Formation of a Stable RuvA Protein Double Tetramer Is Required for Efficient Branch Migration in Vitro and for Replication Fork Reversal in Vivo*

In bacteria, RuvABC is required for the resolution of Holliday junctions (HJ) made during homologous recombination. The RuvAB complex catalyzes HJ branch migration and replication fork reversal (RFR). During RFR, a stalled fork is reversed to form a HJ adjacent to a DNA double strand end, a reaction that requires RuvAB in certain Escherichia coli replication mutants. The exact structure of active RuvAB complexes remains elusive as it is still unknown whether one or two tetramers of RuvA support RuvB during branch migration and during RFR. We designed an E. coli RuvA mutant, RuvA2KaP, specifically impaired for RuvA tetramer-tetramer interactions. As expected, the mutant protein is impaired for complex II (two tetramers) formation on HJs, although the binding efficiency of complex I (a single tetramer) is as wild type. We show that although RuvA complex II formation is required for efficient HJ branch migration in vitro, RuvA2KaP is fully active for homologous recombination in vivo. RuvA2KaP is also deficient at forming complex II on synthetic replication forks, and the binding affinity of RuvA2KaP for forks is decreased compared with wild type. Accordingly, RuvA2KaP is inefficient at processing forks in vitro and in vivo. These data indicate that RuvA2KaP is a separation-of-function mutant, capable of homologous recombination but impaired for RFR. RuvA2KaP is defective for stimulation of RuvB activity and stability of HJ·RuvA·RuvB tripartite complexes. This work demonstrates that the need for RuvA tetramer-tetramer interactions for full RuvAB activity in vitro causes specifically an RFR defect in vivo.

The RuvAB complex is a highly sophisticated molecular machine, which carries out branch migration of Holliday junctions during homologous recombination. RuvA binds specifi-cally to four-armed Holliday junctions (HJ) 4 and guides the assembly of two RuvB hexameric rings onto diametrically opposite arms of the HJ. RuvB, an AAA ϩ ATPase (1), is the motor that drives branch migration of the crossover point (1)(2)(3). After branch migration, a dimer of RuvC resolves the HJ by making two sequence-specific symmetrical cuts, producing either patched or spliced linear products (4 -6). Genetic studies showed that RuvC cannot function in vivo in the absence of RuvAB (7,8), and it has been proposed that an RuvABC complex, known as the resolvasome, allows RuvC to scan for cleavable sequences (3, 9 -11). RuvA binds to HJs in vitro as one tetramer (complex I) or two tetramers that sandwich the junction (complex II) in a concentration-dependent manner; however, it is not clear whether the RuvAB complex contains one or two tetramers of RuvA in vivo (12)(13)(14)(15)(16)(17)(18)(19)(20)(21). A RuvAB branch migration complex made of two RuvA tetramers would prevent access of RuvC to the Holliday junction. Whether to form the resolvasome the RuvC dimer displaces one of the two RuvA tetramers present in the RuvAB complex, or whether the RuvC dimer simply binds opposite a single RuvA tetramer present in the complex is a currently unanswered question.
In addition to processing Holliday junctions in homologous recombination, RuvAB plays an important role upon DNA replication inactivation. In certain E. coli replication mutants, stalled replication forks undergo a process known as replication fork reversal (RFR) (22). As the stalled fork is reversed, the newly synthesized strands are unwound from the daughter duplexes and base pair to form a Holliday junction, known as a reversed fork. Branch migration of the reversed fork by RuvAB drives extrusion of an arm with a duplex end, which allows entry of RecBCD. RecBCD can either reset the fork by degrading the duplex or load RecA, which carries out invasion and strand exchange with the homologous duplex at the replication fork, to reset a new fork. Both pathways result in PriA-dependent replication restart (23). Intriguingly, RuvAB are actually required for replication fork reversal to occur in dnaEts, holD Q10am and rep E. coli mutants (24). Yet in vitro, RuvAB preferentially unwinds synthetic replication forks in a direction that is opposite the direction for fork reversal (25). RuvAB could reverse model replication forks in vitro if RuvB was only allowed to form one hexameric ring on the parental duplex of the fork (25). It has been speculated that in vivo the asymmetric binding of a single RuvA tetramer onto a three-armed fork may result in asymmetric loading of a single RuvB hexamer onto the parental duplex (24). Alternatively, cellular factors may force RuvA to load RuvB in this manner.
In an RuvA octamer, the two tetramers do not only interact with DNA but also with each other through four contacts involving domain II of each monomer. Specifically, six ionic interactions form between the ␣-helix 6 in domain II of each opposite monomer resulting in four points of contact between tetramers (supplemental Fig. 1) (18). The role of RuvA octamers for efficient branch migration has been investigated using RuvA mutants designed to disrupt the tetramer-tetramer interface and prevent complex II formation. A triple Escherichia coli RuvA mutant, RuvA3m, was unable to form complex II at RuvA concentrations of up to 2 M and was deficient in processing synthetic HJs in vitro and in vivo (26). Unexpectedly, RuvA3m helicase activity and branch migration of Y-junctions in vitro appeared unaffected, and it was proposed that complex I was able to support one RuvB hexameric ring, but complex II was needed to assemble two hexameric rings on the Holliday junction (26). A Thermus thermophilus "tetramer-only" RuvA mutant, RuvA(DK) (L125D and E126K) was also studied. Electron microscopy demonstrated that a single RuvA(DK) tetramer formed tripartite complexes containing two hexameric rings of RuvB on the HJ (27). RuvA(DK) displayed reduced ability to promote branch migration of Holliday junctions in vitro and, significantly, could not support branch migration with a single RuvB hexamer (28). The capacity for replication fork reversal of these tetramer-only mutants has not been tested.
Recently, two ruvA mutants called ruvAz3 and ruvAz87 (H29R/K129E/F140S and N79D/N100D, respectively) were isolated and characterized as separation-of-function mutants that can process Holliday junctions but cannot reverse replication forks (29). The RuvAz proteins contain several mutations in different domains of the proteins, making it difficult to ascertain the molecular cause for their phenotypes. One intriguing observation is the inability of RuvAz mutants to form complex II on Holliday junctions, which is common with the RuvA3m and RuvA(DK) mutants discussed above. However, the tetramer-only RuvAz mutants are fully capable of promoting homologous recombination in vivo, whereas the tetramer-only RuvA3m is not.
