Cleavage of Holliday Junctions by the Escherichia coli RuvABC Complex*

The Escherichia coli RuvABC proteins process recombination intermediates during genetic recombination and recombinational repair. Although early biochemical studies indicated distinct RuvAB-mediated branch migration and RuvC-mediated Holliday junction resolution reactions, more recent studies have shown that the three proteins act together as a “resolvasome” complex. In this work we have used recombination intermediates made by RecA to determine whether the RuvAB proteins affect the sequence specificity of the RuvC resolvase. We find that RuvAB proteins do not alter significantly the site specificity of RuvC-dependent cleavage, although under certain conditions, they do affect the efficiency of cleavage at particular sites. The presence of RecA also influences cleavage at some sites. We also show that the RuvAB proteins act upon transient strand exchange intermediates made using substrates that have the opposite polarity of those preferred by RecA. Together, our results allow us to develop further a model for the recombinational repair of DNA lesions that lead to the formation of post-replication gaps during DNA replication. The novel features of this model are as follows: (i) the RuvABC resolvasome recognizes joints made by RecA; (ii) resolution by RuvABC occurs at specific sites containing the RuvC consensus cleavage sequence 5 * -(A/T)TT 2 (G/C)-3 * ; and (iii) Holliday junction resolution often occurs close to the initiating gap without significant heteroduplex DNA formation.

The Escherichia coli RuvABC proteins process recombination intermediates during genetic recombination and recombinational repair. Although early biochemical studies indicated distinct RuvAB-mediated branch migration and RuvC-mediated Holliday junction resolution reactions, more recent studies have shown that the three proteins act together as a "resolvasome" complex. In this work we have used recombination intermediates made by RecA to determine whether the RuvAB proteins affect the sequence specificity of the RuvC resolvase. We find that RuvAB proteins do not alter significantly the site specificity of RuvC-dependent cleavage, although under certain conditions, they do affect the efficiency of cleavage at particular sites. The presence of RecA also influences cleavage at some sites. We also show that the RuvAB proteins act upon transient strand exchange intermediates made using substrates that have the opposite polarity of those preferred by RecA. Together, our results allow us to develop further a model for the recombinational repair of DNA lesions that lead to the formation of post-replication gaps during DNA replication. The novel features of this model are as follows: (i) the RuvABC resolvasome recognizes joints made by RecA; (ii) resolution by RuvABC occurs at specific sites containing the RuvC consensus cleavage sequence 5-(A/T)TT2(G/C)-3; and (iii) Holliday junction resolution often occurs close to the initiating gap without significant heteroduplex DNA formation.
Several processes, including the repair of broken replication forks or DNA lesions, the reassortment of alleles, and the correct segregation of chromosomes during meiosis, employ the mechanism of homologous recombination. Genetic exchange begins with the pairing and alignment of two homologous DNA molecules. After alignment, a single-stranded end from one molecule invades the second duplex; this is followed by further strand transfer between the molecules to produce a fourstranded exchange or Holliday junction. The protein that promotes these pairing and strand exchange activities is highly conserved throughout evolution. Prokaryotes possess one such protein, typified by Escherichia coli RecA, but in eukaryotic organisms two homologues, Rad51 and Dmc1, have been iden-tified (1,2). Biochemical studies of these proteins, isolated from yeast and human cells, have confirmed that the similarity in amino acid sequence to RecA is also retained in their structural and functional characteristics (3)(4)(5)(6)(7)(8).
To maintain cell viability, Holliday junctions formed during recombination or recombinational repair must be resolved to liberate the individual DNA molecules. In E. coli, Holliday junction processing is mediated by the RuvA, RuvB, and RuvC proteins (reviewed in Ref. 9). RuvA is a junction-specific DNAbinding protein (10). The structure of RuvA has been solved both in solution and as a cocrystal with synthetic junction DNA. It binds as a tetramer to either one or both faces of the junction, thereby positioning the arms of the junction in an unfolded, almost square-planar configuration (11)(12)(13)(14)(15). RuvB forms hexameric rings that have low intrinsic affinity for DNA (16,17); through direct interaction with RuvA, however, RuvB rings are targeted to the junction (18,19). The RuvAB complex, which consists of one or two RuvA tetramers and two RuvB rings bound to opposing arms of the junction (15,20), promotes translocation of the junction (10,(21)(22)(23). Although branch migration is an isoenergetic process, the input of energy is required for processivity and directionality, and this is provided by the ATPase activity of RuvB (22).
