MOLECULAR AND CELLULAR BIOLOGY, Oct. 1991, p. 5372-5380 0270-7306/91/105372-09$02.00/0 Copyright C 1991, American Society for Microbiology

Vol. 11, No. 10

Heteroduplex Formation and Mismatch Repair of the "Stuck" Mutation during Mating-Type Switching in Saccharomyces cerevisiae BRYAN L. RAY, CHARLES

I.

WHITE,t

AND

JAMES E. HABER*

Rosenstiel Basic Medical Research Center and Department of Biology, Brandeis University, Waltham, Massachusetts 02254-9110 Received 4 February 1991/Accepted 30 July 1991

We sequenced two alleles of the MATa locus of Saccharomyces cerevisiae that reduce homothallic switching and confer viability to HO rad52 strains. Both the MATa-stk (J. E. Haber, W. T. Savage, S. M. Raposa, B. Weiffenbach, and L. B. Rowe, Proc. Natl. Acad. Sci. USA 77:2824-2828, 1980) and MATa-survivor (R. E. Malone and D. Hyman, Curr. Genet. 7:439-447, 1983) alleles result from a T-*A base change at position Zll of the MAT locus. These strains also contain identical base substitutions at HMRa, so that the mutation is reintroduced when MATa switches to MATa. Mating-type switching in a MATa-stk strain relative to a MATa Z1lT strain is reduced at least 50-fold but can be increased by expression of HO from a galactose-inducible promoter. We confirmed by Southern analysis that the Zl1A mutation reduced the efficiency of double-strand break formation compared with the Z1lT variant; the reduction was more severe in MATa than in MATa. In MATa, the Zl1A mutation also creates a matel (sterile) mutation that distinguishes switches of MATa-stk to either MATa or matkl-stk. Pedigree analysis of cells induced to switch in G, showed that MATa-stk switched frequently (23% of the time) to produce one matod-stk and one MATa progeny. This postswitching segregation suggests that Zil was often present in heteroduplex DNA that was not mismatch repaired. When mismatch repair was prevented by deletion of the PMS1 gene, there was an increase in the proportion of matal -stkIMATa sectors (59%) and in pairs of switched cells that both retained the stk mutation (27%). We conclude that at least one strand of DNA only 4 bp from the HO cut site is not degraded in most of the gene conversion events that accompany MAT switching. was able to survive without switching in HO radS2 genotypes. We cloned and sequenced these alleles to uncover the molecular basis for their unusual properties. We found that both the MATa-stk and MATa-survivor alleles contain identical T-*A changes in the Zll base pair of MAT and that the same mutation is present at HMRa in strains in which the mutant phenotype persists after several rounds of switching. This variant was previously found in HMRa and was presumed to be an innocuous polymorphism between MATa and MATa sequences (1). The precise mechanism by which MATa sequences are replaced by MATa is not fully understood, although it is clear that switching is a site-specific gene conversion event initiated by double-strand cleavage in MAT-Z by the HO endonuclease (33). More recent physical analysis of DNA in synchronized cells undergoing switching (42) has shown that the double-strand cut is processed by a 5'-to-3' exonuclease that leaves a long 3'-ended, single-stranded molecule in the Z region to the right of the cut at MAT. The equivalent Y region is somehow protected from such degradation in a donor-dependent fashion (42). This suggests that only the Z region is initially active in instigating gene conversion by strand invasion. Consistent with this idea was our demonstration that a covalent intermediate joining Ya to MAT-Z occurred 30 min before MAT-WX became ligated to Ya (42). Recently, McGill et al. (21) demonstrated that during switching, heteroduplex DNA, with one strand from the donor locus and one strand from the sequences initially at MAT, is sometimes formed in both the MAT-X and MAT-Z regions. This could be seen by pedigree analysis of the mother and daughter cells from a single switching event.

Mating-type switching in the yeast Saccharomyces cerevisiae involves replacement of a or a DNA sequences at the MAT locus by opposite mating-type sequences derived from one of two silent donor loci, HMRa and HMLax (Fig. 1). cis-acting mutations that reduce or abolish MAT switching have been obtained (9, 33, 38, 41). These mutations prevent efficient cleavage of the MAT locus by HO endonuclease, which initiates replacement of Ya or Ya by a gene conversion mechanism. Two types of mutations have been described. MATa-inc and MATa-inc (inconvertible) mutations are lesions within the HO recognition-cleavage site that are lost (healed) during rare switches to the opposite mating type because the double-strand break repair mechanism invariably replaces the region including these base pair changes. A second class of "stuck" mutations was described by Haber et al. (9). These mutants exhibited reduced switching from MATa to MATa but differed from inc mutants in that they were not healed in switching from MATa-stk to MATa (which switched normally) and back to MATa-stk (9). The MATa-stk mutations appeared to reduce the efficiency of HO endonuclease cleavage because HO MATa-stk strains carrying the rad52-1 mutation did not die, while wild-type MATa strains died when HO endonuclease-induced double-strand breaks were unable to be repaired. Another MATa allele has also been described as HO rad52 resistant. The MATa-survivor allele (18) was described as normal in its switching behavior in HO RADS2 strains, yet it * Corresponding author. t Present address: Institut des Sciences Vegetales, Centre Na-

tional de la Recherche Scientifique, 91198 Gif sur Yvette Cedex, France. 5372

MECHANISM OF MAT SWITCHING

VOL. 11, 1991 HMLa W

X

MATa

Zl Z2

Ya

HMRa

Ya

MATa FIG. 1. Mating type switching is initiated by a double-strand break catalyzed by the product of the HO gene. MATa interacts with HMLa, resulting in unilateral transfer of a information to MAT.

MAT switching normally occurs only in the G1 stage of the cell cycle, so that after DNA replication and cell division, the mother and its new daughter both switch to the same mating type (10). When the MATa locus was modified by restriction sites in the MAT-X or MAT-Z region that are not necessarily replaced along with the new Ya region, some of the MATa progeny exhibited postswitching segregation (PSS), in which one of the two progeny retains the restriction site and the other has the wild-type donor sequence. These PSS events presumably arise from subsequent DNA replication of a switched MAT locus that contains a region of heteroduplex DNA. The frequencies with which a large (-8 bp) XhoI insertion in MAT-X and a single-base-pair substi-

5373

tution in MAT-Z give rise to PSS are significantly different. McGill et al. (21) suggested that these differences probably reflect a difference in the way DNA strands are exchanged in these two regions. It is also possible that there is a difference in the way in which such different mismatched DNA substrates undergo correction. We investigated the fate of a single-base-pair substitution that is only 4 bp to the right of the HO endonuclease cleavage site, in both wild-type cells and an isogenic strain carrying a deletion of the PMSJ gene that is important for mismatch repair of single-base-pair substitutions in both meiotic (5, 43) and mitotic (2, 15) cells. We present pedigree studies that show that this base pair was usually present in heteroduplex DNA and that loss of the stk mutation occurred through efficient strand-specific repair of the heteroduplex. These results are significant in understanding both the mechanism of MAT switching and the mechanisms of mismatch repair in S. cerevisiae. MATERIALS AND METHODS

Strains. The genotypes of the strains used in this study are shown in Table 1. RY400 was constructed by replacing the MAT locus of NR238-23D with matA::LEU2 isolated from plasmid pJH124 (see below). Strain R175t contains the same matA: :LEU2 replacement and also contains a deletion of the HMR locus, hmra-3 (13). RY610, a pmslA derivative of RY530-2BT, was constructed as previously described (16), by using a pmslA derivative of pWBK4 (16) provided by S.

