The role of RuvA octamerization for RuvAB function in vitro and in vivo.

RuvA plays an essential role in branch migration of the Holliday junction by RuvAB as part of the RuvABC pathway for processing Holliday junctions in Escherichia coli. Two types of RuvA-Holliday junction complexes have been characterized: 1) complex I containing a single RuvA tetramer and 2) complex II in which the junction is sandwiched between two RuvA tetramers. The functional differences between the two forms are still not clear. To investigate the role of RuvA octamerization, we introduced three amino acid substitutions designed to disrupt the E. coli RuvA tetramer-tetramer interface as identified by structural studies. The mutant RuvA was tetrameric and interacted with both RuvB and junction DNA but, as predicted, formed complex I only at protein concentrations up to 500 nm. We present biochemical and surface plasmon resonance evidence for functional and physical interactions of the mutant RuvA with RuvB and RuvC on synthetic junctions. The mutant RuvA with RuvB showed DNA helicase activity and could support branch migration of synthetic four-way and three-way junctions. However, junction binding and the efficiency of branch migration of four-way junctions were affected. The activity of the RuvA mutant was consistent with a RuvAB complex driven by one RuvB hexamer only and lead us to propose that one RuvA tetramer can only support the activity of one RuvB hexamer. Significantly, the mutant failed to complement the UV sensitivity of E. coli DeltaruvA cells. These results indicate strongly that RuvA octamerization is essential for the full biological activity of RuvABC.

RuvA plays an essential role in branch migration of the Holliday junction by RuvAB as part of the RuvABC pathway for processing Holliday junctions in Escherichia coli. Two types of RuvA-Holliday junction complexes have been characterized: 1) complex I containing a single RuvA tetramer and 2) complex II in which the junction is sandwiched between two RuvA tetramers. The functional differences between the two forms are still not clear. To investigate the role of RuvA octamerization, we introduced three amino acid substitutions designed to disrupt the E. coli RuvA tetramer-tetramer interface as identified by structural studies. The mutant RuvA was tetrameric and interacted with both RuvB and junction DNA but, as predicted, formed complex I only at protein concentrations up to 500 nM. We present biochemical and surface plasmon resonance evidence for functional and physical interactions of the mutant RuvA with RuvB and RuvC on synthetic junctions. The mutant RuvA with RuvB showed DNA helicase activity and could support branch migration of synthetic four-way and three-way junctions. However, junction binding and the efficiency of branch migration of four-way junctions were affected. The activity of the RuvA mutant was consistent with a RuvAB complex driven by one RuvB hexamer only and lead us to propose that one RuvA tetramer can only support the activity of one RuvB hexamer. Significantly, the mutant failed to complement the UV sensitivity of E. coli ⌬ruvA cells. These results indicate strongly that RuvA octamerization is essential for the full biological activity of RuvABC.
Homologous recombination is a fundamental and ubiquitous cellular process necessary for maintaining the stability of the genome and for generating genetic diversity. It provides a powerful mechanism for the repair of DNA doublestranded breaks and plays a crucial role in the repair and restart of stalled or collapsed replication forks (reviewed in Ref. 1). Holliday junctions are key intermediates in homologous recombination that need to be enzymatically resolved for the cells to survive. The main pathway for processing Holliday junctions in prokaryotes involves the well charac-terized RuvABC system that promotes branch migration and resolution of the junction (2,3). Based on extensive genetic, biochemical, and structural data, elegant models have been proposed for the concerted action of RuvABC on Holliday junctions (4 -7).
The RuvAB complex promotes ATP-dependent branch migration of Holliday junctions (8,9). Resolution of the Holliday junction is catalyzed by the RuvC endonuclease (10,11), which acts as a dimer and introduces coordinated cuts at two symmetrical sites across the junction. Genetic analyses provide strong evidence that junction resolution by RuvC is RuvAB-dependent (12)(13)(14)(15), and a series of biochemical experiments suggest the formation of a RuvABC resolvasome working in a well coordinated manner (16 -18).
There is a wealth of structural information regarding the components of the RuvABC system (reviewed in Ref. 19). RuvA forms a tight tetramer that binds specifically to the Holliday junction (20) and maintains its global structure in square planar conformation, which facilitates branch migration (5). The central part of the tetramer is formed largely by RuvA domains I and II and provides a platform for DNA binding and disruption of base pairing at the cross-over point (21)(22)(23), whereas domain III contacts RuvB and is involved in branch migration (24,25).
Two RuvA-junction complexes have been observed by EMSA 1 : 1) complex I, which contains one RuvA tetramer bound to the junction, and 2) complex II in which the junction is sandwiched between two tetramers (26,27). The crystal structures of RuvA complexes with synthetic junctions corresponding to each of these forms have been obtained (21)(22)(23). The existence of the octameric Mycobacterium leprae RuvA-synthetic junction complex in solution was also demonstrated by neutron-scattering experiments (28). The structures of complex I with RuvA from Escherichia coli showed that the junction did not have a strictly square planar conformation as the crossover point deviated closer to the concave surface of the RuvA tetramer (21,23). In contrast, the junction in complex II obtained with RuvA from M. leprae had a planar conformation (22). However, in both cases, two base pairs located symmetrically at the cross-over point were disrupted. Analytical ultracentrifugation analysis provided evidence for a dynamic equilibrium of M. leprae RuvA tetramers and octamers in solution (29) and that a similar equilibrium between complexes I and II is most probable. Interestingly, in the atomic structure of the Thermus thermophilus RuvAB complex, RuvA was found to be a double tetramer that could be superimposed on the M. leprae RuvA structure (30). The residues involved in the interface between the two tetramers are strongly conserved in the RuvA sequences of many bacteria, including Escherichia coli (22). The ability to form the octameric structure appears to be conserved among RuvA from distant species, which highlights its importance for RuvA function. Despite the structural insights into the two forms of RuvA-junction complexes, their functional role in the RuvABC complex remains unclear.
