Abstract

Maintaining genome stability requires that recombination between repetitive sequences be avoided. Because short, repetitive sequences are the most abundant, recombination between sequences that are below a certain length are selectively restricted. Novel alleles of the RAD3 and SSL1 genes, which code for components of a basal transcription and UV-damage-repair complex in Saccharomyces cerevisiae, have been found to stimulate recombination between short, repeated sequences. In double mutants, these effects are suppressed, indicating that the RAD3 and SSL1 gene products work together in influencing genome stability. Genetic analysis indicates that this function is independent of UV-damage repair and mutation avoidance, supporting the notion that RAD3 and SSL1 together play a novel role in the maintenance of genome integrity.

IN the yeast Saccharomyces cerevisiae, homologous recombination is an important mechanism for repairing DNA damage (Friedberget al. 1991). However, recombination between dispersed, repetitive sequences can result in deleterious genome rearrangements, such as deletions, insertions, inversions, and translocations (Petes and Hill 1988). One way to control these rearrangements is to restrict recombination to only the most similar sequences, thus blocking recombination between dispersed, duplicate sequences that have diverged over time (Shen and Huang 1986; Waldman and Liskay 1988; Bailis and Rothstein 1990; Bailiset al. 1992). This mechanism is under the control of certain mismatch repair genes in yeast (Selvaet al. 1995; Dattaet al. 1996; Negrittoet al. 1997), E. coli (Feinstein and Low 1986; Radman 1988; Rayssiguieret al. 1989; Petitet al. 1991), and mammals (deWindet al. 1995).

Another way to reduce the frequency of genome rearrangement is to selectively restrict recombination between short repeats, because they are the most abundant (Britten and Kohne 1968). In yeast and mammalian cells, repeats below 250-300 bp recombine less well per unit length than longer sequences (Jinks-Robertsonet al. 1993; Rubnitz and Subramani 1984). In yeast, sequences that share less than 30 bp are unable to interact by homologous recombination (Manivaskamet al. 1995). Rothstein and colleagues reported that in S. cerevisiae, deletions by recombination between ∂ elements, a class of short repeats, is selectively stimulated by mutations in the TOP3 gene (Walliset al. 1989). Recently, we showed that a novel mutation in the RAD3 gene also disrupts the control of short-repeat recombination (SRR) in S. cerevisiae (Bailiset al. 1995). Together, these observations suggest that there is a genetic basis for the selective discrimination against recombination between short repeats in yeast. Recent work suggests that a barrier against SRR could be important in humans, as homologous recombination between short repeats may give rise to tumor suppressor gene mutations in several cancers (Guet al. 1994; Schichmanet al. 1994; Soet al. 1997).

The S. cerevisiae RAD3 gene codes for a DNA-DNA and DNA-RNA helicase (Sunget al. 1987) that is a component of a heteropentameric complex at the core of both the transcription initiation factor TFIIH and the nucleotide excision repair (NER) complex (Feaveret al. 1993; Wanget al. 1994). The components of the heteropentamer are structurally and functionally conserved from yeast to humans (Gerardet al. 1991; Fischeret al. 1992; Hoejimakers 1993; Humbertet al. 1994; Schaefferet al. 1994). Many rad3 mutations that confer a broad array of phenotypes have been identified: temperature-sensitive growth (Naumovski and Friedberg 1986; Guzderet al. 1994; Bailiset al. 1995); defective transcription (Guzderet al. 1994); UV sensitivity (Reynolds and Friedberg 1981; Wilcox and Prakash 1981; Songet al. 1990); elevated mutation rate (Monteloneet al. 1988; Songet al. 1990); and elevated frequencies of spontaneous recombination (Monteloneet al. 1988; Songet al. 1990). The existence of alleles that confer some, but not other, phenotypes indicates that some of these phenotypes are genetically distinct. For instance, the rad3-K48R mutation blocks helicase activity in vitro and confers sensitivity to UV light, but does not confer a growth defect (Sunget al. 1988). This separates the growth-control and UV-damage-repair functions of Rad3p and indicates that Rad3p helicase activity is not required for growth control. Further, it suggests that Rad3p may perform more than one biochemical function.

We previously isolated the rad3-G595R mutant on the basis of its temperature-sensitive growth phenotype and elevated levels of SRR (Bailiset al. 1995). Interestingly, the rad3-G595R mutation increased the frequency of recombination between short sequences but not long sequences. This suggests that the effect of rad3-G595R on SRR is distinct from the previously identified rem phenotype where spontaneous recombination between sequences in excess of 300 bp is affected (Monteloneet al. 1988; Songet al. 1990).

Physical studies of recombination in the rad3-G595R mutant suggested a link between the SRR phenotype and defective processing of the ends of broken DNA molecules (Bailiset al. 1995). We found that degradation of both the 5′ and 3′ strands is slower in this mutant. This led us to propose that reduced exonucleolytic processing allows sequences adjacent to a break to persist, increasing the likelihood of recombination with similar sequences in the genome.

We demonstrated the link between the SRR- and DNA-processing phenotypes in a subsequent study (Bailis and Maines 1996). We found that adding heterologous DNA onto the ends of a DNA fragment with little terminal homology to the HIS3 locus blocked insertion into the HIS3 locus in rad3-G595R mutant cells, but not wild-type cells. In this case, blocking degradation of the ends of broken DNA inhibited the removal of heterologous sequences obscuring the homologous sequences on the fragment, preventing recombination. This is consistent with rad3-G595R affecting SRR by changing the processing of broken DNA molecules. Further, it suggests that the amount of degradation in wild-type cells is usually sufficient to degrade any short repeats that may flank a break. This could reduce the incidence of SRR by leaving unique sequence DNA at the break that can be healed by recombination with sequences on a sister chromatid or homolog without rearranging the genome. Because the length of the unique sequences separating dispersed, short repeats tends to exceed the repeat length (Rothstein 1979), breaks terminating in these regions should be repaired relatively efficiently.

