Reshuffling yeast chromosomes with CRISPR/Cas9

Genome engineering is a powerful approach to study how chromosomal architecture impacts phenotypes. However, quantifying the fitness impact of translocations independently from the confounding effect of base substitutions has so far remained challenging. We report a novel application of the CRISPR/Cas9 technology allowing to generate with high efficiency both uniquely targeted and multiple concomitant reciprocal translocations in the yeast genome. Targeted translocations are constructed by inducing two double-strand breaks on different chromosomes and forcing the trans-chromosomal repair through homologous recombination by chimerical donor DNAs. Multiple translocations are generated from the induction of several DSBs in LTR repeated sequences and promoting repair using endogenous uncut LTR copies as template. All engineered translocations are markerless and scarless. Targeted translocations are produced at base pair resolution and can be sequentially generated one after the other. Multiple translocations result in a large diversity of karyotypes and are associated in many instances with the formation of unanticipated segmental duplications. To test the phenotypic impact of translocations, we first recapitulated in a lab strain the SSU1/ECM34 translocation providing increased sulphite resistance to wine isolates. Surprisingly, the same translocation in a laboratory strain resulted in decreased sulphite resistance. However, adding the repeated sequences that are present in the SSU1 promoter of the resistant wine strain induced sulphite resistance in the lab strain, yet to a lower level than that of the wine isolate, implying that additional polymorphisms also contribute to the phenotype. These findings illustrate the advantage brought by our technique to untangle the phenotypic impacts of structural variations from confounding effects of base substitutions. Secondly, we showed that strains with multiple translocations, even those devoid of unanticipated segmental duplications, display large phenotypic diversity in a wide range of environmental conditions, showing that simply reconfiguring chromosome architecture is sufficient to provide fitness advantages in stressful growth conditions.


