Cellular location and activity of Escherichia coli RecG proteins shed light on the function of its structurally unresolved C-terminus

RecG is a DNA translocase encoded by most species of bacteria. The Escherichia coli protein targets branched DNA substrates and drives the unwinding and rewinding of DNA strands. Its ability to remodel replication forks and to genetically interact with PriA protein have led to the idea that it plays an important role in securing faithful genome duplication. Here we report that RecG co-localises with sites of DNA replication and identify conserved arginine and tryptophan residues near its C-terminus that are needed for this localisation. We establish that the extreme C-terminus, which is not resolved in the crystal structure, is vital for DNA unwinding but not for DNA binding. Substituting an alanine for a highly conserved tyrosine near the very end results in a substantial reduction in the ability to unwind replication fork and Holliday junction structures but has no effect on substrate affinity. Deleting or substituting the terminal alanine causes an even greater reduction in unwinding activity, which is somewhat surprising as this residue is not uniformly present in closely related RecG proteins. More significantly, the extreme C-terminal mutations have little effect on localisation. Mutations that do prevent localisation result in only a slight reduction in the capacity for DNA repair.


INTRODUCTION
RecG is a double-stranded DNA translocase associated with the maintenance of genomic integrity in bacteria (1,2). It is present in almost all species examined (3,4). Early genetic studies suggested a role in promoting homologous re-combination, a possibility consistent with the strong synergism between recG and ruv null alleles (1,5) and reinforced when the Escherichia coli RecG protein was shown to unwind Holliday junction structures (6,7). The mode of unwinding proved reminiscent of the branch migration activity displayed by the RuvAB component of the RuvABC Holliday junction resolvase complex (8)(9)(10). This has leant weight to the idea that RecG and RuvABC provide partially overlapping pathways for processing intermediates in homologous recombination (5).
However, subsequent studies showed that RecG targets a range of other substrates, including D-loops and R-loops, raising the possibility of alternative roles. It has a strong affinity for structures mimicking complete or partial replication forks (11)(12)(13)(14)(15)(16), and can convert a fork to a Holliday junction (17)(18)(19)(20)(21)(22)(23). Coupled with evidence of a genetic interaction with PriA protein (24)(25)(26)(27)(28), these studies led to the proposal that RecG might catalyse the reversal or regression of stalled forks in vivo, enabling repair or bypass of the blocking lesion and the restart of replication, thus facilitating the completion of chromosome duplication (17). PriA initiates replisome assembly at branched DNA structures, thereby enabling replication to restart at sites remote from oriC (11,29). Fork regression has been investigated in detail in E. coli (30,31), and models of replication restart invoking such regression have become generally popular, with several eukaryotic helicases having been shown in recent years to have the potential to catalyse such a reaction in vitro (32)(33)(34)(35). However, evidence for such pathways in vivo remains at best indirect (36). Furthermore, recent studies have indicated that much of the recG mutant phenotype is a pathological consequence of stable DNA replication (SDR), a form of unscheduled chromosome replication triggered by PriA-mediated replisome assembly (2,(37)(38). The initiation of SDR is particularly noticeable in the terminus area of the chromosome where forks normally meet to complete replication. It appears that fork collision may frequently generate 3 flap structures that could be targeted by PriA (39). RecG has a high affinity for 3 flaps and in conjunction with single-stranded DNA exonucleases may normally eliminate these structures, preventing their exploitation by PriA (37,39). As with fork regression, this role would place RecG at or near sites of DNA replication.
Here we present evidence consistent with the idea that E. coli RecG localises to sites of active replication and identify residues near the C-terminus of the protein that may facilitate this localisation. In addition, we demonstrate that the extreme C-terminus of RecG, which is not resolved in the crystal structure, is vital for DNA unwinding but not for DNA binding. Indeed, deleting or substituting the very last amino acid is sufficient to reduce unwinding activity to less than 5% of wild-type.

