Chosing the strategy for identification of S. cerevisiae diploid deletion clones displaying the mutator phenotype
The collections of Saccharomyces cerevisiae
strains with knockout of almost every gene present in the genome of this organism (YKO collections) constitute an invaluable and powerful tool enabling diverse functional tests on a genome-wide scale. Those tests can be done not only on individual strains but also on the mixed cell population containing all deletion clones in one culture, since each deletion strain is uniquely bar-coded with two 20 bp DNA sequences. The changes in relative abundance of individual clones in any mixture subjected to selection conditions can be monitored by PCR-amplification and labeling of the barcode sequences followed by comparative hybridization to barcode microarray 
. The collections have also proven to be a powerful tool for studying genetic interactions.
The screen for genes whose deletion results in genome instability holds one major difficulty. The strains deficient in such genes, being genetically unstable are less viable and, further, they will over time accumulate additional changes in their genomes. The strains that we intend to isolate, are at the same time the most difficult to preserve in their original state. Parental BY4743 contains two heterozygous markers MET15/met15Δ and LYS2/lys2Δ that could be conveniently used in LOF screen but in our experience heterozygosity of those loci is often lost, regardless of any defect in genome stability. Moreover, some of the potential mutators are slow growers and might be difficult to score as mutators in a high throughput screen. The barcode microarray-based SLM screen that we have devised establishes an improved method of detecting the mutator phenotype and provides the solution to these and other challenges. The key novelty of this method was the introduction of two new heterozygous markers CAN1/can1Δ and URA3/ura3Δ to the entire YKO collection. Equally important was the choice of the method of marker introduction. In theory the most reliable method of creating the collection of diploids homozygous for the deletion of every yeast gene and containing heterozygous LOF marker would be to introduce the marker into each clone of e.g. MATa deletion collection and then to cross each resulting clone with the respective clone from MATα deletion collection. There are, however, potential dangers that could compromise the quality of the clone set obtained in that way. Some deletion clones may mate inefficiently or not mate at all. One could reasonably expect that some of the clones defective in genome stability will fall into that category and thus will be excluded from the collection from the very beginning. Another obstacle would be the lack of methionine or lysine auxotrophy in some clones from the haploid collection making simple selection of diploids on drop-out medium impossible and necessitating the use of micromanipulator to catch diploid zygotes. Less laborious and less perfect would be to introduce the heterozygous marker into individual homozygous diploid deletion clones. With this approach, the inevitable failure of some difficult clones to transform successfully on the first attempt would require repeating, perhaps several times, the transformation procedure on a subset of the deletion strains. Thus the imperative to bring the derivative collection to perfection would increase time, labor and frustration. Moreover, any of these laborious approaches might turn out to be unproductive if we take into account that the strains we are most interested in are at the same time the least stable. Even collections prepared meticulously could soon become useless for genome instability selection. Thus we came to understand that the most streamlined approach would be the best and decided to introduce the LOF markers in a single transformation reaction done on the mixture of all deletion clones. With that approach it was achievable to prepare two separate derivative homodiploid clone mixtures with CAN1/can1Δ and URA3/ura3Δ markers, allowing whole-genomic estimates of SLM frequencies with more than one locus. Furthermore, we could set the starting point for DNA changes accumulation that was common for all deletion clones, and we could also narrow the time period between marker introduction and SLM assay to as little as 4 days, the equivalent of approximately 30 cell divisions. By optimizing the transformation procedure we could assure a single correctly targeted insertion of marker in as many as 99.9% of cells.
It is worth mentioning that a number of deletion strains clearly identified as mutators in our screens and selected for phenotype confirmation with the individual semi-quantitative test, later turned out to be extremely resistant to individual LOF marker introduction. So in retrospect we can say that in terms of deletion collection coverage and selection accuracy, the strategy chosen was at least as good as other, more laborious alternatives.
