INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

As many as 15 to 20 proteins are thought to be directly involved in homologous recombination in Escherichia coli (for reviews, see references 19 and 23). The biochemical activities of most of these proteins have been extensively characterized. Even so, it is difficult to specify the precise roles of most of the recombination proteins. One main reason for the difficulty is that only one of the known proteins, RecA, is strictly required for recombination to take place. Certain recombination events, involving strand annealing rather than strand invasion, can take place in a recA null mutant, but these require bacteriophage-encoded recombination functions (36). The nearly complete deficiency of recA mutants, blocked at an early step in recombination (5), combined with extensive biochemical characterization of RecA protein have yielded a relatively clear description of RecA's role in recombination: synapsis and strand invasion (for a review, see reference 32.

Several of the recombination proteins of E. coli are helicases. Recombinant formation following conjugation or generalized transduction depends upon one of the helicases, the RecBCD protein, which is also a complex nuclease. In the absence of RecBCD, efficient recombination can take place if the sbcB gene, as well as the sbcC or sbcD gene, encoding other nucleases, is mutated. In this setting, recombination depends upon several helicases, including RecG, RecQ, and RuvAB (19). Another helicase, PriA, is important for recombination in otherwise-wild-type cells (18, 33).

One recombination protein, RuvC, is a junction-specific endonuclease that has the properties expected of a Holliday junction resolvase (8). Recombination in wild-type E. coli is not highly dependent upon RuvC, though, suggesting either that there is another resolvase or that cutting a Holliday junction is only one of two or more pathways to the production of recombinants. The latter explanation appears to be the case: Lloyd (21) found that recombination in a ruvC mutant is dependent on RecG, and biochemical studies of RecG protein have uncovered helicase but not endonuclease activity (24). Rather, Whitby and coworkers (38, 39) have proposed that RecG helicase can process three-stranded junctions in such a way as to generate recombinants without endonucleolytic resolution of a Holliday junction.

The bacteriophage  homologous recombination system, known as Red, normally promotes highly efficient double-strand break repair and recombination transiently in -infected E. coli. The Red-induced "hyper-rec" state of E. coli can be made permanent if the  recombination proteins are expressed in the cell in the absence of infecting phage (29, 31). In addition to operating at higher efficiency than other homologous recombination pathways in E. coli, Red-mediated recombination is possibly simpler in its early steps (diagrammed in Fig. 1). The  Gam protein shuts down all the enzymatic activities of RecBCD (16, 28). Double-stranded ends initiate Red-mediated recombination (35, 37). The  exonuclease (Red) loads onto double-stranded ends and processively degrades the 5'-ended strand, leaving a 3' single-stranded tail (20). RecA and  Red protein then cooperate in promoting invasion of the 3'-ended strand into an unbroken homologous duplex (17, 27, 31). (Red can promote recombination by strand annealing in the absence of RecA protein, but only if a partner with a double-stranded end at a nonallelic site is provided [36].) In the absence of RecBCD, the only  proteins required for efficient double-strand break repair and recombination are Red and Red; host proteins carry out all other steps.

View larger version (14K): [in this window] [in a new window]   FIG. 1.   Possible mechanism of Red-mediated recombination involving strand invasion. The steps, and the roles of the recombination proteins, are described in the text.

An E. coli strain containing  recombination proteins provides an experimental setting in which recombination between nonreplicating  phages occurs at high frequency at the site of a double-strand break (37). Formation of recombinants between physically marked chromosomes can be monitored directly by extraction and analysis of DNA from infected cells. It depends upon (i) double-strand breaks, (ii) RecA, (iii) Red proteins, and (iv) inactivation of RecBCD, either by mutation or by  Gam protein (31).

In this study, we examined the properties of Red-expressing E. coli bearing mutations in ruvC and recG, singly and in combination. We found that the dependence of Red-mediated recombination on RuvC was variable among the different types of crosses tested. Red-mediated recombination is more efficient in a recG mutant than in a recG⁺ strain, in contrast to recombination via other pathways (RecBCD, RecF, or RecE), which is reduced in recG mutant (22). However, as in wild-type E. coli, this Red-mediated recombination was, in most cases, highly dependent upon RuvC in the recG mutant. These observations distinguish between models that have been proposed for the roles of RuvC and RecG in recombination (38, 39).

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacteria. Except as noted, all bacterial strains used in this study were derivatives of E. coli AB1157 (argE3 his-4 leuB6 proA2 thr-1 ara-14 galK2 lacY1 mtl-1 xyl-5 thi-1 rpsL31 tsx-33 supE44); it and JC15329 [(srlR-recA)306::Tn10] were obtained from A. J. Clark. Strains CS85 (eda-51::Tn10 ruvC53) (34) and N2731 (recG258::kan) (22) were obtained from F. W. Stahl. Strain CS85 was reconstructed by transduction of AB1157 with a P1 lysate grown on strain CS85, with selection for tetracycline resistance. A tetracycline-sensitive derivative, TP438, was subsequently selected on the basis of fusaric acid resistance (26). TP440 (ruvC53 recG258::kan) was constructed by transduction of TP438 with a P1 lysate grown on N2731, with selection for kanamycin resistance. Phage recombination functions and the EcoRI restriction-modification system were introduced into these strains by simultaneous transformation with plasmids pTP223 and pMB4, as described previously (31).

