RuvA is a Sliding Collar that Protects Holliday Junctions from Unwinding while Promoting Branch Migration

doi:10.1016/j.jmb.2005.10.075

J. Mol. Biol. (2006) 355, 473–490

RuvA is a Sliding Collar that Protects Holliday Junctions from Unwinding while Promoting Branch Migration Daniel L. Kaplan1* and Mike O’Donnell1,2 1

Rockefeller University, Laboratory of DNA Replication New York, NY 10021, USA 2

Howard Hughes Medical Institute, Laboratory of DNA Replication, New York, NY 10021, USA

The RuvAB proteins catalyze branch migration of Holliday junctions during DNA recombination in Escherichia coli. RuvA binds tightly to the Holliday junction, and then recruits two RuvB pumps to power branch migration. Previous investigations have studied RuvA in conjunction with its cellular partner RuvB. The replication fork helicase DnaB catalyzes branch migration like RuvB but, unlike RuvB, is not dependent on RuvA for activity. In this study, we specifically analyze the function of RuvA by studying RuvA in conjunction with DnaB, a DNA pump that does not work with RuvA in the cell. Thus, we use DnaB as a tool to dissect RuvA function from RuvB. We find that RuvA does not inhibit DnaB-catalyzed branch migration of a homologous junction, even at high concentrations of RuvA. Hence, specific protein–protein interaction is not required for RuvA mobilization during branch migration, in contrast to previous proposals. However, low concentrations of RuvA block DnaB unwinding at a Holliday junction. RuvA even blocks DnaB-catalyzed unwinding when two DnaB rings are acting in concert on opposite sides of the junction. These findings indicate that RuvA is intrinsically mobile at a Holliday junction when the DNA is undergoing branch migration, but RuvA is immobile at the same junction during DNA unwinding. We present evidence that suggests that RuvA can slide along a Holliday junction structure during DnaB-catalyzed branch migration, but not during unwinding. Thus, RuvA may act as a sliding collar at Holliday junctions, promoting DNA branch migration activity while blocking other DNA remodeling activities. Finally, we show that RuvA is less mobile at a heterologous junction compared to a homologous junction, as two opposing DnaB pumps are required to mobilize RuvA over heterologous DNA. q 2005 Elsevier Ltd. All rights reserved.

*Corresponding author

Keywords: DNA replication; DNA recombination; Holliday junction; branch migration; RuvA

Introduction DNA recombination functions in Escherichia coli to repair damaged DNA and rescue stalled replication forks.1 During recombination, a DNA strand is paired with its homolog from a different duplex in a reaction catalyzed by RecA working in concert with other proteins.1 This creates a fourway DNA structure, also called a Holliday junction, Present address: D. L. Kaplan, Vanderbilt University, Department of Biological Sciences, Nashville, TN 37232, USA. Abbreviation used: ssDNA, single-stranded DNA. E-mail address of the corresponding author: [email protected]

which is processed by the RuvABC proteins.2 The RuvA protein initially binds the Holliday junction, and then recruits RuvB protein rings to opposite sides of the junction.3,4 RuvB is a molecular motor that uses the energy derived from ATP binding and hydrolysis to drive unidirectional movement of the four-way junction.5 The RuvAB complex then recruits RuvC to the junction.6 RuvC is a nuclease that cleaves the Holliday junction, thereby resolving it into two duplex DNAs.7 RuvAB has been studied by biochemical and structural techniques. RuvA binds as a tetramer or octamer to a Holliday junction. The RuvA protein has acidic pins that inhibit binding to doublestranded DNA, thereby targeting the protein to Holliday junctions.8–11 RuvB is a hexameric ring

0022-2836/$ - see front matter q 2005 Elsevier Ltd. All rights reserved.

474 protein with a central channel wide enough to encircle double-stranded DNA.11–13 RuvB alone does not bind Holliday junction DNA under physiological conditions, but after RuvA binds the Holliday junction, RuvA facilitates the assembly of two RuvB rings onto opposite sides of the RuvAjunction (Figure 4(a)).14–16 The two RuvB rings are thought to function in concert as DNA pumps that drive branch migration of the Holliday junction.15 DnaB functions in DNA replication and is a member of the class of ring-shaped helicases.17–19 DnaB is the primary replicative helicase of E. coli, and unwinds the parental duplex to provide singlestranded DNA (ssDNA) for the replicative polymerases.20 The DnaB hexamer encircles ssDNA while translocating along it, pumping the strand through the central channel.21–24 Upon encountering a forked duplex structure, the second DNA strand cannot fit into the central channel of DnaB, and therefore the continued advance of DnaB along the original strand forces its separation from the second DNA strand.24,25 It is thought that DnaB, like other hexameric helicases, may act at the replication fork to unwind parental DNA in this manner.26–28 The DnaB hexamer can also operate in a mode in which it encircles both strands of duplex DNA. In this mode, the DnaB does not unwind the DNA.24 However, DnaB actively translocates along the duplex while encircling two DNA strands as it powers branch migration of Holliday junctions.29 Although this reaction is very efficient in vitro, the in vivo role for DnaB-catalyzed branch migration is unclear. In summary, DnaB unwinds DNA when encircling one DNA strand, and drives DNA branch migration while encircling two DNA strands. DnaB and RuvB have several mechanistic features in common. For example, DnaB and RuvB are both hexameric ring proteins that encircle DNA.2 Furthermore, both DnaB and RuvB utilize ATP binding and hydrolysis to unwind DNA with 5 0 to 3 0 polarity.24,30–32 DnaB and RuvB also displace proteins bound to DNA,29,33 and they both drive branch migration of Holliday junctions.15,29 Finally, two DnaB rings can bind to opposite sides of a Holliday junction, and work in concert to drive branch migration of an extended heterologous junction, like RuvB.34–36 Biochemical studies of RuvA in the past have been performed with RuvA in conjunction with its cellular partner, RuvB. The biochemical action of the RuvAB complex is thus well studied. However, some intrinsic properties of RuvA during its action are unclear, as it is most often studied in conjunction with RuvB. This leaves unanswered a number of questions of how RuvA functions. For example, how does RuvA bind tightly to Holliday junctions, yet become activated to move during branch migration? Previous proposals suggest that RuvB must mobilize RuvA bound to a Holliday junction, and that this action is mediated via specific protein– protein interaction between RuvB and RuvA.10 However, it is difficult to study the process that

RuvA Promotes Branch Migration

mobilizes RuvA using only RuvA and RuvB, since the activities of these two proteins are dependent upon one another.2 Unlike RuvB, DnaB does not require RuvA for activation. Thus, we investigated how RuvA moves at a Holliday junction by using DnaB in conjunction with RuvA. In this study, DnaB is used as a tool to study how RuvA functions, since these two proteins do not function together in vivo. There is an additional advantage to using DnaB with RuvA to study RuvA function. RuvB rings bind to opposite arms of a RuvA-bound Holliday junction in either of two orientations (Figure 4(a)). Thus, it is difficult to target RuvB loading to a particular junction arm. In contrast, DnaB loads onto junction-arm DNA only if a 5 0 single-strand extension (5 0 tail) is attached to a particular junction arm. Thus, unlike RuvB, one DnaB hexamer can be loaded onto a particular junction arm by specifically adding a 5 0 loading tail. We find that RuvA binds tightly to Holliday junction DNA and blocks DnaB-catalyzed unwinding of a Holliday junction. Unwinding activity is blocked even when two DnaB hexamers act in concert. However, RuvA does not block DnaBcatalyzed branch migration of a homologous Holliday junction. Hence, RuvA does not need specific protein activation by RuvB to mobilize in the direction of branch migration. We present evidence that suggests that RuvA slides along the Holliday junction during DNA branch migration, but not during DNA unwinding. Interestingly, two DnaB pumps are needed to power migration of RuvA over heterologous DNA, an action that fits nicely with the physiological architecture of two RuvB pumps straddling the RuvA protein at a Holliday junction.

Results Homologous and heterologous Holliday junction DNA substrates used in this study Holliday junctions in the cell are usually homologous, as RecA normally pairs DNA strands of the same sequence to create the junction. By homologous, we mean that the DNA arms contain complementary sequences before and after branch migration. Figure 1(a) shows a homologous Holliday junction substrate used in this study. Note that the substrate is not completely homologous, otherwise the Holliday junction is unstable and can migrate spontaneously. Thus, the homologous substrate used here bears a small degree of heterology (5 bp) to stabilize the structure and render it amenable to experimentation. In the reaction shown in Figure 1(a), the Holliday junction branch-point migrates a distance of 45 bp, of which 40 are complementary, while five are non-complementary. Heterologous junctions contain DNA sequence that is non-complementary, and the DNA arm will therefore become unpaired after branch migration

RuvA Promotes Branch Migration

475

Figure 1. Homologous and heterologous Holliday junction DNA substrates used in this study. The substrates shown are used in this study to assess RuvA mobility at (a) a homologous and (b) heterologous Holliday junction. (a) This homologous substrate is used in Figure 2(a). The 1-2 and 3-4 duplexes are 45 bp in length, and the 1-4 and 2-3 duplexes are 25 bp in length. The 5 0 tail is composed of 30 dT residues. Oligonucleotides used to form this substrate are provided in Table 1. In the reaction shown, the branch-point migrates for 45 bp, of which 40 bp are complementary, and 5 bp are non-complementary. (b) This heterologous substrate is used in Figure 6(a). The 1-2 and 3-4 duplexes are 25 bp in length, and the 1-4 and 2-3 duplexes are 25 bp in length. The 5 0 tail is composed of 30 dT residues. In the reaction shown, the branch-point migrates over 19 non-complementary base-pairs.

