Saccharomyces cerevisiae Rrm3p DNA Helicase Promotes Genome Integrity by Preventing Replication Fork Stalling: Viability of rrm3 Cells Requires the Intra-S-Phase Checkpoint and Fork Restart Activities

Abstract

Rrm3p is a 5′-to-3′ DNA helicase that helps replication forks traverse protein-DNA complexes. Its absence leads to increased fork stalling and breakage at over 1,000 specific sites located throughout the Saccharomyces cerevisiae genome. To understand the mechanisms that respond to and repair rrm3-dependent lesions, we carried out a candidate gene deletion analysis to identify genes whose mutation conferred slow growth or lethality on rrm3 cells. Based on synthetic phenotypes, the intra-S-phase checkpoint, the SRS2 inhibitor of recombination, the SGS1/TOP3 replication fork restart pathway, and the MRE11/RAD50/XRS2 (MRX) complex were critical for viability of rrm3 cells. DNA damage checkpoint and homologous recombination genes were important for normal growth of rrm3 cells. However, the MUS81/MMS4 replication fork restart pathway did not affect growth of rrm3 cells. These data suggest a model in which the stalled and broken forks generated in rrm3 cells activate a checkpoint response that provides time for fork repair and restart. Stalled forks are converted by a Rad51p-mediated process to intermediates that are resolved by Sgs1p/Top3p. The rrm3 system provides a unique opportunity to learn the fate of forks whose progress is impaired by natural impediments rather than by exogenous DNA damage.

DNA replication is a fragile process. The replication machinery is constantly encountering obstacles, such as protein-DNA complexes, DNA secondary structures, and extrinsically and intrinsically induced DNA damage, all of which can inhibit fork progression. If these stalled forks are not restarted, they can irreversibly arrest. This inability to restart replication, often called fork collapse, probably results from damage or disassembly of the replication complex. In addition, stalled forks can break. Thus, the integrity of the genome is critically dependent on the cell's ability to detect and restart stalled forks and to repair broken forks.

Eukaryotes have checkpoints that detect and stabilize stalled replication forks. In Saccharomyces cerevisiae, DNA replication in the presence of sublethal concentrations of hydroxyurea (HU), which depletes nucleotide precursor pools, or methyl methanesulfonate (MMS), which alkylates DNA, results in the slowing of DNA replication, a response known as the intra-S-phase checkpoint (57). Activation of the intra-S-phase checkpoint induces phosphorylation of the DNA polymerase-primase complex (50), inhibits firing of late activated replication origins (63), and increases transcription of genes involved in DNA repair (90). When cells lacking the intra-S-phase checkpoint are treated with HU or MMS, replication forks are irreversibly stalled, unable to progress, even upon removal of the DNA-damaging agent (46, 80). The most critical function of the intra-S-phase checkpoint is probably to stabilize stalled forks so that they can resume replication (81).

In addition to the intra-S-phase checkpoint, eukaryotes have DNA damage checkpoints that inhibit the onset of mitosis in the presence of DNA lesions, such as double-strand breaks (DSBs) (Fig. 1) (24). Stalled replication forks generate an intra-S-phase response, but they can also break, which activates the DNA damage checkpoint and inhibits mitosis. Some investigators classify the response to DNA damage acquired during S phase with the intra-S-phase checkpoint. In this paper, the response to S-phase-induced damage is considered a DNA damage checkpoint response.

FIG. 1.

Intra-S-phase and DNA damage checkpoints. Organization of genes involved in the intra-S-phase and DNA damage checkpoints. The boxed genes were deleted and tested for their ability to confer lethality or slow growth in an rrm3 background. Genes whose deletion had no effect on growth of rrm3 cells are in a dashed box. Genes whose deletion resulted in a modest growth defect in an rrm3 strain are in thin-lined boxes. Genes whose deletion resulted in no or extremely slow growth of rrm3 cells are in boldface and thick-lined boxes. Finally, genes whose deletion resulted in no or extremely slow growth of rrm3 cells at 23°C are marked with an asterisk.

DNA checkpoints are signal transduction pathways, requiring sensors to detect DNA lesions, transducers to integrate and transmit signals, and effectors that set in motion downstream events (Fig. 1). In S. cerevisiae, the ATR-like Mec1p kinase and its binding partner, Ddc2p, are sensors for stalled replication forks and for DNA damage (90). Likewise, Rad53p, a protein kinase that is phosphorylated and activated in a Mec1p-dependent manner, is an effector kinase for both the intra-S-phase and DNA damage checkpoints (90). However, the DNA damage checkpoint has an additional effector kinase, Chk1p (62). Moreover, the key transducers for the two pathways are different: Mrc1p functions in the intra-S-phase checkpoint, and Rad9p functions in the DNA damage checkpoint (1, 21).

There are additional proteins that affect the cell's ability to respond to stalled replication forks but whose functions are not limited to checkpoint roles. Sgs1p, a 3′-to-5′ DNA helicase, is important for full activity of the intra-S-phase checkpoint but does not affect the DNA damage checkpoint (25). Sgs1p is closely associated with the replication fork and may stabilize stalled forks in addition to serving as a sensor (15, 25). In addition, Sgs1p is thought to be involved in fork restart (38). The intra-S-phase checkpoint is also impaired in cells lacking Srs2p, another 3′-to-5′ DNA helicase (45). In vitro, Srs2p displaces Rad51p, a strand-annealing protein, from DNA filaments (42, 86), a result that explains its in vivo recombination-inhibiting activity (61). The Mre11p/Rad50p/Xrs2p (MRX) complex, an endo/exonuclease, functions in both the intra-S-phase and DNA damage checkpoints, as well as in homologous recombination (HR) and nonhomologous end joining (NHEJ) (18, 31, 84).

There are two recombination pathways that are proposed to restart broken replication forks. The first pathway is Rad51p and Rad52p dependent and has the helicase-topoisomerase complex Sgs1p/Top3p and the endonuclease Mus81p/Mms4p in parallel pathways downstream of a Rad51p-initiated event (6, 38). The second pathway is break-induced replication (BIR), which is Rad52p and Rad59p dependent but Rad51p independent (77). In BIR, broken replication forks can be reinitiated by strand invasion and replication can continue to the end of the chromosome or until this replication complex converges with another replication fork (41). Although BIR has been shown to repair DSBs (49), there is as yet no direct evidence for BIR having a role in restart of stalled replication forks.

Rrm3p, a 5′-to-3′ DNA helicase (36), is a member of the Pif1p subfamily of DNA helicases that are highly conserved among eukaryotes (9). RRM3 is not an essential gene, but in its absence, replication forks pause at over 1,000 discrete sites, including multiple sites in each of the ∼150 ribosomal DNA (rDNA) repeats, tRNA genes, centromeres, telomeres, and the silent mating-type loci (35-37). Fork breakage and recombination are both increased at sites of rrm3 pausing (35-37, 39). All sites affected by Rrm3p are assembled into nonnucleosomal protein-DNA complexes. At the silent mating-type loci, tRNA genes, and rDNA, disruption of these complexes eliminates the site's dependence upon Rrm3p during DNA replication (35, 82a). Rrm3p is telomere and rDNA associated in vivo (36, 37), and mutation of an invariant amino acid in an ATP binding motif eliminates its replication functions (36, 37). Thus, Rrm3p likely acts directly and catalytically to promote replication past protein-DNA complexes.

