|Home | About | Journals | Submit | Contact Us | Français|
The Mre11-Rad50-Nbs1 (MRN) complex has many biological functions: processing of double-strand breaks in meiosis, homologous recombination, telomere maintenance, S-phase checkpoint, and genome stability during replication. In the S-phase DNA damage checkpoint, MRN acts both in activation of checkpoint signaling and downstream of the checkpoint kinases to slow DNA replication. Mechanistically, MRN, along with its cofactor Ctp1, is involved in 5′ resection to create single-stranded DNA that is required for both signaling and homologous recombination. However, it is unclear whether resection is essential for all of the cellular functions of MRN. To dissect the various roles of MRN, we performed a structure–function analysis of nuclease dead alleles and potential separation-of-function alleles analogous to those found in the human disease ataxia telangiectasia-like disorder, which is caused by mutations in Mre11. We find that several alleles of rad32 (the fission yeast homologue of mre11), along with ctp1Δ, are defective in double-strand break repair and most other functions of the complex, but they maintain an intact S phase DNA damage checkpoint. Thus, the MRN S-phase checkpoint role is separate from its Ctp1- and resection-dependent role in double-strand break repair. This observation leads us to conclude that other functions of MRN, possibly its role in replication fork metabolism, are required for S-phase DNA damage checkpoint function.
Mre11-Rad50-Nbs1 (MRN) is a heterotrimeric complex composed of Mre11, Rad50, and Nbs1 (Carney et al., 1998 ; Haber, 1998 ; D'Amours and Jackson, 2002 ; Williams et al., 2007 ). Originally identified in budding yeast as recombinational repair proteins, Rad50 and Mre11 are conserved from humans to bacteria and even some bacteriophages. The third MRN subunit found in eukaryotes, Nbs1, is thought to play a regulatory role (Lee et al., 2003 ). The degree of conservation of the MRN complex between species suggests a critical role in DNA transactions. One of its conserved roles is in the S-phase DNA damage checkpoint, in which it is involved in slowing replication in response to DNA damage in both humans and fission yeast (Young and Painter, 1989 ; Lavin and Shiloh, 1997 ; Stewart et al., 1999 ; Chahwan et al., 2003 ).
The MRN complex has many roles in cellular metabolism; it functions not only in the S-phase DNA damage checkpoint but also in recombinational repair, telomere maintenance, processing of double-strand breaks in meiosis and another, as yet undefined, role in normal replication (Carney et al., 1998 ; Haber, 1998 ; D'Amours and Jackson, 2002 ). Its role in normal replication prevents the accumulation of broken replication forks possibly through homologous recombination (Costanzo et al., 2001 ; Maser et al., 2001 ; Chahwan et al., 2003 ; Trenz et al., 2006 ; Wen et al., 2008 ). In vertebrates, the deletion of any member of the MRN complex results in cell lethality (Xiao and Weaver, 1997 ; Luo et al., 1999 ; Zhu et al., 2001 ; Dumon-Jones et al., 2003 ; Difilippantonio et al., 2005 ). The deletion of MRN subunits in budding and fission yeast is characterized by slow growth, S-phase accumulation, and increased chromosome instability. These phenotypes are probably due to defects in repairing spontaneous fork collapses (Hartsuiker et al., 2001 ; Myung and Kolodner, 2002 ). Hypomorphic mutations in Mre11 and Nbs1 lead to two diseases in humans, ataxia-telangiectasia-like disease (ATLD) and Nijmegan breakage syndrome (NBS) (Petrini, 2000 ). Both of these diseases are phenotypically related to ataxia-telangiectasia (AT), which is caused by mutations in the checkpoint kinase AT-mutated (ATM) (Painter and Young, 1980 ; Stewart et al., 1999 ). In addition to neuronal and immunological phenotypes, AT patients display predisposition to cancer, hypersensitivity to double-strand break-inducing agents, chromosome instability, and radioresistant DNA synthesis (RDS). RDS occurs when cells are unable to induce the S-phase DNA damage checkpoint and therefore are unable to slow replication. Cells from AT patients show no slowing of replication in response to ionizing radiation and radiomimetic agents (Lavin and Shiloh, 1997 ). Cells of ATLD and NBS patients show an intermediate overall slowing of replication (Young and Painter, 1989 ; Falck et al., 2002 ). These hypomorphic mutations must maintain some checkpoint-independent function of the MRN complex because the cells are able to survive (Carney et al., 1998 ; Stewart et al., 1999 ). Thus, it has been suggested that they display separation of checkpoint-dependent and -independent functions of MRN.
