Characterization of Recombination Intermediates from DNA Injected ...

Vol. 11, No. 6

MOLECULAR AND CELLULAR BIOLOGY, June 1991, p. 3278-3287 0270-7306/91/063278-10$02.00/0 Copyright C) 1991, American Society for Microbiology

Characterization of Recombination Intermediates from DNA Injected into Xenopus laevis Oocytes: Evidence for a Nonconservative Mechanism of Homologous Recombination ED MARYONt AND DANA CARROLL* Departments of Biochemistry* and Cellular, Viral and Molecular Biology, University of Utah School of Medicine, Salt Lake City, Utah 84132 Received 26 December 1990/Accepted 20 March 1991

Homologous recombination between DNA molecules injected into Xenopus laevis oocyte nuclei is extremely efficient if injected molecules have overlapping homologous ends. Earlier work demonstrated that ends of linear molecules are degraded by a 5'->3' exonuclease activity, yielding 3' tails that participate in recombination. Here, we have characterized intermediates further advanced along the recombination pathway. The intermediates were identified by their unique electrophoretic and kinetic properties. Two-dimensional gel electrophoresis and hybridization with oligonucleotide probes showed that the intermediates had heteroduplex junctions within their homologous overlaps in which strands ending 3' were full length and those ending 5' were shortened. Additional characterization suggested that these intermediates had formed by the annealing of complementary 3' tails. Annealed junctions made in vitro were rapidly processed to products, indicating that they are on the normal recombination pathway. These results support a nonconservative, single-strand annealing mode of recombination. This recombination mechanism appears to be shared by many organisms, including bacteria, fungi, plants, and mammals.

efficiency. A single oocyte can convert 109 substrate molecules, or 10 ng of plasmid DNA, into covalently closed products in several hours of incubation (6, 35). Given this large capacity for recombination of a well-defined substrate and the steady accumulation of products over a convenient time course (35), we set out to identify and characterize the structure of recombination intermediates produced from injected DNA. Our earlier results had shown that both extensive sequence homology and molecular ends are required for efficient recombination in oocytes (6). Ends are acted on by a 5'-+3' exonuclease (35), and the single-stranded 3' tails it generates are on the pathway to recombination products (35a). Molecular models of recombination incorporating these features have been proposed on the basis of studies in other systems (28, 33, 49). In fact, single-stranded 3' tails commonly appear in recombination models (e.g., reference 39), and there is substantial evidence for their involvement in some recombination systems (12, 29, 38, 46, 50, 60). Examples of some recombination models utilizing 3' tails are shown in Fig. 1. They are illustrated for intramolecular recombination of a substrate with a terminal direct repeat, like that used in our experiments. The invasion model (Fig. 1A) depicts invasion of the single-stranded tail into a region of homologous duplex by strand displacement (37, 39). Branch migration can enlarge the displacement loop to produce regions of svmmetric heteroduplex and a Holliday junction. Resolution of this intermediate could proceed by cleavage of strands catalyzed by activities shown to act specifically at junctionlike structures (59) and subsequent ligation, or possibly by the action of topoisomerases (55). Like the invasion model, the priming model (Fig. 1B) begins with invasion by the single-stranded 3' tail. The 3' end is used as a primer for DNA synthesis, with continued strand displacement and perhaps the initiation of lagging-strand synthesis at the replication fork. This mode of recombination occurs during lytic infection of bacteriophage T4 (19, 38).

There are a number of motivations for the study of homologous genetic recombination: it is a ubiquitous natural process; it is required for proper chromosome segregation in meiosis; it is responsible for the relinking of alleles that is the basis of genetic mapping; and it plays a role in other processes of DNA metabolism such as repair and replication (28, 33, 49). In some organisms (and potentially in others), recombination pathways mediate the homologous integration of transfected DNA, a powerful tool for genetic manipulation (28). The term "homologous recombination" includes a variety of different mechanisms that operate in many different organisms (55). Advances in understanding recombination mechanisms have come from a number of experimental approaches. The structure of heteroduplex recombination intermediates has been inferred by genetic analyses of recombination products, particularly in microorganisms (33, 39, 61). Activities catalyzing recombination events have been identified by characterizing products of genes known from mutational analysis to be required for recombination or DNA repair (11, 13, 48). The efficiency with which cells recombine transfected DNA molecules of various structures has also provided information about recombination mechanisms (1, 18, 30, 31, 32, 47, 54). Physical isolation and characterization of recombination intermediates can also be informative (4, 29, 51, 60), but has not generally been feasible, since in most circumstances intermediates are transient, unstable, and heterogeneous and constitute a minute fraction of the total DNA. We have been studying the homologous recombination of DNA molecules injected into Xenopus oocyte nuclei (5, 6, 21, 35). Linear substrates with overlapping homologous ends undergo recombination within the overlaps with exceptional Corresponding author. t Present address: Department of Genetics, University of Wisconsin, Madison, WI 53706. *

3278

INTERMEDIATES OF NONCONSERVATIVE RECOMBINATION

VOL . 1 l, 1991

nisms by analyzing the structure of recombination intermediates recovered from oocytes and by injection of molecules constructed in vitro that have the structure attributed to the intermediates. The experiments provide strong support for the annealing model (Fig. 1C).

I..,

I

/A

B

\C ---

'.Ib

II

qqcrl-

3279

A

I=EII

\ ~~~~I/ FIG. 1. Models of homologous recombination utilizing 3' tails. Three pathways are illustrated for intramolecular recombination of a substrate with terminal direct repeats (thick lines). Plain ends are 5' ends; half-arrowheads denote 3' ends. The first structure shown already has 3' tails generated by the action of 5'-+3' exonuclease, as occurs in oocytes. In pathway A (invasion), a 3' tail invades homologous sequences in the other repeat; this is followed by branch migration to enlarge the region of strand exchange and then resolution. In pathway B (priming), invasion is followed by DNA synthesis (dashed line) primed by the invading 3' end; subsequent steps are not specified. In pathway C (annealing), exonucleolytic degradation continues until complementary sequences are exposed in single strands; these anneal, and the junction is further processed, presumably by nuclease(s) and ligase. See text for additional details.

