Copyright 0 1989 by the Societyof America

The Genetic Control of Direct-Repeat Recombinationin Saccharomyces: The Effect of rad52 and radl on Mitotic Recombination atGALlO, a Transcriptionally Regulated Gene

Barbara J. Thomas’ and Rodney Rothstein

Department of Genetics and Development, Columbia University College of Physicians and Surgeons, New York, New York 10032 Manuscript received May 17, 1989 Accepted for publication September 8, 1989

ABSTRACT We have previously showndirect-repeat recombination events leadingto loss of a plasmid integrated at the GALlO locus in Saccharomyces cerevisiae are stimulated by transcription of the region. We have examined the role of two recombination- and repair-defective mutations, rad1 and rad52, on direct repeat recombination in transcriptionally active and inactive sequences. We show that the RAD52 gene is required for transcription-stimulated recombination events leading to loss of the integrated plasmid. Similarly, Gal+ events between the duplicated repeats that retain the integrated plasmid DNA (Gal+ Ura+ replacement events) are reduced 20-fold in the rad52 mutant in sequences that are constitutively expressed. Incontrast, in sequences that are not expressed, the rad52 mutation reduces plasmid loss events by only twofold and Gal+ Ura+ replacements by fourfold. We also observe an increase in disome-associated plasmidloss events in the rad52 mutant, indicative of chromosome gain. This event is not affected by expression of the region. Plasmid loss events in rad1 mutant strains are reduced only twofold in transcriptionally active sequencesand are not affected in sequences that are repressed. However, the radl and rad52 double mutant shows a decrease inplasmid loss events greater than the sum of the decreases in the rates of this event displayed by either single mutant in both constitutive and repressed DNA, indicating a synergistic interaction between these two genes. The synergism is limited to recombination since the radl rad52 double mutant is no more sensitive when compared with either single mutant inits ability to survive radiation damage. Finally, the recombination pathway that remains in the double mutant is positively affected by transcription of the region.

ENETIC recombination in most organisms oc- orderingthese genes into genetic pathways(GAME G curs by multiple pathways. In prokaryotic sys- and MORTIMER1974; HAYNES and KUNZ198 1; KUNZ tems such as Escherichia coli or the bacteriophage X, and HAYNES198 I ; MALONE1983). the isolation and characterizationof mutations defec- The RAD52 gene was originallyidentified by a tive in various steps of the recombination process has mutation that confers sensitivity to x-rays (RESNICK contributed greatly tothe understanding of the mech- 1969). Based on physical analysis of irradiated cells, it anism of theseevents in vivo (reviewedin SMITH was concluded that the RAD52 gene product is nec- 1987). In the yeast Saccharomyces cerevisiae, mutants essary for the repair of double-strand breaks in the defective in meiotic or mitotic recombination or both DNA(Ho 1975; RESNICKand MARTIN 1976).In have been isolated by their sensitivity to various types addition,the RAD52 geneproduct is requiredto of radiation or by directrecombinational screens repair the double-strand lesion induced in HO-cata- (GAMEand MORTIMER1974; HAYNES and KUNZ1 ; 198 lyzed mating-type interconversion (MALONEand Es- KUNZand HAYNES 1981; ESPOSITOet al. 1984; HOL- POSITO 1980; WEIFFENBACHand HABER1981). De- LINGSWORTH and BYERS1989). Meioticrecombina- tailed analyses of specific mitoticrecombination events tion mutants have also been recovered in screensfor have shown differing effectsof the rad52 mutation in sporulation-defective cells (ESPOSITOand KLAPHOLZ different assay systems. In one case, direct-repeat re- 1981) or in screens for meiotic lethals (ENGEBRECHT ciprocal exchange events leading to loss of a plasmid and ROEDER1989). However, although some of these integrated at the HIS4 locus were found to be only mutations have been studied quite extensively in spe- slightly affected by a mutation in rad52 (JACKSON and cific assay systems, there has been little progress in FINK 1981). However, events described as gene con- versions weredramatically reduced (JACKSON and ’ Currentaddress: Department of Biological Chemistry,University of California, Los Angeles, California 90024. FINK 1981). Recombination events leading to loss of The publication costs of this article were partly defrayed by the payment of page charges. This article must therefore be hereby“advertisement” marked a SUP4-o allele located within a clusterof the repeated in accordance with 18 U.S.C. 51734 solely to indicate this fact. element, delta, on chromosomeX were reduced 100-

Genetics 123 725-738 (December, 1989) 726 B. J. Thomas and R. Rothstein fold by a mutation in the RAD52 gene (ROTHSTEIN, below) are recovered at nearly wild-type levelsin HELMSand ROSENBERG1987). Similarly, recombina- transcriptionally inactive rad52 cells, although this tion between small inverted repeats displays a 100- class of recombinant was previously shown to be ex- fold decrease in the rad52 mutant(WILLIS and KLEIN tremely sensitive to a mutation in rad52 (JACKSON and 1987). Recombination occurring by sister chromatid FINK1981). On the other hand, the rateof this event exchange is dependenton RAD52 (FASULLOand is reduced 20-fold in transcriptionally active cells in DAVIS1987). Integration of a gappedplasmid into the therad52 mutant, although transcriptiondoes not is blocked in therad52 mutant (ORR- affect the rate of this event in RAD wild-type strains. WEAVER,SZOSTAK and ROTHSTEIN198 1). Finally, In contrast to other studies (R. KEIL, personal com- directrepeat recombination events initiated by an munication),a mutation in RADl results in only a HO-induced double-strand-break also require a func- twofold reduction in plasmid loss events in expressed tional RAD52 gene (RUDINand HABER1988; NICKO- sequences, and shows no effect on plasmid loss events LOFF et d. 1989). Ingeneral, the rad52 mutation in repressed DNA. The radl rad52 double mutant results in a striking reduction in what have classically displays a dramatic decreasein the rateof plasmid loss been termedgene conversion events. Incontrast, events. These experiments show that a wild-type copy reciprocal exchange events are reduced by varying of RAD52 can compensate for a defect in radl, but a amounts in rad52 cells. It is clear that the RAD52 functional RAD1 gene only partially rescues the rad52 gene plays a major role in recombination in mitotic mutantphenotype, which suggests that RADl and cells; however, since recombination is not completely RAD52 are components of two overlapping pathways eliminated in rad52 mutant cells, it suggests that an- for mitotic recombination in yeast. Furthermore, other, RAD52-independent recombination pathway based on the different sensitivities of these events to must also beoperating (HABERand HEARN1985; a mutation in rad52 in transcribed versus repressed HOEKSTRA, NAUGHTONand MALONE 1986). DNA, we proposethat transcription-stimulated re- The RADl gene is a component of the excision- combination probably proceeds via a different path- repair pathway for UV damage in yeast (REYNOLDS way fromrecombination in transcriptionally silent and FRIEDBERG198 1 ; WILCOXand PRAKASH198 1). sequences. Finally, since the radl rad52 double mu- Recently, studies of mitotic recombinationevents tantdoes not exhibit a concomitant synergistic in- stimulated by the Pol1 rDNA promoter, HOT1 (KEIL crease in sensitivity to damage by radiation when and ROEDER 1984; ROEDER,KEIL and VOELKEL-MEI- compared with either single mutant, the recombina- MAN 1986), show that a mutation in RADl reduces tion function of these gene products is independent HOT1-stimulated recombination events (R. KEIL, per- of their role in survival of radiation damage. sonal communication). Similarly, the RADl gene has MATERIALS AND METHODS been shown to be required for recombinationbetween direct-repeats, as well as for homologous integration Strains: The strainsused in these experiments are de- of linear plasmids (SCHIESTLand PRAKASH1988). scribed in Table l. A TRPl disruption of rad52 (SCHILDet al. 1983a, b) was introduced into the genome by transform- Based on analysis of recombination events between ing linearized DNA (ROTHSTEIN 1983) into yeastmade heteroallelic duplications, it was suggested that RADl competent by the LiCl method (ITO et al. 1983). The pres- may be involved in the processing of recombination ence of the disruption was confirmed byanalysis of the intermediates containing short stretchesof heterodu- genomic DNA of transformants by blots (SOUTHERN1975), genetic linkage analysis to known mutations, and sensitivity plex DNA (KLEIN 1988). Inaddition, studies of of themutant strain to ionizingradiation (50 krad, in a RAD5'Zindependent recombination demonstrate that cesium source emitting 7.8 krad/hr). The radl disruption exposure of rad52 cells to UV-irradiation results in a was previously constructed in this laboratory in an isogenic wild-type spectrum of recombination events (E. SUC- genetic background (RONNEand ROTHSTEIN1988), and is ARMAN and J. HABER, personal communication). marked with the LEU2 selectable marker. Segregants con- taining either a single rad mutation or both were made by These experiments suggest a role for excision-repair crossing the disrupted strainsto previously constructed iso- gene products in genetic recombination in yeast. genic strains containing null mutations in the GAL regula- We examined the effect of the radl and rad52 tory genes gal4 or gal80 and the plasmid pWJ227 (Figure mutations in transcription-stimulated and transcrip- 1) integrated at the GALlO locus (THOMASand ROTHSTEIN 1989). tion-independent recombination in the hope of eluci- Plasmid integration:The plasmid pWJ227 was described dating the genetic pathways of these recombination previously (THOMASand ROTHSTEIN1989; Figure 1). The events. The results of these experiments show that plasmidwas targeted to integrate at the GALlO locus by the transcription-stimulated recombination we ob- linearizing plasmid DNA at the unique HpaI site within the serve in gal80 constitutive cells is specifically elimi- GALlO fragment on the plasmid (ORR-WEAVER,SZOSTAK and ROTHSTEIN1981), and transforming yeast cells made nated by a mutation in the RAD52 gene. Interestingly, competent by the LiCl method (ITO et al. 1983). Gal+ recombination events that retain the integrated Fluctuation analysis of recombination events:Fluctua- plasmid DNA (Gal+ Ura+replacement events; see tionanalyses were performed following themethods of Direct-Repeat Recombination 727

