Mycobacterium tuberculosis RecG binds and unwinds model DNA substrates with a preference for Holliday junctions

The RecG enzyme, a superfamily 2 helicase, is present in nearly all bacteria. Here we report for the first time that the recG gene is also present in the genomes of most vascular plants as well as in green algae, but is not found in other eukaryotes or archaea. The precise function of RecG is poorly understood, although ample evidence shows that it plays critical roles in DNA repair, recombination and replication. We further demonstrate that Mycobacterium tuberculosis RecG (RecGMtb) DNA binding activity had a broad substrate specificity, whereas it only unwound branched-DNA substrates such as Holliday junctions (HJs), replication forks, D-loops and R-loops, with a strong preference for the HJ as a helicase substrate. In addition, RecGMtb preferentially bound relatively long (≥40 nt) ssDNA, exhibiting a higher affinity for the homopolymeric nucleotides poly(dT), poly(dG) and poly(dC) than for poly(dA). RecGMtb helicase activity was supported by hydrolysis of ATP or dATP in the presence of Mg2+, Mn2+, Cu2+ or Fe2+. Like its Escherichia coli orthologue, RecGMtb is also a strictly DNA-dependent ATPase.


INTRODUCTION
Mycobacterium tuberculosis, the aetiological agent of the reemerging disease tuberculosis (TB), remains a global health threat, killing at least 1.5 million individuals every year (WHO, 2011). The emergence of extensively and extremely drug-resistant M. tuberculosis strains, coupled with the HIV/AIDS pandemic, has exacerbated the risk of TB resurgence, underlining the urgent need to develop interventions that halt the spread of this disease. M. tuberculosis, an intracellular human pathogen, successfully combats many host cell defence mechanisms, including genotoxic stress, using efficient DNA repair pathways that help to maintain its genome integrity, in spite of accumulating DNA damage during infection (Ambur et al., 2009;Dos Vultos et al., 2009;Gorna et al., 2010;Mizrahi & Andersen, 1998;Stallings & Glickman, 2010). It is believed that these DNA repair pathways promote M. tuberculosis survival and increase its pathogenicity and virulence (Gorna et al., 2010;Warner, 2010). Thus, in-depth characterization of the mechanisms that protect the M. tuberculosis genome and promote its virulence and/or capacity to develop drug resistance may lead to novel therapeutic targets and attenuate the increasing risk of global resurgence of TB.
Escherichia coli RecG, a 76 kDa monomeric helicase with a particular affinity for branched-DNA substrates such as replication forks, Holliday junctions (HJs), and D-and Rloops, has been shown to play roles in DNA repair, recombination and replication (Lloyd & Sharples, 1993;McGlynn et al., 1997;McGlynn & Lloyd, 2000Rudolph et al., 2010b;Vincent et al., 1996;Whitby & Lloyd, 1998). E. coli RecG is widely believed to promote regression of stalled replication forks, when fork progression is blocked by lesions in the DNA template strand, thereby facilitating repair or bypass of the lesion . In vitro studies suggest that RecG actively unwinds stalled replication forks, generating a four-way junction product that resembles an HJ (McGlynn & Lloyd, 2000, and also promotes HJ branch migration (Müller & West, 1994;Whitby et al., 1994). Structural characterization of a complex between Thermotoga maritima RecG and a forked-DNA substrate has revealed the mechanism by which RecG recognizes junctions (Singleton et al., 2001). E. coli RecG inhibits inappropriate DNA amplification and aberrant chromosome segregation in cells exposed to UV irradiation (Rudolph et al., 2009a, b), and is also essential in cells lacking 39 ssDNA exonucleases to counteract PriA helicasemediated DNA re-replication (Rudolph et al., 2010a). Furthermore, a recent study has revealed that RecG promotes resolution of intermolecular recombination intermediates that are poorly recognized/resolved by RuvABC (Fonville et al., 2010).
Strains with mutations in recG have been shown to exhibit complex and variable phenotypes, including transformation deficiency in Neisseria meningitidis (Sun et al., 2005), growth defects and reduced radio-resistance in Deinococcus radiodurans (Wu et al., 2009), and sensitivity to UV irradiation and oxidative stress in Pseudomonas aeruginosa (Ochsner et al., 2000). recG mutation has also been suggested to be responsible for the susceptibility of Staphylococcus aureus to quinolone (Niga et al., 1997) and of E. coli to bleomycin, metronidazole and ciprofloxacin (Kosa et al., 2004;Tamae et al., 2008).