To understand how the ability of RuvA to form octamers (complex II) relates to branch migration of HJs and replication fork reversal in vitro and in vivo, we have designed a new tetramer-only RuvA mutant, RuvA2 KaP . RuvA3m, which was used in previous work (26), is inactive in vivo and displays significant nonspecific DNA binding, which could explain some discrepancies compared with studies of RuvA(DK) (28). In contrast, RuvA2 KaP was designed to only disrupt tetramer-tetramer interactions. In this study, we have investigated the ability of RuvA2 KaP to bind and process HJs and synthetic forks in vitro and how these correlate with the ability of the mutant to process different substrates in vivo. We also compare certain biochemical activities of RuvA2 KaP mutant with those of the separation-of function mutants RuvAz3 and RuvAz87.

EXPERIMENTAL PROCEDURES
Mutagenesis of RuvA2 KaP -To test RuvA2 KaP function in vivo, plasmids pGB-ruvA and pGB-ruvAB were mutagenized using mutagenic primers 259F and 260R (supplemental Table  2). Both of these primers incorporated a codon change from GAA to CGC at position 379 in the coding sequence. Site-directed mutagenesis (QuikChange II site-directed mutagenesis kit) was used to produce pGB-ruvA1 KaP and pGB-ruvA1 KaP B, which contained the single amino acid substitution E127R. The thermal cycling procedure employed was as follows: 1 time at 95°C for 30 s, 16 cycles of 95°C for 30 s, 53°C for 45 s, and 68°C for 10 min. To produce pGB-ruvA2 KaP and pGB-ruvA2 KaP B, which contain two amino acid substitutions E127R and K119A, pGB-ruvA1 KaP and pGB-ruvA1 KaP B were used as templates, and the primers 401F and 402R were used to introduce the second amino acid substitution K119A. Both primers 401F and 402R incorporated a codon change from AAA to GCA at position 355 in the coding sequence. The mutagenesis was carried out as described to introduce the first substitution (E127R). Sequencing of the pGB-ruvA1 KaP , pGB-ruvA2 KaP , pGB-ruvA1 KaP B, and pGB-ruvA2 KaP B constructs was carried out by MWG-Biotech using the primers RkF, RkR, and internal primers RkiF and RkiR (supplemental Table 2).
Recombinant Protein Production-All protein fractions were diluted with loading buffer (0.125 M Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 0.01% bromophenol blue, 10% ␤-mercaptoethanol) and analyzed by SDS-PAGE analysis. RuvA or RuvA2 KaP was cloned into pET21a, expressed in BL21-GOLD(DE3), and purified following a protocol described previously (26) with several modifications. For each protein, four frozen pellets (total volume 4 ml) were thawed on ice and resuspended in 21 ml of Lysis buffer (100 mM Tris-HCl, pH 8, 5% glycerol, 2 mM EDTA, 1 mM of DTT, and 1 mg⅐ml Ϫ1 of lysozyme) and incubated for 30 min at 4°C. A final concentration of 1 M NaCl and 0.1% Triton X-100 was added to the solution and incubation continued for a further 10 min on ice. The solution was made up to 0.4% sodium deoxycholate and spun at 42,000 rpm for 60 min at 4°C in a Type 70 Ti rotor in an Optima TM L-100 XP ultracentrifuge (Beckman Coulter). The supernatant was dialyzed against 2 liters of TEGD buffer (20 mM Tris-HCl, pH 8, 1 mM EDTA, 0.5 mM DTT, 10% glycerol).
The crude lysate was loaded onto a DEAE column, and the column was washed with TEGD buffer, and a gradient of 0 -500 mM KCl in TEGD buffer was employed to elute the protein. The eluted protein was dialyzed against 2 liters of H buffer (10 mM KP i , pH 6.8, 150 mM KCl, 0.5 mM DTT, 10% glycerol) at 4°C. Dialyzed protein was loaded onto a hydroxylapatite column, and the column was washed with H buffer. A 100-ml 10 -600 mM gradient of KP i in H buffer was used to develop the column. Eluted RuvA was dialyzed at 4°C against 2 liters of H buffer supplemented with 5 mM ␤-mercaptoethanol, 0.1 mM PMSF, and 100 mM KCl with no DTT. Dialyzed RuvA was loaded onto a HiTrap heparin column that was washed with buffer, and a 10 -600 mM KP i gradient was used to elute the protein that was dialyzed against 2 liters of TEGD buffer at 4°C. RuvA was loaded onto a single strand DNA column, and the column was washed with TEGD buffer and developed with a 0 mM to 1 M KCl gradient. Eluted RuvA protein was dialyzed overnight at 4°C against 2 liters of TEGD buffer. Dialyzed RuvA protein was loaded onto a 1-ml Mono Q column that was washed with TEGD buffer, and a gradient 0 -1.5 M KCl in TEGD buffer was used to elute the RuvA protein. The protein concentrations of the purified RuvA protein were determined using a Bradford assay. RuvA was stored as either 10 or 50% glycerol stocks at Ϫ20°C. RuvA2 KaP protein was produced using an identical procedure except that after purification on a heparin HiTrap column the protein was loaded straight onto a Mono Q column; the single strand DNA column was not used.
A protocol for producing RuvBD113E was modified to produce wild type RuvB (26). RuvB was overexpressed from plasmid pET21a in BL21-GOLD(DE3) cells. Cultures of RuvB-pET21a were grown in LB (100 g⅐ml Ϫ1 ampicillin) at 37°C to an absorbance of 0.6 at 600 nm. The cultures were supplemented with 1 mM isopropyl 1-thio-␤-D-galactopyranoside and incubated for 6 h at 37°C at 250 rpm. The cultures were pelleted at 4000 relative centrifugal force for 10 min in an SLA 3000 rotor in a Sorvall RC6 Plus centrifuge and resuspended in 25 ml of LB. The suspension was further spun for 10 min at 4000 relative centrifugal force at 4°C, and the resulting pellets were frozen at Ϫ20°C. Purification was carried out as described previously (26). RuvC was expressed and purified using a protocol as described previously (5).
