The Shu complex promotes error-free tolerance of alkylation-induced base excision repair products.

Here, we investigate the role of the budding yeast Shu complex in promoting homologous recombination (HR) upon replication fork damage. We recently found that the Shu complex stimulates Rad51 filament formation during HR through its physical interactions with Rad55-Rad57. Unlike other HR factors, Shu complex mutants are primarily sensitive to replicative stress caused by MMS and not to more direct DNA breaks. Here, we uncover a novel role for the Shu complex in the repair of specific MMS-induced DNA lesions and elucidate the interplay between HR and translesion DNA synthesis. We find that the Shu complex promotes high-fidelity bypass of MMS-induced alkylation damage, such as N3-methyladenine, as well as bypassing the abasic sites generated after Mag1 removes N3-methyladenine lesions. Furthermore, we find that the Shu complex responds to ssDNA breaks generated in cells lacking the abasic site endonucleases. At each lesion, the Shu complex promotes Rad51-dependent HR as the primary repair/tolerance mechanism over error-prone translesion DNA polymerases. Together, our work demonstrates that the Shu complex's promotion of Rad51 pre-synaptic filaments is critical for high-fidelity bypass of multiple replication-blocking lesion.

INTRODUCTION

DNA double-strand breaks (DSBs) can arise from both endogenous sources such as cellular metabolites and DNA replication errors as well as from exogenous sources such as radiation and chemical exposure. Once a DSB occurs, the DNA ends can be processed for religation in a repair pathway called non-homologous end joining (NHEJ). Alternatively, the DSB ends are resected to create 3′ single-stranded DNA (ssDNA) overhangs that invade a homologous sequence as a repair template in a process called homologous recombination (HR). Use of the different DSB repair pathways depends on many different factors including the cell cycle stage (1), the DNA damage source and even the spatial relationship between the damaged DNA and its potential homologous donor (2,3). When DSBs are not accurately repaired then mutations, chromosomal rearrangements and even cell death can occur. Furthermore, mutations in genes important in DSB repair are associated with cancer predisposition and various syndromes including Bloom, Werner, Fanconi Anemia and Nijmegen breakage syndromes.

In budding yeast Saccharomyces cerevisiae, HR is the primary mechanism in the repair of DSBs and relies on Rad51 filament formation on the 3′ ssDNA ends (4). Not surprisingly, many proteins have evolved to regulate Rad51 filament formation and disassembly in order to temporally and spatially control HR (5–10). Rad51 filament formation depends upon the activities of Rad52 and its epistasis group of proteins, which include the Rad51 paralogues, Rad55–Rad57 (10–13). The precise mechanism of how Rad52 and Rad55–Rad57 enable Rad51 filaments to form is not completely understood. Recently it was uncovered that the Shu complex (also referred to as the PCSS complex) contains additional Rad51 paralogues (14–16) and is conserved in fission yeast and humans (17,18). The budding yeast Shu complex consists of four proteins, Shu1, Shu2, Psy3 and Csm2, where both Psy3 and Csm2 are structurally similar to each other as well as to Rad51 (14–17,19). We find that the Shu complex physically interacts with Rad51 through Rad55–Rad57, and acts to synergistically stimulate Rad51 filament formation when combined with the Rad51 regulators Rad52 and Rad55–Rad57 (10,20,21). In support of this model, in a direct repeat recombination assay, disruption of the Shu complex leads to a 50% reduction in spontaneous Rad51-mediated repair and a corresponding increase in Rad51-independent single-strand annealing (10,20). Other groups have found that Shu complex disruption leads to gross chromosomal rearrangements and to increased mutation frequency, which depends upon the translesion synthesis polymerases during post-replicative repair (19,21–23).

While disruption of many of the HR genes involved in Rad51 filament formation such as RAD52 or RAD55-RAD57 lead to sensitivity to a broad range of DNA damaging agents, Shu gene disrupted cells are primarily sensitive to DNA damage induced by methylmethane sulfonate (MMS). For example, whereas rad52Δ and rad55Δ cells are sensitive to ionizing radiation (IR), ultra-violet (UV) light and hydroxyurea (HU), Shu complex mutants are resistant to all these reagents (19,20,22,24). This observation has led us to propose that the Shu complex's role in HR may be restricted to the repair of specific types of DNA lesions, especially those that are created by MMS. However, what the nature of these MMS-induced lesions are and what the Shu complex role is in their repair remain unknown.

MMS is a DNA alkylating agent that methylates DNA, leading to adducts that slow replication forks, accumulation of ssDNA gaps behind the replication fork and potentially replication fork collapse (25). MMS specifically leads to N3-methyladenines (N3-MeA) and N7-methylguanines (N7-MeG) where N3-MeAs are thought to be directly toxic lesions that lead to altered replication fork dynamics (26,27). Other lesions can also occur upon MMS exposure including N3-methyl cytosines, which can be a source of hyper-mutability in ssDNA (28). DNA alkylation damage, such as that created by MMS exposure, is primarily repaired through the base excision repair (BER) pathway (Figure 1A; (29)). During BER in budding yeast, N3-MeA and N7-MeG DNA adducts are removed by the DNA N-glycosylase Mag1, which leaves an apyrimidinic/apurinic (AP) site in the DNA (30–33). Subsequently the Apn1 and Apn2 endonucleases nick the phosphodiester bond at the AP site resulting in a 5′ deoxyribose (5′-dRP) (Figure 1A, left side) (34–36). Then, DNA polymerase ϵ fills the gap and the flap endonuclease Rad27 removes the flap containing the 5′-dRP (37). Subsequently, the DNA backbone is sealed by DNA Ligase 1 (Lig1). In an alternative and less frequently used BER sub-pathway in budding yeast, the DNA glycosylase and AP lyase Ntg1 and its paralogue Ntg2 can nick the Mag1-generated AP site resulting in a 3′ deoxyribose (3′-dRP) (Figure 1A, right side)(34). This DNA intermediate can either be further processed by Apn1 and Apn2 or alternatively excised by the flap-endonuclease Rad1-Rad10. Subsequently, DNA polymerase ϵ fills the gap at the DNA damage site and then Lig1 completes repair. Although MMS-induced DNA damage is primarily repaired via BER, if the DNA damage occurs during replication then partial processing of the damage can lead to replication fork stalling, the accumulation of ssDNA gaps and fork collapse (38). It is these types of lesions that are then targeted for repair/bypass by HR. How HR responds to BER intermediates encountered by a replication fork during DNA replication is not well understood but is thought to depend upon the DNA repair proteins Mre11-Rad50-Xrs2 as well as Rad51 and Rad52 (38).

Figure 1.

Schematic of base excision repair (BER) and the Shu complex primary sensitivity to methylmethane sulfonate (MMS) induced DNA damage. (A) Cartoon of BER pathway adapted from (51). The black arrows indicate the major pathway. (B) The indicated strains were 5-fold serially diluted onto YPD medium, YPD medium containing 0.006% or 0.02% MMS, SC medium or SC medium containing 40 μg/ml cisplatin. The plates were incubated for two days at 30°C and photographed. (C) Same as (B) except that the YPD plate was exposed to 70 Gray ionizing radiation (IR) or 50 J/m2 ultra-violet (UV), or contained 100 mM hydroxyurea (HU) or 3 mM H2O2. (D) Same as (B) except that 2 mM etoposide was added to the YPD medium. (E) Same as (B) except that 50 μg/ml CPT was added to the YPD medium.

