Mismatch repair in correction of replication errors and processing of DNA damage


  • Gabriele Aquilina,

    1. Laboratory of Comparative Toxicology and Ecotoxicology, Istituto Superiore di Sanita′, Rome, Italy
    Search for more papers by this author
  • Margherita Bignami

    Corresponding author
    1. Laboratory of Comparative Toxicology and Ecotoxicology, Istituto Superiore di Sanita′, Rome, Italy
    • Laboratory of Comparative Toxicology and Ecotoxicology, Istituto Superiore di Sanita', Viale Regina Elena 299, 00161 Rome, Italy.
    Search for more papers by this author


The primary role of mismatch repair (MMR) is to maintain genomic stability by removing replication errors from DNA. This repair pathway was originally implicated in human cancer through an association between microsatellite instability in colorectal tumors in hereditary nonpolyposis colon cancer (HNPCC) kindreds. Microsatellites are short repetitive sequences which are often copied incorrectly by DNA polymerases because the template and daughter strands in these regions are particularly prone to misalignment. These replication-dependent events create loops of extrahelical bases which would produce frameshift mutations unless reversed by MMR. One consequence of MMR loss is a widespread expansion and contraction of these repeated sequences that affects the whole genome. Defective MMR is therefore associated with a mutator phenotype. Since the same pathway is also responsible for repairing base:base mismatches, defective cells also experience large increases in the frequency of spontaneous transition and transversion mutations. Three different approaches have been used to investigate the function of individual components of the MMR pathway. The first is based on the biochemical characterization of the purified protein complexes using synthetic DNA substrates containing loops or single mismatches. In the second, the biological consequences of MMR loss are inferred from the phenotype of cell lines established from repair-deficient human tumors, from tolerant cells or from mice defective in single MMR genes. In particular, molecular analysis of the mutations in endogenous or reporter genes helped to identify the DNA substrates for MMR. Finally, mice bearing single inactive MMR genes have helped to define the involvement of MMR in cancer prevention. © 2001 Wiley-Liss, Inc.

DNA replication is a highly accurate process (< one error per 1010 nucleotides synthesized) (Kunkel, 1992). The fidelity of base insertion by replicative polymerases is, however, considerably lower. The high level of overall fidelity is attained through DNA polymerase proofreading and post-replicative mismatch repair (MMR) correction. After editing by proofreading, MMR intervenes to correct any remaining errors. To do this, MMR must be able to recognize the newly-synthesized (daughter) strand which contains the incorrect nucleotide. Repair then proceeds through degradation of a long stretch of this error-containing strand, DNA resynthesis and ligation. Other MMR proteins with a more specialized function and involved in a “short-patch” MMR are reviewed in this same issue (Bellacosa, 2001).


In E. coli post-replicative MMR is initiated by binding to the mismatch of the MutS homodimer followed by recruitment of a second homodimer, MutL. Both MutS and MutL functions are retained in mammalian cells. Two mismatch binding complexes formed by MutS homologs, hMutSα and hMutSβ, carry out recognition of replication errors in human cells. Both contain hMSH2, which interacts with hMSH6 in hMutSα (Drummond et al., 1995; Palombo et al., 1995) or with hMSH3 in hMutSβ (Palombo et al., 1996; Genschel et al., 1998). In vitro studies using duplex oligonucleotides indicate that hMutSα preferentially recognizes single base mismatches and loops of 1 base, whereas insertion/deletion loops (IDLs) containing between 2 and 8 extra-helical bases are bound by the hMutSβ complex (Risinger et al., 1996; Marra et al., 1998; Genschel et al., 1998) (Fig. 1). There is some overlap in recognition specificity and both complexes recognize IDLs of 1 base. A similar bias was observed in S. cerevisiae (Iaccarino et al., 1996; Habraken et al., 1996). Repair of large IDLs (from 12 up to 216 unpaired nucleotides) is independent of long-patch mismatch repair in both humans and yeast (Littman et al., 1999; Corrette-Bennett et al., 1999).

Figure 1.

Steps involved in mammalian mismatch repair. Different types of mismatches (a base:base mispair and IDLs of 1 or 2 bases) are recognized by MutSα and MutSβ in combination with heterodimers of MLH1 with PMS2, PMS1 or MLH3. The MMR correction is targeted to the strand containing an interruption in the primer strand, possibly through the interaction of MSH3 and/or MSH6 with PCNA. The strand containing the mismatch is exonucleolytically degraded by a 3′-> 5′ exonuclease and resynthesis is then performed by DNA polymerase δ and PCNA as a processivity factor.

