Aberrant Double-Strand Break Repair Resulting in Half Crossovers in Mutants Defective for Rad51 or the DNA Polymerase {delta} Complex

Homologous recombination is an error-free mechanism for therepair of DNA double-strand breaks (DSBs). Most DSB repair eventsoccur by gene conversion limiting loss of heterozygosity (LOH)for markers downstream of the site of repair and restrictingdeleterious chromosome rearrangements. DSBs with only one endavailable for repair undergo strand invasion into a homologousduplex DNA, followed by replication to the chromosome end (break-inducedreplication [BIR]), leading to LOH for all markers downstreamof the site of strand invasion. Using a transformation-basedassay system, we show that most of the apparent BIR events thatarise in diploid Saccharomyces cerevisiae rad51 ${Delta}$ mutants aredue to half crossovers instead of BIR. These events lead toextensive LOH because one arm of chromosome III is deleted.This outcome is also observed in pol32 ${Delta}$ and pol3-ct mutants,defective for components of the DNA polymerase ${delta}$ (Pol ${delta}$ ) complex.The half crossovers formed in Pol ${delta}$ complex mutants show evidenceof limited homology-dependent DNA synthesis and are partiallyMus81 dependent, suggesting that strand invasion occurs andthe stalled intermediate is subsequently cleaved. In contrastto rad51 ${Delta}$ mutants, the Pol ${delta}$ complex mutants are proficient forrepair of a 238-bp gap by gene conversion. Thus, the BIR defectobserved for rad51 mutants is due to strand invasion failure,whereas the Pol ${delta}$ complex mutants are proficient for strand invasionbut unable to complete extensive tracts of recombination-initiatedDNA synthesis.

INTRODUCTION

Top

INTRODUCTION

DNA double-strand breaks (DSBs) are potentially lethal lesionsthat can occur spontaneously during normal cell metabolism,by treatment of cells with DNA-damaging agents, or during programmedrecombination processes (54). There are two major pathways torepair DSBs: nonhomologous end joining (NHEJ) and homologousrecombination (HR). NHEJ involves the religation of the twoends of the broken chromosome and can occur with high fidelityor be accompanied by a gain or loss of nucleotides at the junction(9). Repair of two-ended DSBs by HR generally occurs by geneconversion resulting from a transfer of information from theintact donor duplex to the broken chromosome (Fig. 1). HR occurspreferentially during S and G₂ when a sister chromatid is availableto template repair (2, 19, 22). Sister-chromatid recombinationevents are genetically silent, whereas gene conversion betweennonsister chromatids associated with an exchange of flankingmarkers can result in extensive loss of heterozygosity (LOH)or chromosome rearrangements (3, 21). One-ended DSBs that ariseby replication fork collapse or by erosion of uncapped telomeresare thought to repair by strand invasion into homologous duplexDNA followed by replication to the end of the chromosome, aprocess referred to as break-induced replication (BIR) (35).BIR appears to be suppressed at two-ended breaks, presumablybecause it can lead to extensive LOH if it occurs between homologuesor to chromosome translocations when strand invasion initiateswithin dispersed repeated sequences (5, 28, 31, 50, 52, 55).

View larger version (11K):
[in this window]
[in a new window]

FIG. 1. Models for gene conversion and BIR. After formation of a DSB, the ends are resected to generate 3' single-strand DNA tails. One end undergoes Rad51-dependent strand invasion to prime DNA synthesis from the invading 3' end templated by the donor duplex. For gene conversion by the synthesis-dependent strand annealing model, the extended invading end is displaced and can anneal to the other side of the break; completion of repair requires DNA synthesis primed from the noninvading 3' end. For a one-ended break, or if the other side of the break lacks homology to the donor duplex, DNA synthesis proceeds to the end of the chromosome. Centromeres are shown as solid ovals and a heterozygous marker centromere distal to the site of repair as A/a.

The strand invasion step of BIR is assumed to be the same asthat for gene conversion based on the requirement for the sameHR proteins: Rad51, Rad52, Rad54, Rad55, and Rad57 (10). However,subsequent steps in BIR are less well defined. Recent studiesof the fate of the invading end during BIR in diploid strainswith polymorphic chromosome III homologues using a plasmid-basedassay have shown that following strand invasion, the invadingend is capable of dissociating from the initial homologous template.Following dissociation, the displaced end subsequently reinvadesinto the same or a different chromosome III homologue by a processtermed template switching (52). One of the interesting featuresof the template switching events is that they occur over a regionof about 10 kb downstream of the site of strand invasion anddo not extend over the entire left arm of chromosome III. Thereare a number of possible mechanisms that could account for thisapparent change in the processivity of BIR. First, it is possiblethat the strand invasion intermediate is cleaved by a structure-specificnuclease and once the invading strand is covalently joined toone of the template strands, the strand invasion process isirreversible. Recent studies of Schizosaccharomyces pombe haveshown an essential role for Mus81, a structure-specific nuclease,in resolution of sister chromatid recombination intermediatesduring repair of collapsed replication forks (48). Another possibilityis that there could be a switch between a translesion DNA polymeraseand a highly processive DNA polymerase during BIR. The translesionpolymerases in budding yeast, polymerase ${zeta}$ (Pol ${zeta}$ ) and Pol ${eta}$ , areencoded by REV3-REV7 and RAD30, respectively (34, 40, 43). Deletionof REV3 has been shown to increase the fidelity of DNA synthesisassociated with HR but has no effect on the overall frequencyof DSB-induced HR (16). Deletion of POL ${eta}$ in chicken DT40 cellsreduces the frequency of DSB-induced gene conversion, and humanPOL ${eta}$ has been shown to extend the invading 3' end of D-loopintermediates in vitro (23, 36). However, this same preferencefor Pol ${eta}$ is not found for Saccharomyces cerevisiae. Instead,DNA synthesis during meiotic and mitotic recombination appearsto be carried out by Pol ${delta}$ , one of the three nuclear replicativepolymerases, which normally functions with Pol ${alpha}$ in Okazaki fragmentsynthesis (13, 32, 33, 44). Pol ${varepsilon}$ is thought to be the primaryleading-strand polymerase (47), but in the absence of the Pol ${varepsilon}$ catalytic domain, Pol ${delta}$ is presumed to carry out leading-strandsynthesis (24). Recent studies by Lydeard et al. (30) have showna requirement for the lagging-strand polymerases, Pol ${delta}$ and Pol ${alpha}$ , to form the initial primer extension product during BIR, andPol ${varepsilon}$ is required to complete replication to the end of the chromosome.In contrast, repair of DSBs by gene conversion does not requirePol ${alpha}$ , and there appears to be functional redundancy betweenPol ${delta}$ and Pol ${varepsilon}$ (56).