Here, we show that the Shu complex mutant cells are additionally sensitive to high doses of camptothecin (CPT), which is a Top1 inhibitor that results in ssDNA breaks that lead to replication fork collapse and one-ended DSBs. Since BER of MMS-induced DNA lesions can also produce ssDNA breaks, this finding raised the question as to what specific DNA lesion(s) produced by MMS are acted upon by the Shu complex. To pinpoint which DNA intermediates produced during BER are acted on by the Shu complex, we generated different BER mutants that block repair of alkylated DNA at specific intermediates, such as abasic sites or 3′-dRPs, and screened for synthetic sensitivity with disruption of the Shu complex gene CSM2. Using this genetic approach, we find that upon exposure to MMS, csm2Δ cells are synthetically sick when combined with mag1Δ or apn1Δ apn2Δ disruptions at low doses of MMS, and that these cells exhibit elevated mutagenesis, altered recombination, impaired progression through S phase, and a heightened reliance upon translesion synthesis DNA polymerases. In addition, we observe a synthetic interaction in csm2Δ rad27Δ cells, and find that this synthetic growth defect is suppressed by over-expression of human DNA polymerase β (Polβ). Together, our work expands upon the types of lesions the Shu complex is known to act on in the context of both endogenous and MMS-induced replicative stress and provides evidence that the lethality caused by MMS is due to a spectrum of lesions that exist concurrently in a replicating cell.

MATERIALS AND METHODS

Strains, plasmids and media

The strains and plasmids utilized here are listed in Supplementary Table S1. All the yeast strains are RAD5+ and isogenic to W303 W9100-2D. Yeast crosses, transformations and tetrad dissections were conducted as described (39). Similarly, the media preparation was done as described except with twice the amount of leucine added (39). MMS (0.002%, 0.006%, 0.012%, 0.02%, Sigma 129925), hydroxyurea (100 mM HU; Sigma H8627), hydrogen peroxide (3 mM H2O2; Fisher H323-500), camptothecin (50 μg/ml CPT; Sigma C9911) and etoposide (2 mM; Sigma E1383) were added to YPD medium at the indicated final concentrations. Cisplatin (40 μg/ml; Sigma ALX-400-040-M250) was made in SC medium. CPT (10 mg/ml), cisplatin (100 mg/ml) and etoposide (50 mM/ml) stock solutions were re-suspended in di-methyl sulfoxide (DMSO). The primers used to make the specific gene knockouts are listed in Supplementary Table S2. APN1 and APN2 genes were disrupted with a hygromycin cassette (pFA6a-hphNT1) using the indicated S1 and S2 primers as described (40). NTG1 and NTG2 genes were disrupted with a clonat cassette (pFA6a-natNT2) using the indicated S1 and S2 primers as described (40). MAG1 were disrupted with histidine using the pFA6a-His3MX6 cassette and the indicated S1 and S2 primers as described (41). Polymerase chain reaction (PCR) colony amplification and genotyping of APN1, APN2, NTG1, NTG2, MAG1 was done using the indicated primers in Supplementary Table S2. The Polβ expression vector was made using the gateway cloning system and sub-cloned into the constitutive pAG414-GPD or galactose inducible pAG414-GAL-ccdB plasmid (Addgene plasmids #14144 and #14143 from Susan Lindquist), which contains TRP1 and Ampicillin selectable markers for yeast and bacteria, respectively.

Growth assays

Yeast strains were incubated in 5 ml YPD medium overnight at 30°C and then diluted to 5 ml OD600 0.2. The cultures were incubated for another 3 h at 30°C and equal cell numbers (5 μl of culture at OD600 0.5) were 5-fold serially diluted onto YPD medium or YPD medium containing 0.002%, 0.006%, 0.012% or 0.02% MMS as indicated. The plates were incubated at 30°C for 2 days and photographed. To examine yeast with Polβ over-expression, the indicated yeast strains were transformed with pAG414-GPD or pAG414-GAL empty or Polβ expressing plasmids. The yeast were grown on minimal SC medium lacking tryptophan overnight at 30°C and then 5-fold serially diluted on YPR + GAL (containing 2% raffinose and galactose) with or without 0.02% MMS or to YPD medium with 0.002% or 0.006% MMS.

Canavanine mutagenesis assay

Three to seven individual CAN1 colonies of wild-type (WT), csm2Δ, apn1Δ apn2Δ, csm2Δ apn1Δ apn2Δ, csm2Δ ntg1Δ ntg2Δ, ntg1Δ ntg2Δ, mag1Δ, csm2Δ mag1Δ, rad5Δ, csm2Δ rad5Δ, ubc13Δ, andcsm2Δ ubc13Δ were grown in 5 ml of YPD medium or YPD medium containing 0.002% MMS (18 h) overnight at 30°C to saturation. The cultures were diluted 100 000-fold and 120 μl was plated onto SC medium or diluted 10-fold and 250 μl was plated onto SC-ARG+CAN medium. The plates were incubated for 2 days at 30°C. Colonies were then counted to measure forward mutation frequency (SC-ARG+CAN) or total cell number (SC). For each trial, a median frequency was obtained and an average of each trial median was reported. The experiment was repeated three to four times with standard deviations plotted on the graph and significance determined by Student's t-test.

Recombination assay

Mitotic recombination assays were performed as described (20). Briefly, nine independent colonies for each strain were grown up overnight at 30°C in 2 ml of SC with or without MMS before being diluted to OD600 0.1. A total of 250 μl of this OD600 0.1 culture was plated onto SC-LEU plates to capture recombination events. The OD600 0.1 culture was further diluted 1:1000 before 250 μl were plated onto SC plates to calculate viable cells. After two days of growth, colonies were counted and the SC-LEU plates were replica plated onto SC-URA and 5′-FOA to observe Rad51-dependent and independent recombination events, respectively. Recombination rates and standard deviations were calculated as described (42) using the online FALCOR analysis tool (43).

Microscopic analysis

Cells were inoculated into 5 ml SC+ADE as described (20) and grown up overnight at room temperature. Cells were diluted 1:10 and allowed to grow for 3 h prior to imaging. Cells were imaged as an 11-step Z-stack with 0.3 µM per step. RAD52-YFP was visualized at 600 ms per image and RFA1-CFP was visualized at 500 ms. The results shown are the average of three individual experiments where at least 100 cells were imaged per strain per trial.

Protein level analysis

Logarithmically growing cells were arrested with 10 μmol alpha factor for 3 h, then released into room temperature YPD. At the indicated time points, equal cell numbers (1 OD600) was removed and TCA prepped and run on a 10% SDS-page gel as previously described (44) where the Kar2 (Santa Cruz sc-33630) and HA antibodies (Roche 12CA5) were used at 1:1000 dilution.