Strains and plasmids
Bacterial strains are listed in Table 1. All constructs used for synthetic lethality assays are based on E. coli K-12 MG1655 ΔlacIZYA strains carrying derivatives of pRC7, an unstable lac + , mini-F plasmid that is easily lost (40). pJJ100 is a derivative of pRC7 encoding recG + (41), while pAM475 is a derivative encoding polA + . The wild-type polA coding sequence plus some 50 bp of upstream sequence was amplified by PCR from MG1655 (42) using primers incorporating flanking ApaI restriction sites and then inserted into pRC7 at the ApaI site within lacI q . pAM475 eliminates the sensitivity to UV light of a strain carrying a C-terminal deletion of polA (27), demonstrating that the cloned polA + gene is functional. pAM210 is a recG + derivative of the expression vector pT7-7 (43). pQW145 is a pAM210 derivative encoding RecG Q640R (20). To construct strains carrying mutant recG genes encoding RecG C-terminal deletions or substitutions, the relevant mutations were first generated by PCR amplification using primers carrying the required sequence alterations. The products were cloned into pQW145, replacing the 3 end of recG with the new mutant sequence. The 3 primer also added a HindIII site immediately downstream of the gene. A gene encoding resistance to kanamycin (kan) was then inserted at this site. The altered 3 end of recG incorporating the desired deletion/substitution plus the downstream kan gene was amplified by PCR and the product used to engineer these features into strain MG1655, replacing the wild-type recG allele, using the protocol of Datsenko and Wanner (44). A strain encoding recG + linked to a downstream kan gene (recG + -kan) was engineered as a control for the C-terminal mutants generated. These recG alleles carrying the kan gene downstream are identified in Table  1 with the relevant alteration of the coding sequence shown in square brackets and with the suffix '-kan'. For ease of presentation, the suffix is omitted in the main text. For the same reason, the gene encoding RecG protein with both R682A and W683S substitutions is abbreviated to recG [RW].
Plasmid constructs expressing fluorescent RecG fusions under control of the arabinose-inducible P araBAD promoter are derivatives of pLau18 (45). A sequence encoding the full length recG gene plus a 5 extension encoding an N-terminal linker sequence (MELYLIDYLEC) was PCR amplified using a 5 primer that also introduced a BsrGI restriction site and a 3 primer that introduced an XbaI site. The PCR product was cloned between the BsrGI and XbaI sites of pLau18 such that the recG gene with its 5 linker extension is inserted in frame downstream of the P araBAD promoter and eYFP coding sequence, generating pDIM071. To create constructs encoding the eYFP-RecG mutant fusions described in the main text, KpnI-HindIII fragments from the pT7-7 constructs encoding the relevant mutant recG genes were cloned between the KpnI and HindIII sites of pDIM071. pDIM083 (eCFP-SeqA) has been described previously (38). The NheI-HindIII fragment from pDIM083 was cloned between the XbaI and HindIII sites of pDIM071, generating pDIM113 (eYFP-RecG eCFP-SeqA).

Media and general methods
LB broth and 56/2 minimal salts media and methods for monitoring cell growth and for strain construction by P1virmediated transduction have been cited (17,24,46). For microscopy, 56/2 salts were solidified with agarose at a final concentration of 1%.

Fluorescence microscopy
Cultures of strains carrying the relevant fusion constructs were incubated in LB broth until they reached an A 650 of 0.2. The encoded fusion proteins were induced by adding arabinose to a final concentration of 0.2% before incubating for a further 60 min. Samples were then transferred to a microscopic slide coated with a thin layer of 56/2 agarose. The cells were visualised using a BX-52 Olympus microscope equipped with a coolSNAP TM HQ camera (Photometrics). eCFP and eYFP foci were visualised using the JP4-CFP-YFP filter set 86002v2 (Chroma). Images were taken and analysed by MetaMorph 6.2 (Universal Imaging) and processed using MetaMorph and Adobe Photoshop.

Measuring sensitivity to DNA damage
Sensitivity to UV light was measured using cultures of cells grown in LB broth to an A 650 of 0.48 (1-2 × 10 8 cells/ml for strain MG1655). Samples of appropriate dilutions were irradiated on the surface of LB agar plates and survivors were scored after 18-24 h incubation. Survival values are means of three to six independent experiments. Error estimates (SE) range from 5% to 15% of the mean. To determine sensitivity to mitomycin C, the cultures were diluted in 10fold steps from 10 −1 to 10 −5 , and 10 l aliquots of each dilution spotted on LB agar with and without mitomycin C at a final concentration of 0.2 or 0.5 g/ml, as indicated. Plates were photographed after 24 h incubation unless stated otherwise.