Size Exclusion Chromatography (SEC)-RuvA, RuvA mutants, and RuvB proteins were dialyzed against TEGD supplemented with 0.1 M NaCl overnight at 4°C. A total volume of 200 l of 250 g of each protein was applied to the 25 ml of Superose 6 TM 10/30 GL column. The proteins were eluted from the column in TEGD buffer supplemented with 0.1 M NaCl at a flow rate of 0.3 ml⅐min Ϫ1 . Molecular weights of species were estimated by comparison with five molecular weight standards (Bio-Rad). Fractions were analyzed by SDS-PAGE analysis, and the UV absorbance profiles of the eluted proteins were recorded.
DNA Substrate Preparation-The DNA substrate X12 (HJ) was constructed (see supplemental Table 1) as described previously (26). Replication fork-like substrates F2 and HJY3m were also constructed. JBM5a and IT01 were synthesized with a fluoro tag IRD700 attached to the 5Ј end.
The required combinations of oligonucleotides were included in annealing reactions, using 1 g of each oligonucleotide in SSC buffer (150 mM sodium chloride and 15 mM trisodium citrate, pH 7.0) incubated at 95°C for 2 min and slowly annealed by cooling the heat block to room temperature. Substrates were purified by electrophoresis on an 8% native gel in TBE buffer run at 15 V⅐cm Ϫ1 for 80 min at 4°C. Gel bands were cut from the gel, and DNA was eluted from the gel pieces by electrophoresis in 0.5ϫ TBE buffer for 1 h using a BioTrap multikit (Schleicher & Schuell). Each DNA substrate was then dialyzed against 2 liters of DNA storage buffer (10 mM Tris-HCl, pH 8, 1 mM EDTA, 50 mM NaCl) and stored at Ϫ20°C.
EMSA-EMSA reactions containing the indicated amount of protein and DNA substrate were incubated in DNA binding buffer (50 mM Tris-HCl, pH 8, 0.5 mM EDTA, 1 mM DTT, 100 g⅐ml Ϫ1 BSA, 6% glycerol). Proteins were diluted using Dilu-tion buffer (20 mM Tris-HCl, pH 8, 150 mM NaCl, 0.5 mM DTT, 10% glycerol). Reactions were incubated for 5 min on ice. For gel loading, 5 l of 80% glycerol was added to each 20-l reaction to allow the samples to sink into the wells; no dye was used as the dye interferes with detection of the signal. In some instances, protein⅐DNA complexes were resolved on native gels of different concentrations of polyacrylamide as follows: between 4 and 10% in TBE buffer or in low ionic TAE buffer (6.7 mM Tris-HCl, pH 8.1, 2 mM EDTA, 3.2 mM sodium acetate) at 6 V⅐cm Ϫ1 at 4°C in a Bio-Rad miniprotean II gel system. The gels were run at 10 V⅐cm Ϫ1 for 4 h at 4°C. Gels were scanned using the Odyssey infrared imaging system (from LI-COR Biosciences) at 700 nm at an intensity of 10.
For EMSAs in Mg 2ϩ , reactions were carried out exactly as for DNA binding assays in EDTA, except EDTA was omitted from the DNA Binding buffer and replaced by 5 mM MgCl 2 . The samples were loaded after addition of 5 l of 80% glycerol and resolved using 6% polyacrylamide native gels with the EDTA omitted from the gel and replaced with 5 mM MgCl 2 . The gels were run in 0.5ϫ TBM buffer (45 mM Tris base, 45 mM boric acid, 200 M MgCl 2 ) at 6 V⅐cm Ϫ1 for 4 h with recircularization of the buffer on ice.
Branch Migration Assays-Reactions were carried out in branch migration buffer (20 mM Tris-HCl, pH 7.5, 15 mM MgCl 2 , 2 mM ATP, 2 mM DTT and 100 g⅐ml Ϫ1 BSA). DNA and then protein were added to the reactions at the required concentrations. The reaction was incubated at 37°C for 30 min, and 5 l of 5ϫ Stop buffer (100 mM Tris-HCl, pH 8, 200 mM EDTA, 2.5% SDS, and 10 mg⅐ml Ϫ1 proteinase K) was then added to each 20-l reaction with incubation at room temperature for 10 min. 5 l of 80% glycerol was added to each sample, and the samples were loaded onto 6% polyacrylamide native gels and run in 0.5ϫ TBE buffer at 7 V⅐cm Ϫ1 for 2 h. DNA products were visualized on the Odyssey infrared imaging system at 700 nm at intensity 10.
RuvA⅐B-DNA Complex Formation Assay-Reactions were incubated at 37°C for 30 min in formation buffer (20 mM triethanolamine-HCl pH 7.5, 10 mM MgCl 2 , 0.25 mM ATP␥S, 1 mM DTT, 50 g⅐ml Ϫ1 BSA). The proteins and DNA were added as required and were fixed by adding 0.2% glutaraldehyde and incubating reactions for 20 min at 37°C. 5 l of 80% glycerol were added to each reaction, and samples were loaded onto a 6% polyacrylamide native gels and run in 1ϫ TAE buffer. Gels were run at 10 V⅐cm Ϫ1 for 2 h at room temperature and scanned on the LI-COR fluoro imaging system at 700 nm intensity 10.
RuvC Cleavage Assay-Reactions were carried out in cleavage buffer (50 mM Tris-HCl, pH 8, 10 mM MgCl 2 , 50 mM KCl, 5 mM ␤-mercaptoethanol, 100 g⅐ml Ϫ1 BSA and 5% glycerol). 2 l of RuvA protein and 2 l of RuvC protein were mixed and incubated with HJ in cleavage buffer at 37°C for 1 h. 5 l of 5ϫ stop buffer was then added to the 20-l reactions, which were incubated at 37°C for a further 10 min. 5 l of 80% glycerol was added to 25-l reactions that were loaded onto 6% native polyacrylamide gels and run in TBE buffer at 10 V⅐cm Ϫ1 at room temperature. DNA products were visualized using the Odyssey LI-COR fluoro imaging system.