In humans and fission yeast, MRN plays roles both upstream and downstream of the checkpoint kinases to mediate the DNA damage response (Lavin, 2004 ; Willis and Rhind, unpublished data). The complex is one of the first sets of proteins recruited to sites of damage both in humans and in Saccharomyces cerevisiae (Nelms et al., 1998 ; Lisby et al., 2004 ). MRN directly activates ATM, although partial activation still occurs without the MRN (Lee and Paull, 2004 ). In addition, MRN is phosphorylated by ATM on Nbs1 (Stewart et al., 1999 ; Gatei et al., 2000 ; Lim et al., 2000 ; Zhao et al., 2000 ). In addition to its role in the activation of checkpoint signaling, MRN is required downstream of the checkpoint kinases to slow replication in response to DNA damage (Falck et al., 2002 ; Willis and Rhind, unpublished data). The downstream checkpoint-dependent role of MRN is not well understood, but it may involve recombinational repair of stalled forks (Rhind and Russell, 2000 ; Willis and Rhind, 2009 ).
In Schizosaccharomyces pombe, S-phase checkpoint signaling requires Rad3, the fission yeast homologue of ATR (ATM- and Rad3-related kinase); the ATM homologue Tel1 is not required for the S-phase DNA damage checkpoint (Bentley et al., 1996 ; Nakamura et al., 2002 ; Willis and Rhind, unpublished data). Nonetheless, MRN is required for full activation of checkpoint signaling during S-phase (Willis and Rhind, unpublished data).
MRN possesses several biochemical activities, including double-strand and single-strand DNA binding, DNA tethering, and ATPase, adenylate kinase, and endo- and exonuclease activities (Williams et al., 2007 ). Mre11 is the nuclease subunit of MRN. It interacts with both Rad50 and Nbs1 independently and can itself form homodimers and homomultimers (D'Amours and Jackson, 2002 ). Mre11 contains DNA binding domains as well as phosphoesterase motifs responsible for the nuclease activities of the complex (Furuse et al., 1998 ; Usui et al., 1998 ; Moreau et al., 1999 ; Figure 1). The nuclease activity of the MRN complex plays a critical role in 5′ resection of double-strand breaks to create single-strand intermediates for repair by homologous recombination (Lewis et al., 2004 ; Jazayeri et al., 2006 ; Myers and Cortez, 2006 ; Chen et al., 2008 ). The resection activity and double-strand break repair function of MRN require a cofactor: Ctp1 in fission yeast, Sae2 in budding yeast, and CtIP in humans (Chen et al., 2007 ; Limbo et al., 2007 ; Sartori et al., 2007 ).
With all the roles that the MRN plays in cellular metabolism, dissecting its role in the S-phase DNA damage checkpoint is difficult. For example, it is not known whether MRN has one major biochemical activity that is responsible for all of its DNA repair roles or whether it has multiple separable biochemical activates that are independently required for different roles. Mutations that separate its role in the checkpoint from its other roles would directly address this question. In particular, analysis of MRN separation-of-function alleles might elucidate which of the various biochemical activities of MRN are required for checkpoint-dependent slowing of replication in response to DNA damage. Mutations in S. cerevisiae Mre11 show separation of function between its meiotic roles and its mitotic roles (Nairz and Klein, 1997 ; Symington, 2002 ). Furthermore, several nuclease dead mutations in budding yeast maintain many mitotic MRN functions, yet they are defective in meiotic function and are sensitive to DNA damage (Bressan et al., 1998 ; Moreau et al., 1999 ; Chen et al., 2001 ; Yazdi et al., 2002 ; Krogh et al., 2005 ). Mutations that separate the checkpoint-dependent from checkpoint-independent roles in fission yeast have not been reported, but based on characterization of human ATLD and NBS phenotypes, which seem to affect checkpoint-dependent but not constitutive functions of MRN, we hypothesized that they would exist. Given the precedent of the ATLD alleles in humans and the nuclease dead alleles in budding yeast, we decided to focus on the fission yeast homologue of Mre11, Rad32. We have phenotypically characterized various alleles of rad32 and found that the role of MRN in checkpoint-dependent slowing of replication is, in fact, genetically separable from its roles in checkpoint signaling and homologous recombination.
Yeast were grown in yeast extract plus supplements (YES) at 30°C and manipulated by standard methods (Forsburg and Rhind, 2006 ). Temperature-sensitive strains were grown at 25°C unless otherwise stated. Strains used for this study are listed in Table 1.
Rad32 C-terminal deletion strain was made using a polymerase chain reaction (PCR) cassette insertion to replace residues 528–649 with a 13Myc tag (Bahler et al., 1998 ). The sequences of the oligonucleotides are listed in Supplemental Table S1. rad32::kanMX4, rad3::ura4, rad2::ura4 have been described previously (Bentley et al., 1996 ; Morishita et al., 2002 ; Chahwan et al., 2003 ).
