Synergistic effect of ATP for RuvA–RuvB–Holliday junction DNA complex formation

The Escherichia coli RuvB hexameric ring motor proteins, together with RuvAs, promote branch migration of Holliday junction DNA. Zero mode waveguides (ZMWs) constitute of nanosized holes and enable the visualization of a single fluorescent molecule under micromolar order of the molecules, which is applicable to characterize the formation of RuvA–RuvB–Holliday junction DNA complex. In this study, we used ZMWs and counted the number of RuvBs binding to RuvA–Holliday junction DNA complex. Our data demonstrated that different nucleotide analogs increased the amount of Cy5-RuvBs binding to RuvA–Holliday junction DNA complex in the following order: no nucleotide, ADP, ATPγS, and mixture of ADP and ATPγS. These results suggest that not only ATP binding to RuvB but also ATP hydrolysis by RuvB facilitates a stable RuvA–RuvB–Holliday junction DNA complex formation.


Results
Labeling RuvB with Cy5. To characterize RuvB binding to Holliday junction DNA using the single molecule fluorescence imaging technique, we constructed and purified a RuvB mutant, RuvB-S39C, and then label the purified RuvB protein with Cy5-maleimide as described in Materials and Methods. We used RuvB-S39C to label RuvB protein by the highly specific conventional reaction between the sulfhydryl group and maleimide group because wild type E. coli RuvB has no Cys residues.
To determine the effect of Ser-Cys mutation, we measured branch migration activity of the purified RuvB mutant in the presence of RuvA using a stopped-flow system (Fig. 1). We used fluorescently labeled Holliday junction DNA, which contained a Cy3 fluorophore and a Cy5 fluorophore, at the same end of the DNA (Fig. 1b). Before Holliday junction DNA branch migration took place, Cy3 and Cy5 were located closely to each other and the fluorescence of Cy3 was suppressed by an energy transfer from Cy3 to Cy5. On the other hand, completion of the Holliday junction DNA branch migration yielded separate Cy3-labeled and Cy5-labeled Y-form DNA, and the Cy3 fluorescence resumed (Fig. 1b). As shown in Fig. 1c, fluorescence intensity from Cy3 started increasing from 5 s introduction of wild-type RuvB into the solutions at 25 °C, indicating that Holliday junction DNAs were unwound (Fig. 1c). Our data demonstrates that RuvB-S39C is slightly defective in Holliday junction DNA branch migration activity compared with wild-type RuvB (Fig. 1c). Because RuvB-S39C is still active in Holliday junction DNA branch migration with RuvA, we labeled RuvB-S39C with Cy5 as described in Materials and Methods.
The labeling ratio of Cy5-labeled RuvB-S39C was 42%. The Holliday junction DNA branch migration activity of Cy5-labeled RuvB-S39C was comparable with that of unlabeled RuvB-S39C (Fig. 1d), indicating that the activity was unaffected by Cy5 labeling. Thus, we used Cy5-RuvB-S39C as Cy5-RuvB to characterize RuvB binding to Holliday junction DNA using the single-molecule fluorescence imaging technique. ZMW fabrication. As described in the methods, 400 nM of Cy5-labeled RuvB-S39C was used to visualize RuvB binding to the RuvA-Holliday junction DNA complex. Because ZMWs enable us to visualize single fluorescently labeled biomolecules at a high concentration of them, we fabricated ZMWs for single molecule Cy5-RuvB observation. Two methods have primarily been reported for ZMW fabrication 26 , the ion-beam milling method 24 and the metal lift-off method 31,32 . In this study, we fabricated ZMWs using the metal lift-off method and obtained nanoscale apertures in aluminum films on the center of fused silica coverslips, as described in Materials and Methods (Fig. 2a,b). After ZMW fabrication, we observed the nanoholes using a scanning electron microscope (SU8000, Hitachi High Technologies) and measured the average diameter of the holes, which was 122 ± 10 nm (Fig. 2c). The hole size was small enough for imaging of Cy5-RuvB binding to a RuvA-Holliday junction DNA complex at the concentration of 400 nM used in this study 24 .

