Two Alternative Pathways of Double-StrandBreak ... - Europe PMC

MOLECULAR AND CELLULAR BIOLOGY, Mar. 1992, p. 1292-1303 0270-7306/92/031292-12$02.00/0 Copyright ©) 1992, American Society for Microbiology

Vol. 12, No. 3

Two Alternative Pathways of Double-Strand Break Repair That Are Kinetically Separable and Independently Modulated JACQUELINE FISHMAN-LOBELL, NORAH RUDIN,t AND JAMES E. HABER* Rosenstiel Center and Department of Biology, Brandeis University, Boston, Massachusetts 02254-9110 Received 20 September 1991/Accepted 23 December 1991

HO endonuclease-induced double-strand breaks in Saccharomyces cerevisiae can undergo recombination by two distinct and competing pathways. In a plasmid containing a direct repeat, in which one repeat is interrupted by an HO endonuclease cut site, gap repair yields gene conversions while single-strand annealing

produces deletions. Consistent with predictions of the single-strand annealing mechanism, deletion formation is not accompanied by the formation of a reciprocal recombination product. Deletions are delayed 60 min when the distance separating the repeats is increased by 4.4 kb. Moreover, the rate of deletion formation corresponds to the time at which complementary regions become single stranded. Gap repair processes are independent of distance but are reduced in rad52 mutants and in Gl-arrested cells. In both mitotic and meiotic cells, double-strand breaks (DSBs) can initiate homologous recombination events. In the yeast Saccharomyces cerevisiae, this process has been characterized by analyzing the outcome of repair of DSBs induced by X rays, transformation of linear substrates into cells, and introduction of site-specific lesions in both mitotic and meiotic cells (1, 18, 36, 37, 44, 54). A mechanism for the repair of DSBs was originally proposed by Resnick (43) on the basis of X-ray induced events and was later extended by Szostak et al. (57) on the basis of transformation studies with linearized and gapped DNA. The gap repair model (Fig. 1A) proposes that a DSB or gap is repaired by a gene conversion event involving the concerted participation of both broken ends, which invade and copy DNA from an intact, homologous donor copy. More recent considerations have prompted a modification of this model to include the creation of extensive single-stranded regions by 5'-to-3' exonucleolytic digestion of the DSB (55, 63). This model predicts that gene conversion can occur both with and without an accompanying crossing over and that these two products occur with equal frequencies (Fig. 1A). Similar recombination events have been described for mammalian cells. Linearized DNA molecules can be used to effect site-directed gene replacement (13, 27, 50) and gap repair of transformed DNA (2). Recently, a system that carries out gap repair in vitro has been developed; however, some of the products obtained had not been predicted by a simple gap repair model (15). A number of studies of recombination in mammalian cells suggest that there is an alternative pathway of recombination, termed single-strand annealing (SSA, Fig. 1B) (3, 6, 21-23, 61). In this pathway (Fig. 1B), deletions are formed between sequences in direct orientation. A DSB produced between the directly repeated sequences is attacked by a 5'-to-3' exonuclease whose degradation produces complementary single strands which then anneal. This mechanism also accounts for products resembling gene conversions that have arisen by the interaction of several DNA molecules (22). Molecular evidence in support of SSA has come from examining molecular inter-

mediates of recombination of DNA injected into Xenopus oocytes (29-31). With yeast, the molecular events associated with recom-

binational repair of DSBs can be observed in vivo by physically monitoring DNA after the synchronous induction of a DSB. With mitotic cells, it has been possible to carry out a detailed analysis of DSB repair by using two highly efficient site-specific endonucleases: the HO endonuclease, which is responsible for initiating mating-type switching (19, 51) and mitochondrial I-SCE1 endonuclease, which is involved in the propagation of the + intron (12, 65). Galactose-induced expression of these endonucleases results in the synchronous cleavage of the target site in a majority of cells in the population, so that it is possible to identify a series of molecular events during the repair process. Recombination initiated either by HO or by I-SCE1 cleavage is surprisingly slow, taking 30 to 60 min for products to be formed once the molecules are cut (5, 35, 39, 40, 47, 48, 63). Analysis of recombination intermediates produced during HO endonuclease-induced switching of the yeast mating type locus has shown that HO endonuclease cut DNA undergoes a slow but extensive 5'-to-3' exonuclease degradation that yields a long 3' single-stranded tail (63). Subsequently, the 3' single-stranded end of the MAT locus has been shown to invade the intact donor locus and begin copying the donor sequences by primer extension; these intermediates have been detected approximately 30 min before the recombination process is complete (63). The slowness of some of these steps may reflect the requirement for new protein synthesis after the DSB is created (47). In this work and a previous study (48), we have analyzed DSB-mediated recombination by using a centromere-containing plasmid with two copies of the Eschenchia coli LacZ gene in direct orientation. One copy contains wild-type sequences but lacks a promotor, and the second is an expressed copy of the gene that is disrupted by a DNA fragment containing the HO endonuclease cutting site (Fig. 2). Initiation of recombination was accomplished by inducing the expression of the HO endonuclease gene, which is under the control of the galactose promoter (14). Both LacZ+ recombinants that are gene conversions without crossing over and deletions (Fig. 2) were recovered; however, physical and genetic analyses of recombination products with this system did not fully conform to the predictions

* Corresponding author. t Present address: Department of Justice, Lawrence Berkeley Laboratory, Berkeley, CA 94720.

1292

VOL. 12, 1992

TWO INDEPENDENTLY MODULATED RECOMBINATION PATHWAYS

A

1293

B LACZ

LACZ

CEN4

LAARC DSB

Gene conversion without crossing-over Gene conversion with crossing-over

i Deletion FIG. 1. The gap repair model and the SSA models of DSB-initiated recombination. (A) In the gap repair mechanism, the 3' ends of the break become single stranded by a 5'-3' exonuclease. The single-stranded ends invade a homologous duplex and fill the gap by DNA synthesis. Cutting and ligating one pair of strands of the Holliday junction intermediate resolve the intermediate and yield products either with or without crossing over. (B) In the SSA mechanism, 5'-3' degradation occurs at the ends of a break and continues until complementary homologous regions become single stranded. The complementary strands then anneal. This pathway produces only one type of product, a deletion product, and removes DNA between the direct repeats.

of the gap repair model (48). On the basis of these results, we suggested that a large fraction of the recombination events which we observed were the result of a "one-ended" event in which only one of the two ends of the DSB participated in a recombination event. An attractive mechanism to account for one-ended recombination events is the SSA mechanism (Fig. 1B) proposed by Lin et al. (21, 23). The nonreciprocal nature of the SSA mechanism can explain the absence of the reciprocal circular product accompanying the formation of deletions. The SSA mechanism also accounts for the highly efficient deletion repair of HO-induced DSBs between directly repeated sequences on the chromosome (47). These deletion events occur even when the homologous sequences are more than 15 kb apart and are separated by sequences in inverted orientation that could also recombine (48). Recently, Ozenberger and Roeder (38) have invoked SSA to explain RAD52independent repair of HO-induced DSBs in tandemly repeated genes. In this paper, we show the kinetics of both gene conversion and deletion product formation of HO-induced DSB in the LacZ plasmid substrate. We found that deletions and gene conversions without crossing over occur at significantly different rates. Moreover, we show that the kinetics and proportions of the two products can be independently modulated in three ways. The first is by changing the distance separating the two repeated sequences, the second is by allowing repair to occur in only one phase of the cell cycle,

and the third is by initiating DSB repair in radS2 mutants. Taken together, these data provide the first direct evidence of two competing pathways of DSB repair (SSA and gap repair) and indicate that a common intermediate is partitioned between them. Our results also argue that gap repair processes between homologous regions in direct orientation are rarely resolved with an associated crossover.

