DNA-damage repair; the good, the bad, and the ugly

Direct reversal of DNA damage

In contrast to other DNA-damage repair pathways, direct reversal of DNA damage is not a multistep process and does not involve multiple proteins (Sedgwick et al, 2007). Furthermore, unlike excision repair, direct reversal of DNA damage does not require the excision of the damaged bases. An example of a DNA lesion that is repaired by direct reversal is the O⁶-alkylguanine. Alkylating agents can transfer methyl or ethyl groups to a guanine, thereby modifying the base and interfering with its pairing with cytosine during DNA replication. The cytotoxic and mutagenic O⁶ alkyl adduct in DNA is repaired by direct reversal, which is mediated by the enzyme Ada in Escherichia coli (E. coli) and the mammalian O⁶-methylguanine-DNA methyltransferase (MGMT). MGMT, also known as AGT, removes the DNA adducts by transferring the alkyl group from the oxygen in the DNA to a cysteine residue in its active site. This reaction leads to the reversal of the base damage; however, the alkylation of MGMT leads to its inactivation and subsequent ubiquitination and proteosomal degradation. MGMT has attracted a great deal of attention, as certain anticancer chemotherapeutic drugs produce O⁶-alkylguanine, further supporting its role in modulating the therapeutic response of tumors to these drugs. Mouse models for Mgmt inactivation have been generated (Tsuzuki et al, 1996b; Glassner et al, 1999). These mutants were viable and showed no increase in spontaneous tumorigenesis (Table I). However, Mgmt homozygous mice and cells were highly sensitive to chemotherapeutic alkylating agents such as methylnitrosourea. Mgmt homozygous mutant females, but not males, developed larger numbers of dimethylnitrosamine-induced liver and lung tumors compared with controls (Iwakuma et al, 1997). Additionally, transgenic mice over-expressing human MGMT or E. coli Ada have also been generated. In response to alkylating carcinogens that produce O⁶-alkylguanine in DNA, these transgenic mice demonstrated a significantly reduced susceptibility to developing cancers, including thymomas (Dumenco et al, 1993), liver tumors (Nakatsuru et al, 1993), and skin tumors (Becker et al, 1997).

AlkB is another enzyme that mediates direct DNA damage reversal in E. coli. This dioxygenase is involved in the repair of alkylation damage, particularly 1-methyladenine (1meA) and 3-methylcytosine (3meC). Two mammalian AlkB homologues, ABH2 and ABH3, have been shown to possess DNA-repair functions similar to the bacterial AlkB (Duncan et al, 2002; Sedgwick et al, 2007). Similar to AlkB, ABH2 and ABH3 have the ability to repair 1meA and 3meC residues. However, whereas ABH2 prefers double-stranded DNA, ABH3 and AlkB favour single-stranded DNA and RNA (Aas et al, 2003; Falnes et al, 2004). Further insight into the function of the mammalian ABH2 and ABH3 came from studies of mice carrying targeted mutations of these genes. Mice deficient in Abh2, Abh3, or both, were viable (Ringvoll et al, 2006). Abh2^−/−, but not Abh3^−/−, mice showed age-dependent accumulation of 1meA in their genomic DNA. As in AlkB mutants in E. coli, mouse embryonic fibroblasts (MEFs) deficient in Abh2 were hypersensitive to methyl methane-sulfonate (MMS) treatment. However, mice deficient in Abh2 or Abh3 did not show increased spontaneous cancer development (Table I). Further studies are required to assess the role of these dioxygenases, and other AlkB homologues, in alkylation damage-induced cancer.

These examples of direct DNA-damage reversal mediated by MGMT/Ada or ABH/AlkB demonstrate the conserved role of this mechanism in DNA repair. In addition, increased tumorigenesis of Mgmt mutants, together with the resistance of MGMT transgenic mice to alkylating carcinogens that produce O⁶-alkylguanine, further demonstrate the important role that direct reversal plays in cancer.

The MMR pathway

The MMR pathway plays an important role in both prokaryotes and eukaryotes in repairing mismatches, which are small insertions and deletions that take place during DNA replication (Figure 1; Jiricny, 2006). Failure of MMR commonly results in microsatellite instability (MSI). Several homologues of the bacterial MMR genes MutS and MutL have been identified in yeast and mammals.

The importance of the MMR pathway became evident upon identification of mutations in certain human MMR genes in hereditary non-polyposis colorectal cancer (HNPCC), a highly penetrant autosomal dominant cancer syndrome (Figure 2; Vasen et al, 2007). HNPCC, also known as Lynch syndrome, is characterized by early-onset colorectal cancer, with elevated levels of MSI in the tumors. Individuals with HNPCC have an approximate 80% lifetime risk for colorectal cancer, and are also predisposed to the development of endometrial, ovarian, gastric, and other types of malignancies.

Open in a separate window

Figure 2

Examples of human syndromes and disorders associated with defective DNA-damage repair. Impaired MMR pathway leads to the hereditary HNPCC. Mutations of certain human NER genes have been associated with syndromes and disorders including the XP, CS, and TTD. MAP, a rare disorder, has been shown associated with mutations of the BER gene MUTYH. Various human syndromes and disorders have been associated with defects of the HR pathway. They include ATLD, NBS, BS, WS and RTS. Mutations of certain human genes involved in NHEJ lead to the SCID or RS-SCID.

