A genomewide screen in Saccharomyces cerevisiae for genes that suppress the accumulation of mutations

Abstract

A genomewide screen of a collection of 4,847 yeast gene deletion mutants was carried out to identify the genes required for suppressing mutations in the CAN1 forward-mutation assay. The primary screens and subsequent analysis allowed (i) identification of 18 known mutator mutants, providing a solid means for checking the efficiency of the screen, and (ii) identification of a number of genes not known previously to be involved in suppressing mutations. Among the previously uncharacterized mutation-suppressing genes were six genes of unknown function including four (CSM2, SHU2, SHU1, and YLR376c) encoding proteins that interact with each other and promote resistance to killing by methyl methanesulfonate, one gene (EGL1) previously identified as suppressing Ty1 mobility and recombination between repeated sequences, and one gene (YLR154c) that was not associated with any known processes. In addition, five genes (TSA1, SOD1, LYS7, SKN7, and YAP1) implicated in the oxidative-stress responses were found to play a significant role in mutation suppression. Furthermore, TSA1, which encodes thioredoxin peroxidase, was found to strongly suppress gross chromosomal rearrangements. These results provide a global view of the nonessential genes involved in preventing mutagenesis. Study of such genes should provide useful clues in identification of human genes potentially involved in cancer predisposition and in understanding their mechanisms of action.

Maintaining the stability of the genome is critical to cell survival and normal cell growth. Inherited or acquired deficiencies in genome maintenance systems contribute significantly to the onset of cancer as evidenced by the observation that a number of the DNA-repair and checkpoint genes are mutated in cancer susceptibility syndromes and sporadic cancers (1–4). This raises the possibility that other genetic defects causing genome instability and mutator phenotypes could contribute to carcinogenesis. The yeast Saccharomyces cerevisiae provides an ideal system for the analysis of mutator phenotypes. Such mutator phenotypes are expected to result from defects in genes with products that act to maintain genome stability. Knowledge obtained from S. cerevisiae about specific DNA-sequence changes, the rates at which they arise, and the influence of different genes and alleles on these changes has yielded insights into the processes that ensure genomic stability. One example in this regard concerns the relationship between DNA mismatch repair (MMR) defects and microsatellite sequence instability (5–7). Mutations in different components of the MMR system can differentially affect the repair of replication errors, and individual MMR-defective mutants display distinct mutation spectra including increased accumulation of frameshift mutations in simple repeat sequences (5–7). In fact, the link between microsatellite sequence instability and MMR defects first demonstrated in Escherichia coli and yeast provided a critical clue in the identification of MMR defects in the cancer-predisposition syndrome hereditary nonpolyposis colorectal carcinoma (8). More recently, S. cerevisiae-based genetic systems have been developed for evaluating genes that function in suppressing gross chromosomal rearrangements (GCRs) (4, 9–12). Studies using these systems have identified a number of genes and pathways that function to suppress genome rearrangements and have established parallels between these pathways and the genes that are mutated in the human chromosomal instability syndromes (4).

Increased mutation rates and modified mutational spectra induced by defects in many S. cerevisiae genes have been studied extensively. However, additional mutation-suppressing genes are still being identified. The Saccharomyces Genome Deletion Project created a set of isogenic mutant strains in which each individual nonessential gene has been deleted (13). This mutant collection has facilitated genomewide studies to identify genes required for resistance to various cellular insults (14–18). Here we report the systematic analysis of the complete set of gene deletion mutants to identify genes required for preventing spontaneous mutations in the CAN1 gene, providing a global view of these nonessential genes in maintaining genome stability as measured by using one type of assay. We found (i) most of the previously known mutator mutants and (ii) a number of genes not reported previously to be involved in suppressing mutations. Among these genes were five genes implicated in the oxidative-stress response that also play a significant role in mutation suppression, providing genetic evidence that the oxidative-stress response functions to prevent genome instability.

Materials and Methods

Yeast Strains and Media. Yeast strains were grown in standard media including yeast extract/peptone/dextrose (YPD) medium or synthetic complete medium (SC) lacking the appropriate amino acid. Canavanine-resistant mutants (Can^r) were selected on SC arginine-dropout plates containing 60 mg/liter canavanine. Hom⁺ revertants were selected for on SC threonine-dropout plates, and Lys⁺ revertants were selected for on SC lysine-dropout plates. Canavanine- and 5-fluoroorotic acid (5FOA)-resistant mutants (Can^r-5FOA^r) were selected on SC arginine- and uracil-dropout plates containing 60 mg/liter canavanine and 1 g/liter 5FOA. G418-resistant colonies were selected for on YPD plates containing 200 mg/liter geneticin.

