Introduction

DNA damage accumulates in cells over time as a result of exposure to exogenous chemicals and physical agents (i.e., benzo[a]pyrene, polychlorinated biphenyls, dioxin, cigarette smoke, asbestos, ultraviolet light, radon), as well as endogenous reactive metabolites including reactive oxygen and nitrogen species (ROS and NOS). Another source of DNA damage is errors that occur during normal DNA metabolism or aberrant DNA processing reactions, including DNA replication, recombination, and repair. Nucleotide misincorporation generates DNA base-base mismatches during DNA synthesis at variable rates, depending on many factors, including the specific DNA polymerases. In general, the replicative DNA polymerases have relatively high replication fidelity (see McCulloch and Kunkel, this issue), while translesion DNA polymerases, which specifically bypass sites of DNA damage, have lower replication fidelity (see Andersen et al. and Gan et al. in this issue). DNA damage, if unrepaired, has the potential to generate mutations in somatic or germline cells, which can alter cellular phenotype and cause dysfunction and disease. To prevent such deleterious effects and safeguard the integrity of the genome, cells possess multiple mechanisms to repair DNA damage and thus prevent mutations. One such system is the critical pathway known as DNA mismatch repair (MMR).

MMR corrects DNA mismatches generated during DNA replication, thereby preventing mutations from becoming permanent in dividing cells 1, 2, 3. Because MMR reduces the number of replication-associated errors, defects in MMR increase the spontaneous mutation rate 4. Inactivation of MMR in human cells is associated with hereditary and sporadic human cancers 1, 3, 5, and the MMR system is required for cell cycle arrest and/or programmed cell death in response to certain types of DNA damage 6, 7. Thus, MMR plays a role in the DNA damage response pathway that eliminates severely damaged cells and prevents both mutagenesis in the short term and tumorigenesis in the long term.

The prototypical Escherichia coli MMR pathway has been extensively studied and is well characterized both biochemically and genetically. Thus, E. coli MMR is a useful and important framework for understanding eukaryotic MMR. In this review, the biochemistry and genetics of E. coli MMR will be described briefly by way of introduction, and the remainder of the discussion will focus on the cellular functions of MMR and their roles in cancer avoidance in mammalian cells. For areas of research on human MMR not discussed in this paper or an additional discussion of MMR in other species, readers are referred to the following excellent reviews: 1, 2, 3, 8, 9, 10, 11.

Mechanism of mismatch correction

DNA MMR in E. coli

E. coli MMR requires the following protein components: MutS, MutL, MutH, DNA helicase II (MutU/UvrD), four exonucleases (ExoI, ExoVII, ExoX, and RecJ), single-stranded DNA binding protein (SSB), DNA polymerase III holoenzyme, and DNA ligase 12, 13. MutS, MutL, and MutH initiate MMR and play specialized biological roles in MMR in E. coli.

MutS recognizes base-base mismatches and small nucleotide insertion/deletion (ID) mispairs, and thus MutS has been called the ā€œmismatch recognitionā€ protein 3. MutS possesses intrinsic ATPase activity. High-resolution structures of MutS bound to DNA have been determined by X-ray crystallography 14, 15. These structures revealed that MutS binds to a mismatch as a homodimer. Interestingly, the mismatch-binding site is comprised of sequence-wise identical but structurally and functionally different domains from the two subunits, indicating asymmetry in the protein-DNA complex. Hence, the MutS homodimer acts as a virtual heterodimer when bound to a DNA mismatch. This characteristic is mimicked by eukaryotic MutS homologs (MSH), which function as heterodimers instead of homodimers (see below). MMR in E. coli is ATP-dependent, and requires the functional MutS ATPase.

MutL interacts physically with MutS, enhances mismatch recognition, and recruits and activates MutH. Defects in MutL completely inhibit MMR in E. coli. Despite the fact that a functional human MutL homolog, MutLĪ±, possesses an endonuclease activity that is essential for mammalian MMR 16, no hydrolytic activity has been detected in MutL. However, MutL may play a role as a molecular matchmaker that facilitates assembly of a functional MMR complex 3, 17, because it stimulates the loading and the processivity of helicase II (or UvrD) at the MMR initiation site 18, 19. Like MutS, MutL functions as a homodimer and possesses ATPase activity 20. Mutations in the ATP-binding domain lead to a dominant negative mutator phenotype 21. MutL mutants that are defective in ATP hydrolysis but proficient in ATP binding can activate MutH but cannot stimulate MutH in response to a mismatch or MutS, suggesting that ATP hydrolysis by MutL is essential for mediating the activation of MutH by MutS 22. Recent studies show that MutL interacts physically with the clamp loader subunits of DNA polymerase III 23, 24, suggesting that MutL may promote binding of DNA polymerase III to MMR intermediates. These observations suggest that MMR is coupled with DNA replication.