Approximately 70–80% of germline mutations identified in HNPCC families are mutations in MLH1 or MSH2, whereas mutations in MSH6 are found in approximately 10% of HNPCC families (Peltomaki and Vasen, 1997). Germline mutations in other human MMR genes, including PMS1, PMS2, MLH3, and exonuclease 1 (EXO1), have also been found in HNPCC families; however, they occur at a much lower frequency (Vasen et al, 2007). In addition, inactivation of MLH1 by mutations at the promoter or coding sequences, or by promoter methylation, has been identified in sporadic colorectal tumors (Kane et al, 1997; Veigl et al, 1998). Recent studies, although very limited, have identified rare patients with homozygous germline mutations for MLH1, MSH2, MSH6, or PMS2 (Felton et al, 2007). Typically, these individuals have a reduced life span and, in contrast to heterozygous MMR individuals, tend to develop juvenile haematological malignancies and brain cancer.

In yeast, Msh2 forms heterodimers with Msh3 and Msh6, proteins that bind DNA mismatches and initiate the MMR process. In Saccharomyces cerevisiae (S. cerevisiae), Msh2, Msh3, and Msh6 mutants are viable (Marsischky et al, 1996). Both Msh2 and Msh6 S. cerevisiae mutants show high frequencies of base substitution, whereas only Msh2 mutants exhibit high frameshift mutations. Msh3 mutations in S. cerevisiae result in low rates of frameshift mutations. However, on Msh6-mutant background, synergistic effects of the dual mutations have been observed, including increased MSI and mutability similar to Msh2 mutants.

Mutant mice for MutS and MutL MMR homologues have also been generated using gene targeting (Table II). Mutant mice for the MutS homologues include Msh2, Msh3, Msh4, Msh5, and Msh6. Mice carrying homozygous mutations for Msh4 or Msh5 did not exhibit cancer phenotypes; however, males and females were infertile, consistent with the role of MutS homologues in processing meiotic recombination intermediates (de Vries et al, 1999; Edelmann et al, 1999). In contrast, homozygous mutants for Msh2 (de Wind et al, 1995; Reitmair et al, 1995), Msh3 (Edelmann et al, 2000), and Msh6 (Edelmann et al, 1997) have increased risk for developing cancers such as lymphoma, gastrointestinal, and skin cancer.

Mutant mice for MutL homologues include Pms1^−/−, Pms2^−/−, Mlh1^−/−, and Mlh3^−/− mice (Table II). These mutants are viable; however, males and females deficient in Mlh1 (Baker et al, 1996; Edelmann et al, 1996) or Mlh3 (Lipkin et al, 2002) and males deficient in Pms2 (Baker et al, 1995) are sterile, demonstrating a requirement for these proteins during meiosis. In addition, mouse MutL homologues are differentially required for cancer suppression. Pms1^−/− mice do not show any increased risk for cancer (Prolla et al, 1998), whereas Mlh1^−/− (Prolla et al, 1998; Chen et al, 2005) and Mlh3^−/− mice (Chen et al, 2005) are predisposed to developing lymphomas and gastrointestinal tumors. Similarly, Pms2-null mutants (Prolla et al, 1998; Chen et al, 2005) are prone for lymphoma development.

EXOI physically interacts with MSH2, MSH3, and MLH1, and is involved in the excision of mismatched bases in DNA (Tishkoff et al, 1997; Schmutte et al, 2001). Mutant mice for ExoI have impaired MMR, accumulate MSI, and exhibit a greater risk for developing lymphomas (Wei et al, 2003). These mutants also have meiotic defects and are sterile, demonstrating the requirement of ExoI in meiosis.

Double mutant mice carrying dual mutations of different MMR genes have also been reported. For example, Msh3-mutant mice develop cancer with low frequency and at a later age, whereas Msh3^−/−Msh6^−/− mice (Edelmann et al, 2000) die prematurely and develop tumors including lymphomas, gastrointestinal, and skin tumors. This phenotypic outcome is similar to that of Msh2^−/− or Mlh1^−/− mice that are the most cancer-prone MMR mutants, as half of these mutants die around 6 months of age. This cooperation between mutations of Msh2 and Msh6 in mice is reminiscent of their collaboration in the maintenance of genomic integrity of S. cerevisiae.

Immunoglobulin (Ig) diversification, an essential process for immunity, involves somatic hypermutation (SHM) of the Ig genes, as well as VDJ recombination and class-switch recombination (CSR), two processes mediated by NHEJ (Maizels, 2005). Interestingly, studies of the various MMR-mutant strains have implicated a role for this pathway in SHM and CSR. Thus, Msh2, Msh6, and ExoI, but not Msh3-mutant mice, have reduced CSR and SHM (Rada et al, 1998; Wiesendanger et al, 2000; Bardwell et al, 2004; Li et al, 2004).

Whereas most HNPCC human individuals carry heterozygous germline mutations of MMR genes, which predisposes them to cancer, mice heterozygous for MMR mutations do not appear to have an increased risk for developing cancer. This difference is not specific for MMR mutations, as heterozygous mutations in certain genes involved in other DNA-damage repair pathways are also able to predispose humans, but not mice, to cancer. The reasons for these differences remain unknown, although species differences in the DNA-damage repair pathways, metabolism, or life span could contribute to these observed human–mouse discrepancies. Despite these differences, mouse models have significantly improved our understanding of the MMR and other repair mechanisms, and their roles in preserving genomic integrity and suppressing cancer.

The NER pathway

The NER pathway is a multistep process that serves to repair a variety of DNA damage, including DNA lesions caused by UV radiation, mutagenic chemicals, or chemotherapeutic drugs that form bulky DNA adducts (Figure 1; Leibeling et al, 2006). Over 30 different proteins are involved in the mammalian NER, whereas only three proteins (UvrA, UvrB, and UvrC) are required by prokaryotes (Truglio et al, 2006).