over the heterologous region. Heterologous Holliday junctions may be found in the cell during episodes of DNA damage, when the integrity and sequence fidelity of the DNA are compromised. Thus, the cell may contain heterologous Holliday junctions under conditions of DNA damage. In this study, the mobility of RuvA at homologous and heterologous DNA structures are investigated, since either may occur in vivo. Figure 1(b) shows a heterologous junction substrate used in this study. For this substrate, 19 bp of contiguous, non-complementary DNA sequence are present in the DNA products after branch migration. RuvA inhibits DnaB-catalyzed unwinding, but not DnaB-catalyzed branch migration, of a homologous junction Here, we use DnaB as a tool to study how RuvA functions at a Holliday junction. In the experiment illustrated by Figure 2(a), DnaB was incubated with a long homologous Holliday junction with a single 5 0 loading tail for DnaB (also illustrated in Figure 1(a)). We have shown that with only a 5 0 tail and no 3 0 tail, DnaB tracks on the ssDNA tail and then, instead of unwinding, it travels onto the duplex by encircling both strands and catalyzes branch migration.35 DnaB was incubated with the

homologous junction for 0.5 min to 4 min at 37 8C. The homologous junction is composed of strands 1, 2, 3, and 4. Strand 1 is radiolabeled (asterisk, strand *1). Reactions were quenched and products were resolved on a native gel, which separates the larger junction substrate from the smaller branch migration product that migrates faster. DnaB promotes substantial branch migration of this substrate, as expected (Figure 1(a), accumulation of *1-4 duplex). The model above each gel is used to orient the reader, and was determined from the experimental gel evidence below. The arrows next to the gel show the migration distance of DNA products as determined by radiolabeled DNA markers that were analyzed in the same gel in the experiments described here. For clarity, however, markers are cropped from the gel images. In Figure 2(b), the substrate was first incubated with RuvA for 1 min, followed by addition of DnaB as in Figure 2(a). Nearly identical results were obtained in the absence or in the presence of RuvA (compare Figure 2(a) with (b), and view quantification of product accumulation in Figure 2(c)). Therefore, RuvA does not inhibit DnaB-catalyzed branch migration of a homologous junction. This result suggests that RuvA is intrinsically mobile on a homologous Holliday junction during DNA branch migration, and does not require activation

476

RuvA Promotes Branch Migration

Figure 2. RuvA inhibits DnaB-catalyzed unwinding, but not DnaB-catalyzed branch migration, of a homologous junction. The model above each gel is used to orient the reader, and was determined from the experimental gel evidence below. (a) DnaB acts on a long homologous Holliday junction with one 5 0 tail. The 1-2 and 3-4 duplexes are 45 bp in length, and the 1-4 and 2-3 duplexes are 25 bp in length. The 5 0 tail is composed of 30 dT residues. Oligonucleotides used to form this substrate are provided in Table 1. DnaB was incubated with the junction illustrated for the time indicated. Native gel analysis of the reaction is shown. The arrows next to the gel show the migration distance of DNA products as determined by radiolabeled DNA markers electrophoresed in the same gel. Markers are cropped from the gel images for clarity. (b) Same as (a), except the substrate is pre-incubated with RuvA for 1 min prior to adding DnaB. (c) Quantification of the accumulation of branch migration product (*1, 4) from (a) and (a). (d) DnaB acts on a long homologous Holliday junction with one fork. DnaB was incubated with the junction DNA illustrated for the time indicated. The substrate is identical with (a), except strand 4 contains a 27 bp double-stranded 3 0 tail. Native gel analysis of the reaction is shown. (e) Same as (d), except the substrate is pre-incubated with RuvA for 1 min prior to adding DnaB. (f) Quantification of the accumulation of unwinding products (*2, 3, 4 and *2) from (a) and (b).

through specific protein–protein contacts with RuvB. The previous experiment demonstrates that RuvA does not inhibit DnaB-catalyzed branch migration of a homologous Holliday junction. Does RuvA inhibit other DNA-modulating reactions at a Holliday junction? To test if RuvA inhibits other DNA remodeling activities, RuvA and DnaB were incubated with a homologous Holliday junction bearing a forked structure at one end (Figure 2(d)). We have shown that DnaB will unwind a Holliday junction if the loading site is

forked.29 DnaB loads onto the 5 0 tail of this forked substrate. The 3 0 tail of the fork is double-stranded, which promotes DnaB-catalyzed unwinding by aiding exclusion of this 3 0 tail from the central channel of DnaB.24 DnaB first unwinds strand 1 of this substrate to produce an intermediate product composed of strand *2, strand 3, and the 4 duplex (see reaction (i) in Figure 2(d), and the gel band at early time-points that corresponds to this intermediate product). Later in the reaction, DnaB unwinds strand *2 of this intermediate product to yield free strand *2 (see reaction (ii) in Figure 2(d),

RuvA Promotes Branch Migration

and the gel band at later time-points corresponding to free strand *2). (Strand 2 of this substrate is labeled instead of strand 1 to definitively distinguish unwinding from branch migration. If strand 1 were labeled here, DnaB could unwind the 1-4 product to yield free strand 1, thereby making it difficult to distinguish branch migration from unwinding. The DnaB enzyme is not fully trapped in these reactions, and it may react with the primary products generated to produce secondary reaction products.) This experiment was then repeated, but the substrate was first incubated with RuvA (Figure 2(e)). The result shows that RuvA inhibits the accumulation of unwinding products substantially (compare Figure 2(d) with (e) and see the quantification of unwinding product accumulation in Figure 2(f)). Thus, RuvA inhibits DnaB-catalyzed unwinding of this homologous Holliday junction substrate. This observation suggests that RuvA blocks DNA unwinding at a Holliday junction. In Figure 3, DnaB was incubated with a Holliday junction that is similar to that of Figure 2(d), except the 3 0 tail is single-stranded instead of doublestranded. With a single-stranded 3 0 tail, DnaB now has a similar probability of catalyzing unwinding or

477 branch migration of this substrate, as observed previously.29 RuvA blocks DnaB-catalyzed unwinding of this substrate, but it does not inhibit DnaB-catalyzed branch migration of this substrate. Thus, RuvA inhibits DnaB unwinding activity, but not branch migration, at a homologous junction. These results suggest that RuvA is free to move at a Holliday junction during DNA branch migration, while RuvA is immobile at the same Holliday junction during DNA unwinding. We next performed several control experiments to test if our RuvAB proteins behave as described by other laboratories. RuvA binds to Holliday junctions, and then recruits two RuvB hexameric rings to opposite sides of RuvA. The RuvAB complex then pumps DNA to drive branch migration. To test if our RuvA and RuvB proteins are able to drive branch migration, as others have reported, the two proteins were incubated with a radiolabeled heterologous Holliday junction (Figure 4(a)). Branch migration of this heterologous junction substrate will result in two smaller products (Figure 4(a)). RuvA was incubated with the heterologous junction for 1 min at 37 8C, followed by RuvB for 0.5 min to 4 min. RuvB rings usually encircle the *1-2 and 3-4 duplexes to drive branch migration of this

Figure 3. RuvA inhibits DnaBcatalyzed unwinding, but not DnaB-catalyzed branch migration, of a homologous junction at a single DNA substrate. The model above each gel is used to orient the reader, and was determined from the experimental gel evidence below. (a) DnaB acts on a long homologous Holliday junction with one fork. The 1-2 and 3-4 duplexes are 45 bp in length, and the 1-4 and 2-3 duplexes are 25 bp in length. Each 5 0 tail is composed of 30 dT, and each 3 0 tail is composed of 30 dT. Oligonucleotides used to form this substrate are provided in Table 1. DnaB was incubated with the junction illustrated for the time indicated. Native gel analysis of the reaction is shown. The arrows next to the gel show the migration distance of DNA products as determined by radiolabeled DNA markers electrophoresed in the same gel. Markers are cropped from the gel images for clarity. (b) Same as (a), except the substrate is pre-incubated with RuvA for 1 min prior to adding DnaB.

478

RuvA Promotes Branch Migration

Figure 4. RuvA does not recruit DnaB to Holliday junctions. The model above each gel is used to orient the reader, and was determined from the experimental gel evidence below. (a) RuvAB activity on a heterologous Holliday junction. The numbers represent names for each DNA strand. Strand 1 is labeled with 32P at the 5 0 terminus (asterisk). Duplexes 1-2 and 3-4 are 45 bp in length, and duplexes 1-4 and 2-3 are 25 bp in length. Oligonucleotides used to form this substrate are provided in Table 1. RuvA was incubated with the heterologous junction illustrated for 1 min, followed by incubation with RuvB for the periods of time indicated. Native gel analysis of the reaction products is shown. The arrows next to the gel show the migration distance of DNA products as determined by radiolabeled DNA markers electrophoresed in the same gel. Markers are cropped from the gel images for clarity. (b) Same as (a), except RuvA is omitted from the reaction. (c) DnaB acts on the heterologous Holliday junction illustrated in (a). DnaB was incubated with the heterologous junction illustrated for the time periods indicated. Native gel analysis of the reaction is shown. (d) Same as (c), except the substrate is pre-incubated with RuvA for 1 min prior to adding DnaB.