RuvC binds specifically to junctions as a dimer (24,25) and cleaves strands of like polarity with perfect symmetry to yield two nicked duplexes (26). Cleavage by RuvC is sequence-specific, occurring preferentially at the partially degenerate tetranucleotide 5Ј-(A/T)TT2(G/C)-3Ј (24,(27)(28)(29). The failure of a class of RuvC mutant proteins that display reduced specificity of cleavage in vitro to complement ruvC mutations emphasizes the biological importance of this sequence specificity (30). Like RuvA, RuvC interacts with RuvB (31) and can target RuvB rings to the Holliday junction, resulting in a complex capable of both limited branch migration and cleavage (32). Analysis of the RuvC G114D mutant protein has indicated that RuvC and RuvA also interact physically on the Holliday junction, presumably by binding opposite faces (33).
These and other data derived from biochemical (34 -36), genetic (37,38), and model building (11,12,39) studies support the proposal that RuvABC proteins form a Holliday junctionprocessing complex that mediates concerted branch migration and resolution. Since the ruvA and ruvB genes form an operon that is regulated by the SOS system and their expression is induced as a result of DNA damage (40,41), it is possible that two types of complex might form in vivo as follows: (i) a Ruv-ABC complex capable of promoting branch migration and resolution, and (ii) a subcomplex consisting of RuvAB alone, which would mediate only branch migration (31,42). An excess of RuvAB proteins during times of cellular stress would ensure that all RuvC molecules, present in limiting amounts under normal growth conditions, are incorporated into a productive resolution complex. Such a complex is likely to be particularly useful in the recombinational repair of post-replication gaps opposite UV-induced lesions. Recent studies also suggest that the RuvAB complex may function independently of RuvC in the repair of arrested replication forks (43); in fact, the cleavage of a Holliday junction generated under these conditions has been proposed to cause a potentially lethal double-strand break.
Previously, we examined the overall characteristics of an in vitro system in which RuvABC proteins were added to an on-going RecA-mediated strand exchange reaction (31). We found that both branch migration and resolution occurred in the presence of RecA protein but did not determine whether the cleavage activity of RuvC or the preferred sites of resolution were affected at the molecular level by the presence of either RuvAB or RecA. In the studies presented below, we have examined in detail whether the association of RuvC with RuvAB affects its cleavage properties.

MATERIALS AND METHODS
Proteins-RecA, RuvA, RuvB, and RuvC proteins were purified as described (31), and concentrations are expressed in terms of monomer. Terminal deoxynucleotidyltransferase was from Amersham Pharmacia Biotech; all other restriction and modification enzymes were from New England Biolabs.
Nucleic Acids-The construction of phagemids pDEA-7Z, pAKE-7Z, and pPB4.3 has been described (27,31,44). Duplex DNA was prepared using a plasmid purification kit (Qiagen). Circular (ϩ) ssDNA 1 and gapped circular DNA were prepared as described (27,31). DNA concentrations were determined using ⑀ 280 ϭ 8784 and 6500 M Ϫ1 cm Ϫ1 for ssand double-stranded DNA, respectively, and are expressed in terms of molar nucleotide (nt) or molecule. DNAs linearized with PstI or XmnI and the 50-bp ladder size marker (Sigma) were 3Ј-end-labeled with [␣-32 P]ddATP (Ͼ5000 Ci/mmol) and terminal deoxynucleotidyltransferase. DNA linearized with BamHI was 3Ј-end-labeled with [␥-32 P]dGTP (3000 Ci/mmol) and exo Ϫ Klenow. DNA linearized with BsaI was dephosphorylated with calf intestinal phosphatase and was 5Ј-end-labeled with [␥-32 P]ATP (3000 Ci/mmol) and T4 polynucleotide kinase. In each case, unincorporated nucleotide was separated from the labeled DNA using a Microspin S-400 column (Amersham Pharmacia Biotech). Linearized DNAs were digested with a second restriction enzyme as noted, and the fragment containing the label at the end that was used to initiate strand exchange was excised from an agarose gel and purified using a gel extraction kit (Qiagen). Labeled linear phagemid DNAs were stored in TE containing 30% ethanol at Ϫ20°C, whereas labeled marker DNA was precipitated, resuspended in formamide/urea loading buffer (1 ml of formamide, 750 mg of urea, 80 l of 50 mM NaOH, 1 mM EDTA, 80 l of 0.5% bromphenol blue, 80 l of 0.5% xylene cyanol), and stored at 4°C.