ATPase Assay-The indicated amounts of protein and DNA substrate were mixed, and reactions were performed in ATPase buffer (50 mM Tris-HCl, pH 7.5, 50 mM NaCl, 15 mM MgCl 2 , 0.1 mg⅐ml Ϫ1 BSA, 1 mM DTT, 15 mM MgCl 2 , and 0.5 mM ATP). A Malachite green kit was used to detect activity; the reactions were preincubated on ice, and a zero time point was taken with 20 l of reaction added to 5 l of 0.5 M EDTA. Reactions were then incubated at 37°C with time points taken at 5, 15, and 30 min. 150 l of ALS mix (Innova Biosciences) was added to each well, and the reactions were incubated for 30 min at room temperature. The 96-well plates were scanned in a Tecan Sunrise fluorometer and analyzed with software by Magellan. A calibration curve of 10 different KP i concentrations was used to determine the amount of phosphate released in each reaction.
Measurement of Recombinational DNA Repair-UV irradiation was performed as described previously (29). The survival ratio was a comparison of colonies that grew on a replica control plate compared with colonies that grew on a plate that was irradiated. For mitomycin C treatment, cells were grown at 37°C in LB to an A 600 ϭ 0.5, and mitomycin C was added to the culture at a final concentration of 2 g/ml, and incubation was continued at 37°C for 90 min. An untreated culture was used as control. Appropriate dilutions were plated on LB plates and incubated overnight at 37°C. Ratios of cfu of mitomycin C treated/cfu of untreated cells were calculated.
Measurement of Conjugational Recombination-Conjugations were performed as described using JJC145 as Hfr donor (29); donor and recipient cells were mixed for 25 min. Selective medium was M9 minimal medium supplemented with leucine, proline, threonine, and arginine (2% final concentration each) and 10 g⅐ml Ϫ1 chloramphenicol.
Measure of Linear DNA by Pulse Field Gel Electrophoresis-Quantification of pulsed field gels was performed using in vivo [ 3 H]thymidine-labeled chromosomes as described previously (22).

RuvA2 KaP Binds Efficiently to Holliday Junctions as a Single
Tetramer-A set of RuvA mutants was generated by introducing amino acid substitutions in ␣-helix-6 in domain II, which is the region involved in tetramer-tetramer interactions. Specifically, the RuvA2 KaP used in this study carried E127R and K119A mutations in the tetramer-tetramer interface (supplemental Fig. 1).The overall electrostatic charge of this region changes only slightly compared with wild type, and the slightly more basic/positive charge does not alter RuvA2 KaP interaction with DNA (Fig. 1A). The mutant RuvA3m used in a previous study formed aberrant complexes on HJs, probably caused by the significant increase in positive charge in ␣-helix-6 (26).
The ability of RuvA2 KaP to form stable complexes was tested in vitro; RuvA2 KaP was analyzed on SDS-polyacrylamide gels and formed tetramers comparable with wild type RuvA, which is stable enough not to dissociate in SDS (supplemental Fig. 2A). When boiled and loaded onto an SDS-polyacrylamide gel, RuvA2 KaP dissociated into monomers, indicating that, like RuvA, the mutant forms stable tetramers that only dissociate upon boiling (supplemental Fig. 2A). RuvA2 KaP was analyzed by SEC on a 25-ml Superose 6 TM 10/30 GL column, and the elution profile of RuvA2 KaP was comparable with RuvA (supplemental Fig. 2B). Additionally, RuvA2 KaP mixed with RuvB was loaded onto a 25-ml Superose 6 TM 10/30 GL column, and the formation of the RuvA2 KaP B complex was equivalent to RuvAB complex formation (supplemental Fig. 2C). In conclusion, RuvA2 KaP forms a stable tetramer and is able to form a complex with RuvB in solution with an efficiency comparable with that of wild type RuvA.
The binding of RuvA2 KaP to a fluoro-tagged synthetic HJ (X12) was tested using EMSAs (Fig. 1A). In EDTA-containing buffer, RuvA formed both complex I and II at lower concentrations but exclusively formed complex II at protein concentrations of 150 nM and higher. Conversely, RuvA2 KaP only formed complex I, even at protein concentrations as high as 2 M. The amount of HJ bound by RuvA2 KaP was similar to wild type RuvA (Fig. 1B), indicating that the mutations did not affect the affinity of RuvA2 KaP for HJs. Holliday junction binding was further tested in the presence of Mg 2ϩ as RuvB requires at least 5 mM of Mg 2ϩ for efficient ATP hydrolysis and branch migration. In 5 mM MgCl 2 , wild type RuvA bound to X12 exclusively as complex II, even at low protein concentrations, leaving a significant amount of free junction (Fig. 1C). RuvA2 KaP formed complex I at concentrations up to 75 nM, but at 250 nM of RuvA2 KaP and above, complex II also formed (Fig. 1C). The overall RuvA2 KaP binding of X12 in Mg 2ϩ was comparable with RuvA ( Fig. 1D). It was clear that RuvA2 KaP complex II was stabilized by Mg 2ϩ , so we checked for a direct stabilizing effect of 5 mM MgCl 2 on tetramer-tetramer interactions using SEC. As the SEC elution profile of RuvA remained unchanged (data not shown), the effect of Mg 2ϩ was therefore dependent on the presence of DNA. Experiments in Mg 2ϩ buffer revealed that a lack of tetramer-tetramer interaction does not fully prevent complex II formation but renders it dependent on a high concentration of RuvA protein.