Most of our understanding of the events associated with replication fork stalling and restart comes from studying the cell's response to exogenous DNA damage. However, in every S phase, replication forks normally encounter nonnucleosomal protein-DNA complexes that affect fork progression. Analysis of the events needed to detect, stabilize, and restart stalled replication forks in rrm3 cells provides a unique opportunity to understand how cells deal with these natural impediments to fork progression. In this paper, we use a genetic approach to determine which pathways are required for viability of rrm3 cells. We report that deletion of RRM3, in combination with genes involved in the intra-S-phase and DNA damage checkpoints, replication fork restart, and homologous recombination, confers lethality or slow growth. Our data suggest that restart of rrm3 stalled forks can occur by a Rad51p-dependent event but only if cells contain Sgs1p/Top3p.

MATERIALS AND METHODS

Yeast methods.

All diploid strains were congenic to YPH501 (71). Mutants were constructed by the one-step gene disruption method (47) with plasmids or PCR fragments. Diploid strains carried the centromere plasmid pIA20, which contains the RRM3, URA3, and ADE3 genes. YPH501-derived diploids that were heterozygous for indicated deletions and carry pIA20 were dissected with a Singer MSM 200 micromanipulator. Genotypes of spore clones were determined by replica plating, assaying for the markers used to delete the genes. For the sectoring assay, nonselective plates containing 0.5% yeast extract, 2% peptone, and 4% glucose were streaked with spore products and the cells were grown at 30°C for 3 days and then maintained at 4°C for 3 days to allow for color development. In the growth assay, plates containing 5-fluoroorotic acid (FOA) were streaked with spore products and the cells were allowed to grow at 30°C for 3 days. For microscopic analyses, freshly sporulated cells from microcolonies were stained with YOYO-1 as described by the manufacturer (Molecular Probes) and imaged by confocal microscopy. To determine sensitivity to agents that damage DNA, freshly sporulated cells were grown to log phase; serially diluted; and applied as spots to 75 or 100 mM HU, 0.01 or 0.02% MMS, or 5 or 10 μg of camptothecin (CPT) per ml or were treated with 20, 30, or 40 J of UV light per m². Cells were assayed after growth at 30°C for 3 days.

Gel assays.

Two-dimensional (2D) neutral-neutral agarose gels (10) were used to examine replication intermediates, using conditions described in reference 37. Western blot analysis to examine Rad53p phosphorylation was carried out as recommended by A. Pelliciloi and M. Foiani using modifications of procedures described in reference 58.

RESULTS

Experimental design.

We used a genetic strategy to identify pathways that allow cells to survive the DNA lesions incurred during DNA replication in the absence of Rrm3p. Specifically, we identified genes whose mutation led to lethality or slow growth in an rrm3 background. We sporulated and dissected diploid strains that were heterozygous for complete open reading frame deletions of both RRM3 and the gene being tested. Thirty to 100 tetrads were dissected from each diploid strain. Because rrm3 cells exhibit increased recombination and Ty1 transposition (35-37, 39, 68), they can accumulate suppressor mutations (our unpublished results). Therefore, the starting diploids were also ade2 ade3 ura3 and carried plasmid pIA20, which contains the RRM3, ADE3, and URA3 genes (35) (Fig. 2A). An ade2 ade3 strain generates white colonies, while an ade2 ADE3 or ade2 ade3 strain bearing pIA20 generates red colonies. If the rrm3 strain is viable without gene X, then the rrm3 gene X-null strain will be viable upon loss of pIA20 and will generate red colonies with white sectors when streaked on nonselective medium (for example, Fig. 2B). Red colonies without white sectors indicate that the strain is not viable upon loss of pIA20. Likewise, genetic backgrounds that require the pIA20-borne RRM3 for survival will not grow on plates containing FOA, which kills cells expressing Ura3p. A genotype that did not generate visible colonies on FOA plates after 3 days at 30°C was considered to be nonviable or extremely slow growing. We also dissected diploids that did not carry pIA20, and for ease of presentation, most of the tetrads shown are from these diploids. For diploids that did not carry pIA20, synthetic phenotypes were inferred from the failure to recover or the slow growth of double mutant spores (for example, Fig. 2C).

FIG. 2.

Methods for determining synthetic lethality: rrm3 sgs1 cells are dead. (A) Schematic of pIA20 plasmid and sectoring assay. The CEN4 plasmid pIA20 contains the RRM3, URA3, and ADE3 genes. While ade2 ade3 cells generate white colonies, ade2 ade3 cells carrying pIA20 generate red colonies. Loss of pIA20 from ade2 ade3 cells is detected by the presence of white sectors in red colonies. (B) Sectoring assay. Low-adenine plates were streaked with freshly sporulated wild-type, rrm3, sgs1, and rrm3 sgs1 cells carrying pIA20, and the cells were allowed to grow at 30°C for 3 days and at 4°C for 3 days to allow red color development. (C) Tetrad dissection of diploids heterozygous for rrm3 and sgs1. In this and subsequent figures, the four spores from a given tetrad are in a vertical line and dissected spores were allowed to grow at 30°C for 3 days and genotyped by replica plating to test media. Assuming 2:2 segregation of the marker inserted into the deleted genes allows one to identify rrm3 sgs1 double mutants (boxed). (D) Confocal microscope images of YOYO-1-stained rrm3 sgs1 cells from microcolonies shown in panel C.

Both the intra-S-phase and DNA damage checkpoints are required for normal growth of rrm3 cells.

Cells lacking Rrm3p exhibit replication fork pausing and DNA breakage at many sites throughout the genome, yet are viable, with a close to wild-type growth rate (35-37). Consistent with their impaired replication and DNA breakage, rrm3 cells phosphorylate and activate Rad53p, suggesting that DNA checkpoints are activated in rrm3 cells (35). The Rad53p activation led us to determine if DNA checkpoints are important for viability of rrm3 cells.

Mec1p and Rad53p are, respectively, sensor and effector protein kinases for both the intra-S-phase and DNA damage checkpoints (Fig. 1). MEC1 and RAD53 are essential genes due to their roles in regulating deoxynucleoside triphosphate (dNTP) pools. However, their essential function is suppressed by deleting SML1, which encodes a repressor of ribonucleotide reductase (95). Thus, mec1 sml1 and rad53 sml1 cells are viable but checkpoint deficient. Recently, we reported that rrm3 mec1 sml1 cells are slow growing at 30°C and dead at 23°C, while rrm3 rad9 cells are alive at all temperatures (35). These data suggest that the intra-S-phase checkpoint but not the DNA damage checkpoint is essential for viability of rrm3 cells. To confirm and extend this interpretation, we deleted other genes important for the intra-S-phase and DNA damage checkpoints in an rrm3 background.

