Epistasis analysis between homologous recombination genes in Saccharomyces cerevisiae identifies multiple repair pathways for Sgs1, Mus81-Mms4 and RNase H2

Abstract

The DNA repair genes SGS1 and MUS81 of Saccharomyces cerevisiae are thought to control alternative pathways for the repair of toxic recombination intermediates based on the fact that sgs1Δ mus81Δ synthetic lethality is suppressed in the absence of homologous recombination (HR). Although these genes appear to functionally overlap in yeast and other model systems, the specific pathways controlled by SGS1 and MUS81 are poorly defined. Epistasis analyses based on DNA damage sensitivity previously indicated that SGS1 functioned primarily downstream of RAD51, and that MUS81 was independent of RAD51. To further define these genetic pathways, we carried out a systematic epistasis analysis between the RAD52-epistasis group genes and SGS1, MUS81, and RNH202, which encodes a subunit of RNase H2. Based on synthetic-fitness interactions and DNA damage sensitivities, we find that RAD52 is epistatic to MUS81 but not SGS1. In contrast, RAD54, RAD55 and RAD57 are epistatic to SGS1, MUS81 and RNH202. As expected, SHU2 is epistatic to SGS1, while both SHU1 and SHU2 are epistatic to MUS81. Importantly, loss of any RNase H2 subunit on its own resulted in increased recombination using a simple marker-excision assay. RNase H2 is thus needed to maintain genome stability consistent with the sgs1Δ rnh202Δ synthetic fitness defect. We conclude that SGS1 and MUS81 act in parallel pathways downstream of RAD51 and RAD52, respectively. The data further indicate these pathways share common components and display complex interactions.

1. INTRODUCTION

Homologous recombination is crucial to the repair of DNA double-strand breaks (DSBs) in living cells [1–3]. HR is an evolutionarily conserved process that uses templated information to repair DSBs in an error-free manner. In humans, HR plays a significant role in maintaining genome stability and tumor suppression [4,5].

An early step of HR involves homology search and DNA strand invasion. This reaction requires a subset of the RAD52 epistasis group genes including RAD51, RAD52, RAD54, RAD55 and RAD57. Rad51 is the eukaryotic RecA homolog that catalyzes homologous search and DNA strand exchange [6] with the help of mediator proteins Rad52 and Rad55-Rad57 complex [7–9]. Rad52 accelerates the removal of RPA from ssDNA by Rad51 [10]. Rad52 is capable of annealing homologous ssDNA [11], which is thought to be required for synthesis-dependent strand annealing (SDSA), single-strand annealing (SSA) and the second-end capture step of the DSB repair [12–14]. Rad54 function is diverse: Rad54 stabilizes the Rad51 presynaptic filament [15,16], it stimulates DNA strand invasion mediated by Rad51 in vitro [17,18], it dissociates Rad51 from heteroduplex DNA to load DNA polymerase [19], and it has branch-migration activity on Holliday junctions [20,21]. Rad55 and Rad57 are Rad51 paralogs in S. cerevisiae and these proteins exhibit mediator activity by forming a heterodimer that accelerates Rad51-mediated recombination in vitro [8,22] together with Rad52.

Consistent with the essential role of Rad51 in strand invasion, rad51Δ mutants display dramatic reductions in most HR reactions. However, some gene conversion events are known to occur in the absence of Rad51. For example, deletion of RAD52 reduces the efficiency of spontaneous gene conversion between inverted repeats by 3,000-fold, whereas deletion of RAD51 reduces it only 4-fold [23]. RAD51 is not essential in ectopic gene conversion [24], and yeast is capable of RAD51-independent DSB repair resulting in interchromosomal gene conversion [25]. Therefore, budding yeast exhibits both Rad51-dependent and Rad51-independent recombination. Moreover, Break-Induced Replication (BIR) includes both Rad51-independent [26,27] and Rad51-dependent events [28,29]. Taken together with the existence of RAD52-independent recombination pathways in budding yeast [30], it is clear that HR pathways have not been completely defined.

SHU1, SHU2, CSM2, and PSY3 (referred to as the SHU genes) were isolated in a top3 suppressor screen [31] and loss of any one of these genes suppresses various defects in sgs1Δ or top3Δ mutants. SHU gene products are thought to form a multimeric complex involved in recombinational repair [31].