Aspect of both models A and B are combined in the double-strand-break repair model of fungal recombination (54). In the annealing model (Fig. 1C), degradation of the 5' ends continues until complementary single strands are exposed. The strands anneal, and redundant sequences are removed by further 5' end degradation or by destruction of the protruding 3' ends. Ultimately, DNA ligase is required to seal the strands. This type of model has been proposed for recombination in general (56), for bacteriophage recombination (8), for a bacterial pathway (53), and for extrachromosomal recombination in mammalian cells (32). This model is nonconservative in that one copy of the original duplication is necessarily lost in the process, and it is supported by findings in a number of systems (1, 2, 18, 30, 31, 44, 45, 47, 57). This list of recombination models is far from exhaustive (23, 24, 42, 54, 55, 57), but the diagrams in Fig. 1 include distinguishing features of several different classes of mechanisms. We have attempted to differentiate among mecha-

MATERIALS AND METHODS Oocytes and injection procedures. Procedures used in the isolation, culturing, and injection of oocytes have been described (35). DNA preparation. pRW4 (35) is a derivative of pRDK41 (15); it consists of pBR322 sequences with a tandem duplication of the tetracycline resistance (Tet) gene. The two Tet genes are separated by a unique XhoI site, so digestion with this enzyme produces a linear molecule having direct terminal repeats of 1,246 bp. pHRS3 is identical to pRW4 except for replacement of two restriction sites that flank the Tet repeats. A KpnI linker has been inserted at the SspI site, and a XbaI linker has been inserted at the StyI site (see Fig. 7). Plasmid DNAs (35) and M13mpl9 replicative form I DNA (3) were prepared as described previously. Preparative and analytical digestions with restriction enzymes were performed essentially as suggested by the manufacturers. The making of 5' and 3' tails with exonuclease with III and T7 gene 6 exonuclease respectively, has been described (35). The lengths of the tails were measured by treating tailed samples with S1 nuclease and sizing the remaining duplexes by native gel electrophoresis. The lengths of individual strands were also measured by alkaline gel electrophoresis. Si nuclease digestions were performed in 20- to 30-pA1 reactions with 1 U of enzyme at room temperature (35). Gel electrophoresis. All gels were 1% agarose (Seakem ME; FMC BioProducts, Rockland, Maine), except the 1.5% low-melting-point agarose (Bethesda Research Laboratories, Gaithersburg, Md.) gel shown in Fig. 5. The lowmelting-point agarose gel was run at 4°C and 1.5 V/cm. The injected DNA sample was run in a lane beside size standards. The standards were cut off, stained, and realigned to the gel. A 1- by 1-cm block including the area of interest was excised from the gel and sectioned by using a stack of razor blades separated by 0.6-mm spacers. Each of the 15 slices was melted and divided, half was treated with S1 nuclease, and the DNAs were analyzed by normal 1% agarose gel electrophoresis. Procedures for performing alkaline and twodimensional (2-D) gel electrophoresis were as described

previously (35). Blot hybridizations. Most gels were transferred in 0.4 N NaOH to Zetaprobe nylon membranes (Bio-Rad Laboratories, Richmond, Calif.) (35), including the 2-D and alkaline gels. Transfer of DNA without prior denaturation (Fig. 6) was performed in 1Ox SSC (lx SSC is 0.15 M NaCl plus 0.015 M sodium citrate). Hybridizations were done at 65°C with random-primed probes and 42°C with oligonucleotide (oligo) probes under conditions described by Church and Gilbert (10). In the nondenatured hybridization experiment (Fig. 6), the first oligo probe was removed by heating to 65°C in 0.1 x SSC-0.1% sodium dodecyl sulfate (SDS), which was sufficient to dissociate the oligo but not to denature long duplex DNA. After the oligo hybridizations, the membrane was soaked in 0.4 N NaOH, neutralized, and rehybridized with random-primed pBR322 probe. Labeling of pBR322 by random priming and labeling of oligos with polynucleotide kinase have been described (35). The oligo probes were 16 or 19 nucleotides (nt) long and correspond to the following

3280

MOL. CELL. BIOL.

MARYON AND CARROLL

B 43o3

Dp

pBR322

PvulI

ill

D

C:EO_ .Eo

C

°i

E

oo

2

-

C

inc.time

0C NC4M I

:0EI.mBlE -.4

_

ow i}'~*

S10

* IL

-o

2mwl _

o

W§iljR

VW

iow-nR3 -KInt.

"'

is--

p

-c

R2

N

1

-PP4363bp e-..S3560

'i&& . _-

4m

Rl

undigested

-.WS2050

+Pvull

FIG. 2. Time course of recombination of linear pRW4. A 10-ng sample of substrate DNA was injected into each oocyte and incubated for the times indicated; M refers to uninjected DNA. Samples were recovered and analyzed by electrophoresis in 1% agarose gels before (C) or after (D) digestion with PvuII. (A and B) Substrate (XhoI-digested pRW4) and expected product. Heavy lines indicate the 1,246-bp terminal direct repeats. Also shown is the location of the PvuII site and the sizes of fragments generated by PvuII digestion. These are the same whether recombination is intramolecular, as suggested by the diagram, or intermolecular. In panel C, the locations of input linear substrate (IL), recombination products (Rl to R3), and some (but not all) kinetic intermediates (Int.) are indicated. By comparison with mobilities of known structures, the major recombinants correspond to supercoiled monomer circles (Ri), nicked monomer and supercoiled dimer circles (R2), and larger intermolecular products (R3). In panel D, the locations of substrate (S) and product (P) fragments are indicated. Positions of apparent recombination intermediates at the 60-min to 4-h time points are shown with arrowheads.

pBR322 sequences (numbered according to Sutcliffe [52]): oligo 1, 63 to 78; oligo 2, 78 to 63; oligo 3, 1259 to 1241; oligo 5, 1955 to 1937; oligo 6, 1937 to 1955; oligo 7, 2171 to 2189; oligo 8, 2189 to 2171. Construction of annealed junctions. Linear pHRS3, at a concentration of 100 ,ug/ml, was treated with T7 gene 6 exonuclease for 1 min at room temperature, which removed 850 to 950 nt from each 5' end. The reaction was diluted into 50 volumes of 1 x SSC-15 mM EDTA, incubated at 65°C for 1 h, and then cooled to room temperature. The DNA was ethanol precipitated and extracted sequentially with phenol, chloroform-isoamyl alcohol (24:1), and ether. After reprecipitation, the structure of the DNA was analyzed by neutral and alkaline gel electrophoresis, before and after Si nuclease treatment. We estimate that the sample consisted of greater than 95% annealed monomer circles, associated through 500to 600-bp duplex regions and having 300- to 400-nt singlestranded 3' tails. RESULTS Recombination kinetics. The structure of the recombination substrate used in these experiments is shown in Fig. 2A. Plasmid pRW4 (35) has the complete sequence of pBR322 and a direct duplication of 1,246 bp in which the copies are separated by a unique XhoI site. Digestion with XhoI produces a linear molecule with homologous overlapping ends. Recombination in oocytes takes place within two overlaps of

same or different molecules, producing monomer circles or various multimers of pBR322. An intramolecular recom-

