The RdgC protein of Escherichia coli binds DNA and counters a toxic effect of RecFOR in strains lacking the replication restart protein PriA

Abstract

PriA protein provides a means to load the DnaB replicative helicase at DNA replication fork and D loop structures, and is therefore a key factor in the rescue of stalled or broken forks and subsequent replication restart. We show that the nucleoid-associated RdgC protein binds non-specifically to single-stranded (ss) DNA and double-stranded DNA. It is also essential for growth of a strain lacking PriA, indicating that it might affect replication fork progression or fork rescue. dnaC suppressors of priA overcome this inviability, especially when RecF, RecO or RecR is inactivated, indicating that RdgC avoids or counters a toxic effect of these proteins. Mutations modifying ssDNA-binding (SSB) protein also negate this toxic effect, suggesting that the toxicity reflects inappropriate loading of RecA on SSB-coated ssDNA, leading to excessive or untimely RecA activity. We suggest that binding of RdgC to DNA limits RecA loading, avoiding problems at replication forks that would otherwise require PriA to promote replication restart. Mutations in RNA polymerase also reduce the toxic effect of RecFOR, providing a further link between DNA replication, transcription and repair.

Introduction

The rdgC gene of Escherichia coli encodes a 34 kDa protein associated with the nucleoid (Ryder et al., 1996; Murphy et al., 1999). Its deletion has no obvious effect except in a nuclease-deficient recBC sbcBC mutant background where it confers a recombination dependent growth phenotype (hence rdg). Thus, a ΔrdgC recBC sbcBC strain is only viable provided the RecA and RecF proteins necessary for recombination in this background are functional (Ryder et al., 1996). Given recombination underpins genome replication (Cox et al., 2000), this raises the possibility that RdgC may normally aid replication fork progression. Replication forks assembled at oriC are inherently processive, but the current view is that their progress is often blocked by lesions in or on the template DNA, with estimates ranging from 10 to 50% of cells requiring some form of fork reactivation (Cox et al., 2000; McGlynn and Lloyd, 2002b). The problem becomes acute in cells exposed to a DNA-damaging agent such as UV light and is complicated by the stalling of transcription complexes at non-pairing lesions in the template strand, creating further obstacles to fork progression (Hanawalt et al., 1994; McGlynn and Lloyd, 2000; van den Boom et al., 2002).

Recent studies in E.coli have suggested that stalled forks might unwind so that the parental strands reanneal and the nascent daughter strands anneal to form a Holliday junction (Seigneur et al., 1998). This may occur spontaneously via release of positive supercoiling ahead of the fork, but is more likely catalysed either by RecG helicase or via the strand exchange activity of RecA, aided perhaps by the RecF, RecO and RecR proteins (Robu et al., 2001; Singleton et al., 2001; McGlynn and Lloyd, 2002a,b). Once formed, the Holliday junction may be driven further from the lesion by the RuvAB branch migration complex (Seigneur et al., 1998). Such reactions would require the replisome complex to have dissociated, which raises the question of how DNA synthesis might subsequently resume. Furthermore, the offending lesion has to be repaired or bypassed for this renewed synthesis to continue without mishap. Fork reversal and Holliday junction formation provide possible solutions to these problems. Backing away from the block may create room for repair, although the timing of repair is still unknown and may depend on how replication is resumed. Creating a Holliday junction provides a substrate that recombination enzymes can exploit to promote restart. At least two general models have been proposed. Both rely on the primosome assembly factor PriA to load the DnaB replicative helicase (Liu and Marians, 1999; Liu et al., 1999; Marians, 2000; Sandler and Marians, 2000), but employ two different DNA structures for this purpose. The first relies on direct restoration of a (corrected) fork structure, the second on recombination to first form a D loop, which is then converted to a fork (Seigneur et al., 1998; Gregg et al., 2002; McGlynn and Lloyd, 2002a,b). The latter may sometimes involve fork breakage, possibly via RuvABC-mediated cleavage of a Holliday junction formed during fork reversal. However, Courcelle and Hanawalt have proposed alternative models for direct fork restoration that do not involve formation of a Holliday junction (Courcelle and Hanawalt, 1999, 2001).

Both direct and indirect models of fork rescue emphasize the critical role of PriA. This is consistent with the fact that inactivation of PriA results in reduced cell viability, defective cell division, sensitivity to DNA damage, chronic SOS induction and recombination deficiency (Nurse et al., 1991; Kogoma et al., 1996). PriA initiates assembly of a primosome at replication fork and D loop structures via a series of defined protein–protein interactions involving PriB, DnaT and possibly PriC, culminating in transfer of DnaB from a DnaB–DnaC complex to the PriA–DNA complex and subsequent binding of DnaG primase (McGlynn et al., 1997; Liu and Marians, 1999; Liu et al., 1999). DNA polymerase III holoenzyme can then be loaded to form a fully functional replisome capable of coupled leading and lagging strand synthesis.

The interaction of PriA with the DnaB–DnaC complex is normally crucial for DnaB loading at replication fork and at D loop structures. However, certain amino acid substitutions in DnaC can circumvent this requirement, enabling DnaB to be loaded without PriA and suppressing the phenotype of priA null strains (Sandler et al., 1996; Xu and Marians, 2000; Gregg et al., 2002). Such dnaC suppressor mutations accumulate rapidly in cultures of priA null strains because of the very substantial improvement in cell viability (Sandler et al., 1996; Gregg et al., 2002).

In this paper, we show that RdgC is a DNA-binding protein that forms stable complexes with both single-stranded (ss) DNA and double-stranded (ds) DNA, and that it is required for growth of priA null strains and for the ability of dnaC suppressor mutations to improve their viability. We also show that the RecF, RecO and RecR proteins are responsible for the low viability of priA dnaC rdgC mutant strains and that this toxic effect can be eliminated by changes to the ssDNA-binding protein, SSB. The results presented provide new insight into the interactions between DNA replication, recombination and repair, and raise the possibility that nucleoid organization is important in maintaining replication fork progression in rapidly dividing cells.