INTRODUCTION
Genetic polymorphisms are not restricted to base substitutions and indels but also include large-scale Structural Variations (SVs) of chromosomes. SVs comprise both unbalanced events, often designated as copy number variations (CNVs) including deletions and duplications, and balanced events that are copy number neutral and include inversions and translocations. Both have a phenotypic impact, however the prevalence and the fitness effect of balanced SVs has been less documented than CNVs, partly because they are much more challenging to map than CNVs and also because quantifying their fitness contribution independently from the confounding effect of base substitutions remains challenging. Natural balanced chromosomal rearrangements result from the exchange of DNA ends during the repair of Double Strand Breaks (DSBs) either through Homologous Directed Repair (HDR) between dispersed repeats or intact chromosomes carrying internal repeat sequences homologous to the DNA ends (Piazza et al. 2017), or through Non-Homologous End Joining (NHEJ) (Branzei and Foiani 2008). Artificial balanced rearrangements are classically engineered by inducing targeted DSBs and promoting repair through both HDR and NHEJ. However, inducing targeted DSBs and engineering scar-less chromosomal rearrangements has remained challenging. In early studies structural variants were obtained through I-SceI-induced DSB repair between split alleles of a selection marker (Fairhead et al. 1996;Richardson and Jasin 2000). In later developments, the use of the I-SceI endonuclease was combined to a "COunter-selectable REporter" or CORE cassette in the frame of the delitto-perfetto technique, allowing the generation of a reciprocal translocation in a scar-less fashion (Storici andResnick, 2003, 2006). Other techniques based on Cre-Lox recombination were used to make the genomes of Saccharomyces cerevisiae and Saccharomyces mikatae colinear and generated interspecific hybrids that produced a large proportion of viable but extensively aneuploid spores (Delneri et al. 2003). Cre/Lox recombination was also used to assess the impact of balanced rearrangements in vegetative growth and meiotic viability (Naseeb and Delneri 2012;Avelar et al. 2013;Naseeb et al. 2016). A novel approach using yeast strains with synthetic chromosomes allowed extensive genome reorganization through CreLox-mediated chromosome scrambling (Annaluru et al. 2014;Hochrein et al. 2018;Jia et al. 2018;Shen et al. 2016). This approach proved to be efficient to generate strains with a wide variety of improved metabolic capacities (Jia et al. 2018;Blount et al. 2018;Luo et al. 2018b;Shen et al. 2018). Muramoto and collaborators recently developed a genome restructuring technology relying on a temperature-dependent endonuclease to conditionally introduce multiple rearrangements in the genome of Arabidopsis thaliana and S. cerevisiae, thus generating strains with marked phenotypes such as increased plant biomass or ethanol production from xylose (Muramoto et al. 2018). Methods using Zinc Finger Nucleases (ZFNs) and Transcription Activator-Like Effector Nucleases (TALENs) were also developed to generate targeted rearrangement in yeast, mammalian and zebrafish cells (Brunet et al. 2009;Piganeau et al. 2013;Richard et al. 2014;Xiao et al. 2013). Although these technologies provide very useful insights, they are often difficult to implement and/or rely on the use of genetic markers. For this reason, the development of the CRISPR/Cas9 (Clustered Regularly Interspaced Short Palindromic Repeats/CRISPR associated) system has boosted the field of genome engineering (Doudna and Charpentier 2014;Fraczek et al. 2018;Alexander 2018). This system, initially derived from immune systems of bacteria, consists of an endonuclease encoded by the Cas9 gene of Streptococcus pyogenes and a short RNA that guides the endonuclease at the targeted genomic locus. The gRNA can be easily designed to target any genomic locus proximal to a "NGG" Promoter Adjacent Motif (PAM). This technology is now routinely used to introduce targeted DSBs in genomes from a wide variety of species (Wang and Qi, 2016).
In yeast, CRISPR/Cas9 induced DSBs can be repaired with high efficiency by providing homologous repair DNA cassettes, allowing a variety of genome editions. Previous studies achieved the introduction of point mutations, single and multiple gene deletions and multiplexed genome modifications at different loci by transforming cells with plasmids bearing single or multiple gRNAs and linear DNA repair templates (DiCarlo et al. 2013;Jakočiūnas et al. 2015;Mans et al. 2015Mans et al. , 2018. CRISPR-based approaches have also been developed to add centromeres and telomeres to chromosome fragments (Sasano et al. 2016) concatenating chromosomes (Luo et al. 2018a;Shao et al. 2018) and for massively parallel genome editing to generate large libraries of genetic variants (Bao et al. 2015;Roy et al. 2018;Sadhu et al. 2017).
CRISPR/Cas9 also opened new avenues in the study of genome structure especially with the engineering of translocations in mammalian cells with high efficiency. The principle is to introduce two DSB in two distinct chromosome with CRISPR, then repair the DNA ends in trans by HDR with donor DNA carrying a selection marker, lost in a second step by Cre/Lox recombination leaving a single loxP element at the chromosomal junction (Vanoli et al. 2017).
In this study, we developed a CRISPR-Cas9 multiplexed genome editing strategy to generate markerless reversible and non-reversible reciprocal translocations in yeast with base-pair precision and high efficiency. We also induce multiple DSBs by targeting scattered Ty3-LTRs, thereby generating patchwork chromosomes resulting from multiple translocations and complex rearrangements. Finally, we quantified the phenotypic impacts of both targeted and multiple rearrangements on meiotic fertility and mitotic growth in various stress conditions.

Rationale for chromosome reshuffling
We used a single-vector, pGZ110 (Bruce Futcher, personal communication), which encodes both the Cas9 nuclease gene and a gRNA expression cassette (fig 1). This cassette allows cloning of either a unique 20 bp fragment corresponding to the target sequence of the gRNA or a synthetic DNA fragment of 460 base-pairs reconstituting two different gRNAs in tandem (fig 1). This system is highly versatile as it allows cloning of any pair of gRNA in a single ligation step. Such pairs of gRNAs can induce two concomitant DSBs which, upon recombination with chimerical donor DNA, generate reciprocal translocations with a single-nucleotide precision when the DSBs are introduced in different chromosomes. Furthermore, a single gRNA that targets Long Terminal Repeats (LTRs) scattered on different chromosomes can induce a higher number of DSBs, eventually resulting in multiple rearrangements upon repair with the uncut LTR copies. All rearrangements are engineered in a scar-less fashion and without integrating any genetic marker in the genome. Figure 1: Experimental strategy to clone in one step a single or a pair of gRNA target sequences. The pGZ110 plasmid encodes the Cas9 nuclease under constitutive promoter TEF1p, the gRNA expression cassette and the LEU2 selection marker. The cassette consists of an SNR52 promoter (blue) separated from the structural component of the gRNA (orange) and a SUP4 terminator (red) by two divergently oriented LguI sites (purple) allowing to clone the gRNA target sequence (green). Upon digestion, the two LguI sites generate non-complementary single strand overhangs of 3 bases. This system is highly versatile as it allows to clone in a single step either a 20 bp oligonucleotide corresponding to the target sequence of a unique gRNA (left) or a synthetic DNA fragment of 460 base-pairs reconstituting two different gRNA expression cassettes in tandem (right).

