... Sciences Research Center and Department of Biology, Brandeis University, Waltham, Massachusetts 02254 ...... J. LOUIS for help with the statistical analysis.
Copyright 0 1989 by the Genetics Society o f America
Length and Distributionof Meiotic Gene Conversion Tracts and Crossovers in Saccharomyces cerevisiae Rhona H. Borts and James E. Haber Rosenstiel Basic Medical Sciences Research Center and Department of Biology, Brandeis University, Waltham, Massachusetts02254 Manuscript received February 2 1, 1989 Accepted for publication May 30, 1989 ABSTRACT We have measured gene conversion tract length instrains of the yeast Saccharomyces cerevisiae containing three to six restriction site heterozygosities in a 9-kb interval. Tetradscontaininga conversion were identified genetically by nonmendelian segregation of a marker in the middle of the interval. Gene conversions accompanied by a crossover have a tract length of 1.4 kb & 0.7 kb, which is indistinguishable from a tract length of 1.6 f 0.8 for conversions without an associated exchange. Among tetrads identified first as having a crossover in the interval, the average gene conversion tracts were apparently significantly shorter (0.71 f 1). We provide evidence that this apparent difference is due to the method of measuring conversion tracts and does not reflect a real difference intract length. We also provide evidence that the number and position of restriction site markers alters the apparent distribution of the conversion tracts. More than ninety percent of the conversion tracts spanning three or more sites were continuous.
HERE appears to be an intimate connection between meiotic crossing over and gene conversion. In fungal systems such as Saccharomyces cerevisiae, where all four products of meiosis can be recovered, gene conversions at a particular locus are associated with crossing over between 30 and 70% of the time (FOCEL,MORTIMER and LUSNAK 1981;FOCELet al. 1979;HURST, FOCELand MORTIMER 1972);fora general review, see (ORR-WEAVER and SZOSTAK 1985). Conversely, when reciprocal exchanges are examined in a small interval, at least 60% areaccompanied by gene conversion of one or more heterologies within the exchange interval (BORTSand HABER 1987; SYMINCTON and PETES1988a). Nevertheless, it is not clear if conversion events that are accompanied by exchange, and those that are not, derive from the same series of recombinational steps. CARPENTER (1984, 1987) has suggested that there may be two related, but separable, processes to generate geneconversions with and without an associated exchange. By this hypothesis, the initial synapsis between homologous chromosomes or homologous segments on different chromosomes involves the formation of a region of heteroduplex DNA that is enzymatically checked for thepresence of non-homologies. If heteroduplex DNA contains too many heterologies, i e . , is sufficiently mismatched, the process is aborted, leaving behind a region of heteroduplex DNA that is not associated with a Holliday junction. Such a structure might still include one or more heterologies in
T
T h e publication costs o f this article were partly defrayed by the payment "advertisement" of pagecharges. This article must therefore be hereby marked in accordance with 18 U.S.C. $1734 solely to indicate this fact. Genetics 123: 69-80 (September, 1989)
heteroduplex DNA that can be corrected by mismatch repair to yield a gene conversion, but it will not be associated with an exchange event. When sufficient homology is encountered, heteroduplex DNA could be extended andisomerization of the recombinational intermediate would permit, if not require, crossingover to accompany gene conversions. A corollary to this proposition is that the extentof a geneconversion event (the length of sites that are co-converted) that is not associated with a crossover might besignificantly shorterthan thoseeventsthat are associated with exchange of flanking markers. Data to support this idea were found in studies of gene conversion of the rosy locus in Drosophila (CARPENTER 1984). However, in another study of conversion tracts at the rosy locus, the converse was found to be true, that conversions associated with a crossover were shorter than those which were not (CURTISet al. 1989). Previous studies of conversion tract length in S. cerevisiae could not address this issue for a number of reasons. In two studies, conversions wereidentifiedamongevents which were identifiedascontaininga crossover (BORTSand HABER 1987; SYMINCTON and PETES 1988a,b) and thereforeonly one of the two classes of conversions were analyzed. In another study UUDD and PETES1988), events were identified as conversions, but whether the events were associated with a crossover was not reported. To explore this question further, we havetaken advantage of a well-defined chromosomal region that undergoes avery high level of crossing over and gene conversion (BORTSand HABER 1987; BORTS,LICHTEN andHABER1986; BORTSet al. 1984). The region
,
70 Kb
R. H. Borts and J. E. Haber ,l,.93
.78
. 4 4 . 6 7 0 7 .73
I I Ill
Bglll
Sa
RHB680
Xh
NcSm Xh
Xh
Bglll
Nd
Xh
MAT a V1X-I
RBT23
r
RBT59
MAT a I
MAT a 1
I
I
Bglll
Xh
I
RHB631 Xh
I I
Bglll
Xh Xh
RBT39-4a
MAT a RHB 555-18~1
MAT a
I
I 1
I
1
I Bglll
Xh
RH9752 Xh
Nc
1 Bglll
Xh
Xh
FIGURE 1 .-The physical structure of the MATURA3-MAT interval is shown. The haploid parents of each strain are indicated next to the map of the interval. Wild-type restriction sites which are heterozygous in each diploid are indicated in bold face. Mutant sites are indicated by a black dot. The distances between sites are given in kilobases.