MATERIALS AND METHODS Strains. The isogenic yeast strains used in this work are derived from strain DBY745. They are heterothallic and have been deleted for MAT and HML; therefore, upon production of the HO endonuclease, no mating type switching occurs. tJF6, tNR109, and tJF406 were obtained by transforming strain R175 (ho HMLa leu2 mat::LEU2 hmr-3A ma12 ura3-52 thr4 trpl GAL and the TRP GAL-HO plasmid pSE271:GAL10:HO [34]) with pJF6, pNR41, and pNR43, respectively, by using lithium acetate (11). A radS2 strain was provided by Neal Sugawara. This strain was created by disrupting the RAD52 gene with the THR4 gene (53). This strain was transformed with both pSE271:GAL10:HO and pJF6 to create tJFL6-71. Plasmids. All strains carry pSE271:GAL:HO. Plasmid pJF3 was constructed by inserting a 117-bp XhoII MATa cut-site fragment into the BclI site of a LacZ gene which contains a composite UAScycl-LEU2 yeast promoter in pJH262 (10). The HindIII fragment of pJF3 containing URA3

FISHMAN-LOBELL ET AL.

1294

A

MOL. CELL. BIOL.

B

pJF6

C

pNR41

CEN4 ARS URA3 S

P

r

A, S.6kb

IacZ

4.4 kb

CEN4

\;; ~~~~PrmtZer tecZ \+

site

pJFL28

D CEN4

ARS

6

SA kb

IacZ

BCRV Bs ii i

URA3

\; 5.6kb

H

i

~~~~~Si

_

4.3 kb

LACZ Sa

ARS

P C

~~~~s;

z

E

CEN4 URA3

acut~~2cusltes 4.4kb

2.2 kb

r

\URA3 \ARS~.OGC

/

r~~~~~~~~

H

pNR43

h

4.0kb

7

c Z. HP

IaCZ

site

ARSi Gene Conversion

Deletion

FIG. 2. Structure of plasmids containing a direct repeat of LacZ sequences. A transcribed LacZ gene is interrupted by the insertion of a 117-bpMA4Ta HO endonuclease cut site at the BclI site (48) (see Materials and Methods). The second LacZ gene lacks a promoter but contains wild-type sequences. (A) Gene conversion and deletion products from the substrate pJF6. The 2osition of restriction sites for PstI (P), HindIII (H), and SmaI (S) are also shown. There are additional restriction sites that are not shown. Placement of the 2.2-kb phage A DNA insert (pNR41) (B) and the 4.4-kb E. coli DNA insert (pNR43) (C) into the direct repeat plasmid (E) are shown. Panel D shows the placement of the ARSI sequence between the LacZ genes, and panel E is a map of the LacZ gene. Bs, Bsu36I, Sa, SacI, Bc, BclI RV, EcoRV. A 1.7-kb Bsu36I-SacI isolated fragment spanning the BclI site was used to probe Southern blots, and a 0.9-kb Bsu36I-EcoRV site was cloned into pGem3 for the single-stranded RNA probes. Arrows indicate orientation of LacZ genes.

UAScycl- and 300 bp of the LEU2 gene promoter and LacZ::CS cut site was inserted into the HindlIl site at the upstream end of a promotorless LacZ gene on plasmid pJH271, (48) producing pJF6. The promoterless LacZ plasmid contains CEN4 and an ARS sequence derived from histone genes. pJF6 contains this LacZ duplication in a direct repeat orientation. Plasmids pNR41 and pNR43 were obtained by partially digesting pNR18 (which is the same as pJF6 only pNR18 contains two tandem HO endonuclease cut sites) with HindIII and inserting HindIll fragments of and bacterial DNA. pNR41 contains the 2.2-kb phage X HindIlI fragment, and pNR43 contains an unknown 4.4-kb E. coliderived fragment between the LacZ duplications. The LacZ::cs promoterless plasmid was made by cutting pJF3 with SmaI and MstII, which cuts 236 bp into the LacZ gene, filling in the ends by using Klenow, and religating. The Hindlll fragment of this new plasmid (pJF11) containing URA3 and promoterless LacZ::cs was ligated into the HindIII site of pJH271, as described above, to make plasmid pJF12. A 1.4-kb EcoRI fragment containing TRPI and ARSI was ligated into the EcoRI site of LacZ to make pJL25. The HindIlI fragment of this plasmid containing URA3, LacZ:: cs, and ARSI was ligated into the HindIII site of pJH271 to obtain pJL28. Media. Cells were grown in YEP (1% [wt/vol] yeast extract, 2% [wt/vol] Bacto Peptone) supplemented with glucose (2% [wt/vol] YEPD) or lactic acid (3.15% [wt/vol] [pH 5.5] YEPL). To induce the production of the HO

endonuclease, galactose (2% [wt/vol]) was added to YEPL medium. Synthetic complete medium lacking uracil and tryptophan (SC-ura-trp) was made according to Sherman et al. (49). Bacto Agar was added (2% [wt/vol]) for solid medium. Induction of recombination. YEPL medium was inoculated with a stationary-'phase SC-ura-trp culture to a final concentration of 5 x 10 cells per ml. The culture was vigorously shaken at 30°C overnight to a final cell density of 1 x 107 to 3 x 107 cells per ml. An aliquot of cells was removed for the zero time point, after which galactose (1/10 volume of 20% [wt/vol]) was added to the remainder and incubation was continued. Cells were diluted and plated onto YEPD plates, anc these colonies were replica plated onto SC-ura and SC-trp plates. Colonies were analyzed for plasmid maintenance and recombinant products. Induction of a synchronous population of cells was carried out by the addition of ax-factor (1 mg/ml in ethanol; Sigma Chemical Co.) to exponentially growing YEPL cultures to a final concentration of 2.5 ,ug/ml. Incubation was continued until >90% of the cells were unbudded, indicating that the majority of the cells were in the G1 phase. The cells were washed out of the a-factor and into galactose medium, effectively releasing the arrest and inducing production of the HO endonuclease. To maintain the cells in the G, phase during induction and repair, additional aliquots of a-factor were added to synchronized cultures every 90 min. Additionally, the cells were monitored microscopically at each time point.


