Replication mutations differentially enhance RecA-dependent and RecA-independent recombination between tandem repeats in Bacillus subtilis

Introduction

Tandemly repeated sequences, ranging in length from a few nucleotides to several kilobasepairs, are found in the genomes of many organisms. Recombination between these repeats can lead to the gain (amplification) or loss (deletion) of genetic information. Rearrangements of this type are of particular interest, as they are involved in several human genetic diseases (for a review, see Djian, 1998).

In Escherichia coli, recombination between homologous sequences is dependent on the RecA protein, which catalyses homologous pairing and strand exchange between DNA molecules (for a review, see Bianco et al., 1998). However, recombination between 100 to 800 bp long tandem direct repeats occurs via RecA-dependent and RecA-independent mechanisms in E. coli, and both can be stimulated independently in different genetic contexts (Lovett et al., 1993; Bierne et al., 1997a; Morag et al., 1999). Thus, RecA-dependent deletions are stimulated by mutations in the DnaB and UvrD helicases (Bierne et al., 1997b; Saveson and Lovett, 1997), whereas RecA-independent deletions are stimulated in mutants of the Rep helicase and of the α (DnaE) and ε (DnaQ) subunits of the DNA polymerase III (Bierne et al., 1997a; Saveson and Lovett, 1997).

Two main routes of RecA-dependent recombination have been described in wild-type E. coli cells, the RecBCD pathway, acting at double-strand breaks, and the RecF pathway, acting mainly at single-strand gaps. It is now thought that the main physiological role of these recombination pathways is to reinitiate replication of forks that have been inactivated by collapse or the encounter of a road block (for recent reviews, see Cox et al., 2000; Kowalczykowski, 2000; Michel, 2000). Stimulation of tandem repeat deletion in dnaB mutant cells was shown to be RecBCD dependent and RecFOR independent (Saveson and Lovett, 1999). This suggests that the dnaB mutation blocks replication forks, which are then subjected to breakage, either directly or after reversal (see Fig. 1), followed by double-strand break repair. During this process, misalignment of the repeats by RecA could lead to the loss of one of the repeats (Fig. 1). In contrast, stimulation of recombination in the uvrD mutant was RecFOR dependent and required induction of the SOS response (Bierne et al., 1997b). It is assumed that, in this mutant, recombination via a gap repair mechanism is stimulated. As in double-strand break repair, mispairing of the repeats by RecA during the process could lead to deletion of one of the repeats (Fig. 2).

In contrast, RecA-independent stimulation of tandem repeat recombination in dnaE, dnaQ and rep mutants was proposed to occur by slipped mispairing of the repeats during DNA replication (Lovett et al., 1993; Bierne et al., 1997a; Saveson and Lovett, 1997), a mechanism supported by in vivo and in vitro experiments (d'Alençon et al., 1994; Canceill and Ehrlich, 1996). According to this model, nascent DNA containing a repeat transiently separates from its template during replication and reanneals with the second repeat, thus generating a deletion in the newly synthesized DNA (Fig. 3). Reannealing with the second repeat can occur on the same arm of the fork (replication slippage model) or on the opposite arm (sister chromomatid exchange model; Lovett et al., 1993). Biochemical analysis of plasmid recombination products suggested the existence of both mechanisms (Saveson and Lovett, 1997; Bzymek et al., 1999).

In the Gram-positive bacterium Bacillus subtilis, amplification of genes as tandem repeat arrays is used to overproduce proteins of industrial interest (see Young and Ehrlich, 1989 and references therein). Tandem repeat recombination is thus important for both production of the amplified structure and its subsequent stability. However, mechanisms of tandem repeat recombination have not yet been investigated in detail. Nevertheless, it is likely that recombination and replication are strongly linked in B. subtilis, as insertion of an actively replicating plasmid in the vicinity of direct repeats in the B. subtilis chromosome stimulates their recombination (Noirot et al., 1987; Petit et al., 1992; Morel-Deville and Ehrlich, 1996).