TABLE 1 Yeast strains

Strain Genotype" W586-6C MATa ga180::LEU2gallO::pWJ227-URA3 W586-8D MATa ga180::LEU2gallO::pWJ227-URA3 W586-9A MATa gal8O::LEU2gallO::pWJ227-URA3 W586-6D MATaga14::LEU2gallO:pWJ227-URA3 W586-8A MATa gal4::LEU2ga110::pWJ227-URA3 W586-9D MATa gal4::LEU2gallO::PWJ227-URA3 W625-17B MATaga14::LEU2 rad52::TRPlgallO::pWJ227-URA3 W625-20D MATaga14::LEU2 rad52::TRPlgallO::pWJ227-URA3 W628-3A MATa gal80::LEU2rad52::TRPl gallO::pWJ227-URA3 W628-5A MATa ga18O::LEU2rad52::TRPl gallO::pWJ227-URA3 W628-13B MATa ga180::LEU2rad52::TRPl gallO::pWJ227-URA3 W672-4B MATaga14::LEU2 radl::LEU2gallO::pWJ227-URA3 W672-14D MATagal4::LEU2radl::LEU2gallO::pWJ227-Uk43 W673-2B MATaga180::LEU2radl::LEU2gallO::pWJ227-URA3 W673-3C MATaga180::LEU2 radl::LEU2gallO::pWJ227-URA3 W673-7C MATa gal80::LEU2radl::LEU2gallO::pWJ22i'-URA3 W674-6A MATa gaEBO::LEU2radl::LEU2 rad5Z::TRPl gaLIO::pWJ227-URA3 W674-1 IC MATa ga180::LEU2radl::LEU2 rad52::TRPl gallO::pWJ227-URA3 W674-12B MATaga180::LEU2radl::LEU2 rad52::TRPl gallO::pWJ227-URA3 W675-7D MATaga14::LEU2radl::LEU2 rad52::TRPl gallO::pWJ227-URA3 W675-8B MATaga14::LEU2radl::LEU2 rad52::TRPl gallO::pWJ227-URA3 W675-1 IC MATa ga14::LEU2radl::LEU2 rad52::TRPl gallO::pWJ227-URA3 W569-1B MATa leu2-3,112 his3-11,15ura3-1 ade2-1 canl-100SUP4-o::HIS3 trp5-2 trpl-1 lys2-1gallO::LEU2 W569-7D MATa leu2-3,112 his3-11,15 ura3-1 canl-100 ade2-1SUP4-o::HIS3 trp5-2 gallO::LEU2 W569-2A MATa leu2-3,112 his3-11,15 ura3-1 canl-100 ade2-1SUP4-o::HIS3 trpl-1 gallO::LEU2 W569-5A MATa leu2-3,112 his3-11,15 ura3-1 canl-100 ade2-1SUP4-o::HIS3 tr81-1 callO::LEU2

a All strains except the W569 segregants are isogenic to W303 as described in MATFXIAIS AND METHOLE, and contain the following markers: leu2-3,112 his3-11,15 trpl-1 canl-100 ade2-1Disruption ura3-1. alleles are followed by "::" and thegene contained in the disruption. ThegallO allele is disrupted with the plasmid pWJ227 containing the URA3 selectable marker, as shown in Figure 1.

LURIAand DELBRUCK(1943) and the data was analyzed by Mating rescue of GAL10 recombinants that arise in a the method of the median of LEAand COULSON(1 949) as gal4 mutant strain:Cells containing a mutation in gal4 that described in THOMASand ROTHSTEIN(1989). Briefly, cells become Gal+at the CALI0 locus cannot be selected directly were plated to single colonies on nonselective medium con- on galactose medium because the strains are constitutively taining a noninducing, nonrepressing carbon source to allow repressed for expression of the GAL genes. GAL10 recom- full expression of the gal80 constitutive allele, which displays binants can be rescued by mating with a GAL4 gall0 tester some glucose repression (TORCHIAet al. 1984). Seven or strain. Since mating rescue does not occur at 100% effi- nine independent colonies were used for each experiment, ciency, it is necessary to determine theproportion of mated and experiments were repeated at least three times. Typi- cells in each experiment. The proportion of mated cells can cally, YPS medium was used (1% yeast extract, 2% bacto- be determined using a color assay system as described pre- peptone, 2% sucrose), although the results were identical if viously (THOMASand ROTHSTEIN1989). This proportion is the carbon source was raffinose (2%)or glycerol-lactate(3% used to determine the frequency of Gal+ recombinants in a each; data not shown). A plug containing an entire colony gal4 population. Cells were mated in suspension by mixing was picked, and cells were resuspended in water, sonicated equal numbers of each parent in 5 ml of YPD pH 4.5 (YPD briefly, and appropriate dilutions were plated to thefollow- is 1% yeast extract, 2% bactopeptone, and 2% dextrose), to ing media: Com, synthetic complete medium, to determine a total cetl concentration of 6 X lo6 cells/ml. After 90 min viable colony-forming units (cfu) per ml; 5-FOA, synthetic of gentle shaking (1 50 rpm in a rotary shaker) at 30 ', cells complete medium containing 750 rg/ml5-fluoro-oroticacid were pelleted gently to force cells of opposite mating type (5-FOA) and 50 rg/ml uracil, to select for Ura- cells; Gal into contact with each other. Cells were gently resuspended, Com, synthetic complete media containing 2% galactose as and were placed at 30" to shake for another 60 min (250- a carbon source, with 0.25% Tween-80 and 7.2 rg/ml 300 rpm). Alternatively, cells were mated on nitrocellulose ergosterol to permit membrane synthesis under anaerobic filters on a YPD plate (DUTCHERand HARTWELL1982). conditions (MORPURCOet al. 1964). Com and 5-FOA plates Equal numbers of cells were mixed to a final concentration were incubated at 30°C for threedays. Gal Com plates were of about 4 X lo' cells in 1 mlof sterile HPO. Cells were incubated at 30" for 5-7 days under anaerobic conditions forced through a sterile filter unit (Gelman Sciences, Inc.) (BBL GasPak Anaerobic System) to reduce nonspecific back- onto a 13-mm nitrocellulose disc. Discs were removed onto ground growth (DOUGLASand HAWTHORNE1966). a YPD plate and incubated at 30" for 2 hr. After mating, 728 B. J. Thomas and R. Rothstein

