Cascades of Genetic Instability Resulting from Compromised Break-Induced Replication

Break-induced replication (BIR) is a mechanism to repair double-strand breaks (DSBs) that possess only a single end that can find homology in the genome. This situation can result from the collapse of replication forks or telomere erosion. BIR frequently produces various genetic instabilities including mutations, loss of heterozygosity, deletions, duplications, and template switching that can result in copy-number variations (CNVs). An important type of genomic rearrangement specifically linked to BIR is half-crossovers (HCs), which result from fusions between parts of recombining chromosomes. Because HC formation produces a fused molecule as well as a broken chromosome fragment, these events could be highly destabilizing. Here we demonstrate that HC formation results from the interruption of BIR caused by a damaged template, defective replisome or premature onset of mitosis. Additionally, we document that checkpoint failure promotes channeling of BIR into half-crossover-initiated instability cascades (HCC) that resemble cycles of non-reciprocal translocations (NRTs) previously described in human tumors. We postulate that HCs represent a potent source of genetic destabilization with significant consequences that mimic those observed in human diseases, including cancer.


Introduction
Double-strand DNA breaks (DSBs) are dangerous because they can lead to chromosomal rearrangements or cell death. DSBs may result from a cell's exposure to various DNA-damaging agents, such as radiation and various chemicals, including anti-cancer drugs. In addition, problems with DNA metabolism can also result in DSB formation. DSB-induced changes to the genome have been implicated in promoting various human diseases, including cancer, which emphasizes the importance of proper repair of such lesions. Multiple pathways of DSB repair have evolved (reviewed in [1,2]). Non-homologous end joining (NHEJ) is a repair mechanism in which two non-homologous broken ends of a DNA molecule fuse together, often producing small DNA insertions and deletions that can be destabilizing. Alternatively, homologous recombination (HR) mechanisms repair DSBs through recombination, where broken DNA ends initiate copying of a homologous sequence elsewhere within the genome. The most efficient pathway of HR is gene conversion (GC), where both ends of a DSB use a homologous sequence to copy lost DNA in order to repair the DSB lesion. Alternatively, break-induced replication (BIR) is an HR mechanism that employs only a single end of a DSB for repair.
During BIR, a single broken DNA end invades a homologous region within the genome to initiate extensive DNA synthesis that can copy large portions of a donor molecule through its telomere (reviewed in [3][4][5][6]). BIR is a primary pathway to repair broken replication forks and eroded telomeres. Also, it has been observed that gap repair can proceed through BIR [7] and, for reasons that are not entirely clear yet; the frequency of BIR is known to increase in aged cells [8]. BIR is initiated by strand invasion, which occurs with kinetics similar to those of the GC pathway [7]. However, after strand invasion, progress stalls and DNA synthesis is delayed by 4 or more hours [9]. The exact reason for this pause is not known, but several possibilities have been proposed, including slow replication fork assembly, unstable D-loop formation, and the existence of a ''recombination checkpoint'', discouraging BIR repair (reviewed in [3,5,6]). The delay in BIR initiation leads to the establishment of a checkpoint-mediated G2/ M arrest, which allows cells to complete BIR prior to cell division. Consistently, it was observed that a defective checkpoint (achieved by deletion of RAD9) decreased BIR efficiency and also increased chromosome loss [9]. Once DNA synthesis associated with BIR is initiated, it is fast and processive, similar to normal S-phase DNA replication [9]. It has been demonstrated that the initiation of BIR DNA synthesis involves the majority of proteins required for initiation of S-phase DNA replication [10]. Also, Pold, a main replicative polymerase, plays a crucial role in DNA synthesis in BIR [11][12][13]. However, the role of two other replicative polymerases, Pole and Pola, in BIR remains unclear [11], (also reviewed in [5,6]). In addition, the mechanism of BIR is drastically different from S-phase replication as it proceeds via migrating DNA bubble leading to conservative inheritance of newly synthesized strands [14,15,16].
While the end result of BIR is repair of the DSBs, the mechanism of BIR increases the likelihood of a variety of deleterious outcomes that may have destabilizing consequences in the genome. Among these are loss of heterozygosity, deletions, duplications, translocations, copy-number variations, and a significantly elevated mutation rate [9,[17][18][19][20][21][22][23][24][25]. In addition, halfcrossovers (HCs), which are chromosome fusions initially identified in rad51D and rad52D mutants [13,[26][27][28][29], were recently demonstrated to occur in wild type and various mutants following initiation of BIR [12,13]. BIR-induced HC formation is initiated by strand invasion, but the resulting intermediate ruptures prior to repair to yield a rearranged chromosome consisting of fused pieces of the recipient and donor molecules, as well as a destabilized fragment of the broken donor. Accordingly, HC formation requires proteins involved in the strand invasion step of BIR, however the impairment of proteins involved in BIR after strand invasion promotes HCs [12]. Thus, HCs are markedly elevated in yeast bearing mutations pol32D and pol3-ct, which interfere with successful initiation of DNA synthesis [12,13]. It was proposed that the failure to initiate DNA synthesis in these mutants promotes resolution of the Holliday junction (HJ) formed during strand invasion. The exact mechanism of HJ resolution remains unknown, though the resolvase Mus81 has been implicated as one protein capable of resolving HJs and therefore may contribute to HC formation [13].
HCs result in the breakage of a previously intact donor chromosome, and this can have deleterious consequences by initiating recurrent cycles of genetic instability. Analogous cycles (called NRTs, for non-reciprocal translocations) have been described in mammalian tumors where broken chromosomes initiate recombination with an intact donor, which in turn leads to breakage of the donor [30]. While the molecular mechanism of NRTs remains undefined, we have previously proposed [12] that cycles of NRTs are mediated by cascades of HCs that continue until the donor fragments are either stabilized or lost. Thus, further investigation into HC formation and the possible cascades of genetic instability that may result is warranted.
To further define mechanisms of HC formation and the effects of HCs on genetic instability, we hypothesized that various factors that interrupt ongoing BIR replication may induce HC formation in a manner similar to mutations that prevent initiation of BIR replication. We show that interruption of BIR synthesis by exposure of cells to DNA damaging agents or due to a defective replisome results in a dramatic increase in HCs. Moreover we demonstrate that a disruption of BIR imposed by premature onset of mitosis increase HC formation. Finally, we document the occurrence of half-crossover-initiated instability cascades (HCCs) that closely resemble NRT cycles observed in cancer cells.

