The majority of recombination events are mediated by a RAD51-dependent pathway ... Furthermore, there is still considerably more recombination in rad1 rad51 ...
Copyright 0 1995 by the Genetics Society of America
Multiple Pathways for Homologous Recombination in Saccharomyces cereuiSiae Alison J. Rattray and Lorraine S. Symington Department of Microbiology and Institutefor Cancer Research, Columbia University College of Physicians and Surgeons,New York, New York 10032 Manuscript received July 11, 1994 Accepted for publication September 12, 1994 ABSTRACT Thegenesinthe RAD52epistasis group ofSaccharomycescevevisiae are necessaryformostmitotic andmeioticrecombinationevents.Usinganintrachromosomalinverted-repeatassay, we previously demonstrated that mitotic recombination of this substrate is dependent upon the RAD52 gene. In the present study the requirement for other genes in this epistasis group for recombination of inverted repeats hasbeenanalyzed,anddoubleandtriplemutantstrainswereexaminedfortheirepistatic relationships. Themajority of recombination events are mediated by a RAD51-dependent pathway, where the RAD54, RAD55 and RAD57genes function downstream ofM 5 1 . Cells mutated in RAD55 or RAD57 aswell as double mutants are cold-sensitive for inverted-repeat recombination, whereas a rad51 rad55 rad57 triple mutant is not. The RAD1 gene is not required for inverted-repeat recombination but is able to process spontaneous DNA lesions to produce recombinant productsin the absence of RAD51. Furthermore, thereis still considerably more recombinationin rad1 rad51mutants than in rad52mutants, indicating the presenceof another, as yet unidentified, recombination pathway.
S
TUDIESof meiotic recombination in ascomycete fungi, where all ofthe productsof a recombination event can be recovered, have showna strongassociation between gene conversion and crossing over (see PETES et al. 1991 for a review). These observations have been accounted for in models for recombination in which heteroduplex DNA is proposed as an intermediate in the formation of crossover products (HOLLIDAY 1964; MESELSONand RADDINC 1975; RADDINC 1982; SZOSTAK et al. 1983). The models predict that gene conversion, the result of heteroduplex repair, is associated with crossing over 50% of the time. However, the isolation of mutations that differentially affect conversion and et al. 1990;ROCKcrossing over in meiosis (ENGEBRECHT MILL and ROEDER 1990) suggests that there might be separate pathways for each type of event. The concurrence of gene conversions and crossovers is difficult to establish in mitotic cells because usually only one product of a recombination eventis recovered. Instead, it is generally inferred from the relative proportions of each type of event among the total recombinant progeny. ROMANand RUZINSKI (1990) found evidence that the majority of gene conversion events occur in the G1 phase of the cell cycle, whereas the majority of crossovers occurred during G2. This temporal separation is suggestive of separate pathways for gene conversions and crossovers. Although all crossovers may require a prior gene conversion event (or heteroduplex precursor), there is evidence thatshortergene conversion tracts are preferentially noncrossover, and longer conCmesponding author: Lorraine S. Symington, Institute of Cancer Research, Columbia University College of Physicians and Surgeons, 701 W. 168th St., Room 912, New York, NY 10032. Genetics 139: 45-56 (Janualy, 1995)
version tracts show a greater association with crossovers (AHN and LMNCSTON 1986; AGUILERA and KLEIN 1989). Therefore, there may be a differentset or subset of gene products that dictate preferentially noncrossover recombination products in G1. Although recombination has been well characterized in Escherichia coli, few studies have examined theassociation between gene conversion and crossing over because in most assays both products of a recombination event are not recovered. To examine theissue of reciprocality of recombination events, SECALLand R o T H (1994) developed a chromosomalinverted-repeated system in Salmonella typhimunum. In this system four classes of recombinant products were recovered: inversions, inversions associated with conversion, apparent gene conversions and double crossovers. In recA mutants all classes of recombination events were reduced by lo4fold. In recB mutants the inversion class was eliminated, and the other classes reduced 2- to 60-fold. These data provide evidence that gene conversions and crossovers can proceed through different pathways in bacteria. The concept of recombination pathways was developed from the genetic analysis of conjugative recombination in E. coli (reviewed inSMITH1989). The RecBCD pathway is the primary pathway for conjugative recombination;the RecE and RecFpathways are activated by suppressor mutations in the absence of recB or recC function. All of the recombination pathways, RecBCD, RecE and RecF, requirethe recA geneproduct,and most of the genes required for theRecE pathway, with the exception of re&, are required for the RecF pathway. Thus there is considerable overlap between these pathways. Although several genes have been identified
A.