Stability of RuvA2 KaP Complex II in Solution-In vivo RuvC cannot function without RuvAB (7,8,29,30); however, it has been shown that formation of RuvA complex II occludes the HJ and prevents RuvC-mediated cleavage in vitro (26,31,32). To assess the stability of the RuvA2 KaP complex II in the presence of Mg 2ϩ , we compared the ability of both wild type and mutant proteins to protect a HJ from cleavage by RuvC (Fig. 2). Varying concentrations of RuvA or RuvA2 KaP were incubated with X12, followed by addition of 100 nM RuvC. Significant inhibition of RuvC HJ cleavage was observed at wild type RuvA concentrations of 100 nM, and cleavage was completely abolished at 300 nM RuvA. The inhibition of RuvC cleavage correlated well with complex II formation in EDTA binding studies but only roughly with the concentration at which complex II forms in the Mg 2ϩ binding studies. HJ cleavage by RuvC was also tested in the presence of up to 2 M of RuvA2 KaP . Cleavage was only slightly inhibited within the range of 100 -500 nM RuvA2 KaP and was still weakly observed in the presence of 2 M of RuvA2 KaP (Fig. 2A). This demonstrates that although RuvA2 KaP forms complex II on Holliday junctions, the stability of RuvA2 KaP complex II is reduced compared with RuvA complex II.
Because the RuvA2 KaP protein confers to E. coli a separationof-function phenotype (see below), we compared the previously isolated separation-of-function mutants RuvAz3 and RuvAz87 to RuvA2 KaP using the same conditions. As with RuvA2 KaP , both RuvAz mutants were unable to inhibit RuvC cleavage of the junction at concentrations up to 300 nM (Fig. 2B). The inability of RuvAz mutants to protect the HJ from RuvC cleavage confirms that they form unstable complex II (29). These results also suggest that a tetramer of RuvA bound to the HJ (complex I) does not inhibit HJ cleavage by RuvC, and it is possible that both RuvA and RuvC bind together to the junction. Alternatively, complex I might transiently dissociate from the junction allowing RuvC access, whereas complex II does not dissociate.
Stimulation of RuvB ATPase Activity by RuvA Mutants-To measure the functional interaction between RuvA2 KaP and RuvB, we tested the ability of RuvA2 KaP to stimulate the DNAdependent ATPase activity of RuvB. Time courses of ATP hydrolysis using 100 nM of RuvA or RuvA2 KaP, 500 nM RuvB, and 5 ng of X12 are shown in Fig. 3A. The DNA-dependent ATPase activity of RuvA2 KaP B was about half of the activity of wild type RuvAB; RuvAB hydrolyzed 80 mol of ATP in 30 min, whereas RuvA2 KaP B hydrolyzed 40 mol of ATP in 30 min (Fig.  3A). The ATPase activity was also measured with higher RuvA concentrations. ATP hydrolysis was 2-3-fold lower with RuvA2 KaP compared with RuvA at all concentrations at which RuvA2 KaP complex II formation was observed by EMSA (250 nM and above) (Fig. 3B). These data indicate that tetramertetramer interactions within complex II are necessary for optimal stimulation of the DNA-dependent ATPase activity of RuvB.
The ability of the RuvA2 KaP mutant to stimulate the DNAdependent ATPase activity of RuvB was compared with that of the separation of function mutants RuvAz3 and RuvAz87.   A, time course of ATP hydrolysis. 100 nM RuvA or RuvA2 KaP was incubated with 500 nM RuvB and X12 at 37°C. ATP hydrolysis is proportional to the release of inorganic phosphate (in micromoles) as quantified using a colorimetric assay. Three separate data sets were quantified, and the data were plotted as the moles of ATP hydrolyzed per mol of RuvB as a function of time. The error bars represent S.D. B, ATPase activity at two different concentrations of RuvA or RuvA2 KaP incubated with 500 nM RuvB and X12 for 30 min. The experiment and graph were produced as described in A. C, RuvA, RuvA2 KaP , RuvAz3, or RuvAz87 were either incubated alone or mixed with 500 nM RuvB. These proteins were then incubated with ATP for 10 min at 37°C in the presence of 200 ng⅐l Ϫ1 X174 virion DNA. ATP hydrolysis was indirectly measured using a colorimetric assay to detect the amount of P i released. The data were used to plot a bar chart of moles of ATP hydrolyzed per min. The error bars represent S.D.
RuvAz3, and RuvAz87 are less efficient in stimulating RuvB ATPase activity than wild type RuvA.
RuvA2 KaP Forms Unstable RuvA⅐RuvB⅐HJ Tripartite Complexes in Solution-The RuvA2 KaP mutant was found to bind RuvB in solution by SEC analysis as wild type RuvA (supplemental Fig. 2, B and C). To analyze the interactions of RuvA2 KaP with RuvB on Holliday junctions (X12), these large protein⅐DNA complexes were cross-linked with 0.25% glutaraldehyde and analyzed by native PAGE. When incubated with 5 ng of X12, 400 nM RuvA alone could be crosslinked on DNA as complex II but 400 nM RuvA2 KaP could not (Fig. 4, lanes 10 and 11). The inefficient glutaraldehyde cross-linking of RuvA2 KaP complex II on X12 reveals the reduced stability of the RuvA2 KaP tetramer-tetramer interaction across the junction. Increasing concentrations of RuvA or RuvA2 KaP were incubated with 300 nM RuvB and X12. RuvB alone could not be cross-linked to X12 under these conditions (Fig. 4, lane 12). RuvAB complexes on X12 are detected when 25 nM RuvA and above is used (Fig. 4,  lanes 1 and 2). At 200, 300, and 400 nM RuvA, complex II is also observed in addition to RuvAB complexes on DNA (Fig.  4, lanes 4 -6). The stoichiometry of RuvA:RuvB on the junction is predicted to be 8:12 monomers, respectively, and thus at equal concentrations of RuvA and RuvB, there is an excess of RuvA that forms complex II. We incubated 300 and 400 nM RuvA with 600 nM RuvB (Fig. 4, lanes 7 and 8). This altered the ratio of RuvA:RuvB to favor formation of RuvAB complexes on DNA rather than complex II (Fig. 4, lanes 7  and 8). Interestingly, only a fraction of X12 could be crosslinked with the mutant RuvA2 KaP B compared with wild type RuvAB, and a significant amount of DNA remained proteinfree (Fig. 4, lanes 13-18). Increasing the concentration of RuvB to 600 nM had no effect on tripartite RuvA2 KaP ⅐RuvB⅐HJ complex formation (Fig. 4, compare lanes 17 and 18 with 19 and 20). The cross-linking experiments demonstrate the reduced stability of RuvA2 KaP complexes with RuvB on DNA. Thus the stability of the RuvA double tetramer is crucial for the formation of a stable tripartite RuvAB⅐HJ complex.
