Proc. Nati. Acad. Sci. USAVol. 88, pp. 7585-7589, September 1991Genetics

Integration of DNA fragments by illegitimate recombination inSaccharomyces cerevisiae

(yeast/nonhomologous recomblnation/restrction enzymes)

ROBERT H. SCHIESTL AND THOMAS D. PETESDepartment of Biology, University of North Carolina, Chapel Hill, NC 27599

Communicated by David Botstein, May 17, 1991

ABSTRACT DNA fragments (generated by BamHI treat-ment) with no homology to the yeast genome were transformedinto Saccharomyces cerevisiae. When the fragments were trans-formed in the presence of the BamfH enzyme, they integratedinto genomic BamHI sites. When the fragments were trans-formed in the absence of the enzyme, they integrated intogenomic G-A-T-C sites. Since the G-A-T-C sequence is presentat the ends ofBamiI fragments, this result indicates that fourbase pairs of homology are sufficient for some types of mitoticrecombination.

Illegitimate recombination events join two DNA molecules(or two noncontiguous parts of a single DNA molecule)without the requirement for extended sequence homology. Ina number of human genetic defects caused by genomicrearrangements (1-6), either no homology or very limitedhomology has been detected at the junctions of the rear-rangements (7, 8). Similar recombination events have alsobeen characterized in bacteria (9).

In most organisms, nonhomologous integration of intro-duced DNA is more common than homologous integration.For example, in mammalian cells, only one in one thousandtransformants contains the transforming DNA in a homolo-gous position (10). In contrast, in Saccharomyces cerevisiae,all transformants analyzed contained the transforming DNAin homologous positions (11, 12). One interpretation of theseresults is that Saccharomyces cerevisiae has a very efficientmechanism of homologous recombination (integration) andan inefficient mechanism of nonhomologous recombination.Alternatively, it is possible that yeast has no mechanismallowing nonhomologous integration of transforming DNA.Below, we describe evidence indicating that the first of thesealternatives is correct. In addition, we show that certainrestriction enzymes can be transformed into yeast cells, andcatalyze the integration of transforming DNA sequences.

MATERIALS AND METHODSPlasmids. The plasmid pJL202 was constructed by Joachim

Li (The Johns Hopkins University) and obtained from JefBoeke (The Johns Hopkins University). In this plasmid, the1.1-kilobase (kb) HindIII URA3 fragment was replaced by theHIS3 gene, leaving sequences homologous to the URA3flanking region on both sides of the HIS3 gene.The URA3 fragment used in our transformation experi-

ments was derived from the plasmid pM20 constructed bySue Jinks-Robertson (Emory University) and Martin Kupiec(Tel Aviv University). This plasmid was constructed by"filling-in" the cohesive ends of the 1.1-kb HindIII fragmentcontaining the URA3 gene and inserting this fragment into theHincIl site in the polylinker of pUC7. Thus, the URA3 gene

in pM20 is flanked by BamHI sites and (outside ofthe BamHIsites) by EcoRI sites.

Strains and Media. The yeast strain OD5 (MATa leu2-3,112his3-11,15) was obtained from Walter Spevak (University ofVienna). An isogenic strain (RSY12) lacking URA3 se-quences was constructed by replacement of the entire openreading frame of URA3 (13) with the HIS3 gene. PlasmidpJL202 (described above) was digested with Xho I and NotI and strain OD5 was transformed to histidine prototrophy.The complete deletion of URAS sequences in strain RSY12was verified by Southern analysis.Growth, minimal medium, and sporulation medium were

prepared as described (14, 15).Genetic and Molecular Techniques. Standard genetic tech-

niques for sporulation, dissection of asci, and determinationof phenotypes (15) and standard methods for molecularbiology (16) were used except as noted below. Small-scaleplasmid preparations from Escherichia coli were carried outas a modification of the boiling method (17). A rapid proce-dure for the preparation of small amounts of high molecularweight yeast DNA was used (18).The yeast transformation procedure was that described by