46
J. Rattray and L. S. Symington
that are required for recombination in Saccharomyces cerevisiae, it has not been clearly demonstrated whether these function in a common pathway. The major group of genes involved in mitotic recombination is the RAD52 epistasis group (GAME1983, 1993). Thesegenes are characterized by their requirement for repair of ionizing radiation-induced DNA damage (GAMEand MORTIMER1974). Studies of mitotic recombination between directly repeated sequences have identified two alternate pathways that require the W 5 2 and RAD1 geneproducts, respectively, indicating thatalternate pathways for recombination exist in yeast (KLEIN 1988; SCHIESTL and PRAKASH1988; THOMAS and ROTHSTEIN 1989; ZEHFUS et al. 1990). The RAD1 gene is required for excision repair of UV-induced DNA damage, but mutants have no otherknown defects in mitotic or meiotic recombination. We recently developed a chromosomal inverted-repeat assay for studying mitotic recombination in haploid yeast cells.Three classes ofrecombination products were recovered: gene conversions, crossovers (inversions) and crossovers associated with gene conversion (Figure 1). In rad52 mutants all classes of events were greatly reduced (>5000-fold), whereas a rad51 mutation reduced gene conversion events 18-fold and crossovers (both with or without an associated conversion) only 2.5-fold (RATTRAY and SYMINGTON 1994). Analysis of recombination in these two mutants was of particular interest because recent studies have shown a physical association between the W 5 1 and RAD52 gene products (SHINOHARA et al. 1992; MILNEand WEAVER 1993) and because Rad51 has substantial homology to the E. coli RecA protein (ABOUSSEKHRA et al. 1992; BASILEet al. 1992; SHINOHARA et al. 1992). The four-fold reduction in the rate of recombination between inverted repeats in rad51 mutants sharply contrasts with the lo4fold reduction observed in S. typhimurium recA mutants. Also the different phenotypes of rad51 and rad52 mutants in this assay suggest that Rad52 has other functions in addition to interaction with Rad51. In the current study this genetic analysis has been extended to include other members of the RAD52 epistasis group, as well as rad1 mutants. In addition, double and triple mutant strains have been examined to determine their epistatic relationships. Our dataprovide evidence for theexistence of at least three distinct RAD52dependent recombination pathways in yeast. The major pathway requiresthe RAD51, RAD54, RAD55 and RAD57 genes, a M I - d e p e n d e n t pathway and at least one other pathway. Furthermore, we show that mutation of genes in the RRD51-dependent pathway results in a significant decrease in the classof events corresponding to gene conversion events unassociated with crossing over. MATERIALSAND METHODS Media, growth conditions and genetic methods: Rich medium (YEPD), synthetic complete (SC) medium lacking the
appropriate aminoacid or nucleic acid base, and 5-fluoroorotic acid (5-FOA) medium were prepared as described (SHERMAN et al. 1986). Cells were grown at 30" unless othemise indicated.Transformations were performed by theLAC method (IT0 et ai. 1983) and tetrad dissection as described previously (SHERMAN et al. 1986). The y-radiation sensitivity of strains was scored by replica plating test strains to YEPD medium and irradiating with 50 krad from a Gammacell-220 irradiator containing "Co (Atomic Energy of Canada, Ottawa, Canada). The UV sensitivity of strains was scored by replica plating test strains to YEPD medium and irradiating with 50 J/m' at 254 nm. For survival curves, cells were grown in liquid YEPD medium to early log phase and dilutions of cells were plated onto solid YEPD and irradiated at various doses. Plasmids: A list of plasmids and their relevant sources is shown in Table 1. Construction of the plasmid used for integrating the inverted-repeat substrate has been described previously (RATTRAY and SYMINGTON 1994). Briefly, one repeat (ade2-n) consists of a 3.1-kilobase (kb) BgtII-SpeIfragment of the wild-type ADE2 gene that has been mutagenized by fill-in of the uniqueNdeI site at codon465 with the Klenow fragment of DNA polymerase I. The second repeat (ade2-5'n) is a 1.8kb EcoRV fragment of ADE2 that lacks the promoter and the first 225 bp of the ADE2 open reading frame. The two copies are separated by a 0.85-kb fragment containingthe TRPl gene. Yeast strains: A list of yeast strains andtheir relevant sources is shown in Table2. All strains are derivatives of strains W3031A or W303-1B (THOMAS and ROTHSTEIN 1989). Strains yAR71 and yAR91 contain the nde2-SA-TRPlade2-n construct at the native H I S 3 locus (Figure ZA). Strains yAR82, yAR86, yAR107,yARlO8 and yARl39 were constructed by one-step transplacement (ROTHSTEIN 1983) of yAR71 o r y m 9 1 (as indicated in Table 2) with the appropriate plasmid DNA fragments to generate disruption alleles of the indicated RAD genes. To construct strain yAR167 the URA3 gene was first excised from ade2::hisGURA3-hisG of strain yAR160 by selection on 5-FOA medium (WINSTON et al. 1983) and then transformed to Ura+ with an EcoRI fragment frompNKY83. Strains yAR188 and yARl90 were constructed by transformation of yAR97 to Leu+ with an Xbal-PstI fragment ofpAM28 or a Hind111 fragment of pL962, respectively. All other strains were constructed by mating appropriate strains (as indicated in Table 2) and selecting forhaploidspores segregating the appropriate rad mutation(s),the ade2-5'A-TRPI-ade2-n inverted-repeat substrate and the ade2::URM disruption at the native ADE2 locus. The presence of the rad mutations was monitored by sensitivity to ionizing radiation (or UV irradiation for radl), and all strain constructions were further verified by Southern blot analysis. We noted thata diploidheterczygous for rad51 rad55 and rad57mutationssporulated poorly at 23" (2-3%) and somewhat more efficiently at 30" (3040%), although the asci were extremely fragile, requiring no glusulase pretreatment before dissection. Single and double heterozygotes for any of these mutations sporulated with wildtype efficiency (>70%). Determination of mitoticrecombmationfrequencies: At least two independent isolates of each strain were used for the determination of recombination rates. Single colonies Of each isolate were grown on YEPD for 2-3 days. Growth for longer than 4 days gave artificially high recombination rates because Ade+ cells continue to divide for longer than Adecells. At least 27 pink colonies of each strain (usually 9 from each independent isolate) were resuspendedin water and plated at the appropriate dilutions to determine total cell number (SC-Trp medium) and the number of Ade+ prototrophs (SCPAde medium). Median mitotic recombination frequencies (A&+ cells/Trp' cells) were determined and rates