To address the roles of Mus81, Pol ${delta}$ , and Pol ${eta}$ in BIR and inparticular template switching, we used the transformation-basedBIR assay with diploids with polymorphic chromosome III homologues.Because the transformation assay can only be used with strainswith viable mutations of replication factors, we used a nullallele of POL32, encoding a nonessential subunit of the Pol ${delta}$ complex (14), and a point mutation in the gene encoding theessential catalytic subunit, POL3. The pol3-ct allele resultsin a truncation removing the last four amino acids of the Pol3protein; the C-terminal region of Pol3 is implicated in interactionwith the other essential subunit of the Pol ${delta}$ complex, Pol31(15, 49). The interesting feature of the pol3-ct allele is thatit decreases the length of gene conversion tracts during mitoticand meiotic recombination, presumably by affecting the processivityof Pol ${delta}$ , but confers no apparent defect in normal DNA synthesis(32, 33). Because BIR requires more-extensive tracts of DNAsynthesis than gene conversion, we expected the pol3-ct mutantto exhibit a BIR defect. We found that in the absence of a fullyfunctional Pol ${delta}$ complex, chromosome fragment (CF) formationproceeds by a half-crossover mechanism associated with lossof the template chromosome, an event with potentially catastrophicconsequences (6, 57). This was also found to occur in rad51mutants, suggesting nonreciprocal translocations arise by failureto undergo strand invasion or because replication followingstrand invasion is inefficient. In contrast to rad51 mutants,the Pol ${delta}$ complex mutants are proficient for repair of a 238-bpgap by gene conversion and fully resistant to ionizing radiation,suggesting there is a unique requirement for Pol ${delta}$ to completeBIR. Consistent with studies of gene conversion in S. cerevisiae(33), we found no role for Pol ${eta}$ in BIR or the process of templateswitching.

MATERIALS AND METHODS

Top

MATERIALS AND METHODS

Media, growth conditions, and genetic methods. Rich medium (yeast extract-peptone-dextrose [YPD]), syntheticdropout and synthetic complete medium (SC) lacking the appropriateamino acids or nucleic acid bases, sporulation medium, and geneticmethods were as described previously (51). Transformation ofyeast cells was performed by the lithium acetate method (20).Mating tests were performed by patching independent Ura⁺ transformantsonto YPD plates, replica plating to lawns of MATa and MAT ${alpha}$ testerstrains to allow mating for 6 h, and then replica plating tosynthetic dropout selection plates.

Yeast strains. The diploid yeast strain with chromosome III markers was madeby crossing haploid strains MW16 (MAT ${alpha}$ trp1 arg4 tyr7 ade6 ura3)and MD249 (MATa leu2-bst ade6 ura3 cha1::hphMX6) (38, 58). Thepol3-ct diploid used was previously described and is isogenicto diploids derived from MW16 and MD249 (32). mus81 ${Delta}$ , pol32 ${Delta}$ ,rad30 ${Delta}$ , and rad51 ${Delta}$ derivatives were constructed by deleting fromthe haploid MW16 and MD249 strains either MUS81, POL32, RAD30,or RAD51 using a PCR fragment containing homologous 5'- and3'-flanking sequences from the BY4742 (MAT ${alpha}$ his3 ${Delta}$ 1 leu2 ${Delta}$ 0 lys2 ${Delta}$ 0ura3 ${Delta}$ 0) deletion strain collection containing either mus81::KanMX4,pol32::KanMX4, rad30::KanMX4, or rad51::KanMX4, resulting inthe integration of the KanMX4 marker and the loss of the wild-typeallele. The mus81 ${Delta}$ pol3-ct double mutant was made by replacementof the MUS81 locus with the KanMX4 marker in the haploid strainsLMP2 and LMP6 containing the pol3-ct allele. The haploid strainswere then crossed, and diploids were selected for on SC mediumlacking Leu and Trp. Strains LSY205 (MATa suc2-437 lys2-802)and LSY206 (MAT ${alpha}$ lys2-802) were used for mating-type tests, orthe MAT locus was scored by PCR (17).

