|Home | About | Journals | Submit | Contact Us | Français|
Homologous recombination (HR) is critical for DNA double-strand break (DSB) repair and genome stabilization. In yeast, HR is catalyzed by the Rad51 strand transferase and its “mediators,” including the Rad52 single-strand DNA-annealing protein, two Rad51 paralogs (Rad55 and Rad57), and Rad54. A Rad51 homolog, Dmc1, is important for meiotic HR. In wild-type cells, most DSB repair results in gene conversion, a conservative HR outcome. Because Rad51 plays a central role in the homology search and strand invasion steps, DSBs either are not repaired or are repaired by nonconservative single-strand annealing or break-induced replication mechanisms in rad51Δ mutants. Although DSB repair by gene conversion in the absence of Rad51 has been reported for ectopic HR events (e.g., inverted repeats or between plasmids), Rad51 has been thought to be essential for DSB repair by conservative interchromosomal (allelic) gene conversion. Here, we demonstrate that DSBs stimulate gene conversion between homologous chromosomes (allelic conversion) by >30-fold in a rad51Δ mutant. We show that Rad51-independent allelic conversion and break-induced replication occur independently of Rad55, Rad57, and Dmc1 but require Rad52. Unlike DSB-induced events, spontaneous allelic conversion was detected in both rad51Δ and rad52Δ mutants, but not in a rad51Δ rad52Δ double mutant. The frequencies of crossovers associated with DSB-induced gene conversion were similar in the wild type and the rad51Δ mutant, but discontinuous conversion tracts were fivefold more frequent and tract lengths were more widely distributed in the rad51Δ mutant, indicating that heteroduplex DNA has an altered structure, or is processed differently, in the absence of Rad51.
DNA repair is critical for genome stability and tumor suppression in higher eukaryotes. DNA double-strand breaks (DSBs) are critical lesions that result from exposure to exogenous agents (ionizing radiation or genotoxic chemicals) and from endogenous sources, such as nucleases and spontaneous DNA lesions that cause replication fork collapse. DSBs can be repaired by homologous recombination (HR) or by nonhomologous end joining (NHEJ). In the yeast Saccharomyces cerevisiae, the primary DSB repair pathway is homologous recombination (HR), and the key HR proteins are members of the Rad52 epistasis group (Rad51, Rad52, Rad54, Rad55, Rad57, and Rad59). The Mre11/Rad50/Xrs2 (MRX) complex has roles in both HR and NHEJ. The most common HR outcome is gene conversion, a conservative and highly accurate mode of DSB repair that occurs with or without an associated crossover. In cells with mild to severe HR defects, other outcomes may occur, including chromosome loss or nonconservative repair by either break-induced replication (BIR) or single-strand annealing (SSA) (10, 16, 25, 27). Inefficient or inaccurate DSB repair by either HR or NHEJ has been linked to the development and progression of cancer (28, 59).
DSB repair by HR involves a series of steps that begins with 5′-to-3′ single-strand resection regulated by MRX, and binding of the single-stranded DNA (ssDNA) tail by replication protein A. Replication protein A is then exchanged with Rad51 to produce Rad51 nucleoprotein filaments that are capable of searching for, and invading, homologous sequences elsewhere in the genome (sister chromatids, homologous chromosomes, or repeated sequences). The Rad51-related protein Dmc1 is also required for meiotic HR (32). Rad51 filament formation is promoted by the “mediator” proteins Rad52, Rad55, and Rad57; Rad54 also mediates Rad51 filament formation in vitro, although filaments form normally in rad54Δ mutants in vivo (19, 51, 54, 60). Genetic and biochemical studies suggest that Rad54 promotes invasion of the intact (donor) duplex by the Rad51 filament and branch migration of the joint molecule producing heteroduplex DNA (hDNA) and dissociates Rad51 from ssDNA to allow repair synthesis by extension of the invading 3′ end (4, 15, 40, 47, 48). The invading 3′ end is extended by repair synthesis past the site of the DSB and can then dissociate from the donor duplex and anneal with the other ssDNA tail in a process termed synthesis-dependent strand annealing, a noncrossover pathway. Alternatively, both ends may invade a donor, producing Holliday junction intermediates that can be resolved with or without crossovers. Mismatches arise in hDNA when interacting duplexes have sequence differences (“markers”). Mismatch repair is typically biased, with the strand from the donor (unbroken) duplex used as a mismatch repair template, resulting in regions of marker loss in the recipient (broken) duplex that are termed conversion tracts (38). Because most mismatch repair proceeds by a long-patch, excision-based mechanism, conversion tracts are usually continuous, i.e., when two markers flanking a central marker are converted, the central marker almost always coconverts. Thus, conversion tracts result from mismatch repair of hDNA produced during strand invasion and branch migration and when an extended 3′ end anneals to ssDNA during synthesis-dependent strand annealing.