Homologous Recombination Pathways in S. cermisiae
47
TABLE 1 Description and sources of disruption plasmids
Name
Description
pL962 pNKY83 pAM28 pSM31 pSTLl1 pSM51 PEI139
rad1::LEUZ disruption plasmid rad50::hisGURA3-hisC disruption plasmid rad51::LEU2 disruption plasmid rad54::LEU2 disruption plasmid rad55::LEUZ disruption plasmid rad57::LEUZ disruption plasmid xrsP::URA? disruption plasmid
(events/cell/generation) werecalculatedaccording to the following formula:rate = (0.4343 X median frequency)/ (log N - log No),where N is the number of Trp+cells present in the colony and No (number of initial cells) = 1 (DRAKE 1970). Neitherthe mating type nor the SPOI? allele affected recombination rates, therefore datafrom isolates differing in either of these genotypes were pooled. Characterization of recombinants: A single Ade+ recombinant was picked from each original colony to ensure analysis of independent events; DNA was isolated (HOFFMAN and WINSTON 1987),digested with NdeI or PstI and analyzed by Southern hybridization to determine the natureof the recombination event.Blots of DNA digested withNdeI were probed with a radioactively labeled fragmentof the TRPl gene, and blots of DNA digested with PstIwere probed with a fragment of the ADE2 gene. The expected fragments from this analysis in strainstransformedwithpAL90-1 are shown in Figure 2B. Statisticalanalysiswas done by using a contingency chisquaredanalysisondataassembledintotables. A P value 0.1). We also examined a rad50 rad51 double mutantstrain and found that the rateof Ade' prototroph formation was similar to a rod51 single mutant (Table 4). The distribution of recombination event5 was not significantly different from a rad51 mutant ( P > 0.1) but did differ from both wild-type and rnd50strains ( P < 0.05). These data suggest that RAD50 and RAD51 function in the same pathway. The W Z gene identifies a second recombination pathway for inverted-repeat recombination:The KAI)1 gene is involved in excision repair of UV-induced DNA damage (see GAME1983). The Rad1 protein forms a complex with the Rad10 protein that specifies a singlestrand (ss) DNA endonuclease (TOMKINSONet nl. 1993). Although the RAD1 gene is not required for spontaneous mitotic heteroallelic recombination, rad1 mutants have been shown to have reduced levels of deletion
Homologous Recornhination I'athwaysin
S. c w ~ ~ i s i r w
TABLE 3 Effect of md51, md55, md57 and rad54 mutations on inverted-repeat recombination
300 26 8 23 20 4 7 76 55 5 14
(24/48) (3/24) (6/24) (3/24) (9/23)E (5/23) (4/24) (8/24) (5/24) (4/24) (2/24)
330 ( 17/48) 140 ( I .5/24) 16 ( l 2 / 2 4 ) 1 I O (l4/24) 18 ( 8 / 2 3 ) I O ( 13/23) 1 1 (6/24) 6 5 (7/24) 63 (8/24) 23 ( 13/24) 95 ( 13/24)
150 (5/48) 3.5 (6/24) 8 (6/24)
.37 (7/24) 13 (6/23) 4 (.5/23) 24 ( 14/24) 87 (9/24) 52 (9/24) 12 (5/24) 68 (9/24)
" Rates arc evcnts/cell/gcner;tion; cxperimcnts were performed a t 30" unlcss othcnvisc st;~tctl;nrml)crs i n parentheses indicate number o f et'ents in class/total events examined. ' Gene conversions associated with a reciprocal crossover. ' Data from three independent experiments; median value o f 81 colonies examined. 'I Data taken from RATI'KAY and SIIiIw;rox (1994). 'Distrihrttion o f events is signilicantly different from wild-type strain ( P < 0.0.3). Distribution of events is not significantly different from md51 strain ( P > 0.1 ). I' One of 24 events examined showed a complex restriction pattern.
events between sequences oriented as directrepeats (KLEIN 1988; SCHIESTL. and PRAKASH1988). We found that disruption of theRADl gene had no effect on the rate of Ade' prototroph formation (Table 4).When 24 Ade' recombinants were examined by Southern blot analysis, the distribution of events was found not to differ significantly from that seen in a wild-type strain ( P > 0.05). We next examined theepistatic relationship of the RAD1 and RA1151 genes by determining the rate of recombination in a rad1 rad51 double mutant. The rate of Ade' prototroph formation was reduced 17-fold (Table 4).Southern blot analysis of Ade' recombinants indicated a synergistic decrease in the rateof crossovers unassociated with a conversion (Table4,class 11), which
is significantly different from arlrd51 mutant strain ( P < 0.025). These data indicate that somelesions normally processed via a IG11151dependent path\vay arc r e p a i d via a RAI11-dependent pathway. Tofurther investigate the relationships betwecn R A D I , RA1)50 and IUI151, double-andtriple-mutant strains were constructed. The rate of Ade' prototroph formation in a rad1 rd50douhlc mutant was intermecliate between either single mutant (Table 4). Southern analysis of recombinants also indicated an intcrmrtliatc phenotype. However, the rate of recombination w a s rcduced 50-fold in the triple mutant,indicating a multiplicative decrease over the r d l md51 double mutant. The distribution of recombination events i n thc triple mutant was significantlv different from that seen in rudl, rad50 and rad1 rctd50 strains (I' < 0.0.5) but was not different from a rd51 mutant (!'> 0.1).
