Homologous recombination involving a large heterology in ...

3 downloads 0 Views 4MB Size Report
and LABAN 1983; FISHEL, JAMES and KOLODNER 198 1 ;. JAMES, MORRISON and KOLODNER 1982; DOHERTY,. MORRISON and KOLODNER 1983; ...
Copyright 0 1988 by the Genetics Societyof America

Homologous Recombination Involving a Large Heterology in Escherichia coli Kenji Yamamoto,* Noriko Takahashi,t Hiroshi Yoshikura* and Ichizo Kobayashi*’t *Department of Bacteriology, Faculty of Medicine, University of Tokyo, Hongo, Tokyo 113, Japan, and ?National Children’s Medical Research Center, Taishido, Tokyo 154, Japan Manuscript receivedJanuary 15, 1988 Accepted April 14, 1988

ABSTRACT Recombination between twodifferent deletion alleles of a gene (neo) for neomycin and kanamycin resistance was studied in an Escherichia coli sbcA- recB-C- strain. The two homologous regions were in an inverted orientation on the same plasmid molecule. Kanamycin-resistant plasmids were selected and analyzed. The rate of recombination to form kanamycin-resistant plasmids was decreased by mutations in the recE, recF and recJ genes, but was not decreased by a mutation in the recA gene. It was found that these plasmids often possessed one wild-type kanamycin-resistant allele(neo+) while the other neo allele was still in its original (deletion) form. Among kanamycin-resistant plasmidswith one wild-type and one parental allele it was often found that the region between the inverted repeats had been flipped (turned around) with respect to sites outside the inverted repeats. These results were interpreted as follows. Gene conversion, analogous togene conversion in eukaryotic meiosis, is responsible for a unidirectional transfer of information from one neo deletion allele to the other. The flipping of the region between the inverted repeats is interpreted as analogous to the crossing over associated with gene conversion in eukaryotic meiosis. In contrast with a rec+ strain, these products cannot beexplained bytwo rounds of reciprocalcrossing over involving a dimeric form as an intermediate. In the accompanying paper we present evidence that gene conversion by double-strand gap repair takes place in the same%. coli strain.

G

ENE conversion, that is, the nonreciprocal transfer of a stretch of sequence information from oneDNA molecule toanotherhomologousDNA molecule,hasbeenobservedinvariousorganisms. Since its discovery, gene conversion has been extensively studied in the meiosis of the fungus Ascomycetes, including Saccharomyces, becauseall the products of meiosis can be recovered and analyzed as a tetrad (for review, seeWHITEHOUSE 1982; ROSSIGNOL et al. 1984). Frequent association of gene conversion and crossing over of the flanking sequences (Figure 1)(KITANI, OLIVEand EL-ANI 1962) and its other features led to the view that gene conversion reflects events at the site of interaction during homologous of analysis, “tetrad recombination. In fact, this type analysis,” of gene conversion in the fungalmeiosis has beenthemajorsource of thecurrentmodelsfor homologous recombination (STAHL1979a; SZOSTAK et al. 1983).IntheHOLLIDAY(1964)model,gene conversion results from mismatch repair on heteroduplex DNA. In MESELSON-RADDING (1975)model (Aviemore model), asymmetric DNA synthesis leading to asymmetric heteroduplex is postulated to be another source of gene conversion. T h e double-strand breakrepairmodel (SZOSTAKet al. 1983)proposes repair of double-stranded gapby copying ofa homologous sequence. Genetics 119: 759-769 (August, 1988)

On the other hand, the molecular components involved in each of the steps of homologous recombination have been more thoroughly characterized in the prokaryotes, such asEscherichia coli and its bacteriophage lambda, than in the eukaryotes. These include recombination proteins such as recA gene product(CLARK a n d MARGULIES1965; KOBAYASHI and IKEDA1978,1983; SHIBATAet al. 1979),RecBCD et enzyme (exonuclease V) (OISHI1969; AMUNDSEN al. 1986), red exonuclease of lambda (LITTLE 1967) and homologous recE exonuclease (exonuclease VIII) (JOSEPH and KOLODNER 1983a) expressedby the sbcA mutations (KUSHNER,NAGAISHIa n d CLARK 1974). T w o special sites for homologous recombination are for the well characterizedinmolecularterms:chi, RecBCD pathway (STAHL1979b; SMITHet al. 1981; KOBAYASHI, STAHLand STAHL 1984; PONTICELLI et al. 1985) and lambda cos for the Red and the RecE pathways (STAHL, KOBAYASHIand STAHL 1985; STAHL1986). Asystem to analyze gene conversionin E. coli, similar to the tetrad analysis system in fungi, would help us to integrate these analyses of molecular components into one global picture. Toward the development of such a system we have constructed a plasmid that allows us to study processes that may be homologous to gene conversion and crossing over in meiosis.

760

K. Yamamoto et al.

M-

a-b+

known mismatch repair systems cannot repair such a large heterology. In the accompanying paper (KOBAYASHI and TAKAHASHI 1988) we present evidence that gene conversion by double-strand gap repair takes place in the E. coli RecE pathway.

N+

X M+

ab''

N-

MATERIALS AND METHODS

j ' + + I

II

-

+

"

-

reciprocal crossing-over

a

I

k+ ++-

-

++

+

geneconversion a t withoutcrossing-over o f f l a n k i n g sequences

l+ -

+

+-

geneconversion

-

at

a

w i t h crossing-over o ff l a n k i n g

sequences

FIGURE 1.-Gene conversion and crossing over in eukaryotes. T h e symbols a and b are close markers. M and N are markers distant from a and b.