Engineering markerless, reversible reciprocal translocations at single base-pair resolution
We first engineered a reciprocal translocation between two reporter genes leading to phenotypes easy to observe upon disruption. Mutation in the ADE2 gene involved in purine nucleotide biosynthesis results in the accumulation of a red pigment while mutating the CAN1 gene which encodes an arginine permease confers canavanine resistance to the cells. We made this translocation reversible such that we can control the alternation between the two phenotypes  2C). Sanger sequencing of 250 bp around the chimerical junctions of these two strains confirmed that the translocation occurred right at the position defined by the sequence of the RT-donors with no additional mutation (supp fig 1). Finally, we cured the plasmid from the two strains before performing the reverse translocation to restore ADE2 and CAN1 (Material and Methods).          (1) and (2) indicate efficient transformation with the Cas9 plasmid and high cutting efficiency of the gRNAs, respectively. Panels (3) and (4)  Sanger-sequenced and were found identical to BY4741 natural junctions (supp fig 1). These results demonstrate that reversible chromosomal translocations can be engineered at base-pair resolution with high efficiency.

Engineering a deletion at the translocation breakpoint
In some instances, reciprocal translocations occur between repeated sequences (micro-homology or homologous sequences) that can also be used as template for the reverse translocation event restoring the WT configuration. Stabilizing a phenotype associated to an engineered translocation would therefore require the concomitant deletion of the homology regions to avoid such reversion events. We used our system to simultaneously generate a deletion of a few nucleotides and a reciprocal translocation. We used the same CAN1.Y and ADE2.Y gRNA target sequences as These experiments demonstrate that a deletion and reciprocal translocation can be concomitantly engineered at base-pair resolution with CRISPR/Cas9 in the yeast genome.

Recapitulating a natural translocation involved in sulphite resistance in wine strains
It was previously reported that a reciprocal translocation between the promoters of ECM34 and SSU1, a sulphite resistance gene, created a chimerical SSU1-R allele with enhanced expression resulting in increased resistance to sulphite in the wine strain Y9 (Perez-Ortin, 2002). This translocation resulted from a recombination event between 4 base-pair micro-homology regions on chromosomes VIII and XVI. We engineered the same translocation into the BY4741 background and tested sulphite resistance.
We first designed the two gRNA target sequences as close as possible to the micro-homology regions (supp fig 3A). The first gRNA targets the SSU1 promoter region, 115 base-pairs upstream of the start codon. The second gRNA targets the promoter region of ECM34, 24 base-pairs upstream of the start codon (supp fig 3A). To mimic the translocated junctions present in the wine strains, we designed double stranded synthetic DNA donors of 90 base-pairs centered on the micro-homology regions but not on the cutting sites. In addition, each donor also contained a point mutation in the PAM sequences to prevent subsequent CRISPR recognition (supp fig 3A).
The transformation with the Cas9 plasmid containing the two gRNAs and the donor DNA yielded on average 202 transformants. We tested natural and chimerical junction by colony PCR for 16 transformants and found the expected chimerical junctions in 15 of them (supp fig 3B). This translocation was not visible by PFGE karyotyping because the size of the translocated chromosomes were too close to the size of the original chromosomes. To further validate the rearrangement, we checked the junctions by southern blot for one translocated strain with probes flanking the two cutting sites on chromosome VIII and XVI ( fig 4A). This experiment clearly shows the presence of the chimerical junction fragments in the rearranged strain as compared to the WT.
Finally, we sequenced the junctions of two translocated strains and found that the rearrangement occurred within the expected micro-homology regions. In addition, the system was designed such that both mutated PAMs ended up into the promoter of ECM34, therefore likely to have no impact on the expression of the sulphite resistance gene SSU1 (supp fig 3C).
We then compared sulphite resistance between the lab strain in which we engineered the translocation, the non-translocated parental strain and wine isolates that either carry or are devoid of the translocation of interest. Surprisingly, we found that the engineered strain was the least resistant of all strains, including the non-translocated reference strain ( fig 4B). This suggests that the promoter of ECM34 in the BY background is weaker than the SSU1 promoter. The wine isolate with the translocation was the most resistant, followed by the wine isolate without the translocation.
It is interesting to note that the most resistant wine strain has tandem repeats of roughly 100bp in the promoter region of SSU1 gene brought by the translocation with the original ECM34 locus (Perez-Ortin, 2002). These repeats are absent from the reference strain.