VOL. 12, 1992

1295

13

A

lboLli-S I .>5 2 I

Ob . 5 -5.6

L1)

4

Lb l)ektion s,48 Sl

a

1)onor

t Site

4.4 1LIb

klb

A4..

iut

1

aeo

1.S Li) cut

2

_

2.6

I I otlolS 1. 2

6,

ILb (ant1 ui'f)lUIsio0

Hours

5.6 kbl) 4.5

1.5 4 6

onor

Lb1) Culst

Site

2.6 Li) ctit

L)

1.8

_lea

I

4.5

1)

2.6

N

C

1

5.6

__

5.8

kb Deletion

4.3

kb)

Gene

conversion

.5

4

6

)onor

(

kl)

ot Site

tit

5.8 Lb

D)eltiion

- eam - - L *F4.3R khl G;t'11 ('IINCI\+'SioIl

I

I

!.. 4L 2

ki) cm

FIG. 3. Kinetics of DSB repair. Cells were grown in YEP-lactate medium to mid-log phase, and the HO endonuclease was induced by the addition of galactose. DNA was then extracted from the cells at 30-min intervals after induction and was digested with PstI, HindIII, and SmaI before being subjected to Southern blot analysis by using the 1.7-kb Bsu36I-SacI internal LacZ fragment as a probe (Fig. 2E). Autoradiograms of a time course experiment from the strain containing the direct repeat plasmid pJF6 (A) and of an induction of a strain containing the direct repeat plasmid pNR41 with a 2.2-kb insert (B) and the direct repeat plasmid pNR43 with a 4.4-kb insert (C). Plasmids pNR41 and pNR43 contain two cut sites inserted into the BcII site. Coincident cutting of both cut sites leads to the doublet bands seen in these blots. We have shown that the inclusion of two cut sites does not change the outcome of the events (this work) (48).

I

exposed to blots without screens was carried out by using a Bio-Rad model 620 video densitometer. 1.8

kl)

ct

2t

Scoring 1-galactosidase activity. Scoring of ,-galactosidase activity was carried out by a method developed by Plessis et al. (39). By this procedure, colonies were replica plated onto a 7-cm-diameter Whatman 3MM filter paper circle, which was placed in a liquid nitrogen bath. The filter was then placed in a petri dish over which Na phosphate assay buffer (pH = 7.0) containing 0.01 M KCl, 0.001 M MgSO4, 2% X-Gal (5-bromo-4-chloro-3-indolyl-o-D-galactopyranoside), and 0.7% agar was poured (32). Colonies were assayed within 2 h for blue color. Southern analysis. DNA extraction was carried out as described by White and Haber (63). Restriction endonuclease-digested DNA was electrophoresed in neutral 40 mM Tris-acetate-2 mM EDTA (26) agarose gels and transferred to nylon membranes (Duralon; Stratagene) by using a model 2016 Vacugene vacuum blotting pump (Pharmacia LKB) and 1 M NH4CI-0.02 M NaOH. The membranes were probed with 32P-labelled probes prepared by using the random primer method (8) with either the Bio-Rad or Boehringer Mannheim kit. DNA restriction fragments for random primer probe synthesis were isolated from agarose gels, and membranes were probed in aqueous solutions as described by Church and Gilbert (4). Densitometry of autoradiograms

RESULTS

Deletion and gene conversion products are formed with different kinetics. To determine the kinetics with which products are produced in the direct repeat construct pJF6 (Fig. 2A), we carried out Southern blot analysis of DNA extracted at intervals after induction of the HO endonuclease. The DNA was digested with PstI, HindIII, and SmaI, in order to separate the donor and HO endonuclease cut site containing LacZ genes, the HO endonuclease cut intermediates, and the deletion and gene conversion products. In approximately 60% of the cells, the 4.4-kb HindIII-SmaI LacZ fragment containing the HO endonuclease cut site was cleaved into 2.6- and 1.8-kb fragments within 30 min of HO induction (Fig. 3A). The 5.8-kb deletion products first appeared 30 min later (Fig. 2A and 3A). In contrast, the 4.3-kb gene conversion product was not visible until 60 min after HO endonuclease cut molecules were formed (Fig. 2A and 3A). Moreover, there was no evidence, at any time, of the formation of the 4-kb reciprocal crossover product that would be expected from a gap repair intermediate resolved with an exchange. With the restriction digest used for Fig. 3, the reciprocal product would be a linear 4-kb molecule containing only LacZ and phage X DNA sequences. The difference in the kinetics of formation of the deletion and gene conversion products is highly reproducible. Densitometric analyses of the kinetics of product formation from four independent experiments are shown in Fig. 4A. As

MOL. CELL. BIOL.


1296

TABLE 1. Fate of Lacz plasmid after HO induction

A

b~~~~~~~~~~~~~~~~~~~

60

40

30 20

-

% of products that wereb:

(plasmid)a

conversions

Deletions

HO-induced plasmid loss' (%)

No Insert (pJF6) 2.2-kb insert (pNR41) 4.4-kb Insert (pNR43)

17 (22/133)

83 (111/133)

21

24 (31/130)

76 (99/130)

17

42 (80/192)

58 (112/192)

23

Direct repeat construct

Gene

20%

a The structures of the Lacz plasmids from HO-induced independent Lac', Ura+, and Trp+ colonies were determined by Southern analysis. b The number of gene conversions or deletions per total number of products is given in parentheses. c The HO-induced Lacz plasmid loss was determined by subtracting the proportion of Ura+ Trp+ colonies at the end of the experiment from the proportion of Ura+ Trp+ colonies at the beginning of the experiment.

10-

B~~~~~~~~~~~~~~

Gene Conversion

0

0.0

4.0

2.0

40

71%

10-

etion -~~~~~~~~~~~

20

29% 10-

GnConversion

0.0

2.0

6.0

4.0

C so

40-

10 Deletion

20

gio~~~~~~~~~~~~3 10

Gene Conversion

0

0.0

~ 2.0

4.0

6.0

time (hours)

FIG. 4. Densitometric analysis of DSB repair. Densitometric of product formation from four independent experiments for each construct were integrated and normalized to the sum of all LacZ-hybridizing DNA, and the mean and standard deviation were plotted (see Materials and Methods). Analysis of pJF6 (A), pNR41 (B), and pNR43 (C). scans

measured by densitometry, 80% of the products were deletions and 20% were gene conversions. This result is consistent with our analysis of the plasmid structure in individual Lac' colonies, in which 83% were deletions and 17% were gene conversions (Table 1) (48). Tests of the SSA model for the formation of deletions between direct repeats. The differences in the time of appearance of the deletion and gene conversion products, the disparity in their relative proportions, and the absence of the expected reciprocal circular crossover product all argue that there might be two different pathways by which this sub-