In this study, we have characterized the recombinant RecG enzyme from M. tuberculosis H37Rv (RecG Mtb ) for its DNA binding, unwinding and ATPase activities in order to delineate its potential roles in the DNA metabolism of M. tuberculosis.
Overexpression and purification of the RecG Mtb protein. The recombinant RecG Mtb protein was purified to homogeneity as follows. E. coli ER2566 harbouring pET28b-recG was inoculated into Luria-Bertani broth (Difco) supplemented with 50 mg kanamycin ml 21 , 2.5 mM betaine hydrochloride and 0.5 M sorbitol, and grown at 37 uC to OD 600~0 .4. The culture was then transferred to 18 uC and induced at OD 600~0 .6 with 0.5 mM IPTG. After overnight growth, cells were harvested, lysed and purified using a nickel-nitrilotriacetic acid agarose column as described in the QIAexpressionist protocol for native purification of His 6 -tagged proteins from E. coli (Qiagen, 2003). b-Mercaptoethanol (5 mM) was added to the lysis, wash and elution buffers as indicated in the protocol. After elution from the column, the eluates containing RecG Mtb were pooled and dialysed overnight against buffer comprising 50 mM NaH 2 PO 4 (pH 7.5), 300 mM NaCl and 1 mM DTT. The N-terminal His 6 -tag was cleaved off using biotinylated thrombin (Novagen) following the manufacturer's protocol. Further purification was carried out on a HiTrap Q HP column (GE Healthcare) after the buffer had been exchanged with 20 mM Bistris (pH 7.2), 100 mM NaCl and 1 mM DTT. Fractions containing pure RecG Mtb were pooled and dialysed against storage buffer [20 mM Bistris (pH 7.5), 300 mM NaCl, 1 mM DTT, 20 % glycerol (w/v)] and stored at 280 uC until use. RecG Mtb protein concentration was determined using the Bradford method (Bio-Rad) using BSA as standard.
Model DNA substrate preparation for DNA binding, unwinding and ATPase assays. DNA substrates were prepared essentially as described by Brosh et al. (2006) with some modifications. Briefly, oligonucleotides were 59 end-labelled using [c-32 P]ATP (PerkinElmer) and T4 PNK enzyme (NEB) for 1 h at 37 uC, followed by enzyme inactivation at 65 uC for 20 min. Unincorporated ATPs were removed using illustra MicroSpin G-25 columns (GE Healthcare). Labelled and unlabelled complementary oligonucleotides were mixed at a molar ratio of 1 : 2.5 in annealing buffer [40 mM Tris/HCl (pH 8.0), 50 mM NaCl], denatured at 95 uC for 5 min, and allowed to cool to room temperature for about 3 h. The annealed products were resolved on an 8 % non-denaturing polyacrylamide gel. The bands containing the completely annealed substrates were excised and DNA was eluted into buffer comprising 10 mM Tris/HCl (pH 8.0) and 0.5 mM EDTA by incubating overnight at 4 uC. The concentrations of the eluted DNA substrates were estimated as described by Brosh et al. (2006) and are given in moles of substrate molecules. For ATPase assays, DNA cofactors employed were prepared by annealing equimolar concentrations of complementary strands. Proper annealing of the prepared DNA cofactors was verified by resolving on a nondenaturing 10 % polyacrylamide gel and staining with SYBR Safe DNA Gel Stain (Invitrogen). The oligonucleotides used and the schematics of the model DNA substrates constructed are presented in Tables 1 and 2, respectively. DNA binding assays. DNA binding assays were carried out as described by Whitby & Lloyd (1998) with some modifications. Assays were performed in reactions (20 ml) containing binding buffer [50 mM Tris/HCl (pH 8.0), 5 mM EDTA, 100 mg BSA ml 21 , 6 % (w/v) glycerol, 2 mM DTT, 50 ng poly(dI-dC) ml 21 (Thermo Scientific)] and 0.1 nM of the indicated DNA substrate. Where indicated, poly(dI-dC) was omitted from the binding buffer. Reactions were initiated by adding indicated concentrations of RecG Mtb . After 15 min incubation on ice, 4 ml of 60 % (w/v) glycerol was added and immediately loaded onto a pre-cooled and pre-run (30 min) 4 % non-denaturing polyacrylamide gel (29 : 1). Electrophoresis was performed using low-ionic-strength buffer at 200 V for 5 min and at 160 V for an additional 85 min in an ice-water bath with buffer recirculation. Gels were dried, exposed to a phosphorimaging screen, visualized using a phosphorimager (Typhoon 9410, Amersham Biosciences) and quantified by ImageQuant TLv 2003.02 software (GE Healthcare).