A collection of 4,847 strains constructed in the BY4741 background representing essentially all haploid MATa nonessential yeast deletion strains were obtained from EUROSCARF (Frankfurt, Germany; see also http://www-sequence.stanford.edu/group/yeast_deletion_project). The deletion strains were transferred to YPD plates by using a multiprong replica-plating device and then screened as described below. For construction of individual mutant strains in the RDKY3615 (MATa, ura3-52, leu2Δ1, trp1Δ63, his3Δ200, lys2ΔBgl, hom3-10, ade2Δ1, ade8, hxt13::URA3) background (10), which is highly related to BY4741, genomic DNA from a BY4741 mutant was used as a template for PCR amplification. The resulting DNA fragment, which contains the kanMX4-selectable marker flanked by the upstream and downstream sequence of each ORF of interest, was transformed into RDKY3615, and proper disruptants were verified by PCR.

Fluctuation Analysis. The rate of accumulation of mutations in cell populations was determined by fluctuation analysis by using the method of the median (19) as described (7, 20). For the fluctuation test of Can^r mutations, Hom⁺ revertants, and Lys⁺ revertants, seven to nine independent cultures were analyzed in each experiment and each fluctuation test was repeated independently at least two times. For determining Can^r-5FOA^r mutation rates, at least 20 independent cultures were analyzed.

Mutation and Breakpoint Spectra. Can^r mutation spectra were determined by PCR amplification of the CAN1 gene from independent Can^r isolates followed by DNA sequencing (7). Sequences of the primers used are available on request. Mutations then were detected by using polyphred and consed software. Rearrangement breakpoints from independent Can^r-5FOA^r mutants were mapped and sequenced as reported (10, 11). The sequences of the breakpoint junctions then were analyzed by performing blast searches against the Saccharomyces Genome Database to determine the location of the breakpoint and the identity of the rearranged chromosomes.

Mutator Mutant Screen. Patches of the wild-type strain BY4741 and each individual deletion strain were made on YPD plates and incubated at 30°C for 2 days. These master plates then were replica-plated to canavanine plates that were incubated for 3 days to detect the presence of Can^r mutants. Each patch was scored as follows: 0 for patches with 0–10 Can^r colonies, similar to the wild-type strain; + for patches with 10–20 colonies; ++ for patches with 21–30 colonies; +++ for patches with 31–40 colonies; and ++++ for patches with >40 colonies or confluent. All mutants with a score of ++ or greater were retested by streaking them to single colonies on YPD plates and then retesting two single colonies from each strain by using the above-described patch-test method. Mutants with a score of ++ or greater then were retested by performing a fluctuation analysis with three to five cultures and visually comparing the plates with those obtained with the wild-type strain, and then analyzed in detail by fluctuation analysis to determine the Can^r mutation rate. Finally, for all mutator mutants identified in which a previously uncharacterized mutationsuppressing gene was implicated, the gene was redisrupted in RDKY3615, and the resulting mutants were analyzed extensively by fluctuation analysis. These previously uncharacterized genes were also analyzed in the opposite mating-type mutant strain, BY4742 (MATα), constructed by the Saccharomyces Genome Deletion Project.

Results

Qualitative Screen. We performed a genomewide screen to identify genes that suppress the accumulation of mutations that inactivate the CAN1 gene; such mutations confer resistance to canavanine. In control experiments the scoring of several known mutants was rad1Δ (+), rad18Δ (++), rad5Δ (+++), rad52Δ (++++), and msh2Δ (++++), which is in accord with the previously reported effect of these mutations on spontaneous Can^r mutation rates (20). In the first round of the screen, which was performed blind with regard to the identity of the deleted ORFs, 335 strains were scored as ++ or greater, and the majority of the other strains were scored as 0. Retesting these 335 strains yielded 96 (including can1Δ) that scored ++ or greater. When the 95 strains (not including can1Δ) were retested by performing a preliminary fluctuation analysis, 34 mutants were identified that yielded an increased number of Can^r colonies compared with the wild-type strain.

Quantitative Determination of Can^r Mutation Rates and Identification of Known and Unknown Mutation-Suppressing Genes. When extensive fluctuation analysis of the 34 candidate mutator mutants was performed, 33 mutants had a significantly increased Can^r mutation rate (Table 1). Eight of the 33 mutations deleted wild-type genes of unknown function. Among the other 25 known mutants for which the relevant wild-type genes have been assigned a function, 18 have previously been identified mutation-suppressing genes. Fifteen of these genes (RAD27, PMS1, MSH2, MLH1, RAD52, RAD54, RAD57, RAD51, MRE11, RAD50, MMS2, MSH6, RAD5, OGG1, and RAD18), when mutated, have been shown to increase Can^r rate (7, 10, 20–24). Three of these (SOD1, PIF1, and UNG1) had not been studied previously by using the Can^r assay but are known to cause mutator phenotypes when disrupted, as revealed by other assays (25–27). The remaining seven known genes (RAD55, TSA1, HIS5, ADR1, LYS7, SKN7, and YAP1), together with eight genes without an assigned function (YMR166c, SHU1, CSM2, YLR376c, SHU2, ELG1, YLR154c, and YDL162c) are identified here as candidate mutation-suppressing genes (Table 1). However, RAD55 was not considered as a novel mutation-suppressing gene and not studied further, because it encodes a protein that is known to function in recombination and repair as a complex with Rad57 (28), a protein encoded by a known mutation-suppressing gene (10).