Schiestl and Gietz (19). Cells from an overnight culture wereresuspended in 300 ml ofYPAD medium (15) and grown to adensity ofabout 5 x 106 cells per ml. The cells were harvestedby centrifugation and resuspended in 10 ml of sterile distilledwater. The washed cells were harvested by centrifugationand resuspended in 1.5 ml of sterile TE/LiOAc (prepared bydilution of 1Ox concentrated stocks: 10x TE = 0.1 MTris-HC/0.01 M EDTA, pH 7.5 and 1Ox LiOAc = 1 MLiOAc adjusted to pH 7.5 with dilute acetic acid). The cellswere then incubated for 1 hr at 30'C with constant agitation.About 50 Al of plasmid DNA solution prepared as describedbelow and 20 p.l of a solution of denatured salmon spermDNA (10 mg/ml) were added to a Microfuge tube containing200 ,.1 of cells in TE/LiOAc. In experiments 2-4, an aliquotof 10x concentrated stock of restriction enzyme buffer(described below) sufficient to yield a final 1 x concentrationwas also added. The resulting mixture was incubated for 30min at 30'C with constant agitation. Sterile 40%o PEG 4000(1.2 ml of40%6 polyethylene glycol 4000 in lx TE/l x LiOAc)was then added, and the cells were incubated for 30 min at30'C with agitation. The suspension was transferred to 420Cfor 15 min and then spun in a Microfuge for 5 sec. Cells werewashed once with 0.5 ml ofTE and resuspended in 1 ml ofTE;200 A.l of the suspension was plated on each selection plate.Plates were incubated at 30'C until colonies appeared.For experiment 1 (Table 1; BamHI present in the original

transforming solution), 20 pug of CsCl-purified plasmid pM20was treated with 100 units of BamHI (obtained from eitherNew England Biolabs or Promega) in 150 mM NaCl/10 mMTris HC, pH 7.9/10 mM MgCl2/1 mM dithiothreitol/100 pugof bovine serum albumin per ml (volume = 50 pl). After theDNA was completely digested, we added 5 p.l of 10x TEbuffer and 5 ,1 of 10x LiOAc buffer. This solution was thenadded to the yeast cells as described above. For experiments

7585

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

7586 Genetics: Schiestl and Petes

Table 1. Frequency of restriction enzyme-mediated(REM) events

Ratio of REM events to total transformants*

BamHI BamHI BamHI removedExp. presentt absent* and readded§

1 29/302 3/10 0/9 10/173 8/10 0/10 17/194 2/20 7/10

Totals 40/50 2/39 34/46*DNA was isolated from yeast transformants, treated with BamHI,and examined by Southern analysis. Transformants that had aURA3 BamHI fragment of 1.2 kb were considered to be the resultof a REM event.tThe BamHI enzyme present in the original digest was included withthe DNA during transformation.*The BamHI enzyme was removed by proteinase K/phenol extrac-tion (experiments 2 and 3) or by heat treatment (experiment 4) priorto transformation.§The BamHI enzyme was removed by proteinase K/phenol extrac-tion or by heat treatment. The enzyme was then readded to thepurified BamHI fragments prior to transformation.

2 and 3, the amount of DNA and volume of the digestionbuffer were scaled up by a factor of 3. After the plasmidDNAwas digested, one-third ofthe sample was removed to be usedin the transformation. The remaining two-thirds was precip-itated with ethanol and resuspended in 200 Al of 0.01 MTris'HCl, pH 7.5/0.05 M EDTA/1% SDS/100 ,ug of protein-ase K per ml. After a 30-min incubation at 370C, the samplewas extracted twice with phenol/chloroform/isoamyl alco-hol, 25:24:1 (vol/vol), precipitated with ethanol, washed with70% ethanol, and air-dried. The pellet was dissolved in water(40 pl). Halfwas added to a Microfuge tube containing 200 ,1of yeast cells in TE/LiOAc buffer, carrier (salmon sperm)DNA, 100 units of BamHI enzyme, and the appropriateconcentration of BamHI restriction buffer. The other halfwas added to an identical solution lacking the BamHI en-zyme. The remainder of the transformation procedure wasperformed as described above. In experiment 4, after diges-tion ofthe plasmid DNA, the BamHI enzyme was inactivatedby heat treatment (70TC for 10 min). The sample was then splitin half. One aliquot was used to transform yeast without anyfurther additions; BamHI (100 units) was added to the secondaliquot before transformation.From several of the yeast transformants with a URA3 gene