The Rad51 filament plays a central role in strand invasion, and DSB repair via gene conversion is severely reduced or eliminated in the rad51Δ mutant, with repair instead occurring by SSA or BIR. With only a single DSB, SSA is limited to interactions between linked, direct repeats as resection to ssDNA tails exposes complementary single strands that anneal to repair the break. SSA may be important for repair when two or more DSBs occur near each other (49). Because SSA does not require strand invasion, it is independent of Rad51, Rad54, Rad55, and Rad57 but requires the strand-annealing activity of Rad52 (10, 27, 49). BIR was initially described in rad51Δ cells (25, 46). Several studies have revealed an inefficient Rad51-independent BIR mechanism that requires Rad59, Tid1, and MRX and a more efficient Rad51-dependent mechanism that requires Rad52, Rad55, Rad57, Rad54, and Ku (6, 16, 26, 46). Even the more efficient BIR mechanism is rare or absent in wild-type cells when gene conversion is possible (25, 35, 46), although BIR appears to be important for telomerase-independent telomere maintenance (24).
Despite the central role of Rad51 in strand invasion, Rad51-independent gene conversion has been detected in certain contexts. Spontaneous gene conversion between inverted repeats is reduced only 4-fold by deletion of RAD51, whereas deletion of RAD52 reduces these events by 3,000-fold (42). Ectopic gene conversion involving repair of plasmid double-strand gaps with plasmid or chromosomal repair templates was readily detected in the rad51Δ mutant, albeit ~100-fold less efficiently than in the wild type; these recombination events were dependent on Rad52 and Rad59 (2). Yeast mating-type switching involves intrachromosomal gene conversion resulting from repair of an HO nuclease-induced DSB at the MAT locus. This is a lethal event in the rad51Δ mutant; however, MAT switching does occur in the rad51Δ mutant when interacting alleles are present on plasmids. These plasmid events were originally explained as reflecting a relaxed requirement for Rad51 when donor sequences are not silenced in heterochromatin, as in normal MAT switching (50), but an alternative explanation is that the plasmid events depend on BIR between closely linked alleles (2). Thus, Rad51-independent DSB repair by gene conversion has been detected only for ectopic events (inverted repeats, or plasmid-chromosome or plasmid-plasmid interactions); there are no prior reports of such repair between chromosomal (allelic) loci in the absence of Rad51 (17).
By using a sensitive DSB repair assay, we demonstrate that yeast is capable of Rad51-independent DSB repair resulting in interchromosomal gene conversion. The efficiency of this repair is similar to that of Rad51-independent BIR, and these processes share similar genetic requirements: both are independent of Rad55, Rad57, and Dmc1 but require Rad52. Spontaneous gene conversion, however, was detected in the absence of Rad51 or Rad52, but not both proteins. Crossovers associated with conversion occur at about the same frequency during Rad51-dependent and -independent events, but Rad51-independent gene conversion tracts were frequently discontinuous and generally longer than in the wild type, suggesting that hDNA forms or is processed differently in the absence of Rad51.
Plasmid pHSS19Trp1rad51 has 849-bp and 1,151-bp fragments from up- and downstream of RAD51 flanking TRP1. Standard procedures were used for yeast culture and chromosome modification (35), and the structures of all modified chromosomes were confirmed by Southern and PCR analyses. All yeast strains were isogenic with diploid JC3598 (37) (MATa-inc/MATα-inc ade2-101/ade2-101 his3-200:HIS3:telV/his3-200 lys2-801::pHSSGALHO::LYS2/lys2-801 trp1-Δ1/trp1-Δ1 leu2-Δ1/leu2-Δ1 RscRI-ura3R-HO432-LEU2/RscBam-ura3-X764-LEU2). GALHO indicates the source of the HO nuclease, which cleaves the HO site at position 432 in ura3 (HO432). All strain genotypes are listed in Table S1 in the supplemental material. RAD51 was replaced with TRP1 by one-step transplacement using pHSS19Trp1rad51. RAD52, RAD55, RAD57, and DMC1 were replaced with KanMX by one-step transplacement with PCR-amplified fragments from appropriate strains from the yeast deletion set. Double and triple mutants were constructed by one-step KanMX replacements or mating/sporulation of appropriate single/double mutants. All markers in and around ura3, and the telomere-proximal HIS3:telV marker, were confirmed in all strains by mapping PCR amplification products.