Coupled Strand Exchange/Holliday Junction Processing Assay-In the standard four-stranded reaction utilizing PstI/BsaI pDEA-7Z gDNA and PstI/ScaI pDEA-7Z linear DNA, gDNA (10 M nt; 1.72 nM molecule) was preincubated with RecA protein (4 M) for 5 min at 37°C in buffer containing 50 mM Tris acetate (pH 8.0), 15 mM Mg(OAc) 2 , 20 mM KOAc, 2 mM ATP, 20 mM phosphocreatine, 2 units/ml phosphocreatine kinase, 2 mM dithiothreitol, and 100 g/ml bovine serum albumin. Following this preincubation, uniquely 3Ј-32 P-end-labeled linear duplex DNA (1.6 M; 0.29 nM molecule) was added to initiate strand exchange. Ten minutes after addition of the linear DNA, Ruv protein(s), premixed in enzyme dilution buffer (20 mM Tris acetate (pH 7.5), 150 mM NaCl, 1 mM EDTA, 0.5 mM dithiothreitol, 50% glycerol), were added as indicated; typically, the reaction was incubated for a further 20 min. When reactions were to be subjected to agarose gel electrophoresis, the products were deproteinized by adding 5ϫ stop buffer (2.5% SDS, 10 mg/ml proteinase K) to a final concentration of 1ϫ and incubating at 37°C for 10 min; 6ϫ gel loading buffer was then added to a final concentration of 1ϫ. When reactions were to be analyzed by denaturing polyacrylamide gel electrophoresis, they were stopped by adding EDTA to 50 mM, extracted once with phenol:chloroform:isoamyl alcohol (25:24:1), backextracted with TE, and precipitated at Ϫ20°C with sodium acetate and ethanol in siliconized tubes. The samples were centrifuged for 15 min at 4°C, dried under vacuum for 10 min, and resuspended in 6 l of formamide/urea loading buffer. Before loading onto the gel, the samples were heated to 95°C for 3 min and then stored on ice.
Three-stranded reactions were performed essentially as described above with the following modifications. Gapped DNA was replaced by 5 M nt (1.72 nM molecule) ssDNA. The concentration of labeled linear DNA was increased to 5 M (0.89 nM molecule). When included, SSB protein (0.45 M) was added after 5 min of preincubation, and the reaction was incubated for another 5 min before addition of the linear DNA.
Preparation of Deproteinized Recombination Intermediates-Deproteinized four-stranded and three-stranded recombination intermediates were prepared as described previously (31) using the reaction conditions outlined above.
Reverse Polarity Strand Exchange-Reactions for agarose gel analysis were conducted in a 65-l volume at 37°C in the buffer described above, with 10 M nt (1.72 nM molecule) PstI/BsaI pDEA-7Z gDNA and 4 M RecA. SSB protein (2 M) was added after 5 min, and 5 min later, 5Ј-32 P-end-labeled, BsaI pDEA-7Z linear duplex DNA (1.6 M nt; 0.27 nM molecule) was added. Ruv protein(s) were added to a final concentration of 200 nM RuvA, 200 nM RuvB, and 100 nM RuvC. Aliquots (10 l) were removed at the indicated times and deproteinized as described above, except that the 5ϫ stop buffer also contained 2.5 g/ml ethidium bromide. Reactions were incubated with the linear DNA for 20 min before addition of the Ruv proteins.
Electrophoresis and Analysis-Reaction products were separated on 1% agarose gels in TAE buffer (40 mM Tris acetate (pH 8.4), 1 mM EDTA) at 4.6 V/cm for 4 h at room temperature with buffer recirculation. Both the gels and the running buffer contained 0.5 g/ml ethidium bromide. For examination of the sites of cleavage, reaction products were separated on either 5 or 8% denaturing polyacrylamide gels in TBE buffer (89 mM Tris borate (pH 8.2), 2 mM EDTA) at 60 watts for 4 -6 h. Gels were dried and exposed to Kodak X-Omat or BioMax MR film.

