Recombinase-independent chromosomal rearrangements between dispersed inverted repeats in Saccharomyces cerevisiae meiosis

Abstract DNA double-strand break (DSB) repair by homologous recombination (HR) uses a DNA template with similar sequence to restore genetic identity. Allelic DNA repair templates can be found on the sister chromatid or homologous chromosome. During meiotic recombination, DSBs preferentially repair from the homologous chromosome, with a proportion of HR events generating crossovers. Nevertheless, regions of similar DNA sequence exist throughout the genome, providing potential DNA repair templates. When DSB repair occurs at these non-allelic loci (termed ectopic recombination), chromosomal duplications, deletions and rearrangements can arise. Here, we characterize in detail ectopic recombination arising between a dispersed pair of inverted repeats in wild-type Saccharomyces cerevisiae at both a local and a chromosomal scale—the latter identified via gross chromosomal acentric and dicentric chromosome rearrangements. Mutation of the DNA damage checkpoint clamp loader Rad24 and the RecQ helicase Sgs1 causes an increase in ectopic recombination. Unexpectedly, additional mutation of the RecA orthologues Rad51 and Dmc1 alters—but does not abolish—the type of ectopic recombinants generated, revealing a novel class of inverted chromosomal rearrangement driven by the single-strand annealing pathway. These data provide important insights into the role of key DNA repair proteins in regulating DNA repair pathway and template choice during meiosis.


INTRODUCTION
DNA damage repair mechanisms exist to ensure genome stability following the numerous endogenous and exogenous insults that the cell is exposed to dail y. DN A doublestrand br eaks (DSBs), wher e both strands of the DNA helix are broken at the same locus, are particularly problematic. Failur e to r epair DSBs a ppropriatel y and accuratel y can lead to a range of mutation outcomes, from singlenucleotide changes to loss, gain and / or rearrangement of large chromosome regions. The two major pathways for repair of DSBs are non-homologous end joining (NHEJ) and homologous recombination (HR) ( 1 ). In NHEJ repair, both ends of the DSB are first processed by nucleases to remove DNA adducts or mismatches before being joined together by ligation, leaving the potential for small nucleotide deletions or additions ( 2 ). By contrast, during HR, DSB ends ar e r esected generating a r egion of single-stranded DNA that invades a homologous template used for repair ( 1 ). Consequentl y, HR typicall y generates error-free repair, although ina ppropriate homolo gous template choice during HR can lead to loss or gain of genetic information, observed as unequal crossing-over events.
Formation of DSBs leads to the activation of DNA damage checkpoint machinery, pre v enting the cell cy cle from progr essing until DSB r epair has occurr ed. Two major DNA damage checkpoint kinases function during meiosis, Mec1 and Tel1 (ATR and ATM orthologues, respecti v ely) ( 13 ). In addition to their conventional cell cycle regulation role, both Mec1 and Tel1 regula te DSB forma tion and repair during meiosis (14)(15)(16)(17). Specifically, Mec1 / Tel1 phosphorylation of the HORMA domain-containing protein Hop1 instigates a repair bias from the homologous chromosome ( 18 ). In doing so, DSB repair during meiosis leads to not only gene conversion events by non-crossover repair from the homologue, but also reciprocal exchanges of genetic information between homologous chromosomes when crossovers occur. Consequently, allelic distribution varies within the haploid gametes.
HR is reliant upon homologous repair templates being available, and the action of RecA orthologues to undertake strand invasion. While it is not known what criterion defines a suitable template, in vitro data suggest that the meiosisspecific RecA orthologue Dmc1 allows for a more imperfect template to be used in repair, supporting the use of the homologous chromosome for repair, which may have allelic differences ( 19 ). DNA sequences with full or partial homology to one another are found throughout genomes, enabling the potential for HR to take place between non-allelic sites, a process r eferr ed to as non-allelic homologous recombination (NAHR) or ectopic recombination. The insertion of the LEU2 gene from strain X2180-1a at the HIS4 locus in the S. cerevisiae SK1 strain background formed a meiotic DSB hotspot, enab ling the inv estiga tion of DSB forma tion and repair dynamics ( 20 ). Howe v er, at the endogenous locus the LEU2 gene was retained, albeit disrupted by insertion of the Salmonella typhimurium hisG gene, providing a perfect (or near-perfect) template for repair and the potential for ectopic recombination between HIS4 :: LEU2 and leu2 :: hisG .
Analyses of Southern blotting at the HIS4::LEU2 locus re v ealed the presence of ectopic recombinants in wild-type strains, a change in the type of ectopic recombinants formed in a dmc1 Δ background and an increase in ectopic recombinant formation observed in DNA damage checkpoint mutants ( 15 , 21-23 ). Subsequent work revealed multiple pathways through which ectopic recombination can occur, with the RecQ helicase Sgs1 ( 24 , 25 ), synaptonemal complex protein Zip1, and recombinase and accessory proteins Rad51 and Tid1 functioning to suppress ectopic recombination ( 22 , 23 ). Howe v er, pre vious analyses hav e only inv estigated a subset of the ectopic recombinants that form from one side of the recombination hotspot studied and described only the local changes arising at that site.
To investigate ectopic HR further, we have probed for additional recombinants at local regions, chromosome-wide and under conditions that allow us to investigate interhomologue / inter-sister bias in ectopic recombinant formation. Our data re v eal nov el ectopic recombinants, differing in frequency based upon the absence of DNA repair or checkpoint proteins, and the formation of acentric and dicentric gr oss chr omosomal rearrangements, e v en under wild-type conditions. In addition, our results identify DNA repair proteins that function to reduce ectopic recombinant formation, not only by suppressing classic NAHR involving strand invasion, but also --unexpectedly --by suppression of the single-strand annealing (SSA) repair pathway, an outcome enabled via the coincident formation of Spo11 DSBs within high-frequency hotspots adjacent to repetiti v e DNA elements.

