An allele of RFA1, the largest subunit of the single-stranded DNA-binding complex RP-A, was identified as a suppressor of decreased direct-repeat recombination in rad1 rad52 double mutants. In this study, we used two LEU2 direct-repeat assays to investigate the mechanism by which the rfa1-D228Y allele increases recombination. We found that both intrachromatid and sister chromatid recombination are stimulated in rfa1-D228Y strains. In a rad1 rad52 background, however, the majority of the increased recombination is caused by stimulation of deletion events by an intrachromatid recombination mechanism that is likely to be single-strand annealing. Studies in which an HO endonuclease cut was introduced between the two leu2 copies indicate that the rfa1-D228Y mutation partially suppresses the rad52 defect in recovering recombination products. Furthermore, molecular analysis of processing and product formation kinetics reveals that, in a rad52 background, the rfa1-D228Y mutation results in increased levels of recombinant products and the disappearance of large single-stranded intermediates characteristic of rad52 strains. On the basis of these results, we propose that in the absence of wild-type Rad52, the interaction of RP-A with single-stranded DNA inhibits strand annealing, and that this inhibition is overcome by the rfa1-D228Y mutation.
RECOMBINATION between repeated sequences is thought to be the major mechanism governing the evolution of multigene families, as well as alterations of genome structure (Edelman and Gally 1970). However, recombination between dispersed homologous sequences can also lead to deleterious genomic rearrangements such as deletions, inversions, translocations, and amplifications (Haluskaet al. 1986; Fukuchiet al. 1989). These types of rearrangements have been suggested to be associated with several human diseases including hypercholesterolemia (Lehrmanet al. 1986), type A severe hemophilia (Lakichet al. 1993), and the expansion of triplet repeats associated with many neurological diseases such as fragile X (Fuet al. 1991), myotonic dystrophy (Fuet al. 1992; Harleyet al. 1992), spinal and bulbar muscular atrophy (La Spadaet al. 1991; Caskeyet al. 1992), and Huntington's disease (Huntington's Disease Collaborative Research Group 1993). In addition, increased genomic instability has been shown to be associated with several human maladies including ataxia telangiectasia, Fanconi anemia, and Bloom and Werner syndromes (Langloiset al. 1989; Monnat 1992; Meyn 1993; D'Andrea and Grompe 1997).
In Saccharomyces cerevisiae, insight into the mechanisms involved in generating rearrangements has been obtained by analysis of recombination between directly repeated sequences (for review see Klein 1995). Genetic analysis of direct-repeat recombination has defined two pathways for deletion events. These alternate pathways are defined by the RAD1 and RAD52 genes. The RAD1 gene, which was identified through a UV-sensitive mutation, is involved in the nucleotide excision repair pathway (Game and Cox 1971; Reynolds and Friedberg 1981). The Rad1 protein has been shown to form a complex with Rad10 that functions as a singlestranded DNA (ssDNA) endonuclease (Bardwellet al. 1992; Tomkinson et al. 1993). This endonuclease function is thought to be required for the removal of nonhomologous sequences in direct repeat recombination (Fishman-Lobell and Haber 1992). A mutation in the RAD52 gene results in X-ray sensitivity (Resnick 1969), and the Rad52 protein has been shown to be involved in the repair of many kinds of double-strand breakinduced recombinational repair events (Resnick and Martin 1976; Malone and Esposito 1980; Orr-Weaveret al. 1981). It has also been demonstrated that the Rad52 protein can promote strand annealing in vitro (Mortensenet al. 1996) as well as stimulate strand exchange by the Rad51 protein (Sung 1997a; Bensonet al. 1998; Newet al. 1998; Shinohara and Ogawa 1998). In direct-repeat recombination, neither rad1 nor rad52 mutants significantly affect deletion events. However, rad1 rad52 double mutants are dramatically decreased for this type of recombination (Klein 1988; Schiestl and Prakash 1988; Thomas and Rothstein 1989b). Although the double mutants exhibit a large decrease in recombination rates, a low level remains, suggesting that additional pathways of recombination exist.
