Branch Migration of Holliday Junctions Promoted by the Escherichia coli RuvA and RuvB Proteins COMPARISON OF RuvAB- AND RuvB-MEDIATED REACTIONS*

The Escherichia coli RuvA and RuvB proteins me- diate the branch migration of Holliday junctions in vitro. In the presence of stoichiometric amounts of RuvB (1 RuvB dimer/l2 nucleotides), branch migration can occur without need for RuvA. However, RuvA is required when the RuvB concentration is reduced 4-fold or more. Under optimal conditions, we found the minimal protein requirement to be 1 RuvB dimer per 500-1100 nucleotides and 1 RuvA tetramer per 600- 1200 nucleotides. To determine the roles of RuvA and RuvB in branch migration, we compared branch mi- gration reactions mediated by RuvB only and by RuvA and RuvB. The time courses of the two reactions were similar, and both required ATP and Mg2+. However, RuvB-mediated branch migration occurred at lower ATP concentrations (2200 PM) and higher Mg2+ concentrations (210 mM MgC12) than the reaction me- diated by RuvA and RuvB (? 1 mM ATP, 2 5 mM MgC12). The Mg2+ requirement for RuvB-mediated branch mi- gration reflects the Mg2+ requirement of RuvB for binding and can be overcome by addition of RuvA. These results indicate that RuvA protein facilitates the interaction of RuvB with generally 26 nM RuvA and 16 nM RuvB (RuvAB-mediated branch migration) or 670 nM RuvB only (RuvB-mediated branch migration). The products of the reactions were analyzed by 0.8% agarose gel electrophoresis following depro- teinization with SDS, EDTA, and proteinase K. Agarose Gel Electrophoresis-Gels were run in TAE buffer (30) at room temperature at 6 V/cm with buffer recirculation. To visualize DNA by autoradiography, the gels were dried and exposed to Kodak XAR films. Autoradiographs were quantitated by densitometry using a Molecular Dynamics model 300 laser densitometer and ImageQuant software. The fraction of 3ZP-labeled linear DNA in recombination intermediates was measured, and the percentage of recombination intermediates converted into branch migration products was determined. ATPase Assay-Assays were 37 mM Tris- mM mM mM and pg/ml bovine serum albumin. Reactions (90 pl) were prewarmed for 2 min and started by addition of either RuvB (670 nM) or RuvA (26 nM) and RuvB (16 nM). Aliquots (9 pl) were stopped prior to the addition of protein or at the indicated times after protein addition by addition of SDS, EDTA, and protein- ase K (to 0.5%, 40 mM, and 2 mg/ml, respectively). Samples were incubated for 10 min at 37 "C, and 1-pl aliquots were applied to a thin (Polygram). Chromatograms in M M correspond-ing to ADP

The biochemical properties of the Ruv proteins have been studied i n uitro, and their likely roles in recombination have * This work was supported by the Imperial Cancer Research Fund.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. $ Supported in part by the Swiss National Science Foundation. been determined. The RuvC protein is a nuclease which resolves recombination intermediates by specific cleavage of the Holliday junction (15)(16)(17). The RuvA protein is a DNA binding protein that interacts with the RuvB ATPase (18,19) to promote the branch migration of Holliday junctions (19)(20)(21). Branch migration has been demonstrated in uitro using (i) synthetic Holliday structures formed by annealing four oligonucleotides (X), and (ii) Holliday junctions made by RecA protein (20,ZZ). The reaction is dependent on ATP and M$+ and occurs without a demonstrable polarity (20).
The interaction of RuvA with RuvB has been demonstrated by glycerol gradient centrifugation (23) and by the formation of specific RuvAB-Holliday junction complexes (24). Since RuvA protein binds specifically to Holliday junctions (21,24), it is likely that one role of RuvA in branch migration is to direct the RuvB ATPase to the junction.