Recombination intermediates are potentially toxic and must be dissolved or displaced. The human BLM-TOPOIII -RMI1-RMI2 complex catalyzes the dissolution of recombination intermediates containing double-Holliday junctions (dHJs) in vitro [32–34]. The corresponding S. cerevisiae complex Sgs1-Top3-Rmi1 [35,36] has the same activity [37]. Mutations in SGS1, TOP3 or RMI1 cause similar phenotypes including sensitivities to DNA damaging agents, hyper-recombination, and synthetic lethality with mutations in MUS81 and SRS2 [35–40]. In addition, sgs1Δ, top3Δ and the catalytic point mutant TOP3^Y356F cells accumulate MMS-induced recombination intermediates based on two-dimensional (2D) gel electrophoresis of DNA replication products [41,42]. Many of the phenotypic detects of sgs1, top3 and rmi1 cells including the synthetic lethality of sgs1Δ mus81Δ double mutants are suppressed by deletion of RAD51, RAD52, RAD54, RAD55, or RAD57 [38,40,43]. These results strongly suggest that SGS1, TOP3 and RMI1 are important for resolving recombination intermediates in living cells.

S. cerevisiae MUS81 and MMS4 were identified in a synthetic-lethal screen using an sgs1Δ mutant [39]. Mus81 and Mms4 form a structure-specific endonuclease that cleaves replication fork-like structures, nicked-Holliday junctions, D-loops and 3′flaps [43–47]. The mus81Δ mutant cells accumulate recombination intermediates during replication as detected by 2D gel electrophoresis [48–50]. Based on DNA damage sensitivity, MUS81 and MMS4 are known to function downstream of RAD52 in S. cerevisiae and RAD22 in S. pombe [51,52].

RNase H2 appears to play a minor role in DNA replication by acting redundantly with Rad27/FEN1 and Dna2 to remove RNA primers from Okazaki fragments [53,54]. In contrast, new results suggest that RNase H2 plays a more direct role in DNA repair by removing mis-incorporated ribonucleotides [55,56]. Interestingly, the loss of SGS1 and any of the three genes encoding RNase H2 (RNH201, RNH202, and RNH203 in budding yeast) results in synthetic-fitness defects [55,57]. Such a result is consistent with the idea that the loss of RNase H2 creates a need for recombinational repair.

Previous epistasis analysis indicated that SGS1 acts primarily downstream of RAD51 and MUS81 was independent of RAD51 [55]. To further characterize the recombinational repair pathways that are utilized by SGS1, MUS81 and RNH202, we performed additional epistasis tests between these genes and a large set of HR genes (RAD52, RAD54, RAD55, RAD57, SHU1, SHU2). Our results indicate that MUS81 functions primarily downstream of RAD52, whereas SGS1 and RNH202 are independent of RAD52. Interestingly, the phenotypes of SGS1, MUS81, and RNH202 and the synthetic-sickness of sgs1Δ rnh202Δ double mutants are suppressed by deletion RAD54, RAD55, and RAD57 suggesting a complex interaction between these two sets of genes. In addition, SHU2 (but not SHU1) is epistatic to SGS1, SHU1 is epistatic to RNH202, and both SHU1 and SHU2 are epistatic to MUS81. Finally, deletion of any of three RNase H2 genes results in an increase of recombination frequency strongly suggesting that lesions created in the absence of RNase H2 require Sgs1 for repair. These results allow us to propose a model that further defines the recombinational repair pathways distinguished by SGS1 and MUS81.

2. MATERIALS AND METHODS

2.1 Media and yeast strains

Standard procedures were used for mating, sporulation, and tetrad dissection [58]. To generate shuΔ mutants, random spore analyses were performed as described previously [59]. All experimental procedures were carried out at 30°C. Yeast strains used in this study are listed in Table 1. All strains are RAD5+ derivatives of W303 unless noted otherwise. The weak rad5-535 mutation found in some W303-derived strains did not account for any MMS sensitivity under the conditions of low (<0.03%) MMS sensitivity used here. In all experiments, multiple meiotic segregants of the same genotype were found to behave similarly.

Table 1.

Strains used in this study.