The highly conserved C-termini of MSH proteins contain ATP binding sites. Binding to DNA mismatches stimulates the intrinsic ATPase activity of hMutSα while binding of ATP provokes a dissociation of the complex from the DNA substrate (Gradia et al., 1997; Gradia et al., 1999; Iaccarino et al., 1998; Blackwell et al., 1998). The formation of a sliding clamp on DNA has been proposed to be the consequence of this mismatch-provoked ADP-> ATP exchange (Gradia et al., 1997).

Upon binding the mismatch, hMutSα associates with another heterodimeric complex (hMutLα) formed by the MutL homologs hMLH1 and hPMS2 (Li and Modrich, 1995). In E. coli and S. cerevisiae gel shift experiments indicate that the presence of MutL or MutLα increases the efficiency of MutS binding to mismatched DNA (Drotschmann et al., 1998; Prolla et al., 1994; Habraken et al., 1997). Co-immunoprecipitation, in an ATP-dependent manner, of hMLH1/hPMS2 and hMSH2 using anti-hMLH1 antibodies suggests that human cells might also organize their hMutSα and hMutLα in a large complex both contributing to mismatch recognition (Gu et al., 1998).

Candidate ATPase domains in yeast MutLα are required for a proper heterodimerization of the two repair proteins and for mutation avoidance (Tran et al., 2000). Since in bacteria the ATP-dependent conformational changes of MutL are considered to act as a switch to coordinate downstream events in MMR it is also possible that eukaryotes use the same strategy (Ban et al., 1999).

Recent evidence suggests that the assembly of the human repairosome might share some of the complexities of the recognition step. Genetic evidence indicates that the main MMR complex in yeast is the heterodimer of Mlh1 with Pms1 (which correspond to the human hPMS2) (for a review, see Kolodner and Marsischky, 1999). However, Mlh1 can form complexes with two other MutL homologs, Mlh2 and Mlh3 (Wang et al., 1999c; Harfe et al., 2000). The Mlh1/Mlh3 and Mlh1/Mlh2 heterodimers appear to interact preferentially with MutSβ to correct a small fraction of IDLs, with the Mlh1/Mlh2 complex being also involved in meiotic recombination (Flores-Rozas and Kolodner, 1998; Harfe et al., 2000; Wang et al., 1999c). The mammalian homolog of the yeast Mlh3 gene has been cloned and the microsatellite instability associated with expression of a dominant-negative MLH3 protein is consistent with its role in MMR (Lipkin et al., 2000). Furthermore, a third complex (hMutLβ) formed by hMLH1 and hPMS1 (closely related to yeast MLH2 and/or MLH3) has been identified in human cells also (Raschle et al., 1999; Leung et al., 2000). Whether human MMR uses different repairosomes to correct different types of mismatches or whether some of these complexes perform some more specialized funtion is unclear.

The strand discrimination signal, the degradation of the newly synthesized strand and the resynthesis of the excised tract are less well-characterized. While E. coli uses the methylation pattern at GATC sequences as the signal to identify the newly synthesized strand, there is no obvious equivalent in human DNA. The interaction of hMSH2 with the replication processivity factor PCNA at a step preceding DNA resynthesis led to the suggestion that human MMR might use the termini of Okazaki fragments as the strand discrimination signal (Umar et al., 1996). This hypothesis is supported by the observation that an anti-hMLH1 antibody is able to co-immunoprecipitate PCNA together with hPMS2 and hMSH2 (Gu et al., 1998) and that yeast PCNA mutated in the binding motifs for MSH3 and MSH6 displays a mutator phenotype that is epistatic with mutations in MMR genes (Clark et al., 2000).

The interaction of human exonuclease I with hMSH2 suggests its involvement in the MMR process and/or DNA recombination (Schmutte et al., 1998). Biochemical evidence indicates that polymerase δ is used in the repair of DNA mismatches (Longley et al., 1997). The recent discovery of several novel eukaryotic DNA polymerases (for a review, see Friedberg et al., 2000) raises the possibility that other polymerases might be used during MMR (see also below).