Site-directed alleles of rad32 were made by oligomediated mutagenesis. Primers were designed to change the desired amino acids by four or less nucleotide changes and also create unique restriction sites. Primer sequences containing mutations and the appropriate restriction site changes are in Supplemental Table S1. The 3′ end of rad32 was cloned into pFA6a-13Myc:natMX6 (Sato et al., 2005 ) by using MPG50 and MPG51 with HindIII and PacI, creating pFS305. A second 3′ piece was cloned into that plasmid by using MPG52 and MPG53 cut with SacI and SacII, creating pFS306. Finally, 5′ mutants were cloned into pFS306 by using MPG3 and MPG6 with NdeI and HindIII as the product of a two-step PCR; each mutant primer was paired with either MPG3 or MPG6 for the first step and then each mutant product pair was used as template for PCR with MPG3 and MPG6 to create each of the six mutants. The resulting plasmids (pFS307-pFS313) were cut with Nde1 and SacII; the rad32 open reading frame fragment was transformed into wild-type (wt) cells (yFS105). Integrants were selected on nourseothricin. Accurate integration was confirmed by PCR and sequencing.
Cells were grown to mid-log phase in liquid YES and fivefold serially diluted. Aliquots of each dilution were plated on solid medium containing 0.01% MMS, 0.03% MMS, 1 mM HU, or 3 mM HU and then grown for 4 d at 30°C. For acute UV exposure, mid-log liquid YES cultures were counted, plated on YES solid media, irradiated in a UV Stratalinker 2400 (Stratagene, La Jolla, CA), and grown for 30°C for 3 or 4 d. X-ray irradiation was done in a Faxitron x-ray RX-650 at 130 kVp at 5 mA (10 Gy/min) at room temperature with cells suspended in liquid YES; cells were immediately plated in triplicate on YES solid media, incubated for 3 or 4 d, and counted.
Genomic DNA was isolated from each strain. DNA was digested overnight with EcoRI and run on a 2% agarose gel. Ethidium bromide staining confirmed equal loading. The gel was transferred by neutral transfer to Hybond N+ (GE Healthcare, Chalfont St. Giles, Buckinghamshire, United Kingdom) and probed with 32P-labeled telomere associated-1 (TAS1) (Nakamura et al., 2002 ).
rad32 mutant strains were crossed to rad2::ura4, and spores were digested with glusulase and plated on YES. The resulting colonies were replica plated to Edinburgh minimal media 2 supplemented with leucine, adenine, and histidine (EMM2 LAH) and either nourseothricin (Werner BioAgents, Jena, Germany) or Geneticin (G418; Invitrogen, Carlsbad, CA) as appropriate. Surviving colonies were counted. Random spore analysis was used because of the poor spore viability of the rad32 mutants. To correct for the low spore viability of rad32 mutants, expected viability (E) was calculated by normalizing the number of rad32 mutant (−) rad2Δ colonies to the number of rad32− rad2+ and rad32+ rad2Δ colonies by using the formula E = A × (B/A) × (C/A), where A is the number of rad32+ rad2+ colonies, B is the number of rad32− rad2+ colonies, and C is the number of rad32+ rad2Δ colonies.
The NHEJ plasmid assay was performed essentially as described previously (Manolis et al., 2001 ). The plasmid pUR19 (Barbet et al., 1992 ) was linearized with SmaI. Logarithmically growing cells were transformed with equal amounts of undigested or linear DNA by using the lithium acetate transformation method (Okazaki et al., 1990 ). Cells were plated on EMM2 LAH plates in triplicate. ura+ colonies were counted, and NHEJ frequency was calculated by the number of colonies from linear DNA over those transformed with undigested DNA. Cells grown in nitrogen-free media were first grown to log phase in YES and then transferred to EMM2-N media (United States Biological, Swampscott, MA) for 36 h before transformation (Ferreira and Cooper, 2004 ).
Cells were grown to mid-log phase in liquid YES, and then 40 OD of cells were collected, washed with water, frozen in liquid nitrogen, and stored at −80°C. Extracts were prepared as described previously (Boddy et al., 1998 ). Cells were lysed with 200 μl of lysis buffer. Cleared cell extracts were incubated with rabbit immunoglobulin G (IgG)-Agarose beads (Sigma-Aldrich, St. Louis, MO). Proteins were separated by SDS-polyacrylamide gel electrophoresis (PAGE), blotted, and visualized with anti-HA antibody directly conjugated to horseradish peroxidase (HRP; Roche Diagnostics, Indianapolis, IN). HRP was quenched with 0.10% sodium azide, the blot was washed, and proteins were visualized with anti-Myc antibody directly conjugated to HRP.