the

bination event is illustrated in Fig. 2A and B. Also shown in Fig. 2 is a time course of recombination of this substrate. When subjected to electrophoresis without prior digestion (Fig. 2C), the uninjected sample consisted entirely of the input linear substrate (labeled IL). After injection and overnight incubation, essentially only recombination products (labeled R) were present. In samples recovered from oocytes after 10 min and up to several hours of incubation, a series of bands was seen at unique mobilities. These bands (one pair is designated Int. in Fig. 2C) had kinetic properties of reaction intermediates: they were always seen when recombinant products were accumulating, they reached steady-state levels which declined as the substrate was depleted, and they were not observed after long incubation when recombination was essentially complete. When the same samples were digested with PvuII prior to electrophoresis, substrate bands (3,560 and 2,050 bp; designated with an S in Fig. 2D) were seen to disappear with time, in favor of recombination products (4,363 bp; designated P). One 4,363-bp fragment is generated by each recombination event, whether it is intra- or intermolecular. At intermediate times, smears were visible trailing the product (P) bands that again showed the kinetic behavior expected of recombination intermediates. The presence of the smears in PvuII digests was correlated with the bands of unique mobilities seen in the undigested samples.

VOL . 1 l, 1991


Structural analysis of intermediates by 2-D gel electrophoresis. We were interested in determining the strand composition of the intermediates. After PvuII digestion, we expected them to be linear molecules with incomplete junctions, having bubbles, branches, or other interruptions, such as illustrated in Fig. 1. We recovered DNA believed to contain intermediates and subjected the sample to 2-D gel electrophoresis. The first dimension was run under normal, nondenaturing conditions, while the second dimension was run in denaturing conditions at pH 12 (14, 35). Hydrogenbonded junctions dissociated in the second dimension into constituent strands, the lengths of which were estimated by comparing them to product and substrate strands run in the same gel. DNA was recovered from oocytes after 90 min of incubation, digested with PvuII, and analyzed (Fig. 3). In the first dimension, this sample had the appearance of those shown previously. It contained intact and partially degraded fragments from the substrate, completed products, and a trailing smear of intermediates. In the second dimension, the completed products and undegraded substrate fragments migrated on a diagonal defined by duplex molecules that have strands of the same length (14, 35). Partially degraded substrate fragments gave rise to two smears that flared away from this diagonal; vertical smears represent full-length strands which had their 3' ends exposed in the oocytes, and smears that angle down to the right represent the shortening of exposed 5' termini. This was shown by the oligo hybridizations described below and was also observed during analysis of nonrecombining linear DNA recovered from oocytes (35). In the second dimension, the trailing smear of recombination intermediates separated into four distinguishable signals (Fig. 3). Each of these corresponds to one strand of the original substrate, as shown by hybridization with specific oligonucleotides. For example, oligo 6 is complementary to strand III, the 3'-ending strand of the 2,050-bp substrate fragment (see diagram in Fig. 3). It hybridized to the 4,363-nt product strands, to the 2,050-nt strands from substrate fragments, and to the vertical smear labelled III derived from the intermediates (Fig. 3). In the second dimension, the mobility of strand III from intermediates was identical to that of (undegraded) strand III from the substrate. Hybridization with oligo 8 showed that the same was true for the other 3'-ending strand (strand I). Thus, the 3'-ending strands in the intermediates were neither shorter nor longer than the input substrate strands. The 5'-ending strands (II and IV) from the intermediates migrated faster in the second dimension than the same strands in unrecombined substrate. Hybridization with oligo 7, for example, showed that the average size of strand II in the intermediates was substantially shorter than the size of the same strand associated with partially degraded substrate molecules (compare the relative mobilities in the second dimension). We estimate that the 5'-ending strands in the intermediates were between 400 and 1,000 bases shorter than the original intact strands. The 2-D gel analysis thus showed that intermediates had full-length 3'-ending strands and shortened 5'-ending strands. There was no evidence that 3' ends were ever extended beyond their original length. This is consistent with the idea that single-stranded tails made by the oocyte 5'--3' exonuclease activity are on the pathway to recombination products, but not with the priming model (Fig. 1B). It was also clear that full-length 5'-ending strands were not present in the intermediates. This is necessarily true in the annealing

3281

alk.

neut.4 aa

4363b55

5

3560

2050

-

P

PROBE

pBR322 r

OLIGO PROBE

S 6.

8B

I

I

S

*1

0

0

-be

FIG. 3. Two-dimensional gel analysis of recombination intermediates. A 90-min sample from a recombination time course like that in Fig. 2 was digested with PvuII. In the normal 1-D analysis, shown at the left in each panel next to size markers, this sample is seen to consist of the usual mixture of partially degraded substrate fragments (3,560 and 2,050 bp), products (4,363 bp), and intermediates (trailing behind the products). In the 2-D portion of the gel, fractionation was at pH 7.8 in the first dimension and then at pH 12 in the second dimension. The gel was transferred to a nylon filter and hybridized sequentially with pBR322 and with specific oligo probes. The diagram shows the locations of the oligo sequences and designates the individual strands of the substrate with Roman numerals. As in Fig. 2, heavy lines indicate the terminal direct repeats that support recombination; 3' ends are designated with half-arrowheads; 5' ends are plain. Each oligo is shown beside the strand to which it is identical; thus, it hybridizes to the opposite strand. For example, oligo 6 hybridizes to strand III, which is the 3'-ending strand of the 2,050-bp substrate fragment. Arrowheads of a given style denote the same signal in the different panels. The conclusions from the oligo hybridizations are included in the top left panel, where each of the strands in the intermediates is identified.

model, but intact 5'-ending strands might have been expected among intermediates from one-ended invasions (Fig. IA) if the invaded end had not suffered degradation. Since many substrate molecules had intact 5'-ending strands (Fig. 3 and 4), the absence of undegraded 5'-ending strands in the intermediates makes simple invasion structures less likely. S1 nuclease sensitivity of intermediates. To extend the structural analysis, a sample like that used in the previous experiment was divided, and half was treated with Si nuclease. After PvuII digestion, the two samples were run in equivalent 2-D gels and hybridized sequentially with whole pBR322 and oligo probes as shown in Fig. 4. In this analysis, the oligo probes hybridized to sequences within the duplication, very near one of the substrate ends (see diagram in Fig. 4A). Results with the sample not treated with Si (Fig. 4B)

MARYON AND CARROLL

3282

MOL. CELL. BIOL.

A

9LMT III

no Si treatment II VI

4

2050

&

I I1 III

*t *

9 Os

It:,

*

a

+

13 -

t

14 slice no. + Si nuclease

-

-

"

Om

40

.9

2050 @

4-2050

a

|

t

t*

0

oligo2

-

IV

4 .

