The RecE recombination pathway mediates recombination between partially homologous DNA sequences: Structural analysis of recombination products

JOURNAL

OF STRUCTURAL

BIOLOGY

104, 97-106 (1990)

The RecE Recombination Pathway Mediates Recombination between Partially Homologous DNA Sequences: Structural Analysis of Recombination Products PAUL KEIM’ Department

of Biology,

AND KARLG.

University

of Utah,

LARK:!

Salt Lake

City,

Utah

84112

Received August 27, 1990

result of discussions at that meeting, I arranged to spend an additional postdoctoral year in Geneva, where I worked, as a “Polio” (NFIP) fellow in Kellenberger’s laboratory on a bacterial cell wall problem originally started by Bonifas. It was a wonderful year. From this postdoctoral experience, I learned about relating structure to function and (by osmosis) a lot about phage-)l. In subsequent years, we collaborated in several studies extending our interaction from Geneva to St. Louis and eventually to Manhattan, Kansas. The work to be described below, may well be the last bacterial study to emerge from my laboratory (which is now almost exclusively concerned with genetic variation and evolution in eukaryotic genomes). In involves themes characteristic of the Kellenberger laboratory during the fifties-a structure-function relationship mediated by a recombinational system of phage-h. In addition, buried in the Material and Methods section (see: Avoiding selection . . .) is a set of experiments which are relevant to current problems in mutation and transposition, but which typify the approach to problems developed during the fifties by members of the phage group.

Escherichia coli generalized recombination,

utilizing the RecA RecB recombination pathway, requires large stretches W-200 bp) of complete DNA sequence homology. In contrast, we have found that the RecE pathway can promote recombination between DNA with only short stretches of homology. A plasmid containing 10 partially homologous direct repeats was linearized by digestion with specific restriction enzymes. After transformation, a RecE+ (&CA) host was able to circularize the plasmid by recombination between partially homologous direct repeat sequences. Recombination occurred in regions of as little as 6 bp of perfect homology. Recombination was enhanced in the regions adjacent to restriction sites used to linearize the plasmid, consistent with a role of doublestrand breaks in promoting recombination. A mechanism is proposed in which the 5’ exonuclease, ExoVIII, produces 3’ single-stranded ends from the linearized plasmid. These pair with other sequences of partial homology. Partial homologies in the sequences flanking the actual join serve to stabilize this recombination intermediate. Recombination is completed by a process of “copy and join.” This recombination mechanism requires less homology to stabilize intermediates than the degree of homology needed for mechanisms involving strand invasion. Its role in nature may be to increase genomic diversity, for example, by enhancing recombination between bacteriophages and regions of the bacterial chromosome. 0 19!JOAcademic Press, Inc.

Gordon Lark INTRODUCTION

Generalized recombination mediated by the RecA RecB system of E. coli requires extensive homology between participating DNA molecules (Gonda and Radding, 1983; King and Richardson, 1986; Shen and Huang, 1986; Watt et al., 1985). Previous, preliminary data from our laboratory (Keim et al., 1984; Liu and Lark, 1982) described recombinational deletions mediated by the A Red, or the Escherichia coli RecE systems, which occurred within a DNA molecule containing DNA sequences which were only partially homologous (direct repeats ca. 70% homology) (Keim and Lark, 1987). These dele-

FOREWORD

In the spring of 1954, I attended a phage meeting in Gottingen, organized by Max Delbruck. At this meeting I met two Swiss, Edouard Kellenberger and Valentin Bonifas. They caught my attention immediately by arriving late. Bonifas’ passport had expired and, unbeknownst to him, valid passports were needed to cross boundaries in Europe. As the 1 Present address: Department of Biology, University, Flagstaff, AZ 86011-5640. ’ To whom correspondence should be sent.

Northern

Arizona

97 1047~8477/90 $3.00 Copyright 0 1990 by Academic Press. Inc. All rights of reproduction in any form reserved.

98

KEIM AND LARK

tions did not occur when the RecA RecB system was substituted for the A Red or E. coli RecE systems. In this paper, we describe experiments which extend these observations to study homology requirements of the RecE recombinational system. The RecE recombination system is related to the A Red function. In E. coli, several recombination pathways have been demonstrated by isolating mutants which restore recombinogenic activity to recBCmutants (Barbour et al., 1970; Kushner et al., 1974). One of these, the sbcA mutation, appears to allow expression of a recombination system, RecE (Barbour et al., 1970; Gillen and Clark, 1974; Gillen et al., 19811, located on a cryptic lambdoid phage, MC (Kaiser and Murray, 1979). RecE is analogous to the A Red system and can substitute for A Red genes in a recombinant phage (A reverse (Gottesman et al., 1974)). (The A Red recombination system is comprised of two components: red cx, the A 5’ exonuclease, and red p whose product may carry out a synaptic function (Kmiec and Holloman, 1981) and a helix destabilizing function (Muniyappa and Radding, 19861.) One gene product of the RecE system, the exonuclease VIII, appears to be analogous to the A 5’exonuclease (Gillen et al., 1981; Joseph and Kolodner, 1983a,b; Kushner et al., 1974). As yet no RecE function analogous to the red p function has been identified. Thaler et al. (1987a,b) have shown that in the absence of DNA replication, A Red recombination is enhanced in regions of DNA adjacent to doublestrand breaks. They have proposed that double strand breaks may be required as cxintermediates in Red-mediated recombination. In RecE + hosts, plasmid recombination (both inter- and intramolecular) occurs with high frequency (Doherty et al., 1983; James et al., 1982; Symington et al., 1985). When intramolecular recombination was studied using dimer plasmid molecules linearized by the action of restriction endonucleases, these linearized plasmids transformed RecE + hosts with a high frequency-nearly that observed with circular dimers (Symington et al., 1985). In contrast, transformation of such linear plasmids into wild type hosts possessing RecA function but lacking RecE function was reduced 200-fold. The ends of linearized plasmid DNA (double-strand breaks) may stimulate recombination-mediated by RecE. Models for such a process were proposed by Symington et al. (1985). In this paper we use the 3.3-kb DNA sequence from the kangaroo rat, Dipodymis ordii, as an in viuo substrate to study the role of homology and of double-strand breaks in RecE-mediated intramolecular recombination. The 3.3-kb D. ordii DNA sequence (shown in Fig. 1) contains 10 partially homologous (ca. 70% homology) tandem repeat units, each of which is about 260 bp in length (Keim and