RBT39-4a RBT59
MAT a
I
I
MAT a I
I J
-
Bglll RHB633
I
RET28
I
Sa
Xh
Nc
Xh
I I
MAT a V/A
VIA
MAT a I
1
Bglll
Nd
S X hX h
MAT a
RBT39-4a
1
1
1
MAT a I
IJ
Bglll
1 NSa d
e
7
Nc
consists of anon-tandem duplication of the yeast mating type (MAT)locus, separated by pBR322 and URA3 sequences (Figure 1). By placing multiple restriction site heterologies in this interval, it is possible to subdivide the region and to measure the extent of gene conversion associated with crossing over (BORTS and HABER1987). A detailed analysis of these events was complicated by the observation that the presence of the heterozygous markers profoundly alters the outcome of the recombination event. When the number of heterologies is large (9 heterologies spreadover 9 kb), normal exchange events are significantly perturbed. Approximately halfof the recombination events were recovered in two classes (MATconversions and intrachromosomal crossovers that yield a Urasegregant with a single MAT locus). These two classes occur at much lower frequencies in the absence of heterologies. Extending models proposed by HASTINCS (1984) and GLICKMAN and RADMAN (1 980),we suggested that a second recombination event, involving a highly recombinogenic intermediate such as a double-strand break or long single stranded region, is induced by the process of correction of multiple heterologies in heteroduplex DNA (BORTS and HABER 1987). This proposal has been substantiated by the recent observation thatthese aberrant events are markedly reduced in diploids homozygous forthe
I
I 1 1 Bglll
pmsl-1 mutationthat is defective in therepair of heteroduplex DNA (R. H. BORTS, W.-Y. LEUNC,M. S . WILLIAMSON, W. KRAMER,B. KRAMER,S . FOCEL and J. E. HABER,unpublished data). We have also suggested thatthefrequency withwhich repair induced recombination occurs is a function of the distance between the heterologies and that the frequency with which it is detected as a MAT conversion or a loopout is a function of the proximity of the heterozygosities tothe MAT locus (R. H. BORTS,W.-Y. W. KRAMER, B. KRAMER, LEUNC,M. S. WILLIAMSON, S . FOCELand J. E. HABER,unpublished data). Because there is little effect on the normalfrequency of crossing over in this interval (26% of all tetrads) if one inserts fewer heterologies (BORTSand HABER1987; R. H. BORTS,W.-Y. LEUNC,M. S . WILLIAMSON, W. KRAMER,B. KRAMER,S . FOCELand J. E. HABER,unpublished data), we have used diploids containing 1, 3 or 6 heterologies. One of the markers is an NcoI mutation within the URA3 gene that creates a ura3 allele, ura3-N (Figure 1). This site is flanked by 0, 2 or 5 heterozygosities. Tetrads were dissected to obtain those in which there had been a gene conversion of URA3 (either 3+:1- or 1+:3-). The conversion tract lengths associated with URA3 conversion events thatwere unaccompanied by a crossing over of the flanking MAT regions were then compared with
71
Gene Conversions and Crossovers the conversion tracts of URA3 conversions associated with crossing over. We present evidence that there is n o significant difference in the extentof gene conversionwith, or without, an associatedcrossingover. Thesestudiesalsoprovideinformationaboutthe in the interkinds of recombination events that occur val when it is not greatly perturbedby a large n u m b e r of heterologies. MATERIALSAND
METHODS
Media: YEPD, synthetic complete,sporulation (SHERFINKand HICKS1986) and 5-fluoro-orotic acid media (BOEKE,LACROUTE and FINK 1984) have been described previously. Plasmids: pRHB15 (not shown) is a pBR322 plasmid in which the SalI and NdeI sites of pBR322 were ablated by cutting with SalI and NdeI, respectively, filling in the 5' overhangs with the Klenow reaction (MANIATIS, FRITSCH and SAMBROOK 1983), followed by ligation with T 4 ligase. The Sal1 ablation is a 4-bp duplication and generates aPvuI site. The NdeI ablation is presumed to be 2-bp duplication. The AvaI site has been changed to a XhoI site by blunt end ligation of 8 bp XhoI linkers to the filled in AvaI site. The and BOTSTEIN1984) 1.2-kb URA3 fragment (ROSE,GRISAFI to which XhoI linkers were added at the filled in HindIII ends, was inserted at this XhoI site. pRHB35 is a Ura3- derivative of pRHB15. The NcoI site at position 436 in URA3 was converted to a NsiI site as described above for the Sal1 and NdeI sites. This ablated site has a four base pair duplication and is referred to as ura3-N. pRHB35 also containsa 2.2 kb LEU2 fragment et al. 1984) inserted at theSsp1 site of pBR322. (ANDREADIS pJH250 is a derivative of pRHB15 to which the 3.4-kb HindIII-EcoRI MATa fragment has been inserted between the EcoRI and HindIII sites of pBR322. pRHB23 is similar to pJH250 except that all pBR322 sequences are wild type. pRHB22 also has wildtype pBR322 sequences. In this plasmid the 1.2-kb URA3 fragment is blunt end ligated directly into the filled in AvaI site. In addition the SmaI site in the 3"noncoding sequences has been destroyed by cutting and filling in an XmaI site. The exact nature of the SmaI ablation is unknown. Strains: The structure of the MAT locus in the diploid strains RHB631, 633, 680 and 752 is illustrated in Figure 1 and the genotypes of the haploid parents of these strains are given in Table 1. With the exception of RHB63 1 these strains are isogenic to strain the strain RHB536 (BORTSand HABER 1987). All MATa parents are isogenic. RBT23andRBT15 (BORTSand HABER 1987) were created by targeted integration of H330 with pRHB23 and pJH250, respectively. In order toinsert the ura3-N mutation, the MATa-URA3-MATa parent haploid strain, RBT15, was transformed (BRUSCHI, COMERand HOWE 1987) with pRHB35. ALeu+ Ura+ transformant (RBT39, not shown) containingatandem duplication of pBR322 sequences was then plated on 5fluoro-orotic acid (BOEKE,LACROUTE and FINK 1984) to select a Leu-Ura- derivative that had "looped out" both LEU2 and URA3, leaving behind one copy of pBR322 containing the u r d - N mutation. The structure of this derivative, RBT39-4a, was confirmed by Southern blot analysis (SOUTHERN 1975). All MATa-URA3-MATa parentsare isogenic (with the exception of the MATa parent of RHB63 1, described below) and were created by targeted integration (ORR-WEAVER, MAN,
SZOSTAKand ROTHSTEIN 1983) of either RBT10-2d or RBT13-le with the MATa-URA3 plasmids, pRHB22 or pRHB23. RBTl0-2d and RBT13-le areisogenic and were derived from H394 by transformation. In order to make RHB63 1, a diploid homozygous for all markers except the ura3-N mutation, RBT15 was crossed to a homothallic, ura3-1, Y55 strain. Tetrads were dissected and an HO, nonmating Ura+ segregant was sporulated. An a-mating Ura+ pBR322' segregant, in which both copies of the mating type cassette had switched from MATa to MATa, was rescued by mating to H394. This diploid was dissected to recoveran ho MATa Ura+ pBR322+ derivative, RHB55518c, which was mated to RBT39-4a to generate RHB63 1 . Genetic and physical analysis of recombination: Cells were grown at30"and sporulated at25" by standard methods (SHERMAN, FINKand HICKS 1986).Meiotic tetrads were dissected on YEPD plates and grown into colonies. The colonies were assayed for gene conversion of the URA3 locus and for crossing-over in the MAT-MAT interval by replica plating tests. Gene conversions were detected as 3+:1- or 1+:3- segregations for growth on synthetic complete medium lacking uracil. Recombination of the flanking markers was detected by the presence of non-mating segregants resulting from the co-dominant expression of MATa and MATa in MATa-MATa or MATa-MATa haploid progeny. The co-conversion of the SalI, XhoI, SmaI and NdeI sites flanking the ura3-N mutation were assessed by Southern blot analysis of DNA from all four spores of tetrads exhibiting a conversion of URA3 or crossing-over in the MATMAT interval, as described by BORTSand HABER (1987). In some cases the presence or absence of the NdeI site was determined by differential hybridization with a 3'P-labeled oligonucleotide homologous to the filled in NdeI (but not and MIYADA1987; WILLIS the wildtype NdeI site) (WALLACE and KLEIN 1987). The