RuvA-RuvB complexes promote branch migration of Holliday junction DNA immobilized on
ZMWs. To confirm that RuvA-RuvB complexes is capable of promoting branch migration of Holliday junction DNA in the nanoholes, Cy3-labeled Holliday junction DNA was immobilized on a streptavidin coated glass surface, as described in Materials and Methods. The ratio of fluorescent spots of Holliday junction DNA to nanoholes was about 90% before the addition of RuvB proteins to the nanoholes (Fig. 3). After the addition of RuvB proteins with ATP and incubation for 5 min at 25 °C, the ratio of the spots was approximately 16% (Fig. 3a). In contrast, without ATP, the ratio of the spots was almost same as that before the addition of RuvB (Fig. 3b). These results indicate that RuvA-RuvB-Holliday junction DNA complex was formed in the nanohole and that the RuvA-RuvB protein complex with ATP could promote branch migration of Holliday junction DNA immobilized on the nanohole, resulting in dissociation of Cy3-labeled Y-form DNA from the nanohole (Fig. 3c).
Number of RuvBs binding to RuvA-Holliday junction DNA, immobilized on ZMWs. As described above, we demonstrated that the RuvA-RuvB complex promoted branch migration of Cy3 labeled Holliday junction DNA immobilized on the nanohole, indicating that the RuvA-RuvB-Holliday junction DNA complex was formed in the nanohole. Next, we performed single molecule characterization of the RuvA-RuvB-Holliday junction DNA complex formation using Cy5-labeled RuvB. In the presence of ATP, the RuvA-RuvB protein complex promotes Holliday junction DNA branch migration, resulting in the disassembly of Holliday junction DNA and the formation of Y-form DNA. Here, we used ATPγ S and ADP as nucleotide cofactors. It was impossible for us to visualize Cy5-RuvB binding to the junction in the presence of ATP. We observed bright spots that emitted stable Cy3 and Cy5 fluorescence from nanoholes on which Cy3-Holliday junction DNA was immobilized. Most of the Cy3 and Cy5 fluorescence intensity decreased in a stepwise manner due to photobreaching (Fig. 4a,b, Supplement Movie 1). The numbers of photobreaching steps corresponded with the number of Cy3-Holliday junction DNA immobilized on the nanohole and Cy5-RuvB binding to the RuvA-Holliday junction DNA complexes, respectively. The mean signal-to-noise ratios of Cy3 and Cy5 were 2.2 and 3.9, respectively. As shown in Fig. 4b, we focused on nanoholes containing single Cy3-Holliday junction DNA and counted the number of photobleaching steps from Cy5-RuvB to determine the number of Cy5-RuvBs binding to the complex in the nanoholes containing single Holliday junction DNA.
We characterized RuvB binding to the complex under conditions without nucleotides, with ADP, ATPγ S or both ADP and ATPγ S (Fig. 4c-f). Because the labeling ratio of Cy5-RuvB was 42%, the number of photobleaching step did not represent the number of RuvB. Thus, to determine the number of RuvBs binding to the complex, we fitted our experimental data with calculated data. As shown in Fig. 1d, our data showed that branch migration activity of Cy5-RuvB was comparable to that of RuvB-S39C. Thus, we regarded affinities to the RuvA-Holliday junction DNA complex of Cy5-RuvB and RuvB-S39C as almost equivalent. Previous biochemical assays indicated that in the absence of nucleotide and divalent cations such as Mg 2+ , RuvBs exist as a monomer and/or dimer 2,33,34 . We considered the RuvB protomer as a monomer and calculated a binominal distribution between 42% of Cy5-RuvB and 58% of nonlabeled RuvB to obtain the calculated data (Supplementary Table S1). A least-squares fitting technique was performed, and we determined the minima of the sum of the square residuals calculated using least-squares fitting technique between our experimental data and the calculated data to fit our experimental data with the calculated data ( Fig. 4c-f and Table 1). The fitted results indicate that in the absence of ADP or ATPγ S, 77% of Holliday junction DNA interacted with RuvB, and that 37% and 40% of the Holliday junction DNA had one or two RuvBs, respectively. In contrast, the results indicate that in the presence of ADP, 90% of Holliday junction DNAs interacted with RuvB, and 30% and 29% of the Holliday junction DNA had three or four RuvBs, respectively. These results indicate that the presence of nucleotide promotes more RuvBs binding to a RuvA-Holliday junction DNA complex. The fitted results also indicate that in the presence of ATPγ S, 92% of Holliday junction DNA interacted with RuvBs, and 31% and 12% of the Holliday junction DNA interacted with four or five RuvBs, respectively. This indicates that ATPγ S promotes more RuvBs binding to a RuvA-Holliday junction DNA complex than ADP. Intriguingly, the fitted results indicate that in the presence of ADP and ATPγ S, 98% of Holliday junction DNA interacted with RuvBs and 31%, 37%, 14%, and 3% of the Holliday junction DNA had three, four, five, and six RuvBs, respectively (Table 1). In the case that the RuvB protomer is a dimer, we considered that the distribution of Cy5-RuvB in dimer was random, and as shown in Supplementary Table S2, similar calculated data was obtained and compared with that based on the RuvB monomer model. We fitted our data with the calculated data to obtain the distribution of the number of RuvBs binding to a RuvA-Holliday junction DNA complex ( Table 2). Even though the numbers of RuvBs binding to a RuvA-Holliday junction DNA complex were only even numbers, the fitted data was almost similar to that from the RuvB monomer model.