52
A. J. Rattray and L. S. Symington
TABLE 4 Effect of rad50, xrs2 and radl mutations on inverted-repeat recombination
Rate of Ade+ products ( X lo7) Genotype
500
RAD rad5G" xrsT rad51 rad50 rad51"' rad 1 radl rad51".' radl rad50' radl rad50 rad51' 'I
Total Ade+ ( X 106)
Class I conversions
Class I1 crossovers
Class I11
100 37 19 22 20 110 6 49 2
(24/48) 162 (21/48) '72 (9/24) 26 (3/24) 34 (4/24) 412 (9/24) 12 (5/24) 264 (13/24) 8 (9/22)"
350 (17/48) 162 (21/48) 87 (11/24) 140 (15/24) 92 ( 1 1/24) 275 (6/24) 10 (4/24) 123 (6/24) 6 (7/22)
150 (7/48) 46 (6/48) 30 (4/24) 55 (6/24) 74 (9/24) 412 (9/24) 38 (15/24) 103 (5/24) 5 (6/22)
both
Data presented as in Table 3. Distribution of events is not significantly different fromwild-type ( P > 0.1). "Distribution of events is not significantly different fromrad51 ( P > 0.1). ' Distribution of events is significantly different from wild-type ( P < 0.05). "Two of 24 events examined had a complex restriction pattern. I'
DISCUSSION
A chromosomal inverted-repeat substrate has been used to investigate the genetic control of homologous recombination in S. cereuisiae. The data presented here provide evidence for the existence of multiple pathways for spontaneous mitotic recombination. This conclusion is based on thefollowing observations. (1) Whereas rad52mutants are reduced3000-fold for inverted repeat recombination, rad51 mutants are only reduced 5-fold (RATTRAYand SYMINGTON 1994). (2) Single mutations in m 5 4 , RAD55 or RADS7 result in a 20- to 30-fold decrease in recombination but reduce recombination only !?-foldin combinationwith a rad51 mutation. These data indicate that all four of these genes function in the same recombination pathway, that RAD51 functions before the other three and that an alternative pathway must function in the absence of RAD51. (3) rad55 and rad57 mutants, aswellas double mutants, are cs for inverted-repeat recombination. (4) Disruption of all three mitotic RecA homologues results in a recombination phenotype indistinguishable from a rad51 single mutant, indicating a lackof functionalredundancy. The triple mutant is not cs for recombination, further supporting an early role for Rad51 in recombination. (5) Although disruption of the R A D 1 gene does not affect inverted-repeat recombination, rad1 rad51 double mutants show a synergistic decrease indicating that some of the M 5 l - i n d e p e n d e n t recombination is processed via a RADldependent pathway. (6) Whereas rad50 and xrs2 mutants show elevated rates of heteroallelic recombination, these mutants have reduced levels of inverted-repeat recombination. The genes in the RAD52 epistasis group were initially identified by the sensitivity ofmutants to ionizing radiation (GAMEand MORTIMER1974). rad52 mutants are the most sensitive to ionizing radiation and areepistatic
to all other genes in this group (MCKEEand LAWRENCE 1980). Although all of the mutants aresensitive to ionizing radiation, considerable heterogeneity exists between members of this group when other phenotypes related to recombinational repair are examined. For example, one expectation for a mutant defective in recombinationalrepair is that haploids and diploids should be equally sensitive to ionizing radiation. rad51, rad52 and rad54 mutants share this property, but xrs2, rad50, rad55 and rad57 mutants all show greater resistance to ionizing radiation in diploids than in haploids (SAEKI et al. 1980; LOVETTand MORTIMER1987;ABOUS SEKHRA et al. 1992; IVANOV et al. 1992).Although this is referred to as diploid-specificrepair, in thecase of rad55 mutants it has been shown to be because of mating type heterozygosity rather than ploidy (LOVETTand MORTIMER 1987). The assay commonly used to measure mitotic recombination is the rate of prototroph formation between heteroalleles of anauxotrophic marker in diploids. These events occur primarily by gene conversion unassociated with crossing over (ROMAN1957; KAKAR1963; HURSTand FOGEL1964; GOLINand ESPOSITO1981). rad51 and rad52 mutants show the greatest defects in spontaneous and induced heteroallelic recombination, whereas rad54, rad55 and rad57mutants show a modest reduction with a substantial induction by y-rays. rad50 and xrs2 mutants share the property of elevated rates of spontaneous heteroallelic recombination. Given that rad50, xrs2, rad55 and rad57 mutants show diploid specific repair, it is possible that recombination in these mutants would be reduced furtherin haploids. Recombination can be measured in haploid strains that contain gene duplications. These are generally arranged as intrachromosomal direct repeats, intrachromosomal inverted repeats or onheterologous chromosomes. Mu-
Homologous Recombination Pathways tation of RAD52 yields a recombination-defective phenotype in all of these assays (JACKSON and FINK1981; KLEIN 1988; SCHIESTL and PRAKASH 1988;AGUILERA and KLEIN 1989;BAILISand ROTHSTEIN 1991; DORNFELD and LMNGSTON1992), but the effects of other genes in the RAD52 epistasis group have previously only been analyzed fordirect-repeatrecombination. Deletion events between direct repeats (pop-outs) are increased in rad51, rad54, ,rad55 and rad57 mutants and occur at wild-type frequency in rad50 mutants (MCDONALD and ROTHSTEIN 1994; H. KLEIN, personal communication). Thus, the genes in the RAD52 group can be further subdivided based on their effects on direct repeat recombination. Our results using a chromosomal inverted-repeat recombination assayshow a similar heterogeneity between the rad mutants. rad52 mutants show >3000-fold decrease in the rate of Ade' prototrophs, and all classes of recombinant products are reduced equally (RATTRAY and SYMINGTON 1994). Mutation of RAD54, RAD55 or RAD57 results in a 20- to 30-fold decrease in the rate of recombination, whereas rad51 mutants show only a 5-fold reduction (Tables 3 and 4). In all of these singly mutated strains, there is a significant decrease in gene conversion events unassociated with a crossover, compared withwild-type strains, andthe distribution of events is not significantly different between all four mutant strains. rad50 and xrs2 mutants show a three- to five-fold reduction in recombination, and the distribution of events is the same as that observed in the wildtype strain. Therefore, using the inverted-repeat assay, the mutants can be divided into three groups. rad52 mutants occupy a uniqueposition in showing the greatest reduction in recombination; rad51, rad54, rad55and rad57 mutants share the property of reduced recombination, specifically of gene