In this manuscript, we extend our studies of branch migration of Holliday junctions mediated by RuvA and RuvB. We confirm that stoichiometric amounts of RuvB can promote branch migration in the absence of RuvA (RuvB-mediated branch migration) and compare this reaction with that catalyzed by RuvA and RuvB (RuvAB-mediated branch migration). We find that the speed and efficiency of RuvB-and RuvAB-mediated branch migration reactions are similar.
However, a systematic investigation of the ATP and M$+ requirements shows that the requirements for branch migration are different and are related to the DNA binding properties of RuvA and RuvB ( 2 5 ) . Although it is unlikely that the RuvB-mediated branch migration reaction occurs in vivo (since ruuA mutants are repair-defective), a comparison of the two i n uitro reactions now allows us to define the individual roles played by the RuvA and RuvB proteins in branch migration and the recombinational repair of DNA.

MATERIALS AND METHODS
Enzymes-RecA protein of E. coli was purified using a modification of a previously published procedure (26). RuvA and RuvB proteins of E. coli were purified as described (27). Both proteins were greater than 99% pure, as judged by SDS-polyacrylamide gel electrophoresis followed by Coomassie Blue staining. The concentrations of RuvA and RuvB were determined by the Bradford (34) and Lowry (35) methods (Bio-Rad and Sigma protein assay kits, respectively) using bovine serum albumin as a standard. In previous studies (20,21,22,27), the protein concentrations were determined using ovalbumin as a standard, which led to an overestimate of the protein concentration. Unless stated otherwise, protein concentrations are defined in moles of protein monomers. DNA Substrates-Gapped duplex DNA (gDNA)' with a defined stranded $X174 (+) DNA with the denatured 5224-base pair PstI-162-nucleotide gap was prepared by annealing circular single-' The abbreviations used are: gDNA, circular 6x174 duplex DNA with a single-stranded gap in the (-) strand between the PstI and AuaI sites; ATP+, adenosine 5'-0-(3-thiotriphosphate).
B r a n c h Migration of Holliday Junctions by RuvAB and RuvB AvaI fragment of duplex $X174 DNA. DNA was prepared and annealing was performed essentially as described (28,29). Homologous linear duplex DNA was produced by PstI cleavage of $X174 RFI DNA, followed by 3'-end-labeling using [ W~~P ]~~A T P and terminal transferase (Amersham). DNA concentrations are expressed in moles of nucleotides.
Preparation of Recombination Intermediates-Strand exchange reactions (100-150 pl) were performed in 20 mM Tris-HC1 (pH 7.5), 15 mM MgClz, 2 mM ATP, 20 mM phosphocreatine, 6 units/ml phosphocreatine kinase, 2 mM dithiothreitol, and 100 pg/ml bovine serum albumin. gDNA (26 pM) and 32P-end-labeled linear duplex $X174 DNA (20 pM) were incubated in the reaction mixture for 2 min at 37 "C, and strand exchange was initiated by the addition of RecA protein (IO pM). Incubation was for 15 min at 37 "C. Reactions were stopped and deproteinized by addition of 1/4 volume of stop mixture (100 mM Tris-HC1 (pH 7.5), 2.5% (w/v) SDS, 200 mM EDTA, and 10 mg/ml proteinase K) followed by incubation for 15 min at 37 "C. The mixture was then applied to a 3.5-ml Sepharose CL-2B column equilibrated with 0.5 mM EDTA (pH 8.0), and the deproteinized DNA was collected in two drop fractions. The DNA concentration was determined by quantitation of 32P-labeled DNA.
Agarose Gel Electrophoresis-Gels were run in TAE buffer (30) at room temperature at 6 V/cm with buffer recirculation. To visualize DNA by autoradiography, the gels were dried and exposed to Kodak XAR films. Autoradiographs were quantitated by densitometry using a Molecular Dynamics model 300 laser densitometer and ImageQuant software. The fraction of 3ZP-labeled linear DNA in recombination intermediates was measured, and the percentage of recombination intermediates converted into branch migration products was determined.