Results

Distribution of rdgC genes in bacteria

BLAST searches identified rdgC genes in the Beta (e.g. Neisseria sp.) and Gamma (e.g. E.coli) subdivisions of the Proteobacteria, but in no other species (Figure 1). Thus, rdgC appears to have arisen in the lineage leading to the Beta and Gamma Proteobacteria. It is present in all fourteen species of these groups for which a complete genome sequence is available. The encoded protein sequences show a high degree of conservation (alignments not shown), but provide no insight as to their function.

Fig. 1. Evolutionary distribution of rdgC. The tree was generated by alignment of 23S rRNA sequences using Clustal_X. An rdgC-like gene is present in the lineage and species highlighted in bold.

Purification and physical properties of RdgC

The 34 kDa E.coli wild-type RdgC protein was expressed from pGS853 and purified to homogeneity as described in Materials and methods (Figure 2A). Gel filtration indicated that the protein might form a dimer in solution (Figure 2B). Glutaraldehyde cross-linking of the purified protein followed by SDS–PAGE analysis revealed a single cross-linked species with an apparent molecular mass of nearly 100 kDa (Figure 2C). This would be consistent with a trimer of RdgC. However, the absence of any intermediate bands resulting from the cross-linking of only two monomers argues against this possibility. We therefore conclude that native RdgC is a dimer with a molecular mass of 68 kDa, although we cannot exclude the possibility of a trimeric state.

Fig. 2. Purification and physical properties of RdgC. (A) SDS–PAGE gel summarizing the recovery of RdgC at different stages of the purification protocol. Lane (a) molecular weight markers, (b) induced cell lysate, (c) 50–70% ammonium sulfate cut, (d–g) peak fractions from heparin, gel filtration, butyl–Sepharose and Q Sepharose columns, respectively. (B) Elution profile of RdgC during gel filtration. The elution volumes of molecular weight standards are indicated by arrowheads. (C) Glutaraldehyde cross-linking of RdgC. Reactions contained 10 µM RdgC and glutaraldehyde at 0 (a), 2 (b), 8 (c), 32 (d), 128 (e) and 512 µM (f), and were analysed by SDS–PAGE. Molecular markers are in (g). Monomeric (i) and cross-linked (ii) RdgC species are identified.

RdgC binds DNA

Two pieces of evidence suggest RdgC might be a DNA-binding protein. First, it is released from E.coli nucleoids by digestion with DNase I (Murphy et al., 1999). Secondly, during attempts at purification, RdgC was found to bind heparin and dsDNA cellulose columns (data not shown). Ryder et al. (1996) suggested RdgC might be an exonuclease, but no nuclease activity of any kind could be detected with the purified protein using a variety of linear and circular DNA substrates under a range of conditions (data not shown). However, band-shift assays confirmed that RdgC binds DNA. Well-defined complexes were detected with linear ssDNA and dsDNA substrates and with a variety of partial duplex and branched molecules (Figure 3A, panels i–vi). Two complexes are formed with a 61 nucleotide single strand (panel i) and four with a 61 bp linear duplex (panel ii). Although, we do not know the precise stoichiometry of these complexes, the results are consistent with the binding of two and four dimers of RdgC, respectively. Complexes formed with the branched substrates vary in number with length of ssDNA and dsDNA available (panels iii–vi). With linear duplexes, the number of complexes increased in proportion to DNA length, indicating that RdgC binds at any point along the DNA and not just to DNA ends (data not shown). The five distinct complexes detected with the Holliday junction substrate, J12, supports this conclusion. A distributive mode of binding is also supported by the fact that RdgC binds circular plasmid DNA and uniformly protects linear duplex DNA from attack by hydroxyl radicals (data not shown).

Fig. 3. DNA-binding activity of RdgC. (A) Gel assays showing binding of RdgC to linear and branched DNA structures. Binding reactions contained 0.1 nM DNA species, (i) 61 nucleotide ssDNA, (ii) 61 bp dsDNA, (iii) flayed duplex, (iv) three-strand junction, (v) Y-DNA, (vi) Holliday junction J12, and RdgC at 0, 0.5, 5, 50 and 500 nM in lanes a–e, f–j and k–o, respectively. (B) Effect of RecG and RdgC on cleavage of χ DNA by RuvC. Reactions contained 0.05 nM χ junction, 5 nM RuvC monomers (lanes b–i), 1, 10 and 100 nM RecG (lanes c–e), and 1, 10, 100 and 1000 nM RdgC dimers (lanes f–i), as indicated. The χ junction is labelled on all four arms and therefore all four possible products of junction resolution are detected (McGlynn and Lloyd, 2000).

The binding data suggested that RdgC does not bind with higher affinity to branch points in DNA as opposed to linear duplex DNA (Figure 3A; data not shown). To confirm this finding, we investigated whether RdgC could interfere with the cleavage by RuvC resolvase of a Holliday junction structure in which the branch point is flanked by long duplex arms (Figure 3B, lane b). RecG protein, which binds with a high affinity to the branch point, clearly interferes with junction resolution (Figure 3B, lanes c–e). RdgC does not, even when in 400-fold molar excess over RuvC (Figure 3B, lanes f–i), indicating that it does not bind the branch point specifically.

RdgC has a higher apparent binding affinity for dsDNA than for ssDNA (Figure 4A). This is particularly noticeable with short substrates. RdgC binds a 25 nucleotide single strand (Figure 4B, lane c), but the complex is unstable and dissociates on electrophoresis. It has a much higher affinity for a 25 bp linear duplex, forming a single, sharply-defined complex (Figure 4B, lane c). Competition binding studies confirmed that RdgC forms an unstable complex with the 25 nucleotide single strand. Hardly any retarded complex could be detected following the addition of poly[dIdC] (Figure 4B, lanes d–j and C). They also showed that the complexes formed with a 25 bp duplex and a 61 nucleotide single strand were much more stable (Figure 4B, lanes d–j and C). Complexes formed with the 61 bp linear duplex proved very stable, with no detectable dissociation during 2 h on ice.