Mutator phenotype

In accordance with a major role of MMR in the correction of replication errors, human tumor cell lines with defective MMR display elevated spontaneous mutation rates (order of magnitude between one and three) (Bhattacharyya et al., 1994; Eshelman and Markowitz, 1996; Branch et al., 1995). Molecular analysis of spontaneous mutations in endogenous (HPRT) or reporter genes (lacI, supF) helped to identify the substrates for MMR. The frequencies of transitions, transversions and frameshift mutations are all greatly increased in the HPRT gene of repair defective human cell lines, although there are major differences in the mutational spectra of hMSH6- and hMLH1- or hPMS2-defective tumor cell lines. Inactivation of hMutLα was associated with a greater increase in frameshifts than loss of hMSH6 (Bhattacharyya et al., 1995; Malkhosyan et al., 1996a; Ohzeki et al., 1997; Glaab et al., 1998a; Lettieri et al., 1999). Thus, hMutSβ appears to support repair of a significant fraction of frameshift intermediates. In addition, differences between mutational spectra of double hMSH6/hPMS2 or hMSH3/hMSH6 versus single hMSH6 or hMSH3 mutant cell lines indicate that the hMutSβ heterodimer also participates in repair of base: base mismatches (Glaab et al., 1998a; Umar et al., 1998). Studies of mutation on the SupF gene replicated episomally in MMR-defective cells also suggest a role for hMutSβ in correcting single base mispairs (Ceccotti et al., 2000). These data partially contradict some of the in vitro studies with purified MMR complexes. Thus, the affinity of MutSα and MutSβ for mismatched DNA might be influenced by several other factors, the most obvious being the interaction with alternative multiple MutL complexes or the sequence context where the mismatches occur.

Within microsatellites, loss of hMSH2, hMLH1 or hPMS2 leads to a pronounced instability at both mono- and dinucleotide repeats while inactivation of hMSH6 is associated only with instability at mononucleotide runs (Bhattacharyya et al., 1994). Both in vitro and in vivo evidence indicates that, in the absence of hMutSα, hMutSβ is sufficient to support repair of 2-base IDLs.

The contribution of the components of the MutLα complex to microsatellite stability has been studied in mice. The degree of mononucleotide instability is 2–3-fold higher in Mlh1−than in Pms2−defective animals (Yao et al., 1999). The larger proportion of−1 frameshifts in Mlh1 mice is suggestive of a residual repair activity in Pms2−/− mice. Thus, other Mlh1 containing complexes might play a role in the repair of frameshift intermediates.

Redundancy in the repair functions suggests that inactivation of more than one MMR gene is required for a strong mutator phenotype. Mouse embryo fibroblasts derived from single MMR-defective mice generally show more moderate mutator phenotypes (5–20-fold increase) than cell lines derived from MMR-defective human tumors (Reitmar et al., 1997; Andrew et al., 1997). Interestingly, most of the human tumor cell lines are indeed mutated in more than one MMR gene. Examples are cell lines defective in hMSH6/MLH1 (SW48), hMSH3/MLH1 (DU145), hPMS2/MSH6 (HEC1A). This would be consistent with the apparent redundancy between the hMutSα and hMutSβ complexes. However, the occurrence of mononucleotide frameshifts in hMSH6 and hMSH3 suggests that they arise as a consequence of a, possibly partial, mutator phenotype conferred by inactivation of hMutLα (Malkhosyan et al., 1996b). The recently revealed redundancy among hMutL complexes may help resolve this apparent paradox.

The mutator phenotype is much more pronounced in tumor tissues than in normal organs of Msh2 mice and mutation rates approach those of repair-defective human tumor cell lines (up to 300-fold) (Baross-Francis et al., 1998). It is currently uncertain whether this reflects inactivation of a second repair locus or whether other metabolic changes in tumors contribute to the increased mutation rates. Possibilities might include altered levels of endogenous mutagenic DNA damage or the involvement of a particularly error-prone DNA polymerase.