HR was analyzed in diploid yeast strains carrying ura3 heteroalleles at the normal chromosome V locus flanked by pUC19 and LEU2 (35). One copy of ura3 (“recipient”) was inactivated by an insertion of a 24-bp HO nuclease site flanked by 11 phenotypically silent mutations that create and/or destroy restriction sites (restriction fragment length polymorphism markers). These markers were used to map conversion tracts in recombinant products. This chromosome also carried a telomere-proximal HIS3 gene 110 kbp upstream from ura3. The second ura3 (“donor”) was inactivated by a frameshift mutation termed X764. An integrated HO gene was regulated by the GAL1 promoter (GALHO). Cells were incubated in rich medium with 2% glycerol for 24 h to deplete intracellular glucose prior to incubation for 6 h in rich medium with 2% galactose (YPGal), or 2% glucose (YPD) as an uninduced control. The depletion of glucose ensures rapid induction of GALHO in YPGal, which leads to a single DSB in the recipient ura3 allele. Both MAT loci had single-base mutations, rendering them insensitive to cleavage by HO. Recombinants with short conversion tracts that extended from the HO site past X764 were Ura positive (Ura+), while those with tracts that included X764 remained Ura negative (Ura−). In HR-proficient cells, both types of products could be analyzed by a nonselective assay in which cells incubated for 6 h in liquid YPGal or YPD were seeded onto nonselective (YPD) plates and the resulting colonies were identified as Ura+ or Ura− recombinants, or parental Ura−, using appropriate selective media. Parental colonies were identified as “reinducible” to Ura+ when transferred first to YPGal plates and then to uracil omission plates (35), whereas Ura− recombinants were homozygous for the X764 frameshift mutation and could not be reinduced to Ura+. In cells with severe HR deficiencies (e.g., rad51Δ), only Ura+ recombinants were analyzed by using a selective assay in which ~107 cells from YPD or YPGal cultures were directly plated on uracil omission medium. In either assay, recombinants were also screened for the His phenotype by replica plating them onto histidine omission medium. In the nonselective assay, HR frequencies were calculated as the number of Ura+ or Ura− recombinants per YPD colony scored, with 1,000 to 1,500 colonies scored in each of three or four independent determinations per strain. In the selective assay, HR frequencies were calculated as the number of Ura+ recombinants per viable cell plated on uracil omission medium, with three or four independent determinations per strain. The number of viable cells was determined by parallel plating of appropriate dilutions on YPD.
Products isolated in nonselective assays may include those arising by HR, BIR, chromosome loss, and half-crossovers (25, 35). The HIS3:telV marker on broken chromosomes was assayed by quantitative real-time PCR to determine chromosome loss among Ura− His− products, as described previously (21). His− products can result from crossovers in G2 cells, and these have an associated product with two copies of HIS3:telV (His++), identified by PCR. His− products arising by BIR or half-crossovers comprise nonloss/noncrossover His− products (35). BIR yields two chromosomes, and half-crossovers yield only one, so distinguishing these requires quantitative analysis of the chromosome number, but this was not performed because we focused on His+ (gene conversion) products. Gene conversion tract spectra were generated by rescuing and mapping donor and recipient alleles from yeast genomic DNA linked to pUC19, as described previously (22, 35). All rad51Δ products were isolated from independent parent cultures. The rad51Δ rad55Δ rad57Δ products were isolated from a limited set of parent cultures, and two pairs of products with identical discontinuous tracts were potential sibling pairs, but only one of each pair was counted to ensure product independence. Limited marker analysis of rad52Δ HR products was performed by mapping PCR products amplified from yeast genomic DNA from subcloned products; subcloning prevents sectored colonies from being scored as heterozygous. In this analysis, loss of heterozygosity (LOH) can be detected but linkage information is unavailable because the two alleles coamplify. Statistics were calculated by using t tests unless otherwise specified.
HR in the absence of Rad51 has been detected for spontaneous events and DSB-induced ectopic events (2, 42, 50), but not DSB-induced allelic (interchromosomal) gene conversions (17, 25, 50). While testing the effects of Rad51 ATPase defects on the repair of HO nuclease-induced DSBs by allelic gene conversion between ura3 heteroalleles in diploid yeast (Fig. (Fig.1),1), we constructed a rad51Δ strain as a control. In a nonselective assay that provided estimates of HR, BIR, and chromosome loss frequencies, HR was not detected in the rad51Δ mutant. In wild-type cells, nearly all DSBs in ura3 were repaired by HR (gene conversion); BIR was not detected, and chromosome loss was very rare (Fig. (Fig.2A)2A) (35). In the rad51Δ mutant, HR was replaced by chromosome loss and BIR (Fig. (Fig.2A2A).