The rrm3 rad53 sml1 cells were viable at both 23 and 30°C but grew more slowly than either single mutant at both temperatures (Fig. 3A and B). Because Rad53p functions in both the intra-S-phase and DNA damage checkpoints, we next examined genes specific to only one checkpoint. Chk1p, a second effector kinase for the DNA damage checkpoint, does not function in the intra-S-phase checkpoint (Fig. 1) (30, 62). Although rrm3 chk1 cells were viable, they grew somewhat more slowly than either single mutant (Fig. 3C and D). RAD24, RAD17, MEC3, and DDC1 are sensors for the DNA damage checkpoint and have at most a minor role in the intra-S-phase checkpoint (Fig. 1) (51). Rad24p interacts with replication factor C (RF-C) and this complex acts as a clamp loader for the Rad17p/Mec3p/Ddc1p complex. To test the importance of this complex, we generated rrm3 rad24 cells. The rrm3 rad24 cells grew at a comparable rate to that of either single mutant (Fig. 3E). Rad9p, which functions downstream of Rad24p, is a transducer for the DNA damage checkpoint but does not function in the intra-S-phase checkpoint (Fig. 1) (51). As demonstrated previously (35), rrm3 rad9 cells were viable. However, the double mutant had a minor growth defect compared to single mutants (Fig. 3F).

FIG. 3.

The intra-S-phase and DNA damage checkpoints are required for normal growth of rrm3 cells. Tetrad dissection of indicated heterozygous diploids. In panels A, C, E, F, and G, hexagons indicate rrm3 cells; circles indicate rad53 sml1, chk1, rad24, rad9, and mrc1 cells; and boxes indicate rrm3 rad53 sml1, rrm3 chk1, rrm3 rad24, rrm3 rad9, and inferred rrm3 mrc1 mutants. In panel A, SML1 was deleted along with RAD53, as deletion of SML1 is required for viability of rad53 cells. In panel E, rrm3 sgs1 spores are in diamonds. For panels F and G, tetrads that could not be genotyped are unmarked. In panels B and D, complete medium was streaked with wild-type (WT), rrm3, rad53 sml1, rrm3 rad53 sml1, chk1, and rrm3 chk1 cells and the cells were allowed to grow for 2 days at 30°C.

MRC1, identified as a gene whose mutation confers sensitivity to HU, is thought to be a transducer for the intra-S-phase checkpoint but to have no role in the DNA damage checkpoint (Fig. 1) (1). So far it is the best candidate for a gene whose checkpoint function is limited to the intra-S-phase checkpoint. The rrm3 mrc1 cells were dead (Fig. 3G). Unlike most synthetically lethal double mutants (see next section), rrm3 mrc1 cells never formed microcolonies.

Taken together, these data indicate that the intra-S-phase checkpoint is important for viability of rrm3 cells. The DNA damage checkpoint, while not essential, also contributes to the normal growth of rrm3 cells.

Genes with roles in checkpoints and fork restart are required for viability of rrm3 cells.

Sgs1p is a member of the highly conserved RecQ subfamily of 3′-to-5′ DNA helicases (93) that is proposed to play roles in sensing, stabilizing, and restarting stalled replication forks (15, 25, 38). To assess if Sgs1p is required to sense the paused and broken forks characteristic of rrm3 cells, we generated rrm3 sgs1 cells. Although rrm3 sgs1 cells were not viable, 90% of the rrm3 sgs1 spores divided at least once and in many cases (35%) formed visible microcolonies (Fig. 2B and C). Similar results are reported in an accompanying paper (67). Microscopic analysis of rrm3 sgs1 cells indicated that most (77%) arrested as large budded cells with DNA near or stretched across the bud neck, indicative of a late S- or G₂/M-phase arrest (Fig. 2D). The synthetic lethality of rrm3 sgs1 cells was also reported in genomewide analyses for genes that are synthetically lethal with sgs1 (54, 82).

Top3p is a type 1 topoisomerase that interacts physically and genetically with Sgs1p (28). Like Sgs1p, Top3p is thought to sense DNA lesions incurred during S phase, but not during the G₁ or G₂/M phases (Fig. 1) (13). As expected, if the Sgs1p/Top3p complex sensed the DNA damage in rrm3 cells, rrm3 top3 cells were dead (Fig. 4A). Although rrm3 top3 cells did not form microcolonies, top3 cells are very slow growing (88) (Fig. 4A). As shown in the figure, this dissection confirmed that deletion of SGS1 relieves the slow growth of top3 cells (28). The rrm3 top3 lethality was specific as rrm3 top1 cells were viable (Fig. 4B). As shown in panel B, this dissection confirmed that sgs1 top1 cells are slow growing (48).

FIG. 4.

Interactions of rrm3 with sgs1 and with genes that are synthetically lethal with sgs1. Tetrad dissection of indicated heterozygous diploids. In panels A through D, circles indicate single deletion of the gene of interest (gene X), boxes indicate rrm3 gene X double mutants, and, for comparison, diamonds indicate sgs1 gene X double mutants. In panels A, B, and D, unlabeled microcolonies are rrm3 sgs1 double mutants. In cases in which only two spores grew, inferred mutants are marked if their genotype could be determined.

Srs2p is a 3′-to-5′ DNA helicase that can dislodge Rad51p molecules from single-stranded DNA (ssDNA) in vitro, consistent with in vivo data showing that recombination is increased in its absence (42, 61, 86). Srs2p is also involved in sensing replicative stress: srs2 cells have a reduced ability to activate Rad53p in response to MMS (45). srs2 is synthetically lethal with both sgs1 and top3 (29, 44). While rrm3 srs2 spore clones were not viable, 35% formed microcolonies (Fig. 4C). Like rrm3 sgs1 cells, rrm3 srs2 cells arrested as large budded cells, suggesting a late S- or G₂/M-phase arrest (data not shown). Similar results are reported in an accompanying paper (67).

The MRX complex, consisting of Mre11p, Rad50p and Xrs2p, has multiple functions, participating in HR, NHEJ, telomere maintenance, and the intra-S-phase checkpoint (17). Deletion of any one or all three of the MRX components confers identical phenotypes, including reduced ability to activate Rad53p and to slow DNA replication in response to HU or bleomycin treatment (18, 31). Cells lacking the MRX complex are synthetically lethal or very slow growing, with sgs1, top3, and srs2 mutations (40, 69). Likewise, rrm3 mrx cells were inviable: 80% of the rrm3 mre11, rrm3 rad50, and rrm3 xrs2 spore clones formed visible microcolonies (for example, see Fig. 4D and 5B for rrm3 mre11 and rrm3 xrs2). Similar results for rrm3 rad50 and rrm3 mre11 are reported in an accompanying paper (67).

FIG. 5.

RRM3 genetic interactions. (A) List of RRM3 genetic interactions. The result “Lethal” indicates that upon dissection, the rrm3 gene X double mutant did not form visible colonies after 3 days at 30°C or formed microcolonies that grew extremely slowly. ++, normal growth rate; +−, moderately reduced growth rate; CS, cold sensitive at 23°C; 1, mutant alone is slow growing, but slow growth is not exacerbated by deletion of RRM3. (B) Assessment of synthetic lethal interactions. In plate sections i to viii, respectively, FOA was streaked with rrm3, rrm3 mrc1, rrm3 mre11, rrm3 rad50, rrm3 xrs2, rrm3 sgs1, rrm3 srs2, and rrm3 top3 spore products carrying pIA20 (RRM3 URA3 ADE3) and the plate was assessed for growth after 3 days at 30°C. Lack of growth on FOA indicates that the strain is inviable in the absence of pIA20.