To identify alternate pathways of recombination that function in deletion formation, mutations that suppress the decreased levels of recombination in rad1 rad52 strains were isolated (Smith and Rothstein 1995). Interestingly, the single mutation that was isolated is an allele of RFA1, the large subunit of the ssDNA-binding complex RP-A (Heyer et al. 1990; Brill and Stillman 1991). The mutation isolated is a missense allele that alters a conserved residue of the protein (D228Y). Biochemical and genetic analysis suggests that the phenotype observed in rfa1-D228Y mutants may be in part caused by the decreased levels of the RP-A complex within the cell. Characterization of its recombination phenotype indicates that the rfa1-D228Y mutation restores wild-type levels of deletion formation in rad1 rad52 strains. In addition, it displays a hyper-recombinogenic phenotype on its own that is independent of RAD52 and slightly dependent on RAD1.
In direct-repeat recombination, deletion events can occur by recombination between repeats located on the same chromosome (intrachromatid) or by an interaction between sister chromatids. Thus, the first approach taken to characterize the role of rfa1-D228Y in deletion formation was to determine whether intrachromatid, sister chromatid, or both types of recombination events were responsible for the elevated levels of recombination observed in an rfa1-D228Y background. Because the direct-repeat assay originally used to identify the rfa1-D228Y mutation does not allow these two events to be distinguished, recombination was examined using a pair of LEU2 direct-repeat assays that separately monitor intrachromatid and sister chromatid recombination. Second, to define the mechanistic role of rfa1-D228Y in deletion formation, recombination was analyzed molecularly using an HO-induced deletion assay that allows the kinetics of intermediate processing and deletion formation to be monitored.
MATERIALS AND METHODS
Media: YPD, synthetic medium supplemented with dextrose (2% w/v), and synthetic medium supplemented with glycerol and lactic acid (3% v/v each) were made as described previously (Shermanet al. 1986; Sherman 1991), with the exception that synthetic medium contains twice the amount of leucine (60 mg/liter). 5-Fluoro-orotic acid medium (5-FOA) is synthetic complete (SC) medium with 50 mg/liter of uracil and 0.75 g of 5-FOA per liter (Boeke et al. 1984). To induce the HO gene, galactose (2% w/v) was added as a 20% solution to SC-Trp glycerol lactate medium.
Yeast strains and plasmids: Standard procedures were used for mating, sporulation, and dissection (Shermanet al. 1986). All S. cerevisiae strains used in this study are derivatives of W303-1A and W303-1B (Thomas and Rothstein 1989a). W303 has recently been shown to contain an allele of RAD5 (rad5-G535R) in which the amino acid residue at position 535 is altered from a glycine (G) to an arginine (R) (Fanet al. 1996). All strains used in this study are from a RAD5 derivative of W303 (R823), kindly provided by Hannah Klein (Table 1).
Both configurations of the leu2 direct-repeat assay were constructed from pWJ567. This plasmid was created by the addition of BamHI linkers to the PvuII site of pWJ543, followed by the insertion of a 117-bp BstYI-BamHI fragment containing the HO endonuclease-cut site from pRK113 (kindly provided by R. Kostricken). pWJ543 contains the leu2ΔBstEII allele inserted between the NheI and SalI sites of YIp5, an integrating vector containing URA3. To create the leu2 duplications, pWJ567 was linearized on either side of the BstEII mutation by complete digestion with HpaI or partial digestion with AseI. The linearized fragments were gel purified and then transformed separately into J539, a strain containing a leu2ΔEcoRI allele. In this manner, strains containing both configurations of leu2 direct repeats were obtained.
A W303 derivative containing rad1::HIS5 was kindly provided by Naz Erdeniz. The RAD1 gene disruption was created by inserting a 2.1-kb SalI fragment containing HIS5 into the ClaI-StuI sites of RAD1 by adding BglII linkers to both fragments. This replaces 2.1 kb of the RAD1 reading frame with the HIS5 disruption in the resulting plasmid, pWJ612. The rad1 disrupted strain (U929) was then obtained by transforming a 2.9-kb SacI-EcoO1091 fragment from pWJ612 into W1088-10D.
A rad52 gene disruption was created by ligating the HIS5 gene contained on a 2.0-kb SalI (end-filled)-AgeI fragment into the SphI (end-filled)-AgeI sites of the RAD52 gene. This replaces 1.5 kb of the RAD52 reading frame with the HIS5 disruption in the resulting plasmid, pWJ600. The rad52-disrupted strain (U900) was then obtained by transforming a 3.8-kb SalI fragment from pWJ600 into W1088-1A.