+

ob

£|

ol go 1

12_

11 -

+ S1 nuclease

0

pBR322

+

Iwkmd* -.4363bp ;Q-3560

350> 3560-*

IV

..4.

4363bp 3560

10 -

4363.4J sectioned

C PROBE

+

2050 bp IV

_-Pvull

alk.

-

_4i

q

FIG. 4. Two-dimensional gel analysis of Si nuclease-treated intermediates. (B) Sample treated just like that in Fig. 3 but hybridized with oligo probes located within the terminal duplication, as illustrated in panel A. (C) The same sample treated with Si nuclease prior to electrophoresis. In panel A, the oligos are shown beside the strands whose sequence they share; oligo 1, for example, hybridizes to strands II and III.

confirmed the previous conclusions. Oligo 1 hybridized to both strand II (partially degraded, 5' ending, from the 3,560-bp fragment) and strand III (full length, 3' ending, from the 2,050-bp fragment) in the intermediates and in the substrate fragments. Oligo 2 hybridized only to strand I (full-length, 3' ending, from the 3,560-bp fragment) in the intermediates, since its complement in strand IV (5' ending, from the 2,050-bp fragment) was removed by the oocyte exonuclease. Si nuclease treatment affected several features of the 2-D pattern (Fig. 4C). First, the smears of degraded substrate fragments spread to greater mobilities, which is expected upon removal of their single-stranded 3' tails. Second, those smears were located on the diagonal, indicating strands of equal length, also as expected. Third, the intermediates were shifted so they migrated with 4,363-bp product in the first dimension, while they still separated into four smears in the second. The identities of the strands in these smears were determined with oligo probes. The smear corresponding to strand II had about equal hybridization intensity in the second dimension before and after Si treatment (oligo 1), while hybridization of oligo 1 to strand III was nearly completely abolished. This indicated that the 3' tip of strand III was largely susceptible to Si. Strand I smeared to heterogeneous smaller sizes after Si treatment (oligo 2), again showing the

FIG. 5. Si nuclease sensitivity of isolated intermediates. A recombining sample was recovered from oocytes after 90 min of incubation as for Fig. 3 and 4. It was digested with PvuII and subjected to electrophoresis in a 1.5% low-melting-point (LMT) agarose gel. The region containing the 4,363-bp product and the intermediates was sectioned; slices with higher numbers contain species of slower mobility. Half of each slice was treated with Si nuclease, and the treated and untreated aliquots were run in a normal 1% agarose gel. Hybridization was with a pBR322 probe.

susceptibility of the 3'-ending strands. The average mobility of the 5'-ending strands also appeared to shift to somewhat smaller size after Si treatment (see Discussion). We were concerned that Si treatment of the intermediates might have cleaved through the junction in some molecules and caused them to run with the substrate fragments in the first dimension, thus obscuring a potentially significant class of molecules. To investigate this possibility, we sectioned the region containing intermediates from a 1-D gel into a series of thin slices. The slices were divided, half of each slice was treated with Si nuclease, and the pairs of samples from each slice were run in a second neutral agarose gel (Fig. 5). Without Si treatment, each sample ran true in the second gel. The molecules in slices 12 to 14 had discrete, retarded mobilities relative to completed products. After Si treatment, these molecules all showed sharp migration at exactly the size of pBR322. Few if any intermediates were cleaved in the junction region by Si nuclease, since little of the material ran with substrate fragments. This experiment confirmed that the mobility of Si-treated intermediates was the same as that of the (pBR322) product, regardless of the mobility prior to Si treatment. We emphasize that although their 5'- and 3'-ending strands were heterogeneous after Si treatment (Fig. 4), the intermediates had the precise mobility of intact pBR322 product molecules. These Si sensitivity patterns are easily explained by the annealing model: slower-migrating intermediates with long tails and faster-migrating ones with shorter tails and longer annealed regions would both be reduced to pBR322 size by removal of the single-stranded tails with Si (see Fig. 1C). In contrast, Si treatment of invasion intermediates (Fig. 1A) would yield branched structures if the Holliday junction were resistant to the nuclease, or substrate-sized fragments if it were susceptible. Single-stranded regions of intermediates. The Si nuclease analysis indicated that 3' tails were exposed as single strands in the recombination intermediates. To test this further, recovered DNA was transferred to nylon membranes and hybridized without denaturation (Fig. 6). In our hands only molecules with exposed single strands bound to the membrane, and these single strands could be identified with oligo

probes.


VOL. 11, 1991 A

1 2 3 4

B1

C1

2 3 4

3 4

2

U 20m 0m 3hr O/N Incl time' U 20m 60m 3hr O/N digsts K S K S K S K S K SK S K S K S K S K S

_1 4363bp

-

3585

U-

-

-

3283 0/N K S -p

'am6 _

2025

-

probe:

D

B

A.

2

pBR322

D

C

-# S.pi

-

#2

._

-stly

2025 bp

Pstl

3585

op

FIG. 6. (A to C) Detection of single-stranded regions in intermediates. A sample containing recombination intermediates was recovered from oocytes as for Fig. 3 to 5. It was run on a gel with various control DNAs, transferred to nylon without denaturation, and hybridized to oligo probes, also without denaturation. The DNA from oocytes is in lanes 4. Control DNAs are as follows: lanes 1, linear pRW4 DNA treated with exonuclease III to leave 350-nt 5' tails; lanes 2, linear pRW4 with 300-nt 3' tails made by treatment with T7 gene 6 exonuclease; lanes 3, pRW4 treated with T7 gene 6 exonuclease to leave 3' tails long enough (750 nt) to anneal within the terminal duplications. All were digested with PstI before electrophoresis. (D) Locations of the oligo probes and the sizes of the strands expected from substrate DNA. After hybridization to the oligo probes without denaturation, the filter was treated with alkali and rehybridized with whole pBR322 probe (A). The pointers in each lane indicate the following species: lanes 1, 3,585-bp fragments with 5' tails; lanes 2, 3,585-bp fragments with 3' tails and some annealed molecules (upper band); lanes 3, annealed molecules with 3' tails; lanes 4 (from bottom to top), 2,025-bp substrate fragments, 3,585-bp substrate fragments, and intermediates (*) from oocytes.