We examined the RecE pathwayof E. coli, which is similar to the Red pathway of bacteriophage lambda, since analysis of recombination enhancement at lambda cos by t h e Red pathway has led t o a picture similar to thedouble-strandbreak repair model (STAHL,KOBAYASHIa n d STAHL 1982, 1985). We used a plasmid with regions of inverted homology such that, if recombination was strictly intramolecular, we would recover both products of the reaction. It was found that the selected recombinantplasmids often possessed one recombinant allele and one parental allele. We interpret this as gene conversion, a unidirectional intramolecular transfer of information from one of the repeats to the other. Among these recombinants itwas often found that the region between the two homologous regions was inverted with respect to outside markers. This is exactly the result that would be expected if recombination were intramolecular and frequently accompaniedby crossing over of flanking markers as in eukaryotic meiosis. If all recombination was intramolecular, then the selected recombination must have been a gene conversion event against a large (283 bp) deletion. T h e

Bacterial strains: The isogenic strains listed in Table 1, JC8679, JC8691, JC9610,and JC9604 (GILLEN, WILLISand CLARK 1981) and JC12123, which has recJ284::TnlO (LovETT andCLARK1984),are from A. J. CLARK. The recJ284::TnlO mutation was transferred by P1 transduction to JC8679 to make BIK814 (Table 1). AB1157 (BACHMAN 1987) is from A. MIURA and H. IKEDA.The recA1 strain, DH 1 (HANAHAN 1985) is from B. HOHN.DH5, a derivative of DH1 (HANAHAN 1985), is from Y. KITAMURA. Plasmidconstruction: Construction of pIK43 is described elsewhere (YAMAMOTOet al. 1988). Its repeated sequence comes from pSV2neo (SOUTHERN and BERG1982), which in turn is derived from Tn5. The left part is from pML2d (LUSKYand BOTCHAN1981),a derivative of pBR322. The right part comes from SV40 origin and from bovine papilloma virus type 1 (CHENet al. 1982). Transformation: The plasmid DNA was introduced into the cells by the standard calcium method (YAMAMOTOet al. 1988). A tube of the competent cells reached saturation level with approximately 100 ng/tube of DNA (pBR322). The experiments were carried out at input of 1-1 0 ng/ tube, well below the saturation level. The competent cells inCaC12 and glycerol were stored in liquid Nz. The transformants were selected with 50 rg/ml ampicillin or 50 rg/ mlof kanamycin.With 1 ng of pIK43, 10' ampicillinresistant colonies were obtained from JC8679. The competent cells of DHl produced l O3 ampicillin-resistant colonies from 1 ng of pIK43. Detection of recombinant clones:Each of the fresh ampicillin-resistantcolonies obtained by transformation of JC8679 with pIK43 was suspended wellin buffer and streaked on kanamycin (50 rg/ml) agar plate for incubation at 37" for 24 hr. Plasmid preparation:In order to avoid accumulation of the rare neo+ recombinants, cultures were prepared from recently isolated colonies. The procedures for small-scale analysis and for large-scale preparation are described in a 1988) andin companion study (KOBAYASHIand TAKAHASHI a previous work (YAMAMOTOet al. 1988). Restriction enzyme analysis:The plasmid DNA was cut with excess ofNaeI (New England Biolab,Toyobo), HindIII, XhoI, EcoRI, EcoRV, or BglIl (Takara). The neo+ monomer plasmids that can be derived from pIK43 by various types of homologous recombination between the repeats were formulated in a computer program (K. YAMAMOTOand I. KOBAYASHI,unpublished data). Their restriction enzyme cleavage patterns were determined (YAMAMOTO et al. 1988). Agarose gel electrophoresis: The intact plasmids were electrophoresed through0.4% agarose gelin a low Na buffer (40 mM Tris-acetate, 2 mM Na2EDTA (pH 8.0)) with 0.5 rg/ml ethidium bromide at 70 V/35 cm for 16-36 hr. The DNA samples were mixed with a dye mixture containing 10 Pg/ml of ethidium bromide just before loading. The restriction enzyme digests were electrophoresed through 0.7% to 1% agarose gel in a high Na buffer (40 mM Na-acetate, 2 mM NazEDTA (pH 8.0)) with or without

76 1

Homologous Recombinationin E. coli TABLE 1 Recombination frequencyin JC8679and related strains

Strain

JC8679 JC869 1 JC9610 BIK814 JC9604 AB1 157 DH 1

per

Recombination frequency cell generation

frequency Recombinant rec genotype

recC22 recB21 sbcA23 sbcA23 recB21 recC22 recEI59 sbcA23 recB21 recC22 recF143 sbcA23 recB21 recC22 recJ284::TnlO sbcA23 recB21 recC22 recA56 rec+ recAl

0.98 f 0.25 X 1.01 0.18 X 1.20 0.29 X 0.90 k 0.39 X 0.89 f 0.20 X

lo-* 10-~ 10-~ lo-' lo-*

1.10 0.33 X 1 0 - ~ 0.70 f 0.35 X

5.43 f 1.43 X 1 0 - ~ 6.76 f 1.15 X 8.07 f 1.83 X 4.93 f 2.17 X 1 0 - ~ 5.63 f 1.32 X loT4 6.92 k 2.48 4.28 f 2.25

X

X

lo-'

Each of the ampicillin-resistant clones from pIK43 was grown with ampicillin selection to saturation in 5 ml. The culture was plated on kanamycin agar and on ampicillin agar. The recombinant frequency ( p ) is [number of kanamycin-resistant colony former]/[number of ampicillin-resistant colony former]. The recombination frequency per cell generation ( a ) was calculated by p = (%)g X a , in which g is the generation number estimated from the number of the ampicillin-resistant colony formers. The top six strains are nearly isogenic. The associated numbers are the standard deviations among 10 clones examined.

ethidium bromide at 100 minigel apparatus.

V/12cm for a short time in a

Transfer of type 1 dimer to cell: The type 1 dimer formed from pIK43 insidea ret+ strain, AB1 157,was transferred into a w c A - strain, DH1,and propagated. Plasmid D N A was prepared after minimal cell generations. The closed circle dimer fraction was purified through agarose gel electrophoresisand used to transform JC8679 or AB1 157.T h e transformant colony on kanamycin agar was grown to saturation in 5 ml of broth with kanamycin for plasmid D N A preparation.

RESULTS Design: Detection of gene conversion requires recovery of both products resulting from interactionof two DNA duplexes and sensitive selection for rare conversion products. Figure 2 shows our starting substrate plasmid for this purpose. This plasmid carries two homologous sequences inan inverted orientation s o that both the products can be recovered on one molecule if recombination is intramolecular. T h e repeated sequence contains thene0 gene, which confers kanamycin resistance to the host bacteria. T h e upper copy has a deletion ('u') removing one end of the neo gene, and thelower copy has a deletion ('b') removing the other end of the neo gene. Between the two copies, there is acontinuous homology to the left of the deletion a, between the two deletions and to the right of the deletion b. Various types of intramolecular homologous recombination events between these two segments would reconstruct a functional neo+ gene and make the host kanamycin resistant. One type (type 1) is reciprocal crossing over between the two deletionsillustrated in Figure 3 andFigure8a below. Others include the types expected by gene conversion at deletion a (types 5 and 6 in Figure 3 and Figure 8a). Figure 3 lists all the possible monomer neo+ forms predictedfromintramolecularrecombination. Homologous recombination events between the upper neo gene of one molecule and the lower ne0 gene of another molecule can produce dimeric neo+ plasmids.