RuvA2 KaP -RuvB-mediated Branch Migration of Synthetic Holliday Junctions-The ability of RuvA2 KaP and RuvB to promote branch migration of HJs was tested using fluoro-tagged X12, increasing concentrations of RuvA or RuvA2 KaP and 250 nM RuvB. As shown in Fig. 5A, X12 was processed efficiently by wild type RuvAB to form two branch migration products. In contrast, RuvA2 KaP was clearly deficient in branch migration with small amounts of products observed at RuvA2 KaP concentrations of 75 nM and above. Around 75% of X12 was processed at 100 nM wild type RuvA, whereas less than 25% of X12 was dissociated at 100 nM RuvA2 KaP (Fig. 5B). We conclude that branch migration was inefficient due to the weak stability of RuvA2 KaP complex II.  Holliday junction branch migration with a single RuvB hexamer was tested using the synthetic junction HJY3-hm (28). HJY3-hm is made of three short 14-bp arms and one 49-bp arm, so that only one RuvB hexamer can load onto the long arm of this junction. As with X12, the binding affinities of wild type RuvA and RuvA2 KaP to HJY3-hm were comparable (data not shown). The assays with HJY3-hm showed highly efficient RuvAB-promoted branch migration of the junction through its short arms, but RuvA2 KaP B complex was inefficient (Fig. 5C). At 25 nM RuvA, HJY3-hm dissociated completely, but in contrast, only 25% of the substrate was processed by 25 nM RuvA2 KaP (Fig. 5C). At 200 nM RuvA2 KaP, only 35% of the HJY3-hm was processed (Fig. 5C). These results show that a single RuvB hexamer assembled on DNA requires stable interactions between two RuvA tetramers for efficient processing of a four-way junction.
Binding and Processing of Synthetic Replication Forks by RuvA2 KaP -We tested the ability of RuvA2 KaP to bind and process a synthetic model replication fork (F2), in which the three branches of the forks are fully double-stranded. In EDTA buffer, RuvA bound to F2 exclusively as complex II at concentrations above 10 nM, but RuvA2 KaP only formed complex I at all concentrations tested (Fig. 6A). Significantly, RuvA2 KaP showed reduced affinity of binding to forks compared with RuvA ( Fig. 6B) even though the binding affinity of RuvA2 KaP to HJs was comparable with that of RuvA. Similar results were obtained in the presence of Mg 2ϩ above 10 nM RuvA, and 100% F2 was in complex II, but even at 250 nM, RuvA2 KaP only bound 40% of F2 as a mixture of complex I and II (data not shown). These results confirm that stable binding of RuvA to the fork requires interactions between the two tetramers. RuvA2 KaP B complex was significantly defective in processing the fork (Fig.  6C). Over 50% of the synthetic fork F2 was dissociated in reactions containing 250 nM RuvA and 300 nM RuvB, whereas less than 10% of the substrate dissociated in parallel reactions using RuvA2 KaP B (Fig. 6D). The majority of the F2 dissociation products represent processing in the direction opposite to fork reversal, as reported previously (25). These results show that the pronounced defect in the ability of RuvA2 KaP to bind to the synthetic fork (F2) resulted in a significant defect in processing.
Processing of Holliday Junctions by RuvA2 KaP in Vivo-As shown above, the mutant RuvA2 KaP was defective in forming stable complex II on model Holliday junctions and replication forks. As a consequence, the mutant was inefficient in branch migration of Holliday junctions and replication forks in vitro. It is important to correlate the observed defects of the RuvA2 KaP mutant with the currently known biological roles of RuvAB in vivo, namely processing of Holliday junctions and RuvAB-dependent reversal of stalled replication forks.
We first tested the ability of RuvA2 KaP to resolve HJ formed in vivo by homologous recombination between intact DNA molecules. The E. coli strain JJC3207 (⌬ruvA100::cat ⌬recG263::kan) (29) has a defect in homologous recombination that causes deficiency in Hfr conjugation. To test if RuvA2 KaP can process HJs in vivo, the ability of RuvA2 KaP to rescue the mutant phenotype of this strain was tested and compared with RuvA. The ruvA2 KaP coding sequence was cloned into the low copy plasmid pGB2 or in combination with the ruvB coding sequence to produce pGB-ruvA2 KaP and pGB-ruvA2 KaP B. The genes were expressed under the control of the native ruvAB promoter. The plasmids pGB2, pGB-ruvA, pGB-ruvA2KaP, or pGB-ruvA2 KaP B were transformed into the recipient JJC3207. As the pGB-ruvA2 KaP B plasmid codes for both RuvA2 KaP and RuvB, both proteins are expressed at the same levels when introduced into E. coli. This reduces any effects of altering the balance between RuvA2 KaP and RuvB, when RuvA2 KaP is overexpressed from the pGB2 plasmid, although RuvB is expressed at low levels from the chromosome. The cells were grown to log phase and then mixed with a His ϩ Hfr donor, plated on chloramphenicol minimal medium devoid of histidine, and incubated for 48 h. If successful homologous recombination had occurred, JJC3207 cells could grow on the plates lacking histidine as the His ϩ gene had been acquired from the Hfr donor via conjugation. Ratios of His ϩ versus total recipient cfu are shown (Fig. 7A). Strain JJC3207 carrying vector pGB2 showed very low conjugation levels, less than 10 Ϫ5 conjugants per cfu. Transformation of JJC3207 with pGB-ruvA resulted in much higher levels of conjugation, ϳ10 Ϫ3 conjugants per cfu. When the cells were transformed with pGBruvA2 KaP or pGBruvA2 KaP B, the conjugation level was restored to nearly 10 Ϫ3 conjugants per cfu, which is comparable with the rescued phenotype demonstrated by pGB-ruvA. Thus RuvA2 KaP is able to process HJs during conjugation as efficiently as wild type RuvA expressed from an exogenous plasmid. The complementation of the homologous recombination defect of JJC3207 by RuvA2 KaP was the same when expressed alone or in combination with RuvB on the pGB2 plasmid. This indicates that expressing extra copies of RuvB was not required for RuvA2 KaP -mediated rescue of the JJC3207 conjugational defect.