B RV S Hp AK E (VA) P PH RV HS K E A P RV 0 I GAL7 gallOAKpnl ga110-3ii GAL 1

2.6 kb 2.6 kb

FIGURE1.-Structure of the GAL locus on chromosome II bearing the integrated plasmid pWJ227. The 2.6 kb of duplicated GAL10 sequence are indicated below the figure. The homology extends from an Ana1 site in GALI, which was destroyed at the plasmid junction in the left repeat, to the SalI site in GALIO. The left repeat is a full-length copy of GAL10 with a 4-bp deletion of a KpnI restriction site, and the right repeat is a 3"truncated copy of GAL10. The wavy line represents yeast chromosomal sequences, the straight line represents plasmid DNA. The URA3 gene is depicted by an open box, and the direction of transcription is shown by an arrow below the gene. The small filled box is the GALI-10 UAS (upstream activating sequence). The arrows indicate the presence of functional GAL promoters, and thearrowhead is in the direction of transcription. S, SalI, K, KpnI, E, EcoRI, A, AvaI; P, PvuII; RV, EcoRV; H, HindIII;B, BgZII; Hp, HpaI;(E/A), a fusion between EcoRI and AvaI that does not restore either site; AK, the KpnI deletion mutation. the discs were washed in 1 ml Hz0 to remove cells. In both If the chromosome becomes disomic, an extra LYS2 gene procedures, appropriate dilutions were made before and segregates in the cross and 4+:0- and 3+:1- tetrads forlysine after mating to determine theproportion of mated cells and prototrophy result in addition to theexpected 2+:2- pattern. to determine the number of cell divisions occurring in the Because GALlO+ recombinants that occur in a gal4 mutant rescued population. The remaining mated cells were pel- background are rescued by mating, these diploids can be leted and resuspended in Hz0 to plate on galactose medium. sporulated directly and analyzed. Lysine segregation could For the suspension mating procedure, the mating efficiency not be analyzed in all of these recombinants due to the was generally on the order of 1-3% diploids formed, while genotype of the rescuing strain; however Gal-Ura linkage for the filter procedure, this value was normally 3-10%. To was determined. avoid expansion of the diploid population during incubation Heteroallelism analysis: The 5-FOARmutants arising in on YPD, assays wereincluded in the analysis only if growth the rad1 rad52 gal80 strain were examined for their ura3 of the rescued haploid strain was limited to 51 doubling. allele by heteroallelism analysis to determinewhether or not Since the rate of recovery of the Gal+ Ura- class of recom- they arose by an ectopic event with the binant is the same after selection on galactose and 5-FOA chromosomal ura3-1 allele. Cells were patched to a YPD (Table 2), we are confident that the mating rescue proce- master plate, and mated to a strain containing the markers dure accurately reflects the Gal+ proportion of the recom- ura3-1 and trp5-2. The mated cells were plated either di- binant population. rectly to sporulation medium to induce meiotic levels of Genomic blot analysis of recombinants: The genomic recombination, or first to medium lacking tryptophan to blot procedure (SOUTHERN1975) was described previously select for diploids, and then to sporulation medium. There (THOMASand ROTHSTEIN1989). was no qualitative difference in the result using either pro- Yeast colony hybridization: Yeast colony hybridization cedure. Control ura3-I/ura3-l diploids produced no pa- (HICKS,HINNEN andFINK 1979) was performed as modified pillae under this regimen. Of 25 of the 5-FOARcells tested, by ROTHSTEIN (1985). Briefly, cells were patched onto a 23 gave rise to Ura+ papillae when mated to a strain that nitrocellulose filter placed on a YPD plate and incubated at was ura3-I. The frequency of Ura+ papillae varied consid- 30" overnight. Colonies were lysed by removing the filter erably from diploid to diploid, probably reflecting the dis- and placing it on a piece of Whatman 3MM paper in a Petri tance between the heteroalleles. Of the two remaining Ura- dish soaked in 3 ml of a solution of 1 M sorbitol, 20 mM derivatives examined, one displayed complementation for EDTA, 5 pg/ml zymolyase-20Tat 37 O overnight. The plate was sealed with plasticto prevent the filter from drying out. growth on medium lacking uracil when mated to a ura3-1 The following day, the filters were processedas for bacterial strain, and likely represents a mutation in another gene in colony hybridization by placing them for 5 min onto What- the uracil biosynthetic pathway, such as ura5 (BOEKE,LAC- man 3" paper soaked in the following solutions: (1) 0.5 ROUTE and FINK 1984). The second derivative did not M NaOH, 1.5 M NaCI; (2) 1.5 M Tris-CI (pH 7.6), 1.5 M produce Ura+ papillae in this assay. NaCI; (3) same as (2). The filters were then air-dried for Analysis of radiation sensitivity in the RAD mutants: several hours, baked and hybridized as usual. Cells were grown to mid-log phase in YPD liquid medium, Genetic analysis of disomes: Putative gene convertants were sonicated to break up any clumps of cells, and appro- were distinguished from disomes by analyzing genetic link- priate dilutions were plated to YPD plates. The plates were age between the Gal+ andUra+ phenotypes. Gal+ Ura+ exposed to the radiation source for the indicated lengths of recombinants recovered in a gal80 mutant background were times, and then were incubated at 30" until colonies were crossed to a gall0::LEUPlys2-I ura3-1 strain of the appro- visible. The gamma source was a cesium source emitting 7.8 priate mating type. The Gal+ and Ura+ phenotypes segre- krad/hr. Viable cfu/ml were plotted as a function of time. gate as unlinked genes if the markers are contained on two Statistical analysis of rates: Significance was determined disomic chromosomes. Additionally, the recombinant by chi-square analysis and by the Fisher exact probability strains contain the wild-type LYSP allele on chromosome ZI. test (SIEGEL1956). Direct-Repeat Recombination 729

TABLE 2 Effect of mutations in radl and rad52 on direct repeat recombination in transcriptionally active and inactive cells

Selected on 5-FOA Selected on galactose

Rate of Ura- Rate of Gal+ Rate of Relative rate Rate of Ura- Gal* Rate of Gal+ Ura- Gal+ Ura+ of Ura-*

wt gal4 2.0 X 10-~ 0.7 X 1.3 X 10-~ 0.9 X 10-~ 4.3 x 1 gal80 29.3 X 10-5 12.5 X 10-~ 14.2 X 10-~ 12.5 X 10-~ 3.5 x 10-6 15 rad 1 gal4 1.7 X 10-5 0.4 X 10-~ 1.5 X 10-~ 1.2 X 10-~ 4.0 X low6 1 gal80 13.4 X 10-~ 6.1 X 10-5 8.6 X 8.0 X 10-~ 5.9 x lo+ 7 rad52 gal4 rad52 1.1 X 10-~ 0.5 X 10-~ 0.5 X 10-~ 0.3 X 10-~ 1.8 X 0.5 gal80 1.8 X 0.7 X 10-~ 1.0 X 10-~ 0.9 X 10-~ 0.8 X 1

radl rad52radlgal4 2.0 X 10-~ ND‘ ND‘ 0.01 ea180 8.0 X 10-~ 5.0 X 10-~ 6.0 X 10-~ 6.0 X 10-~ 6.0 X 10-7 0.04

Experiments were done by fluctuation analysis as described in MATERIALS AND METHODS. The “Rateof Ura”’ is the total rate of 5-FOAR cells in the population; these cells were screened by replica-plating for their galactose phenotype, and the rateof Ura- Gal+ was determined independently. Similarly, the “Rate of Gal+” is the total rate of galactose-positive recombinants. These cells were screened for their uracil phenotype to determine the rates of Gal+ Ura- and Gal+ Ura+. a The full genotypes of the strains used in these experiments is in Table 1. Relative to the rate in RADl RAD52 ea14 cells. ‘ Not determined. -

EXPERIMENTALRESULTS copy of the GALlO gene (Figure 2). Ura- cells will be either Gal+ or Gal-, depending on whether or not the An assay system for direct repeat recombination KfinI-deletion mutation has been left in the chromo- at GAL10 Weused a system described previously some. Alternatively, events that restore GALlO gene (THOMASand ROTHSTEIN1989), in which a plasmid function can be selected. Since the GAL genes are containing the first three-fourths of the GALlO gene constitutively expressedin aguZ80 mutant, Gal+ events and the GALl-10 divergent promoter is targeted to can be selected directly. However, we cannot directly integrate at the chromosomal GALlO locus (ORR- select for GALlO recombinants in the guZ4 mutant WEAVER,SZOSTAK and ROTHSTEIN198 1; Figure 1). background, since gal4 cells are unable to grow on This creates a 2.6-kb duplication of GALlO coding galactose-containing media. GALlO recombinants in sequencesflanking the integrated plasmid andthe the guZ4 mutant background can be rescued by mat- URA3 selectable marker. Neither repeat contains a ing, as described in MATERIALS AND METHODS. Gal+ functional GALlO gene; the repeat on the right is recombinants that occur by loss of the integrated deleted for the 3”portion of the gene, and the repeat plasmid will result in a cell that is Gal+ and Ura-. on the left contains an in vitro constructed frameshift Therefore, theGal+ Ura- class of recombinant can be mutation that results in a translation termination co- independently recovered under both selective condi- don. Expression ofthe repeats is regulated in isogenic strains by null mutations in either the positive activa- tions, ensuring that theselections usedare not biasing the recombination events that we detect. tor, GAL4 (JOHNSTON and HOPPER1982), or the repressor, GAL80 (TORCHIAet ul. 1984). A guZ4 mu- Three types of Gal+Ura+ events are obtained (Fig- tant cannot express the GAL structural genes; a gal80 ure 2). One class is disomic for chromosome ZZ, where mutant constitutively expresses these genes to high one chromosome has lost the plasmid sequences leav- levels(reviewed in OSHIMA1982). Isogenic strains ing a wild-type copy ofthe GALlO gene, and the other either wild type or containing disruption mutations in chromosome contains the original plasmid pWJ227 RADl (YANGand FRIEDBERG1984; RONNEand ROTH- integrated at gallo, which contains URA3. A second STEIN 1988) and RAD52 (SCHILD et ul. 1983a, b) were class is a novel structure containing a triplication of used in all experiments, as described in MATERIALS GALlO sequences, with two copies of the integrated AND METHODS. The relevant genotypes of the strains plasmid and the URA3 gene on a single chromosome. used in this study are listed in Table 1. A third class results from replacement of the KfinI- Because the integrated plasmid results in a cell that deletion mutation withwild-type information from is gull0 and URA3,recombination events between the the truncated repeat without loss of the integrated duplicated GALlO repeats can be detected in twoways. plasmid. This results in a single-copy insertion of the By selecting for loss of the URA3 marker with the plasmid containing a functional GALlO gene and the drug 5-FOA (BOEKE,LACROUTE and FINK 1984), URA3 marker. events are obtained that have deleted the integrated Thisthird classof Gal+ Ura+ recombinants can plasmid sequences, regenerating a single full-length occur by a variety of mechanisms (see Figure 3 of 730 B. J. Thomas and R. Rothstein

