ORIGINAL PAPER

Ori Inbar á Martin Kupiec

Recombination between divergent sequences leads to cell deathin a mismatch-repair-independent manner

Received: 11 January / 13 March 2000

Abstract Homologous recombination is an importantDNA repair mechanism in vegetative cells. During therepair of double-strand breaks, genetic information istransferred between the interacting DNA sequences,thus creating a gene-conversion event. Gene conversionof a functional member of a gene family, which uses aninactive member (such as a pseudogene) as a template,might have deleterious consequences. It is thereforeimportant for the cell to prevent recombination be-tween divergent sequences. We have studied the repairof a double-strand break by recombination in a haploidyeast strain carrying 99% identical alleles located ondi�erent chromosomes. The fate of the broken chro-mosome was followed in the whole cell populationwithout imposing selective constraints. Our resultsshow that all the cells were able to repair the brokenchromosome by gene conversion. During the repair,the cells arrest in the cell cycle with a ``dumbbell''con®guration characteristic of G2/M-arrested cells.Surprisingly, although all the cells repaired the brokenchromosome, 60% of them were unable to resumegrowth and to form colonies after the repair wascompleted. The low level of cell recovery was due to the1% divergence between the alleles, but was not depen-dent on the function of the mismatch-repair system.Cell death, however, could be prevented by the presenceof an alternative source of perfect homology located ona di�erent chromosome.

Key words Gene conversion á Double-strandbreaks á DNA repair á Mismatch repair

Introduction

Homologous recombination is a universal process; itplays a role in generating diversity during meiosis, and isan important DNA repair mechanism in vegetative cells.Recombination results in the transfer of genetic infor-mation from one DNA molecule to a homologous one(gene conversion) and in the reciprocal exchange ofDNA fragments between chromosomes (crossing-over).

Double-strand breaks (DSBs) in the DNA of livingorganisms occur as a consequence of the natural cellmetabolism, or can be created by exogenous sources,such as chemical agents or radiation. In addition, DSBsare generated during certain developmental processes,such as V(D)J recombination in lymphoid cells (re-viewed in Bogue and Roth 1996), meiosis (Zenvirthet al. 1992) and mating-type switch in yeast (White andHaber 1990). If left unrepaired, DSBs result in brokenchromosomes and cell death. Mitotic recombinationserves as an important mechanism able to repair thistype of damage (for a review see Petes et al. 1991). Thegenomic instability resulting from incorrectly repairedDSBs may also lead to carcinogenesis or inborn defects.

A model for DSB-initiated recombination was pro-posed by Szostak et al. [DSB repair model, (Szostaket al. 1983)]. Basically, single-strand degradation at thebroken DNA molecule creates protruding 3¢ ends thatcan invade the homologous regions in the other chro-mosome, forming two cross-stranded structures (Holli-day junctions) surrounded by regions of heteroduplexDNA (hDNA). Mismatch-repair in the hDNA may leadto gene conversion and, depending on the resolution ofthe intermediates, this gene conversion may or may notbe associated with reciprocal exchange. Alternative re-combination models have been proposed in which onlyone of the broken DNA ends invades the homologouschromosome, copying information, and then reannealswith the second broken arm, without forming Hollidayjunctions. Many versions of this Synthesis-DependentStrand Annealing (SDSA) model have been proposed

Curr Genet (2000) 38: 23±32 Ó Springer-Verlag 2000

Communicated by L. A. Grivell

O. Inbar á M. Kupiec (&)Department of Molecular Microbiology and Biotechnology,Tel Aviv University, Ramat Aviv 69978, Israele-mail: martin@ccsg.tau.ac.ilTel.: +972-3-640 9031; Fax: +972-3-640 9407

(summarized in Paques and Haber 1999). In cells un-dergoing meiosis, an intermediate containing two Holli-day junctions has been identi®ed; in addition, mutationsin the MMR genes usually result in a reduction in gene-conversion events during meiosis with a concomitantincrease in post-meiotic segregation and other types ofnon-Mendelian segregation (Alani et al. 1994). Theseresults are completely consistent with the predictions ofthe DSB repair model. However, similar evidence islacking for recombination in vegetative cells.

The mismatch repair (MMR) system is composed ofproteins able to identify and lead to the correction ofmismatched DNA created during replication or recom-bination. The MMR proteins of prokaryotes and euk-aryotes have been extensively analyzed. Homologs of thebacterial MutS protein, which binds to mismatched basepairs, and of the MutL protein, which interacts withMutS, have been identi®ed in many eukaryotic organ-isms. In eukaryotes the MMR proteins form severalalternative complexes that usually include two di�erentMutS homologs and two MutL homologs (reviewed inKolodner 1996). The MMR proteins that have beenimplicated in mitotic recombination include the MutShomologs Msh2p, Msh3p and Msh6p, and the MutLhomologs Pms1p and Mlh1p (Datta et al. 1996, 1997;Selva 1997; Nicholson 2000). The MMR proteins havebeen shown to prevent recombination between divergentDNA sequences in prokaryotes and eukaryotes (re-viewed in Kolodner 1996). This anti-recombinationale�ect is presumably aimed at preventing recombinationbetween divergent sequences and thus may play a centralrole in the creation of new species. In addition, theMMR proteins may prevent recombination betweendivergent members of gene families within a singlegenome and, in this way, they contribute to the integrityof the genome.