In this report we present evidence that the role played by RAD3 in SRR control is distinct from its roles in UV resistance and mutation avoidance. We also discuss the isolation of an allele of the SSL1 gene, SSL1-T242I, that disrupts SRR control, but suppresses the defective transcription, growth, and SRR phenotypes conferred by rad3-G595R. The SSL1 gene product, a zinc finger protein (Yoonet al. 1993; Humbertet al. 1994) and component of the heteropentamer (Feaveret al. 1993), has been shown by others to interact with itself, Rad3p, and Tfb1p (Bardwellet al. 1994; Matsuiet al. 1995), a third subunit of the heteropentamer (Gileadiet al. 1992; Feaveret al. 1993). Interestingly, the results of two hybrid experiments suggest that the rad3-G595R and SSL1-T242I mutations may lead to changes in how subunits of the heteropentamer interact. We speculate that the growth, transcription, and SRR phenotypes of the rad3-G595R and SSL1-T242I mutants may all be due to changes within the heteropentamer.

MATERIALS AND METHODS

Strains: The yeast strains constructed for this study are isogenic, derived from W303-1A and W303-1B (Thomas and Rothstein 1989), and listed in Table 1. The two-hybrid strains CTY10-5d and Y187 were the gifts of Stan Fields and Steve Elledge, respectively. Standard methods were used for the growth, maintenance, and genetic manipulation of yeast (Shermanet al. 1986). Both the spheroplast (Hinnenet al. 1978) and alkali cation (Itoet al. 1983) methods of yeast transformation were employed.

Plasmids: The plasmids used in this study are listed in Table 2 and were built using standard molecular biological methods (Maniatiset al. 1989). All enzymes were obtained from New England Biolabs (Beverly, MA). All chemicals were obtained from Sigma (St. Louis, MO). pRS414 and 416 were kindly provided by Phil Hieter (Christiansenet al. 1991). pGHOT (Nickoloffet al. 1986) was generously provided by Jac Nickoloff and Fred Heffron. p1032 and 1033 were the kind gifts of Tom Donahue (Yoonet al. 1993). The two-hybrid plasmids pBTM116, pGAD-HB, pBTMRAD3, pGADSSL1, pMASSL1, pGADTFB1, and pMATFB1 were the generous gifts of Lee Bardwell (Bardwellet al. 1994).

Mutagenesis and selection for suppressors of the temperature-sensitive growth phenotype of a rad3-G595R mutant: A temperature-sensitive rad3-G595R mutant strain, ABX46-1C, was mutagenized to 20-30% viability with ethyl methanesulfonate. Approximately 200,000 survivors were plated onto YPD (2% dextrose, 2% bacto-peptone, 1% yeast extract) plates and incubated at the nonpermissive temperature of 37° for 5 days to select for high-temperature-resistant mutants.

RNA isolation and Northern blot hybridization: Wild-type and mutant cells were grown to midlog (1-5 × 107 cells/ml) in 10 ml of YPD liquid medium at 30°. The cultures were split and one half incubated at 30° and the other half at 37° for 1 hr. The cells from both cultures were harvested and washed, and total cellular RNA was prepared as described previously (Elion and Warner 1984). RNA (10 μg) was denatured and loaded into each lane of a 1.2% agarose, 3% formaldehyde, 20 mm N-morpholinopropanesulfonic acid (MOPS), and 1 mm EDTA gel and electrophoresed. Size-fractionated RNA was transferred to a nylon membrane and hybridized with both a 1.2-kb HindIII/HindIII URA3 sequence and a 590-bp SalI/SnaBI SAM1 fragment that had been 32P-labeled by random priming. After quantitation of the URA3 and SAM1 hybridization signals with a phosphorimager, the blots were stripped and rehybridized with a 32P-labeled 330-bp double-stranded DNA probe specific for 18S rRNA. These signals were used to normalize the URA3 and SAM1 levels because rRNA levels are unaffected by the rad3-G595R and SSL1-T242I mutations (T. Negritto and A. Bailis, unpublished data).

Growth-rate determination: YPD liquid (5 ml) was inoculated with a single yeast colony and grown to saturation at 30°. The saturated cultures were used to inoculate 25 ml of YPD liquid to a density of 5 × 106 to 1 × 107 cells/ml and were grown at 30°. Culture density was assessed at 30-min intervals by turbidimetry using a Klett-Summerson colorimeter fitted with a red filter.

View this table:
TABLE 1

S. cerevisiae strains

UV sensitivity assays: Five-milliliter YPD liquid cultures were started from single colonies and grown to saturation at 30°. Appropriate dilutions were plated onto YPD plates and exposed to varying doses of UV light in a UV cross-linker (Stratagene, La Jolla, CA). Irradiated cells were incubated in the dark at 30° for 3-5 days and the number of surviving colonies counted. The fraction of cells surviving treatment was determined by dividing the number of colonies formed after exposure by the number of colony-forming units in the original culture.

Mutation frequency determination: YPD medium (10 ml) was inoculated with single colonies of cycloheximide-sensitive yeast and grown to saturation at 30°. The cells were harvested, washed, and resuspended in water. Cells were plated onto YPD agar plates containing 10 μg/ml cycloheximide, a dose that selects primarily for mutations in the CYH2 ribosomal protein gene (Sikorski and Boeke 1991). The number of cycloheximide-resistant colonies was counted after incubation at 30° for 5-7 days. Viable counts were determined by counting the colonies that arose on YPD plates after incubation at 30° for 2-3 days. Mutation frequencies were determined by dividing the number of cycloheximide-resistant mutant colonies by the number of viable cells plated. Median mutation frequencies were determined from at least nine independent trials. Statistical significance was tested by determining the number of trials with each strain that were above and below the grouped median frequency and then performing contingency χ2 analysis and Yate’s correction for continuity (Cochran 1954).