Here, we report that recombination between tandem repeats is stimulated by several mutations affecting DNA replication in B. subtilis, including mutations of primosomal proteins (DnaB, DnaD and the DnaG primase), mutations of putative components of the DNA polymerase III holoenzyme (DnaN, DnaX and PolC), as well as a mutation in a DNA polymerase of unknown function (DnaE). Recombination does not require RecA in wild-type cells as well as in the primosome mutants that could be tested (dnaD and dnaG). In contrast, RecA is required in three other mutants (dnaN, dnaX and dnaE). We conclude that both RecA-dependent and RecA-independent recombination between tandem repeats takes place in B. subtilis.

Results

Tandem repeat recombination occurs in the absence of RecA

To measure recombination between tandem repeats in B. subtilis, a kanamycin resistance (Kn^R) gene carrying an internal 513 bp direct repeat (Kn_dup513) was introduced in the chromosome near the thyB locus (Fig. 4; for details of construction, see Experimental procedures). The repeats inactivate the Kn^R gene, and cells are kanamycin sensitive (Kn^S). Deletion of one of the repeats restores an intact gene and thus confers a Kn^R phenotype to the cells (Fig. 4). The recombination frequency was therefore estimated by measuring the proportion of Kn^R cells in several independent growing cultures (see Experimental procedures).

Recombinant proportions were very similar in two wild-type strains and were not affected significantly by the presence of a recA mutation (Table 1). We conclude that tandem repeat recombination can occur independently of RecA in B. subtilis. It should be noted that some recA mutants had low plating efficiency on rich medium (not shown). We therefore carried out all experiments with recA strains on minimal medium. The wild-type cells had a slightly higher recombination frequency (≈ threefold) on this medium than on rich medium, for unknown reasons.

Table 1. Recombinant proportion is not significantly affected by a recA mutation.

Straina	Relevant genotype	Recombinant proportion (× 104)b	Relative valuec	n
A.
CBB350	rec+	5.7 (± 2.5)	= 1	5
CBB500	recA::tet	3.0 (± 2.2)	0.5	5
B.
CBB383	rec+	3.9 (± 1.6)	= 1	5
CBB506	recA::tet	1.9 (± 0.5)	0.5	5

• a. Subsections A and B correspond to different genetic backgrounds.
  
• b. Recombinant proportions were measured as described in Experimental procedures in minimal medium at 37°C. Average and standard deviation of n independent determinations are shown.
  
• c . Recombinant proportion is expressed relative to that determined for the isogenic wild-type strain, taken as 1.

Discussion

In this work, we studied recombination between 513 bp tandem direct repeats in the B. subtilis chromosome. Wild-type cells showed a recombinant proportion of around 2 × 10⁻⁴, and a recA mutation did not significantly affect this value (Table 1). Recombination was stimulated by mutations in various replication proteins, including the primosomal proteins DnaB, DnaD and the DnaG primase (Table 2), the DNA polymerase III subunits PolC, DnaX and DnaN and the DNA polymerase DnaE (Table 3). No effect was observed in the priA1, polA5, dnaC14 and dnaI2 mutants. RecA dependence of recombination stimulation was examined in some of the mutants that showed a hyper-recombination phenotype. Mutations in the dnaG and dnaD genes stimulated recombination independently of RecA, whereas mutations in the dnaE, dnaX and dnaN genes stimulated recombination in a RecA-dependent manner (Table 4). The effect of recA on stimulation by dnaB and polC mutations could not be determined, as the double mutants were not viable under conditions in which hyper-recombination was observed. These results show that both RecA-dependent and RecA-independent mechanisms can contribute to recombination between tandem repeats in B. subtilis and that each is stimulated by alterations in the DNA replication process.