The mat::HIS3 disruption was created by the PCR-based disruption method (Baudinet al. 1993; Smithet al. 1995). PCR primers were synthesized that contain 45 bp of target homology and 20 bp that is homologous to HIS3. The target homology of primer 1 is located within the W segment of the MAT locus, while primer 2 is homologous to sequences at the right boundary of the Z2 segment: primer 1, 5′-CCTCCAGG CGGAGTTAACAACTAGTAATACGGCATCCATGTTTGCG GATCCGCTGCACGGTCCTG-3′; primer 2, 5′-GATGCTAAG AATTGATTGTTTGCTTGAGTCTGAGTAATATCATATGCC TCGTTCAGAATGACACG-3′. PCR amplification of HIS3 using these primers results in a 1.3-kb fragment that is used directly for target disruption. The presence of the rad and mat disruptions, as well as the configuration of the leu2 duplications, was confirmed by analysis of genomic DNA blots (Southern 1975). Also, the presence of the rad1 and rad52 disruptions was verified by sensitivity of the mutant strains to UV or ionizing radiation, respectively.
The plasmid pJH132 (kindly provided by Jim Haber) contains the GAL10::HO fusion gene cloned into a TRP1 ARS1 CEN4 vector, thus permitting galactose-regulated expression of the HO endonuclease (Jensen and Herskowitz 1984). Riboprobes were made from plasmids pWJ656 and pWJ644 containing the EcoRV-SalI fragment of LEU2 cloned into the SmaI-SalI sites of pGEM-3Z and pGEM-4Z, respectively (Promega, Madison, WI).
DNA manipulations: Standard methods were used for recombinant DNA manipulations (Sambrooket al. 1989). Escherichia coli TG1 was transformed by a calcium chloride (Sambrooket al. 1989). Yeast cells were transformed by the lithium acetate method (Gietzet al. 1992). Genomic DNA from transformants and spontaneous recombinants was prepared from 5-ml YPD cultures according to the method of Hoffman and Winston (1987).
Neutral agarose gels were prepared and run in 0.5 × Trisborate (Sambrooket al. 1989). Alkaline denaturing gels were run as described by McDonnell et al. (1977). Gels were transferred overnight to a Biotrans nylon membrane according to the alkaline technique described by the manufacturer (ICN). Radiolabeled DNA probes were prepared using a random priming kit (Pharmacia, Piscataway, NJ; Feinberg and Vogelstein 1983). RNA probes were prepared by the method of Melton et al. (1984) as modified by Promega. Hybridization and washing of filters were carried out according to standard procedures (Sambrook et al. 1989). Blots were analyzed and quantitated densitometrically on a Molecular Dynamics PhosphorImager 445SI.
Analysis of spontaneous LEU2 direct-repeat recombination: Recombination rates and their standard deviations were calculated using the median method of Lea and Coulson (1949). Each rate was determined from trials of two to three different segregants, six to nine trials in total. Each trial represents a single colony from a YPD plate that was suspended in water, sonicated, and, after appropriate dilutions, plated on SC, synthetic medium lacking leucine (SC –Leu), and 5-FOA plates.
Independent recombinants were obtained by streaking colonies from a YPD plate onto SC –Leu medium and selecting a single colony from each streak. Recombinants that were unable to grow when replica plated to SC –Ura medium were not analyzed further and were categorized as deletion events. The remaining Ura+ recombinants were subjected to molecular analysis to determine their configuration.
HO-induced recombination assay: To monitor the efficiency of HO-induced deletions, cells of the appropriate genotype containing the proximal LEU2 direct-repeat assay were transformed with the HO plasmid pJH132. Transformations were grown to midlog phase in –Trp glycerol lactate medium. This medium selects for the retention of the HO plasmid and derepresses the GAL structural genes. After the removal of an aliquot of cells for the zero time point, galactose was added to a final concentration of 2%, and the incubation was continued for 1 hr. Appropriate dilutions of the cells obtained both before and after HO induction were made in sterile water and plated on YPD and SC –Trp. The resulting colonies were replica plated to SC –Trp, SC –Ura, and SC –Leu media to monitor retention of the HO containing plasmid, deletion formation, and the number of Leu+ recombinants, respectively. In addition, for each genotype, 20 Ura– deletions were analyzed by genomic DNA blots to verify that they contained the proper molecular configuration (Southern 1975).
