|Home | About | Journals | Submit | Contact Us | Français|
Homologous recombination-dependent DNA replication (RDR) of a λ cos site-carrying plasmid is demonstrated in E. coli cells when the cells express λ terminase that introduces a double-stranded break into the cos site. RDR occurs in normal wild-type cells if the plasmid also contains the recombination hotspot χ. χ is dispensable when cells are induced for the SOS response or contain a recD mutation. recBC sbcA mutant cells are also capable of RDR induction. A recN mutation greatly reduces RDR in normal cells, but not in SOS-induced cells. RDR proceeds by the θ mode or rolling circle mode of DNA synthesis, yielding covalently closed circular plasmid monomers or linear plasmid multimers, respectively. Previously described inducible stable DNA replication is considered to be a special type of RDR that starts exclusively from specific sites (or/Ms) on the chromosome.
The major homologous recombination pathway in wild-type Escherichia coli, the RecBCD pathway, requires RecBCD enzyme (exonuclease V), which consists of the products of the recB, recC, and recD genes (for reviews see Smith, 1988; Kowalczykowski et al., 1994). The enzyme binds a flush or nearly flush end of duplex DNA and unwinds it by its helicase activity. When the enzyme encounters a χ sequence (5′-GCTGGTGG-3′) from the 3′ side, it frequently nicks the χ-containing strand 4–6 nt to the 3′ side of χ (Ponticelli et al., 1985). The single-stranded DNA (ssDNA) tail with a 3′ end produced by subsequent unwinding would then be assimilated into a homolog by the action of RecA, yielding a recombination intermediate, i.e., a D loop (Kowalczykowski et al., 1994). Thus, properly oriented χ stimulates homologous recombination. The RecBCD enzyme also possesses a nonspecific nuclease activity that can be attenuated by the interaction between the enzyme and χ (Dixon and Kowalczykowski, 1991, 1993; Taylor and Smith, 1992; Dabert et al., 1992). Subsequent translocation of the nuclease-attenuated RecBCD enzyme could produce ssDNA that remains undigested and can be used by RecA for D loop formation.
The attenuation of the nuclease activity is due either to removal or functional inactivation of the RecD subunit (Dixon et al., 1994). This suggests that the RecBC enzyme found in recD mutant cells may be functionally similar to the wild-type enzyme that has encountered χ (Kowalczykowski et al., 1994). This is consistent with the observations that, while recB and recC mutants are recombination deficient, recD mutants are recombination proficient (Chaudhury and Smith, 1984; Amundsen et al., 1986; Thaler et al., 1989) and that the χ stimulation is not detected in recD mutants (Chaudhury and Smith, 1984). The nuclease activity of RecBCD enzyme is undetectable in cell-free extracts prepared from recD mutants (Chaudhury and Smith, 1984). Furthermore, purified or reconstituted RecBC enzyme possesses a significant level of helicase activity but little nuclease activity (Palas and Kushner, 1990; Boehmer and Emmerson, 1991; Masterson et al., 1992). In vivo experiments also suggest that RecBC enzyme in recD mutant cells is capable of DNA unwinding, but not DNA degradation (Chaudhury and Smith, 1984; Rinken et al., 1992). The χ stimulation of homologous recombination is partially suppressed in SOS-induced cells, although overall recombination levels are not affected (Rinken and Wackernagel, 1992). Since overproduction of the RecD subunit alleviates the suppression, it is proposed that an SOS-inducible factor modifies RecBCD via interaction with the RecD subunit to produce a χ-independent enzyme. The nuclease activity of RecBCD is also attenuated after SOS induction (Kannan and Dharmalingam, 1990, and references therein).
Activation of one of the other two pathways of homologous recombination, i.e., the RecE and RecF pathways, suppresses recombination deficiencies in the RecBCD pathway of recB and recC mutants (Clark, 1991). The RecE pathway is activated by an sbcA mutation and requires the product of the recE gene, exonuclease VIII. Exonuclease VIII degrades one strand of linear duplex DNA in the 5′ to 3′ direction (Joseph and Kolodner, 1983), and thus the enzyme yields a ssDNA tail having a 3′ end that is utilized by RecA for D loop formation. The RecF pathway is activated by mutations in the sbcB and sbcC genes. In this pathway, RecQ helicase (Umezu and Nakayama, 1993) and RecJ nuclease (5′ to 3′ ssDNA exonuclease; Lovett and Kolodner, 1989) are proposed to play important roles in the production of ssDNA (for review see Clark, 1991).