Tumor susceptibility

In humans MMR is associated with HNPCC, a relatively common genetic disease that accounts for about 5% of all colorectal cancers. Most HNPCC individuals inherit a mutation in one allele of either hMLH1 or hMSH2. Mutations in hPMS1, hPMS2 and hMSH6 are quite rare (Liu et al., 1996). With few exceptions, the presence of a wild-type allele appears to be sufficient for normal MMR activity and cancer progression in predisposed individuals results from somatic mutation of this normal copy leading to a mutator phenotype. Large bowel cancer, carcinomas of the endometrium, ovary, small intestine and stomach are present with increased frequency in HNPCC families. In contrast, hematological malignancies (acute myeloid leukemia and non-Hodgkin lymphoma) were found in two families with homozygous inactivation of the hMLH1 gene (Wang et al., 1999b; Ricciardone et al., 1999). MMR deficiency associated with microsatellite instability has also been found in sporadic colon cancer as well as in cancer at other sites. In many cases of sporadic cancer with microsatellite instability MMR inactivation is not the consequence of mutation in MMR genes but is due to epigenetic silencing of the hMLH1 gene through promoter methylation (Kane et al., 1997; Esteller et al., 1998; Veigl et al., 1998).

Tumor susceptibility is generally associated with MMR inactivation in mice. An excellent review has been recently published on this subject (Buermeyer et al., 1999a). Unlike HNPCC individuals, mice heterozygous for MMR defects do not develop cancer at an early age and inactivation of both alleles is required for cancer proneness. The Msh2−/− and Mlh1−/− genotypes—involving the genes most commonly mutated in HNPCC families—are particularly cancer prone and mostly develop lymphoma with some neoplasms of the gastrointestinal tract.

The cancer proneness of various knockout animals may also reflect redundancies among repair factors. Thus, Msh6−/− mice display a mild phenotype and tumor development is not detectably accelerated in Msh3−/− mice. Doubly Msh6/Msh3-defective animals exhibit the extreme tumor susceptibility of Msh2 knockouts (de Wind et al., 1999). The PMS2−/− phenotype is also less severe than that of Mlh1−/− and is not associated with significant cancer proneness. This again might reflect redundancy at the MutL step of correction.


MMR is implicated in the processing of DNA damage and loss of the repair pathway is associated with changes in the toxicity of several DNA damaging agents. Since the role of MMR is to remove DNA mismatches it is not surprising that mispairs formed in DNA by exposure to some DNA damaging agents can be recognized by MMR. The modulation of toxicity by MMR, however, is not easily predictable and some of the data reported in the literature are summarized in Table 1.

Table 1. MMR modulation of toxicity induced by mutagens
  • a

    Contradictory results have been reported for DNA damaging agents listed in this column.

Methylation toleranceγ-raysCCNU Mitomycin C
Temozolomide, ProcarbazineBenzopyreneHalogenated thymidine
6-TG analogs+irradiation
Cisplatin, carboplatin  
Minor groove alkylating agents  

