Mol Gen Genet (1992) 234:1-13 © Springer-Verlag 1992

Gene conversion in the Escherichia coli RecF pathway: a successive half crossing-over model Kenji Yamamoto 1., Kohji Kusano 2, Noriko K. Takahashi 1'3, Hiroshi Yoshikura 1, and Ichizo Kobayashi 1'2'3 1 Department of Bacteriology,Medical School, Universityof Tokyo, 7-3-1 Hongo, Tokyo 113, Japan 2 Department of Molecular Biology,Institute of Medical Science, University of Tokyo, 4-6~1 Shiroganedai,Tokyo 108, Japan a National Children's Medical Research Center, Tokyo 154, Japan Received July 24, 1991

Summary. Gene conversion - apparently non-reciprocal transfer of sequence information between homologous DNA sequences - has been reported in various organisms. Frequent association of gene conversion with reciprocal exchange (crossing-over) of the flanking sequences in meiosis has formed the basis of the current view that gene conversion reflects events at the site of interaction during homologous recombination. In order to analyze mechanisms of gene conversion and homologous recombination in an E s c h e r i c h i a c o l i strain with an active RecF pathway ( r e c B C s b c B C ) , we first established in cells of this strain a plasmid carrying two mutant n e o genes, each deleted for a different gene segment, in inverted orientation. We then selected kanamycin-resistant plasmids that had reconstituted an intact n e o + gene by homologous recombination. We found that all the n e o + plasmids from these clones belonged to the gene-conversion type in the sense that they carried one n e o + gene and retained one of the mutant n e o genes. This apparent gene conversion was, however, only very rarely accompanied by apparent crossing-over of the flanking sequences. This is in contrast to the case in a r e c + strain or in a strain with an active RecE pathway ( r e c B C s b c A ) . Our further analyses, especially comparisons with apparent gene conversion in the r e c + strain, led us to propose a mechanism for this biased gene conversion. This "successive half crossing-over model" proposes that the elementary recombinational process is half crossing-over in the sense that it generates only one recombinant DNA duplex molecule, and leaves one or two free end(s), out of two parental DNA duplexes. The resulting free end is, the model assumes, recombinogenic and frequently engages in a second round of half crossing-ove, r with the recombinant duplex. The products resulting from such interaction involving two molecules of the plasmid would be classified as belonging to the gene-conversion type without crossing-over. We constructed a dimeric molecule * Present address: Department of Bacteriology,The National Institute of Health (Yoken), Tokyo 141, Japan Correspondence to: I. Kobayashi at address 21

that mimics the intermediate form hypothesized in this model and introduced it into cells. Biased gene conversion products were obtained in this reconstruction experiment. The half crossing-over mechanism can also explain formation of huge linear multimers of bacterial plasmids, the nature of transcribable recombination products in bacterial conjugation, chromosomal gene conversion not accompanied by flanking exchange (like that in yeast mating-type switching), and antigenic variation in microorganisms.

Key words: Homologous recombination - Plasmid linear multimer - Yeast mating-type switching - Antigenic variation

Introduction Gene conversion, replicative transfer of a stretch of sequence information from one DNA molecule to another homolog, has been observed in various organisms. Frequent association of gene conversion with reciprocal exchange (crossing-over) of the flanking sequences (Fig. 1A) in fungal meiosis has led to the current view that gene conversion reflects events at the site of interaction during homologous recombination as revealed by distant markers (Stahl 1979; Meselson and Radding 1975). The extent of association of gene conversion and crossing-over varies, however, among different systems (for example, see Klein 1984). It has been suggested that gene convesion and crossing-over represent separate events in meiosis and mitosis (Rossignol et al. 1984; Roman and Fabre 1983). A specialized case of gene conversion not accompanied by crossing-over is mating-type switching in yeasts (Strathern 1988). Such a system of gene conversion has an obvious advantage: it allows transfer of information between homologous sequences at different sites without altering chromosome organization by deletion or inversion. Possible molecular mecha-

A iiilUlllllllUlllllllllllll

/

\ ...................

iiiinlnlllUllll

[..........................

1st half crossing-over gene c o n v e r s i o n without crossing-over

i=

gene c o n v e r s i o n with crossing-over

iiiiilUllllllllllnll~ t

_1B

_U

BB B i 2nd half crossing-over

I ' [

neo~--~ .Iqg p I K 4 3

ikb

~

]]

B RI

C

4~rap+ ~ ~

V

"gene c o n v e r s i o n "gene c o n v e r s i o n at a without at a with crossing-over" crossing-over" (Type 5) (Type 6) Fig. 1. A Gene conversion and crossing-over (reciprocal exchange) of flanking sequences. B Substrate plasmid (pIK43). The two homologous duplex segments are drawn in parallel. The top segment has a 283 bp deletion (deletion a) between two NaeI sites, which removes one (C-terminal encoding) end of the neo gene. The NaeI site was inactivated by insertion of an 8 bp XhoI linker sequence. The bottom segment has a 248 bp deletion (deletion b), which removes the other (N-terminal) end of the neo gene. The two deletions are separated by 506 bp homology and are flanked by 1079 bp homology to their left and 231 bp homology to their right. The left, non-homologous "arm" (2321 bp long), including the ampicillin resistance gene and the replication origin, is derived from pBR322. The entire plasmid length is 14 795 bp. The restriction enzymes are: X, XhoI; RI, EcoRI; RV, E c o R V ; N, NaeI; B, BglII. Details of construction of the plasmid are given elsewhere (Yamamoto et al. 1988a). C "Gene conversion" type products. These two types of neo + plasmid have a structure that could be formed by intramolecular gene conversion at site a. The XhoI site in the top neo segment has been replaced by two NaeI sites. The EcoRI pattern has been changed only in the crossing-over type

nisms for gene conversion without crossing-over have been the subjects of much speculation (Rossignol et al. 1984; Roman and Fabre 1983; Duckett et al. 1988; Strathern 1988; Sobell 1972). We have been analyzing apparent gene conversion involving two mutant alleles of a gene in plasmids in various mutant strains of Escherichia eoli (Yamamoto et al. 1988a, b). We found that apparent gene conversion in a plasmid in an E. coli strain (recBC sbcBC), which is thought to have an activated RecF pathway, is not accompanied by crossing-over. Our further analyses lead us to propose a mechanism for this type of gene conversion. This "successive half crossing-over model" (Fig. 2) proposes that the elementary recombinational process is a half crossing-over in the sense that it generates only one

=F"

l=

gene c o n v e r s i o n without crossing-over

gene c o n v e r s i o n without crossing-over

Fig. 2A and B. Apparent gene conversion without crossing-over by two rounds of half crossing-over. A Half crossing-over between two intact duplex molecules generating one recombinant duplex and two ends. One end engages in a second event of half crossing-over with this recombinant duplex at a site at some distance from that of the first event, presumably after degradation and generation of a single-stranded tail. The final products are one duplex, with two closely spaced double-strand exchanges with heteroduplex regions, and two ends. B Half crossing-over between one intact duplex and one duplex with an end. The resulting end engages in a second event of half crossing-over as in A. The final products are one duplex, with two closely spaced double-strand exchanges, and one end

recombinant DNA duplex molecule out of two parental DNA duplexes and leaves one or two free end(s). The resulting end is, the model assumes, recombinogenic and engages in a second round of half crossing-over with the recombinant duplex. Our reconstruction experiments with a putative intermediate form gave results in accordance with this model. We show that this successive half crossing-over model can resolve several mysteries in homologous recombination. Materials and methods

Bacterial strains and plasmids. The following isogenic E. coli K12 strains were from Dr. A.J. Clark: JC7623 (= recB21 recC22 sbcB15 sbcC201) (Kushner et al. 1971 ; Lloyd and Buckman 1985), JC8111 (as JC7623 but reeF143), JC12190 (as JC7623 but recJ153), and JC7940 (as JC7623 but reeA142) (Horii and Clark 1973; Lovett and Clark 1984). BIK1212, a recN262 tyrA16: :TnlO derivative of JC7623, was made by P1 transduction from SP256 (Picksley et al. 1984), a gift from Dr. R. Lloyd. KEN88, a ruvC53 eda-51::TnlO (Shurvinton et al. 1984; Sharples et al. 1990) derivative of JC7623 was made by P1 transduction from KD2245 (Kusano et al. 1989), a gift from Dr. H. Nakayama. JC5519 (recB21 recC22) (Willetts and Clark 1969) is isogenic with JC7623 and was from Dr. A. Miura. ABl157 rec +) and JC8679 (= recB21 recC22 sbcA23), with similar genetic backgrounds, have been described previously (Yamamo-

to et al. 1988a, b; Kobayashi and Takahashi 1988), as have two recA1 strains, DH1 and DH5, not isogenic with the above strains. The plasmids used are described in the figure legends.