Fig. 4. Affinity of RdgC for ssDNA and dsDNA. (A) Binding isotherm showing relative affinity of RdgC for a 61 bp linear duplex DNA and a 61 nucleotide single strand. Binding reactions contained 0.1 nM labelled DNA and RdgC at 0.025–410 nM. Data are means of two experiments. (B) Competition binding assays. Reactions contained 0.1 nM of the labelled DNA indicated. RdgC was present at 250 nM (25 nucleotide ssDNA), 25 nM (25 bp dsDNA), 20 nM (61 nucleotide ssDNA) and 2 nM (61 bp dsDNA). The RdgC concentration used in each case was the amount needed to achieve a significant bandshift without formation of a substantial fraction of higher order complexes. After 10 min on ice, 50 ng unlabelled poly[dIdC] competitor DNA was added as indicated and the reactions were kept on ice for a further 0, 5, 10, 30, 60, 90 or 120 min (lanes d–j) before electrophoresis, as described in Materials and methods. Horizontal and vertical lines identify substrate DNA and RdgC–DNA complexes, respectively. (C) Relative stability of RdgC complexes with ssDNA and dsDNA. The data from gels of the type shown in (B) were quantified and the complexes detected in the presence of poly[dIdC] expressed as the percentage of the DNA bound in the absence of competitor (B, lane c in each case). Data are averages of two or more experiments.

RdgC is expressed at a high level in dividing cells

Previous studies indicated that RdgC might be a fairly abundant, nucleoid-associated protein (Murphy et al., 1999). Western blots revealed that RdgC can be detected in extracts of wild-type E.coli cells (Figure 5A, lane b). The polyclonal antibodies used are highly specific for RdgC as no signal is detected in extracts from a ΔrdgC strain (Figure 5A, lane c). This enabled us to accurately measure RdgC in cell extracts at different phases of growth (Figure 5B). RdgC was at its highest level during exponential phase, reaching at its maximum ∼1000 dimers per cell. Its level decreased sharply to ∼50 dimers per cell in stationary phase (Figure 5C). This profile suggests RdgC might function during the period of DNA replication.

Fig. 5. Immunodetection of RdgC and effect of growth phase on expression. (A) Western blots of purified RdgC (6 ng, lane a) and cells extracts from wild-type strain MG1655 (lane b) and ΔrdgC strain N4586 (lane c) probed with anti-RdgC polyclonal antibodies. (B and C) Growth phase expression of RdgC. Strain MG1655 was grown in LB broth for 8 h from an initial A₆₅₀ of 0.02. Samples were taken at intervals and assayed by western blotting for RdgC levels in total cell extracts and by microscopy for total cell numbers. (B) Western blots showing immunodetection of RdgC at the intervals shown. (C) Growth-dependent expression of RdgC. The data from blots of the type shown in (B) were quantified by reference to standard concentrations of pure RdgC. Squares indicate the amount of RdgC (dimers) per cell and triangles the total cell number. Data are averages of two experiments.

Inviability of ΔrdgC priA and its suppression by dnaC mutations

The high expression of RdgC during the period of very active DNA synthesis and its ability to bind DNA are consistent with the previous suggestion by Ryder et al. (1996) that RdgC may have a role in promoting chromosome replication. To investigate whether RdgC promotes replication fork progression, we tried to introduce the priA2 null mutation into a ΔrdgC strain. P1 phage grown on the priA2::Km strain, AG181, was used to transduce strain AB1157 and its ΔrdgC derivative, DIM037, selecting resistance to kanamycin. Although a high number of transductants were obtained with AB1157, none were obtained with DIM037, suggesting that a strain lacking both PriA and RdgC is inviable. Attempts to combine priA2 and ΔrdgC in the MG1655 background also failed.

Given that mutations in dnaC such as dnaC212 or dnaC810 suppress priA (Sandler et al., 1996; Gregg et al., 2002), we tried to transduce ΔrdgC::Tm from strain N4586 to the priA2 dnaC212 strain AG181 and the priA2 dnaC810 strain DIM215, this time selecting resistance to trimethoprim. Transductant colonies were obtained in both cases, although they took several days to appear. However, they could be subcultured, indicating that a priA dnaC ΔrdgC construct is viable. One such clone, DIM063, was kept for further analysis.

Strain DIM063 grows slowly in LB broth. The cells are highly filamentous and have grossly distorted nucleoids (Figure 6E), and only 10–20% are able to form colonies on LB agar, suggesting they have a major problem with chromosome replication and segregation. Filamentation is more extensive than in a priA2 single mutant (Figure 6C). By comparison cells of the priA dnaC212 parent, AG181, and of the ΔrdgC strain DIM037 have close to 100% viability. They generally resemble wild-type cells, although ΔrdgC cells appear a little more elongated and some form short filaments (Figure 6A, B and D). Thus, elimination of RdgC reverses the ability of dnaC212 to suppress the low viability of priA2 cells. Indeed, it exacerbates defects in cell division, which may explain the inability to construct a priA2 ΔrdgC strain. However, it does not restore sensitivity to DNA-damaging agents. Strain DIM063 is hardly more sensitive to UV than the priA2 dnaC212 parent and is certainly much more resistant than a priA2 single mutant (Figure 7C). It grows very weakly on LB agar containing mitomycin C (MC) at 0.5 µg/ml, but this can be attributed to slow growth and reduced cell viability (data not shown).

Fig. 6. Filamentous cell morphology of a ΔrdgC priA2 dnaC212 strain and suppression of filamentation by fgv mutations. Phase-contrast and DAPI images are merged to show nucleoid organization within the cell. The strains shown are (A) AB1157 (priA⁺ dnaC⁺ rdgC⁺), (B) DIM037 (ΔrdgC), (C) DIM070 (priA2), (D) AG181 (priA2 dnaC212), (E) DIM063 (priA2 dnaC212 ΔrdgC), (F) DIM061 (priA2 dnaC212 ΔrdgC rpoB), (G) DIM060 (priA2 dnaC212 ΔrdgC ssb) and (H) DIM064 (priA2 dnaC212 ΔrdgC recO). Details of the rpoB, ssb and recO suppressor mutations are in Table I.