The specificity of the hybridizations was shown with various control DNAs (Fig. 6A to C). Lanes 1 contain DNA resected in vitro with exonuclease III, which degrades 3'->5'. Its PstI fragments bound to the membrane, as shown by subsequent hybridization with whole pBR322 probe, but only sequences at the 5' end of the 3,585-bp fragment were available for hybridization (with oligo 2) without denaturation. The signal in this hybridization was weak, but clearly specific. Neither oligo 1 nor oligo 2 hybridized with duplex sequences in the 2,025-bp fragment. Complementary results were seen with DNA resected in vitro with T7 gene 6 exonuclease, which degrades 5'->3' (lanes 2 and 3). Sequences near the 3' end of the large PstI fragment hybridized with oligo 1, whether they were on linear fragments (lanes 2) or in annealed junctions made in vitro (lanes 3), but no hybridization to the small fragment was seen. The results with injected DNA (Fig. 6A to C, lanes 4) were essentially the same as those with control DNA treated with T7 gene 6 exonuclease (lanes 2 and 3). Oligo 1 hybridized to intermediates and to partially degraded 3,585-bp substrate fragments, but no hybridization was seen with oligo 2 or to the small fragment with either oligo. This showed that there were exposed single-stranded regions of 3'-ending strands in the intermediates, confirming the S1 results. Recombination of intermediates made in vitro. All of the results presented above pointed toward intermediates with

Kpn / --t--

Kpnll

4,

Xbal

PsSI

-+

..

FIG. 7. Processing of annealed junctions in oocytes. Pairs of substrates (C and D) were injected into oocytes and incubated for the times indicated. In each case, one DNA (lower structures in panels C or D) was annealed monomer circles of pHRS3, prepared by removing 850 to 950 nt from each 5' end with T7 gene 6 exonuclease and annealing at a low DNA concentration. The other was pRW4, either with flush ends (C, top structure) or 400-nt 3' tails made by T7 gene 6 exonuclease digestion (D, top structure). pRW4 was in fourfold excess over the pHRS3 circles in both samples. DNAs were recovered and digested with PstI-KpnI-XbaI (K) or with PstI-SspI-StyI (S). (C and D) Locations of these sites on the substrate DNAs. The digested DNAs were run in an alkaline agarose gel, transferred to a nylon filter, and hybridized with pBR322 probe. P, Locations of 4,363-bp completed intrasubstrate recombination products, which come from pRW4 in the K digests and from pHRS3 in the S digests. The other letters at the left indicate the positions of terminal fragments from the substrates (T) and internal fragments that are common to multiply digested substrates and products (C). The two samples at the far right are from injection of a mixture of linear pHRS3 and linear pRW4; they show the position of intersubstrate recombinants (IP). The dots in the 20and 60-min samples in panel B are beside bands that resulted from partial digestion by PstI; they do not run at the location expected for intersubstrate products.

annealed junctions as predicted by the annealing model (Fig. 1C). It was our intention to isolate such intermediates from injected oocytes and to reinject them to see if they would be rapidly processed to recombination products. We made some progress with this purification (34) but could not efficiently separate molecules with incomplete junctions from those that had been sealed. Therefore, we constructed molecules in vitro that had the structures deduced for the intermediates. We reasoned that if they were completed rapidly after injection, this would provide evidence that they are on the normal pathway to recombination products. Furthermore, the junctions could be coinjected with distinguishable linear substrate molecules in order to compare their rates of product formation. Circular molecules with annealed junctions were prepared from linear pHRS3, a marked derivative of pRW4 (Fig. 7); 850 to 950 nt were removed from each 5' end with T7 gene 6 exonuclease, and the DNA was annealed at low concentration. The resulting molecules had annealed regions of about 500 to 600 bp and protruding single-stranded 3' tails of about 300 to 400 bases. Their structure was confirmed by neutral

3284

MARYON AND CARROLL

and alkaline gel electrophoresis and by Si nuclease treatment. The annealed circles were mixed with a 4:1 excess of linear pRW4 or of pRW4 with 400-nt single-stranded 3' tails. In addition to providing a direct comparison of rates of product formation in the same oocytes, the use of marked substrates allowed monitoring of intersubstrate recombination, which could occur if the annealed junctions were dissociated in the oocytes. The mixtures were injected; recovered after 20-min, 1-h, 3-h, and overnight incubations; subjected to diagnostic digests; and analyzed by electrophoresis in alkaline agarose gels. The results (Fig. 7) showed that the annealed junctions were converted to covalently closed products considerably faster than flush-ended or tailed linear substrates. In Fig. 7A and B, 4,363-nt product strands in the PstI-SspI-StyI digest (lanes S) arose from annealed pHRS3, while those in the PstI-KpnI-XbaI digest (lanes K) came from pRW4. At 20 min in both mixtures, 4,363-nt products were seen primarily in the S lanes; at 1 h these still predominated, even though pHRS3 constituted a minority of the DNA injected. By 3 h, pRW4 products were equivalent to those of pHRS3; in the overnight samples, pRW4 products predominated. There was no evidence of joint products between pHRS3 and pRW4, showing that the annealed junctions were not opened prior to product formation. In a control experiment in which both substrates were linear molecules with 3' tails, the completed products were mostly intermolecular (Fig. 7B, two rightmost lanes; band IP in the S lane is diagnostic for intermolecular products). These experiments were performed with monomer circle junctions, as would be found in intramolecular recombination events. We have constructed a similar annealed junction in a linear intermediate that models intermolecular events. After injection, this was also sealed very rapidly (34). Mechanism of junction closure. We showed that annealed junctions do not come apart en route to completion, but there remained a number of mechanisms by which they could be matured. For example, the protruding 3' tails could be removed, or degradation at the resected 5' ends could continue until the undegraded 3' tails were fully paired. The difference between these models is whether 3'- or 5'-terminated strands are degraded in the junction. We distinguished these possibilities by examining the fate of each strand after injection of annealed circular molecules. The substrate was the same as in the experiment shown in Fig. 7, but no competing pRW4 was included. Following incubation, recovery, and digestion with PstI, the samples were analyzed by alkaline gel electrophoresis (Fig. 8). Hybridization with strand-specific oligo probes showed that the 5'-ending strands (II and IV) continued to be shortened in oocytes, while the 3'-ending strands (I and III) were not altered. If there had been any degradation of 3' ends, the oligo hybridization signals would have been reduced in comparison with those from the pBR322 probe, since the oligos were complementary to sequences very near the original ends (oligo 1 is complementary to the 3' end of strand I, and oligo 3 is complementary to the 3' end of strand III; Fig. 8). The extent of degradation of 5' ends prior to joining with 3' ends can be estimated from the data in Fig. 8 and from the distribution of strand lengths in intermediates observed in 2-D gels (Fig. 3 and 4). In neither case do strands II and IV appear to be degraded significantly beyond the region of homologous overlap; that is, they do not become more than 1,250 nt shorter than their original lengths. Thus, ligation

MOL. CELL. BIOL. U 20603hO/N

4363b 3585 -2600 2025

20'60 3h

U

U 20 60 3h ON

O/N

m

it N

I

_

III-oi

* III-

-1125

I)

IV-

probe

pBR322

*

vAD oligo 1

oligo 3

3-

pi 2025

1

bp

IV

III

PstI

/

_

II

No

11

II

Pstl

3585 bp

ivIV

4362 D)p

_-

FIG. 8. Strand processing in annealed junctions. Annealed pHRS3 circles, as in Fig. 7, were mixed with carrier linear M13 replicative form I DNA and injected into oocytes. At the indicated times, DNA was isolated, digested with PstI, fractionated by alkaline agarose gel electrophoresis, and transferred to a nylon filter. Hybridization was done sequentially with whole pBR322 and the oligo probes indicated. The locations of these probes on substrate and product sequences are shown in the diagram. Each oligo is drawn beside the strand to which it is identical; oligo 1 hybridizes to strands I and IV, while oligo 3 hybridizes to strands II and II1. Strand IV was only faintly visible with the pBR322 probe at this exposure but was detected easily with oligo 1.

must proceed rapidly after the 3' ends are assimilated, and there are no long-lived single-strand gaps in the intermediates.