Analysis of RRM3 genetic interactions.

The observed synthetic lethal interactions, summarized in Fig. 5A, were confirmed by the inability of double mutants to be viable upon loss of the RRM3-containing plasmid pIA20. FOA was streaked with rrm3, rrm3 mrc1, rrm3 mre11, rrm3 rad50, rrm3 xrs2, rrm3 sgs1, rrm3 srs2, and rrm3 top3 spore products carrying pIA20. Although the rrm3 spore clone grew well on FOA plates, rrm3 mrc1, rrm3 mre11, rrm3 rad50, rrm3 xrs2, rrm3 sgs1, rrm3 srs2, and rrm3 top3 spores did not form visible colonies after 3 days at 30°C (Fig. 5B, plate sections i to viii). These synthetic lethalities cannot be attributed to spore germination defects. After 5 days at 30°C, rrm3 sgs1, rrm3 srs2, and rrm3 mrx strains generated a heterogeneous mix of microcolonies. We interpret this late growth as indicating that rrm3 sgs1, rrm3 srs2, and rrm3 mrx strains grow extremely slowly, and this slow growth allows suppressor mutations to arise.

Unlike rrm3 cells, sgs1 and srs2 cells do not exhibit replication fork pausing in rDNA.

Deletion of RRM3 or SGS1 results in increased rDNA recombination, including an increase in the abundance of recombination-generated rDNA circles (28, 37, 39, 72). In rrm3 cells, increased rDNA recombination is probably a secondary consequence of faulty replication (36, 37). Sgs1p is physically associated with replication forks, even in the absence of DNA damage (15) and also has a genomewide effect on fork movement, with forks moving more quickly in its absence (87). Thus, a possible explanation for the lethality of rrm3 sgs1 strains is that the Sgs1p DNA helicase also has a role in rDNA replication.

To determine if Sgs1p has Rrm3p-like effects on fork progression, we used 2D gel electrophoresis to examine rDNA replication in sgs1 cells. DNA was digested with StuI, which liberates a 5-kb fragment containing the replication fork barrier (RFB) at the left end of the fragment, the 5S rDNA, and an origin of DNA replication (ARS) in the middle, as well as the 5′ end of the 35S rDNA at the right end of the fragment (Fig. 6A and B). As shown previously (37) (Fig. 6C), in rrm3 cells, replication forks paused at specific sites, including the 5S rDNA (pause c), the rDNA ARS (pause d), and the promoter region of the 35S rRNA gene (pause e) (Fig. 6D, panel 2). Compared to wild-type cells, rrm3 cells had more putative Holliday junctions (HJ) as well as increased accumulation of forks stalled or converged (labeled X) at the RFB, and both of these intermediates had an increased probability of breakage (Fig. 6D; compare panels 1 and 2). In contrast, in sgs1 cells, the pattern of rDNA replication was similar to that of the wild type (Fig. 6D; compare panels 1 and 3). These results confirm a recently published 2D gel analysis study that also concluded that replication forks do not stall in rDNA in sgs1 cells (87).

FIG. 6.

sgs1 and srs2 cells have a wild-type rDNA replication pattern. (A) Schematic of a small portion of the ∼150-repeat rDNA array. Positions of StuI sites are indicated. The arrowhead and arrow indicate, respectively, the 5S rRNA and the 35S rRNA coding regions. The position of the ARS or replication origin is also indicated. The ARS is active as an origin in only 10 to 20% of rDNA repeats. (B) Schematic of rDNA replication. The left panel shows the progression of rightward-moving forks, with rrm3-dependent pauses indicated by broken lines and lowercase letters c, d, and e in repeats that do not have an active origin. The panel on the right shows bidirectional replication through the StuI fragment in the subset of repeats that have an active origin. StuI fragments with active origins are labeled “BU” in panel C. Leftward-moving forks arrested at the RFB and rightward-moving forks converging at the RFB generate an “X”-like structure (X in panel C). (C) Replication pattern of StuI-digested rDNA from wild-type and rrm3 cells. Lowercase letters c, d, and e correspond to pauses in the 5S rDNA, ARS, and 35S rRNA initiator sequence, respectively (B). BR, breakage products; BU, bubble arc (visible in darker exposures); HJ, putative Holliday junctions; X, converged forks. (D) 2D gel analysis of rDNA from wild-type (panel 1), rrm3 (panel 2), sgs1 (panel 3), and srs2 (panel 4) cells was digested with StuI. Gel conditions, Southern blotting, and probing were as previously described (37).

Like rrm3, srs2 cells have a hyperrecombination phenotype (37, 39, 61, 89). In addition, Srs2p has been shown to remove protein from DNA (42, 86). Since Rrm3p is needed to move forks past protein-DNA complexes (35), the rrm3 srs2 lethality might reflect overlapping functions of these two helicases in removing proteins from DNA during DNA replication. However, using 2D gels, we found that rDNA replication in srs2 cells was indistinguishable from wild-type replication (Fig. 6D, compare panels 1 and 4). Taken together, these data suggest that the lethality of rrm3 sgs1 and rrm3 srs2 cells is probably not due to overlapping functions of these helicases in promoting fork progression through rDNA.

Rrm3p does not play a critical role in DNA damage repair.

At sublethal concentrations, HU stalls replication forks while CPT and MMS lead to replication-dependent DSB formation (8, 52, 83). UV light creates pyrimidine dimers that can impede replication (27). Deletion of SGS1, SRS2, or the MRX complex confers sensitivity to all of these treatments, phenotypes that are often attributed to defects in DNA repair (7). Cells lacking Mrc1p are also UV, MMS, and HU sensitive (1, 7).

To determine if rrm3 cells have similar repair defects, we tested their sensitivity to HU, MMS, UV, and CPT by spotting plates containing different concentrations of HU, MMS, or CPT with serial dilutions of exponentially growing rrm3 or control cells (see Materials and Methods). Alternatively, cells were exposed to UV light immediately after plating. The rrm3 cells grew as well as wild-type cells after UV treatment or on HU plates (Fig. 7B). While rrm3 cells were modestly MMS and CPT sensitive, their sensitivity to these drugs was considerably lower than that reported for the other mutants (see, for example, comparison of rrm3 with sgs1 in Fig. 7B). These data suggest that, unlike sgs1, srs2, mrc1, mre11, rad50, or xrs2 cells, rrm3 cells have wild-type or nearly wild-type abilities to repair exogenously generated DNA damage.