Table 1. Rank order of the 33 BY4741 Can^r mutator strains.

Gene deleted	Canr rate* (× 10-7)	Mutator phenotype	Wild-type function
Wild type	3.4
RAD27	167.1	CAN1†	DNA replication and repair
PMS1	79.9	CAN1†	MMR
MSH2	64.9	CAN1†	MMR
MLH1	53.0	CAN1†	MMR
RAD52	46.4	CAN1†	Recombinational repair
YMR166c	45.1	Unknown	Unknown function
RAD54	37.9	CAN1†	Recombinational repair
RAD57	37.2	CAN1†	Recombinational repair
RAD55	35.8	Unknown	Recombinational repair
TSA1	35.5	Unknown	Thioredoxin peroxidase
RAD51	34.2	CAN1†	Recombinational repair
MRE11‡	33.3	CAN1†	Recombinational repair, checkpoints
RAD50‡	33.0	CAN1†	Recombinational repair, checkpoints
HIS5	31.1	Unknown	Histidinol-phosphate aminotransferase
MMS2	28.5	CAN1†	Ubiquitin-conjugating enzyme
MSH6	27.7	CAN1†	MMR
ADR1	24.5	Unknown	Transcriptional regulator
SHU1	21.7	Unknown	Unknown function
RAD5	21.5	CAN1†	DNA helicase, postreplication repair
CSM2	21.0	Unknown	Unknown function
YLR376c	20.4	Unknown	Unknown function
SOD1	20.4	trp1-289	Cu/Zn superoxide dismutase
OGG1	17.6	CAN1†	8-Oxoguanine DNA glycosylase
RAD18	17.5	CAN1†	Postreplication repair
SHU2	17.4	Unknown	Unknown function
LYS7	17.0	Unknown	Lysine synthesis, Cu chaperone for Sod1
ELG1	13.6	Unknown	Unknown function
PIF1	12.0	GCR	DNA helicase
SKN7	10.9	unknown	Oxidative stress response
YAP1	9.8	Unknown	Oxidative stress response
YLR154c	8.9	Unknown	Unknown function
YDL162c	8.4	Unknown	Unknown function
UNG1	7.9	SUP4-o	Uracil DNA glycosylase

^*Mutation rates shown are the average of two independent experiments, each with nine cultures.

^†Mutator phenotypes were analyzed previously by the CAN1 forward-mutation assay as well as at least one other assay including reversion assays, mutation of a plasmid-borne SUP4-o locus, microsatellite sequence instability assays, or GCR assays.

^‡The xrs2Δ mutant was not identified as a mutator in the original patch tests, presumably due to its slow growth. Subsequent fluctuation tests showed this mutant had an increased Can^r mutation rate of 23.6 × 10^-7.

Confirmation of the Mutator Phenotypes. To confirm that the Can^r mutator phenotypes of the 14 newly identified candidate mutator mutants (not including rad55Δ) were the result of deletion of the given ORFs, their counterparts in the BY4742 strain were analyzed for mutator phenotypes. All the mutations except his5Δ and adr1Δ caused a similar mutator phenotype in both the BY4741 and BY4742 strain backgrounds, whereas the his5Δ and adr1Δ mutations did not cause an increased mutation rate in the BY4742 strain background. This raised the possibility that the increased Can^r rate of his5Δ and adr1Δ BY4741 mutants was not related to deletion of the indicated gene.

We redisrupted the 14 genes as well as SOD1 in the RDKY3615 background (Table 2). SOD1 was selected for further study because it had not been studied in detail previously (25). The RDKY3615 strain, in addition to the CAN1 locus, contains the hom3-10 and lys2Δ-Bgl alleles and also allows detection of GCRs. The hom3-10 assay measures the reversion of a +1 insertion in the HOM3 gene (7). The lys2Δ-Bgl assay measures the reversion of a +4 insertion in the LYS2 gene (7). The GCR assay measures chromosomal rearrangement due to loss of the region containing CAN1 and URA3 (10, 11, 26, 29). Fluctuation analysis showed that 13 of these mutants had increased Can^r rates compared with the wild-type RDKY3615 (Table 2). The adr1Δ and his5Δ mutations did not cause an increased mutation rate in any of the CAN1, hom3-10, and lys2Δ-Bgl assays. Thus the Can^r mutator phenotype of the adr1Δ and his5Δ BY4741 mutants was not due to deletion of the indicated gene.