Chromosome replication in E. coli is initiated by interactions between DnaA protein and its binding sites, which results in duplex unwinding of the AT-rich region within the minimal oriC (Kornberg and Baker, 1992). Every round of initiation at oriC requires transcription and translation. Homologous recombination functions are normally not essential for chromosome replication. When E. coli cells are induced for the SOS response by DNA damage or DNA replication inhibition, a different type of DNA replication, called inducible stable DNA replication (iSDR), is activated (for review see Asai and Kogoma, 1994). iSDR replicates the entire genome semiconservatively. Major origins for iSDR map in the oriC and terC regions of the chromosome (Magee et al., 1992) and are termed oriM1 and oriM2, respectively. The active oriC site, however, is not necessary tor oriM1 functions (Asai et al., 1994). The initiation mechanism of iSDR is radically different from that of the oriC system; it is independent of normally required DnaA protein, transcription, and translation, but absolutely requires homologous recombination functions. The D loop model for iSDR initiation (Asai et al., 1993) postulates that upon induction of the SOS response, an endonuclease is induced and introduces a double-stranded break (DSB) at or near oriM on a circular DNA molecule, yielding a linear duplex. The subsequent action of a helicase (e.g., RecBC, RecQ) or a nuclease (e.g., RecBCD, exonuclease VIII) or both at the end of the duplex should, in the presence of RecA, produce a D loop, provided that an intact homolog is available. This reaction, i.e., D loop formation, could accomplish duplex unwinding for iSDR initiation (Asai and Kogoma, 1994).
The D loop model predicts that DSBs trigger a novel type of DNA replication that is independent of transcription and translation. We have tested this prediction by placing a λ cos site on plasmid DNA and generating an artificial DSB at the site with λ terminase, which introduces two staggered nicks, 12 bp apart, one on each strand of the duplex (for review see Feiss and Becker, 1983). We have found that regulated expression of the terminase indeed allows replication of the cos-carrying plasmid in the absence of transcription and translation. The replication requires E. coli homologous recombination functions. Unlike iSDR, it occurs in the complete absence of SOS induction, provided that the cos-carrying plasmid also contains a properly oriented χ site. We designate this replication RDR, for homologous recombination-dependent DNA replication. We discuss possible roles of RDR in DSB repair and homologous recombination.
pFM123 (Figure 1; Murialdo, 1988) is a derivative of pBR322, which carries the genes Nu1 and A encoding λ terminase. Expression of the genes is regulated by the λ PR and PL promoters and a mutant repressor gene, c1857, which encodes a temperature-sensitive repressor. This plasmid does not contain cos sites. pHK-cos (Figure 1) was constructed by inserting a cos-containing fragment into the multiple cloning site of pHK (Asai et al., 1994), a derivative of pSC101. When cells harboring both pHK-cos and pFM123 are incubated at intermediate temperatures, a small amount of terminase could be produced. The terminase is expected to linearize some of the pHK-cos molecules by cleaving cos sites (Figure 2, step A; Kuzminov et al., 1994), while the rest of the molecules remain intact. The left end of λ DNA contains a strong terminase-binding site (for review see Becker and Murialdo, 1990). Thus, terminase, after cutting at cos, remains bound to the left end and blocks access to the end by nucleases and helicases, such as RecBCD and exonuclease VIII (Kuzminov et al., 1994). It is, therefore, expected that the terminase also protects one end of the linearized pHK-cos and that ssDNA tails are produced by a helicase, a nuclease, or both, preferentially from the other end, which contains the right end of λ DNA (Figure 2, step B). The ssDNA would be assimilated into an uncut pHK-cos molecule by RecA, generating a D loop (Figure 2, step C). A DNA molecule containing a Holliday junction could then be formed from the D loop (Figure 2, step D). Resolution of the junction (Figure 2, step E or F) leads to the formation of one of the two types of templates for rolling circle replication, depending on cutting sites (e or f, respectively). DNA replication from the D loop should lead to the formation of a template for either rolling circle mode (Figure 2, steps G and H) or θ mode (Figure 2, steps I and J) of DNA synthesis (see below for details). Therefore, DSBs could trigger extensive DNA synthesis that is independent of the plasmid replication origin.