FIG. 1. Restriction map and internal homology comparison for the D. ordii 3.3-kb DNA sequence of KR-1. The KR-1 restriction map is shown at the bottom. Homology is compared as a dot matrix in which the sequence is compared with itself at a criteria of 9 matching bases out of 11. Regions with homology appear as dots on diagonal lines. These diagonal lines are indicative of direct repeats in which the intensities of dots increase with increasing homology. The location of these direct repeats is represented at the left and top diagrammatically (-1. Two unique regions are represented as boxes (0). The nucleotide sequence deleted in KR-6 is indicated by the brackets, [ I, on the restriction map. A scale for 1 kg is shown.

Lark, 1987). The repeats are interrupted by two unique sequences (believed to be insertions) which divide the 10 repeats into two arrays of five (Fig. 1). Clones of this sequence are stable in A phage or in plasmid vectors grown in RecA + RecBC + hosts provided that A Red function or E. coli RecE function is not active. In the presence of functional Red or RecE pathways, these cloned DNAs accumulate deletions (Keim et al., 1984; Liu and Lark, 1982). We describe below results of experiments in which double-strand cuts are introduced at different sites into plasmid clones of the D. ordii 3.3-kb sequence. Subsequently, transformation of this linearized DNA into RecE+ hosts results in recombination, leading to a circular plasmid capable of replication. Two parameters appear to determine the position of the recombinational junction: (a) the position of the double-stranded cut which linearized the plasmid, and (b) the degree of homology between the regions which are joined. We find that RecE can promote recombination between regions of sequence which lack complete homology, increasing genome diversity. MATERIALS

AND METHODS

Bactericzl strains. The bacterial strains used were: AB1157 (F-, thr-1, leu6, thil, lucY1, guZK2, arul4, ~~15, mtll, proA2, his4, urgE3, str31, k-x33, supE44, A-, As) from A. J. Clark; JC8679 (same as AB1157 except recB21, recC22, sbcA23) from A. J. Clark; JC9604 (same as JC8679 except recA1) from A. J. Clark; DH5a (F-, recA1, endAl, gyrA96, thi-1, hsdR17 (rk-, mk+),

RecE-MEDIATED supE44, relAI, phiSOAlacZdM15) from Bethesda Research Laboratories. All strains were grown in L-Tris broth (10 g/liter tryptone, 5 g/liter yeast extract, 5 g/liter NaCl and 10 mM Tris pH = 7.5) or on L-Tris plates (L-tris broth with 15 g/liter agar). Media was supplemented with 50 kg/ml ampicillin when appropriate. The cell density of cultures was determined by particle counting using a Coulter Counter Model ZF (Coulter Electronics, Inc.). Plasmids and plasmid preparation. The molecular cloning and nucleotide sequence of the 3.3-kb KR-1 insert have been reported previously (Liu and Lark, 1982; Keim and Lark, 1987). KR-1 was derived from KR-2 (Keim and Lark, 1987) by recloning the 3.3-kb insert of KR-2 into pUC9. Most experiments reported here involve KR-6, a plasmid derived from KR-1. KR-6 is a deletion (247 bp between bp 2264 and 2511, see Fig. 1 [ 1) isolated from KR-1. All plasmid constructions were transformed (Alexander et al., 1984) into DH5a for routine screening or for isolating DNA for sequence analysis. Avoiding selection of deleted plasmids grown in E. coli AB1157 hosts. When a covalently closed, circular, plasmid containing the 3.3-kb repetitive element from D. ordii is introduced into the E. coli RecE + strains JC8679 (recA + sbcA) or JC9604 (recA-sbcA1 the D. ordii DNA accumulates deletions (Keim et al., 1984; Liu and Lark, 1982). When a colony of transformed RecE + cells is picked from the initial selective plate and grown overnight in broth media, up to 95% of the plasmids can be deletion products (1.29 kb, see Fig. 2A, lane a). This is not seen in hosts which lack