inserted at a nonhomologous site, the URA3 gene and itsflanking DNA sequences were recovered. Genomic DNAfrom the transformants was treated with Bgi II and theresulting fragments were inserted into the BamHI site of theplasmid YEplacl81 (20). The pool of recombinant plasmidswas transformed into the yeast strain RSY12, and Ura'colonies were selected (21). From the Ura' colonies, plasmidDNA was extracted (22) and used to transform E. coli toampicillin-resistance. The sequences flanking the URA3 in-sertions were determined by using 18-base-pair (bp) primershomologous to the insertion 68 bp from the 5' (relative toURA3)junction (5'-CGGAGATTACCGAATCAA) and 42 bpfrom the 3' junction (5'-GAATCTCGGTCGTAATGA). Se-quencing of double-stranded DNA was carried out withSequenase (23) and a Bio-Rad sequencing cell.The resulting sequence information was employed to de-

sign oligonucleotides that could be used to amplify "target"sites (the site without the insertion). DNA was isolated fromstrain OD5 and, in separate experiments, pairs of primersdesigned to produce a DNA fragment 150 bp in length wereused in the polymerase chain reaction (PCR method asdescribed in ref. 24). Following amplification, the resulting

fragments were sequenced directly (protocol of United StatesBiochemical).

RESULTSRestriction Enzyme-Mediated Integration Events. The plas-

mid pM20 contains the yeast URA3 gene on a 1.2-kb BamHIfragment inserted into a pUC7 vector. This plasmid wastreated with BamHI and transformed (19) in the presence ofthe BamHI enzyme into a haploid yeast strain (RSY12) thatlacks URA3 sequences. About five Ura' transformants per,gg of transforming DNA were obtained. Thirty were testedfor stability of the Ura' phenotype; all were stable (10-7frequency of Ura- derivatives). Fifteen of the transformantswere crossed to a ura3 strain of opposite mating type. Fivetetrads were dissected from each sporulated diploid. For all15 transformants examined, most ofthe five tetrads examinedsegregated 2+:2-. This segregation pattern indicates that thewild-type URA3 gene is integrated into a single genomic sitein each transformant. The same 15 transformants werecrossed to a haploid strain with a wild-type URA3 genelocated at its normal location (chromosome V). Each diploidstrain segregated 4+:0-, 3+:1-, and 2+:2- tetrads in the ratiosexpected for unlinked markers. Spore viability was excellent(>90%o), indicating that these transformants did not containchromosomal translocations or large inversions.DNA was isolated from the transformants and examined by

Southern analysis. The hybridization probe was pM20. Whenthe DNA was treated with EcoRI (which does not cut withinthe 1.2-kb URA3 fragment), we found that most transform-ants contained a single strong band of hybridization that wasdifferent in different transformants (Fig. 1). The URA3 genein OD5 (a strain isogenic with RSY12 that contains URA3 atits normal location) was located on an EcoPJ fragment of 13kb, as expected from previous studies (25). The URA3deletion strain RSY12 lacked this band. The observation thatmost of the transformants contain a single strong band ofhybridization at different positions suggests that most of thetransformants have a single URA3 insertion and that theposition of the insertions in the genome is different indifferent transformants.

r r - --

FIG. 1. Southern blot of DNA from the yeast strain OD5, anisogenic derivative of OD5 that lacks URA3 (RSY12), and Ura+derivatives of RSY12 (B1-B5, S1-S5, T1-T5, T9, and T10) obtainedby transformation with the URA3 BamHI fragment (in the presenceof the BamHI enzyme); these transformants were from experiment1 in Table 1. DNA from all strains was treated with EcoRI (whichdoes not cut within the URA3 BamHI fragment) and hybridized witha 32P-labeled plasmid (pM20) containing URA3. The strain OD5contains a single 13-kb fragment characteristic of the URA3 gene atits proper position on chromosome V. Strain RSY12 does not containany strongly hybridizing fragment, and the Ura+ transformants showfragments of different sizes that hybridize to the URA3 probe,indicating that the URA3 fragment integrated into different genomicpositions in different transformants.