DSB and ectopic recombinant analysis
DSB signals were detected using standard techniques by indirect end labelling of specific genomic loci after fractionation and transfer to Nylon membranes. For DSBs and ectopic products at HIS4 :: LEU2 , genomic DNA was digested with PstI, separated on 0.7% agarose in 1 × Tris-acetate-EDTA for ∼18 h at room temper ature, tr ansferred to Nylon membrane under denaturing conditions and then hybridized with a HIS4 LH probe (purple box, A) (Figures 1 F, 2 A and 5 A). For DSBs at leu2 :: hisG , genomic DNA was digested with PstI, separated and transferred as above, and then hybridized with the LEU2 probe (blue box, B) (Figures 1 G, 2 B, 5 B and 6 B).
To quantify full-length acentric and dicentric chromosomes, chromosomal-length DNA was pr epar ed after first immobilizing cells in agarose plugs as described ( 28 ). Chromosomes were separated using a CHEF-DR III pulsedfield gel electrophoresis (PFGE) system (Bio-Rad) using the following conditions: 1% agarose in 0.5 × TBE; 14 • C; 6 V / cm; switch angle 120 • . For Figure 3 , switch times were ramped from 12 to 20 s for 25 h. For Figure 4 , switch times were ramped from 5 to 15 s for 40 h. After transfer to Nylon membrane under denaturing conditions, genomic DNA was hybridized with DNA probes that localize close to the left ( CHA1 ) or right ( GIT1 ) telomere of chromosome III. Radioacti v e signals wer e collected on phosphor scr eens, scanned with a Fuji FLA5100 and quantified using Image-Gauge software (FujiFilm). When quantifying acentric and dicentric chromosomes, we divided the detected signal of these recombinants by 2, due to the presence of two copies of the probed sequence within the molecule to normalize signal relati v e to the rest of the lane. All DSBs and ectopic products are reported as a percentage of the total lane signal after background subtraction.
Values plotted with standard deviation bars are the mean of at least three independent timecourse experiments.