This method has of course its own shortcomings. We were aware that individual deletion strains might behave differently compared to the majority. Some may differ in transformation efficiency. Should it be lower than average, the clone would be underrepresented and the sensitivity of SLM detection for that clone will be lowered accordingly. Higher than average transformation efficiency does not cause any problems provided that marker cassette is still introduced in the right place and in single copy. By comparing the relative abundance of deletion clones before and after marker introduction using the same barcode microarray hybridization technique that was used for determination of SLM, we could assure that the derivative clone mixture containing the selection markers remained representative of the library. Another drawback of this method is the impossibility of performing any quality tests for correct marker insertion into the individual deletion clones. Although, on average, the great majority of Leu+ cells had a single copy of CAN1 replaced by can1Δ and the great majority of Ura+ cells got a single copy of ura3Δ replaced by URA3, some individual clones may display different behavior as a result of specific gene deletion. Since marker insertion involves the mechanisms of homologous DNA recombination, deletion strains defective in aspects of genome stability might be among those with an improperly inserted marker. It seems, however, that any inaccuracies in marker insertion had minor influence on the results obtained with the derivative clone pool. If the LOF marker is inserted at some frequency in the incorrect locus then some cells would still have two wild-type copies of the CAN1 gene and hence the frequency of SLM will be lowered. On the other hand, URA3 inserted randomly but in single copy would likely form a functional marker as good as that when it is inserted in place of ura3Δ. Multiple nonhomologous insertions of URA3 marker cassette would exclude that cell from the 5′-FOA resistance screen, whereas multiple nonhomologous insertions of can1Δ marker cassette would do no harm to the canavanine resistance screen as long as a single CAN1 gene is replaced by can1Δ cassette. It should be borne in mind that our derivative clone pools would contain around fifty independent transformation clones of each original deletion strain. Even should some of them be faulty and do not participate in selection for canavanine or 5-FOA resistance, the remaining ones should still respond as expected. The only effect would be lowered sensitivity of mutator phenotype detection for that strain. If, for any given deletion strain, all transformation clones are incorrect then the relevant gene would be lost to our screen. Yet such problematic strains would likely be missing also from the derivative set composed of strains transformed individually.
To make this method effective as a screen for increased SLM, two important conditions have to be met. Firstly
, the derivative pools heterodiploid with respect to mutagenesis markers must remain representative. To assure this, we prepared CAN1/can1::LEU2
heterodiploid pools with 58- and 42-fold coverage of yeast genome, respectively. The representativeness of both derivative pools was confirmed by comparison, using barcode microarrays, to the original HD+ESS pool. We observed that, despite our effort to assure the balance of the original pool (see Materials and Methods
), less than 3% of all strains consistently gave a signal that was so low as to preclude them from the analyses. Among them could be the strains growing extremely slowly that despite of it were allocated to the homozygous diploid collection rather than to the essential heterodiploid collection. Also, the presence of faulty barcodes in some of the deletion clones resulting in low or no hybridization cannot be excluded 
. Of the remaining over 97% deletion clones, only three were 15 to 10 fold underrepresented and another fifty were 10 to 5 fold underrepresented, relative to the parental pool. A further three hundred were 5 to 2 fold underrepresented. Thus, in our judgment the derivative pools remained sufficiently representative.
, the mixed population subject to canavanine or 5′-FOA selection should contain a sufficient number of cells of each deletion clone. Unlike in typical sensitivity or resistance screens where all tested cells carrying a given gene deletion behave similarly, only a small fraction of cells of each clone, determined by its mutator phenotype, would acquire a mutation at the marker gene locus (CAN1
). Therefore, to make this screen representative, the average number of cells of each clone used in the assay should be several-fold greater than the inverse of mutation frequency of the wild-type strain. Our tests revealed that SLM frequency in BY474X genetic background is 8.2×10−7
in haploid cells, and is approximately two orders of magnitude higher, namely 1.5×10−4
, respectively, in diploid cells. This is in accordance with published data 
. Thus, for the screen to be representative, the initial number of cells per single deletion clone should be at least 105
and the total number of cells in the whole population should be at least 109
(see Supplementary Figures S1
Contribution of our SLM screen data to the genome maintenance field
Much large-scale data pertaining to the genome maintenance in S. cerevisiae
exists in literature, including screens for the mutator phenotype in haploid cells 
, for increased LOH phenotype in diploid cells 
, or for genome instability genes relevant to cancer 
. The results of numerous global screens of sensitivity to various genotoxic stress are also available 
. There is only modest overlap of our gene list with any of the published studies, but they are also quite dissimilar (see Supplementary Table S2
). Although superficially one would expect that screens for related phenotypes should produce similar gene lists, it should be kept in mind that each screen approach is different. In practice dissimilarities of the gene lists contents should be anticipated regardless of which phenotype is assessed or which biological process is explored with genome-wide approaches. To us it is clear indication that, in the case of genome stability, the search for genes involved should continue and that diverse screening conditions may reveal distinct functions related to this biological process. Nonetheless for almost half of genes from our list data exist suggesting the involvement of their gene products in the genome stability (see Supplementary Table S2
Although our approach involved diploid cells, it was not limited to LOH events. Rather than focusing on this phenomenon, already extensively studied in excellent work of Andersen et al. 