In E. coli, DNA is methylated at the N6 position of adenine in dGATC sequences. In replicating DNA, the daughter strand is transiently unmethylated, and it is the presence of hemimethylated dGATC sequences that molecularly distinguishes the newly synthesized daughter strand from the parental DNA strand. In MMR, hemi-methylated dGATC sites determine the strand specificity of repair. MutH, which recognizes hemimethylated dGATC sequences, functions as a monomer and belongs to a family of type-II restriction endonucleases 25, 26. Upon its recruitment and activation by MutS and MutL in the presence of ATP, MutH specifically incises the unmethylated daughter strand of hemimethylated dGATC 3, 22, and this strand-specific nick provides the initiation site for mismatch-provoked excision.

The first step of the MMR pathway is binding of a MutS homodimer to the mismatch. Subsequently, a hemi-methylated dGATC site 5ā€² or 3ā€² to the mismatch is located and cleaved by the concerted action of MutS, MutL, MutH, and ATP. Three models have been proposed to address how mismatch binding by MutS leads to cleavage of the hemimethylated dGATC site (see the section of Unsolved Fundamental Problems in MMR for details). The strand-specific nick generated by MutH at hemimethylated dGATC is a starting point for excision of the mispaired base. In the presence of MutL, helicase II (UvrD) loads at the nick and unwinds the duplex from the nick towards the mismatch 18, generating single-strand DNA, which is rapidly bound by single-stranded DNA-binding protein (SSB) and protected from nuclease attack 27. Depending on the position of the strand break relative to the mismatch, ExoI or ExoX (3ā€²ā†’5ā€² exonuclease), or ExoVII or RecJ (5ā€²ā†’3ā€² exonuclease) excises the nicked strand from the nicked site (the dGATC site) up to and slightly past the mismatch. The resulting single-stranded gap undergoes repair DNA resynthesis and ligation by DNA polymerase III holoenzyme, SSB, and DNA ligase 3.

These early studies on E. coli MMR demonstrate three key features of this important pathway: first, repair is strand specific (i.e., restricted to the newly synthesized DNA strand); second, repair is bi-directional, proceeding 5ā€²ā†’3ā€² or 3ā€²ā†’5ā€² from the nick to the site of the mismatch; and third, MMR has broad substrate specificity including base-base mismatches and small ID mispairs. All of these properties require functional MutS, MutL, and MutH. Because the mechanism of MMR is highly conserved throughout evolution, E. coli MMR is an excellent model for MMR in eukaryotic cells.

MMR in human cells

MMR is a highly conserved biological pathway with strong similarities between human MMR and prototypical E. coli MMR 2, 3. These similarities include substrate specificity, bidirectionality, and nick-directed strand specificity. The role of hemi-methylated dGATC sites as a signal for strand discrimination is not conserved from E. coli MMR to human MMR, but because the hemi-methylated dGATC site directs MutH-dependent nicking, and because human MMR is presumed to be nick-directed in vivo, both systems are thought to discriminate daughter and template strands using a strand-specific nick.

Several human MMR proteins have been identified based on their homology to E. coli MMR proteins (Table 1). These include human homologs of MutS 28, 29, 30, 31, 32, MutL 33, 34, 35, 36, EXO1 37, 38, 39, single-strand DNA-binding protein RPA 27, 40, proliferating cellular nuclear antigen (PCNA) 41, 42, 43, DNA polymerase Ī“ (pol Ī“) 44, and DNA ligase I 45. Although E. coli MutS and MutL proteins are homodimers, human MutS and MutL homologues are heterodimers 32, 34, 46. hMSH2 heterodimerizes with hMSH6 or hMSH3 to form hMutSĪ± or hMutSĪ², respectively, both of which are ATPases that play a critical role in mismatch recognition and initiation of repair 2. hMutSĪ± preferentially recognizes base-base mismatches and ID mispairs of 1 or 2 nucleotides, while hMutSĪ² preferentially recognizes larger ID mispairs. At least 4 human MutL homologs (hMLH1, hMLH3, hPMS1, and hPMS2) have been identified 33, 35, 36, 47. hMLH1 heterodimerizes with hPMS2, hPMS1, or hMLH3 to form hMutLĪ±, hMutLĪ², or hMutLĪ³, respectively 2. hMutLĪ± is required for MMR and hMutLĪ³ plays a role in meiosis, but no specific biological role has been identified for hMutLĪ² 2. hMutLĪ± possesses an ATPase activity and defects in this activity inactivate MMR in human cells. In a reconstituted human MMR system, hMutLĪ± regulates termination of mismatch-provoked excision 45. Recent studies show that MutLĪ± possesses a PCNA/replication factor C (RFC)-dependent endonuclease activity which plays a critical role in 3ā€² nick-directed MMR involving EXO1 16.