A thyA strain, AQ5004 (rec+ sbc+; Table 1), carrying both pHK-cos and pFM123 was grown at 30°C in M9-CAA medium containing [3H]thymine (see Experimental Procedures). At this temperature, only a small amount of terminase would be produced. The exponentially growing cells were collected by filtration, washed, and resuspended in M9-CAA medium lacking required thymine. The cell suspension was then incubated at 30°C for a duration of time equivalent to two doublings (80 min) to induce the SOS response by thymine deprivation. A mixture of thymine, rifampicin (Rif), and chloramphenicol (Cm) was added (time 0) to the suspension at the end of thymine deprivation to inhibit subsequent origin-dependent plasmid replication (Asai et al., 1994). After the addition of Rif and Cm, the suspension was incubated at 37°C. Samples were withdrawn at the indicated times, and total DNA was prepared from each sample. Amounts of DNA samples normalized to give equal 3H counts were loaded for agarose gel electrophoresis. pHK-cos plasmid DNA was detected by Southern blot hybridization using a 32P-labeled plasmid-specific probe.
An autoradiogram of this experiment is shown in Figure 3A. Plasmid DNA steadily increased despite the presence of Rif and Cm. The major product of this origin-independent replication of pHK-cos was covalently closed circular (ccc) monomers, although a small amount of linear plasmid multimers (LPMs), which migrate in agarose gel with a mobility similar to that of chromosome DNA fragments (Cohen and Clark, 1986), was also detected (see below). The intensity of 32P radioactivity of each band corresponding to ccc monomers was quantified, and the relative copy number increase was determined. The result (Figure 3B) shows that the ccc form of pHK-cos increased more than 9-fold after a 5 hr incubation in the presence of Rif and Cm.
When pHK-cos or pFM123 was replaced with the corresponding parental plasmid, i.e., pHK or pBR322, respectively, plasmid copy number increase was limited to ~2-fold (Figure 3B), indicating that origin-independent replication requires both cos and terminase. The 2-fold increase in copy number during hour 1 is due most likely to replication from the plasmid origin. Initiation potential, which allows a single round of plasmid replication in the presence of Rif and Cm, would have been accumulated prior to the addition of the drugs.
The copy number increase of pHK-cos in the presence of Rif and Cm was also monitored without SOS-inducing treatment. In this case, cells carrying pHK-cos and pFM123 were grown and treated as described above except that a mixture of thymine, Rif, and Cm was added (time 0) to cell suspension with no period of thymine deprivation. As shown in Figure 3B, the copy number of pHK-cos increased during hour 1 and then remained constant. The result suggests that some factor(s) that is essential for the origin-independent replication of pHK-cos is not present in normal (uninduced) rec+ sbc+ cells and that the factor(s) can be induced by SOS induction. The expected increase of RecA amount after SOS induction is not the factor because SOS-inducing treatment was still required even when recAo (Con) mutant cells, in which RecA is constitutively overexpressed, were used for host cells (data not shown).
We considered the possibility that the failure to observe origin-independent plasmid replication in normal (uninduced) cells is due to the nuclease activity of RecBCD, which is expected to degrade linear duplexes created by cos cutting (Kuzminov et al., 1994). The SOS induction would be necessary to attenuate the nuclease activity of RecBCD (see Introduction). In normal recB+ recC+ recD+ cells, the nuclease activity can be attenuated by a χ site, but the pHK-cos plasmid contains no χ sequence (data not shown). χ is effective only when RecBCD enzyme encounters it from the 3′ side (Taylor et al., 1985). In our system, the enzyme can only bind to one end of linearized pHK-cos that contains the right end of λ DNA (see Figure 2). This allowed us to analyze the orientation-dependent effect of χ. pHKχ1-cos and pHKχ3-cos (see Figure 1) contain a χ site downstream of the RecBCD entry site in the active and inactive orientations, respectively. When pHKχ1-cos along with pFM123 was present in rec+ sbc+ cells, the copy number of the plasmid increased in the presence of Rif and Cm without SOS-inducing treatment (Figures 4A and 4B). In contrast, the level of the copy number increase of pHKχ3-cos was similar to that of pHK-cos (Figure 4B). The basal level of SOS induction in rec+ sbc+ cells (AQ9789; a LacZ− derivative of AQ5004 carrying an sfiAlacZ+ operon fusion) and the same cells containing pHKχ1-cos and pFM123 (AQ9833) was analyzed by measuring the activity of β-galactosidase expressed by an sfiAlacZ+ operon fusion. The results are 56.3 ± 1.4 and 57.1 ± 0.4 Miller units for AQ9789 and AQ9833, respectively. This implies that χ-dependent plasmid replication occurs in the complete absence of SOS induction.