Ectopic recombination occurs between HIS4::LEU2 and leu2::hisG
The natural LEU2 locus is present on the left arm of chromosome III (S288c YCL018W , position 91324-92418). In strains used by many labs investigating meiotic recombina tion, this na tural LEU2 copy is interrupted by an ∼1.2 kb insertion of exogenous hisG DNA (Figure 1 A) ( 20 ). In addition, ∼3.0 kb of sequence from the natural uninterrupted LEU2 locus is often inserted to the left of the HIS4 gene --an insertion tha t genera tes a pair of strong meiosis-specific Spo11 DSB sites, one at the inserted LEU2 promoter (DSB II) and the other, stronger site, at the left-most insertion junction and coinciding with the fortuitous co-insertion of 72 bp of bacterial DNA (DSB I; Figure 1 A) ( 20 ). Recombination between these LEU2 repeats, intra-chromosomally (Figure 1 B) or inter-chromosomally (Figure 1 C), generates four possible ectopic recombination products (EC1-EC4; Figure 1 D), two of which (EC1 and EC2) have been measured forming during meiosis using physical analyses ( 15 , 21-23 , 25 ).
EC1 and EC2 theoretically differ only by the presence and absence of the hisG sequence, respecti v ely, as do EC3 and EC4 (Figure 1 D). To determine whether this was the case, we generated strains with wild-type LEU2 , without the hisG sequence, and observed both a general decrease in ectopic recombinant le v els forming during meiosis and, im-portantly, a complete loss of EC1 and EC3, consistent with the latter classes of recombinant products forming between HIS4::LEU2 and leu2::hisG and in a manner that retains the hisG sequence (Supplementary Figure S1).
During assembly of these diagrams, we realized that such ectopic r ecombinants ar e expected to produce acentric and dicentric rearrangements of chromosome III (Figure 1 B-D), something not explicitly noted before ( 15 , 21-23 , 25 ). To assess the entire spectrum of ectopic recombination arising between the LEU2 repeats, we designed restriction digests and DNA probes (Figure 1 A-D) capable of assessing the relati v e frequency of all four ectopic products via Southern blotting of genomic DNA isolated fr om synchr onous meiotic cultures of S. cerevisiae .

Dmc1-independent ectopic recombination is increased in rad24 Δ mutants
Pre vious wor k has demonstra ted tha t the forma tion of two of the ectopic recombinants (labelled here as EC3 and EC4) was increased upon deletion of the RAD24 DNA damage clamp loader ( 21 ), and intriguingly that the proportion of products is altered when the meiosis-specific RecA family recombinase, Dmc1, is absent ( 23 ). We corroborated these r esults, r e v ealing an increase in EC3 and EC4 in rad24 Δ cells, and a total loss of EC3 in the absence of Dmc1 (Figure 1 E, F and H). Moreover, our analyses enabled us to also measure the frequency of EC1 and EC2. These products were also increased in rad24 mutants, with formation of EC1 being highly dependent on Dmc1 activity ( Figure 1 G and H). The differential formation of EC1 and EC3, versus EC2 and EC4 in backgrounds with or without DMC1 , respecti v ely, led us to hypothesize that they may arise through mechanistically distinct DNA repair pathways (see below).

Ectopic recombination is suppressed by Sgs1
Sgs1, the S. cerevisiae orthologue of the human BLM helicase, has been implicated at se v eral steps in the meiotic recombina tion pa thway ( 25 , 27 , 29-31 ). Sgs1 is most often termed as an anti-recombinase --promoting the dismantling of nascent recombination intermediates and homologous interactions ( 29 , 30 ). Sgs1 has previously been implicated in suppressing ectopic recombination ( 24 , 25 ), but it has been unclear in which pathways Sgs1 acts. Here, to help determine which ectopic r ecombination r epair pathways Sgs1 is involved in, we assessed the frequency of ectopic recombination between the LEU2 repeats in a yeast strain depleted for Sgs1 expression ( sgs1-md 'meiotic depletion') in the presence and absence of Rad24 and Dmc1 ( Figure 2 ).
Depletion of Sgs1 increased the frequency of all four ectopic products compared to wild-type cells, in most cases exceeding the levels observed in the rad24 Δ single mutant ( ( D ) Ectopic recombinant products generated from repair occurring between HIS4::LEU2 and leu2::hisG with probes identifying recombinants indicated. EC1 and EC3 include hisG sequence in contrast to EC2 and EC4. Ectopic recombinant labelling is consistent throughout, but not with previous publica tions investiga ted due to inconsistencies existing from across articles. ( E ) Southern blot of the HIS4::LEU2 locus using the MXR2 probe previously described in ( 15 ). DSB signal is indicated by arrowheads and brackets. Non-specific signal is indicated by diamond. r esidual ectopic r ecombinants w as shifted tow ards EC2 and EC4 (Figure 2 A-C). This effect was further exacerbated in the sgs1-md rad24 Δdmc1 Δ triple m utant, w hich increased le v els of ectopic recombinants 3-fold compared with the sgs1-md dmc1 Δ mutant and had the heaviest skew towards EC2 and EC4 (Figure 2 A-C). Additionally, a novel ectopic r ecombinant pr edicted to form between the HIS4 :: LEU2 and nuc1 :: LEU2 loci was detected in the sgs1-md rad24 Δ dmc1 Δ background (Figure 2 A, asterisk, and Supplementary Figure S2). We conclude that, as for other types of homologous repair reactions, Sgs1 also appears to suppress the formation of ectopic recombination, as has been suggested elsewhere ( 24 , 25 ), and acts in both the Dmc1dependent and -independent repair pathways.