Fig. 7. Suppressors of the slow growth and poor viability of a ΔrdgC priA dnaC strain. (A) Photograph of an LB agar plate streaked with ΔrdgC priA dnaC strain DIM063 and incubated for 60 h at 37°C. Colonies of fast-growing variants are readily visible against the background of small colonies of the parent strain. (B) Colony morphology of purified fast-growing variants of DIM063 incubated for 48 h on LB agar. The strains shown (all ΔrdgC priA2 dnaC212, except AG181) are (i) DIM063 (ΔrdgC priA dnaC parent strain), (ii) DIM057 (unidentified fgv), (iii) DIM060 (ssb), (iv) DIM061 (rpoB), (v) DIM064 (recO), (vi) AG181 (priA dnaC control strain), (vii) DIM089 (recF), (viii) DIM122 (recF143) and (ix) DIM123 (lexA3). Details of the ssb, rpoB, recF and recO suppressor mutations are in Table I. (C) Effect of priA, dnaC and rdgC mutations on sensitivity to UV light. The strains are AB1157 (wt), DIM037 (rdgC), AG181 (priA dnaC), DIM063 (priA dnaC rdgC) and DIM070 (priA). (D) UV sensitivity of derivates of priA dnaC rdgC strain DIM063 carrying suppressors of the slow-growth phenotype. The strains are HP126 (rpoB), DIM062 (ssb), DIM057 (unidentified fgv), HP125 (recO) and DIM089 (recF). Details of the rpoB, ssb, recO and recF mutations are in Table I.

Previous studies revealed that the SOS response is chronically induced in priA2 strains (Nurse et al., 1991), and that this phenotype is alleviated by dnaC suppressors (Sandler, 1996). The highly filamentous morphology of DIM063 cells is indicative of chronic SOS induction, consistent with ΔrdgC reversing the effect of dnaC212 in a priA null background. To investigate this directly, we made constructs carrying lacZ fused to the SOS-inducible sfiA gene and tested SOS expression by measuring β-galactosidase activity. In cultures grown in LB broth to an A₆₅₀ of 0.2, we detected 333 ± 24 units of enzyme activity in the priA2 construct, DIM173, compared with only 39 ± 1 units in the wild type, N5170. Activity was reduced to 83 ± 7 units in a priA2 dnaC212 construct, DIM175, consistent with the suppression of priA, but adding ΔrdgC partially reversed this effect, increasing activity in strain DIM177 to 164 ± 8 units. However, it is significant that this level is <50% of the activity in the priA2 construct. Taken together, these data indicate that chronic SOS induction is a feature of the priA2 dnaC212 ΔrdgC strain DIM063. However, this chronic induction may not be the only factor responsible for the severe growth defects.

Cultures of a priA2 dnaC212 ΔrdgC strain accumulate fast-growing variants

Strain DIM063 forms small colonies on LB agar and these are quickly overtaken by faster-growing variants. After 3 days incubation, these variants stand out against a background of small colonies (Figure 7A). They are stable and retain their fast-growth phenotype on subculture (Figure 7B). Such variants can also be selected directly by plating DIM063 on LB agar containing MC at 0.5 µg/ml, or rifampicin at 20 µg/ml. We isolated 39 independent isolates of these fast-growth variants (fgv). The gross filamentous phenotype of the parent (Figure 6E) is suppressed in all cases. Typical examples are shown in Figure 6F–H. However, phenotypic and genetic analyses indicated that they fall into at least four different classes. Table I lists examples that we have studied in some detail. All retain the priA2, dnaC212 and ΔrdgC mutations (data not shown), indicating the presence in each case of an additional suppressor mutation. The high frequency with which these suppressors arise suggests that a single suppressor mutation is responsible. This is supported by our mapping of the fgv mutations, which in each case showed that the slow growth phenotype of the parental strain could be fully restored by introducing the wild-type allele for the mutated gene identified.

Table I. Isolation and characterization of fast-growth variants (fgv) of strain DIM063^a.

Strain	Selection	fgv allele	Gene mutation	Protein alteration	UV survivalb
DIM063					0.064
DIM089	LB	fgv-010	recFT451A	RecFW151R	0.000087
DIM064	LB	fgv-007	recOΔ173–362c	RecOΔ58-end	0.00014
HP125	LB	fgv-037	recOΔT332d	RecOΔ110-end	0.00022
HP101	LB	fgv-013	recRΔ59–150e	RecRΔ19-end	0.00014
DIM060	LB + MC	fgv-004	ssbC289T	SSBR97C	0.1
DIM062	LB + MC	fgv-006	ssbΔ345–434f	SSBΔ115–144	0.16
DIM061	LB + MC	fgv-005	rpoBC1565A	RpoBS522Y	0.084
DIM104	LB + Rif	fgv-011	rpoBC1578A	RpoBH526Q	0.11
DIM105	LB + Rif	fgv-012	rpoBA1547G	RpoBD516G	0.039
HP118	LB	fgv-030	rpoBC604T	RpoBR202C	0.27
HP126	LB	fgv-038	rpoBG2542A	RpoBE848K	0.15
DIM057	LB + MC	fgv-001	Unknown	–	0.31
DIM058	LB + MC	fgv-002	Unknown	–	0.23
DIM059	LB + MC	fgv-003	Unknown	–	0.28

^aSamples from independent cultures of DIM063 (priA dnaC ΔrdgC) grown from independent inocula were spread on the indicated plates (MC at 0.5 µg/ml, Rif at 20 µg/ml), and after 48–60 h at 37°C, colonies of fast-growing variants were purified.

^bFraction surviving 30 J/m² UV relative to unirradiated cells. Survival of wild-type strain AB1157 was 0.5. Data are means from at least two experiments.

^cDeletion of a sequence flanked by tctcttgc direct repeats and one of the repeats.