Two NER sub-pathways that have been identified are as follows: the global genome NER (GG-NER) that detects and removes lesions throughout the genome, and the transcription-coupled NER (TC-NER), which repairs actively transcribed genes. NER begins with the recognition of the DNA lesion, followed by incisions at sites flanking the DNA lesion, and culminates in the removal of the oligonucleotide containing the DNA lesion. Ligation of a newly synthesized oligonucleotide, complementary to the pre-existing strand, serves to fill the gap, thus ending the NER process. GG-NER and TC-NER involve several common proteins and proceed through the same repair steps, except during recognition of the DNA lesion. In GG-NER this recognition involves the XPC–RAD23B and DDB1–DDB2/XPE proteins, whereas recognition in TC-NER is mediated by cockayne syndrome group A (CSA) (ERCC8) and CSB (ERCC6). NER has attracted a great deal of attention due to its role in three rare human syndromes characterized by increased cancer frequencies, neurodegeneration and ageing (Figure 2). These syndromes are xeroderma pigmentosum (XP), Cockayne syndrome (CS), and trichothiodystrophy (TTD) (Thoms et al, 2007).

XP individuals show extremely severe skin sensitivity to short intervals of sun exposure and most develop freckles at an early age. In addition, XP individuals may exhibit eye damage as they suffer chronic UV-induced conjunctivitis and keratitis as a consequence of continual sun exposure. XP individuals have greater than 1000-fold increased skin cancer risk, which first appears at an average age of 10 years. Approximately 20% of XP individuals also develop neurological abnormalities. XP is caused by mutations of the NER gene XPA, XPB (ERCC3), XPC, XPD (ERCC2), XPE (DDB2), or XPF; XPG. Whereas XP individuals carrying mutations of XPC or XPE (DDB2) are only deficient in GG-NER, the remaining XP individuals are deficient in both GG-NER and TC-NER (Thoms et al, 2007).

CS is a very rare human autosomal recessive inherited genetic disease (Thoms et al, 2007). Similar to XP individuals, CS individuals suffer excessive sun sensitivity, but without increased predisposition for skin cancer. Common CS symptoms include growth retardation (dwarfism), progressive cognitive impairment, and ophthalmologic disorders such as cataracts or retinitis pigmentosa. CS individuals typically die in the first or second decade of life. CS is caused by mutations of either CSA or CSB, two proteins essential for DNA-damage recognition and initiation of TC-NER. Therefore, although CS individuals are deficient in TC-NER, they remain proficient in GG-NER.

TTD is a rare human autosomal recessive disorder associated with defective NER; the most severe cases are associated with mutations in the XPB or XPD genes. Clinical characteristics of TTD include brittle hair and nails, dwarfism, and ataxia (Thoms et al, 2007). In addition, half of TTD individuals exhibit sensitivity to sunlight; however, skin cancer predisposition has not been linked to this syndrome.

Several mouse models for NER mutations have been generated (Table III). Homozygous mutants for XP genes are viable (Nakane et al, 1995; Sands et al, 1995; Harada et al, 1999; Itoh et al, 2004; Tian et al, 2004a, 2004b; Yoon et al, 2005), with the exception of the pre-implantation embryonic lethality of Xpd mutants (de Boer et al, 1998b). Xpd^R722W-mutant mice carrying an amino-acid substitution that mimics a human XPD allele associated with TTD have been generated (de Boer et al, 1998a). Similarly, Xpd^G602D-mutant mice carrying a substitution at amino acid 602 have been generated to mimic the human combined XP/CS (Andressoo et al, 2006). Both Xpd^R722W and Xpd^G602D mutants have been proven viable and reproduced some of the characteristics of individuals that carry these XPD mutations.

With the exception of Xpe (Ddb2) mutants (Itoh et al, 2004; Yoon et al, 2005), homozygous mutant mouse cells for Xpa (Nakane et al, 1995), Xpc (Sands et al, 1995), Xpf (Tian et al, 2004b), Xpg (Harada et al, 1999; Tian et al, 2004a), Xpd^R722W (de Boer et al, 1998a), or Xpd^G602D (Andressoo et al, 2006) were all UV sensitive, correlating to the results obtained from human mutant cells. Increased predisposition for UV-induced skin cancer was observed with Xpa, Xpc, Xpd^R722W, Xpd^G602D, and Xpe (Ddb2)-mutant mice (Nakane et al, 1995; Sands et al, 1995; de Boer et al, 1999; Itoh et al, 2004; Yoon et al, 2005; Andressoo et al, 2006). Thus, as seen in humans, XP proteins play a major role in murine NER.

DDB1 and HR23B are two proteins involved with XPC in the recognition of DNA damage and the initiation of GG-NER. However, in contrast to Xpc^−/− mice (Sands et al, 1995), Ddb1 mutants die during early embryonic development (Cang et al, 2006). Inactivation of Ddb1 in developing CNS and lens resulted in massive p53-dependent apoptosis of dividing cells and lethality just after birth (Cang et al, 2006). MEFs deficient in Ddb1 showed defective proliferation and were UV-sensitive. A total of 90% of HR23B mutants suffer intrauterine or neonatal death (Ng et al, 2002). The surviving HR23B^−/− mice were growth retarded and males were sterile; however, their NER and UV sensitivity remained normal. In contrast, mutants for HR23A, another homologue of the S. cerevisiae NER gene Rad23, are viable and their cells proficient in UV responses (Ng et al, 2003). However, mutations of both HR23A and HR23B lead to embryonic lethality and increased UV sensitivity, showing a redundancy between these two NER proteins (Ng et al, 2003). It remains to be shown whether inactivation of Ddb1 or mHR23A/B would facilitate spontaneous or UV-induced cancer in mouse models.