FIG. 7.

rrm3 has little or no sensitivity to DNA-damaging agents. (A) Summary of data for rrm3, sgs1, and srs2 sensitivities to HU, MMS, UV light, and CPT. A superscript 1 indicates that our results are in agreement with previously published results for sgs1 (summarized in reference 7). A superscript 2 indicates previously published results (summarized in reference 7 for srs2). Wt, wild type. (B) Wild-type, rrm3, and sgs1 cells were grown to an optical density at 660 nm of 0.8, serially diluted, and plated onto media containing either 75 mM HU, 0.01% MMS, or 10 μg of CPT per ml. Alternatively, cells were plated on complete medium and treated with 30 J of UV light per m². YC cultures represent the same dilutions plated on complete medium without DNA-damaging agents. Cells were grown at 30°C for 3 days. (C) Tetrad dissection of diploids heterozygous for rrm3, sgs1, and rad27. Circles indicate rad27 single mutants, and boxes indicate rrm3 rad27 double mutants. Unlabeled microcolonies indicate rrm3 sgs1 double mutants.

To provide additional evidence that rrm3 sgs1 or rrm3 srs2 lethalities were not due to overlapping functions in DNA repair, we determined the phenotypes of rrm3 rad27 cells. RAD27 encodes an exonuclease that is involved in completion of lagging strand synthesis (33, 60, 74). Deletion of RAD27 leads to lesions that are repaired by the homologous recombination machinery (76), and its absence leads to lethality or very slow growth in sgs1 and srs2 cells as well as in strains lacking the MRX complex (20, 82). The rrm3 rad27 cells were viable with a growth rate similar to that of rad27 cells (Fig. 7C). Taken together, these data suggest that Rrm3p does not play an important role in DNA damage repair. In addition, its synthetic interactions with sgs1, srs2, mrc1, and the MRX complex are unlikely due to its having an overlapping function in DNA repair with these genes.

Inactivation of the S-phase checkpoint does not rescue the lethality of rrm3 sgs1 or rrm3 srs2 cells.

Most of the rrm3 sgs1 and rrm3 srs2 cells arrested with a late S- or G₂/M-phase nuclear morphology (Fig. 2D). Because cells arrested by the DNA damage checkpoint show a similar morphology (91), the lack of division of these double mutant strains might be due to a checkpoint-mediated cell cycle arrest. In addition, Srs2p is required to recover from checkpoint-mediated arrest after DNA damage: in its absence, cells complete DNA repair but do not inactivate Rad53p (85). Since rrm3 cells activate Rad53p (35), the lethality of rrm3 srs2 cells might be a consequence of their inability to turn off the Rad53p-mediated checkpoint. If rrm3 srs2 and rrm3 sgs1cells are checkpoint arrested, rather than dead, elimination of the DNA checkpoints might rescue their growth defects. In contrast to this expectation, rrm3 srs2 (or sgs1) rad9, rrm3 srs2 (or sgs1) rad24, rrm3 srs2 (or sgs1) mec1 sml1, and rrm3 srs2 (or sgs1) mrc1 cells were all dead (Fig. 8A). Thus, the lethality of rrm3 sgs1 and rrm3 srs2 cells is not due solely to their activating the intra-S-phase or DNA damage checkpoints.

FIG. 8.

Deletion of RAD51 (but not RAD52) suppresses some but not all rrm3 synthetic phenotypes. (A) Summary of synthetic lethal phenotypes and their ability to be suppressed by deletion of recombination and checkpoint genes. (B to E) Deletion of RAD51 but not RAD52 rescues the rrm3 sgs1 and rrm3 srs2 lethal phenotypes. Tetrad dissections of indicated heterozygous diploids. Diamonds indicate rrm3 rad51 or rrm3 rad52 double mutants; circles indicate rrm3 sgs1 or rrm3 srs2 double mutants; and boxes indicate rrm3 sgs1 rad51, rrm3 srs2 rad51, rrm3 sgs1 rad52, and rrm3 srs2 rad52 triple mutants. (F) rrm3 rad52 cells have a modest growth defect. Complete medium was streaked with wild-type (WT), rrm3, rad52, and rrm3 rad52 cells, and the cells were allowed to grow for 2 days at 30°C. The right panel shows a close-up of the colonies. (G and H) Deletion of RAD51 does not rescue rrm3 mrc1 lethality and rrm3 rad53 sml1 slow growth. Diploid cells harboring the CEN4 plasmid pIA20, homozygous for deletion of RRM3 and heterozygous for deletion of indicated genes, were dissected. Due to the wild-type copy of RRM3 in pIA20, triple and quadruple mutants were rare. (G) rrm3 mrc1 spores are circled, and rrm3 mrc1 rad51 spores are boxed. Only 1 of 12 rrm3 mrc1 rad51 spores formed a microcolony. (H) rrm3 rad53 sml1 spores are circled, and rrm3 rad53 sml1 rad51 spores are boxed.

Deletion of RAD51 (but not RAD52) rescues some of the rrm3 synthetic lethal phenotypes.

The lethality of sgs1 (or top3) srs2, sgs1 mus81 (or mms4), and srs2 rad54 strains is rescued by deletion of either RAD51 or RAD52 (summarized in reference 23). This rescue suggests that toxic recombination intermediates are responsible for lethality in these genetic backgrounds. To determine if the lethality of rrm3 double mutant strains was also due to accumulation of toxic recombination intermediates, we deleted RAD51 from these strains (Fig. 8A). Deletion of RAD51 rescued the lethality of rrm3 sgs1, rrm3 srs2 (Fig. 8B and C), and rrm3 top3 cells (data not shown). These triple mutant strains grew about as well as each of the single mutants. In contrast, the rrm3 mre11, rrm3 rad50, rrm3 xrs2, and rrm3 mrc1 lethalities were not relieved by deleting RAD51 (Fig. 8G) (data not shown). Likewise, the synergistic slow growth of rrm3 rad53 sml1 cells was not rescued by deletion of RAD51 (Fig. 8H). These results suggest that the synthetic lethality of rrm3 with sgs1, top3, and srs2 is due to accumulation of toxic recombination intermediates. However, the events that impose lethality or slow growth on rrm3 in combination with mrc1, rad53 sml1, mre11, rad50, and xrs2 are likely upstream of recombination. Similar suppression or nonsuppression of rrm3 sgs1, rrm3 srs2, rrm3 rad50, and rrm3 mre11 by deletion of RAD51 is reported in an accompanying paper (67), and for rrm3 sgs1, suppression occurred with deletion of RAD51 (54).

In yeast, some homologous recombination is Rad51p independent, but Rad52p is needed for essentially all homologous recombination (77). As described above, deletion of either RAD51 or RAD52 rescues the synthetic lethality of other double mutants. In contrast, deletion of RAD52 did not rescue the lethality of the rrm3 sgs1 and rrm3 srs2 double mutant strains after 3 days at 30°C (Fig. 8D and E), although some growth was observed after 5 days at 30°C (data not shown). Partial suppression of these synthetic lethalities by deletion of RAD52 is also reported in reference 67. Next we asked if eliminating single-strand annealing (SSA), a Rad51p-independent HR pathway that is important for DSB repair between repeated DNA sequences, including in the rDNA (34, 56, 65) or NHEJ, which can rejoin DNA breaks, can relieve the lethality of rrm3 double mutant strains. We found that deletion of RAD1, a gene needed for SSA, or LIG4, a gene required for NHEJ (64, 79), did not rescue the lethality of these strains (data not shown).

