Genes in the RAD52 epistasis group are required for the formation of duplication-mediated GCRs in mms21-CH mutant strains

We previously showed that the mms21-CH mutation caused a substantial accumulation of GCRs selected in the duplication-mediated GCR (dGCR) assay (also called the yel072w::CAN1/URA3 assay) [4]. In the dGCR assay, non-allelic HR between divergent homologous sequences on chromosome V and chromosomes IV, X, or XIV resulting in the formation of translocations dominate the GCRs selected in most HR-proficient strains [22,26] (S1 Fig). In contrast, the mms21-CH mutation caused only a modest increase in GCR rates in the unique sequence-mediated (uGCR) assay (also called the yel068c::CAN1/URA3 assay) [4], which primarily selects for GCRs mediated by terminal deletions healed by de novo telomere additions and various types of micro- and non-homology mediated translocations [27] (S1 Fig). To explore this further, we combined the mms21-CH mutation with mutations affecting individual genes in the RAD52 epistasis group in strains containing the dGCR assay or the uGCR assay. We then performed fluctuation analysis to measure the GCR rates of these single and double mutant strains (Fig 1 and S1 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. dGCR and uGCR rates caused by mutations of genes in the RAD52 epistasis group and the mms21-CH mutation, a SUMO E3 ligase-null allele.

The number above each bar indicates the fold change normalized to the uGCR rate of wild-type strain. Detailed results used to generate the bar graph are shown in S1 Table.

https://doi.org/10.1371/journal.pgen.1007250.g001

This analysis of the RAD52 epistasis group of genes uncovered three main classes of genetic interactions. Class I mutations included deletions of the RAD51, RAD52, RAD54 and RAD55 genes required for HR [28]. In this case, deletion of each gene caused a drastic reduction of the increased dGCR rate caused by an mms21-CH mutation, indicating a requirement for HR in the formation of GCRs selected in the dGCR assay. Two of the Class I mutations, rad51Δ and rad55Δ, did not cause an increased uGCR rate when combined with the mms21-CH mutation, whereas two of the class I mutations, rad52Δ and rad54Δ (as well as a rad59Δ mutation; see below), caused an increased uGCR rate when combined with the mms21-CH mutation; this is consistent with previous observations that some HR pathways suppress GCRs selected in single copy sequence-mediated GCR assays such as the uGCR assay, presumably by promoting sister chromatid HR [22,29]. In contrast, deletion of RDH54, which encodes a Rad54 paralog with a role in meiotic HR [30], had little effect on the accumulation of GCRs in the mms21-CH mutant.

Class II mutations included deletions of RAD59 and CSM2. Class II mutations partially suppressed the increased GCR rate caused by the mm21-CH mutation in the dGCR assay, but caused an increased GCR rate in the uGCR assay when combined with the mms21-CH mutation. Rad59 is a stimulatory factor for Rad52 and is important for HR involving shorter repeats or when Rad52 is absent [28]. Csm2 is a subunit of the Shu complex [31], which has been implicated as a regulator of HR, possibly by facilitating the formation of Rad51 filaments [28]; other Shu complex mutations were not tested. Consistent with these accessory roles in HR, deletions of RAD59 and CSM2 in the mms21-CH mutant modestly reduced the rate of accumulating GCRs in the dGCR assay (Fig 1, upper panel) and substantially increased the rate of accumulating GCRs in the uGCR assay in the mms21-CH mutant (Fig 1, lower panel).

Class III mutations included mutations in MRE11 and SAE2. Class III mutations caused a modest increase in the increased dGCR rate caused by the mms21-CH mutation, but caused a substantial increase in the uGCR rate when combined with the mms21-CH mutation. The Mre11-Rad50-Xrs2 complex, together with Sae2, performs nucleolytic processing of DNA DSBs, leading to 5’-resection at DSBs and an ordered recruitment of HR proteins [28]. Deletion of MRE11 has been shown to cause substantial increases in the rate of accumulation of GCRs [22], and deletion of MRE11 in combination with the mms21-CH mutation caused an increase in the rate of accumulating GCRs in both the dGCR and uGCR assays compared to the mms21-CH single mutant (Fig 1). A mutation inactivating the endonuclease activity of Mre11, mre11-H125N, alone caused a 30-fold increase and a 122-fold increase in the rate of accumulating GCRs in the dGCR and uGCR assays, respectively (Fig 1). Interestingly, the mre11-H125N mutation did not appreciably affect the dGCR rate of the mms21-CH mutant, but caused a further increase in the uGCR rate of the mms21-CH mutant, suggesting the involvement of the Mre11 endonuclease activity in suppressing the GCRs selected in the uGCR assay. Sae2 participates in DNA DSB processing by specifically stimulating Mre11 endonuclease activity [32,33]. Like the mre11-H125N mutation, deletion of SAE2 only modestly increased the dGCR rate of the mms21-CH mutant, but caused a much larger increase in the uGCR rate of the mms21-CH mutant. Thus, the initial nucleolytic processing by Mre11 endonuclease has a critical role in suppressing the formation of the GCRs selected in the uGCR assay in the mms21-CH mutant. In contrast, deletion of EXO1, which eliminates a key exonuclease that participates in long-range resection of DNA breaks, had little effect on the rate of accumulating GCRs selected in either the dGCR or uGCR assays in the mms21-CH mutant (Fig 1).