Helicase assays. Unless otherwise specified, all helicase assays were conducted in a 20 ml reaction containing helicase buffer [20 mM Tris/ HCl (pH 7.5), 2 mM MgCl 2 , 2 mM ATP, 100 mg BSA ml 21 , 2 mM DTT], 0.5 nM DNA substrate and the indicated concentrations of RecG Mtb . For divalent metal cofactor studies, MgCl 2 in the aforementioned buffer was replaced by 2 mM of the indicated metal chloride. Similarly, in fuel preference studies, 2 mM of the tetrasodium salt of the indicated NTP/dNTP was used in place of ATP in the helicase buffer. Helicase reactions were initiated by adding RecG Mtb and, after incubating at 37 uC for 30 min, were terminated with 10 ml of 36 helicase stop solution [50 mM EDTA, 40 % (w/v) glycerol, 0.9 % SDS, 0.1 % bromophenol blue, 0.1 % xylene cyanol] containing a 10-fold molar excess of trap oligonucleotide. For helicase time-course assays, the reaction was scaled up to 140 ml and RecG Mtb was added into preincubated (3 min) reaction mixture at 37 uC. An aliquot (10 ml) of the reaction mixture was withdrawn at the indicated time points and mixed with 5 ml 36 helicase stop solution. All helicase reaction products were resolved by 10 % non-denaturing polyacrylamide (19 : 1) gel electrophoresis at 150 V for 2 h at room temperature using 16 Tris/borate-EDTA buffer. Gels were dried and analysed as described for DNA binding assays. The proportion of helicase substrate unwound (%) was calculated as described by Brosh et al. (2006). ATPase assays. RecG Mtb ATP hydrolysis activity was monitored by TLC, as described by Kornberg et al. (1978) with some modifications. Briefly, RecG Mtb was added to initiate a 10 ml reaction in ATPase buffer [20 mM Tris/HCl (pH 7.5), 2 mM MgCl 2 , 100 mg BSA ml 21 , 25 mM cold ATP, 0.023 nM [c-32 P]ATP, 2 mM DTT] and the indicated DNA cofactor. The reaction was incubated at 37 uC for the indicated times and terminated by adding 5 ml 0.5 M EDTA (pH 8.0). Samples (2 ml) were spotted onto TLC plates (Cellulose PEI F, Merck) at 1.5 cm intervals and resolved using a solution containing 1 M formic acid and 0.5 M LiCl. The TLC plates were air-dried, exposed to a phosphorimaging screen, imaged and quantified as described above for the DNA binding assays. The proportion of hydrolysed ATP (%) was calculated as {counts for c-32 P i /(counts for c-32 P i +counts for [c-32 P]ATP)}6100. The values obtained from samples lacking RecG Mtb were subtracted from the samples containing RecG Mtb to account for background ATP hydrolysis. An unpaired Student's t test was used to determine statistical significance.
To determine the steady-state kinetic parameters of ATP hydrolysis, a 20 ml reaction was set up with ATPase buffer [20 mM Tris/HCl (pH 7.5), 4 mM MgCl 2 , 100 mg BSA ml 21 , cold ATP, 0.023 nM [c-32 P]ATP, 2 mM DTT], 125 ng plasmid DNA (pET28b) ml 21 and 150 nM RecG Mtb . The cold ATP concentration was varied between 100 and 800 mM. The reactions were incubated at 37 uC for 10 min and quenched with 10 ml 0.5 M EDTA (pH 8.0). The concentration of hot ATP was negligible and thus not considered in the calculations. The velocity data points versus cold ATP concentrations were non-linearly fitted to Michaelis-Menten and Hill equations using Prism 5 software (GraphPad).