Deletion of RAD51 has no effect on rrm3-dependent replication defects or on activation of Rad53p.

Since deletion of RAD51 suppressed the lethality of rrm3 sgs1 cells, we wished to determine if it also suppressed the replication defects, increased recombination and DNA breakage characteristic of rrm3 cells. Using 2D gels (Fig. 9A), we found that rrm3 sgs1 rad51 cells (panel 4) had the same rDNA replication defects as rrm3 cells (Fig. 6D, panel 2), including replication pauses, increased abundance of putative Holliday junctions, and broken replication intermediates. In contrast, the pattern of rDNA replication in rad51 (panel 1) and sgs1 rad51 (panel 3) cells was indistinguishable from that of wild-type cells (Fig. 6D, panel 1). Therefore, the stalled and broken replication forks that typify the rrm3 phenotype are unlikely to be the direct cause of death in rrm3 sgs1 cells. Rather, lethality arises when these stalled or broken forks are processed by Rad51p into an intermediate that is toxic in the absence of Sgs1p/Top3p or Srs2p.

FIG. 9.

Deletion of RAD51 has no effect on rrm3-dependent replication defects or activation of Rad53p. (A) 2D replication gel analysis of StuI-digested rDNA as previously described (37) from rad51, rrm3 rad51, sgs1 rad51, and rrm3 sgs1 rad51 cells. See the legend to Fig. 6 for details. (B) Western blot analysis of Rad53p hyperphosphorylation. Protein samples from wild-type (Wt), rrm3, sgs1, rad51, rrm3 rad51, sgs1 rad51, and rrm3 sgs1 rad51 cells were analyzed by Western blotting. Detection of Rad53p with anti-Rad53p antibodies is as previously described (35, 58). An arrow points to Rad53p, and P indicates hyperphosphorylated forms of Rad53p.

Rad53p is hyperphosphorylated and activated in rrm3 cells (35). Since deletion of RAD51 suppressed the lethality of rrm3 sgs1 cells, we asked whether activation of Rad53p was also suppressed by deletion of RAD51. Phosphorylated Rad53p is detected as a smear of slower-migrating forms of Rad53p on Western blots (58). Protein preparations from wild-type, rrm3, sgs1, rad51, rrm3 rad51, sgs1 rad51, and rrm3 sgs1 rad51 cells were analyzed for their Rad53p phosphorylation status by Western analysis using anti-Rad53p antibodies (Fig. 9B). In wild-type cells, a single sharp band corresponding to Rad53p was detected, while all other strains, including rrm3 sgs1 rad51 cells, had hyperphosphorylated Rad53p (Fig. 9B). These data suggest that the DNA structures that activate the intra-S-phase checkpoint are still present in rrm3 sgs1 rad51 (and rrm3 srs2 rad51) cells. Thus, the structures that activate DNA checkpoints in rrm3 cells are likely the stalled and broken replication forks, not toxic recombination intermediates.

RAD51-independent HR is important for repair of rrm3-dependent DNA lesions.

DSBs that arise from broken replication forks should be excellent substrates for HR. As rrm3 cells have increased fork breakage (35-37), we asked whether specific recombination pathways are essential for viability of rrm3 cells. The RAD51 subgroup (RAD51, RAD54, RAD55, and RAD57) is involved in catalyzing the early steps of strand invasion during HR (77). However, rrm3 rad51, rrm3 rad54, rrm3 rad55, and rrm3 rad57 cells were all viable, with growth rates comparable to that of single mutant cells (for example, Fig. 8B and C). Likewise, neither NHEJ nor SSA is important for repair of rrm3 DNA damage, as rrm3 lig4 and rrm3 rad1 cells grew at the same rate as single mutant cells (Fig. 10E). Rad18p is an ssDNA-binding protein that is required for both error-free and error-prone postreplicative repair (PRR) (94). As rrm3 rad18 cells grew as well as single mutant cells (data not shown), PRR is also not important for growth or viability of rrm3 cells.

FIG. 10.

BIR contributes to rrm3 sgs1 rad51 and rrm3 srs2 rad51 cell growth. Shown are tetrad dissections of diploids homozygous for deletion of RRM3 and heterozygous for deletion of indicated genes. In panels A, B, D, and E, circles indicate rrm3 sgs1 rad51 and rrm3 srs2 rad51 triple mutants and boxes indicate rrm3 sgs1 rad51 rad59 and rrm3 srs2 rad51 rad59 or rrm3 sgs1 rad51 rad1 and rrm3 srs2 rad51 rad1 quadruple mutants. In panel B, diamonds indicate rrm3 rad59 double mutants. In panel E, diamonds indicate rrm3 rad1 double mutants. (C) rrm3 rad59 cells have a modest growth defect. Complete medium was streaked with wild-type (WT), rrm3, rad59, and rrm3 rad59 cells, and the cells were allowed to grow for 2 days at 30°C. The right panel shows a close-up of the colonies.

Rad52p is required for virtually all HR events (77). We confirmed our early finding that rrm3 rad52 cells were viable (37) but found that they grew somewhat more slowly than either single mutant (Fig. 8D, E, and F). Since rrm3 rad59 cells also grew somewhat slower than either single mutant (Fig. 10B and C), the HR that is important for optimal rrm3 growth might be Rad59p dependent (discussed in more detail in the next section).

Taken together, these data indicate that NHEJ, NER, SSA, and PRR are not needed to repair the DNA lesions acquired during DNA replication in the absence of Rrm3p. Rad52p-mediated HR is important but not essential for viability of rrm3 cells.

BIR contributes to repair of DNA lesions incurred in rrm3 cells.

The rrm3 rad52 cells had a slow-growth phenotype, and deletion of RAD52 only partially suppressed the lethality of rrm3 sgs1 or rrm3 srs2 cells (Fig. 8D and E). These data suggest that a Rad52p-dependent recombination pathway was important for repair of the DNA lesions incurred in rrm3 cells. Since rrm3 rad51, rrm3 sgs1 rad51, and rrm3 srs2 rad51 cells grew at normal rates (Fig. 8B and C), the Rad52p-dependent event that is important in the absence of Rrm3p must be Rad51p independent.

BIR and SSA are two Rad52p-dependent repair pathways that do not require Rad51p (77). In BIR, DSBs are repaired by strand invasion followed by replication from the break site to the end of the chromosome or until another replication fork is encountered (41, 93). BIR, a Rad51p-independent and Rad52p- and Rad59p-dependent process, has been proposed to restart collapsed replication forks (24, 77). Rad59p is a Rad52p-related protein that interacts with it and augments its activity (3, 4, 19, 59). Deletion of RAD59 dramatically reduces BIR and SSA events (70, 75). If BIR is responsible for repairing DNA damage incurred in rrm3 sgs1 (or srs2) cells, then eliminating Rad59p should confer lethality or near lethality to otherwise viable rrm3 sgs1 rad51 and rrm3 srs2 rad51 strains. Consistent with this prediction, rrm3 sgs1 rad51 rad59 (Fig. 10A) and rrm3 srs2 rad51 rad59 (Fig. 10B) quadruple mutants were slow growing. In addition, rrm3 rad59 cells grew more slowly than either single mutant (Fig. 10B and C). These data suggest that a Rad59p-dependent pathway is important for repairing the DNA lesions incurred in rrm3 cells.