, we aimed at identifying genes whose deletion or insufficiency (for essential genes) causes increased frequency of any DNA changes that could be detected with the employed markers. Those would include, besides LOH, point mutations, small deletions, epigenetic changes, or poorly characterized events. Rather than assigning mechanistic functions for gene products known for their involvement in genome stability, we were interested in finding new functional interconnections linking genome stability to other cellular processes. To make our screens more far-reaching, thus encompassing new, potentially interesting, functional groups of genes, two of them were performed on exponentially growing cells where any deficiency in genome stability systems will be better exposed than in postdiauxic or stationary phase cells. Both screens were done on the complete YKO collections with newly introduced heterozygous mutagenesis markers, CAN1/can1Δ
. The inclusion of the heterodiploid collection of essential gene deletions allowed us to study gene dosage effects for those genes.
Genes implicated in the genome stability
Several remarkable trends emerged from our SLM screen. Essential genes comprise approximately a quarter of all genes (65 out of 249) that stabilize the genome. This underlines importance to the cell of preservation of genomic integrity. The 249 genes could be allocated to separate groups: 190 (76.3%) are verified genes (even though only 40 have known genome stability associations), 36 (14.46%) are uncharacterized and 23 (9.24%) are considered dubious (see Supplementary Table S2
Nuclear and mitochondrial localization predominates among gene products selected in SLM screen
With respect to intracellular localization, the largest group of gene products can be found in the nucleus (32.12%, see Supplementary Table S2
and Supplementary Table S4
, ). Interestingly, a considerable fraction of these contains proteins located in the nucleolus (14 of 80 genes). This resembles the observation in Caenorhabditis elegans
cells that links genome integrity and post-transcriptional RNA regulation functions via diverse RNA metabolic processes 
. Although the presence of RNAi in S. cerevisiae
cells has not been documented, several lines of evidence indicate the existence of posttranscriptional regulation in yeast cells. It is known that the loss of function of the exosome component Rrp6 leads to stabilization of PHO84
antisense transcripts and subsequent inhibition of PHO84
gene transcription. The data indicate that PHO84
repression is not due to transcription interference, but results from antisense RNA-induced histone deacetylation by the Hda1/2/3 complex 
. In our screen we have found RNA degrading enzymes (RRP46, SKI3
) and different components of histone deacetylating complexes (HDA3, RTX3, SIF2
). Thus, we anticipate the existence in yeast cells of a posttranscriptional mechanism of gene expression modulation that influences genome stability in response of genotoxic stress.
Overrepresentation of GO annotations in the group of 249 genes selected in genomic SLM screen.
Our data also confirmed the observation that abnormalities in ribosome biogenesis, which in turn lead to START delay and affect the cell cycle, can provoke genome instability 
. In our screen we have found not only nucleolar genes responsible for rRNA processing and ribosome assembly (IPI3, LSM4, MPP10, NOP9, POP8, PTI1, RRP46, SLX9, UTP13
), but also genes encoding: ribosomal subunits (RPL4A, RPS22A, RSM24
, especially mitochondrial ones: MRPL7, MRPL15, MRPL16, MRPL28, MRPL39, MRPS16, MRPS5
), proteins engaged in RNA transport (HAS1, MAK21, NUP1
) and necessary for RNA turnover (SUV3
), proteins involved in the synthesis of rRNA (RSC9
) and rDNA silencing (TOF2
) and, finally, START regulators, WHI5
, gene products whose role is tied to sensing the intracellular ribosome level (, ).