Structures of GCRs formed in the wild-type strain and the mre11 and mms21-CH mutant strains

To gain further insight into the effects of the loss of MRE11 and MMS21 function, we investigated the structures of the GCRs selected in the wild-type strain and the mms21-CH, mre11Δ, mre11-H125N, mms21-CH mre11Δ, and mms21-CH mre11-H125N mutant strains. We focused on GCRs selected in the uGCR assay, as the GCRs selected in the dGCR assay are almost exclusively duplication-mediated translocations formed by non-allelic HR between the DSF1-HXT13 segmental duplication on chromosome V and regions of divergent homology on chromosomes IV, X and XIV, consistent with the HR gene dependency observed for GCRs selected in the dGCR assay in the mms21-CH mutant (S1 Fig). We first characterized the GCRs by testing the individual independent GCR-containing isolates for retention of the telomeric hygromycin resistance marker hph located between the telomere and the counter-selectable CAN1/URA3 cassette on the uGCR assay chromosome and by determining the size of the rearranged chromosome V by Pulse Field Gel Electrophoresis (Fig 2; S2 Table). GCRs were divided into three groups: rearranged chromosomes that were larger than the wild-type chromosome V (group 1) and chromosomes that were similar to or slightly shorter than the wild-type chromosome V and either lost (group 2) or retained (group 3) the telomeric hph marker. We classified GCRs in group 2 as de novo telomere addition GCRs, which are formed by the healing of broken chromosomes by the de novo addition of a new telomere [34]. De novo telomere additions are the predominant form of GCRs selected in uGCR assays in strains without telomerase defects [29,35,36] and are always associated with loss of the hph marker [22], although it should be noted that rare interstitial deletion GCRs can be associated with deletion of the hph marker. Similarly, we classified GCRs in group 3 as interstitial deletion GCRs, in which the deletion is typically associated with non-homology or microhomology breakpoint junctions when selected in GCR assays containing only unique sequences in the breakpoint region like the uGCR assay used here [37].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Analysis of the size of the rearranged chromosome V selected in wild-type, mms21-CH, mre11Δ, mre11-H125N, mms21-CH mre11Δ, and mms21-CH mre11-H125N uGCR strains.

Chromosomes were separated by Pulsed-Field Gel Electrophoresis (PFGE) and Southern blotted using a probe for the essential chromosome V gene MCM3. The size of chromosome V in the parental strain (P) is indicated by an arrow. The retention (+) or loss (-) of the hph marker inserted next to the telomere on the left end of chromosome V is indicated above each lane in the gel. Isolates selected for whole genome sequencing are indicated with their WGS data set name above each lane.

https://doi.org/10.1371/journal.pgen.1007250.g002

Strains containing GCRs falling into group 1 were subjected to whole genome paired end sequencing to determine the structures of the GCRs present (S3 Table). In addition to being able to detect all of the mutations and chromosome modifications introduced into the starting strains during strain construction (S2 and S3 Figs), we were also able to extensively characterize the structures of the GCR-containing chromosomes (Fig 3, S4–S9 Figs and S4 Table). We observed two distinct types of group 1 GCRs: microhomology-mediated translocations and hairpin-mediated inverted duplications.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Types of GCRs identified by whole genome sequencing.