The ability of RuvA2 KaP to process HJs in vivo was additionally tested during the DNA single strand gap and double strand break repair, by assessing whether the mutant RuvA2 KaP was able to suppress the UV or mitomycin C (MMC) sensitivity of the mutant strain JJC2971 (⌬ruvA100::cat) (29). JJC2971 was transformed with pGB-ruvA, pGB-ruvAB, pGB-ruvA2 KaP , and pGB-ruvA2 KaP B, and the cells were exposed to increasing doses of UV light. JJC2971 cells transformed with pGB2 alone resulted in survival ratio of 10 Ϫ3 cells at a UV dose of 40 J/m 2 (Fig. 7B). In contrast, the survival of JJC2971 transformed with pGB-ruvA was 100% at 40 J/m 2 UV dose, indicating that the cells were able to repair DNA lesions. JJC2971 cells transformed with either pGB-ruvA2 KaP or pGB-ruvA2 KaP B gave survival profiles comparable with pGB-ruvA transformation at all doses tested, indicating that RuvA2 KaP is fully capable of resolving HJ made during the recombinational repair of UV lesions. The co-expression of RuvA2 KaP with RuvB in JJC2971 cells very slightly improved survival, which suggests that there may be a minor defect in the RuvA2 KaP interaction with RuvB. In an additional assay, the JJC2971 cells were treated with 2 g⅐ml Ϫ1 MMC for 90 min to induce double strand breaks in the DNA. These cells were then plated on LB spectinomycin overnight, and the ratios of colony forming units in treated and untreated cultures were calculated (Fig. 7C). Introduction of pGB2 only resulted in a survival ratio of 10 Ϫ5 cells. Introduction of pGB-ruvA, pGB-ruvA2 KaP , or pGB-ruvA2 KaP B all resulted in a survival ratio of 10 Ϫ2 cells indicating that the RuvA mutants were able to restore the cells resistance to MMC to the levels observed after introducing pGB-ruvA. These data suggest that RuvA2 KaP is able to process Holliday junctions formed during recombination-mediated repair of DNA lesions as efficiently as wild type RuvA.
Stability of RuvA2 KaP on Holliday Junctions in Vivo-The defect of RuvA2 KaP in maintaining stable binding to both sides of junction DNA as a double tetramer was demonstrated in A, RuvA2 KaP was tested for rescue of conjugation ability of the E. coli strain JJC3207 (ruvA100 recG). RuvA2 KaP was expressed from a pGB2 plasmid alone (pGB-ruvA2 KaP ) or in combination with RuvB (pGB-ruvA2 KaP B). The log 10 survival of conjugates/cfu was plotted as a bar graph. B, ability of RuvA2 KaP to rescue UV sensitivity in a JJC2971 (ruvA100::cat) E. coli strain was tested. JJC291 was transformed with pGB2, pGB-ruvA, pGB-ruvA2 KaP , or pGB-ruvA2 KaP B. The ratio of colony-forming units of treated versus untreated cells was calculated to derive log 10 survival ratio. C, JJC2971 cells were transformed with pGB2, pGB-ruvA, pGB-ruvA2 KaP , and pGB-ruvA2 KaP B and incubated with 2 g⅐ml Ϫ1 of mitomycin C for 90 min. The ratio of colonyforming units of treated versus untreated cells was calculated to derive log 10 survival ratio. D, RusA cleavage in vivo. JJC2761 (⌬ruvABC rus-1) E. coli cells were transformed with pGB2, pGB-ruvA, pGB-ruvA2 KaP , or pGB-ruvA2 KaP B and exposed to UV radiation. The experiments were quantified, and the data were used to generate a graph of the log 10 survival ratio of the transformants as a function of the dose of UV (J/m 2 ). E, processing of synthetic forks by RuvA2 KaP in vivo. An E. coli strain, JJC3723 (dnaEts recBCts ruvA100), was transformed with pGB2, pGB-ruvA, pGB-ruvA2 KaP , or pGB-ruvA2 KaP B. The cells were grown at 30°C and then shifted to 42°C for 3 h, after which fork reversal was assessed by the amount of double strand breaks generated, which was measured by the amount of linear chromosomal DNA entering a pulse field gel. The experiments were quantified and used to generate a histogram of the percentage of linear DNA for each strain.
vitro by its inability to inhibit cleavage of the junction by RuvC. We decided to examine this finding in vivo. In E. coli, the rusA gene encodes a Holliday junction resolvase carried on a cryptic prophage, but this gene in not normally expressed (8,29). The E. coli strain JJC2761 (⌬ruvABC rus-1) does not encode RuvABC, but the rus-1 mutation results in rusA expression, and the cells survive UV damage as RusA resolves Holliday junctions formed by recombination-mediated repair (24,29). We used the JJC2761 strain as a tool to test the effect of RuvA2 KaP on Holliday junction resolution by RusA in vivo. If a ruvA-carrying plasmid is added to the strain, RuvA binding occludes the Holliday junction and prevents the action of RusA. Holliday junctions cannot be resolved, resulting in UV sensitivity and cell lethality. JJC2761 cells were transformed with pGB2, pGB-ruvA, pGBruvA2 KaP , or pGBruvA2 KaP B, and the survival of plasmid-harboring cells subjected to different doses of UV-irradiation is shown in Fig. 7D. A similar level of protection against RusA-catalyzed HJ resolution, resulting in UV sensibility, was observed for pGB-ruvA, pGBruvA2 KaP , or pGBruvA2 KaP B at 40 J/m 2 . At higher doses, the survival ratio of JJC2761 cells with pGBruvA2 KaP or pGBruvA2 KaP B was intermediate between those observed with the vector pGB2 and with the control plasmid pGB-ruvA, ϳ15-20-fold more cells survived compared with cells carrying pGB-ruvA at 100 J/m 2 . These data demonstrate that RuvA2 KaP alone or with RuvB could protect the Holliday junction from cleavage by RusA, but not as efficiently as wild type RuvA. The RuvAz3 and RuvAz87 mutants, which were deficient at inhibiting RuvC-mediated cleavage of HJs (Fig. 2B), were also defective in preventing RusA cleaving Holliday junctions in the JJC2761 strain (29).