BIR assay. The chromosome fragmentation vector (CFV), CFV/D8B-tg (CFV1),containing a 5.2-kb insert from the left arm of chromosome III(SGD coordinates 96821 to 102096), and CFV2, containing a 3.3-kbBglII fragment from chromosome III (coordinates 103535 to 106833),were described previously (10, 52). A 0.1- to 1-µg amountof the CFV, digested with SnaBI, was used to transform competentyeast cells, selecting for Ura⁺ transformants. Larger amountsof DNA were necessary to recover transformants from the rad51mutant. Because some Ura⁺ transformants are due to NHEJ of thevector, resulting in an unstable plasmid, Ura⁺ transformantswere struck onto nonselective YPD medium and then replica platedto SC medium lacking Ura to determine the stability of the Ura⁺phenotype. The frequency of BIR presented in Fig. 2 is the numberof mitotically stable Ura⁺ transformants per microgram of linearizedDNA transformed divided by the number of Ura⁺ transformantsper microgram of circular plasmid DNA transformed. The meanBIR frequencies (with standard deviations) presented are fromat least three independent transformations of each strain. Statisticalanalyses were performed using an unpaired t test. DNA analysisof stable Ura⁺ transformants to score restriction site markerswas as described previously (52). Statistical significance forthe loss of chromosome III during CF formation was performedusing a chi-square test.

View larger version (15K):
[in this window]
[in a new window]

FIG. 2. Schematic representation of the chromosome fragmentation assay and frequency of BIR. (A) Following transformation into competent yeast cells, the linear CFV undergoes de novo telomere addition to the TG tract at one end of the vector, and the other end invades the endogenous chromosomal locus duplicating sequences from the region of homology between the vector and the native chromosome to the telomere. The region of homology is shown as a white rectangle; centromeres are shown as solid circles and telomeres as arrowheads. (B) Frequency of BIR. Each frequency is measured as the number of stable Ura⁺ transformants per microgram of cut CFV1 divided by the number of Ura⁺ transformants per microgram with uncut DNA transformed. wt, wild type.

Plasmid DNA gap repair assay. The nonreplicating gap repair plasmid, pSB101, was describedpreviously (4). Plasmid pSB101 was digested with BspEI and EcoNI,and the linear DNA was gel purified. Transformation was performedby the lithium acetate transformation method with 100 ng ofgapped plasmid (gap repair substrate) or 100 ng of uncut replicatingplasmid pSB110 (transformation efficiency control). The transformedcells were diluted and plated on SC medium lacking Ura. Thecolonies were counted after incubation at 30°C for 2 to3 days. The frequency of gap repair is the number of Ura⁺ transformantsper microgram of linearized DNA transformed divided by the numberof Ura⁺ transformants per microgram of circular plasmid DNAtransformed. Statistical significance was calculated using anunpaired t test.

${gamma}$ -irradiation sensitivity tests. Cells were grown in liquid YPD medium at 30°C to exponentialphase. The cultures were serially diluted, and aliquots of eachdilution were spotted onto YPD plates. The plates were irradiatedin a Gammacell-220 irradiator containing ⁶⁰Co and incubatedat 30°C for 3 days.

RESULTS

Top

RESULTS

The polymerase ${delta}$ complex is required for efficient BIR. To evaluate the role of the Pol ${delta}$ complex, Pol ${eta}$ , and Mus81 inBIR, we used the transformation-based assay in conjunction withdiploid strains containing restriction site polymorphisms onthe left arm of chromosome III and homozygous mutations of therelevant genes (Fig. 2A) (10, 52). The CFV contains the URA3selectable marker, SUP11, CEN4, an autonomously replicatingsequence, a tract of (G_1-3T)_n to provide a site for telomereaddition, and a unique DNA segment from the left arm of chromosomeIII. After transformation of yeast with the linearized CFV,a CF is generated by de novo telomere addition at one end ofthe vector and recombination-dependent replication of 97 kbof chromosomal sequences at the other end of the vector (Fig.2A). The rad51 ${Delta}$ diploid was used as a control for a mutant defectivefor BIR (10). The number of transformants with a stable Ura⁺phenotype after being challenged with nonselective growth conditionswas normalized for transformation efficiency of each strain,measured by transformation with uncut CFV. In the case of therad51 ${Delta}$ strain, 25% of the stable Ura⁺ colonies were due to reversionor conversion of the chromosomal ura3 gene, and the BIR frequencywas adjusted accordingly. The pol32 ${Delta}$ diploid exhibited a sixfold-reducedfrequency of BIR compared with that of the wild type (P = 0.01),but there was not a complete dependence on POL32 for CF formation(Fig. 2B). The frequency of BIR was reduced by 20-fold in therad51 ${Delta}$ diploid compared with that in the wild-type diploid (P= 0.007) and was threefold lower than that of the pol32 ${Delta}$ mutant(P = 0.002), consistent with results in previous studies (30).The frequency of BIR was reduced by only twofold in the pol3-ctdiploid (P = 0.02). We had expected a greater reduction in BIRbased on previous studies showing reduced meiotic and mitoticconversion tract lengths in this mutant (32, 33); however, asdescribed below, the events recovered in the pol3-ct strainwere formed by a mechanism different from CF formation in thewild-type diploid. There was no significant decrease in thefrequency of BIR in the rad30 ${Delta}$ and mus81 ${Delta}$ mutants.

In the absence of a fully functional Pol ${delta}$ complex or RAD51, chromosome fragments form by half crossovers. To determine whether the Pol ${delta}$ mutants undergo DNA synthesisduring CF formation, we made use of the restriction site polymorphismspresent on the chromosome III homologues (Fig. 3). These allowus to track which chromosomal template was used for CF formation,giving insight into how replication progresses during BIR. Wepreviously showed that in wild-type cells, the invading strandthat was extended by DNA synthesis could dissociate and invadeinto the other parental homologue, creating a chimeric CF (52).