SSA is also Rad51p independent and Rad52p and Rad59p dependent (77). However, SSA requires the Rad1p/Rad10p heterodimeric endonuclease (5). The rrm3 sgs1 rad51 rad1 and rrm3 srs2 rad51 rad1 quadruple mutants were viable (Fig. 10D and E), with growth rates similar to that of triple mutants. Therefore, SSA is not critical for the growth of rrm3 sgs1 (or srs2) rad51 cells. Rather, the viability of these cells is likely due to repair of rrm3 lesions by BIR.

The MUS81/MMS4 replication fork restart pathway is not required for rrm3 survival.

In the absence of Rrm3p, forks pause and break (35-37). There are two pathways that are thought to act downstream of Rad51p to restart stalled replication forks. One pathway requires SGS1 and TOP3, both of which are synthetically lethal with rrm3 (Fig. 2 and 4A). The other pathway requires MUS81 and MMS4, which encode a heterodimeric, structure-specific endonuclease that cleaves 3′ single-strand tails from duplex DNA (6, 38). The synthetic lethality of sgs1 (or top3) and mus81 (or mms4) is attributed to the absence of both replication fork restart pathways (23, 38). This synthetic lethality is eliminated by deletion of RAD51 or RAD52, suggesting that these pathways are both downstream of recombination (23). To test the contribution of the Mus81p/Mms4p fork restart pathway to the survival of rrm3 cells, we generated rrm3 mus81 cells and found that they were viable, growing about as well as single mutant cells (Fig. 11A). Similar results are reported in an accompanying paper (67). This experiment also confirmed the lethality of sgs1 mus81 cells (Fig. 11A). Therefore, the Mus81p/Mms4p pathway is not important for growth or viability of rrm3 cells.

FIG. 11.

Unlike sgs1, rrm3 is not lethal with mus81 or slx1. Tetrad dissection of diploids heterozygous for deletion of RRM3, SGS1, and MUS81 or SLX1. In panels A and B, circles indicate deletion of the gene of interest (gene X), boxes indicate rrm3 gene X double mutants, and, for comparison, diamonds indicate sgs1 gene X double mutants. In cases in which only two spores grew, inferred mutants are marked if their genotype could be determined.

Deletion of SGS1 or TOP3 is also lethal in cells lacking another structure-specific endonuclease encoded by SLX1 and SLX4 (53, 82). The Slx1p/Slx4p complex, which cleaves 5′ single-strand tails from duplex DNA (26), is thought to act upstream of recombination, since the lethality of sgs1 slx1 cells is not relieved by deleting RAD51 (23). Similar to rrm3 mus81 cells, the rrm3 slx1 cells were viable, with a growth rate similar to that of single mutants (Fig. 11B; note that sgs1 sxl1 spores were dead). Thus, the Slx1p/Slx4p complex is not important for repair of rrm3-dependent damage.

DISCUSSION

Cells lacking the Rrm3p DNA helicase incur dramatic increases in replication fork pausing at over 1,000 sites scattered throughout the genome. Broken replication forks are also readily detected in rrm3 cells (35-37). Here we identify the genes and processes that enable rrm3 cells to divide at close to wild-type rates despite this extensive DNA damage.

Given that Rad53p is hyperphosphorylated and activated in rrm3 cells (35), we anticipated that DNA checkpoints would be essential for rrm3 viability. Consistent with this expectation, rrm3 cells lacking Mec1p, the sensor kinase for both the intra-S-phase and DNA damage checkpoints, are dead at low temperatures and slow growing at 30°C (35). However, rrm3 cells lacking Rad53p, the effector kinase for both checkpoints, were viable, albeit slow growing, at all temperatures (Fig. 3A and B). This greater importance of Mec1p for rrm3 viability is consistent with the finding that recovery from exogenous replication stress is more dependent on Mec1p than on Rad53p (22). Deletion of other genes whose functions are critical for the DNA damage checkpoint had no (rad24) or smaller (chk1) deleterious effects on the growth of rrm3 cells (Fig. 3C, D, and E). However, lack of Mrc1p, a signal transducer limited to the intra-S-phase checkpoint (1), had the strongest synthetic effects of the genes tested. The rrm3 mrc1 cells were dead. Unlike rrm3 sgs1, rrm3 srs2, or rrm3 mrx cells, the rrm3 mrc1 spores never generated microcolonies, suggesting that these cells acquire or detect (or both) lethal damage in virtually all cell cycles (Fig. 3G and 4A). (Note that rrm3 top3 spores also did not generate microcolonies; however, a top3 mutation alone is extremely slow growing.) The rrm3 mrc1, rrm3 mre11, rrm3 rad50, and rrm3 xrs2 synthetic lethalities were not relieved by deletion of RAD51 (Fig. 8A and G), consistent with these lethalities being a consequence of DNA lesions acquired during S phase. Lack of suppression of rrm3 mre11 and rrm3 rad50 lethalities by deletion of RAD51 is also reported in reference 67. Likewise, the slow growth of rrm3 rad53 sml1 cells was not relieved by deleting RAD51 (Fig. 8H).

Even in the absence of DNA-damaging agents, Mrc1p is associated with the replication fork and contributes in an unknown manner to normal DNA replication (1, 55). Upon replication stress, Mrc1p remains fork associated but is phosphorylated, presumably by Mec1p. These phosphorylation events lead to the phosphorylation and activation of Rad53p. If Rrm3p and Mrc1p have partially redundant roles at the replication fork, it would explain why rrm3 mrc1 cells had a more severe synthetic phenotype than rrm3 cells lacking Mec1p or Rad53p. It was recently reported that dna2-2 is synthetically lethal with rrm3 (92). Dna2p is an essential nuclease/helicase involved in Okazaki fragment maturation (2, 12). Since dna2-2 cells exhibit fork progression defects in rDNA, its synthetic lethality with Rrm3p may reflect the deleterious consequence of additive defects in fork progression (92). Alternatively, since Dna2p also functions in DSB repair (11), this synthetic phenotype may reflect a role for Dna2p in repair of the broken forks generated in rrm3 cells.

We also demonstrate that rrm3 was synthetically lethal with sgs1 and srs2, both of which encode DNA helicases, as well as with three genes—mre11, rad50, and xrs2—that encode the MRX endo/exonuclease complex (Fig. 2, 4C and 4D, and 5B). Similar results are described in an accompanying paper (67). In addition, rrm3 was lethal in combination with top3 (Fig. 4A), which encodes a topoisomerase that acts in concert with Sgs1p. By the criterion of 2D gel analysis, neither Sgs1p nor Srs2p affected fork progression in the rDNA (Fig. 6D). In addition, rrm3 cells did not have the same sensitivities to DNA-damaging agents as sgs1, srs2, or mrx strains (Fig. 7). These data suggest that these synthetic phenotypes were not due to overlapping functions of these genes with RRM3 in DNA replication or repair. Previously, we proposed that Rrm3p might be needed for fork restart at sites of protein DNA complexes (36). However, the fact that rrm3 cells had at most modest sensitivity to DNA-damaging agents (Fig. 7) argues against a role for Rrm3p in the restart of stalled forks. More likely, the Rrm3p helicase promotes fork progression past protein-DNA complexes.