Methylation damage and 6-thioguanine

The involvement of MMR in damage recognition was initially unveiled in E. coli where loss of this repair pathway, in a dam background, restored resistance to the cytotoxicity of N-methyl-N′-nir-N-nitrosoquanidine (MNNG) (Karran and Marinus, 1982). Inactivation of MMR also confers high levels of methylating agent resistance (increases up to 100-fold in D37) in eukaryotes, with the single known exception of yeast (for review see ref. Karran and Bignami, 1994). This resistance does not involve increased removal of DNA lesions and is thus a DNA damage tolerance phenomenon. Cells defective in MSH2, MSH6, MLH1 and PMS2 all exhibit methylation tolerance and restoration of MMR by transfer of the appropriate human chromosome (Koi et al., 1994; Karran and Hampson, 1996) and by cloned hMLH1 in Mlh1−/− mouse embryo fibroblasts or human cells restores sensitivity (Buermeyer et al., 1999b; G. Aquilina et al., unpublished data). Furthermore human tumor cells selected for resistance to the toxic effects of methylating agents, either in culture or in xenografts, are found to have MMR defects (Branch et al., 1993; Kat et al., 1993; Friedman et al., 1997). The peculiarity of MMR is, therefore, that an intact MMR function is required for the cytotoxicity of methylating drugs. It is well established that O6-methylguanine (O6-megua) is the main cytotoxic lesion induced by methylating agents (for a review, see Bignami et al., 2000). This adduct is normally repaired through a single-step, error-free repair reaction by the O6-megua-DNA-methyltransferase protein (MGMT). MGMT levels are however low in some human tissues and O6-megua can be left unrepaired. At replication either cytosine or thymine may be incorporated opposite unrepaired O6-megua resulting in the fixation of G to A transitions. Oligonucleotides containing O6-megua paired either with T or, to a lesser extent with C, are bound by hMutSα (but not MutSβ) (Griffin et al., 1994; Duckett et al., 1996) and the mismatch-dependent ATPase activity of hMutSα is stimulated by O6-megua/T pairs (Berardini et al., 2000). While the relationship between MMR and methylation tolerance is well-established, just how processing of these mismatches leads to cell death is still a matter of debate. A model, originally proposed for E. coli, invoked the ambiguous coding properties of O6-megua and the possibility of death as a consequence of incomplete repair events. We have adapted this model to take into account death in human cell induced by methylation damage (Fig. 2). O6-megua-containing mismatches trigger post-replicative MMR that is directed to the newly synthesized strand containing the T or the C. Since the O6-megua in the template strand is not removed, the resynthesis step will again generate O6-megua-containing mismatches. This incomplete processing by MMR of the O6-megua-containing mismatches might lead to repetitive and futile cycles of excision-resynthesis that result in the generation of persisting DNA termini. Replication of these lesions in the following cell cycle elicits cellular death by apoptosis or necrosis. Various experimental evidence fits into the model. Chromosomal aberrations and SCEs induced by methylating agents are largely dependent on active MMR and are mainly formed in the second cycle after treatment (Galloway et al., 1995). Secondly, hMSH6-defective human cells perform less MNNG-induced homologous recombination (Zhang et al., 2000). Finally, in repair competent cells, cell cycle arrest and cell death are observed only after a first, apparently normal, cellular division (Zhukovskaya et al., 1994). Alternative models propose that the assembly of hMutSα and hMutLα at sites of methylation damage or a threshold number of DNA-bound sliding clamps directly trigger a damage-signaling cascade that activates cell cycle checkpoints or cell death (Hawn et al., 1995; Duckett et al., 1996; Carethers et al., 1996; Berardini et al., 2000).

Figure 2.

Model for the role of MMR in toxicity of methylating agents. Recognition, following replication, of O6-megua:C or O6-megua:T mismatches by hMutSα and recruitment of hMutLα initiates futile cycles of excision repair. Replication past a gap left by removal of a mismatched T produces double strand breaks leading to cell death.

In vivo recruitment of the hMutSα complex to the O6-megua mispairs can be inferred by the observations that hMutSα translocates from the cytoplasm to the nucleus when cells are treated with methylating agents (Christmann and Kaina, 2000). Methylating agent-induced apoptosis depends on active MMR in cultured human tumor cells and in the mouse intestine in vivo (D'Atri et al., 1998; Toft et al., 1999). Methylation damage provokes phosphorylation of the tumor suppressor p53 on serine residues 15 and 392 in a hMutSα− (but not MutSβ−) and hMutLa-dependent manner (Duckett et al., 1999; Hickman and Samson, 1999). Although this p53 stabilization might signal the initiation of apoptosis in response to O6-meGua, MMR-dependent apoptosis can be executed in a p53-independent manner (Aquilina et al., 1998; Hickman and Samson, 1999). Interestingly in the murine small intestine Msh2 was required for both a p53-dependent and p53-independent apoptotic response following methylation damage (Toft et al., 1999). More recently, a MMR-dependent stabilization of p73 after cisplatin exposure has involved this p53 homolog in the activation of an apoptotic pathway (Gong et al., 1999). Whether methylation damage and cisplatin use the same p73-dependent pathway to activate apoptosis and whether p73 and p53 are interchangeable in their damage-dependent signaling activities is still unclear.

MMR-defective cells are hypermutable by methylating agents (Glaab et al., 1998b; Bignami et al., 2000). Since O6-meGua is essentially the only pre-mutagenic lesion induced by these agents, it appears that the probability of a mutation at O6-meGua is higher in a MMR-defective background. This observation might be explained by a MMR-mediated recruitment of an error-free DNA polymerase which would preferentially direct the insertion of a C opposite the methylated base.