^dDeletion from a run of 7 T residues.

^eDeletion of a sequence flanked by ccgggcg direct repeats and one of the repeats.

^fDeletion of a sequence flanked by gggtgg direct repeats and one of the repeats.

RecFOR are toxic to ΔrdgC priA dnaC strains

Of the 39 fgv isolates analysed, 16 proved quite sensitive to UV, and remained sensitive to MC. Genetic analyses with 14 of these indicated the presence of an additional mutation in recF (four isolates), recO (two isolates) or recR (eight isolates). Four were sequenced, two in recO and one each in recF and recR. The changes found indicated that protein function is most likely inactivated in each case (Table I). This is consistent with the observed sensitivity to UV (Figure 7B; Table I; data not shown). The recO mutation in strain DIM064 was transferred to wild-type strain AB1157 by exploiting its linkage to pheA. The resulting construct had the same sensitivity to UV as a recO null strain (Mahdi and Lloyd, 1989; data not shown). We also constructed a priA2 dnaC212 ΔrdgC strain carrying the well-characterized recF143 allele. This construct has the same fast-growth, UV-sensitive and non-filamentous cell morphology phenotype as the recF isolate, DIM089 (Figure 7Bviii; data not shown). Taken together, these data demonstrate that the RecF, RecO and RecR proteins are toxic to priA2 dnaC212 ΔrdgC. Furthermore, all three proteins have to be active in order to achieve this effect, suggesting a common basis for the toxicity.

Modifications to SSB protein improve growth of ΔrdgC priA dnaC strains

The remaining mutants grow well on LB agar containing MC. Indeed, several were selected on this basis (Table I). Two were shown by sequencing to carry mutations in ssb, which encodes the SSB protein (Table I, strains DIM060 and DIM062). One has a single nucleotide change encoding an arginine to cysteine substitution at position 97. The other has an in-frame deletion of 90 bp, eliminating 30 amino acid residues from the C-terminal half of SSB. A further three strains were tentatively identified as carrying fgv mutations in ssb on the basis of linkage to malE and the phenotypic similarity to strains DIM060 and DIM062 (data not shown). Given SSB is an essential protein playing crucial roles in DNA replication and repair (Kuzminov, 1999), it is most unlikely that the ssb alleles in these five strains lead to a substantial loss of SSB activity. This conclusion is supported by the fact that the fgv isolates carrying these mutations have a fairly normal cell morphology (Figure 6G), grow very well on LB agar (Figure 7Biii), are resistant to MC and are no more sensitive to UV light than the priA2 dnaC212 rdgC parent (Figure 7D).

RNA polymerase mutations improve viability

Five fgv isolates were shown to carry mutations in the β-subunit (RpoB) of RNA polymerase. These were identified because of their resistance to MC (DIM061) or to rifampicin (DIM104 and DIM105), or simply faster growth on LB agar (HP118 and HP126). The latter are also resistant to MC, but not to rifampicin (data not shown). All five reduce cell filamentation (Figure 6F) and substantially improve viability (data not shown), but have little or no effect on sensitivity to UV light (Figure 7D). However, despite the increased viability, the fgv isolates carrying these mutations tend to form colonies that are generally smaller than those carrying recF, O, R or ssb mutations (Figure 7B; data not shown).

Other fgv suppressors

The final group of fgv isolates are resistant to MC and form smaller colonies, but unlike the RpoB class, they are slightly more resistant to UV light than the parent (Figure 7B and D; Table I; data not shown). Based on the absence of linkage to argE, we can exclude mutations in rpoB or rpoC. We can also exclude recF, recO and recR and we failed to detect linkage of the fgv allele to malE, thereby excluding lexA or ssb (data not shown).

Given the filamentous phenotype of the parent strain, and the chronic SOS induction, we considered the possibility that these strains might carry mutations inactivating the SOS-regulated SulA (SfiA) division inhibitor. We therefore examined the viability and morphology of the sfiA priA2 dnaC212 ΔrdgC construct, DIM177. Although the sfiA mutation reduced filamentation a little, the strain grew just as poorly as the sfiA⁺ control and fast-growing variants appeared just as frequently (data not shown). Thus, sfiA mutations are unlikely to account for this group of fgv isolates.

To investigate whether inactivation of some other SOS-inducible factor might be responsible, we made a lexA3 priA2 dnaC212 ΔrdgC strain. lexA3 encodes a non-cleavable LexA protein, the repressor of SOS, and therefore prevents induction of SOS genes. Figure 7B shows that this lexA3 construct (streak ix) grows better than the lexA⁺ parent, DIM063 (streak i), but not as strongly as derivatives of DIM063 carrying recFOR or ssb suppressors. It is also very UV sensitive (data not shown), as might be expected from its inability to induce SOS. Nevertheless, this observation allows for the possibility that the unidentified fgv mutations might be alleles of some SOS-inducible gene, or genes. However, the UV-resistant phenotype of these isolates would make it unlikely that they carry mutations inactivating recA, ruvAB, uvrA or uvrB.

Discussion

We have shown that the RdgC protein of E.coli binds DNA and is essential for growth of a strain lacking PriA, providing a strong indication that this nucleoid-associated factor might affect replication fork progression or fork rescue. We have also shown that dnaC suppressors of priA overcome this inviability, especially when either RecF, RecO or RecR is inactivated, indicating that RdgC avoids or counters a toxic effect of the RecFOR proteins. Furthermore, we have demonstrated that mutations modifying SSB protein can negate this toxic effect, raising the possibility that binding of RdgC to DNA avoids creating problems at replication forks that would otherwise require PriA to promote replication restart. Mutations in RNAP have a similar effect, providing a connection to transcription.