The B. subtilis dnaX51 and dnaN5 mutations stimulate recombination, which contrasts with data obtained previously in E. coli, in which mutations in the γ–τ- and β-subunits (encoded by the dnaX and dnaN genes respectively) decrease tandem repeat recombination (Saveson and Lovett, 1997). These differences might result from the fact that the mutations tested affect different activities of the proteins in E. coli and B. subtilis. Alternatively, the proteins assumed to be equivalent could play different functions in the two hosts. Thus, dnaX encodes only the τ-subunit in Gram-positive bacteria, whereas it also encodes the γ-subunit in E. coli by translational frameshifting (Bruck and O'Donnell, 2000). Mutations in the α- and ε-subunits of the E. coli DNA polymerase III, encoded by the dnaE and dnaQ genes, respectively, stimulate tandem repeat deletion independently of RecA (Bierne et al., 1997a; Saveson and Lovett, 1997). It is thought that the B. subtilis counterparts of the E. coliα- and ε-subunits are encoded by the same polypeptide, the product of the polC gene (Sanjanwala and Ganesan, 1991). We found that the polC33 mutation stimulated recombination, but could not test the RecA dependence of hyper-recombination in this mutant. The dnaE gene encodes a protein highly similar to the E. coliα-subunit, but its role in DNA replication is currently unknown (Bruck and O'Donnell, 2000). However, it should be noted that the B. subtilis dnaE mutation, in contrast to E. coli dnaE mutations, stimulated recombination in a RecA-dependent manner.

The RecA-dependent mechanisms of stimulation of recombination in the B. subtilis polymerase mutants are not known. Nevertheless, like E. coli, B. subtilis possesses several pathways of RecA-dependent recombination (for a recent review, see Fernández et al., 2000). The counterpart of the RecBCD enzyme of E. coli has been shown, in vivo and in vitro, to be encoded by the B. subtilis addAB operon (Chédin et al., 2000 and references therein), whereas the counterpart of the RecFOR system of E. coli would be encoded by the recFLOR genes (Fernández et al., 1999). Therefore, the models proposed previously to explain RecA-dependent homologous recombination between tandem repeats in E. coli (see Introduction) could apply to B. subtilis as well: mutations in the DNA polymerase subunits could possibly block replication forks, leading to breakage and AddAB-dependent double-strand break repair, or leave small gaps of unreplicated DNA that are repaired by a RecFLOR-dependent mechanism. During these processes, mispairing of the repeats by RecA leads to the deletion of one repeat.

In addition to its roles in recombination, RecA is involved in the induction of the SOS response in B. subtilis as in E. coli (for reviews, see Yasbin et al., 1993; Walker, 1996). Accumulation of single-stranded DNA as a possible consequence of some of the replication defects might induce, via RecA, the SOS response. In E. coli as in B. subtilis, some genes involved in RecA-dependent recombination, in particular RecA itself, are induced during the SOS response, and this probably contributes to stimulate the recombination processes described above. Additionally, several error-prone polymerases involved in DNA repair are activated during the SOS response in E. coli (for reviews, see Goodman, 2000; Goodman and Tippin, 2000). These polymerases could be involved in recombination through slipped mispairing of repeated sequences during DNA synthesis. If such polymerases are present in B. subtilis, they might be responsible for some recombination events. Therefore, further genetic studies are needed to elucidate the different role(s) played by RecA in the process of stimulation of recombination in replication mutants.

In contrast to polymerase mutants, the dnaG20 and dnaD23 mutations stimulated recombination in a RecA-independent manner. In this case, slipped-mispairing during replication is likely to be the appropriate model, because it need not involve RecA at any stage (see Introduction). An alternative is the single-strand annealing model, in which DNA is subject to breakage in one of the repeats. The free DNA ends could then be degraded by exonucleases and generate single-stranded regions containing the repeats. Erroneous annealing of complementary repeats in the absence of any recombination function generates the deletion of one of the repeats. Both these models were proposed previously in bacteria to account for RecA-independent recombination between short homologous sequences (for a review, see Michel, 1999).