Analysis of a microhomology-mediated translocation (A) and hairpin-mediated inverted duplications that were resolved by homologous recombination between Ty-derived elements (B), the homology between URA3 and ura3-52 (C), and a microhomology-mediated translocation (D). Left. Copy number analysis of uniquely mapping regions along a portion of the left arm of the assay-containing chromosome V derived from the read depth of uniquely mapping regions of the genome. Middle. Copy number analysis that, if present, shows other copy number changes elsewhere in the genome. Left and Middle. The path describing the GCR-containing chromosome is illustrated by the hashed thick blue line; the thin dashed blue lines indicate the connectivity between individual fragments that are separated on the reference genome. Homology-mediated translocation junctions are depicted with filled in triangles that point in the direction in which homology element points; junctions involving Ty-related homologies are red and other homologies are blue. Non-homology or micro-homology translocations are shown using two chevrons. Telomeres associated with the GCR (if known) are shown as a series of black vertical lines. Right. Sequences of any novel junctions are with the central line in the alignment corresponding to the novel junction. Sequences at the junction that could have been derived from either sequence are surrounded with colons. For GCRs with hairpin-mediated inversions, the inferred structure of the hairpin intermediate is also shown.

https://doi.org/10.1371/journal.pgen.1007250.g003

In microhomology-mediated translocations, the broken end of a broken chromosome V is fused to another broken chromosome such that the broken chromosome V acquires a fragment of the second broken chromosome that is terminated with a telomere (Fig 3a). Copy number analysis indicated that these fusion events duplicated the non-chromosome V target, and junction sequences revealed only short sequences of identity at the translocation junctions. The copy number analysis was also consistent with the presence on an intact copy of the target chromosome, indicating that the microhomology-mediated translocations were non-reciprocal.

In hairpin-mediated inverted duplications, the broken end of a broken chromosome V is fused to an inverted copy of itself on the left arm of chromosome V at a position between the CAN1/URA3 cassette and the first centromeric essential gene (Fig 3b, 3c and 3d). The inversion site sequences are consistent with a mechanism in which a broken chromosome V is resected to form a 3’ overhang that then pairs with a short stretch of homologous sequence centromeric to the breakpoint that is processed to yield a hairpin-terminated chromosome followed by replication of the hairpin-terminated chromosome (S10 and S11 Figs). As previously observed [36], these inverted dicentric duplication chromosomes (also called isoduplications) all underwent additional rounds of rearrangement that resolved them to the monocentric translocations observed, although other mechanisms for the hairpin formation and resolution are possible. These secondary rearrangements often, but not always, involved HR between the Ty- or PAU gene-related sequences on chromosome V L and a homology elsewhere in the genome (Fig 3b and 3c; S12–S14 Figs). The secondary rearrangements that initially appeared to involve HR between ura3-52 on chromosome V and YLRCdelta21 on chromosome XII actually proved to target an adjacent full-length Ty element on chromosome XII that was not present in the reference sequence (S15 Fig); this full-length Ty element has been previously observed by others [38,39]. A specific secondary rearrangement between a URA3 fragment in the Ty-inactivated ura3-52 on chromosome V L and the URA3 in the yel068c::CAN1/URA3 cassette first observed in GCRs derived from the tel1Δ uGCR strain was also observed here [36]. An additional type of secondary rearrangement was mediated by microhomologies (Fig 3d); microhomology-mediated secondary rearrangements were not observed in GCRs selected in tel1Δ mutants [36]. In most cases, the hairpin-mediated inverted duplications underwent a single secondary rearrangement as described above; however, in a small number of cases multiple rounds of secondary rearrangements were observed leading to the formation of monocentric GCRs (S4 Table).

Distribution of GCRs formed in different mutant strains

GCRs selected in the wild-type uGCR strain were primarily de novo telomere addition GCRs (Fig 4a; S7 Fig), consistent with dominance of de novo telomere addition GCRs among the GCRs selected in the “classical” GCR assay [35,40], which lacks large repetitive sequences in the breakpoint region like the uGCR assay used here. In addition, two interstitial deletions and two hairpin-mediated inverted duplications that were resolved by HR between the ura3-52 allele and the URA3 gene on the terminal chromosome V telomere-containing fragment were recovered. The spectrum of GCRs obtained from the mms21-CH uGCR strain shared this bias towards the formation of de novo telomere addition GCRs, with the other GCRs recovered being translocations involving other chromosomes (Fig 4a; S5 Fig). In contrast, substantially increased numbers of hairpin-mediated inverted duplications and decreased numbers of de novo telomere addition GCRs were selected in the mre11Δ and mre11-H125N single mutant uGCR strains and the mms21-CH mre11Δ and mms21-CH mre11-H125N double mutant uGCR strains (Fig 4a and 4b; S5–S9 Figs). These observations were consistent with role of the Mre11-Rad50-Xrs2 complex in cleaving hairpin structures [33] and the recovery of hairpin-mediated GCRs in strains containing an mre11-H125N mutation [41]. Remarkably, MRE11-deficient strains showed a bias for selection of translocations containing a copy of a long region of chromosome XII R (Fig 4c; S16 Fig), which could reflect either a bias due to increased fragility or accessibility of chromosome XII or due to suppression of mre11-dependent growth defects by duplication of chromosome XII R. We also observed that 8 of the 10 sequenced mms21-CH mre11Δ GCR-containing isolates were disomic for chromosome VIII and 1 of the 10 was disomic for chromosome I (S2 Table, S17 Fig). Taken together, these data are consistent with the idea that the mms21-CH mutation increases the total level of DNA damage without substantially biasing the mechanisms involved in forming GCRs, whereas mre11 defects increase the propensity of damaged DNAs to form hairpin inversions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Distribution of GCR types from each mutant analyzed.