Discussion
To date, RuvB properties on RuvA-RuvB-Holliday junction complex formation or DNA-binding activity have been largely characterized by an electrophoresis mobility shift assay (EMSA) with glutaraldehyde cross-linking [35][36][37] , because of the weak stability of the RuvB-DNA complex. As reported previously, our EMSA data showed that the RuvA-Holliday junction DNA complex formed complexes with RuvB in the presence of ATPγ S 13 . However, we could not measure the number of RuvBs in the complex.
Single fluorescence imaging techniques enabled us to characterize the protein-protein or protein-DNA complex in more detail. We could visualize the assembly or disassembly processes of the complex and count the number of molecules constituting the complex in real time. In this study, to characterize the single molecule formation of RuvA-RuvB-Holliday junction DNA complex, we labeled RuvB with Cy5 and fabricated ZMWs. We measured the number of RuvBs binding to a RuvA-Holliday junction DNA complex under various nucleotide conditions (Fig. 4). Interestingly, our results indicate that in the absence of ATPγ S or ADP, RuvBs formed complexes with RuvA-Holliday junction DNA complexes and all of the complexes contained one or two RuvBs. Our results also indicate that in the presence of ATPγ S or ADP, about 90% of Holliday junction DNA formed complexes with RuvBs.
To date, the crystallographic RuvA-RuvB complex structure containing AMPPNP or ADP has been resolved; however, structural information of the RuvA-RuvB complex without a nucleotide has not been reported 14   Previously, the crystallographic RuvA domain III-RuvB structure revealed that the β -hairpin was partly involved in the interface of RuvB subunits assembly 14 . The interface contains an arginine finger, which senses ATP hydrolysis in the adjacent RuvB subunit 40 . The arginine finger is located between Sensor I and Sensor II 5 , which are also involved in ATP binding and hydrolysis in cooperation with Walker A and B motifs. Our data suggests that ATP or ADP binding to RuvB induces structural changes, not only for a higher oligomeric formation of RuvB, but also for a stable RuvA-RuvB interaction.