Experimental system
Half-crossovers (HCs) are chromosome fusions resulting from aberrant processing of BIR intermediates. It has been proposed that HCs lead to deleterious consequences by initiating cascades of genetic instabilities. Here we aimed to identify genetic factors that promote the channeling of BIR into HC formation.
To assay the efficiency of BIR and the frequency of halfcrossovers in DSB repair, we employed our disomic experimental system in yeast, Saccharomyces cerevisiae, wherein a galactose-induced DSB is initiated at the MATa locus on the truncated copy of chromosome III (recipient chromosome) (Fig. 1A) [12]. The second full copy of chromosome III contains the uncleavable MATa-inc allele and serves as a template for DSB repair (donor chromosome). Upon induction of the DSB, DNA is repaired predominantly by BIR (Fig. 1B) because only one end of the DSB has large homology to the full-length donor copy of chromosome III. The ends of both chromosomes that participate in BIR repair are marked by ADE1, LEU2, ADE3 or HPH; such that repair outcomes can be determined using appropriate selective media. Also, a NAT cassette was used to replace a region 30 kb centromere-proximal to MATa that contained two Ty1 elements (FS2) in the recipient chromosome [21]. Using these genetic markers, it was determined that more than 75% of DSB repair outcomes displayed an Ade + Leu 2 phenotype, indicating BIR repair of the galactose-induced DSB (Fig. 1B, Fig. S1). Approximately 14% of the DSB repair outcomes were Ade + Leu + , indicating the DSB was repaired by gene conversion (GC) (Fig. 1C, Fig. S1). Other colonies had an Ade 2r Leu 2 phenotype (were ADE1-deficient and red (as described in [12]) and resulted from failure of the chromosome to repair the DSB leading to chromosome loss (CL) (Fig. 1D). Also, a small number of colonies were Ade 2w Leu 2 (were ADE3-deficient and white), which represented HC events resulting from fusion of the ADE1containing segment of the recipient chromosome with the HPHcontaining segment of the donor chromosome and concurrent loss of the ADE3 and LEU2 segments of the donor and recipient chromosomes, respectively (Fig. 1E).

Reduced processivity of Pold promotes HC formation
We tested the effect of mutations that impair DNA polymerases on HC formation by plating yeast on a galactose-containing medium [12]. Although each of the Pold mutations tested here had varying effects on BIR efficiency, they all stimulated HC formation. In particular, pol3-Y708A, a mutation that affects the catalytic subunit of Pold [31], dramatically decreased BIR efficiency, and increased chromosome loss (P,0.0001; Fig. S1). Also, pol3-Y708A increased the number of colonies containing HCs

Author Summary
Maintaining genomic stability is important to prevent birth defects, genetic disorders and other diseases, including cancer. Double-strand DNA breaks (DSBs), which can result from exposure of living cells to ionizing radiation and various chemicals, threaten genomic integrity, thus making DSB repair essential. The choice of DSB repair pathway is important because some pathways confer destabilizing consequences. Break-induced replication (BIR) is a mechanism of DSB repair that is often associated with deleterious events that can threaten genetic stability. One such deleterious event is the formation of halfcrossovers (HCs), which occurs when two chromosomes physically interacting during BIR repair fuse. Here we employed a yeast-based system to unravel the genetic factors promoting HC formation. We demonstrate that the interruption of BIR due to problems in DNA synthesis or checkpoint control, promote HCs. Additionally, we document that disruption of BIR promotes half-crossoverinitiated cascades (HCC) that can significantly destabilize the genome and could be accounted as a potential mechanism responsible for cycles of non-reciprocal translocations contributing to cancer in humans.  [12]. A DSB is created at MATa in truncated Chr III (recipient (upper) chromosome) by a galactose-inducible HO endonuclease. The MATa-inc chromosome (donor (lower) chromosome) is full-length and is resistant to cutting by HO. The ends of the recombining chromosomes are marked by ADE1, LEU2, ADE3 and HPH. Two copies of Ty1 elements (Ty1a and Ty1b), comprising the FS2 region located ,30 kb centromere proximal from MATa, are replaced by a NAT cassette in the recipient chromosome. The positions of two other Ty elements (Ty1c and Ty1e) comprising the FS1 region are shown. (B) Schematic representation of an Ade + Leu 2 (Break-Induced Replication (BIR)) outcome. (C) Ade + Leu + (Gene Conversion (GC)) outcome. (D) Ade 2r Leu 2 (Chromosome Loss (CL)) outcome. (E) Ade 2w Leu 2 (Half-Crossover (HC)) outcome. (F) Ade + Leu 2 (Chromosomal rearrangement (CR)) outcome. (G) Ade + Leu 2 (Half-crossover-initiated cascade (HCC)) outcome. doi:10.1371/journal.pgen.1004119.g001 to 17% compared to approximately 5% in wild type (P,0.0001; Fig. 2A; see also Fig. S1; note that Fig. 2 presents the fraction of colonies that are fully or partially HCs, while Fig. S1 shows the fraction of HCs among all DSB repair events). Similarly, the pol31-WRRGW mutation, which disrupts the Pol31-Pol32 interaction [32], displayed similar effects and elevated HCs to 26% (P, 0.0001; Fig. 2A; Fig. S1). These phenotypes were similar to those previously observed in pol32D and pol3-ct mutants [12,13], suggesting that HCs in these mutants are promoted predominantly by failure to initiate BIR. In strains bearing the pol3-t mutation known to compromise the processivity of Pold during S-phase DNA replication [33,34], HCs were also elevated (P,0.0001), even though these cells frequently successfully completed BIR repair ( Fig. 2A; Fig. S1). Therefore, the increase of HCs in pol3-t might be explained by interruptions of ongoing BIR rather than by problems in BIR initiation. Thus, an intact Pold appears to be necessary to prevent HC formation. Conversely, mutations impairing Pole showed no effect on HCs. Thus, no increase compared to wild type was observed in either pol2-1 [35] mutants with a truncated catalytic subunit or in pol2-Y831A mutants [31] with a mutation in the same conserved catalytic motif as pol3-Y708A mutants ( Fig. 2A). Interestingly, the pol1-1 mutation [36], which impairs Pola (a part of the primase complex), decreased HCs to less than 0.3% (P,0.0001; Fig. 2A).