Cells were grown in liquid YES to mid-log phase and treated with 0.03% MMS for 4 h. Then, 10 OD pellets were collected for the kinase assay. In vitro kinase assays were performed as described previously (Kai et al., 2001 ; Dutta et al., 2008 ; Willis and Rhind, 2008). Immunopurified Cds1 was incubated with myelin basic protein and [γ32P]ATP at 30°C for 15 min. Proteins were separated by SDS-PAGE and visualized by autoradiography.
Synchronization, flow cytometry of isolated nuclei, and S-phase progression analysis was performed as described previously (Kommajosyula and Rhind, 2006 ; Willis and Rhind, 2008).
In a search for alleles that would separate the checkpoint functions of MRN from its checkpoint-independent functions, we constructed seven site-directed alleles analogous to nuclease dead alleles in S. cerevisiae (rad32-D25A, rad32-D65N, rad32-H134N, and rad32-H134L/D135V) or to ATLD alleles in humans (rad32-N122S, rad32-W215C, and rad32ΔC) (Figure 1A; Bressan et al., 1998 ; Moreau et al., 1999 ; Stewart et al., 1999 ; Chen et al., 2001 ; Yazdi et al., 2002 ; Fernet et al., 2005 ). These alleles were used to replace the wild-type rad32 allele at its endogenous locus. All of the alleles are tagged with 13 copies of the Myc epitope, so our analyses use rad32-13Myc for the wild-type control. In a variety of phenotypic assays, we found that cells carrying two of our alleles, the class I mutants (rad32-N122S and rad32-W215C), seem similar to wild type; cells carrying three alleles, the class II mutants (rad32-D25A, rad32-D65N, and rad32-H134N), seem similar to rad32Δ cells; cells carrying two other alleles, the class III mutants (rad32-H134L/D135V and rad32ΔC) resemble rad32Δ cells in all assays except the checkpoint assay, in which they display checkpoint proficiency (Table 3). The details of these phenotypic characterizations follow.
The simplest phenotypic assays for rad32 alleles are growth rate and cell morphology. rad32Δ cells accumulate DNA damage, which leads to cell cycle arrest and death in a significant fraction of cells and reduces the growth rate of the culture (Hartsuiker et al., 2001 ; Myung and Kolodner, 2002 ; Chahwan et al., 2003 ). The class II and III mutants resemble rad32Δ cells in both growth rate and heterogeneity of cell morphology; the class I mutants exhibit wild-type morphology and growth rate (Figure 1B). These results suggest that the class I mutants retain the Rad32 constitutive repair and genome stability functions, functions that seem to be missing in the class II and III mutants.
To more carefully investigate the repair and genome stability phenotypes of our mutants, we used a panel of assays. We first analyzed the DNA damage sensitivity of our mutants. We tested the survival of cells treated with UV irradiation, x-ray irradiation, and the alkylating agent MMS. In response to all three treatments, our mutants fall into two broad categories, with the class I mutants being similar to wild-type cells in resistance to DNA damage and the class II and III mutants being similar to rad32Δ cells in sensitivity (Figure 2, A–C). However, within those categories there is some heterogeneity. For example, both of the class I mutants are slightly more sensitive to MMS than wild type. In addition, one of the class II mutants, rad32-D65N, is slightly more resistant to all three DNA-damaging agents than rad32Δ whereas both of the class III mutants are slightly more resistant to UV and one of them, rad32ΔC, is slightly more resistant to MMS.
We also tested our mutants for sensitivity to HU, which blocks replication by inhibiting deoxyribonucleotide synthesis. Again, the class I mutants are similar to wild-type cells, and the class II and III mutants are similar to rad32Δ cells, with some heterogeneity. In this case, several of the class II and III mutants, rad32-H134N, rad32-D65N, and rad32ΔC, show intermediate resistance to HU between that of wild-type and rad32Δ cells.
rad32Δ strains have shortened telomeres compared with wild type, and rad32Δ rad3Δ strains suffer catastrophic telomere loss and are only able to survive by circularizing their chromosomes (Wilson et al., 1999 ; Nakamura et al., 2002 ). We tested our mutant strains for telomere loss in a rad3Δ background. rad3Δ double mutants were constructed and genomic Southern blots were probed with a telomere-associated sequence. rad32-D25A, rad32-H134L/D135V, and rad32ΔC lost their telomeres, whereas all other mutants have maintained their telomeres (Figure 3).
As an independent assay for the role of MRN in DNA repair, we examined the viability of our alleles in combination with rad2Δ. Rad2 is the fission yeast homologue of Fen1, a flap endonuclease involved in the maturation of Okazaki fragments in lagging-strand DNA synthesis (Murray et al., 1994 ; Symington, 1998 ; Moreau et al., 1999 ). Without Rad2, cells use other mechanisms to mature the RNA-containing flap (Symington, 1998 ). rad32Δ is synthetically lethal with rad2Δ, apparently because MRN-dependent homologous recombination is required after replication in the absence of Rad2 (Tavassoli et al., 1995 ; Symington, 1998 ). Because of the low viability of spores from diploids heterozygous for our rad32 alleles, we did our assay using random spore analysis. The class I alleles are not synthetically lethal with rad2Δ; all of the class II and III alleles are lethal (Table 2).