The result described above supports the notion that the attenuation of the nuclease activity of RecBCD is necessary for origin-independent plasmid replication. To test the notion further, we analyzed the copy number increase of pHK-cos (without χ) in recD mutant cells in the presence of pFM123. As shown in Figure 4B, pHK-cos continued replication in this mutant in the presence of Rif and Cm without SOS-inducing treatment. Thus, the attenuation of the nuclease activity of RecBCD is a prerequisite for origin-independent plasmid replication in rec+ sbc+ cells. The copy number increase of pHKχ1-cos in rec+ sbc+ cells was ~50% of that of pHK-cos in recD mutant cells (Figure 4B). This could be explained by the in vivo and in vitro observations that RecBCD has a less than 50% chance of recognizing a particular χ site placed in the active orientation (Yagil and Shtromas, 1985; Stahl et al., 1990; Taylor and Smith, 1992).
The dispensability of SOS induction for origin-independent replication of pHKχ1-cos allowed us to analyze the role of the recombinase activity of RecA in this type of DNA replication. recA mutant cells harboring pFM123 are viable at 30°C (Murialdo, 1988). When pHKχ1-cos and pFM123 were introduced into recA mutant cells, the origin-independent replication of pHKχ1-cos was not observed (Figure 4C). Similarly, the recB and recC mutations blocked the replication (Figure 4C). These results indicate that χ-dependent plasmid replication observed in rec+ sbc+ cells requires the function of the RecBCD pathway of homologous recombination. We designate this type of origin-independent replication RDR.
The RecE pathway of homologous recombination is active in recBC sbcA mutant cells. Since RecBCD is inactive in this strain, attenuation of the nuclease activity of RecBCD by χ or SOS induction was expected to be dispensable for origin-independent plasmid replication. Thus, the copy number increase of pHK-cos in the presence of Rif and Cm was monitored in this strain without SOS-inducing treatment. An autoradiogram of this experiment is shown in Figure 5A. The copy number of ccc monomers increased ~6-fold after a 5 hr incubation (Figure 5C). The increase depends on the presence of pFM123 (Figure 5C). In recBC (sbcA+) (Figure 5D) and recBC sbcA recE (Figure 5B and 5D) mutant cells, in which neither the RecBCD nor RecE pathway of homologous recombination is active, the origin-independent replication of the ccc molecules of pHK-cos was not observed. A recA mutation also greatly reduced the copy number increase of the ccc molecules in recBC sbcA mutant cells (Figure 5D). The residual replication seen in this strain could be accounted for by the activity of RecT protein. RecT is derepressed by sbcA mutations, and some recombination events occurring in recBC sbcA recA mutant cells depend on it (Clark et al., 1993; Hall and Kolodner, 1994, and references therein).
In addition to ccc monomers, LPMs of pHK-cos are also obvious in this strain; ~60% of total plasmid DNA is made up of LPMs (Figure 5A). The copy number of the LPMs increased ~3-fold after a 5 hr incubation in Rif and Cm (Figure 5C). The amount of LPMs was greatly reduced in recBC (data not shown) and recBC sbcA recE (Figure 5B) mutant cells. These results suggest that the efficient formation of templates for rolling circle replication in recBC sbcA mutant cells requires the function of the RecE pathway of homologous recombination. A template for rolling circle replication could also be produced from a circular molecule by introducing a nick (Cohen and Clark, 1986). This mechanism, which does not require homologous recombination functions, could account for the formation of a small amount of LPMs in recombination-deficient cells.
