A Whole Genome Screen for Minisatellite Stability Genes in Stationary-Phase Yeast Cells

Repetitive elements comprise a significant portion of most eukaryotic genomes. Minisatellites, a type of repetitive element composed of repeat units 15−100 bp in length, are stable in actively dividing cells but change in composition during meiosis and in stationary-phase cells. Alterations within minisatellite tracts have been correlated with the onset of a variety of diseases, including diabetes mellitus, myoclonus epilepsy, and several types of cancer. However, little is known about the factors preventing minisatellite alterations. Previously, our laboratory developed a color segregation assay in which a minisatellite was inserted into the ADE2 gene in the yeast Saccharomyces cerevisiae to monitor alteration events. We demonstrated that minisatellite alterations that occur in stationary-phase cells give rise to a specific colony morphology phenotype known as blebbing. Here, we performed a modified version of the synthetic genetic array analysis to screen for mutants that produce a blebbing phenotype. Screens were conducted using two distinctly different minisatellite tracts: the ade2-min3 construct consisting of three identical 20-bp repeats, and the ade2-h7.5 construct, consisting of seven-and-a-half 28-bp variable repeats. Mutations in 102 and 157 genes affect the stability of the ade2-min3 and ade2-h7.5 alleles, respectively. Only seven hits overlapped both screens, indicating that different factors regulate repeat stability depending upon minisatellite size and composition. Importantly, we demonstrate that mismatch repair influences the stability of the ade2-h7.5 allele, indicating that this type of DNA repair stabilizes complex minisatellites in stationary phase cells. Our work provides insight into the factors regulating minisatellite stability.

Various types of repetitive noncoding DNA exist in abundance within eukaryotic genomes. Categorized by repeat unit size, repetitive elements consist of units that can range from one nucleotide to hundreds of nucleotides in length (reviewed in Debrauwere et al. 1997). Minisatellites (also known as variable number of tandem repeats) are classically defined as repetitive tracts of DNA consisting of repeat units that are specifically 152100 bp in size (Vergnaud and Denoeud 2000). These repetitive elements are predominantly stable in actively dividing cells yet change in repeat length as well as in composition during meiosis (Jauert et al. 2002;Richard et al. 2008).
Minisatellites have been shown to perform several important biological functions. These functions include regulating gene transcription (Cohen et al. 1987;Krontiris 1995;Spandidos and Holmes 1987) interfering with gene splicing (Turri et al. 1995), acting as chromosomal fragile sites (Lukusa and Fryns 2008;Yu et al. 1997), and influencing chromosomal pairing during meiosis (Ashley 1994;Chandley 1989). A well-characterized example of minisatellite genomic function is that associated with the human HRAS1 oncogene (Capon et al. 1983). This minisatellite is composed of nonidentical repeat units that are each 28 bp in length and have a high GC content (68%). Each repeat varies, with either a C or G at nucleotides +14 and +22 (numbered relative to the first nucleotide in the repeat). The HRAS1 minisatellite is located 39 of the HRAS1 open reading frame (ORF) and acts as a binding site for the rel/nuclear factor-kB family of transcription factors (Trepicchio and Krontiris 1992). Altered minisatellites exhibit enhancement of HRAS1 transcription (Spandidos and Holmes 1987;Krontiris 1995), indicating minisatellites can significantly influence the expression of nearby genes. Altered human minisatellites are associated with an increased risk of myoclonus epilepsy (Lafreniere et al. 1997;Virtaneva et al. 1997), diabetes mellitus (Kennedy et al. 1995), asthma (Kirkbride et al. 2001), attention deficit-hyperactivity disorder (Faraone et al. 2001;Yang et al. 2007), and several different types of cancer (Calvo et al. 1998;Krontiris 1995;Rosell et al. 1999;Vega et al. 2001;Weitzel et al. 2000). The presence of rare altered alleles of the HRAS1 minisatellite correlates with tumors of the lung, bladder, ovaries, and brain and have been isolated from the primary tumors of patients with breast cancer (Devlin et al. 1993;Ding et al. 1999;Rosell et al. 1999;Vega et al. 2001;Weitzel et al. 2000).
The majority of human cells exist as a population of nondividing, quiescent cells that are contact and growth inhibited. Cancer formation in eukaryotic organisms requires loss of these inhibitory mechanisms. An initial oncogenic event can result in genomic instability within a quiescent cell, promoting uncontrolled re-entry into the cell cycle, leading to tumorigenesis (Jin et al. 2009;Kim et al. 2005;Suda et al. 2005). At present, little is known about genomic instability events in nondividing cells. The yeast Saccharomyces cerevisiae can serve as a model organism for the study of genomic instability in the context of a quiescent cell population, as yeast can enter a nonmitotic state known as stationary phase (Gray et al. 2004;Werner-Washburne et al. 1993) that mimics several key characteristics associated with mammalian quiescent G 0 cells, including a reduced level of gene expression and condensed unreplicated chromosomes.