Effect of RuvAB on the Cleavage Specificity of RuvC-In vitro
recombination intermediates containing a Holliday junction (␣-structures) are formed by RecA-mediated strand exchange between gapped circular and homologous linear duplex DNAs ( Fig. 1). Addition of the RuvABC Holliday junction processing proteins to an ongoing strand exchange reaction results in the earlier formation of branch migration products and also the creation of resolution products (31).
Previous studies identified the degenerate sequence 5Ј-(A/ T)TT2(G/C)-3Ј as the site selected for resolution by RuvC protein (27). In the following experiments, we analyzed whether the presence of RecA and/or the RuvAB proteins affect the cleavage specificity or efficiency of RuvC. To permit this analysis, three pairs of gapped circular and linear duplex DNA molecules were used. Each substrate is designated by the enzymes employed in their construction (Fig. 2). For gDNAs (e.g. PstI/BsaI pDEA-7Z), the enzymes indicate the double digest performed to generate the linear fragment that was annealed to circular homologous ssDNA during the preparation of gapped circular DNA. For linear DNAs (e.g. PstI/ScaI pDEA-7Z), the first enzyme indicates the uniquely 3Ј-end-labeled terminus that initiates strand exchange, whereas the second identifies the digest used to remove label from the distal end. Removal of the distal label permitted unambiguous identification of the sites of cleavage without having to perform primer extension analysis. Because the distal label was removed, each linear DNA is shorter than the gDNA molecule with which it reacts. Therefore, the product of complete strand exchange is a -structure (Fig. 1).
When gapped circular PstI/BsaI pDEA-7Z DNA was reacted with homologous 32 P-labeled PstI/ScaI linear duplex DNA ( To determine the effect of RuvAB on the specificity or efficiency of cleavage at each RuvC cleavage consensus site, the resolution products of reactions similar to those shown in Fig.  3 (lanes e and f) were analyzed as a function of time by denaturing gel electrophoresis (Fig. 4A). Since the label was at the 3Ј-end that initiates strand exchange, the shortest fragments arise from cleavage near the gap, which spans 175 nt in this substrate. With RuvC alone, a series of discrete bands were observed, most of which correspond to predicted consensus cleavage products (Fig. 4A, lanes b-g, consensus sites are indicated by filled arrowheads). The intensity of these bands increased over time in a coordinated fashion (i.e. all sites appeared to be cleaved with similar kinetics), although the intensities of all cleavage products were not equal (compare, for example, the strong 270-and weak 421-nt cleavage products). When RuvAB proteins were present (lanes i-n), a similar cleavage pattern was observed, and the intensities of the products were comparable to those seen in their absence. Under both conditions, some non-consensus sites were cleaved (see band at 725 nt, indicated by open arrowhead).
We had anticipated that the presence of RuvAB might affect the intensities of the cleavage products, given their ability to promote branch migration. The observation that the relative intensities of the cleavage products were not altered when RuvAB were present suggested that the positional distribution of Holliday junctions was not changed by RuvAB-mediated branch migration when the three Ruv proteins were added simultaneously to reactions containing RecA protein. To test this proposal, RuvC was added to reactions at various times after RuvAB. If our interpretation was correct, increasing the amount of time between the two additions would allow RuvAB to branch migrate the junctions in the absence of the resolvase, and a difference in the intensity of the cleavage products would be detected. In Fig. 4B (lanes c-j), it is shown that the intensity of each cleavage product decreases as the time of incubation between the addition of RuvAB and that of RuvC increases. Since the efficiency of cleavage in each reaction was similar (data not shown), these results demonstrate that RuvAB-mediated branch migration is occurring and results in the movement of Holliday junctions beyond our region of analysis.