In the present study we have analyzed the repair of asingle chromosomal DSB in haploid yeast cells by re-combination.We show that a divergence level of only 1%between the interacting sequences does not a�ect the ef-®ciency of repair by gene conversion, but has a majore�ect on the ability of the cells to resume growth andform colonies after successfully completing the repair.TheMMR system is not responsible for this low viability.

Materials and methods

Media and growth conditions

Saccharomyces cerevisiae strains were grown at 30 °C. StandardYEP medium (1% Yeast extract, 2% Bacto peptone) supplementedwith 3% glycerol (YEPGly), 2% galactose (YEPGal) or 2%dextrose (YEPD) was used for non-selective growth (Shermanet al. 1986); 1.8% Bacto-agar was added for solid media.

Yeast strains

All the strains used in the present work are isogenic derivatives ofOI27 (MATa-inc ura3-HOcs-inc ade3::GALHO ade2-1 leu2-3,112

his3-11,15 trp1-1 can1-100) (Inbar and Kupiec 1999). At the LYS2locus on chromosome II OI29 carries a 1.2-kb HindIII fragmentcontaining the ura3-HOcs allele. This allele was created by insertinga 39-bp oligonucleotide at the NcoI site of the URA3 gene. OI31carries a similar 1.2 kb HindIII fragment containing the ura3-HOcsallele, into which changes were introduced by site-speci®c muta-genesis at approximately 100-bp intervals (Sweetser et al. 1994).The di�erent alleles used were inserted at a HpaI site within LYS2sequences in the integrative plasmid pOI5 (TRP1 LYS2 URA3).They were integrated into yeast chromosome II by a two-stepreplacement method. The ura3 allele at chromosome V carries a39-bp insertion containing the HOcs-inc allele, and EcoRI andBamHI restriction sites. These polymorphisms represent singlebase-pair substitutions (G to A and T to C for HOcs-inc andEcoRI, respectively), and a three-consecutive base-pairs substitu-tion (AAT to GGA) for the BamHI site. In strain MK182 a LYS2-containing 4.9-kb XbaI-HindIII DNA fragment was integratedclose to the HIS3 locus on chromosome XV. The deletions of theMMR genes were created by one- or two-step transplacement usingplasmids pSR211 for PMS1, pSR453 for MLH1, pSR395 forMSH2, pRK366 for MSH3, and pRK465 for MSH6 (Alani et al.1994; Datta et al. 1996). For each yeast strain created by trans-formation, at least two independent transformants were chosen forrecombination analysis.

Recombination assays

Single colonies were resuspended in rich YEPGly medium, grownto logarithmic phase, centrifuged and then resuspended in YEPGalor YEPD medium. DNA was extracted from samples at timelyintervals, and subjected to Southern-blot analysis using LYS2 se-quences as probes. The blots were quanti®ed with a Fujix BAS1000phosphorimager. For each strain tested, the experiment wasrepeated at least three times. The e�ciency of cell recovery wasestimated by resuspending at least ten independent colonies inwater and plating on YEPGal and YEPD plates. The statisticalsigni®cance of the results was examined with a two-tailed two-sample Mann-Whitney test. Independent colonies that grew 3 dayson YEPGal plates were then subjected to PCR or Southern-blotanalysis to monitor the presence of the EcoRI or BamHI sitesindicative of a gene-conversion event.

DSB lifespan

We have measured the mean period of time during which the DSBremains unrepaired by the method of Padmore et al. (1991).Brie¯y, the area under the curves in Fig. 1D and E (in units of cellpercentage with DSBs ´ time) was divided by the total percentageof cells undergoing DSB. This calculation takes into account anypossible di�erences in the synchrony of DSB formation between thetwo strains.

Results

Experimental system

In order to study the e�ect of sequence heterology onrecombinational repair, we have developed an assay forDSB-initiated interchromosomal recombination(Fig. 1A). A haploid yeast strain (OI29) bears twocopies of the URA3 gene; one of them carries the rec-ognition site for the yeast HO site-speci®c endonuclease(Nickolo� et al. 1990) inserted as a short oligonucleotide(ura3-HOcs). The second copy, located on anotherchromosome, carries a similar site containing a single-bpmutation that prevents recognition by the endonuclease

24

(ura3-HOcs-inc). In addition, the two ura3 alleles di�erat two restriction sites, located to the right and to the leftof the HOcs-inc insertion; these polymorphisms are usedto follow the transfer of information between the chro-mosomes. In these strains the HO gene is under thetranscriptional control of the GAL1 promoter (Sandelland Zakian 1993). Upon transfer of the cells to galac-tose-containing medium, the HO endonuclease is pro-duced at elevated levels. The enzyme creates a DSB; ifleft unrepaired, such a lesion leads to cell death (Bennettet al. 1993; Moore and Haber 1996). The presence ofhomologous information in the genome can be used torepair the DSB by recombination. The broken chro-mosome shares homology with its allele up to the veryend; therefore, when they interact, no non-homologoustails are created. Such tails have been shown to a�ect therecombinational repair (Sugawara et al. 1997). In this

system, no genetic selection is applied in order to mon-itor recombination, and therefore it is possible to followthe repair of the whole cell population without imposingselective constraints.