DNA damage is synergistic to BIR in promoting HC formation
Due to the increased HCs observed in mutants with decreased polymerase processivity, we hypothesized that BIR pausing induced by damage in the template DNA used for BIR could promote HC formation. This idea was tested by analyzing DSB repair in our BIR system in the presence of DNA damaging agents. Following induction of BIR in galactose-containing liquid cultures, cells were exposed to either the alkylating agent MMS or 4-NQO (UV mimetic) for seven hours while BIR repair occurred. Cells were plated on YEPD and the resulting colonies were analyzed using selective media ( Fig. 2B; Fig. S2). Cell viability was calculated by plating cells on YEPD (see Materials and Methods for details). Both drugs were deactivated (see Materials and Methods) prior to being plated on YEPD. In cells treated with 2.4 or 6 mM MMS, the percent of colonies with HC outcomes was increased approximately 3-and 4-fold, respectively, compared to cells treated with galactose but no drug (P,0.0001), while the percent of HC outcomes approximately doubled in cells treated with 5.3 mM 4-NQO compared to the same control (P,0.0001; Fig. 2B; see also Fig. S2). Significantly, the level of HCs in cells exposed to damage alone (without DSB induction) was not significantly increased in comparison to no-damage control (Fig. 2B, no-DSB). Thus, our data suggest that the observed increase in HCs is promoted not by DNA damage per se, but by base damage in the chromosomal region undergoing BIR repair. Our data also identifies a previously unknown synergy between BIR and DNA damage that dramatically increases the rate of HC formation. We note that the levels of DNA damaging drugs used in these experiments, as expected, decrease viability (see Materials and Methods). Overall, since both MMS and 4-NQO induce base damage capable of blocking DNA polymerases (and therefore replication) [37][38][39], our results suggest that interruption of ongoing BIR leads to the aberrant processing of BIR intermediates resulting in HC formation.

Checkpoint deficiency promotes HC formation
Previously, we demonstrated that initiation of DNA synthesis during BIR is a very slow process (takes up-to 4 hours) and leads to the establishment of a checkpoint-mediated G2/M cell cycle arrest [7,9] that prevents mitotic division and thus allowing cells to complete BIR. Given our data that interruptions in BIR due to decreased processivity of polymerases or DNA damage promote HC formation, we hypothesized that checkpoint deficiency may stimulate HC formation due to an interruption in BIR progression by the premature onset of mitosis. To test this hypothesis, we analyzed BIR outcomes in mutants lacking Rad9 or Rad24, which are required for the DNA damage response (reviewed in [40]). Consistent with previous observations in these mutants [9], checkpoint deficiency led to a high frequency of multi-sectored colonies (colonies containing $3 different repair sectors) (Fig. 3A,  Fig. 4C). Formation of multi-sectored colonies likely resulted from premature onset of mitosis, and DSB repair that occurred after subsequent cell divisions and took place only in a fraction of the daughter calls. Consistent with this idea were the results of FACS analyses that confirmed full G2/M arrest in wild type cells between 4 and 8 hours after DSB induction, with only partial arrest observed at these time points in rad9D and rad24D mutants (Fig. 3B).
We observed that the percentage of colonies with at least one HC sector was extremely high, 71% and 65% among multisectored colonies of rad24D and rad9D mutants respectively. This was a significant increase (P,0.0001) compared to wild type where the frequency of colonies with HC was only 4% (Fig. 4D). Also, when calculated as a fraction of all sectors, HC sectors comprised approximately 20% in rad9D and rad24D mutants ( Fig. 4E; left). Among simple colonies (with no more than 2 sectors or events evident), no notable difference in HC frequency was observed between the checkpoint-deficient and wild type strains (Fig. 4D). We propose that the increase in HCs in checkpointdeficient mutants results from premature onset of mitosis that may occur either during the first cell division following DSB induction or during subsequent cell divisions, as further explored in the following sections.
DSB repair in checkpoint-deficient mutants results in colonies with multiple, complex outcomes Genetic analysis of repair outcomes in checkpoint-deficient mutants revealed increased chromosome loss (P,0.0001; Fig. 4A, 4E; middle) and a decreased level of Ade + Leu 2 outcomes, which normally represent BIR (P,0.0001; Fig. 4A, 4E; right). This was consistent with our previous results [9] and most likely reflected failed DSB repair in these strains. The decreased BIR efficiency in checkpoint-deficient mutants was further supported by PFGE analysis of cells undergoing DSB repair over a 10-hour time course (performed similar to [12]), where the amount of BIR repair product was significantly reduced in rad9D and rad24D mutants compared to the wild type ( Fig. 3C-F)). Based on the decreased efficiency of BIR, we hypothesized that a fraction of Ade + Leu 2 events in checkpoint deficient mutants might in fact represent not BIR, but GCRs resulting from abnormal stabilization of the broken molecules (similar to discussed in [21,41]). To detect possible GCRs and to characterize their contribution to heterogeneity of the colonies, we employed PFGE to analyze repair outcomes in 23 individual rad24D and 11 rad9D Ade +/2 multisectored colonies, which comprised Leu + and/or Leu 2 clones (see, for example, colonies in Fig. 5A, B). In this analysis, we focused on individual colonies with at least one HC sector because they represented the majority of all multi-sectored colonies in both mutants. All Ade + Leu 2 sectors as well as a representative number of Ade + Leu + (GC), Ade 2w Leu 2 (HC) and Ade 2r Leu 2 (CL) sectors from each colony were cloned out and analyzed (see, for example, Fig. 5C for PFGE analysis of all sectors from the colony shown in Fig. 5A). PFGE analyses of colony sectors with Ade + Leu + , Ade 2w Leu 2 and Ade 2r Leu 2 phenotypes confirmed that they resulted from GC, HC, and CL, respectively, as predicted ( Fig. 5C and see also Fig. 4A, B). PFGE analysis of the Ade + Leu 2 outcomes from individual colonies revealed three categories of outcomes (Fig. 5D). In the first group, which comprised 64% and 35% of all Ade + Leu 2 events in rad24D and rad9D, respectively, the chromosome   structure was similar to the one observed in true BIR outcomes. However, the high fraction of HC events in the respective colonies makes it highly likely that many of these Ade + Leu 2 outcomes resulted not from BIR, but from segregation of a HC repair product with an intact copy of the full-length chromosome III (similar to events described in [12,13]; see Materials and Methods for details). Other Ade + Leu 2 outcomes were represented by events where DSB repair resulted in formation of GCRs. Thus, approximately 13% of Ade + Leu 2 from multi-sectored colonies of each of the checkpoint-deficient mutants were chromosomal rearrangements (called CRs), where the broken recipient chromosome was aberrantly stabilized by de novo telomere formation or through ectopic recombination between a Ty or delta element in the MATa-containing chromosome and Ty or delta element located in an ectopic position (Fig. 1F, Fig. 5D; similar to previously demonstrated [21]). These CRs carried an unchanged donor chromosome (a 356 kb band that hybridized to the ADE3specific probe) and a recipient band of any size (different from 346 kb) that hybridized to ADE1 (Fig. 4A, 4B; CR). Previously, the structure of similar CR events was characterized by array-CGH [21] and it was determined that CRs often result from ectopic BIR initiated by strand invasion of Ty or delta elements of the broken recipient chromosome into Ty or delta elements at ectopic positions.
A significant fraction of Ade + Leu 2 outcomes represented a new type of GCR that contained a single BIR-sized (346 kb) recipient chromosome and a rearranged donor (a band other than 356 kb that hybridized to ADE3 (Fig. 4A, 4B; half-crossover-initiated cascades (HCC), Fig. 5D)). We posited that these repair outcomes likely arose from the rupture of the donor chromosome during HC formation, resulting in an ADE3-containing broken fragment that was stabilized by ectopic recombination. The possibility of such HCC events has been previously discussed [12], but never demonstrated. Here we found that 61% and 73% of rad24D and rad9D multi-sectored colonies contained at least one Ade + Leu 2 sector that represented a HCC event (Fig. 5D), the molecular structure of which was further analyzed by array-CGH (see below).
We conclude that premature onset of mitosis resulting from a defective checkpoint leads to aberrant processing of BIR intermediates resulting in frequent HCs and other GCRs. We observed that more than 74% and 91% of all analyzed colonies in rad24Dand rad9D, respectively, contained at least one CR or HCC sector. Also, both HCCs and CR events were frequently observed among Ade + Leu 2 events obtained from unselected (random) colonies in both rad9D and rad24D mutants, but were very rare in the wild type strains (Fig. 6A, 6B, 6F, and [9,12]). In addition, the analysis of strains containing the pol3-t mutation, which increased the frequency of HC formation (discussed in the previous section) also revealed DSB-induced HCC outcomes (Fig. S3).