RecG is conserved in bacteria and is present in vascular plants
The full-length E. coli RecG protein sequence was used as a query to search NCBI protein sequence databases (Sayers et al., 2012) for conserved homologues in bacteria, archaea, and plants and other eukaryotes. Homologues of recG were found to be present in the genomes of most bacteria, except Chlamydiae and Mollicutes, as reported earlier (Rocha et al., 2005;Sharples et al., 1999). However, recG homologues were not found in any of the .90 archaeal genomes in the current version of the database (Sayers et al., 2012). Notably, full-length recG was also present in the genomes of most vascular plants, including Arabidopsis thaliana, the castor oil plant (Ricinus communis), common grape vine (Vitis vinifera), California poplar (Populus trichocarpa) and  (Mahdi et al., 2003) (Fig. 1).

RecG Mtb binds a wide variety of DNA substrates
RecG Mtb bound to a wide variety of model DNA substrates, including partial and complete replication forks, HJ, bubble, and D-and R-loop substrates (Fig. 2a, b, Table 2), with the highest affinity for HJs (Fig. 2c). In contrast, the affinity of RecG Mtb for a linear DNA duplex (49 bp) was very low, and it did not bind DNA substrates containing 20 nt 59 or 39 overhangs, in the presence of poly(dI-dC) competitor (Fig. 2a, c, Table 2). When we analysed the binding affinity of RecG Mtb to homopolymeric nucleotides (40 nt) in the absence of poly(dI-dC) competitor, poly(dA) showed very weak binding compared with poly(dC), poly(dG), poly(dT) and random nucleotides (dN) (Fig. 2d). However, the binding affinity of RecG Mtb to poly(dA) appeared to increase with increasing length of nucleotides as for poly(dT) and poly(dA : dT), yet with less stable protein-DNA complexes ( Fig. 2e; data not shown). The binding activity of RecG Mtb was not influenced by the presence or absence of ATP, ADP or ATPcS (see Fig. S1 available with the online version of this paper).
RecG Mtb unwinds DNA replication forks, D-loops, R-loops and HJ substrates The unwinding activity of RecG Mtb was examined using a variety of DNA substrates, including flayed DNA duplex, and lagging, leading and complete replication fork substrates. RecG Mtb did not exert any unwinding activity on flayed DNA duplex, but demonstrated weak and strong unwinding activities on leading and lagging strand replication forks, respectively (Fig. 3a-c). Moreover, RecG Mtb unwound both strands of a complete replication fork substrate (Fig. 3d). These results suggested that RecG Mtb , like E. coli RecG, requires more than one duplex arm to unwind a three-way junction (Whitby & Lloyd, 1998 RecG Mtb also unwound an HJ substrate with a 12 bp central homologous 'movable core', producing flayed duplexes (Fig. 4a). To determine the direction of RecG Mtb -mediated branch migration of the HJ, RecG Mtb was challenged with an HJ substrate containing a 16 nt extension on one of the four duplex arms (Fig. 4b). Interestingly, RecG Mtb appeared to drive branch migration bidirectionally. The time course of HJ substrate unwinding indicated that nearly half (47 %) of the substrate was converted to flayed duplex by RecG Mtb within 1 min (Fig. 4c).
RecG Mtb also unwound a variety of synthetic D-and Rloop structures. These included D-loops without tail (substrate J), D-loops with 59 or 39 tails (substrates L and K, respectively), D-loops with a hairpin end at the 59 or 39 tail (substrates N and M, respectively), and R-loops with 59 and 39 (29-O-methyl-RNA) tails (substrates P and O, respectively) (Fig. 5a, b). The 29-O-methyl modification was introduced to protect the RNA against nuclease degradation (Inoue et al., 1987). The ability of RecG Mtb to unwind D-and R-loops regardless of the absence or presence of a tail suggests that RecG Mtb may preferentially interact with such DNA structures at the junction. RecG Mtb might also pull one side or a segment of the duplex arm of the D-or R-loop structure through the wedge-containing domain, thereby stripping the invading strand off the loop structure, as previously proposed for unwinding of the lagging strand from a partial replication fork (Rudolph et al., 2010b;Singleton et al., 2001).