Proc. Natl. Acad. Sci. USA 88 (1991)

Proc. Natl. Acad. Sci. USA 88 (1991) 7587

When the DNA of 30 transformants was digested withBamHI and Southern hybridization was carried out, most (29of 30) had a single fragment of about 1.2 kb that hybridizedto URA3 (Fig. 2). This result was unexpected because itindicates that the BamHI sites (G-G-A-T-C-C) at both junc-tions of the integrated URA3 gene were recreated uponinsertion of the transforming fragment into the genome. Todetermine the mechanism of integration, we sequenced thejunctions of nine different URA3 integrated fragments andobtained the "target" sequences (the genomic sequence priorto the integration event) for four of these (Fig. 3). Approxi-mately 200 bp of genomic sequence was obtained for eachjunction and the GenBank data base was searched for ho-mology. The genomic sequences in pRS137 [representing theinsertion in RSY12(B5)] were identical to those of the SGAIgene (26). This "target" gene had aBamHI site at the positionof insertion of the URA3 fragment.To test for the generality of this unexpected result, we

sequenced three other target sites. We used the flankingsequences to design pairs of oligonucleotides with the ap-propriate sequences to yield aDNA fragment of about 150 bpcontaining the "target" after application of the polymerasechain reaction (PCR). Using these primers, we amplified thegenomic target sites representing the insertions in the plas-mids pRS124, pRS125, and pRS126. Sequence analysis indi-cated that all three target sites contained a BamHI site at theposition of insertion of the URA3 fragment. In summary, infour offour transformants, the insertion ofthe BamHI URA3fragment into the genome had the structure predicted for aconservative integration of the fragment into aBamHI site inthe genome.To determine whether these integration events were cat-