Reshuffling chromosomes with multiple rearrangements
We introduced multiple DSBs in a single step using a unique gRNA targeting repeated Ty3 LTR sequences. There are 35 complete copies of Ty3 LTRs dispersed throughout the genome and these sequences are polymorphic ( fig 5A, supp fig 4A). Four of them comprise a region identical to the gRNA target sequence and also contain a PAM. A fifth copy, also flanked by a PAM, differs from the target sequence by a single mismatch at its 5' end and therefore might also be recognized by the gRNA (supp fig 4A). These 5 copies are located in chromosomes IV, VII, XV and XVI ( fig   5A). All the other Ty3 LTRs contain several mismatches or indels and/or are devoid of PAM, suggesting that Cas9 will not cut at these sites (supp fig 4A). Therefore, we expect from 1 to 5 sites to be concomitantly cut upon transformation with the Cas9/gRNA plasmid. In contrast to all experiments described above, we do not provide any donor DNA assuming that DSBs will be repaired by using uncut homologous LTR copies as template.
We  4B). In total 20 strains showed such predicted rearrangements, representing 11 distinct types (B, C, D, F, G, H, J, K, R, T and V in fig 5C). We validated the presence of all expected junctions by colony-PCR in 10 out of the 20 strains and on average 3 junctions out of 5 were also validated by PCR in the remaining 10 strains. We sequenced all chromosomal junctions in 2 strains that show the most frequently observed rearranged karyotype (type J in supp fig 4B). We found that all junctions, from both chimerical and un-rearranged chromosomes, were mutated in their PAM compared to the original target sequence (supp fig 4C).
This shows that during shuffling, all targeted sites were cut and repaired using as donor other Ty3 LTRs that had no PAM. Additional mutations in the region corresponding to the gRNA target sequence were also observed for 3 junctions, but were too few to identify which copy of Ty3 LTR was used as donor (supp fig 4C). We compared the theoretical frequency of each possible junction in the predicted types, considering types B to X being equiprobable, with the observed frequency of the junctions in the predicted strains whose type was assigned by PFGE ( fig 5D). We found very similar distributions, suggesting that DSBs were repaired in a random way. Moreover, we obtained targeted ones (fig 5B and 5C). For instance, chromosomes XI and XIV that have no PAM sequence associated with their Ty3 LTRs (supp fig 4A) are absent from several karyotypes suggesting that they can be rearranged in the absence of DSB ( fig 5B). In addition, some of these unexpected karyotypes showed an apparent genome size increase suggesting the presence of large duplications ( fig 5B). Using Oxford Nanopore MinION, we sequenced and de-novo assembled the genome of a strain showing a DNA content increase in PFGE (black star in fig 5B).
We characterized in this strain an unequal reciprocal translocation between chromosomes VII and XV ( fig 6A). The junctions corresponded to the targeted CRISPR cut site on chromosome XV but not on chromosome VII. In the chimerical chromosome XVtVII, the junction occurred away from the expected site resulting in a 30kb increase in DNA content ( fig 6A). In the chimerical chromosome VIItXV, the junction also occurred away from the expected site but was accompanied by a  6C).
Finally, we also recovered 39 type A strains that had a WT karyotype ( fig 5C). For all of them we validated the presence of the un-rearranged junctions by colony PCR. We sequenced the junctions in 4 independent clones. Surprisingly, we found that these clones gather transformants that underwent two different paths within the same experiment. Firstly, 3 clones had junctions identical to the original LTR sequences with intact PAM, suggesting that Cas9 did not cut the target sites.
Secondly, in the fourth clone, the PAM sequences of three junctions were mutated, showing that the corresponding chromosomes were cut and repaired yet without any rearrangement (supp fig   4C). In this clone, one junction could not be amplified, possibly because of a small-sized indel that could not be observed on the PFGE profile.