Mice mutant for Csa (Ercc8) or Csb (Ercc6) have been generated (van der Horst et al, 1997, 2002). These mutants are viable, but their cells, although competent in GG-NER, are nonetheless impaired in TC-NER and are UV-sensitive. Surprisingly, in contrast to CS individuals, Csa- and Csb-mutant mice are prone to UV-induced skin cancer. This discrepancy is also evident in TTD mouse models, as in contrast to TTD individuals, Xpd^R722W mice are susceptible to UV-induced skin cancer (de Boer et al, 1999).

As mentioned earlier, cross talk exists between DNA-damage repair pathways and certain DNA-damage repair proteins. For example, during NER, the endonuclease ERCC1/XPF cleaves a DNA strand on the 5′-side of the lesion, allowing the repair to proceed. However, ERCC1/XPF is also implicated in interstrand crosslink (ICL) repair and HR, suggesting that the phenotypes observed in Ercc1 or Xpf mutants are unlikely to result only from the impairment of NER. Cells deficient in Ercc1 or Xpf show high sensitivity to UV and to the ICL mitomycin C (MMC), and mice with inactivation of Ercc1 or Xpf are runted, dying at approximately 3–4 weeks of age (McWhir et al, 1993; Weeda et al, 1997; Tian et al, 2004b).

Thus, NER is an important DNA-repair pathway and its impairment is associated with growth defects, excessive UV sensitivity, and in certain cases, increased skin cancer.

The BER pathway

The BER pathway deals with base damage, the most common insult to cellular DNA (Figure 1; Wilson and Bohr, 2007). Two sub-pathways, short-patch BER and long-patch BER, are involved in BER. The short-patch BER sub-pathway typically replaces a single nucleotide, whereas the long-patch sub-pathway results in the incorporation of 2–13 nucleotides. The two sub-pathways progress through different major processes that initially involve the removal of the damaged base by glycosylases such as Ogg1 and Mutyh (Myh). This is followed by strand incision of the apurinic or apurimidinic (AP) site by the endonuclease APE1. The newly generated gap is filled by incorporation of nucleotide(s), mediated by DNA polymerase-β (Polβ) in the case of the short-patch BER sub-pathway, and by Polβ and/or Polɛ or δ in the case of the long-patch BER sub-pathway. Strand ligation is carried out by the XRCC1/ligase III (LigIII) complex in the case of the short-patch sub-pathway. For the long-patch BER sub-pathway, other proteins are involved in the repair process before the ligation takes place. Thus, FEN1, PARP1, and LigI participate in the DNA synthesis–ligation step and displace the 2- to 13-base DNA flap. FEN1, a 5′-flap endonuclease and 5′–3′ exonuclease, excises the flap and the strand ligation is carried out by LigI.

For a period of time, little evidence existed to support the involvement of BER in human cancer or any other disorders. However, recent studies have demonstrated the existence of a human disorder linked to defective BER (Figure 2). This autosomal recessive disorder, referred to as MUTYH-associated polyposis (MAP), is associated with biallelic germline mutations of the human MUTYH, and is characterized by multiple colorectal adenomas and carcinomas (Cheadle and Sampson, 2007). The glycosylase MUTYH, a bacterial mutY homologue, functions in the BER of oxidative DNA damage by excising adenines misincorporated opposite 8-oxoG. Deficient repair of this damage results in G:C → T:A mutations, typically found in the adenomatous polyposis coli (APC) gene in MAP tumors.

In addition to Myh, several mammalian glycosylases are involved in BER, including Nth1, Ogg1, Ung, and Aag. Murine homozygous mutants for the glycosylases Nth1, Ogg1, Ung, Aag, and Mutyh have been generated by gene targeting, and surprisingly, these mutants were viable (Table IV). Nth1^−/−, Ogg1^−/−, and Aag^−/− mice showed no overt abnormalities (Engelward et al, 1997; Klungland et al, 1999; Ocampo et al, 2002; Takao et al, 2002); however after a long latency, Ung^−/− and Mutyh^−/− mice developed B-cell lymphomas and intestinal tumors, respectively (Nilsen et al, 2003; Sakamoto et al, 2007). These findings suggest functional redundancy of certain BER glycosylases. This is further supported by the pronounced phenotypes of Ogg1^−/− Mutyh^−/− mice; these mice had shorter life span and elevated risk for cancer, with 50% of double mutants developing lung tumors, lymphomas, sarcomas, and others tumors by 15 months of age (Xie et al, 2004).

In contrast to glycosylases, targeted inactivation of enzymes that act downstream of the glycosylases in BER resulted in embryonic or post-natal lethality. Thus, homozygous mutants for Ape1 (Ludwig et al, 1998), LigI (Petrini et al, 1995), LigIII (Puebla-Osorio et al, 2006), Xrcc1 (Tebbs et al, 1999), or Flap endonuclease 1 (Fen1) (Kucherlapati et al, 2002) died during embryonic development, whereas homozygous mutants for Polβ died immediately after birth (Gu et al, 1994; Sugo et al, 2000). The death of mutants such as Xrcc1^−/−, LigIII^−/− and Polβ^−/− was preceded by elevated levels of apoptosis (Gu et al, 1994; Tebbs et al, 1999; Sugo et al, 2000; Puebla-Osorio et al, 2006). Interestingly, inactivation of p53 rescued the apoptosis but not the lethality of these mutants, suggesting the contribution of other p53-independent mechanisms in the deaths of these mutants.