Strain	Genotype	Reference or source
HKY579-10A	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 can1-100	Hannah Klein
HKY580-10D	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 can1-100	Hannah Klein
JMY332	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 sgs1-3::TRP1 can1-100 rad5-535	[39]
JMY380	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 mus81-10::KAN can1-100 rad5-535	[39]
HKY614-10B	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad52::TRP1 can1-100	Hannah Klein
HKY624	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad54::HIS3 can1-100	Hannah Klein
HKY597-2C	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad55::LEU2 can1-100	Hannah Klein
HKY598-8B	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad57::LEU2 can1-100	Hannah Klein
JMY372	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 top3-2::HIS3 can1-100 rad5-535	[35]
JMY1918	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rmi1-10::KAN can1-100 rad5-535	[35]
MIY0614	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 shu1::KAN can1-100	This study
MIY0650	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 shu1::hphMX4 can1-100	This study
MIY0632	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 shu2::hphMX4 can1-100	This study
MIY0640	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 shu2::KAN can1-100	This study
MIY1903	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rnh202::NAT can1-100 rad5-535	[55]
MIY2071	MATa ade2-1 ade3::hisG ura3-1 his3-11, 15 trp1-1 leu2-3, 112 sgs1- 20::hphMX4 rnh202::NAT can1-100	[55]
MIY0944	MATα ade2-1 ade3::hisG ura3-1 his3-11, 15 trp1-1 leu2-3, 112 sgs1- 20::hphMX4 rnh202::NAT can1-100	This study
MIY2073	MATa ade2-1 ade3::hisG ura3-1 his3-11, 15 trp1-1 leu2-3, 112 mus81- 10::KAN rnh202::NAT can1-100	[55]
MIY0945	MATα ade2-1 ade3::hisG ura3-1 his3-11, 15 trp1-1 leu2-3, 112 mus81- 10::KAN rnh202::NAT can1-100	This study
NJY1612	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad52::TRP1 sgs1- 20::hphMX4 mus81::KAN can1-100	This study
MIY0929	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad52::TRP1 sgs1::hphMX4 can1-100	This study
MIY0930	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad52::TRP1 mus81::KAN can1-100	This study
MIY0931	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad52::TRP1 rnh202::NAT can1-100	This study
MIY0932	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 mus81::KAN rad52::TRP1 rnh202::NAT can1-100	This study
MIY0933	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 sgs1::hphMX4 rad52::TRP1 rnh202::NAT can1-100	This study
VCY1471	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 sgs1-3::TRP1 rad54::HIS3 mus81::KAN can1-100	This study
MIY0934	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad54::HIS3 sgs1-3::TRP1 can1-100	This study
MIY0936	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad54::HIS3 mus81::KAN can1-100	This study
MIY0938	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad54::HIS3 rnh202::NAT can1-100	This study
MIY0941	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad54::HIS3 sgs1-3::TRP1 rnh202::NAT can1-100	This study
MIY0942	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad54::HIS3 mus81::KAN rnh202::NAT can1-100	This study
MIY0943	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad54::HIS3 sgs1-3::TRP1 mus81::KAN rnh202::NAT can1-100	This study
JMY1441	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad55::LEU2 sgs1- 3::TRP1 can1-100	This study
MIY0948	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad55::LEU2 mus81- 10::KAN can1-100	This study
MIY0950	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad55::LEU2 rnh202::NAT can1-100	This study
MIY0952	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad55::LEU2 sgs1- 3::TRP1 mus81-10::KAN can1-100	This study
MIY0953	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad55::LEU2 sgs1- 3::TRP1 rnh202::NAT can1-100	This study
MIY0954	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad55::LEU2 mus81- 10::KAN rnh202::NAT can1-100	This study
MIY0955	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad55::LEU2 sgs1- 3::TRP1 mus81-10::KAN rnh202::NAT can1-100	This study
VCY1516	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad57::LEU2 sgs1- 3::TRP1 mus81::KAN can1-100	This study
MIY0956	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad57::LEU2 sgs1- 3::TRP1 can1-100	This study
MIY0958	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad57::LEU2 mus81::KAN can1-100	This study
MIY0960	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad57::LEU2 rnh202::NAT can1-100	This study
MIY0963	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad57::LEU2 sgs1- 3::TRP1 rnh202::NAT can1-100	This study
MIY0964	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad57::LEU2 mus81::KAN rnh202::NAT can1-100	This study
MIY0965	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 rad57::LEU2 sgs1- 3::TRP1 mus81::KAN rnh202::NAT can1-100	This study
MIY0685	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 sgs1-20::hphMX4 shu1::KAN can1-100	This study
MIY0763	MATα ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 sgs1-20::hphMX4 shu2::KAN can1-100	This study
MIY0745	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 mus81-10::KAN shu1::hphMX4 can1-100	This study
MIY0825	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 mus81-10::KAN shu2::hphMX4 can1-100	This study
MIY0705	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 shu1::KAN rnh202::NAT can1-100	This study
MIY0777	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 shu2::hphMX4 rnh202::NAT can1-100	This study
MIY0725	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 sgs1-20::hphMX4 rnh202::NAT shu1::KAN can1-100	This study
MIY0766	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 sgs1-20::hphMX4 rnh202::NAT shu2::KAN can1-100	This study
MIY0748	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 mus81-10::KAN rnh202::NAT shu1::hphMX4 can1-100	This study
MIY0698	MATa ade2-1 ura3-1 his3-11, 15 trp1-1 leu2-3, 112 mus81-10::KAN rnh202::NAT shu2::hphMX4 can1-100	This study

2.2 Growth rate determination

The OD₆₀₀ of cultures in exponential growth phase in YPD was determined every hour for 7 hr, and doubling times were determined. Three isolates were analyzed for each strain.

2.3 Sensitivity to DNA damaging agents

To measure sensitivity to DNA damaging agents, spot assays were performed as reported previously [55]. OD₆₀₀ =3 cell suspensions were made and serial 1:10 dilutions were spotted onto indicated plates with or without DNA damaging agents. Plates were photographed after 3 days.