RuvAB Complex Targets RuvC for Cleavage-The data presented in Fig. 4 indicate that RuvAB proteins do not affect the cleavage specificity of RuvC. To analyze whether RuvAB proteins play a role in targeting RuvC to the Holliday junction, we varied the concentration of resolvase to determine whether, at subsaturating RuvC concentrations, RuvAB could stimulate resolution. In the absence of RuvAB (Fig. 5, lanes d-i), cleavage at each consensus site was modestly enhanced with increasing RuvC, although at the highest concentration of protein (640 nM), a reduction in cleavage was observed, presumably due to the aggregation of RuvC (45). In the presence of RuvAB (lanes FIG. 1. In vitro system for coupled strand exchange and Holliday junction processing. Schematic diagram indicating RecA-mediated interactions between gapped circular (gDNA) and linear duplex DNA, forming a recombination intermediate (␣-structure). Since the linear duplex is shorter in length than the gDNA, complete strand exchange by RecA results in formation of a -structure. Addition of RuvAB during strand exchange promotes bi-directional translocation of the Holliday junction, yielding either -structures or the starting substrates. Cleavage of the junction by RuvC generates unique products; depending on the orientation of cleavage, either nicked circular and gapped linear or linear dimer molecules are formed. In the experiments described in this work, the unique label (*) was located at the 3Ј-end of the linear duplex DNA that initiates strand exchange. k-p), however, we found that the intensity of cleavage products was slightly, although reproducibly, greater at lower concentrations (20 -80 nM RuvC) than was seen in their absence, and little variation in activity was observed throughout this concentration range. Enhancement of cleavage at low RuvC concentrations presumably reflects the role that RuvAB plays in (i) targeting RuvC to the junction by assembly of a RuvABC complex, and (ii) promoting branch migration that facilitates the movement of DNA through the RuvABC complex, resulting in the passage of a consensus sequence into the active site of RuvC.
Variability in RuvC-dependent Cleavage at Consensus Sites-In the reactions described above, the Ruv proteins were added 10 min after joint molecule formation was initiated by RecA. At this time, the amount of recombination intermediates (␣-structures) is maximal, and strand transfer products (structures) are hardly detectable. Despite the low amount of -structures at 10 min (Fig. 3, lane b), however, the location of the Holliday junction in these intermediates is expected to be a significant distance from the initiating gap since strand transfer by RecA is unidirectional and occurs at a rate of approximately 3 bp/s (46). Nevertheless, in the experiments shown in Figs. 4, A and B, and 5, one particular site, just 270 nt from the initiating end of the linear duplex (i.e. 95 nt from the 5Ј-side of the gap), appeared to be cleaved very efficiently, particularly when compared with the next closest consensus sites located 421, 470, and 478 nt from the 3Ј-end of the linear DNA. Given the assumption noted above, this observation was unexpected since the region of DNA close to the gap would be expected to have only a low concentration of Holliday junctions.
To investigate the nature of this strong cleavage product, we first determined whether it was derived from cleavage at a consensus site, since it was possible that this site might define a novel sequence with enhanced specificity for RuvC. Comparison with a Maxam-Gilbert sequencing ladder, however, confirmed that cleavage occurred at a known consensus sequence, 5Ј-TTT2C-3Ј (data not shown). Since this site was the first correctly oriented consensus sequence beyond the gap, we next determined whether the location of this site near the gap might be responsible for its enhanced reactivity. This was addressed by constructing a gDNA molecule (XmnI/ScaI pDEA-7Z; Fig. 2 derived from the same phagemid, in which the distance between the gap and the consensus site was increased, and in which there would be seven intervening consensus sites of the correct polarity. Reactions were performed as described above using this gDNA and uniquely 3Ј-end-labeled XmnI/SphI linear duplex pDEA-7Z DNA (Fig. 2), and the products were analyzed on both 5 and 8% denaturing polyacrylamide gels (Fig. 6). We anticipated that if this site, or its sequence context, was responsible for its apparently heightened reactivity, the intensity of this cleavage product (now having a length of 624 nt) would remain greater than that of nearby consensus sites. However, we now found that the intensity of this band (Fig. 6, circled) was no greater than that of other nearby cleavage products (Fig. 6, lanes e and f). Instead, we observed that bands of 134 and 167 nt (15 and 48 nt, respectively, from the end of the gap) were now enhanced. These results show that proximity to the gap, rather than the sequence itself, affects reactivity toward RuvC.