TABLE 1. S. cerevisiae strains used in this study Strain

Genotype'

Source

MH188-4B

HO HMLa MATa hmrA HO HMLa MATa-stk2 hmrA

M. Heam

MH185-1D

HO HMLa MATTa-stkl HMRa-stkl HO HMLa MATa-stkl HMRa-stkl

M. Heam

U11o RY510

HO HMLa-stkl MATa-stkl HMRa-stkl ho HMLa MATa-stk2 hmrA&

J. Haber This work

RY530

ho HMLa MATa-stk2 hmrA HO HMLa MA Ta HMRa

This work

RY530-2B RY530-2BT RY511

ho HMLa MATa-stk2 HMRa ho HMLa MATa-stk2 HMRa pGal-HO ho HMLa MATa-stk2 hmrA

This work This work This work

RY540

ho HMLa MATa HMRa HO HMLa MATa HMRa

This work

RY400 RY610 R175t

HO HMLa mataA::LEU2 HMRa ho HMLa MATa-stk2 HMRa pmslA pGal-HO ho HMLa mataA::LEU2 hmrA

This work This work J. Haber

NR226-7B

HO HMLa MA Ta HMRa HO HMLa MATa HMRa

N. Rudin

NR238-23D

HO HMLa mata-inc HMRa ho HMLa MATa-lethal HMRa rad52-1 ho HMLa MA Ta-survivor HMRa radS2-1 ho HMLa MATa-survivor HMRa rad52-1 HO HMLa MA Ta-survivor HMRa rad52-1 HO HMLa MA Ta-survivor HMRa rad52-1

N. Rudin R. Malone R. Malone R. Malone R. Malone R. Malone

RM124-23Cb RM121-5Ab RM121-6Cb RM37-1Db

RM38-6Ab a

b

The designations stkl and stk2 reflect the origin of the allele. They are, in fact, identical. The HMR locus of this strain has not been sequenced.

5374

RAY ET AL.

MOL . CELL . B IOL .

MAT.

9gW HindI

BgIII MATastk

Eco

Ps"

MAT.

SaRn

EcoR

IilAasail>HIiIII{ 7.5kb Hknd

pBLR4O

W

X

Ya

pBLR41

PSI!

Z1 Z2 I

+RI

7.7kb

1000 bp

I

FIG. 2. Restriction analysis of MATa. Abbreviations: Bg, BglII; H, HindlIl; N, NcoI; R, EcoRI. The XhoI linker insertions ax124 and ax161 have already been described (39). The sequence shown contains the Z1lT allele enclosed by a box. The arrows indicate the positions where HO endonuclease cuts.

Fogel. The presence of pmslA was confirmed by the ability of the strain to papillate on medium containing canavanine, which is indicative of the mutator phenotype associated with pmslA (16). Strains RM37-1D, RM38-6A, RM124-23C, RM121-5A, and RM121-6C were gifts from R. Malone (18). Plasmid constructions. Ligations and fillings in of DNA 5' protruding ends were performed as described by Maniatis et al. (19). Restriction enzymes were from New England BioLabs, Beverly, Mass. DNA fragments were purified by agarose gel electrophoresis, and the DNA was excised and then extracted by using GeneClean (BiolOl, La Jolla, Calif.). DNA purified by GeneClean was phenol extracted and ethanol precipitated before transformation into Escherichia coli by electroporation; transformation efficiency was poor when this treatment was omitted. Plasmid pJH68 is a TRPI ARSI derivative of pBR322 and carries MATa on a Hindlll fragment. In addition, it contains a deletion spanning the region between an XhoI linker in W (oax124) and an XhoI linker in Z2 (otxl6l) (39) (Fig. 2). Plasmid pJH124, used to disrupt the MAT locus, is a derivative of pJH68 and contains the SalI-to-XhoI fragment of LEU2 cloned into the XhoI site. To construct pCW2-3, plasmid pJH68 was digested with XhoI and NcoI and the resulting linear fragment was used to transform MATa-stk strain MH188-4B to clone the MATa region by gap repair (24). The resulting plasmid was isolated from a Trp+ transformant, and the MAT HindIlI fragment was subsequently cloned into pGEM-3, yielding pCW2-3. Transformations. Yeast strains were transformed by using the lithium acetate method of Ito et al. (11). Yeast cotransformations were conducted as described by Rudolph et al. (26), by using 5 to 10 ,ug of plasmid YEp24, which contains the selectable marker URA3, and 20 jig of the DNA fragment which was to be integrated. Bacteria were transformed by electroporation with a Gene Pulser coupled to a Pulse Controller (Bio-Rad, Richmond, Calif.) as described in the Pulse Controller instruction manual. DNA amplification. Genomic DNA was isolated (31) and amplified by using the polymerase chain reaction (PCR) (27, 28, 30). Taq polymerase was purchased from Perkin Elmer Cetus, Norwalk, Conn., and amplification was carried out as described in their GeneAmp DNA Amplification Kit. PCR was performed by using approximately 5 ng of genomic DNA, and the DNA was amplified for 28 cycles, each cycle consisting of 1.5 min at 94°C, 2 min at 55°C, and 3 min at 72°C. Following the last cycle, the reactions were kept at 72°C for 7 min. Samples were then cooled to ambient temperature, ethanol precipitated, resuspended, and electrophoresed on a 0.6% agarose gel. The DNA band was

FIG. 3. Partial restriction maps of plasmids pBLR40 and pBLR41, which were constructed from pCW2-3 and pJH3, which contains the HindIII fragment of the MATa region cloned into the HindIII site of pBR322. In pBLR40, the BgIII-to-BamHI fragment of pCW2-3 was used to replace the BglII-to-BamHI fragment of pJH3. In pBLR41, the BglII-to-BamHI fragment of pJH3 was used to replace the BglII-to-BamHI fragment of pCW2-3.