2.4 Recombination assay

The recombination assay was performed essentially as described previously [35,60]. The loss of three different marker genes (ADE2 and URA3 integrated independently at rDNA loci and CAN1 integrated at LYS2 locus) were measured as follows. Loss of ADE2 was examined by growing the cells in YPD broth supplemented with adenine, plating them on YPD plates for single colonies, and incubating for 3 days. The marker loss rates for ADE2 were determined by dividing the number of red colonies by the total number of colonies on the plates. For the loss of URA3, the same number of cells were plated on –URA and non-selective plates and incubated for 3 days. The number of colonies on –URA plate was subtracted from the number of colonies on non-selective medium plate and that number was divided by the total number of colonies on non-selective plate. To measure the loss of CAN1, about 400 cells were plated on non-selective plates and 100-fold more cells were on canavanine plates. The colonies growing on canavanine were counted and designated as can1 mutants having lost CAN1 through recombination. Thus, the recombination frequency at the LYS2 locus was determined by dividing the number of colonies on the canavanine plate by the total number of plated cells that was derived from the number of colonies on the non-selective plate.

3. RESULTS

3.1 MUS81 functions in a RAD52-dependent pathway whereas RNase H2 does not

Previous epistasis analysis indicated that SGS1 functions primarily in a RAD51-dependent pathway whereas MUS81 was independent of RAD51 [55]. For example, loss of RAD51 had little effect on the doubling time (DT) of an sgs1Δ rnh202Δ double mutant (179 vs 174 min), but it severely inhibited a mus81Δ rnh202Δ double mutant (200 vs 147 min). It was also concluded that RNH202 functioned independently of RAD51, SGS1 and MUS81 since the doubling time of the mus81Δ sgs1Δ rnh202Δ rad51Δ quadruple mutant was significantly slower than the mus81Δ sgs1Δ rad51Δ strain (271 vs 202 min).

To further examine the relationships between other RAD52 epistasis group genes and SGS1, MUS81 and RNH202, we crossed strains bearing complete deletions of each gene to create double, triple and quadruple mutants. We first examined the relationships among the rad52Δ, sgs1Δ, mus81Δ and rnh202Δ mutants. Surprisingly, we found that the sgs1Δ mus81Δ rad52Δ rnh202Δ quadruple mutant is inviable. Since sgs1Δ mus81Δ synthetic lethality was previously shown to be suppressed by deletion of RAD52, we conclude that some portion of SGS1 and/or MUS81 must function downstream of RAD52. The lethality of the sgs1Δ mus81Δ rad52Δ rnh202Δ quadruple mutant is therefore consistent with the idea that RNH202 functions independently of RAD52.

To test this hypothesis, we examined the doubling times of several rad52Δ mutants (Fig. 1A). Although the growth rate of a rad52Δ single mutant was not significantly increased by the additional deletion of MUS81 (147 vs 142 min), the rad52Δ growth defect was exacerbated by deletion of SGS1 (197 vs 142 min), which is consistent with the result reported previously [61]. In addition, we observed a severe synthetic growth defect in the rad52Δ rnh202Δ double mutant (234 vs 142 min). Based on these results, RAD52 is epistatic to MUS81 and independent of SGS1. It had been reported that RAD52 is epistatic to MMS4 [51] therefore our result is consistent with theirs. We also conclude that RNH202 functions in a pathway parallel to RAD52 due to severity of the rad52Δ rnh202Δ synthetic growth defect and the synthetic lethality of the sgs1Δ mus81Δ rad52Δ rnh202Δ mutant.

Figure 1. Genetic interactions between RAD52, SGS1, MUS81, and RNH202.

A. Strains of the indicated genotypes were grown in liquid medium at 30°C and the doubling times were examined to determine genetic interactions between RAD52 and other DNA repair genes. B. Cells of the indicated genotype were spotted in ten-fold serial dilutions onto YPD plates containing the indicated concentrations of DNA damaging drug or no drug. Photographs were taken after 3 days. C. Schematic representation of the simplest relationship between RAD51, RAD52, SGS1, MUS81 and RNH202.

We next examined the sensitivities of each of the strains to DNA damaging agents. This was determined by continuous exposure to camptothecin (CPT), hydroxyurea (HU), and MMS using the spot-dilution assay (Fig. 1B). For each drug, the rad52Δ strain was found to be more sensitive than any other single mutant. In addition, the rad52Δ and rad52Δ mus81Δ strains showed the same level of sensitivity to each of the DNA damaging agents. Thus, MUS81 functions primarily in the same pathway as RAD52 consistent with the above epistasis analysis based on doubling time. Since MUS81 is known to function downstream of HR [51,52], these results suggest that Mus81 acts to resolve recombination-intermediates in the Rad52 pathway. In contrast to rad52Δ mus81Δ epistasis, rad52Δ sgs1Δ and rad52Δ rnh202Δ mutants displayed an additive effect. The double mutants grew more slowly than their respective single mutants on plates containing CPT, HU or MMS (Fig. 1B). This is especially noticeable in the case of rnh202Δ, which eliminated the viability of rad52Δ cells on CPT and HU. This analysis indicates that a portion of SGS1 acts in a pathway parallel to RAD52. In contrast, RNH202 appears to function exclusively in a pathway parallel to RAD52. This latter result is again consistent with the results of the doubling time analysis.