conversion events unassociated with crossing over, and rad50 and xrs2 mutants show a modest reduction in all classes ofrecombination events. Although rad51 mutants show a less severe recombination defect than rad54, rad55 and rad57, they are grouped together based on the phenotype of double and triple mutants. We foundthat rad51 rad54, rad51 rad57 double mutants and rad51 rad55 rad57 triple mutants have a phenotype indistinguishable from a rad51 single-mutant strain. The simplest interpretation of these data is that the RADS1 gene product functions before the RAD54, RAD55 or RALl57 gene products, that one (or more) alternative pathway(s) can be used in the absenceof RAD51 and that once recombination intermediates have beenchannelledintothe RADS1 pathway they can no longer be acted upon by an alternative pathway. In this scenario the alternative pathway(s) must normally process only -5% of the lesions in a wild-type cell, based upon the 20- to 30-fold reduction in recombination by mutants in the downstream genes, butcan process up to 20% of the lesions in a rad51 mutant. Although mutations in all four of
in S. cerevisiae
53
these genesreducegene conversions not associated with a crossover more drastically than crossovers (either with or without an associated conversion), crossovers are also reduced. Further support for an earlier role in recombination for RAD51 comes from analysis of the cs recombination phenotype of rad55 and rad57 mutants. Mutation in either of these genes have previously been shown to result in a cs repair phenotype (GAME and MORTIMER1974; GAME1983,1993; LOVE= and MORTIMER 1987), and we show here that they are also cs for inverted-repeat recombination. We also find that there is no furtherreduction in the double mutant, indicating that it is likely that these two proteins function at the same step in recombination. However, we find that the cs phenotype is largely suppressed in the presence of a rad51 mutation (Figure 4). The modest reduction in recombination observed in rad51 mutants raised two questions. First, is the intrachromosomal inverted-repeat assay representative of recombination between homologues, and second, is there functional redundancy between RecA homologues? Althoughrecombination between heteroalleles occurs primarily by gene conversion unassociated with crossing over, 50% of therecombination events between inverted repeats occur by crossing over. Although mutation of RAD51 results in only a 5-fold reduction in recombination, gene conversion unassociated with crossing over is reduced 18-fold. Spontaneous heteroallelic recombination has been shown to be reduced -20fold in rad51 mutants (SAEKIet al. 1980; ABOUSSEKHRA et al. 1992; SHINOHARA et al. 1992). If only the class of events corresponding to gene conversions is considered, the effects of a RAD51 mutation are quite similar in these two assays. Four yeast genes have been shown to encode homologues of the E. coli RecA protein, RAD51 (ABOUS SEKHRA et al. 1992; BASILEet al. 1992; SHINOHARA et al. 1992), DMCl (BISHOP et al. 1992), RAD55 (LOVETT 1994) and RAD57 (KANS and MORTIMER1991). Of these, Rad51 shows the greatest homology withRecA and has some of the expected biochemical properties, including DNA-dependent ATPase and nucleoprotein filament formationonduplex DNA (OGAWAet al. 1993a,b). We found that a rad51rad55rad57 triplemutant strain had the same rate of recombination as the rad51 single mutant. Thus, the mitotic RecA homologues are not functionally redundant in this assay and appear to function in the same recombination pathway (see above). The lack of redundancy in this assay was not unexpected because mutation of any of the mitotic RecA homologs yields a y-radiation-sensitive phenotype, and mutation of any of the four genes results in sporulation andspore viability defects (PETESet al. 1991). This suggests thatperhapsthe many roles of RecA in bacteria are partitioned in yeast among several different proteins or that the proteins function together in a complex. The formation of a complex would be
54
A. J. Rattray and L. S. Symington
consistent with the cs repair defect of rad55 and rad57 mutants. Furthermore,because there are still considerable levels of mitotic recombination in the absence of all three gene products, it suggests the presenceof (an)other RecA-like function (s). It is possible that theDMCl gene is expressed in a rad51 mutant, and either oneof these genes is sufficient for mitotic recombination. The role of the RADl gene was investigated because mutation of this gene reduces intrachromosomal popout events by up to 10-fold, although it is not required for heteroallelic recombination (KLEIN 1988; SCHIESTL and PRAKASH 1988; THOMAS and ROTHSTEIN 1989). We found that radl mutants do not affect the rate of intrachromosomal inverted-repeat recombination, and the distribution of events is the same as the wild-type strain. This contrasts with the results of AGUILERA and KLEIN (1989) who found a decrease in gene conversion tract length and decreased association with crossing over in radl mutants. Although gene conversion tract length cannot be estimated using the ade2 inverted repeat, altered rates of crossing over are readily detected. The reason for the differentresults obtained is not clear but could be explained by the use of different radl alleles or chromosome context effects. We haveshown that radl rad51 mutants are more recombination deficient thana rad51 single mutant strain, indicating that Radl must be able to act on lesions that are normally processed by the EliiD51 pathway. This interpretation is consistent with the results d aL. (1988) who proposed that Radl of MONTEL~ONE processes lesions generatedin hyper-recombination rad3102 mutants into a RALl52-dependentpathway. The most pronounced defect seen in rudl rad51 strains is a synergistic decrease in crossovers, and the major product recovered is crossovers associated with a gene conversion. The RAD1 pathway cannot be functioning by bypassing the requirement for RAD51 and chmnelling the lesions further downstream in that pathway, as in this case rad51 rad57 double-mutant strains would have the phenotype of a rad57 single mutant. It should be noted that the RADl pathway does not represent the only alternative source of recombination events, because recombination in a radl rad51 double-mutant strain is still 250-fold higher than recombination in a rad52 mutant. We cannot determine whether the RADl pathway requires the RAD52 gene, because the level of recombination in a rad52 mutant is at the limits of detection in our assay, and we would be unlikely to distinguish any synergistic decreases. In contrast to the hyper-recombination phenotype seen in mitotic heteroallelic recombination, we find that inverted-repeat recombination is slightly reduced in both rad50andxrs2strains. The distribution of events is similar to that seen in a wild-type strain. A rad50 rad51 double-mutant strain showed no further reduction in the overall rate of prototroph formation, although the pattern of' events recovered most resembled a rad51
