction in eukaryotic 3 -nick-directed and MutLα-dependent 5 -nick-directed MMR can be directed by 3 -termini that are pre-existing and newly introduced by MutLα, respectively ( Figure 3). In contrast, excision termination in MutLαindependent 5 -nick-directed MMR appears to employ a relatively complicated mechanism, because the termination of the excision reaction is not directed by the terminus of the DNA strand. Excision termination in MutLα-independent 5 -nick-directed MMR is conducted by the inhibitory function of RPA, a single-stranded DNA-binding protein [32]. In mutH-less bacteria, the mechanism for terminating the excision reaction remains unknown. The implication of the involvements of 5 -3 and 3 -5 exonucleases [106] raises the possibility that the excision termination of 5 -and 3 -nickdirected MMR in mutH-less bacteria is directed by the 3and 5 -termini of the DNA strand that are introduced by MutL ( Figure 10). Biochemical and structural studies on the exonucleases are required to further understand the excision reaction. Structural analyses of the RecJ 5 -3 exonuclease from T. thermophilus were successfully performed [20,107]; thus, proteins from T. thermophilus are known to be suitable for physicochemical examinations [108,109]. The structure of RecJ consists of 4 domains that form a ring-like structure with the catalytic site in the center of the ring (Figure 11). Based on this structure, the molecular basis for the high processivity and substrate specificity of this enzyme was discussed [107]. In addition, a detailed biochemical study on the T. thermophilus ExoI was also performed [106]. It is expected that, unlike other model organisms, T. thermophilus possesses only a single set of 5 -3 and 3 -5 exonucleases [106]. Hence, in vitro and in vivo experiments concerning the excision reaction would be relatively straightforward in this bacterium. It will be intriguing to examine whether MutS can stimulate the exonuclease activities of RecJ and ExoI.

Other Functions
Mismatch base pairs can arise not only from replication error but also from other biological processes including homeologous recombination, oxidation, and methylation of bases. Long-patch MMR also has a role in the repair machinery for those mismatches.
The involvement of MMR proteins in the suppression of homeologous recombination, that is, the strand exchange between nonidentical DNA molecules, has been reported [110,111]. Inhibition of homeologous recombination contributes to genome integrity by limiting the invasion of foreign replicons and the excessive intracellular rearrangement of genome. Although the requirement of MutS and MutL homologues for this inhibitory function has been demonstrated, the downstream reactions following the recognition of the mismatch had not been described. Recently, it is suggested that the endonuclease activity of MutLα is required for this system [112].
Oxidative damage is one of the major spontaneously arising forms of DNA damage. Aerobic cells yield reactive oxygen species via respiration events that attack biomolecules such as proteins, lipids, and DNAs [113]. The attack of reactive oxygen species on DNA bases generates oxidized bases including 8-oxoguanine (8OG) (Figure 12) [114]. An 8OG base can pair not only with cytosine but also with adenine [115]. An 8OG:A pair can be converted to a T:A pair through replication, forming a G:C-T:A transversion mutation. To prevent this mutational process, base-excision repair employs MutM (OGG1 in eukaryotes) and MutY (MYH in eukaryotes) glycosylases that excise 8OG and adenine from 8OG:C and 8OG:A pairs, respectively [115]. In vitro and in vivo studies indicated that bacterial and eukaryotic MMR can recognize an 8OG:A pair as a substrate and remove the adenine residue because the mismatched adenine residue exists in the newly synthesized strand [116,117], that is, MMR can perform the same role as the MutY (MYH)-dependent base-excision repair pathway ( Figure 12). O 6 -methylguanine (O 6 MeG) is generated by the action of S N 1-alkylating agents, such as N-methyl-N-nitrosourea and N-methyl-N -nitro-N-nitrosoguanidine that are used in cancer chemotherapy [118]. O 6 MeG can pair with thymine, resulting in a G:C-A:T transition mutation through replication. Although the major repair activity for O 6 MeG is derived from O 6 MeG methyltransferase and/or its homologues [56,119], MutSα can also bind to an O 6 MeG:T mismatch [43,120]. However, if the O 6 MeG is in the template strand of the duplex, MMR does not remove the lesion. The accumulation of the complex of unrepaired O 6 MeG with MMR proteins is thought to result in the induction of apoptosis [121] through the crosstalk between MMR proteins and check point kinases [120].

Conclusions
In this paper, the molecular mechanism of widely conserved human-type MMR has been described. Since the accumulating evidence indicates the similarities of basic features among bacterial and eukaryotic MMR, mutH-less bacteria may serve as a model organism for biochemical and structural studies of MMR proteins.
The structural analyses on initial recognition complex should be required for further understanding of mismatchrecognition mechanism of MutS homologues. It would be beneficial to identify the amino acid residue which is responsible for the kinking of heteroduplex by MutS homologues. Such mutant might be utilized for the crystallographic analysis of the initial recognition complex. In order to understand the ATP-dependent functional cycles of MutL endonucleases, structural analyses on the full-length MutL homolgoues should be necessary. Since MutL endonucleases from several thermophilic bacteria have relatively short linker region between N-terminal and C-terminal domains [34,36,92,105], these proteins are expected to be suitable for crystallographic analyses. However, we should not underestimate the importance of this interdomain linker region. This region might be responsible for the protein-protein interaction or domain-domain interaction just like intrinsically disordered proteins [122]. It also remains to be clarified how MutL endonucleases discriminate the discontinuous strand of mismatched heteroduplex. Further experiments using in vitro reconstituted system may provide the key findings to understand the mechanism. Discrimination between the "moving" and "stationary" models should also be argued in this context.