Table 1. Plasmid recombination by the RecF pathway Bacterial strain

Recombinant frequency (Kanr colony former/Amp r colony

former) x 104 Detection o f recombination. Each ampicillin-resistant (Amp ~) colony obtained by transformation with plK43 was grown with ampicillin selection (50 ~tg/ml with 200 ~tg/ml methicillin) to saturation in 5 rnl. The culture was then plated on kanamycin (Kan) agar (50 gg/ml) and on ampicillin agar. Electrophoresis. Intact plasmids (not cut with enzymes) were electrophoresed through 0.4% agarose in a low sodium buffer (40 m M TRIS-acetate, 2 m M Na2EDTA, pH 8.0) with 0.5 gg/ml ethidium bromide at a low voltage (70 V/35 cm) for 16 36 h. The restriction enzyme digests were electrophoresed through 0.7% agarose in a high sodium buffer (40 m M sodium acetate., 2 m M EDTA, pH 8.0) with 0.5 btg/ml ethidium bromide at a voltage of 100 V/12 cm for 30 min. Linear plasmid multimers were isolated by electrophoresis in 1% agarose in 8 9 m M TRIS-borate, 0.2 m M EDTA, p H 8.0 with 0.5 btg/ml ethidium bromide at 300 V / 2 4 cm for 1 h. Other methods. Substrate plasmid preparation, transformation by the calcium method, plasmid preparation by the boiling method, and other procedures have been described (Yamamoto et al. 1988a, b; Kobayashi and Takahashi 1988; Takahashi and Kobayashi 1990).

Results

recBC recBC recBC recBC recBC recBC recBC

sbcBC sbcBC sbeBC sbcBC sbcBC sbcBC

recA recF recJ recN ruvC

fecal

56.7 1.59 2.56 10.8 2.96 2.80 12.5

+-29.3 +_ 0.47 +_ 0.93 _+ 2.5 ± 1.13 +_ 2.15 +_ 3.3

(n = 5) (n=5) (n=5) (n=5) (n=5) (n=4) (n=5)

0.0696+_0.0348

(n= 10)

Each of n ampicillin-resistant (Amp r) colonies (50 btg/ml ampicillin and 200 btg/ml methicillin) obtained by transformation with pIK43

was grown in liquid medium with ampicillin selection (50 gg/ml ampicillin and 200 gg/ml methicillin) to an OD6oo of 0.3. The culture was diluted and plated on kanamycin agar (50 gg/ml) and on ampicillin agar for overnight incubation. The recombinant frequency is shown together with the standard deviation among n clones examined. The data for the recA1 strain, DH1, is from Yamamoto et al. (1988a). DH1 is not isogenic with the other strains and was used for preparation of substrate plasmid molecules with low background recombination. Relative plating efficiency (kanamycin/ampicillin) of DH 1 carrying a neo + plasmid ("gene conversion at a without crossing-over" type; Fig. 1C, left) was 1.07 and 1.21 in two experiments

plasmid in rec + and recBC sbcA strains (Yamamoto et al., 1988a, b).

Apparent 9ene conversion not accompanied by crossinff-ol)er

Experimental system

Our starting plasmid pIK43 (Fig. 1B) carries two homologous neo segments in inverted orientation. The upper segment has a deletion (deletion a) removing one end (C-terminus) of the neo gene, and the lower segment has a deletion (deletion b) removing the other end. This plasmid was introduced into an E. coli strain with recB recC sbcB sbcC mutations (JC7623) using ampicillin selection. The resulting strain (JC7623 [plK43]) generated Kan ~ clones at a frequency of 6 x 10 -3 (No. of Kan r colony formers/No, of Amp ~ colony formers) (Table 1). Plasmid D N A preparations isolated from these Kan ~ clones as well as the parental strain showed three bands on gel electrophoresis: bands at positions expected for a closed circle and an open circle of approximately the size of the parental plasmid and one broad band migrating more slowly (data not shown). This last species was identified as plasmid linear multimer (Cohen and Clark 1986; Kusano et al. 1989) by its behavior as a long linear molecule under different electrophoresis conditions (described in Materials and methods). We could not detect any circular dimers, closed or open, by ethidium bromide staining. This is in contrast to the presence of many circular dimeric forms of this

We recovered monomer closed circle D N A from these Kanr clones by electrophoresis and analyzed it with the restriction enzymes XhoI, EcoRI, NaeI, and Bg/II. No heterogeneity in plasmid structure was observed within a given Kan ~ clone. We were able to assign each of the clones to one of the predicted forms (Yamamoto et al. 1988a), as summarized in Table 2. The linear multimer fraction produced the restriction patterns expected for head-to-tail multimers of the monomer closed circle plasmid present in the same clone in all the cases examined (6/6). Plasmids in most of the Kanr clones had the restriction patterns shown in Fig. 3 and belonged to the type shown on the left, in Fig. 1C (Table 2). We define this as a gene conversion type (gene conversion at deletion a) because it possesses one recombinant allele and one parental allele. Plasmids in some of the clones showed the restriction patterns expected from gene conversion at deletion b. If the responsible gene conversion events were intramolecular and accompanied by reciprocal exchange (crossing-over) of the flanking sequences, the region between the two homologous segments (the left "arm" in Fig. 1B) would be inverted with respect to the outside markers (the right arm in Fig. 1B) to give the form on the right in Fig. 1C. We found, however, only one clone

4 Table 2. Classification of recombinant (neo +) plasmids

Bacterial

Plasmid

Product

strain

Gene conversion type at s i t e a

at s i t e a

Other

Mixture Total

at s i t e b

without with without crossing- crossing- crossingover over over recBC sbeBC

pIK43 Exp. 1 Exp. 2 Exp. 3 Exp. 4 Exp. 5 Total

8 18 11 9 16 62

0 0 0 0 1 1

1 2 3 0 4 10

0 0 0 0 0 0

0 0 0 0 0 0

9 20 14 9 21 73

recBC sbcA

plK43

60

17

1

8

0

87

rec +

pIK43

(43)

(67)

(0)

(15)

-

(125)

Kan ~clones generated during growth of the cells carrying pIK43 were isolated as colonies. Plasmid DNA (monomer closed circles) from each of them was analyzed with XhoI, NaeI, EeoRI, and BglII and classified (Yamamoto et al. 1988a). In the ree + strain, plasmid molecules from the primary Kan~transformants showed heterogeneous forms. These were introduced into a reeA strain, and the resulting secondary transformants were analyzed and classified. The results for the reeBC sbeA strain and the rec + strain are from earlier publications (Yamamoto et al. 1988a, b). Mixture, mixture of crossing-over and non-crossing-over types