Despite the presumed important role for the BER pathway in maintaining genomic integrity, mutations in this pathway have not significantly predisposed mutant mice for cancer. This is in contrast to mutations of other excision repair pathways, such as NER and MMR.

The HR repair pathway

DSB repair can be mediated by two major repair pathways depending on the context of the DNA damage, HR or NHEJ repair pathways (Figure 1; Kanaar et al, 2008). In bacteria and yeast, DSBs are preferentially repaired by HR, whereas more than 90% of DSB in mammalian cells are repaired by NHEJ. Both pathways are well defined and their impairment is associated with defects and pathologies, including increased cell death, cell-cycle arrest, telomere defects, genomic instability, meiotic defects, immunodeficiency, and cancer (Krogh and Symington, 2004; Sung and Klein, 2006).

HR is a multistep process that requires several proteins and operates at the S or G2 phase of the cell cycle. Although it accounts only for the repair of ∼10% of DSBs in mammalian cells, HR defects can have severe consequences, as demonstrated by the human syndromes AT-like disorder (ATLD) and the NBS (Figure 2; Thompson and Schild, 2002). The predisposition to either syndrome has been linked to mutations in the MRN complex, which is important for the resection of DSBs (Thoms et al, 2007). However, it is important to note MRN functions are not restricted to HR, as is also involved in NHEJ, checkpoint activation, and telomere maintenance (Niida and Nakanishi, 2006).

The very rare human ATLD is associated with hypomorphic mutations in the MRE11 gene (Taylor et al, 2004). Patients with this disorder exhibit clinical features similar to AT, including immunodeficiency and progressive neurological degeneration (Thoms et al, 2007). However, in contrast to AT, ATLD is not associated with ocular telangiectasia and the course of the disease is considerably milder. Similar to AT and NBS, cellular features of ATLD include defective DSB repair and repair-related cell responses, hypersensitivity to IR, as well as increased spontaneous and IR-induced genomic instability.

The NBS is a rare human autosomal recessive disorder caused by hypomorphic NBS1 (NIBRIN) mutations (Thoms et al, 2007). This disorder is characterized by growth retardation, immunodeficiency, microcephaly, and cancer predispositions, particularly lymphomas. Cellular characteristics of NBS include radiosensitivity, increased levels of spontaneous and IR-induced chromosome breakage, and defective cell-cycle checkpoints.

In contrast to both MRE11 and NBS1, no data have yet been reported to support a role for RAD50 in human chromosomal breakage or immunodeficiency syndromes.

The functions of the MRN complex have been studied in various organisms. S. cerevisiae strains carrying null mutations in components of the Mre11p–Rad50p–Xrs2p (MRX) complex have been generated, where Xrs2p is the functional homologue of mammalian NBS1. MRX mutants are viable, hypersensitive to IR and MMS, and are defective in meiotic recombination (Krogh and Symington, 2004). Mutations of the MRN complex have also been assessed in DT40. This chicken B-lymphocyte cell line is deficient in p53, but highly competent for homologous recombination and gene targeting. Deficiency of Mre11 in DT40 is lethal and its conditional inactivation leads to radiosensitivity, defective proliferation, genomic instability, and decreased HR (Yamaguchi-Iwai et al, 1999). DT40 deficient in NBS1 also display increased IR sensitivity, abnormal S-phase checkpoints, and decreased HR (Tauchi et al, 2002).

In contrast to S. cerevisiae, null mutations of Mre11 (Xiao and Weaver, 1997), Nbs1 (Zhu et al, 2001), and Rad50 (Luo et al, 1999) result in early mouse embryonic lethality (Table V). Therefore, to circumvent embryonic lethality, hypomorphic and tissue-specific mutants for the three MRN components were generated. Mre11^ATLD1/ATLD1-mutant mice carrying an A to T substitution at amino acid 1894 of Mre11, which results in a 75-amino-acid truncation, exhibited reduced female fertility, defective ATM functions, and increased genomic instability. However, no cancer phenotypes were observed (Theunissen et al, 2003).

Rad50^s (Rad50^k22M) hypomorphic mutant mice have also been generated (Bender et al, 2002). Approximately 40% of Rad50^s/s mutants die in utero; however, those that are viable show progressive haematopoietic failure, short life span, and increased predisposition to thymic lymphoma. In contrast to rad50^s in yeast, Rad50^s/s did not impair meiotic progression and the mutants were fertile; however, p53-mediated apoptosis was increased in the testes. Inactivation of the Atm–Chk2–p53 pathway by the hypomorphic mutation Mre11^ATLD1 or the Nbs1^ΔB (Nbs1^Δ4−5) mutation, was able to rescue the depletion of haematopoeitic cells in the Rad50^s/s mutants (Morales et al, 2005). Surprisingly, tumorigenesis, senescence, and radiosensitivity associated with Atm mutation were all partially suppressed by the Rad50^s mutation (Morales et al, 2005). Although the exact mechanism for this rescue is not fully understood, it might involve compensatory activation of other checkpoint pathways, such as the ATR pathway.