In S. cerevisiae, the MRN complex is required for NHEJ (Zhang and Paull, 2005 ). Previously, it was shown that the MRN is not required for NHEJ in asynchronous fission yeast (Manolis et al., 2001 ). However, because S. pombe spends most of its time in G2, these experiments did not address the requirement in G1, when NHEJ is most active (Ferreira and Cooper, 2004 ). In fact, there is a requirement for some proteins in G1 that are not required in G2 (Ferreira and Cooper, 2004 ). To look at the role of the MRN in NHEJ in G1, we used a plasmid religation assay in cells arrest in G1 by nitrogen starvation. Figure 4 confirms that the MRN is not required for NHEJ in fission yeast, even in G1.
MRN is required for the processing of double-strand breaks in meiosis and thus for spore viability (Tavassoli et al., 1995 ; Wilson et al., 1998 ; Symington, 2002 ). To test the function of our alleles in meiosis, we measured spore viability by random spore analysis. We tested the viability of spores from zygotes homozygous for our alleles. Class I mutants have spore viabilities similar to that of wild-type cells. Class II and class III mutants show a substantial reduction in spore viability, with all but one showing the same lack of spore viability as rad32Δ zygotes; one class III allele, rad32ΔC, retains some residual spore viability (Table 2).
Observations in S. cerevisiae indicate that some of the nuclease motif mutations disrupt MRN complex formation (Krogh et al., 2005 ). To assay for complex formation in our strains, we tested the interaction of Rad32 with Nbs1. We built doubly tagged strains carrying both a mutant rad32 allele carboxy-terminally tagged with 13 copies of the Myc epitope and a carboxy-terminal HA and protein A affinity (HA-TAP)-tagged nbs1 allele. Cells expressing both tagged proteins were lysed and Nbs1 captured on IgG agarose beads. After washing the beads, bound proteins were separated by SDS-PAGE, blotted, and probed with anti-HA to visualize Nbs1 (Figure 5). The blot was reprobed with anti-Myc to visualize Rad32. Two of the class II protein, Rad32-H134N and Rad32-D65N, as well as one of the class III proteins, Rad32ΔC, expressed well and bound to the beads at wild-type levels. Both of the class I proteins and one of the class III proteins, Rad32-H134L/D135V, expressed at lower levels than wild type. In addition, these proteins bound the beads with variable efficiency; in some preps they showed significant binding, albeit less than wild type, and in others they showed only background levels of binding. These mutations are believed to be disrupted the Nbs1 binding domain and inconsistent Nbs1 binding has been reported for the human homologues of the class I proteins (Stewart et al., 1999 ; Lee et al., 2003 ; Fernet et al., 2005 ). Finally, one class II protein, Rad32-D25A, reproducibly expressed at lower levels and did not bind to the beads above background levels, consistent with its failure to bind Rad50 (Tomita et al., 2003 ). rad32-D25A cells also failed to express Nbs1 at detectable levels in some preparations while expressing at very low levels in others.
We next examined the S-phase checkpoint response to damage of our rad32 mutants. MMS, which produces DNA damage in the form of base adducts, was used to induce the checkpoint. MMS-induced damage is recognized during replication and therefore is a good activator of the S-phase DNA damage checkpoint. Cells were synchronized in G1 by a combination of elutriation and a cdc10-ts block and release, treated with 0.03% MMS and followed through replication by flow cytometry (Figure 6, B–D). Untreated cells replicated between 40 and 80 min after release from G1. MMS-treated wild-type cells fail to complete replication by 180 min. However, checkpoint defective strains, such as rad3Δ, continue replication despite the presence of DNA damage. rad32Δ cells present a partial checkpoint response as they slow their S-phase progression more than rad3Δ but do not slow to the same extent as wild-type cells (Willis and Rhind, unpublished data). The class I mutants, which are wild type in other assays, had intact checkpoints and the class II mutants, which act as nulls in other assays, had defective checkpoints. However, the class III mutants, which act as nulls in other assays, have intact checkpoints and thus display a separation-of-function phenotype.