Ectopic recombination between HIS4::LEU2 and leu2::hisG creates acentric and dicentric products
We noted that the inverted orientation of the homologous LEU2 repeats means that ectopic recombination is expected to generate acentric and dicentric rearrangements of chromosome III --something that had not previously been documented (Figure 1 B-D). To investigate this process, we used PFGE to separate whole chromosomes of S. cerevisiae undergoing synchronous meiosis, and sequentially hybridized the Southern blots using DNA probes to the two termini of chromosome III (Figure 3 ). We carried out these experiments in the DMC1 + background to avoid the complications in the quantification of ectopic products that would arise from most of the chromosome III-associated DSBs r emaining unr epair ed in the absence of Dmc1 activity.
To our surprise, bands migra ting a t the expected position for both acentric and dicentric chromosome III rearrangements were observed at significant levels ( ∼2% of total DNA) e v en in the wild-type control strain (Figure 3 ). Both acentric and dicentric products wer e incr eased 2-3fold upon loss of either Rad24 or Sgs1 activity, with the double mutant displaying similar, or slightly greater, le v els to the Sgs1 single mutant (up to ∼5-6% of total DNA) (Figure 3 ). These frequencies agree well with the frequencies of EC1 + EC2 and EC3 + EC4 measured earlier by standard Southern blotting (Figures 1 and 2 ), supporting the conclusion that they are independent readouts of the same molecular reactions.

Ectopic recombination chromatid repair template choice is altered in sgs1 and rad24 mutants
Ectopic recombinants can form by repairing from the interor intra-sister chromatid or the homologous chromosome. To determine whether there is bias towards one repair template in ectopic r ecombination, we incr eased the length of one copy of chromosome III by insertion of a 7-kb plasmid on the left-hand end at the PRD1 locus and measured acentric recombinant formation in diploid yeast heterozygous for the chromosome III variants (Figure 4 A). In wild type, a total of ∼4% of acentric recombinants form, with a ratio of 3:1 bias towards repair from the homologous chromosome (Figure 4 B-D). The inter-homologue ratio observed is similar to that reported for allelic inter-homologue DSB repair ( 32 ). In both rad24 Δ and sgs1-md single and double mutants, the total frequency of acentric ectopic r ecombinants incr eases ∼2-3-fold compar ed to wild type, and inter-homologue bias decreases, with mutation of rad24 Δ leading to little or no inter-homologue bias (Figure 4 B-D). These results are consistent with the previous observa tion tha t the Mec1-Rad24 checkpoint pa thway is r equir ed for establishment of the inter-homologue repair bias during allelic meiotic recombination ( 18 ). In addition, sgs1-md also lowered the inter-homologue ectopic recombinant ratio compared to wild type, consistent with previous observations that Sgs1 plays a role in establishing interhomologue bias ( 24 ).