RuvB is an ATPase motor protein belonging to the AAA ϩ protein family and consists of three domains (N, M, and C). Two of these (N and M) are known to be well conserved in the AAA ϩ family and are responsible ATP binding and divalent metal-mediated ATP hydrolysis (31,32). In addition, domain N contains a hydrophobic ␤-hairpin proposed to be the interface with domain III of RuvA (33). The third domain of RuvB is thought to contain a DNA binding motif that can play a role in DNA translocation during branch migration (31,32). RuvB undergoes oligomerization in the presence of cofactors such as ATP, Mg 2ϩ , and DNA (34 -36). Electron microscopic observations and biochemical studies indicate strongly that the functional form of RuvB is a hexameric ring assembled on DNA (37).
According to the current models for the RuvAB complex, branch migration is driven by two hexameric RuvB rings assembled on opposite arms of the junction bound by RuvA (5). The model could accommodate one or two tetramers of RuvA with scanning-transmission electron microscopy measurements supporting the double tetramer (38). On the other hand, a RuvABC resolvasome is envisaged to contain a RuvA/RuvC sandwich on the junction flanked by two RuvB rings and to promote coupled branch migration and resolution (4). It has been argued that the hexameric RuvB motors would need a "stator" as they rotate the DNA arms in opposite directions. Therefore, the "twin-engine" RuvAB complex would need a double tetramer RuvA on the junction for stability (22). A double tetramer of RuvA flanked by two RuvB rings, however, would block the access of RuvC to the junction, and it is difficult to envisage how such a tripartite RuvAB complex could be converted into a RuvABC resolvasome. The symmetry of the proposed protein complexes matches the inherent symmetry of the Holliday junction itself, adding undisputed beauty to these models. However, experiments with three-way junctions have demonstrated that a single RuvB hexamer can support "branch migration" with RuvA (39). It is not known whether this reaction is supported by one or two tetramers of RuvA. Both complex I and complex II can form on Y-junctions (40) and replication fork substrates (41). These experiments suggest some functional flexibility of the RuvABC system that may be involved in processing of asymmetric recombination intermediates, such as three-stranded intermediates (42,43). The oligomeric form of RuvA that acts with a single RuvB hexamer (one or two tetramers) has not been assessed.
Here we present an investigation into the role of tetramertetramer interactions for the activity of E. coli RuvAB using a mutant RuvA protein that was unable to form complex II. We provide evidence that complex I exhibits proficient DNA helicase activity and supports branch migration of three-and fourway junctions in vitro. However, the efficiency of branch migration of four-way junctions was affected and the mutant RuvA could not complement the UV-sensitive phenotype of ⌬ruvA E. coli.
Site-directed Mutagenesis-Defined amino acid substitutions in the E. coli RuvA protein at E122R, E127R, and D130R were made by introducing the appropriate codon changes into the coding sequence of ruvA in pET21ruvA using the QuikChange TM kit from Stratagene. The respective codons GAA, GAA, and GAC were changed to CGC in all three cases. The presence of the correct mutations in the resultant pET21ruvA3m was verified by sequencing.
Proteins-Wild-type RuvA protein was overexpressed from the pET21ruvA construct in E. coli BL21-Gold (DE3). The culture was grown at 37°C in LB medium supplemented with ampicillin (100 g/ ml) and 1% glucose to A 600 nm ϭ 0.8 and induced with 1 mM IPTG for 4 h at 37°C. Bacteria were collected by centrifugation, washed in lysis buffer (100 mM Tris-HCl, pH 8.0, 2 mM EDTA, 5% glycerol), resuspended in the same buffer, and frozen. The thawed cell suspension was supplemented with 100 mM NaCl, 5 mM ␤-mercaptoethanol, and 0.1 mM PMSF and lysed using three freeze-thaw cycles with 1 mg/ml lysozyme added after the first cycle. The resulting suspension of lysed cells was cleared by centrifugation in a Beckman 45 Ti rotor at 42,000 rpm for 1 h at 4°C. DNA was removed by stirring the crude lysate with 0.1% polyethylenimine for 1 h on ice followed by centrifugation at 15,000 rpm for 30 min at 4°C in a Sorvall SS-34 rotor. Ammonium sulfate was added to the supernatant to 40% saturation. The precipitated proteins and DNA were pelleted at 15,000 rpm for 30 min at 4°C in a Sorvall SS-34 rotor. RuvA remained in solution at this ammonium sulfate saturation. Solid ammonium sulfate was added to the supernatant to 70% saturation. The precipitated proteins including RuvA were pelleted as above. The pellet was dissolved in TEG buffer (20 mM Tris-HCl, pH 8.5, 1 mM EDTA, 10% glycerol) containing 5 mM ␤-mercaptoethanol, 0.1 mM PMSF, and 0.1 M KCl and dialyzed against the same buffer. The dialyzed protein solution was loaded onto a 5-ml HiTrap Q HP column (Amersham Biosciences), and the column was eluted with a linear gradient of 100 -1000 mM KCl in the same buffer. Fractions containing RuvA were diluted in P200 buffer (10 mM potassium phosphate, pH 6.8, 200 mM KCl, 10% glycerol, 5 mM ␤-mercaptoethanol, 0.1 mM PMSF) and dialyzed against the same buffer. The dialyzed fractions were applied to a Hydroxyapatite Bio-Gel HTP (Bio-Rad) column preequilibrated in P200 buffer. The proteins were eluted with a 10 -600 mM linear gradient of potassium phosphate. Fractions containing RuvA were pooled and dialyzed against P100 buffer (10 mM potassium phosphate, pH 6.8, 100 mM KCl, 10% glycerol, 5 mM ␤-mercaptoethanol, 0.1 mM PMSF). The dialyzed protein solution was loaded onto HiTrap Heparin HP column (Amersham Biosciences), and the column was developed with a linear gradient from 100 to 1000 mM KCl in P100 buffer.