Methylation tolerance and the consequent increased susceptibility to methylation induced mutations might both contribute to the increased susceptibility of MMR defective mice to carcinogenesis by alkylating agents. Exposure to the ethylating or methylating carcinogens accelerates the appearance of lymphomas in Msh2−/− or PMS2−/− mice (de Wind et al., 1998; Qin et al., 1999). Similarly treatment with 1,2-dimethylhydrazine increases the rate of appearance of both lymphoma and colorectal carcinoma in Msh2−/− mice (M. Bignami, unpublished). Thus, in addition to threats posed by increased spontaneous mutagenesis, a defect in MMR confers an extra risk of acquiring transforming mutations through DNA damage induced by exogenous mutagens.

Methylation tolerance consequent to MMR deficiency is regularly associated with cross-resistance to the base analog 6-thioguanine (6-TG). The molecular mechanism of this cross-resistance was clarified some years ago with the discovery that a methylated derivative of 6-TG may be processed by MMR system in a way similar to O6-megua (Swann et al., 1996; Waters and Swann, 1997). After incorporation in DNA, 6-TG undergoes a non-enzymatic methylation by S-adenosylmethionine (SAM) that leads to the formation of S6-methylthioguanine (S6-meG). SAM is a cofactor involved in several biosynthetic reactions, but it may occasionally work as a weak methylating agent. The thiol group of 6-TG is a good substrate for SAM and the product S6-meG is not repaired by the MGMT. During replication S6-meG codes ambiguously to direct insertion of T or C with similar frequency. S6-meG:T is efficiently bound by hMutSα while the stability of the complex on S6-meG:C depends to some extent on the sequence context. By analogy to O6-megua, following recognition by hMutSα, MMR processing of the lesion is lethal. Thus, both human tumor cell lines and mice with MMR defects are tolerant to the toxic effects of 6-TG (Aquilina et al., 1995; de Wind et al., 1995). Although S6-meG has miscoding properties, 6-TG is not a good mutagen, possibly because of a prevailing toxicity. However, in a MMR-defective background, 6-TG mutagenic properties are revealed (Aquilina et al., 1993). Tolerance to 6-TG cells is less pronounced than the concomitant tolerance to SN1 methylating agents, suggesting that a fraction of 6-TG cytotoxicity might be due to other factors such as alterations of purine biosynthesis.

Minor levels of tolerance have been also reported with alkylating agents that alkylate the N3 adenine lying in the minor groove (Colella et al., 1999).


Cisplatin, a widely used chemotherapeutic agent in testicular, ovarian and bladder cancers, is cytotoxic because it forms covalent adducts with DNA. The major reaction products are 1,2-intrastrand cross-links between N7 atoms of adjacent purines, with 1,2 GpG and 1,2 ApG representing 65 and 25%, respectively, and the 1,3-adducts in GpNpGp (6%). The therapeutic effectiveness of cisplatin is limited by both intrinsic and acquired resistance that is multifactorial. Some evidence suggests that MMR deficiency may contribute to drug tolerance. The extent of this contribution and its possible clinical relevance are still under active investigation.

In E. coli, defects in MMR affect sensitivity to both methylating agents and cisplatin in a dam-dependent fashion. The hypersensitivity of dam mutants is abolished by a second mutation in the mutL gene (Fram et al., 1985). In S. cerevisiae the inactivation of several MMR genes (MLH1, MLH2, MSH2, MSH3, MSH6, but not PMS1) is associated with a small increase in cisplatin and carboplatin resistance. Reintroducing a wild type MLH1 gene into the mlh1 mutant strain restores cisplatin sensitivity (Durant et al., 1999). Since there is no parallel MMR-related tolerance for methylating agents in S. cerevisiae, the mechanism for resistance to cisplatin might be mechanistically different from methylation tolerance.

MMR proteins can recognize and bind cisplatin adducts. The MutSα complex binds DNA duplexes containing a 1,2 GpG intrastrand cross-link, although with a lower affinity than those with a G:T mismatch. No binding is observed with 1,3- intrastrand adduct (Duckett et al., 1996; Yamada et al., 1997). Binding comparable to that for a G:T mispair is observed if the duplex contains a mispaired T opposite the 3′ G of the cisplatin cross-link (Yamada et al., 1997). This mismatched duplex is a possible product of mutagenic replication bypass of the 1,2 GpG cisplatin cross-link. In addition, the formation of a DNA-protein complex that contained both hMSH2 and hMLH1 was observed by gel retardation assays using oligonucleotides platinated with cisplatin but not with oxaliplatin, an analog of cisplatin whose toxicity is independent of MMR proficiency (Fink et al., 1996).