Figure 1C uses a schematic to summarize the above results in terms of the relationships between RAD51 and RAD52, and the genes SGS1, MUS81 and RNH202. Previous work indicated that SGS1 functioned downstream of RAD51 whereas MUS81 and RNH202 did not [55]. Thus, RAD51-SGS1 is shown to form a recombinational repair pathway parallel to the pathway composed of RAD52-MUS81. Because neither RAD51 nor RAD52 is epistatic to RNH202, it is shown to function in parallel to RAD51-SGS1 and RAD52-MUS81.

3.2 The synthetic growth defect of sgs1Δ rnh202Δ is partially suppressed by deletion of RAD54, RAD55, or RAD57

To further define these pathways, we searched for genetic interactions between other RAD52 epistasis group genes and SGS1, MUS81 and RNH202. RAD54 is thought to function upstream of SGS1 and/or MUS81 due to the fact that sgs1Δ mus81Δ synthetic lethality is suppressed by deletion of RAD54 [40,43]. Our analyses of RAD54 revealed that the synthetic sickness between sgs1Δ and rnh202Δ is partially suppressed by deletion of RAD54 (Fig. 2A). That is, whereas the doubling time of sgs1Δ rnh202Δ was 182 min, the rad54Δ sgs1Δ rnh202Δ triple mutant grew faster (167 min). This result is consistent with the results of Ooi and Boeke [57], and the idea that RAD54 acts upstream of SGS1 and/or RNH202. In contrast, the synthetic growth defect of mus81Δ rnh202Δ was exacerbated by deletion of RAD54, rather than being suppressed (Fig. 2A: 150 vs 168 min). Our interpretation is that Mus81 and RNase H2 function differently in the repair that requires Rad54. We next examined the relationships with RAD55. In general, we observed similar results to the RAD54 analysis. The synthetic growth defect of sgs1Δ rnh202Δ (DT=182 min) was partially suppressed by deleting RAD55 (DT=168 min), and the synthetic growth defect of mus81Δ rnh202Δ (DT= 150 min) was not suppressed but was exacerbated (DT=161 min). Suppression of the sgs1Δ rnh202Δ growth defect was also observed by deleting RAD57. We note that suppression of sgs1Δ rnh202Δ synthetic sickness was most effective when RAD57 was deleted. Again, no suppression of the mus81Δ rnh202Δ synthetic growth defect was observed by deleting RAD57 (DT=155 min).

Figure 2. Genetic interactions between RAD54, RAD55, RAD57, and SGS1, MUS81, and RNH202.

A. The doubling times of each of the mutants were determined as described in Figure 1. B. Spot assay with single mutants and rad54Δ derivatives. Assay was performed as described in Fig. 1B. C. Spot assay with rad55Δ derivatives. D. Spot assay with rad57Δ derivatives. E. Summary of suppressions of drug sensitivities of rad54Δ, rad55Δ, and rad57Δ mutants by deletion of SGS1, RNH202, or MUS81. This summarizes the results of the spot assays shown in Fig. 2B–D and reveals the similarities between sgs1Δ and rnh202Δ on the drug sensitivities of rad54Δ, rad55Δ and rad57Δ mutants. One plus implies 10-fold better growth compared to the drug sensitivity of the corresponding rad54/55/57 single mutants. Note that deletion of SGS1 or RNH202 resulted in the same general pattern of suppression of DNA damage sensitivity in rad54Δ, rad55Δ and rad57Δ mutants.

As reported previously, sgs1Δ mus81Δ synthetic lethality can be suppressed by deletion of any one of these RAD genes [38,40,43]. This fact and our results suggest that the sgs1Δ rnh202Δ defect lies downstream of these recombination factors. In contrast, deletion of these RAD genes failed to suppress the synthetic growth defects of mus81Δ rnh202Δ mutants. Thus, the mus81Δ rnh202Δ defect appears to lie upstream of these factors.

To further examine the differences between these recombinational repair pathways, we tested the sensitivity of single and double mutants by spotting serial dilutions of them onto plates containing DNA damaging agents. Figure 2B shows the results of spot assays of rad54Δ derivatives. Surprisingly, deletion of SGS1 partially suppressed the sensitivity of rad54Δ to CPT and HU, whereas it did not suppress the sensitivity of rad54Δ to MMS. Similarly, deletion of RNH202 suppressed the sensitivity of rad54Δ to CPT and HU, but not MMS. In contrast, deletion of MUS81 did not significantly affect rad54Δ’s sensitivity to any of these agents.