Analysis of half-crossover-initiated cascades (HCC)
As indicated above, the majority of multi-sectored colonies in rad9D and rad24D mutants contained at least one HCC event characterized by the presence of a 346 kb band that hybridized to ADE1, as well as a second band of varying size (other than 356 kb) that hybridized to an ADE3-specific probe and represented a GCR that resulted from breakage and stabilization of the donor chromosome ( Fig. 4B; HCC). We used comparative genomic hybridization (array-CGH) to further characterize the nature of 13 stabilized donor chromosomes obtained from HCC events identified in rad24D mutants. (Fig. 7A) Based on array-CGH, the stabilized donor chromosomes resulting from HCCs were divided into three main classes that accounted for all 13 analyzed outcomes: isochromosomes (Class I), translocations (Class II), and secondary BIR events (Class III) (Table S2).
Class I rearrangements included 9 of the 13 HCC events analyzed by array-CGH (Fig. 7A, Table S2). Each of these events had in common a deletion of sequences in the right arm of chromosome III and a duplication of sequences from the opposite chromosome arm (Fig. 7B, 7C; Table S2), thus forming an isochromosome. (Class I is subdivided into Class Ia, Ib, Ic, and Id depending on the point of recombination and other details of the process; see Table S2). We propose that the formation of outcomes Ia, Ib, and Ic was initiated by invasion of the broken recipient into the full donor chromosome III, which led to the formation of an HC represented by a 346 kb band hybridized to ADE1-specific probe (Fig. S4, S5). The resulting broken ADE3-containing fragment was then resected and subsequently repaired by nonallelic homologous recombination between a Ty or delta element located in the right arm of chromosome III and a Ty or delta element located in the left arm of chromosome III. For example, in the case of Class Ia outcomes (H7 and H8), the recombination occurred between the Ty1c element in FS1 and a delta element YCLCd1 (Fig. 7A, Table S2, Fig. S4). The predicted size of such an isochromosome (calculated based on [19,42], and also based on the data from SGD) was approximately 245 kb (Table S2), which was consistent with the size of the ADE3-hybridizing band observed by PFGE analysis (Fig. 7A). Our proposed molecular structure was further confirmed through a detailed Southern analysis using the restriction enzymes EcoO1091 and EciI and Probe 1 (FS1-specific) and Probe 4 (specific to the region of chromosome III located centromere-distal to YCLCd1 (Fig. 7B,  Fig. S4, Table S3). Analogously, in the case of Class Ib outcomes (H10, H11, H12 and H13), recombination occurred between delta elements YCRCd6 and YCLWTy1-1 [19], which corresponds to YCLWd15in SGD. The predicted size of such an isochromosome was approximately 219 kb (Table S2), which was consistent with the size of the ADE3-hybridizing band observed by PFGE analysis (Fig. 7A). The structures of H10, H11, and H13 were further confirmed by Southern analysis using the restriction enzymes EciI and PshAI and also Probe 4 and Probe 5 (specific to region of chromosome III located centromere proximal to YCRCd6) (Fig. 7C,  Fig. S5, Table S3). Also see Text S1 and Fig. S6 for a detailed description of Class Ic HCC outcomes (H4 and H5). The formation of the HCC outcome H2 (Class Id) can be explained similarly to other Class I events, but likely involved two halfcrossover events (see Text S1 and Fig. S7).
Class II included only one of the 13 HCC outcomes, outcome H3. H3 was determined to result from a deletion in chromosome III between positions of FS1 (a tandem repeat of Ty1 elements) and MAT, and a duplication of all sequences located on grey and blue arrowheads, respectively. (C) Percentage of multi-sectored colonies (containing $3 repair sectors) formed following DSB induction in wt and checkpoint-deficient mutants. Results of 3 to 4 experiments performed for each strain were used to calculate the average 6 SD. See Materials and Methods for the number of analyzed colonies. (D) Percent of colonies containing HC events in wt and checkpoint-deficient mutants. (E) Percent of various repair outcomes (HC, CL and Ade + Leu 2 ) calculated among all repair events detected in simple and multi-sectored colonies (See Materials and Methods for details). Asterisk indicates that BIR could not be distinguished from cases where a HC chromosome segregated in mitosis with an intact donor chromosome (see text for details). doi:10.1371/journal.pgen.1004119.g004 Southern blot hybridization with ADE3-and ADE1-specific probes, respectively. Note that the size of BIR products in BIR-LR outcomes (lanes 1 and 6) is chromosome V distal to a solo delta element YERCdelta16 (Fig. S8 and Table S2). We propose that the formation of H3 was initiated by HC formation between the left arm of the recipient and the right arm of the donor chromosomes (similar to Class I), which resulted in a broken ADE3-containing fragment. This fragment was subsequently repaired by recombination between a Ty element in FS1 and a delta element located in the right arm of chromosome V, which led to the formation of translocation. (See Text S1, Table S2 and Fig. S8 for the details of structural analysis of H3).
Class III rearrangements included 3 of the 13 analyzed HCC outcomes and were further divided into IIIa and IIIb. Class IIIa was represented by the outcome H9. Array-CGH analysis of this outcome indicated a duplication of chromosome III sequences from MAT through the telomere indicative of BIR repair; however, PFGE analysis revealed that both ADE1 and ADE3hybridizing chromosomes were equal in size (346 kb long; Fig. S9, Table S2). We hypothesized that, similar to other HCC events, the formation of H9 was initiated by HC, which led to the formation of an ADE3-containing broken fragment. This broken donor fragment was stabilized through invasion into the HC product centromere proximal to NAT followed by BIR that copied the right arm of HC (Fig. S8). Therefore, we named this outcome a ''secondary BIR event''. Importantly, 8 of 24 HCCs that were originally identified by PFGE showed a pattern similar to H9 (both ADE1and ADE3-hybridizing bands were approximately 346 kb; data not shown), which suggests that all of them most likely represented secondary BIR events, even though only H9 was analyzed by array-CGH. Therefore, it appears that secondary BIR events are relatively common among BIR outcomes in checkpointdeficient mutants.
The array-CGH analysis of the outcomes H6 and H1 (class IIIb) also showed a duplication of chromosome III sequences from MAT through the telomere indicative of BIR, which made them similar to secondary BIR events. However, these events included additional rearrangements (Table S2; see also Text S1 for details).
Overall, we conclude that interruptions during BIR repair in checkpoint-deficient mutants lead to frequent breakage of the donor chromosome that results in further cascades of DNA instabilities.