In the presence of ATPγ S and ADP, 97%-98% of Holliday junction DNAs interacted with RuvBs and approximately 3% of the complexes contained six RuvBs (Table 1 and 2). This demonstrated that different nucleotide analogs increased the number of RuvBs binding to RuvA-Holliday junction DNA in the following order: no nucleotide, ADP, ATPγ S, and both of ADP and ATPγ S. Because ATPγ S is ATP nonhydrolyzable analogue, our data suggested that RuvB hexamer containing ATP and ADP was a more stable complex compared with other RuvB hexamers. Like F1-ATPase, RuvB hexamer constituting one pair each of ATP-bound, ADP-bound, and nucleotide-free monomers is supposed to be a stable RuvB hexamer [40][41][42] . Our data also showed that in the presence of ATPγ S and ADP, most of RuvA-RuvB-Holliday junction DNA complexes contained three, four, or five RuvBs at 400 nM of the Cy5-RuvB. These complexes might indicate the intermediate RuvB hexameric ring formation, suggesting that RuvB monomers and/or dimers assemble on the RuvA-Holliday junction complex in the presence of ATP to form RuvA-RuvB-Holliday junction DNA complex at a low concentration of RuvB (Fig. 5) 43 . Electron microscopic imaging of RuvA-RuvB-Holliday junction DNA complexes showed that RuvBs formed a hexameric ring on dsDNA in the presence of ATPγ S, suggesting that ATP hydrolysis was not required for hexameric ring formation. However, our data demonstrated that in the presence of ATPγ S and ADP, more RuvBs interacted with RuvA-Holliday junction DNA complexes, compared with that in the presence of ATPγ S only. Furthermore, the stopped flow analysis demonstrated that RuvA-RuvB mediated Holliday junction DNA branch migration started several seconds after mixing RuvA, RuvB, Holliday junction DNA, and ATP (Fig. 1), indicating that hexameric RuvB rings formed on Holliday junction DNA in several seconds. However, as shown in Fig. 4d, less RuvB hexameric rings formed on Holliday junction at 400 nM of RuvB in the presence of ATPγ S. These data indicate that the rate constant of RuvB hexameric ring formation on a RuvA-Holliday junction DNA complex in the presence of ATP was much faster than that in the presence of ATPγ S. These results suggest that not only ATP binding to RuvBs but also ATP hydrolysis by RuvBs facilitated RuvB hexameric ring formation on dsDNA.
The RuvB protomer was recently supposed to be dimer 13 ; however, we could not rule out the possibility that RuvB exits as a monomer at low concentration of RuvB in the absence of ATP and Mg 2+ . Thus, in this study, we determined the distribution of the number of RuvBs binding to a RuvA-Holliday junction DNA based on two models. One model is based on the model that the RuvB protomer is a monomer (Table 1), and another model assumes that the RuvB protomer is a dimer and we considered that Cy5-RuvB existed in RuvB dimers at random ( Table 2). These data were comparable with each other, even though in the case of RuvB dimer model, only even numbers of RuvBs binds to a RuvA-Holliday junction DNA complex. To characterize the RuvB loading process  onto Holliday junction DNA in more detail, we need to visualize the initial steps of RuvB loading to DNA in the presence of ATP. We are currently customizing our ZMWs combining with microfluidic system as reported previously 23,44,45 , which enables us to visualize the initial step of the complex formation process in real time. Not only the customizing system, but also higher labeling ratio of fluorescently labeled RuvB are required; however, in this study, the labeling ratio was 42%. In this study, we constructed two Ser-Cys mutant, RuvB-S39C and RuvB-S9C. RuvB-S9C was defective in Holliday junction DNA branch migration activity and we did not use the RuvB mutant protein (data not shown). However, E. coli RuvB has 11 Ser residues, and we are now constructing other Ser-Cys mutant to obtain fluorescently labeled RuvB with high labeling ratio. As described above, the labeling ratio of RuvB-S39C with Cy5 was 42%. In this study, to determine the number of RuvBs binding to a RuvA-Holliday junction DNA, we considered two possibilities. One possibility is that the RuvB protomer is a monomer ( Table 1) and another possibility is that the RuvB protomer is a dimer and Cy5-RuvBs are distributed throughout RuvB dimer at random (Table 2). Furthermore, we assumed that RuvB stably forms a dimer and only one RuvB in the dimer can be labeled by Cy5. In this case, all of Cy5 labeled RuvB dimers contain a Cy5-RuvB and a non-labeled RuvB. Even though we do not have any data to support this model, we calculated the binominal distribution based on this assumption. As shown in Supplementary Table S3, we assumed that 84% of RuvB dimers contained a Cy5-RuvB and 16% of RuvB dimers were non labeled RuvB dimers. We also calculated the binominal distribution between 84% of Cy5 labeled RuvB dimers and 16% of non labeled RuvB dimers to obtain the calculated data (Supplementary Table 3). We fitted our data with the calculated data to obtain the distribution of the number of RuvBs binding to a RuvA-Holliday junction DNA complex (Table 3). Compared with the data from Table 1 and 2, the number of RuvBs binding to a RuvA-Holliday junction DNA complex increased by approximately two-fold; however, the number of RuvBs binding to the complex was 10 at the maximum. These data suggested that more RuvB was required for the formation of double RuvB hexameric rings with a RuvA-Holliday junction DNA complex in the presence of ADP and ATPγ S. To visualize RuvBs binding to a RuvA-Holliday junction DNA complex at 10 μ M RuvB using ZMWs, the diameter of the nanoholes should be narrow and approximately 50 nm.