Galactose-inducible transcription assays: A version of a standard β-galactosidase assay was used (Miller 1972). Cells freshly transformed with pLGSD5 (Table 2) were grown to midlog (∼2 × 107 cells/ml) at 30° in 20 ml of uracil-less, 3% glycerol, and 3% lactate growth medium. This medium selects for the presence of the plasmid and neither induces nor represses expression of the galactose-inducible GAL::cyc1::lacZ fusion gene on pLGSD5. The cultures were split and 1 ml of 20% (w/v) galactose was added to each culture. One culture was maintained at 30°, while the other was shifted to 37°. Both were incubated for 1 hr before harvest by centrifugation, resuspension with 1 ml of Z buffer (0.1 m tris HCl, pH 8.0, 20% (v/v) glycerol, 1 mm dithiothreitol), disruption with glass beads, and clarification by centrifugation at 4° in a microcentrifuge for 5 min. Another 0.5 ml of Z buffer was added to the clarified extract. Cell extract (100 ml) was added to 0.9 ml of Z buffer and warmed at 37° for 5 min before addition of 0.25 ml of 4 mg/ml o-nitrophenylgalactoside and incubation for 15 min to 1 hr at 37°. The reactions were stopped with the addition of 0.5 ml of 1 m sodium carbonate and the amount of product released was determined spectrophotometrically by the absorbance at 420 nm. Specific activities were determined by calculating the moles of nitrophenol released per minute per milligram of protein. Protein concentration was measured using the Bio-Rad protein assay (Hercules, CA).

Two-hybrid assays: Freshly transformed cells containing the two-hybrid constructs were grown to a density of 1-7 × 107 cells/ml at 30° in 5 ml of medium that selected for the presence of the plasmids and contained 2% glucose. Cell extracts were prepared and assayed as above.

Deletion assay: Integrating pLAY202, pLAY204, or pLAY214 into the HIS3 locus created 415-, 223-, and 103-bp duplications of HIS3 coding sequence flanking the 5-kb plasmid YIp5. The duplication strains were maintained on uracil-less medium, which selects against loss of the URA3 marker in YIp5. Uracilless medium was inoculated with a single colony and grown to saturation at 30°. Maintaining selection during growth reduces the incidence of jackpot events early in the growth of the culture, which skew the determination of recombination frequency. Cells were plated onto uracil-less agar and incubated at 30° for 5 days to determine the number of viable cells in the culture. Cells were also plated onto histidine-less medium and incubated at 30° for 5 days to select for recombinants that had generated an intact HIS3 gene. Loss of the plasmid in the His+ recombinants was confirmed by replica plating to uracil-less medium and Southern blot analysis (S. Maines and A. Bailis, unpublished results). Deletion frequencies are expressed as the number of His+, Ura- recombinants per viable cell plated. We tested for statistically significant differences between the median deletion frequencies using contingency χ 2 analysis and Yate’s correction for continuity (Cochran 1954) as described above for the mutation frequencies.

DNA fragment integration assay: The plasmid pLAY144, which contains the HIS3 gene on a 1.3-kb genomic fragment disrupted by the insertion of the URA3 gene on a 1.2-kb fragment, was digested with a variety of restriction endonucleases to yield fragments with different lengths of HIS3 sequence flanking the URA3 gene. Gel-purified DNA fragments were used to transform His+ Ura- yeast spheroplasts to uracil prototrophy. The number of uracil prototrophs was counted, and all were screened for the ability to grow without histidine to determine whether the DNA fragments had integrated into, and disrupted, the HIS3 locus (Ura+ His-) or gene-converted the ura3-1 marker at the URA3 locus (Ura+ His+). Southern blots of over 100 Ura+ recombinants showed that the DNA fragments either integrated at the HIS3 locus or gene-converted the ura3-1 marker at the URA3 locus (S. Maines and A. Bailis, unpublished results). Percentage insertion of the DNA fragment into the HIS3 locus, versus gene conversion at the URA3 locus, was determined by dividing the number of His- transformants by the total number of transformants (His+ and His-) and multiplying by 100. The efficiency of transformation with these fragments, normalized against the efficiency of transformation with an intact centromere plasmid, varied from 10- to 15-fold with changes in the length of HIS3 homology, but only 1.5- to 4-fold from strain to strain (S. Maines and A. Bailis, unpublished results).

Double-strand break (DSB) processing assay: Stationary cultures grown from single colonies of yeast transformed with pLAY97 (Negrittoet al. 1997) and pGHOT (Nickoloffet al. 1986) were used to inoculate 500 ml of medium that selected for both plasmids (without uracil or tryptophan) and neither induced nor repressed the galactose-inducible GAL::HO fusion gene on pGHOT (3% glycerol, 3% lactate). Cultures were grown to a density of 5 × 106 cells/ml at 30° before a 50-ml aliquot was removed, the cells were pelleted, and the pellet was frozen at -80°. Fifty milliliters of 20% galactose was then added to the culture to induce expression of the GAL::HO gene on pGHOT, which cut at the unique HO recognition site in pLAY97. Induction proceeded for 30 min at 30°, at which time hemacytometer counts were made, another aliquot was removed, and the cells were processed as above. The cells were then filtered free of this medium through sterile 0.4-μm nitrocellulose filters and resuspended in fresh, prewarmed, minimal medium lacking tryptophan and containing 2% glucose. Uracil was provided because pLAY97 was cleaved in more than 50% of the cells exposed to these conditions and less than 1% of the broken plasmids rejoin (Bailis and Maines 1996). Glucose was provided to repress GAL::HO gene expression. Cells were counted and aliquots removed and processed at regular intervals as described above.