SGS1 modulates distribution of repair outcomes in checkpoint-deficient mutants
It has been demonstrated that the absence of Rad9 increases the rate of resection at a DSB, which could contribute to the increased frequency of chromosome loss and GCRs we observed in rad9D mutants [43,44]. Therefore, we tested whether sgs1D and rad50D known to decrease the efficiency of DSB resection) [45][46][47] affected the distribution of repair outcomes in rad9D and rad24D.
We observed that deletion of SGS1, which is required for longrange 59-strand resection, in rad24D, dramatically reduced the frequency of chromosome loss (P,0.0001) (Fig. 4E), and virtually eliminated all multi-sectored colonies (Fig. 4C). The majority of colonies formed in rad24Dsgs1D were fully Ade + Leu 2 , and their PFGE analysis indicated that they contained normal BIR events ( Fig. 4E; Fig. 6D, 6F), even though they can also represent secondary BIR events. Deletion of SGS1 in rad9D also affected the distribution of repair outcomes. Compared to rad9Dalone, the frequency of chromosome loss was decreased (P,0.0001), while HCC were increased in the double mutant (P,0.0001) (Fig. 4E;  Fig. 6E, 6F). In addition, we observed that deletion of RAD50, which is involved in end processing near the DSB site [45], did not affect the distribution of repair outcomes in rad9D or rad24D (Fig. 4E). Importantly, the deletion of SGS1 gene alone (in strains with functional checkpoint response) led to only a modest change in distribution of DSB repair outcomes (Fig. 4C, 4E, 6C, 6F.). Overall, our data suggest that deletion of SGS1 significantly affects the distribution of repair outcomes in the absence of a functional checkpoint response.

Discussion
Decreased quality of BIR DNA synthesis promotes HC formation BIR is a critical mechanism to repair broken chromosomes. Normally, BIR is initiated by a DSB produced in such a way that only one end of the broken molecule is available for repair (Fig. 8A). It thus initiates with a single invasion into a homologous template (Fig. 8B) followed by initiation of DNA synthesis (Fig. 8C) that proceeds to the telomere (Fig. 8D). Increased HC formation and chromosome loss was previously demonstrated in pol32D and pol3-ct mutants during BIR repair where strand invasion was successful, but DNA synthesis could not be (or was poorly) initiated [12,13]. Likewise, here we report a similar phenotype in strains containing other mutations in Pold, pol3-Y708A (a mutation affecting the catalytic subunit of Pold) [31] and pol31-WRRGW (a mutation in the Pol31 subunit of Pold) [32]. For each of these cases, we propose that HC formation results from resolution of HJ structures that persist when BIR DNA synthesis is not initiated (Fig. 8B, 8L). It has been suggested that Mus81 resolves HJs and therefore may contribute to HC formation [13].
We demonstrate that BIR interrupted at various stages of its progression, for example during replication, also stimulates HCs. This most likely occurs in pol3-t mutants because the processivity of Pold is compromised (Fig. 8C, 8L). Similarly, results from another recent study [15] demonstrate that the deletion of PIF1, which encodes a DNA helicase specifically required for DNA synthesis during BIR [48], also leads to more frequent HCs. We speculate that in these mutants, DNA synthesis is successfully initiated but proceeds with frequent stops, thereby promoting HC formation. We postulate that a similar mechanism of paused DNA synthesis can explain the increase in HC formation we observed in wild type cells exposed to the DNA damaging agents (MMS or 4-NQO) during BIR repair (Fig. 8C, 8L). Regardless of the mechanism that leads to paused replication, these data support our hypothesis that interruptions in DNA replication during BIR induce HCs. Interestingly, mutations affecting Pole that were investigated so far did not promote HCs. This might be explained by limited participation of Pole in BIR [11]. Curiously, the mutation in Pola (pol1-1) led to decreased HCs compared to wild type, which may indicate that pol1-1 delays accumulation of BIR intermediates that are resolved to produce HCs.
We previously demonstrated that successful completion of BIR replication requires checkpoint machinery to maintain cell cycle arrest until repair is completed [9]. Consistently, here we observed that the premature onset of mitosis in checkpoint-deficient cells 356 kb instead of 346 kb because they resulted from strand invasion centromere proximal to NAT (see Materials and Methods for details). (D) Summary of the results of PFGE analysis of all sectors from individual colonies in rad24D (left) and rad9D (right). All DSB repair outcomes are abbreviated similarly to Fig. 4A undergoing BIR repair led to an increased frequency of HCs (Fig. 8F, 8L). Formation of HC molecules could result from a signal from the cell to resolve the HJ structure as previously discussed, but mechanical rupture of BIR intermediates initiated by chromosomal segregation is also a possibility. In checkpointdeficient mutants, we also frequently observed chromosome loss and translocations, which we propose result from strand dissociation that can be stimulated by HJ resolution (Fig. 8G, 8H). 59-to-39 DNA resection following dissociation may lead to chromosome loss (Fig. 8K)