mutant. Furthermore, a radl rad50 rad51 triple mutant strain showed an even greater reduction in the rate of prototroph formation than did a rudl rad51 doublemutant strain. One possible explanation for this observation is that the W 5 0 gene product acts early, on a specific subset of spontaneous lesions, and that these are then channelled into either of two pathways. One pathway is acted on by Rad51 or by Radl in a rad51 mutant. The second pathway operates inefficiently in wild-type cellsbut accounts forsome of the recombination events observed in radl rad51 mutants. Thus rad50 mutants have a weak hypo-recombination phenotype because they act on only a subset of DNA lesions. However, a multiplicative effect is observed when combined with mutations in RADl and RAD51 because part of the RAD1 RAD51-independent pathway(s) is inactivated. The rationale for placing RAD50 upstream of RAD51 is based on thedifferent phenotype comparedwith rad54, rad55 and rad57mutants andalso the meiotic data that indicate a role before the initiation of recombination (MALONE and ESPOSITO1981). There are several possible interpretations of the differing recombination phenotypes of rad50 and xrs2 on heteroallelic and inverted-repeat recombination. Heteroallelic recombination must occur by interaction between homologous chromosomes in diploids, whereas inverted-repeat recombination occurs by intrachromatid or sister chromatid interactions (collectively referred toas intrachromosomal events). It is possible that the RAD50 and XRS2 genes areinvolved in regulatinginterchromosomal us. intrachromosomal events. The observation that rad50 and xrs2 haploids lack G2 repair (sister chromatid repair) but show diploid specific repair could be interpreted as retention of the ability for interchromosomalbut notintrachromosomal interactions. These could be competing processes, and in the absence of a system for intrachromosomal recombination the opportunity for interchromosomal interactions may be increased, resulting in a hyper-recombination phenotypefor heteroallelic recombination and hypo-recombination phenotype in intrachromosomal assays. The balance between inter- and intrachromosomal events may also be influenced by heterozygosity at the M A 7 locus in diploids (KADw and H ~ T W E L I . 1992). We present a model based on our data in Figure 5. It is clear that thevast majority of recombination in this system is dependent upon the W 5 2 gene, because rad52 mutants recombine at 0.03% wild-type levels,and the majority of Ade+ prototrophs recovered arise by a nonrecombinational mechanism ( R A n u y a n d S W I N G TON 1994). We cannot be certain that RAD52 initiates the pathway and its placement ismerelyto indicate the importance of this gene in recombination. Meiotic experiments indicate a late role in recombination (MALONE and ESPOSITO1981), anda study of the processing HO-induced double-strand breaks in mitotic cells
Homologous Recombination Pathways in S. cereuisiae
55
Repaired product (Mostly mutagenic)
FIGURE5.-Model for multiple pathways of homologous recombination between inverted repeats in yeast. The vast majority of recombination to produce Ade+ prototrophs is dependent upon theRAD52 gene. Many of the rare events recovered in rad52 mutants showed no physical alteration of the unrecombined substrate. We cannot determine when the RAD52 gene product functions, so its placement in the beginning of the pathway is only meant to denote the RAD52 dependence of recombination events. RAD50 and XRS2 appear to function early in recombination before commitment to a recombination pathway. Because rad50 and xrs2 mutants are only marginally deficient for recombination,it is likely that many lesions bypass the requirement for these gene products. Most of the recombination events appear to be promoted by a RAD51 dependent pathway, which channels lesions downstream to be acted upon by the RAD54, RAD55 and RAD57 gene products. Mutation of any of the four genes in this pathway reduces geneconversions to a greater extent than crossovers, thus we assume that this pathway resolvesrecombination events primarily (but notexclusively) as gene conversions. The R A D 1 gene is able to repair some of the lesions normally processed by the RAD51 pathway, although it may not function in wild-type cells. There is still 250-fold more recombination in radl rad51 cells than in rad52 cells, indicating that there must be at least one other pathway, denoted here as X. The principal product seen in these cells are gene conversions associated with a crossover.
showed that Rad50 acts before Rad52 (SUGARAWA and HABER 1992). RAD50 and XRS2 function early in meiotic recombination, and our data support theidea that RAD50 (and by analogy, XRS2) may function before the commitment to a recombination pathway. These genes are clearly not essential for most inverted-repeat recombination and may only be required for asubset of DNA lesions. The RAD51 gene appears to function early in recombination and channels lesions intoa RAD54, RAD55 and RAD57dependent pathway, which results primarily (althoughnot exclusively) in gene conversions. In the absence of the RAD51 gene, alternative pathways can act upon these lesions and result primarily in crossovers. The RAD51 pathway appears to be able to repair some of these lesions in the absence of RAD51 but is probably not used in its presence. Because radl rad51 double mutants, as well as mutants downstream of RADSZ, still recombine at rates much greater than do rad52 mutants, there is clearly another, as yet unidentified, pathway for recombination. We thank M. AKER, E. IVANOV, R. KEIL, N. KLECKNER, R. ROTHSTEIN and D. SCHILDfor gifts of plasmids and strains. This work was supported by grants from the National Institutes of Health (NIH) (GM41784) and the American Cancer Society (NP-73424) and in part by a grant from theIRMA T. HIRSCHI. Trust. L.S.S. is a Leukemia Society of America Scholar. A.J.R.was supported in part by NIH training grant (NIAID-AI-07161).