A. Table of the number of each type of rearrangement, de novo telomere addition, interstitial deletion, hairpin-mediated inverted duplication, micro- and non-homology-mediated translocations, and homology-mediated translocations. B. The observed rates (purple) of accumulating de novo telomere addition GCRs and hairpin inversion GCRs were calculated by multiplying the observed rate and 95% confidence intervals by the fraction of GCRs in each mutant corresponding to each type of GCR. The expected rates (green) were calculated by multiplying the rates by the fraction of GCRs in the wild-type strain. Statistically significant differences between the observed and expected rates were calculated by the Mann-Whitney test; p-values < 0.05 displayed as ‘*’ and p-values < 0.0005 displayed as ‘**’. C. Venn diagram of sequenced GCRs demonstrates that strains with MRE11 defects tend to accumulate GCRs involving inverted duplications and/or chromosome XII.

https://doi.org/10.1371/journal.pgen.1007250.g004

Roles of Pol32 and DNA helicases in the accumulation of GCRs in mms21-CH mutant strains

The dramatic HR-dependent increase in the dGCR rate caused by the mms21-CH mutation (Fig 1), combined with the fact that the mms21-CH mutation caused at best modest changes in the spectrum of GCRs selected in the uGCR assay (Fig 4 and S5 Table), suggested that the mms21-CH mutation causes an increase in DNA damage that underlies the formation of GCRs without dramatically affecting the DNA repair pathways that act on this DNA damage. Because HR appears to act on this DNA damage to produce GCRs selected in the dGCR assay, we investigated whether BIR or a BIR-related pathway might play a role in the formation of GCRs in mms21-CH dGCR strains.

Previous studies of the repair of HO endonuclease-induced DNA DSBs by BIR showed that Pol32, a subunit of DNA polymerases delta and zeta, is required for BIR [42–44]. However, other studies have found that the pol32Δ mutation only reduced the efficiency of BIR [45]. We previously found that a pol32Δ mutation did not decrease the wild-type dGCR rate nor did the pol32Δ mutation eliminate duplication-mediated GCRs [22], suggesting that the role of POL32 in promoting BIR may be dependent on the nature of the initiating damage or that the GCRs selected in the dGCR assay are not formed by BIR. Remarkably, we found that deletion of POL32 in the mms21-CH mutant caused a drastic reduction of the dGCR rate by about 15-fold and a relatively modest increase in its uGCR rate (Fig 5A), consistent with an important role of POL32-dependent BIR in forming dGCRs in the mms21-CH mutant.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Role of Pol32 and DNA helicases in the formation of GCRs caused by mms21-CH.

A) dGCR and uGCR rates caused by deleting POL32, PIF1, RRM3, MPH1, SRS2 and SGS1 in wild-type and mms21-CH mutant strains. B) Effect of combining sgs1Δ and pol32Δ, rad52Δ or mre11-H125N on the rate of accumulating GCRs. C) Effects of sgs1-3KR (K175R, K621R and K831R) mutations on the rate of accumulating GCRs. The number above each bar indicates the fold change normalized to the uGCR rate of wild-type strain. Detailed results used to generate the bar graph are shown in S5 Table. D) Combining mms21-CH with rad52Δ, sgs1Δ or pol32Δ causes a more severe growth defect in the double mutants. Equal amounts of cells, as measured by cell density, were 10-fold serially diluted, plated on YPD plate and incubated at 30 degree for 2 days before the picture was taken.

https://doi.org/10.1371/journal.pgen.1007250.g005

Pif1 has been shown to be required for BIR initiated by HO endonuclease-induced DSBs and is thought to act by promoting DNA synthesis mediated by a migrating D-loop replication intermediate [42,43]. Pif1 also dissociates telomerase from single-stranded DNA thereby suppressing GCRs mediated by de novo telomere addition at DSBs [29,46]. We found that deleting PIF1 in the mms21-CH mutant caused further increases in both the dGCR and uGCR rates relative to that of the respective single mutants (Fig 5A), which is consistent with the idea that the DNA damage that underlies the formation of GCRs in the mms21-CH mutant is a substrate for de novo telomere additions [34,36].