Total cellular DNA was prepared from the frozen cell pellets by a standard protocol (Hoffman and Winston 1987), digested with ScaI restriction endonuclease, blotted to nylon, and probed with 32P-labeled pBluescript DNA, which is the backbone of pLAY97. The patterns of hybridization were visualized and quantitated with a Molecular Dynamics phosphorimager. DSB processing was quantitated by comparing the levels of the 2.8- and 3.4-kb ScaI and HO cut bands with those of the ScaI cut 6.2-kb band. Signal levels were not much influenced by the outgrowth of cells in which pLAY97 was not cut because all cells required at least 4 hr to double in density (Bailiset al. 1995; S. Maines and A. Bailis, unpublished results), when the broken plasmid signal was reduced at least eightfold (see results). Further, while the wild type and SSL1-T242I doubling times were very similar, the kinetics of DSB processing were very different (see results).

View this table:
TABLE 2

Plasmids

DNA from each time-point was transferred to a positively charged nylon membrane under either denaturing (0.4 N sodium hydroxide) or nondenaturing (3 m sodium chloride, 0.3 m sodium citrate, pH 7.0) conditions using a slot blot manifold (Bio-Rad, Hercules, CA) and fixed to the membrane using a UV cross-linker. The blots were hybridized with a 32P-labeled 413-base RNA species complementary to one side of the HO cut-site in pLAY97, obtained by in vitro transcription of pLAY159. Hybridization signals were quantitated with a phosphorimager. Levels of single-stranded DNA were determined by dividing the nondenatured DNA signals by the denatured DNA signals.

RESULTS

A mutation in SSL1 suppresses the rad3-G595R temperature-sensitive growth defect: Approximately 6 × 105 rad3-G595R mutant cells (ABX46-1C) were mutagenized to 30% survival with EMS, plated onto YPD agar, and incubated at 37° to select for mutants that suppress the temperature-sensitive growth defect. Three high-temperature-resistant clones were isolated and crossed to a wild-type strain (W961-5A). All three of the suppressors segregated as single genes, one linked to the RAD3 locus, and the other two unlinked to RAD3. One of the extragenic suppressors gave variable suppression of the rad3-G595R growth phenotype and was not studied further, while the other extragenic suppressor consistently suppressed the rad3-G595R growth defect. A cross between this strain (ABM43) and a rad3-G595R mutant (ABX81-9D) determined that the suppressor is dominant because the resulting diploid grew at 37°. The suppressor was discovered to confer a mild temperature-sensitive growth defect when it segregated away from the rad3-G595R mutation in crosses to wild type. The isolated suppressor was backcrossed five more times to wild type. All putative suppressor-containing segregants were checked for the ability to suppress the rad3-G595R Ts growth phenotype in crosses.

We reasoned that because the temperature-sensitive phenotype of the rad3-ts14 allele is linked to a transcription defect (Guzderet al. 1994), the same may be true for rad3-G595R. Further, because the RAD3 gene exerts its effect on transcription through TFIIH (Feaveret al. 1993), the suppressor mutation might be in a gene that codes for another subunit of TFIIH. To test this hypothesis we crossed the original suppressed rad3-G595R strain (ABM43) with a strain containing a URA3-marked SSL1 locus (ABT84). We looked for linkage to SSL1 because the Rad3p and Ssl1p proteins interact in vivo and in vitro (Bardwellet al. 1994). We found that the suppressor segregated in repulsion with the URA3 marker, indicating that the suppressor was very closely linked to the SSL1 locus. To further define the location of the suppressor mutation, the mutant SSL1 sequence from a suppressor strain (ABX96-8D) was recovered by gap repair (Rothstein 1991) and sequenced, revealing a single C to T transition at residue 725 that converts the threonine codon at position 242 to an isoleucine codon. Thereafter, the suppressor was referred to as SSL1-T242I because of the dominant nature of the suppressor phenotype.

The rad3-G595R and SSL1-T242I mutant cells display defective growth and transcription phenotypes: We were interested in quantitating the effects of the rad3-G595R mutation on growth and transcription. We determined that the rad3-G595R growth defect is recessive and is more severe at 30° than at 37° (Table 3). We also found that the rad3-G595R mutant has a recessive defect in expression from a galactose-inducible CYC1::lacZ fusion gene (2.5- to 3-fold reduced) and consistently lower steady-state levels of the URA3 and SAM1 mRNAs at both 30° and 37° (Table 4). These results are similar to those observed previously with the rad3-ts14 mutant mentioned above (Guzderet al. 1994), indicating that the rad3-G595R mutation may block transcription and inhibit growth in a similar way.