Table 2. Verification of mutator phenotypes in the RDKY 3615 background.

Gene deleted	Canr (× 10-7)*	hom3-10 (× 10-8)*	lys2Δ-Bgl (× 10-8)*	GCR (× 10-10)†
Wild type	5.5	1.3	2.1	9.5
TSA1	55.1	6.7	14.1	172.8
SOD1	32.0	5.5	No growth	8.8
LYS7	31.7	4.0	No growth	5.0
SKN7	15.1	2.2	3.4	29.1
YAP1	12.2	1.9	3.0	22.0
CSM2	32.8	2.8	6.7	8.5
SHU2	30.0	2.9	6.1	35.5
SHU1	22.8	2.5	6.0	29.5
YLR376c	22.0	2.8	6.3	5.0
YLR154c	17.1	3.6	5.9	29.6
ELG1	16.3	2.3	3.5	74.5
YMR166c‡	42.4	320.6	27.5	ND
YDL162c‡	7.6	1.6	2.7	45.9
ADR1‡	4.9	1.4	1.6	ND
HIS5‡	4.5	1.3	1.9	ND

ND, not determined.

^*Average of two to four independent experiments, each with seven to nine cultures.

^†Determined from the median of at least 20 cultures.

^‡Four mutants eliminated as false positives.

Of the 13 mutations that increased the Can^r mutation rates in RDKY3615, ymr166cΔ also caused a striking increase in Hom⁺ (247-fold) and Lys⁺ (13-fold) reversion rates (Table 2) similar to that of msh2Δ, mlh1Δ, or pms1Δ mutants (7, 21). YMR166c is located 413 bp from MLH1 (YMR167w) on the opposite strand, raising the possibility that deletion of YMR166c could affect the MLH1 promotor. Consistent with this, an ARS CEN MLH1 plasmid largely corrected the mutator phenotype of the ymr166cΔ mutant (data not shown), indicating that the mutator phenotype of ymr166cΔ can be attributed to the perturbation of the neighboring MLH1 gene. A similar situation is likely true for the ydl162cΔ mutant, which had a slightly increased Can^r mutation rate, no increased Hom⁺ and Lys⁺ reversion, and an increase in the GCR rate. YDL162c is located 105 bp upstream of the essential CDC9 gene, and deletion of YDL162c could affect the promotor region of the CDC9 gene. Consistent with this, cdc9 mutations cause a strikingly increased GCR rate (C. Putnam and R.D.K., unpublished results).

Of the 11 confirmed mutation-suppressing genes (including SOD1) (Table 2), TSA1, SOD1, LYS7, SKN7, and YAP1 are all involved in the oxidative-stress response (30–32). Deletion of TSA1, which encodes thioredoxin peroxidase, resulted in elevated mutation rates in all four mutation assays including the GCR assay, indicating the importance of this gene in preventing a broad spectrum of types of genomic instability. Deletion of SOD1 or LYS7, which encode a Cu/Zn superoxide dismutase and its Cu chaperone, respectively, increased both the Can^r mutation rate and the hom3-10 reversion rate but did not increase the GCR rate. As expected, neither mutant could grow on SC plates lacking lysine (32, 33). YAP1 and SKN7 encode transcription factors that control the oxidative-stress response. Deletion of these two genes causes much weaker but consistent increases in the Can^r mutation rate and the GCR rate. The remaining six genes including CSM2, SHU2, SHU1, YLR376c, YLR154c, and ELG1 are genes of unknown function. Their deletion resulted in elevated mutation rates in the Can^r (3- to 6-fold), Hom⁺ (2-fold), and Lys⁺ (2- to 3-fold) assays. The shu2Δ, shu1Δ, ylr154cΔ, and elg1Δ mutants also had a slightly increased GCR rate (3- to 8-fold; Table 2).

Oxidant Defense Systems Revisited. In addition to the five identified mutation-suppressing genes that encode components of the S. cerevisiae oxidant defense system, there are a number of additional factors that function in the oxidant defense system (30, 31, 34) that could have escaped detection by the genetic screen described here. On specific retesting these mutations only caused marginal or no increase in the Can^r mutation rate (Table 3). In addition, these mutations did not increase the GCR rate of a pif1 mutant, whereas a tsa1 mutation did increase the pif1 GCR phenotype (S. Smith and K. Myung, personal communication). These data confirm that Tsa1, Sod1, Lys7, Skn7, and Yap1 are critical members of the oxidant defense system that function as suppressors of mutagenesis.

