To examine the role of the RAD52 recombinational repair pathway in compensating for DNA replication defects in Saccharomyces cerevisiae, we performed a genetic screen to identify mutants that require Rad52p for viability. We isolated 10 mec1 mutations that display synthetic lethality with rad52. These mutations (designated mec1-srf for synthetic lethality with rad-fifty-two) simultaneously cause two types of phenotypes: defects in the checkpoint function of Mec1p and defects in the essential function of Mec1p. Velocity sedimentation in alkaline sucrose gradients revealed that mec1-srf mutants accumulate small single-stranded DNA synthesis intermediates, suggesting that Mec1p is required for the normal progression of DNA synthesis. sml1 suppressor mutations suppress both the accumulation of DNA synthesis intermediates and the requirement for Rad52p in mec1-srf mutants, but they do not suppress the checkpoint defect in mec1-srf mutants. Thus, it appears to be the DNA replication defects in mec1-srf mutants that cause the requirement for Rad52p. By using hydroxyurea to introduce similar DNA replication defects, we found that single-stranded DNA breaks frequently lead to double-stranded DNA breaks that are not rapidly repaired in rad52 mutants. Taken together, these data suggest that the RAD52 recombinational repair pathway is required to prevent or repair double-stranded DNA breaks caused by defective DNA replication in mec1-srf mutants.
THE activities of many different proteins must be coordinated to perform the different steps of DNA synthesis, and defects in these proteins can result in the requirement for compensatory mechanisms to complete DNA synthesis. In addition, damage to the DNA template itself can prevent the cell from executing DNA synthesis in a normal fashion, because DNA lesions can block the progress of the DNA synthesis machinery (Strauss 1985; Berger and Edenberg 1986). When the cell does not properly compensate for defective DNA synthesis or a damaged DNA template, progression through the cell cycle results in broken chromosomes and an accumulation of mutations. In humans, these genomic aberrations can cause disastrous effects, such as cell death or uncontrolled cell proliferation leading to cancer (for reviews see Paulovichet al. 1997; Weinert 1998). Therefore, cells contain several pathways that ensure that a faithful copy of the genome will be completed prior to chromosome segregation during mitosis.
While the RAD52 recombinational repair pathway was originally identified as being important for the repair of DNA damage caused by ionizing radiation (Game and Mortimer 1974), it also appears to be important for compensating for defects in DNA synthesis. Many mutations affecting DNA synthesis proteins, such as FEN-1, thymidylate kinase, Cdc6p, DNA ligase, Polα, and Polδ (Hartwell and Smith 1985; Tishkoffet al. 1997), cause hyper-recombination phenotypes, suggesting they cause defects that are repaired by recombinational mechanisms. Furthermore, many mutations affecting DNA synthesis proteins, such as PCNA, RFC1, FEN-1, DNA ligase, RPA, and Polδ, cause a requirement for the RAD52 recombinational repair pathway for viability (Monteloneet al. 1981; Giotet al. 1997; Tishkoffet al. 1997; Chenet al. 1998; Merrill and Holm 1998). Each of these proteins is important for the in vitro reconstitution of lagging strand DNA synthesis (Ishimiet al. 1988; Tsurimoto and Stillman 1991; Turchiet al. 1994; Waga and Stillman 1994), and mutations affecting the structural genes for PCNA, RFC1, FEN-1, and DNA ligase cause the accumulation of Okazaki-fragment-sized single-stranded DNA (ssDNA) fragments during DNA synthesis in vivo (Johnston and Nasmyth 1978; Merrill and Holm 1998). Inhibition of DNA synthesis by the ribonucleotide reductase inhibitor hydroxyurea (HU) also causes a requirement for the RAD52 recombinational repair pathway (Allenet al. 1994). Interestingly, despite the ample evidence suggesting that Rad52p is required to overcome the consequences of DNA synthesis defects, the mechanisms of this compensatory role for recombinational repair remain unclear.
Proper replication of the genome is also ensured by regulatory mechanisms that must coordinate the activity of DNA synthesis and DNA repair pathways. Under conditions of genotoxic stress, the cell division cycle arrests at the G1 or G2/M checkpoint, DNA synthesis is slowed, and the transcription of damage response genes is stimulated. A number of genes, including RAD9, RAD17, RAD24, MEC3, RFC2, RFC5, and POL2, are required to maintain the G1 and G2/M checkpoints (Weinert and Hartwell 1988; Siedeet al. 1993; Weinertet al. 1994; Sugimoto et al. 1996, 1997; Noskovet al. 1998). Two other checkpoint proteins, Mec1p and Rad53p, are also particularly important for the integrity of the S-phase checkpoint, because they are required for the slowing of DNA synthesis during exposure to genotoxins (Paulovich and Hartwell 1995). While the precise mechanism responsible for the slowing of DNA synthesis has yet to be elucidated, it is clear that Mec1p and Rad53p are required to prevent firing of “late” origins of replication during conditions of genotoxic stress (Santocanale and Diffley 1998; Shirahigeet al. 1998). Mec1p and Rad53p also respond to DNA damage by inducing a transcriptional response that is mediated by the Dun1p protein kinase (Zhou and Elledge 1993; Allenet al. 1994).