Exploring the phenotypic diversity of reshuffled strains
Firstly, we tested the meiotic fertility of diploid strains heterozygous for the chromosomal rearrangements by measuring spore viability in their offspring ( fig 5E). We tested 8 different strains with predicted karyotypes and PCR-validated junctions. These strains comprise various rearrangements including reciprocal translocations between 2 and 3 chromosomes (types C, D, E, F and G, respectively) and transpositions (types J and K). For each cross we dissected 10 tetrads from two independent diploid strains (20 tetrads in total). The control strain without any rearrangement shows 78% of viable spores with most tetrads harbouring 3 to 4 viable spores ( fig   5E). By contrast, all heterozygous diploids had a severely impaired spore viability ranging from 5 to 39% with predominantly only 1 to 2 viable spores per tetrad, except one that shows 66% of spore viability (type E in fig 5E). We observed no clear correlation between the type of rearrangement and the impact on fertility. One strain with a single reciprocal translocation between 2 chromosomes (type F) shows the lowest viability, comparable to that of a more rearranged strain with 2 translocations between 3 chromosomes (type G). By contrast, the strain showing the highest spore viability also has a single translocation between two chromosomes (type E). Note that two strains carrying the same transposition (type J) but from opposite mating type present significantly different spore viabilities (Chi2 Pval=0.004) suggesting additional genotypic heterogeneities between the 2 strains or some interactions between the rearrangements and the genotypes of the parental strains. These results show that knowing the type of rearrangements is not sufficient to predict their quantitative impact on meiotic fertility, although most of them have a drastic negative effect. Secondly, we measured the mitotic growth of rearranged strains in rich medium and under stress conditions including a variety of abiotic factors or drugs interfering with cell division. NaCl exerts an osmotic stress (Blomberg 1997), hydroxyurea (HU), is an inhibitor of nucleic acid synthesis (Koç et al. 2004), and caffeine is a cell-growth inhibitor known to interfere with the TOR pathway (Loewith and Hall 2011). We tested representative strains for 13 distinct predicted karyotype categories and 16 unexpected karyotype categories and indicated the generation time ratio of the rearranged mutants relatively to the WT strain ( fig 5F). Note that the type A strain that has a WT-like karyotype showed either a generation time identical to the WT strain as expected or a slightly increased one (1.2 times, fig 5F). To be conservative, we decided to consider as significant only the values smaller than 0.8 and greater than 1.2. In rich medium, all but one rearranged strain show no significant growth difference than the WT, showing that chromosome reshuffling has little impact on mitotic fitness in this growth condition (from 1.0 to 1.2 times longer generation times) although in one case the rearrangements, involving extensive aneuploidy, proved highly disadvantageous (second strain on the right with a generation time ratio of 1.5). In both caffeine and NaCl conditions, 4 rearranged strains were negatively affected with the most impacted strain being also the slow grower in the rich medium ( fig 5F). By contrast, in HU, the number of mutants positively and negatively impacted are much more balanced with 5 and 3 strains growing faster and slower than the WT, respectively. Interestingly, we found both predicted and unexpected karyotypes showing significant growth differences in the various environmental conditions with both positively and negatively impacted strains. In addition, two strains, one type G and one unexpected karyotype, showed alternatively increased and decrease mitotic fitness depending on the conditions showing genotype by environment interactions.