NaeI

EcoRI

Fig. 3. Form of neo ÷ plasmids. Plasmid DNA was isolated from each of three Kanr colonies obtained by streaking of an Amp r pIK43 transformant of JC7623. Closed circle monomer DNA was purified by agarose gel electrophoresis. It was cut with restriction enzymes and electrophoresed through 0.7% agarose. The NaeI pattern reveals apparent gene conversion at site a. The EcoRI pattern shows absence of apparent crossing-over. The numbers indicate fragment lengths in kb. The white triangle indicates the origin. Parent, pIK43 carrying plasmid molecules of this crossing-over type (Table 2). This strong bias towards the non-crossing-over types is in contrast to the case in a r e c B C s b c A strain and in a rec ÷ strain (Table 2). In a r e c B C s b c A strain we found a significant number of clones carrying the "gene conversion with crossing-over" type ( Y a m a m o t o et al. 1988a). In a rec + strain each Kanr clone contained multiple types of recombinant plasmid molecules. There were m a n y gene conversion with crossing-over "type molecules" (Table 2; Y a m a m o t o et al. 1988a). This bias is not likely to have resulted from a selective advantage of the non-crossing-over types in the r e c B C

s b c B C strain for the following reasons. All of the four

plasmid species - gene conversion at a without crossingover, gene conversion at a with crossing-over, reciprocal crossing-over between a and b, and gene conversion at b without crossing-over - prepared in a recombinationdefective strain (DH1) transformed the host bacterial strain with similar efficiencies. They produced, f r o m 10ng D N A , 1.09x103 ' 1.51x103 , 0.73x103 , and 1.83x 10 a of Kanr transformants, respectively. We analyzed more than 11 transformants from each of the four types but found no clone in which the starting plasmid type had been replaced by another type. We found that the doubling times of bacteria carrying type 5 and type 6 plasmids in the presence of ampicillin selection were not significantly different. JC7623 [type 5] showed a doubling time of 95 +__12 min (n = 3) for A m p r colony former and 91 + 10 min ( n = 3 ) for Kanr colony former. JC7623 [type 6] showed a doubling time of 9 0 + 2 3 min ( n = 3 ) for A m p r colony former and 90_+ 16 min (n = 3) for Kanr colony former.

C o m p a r i s o n with a p p a r e n t g e n e conversion in a rec + strain

Several current models can provide explanations for this biased conversion. These include formation of "patchtype" heteroduplex and double-strand break repair, which will be discussed later. We were led to consider another model after we had analyzed this apparent conversion in comparison with apparent conversion in a rec + strain ( Y a m a m o t o et al. 1988a), for which we demonstrated the mechanism illustrated in Fig. 4A. We here describe our experiments and ideas and will compare this model with other models in more detail in Discussion.

1st half crossing-over L i n e a r multimer

B

A

C

1st half crossing-over

1st crossing-over

~ Type 1 dimer 2nd c r o s s i n g - o v e r ~ 0 0

+

Type 5

Type 6

----~

'd

L i n e a r dimer 2nd half crossing-over ~O(~

Type 5

Fig. 4A-C. Proposed origins of the gene-conversion type products in A ree + and B, C r e c B C s b e B C bacteria. A ree +. Reciprocal crossing-over involving two plasmid molecules at some site between deletion a and deletion b produces a "type 1" dimer. Descendant molecules of this type 1 dimer then produce, by a second, intramolecular crossing-over between direct repeats, various neo + monomers. These include gene conversion type products with and without apparent crossing-over. B r e e B C s b c B C . Half crossing-over between the bottom n e o segment of one mole,cule and the top n e o

In r e c + cells the "bottom" n e o segment of one plasmid molecule and the "top" n e o segment of a second molecule undergo i n t e r - m o l e c u l a r reciprocal crossing-over between site a and site b as illustrated in Fig. 4A. The resulting n e o + dimer, which we call a type 1 dimer (Fig. 4A top right), renders the host cell Kanr. Descendants of this type 1 dimer produce various n e o + monomers by intra-molecular crossing-over within the resulting Kan r clone. Figure 4A indicates two routes for the second crossing-over: the left one leads to a plasmid of the type gene conversion at a without crossing-over, and the right one leads to a plasmid of the type gene conversion at a with crossing-over. This mechanism in the r e c + strain, involving the type 1 dimer, however, cannot account for the apparent biased gene conversion observed in the r e c B C s b c B C strain for the following reasons. First, each of the Kan r clones from the r e c + strain contained the type 1 dimer (Yamamoto et al. 1988a). No circular dimer was, however, detected in the r e c B C s b c B C strain as mentioned above. Second, each of the Kan r clones from the r e c + strain contained multiple types of n e o + plasmid (Yamamoto et al. 1988a). These are proposed to be "cousins" generated by different recombination events from replicating type 1 dimer molecules within one clone (Fig. 4A; Yamamoto et al. 1988a). Each of the Kan r clones from the r e c B C s b c B C strain, however, had only one type of n e o + plasmid. Third, when the type 1 dimer was introduced into the r e c B C s b c B C strain, each of the resulting colonies carried both the non-crossing-over type (gene conversion at a without crossing-over; Fig. 1C left) and the crossing-over type (gene conversion at a with crossing-over; Fig. 1C right) as shown in Fig. 5C, (right-hand lanes of the gel) and summarized in Table 3. The bias toward the non-crossing-over type was not seen.

2nd Linear multimer half crossing-over ~Oq)

Type 6

Type 5

Type 6

segment of another molecule produces a linear dimer intermediate. This linear dimer then produces, by a second, intramoleeular exchange between direct repeats, various types of n e o + monomers including the two gene conversion types. The end favors the route leading to the gene conversion type without apparent crossing-over. C r e c B C s b c B C (involving plasmid linear multimers). The first half crossing-over is initiated by an end of a plasmid linear multimer. The n e o + monomers are produced as in B

A

B

pKEN1 1 / X h o I

f

Type 5

C

pKEN12/XhoI

II

Ii

pKEN10

I

pe

6

Fig. 5A-C. Reconstruction experiments introducing dimeric plasmids into the cells. Top The dimeric plasmids, p K E N I 0 is a type 1 dimer (Fig. 4A), a neo + dimer produced by reciprocal crossing-over between two parental molecules (pIK43) (Yamamoto et al. 1988a). p K E N l l and pKEN12 are its derivatives carrying only one X h o I site. They were obtained by partial digestion of p K E N 10 with X h o I , filling-in by Klenow fragment of D N A polymerase I, and ligation. pKEN11 and pKEN12 were cut with X h o I before transfer into the r e c B C s b c B C strain. Bottom The neo + plasmids in the resulting transformants. Monomer closed circle plasmids were recovered from each of the resulting Kanr colonies. They were cut with E c o R I and electrophoresed to distinguish between the non-crossing-over type (8 and 7 kb) and the crossing-over type (10 and 5 kb). Results for four Kanr clones are shown for each transformation. Numbers indicate fragment lengths in kb. Lane 11 shows small but significant amounts of 7 and 8 kb fragments

Table 3. Classification of recombinant (neo +) plasmids in the reconstruction experiments

Bacterial strain

Plasmid

Product Gene conversion type

Other Mixture Total

at sitea at sitea at siteb without with without crossing- crossing- crossingover over over recBC sbcBC pKENI0 recBC sbcBC pKEN11 recBC sbcBC pKENll/XhoI Exp. 1 Exp. 2 Exp. 3 Total recBC sbeBC pKEN12 reeBC sbeBC pKEN12/Xhol Exp. 1 Exp. 2 Exp. 3 Total