The procedure to monitor intermediates of double-strand, break-induced deletion formation was based on the protocol of Rudin and Haber (1988). Strains containing the HO plasmid pJH132 were grown to late log phase in –Trp –Ura medium, which selects for the retention of the leu2 duplication and the HO plasmid. These cultures were diluted 1:100 in –Trp glycerol lactate medium and were grown at 30° to a final cell density of 0.7 to 1 × 107 cells per milliliter. After the removal of an aliquot of cells for the zero time point, galactose was added to a final concentration of 2%. The incubation was then continued with additional aliquots of cells removed at the given times. For the overnight time point samples, the final aliquot was centrifuged, resuspended in sterile water, and incubated overnight at 30°. DNA was prepared from each 30-ml sample (Hoffman and Winston 1987).
Spontaneous direct-repeat recombination
LEU2 direct-repeat assays: The original assay used to isolate rfa1-D228Y was limited in that only deletion events could be monitored. To extend the analysis, two chromosomal inserts similar to those previously described (Klein 1988; Bollag and Liskay 1991; Sugawara and Haber 1992) were constructed to help differentiate between the multiple mechanisms that have been proposed to account for direct-repeat recombination (Figure 1A). Each insert contains a 2.4-kb duplication of the LEU2 locus separated by 5.3 kb of plasmid sequences that includes the URA3 gene and an HO endonuclease-cut site (Kostrikenet al. 1983). One copy of LEU2, leu2ΔBstEII, contains a 5-bp insertion at the BstEII site at position 34, while the other leu2ΔEcoRI, results from a 4-kp insertion at the EcoRI site at position 637. Both mutations destroy the restriction site and create frameshift mutations that disrupt the open reading frame of LEU2. The two different configurations differ only in the relative positions of the LEU2 alleles. In the “proximal” configuration, leu2ΔEcoRI is in the left repeat and leu2ΔBstEII is in the right repeat; therefore, the mutant sites are closer (proximal). In the “distal” configuration, the alleles are reversed, resulting in a greater distance separating the sites (Figure 1A).
Both configurations can detect three types of spontaneous Leu+ recombinants: replacements (commonly referred to as gene conversions), deletions, and triplications (Figure 1B). All three can be distinguished by genomic DNA blots. Replacement events, which are likely to occur via gene conversion, are unaffected by the position of the mutant sites and, therefore, occur at the same rate in both configurations (Klein 1988). In contrast, the mechanisms that contribute to deletion and triplication events are influenced by the differential positioning of the leu2 alleles (Figure 1B, see Klein 1995 for a discussion of mechanisms). In the proximal configuration, deletions result from both intrachromatid and sister chromatid recombination, while triplications are rarely observed. In the distal configuration, intrachromatid recombination produces deletions and sister chromatid events generate triplications.
The increased levels of recombination in rfa1-D228Y mutants result from a stimulation of both intrachromatid and sister chromatid recombination: The spontaneous rate of leucine prototroph formation was determined in both proximal and distal configurations for each strain. Also, the distribution of deletion, replacement, and triplication events was determined by the molecular characterization of 20–60 independent Leu+ recombinants per genotype (see materials and methods).
Analysis of direct-repeat recombination in wild-type and rfa1-D228Y strains indicates that the rate of Leu+ formation in rfa1-D228Y strains is elevated in both the proximal and distal configurations (12- and 3-fold, respectively) compared to wild type (see Tables 2 and 3). This increase is consistent with the hyper-recombination phenotype of rfa1-D228Y strains described previously (Smith and Rothstein 1995). In the proximal configuration, the majority of the stimulated recombination is caused by an increase in the rate of deletions, which is elevated 18-fold compared to wild-type strains. In the distal configuration, however, rfa1-D228Y strains display an increase in both deletion and triplication events, 30- and 10-fold, respectively. Thus, while the substantial increase in deletion events in both orientations indicates that an intrachromosomal recombination mechanism is stimulated, the increase in triplications in the distal configuration suggests that sister chromatid events also contribute to the elevated recombination observed in rfa1-D228Y strains.