View larger version (18K):
[in this window]
[in a new window]

FIG. 3. CF formation proceeds by a half crossover in polymerase ${delta}$ and rad51 mutants. A schematic representation of the restriction site polymorphisms carried on the chromosome III homologues and the classes of CFs recovered is shown. RCF refers to the events formed by template switching to form chimeric CFs. Coordinates of the markers scored are 91358 (Bs, BstEII), 96182 (Sp, SpeI), 96821 (G, BglII), 99611 (X, XhoI), 102096 (G, BglII), and 103535 (G, BglII). The numbers of each class of event in the wild-type and mutant strains are shown; the numbers on the second line, in parentheses, are percentages of each class.

Total genomic DNA from stable Ura⁺ transformants was analyzedby restriction endonuclease digestion and Southern blottingto determine the composition of the CF. We then selected forloss of the CF by plating cells on medium containing 5-fluorooroticacid (which selects against Ura⁺ cells) and repeated the genomicDNA analysis to confirm that novel restriction fragments werelinked to the CF. In the wild-type diploid, 32 of 161 Ura⁺ transformantsshowed template switching, referred to as recombinant CFs (RCFs)(Fig. 3) (52). Of the 129 CFs that had not undergone templateswitching, 5 were associated with the loss of a chromosome IIIhomologue and the sequences acquired by the CF were derivedfrom the lost copy of chromosome III. These events are mostsimply explained by a crossover between the linear fragmentand the chromosome III homologue invaded, similar to the mechanismproposed for gene targeting except that only one crossover isrequired. This type of recombination event maintains disomyfor sequence distal to the site of strand invasion but resultsin monosomy for most of chromosome III. These events will bereferred to as half crossovers to distinguish them from BIRevents, which maintain both copies of chromosome III in additionto the CF. It should be noted, however, that some events thatmaintain both copies of chromosome III could be formed by ahalf crossover in G₂ followed by segregation of two intact copiesof chromosome III and the CF to the same daughter cell (seeDiscussion). In five of the samples, one or more chromosomalsites had become homozygous, referred to as "other".

In both the pol32 ${Delta}$ and pol3-ct diploids, we found that formationof CFs was frequently by a half-crossover mechanism (Fig. 3).Of 79 CFs analyzed from the pol32 ${Delta}$ diploid, 41 (51%) were associatedwith loss of one of the chromosome III homologues. Of these,23 acquired sequences from homologue P1; the other 18 acquiredsequences from homologue P2. The same pattern was observed forthe pol3-ct diploid. Of 85 CFs analyzed, 39 (46%) were associatedwith chromosome loss. Of these, 23 CFs were formed by acquisitionof sequence from chromosome III homologue P1, while 16 acquiredsequence from P2 (Fig. 3). Like the Pol ${delta}$ complex mutants, therad51 ${Delta}$ diploids also showed elevated rates of chromosome lossassociated with CF formation. In this case, of 41 CFs analyzed31 (75%) were associated with chromosome loss. Chromosome lossin the rad51 ${Delta}$ , pol3-ct, and pol32 ${Delta}$ mutant diploids was significantlyhigher than that in the wild-type diploid (P < 0.0001 forall mutants tested). This suggests the CFs formed in the mutantsare not the result of a true BIR event and instead form by ahalf-crossover mechanism.

Diploids are heterozygous for the mating type (MAT) locus onthe right arm of chromosome III and exhibit a nonmating phenotype.Chromosome III loss, or homozygosis of the MAT locus, resultsin expression of only MATa or MAT ${alpha}$ information and can be detectedby the ability of cells to mate with haploid test strains. Toconfirm loss of one copy of chromosome III in the rad51 ${Delta}$ , pol32 ${Delta}$ ,and pol3-ct diploids transformed with CFV1, mating tests wereperformed on independent Ura⁺ transformants. For all three mutants,>50% of the transformants exhibited a mating phenotype, comparedwith 13% for the wild-type diploid (data not shown). To ensurethat chromosome loss was not due to the transformation process,we also tested the mating phenotype of Ura⁺ transformants derivedfrom uncut CFV1 in the pol32 ${Delta}$ and pol3-ct diploids. Only 1/93pol3-ct transformants and 1/64 pol32 ${Delta}$ transformants were matingproficient. Thus, chromosome loss is associated with CF formationand is not induced by transformation.

Analysis of CFs formed in the rad30 ${Delta}$ (Pol ${eta}$ ) and mus81 ${Delta}$ diploidsrevealed that most were formed by BIR (Fig. 3). Of 37 CFs analyzedfrom the rad30 ${Delta}$ diploid, 8 RCFs were identified; of these, threewere associated with loss of one of the chromosome III homologues.Three other events were associated with conversion of chromosomalmarkers, and one of these also had an RCF. Because the frequencyof template switching events and the distribution of eventswere not significantly different from those for the wild type,we conclude Pol ${eta}$ is not responsible for the change in processivityduring BIR. Similarly, the frequency and distribution of RCFsrecovered from the mus81 ${Delta}$ diploid were not significantly differentfrom those for the wild-type strain (Fig. 3 and data not shown).