Phages. wild type and  nin5 were obtained from F. W. Stahl.  sR1° RFLP381 sR3° sR4° sR5° and sR1° RFLP382 sR3° sR4° sR5°, referred to below as  RFLP381 and  RFLP382, have been described previously (31). Additional RFLP substitutions were crossed into the  chromosome as described previously (31); recombinants were selected as described or on the basis of plaque formation on a phage P2 lysogen (Spi phenotype) and were given the numerical designations of their plasmid parents (e.g.,  RFLP835 was made by crossing  wild type with plasmid pTP835). Phages were propagated in a (recC ptr recB recD)::P_tac-bet-exo cat derivative of W3110 strA594 lac gal, which was constructed by P1 transduction with KM32 as the donor and selection for chloramphenicol resistance.

Crosses. Bacterial host strains were cultured and infected with parent phages at a multiplicity of seven each as described previously (31). When plasmid-bearing strains were used as hosts, retention of the plasmid pMB4 bearing the EcoRI restriction-modification system was measured as described previously (31); in cases in which less than 85% of the cells retained pMB4, the experiment was discontinued.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Red-mediated recombination independent of RuvC and RecG. Genetic and biochemical studies of recombination in E. coli, summarized in the introduction, suggest that a recG ruvC double mutant, which is highly defective in recombination, is specifically blocked at the resolution step. A prediction that follows from this idea is that unresolved recombination intermediates might accumulate in replication-blocked crosses between  phages in a recG ruvC double mutant. To test this, we constructed single- and double-mutant strains bearing plasmids that express red genes,  cI repressor, and the EcoRI restriction-modification system. The somewhat surprising result of phage crosses in these strains (Fig. 2) was that the frequency of recombination relative to that in the wild type was increased by mutation of recG and only slightly affected by mutation of ruvC, either alone or in combination with recG. In seeking to understand this lack of dependence on either RuvC or RecG, we considered and tested a number of possible explanations.

View larger version (20K): [in this window] [in a new window]   FIG. 2.   Recombination (Rec.) between nonreplicating  chromosomes, one of which has a double-strand break. The phages,  RFLP381 (381) and  RFLP382 (382), bear a substitution that replaces  genes between orf194 and ssb with unrelated sequences; the ends of the substitution are marked with different restriction sites: B, BglII, and X, XbaI. The host cells bear two plasmids: pTP223, which expresses  genes cI, gam, red, and red, and pMB4, which encodes the EcoRI restriction-modification system. The phages each have a single EcoRI restriction site in the substitution sequence; in the case of  RFLP382, it was made uncuttable by prior passage through an EcoRI-modifying host. DNA was extracted from infected cells 60 min after infection, digested with BglII, and analyzed on Southern gels. Parental and recombinant bands from three experiments were quantitated; the means and standard errors are indicated.

View larger version (22K): [in this window] [in a new window]   FIG. 3.   (a) Possible mechanism of recombinant formation via the Red pathway, independent of RuvC and RecG. The first steps (not shown) would be the same as in Fig. 1. Branch migration, possibly directionally biased due to the absence of RecG, leads to heteroduplex formation at the site of marker B. Mismatch repair then converts the heteroduplex to a homoduplex, bypassing the need for endonucleolytic processing of a Holliday junction. On the right side, endonucleolytic cleavage and trimming of the displaced strand or, possibly, extension of the invading 3' end would produce a structure that would be counted as a recombinant following extraction, digestion with BglII, and analysis on a Southern gel. (b) A heterologous sequence inserted between the invading end and the marker B is predicted to block the pathway shown in panel a, making recombination more dependent upon RuvC and/or RecG.

Redesign of phages and hosts for detection of resolvase-dependent recombination events. For reasons described above, we constructed new variants of the  RFLP phages (see Fig. 5 [top]). The  sequences replaced by the new RFLP substitutions are the same as in the previous versions, but three main changes were made. (i) In the new phages, the leftmost 1 kb of sequence in the substitution is either the same as before or has been replaced by a heterologous sequence, the cat gene of transposon Tn9. (ii) A unique PaeR7 restriction site has been introduced near the middle of the 5-kb substitution sequence (near the EcoRI site of the previous versions). This change was introduced because other studies had demonstrated the greater efficiency of the PaeR7 restriction enzyme in promoting recombination events in vivo, relative to the EcoRI system employed previously (36). PaeR7-uncuttable partners were constructed by insertion of HindIII linkers into the PaeR7 sites. (iii) The nin region was eliminated by crossing in the deletion nin5.