strate is repaired. Deletions could arise by an SSA pathway (Fig. 1B), which does not produce a reciprocal product, while gene conversions without crossing over could occur by a gap repair pathway (Fig. 1A). The SSA mechanism predicts that extensive singlestranded degradation occurs in both directions from the DSB until complementary single strands are exposed and then anneal. Thus, the kinetics of deletion formation would depend on the time required for the two homologous regions to become single stranded. Since the rate of 5'-to-3' exonuclease digestion is slow, on the order of 1 to 2 nucleotides per s (53, 63), we predicted that if the distances between the direct repeats were increased, the kinetics of deletion product formation would be slowed because of the additional time necessary to degrade the intervening DNA. To test this prediction, we used two plasmids that contained 2.2 and 4.4 kb of DNA inserted between the direct repeats (pNR41 and pNR43, illustrated in Fig. 2B). Southern blots and densitometric analyses of these blots comparing the kinetics of product formation for all three constructs are shown in Fig. 3 and 4. When the distance between the direct repeats was increased, there was a profound change in the kinetics with which deletion products were formed, but there was no change in the kinetics of formation of gene conversion products. In the construct with a 2.2-kb insert, deletion products appeared later than in the construct with no insert, so that instead of appearing 30 min before gene conversion products, deletions were now formed simultaneously with gene conversion products. In the construct with a 4.4-kb insert, the kinetics of deletion product formation were slowed even further, so that deletions appeared 30 min after the gene conversions. These results are consistent with the dependence of deletion formation on the action of a slowly moving (1 to 2 nucleotides per s) 5'-to-3' exonuclease; such an exonuclease should require about 60 min to degrade 4,400 bp of additional DNA. Increasing the distance between the repeats also statistically significantly altered the relative proportions of the two products as measured by the z test. The proportion of gene conversion products in the construct without an insert was 17%, which is significantly different from the 42% gene conversions in the construct with the 4.4-kb insert. The construct with the 2.2-kb insert produced an intermediate level of gene conversions (24%) which is not significantly different from either the larger insert or the no-insert plasmid (Table 1). However, this same intermediate value is also

VOL. 12, 1992


A

O

.5

Hours 2 2.- 4 1 1.

C 6

.5

0

Hours 1 1.5 2 2.5 4

Native

&

*

0

Denalured

l l l

l

l

1297

6

*-

Native

B

O

.5

1 1.5

2 2.5 4

6

Native

D

.5

1 1.5

2

4 2.5

4

6

NN-'IliC' ati

Denatured

ID

0 0 0 0 -.

I..

I.,

(G

FIG. 5. Analysis of single-stranded degradation. Only regions of the plasmid made single stranded by an exonuclease will hybridize with complementary single-stranded probe. If exonuclease activity occurs in the 5'-to-3' direction, only the probe complementary to the strand ending 3' will hybridize to the native DNA, while the probe complementary to the degraded strand will not hybridize. Dot blot analysis of undenatured DNA from HO-induced and uninduced cells probed with single-stranded probes of the 4.4-kb insert (A and B) and the URA3 gene (C and D). Regions probed with strands complementary to those which end 3' at the HO endonuclease cut (A and C) and those which end 5' (B and D). One tenth of the amount of undenatured time course DNA was denatured and spotted onto the filter to control for the amount of DNA in each dot (C and D). Diagrams of probes complementary to the strand ending 3' on opposite sides of the cut (E and G); diagrams of probes complementary to the strand ending 5' (F and H). a

seen by the densitometric analysis of large populations of molecules (Fig. 4). The slowed kinetics of the SSA pathway may allow a larger proportion of a common intermediate to partition through the gene conversion pathway, leading to an increase in the proportion of gene conversion products. Another possibility to explain the change in proportion of gene conversion and deletion products could be that the slowed kinetics of deletion formation in the constructs with inserts could lead to greater plasmid loss. This could happen if exonuclease digestion were to render single stranded some essential region of the plasmid (e.g., the CEN or ARS sequences) or if the diverging exonucleases continued around the plasmid and met. However, there was no change in plasmid loss as a function of the size of the insert; in each case, about 20% of the plasmids were lost after HO induction (Table 1). 5'-to-3' degradation initiating at the DSB is bidirectional. The SSA mechanism requires the formation of extensive single-stranded DNA in both directions. Therefore, it was important to demonstrate bidirectional 5'-to-3' exonucleolytic digestion, in the HO endonuclease cut LacZ plasmid. To do so, we used a simple dot blot method for the detection of single-stranded DNA (Fig. 5). DNA samples in the absence of denaturation are spotted onto a nylon filter, and the DNA is then fixed to the filter by UV irradiation and hybridized with strand-specific RNA probes. Significant hybridization only occurs when a region is single stranded. The direction of exonuclease digestion can, therefore, be determined from the strand that hybridizes to undenatured DNA. When DNA from the plasmid construct with the 4.4-kb insert was probed with single-stranded probes homologous to this insert, only one strand hybridized. Probe complementary to the strand that ends 3' at the DSB

hybridized to samples taken 60 min after induction but failed to hybridize to samples taken at 6 h, when product formation was completed (Fig. 5A). The negligible hybridization that occurred with the opposite strand probe shows that only the 5'-ended strand was significantly degraded (Fig. 5B). Hybridization of DNA samples with single-stranded URA3 probes on the other side of the break (Fig. 6C and D) confirms that single-stranded degradation occurs bidirectionally; the time of appearance of single-stranded DNA is also 60 min after induction and 30 min after the creation of the DSB. Both the 4.4-kb insert and URA3 probes hybridize to sequences that are a similar distance (ca. 2.5 kb) from the break; therefore, the kinetics of single-stranded DNA formation in these regions are expected to be similar. These regions became single stranded 30 min after HO cleavage. This rate of formation of single-stranded DNA is consistent with degradation occurring at 1 to 2 nucleotides per s, as estimated by White and Haber (63) and by our analysis of the time required before deletion products are formed. Effects of substrate structure on the kinetics of product formation. In order to show that the absence of reciprocal product formation was not caused by a feature of the plasmid structure that constrained resolution of branched intermediates, a number of controls were carried out. We have previously shown that the above disparities were not due to the orientation of the HO endonuclease cut site itself (48). Additionally, similar kinetics of deletion formation were observed in LacZ plasmids with an I-SCE1 cut site (39); thus, the pattern of repair we observed was not a specific feature of HO-induced recombination. A number of other possible factors that might bias the outcome of the recombination events are presented below. High rates of transcription have been shown to stimulate

1298


H

iji

-

-

)i Tlt,.'

MOL. CELL. BIOL.

I )

?I

7'

FIG. 6. Lack of formation of the reciprocal crossover product even when both possible products contain ARS sequences. An autoradiogram of a time course experiment with a strain containing pJFL28. DNA was digested as described in the legend to Fig. 3 and analyzed with the same internal LacZ probe. Here, the linearized cut site gene is 4.0 kb and the linearized reciprocal circle would be 3.6 kb.

