Genetic Requirements for RAD51- and RAD54-Independent Break-Induced Replication Repair of a Chromosomal Double-Strand Break

Abstract

Broken chromosomes can be repaired by several homologous recombination mechanisms, including gene conversion and break-induced replication (BIR). In Saccharomyces cerevisiae, an HO endonuclease-induced double-strand break (DSB) is normally repaired by gene conversion. Previously, we have shown that in the absence of RAD52, repair is nearly absent and diploid cells lose the broken chromosome; however, in cells lacking RAD51, gene conversion is absent but cells can repair the DSB by BIR. We now report that gene conversion is also abolished when RAD54, RAD55, and RAD57 are deleted but BIR occurs, as with rad51Δ cells. DSB-induced gene conversion is not significantly affected when RAD50, RAD59, TID1 (RDH54), SRS2, or SGS1 is deleted. Various double mutations largely eliminate both gene conversion and BIR, including rad51Δ rad50Δ, rad51Δ rad59Δ, and rad54Δ tid1Δ. These results demonstrate that there is a RAD51- and RAD54-independent BIR pathway that requires RAD59, TID1, RAD50, and presumably MRE11 and XRS2. The similar genetic requirements for BIR and telomere maintenance in the absence of telomerase also suggest that these two processes proceed by similar mechanisms.

The repair of broken chromosomes by homologous recombination may occur in several ways (reviewed by Pâques and Haber [46]). If both ends of the broken molecule have homology to sequences on an unbroken chromosome that can serve as a template, repair may proceed by gene conversion. For example, in the G₁ stage of the cell cycle, a haploid cell or a diploid with a monosomic chromosome lacks an intact, homologous chromosome that it can use as a template for repair. However, if the centromere-proximal end of the double-strand break (DSB) has homology at or near one end to sequences elsewhere in the genome, a one-ended recombination event may still take place that will repair the broken end of the chromosome by appending a portion of a chromosome arm containing its telomere. A similar situation prevails in cells lacking the telomerase enzyme, which maintains chromosome termini. Here the degrading chromosome ends will only share homology with sequences at a single end of a broken chromosome, and repair again can occur only by a one-ended recombination mechanism. In both of these situations, an alternative mechanism of repair known as break-induced replication (BIR) may be able to restore a telomere to the broken chromosome and thus preserve its integrity.

BIR is a recombination-dependent DNA replication process that has been invoked to explain late DNA replication in phage T4 (36, 43), break-copy-recombination in phage λ (31, 44, 56), and origin-independent DNA replication in Escherichia coli (28, 29). In Saccharomyces cerevisiae, events consistent with BIR following creation of a DSB have been directly demonstrated in several ways. First, when HO endonuclease is used to cleave off the end of one chromosome in a wild-type haploid, repair can occur by BIR, in which sequences centromere proximal to an HO-induced DSB apparently invade a homologous sequence located on another chromosome arm (3). Replication from this site produced a nonreciprocal translocation in which a 30-kb terminal region of a different chromosome arm was found distal to the end produced by HO cleavage. Similar kinds of events have been demonstrated in transformation experiments, where the end of a linearized fragment apparently established a replication fork that could proceed several hundred kilobases to a chromosome end (41).

BIR has also been documented in a diploid experiencing a single HO endonuclease-induced DSB on one chromosome. In wild-type cells, this DSB is efficiently repaired by gene conversion using a homologous chromosome as a template. Gene conversion is abolished by a rad51Δ mutation (38), but surprisingly, in the absence of the Rad51p strand exchange protein, repair of the broken chromosome can still occur by a BIR mechanism that causes all markers distal to the site of the DSB to become homozygous (Fig. 1). Although BIR is RAD51 independent, it is RAD52 dependent (38). In the absence of RAD52, the broken chromosome is almost always lost, producing a 2N − 1 viable monosomic diploid.

FIG. 1.

Repair of a DSB in a diploid. HO endonuclease cleavage at MATa can be repaired by gene conversion (A) using the uncleavable MATα-inc allele on the homologous chromosome as the donor sequence. Gene conversion can also occur, rarely, using the interchromosomal HMLα and HMRa loci as donors (38, 40), but these would be cleaved again if HO is continually expressed. If gene conversion occurs in G₂ cells, crossing over may generate sectored colonies homozygous for THR4 and for thr4 (B). Colonies that are phenotypically identical to but genotypically different from those shown in panel B can occur if there are two independent DNA repair events, one of which is a gene conversion and the other of which is BIR (C). BIR yields Ade1⁺ Thr4⁻ cells (D). Failure to repair the DSB leads to chromosome loss and the formation of Ade1⁺ Thr4⁻ cells (E).