From these results, we conclude that the major part of origin-independent plasmid replication (both the θ mode and the rolling circle mode of DNA synthesis) observed in recBC sbcA mutant cells is also RDR, which depends on the function of the RecE pathway of homologous recombination.
As shown in Figure 6A, the copy number increase of pHKχ1-cos was greatly reduced by a recNTn5 mutation. This indicates that RecN is required for DSB-triggered RDR in recB+ recC+ sbcA+ cells. In contrast, the recN mutation had no effect on RDR when cells were induced for the SOS response (Figure 6B). In this case, induction of some factor(s) might have rendered RDR independent of RecN.
Homologous recombination-dependent DNA replication is essential for bacteriophage T4 genome replication (Mosig, 1983). When replication forks started from origins reach the ends of T4 linear DNA, the 3′ ends of parental DNA remain unreplicated. The resulting ssDNA end invades a homologous region of another or the same T4 DNA molecule by the action of T4 recombinase, UvsX, and primes the leading-strand synthesis. A similar reaction has been reconstituted in vitro using purified T4 recombination and replication proteins (Formosa and Alberts, 1986; for review see Kreuzer and Morrical, 1994). In this study we have demonstrated that plasmid DNA replication can be initiated in normal wild-type E. coli cells in a manner that is independent of plasmid replication origin but is absolutely dependent on E. coli homologous recombination functions. This RDR is promoted by either the RecBCD or RecE homologous recombination pathway (for the RecF pathway, see below). RDR is observed only when a plasmid contains a λ cos site and cells express λ terminase. The in vivo generation of DSBs at the cos sites by λ terminase has clearly been demonstrated in a similar system (Murialdo and Fife, 1984). Thus, RDR is most likely to require a duplex DNA end. This is further supported by the result that, in recBC sbcA mutant cells, initiation of RDR depends on the recE gene product exonuclease VIII (Figure 5), which preferentially acts on duplex DNA ends and appears to be unable to initiate degradation from single-stranded breaks (Joseph and Kolodner, 1983). We conclude that DSBs can trigger extensive DNA replication that depends on homologous recombination functions.
The major product of RDR in rec+ sbc+ cells is ccc monomers (Figure 3A and Figure 4A). This suggests that RDR proceeds preferentially by the θ mode of DNA synthesis. As predicted from our model (Figure 2), LPMs produced by the rolling circle mode of DNA synthesis are also detected in samples prepared from SOS-induced cells (Figure 3A). In contrast, LPMs are undetectable in uninduced cells (Figure 4A). This is probably because terminase molecules eventually leave the cleaved cos sites. χ sites on the LPMs cannot protect them from digestion when RecBCD molecules enter from the ends of the multimers (see Figure 2). In recBC sbcA mutant cells, a significant amount of plasmid DNA replicates as LPMs (Figures 5A and 5C) owing to the absence of the RecBCD nuclease activity. Importantly, the formation of LPMs largely depends on the RecE pathway of homologous recombination (Figures 5B and 5D). Thus, the rolling circle mode of DNA synthesis observed in this mutant is also likely to be RDR. It has been demonstrated that LPMs are efficiently generated in recBC sbcBC mutant cells in a manner dependent on the RecF pathway of homologous recombination (Silberstein and Cohen, 1987; Kusano et al., 1989). The multimer formation does not require the activity of plasmid replication origin (Silberstein and Cohen, 1987), suggesting that RDR is also induced in these mutant cells.