excised, and the DNA was extracted by using GeneClean. To amplify MATa, pYa (AAATAAACGTATGAGATCTA) and pMATdistal (ATGTGAACCGCATGGGCAGT [designated pB in reference 42]) were used, and to amplify MATa, pMATdistal and pYa (GCAGCACGGAATATGGGACT [designated pA in reference 42]) were used. The pMATdistal primer is unique to MAT; hence, HML and HMR are not amplified when this primer is combined with pYa or pYa. Sequencing. Plasmids were sequenced by using the Sequenase kit (United States Biochemical Corp., Cleveland, Ohio) (37), which employs the dideoxy method first described by Sanger et al. (29). PCR-amplified DNA was sequenced as described by Winship (44). Primers for PCR and sequencing were synthesized by using a Cyclone DNA Synthesizer (MilliGen/Biosearch, div. Millipore Corp., Bedford, Mass.). Materials and media. All chemicals were reagent grade or better. Medium supplies were from Difco Laboratories, except for agar, which was from Acumedia Manufacturers. Agarose was from FMC Bioproducts, Rockland, Maine. Restriction enzymes, the DNA polymerase large fragment, and DNA ligase were from New England BioLabs. Nucleotides were purchased from Pharmacia, Piscataway, N.J. Sequenase is a product of United States Biochemical Corp. pGEM-3, T7, and SP6 primers were purchased from Promega Corp., Madison, Wis. YPD and synthetic minimal and sporulation media for S. cerevisiae were prepared as described by Sherman et al. (31). Medium containing 5'fluoro-orotic acid was prepared as previously described (3). YEPL is 1% yeast extract-2% Bacto-Peptone-3.7% (wt/vol) lactic acid adjusted to a final pH of 5.5. LB medium was prepared as described by Miller (22). RESULTS The mutation responsible for the stuck a>a mating phenotype is at the MAT locus. The MATa locus from stuck strain MH188-4B was cloned by using the method of gap repair (24) with XhoI-NcoI-cut plasmid pJH68 (see Materials and Methods). The pCW2-3 HindlIl fragment containing MATa-stk (Fig. 2) was subcloned from the gap-repaired plasmid and used to transform HO matA: :LEU2 strain RY400. This strain also was transformed simultaneously with URA3-containing replicative plasmid YEp24. Ura+ transformants were screened for loss of the Leu+ phenotype; approximately 0.25% of the Ura+ transformants were Leu-, consistent with replacement of the LEU2-marked deletion of MAT with the MATa-stk segment. These transformants proved to have an a>ac mating phenotype, characteristic of MATa-stk strains (9). These

VOL. 11, 1991

MECHANISM OF MAT SWITCHING

TABLE 2. Comparison of MATa sequence of Stk and normally switching strains Straina or

Phenotype or genotype

plasmid

Strains MH185-1D MH188-4B RY501 RM124-23C RM121-6C RM37-1D RM38-6A

a>a a>a

Nonmating a-lethal a-survivor a-survivor a-survivor

Plasmids pJH3 pCW2-3

MA Ta3 MATa-stk2

Sequence of MATa Z6-*Z16

CAGTAAAATTT CAGTAAAATTT CAGTATAATTT CAGTATAATTT CAGTAAAATTT

CAGTAAAATTT CAGTAAAATTT CAGTATAATTT CAGTAAAATTT

a The genotypes of these strains are provided in Table 1.

slowly switching a>a strains also contain nonmating diploid cells resulting from mating; consequently, these colonies were sporulated and tetrads were dissected. All tetrads yielded two nonmating and two a>oa mating spores, expected for a homothallic MATa-stk/MATa diploid. These results demonstrate that the genetic defect responsible for the a>a phenotype is on the gap-repaired, transformed HindIll fragment. Locating and sequencing the stk mutation. To identify the region of MAT that contains the stk mutation, two plasmids (pBLR40 and pBLR41) were created that contain reciprocal halves of the wild-type and stuck MATa alleles (Fig. 3). S. cerevisiae R175t was transformed simultaneously with 4 ,ug of YEp24 and 40 ,ug of either EcoRI and HindlIl-digested pBLR40 or HindIII-digested pBLR41. As before, Ura+ transformants

were

selected and then scored for loss of the

Leu+ phenotype. Because this strain contains a deletion of HMR, the resulting Ura+ Leu- transformants, RY510 from pBLR40 and RY520 from pBLR41, were each mated to a MATa spore from normally switching strain NR226-7B (HO HMLa HMRa). The resulting diploids, RY530 (derived from RY510) and RY540 (derived from RY520), were sporulated and dissected to obtain ho HMLa MATa (or MATa-stk) HMRa strains RY530-2B and RY540-2B for further analysis. To determine whether either of these strains carried MATastk, these spores were each mated to an HO MATa spore from NR226-7B, producing RY531 and RY541. By dissection of these resulting diploids, it was demonstrated that the stk lesion is distal to the BglII site in MATa since HO spores derived from the pBLR40 transformant acquired the a>a phenotype whereas HO spores derived from the pBLR41 transformant did not. The BglII-HindIII fragment of MAT from pCW2-4, a plasmid that contains the MAT HindIII fragment in the orientation opposite to that of pCW2-3, was cloned into pGEM-3, and the resulting plasmid was designated pCW5-2. A T7 primer was used to sequence MAT from the BglII site, and a synthetic oligonucleotide was used to prime synthesis from MAT distal into MAT. Sequencing of the stk allele revealed only one base pair difference from the published sequence of MATa (1); this difference was an A--G transition located 127 bp distal to the MAT-Y-MAT-Z border. Sequencing of the same region from pJH3, a MATa allele that switches efficiently, revealed the same base pair change, suggesting that this is not the mutation responsible for the Stk phenotype. However, pJH3 differed from pCW5-2 at