The mutant RuvA3m protein was expressed and fractionated by the procedures described above up to the first chromatography step. The ammonium sulfate pellet was dissolved in buffer containing 20 mM Tris-HCl, pH 7.0, 1 mM EDTA, 10% glycerol, 5 mM ␤-mercaptoethanol, 0.1 mM PMSF, and 300 mM NaCl, and dialyzed against the same buffer. The dialyzed protein solution was loaded onto HiTrap SP HP column (Amersham Biosciences), and the column was developed with a linear gradient from 0 to 3 M NaCl in the same buffer. Active fractions were pooled and dialyzed against buffer containing 20 mM Tris-HCl, pH 8.5, 1 mM EDTA, 10% glycerol, 5 mM ␤-mercaptoethanol, 0.1 mM PMSF, and 250 mM KCl. The dialyzed material was applied to HiTrap Q HP column, and the protein was eluted with a linear gradient from 100 to 1000 mM KCl. At this step, the RuvA3m protein was 99.9% pure as judged by SDS-PAGE stained with Coomassie Brilliant Blue. The RuvB and RuvC protein were purified as described previously (45,46).
The concentration of RuvB was determined by absorbance at 280-nm wavelength using an extinction coefficient of ⑀ 280,native ϭ 16,900 M Ϫ1 cm Ϫ1 (35). All of the other protein concentrations were determined by the Bradford method (47) using the protein assay reagent from Bio-Rad with BSA as a standard.
Size-exclusion chromatography on Superdex 200 fast protein liquid chromatography column (Amersham Biosciences) was carried out in TEG buffer, pH 7.5, supplemented with 1 M KCl. The column was calibrated using gel-filtration standards from Bio-Rad. 100 g of wildtype RuvA or 200 g of RuvA3m were applied to the column. RuvA proteins were mixed with RuvB at a 4:6 molar ratio, and a total of 155 g of wild-type RuvAB or 278 g of RuvA3m⅐RuvB were applied to the column.
DNA Substrates-The synthetic 62-mer four-way junctions X0 and X12, which contained a 12-bp core of homology three-way Y-junction and a 62-bp linear duplex DNA L0, were described previously (48). Synthetic 40-mer four-way junction Jmb5 was as described previously (49). When appropriate, oligonucleotides were labeled at the 5Ј termini using [␥-32 P]ATP (3000 Ci/mmol, PerkinElmer Life Sciences) and T4 polynucleotide kinase (New England Biolabs). Assembly and gel purification of the synthetic substrates were described previously (48). DNA helicase substrate was prepared by annealing a 32 P-labeled 52-mer oligonucleotide to X174 virion DNA (50).
Electrophoretic Mobility Shift Assay-Binding reactions (20 l final volume) were performed in 50 mM Tris-HCl, pH 8.0, 5 mM EDTA, 1 mM dithiothreitol, 100 g/ml BSA, and 5% glycerol. The reactions contained 1 nM 32 P-labeled substrate and varying concentrations of protein (expressed as monomers). Proteins were diluted in 20 mM Tris-HCl, pH 7.5, 1 mM EDTA, 10% glycerol, 5 mM ␤-mercaptoethanol, 0.1 mM PMSF, 500 mM NaCl, and 100 g/ml BSA. Incubation took place for 15 min on ice. Samples were mixed with 6ϫ DNA loading buffer, and 12 l were loaded directly on neutral 4% polyacrylamide gels in "low" TAE buffer (6.7 mM Tris-HCl, pH 8.0, 3.3 mM sodium acetate, 2 mM EDTA). Electrophoresis was performed for 2 h at 8.5 V/cm at 4°C. Non-labeled 62-bp linear duplex DNA used as a competitor was pre-mixed with X12 junction at a 22-fold molar excess in reactions with wild-type RuvA and at a 4-or 40-fold molar excess in reactions containing 500 or 2500 nM RuvA3m, respectively. The gels were dried, and the bands were visualized using a Fuji FLA-2000 PhosphorImager.