Pulse-field gel electrophoresis analysis

The 50 ml cultures of each indicated strain at OD600 1 were G2-arrested for 3 h with 1.5 μg/ml nocodazole. After 3 h, the cells were pelleted and resuspended in 40 ml of phosphate buffered saline (PBS) with 0.1% MMS at 30°C with shaking for 15 min. At 15 min, 40 ml of 10% sodium thiosulfate was added to each culture to inactivate the MMS, and then the cells were pelleted, washed once in YPD medium, and then resuspended in 25 ml YPD containing 1.5 μg/ml nocodazole. A total of 125 μl of 1.5 mg/ml nocodazole was added at 2.5 and 5 h to maintain G2 arrest. Equal cell numbers (3 OD600) were harvested just prior to MMS treatment for the control sample, and at T = 0, 1, 2, 4, 6 and 8 h following MMS exposure. After harvesting, the cells were washed once in ice cold buffer V (100 mM EDTA, 50 mM Tris pH 7.4) and kept on ice until being made into plugs for PFGE analysis. To make plugs, the cells were briefly warmed at 42°C, resuspended in 100 µl of 0.5% agarose with zymolyase and cast using a biorad plug mold. Plugs were incubated in buffer V for 30 min at 37°C to allow cell wall digestion and then incubated overnight in 1% Tween with 1 mg/ml Proteinase K to lyse cells. Alternatively, for the S-phase progression experiments, logarithmically growing cells were arrested by alpha factor addition for 3 h, released into fresh YPD for 10 min and then treated with 0.033% MMS for 1 h before being washed and resuspended in YPD to start the time course experiment as previously described (45). PFGE was carried out with a CHEF Mapper (BioRad) using the following conditions: 10–90 s pulses ramped linearly for 24 h at 6 V/cm using 120° pulse angle at 14°C.

RESULTS

Disruption of the Shu complex genes lead to sensitivity to MMS, CPT and cisplatin but not to DNA damaging agents such as IR, HU, UV, H2O2 or etoposide

To determine the types of DNA lesions that the Shu complex responds to we sought to directly compare the sensitivity of Shu complex mutants (shu1Δ, shu2Δ, csm2Δ and psy3Δ) and the Rad51 paralogue rad55Δ, to a broad range of DNA damaging agents (Figure 1B–E). Using this approach, DSBs are predicted to form after exposure to IR, MMS (leading to replication fork slowing and collapse or DSBs derived from processing closely opposed lesions), HU (which limits dNTP pools leading to replication fork stalling and collapse), CPT and etoposide (which is a Top2 inhibitor that leads to single- and double-strand breaks). In addition, we analyzed cells for sensitivity to acute UV exposure (primarily thymine dimers), cisplatin (principally intra-strand crosslinks) and hydrogen peroxide (oxidative damage). By serial dilution, as a positive control we confirmed that Shu complex mutants are primarily sensitive to MMS (19,20,22,24), which can cause DSBs and replicative collapse, and not to direct DSB-inducing agents such as IR or etoposide, or replicative stalling agents such as HU (Figure 1B–D) (19,22,24). Interestingly, Psy3 (Platinum Sensitivity 3) was originally identified and named in a genome-wide screen for its sensitivity to oxaliplatin and cisplatin, both drugs that create inter- and intra-strand DNA crosslinks (46). Here, we demonstrate that the rest of the Shu complex mutants, shu1Δ, shu2Δ and csm2Δ are also identically sensitive to cisplatin (Figure 1B).

We do not observe sensitivity of Shu complex mutants to UV or H202 (Figure 1C), although we can not rule out that chronic exposure to UV, which increases reliance upon bypass mechanisms, would cause sensitivity in a Shu complex mutant cell. Finally, we find that rad55Δ cells display increased sensitivity to MMS as well as to a broader range of DNA damaging agents (i.e. cisplatin, IR, HU, etoposide, CPT) than the Shu complex mutants (Figure 1B–E). Interestingly, we also find that the Shu complex mutants exhibit a mild sensitivity to high doses of CPT, a Topoisomerase I inhibitor that leads to the formation of excess ssDNA breaks that are channeled into DSBs at a replication fork (Figure 1E). This finding suggests that the Shu complex may contribute to the repair of one-ended DSBs formed by high doses of CPT at a replication fork. Previous work has suggested that the Shu complex responds to the direct replicative blocks produced by MMS, such as N3-MeAs, but the finding that Shu complex mutants are sensitive to the ssDNA breaks produced by CPT suggest that the Shu complex could be capable of promoting repair of ssDNA breaks produced as intermediates during BER (47).

Upon MMS exposure, the Shu complex mutants exhibit synthetic sensitivity when combined with mag1Δ

To understand the role of the Shu complex in the tolerance of MMS-induced DNA lesions, we combined Shu complex mutants with disruption of BER genes and tested these mutants for synthetic sensitivity upon MMS exposure. Since Mag1 is necessary for processing DNA damage into AP sites and its disruption would block downstream processing of the MMS-induced DNA damage, we began our analysis with mag1Δ cells, which accumulate methylated DNA adducts such as N3-MeAs upon MMS treatment (Figure 2A). When compared to WT, mag1Δ cells are sensitive to 0.012% MMS whereas csm2Δ, shu1Δ, shu2Δ and psy3Δ cells are sensitive to 0.02% MMS (Figure 2B, Supplementary Figure S1). Consistent with a previous study (47), we observe a synthetic sensitivity when mag1Δ is combined with disruption of any of the Shu complex genes (csm2Δ, shu1Δ, shu2Δ, psy3Δ)(0.006% MMS, Figure 2B, Supplementary Figure S1). Furthermore, the synthetic sensitivity on low dose MMS of the double mag1Δ shuΔ mutants is equivalent regardless of which Shu complex component is disrupted (Figure 2B, Supplementary Figure S1). The Shu complex functions as an obligate heterotetramer where Csm2 confers DNA binding activity and most of the physical interactions between the Shu complex and other HR proteins occur through this subunit (15,16,20,21). Therefore, we focused on csm2Δ cells for the rest of our analysis.

Figure 2.

In the absence of the DNA glycosylase MAG1, Shu complex disrupted cells are hypersensitive to MMS-induced alkylation damage. (A) Cartoon demonstrating what DNA processing step is blocked by mag1Δ disruption. (B) The indicated yeast strains were 5-fold serially diluted onto YPD medium or YPD medium containing MMS (0.002%, 0.006%, 0.012% and 0.02%) and incubated for two days at 30°C and subsequently photographed. (C) Mutation frequencies of WT, csm2Δ, mag1Δ and csm2Δ mag1Δ cells were determined at the CAN1 locus both spontaneously and upon MMS exposure. Three experiments with five separate isolates were analyzed and averaged. The graph represents the average of the medians with standard deviations plotted. Significance was determined by t-test. (D) Schematic of a direct-repeat recombination assay, where Rad51-dependent gene conversion (GC; URA3+ LEU2+) and Rad51-independent single-strand annealing (SSA; URA3- LEU2+) repair can be measured. Total recombination rates of the indicated strains are shown before and after MMS treatment. (E) The levels of Rad51-dependent (GC) and Rad51-independent (SSA) repair for each condition shown in (D). For (D) and (E), each condition was repeated 3 to 5 times and the average recombination rates were calculated as described in the methods section, are graphed.

csm2Δ mag1Δ cells exhibit elevated mutation frequencies and recombination rates upon MMS exposure