Rare template switching in pol32 ${Delta}$ and pol3-ct diploids. The decrease in the frequency of BIR and formation of half crossoversin the rad51 ${Delta}$ , pol3-ct, and pol32 ${Delta}$ mutant diploids suggests strandinvasion or that DNA synthesis primed from the invading endis defective. While strand invasion is known to be severelyimpaired in the rad51 ${Delta}$ diploid, we assume strand invasion occursin the pol3-ct and pol32 ${Delta}$ diploids but DNA synthesis is attenuated.If a short tract of DNA was synthesized from the invading endand displacement and reinvasion occurred, we would expect toobserve low-frequency template switching in the Pol ${delta}$ -defectivemutants. Of the 79 CFs screened from the pol32 ${Delta}$ mutant, 4 wereformed by template switching and all were associated with lossof the chromosome III homologue that was used for the second-strandinvasion (Fig. 3). For the RCF shown in Fig. 4, the invadingend would have to be extended by at least 500 nucleotides tocreate the novel restriction fragment diagnostic of templateswitching. Similarly, five RCFs were recovered from the pol3-ctmutant, and three of them were associated with loss of one ofthe chromosome III homologues. For one of the events observedin the pol3-ct mutant, at least 3.4 kb of DNA would need tobe synthesized to produce the RCF. Therefore, short tracts ofDNA synthesis occur in the pol32 ${Delta}$ and pol3-ct mutants prior tocrossover formation. We also noted several events with chromosomerearrangements in the Pol ${delta}$ complex mutants. This class includedevents with conversion of one of the chromosomal markers andone case of a reciprocal exchange between markers in the pol32 ${Delta}$ mutant. One event recovered from the pol3-ct mutant could havebeen due to a template switch event following a reciprocal exchangebetween the BstEII and SpeI markers or could have been due tocopying of P2 followed by a reciprocal exchange between thehomologues. Analysis of 41 CFs from the rad51 ${Delta}$ diploid failedto reveal any evidence of template switching or alterationsof the chromosome III markers consistent with strand invasionfailure.

View larger version (38K):
[in this window]
[in a new window]

FIG. 4. Physical analysis of CFs from the pol32 ${Delta}$ mutant, showing template switching prior to half-crossover formation. (A) Schematic representation of the region of chromosome III showing the heterozygous BstEII (Bs) and SpeI (Sp) sites and the expected restriction fragments for P1, P2, and one of the two possible RCFs. Note that CFs that copy from only P1 or P2 will have the same-size restriction fragments as P1 or P2, respectively. (B) Southern blot showing representative 2N + CF, (2N–1) + CF, and (2N–1) + RCF classes in the left panel; the panel on the right shows the DNA analysis after selection for loss of the CF by plating cells on medium containing 5-fluoroorotic acid.

MUS81-dependent half crossovers. Template switching in the pol3-ct and pol32 ${Delta}$ mutants providesevidence that strand invasion and limited DNA synthesis occur,but due to the defect in DNA synthesis, we imagine the stalledintermediate is then cleaved by an endonuclease to generatea crossover (Fig. 5). The Mus81-Mms4 heterodimeric endonucleasecleaves a variety of branched DNA structures in vitro, includingstrand invasion intermediates (45). We constructed a diploidmus81 ${Delta}$ pol3-ct double mutant to determine whether the half crossoversare the result of Mus81 cleavage of the strand invasion intermediate.Although the frequency of Ura⁺ transformants for the mus81 ${Delta}$ pol3-ctdouble mutant was not significantly reduced from that for thepol3-ct single mutant, there was a significant reduction inthe number of chromosome loss events (P = 0.0002) (Table 1).

View larger version (11K):
[in this window]
[in a new window]

FIG. 5. Model for CF formation in the Pol ${delta}$ mutants. Following strand invasion of wild-type (RAD51) cells, the 3' invading end is extended by DNA synthesis. The invading end can dissociate and undergo strand invasion and extension from the 3' end again (template switching). The strand invasion intermediate is cleaved and one strand from the CFV ligated to the nicked strand of the donor duplex. This transition prevents dissociation (and template switching) of the invading end and is permissive for assembly of the other replication components to initiate processive replication to the chromosome end. In this case, both chromosome III homologues remain intact. Following strand invasion in Pol ${delta}$ mutant diploids, the D-loop is cleaved as described above, but in the absence of extensive synthesis from the invading strand and/or because initiation of lagging-strand synthesis is delayed, a second cleavage event occurs, generating a half crossover. This results in the formation of a CF and broken chromosome III homologue, which is subsequently lost.

View this table:
[in this window]
[in a new window]

TABLE 1. Chromosome loss in pol3-ct and mus81 ${Delta}$ diploids

The Pol ${delta}$ complex mutants are proficient for plasmid gap repair and survival with ionizing radiation. To determine whether the defect in BIR seen in the pol32 ${Delta}$ andpol3-ct mutants was due to a general recombination defect orwas specific to the extensive tracts of DNA synthesis associatedwith BIR, the frequency of plasmid double-strand gap repairwas determined. In the plasmid gap repair assay, Ura⁺ transformantsresult from homology-dependent repair of a 238-bp gap withinthe MET17 open reading frame followed by integration of theplasmid at the chromosomal MET17 locus (Fig. 6A). Plasmid gaprepair occurs with a frequency of 20.7 x 10^–2 in the MW16/MD249background. As expected, the rad51 ${Delta}$ mutant diploid showed a significantdecrease in the frequency of plasmid gap repair (0.42 x 10^–2;P = 0.001), while the pol32 ${Delta}$ mutant diploid had only a twofold(P = 0.01) decrease, and there was no significant decrease forthe pol3-ct diploid (Fig. 6B). These results indicate that thedefect observed for the Pol ${delta}$ mutants in the BIR assay is specificfor BIR and is not due to a general recombination defect ora defect in cellular uptake of linear DNA during transformation.