Twenty four independent colonies that grew on ga-lactose-containing medium were subjected to PCR andDNA-sequence analysis. In all of them, the HOcsinformation on the ura3 allele on chromosome II was

Fig. 1 A Schematic representation of the experimental system used inthis study. Open rectangles represent the ura3 alleles on chromosomesII and V. Striped rectangles represent LYS2 sequences on chromo-some II, which were used as probes for Southern-blot analysis. Astippled box represents the HOcs.; black boxes represent the inactiveHOcs-inc, which also shows two lines, representing polymorphic sitesrecognized by the EcoRI and BamHI restriction enzymes. Transfer ofthe cells to galactose-containing medium induces the production ofHO endonuclease, which recognizes the HOcs and creates a DSB.This break is repaired by a gene conversion event that transfers theHOcs-inc to chromosome II thus preventing further recognition by theHO endonuclease. Unrepaired chromosomes lead to cell death.B Southern-blot analysis of DNA extracted from strain OI29 atintervals. The DNA was digested with BglII and probed with afragment of chromosome II carrying the LYS2 gene. In the schematicmap, the region covered by the LYS2 probe is shown. A squarerepresents the ura3:: HOcs allele. The reference bands, which are usedto quantitate the amount of DNA in each lane, include the5.3-kb band neighbouring the ura3 allele, and two additional bands(5.1 kb and 3.0 kb), which result from hybridization of vectorsequences present in the probe to another chromosomal region. Inthis blot the parental bands and the gene-conversion products migratetogether. C Southern-blot analysis of DNA extracted from theexperiment presented in B. The DNA was digested with BamHI andBglII and probed with a fragment of chromosome II carrying theLYS2 gene. In this blot the gene-conversion products migrate togetherwith the DSB bands. D Kinetics of DSB repair. The Southern blotsshown in B and C were scanned; the intensity of each band ispresented after normalization with the reference bands. E Kinetics ofDSB repair in strain OI31, bearing 1% divergent ura3 alleles

Fig. 1 (Contd.)

25

replaced by HOcs-inc information (from the unbrokenura3 allele on chromosome V ). This transfer occurredtogether with the copying of the EcoRI and BamHIpolymorphic sites present to the right and to the left ofthe HOcs-inc insert. These results con®rm that the e�-cient repair observed is carried out by gene conversion.

The kinetics of recombination of a logarithmic cellculture transferred to galactose-containing medium wasfollowed in real time. DNA samples taken at intervalswere subjected to two di�erent restriction digestionsfollowed by Southern-blot analysis. In the ®rst blot(Fig. 1B) we monitored the appearance and disappear-ance of the DSB. Induction of the HO endonucleasecreated a chromosomal break, which could be followedby the appearance of two new bands (1.9 and 4.2 kb insize), and the concomitant decrease in intensity of theparental band (6.1 kb). Repair of the broken chromo-

some lead to the disappearance of the small DSB bandsand the creation of a band of the same size as the pa-rental. In order to distinguish between the parental bandand the recombinant products the same DNA sampleswere subjected to a second Southern-blot analysis(Fig. 1C). In this blot the conversion products wererecognized by the presence of EcoRI or BamHI sitestransferred during the repair of the DSB. Only unbrokenchromosomes that have not been repaired by geneconversion give a parental band of 4.1 kb. It is apparentfrom Fig. 1C that most of the population has undergonea gene-conversion event by 7 h.

The results of the two blots can be integrated in aquantitative way (Fig. 1D). Within 45 min after transfer,two new bands were seen, representing the brokenchromosome arms. These bands accumulated, reaching amaximal level 2 h after transfer; at this point about halfof the cell population showed a broken chromosome II.The bands representing the DSB then disappeared in aprocess that took several hours. Gene-conversion prod-ucts started to appear simultaneously with the decreasein intensity of the DSB bands, and the repair in thewhole cell population was ®nished 7 h after the cells weretransferred to galactose. During the repair, the donor(unbroken) chromosome remained unchanged (data notshown). No crossing-over was seen associated with therepair of the DSB; in another study we have shown thatthe lack of association was due to the limited length ofhomology between the interacting alleles (1.2 kb).

In order to analyze the e�ect of heterology onchromosomal recombinational repair, we created astrain (OI31) which is identical to strain OI29, exceptthat the 1.2-kb ura3-HOcs inserted at chromosome IIhas been modi®ed by introducing one sequence changeabout every 100 bp. Most of these changes are singlebase-pair substitutions (Sweetser et al. 1994). Thus, thetwo ura3 alleles show 1% sequence divergence and en-able us to study the e�ect of low heterology levels on therecombinational repair of chromosomal DSBs. A cul-ture of strain OI31 was transferred to galactose-con-taining medium, and samples were taken and analyzedby Southern blots. Figure 1E summarizes one experi-ment in a quantitative way; it shows that strain OI31was pro®cient for DSB repair, since approximately 99%of the cells repaired the break by a gene-conversionevent. The kinetics of accumulation of broken DNAmolecules was similar to that of OI29. In both strainsrepair was completed within 7 h after the cells weretransferred to galactose-containing medium, althoughthroughout the experiment the level of cells of strainOI31 with broken chromosomes was higher than thatobserved in strain OI29.

The di�erent kinetics seen in strains OI29 and OI31could be due either to di�erences in the synchrony ofDSB formation between the strains, or to di�erences inthe time required for the repair of the break. From thekinetics of accumulation of the broken chromosomes inexperiments such as shown in Fig. 1 it is possible todeduce the mean period of time during which the DSB

Fig. 1 (Contd.)

26

remains unrepaired (DSB lifespan) (Padmore et al.1991). This calculation takes into account any possibledi�erences in the synchrony of DSB formation betweenthe two strains. The DSB lifespan of strain OI29 is1.9 � 0.2 h whereas that of strain OI31 is 2.9 � 0.4 h.Thus, the presence of a low level of heterology prolongsthe duration of the repair of the broken chromosome. Insummary, the presence of a low level of heterologydelays, but does not prevent, the successful repair of aDSB by homologous recombination.