We also screened other DNA helicases for their role in forming GCRs in mms21-CH mutant strains. RRM3 encodes a DNA helicase that travels with DNA replication fork [47]. A recent study showed that Rrm3 participates in the repair of replication-associated DNA breaks [48], although it is not involved in BIR induced by HO endonuclease. Deletion of RRM3 in the mms21-CH mutant caused a reduction (43-fold) in the dGCR rate without appreciably affecting the uGCR rate compared to that of the respective single mutants (Fig 5A), indicating a requirement of Rrm3 in the formation of duplication-mediated GCRs in the mms21-CH mutant strains. These findings suggest that accumulation of duplication-mediated GCRs in mms21-CH might reflect the formation of replication-associated DNA DSBs that require Rrm3 for BIR-like repair.

The DNA helicase Srs2 acts as an anti-recombinase by disrupting the formation of Rad51 filaments and D-loops [49–51]. In addition, the Smc5/6 complex of which Mms21 is a subunit has been shown to control the recombination activity of the Mph1 helicase [12,52]. Deletion of SRS2 or MPH1 in the mms21-CH mutant did not appreciably alter the dGCR rate, but caused a drastic increase in the uGCR rate relative to that of the respective single mutants (Fig 5A). This latter result could be explained if Srs2 and Mph1 either suppress the formation of a critical intermediate in the formation of the GCRs selected in the uGCR assay or target the initiating DNA damage to sister chromatid HR to an extent that suppresses GCRs selected in the uGCR assay, but not those selected in the dGCR assay.

The Sgs1 helicase has a major role in specifically suppressing dGCRs [16,22], and this has been attributed to its role in preventing crossovers during the resolution of HR intermediates [53]. Interestingly, combining an sgs1Δ with the mms21-CH mutation resulted in synergistic increases in both dGCR and uGCR rates relative to the respective single mutants (Fig 5A), indicating Mms21 and Sgs1 function in distinct pathways to prevent the formation of GCRs. To explore this further, we analyzed the effects of mutating RAD52, MRE11 and POL32 in the sgs1Δ mutant. A deletion of RAD52 and the mre11-H125N mutation caused similar effects in sgs1Δ and mms21-CH mutants (comparing Figs 1 and 5). In contrast, deletion of POL32 caused an increase in the dGCR rate of the sgs1Δ mutant whereas deletion of POL32 in the mms21-CH mutant reduced the dGCR rate more than 10-fold (Fig 5B). Thus, the formation of duplication-mediated GCRs in the mms21-CH and sgs1Δ mutants had distinctly different requirements for Pol32. Because deletion of RRM3 is lethal in an sgs1Δ mutant [54], we could not compare the role of Rrm3 in the mms21-CH and sgs1Δ mutants.

To determine whether the helicase activity or the Top3-binding activity of Sgs1 is involved in the suppressing of dGCRs, we analyzed sgs1-K706A (helicase-dead) and sgs1-E12G, H13S (Top3-binding defective) mutants [55]. We found that both sgs1 mutations caused approximately the same increase in GCR rates compared to that caused by the sgs1Δ mutation, indicating that the function of Sgs1 in suppressing GCRs requires both its helicase activity and interaction with Top3-Rmi1 (Fig 5C). Recent studies showed that Mms21 specifically catalyzes sumoylation of Sgs1 in response to treatment with DNA alkylating agents [10,20,21]. We found that the sgs1-3KR mutation that eliminates the sumoylation sites on Sgs1 did not cause a comparable increase in GCR rates to that seen in the sgs1Δ mutant (Fig 5C). Although we cannot exclude the possibility that a low and undetectable level of Sgs1 sumoylation occurs in the sgs1-3KR mutant, this result indicates that the major DNA damage-induced sumoylation of Sgs1 does not have an appreciable role in preventing spontaneous GCRs.