Previous biochemical analyses and electron microscopic observations demonstrated that RuvBs form a hexameric or heptameric ring in solution with ATP or ATPγ S, suggesting that the RuvB rings directly load onto Holliday junction DNA, resulting in a RuvA-RuvB-Holliday junction DNA complex formation at a high concentration of RuvB 46 . However, the reaction mechanism as to how the RuvB rings directly load onto dsDNA has not yet been clarified (Fig. 5). As discussed above, we are now customizing and improving our ZMWs and further analysis of the formation of RuvA-RuvB-Holliday junction DNA complex is now in progress.
Holliday junction DNA preparation. Two Holliday junction DNAs were prepared as below. Four oligonucleotides (JY21-Cy5, JY22-Cy3, JY23, and JY24) were mixed in a buffer containing 50 mM Tris-HCl (pH 7.5), 100 mM NaCl, and 10 mM MgCl 2 , and Holliday junction DNA was constructed as described previously and used for the branch migration assay 37  ZMW fabrication. Fused silica coverslips were immersed in 4% ammonium and 4.3% hydrogen peroxide for 10 min at 75 °C and then washed thoroughly with deionized water. The coverslips were dried with an air blower, and then baked at 200 °C for 10 min. The coverslips were then cleaned by air plasma at 18 W for 10 min. Before coating with Hexamethyldisilazane (HMDS, AZ Electronic Materials), the coverslips were immersed in 2-butanone and cleaned by sonication for 5 min. A resist film of ma-N 2403 (Micro resist technology) was then coated on the HMDS coated coverslip with a spin coater. The ESPACER (Showa Denko) was then coated on the ma-N 2403 coated coverslip. Electron beam (EB) lithography (Elionix Inc.) was performed with an accelerating voltage of 80 kV, and a beam current of 100 pA. After EB patterning, the coverslips were immersed in deionized water for 30 s, and the pattern was then developed by immersing it in ma-D 525 (Micro resist technology) for 2 min. After development, the coverslips were washed thoroughly with deionized water for 5 min and dried with an air blower. Aluminum coating was performed in a thermal evaporator using a BN composite boat. The thickness of the coated aluminum was monitored using a thickness monitor (Eiko Engineering Co. Ltd.). After aluminum coating, the remaining photoresist was lifted off by immersing it in 2-butanone with sonication for 5 min.
The ZMWs were observed by scanning electron microscopy (SU-8000; Hitachi) and the diameter of each nanohole was measured.
Microscope. Samples were observed at 25 ± 2 °C on an Olympus IX71 inverted microscope with a 100X oil-immersion objective as described previously 47 . An Nd:YAG laser (Compass 315M, Coherent) and a HeNe laser (05-LHP-991, Melles Griot) were used to excite Cy3 at 532 nm and Cy5 at 633 nm, respectively. The fluorescent signals from the samples were passed through dichroic mirrors to separate the fluorescences of Cy3 and Cy5. Barrier filters (580DF30 for Cy3 and 670DF40 for Cy5) were used to eliminate the background light. The filtered fluorescence signals (565-595 nm for Cy3 and 650-690 nm for Cy5) were imaged using a dual view apparatus and recorded with a high-sensitivity CCD camera. The recorded images were analyzed using Image Pro Plus.
l, m, n, p, q, r, and s indicate percentages of 0, 1, 2, 3, 4, 5, and 6 RuvB dimers binding to a RuvA-Holliday junction DNA. The l, m, n, p, q, r, and s satisfied the following conditions. + + + + + + = l m n p q r s 100 l, m, n, p, q, r, and s were nonnegative integers. The inequalities as below were not allowed. > < , > < , > < , > < , > < . l m n m n p n p q p q r q r s