Table 3. Can^r rate of strains mutated in oxidative stress-response genes.

Gene deleted	Rate (× 10-7)*	Wild-type function
Wild type	3.4
cTPx II	5.1	Cytoplasmic thiol peroxidase isoform II
cTPx III	5.6	Cytoplasmic thiol peroxidase isoform III
mTPx	5.3	Mitochondrial thiol peroxidase isoform
nTPx	6.4	Nuclear thiol peroxidase isoform
SOD2	5.8	Manganese superoxide dismutase
TRX1	4.5	Thioredoxin 1
TRX2	4.0	Thioredoxin 2
TRX3	3.3	Mitochondrial thioredoxin
TRR2	4.2	Mitochondrial thioredoxin reductase
GPX1	3.8	Glutathione peroxidase
GPX2	3.8	Glutathione peroxidase
HYR1	7.7	Glutathione peroxidase
GSH1†	0	γ-Glutamylcysteine synthetase
GSH2	4.6	Glutathione synthetase
CCP1	4.2	Cytochrome c peroxidase
CTT1	4.5	Catalase T
CTA1	3.7	Catalase A
ZWF1	6.4	Glucose-6-phosphate dehydrogenase
TKL1	3.4	Transketolase
RPE1	2.7	Ribulose-5-phosphate epimerase

^*Average of two independent experiments, each with nine cultures. Mutations were in the BY4741 background.

^†No colonies grew on SC arginine-dropout plates containing 60 mg/liter canavanine.

Analysis of the Spectrum of Mutations That Accumulate in tsa1Δ, sod1Δ, and lys7Δ Mutants. The sequence of 96 independent Can^r mutations arising in tsa1Δ, sod1Δ, and lys7Δ mutants was determined. In the wild-type strain, single-base substitutions (63.8%), double-base substitutions (1.1%), single-base frameshifts (18.1%), large deletions (10.6%), and complex events (7.4%) were detected (Table 4). This mutation spectrum is similar to that reported in other studies (20, 24), although a larger number of independent Can^r isolates was analyzed here. The data showed that the proportions of different types of mutations in tsa1Δ, sod1Δ, and lys7Δ mutants were globally similar to those of the wild-type strain, whereas the rate of each type of mutation was higher in the mutants than the wild-type (Table 4). In the wild-type spontaneous mutational spectra, the most commonly observed base substitutions were GC-to-TA transversions (20.2%) followed by GC-to-AT transitions (16.0%) and GC-to-CG transversions (16.0%). All three mutants had an altered mutation spectrum reflected in part by an increase in the proportion of GC-to-AT transitions and a decrease in the proportion of GC-to-TA transversions. For example, in the tsa1Δ mutant, the most prominent base substitutions were GC-to-AT transitions (26.9%) followed by GC-to-CG transversions (12.9%) and GC-to-TA transversions (11.8%). The rate of GC-to-AT transitions was increased by ≈17-fold compared with wild type. Overall, the base substitutions observed in the tsa1Δ, sod1Δ, and lys7Δ mutants are comparable to the previously reported oxidant-induced mutation spectra (35).

Table 4. Spectrum of Can^r mutations.