conditions under which hybridization to the mutant site, (4X SSPE, 39") but not the wild-type site, occurred were determined empirically. Southern blot analysis distinguished Ura-segregants that hadresulted from intra-chromosomal "loop-out" or unequal crossingover between the flanking MAT regions on sister or nonsister chromatids from gene conversions. All events leading to loss of pBR322 sequences were identified in RHB63 1 and RHB752 by colony hybridization (SHERMAN, FINKand HICKS1986) with "P-labeled pBR322 (MANIATIS,FRITSCH and SAMBROOK 1983). Statisticalanalysis: Statistical differences between the averages of the minimum of conversion tract length were determined using the Student's t test (SOKALand ROHLF 1969). Differences in the distributions of conversion tracts were calculated using a G test (SOKAL and ROHLF 1969). P values less than 0.05 were considered significant. T o determine if interference was occurring, the probability of the observed number of four-strand double crossovers was calculated by the binomial distribution (SOKALand ROHLF 1969).A G test was used todetermine if the observed number of three-strand double crossovers was consistent with the interference determined for four-strand double crossovers. RESULTS
Thepresence of 3 or 6 heterologies does not stimulate a highfrequency of aberrantevents: A total of 924 tetrads of strainRHB633 a n d 383 tetrads of RHB752 were dissected and analyzed (Table 2). T h e results show that the presenceof 3 or 6 heterol-
72
R. H. Borts and J. E. Haber TABLE 1 Construction of haploid strains
Strain
Transforming plasmid
Genotype
H394
Haploid parent
ura3-1 adel met13-4 cyh2 lys2-d leu2-R MATa
RBT 10-2d"
ura3-1 adel
RBT13-le"
metl3-4 lys2-d ura3-1 adel
pRHBS3
H394
cyh2 leu2-Rl MATa his4-R
pRHB12
H394
cyh2 leu2-Rl his4-R MATa-XhoS (IRA3
pRHB23
RBTS3-le
pRHB22 RBT
10-2d
1332-dleu2-Rl his4-Cl met13-4 MATa cyh2
RBT59'
ura3-1 adel lys2-d met13-4 Xholi-pBR322-MATa
RBT28'
urajr-1 adel lys2-d met13-4 cyh2 leu2-Rl his4-Cl MATa-URA3 (SmaI-)-MATa
H330
ura3-1 canl adel lys2-c me113-2 cyh2 trp5-1 leu2-K MATa
RBT 15
ura3-1 canl adel lys2-c met13-2 cyh2 trpJ-1 leu2-K MATa-Xhol+ (IRA3 Xhol+-pBR322 (SalI-, NdeI-)-MAT&
pJH250
H330
RBT23b
ura3-1 canl adel lys2-c met13-2 cyh2 trp5-I leu2-K MATa-Xhol+ (IRA3 Xhol"pBR322-MATa
pRHB23
H330
RBT39-4a""
u r d - 1 cant adel lys2-c met13-2 cyh2 trp5-1 leu2-K MATcu-Xhol+ ura3-N Xhol+-pBR322 (Sall-, NdeI-)-MATa
RBT39
RHB555-18cb
ura3-1 canl his6 met13-4 trp5-1 lys2-c leu.2-K MATa-URA3-PBR322 (SatI-, NdeI-)-MATa
RBTl5
a RBT10-2d and RBT13-le were derived from H394 by transformation with the IIRA3, his4-Clal- (his4-Cl) and Ura3+, his4-EcoRI(his4-R) plasmids, pRHB13 and 12 respectively. The ClaI and EcoRI sites in HIS4 were destroyed as for other restriction site ablations as described in MATERIALS AND METHODS. His-, Urd- derivatives were selected by looping out of the plasmid sequences on 5-fluoro-orotic acid BORTSand HABER1987). (BOEKE,LACROUTE and FINK 1984). The leu2-K and leu2-R alleles have been described previously (LICHTEN, The diploids derived from these strains are shown in Figure 1.
TABLE 2 Frequency of recombination events in MAT-MAT duplications URA3 conversions
Diploid"
Total reciprocal crossovers
RHB63l 1
65 9 (20%) (2.1%) (4.6%) (2.7%)
RHB752 3
94 22 19 (24.5%) (5.0%)
RHB633 6
208 (5.3%) (22.5%)
RHB680 0
166 (26%)
Without crossover
Other'
13 (4.2%)
3 (0.09%)
1 (0.03%)
20 (5.2%) (0.26%) (3.7%) (5.7%)
14
1
2 (0.52%)
41 43 (4.4%) (0.6%) (3.5%) (4.6%)
9 32
6
With
3+:1-
49 NA'
Aberrant events' With MAT conversion
1+:3-
15
N AN A
crossover
7
NA
pBR322 losses
0
16 (1.3%) (4.1%) N D ~
(1 %) NA
NA
10 (1.6%)
MAT conversions
Total No. of tetrads
8 (2.4%)
329
5
383
19 (2.0%)
924
12
637
(1.9%)
' All MATa strains are isogenic. All MATa strains are also isogenic except the MATa parent of RHB631. The number of heterozygosities in each diploid is given under the strain name. Aberrant events were detected genetically as Ura- or by colony hybridization as Ura-, pBR322- segregants (SHERMAN, FINK and HICKS 1986). Detailed analysis of these events (BORTSand HABER1987) indicates that these events consist of intrachromosomal crossovers and unequal inter- and sister-chromatid exchanges. ' These events had three or four nonmating spores or were conversions detected in aberrant events by Southern analysis. Four of the nine events had two or more conversion tracts and it could not be determined if it was the urajr-N conversion which was associated with the crossover. Not done. Not applicable.
'
ogies yielded little perturbation of thepattern of recombinationfound for strainshaving one or no heterologies (Table 2, and BORTSand HABER1987). Of all tetrads, 23-25% exhibited the pattern of one a-mating,onea-mating and two nonmatingsegregants expected for a reciprocal exchange in the inter-
Val. Another 1.3-2.0% contained a single nonmating segregant, the product of gene conversion of one of the flanking MAT loci. T h e number is not significantly different from what we observed in strains lacking multiple heterologies(Table 2). A second class of aberrant tetrads,in whichthere hadbeen loss of URA3
73
Gene Conversions and Crossovers
and pBR322sequences,couldnot be distinguished genetically from events arisingby gene conversions of URA? to u r d - N . However, colony hybridization with "P-pBR322 detects all such events. All 383 tetrads of RHB752 were screened forloss of pBR322 sequences. Four percent (16/383) of all tetrads contained such an aberrant event. This is also not significantly different from the frequency found with no heterozygosities. The tetrads from RHB633 were not analysed by colony hybridization.However, theSouthern blot analysis of tetrads which were 3-:1+ which was performedto analyze conversion tracts, identifies the aberrant events which involved a MATa-URA3-MATa chromatid. In RHB633 therewere 17 (1.8%)of these. This is a minimum estimate of the frequency of aberrant events because only the aberrantevents involving the MATa Ura+ chromatidswere detected. Some aberrant events, such as intrachromatidalloopoutsand unequal sister chromatid exchanges involving the already Ura- MATa duplications were not detected by this analysis. If we assume that events involving the MATa Ura- chromatids occur at the same frequency as those involving MATa, we estimate that the frequency of pBR322 losses is approximately 3%. Association of crossing over with gene conversion (88/924) of the at uru3-N: Approximately9.5% tetrads of RHB633 exhibited gene conversion at the ura3-N locus (Table 2). These were equally distributed between conversions to yield 3+:1- and 1+:3-. Two of these tetrads contained two ura3-N conversions (one 4+:0- and one 0+:4-) for a total conversion frequency of 9.7% (90/924). Subsequent analysis of the conversion tracts in the 4+:0- and 0+:4-tetrads indicated thatthey were the result of two independent meiotic gene conversions of the ura3-N allele, rather than a mitotic gene conversion. There were no examples of post-meiotic segregation. Because tetrads were directly plated to selective media, post-meiotic segregations are easily detected as sectored colonies (R. H. BORTS, W.-Y. LEUNG,M. S. WILLIAMSON, W. KRAMER,B. KRAMER,S. FOGELand j. E. HABER, unpublished data). The absence of postmeiotic segregation is not unusual (reviewed in FOGEL,MORTIMER and LUSNAK1981). In 17% (15 events) of the cases, it was not possible to determine if there had been an exchange associated with the conversion. Six of these (6.7% of the total conversions) were co-conversions covering the selected URA3 site and one of theflanking MAT regionslocated morethan 3.5 kb away. Southern blot analysis (Figure 2) confirmed that they were indeed long geneconversion events that covered all intervening sites. Five of the ura-N conversion events in the 88 tetrads (6%)were containedin tetrads which hadundergoneanunequalinterchromatidal exchange or a loopout. Four events had two or more conversion tracts on different chromatids, and it was
Xho SmaXho
MAT a
MAT a
PA J
Bgl
Bgl SIMPLE EVENTS m WlTHOlR CROSSOVER cnwsnvm
1
i
, .