We previously developed a colony color segregation assay to monitor minisatellite repeat alterations in yeast (Kelly et al. 2007(Kelly et al. , 2011(Kelly et al. , 2012. We inserted either a minisatellite consisting of three 20bp repeat units and a 5-bp linker (the ade2-min3 allele) or a minisatellite consisting of seven-and-a-half variable repeats of the HRAS1associated minisatellite (the ade2-h7.5 allele) into the ORF of ADE2 (Figure 1, A and B). These insertions create frameshifts, resulting in Ade 2 red colonies that require supplemental adenine for growth. The red pigment is produced as a byproduct of the disruption of the adenine biosynthetic pathway (Smirnov et al. 1967). Loss of a repeat unit or gain of two repeat units within the minisatellite restores the ADE2 reading frame, rendering cells white and Ade + . Minisatellite alterations that occur in stationary phase cells after colony formation is complete lead to the formation of white microcolonies ("blebs") that arise on the surface of the red colony. This phenotype allows us to easily detect minisatellite alterations that occur in cells within the postmitotic cellular population.
We previously used the ade2-min3 reporter construct to identify mutants that increase stationary-phase minisatellite instability. An ultraviolet (UV) mutagenesis screen identified several mutations in END3 (Kelly et al. 2012) and in the zinc homeostasis genes ZRT1 and ZAP1 (Kelly et al. 2007). Minisatellite alterations were specific to the quiescent population of stationary-phase cells and were independent of adenine auxotrophy or chromosomal context (Kelly et al. 2011). Thus, zinc homeostasis is essential for maintenance of minisatellite stability during the stationary phase possibly due to disruption of zinc-dependent DNAbinding proteins involved in DNA stability, specifically in stationaryphase cells. Minisatellite tract alterations in these mutants require homologous recombination, and multiple pathways act to maintain minisatellite stability in stationary phase cells (Kelly et al. 2011(Kelly et al. , 2012. Genes involved in these pathways include END3, RAD27, PKC1, ZRT1, and ZAP1. More recently, we used a modified version of the synthetic genetic array (SGA) protocol (Li et al. 2011;Tong and Boone 2006;Tong et al. 2001) to screen all of the genes annotated for checkpoint function and found that a subset of those genes also influence stationary phase minisatellite stability (Alver et al. 2013). Here, we used our modified SGA protocol to screen the entire yeast genome in an effort to: (1) determine what factors are involved in regulating minisatellite stability in stationary phase cells and (2) determine whether minisatellites varying in repeat composition and size are regulated by different factors in stationary phase cells.

Media and yeast strains
The solid and liquid media used in this study were prepared as stated in (Guthrie and Fink 1991). All media that was used in the SGA analysis was prepared as described in (Tong and Boone 2006;Tong et al. 2001) with the exception of presporulation media that contains 2% agar, 5% dextrose, 1% Difco yeast extract, and 3% Difco nutrient broth. Media consisting of Geneticin (G418) was prepared by adding 200 mg/L G418 sulfate (Cellgro).
The plasmids and S. cerevisiae strains used for this study are listed in Table 1. All strains, except those used to perform the SGA analyses, were derived from DTK271 (MATa his7-2 leu2::HisG trp1-289 ura3-52 ade2-min3) (Kelly et al. 2007), in the AMY125 background. Yeast strains bearing deletions of nonessential genes were constructed by isolating genomic DNA from the G418-resistant nonessential Yeast Deletion Strain Haploid Set. Polymerase chain reaction (PCR) products containing the KANMX4 gene (G418 resistance) and 59 and 39 regions of genomic homology were constructed using the primers listed in Table 2. The parental strains were transformed with the PCR product, and the transformants were selected on yeast extract-peptone-dextrose (YPD) + G418 solid media and verified by PCR.
We constructed the ade2-min3 query strain DTK893 (MATa his3-1 ura3-0 can1::MFA1pr-spHIS5 ade2-min3 -URA3MX) using a two-step PCR process as previously described (Alver et al. 2013). To summarize, the plasmid pDC369 was used to generate a URA3MX PCR product with flanking sequence using primers 14193004 and 14193005. DTK271 genomic DNA was used as a PCR template to isolate the ade2-min3 allele using primers 14193006 and 14193007. To combine the URA3MX and ade2-min3 PCR products, we performed PCR by using primers 14193007 and 14193008 resulting in the ade2-min3 URA3MX-linked cassette that was transformed into DCY2556. Transformants were selected on solid synthetic media lacking uracil (SD-Ura), resulting in DTK893. DTK893 and DCY2557 were mated, resulting in DTK1175, which was sporulated and dissected. Mating, sporulation, and tetrad dissection were performed as described previously (Jauert et al. 2002). Red spores were selected on solid SD-Ura media. Query strains for the SGA analyses were isolates of DTK1175: DTK1189 5a = MATa derivative for the nonessential SGA and DTK1189 2b = MATa derivate for the essential SGA. pDC369, DCY2556, and DCY2557 are from Dr. Duncan Clarke, University of Minnesota.