5375

position Zll in the MATa locus. The published sequence (1) for MATa has an A at this position, as does pCW5-2; however, pJH3 contains a T at this position. To test whether this base change is responsible for the observed phenotype, several normally switching strains and Stk strains were sequenced by using PCR-amplified DNA. All normally switching strains contained a T at position Zll, whereas all Stk strains contained an A (Table 2). Persistence of the stuck phenotype after several rounds of switching requires an additional mutation at HMRa. Two hypotheses can be offered to explain the Stk phenotype. Either the mutation is always present at MAT but has no effect on MATa switching, or the mutation is lost at MAT when switching from a to a and reintroduced from HMRa when switching back to MATa, thus requiring that the mutation be present at HMRa. Work by McGill et al. (21) suggested that a mutation at position Zll frequently would be coconverted during mating-type switching, so that the stk mutation should be lost. To test whether the mutation is lost during switching from MATa-stk to MATa, strain RY530-2BT was induced to switch from a to a by galactose-mediated induction of the HO gene for 45 min (42). The cells were plated for single colonies and then tested for mating type. The MAT region from 14 a maters was amplified by PCR and then sequenced. All 14 a maters contained a T at position Zll, verifying that the Z1lA stk mutation had been lost during mating-type switching. These results implied that strains in which the Stk phenotype persisted after several rounds of switching must also contain the same mutation at HMRa. To determine whether the HMRa locus of Stk strains contained an A at Zll, PCR was used to amplify the HMR locus from Stk strains MH185-1D and U110. The HMRa locus from these strains was found to contain an A at position Zll, whereas the HMRa locus from the normally switching strains DBY745 and NR238-23D each contained a T at position Zll (data not shown). The MATa-survivor mutation is identical to MATa-stk. Malone and Hyman (18) described an allele of MA Ta, designated MATa-survivor, which was viable in an HO rad52 context. When we crossed a MATa-survivor strain into our genetic background, we also found that MATa-survivor gave rise to the same a>a phenotype as we have reported for MATa-stk (data not shown). We used PCR amplification to sequence the Y-Z junction from several MATa-survivor alleles provided by R. Malone (RM37-1D, RM38-6A, RM121-5A, and RM121-6C). Each was found to contain the identical T--A change at position Zll as in MATa-stk. No other mutation was found in the region sequenced, which covered 50 bp of Ya and 150 bp of Zl. The Ya-Z junction at HMRa from these strains also contained this substitution. In contrast, a MATalethal locus from RM124-23C contained a T at Zll. We conclude that MATa-survivor and MATa-stk are identical alleles. The previously reported differences in their apparent switching phenotypes may depend on strain background or on different methods of analysis of mating efficiency. Some switches of MA Ta-stk yield matal -stk, which is sterile. If the A at position Zll in MATa-stk were maintained at MAT during switching, the resulting MATa-stk would presumably be sterile and, hence, nonmating because the al peptide would be prematurely terminated (1, 39). We have confirmed that MATa-stk strains with an A at position Zll are indeed sterile. Strain RY530-2BT cells (MATa-stk) were galactose induced for 30 min, washed, and then plated for single colonies. Nonmating, a mating, and a mating colonies

5376

RAY

MOL. CELL. BIOL.

ET AL.

MATa

MATa-stk

MATui

MATa,i-stk

0 30 6090

0 30 6090

0 30 6090

0 30 6090

TABLE 3. Mating-type switching after 30 min of galactose induction'

min

No. switched/total' (% switch) Expt 1 Expt 2 36/101 (36) 73/108 (68)

Mating type MATa Z11T

WIN

MATa Z11A MATa Z11T MATa Z1lA

MAT

Distal

I i..s

ttion

FIG. 4. Time

RY530-2B,

MAT.

of double-strand break forma

course

haploid strain containing

a

the

MATa-stk

allele and

Galactose

was

was

transformed with

then added for 30

min,

when cells

wer

-ewashed and

to

The dligeoste analysis.estea

ence

of

an

A at

Zll.

intermediate reaches MAT is balanced

lonfycteapres-cu onucleavae-cut

Accumulation of the HO endi

an

grown in YEP-lactate and

in

a

glucose medium. Aliquots (50 ml) wvere taken at intervals over a 2-h period. DNA was prepared from ti as previously described (42), and digested with EcoRI. DNA was then run on a 0.6% agarose gel for Southern, control, a normal MATa derivative of RY530-2BT wa!Ls obtaned by first galactose inducing the strain to obtain a MATa lerivative and then inducing this derivative to obtain a MATa strain. This process resulted in an A--*T base change at position Zll. Arf MATresk derivative was obtained similarly and sequenced to ve transferred

were

resuspension

a

plasmid conta uining the HO endonuclease under galactose regulation (11). The transformant, RY530-2BT, was grown in YEPL to a cell density of 3Xl107. normal HMRa,

47/112 (42) 90/114 (79) 16/114 (14)

galactose induced by pelleting and YEP-galactose. After 30 min, cells were diluted with YEPD and plated for single colonies on YEPD agar plates. One hundred twenty colonies were picked, and mating type was checked by crossing to strains of known mating type. In addition, the colonies were tested for the presence of the Gal::HO plasmid. The results of two separate experiments are given. The cells were the same as those used for Fig. 4. Total number of colonies which maintained the Gal::HO plasmid. Cells

Fragment

i.

.:' L,

12/108 (11) 69/111 (62) 4/115 (3.5)

apparent steady

More

efficient

switching

occurred

inducible HO endonuclease (12)

was

when

galactose-

the

used. We examined in

vivo cutting of the MAT locus from isogenic MA Ta, MA Tastk, MAToa, and matal-stk strains by using the galactosenducible HO endonuclease, which was induced for 30 min. Figure 4 shows that even with the galactose-inducible HO

endonuclease, the MATa-stk allele as

the normal MATa

MAToa

or

not cut

was

as

efficiently

allele. The reduced level of

cutting is consistent with the observed reduced level in switching when galactose-induced cells were plated and colonies were tested for mating type (Table 3). HO endonuclease cutting of the matctl-stk (Z11A) allele

state becaus,

by its subsequent loss during completLion

of switch-

ing.

MATca

was also reduced, compared with that of the (Z11T) allele. In fact, the reduction was much greater for the Z11A

mutation in MAThL than in MATa with were no longer a miating, 5% (4 nonmating. The MATot locus fro)m the four nonmaters was amplified by PCR and sequenceid; an A was found at Zll in all four nonmaters. When primer~ sspecific for MATa were used for PCR, no amplified prodi Luct was obtamned, verifying that the strains were not non,imating as a result of being MATa/MATa, but rather that they were sterile because of the MATat-stk allele. This sti eiephenotype provided a convenient assay for determir quency at which the A at position Zll was repllaced during the mating type switching process (see below). The stk mutation diminishes cutting of MAT biy HO endonuclease. Weiffenbach and Haber (40) have shoNw'n that, like were

obtained. Of those that

of 85)

HO

were

MATa-survivor

strains

are

viable.

radS2

One

strains,

HO

explanation for

were

thein

tested for

experiment, when 50 HO M,~ Ta-stk cells were monitored through two divisions all oftthe colonies retained their a>a mating phenotype, indicating that none of the MATa-stk cells had switched to MAToa. Alltogether, no switches occurred at the first possible cell divisi,ion in any of 75 pedigrees analyzed. In

one

was

that

(Fig. 4). This is consistent galactose-induced HO mating-type

reduced at least 82% for

The difference in HO endonuclease in MATa and in

MATct

mataxl-stk

cutting of the

(Table 3).

same

allele

may reflect the fact that the HO

recognition sites are different (23). Switching of MATa-stk in G, cells yields postswitching segregation indicative of heteroduplex DNA formation. McGill et al. (21) demonstrated the presence of heteroduplex DNA during switching of several mutations in the MAT locus. To determine whether a heteroduplex is ever present at Zll, we used pedigree analysis by monitoring the fate of the mother and daughter cells produced by brief induction of HO in the G, stage of the cell cycle (7). Previous experiments had shown that galactose-induced G, (unbudded) cells nearly

MAT'a-stk radS2 observation,

recognition site (23), is that the MA Ta-stk locusiis cut poorly by the HO endonuclease. Such poor cutting would also explain the a>a phenotype observed for MAILTa-stk. We confirmed that the MATa-stk mutation marke(dly reduced switching by pedigree analysis. GerminatedI MA Ta-stk spores were identified by their inhibition by the a-factor and then were moved away from the pheromone and ,followed by micromanipulation through two cell divisions. Each of the mating type.

switching

finding

this

consistent with the notion that the mutation lie:,s in the HO

colonies from the first four progeny

our

TABLE 4.