a. 5-FOASELECTION b. GALACTOSESELECTION

Gar ura- Gal+ Ura+ events: popou +

replacement

Gal+ Ura- popou

disome

lriplicatbn

FIGURE2.-Types of recombinants obtained under galactose and 5-FOA-selection. a, Selection on 5-FOA results in loss of one copy of the duplicated region, the plasmid sequences and the URA3 selectable marker. The GAL10 gene left in the chromosome is either Gal+ or Gal-, depending on whether or not the KpnI frameshift mutation has been left in the chromosome. b, Galactose selection can also result in loss of the plasmid, producing Gal+ Ura- cell. In addition, three types of Gal+ Ura+ events are recovered. Gal+ Ura+ replacements contain a single copy of chromosome 22, where the KpnI frameshift mutation in the full-length repeat has been replaced with wild-type information from the truncated repeat. Gal+ Ura+ disomes contain one copy of chromosome 22 with the original plasmid pWJ227, which is Gal- and Ura+, and a second copy of chromosome 22 which has undergone a plasmid loss event to become Gal+. The final Gal+ Ura+ event observed is a triplication of GAL10 sequences. The full-length repeat on the left is Gal'.

THOMASand ROTHSTEIN1989). For example, a gene bination eventmust be recovered.Since the reciprocal conversion of the KpnI-deletion mutation with wild- products of these mitotic direct repeat recombination type information contained in the 3"truncated repeat events are generally not recoverable under the selec- in GI or G2 can result in a Gal+ Ura+ chromosome. tive conditions used, this type of Gal+ Ura+ recombi- Alternatively, a double crossover event flanking the nant cannot be rigorously classified as a gene conver- frameshift mutation can restore a functional GAL10 sion. In addition, geneconversion may also be respon- gene while maintaining theintegrated plasmid. Fi- sible forpopout events (ROTHSTEIN,HELMS and nally, a Gal+ Ura+ chromosome can occurby mismatch ROSENBERC1987; SCHIESTL,ICARASHI and HASTINCS repair of heteroduplex DNA covering the mutantsite 1988). Thus this term is ambiguous since it too nar- and subsequent segregation of daughter strands after rowly defines the mechanism. Therefore, we will refer DNA replication. Historically, this type of event has to these Gal' events that retain a single copy of the been referred to as gene conversion (JACKSON and plasmid DNA as Gal+ Ura+ replacements, which is a FINK 1981). The termgene conversion was originally mechanistically neutral description of these events. defined as the nonreciprocal transfer of genetic infor- Transcription-stimulated recombinationevents mation, based on the recovery of non-Mendelian seg- specifically require the RAD52 gene product: We regationpatterns during meiotic recombination in examined recombination in gal4 and gal80 cells in the fungi. Therefore, in order totruly designate an event presence of the rad1 and rad52 disruption mutations. as a gene conversion, all the products of the recom- The results are shown in Table 2. The rate of plasmid Direct-Repeat Recombination Direct-Repeat 731 loss asmeasured by 5-FOA-resistance in RAD1 RAD52 TABLE 3 wild-type cells that do not express the GAL genes is Fraction of 5-FOAR events that are Gal+in rad1 and rad52 2 X 1O-5. Cells that actively transcribe the GAL region strains display a 15-foldincrease in recombination events leading to loss of the plasmid (Table 2; THOMASand Genotype gal4 gal80 ROTHSTEIN 1989). We examined the effect of a dis- Expected“ 0.33 0.33 ruption alleleof the RADl gene on direct repeat wt 0.41 0.42 recombination at GALlO (Table 2). In radl mutant rad 1 0.46 0.24 cells that constitutively express the GAL region, a rad52 0.45 0.41 reproducible twofold reduction in the rate of plasmid Fractions were determined by dividing the rate of Gal+ from the loss isobserved whencompared to cells that are RADl 5-FOARpopulation by the total rate of 5-FOAR. a Expected values were determined by assuming that the proba- (P C 0.0001; Table 2). No effect of the radl mutation bility of a crossover in a region is linearly related to the physical is observed in gal4 cells. Thus, it appears that recom- size of that region. bination events in the rad 1 mutant can occur relatively residual recombination left in the double mutant is efficiently. affected positively by transcription. In contrast, the rad52 disruption mutation com- The rad52 and mutations do not change the pletely abolishes the transcription-dependent stimu- rad1 distribution of crossovers in the duplication: We lation in recombination seen ingaZ80constitutive cells, determined the position of the exchange event that decreasing the rate of Ura- recombinants to 1.8 X leads to plasmid loss by examining the Ura- popula- 10-5 (Table 2). Transcription-independent recombi- tion of recombinants for their Gal phenotype. In nation in the gal4 mutant is reduced only about two- models for lossof the plasmid by a reciprocalex- fold in the rad52 strain (P = 0.0007). Because the change, a crossover to the left of the Kpnl-deletion stimulation in recombination at GALlO is dependent mutation will produce a Ura- cell that is Gal+; a on expressionof the GALlO region (THOMASand crossover to the right will result in a Gal-, 5-FOAR ROTHSTEIN1989), it was possible that the effect of cell. A change in the distribution of crossover events the rad52 mutation was to reduce or eliminate expres- through the GALIO repeat might therefore reflect a sion of the GALlO repeats. We examined steady-state different pattern of recombination initiation or reso- transcripts of the construct in the mu- GALIO rad52 lution in the rad mutants. tant strains, and find no effect of the rad52 mutation In wild-type cells, the distribution of crossovers is on expression of the GALlO region (data not shown). approximately related to the physical size of the re- These results suggesttranscriptional activation of the gions undergoing the recombination event; about GALlO construct results in a recombination interme- 40% of the crossovers occur in the 0.9 kb to the left diate that requires processing by the RAD52 gene of the KpnI-deletion mutation to give a Gal+ Ura- product. recombinant, and about 60%are in the 1.7 kb to the Mutations in rad1 and rad52 display synergism right, producing a Gal- Ura- cell (Table 3). The in their effects on direct repeat recombination: We distribution of crossover events in wild-type cells is examined recombination in a radl rad52 strain and not affected by expression of the region. Ura- recom- found that the double mutant has a dramatic effect binants in the rad52 mutant display apattern of on recombination events leading to 5-FOA-resistance. crossover events identical to wild-type cells, and simi- Of the 5-FOAR events recovered in the radl rad52 larly, the distribution of the exchanges is not affected strains, 98% are mutations of the URA3 selectable by expressionof the repeats. Thispattern is also marker carried on the integrated plasmid. Southern observed in the gal80 radl mutant. A slight decrease analysis (SOUTHERN1975) revealed that these Gal- in the recovery of Gal+ Ura- derivatives is observed Ura- cells still maintain the plasmid insertion at the in the repressed gal4 radl strain, however, thisdiffer- GALlO locus. We examined all the 5-FOARevents in ence is not significantly different from wild-type (P = both gal4 and gal80 cellsby colony hybridization 0.08). (HICKS, HINNENand FINK 1979) with nick-translated Analysis of the Gal+ Ura+ eventsin wild-type and pUCl8 sequences to determine the proportionof cells rad mutant strains: Previous analyses of mitotic re- arising by recombination (pUCl8-) vs. mutation combination in yeast suggested that the rad52 muta- (pUCl8+). After adjusting for mutations to ura3 we tion has a severe effect on replacement events, but find that theradl rad52double mutant reduces direct has a smaller, more variable effecton events resolving repeat recombination relative to wild type by approx- reciprocally (JACKSON and FINK 198 1; HABERand imately100-fold in gal4 cells, to 2 X lo-’, and by HEARN1985; HOEKSTRA,NAUGHTON and MAMNE 400-fold in gal80 cells, to 8 X 10”. The resulting 1986; ROTHSTEIN,HELMS and ROSENBERG1987; Ru- fourfold greater rate of recombination in the tran- DIN and HABER1988; SCHIESTLand PRAKASH1988; scriptionally activerad 1 rad52 mutant implies that the KLEIN1988). We examined Gal+ Ura+ events in wild- 732 B. J. Thomas and R. Rothstein type rad1 and rad52 cells by screening the Gal+ pop- TABLE 4 ulation of cells for their Ura phenotype. These Gal+ Distribution of Gal+ Ura+ events between replacements, disomes Ura+ recombinants were screened by genomic blot and triplications in the rad mutants analysis to detect the three classes of events described below. Rate of re- Rate of di- Rate of tripli- Mutant" placement somy cation The first class of Gal+ Ura+ events we will consider are those that replace the KpnI frameshift mutation wt gal4 4.3 X <0.2 X c0.2 X loe6 with wild-typeinformation. In Rad+ cells, the majority gal80 3.4 X 0.1 X <0.1 X of Gal+ Ura+ events observed are replacements. The radl gal4 3.4 X <0.2 X 0.6 X total rate of Gal+ Ura+ replacements is unaffected by gal80 4.9 X IOe6 1.0 X <0.2 X transcription, and is about 4 X in both gal4 and rad52 gal4 1.0 X 0.6 X 0.2 X gal80 cells (Table 4).The radl mutation has no effect gal80 0.2 X IO-' 0.6 X IOe6 <0.6 X lo-' on the rate of Gal+ Ura+ replacements. In the rad52 The Gal+ Ura+ events were screened by Southern analysis of genomic DNA (SOUTHERN1975) to detect the different classes of mutant, we observe a difference in the recovery of events. Gal+ Ura+ replacements depending on the transcrip- The full genotypes of the strains are given in Table 1. tional state of the region. In transcriptionally active cells, Gal+ Ura+replacements arereduced 20-fold are recovered by mating rescue (see MATERIALS AND relative to wild type. However, in transcriptionally METHODS; gal4 strains that are GALIO+ are phenotyp- inactive cells we observe only a fourfold reduction in ically Gal-). However, we examined Gal+ Ura+ cells the rate of Gal+ Ura+ replacement when compared arising in a radl rad52 gal80strain to see whether the with wild type. rad double mutantalso displays a synergistic reduction Disomes associated with a Gal+ popout were also in the recovery of these events. In one experimentin the radl rad52 gal80 strain, nine Gal+ Ura+ colonies observed (Table 4). Representative disomic recombi- representing a minimum of six independent events nants were also analyzed genetically, showing that the were examined molecularly and genetically for evi- Gal+ and Ura+ phenotypes were not linked in these dence of recombination leading to conversion of the strains (see MATERIALS AND METHODS for a description KpnI deletion mutation (Table 5). Analysisof total of the crosses). The rate of this event in cells wildtype genomic DNA revealedthat three of the colonies for the RAD mutations is about 0.1 X 1O-6. The rad52 (recombinants 2-1, 2-4 and 4-1 in Table 5) contain strains show a rate of disomy associated with plasmid restriction fragments consistent with disomy for chro- lossof 0.6 X sixfold elevatedover the level mosome ZI, where one chromosome contains a single observed in wild-type cells. The rate of this event is copy of the GALlO gene and the other contains the not dependent on the transcriptional state of the GAL original integrated plasmid. The remaining six colo- locus. In radl cells, disomes are recovered at wild- nies (1-1, 2-2, 2-3, 3-1, 3-2, and 3-3) appear to have a type levels in the transcriptionally inactive gal4 strain. single chromosome containing the plasmid insertion, However, in the transcriptionally active radl gal80 which is indicative of a replacement event. However, cells, a tenfold increase relative to wild type in the when these cells were analyzed for thepresence of the rate of disomes associated with plasmid loss isobserved KpnI restriction site, only one of the six (2-3) has the (to 1.O X This increase in therate of disomes KpnI restriction site restored, suggesting that the re- may be associated with a decrease in the rate of a maining five arenot true replacements(data not novel class of recombinant in the gaZ80 radl strain shown). (triplications, see below). Genetic analysis was also performed on each of the The last class of Gal+ Ura+ recombinants observed nine recombinants. One of the putative disomes (2-1) consistsof a novel event,a triplication of GALlO displays no linkage between the Gal+ and Ura+ phe- sequences (Table 4). A total of six triplication events notypes, as anticipated if the markers were segregat- were observed in the transcriptionally repressed gal4 ing independently on disomic chromosomes. In addi- mutant strain: five in the radl mutant, and one in tion, this recombinant contains an extra copy of the rad52. In addition, one event containing four copies LYS2 gene, which is located on chromosome IZ distal of the gall0 repeat was also recovered in a gal4 radl tothe GAL locus, indicatinga duplication of this strain (Figure 3). This event was not observed in cells marker. A second putative disomic event (2-4) lost the wild type for the RAD genes, nor in any other gal80 URA3 containingchromosome before the genetic constitutive strain examined. analysis, and was not analyzed further. The final di- Gal+ Ura+ events are recovered in the transcrip- somic event (4-1) displays an unusual segregation pat- tionally active rad1 rad52 double mutant: No Gal+ tern; although this strain appears disomic when ana- events were recovered in the radl rad52 gal4 strain lyzed by Southern blots, genetic analysis reveals only due to the reducedrecombination rates in these cells one LYS2 wild-type chromosome segregating in the and the low efficiency with which Gal+ recombinants cross. In addition, this strain displays linkage between Direct-Repeat Recombination Direct-Repeat 733 abC and PETES 1986; LICHTEN,BORTS and HABER1987). It is possible that the ura3 mutations recovered in the radl rad52 double mutant arose by an ectopic gene conversion between the endogenous ura3-1 allele on chromosome V and the URA3 selectable marker inte- grated at GALlO. Heteroallelism tests were performed on 25 of the Ura- mutants arising in the radl rad52 22*35L16.65 double mutant as describedin MATERIALS AND METH- 10.95" ODS. We found no evidence for ectopic gene conver- sion in these experiments. The radl rad52 double mutant shows no syner- gism in radiation sensitivity:To determine whether the radl rad52 double mutant displays a synergistic effect in response to radiation damage in addition to its dramatic effect on recombination, we examined the ability ofthe double mutant to survive either gamma-irradiation or UV treatment as compared to either single mutant and the wild-type strain. The results are shown in Figure 4. The double mutant is no more sensitive to either gamma or UV irradiation than either single mutant. We conclude that the syn- ergism displayed by the double mutant is limited to its effect on genetic recombination.