SGA analysis
Nonessential SGA: For our study, we performed a modified SGA analysis as described in our previous work (Alver et al. 2013;Tong and Boone 2006;Tong et al. 2001) ( Figure 1C). To summarize, we inoculated YPD liquid media with a single red colony of query strain DTK1189 5a or DTK1624. Cultures were plated onto YPD solid media after overnight incubation at 30°. The MATa nonessential Yeast Deletion Strain Haploid Set (Invitrogen; from Dr. Robin Wright, University of Minnesota) was pinned onto the query strain in a 96-well format ( Figure 1D). Mated strains were incubated at 30°overnight. The resulting zygotes were pinned to SD-Ura + G418 solid media and incubated at 30°overnight. Diploids were then pinned to presporulation solid media, incubated at 30°overnight, and then pinned to sporulation media and incubated at room temperature (RT) for 6 d. Haploid MATa progeny were selected on synthetic media lacking histidine, arginine, and uracil (SD-His/Arg/Ura) and containing canavanine (US Biological). Strains were incubated at 30°overnight. This step was repeated. Strains were then pinned to SD-His/Arg/Ura + canavanine + G418 media, and this step was repeated. The haploids were then pinned to YPD and left at RT for 5 d. Each plate was pinned in duplicate, and the screen was repeated three independent times. A positive control consisting of a strong blebbing strain (a MATa zrt1Δ mutant; SCD153) was included on each plate (Kelly et al. 2007). The blebbing produced by the resulting double mutants was rated on a scale of + to ++++ ( Figure 1D). Candidate hits were characterized as strains producing a strong degree of blebbing (+++ or ++++). The blebbing phenotype for each hit was verified by restreaking the strain onto YPD and assessing the individual yeast colonies for a blebbing phenotype on a scale of + to ++++.
Scores for the blebbing phenotypic analysis of the ade2-min3 and ade2-h7.5 nonessential SGA independent screens are listed in Supporting Information, File S1 and File S2, respectively. Duplicate plates from each screen are indicated by an "a" or "b" for each individual screen. The scores are represented numerically instead of by "+'s." For example, a "1" indicates that a specific ORF was assigned of score of "+" for a specific screen. A score of "0" indicates that a specific strain did not grow, and a score of "5" indicates that a strain was composed of white cells and therefore could not be assigned a rating for a blebbing phenotype. A hit was defined as a strain (corresponding to a specific ORF) that produced a level of blebbing of "324" (+++ to ++++) in both replicates for at least two of the three independent screens performed. A second criteria used to define a hit was a strain that was scored as a "3" or "4" (+++ or ++++) in both replicates for blebbing in combination with a score of "0" (no growth) or a "5" (white patch) for a least two of the three independent screens. Essential SGA: We followed a similar screening protocol as described previously and in our previous work (Alver et al. 2013). In summary, Figure 1 SGA analysis to screen for minisatellite instability. (A) The ade2-min3 allele consists of three identical 20-bp repeats plus one additional base pair inserted into the XbaI site of ADE2 resulting in a 5-bp overhang. Insertion of the minisatellite disrupts the ADE2 ORF. A gain of two repeats or loss of one repeat restores the ADE2 reading frame. (B) The ade2-h7.5 allele contains seven-and-a-half 28-bp repeats that differ at positions +14nt and + 22nt with a C or a G as well as flanking sequence unique to the HRAS1 locus. The minisatellite is inserted at the XbaI site of ADE2 and throws the gene out of frame. Loss of a repeat restores the open reading frame. (C) Summary of the modified SGA screen. A query strain bearing the ade2-min3 or ade2-h7.5 allele was mated to the entire S. cerevisiae nonessential deletion collection or an essential strain collection containing ts mutant alleles. At different steps, strains were pinned to various selective media to isolate the desired mutant strains. Final haploid strains bearing the minisatellite allele plus a deletion or ts allele were assessed for a blebbing phenotype. (D) Each screen was performed using a 96-well format. Strains on plates from the final double mutant selection step were scored for a blebbing phenotype on a qualitative scale of + to ++++ (++++ = level of blebbing produced by the zrt1D positive control). the query strain DTK1189 2b was mated to a MATa essential temperaturesensitive (ts) strain set containing 455 ts genes (from Dr. Charles Boone, University of Toronto) (Li et al. 2011) and incubated at RT for 2 d. Zygote selection, presporulation, and sporulation were performed as described previously with the exception that strains were incubated at RT. Haploid and double mutant selection were as described previously with strain incubation being carried out at RT for 2 d. After selection, the double mutant haploids bearing the ade2-min3 allele were pinned to five separate YPD solid media plates, and the plates were incubated for 5 d at a range of temperatures: 26°, 30°, 32°, 34°, or 37°. Blebbing was scored as described previously. Strains producing a +++ to ++++ level of blebbing were restruck onto YPD and incubated at the corresponding temperature to verify the blebbing phenotype. Restruck strains were again rated on a scale of + to ++++. As noted previoulsy, a MATa zrt1Δ haploid mutant (SCD202) was used as a positive control (Kelly et al. 2007).