A similar relationship between repair and recombination genes has been found in the survival of strains lacking telomerase. Without telomerase, telomeres slowly degrade until, after many generations, most cells die; however, some survivors appear. Many of these have amplified the subtelomeric Y′ elements to all chromosome ends, whereas others have managed to amplify telomere sequences themselves (37, 62). The appearance of survivors is RAD52 dependent (37). Le et al. (32) found that the appearance of survivors occurred in the absence of RAD51. Their investigation suggested that there were in fact two pathways of telomere maintenance in the absence of telomerase, one of which required RAD51, RAD54, RAD55, and RAD57 and the other of which used RAD50, MRE11, and XRS2. Consequently, there were no survivors in a rad51Δ rad50Δ double mutant. Recent studies (6, 61, 62) have strongly supported this idea, showing that there are indeed two types of telomerase-independent recombination products and that one of the two types is eliminated in rad51Δ and the other in rad50Δ strains.

Although it is likely that the repair processes involved in telomere maintenance in the absence of telomerase are based on BIR, it is not possible to examine telomere repair events in detail to determine if they yield nonreciprocal translocations. Hence we have returned to the system where it was first demonstrated that the BIR could occur in the absence of RAD51 to repair a single DSB in the middle of a chromosome (38). We have examined the effects of deleting most of the other members of the RAD52 epistasis group. Although deletions of these genes cause sensitivity to ionizing radiation and to radiomimetic drugs, it is evident that they fall into several distinctive subgroups (reviewed by Pâques and Haber [46]). RAD52 stands alone in being required for essentially all homologous recombination events, although even without RAD52 there remains a crossover-associated pathway of recombination that is accompanied by chromosome loss (16, 38).

Deletion of RAD51 leads to the loss of gene conversions between chromosomal sites, but gene conversions between inverted repeats on plasmids still occur (20, 58). A rad51Δ strain is not impaired in another homologous recombinational repair process—single-strand annealing (SSA). Spontaneous heteroallelic recombination in S. cerevisiae is also reduced in rad52Δ strains but is only mildly affected by rad51Δ (49, 50). Insofar as they have been tested, deletions of RAD54, RAD55, and RAD57 resemble rad51Δ. In some assays, rad55Δ and rad57Δ strains are defective only at low temperatures and this defect can be suppressed by overexpressing RAD51 (17, 35). This observation and biochemical studies have led to the suggestion that Rad55p and Rad57p help load Rad51p onto DNA for recombination (60). For HO-induced MAT gene switching, RAD55 and RAD57 are required at any temperature, possibly because the recombination process is more demanding when the donor sequences are heterochromatic and are probably more difficult to invade (58).

A third set of proteins includes Mre11p, Rad50p, and Xrs2p, which form a complex (64). Loss of any of these three proteins has a similar effect on both homologous and nonhomologous recombination (4, 19, 21, 40). They are actually hyperrecombinational for spontaneous heteroallelic recombination and only mildly defective in the completion of both gene conversion and SSA.

Recently two other genes have been identified that fall into the RAD52 epistasis group, TID1 and RAD59. The Tid1 (or Rdh54) protein is homologous to Rad54p. A tid1Δ strain is defective in spontaneous mitotic interhomologue recombination between heteroalleles but appears to be unaffected in intrachromosomal or intersister chromatid recombination (27, 54). In some strains, the effect of deleting TID1 can be seen only when RAD54 has been deleted. The role of Tid1p in DSB-induced events has not been tested.

Finally, Rad59p has been shown to be a homologue of Rad52p (1). Overexpression of RAD52 partially complements the defects of rad59Δ, but the converse is not observed. Loss of Rad59p impairs SSA and reduces HO-induced gene conversions on plasmids (2, 22, 48, 57). In spontaneous recombination, rad59Δ has no strong phenotype by itself, but a rad51Δ rad59Δ double mutant is nearly as severely impaired in spontaneous heteroallelic recombination as rad52Δ. This has led to the suggestion that RAD59 functions in a RAD51-independent pathway (1). The role of Rad59p in BIR is unknown. In gene conversions induced by DSBs, rad59Δ has a relatively minor role when the lengths of homology flanking the DSB are several kilobases; however, as the length of homology decreases, Rad59p plays an increasingly important role, suggesting that it may help stabilize the initial encounter between the DSB end and the donor sequence (57).

In this paper we examine the roles of the RAD52 family of homologous recombination genes in gene conversion and BIR in a diploid with a single HO-induced DSB. We provide evidence that BIR may proceed by a RAD52-dependent pathway, independent of RAD51, RAD54, RAD55, and RAD57 but requiring RAD50, RAD59, and TID1.

MATERIALS AND METHODS

Experiments were carried out using two sets of isogenic diploids, closely related to each other. One set of diploids is isogenic to those described by Malkova et al. (38) and is the result of crossing derivatives of haploid strains EI515 and AM133 (Table 1). A second set of diploids was constructed by crossing derivatives of EI515 with those of LS23, which is a segregant of a cross of AM133 with AM9 (MATa ade1-100 his4-519 ura3-52 leu2-3,112).

TABLE 1.