The rfa1-D228Y-stimulated levels of intrachromatid and sister chromatid recombination display a differential dependence on the RAD1 and RAD52 genes: Analysis of recombination in the absence of RAD1 or RAD52 indicates that the increased rate of Leu+ formation observed in rfa1-D228Y strains is partially dependent on both of these genes (Tables 2 and 3). In rad1 rfa1-D228Y double mutants, the rate of Leu+ prototroph formation in the proximal and distal configurations is decreased seven- and twofold, respectively, compared to rfa1-D228Y strains. The decrease is largely caused by a reduction in deletion events (12- and 23-fold for the proximal and distal configurations, respectively). In rad1 rfa1-D228Y strains, deletion events occur at a level similar to that observed in rad1 mutants, indicating that the intrachromatid recombination pathway(s) stimulated by rfa1-D228Y is largely RAD1 dependent.
In rad52 rfa1-D228Y strains, Leu+ prototroph formation is decreased in both configurations compared to rfa1-D228Y: 5-fold in the proximal and 36-fold in the distal configuration (Tables 2 and 3). The rate of deletion events in the proximal and distal configurations in rad52 rfa1-D228Y strains, although decreased compared to rfa1-D228Y (4- and 15-fold, respectively), is still significantly elevated compared to rad52 strains (48- and 12-fold, respectively). In contrast, triplication events in the distal configuration in rad52 rfa1-D228Y double mutants are reduced 27-fold to approximately the same levels observed in rad52 single mutants. This suggests that the sister chromatid recombination pathway(s) stimulated by rfa1-D228Y is RAD52 dependent.
The rfa1-D228Y mutation alters the synergistic decrease in deletions and triplications in rad1 rad52 strains only in the distal configuration: In rad1 rad52 rfa1-D228Y strains, the rate of Leu+ recombinants is 78- and 13-fold higher than that observed in rad1 rad52 strains in the proximal and distal configurations, respectively. This increase is partly caused by an increase in replacement and triplication events (see Tables 2 and 3). However, deletion events display the greatest stimulation, 100-fold in the proximal and ∼200-fold in the distal configuration.
Previous analysis of direct-repeat recombination demonstrated that deletions are synergistically decreased in rad1 rad52 strains (Klein 1988; Schiestl and Prakash 1988; Thomas and Rothstein 1989b). Here we find that triplications are also synergistically reduced in rad1 rad52. Interestingly, rad1 rad52 rfa1-D228Y strains exhibit a synergistic reduction in deletions in the proximal configuration compared with rad1 rfa1-D228Y and rad52 rfa1-D228Y double mutants, while in the distal configuration, neither deletion nor triplication events display a similar reduction. Thus, in the distal configuration, the synergistic decrease of deletion and triplication events in rad1 rad52 strains is suppressed by the rfa1-D228Y allele.
HO-induced direct-repeat recombination
Analysis of HO-induced direct-repeat recombination indicates that the rfa1-D228Y mutation increases the efficiency of repair in a rad52 background: Because deletions are the most stimulated events in an rfa1-D228Y background, we used an assay that specifically monitors deletion formation and permits a physical analysis of the intermediates and products. Using the proximal configuration of the LEU2 assay described previously (Figure 1A), recombination was initiated by a double-stranded break (DSB) at an HO endonuclease-cut site located between the LEU2 repeats. Subsequent repair of this DSB occurs by recombination between the direct repeats and specifically results in a deletion event via an SSA mechanism (Sugawara and Haber 1992). To regulate the induction of the DSB, a plasmid containing a galactose-inducible promoter fused to the HO endonuclease gene was introduced into each strain. Cells that fail to repair the DSB break die, while cells that repair the DSB delete the URA3 gene located between the LEU2 repeats. Therefore, to measure the efficiency of repair, we divided the number of Ura– cells after HO induction by the total number of cells containing the HO plasmid before induction. Last, the generation of Leu+ prototrophs, which likely reflects mismatch repair events in this assay, was also monitored.