Postswitching segregation of switching of G,

Strain

MATa-stk

the stk mutation after

cells'

No. (%) of mother-daughter pairs of switched cells with the following mating phenotype:

MATa and

MATa

MATa and

Matasl-Stk

Matatl-Stk and Matal-Stk

2 (3) RY530-2BT (PMSJ) 57 (74) 18 (23) 24 (59) 11 (27) RY610 (pins)) 6 (15) a MATa-stk cells of strains RY530-2BT (PMSJ) and RY610 (pmsl) were induced to switch by gaiactose induction of HO endonuclease for 45 min as described in Materials and Methods. Individual unbudded (GI) cells were micromanipulated on YEPD plates and allowed to divide once, whereupon the mother and daughter cells were separated and allowed to grow into colonies. mataxl-stk was scored on the basis of its nonmating phenotype. The identity of the mothier and daughter cells in each pair is not specified. Of 122 pairs, 4 yielded one MATa and one MATa colony. These data are not included here.

VOL . 1 l, 1991

always switched in pairs (that is, both progeny switched), just as for normal expression of HO (10, 21). In the present experiments, we confirmed this finding, as only 4 of 122 galactose-induced, unbudded MATa-stk cells gave rise to two progeny cells with opposite mating types. If a heteroduplex were formed in the Zl region during switching and not repaired, one MATa-stk cell should give rise to one MATa cell and one matal-stk cell. Pedigree analysis revealed that when a MATa-stk cell switched, 23% (18 of 77) of the switches yielded one MATa cell and one matal-stk cell (Table 4). In addition, 3% (2 of 77) of the switches yielded two matal-stk progeny. This demonstrates that the DNA in the MAT-Z region was often heteroduplex and furthermore that the mismatched base pair often failed to be repaired. Effect of prevention of mismatch repair on the frequency of PSS events. Switching of MATa-stk in wild-type cells clearly showed that the MAT-Z region contained heteroduplex DNA at least 23% of the time. If heteroduplex DNA had been created in every switching event, then mismatch correction would have had to occur frequently. Moreover, mismatch correction would have to be highly biased in favor of the invading strand, as 74% of the switches were completely wild type and only 3% gave rise to two progeny carrying the stk mutation. Alternatively, it is possible that neither strand of the newly switched MATa contained the stk mutation because both strands of DNA had been chewed back by exonucleases prior to repair of the double-strand break. To determine how often heteroduplex DNA was formed at MAT, we took advantage of a deletion of the PMSJ gene (16), which is required for mitotic and meiotic repair of single-base-pair mismatches. Previous studies had shown that AA and TT mismatches (i.e., those that should be formed between molecules carrying Zl1A [stk] and ZilT [Fig. 2]) are efficiently repaired to both AT and TA (2, 15). The pmsl deletion (16) was introduced into the strain used for the previously described pedigree analysis (Materials and Methods). The results of a pedigree analysis of galatoseinduced switches in pmslA strain RY610 (Table 4) demonstrated that at least 85% of the switching events must have retained at least one of the two MAT DNA strands carrying the stk mutation. Approximately 59% (24 of 41) of the switches exhibited PSS, indicative of the presence of unrepaired heteroduplex DNA, compared with 23% in the Pms+ strain. Surprisingly, the proportion of switches in which both the mother and the daughter retained the stk mutation also increased, from 3 to 27%. The MAT-ZllA mutation is not transferred to HMLa in a pmsl strain. It has been shown (21) that several different restriction site mutations in MAT-X or MAT-Z are not transferred to the silent donor copy during switching. Because a stable duplex must be formed between MATa and HMLa for synthesis to proceed into Ya, it is most likely that such a duplex must involve the Zll base, which is only 4 bases from the HO cut site. The failure to find transfer of the mutation to the donor might be explained by highly asymmetric mismatch repair of any heteroduplex that formed between HML and the invading MAT strand. Consequently, it was important to determine whether the Zl1A mutation was transferred to HML when mismatch correction was suppressed. We focussed our attention on the 15% of cases discussed above in which neither the switched mother nor its daughter cell retained the Zl1A mutation. DNA from colonies of each progeny was amplified by PCR to sequence the HML-Z region. In none of the six cases studied had the Zl1A mutation been transferred to HML. Thus, while 85%

MECHANISM OF MAT SWITCHING

5377

of the pairs of colonies retained the Z1lA mutation at MAT, none of the remaining colonies had experienced a transfer of this site to the donor. DISCUSSION The MATa-stk mutation was originally thought to lie outside of the MAT region that is replaced during matingtype switching, because the mutation was not lost during switching (9). Another attribute of the mutation is that it affects MATa switching but apparently does not affect MATTa switching. Both these anomalies could be explained if the mutation were lost upon switching to MATa and reintroduced upon switching back to MATa. We have now shown that the stk mutation at MAT is frequently lost upon switching from MA Ta-stk to MATa and that persistence of the Stk phenotype through several rounds of switching requires that the mutation be present at both HMRa and MA Ta. It is not clear from the way in which the MATa-stk mutations were isolated whether this site is subject to frequent mutation or the mutation was introduced during crosses with other strains carrying HMRa-stk (9). Apparently, the strains identified by Malone and Hyman as MATa-survivor (18) also carried this variant (Table 2). The finding that MA Ta-stk and MATa-survivor are identical indicates that both MATa ZilT and Zl1A are naturally occurring alleles of MATa. Although the originally published sequence of MATa contains an A at Zll (1), there are no previous data on the effectiveness of switching when an A is present at Zll, but a C or G at Zll has been reported to have no effect on cutting by HO endonuclease (23). The Zl1T polymorphism switches normally (more than 75% of the cells capable of switching do so) (41). In contrast, no switching was detected when more than 75 HO MATa Zl1A cells were monitored through two divisions, although they gave rise to the a>a phenotype characteristic of slowly switching colonies (see also reference 9). Galactose-induced expression of HO endonuclease from the Gal:: HO plasmid increased the mating type switching frequency of the MATa Zl1A-containing strain, but it was still reduced relative to that of the MATa Zl1T-containing strain (Table 3). Physical monitoring of DNA during galactose-induced expression of HO endonuclease (Fig. 4) showed, in vivo, that HO cleavage of MATa-stk was only about 50% of the level of MATa or MATa Zl1T. Together, these results indicate that poor cutting by HO endonuclease is solely responsible for the MATa-stk phenotype. Malone and Hyman (18) found that MATa-survivor showed apparently normal switching but resistance to HO rad52 lethality. In our study, this mutation was more clearly slowly switching, giving nse to a>a colonies. It is possible that the apparent difference in switching efficiency of MA Tasurvivor reflects different levels of HO endonuclease expressed in different strains. The slowly switching phenotype we observe is more consistent with the lack of lethality in HO rad52 backgrounds. One interesting result was that HO cutting of the ZllA mutation was much more efficient in MATa than in MATa (Fig. 4). Since the HO endonuclease recognition sites are different in the two mating types (because sequences in Ya and Ya are involved) (23), this is not necessarily surprising. Moreover, we have previously described a mutation in the HO endonuclease that was much less efficient in switching MATa than in switching MATa (20). Heteroduplex DNA formation at MAT-Z. In Pms+ cells, 23% of the switches produced pairs of cells in which one of the two progeny retained the stk mutation. These results are