substrate, producing the *1-2 product. This reaction is shown in scheme (i) of Figure 4(a), and confirmed by the dark product band in the gel. Less frequently, RuvB encircles the *1-4 and 2-3 duplexes to drive branch migration, producing the radiolabeled *1-4 duplex product (scheme (ii) of Figure 4(a), and faint branch migration product in the gel). Reaction (i) of Figure 4(a) occurs far more frequently than reaction (ii). The reaction preference correlates with the length of the duplex arms. This heterologous junction has arm lengths of 45 bp for duplexes 1-2 and 3-4, and 25 bp for duplexes 1-4 and 2-3. The binding site size for RuvAB is approximately 35 bp on either side of the junction.15 Thus, there is likely competition between these two reactions, and the longer duplex arms act as better binding surfaces for the RuvB rings. In the absence of RuvA, RuvB alone produces very little product (Figure 4(b)). RuvA alone does not produce any product (not shown), and the substrate does not disassemble spontaneously in the absence of

protein (not shown). In summary, the RuvAB proteins used in this study drive branch migration of Holliday junctions as expected from the work of other laboratories.8,15,34,37 Next, we performed an additional control to ensure that RuvA does not influence DnaB loading onto a Holliday junction. We do not expect RuvA to load DnaB, as there is no genetic evidence to suggest that RuvA recruits DnaB. However, we performed the following control experiment just to be sure. DnaB was incubated with the Holliday junction lacking a 5 0 tail for 0.5 min to 4 min (Figure 4(c)). Very little product accumulation is observed (Figure 4(c)), as expected for a substrate with no 5 0 tail. To determine if RuvA recruits DnaB to a Holliday junction, RuvA was incubated with the junction DNA for 1 min at 37 8C, then DnaB was added to the reaction for 0.5 min to 4 min (Figure 4(d)). There is no conversion of substrate to product during this time-frame (Figure 4(d)). Thus, RuvA does not load DnaB at a Holliday

RuvA Promotes Branch Migration

junction, as expected for this control. These results allow one to conclude that in Figures 2 and 3, DnaB self-loads on the 5 0 single-stranded region of the substrate in a manner independent of the presence of RuvA. How does RuvA inhibit DnaB unwinding activity at a Holliday junction in Figures 2 and 3? One explanation is that RuvA binds tightly to the junction substrate, and creates a physical block to DnaB movement. In support of this idea, it has been demonstrated that RuvA binds tightly to Holliday junctions.16 To test if RuvA binds to junction DNA under our reaction conditions, an electrophoretic mobility-shift assay (EMSA) was performed. Radiolabeled junction DNA was incubated with RuvA under conditions that are identical with those used in our branch migration and unwinding assays (Figure 5(a)). The sample was then analyzed by electrophoreses in a native polyacrylamide gel to separate the RuvA–DNA complex from free DNA. As the concentration of RuvA is increased, bands corresponding to RuvA–DNA complexes increase in intensity (Figure 5(b)). (The appearance of several RuvA–junction DNA complexes is consistent with

Figure 5. RuvA binds tightly to Holliday junctions. Electrophoretic mobility-shift analysis of RuvA binding a Holliday junction. (a) Scheme for RuvA binding a heterologous junction. The duplex arm lengths are each 25 bp. (b) After RuvA incubation for 1 min with the radiolabeled junction illustrated in (a), the reaction is analyzed by native gel electrophoresis. (c) The data from (b) are quantified and plotted.

479 published studies, and may correspond to tetrameric RuvA, octameric RuvA, and higher-order species.16) Quantification of the results of this gel is shown in Figure 5(c). RuvA binds half the junction DNA at a protein concentration of approximately 10 nM. Thus, RuvA binds tightly to Holliday junction DNA under our reaction conditions, and this provides a simple explanation for how RuvA inhibits DnaB unwinding activity. Tight binding of RuvA to Holliday junctions and the resulting inhibition of unwinding by DnaB raise an interesting question. Why does RuvA not block DnaB branch migration activity? One possible explanation is that DnaB readily dislodges, or disassembles, the RuvA multimer from the junction DNA during branch migration but not unwinding. If this is the case, then DnaB must be far more active in dislodging RuvA when DnaB encircles two DNA strands during branch migration compared to when DnaB encircles one strand during unwinding. However, previous studies make this hypothesis unlikely. The rate of DnaB-catalyzed protein displacement from DNA is similar for DnaB encircling two DNA strands and for DnaB encircling one DNA strand.29 In addition, DnaB is quite slow to displace tightly bound proteins from DNA relative to the rate of DnaB-catalyzed branch migration activity.29 An alternative hypothesis to account for the observation that RuvA blocks DnaB-catalyzed unwinding but not branch migration is that RuvA can slide along the Holliday junction structure during branch migration, but not unwinding (Figure 10). For DnaB to catalyze unwinding of an RuvA-bound Holliday junction, RuvA is probably displaced, since the four-way junction structure is destroyed during the reaction (Figure 10(a)). In contrast, during DnaB-catalyzed branch migration of an RuvA-homologous junction complex, the junction structure is preserved continuously (until the junction is eventually split when branch migration reaches the end of the substrate). This may enable RuvA to stay bound to DNA throughout the branch migration process (Figure 10(b)). This proposed difference in mechanism, in which RuvA slides during branch migration but is displaced directly off a junction during unwinding may explain the difference in RuvA blocking ability. In other words, RuvA binds Holliday junctions stably, and it may be difficult to dislodge this protein from an internal position, but during branch migration RuvA may slide along with the junction. When branch migration is complete and the junction is split in two, then RuvA would presumably slide off the end of the DNA. We would like to determine directly whether RuvA is displaced from DNA or slides along the junction structure during DnaB-catalyzed branch migration. However, this cannot be determined by examining whether RuvA is on or off DNA, as RuvA must fall off the DNA substrate in both scenarios, whether it is dislodged from the internal junction structure directly, or whether it slides off

480 the end only after branch migration is complete. In both cases, RuvA is bound to DNA only transiently. Furthermore, whether DnaB dislodges RuvA directly, or DnaB slides with RuvA, DnaB and RuvA will be bound to the Holliday junction at the same time when DnaB loads onto the RuvA-bound Holliday junction. Thus, simply showing that DnaB and RuvA are bound to junction DNA at the same time will not distinguish between RuvA sliding and RuvA displacement. We next present an indirect method to examine whether RuvA is displaced or slides along the junction during DNA branch migration.

RuvA Promotes Branch Migration

RuvA inhibits branch migration of a heterologous junction catalyzed by a single DnaB ring In the experiments of Figure 6 we examine RuvA and DnaB on a heterologous junction substrate to help distinguish between the sliding and direct displacement models for branch migration presented above. Studying RuvA action at a heterologous junction also has important biological relevance, since RuvA is likely to encounter regions of heterology during DNA damage. The junction substrate of Figure 6(a)–(c) is similar to that of

Figure 6. RuvA inhibits branch migration of a heterologous junction catalyzed by a single DnaB ring. The model above each gel is used to orient the reader, and was determined from the experimental gel evidence below. (a) DnaB acts on a short heterologous Holliday junction with one 5 0 tail. Each duplex is 25 bp in length, and the 5 0 tail is composed of 30 dT. Oligonucleotides used to form this substrate are provided in Table 1. DnaB was incubated with the junction illustrated for the periods of time indicated. Native gel analysis of the reaction is shown. The arrows next to the gel show the migration distance of DNA products as determined by radiolabeled DNA markers electrophoresed in the same gel. Markers are cropped from the gel images for clarity. (b) Same as (a), except the substrate is pre-incubated with RuvA for 1 min prior to adding DnaB. (c) Data from gels shown in (a) and (b) were quantified and plotted. The products that are quantified in (c) are all derived from the *1-4 branch migration product. (d) Filled triangles show data from an experiment similar to that shown in (b), except the concentration of RuvA is varied and the time of DnaB incubation is fixed at 2 min. Open circles show data from a similar experiment using the substrate from Figure 2(b) (4 min DnaB). DnaB incubation times were chosen to roughly match product accumulation in the absence of RuvA. (e) RuvA and RuvB acting on the junction illustrated in (a). RuvA was incubated with the junction illustrated for 1 min, followed by incubation with RuvB for the periods of time indicated. (f) Same as (e), except RuvA is omitted from the reaction.

RuvA Promotes Branch Migration

Figures 2(a)–(c), except the heterologous junction of Figure 6(a)–(c) will yield unannealed duplex arms once branch migration is complete, unlike the homologous junction of Figure 2(a)–(c). The core region of the junction has an identical DNA sequence for the substrates used in Figures 2–6. Thus, the DNA-binding site for RuvA is similar in all of these experiments. We have shown elsewhere that with the junction in Figure 6(a)–(c), DnaB loads onto strand 1 and travels onto the duplex by encircling strands 1 and 4. DnaB then catalyzes branch migration while encircling strands 1 and 4.29 DnaB encircles two DNA strands and pumps DNA in a similar manner in the junction of Figure 6(a)–(c) and that of Figure 2(a)–(c). Thus, the ability of DnaB to dislodge RuvA from the junction of Figure 6(a)–(c) should be the same as the ability of DnaB to dislodge RuvA from the junction of Figure 2(a)–(c). Hence, if DnaB displaces RuvA rapidly during branch migration on a homologous junction, RuvA should also have little effect on migration of a heterologous junction, since DnaB should simply knock it off the DNA and then proceed on its way. A different result maybe expected if RuvA slides during branch migration. Once branch migration starts, the DNA strands in the junction of Figure 6(a)– (c) will become non-complementary, unlike in Figure 2(a)–(c). Thus, RuvA may be expected to have difficulty sliding over this heterologous junction once branch migration starts, because the junction structure will become distorted (see Figure 10(c)). RuvA is structurally suited to bind to a four-way junction of annealed, duplex DNA.10 If the arms become non-complementary, RuvAwill not be able to optimally accommodate the unannealed duplex arms. Forcing unannealed arms through the RuvA sandwich may be energetically unfavorable. The structures of the homologous and heterologous junctions differ from each other only after branch migration has begun. Hence, if RuvA behaves differently at a heterologous junction compared to a homologous junction, then RuvA is likely recognizing the change in junction structure that arises only once branch migration starts. In other words, if RuvA blocks DnaB at a heterologous but not a homologous junction, then RuvA is likely bound to the junction continuously during branch migration (i.e. RuvA slides). Thus, study of the effect of RuvA on DnaB using the heterologous substrate of Figure 6(a)–(c) may provide indirect evidence of how RuvA acts during DnaB-catalyzed branch migration of a homologous junction (i.e. is RuvA displaced or does it slide?). First, we examined the results of DnaB action on the heterologous junction in the absence of RuvA as a control (Figure 6(a)). DnaB promotes substantial accumulation of the *1-4 branch migration product, as expected. The *1-4 duplex is a branch migration product, and not the secondary result of strand reannealing. (The half-time of strand 1 annealing with strand 4 is 70 min, due to an intramolecular hairpin within the region of strand 1 that base-pairs