Native RdgC protein is a homodimer of 34 kDa subunits. Its high expression during early exponential phase encourages the view that it is associated with replication of the chromosome. Its ability to bind dsDNA in a non-specific manner and its association with the nucleoid might suggest a structural role. However, it also forms stable complexes with ssDNA and might bind single strands unwound during replication fork progression. The maximal level of RdgC indicates it would be among the least abundant of nucleoid proteins. Others, such as HU, H-NS, IHF, Fis and StpA are expressed at 5- to 25-fold higher levels (Azam et al., 1999). Fis has a similar expression profile to RdgC and is thought to control transcription and DNA replication during periods of rapid growth and to help maintain chromosome structure (Finkel and Johnson, 1992; Schneider et al., 2001).

Our inability to make a priA2 ΔrdgC double mutant suggests RdgC may have a role that normally reduces problems with replication fork progression, or that RdgC helps to stabilize or repair damaged forks. However, unlike priA null mutants, ΔrdgC strains are resistant to UV, indicating that they have little or no difficulty in maintaining replication fork progression. Furthermore ΔrdgC recB and ΔrdgC recA strains are viable, suggesting that the absence of RdgC does not increase replication fork breakage (Ryder et al., 1996). Either DNA repair and the maintenance of replication fork progression do not require RdgC or these processes can be managed perfectly well by other proteins.

We could make strains lacking PriA and RdgC only if they carried a dnaC212 or dnaC810 suppressor of the priA2 allele used to knock out PriA, but these constructs grow slowly, have low viability and, from their cellular morphology, appear to have more serious defects in chromosome segregation and cell division than a priA-null single mutant. The mutant DnaC proteins in these strains allow PriA-independent loading of DnaB (Xu and Marians, 2000). Because of the absence of PriA helicase activity to unwind a lagging strand at a fork, such DnaB loading is presumably restricted to D loops and forks without a lagging strand at the branch point (Jones and Nakai, 2001; Gregg et al., 2002). So, do strains lacking PriA and RdgC have a specific problem with damaged forks that have a lagging strand obstructing DnaB loading? This is unlikely as a ΔrdgC strain carrying priA300, which encodes a helicase-defective PriA K230R protein (Zavitz and Marians, 1992), is indistinguishable from the single mutants (data not shown).

A more likely explanation stems from our discovery that an activity dependent upon RecF, RecO and RecR is toxic to priA dnaC ΔrdgC strains, and that this toxic effect can be avoided by mutations that modify SSB protein. Derivatives of a priA dnaC ΔrdgC strain carrying these ssb mutations, or mutations that inactivate recF, recO or recR, are very healthy. Although the precise function of the RecF, RecO and RecR proteins remains elusive, the current view is that they act as accessory proteins to facilitate loading of RecA on ssDNA coated with SSB and to help maintain stable RecA–nucleoprotein filaments (Kuzminov, 1999). Thus, the toxicity of RecFOR may reflect some deleterious effect of the RecA filament formed on ssDNA.

Genetic studies indicate that RecF, RecO and RecR function as a complex, and this is supported by evidence of interaction between these proteins in vitro (Umezu et al., 1993; Umezu and Kolodner, 1994; Webb et al., 1997; Bork et al., 2001). Inactivation of any one of the three proteins might therefore limit the loading of RecA at single-strand gaps in DNA. Efficient RecA loading might be restricted largely to situations in which RecBCD enzyme acts to facilitate binding of RecA to ssDNA (Anderson and Kowalczykowski, 1997). That is, RecA might be loaded only at DNA ends processed by RecBCD enzyme. RecO also interacts with SSB, and both RecO and RecR bind SSB-coated ssDNA (Umezu and Kolodner, 1994). Such interactions are crucial for replacing SSB with RecA. The ssb suppressors we identified might prevent these interactions by removing the residues in SSB interacting with Rec(F)OR, or possibly by increasing the affinity of SSB for ssDNA. The former possibility is likely in strain DIM062, which carries a C-terminal deletion of SSB (Table I). This deletion affects a region of SSB implicated in several protein–protein interactions, leaving the N-terminal DNA-binding region unaffected. Whether one might expect such ssb mutations to confer sensitivity to UV, which these particular alleles do not, is an intriguing question. It also begs the question why the toxic effect of RecFOR occurs in a priA2 dnaC212 strain only when RdgC is missing. One tantalizing possibility is that, like SSB, RdgC binds ssDNA exposed during replication, thereby limiting the loading of RecA. SSB is limited to ∼800 tetramers per cell (S.Kowalczykowski, personal communication), whereas our studies indicate that up to 1000 dimers of RdgC may be available during the period of rapid DNA replication. Alternatively, by binding to dsDNA, RdgC may directly or indirectly reduce the amount of ssDNA exposed.

The inactivation of RecFOR would be expected to interfere with many RecA-dependent processes, including post-replication repair of gaps in newly synthesized DNA, SOS induction, replication fork repair and replisome reactivation (Kuzminov, 1999; Courcelle and Hanawalt, 2001; Rangarajan et al., 2002). Which of these RecA-dependent reactions might be toxic in priA2 dnaC212 ΔrdgC cells is unclear, although our studies indicate that it cannot be attributed entirely to SOS induction. Excessive RecA-mediated strand exchange is a distinct possibility. Petit and Ehrlich (2002) found that the low viability of a rep uvrD strain can be suppressed by recFOR mutations. This would be consistent with the hyper-rec phenotype of uvrD strains and the idea that UvrD helicase might normally eliminate recombination intermediates (Petit and Ehrlich, 2002). Inactivation of RecFOR also increases the recovery of recombinants in crosses with recG ruv strains (Ryder et al., 1994), which could be explained if, in the absence of RecG and RuvAB, ‘excessive’ RecFOR-mediated exchanges at gaps interfered with the viability of recombinants generated by RecBCD. Similarly, inactivation of RecF suppresses the low viability of a rep ruv dif strain (Michel et al., 2000), which would otherwise have difficulty eliminating chromosome dimers generated by recombination.

Our work has established a link between RdgC and the ability of RecFOR to stimulate homologous recombination. Therefore, it may be significant that the Neisseria gonorrhoeae homologue of RdgC, along with RecA and RecO proteins, is required for the locus-specific recombination associated with pilin antigenic variation (Mehr et al., 2000).