Ectopic recombination is both Rad51-and Dmc1independent in rad24 mutants
The shift in ectopic product formation from EC1 to EC2 and from EC3 to EC4 that arises upon loss of Dmc1 activity led us to consider whether EC2 and EC4 products can arise from a novel --potentially recombinase-independent (i.e. Dmc1-and Rad51-independent) --molecular reaction.
Before taking this further, we first tested the r equir ement of Rad51 in the formation of the four ectopic recombinants ( Figure 5 ). Consistent with prior observations made in single dmc1 Δ (Figure 1 ) or single rad51 mutants ( 21 , 23 ), we observed that the shift towards EC2 and EC4 that arises in dmc1 Δ mutants was also independent of Rad51 (Figure 5 ) --indica ting tha t these ectopic products were likely to be forming without a r equir ement for the canonical DNA strand invasion and homolo gy reco gnition steps usually associated with homology-directed repair.
The main homology-directed repair process that is independent of RecA orthologue activity is SSA ( 33 ). SSA is capable of stitching together a pair of repeats --deleting the intervening region --in response to the two regions becom-ing single-stranded. In genetic assays, this reaction most often involves bidir ectional r esection from a single DSB that forms between a pair of direct repeats [e.g. ( 34 , 35 )]. Howe v er, importantly, resection from a single DSB is unable to create single-stranded DNA with the r equir ed complementary polarity to enable the annealing of a pair of inverted r epeats. Mor eover, the LEU2 inverted repeats used in our assay are > 25 kb a part, significantl y farther than meiotic resection has been measured to traverse ( 11 , 12 ).

leu2::hisG is a frequently breaking meiotic DSB hotspot and provides homology for repair by SSA
Gi v en that DSBs arising within the HIS4 :: LEU2 locus would be incapable --alone --of enabling an SSA r epair r eaction involving leu2 :: hisG , we sought to assess the frequency of DSBs forming across the leu2 :: hisG region (Figur e 6 ). Pr evious w hole-genome anal ysis of Spo11 DSB forma tion indica ted tha t a significant number of reads mapped to within the exogenous hisG insertion ( 5 ). To confirm this observation, we used Southern blotting to probe the leu2 :: hisG locus in meiotic genomic DNA pr epar ed from a sae2 Δ strain where DSBs accumulate without repair and also in the set of strains used in our ectopic analyses (Figure  6 B). Consistent with the whole-genome mapping data, we detected a strong meiosis-specific DSB site located within the hisG insertion (strength ∼10% of total DNA, comparable to other strong sites like ARE1 ), and consistent with the genome-wide mapping data (Figure 6 C) ( 5 ).
We then explored the possible outcomes for SSAmediated repair of two DSBs --one arising at hisG and the other at either of the two hotspots within the HIS4 :: LEU2 insertion ( Figure 6D). Due to the location of homologies relati v e to the major DSB sites, we predict there to be only two major products possible via a process of SSA --both of which are expected to lack the hisG sequence itself because this will be trimmed away in the 3 flap removal stage of SSA. SSA product 1 involves repair between LEU2 homology to the right of hisG and LEU2 homology to the right of HIS4 :: LEU2 DSB I. Because both homologies are centr omere-pr oximal, the expected pr oduct is a dicentric identical in structure to EC2. By contrast, SSA product 2 involves repair between LEU2 homology to the left of hisG and LEU2 homology to the left of DSB II. The expected product in this case is an acentric product identical in structure to EC4.
Taken together, these analyses strongly suggest that SSA is another --previously unconsidered --mechanism that can generate gr oss chr omosomal rearrangements and deletions during meiosis between dispersed genetic elements. While a less than ideal repair outcome, the potential to form ectopic recombinants by SSA is likely more favourable than unr epair ed DSBs, which would provide a greater potential for chromosomal loss or gain during chromosome segr egation. Mor eover, the fact that ectopic recombination outcome skews strongly towards those compatible with SSA when Dmc1 recombination activity is abrogated suggests a hitherto unforeseen role for the canonical strand in vasion-dependent HR pathwa y in suppressing deleterious outcomes --something that may be especially important in meiotic cells that experience a high frequency of DSBs forming and repairing across the genome within the same window of time (see the next section for further details). Rad24 and Sgs1 function to promote interhomologue HR repair and inhibit inter-and intra-chromatid HR. DSB repair can occur via recombinase-dependent and -independent SSA and breakinduced replication (BIR), but the relationship between Rad24 and Sgs1 regulation is unknown.