ATPase Assay-Assays were carried out at 37 "C in 20 mM Tris-HCI (pH 7.5), 15 mM MgC12,l mM ATP, 50 pCi/rnl [cx-~'P]ATP (3000 Ci/mmol), 2 mM dithiothreitol, and 100 pg/ml bovine serum albumin. Reactions (90 pl) were prewarmed for 2 min and started by addition of either RuvB (670 nM) or RuvA (26 nM) and RuvB (16 nM). Aliquots (9 p l ) were stopped prior to the addition of protein or at the indicated times after protein addition by addition of SDS, EDTA, and proteinase K (to 0.5%, 40 mM, and 2 mg/ml, respectively). Samples were incubated for 10 min at 37 "C, and 1-pl aliquots were applied to a CEL 300PEI/UVZ5, thin layer chromatography plate (Polygram). Chromatograms were developed in 1 M formic acid, 0.5 M LiCI. The plates were dried and exposed to x-ray film, and the spots corresponding to ATP and ADP were excised. The amount of radioactivity in each spot was determined by scintillation counting using Aquasol scintillation fluid (Du Pont-New England Nuclear).
Preparation of Antibodies-Rabbit polyclonal antibodies were raised against homogeneous RuvA and RuvB proteins (27). Preimmune sera were taken prior to immunization. Ruv proteins (200 pg in Freund's adjuvant) were injected subcutaneously at 2-week intervals. Rabbit IgGs were isolated from serum taken after the third and fifth injection using protein A-Sepharose (Pharmacia LKB Biotechnology Inc.) as described (31). The antibodies were then precipitated by the addition of 0.5 volume of saturated ammonium sulfate, collected by centrifugation (30 min, 3000 X g), resuspended in phosphate-buffered saline (171 mM NaCl, 3.3 mM KCI, 10.1 mM Na2HPO4, 1.8 mM KHzP04 at pH 7.2), and dialyzed against 3 changes of 600 ml of phosphate-buffered saline (31). If necessary, antibodies were concentrated by centrifugation (2000 X g) using Centricon-30 microconcentrators (Amicon). The antibody concentrations were determined assuming an Am of 1.35 for a concentration of 1 mg/ml.

RESULTS
In previous studies, we used recombination intermediates made by RecA to show that the RuvA and RuvB proteins catalyze the branch migration of Holliday junctions (20, 22). Two reactions were observed (i) a RuvB-mediated reaction which required stoichiometric amounts of RuvB protein but was independent of RuvA protein, and (ii) a RuvA/RuvBmediated reaction which required much lower concentrations of RuvB protein.
To understand the RuvAB-and RuvB-mediated branch migration reactions further, we used the same branch migration assay. In the following experiments, RuvB-mediated reactions will be carried out using 670 nM RuvB, whereas the RuvAB-mediated reactions will be performed at 26 nM RuvA and 16 nM RuvB.
To prepare recombination intermediates, gDNA was incubated with homologous 32P-end-labeled linear duplex DNA in the presence of RecA protein, and the resulting DNA species were deproteinized and purified by gel filtration. The recombination intermediates (which form a diffuse band on a gel; Fig. lA, lune a) look like a-structures in the electron microscope and are stable for several hours at 37 "C in the absence of added protein ( 2 2 ) . When incubated with 26 nM RuvA and 16 nM RuvB, we observed a loss of label at the position of intermediates, indicating their dissociation with time ( Fig.  lA, lunes b-g). This loss of label coincided with an increase in the amounts of label at the position of circular and linear duplex DNA, as expected from branch migration and as described previously (20). At this RuvB concentration, the branch migration reaction shows an absolute requirement for RuvA (20). However, at much higher RuvB concentrations, the need for RuvA is overcome, and a similar time course was observed in the presence of 670 nM RuvB alone (Fig. lA, lunes j-0). The two time courses, which appeared identical by visual analysis (Fig. M), were quantitated using a laser densitometer, and the amounts of 32P-labeled intermediates were determined at each time point. Fig. 1B shows the mean values calculated from five independent experiments and indicates that the rates of dissociation are similar within the limits of experimental variation.