We also determined the effects of the SSL1-T242I allele on growth and transcription. We found that the mild temperature-sensitive growth phenotype conferred by SSL1-T242I is recessive, while the rad3-G595R slow growth suppressor phenotype is semi-dominant because two copies of the SSL1-T242I allele suppress better than one (Table 3). We also found that the SSL1-T242I mutant exhibits small, but significant, increases in CYC1::lacZ fusion gene expression, but no significant changes in the steady-state levels of either the SAM1 or URA3 mRNAs (Table 4). Interestingly, however, the rad3-G595R SSL1-T242I double mutant displays fusion gene expression and steady-state mRNA levels that are near wild type (Table 4), indicating that the SSL1-T242I mutation also suppresses the transcription defect conferred by rad3-G595R.

rad3-G595R and SSL1-T242I have minimal effects on mutation avoidance and UV resistance: Because other hyper-rec rad3 mutants display mutator phenotypes (Monteleone et al. 1988; Songet al. 1990), we investigated whether the rad3-G595R mutant also displays elevated levels of spontaneous mutation. We determined that the mutation frequency in rad3-G595R cells (4.2 × 10-9 mutations/viable ABX81-9D cell) is not statistically different (P = 0.35) from that of wild-type cells (2.7 × 10-9 mutations/viable W961-5A cell). This suggests that the recombination phenotype observed in rad3-G595R mutants is not related to the mutation avoidance function of RAD3. The mutation frequency in the SSL1-T242I mutant (2.0 × 10-9 mutations/viable ABX108-5B cell) is also not different from wild type (P = 0.74).

View this table:
TABLE 3

Growth rates

Several rad3 and ssl1 mutants that display extreme sensitivity to UV light have been isolated (Reynolds and Friedberg 1981; Wilcox and Prakash 1981; Yoonet al. 1993). We found that the rad3-G595R and SSL1-T242I mutant strains are nearly as resistant to UV light as wild type (Figure 1). In contrast, rad3-20 mutant cells (Naumovski and Friedberg 1986) that possess a normal recombination phenotype (Bailis and Maines 1996) are highly UV sensitive. These results suggest that the phenotypes conferred by the rad3-G595R and SSL1-T242I mutations are not due to an NER defect.

View this table:
TABLE 4

Transcription induction and steady-state mRNA levels in wild-type and mutant cells

rad3-G595R and SSL1-T242I act together to influence short-repeat recombination: As discussed above, our work suggests that the rad3-G595R mutation disrupts the barrier against recombination between sequences below 250-300 bp in length (Bailiset al. 1995; Bailis and Maines 1996). To extend our observations we determined the frequencies of spontaneous deletion by recombination between 103-, 223-, and 415-bp nontandem repeats in wild-type and rad3-G595R mutant cells. If the rad3-G595R mutation specifically affects the control of SRR we expected that the frequency of recombination between the 415-bp repeats, which are longer than the 250- to 300-bp threshold between long- and short-repeat recombination (Jinks-Robertsonet al. 1993), would be minimally altered. Concomitantly, we expected that the frequency of recombination between the 223-bp repeats, which are just below the threshold length, would be more increased, while the frequency of recombination between the 103-bp repeats, which are considerably below the threshold length, would be the most increased. As expected, we found that the rad3-G595R mutation increases deletion frequencies 2-fold for the 415-bp repeats (P = 0.0002), 4-fold for the 223-bp repeats (P = 0.0001), and 10-fold for the 103-bp repeats (P = 0.0001) (Figure 2). The SSL1-T242I mutation has a similar but smaller effect on the control of SRR as deletion frequencies are unchanged from wild type when the repeats are 415 bp (P = 0.73), but are increased 2-fold for 223-bp repeats (P = 0.02) and nearly 3-fold for 103-bp repeats (P = 0.005; Figure 2). Most interestingly, however, the rad3-G595R SSL1-T242I double mutant displays deletion frequencies for the 415-bp (P = 0.65), 223-bp (P = 0.12), and 103-bp repeats (P = 0.20) that are not statistically different from wild type (Figure 2), indicating that the rad3-G595R and SSL1-T242I mutations suppress each other. These results indicate that the rad3-G595R and SSL1-T242I mutations can act together to influence SRR.

Figure 1.

—Sensitivity of wild-type and mutant S. cerevisiae cells to UV radiation. The fraction of isogenic wild-type (solid boxes), SSL1-T242I (solid circles), rad3-G595R (solid triangles), and rad3-20 (open squares) mutant yeast cells surviving 10, 30, or 100 J/m2 doses of UV radiation was determined as discussed in materials and methods. The means ±2 SE from at least three separate determinations are reported. Mean values that are different by >2 SE are different at the 95% confidence level.

We also used an assay of DNA fragment insertion into homologous genomic sequences (Bailis and Maines 1996) to study the effects of the rad3-G595R and SSL1-T242I mutations on a different type of recombination event. Unlike the spontaneous deletion events that involve sequences that can be on the same or different molecules and are initiated at random, the insertion events involve sequences on different molecules and are initiated by the ends of the DNA fragments. However, these events may be mechanistically similar, involving the formation and resolution of heteroduplex DNA (Klein 1995; Leunget al. 1997; Negrittoet al. 1997). Interestingly, we found that DNA fragment insertion increases in both rad3-G595R and SSL1-T242I mutant cells and that the effect is greatest when the lengths of sequence shared by the fragment and genomic target were shorter than 300 bp (Figure 3). We also found that the insertion of the smallest DNA fragment in rad3-G595R SSL1-T242I double-mutant cells is more like wild type than either of the single mutants (Figure 3). The observation that the rad3-G595R and SSL1-T242I mutations have very similar effects on deletion and insertion suggests that they affect recombination between sequences on the same and on different molecules and, perhaps, between linked and unlinked repeats (Bailiset al. 1995).

Figure 2.

—Relationship between HIS3 duplication length and frequency of deletion by direct-repeat recombination in wild-type and mutant cells. Recombination between duplicate 103-, 223-, or 415-bp HIS3 coding segments in wild-type (solid boxes), rad3-G595R (solid triangles), SSL1-T242I (solid circles), and rad3-G595R SSL1-T242I (open triangles) recreates an intact HIS3 gene and deletes a URA3 marker as described in the text. Frequencies are expressed as the number of deletions per viable cell plated. Assays were conducted at 30°. The median frequency from a minimum of 10 separate determinations was plotted against the length of the repeat in base pairs. The statistical relationships between the median frequencies were determined using contigency χ2 analysis with Yate’s continuity correction and are discussed in the text.