Another considerable group of gene products is localized in the mitochondria. This can be explained in several ways, but most probably abnormal reactive oxygen species (ROS) production connected with deletion of a variety of mitochondrial genes results in an increase in endogenous premutagenic lesion formation 
. An alternative explanation involves the essential role of mitochondria in the formation of iron-sulfur clusters, which perform catalytic and structural functions in many cellular proteins, among them DNA repair proteins, and as was recently shown, the maturation step of these proteins is required for the maintenance of nuclear genome integrity 
. It is also possible that the imbalance in cytosolic dNTP pools due to mitochondrial dysfunction leads to chromosomal instability, as shown in human cells by Desler et al. 
. In agreement with the last explanation is the observation that among deletion strains displaying genome instability is a group defective in dNTP biosynthetic pathways (ADE3, ADE8, HIS1, RNR3
). Whatever the mechanism, the experimental data show that intact mitochondria are crucial for preservation of genomic integrity.
Many genes identified in the screen encode molecules located in vesicles, suggesting the participation of a vesicular path in the response to endogenous genotoxic stress. It is possible that response to stress requires the redistribution of protein(s) to an appropriate compartment. A number of genes whose products were connected with spindle pole body, bud neck, cytoskeleton and cellular wall were also found; these are likely to be engaged in proper cell division.
Genome-wide SLM screen reveals genes whose products are involved in various mechanisms assuring genome stability as well as numerous genes unassigned to any biological process within the cell
The Gene Ontology (GO) annotations indicate that the most abundant group identified in our screen has not been assigned previously to any biological process (). This suggests that our knowledge concerning the maintenance of genome stability in diploid cells is rather incomplete and substantiates the motives that encouraged us to undertake this study. On the other hand, the known annotations of the remaining gene groups confirm the correctness of our experimental approach. Our data point to numerous molecular processes engaged in genome maintenance. As was expected, many genes encoding proteins engaged in DNA replication and repair (ABF2, CGI121, DPB3, DUT1, KRE29, MPH1, MSH6, PBP2, RAD1, RAD5, RAD9, RAD24, RFC5), cell cycle regulation (BFA1, CDC16, HSL7, MAD1, NDD1, VHS1) and cell division (AKL1, BUD3, DDC1, DOM34, IML3, MCD1, LGE1, MPS3) have been revealed. We have also identified a significant group of gene deletions that influence the chromatin state (ELF1, RLF2, RSC4, RSC9, SIF2, SWR1, VPS72), which in turn destabilizes genome integrity, because maintenance of chromatin assures chromosome stability.
Another interesting group of genes revealed by our screen are DDC1
. These genes encode proteins that are either already documented or potential substrates for Cdc28p cyclin-dependent kinase, which, as recently shown by Enserink et al. 
, regulates proteins involved in DNA damage response and genome maintenance.
In addition, we have found a sizable group of genes whose products are involved in cellular stress responses (FRT2, SGD1, AHP1, ALO1, GPX2, OCA1, RIM15, YBR014C, AFG2, BLM10, PHM6, SSD1, TPS1, PRM9, GCN2, HSP26, SSA2,). Dysfunction in the stress response affects the ability of the cell to deal effectively with emerging problems that, as a natural consequence, manifests in genome destabilization.
The genome-wide SLM screen reveals the components of ‘structural maintenance of chromosome’ (SMC) complexes
Among the gene products revealed by our genome-wide approach we found some that have especially drawn our attention. We found MCD1
genes on microarray output list. These three essential genes encode subunits of three different complexes involved in assembling proper chromosome structure: cohesion complex, condensin complex and Smc5,6 complex, respectively. Two of these three ‘structural maintenance of chromosome’ (SMC) complexes directly regulate chromosome dynamics. The third, Smc5/6, functions mainly in homologous recombination and in completing DNA replication 
. However, upon a double-strand break (DSB), cohesin complex is recruited to the DSB region through phosphorylation of H2AX and binding of another SMC complex, MRX (Mre11, Rad50, Xrs2) to the break site 
. As can be expected, mutations affecting these complexes lead to chromosome aberrations. This phenotype has been shown mainly in meiotic cells, which demonstrate unequal division of genetic material, but for some mutations in SMC related genes, it has been also shown that they may cause aneuploidy in mitotic cells 