DISCUSSION We have previously shown that direct-repeat recom- binationleading to loss of a plasmid integrated at FIGURE3.-Genomic blot analysis of Gal+ Ura+ recombinants GALlO is stimulated 15-fold when the region is ex- containing multiple copies of thegall0 repeat. Genomic DNA was pressed (THOMASand ROTHSTEIN1989). We show isolated as described (THOMASand ROTHSTEIN1989) and was here that the transcription-stimulated recombination digested with Bglll, which cleaves outside of the integrated con- that we observe is dependent on a functional RAD52 struct (see Figure 1). a, Contains two copies of the gall0repeat; b, gene. A mutation in RAD1 alone has only a small contains three copies; c, contains four copies of the gall0 repeat. The size of each fragment is indicated next to the bands in kilobases. effect on the rate of plasmid loss events in constitu- tively transcribed sequences, and shows no effect on the Gal+ and Ura+ phenotypes. No simple model for recombination in repressed sequences. However, we this recombinant can be proposed. observe synergism between the radl and rad52 mu- The six single copy integrants display 2:2 segrega- tations on recombination events in both constitutive tion for lysine prototrophy, demonstrating that they and repressed sequencesin this system in that the rate contain a single copy of chromosome ZI. In addition, of plasmid loss when both mutations are present is the Gal+ and Ura+ phenotypessegregate as linked much lower than the sum of the rates of either single genes. However, we noticed that after crossing four mutation. of thesestrains (3-1, 3-2, 3-3 and 4-1), theUra+ Replacements: We were surprised to see very little segregants are eitherweakly Gal+, or only papillate to effect of the rad52 mutation on Gal" Ura+ replace- Gal+ after several days incubation on galactose me- mentevents in transcriptionally inactive cells. Pre- dium. Because these events did not restore the KpnI vious studies of intrachromosomal direct repeat re- deletion mutation to wild type, it is likely that these combination have shown this event to be extremely cells represent mutation events that partially restore sensitive to rad52 (JACKSON and FINK 1981; ROTH- GALlO function. Therefore, taking into account these STEIN,HELMS and ROSENBERG1987; SCHIESTL and mutation events, we estimate the rate of replacement PRAKASH1988; KLEIN 1988). In contrast, the rate of to Gal+ Ura+ to be about 6 X lo-* in the transcrip- Gal+ Ura+ replacements in the gal80 constitutive tionally active radl rad52 double mutant. strain was reduced 20-fold relative to wild type. This The ura3 mutants arising in the radl rad52 dou- is consistent with the previousstudies cited above, ble mutant do not occurby ectopic gene conversion: which have reported up to a 100-fold reduction in Ectopic gene conversion occurs between homologous replacement events in the rad52 mutant. Therefore, sequences located at nonhomologous positions in the although the absolute rate of Gal+ Ura+ replacement genome (MIKUS and PETES 1982; JINKS-ROBERTSON is the same in gal4 and gal80 RAD52 wild-type cells 734 B.and J. Thomas R. Rothstein