the formation of deletions in spontaneous direct repeat recombination events (58, 60). It was possible that transcription through the LacZ gene containing the cut site might enhance the reactivity of the cut end that could still be transcribed. Indeed, it is the end linked to the promoter that participates in a one-ended event to create the deletion without yielding the reciprocal circular product. To investigate this possibility, the LacZ plasmid was modified by deleting the CYC1::LEU2 composite promoter region. This plasmid yielded results identical to those shown in Fig. 3A (data not shown). As described above, we failed to observe the crossover product expected from a reciprocal exchange event. Since this product would contain only sequences from E. coli LacZ and phage X and lacks yeast chromosomal maintenance sequences, we considered the possibility that a DNA circle lacking an ARS might be lost during DNA extraction or rapidly eliminated after recombination. To address these issues, we inserted the ARSI sequence between the two LacZ repeats so that if the reciprocal circle was formed, it would be maintained. When this plasmid, pJFL28, was analyzed, there was again no change in the kinetics of product formation, nor was the LacZ-ARS1 circle (which, with this restriction digest, would be a 3.6-kb linear molecule) recovered (Fig. 6). These results strongly argue that most of the deletion product is not formed through the gap repair pathway and that gene conversions between direct repeats are rarely resolved to yield reciprocal products. Pattern of repair in cells arrested in the G1 phase of the cell cycle. In unsynchronized cells, the kinetics of product formation or resolution of intermediates might also depend on the stage of the cell in the cell cycle. Thus, we determined whether this pattern of repair was affected by synchronizing cell growth. If the pattern of repair is dependent on the stage of the cell within the cell cycle, this pattern should be different in synchronized cells. Cells arrested in G1 with a-factor were released and simultaneously induced, as described previously for studies of mating type switching (5, 63). No change in the kinetics or proportion of products was

FIG. 7. Analysis of the kinetics of product formation of synchronized cells. Cells were arrested in the G, phase of the cell cycle with a-factor until the proportion of unbudded cells was >95%. The cells were released from G1 and simultaneously induced for the HO endonuclease by being resuspended in galactose medium. DNA was extracted and analyzed as described in the legend to Fig. 3 and Materials and Methods. A Southern blot (A) and densitometric analysis (B) are shown.

observed (Fig. 7). Thus, the different kinetics and proportions with which the two products are formed are not due to populations of cells in different phases of the cell cycle or to any functional elements of the DNA. Additionally, the above experiment rules out the possibility of gene conversions occurring by a bimolecular mechanism in G2, after replication of the centromere-containing plasmid. Previously, we had shown that HO-induced mating type switching occurred with the same kinetics in cells passing through different phases of the yeast cell cycle (5). Galactose-induced mating type switching can also occur, although at apparently diminished efficiency, in Gl-arrested cells (40). Deletion formation and gene conversion occur with the same kinetics in logarithmically growing cells and in synchronized cells released from G1 arrest (Fig. 7). However, the pattern of repair in cells that are not allowed to traverse the cell cycle was found to be different. When the HO endonuclease was induced in cells arrested and maintained in the G1 phase of the cell cycle, there was a significant effect on the kinetics and proportions of products (Fig. 8). While the kinetics of the deletion formation was largely unaffected, the kinetics of the gene conversion formation was delayed by 30 min. In addition, the amount of gene conversion products decreased from 20 to 5%, while the proportion of deletions correspondingly increased to 95%. The impaired kinetics of gene

VOL. 12, 1992


A

0

5.6 kh

.1

5.8

Donor

4.4 kb Cut Site

Hours

A

5 2.5 2.

1

kb Deletion

1299

0 .5 I 1-4 2 2.5 4 6

.6 Lb Donor

_

_

-

b Deletion

4.4 kb Cut Site

4.3 kb Gene Conversion

--M

2.6 Lb cut I

2.6 kb cut 1

1.8 kb cut 2

1.8 kl) cut 2

a

B

B

6.0

lqime (hours) time hours

3.0

FIG. 8. Analysis of the kinetics of product formation during the G1 phase of the cell cycle. Cells were arrested in G1 with a-factor and held in G1 with additional pulses of a-factor during the induction and repair process. The proportion of unbudded cells throughout the experiment remained at >95%. DNA was extracted and analyzed as described in the legend to Fig. 3. A Southern blot (A) and densitometric analysis (B) are shown.

conversions may allow a common intermediate to partition more often through the SSA pathway, leading to an increase in deletions, as is the case when deletions are delayed and there is an increase in conversion products. DSB repair in the absence of RAD52. The RAD52 gene was originally identified by a mutation that confers sensitivity to X rays (9, 42). Evidence showing that a functional RAD52 gene product is required for almost all repair of DSBs, including mating type switching (25, 62), transformation with linear fragments (37), and the repair of HO-induced DSB (35, 46), has since accumulated. Recently, Ozenberger and Roeder (38) have shown that HO-induced deletion formation occurs in the absence of RAD52 in the repeated rDNA and CUPI arrays and the efficiency of deletion repair decreases with decreasing numbers of repeats. We examined both deletion formation and gap repair in the absence of the RAD52 gene product. A kinetic analysis shows that deletion formation occurs with the same kinetics as that of the wild type, although the amount of product produced was decreased approximately fivefold (Fig. 9). There was also a significant increase in HO-induced plasmid loss, from 21% in wild-type cells to 81% in radS2 cells. The plasmid is lost presumably because the cell is unable to repair the break. Moreover, gap repair products are not formed (Fig. 9). If gene conversions were also decreased fivefold, like dele-

FIG. 9. Pattern of repair in rad52 cells. (A) Autoradiogram of DSB repair in rad52 strain containing plasmid pJF6. DNA was extracted and digested as described in the legend to Fig. 3 and analyzed with the same internal LacZ probe. (B) Densitometric analysis of this autoradiogram.

tions, they should still represent 2% of the total DNA. We know we can detect this level of gene conversions, as this is the same level produced in Gl-arrested cells (Fig. 8). The absence of gene conversion products in radS2 cells is supported by previous studies showing that no mating type switching or PCR-amplified intermediates are produced in radS2 cells (63). These results show that deletion formation occurs, although at reduced efficiency, by a RAD52-independent pathway, while gene conversions occur via a RAD52dependent pathway. DISCUSSION

The results presented in this paper strongly argue that DSBs can be repaired by two competing and very different recombination pathways. DSBs within directly repeated sequences will frequently be repaired through a nonconservative SSA pathway, leading to a deletion between the direct repeats which removes the information between them. An alternative gap repair process yields gene conversions, nearly all of which are apparently resolved without crossing over (see below). Our results show that changes in substrate structure and the cellular growth conditions alter the kinetics and proportion of products and demonstrate that these two pathways are independent, although they compete for the same substrate. The formation of single-stranded DNA (which exposes complementary strands) is an essential step in the SSA