Table 1 MMR components and their functions
Full size table

PCNA interacts with MSH2 and MLH1 and is thought to play roles in the initiation and DNA resynthesis steps of MMR 41, 43. PCNA also interacts with MSH6 and MSH3 48, 49, 50, 51 via a conserved PCNA interaction motif termed the PIP box 52. It has been proposed that PCNA may help localize MutSĪ± and MutSĪ² to mispairs in newly replicated DNA 53, 54. Although PCNA is absolutely required during 3ā€² nick-directed MMR, it is not essential during 5ā€² nick-directed MMR 55. This observation might be explained by the fact that EXO1, a 5ā€²ā†’3ā€² exonuclease, is involved in both 5ā€² and 3ā€² directed MMR. Like PCNA, EXO1 also interacts with MSH2 and MLH1 37, 38, 39, 56, 57, 58. While EXO1 can readily carry out 5ā€² directed mismatch excision in the presence of MutSĪ± or MutSĪ² and RPA 45, 59, its role in catalyzing 3ā€² nick-directed excision requires the MutLĪ± endonuclease, which is activated by PCNA and RFC 16, 60. Although it has been suggested that EXO1 possesses a cryptic 3ā€²ā†’5ā€² exonuclease activity 60, 61, current data do not support that hypothesis. Instead, recent studies suggest the following steps for EXO1-catalyzed 3ā€² nick-directed repair: (1) after recognition of the 3ā€² nick and the mismatch, MutLĪ± endonuclease makes an incision 5ā€² to the mismatch in a manner dependent on PCNA and RFC; and (2) EXO1 performs 5ā€²ā†’3ā€² excision from the MutLĪ±-incision site through and beyond the site of the mismatch 16. However, exo1 null mutants in mice and yeast have a weak mutator phenotype 56, 62; thus, it is likely that additional as yet unidentified exonucleases are involved in eukaryotic MMR.

Other protein components involved in human MMR include single-strand DNA-binding protein RPA, RFC, high mobility group box 1 protein (HMGB1), and DNA pol Ī“. RPA seems to be involved in all stages of MMR: it binds to nicked heteroduplex DNA before MutSĪ± and MutLĪ±, stimulates mismatch-provoked excision, protects the ssDNA gapped region generated during excision, and facilitates DNA resynthesis 27, 45, 60, 63. Furthermore, RPA is phosphorylated after pol Ī“ is recruited to the gapped DNA substrate. Recent studies indicate that phosphorylation reduces the affinity of RPA for DNA, that unphosphorylated RPA stimulates mismatch-provoked DNA excision more efficiently than phosphorylated RPA, and that phosphorylated RPA facilitates MMR-associated DNA resynthesis more efficiently than unphosphorylated RPA 63. These results are consistent with the fact that a high-affinity RPA-DNA complex might be required to protect nascent ssDNA and to displace DNA-bound MutSĪ±/MutLĪ± 27, 45, while a lower-affinity RPA-DNA complex might facilitate DNA resynthesis by pol Ī“ 63. HMGB1 is a mismatch-binding protein and has a DNA-unwinding activity 64, 65, 66. It interacts with MSH2 and MSH6 in vitro 67. Recent studies show that HMGB1 can substitute for RPA in an in vitro reconstituted MMR system 45. Additional studies are needed to precisely define the function of HMGB1 in MMR.

Unsolved fundamental problems in MMR

Despite great progress in identifying MMR proteins and genes and application of state-of-the-art biochemical and genetic approaches to analyze the mechanism of MMR in prokaryotic and eukaryotic cells, several key questions about this pathway remain unanswered. One of these questions concerns the mechanism by which MMR proteins facilitate the communication between two physically distant DNA sites: the mismatch and the strand discrimination signal. It is generally agreed that the strand discrimination signal is a strand-specific nick in both prokaryotic and eukaryotic cells (see above), although the source of the nicking activity, at least for the leading strand, is not known in eukaryotic cells. Previous studies have proposed several alternative models for this process, which can be classified into ā€œcis-ā€ or ā€œmovingā€ and ā€œtrans-ā€ or ā€œstationaryā€ models (Figure 1). The ā€œstationaryā€ model (Figure 1, right) proposes that interactions among MMR proteins induce DNA bending or looping that brings the two distant sites together, while MutS (or the MSH heterodimers, i.e. MutSĪ± and MutSĪ²) remains bound at the mismatch 19, 22. In this model, the MutS (or MSH heterodimers) ATPase activity acts in a proofreading role to verify mismatch binding and authorize the downstream excision 22. Support for the stationary model came from the following experiments. Junop et al. showed that recognition of a mismatch by MutS on a DNA molecule activated MutH cleavage of a GATC site located on a separate DNA molecule without a mismatch 22. Consistent with this observation, a second study demonstrated that mismatch-provoked excision could be initiated when a biotin-streptavidin blockade was placed between the mismatch and pre-existing nick 68. The ā€œcisā€ or ā€œmovingā€ models suggest that MutS-MutL (or MutSĪ±/Ī²-MutLĪ±) complexes load at a mismatch site and then move away from the site to search for the strand break, where exonucleases can be recruited to initiate excision.