Since RuvB protein was purified from an overexpression vector carried by a bacterial strain from which the ruuA gene was inactivated by transposon insertion (27), it is unlikely that trace amounts of contaminating RuvA protein could be contributing to the RuvB-mediated branch migration reaction. However, to unequivocally rule out this possibility, we tested the effect of polyclonal antibodies raised against RuvA and RuvB on branch migration reactions catalyzed by RuvAB (Fig. 2, lunes b-g) or by RuvB alone (lunes h-1). We observed that branch migration reactions catalyzed by RuvAB (lunes b-g) were blocked by preincubation with antibodies raised against either RuvA (lune d ) or RuvB (lune f ) . Similar reactions to which preimmune serum was added gave rise to branch migration products (lunes e and g), as did controls without antibody addition (lunes b and c). However, in RuvBmediated branch migration reactions (lunes h-l), only the incubation with RuvB antibodies (lune k ) prevented branch migration, since the RuvA antibodies had no effect on this reaction ( l a n e i). These results confirm that the branch migration reaction which occurs in the presence of stoichiometric amounts of RuvB is independent of RuvA.
To determine the effect of the RuvA and RuvB concentrations on branch migration, reactions containing different amounts of RuvA and RuvB protein were performed (Fig. 3). As shown above, reactions carried out with 670 nM RuvB (corresponding to a ratio of 1 RuvB monomer per 6 nucleotides) showed no requirement for RuvA. When the RuvB concentration was reduced to 160 or 16 nM, RuvA protein (7 nM) was required for branch migration. A plateau was reached at 14 nM RuvA, and, at very high concentrations (1.7 p M ) , we

Branch Migration of Holliday Junctions by RuvAB and RuvB
observed that RuvA protein inhibited branch migration in the presence of 160 and 670 nM RuvB. We presume that the excess RuvA protein inhibits the reaction by competing with the RuvAB protein complex for DNA. In the experiment shown in Fig. 3, efficient branch migration at a low RuvB concentration (16 nM) required 214 nM RuvA (at a DNA concentration of 4.2 PM, expressed in nucleotides). Other experiments performed at 8 nM RuvB required 27 nM RuvA (data not shown). These results indicate that RuvAB-mediated branch migration requires approximately 1 RuvB dimer per 500-1100 nucleotides and 1 RuvA tetramer per 600-1200 nucleotides. These estimates are based on observations which show that RuvA and RuvB are tetrameric and dimeric, respectively (23,27).
The addition of RuvA and RuvB to RecA-mediated in vitro recombination reactions facilitates branch migration past . To determine whether RuvA was specifically required for this reaction, we tested the effect of UV-irradiation on the branch migration reaction catalyzed by RuvB or by RuvAB. We found that both reactions were slowed down to a similar extent, and the inhibitory effect was proportional to the dose of UV irradiation (data not shown). Thus, branch migration past UV-lesions does not specifically require the presence of RuvA protein.
To determine whether RuvAB-and RuvB-mediated branch migration reactions have different cofactor requirements, we varied the ATP and Mg2+ concentrations. Both RuvAB-and RuvB-mediated branch migration reactions were dependent upon the presence of ATP (Fig. 4). However, the ATP requirements were different, with the RuvB-mediated reaction requiring less ATP (20.2 mM) than the RuvAB-mediated reaction (21 mM).
Both reactions were dependent on the presence of Mg2+ (Fig. 5A), but again, the optimal M e concentrations were found to differ. RuvAB-mediated branch migration occurred with low efficiency at 5 mM M&12 and was maximally efficient at 10 mM MgC12. In contrast, higher MgC12 concentrations were required for RuvB-mediated branch migration. In this case, no reaction was observed at 5 mM MgC12, and the optimal concentration was found to be 215 mM M&l2. At low M$12, the inability of RuvB to promote branch migration (Fig. 5B, lane c ) was overcome by addition of RuvA ( l a n e b).