Why RNA polymerase mutations should alleviate growth defects of a priA dnaC rdgC strain is a mystery. McGlynn and Lloyd (2000) found that certain RNAP mutations can alleviate the UV sensitivity of ruv mutants. The available evidence indicates that these changes to RNAP may reduce transcriptional roadblocks to replication fork progression. Whether this has any relationship to the toxicity of RecFOR remains to be seen, but the observations reported here provide yet another link between DNA replication, transcription and repair. It remains to be seen how they relate to the original observation that RecA is essential for the growth of a recBC sbcBC strain lacking RdgC (Ryder et al., 1996).

Materials and methods

Strains and plasmids

E.coli K-12 strains are listed in Tables I and II. The sfiA::lacZ fusion in strain N5170 is from strain DM4000 (Sandler, 1996). Escherichia coli B strain BL21 DE3 pLysS was used for overexpression of RdgC. pGS853 is an rdgC⁺ derivative of pT7-7 made by PCR amplification of rdgC from pGS830 (Ryder et al., 1996). Primers 5′-TGGAAATTCATATGC TGTGG-3′ and 5′-GGCATTAAAGCTTAATCAGC-3′ were designed to amplify the coding region and incorporated NdeI and HindIII sites (underlined) for cloning the PCR product in pT7-7.

Table II. Strains used.

Strain	Relevant genotypea	Source or reference
DIM167	priA2::Km dnaC810 zji202::Tn10	P1.AG127 × JC19009 to Tcr
JC12334	tna-300::Tn10 recF143	A.J.Clark
JC19009	priA2 dnaC810	Sandler et al. (1996)
MG1655	F–, wild type	Bachmann (1996)
N2247	malE::Tn10 lexA3	Laboratory strain
N4586	ΔrdgC::Tm	P1.JP947 × MG1655 to Tmr
PN109	priA2 sulA	Nurse et al. (1991)
W3110	Wild type	Bachmann (1996)
AB1157 backgroundb
AB1157	F– priA+ rdgC+ rec+	Bachmann (1996)
AG127	zji202::Tn10	A.V.Gregg
AG181	priA2 dnaC212	Gregg et al. (2002)
AG291	priA2 dnaC212 zji-202	A.V.Gregg
DIM037	ΔrdgC	P1.N4586 × AB1157 to Tmr
DIM063	priA2 dnaC212 ΔrdgC	P1.N4586 × AG181 to Tmr
DIM070	priA2	P1.PN109 × AB1157 to Kmr
DIM095	argE+ dnaC212 ΔrdgC	P1.W3110 × DIM061 to Arg+, and Rifs and Kms
DIM112	argE+ dnaC212 ΔrdgC tna recF143	P1.JC12334 × DIM095 to Tcr
DIM113	argE+ dnaC212 ΔrdgC malE lexA3	P1.N2247 × DIM095 to Tcr
DIM122	priA2 dnaC212 ΔrdgC tna recF143	P1.AG181 × DIM112 to Kmr
DIM123	priA2 dnaC212 ΔrdgC malE lexA3	P1.AG181 × DIM113 to Kmr
DIM152	sfiA::lacZ ΔrdgCc	P1.N4586 × N5170 to Tmr
DIM156	sfiA::lacZ dnaC212 zji-202c	P1.AG291 × N5170 to Tcr
DIM158	sfiA::lacZ ΔrdgC dnaC212 zji-202c	P1.AG291 × DIM152 to Tcr
DIM173	sfiA::lacZ priA2c	P1.AG291 × N5170 to Kmr
DIM175	sfiA::lacZ priA2 dnaC212 zji-202c	P1. AG291 × DIM156 to Kmr
DIM177	sfiA::lacZ priA2 dnaC212 zji-202 ΔrdgCc	P1. AG291 × DIM158 to Kmr
DIM211	dnaC810 zji202	P1.DIM167 × AB1157 to Tcr
DIM215	priA2 dnaC810 zji202	P1.AG181 × DIM211 to Kmr
JP947	ΔrdgC::Tm	J.Peters
N5170	sfiA::lacZc	This work

^aAfter the first full listing, insertion and deletion/insertion mutations are referred to by the allele designation only.

^bAlso thi-1 his-4 Δ(gpt-proA)62 argE3 thr-1 leuB6 kdgK51 rfbD1(?) ara-14 lacY1 galK2 xyl-5 mtl-1 tsx-33 supE44 rpsL31, except as indicated.

^cAlso thr+ leu+ Δ(pro-lac)_χ111 and sfiA::lacZ = sfiA::Mu-d(Ap, lac, B::Tn9).

Media and general methods

LB broth and 56/2 salts media have been described previously (Lloyd et al., 1974) and contained 20 µg chloramphenicol (Cm), 40 µg ampicillin (Ap), 25 µg kanamycin (Km), 20 µg tetracycline (Tc), 20 µg trimethoprim (Tm) and 10–100 µg rifampicin (Rif) per ml, as required. MC was incorporated into LB agar at 0.2 and 0.5 µg/ml for plate sensitivity tests. Strain construction by transduction with phage P1vir and methods for determining sensitivity to UV light and MC followed the procedures cited (Lloyd and Buckman, 1991). Suppressors (fgv) of the slow-growth phenotype of priA2 dnaC212 ΔrdgC strain DIM063 were located by P1 transductional crosses. Transductants of fgv strains inheriting markers linked to candidate genes were tested for segregation of the slow-growth phenotype. β-galactosidase activity in sfiA::lacZ fusion strains was measured as described previously (Miller, 1972) and expressed in Miller units. Cell viability was determined by measuring the ratio of colony forming units to total cells detected microscopically in LB broth cultures grown to an A₆₅₀ of 0.4. For DAPI staining, cells were concentrated in TBS buffer (20 mM Tris–HCl pH 7.6, 137 mM NaCl), mixed with an equal volume of DAPI (10 µg/ml in TBS) and spread on polysine-coated slides (BDH). Phase-contrast and DAPI fluorescence images were captured at 1000× magnification.