A

abcde FIG. 2. Restriction fragment analysis of plasmids prepared on hosts with or without RecE activity. (A) Ethidium bromide stained agarose gels containing DNA fragments from KR-1 or KR-6 preparations digested with BamHI. Bacteria were transformed with KR-1 or KR-6 and grown on ampicillin L-Tris agar. Colonies again were transferred to ampicillin plates and after growth used to inoculate 2-ml cultures in ampicillin GTris broth. These were grown overnight and plasmid was prepared, digested with BamHI, and analyzed by electrophoresis through 1.0% agarose: (a) KR-1 was used to transform the RecE + host JC8679. (b and c) KR-1 was used to transform RecE negative hosts (b) AB1157 and (c) DH5a. (d) KR-6 was used to transform the RecE + host JC8769 or (e) the RecE negative host AB1157. (B) Plasmid inserts (including deletion products) from KR-1 plasmids prepared (lane 1) from colonies derived directly from transformed cells or (lane 2) derived from previously selected colonies. Autoradiographs of Southern transfers hybridized with radioactive KR-1 D. ordii insert DNA. JC8679 (RecE+) was transformed with KR-1 and cells plated on an ampicillin plate. All of the colonies were harvested into broth and collected by centrifugation. An aliquot of this cell suspension was replated on an ampicillin plate while plasmid was extracted from the rest (lane 1). The colonies from the second plating also were harvested and plasmid extracted (lane 2). Plasmids were digested with BamHI, electrophoresed, and analyzed by Southern transfer and hybridization with the 3.3-kb insert from KR-1. The sizes of the undeleted insertions were KR-1 = 3.3 kb and KR-6 = 3.05 kb. The size of pUC9 vector is 2.7 kb.

RECOMBINATION

99

RecE function, such as E. coli AB1157 (Fig. 2A, lane b) or DH5u (Fig. 2A, lane c). The following experiments suggest that deletion products accumulate because of selection against cells containing the intact plasmid KR-1. They demonstrate that deletion products are not selected when the plasmid lacks a small region of the D. ordii DNA, or when plasmid is prepared from cells which have not been passed through cycles of growth separated by a prolonged stationary growth phase. A small sequence between the SphI and BstEII sites within the KR-1 plasmid appears to play an important role in the accumulation of deletions. When this sequence is removed, as in clone KR-6 (see [ I in Fig. 1, there is a more than loo-fold reduction in the frequency of deletion products (compare Fig. 2A, lane d with lane e). Plasmids constructed in vitro from KR-1, using Ba131 digestion from the BstEII restriction site, have suggested that a specific sequence influences the frequency of deletions which are observed. This was demonstrated (data not shown) by using Ba131 to construct two plasmids which define the region of nucleotide sequence affecting the accumulation of further deletions in a RecE + host. One deletion destroys the promoter and reduces the deletion frequency (as in KR-6); the other deletion does not destroy the promoter and undergoes deletion at a frequency similar to KR-1. The sequence of this small region (bp 2429 to 2468 of KR-1) is: CTGTTATTGTTGACAGTATGTGGAACAGGAZ’TA-r which contains a strong transcriptional promoter (underlined). Growth conditions also can affect deletion frequency. KR-1 plasmids isolated directly from RecE + bacterial colonies growing on the initial ampicillin selection plate are seldom deleted (deletions are detected in less than 1.5% of the plasmids isolated, Fig. 2B, lane 1). Thus, deletion does not occur immediately upon transformation. However, a dramatic increase in deletion products of different sizes occurs after harvesting and replating a mixture of these same cells (Fig. 2A, lane 2). When single colonies from the transformation KR-1 into JC8679 (RecE+) were replated and plasmids isolated and analyzed (dat.a not shown), single classes of deletions were predominant in each, the type of deletion varying from colony to colony. Since all colonies give rise to some type of deleted plasmid, this suggested that in any one colony a few rare recombination events have occurred and the resulting deletion products are subsequently selected. The transition of cells into stationary phase and then back through lag into exponential growth appears to be important in this process. This is shown in the experiment in Figure 3, in which pure cultures of AB1157 (used to avoid deletion of KR-1) containing either KR-6 or KR-1 were monitored through two growth cycles by particle counting and by viable cell count (colony forming ability). Upon entering stationary phase, cells carrying KR-1 showed a decrease in plating efficiency (15- to 20-fold over a lo-hr period, shown as E.O.P. at top of Fig. 3). Little loss in viability was observed in cells carrying KR-6. The drop in E.O.P. observed when cells contain KR-1 also was consistent with the longer lag observed when the KR-1 culture was diluted into fresh medium. Thus, cells carrying the intact 3.3-kb plasmid are at a selective disadvantage during the stationary-lag phases of growth. In the experiments below, we have avoided selection of deletion products by using the plasmid KR-6. Analysis of deletionproducts. Deletions described in Fig. 6 were produced when linearized, KR-6, DNA was used to transform JC8679 (RecE + ) in the experiment described in Fig. 4. Colonies from the initial selective plating were scraped off the agar surface and plasmid DNA isolated (Maniatis et al., 1982). Plasmid DNAs were analyzed for deletion products by digestion with BamHI, agarose gel electrophoresis (Maniatis et al., 1982), Southern transfer (Southern, 1975), and hybridization with radioactive nick-translated 3.3-kb D. ordii DNA (Rigby et al., 1977). Autoradiographs were analyzed (see Table I) by densitometer scanning (Quick Scan, Helena Laboratories Corp.). This “batch” DNA also was used to transform DH5a. Individual colonies from this second

100

KEIM AND LARK

- O?

Linear

A

EOP

1.4

,

B

KB 3.03

-

2.05

-

1.551.30-

,o7p -‘I

, 200

I 400

/..