To test whether the checkpoint phenotypes correlate with defects in the role of MRN in S-phase DNA damage checkpoint signaling, we checked to see whether the Cds1 kinase activity is compromised in our mutants. MRN mutants in human cells display a partial defect in ATM checkpoint signaling (Lee and Paull, 2004 ). Similarly, we have seen a partial defect in Cds1 S-phase checkpoint kinase activity in nbs1Δ cells (Willis and Rhind, unpublished data). We found that the class I alleles have wild-type levels of kinase activity, whereas class II and III alleles have significantly lower kinase activity levels (Figure 7). One class II allele, rad32-D25A, is as defective for signaling as rad32Δ. Another class II allele, rad32-H134N, and both class III alleles are severely defective but show some residual signaling. The third class II allele, rad32-D65N, shows intermediate signaling.
To test the requirement of 5′ resection in the various functions of MRN, we examined the phenotypes of ctp1Δ cells. Ctp1 has recently been shown to function with the MRN complex in 5′ resection (Limbo et al., 2007 ; Sartori et al., 2007 ; Takeda et al., 2007 ). In our panel of phenotypic assays, ctp1Δ behaves as a class III separation of function allele (Table 3). In most assays, deletion of ctp1 causes a similar phenotypes to those of rad32Δ (Limbo et al., 2007 ; Akamatsu et al., 2008 , Figure 6A; data not shown). In particular, ctp1Δ cells display a significant defect in activation of the Cds1 checkpoint kinase in response to MMS-induced DNA damage during S phase, although not as strong a defect as rad32Δ cells (Figure 7). However, when tested for their ability to slow replication in response to DNA damage, cpt1Δ cells display a fully functional S-phase DNA damage checkpoint (Figure 8B). Therefore, ctp1Δ and the class III separation-of-function mutants define a resection- and signaling-independent role for MRN in the S-phase DNA damage checkpoint.
To dissect the role of the MRN complex in the S-phase DNA damage checkpoint, we created site-directed mutations in the Rad32 (Mre11) subunit that disrupt nuclease function or mimic human ATLD mutations. As a result, we have defined a function for MRN in S-phase checkpoint role that is independent of its Ctp1-dependent roles in double-strand break resection, constitutive repair, and checkpoint signaling. Specifically, we identified three classes of mutants: class I mutants are similar to wild type in all assays, class II mutants are similar to rad32 null mutants, and class III mutants behave like null mutants in all assays except the S-phase DNA damage checkpoint assay. In this respect, the class III mutants phenocopy ctp1 null mutants and genetically separate the Ctp1-dependent functions of MRN from its S-phase checkpoint function. This genetic separation of function demonstrates that at least one of the defining roles of MRN is independent of its well-studied involvement in DNA damage recognition and checkpoint signaling. Table 3 summarizes our results.
The class I alleles rad32-N122S and rad32-W215C were both designed to recapitulate human ATLD alleles (Stewart et al., 1999 ; Fernet et al., 2005 ). The sites mutated are conserved residues in conserved regions of the protein. However, the fission yeast alleles do not show the strong checkpoint phenotypes characteristic of the human alleles. Nonetheless, the fission yeast mutants do display mild HU sensitivity and mild spore viability defects, indicating that they somewhat impair MRN function (Figure 2). Furthermore, they display reduced MRN complex formation (Figure 5), consistent with the fact that they are thought to compromise Nbs1 binding in humans (Hopfner et al., 2001 ; Lee et al., 2003 ).
The class II alleles rad32-D25A, rad32-D65N, and rad32-H134N affect conserved nuclease residues and disrupt in vitro nuclease activity in the budding yeast protein (Furuse et al., 1998 ; Usui et al., 1998 ; Moreau et al., 1999 ). They are phenotypically similar to the null allele. However, two of them, rad32-D65N and rad32-H134N, are slightly more resistant to DNA damage than the null, suggesting they retain some residual MRN function (Figure 2). Furthermore, these alleles are functional in telomere maintenance (Figure 3). This apparent nuclease-independent function at the telomere is consistent with the recent demonstration that MRN nuclease activity is not required for ATM(Tel1) activation in mice, the presumptive function of MRN in telomere maintenance (Buis et al., 2008 ). The third allele, rad32-D25A, is indistinguishable for rad32Δ. Furthermore, it expresses poorly and fails to form MRN complexes (Figure 5), suggesting that the D25A mutation disrupts overall protein structure, consistent with behavior of the budding yeast protein (Krogh et al., 2005 ). The fact that the D25A mutation leads to significant reduction in Nbs1 expression suggests that lack of Mre11 binding destabilizes Nbs1, in agreement with reduced Xrs2 expression in budding yeast D25A cells (Krogh et al., 2005 ). The fact that our stable nuclease-dead alleles behave mostly as nulls contrasts with the behavior of the analogous alleles in budding yeast, which retain significant resistance to DNA damage (Krogh et al., 2005 ). This result highlights differences between the roles of fission yeast MRN and the budding yeast MRX in DNA damage metabolism, which are mirrored by differences in their roles in replication slowing (Chahwan et al., 2003 ; Andrews and Clarke, 2005 ).