FIGURE2,"Parental plasmid (pIK43). The two homologous duplex segments are drawn as parallel lines. The top segment has a 283-bp long deletion (deletion a ) between the two NaeI sites, which removed one end of the neo gene. The NaeI site has been inactivated by insertion of the XhoI linker sequence, 5' C-C-T-C-GA-G-G. The bottom sequence has a 248-bp long deletion (deletion b), which removed the other endof the neo gene. The two deletions are separated by 506-bp long homology and are flanked by 1079bp long homology to their left and 231-bp long homology to their right. The left, nonhomologous part (2321-bp long), including the ampicillin-resistancegene and thereplication origin, is derived from pBR322. The right, nonhomologous part (8297-bp long) is from SV40 and BPV-1. The entire length of pIK43 is 14795 bp. The restriction enzymes and their site coordinates are: X: XhoI, 1082; B: BglII, 1750, 4215, 4241, 4911, 7426; RI: EcoRI, 8024, 14793; RV: EcoRV, 10367; N: NaeI, 11 109, 11392, 12590, 12750.

Each of these forms can be distinguished from the others by cleavage with suitable restriction enzymes (YAMAMOTOet ul. 1988). We introduced this plasmid intoan E. coli K12 strain JC8679 (GILLEN,WILLISand CLARK 1981)by calcium transformation followed by ampicillin selection. This strain carries sbcA23 mutation, which activates the RecE system of recombination, and recB21 and recC22 mutations, which inactivate RecBCD enzyme (exonuclease V). Detection of gene conversion type products: JC8679 carrying this plasmid produced kanamycinresistant cells at a low frequency as shown in Table 1. This recombinationfrequency was decreased by a

K. Yamamoto et al.

762

V.A

: delelion.

1

2

0

2

3

I

9

n

n

n

n

n

I

0

0

0

0

0

0

0

0

FIGURE 3.-Classification of the neo* plasmids formed. The left columns show all the possible neo+ plasmid types formed by various kinds of homologous recombination between the two neo segments in the same plasmid. The plasmid DNA preparation from each of the kanamycin-resistant clones derived from JC8679[plK43] was analyzed with XhoI, Nael, Rglll, and EcoRI, and classified unambiguously into one of the 12 predicted forms.

V. A : non.#mletion.

factor of 10 by a recE- mutation, a r e c F mutation, and a r e c s mutation. It was not decreased by a recAmutation in this sbcA- recB- recC- background. In a nonisogenic recA I strain, DH 1, the recombination frequency was much lower (Table 1). When we selected for KanR colonies immediately after transformation with pIK43,theratio of the kanamycin-resistant colonies to the ampicillin-resistant colonies was less than This demonstrates that the recombinant neo+ plasmids were produced during growth on agar and in liquid and were not present in the starting plasmid preparation nor produced in the process of transformation. Each of these kanamycin-resistant (Neo+) colonies from JC8679 [pIK43] was propagated and plasmid DNAs were prepared. As shown in Figure 4, most of the plasmid DNA molecules retained the original size inmost of the clones. T h e clones also contained variable but usually small amounts of dimeric circles. We cut these plasmid DNA preparations with restriction enzymes, XhoI, EcoRI, NueI and BglII, and were able to assign one of the predicted forms to each of the clones. We did not detect any heterogeneity within one clone (see also Table 3 below). We did not detect, by gel electrophoresis, any interconversion among these forms in JC8679 even after subculture. As summarized in Figure 3, the majority of the neo+ plasmids turned out to be the type expected from gene conversion at deletion a, although we found some plasmids of thereciprocal crossing-over type. As exemplified by clones 1 through 4 and 7 through 12 in Figure 5 , their two ne0 segments both have the

FIGURE4.-Forms of the recovered plasmid DNA. The plasmid preparations from the following sources were electrophoresed. Amp', an ampicillin-resistant clone made by transformation of JC8679 with the parental plasmid, pIK43; Kan', a kanamycinresistant (Neo+)clone, isolated from the primary ampicillin-resistant transformant colony; marker, mixture of pIK43 monomer and pIK43 dimer (type 1 dimer): ori, origin of electrophoresis; 2 mer, dimer: 1 mer, monomer: o.c., open circle; c.c., closed circle.

NueI sites. This pattern is expected from intramolecular gene conversion from the lower ne0 segment to the upper ne0 segment of the parental plasmid (see Figure 2). We also detectedthe type expectedfromgene conversion at site b by appearance of a BglII site (data

in E. coli

Recombination Homologous

763

Eco R I 1 2

3

4

5

6

7

8

9101112

‘torigin

3kbp

41.2kbp

FIGURE5.-Apparent gene conversion revealed by A’ael restriction analysis. The plasmid preparations from the kanamycin-resistant clones from JC8679[pIK43] were cut with Nael and electrophoresed. The plasmids from the clones I , 2, 3, 4, 7, 8, 9. 10, 1 1 and 12 had obtained new Nael sites by gene conversion: marker, lambda D N A cut with Hindlll: parent, pIK43: kbp, kilobase pair.

not shown). We do not know why the events at b are a (Figure 3). One less frequentthantheeventsat possible cause might be the difference in the length of the flanking homology. Association of inversion: Among the gene conversion type products, it was found that the region between the two homologousregions (left “arm” in Figure 2) was inverted with respect to outside markers (right “arm” in Figure 2). The cleavage with EcoRI is diagnostic for this inversion: a pair of the fragments of 8 kbp and 7 kbp would be changed to a pair of fragments of approximately 10 kbp and 5 kbp (Figure 2). The plasmids 1, 2, 9,and 11 in Figure 6 had experienced such inversion. This is exactly the result that would beexpected if this recombination was intramoleculargene conversion and frequently accompanied by crossing over of flanking markers asin eukaryotic meiosis. This association of the inversion and the apparent gene conversion is not accidental. That each kanamycin-resistant clone had only one type of plasmid shows that the inversion does not occur frequently once the plasmid is established inside the cell. In a control experiment, we prepared the type 5 plasmid (apparent gene conversion at site a without crossing over) and the type 6 plasmid (apparent gene conversion at site a with crossing over) in a recA strain, which I), and shows quitereducedrecombination(Table introduced each back into the JC8679cells. As shown in Figure 7, we detected no sign of interconversion between the non-crossing-over type and the crossingover type. T h e starting plasmid, pIK43, in JC8679 didnot show the inversion asdetected by the gel electrophoresis during growth in the presence of ampicillin. We therefore concluded that this association