DISCUSSION

The results presented above are all consistent with an annealing model of homologous recombination in oocytes. A more detailed version of that mechanism is shown in Fig. 9. We should emphasize that although we continue to diagram intramolecular events for simplicity, both intra- and intermolecular events actually occur in oocytes (6, 35). The intermediates and products from intra- and intermolecular recombination events are indistinguishable after digestion with enzymes having sites outside the terminal repeats. A number of experiments indicate that intra- and intermolecular recombination events occur via the same pathway (34). We identified apparent intermediates in the recombination 1-

I OW.

-%.

2> 13

Il1

5

FIG. 9. Steps in the annealing pathway of recombination. As in all previous figures, thick lines represent terminal direct repeats that support homologous recombination. 3' ends have half-arrowheads; 5' ends are plain. See text for details.

VOL.

11, 1991


process by their unique mobility in agarose gels. Following digestion with a restriction enzyme that cuts once outside the terminal duplication, we expected that the intermediates would be molecules with heteroduplex junctions. The electrophoretic mobility of these molecules was slower than that of completed products, consistent with their having branches and/or bubbles. We showed by 2-D gel analysis that the intermediates have the strand composition predicted by their derivation from 3'-tailed substrates. That is, they have full-length 3'-terminated strands and shortened 5'-terminated strands. There was no evidence for extension of the 3' ends as predicted by the priming model (Fig. 1B). The extent of degradation of 5'-ending strands in the intermediates (Fig. 3) suggests that a minimum length of degradation must be achieved before an end becomes associated with intermediates. That minimum is about 400 to 500 nt out of a total homologous overlap of 1,246 bp. This is as predicted by the annealing model, allowing for the fact that the two interacting ends need not be degraded to exactly the same extent. Two types of experiment showed that the 3'-terminated strands in the intermediates were single stranded. 3' ends were digested by Si nuclease, as shown in Fig. 4, and these 3'-ended strands were accessible to hybridization probes when the intermediates were bound to a filter without denaturation (Fig. 6). The modest level of Si sensitivity of 5' ends is very likely due to some isomerization in vitro by strand displacement of junctions like that shown in Fig. 9 (between steps 3 and 4). These data are all consistent with the existence of annealed junctions in the intermediates. To test whether such junctions were readily processed in oocytes, we constructed annealed circular molecules in vitro and injected them. The junctions were sealed more rapidly than recombination products appeared from coinjected substrates not containing annealed junctions, even when those substrates carried single-stranded 3' tails, which are known to accelerate recombination relative to flush ends (35a). This supports the idea that annealed junctions are further advanced on the recombination pathway (Fig. 9). The absence of intermolecular recombination products in the experiment showed that the junctions were not opened in the oocytes and processed to recombinants via a different pathway. This experiment is important for another reason. A priori it is possible that step 3 (Fig. 9) is not accomplished in oocytes, but occurs during the recovery of DNA, which involves phenol extraction and ethanol precipitation, both of which might encourage annealing of complementary single strands (26). Showing that annealed junctions are completed without dissociation makes this an unlikely hypothesis. We have also shown that DNA analyzed without extraction or precipitation contains identical intermediates, as judged by electrophoretic analysis (34). Furthermore, since the ends of molecules must interact in some way to form recombinant products, there is no reason to think that complementary tails would fail to anneal in vivo. The evidence against alternative mechanisms of recombination in the oocytes can be summarized as follows. As mentioned above, we have never seen any indication that 3' ends are extended in recombination intermediates, as predicted by the priming model (Fig. 1B). The data suggest that 3' ends are stable until they are joined into full-length strands in the final step of product formation. Si treatment of intermediates reduced them to a discrete size exactly that of the final product, regardless of their mobility prior to Si treatment. This is a clear prediction of annealed intermedi-

3285

ates, since all redundant sequences are by necessity single stranded (Fig. 9). Invasion intermediates (Fig. 1A), in contrast, are not expected to be Si sensitive at crossed-strand junctions (40), nor are they necessarily expected to resolve to a discrete size after Si treatment, since redundant sequences are likely to be at least partly in duplex form (Fig. 1A). Finally, there were no intact 5'-ending strands in intermediates, an observation that is unexpected for one-end invasion events but predicted by the annealing model. The intermediates predicted by the invasion model may be unstable during isolation, and branch migration can give rise to structures with some similarities to annealing model intermediates. To generate structures via the invasion pathway that reflect our experimental observations, however, requires multiple additional ad hoc assumptions that are ultimately much less compelling than the straightforward predictions of the annealing model. We have also found that a single terminal nonhomology interferes severely with recombination in oocytes (25a), which is contrary to the simple predictions of the invasion model. Likely steps in the annealing pathway are diagrammed in Fig. 9. In oocytes, the process begins with molecular ends; circular molecules do not participate in recombination (6). Steps 1 and 2 are performed by a 5'--3' exonuclease activity, which has been demonstrated in oocytes (35). It has also been shown that the 3' tails thus generated are on the recombination pathway (35a). Step 3 is the pairing of complementary single strands; it could proceed passively, or it may be catalyzed by oocyte activities. These could be of the types represented by the bacterial RecA protein (41), eukaryotic strand-exchange activities (17, 20, 25, 27), prokaryotic single-strand-binding proteins (9)-all of which have been shown to facilitate strand pairing-and/or topoisomerases (36, 58) which are capable of resolving interwound structures. Settling this issue and identifying catalysts of step 3, if they exist, is a focus of our continuing research. Step 4 is very likely due to continuing 5'--3' exonuclease digestion (Fig. 8). When X exonuclease removes a redundant strand in just this fashion, it stops at the point where a ligatable nick remains (7). We do not know whether the oocyte exonuclease shares this property. We have some evidence that when resolution is prevented by a terminal nonhomology, digestion can continue beyond the point where strand displacement is possible; however, the resulting structures are not destroyed, but remain for a very long time in oocytes (25a). Thus, step 5 may be accomplished by DNA ligase alone, or DNA polymerase or some other activity may be required to mature the joints for ligation. The annealing pathway is a nonconservative mechanism of recombination, since one of the participating homologous duplexes is lost in the process. Meiotic recombination is distinctly different in that both crossover and conversion events are conservative. The ability to perform nonconservative homologous recombination is, nonetheless, exhibited by a wide variety of organisms and cell types. It has been demonstrated in cultured mammalian cells (1, 18, 32, 47, 57), plant cells (2), bacteria (53), and yeast cells (44, 45). A unique feature of the oocyte system is the extreme sensitivity of recombination to terminal nonhomologies (6, 25a). Oocytes may lack exonuclease or debranching endonuclease activities that, in other organisms, allow elimination of substantial nonhomologies (e.g., references 44 and 45). What might be the normal function of the nonconservative annealing pathway? In oocytes, it might be responsible for the resolution of amplified ribosomal DNA, generated as linear molecules by a rolling circle mechanism (43), into the