DISCUSSION
Pre vious inv estiga tions of ectopic recombina tion during S. cerevisiae meiosis have focused on two of the four products generated between the HIS4 :: LEU2 and leu2 :: hisG loci. Our analysis has investigated the mechanisms involved and characterizes in detail all ectopic recombination outcomes at the local and chromosomal scales.
Consistent with previously published da ta, our da ta show tha t muta tion of Rad24 and Sgs1 incr eases ectopic r ecombination ( Figure 7 ) ( 15 , 21-25 ). Interestingly, our data reveal that when combined with dmc1 Δ, the types and levels of ectopic recombinants that form change (summarized in Supplementary Table S1). Using a strain lacking the hisG insertion at LEU2 , we re v ealed not only a loss of ectopic recombinants that contain the hisG sequence (EC1 and EC3), but also a general decrease in ectopic recombinant le v els, suggesting that the main pathway for ectopic recombinant formation depends on the hisG insertion --itself a strong meiotic DSB hotspot. The le v els of ectopic recombinants observed in the rad24 Δ dmc1 Δ, sgs1-md dmc1 Δ and their respecti v e single mutants lead us to conclude that inhibition of an ectopic repair mechanism is occurring in the sgs1-md dmc1 Δ background, which is not present in the rad24 Δ dmc1 Δ background. Because an increase in ectopic recombination is observed when removing Rad24 function from the sgs1-md dmc1 Δ strain, we propose that it is Rad24 that functions to inhibit ectopic recombinant formation in this strain. The mechanism by which Rad24 inhibits ectopic recombination is unclear, but may be related to either its checkpoint role or the increased resection observed in Rad24 mutants (summarized in Supplementary Table S1) ( 36 , 37 ).
The DNA checkpoint clamp loader Rad24 and the helicase Sgs1 are known to function throughout meiotic recombination ( 15 , 22-24 , 27 , 29 , 30 , 36-38 ). As such, Rad24 and Sgs1 aid in defining repair template choice, both allelic and non-allelic, and, as our data show, sister chromatid and homologous chromosome choice (Figure 4 ) ( 18 ). Gi v en the previously described actions of Rad24 and Sgs1, one model for pre v enting ectopic recombination by these proteins may be that Rad24 functions early, defining correct template choice, and Sgs1 functions later to unwind ectopic strand invasion e v ents. Layered on top of these mechanisms are the activities of Rad51 and Dmc1 that promote HR repair, enabling DNA invasion at homologous templates but also tolerating the allelic variation that may exist between homologs. Howe v er, while this model seems appealing, the decrease in inter-homologue bias observed in the sgs1-md backgr ound pr ovides further insight into Sgs1 activity in regulating ectopic recombinant forma tion a t an early stage.
Previous analysis of inter-homologue repair at HIS4 :: LEU2 relies upon detecting differentially sized DNA molecules generated by recombination between chromosomes containing restriction site polymorphisms.
Howe v er, non-crossov er repair products generated via inter-sister repair products cannot be distinguished from the parental DNA. Gi v en that all ectopic crossover recombinants formed lead to a product of a different size compared to the parental chromosomes, using two differentially sized chromosome III alleles allowed us to distinguish all acentric molecules and determine the ratio of inter-sister to inter-homologue template use. In doing so, we have revealed that in the rad24 Δ background ectopic recombinants form at similar le v els between sister chromatid and homologous chromosome. Strikingly, we show that sgs1-md also has a reduction in inter-homologue bias in ectopic recombinant formation. Previous observations have defined Sgs1 as playing a role in disassembling D-loop formation, promoting synthesis-dependent strand annealing (SDSA) and class I crossover repair ( 24 , 27 , 29 , 30 ). The r esults pr esented her e, and consistent with published data, support a role of Sgs1 in promoting inter-homologue repair ( 24 ).
Prior work has shown that ectopic recombination does not r equir e Ndt80 expr ession, a transcription factor that r egulates expr ession of middle meiosis genes, some of which ar e r equir ed f or Hollida y junction resolution ( 21 ). Gi v en this, ectopic recombinants may be generated by the activity of Ndt80-independent structure-specific nucleases, such as Mus81, long-range SDSA, BIR or SSA, although we cannot rule out canonical Holliday junction resolution mechanisms also playing a role. In our anal ysis, m utation of DMC1 decreases the frequency of ectopic recombinants formed that retain the artificially inserted hisG sequence (EC1 and EC3) with the remaining ectopic recombinants (EC2 and EC4) still forming in the absence of both