Synthetic lethality assays
The assays were conducted as described (40,41). Essentially, a wild-type gene of interest is cloned in pRC7, an unstable lac + plasmid that is easily lost from cells, which is then used to cover a mutation of the same gene in the chromosome in a  Δrep::dhfr pDIM113 pDIM113 × AM1969 to Ap r RCe452 ΔrecG::apra MG1655 × P1.N6052 to Apra r RCe571 Δrep::dhfr MG1655 × P1.RCe371 to Tm r RCe575 Δrep::dhfr pDIM071 pDIM071 × RCe571 to Ap r TB28 ΔlacIZYA (40) a The abbreviations apra, cat, dhfr and kan refer to insertions conferring resistance to apramycin (Apra r ), chloramphenicol (Cm r) ), trimethoprim (Tm r ) and kanamycin (Km r ), respectively. Ap r refers to ampicillin resistance. b From white colonies on minimal agar supplemented with X-gal and IPTG. c The construct was generated by fusing the open reading frame for expressing eYFP to the 5 end of a wild-type recG gene linked to a downstream kan sequence (see the Materials and Methods section) to give pJG146. The entire fusion linked to kan was amplified from pJG146 by PCR and directed by recombineering into N6052, selecting Km r and screening for sensitivity to apramycin.
Δlac background. Plasmid loss is readily scored by spreading samples on agar plates supplemented with X-gal and IPTG. Plasmid-free cells form white colonies whereas cells retaining the plasmid form blue colonies with white sectors as plasmid loss continues during colony growth. If a mutation in another gene is introduced and the double mutant is viable, white colonies/sectors of plasmid-free cells still appear. However, if synthetically lethal, the plasmidfree cells fail to grow and only non-sectored blue colonies formed by cells retaining the plasmid are observed. The size of any Lac − colonies relative to the Lac + colonies also gives some indication of the viability of the plasmid-free cells. In our standard assays, cultures of strains carrying the relevant pRC7 derivatives were grown overnight in LB broth containing ampicillin to maintain plasmid selection, diluted 80-fold in fresh broth without ampicillin and incubation continued to an A 650 of 0.48 before spreading samples of various dilutions on LB agar or 56/2 glucose minimal salts agar supplemented with X-gal and IPTG. Plates were photographed and scored after 48 h (LB agar) or 72 h (minimal agar) at 37 • C. Plasmid-free cells forming small white colonies were re-streaked to see if they could be subcultured, and the streaked plates were photographed after incubation at 37 • C for 24-48 h (LB agar).

Purification of RecG
All chromatography was performed at 4 • C and has been described previously (21). RecG proteins were expressed by IPTG induction from pT7-7 constructs in strain AU1118.

DNA substrates
DNA substrates were made by annealing the oligonucleotides, one of which was labelled with 32 P at its 5 end, as described (6). DNA concentrations are in moles of the molecular structure. J12 is a Holliday junction structure with a homologous core of 12 bp flanked by 19-20 bp heterologous arms (6). The fork structure mimics a replication fork lacking a lagging strand (21).

DNA binding and unwinding assays
DNA binding by RecG was measured using a band-shift assay (6). RecG and 32 P-labelled DNA were mixed in binding buffer (50 mM Tris-HCl pH 8.0, 5 mM EDTA, 1 mM DTT, 100 mg/ml BSA and 6% glycerol) and incubated on ice for 20 min before loading onto a pre-chilled 4% native polyacrylamide gel in a low ionic strength buffer (6.7 mM Tris-HCl pH 8.0, 3.3 mM sodium acetate and 2 mM EDTA). Electrophoresis was at 160 V for 75 min. Gels were then dried and analysed by autoradiography (X-Omat UV Plus film, Kodak) and phosphorimaging (STORM scanner system and ImageQuant 5.2, Molecular Dynamics). DNA unwinding was assayed essentially as described (12). The rates of unwinding were measured using bulk reactions. RecG at 0.5 nM was mixed in helicase buffer (20 mM Tris-HCl pH7.5, 2 mM DTT, 100 mg/ml BSA, 5 mM ATP and 5 mM MgCl 2 ) and kept on ice prior to addition of labelled substrate DNA (0.2 nM). An aliquot was removed immediately and deproteinized by the addition of 0.2 volume of stop buffer (2.5% SDS, 200 mM EDTA and 10 mg/ml proteinase K) and incubating for a further 10 min at 37 • C; this was taken as the time zero sample. The reaction was then placed at 37 • C and samples subsequently removed at intervals and deproteinized. Samples were analysed by electrophoresis using a 10% polyacrylamide gel and a Trisborate buffer system before processing as above.