I;IC;URE 6.--hpparent association of crossing o v e r with the putative gene conversion. The plasmid preparations from 12 kanamycin-resistant clones from JC8679[pIK43] were cut with EcoRl and electrophoresed. The plasmids from the clones 1. 2, 5 . 6, 9 and 1 1 have experienced crossing over of the flanking sequences. Among these, the clones 1, 2, 9 and 1 1 have experienced apparent gene conversion as revealed by Nael analysis (see Figure 5 ) . kbp, kilobase pair.

FIGURE7.“Stability of notl-crossitlg-ovt.1.-1yl)e and crossingover-type plasmids. The type 5 or type 6 plasmid was propagated in a recA- strain, DH5, and transferred back into JC8679. The plasmid preparations from the resulting kanamycin-resistant transformant clones were cut with EcoRl and electrophoresed. In no clone, interconversion between the crossing-over-typeplasmid and the non-crossing-overtype was detected. Marker, lambda DNA cut with Hindlll; parent, pIK43; type 5 , type 5 plasmid in Figure 3. Gene conversion at site a without crossing over; 5-1 through 5-6, plasmids from six JC8679 clones that had received the t y p e 5 plasmid; type 6, type 6 plasmid in Figure 3. Gene conversion at site a with crossing over; 6-1 through 6-6, plasmids from six JC8679 clones that had received the type 6 plasmid.

of the inversion is characteristic of the apparent gene conversion process. The recA mutation did not decrease the frequency of recombination generating the Neo+ cells as mentioned above. T h e two types of apparent gene conversion products were both detected in a sbcA- recBr e d - recA- strain, JC9604: 6 clones of the inversion

K. Yamamoto et al.

Parent crossing-over

gene conversion a t a withoutcrossing-over

im-+C=3 gene conversion a t b

withoutcrossing-over

( b ) rec‘

CZ5-J

~

~

~

A

A

parent

3 t e m ~reciprocal i c u b + crossing-over

A

A

I

i

type 1d i n e r

type and 13 clones of the noninversion type were found among the 19 clones examined. Evidenceagainst one nonconversionroute: Were these neo+ plasmids produced by gene conversion between the two ne0 segments of one plasmid molecule asillustrated in Figure8a? Or, alternatively, could they be products of intermolecular and/or multiple rounds O f recombination? We have evidence against one particular nonconversion route, which was demonstrated in a ret+ strain, AB1 157, with the same

A

FIGURE8.-Possible origins of the gene conversion type products. (a) JC8679, a sbcA23 recB21 recC22 strain. 6 Events of homologous recombination between the two homologous segments within one plasmid molecule are assumed to be responsible for formation of the monomer neo+ plasmids. These events include gene conversion event at deletion a with or without crossing over of the flanking sequences, gene conversion event at deletion b, and a reciprocal crossing over at some site between deletion a and deletion 6. (b) AB1 157, a rec’ strain. A similar reciprocal crossing over at some site between deletion a and deletion b, but involving two plasmid molecules, produces the “type 1” dimer. This type 1 dimer then produces, by a second, 1 intramolecular crossing over between thedirect repeats, those various neo‘ recombinant monomers including those apparent gene conversion products. An alternative origin of the type 1 dimer is resolution of intramolecular HOLLIDAY 5 structure by replication as discussed previously (YAMAMOTO et al. 1988).

A

plasmid (YAMAMOTO et al. 1988). We will discuss later the other non-gene-conversion routes. The nonconversion route is illustrated in Fig,ure 8b. The bottom segment of one plasmid molecule and the top segment of a second molecule undergo reciprocal crossing over. The resulting neo+ dimer, which we call the ,, dimer, makes the cell kanamycin resistant. (For another possible origin of this type 1 dimer, see Figure 8b legend.) This ‘‘type 1 dimer”frequently produces various nee+ monomers,includingthose gene conversion types, by asecond,intramolecular

Homologous Recombination in E. coli

765

TABLE 2 Analysis of dimer plasmidsby their transferto recA- strain In JC8679

Type

Clone D (type 1 monomer).

Clone F (type 6 monomer).

Clone G (type 9 monomer).

1 X 2 (tandem dimer of type 1 monomer) 5 X 2 (tandem dimer of type 5 monomer) 6 X 2 (tandem dimer of type 6 monomer) 9 X 2 (tandem dimer of type 9 monomer) Type 1 dimer (Figure 7) Others

4 0 0 0 0 0

0

0

0

0

0 0

0 0 4 0 0

0 0 0 63 0

4

6

4

63

Total dimer subclones examined

6 0

In AB1 157

Plasmid DNA prepared from the kanamycin-resistant clones from JC8679[pIK43] or AB1 157[pIK43] was electrophoresed through agarose gel as described for Figure 4. The few dimer closed circles were recovered and used to transform a recA- strain, D H I , to kanamycin resistance. Plasmid DNA from each of these transformants were classified by restriction enzyme analysis. a Type of the monomer plasmids in the primary kanamycin-resistant clone.