When rad51Δ colonies were transferred from YPD to YPGal plates (which induces GALHO), incubated for 1 to 2 days, and then transferred to uracil omission medium, many colonies showed Ura+ papillae. Such Ura+ products could arise by BIR, which is known to occur in the absence of Rad51 (14, 25, 46). However, BIR produces only His− products (Fig. (Fig.1),1), yet a substantial fraction (~50%) of the Ura+ papillae were His+ (data not shown) and potentially arose by allelic gene conversion. Because Ura+ papillae arose far less frequently when colonies were maintained on YPD (with GALHO repressed) prior to transfer to uracil omission medium, the Ura+ His+ colonies were apparently products of DSB repair. Because these products arose at low frequencies, we employed a sensitive, selective assay to measure Ura+ His+ induction levels in the rad51Δ mutant and in a wild-type control strain. Direct selection of Ura+ products detects only short-tract products; long-tract gene conversion products remain Ura− and are not detected. The selective assay also fails to detect chromosome loss and some BIR events (see Materials and Methods). As shown in Fig. Fig.2B,2B, GALHO induction increased the frequency of Ura+ His+ products by 190-fold above spontaneous (uninduced) levels in wild-type cells. The induced Ura+ His+ frequency in the rad51Δ mutant was far lower than in the wild type (6.6 × 10−6 versus 9.5 × 10−2), but spontaneous levels were also reduced, and there was a 33-fold increase in Ura+ His+ products upon GALHO induction.
Although the absolute frequency of induced Ura+ His+ products in the rad51Δ mutant was low, it was unlikely that these products arose by imprecise NHEJ, because this occurs at very low frequencies in haploids and even lower frequencies in diploids (5, 29). To rule out NHEJ as a source of the induced Ura+ His+ products, we performed several control experiments. First, we induced GALHO in the haploid parent of our wild-type diploid carrying ura3 with the HO site (strain JC3441) and a rad51Δ derivative (strain JC3702), but no Ura+ products were detected in either strain (<2.5 × 10−8). Next, we induced GALHO in a diploid that carried the ura3-HO site allele but lacked a ura3 donor in the homologous chromosome V (strain TP3826); again, no Ura+ products were detected (<2.5 × 10−8). Finally, we used PCR to amplify the ura3 loci from several Ura+ His+ products of the original rad51Δ diploid and mapped these products with NcoI (the natural restriction site in the donor opposite the HO site insertion). In all cases, the products were homozygous NcoI+, consistent with nonreciprocal information transfer from the donor to the recipient ura3. Together, these results indicate that the induced Ura+ His+ colonies that arose in the absence of Rad51 were bona fide gene conversion products.
Yeasts encode two mitotic proteins related to Rad51, the paralogs Rad55 and Rad57. Like Rad51, Rad55 and Rad57 bind DNA and are ATPases, raising the possibility that the paralogs are responsible for the residual gene conversion in the absence of Rad51. Because Rad55 and Rad57 function as a heterodimer (53), we first tested a rad51Δ rad55Δ double mutant. As shown in Fig. Fig.3A,3A, GALHO induction led to similar levels of Ura+ His+ products in the rad51Δ mutant and the double mutant. To test whether Rad57 was responsible for the induced gene conversion in the rad51Δ rad55Δ double mutant, we tested a rad51Δ rad55Δ rad57Δ strain. As shown in Fig. Fig.3A,3A, DSB repair by gene conversion in the triple mutant was similar to that in the rad51Δ mutant and the rad51Δ rad55Δ double mutant. Dmc1 is more closely related to Rad51 than to Rad55/Rad57, but Dmc1 is expressed only in meiosis (induced by nitrogen starvation). However, because allelic DSB repair by gene conversion occurs at such low levels, it was possible that the rare HR events were occurring in a subpopulation of cells experiencing metabolic stress, resulting in transient Dmc1 expression and an HR-competent state. We tested this with a rad51Δ dmc1Δ double mutant and again found that the efficiency of DSB-induced gene conversion was similar to that in the rad51Δ mutant (Fig. (Fig.3A).3A). Thus, Rad51-independent, DSB-induced allelic gene conversion is independent of Rad55, Rad57, and Dmc1.
DSB repair deficiencies caused by mutations in RAD55 or RAD57 are more severe at 23°C than at 30°C. These cold-sensitive phenotypes are thought to reflect a greater requirement for Rad55/Rad57 mediator function at lower temperatures because they are suppressed by overexpression of Rad51 and by mutations in Rad51 that increase the affinity of Rad51 for DNA (7, 13, 23). Because Rad55/Rad57 are thought to function only in Rad51-dependent DSB repair, we expected that Rad51-independent DSB repair by gene conversion would be insensitive to reduced temperature. We tested this by measuring DSB-induced Ura+ His+ frequencies at 23°C in the rad51Δ mutant alone and in combination with the rad55Δ and rad57Δ mutants. As shown in Fig. Fig.3B,3B, the lower temperature had no significant effect on the level of Rad51-independent gene conversion in rad51Δ or rad51Δ rad55Δ strains, and while there was a twofold reduction in the rad51Δ rad55Δ rad57Δ triple mutant compared to the rad51Δ mutant, DSB-induced gene conversion in the triple mutant was significantly higher than the spontaneous level. These results indicate that Rad55, and to a large extent Rad57, functions mainly in the Rad51-dependent HR pathway.