Strains

Strain	Genotype
Haploid strains
EI515	MATaTHR4 hmlΔ::ADE1 hmrΔ::ADE1 ade1 ura3-52 leu2-3,112 lys5 GAL3::HO
AM133	MATα-inc leu2-3,112 thr4 met13 trp1 ura3-52 ade1 HMLα HMRa
LS23	MATα-inc his4 leu2-3,112 thr4 met13 trp1 ura3-52 ade1 HMLα HMRa
Diploid strains
YLS100	EI515 × LS23
YLS101	EI515 rad51Δ × LS23 rad51Δ
YLS102	EI515 rad54Δ × LS23 rad54Δ
YLS103	EI515 rad55Δ × LS23 rad55Δ
YLS104	EI515 rad57Δ × LS23 rad57Δ
YLS105	EI515 rad1Δ × LS23 rad1Δ
YLS106	EI515 rad1Δ rad51Δ × LS23 rad1Δ rad51Δ
JZ605	EI515 sgs1Δ × LS23 sgs1Δ
JZ605	EI515 sgs1Δ rad51Δ × LS23 sgs1Δ rad51Δ
MLN120	EI515 rad51Δ tid1Δ × AM133 rad51Δ tid1Δ
MLN123	EI515 rad51Δ × AM133 rad51Δ
MLN126	EI515 tid1Δ × AM133 tid1Δ
MLN127	EI515 rad54Δ tid1Δ × AM133 rad54Δ tid1Δ
MLN128	MN130 rad51Δ rad59Δ × AM482 rad51Δ rad59Δ
MLN154	AM483 rad51Δ rad50Δ × AM484 rad51Δ rad50Δ

Deletions of the RAD genes were introduced into strain EI515, AM133, or LS23 by a one-step gene disruption method (5) using the plasmid or PCR fragment listed below. Details of the deletions of RAD50, RAD54, RAD55, RAD57, RAD59, and TID1 are available on request. Linearized DNA fragments derived from the following plasmids or PCR amplifications were used to disrupt RAD genes: pJH683 (rad51Δ::URA3) or pJH1079 (rad51Δ::LEU2); pJH573-(pXRAD) (rad54Δ::LEU2), a gift of L. Symington (Columbia University); pSTL11 (rad55Δ::LEU2) (35); pSM57 (rad57Δ::LEU2), a gift of David Schild (University of California, Berkeley); pL962 (rad1Δ::LEU2), a gift of Ralph Keil (Pennsylvania State University Medical School, Hershey). pJH1340 (sgs1Δ::URA3) was disrupted by F. Pâques. tid1Δ::URA3 was disrupted as previously described for tid1Δ::HIS3 (27), and rad59Δ::KAN was created in strain JKM179 by Qijun Chen and Carol Greider.

All rad deletion strains were checked for UV or methyl methanesulfonate (MMS) sensitivity and were verified by Southern blot analysis. When double mutants were constructed with rad51Δ by transformation, the two haploid parents were first subjected to deletion of the second gene and then RAD51 was deleted. In other cases, the double mutants were obtained from meiotic segregants of appropriate crosses. The rad50Δ rad51Δ double mutant (MLN154) was obtained by a cross of two segregants (AM484 and AM483) from a cross of a rad50Δ derivative of EI515 and a rad51Δ derivative of AM133. The rad51Δ rad59Δ diploid MLN128 was obtained from a cross of two segregants (AM482 and MN130) from crosses between rad59Δ derivatives of strain JKM111, a MATα strain otherwise identical to EI515 (38, 40), and a rad51Δ derivative of AM133.

Media and growth conditions.

Rich medium (yeast extract-peptone-dextrose [YEP-dextrose]) and synthetic complete medium with bases and amino acids omitted as specified were used as described previously (25). YEP-glycerol and YEP-galactose (YEP-Gal) consisted of 1% yeast extract–2% Bacto Peptone medium supplemented with 3% (vol/vol) glycerin and 2% (wt/vol) galactose, respectively. YEP-dextrose medium containing 0.015% (vol/vol) MMS was used to assess Rad⁻ phenotypes. Cultures were incubated at 30°C.

Analysis of DNA repair.

Logarithmically growing cells grown in YEP-glycerol were plated on YEP-Gal and grown into colonies. The colonies were then replica plated onto nutritional dropout media. Colonies containing Ade⁺ Thr⁻ sectors against an Ade⁻ Thr⁻ background were counted in two different ways during the course of this work. In experiments shown in Table 2, group B and C Ade⁺ Thr⁻ sectors smaller than one-fourth of the total colony were not included in this category but were classified as Ade⁻ Thr⁻ colonies. In other experiments, such colonies were included. There is thus a quantitative but not qualitative difference in the number of sectored colonies enumerated.

TABLE 2.