1300


mechanism of recombination. Hence, the addition of more DNA between the two repeats requires that the 5'-to-3' exonuclease travel through the insert before exposing a complementary single strand and should delay deletion formation. Based on the estimated rate of 1 to 2 nucleotides per s for this exonuclease, the additional time required for the exonuclease to traverse a 4.4-kb insert is estimated to be 35 to 75 min. This is entirely consistent with the observed delay in deletion product formation of 60 min (Fig. 3C and 4C). The increased time it takes to commit to the SSA pathway permits a larger proportion of the intermediates to enter the gap repair pathway, as the proportion of gene conversion events increased from 20 to more than 40%. It should be pointed out that the level of plasmid loss is the same with all the constructs; hence, the change in the proportion of products represents a shift in the partitioning of a common intermediate between these two pathways rather than preferential recovery of one of the products. Single-stranded DNA produced by bidirectional 5'-3' exonuclease degradation occurs after the DSB is produced and disappears as products are formed (Fig. 5). This pattern of single-stranded degradation also has been observed for HOinduced events in which the cut site is integrated in a chromosome (53) and in the processing of a DSB generated at the ARG4 locus during meiosis (55). Bidirectional 5'-to-3' degradation also occurs at the ends of linear DNA injected into Xenopus oocytes (29). The role of 3'-ended single strands has additionally been implicated to be on the pathway of recombination by injection of partially singlestranded molecules (31). The one exception to the bidirectional recision of DNA from a DSB is found in yeast mating type switching, in which 5'-to-3' single-stranded degradation occurs on only one side of the DSB (63). This restriction on degradation is a special feature of the interaction of M4T with its donor sequences, as deletion of the donors leads to bidirectional degradation (63). Theoretically, it should be possible to observe the time at which the donor LacZ gene, several kb from the HO endonuclease cut, becomes single stranded, prior to annealing. This LacZ region should become single stranded on the DNA strand complementary to that of the HO endonuclease cut LacZ gene. When we carried out dot blot hybridization of samples with probes complementary to each strand, we observed the expected formation of single strandedness in the HO endonuclease cut LacZ gene but little hybridization to the donor LacZ gene (data not shown). This result might suggest that reannealing occurs rapidly, once the complementary single strands have been exposed. However, it is equally likely that such complementary single-stranded regions will have reannealed in vitro, during DNA preparation. Nevertheless, on the basis of the kinetics of 5'-to-3' exonucleolytic digestion and the kinetics of deletion product formation, we conclude that reannealing is probably not a rate-limiting step in the process. We recognize that deletion products could also be formed by a one-ended invasion process which does not require both homologous regions to become single stranded. Such oneended events are the likely mechanism for the integration of a linearized fragment into DNA during gene replacement (45), especially when the integrating DNA contains ends complementary to two very distant sites (56). This process does not appear to be the mechanism by which intrachromosomal repeated sequences induced by a DSB form deletions. A one-ended recombination event that does not require both homologous regions to become single stranded should not be any more dependent on the distance separating

MOL. CELL. BIOL.

the two direct repeats than a two-ended gap repair mechanism. It is clear from our results that deletion formation (by SSA) is much more sensitive to the distance separating the complementary regions than is gap repair. Thus, our results are more consistent with the SSA pathway. We have yet to analyze structural intermediates for either pathway, although Rudin et al. (48) have described several possible structures. Physical evidence of an annealed intermediate containing long single-stranded tails has recently been provided by Maryon and Carroll (31) from their study of DNA molecules injected into Xenopus oocytes. In general, the yeast and vertebrate systems seem quite similar; there are, however, some interesting differences among these organisms. In Xenopus oocytes, SSA events are significantly impaired when nonhomologous sequences are at the very ends of the linearized molecules, so that reannealing of the ends themselves is not possible (30). This is clearly not the case in mammalian cells (21, 61) or yeast cells, since all of the reactions we have studied involve the removal of at least 50 bp of MA Ta DNA on either side of the HO cleavage site before LacZ homology is encountered. Moreover, in previous studies of deletion repair on the chromosome, HO endonuclease induced efficient deletion formation that required the removal of several thousand base pairs (47, 48). How important is SSA versus gap repair in the repair of DSBs? These data show that the SSA pathway is an effective recombination pathway that efficiently competes with gap repair pathways in yeast, both in plasmid substrates and in chromosomes (47, 48). Recently, Ozenberger and Roeder (38) have invoked SSA to account for the repair of HOinduced DSBs in the midst of repeated arrays (rDNA and CUPI), even in the absence of the important recombination gene RAD52. We have extended this analysis by showing that deletion formation but not gene conversion occurs in the LacZ plasmid in the absence of RAD52. Recently, we have further characterized our analysis of deletion formation by showing that as the amount of homology increases, the efficiency of deletion formation increases in a linear fashion from 0.09 to 1.2 kb (53). In mammalian cells, SSA represents a major pathway by which extrachromosomal elements recombine (3, 6, 21, 23, 61). Deletion products which could be formed via SSA appear to be less prevalent for spontaneous gene conversion events when direct repeats are integrated into a chromosome (20, 24, 52). We note, however, that the nature of the lesions that initiate recombination in these mammalian cells is not known. If DSBs occur less often in chromosomes than on extrachromosomal DNA, this fact would account for the difference in the ratios of gene conversions without exchange and deletion formation. No studies using defined DSBs in mammalian chromosomes have yet been reported. For a haploid organism such as S. cerevisiae and even in diploid mammalian cells, it might seem that deletion formation is an unattractive route for repairing DSBs, as the distribution of repeated sequences greater than ca. 100 bp is likely to cause the deletion of essential genes. This deletion would be lethal in a haploid cell and also in a diploid cell in which one functional copy of the gene is insufficient. Moreover, such deletions might reveal deleterious recessive mutations. We suggest that SSA does occur and that deletions in mammals associated with the loss of tumor suppressor genes (reviewed by Marshall [281) and other genes that lead to abnormalities, such as muscular dystrophy, and deficiencies in human growth hormone or steroid sulfatases (33, 59, 64) may very well arise by this pathway. In fact, SSA may be less likely than interchromosomal gap repair processes to

VOL. 12, 1992


reveal recessive mutations, because interchromosomal gap repair has a high (approximately 30 to 50%) probability of being accompanied by crossing over (16 and 18), thus leading to homozygosis of all the more distal markers. Additionally, deletions occurring within repeated arrays by SSA will lead to loss of a repeated unit and will not be lethal. A key question is whether a DSB is more likely to be repaired by intrachromosomal SSA between flanking homologous regions or by interchromosomal gap repair. In one study in which an HO endonuclease-induced break could be repaired by intrachromosomal deletion or interchromosomal gap repair, Rudin (46) showed that a MA4Ta gene integrated between two URA3 genes on chromosome V was more likely to experience deletion repair than to recombine with its normal donors on chromosome III. Additionally, when a MA4Ta gene was integrated between two LEU2 genes on chromosome III, approximately 20% of the HO endonuclease-induced events were deletions of A4T instead of a normal switching event using the HMLa donor on the same chromosome. Given that MAT appears to have a genetically determined preferential pairing system to facilitate a choice of donors between HML and HMR (17), these results argue that SSA is a major repair pathway that efficiently competes with other pathways. In many instances, SSA may be a costly and unavoidable consequence of the same machinery that generates the more intuitively attractive gap repair events.