3286

MARYON AND CARROLL

circular multimers found in oocyte nucleoli, although that amplification occurs at much earlier oocyte stages than we use for injection. More generally, it may represent an available means to repair double-strand breaks. It would do this very efficiently if the breaks occurred in tandemly repeated sequences but less well in randomly placed breaks. We have found that short sequence matches near molecular ends can also support recombination in oocytes, albeit at reduced efficiency (21), so it is conceivable that adventitious short repeats in chromosomal DNA might support repair by this means. Although large oocytes are arrested in prophase I of meiosis, they are many weeks or months beyond the pachytene stage when normal meiotic recombination occurs (16). During this period, the oocytes undergo tremendous growth, stockpiling materials that will be used in the early stages of embryogenesis. Because very rapid DNA synthesis is a feature of the first hours after fertilization, it is not surprising that enzymes of DNA metabolism are stored in oocytes in large quantities (22, 62). Perhaps this is also true of the activities that catalyze the recombination events we observe with injected DNAs. Whatever the function of the activities that catalyze the annealing pathway in frogs and other organisms, elucidating the details of this mechanism promises to facilitate the genetic manipulation of higher organisms. The annealing pathway appears to be operative in mammalian somatic cells. A thorough understanding of its machinery and re-

quirements could allow improvement of the efficiency of targeted integration events in the chromosomes of such cells. Based on what we now know, the ability to make targeted breaks in the chromosomal site(s) of interest may be critical in this process, but there may be other ways to increase the efficiency once the mechanism is better understood. Because of their manipulability and large capacity for recombination, Xenopus oocytes offer a particularly favorable system in which to pursue mechanistic and biochemical studies of a widespread mode of homologous recombination. ACKNOWLEDGMENTS We thank Jeff Strathern, Steve West, Jim Haber, Miro Radman, and members of this laboratory for helpful ideas and comments. The manuscript was improved by the suggestions of Phil Anderson, Rende Dawson, Richard Kolodner, Steve Kowalczykowski, Chris Lehman, and Genevieve Pont-Kingdon. This work was supported by grant DCB-8718227 from the National Science Foundation. E.M. was supported in part by a training grant in genetics (T32-GM07464) from the National Institutes of Health. REFERENCES 1. Anderson, R. A., and S. L. Eliason. 1986. Recombination of homologous DNA fragments transfected into mammalian cells occurs predominantly by terminal pairing. Mol. Cell. Biol. 6:3246-3252. 2. Baur, M., I. Potrykus, and J. Paszkowski. 1990. Intermolecular homologous recombination in plants. Mol. Cell. Biol. 10:492500. 3. Berger, S. L., and A. R. Kimmel. 1987. Guide to molecular cloning techniques. Methods Enzymol. 152:105-126. 4. Cao, L., and N. Kleckner. 1990. A pathway for generation and processing of double-strand breaks during meiotic recombination in S. cerevisiae. Cell 61:1089-1101. 5. Carroll, D. 1983. Genetic recombination of bacteriophage X DNAs in Xenopus oocytes. Proc. Natl. Acad. Sci. USA 80: 6902-6906. 6. Carroll, D., S. H. Wright, R. K. Wolff, E. Grzesiuk, and E. B. Maryon. 1986. Efficient homologous recombination of linear

MOL. CELL. BIOL. DNA substrates after injection into Xenopus laevis oocytes. Mol. Cell. Biol. 6:2053-2061. 7. Cassuto, E., T. Lash, D. S. Sriprakash, and C. M. Radding. 1971. Role of exonuclease and e protein of phage A in genetic recombination. V. Recombination of X DNA in vitro. Proc. Natl. Acad. Sci. USA 68:1639-1643. 8. Cassuto, E., and C. M. Radding. 1971. Mechanism for the action of X exonuclease in genetic recombination. Nature (London) New Biol. 229:13-16. 9. Chase, J. W., and K. R. Williams. 1986. Single-stranded DNA binding proteins required for DNA replication. Annu. Rev. Biochem. 55:103-136. 10. Church, G. M., and W. Gilbert. 1984. Genomic sequencing. Proc. Natl. Acad. Sci. USA 81:1991-1995. 11. Clark, A. J. 1973. Recombination deficient mutations of E. coli and other bacteria. Annu. Rev. Genet. 7:67-86. 12. Clark, A. J., and K. B. Low. 1988. Pathways and systems of homologous recombination in Escherichia coli, p. 155-215. In K. B. Low (ed.), The recombination of genetic material. Academic Press, San Diego. 13. Cox, M. M., and I. R. Lehman. 1987. Enzymes of general recombination. Annu. Rev. Biochem. 56:229-262. 14. Craigie, R., and K. Mizuuchi. 1985. Mechanism of transposition of bacteriophage Mu: structure of a transposition intermediate. Cell 41:867-876. 15. Doherty, M. J., P. T. Morrison, and R. Kolodner. 1983. Genetic recombination of bacterial plasmid DNA. J. Mol. Biol. 167:539560. 16. Dumont, J. N. 1972. Oogenesis in Xenopus laevis (Daudin). 1. Stages of oocyte development in laboratory maintained animals. J. Morphol. 136:153-180. 17. Fishel, R. A., K. Detmer, and A. Rich. 1988. Identification of homologous pairing and strand exchange activity from a human tumor cell line based on Z-DNA affinity chromatography. Proc. Natl. Acad. Sci. USA 85:36-40. 18. Folger, K. R., K. Thomas, and M. R. Capecchi. 1985. Nonreciprocal exchanges of information between DNA duplexes coinjected into mammalian cell nuclei. Mol. Cell. Biol. 5:59-69. 19. Formosa, T., and B. M. Alberts. 1986. DNA synthesis dependent on genetic recombination: characterization of a reaction catalyzed by purified bacteriophage T4 proteins. Cell 47:793806. 20. Ganea, D., P. Moore, L. Chekuri, and R. Kucherlapati. 1987. Characterization of an ATP-dependent strand transferase from human cells. Mol. Cell. Biol. 7:3124-3130. 21. Grzesiuk, E., and D. Carroll. 1987. Recombination of DNAs in Xenopus oocytes based on short homologous overlaps. Nucleic Acids Res. 15:971-985. 22. Gurdon, J. B., and D. A. Melton. 1981. Gene transfer in amphibian eggs and oocytes. Annu. Rev. Genet. 15:189-218. 23. Holliday, R. 1989. Untwisting B-Z DNA. Trends Genet. 5:355356. 24. Hsieh, P., and R. D. Camerini-Otero. 1989. Formation of joint DNA molecules by two eukaryotic strand exchange proteins does not require melting of a DNA duplex. J. Biol. Chem. 264:5089-5097. 25. Hsieh, P., M. S. Meyn, and R. D. Camerini-Otero. 1986. Partial purification and characterization of a recombinase from human cells. Cell 44:885-894. 25a.Jeong-Yu, S., and D. Carroll. Unpublished data. 26. Kohne, D. E., S. A. Levinson, and M. J. Byers. 1977. Room temperature method for increasing the rate of DNA reassociation many thousandfold: the phenol emulsion reassociation technique. Biochemistry 16:5329-5341. 27. Kolodner, R., D. H. Evans, and P. T. Morrison. 1987. Purification and characterization of an activity from Saccharomyces cerevisiae that catalyzes homologous pairing and strand exchange. Proc. Natl. Acad. Sci. USA 84:5560-5564. 28. Kucherlapati, R., and G. R. Smith (ed.). 1988. Genetic recombination. American Society for Microbiology, Washington, D.C. 29. Lichten, M., C. Goyon, N. P. Schultes, D. Treco, J. W. Szostak, J. E. Haber, and A. Nicolas. 1990. Detection of heteroduplex