View larger version (19K):
[in this window]
[in a new window]

FIG. 6. Pol ${delta}$ mutants are proficient for limited DNA synthesis following strand invasion. (A) Schematic representation of the gap repair assay. Following transformation of a gapped plasmid into competent yeast, both ends invade the endogenous chromosomal locus. Repair has to occur by integration to yield a stable Ura⁺ transformant. Repair by integration leads to duplication of sequences from the region of homology between the vector and the native chromosome. (B) Frequency of repair of a gapped plasmid substrate. Frequencies are measured as the number of Ura⁺ transformants per 100 ng of cut DNA divided by the number of transformants per 100 ng uncut DNA transformed. (C) Tenfold serial dilutions of each strain were spotted onto YPD plates and left unirradiated or were irradiated with 400 Gy.

To confirm this result, we tested wild-type and mutant strainsfor their ability to repair DNA damage caused by ionizing radiation.It is known that mutation of genes in the RAD52 epistasis groupconfers high sensitivity to ionizing radiation due to defectsin homology-dependent repair (54). To determine whether thesame was true for the Pol ${delta}$ mutants, diploid strains were grownto exponential phase and then exposed to 400 Grays of ionizingradiation. As expected, the rad51 ${Delta}$ mutant was highly sensitiveto ionizing radiation while the pol32 ${Delta}$ and pol3-ct mutants wereresistant (Fig. 6C). These results confirm that the BIR defectobserved in the Pol ${delta}$ mutants is specific for the homology-dependentrepair that requires long stretches of DNA synthesis.

DISCUSSION

Top

DISCUSSION

DNA synthesis is an essential step in homologous recombinationto replace nucleotides lost by excision of sequence during therepair process. Homology-dependent repair of DSBs initiatesfrom single-stranded DNA (ssDNA) generated by 5'-3' resectionof the DNA ends (25). Rad51 binds to the resulting ssDNA tailsto initiate pairing and strand invasion with homologous duplexDNA (25). The 3' end from the broken chromosome is used to primeleading-strand DNA synthesis templated by the donor duplex.For gene conversion by the synthesis-dependent strand annealingmodel (12, 42), the invading strand that has been extended byDNA synthesis is displaced and anneals to complementary sequenceexposed by 5'-3' resection of the other side of the break (Fig.1). Thus, only short tracts of DNA synthesis are required forrepair of DSBs by gene conversion. DSBs that present only oneend for repair, such as those produced by unprotected telomeres,undergo strand invasion and replication to the end of the chromosome(BIR). Because this process can lead to the amplification ofthe 6.7-kb subtelomeric Y' repeats, the tracts of DNA synthesisare likely to be considerably longer than those for gene conversion(29, 35). BIR can also occur at two-ended DSBs that have onlyone end with homology to a donor sequence and initiate extensivetracts (>100 kb) of DNA synthesis (31, 41). BIR from a two-endedDSB can result in extensive LOH or chromosome rearrangementsif it occurs between chromosome homologues or dispersed repeats,a common phenotype of tumor cell lines (6, 28, 52, 53, 55, 57).Furthermore, aging yeast cells more frequently repair spontaneousdamage by a mechanism resulting in LOH than young cells, andit has been suggested this mechanism is BIR (37). Thus, regulationof BIR at two-ended breaks is important for maintaining genomeintegrity. One proposed mechanism to control BIR is to delaystrand invasion or DNA synthesis from the invading end at aone-ended break until the cell senses the absence of the secondend; this implies coordination of the two ends of the breakat an early step (31). We recently suggested that another mechanismto control extensive replication at two-ended DSBs is the abilityof the invading end to displace from the template DNA. Assumingthe second end of a DSB is available to capture the displacedinvading end, this terminates replication, leading to limitedLOH (52). When the second end is absent, the displaced end invadesthe same or a different homologous template, a process referredto as template switching. Because template switching is limitedto about 10 kb downstream of the site of strand invasion, thereappears to be a change in the nature of the strand invasionintermediate or mode of DNA replication during BIR. This studywas initiated to gain insight into the replication factors thatextend the invading 3' end during BIR and how template switchingis regulated.