Phenotypic scoring for each independent screen is shown in File S3. As described previously, the scores are represented numerically rather than by "+s." Each score represents the maximum level of blebbing assigned to a specific ORF. Specifically, all of the scores assigned to each individual ts allele of a particular ORF were compared at each individual temperature. The highest blebbing score of all of the alleles and all of the incubation temperatures was assigned to the corresponding ORF for each independent screen. As described previoulsy, a hit was defined as an ORF that produced a level of blebbing of "324" (+++ to ++++) for at least two of the three independent screens performed or as an ORF that produced a level of blebbing of "3" or "4" (+++ or ++++) in combination with a score of "0" (no growth) or "5" (solid white patch) for at least two out of the three independent screens.
Gene Ontology (GO) term analysis GO data used ontology version 1.2 with annotations downloaded on January 22, 2012. Enrichment calculations are based on a hypergeometric overlap test and reported P-values have been Bonferronicorrected for multiple hypotheses (Boyle et al. 2004).

Blebbing quantification assay
Yeast strains used for the quantification assay were streaked onto solid YPD media and incubated at 30°for 2 d. A single red colony was used to inoculate 5 mL of liquid YPD media, and the cultures were incubated at 30°for 4 h. Cultures were diluted and plated onto solid YPD media. Strains were incubated at 30°for 2 d and then left at RT for 6 d to allow for bleb formation. For each strain, blebs were counted on at least 100 colonies ranging in size from 1.26 to 1.32 mm in diameter. This assay was repeated three independent times whereupon the average number of blebs per colony as well as the 95% confidence interval of the mean was calculated.
Our assay allows us to distinguish between minisatellite alterations that take place during mitotic growth and those that occur during the stationary phase. Alterations occurring in actively dividing cells lead to a red/white sectoring color segregation phenotype, whereas those occurring in stationary-phase cells result in a novel phenotype known as blebbing (Kelly et al. 2007) in which white microcolonies form on the surface of a red colony. This phenotype allows us to directly identify factors that regulate the stability of a minisatellite in a postmitotic population of yeast cells. We recently used a modification of the SGA protocol to examine genes annotated as having a checkpoint function in yeast and determined that a subset of these proteins affect the stability of the ade2-min3 minisatellite tract in stationary-phase cells (Alver et al. 2013). This pilot study showed that an automated approach was not possible, as any white Ade + cells that arose during the procedure rapidly overwhelmed the red Ade 2 cells. To compensate, we modified the SGA protocols to manually screen for mutants that produced a blebbing phenotype ( Figure 1C) (Alver et al. 2013;Li et al. 2011;Tong and Boone 2006;Tong et al. 2001).
Using the ade2-min3 reporter, we identified 102 candidate genes that, when mutated, resulted in a strong blebbing phenotype (Table 3). These included the zinc homeostasis genes ZRT1 and ZAP1 identified in our original UV mutagenesis screen (Kelly et al. 2007), demonstrating that our SGA analysis was working correctly. Also included were the genes PKC1, an essential gene that encodes for a serine/ threonine kinase that regulates cell wall modifications (Levin et al. 1990), and RAD27, which encodes for a nuclease that processes Okazaki fragments (reviewed in Liu et al. 2004a,b). Both genes were shown by our laboratory to be important for maintaining minisatellite stability in stationary phase cells (Kelly et al. 2012). Finally, we previously identified checkpoint-related genes including MRC1, CSM3, and TOF1 by SGA analysis (Alver et al. 2013); those genes are marked in Table 3.
To determine whether any enriched GO terms were represented within the 102 candidate hits, we performed GO term analysis using Ontology version 1.2. We find that genes associated with GO terms representing chromosomal maintenance and DNA replication are overrepresented within our list of 102 hits (Table 4), indicating these cellular processes are likely to be important for regulating the stationary phase stability of the ade2-min3 minisatellite.