alyzed by BamHI, we examined the transformation proper-ties of the URA3 fragment in the presence and the absence ofBamHI. The efficiency of transformation was about 7-foldhigher in the presence of the enzyme than in its absence (101transformants versus 15 transformants; 20 pug ofDNA in eachtransformation). The cell viability after transformation wasunaffected by the presence of the enzyme. DNA sampleswere prepared from the transformants, treated with BamHI,and examined by Southern analysis. The frequencies oftransformants with URA3 genes with two flanking BamHIsites (see Table 1) were: 80% (40 of 50) for BamHI URA3fragments in the original BamHI digestion buffer (BamHI

~~~- -> -X~ rZ r- ' -

FIG. 2. Southern blot of DNA from yeast strains OD5, RSY12(lacking URA3 sequences), and Ura+ transformants RSY12 (S6-SI,T1-T10) obtained with the URA3 BamHI fragment in the presence ofBamHI enzyme (experiment 1 in Table 1). The genomic DNA wastreated with BamHI and hybridized with 32P-labeled plasmid pM20(containing URA3). The DNA of most of the transformants hybrid-ized to a single fragment 1.2 kb in size, the same size as the originalBamHI fragment used in transformation. The size of the genomicBamHI URA3 fragment in OD5 is about 5.5 kb, as expected (25).

pRS124 CCGGTGGATGTGATG GATCCGTTTAATGCTGATCRSY12(Bl) GGCCACCTACACTACCTAG GCAAATTACGACTAG

pRS125 TTTACGGGATTTATG GATCCGTCAGAACAAAAAGRSY12(B2) AAATGCCCTAAATACCTAG GCAGTCTTGTTTTTC

pRS126 TTCATAGGAAAATGGRSY12(T3) AAGTATCCTTTTACCCTAG

GATCCATCCATCCCAACAAGTAGGTAGGGTTGTT

pRS137 CATCTCTTGGCTCTG GATCCGTTATCTGTTCTGTRSY12(B5) GTAGAGAACCGAGACCTAG GCAATAGACAAGACA

pRS138 GGCTTCTCTTTTCCGRSY12(B7) CCGAAGAGAAAAGGCCTAG

pRS139 CTCCGGGAGCGGGCGRSY12(S4) GAGGCCCTCGCCCGCCTAG

GATCCGTGATTGCGGTACGGCACTAACGCCATGC

GGTGATTGCCGGGTC

pRS 140 TGTTTTTTCACCTCG GATCCGCCCGACAATATCTRSY12(85) ACAAAAAAGTGGAGCCTAG GCGGGCTGTTATAGA

pRS141 GGAAAGAAAATTTCG GATCCGAAAATTTTTGGCGRSY12(T5) CCTTTCTTTTAAAGCCTAG GCTTTTAAAAACCGC

pRS142 GTGGCTACAGAATGG GATCCTGAATTTATAACTARSY12(T6) CACCGATGTCTTACCCTAG GACTTAAATATTGAT

FIG. 3. DNA sequences from both junctions of the integratedBamHI URA3 fragments in nine different Ura+ transformants. Theintegrated URA3 genes from nine restriction enzyme-mediated trans-formants were cloned and sequenced. The letters and numbersdenoting the transformants are indicated below the name of theplasmid containing the integrated gene. Primers homologous tosequences within the insert were used to obtain the sequence of theflanking genomic DNA. The boxed region represents sequences ofthe URA3 fragment (the 5' end of the gene is on the left side). For theflanking sequences, the gap represents the position ofinsertion oftheURA3 gene. Although only 15 bp offlanking genomic DNA is shown,we obtained about 200 bp of flanking sequences from each junction.In the sequence of the URA3 fragment, 5 bp flanking the restrictionsite are underlined; in the flanking DNA sequences, bases homolo-gous to these 5 bp are underlined. As discussed in the text, there issignificant homology between the bases at the left side of the URA3fragment and the right junction of the flanking sequences.

present), 5% (2 of 39) for BamHI URA3 fragments in theabsence of BamHI enzyme, and 74% (34 of 46) for BamHIURA3 fragments with the BamHI enzyme removed and thenadded back. These results indicate that the integration of theBamHI URA3 fragment into BamHI sites in the genome iscatalyzed by the BamHI restriction enzyme, a restrictionenzyme-mediated event.The simplest explanation of the restriction enzyme-

mediated event is that the BamHI enzyme enters the cells,perhaps in association with the transforming DNA. Theenzyme cleaves the chromosomal DNA at BamHI sites, thecohesive ends of the URA3 fragment pair with those of thechromosomal DNA, and the resulting nicks are sealed byDNA ligase (Fig. 4a). An observation that complicates thissimple model is that sequences adjacent to the BamHI site atthe right junction (Fig. 3) appear nonrandomly related tosequences at the left end (the 5' end of URA3) of the BamHIURA3 fragment. Beginning after the terminal C in the BamHIsite at the right junction, the loose consensus sequence is: G

Genetics: Schiestl and Petes

7588 Genetics: Schiestl and Petes

(6 of 9 inserts), T (5 of 9), C (4 of 9), and A (5 of 9). The samefour bases are present adjacent to the BamHI site at the left