Branch Migration Assay-Branch migration reactions (20 l final volume) were performed in 50 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , 50 mM KCl, 5 mM ␤-mercaptoethanol, 100 g/ml BSA, and 2 mM ATP. The reactions contained ϳ1 nM 32 P-labeled synthetic junction and varying concentrations of proteins (expressed as monomers). Proteins were diluted in 20 mM Tris-HCl, pH 7.5, 1 mM EDTA, 10% glycerol, 5 mM ␤-mercaptoethanol, 0.1 mM PMSF, 500 mM NaCl, and 100 g/ml BSA. Incubation was for 30 min at 37°C. Reactions were stopped and deproteinized by the addition of 5 l of stop mixture (10 mg/ml proteinase K and 2.5% SDS) followed by incubation for 10 min. Samples were mixed with 5 l of 6ϫ DNA loading buffer, and 15 l were loaded onto neutral 6% polyacrylamide gels in 1ϫ Tris borate-EDTA buffer. Electrophoresis was performed for 1 h at 10 V/cm at 25°C in 1ϫ Tris borate-EDTA buffer. Reaction products were analyzed as described above.
Assays for Resolution with RuvC in the Presence of RuvA-Reactions (20 l final volume) contained 50 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , 50 mM KCl, 5 mM ␤-mercaptoethanol, 100 g/ml BSA, 1 nM 32 P-labeled junction X12, and varying concentrations of proteins (expressed as monomers). Incubation was for 60 min at 37°C. Reactions were stopped and deproteinized by the addition of 5 l of stop mixture (10 mg/ml proteinase K and 2.5% SDS) followed by incubation for 10 min, and the labeled DNA products were analyzed as described above.
DNA Helicase Assay-Reaction mixtures (20 l) contained 5 nM substrate DNA in 50 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , 50 mM KCl, 5 mM ␤-mercaptoethanol, 100 g/ml BSA, 2 mM ATP, and varying concentrations of proteins (expressed as monomers) as indicated. Reactions were stopped and deproteinized by the addition of 5 l of stop mixture (10 mg/ml proteinase K and 2.5% SDS) followed by incubation for 10 min. Reaction products were analyzed by electrophoresis in 1% agarose gel in 1ϫ TAE buffer. Gels were dried on Whatman DE81 paper, and the labeled DNA products were analyzed as described above.
SPR Experiments-The interactions between Holliday junction and WT RuvA, RuvA3m, RuvB, and the RuvC were analyzed using BIAcore 1000 instrument (Biacore AB, Uppsala, Sweden). Holliday junction Jbm5 containing one biotinylated strand was diluted in HBS-EP buffer (10 mM HEPES-KOH, pH 7.4, 150 mM NaCl, 3.4 mM EDTA, 0.005% (v/v) polysorbate 20) supplemented by NaCl up to 500 mM final concentration as specified. The junction (0.1 ng/l) was injected at a flow rate of 5 l/min onto a streptavidin-coated sensor chip (SA sensor chip, BIAcore). The amount of immobilized DNA was found to be ϳ400 RU. Different concentrations of WT RuvA, RuvA3m, RuvB, and the RuvC proteins were then injected at a flow rate of 10 l/min in HBS-EP buffer, and at the end of the association time, dissociation occurred by adding HBS-EP buffer.
UV Sensitivity Test-Cultures of E. coli cells carrying plasmid vector pET21b(ϩ) or pET21b(ϩ) constructs with wild-type RuvA or RuvA3m were grown at 37°C for 2 h in LB containing 100 g/ml ampicillin. At A 600 ϭ 0.5, the cells were induced with 0.1 mM IPTG for an additional hour. Appropriate dilutions were made in sterile water, and 100 l of the diluted cultures were plated onto Petri dishes containing 25 ml of LB-agar medium supplemented with 100 g/ml ampicillin. The plates were exposed to UV radiation using a herbicidal ultraviolet lamp (BDH, 254 nm, 2 ϫ 8W) from a distance of 35 cm. Colonies were counted following overnight incubation at 37°C. The values given represent five separate experiments. For Western blot analysis, an equal number of non-induced and induced cells were lysed and denatured in SDS sample buffer, fractionated by SDS-PAGE, and transferred to nitrocellulose membranes. The blots were immunostained using rabbit polyclonal antibodies raised against M. leprae RuvA (40) followed by alkaline phosphatase-conjugated secondary antibody and color detection.

Design of Mutations in the E. coli RuvA Tetramer-Tetramer
Interface-The crystal structure of M. leprae RuvA complexed with a synthetic junction showed two RuvA tetramers that formed an octameric shell encasing the junction (22). The two tetramers made direct protein-protein contacts at four points that involved side chain interactions of residues 117-129 of two ␣-helices lying in almost antiparallel orientation that allowed them to interact over several helical turns. Six ion interactions were formed among residues Arg 118 -Asp 129 , Glu 121 -Arg 128 , and Arg 122 -Glu 126 , whereas the interaction between Leu 125 and Leu 125 was hydrophobic (22). To disrupt these interactions, three negatively charged residues predicted to form salt bridges in E. coli RuvA were replaced by Arg (E122R, E127R, and D130R). It was expected that the loss of ionic interactions plus repulsion by positive charge would prevent the formation of RuvA octamers in the triple mutant RuvA (RuvA3m).