DISCUSSION
In this study, we developed a versatile CRISPR/Cas9-based method allowing to engineer, with the Considering that these predictable rearrangements were only represented by a small number of strains, some being not represented at all (supp fig 4B) and that the unexpected karyotypes were all nearly unique, the number of possible rearrangements accessible by our method is probably very large.
One reason that could explain some of the unexpected karyotypes would be that CRISPR/Cas9 may recognize and cut the two Ty3 LTRs on chromosome XII that contain a PAM despite small indels in the gRNA target sequence (supp fig 4A). Such types would be difficult to identify by PFGE because of the large size of chromosome XII and its frequent size variations due to the rDNA cluster dynamics. One cannot exclude that additional LTRs were cut by CRISPR/Cas9 even if this seems very unlikely because all other LTRs are devoid of PAM. A more plausible explanation would be multipartite ectopic recombination between cut and uncut chromosomes, as in (Piazza et al. 2017). In this study, the authors report that the single end of a DSB can invade multiple sequences located on intact chromosomes during its search for homology. This leads to translocations between intact chromosomes with insertions of fragments of the triggering singleend sequence in the translocation junctions. Consistently with this possibility we found that one unexpected chimerical junction resulted from Ty3-LTRs being "outcompeted" during repair by the full-length Ty2 elements found in the vicinity of the CRISPR-induced DSBs (fig 6). We also found several karyotypes with rearrangements in untargeted chromosomes suggesting that rearrangements can occur in intact chromosome in the absence of DSB.
Our procedure of genome reshuffling is reminiscent of the restructuring technology developed by Muramoto and collaborators (2018) that uses a temperature-dependent endonuclease to conditionally induce multiple DSBs in the genome of yeast and A.thaliana. However, our reshuffling procedure is potentially more versatile because we can target different types of repeated sequences with variable locations (within or between genes, subtelomeric or internal, etc) and variable copy numbers. By analyzing the genome of S. cerevisiae for repeated gRNA target sequences we identified 39,374 repeated targets with varying number of occurrences in the genome (from 2 to 59). Repeated targets were mostly found within Ty transposable elements (11,503) and protein coding genes (10,368 in 479 genes located in all 16 chromosomes).
Interestingly, we also found 3,438 repeated target gRNA located in intergenic regions which should allow to rearrange chromosomes without disrupting coding sequences. In addition, our CRISPR/Cas9 shuffling procedure is doable in any genetic background and therefore potentially more versatile than the Cre/Lox based SCRaMbLE techniques that can be used only in yeast strains with synthetic chromosomes.
Our approach provides the possibility to quantify the role of structural variants in organismal phenotypes. The translocation between the SSU1 and ECM34 genes provides increased sulphite resistance to wine isolates and is believed to be a consequence of sulphite supplementation in wine-making (Perez-Ortin, 2002). Surprisingly, the translocation we engineered in the reference strain resulted in decreased sulphite resistance. We hypothesize that this phenotype is linked to the absence of repeats in the promoter of ECM34 in the BY4741 background in contrast to wine strains (Perez-Ortin, 2002), implying that the promoter of ECM34 is weaker than the SSU1 promoter in the BY background. Therefore, the translocation itself would not be sufficient to explain increased sulphite resistance. These findings provide a striking example of the advantages brought by our technique to untangle the phenotypic impact of SVs from that of the genetic background.
We also measured the phenotypes of strains carrying multiple rearrangements both in mitotic and meiotic conditions. We showed that crosses between rearranged and parental strains produced very few viable spores, supporting earlier studies revealing that meiotic pairing and segregation is impaired in diploid strains bearing heterozygous translocations (Avelar et al. 2013;Loidl et al. 1998;Liti et al. 2006). In vegetative growth, reshuffling provided a great phenotypic diversity with strains showing fitness advantages under specific environmental conditions. Similar findings were previously described both in S. cerevisiae (Naseeb et al. 2016;Colson et al. 2004) and Schizosaccharomyces pombe (Avelar et al. 2013). However, in our study all rearrangements are completely markerless and scarless (no CRE/Lox site) and no gene was disrupted nor duplicated (at least for the predicted karyotypes) suggesting that balanced rearrangements between repeated sequences simply reconfiguring the chromosome architecture are sufficient to create fitness diversity. Therefore, we believe that CRISPR/Cas9 chromosome reshuffling is a powerful new tool complementing the features of pre-existing genome restructuring technologies and will prove useful in the study of genotype to phenotype relationship and for isolating new chromosomal combinations of biotechnological interests.

Sulphite resistance spot tests
Sulphite resistance of two wine strains and two engineered lab strains (YAF082, YAF083) was assessed on YPD plates buffered at pH = 3.2 and supplemented with sulphites according to an established procedure (Park et al. 1999) with minor modifications: buffered plates were prepared by adding tartaric acid / potassium, sodium tartrate buffer to a final concentration of 38 mM to freshly autoclaved agar YPD. Sulphites plates were prepared by spreading a 1M stock solution of Na 2 SO 3 on 75 ml buffered square plates to a final concentration of 2 mM or 4 mM. Sulphites were let to diffuse overnight at room temperature. The next day, for each experiment, we dispensed in triplicate 10 5 , 10 4 , 10 3 , 10 2 cells of each strain in 50 µL sterile water drops. The position of each strain in the three replicates was different to minimize neighbouring and edge effects. The plates were incubated for 4 days at 30 °C before being scanned. The whole experiment was repeated 3 times.