Similar results were obtained in the analyses of rad55Δ and rad57Δ derivatives as shown in Figures 2C and D. Deletion of SGS1 and RNH202 suppressed the sensitivities of rad55Δ and rad57Δ to CPT and HU to various degrees, but did not affect their sensitivities to MMS. As above, deletion of MUS81 did not have a significant affect on the sensitivities of rad55Δ or rad57Δ to these drugs.

Figure 2E summarizes the above results and highlights the similarities between sgs1Δ and rnh202Δ. For example, deletion of SGS1 or RNH202 resulted in the same general pattern of suppression of DNA damage sensitivity in rad54Δ, rad55Δ and rad57Δ mutants. MUS81 behaved differently as its deletion suppressed only the HU sensitivity of rad57Δ. As described above, deletion of RAD54, RAD55 or RAD57 partially suppressed the growth defect of sgs1Δ rnh202Δ mutants, but could not suppress the growth defect of mus81Δ rnh202Δ cells. One interpretation of these results is that the functions of RNase H2 and Sgs1 are partially overlapping upstream of RAD54, RAD55, and RAD57. In addition, it suggests that Sgs1 and RNase H2 contribute to the toxicity of agents that arrest replication forks in the absence of these HR factors.

3.3 Synthetic lethality between rnh202Δ and top3Δ reveals an overlapping function between RNase H2 and Top3-Rmi1 complex

Since Sgs1 is thought to function together with Top3 and Rmi1 to resolve recombination intermediates [35,36,41,42], we examined whether top3Δ and rmi1Δ displayed a synthetic lethal/sickness interaction with rnh202Δ. A top3Δ mutant was crossed to an rnh202Δ mutant and sporulated prior to dissection of ascospores. As shown in Figure 3A, top3Δ rnh202Δ double mutants were inviable. Haploid rmi1Δ and rnh202Δ strains were similarly mated and dissected. Shown in Figure 3B are the results from six rmi1Δ rnh202Δ spores that were obtained from five tetrads. Two of these double mutants were inviable and, although four others germinated on the YPD plate, they could not be passaged. These synthetic-lethal interactions suggest that Top3-Rmi1 and RNase H2 act in parallel pathways similar to Sgs1 and RNase H2.

Figure 3. Synthetic-lethal interactions between TOP3, RMI1 and RNH202.

A. Tetrads from a cross between strains JMY372 (top3Δ) and MIY1903 (rnh202Δ) were dissected and incubated for 4 days at 30°C. Subsequent analysis revealed that small colonies are top3Δ mutants and all dead spores are top3Δ rnh202Δ double mutants. B. Tetrads from a cross between strains JMY1918 (rmi1Δ) and MIY1903 (rnh202Δ) were dissected and incubated for 4 days at 30°C. Subsequent analysis revealed that both two dead spores and four small colonies are rmi1Δ rnh202Δ double mutants. Note that four rmi1Δ rnh202Δ spores were germinated but could not be passaged.

3.4 Genetic interactions between SHU1 and SHU2 and the downstream genes SGS1, MUS81 and RNH202

SHU1 and SHU2 are known to function in recombination together with RAD52 upstream of SGS1 and TOP3 [31]. To further characterize these recombinational pathways, we searched for genetic interactions between SHU1 and SHU2 and these downstream genes. Strains bearing complete deletions of each gene were created, and double or triple mutants were obtained from genetic crosses. As shown in Figure 4, deletion of SHU1 in the sgs1Δ background resulted in a slow-growth phenotype (160 vs 142 min). In contrast, deletion of SHU2 had no effect on sgs1Δ (137 vs 142 min). These results suggest that SGS1 functions primarily downstream of SHU2, but not SHU1. Deletion of SHU1 and SHU2 had no effect on the growth rate of mus81Δ cells. However, deletion of SHU2 in the rnh202Δ background resulted in a slow-growth phenotype (155 vs 137 min) and deletion of SHU1 in did not. Thus, in contrast to SGS1, these results suggest that RNH202 functions downstream of SHU1 but not SHU2. Finally, the synthetic growth defect of sgs1Δ rnh202Δ was suppressed by deletion of SHU2 (153 vs 174 min), but was exacerbated by a deletion of SHU1 (184 vs 174 min). These results are consistent with the idea that SGS1, MUS81, and RNH202 act in parallel pathways. Although these pathways are downstream of both SHU1 and SHU2 they display somewhat different dependencies on SHU1 and SHU2.

Figure 4. Genetic interactions between SHU1, SHU2 and SGS1, MUS81, and RNH202.