As a final test of this hypothesis, we ruled out the possibility that the results were substrate-specific. To do this, a third gDNA (BamHI/NcoI pPB4.3) and homologous linear duplex fragment (BamHI/ScaI pPB4.3) from a different phagemid (Fig. 2) were constructed. In this gDNA, the location of the gap was chosen to provide many consensus sites within the first 200 bp of strand exchange, so that the distance over which this putative positional effect was exerted might be determined. Contrary to our expectation, the efficiency of cleavage at six consensus sites within 250 bp of the gap was similar (Fig. 6,  lanes g-i). Taken together, these results show that not all consensus sites are cleaved equally well by RuvC, but the basis of this variation, whether context-, structure-, or sequence-dependent, remains to be determined.

RuvAB-mediated Branch Migration Is More Efficient in the
Absence of RecA-The experiments described above were conducted in the presence of RecA-promoted strand exchange. To determine whether RecA had an effect on cleavage, recombination intermediates were formed in vitro, deproteinized by incubation with SDS and proteinase K, and purified by gel filtration chromatography. The protein-free intermediates were then incubated with increasing concentrations of RuvC in the absence or presence of RuvAB (Fig. 7, lanes b-g); for comparison, parallel reactions were conducted with ongoing strand exchange, as in previous experiments (lanes i-n). Several differences were immediately apparent. First, cleavage by RuvC alone was more efficient in the absence of RecA, as evidenced by the lower concentration of RuvC required to give similar amounts of cleavage (compare lanes b-d and i-k, and data not shown). Second, the cleavage pattern differed according to the presence or absence of RecA. Third, in the absence of RecA, the presence of RuvAB resulted in a significant reduction in the cleavage of observable consensus sites (i.e. those located between 400 and 900 nt from the gap; compare lanes b-d and e-g). Since this result was not seen in the presence of RecA (compare lanes i-k and l-n), we conclude that RuvAB-mediated branch migration occurs more efficiently after RecA removal. Four-and Three-stranded Intermediates Display Different Patterns of Cleavage-The intermediates used above can be described as "four-stranded" since they involve interactions between two duplex molecules. However, strand exchange by RecA can also occur between duplex linear and single-stranded circular DNA (47). Previous studies explored the possibility that RuvC could cleave three-stranded intermediates and concluded that they could be efficiently cut only in the absence of RecA (45,48). We therefore wanted to determine whether RuvAB would affect the ability of RuvC to cleave three-way junctions in reactions containing RecA. Because SSB protein greatly enhances pairing and strand exchange in the three-stranded reaction (49), reactions were performed in both the presence and absence of SSB. By using denaturing gel electrophoresis, a more sensitive assay than that used previously (45,48), we found that RuvC was capable of cleaving three-stranded intermediates in the presence of RecA, although cleavage was less efficient than that observed in four-strand reactions (Fig. 8, compare lanes e and f with h  and i). Whereas the cleavage of most sites was reduced in the three-strand reaction, one or two consensus sites (at 527/531 nt) and a non-consensus site (at ϳ725 nt) were cleaved equally well in either four-or three-stranded intermediates. With the exception of the 527/531-nt cleavage product(s), the presence of SSB reduced cleavage further (lanes k and l). RuvAB-mediated branch migration appeared to have only a slight effect in reducing the intensity of the fragments. Similar reactions carried out in the absence of RecA showed that three-stranded intermediates were again cleaved less efficiently than those containing four strands (data not shown).
Formation and Resolution of Reverse Polarity Intermedi-ates-The four-strand in vitro system used in this paper was originally developed as a model system for the recombinational repair of post-replication gaps occurring opposite DNA lesions (50, 51). Previous studies showed that RecA-mediated reactions between gapped and linear duplex DNA only take place when the 3Ј terminus of the complementary strand of the linear duplex lies opposite the gap (51). Due to the polarity of RecAmediated strand exchange, reactions fail to go to completion when the linear duplex contains a 5Ј-ended complementary strand. This directionality, imposed by RecA, has been incorporated into models for post-replication repair in E. coli (52,53). The ability of RuvAB to promote bi-directional branch migration, however, raises the question whether these proteins might be capable of recognizing transient (i.e. "wrong polarity") joints made by RecA. A positive result would have important implications for the mechanism of post-replication repair since it would remove any polarity limitation on strand exchange away from the gap. The substrates were those used in Fig. 4, and the reaction volume was 50 l. Varying amounts of RuvC (0, 20, 40, 80, 160, 320, or 640 nM), without (lanes d-i) or with (lanes k-p) 160 nM RuvA and 240 nM RuvB, were added 10 min after the linear DNA, and the reaction was incubated a further 20 min. Products were separated on a 5% denaturing polyacrylamide gel. Lane a, no gDNA or Ruv proteins added; lane b, as lane a except that 160 nM RuvA, 240 nM RuvB, and 320 nM RuvC were included. Lane M, size markers; filled arrowheads, location of consensus cleavage sites.