In wild-type strains, after an HO-induced DSB, 72% of the cells have undergone a deletion event (Table 4). In an rfa1-D228Y background, deletion formation after HO induction was 73%, a level similar to wild type. In rad1 and rad1 rfa1-D228Y strains, similar levels of deletions were observed after HO induction, 34 and 35%, respectively. The reduced levels observed in the absence of RAD1 likely result from the requirement for the Rad1/Rad10 endonuclease in the removal of nonhomology during recombination (Fishman-Lobell and Haber 1992). Analysis of rad52 strains indicates that only 3% of the cells have undergone a deletion event after HO induction. This decrease is consistent with the previously described requirement for the RAD52 gene product in the repair of almost all DSBs, including mating type switching and transformation with linear fragments (Resnick and Martin 1976; Malone and Esposito 1980; Orr-Weaver et al. 1981). Interestingly, in rad52 rfa1-D228Y strains, deletion formation increases to 21%. These results demonstrate that the rfa1-D228Y mutation can partially rescue the DSB repair defect observed in rad52 mutants. The partial rescue phenotype was also observed in rad1 rad52 rfa1-D228Y triple mutant strains, as we found an 80-fold increase in deletion events (8%) compared to rad1 rad52 strains (0.1%). Finally, in all genetic backgrounds tested, the level of Leu+ prototrophs was between 15 and 22%, consistent with the idea that these arise from mismatch repair of the heteroduplex formed during SSA.
The rfa1-D228Y mutation alters processing of recombination intermediates in a rad52 background: To examine the mechanism of rfa1-D228Y suppression of rad52-dependent recombination, we analyzed the processing of DSBs and the kinetics of deletion formation at the molecular level. After HO endonuclease induction, DNA was isolated from the cells at specific time points, digested with the restriction enzymes that flank the HO site, and electrophoresed on a denaturing gel. Hybridization of the resulting genomic blot with a LEU2 sequence-specific probe allows the detection of fragments representing the intact assay, HO-cut intermediates, and the recombinant product (Figure 2). Previous studies of HO-induced recombination have demonstrated that after DSB formation, 3′ ssDNA tails are extended on both sides of the HO-cut site by 5′ to 3′ exonucleolytic degradation (White and Haber 1990; Sugawara and Haber 1992). As the degradation proceeds past the flanking restriction enzyme sites, cleavage is blocked and results in a longer single-stranded intermediate that is easily detected with a strand-specific probe (Figure 2).
DNA is isolated from wild-type strains before HO induction, and two fragments (3.9 and 9.5 kb) representing the intact, unrearranged assay are observed (Figure 3A). Thirty minutes after HO induction, an 8.3-kb HO-cut fragment is detected. This fragment is processed into a 5.7-kb recombinant product that is first detected after 1 hr. As a result of the loss of the adjacent StuI site, a 5.1-kb single-stranded intermediate is observed at the 30-min time point. The level of this fragment reaches a maximum by 60 min and then slowly decreases. By 5 hr after induction, the reaction is complete, with the final amount of product reaching 77%. Virtually identical kinetics of deletion formation were also observed for rfa1-D228Y strains (Figure 3B).
In rad52 strains, the level of product formation observed is significantly reduced, representing at most 10% of the total amount of signal detected (Figure 3C). The decreased efficiency of repair is also evidenced by the prolonged presence of both the ssDNA intermediate and the HO-cut fragment, which are detectable up to 6 hr after induction. In fact, the gradual decrease of these intermediates at later time points most likely results from degradation of ssDNA and/or double-stranded DNA. The kinetics of product formation are also delayed in rad52 strains, as detectable levels of product are observed 2 hr later than in wild type (180 min). In contrast, repair of an HO-induced DSB is relatively efficient in rad52 rfa1-D228Y strains (Figure 3D). By 6 hr after induction, the product represents 40% of the total signal detected. However, there is a delay in processing in these double mutants because product formation occurs later (90–120 min after induction) and the intermediates persist longer.
The kinetics of strand degradation in rad52 and rad52 rfa1-D228Y strains were examined in more detail by using an alternative restriction enzyme with a greater number of recognition sites around the HO-cut site (Figure 4). This permits the detection of a ladder of single-stranded intermediates caused by 5′ to 3′ degradation after the HO-induced DSB, which is characteristic of rad52 strains (Sugawara and Haber 1992). At 30 min after an HO-induced DSB in a rad52 strain, the first single ssDNA intermediate is observed (4.4 kb). By 2 hr after induction, a ladder of additional intermediates can be visualized. In rad52 rfa1-D228Y strains, the larger intermediates were not observed, and, throughout the time course, the only detectable ssDNA intermediate was the 4.4-kb fragment (Figure 4). This intermediate was first observed at the 30-min time point and persisted until ∼5 hr after induction. The increased product formation seen in rad52 rfa1-D228Y, combined with the failure to observe large ssDNA intermediates in these double mutants, indicates that the rfa1-D228Y mutation suppresses both aspects of the rad52 DSB repair defect.