Analysis of the conversion tracts at the brokenchromosomal ends

The presence of restriction length polymorphismsbetween the two interacting ura3 alleles in strain OI31enabled us to study the extent of genetic informationtransferred during the repair process (conversion tractlength). Figure 2 shows the results obtained from theanalysis of 20 independent repair events. The results canbe summarized as follows. (1) In all the cases analyzed,genetic information from both sides of the DSB wasreplaced by the homologous sequences: in all the events,both the inc mutation, to the left of the DSB, and theBamHI site, to its right, were transferred. (2) In 30% ofthe events the conversion tracts were shorter than 97 bpin length (they did not include the PmlI and StuI sites).(3) The transfer of information was asymmetric: al-though both ends participated in the repair, in mostcases the conversion tract was very short on one of thearms, and longer on the other. For example, 30% of theconversion tracts included the inc mutation, but did notreach the EcoRI site located 15-bp away from it on theleft arm. Moreover, 60% of the conversion tracts did notreach the PmlI site, located 47-bp away on the samearm. Similarly, on the other side, 45% of the conversionevents analyzed did not include the StuI site located

46-bp away from the DSB. In only 15% of the cases theconversion tract extended more than about 50 bp oneach side and co-transferred the PmlI and the StuI sites.The involvement of both broken chromosomal ends,and the asymmetry of the conversion tracts, should betaken into consideration when trying to explain themechanism of recombinational repair of DSBs, asdiscussed below (see Discussion).

Divergence between the interacting alleles causesa reduction in the level of cell recovery

When strains OI29 and OI31 are plated on galactose-containing medium, the HO endonuclease is constitu-tively expressed, and a DSB is created in essentially thewhole cell population. In order for the cells to formcolonies, the DSB has to be repaired; thus, the e�ciencyof DSB repair can be estimated by comparing the abilityto form colonies on galactose-containing medium (con-stitutive HO expression) to that which is seen in glucose-containing medium (no HO expression). For example,an isogenic strain lacking sequences homologous tothose ¯anking the DSB shows a plating e�ciency of only0.05% of the cells on galactose-containing plates. Incomparison, 80% of the OI29 cells plated on galactose-containing medium are able to form colonies, demon-strating that the homology dependent repair is verye�cient.

Surprisingly, only 37% of the cells of strain OI31,bearing the 1% divergent ura3 alleles, were able to formcolonies on galactose-containing medium (in which theHO endonuclease is constitutively expressed). The per-centage of cells of this strain unable to resume growth isthree-fold higher than that seen in the isogenic controlstrain OI29; the di�erence between the results of the twostrains is statistically signi®cant (P < 0.0002). Thesimplest interpretation of these results would be that thelow recovery e�ciency observed is due to a failure torepair the DSB. We can overrule this possibility, how-ever, since Southern-blot analyses demonstrate that theall DSBs (>99%) are e�ciently repaired by gene con-version, and that no broken chromosomes can be seenlater than 7 h after transfer to galactose (Fig. 1E,Fig. 3). It is still possible that the results obtained inSouthern-blot analyses monitor only part of the cellpopulation, and that the DNA of about 60% of the cellscannot be seen in the blots, because of extensive degra-dation of the broken chromosomal ends. The results ofSouthern-blot analyses using DNA probes from ¯ankingchromosomal regions, however, rule out this possibility.(1) The DSB bands were always discrete and show nosign of double-stranded degradation. (2) The 4.2-kb``Reference'' band seen in Fig. 3 represents sequenceslocated 2.5 kb from the DSB. No decrease in the size orintensity of this band was observed during the experi-ments. (3) Additional probes to either side of the DSB,at various distances from the break, also failed to showany evidence for DNA degradation around the break.

Fig. 2 Distribution of conversion tracts among cells of strain OI31that repaired a broken chromosome. The ®ve RFLP markers close tothe DSB are presented, with the distance from the DSB given in bp.The EcoRI, inc, and BamHI markers are present in the donorchromosome, whereas the PmlI and StuI markers (between brackets),are present in the recipient chromosome. Black bars indicate theminimal conversion tract length for each category

27

Finally, denaturing Southern-blot analysis of cells takenafter 10-h incubation in galactose failed to show anyunligated nicks in the repaired chromosome (data notshown). We therefore conclude that in cells carryingdivergent ura3 alleles the DSB is e�ciently repaired, butmost of the cells nevertheless die and cannot resumegrowth to create a colony. This cell death is dependenton the presence of a low divergence level (1%) betweenthe interacting alleles.

The reduced cell recovery of strains carrying divergentalleles is due to the inability to exit the cell-cycle arrest

In order to understand the reasons for the low colonyforming ability of OI31 cells under conditions of con-stitutive HO expression, we monitored individual cellsunder the microscope. We could thus follow the fate ofthose cells that were unable to form colonies under theseconditions. Unbudded (G1) cells of strains OI29 andOI31 were micromanipulated onto a grid on galactose-containing plates. By 5 h after transfer, the majority ofthe cells (approximately 80%) in both strains had ar-rested at the two-cell stage characteristic of G2/M arrestand had become enlarged (Fig. 4A). Most of the cells ofstrain OI29 were able to resume growth by 8±10 h (datanot shown), and by 24 h after transfer 86% of the cells

had divided many times and eventually developed into acolony (Fig. 4B). In contrast, only 39% of the cells ofstrain OI31 were able to do so, and 50% of them werestill arrested by 24 h (Fig. 4B). Most of the arrested cellsstill exhibited the G2/M ``dumbbell'' shape at this time,although in some of them one or two small buds wereobserved, which did not develop further. The percent ofmicromanipulated cells that were able to exit the cell-cycle arrest (86% and 39% for strains OI29 and OI31,respectively), correlates well with the plating e�ciency ofthese strains on galactose plates (80% and 37%, re-spectively). We thus conclude that the low recoveryobserved under conditions of continuous HO expressionwas due to the inability of the cells to leave the G2/Mcell-cycle arrest.