Disruption of the Shu complex mutants results in increased mutation frequencies that are measured by analyzing cells for mutations in the arginine permease gene (CAN1)(19,22). WT CAN1 cells die when grown in the presence of canavanine; however, mutations in CAN1 prevent this lethality and can readily be scored by plating on medium containing canavanine. To determine how the un-excised MMS damage contributes to the increased mutations observed in csm2Δ cells, we analyzed the csm2Δ mag1Δ cells for spontaneous and MMS-induced mutations in the CAN1 gene (Figure 2C). csm2Δ cells, like other HR mutants, have an elevated basal mutation frequency compared to WT cells (Figure 2C) that is exacerbated by growth in low dose MMS (Figure 2C) (19,22). Interestingly, we do not observe an increase in the mutation frequency of mag1Δ cells compared to WT in the absence of MMS, although a modest increase is evident in MMS-treated mag1Δ cells (Figure 2C, P < 0.05). Although spontaneous mutation frequencies in untreated csm2Δ mag1Δ are similar to the csm2Δ single mutant (Figure 2C; P > 0.5), upon MMS exposure we observe a synergistic increase in mutation frequency comparing the csm2Δ and csm2Δ mag1Δ mutants (Figure 2C, P < 0.05 comparing csm2Δ and csm2Δ mag1Δ after MMS treatment). Due to the high sensitivity of csm2Δ mag1Δ cells, chronic low-dose MMS exposure was used in liquid culture experiments to maintain viability of the double mutants. These results suggest that in the absence of MAG1, the Shu complex contributes to high-fidelity bypass of alkylated DNA.

To test the hypothesis that MMS lesions are bypassed using high-fidelity mechanisms such as HR, rather than error-prone TLS, we examined MMS-treated mag1Δ cells for increased recombination rates (48,49). To do this, we utilized a direct repeat recombination assay that enabled us to measure total recombination rates as well as the type of recombination events that occurred. As pictured (Figure 2D), recombination events that produce a LEU2 allele can occur via Rad51-dependent gene conversion (GC), which retains the URA3 allele, or by Rad51-independent single-strand annealing (SSA) where the intervening URA3 allele is deleted. SSA is usually considered to require a break, as the DNA needs to be resected back to reveal homology between the two LEU2 alleles. Alternatively, it has been shown using a similar assay that SSA-like colonies can be generated by a Rad51-independent, inter-sister recombination event (50). Consistent with our hypothesis that HR bypasses the majority of the MMS-induced lesions, we find that MMS treated WT and mag1Δ cells exhibit a 3- to 4-fold induction in recombination rates (Figure 2D, P value < 0.05). Interestingly, the MMS treated WT and mag1Δ cells have a significant bias toward Rad51-mediated GC that is dependent upon CSM2 (Figure 2E, compare WT to csm2Δ, P < 0.005 and mag1Δ to csm2Δ mag1Δ P < 0.001). This unexpected finding indicates that MMS-induced lesions are preferentially channeled into Rad51-dependent GC, whereas spontaneous lesions do not show any bias in the HR pathway (Figure 2E, compare untreated and treated WT). Surprisingly, we also find that the accumulation of lesions in MMS-treated mag1Δ cells also leads to a significant increase in SSA relative to treated WT cells (Figure 2E, P value < 0.05). Together, these data suggest that the majority of MMS lesions encountered by a replication fork are channeled into high-fidelity mechanisms of damage tolerance and bypass.

CSM2 disrupted cells exhibit a synthetic sensitivity with apn1Δ apn2Δ upon MMS exposure

Since we observe a synthetic MMS sensitivity between the Shu complex mutant csm2Δ and mag1Δ, we asked whether this phenotype is due to a general defect in BER, an accumulation of methylated DNA, or to the accumulation of specific BER intermediates. To address this question, we first determined if disruption of the down-stream AP site processing factors, APN1 or APN2, also exhibit elevated sensitivity when combined with CSM2 disruption. Upon MMS exposure, deletion of the APN1 and APN2 nucleases results in accumulation of AP sites, 3′-dRPs and gapped ssDNA (Figure 3A). Interestingly, APN1 disruption in a csm2Δ cell leads to a synthetic sensitivity upon 0.006–0.012% MMS exposure (Figure 3B). In contrast, csm2Δ apn2Δ cells exhibit a very modest synthetic sensitivity at 0.02% MMS (Figure 3C). Since both apn1Δ and apn2Δ act independently to process abasic sites, their combined deletion creates a stronger MMS sensitivity (33), which we also observe (Figure 3D). When CSM2 is deleted in apn1Δ apn2Δ cells there is a further increase in sensitivity, with cells unable to survive at 0.002% MMS (Figure 3D). Importantly, we find that MAG1 is epistatic to APN1 and APN2 since csm2Δ mag1Δ cells exhibit the same growth as a csm2Δ mag1Δ apn1Δ apn2Δ quadruple mutant at 0.002% MMS (Figure 3D). These results are consistent with Mag1 functioning upstream to Apn1 and Apn2 (33). Furthermore, we find that not only are csm2Δ apn1Δ apn2Δ cells more sensitive to MMS damage but they also exhibit increased spontaneous mutation frequencies (Table 1; P ≤ 0.0007 comparing csm2Δ to csm2Δ apn1Δ apn2Δ). Due to the severe MMS sensitivity of apn1Δ apn2Δ csm2Δ cells, we are unable to analyze mutation frequencies upon MMS exposure. These results suggest that accumulation of AP sites, 3′-dRPs and/or ssDNA gaps that occur in apn1Δ apn2Δ cells are both toxic and mutagenic in the absence of the Shu complex, thereby increasing the hypermutable effects of the ssDNA that is formed (28,38,51,52).

Figure 3.

CSM2 disrupted cells are sensitive to AP sites and/or 3′-dRPs as demonstrated by the synthetic sensitivity of csm2Δ with apn1Δ apn2Δ on MMS containing medium. (A) Cartoon demonstrating what DNA processing steps are blocked by apn1Δ, apn2Δ disruption and accumulation of AP sites and 3′-dRPs. (B and C) The indicated yeast strains were 5-fold serially diluted onto rich medium (YPD) or YPD medium containing the indicated dose of MMS. The plates were incubated at 30°C for two days and photographed. (D) In CSM2 disrupted cells, Mag1 is epistatic to Apn1 and Apn2. The indicated yeast strains were 5-fold serially diluted onto YPD or YPD with MMS as described for (B and C). (E) WT, apn1Δ apn2Δ, csm2Δ and apn1Δ apn2Δ csm2Δ cells expressing Rad52-YFP and Rfa1-CFP were analyzed by fluorescent microscopy for Rad52 and Rfa1 focus formation (indicated by a white arrow). A single z-plane is shown for WT and apn1Δ apn2Δ csm2Δ cells. (F) The percentage of WT, apn1Δ apn2Δ, csm2Δ and apn1Δ apn2Δ csm2Δ S/G2/M cells exhibiting a spontaneous nuclear Rad52-YFP or Rfa1-CFP focus. Three individual experiments totaling 300–400 cells were analyzed per genotype and standard error of the means plotted. (G) The cell cycle distribution of the indicated genotypes in (E) were measured as a percentage of G1 cells (unbudded) and S/G2/M cells (budded) from the images acquired in (E).

Table 1.