0 0 11 12 12 35 0 0 0 0 0

0 0 0 0 0 0 0 12 12 10 34

0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0

3 4 1 0 0 1 4 0 0 0 0

3 4 12 12 12 36 4 12 12 10 34

Kan ~colonies were selected immediately after expression following transformation with the dimeric plasmids pKEN 10, pKEN 11, and pKEN 12. The plasmids were classified as in Table 2. In a fourth experiment the plasmids were analyzed only with EeoRI. All the clones derived from pKEN11/XhoI (6/6) showed the non-crossing-over pattern (as in Fig. 5, bottom left). All the clones from pKEN12/XhoI (6/6) showed the crossing-over pattern (as in Fig. 5, bottom middle), but one of them showed an extra band of approximately 6 kb. All the clones from intact pKEN10 (6/6) showed the mixture pattern (as in Fig. 5, bottom right). Mixture, mixture of the crossing-over type and the non-crossing-over type as analyzed with EcoRI. pKEN11/XhoI, pKEN11 cut with XhoI. pKEN12/XhoI, pKEN12 cut with XhoI

I f the type 1 dimer were an intermediate as illustrated in Fig. 4A, both the crossing-over type and the non-crossingover type would result. We observed only the non-crossing-over type. Therefore, the type 1 dimer is not likely to be an intermediate form.

the end, however, causes bias toward the former route (Fig. 4B, left). The order of the two exchanges m a y be reversed in producing gene conversion at a without crossing-over. Generation of gene conversion at b without crossing-over can be explained in a similar way.

A mechanism: Two rounds o f half crossino-over Reconstruction f r o m the intermediate f o r m A modification (Fig. 4B) of the above scheme (Fig. 4A) can, however, explain the biased gene conversion we observed in the reeBC sbeBC strain. Here we assume that the elementary recombination process in the recBC sbcBC strain is half crossing-over rather than full crossing-over. We define half crossing-over as a homologous recombination process generating only one recombinant D N A duplex, and leaving two ends (or one end) free, out of two parental D N A duplexes (Fig. 2A and B). (This is a specific type of non-conservative recombination, which is defined here as recombination between two duplexes generating only one recombinant duplex.) The resulting end is, we assume, recombinogenic and engages in a second round of exchange with the recombinant duplex at some distance f r o m the first exchange. The net outcome is a duplex with two close double-stranded exchanges (Fig. 2A). In the modified scheme in Fig. 4B, a half crossing-over between the b o t t o m neo segment of one plasmid and the top neo segment o f another plasmid generates a linear dimer instead of a circular dimer. Two alternative routes of second exchange lead to the non-crossing-over type product and to the crossing-over type product as in the case with type 1 dimer. The recombinogenic nature of

The mechanism in Fig. 4B predicts that the dimeric intermediate form will produce only "gene conversion without crossing-over" type m o n o m e r s and no "gene conversion with crossing-over" type m o n o m e r in the recBC sbcBC strain. This prediction was borne out in the reconstruction experiments shown in Fig. 5. We prepared a linear dimer molecule that mimics the proposed intermediate form as follows. The type 1 dimer (pKEN10) carries two XhoI sites (compare Fig. 1B and Fig. 5C). We inactivated the upper one (see Fig. 5 legend) and the resulting plasmid (pKEN11) was cut with J(hoI (Fig. 5A). This linear dimer resembles the hypothetical intermediate in Fig. 4B in that a double-strand break is present opposite to the recombinant segment. We introduced this linear dimer form into the recBC sbcBC cells. Plasmid preparations from these transformants contained m o n o m e r s and linear multimers but no detectable circular dimers. We analyzed m o n o m e r closed circles and found only the non-crossing-over type as shown in Fig. 5A (bottom) in all the clones examined (summarized in Table 3 and its legend). The opposite result was obtained when a doublestrand break was present on a segment not involved in

Table 4. Effect of sbcBC mutations in the reconstruction experi-

ments Plasmid

Selection

Exp. 1 pKEN11/J(hoI Amp pKEN11/XhoI Kan pKEN 11

Amp

Exp. 2 pKENll/XhoI

Kan

pKEN11

Kan

Colonyformingunits recBC sbcBC

recBC sbc +

300 340 280 260 28000 36000

48 43 40 35 39000 37000

400 470 18500 21000

64 56 26000 24000

Equal aliquots of a plasmid DNA preparation were incubatedwith or without J(hoI. Approximately0.25 gg was introduced into the bacteria in duplicate. Kan, kanamycinselection. Amp, ampicillin selection, pKEN11/J(hoI, pKEN11 cut with XhoI the hypothetical first exchange (Fig. 5B). In this case we inactivated the other XhoI site of pKEN10 to construct pKEN 12. pKEN 12 cut with J(hoI produced clones carrying only the crossing-over type plasmids (Fig, 5B and Table 3). The intact form of these two plasmids (pKENll and pKEN12) produced clones containing both the non-crossing-over type and the crossing-over type (as did the type 1 dimer) when introduced into the cells (Table 3). The ratio of Kanr transformants to Amp ~ transformants from pKEN11 cut with J(hoI was close to unity (Table 4). This is consistent with our hypothesis that this form resembles one of the recombination intermediate forms. Effect o f sbc and rec mutations

The strain we used in this study (JC7623) carries recB21, recC22, sbcB15 and sbcC201 mutatio~s. The recB21 and recC22 mutations inactivate the RecBCD enzyme and decrease recombination between bacterial chromosomes following conjugation. The sbcB- and sbcC- mutations suppress the latter defect (Kushner et al. 1971 ; Lloyd and Buckman 1985). Mutations in the recA, recF, recJ, recN, and ruvC genes again decrease recombination associated with conjugation in this recBC sbcBC background (Horii and Clark 1973; Lovett and Clark 1984; Lloyd et al. 1983, 1984). It is assumed that the products of these genes constitute the RecF pathway of recombination in this background (Clark and Low 1988). The recombination frequency of our plasmid (pIK43) in the recBC- s b c B C - background we described was also decreased by mutations in the recA, recF, recJ, recN, and ruvC genes (Table 1) although the decrease was not so great as in conjugtion. These results indicate that the RecF pathway as defined by bacterial recombination is responsible for the gene conversion we observed.

We also found that the recombination frequency of our plasmid was lower in a recBC sbc + strain than in our recBC sbcBC strain (Table 1) in accord with the observations on conjugation. The effect of the sbcBC mutations was also observed in the reconstruction experiments shown in Fig. 5. A recBC- sbc ÷ strain and a recBC- sbcBC- strain were transformed with approximately the same efficiency by the circular form of the dimer. However, when the linearized form was introduced, the recBC sbc + cells produced fewer transformants than did recBC sbc- cells (Table 4).

Discussion

The successive half crossing-over mechanism illustrated in Fig. 2 can explain production of "gone-conversion without crossing-over" type plasmids. If half crossingover takes place betwen two plasmid molecules as illustrated in Fig. 4B, the resulting intermediate form should produce biased conversion products as shown by the reconstruction experiments (Fig. 5 and Table 3). The hypothesis of half crossing-over implies that recombination is non-conservative in the sense that recombination between two duplexes generates only one recombinant duplex. We have demonstrated that recombination is non-conservative in the recBC sbcBC strain that we used in the present study (Takahashi et al. 1992). The ends generated by restriction enzymes and by )~ terminase at cos are recombinogenic in this genetic background (Luisi-DeLuca et al. 1989; Takahashi et al. 1992; Thaler et al. 1989). The hypothesis of half crossing-over also implies that non-conservative recombination creates DNA molecules with one or two broken end(s). We are now testing this prediction. We did not detect linear dimers in the present experiment. This is not surprising for the following reasons: (1) the recombination rate of circular plasmids is relatively low. It is approximately 10 .3 per generation according to Luisi-DeLuca et al. (1989), and our results with transformation with a circular dimer in Fig. 5C are consistent with this order of magnitude. (2) This intermediate might be very short lived. Luisi-DeLuca et al. (1989) have reported that the recombination rate observed with a plasmid carrying direct repeats is much higher in a recBC sbcBC strain than in a rec ÷ strain. With our plasmid the recombination rates do not differ greatly in the recBC sbcBC and rec + strains (Table 1 and Yamamoto et al. 1988a). This is readily explained by the successive half crossing-over model. Only one round of half crossing-over is sufficient to produce a recombinant monomer from direct repeats, whereas successive rounds of half crossing-over are necessary to produce a recombinant monomer from inverted repeats (Fig. 4B, C). In our reconstruction experiments many fewer transformants were produced with the linear form than with the intact form of the plasmid (Table 4). Luisi-DeLuca et al. (1989) found that transformation by a linear form of a head-to-tail dimer of a plasmid is as efficient as that

A (i)

Repcilan itg {f~'~~-.