. The fact that strains depleted in genes encoding essential subunits of different SMC complexes appeared in the screen for LOF mutator genes made us curious why other subunits engaged in building these complexes did not appear. Examination of the whole dataset revealed that some of the genes were missing because the strength of the deletion phenotypes caused the disappearance of the respective clones from the analyzed population. Others were present and displayed a mutator phenotype in high throughput screens, but at lower significance than the selected cut-off value. Comparison of the microarray data with the individual tests done on a small sample of clones that had a high mutator score in the microarray screen, but with too high a p-value, indeed revealed a quite good correlation. Hence, we decided to search all our microarray data, including those rejected because of a high p-value, for other components of SMC complexes. The results are presented in supplementary Figure S8
. One can see the representation of all known SMC complexes, which regulate higher-order chromosome structure: cohesion complex (MCD1, SMC1, SCC3
), condensin complex (BRN1, SMC4, YCG1, YCS4
), Smc5,6 complex (KRE29, NSE3, NSE5, SMC5
) and finally MRX complex (XRS2, RAD50
) engaged in DSB repair. Further analysis revealed also other genes from SLM screen, encoding proteins responsible for physical interaction with cohesion Ctf4 protein, which binds also to Pol1 allowing it to access DNA (CTF4, POL1
) and Ctf18-replication factor C (CTF18, CTF8, RFC5
), which loads proliferating cell nuclear antigen (PCNA) on DNA. PCNA functions as a sliding clamp for replicative DNA polymerase and as a docking site for other proteins required for DNA replication and repair. We also noted the Rad24-replication factor C and its DNA binding partner from the 9-1-1 complex (RAD24, RFC5, DDC1
), which form a platform enabling DNA polymerases to access the DNA template at the site of damage. We also observed DPB3
encoding DNA polymerase-ε major subunit. Depletion of this gene is known already to have a mutator phenotype. These results show not only the involvement of SMC complexes in the maintenance of genome stability but, in addition, through their various interactions, suggest possible mechanisms of emergence of DNA alterations.
Escape from rearrangement catastrophe through conversion to haploid
In light of these remarks the appearance of CTF18 among the genes whose deletion shows the strongest mutator phenotype was not surprising. Unexpectedly, many of those deletion strains appeared as haploids residing within the homodiploid collection. If those arose as false positives due to strain misplacement their presence in our dataset would undermine the credibility of our results. However, we were able to prove that the lack of those genes in diploid yeast cells does result in the mutator phenotype. We also showed that the mutator phenotype of the deletion of CTF18 is manifested by the conversion of diploid strain into a haploid. Thus it is likely that the absence in diploid yeast cell of genes such as CTF8, TED1, MTO1 and PHM6 (and possibly as yet undiscovered genes), leads to diploid to haploid conversion by the same unknown mechanism. Now the most important question is what is that mechanism?
The mutator phenotype arising from the absence of MTO1, TED1 and PHM6 genes, and the existence of respective deletion strains as haploid in homodiploid collection indicates the excessive incidence of genomic DNA abnormalities when those genes are missing. Remarkably, they have not been previously linked to genome maintenance processes.
gene encodes a phosphoesterase domain-containing protein that acts in endoplasmic reticulum to Golgi vesicle-mediated transport 
. It is one of many genes engaged in vesicular trafficking that appeared in our screens and we discussed this matter earlier.
Phm6 is a protein of unknown function, whose expression is regulated by phosphate levels. While the link between phosphate homeostasis and genome stability is unclear at the moment it was shown that several phosphate regulated proteins, like Pho80, Pho85 and Pho4, together with Rad9, Rad53 and Cdc28, are employed in activation of checkpoint response on DNA damage in G1 phase of the cell cycle 
. In one of the early transcriptome studies PHM6
, together with CTF19
encoding the component of the kinetochore, were listed as responding to PHO regulatory pathway and possessing Pho4 binding sites on their promoters 
. It might also be that phosphate metabolism influences the levels of intracellular nucleotide triphosphate pools 
or that there is an interconnection between phosphate levels and the synthesis of pyridoxal 5′-phosphate (PLP). The results of a recent genome-wide study showed clearly that PLP levels are crucial for GCR suppression by curtailing the appearance of DNA lesions during the cell cycle 
. In any case our data support the hypothesis that there is a functional link between the metabolism of this crucial nutrient and the genome stability.