To investigate the role of recombinational repair in compensating for defects in DNA synthesis, we performed a genetic screen to isolate mutations that cause a requirement for an intact RAD52 recombinational repair pathway. Of the 15 mutants recovered, 10 were mec1 mutants. Although these mec1-srf mutations confer a checkpoint defect, we found that this checkpoint defect does not cause the requirement for Rad52p; instead, the RAD52 recombinational repair pathway must compensate for defects in DNA synthesis caused by mec1-srf mutations. In addition, we show that inhibiting DNA synthesis with HU in rad52 mutants results in the appearance of double-stranded DNA (dsDNA) breaks, suggesting that the RAD52 recombinational repair pathway may be required to repair dsDNA breaks created during defective DNA synthesis.
MATERIALS AND METHODS
Strains, plasmids, and growth conditions: The yeast strains used in this study have an S288C background and are listed in Table 1. Standard genetic techniques were used for the construction and growth of each strain (Shermanet al. 1986). Parent strain CH2550 (ade2 ade3 leu2 ura3 his3 rad52::LEU2), which was mutagenized for the synthetic lethal screen, was constructed by transforming strain CH1462 (ade2 ade3 leu2 ura3 his3) with a rad52::LEU2 disruption fragment generated by digestion of plasmid pCH1619 (kindly provided by D. Schild) with BamHI. To disrupt the SML1 gene, 45 bp of homology adjacent to the SML1 gene was added to the kanMX4 cassette (Wachet al. 1994) by PCR using two primers: 5′-CTTACGGTCTCACTAACCTCTCTTCAATGCTCAATAAT TTCCCGCAGGTGAAGCTTCGTACGC-3′ and 5′-GTATGA AAGGAACTTTAGAAGTCCATTTCCTCGACCTTACCCTGG GCATAGGCCACTAGTGGATCTG-3′. G418-resistant transformants were selected as described (Wachet al. 1994). All gene disruptions were confirmed by PCR.
Cells were grown in YEPD (rich) or SD (minimal) medium. YEPD is 1% yeast extract, 2% bacto-peptone, and 2% dextrose; SD is 0.67% yeast nitrogen base and 2% dextrose. For synthetic complete (SC) medium, 20 mg of uracil, adenine, tryptophan, and histidine, 30 mg of lysine, and 60 mg of leucine were added to 1 liter of SD medium. Plates of 5-fluoroorotic acid (5-FOA) were made as described previously (Ausubelet al. 1988). All plates were made with 2% bacto-agar. Unless otherwise noted, cells were grown at the permissive temperature of 30°.
Conventional techniques of molecular biology were used in this study (Ausubelet al. 1988). The screening plasmid for the srf screen was constructed by a two-step process. To make plasmid pCH1676 (RAD52 URA3 CEN/ARS), a SalI-SalI RAD52 fragment from plasmid pCH1624 (RAD52 LEU2 2μ; kindly provided by D. Schild) was subcloned into the XhoI site of plasmid pRS316 (URA3 CEN/ARS). The ADE3 gene was then subcloned into the SalI-NotI sites of plasmid pCH1676 with a 3.5-kb SalI-NotI fragment from plasmid pCH1675 (ADE3 URA3 CEN/ARS) to make plasmid pCH1677 (RAD52 ADE3 URA3 CEN/ARS). To make plasmid pCH1674 (RAD52 HIS3 CEN/ARS), a SalI-EcoRI fragment containing RAD52 from plasmid pCH1624 (RAD52 LEU2 2μ) was ligated into the SalI-EcoRI sites of plasmid pRS313 (HIS3 CEN/ARS).