{(~'~ )) plasmid ~HF°rk

~

'~ ~__x.~

~ alerl °$sing-0°ver

[ ~ recBCD+

/

recA+

tst half x~ossing_ove r recJ +recF+

(ii)

( i ~

recO" --'%~ dii~

"transcribable recombination product"

,~nd fver crosstn ilag-lo

sbcB-c-

-%

(iv)

J-

recN +

ruvB+ viable recombinant progeny

Rollincigrcle (ii~

0

dii)

(iv)

0

l

@

@*

"O

O

0 l

0

©

©

Fig. 6A-C. Implications of the half crossing-over model. A Possible contribution of half crossing-over mechanism to plasmid linear multimer formation. (i) Half crossing-over between a replicating circle and another circle would generate a rolling circle. (ii) Half crossing-over between the end of a linear multimer and a replicating circle would generate a double rolling circle. (iii) Half crossing-over between the end of one rolling circle and the tail of another rolling circle would also generate a double rolling circle. (iv) Half crossingby its circular form. This contrast can be explained by the nature of the recombination event being selected. In their experiment recombination between any pair o f homologous sequences could produce a monomer circle. Our experiments selected recombination between a pair of specific regions (neo regions). Homologous recombination elsewhere would not produce a circular molecule. We do not know what triggers the first exchange in the model in Fig. 2A. One o f the duplexes may first suffer a double-strand break, which would result in two recombinogenic ends. Alternatively, one parent might carry only one end, to begin with, as illustrated in Fig. 2B. This end might be produced by rolling-circle replication. A large amount of huge plasmid linear multimers are formed in recBC sbcBC strains (Cohen and Clark 1986). The ends of these linear multimers may initiate half crossing-over. This and a second half crossing-over, which is intramolecular, might lead to monomers of the non-crossing-over gene conversion type as illustrated in Fig. 4C.

1.

0 over between the end of a linear multimer and a circle would elongate the multimer. B Transcribable recombination products in bacterial conjugation. C Gene conversion as explained by successive half crossing-over. The donor chromosome is assumed to be partially replicated prior to the first exchange. (i) Chromosomal gene conversion without crossing-over. (ii) Chromosomal gene conversion with crossing-over. (iii) Apparent gap repair without crossingover. (iv) Aparent gap repair with crossing-over

Possible contribution of half crossing-over to plasmid muhimer formation On the other hand, formation of these linear multimers depends on gene functions involved in the ReeF pathway of recombination (recA, recF, recJ, and reeO) (Cohen and Clark 1986; Silberstein and Cohen 1987; Kusano et al. 1989). Half crossing-over between a replicating circle and another circle might generate a rolling circle as illustrated in Fig. 6A (i). [Another possible role for the recombination functions is protection of the end of a rolling circle by homologous pairing (Niki et al. 1990).] Half crossing-over between an end of a rolling circle and a replicating circle will generate a double rolling circle as illustrated in Fig. 6A (ii). Half crossing-over between an end of a rolling circle and a tail of another rolling circle will have the same consequence (Fig. 6A (iii)). These double rolling circles may be the exonuclease-resistant multimers described by Kusano et al. (1989). Elongation of the linear multimers can be explained by the rolling-circle replication mechanism

(Cohen and Clark 1986). Half crossing-over between an end of a linear multimer and a circle is, however, another way of causing elongation of the multimer as illustrated in Fig. 6A (iv). Kinetic experiments (Cohen and Clark 1986) and density-shift experiments (Silberstein et al. 1990) indicate that the former route of elongation is the major route (although the genetic ba,ckgrounds of their strains are not identical with ours). But they cannot exclude a contribution of the latter route. If these kinds of intermolecular half crossing-over events happen to take place frequently between the inve,rted repeats of our plasmids, this would result in characteristic restriction patterns (head-to-head instead of head-to-tail). We did not see these after ethidium bromide staining (see Results), but we may need a more sensitive assay for their detection.

Roles of rec and sbc gene products We found parallelism in genetic requirements between bacterial conjugative recombination through the ReeF pathway and our plasmid recombination. Mutations in the recA, recF, reeJ, recN, and ruvC genes decreased recombination in a recBC sbcBC background. The reeBC sbcBC strain showed higher recombination than the recBC sbc + strain. We propose that the sbcBC mutations affect at least the second recombination event in the two-step model (in Fig. 2 and Fig. 4B) because on effect of sbcBC mutations was observed in the reconstruction experiments (Table 4). Homologous recombination between direct repeats on a linear molecule has been reporl:ed to need only a subset of these gene functions (recA, recF, recJ, and recO) but not recN or ruvB functions (Luisi-DeLuca et al. 1989). We observed an effect of recN and ruvC mutations on apparent gene conversion with a plasmid with inverted repeats (Table 1). If we assume that the two steps in Fig. 2 have different genetic requirements, these contrasting results can be explained. The first step might need only the primary subset (recA, recF, recJ, and recO) of the RecF pathway genes. The second step might need the second subset (recN and ruvC) (and possibly the primary subset as well). Recombination between direct repeats might require only one event of intramolecular half crossing-over and hence only the first subset of genes. This hypothesis needs to be tested experimentally. Examination of the effects of rec mutations in our reconstruction experiments, however, has been difficult for technical reasons (including low transformation efficiency of the strains). The known and inferred properties of these and other gene products revealed by many researchers have helped us to speculate on the details of the hypothesized twostep half crossing-over process. Replication or some other process may produce double-strand ends in a DNA duplex. Distribution of crossing-over along the lambda chromosome is more or less uniform in the presence of replication but is concentrated at the end (cos) in the

absence of replication (Thaler et al. 1989). The recQ helicase (Umezu et al. 1990) may expose two singlestrand tails at the end. The recJ gene product, a 5' to 3' single-strand specific exonuclease (Lovett and Kolodner 1989), may digest the 5' tailed end. The remaining 3' tailed single-strand may stimulate recombination. Exonuclease I (ExoI), the sbcB gene product, when it is present, may degrade this 3' tailed single strand (Lehman 1960). The RecA and RecF proteins may help the 3' tailed single strand to pair with a homologous duplex (Shibata et al. 1979; Madiraju et al. 1988). The RuvA/ RuvB complex renatures cruciform structures (Shiba et al. 1991), and RuvC protein resolves the Holliday structure (Connolly et al. 1991; Iwasaki et al. 1991). We propose that proper processing of the above intermediate form by these Ruv proteins produces one recombinant duplex and one 3' tailed recombinogenic single-stranded end. The latter is ready for another round of recombination unless erased by ExoI, the sbcB gene product. The former is completed by repair synthesis and ligation at the single-strand interruptions. Processing other than by these Ruv gene products would somehow produce one recombinant duplex but would not leave a recombinogenic end. The resulting recombinant duplex carries a heteroduplex with a 3' overhang as demonstrated (Siddiqi et al. 1991).