The effects of NaCl and pH on RuvAB-and RuvB-mediated branch migration were also investigated (at 15 mM MgC12 and 2 mM ATP). Both reactions were inhibited in a similar way by NaCl, with 50% inhibition at 200 mM (data not shown). Branch migration occurred efficiently between pH 5.5 and pH 10.5, the highest pH tested (data not shown).
Above, we showed that ATP was required for RuvAB-and  RuvB-mediated branch migration. Analysis of the RuvB ATPase has shown that ADP is an effective competitor for ATP binding (18). In addition, ADP cannot replace ATP for RuvAB-mediated branch migration (20). To test whether ADP inhibits ATP-dependent branch migration, a series of reactions were performed in the presence of 2 mM ATP and various amounts of ADP. Both RuvAB-and RuvB-mediated reactions were found to be sensitive to the presence of ADP. However, the two reactions were inhibited to a different extent (Fig. 6). For example, the RuvB-mediated branch migration reaction was severely inhibited by 2 mM ADP (ADP:ATP = l:l), whereas 6 mM ADP was requiredto inhibit the RuvAB-mediated reaction to the same extent (ADP:ATP = 3:l).
We found previously that ATPrS, a nonhydrolyzable analog of ATP, could not replace ATP in the RuvAB-mediated branch migration reaction (20), indicating that ATP hydrolysis was required. To test whether ATP hydrolysis was also required for RuvB-mediated branch migration, the two reactions were performed in the presence of 2 mM ATP and various amounts of ATPrS. As expected, we found that both the RuvAB-and RuvB-mediated branch migration reactions were sensitive to the presence of ATP-yS (Fig. 7). RuvABmediated branch migration occurred in the presence of ATPyS, up to a concentration of 3 mM (ATPyS:ATP = 3:2). In this reaction, a complete block was observed at 5 mM ATPyS (ATP7S:ATP = 52). By comparison, we observed that RuvB-mediated branch migration was significantly more sensitive and was severely inhibited in the presence of 0.5 mM ATPyS and blocked by 1 mM ATPrS (ATP7S:ATP = 1:2).
The competition experiments with ADP and ATPyS confirm that ATP binding and hydrolysis are required for branch migration and show that the RuvB-mediated branch migration reaction is more sensitive to competitors than the RuvAB-mediated reaction, Interestingly, when we measured the amounts of ATP hydrolyzed during the RuvAB and RuvB reactions, we found that the RuvAB reaction was more energy efficient than the reaction catalyzed by RuvB.
The data presented in Fig. 8 show time courses of ATP hydrolysis in RuvAB-and RuvB-mediated branch migration reactions. In the RuvB-mediated reaction, significant amounts of ATP were hydrolyzed. However, in this experiment, we have been unable to correlate ATP hydrolysis with branch migration since the time course shows that ATP hydrolysis continues after 10 min, the time at which the majority of intermediates had been dissociated (Figs. 1 and 8). In contrast, the level of ATP hydrolysis in the RuvABmediated reaction hardly exceeded the background.

DISCUSSION
In this study, we have used deproteinized recombination intermediates to investigate the branch migration of Holliday junctions by the E. coli RuvA and RuvB proteins. We reported previously that RuvB protein is involved in two different types of in vitro branch migration reactions. In the presence of stoichiometric amounts of RuvB, branch migration occurred in the absence of RuvA, while RuvA was required for branch migration in the presence of lower amounts of RuvB (20). In this paper, we confirm that branch migration reactions can be mediated in vitro by RuvB alone and have compared the time courses and cofactor requirements of the RuvAB-and RuvB-mediated branch migration reactions. Although we observe a number of similarities, we also find conditions in which the two reactions can be distinguished from each other.

100'1
A RuvA + RuvB I The time courses of branch migration and the final yield of RuvAB-and RuvB-mediated products were very similar (Fig.  1). We determined the protein requirements for the two reactions and found that RuvB-mediated branch migration occurred at a ratio of 1 RuvB dimer/l2 nucleotides of DNA. Interestingly, at this RuvB to nucleotide ratio, duplex DNA is maximally protected by RuvB (in the presence of ATP+) from attack by DNase I and phosphodiesterase I (25). The need for high RuvB concentrations is likely to be due to the low affinity for DNA shown by RuvB protein. Alternatively, it is possible that RuvB-mediated branch migration requires the formation of a continuous nucleoprotein filament, as has been observed with RecA protein (32).