TABLE 5 Analysis of the Gal' Ura' events recovered in the radl rad52 gal80 strain

Plate#- LYS2 Gal-Ura recombinant" Chromosome structureb Kpnl' segregationd segregation Conclusion 1-1 Integrant - 2:2 Linked Reversion of frameshift mutation at GALlU 2- 1 Integrant + wild type Disome Multiple Unlinked Disome copies 2-2 Integrant - 2:2 Linked Reversion of frameshift mutation at GALlO 2-3 Integrant + 2: 2 Linked Replacement 2-4 Integrant + wild type Disome 2:2 All segregants Ura- Putative disome; lost the Ura+ chromosome before analysis 3- 1 Integrant - 2:2 Linked; Gal +/- or pap Reversion of frameshift mutation at GALlU 3-2 Integrant - 2:2 Linked; Gal +/- or pap Reversion of frameshift mutation at GALlU 3-3 Integrant ND 2:2 Linked; Gal +/- Reversion of frameshift mutation at GALlO 4- 1 Integrant + wild type Disome 2:2 Linked; Gal +/- or pap Multiple events? Cells were screened by analysis of genomic DNA by the method of SOUTHERN(1975) as described (THOMASand ROTHSTEIN 1989). Cleavage with PvuII reveals an extra disomic copy of chromosome Zl, while cleavage with KpnI detects the presence of a restored KpnI site in the full-length repeat. Chromosome number was also determined genetically by crossing the cells witha gull0 urn3 lys2 strain as described in MATERIALS AND METHODS and analyzing segregants from this cross. a Refers to the origin of each recombinant. Recombinants from the same plate may be clonally related. * "Integrant" is a single copy of the integrated plasmid at GALlO; "integrant + wild type" is one copy of the integrated plasmid and one copy of the wild-type parental GALlO gene. ' A "-"indicates no cutting of the full-length repeat by KpnI; a "+" indicates that the full-length repeat is sensitive to KpnI digestion; "disome" refers to the presence of fragments indicative to chromosome ZZ disomy; ND is not done. Cells containing one copy of chromosome ZZ will display 2:2 segregation for lysine prototrophy. Cells disomic for chromosome ZZ will contain two copies of LYSZ, and will display 4:0, 3: 1 and 2:2 tetrads for lysine prototrophy.

(3-4 X Table 4), this result implies thatthe Table 4). Therefore, it is likely that therecombination pathways for this event are different depending on event leading to loss of the plasmid is linked mechan- whether the DNA is expressed. We speculate that the istically to the chromosome nondisjunction in these variations in recombination frequencies observed be- cells. One prediction of this model is that at least a tween different loci in the rad52 mutant may reflect fraction of the plasmid loss events we observe occur the level of transcription of the regions being meas- in GP. ured. In addition to defectsin recombination and repair, Disomes: A second class of Gal+ Ura+ recombinants the rad52 mutation has also been associated with an recovered in these experiments consists of a disome increase in chromosome loss (MORTIMER, CONTOPOU- associated with a plasmid loss (Figure 2 and Table 4). LOU and SCHILD1981; HABERand HEARN1985). One mechanism to generate a disome is by chromo- HABERand HEARN(1985) proposed that RAD52-in- some nondisjunction. The rate of chromosome non- dependent recombination proceeds via events initi- disjunction as measured by chromosome loss in yeast ated by single-strand nicks. They suggested that chro- is about (HARTWELLand SMITH1985). In RAD mosome loss was mediated by the inability of the wild-typecells therate of plasmid lossis 2 X in RAD52-independent recombination pathway to re- gal4 cells and 3 X in gal80 cells. Therefore, if solve a half-chiasma structuregenerated by single- plasmidloss and chromosomenondisjunction were stranded invasion, resulting in a chromosome contain- two independent events, we would expect the rate of ingunrepaired breaks. Intheir model, the broken Gal+ Ura+ disomy to be the product of the rates of chromosome is lethal and is therefore not recovered. these two events (lo-" in the gal4 mutant and 2 X This contrasts with our observation that the disomic 10"O in gal80). However, in these experimentsdisomy events representa chromosome gain. It is possible occurs at a rate of about 0.05-0.1 X in consti- that heteroallelic recombination in diploids proceeds tutive and repressed RAD wild-type cells and thegaZ4 by adifferent mechanism in rad52 cells fromthe radl mutant (Table 4,and B. J. T. and R. R., unpub- haploid direct repeat recombination events presented lished observations), and is sixfold higher in the rad52 here. mutant in both transcriptional states. The transcrip- Triplications: A third, novel class of Gal+ Ura+ tionally active gaZ8U radl strain shows a tenfold ele- recombinants was also observed at a low frequency in vation in the rateof disome recombinants (1 .O X transcriptionally inactive radl and rad52strains. This Direct-Repeat Recombination 735 a.

109 108

wild type rad 1 rad52 radl rad52

101 FIGURE4.-Comparison of the 100 I'I'I'I. I .I"'''' sensitivities to gamma- and UV-irra- 270240210180 150 1200 90 60 30 diation of the rad mutant strains with an isogenic wild-type strain. a, Viable colony forming units were deter- time (minutes) mined by plating cells and exposing the plates for varying times to a ce- sium source emitting 7.8 krad/hr of b. ionizing radiation. b, Viable colony forming units were determined by 108 plating cells and exposing the plates 1 for varying times to UV-irradiation. 107

106

lo5 wild-type rad 1 -Q, lo4 n rad52 .-a radl rad52 > lo3

102

lo1 0 10 20 30 40 time (seconds)