Bacterial recombination: Transcribable recombination products Our model of successive half crossing-over (Fig. 2) can explain recombination following bacterial conjugation, and especially the nature of transcribable recombination products (Birge and Low 1974). Although recB- sbcBC + mutants do not produce recombinant colonies, they synthesize products of recombinant type genes and, therefore, transcribable recombination products (Birge and Low 1974). Their formation depends on the function of a subset of the above genes (recA, reeF, reeJ, and reeO) but not on the function of recN or ruv (Lloyd et al. 1987). We propose that this transcribable recombination product is a linearized bacterial chromosome produced by the first half crossing-over in our model as illustrated in Fig. 6B. In the recB strain, a first half crossing-over by reeA, reeF, reeJ, and recO functions would generate a linear chromosome of the recombinant genotype, which would produce recombinant type protein but would not replicate to produce a recombinant colony. Some special feature of the incoming DNA might make this recombination resistant to ExoI. We propose that, in the presence of sbeBC mutations only, a second half crossing-over would restore a circular chromosome of the recombinant genotype, which can generate a recombinant colony. The slow time course and the clustering of exchanges in reeBC sbcBC cells (Mahajan and Datta 1988) are consistent with our two-step model.

10 Alternative mechanisms

We briefly discuss here three other models that might provide an explanation for the bias in gene conversion that we observe. The first is formation of a heteroduplex of the patch type. A patch covering one of the deletion mutations would replicate to give the "gene-conversion without crossing-over" type plasmid. A large region of heterology can be included in a heteroduplex at least in a 9am + rec ÷ background (Lichten and Fox 1983, 1984). The bias toward the non-crossing-over type as opposed to the crossing-over type [patch vs splice (Stahl 1979)] can be explained if an end of a linear multimer attacks a circular molecule. Splice formation would result in an elongated linear multimer. Patch formation, however, could result in a neo ÷ monomer after replication. Alternatively a heteroduplex is formed within one plasmid molecule, and some topological constraint on crossing-over leads primarily to the patch type. This model and our model are difficult to distinguish because a patch and a pair of close double-stranded exchanges are difficult to distinguish (Stahl 1979). Our recent demonstration of the non-conservative nature of exchange in the recB recC sbcB sbcC strain (Takahashi et al. 1992), however, implies that patch formation must be non-reciprocal even if it is part of the mechanism. The second type of model, involving double-strand break repair (Szostak et al. 1983; Kobayashi 1991), can explain apparent conversion of deletions in plasmids like the one we observed (Kobayashi 1991; Kobayashi and Takahashi 1988; Takahashi and Kobayashi 1990). We, however, failed to detect double-strand gap repair reaction or double-strand break repair reaction following transfer of a gapped form of the plasmid into the recBC sbcBC strain (N. Takahashi, K. Kusano, Y. Kitamura, H. Yoshikura, I. Kobayashi unpublished) while such repair can be observed with the E. coli RecE (Kobayashi and Takahashi 1988) and the )~ Red pathways (Takahashi and Kobayashi 1990). These latter gap repair events were often accompanied by crossing-over (Kobayashi and Takahashi 1988; Takahashi and Kobayashi 1990). These models must be modified in some way in order to explain the bias toward the non-crossing-over type (see Strathern 1988). The third type of model we consider emerged from analysis of recombination between homologous sequences on the Salmonella chromosome (Mahan and Roth 1989, 1990). Homologous recombination between direct repeats on this chromosome was not decreased by a recB- mutation. Some pairs of repeats in inverted orientation did not undergo crossing-over of flanking sequences, which should result in chromosomal inversion. The latter observation is similar to our present result in the plasmid system. The observations in Salmonella can he explained well by our two-route theory: non-conservative recombination for the RecF pathway and conservative recombination for the RecBCD pathway. We assume that the non-conservative RecF pathway rather than the apparently conservative (Kobayashi et al. 1984a) RecBCD pathway makes a large contribu-

tion to homologous recombination in some regions along the chromosome. Mahan and Roth (1989, 1990), however, have proposed a one-pathway, two-step, model to explain their findings. Its first step is a recBC-independent "half reciprocal" recombination, which produces one recombinant molecule and two ends. In the second step, they propose, the RecBCD enzyme enters DNA from the end, as in the case of Z activation (Kobayashi et al. 1982, 1984a, b; Taylor et al. i985). Its action completes the second recombinant molecule. Thus the overall process is "fully reciprocal" recombination. We think this model is unlikely for the following reasons: (i) there has been no evidence for recombination enhancement in the immediate vicinity of double-strand breaks by the RecBCD enzyme. The RecBCD enzyme rather appears to inhibit enhancement at double-strand breaks, at least, by the £ Red pathway (Stahl et al. 1985; Thaler et al. 1987b). (ii) In this model the end must be similar to a blunt end in shape in order to serve as an entry site for the RecBCD enzyme. An end with a long single-strand tail cannot serve as an entry site (Kobayashi et al. 1982; Taylor and Smith 1985). Then the first exchange is likely to leave a double-strand gap around the break. If this is the case, this one-pathway model has to assume a conservative double-strand gap repair reaction by the RecBCD system in the second step. We have failed to detect such conservative double-strand gap repair in a recBC + strain (N. Takahashi, K. Kusano, H. Yoshikura, I. Kobayashi, unpublished). (iii) They found fully reciprocal exchange (by their definition) in a recB strain. This is in contrast to an economical prediction of their one-pathway model (Mahan and Roth 1989).

Other organisms: yeasts, mammals, Trypanosoma and

Borrelia Our successive half crossing-over model can explain chromosomal gene conversion if we assume that donor DNA is partially replicated prior to the first exchange (Fig. 6C, i, ii). This version is formally equivalent to the uni-sex circle model proposed earlier (Stahl 1979, 1969) although emphasis in our model is on generation of recombinogenic end(s) by a non-conservative exchange. We applied this concept to mating-type switching in the yeast Saccharomyces (Fig. 7), which has been explained by various versions of the double-strand break repair model (Strathern 1988). We are testing the possibility that our plasmid gene conversion takes place within a replicating molecule as in this scheme. Previous results interpreted in terms of the doublestrand break repair models in vegetatively growing yeast cells (Orr-Weaver et al. 1981; Orr-Weaver and Szostak 1983; Jayaram 1985; Rudin et al. 1989) can also be explained by this model, as illustrated in Fig. 6C (Campbell 1984) and in Fig. 4 of our earlier publication (Takahashi and Kobayashi 1990). Although £ Red recombination initiated by a double-strand break often involves invasion of different homologs by the two ends (Thaler et al. 1987a), it can carry out true double-strand break

11 A

E (ars)

Ya

X

(a!s)

Zl

HMRa . . . .

-

=

X

~

--

MAT

(~

Y(t Zl

HOcut

Acknowledgments. We thank Drs. A.J. Clark, R. Lloyd, H. Nakayama, A. Miura and their colleagues for gifts of bacterial strains. T. Tohda and Y. Sunohara helped us in preparation of the manuscript. We are grateful to Drs. Y. Kitamura and I. Saito for advice in experiments, and to Drs. H. Shinagawa, H. Nakayama, S. Hiraga, C.I. White and J. Haber for communication of unpublished results. Drs. J. Roth, J. Haber, D. Carrol, I. Herskowitz, A.J. Clark, and J. Rine provided critical and helpful discussion. Drs. M.S. Fox and F.W. Stahl provided comments on the manuscript. Drs. T. Takeda and H. Ikeda supported us in part. K.Y. was supported by a fellowship from the Japan Society for the Promotion of Science.