RecG is located to sites of active DNA replication
We fused eYFP to the N-terminus of RecG and investigated whether the fusion protein forms foci that might reflect the role of RecG in vivo. A linker peptide was employed to construct the fusion protein because previous attempts to create direct fusions to the N-terminus of E. coli RecG failed, suggesting that such fusions might be lethal (47). The linker used is based on the peptide sequence linking the DNA binding domain of Thermotoga maritima RecG to an Nterminal domain that does not exist in the E. coli protein (22). We reasoned that the linker might keep the eYFP from interfering with RecG function and generally out of harm's way. We replaced the chromosomal copy of recG with our fusion construct. Strains expressing the fusion protein from the native recG promoter proved as resistant to mitomycin C and to UV irradiation as a strain carrying wild-type recG, showing none of the sensitivity of a ΔrecG strain and confirming that the eYFP-RecG fusion protein is indeed functional in vivo ( Figure 1). However, we were unable to detect any foci by fluorescence microscopy (data not shown).
RecG is present at very low levels (less than 10 molecules per cell) (48), which might explain the absence of a fluorescence signal. Indeed, a recent study in which a more sensitive fluorescence microscopy system was used also revealed that a RecG fusion protein expressed from its native promoter does not form any detectable foci (49). Therefore, we over-expressed the fusion protein from a multi-copy plasmid using the P araBAD promoter. Its expression eliminated the sensitivity of a ΔrecG strain to UV irradiation and mitomycin C (data not shown). It also resulted in the formation of clearly defined foci in the majority of the cells analysed, anything from 1 to 6 per cell (Figure 2A(i)). A sample of 382 cells scored from three independent experiments revealed that 65% had 1 or 2 foci, 14.7% had 3 or 4 and 1.3% had 5 or 6. The remaining 19% had none. We next examined whether RecG foci co-localise with foci formed by SeqA protein tagged with eCFP (Figure 2A(i)). Of 323 RecG foci analysed in 164 cells from three independent ex-periments, 296 co-localised with SeqA foci. SeqA binds to hemi-methylated DNA, as found immediately behind replication forks (50,51), and therefore provides a marker for areas of the nucleoid in which replication is taking place. To gain further evidence that the location of RecG foci corresponds to sites of replication, we exploited the fact that chromosome copy number is increased in cells lacking the DNA helicase Rep (52). If RecG does accumulate at sites of replication, there should be an increase in foci following expression of the fusion protein in these cells. This is what we observed (Figure 2A(ii)). An analysis of 186 cells sampled from two independent experiments with a Δrep strain revealed that only 38.5% of the cells had 1 or 2 foci. Sixty percent had 3 or more, with 12% showing 5 or 6. Only 1.5% of the cells had no foci at all, which is less than a tenth of the number seen with rep + cells.
Both the cellular location and number of these RecG foci are consistent with the idea that the protein associates with replisome complexes, as RecG is reported to do in B. subtilis. Lecointe et al. (53) demonstrated that B. subtilis RecG co-localises with the replisome protein DnaX via an interaction with the C-terminus of the single-stranded DNA binding protein SSB (53). Recent studies have shown that E. coli RecG also interacts with the C-terminus of E. coli SSB in vitro (27,(54)(55). Given that SSB is concentrated at sites of active replication (56), this interaction might account for the RecG foci we observed, and for their co-localisation with SeqA. We can exclude localisation via a protein:DNA interaction as we found that an eYFP-RecG construct based on a RecG derivative lacking the 'wedge domain' needed for high affinity DNA binding still localises into foci (48) ( Figure 2B).
Co-localisation of RecG with the replisome via an interaction with SSB might explain the previously observed lethal effect of a direct fusion of maltose binding protein (MBP) to the N-terminus of RecG (47). The fusion could be lethal because the MBP tag might disturb vital replisome interactions. Since the eYFP-RecG described here was expressed without causing loss of viability, we assumed the extra linker region deployed between the two globular proteins might be holding the eYFP tag in a position where it does little or no harm. Significantly, the failure to create a direct fusion of MBP to RecG was overcome by deleting the last 32 amino acid residues from the C-terminus of RecG. The MBP-RecG C32 mutant protein still bound branched DNA in vitro with an affinity similar to that of wild-type RecG (47). However, localisation of the construct in vivo was never investigated. If some feature of the C-terminus of RecG mediates the interaction with SSB, mutant RecG proteins with deletions or substitutions eliminating this feature would no longer be expected to exhibit localisation.