RuvA2 KaP Is Deficient for Replication Fork Reversal in Vivo-Finally, the ability of the mutant RuvA2 KaP to reverse stalled replication forks in conjunction with RuvB was tested in vivo using the strain JJC3723 (dnaEts recB ruvA100). In cells carrying the temperature-sensitive replication mutant dnaEts, RuvA is required for RFR (24). In the strain JJC3723, RFR can be measured by re-introducing RuvA on a plasmid and measuring the linearized DNA formed in this recB mutant by RuvC cleavage of reversed forks (29).
JJC3723 cells transformed with pGB2, pGB-ruvA, pGB-ruvA2 KaP , and pGB-ruvA2 KaP B constructs were grown at 30°C and then shifted to 42°C for 3 h. The amount of linearized DNA in the cells was analyzed by pulse field gel electrophoresis and quantified (Fig. 7E). Cells carrying the wild type RuvA plasmid, pGB-ruvA, resulted in 55% linearization of DNA compared with 11% DNA linearization in control cells with empty vector pGB2. Cells transformed with pGB-ruvA2 KaP resulted in about 11% DNA linearization, similar to the control cells transformed with pGB2 alone, and thus RuvA2 KaP could not promote replication fork reversal. When the JJC3723 cells were transformed with RuvA2 KaP and RuvB (pGB-ruvA2 KaP B), the amount of double strand breaks increased from 11 to ϳ28%, but it was still significantly lower than the levels of 55% observed with pGB-ruvA and the ϳ70% breakage conferred by pGB-ruvAB in a similar strain (33). These data indicate that RuvA2 KaP , which binds forks but is deficient in complex II formation in vitro, is unable to reverse forks in vivo. Thus, a single RuvA tetramer that cannot interact with a second RuvA tetramer on DNA is not sufficient to reverse forks in concert with RuvB. Because RuvA2 KaP is proficient in processing Holliday junctions in vivo during homologous recombination but is unable to reverse stalled replication forks, RuvA2 KaP is a separation-of-function mutant similar to RuvAz3 and RuvAz87.

DISCUSSION
The aim of this study is to understand the molecular mechanisms of RuvAB acting on two different branched substrates in vivo, Holliday junctions and stalled replication forks. To this end, we used a defined RuvA mutant, RuvA2 KaP , designed as tetramer-only by replacing two amino acid residues (E127R and K119A) engaged in ionic pairs that stabilize interactions between two RuvA tetramers on the Holliday junction (supplemental Fig. 1). Several lines of evidence argue that the two substitutions in RuvA2 KaP do not affect the quaternary structure of the protein. The size exclusion data indicated that RuvA2 KaP forms the correct complexes in the correct amounts in solution (supplemental Fig. 2A). The protein is stable compared with wild type RuvA when analyzed on SDS-PAGE (no degradation products were detected). Furthermore, like RuvA, RuvA2 KaP remains a tetramer when run in SDS, indicating that interactions within the tetramer are stable (supplemental Fig. 2A). Complex formation with RuvB analyzed by SEC (supplemental Fig. 2C) indicates no defect in RuvA2 KaP -RuvB protein interaction, and finally RuvA2 KaP HJ binding affinity is as wild type (Fig. 1B). Therefore, we consider it unlikely that the two substitutions in RuvA2 KaP affect the protein structure beyond destabilizing the double tetramer. As expected, RuvA2 KaP was found to bind predominantly to HJs as complex I.
Tetramer Only Mutants Are Impaired in Vitro for the Formation of an RuvA⅐RuvB⅐HJ Tripartite Complex for the Stimulation of RuvB ATPase and Branch Migration Activities; Nevertheless, They Promote Homologous Recombination in Vivo-We observed that in Mg 2ϩ buffer the affinity of RuvA2 KaP for DNA remains similar to that of the wild type protein, but binding of the second tetramer is affected. This result suggests that Mg 2ϩ has a greater effect on tetramer-tetramer interactions compared with modulating the affinity of RuvA for DNA. Interestingly, this result also reveals that complex II devoid of tetramertetramer interactions might form if the affinity for DNA of each of the two tetramers is high enough, albeit less stably than wild type. It supports the idea of RuvA co-existing on a HJ with RuvC (34). In fact, proteins are able to bind the opposite faces of a HJ without protein-protein interactions, as simultaneous binding to HJ was observed using two unrelated proteins, RuvA and the yeast mitochondrial protein YDC2 (35). The formation of unstable complex II devoid of tetramer-tetramer interactions correlates with the reduced ability of RuvA2 KaP to stimulate RuvB ATPase activity in vitro and a defect in tripartite complex formation. Even at concentrations that allowed RuvA2 KaP complex II formation on HJs in Mg 2ϩ buffer, RuvA2 KaP could not stimulate RuvB ATPase activity to wild type levels. The loss of tetramer-tetramer interaction could lead to a deficiency in functional interactions with RuvB on DNA. SEC analysis confirmed that physical interactions between RuvA2 KaP and RuvB in the absence of DNA were not compromised, suggesting that RuvA2 KaP -RuvB interactions were not responsible for the stabilizing this complex on a three-way fork junction is a double tetramer of RuvA with wild type tetramer-tetramer interactions, whereas branch migration of a HJ can be carried out by an unstable double tetramer with two RuvB hexamers providing additional stabilization through their contacts with RuvA.
Separation-of-function phenotype might result from various RuvAB defects, because different separation-of-function mutants were isolated in the ruvA as in the ruvB gene (29,33). Studies of ruvB mutants showed that mutations that presumably affect the ATPase activity of RuvB also confer a separation-of-function phenotype (33). This study identifies the tight binding of two RuvA tetramers via tetramer-tetramer interactions as a property crucial for replication fork reversal but not for homologous recombination.