View larger version (8K): [in this window] [in a new window]   FIG. 4.   Map of the (recC-recD)::red-pae-cI substitution in strain TP507 and derivatives. Sequences between the XmaI site upstream from recC and the SspI site near the C-terminus-encoding end of recD are replaced. The block of sequence designated "red" includes (from left to right) the promoter Ptac and  genes gam, red, and red. pae includes the restriction endonuclease and methylase genes of Pseudomonas aeruginosa R7, transcribed rightward. The  cI gene is fused to the promoter PlacUV5 and is also transcribed rightward.

Resolvase-dependent Red-mediated recombination. Crosses between the new  RFLP phages were predicted to have different outcomes, depending upon which partner was cut, as diagrammed in Fig. 5 (top). Cutting the non-cat-bearing phage should result in the production of recombinants independent of resolution functions, as seen previously. However, according to the scheme outlined in Fig. 3, a double-strand break in the cat-bearing phage should lead to the production of a recombination intermediate that can be processed into a recombinant only by the action of a resolvase.

View larger version (69K): [in this window] [in a new window]   FIG. 5.   (Top) Recombination between  RFLP phages bearing a heterologous sequence. As indicated, production of a recombinant follows different potential pathways, depending upon which phage is cut. On the left, the invading partner has homology (1 kb) at the end but has a heterologous sequence (the cat gene; 1 kb) between the end and the restriction site marker B (BglII). Formation of recombinants by the mechanism diagrammed in Fig. 3 is blocked. On the right, the invading strand encounters no substantial heterology, only the restriction site polymorphism. Formation of recombinants by the mechanism diagrammed in Fig. 3 is permitted. The cross on the left involves phages  RFLP835 nin5 (above) and  RFLP864 nin5 (below); that on the right involves  RFLP863 nin5 (above) and  RFLP836 nin5 (below). (Bottom) Autoradiographs of Southern gels of BglII-digested DNA extracted from host cells 60 min after infection and probed with cat sequences. The positions of parental (P) and recombinant (R) bands are indicated. The lowest band results from cutting with PaeR7 in vivo and with BglII in vitro. The bacterial strains were all (recC-recD)::red-pae-cI. +, present; , absent.

View this table: [in this window] [in a new window]   TABLE 1.   Effects of ruvC and recG mutations on  Red-mediated recombination between phage chromosomes

Biological-recombinant formation. We sought to address the question of whether the formation of recombinant DNA molecules in  RFLP crosses correlates with the production of biological recombinants. We constructed a variant of the RFLP phages,  lac::cat819, which, upon infecting an immune host with the PaeR7 restriction-modification system, is cut by the PaeR7 nuclease, releasing a 4-kb fragment with lacZ and lacY sequences at its ends and the cat gene in the middle (Fig. 6a). Red-mediated recombination results in replacement of the chromosomal lacZ gene with a mutant allele that is defective for lacZ function but confers chloramphenicol resistance on the recombinant. According to the rationale diagrammed in Fig. 3, production of chloramphenicol-resistant recombinants would be expected to require the same functions as production of recombinant DNA molecules in  RFLP crosses in which the cat-bearing chromosome is cut (Fig. 5, top left).

View larger version (18K): [in this window] [in a new window]   FIG. 6.   (a) Recombination between  lac::cat819 nin5 and the bacterial chromosome. Phage repressor in the infected cell prevents transcription of phage genes. PaeR7 restriction endonuclease releases a DNA fragment that recombines with the chromosome, generating a chloramphenicol-resistant, Lac recombinant. (b) Production of recombinants following infection with  lac::cat819 nin5. Log-phase cells (approximately 108/ml) were infected at various multiplicities and, after 1 h of further aeration, were plated for total and recombinant (cat+ lac) titers. Open circles, (recC-recD)::red-pae-cI recG258; solid circles, (recC-recD)::red-pae-cI. Total cell counts increased approximately fourfold during the experiment.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The results of experiments examining Red-mediated recombination between linear and circular partners suggest two main interpretations concerning the roles of RecG and RuvC. (i) the main pathway to recombinant formation involves RuvC and (ii) RecG impedes this main recombination pathway. However, RecG may also stimulate recombination via a RuvC-independent pathway.

In the presence of RecG function, the degree of dependence of Red-mediated recombination on RuvC varied substantially among the three types of crosses described in the preceding section. The dependence was greatest in crosses between nonreplicating phage chromosomes in non-plasmid-bearing cells (Fig. 5). In the other cases, recombination appeared to be nearly independent of RuvC.