5378

RAY ET AL.

similar to those obtained by McGill et al. (21), who found that 25% of pairs of switched cells exhibit PSS for a T-*C base pair substitution mutation at Z26. Furthermore, the proportion of events in which the Zll site is in heteroduplex DNA was underestimated in Pms+ cells. When similar experiments were done with an isogenic pmslA strain, in which mismatch correction is prevented more than 85% of the time in mitotic cells (2, 15), the proportion of PSS events increased from 23 to 59%. The appearance of PSS events is generally taken to be an indication that heteroduplex DNA formed in the region undergoing recombination and was left uncorrected. Sectored colonies could, of course, also arise if switching occurred independently on two chromatids, in the G2 stage of the cell cycle, after DNA replication. We previously demonstrated that galactose-induced HO expression in unbudded cells nearly always gives rise to the products expected for switching only in the G1 stage of the cell cycle (7). A further indication that the sectoring we observed was the result of heteroduplex formation comes from the fact that deletion of the PMSJ gene has such a profound effect on the outcome. The pmsl deletion would not be expected to affect the time or extent of HO endonuclease induction, but it should (and did) significantly change the proportion of sectored colonies if there is mismatch correction of the

heteroduplex. Previous analysis of MAT switching suggested that there was extensive 5'-to-3' exonucleolytic recision of one strand but that the 3' end distal to the HO-induced double-strand break appeared to remain intact (42). Extensive 5'-to-3' degradation has also been observed in systems in which the HO endonuclease cut site has been inserted into other chromosomal locations (8, 34), and it has been observed at the site of double-strand breaks produced during meiosis at the ARG4 locus (35). Our present genetic data show that fewer than four base pairs were removed from this 3' end in at least 85% of the switching events. Had both strands been degraded to form a gap, the stk allele could not have been recovered in the switched progeny. The 15% of pmslA cells in which both progeny had lost the stk allele might represent instances in which either at least four base pairs were deleted prior to invasion of HML or mismatch correction of heteroduplex occurred despite the pmsl deletion, or it might represent cases in which molecular intermediates of switching were resolved in a different way. It is unlikely that pmslA altered the resolution of intermediates because, consistent with the results of McGill et al. (21), when MAT-ZllA was lost in pairs of switched cells, it was never transferred to the HMLa donor locus in the six pairs of cells investigated. Both the Z1lA and Z26C mutations were recovered as PSS events in about 25% of switching events. However, in contrast to the Z26C mutation, in which 51% of switched cells produce two progeny that both preserved the mutation (21), only 3% of switched pairs both inherited the ZilA mutation. This difference could be explained if 5'-to-3' exonucleolytic processing of the HO-cleaved end often stopped before Z26, so that Zll was more often single stranded (to form heteroduplex DNA) while Z26 more frequently remained double stranded. However, this is unlikely, given the physical evidence that the entire MAT-Z region is frequently rendered single stranded by a 5'-to-3' exonuclease (42). Alternatively, the difference between the frequencies with which a mutation is retained at Zll or Z26 may be explained by variations in the efficiency of mismatch repair. However, the two types of mismatches formed at these two positions exhibit similar efficiencies of mismatch

MOL. CELL. BIOL.

correction (2, 15). Instead, we suggest that the Zll position is within the region which must invade the cassette to form a stable template for copying of the cassette and that Z26 is outside this region. If branch migration is minimal, then MATa Z26 will rarely be present in duplex with HMLa (see Fig. 5). Thus, Zll would be more likely to experience mismatch correction when it is in heteroduplex DNA with the HML donor sequences. A molecular model of MAT switching involving frequent formation of heteroduplex DNA in MAT-Z. To account for our observations of heteroduplex DNA and repair and previous studies of MAT switching (21, 42), we suggest the model shown in Fig. 5. Physical monitoring of MAT switching has demonstrated that the 3' end to the right of the HO cutting site initiates MAT switching by strand invasion and primer extension (42), leading to formation of the structure shown in Fig. SA. As with previous models (21, 36, 42), we presume that the strand displacement at HML leads to initiation of copying of a second new strand, beginning in the MAT-Z region (Fig. SC). This structure initially contains heteroduplex (stkl+) DNA at HML, which may undergo

several fates (Fig. SA). First, the heteroduplex at HML may undergo mismatch repair, leading to preferential correction of the invading strand (Fig. SB). Following completion of the repair events, this leads to two MATa progeny, both lacking the stk allele (Fig. SE). A highly preferential mismatch correction of the invading strand at HML, presumably because it is near a 3' end, also helps to account for the great disparity in the number of events in which both cells switched to lose the Zll mutation (74%) compared with those in which both have retained it (3%). It is more difficult to account for such disparity if heteroduplex DNA occurs only after the two newly synthesized strands are displaced back to MAT. Such a disparity is inconsistent with published observations of artificial heteroduplexes (2, 15). Alternatively, the mismatch may not be corrected. The newly copied strands are then displaced back to MAT, either catalyzed by topoisomerase (21) or by passive branch migration. The configuration of Z1lA and Z1lT alleles at MAT depends on the timing of this strand displacement and on completion of copying of the second strand of new information. If the displacement of the 3' strand containing the stk allele occurs after copying of the same region from the HML donor is completed (Fig. 5C), the MAT region will contain heteroduplex DNA (Fig. 5F). If the displacement occurs prior to copying of this region (Fig. 5D), DNA polymerase will use the stk-containing strand as a template to place the stk variant in both strands (Fig. 5G). This model presumes that the DNA polymerase is able to switch from using HML as a template to using MAT. Previous studies of DNA polymerases indicate that such template jumping is feasible (6, 17). The model also accounts for the effect of inhibition of mismatch repair by the pmslA mutation. In addition to an increase in PSS events, there is also an increase in events in which both switched strands retain the stk allele. This can be explained if PMSJ inhibits branch migration (Fig. SD), perhaps as a result of binding to the heteroduplex. Thus, there is a substantial increase in the proportion of events in which both MAT strands apparently retain the stk allele (Fig. SG). Formally, these events appear to be corrections, by restoration, of heteroduplex DNA; however, we believe that they are best accounted for by this proposed mechanism, in which they arise without formation of a heteroduplex. An alternative explanation would be that in the absence of the

A HML a

___-

MATa.

i

' a Y

Z

\

......