Figure 1
figure 1

Models for signaling downstream MMR events following mismatch recognition. A schematic diagram for signaling between the mismatch and the strand discrimination signal is shown. Here, a 5ā€² nick is the strand discrimination signal. Similar models apply for 3ā€² nick-directed MMR. The ā€œstationaryā€ or ā€œtransā€ model (right) emphasizes that MutS or its homolog (MSH) proteins remain bound at the mismatch. It is the protein-protein interactions that induce DNA bending or looping that brings the two distant sites together. The two DNA sites can cooperate in a ā€œtransā€ configuration. In two ā€œcisā€ or ā€œmovingā€ models, one called the ā€œtranslocationā€ model (left) and the other called the ā€œmolecular switchā€ or ā€œsliding clampā€ model (middle), it is hypothesized that the MSH proteins bind to the mismatch and then move away from the site to search for the strand discrimination signal. The translocation model suggests that ATP hydrolysis drives unidirectional movement of the MSH proteins, resulting in the formation of an Ī±-like loop. In the molecular switch model (center), binding of an MSH protein (in its ADP-bound state) to the mismatch triggers an ADP to ATP exchange that promotes bi-directional sliding of the protein away from the mismatch, thereby emptying the mismatch site for an incoming MSH protein. Mismatch excision begins when an MSH protein reaches the strand break.

Full size image

There are two moving models, one called the ā€œtranslocationā€ model and the other called the ā€œmolecular switchā€ or ā€œsliding clampā€ model (Figure 1). In the translocation model 69, ATP reduces the mismatch-binding affinity of MutS or the MSH heterodimers, and ATP hydrolysis drives unidirectional translocation of MutS proteins along the DNA helix. DNA is threaded through the protein complex until the latter reaches a strand discrimination signal in either orientation, a process that forms a DNA loop (Figure 1, left). In the molecular switch model, MutS or the MSH heterodimer binds to mismatched DNA in an ADP-bound state. The mismatch binding by MutS or the MSH heterodimer triggers a conformational change that allows an ADP to ATP exchange, which promotes a second conformational change that allows MutS or the MSH heterodimers to form a sliding clamp 70, 71, 72, 73. In this model (Figure 1, middle), it is the binding of ATP, not ATP hydrolysis, that signals downstream events including formation of ternary complex with MutL (or MLH heterodimers) and sliding of the ternary complex from the mismatch to the strand break 70, 71, 72, 73.

Recent studies by Pluciennik and Modrich 74 argue in favor of a moving rather than a stationary mechanism, because their data demonstrate that a dsDNA break 75 or a protein ā€œroadblockā€ between the mismatch and the nick inhibits in vitro MMR with recombinant E. coli proteins. It is not clear why two ā€œroadblockā€ experiments 68, 74 obtained distinct results. In a reconstituted human MMR reaction, Zhang et al. 45 show that multiple MutSĪ±-MutLĪ± complexes are essential for processing a single mismatch, providing evidence to support the molecular switch model. Additional studies are needed to address these unresolved questions about the molecular mechanism of MMR.

MMR mediates DNA damage signaling

MMR deficiency and drug resistance

DNA-damaging agents such as the alkylating agents N-methyl-Nā€²-nitro-N-nitrosoguanidine (MNNG), temozolomide, or procarbazine are cytotoxic agents that kill most of the replicating cells. Many cancer therapeutics are genotoxic and cytotoxic agents that induce apoptotic cell death. Interestingly, many cells that acquire resistance to such agents are deficient in MMR. For example, the human lymphoblastoid cell line MTI, which has a defect in hMSH6, was derived by culturing TK6 cells in the presence of a high concentration of MNNG. The resulting MNNG-resistant MT1 cells are defective in strand-specific MMR 76. Many human colorectal cancer cell lines are also resistant to alkylating agents and have associated defects in MMR. The causal relationship between drug resistance and MMR is demonstrated by the fact that hMLH1-defective MNNG-resistant cells lose drug resistance when the hMLH1 defect is genetically complemented with wild-type hMLH1 on chromosome 3 77. It has also been observed that defects in MSH2 and PMS2 confer resistance to alkylating agents (reviewed in 78). The mechanism by which MMR influences drug cytotoxicity is discussed further below.

Resistance to methotrexate (MTX) has also been associated with phenotypic changes in MMR in human cells. This occurs by the unusual mechanism of co-amplification of the human chromosomal region that encodes dihydrofolate reductase (DHFR, the target of MTX) and hMSH3 79, 80. Amplification of DHFR lowers sensitivity to MTX by overexpressing the target of the drug. However, overexpression of hMSH3 sequesters hMSH2 in the hMutSĪ² heterodimer, effectively preventing formation of the hMutSĪ± (hMSH2/hMSH6) heterodimer, which leads to degradation of uncomplexed hMSH6, significant dysregulation of MMR and hypermutability 81, 82. Overexpression of DHFR combined with genome-wide hypermutability and defective MMR are likely responsible for the MTX resistance of HL60 and other tumor cells.