Gene conversion, BIR, and chromosome loss following an HO-induced DSB^a

Strain	Temp tested (°C)	No. tested	% Event for phenotype of colonies
			Ade+ Thr+ (GC)	Ade+ Thr+ Ade+ Thr− (GC + CO, GC + BIR)	Ade+ Thr− (BIR)	Ade+ Thr− Ade− Thr− (BIR + chrom. loss)	Ade− Thr− (chrom. loss)	Ade+ Thr+ Ade− Thr− (GC + chrom. loss)
Group A
Wild-type YLS100	30	638	70.6	28.0	1.0	0.3	0	0.1
	18	314	87.0	12.0	1.0	0	0	0
rad51Δ YLS101	30	706	2.0	0	13.0	70.0	15.0	0
	18	205	3.0	0	20.0	42.0	35.0	0
rad54Δ YLS102	30	383	1.0	0	14.0	30.0	55.0	0
	18	97	2.0	0	3.0	66.0	29.0	0
rad55Δ YLS103	30	78	55.0	10.0	3.0	10.0	5.0	17.0
	18	119	3.0	0	15.0	55.0	27.0	0
rad57Δ YLS104	30	764	52.0	13.0	4.0	8.0	8.0	15.0
	18	84	8.0	0	13.0	51.0	27.0	0
rad52Δ EI524b	30				3.0		97.0
Group B
sgs1Δ JZ604	30	155	90.0	3.9	4.1		2.0
sgs1Δ rad51Δ JZ605	30	187	5.9		34.2	26.7	33.2
Group C
rad1Δ YLS105	30	977	94.7	4.5	0.8
rad1Δ rad51Δ YLS106	30	641	2.0		27.9	21.9	45.2

^aDiploid cells illustrated in Fig. 1 were plated on YEP-Gal at the temperature indicated to induce expression of HO endonuclease, and colonies were then analyzed by replica plating to assess the retention of the ADE1 and THR4 alleles marking the left and right arms of chromosome III, respectively, as shown in Fig. 1. The types of events associated with each phenotypic class are also shown. GC, gene conversion. GC + CO, a sectored colony arising by reciprocal crossing over in G₂ cells. GC + BIR, a sectored colony showing both gene conversion and BIR. This category cannot be distinguished from GC + CO without further analysis. chrom. loss, chromosome loss. Sectored colonies with both gene conversion and loss (GC + chrom. loss) or BIR + chromosome loss (BIR + chrom. loss) are also indicated. 

^bData for rad52Δ diploid EI524 are taken from Malkova et al. (38). 

Induction of recombination.

The induction of DSBs by HO and the analysis of colonies were carried out as described previously (38). Cells were grown overnight in 50 ml of YEP-glycerol to a density of 1 × 10⁷ to 5 × 10⁷ cells per ml. Appropriate dilutions of cells were plated on YEP-Gal, grown to colonies, and analyzed.

RESULTS

rad54Δ, rad55Δ, and rad57Δ resemble rad51Δ, preventing gene conversions but allowing BIR.

The ability of diploids homozygous for deletion of rad54Δ, rad55Δ, or rad57Δ to repair an HO-induced DSB was compared with that of wild-type strains. As illustrated in Fig. 1, HO endonuclease will cleave the MATa locus but not the MATα-inc allele, so that a single broken chromosome is created. If there is no repair, this broken chromosome will be lost, creating an Ade1⁻ Thr4⁻ colony that is 2N − 1 monosomic for chromosome III (Fig. 1E). If repair occurs by BIR, then the cells will be Ade1⁺ but Thr4⁻ (Fig. 1D). If the DSB is repaired by gene conversion without crossing over, the colony will be Ade1⁺ and Thr4⁺ (Fig. 1A). Crossing over of a cell in the G₂ stage of the cell cycle may also produce colonies fully Ade1⁺ but sectored for Thr4⁺ and Thr4⁻ (Fig. 1B), but similar colonies could arise if there were two independent repair events, one leading to gene conversion and one to BIR (Fig. 1C). Some colonies are expected to be sectored because recombination was initiated in asynchronous cells, so that G₂ cells will have two genomes that can be repaired independently. Furthermore, in the case of repair-defective mutant cells, sectored colonies may also arise if a broken chromosome is replicated and segregated, so that repair occurs only in a later generation (33, 52, 63).

As shown in Table 2, rad51Δ cells gave rise to sectored colonies displaying a mixture of Ade⁻ Thr⁻ and Ade⁺ Thr⁻ sectors that, as researchers previously demonstrated, corresponded to chromosome loss and BIR events, respectively (38). In nearly all cases, the mating type of the colonies arising in rad51Δ strains was changed from nonmating (MATa/MATα-inc) to α-mating, either hemizygous or homozygous for MATα-inc. Indeed, this is the predominant type of repair seen for rad51Δ strains (Table 2). These results are very different from what is found in cells lacking RAD52, which produce almost exclusively Ade⁻ Thr⁻ colonies, indicative of chromosome loss (38). The small proportion of Ade⁺ Thr⁻ cells in rad52Δ strains proved to result from nonreciprocal crossover events associated with a loss of the other chromosome, different from what is seen in rad51Δ cells (38). We note that in the absence of HO induction, more than 99% remain Ade⁺ Thr⁺ and that there are occasional chromosome loss events that do not favor the MATa-containing chromosome compared to the MATα-inc chromosome (data not shown).