481 with strand 4 and slows the rate of strand annealing.29) The additional faint bands in Figure 6(a) arise from DnaB-catalyzed unwinding of the branch migration product, followed by strand reannealing.29 The effect of RuvA on DnaB-catalyzed branch migration of the heterologous junction is shown in Figure 6(b). The result shows that the branch migration product is produced far more slowly in the presence of RuvA (compare Figure 6(a) with (b)). Quantification of all products produced in Figure 6(a) and (b) shows that RuvA substantially inhibits DnaB activity at this heterologous junction (Figure 6(c)). Thus, RuvA inhibits DnaB-catalyzed branch migration at the heterologous junction of Figure 6. The concentration-dependence of RuvA inhibition of DnaB activity on this heterologous junction was studied (Figure 6(d), filled triangles). There is substantial inhibition of DnaB-catalyzed branch migration over a wide range of concentrations of RuvA. This result contrasts markedly with a similar experiment performed with the homologous junction substrate from Figure 1(a) (Figure 6(d), open circles), where RuvA does not inhibit DnaB over the same range of concentrations. Thus, RuvA blocks DnaBcatalyzed branch migration of a heterologous junction, but not of a homologous junction. If DnaB dislodges RuvA from the homologous substrate, it should also dislodge RuvA rapidly from the heterologous substrate. However, RuvA inhibits DnaB greatly on a heterologous junction. Thus, the rapid rate of branch migration of the homologous junction in the presence of RuvA supports the model whereby RuvA slides over the homologous Holliday junction when DnaB pushes it (Figure 10(a)). As an additional control, we tested RuvAB on the heterologous junction, which should result in branch migration, as the work of other laboratories has shown.8,15,34,37 The results demonstrate that RuvAB is active on this substrate, with roughly equal accumulation of each branch migration product (Figure 6(e)). Each duplex arm of this substrate has a length of 25 bp. The roughly equal accumulation of each branch migration product correlates with the equal size of the duplex arms. Thus, the hypothesis that the length of the arms sets up a competition between the two branch migration reactions is supported by these data (i.e. results in Figure 4). There is very little activity when RuvA is excluded from the reaction (Figure 6(f)). RuvA inhibits unwinding of a Holliday junction catalyzed by a single DnaB ring We next studied the mobility of RuvA at heterologous junctions during DNA unwinding. We incubated DnaB with a heterologous junction bearing both 5 0 and 3 0 tails (a fork) (Figure 7(a)). The result shows that DnaB unwinds the strand bearing the 5 0 tail from the rest of the junction (Figure 7(a), scheme (i)). We have shown elsewhere that DnaB drives branch migration of this substrate (Figure 7(a), scheme (ii)),29 and this product is observed, as

482

RuvA Promotes Branch Migration

Figure 7. RuvA inhibits unwinding of a Holliday junction catalyzed by a single DnaB ring. The model above each gel is used to orient the reader, and was determined from the experimental gel evidence below. (a) DnaB acts on a heterologous Holliday junction with one fork. Each duplex is 25 bp in length, and the 5 0 and 3 0 tails are each composed of 30 dT. Oligonucleotides used to form this substrate are provided in Table 1. DnaB was incubated with the junction illustrated for the periods of time indicated. Native gel analysis of the reaction is shown. The arrows next to the gel show the migration distance of DNA products as determined by radiolabeled DNA markers electrophoresed in the same gel. Markers are cropped from the gel images for clarity. (b) Same as (a), except the substrate is incubated with RuvA for 1 min before adding DnaB. (c) Data from gels shown in (a) and (b) were quantified and plotted. (d) Experiments similar to those shown in (b) were performed, except the concentration of RuvA is varied and the DnaB incubation time is held constant at 2 min. Data from native gel analyses were quantified and plotted as a function of RuvA concentration (open circles). The experiment was then conducted using RuvB in place of DnaB (filled squares).

indicated in Figure 7(a). (The faint band that accumulates between these products at later timepoints corresponds to the *1-2 duplex that results from unwound strand 1 reannealing with unwound strand 2.29) To determine if RuvA inhibits DnaB-catalyzed unwinding of a heterologous junction, RuvA was incubated with the heterologous junction before adding DnaB (Figure 7(b)). Only faint products are observed for both unwinding (strand *1) and branch migration (*1-4 duplex). Therefore, in the presence of RuvA, both of these products are produced at a slower rate (compare Figure 7(a) with (b)). Quantification of all products shows that RuvA substantially inhibits DnaB activity on this forked heterologous Holliday junction (Figure 7(c)). Thus, RuvA inhibits both DnaB-catalyzed unwinding and branch migration of the heterologous junction, while RuvA activates RuvB-catalyzed branch migration on this same substrate (not shown, but similar to the result in Figure 6(e)). Next, we examined the concentration-dependence of RuvA for inhibition of DnaB, and for activation of RuvB. Experiments were performed as described above, except the concentration of RuvA was varied and the time was held constant

(Figure 7(d)). Low concentrations of RuvA activate RuvB-catalyzed branch migration (Figure 7(d), filled squares). Half-maximal stimulation of RuvB activity occurs at approximately 3 nM RuvA. Halfmaximal inhibition of single-DnaB ring activity occurs at a similar concentration of RuvA (between 3 nM and 10 nM, Figure 7(d), open circles). Thus, a similar concentration of RuvA that activates RuvB also inhibits DnaB. This is the concentration range in which RuvA binds to the Holliday junction (Figure 5). Thus, binding of RuvA to the Holliday junction at low concentrations likely accounts for activation of RuvB and inhibition of DnaB. RuvA inhibits double-DnaB ring-catalyzed unwinding, but not branch migration RuvA inhibits branch migration of a heterologous junction catalyzed by a single DnaB ring. Does RuvA inhibit branch migration of a heterologous junction catalyzed by two opposing DnaB rings? RuvA loads two opposing RuvB rings onto junction DNA; thus, the effect of RuvA on two opposing DnaB rings may provide further insight into RuvA mechanism during RuvAB function. Recombination processes must sometimes occur over

RuvA Promotes Branch Migration

regions of unmatched sequences, such as during DNA damage, and it is therefore relevant to test the mobility of RuvA over heterologous regions. Others have demonstrated that RuvAB can catalyze branch migration over regions of DNA heterology,34,36,38 but the role of RuvA has not been distinguished from that of RuvB in these publications. Below, we once again use DnaB in conjunction with RuvA to study the role of RuvA in heterologous branch migration. DnaB catalyzes branch migration of a long heterologous junction if two opposing DnaB rings are loaded onto the substrate.35 To load two DnaB rings onto the DNA, a long heterologous junction was constructed that contains two 5 0 tails on opposite sides of the junction (Figure 8(a)). DnaB rapidly catalyzes branch migration of this substrate, as expected, producing the *1,4 duplex product (Figure 7(a), scheme (i), and gel). There is a secondary reaction, where the *1-4 duplex is unwound by DnaB to yield free strand *1

483 (Figure 8(a), scheme (ii) and gel). Free strand *1 increases in intensity for the first 4 min of the reaction as the *1,4 duplex is unwound by DnaB (Figure 8(a), filled squares in graph). The levels of *1,4 duplex are roughly constant in the first 4 min of the reaction (Figure 8(a), open circles in graph). This result is explained if the rate of branch migration, shown in scheme (i), is similar to the rate of secondary unwinding, shown in scheme (ii). To determine if RuvA blocks this reaction, the substrate was incubated with RuvA before adding DnaB (Figure 8(b)). The opposing DnaB rings catalyze branch migration of this substrate rapidly, even in the presence of RuvA, producing the *1-4 duplex product (Figure 8(b), scheme (i) and gel). The concentration of *1-4 duplex continues to increase during the first 4 min of the reaction (Figure 8(b), open circles in the graph). In fact, the levels of branch migration product rise higher in the presence of RuvA than in the absence of RuvA

Figure 8. RuvA inhibits doubleDnaB ring-catalyzed unwinding, but not branch migration. The model above each gel is used to orient the reader, and was determined from the experimental gel evidence below. (a) DnaB acts on a long heterologous Holliday junction with two 5 0 tails. The 1-2 and 3-4 duplexes are 45 bp in length, and the 1-4 and 2-3 duplexes are 25 bp in length. Each 5 0 tail is composed of 30 dT. Oligonucleotides used to form this substrate are provided in Table 1. DnaB was incubated with the junction illustrated for the periods of time indicated. Native gel analysis of the reaction is shown. The arrows next to the gel show the migration distance of DNA products as determined by radiolabeled DNA markers electrophoresed in the same gel. Markers are cropped from the gel images for clarity. The accumulation of the *1,4 duplex (open circles) and free strand *1 (filled squares) are quantified and plotted. (b) Same as (a), except the substrate is incubated with RuvA for 1 min before adding DnaB. (c) Total product accumulation from gels shown in (a) (open diamonds) and (b) (filled circles) were quantified and plotted. (d) Experiments similar to those shown in (b) were performed, except the concentration of RuvA is varied and the DnaB incubation time is held constant at 2 min. Data from native gel analyses were quantified and plotted as a function of RuvA concentration (filled triangles). A similar experiment was conducted using a long heterologous Holliday junction with two forks. The concentration of RuvA was varied and the DnaB incubation time is held constant at 30 s (open circles). DnaB primarily catalyzes unwinding of this two-forked substrate (not shown). DnaB incubation times were chosen to roughly match product accumulation in the absence of RuvA.