Similar to Rad50 and Mre11 mutations, Nbs1-null mutations resulted in early embryonic lethality (Zhu et al, 2001; Dumon-Jones et al, 2003). In contrast, Nbs1^Δ2−3and Nbs1^Δ4−5 hypomorphic mutants were viable (Kang et al, 2002; Williams et al, 2002). Cells from these mutants were defective in the intra-S-phase and G2/M checkpoints. Heterozygous mice carrying a null Nbs1 mutation (Dumon-Jones et al, 2003), as well as homozygous mice carrying Nbs1^Δ2−3or Nbs1^Δ4−5 hypomorphic mutations (Kang et al, 2002; Williams et al, 2002), demonstrated a mild predisposition for cancer. Interestingly, whereas p53 inactivation shortened the tumor latency of Nbs1^Δ4−5 mutants, Atm inactivation, or its impaired function on Mre11^ATLD1/ATLD1-mutant background, resulted in synthetic embryonic lethality of Nbs1^Δ4−5 mutants (Williams et al, 2002; Morales et al, 2005).

In addition to the previous models, a conditional mutant strain for Nbs1 has also been generated (Frappart et al, 2005). Neuronal inactivation of Nbs1 in these mice leads to chromosomal breaks, microcephaly, growth retardation, cerebellar defects, and ataxia, representing a combination of features characteristic of NBS, AT, and ATLD. p53 inactivation in this model also significantly rescued the neurological defects associated with Nbs1 mutations (Frappart et al, 2005).

Thus, the MRN complex is essential for maintaining genomic integrity, cell viability, and checkpoint activation. Moreso, its requirement in various species demonstrates its essential conserved functions.

Similar to Rad50, Rad52 is a member of the Rad52 epistasis group of proteins originally identified by their requirement for the repair of IR-induced DNA damage (Krogh and Symington, 2004). Inactivation of rad52 in S. cerevisiae results in increased radiation sensitivity and decreased recombination (Symington, 2002). However, its inactivation in DT40 reduced gene-targeting frequency, but did not lead to increased IR sensitivity and viability and growth of the mutant cells were not affected (Yamaguchi-Iwai et al, 1998). The effects of inactivation of mouse Rad52 were also investigated. Null Rad52 mouse embryonic stem (ES) cells obtained by gene targeting demonstrated a 30–40% decrease in the frequency of HR compared with controls (Rijkers et al, 1998). In contrast to rad52 mutants in S. cerevisiae, Rad52-deficient ES cells were not hypersensitive to IR or agents that induce DSBs. Rad52^−/− mice, were viable, fertile, and showed no overt abnormalities (Rijkers et al, 1998). Rad52 inactivation was shown to extend the life span of Atm^−/− mice, partially rescue their T-cell development, and suppress/delay their tumorigenesis (Treuner et al, 2004). However, growth defects, infertility, and radiosensitivity of Atm^−/− mice were not rescued on by a Rad52-mutant background. The reasons for the rescue of certain Atm^−/− phenotypes by Rad52 inactivation are not clear, but nevertheless this is reminiscent of the rescue mediated by Rad50^s/s mutation of tumorigenesis, senescence, and radiosensitivity associated with Atm inactivation (Morales et al, 2005).

The mammalian Rad51 family is also important for HR. This family is composed of RAD51, disrupted meiotic cDNA 1 (Dmc1), and five RAD51 paralogues, RAD51B, RAD51C, RAD51D, XRCC2, and XRCC3. RAD51, a homologue of the bacterial DNA-dependent ATPase, RecA, is central for the homology search and strand exchange during HR.

Mutants rad51 in S. cerevisiae are viable, IR sensitive, have meiotic defects, and accumulate meiosis-specific double-strand breaks (Shinohara et al, 1992). Rad51 deficiency in DT40 leads to chromosomal breakage, G2/M arrest, and cell death (Sonoda et al, 1998). Rad51-null mutation in mice results in early embryonic lethality associated with chromosomal loss, radiosensitivity, decreased cell proliferation, and increased apoptosis (Lim and Hasty, 1996; Tsuzuki et al, 1996a).

Mutants for Rad51 paralogues were also generated. Thus, for example, inactivation of Rad51 paralogues in DT40 did not compromise cell survival (Takata et al, 2000, 2001). However, decreased gene-targeting efficiency, increased spontaneous, and induced cell death in response to IR or MMC, as well as elevated chromosomal aberrations, were observed in these mutants. Mutant mice for the Rad51 paralogue Rad51B (Shu et al, 1999), Rad51D (Pittman and Schimenti, 2000), or Xrcc2 (Deans et al, 2003) were also generated. These mutations resulted in lethality at different stages of embryonic development and their effects on cell growth were variable. Consistent with the role of Rad51 and its paralogues in HR, their mutations result in accumulation of DNA damage and activation of cellular checkpoints, including p53. Consequently, the embryonic lethality of Rad51 (Lim and Hasty, 1996), Rad51 B (Shu et al, 1999), and Rad51 D (Smiraldo et al, 2005) mutants was delayed on p53-null background. The role of p53 in the embryonic lethality of Xrcc2^−/− mutants remains controversial (Orii et al, 2006; Adam et al, 2007).

The meiosis specific the Rad51 paralogue Dmc1 is essential for meiotic recombination in S. cerevisiae (Bishop et al, 1992). In mice, Dmc1 expression is restricted to the testis and ovaries, and null Dmc1 mutants are sterile and exhibit arrested gametogenesis in the first meiotic prophase (Pittman et al, 1998; Yoshida et al, 1998). These results demonstrate the conservation of the essential meiotic function of Dmc1.