Because the DNA lesions accumulated in the mms21-CH mutant are repaired by a variety of DNA repair pathways, especially the HR and more specifically BIR pathway, we reason that inactivating these DNA repair pathways in the mms21-CH mutant could cause a synergistic growth defect. Consistent with this view, we found that mutations of rad52Δ, sgs1Δ and pol32Δ caused a significant slower growth when they are combined with the mms21-CH mutation (Fig 5D). Thus, a failure in properly repairing the spontaneous DNA lesions accumulated in the mms21-CH mutant is detrimental to cell growth.

The mms21-CH mutation induces spontaneous DNA lesions and activates the DNA damage checkpoint to promote dGCRs

The above findings are consistent with the idea that the mms21-CH mutation causes accumulation of spontaneous DNA lesions that underlie the formation of a diverse range of GCRs. Cells have evolved a signal transduction pathway, the DNA damage checkpoint, to detect endogenous DNA lesions [56]. We therefore examined the Rad53 kinase, which becomes hyperphosphorylated in the presence of such DNA damage and migrates with a slower electrophoretic mobility than non-phosphorylated Rad53. Treatment by the DNA alkylating agent methyl methane sulfonate (MMS) caused a pronounced electrophoretic mobility shift of fully activated Rad53 to slower migrating species (Fig 6A). An increased but sub-stoichiometric amount of Rad53 was found to show slower gel mobility in an mms21-CH mutant that was not treated with MMS compared to untreated wild-type, indicating that Rad53 is partially activated in mms21-CH mutants. Deletion of RAD9, which encodes an adaptor protein that acts to promote DNA damage-induced activation of Rad53, reduced the amount of the slower migrating species of Rad53 to an undetectable level in the mms21-CH mutant that was not treated with MMS; note that MMS-induced Rad53 phosphorylation still occurs in a rad9Δ mutant due to the redundant role of Mrc1 in mediating Rad53 activation [57–59]. Ddc2, together with the Mec1 kinase, is recruited to RPA-coated single stranded DNA at the sites of DNA damage where it can be visualized as sub-nuclear foci [60]. A higher incidence of Ddc2 foci was seen in the untreated mms21-CH mutant compared to untreated wild-type cells (Fig 6B). We also examined the localization of Rad52, which forms foci at the sites of DNA lesions that undergo DNA repair by HR [60]. Similar to Ddc2 foci, an elevated incidence of Rad52 foci was detected in the mms21-CH mutant compared to wild-type cells (Fig 6B). Together, these results suggest that elevated levels of endogenous DNA lesions occur in the mms21-CH mutant that activate the DNA damage checkpoint and are and processed by Rad52-mediated HR.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Role of DNA damage checkpoint in the formation of GCRs.

A) Rad53 gel shift assay to examine Rad53 activation in WT, mms21-CH, rad9Δ and rad9Δ mms21-CH mutants. B) Spontaneous Ddc2 foci in mms21-CH mutant and WT, which also contain Nup49-mCherry to mark the nuclear envelope. Bar graph indicates the percentage of Ddc2-foci within the nuclear envelope (marked by Nup49-mCherry). Error bars represent the standard deviation from three replicate experiments using two biological isolates per strain per replicate. 200–400 cells were imaged and counted for each experiment. C) dGCR and uGCR rates caused by mutations of DNA damage checkpoint genes with or without mms21-CH. The number above each bar indicates the fold change normalized to the uGCR rate of wild-type strain. Detailed results used to generate the bar graph are shown in S6 Table.

https://doi.org/10.1371/journal.pgen.1007250.g006

We next asked whether the DNA damage checkpoint influences the formation of GCRs in the mms21-CH mutant (Fig 6C and S6 Table). The DNA damage checkpoint involves two partially redundant protein kinases, Mec1 and Tel1. While Mec1 has a major role in controlling the DNA damage response, Tel1 has an important role in telomere length maintenance in wild-type cells and in checkpoint responses in mec1 mutants [61,62]. We found that deletion of MEC1 caused a 5-fold reduction in the dGCR rate of the mms21-CH mutant (Fig 6C, upper panel) and a substantial increase in the uGCR rate of the mms21-CH mutant (Fig 6C, lower panel). Unlike the mec1Δ mutation, deletion of TEL1 caused little of no change in the dGCR and uGCR rates of the mms21-CH mutant. Like the deletion of MEC1, deletions of the MEC3, RAD24 and RAD9 genes involved in the DNA damage checkpoint, caused varying degrees of reduction of the dGCR rate of the mms21-CH mutant with the rad9Δ mutation causing the greatest reduction (~ 24-fold). In contrast, deletions of the MEC3, RAD24 and RAD9 genes in the mms21-CH mutant resulted in increases in the uGCR rate, consistent with the known role of DNA damage checkpoint in suppressing GCRs mediated by single-copy sequences [63].