Constraints on crossing over in DSB-induced gene conversions between direct repeats. We have found a striking difference between HO-induced gene conversions, depending on whether the genes are in a direct or inverted orientation. When the two homologous regions are in inverted orientation, gene conversions with and without crossing over are recovered in equal proportions (48). This result is consistent with targeted gap repair studies involving a plasmid and a chromosome (36). In contrast, gene conversions with direct repeats are apparently always resolved without

exchange. Even when the proportion of conversion products without a crossover increased to 42%, there was still no reciprocal circular product expected from a gap repair intermediate resolved with crossing over. We have ruled out the possibility that failure to recover the reciprocal crossover product could be attributed to the lack of an ARS sequence. Thus, this constraint on crossing over seems to be limited to intrachromosomal events involving direct repeats, as both intrachromosomal events between inverted repeats and interchromosomal repair of HO-induced DSB are frequently associated with crossing over (16 and 18). In general, studies of intrachromosomal HO-induced events could not distinguish between deletions arising by a conversion with crossing over from those arising by SSA (34, 37, 47). Consistent with our results, in one study that could distinguish between these two pathways (41), it was also found that most of the gene conversions (94%) were resolved without a reciprocal crossover. Inconsistent with our results, however, was the low proportion (12%) of deletion formation. However, the strain used by Ray et al. (41) exhibited a significant loss of viability upon induction, which could be linked to its poor ability to carry out deletion

repair. Independent modulation of deletion formation and gene conversions. One of the surprising aspects of HO-induced recombination is the long delay between production of HO endonuclease cut molecules and product formation. This delay may have several causes. Clearly, the rate of 5'-to-3' exonuclease can be rate limiting. In other circumstances,

1301

however, the limitation may reflect the low abundance of some essential recombination protein. Indeed, we have previously provided evidence that the completion of deletion formation requires new protein synthesis (47). Thus, holding cells at one point in the cell cycle may limit the abundance of a protein whose synthesis is cell cycle regulated. In this paper, we present physical evidence that DSB repair is affected when the cells are held in G1. Under these conditions, the kinetics of conversions were slowed by 30 min, while the kinetics of deletion formation were unaffected (Fig. 8). Additionally, the proportion of conversion products decreased by 75%. We do not believe that this decrease implies that the completion of gene conversion events requires the initiation of chromosomal DNA replication (as might be predicted from the model of Esposito [7]), because the kinetics and proportion of gene conversions were not affected when cells were G1 arrested and allowed to enter S phase as the HO endonuclease was induced. We have presented evidence that the kinetics of deletions and gene conversions can be independently modulated, with the consequence that the process that is retarded is also less frequently represented among the final products. This finding argues that the two pathways compete for the common DSB intermediate. It should be possible, by using this system, to assess the relative importance of many gene products to each of these two processes. For example, which DNA polymerase(s) is required for SSA and for gap filling? Are some radiation-sensitive genes needed for one but not the other? Experiments along these lines are currently under way. ACKNOWLEDGMENTS We thank M. Lichten, S. Lovett, C. Boles, B. Ray, N. Sugawara, and J. J. Kupiec for helpful comments on the manuscript. This work was supported by NIH grant GM20056.

REFERENCES 1. Alani, E., R. Padmore, and N. Kleckner. 1990. Analysis of wild-type and rad50 mutants of yeast suggests an intimate relationship between meiotic chromosome synapses and recombination. Cell 61:419-436. 2. Brenner, D., A. Smigocki, and D. Camerini-Otero. 1986. Doublestrand gap repair results in homologous recombination in mouse L cells. Proc. Natl. Acad. Sci. USA 83:1762-1766. 3. Chakrabarti, S., and M. Seidman. 1986. Intramolecular recombination between transfected repeated sequences in mammalian cells is nonconservative. Mol. Cell. Biol. 6:2520-2526. 4. Church, G. M., and W. Gilbert. 1984. Genomic sequencing. Proc. Natl. Acad. Sci. USA 81:1991-1995. 5. 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. 6. Desautels, L., S. Brouillette, J. Wallenburg, A. Belmaaza, N. Gusew, P. Trudel, and P. Chartrand. 1990. Characterization of nonconservative homologous junctions in mammalian cells. Mol. Cell. Biol. 10:6613-6618. 7. Esposito, M. 1978. Evidence that spontaneous mitotic recombination occurs at the two-strand stage. Proc. Natl. Acad. Sci. USA 75:4436-4440. 8. Feinberg, A. P., and B. Vogelstein. 1983. A technique for radiolabeling DNA restriction endonuclease fragments to high specific activity. Anal. Biochem. 132:6-13. 9. Game, J. C., and R. K. Mortimer. 1974. A genetic study of x-ray sensitive mutants in yeast. Mutat. Res. 24:218-292. 10. Guarente, L., R. R. Yocum, and P. Gifford. 1982. A GallO-CYCI hybrid yeast promoter identifies the Gal4 regulatory region as an upstream site. Proc. Natl. Acad. Sci. USA 79:7410-7414.

1302


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. Jacquier, A., and B. Dujon. 1985. An intron-encoded protein is active in a gene conversion process that spreads an intron into a mitochondrial gene. Cell 41:383-394. 13. Jasin, M., and P. Berg. 1988. Homologous integration in mammalian cells without target gene selection. Genes Dev. 2:13531363. 14. Jensen, R., and I. Herskowitz. 1984. Directionality and regulation of cassette substitution in yeast. Cold Spring Harbor Symp. Quant. Biol. 49:97-104. 15. Jessberger, R., and P. Berg. 1991. Repair of deletions and double-strand gaps by homologous recombination in a mammalian in vitro system. Mol. Cell. Biol. 11:445-457. 16. Klar, A., and J. N. Strathern. 1984. Resolution of recombination intermediates generated during yeast mating type switching. Nature (London) 310:744-748. 17. Klar, A. J., J. N. Strathern, and J. B. Hicks. 1981. A positioneffect control for gene transposition: state of expression of yeast mating-type genes affects their ability to switch. Cell 25:517524. 18. Kolodkin, A., A. J. Klar, and F. Stahl. 1986. Double-strand breaks can initiate meiotic recombination in Saccharomyces cerevisiae. Cell 46:733-740. 19. Kostriken, R., J. N. Strathern, A. J. Klar, J. B. Hicks, and F. Heffron. 1983. A site-specific endonuclease essential for matingtype switching in Saccharomyces cerevisiae. Cell 35:167-174. 20. Letsou, A., and M. R. Liskay. 1986. Intrachromosomal recombination in mammalian cells, p. 383-409. In R. Kucherlapati (ed.), Gene transfer. Plenum Press, New York. 21. Lin, F., K. Sperle, and N. Sternberg. 1990. Intermolecular recombination between DNAs introduced into mouse L cells is mediated by a nonconservative pathway that leads to crossover products. Mol. Cell. Biol. 10:103-112. 22. Lin, F., K. Sperle, and N. Steirnberg. 1990. Repair of doublestranded DNA breaks by homologous DNA fragments during transfer of DNA into mouse L cells. Mol. Cell. Biol. 10:113-119. 23. Lin, F. L., K. Sperle, and N. Sternberg. 1984. Model for homologous recombination during transfer of DNA into mouse L cells: role for DNA ends in the recombination process. Mol. Cell. Biol. 4:1020-1034. 24. Liskay, R. M., J. L. Stachelek, and A. Letsou. 1984. Homologous recombination between repeated chromosomal sequences in mouse cells. Cold Spring Harbor Symp. Quant. Biol. 49:183189. 25. Malone, R., and R. Esposito. 1980. The RAD52 gene is required for homothallic interconversion of mating types and spontaneous mitotic recombination in yeast. Proc. Natl. Acad. Sci. USA 77:503-507. 26. Maniatis, T., E. F. Fritsch, and J. SambrooL 1982. Molecular cloning: a laboratory manual. Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y. 27. Mansour, S. L., K. R. Thomas, and M. R. Capecchi. 1988. Disruption of the proto-oncogene INT-2 in mouse embryoderived stem cells: a general strategy for targeting mutations to non-selectable genes. Nature (London) 336:348-352. 28. Marshall, C. J. 1991. Tumor suppressor genes. Cell 64:313-326. 29. Maryon, E., and D. Carroll. 1989. Degradation of linear DNA by a strand-specific exonuclease activity in Xenopus laevis oocytes. Mol. Cell. Biol. 9:4862-4871. 30. Maryon, E., and D. Carroll. 1991. Involvement of singlestranded tails in homologous recombination of DNA injected into Xenopus laevis oocyte nuclei. Mol. Cell. Biol. 11:32683277. 31. Maryon, E., and D. Carroll. 1991. Characterization of recombination intermediates from DNA injected into Xenopus laevis oocytes: evidence for a nonconservative mechanism of homologous recombination. Mol. Cell. Biol. 11:3278-3287. 32. Miller, J. H. 1972. Experiments in molecular genetics, p. 23-56. Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y. 33. Nicholls, R. D., N. Fischel-Ghodsian, and D. R. Higgs. 1987. Recombination at the human ox-globin gene cluster: sequence