Half-crossovers initiate cascades of genetic instability
An important outcome of this study is the discovery of HCinduced cascades (HCC). The existence of HCCs has been previously hypothesized [12], but, until now, had not been demonstrated. HCCs represent DSB repair outcomes that contain a HC product along with a rearranged donor chromosome (Fig. 8N). We propose that HCCs are initiated by a single HC that leads to breakage of the donor chromosome (Fig. 9A,B). The new DSB in the donor molecule undergoes 59 to 39 resection (Fig. 9C), and the resulting 39 DNA end invades a homologous DNA molecule at an ectopic position (Fig. 9E) in the newly formed HC (Fig. 9F), or in the sister chromatid (Fig. 9G). This initiates recombination and can stabilize the broken donor chromosome if repair proceeds through BIR; conversely, this intermediate may also result in HC formation, thereby continuing the cascade of genetic instability. Even when the donor fragment successfully stabilizes through BIR, if an ectopic site such as a Ty or delta element is used for recombination, translocations will occur ( Fig. 9E; see also Figs. S4, S5, S6, S7, S8).
In our system, repair of the broken donor chromosome often proceeded by BIR using the recently formed HC (Fig. 9F) or the sister chromatid (Fig. 9G) as a template. We termed these events ''secondary BIR''. Among all HCC events analyzed by PFGE, approximately half showed a pattern suggestive of HCC resulting in invasion of the broken donor into the initial HC, even though only one of these cases was analyzed by array-CGH (case H9; Fig. S9). Additionally, 15% of the HCCs analyzed by CGH could be explained by secondary BIR associated with complex rearrangements (cases H1 and H6). It should be noted that all identified secondary BIR events were initiated by strand invasion that occurred centromere proximal to FS2 which resulted in a change in the size of the donor and therefore allowed the detection of these events. It remains possible; however, that many additional secondary BIR events are initiated by strand invasion between FS2 and MAT. This is expected to result in chromosome structures and phenotypes indistinguishable from classic BIR. Therefore, we propose that the actual frequency of secondary BIR events maybe higher than currently estimated. This is significant because secondary BIR events could be more deleterious than classic BIR events. For example, we speculate that secondary BIR may result in homozygous mutations that result when a mutation occurs during DNA synthesis associated with HC formation (Fig. 9F) and is then copied during the repair of the broken donor using the initial HC as a template. Homozygous mutations could be more deleterious than heterozygous ones (reported in association with classic BIR [18]) because they can lead to the manifestation of recessive phenotypes including those leading to cancer.
A significant finding was the formation of multi-sectored colonies consisting of broad genotypic variations by checkpointdeficient mutants. More than 70% of the multi-sectored colonies contained at least one sector with chromosomal rearrangements of recipient or donor chromosomes, with HCC being a major class of these rearrangements. In addition, we observed that deletion of SGS1, which is known to reduce long-range DSB resection, led to a significant increase in HCC frequency in rad9D mutants. Possibly, reduced resection stabilizes the broken donor chromosome, thus giving it more chances to repair by invading a homologous template (Fig. 9E-G). Alternatively, it is possible that faster initiation and/or progression of BIR that was previously documented in the absence of Sgs1 [7] contributes to the increased stabilization of the donor chromosome and therefore to the increased level of HCC. Interestingly, in rad24Dsgs1D, the multiple sectoring of colonies was completely eliminated, and the majority of outcomes were indistinguishable from normal BIR. We propose that these events are likely to be secondary HCC resulting from secondary BIR.
In addition to being frequently observed when BIR is induced in checkpoint-defective mutants, HCCs occurred in cases of compromised BIR in polymerase-deficient mutants (pol3-t) and when BIR proceeds in the presence of MMS (Sakofsky et al, manuscript submitted). HCCs were also observed in BIR-defective pif1D mutants [15]. Overall, we propose that ongoing cycles of genetic instability are a ubiquitous outcome of HC formation.

HC-induced cascades: Potential for promoting genetic instability in humans
We propose that HCs and HCCs may be a mechanism for genetic destabilization leading to various diseases in humans. In particular, we propose HCCs to be a mechanism capable of producing non-reciprocal translocations (NRTs) that have been described in mammalian tumor cells. NRT is a pathway of telomere acquisition by broken chromosomes that results in the donor molecule losing genetic information, including its telomere, and becoming unstable [30]. This destabilization of the donor makes NRTs especially devastating because the events are selfperpetuating and result in cascades of genomic destabilization, including chromosome loss and multiple rearrangements. We Figure 6. Structural analysis of random Ade + Leu 2 DSB repair outcomes. PFGE analysis of randomly selected Ade + Leu 2 DSB repair events obtained in (A) rad24D, (B) rad9D, (C) sgs1D (D) rad24Dsgs1D, and (E) rad9Dsgs1D. On the top of each panel: Ethidium bromide-stained PFGE gel; in the middle and on the bottom: Southern blot analysis of the PFGE gel using ADE3-and ADE1-specific probes, respectively. Lanes labeled ''C'': genomic DNA from rad24D before DSB induction. Lanes labeled ''BIR'' depict DNA from a BIR outcome. See Fig. 4A and 4B for the structure of BIR, CR, and HCC repair events. Note that the size of BIR product in BIR-LR outcomes is 356 kb instead of 346 kb because they result from strand invasion centromere proximal to NAT (See Materials and Methods for details). For rad24D (shown in A), rad9D (shown in B) and rad9Dsgs1D (shown in E), lanes 1-13, lanes 14-26 and lanes 53-65 represent Ade + Leu 2 sectors from multi-sectored colonies. For sgs1D (C) and rad24Dsgs1D (D) lanes 27-39 and 40-52 represent Ade + Leu 2 sectors from simple colonies. (F) The distribution of repair events among random Ade + Leu 2 outcomes in simple and multisectored colonies calculated based on analyses presented in A-E. BIR or HC classes represent Ade + Leu 2 outcomes that have chromosome structure similar to BIR (or BIR-LR), but could also represent instances of HC co-segregation with an intact donor chromosome during mitosis (see Results and Materials and Methods for details). doi:10.1371/journal.pgen.1004119.g006 Array-CGH analysis shows a deletion (red) in Chr. III (between YCRCd6; 124250 bp position) and MAT (200142 bp position)) and a duplication (blue) of sequences in the same Chr. III (located centromere-distal to YCLWTy1-1, which corresponds to YCLWd15 in SGD (83110 bp position)). Underlined numbers in the schematic diagram (4 and 5) indicate the positions of probes used for Southern hybridization. The positions of EciI (Ec) and PshAI (Ps) restriction sites are indicated. As expected from the maps, the size of the DNA fragment corresponding to the HCC junction (H) was approximately 8 kb following digestion with EciI and 11 kb following digestion with PshAI. doi:10.1371/journal.pgen.1004119.g007 propose that the cycles of NRTs can be explained by initiation of BIR followed by its interruption leading to HCCs in tumor cells. Importantly, we suggest that initiation of HCC can be facilitated by checkpoint deficiency, which is frequent in cancer cells. [49,50]. In addition, our data suggest that cycles of HCCs could also contribute to clonal variations in pre-cancerous cells. Finally, our observation of increased HCs and HCCs resulting from the exposure of cells undergoing BIR to DNA damaging agents, could be of significant relevance to current cancer therapeutic strategies where anti-replication drugs are used in combination with agents that induce DSBs (for example, X-ray and gamma irradiation). We propose that DSBs induced by these agents may initiate BIR, which will frequently lead to NRT events in the presence of cancer drugs that inhibit replication. Because NRT events often initiate cascades of genetic instability, promoting such events in cancer cells could be one mechanism of rapid GCR formation that could result in negative oncotherapy outcomes such as secondary tumorigenesis and drug resistance.