Of note were three hits from the ade2-min3 SGA analyses that were represented in both the nonessential and the essential mutant strain sets. These hits included the stress response kinase gene DBF2 (Johnston and Thomas 1982), the checkpoint gene SLI15 (Kim et al. 1999;Ruchaud et al. 2007;Sandall et al. 2006), and the SNARE complex gene SEC22 (Liu and Barlowe 2002). Deletion of DBF2 and SLI15 in the nonessential strain set resulted in strains that produced a low level of blebbing. However, ts alleles of both strains were identified as hits in our screen. Our results suggest that aberrant alleles of either gene are more detrimental to minisatellite stability than complete loss of the gene itself. We found the opposite effect for the gene SEC22, which was identified as a hit when deleted but not as a ts allele. Therefore, variations in gene product expression or function appear to differentially affect ade2-min3 minisatellite stability.
Of the 102 genes identified in our screen, only three produced a high level of blebbing (++++) for at least two of the three independent analyses that were performed (Table 3). These included the zinc transporter gene ZRT1 (Zhao and Eide 1996), the mitochondrial transport gene ODC1 (Palmieri et al. 2001), and SAC1, a gene encoding for a lipid phosphatase involved in protein trafficking and cell wall maintenance (reviewed in Strahl and Thorner 2007). These results indicate that aberrant cellular trafficking, such as that involved in regulating intracellular protein and zinc levels, can drastically affect the stability of a minisatellite allele in stationary phase yeast cells.
Identification of factors that maintain the stability of the ade2-h7.5 HRAS1-associated minisatellite allele in stationary-phase yeast cells Although the ade2-min3 allele allows us to readily assess mutant strains for minisatellite instability, its composition does not mimic most minisatellites found within human cells. The majority of human minisatellites are composed of long repeat units that are variable in sequence composition and are significantly enriched for GC content (Vergnaud and Denoeud 2000). The ade2-min3 minisatellite is comprised of identical repeats with only 50% GC content. Because of these differences, human minisatellites may be regulated by different factors than those associated with the ade2-min3 allele, a short minisatellite with identical repeat units.
To examine these potential issues, we used a second minisatellite reporter (the ade2-h7.5 allele; Figure 1B), which incorporates a portion of the complex human HRAS1 minisatellite into the ADE2 gene (Kelly et al. 2007(Kelly et al. , 2011. This tract is composed of seven-and-a-half repeat units that are 28 bp in size, is variable in base content at positions +14nt and +22nt, and has a GC content of 68% (Kelly et al. 2011). Similar to the ade2-min3 assay system, insertion of the tract disrupts the ADE2 ORF, and loss of a 28-bp repeat (or gain of two repeats) will restore the proper reading frame. To identify the factors involved in regulating the stability of a complex human-associated minisatellite during stationary phase, we used the same SGA methodology described previously, screening the strain set three independent times. A caveat to using the ade2-h7.5 allele as a query strain for this screen is that the overall level of blebbing associated with this allele is approximately fourfold lower than that produced by ade2-min3 strains (Kelly et al. 2011); this increase in minisatellite stability makes phenotypic scoring more difficult.
We identified 157 candidate hits that result in a high level of blebbing when mutated (Table 5) and performed GO term analysis as described previously. Intriguingly, only one enriched GO term (GO:0016831 carboxy-lyase activity) was associated with this data set. Within this particular set of 157 hits, we found that four mutant strains maintained a very high level of blebbing throughout our ade2-h7.5 SGA analyses (++++ for at least two of the three screens performed; Table 5). These included the previously characterized zinc transporter genes ZRT1 and ZAP1 (Kelly et al. 2007(Kelly et al. , 2011Eide 1996, 1997), plus the uncharacterized gene encoded by the open reading frame YHR022C and the mismatch repair associated gene, PMS1 (Prolla et al. 1994a,b). We conclude that zinc homeostasis plays an important role in preventing ade2-h7.5 minisatellite instability and that mismatch repair influences minisatellite stability in stationary-phase cells.
Analysis of factors involved in maintaining the stability of both the ade2-min3 and ade2-h7.5 minisatellite alleles in stationary-phase cells Although the overlap of candidate hits between the ade2-min3 and the ade2-h7.5 screens was statistically significant (P = 0.047), only seven genes were identified as hits in both SGA screens (Figure 2A): the genes BUD28, COT1, MON1, YGL217C, YLR125W, ZAP1, and ZRT1. BUB28 and YGL217C are both dubious ORFs that are unlikely to code for a protein. Each ORF, however, overlaps the coding sequence of an adjacent characterized gene. Approximately 98% of the BUD28 ORF overlaps the ribosomal subunit gene RPL22A (Planta and Mager 1998) whereas YGL217C overlaps roughly 10% of KIP3, a gene that encodes for a kinesin-like protein involved in mitotic spindle positioning (Cottingham and Hoyt 1997). RPL22A was identified as a hit in our ade2-min3 SGA screen but not in the ade2-h7.5 screen. Nevertheless, it is possible that this gene was not identified in the ade2-h7.5 screen due to the difficulty in scoring each strain. Deletion of the ORF YGL217C is represented twice in the nonessential deletion haploid set; one deletion strain was identified as a hit in both screens, and the second was not. No other hits were duplicated within our strain sets. PCR analysis revealed that both of the YGL217C isolates from the strain collection and our SGA analyses are correct deletion mutants. Therefore, one or both isolates could contain a secondary mutation that enhances or suppresses the blebbing phenotype.