MMR proteins promote DNA damage-induced cell cycle arrest and apoptosis

Cell cycle arrest is an important mechanism for preventing DNA damage-induced genomic instability. A large number of studies have characterized the so-called G2 or S phase checkpoints, and identified proteins required for cell cycle arrest, including ATM, ATR, p53, p73, Chk1, and Chk2. However, it was a somewhat unexpected finding that hMutSĪ±- and hMutLĪ±-deficient cells are defective in cell cycle arrest in response to multiple types of DNA damaging agents 6, 7, 83. While the molecular basis of this effect is not precisely known, it has been reported that MMR-deficient cells fail to phosphorylate p53 and p73 in response to DNA damage 84, 85. This implicates ATM, ATR, and/or c-Abl, because these kinases phosphorylate p53 and p73 during the response to DNA damage 85, 86. In support of this, it has been reported that hMutSĪ± and hMutLĪ± interact physically with ATM, ATR-ARTIP, c-Abl, and p73 in cells treated with DNA damaging agents/drugs 83, 87, 88, 89. These observations implicate hMutSĪ± and hMutLĪ± in a signaling cascade that leads from DNA damage to cell cycle arrest and/or apoptosis. They also at least in part explain the fact that drug-induced cytotoxicity is lost in MMR-deficient cells, as discussed above 6. Very recently, EXO1 has been shown to be essential for upstream induction of DNA damage response, possibly by reducing ssDNA formation and recruiting RPA and ATR to the damage site 90. It remains to be seen if MutSĪ± and/or MutLĪ± act to recruit EXO1 in DNA damage response as they do in MMR.

Two models have been proposed to describe the role of MMR in DNA damage signaling. The ā€œfutile DNA repair cycleā€ model (Figure 2, left) proposes that strand-specific MMR, which targets only newly replicated DNA, engages in a futile DNA repair cycle when it encounters DNA lesions in the template strand, and this futile cycling activates DNA damage signaling pathways to induce cell cycle arrest and apoptosis 6. Support for this model came from both in vivo and in vitro experiments. Stojic et al. 86 showed that exposure to MNNG induces DNA breaks/gaps, cell cycle arrest, and persistent nuclear foci at sites of DNA damage. The DNA damage-associated repair foci contain both damage signaling and DNA repair proteins, including ATR, Ī³-H2AX, and RPA. York and Modrich 91 showed that nicked circular heteroduplex plasmid DNA containing a single O6-methylguanine (O6-me-G)-thymine (T) mispair cannot be repaired by the MMR system when the lesion (O6-me-G) and the nick are on opposite strands; this suggests a futile repair process. An alternative model, referred to as the direct signaling model (Figure 2, right), argues that hMutSĪ±/hMutLĪ± directly trigger DNA damage signaling by recruiting ATM or ATR/ARTIP to the lesion, which activates a checkpoint response. This model is supported by an elegant study from the Hsieh laboratory showing that ATR and ATRIP form a complex with MutSĪ±/MutLĪ± in the presence of O6-me-G/T, which activates the ATR kinase and phosphorylates Chk1 89. Because mammalian MMR proteins interact with a broad spectrum of DNA lesions 6, this model is consistent with the notion that MutSĪ±/MutLĪ± acts as a sensor for DNA damage in mammalian cells. Both models provide a reasonable explanation for decreased DNA damage-induced apoptotic signaling and increased drug resistance in MMR-deficient cells.

Figure 2
figure 2

Models for MMR-dependent DNA damage signaling. The ā€œfutile DNA repair cycleā€ model (left) suggests that DNA adducts (solid black circle) induce misincorporation, which triggers the strand-specific MMR reaction. Since MMR only targets the newly synthesized strand for repair, the offending adduct in the template strand cannot be removed, and will provoke a new cycle of MMR upon repair resynthesis. Such a futile repair cycle persists and activates the ATR and/or ATM damage signaling network to promote cell cycle arrest and/or programmed cell death. The direct signaling model proposes that recognition of DNA adducts by MSH-MLH complexes allows the proteins to recruit ATR and/or ATM to the site, activating the downstream damage signaling.

Full size image

Role of MMR in other DNA metabolic pathways

MMR proteins have also been implicated in homeologous recombination, immunoglobulin class switching, somatic hypermutation, interstrand-crosslink repair, and trinucleotide repeat (TNR) expansion. Homeologous recombination (recombination between related but non-identical DNA sequences) generates mispairs/heteroduplexes, and induces genomic instability via chromosomal translocations, deletions, or inversions 92, 93. The frequency of homeologous recombination is much lower than that of homologous recombination in normal cells, but the frequency of homeologous recombination is dramatically elevated in MMR-deficient cells, suggesting that MMR suppresses homeologous recombination 3, 94. MutS and MutL inhibit DNA strand exchange between divergent sequences in vitro, most likely by binding to the mismatches generated during strand exchange 95. Recent studies in yeast reveal that suppression of homeologous recombination is mediated by MutSĪ± and a RecQ family helicase, SGS1 96, 97. Consistent with this notion, yeast strains defective in sgs1 fail to suppress homeologous recombination 97, 98. It has been postulated that MutSĪ± recruits SGS1 to DNA mismatches, where it unwinds the heteroduplex and blocks homeologous recombination 96. Although suppression of homeologous recombination by MMR proteins in human cells is less well understood, two human SGS1 homologous proteins, BLM and RECQ1, interact with MutSĪ± 99, 100. This suggests that a similar mechanism is used to suppress homeologous recombination in yeast and human cells.