The class III alleles display a separation-of-function phenotype; they disrupt all of the functions of MRN except for its checkpoint-dependent role in slowing replication in response to DNA damage. One of the class III alleles, rad32ΔC, is analogous to a human ATLD allele (Stewart et al., 1999 ). Rad32ΔC deletes the C-terminal DNA binding domains of Rad32. A plausible explanation for the phenotype of rad32ΔC cells is that Rad32ΔC compromises Ctp1 recruitment to sites of MRN function (see below). The phenotypes of the other class III allele, rad32-H134L/D135V, are harder to explain. This allele, originally described as mre11-3 in budding yeast, affects two conserved nuclease residues and has been characterized as nuclease dead in the context of the human protein (Arthur et al., 2004 ). Nonetheless, it retains DNA damage recognition functions in human cells and replication slowing functions in fission yeast (Arthur et al., 2004 ; Figure 5). In particular, it behaves differently from the other nuclease-dead alleles in fission yeast, including one allele, rad32-H134N, that affects one of the same residues. One possibility is that there is sufficient binding of manganese in the Rad32-H134L/D135V active site to retain residual nuclease function that would support the S-phase DNA damage checkpoint role, but not other roles, even though that level of nuclease activity would be immeasurable in an in vitro assay (Lewis et al., 2004 ). In a similar vein, Rad32-H134L/D135V may disrupt only a subset of the various endo- and exonuclease activities of MRN, perhaps disrupting its 3′ exonuclease activity, the activity that has been assayed, but not its endonuclease activity (Williams et al., 2008 ). Alternatively, it is possible that Rad32-H134L/D135V disrupts all nuclease activity, but the Rad32-H134N and the other nuclease-dead mutations disrupt additional functions, such as ATP binding or DNA binding, required for replication slowing. Recently, the Russell laboratory has made a rad32-H134S mutant, which is not highly sensitive to DNA-damaging agents (Williams et al., 2008 ). They suggested that the nuclease active site in this mutant still forms Rad32/Mre11 dimers and that dimerization of Mre11 is critical for MRN function. It is possible that rad32-H134L/D135V is able to form Rad32 dimers, whereas rad32-H134N cannot.
The class III alleles phenocopy ctp1Δ. Ctp1, the homologue of budding yeast Sae2 and human CtIP, is an MRN cofactor that is recruited to sites of DNA damage by MRN and is required for MRN-dependent resection at double-stand breaks (Clerici et al., 2005 ; Limbo et al., 2007 ; Sartori et al., 2007 ; Akamatsu et al., 2008 ). The similarity between ctp1 and the class III alleles suggests that these alleles interfere specifically with the Ctp1-dependent functions of MRN. Ctp1 and MRN have not been shown to directly interact, but MRN is required for Ctp1 to be recruited to a double-strand break, leading to the suggestion that the two may interact on DNA (Limbo et al., 2007 ). That Rad32ΔC lacks the C-terminal DNA binding domain suggests that this region may be involved in Ctp1 interaction. However, that the rad32ΔC mutant has a less severe meiotic phenotype but more severe signaling phenotype than ctp1Δ, and that both class III alleles lack a telomere function that ctp1Δ retains, suggests that the class III alleles effects other functions of MRN, beyond its Ctp1-dependent functions.
The alleles modeled on the human ATLD alleles behave differently in our assays, with two alleles, rad32-N122S and rad32-W215C, being class I wild-type alleles and one allele, rad32ΔC, being a class III separation-of-function allele. However, the class I alleles do show some defects, particularly in complex formation, and the class III allele does show some resistance to DNA damage and HU. These results suggest that the ATLD alleles are not qualitative separation-of-function alleles but rather hypomorphs of various strengths that disrupt some MRN functions but not others due to threshold effects. In vitro biochemical analyses of these alleles have lead to a similar conclusion (Paull, personal communication). It is worth noting that the inferred separation-of-function phenotype of the human ATLD alleles—checkpoint defective but constitutive repair competent—is the opposite of the separation-of-function phenotype we observe for our class III alleles.
Phenotypic analysis of our rad32 alleles provides a basis for speculation on the mechanistic role of MRN in the S-phase DNA damage checkpoint. Our first conclusion is that the role of MRN in replication slowing is independent from its role in checkpoint signaling. We draw this conclusion from the fact that the class III separation-of-function mutants have strong signaling defects and yet display fully functional replication slowing. Furthermore, one of the class II mutants, rad32-D65N, is as defective as ctp1Δ for signaling, but it fails to slow, showing that there is no correlation between signaling and slowing. We have recently reached the same conclusion based on the constitutive activation of checkpoint signaling not suppressing the slowing defect in nbs1Δ cells (Willis and Rhind, unpublished data). This conclusion is also consistent with two observations from human NBS cells. First, these cell display slowing defects in conditions in which they seem to activate ATM normally (Difilippantonio et al., 2005 ; Kanu and Behrens, 2007 ). Second, they degrade Cdc25A, a downstream target of the checkpoint, demonstrating that signaling to that branch of the S-phase DNA damage checkpoint is intact (Yazdi et al., 2002 ).