I 1000 Time (mid

600”

I 1200

/ 1400

I 1600

FIG. 3. Viability of AR1157 transformed with either KR-1 or KR-6 plasmid during successive cycles of growth. AB1157, transformed with either KR-1 or KR-6, was inoculated into 25 ml broth cultures to a density of 1 x lo7 cells/ml. These cultures were incubated at 37°C with aeration. Growth of the cells was monitored by particle counting (see Methods and Materials) and by colony count after dilution and plating onto L-Tris plates. Efficiency of plating (E.O.P.) was calculated by dividing the viable cell titer (colony count) by the particle count. The E.O.P. for each culture is written at the top of the figure above arrows ( J ) indicating the sampling time. Each culture was diluted into fresh medium after 20 hr (1200 min). Note the decrease in E.O.P. during the stationary phase for cells containing KR-1. transformation were screened by restriction analysis for unique products. Thirty of these cloned plasmids were analyzed for the sequence of the recombination junction. (Of these, only those known to represent independent isolates are presented in Fig. 6). Plasmids were prepared and restriction enzymes were used to determine the junction point to within several hundred nucleotides. The nucleotide sequence of the junction was then determined by the method of Maxam and Gilbert (1980). The analyses of nucleotide sequence homology in Figs. 1 and 5 utilized computer programs kindly provided by Dr. John Shepherd (Biocenter of the University of Basel, Basel, Switzerland).

Linear

FIG. 4. Deletions generated by transforming RecE+ hosts with linearized plasmids. (A) KR-6 DNA was digested with the three different restriction enzymes shown and then transformed into JC8679 (RecE + 1. Transformed cells were plated on ampicillin L-Tris agar and grown overnight. Colonies were harvested and plasmid DNA was prepared (see Methods and Materials and Fig. 2B). In addition, JC8679 was transformed with covalently closed circular DNA (CCC) from each plasmid and DNA subsequently prepared. Plasmid DNA was isolated from transformed cells, digested with BamHI, separated by electrophoresis, and then transferred to nitrocellulose membranes (Southern, 1975). Radioactive D. ordii 3.3-kb DNA from KR-1 was hybridized to the transfer which then was visualized by autoradiography. (B) Two RecE + hosts, JC8679 (RecA+ 1 and JC9604 (RecA - ), were transformed with KR-6 DNA digested with BgZII. Plasmid DNA was prepared and analyzed as described in A.

JC8679 (RecE + ) with about 10% of the efficiency of supercoiled molecules. Transformation of a wild type host (AB1157) by such linear molecules was very inefficienti.e., about 0.1%. (Most of the few products obtained from AB1157 were undeleted plasmids, suggesting that they may have originated from transformation by residual circular plasmids which had not been cut by the restriction enzymes.)

RESULTS

Deletion products from linearized plasmids. Linearized plasmid molecules must be circularized prior to replication. Therefore, after transformation, each ampicillin-resistant colony should contain a population derived from a single recombination event occurring shortly after transformation. Linearized plasmids were used to study recombination to determine whether the location of doublestrand cuts had specific effects on single recombination events (Symington et al., 1985). KR-6 and KR-1 were cleaved at each of three unique restrictions sites: BgZII, SphI, and BstEII (see Fig. 1). BgZII and BstEII digestions create two ends, both of which have some homology with other regions within the 3.3-kb sequence. The SphI cut adjoins a 500-bp unique region (see Fig. 1) and therefore creates only one end (3’ to the restriction cut) with some homology to other regions in the 3.3-kb sequence. These linear plasmids were successfully transformed into

TABLE I Frequency

of Individual Deletion Products Derived from Linearized KR-6 Transformed into and Replicated in RecE + Bacteria

Insert size

Relative frequency as percent of total BgZII

3.05

4.9

2.8 2.55 2.3 2.05 1.8 1.55 1.3

22.2 5.5 13.9 53.6

SphI

6.6 co.1 0.8 3.6 43.9 45.1

BstEII

ccc

24.7 35.5 4.1 0.5 6.2 co.1 29.0

100 -

Note. Data are from densitometer measurements of autoradiographs of restriction fragments in the experiment in Fig. 4. RecE + cells (E. coli JC8679) were transformed with linearized or covalently closed circular (CCC) KR-6 plasmid (see Fig. 4 for details). The enzymes used to linearize the plasmid head each column. (-), Not detected. CO.1 =
RecE-MEDIATED

5’

Left flanking sequence FIG. 5. Comparison of the nucleotide sequence homology between DNA regions flanking the restriction sites used to linearize plasmids prior to transformation into RecE + hosts (see Fig. 4). This comparison is similar to that described for Fig. 1. As in Fig. 1, the criteria for homology was 9 matching bases out of 11. The restriction site used in linearizing the plasmids is located at the lower right hand corner. All of the sequence 5’ to this site (abscissa) is compared with all of the sequence 3’ to the site (ordinate). The sequence of KR-6 (A,B, or Cl reads from left to right and continues from bottom to top. This is shown diagrammatically in (A). (A) The KR-6 nucleotide sequence 5’ to the BglII is compared with sequence 3’ to the BglII site. (B) The KR-6 nucleotide sequences 5’ and 3’ to the SphI site are compared. (Cl The KR-6 nucleotide sequences 5’ and 3’ to the BstEII site are compared. Deletions which result from recombination between regions ofhomology will yield products of defined size (i.e., 1.3,2.05, 2.3 kbl. The size of these products is indicated on the lines corresponding to regions of homology between which deletions can occur. This notation facilitates a comparison between homology in this figure and frequency of deletion products tabulated in Table I. Not all deletion products are noted. Thus there is no notation of the 1.55- and 1.6kb products (corresponding to the two lines of homology between the (1.30) and (2.05) products notations). Examples of deletions A[ - - - ]A and (A - - - A), yielding products of size 1.3 kb ( t A) and 1.8 kb (A t ), respectively, are shown in (A). Deletion products examined in Fig. 6 are indicated by arrows ( f 1. The asterisk (*) in (B) denotes the region of recombination yielding a high frequency of 1.55-kb products. As in Figure 1, the ordinate and abscissa have different scales: the distance from BamHI to Sal1 on the abscissa corresponds to ca. 1 kb; the distance from B&E11 to BamHI on the ordinate corresponds to ca. 0.75 kb.