Screen for srf mutants: The screen to identify mutations that display synthetic lethality with rad fifty-two (srf) was based on the plasmid dependence assay developed by Kranz and Holm (1990). Strain CH2550 (ade2 ade3 leu2 ura3 his3 rad52::LEU2) was transformed with screening plasmid pCH1677 (RAD52 URA3 ADE3 CEN/ARS), and stationary cultures were grown from individual transformants in synthetic medium to select for the pCH1677 plasmid. Sonicated cell suspensions were mutagenized by EMS to 80% viability and spread onto YEPD plates. Approximately 150,000 EMS-mutagenized colonies were assayed for sectoring after 7 days growth at 30°. Putative srf mutants (solid red colonies) were restreaked onto fresh YEPD plates (to retest the sectoring phenotype) and 5-FOA plates [to provide a secondary test for dependence on plasmid pCH1677 (RAD52 URA3 ADE3 CEN/ARS)]. Only one srf mutant from each individual culture was kept for further characterization, to prevent the accumulation of mutations that could be identical by descent.
The following tests were performed to determine if plasmid dependence reflected dependence on the RAD52 gene on plasmid pCH1677, and to determine if plasmid dependence was caused by mutations affecting a single gene. Nonsectoring strains were transformed with testing plasmids pCH1674 (RAD52 HIS3 CEN/ARS) and pCH1093 (HIS3 CEN/ARS), and transformants were tested for sectoring on YEPD and for growth on 5-FOA plates. We kept only the putative srf mutants that (1) regained sectoring and became 5-FOA resistant when transformed with pCH1674 (RAD52 HIS3 CEN/ARS), and (2) failed to regain sectoring and remained 5-FOA sensitive when transformed with plasmid pCH1093 (HIS3 CEN/ARS). To determine if the Srf- phenotype was caused by mutations affecting a single nuclear gene, each putative srf mutant (MATα ade2 ade3 leu2 ura3 his3 rad52::LEU2 srf(1, 2, 3, or 4) [pCH1677 (RAD52 URA3 ADE3 CEN/ARS)]) was backcrossed with unmutagenized parent strain CH2551 (MATa ade2 ade3 leu2 ura3 lys2 rad52 SRF). Diploids were sporulated to obtain tetrads that were subsequently dissected and analyzed genetically. srf mutants in which the Srf- phenotype segregated in a 2:2 fashion were considered to have mutations affecting a single gene, and they were kept for further characterization. Analysis of backcrosses involving srf4 mutants indicated that the srf4 mutation was linked to the LYS2 locus (31PD:2NPD:8T). Spores in which the srf mutation had cosegregated with the rad52 mutation (and not the rad52::LEU2 mutation) were subsequently used for cloning by plasmid suppression.
Cloning of srf mutations by plasmid suppression: To clone the SRF genes, mutants with the general genotype ade2 ade3 leu2 ura3 his3 rad52 srf(1, 2, 3, or 4) [pCH1677(URA3 ADE3 RAD52 CEN/ARS)] were transformed with pCH1132 (LEU2 CEN/ARS genomic library; kindly provided by P. Hieter). To allow for the spontaneous loss of plasmid pCH1677 (URA3 ADE3 RAD52 CEN/ARS), transformants were grown for 3 days at 30° on synthetic dextrose plates that contained uracil, adenine, and histidine. To select for transformants harboring library plasmids that complemented the srf mutation, library transformants were transferred to 5-FOA plates by replica plating. Plasmid DNA was isolated from 5-FOA resistant colonies as described (Robzyk and Kassir 1992) and retransformed into the original srf mutant. Partial sequence was obtained for library plasmid inserts that complemented the Srf- phenotype upon retransformation, and it was submitted to the Saccharomyces Genome Database (http://genome-www.stanford.edu) for BLAST analysis. Conventional subcloning techniques were used to identify open reading frames (ORF) on the library plasmids that were necessary and sufficient for complementation.
Isolating library plasmids that complemented the Srf- phenotype successfully identified all four srf mutations. The Srf- phenotype of strains CH2568 (srf1-1) and CH2569 (srf1-2) was complemented by a library plasmid (pCH1750) that contained the RAD27/RTH1 gene. The identity of the srf1 mutations was confirmed by the failure of the two srf1 mutants, strains CH2572 (srf1-1) and CH2573 (srf1-2), to complement the temperature sensitivity of strain CH2363 (rth1Δ). The Srf- phenotype of strains CH2570 (srf2-1) and CH2571 (srf2-2) was complemented by a library plasmid (pCH1751) containing an insert with the NUP84 gene. Conventional subcloning was used to determine that the NUP84 sequence of the insert was necessary and sufficient for complementation of the Srf- phenotype. The srf3 mutation affecting strain CH2574 (srf3) caused sensitivity to HU in addition to the Srf- phenotype. A library plasmid (pCH1752) that complemented both the Srf- phenotype and the HU sensitivity of CH2574 (srf3) contained an insert with three predicted open reading frames. Subcloning of this insert demonstrated that the novel gene YDR499w was necessary and sufficient to suppress the defects caused by the srf3 mutation. The identification of srf4 mutations was determined using similar methods (see results).