Spontaneous mutation frequencies of the indicated strains were determined at the CAN1 locus

Genotype	Frequency (Mutations/Cell)	Fold Change
WT	2.97 × 10−7	1
ntg1Δ ntg2Δ	4.01 × 10−7	1.4
apn1Δ apn2Δ	3.02 × 10−6	10.2
csm2Δ	4.05 × 10−6	13.6
csm2Δ ntg1Δ ntg2Δ	3.62 × 10−6	12.2
csm2Δ apn1Δ apn2Δ	3.72 × 10−5	125.4
ntg1Δ ntg2Δ apn1Δ apn2Δ	3.2 × 10−5	108
csm2Δ ntg1Δ ntg2Δ apn1Δ apn2Δ	3.19 × 10−4	1074.6

We performed fluorescent microscopy to analyze Rfa1-CFP and Rad52-YFP expressing cells for focus formation indicative of increased ssDNA and HR, respectively, in untreated WT, apn1Δ apn2Δ and csm2Δ apn1Δ apn2Δ cells, as loss of APN1 and APN2 results in elevated ssDNA and recombination (28,38,51,52). Indeed, we find Rad52-YFP and Rfa1-CFP foci are significantly elevated in untreated apn1Δ apn2Δ budded-cells (indicative of S/G2/M phase) compared to WT (Figure 3E and F, P-value < 0.05). Additionally, further deletion of CSM2 in apn1Δ apn2Δ cells results in a significant increase in both Rfa1 and Rad52 foci (Figure 3E and F, P-value < 0.05). These results are consistent with slower resolution of Rad52 focus formation occurring in MMS-treated shu1Δ cells (19). Furthermore, since we observed an increase in the percentage of budded cells in untreated csm2Δ apn1Δ apn2Δ, the slow growth phenotype of this mutant (Figure 3D, YPD plate) may be due to a further accumulation of S/G2/M phase cells than observed in an apn1Δ apn2Δ double mutant (Figure 3G, P-value < 0.05).

In cells lacking APN1 and APN2, repair of MMS-induced damage during G2 occurs normally in the absence of the Shu complex

Shu complex proteins Csm2 and Psy3 preferentially bind to forked DNA intermediates (20) and its disruption leads to sensitivity to replicative stress, such as that induced by MMS (Figure 1), suggesting that the Shu complex predominately functions in repair and/or tolerance of replicative damage. One possibility is that the Shu complex primarily functions during S phase and, therefore, may be cell cycle regulated. To determine if the Shu complex protein Csm2 is cell cycle regulated, we used alpha factor to arrest Csm2-6HA tagged cells in G1 and then monitored Csm2 protein levels every 20 min after removal of alpha factor. Unlike the G2 cyclin, Clb2, Csm2 protein levels remained relatively constant throughout the time course (Figure 4A). Therefore, Csm2 protein levels are not cell cycle regulated. Similarly, Shu2 and Psy3 protein levels remain constant throughout the cell cycle (Supplementary Figure S3). However, we cannot rule out the possibility that the Shu complex's activity could be regulated by post-translational modification or levels of Shu1, which we have been unable to tag.

Figure 4.

CSM2 is important for repair of MMS-induced damage during S phase but not G2 phase. (A) Protein levels of Csm2, Clb2 (S/G2-phase control) and Kar2 (loading control) in asynchronous (ASN) cells or after release from G1 arrest by alpha factor. (B) Chromosome shattering and restitution in WT, apn1Δ apn2Δ and apn1Δ apn2Δ csm2Δ cells arrested in G2 cell cycle by nocodazole. Chromosomes are harvested at the indicated time points before and after treatment with MMS and separated by PFGE as described in the methods. (C) Chromosome restitution following progression through S phase in WT or csm2Δ cells treated with MMS.

To determine if Csm2 may have a role in the repair of MMS-induced DNA lesions during the G2/M cell cycle phase, we analyzed cells for chromosome restitution by pulse-field gel electrophoresis. First, we used nocodazole to arrest WT, apn1Δ apn2Δ or csm2Δ apn1Δ apn2Δ cells in G2 (‘C’ in Figure 4B). The arrested cells were then treated with 0.1% MMS for 15 min, leading to global chromosomal changes in apn1Δ apn2Δ cells due to the accumulation of gapped ssDNA, where T = 0 is taken directly after washing away the MMS (Figure 4B; (38)). The chromosomal changes are revealed as loss of discrete chromosomal bands and the appearance of slow moving DNA (SMD), which is thought to represent chromosomal DNA with excess ssDNA generated during repair (38,51). Chromosome restitution is evident by 8 h following MMS exposure (Figure 4B; compare WT and apn1Δ apn2Δ and results in (38,51)). In contrast to previous work showing that chromosome restitution depends on Rad52 and Rad51 (38), we find that the faint chromosome restitution visible in apn1Δ apn2Δ cells occurs independently of Csm2 during G2 across multiple trials (Figure 4B).

We next asked if the Shu complex would be required for the bypass of MMS-induced lesions during S-phase. WT and csm2Δ cells were arrested in G1 with alpha factor. Cells were released from alpha factor for 10 min before being treated with 0.033% MMS for 1 h (45). As cells progress into S-phase, their chromosomes are unable to be visualized by PFGE (Figure 4C, T = 0) due to accumulation in the well, but upon entry into G2, their chromosomes become visible (45). In WT cells, chromosome restitution can be visualized by 60 min (Figure 4C). In contrast, chromosome restitution in csm2Δ cells is delayed to ∼4 h (Figure 4C). These results are consistent with cell cycle profiles of an MMS-treated PSY3 mutant (22). These results indicate that while the Shu complex is present throughout the cell cycle, it functions predominately during S phase to bypass MMS-induced damage and has no significant contribution to the repair of similar lesions in G2 cells.

NTG1 and NTG2 generate toxic BER intermediates in the absence of APN1 and APN2

To determine if the Shu complex is responding to both abasic sites and ssDNA breaks in apn1Δ apn2Δ cells, we also deleted NTG1 and NTG2 to block the formation of ssDNA breaks (Figure 3A). Ntg1 and Ntg2 are bifunctional DNA glycosylases with associated apurinic/apyrimidinic (AP) lyase activities that cleave abasic sites to generate 3′-dRPs and their disruption results in AP sites being primarily processed by Apn1 and Apn2 (Supplementary Figure S2A). We sought to determine how the Shu complex responds to abasic sites, using various combinations of mutants. In contrast to csm2Δ mag1Δ or csm2Δ apn1Δ apn2Δ, we do not observe an elevated sensitivity in csm2Δ ntg1Δ (Supplementary Figure S2B), csm2Δ ntg2Δ (Supplementary Figure S2C) or csm2Δ ntg1Δ ntg2Δ cells upon MMS exposure (Supplementary Figure S2D). Nor do we observe an increase in spontaneous mutation frequencies in csm2Δ ntg1Δ ntg2Δ cells (Table 1; P ≥ 0.5 comparing csm2Δ to csm2Δ ntg1Δ ntg2Δ). Furthermore, we find that MAG1 is epistatic to NTG1 and NTG2 where a csm2Δ mag1Δ exhibits the same defect as a csm2Δ mag1Δ ntg1Δ ntg2Δ cell (Supplementary Figure S2D).