484 (compare open circles in the graphs in Figure 8(a) and (b)). This may be due to RuvA inhibiting the secondary reaction of *1-4 duplex unwinding as the accumulation of free strand *1 is much slower in the presence of RuvA (see scheme (ii) and compare filled squares in the graphs in Figure 8(a) and (b)). To test if RuvA inhibits the unwinding reaction of scheme (ii) in Figure 8(b), the *1-4 duplex was incubated with DnaB in the absence and in the presence of 300 nM RuvA, the same concentration of RuvA as that used in Figure 8(b) (Figure 9). This high concentration of RuvA slightly inhibits DnaBcatalyzed unwinding of this duplex. Thus, RuvA partially blocks reaction (ii) in Figure 8(b), explaining why in the presence of RuvA, free strand *1 levels are low and *1-4 duplex product accumulates substantially. The rate of accumulation of all products in Figure 8(a) and (b) is plotted in the graph shown in Figure 8(c). RuvA does not inhibit the rate of accumulation of all products. All products in Figure 8(a) and (b) are derived from branch migration of the junction substrate (reaction (i)).29 Thus, RuvA inhibits DnaB unwinding of the *1-4 duplex, but it does not inhibit DnaB branch migration of the heterologous junction substrate with two 5 0 tails. This observation indicates that two DnaB pumps overcome RuvA inhibition of branch migration of a heterologous junction.

Figure 9. RuvA inhibits DnaB-catalyzed unwinding of forked-duplex DNA. (a) DnaB unwinds forked-duplex DNA. The substrate is the primary branch migration product of Figure 8(a) and (b). Oligonucleotides used to form this substrate are provided in Table 1. DnaB was incubated with the junction illustrated for the time indicated. Native gel analysis of the reaction is shown. (b) Same as (a), except the substrate is pre-incubated with RuvA for 1 min prior to adding DnaB.

RuvA Promotes Branch Migration

The ability of RuvA at high concentrations to bind to a free forked-duplex structure and slightly block DnaB activity, as demonstrated in Figure 9, raises the question of why RuvA does not bind to forked DNA structures when they are part of junction DNA in other experiments in this study (Figures 3 and 7), thereby partially inhibiting DnaB activity. However, the single-stranded tail regions of strands 1 and 4 in the forked-duplex of Figure 9 are not homopolymeric, and thus may form secondary structures to which RuvA has some affinity. DnaB can unwind this substrate because strand 4 retains enough single-stranded character for DnaB-loading. In contrast, the single-stranded regions of the substrates in Figures 3 and 7 are composed entirely of dT. These sequences form no secondary structure, and RuvA cannot bind to them (data not shown). Thus, RuvA inhibition of unwinding in Figures 3 and 7 is presumed to be the result of RuvA binding to the DNA branchpoint. RuvA inhibits single-DnaB ring unwinding and branch migration activity of a heterologous junction, but it does not inhibit double-ring branch migration activity. We next investigated if RuvA inhibits double-DnaB ring unwinding activity. We incubated DnaB with a heterologous junction with forks positioned on either side of the junction (Figure 8(d)). DnaB primarily catalyzes unwinding of this substrate (not shown). After 1 min incubation with RuvA, DnaB was added for 30 s. The concentration of RuvA was varied from 0 nM to 300 nM. Quantification of the reaction is shown in Figure 8(d) (open circles), and compared to a similar experiment using the substrate in Figure 8(b) (filled triangles, 2 min DnaB). The result shows that RuvA still inhibits double-DnaB activity on the unwinding substrate, but not the branch-migration substrate. We have shown elsewhere that two DnaB rings work in conjunction at this heterologous junction to power branch migration. The combined power of the two DnaB pumps likely also push RuvA over non-complementary base-pairs during branch migration (Figure 10(d)). However, the need to dislodge RuvA presumably underlies inhibition of helicase activity, even when two DnaB rings are present. Although we cannot rule out the possibility that RuvA is dislodged during branch migration over DNA heterology (Figure 8(d), filled triangles), it is unlikely, given the weak ability of two DnaB pumps to dislodge RuvA during unwinding (Figure 8(d), open circles).

Discussion RuvA blocks unwinding, but not branch migration RuvA binds Holliday junction DNA tightly in vitro (Figure 5). Once bound to a Holliday junction, RuvA inhibits DnaB unwinding activity (Figures 2(f) and 7(c)). Even when two DnaB

RuvA Promotes Branch Migration

485

Figure 10. Mobilization of the RuvA sliding collar at a Holliday junction by either one or two DNA pumps. (a) RuvA inhibits DNA unwinding at a Holliday junction. If the junction were to be unwound by a helicase, the Holliday junction will be destroyed and RuvA will likely be displaced. Thus, tight binding of the RuvA collar to the Holliday junction blocks unwinding by DnaB when it encircles one DNA strand. (b) RuvA does not inhibit homologous branch migration catalyzed by a single duplex DNA pump like DnaB encircling two DNA strands. During branch migration of a homologous junction, the junction structure is preserved as it translocates. The RuvA collar may slide along with the junction at little energetic cost. Thus, the RuvA sliding collar is readily mobilized for branch migration at a homologous junction, and does not require specific protein activation from RuvB. (c) During branch migration of a heterologous junction, two of the duplex arms will change from annealed duplex to unannealed duplex. The RuvA collar is optimized for binding to fully annealed duplex arms; thus, RuvA junction sliding is inhibited by unannealed DNA. (d) Two opposing DNA pumps mobilize the RuvA sliding collar at a heterologous junction, overcoming resistance of RuvA sliding over unannealed DNA. The RuvA sliding collar works in conjunction with two opposing RuvB rings in vivo. These two RuvB rings may enable the RuvA collar to slide over regions of DNA heterology or DNA lesions.

hexamers act on opposite sides of the RuvAjunction, RuvA continues to block DnaB-catalyzed unwinding activity (Figure 8(d)). Thus, the RuvAjunction is quite stable and resistant to DNA remodeling activity that involves unwinding. In contrast, RuvA does not block DnaB-catalyzed branch migration activity on a homologous Holliday junction (Figure 2(c)). At a heterologous junction, RuvA blocks single-DnaB ring branch migration activity (Figure 6(c)), but not double ring activity (Figure 8(d)). Thus, RuvA blocks unwinding activity but not branch migration activity. This result is supported further by the experiment illustrated by Figure 2, in which RuvA blocks DnaB-catalyzed unwinding, but not branch migration, at the same junction. At a homologous junction, RuvA blocks DnaB activity when the helicase surrounds one strand, but RuvA does not block DnaB when the hexameric ring surrounds two strands (Figures 2 and 3). Thus, RuvA is highly specific in its inhibition, as RuvA

recognizes whether DnaB is surrounding one DNA strand or two. RuvA likely evolved this strand specificity to help ensure that RuvA is mobile only in the correct context of RuvB-catalyzed branch migration. RuvB usually surrounds two strands in vivo, and RuvA does not block RuvB when RuvB surrounds two strands to power branch migration. However, if RuvB were to mistakenly surround one strand, RuvA would probably block RuvB action just as it blocks DnaB action. RuvB may mistakenly surround one DNA strand during extensive DNA damage with unannealed DNA strands. RuvB has been shown to act as a DNA helicase in vitro when surrounding one DNA strand,31 and RuvA may block this aberrant function at a Holliday junction in vivo. RuvA may block other helicases at a Holliday junction by this mechanism as well. The data presented here suggest that RuvA slides along homologous junction DNA during branch migration, but not unwinding. Why does RuvA not slide along the Holliday junction during

486 unwinding? In other words, how can RuvA distinguish whether DnaB is surrounding one or two DNA strands? Our previous studies of DnaB suggest that the ring-shaped protein functions by a similar mechanism, whether the helicase ring surrounds one strand during unwinding, or two DNA strands during branch migration. In either case, DnaB primarily pumps one DNA strand through the central channel. The primary difference between the two modes of action is that, during unwinding, the strands become separated and one strand passes outside of the DnaB ring, whereas during branch migration, both strands pass through the DnaB ring. While it is unclear at present how RuvA can discriminate between the two modes of DnaB action, this puzzle may have a basis similar to that of the replication termination system in bacteria, the Tus–Ter complex. The Tus–Ter complex blocks DnaB when the helicase approaches from one direction (the permissive direction), but not the other (the non-permissive direction).39 The Tus–Ter system is remarkably similar to the findings here for RuvA, in that, even for helicases that the Tus protein does not interact with in vivo, Tus blocks helicase action from only one direction.40,41 Although this mechanism is not understood completely, an interesting model has been proposed by the Dixon laboratory.39 In this model, the ssDNA produced by unwinding as DnaB approaches adheres to the ssDNA pocket in Tus, holding it tight to DNA and preventing DnaB from displacing it. Since Tus is asymmetric, the ssDNA pocket faces the duplex in one direction but not the other. Hence, DnaB approaching from the opposite side can displace Tus, since the ssDNA produced by the helicase does not interact with the ssDNA pocket on the other side of Tus. Perhaps RuvA acts similarly and binds ssDNA, allowing it to grip and not slide (RuvA is symmetric, and the hypothetical ssDNA pocket would be seen by all directions). This ssDNA is produced only when DnaB acts as a helicase, allowing RuvA to bind tightly to the DNA side and not slide. But when DnaB encircles duplex DNA, the ssDNA is not produced and RuvA develops no extra affinity for DNA and thereby slides as DnaB pushes on it. Ultimately, this speculation may or may not be correct, but at least it provides a mechanistic basis for understanding why RuvA slides while DnaB encircles two strands, and does not slide when it acts as a helicase. RuvA is intrinsically able to slide during branch migration As described above, RuvA likely slides during branch migration, but not during unwinding. For DnaB to catalyze unwinding of an RuvA-bound Holliday junction, RuvA must be displaced as the four-way junction structure is destroyed during the