C-terminal mutations reduce the ability of RecG to form foci
We generated a set of recG alleles encoding eYFP-RecG fusions lacking the last 5, 10, 15, 20 or 30 amino acids (Figure 2C(ii)). Proteins lacking the last 5 or 10 residues still form discrete foci, but a protein lacking the last 15 was mostly distributed throughout the cell, with foci barely visible ( Figure 2B). More extensive deletions eliminated focus formation altogether (data not shown). An alignment of the C-terminus of RecG proteins that have C-termini of similar length shows that a tryptophan (W) and an arginine (R), located 11 and 12 residues respectively from the Cterminus of E. coli RecG, are conserved ( Figure 2C(i)). We created mutant proteins with substitutions at one or both of these positions and examined eYFP derivatives to see if they would still localise. Focus formation was still evident with R682L and W683F proteins, but less so with W683L and W683S derivatives, and hardly detectable with a protein carrying both R682A and W683S substitutions ( Figure 2B). This double mutant, which we designate as RecG[RW], was clearly distributed throughout the cell.
These data indicate that arginine 682 and tryptophan 683 might indeed be key to focus formation. They are consistent with the notion that the foci we observe with eYFP fused to wild-type RecG reflect localisation of RecG to specific sites rather than being caused by aggregation of the fluorescent fusion. Such localisation might be facilitated via an interaction between the C-terminus of RecG and that of SSB. The fact that the arginine and the tryptophan are conserved in RecG proteins from diverse species indicates that localisation of RecG might be generally important.

Effect of localisation on the in vivo activity of RecG
To assess whether localisation has any effect on RecG function, we analysed the phenotype conferred by the Cterminal mutations. Each mutation was engineered at the chromosomal recG locus under its native promoter, linked to a downstream kan insertion. A wild-type recG sequence was similarly linked to kan to provide an appropriate control, and is designated recG + -kan. Replacing recG + with recG + -kan does not increase sensitivity to mitomycin C ( Figure 3A), a major feature of recG null strains (1). Furthermore, it does not increase the sensitivity of a ΔruvABC strain to UV light ( Figure 3B(i)), a particularly sensitive test of RecG activity given the very strong synergism between recG and ruv null alleles (5,57). Thus, it seems the kan insertion engineered downstream of recG has little or no effect on the activity of the gene.
Strains carrying ΔC20 and ΔC30 alleles of recG proved sensitive to mitomycin C and UV light, and each allele conferred extreme UV sensitivity when combined with ruvABC (data not shown), consistent with these alleles retaining little or no RecG activity. However, we cannot attribute this to the failure of localisation as previous studies suggested that substantial deletions from the C-terminus severely curtail in vitro DNA unwinding activity (43). We therefore investigated the phenotype of an R682A, W683S double mutant, which greatly reduces the ability to form foci. Despite this failure to localise, the RecG[RW] protein appears to retain much of its ability to function. Thus, a strain carrying the mutant allele (recG[RW]) grows robustly on agar plates supplemented with mitomycin C (Figure 3A). On its own, the recG[RW] allele confers no sensitivity to UV light and although it does increase sensitivity when combined with ΔruvABC, it does so only to a slight extent ( Figure 3B(ii)). To gain further measures of activity, we exploited the fact that a recG null allele confers lethality when combined with rnhA or polA deletions (27,37,58). An assay for synthetic lethality based on unstable pRC7 derivatives carrying either recG + or polA + revealed that recG [RW] ΔrnhA and recG[RW] polA2 double mutant cells are quite viable, establishing robust colonies without a covering plasmid ( Figure 3C, panels (iii) and (vi)), colonies that can be subcultured without difficulty ( Figure  3D). Thus, RecG[RW] does not appear severely compromised by the reduced ability to form foci.
Genetic analyses of strains carrying recG[ΔC5] or recG [ΔC15] revealed that while they appear to differ in their ability to form foci, they confer equally strong mutant phenotypes, much stronger than is conferred by recG [RW]. Both confer more resistance to mitomycin C and UV light than does a recG null allele (ΔrecG) ( Figure 3A and B). However, when combined with ΔruvABC, they both confer extreme sensitivity to UV light, almost as extreme as that of a ΔrecG ΔruvABC strain ( Figure 3B). Unlike recG [RW], both alleles also proved synthetically lethal with ΔrnhA and also with ΔpolA2 ( Figure 3C and data not shown). The small white colonies seen in the ΔpolA2 assay ( Figure  3C(v)) could not be subcultured.
Taken together, these data indicate that there is no strong correlation between RecG focus formation and RecG activity. However, the slightly increased sensitivity to UV light of the recG[RW] ruvABC double mutant means we cannot dismiss entirely the possibility that the localisation of RecG confers some advantage.