The mutant RuvA3m was readily purified following the modified wild-type RuvA purification procedure described under "Experimental Procedures." RuvA3m eluted from the SP-Sepharose ion-exchange column at a higher salt concentration than WT RuvA, which is consistent with the expected overall increase in positive charge at pH 8. Size-exclusion chromatography on Superdex 200 showed that RuvA3m eluted in a single peak at the same elution position as WT RuvA, corresponding to a molecular mass of 88 kDa or a tetramer of RuvA (Fig. 1A). The mutant RuvA3m was binding to RuvB in solution and formed a RuvA3m⅐RuvB complex normally as observed by sizeexclusion chromatography (Fig. 1B).
DNA Binding Activity of RuvA3m-The binding of the RuvA3m mutant to synthetic Holliday junctions was investigated using a standard EMSA. The mutant RuvA3m protein retained the ability to bind the synthetic junction and readily formed complex I consisting of a single RuvA tetramer ( Fig.  2A). However, no band corresponding to complex II could be detected at a RuvA3m concentration of up to 500 nM, whereas all of the junction DNA was found in complex II using WT RuvA at this protein concentration. Instead, RuvA3m formed a novel aberrant band with electrophoretic mobility that was intermediate between complex II and complex I. At very high protein concentrations, RuvA3m formed a smeared ladder of slowly moving bands but none of them corresponded to complex II. The binding pattern observed in the presence of Mg 2ϩ was very similar (data not shown). The molecular architecture of the aberrant complexes observed at high protein concentrations cannot be deduced from the EMSA experiments. However, similar ladders of complexes on synthetic junctions were shown in previous studies of RuvA "pin" mutants that bound to the duplex arms of junction as well as to the cross-over point (51) and of the archaeal Holliday junction resolvase Hjc (52). It seemed possible that the mutant RuvA3m was binding to the duplex arms of the junction due to the high concentration of positively charged residues in the RuvA3m tetramer interface. To test whether the aberrant complexes could constitute nonspecific binding of RuvA3m to the arms of the junction, binding reactions were carried out in the presence of a 62-bp duplex competitor DNA. As shown in Fig. 2B, the aberrant band above complex I at 500 nM RuvA3m was efficiently competed by a 4-fold excess of non-labeled linear dsDNA competitor, whereas complex I remained unchanged (lane e). At a high concentration of RuvA3m, all of the junction DNA was shifted into aberrant complexes in the absence of competitor DNA (Fig. 2B,  lane f). However, both the smear of aberrant multimeric complexes and the aberrant band were efficiently competed by a 40-fold excess of competitor dsDNA, leaving all of the junction in complex I (Fig. 2B, lane g). The ability of duplex DNA to compete out the aberrant band in particular shows that, although this complex most probably contains two RuvA3m tetramers, they do not form a stable sandwich on the junction but rather involve the binding of one RuvA3m tetramer to junction arms. In contrast, linear dsDNA could not compete the band corresponding to complex I (Fig. 2B, lanes e and g). Neither complex I nor complex II could be competed by dsDNA in control reactions with WT RuvA as expected (Fig. 2B, lane c). EMSA experiments were also conducted using labeled linear duplex DNA directly. No binding of WT RuvA to duplex DNA was detected, but the mutant RuvA showed some binding to duplex DNA at protein concentrations of 500 nM and higher (data not shown). The binding experiments demonstrated that the RuvA3m mutant was able to bind specifically to the synthetic junction but, as predicted, the mutant was unable to sandwich the junction at the cross-over point, so complex I was the predominant species at moderate protein concentrations.
The overall efficiency of RuvA3m binding to the junction was lower compared with WT RuvA. The differences were less pronounced at higher protein concentrations (Fig. 3). The inability of mutant RuvA3m to sandwich the junction through tetramertetramer interactions is likely to affect the overall efficiency of binding measured by EMSA due to an inherently lower stability of complex I during gel electrophoresis. As shown previously, complex I had lower stability in the presence of divalent cations compared with complex II and the binding pattern of M. leprae RuvA to Holliday junctions suggested that binding might be cooperative (40).
Biochemical Activities of RuvA3m with RuvB-The biochemical activity of RuvA3m with RuvB was first tested in a DNA helicase assays that followed the displacement of a 52-mer oligonucleotide annealed to X174 virion single-stranded DNA (Fig. 4). The mutant RuvA3m⅐RuvB complex showed DNA helicase activity as efficient as the wild-type RuvAB complex at all of the protein concentrations. At high protein concentration, the reaction showed a slight inhibition with both proteins, as observed previously for WT RuvA (53).
The ability of the mutant RuvA3m to promote branch migration with RuvB was tested using synthetic Holliday junctions X12 and X0. The RuvA3m mutant was able to support branch migration with RuvB using both substrates, as shown in Fig. 5. However, branch migration with RuvA3m was less efficient compared with WT RuvA (Fig. 5). Higher concentrations of RuvA3m were required compared with WT RuvA, and the reaction was inhibited at high protein concentrations (above 500 nM RuvA3m). Inhibition of branch migration by high concentrations of WT RuvA has not been observed (26,54). A comparison with the binding experiments in Fig. 2 suggests that branch migration occurred at RuvA3m concentrations that showed the formation of complex I but was inhibited at concentrations that produced aberrant complexes.