Mto1 is a mitochondrial protein. It forms a heterodimer complex with Mss1 that performs the 5-carboxymethylaminomethyl modification of the wobble uridine base of mitochondrial tRNAs 
. In mto1Δ
strain the levels of many classes of mitochondrial tRNA are significantly lowered. The critical role of Mto1 in modifications at U34 of tRNA-Lys, tRNA-Glu, and tRNA-Gln, in mitochondrial 21S and 25S rRNA stability, in translation of COX1
mRNAs, in the maintenance of mitochondrial genome, and subsequently in respiratory competence, has recently been demonstrated 
. The chain of events starting with wobbling tRNA deficiency causing the absence of crucial mitochondrial proteins ultimately results in the loss of mitochondrial DNA. This, as we discussed earlier, would compromise the stability of the nuclear genome. Even though the increase of SLM in freshly made diploid mto1Δ
strain is modest (see ), it increases with time (data not shown). Moreover, we saw a decrease in the sporulation frequency (see ) and we noticed the increased frequency of petite colonies during the construction of mto1Δ
strains (data not shown). So it is conceivable that this phenotype, relatively weak soon after the deletion of the gene, may grow stronger leading to chromosomal rearrangements and haploidization after sufficient number of generations.
was previously shown to lose individual chromosomes easily 
. It has also been shown that CTF18
deficient strain is unable to grow as a tetraploid at restrictive temperature so it was referred to as ploidy-specific lethal mutation 
. Recently it has been shown that Ctf18 interacts physically with DNA polymerase ε, origin recognition complex, Cdt1 and minichromosome maintenance proteins, which suggests important role of Ctf18 in regulating the initiation of DNA replication 
encodes a major subunit of the Ctf18-replication factor C (see supplementary Figure S8
) that loads PCNA sliding clamp on DNA, interacts with cohesion complex and is involved in chromosome segregation during cell division 
. Thus, the absence of Ctf18p will likely cause severe chromosomal aberrations 
. Yet to our knowledge, the phenomenon of losing an entire chromosome set from a diploid cell as a consequence of lack of CTF18
, or any other gene, was never reported.
The phenomenon of ploidy loss was in fact reported but for tetraploid strains of C.albicans 
. Recently, it has also been shown that after several hundred generations, ploidy reduction towards diploidy occurs also in both triploid and tetraploid lines of S. cerevisiae 
. The data presented in those papers suggest that the chromosome loss was not random but rather that full sets of chromosomes were lost at once. These results imply the existence of a mitotic mechanism allowing the elimination of an entire set of chromosomes in yeast, thereby reducing the ploidy level. Interestingly, polyploidy reduction observed in those studies always led to diploid cells. On the other hand, it has been shown that after sufficient number of generations haploid strains can also convert into diploids. In that case the conversion process requires more time, occurring after about 1800 generations 
. The results of those studies clearly show that the diploid state is a favorable one for standard laboratory S. cerevisiae
strain maintained in typical conditions.
The phenomenon that we have found for ctf18Δ/ctf18Δ strain is quite different. One can notice two alternative routes that differ in cell destiny: either the cells reduce the ploidy of their genome to the 1c level, which seems to be stable, or GCR in the genome will continue resulting in a very heterogeneous population of cells varying in their level of polyploidy or aneuploidy as well as in their viability. Remarkably, these changes are accompanied by an additional phenotype regarding SLM. When the cells continue to accumulate the rearrangements their average genome size increases and SLM remains high. Whereas, when the cells manage to reduce the ploidy of their genetic material (thus minimizing the possibility of rearrangements), SLM is diminished thereby increasing their chances for survival (see and ). This is documented by the domination of the population by haploid cells in two clones of ctf18Δ/ctf18Δ genotype and two clones of ndt80Δ/ndt80Δ ctf18Δ/ctf18Δ genotype. It is further substantiated by our observation that the cells in those cultures had on average shorter doubling time and higher survival rate than the cells from the remaining cultures.
Therefore we postulate that the reduction in ploidy from 2c to 1c by the cells devoid of functional Ctf18 is not accidental but rather is a new mechanism of avoiding the severe condition of genomic instability. We envision this phenomenon of conversion into haploid as a route to escape from rearrangement catastrophe. The mechanism governing this process remains to be explained, but our data clearly indicate that it is triggered by the deficiency of Ctf18 protein. Several conjectures can be made about this phenomenon on the basis of our current knowledge. S. cerevisiae can grow vegetatively both as haploids and diploids. The fact that the rate of GCR events in diploids is so much higher than in haploids suggests that under the risk of severe DNA damage discarding of the extra genome may act in favor of the genome preservation and sufficiently outweigh the disadvantage of short term lack of genetic heterogeneity and other benefits of diploidy. This can easily be reestablished by conjugation once the stress conditions disappear. Building up of GCR during prolonged exposure to environmental stress would lead to so extensive rearrangements and aneuploidy, such that the disposal of precisely one chromosome set would be impossible. Therefore successful escape from rearrangement catastrophe should be undertaken soon after the conditions that triggered it as suggested by our results. Since haploid cells dominate the ctf18Δ/ctf18Δ population after as little as 50 to 100 generations they must have appeared quite early.