LITERATURECITED ABOUSSEKHRA, A,, R. CHANET, A. ADJIRI and F. FABRE,1992 Semidominant suppressors of Srs2 helicase mutations of Saccharomyces cereuisiae map in the RAD51 gene, whose sequence predicts a
protein with similarities to procaryotic RecA proteins. Mol. Cell. Biol. 1 2 3224-3234. AGUILERA, A,, and H. L. KLEIN,1989 Yeast intrachromosomal recombination: long geneconversion tracts are preferentially ass@ ciated with reciprocal exchange and require the R A D 1 and RAD3 gene products. Genetics 123: 683-694. AHN,B.-Y, and D. M. LMNGSTON,1986 Mitotic gene converstion lengths, conconversion patterns and the incidence of reciprocal recombination in a Saccharomycescereuisiae plasmid system. Mol. Cell. Biol. 6: 3685-3693. A I A N I , E., S. SUBBIAH and N. KLECKNER, 1989 The yeast RAD50 gene encodes a predicted 153-kd protein containing a purine nucleotide binding domain andtwo large heptad-repeat regions. Genetics 122: 47-57. BAILIS,A. M., and R. ROTHSTEIN, 1991 A defect in mismatch repair in Saccharomycescereuisiae stimulates ectopic recombination between homeologous genes by an excision repair dependent process. Genetics 126: 535-547. G., M. AKER and R. K. MORTIMER,1992 Nucleotide sequence BASILE, and transcriptional regulation of the yeast recombinational repair gene RAD51. Mol. Cell. Biol. 12: 3235-3246. 1992 DMCI, A BISHOP,D. R , D. PARK,L. Xu and N. KLECKNER, meiosis-specificyeast homolog of E. coli recA required forrecombination, synaptonemal complex formation, and cell cycle progression. Cell 69: 439-456. DORNFELD, K. and D LMNCSTON,1992 Plasmid recombination in a rad52 mutant of Saccharomyces cereuisiae. Genetics 131: 261-276. D m , J. W.,1970 The Molecular Basis of Mutation. Holden-Day, San Francisco, CA. ENGEBRECHT, J., J. HIRSCHand G. S. ROEDER,1990 Meiotic gene conversion and crossing over: their relationship to each other and to chromosome synapsis and segregation. Cell 62: 927-937, GAME,J. C., 1983 Radiation sensitive mutants and repair in yeast, pp. 109-139 in Yeast Cmetirs: Fundamental and Applied Aspects, edited by J. F. T. SPENCER, D. SPENCER and A.R. W. SMITH. Springer-Verlag, New York. J. C., 1993 DNA double-strand breaks and the RAD5@RALl57 GAME, genes in Saccharomyces. Semin. Cancer Biol. 4 73-83. GAME, J., and R. K. MORTIMER, 1974 A genetic study of X-ray sensitive mutants in yeast. Mutat. Res. 24: 281-292. COLIN,J. E. and M. S. ESPOSITO,1981 Mitotic recombination: mismatch correction and replicational resolution of Holliday struc-
56
A. J. Rattray and L. S. Symington
tures formed at the two-strand stage i l l Snrrharomyres t.rrp~i.ti~e. Mol. Geu. Genet. 183: 252-263. HOFFMAN, C . , and F. Wmsroh., 1987 A ten-minute DNA preparation from yeast efficiently releases autonomous plasmids for trausformation of Esrhm'chia rob. Geue 57: 262-272. HOI.I.IIM\', R., 1964 A mechanism for gene conversion i n fungi. Genet. Res. 5: 284-SO4. HLIRSI,D. D., and S . Fo(:EI., 1964 Mitotic recombination aut1 heteroallelic repair i u Sarrhwovyu\ rrrr?jtsiar. Genetics 50: 435458. 110, H., Y. FL"I).A, K. Mt1wr.A a d A. KIML~KA, 1983 Transformatiou of intact yeast cells treated with alkali cations. J. Bacteriol. 153: 163-168. Iw\ruov, E., V. KOKOI.EVand F. FAHKE, 1992 XILS2, a DNA repair geue of Sarrhnnnoyre~rf?#i.siae, is needed for meiotic rccombination. Genetics 132: 651-664. IVANOV, E. I.., N. SU(;AM'A&\, C. 1. WHITE, F. FA~RF. a1ld.1. E. HABER, 1994 Mutations in XRSZ and KAr150 delay but do not prevent mating-type switching in Saccharomycu.y t.mmiriar. Mol. Cell. Biol. 14: 3414-3425. J . U : K X ~J. , A., and G. R. FINK,1981 Gcne conversion between duplicated genetic elements i n yeast. Nature 292: 306311. KAnx, L. C . and I,. H. HAR'I'UTII . , 1992 Sister chromatids are preferred over homologs as substrates for recornhinational repair in Sacchammycus cermisiae. Genetics 132: 387402. KAKAR, S. N., I963 Allelic recombination and its relation to recomhination of outside markers. Genetics 48: 957-966. KANS,.!. A. and R. K. MORTIMER, 1991 Nucleotide sequence of the RAD57 gene of .Sacd~,aromyrerrenwiszar. Gene 105: 139-1 40. KI.EIN,H. L., 1988 Different types of recombination eveuts are corn trolled by the R A D Z and RAD52 geues of .krhn?otnytn rprmisiar. Genetics 120: 367-377. Lo\%TI, S. T., 1994 Sequence of the M 5 5 gene o f Sarchuromyres cmmzsiar: similarity ofRAD55 to prokaryotic RecA and other RecA-like proteins. Gene 142: 103-106. I.ov~;TT, S . T. and K. K. MORTIMIX, 1987 Characterizatiou of n u l l mutants of the KAD55 geue of Sarrharomya rwmisiau: effects of temperature, osmotic strength and mating type. Genetics 116: 547-553. MAILINE, R. E. and R. E. ESPOSITO,1981 Recombinationless meiosis i ~ ~ Cell. . Biol. 1: 891-901. in Saccharomym r m r ? ~ i ~Mol. MAI.oN~., R. E., T. WARD,S. LIN a n d J . WARIN(,,1990 The RAD50 gene, amember of the douhle straudbreakrepair epistasis group, is not required for spontaneous mitotic recombination in yeast. Curr. Genet. 