RAD51 and DMC1 (summarized in Supplementary Table S1 and Figure 7 ) . These results support the idea that ectopic recombinants formed in the dmc1 Δ/ rad51 Δ dmc1 Δ background are generated by a strand invasion-independent mechanism, ther efor e ruling out SDSA as a potential repair mechanism. We propose that the remaining ectopic recombinants form by SSA (see below) due to the otherwise e xtensi v e (chromosome-length) DNA synthesis that would be r equir ed to generate ectopic products by Rad51 / Dmc1independent BIR. On this point, we further note that acentric and dicentric gross chromosomal rearrangements arise with similar kinetics (Figure 3 ) despite, if they were forming via BIR, the latter requiring the generation of ∼10 times the amount of newly synthesized DNA. Although further investigations to unambiguously elucidate the repair mechanism are theoretically possible, such experiments are technically challenging due to the low cell viability of the multiple mutant strain r equir ed to test this question [a DN A repair m utant ( rad24 Δ or sgs1-md ) in a recombinasedefecti v e background ( rad51 Δ dmc1 Δ) in combination with a BIR mutant or SSA mutant].
Combined, our data lead us to conclude that the SSA repair pathway is utilized in recombinase-defecti v e backgrounds. Interestingl y, it is onl y through insertion of the bacterial hisG sequence, leading to the formation of a second strong Spo11 DSB hotspot, that the formation of single-stranded DN A homolo gy r equir ed for r epair by SSA takes place. Thus, we infer that ectopic SSA products arise through the coincident formation of Spo11 DSBs at both HIS4 :: LEU2 and leu2 :: hisG , as has been demonstrated to be possible ( 14 ), and consistent with the reduction in ectopic recombinants that occurs when the hisG insert is removed (Supplementary Figure S1).
While the ectopic repair characterized in this study occurs via the LEU2 repeats adjacent to the Spo11 hotspots described, it should be noted that insertion of hisG sequence has been used widely to disrupt gene function in meiotic SK1 S. cerevisiae strains (observed in strains defecti v e in the HO endonuclease, his3 gene and trp1 gene, for example) and, as such, is likely to have introduced novel Spo11 DSB hotspots within each ectopic hisG repeat. Consequently, the hisG insertion may have altered the local recombination landscape, enabling ectopic recombination between dispersed hisG repeats via HR and SSA repair pathways.
Mutation of DNA repair proteins also contributes to other factors that may influence ectopic recombination (summarized in Supplementary Table S1). For example, mutations of Rad24 and Rad51 / Dmc1 are known to incr ease r esection lengths ( 37 , 39 , 40 ) and rad24 Δ also incr eases the pr efer ence for Spo11 DSBs to arise at DSB1 versus DSB2 within the HIS4 :: LEU2 hotspot ( 15 ). Due to the location of the DN A homolo gies involved, we would predict that increased resection tract lengths would favour EC1 and EC3 products, whereas an increased bias of DSB1 o ver DSB2 w ould promote the formation of EC1 and EC2. While it is likely that these effects have an impact on the outcome of ectopic recombination, the ectopic recombinants formed in these backgrounds do not fit these predictions. Ther efor e, our data suggest that the absence of these DNA repair proteins directly alters DSB repair pathway choice, and that this has a dominant role in the amount and type of ectopic recombinants formed.
Taken together, our data identify differences in ectopic recombinants that form across mutant backgrounds originating through differ ent DSB r epair pathways. For the first time, we characterize the formation of large chromosomal translocations arising by ectopic recombination in S. cerevisiae meiosis, leading to the generation of acentric and dicentric chromosome fragments. Such rearrangements affect copy number variation across the genomic region and dicentric chromosomes will cause chromosome segregation problems, affecting cell survival. While we have identified these specific ectopic recombination outcomes, we recognize that gaining greater understanding of the different types of recombinants and rearrangements that form across wild-type and mutant backgrounds both within and outside of meiotic prophase will provide important and generalized insight into the mechanisms that control repair pathway choice and maintain genome stability.

DA T A A V AILABILITY
The data that support the findings of this study are contained within the article and the supporting information. All source data generated for this study are available from the corresponding author (Stephen Gray; stephen.gray@nottingham.ac.uk) upon reasonable request.

SUPPLEMENT ARY DA T A
Supplementary Data are available at NAR Online.