PCR analysis of the YLR125W deletion mutant from the nonessential haploid strain set as well as of the ade2-min3 YLR125W mutant from the final step of the SGA analysis revealed that each strain was wild-type (WT) at the YLR125W locus. Because both strains are G418 R , the KANMX PCR product used to construct the deletion collection parental strain must be located elsewhere in the genome, likely at an area of homology to the YLR125W locus. Future determination of the deletion's genomic location may give insight into the source of blebbing observed in this strain.
To determine whether any sets of genes were overrepresented in both the ade2-min3 and ade2-h7.5 hit lists, we performed GO term analyses on the combined results of each screen (Table 6). Each enriched GO term was primarily associated with genes identified in the ade2-min3 screen rather than those of the ade2-h7.5 screen. Therefore, the GO terms represented general terms associated with cellular chromosomes and DNA replication as discussed previously. Based upon the low number of overlapping hits between each screen and the results from our GO term analyses, we conclude that each minisatellite tract is regulated by a distinct set of genes that do not share many overlapping functions or processes.
To verify the results of the SGA analysis, we deleted these genes in a separate genetic background-our well-characterized ade2-min3 (DTK271) and ade2-h7.5 (DTK1188) strain background (Alver et al. 2013;Kelly et al. 2007Kelly et al. , 2011. We then quantified the blebbing phenotype in the resulting mutants. As previously reported, deletion of ZAP1 and ZRT1 in both ade2-min3 and ade2-h7.5 strain backgrounds resulted in a high level of blebbing compared with the WT parental strains (Figure 2, B and C) (Kelly et al. 2007(Kelly et al. , 2011. However, unlike the results from the SGA analyses, strains bearing a deletion of COT1, MON1, or RPL22A did not result in a dramatic increase in blebbing in either strain background. The deletion of COT1, MON1, or RPL22A in a strain bearing the ade2-min3 allele produced a level of blebbing that, although significantly greater than that of the WT strain, was only 30% of that displayed by the zrt1Δ strain (cot1Δ at 8.8 blebs/colony; mon1Δ at 9.7 blebs/ colony; rpl22aΔ at 6.7 blebs/colony vs. WT at 4.7 blebs/colony and zrt1Δ at 24.5 blebs/colony). An ade2-h7.5 strain bearing a deletion of COT1, MON1, or RPL22A did not produce a significant increase in blebbing compared with the WT strain (cot1Δ at 2.6 blebs/colony; mon1Δ at 2.6 blebs/colony; rpl22aΔ at 0.3 blebs/colony vs. WT at 2.0 blebs/colony). We suspect that these results could be due to a secondary mutation in the SGA strain background that could act as an enhancer, or in our laboratory strain background which could act as a suppressor of the blebbing phenotype.
Mismatch repair regulates ade2-h7.5 minisatellite stability in stationary-phase cells Previous work in actively dividing cells suggested that mismatch repair is associated with preventing microsatellite, rather than minisatellite, alterations (Sia et al. 1997). However, the deletion of PMS1 in the ade2-h7.5 strain resulted in a strong blebbing phenotype, indicating a potential role for mismatch repair in stationary phase cells (Table 5). PMS1 encodes a mismatch repair protein that, together with Mlh1p, repairs multiple forms of damaged DNA (Prolla et al. 1994a, b). To verify the results of our screen, we deleted PMS1 in our laboratory strain background (Kelly et al. 2011) and quantified the average number of blebs/colony. Deletion of PMS1 confirmed the results from the SGA analysis, as pms1Δ produced a significantly higher level of blebbing (13.4 blebs/colony) compared with the WT strain (2.0 blebs/colony; Figure 3).
To determine whether other well-characterized mismatch repair genes (reviewed in Marti et al. 2002) were involved in maintaining stationary phase minisatellite stability, we quantified the level of blebbing in several mismatch repair mutant strains bearing the ade2-h7.5 allele (Figure 3). Strains with a deletion of MLH1 (10.0 blebs/colony), MSH2 (13.8 blebs/colony), or MSH6 (14.1 blebs/colony) produced notably greater levels of blebbing compared with the WT strain. Deletion of EXO1, MLH2, MLH3, or MSH3 did not result in a level of blebbing significantly different from that of the WT strain. Together our results indicate that a specific subset of mismatch repair genes maintains the stationary phase stability of the variable-repeat HRAS1associated minisatellite. The MLH1, MSH2, and MSH6 results also demonstrate that our SGA screens may not have identified all genes that contribute to minisatellite stability, possibly due to issues of sensitivity or strain background.