crossing over using homology between the direct repeats. T h e type 1 dimer is rapidly replaced by these monomers within one clone. We have three lines of evidence against this route in our sbcA- recB- recC- strain, JC8679. The first is the absence of the type 1 dimer. With the ret+ strain, AB1 157, all the dimeric species examined in the kanamycin-resistant clones were the type 1 dimer. With JC8679, we detected variable but usually small amounts of dimeric forms in the kanamycin-resistant clones (Figure 4). We chose clones which have type 1 , 5, 6 or 9 monomer, transferred their dimer fractions into a recombination-defective recA2 strain, DH 1 , and analyzed the plasmid in several of these subclones. As shown in Table 2, we found only “tandem” (head-to-tail) dimer of the monomer present in the primary clone (type 1 , 5, 6 or 9 monomer) but not the type 1 dimer in all the subclones examined. The second line of evidence against this route is homogeneity of the plasmid form as mentioned above. In AB1 157 (Tee+), each of the recombinantclones had mixed types of the monomer plasmids. This heterogeneity cannot be explained by the difference in the recombination frequency because AB1 157 shows a lower recombination frequency than JC8679 (Table 1). These mixed types are presumably the sisters derived from the type l dimer which had founded the kanamycin-resistant clone. In contrast, with JC8679, inall the recombinant clones examined, no signof heterogeneity in the plasmid type was detected by gel electrophoresis (Figures 5 and 6). We confirmed this homogeneity by analyzing the individual plasmids of the representative clones by transfer tothe recAstrain, DH 1 . As shown in the centralcolumns of Table 3, no sign of heterogeneity was detected by this procedure. T h e third line of evidence against this nonconver-

sion route is the stability of the type 1 dimer in JC8679, which was revealed in the following reconstruction experiments. We prepared the type 1 dimer molecules from a recA- strain and transferred them to AB1 157 (Tee+) and JC8679. As shown in Figure 9, the type 1 dimer was rapidly converted to the monomers in the AB1 157 cells. In JC8679, however, the vast majority of the type 1 dimer molecules stayed as a dimer in the time scale we employed. We detected appearance of only a very small amount (approximately 1/100) of monomer plasmids by gel electrophoresis. We analyzed these few monomers by transfer to a recA- strain. As shown in Table 3 (right columns), the monomer plasmids were heterogeneous inall the clones examined as in the recombination experiments with AB1 157. T h e majority in each of these clones was the reciprocal crossing-over type. The conversion type plasmids were rare. This is in contrast with the above recombination experiments with JC8679, where mostof the clones had the conversion type plasmid (Figure 3) as the great majority [Figures 5 and 6 and Table 3 (central columns)]. DISCUSSION

Observations and interpretations:Most of the selected recombinant plasmids carried one recombinant allele and one parentalallele. Among these the region between the repeats was often inverted with respect to sites outside the repeats. We interpreted these observations as follows. Gene conversion, homologous to gene conversion in eukaryotic meiosis, is responsible foraunidirectionalintramoleculartransfer of informationfrom one parental allele totheother allele. We interpretedthe inversion of the region between the repeats as homologous to the crossing over associated with gene conversion in eukaryotic meiosis.

766

K. Yamamoto et al. TABLE 3 Analysis of monomer plasmids by their transfer to recA- strain From parent (pIK45)

Type

Clone

Clone A (type 1 monomer). monomer).

1 (reciprocal crossing over) 5 (gene conversion at site a without crossing over) 6 (gene conversion at site a with crossing over) 9 (gene conversion at site b without crossing over) Others 23

Total subclones 30 examined

19

Clone B (type 5 monomer).

From type I dimer Clone C (type 6 1

Clone 2

Clone 5

23 0

0 19

0 0

17 1

16 1

18

0

0

30

5

4

6

0

0

0

0

0

0

0

0

0

0

0

0

21

24

23

0

From parent (pIK43): Plasmid DNA prepared from the kanamycin resistant clones from JC8679[pIK43] as described for Figure 4 was used to transform a recA- strain, DHl (clone C ) or DH5 (clone A, B), to kanamycin resistance. Plasmid DNA from each of these transformants was classified by restriction enzyme analysis. From rype J dimer: Plasmid DNA from JC8679[type 1 dimer] was prepared and electrophoresed through agarose gel as described for Figure 9. The few monomer closed circles were recovered and used to transform a recA- strain, DH 1, to kanamycin resistance. Plasmid DNA from each of these transformants was classified by restriction enzyme analysis. a T y p e of the monomer plasmids in the primary kanamycin-resistant clone.

-origin

4

2mer oc

-

4lrneroc 2mercc

lrnercc

FIGURE9.-Fate of the type 1 dimer in rec+ and sbcA- r e d r e c C strains. The closed circle dimers of type 1 dimer (Figure 8) were purified through agarose gel electrophoresis and used to transform sbcA- re&-C- strain, JC8679, and ret+ strain, AB1 157, to kanamycin resistance. The plasmid preparations from these transformants were electrophoresed. Parent, pIK43; type 1 dimer, see Figure 7b. Prepared from the recA- strain, DH 1; ret+, plasmids from AB1 157 (ret+) that had received type 1 dimer;sbcA- re&-C1, plasmids from one clone of JC8679 that had received the type 1 dimer; sbcA- recB-C--2, plasmids from another clone of JC8679 that had received the type 1 dimer.

Comparison with tetrad analysis: Before we assess validity of this interpretation, let us compare the present plasmid system with the tetrad analysis system of the meiosis of Ascomycetes. We admit that our design is inferior to tetrad analysis in three aspects: (1) Inter-

the scored molecular reaction. It is notcertainthat recombination is intramolecular. Therefore it is unclear whether both partners in a recombination reaction will be recovered on the same molecule. Intermolecular recombination between heteroalleles on plasmids was shown in the strain used in our work and its close relatives (JAM=, MORRISONand KOLODNER 1982; LABANand COHEN1981). (2) Multiple rounds of recombination intervened by replication. In tetrad analysis, the meiotic recombination events are limited to one particular stage with respect to DNA replication. In our plasmid system a second recombination can take place after many rounds of replication. This uncertainty in the system is analogous to problems in the study of mitotic rather than meiotic recombination in fungi. (3) Two-strand vs. four-strand analysis. Recombination between two DNA duplexes involves four polynucleotide chains. In tetrad analysis all four are recovered. Even if all recombination is intramolecular in our system, only two of the original chains are recovered. Alternative interpretations: Now we consider explanations alternative to our interpretation as intramolecular gene conversion. T h e gene conversion types without crossing over (type 5 and type 9 in Figure 3) can be explained by one event of intermolecular, rather than intramolecular, recombination. T h e recombination could be physically nonreciprocal or physically reciprocal. In the latter case it can be genetically reciprocal (reciprocal patches, for example) or genetically nonreciprocal (gene conversion). The gene conversion type with crossing over (type 6 in Figure 3j, however, cannot be explained by one