CF formation proceeds by a half-crossover mechanism in mutants defective for Pol ${delta}$ . In the transformation-based BIR assay, we found a sixfold decreasein the recovery of stable Ura⁺ transformants in the pol32 ${Delta}$ mutant;however, the CFs formed were frequently ( $~$ 50%) associated witha loss of the chromosome that was used as the donor for strandinvasion (Fig. 3). Thus, in the pol32 ${Delta}$ mutant, CF formation proceedsby a half-crossover mechanism and not by BIR. This process leadsto a nonreciprocal transfer of information from the donor tothe recipient, in this case from chromosome III to the CFV,leaving the cell with a broken chromosome III arm that is subsequentlydegraded. The donor chromosome is unchanged in true BIR events.These events were detected at a lower frequency in wild-typecells and are mechanistically similar to the processing of endsduring gene targeting (26, 46). Using a different assay systemfor BIR, Deem et al. (11) also showed generation of half-crossoverevents in the pol32 ${Delta}$ mutant, though the frequency of events waslower than that reported here and they found a more-severe defectin gene conversion than BIR in the pol32 ${Delta}$ mutant. POL32 was recentlyidentified in a screen of diploid yeast for mutants with elevatedrates of spontaneous LOH (1). Most of the LOH events observedin the pol32 ${Delta}$ mutant were due to reciprocal recombination, presumablybecause these initiate at two-ended DSBs or ssDNA gaps or becausethe BIR pathway is inefficient. The increased rate of spontaneousrecombination is most likely due to defects in DNA synthesis,resulting in more initiating lesions. Although the frequencyof CF formation was only modestly decreased in the pol3-ct mutant,the events recovered were also associated with high-frequencychromosome loss, suggesting strand invasion is proficient andthe defect occurs at a subsequent step (Fig. 3). In the caseof the pol32 ${Delta}$ mutant, it appears that the defect is at an earlierstep than in the case of the pol3-ct mutant because the recoveryof CFs is lower. We consider it unlikely for strand invasionto be defective in the pol32 ${Delta}$ mutant and assume the defect isdue to an inability to efficiently extend the 3' end of theinvading strand. These results are consistent with those inprevious studies showing an important role for Pol ${delta}$ in the extensionof the invading strand during gene conversion and BIR (30, 32,33). However, our data do not exclude the possibility that anotherpolymerase extends the invading end and the BIR failure of thePol ${delta}$ mutants is due to an inability to initiate lagging-strandsynthesis. We did not find any role for Pol ${eta}$ in these studies,suggesting Pol ${delta}$ is required for both the initial stage of DNAsynthesis and the transition to more processive DNA synthesis.

Short tracts of DNA synthesis in Pol ${delta}$ complex mutants. In the studies of Lydeard et al. (30), DNA synthesis of $~$ 250nucleotides from the invading end was undetectable in the pol32 ${Delta}$ mutant; however, mating-type switching, which requires at least700 bp of DNA synthesis, appeared unaffected. These resultssuggest Pol32 is required specifically to extend one-ended strandinvasion intermediates, a conclusion bolstered by the observationthat telomerase-independent survivors were absent from the pol32 ${Delta}$ mutant (30). To address whether any DNA synthesis occurs duringCF formation in the pol32 ${Delta}$ mutant, we made use of the polymorphicmarkers on chromosome III. If template switching occurs, weassume DNA was synthesized from the invading end prior to dissociationand invasion into the other chromosome homologue. In both pol32 ${Delta}$ and pol3-ct mutants, we obtained evidence of template switching;in most cases the chimeric CF was found in a 2N–1 diploid,indicating a half crossover occurred after the second strandinvasion event (Fig. 4). These results show DNA synthesis fromone-ended strand invasion intermediates is not eliminated bythe pol32 ${Delta}$ mutation, but the frequency of these events is greatlyreduced. This low level of DNA synthesis may be undetectableby physical assays (30). Because the recovery of CFs is higherin the pol3-ct mutant than in the pol32 ${Delta}$ mutant, we suggest strandinvasion occurs in both mutants but DNA synthesis is initiatedless frequently in the pol32 ${Delta}$ mutant, destabilizing the strandinvasion intermediate. The plasmid gap repair assay requiressynthesis of at least 238 bp. In this assay, only a twofolddecrease in the frequency of gap repair and integration wasfound in the pol32 ${Delta}$ mutant compared with results for the wildtype and no defect was found for the pol3-ct mutant (Fig. 6).Furthermore, the pol32 ${Delta}$ and pol3-ct mutants exhibit no sensitivityto ionizing radiation, even at a dose that generates about 40DSBs per diploid genome. These results suggest that strand invasionoccurs in the Pol ${delta}$ complex mutants and a short tract of DNAsynthesis ensues, followed by dissociation of the invading strand;this is sufficient for repair of two-ended DSBs by gene conversion.For successful BIR, more extensive DNA synthesis is required,presumably coupled to assembly of the full replisome. In orderto replicate 100 kb, we imagine the replicative helicase (MCMcomplex) would need to be recruited, as well as all the componentsof leading- and lagging-strand DNA synthesis. The MCM complexis not required for the DNA synthesis associated with gene conversion(56). Whether replisome assembly needs to occur before DNA synthesisfrom a one-ended break initiates or occurs after gene conversionfails is an open question.

Models for half-crossover formation. We suggest that following introduction of the CFV into the cell,strand invasion occurs, followed by a minimal amount of DNAsynthesis. The invading strand may dissociate and undergo asecond strand invasion event, giving rise to template switching.For replication of the entire chromosome arm (BIR), we suggestthe D-loop intermediate is cleaved and one strand of the vectorbecomes covalently joined to one strand of the donor duplex.This would prevent dissociation of the invading strand and createa structure for assembly of the full replisome, leading to processiveDNA synthesis. In the Pol ${delta}$ complex mutants, we suggest thatextension of sequence from the invading end is inefficient and/orthere is a defect in recruitment of lagging-strand replicationfactors. The D-loop would then be susceptible to a second cleavageevent, so that sequence distal to the strand invasion intermediatewould be transferred to the CFV, leaving a broken chromosomeIII arm that is subsequently degraded (Fig. 5). The number ofchromosome loss events in the pol3-ct diploid was reduced to24% by mutation of MUS81, suggesting at least half of the strandinvasion intermediates are cleaved by the Mus81-Mms4 nuclease(Table 1). The residual events could be due to cleavage by othernucleases, such as Rad1-Rad10 or Yen1 (18, 27), or could beformed by single-strand annealing (see below).