MMR deficiency is associated with increased cisplatin resistance in several human tumor cell lines, defective in either hMutSα or hMutLα. Selection of human ovarian carcinoma cells for resistance to cisplatin frequently leads to the isolation of clones defective in the expression of hMLH1 (Anthoney et al., 1996; Drummond et al., 1996). The resistance factor in clones selected by multiple exposure to cisplatin varies greatly (between 2 and 50-fold), whereas cisplatin resistance in cisplatin-naïve isogenic pairs of MMR-proficient or MMR-defective cell lines is never greater than 2-fold (Fink et al., 1996; Aebi et al., 1996). The hMLH1-defective human colon carcinoma line HCT116 is 2-fold more resistant to cisplatin than the MMR-proficient subline where the MMR defects is corrected by a normal copy of chromosome 3. A two-fold increase in cisplatin resistance was also observed in the hMSH2-defective human endometrial cancer line HEC59 compared to an hMSH2-proficient line bearing an extra chromosome. Two MMR-defective cell lines are also resistant to Carboplatin which forms similar types of DNA adducts to cisplatin.

Independent introduction of either an hMLH1 or a p53 defect into the model A2780 human ovarian tumor cells did not provide evidence for a major contribution of defective MMR to cisplatin resistance. In this case, p53 was revealed as a more potent determinant of drug sensitivity (Branch et al., 2000). Overall, it appears likely that large increases in cisplatin tolerance are due to factors other than MMR. These might include altered levels of drug metabolizing enzymes that reduce the levels of DNA reaction or alteration in pathways controlling DNA damage responses or apoptosis.

Part of the cellular response to cisplatin is mediated by the activation of JNK (c-Jun NH2-terminal kinase) and c-Abl kinase. In MMR-defective human tumor cells c-Abl activation is undetectable and JNK activation is 3.8-fold less efficient than in the MMR-proficient counterpart (Nehme' et al., 1997). The increased resistance to cisplatin of Mlh1- or c-Abl -defective mouse embryo fibroblasts suggests that both c-Abl and p73 are components of an MMR-dependent apoptotic pathway (Gong et al., 1999).

Oxidative DNA damage

The role of MMR in sensitivity to ionizing radiation is controversial. Contradictory consequences of MMR deficiency have been reported in human and mouse cells, although the effects are generally minor (Fritzell et al., 1997; Davis et al., 1998; Aquilina et al., 1999; Zeng et al., 2000). Similar discrepancies were found in the ability of MMR-defective cells to activate the G2/M checkpoint in the presence of oxidative DNA damage. In cases where loss of MMR increases cell survival, a concomitant decrease in the level of apoptosis induced by oxidative damage is observed (DeWeese et al., 1998; Zeng et al., 2000). Doubly defective p53 and Pms2 mice suggest that p53 and MMR mediate X-ray cytotoxicity via non-overlapping pathways (Zeng et al., 2000).

Ionizing radiation induces numerous DNA modifications such as single and double strand breaks, abasic sites and several modified bases. The oxidized bases thymine glycol and 8-oxoguanine (8-oxoG) accumulate selectively in the DNA of Msh2-defective mouse embryo fibroblasts exposed to low dose rates of ionizing radiation (DeWeese et al., 1998; M. Bignami unpublished observations). This increased level of oxidative lesions was accompanied by increased mutagenesis. 8-oxoG is removed from DNA by base excision repair (BER). A specific glycosylase, OGG1 removes 8-oxoG from 8-oxoG:C pairs and the resulting AP site is repaired predominantly through a 1 nucleotide gap-filling reaction by DNA polymerase β (Fortini et al., 1999). 8-oxoG is a miscoding lesion and forms 8-oxoguanine: A mismatches during replication. If uncorrected, these give rise to G − > T transversions. A second DNA glycosylase, MYH, the mammalian homolog of the E. coli MutY, removes A from 8-oxoG:A mismatches thereby avoiding mutations. It is not clear which are the mismatches recognized by MMR and whether MMR provides a proper repair function or a tolerance mechanism is put in action. In S. cerevisiae genetic and biochemical evidence implicates MSH2 and MSH6 in the removal of A opposite 8-oxoguanine following DNA replication (Earley and Crouse, 1998; Ni et al., 1999). Following the removal of A mispaired with 8-oxoG by MMR, the error-free DNA polymerase Polη correctly pairs C with 8-oxoG. It has been proposed that the interaction between Msh2/Msh6 and polη might target this polymerase to 8-oxoG:A mismatches (Haracska et al., 2000). Since yeast apparently lacks MYH it is not clear whether the yeast model can be straigthforwardly applied to mammalian cells.