Homologous Recombination in E. coli

event of intermolecular recombination. It would be explained by two rounds of interaction, both resulting in crossing over of the flanking sequences. This route couldbe classified into two groupsdependingon whether the first exchange is physically reciprocal or physically nonreciprocal. In the first category the product of the first exchange is a circle and can replicate before a second exchange. The most likely route of this kind (first exchange physically reciprocal) is the one illustrated in Figure 8b.Thisrouteproduces,afterthe first reciprocal crossing over, a neo+ dimeric plasmid (the type 1 dimer), which would make the host kanamycin resistant. We have shown that in a rec+ strain, AB1 157, this is the origin of the apparent gene conversion products (YAMAMOTOet al. 1988). We have excluded, however, this nonconversion route in this work with an sbcA- recB- C- strain. We infer that the otherroutes of this category (two exchanges,first exchange physically reciprocal) are unlikely since the most likely one was excluded. The second category of two-exchange route is the one in which the first exchange is physically nonreciprocal and, therefore, produces a linear molecule. Such linear forms cannot replicate before second exchange in contrast to the intermediates(type 1 dimer, for example) in the first category. These linear forms might be more recombigenic than the circles, though (SYMINGTON, MORRISON and KOLODNER1985; KOBAYASHI and TAKAHASHI 1988).Longlinearforms were detected in recB- C- strains (COHENand CLARK 1986; SILVERSTEIN and COHEN1987). We feel this category unlikely but cannot excludeit at the present stage. In these two categories we assumed that the recombining plasmids are circles. If we assume thatthe recombining plasmid is linear, the product of the first exchange will be linear, as in the above category two, whether the exchange is physically reciprocal or not. T h e primary ampicillin-resistant clones contained a small amount of the head-to-tail dimer of the parental plasmid. This type of dimer, however, cannot produce the neo+ monomers by a second single event of recombination. We will not discuss the possibility that three events of recombination lead to the neo+ monomers in this case and in general for the sake of economy. Possible mechanisms: If the recombination we observed is, in fact, gene conversion against a large (283 bp) deletion, what can be the underlying mechanism? The frequency of producing the Nee+ cells was decreased by a recE mutation, a recF mutation, and a recJ mutation by a factor of ten. Since the majority of the neo+ recombinant plasmids belonged to the gene conversion types, we conclude that the recE+, recF+, and r e c y functions wereinvolved in the putative gene conversion events. T h e frequency of generation of

767

thegene conversion type plasmids, either without apparent crossing over or with apparent crossing over, was independent of recA+ function. These recE-, recFand recJ mutations as well as the recA- mutation decreaserecombination in bacterial conjugation in sbcA- recB- r e d - background (GILLEN, WILLISand CLARK1981; LOVETTandCLARK1984). Recombination of replicating lambdain sbcA- recB- r e d - background is decreased by a recE- mutation butnot decreased by a recA- or r e c F mutation (GILLENand CLARK1974). T h e effects of the rec mutations on plasmid homologous recombination in sbcA- recBrecC- background are somewhat variable among the previous reports (LABAN and COHEN1981; COHEN and LABAN 1983; FISHEL, JAMES and KOLODNER1981; JAMES,MORRISON and KOLODNER 1982; DOHERTY, MORRISONand KOLODNER 1983; SYMINGTON, MORRISON and KOLODNER1986). T h e recE mutation has a negative effect inall the cases. The recA and recF mutations have negative effects in some systems but not in other systems. One possible route to these gene conversion type plasmids is heteroduplex formation between thedeletion and the nondeletion. Heteroduplex between a large (700 bp) deletionand nondeletion was detected for the Red pathway of lambda, which is similar to the RecE pathway operating in JC8679 (LICHTEN and FOX 1984). Replication might segregate such a heteroduplex and produce a monomer of the gene conversion type. Another consequence from these heteroduplexes, that is, mismatch repair leading to gene conversion type plasmids, is unlikely in our case. T h e known mismatch repair systemsin E. coli (WILDENBERG and MESELSON1975; LIEB 1983) cannot act on heteroduplexes involving large (90 bp, 700 bp and 800 bp) heterology (KRAMER,SCHUGHART and FRITZ 1982; WAGNER et al. 1984; DOHETet al. 1987; RAPOSA and FOX 1987). In transfection of artificial heteroduplexes including a huge heterology (TnlO), theparity of repair was such that the extra DNA was removed (RAPOSA and FOX 1987). This parity is opposite to the parity observed in this work. A contrasting category of mechanism is replicative transfer of duplex information (STAHL1969; BOON and ZINDER 1969; SZOSTAK et al. 1983). For our gene conversion, we favor this category, especially, the double-strandbreakrepair model (SZOSTAKet al. 1983; STAHL,KOBAYASHIand STAHL1985; THALER, STAHLand STAHL 1987b) forseveral reasons. Repair of DNA double-stranded breaks is known in E. coli (TOMIZAWA and OGAWA 1967, 1968). For lambda's Red system, which is similar tothe RecE system, analyses of stimulation of recombination by cos suggested that a double-strand break initiates homologous exchanges(STAHL,KOBAYASHIand STAHL1982, 1985; THALER, STAHLand STAHL 1987a, b). In this

768

K. Yamamoto et al.

scheme the role of the recE protein will be to create an invasive single-stranded tail by its 5’ to 3’ exonuclease action from the double-stranded end UOSEPH and KOLODNER1983b) as proposed for lambda’s red exonuclease (STAHL,KOBAYASHIand STAHL1985). In a companion study, we directly demonstrated repair of a double-stranded gap by a homologous sequence through gene conversion with the same plasmid and in the same bacterial strain (KOBAYASHIand TAKAHASHI 1988). Related work There is an observation of gene conversion-type products in a plasmid molecule with direct repeats (DOHERTY,MORRISONand KOLODNER 1983). Several aspects of homologous recombination reminiscent of eukaryotic gene conversion were observed in E. coli. These include production of one parental type and one recombinant type in a single burst of bacteriophage (BOON and ZINDER 1969; ENEA and ZINDER 1976; SARTHY and MESELSON 1976),high negative interference(clustering of exchanges) in lambda (AMATIand MESELSON1965; JONES, WAGNER and RADMAN1987), andunequal marker recovery in incorporation of plasmid into lambda (THALER, STAHL and STAHL 1987b). We are grateful to A. J. CLARK, B. HOHN, Y. KITAMURA,A. IKEDA for the generous gift of bacterial strains. We MIURA and H. appreciate discussion and encouragementby A. NOMOTO,1. SAITO, M. S. Fox, H. IKEDA, ROSSIGNOL J. and F. W. STAHL. We appreciate and J. help in the course of publicationby G. MOSIG,N. KLECKNER W. DRAKE. This work was supported by grants to I.K., H.Y., K.Y., and A. NOMOTOfrom the Ministry of Education, the Ministry of Health, and Mochida Pharmaceutical Foundation.