We determined the efficiency with which RecG Mtb unwound HJ, replication fork and D-loop substrates, and identified HJ as the preferred DNA substrate of RecG Mtb helicase, followed by the lagging strand replication fork (Fig. 5c). These two DNA substrates were also preferentially bound by RecG Mtb (Fig. 2c). On the other hand, RecG Mtb did not unwind blunt-end DNA duplex (Fig. S2a), 39-or 59-tailed DNA duplex (Fig. S2b, c), or a bubble substrate (Fig. 5b).

Divalent metal ion and nucleotide requirements for RecG Mtb unwinding activity
The RecG Mtb helicase was active in the presence of magnesium, manganese, copper, iron or cobalt ions, but completely inactive in the reactions lacking divalent cations (Fig. 6a). No significant difference was observed in RecG Mtb unwinding activity in the presence of Mn 2+ or Mg 2+ (P50.125). RecG Mtb unwound DNA substrates in the presence of ATP or dATP, but was inactive in the presence of other NTP/dNTPs (Fig. 6b). RecG Mtb unwinding activity was significantly higher in the presence of ATP than dATP (P50.023). Furthermore, ADP and the slowly  hydrolysable ATP analogue ATPcS did not support the unwinding activity of RecG Mtb (Fig. S3).

RecG Mtb is a DNA-dependent ATPase
The ATPase activity of RecG Mtb was measured in the presence of 50 nM ssDNA (49 nt), dsDNA (49 bp) or an HJ substrate (assembled from four~49 nt oligos). The efficiency of ATP hydrolysis was 68, 32 and 0 % in the presence of HJ, dsDNA and ssDNA, respectively (Fig. 7a). Moreover, no ATP hydrolysis was observed in the reactions containing 20-100 nt poly(dT) (data not shown). We also tested ATP hydrolysis in the presence of 10 nM circular DNA cofactors, M13mp18 (ssDNA) or pET28b (dsDNA); pET28b (51 %) stimulated threefold higher ATP hydrolysis than M13mp18 (16 %) (Fig. 7b). To avoid intramolecular duplex formation, all the ssDNA cofactors tested in this study were heated to 95 u C for 5 min immediately before use.
Steady-state kinetic analysis of RecG Mtb ATPase activity was performed by titrating ATP in the presence of saturating circular dsDNA (pET28b). Under these conditions, ATP hydrolysis was linear for at least 15 min (data not shown). Michaelis-Menten (hyperbolic) curve-fitting

DISCUSSION
The M. tuberculosis genome is susceptible to the effects of genotoxic and general cellular stress, including nitrosative and oxidative damage to DNA, RNA and other biomolecules (Stallings & Glickman, 2010;Warner & Mizrahi, 2006), owing to the harsh internal environment, the human macrophage, in which it usually resides. Mechanisms that promote genome maintenance and function are likely to be essential for M. tuberculosis survival and virulence, because persistent unrepaired DNA damage can completely block replication of the genome (Masai et al., 2010;Mirkin & Mirkin, 2007). RecG is an important enzyme widely thought to play a role in remodelling replication forks stalled at DNA lesions, mediating replication restart via fork regression . In the present study, the biochemical activities of RecG Mtb are characterized, providing considerable insight into the potential role(s) of RecG in DNA/nucleic acid metabolism in an intracellular pathogen.
This study demonstrated that RecG Mtb binds and unwinds a variety of DNA substrates that mimic intermediates in DNA replication, recombination and repair, like its E. coli orthologue. Among the substrates examined here, RecG Mtb had the highest affinity for HJ, while 39-and 59-overhang DNA substrates and blunt-end duplex DNA substrates were bound very poorly (Table 2). In general, the binding and unwinding activity of RecG Mtb required protein concentrations that were considerably higher than those described for E. coli RecG. Whereas E. coli RecG shifted an HJ substrate at protein concentrations of 0.1 nM (Briggs et al., 2005), the same shift was only observed with 0.25 mM RecG Mtb . This might be due to suboptimal assay conditions. Generally, M. tuberculosis helicases exerted their activities at concentrations .100 nM (Balasingham et al., 2012;Biswas et al., 2009;Curti et al., 2007).