Primary characterization of mec1-srf mutants: To determine if the mec1 mutations confer growth defects and/or sensitivity to DNA-damaging agents, serial dilutions of cultures of mec1-srf mutants were spotted onto YEPD plates and grown under various conditions: (1) at different temperatures (37°, 35°, 30°, 25°, 23°), (2) UV (20, 40, 60, 80, 100, 150, or 200 J/m 2), (3) methyl methanesulfonate (MMS; 0.005%, 0.01%, or 0.02%), or (4) HU (0.005 m, 0.01 m, 0.02 m, or 0.04 m). Growth on each plate was scored after 2 days.
Assessment of checkpoint defects: To determine if mec1-srf mutations cause defects in the checkpoint function of Mec1p, log-phase cultures of cells grown at 30° were treated with 200 mm HU. Aliquots were removed every hour, and cells were plated for viability or fixed for 4′,6-diamidino-2-phenylindole (DAPI) staining. Cell suspensions were diluted and sonicated so that individual cells could be spread onto YEPD plates. After 20 hr at 30°, the growth of individual cells was determined by scoring the number of cells in each of 100 microcolonies. Microcolonies consisting of at least 20 cells were scored as viable. To determine if cells had inappropriately passed through the G2/M checkpoint during treatment with HU, nuclear morphologies were scored; “normal” morphology consisted of cells with a single round nucleus, and “elongated” morphology consisted of large budded cells with either two separate nuclei or a single elongated nucleus.
Sucrose velocity sedimentation gradients: To determine if mec1-srf mutations cause DNA synthesis defects, DNA fragments from pulse-labeled cultures were resolved by size through sucrose velocity sedimentation gradients. To focus our analysis solely on chromosomal DNA replication, all gradient experiments were performed using [rho0] strains generated as described (Shermanet al. 1986). Briefly, 1-ml cultures of [rho0] strains were grown overnight in the presence of 0.08 μCi [14C]uracil for at least five generation times to chronically label genomic DNA with 14C. The log-phase cultures were concentrated to 0.5 ml in 1.5-ml tubes and preincubated for 10 min at 30°. For experiments that involved treatment with HU, 25 μl of a 2 m HU stock solution was added to the 1-ml culture immediately prior to preincubation. Next, newly synthesized DNA was labeled with a pulse of 100 μCi [3H]uracil for 35 min at 30°. Labeling was terminated as described (Johnston and Williamson 1978), and fixed cells were gently lysed as described (McAlearet al. 1996; Merrill and Holm 1998).
For alkaline gradients, 25 μl of 5 m NaOH was added to denature dsDNA just prior to floating lysed cells on top of a sucrose gradient (5-20% sucrose containing 0.7 m NaCl, 0.03 m EDTA, and 0.3 m NaOH). For neutral gradients, lysed cells were floated on top of a 15-30% sucrose gradient containing 0.7 m NaCl and 0.03 m EDTA. DNA molecules were resolved by velocity sedimentation through the sucrose gradients by centrifugation as described (McAlearet al. 1996; Merrill and Holm 1998). After centrifugation, each gradient was separated into 20, 250-μl fractions. For neutral gradients, fractions were treated with an equal volume of 1.0 m NaOH overnight at room temperature. Acid precipitable counts were determined for each fraction (McAlearet al. 1996; Merrill and Holm 1998). Gradient fraction 1 represents the top of the gradient (where smaller DNA molecules sediment), and fraction 20 is the bottom of the gradient (where larger DNA molecules sediment).
mec1 mutations display synthetic lethality with rad52: To identify mutations that cause a requirement for an intact RAD52 recombinational repair pathway, we performed a genetic screen to recover mutations that displayed synthetic lethality with a rad52 mutation (srf mutants). By screening ∼150,000 EMS-mutagenized colonies, we recovered 15 srf mutants that fell into four complementation groups (srf1, 2, 3, and 4), two of which (srf3 and srf4) proved to be sensitive to low levels of HU. For each complementation group, the Srf- phenotype was caused by a recessive mutation that segregated 2:2 in tetrads from diploids constructed by backcrossing srf mutant strains with the unmutagenized parent strain CH2551 (MATa ade2 ade3 leu2 ura3 lys2 rad52-pd1 SRF [pCH1677 (URA3 ADE3 RAD52)]). This result indicates that the Srf- phenotype in each strain is caused by a srf mutation affecting a single gene.