The low level of cell recovery of strain OI31is not due to the action of the MMR system

During homologous recombination between non-iden-tical sequences, heteroduplex DNA containing mis-matched base pairs may be formed. Studies in bacteria,yeast and mammalian cells have shown that sequencedivergence acts as a potent barrier to recombination.The MMR proteins have been shown to participate inthis anti-recombinational mechanism (summarized inKolodner 1996).

It is possible that the low level of recovery followingDSB creation shown by the strain with 1% divergedura3 alleles is due to the action of the MMR system. Wehave therefore created a series of isogenic strains, eachone defective in one or more of the MMR proteins.Table 1 shows that deletion of the MMR genes did notimprove the low recovery seen in strain OI31. A com-parison of each MMR) mutant in OI29 derivatives, vsthe isogenic polymorphic strain, always showed a sig-ni®cant result, at levels of <0.0004. This is in contrast tothe comparison between the di�erent OI31-derivatives instrains carrying di�erent combinations of MMR muta-

Fig. 3 A,B DSB repair in MMR+ and MMR) strains carryingdivergent ura3 alleles. Southern-blot analysis of DNA from strainsOI31 (MMR+) and OI31 msh2D msh3D msh6D pms1D (MMR))extracted at intervals. A The DNA was digested with XbaI andHindIII and probed with a fragment of chromosome II carrying theLYS2 gene. The reference bands are neighboring invariant DNArestriction fragments which are used to quantitate the amount ofDNA in each lane. In this blot the parental bands and the gene-conversion products migrate together. B The DSB was repaired bya gene-conversion event that transferred information from theura3HOcs-inc allele. DNA of the same strains was extracted 24 hafter transfer to galactose-containing medium, digested withBamHI and XbaI and probed with a fragment of chromosome IIcarrying the LYS2 gene

28

tions, which showed results that were not statisticallysigni®cant in most cases, and questionably signi®cant inthe single case of mlh1D. Even in the quadruple mutantsmsh2D msh3D msh6D pms1D and msh2D msh3D msh6Dmlh1D only about 40% of the cells of OI31 derivativeswere able to form colonies, compared to about 80% ofthose derived from strain OI29. Southern-blot analysesshowed that the quadruple MMR) derivatives of bothstrains were able to repair the broken chromosomes bygene conversion with an e�ciency and kinetics similar tothose of the parental strains (Fig. 3). We concludetherefore that the low ability of OI31 cells to form viablecolonies after DSB repair is not due to the function ofthe MMR genes.

The low level of cell recovery of strain OI31can be prevented by supplying an alternative sourceof homology

We have shown that cells of strain OI31 (1% divergentalleles) repair the DSB in their chromosomes e�ciently;however, most of them are unable to resume growth andform colonies. We wanted to test whether it is possible torescue the cells by supplying an alternative source ofhomology, which may serve as a template to repair thebroken chromosome. On the other hand, it is possiblethat the presence of divergent DNA ¯anking the DSBmay lead to cell death, despite the presence of an alter-native option. A second copy of the LYS2 sequences¯anking the site of insertion of the ura3-HOcs allele wasinserted on chromosome XV of strain OI31, as shown inFig. 5. When a DSB is created in this strain (MK182) atthe ura3-HOcs allele, it can be repaired either by a gene-conversion event involving the divergent ura3-HOcs-incallele on chromosome V, or by gene conversion with theLYS2 fragment on chromosome XV. In the latter case,

Fig. 4 Cell-cycle arrest in cells repairing a broken chromosome. One-hundred unbudded (G1) cells of strains OI29 and OI31 grown onYEPGly were micromanipulated onto a grid on YEPGal plates. Thenumber of cells at each position was scored at timely intervals (2, 5, 8,12 and 24 h after transfer). Data from times: 5 h and 24 h arepresented

Table 1 Cell recovery of the di�erent MMR defective strains

Genotype OI29-derivativea OI31-derivativea

Wild-type 0.80 � 0.12 0.37 � 0.05msh2D 0.69 � 0.07 0.27 � 0.04msh3D 0.99 � 0.17 0.41 � 0.07msh6D 0.90 � 0.23 0.46 � 0.04pms1D 0.97 � 0.12 0.49 � 0.07mlh1D 0.84 � 0.11 0.53 � 0.07msh2D msh3D msh6D 0.82 � 0.15 0.44 � 0.08msh2D msh3D msh6D pms1D 0.80 � 0.14 0.36 � 0.05msh2D msh3D msh6D mlh1D 0.74 � 0.14 0.37 � 0.07

aRatio of the number of colonies on galactose-containing medium/number of colonies on glucose-containing medium. Strains derivedfrom OI29 carry near-identical ura3 alleles, whereas strains derivedfrom OI31 carry 1% divergent ura3 alleles. From each strain atleast 700 colonies were plated on each medium. The di�erencebetween pairs of OI29 and OI31-derivatives for each MMR)

mutant combination was signi®cant in all cases with a Pvalue <0.0004

Fig. 5 Prevention of cell death by an alternative source of perfecthomology. Symbols are as in Fig. 1A. The ten sequence polymor-phisms in the ura3-HOcs allele of strain OI31 are symbolized byvertical lines in the ura3 allele (open square) on chromosome II

29

the 1.2-kb ura3 information present at the ends of thebroken chromosome has to be removed (Inbar andKupiec 1999).