Ya

D

C

B

4

5379

Branch Migration

Mismatch Repair

-4

.

MECHANISM OF MAT SWITCHING

VOL. 11 1991

D

-Iv

......

I

Branch Migration or

F

E

Topolsomerase Unwinding

G

~~~.........

.......

I

.............

MATa

_0

,.

HML a *

_

.

Mismatch Repair

Mismatch Repair

__ o

HML a b

mat a 1

FIG. 5. Molecular model for heteroduplex formation during mating-type switching. Mating-type switching is initiated by a double-strand break at the MATa-stk locus. The HO endonuclease-cut DNA in the Z region is digested by a 5'-to-3' exonuclease. Concomitantly, the single-stranded 3' end invades the HMLa cassette (A). Primer extension adds Ya DNA to the MAT strand (C). The heteroduplex DNA at HML is subject to several alternative fates: mismatch correction preferentially converts the mutation to the wild type on the invading strand (B); branch migration may displace the unrepaired mutation back to MAT (D) prior to completion of copying of the second new strand. Completion of switching requires removal of MAT Ya sequences; this process remains uncharacterized, except that it appears to occur late in the switching process (42). A second new strand of MAT Ya DNA is initiated from the Z region. Depending on when branch migration has displaced the MAT-Z region, this replication event uses either HML or MAT DNA as a template in this region. As a consequence, the switched MAT region contains either two strands that have both have lost the original mutation (E), heteroduplex DNA (F), or two strands that both retain the mutation (G). A switched MAT that contains heteroduplex DNA may also be subject to mismatch correction prior to DNA replication of the chromosomal region. In a Pms- strain in which mismatch repair is greatly reduced, more of the products are of types F and G. Hatched arrows represent newly synthesized DNA.

PMS repair system, heteroduplex DNA that would normally be converted is more likely to be restored. This model accounts for all of the previous molecular (25, 42) and genetic (21) data on the fate of individual strands during MAT switching. We note, however, that while events involving the 3' strand in MAT-Z have been monitored, there is still no evidence on the timing and manner by which the Ya region DNA is removed. Ya DNA is not degraded by a 5'-to-3' exonuclease, as is the MAT-Z region, and it disappears only concomitantly with completion of switching (7, 25, 42). It seems likely that at least one MAT-Y strand must be removed to allow replicative copying of the second strand that is introduced at MAT, but other models, including reciprocal recombination of a replicated bubble of MAT DNA prompted by strand invasion of MAT-Z (an extension of the model of Stahl [32]), cannot be excluded. In the model of McGill et al. (21), resolution of the symmetrical complex between MAT and HML was accomplished by unwinding of the heteroduplex DNA by topoisomerase. Another possibility is that this displacement occurs by the propensity of the D loop created by strand invasion to remain small by rewinding the displaced donor strand behind the processive DNA polymerase complex copying the donor (36). Our data are inconsistent with the latter explanation. To form a heteroduplex at Zll, synthesis of the Y-Z junction on the 5'-to-3'-oriented strand at MAT must use HML as a template (Fig. SC). Since this step presumably occurs late in the mating-type switching process (42), long after the strand of

opposite polarity has invaded HML and copied much of Ya, the D loop formed must remain quite large. Finally, on the basis of observed differences in the loss and sectoring of XhoI linker insertions in the MAT-X region and the Z26 point mutation, McGill et al. (21) suggested that there might be a fundamental difference in the way molecular intermediates were formed in the MAT-Z and MAT-X regions. Given a significant variation in the efficiency of mismatch correction of different heteroduplexes (2, 15, 43), we believe it is difficult to extrapolate from the proportions of repaired and unrepaired products in different mismatches an argument for dissimilar roles of the MAT-X and MAT-Z regions during MAT conversion. Studies of mismatch correction involving synthetic heteroduplexes introduced into mitotic cells (2, 15) suggest that the system of mismatch repair including the PMSJ gene product probably does not act on the types of large insertions and substitution bubbles created by the XhoI linkers. Moreover, repair of these larger mismatches may sometimes yield lesions that themselves cause mismatch repair-induced secondary recombination events (4, 5). Secondary events of this sort have previously been found during MAT switching involving large heterologies (14); in the study of McGill et al. (21), as many as 15% of the pedigrees showed granddaughter sectors that may have arisen in this way. Clarification of this issue will require introduction of the same base pair substitutions into MAT-X that have been studied in MAT-Z.

5380

RAY ET AL. ACKNOWLEDGMENTS

Robert Malone generously provided MA Ta-survivor strains. Wil-

fred Kramer provided the pmslA plasmid. We thank John Wilson, Jacqueline Lobell, and Neal Sugawara for comments on the manuscript. This research was supported by grant GM20056. REFERENCES 1. Astell, C. R., L. Ahlstrom-Jonasson, M. Smith, K. Tatchell, K. A. Nasmyth, and B. D. Hall. 1981. The sequence of the DNAs coding for the mating-type loci of Saccharomyces cerevisiae. Cell 27:15-23. 2. Bishop, D. K., M. S. Williamson, S. Fogel, and R. D. Kolodner. 1987. The role of heteroduplex correction in gene conversion in Saccharomyces cerevisiae. Nature (London) 328:362-364. 3. Boeke, J. D., F. Lacroute, and G. R. Fink. 1984. A positive selection for mutants lacking orotidine 5'-phosphate decarboxylase activity in yeast. Mol. Gen. Genet. 197:345-346. 4. Borts, R. H., and J. E. Haber. 1987. Meiotic recombination in yeast: alteration by multiple heterozygosities. Science 237: 1459-1465. 5. Borts, R. H., W.-Y. Leung, W. Kramer, B. Kramer, M. Williamson, S. Fogel, and J. E. Haber. 1990. Mismatch repair-induced meiotic recombination requires the PMSJ gene product. Genetics 124:573-584. 6. Bryant, F. R., K. A. Johnson, and S. J. Benkovik. 