Recent studies also implicate MMR in repair of inter-strand crosslinks (ICLs), in a process that involves protein components from homologous recombination, double-strand break repair, and nucleotide excision repair 101, 102, 103. The precise nature of this involvement is not yet clear, and the specific MMR proteins that participate remain somewhat controversial. However, hMutSĪ² appears to directly bind ICLs in vitro 102, and hMutLĪ± interacts specifically with the helicase domain of Fanconi Anemia protein FANC-J to facilitate ICL repair 101.

The studies discussed above suggest that MMR promotes genomic stability. However, during immunoglobulin class switching and somatic hypermutation, it appears that MMR proteins play a highly specialized role in promoting genetic variation. Immunoglobulin class switching and somatic hypermutation are mechanisms for increasing antibody diversity during antigen-stimulated B-cell differentiation. During this process, activation-induced cytidine deaminase (AID) deaminates cytosine residues to uracil, generating G:U mispairs, which can be recognized and processed by MMR 104, 105. However, during repair resynthesis, the high-fidelity replicative pol Ī“ and ɛ are thought to be replaced by the translesion polymerases, which are error-prone and crucially introduce base substitutions and frameshift mutations 106. Additionally, MMR proteins play an important role in class switch recombination, an event where the IgM constant region is substituted by downstream constant sequences. In this capacity, MMR proteins utilize strand breaks generated by uracil DNA glycosylase to repair the AID-induced G:U mispairs in a strand-indiscriminate manner, leading to double-strand DNA breaks. It is these breaks that stimulate class switch recombination 107. B cells from mice deficient in MMR genes (MSH2, MSH6, MLH1, PMS2, or EXO1) display a low level of somatic hypermutation and reduced class switch recombination 8, 62, 108, 109.

MMR proteins also promote TNR expansion, a phenomenon associated with a number of neurological disorders in humans, including Huntington's disease, myotonic dystrophy, and fragile X syndrome 110 (also see Kovtun and McMurray in this issue). TNRs such as (CAG)n form single-stranded DNA loop/hairpin structures in vitro 111. Surprisingly, transgenic CAG repeats undergo expansion in wild-type, but not in knockout mice defective in MMR genes MSH2 and MSH3 112, 113, suggesting that the expansion mutations of the transgenic CAG repeats require functional MMR proteins MSH2 and MSH3. In vitro biochemical studies indeed show that MutSĪ² (the MSH2-MSH3 heterodimer) specifically binds to the (CAG)n hairpin structure 113. One model proposes that MutSĪ² inhibits repair or resolution of the (CAG)n hairpin, thus stimulating (CAG)n expansion 113. However, many of these studies were conducted using transgenic mouse models for TNR expansion. Thus, the role of MMR in human neurological diseases involving TNR expansion is at present unclear, and additional genetic and biochemical studies are needed to define the mechanism of TNR expansion in human cells.

MMR deficiency leads to cancer development

MMR defects in hereditary non-polyposis colorectal cancer (HNPCC) and other cancers

In the early 1990s, it was shown that HNPCC and some cases of sporadic colon cancer are caused by defects in human MMR 3. For HNPCC, Kolodner and co-workers, and Vogelstein and co-workers independently identified germ-line mutations in hMSH2 at chromosome 2p16-p21 in HNPCC families 29, 30. Genetic analyses of HNPCC kindreds revealed a large increase in frequency of insertion and deletion mutations in simple repeat (microsatellite) sequences, a phenomenon known as microsatellite instability (MSI) 114. MSI was also observed at lower incidence in sporadic colon cancers 114, 115, 116. Additional studies showed 100- to 700-fold decreased stability of poly(GT) tracts in yeast strains with single or double knockouts in MSH2, MLH1, or PMS1 117. Biochemical studies by Modrich and co-workers 118 and Kunkel and co-workers 119 also showed that extracts of MSI-positive tumor cells were severely defective in repair of base-base and ID mispairs. Thus, genetic and biochemical evidence converged to support the hypothesis that defective MMR plays a causal role in carcinogenesis leading to HNPCC, and strongly implicated such defects in some sporadic human cancers such as colorectal cancer.

The second locus linked to HNPCC was hMLH1 at 3p21-p23 33, 36, 120. Furthermore, HNPCC has also been linked to mutations in two additional MutL homolog genes hPMS1 and hPMS2, on 2q and 7p, respectively 35. Defects in hMLH1 represent the majority of all HNPCC cases 35, 36, with mutations in hMSH2 accounting for a large fraction of all the remaining HNPCC cases for which a genetic defect has been identified. In contrast, germ-line mutations in hMSH3 have not yet been linked to HNPCC, and mutations in hPMS1, hPMS2, and hMSH6 are relatively rare in HNPCC patients 1, 2, 3. These observations are consistent with the fact that hMSH2 and hMLH1 are essential MMR components, while hPMS1, hPMS2, hMSH6, and hMSH3 play important but partially redundant and/or dispensable roles in MMR.