LITERATURE CITED AMATI, P., and M. MESELSON,1965 Localizednegativeinterference in bacteriophage X. Genetics 51: 369-379. S. K . , A. F. TAYLOR, A. M. CHAUDHURY and G. R. AMUNDSEN, SMITH,1986 recD: the gene for an essential third subunit of exonuclease V. Proc. Natl. Acad. Sci. USA 83: 5558-5562. BACHMAN, B. J., 1987 Derivation and genotypes of some mutant derivatives of Escherichia coli K-12. pp. 1190-1219. In: Escherichia coli and Salmonella typhimurium. Cellular and Molecular Biology, Edited by F. C. NEIDHARDT,J. L. INGRAHAM,K. B. Low, B. MAGASANIK, M. SCHAECHTER. ASM, Washington,D.

c.

BOON,T., and N. ZINDER, 1969 A mechanism for genetic recombination generating one parent and one recombinant. Proc. Natl. Acad. Sci. USA 64: 573-574. A. D. LEVINSON and P. H. SEEBURG, CHEN,E. Y., P. M. HOWLEY, 1982 The primary structure and genetic organization of the bovine papillomavirus type 1 genome. Nature 2 9 9 529-534. CLARK, A. J., and A. D. MARGULIES, 1965 Isolation and characterization of recombination deficient mutants of E. coli K12. Proc. Natl. Acad. Sci. USA 53: 451-459. COHEN,A,,andA.LABAN,1983Plasmidicrecombinationin Escherichia coli K12:therole of therecFgene. Mol. Gen. Genet. 189: 471-474. COHEN, A,, and A. J. CLARK,1986Synthesisoflinearplasmid multimers in Escherichia coli K12. J. Bacteriol. 167: 327-335. M . J., P. T. MORRISON and R. KOLODNER, 1983 Genetic DOHERTY,

recombination of bacterial plasmid DNA:physical and genetic analyses of the products of plasmid recombination Escherichia in coli. J. Mol. Biol. 167: 539-560. DOHET,C., S. DZIDIE,R.WAGER and M. RADMAN, 1987 Large nonhomology in heteroduplex DNA is processed differently than single base pair mismatches. Mol. Gen. Genet. 206: 181184. ENEA, V., andN. D. ZINDER,1976 Heteroduplex DNA: a recombinational intermediate in bacteriophage f l . J .Mol. Biol. 101: 25-38. FISHEL,R.A., A. A. JAMES and R. KOLODNER,1981 recA-independent general genetic recombination of plasmids. Nature 294: 184-186. GILLEN,J. R.,and A. J.CLARK,1974TherecEpathwayof MechanismsinGebacterial recombination. pp. 123-135. In: New York. netic Recombination, Edited by R. F. GRELL. Plenum, GILLEN, J. R., D. K. W u m s a n d A. J. CLARK, 1981 Geneticanalysis of the RecE pathway of genetic recombination in Escherichia coli K12. J. Bacteriol. 157: 521-532. HANAHAN, D., 1985 Techniques for transformation of E. coli. pp. 109-135. In: DNA Cloning, Vol. I , Edited by D. M. GLOVER. IRL Press, Oxford. HOLLIDAY, R., 1964 A mechanism for gene conversion in fungi. Genet. Res. 5: 282-304. JAMES,A. A,, P. T . MORRISONand R. KOLODNER, 1982 Genetic recombination of bacterial plasmid DNA; Analysis of the effect ofrecombination-deficientmutationonplasmidrecombination. J. Mol. Biol 1 6 0 411-430. JONES,M.,R.WAGNERand M. RADMAN,1987Mismatchrepair and recombination in E. coli. Cell 5 0 62 1-626. JOSEPH,J. W . , andR. KOLODNER,1983aExonuclease VI11 of Escherichia coli. I. Purification and physical properties. J . Biol. Chem. 258: 10411-10417. JOSEPH, J.W.,andR. KOLODNER,1983bExonuclease VI11 of Escherichia coli. 11. Mechanism of action. J. Biol. Chem. 258: 10418-10424. KITANI,Y . , L. S. OLIVEandA. S. EL-ANI, 1962Geneticsof SordariaEmicola. V. Aberrant segregation at the g locus. Am. J. Bot. 4 9 697-706. recA gene KOBAYASHI, I.,andH.IKEDA,1978Ontheroleof product in genetic recombination: an analysis by in vitro packaging of recombinant DNA molecules formed in the absence of protein synthesis. Mol. Gen. Genet. 1 6 6 25-29. KOBAYASHI, I., and H. IKEDA, 1983 Double Holliday structure: a possible in vivo intermediate form of general recombination in Escherichia coli. Mol. Gen. Genet. 191: 213-220. KOBAYASHI, l.,andN. TAKAHASHI, 1988Double-strandedgap repair of DNA by gene conversion in Escherichia coli. Genetics 1 1 9 751-757. KOBAYASHI, I . , M. M. STAHLand F. W. STAHL, 1984 T h e mechanism of the chi-cos interaction in RecA-RecBC-mediated recombination in phagelambda.ColdSpringHarborSymp. Quant. Biol. 49: 497-506. KRAMER,W., K. SCHUCHART and H. J. FRITZ,1982Directed mutagenesis of DNA cloned in filamentous phage: influence of hemimethylated GATC sites on marker recovery from restriction fragments. Nucleic Acid Res. 10: 6475-6485. S. R., H . NAGAISHI and A. J. CLARK, 1974Isolation of KUSHNER, exonuclease VIII: the enzyme associated with the sbcA indirect suppressor. Proc. Natl. Acad. Sci. USA 71: 3593-3597. LABAN,A,,and A. COHEN,1981Interplasmidicandintraplasmidic recombination in Escherichia coli K-12. Mol. Gen. Genet. 184: 200-207. LIEB,M., 1983 Specificmismatchcorrection in bacteriophage lambda crosses by very short patch repair. Mol. Gen. Genet. 191: 118-125. LICHTEN,M. J., and M. S. Fox, 1984 Evidence for inclusionof region of non-homology in heteroduplex products of bacteri-