Rad54 is another member of the Rad52 epistasis group involved in HR. Similar to rad51 and rad52 mutants, S. cerevisiae deficient in rad54 are highly sensitive to IR. DT40 cells mutant for Rad54 are viable, hypersensitive to IR, exhibit slow growth, decreased rate of Ig gene conversion, and show a drastic decrease in gene-targeting efficiency (Bezzubova et al, 1997). Rad54 paralogues exist and include rdh54 in S. cerevisiae and Rad54B in mammalian cells. In contrast to rad54 mutants in S. cerevisiae, rdh54 mutants are not sensitive to IR and show no defects in mitosis (Shinohara et al, 1997). However, meiotic recombination was affected in rdh54 mutants and was completely defective in S. cerevisiae lacking the two Rad54 orthologues (Shinohara et al, 1997).

Homozygous mutant mice lacking Rad54 and/or its paralogue Rad54B are viable and ES cells deficient in Rad54 or Rad54B are hypersensitive to IR, MMS, and MMC (Essers et al, 1997; Bross et al, 2003; Wesoly et al, 2006). Rad54^−/−, but not Rad54B^−/−, ES cells have a 3- to 40-fold decrease in HR as assessed by gene targeting. Dual inactivation of both paralogues further impairs HR, as double mutant ES cells show a further 10-fold reduction in their gene targeting efficiency compared with Rad54-null ES cells. Although Rad54 is important for HR, mice deficient in Rad54 and/or Rad54B do not have a predisposition to cancer.

BRCA1 and BRCA2, the early-onset breast cancer-susceptibility genes, have also been demonstrated to partake in HR-mediated DSB repair. Germline mutations of BRCA1 or BRCA2 predispose women to familial human breast and ovarian cancer (Narod and Foulkes, 2004). Individuals with germline mutations for BRCA1 or BRCA2 also have increased risk for other cancers, including prostate cancer. Although the BRCA1 gene has important functions in DNA-damage signalling/repair it is also involved in other cellular processes, including transcription, ubiquitination, oestrogen receptor signalling, and chromatin remodelling. In response to DSBs, BRCA1 is phosphorylated by ATM, ATR, and Chk2, underscoring its functional regulation by other molecules involved in DNA-damage checkpoints. BRCA2 functions in the loading of the HR protein RAD51 during filament formation. It directly binds to RAD51 and its phosphorylation on Ser3291 inhibits this binding (Esashi et al, 2005). In addition to its role in HR, recent studies have suggested a role for BRCA2 in the stabilization of stalled RFs (Lomonosov et al, 2003).

Homozygous mutations of murine Brca1 resulted in embryonic lethality. (Gowen et al, 1996; Hakem et al, 1996; Liu et al, 1996; Ludwig et al, 1997; Xu et al, 2001). Premature death of Brca1-mutant embryos is likely due to accumulation of damaged DNA and activation of DNA-damage checkpoints, including Chk2–p53. Inactivation of p53 delayed the embryonic lethality of Brca1-null mutants (Hakem et al, 1997; Ludwig et al, 1997) and completely rescued survival of Brca1 hypomorphic mutants (Xu et al, 2001). Studies of hypomorphic and conditional null Brca1 mutant females demonstrated that similar to humans, Brca1 mutations increase the risk for mammary cancer (Xu et al, 1999; McPherson et al, 2004b). Interestingly, inactivation of p53, or Chk2, drastically facilitates development of mammary tumors on Brca1-mutant backgrounds (Xu et al, 1999; McPherson et al, 2004b). A targeted Brca1-null mutation to the T-cell lineage resulted in increased genomic instability, apoptosis, cell-cycle arrest, and a drastic depletion of the T-cell lineage (Mak et al, 2000). Interestingly, development of Brca1-null T-cells was completely rescued in p53- or Chk2-mutant backgrounds, and this rescue was at the expense of increased genomic instability and increased risk for tumorigenesis (McPherson et al, 2004b).

BRCA1 plays a major role in the DNA-damage response as it interacts with the MRN complex and γ-H2AX at the site of DSB. Likewise, it is also required for the recruitment of other proteins such as Rad51, BRCA2, and BARD1 to sites of DSBs (Greenberg et al, 2006). HR, as measured by the efficiency of gene targeting in ES cells, was drastically reduced in Brca1 hypomorphic mutants (Moynahan et al, 1999). Other defects, including compromised telomeres integrity and male sterility, were also associated with Brca1 mutations (Xu et al, 2001; McPherson et al, 2006).

Similar to Brca1 and Rad51 mutants, null mutations for Brca2 result in early mouse embryonic lethality and impaired HR (Ludwig et al, 1997; Suzuki et al, 1997; Moynahan et al, 2001). Furthermore, loss of Brca2 in murine cells resulted in increased genomic instability and activation of p53. A small subset of Brca2 hypomorphic mutants survived embryonic development (Connor et al, 1997; Friedman et al, 1998). These mice were growth defective, sterile, and predisposed for thymic lymphoma, and their cells were sensitive to DNA damaging agents, including IR, MMS, and UV radiation. Conditional mutant strains have also been generated using the Cre/loxP system and similar to Brca1 mutation, dual inactivation of Brca2 and p53 in mammary epithelial cells resulted in increased frequency and decreased latency for mammary tumors (Jonkers et al, 2001).

Beside the proteins described above, efficient HR requires other proteins such as members of the family of RecQ helicases (Hickson, 2003). The human family of RecQ helicases includes BLM, REQ, WRN, RECQL4, and RECQL5, and is important for unwinding DNA, repairing stalled RFs, promoting HR, and maintaining overall genomic integrity. Mutations of some of these helicases are associated with rare human syndromes (Hickson, 2003). Thus, the Werner syndrome (WS) and the Rothmund Thomson syndrome (RTS) are associated with mutations of human WRN and RECQL4 genes, respectively (Figure 2). WS features include premature ageing, short stature, and cancer predisposition. Clinical features of RTS include short stature, skin pigmentation changes, skin atrophy, and increased cancer predisposition.