The difference between WT RuvA and RuvA3m was more pronounced when junction X0 was used as a substrate (Fig.  5B). The time course of the reaction with junction X0 showed that both the rate and the overall efficiency of branch migration with RuvA3m were significantly reduced compared with WT RuvA (Fig. 6A). Whereas Ͼ60% junction was dissociated in 15 min using 50 nM WT RuvA, Ͻ40% junction branch-migrated with RuvA3m at this concentration. The differences were even greater at the earlier time points. These results suggest that the mutant RuvA3m⅐RuvB complex might have a problem in promoting branch migration through heterologous arms. Whereas both junctions have duplex arms of the same length (26 bp), the region of heterology is shorter in junction X12, which has a 12-bp homologous core.
The lower efficiency of branch migration observed with RuvA3m could be the result of impaired processivity due to less stable binding of the mutant to the junction. It could also reflect structural changes that affect the intrinsic properties of the branch migration complex. A complex that contains a single RuvA3m tetramer flanked by two RuvB hexameric rings is likely to be unstable and could be stalling and falling apart at regions of heterology. Alternatively, it could be that the RuvA3m mutant was unable to assemble a tripartite branch migration complex on the junction at all and, similar to reactions with Y-junctions, branch migration under these conditions was driven by a single RuvB hexamer. To test this latter assumption directly, we examined the branch migration activity of RuvA3m and RuvB with Y-junctions. Both the RuvA3m mutant and wild-type RuvA were able to displace one strand of the Y-junction with similar efficiency as shown by the time course of the reactions (Fig. 6B). The mutations introduced in RuvA3m therefore had little effect in reactions with Y-junctions, whereas they seriously affected branch migration of four-way junctions (Fig. 6A). These results are consistent with the idea that branch migration driven by a single RuvB hexametric ring only needs a tetramer of RuvA and would not be affected by the mutations in RuvA3m, whereas a more efficient tripartite branch migration complex, containing two hexametric RuvB rings, would require stable RuvA tetramer-tetramer interactions. Branch migration of short synthetic Holliday junctions observed in our experiments with RuvA3m could well be driven by a single RuvB hexamer, which could explain the lower efficiency of these reactions.
Interaction of RuvA3m with RuvC on Holliday Junction-It has been shown that a sandwich complex of RuvA and RuvC can form at the junction cross-over (27,40,55). This complex is thought to be an intermediate in the formation of a RuvABC resolvasome that can carry coupled branch migration and resolution of the Holliday junction (4,16). However, junction resolution by RuvC in vitro is inhibited by increasing concentrations of RuvA (27), which can be attributed to a RuvA octameric sandwich blocking the access of RuvC to the junction. It was interesting to test whether preventing the octamer formation would affect the ability of RuvA3m to inhibit junction resolution by RuvC. As reported previously (27), junction cleavage by RuvC was inhibited by increasing concentrations of WT RuvA (Fig. 7). In the presence of RuvA3m, inhibition of RuvC cleavage was still observed but it required much higher concentrations of the mutant. Approximately 3 times higher concentrations of RuvA3m compared with WT RuvA were required to achieve 50% inhibition of junction resolution by RuvC (Fig. 7). Interestingly, at very low protein concentrations, a modest but reproducible stimulation of RuvC cleavage was observed in the presence of both WT RuvA and RuvA3m (Fig. 7, inset). This observation supports the idea that RuvC is more efficient in resolving a junction that is already bound by RuvA, in agreement with the proposed function of a RuvABC resolvasome. However, the mutant RuvA3m did not stimulate cleavage more efficiently than the wild type and still inhibited the reaction at high concentrations. It should be noted that junction cleavage by RuvC is sequence-specific. The 12-bp homologous core of junction X12 allows some branch migration, and the enzyme could choose from a population of junction molecules for those that present a cleavable sequence at the cross-over. The binding of RuvA (in the absence of RuvB) is likely to prevent spontaneous branch migration of the junction. This could limit the choice of RuvC for cleavable sequences and in this way inhibit resolution.

Comparative SPR Analysis of RuvA3m and WT RuvA Interactions with Holliday Junction and RuvB and RuvC Protein-
The binding of WT RuvA and RuvA3m to Holliday junctions and their interactions with RuvB and RuvC on the junction were tested using the SPR technique that allows real-time monitoring of the interaction of biotinylated DNA immobilized on the surface of streptavidin-coated sensor chip with proteins in solution passed over the chip. Initial experiments were designed to investigate the binding of WT or mutant RuvA proteins to Holliday junctions and RuvB. The synthetic junction Jbm5 used in these experiments contained a homologous core 20 bp/arm (49) and one biotinylated strand for attachment to the sensor chip. Both WT RuvA and RuvA3m readily bound to junction Jbm5 (Fig. 8) and were retained on the junction for the dissociation time given (600 s). However, the binding profiles of WT RuvA (dashed line) and RuvA3m (solid line) were strikingly different, reflecting their different affinities for Holliday junctions as shown by gel assays (Fig. 3). Moreover, the amount of protein retained on the chip, expressed in RU, was also different for WT RuvA and RuvA3m at the same protein concentrations, displaying a difference in the stoichiometry of protein binding to DNA. The stoichiometry was calculated based on the value of 1 RU corresponding to 0.73 pg⅐mm Ϫ2 for bound DNA (56) and 1 pg⅐mm Ϫ2 for protein (57). At the protein concentrations used, a single Holliday junction bound ϳ8 monomers (2 tetramers) of WT RuvA but up to ϳ25 monomers (6 tetramers) of RuvA3m. To examine the formation of Holliday junction-RuvA⅐RuvB ternary complex, RuvB protein was injected over the chip surface already containing Jbm5⅐RuvA or Jbm5⅐RuvA3m complexes (Fig. 8A). Again, a clear difference in the sensograms could be seen. The addition of RuvB to the Jbm5⅐RuvA preformed complex gave a binding pattern corresponding to fast association. After the end of the association phase, some of the RuvB protein was retained to the Jbm5⅐wtRuvA complex. However, when RuvB at the same concentration was added to Jbm5⅐RuvA3m, a very unusual but reproducible "hammock-shaped" sensogram was obtained. The most obvious interpretation of this observation would be a biphasic interaction of RuvB with the preformed Jbm5⅐RuvA3m complex. The first phase might reflect dissociation of RuvA3m resulting from possible interactions of RuvB with DNA and/or RuvA3m. The second phase could reflect the binding of RuvB to RuvA and the formation of a ternary DNA⅐RuvA3m⅐RuvB complex. However, the sensogram did not show the increase in RU level as observed with WT RuvA. Since gel-filtration experiments clearly demonstrated stable interactions between RuvA3m and RuvB, this result suggests that inappropriately bound RuvA3m is substituted by RuvB.