event consists of a triplication of GALlO sequences. Triplications are normally produced during direct repeat recombination by an unequal sister chromatid exchange (Figure 5). However, using this construct, a simplecrossover between unequally aligned sister chromatids is not sufficient to generate a functional full-length GALlO gene in the leftmost repeat (Figure 5). One possibility is that two,possibly concerted, events are required to generate the triplication class. However, as argued for the disome recombinants above, the rate of this event in the gal4 radl strain is FIGURE 5.--Illustration of a sister chromatid exchange event in too high to be the productof twoindependent events. the gall0 duplication. Unequal pairing occurs after DNA replica- The majority of the triplications observed occur in tion between the truncated repeat on one chromatid and the full- the transcriptionally inactive gaZ4 radl strains. In the length repeat containing the AKpnI mutation from the otherchro- guZ80 radl mutant, triplications are not observed; matid. A recombination event, indicated by the heavy zig-zag line, can occur between these repeats. A recombination event that re- however, these strains exhibit a tenfold elevated rate solves by a crossover will produce the triplicated structure, shown of popout-associateddisomes (to 1.O X Table 4). below. However, the final structure will always contain the frame- It appears that, in the transcriptionally inactive radl shift mutation in the full-length repeat. 736 B. J. ThomasRothstein and R. strains, Gal+ triplications occur at the expense of the RADI. It is possible that single-stranded nicks left in disome recombinants. One explanation for this obser- the DNA after replication in the cdc9 mutant are vation is that the two events possess a common inter- recombinogenic. This is consistent with the results mediate and that transcription drives the resolution presented here that show that only the transcription- of this intermediate towards the popout disome re- stimulated recombination events are lost in the rad52 combinant. A second model proposes that amplifica- mutant, which we propose initiate by a double-strand tion of the region through multiple rounds of DNA lesion. replication causes the triplicated structures (reviewed One possible model for the roleof RADI in genetic in STARKand WAHL 1984). replicativeA model would recombination can be drawnby analogy with excision explain the recovery of the recombinant containing repair in E. cold. In this process, known as “short- four copies of the gall0 repeat. patch” repair, one toseveral nucleotides surrounding The role of rad1 and rad52 on direct repeat re- the damage are removed and replaced by a polymer- combination in transcriptionally active and inactive ase activity using the intact strand as a template (for DNA: In contrast to the rad52 mutation, the effects review, see MODRICH1987). The single-strand gap in of radl alone on recombination at GALlO are subtle. the DNA formed after the removal of the lesion may In transcriptionally active cells in the radl mutant, we be a target for a single-strand invasion, such as pro- see areproducible twofold decrease in the rate of posed in the Aviemore model of recombination (ME- plasmid loss when compared with RADl. The distri- SELSON and RADDINC1975). In yeast, RADl might be bution of the crossover events through the GALlO involved in the formation or processing of such an gene is unaffected by a mutation in rad 1,as is the rate intermediate, for example, by facilitating the strand of Gal+ Ura+ replacement. In transcriptionally inac- invasion process. A similar model for the roleof RADl tive cells, the rate of plasmid loss is not affected by a in geneticrecombination has been proposed by rad1 mutation. Similarly, the rate of Gal+ Ura+ re- SCHIESTLand PRAKASH (1 988). placement is no different from wild type in gal4 radl The RAD52 gene is thought to be involved in the strains. These results suggest that, in the absence of repair of double-strand breaks, probably by a recom- rad I function,recombination events proceed effi- binational-repair mechanism (Ho 1975; RESNICK ciently by arecombination pathway that probably 1976; RESNICKand MARTIN 1976). We have proposed involves RAD52. a model whereby the transcription-stimulated recom- The results of these experimentsare similar to other bination events in gal80 constitutive cells are initiated reportsexamining the role of excision-repair gene by double-strand breaks within the duplicated gall0 products in genetic recombination in yeast (SCHIESTL repeats or the integratedplasmid sequences (THOMAS and PRAKASH 1988;KLEIN 1988). In the experiments and ROTHSTEIN1989). These breaks couldbe induced reported by SCHIESTLand PRAKASH(1 988),duplica- by a specific recombination enzyme or alternatively tions containing a small region of homology, ca. 400 by a component of the transcription machinery, such bp, showed a four- tosevenfold reduction in the radl as a polymerase-associated type I1 topoisomerase. mutant; however, a much larger duplication contain- Since double-strand breaks in the rad52 mutant are ing his4 heteroalleles showed eitherno effect or lethal (RESNICKand MARTIN 1976; MALONEand Es- slightly higher recombination frequencies when com- POSITO 1980; WEIFFENBACHand HABER 1981), such pared to wild type. Similarly, KLEIN (1 988)saw little transcription-stimulated, double-strand break-in- effect of a radl point mutation on popout frequencies duced events would be lost in the constitutive gal80 between duplications of leu2 and his3 containing het- rad52 mutant. Transcriptional activation of a region eroalleles in the same linear arrangement as the his4 has been shown to be important for accessibility of repeats examined by SCHIESTLand PRAKASH(1 988). DNA repair enzymes in mammalian cells (BOHRet al. The RADI gene has been shown to be involved in 1985; MADHANI,BOHR and HANAWALT 1986;MEL- the excision repair of UV-induced damage (REYNOLDS LON et al. 1986). TranscribedDNA shows an increased and FRIEDBERG198 1;WILCOX and PRAKASH 1981). rate of removal of pyrimidine dimers compared to Based on physical analysis of irradiated DNA, RADl nontranscribed DNA or bulk chromatin. In addition, has been grouped with several genes thought to be MELLON,SPIVAK and HANAWALT (1987)have shown required for the incision step in the repair process. that UV-induced dimers are preferentially removed Recent work by SCHIESTLand PRAKASH(1 988) has from the transcribed strand in hamster cells. These shown that RADl is required for integrationof a linear results support the notion that transcription renders plasmid into the chromosome, suggesting that RADl the DNA more accessible to recombination and repair acts at astep after the initiating lesion is formed. functions in the cell. Furthermore, they show that the hyperrecombination The radlrad52 doublemutant displays amuch phenotype of cdc9, which containsa temperature- lower recombination rate in all the assays examined sensitive mutation in DNA ligase, is dependent on than that expectedif the two genes actedin a common Direct-Repeat Recombination 737 pathway. This implies that RADl and RAD52 act in GAME,J. C., and R. K. MORTIMER,1974 A genetic study of x-ray sensitive mutants in yeast. Mutat. Res. 24 281-292. two overlapping pathways for recombination in yeast. HABER,J. E., and M. HEARN, 1985 RAD52-independent mitotic Interestingly, this synergistic interaction of the two gene conversion in Saccharomyces cerevisiaefrequently results in mutations does not extend to survival after exposure chromosomal loss. Genetics 111: 7-22. to radiation; the double mutantwas no more sensitive HARTWELL,L. H., and D. SMITH,1985 Altered fidelity of mitotic to either gamma-or UV-irradiation than either single chromosome transmission in cell cyclemutants of Saccharomyces 4). cerevisiae. Genetics 110 381-395. mutant (Figure In fact, RADl and RAD52 were HAYNES,R. H., and B. A. KUNZ, 1981 DNA repair and mutagen- originally placed into separate groups on the basis of esis in yeast, pp. 371-414 in The Molecular Biology of the Yeast their Rad phenotypes.This result suggests that notall Saccharomyces:Life Cycle and Inheritance, edited by J. N. the lesions induced by exposure to radiation are proc- STRATHERN,E. W. JONES and J. R.BROACH. Cold Spring essed by a recombination-repair type pathway. We are Harbor Laboratory, Cold Spring Harbor, N.Y. currently testing the hypothesis that the residual re- HICKS,J. B., A. HINNEN andG. R. FINK,1979 Properties of yeast transformation. Cold Spring Harbor Symp. Quant. Biol. 43: combination seen in the radl rad52 double mutant is 1305-1313. mediated by yet another RAD epistasis group, such as Ho, K. S. Y., 1975 Induction of DNA double strand breaks by x- the RAD18 group. The genes in this group define an rays in a radio-sensitive strain of the yeast Saccharomyces cerevis- “error-prone” pathway for response to DNA damage iae. Mutat. Res. 30 327-334. [reviewed by KUNZ and HAYNES(1 98 1) and HAYNES HOEKSTRA,M. F., T. NAUGHTON and E.R. MALONE, 1986 Properties of spontaneous mitotic recombination occur- and KUNZ(1 98 l)]. Finally, it is interesting to note that ring in the presence of the rad52-1 mutation of Saccharomyces the residual recombination in the radl rad52 double cerevisiae. Genet. Res. 48: 9-17. mutant is still affected positively by transcription. HOLLINGSWORTH,N. M., and B. BYERS,1989 HOPI: a yeast meiotic pairing gene. Genetics 121: 445-462. We would like to thank HAMISHYOUNG and DAVIDSHORE for ITO, H., FUKADA,K. MURATA and A. KIMURA, 1983 Trans- critical reading of this manuscript, and RALPHKEIL, ELAINE SUG- formation of intact yeast cells treated with alkali cations. J. ARMAN and JIM HABERfor communicating unpublished observa- Bacteriol. 153: 163-168. tions. We also thank DAVID SCHILD the for rad52 disruption muta- JACKSON, J. A., and G. R. FINK,1981 Gene conversion between tion and JURG OTT for help with the statistical analyses. B. J. T. duplicated genetic elements in yeast. Nature 292: 306-31 1. was supported by grant CA09503 from the National Institutes of JINKS-ROBERTSON,S., and T. D. PETES, 1986 Chromosomal trans- Health, and a grant from the MacArthur Foundation. R. R. is an locations generated by high frequency meiotic recombination American Heart Association Established Investigator. This work between repeated yeast genes. Genetics 114 731-752. was also supported by National Institutes of Health grant GM34587 JOHNSTON,S. A., and J. E. HOPPER,1982 Isolation of the yeast and a grant from the Irma T. Hirschl Trust. regulatory gene GAL4 and analysis of its dosage effects on the galactose/melibiose regulon. Proc. Natl. Acad.Sci. USA 79 LITERATURECITED 6971-6975. KEIL,R. L., and G. S. ROEDER,1984 Cis-acting recombination- BOEKE,J. D.,F. LACROUTEand G. R. FINK,1984 A positive stimulating activity in a fragment of the ribosomal DNA of S. selection for mutants lacking orotidine-5”phosphate decarbox- cerevisiae. Cell 39 377-386. ylase activity in yeast: 5-fluoro-orotic acid resistance. Mol. Gen. KLEIN,H. L., 1988 Different types of recombination events are Genet. 197: 345-346. controlled by the RADl and RAD52 genes of Saccharomyces BOHR,V. A,, C. A. SMITH,D. S. OKUMOTOand P. C. HANAWALT, cermisiae. Genetics 120: 367-377. 1985 DNA repair in an active gene: removal of pyrimidine KUNZ,B. A., and R. H.HAYNES, 1981 Phenomenology and ge- dimers from the DHFR gene of CHO cells is much more netic control of mitotic recombination inyeast. Annu. Rev. efficient than in the genome overall. Cell 40 359-369. Genet. 15: 57-89. DOUGLAS,H. C., and D. C. HAWTHORNE,1966 Regulation of LEA, D. E., and C. A. COULSON,1949 The distribution of the genes controlling synthesis of the galactose pathway enzymes numbers of mutants in bacterial populations. J. Genetics 49 in yeast. Genetics 5491 1-916. 264-285. DUTCHER,S. K., and L. H. HARTWELL,1982 The role of S. LICHTEN,M., R. H. BORTSand J. E. HABER, 1987 Meiotic gene cerevisiae cell division cycle genes in nuclear fusion. Genetics conversion and crossing over between dispersed homologous 100 175-184. sequences occurs frequently in Saccharomycescereoisiae. Ge- ENGEBRECHT,J., and G. S. ROEDER,1989 Yeastmer1 mutants netics 115 233-246. display reduced levels of meiotic recombination. Genetics 121: LURIA,S. E., and M. DELBRUCK,1943 Mutations of bacteria from 237-247. virus sensitivity to virus resistance. Genetics 28: 491-51 1. ESPOSITO, R. E., and S. KLAPHOLZ,1981 Meiosis and ascospore MADHANI, H. D., V. A. BOHR and P.C. HANAWALT,1986 development, pp. 21 1-287 in The Molecular Biology of the Yeast Differential DNA repair in transcriptionally active and inactive Saccharomyces: Life Cycle and Inheritance, edited by J. N. Strath- proto-oncogenes: c-ab1 and c-mos. Cell 45 417-423. ern, E.W. Jones andJ. R. BROACH.Cold Spring Harbor MALONE,R. E., 1983 Multiple mutant analysis of recombination Laboratory, Cold Spring Harbor, N.Y. in yeast. Mol. Gen. Genet. 189 405-412. ESNSITO, M. S., J. HOSODA,J. GOLIN, H. MOISE,K. BJORNSTAD MALONE, R., and R. E. ESPOSITO,1980 The RAD52 gene is and D. MALEAS, 1984 Recombination inSaccharomyces cere- required for homothallic interconversion of mating types and visiae: REG-gene mutants and DNA-binding proteins. Cold spontaneous mitotic recombination in yeast. Proc. Natl. Acad. Spring Harbor Symp. Quant. Biol. 49: 41-48. Sci. USA 77: 503-507. FASULLO,M. T., and R.W. DAVIS,1987 Recombinational sub- MELLON, I., G. SPIVAKand P. C. HANAWALT,1987 Selective strates designed to study recombination between unique and removal of transcription-blocking DNA damage from the tran- repetitive sequences invivo. Proc. Natl. Acad. Sci.USA 84: scribed strand of the mammalian DHFR gene. Cell 51: 241- 62 15-62 19. 249. 738 B.and J. Thomas R. Rothstein