We chose to concentrate on the largest complementation group of srf mutations (srf4), which consisted of 10 of the 15 total srf mutations, although the identity of each of the four srf complementation groups was determined (see materials and methods). To identify the srf4 mutation, we cloned library plasmids that suppress both the Srf- phenotype (dependency on RAD52) and the HU sensitivity of srf4 mutants. Partial sequencing of one such plasmid, pCH1753, revealed that it contains the MEC1 gene. To determine if SRF4 and MEC1 are the same gene, we performed several experiments. We demonstrated that MEC1 specifically suppresses both phenotypes using an unrelated plasmid containing the MEC1 gene pEF212 (MEC1 HIS3) kindly provided by L. Hartwell; when transformed into strain CH2552 (srf4), this MEC1 plasmid complements both the Srf- phenotype and the HU sensitivity (data not shown). Mapping data further support the identity of SRF4 and MEC1; mutations in each gene lie 24 cM from LYS2 (see materials and methods). Finally, srf4 mutations fail to complement the HU sensitivity of mec1-1 when strain CH2552 (srf4) is crossed with strain CH2097 (mec1-1). Thus, we conclude that the Srf- phenotype of srf4 mutants is caused by mutations affecting the MEC1 gene. The 10 srf4 mutants were renamed mec1-100srf to mec1-109srf, and collectively they are referred to as mec1-srf mutants.
Mec1-srf mutants have a checkpoint phenotype: To begin to characterize their effects on DNA metabolism, we examined the general growth defects and DNA-damage sensitivities caused by each of the 10 mec1-srf mutations. While most of the mutants (8 of 10) have a similar degree of sensitivity to UV and MMS-induced DNA damage (Table 2), strain CH2559 (mec1-107srf) is more sensitive, and strain CH2560 (mec1-108srf) is less sensitive than the rest of the mec1-srf mutants. In addition, each of the mutants is sensitive to inhibition of DNA synthesis by HU. None of the mec1-srf mutations cause a temperature-sensitive growth defect.
To determine if mec1-srf mutations cause a defect in the checkpoint function of Mec1p, we tested the ability of the mec1-100srf mutant to arrest in the cell division cycle when DNA synthesis was inhibited by HU. When strain CH1462 (MEC1) is treated with 200 mm HU, it undergoes a cell-cycle arrest (only 19% of nuclei were elongated; Table 3), and it maintains good viability throughout the duration of the experiment (Figure 1). In contrast, the control checkpoint-defective strain CH2097 (mec1-1) fails to arrest (50% of nuclei were elongated), and it loses viability in HU (Figure 1). Similar to the mec1-1 control, strain CH2552 (mec1-100srf) fails to arrest cell division (51% of nuclei were elongated), and it rapidly loses viability in HU (Figure 1). Similar results were obtained with strain CH2553 (mec1-101srf). Taken together, these data indicate that mec1-srf mutations cause checkpoint defects and sensitivity to DNA-damaging agents, in addition to causing dependence on the RAD52 recombinational repair pathway.
Defects in the checkpoint function of Mec1p do not cause the requirement for RAD52: The biochemical functions of Mec1p during cell division are complex. Not only is Mec1p required to elicit a response to DNA damage and arrest cell division (checkpoint function), but it is also required for viability even in the absence of stress (essential function). While it is unclear whether these two functions reflect quantitative or qualitative differences in biochemical activities of Mec1p in vivo, these two functions can be genetically separated by suppressor mutations. sml1 mutations suppress inviability caused by mec1 mutations, including a mec1Δ mutation, but they do not suppress the defective damage-induced cell division arrest of mec1 mutants (data not shown; Paulovich and Hartwell 1995; Zhaoet al. 1998). Therefore, if a failure to arrest cell division in mec1 mutants (checkpoint defect) causes the requirement for Rad52p, then the sml1 mutation should not affect the viability of the mec1-srf rad52 double mutants, because sml1 mutations do not suppress this defect. In contrast, if a defect in the essential function of Mec1p causes the requirement for RAD52, then the sml1 mutation should suppress the synthetic lethality displayed between rad52 and mec1 mutations, because sml1 mutations suppress defects in the essential function of Mec1p.