In rad51Δ diploids and in the rad54Δ, rad55Δ, and rad57Δ diploids discussed below, there were a small number of α-mating Ade⁺ Thr⁺ colonies that could be the result of authentic Rad51p-independent gene conversions (Table 2). These might occur either by a conventional gene conversion pathway or by a combination of two RAD51-independent pathways, an interrupted BIR event that would copy across MATα-inc but would then dissociate, followed by SSA (26). It is also possible that these represent large deletions of the HO-cleaved MATa locus, so that only MATα-inc on the other chromosome is expressed. This seems likely, because there were an equal number of nonmating (but no longer HO-cleavable) Ade⁺ Thr⁺ cells appearing to represent nonhomologous end-joinings that remove the HO cleavage site of MATa without deleting the MATa1 gene (40). In any case, these represent a very small proportion of recombinants in rad51Δ.

Diploids homozygous for rad54Δ behaved nearly identically to rad51Δ (Table 2). As with rad51Δ, the rad54Δ diploids produced mostly colonies that had either completely lost the broken chromosome or repaired the break by BIR (Ade⁺ Thr⁻). At 30°C the rad54Δ strain produced fewer sectored colonies that had undergone both BIR and chromosome loss events, and there were more colonies with only chromosome loss than found with rad51Δ. The small number of gene conversions in rad54Δ strains is consistent with the finding of rare gene conversion repair of HO-induced DSBs during MAT switching (53). Thus BIR is RAD54 independent as well as RAD51 independent in a situation where efficient gene conversion requires both genes.

Diploids homozygous for rad55Δ and rad57Δ were tested at both 18 and 30°C, because in some assays, such as MMS treatment or gamma irradiation, the repair defects of these mutants are seen only at lower temperatures (17, 35). Indeed, both of these mutants showed a striking temperature dependence in terms of the types of repair that were obtained. At 30°C, both mutants were nearly wild type in their outcomes, as more than half of the cells plated gave rise to gene conversion (Ade⁺ Thr⁺) events (Table 2). However, in a significant number of cases, the cell that was plated, which may have been in G₂, gave rise to a mixed colony containing one half in which gene conversion was successful and the other half in which loss of the broken chromosome had apparently occurred. Thus, the two mutants were not fully wild type at 30°C. In contrast, at 18°C, rad55Δ and rad57Δ strongly resembled rad51Δ and rad54Δ (Table 2).

RAD51-independent DSB repair mostly depends on RAD50.

A recent study showed that the maintenance of telomeres in the absence of the telomerase TLC1 gene could occur in the absence of RAD51, RAD54, RAD55, and RAD57, as well as without RAD50, MRE11, or XRS2 (32). However, survivors that could maintain telomeres without telomerase were prevented both in a rad52Δ strain and in a rad51Δ rad50Δ double mutant. By itself, a rad50Δ mutation behaved very similarly to the wild-type diploid (Table 3). More than 80% of the repair events were gene conversions, although there were a small number of instances of chromosome loss to produce Ade⁻ Thr⁻ sectors of colonies. When we compared a rad51Δ rad50Δ double mutant isogenic with the rad51Δ strain discussed above, we found that BIR was severely reduced and the great majority of colonies showed only chromosome loss. The Ade⁺ Thr⁻ colonies, which we presume to arise by BIR, were reduced from 85% of colonies in rad51Δ to 20% in rad51Δ rad50Δ (Table 3). In contrast to rad52Δ diploids, where BIR was eliminated (38), 20% of the colonies derived from the HO-induced rad51Δ rad50Δ diploid still were Ade⁺ Thr⁻. We confirmed that all nine out of nine colonies tested were still heterozygous for the ADE1 gene inserted on the left arm of chromosome III; hence these appear to be BIR events and not a nonreciprocal exchange event associated with chromosome loss that is seen in a small proportion of rad52Δ cells (38). This suggests that there is yet another pathway, which is RAD52 dependent but independent of RAD51 and RAD50, to generate these events (see Discussion).

TABLE 3.

Effects of recombination mutations on repair of HO-induced DSBs.

Strain	No. tested	% Event for phenotype of coloniesa
		Ade+ Thr+ (GC)	Ade+ Thr+ Ade+ Thr− (GC + CO, GC + BIR)	Ade+ Thr− (BIR)	Ade+ Thr− Ade− Thr− (BIR + chrom. loss)	Ade− Thr− (chrom. loss)	Ade+ Thr+ Ade− Thr− (GC + chrom. loss)	Ade− Thr+ (not analyzed)
Wild type MLN155	60	78.3	15.0	1.7		1.7		3.3
rad51Δ MLN123	732	1.2	0.5	27.9	57.1	12.4	0.7	0.2
rad59Δ MLN147	135	94.8		2.2	1.5	1.5
rad50Δ MLN113	490	83.4	13.9		2.3	0.2	0.2	0.2
rad51Δ rad50Δ MLN114	157	2.5		8.9	11.5	77.1
rad51Δ rad59Δ MLN128	169	3.6	1.2	4.7	5.9	84.0		0.6
rad54Δ YLS102	383	1.0	0	14.0	30.0	55.0	0
tid1Δ MLN126	1,059	70.4	19.6	1.3	0.1	0.5	2.0	0.2
rad54Δ tid1Δ MLN127	576	1.0		13.7	6.3	78.6	0.2	0.2
rad51Δ tid1Δ MLN120	1,110	2.2	1.0	15.9	36.1	44.1	0.7

^aColonies grown at 30°C were analyzed as described in Materials and Methods and shown in Table 2. All strains in this table were derived from crosses of isogenic derivatives of EI515 and AM133. Data for rad54Δ are the same as shown in Table 2. Types of events are those listed in Table 2. 