25
31
COMPLEX EVENTS class 1 1 1 1 1 3 1
,
;
,
I I
I
A A
A
B B C
c.0
1 1 1 1 1
a
i
,
ig
8 MATCONVERSIONS 1 1 1 1 1
6
I I
I
S
2
,
,
3
,
,
/
>\$W \*
2
;
I I I
I 1
:
h\\V
FIGURE2.-Position and length of NcoI conversions in RHB633. Solid lines indicate the minimumconversion tract for each Ncol conversion. For the simple events the numbers with and without NcoI crossover are indicated to the left. In the complex events, conversions accompanied by a crossover are indicated by a solid line with an X . Open boxes indicate conversions which occurred on one of the chromatids involved in the NcoI conversion. In the cases where the ura3-N conversion was accompanied by a crossover, the open box indicates a conversion on the other chromatid that was involved in the crossover. An X with a square around it indicates a crossover which involved one of the two strands involved in the Ncol conversion and a third strand. An X with a circle or a diamond around it indicates crossovers which occurred on strands not involved with the NcoI conversion. Hatched lines represent conversion tracts that are on strands other than the two that participated in the NcoI conversion. There are four cases (class G) where it is not possible to determine which of two conversion tracts are associated with the crossover. These events are indicated by the presence of ( 9 ' s on both conversion tracts. For many of the complex events, several interpretations were possible, only one of which is presented here.
not possible to determine which tract was associated with the crossover. In73tetrads(75 events), the occurrence of the crossover could be unambiguously determined.43/75(57%) of these conversions at ura3-N were accompanied by an exchange of flanking markers. This frequency is in good agreement with previous measurements showing 30-70% association of crossing over with gene conversions of different loci (FOGEL,MORTIMERand LUSNAK, 1981). Co-conversion events accompanying uru3-N conversions: DNA from all four spores of 83 of the 88
74
R. H. Borts and J. E. Haber
tetrads from RHB633 that displayed a gene conversion at the ura3-N site (Table 2) were analyzed by Southern blot analysis todetermine whichof the other, unselected, restriction site markers were also converted. The five other events, in which one spore was pBR322-, were not examined for this analysis. The results of this analysis are shown in Figure 2. The majority (62/83) of these events appeartobethe result of simple events, in which there was a single, continuous,gene conversion tract(Figure 2). 21 events were morecomplex. Two events hadinterrupted conversion tracts, while a third event had a crossover separated from the ura3-N conversion by unconverted sites (Figure 2, class A). Six of the ura3N conversions which were associated with a crossover had a second conversion on a third strand (Figure 2, class B). Two of these second conversions were of the ura3-N site. In classes C , D, F, and H, the ura3-N conversion was not associated with a crossover. In the two class C events asecond conversion occurred which was associated with a crossover. The class D events also had a second conversion tract. However, in these two cases, the second conversion was not associated with a crossover and occurred on one of the homologues of the strand with the ura3-N conversion. T h e three class E events are three-strand double crossovers. One crossover is associated with the ura3-N conversion and the otheris between one of the strands involved in the ura3-N conversion and a third strand. The class F event had acrossover between two strands not involved in the ura3-N conversion. The four class G events had two or more conversion tracts. Because one of the conversion tracts occuron each homologue, and because there are no unconverted sites between the endpoints of the conversion tracts, which conversion is associated with the crossover cannot be determined. The class H event is very complex. A simple ura3-N conversion occurred on one strand. A second conversion occurred on its sister strand. This conversion was separatedfroma crossover by anunconverted site. A second crossover occurred between this strand and the fourth strand.The class I event is also very complex. In addition to the ura3-N conversion associated with a crossover, the other two chromatids were involved in what appearstobeatwo-strand double crossover. The events obtained fromRHB752 (described below, Figure 4), are similar to the ones obtained in RHB633. Most of the conversion tracts detected in RHB633 covered more thanone site. Of the43 events in which gene conversion of ura3-N was accompanied by a crossover, all but 5 (the 2 simple NcoI conversions, the two XhoI to NcoI and the simple NcoI to Smal conversion) converted at least three sites. Three of these 38 events (two of the class A and oneof the class G events) had conversion tracts which were discontin-
uous. Of the32 conversions which were not associated with a crossover all but two (the class C and class H events) covered three or more sites. None of these thirty conversions (the 25 simple conversions, 2 class D, classF, the class B eventwhich had two conversions spanning the NcoI site and the one, E and B event) were discontinuous. In the three cases where there were apparently discontinuous conversion tracts, it is not possible to determine if these resulted from two independent events or from independent restoration of one region of heteroduplex DNA surrounded by conversion of other markers. Association of geneconversionwithcrossing over: We have previously shown that approximately 60% of the tetrads that contained a reciprocal exchangeofflankingmarkers also harbored agene conversion of one or morerestriction site markers in the interval. These studies used a diploid with 9 heterologies that yielded a high proportion of aberrant events (BORTSand HABER 1987).We have reinvestigated the relationship between crossing-over and gene conversion with a strain that shows fewer aberrant events (RHB633). An additional 328 tetrads from RHB633 (data not shown in Table 2) were dissected in order to identify tetrads containing a reciprocal crossover. A totalof 61 tetrads exhibited one a-mater, one a-mater and two nonmaters (that is, a crossover in the interval) were identified (19%). DNA from each segregant of 54 of the 61 tetrads were analyzed by Southern blots. The results, summarized in Figure 3, demonstrate that approximately 6 1% (33/54, classes A, B, D, E and H) of the crossovers were associated with a detectable gene conversion of one or more sites. Three events (Figure3, class B) had anadditional conversion tract on the same chromatid as the