We determined the amounts of RuvA and RuvB required for RuvAB-mediated branch migration and found that the minimal protein requirement was 1 RuvB dimer per 500-1100 nucleotides and 1 RuvA tetramer per 600-1200 nucleotides.
Previously, we demonstrated that RuvA protein interacts with synthetic Holliday junctions to form specific protein-DNA complexes (21). More recently, we have used bandshift assays to detect the formation of a RuvA-RuvB-Holliday junction complex (24). In the experiments described here, we show that the requirement for RuvB is substantially reduced by the presence of RuvA. Presumably, the RuvA protein specifically targets RuvB to the region of the junction. This hypothesis is supported by recent observations which show that RuvA and RuvB interact directly in solution to form a RuvAB protein complex (23). The specific interaction of RuvA with the Holliday junction might also explain the observation that branch migration reactions are inhibited by excess RuvA (1 RuvA tetramer/lO nucleotides) (Fig. 3). Most likely, inhibition was due to competition between RuvA and RuvAB for the Holliday junction.
In addition to their protein requirements, branch migration reactions mediated by RuvAB or RuvB can be distinguished by their cofactor requirements. For example, RuvAB-mediated branch migration needed higher concentrations of ATP than the reaction performed by RuvB only (Fig. 4). Moreover, the addition of ADP (Fig. 6) or ATPrS (Fig. 7) to reactions containing ATP was found to inhibit branch migration by RuvAB or RuvB to different extents. Reflecting its low requirement for ATP, RuvB-mediated branch migration was more sensitive to the competitive effects of ADP and ATP-@. We cannot rule out the possibility that the difference in nucleotide cofactor requirements between the RuvAB-and RuvB-mediated reactions was caused by the change in RuvB concentration. However, we feel this is unlikely since RuvA directly stimulates the RuvB ATPase in the presence of DNA (19).
RuvAB-and RuvB-mediated branch migration reactions also had different M$+ requirements (Fig. 5 ) , with the RuvBmediated reaction requiring concentrations of MgClz 2 10 mM. The need for 210 m M MgClZ correlates with the M e dependence of duplex DNA binding by RuvB, as observed by gel analyses and stimulation of the RuvB ATPase by DNA (25). In these experiments, we observed a direct interaction between RuvB and duplex DNA in the presence of ATP at MgClz concentrations above 10 mM, indicating that M$+ affects the binding of RuvB to DNA.
The presence of RuvA protein allows branch migration at lower M$+ concentrations. In experiments that analyzed the interaction of RuvA and RuvB with DNA (25), we found that RuvA facilitates the interaction of RuvB with duplex DNA a t low Mg2+ concentrations, leading to the formation of a complex consisting of RuvA, RuvB, and duplex DNA. From these results, we conclude that RuvA protein is able to facilitate the interaction of RuvB with DNA and more specifically with Holliday junctions.
In conclusion, our experiments confirm that RuvB-mediated branch migration can occur in the absence of RuvA, provided stoichiometric amounts of RuvB are present (20). This supports the notion that RuvB, which is an ATPase and contains structural motifs characteristic of DNA helicases (33), is the motor of branch migration. In the presence of lower concentrations of RuvB, RuvA is required for branch migration, and this reaction is likely to be more physiologically relevant than the RuvB-mediated reaction. Our study of RuvAB-mediated branch migration indicates that RuvA facilitates the interaction of RuvB with DNA, by the formation of a RuvA-RuvB-DNA complex, and directs RuvB to the Holliday junction (21,24,25). The formation of a RuvAB-Holliday junction complex leads to movement of the joint by ATP-driven branch migration. The mechanism of this reaction is presently unknown, but is likely to involve strand separation followed by reannealing since RuvAB protein exhibits DNA helicase activity in vitro (33).