aGATCC URA3 G

G URA3CCTAG

CCTAQG_

b

1

2

URA3

KZ)ATC

I~CA

3

I

end of the URA3 fragment (Fig. 3). The plasmid pRS125 hasthe complete consensus at the 3' junction plus one additionalbase in common with the URA3 fragment. The likelihood ofa random occurrence of this sequence (G-T-C-A-G) at the 3'junction (calculated based on the 39%o G+C content of yeastDNA) is (0.2)3(0.3)2 or 0.00072; the probability of observingthis sequence in one of nine transformants is 0.0065.

Integration Events in the Absence of BamHI: legitimateIntegration. As described above, we found that the integra-tion events of the BamHI URA3 fragments that occurred inthe absence of the enzyme usually did not regenerate BamHIsites flanking the insertion (Table 1). To determine the natureof these enzyme-independent integration events, we se-quenced three of the integrated fragments. We found that theG-A-T-C sequence at the ends of the URA3 fragment wasconserved; however, there was a BamHI site at only one ofthe six junctions (Fig. 5). When the target sites for two ofthese integration events were determined, we observed thatthe BamHI URA3 fragment had integrated into G-A-T-Csequences in the chromosome. Since the G-A-T-C targetsequence is the same as that of the cohesive end in BamHI-treated DNA, these integration events have the structureexpected for a conservative insertion of the BamHI URA3fragment into genomic DNA with a staggered cleavage of aG-A-T-C site. These results indicate the existence of arecombination mechanism in yeast that utilizes extremelysmall (4 bp) sequence homologies (see Fig. 4b). Since re-combination events involving little or no sequence homologywere defined as "illegitimate" recombination (27), we callthis type of event "illegitimate" integration.

DISCUSSIONRestriction Enzyme-Mediated Events. The observation of

restriction enzyme-mediated events indicates that theBamHI restriction enzyme can enter the cell under theconditions used in the yeast transformation protocol. Al-though the entry of restriction enzymes into S. cerevisiaecells in vivo has not been described previously to our knowl-edge, it has been shown that, when the gene encoding EcoRIwas expressed in yeast, the chromosomalDNA was digested,resulting in cell death (28). A number of studies indicate thatrestriction enzymes can enter mammalian cells that had beenmade permeable by electroporation (29). These enzymes

4

FIG. 4. Models for restriction enzyme-mediated (REM) (a) andillegitimate (b) integration events in yeast. (a)REM events. TheBamHIenzyme enters the cell together with the transforming URA3 BamHIDNA fragments. As discussed in the text, there is likely to be ahomology search involving the 5' end of the URA3 fragment, followedby a metastable association between the chromosome and the trans-formingfragment. Following this association, the chromosomal DNA iscleaved by BamHI (which is possibly associated with the transformingDNA), and the cohesive ends ofthe URA3 BamHI fragment are ligatedto the chromosomal BamHI ends. Other related mechanisms are alsopossible. For example, it is possible that the transforming DNAassociates with the chromosomalDNA after the genomic sequences arecut by BamHI. The net result ofthe REM event is an integrated URA3gene flanked by BamHI sites. (b) Illegitimate integration. The BamHIURA3 fragment enters the cell in the absence of BamHI. The 5'single-stranded G-A-T-C ends of the URA3 fragment (indicated by thinlines and thin letters) invade the chromosome (indicated by thick linesand bold letters), pairing at chromosomal G-A-T-C sites (step 2). Thechromosomal DNA is nicked at two sites (indicated by arrows). Theresulting 3' strands derived from the chromosome are ligated to the 5'strands derived from the URA3 fragment (step 3). The net result of thismechanism is an integrated URA3 sequence flanked by G-A-T-C-C andG-G-A-T-C sequences (step 4).

pRS159 CCAACGATAATAACG GATCAACTTTCAATTCCAARSY12(pM20 GGTTGCTATTATTGCCTAG TTGAAAGTTAAGGTTBIIB7)

pRS 161RSY12(pM2OBIIB9)

pRS162RSY12(pM20EE7)

TACCTACA&TTAAAT GATCACAATjTATGGATGCATGGATGTTAATTTACTAG TGTTACATACCTACG

GAAGCTGAACTAAAA GATCTAAATCAGATGCTGCCTTCGACTTGATTTTCTAG ATTTAGTCTACGACG