RAD51-independent DSB repair involves RAD59.

Recently Bai and Symington (1) showed that a rad51Δ rad59Δ strain was nearly as impaired in spontaneous heteroallelic recombination as rad52Δ and much more severely impaired than either single mutant. By itself, a rad59Δ derivative of our diploid yielded results very similar to those for the wild-type isogenic strain; nearly all repair events were gene conversions (Table 3). However, DSB repair in a rad51Δ rad59Δ double mutant is much more severely affected than in rad51Δ alone (Table 3). Only 11% of colonies were fully or partially Ade⁺ Thr⁻, with the rest being Ade⁻ Thr⁻, reflecting the failure to repair the broken chromosome in almost 85% of cases. Thus, RAD59 appears to play an important role in the RAD51-independent repair of DSBs by BIR. As with spontaneous recombination (1), however, the rad51Δ rad59Δ double mutant was not as severely blocked in recombination as a rad52Δ strain. Indeed, 12 of 14 of the remaining Ade⁺ Thr⁻ colonies recovered from the rad51Δ rad59Δ diploid were shown by Southern blot analysis to have undergone BIR events. The remaining two were likely to have undergone chromosome loss events in which there was a reversion of the ade1 mutation on chromosome I.

A TID1 deletion impairs BIR in rad51Δ and rad54Δ strains.

Recent studies of spontaneous recombination have shown that deletion of TID1 affects spontaneous interchromosomal gene conversions between heteroalleles but not intrachromosomal or sister chromatid interactions. In some cases the inhibition of interchromosomal recombination by deleting TID1 could be seen in comparison to wild-type strains (27); in other strains, the tid1Δ effect was evident only when RAD54 was also deleted (56). We therefore assessed how a tid1Δ diploid would carry out repair of an HO-induced DSB. We found that the absence of Tid1p had no significant effect on the repair of DSBs (Table 2). The great majority of events were gene conversions (Ade1⁺ Thr4⁺). Thus, in our strains Tid1p does not play a key role in interchromosomal DSB-induced gene conversions in an otherwise wild-type strain.

A rad54Δ tid1Δ strain was more severely blocked in DSB repair than was either single mutant alone (Table 3). The rad54Δ tid1Δ double mutant gave results virtually identical to those for rad51Δ rad50Δ, reducing but not completely eliminating Ade⁺ Thr⁻ recombinants. Southern blot analysis showed that 10 of 19 derivatives were consistent with BIR events. Four may be chromosome losses with a reversion of ade1, and four may be similar to rad52Δ-independent events in which there is a crossover chromosome associated with loss of the reciprocal partner (15, 38).

Interestingly, the rad51Δ tid1Δ double mutant showed intermediate behavior compared to rad51Δ alone or rad54Δ tid1Δ (Table 3). The number of colonies in which there was no repair (Ade⁻ Thr⁻) increased approximately fourfold, from 11% in rad51Δ to 44%, still much less than the almost 80% lack of repair in rad54Δ tid1Δ. One explanation for this result is that Rad54p is able to substitute in part for Tid1p in a RAD51-independent pathway of BIR.

SGS1 does not play a significant role in BIR.

The sgs1Δ mutation has been shown to increase spontaneous mitotic recombination both in ribosomal DNA and for sister chromatid exchange (45, 55, 67). The Sgs1 helicase has also been implicated as an important helicase during DNA replication (34). Hence it was of interest to know if an sgs1Δ mutation would affect HO-induced recombination. As shown in Table 2, group B, the absence of Sgs1p had no effect on predominantly gene conversion repair compared to what was found with wild-type cells. Likewise, a rad51Δ sgs1Δ strain was able to carry out BIR in a manner similar to that of rad51Δ alone. Hence it appears that RAD51-independent BIR does not require Sgs1p.

RAD1 is not required for RAD51-independent BIR.

The Rad1-Rad10 endonuclease has been shown to play an important role in recombination in which nonhomologous, 3′-ended single-stranded tails at the ends of a DSB must be excised before a 3′ end can be generated to prime new DNA synthesis (7, 10, 47). When there is nonhomology on both DNA ends, RAD1 is essential, but when there is nonhomology on only one side, as there is when HO cleaves MATa and MATα-inc is the donor, a rad1 deletion delays but allows the completion of the process (7, 18). We investigated whether RAD51-independent BIR also requires RAD1. In an otherwise wild-type strain, rad1Δ has no significant effect: most repair events are gene conversions, replacing MATa with MATα-inc. Perhaps more surprising, in a rad51Δ rad1Δ double mutant, there was also no significant effect relative to rad51Δ alone. Thus, even though at least 700 bp of the Ya region in MATa must be removed before a DNA end is homologous to the template chromosome and BIR can be initiated, Rad1p does not appear to be critical for this process (Table 2, group C). Previously, it has been shown that there is an alternative, though less efficient, mechanism for removing nonhomologous tails from DNA ends (6, 38), but the genes encoding elements of this process have yet to be identified.