conversion which was associated with the crossover. The single class C event had a crossover separated from the conversion tract by unconverted sites. These four events may represent discontinuous conversion tracts in which the intervening sites were restored, or they may result from independent events. T h e four class D events had a second conversion on the other chromatid which participated in the crossover. Two events (Figure 3, class E) had a second conversion tract on a third strand not involved in the crossover. T h e class F event had two conversion tracts, one on each of the two chromatids participating in the crossover. Both of these are separated from the crossover by unconverted sites. The four class G events had conversion tracts on a third strand. The class H event is apparently a three-strand triple crossover, with one of the crossovers associated with a conversion. T h e 15 class I crossovers had no detectable conversion tract. Whetherthe cases thatdidnotexhibit any gene conversion associated with the crossover (classes F, G and I) reflect events where heteroduplex DNA did
and
Conversions
Gene MAT a
Xho Sma Xho
Crossovers
FOCEL 1974). In the MAT-MAT interval, most twoandthree-stranddouble crossovers are genetically undetectable because two-strand double crossovers give four mating segregants,while three strand double crossovers appear as single crossovers. Four-strand double crossovers, however, are detectable as tetrads with four nonmating spores. Assuming no chromatid interference, the extent of chiasma interference can be estimated from the observed value of four-strand double crossovers compared to the expected value. The expected value of all double crossovers is the square of the frequency of single crossovers. One fourth of these are expected to be four-strand double crossovers. Therefore, for the 1252 tetrads of RHB633 the expected value of four-strand double crossovers in 0.215'/4 = 0.012. The observed value is 0.0024. Thishighly significant difference (P < 0.01) is indicative of a high degree of interference. How interference affects the interpretation of the complex events is addressed in the discussion. Conversiontractlength: We have calculated the conversion tract length for the threeclasses of events identified in RHB633; (1) NcoI conversions with an associated exchange, (2) NcoI conversions without an associated exchange and (3) conversions identified as being associated with a crossover. Single site conversions were assigned a value of 0.005 kb and conversions covering two or more sites were assigned a length of the distance between the outer most converted sites. Simple crossovers were assigned a value of 0 or were not included. We have calculated the minimum average conversion tract of the 31 simple uninterrupted conversion tracts in RHB633 which are accompanied by a crossover (Figure 2) as 1.4 kb & 0.7 kb (mean & SD). This value is 1.3 & 1 forRHB752. All events inwhich there was a second conversion tract, or asecond crossover were excluded from the analysis. Similarly, the conversion tract length of the 25 simple conversions of the NcoI site without an accompanying exchange of flanking markers is 1.6 kb & 0.8 kb for RHB633 and 1.5 f 0.9 for RHB752. These values are not significantly different from those determined for conversions which were associated with crossing over. The average tract length for all conversions is 1.5 kb f 0.7 kb in RHB633 and 1.4 f 1 in RHB752. This is a minimum estimate, as many events could have extended more than 1 kb further without encountering another restriction site. There were five instances in which conversions at NcoI were accompaniedby coconversion of one of the MAT loci (6 cM away, see below) and all intervening sites (Figure 2). The end points of these events are unknown and it cannot be determined if they were accompanied by a crossover. However, their inclusion in the analysis may allow a more accurate estimate of
j MAT a
MAT a
LIRA3
Bgl
Nde
Class 7 1 1
3 2 1
1 1 6 1 1 1 1 1 1 1
1 1 1 1 1
1 1 1
1 7
1 1
6 54
A A
A A A
A A A A B B B C D D D
D E E F G G G
G H I I I I
FIGURE3,"Position and length of conversions from tetrads of RHB633 which were identified on the basis of their having undergone a crossover. Solid lines indicate the minimum extent of conversion associated with each crossover. Crossovers not associated with a conversion tract are indicated by an X without a black bar. Open boxes indicate conversions on the other strand involved in the crossover-associatedconversion. One class (F) had noconversion directly associated with the crossover, but had separate conversion tracts on each of the interacting homologues. These conversion tracts are indicated by light dotted lines. Hatched lines indicate a conversion tract located on a strand not involved in the crossover. X's with squares around them indicate a three strand double crossover. One strand of this crossover is one of the strands of the conversion associated crossover. The number of times each event occurred is indicated to the left.
not cover one of the restriction sites (for example, in the 3-kb interval between NdeI and the distal BgZII site in MATa) or represent instances where the heteroduplex DNA was restored to the parental genotype cannot be assessed from these data. Interferenceoccurs in the MAT-MAT interval: Positive chiasma interference has been defined as the reduction in the expected number of double crossovers. Such interference has been seen in S. cerevisiae (MORTIMERand FOCEL1974) and Neurospora (STADLER 1959). Chromatid interference, where two-, three- and four-strand doubles differ from the expected 1:2: 1, ratio does not occur (MORTIMERand
75
76
R. H. Borts and J. E. Haber TABLE 3
Distribution of the simple conversion tracts of the ura3-N identified conversions NcolStrain
Sall-Ncol“
SalI-Ndelb
Ncol‘
NdeId
Total
RHB633 RHB752
37
11 10
11 8
2
4
924 382
9
T h e conversion tracts of the u r d - N conversions of RHB633 were analyzed as if the XhoI and SmaI sites were not present in order that the distribution ofconversion tracts of the RHB633 could be compared to those of RHB752. Five of thecomplex events of RHB633 (Figure 3, the three class A, one of the class E, and the class 1) were included in this analysis because without the XhoI and SmaI sites they would bescored assimpleevents. None of the complex events from RHB752 were included in the analysis. a Number of conversions that extend from the SalI site to the NcoI site. Number of conversions that extend from the SalI site to the NdeI site. Number of conversions that cover only the Ncol site. Number of conversions that extend from the NcoI site to the NdeI site.