18: 111-1 16. MCDOYAI,~, J. P. and R. ROTIISTEIN, 1994 Unrepaired heteroduplex DNA in Sar.rharomycu,t ~WYUZ.SIICis decreased in K4DZ MD52-independent recombination. Geuetics 137: 393-405. McK~F:,R. H. a d C . U'. LAWRLNLII, 1980 ( h e t i c analysis of y-ray mutagenesis i u yeast. I l l . Double-mutant strains. Mutat. Res. 70: 37-48. MESEI.SON, M. and C. R\L)I)IN(;, 1975 A geueral model for recombinatiou. Proc. Natl. Acad. Sci. USA 72: 358-361. MII.NF,, (;. and D. U'F.A\XR,I993 Dominant uegative alleles of W 5 2 reveal a DNA repair/reconlbination complex iucludiug Rad51 and Rad52. Geues. Drv. 7: 1755-1765. MOF;TFI.ONI:., B. A,, M. F. Hot.k3TRr\\lld R. E. Mt\I,ONb', 1988 spoutaneolts mitotic recombination iu yeast:the hyper-recomhinatiollal reml Inutations are alleles of' the K403 gene. Genetics 119: 289301. OCAWA, T., A. SHINOHARA, A. NARWA N I , T. I K I ~ , \ , X. YL rl nl., 199% Rec.4 like recombinationproteins in eukaryotes: functions Of & E 5 1 and RAD52 genes of SnrrhnromyrPs t.prr?Jisiar.Cold Spring Harhor Symp. Quant. Biol. 58: 109-137. OGAU~A, T., X. Ycl, A. SHINOIIAKA E. and H. E(;k,iMs\N, 199% Similar-
ity of the yeast Rad51 filament to the bacterial RecA filament. Science 259: 18961899. PRIES, T. D., R. E. MALONE and L. S. SYMINGTON, 1991 Recomhination in yeast, pp. 407-521in The M o k m l a r and Cellular Biology of the Yeast Saccharomycm: Genome Dynamics, Protein Synthesis and Entlgrrzcr, edited byJ. R. BROACH, J. R. PRING1.E. and E. W. JONES. Cold Spriug Harbor Laboratory Press, Cold Spring Harbor, NY. ~ \ l ) l ) l N G , C . , 1982 Homologouspairing andstrand exchange in genetic recombination. Annu. Rev. Genet. 16: 405-457. RVL'TMY, A. J., and L. S. SWINGTON,1994 Use of a chromosomal inverted repeat to demonsfrate that theRAD51 and RAD52genes of Sacchnromycp.~cprmisiae have different functious in mitotic recombiuation. Genetics 138: 587-595. K ~ ~ : K M I I .B. I . , and G. S. ROEDER, 1990 Meiosisin asvnaptic yeast. Genetics 126: 563-574. ROMAN,H. L., 1957 Studies of recombination in yeast. Cold Spring Harbor Symp. Quaut. Biol. 21: 175-183. ROMAN,H. and M. RUZINSKI, 1990 Mechanisms of gene conversion in Saccharomyces cermisiae. Genetics 124: 7-25. ROTHSTEIN, R. J., 1983 One-step gene disruption in yeast. Methods Enzymol. 101: 202-21 1. SAF.KI, T.. I . M~C:HIUA and S. NAKAI, 1980 Genetic control of diploid recovely after y-irradiation in the yeast Sarcharomyces rmmisiae. Mutat. Res. 73: 251-265. SCHIESTI., R. and S . PRAuw, 1988 RADl, an excision repair gene of Saccharomyces cermisiae, is also involved in recombination. Mol. Cell. Biol. 8: 3619-36'26. Scm.L), D., 8.J. GIASSNER, R. K. MORTIMER, M. CARLSON and B.C. LU'R~NT, 1992 Identification of RAD16, a yeast excision repair gene homologous totherecornhinationalrepair gene RAD54 and to the SNF2gene involved in transcriptional activation. Yeast 8: 385-395. SEC:AI.I., A. M., and J. R. RorH, 1994 Approaches to half-tetrad analysis iu bacteria: recombination between repeated, inverse-order chromosomal sequences. Genetics 136: 27-39. F., G. FINKand J. HICKS,1986 Methods in Yeast Genetin. SHERMAN, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY. SHIKOHARA, A,, H. OGAWA and T. OGAWA, 1992 Rad51 protein involved in repair and recombination in S. cermisiar is a RecA-like protein. Cell 69: 457-470. SMITII, G. R.. 1989 Homologous recombination in E. coli: multiple pathways for multiple reasons. Cell 5 8 807-809. SUGAW.,\M,N., and J. E. HAWR, 1992 Characterization of douhlestrandbreak-inducedrecombination: homology requirements and single-stranded DNA formation. Mol. Cell. Biol. 12: 56% 575. Szosr,w, J. W., T. I,. ORR-WEAVER, R. J. ROTHSTEIN and F. W. STAHI., 1983 The double-strand-break repair model for recombination. Cell 33: 25-35. THOMAS, B. J. and R. ROTHSTEIN, 1989 The genetic control of directrepeat recombination in Saccharomyces: the effect of rad52 and m d Z on mitotic recomhination at GALIO, a transcriptionally regulated gene. Genetics 123: 725-738. TOMKINSON, A. E., A. J . BARDWEJ.~., L. BARDWEL.~., N. J . TMPEand E.