1983. Elementary steps in the DNA polymerase I reaction pathway. Biochemistry 22:3537-3546. 7. Connolly, B., C. I. White, and J. E. Haber. 1988. Physical monitoring of mating type switching in Saccharomyces cerevisiae. Mol. Cell. Biol. 8:2342-2349. 8. Fishman-Lobell, J., and J. E. Haber. Unpublished data. 9. Haber, J. E., W. T. Savage, S. M. Raposa, B. Weiffenbach, and L. B. Rowe. 1980. Mutations preventing transpositions of yeast mating type alleles. Proc. Natl. Acad. Sci. USA 77:2824-2828. 10. Hicks, J., and I. Herskowitz. 1976. Interconversion of yeast mating types. I. Direct observation of the action of the homothallism (HO) gene. Genetics 83:245-258. 11. Ito, H., Y. Fukuda, K. Murata, and A. Kimura. 1983. Transformation of intact yeast cells treated with alkali cations. J. Bacteriol. 153:163-168. 12. Jensen, R. E., and I. Herskowitz. 1984. Directionality and regulation of cassette substitution in yeast. Cold Spring Harbor Symp. Quant. Biol. 49:97-104. 13. Klar, A. J. S., J. B. Hicks, and J. N. Strathern. 1982. Directionality of yeast mating-type interconversion. Cell 28:551-561. 14. Klar, A. J., and J. N. Strathern. 1984. Resolution of recombination intermediates generated during yeast mating type switching. Nature (London) 310:744-748. 15. Kramer, B., W. Kramer, M. S. Williamson, and S. Fogel. 1989. Heteroduplex DNA correction in Saccharomyces cerevisiae is mismatch specific and requires functional PMS genes. Mol. Cell. Biol. 9:4432 4440. 16. Kramer, W., B. Kramer, M. S. Williamson, and S. Fogel. 1989. Cloning and nucleotide sequence of DNA mismatch repair gene PMSJ from Saccharomyces cerevisiae: homology of PMS1 to procaryotic mutL and hexB. J. Bacteriol. 171:5339-5346. 17. Kunkel, T. A., and A. Soni. 1988. Mutagenesis by transient misalignment. J. Biol. Chem. 263:14784-14789. 18. Malone, R. E., and D. Hyman. 1983. Interactions between the MAT locus and the rad52-1 mutation in yeast. Curr. Genet. 7:439-447. 19. Maniatis, T., E. F. Fritsch, and J. Sambrook. 1983. Molecular cloning: a laboratory manual. Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y. 20. Mascioli, D. W., and J. E. Haber. 1980. A cis-dominant mutation within the MATa locus of Saccharomyces cerevisiae that prevents efficient homothallic mating type switching. Genetics 94:341-360. 21. McGill, C., B. Shafer, and J. Strathern. 1989. Coconversion of flanking sequences with homothallic switching. Cell 57:459-467. 22. Miller, J. H. 1972. Experiments in molecular genetics. Cold

MOL. CELL. BIOL.

Spring Harbor Laboratory, Cold Spring Harbor, N.Y. 23. Nickoloff, J. A., J. D. Singer, and F. Heffron. 1990. In vivo analysis of the Saccharomyces cerevisiae HO nuclease recognition site by site-directed mutagenesis. Mol. Cell. Biol. 10: 1174-1179. 24. Orr-Weaver, T. L., J. W. Szostak, and R. J. Rothstein. 1981. Yeast transformation: a model system for the study of recombination. Proc. Natl. Acad. Sci. USA 78:6354-6358. 25. Raveh, D., S. H. Hughes, B. K. Shafer, and J. N. Strathern. 1989. Analysis of the HO-cleaved MAT DNA intermediate generated during the mating-type switch in the yeast Saccharomyces cerevisiae. Mol. Gen. Genet. 220:33-42. 26. Rudolph, H., I. Koenig-Rauseo, and A. Hinnen. 1985. One-step gene replacement in yeast by cotransformation. Gene 36:87-95. 27. Saiki, R. K., D. H. Gelfand, S. Stoffel, S. J. Scharf, R. Higuchi, G. T. Horn, K. B. Mullis, and H. A. Erlich. 1988. Primerdirected enzymatic amplification of DNA with a thermostable DNA polymerase. Science 239:487-491. 28. Saiki, R. K., S. J. Scharf, F. Faloona, K. B. Mullis, G. T. Horn, H. A. Erlich, and N. Arnheim. 1985. Enzymatic amplification of ,-globin genomic sequences and restriction site analysis for diagnosis of sickle-cell anaemia. Science 230:1350-1354. 29. Sanger, F., S. Nicklen, and A. R. Coulson. 1977. DNA sequencing with chain-terminating inhibitors. Proc. Natl. Acad. Sci. USA 74:5463-5467. 30. Scharf, S. J., G. T. Horn, and H. A. Erlich. 1986. Direct cloning and sequence analysis of enzymatically amplified genomic sequences. Science 233:1076-1078. 31. Sherman, F., G. R. Fink, and J. B. Hicks. 1986. Laboratory course manual for methods in yeast genetics. Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y. 32. Stahl, F. W. 1979. Genetic recombination. W. H. Freeman & Co., San Francisco. 33. Strathern, J. N., A. J. S. Klar, J. B. Hicks, J. A. Abraham, J. M. Ivy, K. A. Nasmyth, and C. McGill. 1982. Homothallic switching of yeast mating type cassettes is initiated by a double-stranded cut in the MAT locus. Cell 31:183-192. 34. Sugawara, N., and J. E. Haber. Unpublished data. 35. Sun, H., D. Treco, and J. W. Szostak. 1991. Extensive 3'overhanging, single-stranded DNA associated with the meiosisspecific double-strand breaks at the ARG4 recombination initiation site. Cell 64:1155-1161. 36. Szostak, J. W., T. L. Orr-Weaver, R. J. Rothstein, and F. W. Stahl. 1983. The double-strand break repair model for recombination. Cell 33:25-35. 37. Tabor, S., and C. C. Richardson. 1987. Proc. Natl. Acad. Sci. USA 84:4767-4771. 38. Takano, I., T. Kusumi, and Y. Oshima. 1973. An a mating type allele insensitive to the mutagenic action of the homothallic gene system in Saccharomyces diastaticus. Mol. Gen. Genet. 126: 19-28. 39. Tatcheli, K., K. A. Nasmyth, B. D. Hall, C. Astell, and M. Smith. 1981. In vitro mutation analysis of the mating-type locus in yeast. Cell 27:25-35. 40. Weiffenbach, B., and J. E. Haber. 1981. Homothallic mating type switching generates lethal chromosome breaks in rad52 strains of Saccharomyces cerevisiae. Mol. Cell. Biol. 1:522534. 41. Weiffenbach, B., D. T. Rogers, J. E. Haber, M. Zoller, D. W. Russell, and M. Smith. 1983. Deletions and single base pair changes in the yeast mating type locus that prevent homothallic mating type conversions. Proc. Natl. Acad. Sci. USA 80:34013405. 42. White, C. I., and J. E. Haber. 1990. Intermediates of recombination during mating type switching in Saccharomyces cerevisiae. EMBO J. 9:663-673. 43. Williamson, M. S., J. C. Game, and S. Fogel. 1985. Meiotic gene conversion mutants in Saccharomyces cerevisiae. I. Isolation and characterization of pmsl-l and pmsl-2. Genetics 110:609646. 44. Winship, P. R. 1989. An improved method for directly sequencing PCR amplified material using dimethyl sulphoxide. Nucleic Acids Res. 17:1266.