the average conversion tract length. If the BglII site is used as an end point for the conversion tract, a minimal averagetractlengthfor all of the simple ura3-N conversion tracts is 1.7 2 1 kb. We also determined the conversion tract length in the 54 tetrads of RHB633identified solely on the basis of having had a crossover. Eight of these events also contained a ura3-N conversion. The minimum conversion tract length of conversions contained in these 54 tetrads was calculated in two ways; either including crossovers not associated with a conversion event as having a length of 0 kb or not including these “simple” crossovers at all. When the simple crossovers are not included, the minimum average conversion tract for the 23 uninterrupted events (Figure 3, class A events)is 0.7 & 1.O. If the simple crossovers (Figure 3, class H) are included, the tract length is 0.4 kb f 0.8 kb. Both of these values are significantly less ( P 0.01) than that calculated for the NcoI conversions associated with a crossover. However, to make a more direct comparison, we determined the minimum average conversion tract length for the subset of these events that converted the NcoI site. This value, 1.8 k 0.9 kb, is similar to that observed for events which were identified on the basis of having a ura3-N conversion. The reason for the apparent discrepancy in tract length is discussed below. Distribution of coconversion events: The distribution of the ura-N conversion tracts for RHB633 is shown in Figure 2. An analysis ofthe conversion tracts shown in Figure 2 revealed an interesting difference in the frequency with which the SalI and NdeI sites, located approximately equally distant on either side of the NcoI site, were coconverted with NcoI (summarized in Table 3). The SalI site was coconverted 79%(48/61) of thetime, while the NdeI site was
converted only 21 % (13/61) of the time. If one removes the 1 1instances (18%)in which both flanking sites were converted, in 61 % of the events Sal1 was converted withoutNdeI, while in only 3% of theura3N conversions was NdeI converted without SalI. This marked asymmetry is not evident in RHB752 (Figure 4, Table 3)in which the only heterozygosities present are the NcoI, SalI and NdeI sites. Equal numbers of events convert Sal1 but not NdeI as well as the converse, NdeI but not SalI. We conclude that this difference is attributable to the presenceof the three additional heterozygosities present in RHB633. Symmetry of crossing over eventsrelated to ura3N: If gene conversions frequentlyextendtocover markers on one side of the interval and not the other, one might expect to find an equivalent asymmetry in the location of crossings-over. For the conversions of the NcoI site itself, it is not possible to assess if the crossover occurred to the right or the left of the converted locus; therefore one must examine tetrads in which there has not been a conversion of the NcoI site. A total of 36 tetrads from RHB633 containing two non-mating spores and with 2 Ura+ and 2 Uraspores were analyzed by Southernblot analysis to determine the linkage of MATa, MATa and ura3-N in the two non-matingsegregants. There wereequal numbers of crossovers to the left of the NcoI site (17) as to the right (19). Thus, the asymmetry seen in the conversion tracts of NcoI is not reflected in the position of crossovers in tetrads lacking such a conversion. DISCUSSION
We have shown that gene conversion tract lengths are on the order of 1.5 kb whether or not they are associated with a crossing over. Previously, CARPENTER (1 984) presented evidence fromDrosophila that suggested thatgene conversion tracts that are not accompanied by crossing over might be shorter than those that are. If the length and degree of homozygosity of heteroduplex DNA formed early in synapsis were a mechanism for determining if the interacting DNA belonged to homologous chromosomes (CARPENTER 1984, 1987), only those heteroduplexes that were sufficiently homologous might be permitted to extend and be resolved with crossing over. Less perfectly matchedheteroduplexes would therefore be preferentially resolved as conversion events without an associated crossing-over, with shorter average conversion tracts. Recently, a somewhat different homology checking mechanism has been proposed. Based on studies of bacterial mismatch repair systems, RADMAN (1988) suggests that the mismatch repair system may “unwind”heteroduplex DNA containingtoo many mismatches. By this model, one might expect gene conversions, both with and without crossing over, to have few coconversion events. Both of these
Gene Conversions and Crossovers MAT a
77
MAT a
vzi
I
7 MAT a
Sal
Bgl
WlTH CROSSOVER CROSSOVER
4
5
I
3
I
I
I
I
I
I
5
!
8
i
i I !
I i
i
2
I
1
0
I
I
I
!
I
I
I
I
I !
FIGURE4.-Position and length of NcoI conversions from tetrads of RHB752. All symbols are the same as those of Figure 2. The classes listed are the same as those in Figure 2.
I
Class 1
Bgl
!
&
5
I
6
Nde
N ~ O
WrnOUT
A
I
X
' i I !
1
O
B
I I I I
2
O
B
0
1
F
-
-
22
14
i I
; !
I A
! I
l
I
models require that the monitoringof homology permit at least some heterologies to be included in heteroduplex DNA. If no heterology were permitted, crossings over would notbeaccompanied by gene conversions. T h e degree of heterozygosity that would provoke discouragement of a crossover remains unspecified in these hypotheses. The heterozygosities surrounding the ura3-N allele that we have studied include six 4-8-bp insertions or deletions in a 3-kb interval.A similar spectrum of nine heterologies in the entire 9-kb MAT-MAT interval has previously been shown to be sufficient to elicit a significant perturbation of recombination (BORTSand HABER 1987). More recently, we have shown that this perturbation is dependent onthe PMSl mismatch repairfunction (R. H. BORTS,W.-Y. LEUNG,M. S. WILLIAMSON, W. KRAMER,B. KRAMER,S. FOCELand J. E. HABER,unpublisheddata). Inthe absence of heterozygosities close to the flanking MAT regions, there arefew detectable aberrant events; nevertheless we find evidence (see below) that mismatch repair-
induced secondary recombination occurs in the middle of the interval, surrounding the ura3-N allele. Thus, we assume that the level of heterozygosity we have included is sufficient to attract the attention of the mismatch repair system and is therefore probably on the level of that envisaged in the homology monitoring models of CARPENTER (1984, 1987) and RADMAN (1988). If this is the case, we conclude that there is no distinction in the lengths of the gene conversion tracts whetherthey are accompanied by crossing-over. Conversion tract length We have found that conversion tracts are apparently shorter when the event is identified on the basis of containing a crossover. When crossovers without a detected conversion are included as having a tract lengthof zero, the minimum average conversion tract length is 0.4 kb 0.8 kb. If they are excluded from the analysis, the conversion tract length is 0.7 1 kb f 1.O kb. Both of these values are apparently less (P < 0.01)thanthatfortracts identified on the basis of a ura3-N conversion. We suggest that this is due to a difference in the kinds of
78
and R. H. Borts
events which can be recovered in the two separate analyses, and that shorter events can be recovered when crossovers are the identified event. When conversions of the ura3-N site are chosen, they must, by definition, cover at least one site. T h e probability of converting either the neighboring XhoI proximal or SmaI sites is dependent on theconversion tract length. If this is on the order of 1.5 kb, as we suggest, then the probability of recovering a single site conversion in this region is low. In fact, we observed only 4/83 such events amongthe NcoI conversion tracts in RHB633. This hypothesis is supported by the observation that single site conversions of the NcoI site are also rare (0/54) in the tetrads identified by the presence of a crossover, while a quarter (13/54) of the events are single site conversions at SalI or NdeI. One might further postulate that if heteroduplex tracts are distributedrandomly across the interval,then the longer oneshave a greater probability of covering the NcoI site, and a population identified as having an NcoI conversion is therefore enriched forlong events. When crossovers are the identifiedevent, no constraints are placed on the location of an associated conversion (except that it not be one of the flanking MAT loci). Thus, a significant proportion (30%)have no apparent conversion, while another 25% convert a single site. There aretwo other reasons that the probability of recovering single site conversions at the SalI or NdeI sites is higher than the probability of recovering single site conversions of the NcoI site. Firstly, both the SalI and NdeI sites have only one nearest neighbor (the XhoI sites) which could be coconvertedwhile the NcoI site has two, the proximal XhoI site and the SmaI site. Thus, the probability of a coconversion is less for the Sal1 or NdeI site relative to that for the NcoI site. Secondly, as discussed above, the nearest sites to the SalI and NdeI sites are 0.7 and 0.8 kb away respectively, while the nearest sites to the NcoI site are 0.4 and 0.6 kb away. Thus, if the averageconversion tract is 1.5 kb and is randomly distributed, then sites 0.7 and 0.8 kb away are not as likely to be included in a conversion tract as a site 0.4 kb away. Forthese reasons, we suggest that conversion tracts identified as being associated with a crossover are not shorter, rather, measurements of conversion tract length are affected by the method of selection and the distribution of the sites used to measure theevent.This hypothesis is supported by the fact that theconversion tract length of crossover-identified conversions which cover the NcoI site is 1.8 kb k 0.9 kb, which is not significantly different from the conversion tract length of the NcoI identified conversions. In a study of conversion tracts at the rosy locus of Drosophila (CURTISet al. 1989), have foundthat conversion tracts associated with crossovers were ap-