Unlike Rad51, Rad52 has been shown to be essential for HR repair of DSBs in many substrates (17). We employed the selective ura3 HR assay to determine whether we could detect a low level of chromosomal DSB repair resulting in gene conversion in the rad52Δ mutant. Initial measurements indicated that GALHO induction led to a slight increase (~2-fold) in Ura+ His+ products, but the difference was not statistically significant (Fig. (Fig.3A).3A). There was wide variation in the frequencies of Ura+ His+ products under both growth conditions, with standard deviations approaching 100% of the averages. This is consistent with most products arising spontaneously, with variability reflecting differential timing of spontaneous events during population expansion (“jackpots”). Fluctuation analysis is appropriate for measuring spontaneous HR rates, as it effectively eliminates problems associated with jackpots, but it is not suitable for measuring DSB-induced HR. In subsequent experiments with the original rad52Δ strain (TP3828) and with an independently constructed rad52Δ strain (RS3830), we continued to observe modest increases in Ura+ His+ frequencies upon GALHO induction (data not shown). In cell populations with a low background of spontaneous products, GALHO induction led to statistically significant increases in both total Ura+ and Ura+ His+ products (Fig. (Fig.3C).3C). Thus, there appears to be a low level of Rad52-independent DSB repair by gene conversion, but it is sometimes masked by spontaneous jackpots. This result is consistent with our previous report showing a small (statistically significant) increase in DSB-induced gene conversion between ura3 direct repeats in a rad52-1 mutant (P < 0.02 in both MATΔ and MATα strains) (34). Both spontaneous and DSB-induced events are dependent on Rad51, as none were observed in the rad51Δ rad52Δ double mutant (<6 × 10−9) at 30°C (Fig. (Fig.3A)3A) or 23°C (data not shown). Together, the results indicate that Rad51 is required for both spontaneous and DSB-induced gene conversion in the absence of Rad52, and vice versa.
In the absence of Rad52, DSBs at MAT in haploid cells are usually lethal, with rare survivors resulting from imprecise NHEJ (29). In diploids, DSBs at MAT lead primarily to loss of the broken chromosome, but rare repair products arise via nonreciprocal crossing over (half-crossover) with loss of the second chromosome (25). In contrast, spontaneous gene conversion with retention of both chromosomes has been detected in rad52 mutants (8). We next investigated how Ura+ His+ and Ura+ His− products arise in the absence Rad52, with or without GALHO induction. With this ura3 allele arrangement, Ura+ His+ products cannot arise by half-crossovers without associated gene conversion (Fig. (Fig.4A),4A), and Ura+ His+ products induced by DSBs cannot arise by BIR (Fig. (Fig.1),1), suggesting that this class forms by a different mechanism. To determine whether Ura+ products arose by conservative gene conversion, we analyzed HIS3:telV, HO, X764, and B3′ markers in 18 rad52Δ products and found a variety of product types (Fig. (Fig.4B).4B). The majority of products from both uninduced and GALHO-induced cultures were heterozygous at one or more ura3 markers and/or HIS3 (7 of 8 and 8 of 10, respectively), ruling out the half-crossover/chromosome loss mechanism described by Malkova et al. (25). Note that BIR and half-crossovers differ in that the latter yields only one full-length chromosome. Unlike wild-type cells, in which 100% of GALHO-induced products convert the HO site (35), there was a modest (2:1) bias in favor of conversion of the HO allele in GALHO-induced rad52Δ cultures, consistent with the relatively low level of DSB-induced gene conversion in the rad52Δ mutant. Although a similar twofold bias in HO site conversion was observed in the rad52Δ diploid when GALHO was not induced (Fig. (Fig.4B),4B), this may reflect leaky GALHO expression or sampling error. In both rad51Δ and rad52Δ strains, DSBs are usually not repaired and chromosome loss (yielding Ura− His− products) is the most common outcome.
In wild-type cells, the majority of allelic DSB repair occurs by gene conversion without an associated crossover. In the ura3 heteroallele system employed in this study, conversion without crossing over produces Ura+ or Ura− recombinants that retain heterozygosity at the HIS3:telV marker (His+/−; phenotypically His+). Crossovers can be detected as either His− products or, among His+ products as those that gain a second copy of HIS3:telV (His++). Only His++ products can be unambiguously attributed to crossovers because BIR or half-crossovers can also produce His− products. Crossovers are relatively rare among mitotic gene conversions in wild-type cells (25, 35), and only ~5% of induced Ura+ colonies are His− (Fig. (Fig.5A).5A). In the rad51Δ mutant, this percentage increased by sevenfold, and His− fractions were similarly increased in double and triple mutants with rad51Δ, rad55Δ, rad57Δ, and dmc1Δ (Fig. (Fig.5A).5A). To distinguish whether the His− products arose by full crossovers or BIR/half-crossovers, we analyzed DSB-induced Ura+ His+ products from the rad51Δ and rad51Δ rad55Δ rad57Δ strains and found that 1 of 17 and 1 of 21 products, respectively, were His++ (data not shown). In the wild type, a similar fraction of Ura+ His+ products were His++ (2 of 20) (35). Thus, crossovers are associated with gene conversions at similar frequencies in the wild type and the rad51Δ mutant. In addition, because only ~5% of induced Ura+ His+ products are His++ crossovers in the rad51Δ background, ~95% of induced Ura+ His− products result from BIR or half-crossovers. As with gene conversion, Ura+ His− products arose at similar frequencies in the rad51Δ single mutant and in double/triple mutant combinations with rad55Δ, rad57Δ, and dmc1Δ (Fig. (Fig.5B).5B). Although the Ura+ His− level was a few times lower in the rad51Δ mutant than the rad51Δ rad55Δ double mutant (Fig. (Fig.5B),5B), in a separate experiment in which we tested these strains in parallel, the levels were not significantly different (P = 0.23) and DSB-induced Ura+ His− products arose in the rad51Δ mutant at a level significantly higher than the spontaneous level (P < 0.0001) (data not shown).