Further experiments were performed by binding either WT RuvA or RuvA3m to the junction first followed by the addition of RuvC (Fig. 8B). The concentrations of both WT RuvA and RuvA3m protein were increased 2-fold in these experiments to improve the sensitivity of detecting interactions of RuvC with the RuvA-Holliday junction complexes. Again, a clear difference can be seen in the binding of WT RuvA and RuvA3m with the immobilized DNA junction. Neither the profile of binding nor the amount of WT RuvA bound to DNA changed when the protein concentration was raised up to 1000 nM (Fig. 8B). In contrast, when RuvA3m concentration was increased, both the profile of binding and the amount of bound protein changed  (Fig. 8B). The binding profile for RuvA3m clearly showed a two-phase association. The initial steep phase could reflect interaction of either a single protein tetramer (or a stack of tetramers) with DNA, whereas the second phase could result from interactions between free RuvA3m tetramers and preformed RuvA3m-junction complexes. Subsequent addition of RuvC to the WT RuvA-Holliday junction complex resulted in the association of RuvC protein with the complex (Fig. 8B). However, there was no difference in the RU level before and after the association phase, indicating that no RuvC protein remained bound to the complex. In contrast, the interaction of RuvC with the RuvA3m-Holliday junction complex showed a steeper association pattern. Moreover, after the end of the association phase, some RuvC protein remained bound to RuvA3M-Holliday junction complex. This suggests the formation of a ternary complex of junction-RuvA3m⅐RuvC complex in vitro.
Complementation Test of the RuvA3m in Vivo-The biological activity of RuvA3m was tested by measuring its ability to rescue from a multicopy plasmid the UV sensitivity of the ruvA⌬ E. coli strain HRS23000 (44). The pET21b(ϩ)-derived expression plasmids for WT RuvA or RuvA3m were introduced in ruv ϩ wild-type (AB1157) or ruvA⌬ (HRS2300) E. coli cells.
To derepress the region upstream of ruvA and allow expression by E. coli RNA polymerase, the cells were treated by 0.1 mM IPTG for 1 h at 37°C before plating and UV irradiation. The ruvA⌬ mutants showed a pronounced UV sensitivity as expected (Fig. 9A). The UV sensitivity was rescued in HRS23000 cells harboring the wild-type expression plasmid pET21ruvA with survival levels reaching those of the ruv ϩ AB1157 cells with vector plasmid (Fig. 9). In contrast, cells harboring pET21ruvA3m remained as UV-sensitive as HRS2300 cells with vector. Western blot analysis of cell lysates showed that both RuvA ϩ and RuvA3m were expressed at comparable levels after induction with IPTG (Fig. 9B). Thus, it does not seem likely that the inability of RuvA3m to complement the repair deficiency was due to differences in protein expression or stability in vivo. The lack of complementation could be due to the decreased efficiency of RuvA3m binding to Holliday junctions. Achieving a higher level of expression of the RuvA mutant might partially overcome this defect, but nonspecific binding to DNA would cause a concern. Moreover, the high level of RuvA expression would be deleterious in itself. It seems more likely that the inability of the RuvA3m mutant to rescue the UV sensitivity was due to inefficient branch migration, as seen with synthetic junctions in vitro. In a similar study (51), mutants in the acidic pin region of RuvA showed an ϳ50% efficiency of branch migration in vitro but were severely deficient in complementing the DNA repair defect of ruvA mutants.