In this study, we have used two approaches to define the mechanism by which direct-repeat recombination is stimulated in rfa1-D228Y mutant strains. First, we have examined spontaneous recombination using a pair of LEU2 direct-repeat assays. This analysis indicates that the elevated levels of direct-repeat recombination observed in rfa1-D228Y strains result from an increase in both intrachromatid and sister chromatid recombination. In addition, suppression of the decreased recombination in rad1 rad52 strains by the rfa1-D228Y mutation is mostly caused by stimulation of deletion events by intrachromatid recombination. Second, to determine the mechanistic role of rfa1-D228Y in intrachromatid recombination, processing of recombinants was monitored using an HO-induced deletion assay. This analysis demonstrates that the rfa1-D228Y mutation can partially rescue the DSB repair defect observed in a rad52 background. Molecular characterization of intermediate and product formation in rad52 rfa1-D228Y strains indicates that in addition to increasing the efficiency of product formation, accumulation of large, single-stranded intermediates characteristic of rad52 strains are no longer observed.
Analysis of spontaneous direct-repeat recombination using the distal and proximal configurations of the LEU2 assay indicates that, in an rfa1-D228Y strain, both intrachromatid and sister chromatid recombination are stimulated. Additionally, these two types of recombination display a differential dependence on RAD1 and RAD52. Sister chromatid recombination in rfa1-D228Y strains is largely dependent on RAD52, while the elevated levels of deletion events display a strong dependence on RAD1. It is likely that the majority of the stimulated sister chromatid events observed in an rfa1-D228Y background are associated with gene conversions because it has been demonstrated previously that RAD52 is required for most gene conversion events (Jackson and Fink 1981; Haber and Hearn 1985; Hoekstraet al. 1986; Maloneet al. 1988; Gangloffet al. 1996). However, it is interesting to note that replacement events are not significantly increased in a rfa1-D228Y background; thus, if the rfa1-D228Y allele is stimulating gene conversion, this effect is limited to events associated with a crossover.
On the other hand, the stimulation of sister chromatid recombination may result from synthesis-dependent strand annealing (SDSA, Nassifet al. 1994; Paâqueset al. 1998). In this process, a 3′ end created after a DSB invades the homologous sequences on the sister chromatid and primes DNA synthesis. Resolution subsequently occurs by unwinding the invading strand from the template and annealing with another DNA molecule from the other side of the DSB. In the case of direct repeats, triplications could be generated by repair of a DSB in the distal repeat that invades the proximal repeat on the sister chromatid. Although it has not yet been demonstrated, this mechanism is likely to depend on RAD52, as it involves strand invasion. Thus, SDSA can explain some of the stimulated events observed in a rfa1-D228Y background.
In contrast, the increased level of deletion events observed in an rfa1-D228Y background displays a strong dependence on RAD1, but only a partial dependence on RAD52. Previous studies have shown that RAD1 is required for the removal of nonhomologous sequences in deletion formation, an important step in the recombination process of single-strand annealing (SSA, Fishman-Lobell and Haber 1992). It has also been demonstrated that SSA is only partially RAD52 dependent (Ozenberger and Roeder 1991; Fishman-Lobellet al. 1992; Prado and Aguilera 1995). Thus, an SSA mechanism is most compatible with the stimulated deletion events observed in rfa1-D228Y strains.
The most dramatic effect of the rfa1-D228Y mutation was observed in a rad1 rad52 background. Analysis of individual recombinants indicates that, in contrast to rad1 rad52 strains, a synergistic decrease in triplications and deletions is not observed in rad1 rad52 rfa1-D228Y strains in the distal configuration. Of the two types of events, however, deletions display the greatest stimulation, 100–200-fold compared to rad1 rad52 strains. This suggests that the suppression of decreased recombination in rad1 rad52 strains by the rfa1-D228Y mutation is mostly caused by a stimulation of deletion events by an SSA mechanism.