Growth curves in stress conditions
Generation time of strains were measured under the following conditions: rich medium (YPD), salt stress (YPD + 1.5 M NaCl), inhibition of nucleic acid synthesis (YPD + hydroxyurea 20 mg/ml), presence of caffeine known as a toxic purine analogue (YPD + caffeine 16 mM). These drug concentrations impeding but not stopping cell growth were determined as the concentration required to generate an approximately three-fold increase in the generation time of BY4741 as compared to the YPD growth condition. Compared to the observed generation time of 80 minutes in YPD, there was a fold increase of 3.7, 3.85 and 2.8 in NaCl, Caffeine and HU conditions respectively. Cells were diluted from a saturated YPD culture to 2.10 6 cells/ml and 10 µl corresponding to 2.10 4 cells were used to inoculate a new plate with 90µl of fresh medium per well.
OD 600 measurement was performed in precision mode every 15 minutes for 3 days using the TECAN Sunrise micro-plate reader and Magellan v7.2 software. Constant high intensity "interior mode" agitation and 30°C temperature were maintained during all experiments. In addition, a few Ø 500 µm glass beads (Sigma G8772) were added to each well to prevent cell aggregation.
For each strain and condition, growth curves were carried out in triplicate and the generation time was extracted using the R growthcurver package and averaged on the three measurements.

Spore viability
Colonies of rearranged strains originating from the BY background and SK1 strains of the opposed mating type were mixed and spread on the same YPD plate and left overnight at 30°C. The next day, the cells were re-suspended in distilled water and single cells were picked using the Sanger MSM 400 micro-manipulator and left to grow on YPD for 2 days at 30°C until a colony appeared.
Colonies originating from single cells were replicated on sporulation medium and left for 4 to 7 days at 30°C until tetrad appeared. For each cross, 10 tetrads from two diploid strains were dissected on YPD and left to grow for 3 days before counting viable spores. We then manually selected a suitable gRNA sequence targeting five Ty3-LTR elements and looked for off-targets to this target sequence with CRISPOR. Predicted off-targets had either mismatches with the chosen guide and/or were devoid of PAM indicating that they would not be recognized by CRISPR/Cas9.

Construction of CRISPR/Cas9 plasmids with one or two guides
All plasmids used in this study were obtained by cloning either 20bp, corresponding to the target sequence of a single gRNA, or a 460 bp synthetic DNA fragment allowing to reconstitute two gRNA expression cassettes in pGZ110 (fig 1). All oligonucleotides are listed in supp table 1. The plasmid pGZ110 was kindly provided by Gang Zhao and Bruce Futcher (Stony Brook University). We linearised pGZ110 with the enzyme LguI (ThermoFischer FD1934) and gel purified the backbone.
In order to clone one target sequence we first annealed two oligonucleotides of 23 bases with adequate 5' overhangs of 3 bases to obtain the insert. Annealing was performed by mixing equimolar amounts of forward and reverse oligonucleotides at 100pmol/µl with NEBuffer 4 (New England Biolabs), heating 5 minutes at 95°C and allowing the mix to cool down slowly to room temperature. We then mixed 100ng of backbone with 20pmol of double stranded insert and performed the ligation with the Thermo Fischer Rapid DNA ligation Kit (K1422) according to manufacturer instructions. In order to obtain plasmids with two gRNA expression cassettes, we first digested and gel-purified the 460bp synthetic DNA fragments with LguI to obtain inserts with adequate 5' overhangs of 3 bases for ligation in pGZ110. We then mixed 100ng of backbone and 20ng of insert and performed ligation as explained previously. Plasmids were amplified in E. coli.
Oligonucleotides and synthetic DNA fragments were ordered from Eurofins. Refer to supplementary material for oligonucleotides and plasmid sequences.

Yeast transformation
Yeast cells were transformed using the standard lithium acetate method (

Plasmid Stability
The pGZ110 plasmid is highly unstable when selection is removed. To cure the plasmid cell were grown overnight in YPD at 30°C. Ten individual cells were micromanipulated on YPD plates with the MSM400 micromanipulator (Singer) and grown at 30°C for 2 days. Colonies were then serially replicated on CSM-Leu and YPD. All 10 colonies lost the ability to grow on CSM-Leu.