The majority of Ura+ products in the rad52Δ mutant were His− (Fig. (Fig.5A)5A) and could have arisen by BIR, half-crossovers, or chromosome loss after conversion, but similar to Rad52-independent conversions, these events were very rare (Fig. (Fig.5B).5B). The data in Fig. Fig.5B5B were generated from cultures incubated at 30°C, and similar results were obtained at 23°C (data not shown). The extremely low level of chromosomal BIR in the absence of Rad52 is consistent with results obtained in plasmid-directed BIR assays (6, 18). No Ura+ His− products were detected in the rad51Δ rad52Δ double mutant. We conclude that Rad51-independent BIR/half-crossover shows the same genetic dependencies as Rad51-independent DSB repair by gene conversion (Fig. (Fig.3A3A and and5B)5B) and that both conservative (conversion) and nonconservative DSB repair mechanisms are more strongly reduced in the absence of Rad52 than in the absence of Rad51.
To further investigate the mechanism of Rad51-independent DSB repair by allelic gene conversion, we analyzed donor and recipient alleles in 17 DSB-induced Ura+ His+ products from the rad51Δ mutant and 21 products from the rad51Δ rad55Δ rad57Δ triple mutant; not surprisingly, these spectra were quite similar, and the two sets were included in a “combined rad51Δ” spectrum for several of the subsequent analyses. Conversion tracts of rad51Δ products were compared to tracts in 45 wild-type Ura+ products (Fig. (Fig.6).6). As in the wild type, donor loci remained largely unchanged: donor conversions occurred in 3 of 17 rad51Δ products, 0 of 21 triple-mutant products, and 1 of 45 wild-type products (there were no significant differences between the wild-type and rad51Δ, the triple mutant, or the combined rad51Δ sets [all P > 0.05; Fisher exact tests]). The average tract length (± standard deviation [SD]) of the combined rad51Δ spectrum was 1,320 ± 1,237 bp; this is somewhat longer than the 895 ± 916 bp in the wild type, but the difference is not statistically significant (P = 0.076). Although there was a trend toward longer tracts in the rad51Δ mutant, a significantly higher fraction of the combined rad51Δ products (8 of 38) had very short tracts (<200 bp), whereas none of the 45 wild-type tracts were <227 bp (P = 0.012; Fisher exact test). Four rad51Δ products, but no wild-type products, converted just the HO site (tract length, <53 bp), which is also a significant difference (P = 0.04; Fisher exact test) (Fig. (Fig.66 and Table Table1).1). Thus, the rad51Δ tracts were more widely distributed, with both very short and very long tracts, whereas wild-type tracts were more tightly clustered around the mean (Fig. (Fig.7A).7A). Given that rad51Δ products included both shorter and longer tracts, it is not surprising that frequencies of marker conversions as a function of distance from the DSB were similar for wild-type and rad51Δ products (Fig. (Fig.7B).7B). The two tract spectra showed similar frequencies of unidirectional and bidirectional tracts, and both lacked 3′ unidirectional tracts (Table (Table11).
A striking feature of the Rad51-independent products was the high frequency of discontinuous conversion tracts. In wild-type cells, only 4.4% were discontinuous (2 of 45), but 20 to 30% were discontinuous in the rad51Δ, triple-mutant, and combined rad51Δ spectra (Fig. (Fig.7C).7C). In most cases, tracts had a single discontinuity, but two of nine rad51Δ discontinuous tracts had two discontinuities (types D10 and D11) (Fig. (Fig.6).6). Most tract discontinuities were present in the recipient allele, with the donor allele retaining its parental markers. In three rad51Δ products, tracts were discontinuous because markers converted in separate regions in donor and recipient chromosomes (types D9 to D11) (Fig. (Fig.6).6). However, donor conversion did not always result in a tract discontinuity. For example, in the rad51Δ product D11 (Fig. (Fig.6),6), the conversion of the donor Stu667 marker was continuous with the recipient tract. Conversions of donor and recipient markers reflect mismatch repair in hDNA, and while conversion tracts may not reflect the full extent of hDNA (due to restoration-type mismatch repair), tracts do represent the minimum extent of hDNA. Tract discontinuities can result from three sources: discontinuous mismatch repair in a continuous region of hDNA, repair in multiple hDNA regions arising during a single HR event, and independent HR events. The preponderance of discontinuous tracts in the rad51Δ spectra, however, cannot be explained by independent HR events (see below).