When a DSB is introduced into a circular DNA molecule in normal wild-type cells, RecBCD enzyme would bind to each end of the linearized molecule and begin unwinding it, with simultaneous degradation of the unwound ssDNA. When the enzyme encounters a properly oriented χ site on the linear molecule, the nuclease activity is attenuated, and thus ssDNA produced by the subsequent helicase action of the enzyme remains undigested. This ssDNA would be assimilated into an intact homolog by the action of RecA recombinase. This reaction, i.e., D loop formation, could accomplish initial duplex opening, which is a prerequisite for replication initiation of duplex DNA molecules (Figure 7A). In normal wild-type cells, therefore, DSB-triggered DNA replication, RDR, could occur preferentially at or downstream of χ sites. When a DSB occurs in recD mutant or SOS-induced wild-type cells, in which the nuclease activity of RecBCD is inactivated, ssDNA generated by RecBC(D) helicase would remain intact unless other ssDNA-specific nucleases destroy it. Therefore, in these cells, D loop formation and subsequent initiation of DNA replication could occur at sites independent of χ. In place of RecBC(D) helicase, RecQ helicase, together with RecJ exonuclease, might also produce invasive ssDNA independent of χ. In recBC sbcA mutant cells, exonuclease VIII would recognize the ends of linear DNA and produce ssDNA with 3′ ends that are utilized by RecA. RDR thus could also occur in this mutant in a manner independent of χ.
The crucial step in replication initiation following duplex unwinding is the loading of the replication fork helicase, a DnaB hexamer, onto ssDNA in the unwound region (for reviews see Kornberg and Baker, 1992; Marians, 1992). In RDR, D loop formation could expose a signal for DnaB loading (e.g., n′-pas) on the displaced ssDNA (Figure 7A). DnaG primase would then interact with DnaB and synthesize an RNA primer on the ssDNA (Figure 7B). DNA synthesis from the primer by DNA polymerase III holoenzyme would reach to the end of the unwound region (Figure 7B). Translocation of DnaB in the 5′ to 3′ direction on the ssDNA would result in the formation of a replication fork that leads to unidirectional replication (Figure 7C). For bidirectional replication, loading of another DnaB hexamer on the other strand is necessary. The priming mechanism at D loops formed at oriM for iSDR is most likely to involve PriA protein, which binds to n′-pas and delivers DnaB through the formation of the X-type primosome (H. Masai, T. A., Y. Kubota, K. Arai, and T. K., submitted). The extreme similarities between the two replication systems (see Introduction and below) make it likely that a PriA-dependent priming mechanism might also accomplish DnaB loading at D loops for RDR.
Genetic requirements for the initiation of iSDR, which replicates the entire E. coli genome, are very similar to those for RDR induction described here in the plasmid system. Thus, the initiation mechanisms of these replication systems are most likely to share common features. iSDR induction in recD and recBC sbcA mutant cells requires an SOS-inducing treatment (Asai et al., 1993), although the nuclease activity of RecBCD is absent in these mutants. In this study we have shown that, when DSBs are artificially provided in these mutant cells, RDR is initiated without SOS-inducing treatments. It follows, therefore, that SOS induction is necessary for iSDR not only to attenuate the RecBCD nuclease but also to generate DSBs at oriMs. Although initiated in similar manners, RDR and iSDR are distinct from each other in that the former starts from any site on the chromosome where a D loop can be formed, whereas the latter is restricted to specific sites on the chromosome. We envisage that iSDR is a special type of DSB-triggered RDR that occurs exclusively at oriMs.
Repair of DSBs is critical for cell survival (see Smith, 1991; Cox, 1993, and references therein). DSBs in intracellular DNA are created by the action of various physical and chemical agents. DSBs also result from enzymatic activities during the repair of DNA damage. Moreover, a single-stranded break can readily be converted to a DSB when a replication fork encounters it (see below). Since single-stranded breaks can be generated by reactive oxygen species, the chromosome of aerobically growing cells would suffer DSBs constantly, even in the absence of external DNA-damaging agents. DSB repair may be also involved in meiotic recombination in yeast. Various observations obtained with yeast strongly suggest that meiosis-specific DSBs occurring transiently at many locations in the genome initiate meiotic gene conversion and crossing over (Sun et al., 1989; Bishop et al., 1992; Wu and Lichten, 1994). DSB repair requires homologous recombination functions (Krasin and Hutchinson, 1977; Bishop et al., 1992; Shinohara et al., 1992). In fact, in the absence of the functions, a single DSB in the E. coli chromosome is lethal (Murialdo, 1988). The current leading models for DSB repair proposed for both yeast and E. coli systems (Thaler and Stahl, 1988; Sun et al., 1991; Kobayashi, 1992), therefore, emphasize the role of homologous recombination in the process. The possible involvement of extensive DNA replication in DSB repair has not been addressed.