Yeast strains
All yeast strains (Table S1) were isogenic to AM1003 [12] which is a chromosome III disome with the following genotype: hmlD::ADE1/hmlD::ADE3 MATa-LEU2-tel/MATa-inc hmrD::HPH FS2D::NAT/FS2 leu2/leu2-3,112 thr4 ura3-52 ade3::GAL::HO ade1 met13. In this strain, the HO endonuclease-induced DSBs introduced at MATa are predominantly repaired by BIR because the portion of the chromosome centromere-distal to MATa is truncated to leave only 46 bp of homology with the donor sequence [9,12]. The majority of single-gene deletion mutants were constructed by transformation with a PCR-derived KAN-MX module flanked by terminal sequences homologous to the sequences flanking the open reading frame of each gene [51]. The resulting constructs were confirmed by PCR and by phenotype. sgs1D mutants were constructed by PCR-amplification of sgs1DURA3 from the strain yWH239 [48] using oligonucleotides complementary to sequences flanking SGS1. To disrupt RAD50, plasmid pNKY83 [52] was digested simultaneously with BglII and EcoRI and transformed into recipient strains that were subsequently screened for a Ura + phenotype (rad50::hisG::URA3::hisG).

Media and growth conditions
Rich medium (yeast extract-peptone-dextrose [YEPD]) and synthetic complete medium, with bases and amino acids omitted as specified, were made as described [53]. YEP-lactate (YEP-Lac) and YEP-galactose (YEP-Gal) media contained 1% yeast extract and 2% Bacto peptone supplemented with 3.7% lactic acid (pH 5.5) or 2% (w/v) galactose, respectively. Yeast cultures were grown at 30uC or at 20uC (in the case of yeast strains bearing polymerase mutations, which rendered them temperaturesensitive). As indicated, MMS or 4-NQO (Sigma Aldrich) was added to rich medium for some experiments.

Kinetics of DSB repair
The kinetics of DSB repair was examined in time-course experiments as described previously [12]. YEP-Lac (500 to 1000 ml was inoculated with approximately 2610 6 cells/ml. Cultures were grown at 30uC overnight to reach a concentration of approximately 5610 6 cells/ml. HO endonuclease was induced by the addition of galactose to achieve a final concentration of 2%. For PFGE gel electrophoresis, 50 ml aliquots were removed, and sodium azide was added to achieve a concentration of 0.1% to stop DNA repair processes. Extraction of DNA embedded in 0.55% agarose plugs was performed as described [9]. For fluorescence-activated cell sorter (FACS) analyses, 5 ml aliquots were removed, cells were spun, diluted, fixed by the addition of 70% ethanol, and stored at 4uC. FACS analysis was performed using propidium iodide with a Becton Dickinson fluorescenceactivated cell analyzer, similar to [9]. PFGE was performed by running genomic DNA embedded in agarose plugs at 6 V/cm, for 40 hours (initial switch time 10 s; final switch time 35 s) followed by Southern blotting and hybridization using 32 P-labeled DNA probes containing either an ADE1 (SalI fragment from pJH879) (similar to [9,12]) or ADE3 (obtained by PCR amplification of chromosome VII from 907979-908735 bp) sequence. The images were analyzed using GE Healthcare Typhoon FLA 9500. The kinetics of accumulation of BIR product was measured using an ADE1-specific fragment as a probe. To account for variation in DNA loads, intensities of the bands corresponding to the intact chromosome III, as well as to the repaired chromosome III, were normalized to intensities of the bands corresponding to chromosome I, which also hybridizes to the ADE1-specific probe. The efficiency of BIR repair, presented as the percentage of truncated chromosome III that was converted to BIR product, was calculated by dividing the normalized intensity of a repair band by the normalized intensity of uncut, truncated chromosome III. Results of three time-course experiments were used to calculate the average 6 SD BIR efficiency for each strain. Cell viability following exposure to the DNA damaging agents was determined as a ratio of the number of colony forming units (CFU) observed experimentally and the number that was predicted based on the cell concentration determined using hemocytometer prior to plating. The viability of cells following treatment with DNA damaging agents was as follows:

Analyses of distribution of DSB repair events
To monitor the repair of HO-induced DSBs in individual colonies, we harvested logarithmically growing cells grown in YEP-Lac at 30uC and plated them on YEP-Gal. The resulting colonies were then replica plated onto omission media to examine the ADE1, ADE3, LEU2, and NAT markers. When temperature-sensitive strains bearing pol3-t or pol1-1 polymerase mutations were used, the cells were grown in YEP-Lac at 20uC. Following plating on YEP-Gal, the cells were incubated at 30uC for 24 hours (a length of time sufficient to complete BIR), and then incubated at 20uC until the colonies were full-grown. To test the effect of DNA damage, the cells were grown to log phase in YEP-Lac medium at 30uC, incubated in galactose-containing media for 30 minutes (to induce DSB), and then incubated with or without a DNA damaging agent (MMS or 4-NQO) for 7 hours. (The time for incubation with DNA damaging agents was selected based on the known kinetics of BIR [12]). The DNA damaging agents were then deactivated by treatment with 10% sodium thiosulfate prior to serial dilution and plating of cells onto YEPD. Repair events were identified by a phenotypic analysis after replica plating onto omission media, and also by PFGE. Gene conversion (GC) outcomes displayed an Ade + Leu + phenotype and contained two copies of chromosome III: a 356 kb chromosome that hybridized to an ADE3-specific probe and a short (217 kb) chromosome that hybridized to an ADE1specific probe (Fig. 1C). The absence of repair led to CL, which was detected by formation of Ade 2r Leu 2 colonies containing a single, 356 kb chromosome III, which hybridized to the ADE3-specific probe (Fig. 1D). Formation of Ade 2w Leu 2 colonies or colony sectors indicated formation of HCs. These cells contained a single, 346 kb chromosome III that hybridized to the ADE1-specific probe (Fig. 1E). Ade + Leu 2 phenotypes could result from several repair outcomes: BIR, HC (when it co-segregates with an intact copy of the donor chromosome during mitosis [12]), from CRs [21] or from HCC events. CRs and HCC events were identified by PFGE. CRs carried a 356 kb band that hybridized to the ADE3-specific probe and a band of any size (different from 346 kb) that hybridized to ADE1 (Fig. 1F). HCC contained a single, 346 kb band that hybridized to an ADE1-specific probe in addition to a band of varying size that hybridized to ADE3 (Fig. 1G). BIR usually carried a 356 kb band that hybridized to ADE3 and a 346 kb band that hybridized to ADE1 (Fig. 1B). Occasionally, BIR resulted from long 59-39 resection beyond NAT, followed by strand invasion that occurred centromere proximal to NAT. This led to the formation of BIR outcomes that were Ade + Nat 2 Leu 2 (BIR-LR) with donor and recipient chromosomes of identical (356 kb) sizes (Fig. 5C). PFGE could not distinguish between BIR and events where a HC cosegregated with an intact donor chromosome. Therefore, we assumed the number of Ade + Leu 2 HCs to be equal to the number of Ade 2w Leu 2 HCs based on the idea that an HC product should co-segregate with an intact copy of the donor chromosome in half of the cases of HC formation. Overall, the formula to calculate the number of BIR events was as follows: BIR = (number of Ade + Leu 2 )2(GCR+HCC+HC).
The distribution of various types of repair among all repair events was determined differently for simple colonies (containing ,3 repair sectors) and for multi-sectored colonies (containing $3 repair sectors). The frequency among simple colonies was determined as previously described in [9]. The frequency of each repair outcome in multi-sectored colonies was determined as the sum of all sectors belonging to this phenotypic class divided by the total number of sectors analyzed.
Importantly, since yeast strains used in this study did not clump during plating, we proposed that multi-sectored colonies originated from single cells, where DSB repair was preceded by mitotic divisions. This idea was also supported by our experiments where individual checkpoint-deficient cells were micro-manipulated on galactose containing plates, which led to the formation of multisectored colonies (data not shown).
In total, the following number of colonies were scored in experiments aimed to determine the effect of defective polymerases on half-crossover formation for each subsequent strain: Pol + (wt) -1192 colonies; pol3Y-708A -2428 colonies; pol3-t -1240 colonies; pol31-WRRGW -776 colonies; pol2-Y831A -2491 colonies; pol2-1 -896 colonies; and pol1-1 -287 colonies. In addition, the following number of colonies were scored in experiments aimed to determine the effect of DNA damage on half-crossover formation for each subsequent condition: DSB+no damage: -2583 colonies;