Genotype	Mutation	Frequency (%)	Rate*
Wild type	Single BS	60/94 (63.8)	3.5 × 10-7
	GC to TA	19/94 (20.2)	1.1 × 10-7
	GC to AT	15/94 (16.0)	8.8 × 10-8
	GC to CG	15/94 (16.0)	8.8 × 10-8
	AT to TA	7/94 (7.4)	4.1 × 10-8
	AT to CG	2/94 (2.1)	1.2 × 10-8
	AT to GC	2/94 (2.1)	1.2 × 10-8
	Double BS	1/94 (1.1)	6.1 × 10-9
	-1 Frameshift	15/94 (16.0)	8.8 × 10-8
	+1 Frameshift	2/94 (2.1)	1.2 × 10-8
	Large deletion	10/94 (10.6)	5.8 × 10-8
	Complex	6/94 (6.4)	3.5 × 10-8
tsa1 Δ	Single BS	63/93 (67.7)	37.3 × 10-7
	GC to TA	11/93 (11.8)	6.5 × 10-7
	GC to AT	25/93 (26.9)	14.8 × 10-7
	GC to CG	12/93 (12.9)	7.1 × 10-7
	AT to TA	8/93 (8.6)	4.7 × 10-7
	AT to CG	6/93 (6.5)	3.6 × 10-7
	AT to GC	1/93 (1.1)	5.9 × 10-8
	Double BS	2/93 (2.2)	1.2 × 10-7
	-1 Frameshift	18/93 (19.4)	10.7 × 10-7
	+1 Frameshift	1/93 (1.1)	5.9 × 10-8
	Large deletion	3/93 (3.2)	1.8 × 10-7
	Complex	6/93 (6.5)	3.6 × 10-7
sod1 Δ	Single BS	65/96 (67.7)	21.7 × 10-7
	GC to TA	14/96 (14.6)	4.7 × 10-7
	GC to AT	18/96 (18.8)	6.0 × 10-7
	GC to CG	17/96 (17.7)	5.7 × 10-7
	AT to TA	6/96 (6.3)	2.0 × 10-7
	AT to CG	6/96 (6.3)	2.0 × 10-7
	AT to GC	4/96 (4.2)	1.3 × 10-7
	Double BS	1/96 (1.0)	3.3 × 10-8
	-1 Frameshift	16/96 (16.7)	5.3 × 10-7
	+1 Frameshift	1/96 (1.0)	3.3 × 10-8
	Large deletion	7/96 (7.3)	2.3 × 10-7
	Duplication	1/96 (1.0)	3.3 × 10-8
	Complex	5/96 (5.2)	1.7 × 10-7
lys7 Δ	Single BS	60/96 (62.5)	19.8 × 10-7
	GC to TA	17/96 (17.7)	5.6 × 10-7
	GC to AT	17/96 (17.7)	5.6 × 10-7
	GC to CG	12/96 (12.5)	4.0 × 10-7
	AT to TA	4/96 (4.2)	1.3 × 10-7
	AT to CG	6/96 (6.3)	2.0 × 10-7
	AT to GC	4/96 (4.2)	1.3 × 10-7
	Double BS	2/96 (2.1)	6.6 × 10-8
	-1 Frameshift	15/96 (15.6)	5.0 × 10-7
	Large deletion	10/96 (10.4)	3.3 × 10-7
	Duplication	2/96 (2.1)	6.6 × 10-8
	Complex	7/96 (7.3)	2.3 × 10-7

BS, base substitution.

^*The rate of each type of mutation was obtained by multiplying the total Can^r mutation rate by the proportion of each mutation.

To study the tsa1Δ mutant further, the breakpoint sequences of 10 independent GCRs were determined and classified (Table 5). The majority of the GCR events (8/10) were deletions of an arm of chromosome V combined with addition of a new telomere (referred to as telomere additions). One translocation and one large deletion were also observed, and in both cases the rearrangement breakpoint was at a region of microhomology.

Table 5. The structure of rearrangement breakpoints isolated in a tsa1 Δ mutant.

Breakpoint type/breakpoint sequences
Telomere addition
GTGTTACTACTAGGATTTGGCGTGG: gtgtgggtgtgggtgtg (34867)
TGTGTATGGGCACAAACCCTTG: ggtgtgggtgtggtgtgggtgtgggtg (34422)
ACTACTAGGATTTGGCGTGG: gtgtggtgtgtgggtgtgggtgtgg (34867)
ACTAGGATTTGGCGTGGATG: ggtgtgggtgtggtgtgggtgtggtgtgtg (34864)
TTTTTGTATGGTTTGTGGTGCTGGG: tgtggtgtgggtg (32708)
AGGGTTTCTGTGTGGTTTCCGGGTG: tggtgtggtg (33820)
ATGGCTATTAAATATCACTG: ggtgtgggtgtgggtgtggtgtggtgtggt (32098)
CGGTATGATGTAGTTTTTAAATGTG: ggtgtgggtgtggtgtgg (34538)
Large deletion
V (34241)	(34184)
ATACGATTACTCCAGTTCCTCTTACAAGAAAT: GCATAAAAATAGTTACAATTAATTAG
ACTAAGGTTTAGTTTCGACTCTTACAAGAAAT:AAACGTTAAAAGTGGGAGACTGAGTA
V (18497)	(18440)
Translocation
V (41740)	(41685)
GGCCGCAAGGGCCAAGACAAG GAGTCTCCG: GAATTCAACGGTAAACGTGCAAGTG G
GATTTGTGTGGACTTCCTTAGAAGTAACCG:AAGCACAGGCGCTACCATGAG AAATG
XVI (856165)	(856220)

For telomere addition, the number indicates the Saccharomyces Genome Database nucleotide coordinate of the last recognizable nucleotide of chromosome V before the added telomere sequences. For deletions and translocations, the numbers given above and below the sequences are the standard Saccharomyces Genome Database nucleotide coordinates for the first and last nucleotide listed with the Roman numeral indicating the chromosome number. The underlined nucleotides indicate those present in the translocation chromosome. The nucleotides in bold indicate identities.