FIG. 5. DNA sequences from both junctions of illegitimate inte-gration events of the BamHI URA3 fragment. The BamHI URA3fragment was transformed into RSY12 in the absence of BamHIenzyme. The integrated URA3 gene and flanking DNA sequenceswere cloned from three transformants, and their sequences wereanalyzed. These sequences are depicted in the same way as those inFig. 3. The target sites for the first two insertions were analyzed andfound to contain a G-A-T-C sequence at the point of insertion. Thesesequences indicate a conservative integration ofthe BamHI fragmentinto a genomic G-A-T-C sequence.

Proc. Natl. Acad. Sci. USA 88 (1991)

Proc. Natl. Acad. Sci. USA 88 (1991) 7589

reduced the viability of the cells and induced chromosomalaberrations. W. Morgan (personal communication) has ob-served deletions and rearrangements of restriction fragmentscontained within plasmids in mammalian cells as a result ofthe enzyme treatment. In E. coli, it has been shown thatexpression ofEcoRI can promote site-specific recombinationin vivo (30).The simplest explanation for restriction enzyme-mediated

events is that the chromosomal DNA is broken at manyBamHI sites by the BamHI enzyme. Although most of thesebreaks are religated without incorporation of transformingDNA, occasionally the BamHI URA3 gene is inserted intothe chromosomal site. Since a single unrepaired chromosomebreak would be lethal, this mechanism would require yeast tobe very proficient in the repair of BamHI-induced breaks,since no decrease in viability was found after transformationin the presence of BamHI. Potential uses of restrictionenzyme-mediated events include: insertional mutagenesis,generation ofchromosome aberrations, stimulation of mitoticrecombination at specific sites, and improvements in theefficiency of gene targeting in organisms with high levels ofnonhomologous integration. In addition, since there is noreason to assume that the uptake of proteins during trans-formation is limited to restriction enzymes, one might be ableto observe the effects of other proteins (topoisomerases,helicases, etc.) on recombination in vivo.

Illegitimate Integration Events. Previous transformationstudies in yeast indicated that most integration events werethe result of recombination involving more than severalhundred base pairs of sequence homology between the trans-forming DNA and the yeast chromosome (11, 12). It has beenfound that integration was not detectable when the trans-forming plasmids had only 31 and 41 bp of homology to thetarget chromosomal locus (31). A low frequency (<10 trans-formants per 85 ,ug) transformation at the CYCI locus with a20-bp oligonucleotide has been observed (32), although notransformants were detected when using 10- or 15-bp oligo-nucleotides.Although previous investigators have shown that repeats

of about 0.2-2 kb recombine efficiently in mitosis (33), thelower limit of homologous recombination in yeast has notbeen defined. In experiments involving duplicated sequenceson a yeast plasmid, a frequency of recombination of 10-10 fora 26-bp repeat has been found (34). From these results,recombination events involving 4 bp of homology would notbe expected. A low frequency (10-8 per mitotic division) ofprecise excision of a TnS element (flanked by 9-bp repeats)from multicopy yeast plasmids has been detected (35). Oneway of rationalizing the rarity of recombination events in-volving small repeated sequences with the ready detection ofillegitimate integration events is that the yeast recombinationmachinery does not function efficiently with very shortregions of homology unless the homology is located at theends of one of the substrate molecules.Recombination events, as described in various recombi-

nation models, often initiate by the invasion ofa single-strandof DNA derived from one chromosome into the duplex of ahomologous chromosome. In the Meselson-Radding model(36), the 5' end of one strand invades the homologous duplex,whereas in the original version of the double-strand break-repair model, there is invasion ofthe 3' end ofone strand (37).Our results indicate that 5' ends are capable of invading aDNA duplex.

In E. coli, it is not clear whether deletions at short repeatsequences are the result of a RecA-independent recombina-tion pathway, a consequence of DNA polymerase slippageduring replication, or a mechanism involving the activity ofDNA gyrase (9). The illegitimate integration observed in ourexperiments is clearly the consequence of recombination. A