The minimal requirement for RDR in wild-type cells, besides homologous recombination functions, is a DSB (or a duplex end), a homolog, and attenuation of the RecBCD nuclease. Since the χ sequence is frequently found in the E. coli chromosome (about once every 5 kb on the average; Faulds et al., 1979) and actively growing cells normally contain more than one replicating chromosome, it is quite reasonable to expect that RDR can be manifested when a DSB is introduced into the chromosome. This expectation immediately suggests one function of RDR: DSB repair. We wish to propose the following mechanism for DSB repair, which involves extensive DNA replication. When a chromosome suffers a DSB (Figure 8, step B), RecBCD enzyme begins to degrade DNA from both ends, converting the DSB to a double-stranded gap (step C). Upon encountering χ, the enzyme starts producing ssDNA (Figure 8, step D). Invasion of 3′-ended ssDNA strands in the respective regions of an intact homolog yields a structure (Figure 8, step E) that contains two D loops. Loading of a DnaB hexamer and subsequent priming at each D loop produce two replication forks that proceed toward each other to repair the gap. When the two replication forks meet and replication is completed, the two homologs are now joined by two Holliday junctions (Figure 8, step F). Resolution of the junctions yields two monomers (Figure 8, step A) or one cointegrate (step G), depending on whether cutting occurs in the same sense (i.e., a north–south or east–west cutting at both junctions) or in the opposite sense, respectively. The cointegrate can be resolved into two monomers. The speculation that the ends generated by DSBs may trigger DNA replication leading to DSB repair has been presented (Thaler and Stahl, 1988; Smith, 1991).
The efficiency of DSB repair in wild-type cells is dramatically reduced by recN mutations (Picksley et al., 1984; Sargentini and Smith, 1986). Consistent with our proposal that RDR is involved in DSB repair, we have found that a recN mutation greatly reduces RDR in rec+ sbc+ cells. In contrast, double-stranded gap repair in recBC sbcA mutant cells studied by Kobayashi and coworkers does not require RecN (Kusano et al., 1993). The requirement of RecN for RDR in recBC sbcA mutant cells remains to be tested.
The feature described in the model seems ideally suited for restoration of a replication fork that is destroyed because of a DSB (Figure 9). When a replication fork encounters a single-stranded break, a DSB should result and the replication fork would be destroyed (Figure 9A). However, the end of the chromosome arm that is broken off could be recognized by RecBCD and the arm would be degraded up to the nearest χ site (Figure 9B). Formation of a D loop at or downstream of the χ site and subsequent loading of DnaB helicase followed by replication protein assembly should effectively restore the replication fork (Figures 9C and 9D). Finally, resolution of a Holliday junction would regenerate the original chromosome structure (Figure 9E). A consideration on the distribution and orientation of χ sites on the E. coli chromosome has led to speculation about the possible involvement of χ sites in the repair of collapsed replication forks (Kuzminov et al., 1994).
Reciprocal recombination between a recipient chromosome and a linear duplex fragment introduced into a recipient cell by Hfr conjugation or P1 transduction would regenerate a duplex fragment that could recombine with the resultant recombinant chromosome. This endless recombination event, as proposed by Smith (1991), can be prevented if the ends of the initially introduced fragment prime DNA replication on the recipient chromosome, yielding a “recombinant” chromosome. Our results strongly support this novel mechanism of homologous recombination. We have recently found that a priAkan mutation reduces the frequency of P1 transduction drastically and the recombination frequency after Hfr conjugation moderately both in wild-type and recBC sbcA mutant cells (T. K., G. W. Cad-well, and T. A., unpublished data). These results are also consistent with the idea that RDR is involved in homologous recombination.
Unless otherwise stated, cells were grown at 30°C, with aeration by shaking, in M9 salts–glucose medium (Miller, 1972) supplemented with Casamino acids (0.2%; Difco Laboratories, Detroit, Ml), required amino acids (50 µg/ml), and thiamine hydrochloride (2 µg/ml) (M9-CAA medium). For thyA deoB (or deoC) mutant cells, thymine (8 µg/ml) was also added. For the selection of plasmid-containing cells, antibiotics were added at the following concentrations: ampicillin, 40 µg/ml; kanamycin, 50 µg/ml: spectinomycin, 40 µg/ml.