MELLON, I., V. A. BOHR,C. A. SMITHand P.C. HANAWALT, ROTHSTEIN,R., 1985 Cloning in yeast, pp. 45-66 in DNA Cloning, 1986 Preferential DNA repair of an active gene in human Vol. 11: APractical Approach, edited byD. M. GLOVER. IRL cells. Proc. Natl. Acad. Sci. USA 83: 8878-8882. Press, Washington, D.C. MESELSON, M. S., and C. M. RADDING,1975 A general model for ROTHSTEIN, R.J., C. HELMSand N. ROSENBERG,1987 Concerted genetic recombination. Proc. Natl. Acad. Sci.USA 72: 358- deletions and inversions are caused by mitotic recombination 361. between delta sequences in Saccharomyces cerevisiae. Mol. MIKUS,M. D., and T. D. PETES,1982 Recombination between CelLBiol. 7: 1198-1207. genes located on nonhomologous chromosomes in Saccharo- RUDIN,N., and J. E. HABER,1988 A novel HO-endonuclease- myces cerevisiae. Genetics 101: 369-404. initiated recombination event between repeated sequences. MODRICH, P., 1987 DNAmismatch correction.Annu. Rev. Mol. Cell. Biol. 8: 3918-3928. Biochem. 56: 435-466. SCHIESTL,R. H., and S. PRAKASH,1988 RADI, an excision repair MORPURGO,G., G. SERLUPI-CRESCENZI,G. TECCE, F. VALENTEand gene of Saccharomyces cerevisiae, is also involved in recombina- D. VENTTACCI, 1964Influence of ergosterol on the physiol- tion. Mol. Cell. Biol. 8: 3619-3626. ogy and the ultra-structure of Saccharomyces cerevisiae. Nature SCHIESTL,R. H., S. IGARASHIand P. J. HASTINGS,1988 Analysis 201: 897-899. of the mechanism for reversion of a disrupted gene. Genetics MORTIMER,R. K., R. CoNTOPoULOUand D. SCHILD,1981 Mitotic 119: 237-247. chromosome loss in a radiation-sensitive strain of the yeast SCHILD,D., B. KONFORTI, C. PEREZ,W. GISHand R. K. MORTIMER, Saccharomyces cerevisiae. Proc. Natl. Acad. Sci. USA 78 5778- 1983a Isolation and characterization of yeastDNA repair 5782. genes. 1. Cloning of the RAD52 gene. Curr. Genet. 7: 85-92. NICKOLOFF,J. A., J. D. SINGER,M. F. HOEKSTRAand F. HEFFRON SCHILD,D., I. L. CALDERON,R. CONTOPOULOUand R. K. MORTI- 1989 Double-strand breaks stimulate alternative mechanisms MER, 1983b Cloning of yeast recombination repair genes and of recombination repair. J. Mol. Biol. 207: 527-541. evidence that several are nonessential genes, pp. 417-427 in ORR-WEAVER,T., J. W. SZOSTAKand R. J. ROTHSTEIN, Cellular Responses to DNA Damage, edited by E. C. FRIEDBERG 1981 Yeast transformation: a model system for the study of and B. A. BRIDGES. AlanR. Liss, N.Y. recombination. Proc. Natl. Acad. Sci. USA 78: 6354-6358. SIEGEL,S., 1956 Nonparametric Statisticsfor the Behavioral Sciences, OSHIMA,Y., 1982 Regulatory circuits forgene expression: the pp. 95-1 15. McGraw-Hill, N.Y. metabolism of galactose and phosphate, pp. 159-180 in The SMITH,G. R., 1987 Mechanism and control of homologous recom- Molecular Biology of the Yeast Saccharomyces: Metabolism and Gene bination in Escherichia coli. Annu. Rev. Genet. 21: 179-201. Expression, edited by J. N. STRATHERN,E. W. JONES and J. R. SOUTHERN,E. M., 1975 Detection of specific sequences among BROACH.Cold Spring Harbor Laboratory, Cold Spring Harbor, DNA fragments separated by gel electrophoresis. J. Mol. Biol. N.Y. 98: 503-517. RESNICK,M. A., 1969 Genetic control of radiation sensitivity in STARK,G. R., and G. M. WAHL, 1984 . Annu. Saccharomyces cerevisiae. Genetics 62: 5 19-53 1. Rev. Biochem. 53: 447-491. RESNICK,M. A,, 1976 The repair of double-strand breaks in DNA: THOMAS,B. J., and R. ROTHSTEIN,1989 Elevated recombination a model involving recombination. J. Theor. Biol. 59: 97-106. rates in transcriptionally active DNA. Cell 56: 619-630. RESNICK,M. A., and P. MARTIN,1976 The repair of double- TORCHIA,T. E., R. W. HAMILTON,C. L. CANOand J. E. HOPPER, strand breaks in the nuclear DNA of Saccharomyces cerevisiae 1984 Disruption of regulatory gene GAL80 in Saccharomyces and its genetic control. Mol. Gen. Genet. 143: 119-129. cerevisiae: effects on carboncontrolled regulation of the galac- REYNOLDS,R. J., and E. C. FRIEDBERG,1981 Molecular mecha- tose/melibiose pathway genes. Mol. Cell. Biol. 4: 1521-1527. nisms of pyrimidine dimer excision in Saccharomyces cerevisiae: WEIFFENBACH,B., and J. E. HABER,1981 Homothallic mating incision of ultraviolet-irradiated deoxyribonucleic acid in vivo. type switching generates lethal chromosome breaks in rad52 J. Bacteriol. 146: 692-704. strains of Saccharomyces cereuisiae. Mol. Cell. Biol. 1: 522-534. ROEDER,G. S., R. L. KEIL and K. A. VOELKEL-MEIMAN,1986 A WILCOX,D. R., and L. PRAKASH,1981 Incision and postincision recombination-stimulating sequence in the ribosomal RNA steps of pyrimidine dimer removal in excisiondefective mu- of yeast, pp. 29-33 in Current Communications in tants of Saccharomyces cereuisiae. J. Bacteriol. 148: 618-623. WILLIS,K. K., and H. L. KLEIN,1987 Intrachromosomal recom- Molecular Biology: Mechanisms of Yeast Recombination, edited by A. KLAR and J. STRATHERN.Cold Spring Harbor Laboratory, bination in Saccharomyces cerevisiae: reciprocal exchange in an and associated gene conversion. Genetics 117: Cold Spring Harbor, N.Y. 633-643. RONNE,H., and R. ROTHSTEIN,1988 Mitotic sectored colonies: YANG,E., and E.C. FRIEDBERG,1984 Molecular cloning and evidence of heteroduplex DNA formation during direct repeat nucleotide sequence analysis of the Saccharomyces cereuisiae recombination. Proc. Natl. Acad. Sci. USA 85: 2696-2700. RAD1 gene. Mol. Cell. Biol. 4 2161-2169. ROTHSTEIN,R., 1983 One-step gene disruption in yeast. Methods Enzymol. 101: 202-211. Communicating editor: G. S. ROEDER