second type of evidence for mechanisms recognizing shortregions of homology is the result of transformation studies(7). It has been found, for example, that when plasmids thatwere linearized in vitro to generate noncompatible ends weretransformed into mammalian cells, the ends rejoined bymechanisms sensitive to very small regions ofhomology (4 bpor less) (38). In conclusion, a fairly diverse collection ofrecombination events involving very small regions of se-quence homology has been detected in a diverse collection oforganisms. Mutational studies that can be easily carried outin yeast are likely to be useful in clarifying the relationshipsbetween different classes of events.

We thank Margareth Dominska for excellent and cheerful techni-cal assistance. We thank B. Byers and J. Boeke for suggestionsconcerning the restriction enzyme-mediated mechanism; D. Gorde-nin, T. Orr-Weaver, J. Gafner, and W. Morgan for providingunpublished information; and members of the Petes laboratory forcomments on the manuscript. We thank R. D. Gietz for collaboratingwith us on relevant preliminary studies. This research was supportedby National Institutes of Health Grant GM 24110 and AmericanCancer Society Grant NP 712.

1. Cairns, J. (1981) Nature (London) 289, 353-357.2. Haluska, F. G., Tsujimoto, Y. & Croce, C. M. (1987) Annu. Rev. Genet.

21, 321-345.3. Klein, G. (1981) Nature (London) 294, 313-318.4. Marx, J. L. (1982) Science 218, 983-985.5. Pall, M. L. (1981) Proc. Natl. Acad. Sci. USA 78, 2465-2468.6. Wintersberger, U. (1982) Naturwissenschaften 69, 107-113.7. Meuth, M. (1989) in Mobile DNA, eds. Berg, D. E. & Howe, M. (Am.

Soc. Microbiol., Washington), pp. 833-860.8. Murnane, J. P. (1990) BioAssays 12, 1-5.9. Ehrlich, S. D. (1989) in Mobile DNA, eds. Berg, D. E. & Howe, M. (Am.

Soc. Microbiol., Washington), pp. 799-832.10. Capecchi, M. R. (1989) Science 244, 1288-1292.11. Hinnen, A., Hicks, J. B. & Fink, G. R. (1978) Proc. NatI. Acad. Sci.

USA 75, 1929-1933.12. Hicks, J. B., Hinnen, A. & Fink, G. R. (1978) Cold Spring Harbor Symp.

Quant. Biol. 43, 1305-1313.13. Rose, M., Grisafi, P. & Botstein, D. (1984) Gene 29, 113-124.14. Schiestl, R. H. & Wintersberger, U. (1982) Mol. Gen. Genet. 186,

512-517.15. Sherman, F., Fink, G. R. & Hicks, J. B. (1986) Methods in Yeast

Genetics: A Laboratory Manual (Cold Spring Harbor Lab., Cold SpringHarbor, NY).

16. Maniatis, T., Fritsch, E. F. & Sambrook, J. (1982) Molecular Cloning:ALaboratory Manual (Cold Spring Harbor Lab., Cold Spring Harbor, NY).

17. Holmes, D. S. & Quigley, M. (1981) Anal. Biochem. 114, 193-197.18. Denis, C. L. & Young, E. T. (1983) Mol. Cell. Biol. 3, 360-370.19. Schiestl, R. H. & Gietz, R. D. (1989) Curr. Genet. 16, 339-346.20. Gietz, R. D. & Sugino, A. (1988) Gene 74, 527-534.21. Gietz, R. D. & Schiestl, R. (1991) Yeast 7, 253-263.22. Hoffman, C. S. & Winston, F. (1987) Gene 57, 267-272.23. Kraft, R., Tardiff, J., Krauter, K. S. & Leinwand, L. A. (1988) Biotech-

niques 6, 544-545.24. Innis, M. A., Gelfand, D. H., Sninsky, J. J. & White, T. J. (1990) PCR

Protocols:A Guide to Methods andApplications (Academic, New York).25. Judd, S. R. & Petes, T. (1988) Genetics 118, 401-410.26. Yamashita, I., Nakamura, M. & Fukui, S. (1987) J. Bacteriol. 169,

2142-2149.27. Franklin, N. (1971) in The Bacteriophage Lambda, ed. Hershey, A. D.

(Cold Spring Harbor Lab., Cold Spring Harbor, NY), pp. 175-194.28. Barnes, G. & Rine, J. (1985) Proc. Natl. Acad. Sci. USA 82, 1354-1358.29. Winegar, R. A., Phillips, J. A., Youngblom, J. H. & Morgan, W. F.

(1989) Mutat. Res. 225, 49-53.30. Chang, S. & Cohen, S. N. (1977) Proc. Natl. Acad. Sci. USA 74,

4811-4815.31. Orr-Weaver, T., Szostak, J. & Rothstein, R. (1983) Methods Enzymol.

101, 228-245.32. Moerschell, R. P., Tsunasawa, S. & Sherman, F. (1988) Proc. Natl.

Acad. Sci. USA 85, 524-528.33. Petes, T. D. & Hill, C. W. (1988) Annu. Rev. Genet. 22, 147-168.34. Ahn, B.-Y. & Livingston, D. M. (1986) Mol. Cell. Biol. 6, 3685-3693.35. Gordenin, D. A., Trofimova, M. V., Shaburova, 0. N., Pavlov, Y. I.,

Chernoff, Y. O., Chekuolene, Y. V., Proscyavichus, Y. Y., Sasnauskas,K. V. & Janulaitis, A. A. (1988) Mol. Gen. Genet. 213, 388-393.

36. Meselson, M. & Radding, C. (1975) Proc. Natl. Acad. Sci. USA 72,358-361.

37. Szostak, J., Orr-Weaver, T., Rothstein, R. & Stahl, F. (1983) Cell 33,25-35.

38. Roth, D. B. & Wilson, J. H. (1986) Mol. Cell. Biol. 6, 4295-4304.

Genetics: Schiestl and Petes