RecE + linearized extracted deletions,

cells were transformed with circular or plasmids and the transformation products and analyzed. To minimize the selection of plasmids were extracted directly from col-

RECOMBINATION

101

onies growing on the first ampicillin selection plates (see Methods). (Several hundreds of colonies were pooled, by washing them off the plates.) These plasmid preparations were analyzed by restriction and electrophoresis (Fig. 4). It can be seen that in contrast to circular DNA (Fig. 4, CCC), transformation with linear DNA gave rise to various deletion products (Fig. 4, linear). Analysis of plasmid obtained from single colonies demonstrated that each such colony contained only one size of deletion product (data not shown) characterized by a unique set of restriction fragments defining the recombination region (see sequence data below). This is consistent with a single early deletion event in the cells transformed with linear DNA. We have used RecA +RecE + and RecA-RecE + hosts and found no difference in the frequency or distribution of deletion products (as in Fig. 4B). This is not unexpected since the deletion process in covalently closed circular plasmid does not require RecA (Keim et al., 1984; Liu and Lark, 1982). Table I presents the frequencies of the different deletion products which arise when KR-6 is circularized. It is clear that the frequency of various types of deletions differ, depending on the restriction site used for linearization (Fig. 4, Table I). Plasmids linearized by BgZII digestion frequently generate 2.05and 2.3-kb classes of deletion products which rarely arise from SphI or B&E11 digested plasmids. The 1.55-kb class of deletion products found in KR-6 are generated frequently from SphI digested plasmid but infrequently when either the BgZII or B&E11 digested plasmid are used for transformation (Table I). Finally, small deletions (which give rise to the 2% kb deletion products in Table I) commonly arise from plasmid linearized by BstEII digestion but not from plasmid linearized by digestion with BgZII or SphI. The classes of deletion products described in Table I correspond to pairings formed between regions of partial homology. This is seen in the dot matrix representations in Fig. 5, in which homology is compared for the sequences flanking the restriction sites used to linearize the plasmid. In these graphs, the lower right hand corner is the site at which the restriction enzyme cuts the plasmid (BgZII, SphI, or BstEII). The sequence of the KR-6 insert (3.05 kb) is represented from left to right and continues past the restriction cut from bottom to top. Lines of dots represent homologies between regions separated by 1.75, 1.5, 1.25 kb, etc. of sequence. These distances also are correlated with the sizes of different classes of deletions which have been observed. Thus, deletion products of a particular size (e.g., 1.3 kb) are formed when deletion of a corresponding fragment (e.g., 1.75 kb) occurs. The 1.3-kb deletion products can be formed by removal of 1.75 kb from any region of the 3.05-kb KR-6 sequence. Thus, the different

102

KEIM AND LARK

1.3-kb products correspond to the homology line (labeled 1.3) representing homology between sequences 1.75 kb apart. (A particular 1.3-kb product, shown as ( t A) in Figure 5A occurs as a result of a deletion of 1.75 kb (represented on the ordinate and abscissa of Fig. 5A as A[ - - - IA): it consists of sequence on the abscissa to the right (5’ of 1) joined to sequence on the ordinate above the deletion (3’ of I).) In contrast, a particular 1.8-kb deletion product (a t ) is formed by a deletion of 1.25 kb (A - - - A). Therefore this group of deletion products corresponds to a line representing homology between sequences 1.25 kb apart. More frequent classes of deletions, as determined by restriction mapping are shown by (*>. In general, higher homologies correspond to more frequent deletion products (e.g., compare the 1.3-kb class in Table I and the homology in Fig. 5). However, that is not always the case. For example, the frequency of 1.55kb deletion products, formed from KR-6 made linear by digestion with SphI, is much greater than might be expected from the homology which corresponds to the pairing of regions 1.5 kb apart (3.05 kb minus 1.55 kb). The comparison in Fig. 5B shows that one region of homology occurs between the DNA close to the 3’ side of the SphI site (vertical axis) and a region located (on the horizontal axis) about 600 bp from the 5’ BamHI site (see the (*) in Fig. 5B). We have found that more than 90% of these 1.55-kb deletion products contain the B&E11 site (data not shown). This suggests that recombination has occurred between the SphI cut and the B&E11 site on the one hand and the region of homology located at about 600 bp (between the Hind III and the Sal1 sites) on the other hand (see Fig. 5B (*) and Fig. 6: clone 7-25). Thus recombination is much more frequent in the neighborhood of the double-strand cut generated by SphI digestion. All of the results suggest that when two DNA sequences have small regions of homology, a double-strand break near to, or in, one of these regions will increase the chance of a recombinational event leading to circularization of the linear plasmid. To further study the role of double-strand breaks, specific deletion products were isolated and characterized by sequence analysis. Sequence of deletion products. Examples of plasmids containing deletion products representing the different deletion classes were isolated (see Materials and Methods). These were restriction mapped, and the recombination region was determined by DNA sequencing. These examples are marked by arrows on the homology matrices in Fig. 5. For each of these products, the sequence of the recombination region and of the flanking regions are given in Fig. 6. The sequence of the deletion product was compared to the sequence of the flanking regions in the parental molecule from which the deletion was de-

A DELETION

CLONE

l-14

(1.293kb)

(Bgl

II

digest)

1083

DELETIONS

l-18,

7-33.