In the absence of APN1 and APN2, Ntg1 and Ntg2 cleave abasic sites to generate 3′-dRPs, which can be converted into toxic one-ended DSBs upon replication (Figure 5A) (35). Consistently, we observe a synthetic rescue of the MMS sensitivity of an apn1Δ apn2Δ double mutant in apn1Δ apn2Δ ntg1Δ ntg2Δ cells at 0.006% MMS (Figure 5B) (38). Interestingly, csm2Δ apn1Δ apn2Δ ntg1Δ ntg2Δ cells also grow better than csm2Δ apn1Δ apn2Δ cells at 0.002% MMS (Figure 5B), indicating that the Shu complex may act on abasic sites as well as one-ended DSBs that could be generated by replication forks encountering a ssDNA break. Quite surprisingly, although deletion of NTG1 and NTG2 suppresses the severe MMS sensitivity of apn1Δ apn2Δ cells by removing the toxic ssDNA breaks formed, we find that the spontaneous mutation frequency of apn1Δ apn2Δ ntg1Δ ntg2Δ cells is approximately 10-fold higher than apn1Δ apn2Δ cells (Table 1). These unexpected results suggest that the high mutation frequency caused by abasic sites is significantly less toxic than formation of ssDNA breaks. Furthermore, we find that disruption of CSM2 in apn1Δ apn2Δ ntg1Δ ntg2Δ cells can further increase the spontaneous mutation frequency over 1000-fold compared to WT, underscoring a novel and important new role for the Shu complex in the error-free bypass of abasic sites (Table 1). Consistent with these results, increased spontaneous mutation frequencies are also observed in ntg1Δ ntg2Δ apn1Δ rad52Δ cells (150-fold; (53)). To determine if the abasic sites are bypassed by Shu complex dependent HR, we measured the rate of spontaneous recombination using the direct repeat recombination assay described above (Figure 2D). In agreement with our finding that apn1Δ apn2Δ cells exhibit elevated Rad52 foci and slow growth in combination with csm2Δ (Figure 3F and D, respectively), we find a nearly 10-fold increased recombination rate in apn1Δ apn2Δ cells compared to WT (Figure 5B and C, P-value < 0.001). Over half the recombination increase relies upon CSM2 based on the 50% reduction in csm2Δ apn1Δ apn2Δ cells (Figure 5B, P-value < 0.005). Similarly, the elevated recombination of apn1Δ apn2Δ ntg1Δ ntg2Δ cells is suppressed by csm2Δ (Figure 5B, P-value < 0.01). Surprisingly, while deletion of NTG1 and NTG2 promotes increased MMS resistance of apn1Δ apn2Δ cells, we find that deletion of NTG1 and NTG2 do not rescue the elevated recombination rates of apn1Δ apn2Δ cells. Unfortunately, fluorescently tagged Rad52 and Rfa1 in apn1Δ apn2Δ ntg1Δ ntg2Δ and csm2Δ apn1Δ apn2Δ ntg1Δ ntg2Δ cells resulted in extremely poor growth and lethality, respectively. Therefore, we were unable to measure the levels of Rad52 and Rfa1 foci in these cells.

Figure 5.

CSM2 disrupted cells are sensitive to AP sites as demonstrated by the synthetic sensitivity of csm2Δ with apn1Δ apn2Δ ntg1Δ ntg2Δ on MMS containing medium. (A) Cartoon demonstrating what DNA processing steps are blocked by apn1Δ, apn2Δ, ntg1Δ and ntg2Δ disruption and accumulation of AP sites. (B) The indicated yeast strains were 5-fold serially diluted onto rich medium (YPD) or YPD medium containing the indicated dose of MMS. The plates were incubated at 30°C for two days and photographed. (C) The total recombination rates, as in Figure 2D. (D) The levels of Rad51-dependent and independent repair as in Figure 2E.

Next, we measured mutation frequencies in apn1Δ apn2Δ and apn1Δ apn2Δ ntg1Δ ntg2Δ cells in the presence or absence of CSM2. Deletion of CSM2 resulted in an approximate 10-fold increase in mutation frequencies (Table 1), indicating that the majority of spontaneous lesions are preferentially bypassed by high-fidelity rather than error-prone mechanisms. As expected, the reduced recombination rate resulting from CSM2 deletion is due to a decrease in Rad51-dependent gene conversion while Rad51-independent SSA remains constant (Figure 5D). While a csm2Δ cell exhibits a significant increase in Rad51-independent SSA (20)(Figure 5D, P-value < 0.001), we do not observe an increase in SSA if CSM2 is disrupted in apn1Δ apn2Δ or apn1Δ apn2Δ ntg1Δ ntg2Δ cells (Figure 5D). Perhaps, cells are unable to perform SSA or an SSA-like event in a replicative context (50).

While apn1Δ apn2Δ and apn1Δ apn2Δ ntg1Δ ntg2Δ cells exhibit similar overall levels of recombination rates, the distribution of GC and SSA-like events differs significantly. For example, the spontaneous recombination events caused by abasic sites in an apn1Δ apn2Δ ntg1Δ ntg2Δ strain are significantly biased toward Rad51-dependent repair similar to MMS treated WT or mag1Δ cells (compare Figures 2E and 5D), whereas the spontaneous recombination events caused by the abasic sites and ssDNA breaks in apn1Δ apn2Δ cells give rise to equivalent levels of Rad51-dependent and -independent repair, similar to untreated WT cells but at ∼10-fold higher rates (Figure 5D). The different recombination outcomes observed in apn1Δ apn2Δ and apn1Δ apn2Δ ntg1Δ ntg2Δ cells indicate that the distinct lesions that accumulate in these cells result in divergent HR outcomes. Since it has been shown that there may be more 3′-dRPs and fewer abasic sites in apn1Δ apn2Δ cells, as Ntg1 and Ntg2 can directly process abasic sites into 3′-dRPs, our data indicate that abasic sites preferentially activate Rad51-dependent recombination while 3′-dRPs trigger a higher percentage of Rad51-independent recombination (35). Importantly, the strong bias toward Rad51-dependent GC that we observe here suggest that the primary recombinogenic lesion in MMS treated cells with functional BER is likely to be an N3-MeA or an abasic site rather than an ssDNA break that encounters a replication fork. Critically, in each case observed, the Shu complex is essential to ensure this bias toward Rad51-dependent HR.

Csm2 physical interaction with Rad55 is essential for promoting repair of N3-MeAs and abasic sites

Recently, we uncovered that a single point mutation in Csm2, csm2-F46A, abrogates Csm2's physical interaction with Rad55, resulting in impaired Rad51 filament formation (10). To determine if the ability of the Shu complex to associate with Rad55 is important for promoting tolerance to different MMS-induced lesions, we compared the MMS sensitivities of mag1Δ, apn1Δ apn2Δ and apn1Δ apn2Δ ntg1Δ ntg2Δ cells (Figure 6) with cells that are also csm2Δ or csm2-F46A mutants. Like csm2Δ cells, we observe similar DNA damage sensitivities when a csm2-F46A mutant is combined with mag1Δ (Figure 6A), apn1Δ apn2Δ (Figure 6B) or apn1Δ apn2Δ ntg1Δ ntg2Δ (Figure 6C). Note that csm2-F46A apn1Δ apn2Δ cells are less MMS sensitive than csm2Δ apn1Δ apn2Δ cells (Figure 6B). Perhaps the residual interaction between csm2-F46A and Rad55 (<5% of WT, (10)) is sufficient to promote limited repair in this context. Together, these data demonstrate that the ability of the Shu complex to associate with Rad55 is essential to promote cell viability in response to each BER lesion examined.

Figure 6.

Cells containing csm2-F46A are defective for tolerance of BER lesions. (A–C) The indicated yeast strains were 5-fold serially diluted onto rich medium (YPD) or YPD medium containing the indicated doses of MMS.