To determine if the inability to arrest cell division in mec1-srf mutants causes a requirement for Rad52p, we tested whether sml1 mutations could suppress the requirement for Rad52p in mec1-srf mutants. Briefly, the SML1 gene was deleted from strain CH2552 (mec1-100srf rad52 [pCH1677 (RAD52 URA3 ADE3)]) and strain CH2553 (mec1-101srf rad52 [pCH1677 (RAD52 URA3 ADE3)]). The resulting mec1-srf rad52 sml1Δ [pCH1677 (RAD52 URA3 ADE3)] triple mutants were streaked onto 5-FOA plates to select for loss of the plasmid copy of RAD52 (Figure 2). In contrast to parental strains CH2552 (mec1-100srf rad52) and CH2553 (mec1-101srf rad52), strains CH2577 (mec1-100srf rad52 sml1Δ) and CH2578 (mec1-101srf rad52 sml1Δ) are viable even without the plasmid copy of RAD52. In addition to growing on 5-FOA plates, the triple mutant strains also regain the ability to form sectored colonies on YEPD plates (data not shown), indicating that in the presence of the sml1 suppressor, mec1-srf strains no longer require Rad52p for survival. Combined with the retention of the checkpoint defect (failure to arrest cell division in HU) in mec1-srf mutants bearing a sml1 mutation (data not shown), this result indicates that it is not the loss of the checkpoint function of Mec1p that produces the need for Rad52p; instead, it is a defect in the essential function of Mec1p that causes the requirement for Rad52p. Since sml1 mutations cause an increase in the concentration of dNTPs, the essential function of Mec1p likely involves DNA synthesis.
A DNA synthesis defect in mec1-srf mutants causes the requirement for RAD52: Given the recent findings implicating Mec1p and Rad53p in DNA replication (Santocanale and Diffley 1998; Shirahigeet al. 1998; Zhaoet al. 1998), we next tested whether the essential function of DNA synthesis is perturbed in mec1-srf mutants. Previously, we found that mutations (pol30, rth1, and rfc1) that display synthetic lethality with rad52 cause an accumulation of small ssDNA fragments during DNA synthesis in vivo (Merrill and Holm 1998). To examine the state of newly replicated DNA in mec1-srf mutants, we used alkaline sucrose gradients to examine the integrity of newly replicated DNA. All gradient experiments were performed with [rho0] cells to eliminate consideration of mitochondrial DNA from the analyses. Briefly, exponentially growing cultures were labeled for approximately five generation times with [14C]uracil to label the bulk of genomic DNA. Next, a short pulse (35 min) of [3H]uracil was added to the cultures to differentially label newly synthesized DNA. Sedimentation velocity through an alkaline sucrose gradient was used to resolve ssDNA molecules by size. With this labeling regimen, strain CH2579 (MEC1) incorporates most of the pulse label into ssDNA fragments that are similar in size to chronically labeled fragments (Figure 3); very little of the pulse-labeled DNA is smaller than 4 kb, which sediments to fraction 4. In contrast, strain CH2581 (mec1-100srf) accumulates small ssDNA synthesis intermediates at the top of the sucrose gradients. This accumulation of small ssDNA fragments during DNA synthesis is similar to, but less severe than, that seen in pol30, rfc1, rth1, and cdc9 mutants (Johnston and Nasmyth 1978; Merrill and Holm 1998). A similar accumulation of small ssDNA synthesis intermediates also occurs in the other mec1-srf mutant that we examined, strain CH2583 (mec1-101srf; data not shown). These results are consistent with the hypothesis that mec1-srf mutations cause an accumulation of small ssDNA synthesis intermediates that are similar in size to unligated (or partially ligated) Okazaki fragments.
If the DNA synthesis defect of mec1-srf mutants causes the requirement for Rad52p, then a mutation that suppresses the requirement for Rad52p should also suppress the DNA synthesis defect. This hypothesis was tested by examining newly replicated DNA in strain CH2581 (mec1-100srf; requires Rad52p) and strain CH2582 (mec1-100srf sml1Δ; does not require Rad52p; Figure 3). Whereas strain CH2581 (mec1-srf) shows a clear accumulation of small ssDNA products, the sedimentation profiles from strain CH2582 (mec1-100srf sml1Δ) are indistinguishable from those of strain CH2579 (MEC1) or strain CH2580 (MEC1 sml1Δ; Figure 3). Thus, the loss of Sml1p concomitantly suppresses the accumulation of small ssDNA synthesis intermediates and the requirement for Rad52p in mec1-srf mutants.
Inhibition of DNA synthesis causes double-strand DNA breaks in rad52 mutants: One possible explanation for the requirement for Rad52p in DNA synthesis mutants is that some ssDNA breaks occurring during DNA synthesis could be converted to dsDNA breaks, which would be lethal to rad52 mutants. A similar hypothesis recently has been proposed in E. coli to explain the requirement for RecBCD in rep and dnaB mutants (Michelet al. 1997; Seigneuret al. 1998). It is possible that an analogous situation occurs in eukaryotes, and dsDNA breaks may arise from DNA synthesis defects. Unfortunately, it is not possible to examine the DNA from strains harboring mutations that cause both defective DNA synthesis and defective recombinational repair because these strains are inviable. However, yeast strains treated with 50 mm HU during DNA synthesis accumulate ssDNA damage (Johnston 1983; Walmsleyet al. 1984) that is similar to what is seen in mec1-srf (present study), pol30, rfc1, rth1 (Merrill and Holm 1998), and cdc9 (Johnston and Nasmyth 1978) mutants. Other studies of the effects of HU suggest that these DNA synthesis fragments are replication intermediates caused by stalled replication forks (Santocanale and Diffley 1998).