2-31

(1.294kb)

(Bglll.

Sphl.

and

BsfEll

digest)

C DELETION 810

DELETIONS

2-23

2-27

(1.293kb)

and

(SsfEII

l-17

digest)

(1.293kb)(BsfEll

and

Bglll

digest)

I

DELETION

l-13

(2.056kb)

(&ill

digest)

513 C*GCCCTCCTTCCCGChCCCCCTIAd

K

FIG. 6. The nucleotide sequence of recombination regions within KR-6 deletion products produced in the experiment in Fig. 4. Deleted plasmids resulting from transformation of JC8679 with linearized KR-6 DNA (Fig. 4) were isolated and the recombination region was determined by nucleotide sequencing. The figure shows the nucleotide sequence of each of the parental flanking sequences with which the recombinant is 100% homologous. They overlap in the region of the deletion endpoints. Homology between the parental sequences is indicated (*). The deletion clone number (e.g., l-14), size of the deletion product (e.g., 1.293 kb), and enzyme used to linearize the DNA prior to transformation (e.g., BstEII digest) are noted (e.g., 1083 and 2930) as well as the distance to the restriction site of the enzyme used to linearize the DNA (e.g., BgZII-237 bp).

RecE-MEDIATED

rived. As expected, the two parental regions were always partially homologous to each other. Comparing the deletion product to the parental sequences demonstrated that the 5’ end was identical to the 5’ parent and the 3’ end was identical to the 3’ parent. The join occurred in a region in which the two parental molecules were completely homologous. This could be as long as 21 bp (Fig. 6D) or as short as 6 bp (Fig. 6F, CCTGAG). The most frequent class of deletion products is 1.3kb shown in Fig. 6 A,B,C, and D. These examples represent seven independently derived clones in which sequence was analyzed. They all contain tracts of 21 or 28 base pairs of perfect parental homology. In five out of seven of these clones (e.g., Fig. 6 A,B, and C), the parental regions appear to have been joined within a short stretch of perfect homology (7-10 bp) bordered on the one side by a larger (21-28 bp) region of perfect homology and on the other by a region of poorer homology. In only two instances (Fig. 6D) had recombination taken place

a !I

Double

strand

(d.s.1

103

RECOMBINATION

inside of the larger continuous stretch of perfect homology. This suggests that regions of homology may be used to achieve stable pairing, and that recombination may occur adjacent to such a region (see Discussion and Fig. 7). This is supported by the finding that in other examples (Fig. 6, examples E,F,H, or I) the region of recombination is flanked on one side by stretches of reasonably good homology, while the other side contains many more mismatches. Finally, we have observed that recombination can occur in stretches with as little as 6 bp of perfect homology imbedded in regions of relatively weak homology (Fig. 7F). In most of these cases, recombination has occurred close to the restriction site used to create the linear plasmid. These data suggest that recombination near the end of a nucleotide strand (derived from a double-strand break) required less homology, possibly because primer extension of the nucleotide 3’ end provides additional pairing which stabilizes the recombinational intermediate (see Fig. 7 and Discussion below).

break

5’ 3’

5’ exonuclease

+

Adjacent

to d.s.

break

1 (DNA

polymerase,

Repair

etc.)

G DNA

Further

repair

*-----

replication f

/+

I -----*

4

heterdduplex

text

FIG. 7. A mechanism for RecE-mediated for a detailed discussion.

recombination

between

homologous

sequences

separated

by a double-strand

break.

See the

104

KEIM DISCUSSION

Selective advantage of KR-6, a deletion product of KRl. Before discussing RecE-mediated recombination, it is worthwhile to consider the effect of the plasmid KR-1 on the growth of its host. KR-1 can be viewed as KR-6 containing an inserted sequence with a strong promoter. This additional sequence appears to exert a strong selection against the survival of the host as these bacteria enter the stationary phase of growth. Selection during the stationary phase of growth has not been as well studied as selective effects occurring during the exponential growth phase. However, such selections may well play a role in the accumulation of transpositions (Arber et al., 1978) and in mutational studies such as those described recently by Cairns et al. (1988). Observations such as our suggest that a detailed study of the stationary phase of bacterial growth may be richly rewarding to those interested in the evolution of bacteria. RecE-mediated recombination. The D. ordii, repeated, DNA sequence, is subject to deletion events which occur as the result of recombination between partially homologous sequences. These events are catalyzed by the Red and RecE recombination systems but not by the RecA RecBC system. We have linearized KR-6 plasmids carrying most of this sequence and analyzed the deletion products formed when the plasmid circularizes by recombination between different partial repeats. Like deletions occurring in the covalently closed plasmid, these circularizing events are mediated by the RecE system. The size of the deletions are determined by the homologies used to circularize and are strongly affected by the position at which the sequence was cut to linearize the plasmid. Linearization with different restriction enzymes altered the location of the double strand cut and our results indicate that recombination frequently occurs near to, or at the end formed by the restriction cut (see Fig. 5). This is consistent with the results of Thaler et al. (1987a,b). Recombination itself does not have to occur in large regions of complete homology. Instead, the exchange more frequently occurs adjacent to such regions. When this happens, the recombinational join is most often flanked on one side by a region of great homology, but on the other by a region of poor homology (for example, Fig. 6h). A simple mechanism to explain our observations is presented in Fig. 7. This mechanism (similar to one proposed by Symington et al. (1985)) postulates an essential role in the deletion process for a 5’ exonuclease, such as the ExoVIII, induced by the sbcA mutation (Gillen et al., 1981; Liu and Lark, 1982). In a first step the ExoVIII digests the 5’ strand of the DNA from the site of the restriction cut (Figs. 7a and 7b). The resulting, oppositely oriented, single