RuvA Promotes Branch Migration

reaction (Figure 10(a)). In contrast, during DnaBcatalyzed branch migration of an RuvA-homologous junction complex, the fully annealed junction structure is continuously preserved, and thus RuvA can stay bound throughout the process (Figure 10(b)). The RuvA collar can simply slide along all four duplex arms of the junction while DnaB drives branch migration. Thus, RuvA binds Holliday junctions stably, and it is difficult to dislodge this protein completely from the DNA substrate. However, RuvA may simply slide along the junction during branch migration. The homologous junction used here bears a slight degree of heterology (five bases); thus, RuvA can tolerate a small degree of heterology and still slide freely. There is no evidence that DnaB and RuvA function together in vivo. Thus, the observation that RuvA is readily mobilized by DnaB for branch migration suggests that RuvA has an intrinsic capacity for movement on DNA during branch migration. Moreover, our results show that RuvA does not require specific contacts with its in vivo protein partner, RuvB, to move readily on DNA branch migration. It has been proposed that RuvA binds a Holliday junction tightly and, upon interaction with RuvB, RuvA is transformed to a weak DNA binder, capable of movement during branch migration.10 Evidence presented here suggests that RuvA binds to Holliday junctions tightly but yet is freely mobile in the direction of branch migration, without the requirement for activation by RuvB. Why does RuvA have an intrinsic capacity to slide in the direction of branch migration? RuvA needs to slide rapidly when functioning with RuvB during branch migration. Since RuvA has evolved to slide freely during branch migration, all of the energy of the RuvB motor can be utilized for branch migration. Thus, RuvA does not impede the function of its in vivo partner, RuvB. Furthermore, since RuvA remains bound tightly to Holliday junction DNA during branch migration, RuvA can continue to function as a sliding collar to preserve Holliday junction structure, and ultimately aid in the recruitment of the RuvC resolvase. Finally, this work suggests that RuvA is part of a family of nucleic acid sliding proteins that grip DNA tightly yet are also highly mobile (see below). Comparison of the RuvA collar with other DNA sliding proteins RuvA is mobilized readily during branch migration, even though it is not displaced easily from DNA. How can RuvA bind tightly to DNA, but also be readily pushed to slide along it? In this respect, RuvA has some similarity to the polymerase processivity clamp. In E. coli, the b-sliding clamp confers processivity to the replicative polymerase.42 During replication, the b-clamp remains bound to the DNA duplex, sliding along the DNA at a rate of 500–1000 bp/s. However, it is difficult to dislodge the b-clamp from the DNA, and an

RuvA Promotes Branch Migration

accessory protein is required to unload the b-clamp.43 The b-clamp forms a ring around the DNA, which enables tight DNA binding while retaining mobility to slide along DNA.44,45 Similar to the b-clamp, RuvA binds DNA topologically by forming a collar or sandwich structure around all four arms of the Holliday junction.46 Thus, the topological binding of RuvA at a Holliday junction ensures that RuvA does not dissociate readily from DNA during branch migration, while still allowing mobility on the DNA arms by sliding. However, there are important differences between the b-clamp and the RuvA collar. While the b-clamp surrounds one DNA duplex, RuvA is a four-way collar that encircles all four duplex arms simultaneously. Thus, for RuvA, all four duplexes must slide within it at the same time (Figure 10(b)). Moreover, the b-dimer ring is a more stable multimer than the RuvA-octamer collar,47 explaining why the b-dimer requires an accessory protein to crack open the dimer to load around DNA, while the RuvA collar can assemble spontaneously around DNA from two tetramers. The crystal structure of the b-clamp bound to DNA has not been solved, but the structure of the b-clamp alone reveals conserved positive and polar residues within the central channel.44 These residues may make direct or water-mediated contact with the DNA duplex, further ensuring that the protein remains bound to DNA while sliding for millions of base-pairs during genome replication. Although RuvA–DNA crystal structures have not been solved at sufficiently high resolution to map protein–DNA interactions unambiguously, RuvA has conserved positive and polar residues that may make direct and water-mediated contact with Holliday junction DNA.10 Water-mediated hydrogen bonds are likely to be quite mobile in an aqueous environment. Thus, the nature of chemical contact between RuvA or the b-clamp and DNA appears designed for DNA sliding. Moreover, RuvA primarily contacts the phosphate backbone of the DNA, which will maintain a uniform structure as RuvA slides along the Holliday junction. However, the b-clamp may have fewer direct contacts with DNA than RuvA, explaining how the b-clamp may slide more freely along DNA than RuvA. In addition, the RuvA sliding collar may not slide on its own, but may need a push by a duplex DNA pump for sliding motion. Other DNA-metabolizing enzymes, such as restriction enzymes, have been proposed to slide along DNA.48,49 Thus, protein sliding along DNA in a non-sequence-specific manner may be used by proteins to accomplish a variety of functions in the cell. Many nucleic acid-binding proteins must traverse along DNA at a very fast rate to accomplish their function, and DNA sliding is an elegant mechanism for proteins to move quickly along DNA without dissociation.

487 RuvA slides over a heterologous junction with two opposing DNA pumps Recombination processes must sometimes occur over regions of unmatched sequences, such as during DNA damage, and it is therefore important to understand the mobility of RuvA over heterologous regions. It has been shown that RuvAB catalyzes branch migration over regions of DNA heterology with less efficiency compared to homologous DNA,34,36,38 but the role of RuvA in this process was not dissected from RuvB in these earlier studies. We show in this study that if the junction is homologous, RuvA is freely mobile to slide over the junction, as discussed above. However, we show also that if the junction is heterologous, the unannealed strands within the RuvA collar inhibit RuvA sliding. DnaB cannot mobilize RuvA at a heterologous junction if the protein is loaded on one junction arm. However, if DnaB is loaded onto opposing junction arms, RuvA no longer inhibits DnaB-catalyzed branch migration. Thus, two opposing DNA pumps can overcome the inhibition to RuvA sliding that DNA heterology creates. RuvA normally functions with opposing RuvB rings. The two opposing RuvB rings may enable the RuvA collar to slide over regions of heterology, consistent with the published mobility of the RuvAB complex over regions of heterology.34,36,38 Thus, in vivo, two RuvB pumps may function in coordination to mobilize RuvA through regions of DNA heterology or DNA lesions. RuvA may function in vivo to stabilize Holliday junction structure and limit action to branch migration In vivo, a Holliday junction arises after RecA-mediated DNA recombination. The Holliday junction structure must be processed to linear duplexes to restore genomic integrity. The RuvABC proteins are responsible for this important cellular function, and RuvA is the first of these three proteins to bind the Holliday junction. We demonstrate here that RuvA, acting alone, inhibits DnaB unwinding activity substantially by binding tightly to a Holliday junction. The collision of a replication fork with an RuvA-bound Holliday junction may not be a frequent event. However, the RuvA collar may function in vivo to protect the Holliday junction from the deleterious action of other DNAmetabolizing enzymes until RuvB is recruited to the Holliday junction. RuvA may thus protect a Holliday junction from helicases, nucleases, and other enzymes that would destroy the Holliday junction structure. Furthermore, once RuvB is bound to RuvA and branch migration ensues, the RuvA can work as a sliding collar to allow efficient RuvB-catalyzed branch migration while preserving the structure of the Holliday junction from the deleterious effect of other DNA metabolizing enzymes.

488

RuvA Promotes Branch Migration

Materials and Methods Proteins and DNA Proteins were expressed in E. coli and purified as described: RuvA, RuvB,50 and DnaB.51 DNA oligonucleotides used to construct the substrates in this work are given in Table 1, and were synthesized as described.24 Oligonucleotide strands were labeled with 32P at the 5 0 -end as described.24 Branch migration assays All manipulations were performed in microfuge tubes on ice, and then shifted to 37 8C, unless stated otherwise. Oligonucleotides were annealed to form DNA substrates as described.35 Enzyme reactions were incubated at 37 8C for the lengths of time indicated, and contained 1 nM DNA substrate (concentration of labeled strand) in 20 mM Tris–HCl (pH 7.5) 5 mM ATP, 5 mM creatine phosphate, 20 mg/ml of creatine kinase, 10 mM magnesium acetate, 20% (v/v) glycerol, 100 mM EDTA, 40 mg/ml of bovine serum albumin, 5 mM DTT, in a final volume of 10 ml. Protein concentrations were 500 nM DnaB (hexamer), 300 nM RuvA (monomer), and/or 1 mM RuvB (monomer),

unless stated otherwise. RuvA was incubated with substrate DNA for 1 min at 37 8C before addition of RuvB or DnaB, and then incubated further as indicated in the Figure or described in the Figure legend. Reactions were quenched upon adding 1 ml of proteinase K (10 mg/ml), and incubated at 37 8C for an additional 1 min, followed by addition of 5 ml of 2% (w/v) SDS, 80 mM EDTA. For gel analysis, 5 ml of 15% (v/v) Ficoll (type 400; Pharmacia) and 0.25% (w/v) xylene cyanol FF were added. Samples were snapfrozen in a dry ice/ethanol bath and stored at K20 8C. Before gel electrophoresis, each reaction tube was incubated at room temperature in a waterbath for 4 min to reduce intrastrand base-pairing. For some experiments, the rate of spontaneous DNA reannealing is fast relative to the time-course of the experiment. For these reactions, unlabeled oligonucleotides were added in tenfold excess to inhibit spontaneous DNA reannealing. The DNA traps used in each experiment are listed in Table 1. Unlabeled DNA trap was added immediately before protein addition. High concentrations of DNA (O500 nM) are required to fully trap DnaB in these reactions. At these high concentrations, the DNA trap inhibits DnaB activity at the Holliday junction. Thus, we do not fully trap DnaB in this study, but we do use DNA at lower concentrations to trap DNA products generated during the reaction to prevent DNA reannealing.