Unlike deletion of RAD9, deletion of MRC1 caused a modest (2-fold) increase in the dGCR rate of the mms21-CH mutant (Fig 6C, upper panel). Interestingly, deletion of MRC1 caused a drastic increase in the uGCR rate of the mms21-CH mutant (Fig 6C, lower panel). Because Mrc1 also has a role in DNA replication [64], we next examined the mrc1-AQ mutant, all of whose Mec1 consensus phosphorylation sites are mutated to non-phosphorylatable alanines and is thus unable to mediate Rad53 activation [65]. We found that the mrc1-aq mutation did not appreciably alter the dGCR rate of the mms21-CH mutant although it did cause an increase in the uGCR rate of the mms21-CH mutant, but not to the extent seen with the mrc1Δ mutation. Rad53, Chk1, and Dun1 are the downstream effector kinases of the checkpoint pathways. Deletion of RAD53 reduced the dGCR rate of the mms21-CH mutant by about 9-fold (Fig 6C, upper panel), while deletion of CHK1 did not appreciably alter the dGCR rate of the mms21-CH mutant and caused a small increase in the uGCR rate of the mms21-CH mutant. Although deletion of DUN1 did not appreciably alter the dGCR rate of the mms21-CH mutant, it caused a synergistic increase in the uGCR rate of the mms21-CH mutant (Fig 6C, lower panel). These data support the idea that the Mec3/Rad24-Rad9-Mec1-Rad53 pathway plays a role in promoting the GCRs selected in the dGCR assay in the mms21-CH mutant while suppressing the GCRs selected in the uGCR assay in the mms21-CH mutant.

Because the mms21-CH mutation appears to cause increased levels of some type of DNA damage, we tested the affect of the mms21-CH mutation on HR between ade2 heteroalleles in a mitotically growing diploid strain [66]. The rate of accumulating Ade+ recombinants in the wild-type strain was 5.0 x10^-5 [95% confidence interval = 1.8 x10-5–8.5 x10^-5] and the rate in the mms21-CH mutant was 3.2 x10^-5 [95% confidence interval = 2.0 x10-5–1.0 x10^-4] indicating that there was no change in the rate of HR between ade2 heteroalleles in the mms21-CH mutant. This is unlike the effect of the mms21-CH mutation on the formation of GCRs. General DNA damage such as γ-rays is known to induce HR between heteroalleles [67]. A possible explanation for the difference between the effect of mms21-CH on heteroallelic HR and the formation of GCRs is that the DNA damage that occurs in the mms21-CH mutant is only subjected to one-ended HR such as BIR. This damage, most likely a DSB, would rarely result in HR between heteroalleles as the damage would need to occur between the heteroalleles to yield recombinants, and this is a much smaller target than the region in which a DSB would yield a GCR [22].

S. cerevisiae strain construction and genetic methods

Standard S. cerevisiae genetics method was used to introduce mutations. S. cerevisiae strains used in this study are listed in S7 Table. Generation of the sgs1-3KR mutation involved two steps by first deleting the N-terminal region of Sgs1 from -425bp to 2600bp and then repairing it using PCR products containing the sgs1-3KR (K175R, K621R and K831R) mutations with a HIS3 marker located at 435bp upstream of the starting codon of Sgs1 to preserve its native promoter. DNA sequencing was used to confirm the integration of sgs1 mutations. Methods used for fluctuation analysis to determine GCR rates have been described previously [70]. Briefly, 14–16 rate measurements were performed using 7–8 independent cultures derived from 2 independently isolated strains and the GCR rate was determined by the method of the median. Fluctuation analysis was also used to determine the rate of HR between ade2 heteroalleles. Diploids were generated from wild-type and mms21-CH derivatives of previously published haploid strains containing different ade2 alleles [66], one due to an insertion of the I-SceI cut site (ade2-I) and another due to a 2bp deletion at the NdeI site (ade2-n). The I-SceI endonuclease was not expressed, allowing spontaneous HR to be measured. Cells were grown from single colonies overnight in YPD until cultures reach an OD₆₀₀ of 0.5 and plated onto CSM complete or CSM minus adenine medium and grown for 3–4 days before counting the resulting colonies.

Microscopy

Cells were grown in CSM medium to log phase and examined by live imaging using Olympus BX43 fluorescence microscope with a 60x, 1.42 PlanApo N Olympus Oil immersion objective. GFP and mCherry fluorescence were detected using a Chroma FITC filter set and a TxRed filter set respectively and captured with a Qimaging QIClick CCD camera. Images were captured using Meta Morph Advanced 7.7 imaging software. Figures were prepared in Adobe Photoshop, keeping processing parameters constant within each experiment.