J. E. Haber
parently shorter than those which were not. Based on the data presented here, we suggest that it is possible that this difference in tract length reflects a difference in the criteria forrecovery of the two classes ofevents. Conversion tract length does not appear to be a function of the number of heterozygosities present. When the conversion tract length of events identified on the basisof an NcoI conversion is compared in diploids with three (RHB752) and six heterozygosities (RHB633), there is no difference in conversion tract length (1.4 f 1, 1.5 k 0.7, respectively). This is also true when thetractlength is measuredforevents identified on the basis of a crossover rather than a conversion. Diploids with nine heterozygosities (BORTS and HABER1987) and with sixheterozygosities both give minimumtractlengths of 0.4 or 0.7 kb (depending on whether simple crossovers are included or not) for conversions associated with a crossover. Conversion tract lengths have now been measured in three different intervals of yeast, by two different methods of identification of events and with varying numbers of heterozygosities. SYMINGTON and PETES (1988a) measured conversion tract lengths associated with crossovers which occurred in the 20 kb interval between LEU2 and CEN3. T h e average conversion tract length in this interval was 2.3 kb f 2.9 kb. This value is significantly larger ( P < 0.016, Student’s t test) than the value obtained for theconversion tracts found to be associated with crossovers in the MATMAT interval. JUDD and PETES(1988) measured the tractlength of conversions of the ura3-3 allele at URA3. T h e minimum average conversion tract length was 0.9 kb k 0.8 kb. Although this value is apparently shorter than for conversions of the ura3-N allele in the MAT-MAT interval, it may be an underestimateof the tract length. Because of the distribution of the sites surrounding the ura3-3 allele, conversion tracts between 0.5 kb and 2 kb could not be detected. Marker effects: Evidence that thechoice of markers plays other roles in the analysis of recombination comes from a comparison of the distribution of conversion tracts in RHB752 (Figure 4) with those of RHB633 (Figure 2).RHB633, which contains six heterozygosities, shows a marked asymmetry of the location of conversion tracts which is not evident in RHB752, which has only three heterozygosities. We suggest that the presence of the three additional heterozygosities surrounding the NcoI site in RHB633 affects the repair of heteroduplex DNA. One possibility is that the Sal1 site is more frequently included in asecondaryrepairevent, andthus appears to convert in a high proportion of events. For example, the heteroduplex DNA which yields apparent single site conversions of the NcoI allele in RHB752 may normally extendthroughthe XhoI proximal site (which cannot be determined without heterozygosity
Crossovers
and Gene Conversions
atthe site), butnottothe SalI site. If repair of heterozygosity at theXhoI proximal site often extends to the SalI site or leads to a second repair induced recombination event which includes the SalI site, a very different spectrum of conversion tracts will be recovered when the XhoI site is heterozygous as compared to when it is homozygous. In fact, if only half of the NcoI conversions of RHB752 extend as far as the XhoI proximal site and are secondarily extended to the SalI site, the distribution of events found in RHB633 would be generated. A second possibility is that theremay be as many heteroduplex tracts formed between the NcoI site and theNdeI site as between the NcoI site and the SalI site in RHB633 as in RHB752, but the inclusion of the distal XhoI site or the SmaI sites may lead to restoration of these tracts,rather than conversion. Marker effects in which the repair of one allele changes the repair pattern of an adjacent allele have been well established in yeast (FOGEL,MORTIMER andLUSNAK1981)andother fungi(LEBLON and ROSSICNOL 1973; NICOLAS and ROSSIGNOL 1983). A third explanation for the asymmetry is that the closely spaced SmaI and XhoI heterologies interfere with the formation of heteroduplex DNA at the initiation of the event. SYMINCTON and PETES(1988a) have also noted a nonrandom in the distribution of conversion tracts and crossovers in the LEU2-CEN3 interval. The clustering of crossover points was interpreted as evidence for “hotspots” of recombination. It is possible that such clustering of crossovers is a result of the placement andrepair of the heterozygosities employed in that study. Complex events: In RHB633 and RHB752, a significant proportion of the events recovered are complex. In RHB633, 16/54 (30%) events are complex, when crossovers are the identified event. When ura3N conversions are the identified event 21/83 (25%) events are complex. T h e value is not significantly lower (5/36, 14%) in RHB752, the diploid with only three heterozygosities. One explanation of the complex events is that they resultfrommultiple, independent, events. The complex events which consist of multiple conversion tracts are consistent with classical geneticobservations (MORTIMER and FOGEL1974) that one conversion event without crossing over does notinterfere with anindependent crossover. However, classical geneticdata (MORTIMER and FOGEL 1974) do show interference for two intergenic crossings over. The frequency withwhich four-strand crossovers are recovered in RHB633 (0.24%) is less than the expected frequency forno interference ( P < 0.01) indicating that there is interference. Although the numbers aresmall, the frequency with which twoand three-strand doublecrossovers (4, oneof the class A, and the three class E) are recovered among the 83 ura3-N conversion-identified events does not show the
79
same degree of interference as the four-strand double crossovers. This implies that the population chosen as having a conversion is enriched for doublecrossovers ( i e . , shows no interference). We suggest that this outcome is a prediction of the model inwhich the repair of heteroduplex DNA containing multiplemismatches leads to a second round of recombination. Complex events such as those found in RHB633 have also been found in an analysis of recombination in the LEUZ-CEN3 interval (SYMINGTON and PETES1988a) at a similar frequency(approximately 20%). These events may also not be independent and may arise as a result of multiple rounds of recombination induced by the multiple heterologies used in the study. Interrupted conversion tracts and crossovers separate from the conversion tracts are another class of complexevents. In this study, three out of the 13 (23%) events (Figure 3, class B) and 6 out of the 67 events ( l o % , Figure 2, two of the class A, E, and one of the class G events) which could have shown an interrupted conversion tractwere, in fact,interrupted. In a previous study (BORTSand HABER 1987) fourout of 12events (25%) were interrupted. In other studies (JUDD and PETES1988; SYMINGTON and PETES1988a), the number of interrupted conversion tracts was 3/33 and 012, respectively. The independence of these eventsis difficult to assess without knowing the frequency of conversion for each site. However, thefrequency with which one of the other three strands has a second conversion, which is presumably independent, can be determined. There are 18 such events (Figure 2, classes B, C, D, G, H and two of the MAT conversions) among the 83 ura3-N conversion events. If the probability of a second conversion is random, then one would expect three times as many second conversions on a non-ura3-N conversion strand as on the ura3-N conversion strand. There are three events (Figure 2, two of the class A events and one of the class G events) where there is a second conversion tract on the same strand as the ura3-N conversion. The observed ratio of 18:3 is consistent with the conversion tracts being independent events. There are two complex events (Figure 3, class H and Figure 2, class I) which can be explained as closely spaced double crossovers. Alternatively, they may result from the formation of symmetric heteroduplex DNA followed by independent repair on both homologues. Althoughthese two possibilities cannotbe distinguished, such events have classically been interpreted as evidence for the formation of symmetric heteroduplex (reviewed in FOGEL,MORTIMER and LUSNAK 1981).We also note that the class D and G events (Figure 2) which have a second conversion on the interacting homologue could also arise from symmetric heteroduplex. The data presented shed light on a number of the
R. H. Borts and J. E. Haber
80
current problems in recombination. Evidence is presented that conversion tract length is not different for conversions that are associated with crossover and for those which are not. It is, however, a function of the method of selection of the events measured. Conversions events which are identified first on the basis of a crossover appear shorter than those identified on the basis of conversion of a particular site. Finally, evidence is presented that some apparently complex events result from two independent events,while others do not.Interrupted conversion tracts are probably the result of independently initiated heteroduplex tracts. However, many of thecomplex crossover events found in strains with multiple heterozygosities may notrepresentindependentevents. We suggest that they are the result of a second roundof mismatch repair induced recombination stemming froma single initiation event. We thank EDWARD J. LOUISfor help with the statistical analysis and for helpful discussions. We thank MICHAEL LICHTEN,DOUG J. Lours for critically reading BISHOP,FRANKSTAHLand EDWARD the manuscript. We also thank the anonymous reviewers for their careful reading of the manuscript. This work was supported by grant GM29736 from the National Institutes of Health and grant 1-968 from the March of Dimes Birth Defects Foundation.