To determine if dsDNA breaks can accumulate in rad52 mutants because of defects in DNA synthesis, we looked for the formation of dsDNA breaks in newly synthesized DNA when DNA replication was inhibited by HU. We treated strains CH2579 (RAD52) and CH2584 (rad52) with 50 mm HU during a pulse-labeling period. Dual-labeled dsDNA fragments were separated by size by sedimentation through a neutral sucrose gradient (Figure 4). In the absence of HU treatment, the sedimentation profiles of dsDNAs from both strain CH2579 (RAD52) and strain CH2584 (rad52) show that most of the chronic and pulse label is found in large dsDNA molecules (>160 kb, which sediments to fraction 5). When DNA synthesis was inhibited by 50 mm HU, a striking accumulation of small dsDNA molecules (50 kb and smaller) containing newly synthesized DNA is observed in the rad52 mutant strain. These results show that rad52 mutants accumulate dsDNA breaks when DNA synthesis is perturbed, and they are consistent with the hypothesis that the formation of dsDNA breaks causes rad52 mutants to be sensitive to the effects of inhibiting DNA synthesis.
To examine the role of recombinational repair in ensuring complete replication of the genome, we performed a genetic screen to isolate mutations that display synthetic lethality with a rad52 mutation. Surprisingly, we recovered 10 new mec1 mutants that require the RAD52 recombinational repair pathway for viability. Sedimentation analysis of newly synthesized DNA from mec1-srf mutants indicates that the mec1-srf mutations cause defects in DNA synthesis that are similar to defects caused by other mutations (pol30, rfc1, rth1, and cdc9) displaying synthetic lethality with rad52. The single-stranded nicks and breaks observed in each of these mutants may be converted to dsDNA breaks, because dsDNA breaks accumulate during DNA synthesis in rad52 mutants treated with the DNA synthesis inhibitor HU. These data suggest that dsDNA break formation and recombinational repair play an important role in maintaining genomic stability when there are defects in the normal mode of DNA replication.
While the genetic screen for srf mutants was not saturated, it was striking that two-thirds of all mutants isolated had mutations affecting Mec1p. We recovered two mutations affecting rad27/rth1, which has previously been shown to display synthetic lethality with rad52 (Tishkoffet al. 1997). The screen was not saturated, however, because we did not recover mutations affecting several other genes (pol30, rfc1, rad3, rpa1, pol3, and cdc9) previously shown to display synthetic lethality with rad52 (Monteloneet al. 1981; Malone and Hoekstra 1984; Giotet al. 1997; Chenet al. 1998; Merrill and Holm 1998). It is somewhat surprising that so many of the srf mutants recovered (10 of 15) were mec1 mutants. However, MEC1 is a large gene for yeast (>7000 bp), and it may be that many different sites in the ORF can be mutated to cause the partial defects leading to the Srf- phenotype. Therefore, it is possible that MEC1 simply provides a larger target for EMS mutagenesis than other genes that can mutate to cause Rad52p dependency.
A connection between recombinational repair and Mec1p-like proteins has also been suggested in humans with the cloning of genes responsible for Nijmegen breakage syndrome and ataxia telangiectasia. These diseases are phenotypically similar to each other, sharing common cell biological defects (chromosomal rearrangements, sensitivity to ionizing radiation, and radio-resistant DNA synthesis) and clinical features (immunodeficiency and predisposition to hematopoietic malignancy; Nagasawaet al. 1985; Taalmanet al. 1989). In light of these similarities, the cloning of the ATM gene (a homologue of Mec1p) (Savitskyet al. 1995) and NBS1/nibrin (a 95-kD protein found in a complex with RAD50 and MRE11 recombinational repair proteins; Carneyet al. 1998; Varonet al. 1998) suggests that these two proteins may either have similar cellular functions, or perform separate steps in a single cellular pathway. The identification of mec1 mutations that cause a requirement for recombinational repair and the DNA synthesis defects caused by these mec1 mutations may have relevance to these diseases. It is possible that the common defects associated with A-T and NBS are caused by DNA synthesis defects and the inability to repair these defects, respectively. Further research into the functions of the Atm protein is required to determine if atm mutant cells have defects in DNA synthesis that are analogous to the defects in mec1-srf mutants.