PFGE and Southern blotting

DNA plugs for PFGE were prepared as described [71]. Electrophoresis was performed using a Bio-Rad CHEF-DRII apparatus at 6 V/cm, with a 60 to 120 s switch time for 25 h. The gels were stained with ethidium bromide and imaged. The DNA in the gel was transferred to Hybond-XL membranes by neutral capillary blotting. The DNA was crosslinked to the membrane by UV irradiation in a Stratalinker (Stratagene) apparatus at maximum output for 60 seconds. The MCM3 probe was generated by amplifying MCM3 from genomic DNA using the primers 5’-CTGTGCAAGAAATGCCCGAAATG-3’ and 5’-GCCCCGGAGTTGGAATGCTC-3’ followed by random primer labeling of the PCR product with the Biotin DecaLabel DNA Labeling Kit (Thermo Scientific). Probe hybridization was performed at 50°C for 1 hr. Biotin signal was detected using Chemiluminescent Nucleic Acid Detection Module Kit (Thermo Scientific).

Whole genome paired-end sequencing

Multiplexed paired-end libraries were constructed from 2 μg of genomic DNA purified using the Purgene kit (Qiagen). The genomic DNA was sheared using M220 focused-ultrasonicator (Covaris) and end-repaired using the End-it DNA End-repair kit (Epicentre Technologies). Common adaptors from the Multiplexing Sample Preparation Oligo Kit (Illumina) were then ligated to the genomic DNA fragments, and the fragments were then subjected to 18 cycles of amplification using the Library Amplification Readymix (KAPA Biosystems). The amplified products were fractionated on an agarose gel to select 600 bp fragments, which were subsequently sequenced on an Illumina HiSeq 4000 using the Illumina GAII sequencing procedure for paired-end short read sequencing. Reads from each read pair were mapped separately by bowtie version 2.2.1 [72] to a reference sequence that contained revision 64 of the S. cerevisiae S288c genome (http://www.yeastgenome.org), hisG from Samonella enterica, and the hphMX4 marker. Sequence data is available from National Center for Biotechnology Information Sequence Read Archive under accession number: SRP106876.

Rearrangement and copy number analysis of paired-end sequencing data

Chromosomal rearrangements were identified after bowtie mapping by version 0.6 of the Pyrus suite (http://www.sourceforge.net/p/pyrus-seq) [36]. Briefly, after removal of PCR duplicates, read pairs in which both reads uniquely mapped were used to generate the read depth and span depth copy number distributions. The read depth copy number distribution is the number of times each base pair was read in a sample; read depth distributions were the distributions plotted to examine copy number (S4–S9 Figs) as this distribution is less distorted than the span depth distribution in regions adjacent to repetitive elements. The span depth copy number distribution is the number of times each base pair in a sample was contained in a read or spanned by a pair of reads; span depth distributions were used to statistically distinguish real rearrangements identified by junction-defining discordant read pairs from discordant read pairs that were noise in the data. Read pair data were then analyzed for junction-defining discordant read pairs that indicated the presence of structural rearrangements relative to the reference genome. Identified rearrangements included junctions produced during strain construction, such as the his3Δ200 deletion (see S2 and S3 Figs), or GCR-related rearrangements (see S4–S9 Figs). Associated junction-sequencing reads, which were reads that did not map to the reference but were in read pairs in which one end was adjacent to discordant reads defining a junction, were used to sequence novel junctions. Most hairpin-generated junctions (S11 Fig) could be determined using alignments of junction-sequencing reads. For junctions formed by HR between short repetitive elements (S12–14 Figs) and for problematic hairpin-generated junctions (S11 Fig), the junction sequence could be derived by alignment of all reads in read pairs where one read was present in an “anchor” region adjacent to the junction of interest and the other read fell within the junction to be sequenced. Similar strategies involving the alignment of reads paired with reads present in “anchor” regions also were used to sequence de novo telomere addition junctions (S4–S9 Figs) and to identify the “YLRWTy1-4” Ty element that was not present in the reference genome (S15 Fig).

Rad53 gel shift analysis

Protein extracts for Western blot analysis was prepared using a TCA (trichloroacetic acid) extraction. To examine Rad53 electrophoretic mobility we used an anti-Rad53 monoclonal antibody (EL7E1 serum) from mouse, a gift from Dr. Marco Foiani.