Genetic evidence described above shows that defects in MMR correlate with HNPCC, and biochemical studies provide compelling additional evidence for this hypothesis. In particular, cell lines from HNPCC patients and sporadic tumors with MSI are defective in strand-specific MMR, and these cells can be divided into at least two complementation groups, corresponding to hMutLĪ± 34 and hMutSĪ± 32. Importantly, purified hMutLĪ± or hMutSĪ± specifically complements the biochemical defect in these cells. Similar evidence was obtained from chromosome or gene transfer experiments: expressions of the exogenous MMR genes complement the biochemical defect and stabilize simple repetitive sequences in the transfected cells (reviewed in 121).

These genetic and biochemical complementation studies essentially prove the causal role of MMR defects in HNPCC. However, many questions remain, including whether or not mutations in hMLH3 or hEXOI also cause HNPCC, a possibility that remains controversial 47, 122, 123. Mice defective in EXOI exhibit a mild defect in genomic stability, are partially defective in strand-specific MMR, and have increased rates of some cancers 62. Thus, it seems possible that germ-line mutations in hEXOI could potentially have similar phenotypic effects in humans. Additional studies are needed to clarify this point.

Although MSI was first correlated with MMR defects in tumors from HNPCC patients, MSI is also associated with a wide variety of non-HNPCC and non-colonic tumors (reviewed in Ref. 124). These include endometrial, ovarian, gastric, cervical, breast, skin, lung, prostate, and bladder tumors as well as glioma, leukemia, and lymphoma. Biochemical studies confirmed that MSI cell lines from sporadic leukemia, endometrial, ovarian, prostate, and bladder cancers are defective in strand-specific MMR 32, 125, 126. Interestingly, MSI in sporadic non-colonic tumors is often associated with hypermethylation of the promoter of hMLH1 (see below for details), and few mutations in MMR genes have been identified in these cells. These findings suggest that MMR defects are a likely cause of MSI in non-colonic sporadic cancers, although other mechanisms may also be involved in causing the MSI mutator phenotype.

Mouse models for MMR demonstrate roles in cancer and meiosis

Knockout mouse models have been developed for MSH2, MSH3, MSH6, MLH1, MLH3, PMS1, PMS2, and EXO1 (reviewed in 8, 62, 127, 128) and their phenotypes have been somewhat informative. Most of the knockout mice have a mutator phenotype, are MSI-positive, and are cancer-prone. However, the primary cancer susceptibility of MSH2, MLH1, and PMS2 knockout mice is lymphoma, not colorectal cancer as in humans, and secondary cancer susceptibilities are to gastrointestinal tumors, skin neoplasms, and/or sarcomas (Table 2).

Table 2 Phenotypes of MMR-deficient knockout mice
Full size table

MSH2āˆ’/āˆ’ deficient mice are fertile 129, 130, are MSI-positive, develop lymphoma within 1 year of age, and have a significantly shorter lifespan than wild-type mice (i.e., 50% mortality by 6 months of age). The phenotype of MSH6āˆ’/āˆ’ deficient mice is similar to that of MSH2āˆ’/āˆ’ deficient mice, but lacking MSI 129, 131, a phenotype resembling that of atypical HNPCC with an hMSH6 defect as the tumors in these MSH6-defective individuals have longer latency and low MSI 132. Cells from MSH3āˆ’/āˆ’ mice are defective in repair of ID mispairs but can repair base-base mismatches. MSH3āˆ’/āˆ’ mice are either tumor free 133 or develop tumors at a very late age 134, essentially consistent with the fact that no MSH3 mutations have been identified in HNPCC patients. However, in MSH3āˆ’/āˆ’ and MSH6āˆ’/āˆ’ double deficient mice, the tumor predisposition phenotype is indistinguishable from MSH2āˆ’/āˆ’ or MLH1āˆ’/āˆ’ mice 133, 134, suggesting that MSH3 cooperates with MSH6 in tumor suppression.

Sterility is a characteristic feature of MLH mutant mice (except PMS1) 135, 136, 137. These animals are also susceptible to cancer and display genomic instability, reflecting defective MMR. However, male and female MLH1 and MLH3 knockout mice 135, 136, 137, and male PMS2 knockout mice are completely sterile 138. PMS1 knockout mice are exceptional, because they are fertile, they lack cancer susceptibility, and, apart from a very small increase in mutations in mononucleotide repeats, they appear to be MSI-negative 127. EXO1 defective mice are also sterile 62. It is clear that the loss of fertility in these knockout mice is caused by abnormal meiosis 62, 128, 135, 136, 137.

The characteristics of all MMR knockout mouse models are summarized in Table 2. These data strongly support the ideas that MMR is a basic genome surveillance mechanism and that defects in MMR can promote cancer development. The effects of MMR defects on carcinogenesis appear to be tissue- and species-specific, in a manner that is poorly understood. The effects of MMR defects on meiosis in humans remain poorly characterized. However, MMR clearly plays a critical role during meiosis and/or gamete formation in mice.