While it is clear that Rad52p must compensate for defects in the essential function of Mec1p, the precise function of Mec1p in DNA synthesis remains unclear. Mec1p may be required to synthesize sufficient amounts of dNTPs to allow replication of the entire genome (Zhaoet al. 1998). Alternatively, Mec1p may be required to coordinate the firing of origins of replication during S phase (Santocanale and Diffley 1998; Shirahigeet al. 1998). Deficits in either of these functions could lead to defects such as stalled replication forks or defective Okazaki fragment maturation. Such defects would cause the accumulation of small ssDNA synthesis intermediates, such as those observed in Figure 3. Thus, either stalled replication forks or an abundance of ssDNA breaks in the nascent DNA strand could cause the requirement for the RAD52 recombinational repair pathway.
If mec1-srf mutations indeed cause replication forks to stall, the requirement for Rad52p could be analogous to the requirement for the RecBCD complex in rep and dnaB mutants of Escherichia coli. Mutations affecting replicative helicases (rep and dnaB) cause a requirement for the recombinational repair complex RecBCD for viability, and the requirement for RecBCD is caused by dsDNA break formation at stalled replication forks (Michelet al. 1997; Seigneuret al. 1998). Interestingly, the requirement for RecBCD and the formation of dsDNA breaks are concomitantly suppressed by mutations affecting the RuvAB enzyme, which processes Holliday junctions (Michelet al. 1997; Seigneuret al. 1998). One interpretation is that RuvAB processes stalled replication forks into Holliday junctions for recombination-dependent repriming of DNA synthesis, which prevents dsDNA break formation. While the yeast genome does not contain sequences similar to ruvA or ruvB, it may encode proteins that have activities analogous to RuvAB with respect to processing stalled replication forks. If so, the accumulation of dsDNA fragments in rad52 strains treated with HU could be explained if the RAD52 pathway is important for preventing dsDNA break formation resulting from stalled replication forks.
Results from other studies are consistent with the hypothesis that the recombinational repair pathway could be necessary to allow complete replication of the genome when there are stalled replication forks. For example, DNA synthesis of long stretches of DNA (virtually entire chromosome arms) can be stimulated by recombinational repair in yeast (Malkovaet al. 1996; Morrowet al. 1997; Bosco and Haber 1998). Other results indicate that recombinational mechanisms are required to accommodate ssDNA lesions during S phase by allowing replication of DNA that contains UV-induced damage (Kadyk and Hartwell 1993; Paulovichet al. 1998). In addition, after the first origin-initiated round of replication of the T4 bacteriophage genome following infection, recombinational mechanisms are responsible for late rounds of DNA synthesis (Mosiget al. 1995). Thus, it is possible that recombination-dependent DNA synthesis could take over a significant amount of DNA synthesis when normal origin-initiated DNA replication fails due to stalled replication forks.
Although stalled replication forks could cause the requirement for Rad52p, a simpler hypothesis explaining the requirement for Rad52p is that an abundance of ssDNA breaks in mec1-srf mutants leads to an elevated level of dsDNA breaks. Consistent with this hypothesis, mec1 mutations have previously been shown to increase the frequency of mitotic recombination (Vallen and Cross 1995). In addition, we have shown that the DNA synthesis defect of mec1-srf mutants causes an accumulation of small ssDNA fragments that can be observed with pulse-labeled DNA resolved in alkaline sucrose gradients. This type of sedimentation profile has also been observed in other mutants (pol30, rth1, rfc1, and cdc9) that require Rad52p for survival (Johnston and Nasmyth 1978; Merrill and Holm 1998). In addition, treatment with HU, which is toxic to rad52 mutants, also causes an accumulation of ssDNA synthesis intermediates (Johnston 1983; Walmsleyet al. 1984). The presence of many ssDNA breaks in one DNA strand could potentiate dsDNA break formation, because stochastic breaks affecting the template strand would be more likely to occur across from a ssDNA break in the nascent strand. Since Rad52p is required for homologous recombinational repair of dsDNA breaks, these dsDNA breaks would be lethal in rad52 mutants; even a single dsDNA break causes inviability in these mutants (Resnick and Martin 1976). Thus an economical explanation for why mec1-srf mutations cause inviability in rad52 mutants is that normally nonlethal forms of damage (ssDNA breaks affecting the template) are converted to dsDNA breaks, which are lethal in the absence of Rad52p.
We thank Scott Oh for his excellent assistance in cloning srf mutants. This work was supported by grant GM-36510 from the National Institutes of Health (NIH). B.J.M. was partially supported by NIH training grant CA-67754.
Communicating editor: M. Lichten
- Received February 10, 1999.
- Accepted June 7, 1999.
- Copyright © 1999 by the Genetics Society of America