Discussion

We have taken a genomewide approach to the analysis of mutationsuppressing genes by screening the complete set of yeast deletion strains for mutants with a Can^r mutator phenotype. This functional genomic screen identified 33 genes potentially associated with the suppression of the accumulation of mutations (Table 1). Most of the known mutation-suppressing genes are involved in a variety of aspects of DNA metabolism. Fourteen potential mutator mutants that had not been identified previously and sod1Δ, which had only been shown to cause weak effects in a reversion assay (25), were subjected to further analysis. Four of these mutants were eliminated as false positives. Of the confirmed genes, five encode components of the oxidative-stress response, and six are genes of unknown function. We believe the data presented here define a nearly complete collection of nonessential genes involved in suppression of mutations in the CAN1 forward-mutation assay and define several previously unappreciated mutation-suppression pathways.

The screen described here was performed blind with regard to the identity of the deleted ORFs. The efficiency of this screen was evaluated by considering the known mutation-suppressing genes identified. Apparently, all the well known strong and moderate mutation-suppressing genes that function under normal growth conditions were identified (Table 1). Some known weak mutator mutants were retained in the first round of screening but eliminated in subsequent analysis due to their failure to match our selection criteria, which were designed to identify strong and moderate mutator mutations. To the best of our knowledge, only the exo1Δ mutant reported in the literature to have an increased Can^r rate was not obtained (36, 37). However, we determined the Can^r rate of the exo1Δ mutant present in the deletion-strain collection and found no difference in rate compared with that of the isogenic wild-type. The reason could be the strain background or the possible presence of a suppressor. Finally, we note that some mutants were reported to display modest mutator phenotypes in mutation assays other than the Can^r assay, but a comparison is difficult because of the difference in assay systems. This analysis suggests that the screen is close to saturated for deletion mutations in nonessential genes that cause increased mutation rates in the Can^r assay.

Our primary screen identified 14 genes that had not been implicated previously in suppressing the accumulation of mutations. However, after subsequent analysis four of these genes proved to be false positives. Two of the mutations caused a mutator phenotype in the BY4741 background but not the BY4742 background and the redisruption in a different MATa strain did not cause a mutator phenotype. Thus, in these two cases the mutator phenotype was probably due to an unlinked mutation. Previous studies have observed such a difference between mutations in the BY4741 and BY4742 backgrounds and have attributed this to differences in the affect of mating type on phenotype (16), whereas our results suggest that such affects are an indication that the phenotype is due to an unlinked mutation. In two other cases, the phenotype caused by the mutation was most likely due to an effect of the mutation on the expression of a neighboring gene. Such effects have generally not been considered in global analyses using the deletion-mutant collection. The false-positive rate observed was 12% when all the mutator mutants were considered. However, if only mutations in previously unidentified genes (14 candidate genes) are considered, the false-positive rate was 29%. This high false-positive rate has important implications for the results of other global analyses using the yeast-deletion collection where the goal is to identify previously uncharacterized genes and interactions, because these analyses have generally not excluded false positives.

The functional genomic screen for Can^r mutator phenotypes reported here has provided a more global view of the overall genetic control of spontaneous mutagenesis. This analysis has identified six groups of genes or pathways that have been implicated previously in suppression of mutations and in some cases has provided insights into these genes/pathways. Among the known genes and pathways identified were (i) RAD27, which encodes a replication factor that functions to suppress a broad spectrum of mutations and genome rearrangements (10, 24), (ii) the MMR genes, which function to suppress the consequences of misincorporation errors during DNA replication (5–7, 21), (iii) members of the RAD52 epistasis group of recombination genes, which suppress both Rev3- and Pol32-dependent mutations as well as a variety of genome rearrangements (10, 20, 23, 38), and (iv) members of the RAD6 postreplicative repair pathway, which suppress mutations that are Pol32-dependent and largely Rev3-dependent (20, 23, 38). The screen identified OGG1, encoding 8-oxoguanine DNA glycosylase, and UNG1, encoding uracil DNA glycosylase, two of the seven known DNA glycosylases that function in base excision repair (39). These results suggest that the major contributions to spontaneous mutations that are suppressed by base excision repair are oxidation of guanine and the presence of uracil in DNA either due to misincorporation of UTP or deamination of cytosine. Finally, PIF1, a gene that was implicated previously in suppressing genome rearrangements (26, 40), was identified, although we did not determine whether pif1 mutations induce other types of mutations besides genome rearrangements. Among these groups of genes, members of the MMR gene family and the recombinational repair group have been shown to have human homologues that are human disease genes (1–4).