Transcription coupled repair and repair of UV damage and bulky adducts

In bacteria both MutS and MutL are required for transcription coupled repair (TCR) of actively transcribed genes (Mellon and Champe, 1996a). While no defects in TCR of pyrimidine dimers were found in yeast with mutations in MSH2, MLH1, PMS1, and MLH1/PMS1 (Sweder et al., 1996; Leadon and Avrutskaya, 1998), human tumor cell lines defective in hMSH2 or hMLH1 performed less TCR of UV damage (Mellon et al., 1996; Leadon et al., 1997).

Although some binding by hMutSα was observed with T<>T/AG, (6-4)T/AG photoproducts (Wang et al., 1999a), in vitro assays of nucleotide excision repair (NER) with purified proteins (where no TCR occurs) showed that the human long-patch MMR does not influence significantly the repair of T<>T dimer and (6-4) photoproducts (Mu et al., 1997). MMR-defective cells are not generally sensitive to UV-damage (Afzal et al., 1995; Leadon and Avrutskaya, 1997; Humbert et al., 1999) and the combined deficiency in MMR (hMSH2) and NER (XPA) does not modify the typical sensitivity of XPA cells to killing and SCE induction by UV (O'Driscoll et al., 1999). Collectively, these data suggest that MMR does not contribute significantly to general NER of UV photoproducts. A role in TCR has not, however, been excluded.

The biological significance of hMutSα binding to 2-aminofluorene and N-acetyl-2-aminofluorene DNA adducts (Li et al., 1996) is not completely understood and contrasting results have been reported on the role of MMR in modifying the toxicity of bulky adducts (Wu et al., 1999; Zhang et al., 2000).

MMR is involved in TCR of oxidative damage and hMutSα-defective (but not hMLH1-defective) human cells are unable to remove thymine glycol from the transcribed strand of an active gene (Leadon and Avrutskaya, 1997). In this case yeast resembles human cells since mutations in MSH2, but not in MLH1 or PMS1, lead to a defect in the removal of thymine glycols. Since the double mutants Mlh1/Pms1 were deficient in TCR of thymine glycols, it has been proposed that the two MutL homologs have overlapping roles of in TCR of oxidative DNA damage (Leadon and Avrutskaya, 1998).

Loss of MMR and sensitivity to chemical agents

Deficiency in MMR is not always associated with cellular resistance to DNA damaging agents. Hypersensitivity to N-(2-chloroethyl)-N′-cyclohexyl-N-nitrosourea or mitozolomide, both of which introduce DNA interstrand crosslinks, has been observed in methylation tolerant cells (Green et al., 1989) and cells with defined defects in either hMutSα or hMutLα (Aquilina et al., 1998). Sensitivity conferred by MMR loss is extended to another cross-linking agent mitomycin C both in vitro and in vivo (Fiumicino et al., 2000). Although the role of MMR in lethality has not been identified, the increased level of chromosomal damage induced by CCNU in MMR-defective cells suggests that this pathway removes some clastogenic DNA lesions induced by this drug. Although limited to the comparison of the hMLH1-HCT116 ± chromosome 3, MMR-deficient cells showed also an increased susceptibility to killing and mutagenesis following exposure to the intercalating frameshift mutagen ICR191 (Chen et al., 2000).

Finally, the radiosensitizing effect of the halogenated thymidine analogs idodeoxyuridine and bromodeoxyuridine is particularly evident in hMLH1- and Msh2-defective cells (Berry et al., 1999). This probably reflects a higher level of incorporation of the dThd analog into DNA of MMR-defective cells (Berry et al. 2000).

Since a significant fraction of human cancers are MMR-defective an understanding of the mechanisms underlying MMR-related sensitivity or tolerance to chemotherapeutic drugs provided by loss of MMR is an important clinical goal. The ultimate aim of these studies is to provide information on which more rational clinical intervention can be based.