DISCUSSION

From bacteriophages to eukaryotes, recombination-dependent DNA replication plays an important role in the repair of DSBs. Break-induced DNA replication was first invoked to account for recombination in phage λ (56) and to explain late DNA replication in phage T4 (36, 43). Recently, extensive DNA replication during recombination has been observed by George and Kreuzer (12) in phage T4-mediated recombination and by Kuzminov and Stahl (31) and Motamedi et al. (44) while studying recombination in phage λ. The work of Kogoma (29) extended this concept to account for recombination-dependent, origin-independent replication in E. coli. More recently, BIR has been proposed to explain how broken replication forks can be restarted (reviewed in references 14 and 30). BIR has also been proposed to explain how the ends of a linear transforming DNA segment can establish the complete replication of a copy of the E. coli genome (8), and a similar proposal has been made by Morrow et al. (41) to account for the duplication of an entire yeast chromosome that received a linearized fragment of transforming DNA.

BIR also appears to account for very long gene conversion tracts, apparently extending to the end of a chromosome, which have been observed in several studies of meiotic and mitotic recombination in S. cerevisiae (9, 24, 65). A demonstration of such repair following a DSB in mitotic cells was provided by Bosco and Haber (3), who used HO endonuclease to cleave one chromosome in such a way that it shared extensive homology with a homologous chromosome only centromere proximal to the DSB. This study also suggested that BIR could occur prior to normal S phase, so that both progeny of a cell suffering a broken chromosome in G₁ survived with the same BIR repair of the chromosome.

One key process in which BIR appears to be of critical importance in eukaryotic cells is in the repair of chromosomes with telomeres that are too short to retain their normally end-protected state. Lundblad and Blackburn (37) first showed that survivors in S. cerevisiae that lacked the telomerase component Est1p required the key recombination gene RAD52. A similar observation was made by McEachern and Blackburn (39) for telomerase-defective Kluyveromyces lactis. These observations were extended by Le et al. (32) to show that there seem to be two distinct pathways for survival in the absence of telomerase, one apparently requiring RAD51, RAD54, and RAD57 (and presumably RAD55) and the other involving RAD50, MRE11, and XRS2. Survivors after deletion of the telomerase TLC1 gene failed to arise in both rad52Δ strains and in rad51Δ rad50Δ double mutants. The idea that there may be two different pathways of telomere repair in the absence of telomerase was supported by the work of Teng and Zakian (62), who found two distinct types of recombination events at telomeres, both RAD52 dependent, and by more recent work showing that the two different types are eliminated in rad51Δ and rad50Δ mutants (61).

Because it is not possible to examine individual telomere repair events in detail, for example to determine if they yield nonreciprocal translocations, we have examined BIR repair of a single DSB in the middle of a chromosome. We showed that while an efficient gene conversion pathway requires RAD51, RAD54, RAD55, and RAD57, BIR can occur in the absence of any of these genes. Moreover, this RAD51- and RAD54-independent BIR process is dependent on RAD59, TID1, RAD50, and presumably MRE11 and XRS2.

In this paper we have focused on the RAD51- and RAD54-independent pathway of BIR that can be seen because the more efficient gene conversion pathway has been eliminated. But we have several indications that there is also a more efficient RAD51-dependent pathway of BIR that is usually masked by gene conversion. First, we have examined BIR events in RAD51 cells, but not at the same location where these studies have been conducted (3). In that study it appeared that BIR in the presence of RAD51 is more efficient than in rad51Δ cells, where repair may not occur for several generations after initiation of the DSB (38). Second, researchers have carried out preliminary experiments, using a diploid with the HO-induced DSB at the same site used in these studies but with only a small amount of homology distal to the cleavage site. Here, where gene conversion is nearly absent, it appears that RAD51-dependent BIR is indeed more efficient than rad51Δ events (M. L. Naylor, A. Malkova, and J. E. Haber, unpublished observations).

One suggestion that Rad54p may also be important for BIR comes from the fact that a rad51Δ tid1Δ strain is much less severely affected than rad51Δ rad59Δ, rad51Δ rad50Δ, or rad54Δ tid1Δ. The simplest interpretation of the fact that this particular double mutant is less severe is that Rad54p can substitute for Tid1p in Rad51p-independent BIR. Moreover, a rad54Δ mutant had a higher incidence of chromosome loss and fewer BIR events that did rad51Δ (Table 2). Since none of the double mutations that we have tested are as severe as rad52Δ, our findings might further suggest that there are still other redundancies to be revealed by further multiple mutant analysis.