AND

LARK

strands pair at a region of partial homology. If this region includes an existing 3’ end (Fig. 7d), chain elongation and ligation complete the join to the 5’ end (Fig. 7e). If the pairing leaves an overhanging 3’ strand (Fig. 7~1, at least one of these would have to be digested before subsequent elongation and ligation (Figs. 7d, 7e) would again complete the process. Progeny could be derived from both strands of the recombinant heteroduplex molecule (Fig. 70 or from only one strand. (Symington et al. (1985) have reported evidence for progeny derived from both strands of such recombinant molecules.) The mechanism proposed postulates an initial pairing step in which intermediates (Figs. 7c or 7d) are stabilized according to the homology between paired strands. Since repair and ligation complete the join, the region in which covalent joining occurs need only be flanked on one side by homologous pairing to stabilize the molecular intermediate (see for examples Fig. 6). Presumably the region of poor homology is copied during the repair process using as a primer the strongly paired homologous 3’ strand (see Stahl and Stahl, 1986). This often places the junction of information adjacent to the strong homology. This mechanism explains why RecE can promote efficient circularization of linear DNA molecules with imperfectly homologous direct repeats. According to the mechanism proposed in Fig. 7, the proximity of regions of homology to double strand cuts will stabilize joint molecules already prepared to initiate repair from a 3’ primer end. Therefore, homologies close to the double-strand cut should favor particular deletion products. In Table I it was seen that the BgZII cut favored 2.05-kb products, whereas the SphI cut favored 1.55-kb products. This can be explained by the close proximity of a sequence 3’ to the BgZII cut which is homologous to sequence near to the HincfIII site (see Fig. 5A (“1 and Fig. 6i). A similar explanation can account for the frequency of 1.55-kb products after transformation by plasmid cut with SphI (see Fig. 5B (*) and Fig. 6e). In both cases, intermediates of the type shown in Fig. 7d would be rapidly stabilized by primer extension as in Fig. 7e. Our results and the proposed mechanism in Fig. 7 are consistent with results implicating doublestrand breaks as intermediates in A Red recombination (Thaler et al., 1987a,b). Moreover, the mechanism in Fig. 7 would explain the failure to obtain recombinants between plasmid and A phage by Shen and Huang (1986), and King and Richardson (1986). In their experiments, a very low frequency of recombinants was observed using the RecE pathway. If RecE acts exclusively on double-stranded breaks (as in Fig. 71, recombination between a plasmid and X phage would generate a nonviable phag*plasmid hybrid (e.g., containing only one arm of the phage).

RecE-MEDIATED

It would take a second, and presumably independent, recombination event to produce a hybrid phage which could be successfully packaged. In contrast, a single intramolecular recombination event within plasmid or chromosomal replicons would be viable, as would intermolecular recombinations between two A phage particles. Recently, measurements have been made of intermolecular recombination between portions of the 3.3-kb D. ordii sequence, cloned into different A phage parents (Roth, 1989). Those results confirm that the Red and the RecE recombination systems can promote intermolecular recombination between partially homologous sequences which are not recombined by the RecA or RecF systems. The frequencies of recombination that were observed were similar to those reported here for intramolecular recombination. Those frequencies were about 10% of the frequency observed with very large regions of complete homology (capable of supporting the RecARecF process). However, RecE could mediate recombination between homologies too low to support the RecARecF system. RecA-mediated recombination events require long sequences (75-300 bp) of perfect homology for optimal recombination (Gonda and Radding, 1983; King and Richardson, 1986; Shen and Huang, 1986; Watt et al., 1985), below which recombination frequency decreases rapidly. In contrast, regions of relatively poor homology appear to suffice for RecE-mediated deletions (Fig. 6). For RecA-mediated recombination, strand invasion (Meselson and Radding, 1975; Radding et al., 1982; Radding, 1982) requires displacement of a homologous strand (i.e., competition with homologous DNA for basepairing) to produce a D-loop. As presented in Fig. 7, the action of a 5’ exonuclease can remove the competitive aspect of strand invasion and, therefore, only requires partial homology with which to stabilize the paired intermediate. This basic difference between the two mechanisms, strand invasion vs pairing of single strand ends, can reduce the degree of homology required to stabilize recombination intermediates. It would, therefore, differentiate the homology requirements of the RecARecBC and the RecE processes. We have noted above that the mechanism described in Fig. 7 will generate viable recombinant progeny between A phages, and between portions of the E. coli chromosome. Such a mechanism can create diversity in the region of the junction and, eventually, permit the joining of sequences that initially could not be joined. It seems reasonable to assume that the generation of diversity outlined above may occur in populations of phage or bacteria and allow mixing of genetic material. These observations suggest that RecE or Red-mediated recombination between partially homologous DNA sequences may lead to the rapid evolution of divergent sequence;

105

RECOMBINATION

whereas the stringent homology requirements of the RecA RecBC system tend to conserve sequences during DNA recombination and repair. We thank Lisa Baird and Louella Apuya for technical assistance. We thank John Clark for bacterial strains and John Shephard for homology comparison programs. This research was supported by a grant (AI 10056) from the National Institute of Allergy and Infectious Disease, and by a grant (BRSG Grant SO7RR07092) from the National Institutes of Health. REFERENCES ALEXANDER, D. C., MCKNIGHT. (1984) Gene 31, 79989.