Unexpectedly, RecG Mtb had very low affinity for poly(dA), and much higher affinity for poly(dT), poly(dG) and poly(dC). Generally, RecG Mtb and other superfamily 2 helicases make extensive contact with the sugar-phosphate DNA backbone and this interface is the dominant functional mode of interaction (Büttner et al., 2007;Kim et al., 1998;Pyle, 2008;Singleton et al., 2001). DNA helicases also tend to exhibit low DNA sequence specificity, presumably because higher sequence specificity might hinder the translocation and/or processivity of the helicase (Rocak & Linder, 2004;Tanner & Linder, 2001;Tuteja & Tuteja, 2004). However, a recent report has revealed that Vaccinia viral helicase NPH-II, a superfamily 2 helicase, favours purine-rich over pyrimidine-rich dsDNA helicase substrates (Taylor et al., 2010). The sequence bias of RecG Mtb reported here might be an intrinsic property of this helicase; however, this conclusion is preliminary and requires additional investigation.
Notably, another interesting observation reported here is that the affinity of RecG Mtb for ssDNA and dsDNA is length-dependent. RecG Mtb had higher affinity for longer oligomers (¢40 nt) than for shorter oligomers (20 nt RecG Mtb required Mg 2+ for its optimal activity, although unwinding activity is also supported by Mn 2+ , Cu 2+ , Co 2+ and Fe 2+ . A similar observation was also reported for another mycobacterial helicase, UvrD (Curti et al., 2007). This property, shared by these two M. tuberculosis helicases, could contribute to the pathogenicity of M. tuberculosis, because Mg 2+ is scarce in the phagosomes of macrophages (Groisman, 1998). The observation that the ATPase activity of RecG Mtb was stimulated to a greater extent in the presence of dsDNA than of ssDNA suggests that the enzyme may translocate on dsDNA, as does E. coli RecG (Mahdi et al., 2003).
Evidence shows that expression of recG in M. tuberculosis is upregulated in infected human cells and mouse macrophages, suggesting that RecG may actively promote virulence and/or pathogenicity during infection of mammalian cells (Davis & Forse, 2009;Rachman et al., 2006;Schnappinger et al., 2003). It is also interesting that recG is conserved in the related human pathogen Mycobacterium leprae, in which there is an extreme case of reductive evolution (Vissa & Brennan, 2001). This suggests a potentially important metabolic role for RecG in other mycobacteria also. The genotoxic stress that M. tuberculosis encounters inside the macrophage with reactive nitrogen and oxygen species is very different from that to which E. coli cells are exposed in their various environmental niches. Thus, the metabolic conditions inside an intracellular pathogen such as M. tuberculosis might be considerably different from those of E. coli cells and other model species. This is exemplified by the facts that the genome of M. tuberculosis comprises an unusually high number of genes involved in lipid metabolism (.233), that its genome has a high G+C content (Cole et al., 1998), and that there is a lack of MutS-based mismatch repair in M. tuberculosis (Mizrahi & Andersen, 1998). The existence of a nonhomologous end-joining pathway (Della et al., 2004) as well as an alternative regulatory mechanism for DNA damage-inducible genes (Davis et al., 2002) have also been indicated in M. tuberculosis. A recent study further has indicated that RuvAB of M. tuberculosis, unlike E. coli RuvAB, can convert replication forks to HJs (Khanduja & Muniyappa, 2012). Moreover, biochemical characterization of DNA repair components indicates that oxidative DNA glycosylases of M. tuberculosis exhibit substrate preferences different from their E. coli counterparts (Guo et al., 2010). Taken together, these findings suggest that the DNA metabolism of M. tuberculosis might differ considerably from that of E. coli.

Conclusions
The novel findings presented here that RecG exists in vascular plants and algae, in addition to eubacteria, and that RecG Mtb preferentially binds relatively long ssDNA, exhibiting a higher affinity for poly(dT), poly(dG) and poly(dC) than for poly(dA), shed new light on the occurrence and role of RecG in nature. Furthermore, the finding that the preferred helicase substrate for RecG Mtb is HJ, a key intermediate in DNA repair, recombination and replication fork restart (Kepple et al., 2005;Liu & West, 2004), may suggest that RecG Mtb is preferentially involved in such processes in vivo. However, future studies involving M. tuberculosis recG-null mutants are needed to clarify the precise role of RecG in the DNA metabolism, survival, fitness and virulence of M. tuberculosis and possibly of other related mycobacteria.