MOL. CELL. BIOL. features and topological constraints. Cell 49:369-378. 34. Nickoloff, J. A., E. Y. Chen, and F. Heffron. 1986. A 24-basepair DNA sequence from the MA4T locus stimulates intergenic recombination in yeast. Proc. Natl. Acad. Sci. USA 83:78317835. 35. Nickoloff, J. A., J. D. Singer, M. F. Hoekstra, and F. Heffron. 1989. Double-strand breaks stimulate alternative mechanisms of recombination repair. J. Mol. Biol. 207:527-541. 36. Orr-Weaver, T. L., and J. W. Szostak. 1983. Yeast recombination: the association between double-strand gap repair and crossing-over. Proc. Natl. Acad. Sci. USA 80:4417-4421. 37. 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. 38. Ozenberger, B., and G. S. Roeder. 1991. A unique pathway of double-strand break repair operates in tandemly repeated genes. Mol. Cell. Biol. 11:1222-1231. 39. Plessis, A., A. Perrin, J. E. Haber, and B. Dujon. Site-specific recombination determined by I-Scel, a mitochondrial group I intron-encoded endonuclease expressed in the yeast nucleus. Genetics, in press. 40. Raveh, D., S. Hughes, B. 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. 41. Ray, A., I. Siddiqi, A. L. Kolodkin, and F. W. Stahl. 1988. Intrachromosomal gene conversion induced by a DNA doublestrand break in Saccharomyces cerevisiae. J. Mol. Biol. 201: 247-260. 42. Resnick, M. A. 1969. Genetic control of radiation sensitivity in Saccharomyces cerevisiae. Genetics 62:519-531. 43. Resnick, M. A. 1976. The repair of double-strand breaks in DNA: a model involving recombination. J. Theor. Biol. 59:97106. 44. Resnick, M. A., and P. M. Martin. 1976. The repair of doublestrand breaks in the nuclear DNA of Saccharomyces cerevisiae and its genetic control. Mol. Gen. Genet. 143:119-129. 45. Rothstein, R. 1983. One-step gene disruption in yeast. Methods Enzymol. 101:202-211. 46. Rudin, N. 1988. Ph.D. thesis. Brandeis University, Waltham, Mass. 47. Rudin, N., and J. E. Haber. 1988. Efficient repair of HO-induced chromosomal breaks in Saccharomyces cerevisiae by recombination between flanking homologous sequences. Mol. Cell. Biol. 8:3918-3928. 48. Rudin, N., E. Sugarman, and J. E. Haber. 1989. Genetic and physical analysis of double-strand break repair and recombination in Saccharomyces cerevisiae. Genetics 122:519-534. 49. Sherman, F., G. R. Fink, and J. B. Hicks. 1983. Methods in yeast genetics. Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y. 50. Smithies, O., R. G. Gregg, S. S. Boggs, M. A. Koralewski, and R. S. Kucherlapati. 1985. Insertion of DNA sequences into the human chromosomal 1-globin locus by homologous recombination. Nature (London) 317:230-234. 51. 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 strand cut in the MAT locus. Cell 31:183-192. 52. Subramani, S., and J. Rubnitz. 1985. Recombination events after transient infection and stable integration of DNA in mouse cells. Mol. Cell. Biol. 5:659-666. 53. Sugawara, N., and J. Haber. 1992. Characterization of doublestrand break-induced recombination: homology requirements and single-stranded DNA formation. Mol. Cell. Biol. 12:563575. 54. Sun, H., D. Treco, N. P. Schultes, and J. W. Szostak. 1989. Double-strand breaks at an initiation site for meiotic gene conversion. Nature (London) 338:87-90. 55. Sun, H., D. Treco, and J. Szostak 1991. Extensive 3Y-overhanging, single-stranded DNA associated with the meiosis-specific double-strand breaks at the ARG4 recombination initiation site. Cell 64:1155-1161.

VOL. 12, 1992


56. Surosky, R. T., and B. K. Tye. 1985. Construction of telocentric chromosomes in Saccharomyces cerevisiae. Proc. Natl. Acad. Sci. USA 82:2106-2110. 57. 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. 58. Thomas, B. J., and R. Rothstein. 1989. Elevated recombination rates in transcriptionally active DNA. Cell 56:619-630. 59. Vnencak-Jones, C. L., and J. A. I. Phillips. 1990. Hot spots for growth hormone gene deletions in homologous regions outside of Alu repeats. Science 250:1745-1748. 60. Voelkel-Meiman, K., R. Keil, and G. S. Roeder. 1987. Recombination-stimulating sequences in yeast ribosomal DNA correspond to sequences regulating transcription by RNA polymerase I. Cell 48:1071-1079. 61. Wake, C., F. Vernaleone, and J. Wilson. 1985. Topological requirements for homologous recombination among DNA mol-

62.

63. 64.

65.

1303

ecules transfected into mammalian cells. Mol. Cell. Biol. 5:2080-2089. Weiffenbach, B., and J. E. Haber. 1981. Homothallic mating type switching generates lethal chromosome breaks in radS2 strains of Saccharomyces cerevisiae. Mol. Cell. Biol. 1:522534. White, C. I., and J. E. Haber. 1990. Intermediates of recombination during mating type switching in Saccharomyces cerevisiae. EMBO J. 9:663-673. Yen, P. H., M. X. Li, S. P. Tsaim, C. Johnson, T. Mohandas, and L. J. Shapiro. 1990. Frequent deletions of the human X chromosome distal short arm result from recombination between low copy repetitive elements. Cell 61:603-610. Zinn, A. R., and R. A. Butow. 1985. Nonreciprocal exchange between alleles of the yeast mitochondrial 21S rRNA gene: kinetics and involvement of a double-strand break. Cell 40:887895.