In addition mutations of the human BLM gene are associated with the Bloom syndrome (BS), a rare autosomal recessive disorder characterized by growth defects, immune deficiency, reduced fertility, and predisposition to a large spectrum of cancers. Cells from BS patients have elevated sister-chromatid exchange (SCE) and genomic instability. Three mouse models for Blm mutation have been generated, but only one was viable (Chester et al, 1998; Luo et al, 2000; Goss et al, 2002). Similar to BS patients, viable Blm mice are prone to a wide range of tumors, and cells from these mutant mice show elevated levels of SCE, a hallmark cellular phenotype of BS. In addition, Blm deficiency leads to increased mitotic recombination and somatic loss of heterozygosity (Luo et al, 2000).

While most steps essential for HR in eukaryotes have been well characterized, the identification of the resolvase(s) required for the resolution of Holliday junctions (HJs) during mitotic and meiotic recombination turned out to be very challenging. Resolution of HJs is critical for HR and is mediated in E. coli by the resolvase RuvC (West, 1997); however, the search for true HJ resolvase(s) in eukaryotes is still ongoing.

Yeast Mus81 (MMS, UV-sensitive, clone 81) with its partner Eme1 (Schizosaccharomyces pombe) or MMS4 (S. cerevisiae) forms a structure-specific endonuclease important for DNA-damage repair (Osman and Whitby, 2007). Yeast mus81, eme1, or mms4 mutants show increased sensitivity to DNA-damaging agents that interfere with normal progression of RFs caused by agents such as UV radiation, MMS, and camptothecin, but their sensitivity to IR is not affected (Osman and Whitby, 2007) . These mutants also show meiotic defects supporting a role for this endonuclease in this process.

In vitro studies indicate that yeast and mammalian Mus81, with their partner Eme1/Mms4, efficiently cleaves various DNA structures that mimic RFs, D-loops, and nicked HJs, but the cleavage activity of intact HJs was weak (Osman and Whitby, 2007). Thus, at least in mammalian cells, Mus81 with Eme1 or with Eme2, another eme1 homologue, forms a heterodimeric 3′-flap/RF endonucleases that process stalled RFs and recombination intermediates, but does not possess typical characteristics of an HJ resolvase (Abraham et al, 2003; Ciccia et al, 2003, 2007; Ogrunc and Sancar, 2003).

Gene-targeted inactivation of Mus81 or Eme1 in mouse and human cells increased sensitivity to ICL, hydroxyurea, but not UV or IR and enhanced spontaneous and MMC-induced genomic instability (Abraham et al, 2003; McPherson et al, 2004a). Mus81 was shown to be important for generating ICL-induced DSBs, as well as for mediating the restart of stalled or blocked RFs (Hanada et al, 2006). Our studies of a mouse model for Mus81-null mutation demonstrated that Mus81 is a haploinsufficient tumor suppressor (McPherson et al, 2004a). Heterozygous and homozygous Mus81 mutants show cancer predisposition, particularly to T- and B-cell lymphomas. The MMC hypersensitivity of Mus81-mutant cells and mice is p53-dependent (Pamidi et al, 2007). Interestingly, on a p53-null background, Mus81-mutant mice have more genomic instability and develop sarcomas and multiple different tumors with short latency and high penetrance, demonstrating a role for p53 in suppressing cancer associated with Mus81 mutation (Pamidi et al, 2007). Both Mus81- and Mus81p53-mutant mice are fertile, thus failing to support a requirement for Mus81 in meiotic recombination. A second Mus81-mutant strain (Mus81^Δ9−12) showed increased genomic instability but failed to show increased tumorigenesis (Dendouga et al, 2005). A role in cancer for Eme1 or Eme2 requires further investigations.

Recent studies have implicated the mammalian RAD51 paralogues RAD51C–XRCC3 in HJ resolution (Liu et al, 2004). Rad51C–XRCC3 binds HJs in vitro and cell extracts from Rad51C- or XRCC3-deficient hamster cells, exhibit low HJ resolvase activities. In addition, depletion of RAD51C from HeLa cell extracts strongly impairs in vitro HJ branch migration and resolution. Similar to other Rad51 paralogues, null mutation of mouse Rad51C results in embryonic lethality. However, a viable mouse model carrying intronic integration of a Neomycine selection cassette resulting in decreased Rad51 expression has been recently reported (Kuznetsov et al, 2007). About 40% of mutant males and 10% of mutant females were infertile. Cell extracts from MEFs null for Rad51C and p53 show reduced in vitro HJ resolvase activities compared with controls. Thus, the current data support a role for the mammalian RAD51C in HJ resolution; however, further studies are still required to establish RAD51C as a typical HJ resolvase and to demonstrate whether other mammalian HJ resolvases also exist.

Thus, the human syndromes and early onset of breast cancer, and the developmental defects, increased genomic instability and tumorigenesis associated with impaired HR in mouse models, all demonstrate the in vivo requirement for HR-mediated DNA-damage repair.

We thank the members of the Hakem laboratory for helpful discussions, and Amanda Fenton, Anne Hakem, Jacinth Abraham, and Renato Cardoso for reviewing the paper. Research in the Hakem laboratory was supported by the Canadian Institutes of Health Research, the National Cancer Institute of Canada, the Cancer Research Society, and the Leukemia & Lymphoma Society Of Canada. We apologize to those whose work was not cited directly owing to space limitations.