Low-level spontaneous HR occurs in both rad51Δ and rad52Δ mutants. Because spontaneous HR is eliminated in the rad51Δ rad52Δ double mutant, spontaneous Rad52-independent HR depends on Rad51, and vice versa. Spontaneous HR has been proposed to initiate at spontaneous DSBs, such as when blocked replication forks collapse at single-strand damage. However, Lettier et al. (18) characterized a class of rad52 mutants that cannot repair DSBs by gene conversion, SSA, or BIR but show wild-type or higher spontaneous HR rates. These rad52 mutants also show wild-type levels of HR stimulated by single-strand damage induced by UV. Because Rad52 foci are seen most often during S phase (18, 20) and spontaneous HR is markedly reduced in rad52Δ (17) (compare Fig. Fig.2B2B and and3A),3A), it appears that Rad52 plays a critical role in HR-mediated restart of blocked or stalled replication forks. A role for Rad52 in replication-associated HR is further indicated by the observation that Holliday junctions form spontaneously in S phase in a Rad52-dependent manner but independently of Rad51 and its paralogs (62).
There is strong sequence conservation between yeast Rad52 and homologs in higher eukaryotes, yet mammalian RAD52 is not critical for DSB repair by HR, perhaps because its DSB repair function has been taken over by BRCA2 (43, 56, 61). However, recent evidence indicates that human RAD52 is highly responsive to hydroxyurea-induced replication fork stalling (J. Wray, J. Liu, J. A. Nickoloff, and Z. Shen, submitted for publication), suggesting that its role in replication may be conserved through evolution. Prokaryotes also express proteins with Rad52-like activities, including Escherichia coli RecT (9, 36) and RecFOR (30) and other prokaryotic/viral Rad52 homologs (11). It will be interesting to determine if these proteins also function in restarting blocked replication forks.
Given the evidence that spontaneous HR initiates when replication forks encounter single-strand lesions, such as nicks or gaps (18, 44), how might Rad51 and Rad52 promote these HR events when both or only one of these proteins is present? Stalled or collapsed forks may be restarted by mechanisms that can result in HR, including sister strand invasion or fork regression, which forms a “chicken foot” structure (45). In the absence of Rad51, Rad52 could promote strand annealing between single-stranded regions (2), and in the absence of Rad52, Rad51 may load onto ssDNA inefficiently but, once loaded, may efficiently catalyze strand invasion (52), thus accounting for the low-level spontaneous HR in rad51Δ and rad52Δ single mutants, but not in the double mutant. The extremely low level of DSB-induced HR in the rad52Δ mutant also depends on Rad51 and may similarly reflect the low efficiency of Rad51 loading onto ssDNA in the absence of Rad52.
Rad51 plays a central role in mediating strand invasion during HR. Although there have been a number of reports of Rad51-independent ectopic HR, in most cases, these events can be explained as resulting from SSA, BIR, or a combination of these processes (1, 2, 14, 50). Rad51 was reported to be essential for DSB-induced allelic gene conversion at MAT (25, 50). Although DSB induction at MAT is very efficient, MAT switching is not selectable and therefore has low sensitivity for detecting rare HR events in mutant cells. With a selective ura3 assay, we found that Rad51-independent, DSB-induced allelic gene conversion was >30-fold higher than the spontaneous level. This increase was observable because rad51Δ also markedly reduces spontaneous HR (Fig. (Fig.2B),2B), highlighting the importance of Rad51 for all HR events. We observed ~30% cleavage of the HO site in ura3 after a 6-h GALHO induction (unpublished results), which probably reflects limited accessibility of HO nuclease to its site in the ura3 chromatin context. The more efficiently cleaved MAT locus lacks nucleosomes at the HO recognition site (57). If ura3 were cleaved with 100% efficiency, detectable (Ura+) HR would increase ~3-fold to ~2 ×10−5 in the rad51Δ mutant; this is 100-fold above the spontaneous level. Because a selective assay is required to detect these low-frequency events, coconversion of the HO site and X764 (Ura−) is not detected. In the wild type, long-tract Ura− products comprise ~66% of gene conversions (35), and our results indicate that long tracts may comprise an even greater fraction of Rad51-independent conversions. Thus, the absolute efficiency of Rad51-independent gene conversion may exceed 10−4, but this is still several orders of magnitude less efficient than the wild type.