LITERATURE CITED ANDREADIS,H., Y-P. Hsu, G. HEMODSON,G . KOHLAWandP. SCHIMMEL, 1984 Yeast LEU2: repression of mRNA levels by leucine and primary structure of the gene product. J. Biol. Chem. 2 5 9 8059-8062. BOEKE,J. D., F. LACROUTEand G. R. FINK, 1984 A positive selection for mutants lacking orotidine 5”phosphate decarboxylase activity in yeast. Mol. Gen. Genet. 197: 345-346. BORTS,R. H., and J. E. HABER,1987 Meiotic recombination in 237: yeast: alteration by multipleheterozygosities.Science 1459-1463. BORTS,R. H., M. LICHTENand J. E. HABER,1986 Analysis of meiosis-defective mutations in yeast by physical monitoring of recombination. Genetics 113: 55 1-567. and J. E. BORTS,R. H., M. LICHTEN,M. HEARN,L. S. DAVIDOW HABER,1984 Physical monitoring ofmeiotic recombination in Saccharomyces cerevisiae. Cold Spring Harbor Symp. Quant. Biol. 49: 67-76. BRUSCHI,C., A. R. COMERand G. A. HOWE, 1987 Specificity of DNA uptake during whole cell transformation of S. cerevisiae. Yeast 3: 131-137. CARPENTER, A. T. C., 1984 Meiotic roles of crossing-over and of gene conversion. Cold Spring Harbor Symp. Quant. Biol. 49: 23-29. CARPENTER, A. T. C., 1987 Gene conversion, recombination nodules, and the initiation of meiotic synapsis. BioEssays 6: 232236. CLARK, A. CHOVNICK and W. BENDER, CURTIS, D., S. H. 1989 Molecular analysis of recombination events in Drosophila. Genetics 122: 653-661. and K. LUSNAK,1981 Mechanisms of FOGEL,S., R. K. MORTIMER meiotic gene conversion, or “wandering on a foreign strand,” pp. 289-293 in The Molecular Biology of the Yeast Saccharomyces: Life Cycle and Inheritance, edited by J. N. STRATHERN,E. W. JONESand J.R. BROACH. Cold Spring Harbor Laboratory,Cold Spring Harbor, N. Y. FOGEL, S., R. K. MORTIMER, K. LUSNAKand F. TAVARES, 1979 Meiotic gene conversion: a signal of the basic recombi-
nation event in yeast. Cold Spring Harbor Symp. Quant. Biol. 43: 1325-1341. GLICKMAN, B. W., and M. RADMAN, 1980Escherichiacoli mutator mutants defective in methylation-instructed DNAmismatch correction. Proc. Natl. Acad. Sci. USA 77: 1063-1067. HASTINGS,P. J.. 1984Measurement of restoration and conversion: its meaning for themismatch repair hypothesis of conversion. Cold Spring Harbor Symp. Quant. Biol. 49: 49-54. HURST,D. D., S. FOGELand R. K. MORTIMER,1972 Conversion associated recombination in yeast. Proc. Natl. Acad. Sci. USA 6 9 101-105. JUDD, R. S., and T. D. PETES, 1988 Physical lengths ofmeiotic and mitotic gene conversion tracts in Saccharomyces cerevisiae. Genetics 118: 40 1-4 10. LEBLON,G.,and J.-L. ROSSIGNOL,1973 Mechanismsof gene conversion in Ascobolus immersus. 111. T h e interaction of heteroalleles in the conversionprocess. Mol. Gen.Genet. 122: 165-182. LICHTEN,M., R. H. BORTSand J. E. HABER, 1987 Meiotic gene conversion and crossing-over between dispersed homologous sequences occur frequently in Saccharomyces cerevisiae. Genetics 115: 233-246. MANIATIS,T., E. F. FRITSCHand J. SAMBROOK, 1983 Molecular Cloning: A Laboratory Manual.Cold Spring Harbor Laboratory, Cold Spring Harbor, N. Y . MORTIMER, R. K., and S. FOGEL,1974 Genetical interference and gene conversion, pp. 263-275 in Mechanisms of Recombination, edited by R. F. GRELL. Plenum,New York. NICOLAS,A., and J.-L. ROSSIGNOL, 1983 Gene conversion: point mutation heterozygosities lower heteroduplex formation. EMBO J. 2: 2265-2270. ORR-WEAVER, T. L., and J. W. SZOSTAK,1985 Fungal recombination. Microbiol. Rev. 49: 33-58. SZOSTAK and R. J. ROTHSTEIN, ORR-WEAVER,T. L., J. W. 1983 Genetic applications of yeast transformation with linear and gapped plasmids. Methods Enzymol. 101: 228-245. RADMAN, M., 1988 Mismatch repair and genetic recombination, pp. 169-192 in GeneticRecombination, edited by R. KUCHERLAPATI and G. R. SMITH. American Society for Microbiology, Washington, D. C . ROSE,M., P. GRISAFIand D. BOTSTEIN,1984 Structure and function of the yeast URA3 gene: expression in Escherichia coli. Gene 29: 1 13- 124. SHERMAN, F., G. R. FINKand J. B. HICKS,1986 Methodsin Yeast Genetics. Cold Spring Harbor Laboratory,Cold Spring Harbor, N. Y. SOKAL,R. R., and F. J. ROHLF, 1969 Biometrics. W. H. Freeman, San Francisco. SOUTHERN, E. M., 1975 Detection of specific sequences among DNA fragments separated by gel electrophoresis. J. Mol. Biol. 98: 503-517. STADLER,D. R.,1959The relationship of gene conversion to crossing-over in Neurospora. Proc. Natl. Acad. Sci. USA 45: 1625-1629. SYMINGTON, L. S., and T. D. PETES,1988a Expansions and contractions of the genetic map relative to the physical map of yeast chromosome 111. Mol. Cell. Biol. 8: 595-604. SYMINGTON, L. S., and T. D. Petes, 1988b Meiotic recombination within the centromere of a yeast chromosome. Cell 52: 237240. WALLACE, R. B., and C. G. MIYADA, 1987 Oligonucleotide probes for the screening of recombination DNA libraries. Methods Enzymol. 152: 432-442. WILLIS,K.,and H. KLEIN, 1987 Intrachromosomal recombination in Saccharomyces cerevisiae: reciprocal exchange in an inverted repeat. Genetics 117: 633-643. Communicating editor: G. S. ROEDER