When recombination between short sequences in SSL1-T242I mutant strains was studied, we observed unusual genetic interactions when more than one copy of the SSL1-T242I allele was present in the cell. First, we found that DNA fragment insertion with short homologies is elevated in SSL1-T242I haploids and SSL1/SSL1-T242I heterozygous diploids, but wild type in SSL1-T242I/SSL1-T242I homozygous mutant diploids (Table 5). Therefore, one SSL1-T242I mutant allele in a diploid cell confers a mutant recombination phenotype, indicating that this allele is dominant, while the presence of a second allele suppresses the dominant effect. This unusual gene dosage effect is also observed in SSL1-T242I haploid cells that contain another SSL1-T242I gene on a single-copy plasmid (Table 5). Interestingly, SSL1-3, an allele isolated on the basis of its effect on translational pausing at RNA secondary structure (Yoonet al. 1993), also confers an elevated insertion phenotype in a haploid mutant (Table 5) and a wild-type phenotype in the homozygous mutant diploid (strain X73; see Bailiset al. 1995). Results described below suggest that this unusual phenotypic display may be due to altered interactions between subunits of the heteropentamer. This property of the SSL1-T242I and SSL1-3 alleles will be discussed further in a later section.

Figure 3.

—An assay of the insertion of DNA fragments into the genome by homologous recombination. Isogenic wild-type (solid boxes), rad3-G595R (solid triangles), SSL1-T242I (solid circles), and rad3-G595R SSL1-T242I (open triangles) mutant Ura- His+ spheroplasts were transformed to uracil prototrophy with 0.01-10 μg of one of several DNA fragments containing the complete URA3 gene on a 1.2-kb fragment flanked by the following lengths of HIS3 DNA in base pairs: 1271 (866-bp 5′ flank + 405-bp 3′ flank), 583 (178-bp 5′ flank + 405-bp 3′ flank), 345 (178-bp 5′ flank + 167-bp 3′ flank), 238 (71-bp 5′ flank + 167-bp 3′ flank), or 127 (91-bp 5′ flank + 36-bp 3′ flank). All assays were conducted at 30° as discussed in materials and methods. The percentage of insertions at HIS3 (versus gene conversions at URA3) was determined from a minimum of 200 transformants and plotted against the total length of HIS3 homology in base pairs on the DNA fragment. All percentages were determined to be significantly different from wild type by contigency χ2 analysis except for the SSL1-T242I percentages obtained with the two largest fragments. Inset: Fold difference between percentages of integration events at the HIS3 locus in wild-type and mutant strains vs. length of HIS3 homology.

View this table:
TABLE 5

DNA fragment insertion into a genomic target by homologous recombination in wild-type and mutant cells

Changes in DSB processing in rad3-G595R and SSL1-T242I mutant cells are consistent with a degradative mechanism for the control of short-repeat recombination: We have previously discussed the link between the SRR- and DSB-processing phenotypes in rad3-G595R mutants (Bailiset al. 1995; Bailis and Maines 1996). Here we show that rad3-G595R and SSL1-T242I mutants display nearly identical DSB-processing defects because the half-life of the in vivo linearized plasmid in both mutants (85 min) is more than twice the half-life in wild type (40 min; Figure 4B). When we monitored the appearance and persistence of 3′ single-stranded DNA tails at the ends of the broken plasmids, we found a reduced rate of degradation of both the 5′ and 3′ strands in rad3-G595R and SSL1-T242I mutant cells (Figure 4C). For instance, 60 min are required to see the maximum amount of 3′ single-stranded DNA in the rad3-G595R and SSL1-T242I mutants while only 30 min are required in wild type. Similarly, the half-life of the 3′ single-stranded DNA is 150 min in wild type, but 215 min in the SSL1-T242I mutant and more than 240 min in the rad3-G595R mutant.

Most interestingly, however, we found that the DSB-processing phenotype in the rad3-G595R SSL1-T242I double mutant is very similar to wild type, as HO-digested plasmid DNA has a half-life of only 40 min in both strains (Figure 4B). Similarly, the amount of time required to see the maximum amount of 3′ single-stranded DNA in the double mutant (30 min) is the same as in wild-type cells and half the time required in both single mutants (Figure 4C), indicating that 5′ strand rescission is like wild type in the double mutant. We also observed that the half-life of the 3′ single-stranded DNA requires less time (175 min) in the double mutant than in either single mutant (Figure 4C), indicating that degradation of the 3′ single strand is nearly normal. These data indicate that rad3-G595R and SSL1-T242I suppress each other’s effects on DSB processing. Further, we suggest that rad3-G595R and SSL1-T242I work together to influence SRR by affecting the stability of broken DNA sequences.

Figure 4.