Table 1. Sequences of DNA oligonucleotides Figure 1(a) 1(b) 2(a) and (b) 2(d) and (e) 3(a) and (b) 4(a) and (d) 5(b) 6(a) and (b) 6(e) and (f) 7(a) and (b) 8(a) and (b) 8(d) 9(a) and (b) 1 1E 1T 1TE 2 2E 2TE 3 3E 3O 3TE 4 4E 4O 4ODS 4OT 4TE 5DS

Oligonucleotides used

Cold DNA trap

1TE, 2E, 3O, and 4O 1T, 2, 3, and 4 1TE, 2E, 3O, and 4O 1TE-4O 1TE, 2E, 3O, 4ODS, 5DS 2E 1TE, 2E, 3O, and 4OT 2E-3O, and 2E 1E, 2E, 3E, and 4E 1E-2E, and 1E-4E 1, 2, 3, and 4 None 1T, 2, 3, and 4 1T-4 1T, 2, 3, and 4 1T-4, and 1T-2 1T, 2, 3, and 4T 1T-4T, and 1T 1TE, 2E, 3TE, and 4E 1TE-4E 1TE, 2E, 3TE, and 4E 1TE-4E 1TE, 2TE, 3TE, and 4TE 1TE-4TE 1TE and 4E None 5 0 -GAC GCT GCC GAA TTC TGG CTT GCT AGG ACA TCT TTG CCC ACG TTG ACC CG-3 0 5 0 -GAC GCT GCC GAA TTC TGG CTT GCT AGG ACA TCT TTG CCC ACG TTG ACC CGA TCG CTT AGG TAC GTT AAC C-3 0 5 0 -TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT GAC GCT GCC GAA TTC TGG CTT GCT AGG ACA TCT TTG CCC ACG TTG ACC CG-3 0 5 0 -TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT GAC GCT GCC GAA TTC TGG CTT GCT AGG ACA TCT TTG CCC ACG TTG ACC CGA TCG CTT AGG TAC GTT AAC C-3 0 5 0 -CGG GTC AAC GTG GGC AAA GAT GTC CTA GCA ATG TAA TCG TCT ATG ACG TC-3 0 5 0 -GGT TAA CG T ACC TAA GCG ATC GGG TCA ACG TGG GCA AAG ATG TCC TAG CAA TGT AAT CGT CTA TGA CGT C-3 0 5 0 -GGT TAA CG T ACC TAA GCG ATC GGG TCA ACG TGG GCA AAG ATG TCC TAG CAA TGT AAT CGT CTA TGA CGT C TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT-3 0 5 0 -GAC GTC ATA GAC GAT TAC ATT GCT AGG ACA TGC TGT CTA GAG ACT ATC GC-3 0 5 0 -GAC GTC ATA GAC GAT TAC ATT GCT AGG ACA TGC TGT CTA GAG ACT ATC GCA CGC TTT CGA ACG AGT CTT A-3 0 5 0 -GAC GTC ATA GAC GAT TAC ATT GCT AGG ACA TGC TGT CTC ACG TTG ACC CGA TCG CTT AGG TAC GTT AAC C-3 0 5 0 -TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT GAC GTC ATA GAC GAT TAC ATT GCT AGG ACA TGC TGT CTA GAG ACT ATC GCA CGC TTT CGA ACG AGT CTT A-3 0 5 0 -GCG ATA GTC TCT AGA CAG CAT GTC CTA GCA AGC CAG AAT TCG GCA GCG TC-3 0 5 0 -TAA GAC TCG TTC GAA AGC GTG CGA TAG TCT CTA GAC AGC ATG TCC TAG CAA GCC AGA ATT CGG CAG CGT C-3 0 5 0 -GGT TAA CGT ACC TAA GCG ATC GGG TCA ACG TGA GAC AGC ATG TCC TAG CAA GCC AGA ATT CGG CAG CGT C-3 0 5 0 -GGT TAA CGT ACC TAA GCG ATC GGG TCA ACG TGA GAC AGC ATG TCC TAG CAA GCC AGA ATT CGG CAG CGT C TTT TCG ATA TCC ATC CAT CCA TCC ATG CAC-3 0 5 0 -GGT TAA CGT ACC TAA GCG ATC GGG TCA ACG TGA GAC AGC ATG TCC TAG CAA GCC AGA ATT CGG CAG CGT C TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT-3 0 5 0 -TAA GAC TCG TTC GAA AGC GTG CGA TAG TCT CTA GAC AGC ATG TCC TAG CAA GCC AGA ATT CGG CAG CGT C TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT-3 0 5 0 -GTG CAT GGA TGG ATG GAT GGA TAT CGA-3 0

RuvA Promotes Branch Migration

DNA products were separated from DNA substrate in a native polyacrylamide gel as described.29 The gels were dried, exposed to a Phosphor-imaging screen, and quantified as described.29 The arrows next to the gel show the migration distance of DNA products as determined by radiolabeled DNA markers in the same gel. Markers are cropped from the gel images for clarity. Electrophoretic mobility-shift assay RuvA was incubated with radiolabeled junction DNA for 1 min at 37 8C under conditions identical with those used for branch migration assays. Then 5 ml of 15% Ficoll (type 400; Pharmacia) and 0.25% xylene cyanol FF was added to the sample. Protein–DNA complexes were separated from free DNA complexes by native gel electrophoresis. The gel was composed of 4% (w/v) polyacrylamide (29:1 (w/w) acrylamide/bis acrylamide) in a buffer of TBE (90 mM Tris–HCl/borate, 1 mM EDTA, pH 8.0). The running buffer was TBE. The gel was developed at 150 V at room temperature until the free DNA was near the bottom of the gel. The gels were dried, exposed to a Phosphor-imaging screen, and quantified as described.29

Acknowledgements We thank Nicholas Dixon, Taekjip Ha, and Nancy Horton for useful discussions regarding this work. We thank Steve West and Ken Marians for providing expression vectors for the Ruv proteins. Thanks to everyone in the O’Donnell laboratory. This research was supported by grant GM38839 from the NIH and by HHMI. D.L.K. was the Leon and Toby Cooperman Fellow of the Damon Runyon Cancer Research Foundation (DRG # 1663).

References 1. Lusetti, S. L. & Cox, M. M. (2002). The bacterial RecA protein and the recombinational DNA repair of stalled replication forks. Annu. Rev. Biochem. 71, 71–100. 2. West, S. C. (1997). Processing of recombination intermediates by the RuvABC proteins. Annu. Rev. Genet. 31, 213–244. 3. Parsons, C. A., Stasiak, A., Bennett, R. J. & West, S. C. (1995). Structure of a multisubunit complex that promotes DNA branch migration. Nature, 374, 375–378. 4. Yu, X., West, S. C. & Egelman, E. H. (1997). Structure and subunit composition of the RuvAB-Holliday junction complex. J. Mol. Biol. 266, 217–222. 5. Tsaneva, I. R., Muller, B. & West, S. C. (1992). ATPdependent branch migration of Holliday junctions promoted by the RuvA and RuvB proteins of E. coli. Cell, 69, 1171–1180. 6. Eggleston, A. K., Mitchell, A. H. & West, S. C. (1997). In vitro reconstitution of the late steps of genetic recombination in E. coli. Cell, 89, 607–617. 7. Eggleston, A. K. & West, S. C. (2000). Cleavage of Holliday junctions by the Escherichia coli RivABC complex. J. Biol. Chem. 275, 26467–26476.