Epigenetic silencing of MMR gene expression leads to cancers

Mutations in MMR genes cause genomic instability and MSI in HNPCC and in a subset of sporadic colorectal cancers. However, in a significant fraction of MSI-positive sporadic colon tumors that have an MMR defect, mutations have not been identified in MMR genes. Epigenetic silencing of hMLH1 via promoter hypermethylation strongly down-regulates MMR in many of these cases 139, 140. In contrast, hypermethylation of the hMSH2 gene is rarely observed in tumors with MSI. In fact, it has been reported that more than 95% of MSI-H sporadic tumors demonstrate mutation and/or epigenetic silencing of hMLH1 141. While most studies demonstrate epigenetic silencing of the hMLH1 promoter in sporadic tumors, hypermethylation of the hMLH1 promoter was also recently demonstrated in an HNPCC patient who does not have a germ-line mutation in any MMR gene 142. Interestingly, recent studies suggest that this effect may be heritable 143, 144, 145. Direct evidence that hMLH1 promoter hypermethylation down-regulates hMLH1 gene expression was obtained by treating cells with 5-aza-deoxycytidine, which reversed promoter hypermethylation, and restored hMLH1 gene expression and normal MMR capacity 146, 147.

MMR deficiency and mutations in coding repeat sequences

Previous studies demonstrate that defects in MMR increase the mutation rate in genes containing a simple repeat sequence in coding regions, often referred to as target genes 148. Thus, defects in MMR confer a mutator phenotype. It is presumed that such a mutator phenotype has genome-wide consequences and could increase the frequency of additional genome-destabilizing and cancer-promoting mutations; however, this is a difficult hypothesis to test experimentally. One approach is to selectively analyze the stability of di- and tri-nucleotide tracts within coding regions. For example, Markowitz et al. 149 reported two mutation ā€œhotspotsā€ in the type II transforming growth factor-Ī² receptor (TGF-Ī² RII) gene in MMR-deficient tumor cells from a patient with sporadic colorectal cancer. One of these mutational hotspots fell within a 6-bp GT dinucleotide repeat and the other fell within an (A)10 mononucleotide repeat 149. Both of these hotspots were sites of frequent frameshift mutations that truncated the TGF-Ī² RII gene product. Similar observations have been made in other colorectal tumor cells and many other MSI-positive tumor cells including glioma, gastric, uterine, cervical and squamous head and neck tumors, as well as ulcerative colitis-associated cancer and cecum cancer. Furthermore, in some of these tumor cells, somatic frameshift mutations have been documented in many other genes including Bax, insulin-like growth factor 2 receptor (IGF2-R), transcription factor E2F-4, APC, PTEN, hMSH3, hMSH6, Mre11, MBD4/MED, ACTRII, AIM2, APAF-1, AXIN-2, BCL-10, BLM, Caspase-5, CDX-2, CHK-1, FAS, GRB-14, cell cycle protein hG4-1, KIAA0977, ubiquinone oxidoreductase gene NADH, OGT, Rad50, RHAMM, RIZ, SEC63, SLC23AT, TCF-4, and WISP-3 (reviewed in Ref. 150). These data are consistent with the idea that similar mutations occur on a genome-wide basis and at a much higher rate in MMR-deficient cells than in wild-type cells. Because the genes noted above play critical roles in regulating cell growth or genomic stability, loss-of-function mutations in these genes may be crucial steps in the multi-step pathway of carcinogenesis.

Conclusion and perspectives

The discovery that defects in MMR play a causal role in HNPCC and many MSI-positive sporadic cancers brought immediate clinical relevance to research in the field of eukaryotic MMR. This discovery led to intensive research on and better understanding of the biological roles of MMR in eukaryotic cells, which relate to cancer prevention and therapy. Although the primary role of MMR is to improve replication fidelity by correcting replication-associated base-base mismatches or ID mispairs, important secondary roles are to modulate DNA recombination and facilitate DNA damage signaling. Thus, it is abundantly clear that defects in MMR are ā€œpermissiveā€ for carcinogenesis.

The identification that MMR-deficient cells are resistant to certain chemotherapeutic drugs such as temozolomide, procarbazine, or cisplatin has significant impacts on cancer treatments, especially for patients with tumors defective in MMR. It is also known that MMR deficiency can be acquired during chemotherapy by selective mutations in MMR genes 6. Therefore, the risks for chemotherapy are two-fold. First, for patients with MMR-deficient tumors, chemotherapeutic treatments could selectively kill patients' MMR-proficient cells (e.g., blood cells) that undergo proliferation, thereby leading to rapid deaths of the cancer patients. Second, if a patient's tumor is not caused by loss of MMR function, chemotherapeutic treatments may kill the tumor cells; at the same time, the treatment may induce or select for a mutation in MMR genes, which could lead to a secondary cancer. Therefore, novel chemotherapeutic or alternative approaches are needed for cancer patients with or without MSI-positive tumors. Such approaches might include targeted gene therapy, which could selectively restore drug sensitivity in tumor cells defective in MMR or treatment with agents that stimulate apoptosis downstream of MMR in tumor cells. Additional research on mechanisms that selectively kill MMR-deficient cells is also warranted. Such efforts should also include more basic research on the molecular mechanisms of eukaryotic MMR. Understanding these mechanisms will support efforts for developing new therapeutic approaches for patients with HNPCC or other MSI-positive MMR-deficient tumors.