—Physical stability of a plasmid linearized in vivo. (A) Autoradiographs of plasmid DNA linearized in vivo. DNA was prepared from isogenic wild-type and mutant strains containing the plasmids pLAY97 and pGHOT either before a 30-min induction of HO-endonuclease expression (-30 min) or after HO-endonuclease expression was shut off (0-360 min). The autoradiogram pictures pLAY97 uncleaved by HO-endonuclease (6.2-kb band) and after HO-endonuclease cleavage (2.8- and 3.4-kb bands) as described in the text. The stability of the HO-cleaved plasmid DNA over time is indicated by changes in the intensity of the 2.8- and 3.4-kb bands relative to the 6.2-kb band. (B) Kinetics of processing linearized plasmid DNA. The kinetics of the loss of pLAY97 sequences over time from isogenic wild-type (solid squares), rad3-G595R (solid triangles), SSL1-T242I (solid circles), and rad3-G595R SSL1-T242I (open triangles) mutant strains was determined by dividing the sum of the 2.8- and 3.4-kb HO-endonuclease cut pLAY97 signals by the total pLAY97 signal obtained by summing the 6.2-kb uncut pLAY97 signal with the 2.8- and 3.4-kb cut signals and plotting the values (broken fraction) relative to elapsed time. Signal levels were determined using a phosphorimager. (C) Kinetics of single-stranded DNA formation and degradation. Slot blots were prepared from native and alkaline denatured DNA obtained from wild-type and mutant cells as described in materials and methods. The DNA was hybridized with a 413-bp single-stranded RNA probe as described in materials and methods. The signals obtained by hybridization were measured directly with a phosphorimager. The level of ssDNA at each time-point was determined by dividing the signal from the nondenatured DNA by the signal from the denatured DNA. These data were plotted vs. the time that the sample was collected.

The rad3-G595R and SSL1-T242I mutations alter interactions between Rad3, Ssl1, and Tfb1 fusion proteins in two-hybrid experiments: Bardwell et al. (1994) used the two-hybrid system (Fields and Song 1989) to study interactions between wild-type Rad3p, Ssl1p, and Tfb1p and found that Ssl1p interacts with Rad3p, Tfb1p, and itself. We investigated whether the rad3-G595R and SSL1-T242I mutations alter these interactions by swapping restriction fragments containing the rad3-G595R and SSL1-T242I mutations into the Rad3 and Ssl1 two-hybrid fusion plasmids. Our data indicate that both the rad3-G595R and SSL1-T242I mutations alter interactions between the fusion proteins. First, we found that the rad3-G595R mutation abolishes the interaction between the Rad3 and Ssl1 fusion proteins (Table 6). Western blots of protein extracts from these cells revealed that the loss of β-galactosidase activity caused by the rad3-G595R mutation is not due to an inability to express the lexA::rad3-G595R fusion gene (G. Manthey and A. Bailis, unpublished results). This suggests that the phenotypes conferred by the rad3-G595R mutation may result from a change in the interaction between Rad3p and Ssl1p. Interestingly, Ssl1-T242I mutant fusion protein does not interact with Rad3-G595R mutant fusion protein, indicating that the suppressor mutation cannot suppress the failure of Rad3-G595R mutant fusion protein to interact with Ssl1 fusion protein.

View this table:
TABLE 6

Quantitative assessment of protein interaction as represented by β-galactosidase activity

However, the SSL1-T242I mutation does have several interesting effects on the behavior of the Ssl1 fusion protein (Table 6). The Ssl1-T242I mutant and wild-type Rad3 fusion proteins interact twofold better than the wild-type Ssl1 and Rad3 fusion proteins. The Ssl1-T242I fusion protein also interacts with the wild-type Tfb1 fusion protein fourfold better than does the wild-type Ssl1 fusion protein. However, the most interesting effect of the SSL1-T242I mutation is the interaction between two Ssl1-T242I mutant fusion proteins, sixfold greater than that between two wild-type Ssl1 fusion proteins or between wild-type Ssl1 and mutant Ssl1-T242I fusion proteins. This suggests that Ssl1-T242I mutant proteins have a significantly higher affinity for each other than do wild-type Ssl1 proteins. These results suggest that the mechanism by which the SSL1-T242I mutation suppresses the phenotypes conferred by the rad3-G595R mutation may be the result of changes in several protein-protein interactions within the heteropentamer. These interactions may also help explain the unusual gene dosage effect of the SSL1-T242I allele observed in the DNA fragment insertion assays (Table 5) as discussed below.

DISCUSSION

Data from our laboratory indicate that recombination between short, repeated sequences is controlled differently from recombination between longer sequences in S. cerevisiae. In this article we show that deletions by recombination between short repeats are more stimulated by the rad3-G595R mutation than deletions between longer sequences (Figure 2). In addition, we show that a newly isolated allele of the SSL1 gene, SSL1-T242I, also selectively stimulates deletions (Figure 2) and DNA fragment insertions by SRR (Figure 3), but together with rad3-G595R it suppresses SRR (Figures 2 and 3). Together these observations show that the control of different recombination events involving short sequences can be affected by a dialog between the RAD3 and SSL1 genes. It is, perhaps, also indicative of the involvement of the heteropentameric core of both TFIIH and the NER complex in the control of SRR.

The concordance of the DNA fragment insertion and deletion data indicates that they are under the common control of RAD3 and SSL1 and that this control may impinge upon all SRR. These results were presaged by the work of Jinks-Robertson et al. (1993), which showed that the minimum length of DNA sequence necessary to give efficient recombination is similar regardless of where the repeats are located. One clue as to how this may be accomplished was revealed by the DSB-processing phenotype observed in the rad3-G595R and SSL1-T242I mutants. We found that the ends of broken DNA molecules are more stable in rad3-G595R and SSL1-T242I single mutants than in wild type or the rad3-G595R SSL1-T242I double mutant (Figure 4). This is coincident with the increased SRR observed in the single mutants and lower levels of SRR in wild-type and double-mutant cells (Figures 2 and 3). More detailed analysis showed that the stability of both the 5′ and 3′ strands is increased in the rad3-G595R and SSL1-T242I mutants (Figure 4), indicating, as discussed previously (Bailiset al. 1995), that reduced exonucleolytic processing delays degradation of short-repeat sequences adjacent to these ends, increasing the likelihood that they will be rescued by recombination.