The contribution of RecG to Red-mediated recombination was less variable among experiments: in all cases, elimination of RecG function increased the frequency of recombination. In both types of crosses that were carried out in non-plasmid-bearing cellsbetween nonreplicating  chromosomes and between short linear DNA fragments and the bacterial chromosomethe resulting higher-frequency recombination was also more dependent on RuvC (except when recombinant formation could in theory proceed via branch migration, heteroduplex formation, and mismatch repair, as pictured in Fig. 3). The simplest interpretation of these observations is that RecG opposes the main pathway of Red-mediated recombination, which is RuvC dependent. The data in Table 2 suggest that RecG may also promote an alternative, RuvC-independent recombination pathway: recombination frequencies in the ruvC mutant were slightly higher than in the ruvC recG double mutant.

An alternative interpretation of the effects on Red-mediated recombination of the recG258 allele is that they do not result from inactivation of recG but rather from a function expressed by the mini-Tn10 used to generate the allele (22). This interpretation is ruled out by the observation that a simple deletion allele of recG exhibits the same stimulation of Red-mediated recombination as recG258 (unpublished observations).

Our description of the roles of RecG and RuvC is similar to a model proposed by Whitby and coworkers (39), in which RecG actually aborted genetic exchanges resulting from RecA-mediated strand invasion but then allowed RecBCD to catalyze exchanges at the ends of the incoming DNA by an unspecified mechanism. Although these researchers found biochemical evidence for RecG's ability to act in this way, the fact that a recG mutant is recombination deficient relative to wild type apparently led them to favor other interpretations of RecG's activity in recombination (1, 38).

The generation of late-arising revertants of lacZ mutants under conditions of selection is another cellular activity that apparently is impeded by RecG (10, 13). This activity, sometimes called "adaptive mutability," is dependent upon recombination functions (6, 13). Its dependence upon recombination functions may be a consequence of the involvement of gene amplification in the process (3). Harris et al. (13) proposed a model for recombination associated with adaptive mutation, in which RecG aborts 3'-end invasion and promotes 5'-end invasion.

The initial steps of Red-mediated recombination are perhaps less complex than those of RecBCD-mediated recombination or of RecBCD-dependent adaptive mutation. The  exonuclease specifically and processively degrades the 5'-ended strand of double-stranded DNA (20), leaving exclusively 3'-ended single strands for synapsis and strand invasion. Production of these 3'-ended single-strand tails in -infected cells has been observed directly (14). The observation that RecG inhibits Red-mediated recombination that is constrained to proceed via strand invasion favors models in which RecG tends to push out invading 3' ends.

The model diagrammed in Fig. 1 and 3 accounts reasonably well for recombination between linear and circular partners in a red-expressing cell lacking recBCD and recG functions: double-stranded ends are channeled nearly exclusively through a pathway that involves the creation of 3'-ended single-strand tails, which invade an unbroken homologous duplex. If and only if branch migration is impeded by a significant nonhomology, recombinant formation is dependent upon nucleolytic resolution of the resulting Holliday junction by RuvC. In the absence of RuvC, presumably, a recombination intermediate something like the structure diagrammed in Fig. 3b accumulates. We have not observed such an intermediate, but the methods employed in the extraction and restriction enzyme digestion of DNA from -infected cells might be expected to favor its dissociation by spontaneous branch migration.

The model diagrammed in Fig. 1 and 3 does not account so well, by itself, for what happens in a recG⁺ cell. A more complete model would account for three questions raised by the data in Fig. 2 and Tables 1 and 2. (i) Why was recombination between nonreplicating  chromosomes, with no substantial nonhomologies, dependent upon RuvC in one set of crosses (Table 1, non-cat-bearing phage cut) and independent in another (Fig. 2)? One possible explanation is that the interacting DNA sequences in question were not identical in the two experiments and branch migration could proceed all the way past the restriction site polymorphism in the face of opposition by RecG in one case but not in the other. (ii) The crosses diagrammed in Fig. 5 (left side) and Fig. 6 both involve recombination of a linear cat-bearing chromosome with a circular chromosome. Why was recombinant formation dependent upon RuvC in the former case and independent in the latter? A key difference between the two crosses is that replication of both partners is blocked in the  cross (Fig. 5) whereas replication of only the linear partner is blocked in the linear-by-host chromosome cross (Fig. 6). The passage of a replication fork may potentiate alternative pathways for the resolution of recombination intermediates. (iii) How does RecG promote, rather than impede, RecBCD-mediated recombination? One possibility is that RecBCD-mediated recombination proceeds primarily via 5'-ended strand invasion, but extensive studies of the activities of RecBCD, both in vitro (2) and in vivo (11), favor the view that RecBCD generates invading 3'-ended strands. Together, these observations are consistent with the idea that RecBCD may function in the processing or resolution of recombination intermediates that it participates in generating (5, 30). Perhaps the normal role of RecG is to facilitate this function of RecBCD.