489 8. Ingleston, S. M., Sharples, G. J. & Lloyd, R. G. (2000). The acidic pin of RuvA modulates Holliday junction binding and processing by the RuvABC resolvasome. EMBO J. 19, 6266–6274. 9. Hargreaves, D., Rice, D. W., Sedelnikova, S. E., Artymiuk, P. J., Lloyd, R. G. & Rafferty, J. B. (1998). Crystal structure of E. coli RuvA with bound DNA ˚ resolution. Nature Struct. Biol. Holliday junction at 6 A 5, 441–446. 10. Ariyoshi, M., Nishino, T., Iwasaki, H., Shinagawa, H. & Morikawa, K. (2000). Crystal structure of the Holliday junction DNA in complex with a single RuvA tetramer. Proc. Natl Acad. Sci. USA, 97, 8257–8262. 11. Yamada, K., Miyata, T., Tsuchiya, D., Oyama, T., Fujiwara, Y., Ohnishi, T. et al. (2002). Crystal structure of the RuvA–RuvB complex: a structural basis for the Holliday junction migrating motor machinery. Mol. Cell, 10, 671–681. 12. Yamada, K., Kunishima, N., Mayanagi, K., Ohnishi, T., Nishino, T., Iwasaki, H. et al. (2001). Crystal structure of the Holliday junction migration motor protein RuvB from Thermus thermophilus HB8. Proc. Natl Acad. Sci. USA, 98, 1442–1447. 13. Putnam, C. D., Clancy, S. B., Tsuruta, H., Gonzalez, S., Wetmur, J. G. & Tainer, J. A. (2001). Structure and mechanism of the RuvB Holliday junction branch migration motor. J. Mol. Biol. 311, 297–310. 14. Parsons, C. A., Tsaneva, I., Lloyd, R. G. & West, S. C. (1992). Interaction of Escherichia coli RuvA and RuvB proteins with synthetic Holliday junctions. Proc. Natl Acad. Sci. USA, 89, 5452–5456. 15. Hiom, K. & West, S. C. (1995). Branch migration during homologous recombination: assembly of a RuvAB Holliday junction complex in vitro. Cell, 80, 787–793. 16. Muller, B., Tsaneva, I. R. & West, S. C. (1993). Branch migration of Holliday junctions promoted by the Escherichia coli RuvA and RuvB proteins. II. Interaction of RuvB with DNA. J. Biol. Chem. 268, 17185–17189. 17. Patel, S. S. & Picha, K. M. (2000). Structure and function of hexameric helicases. Annu. Rev. Biochem. 69, 651–697. 18. Sawaya, M. R., Guo, S., Tabor, S., Richardson, C. C. & Ellenberger, T. (1999). Crystal structure of the helicase domain from the replicative helicase-primase of bacteriophage T7. Cell, 99, 167–177. 19. Singleton, M. R., Sawaya, M. R., Ellenberger, T. & Wigley, D. B. (2000). Crystal structure of T7 gene 4 ring helicase indicates a mechanism for sequential hydrolysis of nucleotides. Cell, 101, 589–600. 20. Baker, T. A., Sekimizu, K., Funnell, B. E. & Kornberg, A. (1986). Extensive unwinding of the plasmid template during staged enzymatic initiation of DNA replication from the origin of the Escherichia coli chromosome. Cell, 45, 53–64. 21. Bujalowski, W., Klonowska, M. M. & Jezewska, M. J. (1994). Oligomeric structure of Escherichia coli primary replicative helicase DnaB protein. J. Biol. Chem. 269, 31350–31358. 22. Yang, S., Yu, X., VanLoock, M., Jezewska, M., Bujalowski, W. & Egelman, E. (2002). Flexibility of the rings: structural asymmetry in the DnaB hexameric helicase. J. Mol. Biol. 321, 839–849. 23. Jezewska, M. J., Rajendran, S., Bujalowska, D. & Bujalowski, W. (1998). Does single-stranded DNA pass through the inner channel of the protein hexamer

490

24.

25.

26.

27. 28. 29. 30.

31.

32. 33. 34.

35.

36. 37.

RuvA Promotes Branch Migration

in complex with the Escherichia coli DnaB helicase? Fluorescence energy transfer studies. J. Biol. Chem. 273, 10515–10529. Kaplan, D. L. (2000). The 3 0 -tail of a forked-duplex sterically determines whether one or two DNA strands pass through the central channel of a replication-fork helicase. J. Mol. Biol. 301, 285–299. Jezewska, M. J., Rajendran, S. & Bujalowski, W. (1998). Complex of Escherichia coli primary replicative helicase DnaB protein with a replication fork: recognition and structure. Biochemistry, 37, 3116–3136. Delagoutte, E. & von Hippel, P. H. (2003). Helicase mechanisms and the coupling of helicases within macromolecular machines. Part II: integration of helicases into cellular processes. Quart. Rev. Biophys. 36, 1–69. Richardson, R. W. & Nossal, N. G. (1989). Characterization of the bacteriophage T4 gene 41 DNA helicase. J. Biol. Chem. 264, 4725–4731. Cunningham, E. L. & Berger, J. M. (2005). Unraveling the early steps of prokaryotic replication. Curr. Opin. Struct. Biol. 15, 68–76. Kaplan, D. L. & O’Donnell, M. (2002). DnaB drives DNA branch migration and dislodges proteins while encircling two DNA strands. Mol. Cell, 10, 647–657. Rajendran, S., Jezewska, M. J. & Bujalowski, W. (2000). Multiple-step kinetic mechanism of DNA-independent ATP binding and hydrolysis by Escherichia coli replicative helicase DnaB protein: quantitative analysis using the rapid quench-flow method. J. Mol. Biol. 303, 773–795. Tsaneva, I. R., Muller, B. & West, S. C. (1993). RuvA and RuvB proteins of Escherichia coli exhibit DNA helicase activity in vitro. Proc. Natl Acad. Sci. USA, 90, 1315–1319. LeBowitz, J. H. & McMacken, R. (1986). The Escherichia coli dnaB replication protein is a DNA helicase. J. Biol. Chem. 261, 4738–4748. Grigoriev, M. & Hsieh, P. (1998). Migration of a Holliday junction through a nucleosome directed by the E. coli RuvAB motor protein. Mol. Cell, 2, 373–381. Iype, L. E., Wood, E. E., Inman, R. B. & Cox, M. M. (1994). RuvA and RuvB proteins facilitate the bypass of heterolgous DNA insertions during RecA proteinmediated DNA strand exchange. J. Biol. Chem. 269, 24967–24978. Kaplan, D. L. & O’Donnell, M. (2004). Twin DNA pumps of a hexameric helicase provide power to simultaneously melt two duplexes. Mol. Cell, 15, 453–465. Adams, D. E. & West, S. C. (1996). Bypass of DNA heterologies during RuvAB-mediated three and fourstrand branch migration. J. Mol. Biol. 263, 582–596. Hishida, T., Iwasaki, H., Han, Y. W., Ohnishi, T. & Shinagawa, H. (2003). Uncoupling of the ATPase activity from the branch migration activity of RuvAB protein complexes containing both wild-type and ATPase-defective RuvB proteins. Genes Cells, 8, 721–730.

38. Dennis, C., Fedorov, A., Kas, E., Salome, L. & Grigoriev, M. (2004). RuvAB-directed branch migration of individual Holliday junctions is impeded by sequence heterology. EMBO J. 23, 2413–2422. 39. Mulugu, S., Shamsuzzaman, A. P. A., Taylor, J., Alexander, K. & Bastia, D. (2001). Mechanism of termination of DNA replication of Escherichia coli involves helicase–contrahelicase interaction. Proc. Natl Acad. Sci. USA, 98, 9569–9574. 40. Lee, E. H., Kornberg, A., Hidaka, M., Kobayashi, T. & Horiuchi, T. (1989). Escherichia coli replication termination protein impedes the action of helicases. Proc. Natl Acad. Sci USA, 86, 9104–9108. 41. Bedrosian, C. L. & Bastia, D. (1991). Escherichia coli replication terminator protein impedes simian virus 40 (SV40) DNA replication fork movement and SV40 large tumor antigen helicase activity in vitro at a prokaryotic terminus sequence. Proc. Natl Acad. Sci. USA, 88, 2618–2622. 42. Kelman, Z. & O’Donnell, M. (1995). DNA polymerase III holoenzyme: structure and function of a chromosomal replicating machine. Annu. Rev. Biochem. 64, 171–200. 43. Leu, F. P., Hingorani, M. M., Turner, J. & O’Donnell, M. (2000). The delta subunit of DNA polymerase III holoenzyme serves as a sliding clamp unloader in Escherichia coli. J. Biol. Chem. 275, 34609–34618. 44. Kong, X. P., Onrust, R. O., O’Donnell, M. & Kuriyan, J. (1992). Three-dimensional structure of the beta subunit of E. coli DNA polymerase III holoenzyme is a sliding DNA clamp. Cell, 69, 425–437. 45. Stukenberg, P. T., Studwell-Vaughan, P. S. & O’Donnell, M. (1991). Mechanism of the sliding beta-clamp of DNA polymerase III holoenzyme. J. Biol. Chem. 266, 11328–11334. 46. Roe, S. M., Barlow, T., Brown, T., Oram, M., Keeley, A., Tsaneva, I. R. & Pearl, L. H. (1998). Crystal structure of an octameric RuvA-Holliday junction complex. Mol. Cell, 2, 361–372. 47. Lee, Y. C., Flora, R., McCafferty, J. A., Gor, J., Tsaneva, I. R. & Perkins, S. J. (2003). A tetrameroctamer equilibrium in Mycobacterium leprae and Escherichia coli RuvA by analytical ultracentrifugation. J. Mol. Biol. 333, 677–682. 48. Viadiu, H. & Aggarwal, A. K. (2000). Structure of BamHI bound to nonspecific DNA: a model for DNA sliding. Mol. Cell, 5, 889–895. 49. Kampmann, M. (2004). Obstacle bypass in protein motion along DNA by two-dimensional rather than one-dimensional sliding. J. Biol. Chem. 279, 38715–38720. 50. Tsaneva, I., Illing, G., Lloyd, R. & West, S. (1992). Purification and properties of the RuvA and RuvB proteins of Escherichia coli. Mol. Gen Genet. 235, 1–10. 51. Yuzhakov, A., Turner, J. & O’Donnell, M. (1996). Replisome assembly reveals the basis for asymmetric function in leading and lagging strand replication. Cell, 86, 877–886.

Edited by R. Ebright (Received 1 September 2005; received in revised form 25 October 2005; accepted 26 October 2005) Available online 16 November 2005