FIG. 6. Effect of location of the efficiency of cleavage at consensus sequences. Three sets of gDNA/linear duplex substrate pairs were used (Fig. 2). Reactions were performed as described in Fig. 3  In the following experiment we tested this proposal by asking whether RuvAB could recognize 5Ј termini RecA-initiated joints and promote branch migration leading to the formation of complete strand transfer products. To do this BsaI-linearized, 5Ј-end-labeled pDEA-7Z was reacted with PstI/BsaI pDEA-7Z gDNA. With this linear duplex DNA, pairing initiates at the opposite end of the gap to that used in the standard reaction, such that RecA-mediated strand transfer is only permitted in the "reverse" direction. In the presence of RecA and SSB proteins, we observed the formation of recombination intermediates (Fig. 9, lanes a-f). Compared with the standard reaction (see Fig. 3), however, the reaction was of low efficiency, as observed previously (51), and few complete strand exchange products were observed (ϳ1% of the 32 P label was found in the position of nicked circular DNA). In addition, the kinetic profile of this reaction is significantly slower than the standard profile of a 3Ј-initiated reaction.
The addition of RuvAB to a similar reaction resulted in a 10-fold stimulation in the formation of 32 P-labeled nicked circular DNA (Fig. 9, lanes g-l). When RuvC was added to the reverse polarity reaction in the absence of RuvAB, we observed the formation of Holliday junction resolution products ( 32 Plabeled linear dimer DNA; Fig. 9, lanes m-r). In the presence of all five proteins (RecA, RuvABC, and SSB), both branch migration and resolution products were observed (lanes s-x).

DISCUSSION
In these experiments, we have utilized a system in which the translocation and cleavage properties of the RuvABC Holliday junction processing complex could be examined during ongoing strand exchange promoted by RecA protein. These studies expand upon previous work examining the ability of RuvC alone to cleave RecA-bound or protein-free recombination intermediates (27,45,48) and extend our earlier findings (31) by determining whether the presence of other proteins (i.e. RecA and RuvAB) influences the cleavage activity of RuvC at specific sites. We observed that RuvC cleaved Holliday junctions at specific sites and that these were essentially the same in the absence or presence of RuvAB. Although the RuvAB complex readily branch migrates Holliday junctions present within protein-free recombination intermediates, we and others (22,31) have noted that the presence of a RecA filament encasing the interacting DNA molecules slows branch migration. We anticipated that if significant RuvAB-mediated branch migration were to occur even in the presence of RecA, the distribution of junctions in the recombination intermediates would be altered, and the intensity of specific cleavage products would change in response. However, we found no difference in the intensity of cleavage products when RuvABC were added simultaneously (Fig. 4A). From this result, we conclude that branch migration by RuvAB on RecA-bound intermediates is hampered (although it still proceeds more rapidly than strand transfer catalyzed by RecA alone; Fig. 3). This finding suggests that the incorporation of RuvC into a RuvAB complex is fast relative to the rate of translocation of the junction, for only by separating the time of addition of RuvAB and RuvC proteins was it possible to detect a low level of branch migration, as measured by a decrease in the intensity of cleavage products (Fig. 4B). By targeting RuvC to the Holliday junction, cleavage was stimulated by RuvAB at low concentrations of RuvC (Fig. 5).
Our work also shows that RecA influences the cleavage of junctions by RuvC (Fig. 7). This appears to occur indirectly in two ways. First, not only does RecA affect the ability of RuvAB to translocate junctions, but it also influences the amount of RuvC required to effect cleavage. For example, deproteinized recombination intermediates were more readily cleaved at lower concentrations of RuvC. Second, since we observed that consensus sites farther from the gap were cleaved more efficiently on deproteinized intermediates, than on RecA-coated molecules, we conclude that RuvAB-mediated branch migration is responsible for this movement. Indeed, the analysis of deproteinized intermediates showed a dramatic effect of RuvAB addition in reducing the levels of many cleavage prod-ucts, suggesting that RuvAB-promoted branch migration is much more rapid in the absence of RecA. The overall effect of RecA, therefore, is to reduce the rate of branch migration such that RuvC is incorporated in a RuvAB complex near the initial site of RuvAB binding, resulting in the junction being cleaved without the formation of a significant length of heteroduplex DNA. The fact that some sites are cleaved equally well in the presence or absence of RuvAB indicates that there may be two populations of Holliday junctions, but the nature of this difference remains obscure.