The doubling times of each of the mutants were determined as described in Fig. 2. Note that synthetic-fitness phenotype of sgs1Δ rnh202Δ is substantially suppressed by loss of SHU2.

3.5 Loss of RNase H2 increases recombination frequency

To determine whether loss of RNase H2 affects genome stability through recombination, a strain was used to assay recombination frequencies in the genome. The strain K1875 carries two selectable markers (ADE2 and URA3) independently integrated at the rDNA loci and the third marker (CAN1) integrated at the LYS2 locus [60]. In all cases the marker is flanked by direct repeats, and it is known that they are most often lost by excision recombination that results in ade2, ura3, or can1 phenotypes. As shown in Figure 5, loss of any one of RNase H2 subunits resulted in an increase in recombination frequency for all the markers ADE2, URA3 and CAN1. Compared to wt cells, all the rnh mutants displayed a 3 to 4-fold increase in marker excision rate for all the loci. Thus, the loss of RNase H2 causes hyper-recombination phenotype. As control, we determined recombination rates in sgs1Δ mutants and confirmed that loss of Sgs1 results in a 5 to 6-fold increase in marker excision rate at all the loci as reported previously [35]. Loss of Mus81 did not affect the marker excision rate at LYS2, and resulted in a small (1.4 – 1.9 fold) increase at the rDNA. Taken together, we conclude that RNase H2 is important for genome stability, presumably due to the inability to remove mis-incorporated ribonucleotides.

Figure 5. Loss of RNase H2 causes an increase of recombination frequency in a whole genome.

Recombination reporter strains were constructed that contained the indicated genotypes in addition to three different marker genes flanked by direct repeats: ADE2 and URA3 integrated independently at rDNA loci, and CAN1 integrated at the LYS2 locus. The frequency with which these markers were lost from the genome was determined as described in the Materials and Methods. Shown are the average recombination frequencies of three experiments ± SD. Also shown is the change in recombination frequency relative to WT.

4. DISCUSSION

One of the major findings of this study is that MUS81 acts primarily in a RAD52-dependent pathway for DNA repair, whereas SGS1 is primarily independent of RAD52. Previously we found that SGS1 primarily functioned in a RAD51-dependent pathway whereas MUS81 primarily acted independently of RAD51 [55]. By taking these results into account, we propose that there are at least two major recombinational repair pathways distinguished by RAD51-SGS1 and RAD52-MUS81. MUS81-MMS4 has functional overlap with SGS1-TOP3-RMI1 in recombinational repair [39], however there are differences between them. For example, sgs1Δ, top3Δ, and rmi1Δ mutants have been reported to exhibit a hyper-recombination phenotype [62–64], whereas mus81Δ and mms4Δ mutants have been reported to display wild-type levels of mitotic recombination [51,65,66]. One interpretation of the current results is that the similarities between RAD51-SGS1 and RAD52-MUS81 derive from their overlapping functions in recombinational repair, and their differences derive from the fact that the proteins act on different substrates. Previous studies have indicated that RAD52 is epistatic to MMS4, which encodes a subunit of the Mus81/Mms4 complex, and that RAD22 (a RAD52 homolog) is epistatic to MUS81 in S. pombe [51,52]. Our result is consistent with these earlier reports and indicates that this epistatic relationship is conserved in yeasts. In addition, it has been reported that SGS1 functions together with RAD52 under certain circumstances [38,67,68], however the relative roles of the RAD52 epistasis genes had not been tested in a comprehensive manner. Our systematic analyses of these genetic interactions provide new information on the functions of these multiple recombinational repair pathways.

These analyses indicate that RNase H2 functions outside of the RAD52-dependent pathway and that it is partly epistatic to RAD51. This conclusion is based on the severe growth defect of rad52Δ rnh202Δ cells and the exacerbation of the sgs1Δ rnh202Δ growth phenotypes by deletion of RAD52. Further, loss of RAD51 had a small negative effect on rnh202Δ cells, and no effect on the growth of sgs1Δ rnh202Δ cells [55]. This suggests that repair in the absence of RNase H2 relies greatly on RAD52.

Interestingly, the loss of RAD54, RAD55, or RAD57 partially suppressed the synthetic growth defects of sgs1Δ rnh202Δ cells. This result is consistent with the earlier finding that loss of RAD51 partially suppressed the growth defects of sgs1Δ rnhΔ cells [55,57]. We previously failed to observed suppression of the growth defects of sgs1 rnh202Δ cells by deleting RAD51 [55,57]. We cannot explain this difference other than to suspect that there were differences in the strain backgrounds that were used in the two studies. The current data suggest that the synthetic sickness of sgs1Δ rnh202Δ double mutants is due to the need for Sgs1 to repair lesions that arise in the absence of RNase H2. The source of these lesions appears to be single ribonucleotides that are mis-incorporated during DNA synthesis since a major role of RNase H2 is to eliminate these ribonucleotides [56]. In the absence of Sgs1 and RNase H2, toxic recombination intermediates arise due to the action of RAD54, RAD55, RAD57 and RAD51 [55,57].