One of the results of the present study is the demonstration that a number of genes involved in the oxidative-stress response have a significant effect on suppressing mutagenesis. Reactive oxygen species such as H₂O₂, alkyl hydroperoxides, and superoxide anion induce many types of DNA damage including base and sugar modifications, single- and double-stranded DNA breaks, abasic sites, and DNA–protein crosslinks (41, 42). In S. cerevisiae, the base excision repair and Msh2-Msh6-dependent MMR pathways are believed to be the primary mechanisms against mispairs due to oxidative damage (22, 43). But the mutagenic consequences of a defective antioxidant system that prevents formation of DNA damage by inactivating reactive oxygen species have not been studied, with the exception that sod1Δ caused a 3-fold increase in the rate of reversion of the trp1-289 allele (25). Our results revealed Tsa1, which is a key enzyme for eliminating H₂O₂ and alkyl hydroperoxide, Sod1, the major enzyme involved in removing superoxide anions and its Cu chaperone Lys7, and Skn7 and Yap1, which control the oxidative-stress response, to be important in suppressing mutations. The tsa1Δ mutation caused the greatest increase in Can^r mutation rate, and the mutation spectrum in tsa1Δ, sod1Δ, and lys7Δ mutants reflected the types of mutations induced by oxidative damage to DNA. Tsa1 also suppresses the accumulation of GCRs, whereas Skn7 and Yap1 seem to have weaker effects (probably through regulation of Tsa1), and Sod1 or Lys7 have no effect. This is a demonstration of a role for oxidative-stress response-related genes in suppressing large genomic rearrangements. Three types of tsa1Δ-induced GCRs were seen including translocations, deletions, and telomere addition. The mechanism of tsa1Δ-induced GCRs is not clear. A notable feature of the spectrum for tsa1Δ-induced GCRs is that they are similar to those that occur spontaneously in wild type and various mutants that have high GCR rates (10, 11, 29) and those induced by methyl methanesulfonate (44). It is possible that DNA modifications from endogenous or environmental sources may cause similar mutagenic damage to DNA that, in the absence of effective GCR-suppressing mechanisms, leads to GCR formation. Our data suggest that Tsa1 is a key antioxidant for preventing DNA damage that can lead to mutations. Although the induction of Tsa1 is Yap1- and Skn7-dependent (45, 46), the fact that the tsa1Δ mutator phenotype was much stronger than that of yap1Δ and skn7Δ suggests that the basal level of Tsa1 plays a major role in the detoxification of reactive oxygen species. Inactivation of other components of the oxidative-stress response system had no significant effect on suppressing mutations, but we cannot rule out the possibility of functional redundancy between the members of the oxidative-stress pathway tested.

Six genes (CSM2, SHU2, SHU1, YLR376c, YLR154c, and ELG1) identified and confirmed to be involved in mutation suppression in our study are genes of unknown function. Interestingly, four of these gene products, Csm2, Shu2, Shu1, and Ylr376c, are interacting proteins, and the data presented here indicate that mutations in the genes encoding these proteins cause the same type and magnitude of mutator phenotype by using Can^r, Hom⁺, and Lys⁺ assays. Shu2 and Shu1 seem to have a weak effect in suppressing GCRs. Shu2, Shu1, and Ylr37c associate together, whereas Csm2 only interacts with Ylr376c (47, 48). The csm2Δ mutant exhibits meiotic chromosome missegregation with reduced spore viability (49). YLR376c is a member of the RAD52 epistasis group (18). All four null mutants are methyl methanesulfonate-sensitive (17, 18, 50), and three of them (csm2Δ, shu1Δ, and ylr376cΔ) are 4-nitroquinoline 1-oxide-sensitive (50). Taken together, these observations suggest that Shu1, Shu2, and Ylr376c, together with Csm2, may be involved in recombination-repair processes that prevent mutations, although much remains to be learned about their function. Mutation of ELG1 causes sensitivity to both methyl methanesulfonate and 4-nitroquinoline 1-oxide (50). ELG1 mutations result in increased mobility of a chromosomal Ty1 element, a phenotype similar to the deletion of several genes with characterized roles in DNA metabolism, including EST2, TEL1, MRE11, RAD50, SGS1, RAD57, and SAE2 (51). Furthermore, Elg1 has been reported to suppress DNA recombination between repeated sequences, similar to Sgs1 and Rrm3 (51). Together with these observations, our finding that Elg1 suppresses mutations and genome rearrangements establishes a role for this protein in maintaining genomic stability, possibly by acting in some type of recombination-repair pathway. Finally, ylr154cΔ showed an increased Can^r rate and hom3-10 and lys2Δ-Bgl reversion rates and a slight increase in GCR rate. The YLR154c gene product physically associates with Top2 and Kss1 (52), which are an essential DNA topoisomerase (53) and mitogen-activated protein kinase (54), respectively. Based on what is known about YLR154c, it is difficult to speculate on how it might function to prevent mutations.