The E. coli strains used in this study are listed in Table 1. Strains were constructed by phage P1-mediated transduction. The ΔrecA306, lacZU118, and thyA mutations were cotransduced with the srlCTn10, lacl42Tn10, and zgaTn10 mutations, respectively. Transductants carrying the lacZU118 mutation were detected as white colonies on plates containing X-Gal and IPTG. A “low thymine” strain (AQ9717) was obtained from a thyA mutant by a spontaneous mutation in the deo operon (Miller, 1972). λp[sfiAlacZ+ cl(lnd−)] was previously described (Huisman and D’Ari, 1983) and provided by S. Sommer.
The Pstl fragment (~450 bp) of pMUA10 (Meyerowitz et al., 1980) contains the λ Hincll fragment (Hincll [coordinate 48298]–cosN (48502/1]–Hincll ) carrying a cos site. pHK-cos (Figure 1) was constructed by inserting the Pstl fragment into the Pstl site of pHK (Asai et al., 1994).
The following oligonucleotides (TA103 and TA106) containing a χ sequence and their complementary fragments (TA104 and TA105, respectively) were used to construct χ-containing plasmids: TA103, 5′-AGCTTAGCGCTGGTGGCTGCA-3′; TA104, 5′-GCCACCAGCG-CTA-3′, TA105, 5′-AGCTTAGCCCACCAGCCTGCA-3′; TA106, 5′-GGC-TGGTGGGCTA-3′. Construction of pHKχ1-cos (Figure 1) was accomplished by first annealing TA103 and 104. One end of the resultant double-stranded DNA fragment carries the ssDNA tail compatible with a Hindlll site, and the other end is compatible with a Pstl site. The fragment was inserted between the Hindlll and the Pstl sites of pUC19, generating pUC19-χ. A Hindlll fragment (2014 bp) that contains the Spc′ gene of R100.1 (Ω fragment; Prentki and Krisch, 1984) was then inserted, after treatment with Klenow enzyme, into the Hincll site of pUC19-χ, generating pUC19-χΩ. The Ω fragment does not contain χ. Finally, the EcoRI-digested pHK-cos fragment and the Hindlll and BamHI fragment of pUC19-χΩ containing χ and Ω were ligated after treating the fragments with Klenow enzyme. One of the two possible constructs, which contains the χ site in the active orientation, was named pHKχ1-cos. pHKχ3-cos (Figure 1), carrying a χ site in the inactive orientation, was constructed by the same procedure with oligomers TA105 and 106.
The copy number increase of pHK and its derivatives in the presence of Rif and Cm were measured as described in the text and figure legends.
Extraction of total DNA and Southern blot hybridization were carried out as described previously (Magee et al., 1992; Asai et al., 1993). The BamHI fragment (1.7 kb) of pHK carrying the kanamycin gene was used to prepare 32P-labeled probe for detection of pHK and its derivatives. The intensity of 32P radioactivity in each DNA band was determined with a phosphorimager (425E-120, Molecular Dynamics, Sunnyvale, CA). Open circular plasmid molecules were generated in the course of preparation of total DNA. When a significant amount of open circular DNA was detected, the intensity of 32P radioadivity in each open circular DNA band was also measured and added to that of the corresponding ccc DNA. The averages of standard error of means of copy number determinations at each sampling time were obtained from a number of independent experiments. They are as follows: time 0 sample, ~8%; time 1 sample, ~15%; time 2 and 3 samples, ~10%; time 4 and 5 samples, ~7%.
We are grateful to S. Sommer for preparing λsfiAlacZ+ phage, to N. Tsurushita for synthesizing DNA oligomers, and to A. J. Clark, R. D. Kolodner, H. Murialdo, and R. S. Myers for sending us E. coli strains and plasmids. We are also grateful to K. N. Kreuzer for helpful discussions and to E. Boye, S. C. Kowalczykowski, A. Kuzminov, T. R. Magee, R. S. Myers, G. R. Smith, and F. W. Stahl for critical reading of the manuscript. T. A. is especially thankful to C. L. Squires for her support in finishing this work. This work was supported by Public Health Service grant GM22092 from the National Institutes of Health and by National Science Foundation grant BIR-9218818.