The extreme C-terminus of RecG is necessary for function
The strong phenotype conferred by recG[ΔC5] prompted us to further dissect the C-terminus. We first examined the effect of adding three alanine residues to the very end. Tests for sensitivity to mitomycin C and UV light, and for viability in the presence of ΔrnhA or ΔpolA2, revealed no substantial reduction in RecG functionality in vivo (data not shown). However, removing one, two or three residues from the C-terminus clearly reduced the ability of the protein to function. The phenotype conferred in each case proved essentially identical and so we restrict our data presentation to recG[ΔC1], which removes the terminal alanine ( Figure  2C). We also report on two missense alleles, one substituting the conserved tyrosine four residues from the end with an alanine (recG[Y690A]) and the other substituting the terminal alanine with a glutamine (recG[A693Q]). Single mutants carrying any one of these three alleles proved only very mildly sensitive to mitomycin C and quite resistant to UV ( Figure 4A and B). But a substantially reduced RecG activity was uncovered when these alleles were combined with ΔruvABC. In each case, the double mutant proved very Nucleic Acids Research, 2014, Vol. 42, No. 9 5709    much more sensitive to UV light than a ΔruvABC control ( Figure 4B). Synthetic lethality assays also revealed a loss of functionality. The assays showed that strains carrying these alleles require wild-type rnhA to grow on LB agar ( Figure  5A). Similar assays testing viability with ΔpolA2 proved deceptive. In this case, the double mutant cells form what appear to be reasonably robust (white) colonies on the LB indicator plates ( Figure 5B(i)-(iii)), giving the impression that DNA polymerase I is dispensable. However, when these colonies are streaked on LB agar, it becomes apparent that much of the colony growth can be attributed to abortive growth of inviable cells and the outgrowth of suppressors ( Figure 5B(iv)). Further inspection revealed that the double mutants are viable on minimal agar and can be subcultured without being outgrown by suppressors, as evident when these subcultures are subsequently tested for colony formation (the same is true for ΔrecG ΔpolA2 cells). The vast majority of the cells fail to establish colonies on LB agar, but grow fine on minimal agar ( Figure 5C), demonstrating that DNA polymerase I is in fact needed to sustain viability under conditions supporting rapid growth and division.

RecG C-terminal mutations compromise biochemical activity
Previous studies established that the helicase domains of RecG are located towards the C-terminus, with the Nterminal wedge domain being involved with DNA binding (20)(21)(22)43,48). Therefore, we suspected that the substantial in vivo reduction in RecG activity observed with each of the three C-terminal mutants described in the previous section is due to a reduction in the ability to unwind DNA. Unfortunately, the extreme C-terminus of RecG is not resolved in the only published crystal structure of RecG (22), indicating a degree of flexibility. Therefore, its precise disposition relative to the conserved helicase domains is not known. We purified some of the mutant RecG proteins and investigated their ability to bind and unwind branched DNA substrates. The C1, Y690A and A693Q proteins retain the ability to bind the Holliday junction and fork DNA structures tested, and with an affinity indistinguishable from wild-type RecG ( Figure 6A and data not shown). They also retain some unwinding activity, but this is substantially reduced ( Figure 6B). RecG[ C1] has the most extreme deficiency in that it unwinds very little of either of the two substrates. RecG[A693Q] also shows little unwinding of the fork, but retains some activity on the Holliday junction. RecG[Y690A] has an intermediate activity with both substrates. Taken together with the in vivo properties, these data demonstrate that the extreme C-terminus of RecG down to the very last amino acid is necessary for the protein to be fully functional. Given this fact, we were surprised that adding three alanine residues to the C-terminus has no negative effect that we could detect. As with the in vivo studies, the extended protein is as active as wild-type RecG in the binding and unwinding assays (data not shown). rnhA recG + -kan