Rad55 and Rad57 probably evolved from Rad51 via gene duplication. Our results indicate that they have lost most or all strand invasion activity and function mainly within the Rad51 HR pathway, consistent with a prior genetic analysis (41). We also ruled out a contribution from Dmc1 in Rad51-independent HR. Thus, Dmc1 appears to function only in meiosis, and the Rad51-independent mitotic gene conversions described here are not mediated by leaky Dmc1 expression in mitotic cells. Together, these results indicate that DSB repair by gene conversion can occur independently of the two known yeast strand invasion proteins and their family members, albeit with low efficiency.
BIR is a nonconservative DSB repair process that involves single-ended strand invasion and extension to the end of the chromosome. Two BIR pathways have been defined (16, 46), and the more efficient pathway requires the same strand invasion proteins as gene conversion (Rad51, Rad52, Rad55, Rad57, and Rad54) (6). In wild-type cells, this single-end pathway is far less efficient than gene conversion (25, 35), despite the fact that conversion requires coordination of two broken ends with a donor duplex. DSB repair by the less efficient Rad51-independent BIR pathway appears to have an efficiency similar to that of Rad51-independent gene conversion (Fig. (Fig.3A3A and and5B).5B). This indicates that in the absence of Rad51, single-ended and coordinated two-ended events occur with roughly equal probabilities. Rad51-independent gene conversion and BIR are both independent of Rad55, Rad57, and Dmc1 but dependent on Rad52 (Fig. (Fig.3A3A and and5B)5B) and probably Rad59 (2). The MRX complex regulates end resection at DSBs (38) and is required for Rad51-independent BIR (16, 46), so it is likely that MRX is also required for Rad51-independent gene conversion. Rad51-independent gene conversion and BIR are therefore likely to reflect Rad52-/Rad59-mediated strand annealing (2), perhaps between resected ssDNA at the DSB and transient single-stranded donor sequences formed during replication, transcription, or natural DNA “breathing.” If Rad52/Rad59 efficiently anneal ssDNA, the rate-limiting step may be the rare occurrence of ssDNA in duplex donor DNA. If these events depend on ssDNA forming during replication or transcription, Rad51-independent DSB repair may be S-phase dependent or more efficient in highly transcribed genes.
A striking feature of Rad51-independent gene conversion is the prevalence of discontinuous conversion tracts. These tracts reflect the extent of hDNA formation and repair of mismatches within hDNA (35, 39, 58). Tract discontinuities may form by several mechanisms, including discontinuous mismatch repair in a single stretch of hDNA, with repair involving opposite strands as templates or with alternating patches of repaired and unrepaired mismatches. In the latter case, unrepaired mismatches segregate in the next mitotic division and one daughter cell will display a discontinuous tract. Mismatches may escape repair if hDNA is partially unpaired (3). It is unlikely that discontinuities seen in the absence of Rad51 reflect multiple tracts from independent HR events because each event is extremely rare (≤10−5) and independent events are expected at <10−10. The increase in discontinuous tracts coupled with increased tract length in the rad51Δ mutant is reminiscent of tracts in XRCC3-defective mammalian cells (3). We proposed that XRCC3 does not actively suppress long conversion tracts but rather that XRCC3 stabilizes short hDNA intermediates and that in the absence of XRCC3, long hDNAs are selected because of their increased stability (3). The long discontinuous tracts in the absence of Rad51 may result from similarly unstable hDNA. However, this does not account for the very short tracts that arose in the absence of Rad51. It is possible that the very short- and long-tract products arise by distinct mechanisms. Interestingly, a similar bimodal tract length distribution was observed in Rad51C-defective mammalian cells (31).
Crossovers pose a significant risk of large-scale genetic change, including LOH of entire chromosome arms, deletions, inversions, and translocations (33), and crossovers are more frequently associated with long gene conversion tracts (12, 22, 55). Although the frequencies of crossovers associated with gene conversions were similar in the wild type and the rad51Δ mutant, only short (and long discontinuous) Ura+ tracts could be analyzed in the rad51Δ mutant. If it were practical to analyze long-tract Ura− products, the total crossover rate might be higher in the rad51Δ mutant than in the wild type. It appears that Rad51 contributes to genome stability by enhancing the efficiency of DSB repair by conservative gene conversion that produces relatively short, continuous conversion tracts. This minimizes localized LOH, as well as large-scale LOH and chromosome rearrangements associated with crossovers.
We thank Rosa Sterk, Jennifer Clikeman, Kimberly Spitz, and Yi-Chen Lo for expert technical assistance and James Haber and Justin Wray for helpful comments.
This research was supported by grant CA118357 (awarded to J.A.N. and Mary Ann Osley) and by the University of New Mexico Initiatives to Maximize Student Diversity program, funded by NIH NIGMS grant GM060201 (directed by M. Werner-Washburne).
Published ahead of print on 26 November 2007.
†Supplemental material for this article may be found at http://mcb.asm.org/.