The csm2Δ rad27Δ growth defect is complemented by human Polβ

To further investigate the role of the Shu complex in repairing ssDNA breaks, we also evaluated the impact of Rad27 on repair of MMS associated breaks. Rad27/FEN1 is a flap endonuclease, which is important for Okazaki fragment processing as well as excising the 5′-dRPs created by Apn1 and Apn2 (Figure 7A). To examine if the Shu complex repairs ssDNA breaks generated in a rad27Δ cell similar to those generated by apn1Δ apn2Δ, we compared the growth of WT, csm2Δ, rad27Δ and csm2Δ rad27Δ cells in the presence or absence of MMS (Figure 7B). The double csm2Δ rad27Δ mutant cells exhibit a slow growth phenotype that is exacerbated upon MMS exposure (Figure 7B). The survival in response to 0.006% MMS is decreased in csm2Δ rad27Δ compared to rad27Δ. Interestingly, the MMS sensitivity of rad27Δ cells is not rescued by deletion of MAG1 as would be expected if Mag1 functioned upstream of Rad27 (Figure 7B), indicating that the MMS sensitivity of a rad27 mutant may not solely reflect a defect in the BER pathway. In agreement with this observation, the slow growth of csm2Δ rad27Δ is not rescued by MAG1 deletion (0.002% MMS; Figure 7B).

Figure 7.

Expression of human Polβ partially suppresses the MMS sensitivity of rad27Δ cells and the synthetic growth defect observed in csm2Δ rad27Δ double mutant cells. (A) Cartoon indicating what processing step is blocked by rad27Δ disruption. (B) The indicated yeast strains were 5-fold serially diluted onto YPD medium or YPD medium containing 0.002%, 0.006%, 0.012% or 0.02% MMS where indicated. The plates were incubated at 30°C for two days and photographed. (C) WT cells or rad27Δ cells which also express either an empty vector (pAG414-GAL) or the human Polβ in an expression plasmid (pAG414-GAL-Polβ) were 5-fold serially diluted and plated onto YPRaffinose with galactose (YPR + GAL) medium or YPR + GAL medium containing 0.02% MMS. The cells were incubated at 30°C for two days and photographed. (D) WT, shu1Δ, shu2Δ, psy3Δ, csm2Δ cells were transformed with pAG414-GAL-Polβ, 5-fold serially diluted and plated onto YPR + GAL medium or YPR + GAL medium containing 0.02% MMS. The plates were incubated at 30°C for two days and photographed. (E) WT, rad27Δ, csm2Δ, csm2Δ rad27Δ cells were transformed with either the pAG414-GPD (Empty Vector) or the pAG414-GPD-Polβ (Polβ) plasmid and 5-fold serially diluted onto YPD or YPD medium containing 0.002% or 0.006% MMS. The plates were incubated at 30°C for two days and photographed.

It was recently shown that expression of human Polβ partially complements the MMS sensitivity of rad27Δ cells (54). We confirmed this observation by expressing human Polβ under a GAL or GPD promoter on a CEN plasmid (Figures 6E and 7C). In contrast to rad27Δ cells, expression of human Polβ does not suppress the MMS sensitivity of Shu complex mutants (Figure 7D), indicating that accumulation of 5′ dRPs is not the primary cause of toxicity in MMS-treated csm2Δ cells. However, human Polβ does rescue the slow growth of a csm2Δ rad27Δ double mutant as well as its MMS sensitivity at 0.002% MMS (Figure 7E). The observation that overexpression of Polβ, but not deletion of MAG1, suppresses the MMS sensitivity of a rad27Δ suggests that rad27Δ cells are not necessarily sensitive to MMS due to the lesions produced by Mag1-dependent repair of methylated DNA. One possibility may be that spontaneous loss of MMS-induced base lesions, such as N7-MeG, contributes to the sensitivity that is suppressed by overexpression of Polβ. Although it is more likely that the incomplete rescue of rad27Δ by mag1Δ suggest additional repair mechanisms, independent of BER, underlie the MMS sensitivity of rad27Δ cells. Perhaps the slowed replication fork progression in MMS-treated cells generates shorter, more frequent Okazaki fragments, which are normally processed by Rad27.

Translesion synthesis is required for bypassing MMS-induced DNA damage in csm2Δ apn1Δ apn2Δ and csm2Δ mag1Δ cells

Disruption of the Shu complex leads to increased mutation frequencies, which depends on REV7/REV3, the DNA translesion polymerase zeta (Polζ) (19,22). To determine if Pol ζ is required to bypass the DNA damage intermediates that accumulate in csm2Δ, csm2Δ apn1Δ apn2Δ or csm2Δ mag1Δ cells, we examined the MMS sensitivity of different mutant combinations in the presence or absence of REV3, the catalytic subunit of polymerase ζ. The csm2Δ rev3Δ double mutants exhibit synthetic sensitivity upon exposure to 0.006% MMS, similar to psy3Δ rev3Δ cells (Figure 8A; (22)). We find that the synthetic lethality in MMS-exposed csm2Δ rev3Δ cells is further exacerbated by deletion of APN1 and APN2 (csm2Δ apn1Δ apn2Δ rev3Δ 0.002% MMS; Figure 8B) or MAG1 disruption (csm2Δ mag1Δ rev3Δ 0.002% MMS; Figure 8C). Interestingly, while mag1Δ and rev3Δ are synthetic sensitive (Figure 8C), apn1Δ apn2Δ and rev3Δ are not as long as Csm2-associated HR is functional (Figure 8B). These results suggest that abasic sites are preferentially bypassed by error-free post-replicative repair and TLS is only initiated if HR is impaired.

Figure 8.

The translesion DNA polymerase ζ is necessary for lesion bypass in MMS-exposed csm2Δ apn1Δ apn2Δ or csm2Δ mag1Δ cells. (A–C) The indicated genotypes were 5-fold serially diluted onto YPD medium or YPD medium containing 0.002%, 0.006%, 0.012% or 0.02% MMS. The plates were incubated at 30°C for two days and photographed.

Epistasis analysis of csm2Δ and the error-free post-replicative repair genes rad5Δ, ubc13Δ, mms2Δ

Since we find that TLS likely bypasses the lesions created in the absence of CSM2, one possibility is that the Shu complex functions in the error-free branch of the post-replicative repair (PRR) pathway. To address this possibility, we combined csm2Δ with deletion of the three genes involved in initiating error-free PRR (rad5Δ, ubc13Δ or mms2Δ) (Figure 9). We find a slight, but reproducible, synthetic growth defect between csm2Δ and disruption of these three genes upon MMS exposure (Figure 9A). The growth effect is small at best at a higher level of MMS. Please note the rad5Δ mutants are significantly more sensitive to MMS than either ubc13Δ or mms2Δ mutants (Figure 9A). A further synthetic effect was observed for spontaneous mutation frequencies and recombination rates. We find that mutation frequencies in the csm2Δ rad5Δ and csm2Δ ubc13Δ double mutants are elevated compared to the corresponding single mutants (Figure 9B; P-value 0.009 and 0.01, respectively). Similarly, we observe decreased rates of GC in these same csm2Δ rad5Δ and csm2Δ ubc13Δ double mutants relative to their single mutants. However, only csm2Δ ubc13Δ is significant value (P-value 0.03), whereas csm2Δ rad5Δ trended toward significance (Figure 9C). In contrast, we do not observe a significant change in Rad51-independent SSA. Surprisingly, these results suggest that the Shu complex may have functions outside of error-free PRR.