VOL . 1 l, 1991


DNA molecules among the products of Saccharomyces cerevisiae meiosis. Proc. Natl. Acad. Sci. USA 87:7653-7657. 30. Lin, F.-L., K. Sperle, and N. Sternberg. 1990. Repair of doublestranded breaks by homologous DNA fragments during transfer of DNA into mouse L cells. Mol. Cell. Biol. 10:113-119. 31. Lin, F.-L., 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. 32. Lin, F. W., 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. 33. Low, K. B. (ed.). 1988. The recombination of genetic material. Academic Press, San Diego. 34. Maryon, E. 1990. The role of molecular ends in homologous recombination of DNA injected into Xenopus oocytes. Ph.D. thesis, University of Utah, Salt Lake City. 35. 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. 35a.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. 36. Maxwell, A., and M. Gellert. 1986. Mechanistic aspects of DNA topoisomerases. Adv. Protein Chem. 38:69-107. 37. Meselson, M. S., and C. M. Radding. 1975. A general model for genetic recombination. Proc. Natl. Acad. Sci. USA 72:358-361. 38. Mosig, G. 1987. The essential role of recombination in phage T4 growth. Annu. Rev. Genet. 21:347-371. 39. Orr-Weaver, T. L., and J. W. Szostak. 1985. Fungal recombination. Microbiol. Rev. 49:33-58. 40. Parsons, C. A., and S. C. West. 1988. Resolution of model Holliday junctions by yeast endonuclease is dependent upon homologous DNA sequences. Cell 52:621-629. 41. Radding, C. M. 1988. Homologous pairing and strand exchange promoted by Escherichia coli RecA protein, p. 193-229. In R. Kucherlapati and G. R. Smith (ed.), Genetic recombination. American Society for Microbiology, Washington, D.C. 42. Resnick, M. A. 1976. The repair of double-strand breaks in DNA: a model involving recombination. J. Theor. Biol. 59:97106. 43. Rochaix, J.-D., A. P. Bird, and A. Bakken. 1974. Ribosomal RNA gene amplification by rolling circles. J. Mol. Biol. 87:473487. 44. 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.

3287

45. 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. 46. Sadowski, P. D. 1985. Role of nucleases in genetic recombination, p. 23-40. In S. M. Linn and R. J. Roberts (ed.), Nucleases. Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y. 47. Seidman, M. M. 1987. Intermolecular homologous recombination between transfected sequences in mammalian cells is primarily nonconservative. Mol. Cell. Biol. 7:3561-3565. 48. Smith, G. R. 1988. Homologous recombination in procaryotes. Microbiol. Rev. 52:1-28. 49. Stahl, F. W. 1979. Genetic recombination: thinking about it in phage and fungi. W. H. Freeman, San Francisco. 50. Stahl, F. W., I. Kobayashi, and M. M. Stahl. 1985. In phage X, cos is a recombinator in the Red pathway. J. Mol. Biol. 181:199-209. 51. 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. 52. Sutcliffe, J. G. 1979. Complete nucleotide sequence of the Escherichia coli plasmid pBR322. Cold Spring Harbor Symp. Quant. Biol. 43:77-90. 53. Symington, L., P. Morrison, and R. Kolodner. 1985. Intracellular recombination of linear DNA catalyzed by the Escherichia coli RecE recombination system. J. Mol. Biol. 186:515-525. 54. 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. 55. Thaler, D., and F. W. Stahl. 1988. DNA double chain breaks in the recombination of phage X and of yeast. Annu. Rev. Genet. 22:169-197. 56. Thomas, C. A., Jr. 1966. Recombination of DNA molecules. Prog. Nucleic Acid Res. Mol. Biol. 5:315-337. 57. Wake, C. T., F. Vernaleone, and J. H. Wilson. 1985. Topological requirements for homologous recombination among DNA molecules transfected into mammalian cells. Mol. Cell. Biol. 5:2080-2089. 58. Wang, J. C. 1985. DNA topoisomerases. Annu. Rev. Biochem. 54:665-697. 59. West, S. C. 1990. Processing of recombination intermediates in vitro. Bioessays 12:151-154. 60. White, C. I., and J. E. Haber. 1990. Intermediates of recombination during mating type switching in Saccharomyces cerevisiae. EMBO J. 9:663-673. 61. Whitehouse, H. L. K. 1982. Genetic recombination: understanding the mechanisms. John Wiley & Sons, New York. 62. Zierler, M. K., N. J. Marini, D. J. Stowers, and R. M. Benbow. 1985. Stockpiling of DNA polymerase during oogenesis and embryogenesis in the frog, Xenopus laevis. J. Biol. Chem. 260:974-981.