Two possibilities present themselves. This phenomenon might occur purely by chance, starting with an early sporadic event of losing an exact chromosome set as a direct result of the absence of CTF18
gene. Alternatively, it may be an adaptive mechanism, encoded by some other genes, that increases the likelihood of survival of a cell subject to severe DNA abnormalities caused by the absence of CTF18
gene. Ctf18 is engaged in double-strand break repair by homologous recombination 
, a biological process involving mitotic sister chromatid cohesion 
. Absence of this protein leads to extensive aneuploidy clearly documented by our DNA content analysis. It is difficult to imagine how the diploid cell devoid of Ctf18 could lose whole chromosome set at once accidentally. A more likely possibility would be the gradual decrease of DNA content in such cells, but this is not what we see; there is either rapid conversion to haploid or gradual randomization of the DNA content drifting to values higher than 2n. While at first it seems difficult to accept that the mechanism of escape from rearrangement catastrophe through haploidization is adaptive, to us it is not unlikely and moreover, it sounds very appealing, especially considering that haploidization occurred by exactly the same means in separate cultures of clones lacking Ctf18. The ultimate mechanism must be based on experimental evidence; if one assumes that haploidization is adaptive, then it must have evolved in response to natural DNA abnormalities. What kind of naturally occurring stress resulting in conversion of diploid into haploid is imitated by CTF18
and possibly also by MTO1
gene deletions? Is this phenomenon unique to diploid S. cerevisiae
cells lacking Ctf18 protein or it is more general strategy of survival of diploid microorganisms in a hostile environment? These are important questions that should be resolved experimentally in a separate study.
Despite the distinctive phenotypes of their deletions MTO1
, identified with our approach, did not show up among the genes selected in two other genome-wide screens aiming at similar phenotypes, both employing crosses with diploid YKO collection strains: searching for diploid bimater strains 
and looking for gene deletions that restore mating competence to diploid strains 
. Only ctf8Δ/ctf8Δ
from our list of haploid strains in diploid YKO collection were reported in those studies. It is therefore possible that the list of genes whose deletion results in 2c to 1c conversion is incomplete. Further genome-wide screens designed specifically for selection of haploids within homodiploid collection may reveal more genes with a role in genome stability, whose deletion results in a specific ploidy reduction. In addition, they will help to determine the overall quality of the homodiploid S. cerevisiae
knock-out collection. Such experiment would certainly be useful for anyone using the collections. Regardless of the results of those screens, the performance of diploid ctf18Δ/ctf18Δ
and other deletion strains of similar phenotype strongly suggest the need for redefining the ‘essential’ gene attribute. For practical reasons this category should also encompass the genes like CTF18
. After several generations, strains carrying such a gene deletion accumulate so many secondary changes in its genome they are no longer the same strain. Effectively, the deletion of such genes does not permit the strain to exist in its original state, so in a sense that gene could be called ‘essential’. Alternatively, separate category could be established e.g. ‘genetically unstable’ to emphasize the characteristic of those deletion strains.
In summary, the genome-wide SLM screen that we have designed is a powerful tool for investigating genome stability. We were able to find genes responsible for maintaining genome integrity of diploid cells. Our screen revealed a genetic instability phenotype of 59 strains associated with the deletions of uncharacterized or dubious ORFs. This implies the existence of new molecular functions and possibly new processes involved in genome maintenance. We have also found functional associations with genome integrity of many well characterized genes that were not previously linked to this process; the suggested mutator phenotype of the deletion had never been shown in a direct assay. Moreover we showed that the lack of some genes made the diploid yeast cells to display an exceptional phenotype, a tendency of conversion to haploid. We believe that our results revealed novel mechanism involved in the genome stability that helps the cell to survive the rearrangement catastrophe.