The idea that there are two alternative BIR pathways is supported by studies of telomerase-independent telomere maintenance (6, 61). In the absence of RAD51, the great majority of survivors prove to be one of the two types (type II) described by Teng and Zakian (62), in which the TG_1–3 telomere ends themselves are lengthened by recombination. In the absence of RAD59, type I prevails, in which tandem arrays of Y′ subtelomeric genes separated by short stretches of telomere sequence are found at most chromosome ends. One might suggest that type I events involve BIR events in which a telomere end is resected into subtelomeric regions, such as X or Y′, and that there is extensive homology shared between that single-stranded region and other X or Y′ elements to promote BIR. In contrast, type II events appear to involve the invasion of TG_1–3 telomeric sequences into the same other TG_1–3 telomere sequences. This might occur by an intrachromosomal strand invasion, forming a T loop (13, 62, 66), and leading to rolling-circle replication.

One additional interesting finding is that rad55Δ and rad57Δ mutants are cold sensitive for the interchromosomal repair of an HO-induced DSB in a diploid, the same phenotype seen for X-ray or MMS sensitivity for these deletions (35). At 30°C or above, the rad55Δ and rad57Δ strains are resistant to ionizing radiation, whereas they are sensitive at 23°C. It is thought that Rad55p and Rad57p act as auxiliary factors to help load Rad51p onto single-stranded DNA created at DSB ends (11, 23, 60). However, not all HO-induced recombination events show cold sensitivity. In HO-induced MAT switching (where the donor sequences are heterochromatic), rad55Δ and rad57Δ strains are unable to complete recombination at 30 or even 34°C. Alternatively, HO-induced gene conversion of inverted repeated LacZ sequences on a plasmid proved to be independent of RAD51, RAD54, RAD55, and RAD57 at 23°C (20, 58). The repair event that we are studying here, an interchromosomal recombination gene conversion repair of a DSB, is the only HO-induced event that we have examined in which rad55Δ and rad57Δ mutants have the same cold sensitivity which they display for ionizing radiation and MMS sensitivity. This suggests to us that the system which we are using is in fact an excellent model system for the consequences of ionizing radiation.

How does BIR occur without Rad51p?

How BIR occurs in the absence of the strand invasion protein Rad51p remains a mystery. Based on our previous studies it seems very likely that the ends of the DSB are resected by a 5′-to-3′ exonuclease, producing long 3′-ended single-strand-DNA (ssDNA) tails (68). Special, accessible sites where Rad52p mediates strand annealing in the absence of Rad51p and Rad54p (43, 51, 59) between the ssDNA tail and a partially denatured region of the template chromosome would permit the necessary formation of heteroduplex DNA that could be used to prime new DNA replication. SSA in S. cerevisiae does not require Rad51p, Rad54p, Rad55p, or Rad57p (20) but does in fact require both Rad52p and Rad59p (57). However, in SSA the absence of Rad50p delays the event without preventing it (20). The role of Tid1p in this process has not been assessed. It is possible that these proteins play a role in the creation of a stable intermediate that can be converted into a replication fork. Support for the general idea that RAD51-independent BIR can occur only at particularly accessible sites comes from our recent discovery that BIR does not initiate anywhere (and hence does not retain genetic markers) in the first 10 kb centromere proximal to MAT (A. Malkova, L. Signon, C. B. Schaefer, M. L. Naylor, J. F. Theis, C. S. Newlon, and J. E. Haber, unpublished observations). Preferential sites of repair could also represent sites where the repair-replication fork could acquire processivity factors that would permit the repair-replication fork to move more than 130 kb down the chromosome.

In the experiments reported here, with the use of rad51Δ and other mutant cells, the efficiency of BIR is high enough that most colonies derived after plating single cells onto medium to induce HO cleavage can complete BIR; however, in the majority of cases, one observes one or more Ade⁺ Thr⁻ sectors against a background of Ade⁻ Thr⁻ cells that have lost the broken chromosome. In many colonies there are multiple Ade⁺ sectors, suggesting that DNA repair occurred several generations after the creation of the DSB. It is now well established that cells carrying a single DSB will arrest but then adapt and resume cell cycle progression even though they carry a broken chromosome (33, 52). The experiments of Malkova et al. (38) and those reported here demonstrate that this broken chromosome can then be repaired in approximately 10 to 20% of cell divisions. Whether BIR becomes an even more efficient event after cells have adapted, where some normal DNA damage checkpoint controls have apparently been turned off, is an important question that we are pursuing.

Finally, we note that BIR shares many common features with gene conversion. RAD51-dependent gene conversion initiated by a DSB is likely to involve both leading- and lagging-strand DNA polymerases (18), and scholars have suggested that the ways in which new DNA synthesis is initiated in gene conversion and in BIR are likely to be very similar or identical (18, 46). Gene conversion would occur if the second end of a DSB becomes engaged in the recombination event, whereas in the absence of a second end, the repair-replication fork would proceed to the chromosome end.