D

E

X -

-

~

Ya

z1

~

~

HMRa

MAT a X

Ya

half crossing-over model rather than conservative mechanisms (including the double-strand break repair models) (Fig. 6C) (Kobayashi 1992). Antigenic variation in unicellular microorganisms can be explained by post-replicational half crossing-over (Kobayashi 1992). Antigenic variation can be effected by only one half crossing-over event between mini-chromosomes in Trypanosoma ("telomere conversion") (Pays etal. 1985) and between linear plasmids in Borrelia (Plasterk et al. 1985).

Zl

Fig. 7A-E. A successive half crossing-over model for Saccharomyces mating-type switching. A HMRa is a donor duplex. MATc~ is the recipient duplex. Y a and Ya are non-homologous. X and Z1 are common to the donor and the recipient. The donor HMR is flanked by silencers, E and I, which have ars (plasmid replication origin) activity. B HO endonuc][ease makes a doublestrand break at the Y-Z junction of the recipient. The donor duplex is partially replicated by the ars activity of the silencers. Cell type (a vs a) determines which donor (HMLa or HMRa) is to be replicated (directionality). At least one of the daughter duplexes is recombinogenic. C The end in the Z region initiates a first half crossing-over with one of the daughter duplexes of the donor. The other end is degraded to a variable extent into the X region. A single-strand tail is produced. D This single-strand tail initiates a second half crossing-over with the same daughter. The extra DNA of the donor is trimed off. E The recipient has a coconversion and heteroduplex pattern as observed (McGill et al. 1989). The donor genotype remains unchanged. Scheduled re.plication of the chromosome and cell division produce two cells of the switched mating type. This model can explain the observed intermediates (White and Haber 1990)

repair by gene conversion as demonstrated by our control experiments with marked plasmids (Takahashi and Kobayashi 1990). Various aspects of mammalian homologous recombination can be explained by the half crossing-over mechanism (see Kobayashi 1992 for review). (i) Homologous recombination events between incoming DNA molecules in mammalian somatic cells have been explained by n o n - c o n s e r v a t i v e m e c h a n i s m s ( C h a k r a b a r t i a n d S e i d m a n 1986; Lin et al. 1987; Brouillette a n d C h a r t r a n d 1987; K i t a m u r a et al. 1990). These n o n - c o n s e r vative r e c o m b i n a t i o n events m i g h t generate D N A ends a n d thus m i g h t be similar to h a l f crossing-over as defined here. (ii) Previous results o f h o m o l o g o u s r e c o m b i n a t i o n i n v o l v i n g c h r o m o s o m a l D N A can be explained by the

References Birge EA, Low KB (1974) Detection of transcribable recombination products following conjugation in rec +, recB-, and recC strains of Escherichia coli K-12. J Mol Biol 83:447-457 Brouillette S, Chartrand P (1987) Intermolecular recombination assay for mammalian cells that produces recombinants carrying both homologous and nonhomologousjunctions. Mol Cell Biol 7 : 2248-2255 Campbell A (1984) Types of recombination: common problems and common strategies. Cold Spring Harbor Syrup Quant Biol 49: 839-844 Chakrabarti S, Seidman MM (1986) Intermolecular recombination between transfected repeated sequences in mammalian cells is nonconservative. Mol Cell Biol 6:2520-2526 Clark AJ, Low KB (1988) Pathways and systems of general recombination in Escherichia coll. In: Low KB (ed) The recombination of genetic material, Academic Press, New York, pp 155-215 Cohen A, Clark AJ (1986) Synthesis of linear plasmid multimers in Escheriehia coli K12. J Bacteriol 167:327-335 Connolly B, Parsons CA, Benson FE, Dunderdale H J, Sharples G J, Lloyd RG, West SC (1991) Resolution of Holliday junction in vitro requires the Escherichia coli ruvC gene product. Proc Natl Acad Sci USA 88:6063~i067 Duckett DR, Murchie AIH, Diekmann S, Kitzinf E, Kemper B, Lilley DMJ (1988) The structure of the Holliday junction, and its resolution. Cell 55: 79-89 Horii Z, Clark AJ (1973) Genetic analysis of the reeF pathway to genetic recombination in Escherichia coli K-12: Isolation and characterization of mutants. J Mol Biol 80:327-344 Iwasaki H, Takahagi M, Shiba T, Nakata A, Shinagawa H (1991) Escherichia coli RuvC protein is an endonuclease that resolves the Holliday structure. EMBO J 10:4381M389 Jayaram M (1985) Association of reciprocal exchange with gene conversion between the repeated segments of the 2-pm circle. J Mol Biol 191:341-352 Kitamura Y, Yoshikura H, Kobayashi I (1990) Homologous recombination in a mammalian plasmid. Mol Gen Genet 222:185-191 Klein HL (1984) Lack of association between intrachromosomal gene conversion and reciprocal exchange. Nature 310:748-753 Kobayashi I (1992) Mechanisms for gene conversion and homologous recombination: The double-strand break repair model and the successive half crossing-over model. Adv Biophys 28: in press

12 Kobayashi I, Takahashi N (1988) Double-stranded gap repair of DNA by gene conversion in Escherichia coli. Genetics 119: 751-757 Kobayashi I, Murialdo H, Crasemann JM, Stahl MM, Stahl FW (1982) Orientation of cohesive end site cos determines the active orientation of chi sequence in stimulating recA-recBC-mediated recombination in phage lambda lytic infections. Proc Natl Acad Sci USA 79:5981-5985 Kobayashi I, Stahl MM, Fairfield FR, Stahl FW (1984a) Coupling with packaging explains apparent nonreciprocality of chistimulated recombination of bacteriophage lambda by recA and recBC functions. Genetics 108:773-794 Kobayashi I, Stahl MM, Stahl FW (1984b) The mechanism of the chi-cos interaction in RecA-RecBC-mediated recombination in phage lambda. Cold Spring Harbor Symp Quant Biol 49: 497-506 Kusano K, Nakayama K, Nakayama H (1989) Plasmid-mediated lethality and plasmid multimer formation in an Escherichia coli recBC sbcBC mutant. J Mol Biol 209: 623-634 Kushner SR, Nagaishi H, Templin A, Clark AJ (1971) Genetic recombination in Escherichia coli: The role of exonuclease I. Proc Natl Acad Sci USA 68:824-827 Lehman IR (1960) The deoxyribonucleases of Escherichia coli. I. Purification and properties of a phosphodiesterase. J Biol Chem 235:1479-1487 Lichten M, Fox MS (1983) Effects of nonhomology on bacteriophage lambda recombination. Genetics 103:5-22 Lichten M, Fox MS (1984) Evidence for inclusion of region of nonhomology in heteroduplex products of bacteriophage lambda recombination. Proc Natl Acad Sci USA 81:7180-7184 Lin FL, Sperle KM, Sternberg NL (1987) Extrachromosomal recombination in mammalian cells as studied with single- and double-stranded DNA substrate. Mol Cell Biol 7:129-140 Lloyd RG, Buckman C (1985) Identification and genetic analysis of sbcC mutations in commonly used recBC sbcB strains of Escherichia coli K-12. J Bacteriol 164:836-844 Lloyd RG, Picksley SM, Prescott C (1983) Inducible expression of a gene specific to the RecF pathway for recombination in Escherichia coli K12. Mol Gen Genet 190:162-167 Lloyd RG, Benson FE, Shurvinton CE (1984) Effect of ruv mutations on recombination and DNA repair in Escherichia coli K 12. Mol Gen Genet 194:303-309 Lloyd RG, Evans NP, Buckman C (1987) Formation of recombinant lacZ + DNA in conjugational crosses with a recB mutant of Escherichia coli K12 depends on recF, recJ, and recO. Mol Gen Genet 209:135-141 Lovett S, Clark AJ (1984) Genetic analysis of the recJ gene of Escherichia coli K-12. J Bacteriol 157:190-196 Lovett ST, Kolodner RD (1989) Identification and purification of a single-stranded-DNA-specific exonuclease encoded by the recJ gene of Escherichia coli. Proc Natl Acad Sci USA 86:2627-2631 Luisi-DeLuca C, Lovett ST, Kolodner RD (1989) Genetic and physical analysis of plasmid recombination in Escherichia coli K-12 mutants. Genetics 122: 269-278 Madiraju MVVS, Templin A, Clark AJ (1988) Properties of a mutant recA-encoded protein reveal a possible role Escherichia coli recF-encoded protein in genetic recombination. J Bacteriol 85: 6592-6596 Mahajan SK, Datta AR (1988) Pathways of homologous recombination in Escherichia coli. In: Kucherlapati R, Smith GR (ed) Genetic recombination. American Society for Microbiology, Washington DC, pp 87-140 Mahan MJ, Roth JR (1989) Role of recBC function in formation of chromosomal rearrangements: A two-step model for recombination. Genetics 121:433-443 Mahan MJ, Roth JR (1990) Rearrangement of the bacterial chromosome: Forbidden inversions. Science 241:1314-1318 McGill C, Shafer B, Strathern J (1989) Coconversion of flanking sequences with homothallic switching. Cell 57:459-467