Localisation of E. coli RecG
Several lines of evidence have led to the suggestion that RecG might have a particularly important role to play in securing faithful genome duplication (2,39). The observation that Bacillus subtilis RecG co-localises with the replisome lends support to this idea (53), suggesting that the function of RecG is required specifically where DNA replication is taking place.
The data presented in this study support the idea that E. coli RecG also localises to the replisome (Figure 2). We found that an eYFP-RecG fusion forms foci that co-localise with SeqA, a marker for sites of active DNA synthesis. We were concerned initially that the foci observed might be the result of fluorescent protein dimerization and subsequent aggregation due to over-expression (59). However, the correlation with chromosome copy number revealed using a strain lacking Rep helicase and the elimination of foci by specific C-terminal mutations indicates that they are instead the result of a specific localisation to sites of chromosome replication.
Two conserved and adjacent amino acid residues, Arginine 682 and Tryptophan 683, were identified as being particularly important effectors of this localisation. This is significant as these residues lie close to the very end of the protein in a region that is not resolved in the RecG crystals described to date (22). Therefore, although the structure and disposition of this region has not been established, it is most likely flexible and free to interact with other factors. Given E. coli RecG interacts with SSB protein in vitro (27,54), and SSB coats the unwound lagging strand at forks, it is tempting to speculate that the observed localisation of RecG reflects this interaction, with the two proteins making contact via their C-termini. However, our analyses of C-terminal RecG mutants provided no evidence of a strong correlation between localisation and function. Our C-terminal deletion mutants were not particularly informative in this respect as all those constructed severely curtail the activity of the protein in vivo. With hindsight, this is perhaps not surprising as we found that deleting just the very last amino acid is sufficient to compromise activity both in vivo and in vitro, as we discuss below. What we did find is that the RecG[RW] substitution mutant localises very poorly compared with the wild-type protein despite otherwise retaining near wild-type activity in vivo. The properties associated with this protein question the significance of the observed localisation. Does it really signify a role for RecG in promoting rescue of stalled forks as suggested (17,25,60)? The fact that RecG interacts with SSB allows for alternative interpretations.
During balanced growth, SSB is particularly concentrated at replication forks where it associates with the unwound lagging strand template (56,61). The observed localisation of the over-expressed E. coli RecG to the replisome may simply reflect this fact. Given the very high affinity of SSB for ssDNA and assuming it does act as a hub for DNA repair proteins in vivo, as suggested (55), there is no reason why the bound RecG should not re-locate to wherever ss-DNA is exposed and covered with SSB. Thus, RecG could be targeted to any number of branched DNA substrates (2), only some of which may be the 'true' substrate at which RecG function is required. However, the conservation of the RW motif near the C-terminus of RecG, coupled with the slight loss of RecG function when this motif is mutated, indicates that there may be some selective advantage to the co-localisation with SSB. For instance, the SSB interaction might enhance RecG helicase activity, as recently reported (62).

RecG helicase activity
We found that the extreme C-terminus of RecG is crucial for helicase activity in vitro. Previous analysis of deletion mutants had established the importance of the C-terminus, but the shortest deletion tested removed 32 residues from the end of the protein (43). Here we describe how eliminating just the terminal alanine is sufficient to make the protein almost non-functional in vitro, despite showing no reduction in substrate affinity (Figure 6). Substituting a glutamine for this alanine has almost as drastic an effect, especially with a fork substrate. This was rather surprising, as the respective single mutant strains exhibited only very mild sensitivity to mitomycin C and appeared fully resistant to UV light ( Figure 4), which seemed to indicate that the proteins retained substantial activity. The presence of SSB, which has been reported to stimulate RecG activity (62), may account for this discrepancy. However, more sensitive assays based on double mutants carrying additional mutations inactivating RuvABC, ribonuclease HI or DNA polymerase I established that there is in fact a substantial loss of function in vivo (Figures 4 and 5). Nevertheless, we note that this alanine is not uniformly present in closely related RecG proteins ( Figure 2C).
In conclusion, our data demonstrate that the C-terminus of RecG is vital for both its helicase activity and for its localisation. However, while a reduced helicase activity is clearly associated with a distinct mutant phenotype, there appears to be no strong correlation between localisation and functionality.