Only seven hits overlapped both screens. Finally, we demonstrated that mismatch repair genes regulate ade2-h7.5 minisatellite stability.
We investigated the hits of each screen independently to characterize candidate genes associated with each individual minisatellite. Several hits identified in the ade2-min3 screen were genes associated with checkpoint function and were described previously (Alver et al. 2013). Other identified genes involve DNA replication and repair. These included POL31, a subunit of Pold (Giot et al. 1997;Hashimoto et al. 1998;Sugimoto et al. 1995) and the Pole subunits DPB3 or DPB4 (Araki 1994;Araki et al. 1991;Lou et al. 2008) (Table 3). Also included were the RFC2 and RFC4 subunits of replication factor C (Cullmann et al. 1995;Noskov et al. 1994;Yao et al. 2003). Replication factor C is a clamp loader of the proliferating cell nuclear antigen, a sliding clamp for Pold and Pole (Chilkova et al. 2007). Both Pold and Pole have been implicated in DNA repair mechanisms. Pold has been shown to be involved in base excision repair (Blank et al. 1994), repair of UV-damaged DNA (Torres-Ramos et al. 1997), and template switching after DNA damage occurs (Vanoli et al. 2010). Pole has been implicated in nucleotide excision repair (Shivji et al. 1995), base excision repair (Wang et al. 1993), and double-strand break repair (Holmes and Haber 1999). GO term analysis of the 102 hits from the ade2-min3 screen revealed that the majority of enriched GO terms were associated with chromosomal regulation and DNA replication. Although it might seem surprising to find such strong evidence of DNA replication in stationary-phase cells, n Table 5 Summary of hits from the ade2-h7.5 SGA analysis of the yeast nonessential strain set Genes appearing in bold typeface are the strongest hits from the ade2-h7.5 screen (scored as ++++ for at least two of three independent screens). GO, Gene Ontology.
as bulk DNA synthesis does not take place within this population, previous work has shown that discrete areas of DNA replication do occur, likely at regions of localized DNA repair (de Morgan et al. 2010). Our results suggest that DNA replication is involved in preventing minisatellite alterations in stationary phase, potentially at sites of genomic repair. Additional replication and repair genes identified in our screen included the sumo-ligase gene MMS21 (Montelone and Koelliker 1995;Prakash and Prakash 1977) and RAD27, a flap endonuclease involved in base excision repair and double-strand break repair (Tseng and Tomkinson 2004;Wu et al. 1999). RAD27 has previously been implicated in minisatellite stability by a number of research groups (Kelly et al. 2012;Lopes et al. 2002). Together, our data suggest that components of DNA replication and repair mechanisms prevent ade2-min3 minisatellite alterations in stationary phase cells.
GO term analysis of candidate hits from the ade2-h7.5 SGA analysis showed that only one term (carboxy-lyase activity) was overrepresented within our data set. This finding could be a consequence of factors that regulate the stability of the ade2-h7.5 minisatellite representing a wide range of cellular functions, or the inherent low level of blebbing associated with the ade2-h7.5 allele (Kelly et al. 2011) complicating the accuracy of scoring this screen. Also, it is possible that uncharacterized candidate hits could be specific to stationary phase gene expression, but a comparison of the uncharacterized hits with genes known to be expressed in stationary phase cells revealed no strong correlation (Aragon et al. 2008;Davidson et al. 2011).
Interestingly, only seven gene hits overlapped between the ade2-min3 and ade2-h7.5 screens (Figure 2A): BUD28, COT1, MON1, YGL217C, YLR125W, ZAP1, and ZRT1. Deletion of each gene in our laboratory strain background revealed that, like the SGA analyses, deletion of ZRT1 or ZAP1 resulted in a dramatic increase in minisatellite instability in both the ade2-min3 and ade2-h7.5 alleles (Kelly et al. 2007(Kelly et al. , 2011. Deletion of COT1, MON1, or RPL22A (BUD28) led to a moderate increase in ade2-min3 instability but did not affect ade2-h7.5 instability. Thus, differences in strain backgrounds, such as the presence of a secondary enhancer or suppressor mutation, can significantly affect minisatellite stability in stationary phase cells. In support of this, the nonessential gene deletion strain collection previously has been shown to harbor secondary mutations (Lehner et al. 2007). The small degree of overlap between the screens further suggests that sequence differences between the ade2-min3 and ade2-h7.5 minisatellites may govern repeat tract stability and alteration. Our laboratory, as well as others, has previously shown that tract length and sequence variability within a minisatellite allele can greatly affect repeat stability (Denoeud et al. 2003;Jauert and Kirkpatrick 2005;Legendre et al. 2007). This study supports these findings and extends it to stationary phase cells, suggesting that the composition and size of the minisatellite affect which cellular components and mechanisms govern tract stability.