We found that the DNA damage sensitivities of a subset of RAD mutants were suppressed by deleting SGS1 or RNH202 (Fig. 2B–E). Interestingly, similar levels of suppression were obtained by deleting SGS1 or RNH202, suggesting that they were behaving similarly. But, RNase H2 has functions that are clearly independent of Sgs1-Top3-Rmi1, so this is unlikely to be the correct interpretation. The rad54Δ, rad55Δ, and rad57Δ mutants are sensitive to CPT and HU due to the need for HR-dependent repair of replication forks stalled by these compounds. Perhaps the suppression of their drug-sensitivities is due to the effect of the sgs1Δ and rnh202Δ mutations on fork movement. For instance, the loss of Sgs1 or RNase H2 may result in lesions that slow the rate of fork movement or stall forks under conditions that are less toxic than those arising due to CPT or HU treatment.

In contrast to the results with RAD54, RAD55 and RAD57, growth rates indicate that SGS1 is episatic to SHU2, RNH202 is epistatic to SHU1 and MUS81 is epistatic to both SHU1 and SHU2. These differences are interesting because the SHU genes are thought to function as a complex [31]. Although we are currently unable to establish a mechanistic explanation for this result, we hypothesize that these differences reflect the dependency of the RAD51-SGS1, RAD52-MUS81 and RNase H2 pathways on SHU1 and SHU2 and thus reflect the differences in the substrates that Shu1 and Shu2 act on in each pathway.

Finally, RNase H2 was found to suppress genome instability. All three RNase H2 mutants displayed a hyper-recombination phenotype using an assay system that employs three related reporter constructs. We suggest that this phenotype is directly related to the role of RNase H2 in removing ribonucleotides mistakenly incorporated into the genome. In its absence, the ribonucleotides may block the progress of replication forks leading to fork arrest or collapse. These damaged forks can presumably be restarted through an Sgs1-Top3-Rmi1-dependent recombination pathway. Thus, we suggest that loss of RNase H2 creates a need for recombinational repair due to an increase in the rate of stalled or collapsed replication forks.

Figure 6 presents a model summarizing the relationship between two major recombinational repair pathways. The left-hand pathway consists of Rad52-54-55-57, Shu1-2, and Mus81-Mms4, while the right-hand pathway consists of Rad51-54-55-57, Shu1-2, and Sgs1-Top3-Rmi1. The sizes of circles enclosing the protein names reflect the apparent importance of each protein that derived from the results of doubling times. The right-hand pathway appears to depend equally on Rad51, Rad54, Rad55 and Shu2. The Rad57 in the right-hand pathway is a little smaller whereas Shu1 is much smaller. The left-hand pathway depends primarily on Rad52 and Rad54 and somewhat less on Rad55 and Rad57. By taking previous and current studies into account, we suggest that bypass of a given pathway can occur by way of crosstalk between these two pathways indicated by the “shake-hands” symbol between Rad51 and Rad52. To account for the suppression results in Fig. 2B–D, we propose that both the Sgs1-Top3-Rmi1 and the RNase H2 complexes function upstream of recombination. A possible role for Sgs1-Top3-Rmi1 here is in the 5′-end resection of DNA ends to generate the 3′-single stranded overhangs [69]. We note that a positive role for RNase H2 upstream of recombination is currently unknown. However, it is possible that RNase H2 functions similarly. For example, RNase H2 may process DNA lesions that contain ribonucleotides to generate a single-strand break so that they can be acted upon by the HR machinery.

Figure 6. A model representing the relationships between RAD51, RAD52, RAD54, RAD55, RAD57, SHU1, SHU2 and SGS1, MUS81 and RNH202.

The model proposes that there are two major recombinational repair pathways consisting of Rad52-54-55-57, Shu1-2, and Mus81-Mms4 (left), and Rad51-54-55-57, Shu1-2, and Sgs1-Top3-Rmi1 (right). These pathways are derived from the results of genetic interaction analyses shown in Fig. 2A–D and Fig. 4. Based on the results that the drug sensitivities of rad54Δ, rad55Δ and rad57Δ were suppressed by deletion of SGS1 or RNH202 as shown in Fig. 2B–E, we propose that both Sgs1-Top3-Rmi1 and RNase H2 complexes function upstream of recombination as indicated at the top of the model.

In this study, we have examined the roles of the recombinational repair pathways distinguished by SGS1, MUS81 and RNH202. Clearly, a full understanding of the pathways will require additional studies. However, our systematic genetic analyses provide insight into the composition of the pathways and provide a framework to understand their genetic interactions.