Homologous Recombination in E. coli ophage lambda recombination. Proc. Natl. Acad. Sci. USA 81: 7180-7184. LITTLE,J. W.,1967 An exonuclease induced by bacteriophage lambda. 11. Nature of the enzymatic reaction. J. Biol. Chem. 242: 679-686. 1984 Genetic analysis of the recJ LOVETT,S. T . , and A. J. CLARK, gene of Escherichia coli K-12. J. Bacteriol. 157: 190-196. LUSKY, M., and M. BOTCHAN, 1981 Inhibition of SV40 replication insimiancells by specific pBR322 DNA sequences. Nature 293: 79-8 1. MESELSON, M. S., and C. M. RADDING, 1975A general model for genetic recombination. Proc. Natl. Acad. Sci.USA 72: 358361. OISHI, M., 1969 An ATP-dependent deoxyribonuclease from Escherichia coli with a possible role in genetic recombination. Proc. Natl. Acad. Sci. USA 64: 1292-1299. PONTICELLI, A . S., D. W. SCHULTZ, A. F. TAYLOR and G. R. SMITH, 1985 Chi dependent DNA strand cleavage by RecBC enzyme. Cell 41: 145-151. RAPOSA,S., and M. S. Fox, 1987 Some features of base pair mismatch and heterology repair in Escherichia coli. Genetics 117: 381-390. ROSSIGNOL,J. L., A. NICOLAS,H.HAMZAand T . LANGIN, 1984 Origins of gene conversion and reciprocal exchange in Ascobolus. Cold Spring Harbor Symp. Quant. Biol. 49: 13-2 1. 1976 Single burst study of recSARTHY, P. V., and M. MESELSON, and red-mediated recombination in bacteriophage lambda. Proc. Natl. Acad. Sci. USA 73: 461 3-4617. SHIBATA, T., C. DASGUPTA, R.P. CUNNINGHAM and C. M. RADDING, 1979 Purified Escherichia coli recA protein catalyzes homologous pairing of superhelical DNA and single-stranded fragments. Proc. Natl. Acad. Sci. USA 76: 1638-1642. SILVERSTEIN, Z., and A. COHEN,1987 Synthesis of linear multimers of oriC and pBR322 derivatives in Escherichia coli K-12: Role of recombination and replication functions. J. Bacteriol. 1 6 9 3131-3137. SMITH,G. R., S. M. KUNES, D. W. SCHULTZ, A. TAYLOR and K. L. TRIMAN,1981 Structure of Chi hot spots of generalized recombination. Cell 24: 249-438. SOUTHERN, P. J.. and P. BERG,1982 Transformation of mammalian cells to antibiotic resistance with a bacterial gene under control of the SV40 early region promoter. J. Mol. Appl. Genet. 1: 327-341. STAHL,F. W., 1969One way to think about gene conversion. Genetics 61 (Suppl.): sl-s13. STAHL,F. W.,1979a GeneticRecombination. Freeman, San Francisco.

769

STAHL,F. W., 1979b Specialsitesin generalized recombination. Annu. Rev. Genet. 13: 7-24. STAHL,F. W., 1986 Roles of double-strand breaks in generalized genetic recombination. Progr. Nucl. Acid Res. Mol. Biol. 33: 169-194. STAHL,F. W., I. KOBAYASHIand M.M. STAHL,1982 Distance from cohesive end site, cos, determines the replication requirement for recombination in phage lambda. Proc. Natl. Acad. Sci. USA 79: 63 18-632 1. STAHL,F. W., I. KOBAYASHIand M.M. STAHL,1985 In phage lambda, cos is a recombinator for Red pathway. J. Mol. Biol. 181: 199-209. SYMINGTON,L. S., P. MORRISON and R. KOLODNER, 1985 Intramolecular recombination of linear DNA catalyzed by the Escherichia coli RecE recombination system. J. Mol. Biol. 186: 515-525. J. W., T . L. ORR-WEAVER, R. J. ROTHSTEINand F. W. SZOSTAK, STAHL, 1983 The double-strand break repair model for recombination. Cell 33: 25-33. THALER, D. S., M.M. STAHLand F.W. STAHL,1987a Doublechain-cut-sitesare recombination hot spots in the Red pathway of phage lambda. J. Mol. Biol. 195: 75-87. D. S., M. M. STAHLand F.W. STAHL,1987b Tests of THALER, the double-strand-break repair model for Red-mediated recombination of phage lambda and plasmid lambda dv. Genetics 116: 501-511. J., and H. OGAWA, 1967 Breakage of polynucleotide TOMIZAWA, strands by disintegration of radiophosphorus atoms in DNA molecules and their repair. 11. Simultaneous breakage of both strands. J. Mol. Biol. 30: 7-15. J., and H. OGAWA,1968 Breakage of DNAin ret+ TOMIZAWA, and rec- bacteria by disintegration of radioactive phosphorus atoms in DNA and possible cause of pleiotropic effects of recA mutation. Cold Spring Harbor Symp. Quant. Biol. 33: 243251. WAGNER, R., C. DOHET,M. JONES, M. P. DOUTRIAUX, F. HUTCHINSON and M. RADMAN, 1984Involvement of Escherichia coli mismatch repair in DNA replication and recombination. Cold Spring Harbor Symp. Quant. Biol. 4 9 61 1-615. H. L. K . , 1982 GeneticRecombination. Wiley, New WHITEHOUSE, York. J., and M. MESELSON, 1975 Mismatch repair in WILDENBERG, heteroduplex DNA. Proc. Natl. Acad. Sci. USA 72: 22022206. and I. KOBAYASHI, YAMAMOTO, K., H. YOSHIKURA, N. TAKAHASHI 1988Apparentgene conversion in an Escherichia coli rec+ strain is explained by multiple rounds of reciprocal crossing over. Mol. Gen. Genet. 213: 393-404. Communicating editor: F. W. STAHL