The finding of prominent RuvC-dependent cleavage products that do not correspond to consensus sites was surprising. Although it is known that substitution of Mn 2ϩ for Mg 2ϩ (54) or mutation of certain residues in RuvC (30) causes relaxation of the specificity of cleavage, it does not appear to be the case that interaction of RuvC with RuvAB is mimicking this effect, since these sites were also cleaved in the absence of the branch migration proteins (e.g. Fig. 8). Furthermore, we found that the consensus sites themselves were not cleaved to the same extent. For example, we noted that one (270 nt) cleavage product displayed enhanced intensity (Fig. 5). The reason why this site, which corresponds to a consensus site, was intrinsically more reactive with RuvC remains unknown. In experiments in which the site was moved away from the initiating gap, we observed that reactivity with RuvC was reduced to normal levels. The possibility that the proximity of the site to the gap placed it in a region where the junction was more accessible to RuvC was not fully supported with other substrates containing consensus sequences. Thus, it remains an empirical observation that some consensus sites, when located in certain regions, are better targets for cleavage by RuvC. One possibility to be considered is that structural and energetic constraints influence the positioning of Holliday junctions in a non-random fashion along the DNA. These factors may cause the complex to stall at certain (non-consensus) sequences, with cleavage subsequently occurring in a sequence-independent manner.
In the experiments described here, we also compared the cleavage of junctions using three-or four-stranded DNA substrates. In three-strand reactions, we observed that cleavage at many consensus sites was eliminated, and this effect was exacerbated by the presence of RuvAB (Fig. 8). Under the same conditions, however, cleavage at one particular site occurred in three-or four-stranded reactions with equal efficiency. Why cleavage at some sites is reduced yet others can be cleaved efficiently remains a puzzle.
As a corollary to these studies, we examined the ability of RuvAB to promote strand transfer on intermediates formed with a polarity bias that is opposite to that preferred by RecA. With RecA alone, interactions with such substrates are inefficient, and strand exchange is limited (51). However, we found that the presence of RuvAB removed the polarity bias from RecA, resulting in the formation of products that had undergone complete strand exchange (Fig. 9).
The results presented in this work are incorporated into a revised model for the recombinational repair of post-replication gaps formed opposite DNA lesions. As shown in Fig. 10, singlestrand gap formation occurs when a replication fork is impeded by the presence of a pyrimidine dimer or other photo-induced lesion (A). Joint formation with the sister duplex (B) leads to gap filling by RecA-mediated strand transfer (C). This reaction is directional, with RecA-mediated transfer occurring with a 3Ј-5Ј polarity relative to the strand that fills the gap. The nuclease that catalyzes nick formation has not as yet been identified although in vivo data with damaged DNA templates suggest its existence (55,56). In the next step, unidirectional RecA-mediated strand transfer leads to the formation of a Holliday junction as the fourth DNA strand is encountered (D). DNA exchange beyond the lesion then provides a template strand that enables UvrABC-mediated excision repair, followed by gap filling (D and E). Finally, Holliday junctions are recognized by the RuvABC complex and are driven along the DNA by branch migration. This reaction can occur in either direction until a consensus site for RuvC is encountered (E; alternative consensus sites are indicated by hatched boxes). Finally, the repaired molecules are resolved by the RuvABC complex, and the nicks are sealed by DNA ligase (F). The data presented in this paper indicate that junctions can be cleaved at RuvC consensus sequences that are located close to the initiating single-strand gap. Our results suggest that only limited branch migration occurs in the presence of the RecA filament and that RuvC is incorporated into a RuvAB complex rapidly. Consequently, resolution is likely to occur at the first consensus site encountered during branch migration. The Ruv-ABC complex therefore makes repair a conservative process that is focused immediately to the site of damage, without significant heteroduplex formation in tracts of undamaged DNA.