When cells of strain MK182 were plated on galac-tose-containing medium, the DSB was e�ciently re-paired and 79% of the cells were now able to formcolonies. These results are signi®cantly di�erent(P < 0.0002) from those obtained with the parentalstrain OI31 (37%). Thus, the presence of an alternativesource of homology was able to increase recovery of thecells after DSB repair.

Southern-blot analysis showed that most of the cells(94.3%) repaired the DSB by gene conversion with theLYS2 sequences. The lifespan of the DSB in this strain,however, was not signi®cantly shortened, indicating thatthe long DSB lifespan of strain OI31 is not the reasonfor its low recovery after DSB repair.

Discussion

Double-strand breaks are a common type of DNAdamage; if left unrepaired, DSBs result in brokenchromosomes and cell death. Homologous recombina-tion is an important DNA repair mechanism in vegeta-tive cells. When a chromosome is broken in more thanone site, random joining of the broken fragments cancreate translocations, inversions, and other chromoso-mal aberrations. By using homologous sequences astemplates during repair, recombination is able to restorethe original chromosomal structure. In addition, thetransfer of genetic information during the recombina-tional repair of DSBs prevents mutagenesis by restoringgenetic information that may have been eliminated atthe site of the DNA damage. On the other hand, geneconversion of a functional member of a gene familyusing an inactive gene (such as a pseudogene) as atemplate might have deleterious consequences. There-fore, it is important for the cell to be able to di�erentiatebetween identical information (such as that present inthe homologous chromosome or in the sister chromatid)and divergent genetic information located elsewhere inthe genome.

The study of the role played by sequence divergencein recombination is hampered by the fact that moni-toring recombination usually requires the selection ofrare genetic products. This introduces biases in the typeof events studied, and precludes an understanding of theprocesses that take place in the majority of the cells. Inorder to overcome these problems, we have developed arecombination system that monitors the fate of cells thathave su�ered a DSB in their chromosomes, withoutimposing selective constraints. The system thereforeenables us to test the e�ect of low levels of heterology onrecombinational repair.

We have analyzed the extent of genetic informationtransferred during the repair of a broken chromosome invegetative cells. Our results clearly show that, during therepair, genetic information is always transferred from

the intact homologous sequence to both sides of thebroken chromosome (i.e. the inc mutation and theBamHI polymorphism are always transferred). Thus,both broken arms participate in the recombinationprocess. The role played by the two broken arms,however, is not identical: the great majority of theconversion tracts are asymmetric, with very short con-version tracts on one of the arms, and longer on theother. Similar results were shown before for plasmid-chromosome interactions (Nelson et al. 1996). Severalmodels of DSB-initiated recombination have been pro-posed. The DSB repair model proposes a symmetricalinvasion of the homologous sequences by the two bro-ken arms; in contrast, the SDSA models propose thatgene conversion is initiated by invasion of only one ofthe broken arms (reviewed in Paques and Haber 1999).The asymmetric distribution of the conversion tractssuggests that the recombinational repair takes place byan SDSA-type of mechanism.

Mutations in the MMR genes did not a�ect theability of strains carrying near-identical (OI29) or 1%diverged (OI31) alleles to repair their broken DNA bygene conversion (Fig. 3 and data not shown). Our re-sults thus directly show that MMR) mutants are per-fectly capable of carrying out gene-conversion events invegetative cells. Hence, the repair of a broken chromo-some by gene conversion does not involve the repair ofmismatches in heteroduplex DNA by the MMR system.In this respect too, our results are consistent with anSDSA-type of recombination, and not with otherproposed models.

We have found that in OI29, a strain carrying onlythree heterologies in a region spanning 1.2 kb (99.75%identity), the whole cell population (>99%) is able torepair the break by a gene-conversion mechanism, asseen by examining DNA samples throughout the pro-cess. The e�ciency of DNA repair of a strain carryingura3 alleles with 1% heterology (OI31) is identical: allthe cells are able to repair the break, albeit with slowerkinetics. Thus, the low divergence levels do not a�ect theability of the broken DNA molecules to undergo re-combinational repair. Surprisingly, however, 60% of thecells of the strain carrying ura3 alleles with 1% hete-rology are unable to form colonies after the DNA isrepaired.

The MMR proteins can interact with mismatchedDNA in vitro, and have been shown to exhibit anti-recombinogenic e�ects in vivo (reviewed in Kolodner1996). They were therefore good candidates for proteinsthat may be responsible for the inability of the cells torecover. The MMR proteins could exert this e�ect byrecognizing the presence of mismatches in the hetero-duplex DNA created during recombination. Our results,however, show that the low cell recovery in strains car-rying divergent alleles is not due to the action of theMMR proteins. MMR+ or MMR) strains exhibit thesame low cell-recovery level, when the recombiningsequences are divergent (Table 1). We also show that theMMR genes do not play a signi®cant role in preventing

30

recombination between 1% diverged sequences. Thisresult seems to contradict some of those obtained byother laboratories. In most of the studies carried out invegetative cells, in which a strong e�ect of mutations inthe MMR system was observed, the interacting se-quences were located in close proximity on the chromo-some as direct or inverted repeats (Datta et al. 1996,1997; Selva et al. 1997; Chen and Jinks-Robertson 1999).Recombination between these sequences can take placeby di�erent mechanisms such as single-strand annealingor recombination between sister chromatids (Schiestlet al. 1988; Chen and Jinks-Robertson 1999). In oursystem the broken chromosome is repaired using infor-mation present on another chromosome. It is possiblethat the MMR proteins play no role in this type of repair.Similar results were obtained in other inter-molecularrecombination systems (Priebe et al. 1994; Porter et al.1996), although not in all of them (Negritto et al. 1997).