The synthetic lethality of rrm3 with sgs1 and srs2 was suppressed by deletion of RAD51, and these triple mutants had no evident growth defect compared to single mutant strains (Fig. 8A to C), a result also reported in reference 67. Thus, while Sgs1p and Srs2p affect the intra-S-phase checkpoint, their lethal interactions with rrm3 are not due to an impaired checkpoint response. Since paused and broken replication forks were similarly abundant in rrm3 sgs1 rad51 and rrm3 cells (Fig. 9A), paused and broken forks are also unlikely to be the cause of the rrm3 sgs1 lethality. However, the paused and broken forks likely provide the signal that activates DNA checkpoints, as Rad53p hyperphosphorylation was not suppressed by deletion of RAD51 (Fig. 9B). Taken together, these data suggest that the stalled or broken forks generated in the absence of Rrm3p are acted upon by Rad51p to generate an intermediate that is toxic in the absence of Sgs1p/Top3p or Srs2p. In agreement with others, we propose that Sgs1p and Top3p are needed to process these otherwise toxic recombination intermediates, while Srs2p limits the number of such intermediates by disrupting Rad51p filaments (42, 86).

Other published synthetic lethal phenotypes, such as those represented by the sgs1 srs2 and sgs1 mus81 genotypes, are alleviated by deletion of either RAD51 or RAD52 (23). However, deletion of RAD52 in rrm3 sgs1 and rrm3 srs2 did not completely suppress the lethal phenotype (Fig. 8D and E), a result also reported in reference 67. These results suggest that some HR is needed for viability in these genetic backgrounds. HR was also important in cells lacking only Rrm3p, as rrm3 rad52 and rrm3 rad59 cells grew more slowly than single mutants (Fig. 8D to F and Fig. 10B and C). The residual recombination had the appropriate genetic dependencies to be BIR, as it was Rad59p-dependent but Rad1p and Rad51p independent (Fig. 10A, B, D, and E). These data suggest that BIR plays a role in restart of the rrm3-generated stalled or broken replication forks, especially in the absence of Rad51p. However, we cannot exclude that an as yet uncharacterized Rad52p-dependent pathway (for example, template switching) helps restart the rrm3 stalled forks.

The synthetic lethality of rrm3 with mrx was not suppressed by deletion of RAD51 or RAD52 (data not shown) (67). As the MRX complex is necessary for complete activation of the intra-S-phase checkpoint (18, 31), it is possible that this lethal interaction is due to an impaired checkpoint response. However, the MRX complex is also necessary for the formation of type II survivors in the absence of telomerase (14, 43, 78). Formation of type II survivors is Rad51p independent but Rad50p and Rad59p dependent and is proposed to occur by a BIR-like mechanism (14, 43, 70, 78). Because Rrm3p is needed for normal fork progression through telomeric and subtelomeric DNAs (36), it is possible that the rrm3 mrx lethality is due to a defect in telomere maintenance. Finally, in higher organisms, the absence of MRE11 leads to the accumulation of DSBs during replication (16). Therefore, the rrm3 mrx synthetic lethality could be due to an as yet uncharacterized role of the MRX complex during replication.

Surprisingly, rrm3 displayed no synthetic phenotype with mus81 (Fig. 11A; the same result is reported in reference 67). The Mus81p/Mms4p structure-specific endonuclease is thought to define a pathway for fork restart that is parallel to the Sgs1p/Top3p pathway (6, 38) (Fig. 12). During a normal S phase, either pathway is sufficient for viability but loss of both is lethal (23, 53). Likewise, rrm3 did not show synthetic defects with slx1 (Fig. 11B). Slx1p/Slx4p genes also encode a structure-specific endonuclease that is thought to act redundantly with Sgs1p/Top3p to process stalled or converging replication forks (26). These data suggest that the stalled and broken forks generated in rrm3 cells are not substrates for the Mus81p/Mms4p or Slx1p/Slx4p nucleases. We propose that the structure of forks stalled and broken as a result of encounters with natural protein-DNA complexes is different from that of forks arrested by exogenous DNA damage. For example, forks stalled by protein complexes may have less single-stranded character, since a protein complex that impedes fork progression will probably also prevent unwinding of the parental duplex. If the parental duplex is not unwound, replication of the leading and lagging strands cannot be uncoupled, an event that generates ssDNA at the fork (73). Consistent with the possibility that forks stalled at protein barriers lack extensive single-stranded regions, forks arrested at the Fob1p-dependent RFB in yeast rDNA have only a few bases of ssDNA (32).

FIG. 12.

Model for how cells detect and repair rrm3-dependent DNA damage. Rrm3p functions during DNA replication to allow forks to move past nonnucleosomal protein-DNA complexes. Absence of Rrm3p leads to fork stalling and fork breakage. Stalled forks activate the intra-S-phase checkpoint. Stalled forks are also prone to breakage, which can activate the DNA damage checkpoint. Collapsed and broken forks are restarted or repaired by a Rad51p-dependent and Srs2p-inhibited recombination pathway. The Sgs1p/Top3p complex is needed to resolve the Rad51p-generated intermediates. However, the Mus81p/Mms4p endonuclease does not act on rrm3-induced recombination intermediates. BIR is also capable of restarting broken replication forks.

The genetic interactions reported here lead us to the following working model (Fig. 12). By virtue of its helicase activity, Rrm3p promotes fork progression past specific nonnucleosomal protein-DNA complexes. A role for Rrm3p in fork progression is consistent with the demonstration that Rrm3p interacts with PCNA by two-hybrid and in vitro criteria (66). In the absence of Rrm3p, forks often stall and sometimes break at these sites. These stalled and broken forks activate the Mec1p-, Mrc1p-, and Rad53p-dependent intra-S-phase checkpoint and in some cases the DNA damage checkpoint. The actions of the intra-S-phase checkpoint, which prevent the collapse of the rrm3 stalled forks and promote their restart and repair, ensure the viability of rrm3 cells. Stalled and broken forks are acted upon by Rad51p to generate an intermediate that is processed by Sgs1p and Top3p. The intermediates generated in rrm3 cells are poor substrates for the Mus81p/Mms4p fork restart pathway, which is thus unimportant for rrm3 viability. Because the Rad51p-generated intermediate is toxic, strains that fail to process it die. In the absence of Rrm3p, there are more Rad51p substrates, and this abundance is increased to a lethal level in the absence of Srs2p, due to loss of its inhibition of Rad51p-mediated events (42, 86). Thus, rrm3 is lethal with srs2 because the number of toxic recombination intermediates overwhelms the pathway for their resolution. Its lethality with sgs1 and top3 is also attributed to accumulation of toxic recombination intermediates, as these genes are required for processing the intermediate. The stalled and broken forks generated in rrm3 cells can also be restarted by BIR, and this pathway plays a role in the absence of Rad51p.