Figure 9.

Analysis of CSM2-disrupted cells with error-free PRR disruption (rad5Δ, ubc13Δ, mms2Δ) by MMS sensitivity, mutation frequencies and direct repeat recombination rates. (A) The indicated yeast strains were 5-fold serially diluted onto rich medium (YPD) and YPD medium containing the indicated dose of MMS. The plates were incubated for two days at 30°C and photographed. (B) Mutation frequencies were determined in the indicated strains at the CAN1 locus. Three experiments with five separate isolates were analyzed and averaged. The graph represents the average of the medians with standard deviations plotted. Significance was determined by t-test. (C) The rate of gene conversion and SSA-like events were measured as described in Figure 2E. Three to four experiments were done per strain and standard deviations plotted. Significance was determined by t-test.

DISCUSSION

MMS-induced DNA damage is repaired through both the BER and HR machineries (29,55). Although the DNA lesions created by MMS are primarily repaired through BER, if the DNA damage occurs during S phase and repair is not complete by the time the replication fork encounters the lesion, then a subset of these DNA repair intermediates can lead to replication fork stalling and collapse (25,38,56,57). Stalled forks may bypass blocking lesions such as N3-MeA through potentially mutagenic translesion DNA polymerases (58). Alternatively, HR can be used to bypass a lesion at a stalled replication fork or to repair a break after a replication fork collapses. The handoff from one repair pathway to another and which DNA intermediates are substrates for the HR machinery is not well understood.

Here, we describe how the Shu complex acts to channel normal BER intermediates into the HR pathway, which also reduces the risk of mutagenesis. The major phenotypes observed in cells disrupted for the Shu complex are an elevated mutation frequency and a heightened sensitivity to the model alkylating agent MMS, which produces replicative blocking lesions such as N3-MeA and N7-MeG (Figure 10)(19,22,24). In cells disrupted for MAG1, the primary glycosylase that recognizes and excises MMS-induced base lesions (31,48,49), loss of the Shu genes causes a synthetic sensitivity to MMS. These results indicate that irreparable ss breaks lead to DSBs, while AP sites or methylated bases lead to ss gaps that are substrates for error-free bypass and the Shu complex (47). Interestingly, disruption of the abasic site endonucleases APN1 (59) and APN2 (60) causes an even more pronounced synthetic growth defect when combined with Shu gene disruptions, suggesting that the Shu complex also participates in repair of abasic sites during aborted or interrupted BER. Furthermore, our genetic evidence indicates that the Shu complex also repairs DNA with 3′-dRPs that are produced during BER and are incredibly toxic due to their inability to be acted upon by DNA polymerases (61).

Figure 10.

Overview repair model of MMS-induced DNA damage in budding yeast by BER, translesion DNA synthesis (TLS) and HR (Adapted from (55)). MMS damage leads to base lesions excised by the DNA glycosylase Mag1 to create an AP site. Both the MMS-induced base lesions and AP sites can lead to a replication fork block, which can either be bypassed by the TLS pathway (requiring Rev3) resulting in increased mutation frequency and therefore, error-prone repair (69). During BER, Apn1/Apn2 further processes the AP site leading to 5′-dRPs, which are filled in by DNA polymerase ϵ and the 5′-dRP flap is cleaved by Rad27. Lig1 ligates the resulting nick. The 5′-dRP and flap as well as the nicked DNA can lead to replication fork collapse and DSB formation. The resulting DSB can be repaired by the high-fidelity HR pathway via the Shu complex, which is important for the critical Rad51 filament formation step of HR.

Moving forward, it will be interesting to examine if Ogg1, which can also act on abasic sites to generate ssDNA breaks, will similarly suppress an apn1Δ apn2Δ csm2Δ cell resulting in elevated mutagenesis and recombination. Surprisingly, we provide evidence that the Shu complex may have a function outside of the error-free PRR pathway. Finally, disruption of Rad27/FEN1, the flap endonuclease that removes a stretch of nucleotides to allow repair during BER (62), also causes an extreme synthetic sensitivity to MMS when the Shu complex is disrupted. This sensitivity is alleviated when human Polβ is expressed indicating that the Shu complex is required to resolve the numerous ssDNA breaks left when Rad27/FEN1 is missing (54). Taken together, our genetic evidence demonstrates that there is a close crosstalk between the BER and HR repair pathways, where normal repair intermediates produced during BER are repaired and resolved faithfully by HR to minimize mutagenesis during DNA replication.

Despite its role in regulating HR through Rad51 filament formation, loss of the Shu complex creates a more modest phenotype when compared to loss of its binding partners RAD55 and RAD57 (10,20). For instance, rad55Δ cells are sensitive to a broader range of damaging agents, such as UV and H2O2, both of which would cause replication fork stalling base modifications. Why the Shu complex is not sensitive to these agents, while Rad55 mutants are, remains an open question. One hypothesis is that, while Rad55-Rad57 is known to function during S and G2 phase (45,64), it has been suggested that the Shu complex functions primarily during replicative stress in S phase (21,22,24). Here, we directly test the hypothesis that the Shu complex has a role in repair during replicative stress. We find no compelling evidence that the Shu complex functions during G2/M. Although the protein levels of the Shu complex members Shu2, Csm2 and Psy3 remain constant throughout the cell cycle, we find that in G2 arrested cells treated with MMS, chromosome restitution is independent of the Shu complex.

Together, our data suggest a model whereby the Shu complex responds to numerous BER intermediates, including the replicative blocking MMS-induced base lesions such as N3-MeA, the highly mutagenic abasic site, cytotoxic ssDNA breaks and blocked ends that generate a full DSB during replication, all in the specific context of replicative stress. These results may suggest that the lethality of rad55 mutants in UV or H2O2 exposed cells may occur due to DSBs arising from lesions outside of S phase, when the Shu complex is not active. Treatment of cells disrupted for the Shu complex with chronic UV exposure, which increases a cell's reliance on bypass mechanisms, may reveal a previously unknown UV sensitivity in these cells. Alternatively, the lesions produced by UV and H2O2 may not be channeled into Shu complex dependent HR, indicating that different replication fork blocking lesions may initiate different forms of HR.

Recent work has clearly demonstrated that numerous human conditions such as cancer predisposition and Fanconi anemia arise from defects in HR proteins such as BRCA1/2, RAD51 and the RAD51 paralogues (65,66). Similarly, aging and cancer have also been attributed to defects in BER (67). It has been shown that cancers with defective HR are exquisitely sensitive to PARP1 inhibitors, which block PARP1 function and increase the amount of ssDNA breaks that give rise to DSBs and require repair by HR (68). Our work in budding yeast creates a framework for understanding and studying the crosstalk between BER intermediates and HR that could have important implications for cancer treatment. For example, our work suggests that there are multiple other targets worth inhibiting for the treatment of cancers with defective HR. Inhibition of human APN1/2 (APE1/2), as well as human Rad27 (FEN1), would likely result in extreme synthetic lethality in such tumors just like PARP1 inhibition. Finally, our work strongly supports the importance of investigations of combinations of PARP1 inhibitors and chemotherapeutic agents. Clinical trials addressing such combinations are currently underway.