If the broken chromosomes are e�ciently repaired instrains with divergent alleles, and the MMR systemplays no role in preventing this repair, what might be thereason for the inability of most of the cells to resumegrowth and form colonies after the repair is completed?

When cells are subjected to DNA damage, speci®ccheckpoint mechanisms are elicited that cause cell-cyclearrest, presumably to allow more time for DNA repair(Elledge 1996). Eventually, after the damage is repaired,the cells resume cell-cycle progression (Sandell andZakian 1993; Toczyski et al. 1997). Recent results haveshown that the processing of a broken chromosome af-fects the ability of the cells to resume growth and formcolonies (Bennett et al. 1993; Lee et al. 1998). WhenDSBs that are irreparable by homologous recombina-tion are created, the recovery is dependent on the extentof ssDNA generated at the broken chromosomal ends.Mutations that increase the rate of ssDNA resection,such as hdf1 (ku70), cause a permanent arrest, whereasmutations that retard the degradation (e.g. mre11) sup-press this arrest. Thus, the authors have suggested thatthe permanent arrest of these cells is a consequence ofthe high level of a signal that depends on DSB pro-cessing (Lee et al. 1998). We have shown that cells with abroken chromosome in which the damage can be re-paired by homologous recombination also arrest at theG2/M checkpoint (Fig. 4). Whereas this arrest is tran-sient for most cells of a strain carrying near-identicalalleles, most of the cells of a strain carrying alleles with1% divergence are unable to exit from this stage.

It is possible that the permanent arrest of the majorityof OI31 cells is due to higher levels of a signal such as theone proposed by Lee et al. (1998) to explain the arrest ina system in which the repair by homologous recombi-nation was not possible. We have shown that cells ofstrain OI31 remain with broken chromosomes for alonger period of time than cells carrying nearly identicalura3 alleles (strain OI29). The persistence of the DSB inthe cells could generate higher amounts of the arrestingsignal, preventing 60% of the cells of strain OI31 fromresuming growth after the DSB repair is completed.

This, however, does not seem to be the case, since in astrain carrying an alternative source of homology, re-covery is improved, without shortening the period oftime during which the chromosomes remain broken.

We therefore propose that the low recovery seen instrain OI31 is due to a sensing mechanism that operatesduring the process of recombinational repair. In contrastto the study discussed above (Lee et al. 1998), which wasperformed under conditions that did not allow the DSBto be repaired, in the present work the broken chro-mosome was e�ciently repaired by recombination. Wesuggest that during the recombinational process a signalis generated that is responsible for the resumption of cellgrowth and division. The presence of sequence diver-gence between the interacting DNA molecules a�ects thelevel of this signal, causing more cells to stay arrested,even if the repair of the DSB is eventually completed.The sequence divergence could be sensed as mismatchesin short heteroduplex DNA regions, such as those ex-pected to be created during the invasion or the annealingsteps in the recombination process. The mismatchsensing between the interacting DNA molecules is notdependent on the function of the MMR proteins; theidentity of the sensing mechanism is still to be estab-lished. It should be noted that about 20% of the cells ofstrain OI29 are unable to resume growth after they havesuccessfully repaired the broken chromosome (Table 1).The three heterologies present in the ura3 allele in OI29may be responsible for this e�ect.

Our results suggest that recombination is monitoredduring the repair itself, and not after it is completed. Inthis way, while the interacting molecules are aligned, it ispossible to distinguish between identical information(such as that present in the homologous chromosome)and divergent genetic information located elsewhere inthe genome (e.g. pseudogenes). In the latter case, geneconversion might lead to gene inactivation with delete-rious consequences. The prolonged cell-cycle arrest mayallow an additional round of homology search that canlead to the ®nding of a more suitable partner. An examplefor such a process can be seen in the strains carrying analternative source of homology: recombination with theidentical sequences restored a high viability level.

Acknowledgements We thank J. Nickolo�, S. Jinks-Robertson andR. Kolodner for the generous gift of plasmids. This work wassupported by a grant of the Israel Science Foundation to M.K. O.I.was partially supported from a scholarship from the ConstantinerInstitute for Molecular Genetics.