Array-CGH analysis of individual HCC events
The CGH analysis was conducted as described recently [54]. Briefly, genomic DNA was prepared from the agarose-embedded full length chromosome material. DNA from the parental strain was labeled with dUTP-Cy3 and DNA from the derivative strains carrying genome rearrangements was labeled with dUTP-Cy5. The labeled DNAs were mixed and competitively hybridized to custom Agilent 60-mer oligonucleotide microarrays. The arrays were scanned, the images were analyzed, and the CNV regions were identified using GenePix 6.0 and Nexus Copy Number software, respectively.

Southern analysis of HCC outcomes
For Southern analysis, genomic DNA of DSB repair outcomes was purified by glass bead/phenol method as described [55], digested with the appropriate restriction enzymes, and the resulting fragments were separated on a 0.8% agarose gel. Southern blotting was carried out by standard procedures using 32 P-labeled DNA probes that were generated by PCR amplification using 24 bp primers (available as Table S3) and genomic DNA of AM1003 as a template. The locations of these probes on chromosome III are as follows: (1)

Statistical analysis
All mutants were analyzed for their effect on BIR repair in at least three independent plating experiments. Results from these independent experiments were pooled if it was determined that the distributions of all events were statistically similar to each other using a Chi-square test. The effects of individual mutations on DSB repair were determined by comparing the resulting pooled distributions of repair outcomes obtained for mutants to the distribution obtained for the wild-type strain (AM1003) by Chi-square tests. Specifically, to determine the effect of various mutations on the frequency of HC, all repair outcomes were divided into two groups: HC (Ade w Leu 2 outcomes) and others (combining all other groups). Comparison of the distributions between these two classes in specific mutants vs. wild type was used to determine whether a mutation affected the frequency of HC. The effect of mutations on other DSB repair outcomes and the effects of exposure to various DNA damaging agents were determined similarly. Figure S1 The effect of polymerase defects on the distribution of DSB repair outcomes. The fraction (%) of various DSB repair outcomes were determined similarly to [12] following DSB induction in strains containing different mutations affecting Pold, Pole or Pola. For each strain, the data are based on 3-14 independent experiments. Asterisks indicate a statistically significant change as compared to wild type (Pol + ) cells. The efficiency of BIR was reduced as compared to wild type in pol3Y-708A, pol3-t, pol31-WRRGW, pol2-1, pol1-1, pol32D (P,0.0001) and in pol2-Y831A (P = 0.0005)). The frequency of chromosome loss (CL) was increased in pol3Y-708A, pol3-t, pol31-WRRGW, pol2-1, pol2-Y831A, and pol32D (P,0.0001). The frequency of half-crossovers (HC) was increased in pol3Y-708A, pol3-t, pol31-WRRGW, and pol32D (P,0.0001). Note: While this figure shows the fraction of HCs among all DSB repair events, Fig. 2 presents the fraction of colonies that are fully or partially HCs. (TIF) Figure S2 The effect of DNA damaging agents on the distribution of DSB repair outcomes. The fraction (%) of various DSB repair outcomes was determined similarly to [12] following DSB induction in AM1003 and its repair in the presence of MMS or 4-NQO. The data are based on 3 to 4 independent experiments for different conditions. A statistically significant difference from the distribution of DSB repair events in the absence of DNA damage is indicated by asterisk. The frequency of HC was significantly increased following exposure to MMS and 4-NQO (P,0.0001). For no-DSB control, the experiments were performed similarly to [12], but without addition of galactose. Asterisks indicate a statistically significant change as compared to the nodamage control. Note: While this figure shows the fraction of HCs among all DSB repair events, Fig. 2 presents the fraction of colonies that are fully or partially HCs. (TIF) Figure S3 Structural analysis of Ade + Leu 2 DSB repair outcomes in pol3-t mutants. PFGE analysis of Ade + Leu 2 DSB repair events obtained from multi-sectored colonies in pol3-t mutants. Ethidium bromide-stained gel (top), Southern blot analysis of PFGE gel using ADE3-specific (middle) and ADE1-specific Table S1 Strain list. The list of strains of yeast Saccharomyces cerevisiae that were used in this study.

Table S2
The analysis of HCC outcomes in rad24D cells. Catalog of half-crossover-initiated cascade DSB repair outcomes that were identified in rad24D and characterized by array-CGH and PFGE. BFB: breakage-fusion-bridge cycle. (TIF)

Table S3
The list of primers used to prepare hybridization probes. The 59-39 sequences of primers use to prepare hybridization probes by PCR are presented. (TIF) Text S1 Supplemental Results and Discussion. Description of HCC outcomes that were identified in rad24D and characterized by array-CGH and PFGE. The discussion of the mechanisms of their formation. (DOCX)