Surprisingly, a strong candidate hit identified in the ade2-h7.5 screen was the mismatch repair gene PMS1 (Prolla et al. 1994a,b). The mismatch repair system has usually been associated with instability in microsatellite tracts, rather than minisatellite tracts (Johnson et al. 1996;Strand et al. 1993Strand et al. , 1995. Minisatellite instability in actively dividing cells was not significantly affected by mutations in mismatch repair mutant strains (Sia et al. 1997), in an analysis that used a minisatellite tract identical to the ade2-min3 allele. Deletion of PMS1 affected only the stability of the ade2-h7.5 minisatellite allele rather than the ade2-min3 allele. Only a subset of the mismatch repair components (PMS1, MLH1, MSH2, and MSH6) affected the stability (Figure 3) of the ade2-h7.5 minisatellite tract. Previous work in actively growing cells demonstrated that mutating the mismatch repair genes MLH1, MSH2 or PMS1 resulted in a high degree of microsatellite instability, while mutating MSH3 or MSH6 had a less drastic effect (Johnson et al. 1996;Sia et al. 1997;Strand et al. 1993Strand et al. , 1995. This and other work suggested that yeast contain distinct mismatch repair complexes; one complex contains the Mhl1-Pms1 and Msh2-Msh6 heterodimers, whereas another incorporates Mlh1-Pms1 and Msh2-Msh3 (Johnson et al. 1996;Marsischky et al. 1996). Each complex has been implicated in the repair of different substrates; the Msh2-Msh6 heteromer is involved in base:base mismatch repair as well as repair of single-base loops generated by insertion or deletion mispairing (Alani 1996;Marsischky et al. 1996) whereas Msh2-Msh3 primarily targets small loops (Habraken et al. 1996; Marsischky et al. Figure 2 Summary of overlapping hits from the ade2-min3 and ade2-h7.5 SGA Screens. (A) Of the 102 hits obtained from the ade2-min3 SGA analysis and 157 hits obtained from the ade2-h7.5 SGA analysis, seven hits overlapped both screens. (B) Blebbing quantification of strains bearing the ade2-min3 allele. YPD cultures were inoculated with a single red colony and grown for 4 h at 30°. Each culture was diluted and plated onto solid YPD media. Strains were incubated at 30°for 2 d and then at RT for 6 d. Blebs were counted on 100 colonies. The average number of blebs +/2 the 95% confidence interval were calculated for each strain. This experiment was repeated three independent times. (C) Blebbing quantification of strains bearing the ade2-h7.5 allele was performed as in (B). 1996). Based upon our data, we predict that the Mlh1-Pms1, Msh2-Msh6 complex prevents the instability of the ade2-h7.5 minisatellite by targeting and repairing mismatches that can occur at two positions within a repeat. If there is misalignment of repeat units during recombinational repair (shown to be required for tract alterations) (Kelly et al. 2011), there will be potential C/C or G/G mispairs at nucleotides 14 and 22 in each repeat; these would be substrates for mismatch repair that could also affect tract length during the repair process. Our results lend further support to the idea that the composition of a minisatellite may dictate which factors are involved in preventing tract alterations within stationary phase cells.
In summary, we have conducted the first whole-genome screens to identify factors that regulate minisatellite stability in stationary phase cells, uncovering more than 250 genes that strongly affect the stability of minisatellite tracts in stationary-phase cells. We provide evidence that factors involved in regulating stationary-phase minisatellite stability are affected by minisatellite repeat length or sequence, as we found only a small overlap in detected genes when using significantly different minisatellite tracts. We find that disruption of DNA replication and repair components result in a dramatic increase in instability of a simple minisatellite tract, whereas loss of a subset of mismatch repair proteins specifically influences variable-repeat minisatellite instability. Thus, our work lends support to the argument that the composition of the repeat tract within a minisatellite greatly affects minisatellite stability and regulatory mechanisms.

ACKNOWLEDGMENTS
We are grateful to Robin Wright and Charlie Boone for the nonessential and essential yeast haploid strain sets, respectively. We thank Duncan Clarke for the plasmid pDC369 and the yeast strains DCY2556 and DCY2557. This work was supported by a grant from the National Institutes of Health (5RO1-GM072598) to D.T.K. B.V. and C.L.M. are partially supported by grants from the National Institutes of Health (1R01HG005084-01A1, 1R01HG005853-01) n Table 6 Enriched GO terms of hits from the ade2-min3 SGA analysis of the yeast nonessential and essential strain sets and the ade2-h7.5 SGA analysis of the yeast nonessential strain set

Figure 3
Specific mismatch repair components stabilize minisatellites in stationary phase. Blebbing quantification for ade2-h7.5 strains bearing a deletion of a nonessential mismatch repair gene was performed as described in Figure 2 and in Materials and Methods.