The dnaG20 mutation lies in the polymerase domain of the primase (Price and Doi, 1985; Wang et al., 1985; Keck et al., 2000) and therefore probably affects the efficiency of primer synthesis. In an in vitro reconstituted E. coli replication system, decreasing the efficiency of primer synthesis resulted in longer Okasaki fragments, which reflect an increase in the distance between consecutive primers on the lagging strand (Wu et al., 1992; Zechner et al., 1992). If the dnaG20 mutation has the same effect in B. subtilis in vivo, it might result in large single-stranded regions on the lagging strand template, thus increasing the probability that both repeats are exposed in a single-stranded form at the same time, a prerequisite for slipped-mispairing. An alternative possibility is that these large regions of single-stranded DNA are susceptible to cleavage by cellular endonucleases, thus favouring the single-strand annealing model. In E. coli, the primase mutant dnaG3903 did not show any stimulation of tandem repeat recombination (Saveson and Lovett, 1999). However, as suggested by the authors, this could result from the fact that this mutant is not affected in primer synthesis but is rather a partition mutant, whereas the B. subtilis dnaG20 mutant tested here was isolated as a replication mutation (Karamata and Gross, 1970).

The RecA-independent stimulation of tandem repeat deletion in the dnaD23 mutant is more difficult to interpret. In B. subtilis, DnaD is involved, together with DnaB and DnaI, in the loading of the DnaC helicase onto DNA, either in a DnaA-dependent manner during initiation of replication at oriC (for a review, see Moriya et al., 1999) or in a PriA-dependent manner during reinitiation of inactivated replication forks (Bruand et al., 1995b; P. Polard and C. Bruand, unpublished). The initiation defect at oriC is unlikely to be responsible for the observed hyper-recombination phenotype. However, the priA1 mutation did not stimulate recombination, which shows that the defect in reinitiation of inactivated forks is not primarily responsible for hyper-recombination. What could be the reason for the stimulation of recombination in the dnaD23 and dnaB19 mutants, but not in the priA1 mutant? The DnaB and DnaD proteins have single-stranded DNA-binding activities (S. Marsin, S. McGovern, C. Bruand and P. Polard, unpublished) and could thus protect single-stranded DNA against nucleases or impair DNA polymerase slippage and, consequently, prevent recombination. Another possibility is that, in the absence of DnaD or DnaB, PriA binding to inactivated forks does not lead to primosome assembly. Instead, in these conditions, if the helicase activity of PriA is activated, the newly synthesized lagging strand may be unwound. Such an activity has been observed in vitro on forked DNA substrates with the E. coli PriA protein in the absence of one or more other primosomal proteins (McGlynn et al., 1997; Jones and Nakai, 1999). This unwinding would generate large regions of single-stranded DNA on the lagging-strand template, a situation similar to that described above in the dnaG20 mutant. Although the RecA dependence of recombination in the dnaB19 mutant could not be tested, we predict that hyper-recombination in this mutant would be as in dnaD23, i.e. RecA independent. Interestingly, the priA mutation in E. coli stimulated tandem repeat recombination 10-fold (Saveson and Lovett, 1997). It has been shown that the loading of the replication fork helicase in E. coli and B. subtilis occurs through different mechanisms (Bruand et al., 1995b; Sandler, 2000; C. Bruand, unpublished). This could be the source of the different recombination phenotypes of E. coli and B. subtilis priA mutants.

No stimulation of recombination was observed in the dnaC14 mutant, which encodes an elongation-deficient replication fork helicase (Gross et al., 1968; Sakamoto et al., 1995). This is rather surprising as, in E. coli, the dnaB107 mutant showed a stimulation of tandem repeat deletion. However, another elongation-deficient allele of E. coli dnaB did not (Saveson and Lovett, 1999). Such an allele specificity could apply to the dnaC mutants of B. subtilis as well, and dnaC14 would not be a hyper-recombinogenic allele.

Experimental procedures

Acknowledgements

We thank E. Dervyn for providing the dnaE1 mutant, A. Schmautz for technical help, M.-A. Petit, B. Michel, P. Polard and S. Marsin for stimulating discussions and/or critical reading of the manuscript, and P. Bellenand for help with statistical analyses of some results. This work was supported, in part, by a grant from the European Commission (BIO4-CT98-0250).