T. D.,

AND

WILLIAMS,

ARBER, W., IIDA, S., JUNE, H., CASPERS, P.. MEYER, HANNI, C. (1978) Cold Spring Harbor S’ymp. Quant. 1197-1208.

B. G. J., AND Biol. 43,

BARBOUR, S. D., NAGAISHI, H., TEMPLIN, H., AND CLARK, A. J. (1970) Proc. Natl. Acad. Sci. USA 67, 128135. CAIRNS, J., OVERBAUGH, J., AND MILLER. S. I 1988) Nature (Lendon) 335, 142-145. DOHERTY, M. J., MORRISON, P. T., AND KOLODNER, R. (1983) 6. Mol. Biol. 167, 539-560. GILLEN, J. R., AND CLARK, A. J. (1974) in GRELL, R. F. (Ed.), The RecE Pathway of Bacterial Recombination, Mechanisms in Recombination, pp. 123-136, Plenum, New York. GILLEN, J. R., WILLIS, D. K., AND CLARK, A. J. ( 1981) J. Bacterial. 145, 521-532. GONDA, D. K., AND RADDING, C. M. (1983) Cell 34, 647-654. GO’ITESMAN, M. M., GOTTESMAN, GELLERT, M. (1974) J. Mol. Biol. JAMES, A. A., MORRISON, P. T., AND Biol. 160, 411430. JOSEPH, J. W., AND KOLODNER, R. 10,411-10,417. JOSEPH, J. W., 10,418-10,424. KAISER, 174.

K.,

AND

KOLODNER,

AND MURRAY,

M. E., GOTTESMAN, S., AND 88, 471-487. KOLODNER. R. (1982) J. Mol. (1983a)

R. (1983b3

N. (1979)

Mol.

J. Bf.01. Chem.

258,

J. Blol.

Chem.

258,

175,

159-

Gen

Genet.

KEIM, P., THLIVERIS, A. T., MEENEN, E. E., AND LARK, K. G. ( 1984) in CANTOR, A., CHESS, L., AND SERCARZ, E. (Eds.), Regulation of the Immune System, Vol. 18, pp. 511-526, UCLA Symposia on Molecular and Cellular Biology, New Series. KEIM, P., AND LARK, K. G. (1987) J. Mol. Euol. KING, S. R., AND RICHARDSON, J. P. (1986) Mol. 141-147. KMIEC, E., AND 12,63612,639.

HOLLOMAN,

W. K.

(1981)

25, 65-73. Gen. Genet.

.J Biol.

Chem.

KUSHNER, S., NAGAISHI, R., TEMPLIN, A., AND CLARK, (1971) Proc. Natl. Acad. Sci. USA 68, 824-928. KUSHNER, S., NAGAISHI, R., AND CLARK, A. J. 11974) hoc. Acad. Sci. USA 71, 3593.

204, 256, A. J. Natl.

LIU, L., AND LARK, K. G. (1982) Mol. Gen. Genet. 188, 2736. MANIATIS, T., FRITSCH, E. F., AND SAMBROOK, J. (1982) Molecular Cloning: A Laboratory Manual. Cold Spring Laboratory, Cold Spring Harbor, New York. MAXAM, A. M., AND GILBERT, W. (1980) in GROSSMAN, L. (Ed.), Methods in Enzymology, Vol. 65, pp. 499-560. MESELSON, M., AND RADDING, C. M. (1975) Proc. Natl. Acad. Sci. USA 72, 358-361. MUNIYAPPA, K., AND RADDING, C. (1986) J. Biol. Chem. 261, 7472-7478. RADDING,

C. M.,

FLORY,

J., Wu,

A.,

KAHN.

R.,

DASGUFTA,

C.,

106

KEIM AND LARK

GONDA, Harbor RADDING,

D., BIANCHA, Symp. Quark C.

M., AND TSANG, S. S. (1982) Biol. 47, 821-825.

M. (1982) Annu.

Rev.

ROTH, E. J. (1989) Low Homology Red. Master of Science thesis, City, UT. SHEN,

P., AND HUANG,

C., AND BERG,

P. (1977)

Recombination Mediated by A University of Utah, Salt Lake

H. V. (1986)

SMITH, G. (1983) in HENDRIX, F. W., AND WEISBERG, R. A.

Spring

16, 405437.

Genet.

RIGBY, P. W. J., DIEKMANN, M., RHODES, J. Mol. Biol. 113, 237-251.

Cold

Genetics

R. W., (Eds.),

112, 441-457.

ROBERTS, J. W., STAHL, General Recombination,

Lambda Spring

II, pp. 175-209, Harbor, NY.

Cold

SOUTHERN, E. (1975) J. Mol. STAHL, F. W., AND STAHL, M. SYMINGTON, L. S., MORRISON, Mol. Biol. 186, 515625. THALER, D. S., STAHL, M. M., Biol. 195, 75-87. THALER, D. S., STAHL, M. M.,

Spring

Harbor

Laboratory,

Biol. 98, 403-417. M. (1986) Genetics P., AND KOLODNER, AND

STAHL,

AND STAHL,

Cold

113, 1-12. R. (1985).

F. W. (1987a)

J.

J. Mol.

F. W. (1987b)EMBOJ.

6, 3171-3176. WAIT, V. M., (1985) Proc.

INGLES, C. J., URDEA, Natl. Acad. Sci. USA

M. S., AND 82, 4768-4772.

RULER,

W. J.