Estimation of CRISP/Cas9 cutting and repair efficiencies
The ADE2/CAN1 experiment was used to quantify the translocation yields. We performed 16 independent transformation experiments.  2).
For the backward translocation restoring the ADE2/CAN1 original chromosomal configuration the transformation of two translocated strains with the plasmid bearing no gRNA produced on average 4010 colonies while cells transformed with the gRNAs plasmid but without donor DNAs produced on average 13 colonies, indicating efficient cutting of the chimerical junctions. Co-transformation with the gRNAs plasmid and donor fragments provided on average 207 colonies, 94.2% of which had the restored white phenotype. We replicated 16 white transformants on canavaninesupplemented medium and all strains were sensitive to canavanine.

Pulse Field Gel Electrophoresis and colony PCR for karyotyping rearranged strains
Whole yeast chromosomes agarose plugs were prepared according to a standard method (Török et al. 1993) and sealed in a 1% Seakem GTC agarose and 0.5x TBE gel. Pulse Field Gel Electrophoresis (PFGE) was conducted in with the CHEF-DRII (BioRad) system with the following program: 6V/cm for 10 hours with a switching time of 60 seconds followed by 6V/cm for 17h with switching time of 90 seconds. The included angle was 120° for the whole duration of the run. We compared observed karyotypes with expected chromosome sizes and tested the chromosomal junctions by colony PCR with ThermoFischer DreamTaq DNA polymerase.

Southern blot validation of ECM34/SSU1 translocation
Southern blot was used to validate the translocation between ECM34 (ch. VIII) and SSU1 (ch. XVI).
Genomic DNA was extracted from BY4741 and the engineered strain YAF082 using the Qiagen DNA buffer set (19060) and Genomic-tip 100/G (10243) according to manufacturer instructions and further purified and concentrated by isopropanol precipitation. Digestion of 10µg of genomic DNA per strain/probe assay was carried out using FastDigest EcoRI (ThermoFischer FD0274).
Electrophoresis, denaturation and neutralisation of the gel were performed according to established procedure (Sambrook et al. 1989). Transfer on nylon membrane (Amersham Hybond XL) was performed using the capillarity setup (Khandjian, 1987;Southern, 1975). The membrane was UV-crosslinked with the Stratalinker 1800 device in automatic mode. Probes targeting the genes YHL044W and ARN1 located upstream and downstream of the cutting site in the ECM34 promoter respectively and the genes NOG1 and SSU1 located upstream and downstream of the cutting site in the SSU1 promoter respectively were amplified with ThermoFischer DreamTaq, DIG-11-dUTP deoxyribonucleotides (Roche 11 175 033 910) and gel purified. Blotting and revelation were conducted using the Roche DIG High Prime DNA Labelling and Detection Starter kit II (11 585 614 910) according to manufacturer instructions. Imaging was performed using the G:BOX Chemi XT4 (Syngene) with CSPD chemiluminescence mode. All oligonucleotides are described in supp table 1.
Oxford Nanopore de-novo genome assembly DNA from strains YAF019 and YAF064 was extracted using QIAGEN Genomic-tip 20/G columns and sheared using covaris g-TUBEs for average reads lengths of 8kb and 15kb respectively. DNA was repaired and dA-tailed using PreCR and FFPE kits (New England Biolabs) and cleaned with Ampure XP beads (Beckman Coulter). SQK-LSK108 adapters were ligated and libraries run on FLO-MIN107 R9.5 flowcells. Raw signals were basecalled locally using Albacore v2.0.2 with default quality filtering. Flowcell outputs are shown in supp table 2.
Due to only 19x coverage, the correctedErrorRate for Canu assembly was increased to 0.16.
Assembly data are shown in supp table 3.
YAF064 was processed using the LRSDAY v1 pipeline. Linear chromosomes were assembled by SMARTdenovo v1 using 40x coverage of the longest Canu-corrected reads and combined with a Canu assembled mitochondrial genome. For canu correction, due to 200x coverage, correctedErrorRate was set at 0.75. All corrected reads were aligned against a reference-quality S288C genome, assembled with PacBio reads (Yue et al. 2017) using LAST-921. Split reads were used to highlight structural variations not apparent in the SMARTdenovo assembly. Read coverage was used to calculate an increase in the number of copies of particular regions within the rearranged genome. Evidence of rearrangements defined by overlapping reads and changes in copy number were used to manually adjust the assembly prior to nanopolish v0.8.5 and Pilon v1.22 error correction.

DATA ACCESS
The Oxford Nanopore sequencing data are deposited in the Sequence Read Archive under the project number (accession number pending).