Our second conclusion is that the role of MRN in replication slowing is independent from its resection activity and its role homologous recombination. The conclusion that MRN does not need to resect DNA to fulfill its role in replication slowing is supported by the observation that Ctp1, which is required for resection, is not required for replication slowing (Clerici et al., 2005 ; Limbo et al., 2007 ; Sartori et al., 2007 ; Akamatsu et al., 2008 ; Figure 6B). The conclusion that homologous recombination is not required for replication slowing is supported by several lines of evidence. First, our class III separation-of-function alleles are synthetically lethal with rad2Δ (Table 2). Synthetic lethality with rad2Δ is characteristic of mutations that disrupt homologous recombination (Symington, 1998 ). In addition, Ctp1 is also required for homologous recombination (Limbo et al., 2007 ). Finally, we have recently shown that Rad51-dependent homologous recombination is not required for checkpoint-dependent replication slowing in fission yeast (Willis and Rhind, 2009 ).
Our third conclusion is that the role of MRN in replication slowing may be independent from its nuclease activity. Conclusions regarding the role of nuclease activity of MRN in replication slowing are complicated by the fact that three of the nuclease-dead mutants, rad32-D25A, rad32-D65N, and rad32-H134N, are class II alleles and fail to slow replication, but a fourth nuclease-dead mutants, rad32-H134L/D135V, is a class III allele and is proficient for slowing. As described above, these results indicate that there are biochemical differences between the class II nuclease-dead alleles and rad32-H134L/D135V, but whether the later allele has some residual nuclease activity or the former alleles disrupt functions in addition to nuclease activity awaits a thorough biochemical characterization of all four proteins.
What role does this leave for MRN in checkpoint-dependent slowing of replication? It has been suggested that the slowing of replication is a consequence of DNA repair (Rhind and Russell, 2000 ). Our results support this conclusion, in that the role of MRN is downstream of signaling, but the suggest that this repair is some MRN repair function other than homologous recombination. We have recently shown that Mus81 and Rqh1 are both required downstream of checkpoint signaling in the S phase DNA damage checkpoint (Willis and Rhind, 2009 ). Both of these proteins are implicated in replication fork metabolism: Mus81 as an endonuclease that can cleave stalled forks and promote sister recombination and Rqh1 as a helicase involved in the regulation of fork regression (Murray et al., 1997 ; Stewart et al., 1997 ; Roseaulin et al., 2008 ). MRN can tether DNA and has been suggested to be involved in coordinating recombinational repair at stalled forks (Lisby and Rothstein, 2004 ; Lisby et al., 2004 ). Because stalled forks do not present double-stand ends, the role of MRN at stalled forks may not involve its resection activity and thus may be independent of Ctp1. In such a role, it may support the functions of Mus81 and Rqh1 in checkpoint-dependent slowing of replication. However, once forks collapse, double-strand breaks are produced, and the resection activity of MRN may be required for homologous-recombinational restart mechanisms (Trenz et al., 2006 ). This distinction between stalled forks that are stabilized by the checkpoint and collapsed forks that need to be restarted by homologous recombination may explain why the Ctp1-dependent resection activity of MRN is dispensable for checkpoint-dependent replication slowing but required for checkpoint-independent repair of spontaneous S-phase damage (Trenz et al., 2006 ; Wen et al., 2008 ).
Regardless of the mechanism of the role of MRN in the S-phase DNA damage checkpoint, our results demonstrate that MRN does more in the checkpoint than simply recognize DNA damage and activate checkpoint signaling. Furthermore, they demonstrate that the checkpoint role of MRN is independent from its resection and homologous recombination activities. Therefore, MRN has at least two biochemically distinct repair activities. Understanding these diverse activities will be essential for understanding the role MRN plays in protecting the genome from both exogenous and endogenous DNA damage and in preventing cancer.
We thank P. Russell (The Scripps Research Institute), P. Jeggo (University of Sussex), and Y. Tsutsui (Osaka University) for providing strains used in this work, T. Wang (Stanford University) for the anti-Cds1 antibody, T. Nakamura (University of Illinois, Chicago) for the Tas1 plasmid; and members of the Rhind laboratory, especially Prasanta Patel, for helpful discussions and experimental assistance. This work was funded by National Institutes of Health grant GM-069957 (to N. R.).
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E08-09-0986) on February 11, 2009.