References

Alani E, Reenan RAG, Kolodner RD (1994) Interaction betweenmismatch repair and genetic recombination in Saccharomycescerevisiae. Genetics 137: 19±39

Bennett CB, Lewis AL, Baldwin KK, Resnick MA (1993) Lethalityinduced by a single site-speci®c double-strand break in a dis-pensable yeast plasmid. Proc Natl Acad Sci USA 90: 5613±5617

Bogue M, Roth DB (1996) Mechanism of V(D)J recombination.Curr Opin Immunol 8: 175±180

31

Chen W, Jinks-Robertson S (1999) The role of the mismatch repairmachinery in regulating mitotic and meiotic recombinationbetween diverged sequences in yeast. Genetics 151: 1299±1313

Datta A, Adjiri A, New L, Crouse GF, Jinks-Robertson S (1996)Mitotic crossovers between diverged sequences are regulated bymismatch repair proteins in Saccharomyces cerevisiae. Mol CellBiol 16: 1085±1093

Datta A, Hendrix M, Lipsitch M, Jinks-Robertson S (1997) Dualroles for DNA sequence identity and the mismatch repair sys-tem in the regulation of mitotic crossing-over in yeast. ProcNatl Acad Sci USA 94: 9757±9762

Elledge SJ (1996) Cell-cycle checkpoints: preventing an identitycrisis. Science 274: 1664±1672

Inbar O, Kupiec M (1999) Homology search and choice ofhomologous partner during mitotic recombination. Mol CellBiol 19: 4134±4142

Kolodner R (1996) Biochemistry and genetics of eukaryotic mis-match repair. Genes Dev 10: 1433±1442

Lee SE, Moore JK, Holmes A, Umezu K, Kolodner RD, Haber JE(1998) Saccharomyces Ku70, mre11/rad50 and RPA proteinsregulate adaptation to G2/M arrest after DNA damage. Cell 94:399±409

Moore JK, Haber JE (1996) Cell-cycle and genetic requirements oftwo pathways of nonhomologous end-joining repair of double-strand breaks in Saccharomyces cerevisiae. Mol Cell Biol 16:2164±2173

Negritto MT, Wu X, Kuo T, Chu S, Bailis AM (1997) In¯uence ofDNA sequence identity on e�ciency of targeted gene replace-ment. Mol Cell Biol 17: 278±286

Nelson HH, Sweetser DB, Nickolo� J (1996) E�ects of terminalnonhomology and homeology on double-strand-break-inducedgene-conversion tract directionality.Mol Cell Biol 16: 2951±2957

Nicholson A, Hendrix M, Jinks-Robertson S, Crouse GF (2000)Regulation of mitotic homeologous recombination in yeast:functions of mismatch repair and nucleotide-excision repairgenes. Genetics 154: 133±146

Nickolo� JA, Singer JD, He�ron F (1990) In vivo analysis of theSaccharomyces cerevisiae HO nuclease recognition site by site-directed mutagenesis. Mol Cell Biol 10: 1174±1179

Padmore R, Cao L, Kleckner N (1991) Temporal comparison ofrecombination and synaptonemal complex formation duringmeiosis in S. cerevisiae. Cell 66: 1239±1256

Paques F, Haber JE (1999) Multiple pathways of recombinationinduced by double-strand breaks in Saccharomyces cerevisiae.Microbiol Mol Biol Rev 63: 349±404

Petes TD, Malone RE, Symington LS (1991) Recombination inyeast. In: Broach JR, Pringle JR, Jones EW (eds) Molecularand cellular biology of the yeast Saccharomyces cerevisiae. ColdSpring Harbor Lab Press, Plainview, N.Y., pp 407±521

Porter G, Westmoreland J, Priebe S, Resnick M (1996) Homolo-gous and homeologous intermolecular gene conversion are notdi�erentially a�ected by mutations in DNA-damage or mis-match repair genes RAD1, RAD50, RAD51, RAD52, RAD54,PMS1 and MSH2. Genetics 143: 755±767

Priebe SD, Westmoreland J, Nilson-Tillgren T, Resnick M (1994)Induction of recombination between homologous and divergedDNAs by double-strand gaps and breaks and role of mismatchrepair. Mol Cell Biol 14: 4802±4814

Sandell LL, Zakian VA (1993) Loss of a yeast telomere: arrest,recovery, and chromosome loss. Cell 75: 729±739

Schiestl RH, Igarashi S, Hastings PJ (1988) Analysis of themechanism for reversion of a disrupted gene. Genetics 119:237±247

Selva EM, Maderazo AB, Lahue RS (1997) Di�erential e�ects ofthe mismatch repair genesMSH2 andMSH3 on homoeologousrecombination in Saccharomyces cerevisiae. Mol Gen Genet257: 71±82

Sherman F, Fink GR, Hicks JB (1986) Methods in yeast genetics.Cold Spring Harbor Laboratory Press, Plainview, N.Y.

Sugawara N, Paques F, Colaiacovo M, Haber JE (1997) Role ofMsh2 and Msh3 repair proteins in double-strand break inducedrecombination. Proc Natl Acad Sci USA 94: 9214±9219

Sweetser DB, Hough H, Whelden JF, Arbuckle M, Nickolo� JA(1994) Fine-resolution mapping of spontaneous and double-strand break-induced gene conversion tracts in Saccharomycescerevisiae reveals reversible mitotic conversion polarity. MolCell Biol 14: 3863±3875

Szostak JW, Orr-Weaver TL, Rothstein R, Stahl FW (1983) Thedouble-strand break repair model for recombination. Cell 3:725±738

Toczyski DP, Galgoczy DJ, Hartwell LH (1997) CDC5 and CKIIcontrol adaptation to the yeast DNA damage checkpoint. Cell90: 1097±1106

White CI, Haber JE (1990) Intermediates of recombination duringmating-type switching in Saccharomyces cerevisiae. EMBO J 9:663±674

Zenvirth D, Arbel T, Sherman A, Goldway M, Klein S, Simchen G(1992) Multiple sites for double-strand breaks in whole meioticchromosomes of Saccharomyces cerevisiae. EMBO J 11: 3441±3447

32