Because rad51 ${Delta}$ mutants are defective for strand invasion, wepropose a different mechanism for half crossovers in this background.In the rad51 ${Delta}$ strains, we propose that following transformationof the linear CFV into the cell, the CFV fuses with a fragmentof chromosome III. Rare chromosome breaks will persist in therad51 ${Delta}$ mutant because of the strand invasion defect. If a randomDSB occurs to the right of the region of homology shared withthe CFV (this would include all of the right arm of chromosomeIII) and the ends of the DSB and the CFV are processed, leadingto the formation of 3' ssDNA tails, the complementary ssDNAexposed could then be annealed, leading to the formation ofa full-length CF and a broken chromosome III arm (Fig. 7). Thislikely proceeds by RAD52-dependent single-strand annealing,a process known to be independent of RAD51 (46). Notably, CFsare not formed in rad52 ${Delta}$ mutants, and the frequency of CF formationin the rad51 mutant is reduced by mutation of RAD59, consistentwith the idea that this rare class of RAD51-independent eventsforms by single-strand annealing (10). Coic et al. (7) estimatedthe probability of a single DSB on the right arm of chromosomeIII to be 2 x 10^–3. Based on the transformation efficiencyof the rad51 ${Delta}$ mutants using a replicating plasmid, we estimate $~$ 5 x 10⁴ cells take up the linear CFV; thus, the number of spontaneousbreaks is sufficient to account for generation of the rare Ura⁺transformants by single-strand annealing. The other predictionfrom this model is that the number of Ura⁺ transformants recoveredfrom the rad51 ${Delta}$ mutant should be decreased by mutations thateliminate resection. Although CFs are still recovered in rad51 ${Delta}$ rad50 ${Delta}$ mutants, the rad50 ${Delta}$ mutation decreases the initiation ofresection and not the extensive resection suggested by the model(10, 39, 59). Zhu et al. (59) found that extensive resectionis reduced by mutation of SGS1, and consistent with this, weobserved a 20-fold decrease in the recovery of stable Ura⁺ transformantsin the rad51 ${Delta}$ sgs1 ${Delta}$ double mutant compared with that in the rad51 ${Delta}$ single mutant (data not shown). The sgs1 ${Delta}$ mutation by itselfdoes not decrease BIR (data not shown). Deem et al. did notrecover half-crossover events in rad51 ${Delta}$ mutants using their BIRassay (11). One possible explanation is that in the transformationBIR assay, we are able to select rare, low-frequency events,whereas the survival rate of the rad51 mutant is very high inthe homothallic switching endonuclease-induced break system,but most of the survivors lose the broken chromosome.

View larger version (10K):
[in this window]
[in a new window]

FIG. 7. Model for CF formation in the rad51 ${Delta}$ mutant. In rad51 ${Delta}$ diploids, following formation of a random chromosomal break, the break end and CFV are processed by nucleases. This results in complementary ssDNA tailed duplexes that are subsequently annealed by Rad52, leading to the formation of a CF and broken chromosome III (Chrom. III) homologue that is subsequently lost.

Though 25 to 50% of the Ura⁺ transformants recovered from rad51 ${Delta}$ ,pol32 ${Delta}$ , and pol3-ct diploids retained both chromosome III homologues,these are unlikely to be true BIR events. We do not know inwhich phase of the cell cycle the half crossovers occur. Ifstrand invasion initiates during the S or G₂ phase of the cellcycle, by random segregation the broken chromosome arm shouldsegregate with the CF 50% of the time (Fig. 8). About 50% ofthe events observed in pol32 ${Delta}$ and pol3-ct diploids are due tosegregation of the CF with the broken chromosome III (2N–1events). The elevated rates of 2N–1 plus CF events observedin the rad51 ${Delta}$ diploid may be caused by half-crossover formationbefore the chromosomes have replicated or by replication ofa broken chromosome followed by G₂ half crossover. In diploidcells, we are able to recover events in which the CF formedby a half-crossover segregates with the intact or broken sisterchromatid. In haploids, only the half-crossover events in whichthe CF segregates with an intact sister chromatid would be viable.CFs are recovered at a low frequency from the haploid rad51mutant, and these are most likely due to a half crossover inwhich the CF segregates with the intact sister chromatid, appearingas a BIR event. We suggest many of the LOH events previouslydescribed for rad51 ${Delta}$ mutants are due to half crossovers and notBIR (8, 55).

View larger version (8K):
[in this window]
[in a new window]

FIG. 8. Model for 2N + CF genotype when CF formation proceeds by a half crossover. Formation of a half crossover in G₂ results in a CF and broken chromatid. Following mitotic division, chromatids 1 and 3 can segregate with the CF while chromatid 2 and the broken chromatid 4 cosegregate. Following selection for uracil prototrophy, only the daughter with the CFV survives, leading to a 2N + CF genotype. If chromatid 1 and the broken chromatid cosegregate with the CF, then selection for uracil prototrophy allows only for growth of a 2N–1 + CF colony. Black corresponds to chromosome III sequences; gray corresponds to the CFV.

ACKNOWLEDGMENTS

We thank L. Maloisel for yeast strains and W. K. Holloman forcritical reading of the manuscript.

This work was supported by grants from the National Institutesof Health (GM041784 and GM054099).

FOOTNOTES

* Corresponding author. Mailing address: Columbia University Medical Center, Department of Microbiology, 701 W. 168th St., Rm. 916, New York, NY 10032-2704. Phone: (212) 305-4793. Fax: (212) 305-1741. E-mail: lss5{at}columbia.edu

Published ahead of print on 12 January 2009.

Present address: Ludwig Institute for Cancer Research, UCSDSchool of Medicine, La Jolla, CA 92130.

REFERENCES

Top