No effects on UV survival were observed by expressing either RuvA or RuvA3m in ruvA ϩ E. coli cells (data not shown). The ruvAB operon would be induced by UV in these cells, and the concentration of RuvA ϩ could well be much higher than that of mutant RuvA3m. In addition, the ability of mixed wild-type RuvA3m tetramers to form double tetramers on junctions may be partially restored. Overall, the results of the in vivo experiments reveal a crucial role for RuvA octamer formation in DNA repair. DISCUSSION The triple mutant RuvA3m used in this study was designed to prevent tetramer-tetramer interactions to investigate the role of double tetramer formation for the activity of RuvA on Holliday junctions. Single substitutions of amino acid residues involved in hydrophobic interactions (V126R) or ionic interactions (E127R) as well as a double mutant RuvA2m (E127R,D130R) were constructed and tested, but all of the mutant RuvA proteins were found to form complexes I and II normally and to supported efficient branch migration in vitro. 2 These preliminary experiments suggested that disabling the ability of RuvA to form complex II would require not only the loss of specific interactions in the tetramer-tetramer interface but also the introduction of charge repulsion. This would destabilize a complex of two tetramers, each binding to opposite sides of the junction without any protein-protein interactions such as the mixed complexes observed with RuvA and the unrelated fission yeast Holliday junction resolvase Ydc2 (58). The concentration of positively charged residues introduced in the triple RuvA3m mutant, however, most probably created a nonspecific DNA binding site, resulting in its ability to bind duplex DNA and to form aberrant complexes on the junction at 2 A. Keeley and I. Tsanera, unpublished data. high protein concentrations (Fig. 2). Up to six RuvA3m tetramers could be accommodated on the junction according to the BIAcore results (Fig. 8). The aberrant complexes seen in gel assays could be competed effectively by short dsDNA, whereas the junction-specific complex I could not. Therefore, RuvA3m retained its high structure specificity of binding to the junction. However, RuvA3m showed a reduced efficiency of complex formation (Fig. 3). Because the binding curves were obtained from EMSA data, the results most probably reflect the lower stability of complex I compared with complex II during electrophoresis. The affinity of RuvA3m for the junction is probably higher than that suggested from Fig. 3.
RuvA3m formed an active enzymatic complex with RuvB, which displayed DNA helicase and branch migration activities. Therefore, the disruption of the tetramer-tetramer interface did not inhibit the activity of RuvAB in vitro. However, the rate FIG. 9. In vivo assay for survival of UV-irradiated ruvA⌬ strain HRS2300 expressing RuvA or RuvA3m from multicopy plasmids. A, survival curves containing values from 3 to 5 independent experiments. B, Western blot analysis of lysed cells of HRS2300 or AB1157 carrying plasmids expressing RuvA or RuvA3m as indicated. Control lanes contain purified RuvA or RuvA3m. and overall efficiency of branch migration of synthetic four-way junctions were lower in reactions with the mutant RuvA3m⅐RuvB complex. In contrast, DNA helicase activity and branch migration of three-way junctions appeared unaffected. Branch migration of Y-junctions is driven by a single RuvB hexamer (39). The defined polarity of the DNA helicase activity (50) also implies the involvement of a single RuvB hexamer in this reaction. These results lead us to propose that one RuvA tetramer can only support the activity of one RuvB hexamer. Such a "single-engine" RuvAB complex would be capable of unwinding a short region of duplex DNA, unwinding one strand of a synthetic Y-junction, and branch migration of small synthetic junctions. However, the efficiency of a single RuvB hexamer would be significantly reduced compared with two opposed RuvB hexamers operating across a junction, as shown experimentally on model replication fork substrates (41). The altered efficiency of branch migration observed in reactions with the RuvA3m mutant could reflect the activity of such a "single-engine" RuvAB complex. However, the rate and processivity of branch migration would be compromised and this could have serious consequences in vivo. It could explain why the RuvA3m mutant was unable to complement the DNA repair deficiency of ruvA⌬ E. coli cells (Fig. 9). Fast processive branch migration over long stretches of DNA and through damaged DNA and heterology could be crucial for DNA repair. The short synthetic DNA junctions used in vitro clearly cannot reveal these aspects of RuvAB activity.
The inability of RuvA3m to rescue the UV sensitivity of ruvA⌬ cells could also reflect impaired interactions within the RuvABC resolvasome complex. The results obtained in this study provide evidence that the mutant RuvA3m can interact with RuvC on the Holliday junction. RuvA3m at low concentrations was able to stimulate cleavage of the junction by RuvC to a similar level as wild-type RuvA (Fig. 7, inset), and the SPR results obtained with RuvA3m show evidence for the formation of a ternary complex of junction-RuvA3m⅐RuvC. However, functional interactions could be impaired between RuvA3m and RuvC and within a mutant RuvA3mBC complex, which could affect its ability to perform coupled branch migration and resolution. The activity of RuvC in vivo is RuvAB-dependent, and interactions with RuvB are particularly important. RuvB stimulates greatly junction cleavage by RuvC, and the assembly of the RuvB rings also dictates the orientation of junction resolution (17,59). RuvC acts as a dimer, and according to the RuvABC resolvasome model, each RuvC monomer interacts with one RuvB hexamer on opposite arms of the junctions. If RuvA3m was unable to assemble two hexameric RuvB rings across the junction, this might have serious implications for the activity of RuvC. It would be interesting to test whether RuvC cleavage could be stimulated by a single RuvB hexameric ring. Clearly, the coordinated action of the RuvABC system in vivo involves specific interactions among all three proteins. Moreover, there may be important interactions between RuvABC and other proteins in the cell, which could be affected. The results of this investigation give a strong indication that interactions between the two RuvA tetramers are likely to be critical for the full biological activity of the RuvABC system, as suggested by the high evolutionary conservation of the amino acid residues involved in these interactions.