Enzymes and protein analysis

RecG and RuvC were purified as cited (McGlynn and Lloyd, 2000). RdgC was purified from strain BL21 DE3 cells carrying pGS853 and pLysS following induction with IPTG. Induced cells (9 g wet weight) were lysed on ice by sonication in buffer A [50 mM Tris–HCl pH 7.5, 1 mM EDTA, 1 mM dithiothreitol (DTT)] before adding NaCl to 1 M and collecting soluble proteins by centrifugation. RdgC was extracted using a 50–70% ammonium sulfate cut and the precipitate resuspended in buffer A + 0.1 M NaCl, dialysed against the same buffer and applied to a 40 ml heparin–Sepharose 6 fast flow column. Bound proteins were eluted with a linear gradient of 0.1–1.0 M NaCl in buffer A and peak RdgC fractions eluting between 0.6 and 0.8 M NaCl were pooled, precipitated with 80% ammonium sulfate, resuspended in buffer A + 0.1 M NaCl and applied to a 2.5 × 83 cm Sephacryl S200HR gel filtration column equilibrated in the same buffer. The RdgC peak fractions, eluting at a position corresponding to 70 kDa, were pooled and applied to a 4 ml dsDNA cellulose column equilibrated in buffer A + 0.1 M NaCl. The column was washed with buffer A + 0.1 M NaCl and then with buffer A + 1.0 M NaCl. RdgC eluted in the flow through and the high salt wash. Both fractions were therefore pooled, dialysed against buffer A + 1.5 M ammonium sulfate, applied to a 10 ml butyl–Sepharose 4B column and eluted with a linear 1.5–0 M gradient of ammonium sulfate in buffer A. RdgC eluted as a single peak between 0.9 and 0.65 M ammonium sulfate. These fractions were pooled, dialysed into buffer B (50 mM Tris–HCl pH 8.0, 1 mM EDTA, 1 mM DTT) + 0.1M NaCl, and applied to a 3.0 ml Q Sepharose fast flow column. The column was washed with 30 ml buffer B + 1 M NaCl. RdgC remains bound under these conditions. The bound protein was eluted with buffer A + 2 M NaCl, dialysed into storage buffer [buffer A + 0.1 M NaCl, 50% (v/v) glycerol] and stored in aliquots at –80°C. N-terminal sequencing of the first 10 residues confirmed the purified protein to be RdgC. Protein concentrations were determined using a modified Bradford assay (Bio-Rad), and are expressed as moles of the native dimer for RdgC, or of the monomeric species for RecG and RuvC. Protein cross-linking was conducted as described (Chan et al., 1997), using a final concentration of 10 µM RdgC and glutaraldehyde as indicated, before analysis of the cross-linked samples by SDS–PAGE. Estimation of the molecular weight of RdgC was made by passing 40 µg of the protein, in a volume of 50 µl, through a Superose 12 column (Amersham Pharmacia Biotech. Ltd) equilibrated in buffer A + 0.2 M NaCl. Molecular weight standards were thyroglobulin, γ-globulin, ovalbumin and myoglobulin.

Immunodetection of RdgC

Polyclonal, RdgC antibodies were obtained from rabbits inoculated with RdgC protein and were purified using an RdgC affinity column. Cellular RdgC was detected by SDS–PAGE analysis of lysed cell extracts and western blotting using purified RdgC antibodies, anti-rabbit-IgG:HRP conjugate (Bio-Rad) and ECL detection reagents (Amersham). RdgC levels were quantified using a LAS-1000 fluorescence image analyser (Fujifilm) and normalized to cell numbers detected microscopically.

DNA substrates

χ DNA, a branched DNA structure containing a Holliday junction within a 300 bp core of homology flanked by duplex arms from 0.8 to 1.6 kb in length, was made as described previously (McGlynn and Lloyd, 2000). Other DNA structures were made by annealing combinations of oligonucleotides 1–4, 6 and 7 listed by McGlynn et al. (1997), 8 (5′-GGGTGAACCTGCAGGTGGGCAAAGATGTCCTAGCAAGGCAC TGGTAGAATTCGGACAGCGTC-3′), 9 (5′-AACGTCATAGACGATT ACATTGCTA-3′) and 10 (5′-TAGCAATGTAATCGTCTATGAC GTT-3′), one of which was ³²P-labeled at the 5′ end before annealing. J12 was made from oligonucleotides 1–4, the three-strand junction from 1+2+4, the Y structure from 1+4+6 and the flayed-duplex from 1+2. All these structures were made of oligonucleotides of 49–51 residues. Sixty-one base pair and 25 bp linear duplexes were made from oligonucleotides 7+8 and 9+10, respectively. The 25 and 61 nucleotide ssDNA substrates were oligonucleotides 7 and 9, respectively. Poly[dIdC] was used as competitor DNA.

Enzyme assays

Resolution of χ DNA by RuvC was measured as described previously (McGlynn and Lloyd, 2000). Reaction mixtures (10 µl) were incubated for 30 min at 37°C before deproteinization and electrophoresis through 0.8% agarose gels. For DNA-binding assays, reactions (20 µl) contained 0.1 nM ³²P-labelled DNA substrate mixed with the indicated amount of protein in 50 mM Tris–HCl pH 8.0, 1 mM DTT, 100 µg/ml BSA, 6% (v/v) glycerol and 2 mM MgCl₂. After 10 min on ice, samples were loaded on a chilled (4°C) 4% polyacrylamide gel in 6.7 mM Tris–HCl pH 8.0, 3.3 mM sodium acetate, 2 mM MgCl₂. Gels were run at 160 V for 90 min, dried and analysed by autoradiography and phosphoimaging. To measure the stability of RdgC bound to ssDNA or dsDNA, binding reactions containing 0.1 nM ³²P-labelled DNA and sufficient RdgC to bind ∼20–50% of the substrate were mixed on ice for 10 min before adding 50 ng poly[dIdC] and incubating for the times indicated before gel analysis of bound complexes.