To analyze further the role of rfa1-D228Y in deletion events, we examined the repair of an HO-inducted DSB located between the leu2 direct repeats. Repair of this DSB occurs by recombination between the direct repeats and specifically results in a deletion event via SSA (Sugawara and Haber 1992). The rfa1-D228Y mutation partially rescues the DSB repair defect observed in rad52 mutants, and it increases the efficiency of repair in a rad1 rad52 background. These results suggest that the rfa1-D228Y mutation has a direct effect on recombination processing/resolution.
Molecular analysis of intermediate and product formation during processing of HO-induced DSB indicates that product is observed at an earlier time point in rad52 rfa1-D228Y strains than in rad52 strains. In contrast to rad52 strains, the double mutants do not display an accumulation of large ssDNA intermediates. One hypothesis to explain the decrease in ssDNA intermediates is that the rfa1-D228Y mutation directly inhibits the exonucleolytic activity responsible for its formation, perhaps via an interaction between an exonuclease and Rfa1. This explanation seems unlikely, however, because the formation of the 5.1-kb ssDNA intermediate does not appear to be delayed in rad52 rfa1-D228Y strains compared to rad52 strains, suggesting that the exonuclease activity is not dramatically reduced in these strains. In addition, mutations in the RAD50 gene decrease ssDNA formation in a rad52 background without affecting the level of product formation (Sugawara and Haber 1992). Thus, the rfa1-D228Y mutation likely increases the efficiency of a processing step that occurs subsequently to ssDNA degradation, thereby preventing the appearance of the ssDNA intermediates seen in rad52 mutants.
Previous studies have shown that rad52 mutants display a significant decrease in viability after induction of a DSB (White and Haber 1990; Fishman-Lobellet al. 1992; Sugawara and Haber 1992). However, there are some cases where repair of a DSB occurs efficiently. Although Ozenberger and Roeder (1991) have demonstrated that an HO-cut site could be efficiently repaired when embedded in the tandemly arrayed rDNA cluster and, to a lesser extent, within a CUP1 array, a DSB located between two repeats is only repaired ∼10–20% of the time (Fishman-Lobellet al. 1992; Ivanov and Haber 1995; this study). Thus, the efficiency of repair appears to be dependent on the number of repeats surrounding the break. These results suggest that rad52 mutants may possess a reduced ability to find homologous sequences. This hypothesis is supported by the observation that the Rad52 protein can promote the annealing of two complementary DNA strands in vitro (Mortensenet al. 1996).
A decreased efficiency of pairing in rad52 mutants may account for both decreased level of product formation and increased ssDNA degradation. One explanation for the ability of rfa1-D228Y to partially suppress both decreased product formation and increased degradation is that the mutant protein increases the efficiency of homologous pairing. In fact, E. coli ssDNA-binding protein has been shown to stimulate DNA renaturation in vitro (Christiansen and Baldwin 1977). A direct role for RP-A in pairing seems unlikely, however, because the conditions required for E. coli ssDNA-binding protein activity are distinctly nonphysiological, requiring high concentrations of spermidine and magnesium ion. Alternatively, the role of wild-type RP-A may be to inhibit pairing, and the rfa1-D228Y mutant protein may relieve this inhibition. This view is supported by recent in vitro experiments in which ssDNA substrates preincubated with S. cerevisiae RP-A inhibited subsequent homologous pairing by the strand exchange protein Rad51 (Sung 1997b). Interestingly, this inhibition can be overcome by addition of the Rad52 protein (Sung 1997a; Newet al. 1998; Shinohara and Ogawa 1998).
On the basis of our molecular results and these in vitro experiments, we suggest that in the absence of Rad52, single-strand annealing occurs inefficiently because RP-A inhibits the access of a pairing or annealing protein (complex) to the ssDNA. In strains containing the rfa1-D228Y mutation, the level of the RP-A complex is reduced and/or its interaction with DNA is destabilized, permitting pairing between homologous sequences. This view is consistent with the idea that the role of Rad52 in homologous pairing is to alter the binding of RP-A to allow a pairing or strand exchange protein (complex) to access the ssDNA.
We thank Adam Bailis and Serge Gangloff for comments on the manuscript. We also thank members of the Rothstein laboratory for helpful discussions concerning this work. This research was supported by National Institutes of Health grants GM07088 (J.S.), CA09503 (J.S.), and GM50237 (R.R.).
Communicating editor: M. Lichten
- Received March 20, 1998.
- Accepted October 13, 1998.
- Copyright © 1999 by the Genetics Society of America