Meselson MS, Radding CM (1975) A general model for genetic recombination. Proc Natl Acad Sci USA 72:358-361 Niki H, Ogura T, Hiraga S (1990) Linear multimer formation of plasmid DNA in Escherichia coli hopE (recD) mutants. Mol Gen Genet 224:1-9 Orr-Weaver TL, Szostak JW (1983) Yeast recombination: The association between double-strand gap repair and crossing over. Proc Natl Acad Sci USA 80:4417-4421 Orr-Weaver TL, Szostak JW, Rothstein RJ (1981) Yeast transformation: A model system for the study of recombination. Proc Natl Acad Sci USA 78:6354-6358 Pays E, Houard S, Pays A, Van Assel S, Dupont F, Aerts D, Huet-Duvillie G, Gomes V, Richet C, Degand P, Van Meirvenne N, Steinert M (1985) Trypanosoma brucei: The extent of conversion in antigen genes may be related to the DNA coding specificity. Cell 42:821-829 Picksley SM, Attfield PV, Lloyd RG (1984) Repair of DNA doublestrand breaks in Escherichia coli K 12 requires a functional recN product. Mol Gen Genet 195:267-274 Plasterk RHA, Simon MI, Barbour AG (1985) Transposition of structural genes to an expression sequence on a linear plasmid causes antigenic variation in the bacterium Borrelia hermsii. Nature 318: 257-263 Roman H, Fabre F (1983) Gene conversion and associated reciprocal recombination are separable events in vegetative cells of Saccharomyces cerevisiae. Proc Natl Acad Sci USA 80:6912-6916 Rossignol JL, Nicolas A, Hamza H, Langin T (1984) Origins of gene conversion reciprocal exchange in Ascobolus. Cold Spring Harbor Symp Quant Biol 49:13-21 Rudin N, Sugarman E, Haber J (1989) Genetic and physical analysis of double-strand break repair and recombination in Saccharornyces cerevisiae. Genetics 122:519-534 Sharples GJ, Benson FE, Illing GT, Lloyd RG (1990) Molecular and functional analysis of the ruv region of Escherichia coli K-12 reveals three genes involved in DNA repair and recombination. Mol Gen Genet 221:219-226 Shiba T, Iwasaki H, Nakata A, Shinagawa H (1991) The SOSinducible DNA repair proteins, RuvA and RuvB, of Escherichia coli: Functional interactions between RuvA and RuvB for ATP hydrolysis and renaturation of cruciform structure in supercoiled DNA. Proc Natl Acad Sci USA, in press Shibata T, DasGupta C, Cunningham RP, Radding CM (1979) Purified E. coli recA protein catalyzes homologous pairing of superhelical DNA and single-stranded fragments. Proc Natl Acad Sci USA 76:1638-1642 Shurvinton CE, Lloyd RG, Benson FE, Attfield PV (1984) Genetic analysis and molecular cloning of the Escherichia coli ruv gene. Mol Gen Genet 194:322-329 Siddiqi I, Stahl MM, Stahl FW (1991) Heteroduplex chain polarity in recombination of phage lambda by the Red, RecBCD, RecBC(D-) and RecF pathways. Genetics 128:7-22 Silberstein Z, Cohen A (1987) Synthesis of linear multimers of oriC and pBR322 derivatives in Escherichia coli K-12: Role of recombination and replication functions. J Bacteriol 169:3131-3137 Silberstein Z, Maor S, Berger I, Cohen A (1990) Lambda Redmediated synthesis of plasmid linear multimers in Escherichia coli K12. Mol Gen Genet 223:496-507 Sobell HM (1972) Molecular mechanism for genetic recombination. Proc Natl Acad Sci USA 68:2483-2487 Stahl FW (1969) One way to think about gene conversion. Genetics [Suppl] 61 : i-13 Stahl FW (1979) Genetic recombination. Freeman, San Francisco Stahl FW, Kobayashi I, Stahl MM (1985) In phage lambda, cos is a recombinator for Red pathway. J Mol Biol 181:199-209 Strathern JN (1988) Control and execution of homothallic switching in Saccharomyces cerevisiae. In: Kucherlapati R, Smith GR (eds) Genetic recombination. American Society for Microbiology, Washington DC, pp 445-464

13 Szostak JW, Orr-Weaver TL, Rothstein RJ, Stahl FW (1983) The double-strand break repair model for recombination. Cell 33:25-33 Takahashi N, Kobayashi I (1990) Evidence for the double-strand break repair model of bacteriophage lambda recombination. Proc Natl Acad Sci USA 89:2790-2794 Takahashi NK, Yamamoto K, Kitamura Y, Luo S, Yoshikura It, Kobayashi I (1992) Nonconservative recombination in Escherichia coli. Proc Natt Acad Sci USA 89 (in press) Taylor AF, Smith GR (1985) Substrate specificity of the DNA unwinding activity of the RecBC enzyme of Escherichia coli. J Mol Biol 185:431~33 Taylor AF, Schultz DW, Ponticelli AS, Smith GR (1985) R e c B C enzyme nicking at chi sites during DNA unwinding: Location and orientation dependence of the cutting. Cell 41 : 153-163 Thaler DS, Stahl MM, Stahl FW (1987a) Double-chain-cut-sites are recombination hot spots in the Red pathway of phage lambda. J Mol Biol 195:75-87 Thaler DS, Stahl MM, Stahl FW (1987b) Evidence that the normal route of replication-allowed red-mediated recombination involves double-chain ends. EMBO J 6:3171-3176

Thaler DS, Sampson E, Siddiqi I, Rosenberg SM, Thomason LC, Stahl FW, Stahl MM (1989) Recombination of bacteriophage lambda in recD mutants of Escherichia coli. Genome 31:53-67 Umezu K, Nakayama K, Nakayama H (1990) Escherichia coli RecQ protein is a DNA helicase. Proc Natl Acad Sci USA 87: 5363-5367 White CI, Haber JE (1990) Intermediates of recombination during mating type switching in Saccharomyces cerevisiae. EMBO J 9 : 663-679 Willetts NS, Clark AJ (1969) Characteristics of some multiply recombination-deficient strains of Escherichia coll. J Bacteriol 100:231-239 Yamamoto K, Yoshikura H, Takahashi N, Kobayashi I (1988a) Apparent gene conversion in an Escherichia coli rec ÷ strain is explained by multiple rounds of reciprocal crossing-over. Mol Gen Genet 212:393-404 Yamamoto K, Takahashi N, Yoshikura H, Kobayashi I (1988b) Homologous recombination involving a large heterology in Escherichia coli. Genetics 119:759-769 C o m m u n i c a t e d b y M. Sekiguchi