By what mechanism are the rad3-G595R and SSL1-T242I mutations affecting the processing of DNA ends and, thereby, the control of SRR? Sancar (1994) has speculated that the helicase activity of the RAD3 gene product may be required to open DNA duplexes around DNA lesions to provide access to nucleases that cleave on either side. Perhaps the heteropentamer is involved in separating DNA strands at the site of a nick or a break, facilitating processing by an exonuclease(s). A loss of strand-separating activity in rad3-G595R and SSL1-T242I mutant cells may explain their defective control of SRR. Because the rad3-G595R and SSL1-T242I mutants display evidence of altered transcription (Table 4) it is possible that this is a response by TFIIH to lesions encountered during transcription of a damaged DNA template. Restoration of wild-type SRR in the rad3-G595R SSL1-T242I double mutant suggests that this is a function of the heteropentamer and that the mutations may make compensatory structural changes to the complex. Two-hybrid analysis with wild-type and mutant Rad3 and Ssl1 fusion proteins suggest that rad3-G595R and SSL1-T242I may alter intersubunit interactions, supporting this hypothesis.

Alternatively, the dominant SRR phenotypes of the rad3-G595R and SSL1-T242I mutants (Table 5) may indicate that the rad3-G595R and SSL1-T242I gene products actively inhibit the degradation of the ends of broken DNA molecules by failing to release them and interfering with exonuclease access. The fact that the mutation in rad3-G595R alters the putative DNA binding domain (Gorbalenya et al. 1995) of the helicase supports this hypothesis. Further, the compensatory effects of the rad3-G595R and SSL1-T242I mutations in the double mutant suggest that Rad3p and Ssl1p together control the access of exonucleases to the DSBs. Interestingly, Garfinkel and colleagues have recently isolated alleles of RAD3 and SSL2 that may stimulate Ty transposition by stabilizing Ty cDNA (Leeet al. 1998). Together, these analyses indicate that a complex involving several NER and TFIIH-associated proteins influences genome stability by determining the stability of the ends of DNA molecules.

Another possibility is that the SRR phenotypes of the rad3-G595R and SSL1-T242I mutants may result from defective transcription of genes that encode SRR control factors. Consistent with this suggestion, the SSL1-T242I allele suppresses both the gene expression (Table 4) and SRR (Figures 2 and 3; Table 5) phenotypes conferred by rad3-G595R. However, the SRR phenotype of the rad3-G595R mutant is dominant (Table 5) while its gene expression phenotype is recessive (Table 4), indicating that Rad3p has gained a function that alters SRR while it has lost an activity required for proper gene expression. This suggests that the relationship between the SRR and transcription control functions of Rad3p may not be at the level of gene expression.

Our analysis of the effects of the SSL1-T242I and SSL1-3 mutations on SRR revealed an unusual pattern of genetic control. The SSL1-T242I and SSL1-3 alleles can suppress their own effect on SRR in homozygous mutant diploid cells and haploid cells that have one chromosomal copy and one centromere-plasmid copy of the SSL1-T242I allele (Table 5). One possible explanation for this effect could be that having two copies of these mutant alleles alters SSL1 expression in a way that nullifies their individual effects on SRR. We found, however, that the suppression is not due to altered transcription of SSL1 because steady-state levels of SSL1 mRNA were the same in wild-type, SSL1-T242I/SSL1 heterozygous, and SSL1-T242I/SSL1-T242I homozygous mutant diploid cells (T. Negritto and A. Bailis, unpublished results).

Another possible explanation for the self-suppressing effect of SSL1-T242I is suggested by the interactions of the Ssl1-T242I mutant fusion proteins in our two-hybrid experiments. We found that the Ssl1-T242I fusion protein displays twofold and fourfold enhanced ability to interact with wild-type Rad3 and Tfb1 fusion proteins, respectively (Table 6), perhaps indicating that the altered SRR phenotype of the SSL1-T242I mutant is a result of changes within the heteropentamer. Our data also indicate a Ssl1-T242I fusion protein interaction stronger by sixfold than interactions of Ssl1 wild-type or Ssl1 wild-type and Ssl1-T242I mutant fusion proteins (Table 6). This indicates that Ssl1-T242Ip mutant subunits may tend to self-associate, which could interfere with the assembly of the heteropentamer if a significant proportion of Ssl1-T242Ip is sequestered. Increasing the copy number of the SSL1-T242I allele might lead to the sequestration of a greater proportion of Ssl1-T242Ip, separating it from the complex that brings about altered SRR. Previously, it was shown that an allele of TFB1 that blocks the interaction between Tfb1p and Ssl1p confers temperature-sensitive growth and UV sensitivity phenotypes (Matsuiet al. 1995; Swederet al. 1996). Together with our observations this work supports the idea that specific interactions between the subunits of the heteropentamer control its various biological functions, allowing diverse cellular processes to be controlled by a limited set of proteins.

Acknowledgments

We thank J. Nickoloff, D. Garfinkel, A. Rattray, S. Gangloff, and anonymous reviewers for critiquing the manuscript. We also thank P. Hanawalt, L. Bardwell, A. Lehmann, J. Termini, B. Shen, R.-J. Lin, T. Krontiris, and J. Rossi for helpful discussions. In addition we thank J. McDonald, L. Bardwell, L. Scherer, T. Donahue, H. Klein, S. Elledge, S. Fields, J. Feaver, J. Nickoloff, F. Heffron, and P. Hieter for providing yeast strains and plasmids. This work was supported by U.S. Public Health Service grants GM-57484 and CA-33572 and funds from the Beckman Research Institute and the City of Hope National Medical Center.

Footnotes

  • This work is dedicated to the memory of Brenda Knowles, scientist and friend.

  • Communicating editor: M. Lichten

  • Received April 19, 1998.
  • Accepted June 24, 1998.

LITERATURE CITED

View Abstract